0 files changed, 0 insertions, 0 deletions
diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig
index ba3fb04bb691..7e6bc0b3a589 100644
--- a/drivers/gpu/drm/Kconfig
+++ b/drivers/gpu/drm/Kconfig
@@ -16,8 +16,7 @@ menuconfig DRM
 # gallium uses SYS_kcmp for os_same_file_description() to de-duplicate
 # device and dmabuf fd. Let's make sure that is available for our userspace.
 	select KCMP
-	select VIDEO_CMDLINE
-	select VIDEO_NOMODESET
+	select VIDEO
 	help
 	  Kernel-level support for the Direct Rendering Infrastructure (DRI)
 	  introduced in XFree86 4.0. If you say Y here, you need to select
@@ -27,6 +26,13 @@ menuconfig DRM
 	  details.  You should also select and configure AGP
 	  (/dev/agpgart) support if it is available for your platform.
 
+menu "DRM debugging options"
+depends on DRM
+source "drivers/gpu/drm/Kconfig.debug"
+endmenu
+
+if DRM
+
 config DRM_MIPI_DBI
 	tristate
 	depends on DRM
@@ -36,73 +42,99 @@ config DRM_MIPI_DSI
 	bool
 	depends on DRM
 
-config DRM_DEBUG_MM
-	bool "Insert extra checks and debug info into the DRM range managers"
-	default n
-	depends on DRM=y
-	depends on STACKTRACE_SUPPORT
-	select STACKDEPOT
+config DRM_KMS_HELPER
+	tristate
+	depends on DRM
+	select FB_CORE if DRM_FBDEV_EMULATION
 	help
-	  Enable allocation tracking of memory manager and leak detection on
-	  shutdown.
-
-	  Recommended for driver developers only.
-
-	  If in doubt, say "N".
+	  CRTC helpers for KMS drivers.
 
-config DRM_USE_DYNAMIC_DEBUG
-	bool "use dynamic debug to implement drm.debug"
-	default n
-	depends on BROKEN
+config DRM_DRAW
+	bool
 	depends on DRM
-	depends on DYNAMIC_DEBUG || DYNAMIC_DEBUG_CORE
-	depends on JUMP_LABEL
-	help
-	  Use dynamic-debug to avoid drm_debug_enabled() runtime overheads.
-	  Due to callsite counts in DRM drivers (~4k in amdgpu) and 56
-	  bytes per callsite, the .data costs can be substantial, and
-	  are therefore configurable.
 
-config DRM_KUNIT_TEST_HELPERS
-	tristate
-	depends on DRM && KUNIT
+config DRM_PANIC
+	bool "Display a user-friendly message when a kernel panic occurs"
+	depends on DRM
+	select FONT_SUPPORT
+	select DRM_DRAW
 	help
-	  KUnit Helpers for KMS drivers.
-
-config DRM_KUNIT_TEST
-	tristate "KUnit tests for DRM" if !KUNIT_ALL_TESTS
-	depends on DRM && KUNIT
-	select PRIME_NUMBERS
-	select DRM_DISPLAY_DP_HELPER
-	select DRM_DISPLAY_HELPER
-	select DRM_LIB_RANDOM
-	select DRM_KMS_HELPER
-	select DRM_BUDDY
-	select DRM_EXPORT_FOR_TESTS if m
-	select DRM_KUNIT_TEST_HELPERS
-	default KUNIT_ALL_TESTS
+	  Enable a drm panic handler, which will display a user-friendly message
+	  when a kernel panic occurs. It's useful when using a user-space
+	  console instead of fbcon.
+	  It will only work if your graphic driver supports this feature.
+	  To support Hi-DPI Display, you can enable bigger fonts like
+	  FONT_TER16x32
+
+config DRM_PANIC_FOREGROUND_COLOR
+	hex "Drm panic screen foreground color, in RGB"
+	depends on DRM_PANIC
+	default 0xffffff
+
+config DRM_PANIC_BACKGROUND_COLOR
+	hex "Drm panic screen background color, in RGB"
+	depends on DRM_PANIC
+	default 0x000000
+
+config DRM_PANIC_DEBUG
+	bool "Add a debug fs entry to trigger drm_panic"
+	depends on DRM_PANIC && DEBUG_FS
 	help
-	  This builds unit tests for DRM. This option is not useful for
-	  distributions or general kernels, but only for kernel
-	  developers working on DRM and associated drivers.
-
-	  For more information on KUnit and unit tests in general,
-	  please refer to the KUnit documentation in
-	  Documentation/dev-tools/kunit/.
-
+	  Add dri/[device]/drm_panic_plane_x in the kernel debugfs, to force the
+	  panic handler to write the panic message to this plane scanout buffer.
+	  This is unsafe and should not be enabled on a production build.
 	  If in doubt, say "N".
 
-config DRM_KMS_HELPER
-	tristate
-	depends on DRM
+config DRM_PANIC_SCREEN
+	string "Panic screen formatter"
+	default "user"
+	depends on DRM_PANIC
 	help
-	  CRTC helpers for KMS drivers.
+	  This option enable to choose what will be displayed when a kernel
+	  panic occurs. You can choose between "user", a short message telling
+	  the user to reboot the system, or "kmsg" which will display the last
+	  lines of kmsg.
+	  This can also be overridden by drm.panic_screen=xxxx kernel parameter
+	  or by writing to /sys/module/drm/parameters/panic_screen sysfs entry
+	  Default is "user"
+
+config DRM_PANIC_SCREEN_QR_CODE
+	bool "Add a panic screen with a QR code"
+	depends on DRM_PANIC && RUST
+	select ZLIB_DEFLATE
+	help
+	  This option adds a QR code generator, and a panic screen with a QR
+	  code. The QR code will contain the last lines of kmsg and other debug
+	  information. This should be easier for the user to report a kernel
+	  panic, with all debug information available.
+	  To use this panic screen, also set DRM_PANIC_SCREEN to "qr_code"
+
+config DRM_PANIC_SCREEN_QR_CODE_URL
+	string "Base URL of the QR code in the panic screen"
+	depends on DRM_PANIC_SCREEN_QR_CODE
+	help
+	  This option sets the base URL to report the kernel panic. If it's set
+	  the QR code will contain the URL and the kmsg compressed with zlib as
+	  a URL parameter. If it's empty, the QR code will contain the kmsg as
+	  uncompressed text only.
+	  There is a demo code in javascript, to decode and uncompress the kmsg
+	  data from the URL parameter at https://github.com/kdj0c/panic_report
+
+config DRM_PANIC_SCREEN_QR_VERSION
+	int "Maximum version (size) of the QR code."
+	depends on DRM_PANIC_SCREEN_QR_CODE
+	default 40
+	help
+	  This option limits the version (or size) of the QR code. QR code
+	  version ranges from Version 1 (21x21) to Version 40 (177x177).
+	  Smaller QR code are easier to read, but will contain less debugging
+	  data. Default is 40.
 
 config DRM_DEBUG_DP_MST_TOPOLOGY_REFS
         bool "Enable refcount backtrace history in the DP MST helpers"
 	depends on STACKTRACE_SUPPORT
         select STACKDEPOT
-        depends on DRM_KMS_HELPER
+        select DRM_KMS_HELPER
         depends on DEBUG_KERNEL
         depends on EXPERT
         help
@@ -128,56 +160,16 @@ config DRM_DEBUG_MODESET_LOCK
 
 	  If in doubt, say "N".
 
-config DRM_FBDEV_EMULATION
-	bool "Enable legacy fbdev support for your modesetting driver"
-	depends on DRM_KMS_HELPER
-	depends on FB=y || FB=DRM_KMS_HELPER
-	select FB_CFB_FILLRECT
-	select FB_CFB_COPYAREA
-	select FB_CFB_IMAGEBLIT
-	select FB_DEFERRED_IO
-	select FB_SYS_FOPS
-	select FB_SYS_FILLRECT
-	select FB_SYS_COPYAREA
-	select FB_SYS_IMAGEBLIT
-	select FRAMEBUFFER_CONSOLE if !EXPERT
-	select FRAMEBUFFER_CONSOLE_DETECT_PRIMARY if FRAMEBUFFER_CONSOLE
-	default y
-	help
-	  Choose this option if you have a need for the legacy fbdev
-	  support. Note that this support also provides the linux console
-	  support on top of your modesetting driver.
-
-	  If in doubt, say "Y".
-
-config DRM_FBDEV_OVERALLOC
-	int "Overallocation of the fbdev buffer"
-	depends on DRM_FBDEV_EMULATION
-	default 100
-	help
-	  Defines the fbdev buffer overallocation in percent. Default
-	  is 100. Typical values for double buffering will be 200,
-	  triple buffering 300.
-
-config DRM_FBDEV_LEAK_PHYS_SMEM
-	bool "Shamelessly allow leaking of fbdev physical address (DANGEROUS)"
-	depends on DRM_FBDEV_EMULATION && EXPERT
-	default n
+config DRM_CLIENT
+	bool
+	depends on DRM
 	help
-	  In order to keep user-space compatibility, we want in certain
-	  use-cases to keep leaking the fbdev physical address to the
-	  user-space program handling the fbdev buffer.
-	  This affects, not only, Amlogic, Allwinner or Rockchip devices
-	  with ARM Mali GPUs using an userspace Blob.
-	  This option is not supported by upstream developers and should be
-	  removed as soon as possible and be considered as a broken and
-	  legacy behaviour from a modern fbdev device driver.
+	  Enables support for DRM clients. DRM drivers that need
+	  struct drm_client_dev and its interfaces should select this
+	  option. Drivers that support the default clients should
+	  select DRM_CLIENT_SELECTION instead.
 
-	  Please send any bug reports when using this to your proprietary
-	  software vendor that requires this.
-
-	  If in doubt, say "N" or spread the word to your closed source
-	  library vendor.
+source "drivers/gpu/drm/clients/Kconfig"
 
 config DRM_LOAD_EDID_FIRMWARE
 	bool "Allow to specify an EDID data set instead of probing for it"
@@ -196,11 +188,35 @@ source "drivers/gpu/drm/display/Kconfig"
 config DRM_TTM
 	tristate
 	depends on DRM && MMU
+	select SHMEM
 	help
 	  GPU memory management subsystem for devices with multiple
 	  GPU memory types. Will be enabled automatically if a device driver
 	  uses it.
 
+config DRM_EXEC
+	tristate
+	depends on DRM
+	help
+	  Execution context for command submissions
+
+config DRM_GPUVM
+	tristate
+	depends on DRM
+	select DRM_EXEC
+	help
+	  GPU-VM representation providing helpers to manage a GPUs virtual
+	  address space
+
+config DRM_GPUSVM
+	tristate
+	depends on DRM && DEVICE_PRIVATE
+	select HMM_MIRROR
+	select MMU_NOTIFIER
+	help
+	  GPU-SVM representation providing helpers to manage a GPUs shared
+	  virtual memory
+
 config DRM_BUDDY
 	tristate
 	depends on DRM
@@ -217,18 +233,27 @@ config DRM_TTM_HELPER
 	tristate
 	depends on DRM
 	select DRM_TTM
+	select DRM_KMS_HELPER if DRM_FBDEV_EMULATION
+	select FB_CORE if DRM_FBDEV_EMULATION
+	select FB_SYSMEM_HELPERS_DEFERRED if DRM_FBDEV_EMULATION
 	help
 	  Helpers for ttm-based gem objects
 
 config DRM_GEM_DMA_HELPER
 	tristate
 	depends on DRM
+	select DRM_KMS_HELPER if DRM_FBDEV_EMULATION
+	select FB_CORE if DRM_FBDEV_EMULATION
+	select FB_DMAMEM_HELPERS_DEFERRED if DRM_FBDEV_EMULATION
 	help
 	  Choose this if you need the GEM DMA helper functions
 
 config DRM_GEM_SHMEM_HELPER
 	tristate
 	depends on DRM && MMU
+	select DRM_KMS_HELPER if DRM_FBDEV_EMULATION
+	select FB_CORE if DRM_FBDEV_EMULATION
+	select FB_SYSMEM_HELPERS_DEFERRED if DRM_FBDEV_EMULATION
 	help
 	  Choose this if you need the GEM shmem helper functions
 
@@ -240,7 +265,7 @@ config DRM_SCHED
 	tristate
 	depends on DRM
 
-source "drivers/gpu/drm/i2c/Kconfig"
+source "drivers/gpu/drm/sysfb/Kconfig"
 
 source "drivers/gpu/drm/arm/Kconfig"
 
@@ -250,8 +275,12 @@ source "drivers/gpu/drm/amd/amdgpu/Kconfig"
 
 source "drivers/gpu/drm/nouveau/Kconfig"
 
+source "drivers/gpu/drm/nova/Kconfig"
+
 source "drivers/gpu/drm/i915/Kconfig"
 
+source "drivers/gpu/drm/xe/Kconfig"
+
 source "drivers/gpu/drm/kmb/Kconfig"
 
 config DRM_VGEM
@@ -263,19 +292,7 @@ config DRM_VGEM
 	  as used by Mesa's software renderer for enhanced performance.
 	  If M is selected the module will be called vgem.
 
-config DRM_VKMS
-	tristate "Virtual KMS (EXPERIMENTAL)"
-	depends on DRM && MMU
-	select DRM_KMS_HELPER
-	select DRM_GEM_SHMEM_HELPER
-	select CRC32
-	default n
-	help
-	  Virtual Kernel Mode-Setting (VKMS) is used for testing or for
-	  running GPU in a headless machines. Choose this option to get
-	  a VKMS.
-
-	  If M is selected the module will be called vkms.
+source "drivers/gpu/drm/vkms/Kconfig"
 
 source "drivers/gpu/drm/exynos/Kconfig"
 
@@ -295,9 +312,7 @@ source "drivers/gpu/drm/armada/Kconfig"
 
 source "drivers/gpu/drm/atmel-hlcdc/Kconfig"
 
-source "drivers/gpu/drm/rcar-du/Kconfig"
-
-source "drivers/gpu/drm/shmobile/Kconfig"
+source "drivers/gpu/drm/renesas/Kconfig"
 
 source "drivers/gpu/drm/sun4i/Kconfig"
 
@@ -331,6 +346,8 @@ source "drivers/gpu/drm/v3d/Kconfig"
 
 source "drivers/gpu/drm/vc4/Kconfig"
 
+source "drivers/gpu/drm/loongson/Kconfig"
+
 source "drivers/gpu/drm/etnaviv/Kconfig"
 
 source "drivers/gpu/drm/hisilicon/Kconfig"
@@ -357,23 +374,34 @@ source "drivers/gpu/drm/lima/Kconfig"
 
 source "drivers/gpu/drm/panfrost/Kconfig"
 
+source "drivers/gpu/drm/panthor/Kconfig"
+
 source "drivers/gpu/drm/aspeed/Kconfig"
 
 source "drivers/gpu/drm/mcde/Kconfig"
 
 source "drivers/gpu/drm/tidss/Kconfig"
 
+source "drivers/gpu/drm/adp/Kconfig"
+
 source "drivers/gpu/drm/xlnx/Kconfig"
 
 source "drivers/gpu/drm/gud/Kconfig"
 
+source "drivers/gpu/drm/sitronix/Kconfig"
+
 source "drivers/gpu/drm/solomon/Kconfig"
 
 source "drivers/gpu/drm/sprd/Kconfig"
 
+source "drivers/gpu/drm/imagination/Kconfig"
+
+source "drivers/gpu/drm/tyr/Kconfig"
+
 config DRM_HYPERV
 	tristate "DRM Support for Hyper-V synthetic video device"
-	depends on DRM && PCI && MMU && HYPERV
+	depends on DRM && PCI && HYPERV_VMBUS
+	select DRM_CLIENT_SELECTION
 	select DRM_KMS_HELPER
 	select DRM_GEM_SHMEM_HELPER
 	help
@@ -384,32 +412,8 @@ config DRM_HYPERV
 
 	 If M is selected the module will be called hyperv_drm.
 
-# Keep legacy drivers last
-
-menuconfig DRM_LEGACY
-	bool "Enable legacy drivers (DANGEROUS)"
-	depends on DRM && MMU
-	help
-	  Enable legacy DRI1 drivers. Those drivers expose unsafe and dangerous
-	  APIs to user-space, which can be used to circumvent access
-	  restrictions and other security measures. For backwards compatibility
-	  those drivers are still available, but their use is highly
-	  inadvisable and might harm your system.
-
-	  You are recommended to use the safe modeset-only drivers instead, and
-	  perform 3D emulation in user-space.
-
-	  Unless you have strong reasons to go rogue, say "N".
-
-if DRM_LEGACY
-# leave here to list legacy drivers
-endif # DRM_LEGACY
-
-config DRM_EXPORT_FOR_TESTS
-	bool
-
-# Separate option because drm_panel_orientation_quirks.c is shared with fbdev
-config DRM_PANEL_ORIENTATION_QUIRKS
+# Separate option as not all DRM drivers use it
+config DRM_PANEL_BACKLIGHT_QUIRKS
 	tristate
 
 config DRM_LIB_RANDOM
@@ -419,3 +423,9 @@ config DRM_LIB_RANDOM
 config DRM_PRIVACY_SCREEN
 	bool
 	default n
+
+endif
+
+# Separate option because drm_panel_orientation_quirks.c is shared with fbdev
+config DRM_PANEL_ORIENTATION_QUIRKS
+	tristate
diff --git a/drivers/gpu/drm/Kconfig.debug b/drivers/gpu/drm/Kconfig.debug
new file mode 100644
index 000000000000..05dc43c0b8c5
--- /dev/null
+++ b/drivers/gpu/drm/Kconfig.debug
@@ -0,0 +1,117 @@
+config DRM_USE_DYNAMIC_DEBUG
+	bool "use dynamic debug to implement drm.debug"
+	default n
+	depends on BROKEN
+	depends on DRM
+	depends on DYNAMIC_DEBUG || DYNAMIC_DEBUG_CORE
+	depends on JUMP_LABEL
+	help
+	 Use dynamic-debug to avoid drm_debug_enabled() runtime overheads.
+	 Due to callsite counts in DRM drivers (~4k in amdgpu) and 56
+	 bytes per callsite, the .data costs can be substantial, and
+	 are therefore configurable.
+
+config DRM_WERROR
+	bool "Compile the drm subsystem with warnings as errors"
+	depends on DRM && EXPERT
+	depends on !WERROR
+	default n
+	help
+	  A kernel build should not cause any compiler warnings, and this
+	  enables the '-Werror' flag to enforce that rule in the drm subsystem.
+
+	  The drm subsystem enables more warnings than the kernel default, so
+	  this config option is disabled by default.
+
+	  If in doubt, say N.
+
+config DRM_HEADER_TEST
+	bool "Ensure DRM headers are self-contained and pass kernel-doc"
+	depends on DRM && EXPERT && BROKEN
+	default n
+	help
+	  Ensure the DRM subsystem headers both under drivers/gpu/drm and
+	  include/drm compile, are self-contained, have header guards, and have
+	  no kernel-doc warnings.
+
+	  If in doubt, say N.
+
+config DRM_DEBUG_MM
+	bool "Insert extra checks and debug info into the DRM range managers"
+	default n
+	depends on DRM
+	depends on STACKTRACE_SUPPORT
+	select STACKDEPOT
+	help
+	  Enable allocation tracking of memory manager and leak detection on
+	  shutdown.
+
+	  Recommended for driver developers only.
+
+	  If in doubt, say "N".
+
+config DRM_KUNIT_TEST_HELPERS
+	tristate
+	depends on DRM && KUNIT
+	select DRM_KMS_HELPER
+	help
+	  KUnit Helpers for KMS drivers.
+
+config DRM_KUNIT_TEST
+	tristate "KUnit tests for DRM" if !KUNIT_ALL_TESTS
+	depends on DRM && KUNIT && MMU
+	select DRM_BRIDGE_CONNECTOR
+	select DRM_BUDDY
+	select DRM_DISPLAY_DP_HELPER
+	select DRM_DISPLAY_HDMI_STATE_HELPER
+	select DRM_DISPLAY_HELPER
+	select DRM_EXEC
+	select DRM_EXPORT_FOR_TESTS if m
+	select DRM_GEM_SHMEM_HELPER
+	select DRM_KUNIT_TEST_HELPERS
+	select DRM_LIB_RANDOM
+	select DRM_SYSFB_HELPER
+	select PRIME_NUMBERS
+	default KUNIT_ALL_TESTS
+	help
+	  This builds unit tests for DRM. This option is not useful for
+	  distributions or general kernels, but only for kernel
+	  developers working on DRM and associated drivers.
+
+	  For more information on KUnit and unit tests in general,
+	  please refer to the KUnit documentation in
+	  Documentation/dev-tools/kunit/.
+
+	  If in doubt, say "N".
+
+config DRM_TTM_KUNIT_TEST
+	tristate "KUnit tests for TTM" if !KUNIT_ALL_TESTS
+	default n
+	depends on DRM && KUNIT && MMU && (UML || COMPILE_TEST)
+	select DRM_TTM
+	select DRM_BUDDY
+	select DRM_EXPORT_FOR_TESTS if m
+	select DRM_KUNIT_TEST_HELPERS
+	default KUNIT_ALL_TESTS
+	help
+	  Enables unit tests for TTM, a GPU memory manager subsystem used
+	  to manage memory buffers. This option is mostly useful for kernel
+	  developers. It depends on (UML || COMPILE_TEST) since no other driver
+	  which uses TTM can be loaded while running the tests.
+
+	  If in doubt, say "N".
+
+config DRM_SCHED_KUNIT_TEST
+	tristate "KUnit tests for the DRM scheduler" if !KUNIT_ALL_TESTS
+	select DRM_SCHED
+	depends on DRM && KUNIT
+	default KUNIT_ALL_TESTS
+	help
+	  Choose this option to build unit tests for the DRM scheduler.
+
+	  Recommended for driver developers only.
+
+	  If in doubt, say "N".
+
+config DRM_EXPORT_FOR_TESTS
+	bool
diff --git a/drivers/gpu/drm/Makefile b/drivers/gpu/drm/Makefile
index a33257d2bc7f..0e1c668b46d2 100644
--- a/drivers/gpu/drm/Makefile
+++ b/drivers/gpu/drm/Makefile
@@ -5,23 +5,50 @@
 
 CFLAGS-$(CONFIG_DRM_USE_DYNAMIC_DEBUG)	+= -DDYNAMIC_DEBUG_MODULE
 
+# Unconditionally enable W=1 warnings locally
+# --- begin copy-paste W=1 warnings from scripts/Makefile.warn
+subdir-ccflags-y += -Wextra -Wunused -Wno-unused-parameter
+subdir-ccflags-y += $(call cc-option, -Wrestrict)
+subdir-ccflags-y += -Wmissing-format-attribute
+subdir-ccflags-y += -Wold-style-definition
+subdir-ccflags-y += -Wmissing-include-dirs
+subdir-ccflags-y += $(call cc-option, -Wunused-but-set-variable)
+subdir-ccflags-y += $(call cc-option, -Wunused-const-variable)
+subdir-ccflags-y += $(call cc-option, -Wpacked-not-aligned)
+subdir-ccflags-y += $(call cc-option, -Wformat-overflow)
+# FIXME: fix -Wformat-truncation warnings and uncomment
+#subdir-ccflags-y += $(call cc-option, -Wformat-truncation)
+subdir-ccflags-y += $(call cc-option, -Wstringop-truncation)
+# The following turn off the warnings enabled by -Wextra
+ifeq ($(findstring 2, $(KBUILD_EXTRA_WARN)),)
+subdir-ccflags-y += -Wno-missing-field-initializers
+subdir-ccflags-y += -Wno-type-limits
+subdir-ccflags-y += -Wno-shift-negative-value
+endif
+ifeq ($(findstring 3, $(KBUILD_EXTRA_WARN)),)
+subdir-ccflags-y += -Wno-sign-compare
+endif
+# --- end copy-paste
+
+# Enable -Werror in CI and development
+subdir-ccflags-$(CONFIG_DRM_WERROR) += -Werror
+
 drm-y := \
-	drm_aperture.o \
 	drm_atomic.o \
 	drm_atomic_uapi.o \
 	drm_auth.o \
 	drm_blend.o \
 	drm_bridge.o \
 	drm_cache.o \
-	drm_client.o \
-	drm_client_modeset.o \
 	drm_color_mgmt.o \
+	drm_colorop.o \
 	drm_connector.o \
 	drm_crtc.o \
 	drm_displayid.o \
 	drm_drv.o \
 	drm_dumb_buffers.o \
 	drm_edid.o \
+	drm_eld.o \
 	drm_encoder.o \
 	drm_file.o \
 	drm_fourcc.o \
@@ -39,6 +66,7 @@ drm-y := \
 	drm_prime.o \
 	drm_print.o \
 	drm_property.o \
+	drm_rect.o \
 	drm_syncobj.o \
 	drm_sysfs.o \
 	drm_trace_points.o \
@@ -46,18 +74,11 @@ drm-y := \
 	drm_vblank_work.o \
 	drm_vma_manager.o \
 	drm_writeback.o
-drm-$(CONFIG_DRM_LEGACY) += \
-	drm_agpsupport.o \
-	drm_bufs.o \
-	drm_context.o \
-	drm_dma.o \
-	drm_hashtab.o \
-	drm_irq.o \
-	drm_legacy_misc.o \
-	drm_lock.o \
-	drm_memory.o \
-	drm_scatter.o \
-	drm_vm.o
+drm-$(CONFIG_DRM_CLIENT) += \
+	drm_client.o \
+	drm_client_event.o \
+	drm_client_modeset.o \
+	drm_client_sysrq.o
 drm-$(CONFIG_DRM_LIB_RANDOM) += lib/drm_random.o
 drm-$(CONFIG_COMPAT) += drm_ioc32.o
 drm-$(CONFIG_DRM_PANEL) += drm_panel.o
@@ -71,13 +92,25 @@ drm-$(CONFIG_DRM_PRIVACY_SCREEN) += \
 	drm_privacy_screen.o \
 	drm_privacy_screen_x86.o
 drm-$(CONFIG_DRM_ACCEL) += ../../accel/drm_accel.o
+drm-$(CONFIG_DRM_PANIC) += drm_panic.o
+drm-$(CONFIG_DRM_DRAW) += drm_draw.o
+drm-$(CONFIG_DRM_PANIC_SCREEN_QR_CODE) += drm_panic_qr.o
 obj-$(CONFIG_DRM)	+= drm.o
 
 obj-$(CONFIG_DRM_PANEL_ORIENTATION_QUIRKS) += drm_panel_orientation_quirks.o
+obj-$(CONFIG_DRM_PANEL_BACKLIGHT_QUIRKS) += drm_panel_backlight_quirks.o
 
 #
 # Memory-management helpers
 #
+#
+obj-$(CONFIG_DRM_EXEC) += drm_exec.o
+obj-$(CONFIG_DRM_GPUVM) += drm_gpuvm.o
+
+drm_gpusvm_helper-y := \
+	drm_gpusvm.o\
+	drm_pagemap.o
+obj-$(CONFIG_DRM_GPUSVM) += drm_gpusvm_helper.o
 
 obj-$(CONFIG_DRM_BUDDY) += drm_buddy.o
 
@@ -87,6 +120,7 @@ drm_dma_helper-$(CONFIG_DRM_KMS_HELPER) += drm_fb_dma_helper.o
 obj-$(CONFIG_DRM_GEM_DMA_HELPER) += drm_dma_helper.o
 
 drm_shmem_helper-y := drm_gem_shmem_helper.o
+drm_shmem_helper-$(CONFIG_DRM_FBDEV_EMULATION) += drm_fbdev_shmem.o
 obj-$(CONFIG_DRM_GEM_SHMEM_HELPER) += drm_shmem_helper.o
 
 drm_suballoc_helper-y := drm_suballoc.o
@@ -96,6 +130,7 @@ drm_vram_helper-y := drm_gem_vram_helper.o
 obj-$(CONFIG_DRM_VRAM_HELPER) += drm_vram_helper.o
 
 drm_ttm_helper-y := drm_gem_ttm_helper.o
+drm_ttm_helper-$(CONFIG_DRM_FBDEV_EMULATION) += drm_fbdev_ttm.o
 obj-$(CONFIG_DRM_TTM_HELPER) += drm_ttm_helper.o
 
 #
@@ -105,10 +140,9 @@ obj-$(CONFIG_DRM_TTM_HELPER) += drm_ttm_helper.o
 drm_kms_helper-y := \
 	drm_atomic_helper.o \
 	drm_atomic_state_helper.o \
-	drm_bridge_connector.o \
+	drm_bridge_helper.o \
 	drm_crtc_helper.o \
 	drm_damage_helper.o \
-	drm_encoder_slave.o \
 	drm_flip_work.o \
 	drm_format_helper.o \
 	drm_gem_atomic_helper.o \
@@ -117,13 +151,11 @@ drm_kms_helper-y := \
 	drm_modeset_helper.o \
 	drm_plane_helper.o \
 	drm_probe_helper.o \
-	drm_rect.o \
 	drm_self_refresh_helper.o \
-	drm_simple_kms_helper.o
+	drm_simple_kms_helper.o \
+	drm_vblank_helper.o
 drm_kms_helper-$(CONFIG_DRM_PANEL_BRIDGE) += bridge/panel.o
-drm_kms_helper-$(CONFIG_DRM_FBDEV_EMULATION) += \
-	drm_fbdev_generic.o \
-	drm_fb_helper.o
+drm_kms_helper-$(CONFIG_DRM_FBDEV_EMULATION) += drm_fb_helper.o
 obj-$(CONFIG_DRM_KMS_HELPER) += drm_kms_helper.o
 
 #
@@ -135,12 +167,15 @@ obj-y			+= tests/
 obj-$(CONFIG_DRM_MIPI_DBI) += drm_mipi_dbi.o
 obj-$(CONFIG_DRM_MIPI_DSI) += drm_mipi_dsi.o
 obj-y			+= arm/
+obj-y			+= clients/
 obj-y			+= display/
 obj-$(CONFIG_DRM_TTM)	+= ttm/
 obj-$(CONFIG_DRM_SCHED)	+= scheduler/
 obj-$(CONFIG_DRM_RADEON)+= radeon/
 obj-$(CONFIG_DRM_AMDGPU)+= amd/amdgpu/
+obj-$(CONFIG_DRM_AMDGPU)+= amd/amdxcp/
 obj-$(CONFIG_DRM_I915)	+= i915/
+obj-$(CONFIG_DRM_XE)	+= xe/
 obj-$(CONFIG_DRM_KMB_DISPLAY)  += kmb/
 obj-$(CONFIG_DRM_MGAG200) += mgag200/
 obj-$(CONFIG_DRM_V3D)  += v3d/
@@ -149,6 +184,7 @@ obj-$(CONFIG_DRM_VMWGFX)+= vmwgfx/
 obj-$(CONFIG_DRM_VGEM)	+= vgem/
 obj-$(CONFIG_DRM_VKMS)	+= vkms/
 obj-$(CONFIG_DRM_NOUVEAU) +=nouveau/
+obj-$(CONFIG_DRM_NOVA) += nova/
 obj-$(CONFIG_DRM_EXYNOS) +=exynos/
 obj-$(CONFIG_DRM_ROCKCHIP) +=rockchip/
 obj-$(CONFIG_DRM_GMA500) += gma500/
@@ -156,8 +192,7 @@ obj-$(CONFIG_DRM_UDL) += udl/
 obj-$(CONFIG_DRM_AST) += ast/
 obj-$(CONFIG_DRM_ARMADA) += armada/
 obj-$(CONFIG_DRM_ATMEL_HLCDC)	+= atmel-hlcdc/
-obj-y			+= rcar-du/
-obj-$(CONFIG_DRM_SHMOBILE) +=shmobile/
+obj-y			+= renesas/
 obj-y			+= omapdrm/
 obj-$(CONFIG_DRM_SUN4I) += sun4i/
 obj-y			+= tilcdc/
@@ -172,25 +207,49 @@ obj-$(CONFIG_DRM_INGENIC) += ingenic/
 obj-$(CONFIG_DRM_LOGICVC) += logicvc/
 obj-$(CONFIG_DRM_MEDIATEK) += mediatek/
 obj-$(CONFIG_DRM_MESON)	+= meson/
-obj-y			+= i2c/
 obj-y			+= panel/
 obj-y			+= bridge/
 obj-$(CONFIG_DRM_FSL_DCU) += fsl-dcu/
 obj-$(CONFIG_DRM_ETNAVIV) += etnaviv/
 obj-y			+= hisilicon/
 obj-y			+= mxsfb/
+obj-y			+= sysfb/
 obj-y			+= tiny/
 obj-$(CONFIG_DRM_PL111) += pl111/
 obj-$(CONFIG_DRM_TVE200) += tve200/
+obj-$(CONFIG_DRM_ADP) += adp/
 obj-$(CONFIG_DRM_XEN) += xen/
 obj-$(CONFIG_DRM_VBOXVIDEO) += vboxvideo/
 obj-$(CONFIG_DRM_LIMA)  += lima/
 obj-$(CONFIG_DRM_PANFROST) += panfrost/
+obj-$(CONFIG_DRM_PANTHOR) += panthor/
+obj-$(CONFIG_DRM_TYR) += tyr/
 obj-$(CONFIG_DRM_ASPEED_GFX) += aspeed/
 obj-$(CONFIG_DRM_MCDE) += mcde/
 obj-$(CONFIG_DRM_TIDSS) += tidss/
 obj-y			+= xlnx/
 obj-y			+= gud/
 obj-$(CONFIG_DRM_HYPERV) += hyperv/
+obj-y			+= sitronix/
 obj-y			+= solomon/
 obj-$(CONFIG_DRM_SPRD) += sprd/
+obj-$(CONFIG_DRM_LOONGSON) += loongson/
+obj-$(CONFIG_DRM_POWERVR) += imagination/
+
+# Ensure drm headers are self-contained and pass kernel-doc
+hdrtest-files := \
+	$(shell cd $(src) && find . -maxdepth 1 -name 'drm_*.h') \
+	$(shell cd $(src) && find display lib -name '*.h')
+
+always-$(CONFIG_DRM_HEADER_TEST) += \
+	$(patsubst %.h,%.hdrtest, $(hdrtest-files))
+
+# Include the header twice to detect missing include guard.
+quiet_cmd_hdrtest = HDRTEST $(patsubst %.hdrtest,%.h,$@)
+      cmd_hdrtest = \
+		$(CC) $(c_flags) -fsyntax-only -x c /dev/null -include $< -include $<; \
+		 PYTHONDONTWRITEBYTECODE=1 $(PYTHON3) $(KERNELDOC) -none $(if $(CONFIG_WERROR)$(CONFIG_DRM_WERROR),-Werror) $<; \
+		touch $@
+
+$(obj)/%.hdrtest: $(src)/%.h FORCE
+	$(call if_changed_dep,hdrtest)
diff --git a/drivers/gpu/drm/adp/Kconfig b/drivers/gpu/drm/adp/Kconfig
new file mode 100644
index 000000000000..9fcc27eb200d
--- /dev/null
+++ b/drivers/gpu/drm/adp/Kconfig
@@ -0,0 +1,17 @@
+# SPDX-License-Identifier: GPL-2.0-only OR MIT
+config DRM_ADP
+	tristate "DRM Support for pre-DCP Apple display controllers"
+	depends on DRM && OF && ARM64
+	depends on ARCH_APPLE || COMPILE_TEST
+	select DRM_KMS_HELPER
+	select DRM_BRIDGE_CONNECTOR
+	select DRM_DISPLAY_HELPER
+	select DRM_KMS_DMA_HELPER
+	select DRM_GEM_DMA_HELPER
+	select DRM_PANEL_BRIDGE
+	select VIDEOMODE_HELPERS
+	select DRM_MIPI_DSI
+	help
+	  Chose this option if you have an Apple Arm laptop with a touchbar.
+
+	  If M is selected, this module will be called adpdrm.
diff --git a/drivers/gpu/drm/adp/Makefile b/drivers/gpu/drm/adp/Makefile
new file mode 100644
index 000000000000..8e7b618edd35
--- /dev/null
+++ b/drivers/gpu/drm/adp/Makefile
@@ -0,0 +1,5 @@
+# SPDX-License-Identifier: GPL-2.0-only OR MIT
+
+adpdrm-y := adp_drv.o
+adpdrm-mipi-y := adp-mipi.o
+obj-$(CONFIG_DRM_ADP) += adpdrm.o adpdrm-mipi.o
diff --git a/drivers/gpu/drm/adp/adp-mipi.c b/drivers/gpu/drm/adp/adp-mipi.c
new file mode 100644
index 000000000000..cba7d32150a9
--- /dev/null
+++ b/drivers/gpu/drm/adp/adp-mipi.c
@@ -0,0 +1,277 @@
+// SPDX-License-Identifier: GPL-2.0-only
+
+#include <linux/component.h>
+#include <linux/iopoll.h>
+#include <linux/of.h>
+#include <linux/platform_device.h>
+
+#include <drm/drm_bridge.h>
+#include <drm/drm_mipi_dsi.h>
+
+#define DSI_GEN_HDR 0x6c
+#define DSI_GEN_PLD_DATA 0x70
+
+#define DSI_CMD_PKT_STATUS 0x74
+
+#define GEN_PLD_R_EMPTY BIT(4)
+#define GEN_PLD_W_FULL BIT(3)
+#define GEN_PLD_W_EMPTY BIT(2)
+#define GEN_CMD_FULL BIT(1)
+#define GEN_CMD_EMPTY BIT(0)
+#define GEN_RD_CMD_BUSY BIT(6)
+#define CMD_PKT_STATUS_TIMEOUT_US 20000
+
+struct adp_mipi_drv_private {
+	struct mipi_dsi_host dsi;
+	struct drm_bridge bridge;
+	struct drm_bridge *next_bridge;
+	void __iomem *mipi;
+};
+
+#define mipi_to_adp(x) container_of(x, struct adp_mipi_drv_private, dsi)
+
+static int adp_dsi_gen_pkt_hdr_write(struct adp_mipi_drv_private *adp, u32 hdr_val)
+{
+	int ret;
+	u32 val, mask;
+
+	ret = readl_poll_timeout(adp->mipi + DSI_CMD_PKT_STATUS,
+				 val, !(val & GEN_CMD_FULL), 1000,
+				 CMD_PKT_STATUS_TIMEOUT_US);
+	if (ret) {
+		dev_err(adp->dsi.dev, "failed to get available command FIFO\n");
+		return ret;
+	}
+
+	writel(hdr_val, adp->mipi + DSI_GEN_HDR);
+
+	mask = GEN_CMD_EMPTY | GEN_PLD_W_EMPTY;
+	ret = readl_poll_timeout(adp->mipi + DSI_CMD_PKT_STATUS,
+				 val, (val & mask) == mask,
+				 1000, CMD_PKT_STATUS_TIMEOUT_US);
+	if (ret) {
+		dev_err(adp->dsi.dev, "failed to write command FIFO\n");
+		return ret;
+	}
+
+	return 0;
+}
+
+static int adp_dsi_write(struct adp_mipi_drv_private *adp,
+			 const struct mipi_dsi_packet *packet)
+{
+	const u8 *tx_buf = packet->payload;
+	int len = packet->payload_length, pld_data_bytes = sizeof(u32), ret;
+	__le32 word;
+	u32 val;
+
+	while (len) {
+		if (len < pld_data_bytes) {
+			word = 0;
+			memcpy(&word, tx_buf, len);
+			writel(le32_to_cpu(word), adp->mipi + DSI_GEN_PLD_DATA);
+			len = 0;
+		} else {
+			memcpy(&word, tx_buf, pld_data_bytes);
+			writel(le32_to_cpu(word), adp->mipi + DSI_GEN_PLD_DATA);
+			tx_buf += pld_data_bytes;
+			len -= pld_data_bytes;
+		}
+
+		ret = readl_poll_timeout(adp->mipi + DSI_CMD_PKT_STATUS,
+					 val, !(val & GEN_PLD_W_FULL), 1000,
+					 CMD_PKT_STATUS_TIMEOUT_US);
+		if (ret) {
+			dev_err(adp->dsi.dev,
+				"failed to get available write payload FIFO\n");
+			return ret;
+		}
+	}
+
+	word = 0;
+	memcpy(&word, packet->header, sizeof(packet->header));
+	return adp_dsi_gen_pkt_hdr_write(adp, le32_to_cpu(word));
+}
+
+static int adp_dsi_read(struct adp_mipi_drv_private *adp,
+			const struct mipi_dsi_msg *msg)
+{
+	int i, j, ret, len = msg->rx_len;
+	u8 *buf = msg->rx_buf;
+	u32 val;
+
+	/* Wait end of the read operation */
+	ret = readl_poll_timeout(adp->mipi + DSI_CMD_PKT_STATUS,
+				 val, !(val & GEN_RD_CMD_BUSY),
+				 1000, CMD_PKT_STATUS_TIMEOUT_US);
+	if (ret) {
+		dev_err(adp->dsi.dev, "Timeout during read operation\n");
+		return ret;
+	}
+
+	for (i = 0; i < len; i += 4) {
+		/* Read fifo must not be empty before all bytes are read */
+		ret = readl_poll_timeout(adp->mipi + DSI_CMD_PKT_STATUS,
+					 val, !(val & GEN_PLD_R_EMPTY),
+					 1000, CMD_PKT_STATUS_TIMEOUT_US);
+		if (ret) {
+			dev_err(adp->dsi.dev, "Read payload FIFO is empty\n");
+			return ret;
+		}
+
+		val = readl(adp->mipi + DSI_GEN_PLD_DATA);
+		for (j = 0; j < 4 && j + i < len; j++)
+			buf[i + j] = val >> (8 * j);
+	}
+
+	return ret;
+}
+
+static ssize_t adp_dsi_host_transfer(struct mipi_dsi_host *host,
+				     const struct mipi_dsi_msg *msg)
+{
+	struct adp_mipi_drv_private *adp = mipi_to_adp(host);
+	struct mipi_dsi_packet packet;
+	int ret, nb_bytes;
+
+	ret = mipi_dsi_create_packet(&packet, msg);
+	if (ret) {
+		dev_err(adp->dsi.dev, "failed to create packet: %d\n", ret);
+		return ret;
+	}
+
+	ret = adp_dsi_write(adp, &packet);
+	if (ret)
+		return ret;
+
+	if (msg->rx_buf && msg->rx_len) {
+		ret = adp_dsi_read(adp, msg);
+		if (ret)
+			return ret;
+		nb_bytes = msg->rx_len;
+	} else {
+		nb_bytes = packet.size;
+	}
+
+	return nb_bytes;
+}
+
+static int adp_dsi_bind(struct device *dev, struct device *master, void *data)
+{
+	return 0;
+}
+
+static void adp_dsi_unbind(struct device *dev, struct device *master, void *data)
+{
+}
+
+static const struct component_ops adp_dsi_component_ops = {
+	.bind	= adp_dsi_bind,
+	.unbind	= adp_dsi_unbind,
+};
+
+static int adp_dsi_host_attach(struct mipi_dsi_host *host,
+			       struct mipi_dsi_device *dev)
+{
+	struct adp_mipi_drv_private *adp = mipi_to_adp(host);
+	struct drm_bridge *next;
+	int ret;
+
+	next = devm_drm_of_get_bridge(adp->dsi.dev, adp->dsi.dev->of_node, 1, 0);
+	if (IS_ERR(next))
+		return PTR_ERR(next);
+
+	adp->next_bridge = next;
+
+	drm_bridge_add(&adp->bridge);
+
+	ret = component_add(host->dev, &adp_dsi_component_ops);
+	if (ret) {
+		pr_err("failed to add dsi_host component: %d\n", ret);
+		drm_bridge_remove(&adp->bridge);
+		return ret;
+	}
+
+	return 0;
+}
+
+static int adp_dsi_host_detach(struct mipi_dsi_host *host,
+			       struct mipi_dsi_device *dev)
+{
+	struct adp_mipi_drv_private *adp = mipi_to_adp(host);
+
+	component_del(host->dev, &adp_dsi_component_ops);
+	drm_bridge_remove(&adp->bridge);
+	return 0;
+}
+
+static const struct mipi_dsi_host_ops adp_dsi_host_ops = {
+	.transfer = adp_dsi_host_transfer,
+	.attach = adp_dsi_host_attach,
+	.detach = adp_dsi_host_detach,
+};
+
+static int adp_dsi_bridge_attach(struct drm_bridge *bridge,
+				 struct drm_encoder *encoder,
+				 enum drm_bridge_attach_flags flags)
+{
+	struct adp_mipi_drv_private *adp =
+		container_of(bridge, struct adp_mipi_drv_private, bridge);
+
+	return drm_bridge_attach(encoder, adp->next_bridge, bridge, flags);
+}
+
+static const struct drm_bridge_funcs adp_dsi_bridge_funcs = {
+	.attach	= adp_dsi_bridge_attach,
+};
+
+static int adp_mipi_probe(struct platform_device *pdev)
+{
+	struct adp_mipi_drv_private *adp;
+
+	adp = devm_drm_bridge_alloc(&pdev->dev, struct adp_mipi_drv_private,
+				    bridge, &adp_dsi_bridge_funcs);
+	if (IS_ERR(adp))
+		return PTR_ERR(adp);
+
+	adp->mipi = devm_platform_ioremap_resource(pdev, 0);
+	if (IS_ERR(adp->mipi)) {
+		dev_err(&pdev->dev, "failed to map mipi mmio");
+		return PTR_ERR(adp->mipi);
+	}
+
+	adp->dsi.dev = &pdev->dev;
+	adp->dsi.ops = &adp_dsi_host_ops;
+	adp->bridge.of_node = pdev->dev.of_node;
+	adp->bridge.type = DRM_MODE_CONNECTOR_DSI;
+	dev_set_drvdata(&pdev->dev, adp);
+	return mipi_dsi_host_register(&adp->dsi);
+}
+
+static void adp_mipi_remove(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct adp_mipi_drv_private *adp = dev_get_drvdata(dev);
+
+	mipi_dsi_host_unregister(&adp->dsi);
+}
+
+static const struct of_device_id adp_mipi_of_match[] = {
+	{ .compatible = "apple,h7-display-pipe-mipi", },
+	{ },
+};
+MODULE_DEVICE_TABLE(of, adp_mipi_of_match);
+
+static struct platform_driver adp_mipi_platform_driver = {
+	.driver = {
+		.name = "adp-mipi",
+		.of_match_table = adp_mipi_of_match,
+	},
+	.probe = adp_mipi_probe,
+	.remove = adp_mipi_remove,
+};
+
+module_platform_driver(adp_mipi_platform_driver);
+
+MODULE_DESCRIPTION("Apple Display Pipe MIPI driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/gpu/drm/adp/adp_drv.c b/drivers/gpu/drm/adp/adp_drv.c
new file mode 100644
index 000000000000..4554cf75565e
--- /dev/null
+++ b/drivers/gpu/drm/adp/adp_drv.c
@@ -0,0 +1,614 @@
+// SPDX-License-Identifier: GPL-2.0-only
+
+#include <linux/component.h>
+#include <linux/iopoll.h>
+#include <linux/of.h>
+#include <linux/platform_device.h>
+
+#include <drm/drm_atomic.h>
+#include <drm/drm_atomic_helper.h>
+#include <drm/drm_bridge.h>
+#include <drm/drm_bridge_connector.h>
+#include <drm/drm_drv.h>
+#include <drm/drm_fb_dma_helper.h>
+#include <drm/drm_framebuffer.h>
+#include <drm/drm_gem_atomic_helper.h>
+#include <drm/drm_gem_dma_helper.h>
+#include <drm/drm_gem_framebuffer_helper.h>
+#include <drm/drm_of.h>
+#include <drm/drm_print.h>
+#include <drm/drm_probe_helper.h>
+#include <drm/drm_vblank.h>
+
+#define ADP_INT_STATUS 0x34
+#define ADP_INT_STATUS_INT_MASK 0x7
+#define ADP_INT_STATUS_VBLANK 0x1
+#define ADP_CTRL 0x100
+#define ADP_CTRL_VBLANK_ON 0x12
+#define ADP_CTRL_FIFO_ON 0x601
+#define ADP_SCREEN_SIZE 0x0c
+#define ADP_SCREEN_HSIZE GENMASK(15, 0)
+#define ADP_SCREEN_VSIZE GENMASK(31, 16)
+
+#define ADBE_FIFO 0x10c0
+#define ADBE_FIFO_SYNC 0xc0000000
+
+#define ADBE_BLEND_BYPASS 0x2020
+#define ADBE_BLEND_EN1 0x2028
+#define ADBE_BLEND_EN2 0x2074
+#define ADBE_BLEND_EN3 0x202c
+#define ADBE_BLEND_EN4 0x2034
+#define ADBE_MASK_BUF 0x2200
+
+#define ADBE_SRC_START 0x4040
+#define ADBE_SRC_SIZE 0x4048
+#define ADBE_DST_START 0x4050
+#define ADBE_DST_SIZE 0x4054
+#define ADBE_STRIDE 0x4038
+#define ADBE_FB_BASE 0x4030
+
+#define ADBE_LAYER_EN1 0x4020
+#define ADBE_LAYER_EN2 0x4068
+#define ADBE_LAYER_EN3 0x40b4
+#define ADBE_LAYER_EN4 0x40f4
+#define ADBE_SCALE_CTL 0x40ac
+#define ADBE_SCALE_CTL_BYPASS 0x100000
+
+#define ADBE_LAYER_CTL 0x1038
+#define ADBE_LAYER_CTL_ENABLE 0x10000
+
+#define ADBE_PIX_FMT 0x402c
+#define ADBE_PIX_FMT_XRGB32 0x53e4001
+
+static int adp_open(struct inode *inode, struct file *filp)
+{
+	/*
+	 * The modesetting driver does not check the non-desktop connector
+	 * property and keeps the device open and locked. If the touchbar daemon
+	 * opens the device first, modesetting breaks the whole X session.
+	 * Simply refuse to open the device for X11 server processes as
+	 * workaround.
+	 */
+	if (current->comm[0] == 'X')
+		return -EBUSY;
+
+	return drm_open(inode, filp);
+}
+
+static const struct file_operations adp_fops = {
+	.owner          = THIS_MODULE,
+	.open           = adp_open,
+	.release        = drm_release,
+	.unlocked_ioctl = drm_ioctl,
+	.compat_ioctl   = drm_compat_ioctl,
+	.poll           = drm_poll,
+	.read           = drm_read,
+	.llseek         = noop_llseek,
+	.mmap           = drm_gem_mmap,
+	.fop_flags      = FOP_UNSIGNED_OFFSET,
+	DRM_GEM_DMA_UNMAPPED_AREA_FOPS
+};
+
+static int adp_drm_gem_dumb_create(struct drm_file *file_priv,
+					struct drm_device *drm,
+					struct drm_mode_create_dumb *args)
+{
+	args->height = ALIGN(args->height, 64);
+	args->size = args->pitch * args->height;
+
+	return drm_gem_dma_dumb_create_internal(file_priv, drm, args);
+}
+
+static const struct drm_driver adp_driver = {
+	.driver_features = DRIVER_GEM | DRIVER_MODESET | DRIVER_ATOMIC,
+	.fops = &adp_fops,
+	DRM_GEM_DMA_DRIVER_OPS_VMAP_WITH_DUMB_CREATE(adp_drm_gem_dumb_create),
+	.name = "adp",
+	.desc = "Apple Display Pipe DRM Driver",
+	.major = 0,
+	.minor = 1,
+};
+
+struct adp_drv_private {
+	struct drm_device drm;
+	struct drm_crtc crtc;
+	struct drm_encoder *encoder;
+	struct drm_connector *connector;
+	struct drm_bridge *next_bridge;
+	void __iomem *be;
+	void __iomem *fe;
+	u32 *mask_buf;
+	u64 mask_buf_size;
+	dma_addr_t mask_iova;
+	int be_irq;
+	int fe_irq;
+	struct drm_pending_vblank_event *event;
+};
+
+#define to_adp(x) container_of(x, struct adp_drv_private, drm)
+#define crtc_to_adp(x) container_of(x, struct adp_drv_private, crtc)
+
+static int adp_plane_atomic_check(struct drm_plane *plane,
+				    struct drm_atomic_state *state)
+{
+	struct drm_plane_state *new_plane_state;
+	struct drm_crtc_state *crtc_state;
+
+	new_plane_state = drm_atomic_get_new_plane_state(state, plane);
+
+	if (!new_plane_state->crtc)
+		return 0;
+
+	crtc_state = drm_atomic_get_crtc_state(state, new_plane_state->crtc);
+	if (IS_ERR(crtc_state))
+		return PTR_ERR(crtc_state);
+
+	return drm_atomic_helper_check_plane_state(new_plane_state,
+						   crtc_state,
+						   DRM_PLANE_NO_SCALING,
+						   DRM_PLANE_NO_SCALING,
+						   true, true);
+}
+
+static void adp_plane_atomic_update(struct drm_plane *plane,
+				    struct drm_atomic_state *state)
+{
+	struct adp_drv_private *adp;
+	struct drm_rect src_rect;
+	struct drm_gem_dma_object *obj;
+	struct drm_framebuffer *fb;
+	struct drm_plane_state *new_state = drm_atomic_get_new_plane_state(state, plane);
+	u32 src_pos, src_size, dst_pos, dst_size;
+
+	if (!plane || !new_state)
+		return;
+
+	fb = new_state->fb;
+	if (!fb)
+		return;
+	adp = to_adp(plane->dev);
+
+	drm_rect_fp_to_int(&src_rect, &new_state->src);
+	src_pos = src_rect.x1 << 16 | src_rect.y1;
+	dst_pos = new_state->dst.x1 << 16 | new_state->dst.y1;
+	src_size = drm_rect_width(&src_rect) << 16 | drm_rect_height(&src_rect);
+	dst_size = drm_rect_width(&new_state->dst) << 16 |
+		drm_rect_height(&new_state->dst);
+	writel(src_pos, adp->be + ADBE_SRC_START);
+	writel(src_size, adp->be + ADBE_SRC_SIZE);
+	writel(dst_pos, adp->be + ADBE_DST_START);
+	writel(dst_size, adp->be + ADBE_DST_SIZE);
+	writel(fb->pitches[0], adp->be + ADBE_STRIDE);
+	obj = drm_fb_dma_get_gem_obj(fb, 0);
+	if (obj)
+		writel(obj->dma_addr + fb->offsets[0], adp->be + ADBE_FB_BASE);
+
+	writel(BIT(0), adp->be + ADBE_LAYER_EN1);
+	writel(BIT(0), adp->be + ADBE_LAYER_EN2);
+	writel(BIT(0), adp->be + ADBE_LAYER_EN3);
+	writel(BIT(0), adp->be + ADBE_LAYER_EN4);
+	writel(ADBE_SCALE_CTL_BYPASS, adp->be + ADBE_SCALE_CTL);
+	writel(ADBE_LAYER_CTL_ENABLE | BIT(0), adp->be + ADBE_LAYER_CTL);
+	writel(ADBE_PIX_FMT_XRGB32, adp->be + ADBE_PIX_FMT);
+}
+
+static void adp_plane_atomic_disable(struct drm_plane *plane,
+				     struct drm_atomic_state *state)
+{
+	struct adp_drv_private *adp = to_adp(plane->dev);
+
+	writel(0x0, adp->be + ADBE_LAYER_EN1);
+	writel(0x0, adp->be + ADBE_LAYER_EN2);
+	writel(0x0, adp->be + ADBE_LAYER_EN3);
+	writel(0x0, adp->be + ADBE_LAYER_EN4);
+	writel(ADBE_LAYER_CTL_ENABLE, adp->be + ADBE_LAYER_CTL);
+}
+
+static const struct drm_plane_helper_funcs adp_plane_helper_funcs = {
+	.atomic_check = adp_plane_atomic_check,
+	.atomic_update = adp_plane_atomic_update,
+	.atomic_disable = adp_plane_atomic_disable,
+	DRM_GEM_SHADOW_PLANE_HELPER_FUNCS
+};
+
+static const struct drm_plane_funcs adp_plane_funcs = {
+	.update_plane = drm_atomic_helper_update_plane,
+	.disable_plane = drm_atomic_helper_disable_plane,
+	DRM_GEM_SHADOW_PLANE_FUNCS
+};
+
+static const u32 plane_formats[] = {
+	DRM_FORMAT_XRGB8888,
+};
+
+#define ALL_CRTCS 1
+
+static struct drm_plane *adp_plane_new(struct adp_drv_private *adp)
+{
+	struct drm_device *drm = &adp->drm;
+	struct drm_plane *plane;
+
+	plane = __drmm_universal_plane_alloc(drm, sizeof(struct drm_plane), 0,
+					     ALL_CRTCS, &adp_plane_funcs,
+					     plane_formats, ARRAY_SIZE(plane_formats),
+					     NULL, DRM_PLANE_TYPE_PRIMARY, "plane");
+	if (IS_ERR(plane)) {
+		drm_err(drm, "failed to allocate plane");
+		return plane;
+	}
+
+	drm_plane_helper_add(plane, &adp_plane_helper_funcs);
+	return plane;
+}
+
+static void adp_enable_vblank(struct adp_drv_private *adp)
+{
+	u32 cur_ctrl;
+
+	writel(ADP_INT_STATUS_INT_MASK, adp->fe + ADP_INT_STATUS);
+
+	cur_ctrl = readl(adp->fe + ADP_CTRL);
+	writel(cur_ctrl | ADP_CTRL_VBLANK_ON, adp->fe + ADP_CTRL);
+}
+
+static int adp_crtc_enable_vblank(struct drm_crtc *crtc)
+{
+	struct drm_device *dev = crtc->dev;
+	struct adp_drv_private *adp = to_adp(dev);
+
+	adp_enable_vblank(adp);
+
+	return 0;
+}
+
+static void adp_disable_vblank(struct adp_drv_private *adp)
+{
+	u32 cur_ctrl;
+
+	cur_ctrl = readl(adp->fe + ADP_CTRL);
+	writel(cur_ctrl & ~ADP_CTRL_VBLANK_ON, adp->fe + ADP_CTRL);
+	writel(ADP_INT_STATUS_INT_MASK, adp->fe + ADP_INT_STATUS);
+}
+
+static void adp_crtc_disable_vblank(struct drm_crtc *crtc)
+{
+	struct drm_device *dev = crtc->dev;
+	struct adp_drv_private *adp = to_adp(dev);
+
+	adp_disable_vblank(adp);
+}
+
+static void adp_crtc_atomic_enable(struct drm_crtc *crtc,
+				   struct drm_atomic_state *state)
+{
+	struct adp_drv_private *adp = crtc_to_adp(crtc);
+
+	writel(BIT(0), adp->be + ADBE_BLEND_EN2);
+	writel(BIT(4), adp->be + ADBE_BLEND_EN1);
+	writel(BIT(0), adp->be + ADBE_BLEND_EN3);
+	writel(BIT(0), adp->be + ADBE_BLEND_BYPASS);
+	writel(BIT(0), adp->be + ADBE_BLEND_EN4);
+	drm_crtc_vblank_on(crtc);
+}
+
+static void adp_crtc_atomic_disable(struct drm_crtc *crtc,
+				    struct drm_atomic_state *state)
+{
+	struct adp_drv_private *adp = crtc_to_adp(crtc);
+	struct drm_crtc_state *old_state = drm_atomic_get_old_crtc_state(state, crtc);
+
+	drm_atomic_helper_disable_planes_on_crtc(old_state, false);
+
+	writel(0x0, adp->be + ADBE_BLEND_EN2);
+	writel(0x0, adp->be + ADBE_BLEND_EN1);
+	writel(0x0, adp->be + ADBE_BLEND_EN3);
+	writel(0x0, adp->be + ADBE_BLEND_BYPASS);
+	writel(0x0, adp->be + ADBE_BLEND_EN4);
+	drm_crtc_vblank_off(crtc);
+}
+
+static void adp_crtc_atomic_flush(struct drm_crtc *crtc,
+				  struct drm_atomic_state *state)
+{
+	u32 frame_num = 1;
+	unsigned long flags;
+	struct adp_drv_private *adp = crtc_to_adp(crtc);
+	struct drm_crtc_state *new_state = drm_atomic_get_new_crtc_state(state, crtc);
+	u64 new_size = ALIGN(new_state->mode.hdisplay *
+			     new_state->mode.vdisplay * 4, PAGE_SIZE);
+
+	if (new_size != adp->mask_buf_size) {
+		if (adp->mask_buf)
+			dma_free_coherent(crtc->dev->dev, adp->mask_buf_size,
+					  adp->mask_buf, adp->mask_iova);
+		adp->mask_buf = NULL;
+		if (new_size != 0) {
+			adp->mask_buf = dma_alloc_coherent(crtc->dev->dev, new_size,
+							   &adp->mask_iova, GFP_KERNEL);
+			memset(adp->mask_buf, 0xFF, new_size);
+			writel(adp->mask_iova, adp->be + ADBE_MASK_BUF);
+		}
+		adp->mask_buf_size = new_size;
+	}
+	writel(ADBE_FIFO_SYNC | frame_num, adp->be + ADBE_FIFO);
+	//FIXME: use adbe flush interrupt
+	if (crtc->state->event) {
+		struct drm_pending_vblank_event *event = crtc->state->event;
+
+		crtc->state->event = NULL;
+		spin_lock_irqsave(&crtc->dev->event_lock, flags);
+
+		if (drm_crtc_vblank_get(crtc) != 0)
+			drm_crtc_send_vblank_event(crtc, event);
+		else
+			adp->event = event;
+
+		spin_unlock_irqrestore(&crtc->dev->event_lock, flags);
+	}
+}
+
+static const struct drm_crtc_funcs adp_crtc_funcs = {
+	.destroy = drm_crtc_cleanup,
+	.set_config = drm_atomic_helper_set_config,
+	.page_flip = drm_atomic_helper_page_flip,
+	.reset = drm_atomic_helper_crtc_reset,
+	.atomic_duplicate_state = drm_atomic_helper_crtc_duplicate_state,
+	.atomic_destroy_state = drm_atomic_helper_crtc_destroy_state,
+	.enable_vblank = adp_crtc_enable_vblank,
+	.disable_vblank = adp_crtc_disable_vblank,
+};
+
+
+static const struct drm_crtc_helper_funcs adp_crtc_helper_funcs = {
+	.atomic_enable = adp_crtc_atomic_enable,
+	.atomic_disable = adp_crtc_atomic_disable,
+	.atomic_flush = adp_crtc_atomic_flush,
+};
+
+static int adp_setup_crtc(struct adp_drv_private *adp)
+{
+	struct drm_device *drm = &adp->drm;
+	struct drm_plane *primary;
+	int ret;
+
+	primary = adp_plane_new(adp);
+	if (IS_ERR(primary))
+		return PTR_ERR(primary);
+
+	ret = drm_crtc_init_with_planes(drm, &adp->crtc, primary,
+					NULL, &adp_crtc_funcs, NULL);
+	if (ret)
+		return ret;
+
+	drm_crtc_helper_add(&adp->crtc, &adp_crtc_helper_funcs);
+	return 0;
+}
+
+static const struct drm_mode_config_funcs adp_mode_config_funcs = {
+	.fb_create = drm_gem_fb_create_with_dirty,
+	.atomic_check = drm_atomic_helper_check,
+	.atomic_commit = drm_atomic_helper_commit,
+};
+
+static int adp_setup_mode_config(struct adp_drv_private *adp)
+{
+	struct drm_device *drm = &adp->drm;
+	int ret;
+	u32 size;
+
+	ret = drmm_mode_config_init(drm);
+	if (ret)
+		return ret;
+
+	/*
+	 * Query screen size restrict the frame buffer size to the screen size
+	 * aligned to the next multiple of 64. This is not necessary but can be
+	 * used as simple check for non-desktop devices.
+	 * Xorg's modesetting driver does not care about the connector
+	 * "non-desktop" property. The max frame buffer width or height can be
+	 * easily checked and a device can be reject if the max width/height is
+	 * smaller than 120 for example.
+	 * Any touchbar daemon is not limited by this small framebuffer size.
+	 */
+	size = readl(adp->fe + ADP_SCREEN_SIZE);
+
+	drm->mode_config.min_width = 32;
+	drm->mode_config.min_height = 32;
+	drm->mode_config.max_width = ALIGN(FIELD_GET(ADP_SCREEN_HSIZE, size), 64);
+	drm->mode_config.max_height = ALIGN(FIELD_GET(ADP_SCREEN_VSIZE, size), 64);
+	drm->mode_config.preferred_depth = 24;
+	drm->mode_config.prefer_shadow = 0;
+	drm->mode_config.funcs = &adp_mode_config_funcs;
+
+	ret = adp_setup_crtc(adp);
+	if (ret) {
+		drm_err(drm, "failed to create crtc");
+		return ret;
+	}
+
+	adp->encoder = drmm_plain_encoder_alloc(drm, NULL, DRM_MODE_ENCODER_DSI, NULL);
+	if (IS_ERR(adp->encoder)) {
+		drm_err(drm, "failed to init encoder");
+		return PTR_ERR(adp->encoder);
+	}
+	adp->encoder->possible_crtcs = ALL_CRTCS;
+
+	ret = drm_bridge_attach(adp->encoder, adp->next_bridge, NULL,
+				DRM_BRIDGE_ATTACH_NO_CONNECTOR);
+	if (ret) {
+		drm_err(drm, "failed to init bridge chain");
+		return ret;
+	}
+
+	adp->connector = drm_bridge_connector_init(drm, adp->encoder);
+	if (IS_ERR(adp->connector))
+		return PTR_ERR(adp->connector);
+
+	drm_connector_attach_encoder(adp->connector, adp->encoder);
+
+	ret = drm_vblank_init(drm, drm->mode_config.num_crtc);
+	if (ret < 0) {
+		drm_err(drm, "failed to initialize vblank");
+		return ret;
+	}
+
+	drm_mode_config_reset(drm);
+
+	return 0;
+}
+
+static int adp_parse_of(struct platform_device *pdev, struct adp_drv_private *adp)
+{
+	struct device *dev = &pdev->dev;
+
+	adp->be = devm_platform_ioremap_resource_byname(pdev, "be");
+	if (IS_ERR(adp->be)) {
+		dev_err(dev, "failed to map display backend mmio");
+		return PTR_ERR(adp->be);
+	}
+
+	adp->fe = devm_platform_ioremap_resource_byname(pdev, "fe");
+	if (IS_ERR(adp->fe)) {
+		dev_err(dev, "failed to map display pipe mmio");
+		return PTR_ERR(adp->fe);
+	}
+
+	adp->be_irq = platform_get_irq_byname(pdev, "be");
+	if (adp->be_irq < 0)
+		return adp->be_irq;
+
+	adp->fe_irq = platform_get_irq_byname(pdev, "fe");
+	if (adp->fe_irq < 0)
+		return adp->fe_irq;
+
+	return 0;
+}
+
+static irqreturn_t adp_fe_irq(int irq, void *arg)
+{
+	struct adp_drv_private *adp = (struct adp_drv_private *)arg;
+	u32 int_status;
+	u32 int_ctl;
+
+	int_status = readl(adp->fe + ADP_INT_STATUS);
+	if (int_status & ADP_INT_STATUS_VBLANK) {
+		drm_crtc_handle_vblank(&adp->crtc);
+		spin_lock(&adp->crtc.dev->event_lock);
+		if (adp->event) {
+			int_ctl = readl(adp->fe + ADP_CTRL);
+			if ((int_ctl & 0xF00) == 0x600) {
+				drm_crtc_send_vblank_event(&adp->crtc, adp->event);
+				adp->event = NULL;
+				drm_crtc_vblank_put(&adp->crtc);
+			}
+		}
+		spin_unlock(&adp->crtc.dev->event_lock);
+	}
+
+	writel(int_status, adp->fe + ADP_INT_STATUS);
+
+
+	return IRQ_HANDLED;
+}
+
+static int adp_drm_bind(struct device *dev)
+{
+	struct drm_device *drm = dev_get_drvdata(dev);
+	struct adp_drv_private *adp = to_adp(drm);
+	int err;
+
+	writel(ADP_CTRL_FIFO_ON, adp->fe + ADP_CTRL);
+
+	adp->next_bridge = drmm_of_get_bridge(&adp->drm, dev->of_node, 0, 0);
+	if (IS_ERR(adp->next_bridge)) {
+		dev_err(dev, "failed to find next bridge");
+		return PTR_ERR(adp->next_bridge);
+	}
+
+	err = adp_setup_mode_config(adp);
+	if (err < 0)
+		return err;
+
+	err = request_irq(adp->fe_irq, adp_fe_irq, 0, "adp-fe", adp);
+	if (err)
+		return err;
+
+	err = drm_dev_register(&adp->drm, 0);
+	if (err)
+		return err;
+
+	return 0;
+}
+
+static void adp_drm_unbind(struct device *dev)
+{
+	struct drm_device *drm = dev_get_drvdata(dev);
+	struct adp_drv_private *adp = to_adp(drm);
+
+	drm_dev_unregister(drm);
+	drm_atomic_helper_shutdown(drm);
+	free_irq(adp->fe_irq, adp);
+}
+
+static const struct component_master_ops adp_master_ops = {
+	.bind	= adp_drm_bind,
+	.unbind = adp_drm_unbind,
+};
+
+static int compare_dev(struct device *dev, void *data)
+{
+	return dev->of_node == data;
+}
+
+static int adp_probe(struct platform_device *pdev)
+{
+	struct device_node *port;
+	struct component_match *match = NULL;
+	struct adp_drv_private *adp;
+	int err;
+
+	adp = devm_drm_dev_alloc(&pdev->dev, &adp_driver, struct adp_drv_private, drm);
+	if (IS_ERR(adp))
+		return PTR_ERR(adp);
+
+	dev_set_drvdata(&pdev->dev, &adp->drm);
+
+	err = adp_parse_of(pdev, adp);
+	if (err < 0)
+		return err;
+
+	port = of_graph_get_remote_node(pdev->dev.of_node, 0, 0);
+	if (!port)
+		return -ENODEV;
+
+	drm_of_component_match_add(&pdev->dev, &match, compare_dev, port);
+	of_node_put(port);
+
+	return component_master_add_with_match(&pdev->dev, &adp_master_ops, match);
+}
+
+static void adp_remove(struct platform_device *pdev)
+{
+	component_master_del(&pdev->dev, &adp_master_ops);
+	dev_set_drvdata(&pdev->dev, NULL);
+}
+
+static const struct of_device_id adp_of_match[] = {
+	{ .compatible = "apple,h7-display-pipe", },
+	{ },
+};
+MODULE_DEVICE_TABLE(of, adp_of_match);
+
+static struct platform_driver adp_platform_driver = {
+	.driver = {
+		.name = "adp",
+		.of_match_table = adp_of_match,
+	},
+	.probe = adp_probe,
+	.remove = adp_remove,
+};
+
+module_platform_driver(adp_platform_driver);
+
+MODULE_DESCRIPTION("Apple Display Pipe DRM driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/gpu/drm/amd/acp/include/acp_gfx_if.h b/drivers/gpu/drm/amd/acp/include/acp_gfx_if.h
index feab8eb7f2a8..b26710cae801 100644
--- a/drivers/gpu/drm/amd/acp/include/acp_gfx_if.h
+++ b/drivers/gpu/drm/amd/acp/include/acp_gfx_if.h
@@ -19,7 +19,7 @@
  * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
  * OTHER DEALINGS IN THE SOFTWARE.
  *
-*/
+ */
 
 #ifndef _ACP_GFX_IF_H
 #define _ACP_GFX_IF_H
diff --git a/drivers/gpu/drm/amd/amdgpu/Kconfig b/drivers/gpu/drm/amd/amdgpu/Kconfig
index 12adca8c7819..7f515be5185d 100644
--- a/drivers/gpu/drm/amd/amdgpu/Kconfig
+++ b/drivers/gpu/drm/amd/amdgpu/Kconfig
@@ -2,10 +2,13 @@
 
 config DRM_AMDGPU
 	tristate "AMD GPU"
-	depends on DRM && PCI && MMU
+	depends on DRM && PCI
 	depends on !UML
 	select FW_LOADER
+	select DRM_CLIENT
+	select DRM_CLIENT_SELECTION
 	select DRM_DISPLAY_DP_HELPER
+	select DRM_DISPLAY_DSC_HELPER
 	select DRM_DISPLAY_HDMI_HELPER
 	select DRM_DISPLAY_HDCP_HELPER
 	select DRM_DISPLAY_HELPER
@@ -17,10 +20,13 @@ config DRM_AMDGPU
 	select HWMON
 	select I2C
 	select I2C_ALGOBIT
+	select CRC16
 	select BACKLIGHT_CLASS_DEVICE
 	select INTERVAL_TREE
 	select DRM_BUDDY
 	select DRM_SUBALLOC_HELPER
+	select DRM_EXEC
+	select DRM_PANEL_BACKLIGHT_QUIRKS
 	# amdgpu depends on ACPI_VIDEO when ACPI is enabled, for select to work
 	# ACPI_VIDEO's dependencies must also be selected.
 	select INPUT if ACPI
@@ -37,14 +43,16 @@ config DRM_AMDGPU_SI
 	bool "Enable amdgpu support for SI parts"
 	depends on DRM_AMDGPU
 	help
-	  Choose this option if you want to enable experimental support
+	  Choose this option if you want to enable support
 	  for SI (Southern Islands) asics.
 
-	  SI is already supported in radeon. Experimental support for SI
-	  in amdgpu will be disabled by default and is still provided by
-	  radeon. Use module options to override this:
+	  SI (Southern Islands) are first generation GCN GPUs,
+	  supported by both drivers: radeon (old) and amdgpu (new).
+	  By default, SI dedicated GPUs are supported by amdgpu.
 
-	  radeon.si_support=0 amdgpu.si_support=1
+	  Use module options to override this:
+	  To use radeon for SI,
+	  radeon.si_support=1 amdgpu.si_support=0
 
 config DRM_AMDGPU_CIK
 	bool "Enable amdgpu support for CIK parts"
@@ -53,22 +61,48 @@ config DRM_AMDGPU_CIK
 	  Choose this option if you want to enable support for CIK (Sea
 	  Islands) asics.
 
-	  CIK is already supported in radeon. Support for CIK in amdgpu
-	  will be disabled by default and is still provided by radeon.
-	  Use module options to override this:
+	  CIK (Sea Islands) are second generation GCN GPUs,
+	  supported by both drivers: radeon (old) and amdgpu (new).
+	  By default,
+	  CIK dedicated GPUs are supported by amdgpu
+	  CIK APUs are supported by radeon
 
+	  Use module options to override this:
+	  To use amdgpu for CIK,
 	  radeon.cik_support=0 amdgpu.cik_support=1
+	  To use radeon for CIK,
+	  radeon.cik_support=1 amdgpu.cik_support=0
 
 config DRM_AMDGPU_USERPTR
 	bool "Always enable userptr write support"
 	depends on DRM_AMDGPU
-	depends on MMU
 	select HMM_MIRROR
 	select MMU_NOTIFIER
 	help
 	  This option selects CONFIG_HMM and CONFIG_HMM_MIRROR if it
 	  isn't already selected to enabled full userptr support.
 
+config DRM_AMD_ISP
+	bool "Enable AMD Image Signal Processor IP support"
+	depends on DRM_AMDGPU && ACPI
+	select MFD_CORE
+	select PM_GENERIC_DOMAINS if PM
+	help
+	Choose this option to enable ISP IP support for AMD SOCs.
+	This adds the ISP (Image Signal Processor) IP driver and wires
+	it up into the amdgpu driver.  It is required for cameras
+	on APUs which utilize mipi cameras.
+
+config DRM_AMDGPU_WERROR
+	bool "Force the compiler to throw an error instead of a warning when compiling"
+	depends on DRM_AMDGPU
+	depends on EXPERT
+	depends on !COMPILE_TEST
+	default n
+	help
+	  Add -Werror to the build flags for amdgpu.ko.
+	  Only enable this if you are warning code for amdgpu.ko.
+
 source "drivers/gpu/drm/amd/acp/Kconfig"
 source "drivers/gpu/drm/amd/display/Kconfig"
 source "drivers/gpu/drm/amd/amdkfd/Kconfig"
diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile
index 415a7fa395c4..c88760fb52ea 100644
--- a/drivers/gpu/drm/amd/amdgpu/Makefile
+++ b/drivers/gpu/drm/amd/amdgpu/Makefile
@@ -1,5 +1,5 @@
 #
-# Copyright 2017 Advanced Micro Devices, Inc.
+# Copyright 2017-2024 Advanced Micro Devices, Inc. All rights reserved.
 #
 # Permission is hereby granted, free of charge, to any person obtaining a
 # copy of this software and associated documentation files (the "Software"),
@@ -23,7 +23,7 @@
 # Makefile for the drm device driver.  This driver provides support for the
 # Direct Rendering Infrastructure (DRI) in XFree86 4.1.0 and higher.
 
-FULL_AMD_PATH=$(srctree)/$(src)/..
+FULL_AMD_PATH=$(src)/..
 DISPLAY_FOLDER_NAME=display
 FULL_AMD_DISPLAY_PATH = $(FULL_AMD_PATH)/$(DISPLAY_FOLDER_NAME)
 
@@ -37,12 +37,17 @@ ccflags-y := -I$(FULL_AMD_PATH)/include/asic_reg \
 	-I$(FULL_AMD_DISPLAY_PATH)/modules/inc \
 	-I$(FULL_AMD_DISPLAY_PATH)/dc \
 	-I$(FULL_AMD_DISPLAY_PATH)/amdgpu_dm \
-	-I$(FULL_AMD_PATH)/amdkfd
+	-I$(FULL_AMD_PATH)/amdkfd \
+	-I$(FULL_AMD_PATH)/ras/ras_mgr
+
+# Locally disable W=1 warnings enabled in drm subsystem Makefile
+subdir-ccflags-y += -Wno-override-init
+subdir-ccflags-$(CONFIG_DRM_AMDGPU_WERROR) += -Werror
 
 amdgpu-y := amdgpu_drv.o
 
 # add KMS driver
-amdgpu-y += amdgpu_device.o amdgpu_kms.o \
+amdgpu-y += amdgpu_device.o amdgpu_doorbell_mgr.o amdgpu_kms.o \
 	amdgpu_atombios.o atombios_crtc.o amdgpu_connectors.o \
 	atom.o amdgpu_fence.o amdgpu_ttm.o amdgpu_object.o amdgpu_gart.o \
 	amdgpu_encoders.o amdgpu_display.o amdgpu_i2c.o \
@@ -50,7 +55,8 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \
 	amdgpu_cs.o amdgpu_bios.o amdgpu_benchmark.o \
 	atombios_dp.o amdgpu_afmt.o amdgpu_trace_points.o \
 	atombios_encoders.o amdgpu_sa.o atombios_i2c.o \
-	amdgpu_dma_buf.o amdgpu_vm.o amdgpu_vm_pt.o amdgpu_ib.o amdgpu_pll.o \
+	amdgpu_dma_buf.o amdgpu_vm.o amdgpu_vm_pt.o amdgpu_vm_tlb_fence.o \
+	amdgpu_ib.o amdgpu_pll.o \
 	amdgpu_ucode.o amdgpu_bo_list.o amdgpu_ctx.o amdgpu_sync.o \
 	amdgpu_gtt_mgr.o amdgpu_preempt_mgr.o amdgpu_vram_mgr.o amdgpu_virt.o \
 	amdgpu_atomfirmware.o amdgpu_vf_error.o amdgpu_sched.o \
@@ -60,7 +66,8 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \
 	amdgpu_umc.o smu_v11_0_i2c.o amdgpu_fru_eeprom.o amdgpu_rap.o \
 	amdgpu_fw_attestation.o amdgpu_securedisplay.o \
 	amdgpu_eeprom.o amdgpu_mca.o amdgpu_psp_ta.o amdgpu_lsdma.o \
-	amdgpu_ring_mux.o
+	amdgpu_ring_mux.o amdgpu_xcp.o amdgpu_seq64.o amdgpu_aca.o amdgpu_dev_coredump.o \
+	amdgpu_cper.o amdgpu_userq_fence.o amdgpu_eviction_fence.o amdgpu_ip.o
 
 amdgpu-$(CONFIG_PROC_FS) += amdgpu_fdinfo.o
 
@@ -71,20 +78,23 @@ amdgpu-$(CONFIG_DRM_AMDGPU_CIK)+= cik.o cik_ih.o \
 	dce_v8_0.o gfx_v7_0.o cik_sdma.o uvd_v4_2.o vce_v2_0.o
 
 amdgpu-$(CONFIG_DRM_AMDGPU_SI)+= si.o gmc_v6_0.o gfx_v6_0.o si_ih.o si_dma.o dce_v6_0.o \
-	uvd_v3_1.o
+	uvd_v3_1.o vce_v1_0.o
 
 amdgpu-y += \
 	vi.o mxgpu_vi.o nbio_v6_1.o soc15.o emu_soc.o mxgpu_ai.o nbio_v7_0.o vega10_reg_init.o \
 	vega20_reg_init.o nbio_v7_4.o nbio_v2_3.o nv.o arct_reg_init.o mxgpu_nv.o \
-	nbio_v7_2.o hdp_v4_0.o hdp_v5_0.o aldebaran_reg_init.o aldebaran.o soc21.o \
+	nbio_v7_2.o hdp_v4_0.o hdp_v5_0.o aldebaran_reg_init.o aldebaran.o soc21.o soc24.o \
 	sienna_cichlid.o smu_v13_0_10.o nbio_v4_3.o hdp_v6_0.o nbio_v7_7.o hdp_v5_2.o lsdma_v6_0.o \
-	nbio_v7_9.o
+	nbio_v7_9.o aqua_vanjaram.o nbio_v7_11.o lsdma_v7_0.o hdp_v7_0.o nbif_v6_3_1.o \
+	cyan_skillfish_reg_init.o
 
 # add DF block
 amdgpu-y += \
 	df_v1_7.o \
 	df_v3_6.o \
-	df_v4_3.o
+	df_v4_3.o \
+	df_v4_6_2.o \
+	df_v4_15.o
 
 # add GMC block
 amdgpu-y += \
@@ -93,11 +103,12 @@ amdgpu-y += \
 	gfxhub_v1_0.o mmhub_v1_0.o gmc_v9_0.o gfxhub_v1_1.o mmhub_v9_4.o \
 	gfxhub_v2_0.o mmhub_v2_0.o gmc_v10_0.o gfxhub_v2_1.o mmhub_v2_3.o \
 	mmhub_v1_7.o gfxhub_v3_0.o mmhub_v3_0.o mmhub_v3_0_2.o gmc_v11_0.o \
-	mmhub_v3_0_1.o gfxhub_v3_0_3.o gfxhub_v1_2.o mmhub_v1_8.o
+	mmhub_v3_0_1.o gfxhub_v3_0_3.o gfxhub_v1_2.o mmhub_v1_8.o mmhub_v3_3.o \
+	gfxhub_v11_5_0.o mmhub_v4_1_0.o gfxhub_v12_0.o gmc_v12_0.o
 
 # add UMC block
 amdgpu-y += \
-	umc_v6_0.o umc_v6_1.o umc_v6_7.o umc_v8_7.o umc_v8_10.o
+	umc_v6_0.o umc_v6_1.o umc_v6_7.o umc_v8_7.o umc_v8_10.o umc_v12_0.o umc_v8_14.o
 
 # add IH block
 amdgpu-y += \
@@ -109,7 +120,9 @@ amdgpu-y += \
 	vega10_ih.o \
 	vega20_ih.o \
 	navi10_ih.o \
-	ih_v6_0.o
+	ih_v6_0.o \
+	ih_v6_1.o \
+	ih_v7_0.o
 
 # add PSP block
 amdgpu-y += \
@@ -120,12 +133,12 @@ amdgpu-y += \
 	psp_v11_0_8.o \
 	psp_v12_0.o \
 	psp_v13_0.o \
-	psp_v13_0_4.o
+	psp_v13_0_4.o \
+	psp_v14_0.o
 
 # add DCE block
 amdgpu-y += \
 	dce_v10_0.o \
-	dce_v11_0.o \
 	amdgpu_vkms.o
 
 # add GFX block
@@ -141,7 +154,9 @@ amdgpu-y += \
 	imu_v11_0.o \
 	gfx_v11_0.o \
 	gfx_v11_0_3.o \
-	imu_v11_0_3.o
+	imu_v11_0_3.o \
+	gfx_v12_0.o \
+	imu_v12_0.o
 
 # add async DMA block
 amdgpu-y += \
@@ -153,13 +168,17 @@ amdgpu-y += \
 	sdma_v4_4_2.o \
 	sdma_v5_0.o \
 	sdma_v5_2.o \
-	sdma_v6_0.o
+	sdma_v6_0.o \
+	sdma_v7_0.o
 
 # add MES block
 amdgpu-y += \
 	amdgpu_mes.o \
-	mes_v10_1.o \
-	mes_v11_0.o
+	mes_v11_0.o \
+	mes_v12_0.o \
+
+# add GFX userqueue support
+amdgpu-y += mes_userqueue.o
 
 # add UVD block
 amdgpu-y += \
@@ -183,19 +202,39 @@ amdgpu-y += \
 	vcn_v2_5.o \
 	vcn_v3_0.o \
 	vcn_v4_0.o \
+	vcn_v4_0_3.o \
+	vcn_v4_0_5.o \
+	vcn_v5_0_0.o \
+	vcn_v5_0_1.o \
 	amdgpu_jpeg.o \
 	jpeg_v1_0.o \
 	jpeg_v2_0.o \
 	jpeg_v2_5.o \
 	jpeg_v3_0.o \
-	jpeg_v4_0.o
+	jpeg_v4_0.o \
+	jpeg_v4_0_3.o \
+	jpeg_v4_0_5.o \
+	jpeg_v5_0_0.o \
+	jpeg_v5_0_1.o
+
+# add VPE block
+amdgpu-y += \
+	amdgpu_vpe.o \
+	vpe_v6_1.o
+
+# add UMSCH block
+amdgpu-y += \
+	amdgpu_umsch_mm.o \
+	umsch_mm_v4_0.o
 
+#
 # add ATHUB block
 amdgpu-y += \
 	athub_v1_0.o \
 	athub_v2_0.o \
 	athub_v2_1.o \
-	athub_v3_0.o
+	athub_v3_0.o \
+	athub_v4_1_0.o
 
 # add SMUIO block
 amdgpu-y += \
@@ -203,7 +242,9 @@ amdgpu-y += \
 	smuio_v11_0.o \
 	smuio_v11_0_6.o \
 	smuio_v13_0.o \
-	smuio_v13_0_6.o
+	smuio_v13_0_3.o \
+	smuio_v13_0_6.o \
+	smuio_v14_0_2.o
 
 # add reset block
 amdgpu-y += \
@@ -216,6 +257,8 @@ amdgpu-y += \
 # add amdkfd interfaces
 amdgpu-y += amdgpu_amdkfd.o
 
+# add gfx usermode queue
+amdgpu-y += amdgpu_userq.o
 
 ifneq ($(CONFIG_HSA_AMD),)
 AMDKFD_PATH := ../amdkfd
@@ -228,9 +271,11 @@ amdgpu-y += \
 	amdgpu_amdkfd_gfx_v9.o \
 	amdgpu_amdkfd_arcturus.o \
 	amdgpu_amdkfd_aldebaran.o \
+	amdgpu_amdkfd_gc_9_4_3.o \
 	amdgpu_amdkfd_gfx_v10.o \
 	amdgpu_amdkfd_gfx_v10_3.o \
-	amdgpu_amdkfd_gfx_v11.o
+	amdgpu_amdkfd_gfx_v11.o \
+	amdgpu_amdkfd_gfx_v12.o
 
 ifneq ($(CONFIG_DRM_AMDGPU_CIK),)
 amdgpu-y += amdgpu_amdkfd_gfx_v7.o
@@ -272,4 +317,17 @@ amdgpu-y += $(AMD_DISPLAY_FILES)
 
 endif
 
+# add isp block
+ifneq ($(CONFIG_DRM_AMD_ISP),)
+amdgpu-y += \
+	amdgpu_isp.o \
+	isp_v4_1_0.o \
+	isp_v4_1_1.o
+endif
+
+AMD_GPU_RAS_PATH := ../ras
+AMD_GPU_RAS_FULL_PATH := $(FULL_AMD_PATH)/ras
+include $(AMD_GPU_RAS_FULL_PATH)/Makefile
+amdgpu-y += $(AMD_GPU_RAS_FILES)
+
 obj-$(CONFIG_DRM_AMDGPU)+= amdgpu.o
diff --git a/drivers/gpu/drm/amd/amdgpu/aldebaran.c b/drivers/gpu/drm/amd/amdgpu/aldebaran.c
index 2b97b8a96fb4..daa7b23bc775 100644
--- a/drivers/gpu/drm/amd/amdgpu/aldebaran.c
+++ b/drivers/gpu/drm/amd/amdgpu/aldebaran.c
@@ -35,7 +35,7 @@ static bool aldebaran_is_mode2_default(struct amdgpu_reset_control *reset_ctl)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle;
 
-	if ((adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 2) &&
+	if ((amdgpu_ip_version(adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 2) &&
 	     adev->gmc.xgmi.connected_to_cpu))
 		return true;
 
@@ -48,59 +48,64 @@ aldebaran_get_reset_handler(struct amdgpu_reset_control *reset_ctl,
 {
 	struct amdgpu_reset_handler *handler;
 	struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle;
+	int i;
+
+	if (reset_context->method == AMD_RESET_METHOD_NONE) {
+		if (aldebaran_is_mode2_default(reset_ctl))
+			reset_context->method = AMD_RESET_METHOD_MODE2;
+		else
+			reset_context->method = amdgpu_asic_reset_method(adev);
+	}
 
 	if (reset_context->method != AMD_RESET_METHOD_NONE) {
 		dev_dbg(adev->dev, "Getting reset handler for method %d\n",
 			reset_context->method);
-		list_for_each_entry(handler, &reset_ctl->reset_handlers,
-				     handler_list) {
+		for_each_handler(i, handler, reset_ctl) {
 			if (handler->reset_method == reset_context->method)
 				return handler;
 		}
 	}
 
-	if (aldebaran_is_mode2_default(reset_ctl)) {
-		list_for_each_entry(handler, &reset_ctl->reset_handlers,
-				     handler_list) {
-			if (handler->reset_method == AMD_RESET_METHOD_MODE2) {
-				reset_context->method = AMD_RESET_METHOD_MODE2;
-				return handler;
-			}
-		}
-	}
-
 	dev_dbg(adev->dev, "Reset handler not found!\n");
 
 	return NULL;
 }
 
+static inline uint32_t aldebaran_get_ip_block_mask(struct amdgpu_device *adev)
+{
+	uint32_t ip_block_mask = BIT(AMD_IP_BLOCK_TYPE_GFX) |
+				 BIT(AMD_IP_BLOCK_TYPE_SDMA);
+
+	if (adev->aid_mask)
+		ip_block_mask |= BIT(AMD_IP_BLOCK_TYPE_IH);
+
+	return ip_block_mask;
+}
+
 static int aldebaran_mode2_suspend_ip(struct amdgpu_device *adev)
 {
+	uint32_t ip_block_mask = aldebaran_get_ip_block_mask(adev);
+	uint32_t ip_block;
 	int r, i;
 
+	/* Skip suspend of SDMA IP versions >= 4.4.2. They are multi-aid */
+	if (adev->aid_mask)
+		ip_block_mask &= ~BIT(AMD_IP_BLOCK_TYPE_SDMA);
+
 	amdgpu_device_set_pg_state(adev, AMD_PG_STATE_UNGATE);
 	amdgpu_device_set_cg_state(adev, AMD_CG_STATE_UNGATE);
 
 	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
-		if (!(adev->ip_blocks[i].version->type ==
-			      AMD_IP_BLOCK_TYPE_GFX ||
-		      adev->ip_blocks[i].version->type ==
-			      AMD_IP_BLOCK_TYPE_SDMA))
+		ip_block = BIT(adev->ip_blocks[i].version->type);
+		if (!(ip_block_mask & ip_block))
 			continue;
 
-		r = adev->ip_blocks[i].version->funcs->suspend(adev);
-
-		if (r) {
-			dev_err(adev->dev,
-				"suspend of IP block <%s> failed %d\n",
-				adev->ip_blocks[i].version->funcs->name, r);
+		r = amdgpu_ip_block_suspend(&adev->ip_blocks[i]);
+		if (r)
 			return r;
-		}
-
-		adev->ip_blocks[i].status.hw = false;
 	}
 
-	return r;
+	return 0;
 }
 
 static int
@@ -124,9 +129,9 @@ static void aldebaran_async_reset(struct work_struct *work)
 	struct amdgpu_reset_control *reset_ctl =
 		container_of(work, struct amdgpu_reset_control, reset_work);
 	struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle;
+	int i;
 
-	list_for_each_entry(handler, &reset_ctl->reset_handlers,
-			     handler_list) {
+	for_each_handler(i, handler, reset_ctl)	{
 		if (handler->reset_method == reset_ctl->active_reset) {
 			dev_dbg(adev->dev, "Resetting device\n");
 			handler->do_reset(adev);
@@ -157,7 +162,7 @@ aldebaran_mode2_perform_reset(struct amdgpu_reset_control *reset_ctl,
 	if (reset_device_list == NULL)
 		return -EINVAL;
 
-	if (adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 2) &&
+	if (amdgpu_ip_version(adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 2) &&
 	    reset_context->hive == NULL) {
 		/* Wrong context, return error */
 		return -EINVAL;
@@ -210,8 +215,10 @@ aldebaran_mode2_perform_reset(struct amdgpu_reset_control *reset_ctl,
 static int aldebaran_mode2_restore_ip(struct amdgpu_device *adev)
 {
 	struct amdgpu_firmware_info *ucode_list[AMDGPU_UCODE_ID_MAXIMUM];
+	uint32_t ip_block_mask = aldebaran_get_ip_block_mask(adev);
 	struct amdgpu_firmware_info *ucode;
 	struct amdgpu_ip_block *cmn_block;
+	struct amdgpu_ip_block *ih_block;
 	int ucode_count = 0;
 	int i, r;
 
@@ -249,10 +256,22 @@ static int aldebaran_mode2_restore_ip(struct amdgpu_device *adev)
 		dev_err(adev->dev, "Failed to get BIF handle\n");
 		return -EINVAL;
 	}
-	r = cmn_block->version->funcs->resume(adev);
+	r = amdgpu_ip_block_resume(cmn_block);
 	if (r)
 		return r;
 
+	if (ip_block_mask & BIT(AMD_IP_BLOCK_TYPE_IH)) {
+		ih_block = amdgpu_device_ip_get_ip_block(adev,
+							 AMD_IP_BLOCK_TYPE_IH);
+		if (unlikely(!ih_block)) {
+			dev_err(adev->dev, "Failed to get IH handle\n");
+			return -EINVAL;
+		}
+		r = amdgpu_ip_block_resume(ih_block);
+		if (r)
+			return r;
+	}
+
 	/* Reinit GFXHUB */
 	adev->gfxhub.funcs->init(adev);
 	r = adev->gfxhub.funcs->gart_enable(adev);
@@ -285,15 +304,10 @@ static int aldebaran_mode2_restore_ip(struct amdgpu_device *adev)
 		      adev->ip_blocks[i].version->type ==
 			      AMD_IP_BLOCK_TYPE_SDMA))
 			continue;
-		r = adev->ip_blocks[i].version->funcs->resume(adev);
-		if (r) {
-			dev_err(adev->dev,
-				"resume of IP block <%s> failed %d\n",
-				adev->ip_blocks[i].version->funcs->name, r);
-			return r;
-		}
 
-		adev->ip_blocks[i].status.hw = true;
+		r = amdgpu_ip_block_resume(&adev->ip_blocks[i]);
+		if (r)
+			return r;
 	}
 
 	for (i = 0; i < adev->num_ip_blocks; i++) {
@@ -307,7 +321,7 @@ static int aldebaran_mode2_restore_ip(struct amdgpu_device *adev)
 
 		if (adev->ip_blocks[i].version->funcs->late_init) {
 			r = adev->ip_blocks[i].version->funcs->late_init(
-				(void *)adev);
+				&adev->ip_blocks[i]);
 			if (r) {
 				dev_err(adev->dev,
 					"late_init of IP block <%s> failed %d after reset\n",
@@ -319,8 +333,6 @@ static int aldebaran_mode2_restore_ip(struct amdgpu_device *adev)
 		adev->ip_blocks[i].status.late_initialized = true;
 	}
 
-	amdgpu_ras_set_error_query_ready(adev, true);
-
 	amdgpu_device_set_cg_state(adev, AMD_CG_STATE_GATE);
 	amdgpu_device_set_pg_state(adev, AMD_PG_STATE_GATE);
 
@@ -333,12 +345,13 @@ aldebaran_mode2_restore_hwcontext(struct amdgpu_reset_control *reset_ctl,
 {
 	struct list_head *reset_device_list = reset_context->reset_device_list;
 	struct amdgpu_device *tmp_adev = NULL;
+	struct amdgpu_ras *con;
 	int r;
 
 	if (reset_device_list == NULL)
 		return -EINVAL;
 
-	if (reset_context->reset_req_dev->ip_versions[MP1_HWIP][0] ==
+	if (amdgpu_ip_version(reset_context->reset_req_dev, MP1_HWIP, 0) ==
 		    IP_VERSION(13, 0, 2) &&
 	    reset_context->hive == NULL) {
 		/* Wrong context, return error */
@@ -346,8 +359,12 @@ aldebaran_mode2_restore_hwcontext(struct amdgpu_reset_control *reset_ctl,
 	}
 
 	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
+		amdgpu_set_init_level(tmp_adev,
+				AMDGPU_INIT_LEVEL_RESET_RECOVERY);
 		dev_info(tmp_adev->dev,
 			 "GPU reset succeeded, trying to resume\n");
+		/*TBD: Ideally should clear only GFX, SDMA blocks*/
+		amdgpu_ras_clear_err_state(tmp_adev);
 		r = aldebaran_mode2_restore_ip(tmp_adev);
 		if (r)
 			goto end;
@@ -358,7 +375,30 @@ aldebaran_mode2_restore_hwcontext(struct amdgpu_reset_control *reset_ctl,
 		 */
 		amdgpu_register_gpu_instance(tmp_adev);
 
-		/* Resume RAS */
+		/* Resume RAS, ecc_irq */
+		con = amdgpu_ras_get_context(tmp_adev);
+		if (!amdgpu_sriov_vf(tmp_adev) && con) {
+			if (tmp_adev->sdma.ras &&
+				tmp_adev->sdma.ras->ras_block.ras_late_init) {
+				r = tmp_adev->sdma.ras->ras_block.ras_late_init(tmp_adev,
+						&tmp_adev->sdma.ras->ras_block.ras_comm);
+				if (r) {
+					dev_err(tmp_adev->dev, "SDMA failed to execute ras_late_init! ret:%d\n", r);
+					goto end;
+				}
+			}
+
+			if (tmp_adev->gfx.ras &&
+				tmp_adev->gfx.ras->ras_block.ras_late_init) {
+				r = tmp_adev->gfx.ras->ras_block.ras_late_init(tmp_adev,
+						&tmp_adev->gfx.ras->ras_block.ras_comm);
+				if (r) {
+					dev_err(tmp_adev->dev, "GFX failed to execute ras_late_init! ret:%d\n", r);
+					goto end;
+				}
+			}
+		}
+
 		amdgpu_ras_resume(tmp_adev);
 
 		/* Update PSP FW topology after reset */
@@ -368,6 +408,8 @@ aldebaran_mode2_restore_hwcontext(struct amdgpu_reset_control *reset_ctl,
 							tmp_adev);
 
 		if (!r) {
+			amdgpu_set_init_level(tmp_adev,
+					      AMDGPU_INIT_LEVEL_DEFAULT);
 			amdgpu_irq_gpu_reset_resume_helper(tmp_adev);
 
 			r = amdgpu_ib_ring_tests(tmp_adev);
@@ -395,6 +437,12 @@ static struct amdgpu_reset_handler aldebaran_mode2_handler = {
 	.do_reset		= aldebaran_mode2_reset,
 };
 
+static struct amdgpu_reset_handler
+	*aldebaran_rst_handlers[AMDGPU_RESET_MAX_HANDLERS] = {
+		&aldebaran_mode2_handler,
+		&xgmi_reset_on_init_handler,
+	};
+
 int aldebaran_reset_init(struct amdgpu_device *adev)
 {
 	struct amdgpu_reset_control *reset_ctl;
@@ -408,10 +456,9 @@ int aldebaran_reset_init(struct amdgpu_device *adev)
 	reset_ctl->active_reset = AMD_RESET_METHOD_NONE;
 	reset_ctl->get_reset_handler = aldebaran_get_reset_handler;
 
-	INIT_LIST_HEAD(&reset_ctl->reset_handlers);
 	INIT_WORK(&reset_ctl->reset_work, reset_ctl->async_reset);
 	/* Only mode2 is handled through reset control now */
-	amdgpu_reset_add_handler(reset_ctl, &aldebaran_mode2_handler);
+	reset_ctl->reset_handlers = &aldebaran_rst_handlers;
 
 	adev->reset_cntl = reset_ctl;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 02b827785e39..9f9774f58ce1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -53,7 +53,6 @@
 
 #include <drm/ttm/ttm_bo.h>
 #include <drm/ttm/ttm_placement.h>
-#include <drm/ttm/ttm_execbuf_util.h>
 
 #include <drm/amdgpu_drm.h>
 #include <drm/drm_gem.h>
@@ -64,6 +63,7 @@
 #include "kgd_pp_interface.h"
 
 #include "amd_shared.h"
+#include "amdgpu_utils.h"
 #include "amdgpu_mode.h"
 #include "amdgpu_ih.h"
 #include "amdgpu_irq.h"
@@ -80,6 +80,8 @@
 #include "amdgpu_vce.h"
 #include "amdgpu_vcn.h"
 #include "amdgpu_jpeg.h"
+#include "amdgpu_vpe.h"
+#include "amdgpu_umsch_mm.h"
 #include "amdgpu_gmc.h"
 #include "amdgpu_gfx.h"
 #include "amdgpu_sdma.h"
@@ -106,27 +108,33 @@
 #include "amdgpu_smuio.h"
 #include "amdgpu_fdinfo.h"
 #include "amdgpu_mca.h"
+#include "amdgpu_aca.h"
 #include "amdgpu_ras.h"
+#include "amdgpu_cper.h"
+#include "amdgpu_xcp.h"
+#include "amdgpu_seq64.h"
+#include "amdgpu_reg_state.h"
+#include "amdgpu_userq.h"
+#include "amdgpu_eviction_fence.h"
+#if defined(CONFIG_DRM_AMD_ISP)
+#include "amdgpu_isp.h"
+#endif
 
-#define MAX_GPU_INSTANCE		16
+#define MAX_GPU_INSTANCE		64
 
-struct amdgpu_gpu_instance
-{
+#define GFX_SLICE_PERIOD_MS		250
+
+struct amdgpu_gpu_instance {
 	struct amdgpu_device		*adev;
 	int				mgpu_fan_enabled;
 };
 
-struct amdgpu_mgpu_info
-{
+struct amdgpu_mgpu_info {
 	struct amdgpu_gpu_instance	gpu_ins[MAX_GPU_INSTANCE];
 	struct mutex			mutex;
 	uint32_t			num_gpu;
 	uint32_t			num_dgpu;
 	uint32_t			num_apu;
-
-	/* delayed reset_func for XGMI configuration if necessary */
-	struct delayed_work		delayed_reset_work;
-	bool				pending_reset;
 };
 
 enum amdgpu_ss {
@@ -136,8 +144,15 @@ enum amdgpu_ss {
 	AMDGPU_SS_DRV_UNLOAD
 };
 
-struct amdgpu_watchdog_timer
-{
+struct amdgpu_hwip_reg_entry {
+	u32		hwip;
+	u32		inst;
+	u32		seg;
+	u32		reg_offset;
+	const char	*reg_name;
+};
+
+struct amdgpu_watchdog_timer {
 	bool timeout_fatal_disable;
 	uint32_t period; /* maxCycles = (1 << period), the number of cycles before a timeout */
 };
@@ -195,8 +210,9 @@ extern uint amdgpu_dc_feature_mask;
 extern uint amdgpu_freesync_vid_mode;
 extern uint amdgpu_dc_debug_mask;
 extern uint amdgpu_dc_visual_confirm;
-extern uint amdgpu_dm_abm_level;
+extern int amdgpu_dm_abm_level;
 extern int amdgpu_backlight;
+extern int amdgpu_damage_clips;
 extern struct amdgpu_mgpu_info mgpu_info;
 extern int amdgpu_ras_enable;
 extern uint amdgpu_ras_mask;
@@ -207,16 +223,21 @@ extern int amdgpu_async_gfx_ring;
 extern int amdgpu_mcbp;
 extern int amdgpu_discovery;
 extern int amdgpu_mes;
+extern int amdgpu_mes_log_enable;
 extern int amdgpu_mes_kiq;
+extern int amdgpu_uni_mes;
 extern int amdgpu_noretry;
 extern int amdgpu_force_asic_type;
 extern int amdgpu_smartshift_bias;
 extern int amdgpu_use_xgmi_p2p;
+extern int amdgpu_mtype_local;
+extern int amdgpu_enforce_isolation;
 #ifdef CONFIG_HSA_AMD
 extern int sched_policy;
 extern bool debug_evictions;
 extern bool no_system_mem_limit;
 extern int halt_if_hws_hang;
+extern uint amdgpu_svm_default_granularity;
 #else
 static const int __maybe_unused sched_policy = KFD_SCHED_POLICY_HWS;
 static const bool __maybe_unused debug_evictions; /* = false */
@@ -239,12 +260,22 @@ extern int amdgpu_cik_support;
 extern int amdgpu_num_kcq;
 
 #define AMDGPU_VCNFW_LOG_SIZE (32 * 1024)
+#define AMDGPU_UMSCHFW_LOG_SIZE (32 * 1024)
 extern int amdgpu_vcnfw_log;
 extern int amdgpu_sg_display;
+extern int amdgpu_umsch_mm;
+extern int amdgpu_seamless;
+extern int amdgpu_umsch_mm_fwlog;
+
+extern int amdgpu_user_partt_mode;
+extern int amdgpu_agp;
+extern int amdgpu_rebar;
+
+extern int amdgpu_wbrf;
+extern int amdgpu_user_queue;
 
 #define AMDGPU_VM_MAX_NUM_CTX			4096
 #define AMDGPU_SG_THRESHOLD			(256*1024*1024)
-#define AMDGPU_DEFAULT_GTT_SIZE_MB		3072ULL /* 3GB by default */
 #define AMDGPU_WAIT_IDLE_TIMEOUT_IN_MS	        3000
 #define AMDGPU_MAX_USEC_TIMEOUT			100000	/* 100 ms */
 #define AMDGPU_FENCE_JIFFIES_TIMEOUT		(HZ / 2)
@@ -274,6 +305,12 @@ extern int amdgpu_sg_display;
 #define AMDGPU_RESET_VCE			(1 << 13)
 #define AMDGPU_RESET_VCE1			(1 << 14)
 
+/* reset mask */
+#define AMDGPU_RESET_TYPE_FULL (1 << 0) /* full adapter reset, mode1/mode2/BACO/etc. */
+#define AMDGPU_RESET_TYPE_SOFT_RESET (1 << 1) /* IP level soft reset */
+#define AMDGPU_RESET_TYPE_PER_QUEUE (1 << 2) /* per queue */
+#define AMDGPU_RESET_TYPE_PER_PIPE (1 << 3) /* per pipe */
+
 /* max cursor sizes (in pixels) */
 #define CIK_CURSOR_WIDTH 128
 #define CIK_CURSOR_HEIGHT 128
@@ -282,6 +319,10 @@ extern int amdgpu_sg_display;
 #define AMDGPU_SMARTSHIFT_MAX_BIAS (100)
 #define AMDGPU_SMARTSHIFT_MIN_BIAS (-100)
 
+/* Extra time delay(in ms) to eliminate the influence of temperature momentary fluctuation */
+#define AMDGPU_SWCTF_EXTRA_DELAY		50
+
+struct amdgpu_xcp_mgr;
 struct amdgpu_device;
 struct amdgpu_irq_src;
 struct amdgpu_fpriv;
@@ -317,7 +358,6 @@ enum amdgpu_kiq_irq {
 	AMDGPU_CP_KIQ_IRQ_DRIVER0 = 0,
 	AMDGPU_CP_KIQ_IRQ_LAST
 };
-#define SRIOV_USEC_TIMEOUT  1200000 /* wait 12 * 100ms for SRIOV */
 #define MAX_KIQ_REG_WAIT       5000 /* in usecs, 5ms */
 #define MAX_KIQ_REG_BAILOUT_INTERVAL   5 /* in msecs, 5ms */
 #define MAX_KIQ_REG_TRY 1000
@@ -332,10 +372,15 @@ void amdgpu_device_ip_get_clockgating_state(struct amdgpu_device *adev,
 					    u64 *flags);
 int amdgpu_device_ip_wait_for_idle(struct amdgpu_device *adev,
 				   enum amd_ip_block_type block_type);
-bool amdgpu_device_ip_is_idle(struct amdgpu_device *adev,
+bool amdgpu_device_ip_is_hw(struct amdgpu_device *adev,
+			    enum amd_ip_block_type block_type);
+bool amdgpu_device_ip_is_valid(struct amdgpu_device *adev,
 			      enum amd_ip_block_type block_type);
+int amdgpu_ip_block_suspend(struct amdgpu_ip_block *ip_block);
 
-#define AMDGPU_MAX_IP_NUM 16
+int amdgpu_ip_block_resume(struct amdgpu_ip_block *ip_block);
+
+#define AMDGPU_MAX_IP_NUM AMD_IP_BLOCK_TYPE_NUM
 
 struct amdgpu_ip_block_status {
 	bool valid;
@@ -353,12 +398,10 @@ struct amdgpu_ip_block_version {
 	const struct amd_ip_funcs *funcs;
 };
 
-#define HW_REV(_Major, _Minor, _Rev) \
-	((((uint32_t) (_Major)) << 16) | ((uint32_t) (_Minor) << 8) | ((uint32_t) (_Rev)))
-
 struct amdgpu_ip_block {
 	struct amdgpu_ip_block_status status;
 	const struct amdgpu_ip_block_version *version;
+	struct amdgpu_device *adev;
 };
 
 int amdgpu_device_ip_block_version_cmp(struct amdgpu_device *adev,
@@ -379,6 +422,7 @@ bool amdgpu_get_bios(struct amdgpu_device *adev);
 bool amdgpu_read_bios(struct amdgpu_device *adev);
 bool amdgpu_soc15_read_bios_from_rom(struct amdgpu_device *adev,
 				     u8 *bios, u32 length_bytes);
+void amdgpu_bios_release(struct amdgpu_device *adev);
 /*
  * Clocks
  */
@@ -393,7 +437,6 @@ struct amdgpu_clock {
 	uint32_t default_mclk;
 	uint32_t default_sclk;
 	uint32_t default_dispclk;
-	uint32_t current_dispclk;
 	uint32_t dp_extclk;
 	uint32_t max_pixel_clock;
 };
@@ -429,9 +472,6 @@ struct amdgpu_sa_manager {
 	void				*cpu_ptr;
 };
 
-int amdgpu_fence_slab_init(void);
-void amdgpu_fence_slab_fini(void);
-
 /*
  * IRQS.
  */
@@ -451,7 +491,6 @@ struct amdgpu_flip_work {
 	bool				async;
 };
 
-
 /*
  * file private structure
  */
@@ -460,9 +499,17 @@ struct amdgpu_fpriv {
 	struct amdgpu_vm	vm;
 	struct amdgpu_bo_va	*prt_va;
 	struct amdgpu_bo_va	*csa_va;
+	struct amdgpu_bo_va	*seq64_va;
 	struct mutex		bo_list_lock;
 	struct idr		bo_list_handles;
 	struct amdgpu_ctx_mgr	ctx_mgr;
+	struct amdgpu_userq_mgr	userq_mgr;
+
+	/* Eviction fence infra */
+	struct amdgpu_eviction_fence_mgr evf_mgr;
+
+	/** GPU partition selection */
+	uint32_t		xcp_id;
 };
 
 int amdgpu_file_to_fpriv(struct file *filp, struct amdgpu_fpriv **fpriv);
@@ -472,12 +519,63 @@ int amdgpu_file_to_fpriv(struct file *filp, struct amdgpu_fpriv **fpriv);
  */
 #define AMDGPU_MAX_WB 1024	/* Reserve at most 1024 WB slots for amdgpu-owned rings. */
 
+/**
+ * amdgpu_wb - This struct is used for small GPU memory allocation.
+ *
+ * This struct is used to allocate a small amount of GPU memory that can be
+ * used to shadow certain states into the memory. This is especially useful for
+ * providing easy CPU access to some states without requiring register access
+ * (e.g., if some block is power gated, reading register may be problematic).
+ *
+ * Note: the term writeback was initially used because many of the amdgpu
+ * components had some level of writeback memory, and this struct initially
+ * described those components.
+ */
 struct amdgpu_wb {
+
+	/**
+	 * @wb_obj:
+	 *
+	 * Buffer Object used for the writeback memory.
+	 */
 	struct amdgpu_bo	*wb_obj;
-	volatile uint32_t	*wb;
+
+	/**
+	 * @wb:
+	 *
+	 * Pointer to the first writeback slot. In terms of CPU address
+	 * this value can be accessed directly by using the offset as an index.
+	 * For the GPU address, it is necessary to use gpu_addr and the offset.
+	 */
+	uint32_t		*wb;
+
+	/**
+	 * @gpu_addr:
+	 *
+	 * Writeback base address in the GPU.
+	 */
 	uint64_t		gpu_addr;
-	u32			num_wb;	/* Number of wb slots actually reserved for amdgpu. */
+
+	/**
+	 * @num_wb:
+	 *
+	 * Number of writeback slots reserved for amdgpu.
+	 */
+	u32			num_wb;
+
+	/**
+	 * @used:
+	 *
+	 * Track the writeback slot already used.
+	 */
 	unsigned long		used[DIV_ROUND_UP(AMDGPU_MAX_WB, BITS_PER_LONG)];
+
+	/**
+	 * @lock:
+	 *
+	 * Protects read and write of the used field array.
+	 */
+	spinlock_t		lock;
 };
 
 int amdgpu_device_wb_get(struct amdgpu_device *adev, u32 *wb);
@@ -496,14 +594,42 @@ struct amdgpu_allowed_register_entry {
 	bool grbm_indexed;
 };
 
+/**
+ * enum amd_reset_method - Methods for resetting AMD GPU devices
+ *
+ * @AMD_RESET_METHOD_NONE: The device will not be reset.
+ * @AMD_RESET_LEGACY: Method reserved for SI, CIK and VI ASICs.
+ * @AMD_RESET_MODE0: Reset the entire ASIC. Not currently available for the
+ *                   any device.
+ * @AMD_RESET_MODE1: Resets all IP blocks on the ASIC (SDMA, GFX, VCN, etc.)
+ *                   individually. Suitable only for some discrete GPU, not
+ *                   available for all ASICs.
+ * @AMD_RESET_MODE2: Resets a lesser level of IPs compared to MODE1. Which IPs
+ *                   are reset depends on the ASIC. Notably doesn't reset IPs
+ *                   shared with the CPU on APUs or the memory controllers (so
+ *                   VRAM is not lost). Not available on all ASICs.
+ * @AMD_RESET_LINK: Triggers SW-UP link reset on other GPUs
+ * @AMD_RESET_BACO: BACO (Bus Alive, Chip Off) method powers off and on the card
+ *                  but without powering off the PCI bus. Suitable only for
+ *                  discrete GPUs.
+ * @AMD_RESET_PCI: Does a full bus reset using core Linux subsystem PCI reset
+ *                 and does a secondary bus reset or FLR, depending on what the
+ *                 underlying hardware supports.
+ *
+ * Methods available for AMD GPU driver for resetting the device. Not all
+ * methods are suitable for every device. User can override the method using
+ * module parameter `reset_method`.
+ */
 enum amd_reset_method {
 	AMD_RESET_METHOD_NONE = -1,
 	AMD_RESET_METHOD_LEGACY = 0,
 	AMD_RESET_METHOD_MODE0,
 	AMD_RESET_METHOD_MODE1,
 	AMD_RESET_METHOD_MODE2,
+	AMD_RESET_METHOD_LINK,
 	AMD_RESET_METHOD_BACO,
 	AMD_RESET_METHOD_PCI,
+	AMD_RESET_METHOD_ON_INIT,
 };
 
 struct amdgpu_video_codec_info {
@@ -565,7 +691,7 @@ struct amdgpu_asic_funcs {
 	/* PCIe replay counter */
 	uint64_t (*get_pcie_replay_count)(struct amdgpu_device *adev);
 	/* device supports BACO */
-	bool (*supports_baco)(struct amdgpu_device *adev);
+	int (*supports_baco)(struct amdgpu_device *adev);
 	/* pre asic_init quirks */
 	void (*pre_asic_init)(struct amdgpu_device *adev);
 	/* enter/exit umd stable pstate */
@@ -573,6 +699,12 @@ struct amdgpu_asic_funcs {
 	/* query video codecs */
 	int (*query_video_codecs)(struct amdgpu_device *adev, bool encode,
 				  const struct amdgpu_video_codecs **codecs);
+	/* encode "> 32bits" smn addressing */
+	u64 (*encode_ext_smn_addressing)(int ext_id);
+
+	ssize_t (*get_reg_state)(struct amdgpu_device *adev,
+				 enum amdgpu_reg_state reg_state, void *buf,
+				 size_t max_size);
 };
 
 /*
@@ -591,7 +723,7 @@ int amdgpu_cs_wait_fences_ioctl(struct drm_device *dev, void *data,
 /* VRAM scratch page for HDP bug, default vram page */
 struct amdgpu_mem_scratch {
 	struct amdgpu_bo		*robj;
-	volatile uint32_t		*ptr;
+	uint32_t			*ptr;
 	u64				gpu_addr;
 };
 
@@ -607,15 +739,22 @@ void amdgpu_cgs_destroy_device(struct cgs_device *cgs_device);
 typedef uint32_t (*amdgpu_rreg_t)(struct amdgpu_device*, uint32_t);
 typedef void (*amdgpu_wreg_t)(struct amdgpu_device*, uint32_t, uint32_t);
 
+typedef uint32_t (*amdgpu_rreg_ext_t)(struct amdgpu_device*, uint64_t);
+typedef void (*amdgpu_wreg_ext_t)(struct amdgpu_device*, uint64_t, uint32_t);
+
 typedef uint64_t (*amdgpu_rreg64_t)(struct amdgpu_device*, uint32_t);
 typedef void (*amdgpu_wreg64_t)(struct amdgpu_device*, uint32_t, uint64_t);
 
+typedef uint64_t (*amdgpu_rreg64_ext_t)(struct amdgpu_device*, uint64_t);
+typedef void (*amdgpu_wreg64_ext_t)(struct amdgpu_device*, uint64_t, uint64_t);
+
 typedef uint32_t (*amdgpu_block_rreg_t)(struct amdgpu_device*, uint32_t, uint32_t);
 typedef void (*amdgpu_block_wreg_t)(struct amdgpu_device*, uint32_t, uint32_t, uint32_t);
 
 struct amdgpu_mmio_remap {
 	u32 reg_offset;
 	resource_size_t bus_addr;
+	struct amdgpu_bo *bo;
 };
 
 /* Define the HW IP blocks will be used in driver , add more if necessary */
@@ -641,6 +780,7 @@ enum amd_hw_ip_block_type {
 	JPEG_HWIP = VCN_HWIP,
 	VCN1_HWIP,
 	VCE_HWIP,
+	VPE_HWIP,
 	DF_HWIP,
 	DCE_HWIP,
 	OSSSYS_HWIP,
@@ -654,24 +794,53 @@ enum amd_hw_ip_block_type {
 	XGMI_HWIP,
 	DCI_HWIP,
 	PCIE_HWIP,
+	ISP_HWIP,
 	MAX_HWIP
 };
 
-#define HWIP_MAX_INSTANCE	28
+#define HWIP_MAX_INSTANCE	44
 
 #define HW_ID_MAX		300
-#define IP_VERSION(mj, mn, rv) (((mj) << 16) | ((mn) << 8) | (rv))
-#define IP_VERSION_MAJ(ver) ((ver) >> 16)
-#define IP_VERSION_MIN(ver) (((ver) >> 8) & 0xFF)
-#define IP_VERSION_REV(ver) ((ver) & 0xFF)
+#define IP_VERSION_FULL(mj, mn, rv, var, srev) \
+	(((mj) << 24) | ((mn) << 16) | ((rv) << 8) | ((var) << 4) | (srev))
+#define IP_VERSION(mj, mn, rv)		IP_VERSION_FULL(mj, mn, rv, 0, 0)
+#define IP_VERSION_MAJ(ver)		((ver) >> 24)
+#define IP_VERSION_MIN(ver)		(((ver) >> 16) & 0xFF)
+#define IP_VERSION_REV(ver)		(((ver) >> 8) & 0xFF)
+#define IP_VERSION_VARIANT(ver)		(((ver) >> 4) & 0xF)
+#define IP_VERSION_SUBREV(ver)		((ver) & 0xF)
+#define IP_VERSION_MAJ_MIN_REV(ver)	((ver) >> 8)
+
+struct amdgpu_ip_map_info {
+	/* Map of logical to actual dev instances/mask */
+	uint32_t 		dev_inst[MAX_HWIP][HWIP_MAX_INSTANCE];
+	int8_t (*logical_to_dev_inst)(struct amdgpu_device *adev,
+				      enum amd_hw_ip_block_type block,
+				      int8_t inst);
+	uint32_t (*logical_to_dev_mask)(struct amdgpu_device *adev,
+					enum amd_hw_ip_block_type block,
+					uint32_t mask);
+};
+
+enum amdgpu_uid_type {
+	AMDGPU_UID_TYPE_XCD,
+	AMDGPU_UID_TYPE_AID,
+	AMDGPU_UID_TYPE_SOC,
+	AMDGPU_UID_TYPE_MAX
+};
+
+#define AMDGPU_UID_INST_MAX 8 /* max number of instances for each UID type */
+
+struct amdgpu_uid {
+	uint64_t uid[AMDGPU_UID_TYPE_MAX][AMDGPU_UID_INST_MAX];
+	struct amdgpu_device *adev;
+};
 
 struct amd_powerplay {
 	void *pp_handle;
 	const struct amd_pm_funcs *pp_funcs;
 };
 
-struct ip_discovery_top;
-
 /* polaris10 kickers */
 #define ASICID_IS_P20(did, rid)		(((did == 0x67DF) && \
 					 ((rid == 0xE3) || \
@@ -722,7 +891,14 @@ struct amdgpu_mqd_prop {
 	uint64_t eop_gpu_addr;
 	uint32_t hqd_pipe_priority;
 	uint32_t hqd_queue_priority;
+	bool allow_tunneling;
 	bool hqd_active;
+	uint64_t shadow_addr;
+	uint64_t gds_bkup_addr;
+	uint64_t csa_addr;
+	uint64_t fence_address;
+	bool tmz_queue;
+	bool kernel_queue;
 };
 
 struct amdgpu_mqd {
@@ -731,15 +907,45 @@ struct amdgpu_mqd {
 			struct amdgpu_mqd_prop *p);
 };
 
+struct amdgpu_pcie_reset_ctx {
+	bool in_link_reset;
+	bool occurs_dpc;
+	bool audio_suspended;
+	struct pci_dev *swus;
+	struct pci_saved_state *swus_pcistate;
+	struct pci_saved_state *swds_pcistate;
+};
+
+/*
+ * Custom Init levels could be defined for different situations where a full
+ * initialization of all hardware blocks are not expected. Sample cases are
+ * custom init sequences after resume after S0i3/S3, reset on initialization,
+ * partial reset of blocks etc. Presently, this defines only two levels. Levels
+ * are described in corresponding struct definitions - amdgpu_init_default,
+ * amdgpu_init_minimal_xgmi.
+ */
+enum amdgpu_init_lvl_id {
+	AMDGPU_INIT_LEVEL_DEFAULT,
+	AMDGPU_INIT_LEVEL_MINIMAL_XGMI,
+	AMDGPU_INIT_LEVEL_RESET_RECOVERY,
+};
+
+struct amdgpu_init_level {
+	enum amdgpu_init_lvl_id level;
+	uint32_t hwini_ip_block_mask;
+};
+
 #define AMDGPU_RESET_MAGIC_NUM 64
 #define AMDGPU_MAX_DF_PERFMONS 4
-#define AMDGPU_PRODUCT_NAME_LEN 64
 struct amdgpu_reset_domain;
+struct amdgpu_fru_info;
 
-/*
- * Non-zero (true) if the GPU has VRAM. Zero (false) otherwise.
- */
-#define AMDGPU_HAS_VRAM(_adev) ((_adev)->gmc.real_vram_size)
+enum amdgpu_enforce_isolation_mode {
+	AMDGPU_ENFORCE_ISOLATION_DISABLE = 0,
+	AMDGPU_ENFORCE_ISOLATION_ENABLE = 1,
+	AMDGPU_ENFORCE_ISOLATION_ENABLE_LEGACY = 2,
+	AMDGPU_ENFORCE_ISOLATION_NO_CLEANER_SHADER = 3,
+};
 
 struct amdgpu_device {
 	struct device			*dev;
@@ -750,6 +956,7 @@ struct amdgpu_device {
 	struct amdgpu_acp		acp;
 #endif
 	struct amdgpu_hive_info *hive;
+	struct amdgpu_xcp_mgr *xcp_mgr;
 	/* ASIC */
 	enum amd_asic_type		asic_type;
 	uint32_t			family;
@@ -763,9 +970,9 @@ struct amdgpu_device {
 	bool				need_swiotlb;
 	bool				accel_working;
 	struct notifier_block		acpi_nb;
+	struct notifier_block		pm_nb;
 	struct amdgpu_i2c_chan		*i2c_bus[AMDGPU_MAX_I2C_BUS];
-	struct debugfs_blob_wrapper     debugfs_vbios_blob;
-	struct debugfs_blob_wrapper     debugfs_discovery_blob;
+	struct debugfs_blob_wrapper debugfs_vbios_blob;
 	struct mutex			srbm_mutex;
 	/* GRBM index mutex. Protects concurrent access to GRBM index */
 	struct mutex                    grbm_idx_mutex;
@@ -797,8 +1004,12 @@ struct amdgpu_device {
 	amdgpu_wreg_t			pcie_wreg;
 	amdgpu_rreg_t			pciep_rreg;
 	amdgpu_wreg_t			pciep_wreg;
+	amdgpu_rreg_ext_t		pcie_rreg_ext;
+	amdgpu_wreg_ext_t		pcie_wreg_ext;
 	amdgpu_rreg64_t			pcie_rreg64;
 	amdgpu_wreg64_t			pcie_wreg64;
+	amdgpu_rreg64_ext_t			pcie_rreg64_ext;
+	amdgpu_wreg64_ext_t			pcie_wreg64_ext;
 	/* protects concurrent UVD register access */
 	spinlock_t uvd_ctx_idx_lock;
 	amdgpu_rreg_t			uvd_ctx_rreg;
@@ -830,7 +1041,7 @@ struct amdgpu_device {
 	dma_addr_t			dummy_page_addr;
 	struct amdgpu_vm_manager	vm_manager;
 	struct amdgpu_vmhub             vmhub[AMDGPU_MAX_VMHUBS];
-	unsigned			num_vmhubs;
+	DECLARE_BITMAP(vmhubs_mask, AMDGPU_MAX_VMHUBS);
 
 	/* memory management */
 	struct amdgpu_mman		mman;
@@ -851,6 +1062,9 @@ struct amdgpu_device {
 		u32			log2_max_MBps;
 	} mm_stats;
 
+	/* discovery*/
+	struct amdgpu_discovery_info discovery;
+
 	/* display */
 	bool				enable_virtual_display;
 	struct amdgpu_vkms_output       *amdgpu_vkms_output;
@@ -919,6 +1133,13 @@ struct amdgpu_device {
 	/* jpeg */
 	struct amdgpu_jpeg		jpeg;
 
+	/* vpe */
+	struct amdgpu_vpe		vpe;
+
+	/* umsch */
+	struct amdgpu_umsch_mm		umsch_mm;
+	bool				enable_umsch_mm;
+
 	/* firmwares */
 	struct amdgpu_firmware		firmware;
 
@@ -928,8 +1149,8 @@ struct amdgpu_device {
 	/* GDS */
 	struct amdgpu_gds		gds;
 
-	/* KFD */
-	struct amdgpu_kfd_dev		kfd;
+	/* for userq and VM fences */
+	struct amdgpu_seq64		seq64;
 
 	/* UMC */
 	struct amdgpu_umc		umc;
@@ -937,11 +1158,30 @@ struct amdgpu_device {
 	/* display related functionality */
 	struct amdgpu_display_manager dm;
 
+#if defined(CONFIG_DRM_AMD_ISP)
+	/* isp */
+	struct amdgpu_isp		isp;
+#endif
+
 	/* mes */
 	bool                            enable_mes;
 	bool                            enable_mes_kiq;
+	bool                            enable_uni_mes;
 	struct amdgpu_mes               mes;
 	struct amdgpu_mqd               mqds[AMDGPU_HW_IP_NUM];
+	const struct amdgpu_userq_funcs *userq_funcs[AMDGPU_HW_IP_NUM];
+
+	/* xarray used to retrieve the user queue fence driver reference
+	 * in the EOP interrupt handler to signal the particular user
+	 * queue fence.
+	 */
+	struct xarray			userq_xa;
+	/**
+	 * @userq_doorbell_xa: Global user queue map (doorbell index → queue)
+	 * Key: doorbell_index (unique global identifier for the queue)
+	 * Value: struct amdgpu_usermode_queue
+	 */
+	struct xarray userq_doorbell_xa;
 
 	/* df */
 	struct amdgpu_df                df;
@@ -949,6 +1189,12 @@ struct amdgpu_device {
 	/* MCA */
 	struct amdgpu_mca               mca;
 
+	/* ACA */
+	struct amdgpu_aca		aca;
+
+	/* CPER */
+	struct amdgpu_cper		cper;
+
 	struct amdgpu_ip_block          ip_blocks[AMDGPU_MAX_IP_NUM];
 	uint32_t		        harvest_ip_mask;
 	int				num_ip_blocks;
@@ -962,16 +1208,13 @@ struct amdgpu_device {
 
 	/* soc15 register offset based on ip, instance and  segment */
 	uint32_t		*reg_offset[MAX_HWIP][HWIP_MAX_INSTANCE];
+	struct amdgpu_ip_map_info	ip_map;
 
 	/* delayed work_func for deferring clockgating during resume */
 	struct delayed_work     delayed_init_work;
 
 	struct amdgpu_virt	virt;
 
-	/* link all shadow bo */
-	struct list_head                shadow_list;
-	struct mutex                    shadow_list_lock;
-
 	/* record hw reset is performed */
 	bool has_hw_reset;
 	u8				reset_magic[AMDGPU_RESET_MAGIC_NUM];
@@ -981,6 +1224,7 @@ struct amdgpu_device {
 	bool				in_s3;
 	bool				in_s4;
 	bool				in_s0ix;
+	suspend_state_t			last_suspend_state;
 
 	enum pp_mp1_state               mp1_state;
 	struct amdgpu_doorbell_index doorbell_index;
@@ -995,6 +1239,7 @@ struct amdgpu_device {
 	long				sdma_timeout;
 	long				video_timeout;
 	long				compute_timeout;
+	long				psp_timeout;
 
 	uint64_t			unique_id;
 	uint64_t	df_perfmon_config_assign_mask[AMDGPU_MAX_DF_PERFMONS];
@@ -1004,22 +1249,23 @@ struct amdgpu_device {
 	bool                            has_pr3;
 
 	bool                            ucode_sysfs_en;
-	bool                            psp_sysfs_en;
-
-	/* Chip product information */
-	char				product_number[20];
-	char				product_name[AMDGPU_PRODUCT_NAME_LEN];
-	char				serial[20];
 
+	struct amdgpu_fru_info		*fru_info;
 	atomic_t			throttling_logging_enabled;
 	struct ratelimit_state		throttling_logging_rs;
 	uint32_t                        ras_hw_enabled;
 	uint32_t                        ras_enabled;
+	bool                            ras_default_ecc_enabled;
 
 	bool                            no_hw_access;
 	struct pci_saved_state          *pci_state;
 	pci_channel_state_t		pci_channel_state;
 
+	struct amdgpu_pcie_reset_ctx	pcie_reset_ctx;
+
+	/* Track auto wait count on s_barrier settings */
+	bool				barrier_has_auto_waitcnt;
+
 	struct amdgpu_reset_control     *reset_cntl;
 	uint32_t                        ip_versions[MAX_HWIP][HWIP_MAX_INSTANCE];
 
@@ -1027,31 +1273,74 @@ struct amdgpu_device {
 
 	struct list_head                ras_list;
 
-	struct ip_discovery_top         *ip_top;
-
 	struct amdgpu_reset_domain	*reset_domain;
 
 	struct mutex			benchmark_mutex;
 
-	/* reset dump register */
-	uint32_t                        *reset_dump_reg_list;
-	uint32_t			*reset_dump_reg_value;
-	int                             num_regs;
-#ifdef CONFIG_DEV_COREDUMP
-	struct amdgpu_task_info         reset_task_info;
-	bool                            reset_vram_lost;
-	struct timespec64               reset_time;
-#endif
-
 	bool                            scpm_enabled;
 	uint32_t                        scpm_status;
 
 	struct work_struct		reset_work;
 
-	bool                            job_hang;
 	bool                            dc_enabled;
+	/* Mask of active clusters */
+	uint32_t			aid_mask;
+
+	/* Debug */
+	bool                            debug_vm;
+	bool                            debug_largebar;
+	bool                            debug_disable_soft_recovery;
+	bool                            debug_use_vram_fw_buf;
+	bool                            debug_enable_ras_aca;
+	bool                            debug_exp_resets;
+	bool                            debug_disable_gpu_ring_reset;
+	bool                            debug_vm_userptr;
+	bool                            debug_disable_ce_logs;
+	bool                            debug_enable_ce_cs;
+
+	/* Protection for the following isolation structure */
+	struct mutex                    enforce_isolation_mutex;
+	enum amdgpu_enforce_isolation_mode	enforce_isolation[MAX_XCP];
+	struct amdgpu_isolation {
+		void			*owner;
+		struct dma_fence	*spearhead;
+		struct amdgpu_sync	active;
+		struct amdgpu_sync	prev;
+	} isolation[MAX_XCP];
+
+	struct amdgpu_init_level *init_lvl;
+
+	/* This flag is used to determine how VRAM allocations are handled for APUs
+	 * in KFD: VRAM or GTT.
+	 */
+	bool                            apu_prefer_gtt;
+
+	bool                            userq_halt_for_enforce_isolation;
+	struct work_struct              userq_reset_work;
+	struct amdgpu_uid *uid_info;
+
+	/* KFD
+	 * Must be last --ends in a flexible-array member.
+	 */
+	struct amdgpu_kfd_dev		kfd;
 };
 
+static inline uint32_t amdgpu_ip_version(const struct amdgpu_device *adev,
+					 uint8_t ip, uint8_t inst)
+{
+	/* This considers only major/minor/rev and ignores
+	 * subrevision/variant fields.
+	 */
+	return adev->ip_versions[ip][inst] & ~0xFFU;
+}
+
+static inline uint32_t amdgpu_ip_version_full(const struct amdgpu_device *adev,
+					      uint8_t ip, uint8_t inst)
+{
+	/* This returns full version - major/minor/rev/variant/subrevision */
+	return adev->ip_versions[ip][inst];
+}
+
 static inline struct amdgpu_device *drm_to_adev(struct drm_device *ddev)
 {
 	return container_of(ddev, struct amdgpu_device, ddev);
@@ -1067,6 +1356,11 @@ static inline struct amdgpu_device *amdgpu_ttm_adev(struct ttm_device *bdev)
 	return container_of(bdev, struct amdgpu_device, mman.bdev);
 }
 
+static inline bool amdgpu_is_multi_aid(struct amdgpu_device *adev)
+{
+	return !!adev->aid_mask;
+}
+
 int amdgpu_device_init(struct amdgpu_device *adev,
 		       uint32_t flags);
 void amdgpu_device_fini_hw(struct amdgpu_device *adev);
@@ -1081,13 +1375,27 @@ size_t amdgpu_device_aper_access(struct amdgpu_device *adev, loff_t pos,
 
 void amdgpu_device_vram_access(struct amdgpu_device *adev, loff_t pos,
 			       void *buf, size_t size, bool write);
+uint32_t amdgpu_device_wait_on_rreg(struct amdgpu_device *adev,
+			    uint32_t inst, uint32_t reg_addr, char reg_name[],
+			    uint32_t expected_value, uint32_t mask);
 uint32_t amdgpu_device_rreg(struct amdgpu_device *adev,
 			    uint32_t reg, uint32_t acc_flags);
+u32 amdgpu_device_indirect_rreg_ext(struct amdgpu_device *adev,
+				    u64 reg_addr);
+uint32_t amdgpu_device_xcc_rreg(struct amdgpu_device *adev,
+				uint32_t reg, uint32_t acc_flags,
+				uint32_t xcc_id);
 void amdgpu_device_wreg(struct amdgpu_device *adev,
 			uint32_t reg, uint32_t v,
 			uint32_t acc_flags);
+void amdgpu_device_indirect_wreg_ext(struct amdgpu_device *adev,
+				     u64 reg_addr, u32 reg_data);
+void amdgpu_device_xcc_wreg(struct amdgpu_device *adev,
+			    uint32_t reg, uint32_t v,
+			    uint32_t acc_flags,
+			    uint32_t xcc_id);
 void amdgpu_mm_wreg_mmio_rlc(struct amdgpu_device *adev,
-			     uint32_t reg, uint32_t v);
+			     uint32_t reg, uint32_t v, uint32_t xcc_id);
 void amdgpu_mm_wreg8(struct amdgpu_device *adev, uint32_t offset, uint8_t value);
 uint8_t amdgpu_mm_rreg8(struct amdgpu_device *adev, uint32_t offset);
 
@@ -1095,12 +1403,17 @@ u32 amdgpu_device_indirect_rreg(struct amdgpu_device *adev,
 				u32 reg_addr);
 u64 amdgpu_device_indirect_rreg64(struct amdgpu_device *adev,
 				  u32 reg_addr);
+u64 amdgpu_device_indirect_rreg64_ext(struct amdgpu_device *adev,
+				  u64 reg_addr);
 void amdgpu_device_indirect_wreg(struct amdgpu_device *adev,
 				 u32 reg_addr, u32 reg_data);
 void amdgpu_device_indirect_wreg64(struct amdgpu_device *adev,
 				   u32 reg_addr, u64 reg_data);
+void amdgpu_device_indirect_wreg64_ext(struct amdgpu_device *adev,
+				   u64 reg_addr, u64 reg_data);
 u32 amdgpu_device_get_rev_id(struct amdgpu_device *adev);
-bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type);
+bool amdgpu_device_asic_has_dc_support(struct pci_dev *pdev,
+				       enum amd_asic_type asic_type);
 bool amdgpu_device_has_dc_support(struct amdgpu_device *adev);
 
 void amdgpu_device_set_sriov_virtual_display(struct amdgpu_device *adev);
@@ -1111,6 +1424,8 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
 int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 			 struct amdgpu_reset_context *reset_context);
 
+int amdgpu_device_reinit_after_reset(struct amdgpu_reset_context *reset_context);
+
 int emu_soc_asic_init(struct amdgpu_device *adev);
 
 /*
@@ -1122,8 +1437,8 @@ int emu_soc_asic_init(struct amdgpu_device *adev);
 #define RREG32_NO_KIQ(reg) amdgpu_device_rreg(adev, (reg), AMDGPU_REGS_NO_KIQ)
 #define WREG32_NO_KIQ(reg, v) amdgpu_device_wreg(adev, (reg), (v), AMDGPU_REGS_NO_KIQ)
 
-#define RREG32_KIQ(reg) amdgpu_kiq_rreg(adev, (reg))
-#define WREG32_KIQ(reg, v) amdgpu_kiq_wreg(adev, (reg), (v))
+#define RREG32_KIQ(reg) amdgpu_kiq_rreg(adev, (reg), 0)
+#define WREG32_KIQ(reg, v) amdgpu_kiq_wreg(adev, (reg), (v), 0)
 
 #define RREG8(reg) amdgpu_mm_rreg8(adev, (reg))
 #define WREG8(reg, v) amdgpu_mm_wreg8(adev, (reg), (v))
@@ -1133,12 +1448,18 @@ int emu_soc_asic_init(struct amdgpu_device *adev);
 #define WREG32(reg, v) amdgpu_device_wreg(adev, (reg), (v), 0)
 #define REG_SET(FIELD, v) (((v) << FIELD##_SHIFT) & FIELD##_MASK)
 #define REG_GET(FIELD, v) (((v) << FIELD##_SHIFT) & FIELD##_MASK)
+#define RREG32_XCC(reg, inst) amdgpu_device_xcc_rreg(adev, (reg), 0, inst)
+#define WREG32_XCC(reg, v, inst) amdgpu_device_xcc_wreg(adev, (reg), (v), 0, inst)
 #define RREG32_PCIE(reg) adev->pcie_rreg(adev, (reg))
 #define WREG32_PCIE(reg, v) adev->pcie_wreg(adev, (reg), (v))
 #define RREG32_PCIE_PORT(reg) adev->pciep_rreg(adev, (reg))
 #define WREG32_PCIE_PORT(reg, v) adev->pciep_wreg(adev, (reg), (v))
+#define RREG32_PCIE_EXT(reg) adev->pcie_rreg_ext(adev, (reg))
+#define WREG32_PCIE_EXT(reg, v) adev->pcie_wreg_ext(adev, (reg), (v))
 #define RREG64_PCIE(reg) adev->pcie_rreg64(adev, (reg))
 #define WREG64_PCIE(reg, v) adev->pcie_wreg64(adev, (reg), (v))
+#define RREG64_PCIE_EXT(reg) adev->pcie_rreg64_ext(adev, (reg))
+#define WREG64_PCIE_EXT(reg, v) adev->pcie_wreg64_ext(adev, (reg), (v))
 #define RREG32_SMC(reg) adev->smc_rreg(adev, (reg))
 #define WREG32_SMC(reg, v) adev->smc_wreg(adev, (reg), (v))
 #define RREG32_UVD_CTX(reg) adev->uvd_ctx_rreg(adev, (reg))
@@ -1194,6 +1515,7 @@ int emu_soc_asic_init(struct amdgpu_device *adev);
 #define WREG32_FIELD_OFFSET(reg, offset, field, val)	\
 	WREG32(mm##reg + offset, (RREG32(mm##reg + offset) & ~REG_FIELD_MASK(reg, field)) | (val) << REG_FIELD_SHIFT(reg, field))
 
+#define AMDGPU_GET_REG_FIELD(x, h, l) (((x) & GENMASK_ULL(h, l)) >> (l))
 /*
  * BIOS helpers.
  */
@@ -1204,7 +1526,8 @@ int emu_soc_asic_init(struct amdgpu_device *adev);
 /*
  * ASICs macro.
  */
-#define amdgpu_asic_set_vga_state(adev, state) (adev)->asic_funcs->set_vga_state((adev), (state))
+#define amdgpu_asic_set_vga_state(adev, state) \
+    ((adev)->asic_funcs->set_vga_state ? (adev)->asic_funcs->set_vga_state((adev), (state)) : 0)
 #define amdgpu_asic_reset(adev) (adev)->asic_funcs->reset((adev))
 #define amdgpu_asic_reset_method(adev) (adev)->asic_funcs->reset_method((adev))
 #define amdgpu_asic_get_xclk(adev) (adev)->asic_funcs->get_xclk((adev))
@@ -1217,11 +1540,6 @@ int emu_soc_asic_init(struct amdgpu_device *adev);
 #define amdgpu_asic_read_bios_from_rom(adev, b, l) (adev)->asic_funcs->read_bios_from_rom((adev), (b), (l))
 #define amdgpu_asic_read_register(adev, se, sh, offset, v)((adev)->asic_funcs->read_register((adev), (se), (sh), (offset), (v)))
 #define amdgpu_asic_get_config_memsize(adev) (adev)->asic_funcs->get_config_memsize((adev))
-#define amdgpu_asic_flush_hdp(adev, r) \
-	((adev)->asic_funcs->flush_hdp ? (adev)->asic_funcs->flush_hdp((adev), (r)) : (adev)->hdp.funcs->flush_hdp((adev), (r)))
-#define amdgpu_asic_invalidate_hdp(adev, r) \
-	((adev)->asic_funcs->invalidate_hdp ? (adev)->asic_funcs->invalidate_hdp((adev), (r)) : \
-	 ((adev)->hdp.funcs->invalidate_hdp ? (adev)->hdp.funcs->invalidate_hdp((adev), (r)) : (void)0))
 #define amdgpu_asic_need_full_reset(adev) (adev)->asic_funcs->need_full_reset((adev))
 #define amdgpu_asic_init_doorbell_index(adev) (adev)->asic_funcs->init_doorbell_index((adev))
 #define amdgpu_asic_get_pcie_usage(adev, cnt0, cnt1) ((adev)->asic_funcs->get_pcie_usage((adev), (cnt0), (cnt1)))
@@ -1233,9 +1551,12 @@ int emu_soc_asic_init(struct amdgpu_device *adev);
 	((adev)->asic_funcs->update_umd_stable_pstate ? (adev)->asic_funcs->update_umd_stable_pstate((adev), (enter)) : 0)
 #define amdgpu_asic_query_video_codecs(adev, e, c) (adev)->asic_funcs->query_video_codecs((adev), (e), (c))
 
-#define amdgpu_inc_vram_lost(adev) atomic_inc(&((adev)->vram_lost_counter));
+#define amdgpu_inc_vram_lost(adev) atomic_inc(&((adev)->vram_lost_counter))
 
-#define MIN(X, Y) ((X) < (Y) ? (X) : (Y))
+#define BIT_MASK_UPPER(i) ((i) >= BITS_PER_LONG ? 0 : ~0UL << (i))
+#define for_each_inst(i, inst_mask)        \
+	for (i = ffs(inst_mask); i-- != 0; \
+	     i = ffs(inst_mask & BIT_MASK_UPPER(i + 1)))
 
 /* Common functions */
 bool amdgpu_device_has_job_running(struct amdgpu_device *adev);
@@ -1246,8 +1567,8 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 void amdgpu_device_pci_config_reset(struct amdgpu_device *adev);
 int amdgpu_device_pci_reset(struct amdgpu_device *adev);
 bool amdgpu_device_need_post(struct amdgpu_device *adev);
+bool amdgpu_device_seamless_boot_supported(struct amdgpu_device *adev);
 bool amdgpu_device_should_use_aspm(struct amdgpu_device *adev);
-bool amdgpu_device_aspm_support_quirk(void);
 
 void amdgpu_cs_report_moved_bytes(struct amdgpu_device *adev, u64 num_bytes,
 				  u64 num_vis_bytes);
@@ -1257,15 +1578,17 @@ void amdgpu_device_program_register_sequence(struct amdgpu_device *adev,
 					     const u32 array_size);
 
 int amdgpu_device_mode1_reset(struct amdgpu_device *adev);
-bool amdgpu_device_supports_atpx(struct drm_device *dev);
-bool amdgpu_device_supports_px(struct drm_device *dev);
-bool amdgpu_device_supports_boco(struct drm_device *dev);
-bool amdgpu_device_supports_smart_shift(struct drm_device *dev);
-bool amdgpu_device_supports_baco(struct drm_device *dev);
+int amdgpu_device_link_reset(struct amdgpu_device *adev);
+bool amdgpu_device_supports_atpx(struct amdgpu_device *adev);
+bool amdgpu_device_supports_px(struct amdgpu_device *adev);
+bool amdgpu_device_supports_boco(struct amdgpu_device *adev);
+bool amdgpu_device_supports_smart_shift(struct amdgpu_device *adev);
+int amdgpu_device_supports_baco(struct amdgpu_device *adev);
+void amdgpu_device_detect_runtime_pm_mode(struct amdgpu_device *adev);
 bool amdgpu_device_is_peer_accessible(struct amdgpu_device *adev,
 				      struct amdgpu_device *peer_adev);
-int amdgpu_device_baco_enter(struct drm_device *dev);
-int amdgpu_device_baco_exit(struct drm_device *dev);
+int amdgpu_device_baco_enter(struct amdgpu_device *adev);
+int amdgpu_device_baco_exit(struct amdgpu_device *adev);
 
 void amdgpu_device_flush_hdp(struct amdgpu_device *adev,
 		struct amdgpu_ring *ring);
@@ -1277,9 +1600,15 @@ u32 amdgpu_device_pcie_port_rreg(struct amdgpu_device *adev,
 				u32 reg);
 void amdgpu_device_pcie_port_wreg(struct amdgpu_device *adev,
 				u32 reg, u32 v);
+struct dma_fence *amdgpu_device_get_gang(struct amdgpu_device *adev);
 struct dma_fence *amdgpu_device_switch_gang(struct amdgpu_device *adev,
 					    struct dma_fence *gang);
+struct dma_fence *amdgpu_device_enforce_isolation(struct amdgpu_device *adev,
+						  struct amdgpu_ring *ring,
+						  struct amdgpu_job *job);
 bool amdgpu_device_has_display_hardware(struct amdgpu_device *adev);
+ssize_t amdgpu_get_soft_full_reset_mask(struct amdgpu_ring *ring);
+ssize_t amdgpu_show_reset_mask(char *buf, uint32_t supported_reset);
 
 /* atpx handler */
 #if defined(CONFIG_VGA_SWITCHEROO)
@@ -1287,23 +1616,15 @@ void amdgpu_register_atpx_handler(void);
 void amdgpu_unregister_atpx_handler(void);
 bool amdgpu_has_atpx_dgpu_power_cntl(void);
 bool amdgpu_is_atpx_hybrid(void);
-bool amdgpu_atpx_dgpu_req_power_for_displays(void);
 bool amdgpu_has_atpx(void);
 #else
 static inline void amdgpu_register_atpx_handler(void) {}
 static inline void amdgpu_unregister_atpx_handler(void) {}
 static inline bool amdgpu_has_atpx_dgpu_power_cntl(void) { return false; }
 static inline bool amdgpu_is_atpx_hybrid(void) { return false; }
-static inline bool amdgpu_atpx_dgpu_req_power_for_displays(void) { return false; }
 static inline bool amdgpu_has_atpx(void) { return false; }
 #endif
 
-#if defined(CONFIG_VGA_SWITCHEROO) && defined(CONFIG_ACPI)
-void *amdgpu_atpx_get_dhandle(void);
-#else
-static inline void *amdgpu_atpx_get_dhandle(void) { return NULL; }
-#endif
-
 /*
  * KMS
  */
@@ -1312,13 +1633,13 @@ extern const int amdgpu_max_kms_ioctl;
 
 int amdgpu_driver_load_kms(struct amdgpu_device *adev, unsigned long flags);
 void amdgpu_driver_unload_kms(struct drm_device *dev);
-void amdgpu_driver_lastclose_kms(struct drm_device *dev);
 int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv);
 void amdgpu_driver_postclose_kms(struct drm_device *dev,
 				 struct drm_file *file_priv);
 void amdgpu_driver_release_kms(struct drm_device *dev);
 
-int amdgpu_device_ip_suspend(struct amdgpu_device *adev);
+int amdgpu_device_prepare(struct drm_device *dev);
+void amdgpu_device_complete(struct drm_device *dev);
 int amdgpu_device_suspend(struct drm_device *dev, bool fbcon);
 int amdgpu_device_resume(struct drm_device *dev, bool fbcon);
 u32 amdgpu_get_vblank_counter_kms(struct drm_crtc *crtc);
@@ -1348,6 +1669,12 @@ struct amdgpu_afmt_acr amdgpu_afmt_acr(uint32_t clock);
 
 /* amdgpu_acpi.c */
 
+struct amdgpu_numa_info {
+	uint64_t size;
+	int pxm;
+	int nid;
+};
+
 /* ATCS Device/Driver State */
 #define AMDGPU_ATCS_PSC_DEV_STATE_D0		0
 #define AMDGPU_ATCS_PSC_DEV_STATE_D3_HOT	3
@@ -1363,22 +1690,44 @@ int amdgpu_acpi_pcie_performance_request(struct amdgpu_device *adev,
 						u8 perf_req, bool advertise);
 int amdgpu_acpi_power_shift_control(struct amdgpu_device *adev,
 				    u8 dev_state, bool drv_state);
-int amdgpu_acpi_smart_shift_update(struct drm_device *dev, enum amdgpu_ss ss_state);
+int amdgpu_acpi_smart_shift_update(struct amdgpu_device *adev,
+				   enum amdgpu_ss ss_state);
 int amdgpu_acpi_pcie_notify_device_ready(struct amdgpu_device *adev);
+int amdgpu_acpi_get_tmr_info(struct amdgpu_device *adev, u64 *tmr_offset,
+			     u64 *tmr_size);
+int amdgpu_acpi_get_mem_info(struct amdgpu_device *adev, int xcc_id,
+			     struct amdgpu_numa_info *numa_info);
 
 void amdgpu_acpi_get_backlight_caps(struct amdgpu_dm_backlight_caps *caps);
 bool amdgpu_acpi_should_gpu_reset(struct amdgpu_device *adev);
 void amdgpu_acpi_detect(void);
+void amdgpu_acpi_release(void);
 #else
 static inline int amdgpu_acpi_init(struct amdgpu_device *adev) { return 0; }
+static inline int amdgpu_acpi_get_tmr_info(struct amdgpu_device *adev,
+					   u64 *tmr_offset, u64 *tmr_size)
+{
+	return -EINVAL;
+}
+static inline int amdgpu_acpi_get_mem_info(struct amdgpu_device *adev,
+					   int xcc_id,
+					   struct amdgpu_numa_info *numa_info)
+{
+	return -EINVAL;
+}
 static inline void amdgpu_acpi_fini(struct amdgpu_device *adev) { }
 static inline bool amdgpu_acpi_should_gpu_reset(struct amdgpu_device *adev) { return false; }
 static inline void amdgpu_acpi_detect(void) { }
+static inline void amdgpu_acpi_release(void) { }
 static inline bool amdgpu_acpi_is_power_shift_control_supported(void) { return false; }
 static inline int amdgpu_acpi_power_shift_control(struct amdgpu_device *adev,
 						  u8 dev_state, bool drv_state) { return 0; }
-static inline int amdgpu_acpi_smart_shift_update(struct drm_device *dev,
-						 enum amdgpu_ss ss_state) { return 0; }
+static inline int amdgpu_acpi_smart_shift_update(struct amdgpu_device *adev,
+						 enum amdgpu_ss ss_state)
+{
+	return 0;
+}
+static inline void amdgpu_acpi_get_backlight_caps(struct amdgpu_dm_backlight_caps *caps) { }
 #endif
 
 #if defined(CONFIG_ACPI) && defined(CONFIG_SUSPEND)
@@ -1389,13 +1738,10 @@ static inline bool amdgpu_acpi_is_s0ix_active(struct amdgpu_device *adev) { retu
 static inline bool amdgpu_acpi_is_s3_active(struct amdgpu_device *adev) { return false; }
 #endif
 
-#if defined(CONFIG_DRM_AMD_DC)
-int amdgpu_dm_display_resume(struct amdgpu_device *adev );
-#else
-static inline int amdgpu_dm_display_resume(struct amdgpu_device *adev) { return 0; }
+#if defined(CONFIG_DRM_AMD_ISP)
+int amdgpu_acpi_get_isp4_dev(struct acpi_device **dev);
 #endif
 
-
 void amdgpu_register_gpu_instance(struct amdgpu_device *adev);
 void amdgpu_unregister_gpu_instance(struct amdgpu_device *adev);
 
@@ -1433,4 +1779,30 @@ static inline bool amdgpu_is_tmz(struct amdgpu_device *adev)
 
 int amdgpu_in_reset(struct amdgpu_device *adev);
 
+extern const struct attribute_group amdgpu_vram_mgr_attr_group;
+extern const struct attribute_group amdgpu_gtt_mgr_attr_group;
+extern const struct attribute_group amdgpu_flash_attr_group;
+
+void amdgpu_set_init_level(struct amdgpu_device *adev,
+			   enum amdgpu_init_lvl_id lvl);
+
+static inline int amdgpu_device_bus_status_check(struct amdgpu_device *adev)
+{
+       u32 status;
+       int r;
+
+       r = pci_read_config_dword(adev->pdev, PCI_COMMAND, &status);
+       if (r || PCI_POSSIBLE_ERROR(status)) {
+		dev_err(adev->dev, "device lost from bus!");
+		return -ENODEV;
+       }
+
+       return 0;
+}
+
+void amdgpu_device_set_uid(struct amdgpu_uid *uid_info,
+			   enum amdgpu_uid_type type, uint8_t inst,
+			   uint64_t uid);
+uint64_t amdgpu_device_get_uid(struct amdgpu_uid *uid_info,
+			       enum amdgpu_uid_type type, uint8_t inst);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c
new file mode 100644
index 000000000000..9b3180449150
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c
@@ -0,0 +1,984 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/list.h>
+#include "amdgpu.h"
+#include "amdgpu_aca.h"
+#include "amdgpu_ras.h"
+
+#define ACA_BANK_HWID(type, hwid, mcatype) [ACA_HWIP_TYPE_##type] = {hwid, mcatype}
+
+typedef int bank_handler_t(struct aca_handle *handle, struct aca_bank *bank, enum aca_smu_type type, void *data);
+
+static struct aca_hwip aca_hwid_mcatypes[ACA_HWIP_TYPE_COUNT] = {
+	ACA_BANK_HWID(SMU,	0x01,	0x01),
+	ACA_BANK_HWID(PCS_XGMI, 0x50,	0x00),
+	ACA_BANK_HWID(UMC,	0x96,	0x00),
+};
+
+static void aca_banks_init(struct aca_banks *banks)
+{
+	if (!banks)
+		return;
+
+	memset(banks, 0, sizeof(*banks));
+	INIT_LIST_HEAD(&banks->list);
+}
+
+static int aca_banks_add_bank(struct aca_banks *banks, struct aca_bank *bank)
+{
+	struct aca_bank_node *node;
+
+	if (!bank)
+		return -EINVAL;
+
+	node = kvzalloc(sizeof(*node), GFP_KERNEL);
+	if (!node)
+		return -ENOMEM;
+
+	memcpy(&node->bank, bank, sizeof(*bank));
+
+	INIT_LIST_HEAD(&node->node);
+	list_add_tail(&node->node, &banks->list);
+
+	banks->nr_banks++;
+
+	return 0;
+}
+
+static void aca_banks_release(struct aca_banks *banks)
+{
+	struct aca_bank_node *node, *tmp;
+
+	if (list_empty(&banks->list))
+		return;
+
+	list_for_each_entry_safe(node, tmp, &banks->list, node) {
+		list_del(&node->node);
+		kvfree(node);
+		banks->nr_banks--;
+	}
+}
+
+static int aca_smu_get_valid_aca_count(struct amdgpu_device *adev, enum aca_smu_type type, u32 *count)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+	const struct aca_smu_funcs *smu_funcs = aca->smu_funcs;
+
+	if (!count)
+		return -EINVAL;
+
+	if (!smu_funcs || !smu_funcs->get_valid_aca_count)
+		return -EOPNOTSUPP;
+
+	return smu_funcs->get_valid_aca_count(adev, type, count);
+}
+
+static struct aca_regs_dump {
+	const char *name;
+	int reg_idx;
+} aca_regs[] = {
+	{"CONTROL",		ACA_REG_IDX_CTL},
+	{"STATUS",		ACA_REG_IDX_STATUS},
+	{"ADDR",		ACA_REG_IDX_ADDR},
+	{"MISC",		ACA_REG_IDX_MISC0},
+	{"CONFIG",		ACA_REG_IDX_CONFIG},
+	{"IPID",		ACA_REG_IDX_IPID},
+	{"SYND",		ACA_REG_IDX_SYND},
+	{"DESTAT",		ACA_REG_IDX_DESTAT},
+	{"DEADDR",		ACA_REG_IDX_DEADDR},
+	{"CONTROL_MASK",	ACA_REG_IDX_CTL_MASK},
+};
+
+static void aca_smu_bank_dump(struct amdgpu_device *adev, int idx, int total, struct aca_bank *bank,
+			      struct ras_query_context *qctx)
+{
+	u64 event_id = qctx ? qctx->evid.event_id : RAS_EVENT_INVALID_ID;
+	int i;
+
+	if (adev->debug_disable_ce_logs &&
+	    bank->smu_err_type == ACA_SMU_TYPE_CE &&
+	    !ACA_BANK_ERR_IS_DEFFERED(bank))
+		return;
+
+	RAS_EVENT_LOG(adev, event_id, HW_ERR "Accelerator Check Architecture events logged\n");
+	/* plus 1 for output format, e.g: ACA[08/08]: xxxx */
+	for (i = 0; i < ARRAY_SIZE(aca_regs); i++)
+		RAS_EVENT_LOG(adev, event_id, HW_ERR "ACA[%02d/%02d].%s=0x%016llx\n",
+			      idx + 1, total, aca_regs[i].name, bank->regs[aca_regs[i].reg_idx]);
+
+	if (ACA_REG__STATUS__SCRUB(bank->regs[ACA_REG_IDX_STATUS]))
+		RAS_EVENT_LOG(adev, event_id, HW_ERR "hardware error logged by the scrubber\n");
+}
+
+static bool aca_bank_hwip_is_matched(struct aca_bank *bank, enum aca_hwip_type type)
+{
+
+	struct aca_hwip *hwip;
+	int hwid, mcatype;
+	u64 ipid;
+
+	if (!bank || type == ACA_HWIP_TYPE_UNKNOW)
+		return false;
+
+	hwip = &aca_hwid_mcatypes[type];
+	if (!hwip->hwid)
+		return false;
+
+	ipid = bank->regs[ACA_REG_IDX_IPID];
+	hwid = ACA_REG__IPID__HARDWAREID(ipid);
+	mcatype = ACA_REG__IPID__MCATYPE(ipid);
+
+	return hwip->hwid == hwid && hwip->mcatype == mcatype;
+}
+
+static int aca_smu_get_valid_aca_banks(struct amdgpu_device *adev, enum aca_smu_type type,
+				       int start, int count,
+				       struct aca_banks *banks, struct ras_query_context *qctx)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+	const struct aca_smu_funcs *smu_funcs = aca->smu_funcs;
+	struct aca_bank bank;
+	int i, max_count, ret;
+
+	if (!count)
+		return 0;
+
+	if (!smu_funcs || !smu_funcs->get_valid_aca_bank)
+		return -EOPNOTSUPP;
+
+	switch (type) {
+	case ACA_SMU_TYPE_UE:
+		max_count = smu_funcs->max_ue_bank_count;
+		break;
+	case ACA_SMU_TYPE_CE:
+		max_count = smu_funcs->max_ce_bank_count;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if (start + count > max_count)
+		return -EINVAL;
+
+	count = min_t(int, count, max_count);
+	for (i = 0; i < count; i++) {
+		memset(&bank, 0, sizeof(bank));
+		ret = smu_funcs->get_valid_aca_bank(adev, type, start + i, &bank);
+		if (ret)
+			return ret;
+
+		bank.smu_err_type = type;
+
+		/*
+		 * Poison being consumed when injecting a UE while running background workloads,
+		 * which are unexpected.
+		 */
+		if (type == ACA_SMU_TYPE_UE &&
+		    ACA_REG__STATUS__POISON(bank.regs[ACA_REG_IDX_STATUS]) &&
+		    !aca_bank_hwip_is_matched(&bank, ACA_HWIP_TYPE_UMC))
+			continue;
+
+		aca_smu_bank_dump(adev, i, count, &bank, qctx);
+
+		ret = aca_banks_add_bank(banks, &bank);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+static bool aca_bank_is_valid(struct aca_handle *handle, struct aca_bank *bank, enum aca_smu_type type)
+{
+	const struct aca_bank_ops *bank_ops = handle->bank_ops;
+
+	/* Parse all deferred errors with UMC aca handle */
+	if (ACA_BANK_ERR_IS_DEFFERED(bank))
+		return handle->hwip == ACA_HWIP_TYPE_UMC;
+
+	if (!aca_bank_hwip_is_matched(bank, handle->hwip))
+		return false;
+
+	if (!bank_ops->aca_bank_is_valid)
+		return true;
+
+	return bank_ops->aca_bank_is_valid(handle, bank, type, handle->data);
+}
+
+static struct aca_bank_error *new_bank_error(struct aca_error *aerr, struct aca_bank_info *info)
+{
+	struct aca_bank_error *bank_error;
+
+	bank_error = kvzalloc(sizeof(*bank_error), GFP_KERNEL);
+	if (!bank_error)
+		return NULL;
+
+	INIT_LIST_HEAD(&bank_error->node);
+	memcpy(&bank_error->info, info, sizeof(*info));
+
+	mutex_lock(&aerr->lock);
+	list_add_tail(&bank_error->node, &aerr->list);
+	aerr->nr_errors++;
+	mutex_unlock(&aerr->lock);
+
+	return bank_error;
+}
+
+static struct aca_bank_error *find_bank_error(struct aca_error *aerr, struct aca_bank_info *info)
+{
+	struct aca_bank_error *bank_error = NULL;
+	struct aca_bank_info *tmp_info;
+	bool found = false;
+
+	mutex_lock(&aerr->lock);
+	list_for_each_entry(bank_error, &aerr->list, node) {
+		tmp_info = &bank_error->info;
+		if (tmp_info->socket_id == info->socket_id &&
+		    tmp_info->die_id == info->die_id) {
+			found = true;
+			goto out_unlock;
+		}
+	}
+
+out_unlock:
+	mutex_unlock(&aerr->lock);
+
+	return found ? bank_error : NULL;
+}
+
+static void aca_bank_error_remove(struct aca_error *aerr, struct aca_bank_error *bank_error)
+{
+	if (!aerr || !bank_error)
+		return;
+
+	list_del(&bank_error->node);
+	aerr->nr_errors--;
+
+	kvfree(bank_error);
+}
+
+static struct aca_bank_error *get_bank_error(struct aca_error *aerr, struct aca_bank_info *info)
+{
+	struct aca_bank_error *bank_error;
+
+	if (!aerr || !info)
+		return NULL;
+
+	bank_error = find_bank_error(aerr, info);
+	if (bank_error)
+		return bank_error;
+
+	return new_bank_error(aerr, info);
+}
+
+int aca_error_cache_log_bank_error(struct aca_handle *handle, struct aca_bank_info *info,
+				   enum aca_error_type type, u64 count)
+{
+	struct aca_error_cache *error_cache = &handle->error_cache;
+	struct aca_bank_error *bank_error;
+	struct aca_error *aerr;
+
+	if (!handle || !info || type >= ACA_ERROR_TYPE_COUNT)
+		return -EINVAL;
+
+	if (!count)
+		return 0;
+
+	aerr = &error_cache->errors[type];
+	bank_error = get_bank_error(aerr, info);
+	if (!bank_error)
+		return -ENOMEM;
+
+	bank_error->count += count;
+
+	return 0;
+}
+
+static int aca_bank_parser(struct aca_handle *handle, struct aca_bank *bank, enum aca_smu_type type)
+{
+	const struct aca_bank_ops *bank_ops = handle->bank_ops;
+
+	if (!bank)
+		return -EINVAL;
+
+	if (!bank_ops->aca_bank_parser)
+		return -EOPNOTSUPP;
+
+	return bank_ops->aca_bank_parser(handle, bank, type,
+					 handle->data);
+}
+
+static int handler_aca_log_bank_error(struct aca_handle *handle, struct aca_bank *bank,
+				      enum aca_smu_type type, void *data)
+{
+	int ret;
+
+	ret = aca_bank_parser(handle, bank, type);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static int aca_dispatch_bank(struct aca_handle_manager *mgr, struct aca_bank *bank,
+			     enum aca_smu_type type, bank_handler_t handler, void *data)
+{
+	struct aca_handle *handle;
+	int ret;
+
+	if (list_empty(&mgr->list))
+		return 0;
+
+	list_for_each_entry(handle, &mgr->list, node) {
+		if (!aca_bank_is_valid(handle, bank, type))
+			continue;
+
+		ret = handler(handle, bank, type, data);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+static int aca_dispatch_banks(struct aca_handle_manager *mgr, struct aca_banks *banks,
+			      enum aca_smu_type type, bank_handler_t handler, void *data)
+{
+	struct aca_bank_node *node;
+	struct aca_bank *bank;
+	int ret;
+
+	if (!mgr || !banks)
+		return -EINVAL;
+
+	/* pre check to avoid unnecessary operations */
+	if (list_empty(&mgr->list) || list_empty(&banks->list))
+		return 0;
+
+	list_for_each_entry(node, &banks->list, node) {
+		bank = &node->bank;
+
+		ret = aca_dispatch_bank(mgr, bank, type, handler, data);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+static bool aca_bank_should_update(struct amdgpu_device *adev, enum aca_smu_type type)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+	bool ret = true;
+
+	/*
+	 * Because the UE Valid MCA count will only be cleared after reset,
+	 * in order to avoid repeated counting of the error count,
+	 * the aca bank is only updated once during the gpu recovery stage.
+	 */
+	if (type == ACA_SMU_TYPE_UE) {
+		if (amdgpu_ras_intr_triggered())
+			ret = atomic_cmpxchg(&aca->ue_update_flag, 0, 1) == 0;
+		else
+			atomic_set(&aca->ue_update_flag, 0);
+	}
+
+	return ret;
+}
+
+static void aca_banks_generate_cper(struct amdgpu_device *adev,
+				    enum aca_smu_type type,
+				    struct aca_banks *banks,
+				    int count)
+{
+	struct aca_bank_node *node;
+	struct aca_bank *bank;
+	int r;
+
+	if (!adev->cper.enabled)
+		return;
+
+	if (!banks || !count) {
+		dev_warn(adev->dev, "fail to generate cper records\n");
+		return;
+	}
+
+	/* UEs must be encoded into separate CPER entries */
+	if (type == ACA_SMU_TYPE_UE) {
+		struct aca_banks de_banks;
+
+		aca_banks_init(&de_banks);
+		list_for_each_entry(node, &banks->list, node) {
+			bank = &node->bank;
+			if (bank->aca_err_type == ACA_ERROR_TYPE_DEFERRED) {
+				r = aca_banks_add_bank(&de_banks, bank);
+				if (r)
+					dev_warn(adev->dev, "fail to add de banks, ret = %d\n", r);
+			} else {
+				if (amdgpu_cper_generate_ue_record(adev, bank))
+					dev_warn(adev->dev, "fail to generate ue cper records\n");
+			}
+		}
+
+		if (!list_empty(&de_banks.list)) {
+			if (amdgpu_cper_generate_ce_records(adev, &de_banks, de_banks.nr_banks))
+				dev_warn(adev->dev, "fail to generate de cper records\n");
+		}
+
+		aca_banks_release(&de_banks);
+	} else {
+		/*
+		 * SMU_TYPE_CE banks are combined into 1 CPER entries,
+		 * they could be CEs or DEs or both
+		 */
+		if (amdgpu_cper_generate_ce_records(adev, banks, count))
+			dev_warn(adev->dev, "fail to generate ce cper records\n");
+	}
+}
+
+static int aca_banks_update(struct amdgpu_device *adev, enum aca_smu_type type,
+			    bank_handler_t handler, struct ras_query_context *qctx, void *data)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+	struct aca_banks banks;
+	u32 count = 0;
+	int ret;
+
+	if (list_empty(&aca->mgr.list))
+		return 0;
+
+	if (!aca_bank_should_update(adev, type))
+		return 0;
+
+	ret = aca_smu_get_valid_aca_count(adev, type, &count);
+	if (ret)
+		return ret;
+
+	if (!count)
+		return 0;
+
+	aca_banks_init(&banks);
+
+	ret = aca_smu_get_valid_aca_banks(adev, type, 0, count, &banks, qctx);
+	if (ret)
+		goto err_release_banks;
+
+	if (list_empty(&banks.list)) {
+		ret = 0;
+		goto err_release_banks;
+	}
+
+	ret = aca_dispatch_banks(&aca->mgr, &banks, type,
+				 handler, data);
+	if (ret)
+		goto err_release_banks;
+
+	aca_banks_generate_cper(adev, type, &banks, count);
+
+err_release_banks:
+	aca_banks_release(&banks);
+
+	return ret;
+}
+
+static int aca_log_aca_error_data(struct aca_bank_error *bank_error, enum aca_error_type type, struct ras_err_data *err_data)
+{
+	struct aca_bank_info *info;
+	struct amdgpu_smuio_mcm_config_info mcm_info;
+	u64 count;
+
+	if (type >= ACA_ERROR_TYPE_COUNT)
+		return -EINVAL;
+
+	count = bank_error->count;
+	if (!count)
+		return 0;
+
+	info = &bank_error->info;
+	mcm_info.die_id = info->die_id;
+	mcm_info.socket_id = info->socket_id;
+
+	switch (type) {
+	case ACA_ERROR_TYPE_UE:
+		amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, count);
+		break;
+	case ACA_ERROR_TYPE_CE:
+		amdgpu_ras_error_statistic_ce_count(err_data, &mcm_info, count);
+		break;
+	case ACA_ERROR_TYPE_DEFERRED:
+		amdgpu_ras_error_statistic_de_count(err_data, &mcm_info, count);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int aca_log_aca_error(struct aca_handle *handle, enum aca_error_type type, struct ras_err_data *err_data)
+{
+	struct aca_error_cache *error_cache = &handle->error_cache;
+	struct aca_error *aerr = &error_cache->errors[type];
+	struct aca_bank_error *bank_error, *tmp;
+
+	mutex_lock(&aerr->lock);
+
+	if (list_empty(&aerr->list))
+		goto out_unlock;
+
+	list_for_each_entry_safe(bank_error, tmp, &aerr->list, node) {
+		aca_log_aca_error_data(bank_error, type, err_data);
+		aca_bank_error_remove(aerr, bank_error);
+	}
+
+out_unlock:
+	mutex_unlock(&aerr->lock);
+
+	return 0;
+}
+
+static int __aca_get_error_data(struct amdgpu_device *adev, struct aca_handle *handle, enum aca_error_type type,
+				struct ras_err_data *err_data, struct ras_query_context *qctx)
+{
+	enum aca_smu_type smu_type;
+	int ret;
+
+	switch (type) {
+	case ACA_ERROR_TYPE_UE:
+		smu_type = ACA_SMU_TYPE_UE;
+		break;
+	case ACA_ERROR_TYPE_CE:
+	case ACA_ERROR_TYPE_DEFERRED:
+		smu_type = ACA_SMU_TYPE_CE;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	/* update aca bank to aca source error_cache first */
+	ret = aca_banks_update(adev, smu_type, handler_aca_log_bank_error, qctx, NULL);
+	if (ret)
+		return ret;
+
+	/* DEs may contain in CEs or UEs */
+	if (type != ACA_ERROR_TYPE_DEFERRED)
+		aca_log_aca_error(handle, ACA_ERROR_TYPE_DEFERRED, err_data);
+
+	return aca_log_aca_error(handle, type, err_data);
+}
+
+static bool aca_handle_is_valid(struct aca_handle *handle)
+{
+	if (!handle->mask || !list_empty(&handle->node))
+		return false;
+
+	return true;
+}
+
+int amdgpu_aca_get_error_data(struct amdgpu_device *adev, struct aca_handle *handle,
+			      enum aca_error_type type, struct ras_err_data *err_data,
+			      struct ras_query_context *qctx)
+{
+	if (!handle || !err_data)
+		return -EINVAL;
+
+	if (aca_handle_is_valid(handle))
+		return -EOPNOTSUPP;
+
+	if ((type < 0) || (!(BIT(type) & handle->mask)))
+		return  0;
+
+	return __aca_get_error_data(adev, handle, type, err_data, qctx);
+}
+
+static void aca_error_init(struct aca_error *aerr, enum aca_error_type type)
+{
+	mutex_init(&aerr->lock);
+	INIT_LIST_HEAD(&aerr->list);
+	aerr->type = type;
+	aerr->nr_errors = 0;
+}
+
+static void aca_init_error_cache(struct aca_handle *handle)
+{
+	struct aca_error_cache *error_cache = &handle->error_cache;
+	int type;
+
+	for (type = ACA_ERROR_TYPE_UE; type < ACA_ERROR_TYPE_COUNT; type++)
+		aca_error_init(&error_cache->errors[type], type);
+}
+
+static void aca_error_fini(struct aca_error *aerr)
+{
+	struct aca_bank_error *bank_error, *tmp;
+
+	mutex_lock(&aerr->lock);
+	if (list_empty(&aerr->list))
+		goto out_unlock;
+
+	list_for_each_entry_safe(bank_error, tmp, &aerr->list, node)
+		aca_bank_error_remove(aerr, bank_error);
+
+out_unlock:
+	mutex_destroy(&aerr->lock);
+}
+
+static void aca_fini_error_cache(struct aca_handle *handle)
+{
+	struct aca_error_cache *error_cache = &handle->error_cache;
+	int type;
+
+	for (type = ACA_ERROR_TYPE_UE; type < ACA_ERROR_TYPE_COUNT; type++)
+		aca_error_fini(&error_cache->errors[type]);
+}
+
+static int add_aca_handle(struct amdgpu_device *adev, struct aca_handle_manager *mgr, struct aca_handle *handle,
+			  const char *name, const struct aca_info *ras_info, void *data)
+{
+	memset(handle, 0, sizeof(*handle));
+
+	handle->adev = adev;
+	handle->mgr = mgr;
+	handle->name = name;
+	handle->hwip = ras_info->hwip;
+	handle->mask = ras_info->mask;
+	handle->bank_ops = ras_info->bank_ops;
+	handle->data = data;
+	aca_init_error_cache(handle);
+
+	INIT_LIST_HEAD(&handle->node);
+	list_add_tail(&handle->node, &mgr->list);
+	mgr->nr_handles++;
+
+	return 0;
+}
+
+static ssize_t aca_sysfs_read(struct device *dev,
+			      struct device_attribute *attr, char *buf)
+{
+	struct aca_handle *handle = container_of(attr, struct aca_handle, aca_attr);
+
+	/* NOTE: the aca cache will be auto cleared once read,
+	 * So the driver should unify the query entry point, forward request to ras query interface directly */
+	return amdgpu_ras_aca_sysfs_read(dev, attr, handle, buf, handle->data);
+}
+
+static int add_aca_sysfs(struct amdgpu_device *adev, struct aca_handle *handle)
+{
+	struct device_attribute *aca_attr = &handle->aca_attr;
+
+	snprintf(handle->attr_name, sizeof(handle->attr_name) - 1, "aca_%s", handle->name);
+	aca_attr->show = aca_sysfs_read;
+	aca_attr->attr.name = handle->attr_name;
+	aca_attr->attr.mode = S_IRUGO;
+	sysfs_attr_init(&aca_attr->attr);
+
+	return sysfs_add_file_to_group(&adev->dev->kobj,
+				       &aca_attr->attr,
+				       "ras");
+}
+
+int amdgpu_aca_add_handle(struct amdgpu_device *adev, struct aca_handle *handle,
+			  const char *name, const struct aca_info *ras_info, void *data)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+	int ret;
+
+	if (!amdgpu_aca_is_enabled(adev))
+		return 0;
+
+	ret = add_aca_handle(adev, &aca->mgr, handle, name, ras_info, data);
+	if (ret)
+		return ret;
+
+	return add_aca_sysfs(adev, handle);
+}
+
+static void remove_aca_handle(struct aca_handle *handle)
+{
+	struct aca_handle_manager *mgr = handle->mgr;
+
+	aca_fini_error_cache(handle);
+	list_del(&handle->node);
+	mgr->nr_handles--;
+}
+
+static void remove_aca_sysfs(struct aca_handle *handle)
+{
+	struct amdgpu_device *adev = handle->adev;
+	struct device_attribute *aca_attr = &handle->aca_attr;
+
+	if (adev->dev->kobj.sd)
+		sysfs_remove_file_from_group(&adev->dev->kobj,
+					     &aca_attr->attr,
+					     "ras");
+}
+
+void amdgpu_aca_remove_handle(struct aca_handle *handle)
+{
+	if (!handle || list_empty(&handle->node))
+		return;
+
+	remove_aca_sysfs(handle);
+	remove_aca_handle(handle);
+}
+
+static int aca_manager_init(struct aca_handle_manager *mgr)
+{
+	INIT_LIST_HEAD(&mgr->list);
+	mgr->nr_handles = 0;
+
+	return 0;
+}
+
+static void aca_manager_fini(struct aca_handle_manager *mgr)
+{
+	struct aca_handle *handle, *tmp;
+
+	if (list_empty(&mgr->list))
+		return;
+
+	list_for_each_entry_safe(handle, tmp, &mgr->list, node)
+		amdgpu_aca_remove_handle(handle);
+}
+
+bool amdgpu_aca_is_enabled(struct amdgpu_device *adev)
+{
+	return (adev->aca.is_enabled ||
+		adev->debug_enable_ras_aca);
+}
+
+int amdgpu_aca_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+	int ret;
+
+	atomic_set(&aca->ue_update_flag, 0);
+
+	ret = aca_manager_init(&aca->mgr);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+void amdgpu_aca_fini(struct amdgpu_device *adev)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+
+	aca_manager_fini(&aca->mgr);
+
+	atomic_set(&aca->ue_update_flag, 0);
+}
+
+int amdgpu_aca_reset(struct amdgpu_device *adev)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+
+	atomic_set(&aca->ue_update_flag, 0);
+
+	return 0;
+}
+
+void amdgpu_aca_set_smu_funcs(struct amdgpu_device *adev, const struct aca_smu_funcs *smu_funcs)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+
+	WARN_ON(aca->smu_funcs);
+	aca->smu_funcs = smu_funcs;
+}
+
+int aca_bank_info_decode(struct aca_bank *bank, struct aca_bank_info *info)
+{
+	u64 ipid;
+	u32 instidhi, instidlo;
+
+	if (!bank || !info)
+		return -EINVAL;
+
+	ipid = bank->regs[ACA_REG_IDX_IPID];
+	info->hwid = ACA_REG__IPID__HARDWAREID(ipid);
+	info->mcatype = ACA_REG__IPID__MCATYPE(ipid);
+	/*
+	 * Unfied DieID Format: SAASS. A:AID, S:Socket.
+	 * Unfied DieID[4:4] = InstanceId[0:0]
+	 * Unfied DieID[0:3] = InstanceIdHi[0:3]
+	 */
+	instidhi = ACA_REG__IPID__INSTANCEIDHI(ipid);
+	instidlo = ACA_REG__IPID__INSTANCEIDLO(ipid);
+	info->die_id = ((instidhi >> 2) & 0x03);
+	info->socket_id = ((instidlo & 0x1) << 2) | (instidhi & 0x03);
+
+	return 0;
+}
+
+static int aca_bank_get_error_code(struct amdgpu_device *adev, struct aca_bank *bank)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+	const struct aca_smu_funcs *smu_funcs = aca->smu_funcs;
+
+	if (!smu_funcs || !smu_funcs->parse_error_code)
+		return -EOPNOTSUPP;
+
+	return smu_funcs->parse_error_code(adev, bank);
+}
+
+int aca_bank_check_error_codes(struct amdgpu_device *adev, struct aca_bank *bank, int *err_codes, int size)
+{
+	int i, error_code;
+
+	if (!bank || !err_codes)
+		return -EINVAL;
+
+	error_code = aca_bank_get_error_code(adev, bank);
+	if (error_code < 0)
+		return error_code;
+
+	for (i = 0; i < size; i++) {
+		if (err_codes[i] == error_code)
+			return 0;
+	}
+
+	return -EINVAL;
+}
+
+int amdgpu_aca_smu_set_debug_mode(struct amdgpu_device *adev, bool en)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+	const struct aca_smu_funcs *smu_funcs = aca->smu_funcs;
+
+	if (!smu_funcs || !smu_funcs->set_debug_mode)
+		return -EOPNOTSUPP;
+
+	return smu_funcs->set_debug_mode(adev, en);
+}
+
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_aca_smu_debug_mode_set(void *data, u64 val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	int ret;
+
+	ret = amdgpu_ras_set_aca_debug_mode(adev, val ? true : false);
+	if (ret)
+		return ret;
+
+	dev_info(adev->dev, "amdgpu set smu aca debug mode %s success\n", val ? "on" : "off");
+
+	return 0;
+}
+
+static void aca_dump_entry(struct seq_file *m, struct aca_bank *bank, enum aca_smu_type type, int idx)
+{
+	struct aca_bank_info info;
+	int i, ret;
+
+	ret = aca_bank_info_decode(bank, &info);
+	if (ret)
+		return;
+
+	seq_printf(m, "aca entry[%d].type: %s\n", idx, type ==  ACA_SMU_TYPE_UE ? "UE" : "CE");
+	seq_printf(m, "aca entry[%d].info: socketid:%d aid:%d hwid:0x%03x mcatype:0x%04x\n",
+		   idx, info.socket_id, info.die_id, info.hwid, info.mcatype);
+
+	for (i = 0; i < ARRAY_SIZE(aca_regs); i++)
+		seq_printf(m, "aca entry[%d].regs[%d]: 0x%016llx\n", idx, aca_regs[i].reg_idx, bank->regs[aca_regs[i].reg_idx]);
+}
+
+struct aca_dump_context {
+	struct seq_file *m;
+	int idx;
+};
+
+static int handler_aca_bank_dump(struct aca_handle *handle, struct aca_bank *bank,
+				 enum aca_smu_type type, void *data)
+{
+	struct aca_dump_context *ctx = (struct aca_dump_context *)data;
+
+	aca_dump_entry(ctx->m, bank, type, ctx->idx++);
+
+	return handler_aca_log_bank_error(handle, bank, type, NULL);
+}
+
+static int aca_dump_show(struct seq_file *m, enum aca_smu_type type)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct aca_dump_context context = {
+		.m = m,
+		.idx = 0,
+	};
+
+	return aca_banks_update(adev, type, handler_aca_bank_dump, NULL, (void *)&context);
+}
+
+static int aca_dump_ce_show(struct seq_file *m, void *unused)
+{
+	return aca_dump_show(m, ACA_SMU_TYPE_CE);
+}
+
+static int aca_dump_ce_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, aca_dump_ce_show, inode->i_private);
+}
+
+static const struct file_operations aca_ce_dump_debug_fops = {
+	.owner = THIS_MODULE,
+	.open = aca_dump_ce_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = single_release,
+};
+
+static int aca_dump_ue_show(struct seq_file *m, void *unused)
+{
+	return aca_dump_show(m, ACA_SMU_TYPE_UE);
+}
+
+static int aca_dump_ue_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, aca_dump_ue_show, inode->i_private);
+}
+
+static const struct file_operations aca_ue_dump_debug_fops = {
+	.owner = THIS_MODULE,
+	.open = aca_dump_ue_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = single_release,
+};
+
+DEFINE_DEBUGFS_ATTRIBUTE(aca_debug_mode_fops, NULL, amdgpu_aca_smu_debug_mode_set, "%llu\n");
+#endif
+
+void amdgpu_aca_smu_debugfs_init(struct amdgpu_device *adev, struct dentry *root)
+{
+#if defined(CONFIG_DEBUG_FS)
+	if (!root)
+		return;
+
+	debugfs_create_file("aca_debug_mode", 0200, root, adev, &aca_debug_mode_fops);
+	debugfs_create_file("aca_ue_dump", 0400, root, adev, &aca_ue_dump_debug_fops);
+	debugfs_create_file("aca_ce_dump", 0400, root, adev, &aca_ce_dump_debug_fops);
+#endif
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h
new file mode 100644
index 000000000000..38c88897e1ec
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h
@@ -0,0 +1,232 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __AMDGPU_ACA_H__
+#define __AMDGPU_ACA_H__
+
+#include <linux/list.h>
+
+struct ras_err_data;
+struct ras_query_context;
+
+#define ACA_MAX_REGS_COUNT	(16)
+
+#define ACA_REG_FIELD(x, h, l)			(((x) & GENMASK_ULL(h, l)) >> l)
+#define ACA_REG__STATUS__VAL(x)			ACA_REG_FIELD(x, 63, 63)
+#define ACA_REG__STATUS__OVERFLOW(x)		ACA_REG_FIELD(x, 62, 62)
+#define ACA_REG__STATUS__UC(x)			ACA_REG_FIELD(x, 61, 61)
+#define ACA_REG__STATUS__EN(x)			ACA_REG_FIELD(x, 60, 60)
+#define ACA_REG__STATUS__MISCV(x)		ACA_REG_FIELD(x, 59, 59)
+#define ACA_REG__STATUS__ADDRV(x)		ACA_REG_FIELD(x, 58, 58)
+#define ACA_REG__STATUS__PCC(x)			ACA_REG_FIELD(x, 57, 57)
+#define ACA_REG__STATUS__ERRCOREIDVAL(x)	ACA_REG_FIELD(x, 56, 56)
+#define ACA_REG__STATUS__TCC(x)			ACA_REG_FIELD(x, 55, 55)
+#define ACA_REG__STATUS__SYNDV(x)		ACA_REG_FIELD(x, 53, 53)
+#define ACA_REG__STATUS__CECC(x)		ACA_REG_FIELD(x, 46, 46)
+#define ACA_REG__STATUS__UECC(x)		ACA_REG_FIELD(x, 45, 45)
+#define ACA_REG__STATUS__DEFERRED(x)		ACA_REG_FIELD(x, 44, 44)
+#define ACA_REG__STATUS__POISON(x)		ACA_REG_FIELD(x, 43, 43)
+#define ACA_REG__STATUS__SCRUB(x)		ACA_REG_FIELD(x, 40, 40)
+#define ACA_REG__STATUS__ERRCOREID(x)		ACA_REG_FIELD(x, 37, 32)
+#define ACA_REG__STATUS__ADDRLSB(x)		ACA_REG_FIELD(x, 29, 24)
+#define ACA_REG__STATUS__ERRORCODEEXT(x)	ACA_REG_FIELD(x, 21, 16)
+#define ACA_REG__STATUS__ERRORCODE(x)		ACA_REG_FIELD(x, 15, 0)
+
+#define ACA_REG__IPID__MCATYPE(x)		ACA_REG_FIELD(x, 63, 48)
+#define ACA_REG__IPID__INSTANCEIDHI(x)		ACA_REG_FIELD(x, 47, 44)
+#define ACA_REG__IPID__HARDWAREID(x)		ACA_REG_FIELD(x, 43, 32)
+#define ACA_REG__IPID__INSTANCEIDLO(x)		ACA_REG_FIELD(x, 31, 0)
+
+#define ACA_REG__MISC0__VALID(x)		ACA_REG_FIELD(x, 63, 63)
+#define ACA_REG__MISC0__OVRFLW(x)		ACA_REG_FIELD(x, 48, 48)
+#define ACA_REG__MISC0__ERRCNT(x)		ACA_REG_FIELD(x, 43, 32)
+
+#define ACA_REG__SYND__ERRORINFORMATION(x)	ACA_REG_FIELD(x, 17, 0)
+
+/* NOTE: The following codes refers to the smu header file */
+#define ACA_EXTERROR_CODE_CE			0x3a
+#define ACA_EXTERROR_CODE_FAULT			0x3b
+
+#define ACA_ERROR_UE_MASK		BIT_MASK(ACA_ERROR_TYPE_UE)
+#define ACA_ERROR_CE_MASK		BIT_MASK(ACA_ERROR_TYPE_CE)
+#define ACA_ERROR_DEFERRED_MASK		BIT_MASK(ACA_ERROR_TYPE_DEFERRED)
+
+#define mmSMNAID_AID0_MCA_SMU		0x03b30400	/* SMN AID AID0 */
+#define mmSMNAID_XCD0_MCA_SMU		0x36430400	/* SMN AID XCD0 */
+#define mmSMNAID_XCD1_MCA_SMU		0x38430400	/* SMN AID XCD1 */
+#define mmSMNXCD_XCD0_MCA_SMU		0x40430400	/* SMN XCD XCD0 */
+
+#define ACA_BANK_ERR_IS_DEFFERED(bank)                                \
+	(ACA_REG__STATUS__POISON((bank)->regs[ACA_REG_IDX_STATUS]) || \
+	 ACA_REG__STATUS__DEFERRED((bank)->regs[ACA_REG_IDX_STATUS]))
+
+enum aca_reg_idx {
+	ACA_REG_IDX_CTL			= 0,
+	ACA_REG_IDX_STATUS		= 1,
+	ACA_REG_IDX_ADDR		= 2,
+	ACA_REG_IDX_MISC0		= 3,
+	ACA_REG_IDX_CONFIG		= 4,
+	ACA_REG_IDX_IPID		= 5,
+	ACA_REG_IDX_SYND		= 6,
+	ACA_REG_IDX_DESTAT		= 8,
+	ACA_REG_IDX_DEADDR		= 9,
+	ACA_REG_IDX_CTL_MASK		= 10,
+	ACA_REG_IDX_COUNT		= 16,
+};
+
+enum aca_hwip_type {
+	ACA_HWIP_TYPE_UNKNOW = -1,
+	ACA_HWIP_TYPE_PSP = 0,
+	ACA_HWIP_TYPE_UMC,
+	ACA_HWIP_TYPE_SMU,
+	ACA_HWIP_TYPE_PCS_XGMI,
+	ACA_HWIP_TYPE_COUNT,
+};
+
+enum aca_error_type {
+	ACA_ERROR_TYPE_INVALID = -1,
+	ACA_ERROR_TYPE_UE = 0,
+	ACA_ERROR_TYPE_CE,
+	ACA_ERROR_TYPE_DEFERRED,
+	ACA_ERROR_TYPE_COUNT
+};
+
+enum aca_smu_type {
+	ACA_SMU_TYPE_INVALID = -1,
+	ACA_SMU_TYPE_UE = 0,
+	ACA_SMU_TYPE_CE,
+	ACA_SMU_TYPE_COUNT,
+};
+
+struct aca_hwip {
+	int hwid;
+	int mcatype;
+};
+
+struct aca_bank {
+	enum aca_error_type aca_err_type;
+	enum aca_smu_type smu_err_type;
+	u64 regs[ACA_MAX_REGS_COUNT];
+};
+
+struct aca_bank_node {
+	struct aca_bank bank;
+	struct list_head node;
+};
+
+struct aca_banks {
+	int nr_banks;
+	struct list_head list;
+};
+
+struct aca_bank_info {
+	int die_id;
+	int socket_id;
+	int hwid;
+	int mcatype;
+};
+
+struct aca_bank_error {
+	struct list_head node;
+	struct aca_bank_info info;
+	u64 count;
+};
+
+struct aca_error {
+	struct list_head list;
+	struct mutex lock;
+	enum aca_error_type type;
+	int nr_errors;
+};
+
+struct aca_handle_manager {
+	struct list_head list;
+	int nr_handles;
+};
+
+struct aca_error_cache {
+	struct aca_error errors[ACA_ERROR_TYPE_COUNT];
+};
+
+struct aca_handle {
+	struct list_head node;
+	enum aca_hwip_type hwip;
+	struct amdgpu_device *adev;
+	struct aca_handle_manager *mgr;
+	struct aca_error_cache error_cache;
+	const struct aca_bank_ops *bank_ops;
+	struct device_attribute aca_attr;
+	char attr_name[64];
+	const char *name;
+	u32 mask;
+	void *data;
+};
+
+struct aca_bank_ops {
+	int (*aca_bank_parser)(struct aca_handle *handle, struct aca_bank *bank, enum aca_smu_type type, void *data);
+	bool (*aca_bank_is_valid)(struct aca_handle *handle, struct aca_bank *bank, enum aca_smu_type type,
+				  void *data);
+};
+
+struct aca_smu_funcs {
+	int max_ue_bank_count;
+	int max_ce_bank_count;
+	int (*set_debug_mode)(struct amdgpu_device *adev, bool enable);
+	int (*get_valid_aca_count)(struct amdgpu_device *adev, enum aca_smu_type type, u32 *count);
+	int (*get_valid_aca_bank)(struct amdgpu_device *adev, enum aca_smu_type type, int idx, struct aca_bank *bank);
+	int (*parse_error_code)(struct amdgpu_device *adev, struct aca_bank *bank);
+};
+
+struct amdgpu_aca {
+	struct aca_handle_manager mgr;
+	const struct aca_smu_funcs *smu_funcs;
+	atomic_t ue_update_flag;
+	bool is_enabled;
+};
+
+struct aca_info {
+	enum aca_hwip_type hwip;
+	const struct aca_bank_ops *bank_ops;
+	u32 mask;
+};
+
+int amdgpu_aca_init(struct amdgpu_device *adev);
+void amdgpu_aca_fini(struct amdgpu_device *adev);
+int amdgpu_aca_reset(struct amdgpu_device *adev);
+void amdgpu_aca_set_smu_funcs(struct amdgpu_device *adev, const struct aca_smu_funcs *smu_funcs);
+bool amdgpu_aca_is_enabled(struct amdgpu_device *adev);
+
+int aca_bank_info_decode(struct aca_bank *bank, struct aca_bank_info *info);
+int aca_bank_check_error_codes(struct amdgpu_device *adev, struct aca_bank *bank, int *err_codes, int size);
+
+int amdgpu_aca_add_handle(struct amdgpu_device *adev, struct aca_handle *handle,
+			  const char *name, const struct aca_info *aca_info, void *data);
+void amdgpu_aca_remove_handle(struct aca_handle *handle);
+int amdgpu_aca_get_error_data(struct amdgpu_device *adev, struct aca_handle *handle,
+			      enum aca_error_type type, struct ras_err_data *err_data,
+			      struct ras_query_context *qctx);
+int amdgpu_aca_smu_set_debug_mode(struct amdgpu_device *adev, bool en);
+void amdgpu_aca_smu_debugfs_init(struct amdgpu_device *adev, struct dentry *root);
+int aca_error_cache_log_bank_error(struct aca_handle *handle, struct aca_bank_info *info,
+				   enum aca_error_type type, u64 count);
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c
index 6d72355ac492..381ef205b0df 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c
@@ -98,9 +98,9 @@ enum {
 	ACP_TILE_DSP2,
 };
 
-static int acp_sw_init(void *handle)
+static int acp_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->acp.parent = adev->dev;
 
@@ -112,9 +112,9 @@ static int acp_sw_init(void *handle)
 	return 0;
 }
 
-static int acp_sw_fini(void *handle)
+static int acp_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (adev->acp.cgs_device)
 		amdgpu_cgs_destroy_device(adev->acp.cgs_device);
@@ -140,7 +140,7 @@ static int acp_poweroff(struct generic_pm_domain *genpd)
 	 * 2. power off the acp tiles
 	 * 3. check and enter ulv state
 	 */
-	amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, true);
+	amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, true, 0);
 	return 0;
 }
 
@@ -157,7 +157,7 @@ static int acp_poweron(struct generic_pm_domain *genpd)
 	 * 2. turn on acp clock
 	 * 3. power on acp tiles
 	 */
-	amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, false);
+	amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, false, 0);
 	return 0;
 }
 
@@ -219,10 +219,10 @@ static const struct dmi_system_id acp_quirk_table[] = {
 /**
  * acp_hw_init - start and test ACP block
  *
- * @handle: handle used to pass amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  */
-static int acp_hw_init(void *handle)
+static int acp_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
 	u64 acp_base;
@@ -230,19 +230,13 @@ static int acp_hw_init(void *handle)
 	u32 count = 0;
 	struct i2s_platform_data *i2s_pdata = NULL;
 
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	const struct amdgpu_ip_block *ip_block =
-		amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_ACP);
-
-	if (!ip_block)
-		return -EINVAL;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = amd_acp_hw_init(adev->acp.cgs_device,
 			    ip_block->version->major, ip_block->version->minor);
 	/* -ENODEV means board uses AZ rather than ACP */
 	if (r == -ENODEV) {
-		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, true);
+		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, true, 0);
 		return 0;
 	} else if (r) {
 		return r;
@@ -308,17 +302,19 @@ static int acp_hw_init(void *handle)
 		adev->acp.acp_res[2].end = adev->acp.acp_res[2].start;
 
 		adev->acp.acp_cell[0].name = "acp_audio_dma";
+		adev->acp.acp_cell[0].id = 0;
 		adev->acp.acp_cell[0].num_resources = 3;
 		adev->acp.acp_cell[0].resources = &adev->acp.acp_res[0];
 		adev->acp.acp_cell[0].platform_data = &adev->asic_type;
 		adev->acp.acp_cell[0].pdata_size = sizeof(adev->asic_type);
 
 		adev->acp.acp_cell[1].name = "designware-i2s";
+		adev->acp.acp_cell[1].id = 1;
 		adev->acp.acp_cell[1].num_resources = 1;
 		adev->acp.acp_cell[1].resources = &adev->acp.acp_res[1];
 		adev->acp.acp_cell[1].platform_data = &i2s_pdata[0];
 		adev->acp.acp_cell[1].pdata_size = sizeof(struct i2s_platform_data);
-		r = mfd_add_hotplug_devices(adev->acp.parent, adev->acp.acp_cell, 2);
+		r = mfd_add_devices(adev->acp.parent, 0, adev->acp.acp_cell, 2, NULL, 0, NULL);
 		if (r)
 			goto failure;
 		r = device_for_each_child(adev->acp.parent, &adev->acp.acp_genpd->gpd,
@@ -416,30 +412,34 @@ static int acp_hw_init(void *handle)
 		adev->acp.acp_res[4].end = adev->acp.acp_res[4].start;
 
 		adev->acp.acp_cell[0].name = "acp_audio_dma";
+		adev->acp.acp_cell[0].id = 0;
 		adev->acp.acp_cell[0].num_resources = 5;
 		adev->acp.acp_cell[0].resources = &adev->acp.acp_res[0];
 		adev->acp.acp_cell[0].platform_data = &adev->asic_type;
 		adev->acp.acp_cell[0].pdata_size = sizeof(adev->asic_type);
 
 		adev->acp.acp_cell[1].name = "designware-i2s";
+		adev->acp.acp_cell[1].id = 1;
 		adev->acp.acp_cell[1].num_resources = 1;
 		adev->acp.acp_cell[1].resources = &adev->acp.acp_res[1];
 		adev->acp.acp_cell[1].platform_data = &i2s_pdata[0];
 		adev->acp.acp_cell[1].pdata_size = sizeof(struct i2s_platform_data);
 
 		adev->acp.acp_cell[2].name = "designware-i2s";
+		adev->acp.acp_cell[2].id = 2;
 		adev->acp.acp_cell[2].num_resources = 1;
 		adev->acp.acp_cell[2].resources = &adev->acp.acp_res[2];
 		adev->acp.acp_cell[2].platform_data = &i2s_pdata[1];
 		adev->acp.acp_cell[2].pdata_size = sizeof(struct i2s_platform_data);
 
 		adev->acp.acp_cell[3].name = "designware-i2s";
+		adev->acp.acp_cell[3].id = 3;
 		adev->acp.acp_cell[3].num_resources = 1;
 		adev->acp.acp_cell[3].resources = &adev->acp.acp_res[3];
 		adev->acp.acp_cell[3].platform_data = &i2s_pdata[2];
 		adev->acp.acp_cell[3].pdata_size = sizeof(struct i2s_platform_data);
 
-		r = mfd_add_hotplug_devices(adev->acp.parent, adev->acp.acp_cell, ACP_DEVS);
+		r = mfd_add_devices(adev->acp.parent, 0, adev->acp.acp_cell, ACP_DEVS, NULL, 0, NULL);
 		if (r)
 			goto failure;
 
@@ -503,18 +503,18 @@ failure:
 /**
  * acp_hw_fini - stop the hardware block
  *
- * @handle: handle used to pass amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  */
-static int acp_hw_fini(void *handle)
+static int acp_hw_fini(struct amdgpu_ip_block *ip_block)
 {
 	u32 val = 0;
 	u32 count = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* return early if no ACP */
 	if (!adev->acp.acp_genpd) {
-		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, false);
+		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, false, 0);
 		return 0;
 	}
 
@@ -565,67 +565,50 @@ static int acp_hw_fini(void *handle)
 	return 0;
 }
 
-static int acp_suspend(void *handle)
+static int acp_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* power up on suspend */
 	if (!adev->acp.acp_cell)
-		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, false);
+		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, false, 0);
 	return 0;
 }
 
-static int acp_resume(void *handle)
+static int acp_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* power down again on resume */
 	if (!adev->acp.acp_cell)
-		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, true);
-	return 0;
-}
-
-static int acp_early_init(void *handle)
-{
+		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, true, 0);
 	return 0;
 }
 
-static bool acp_is_idle(void *handle)
+static bool acp_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	return true;
 }
 
-static int acp_wait_for_idle(void *handle)
-{
-	return 0;
-}
-
-static int acp_soft_reset(void *handle)
-{
-	return 0;
-}
-
-static int acp_set_clockgating_state(void *handle,
+static int acp_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 				     enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int acp_set_powergating_state(void *handle,
+static int acp_set_powergating_state(struct amdgpu_ip_block *ip_block,
 				     enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_PG_STATE_GATE);
 
-	amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, enable);
+	amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, enable, 0);
 
 	return 0;
 }
 
 static const struct amd_ip_funcs acp_ip_funcs = {
 	.name = "acp_ip",
-	.early_init = acp_early_init,
-	.late_init = NULL,
 	.sw_init = acp_sw_init,
 	.sw_fini = acp_sw_fini,
 	.hw_init = acp_hw_init,
@@ -633,8 +616,6 @@ static const struct amd_ip_funcs acp_ip_funcs = {
 	.suspend = acp_suspend,
 	.resume = acp_resume,
 	.is_idle = acp_is_idle,
-	.wait_for_idle = acp_wait_for_idle,
-	.soft_reset = acp_soft_reset,
 	.set_clockgating_state = acp_set_clockgating_state,
 	.set_powergating_state = acp_set_powergating_state,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c
index aeeec211861c..d31460a9e958 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c
@@ -26,6 +26,7 @@
 #include <linux/acpi.h>
 #include <linux/backlight.h>
 #include <linux/slab.h>
+#include <linux/xarray.h>
 #include <linux/power_supply.h>
 #include <linux/pm_runtime.h>
 #include <linux/suspend.h>
@@ -38,6 +39,45 @@
 #include "amd_acpi.h"
 #include "atom.h"
 
+/* Declare GUID for AMD _DSM method for XCCs */
+static const guid_t amd_xcc_dsm_guid = GUID_INIT(0x8267f5d5, 0xa556, 0x44f2,
+						 0xb8, 0xb4, 0x45, 0x56, 0x2e,
+						 0x8c, 0x5b, 0xec);
+
+#define AMD_XCC_HID_START 3000
+#define AMD_XCC_DSM_GET_NUM_FUNCS 0
+#define AMD_XCC_DSM_GET_SUPP_MODE 1
+#define AMD_XCC_DSM_GET_XCP_MODE 2
+#define AMD_XCC_DSM_GET_VF_XCC_MAPPING 4
+#define AMD_XCC_DSM_GET_TMR_INFO 5
+#define AMD_XCC_DSM_NUM_FUNCS 5
+
+#define AMD_XCC_MAX_HID 24
+
+struct xarray numa_info_xa;
+
+/* Encapsulates the XCD acpi object information */
+struct amdgpu_acpi_xcc_info {
+	struct list_head list;
+	struct amdgpu_numa_info *numa_info;
+	uint8_t xcp_node;
+	uint8_t phy_id;
+	acpi_handle handle;
+};
+
+struct amdgpu_acpi_dev_info {
+	struct list_head list;
+	struct list_head xcc_list;
+	uint32_t sbdf;
+	uint16_t supp_xcp_mode;
+	uint16_t xcp_mode;
+	uint16_t mem_mode;
+	uint64_t tmr_base;
+	uint64_t tmr_size;
+};
+
+struct list_head amdgpu_acpi_dev_list;
+
 struct amdgpu_atif_notification_cfg {
 	bool enabled;
 	int command_code;
@@ -107,6 +147,7 @@ static union acpi_object *amdgpu_atif_call(struct amdgpu_atif *atif,
 					   struct acpi_buffer *params)
 {
 	acpi_status status;
+	union acpi_object *obj;
 	union acpi_object atif_arg_elements[2];
 	struct acpi_object_list atif_arg;
 	struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL };
@@ -129,16 +170,24 @@ static union acpi_object *amdgpu_atif_call(struct amdgpu_atif *atif,
 
 	status = acpi_evaluate_object(atif->handle, NULL, &atif_arg,
 				      &buffer);
+	obj = (union acpi_object *)buffer.pointer;
 
-	/* Fail only if calling the method fails and ATIF is supported */
-	if (ACPI_FAILURE(status) && status != AE_NOT_FOUND) {
+	/* Fail if calling the method fails */
+	if (ACPI_FAILURE(status)) {
 		DRM_DEBUG_DRIVER("failed to evaluate ATIF got %s\n",
 				 acpi_format_exception(status));
-		kfree(buffer.pointer);
+		kfree(obj);
 		return NULL;
 	}
 
-	return buffer.pointer;
+	if (obj->type != ACPI_TYPE_BUFFER) {
+		DRM_DEBUG_DRIVER("bad object returned from ATIF: %d\n",
+				 obj->type);
+		kfree(obj);
+		return NULL;
+	}
+
+	return obj;
 }
 
 /**
@@ -343,6 +392,12 @@ static int amdgpu_atif_query_backlight_caps(struct amdgpu_atif *atif)
 			characteristics.min_input_signal;
 	atif->backlight_caps.max_input_signal =
 			characteristics.max_input_signal;
+	atif->backlight_caps.ac_level = characteristics.ac_level;
+	atif->backlight_caps.dc_level = characteristics.dc_level;
+	atif->backlight_caps.data_points = characteristics.number_of_points;
+	memcpy(atif->backlight_caps.luminance_data,
+	       characteristics.data_points,
+	       sizeof(atif->backlight_caps.luminance_data));
 out:
 	kfree(info);
 	return err;
@@ -452,7 +507,6 @@ static int amdgpu_atif_handler(struct amdgpu_device *adev,
 				pm_runtime_get_sync(adev_to_drm(adev)->dev);
 				/* Just fire off a uevent and let userspace tell us what to do */
 				drm_helper_hpd_irq_event(adev_to_drm(adev));
-				pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 				pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 			}
 		}
@@ -666,7 +720,7 @@ int amdgpu_acpi_pcie_performance_request(struct amdgpu_device *adev,
 
 	atcs_input.size = sizeof(struct atcs_pref_req_input);
 	/* client id (bit 2-0: func num, 7-3: dev num, 15-8: bus num) */
-	atcs_input.client_id = adev->pdev->devfn | (adev->pdev->bus->number << 8);
+	atcs_input.client_id = pci_dev_id(adev->pdev);
 	atcs_input.valid_flags_mask = ATCS_VALID_FLAGS_MASK;
 	atcs_input.flags = ATCS_WAIT_FOR_COMPLETION;
 	if (advertise)
@@ -736,7 +790,7 @@ int amdgpu_acpi_power_shift_control(struct amdgpu_device *adev,
 
 	atcs_input.size = sizeof(struct atcs_pwr_shift_input);
 	/* dGPU id (bit 2-0: func num, 7-3: dev num, 15-8: bus num) */
-	atcs_input.dgpu_id = adev->pdev->devfn | (adev->pdev->bus->number << 8);
+	atcs_input.dgpu_id = pci_dev_id(adev->pdev);
 	atcs_input.dev_acpi_state = dev_state;
 	atcs_input.drv_state = drv_state;
 
@@ -749,24 +803,25 @@ int amdgpu_acpi_power_shift_control(struct amdgpu_device *adev,
 		return -EIO;
 	}
 
+	kfree(info);
 	return 0;
 }
 
 /**
  * amdgpu_acpi_smart_shift_update - update dGPU device state to SBIOS
  *
- * @dev: drm_device pointer
+ * @adev: amdgpu device pointer
  * @ss_state: current smart shift event
  *
  * returns 0 on success,
  * otherwise return error number.
  */
-int amdgpu_acpi_smart_shift_update(struct drm_device *dev, enum amdgpu_ss ss_state)
+int amdgpu_acpi_smart_shift_update(struct amdgpu_device *adev,
+				   enum amdgpu_ss ss_state)
 {
-	struct amdgpu_device *adev = drm_to_adev(dev);
 	int r;
 
-	if (!amdgpu_device_supports_smart_shift(dev))
+	if (!amdgpu_device_supports_smart_shift(adev))
 		return 0;
 
 	switch (ss_state) {
@@ -801,6 +856,347 @@ int amdgpu_acpi_smart_shift_update(struct drm_device *dev, enum amdgpu_ss ss_sta
 	return r;
 }
 
+#ifdef CONFIG_ACPI_NUMA
+static inline uint64_t amdgpu_acpi_get_numa_size(int nid)
+{
+	/* This is directly using si_meminfo_node implementation as the
+	 * function is not exported.
+	 */
+	int zone_type;
+	uint64_t managed_pages = 0;
+
+	pg_data_t *pgdat = NODE_DATA(nid);
+
+	for (zone_type = 0; zone_type < MAX_NR_ZONES; zone_type++)
+		managed_pages +=
+			zone_managed_pages(&pgdat->node_zones[zone_type]);
+	return managed_pages * PAGE_SIZE;
+}
+
+static struct amdgpu_numa_info *amdgpu_acpi_get_numa_info(uint32_t pxm)
+{
+	struct amdgpu_numa_info *numa_info;
+	int nid;
+
+	numa_info = xa_load(&numa_info_xa, pxm);
+
+	if (!numa_info) {
+		struct sysinfo info;
+
+		numa_info = kzalloc(sizeof(*numa_info), GFP_KERNEL);
+		if (!numa_info)
+			return NULL;
+
+		nid = pxm_to_node(pxm);
+		numa_info->pxm = pxm;
+		numa_info->nid = nid;
+
+		if (numa_info->nid == NUMA_NO_NODE) {
+			si_meminfo(&info);
+			numa_info->size = info.totalram * info.mem_unit;
+		} else {
+			numa_info->size = amdgpu_acpi_get_numa_size(nid);
+		}
+		xa_store(&numa_info_xa, numa_info->pxm, numa_info, GFP_KERNEL);
+	}
+
+	return numa_info;
+}
+#endif
+
+/**
+ * amdgpu_acpi_get_node_id - obtain the NUMA node id for corresponding amdgpu
+ * acpi device handle
+ *
+ * @handle: acpi handle
+ * @numa_info: amdgpu_numa_info structure holding numa information
+ *
+ * Queries the ACPI interface to fetch the corresponding NUMA Node ID for a
+ * given amdgpu acpi device.
+ *
+ * Returns ACPI STATUS OK with Node ID on success or the corresponding failure reason
+ */
+static acpi_status amdgpu_acpi_get_node_id(acpi_handle handle,
+				    struct amdgpu_numa_info **numa_info)
+{
+#ifdef CONFIG_ACPI_NUMA
+	u64 pxm;
+	acpi_status status;
+
+	if (!numa_info)
+		return_ACPI_STATUS(AE_ERROR);
+
+	status = acpi_evaluate_integer(handle, "_PXM", NULL, &pxm);
+
+	if (ACPI_FAILURE(status))
+		return status;
+
+	*numa_info = amdgpu_acpi_get_numa_info(pxm);
+
+	if (!*numa_info)
+		return_ACPI_STATUS(AE_ERROR);
+
+	return_ACPI_STATUS(AE_OK);
+#else
+	return_ACPI_STATUS(AE_NOT_EXIST);
+#endif
+}
+
+static struct amdgpu_acpi_dev_info *amdgpu_acpi_get_dev(u32 sbdf)
+{
+	struct amdgpu_acpi_dev_info *acpi_dev;
+
+	if (list_empty(&amdgpu_acpi_dev_list))
+		return NULL;
+
+	list_for_each_entry(acpi_dev, &amdgpu_acpi_dev_list, list)
+		if (acpi_dev->sbdf == sbdf)
+			return acpi_dev;
+
+	return NULL;
+}
+
+static int amdgpu_acpi_dev_init(struct amdgpu_acpi_dev_info **dev_info,
+				struct amdgpu_acpi_xcc_info *xcc_info, u32 sbdf)
+{
+	struct amdgpu_acpi_dev_info *tmp;
+	union acpi_object *obj;
+	int ret = -ENOENT;
+
+	*dev_info = NULL;
+	tmp = kzalloc(sizeof(struct amdgpu_acpi_dev_info), GFP_KERNEL);
+	if (!tmp)
+		return -ENOMEM;
+
+	INIT_LIST_HEAD(&tmp->xcc_list);
+	INIT_LIST_HEAD(&tmp->list);
+	tmp->sbdf = sbdf;
+
+	obj = acpi_evaluate_dsm_typed(xcc_info->handle, &amd_xcc_dsm_guid, 0,
+				      AMD_XCC_DSM_GET_SUPP_MODE, NULL,
+				      ACPI_TYPE_INTEGER);
+
+	if (!obj) {
+		acpi_handle_debug(xcc_info->handle,
+				  "_DSM function %d evaluation failed",
+				  AMD_XCC_DSM_GET_SUPP_MODE);
+		ret = -ENOENT;
+		goto out;
+	}
+
+	tmp->supp_xcp_mode = obj->integer.value & 0xFFFF;
+	ACPI_FREE(obj);
+
+	obj = acpi_evaluate_dsm_typed(xcc_info->handle, &amd_xcc_dsm_guid, 0,
+				      AMD_XCC_DSM_GET_XCP_MODE, NULL,
+				      ACPI_TYPE_INTEGER);
+
+	if (!obj) {
+		acpi_handle_debug(xcc_info->handle,
+				  "_DSM function %d evaluation failed",
+				  AMD_XCC_DSM_GET_XCP_MODE);
+		ret = -ENOENT;
+		goto out;
+	}
+
+	tmp->xcp_mode = obj->integer.value & 0xFFFF;
+	tmp->mem_mode = (obj->integer.value >> 32) & 0xFFFF;
+	ACPI_FREE(obj);
+
+	/* Evaluate DSMs and fill XCC information */
+	obj = acpi_evaluate_dsm_typed(xcc_info->handle, &amd_xcc_dsm_guid, 0,
+				      AMD_XCC_DSM_GET_TMR_INFO, NULL,
+				      ACPI_TYPE_PACKAGE);
+
+	if (!obj || obj->package.count < 2) {
+		acpi_handle_debug(xcc_info->handle,
+				  "_DSM function %d evaluation failed",
+				  AMD_XCC_DSM_GET_TMR_INFO);
+		ret = -ENOENT;
+		goto out;
+	}
+
+	tmp->tmr_base = obj->package.elements[0].integer.value;
+	tmp->tmr_size = obj->package.elements[1].integer.value;
+	ACPI_FREE(obj);
+
+	DRM_DEBUG_DRIVER(
+		"New dev(%x): Supported xcp mode: %x curr xcp_mode : %x mem mode : %x, tmr base: %llx tmr size: %llx  ",
+		tmp->sbdf, tmp->supp_xcp_mode, tmp->xcp_mode, tmp->mem_mode,
+		tmp->tmr_base, tmp->tmr_size);
+	list_add_tail(&tmp->list, &amdgpu_acpi_dev_list);
+	*dev_info = tmp;
+
+	return 0;
+
+out:
+	if (obj)
+		ACPI_FREE(obj);
+	kfree(tmp);
+
+	return ret;
+}
+
+static int amdgpu_acpi_get_xcc_info(struct amdgpu_acpi_xcc_info *xcc_info,
+				    u32 *sbdf)
+{
+	union acpi_object *obj;
+	acpi_status status;
+	int ret = -ENOENT;
+
+	obj = acpi_evaluate_dsm_typed(xcc_info->handle, &amd_xcc_dsm_guid, 0,
+				      AMD_XCC_DSM_GET_NUM_FUNCS, NULL,
+				      ACPI_TYPE_INTEGER);
+
+	if (!obj || obj->integer.value != AMD_XCC_DSM_NUM_FUNCS)
+		goto out;
+	ACPI_FREE(obj);
+
+	/* Evaluate DSMs and fill XCC information */
+	obj = acpi_evaluate_dsm_typed(xcc_info->handle, &amd_xcc_dsm_guid, 0,
+				      AMD_XCC_DSM_GET_VF_XCC_MAPPING, NULL,
+				      ACPI_TYPE_INTEGER);
+
+	if (!obj) {
+		acpi_handle_debug(xcc_info->handle,
+				  "_DSM function %d evaluation failed",
+				  AMD_XCC_DSM_GET_VF_XCC_MAPPING);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	/* PF xcc id [39:32] */
+	xcc_info->phy_id = (obj->integer.value >> 32) & 0xFF;
+	/* xcp node of this xcc [47:40] */
+	xcc_info->xcp_node = (obj->integer.value >> 40) & 0xFF;
+	/* PF domain of this xcc [31:16] */
+	*sbdf = (obj->integer.value) & 0xFFFF0000;
+	/* PF bus/dev/fn of this xcc [63:48] */
+	*sbdf |= (obj->integer.value >> 48) & 0xFFFF;
+	ACPI_FREE(obj);
+	obj = NULL;
+
+	status =
+		amdgpu_acpi_get_node_id(xcc_info->handle, &xcc_info->numa_info);
+
+	/* TODO: check if this check is required */
+	if (ACPI_SUCCESS(status))
+		ret = 0;
+out:
+	if (obj)
+		ACPI_FREE(obj);
+
+	return ret;
+}
+
+static int amdgpu_acpi_enumerate_xcc(void)
+{
+	struct amdgpu_acpi_dev_info *dev_info = NULL;
+	struct amdgpu_acpi_xcc_info *xcc_info;
+	struct acpi_device *acpi_dev;
+	char hid[ACPI_ID_LEN];
+	int ret, id;
+	u32 sbdf;
+
+	INIT_LIST_HEAD(&amdgpu_acpi_dev_list);
+	xa_init(&numa_info_xa);
+
+	for (id = 0; id < AMD_XCC_MAX_HID; id++) {
+		sprintf(hid, "%s%d", "AMD", AMD_XCC_HID_START + id);
+		acpi_dev = acpi_dev_get_first_match_dev(hid, NULL, -1);
+		/* These ACPI objects are expected to be in sequential order. If
+		 * one is not found, no need to check the rest.
+		 */
+		if (!acpi_dev) {
+			DRM_DEBUG_DRIVER("No matching acpi device found for %s",
+					 hid);
+			break;
+		}
+
+		xcc_info = kzalloc(sizeof(struct amdgpu_acpi_xcc_info),
+				   GFP_KERNEL);
+		if (!xcc_info) {
+			DRM_ERROR("Failed to allocate memory for xcc info\n");
+			return -ENOMEM;
+		}
+
+		INIT_LIST_HEAD(&xcc_info->list);
+		xcc_info->handle = acpi_device_handle(acpi_dev);
+		acpi_dev_put(acpi_dev);
+
+		ret = amdgpu_acpi_get_xcc_info(xcc_info, &sbdf);
+		if (ret) {
+			kfree(xcc_info);
+			continue;
+		}
+
+		dev_info = amdgpu_acpi_get_dev(sbdf);
+
+		if (!dev_info)
+			ret = amdgpu_acpi_dev_init(&dev_info, xcc_info, sbdf);
+
+		if (ret == -ENOMEM)
+			return ret;
+
+		if (!dev_info) {
+			kfree(xcc_info);
+			continue;
+		}
+
+		list_add_tail(&xcc_info->list, &dev_info->xcc_list);
+	}
+
+	return 0;
+}
+
+int amdgpu_acpi_get_tmr_info(struct amdgpu_device *adev, u64 *tmr_offset,
+			     u64 *tmr_size)
+{
+	struct amdgpu_acpi_dev_info *dev_info;
+	u32 sbdf;
+
+	if (!tmr_offset || !tmr_size)
+		return -EINVAL;
+
+	sbdf = (pci_domain_nr(adev->pdev->bus) << 16);
+	sbdf |= pci_dev_id(adev->pdev);
+	dev_info = amdgpu_acpi_get_dev(sbdf);
+	if (!dev_info)
+		return -ENOENT;
+
+	*tmr_offset = dev_info->tmr_base;
+	*tmr_size = dev_info->tmr_size;
+
+	return 0;
+}
+
+int amdgpu_acpi_get_mem_info(struct amdgpu_device *adev, int xcc_id,
+			     struct amdgpu_numa_info *numa_info)
+{
+	struct amdgpu_acpi_dev_info *dev_info;
+	struct amdgpu_acpi_xcc_info *xcc_info;
+	u32 sbdf;
+
+	if (!numa_info)
+		return -EINVAL;
+
+	sbdf = (pci_domain_nr(adev->pdev->bus) << 16);
+	sbdf |= pci_dev_id(adev->pdev);
+	dev_info = amdgpu_acpi_get_dev(sbdf);
+	if (!dev_info)
+		return -ENOENT;
+
+	list_for_each_entry(xcc_info, &dev_info->xcc_list, list) {
+		if (xcc_info->phy_id == xcc_id) {
+			memcpy(numa_info, xcc_info->numa_info,
+			       sizeof(*numa_info));
+			return 0;
+		}
+	}
+
+	return -ENOENT;
+}
+
 /**
  * amdgpu_acpi_event - handle notify events
  *
@@ -884,9 +1280,7 @@ void amdgpu_acpi_get_backlight_caps(struct amdgpu_dm_backlight_caps *caps)
 {
 	struct amdgpu_atif *atif = &amdgpu_acpi_priv.atif;
 
-	caps->caps_valid = atif->backlight_caps.caps_valid;
-	caps->min_input_signal = atif->backlight_caps.min_input_signal;
-	caps->max_input_signal = atif->backlight_caps.max_input_signal;
+	memcpy(caps, &atif->backlight_caps, sizeof(*caps));
 }
 
 /**
@@ -1012,14 +1406,11 @@ void amdgpu_acpi_detect(void)
 	struct pci_dev *pdev = NULL;
 	int ret;
 
-	while ((pdev = pci_get_class(PCI_CLASS_DISPLAY_VGA << 8, pdev)) != NULL) {
-		if (!atif->handle)
-			amdgpu_atif_pci_probe_handle(pdev);
-		if (!atcs->handle)
-			amdgpu_atcs_pci_probe_handle(pdev);
-	}
+	while ((pdev = pci_get_base_class(PCI_BASE_CLASS_DISPLAY, pdev))) {
+		if ((pdev->class != PCI_CLASS_DISPLAY_VGA << 8) &&
+		    (pdev->class != PCI_CLASS_DISPLAY_OTHER << 8))
+			continue;
 
-	while ((pdev = pci_get_class(PCI_CLASS_DISPLAY_OTHER << 8, pdev)) != NULL) {
 		if (!atif->handle)
 			amdgpu_atif_pci_probe_handle(pdev);
 		if (!atcs->handle)
@@ -1054,6 +1445,36 @@ void amdgpu_acpi_detect(void)
 	} else {
 		atif->backlight_caps.caps_valid = false;
 	}
+
+	amdgpu_acpi_enumerate_xcc();
+}
+
+void amdgpu_acpi_release(void)
+{
+	struct amdgpu_acpi_dev_info *dev_info, *dev_tmp;
+	struct amdgpu_acpi_xcc_info *xcc_info, *xcc_tmp;
+	struct amdgpu_numa_info *numa_info;
+	unsigned long index;
+
+	xa_for_each(&numa_info_xa, index, numa_info) {
+		kfree(numa_info);
+		xa_erase(&numa_info_xa, index);
+	}
+
+	if (list_empty(&amdgpu_acpi_dev_list))
+		return;
+
+	list_for_each_entry_safe(dev_info, dev_tmp, &amdgpu_acpi_dev_list,
+				 list) {
+		list_for_each_entry_safe(xcc_info, xcc_tmp, &dev_info->xcc_list,
+					 list) {
+			list_del(&xcc_info->list);
+			kfree(xcc_info);
+		}
+
+		list_del(&dev_info->list);
+		kfree(dev_info);
+	}
 }
 
 #if IS_ENABLED(CONFIG_SUSPEND)
@@ -1086,22 +1507,59 @@ bool amdgpu_acpi_is_s0ix_active(struct amdgpu_device *adev)
 	if (adev->asic_type < CHIP_RAVEN)
 		return false;
 
+	if (!(adev->pm.pp_feature & PP_GFXOFF_MASK))
+		return false;
+
 	/*
 	 * If ACPI_FADT_LOW_POWER_S0 is not set in the FADT, it is generally
 	 * risky to do any special firmware-related preparations for entering
 	 * S0ix even though the system is suspending to idle, so return false
 	 * in that case.
 	 */
-	if (!(acpi_gbl_FADT.flags & ACPI_FADT_LOW_POWER_S0))
-		dev_warn_once(adev->dev,
+	if (!(acpi_gbl_FADT.flags & ACPI_FADT_LOW_POWER_S0)) {
+		dev_err_once(adev->dev,
 			      "Power consumption will be higher as BIOS has not been configured for suspend-to-idle.\n"
 			      "To use suspend-to-idle change the sleep mode in BIOS setup.\n");
+		return false;
+	}
 
 #if !IS_ENABLED(CONFIG_AMD_PMC)
-	dev_warn_once(adev->dev,
+	dev_err_once(adev->dev,
 		      "Power consumption will be higher as the kernel has not been compiled with CONFIG_AMD_PMC.\n");
-#endif /* CONFIG_AMD_PMC */
+	return false;
+#else
 	return true;
+#endif /* CONFIG_AMD_PMC */
 }
-
 #endif /* CONFIG_SUSPEND */
+
+#if IS_ENABLED(CONFIG_DRM_AMD_ISP)
+static const struct acpi_device_id isp_sensor_ids[] = {
+	{ "OMNI5C10" },
+	{ }
+};
+
+static int isp_match_acpi_device_ids(struct device *dev, const void *data)
+{
+	return acpi_match_device(data, dev) ? 1 : 0;
+}
+
+int amdgpu_acpi_get_isp4_dev(struct acpi_device **dev)
+{
+	struct device *pdev __free(put_device) = NULL;
+	struct acpi_device *acpi_pdev;
+
+	pdev = bus_find_device(&platform_bus_type, NULL, isp_sensor_ids,
+			       isp_match_acpi_device_ids);
+	if (!pdev)
+		return -EINVAL;
+
+	acpi_pdev = ACPI_COMPANION(pdev);
+	if (!acpi_pdev)
+		return -ENODEV;
+
+	*dev = acpi_pdev;
+
+	return 0;
+}
+#endif /* CONFIG_DRM_AMD_ISP */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_afmt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_afmt.c
index a4d65973bf7c..80771b1480ff 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_afmt.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_afmt.c
@@ -100,6 +100,7 @@ struct amdgpu_afmt_acr amdgpu_afmt_acr(uint32_t clock)
 	amdgpu_afmt_calc_cts(clock, &res.cts_32khz, &res.n_32khz, 32000);
 	amdgpu_afmt_calc_cts(clock, &res.cts_44_1khz, &res.n_44_1khz, 44100);
 	amdgpu_afmt_calc_cts(clock, &res.cts_48khz, &res.n_48khz, 48000);
+	res.clock = clock;
 
 	return res;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 0385f7f69278..a2879d2b7c8e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -28,6 +28,7 @@
 #include "amdgpu.h"
 #include "amdgpu_gfx.h"
 #include "amdgpu_dma_buf.h"
+#include <drm/ttm/ttm_tt.h>
 #include <linux/module.h>
 #include <linux/dma-buf.h>
 #include "amdgpu_xgmi.h"
@@ -53,7 +54,6 @@ int amdgpu_amdkfd_init(void)
 	amdgpu_amdkfd_total_mem_size *= si.mem_unit;
 
 	ret = kgd2kfd_init();
-	amdgpu_amdkfd_gpuvm_init_mem_limits();
 	kfd_initialized = !ret;
 
 	return ret;
@@ -133,16 +133,45 @@ static void amdgpu_amdkfd_reset_work(struct work_struct *work)
 
 	reset_context.method = AMD_RESET_METHOD_NONE;
 	reset_context.reset_req_dev = adev;
+	reset_context.src = adev->enable_mes ?
+			    AMDGPU_RESET_SRC_MES :
+			    AMDGPU_RESET_SRC_HWS;
 	clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
 
 	amdgpu_device_gpu_recover(adev, NULL, &reset_context);
 }
 
+static const struct drm_client_funcs kfd_client_funcs = {
+	.unregister	= drm_client_release,
+};
+
+int amdgpu_amdkfd_drm_client_create(struct amdgpu_device *adev)
+{
+	int ret;
+
+	if (!adev->kfd.init_complete || adev->kfd.client.dev)
+		return 0;
+
+	ret = drm_client_init(&adev->ddev, &adev->kfd.client, "kfd",
+			      &kfd_client_funcs);
+	if (ret) {
+		dev_err(adev->dev, "Failed to init DRM client: %d\n",
+			ret);
+		return ret;
+	}
+
+	drm_client_register(&adev->kfd.client);
+
+	return 0;
+}
+
 void amdgpu_amdkfd_device_init(struct amdgpu_device *adev)
 {
 	int i;
 	int last_valid_bit;
 
+	amdgpu_amdkfd_gpuvm_init_mem_limits();
+
 	if (adev->kfd.dev) {
 		struct kgd2kfd_shared_resources gpu_resources = {
 			.compute_vmid_bitmap =
@@ -162,8 +191,8 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev)
 		 * clear
 		 */
 		bitmap_complement(gpu_resources.cp_queue_bitmap,
-				  adev->gfx.mec.queue_bitmap,
-				  KGD_MAX_QUEUES);
+				  adev->gfx.mec_bitmap[0].queue_bitmap,
+				  AMDGPU_MAX_QUEUES);
 
 		/* According to linux/bitmap.h we shouldn't use bitmap_clear if
 		 * nbits is not compile time constant
@@ -171,7 +200,7 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev)
 		last_valid_bit = 1 /* only first MEC can have compute queues */
 				* adev->gfx.mec.num_pipe_per_mec
 				* adev->gfx.mec.num_queue_per_pipe;
-		for (i = last_valid_bit; i < KGD_MAX_QUEUES; ++i)
+		for (i = last_valid_bit; i < AMDGPU_MAX_QUEUES; ++i)
 			clear_bit(i, gpu_resources.cp_queue_bitmap);
 
 		amdgpu_doorbell_get_kfd_info(adev,
@@ -219,38 +248,53 @@ void amdgpu_amdkfd_interrupt(struct amdgpu_device *adev,
 		kgd2kfd_interrupt(adev->kfd.dev, ih_ring_entry);
 }
 
-void amdgpu_amdkfd_suspend(struct amdgpu_device *adev, bool run_pm)
+void amdgpu_amdkfd_suspend(struct amdgpu_device *adev, bool suspend_proc)
 {
-	if (adev->kfd.dev)
-		kgd2kfd_suspend(adev->kfd.dev, run_pm);
+	if (adev->kfd.dev) {
+		if (adev->in_s0ix)
+			kgd2kfd_stop_sched_all_nodes(adev->kfd.dev);
+		else
+			kgd2kfd_suspend(adev->kfd.dev, suspend_proc);
+	}
 }
 
-int amdgpu_amdkfd_resume_iommu(struct amdgpu_device *adev)
+int amdgpu_amdkfd_resume(struct amdgpu_device *adev, bool resume_proc)
 {
 	int r = 0;
 
-	if (adev->kfd.dev)
-		r = kgd2kfd_resume_iommu(adev->kfd.dev);
+	if (adev->kfd.dev) {
+		if (adev->in_s0ix)
+			r = kgd2kfd_start_sched_all_nodes(adev->kfd.dev);
+		else
+			r = kgd2kfd_resume(adev->kfd.dev, resume_proc);
+	}
 
 	return r;
 }
 
-int amdgpu_amdkfd_resume(struct amdgpu_device *adev, bool run_pm)
+void amdgpu_amdkfd_suspend_process(struct amdgpu_device *adev)
+{
+	if (adev->kfd.dev)
+		kgd2kfd_suspend_process(adev->kfd.dev);
+}
+
+int amdgpu_amdkfd_resume_process(struct amdgpu_device *adev)
 {
 	int r = 0;
 
 	if (adev->kfd.dev)
-		r = kgd2kfd_resume(adev->kfd.dev, run_pm);
+		r = kgd2kfd_resume_process(adev->kfd.dev);
 
 	return r;
 }
 
-int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev)
+int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev,
+			    struct amdgpu_reset_context *reset_context)
 {
 	int r = 0;
 
 	if (adev->kfd.dev)
-		r = kgd2kfd_pre_reset(adev->kfd.dev);
+		r = kgd2kfd_pre_reset(adev->kfd.dev, reset_context);
 
 	return r;
 }
@@ -344,15 +388,18 @@ allocate_mem_reserve_bo_failed:
 	return r;
 }
 
-void amdgpu_amdkfd_free_gtt_mem(struct amdgpu_device *adev, void *mem_obj)
+void amdgpu_amdkfd_free_gtt_mem(struct amdgpu_device *adev, void **mem_obj)
 {
-	struct amdgpu_bo *bo = (struct amdgpu_bo *) mem_obj;
+	struct amdgpu_bo **bo = (struct amdgpu_bo **) mem_obj;
 
-	amdgpu_bo_reserve(bo, true);
-	amdgpu_bo_kunmap(bo);
-	amdgpu_bo_unpin(bo);
-	amdgpu_bo_unreserve(bo);
-	amdgpu_bo_unref(&(bo));
+	if (!bo || !*bo)
+		return;
+
+	(void)amdgpu_bo_reserve(*bo, true);
+	amdgpu_bo_kunmap(*bo);
+	amdgpu_bo_unpin(*bo);
+	amdgpu_bo_unreserve(*bo);
+	amdgpu_bo_unref(bo);
 }
 
 int amdgpu_amdkfd_alloc_gws(struct amdgpu_device *adev, size_t size,
@@ -427,14 +474,26 @@ uint32_t amdgpu_amdkfd_get_fw_version(struct amdgpu_device *adev,
 }
 
 void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device *adev,
-				      struct kfd_local_mem_info *mem_info)
+				      struct kfd_local_mem_info *mem_info,
+				      struct amdgpu_xcp *xcp)
 {
 	memset(mem_info, 0, sizeof(*mem_info));
 
-	mem_info->local_mem_size_public = adev->gmc.visible_vram_size;
-	mem_info->local_mem_size_private = adev->gmc.real_vram_size -
+	if (xcp) {
+		if (adev->gmc.real_vram_size == adev->gmc.visible_vram_size)
+			mem_info->local_mem_size_public =
+					KFD_XCP_MEMORY_SIZE(adev, xcp->id);
+		else
+			mem_info->local_mem_size_private =
+					KFD_XCP_MEMORY_SIZE(adev, xcp->id);
+	} else if (adev->apu_prefer_gtt) {
+		mem_info->local_mem_size_public = (ttm_tt_pages_limit() << PAGE_SHIFT);
+		mem_info->local_mem_size_private = 0;
+	} else {
+		mem_info->local_mem_size_public = adev->gmc.visible_vram_size;
+		mem_info->local_mem_size_private = adev->gmc.real_vram_size -
 						adev->gmc.visible_vram_size;
-
+	}
 	mem_info->vram_width = adev->gmc.vram_width;
 
 	pr_debug("Address base: %pap public 0x%llx private 0x%llx\n",
@@ -442,9 +501,7 @@ void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device *adev,
 			mem_info->local_mem_size_public,
 			mem_info->local_mem_size_private);
 
-	if (amdgpu_sriov_vf(adev))
-		mem_info->mem_clk_max = adev->clock.default_mclk / 100;
-	else if (adev->pm.dpm_enabled) {
+	if (adev->pm.dpm_enabled) {
 		if (amdgpu_emu_mode == 1)
 			mem_info->mem_clk_max = 0;
 		else
@@ -463,41 +520,17 @@ uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct amdgpu_device *adev)
 uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct amdgpu_device *adev)
 {
 	/* the sclk is in quantas of 10kHz */
-	if (amdgpu_sriov_vf(adev))
-		return adev->clock.default_sclk / 100;
-	else if (adev->pm.dpm_enabled)
+	if (adev->pm.dpm_enabled)
 		return amdgpu_dpm_get_sclk(adev, false) / 100;
 	else
 		return 100;
 }
 
-void amdgpu_amdkfd_get_cu_info(struct amdgpu_device *adev, struct kfd_cu_info *cu_info)
-{
-	struct amdgpu_cu_info acu_info = adev->gfx.cu_info;
-
-	memset(cu_info, 0, sizeof(*cu_info));
-	if (sizeof(cu_info->cu_bitmap) != sizeof(acu_info.bitmap))
-		return;
-
-	cu_info->cu_active_number = acu_info.number;
-	cu_info->cu_ao_mask = acu_info.ao_cu_mask;
-	memcpy(&cu_info->cu_bitmap[0], &acu_info.bitmap[0],
-	       sizeof(acu_info.bitmap));
-	cu_info->num_shader_engines = adev->gfx.config.max_shader_engines;
-	cu_info->num_shader_arrays_per_engine = adev->gfx.config.max_sh_per_se;
-	cu_info->num_cu_per_sh = adev->gfx.config.max_cu_per_sh;
-	cu_info->simd_per_cu = acu_info.simd_per_cu;
-	cu_info->max_waves_per_simd = acu_info.max_waves_per_simd;
-	cu_info->wave_front_size = acu_info.wave_front_size;
-	cu_info->max_scratch_slots_per_cu = acu_info.max_scratch_slots_per_cu;
-	cu_info->lds_size = acu_info.lds_size;
-}
-
 int amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int dma_buf_fd,
 				  struct amdgpu_device **dmabuf_adev,
 				  uint64_t *bo_size, void *metadata_buffer,
 				  size_t buffer_size, uint32_t *metadata_size,
-				  uint32_t *flags)
+				  uint32_t *flags, int8_t *xcp_id)
 {
 	struct dma_buf *dma_buf;
 	struct drm_gem_object *obj;
@@ -541,54 +574,14 @@ int amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int dma_buf_fd,
 		if (bo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)
 			*flags |= KFD_IOC_ALLOC_MEM_FLAGS_PUBLIC;
 	}
+	if (xcp_id)
+		*xcp_id = bo->xcp_id;
 
 out_put:
 	dma_buf_put(dma_buf);
 	return r;
 }
 
-uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct amdgpu_device *dst,
-					  struct amdgpu_device *src)
-{
-	struct amdgpu_device *peer_adev = src;
-	struct amdgpu_device *adev = dst;
-	int ret = amdgpu_xgmi_get_hops_count(adev, peer_adev);
-
-	if (ret < 0) {
-		DRM_ERROR("amdgpu: failed to get  xgmi hops count between node %d and %d. ret = %d\n",
-			adev->gmc.xgmi.physical_node_id,
-			peer_adev->gmc.xgmi.physical_node_id, ret);
-		ret = 0;
-	}
-	return  (uint8_t)ret;
-}
-
-int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct amdgpu_device *dst,
-					    struct amdgpu_device *src,
-					    bool is_min)
-{
-	struct amdgpu_device *adev = dst, *peer_adev;
-	int num_links;
-
-	if (adev->asic_type != CHIP_ALDEBARAN)
-		return 0;
-
-	if (src)
-		peer_adev = src;
-
-	/* num links returns 0 for indirect peers since indirect route is unknown. */
-	num_links = is_min ? 1 : amdgpu_xgmi_get_num_links(adev, peer_adev);
-	if (num_links < 0) {
-		DRM_ERROR("amdgpu: failed to get xgmi num links between node %d and %d. ret = %d\n",
-			adev->gmc.xgmi.physical_node_id,
-			peer_adev->gmc.xgmi.physical_node_id, num_links);
-		num_links = 0;
-	}
-
-	/* Aldebaran xGMI DPM is defeatured so assume x16 x 25Gbps for bandwidth. */
-	return (num_links * 16 * 25000)/BITS_PER_BYTE;
-}
-
 int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct amdgpu_device *adev, bool is_min)
 {
 	int num_lanes_shift = (is_min ? ffs(adev->pm.pcie_mlw_mask) :
@@ -673,7 +666,7 @@ int amdgpu_amdkfd_submit_ib(struct amdgpu_device *adev,
 		goto err;
 	}
 
-	ret = amdgpu_job_alloc(adev, NULL, NULL, NULL, 1, &job);
+	ret = amdgpu_job_alloc(adev, NULL, NULL, NULL, 1, &job, 0);
 	if (ret)
 		goto err;
 
@@ -706,12 +699,20 @@ err:
 
 void amdgpu_amdkfd_set_compute_idle(struct amdgpu_device *adev, bool idle)
 {
-	/* Temporary workaround to fix issues observed in some
-	 * compute applications when GFXOFF is enabled on GFX11.
-	 */
-	if (IP_VERSION_MAJ(adev->ip_versions[GC_HWIP][0]) == 11) {
+	enum amd_powergating_state state = idle ? AMD_PG_STATE_GATE : AMD_PG_STATE_UNGATE;
+	if ((IP_VERSION_MAJ(amdgpu_ip_version(adev, GC_HWIP, 0)) == 11 &&
+	    ((adev->mes.kiq_version & AMDGPU_MES_VERSION_MASK) <= 64)) ||
+		(IP_VERSION_MAJ(amdgpu_ip_version(adev, GC_HWIP, 0)) == 12)) {
 		pr_debug("GFXOFF is %s\n", idle ? "enabled" : "disabled");
 		amdgpu_gfx_off_ctrl(adev, idle);
+	} else if ((IP_VERSION_MAJ(amdgpu_ip_version(adev, GC_HWIP, 0)) == 9) &&
+		(adev->flags & AMD_IS_APU)) {
+		/* Disable GFXOFF and PG. Temporary workaround
+		 * to fix some compute applications issue on GFX9.
+		 */
+		struct amdgpu_ip_block *gfx_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GFX);
+		if (gfx_block != NULL)
+			gfx_block->version->funcs->set_powergating_state((void *)gfx_block, state);
 	}
 	amdgpu_dpm_switch_power_profile(adev,
 					PP_SMC_POWER_PROFILE_COMPUTE,
@@ -726,47 +727,186 @@ bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid)
 	return false;
 }
 
-int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct amdgpu_device *adev,
-				     uint16_t vmid)
+bool amdgpu_amdkfd_have_atomics_support(struct amdgpu_device *adev)
 {
-	if (adev->family == AMDGPU_FAMILY_AI) {
-		int i;
+	return adev->have_atomics_support;
+}
 
-		for (i = 0; i < adev->num_vmhubs; i++)
-			amdgpu_gmc_flush_gpu_tlb(adev, vmid, i, 0);
-	} else {
-		amdgpu_gmc_flush_gpu_tlb(adev, vmid, AMDGPU_GFXHUB_0, 0);
-	}
+void amdgpu_amdkfd_debug_mem_fence(struct amdgpu_device *adev)
+{
+	amdgpu_device_flush_hdp(adev, NULL);
+}
+
+bool amdgpu_amdkfd_is_fed(struct amdgpu_device *adev)
+{
+	return amdgpu_ras_get_fed_status(adev);
+}
+
+void amdgpu_amdkfd_ras_pasid_poison_consumption_handler(struct amdgpu_device *adev,
+				enum amdgpu_ras_block block, uint16_t pasid,
+				pasid_notify pasid_fn, void *data, uint32_t reset)
+{
+	amdgpu_umc_pasid_poison_handler(adev, block, pasid, pasid_fn, data, reset);
+}
+
+void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev,
+	enum amdgpu_ras_block block, uint32_t reset)
+{
+	amdgpu_umc_pasid_poison_handler(adev, block, 0, NULL, NULL, reset);
+}
+
+int amdgpu_amdkfd_send_close_event_drain_irq(struct amdgpu_device *adev,
+					uint32_t *payload)
+{
+	int ret;
+
+	/* Device or IH ring is not ready so bail. */
+	ret = amdgpu_ih_wait_on_checkpoint_process_ts(adev, &adev->irq.ih);
+	if (ret)
+		return ret;
+
+	/* Send payload to fence KFD interrupts */
+	amdgpu_amdkfd_interrupt(adev, payload);
 
 	return 0;
 }
 
-int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
-				      uint16_t pasid, enum TLB_FLUSH_TYPE flush_type)
+int amdgpu_amdkfd_check_and_lock_kfd(struct amdgpu_device *adev)
 {
-	bool all_hub = false;
+	return kgd2kfd_check_and_lock_kfd(adev->kfd.dev);
+}
 
-	if (adev->family == AMDGPU_FAMILY_AI ||
-	    adev->family == AMDGPU_FAMILY_RV)
-		all_hub = true;
+void amdgpu_amdkfd_unlock_kfd(struct amdgpu_device *adev)
+{
+	kgd2kfd_unlock_kfd(adev->kfd.dev);
+}
 
-	return amdgpu_gmc_flush_gpu_tlb_pasid(adev, pasid, flush_type, all_hub);
+
+u64 amdgpu_amdkfd_xcp_memory_size(struct amdgpu_device *adev, int xcp_id)
+{
+	s8 mem_id = KFD_XCP_MEM_ID(adev, xcp_id);
+	u64 tmp;
+
+	if (adev->gmc.num_mem_partitions && xcp_id >= 0 && mem_id >= 0) {
+		if (adev->gmc.is_app_apu && adev->gmc.num_mem_partitions == 1) {
+			/* In NPS1 mode, we should restrict the vram reporting
+			 * tied to the ttm_pages_limit which is 1/2 of the system
+			 * memory. For other partition modes, the HBM is uniformly
+			 * divided already per numa node reported. If user wants to
+			 * go beyond the default ttm limit and maximize the ROCm
+			 * allocations, they can go up to max ttm and sysmem limits.
+			 */
+
+			tmp = (ttm_tt_pages_limit() << PAGE_SHIFT) / num_online_nodes();
+		} else {
+			tmp = adev->gmc.mem_partitions[mem_id].size;
+		}
+		do_div(tmp, adev->xcp_mgr->num_xcp_per_mem_partition);
+		return ALIGN_DOWN(tmp, PAGE_SIZE);
+	} else if (adev->apu_prefer_gtt) {
+		return (ttm_tt_pages_limit() << PAGE_SHIFT);
+	} else {
+		return adev->gmc.real_vram_size;
+	}
 }
 
-bool amdgpu_amdkfd_have_atomics_support(struct amdgpu_device *adev)
+int amdgpu_amdkfd_unmap_hiq(struct amdgpu_device *adev, u32 doorbell_off,
+			    u32 inst)
 {
-	return adev->have_atomics_support;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[inst];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
+	struct amdgpu_ring_funcs *ring_funcs;
+	struct amdgpu_ring *ring;
+	int r = 0;
+
+	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
+		return -EINVAL;
+
+	if (!kiq_ring->sched.ready || amdgpu_in_reset(adev))
+		return 0;
+
+	ring_funcs = kzalloc(sizeof(*ring_funcs), GFP_KERNEL);
+	if (!ring_funcs)
+		return -ENOMEM;
+
+	ring = kzalloc(sizeof(*ring), GFP_KERNEL);
+	if (!ring) {
+		r = -ENOMEM;
+		goto free_ring_funcs;
+	}
+
+	ring_funcs->type = AMDGPU_RING_TYPE_COMPUTE;
+	ring->doorbell_index = doorbell_off;
+	ring->funcs = ring_funcs;
+
+	spin_lock(&kiq->ring_lock);
+
+	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size)) {
+		spin_unlock(&kiq->ring_lock);
+		r = -ENOMEM;
+		goto free_ring;
+	}
+
+	kiq->pmf->kiq_unmap_queues(kiq_ring, ring, RESET_QUEUES, 0, 0);
+
+	/* Submit unmap queue packet */
+	amdgpu_ring_commit(kiq_ring);
+	/*
+	 * Ring test will do a basic scratch register change check. Just run
+	 * this to ensure that unmap queues that is submitted before got
+	 * processed successfully before returning.
+	 */
+	r = amdgpu_ring_test_helper(kiq_ring);
+
+	spin_unlock(&kiq->ring_lock);
+
+free_ring:
+	kfree(ring);
+
+free_ring_funcs:
+	kfree(ring_funcs);
+
+	return r;
+}
+
+/* Stop scheduling on KFD */
+int amdgpu_amdkfd_stop_sched(struct amdgpu_device *adev, uint32_t node_id)
+{
+	if (!adev->kfd.init_complete)
+		return 0;
+
+	return kgd2kfd_stop_sched(adev->kfd.dev, node_id);
 }
 
-void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev, bool reset)
+/* Start scheduling on KFD */
+int amdgpu_amdkfd_start_sched(struct amdgpu_device *adev, uint32_t node_id)
 {
-	amdgpu_umc_poison_handler(adev, reset);
+	if (!adev->kfd.init_complete)
+		return 0;
+
+	return kgd2kfd_start_sched(adev->kfd.dev, node_id);
 }
 
-bool amdgpu_amdkfd_ras_query_utcl2_poison_status(struct amdgpu_device *adev)
+/* check if there are KFD queues active */
+bool amdgpu_amdkfd_compute_active(struct amdgpu_device *adev, uint32_t node_id)
 {
-	if (adev->gfx.ras && adev->gfx.ras->query_utcl2_poison_status)
-		return adev->gfx.ras->query_utcl2_poison_status(adev);
-	else
+	if (!adev->kfd.init_complete)
 		return false;
+
+	return kgd2kfd_compute_active(adev->kfd.dev, node_id);
+}
+
+/* Config CGTT_SQ_CLK_CTRL */
+int amdgpu_amdkfd_config_sq_perfmon(struct amdgpu_device *adev, uint32_t xcp_id,
+	bool core_override_enable, bool reg_override_enable, bool perfmon_override_enable)
+{
+	int r;
+
+	if (!adev->kfd.init_complete)
+		return 0;
+
+	r = psp_config_sq_perfmon(&adev->psp, xcp_id, core_override_enable,
+					reg_override_enable, perfmon_override_enable);
+
+	return r;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index 01ba3589b60a..8bdfcde2029b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -25,15 +25,18 @@
 #ifndef AMDGPU_AMDKFD_H_INCLUDED
 #define AMDGPU_AMDKFD_H_INCLUDED
 
+#include <linux/list.h>
 #include <linux/types.h>
 #include <linux/mm.h>
 #include <linux/kthread.h>
 #include <linux/workqueue.h>
 #include <linux/mmu_notifier.h>
+#include <linux/memremap.h>
 #include <kgd_kfd_interface.h>
-#include <drm/ttm/ttm_execbuf_util.h>
+#include <drm/drm_client.h>
 #include "amdgpu_sync.h"
 #include "amdgpu_vm.h"
+#include "amdgpu_xcp.h"
 
 extern uint64_t amdgpu_amdkfd_total_mem_size;
 
@@ -44,6 +47,8 @@ enum TLB_FLUSH_TYPE {
 };
 
 struct amdgpu_device;
+struct kfd_process_device;
+struct amdgpu_reset_context;
 
 enum kfd_mem_attachment_type {
 	KFD_MEM_ATT_SHARED,	/* Share kgd_mem->bo or another attachment's */
@@ -66,11 +71,10 @@ struct kgd_mem {
 	struct mutex lock;
 	struct amdgpu_bo *bo;
 	struct dma_buf *dmabuf;
-	struct hmm_range *range;
+	struct amdgpu_hmm_range *range;
 	struct list_head attachments;
 	/* protected by amdkfd_process_info.lock */
-	struct ttm_validate_buffer validate_list;
-	struct ttm_validate_buffer resv_list;
+	struct list_head validate_list;
 	uint32_t domain;
 	unsigned int mapped_to_gpu_memory;
 	uint64_t va;
@@ -82,6 +86,7 @@ struct kgd_mem {
 
 	struct amdgpu_sync sync;
 
+	uint32_t gem_handle;
 	bool aql_queue;
 	bool is_imported;
 };
@@ -97,10 +102,18 @@ struct amdgpu_amdkfd_fence {
 
 struct amdgpu_kfd_dev {
 	struct kfd_dev *dev;
-	int64_t vram_used;
-	uint64_t vram_used_aligned;
+	int64_t vram_used[MAX_XCP];
+	uint64_t vram_used_aligned[MAX_XCP];
 	bool init_complete;
 	struct work_struct reset_work;
+
+	/* Client for KFD BO GEM handle allocations */
+	struct drm_client_dev client;
+
+	/* HMM page migration MEMORY_DEVICE_PRIVATE mapping
+	 * Must be last --ends in a flexible-array member.
+	 */
+	struct dev_pagemap pgmap;
 };
 
 enum kgd_engine_type {
@@ -143,28 +156,28 @@ struct amdkfd_process_info {
 int amdgpu_amdkfd_init(void);
 void amdgpu_amdkfd_fini(void);
 
-void amdgpu_amdkfd_suspend(struct amdgpu_device *adev, bool run_pm);
-int amdgpu_amdkfd_resume_iommu(struct amdgpu_device *adev);
-int amdgpu_amdkfd_resume(struct amdgpu_device *adev, bool run_pm);
+void amdgpu_amdkfd_suspend(struct amdgpu_device *adev, bool suspend_proc);
+int amdgpu_amdkfd_resume(struct amdgpu_device *adev, bool resume_proc);
+void amdgpu_amdkfd_suspend_process(struct amdgpu_device *adev);
+int amdgpu_amdkfd_resume_process(struct amdgpu_device *adev);
 void amdgpu_amdkfd_interrupt(struct amdgpu_device *adev,
 			const void *ih_ring_entry);
 void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev);
 void amdgpu_amdkfd_device_init(struct amdgpu_device *adev);
 void amdgpu_amdkfd_device_fini_sw(struct amdgpu_device *adev);
+int amdgpu_amdkfd_check_and_lock_kfd(struct amdgpu_device *adev);
+void amdgpu_amdkfd_unlock_kfd(struct amdgpu_device *adev);
 int amdgpu_amdkfd_submit_ib(struct amdgpu_device *adev,
 				enum kgd_engine_type engine,
 				uint32_t vmid, uint64_t gpu_addr,
 				uint32_t *ib_cmd, uint32_t ib_len);
 void amdgpu_amdkfd_set_compute_idle(struct amdgpu_device *adev, bool idle);
 bool amdgpu_amdkfd_have_atomics_support(struct amdgpu_device *adev);
-int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct amdgpu_device *adev,
-				uint16_t vmid);
-int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
-				uint16_t pasid, enum TLB_FLUSH_TYPE flush_type);
 
 bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid);
 
-int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev);
+int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev,
+			    struct amdgpu_reset_context *reset_context);
 
 int amdgpu_amdkfd_post_reset(struct amdgpu_device *adev);
 
@@ -176,15 +189,20 @@ int amdgpu_queue_mask_bit_to_set_resource_bit(struct amdgpu_device *adev,
 struct amdgpu_amdkfd_fence *amdgpu_amdkfd_fence_create(u64 context,
 				struct mm_struct *mm,
 				struct svm_range_bo *svm_bo);
+
+int amdgpu_amdkfd_drm_client_create(struct amdgpu_device *adev);
 #if defined(CONFIG_DEBUG_FS)
 int kfd_debugfs_kfd_mem_limits(struct seq_file *m, void *data);
 #endif
 #if IS_ENABLED(CONFIG_HSA_AMD)
 bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm);
 struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f);
-int amdgpu_amdkfd_remove_fence_on_pt_pd_bos(struct amdgpu_bo *bo);
+void amdgpu_amdkfd_remove_all_eviction_fences(struct amdgpu_bo *bo);
 int amdgpu_amdkfd_evict_userptr(struct mmu_interval_notifier *mni,
 				unsigned long cur_seq, struct kgd_mem *mem);
+int amdgpu_amdkfd_bo_validate_and_fence(struct amdgpu_bo *bo,
+					uint32_t domain,
+					struct dma_fence *fence);
 #else
 static inline
 bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm)
@@ -199,9 +217,8 @@ struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f)
 }
 
 static inline
-int amdgpu_amdkfd_remove_fence_on_pt_pd_bos(struct amdgpu_bo *bo)
+void amdgpu_amdkfd_remove_all_eviction_fences(struct amdgpu_bo *bo)
 {
-	return 0;
 }
 
 static inline
@@ -210,12 +227,19 @@ int amdgpu_amdkfd_evict_userptr(struct mmu_interval_notifier *mni,
 {
 	return 0;
 }
+static inline
+int amdgpu_amdkfd_bo_validate_and_fence(struct amdgpu_bo *bo,
+					uint32_t domain,
+					struct dma_fence *fence)
+{
+	return 0;
+}
 #endif
 /* Shared API */
 int amdgpu_amdkfd_alloc_gtt_mem(struct amdgpu_device *adev, size_t size,
 				void **mem_obj, uint64_t *gpu_addr,
 				void **cpu_ptr, bool mqd_gfx9);
-void amdgpu_amdkfd_free_gtt_mem(struct amdgpu_device *adev, void *mem_obj);
+void amdgpu_amdkfd_free_gtt_mem(struct amdgpu_device *adev, void **mem_obj);
 int amdgpu_amdkfd_alloc_gws(struct amdgpu_device *adev, size_t size,
 				void **mem_obj);
 void amdgpu_amdkfd_free_gws(struct amdgpu_device *adev, void *mem_obj);
@@ -224,23 +248,27 @@ int amdgpu_amdkfd_remove_gws_from_process(void *info, void *mem);
 uint32_t amdgpu_amdkfd_get_fw_version(struct amdgpu_device *adev,
 				      enum kgd_engine_type type);
 void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device *adev,
-				      struct kfd_local_mem_info *mem_info);
+				      struct kfd_local_mem_info *mem_info,
+				      struct amdgpu_xcp *xcp);
 uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct amdgpu_device *adev);
 
 uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct amdgpu_device *adev);
-void amdgpu_amdkfd_get_cu_info(struct amdgpu_device *adev,
-			       struct kfd_cu_info *cu_info);
 int amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int dma_buf_fd,
 				  struct amdgpu_device **dmabuf_adev,
 				  uint64_t *bo_size, void *metadata_buffer,
 				  size_t buffer_size, uint32_t *metadata_size,
-				  uint32_t *flags);
-uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct amdgpu_device *dst,
-					  struct amdgpu_device *src);
-int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct amdgpu_device *dst,
-					    struct amdgpu_device *src,
-					    bool is_min);
+				  uint32_t *flags, int8_t *xcp_id);
 int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct amdgpu_device *adev, bool is_min);
+int amdgpu_amdkfd_send_close_event_drain_irq(struct amdgpu_device *adev,
+					uint32_t *payload);
+int amdgpu_amdkfd_unmap_hiq(struct amdgpu_device *adev, u32 doorbell_off,
+				u32 inst);
+int amdgpu_amdkfd_start_sched(struct amdgpu_device *adev, uint32_t node_id);
+int amdgpu_amdkfd_stop_sched(struct amdgpu_device *adev, uint32_t node_id);
+int amdgpu_amdkfd_config_sq_perfmon(struct amdgpu_device *adev, uint32_t xcp_id,
+	bool core_override_enable, bool reg_override_enable, bool perfmon_override_enable);
+bool amdgpu_amdkfd_compute_active(struct amdgpu_device *adev, uint32_t node_id);
+
 
 /* Read user wptr from a specified user address space with page fault
  * disabled. The memory must be pinned and mapped to the hardware when
@@ -270,16 +298,13 @@ int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct amdgpu_device *adev, bool is_
 	(&((struct amdgpu_fpriv *)					\
 		((struct drm_file *)(drm_priv))->driver_priv)->vm)
 
-int amdgpu_amdkfd_gpuvm_set_vm_pasid(struct amdgpu_device *adev,
-				     struct amdgpu_vm *avm, u32 pasid);
 int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev,
 					struct amdgpu_vm *avm,
 					void **process_info,
 					struct dma_fence **ef);
-void amdgpu_amdkfd_gpuvm_release_process_vm(struct amdgpu_device *adev,
-					void *drm_priv);
 uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *drm_priv);
-size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev);
+size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev,
+					uint8_t xcp_id);
 int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 		struct amdgpu_device *adev, uint64_t va, uint64_t size,
 		void *drm_priv, struct kgd_mem **mem,
@@ -291,37 +316,52 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu(struct amdgpu_device *adev,
 					  struct kgd_mem *mem, void *drm_priv);
 int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
 		struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv);
+int amdgpu_amdkfd_gpuvm_dmaunmap_mem(struct kgd_mem *mem, void *drm_priv);
 int amdgpu_amdkfd_gpuvm_sync_memory(
 		struct amdgpu_device *adev, struct kgd_mem *mem, bool intr);
 int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_mem *mem,
 					     void **kptr, uint64_t *size);
 void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct kgd_mem *mem);
 
-int amdgpu_amdkfd_map_gtt_bo_to_gart(struct amdgpu_device *adev, struct amdgpu_bo *bo);
+int amdgpu_amdkfd_map_gtt_bo_to_gart(struct amdgpu_bo *bo, struct amdgpu_bo **bo_gart);
 
 int amdgpu_amdkfd_gpuvm_restore_process_bos(void *process_info,
-					    struct dma_fence **ef);
+					    struct dma_fence __rcu **ef);
 int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct amdgpu_device *adev,
 					      struct kfd_vm_fault_info *info);
-int amdgpu_amdkfd_gpuvm_import_dmabuf(struct amdgpu_device *adev,
-				      struct dma_buf *dmabuf,
-				      uint64_t va, void *drm_priv,
-				      struct kgd_mem **mem, uint64_t *size,
-				      uint64_t *mmap_offset);
+int amdgpu_amdkfd_gpuvm_import_dmabuf_fd(struct amdgpu_device *adev, int fd,
+					 uint64_t va, void *drm_priv,
+					 struct kgd_mem **mem, uint64_t *size,
+					 uint64_t *mmap_offset);
 int amdgpu_amdkfd_gpuvm_export_dmabuf(struct kgd_mem *mem,
 				      struct dma_buf **dmabuf);
+void amdgpu_amdkfd_debug_mem_fence(struct amdgpu_device *adev);
 int amdgpu_amdkfd_get_tile_config(struct amdgpu_device *adev,
 				struct tile_config *config);
 void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev,
-				bool reset);
-bool amdgpu_amdkfd_bo_mapped_to_dev(struct amdgpu_device *adev, struct kgd_mem *mem);
+			enum amdgpu_ras_block block, uint32_t reset);
+
+void amdgpu_amdkfd_ras_pasid_poison_consumption_handler(struct amdgpu_device *adev,
+			enum amdgpu_ras_block block, uint16_t pasid,
+			pasid_notify pasid_fn, void *data, uint32_t reset);
+
+bool amdgpu_amdkfd_is_fed(struct amdgpu_device *adev);
+bool amdgpu_amdkfd_bo_mapped_to_dev(void *drm_priv, struct kgd_mem *mem);
 void amdgpu_amdkfd_block_mmu_notifications(void *p);
 int amdgpu_amdkfd_criu_resume(void *p);
-bool amdgpu_amdkfd_ras_query_utcl2_poison_status(struct amdgpu_device *adev);
 int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
-		uint64_t size, u32 alloc_flag);
+		uint64_t size, u32 alloc_flag, int8_t xcp_id);
 void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
-		uint64_t size, u32 alloc_flag);
+		uint64_t size, u32 alloc_flag, int8_t xcp_id);
+
+u64 amdgpu_amdkfd_xcp_memory_size(struct amdgpu_device *adev, int xcp_id);
+
+#define KFD_XCP_MEM_ID(adev, xcp_id) \
+		((adev)->xcp_mgr && (xcp_id) >= 0 ?\
+		(adev)->xcp_mgr->xcp[(xcp_id)].mem_id : -1)
+
+#define KFD_XCP_MEMORY_SIZE(adev, xcp_id) amdgpu_amdkfd_xcp_memory_size((adev), (xcp_id))
+
 
 #if IS_ENABLED(CONFIG_HSA_AMD)
 void amdgpu_amdkfd_gpuvm_init_mem_limits(void);
@@ -352,6 +392,17 @@ void amdgpu_amdkfd_release_notify(struct amdgpu_bo *bo)
 {
 }
 #endif
+
+#if IS_ENABLED(CONFIG_HSA_AMD_SVM)
+int kgd2kfd_init_zone_device(struct amdgpu_device *adev);
+#else
+static inline
+int kgd2kfd_init_zone_device(struct amdgpu_device *adev)
+{
+	return 0;
+}
+#endif
+
 /* KGD2KFD callbacks */
 int kgd2kfd_quiesce_mm(struct mm_struct *mm, uint32_t trigger);
 int kgd2kfd_resume_mm(struct mm_struct *mm);
@@ -364,14 +415,26 @@ struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf);
 bool kgd2kfd_device_init(struct kfd_dev *kfd,
 			 const struct kgd2kfd_shared_resources *gpu_resources);
 void kgd2kfd_device_exit(struct kfd_dev *kfd);
-void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm);
-int kgd2kfd_resume_iommu(struct kfd_dev *kfd);
-int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm);
-int kgd2kfd_pre_reset(struct kfd_dev *kfd);
+void kgd2kfd_suspend(struct kfd_dev *kfd, bool suspend_proc);
+int kgd2kfd_resume(struct kfd_dev *kfd, bool resume_proc);
+void kgd2kfd_suspend_process(struct kfd_dev *kfd);
+int kgd2kfd_resume_process(struct kfd_dev *kfd);
+int kgd2kfd_pre_reset(struct kfd_dev *kfd,
+		      struct amdgpu_reset_context *reset_context);
 int kgd2kfd_post_reset(struct kfd_dev *kfd);
 void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry);
 void kgd2kfd_set_sram_ecc_flag(struct kfd_dev *kfd);
 void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint64_t throttle_bitmask);
+int kgd2kfd_check_and_lock_kfd(struct kfd_dev *kfd);
+void kgd2kfd_unlock_kfd(struct kfd_dev *kfd);
+int kgd2kfd_start_sched(struct kfd_dev *kfd, uint32_t node_id);
+int kgd2kfd_start_sched_all_nodes(struct kfd_dev *kfd);
+int kgd2kfd_stop_sched(struct kfd_dev *kfd, uint32_t node_id);
+int kgd2kfd_stop_sched_all_nodes(struct kfd_dev *kfd);
+bool kgd2kfd_compute_active(struct kfd_dev *kfd, uint32_t node_id);
+bool kgd2kfd_vmfault_fast_path(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry,
+			       bool retry_fault);
+
 #else
 static inline int kgd2kfd_init(void)
 {
@@ -399,21 +462,26 @@ static inline void kgd2kfd_device_exit(struct kfd_dev *kfd)
 {
 }
 
-static inline void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm)
+static inline void kgd2kfd_suspend(struct kfd_dev *kfd, bool suspend_proc)
 {
 }
 
-static int __maybe_unused kgd2kfd_resume_iommu(struct kfd_dev *kfd)
+static inline int kgd2kfd_resume(struct kfd_dev *kfd, bool resume_proc)
 {
 	return 0;
 }
 
-static inline int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm)
+static inline void kgd2kfd_suspend_process(struct kfd_dev *kfd)
+{
+}
+
+static inline int kgd2kfd_resume_process(struct kfd_dev *kfd)
 {
 	return 0;
 }
 
-static inline int kgd2kfd_pre_reset(struct kfd_dev *kfd)
+static inline int kgd2kfd_pre_reset(struct kfd_dev *kfd,
+				    struct amdgpu_reset_context *reset_context)
 {
 	return 0;
 }
@@ -437,5 +505,46 @@ static inline
 void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint64_t throttle_bitmask)
 {
 }
+
+static inline int kgd2kfd_check_and_lock_kfd(struct kfd_dev *kfd)
+{
+	return 0;
+}
+
+static inline void kgd2kfd_unlock_kfd(struct kfd_dev *kfd)
+{
+}
+
+static inline int kgd2kfd_start_sched(struct kfd_dev *kfd, uint32_t node_id)
+{
+	return 0;
+}
+
+static inline int kgd2kfd_start_sched_all_nodes(struct kfd_dev *kfd)
+{
+	return 0;
+}
+
+static inline int kgd2kfd_stop_sched(struct kfd_dev *kfd, uint32_t node_id)
+{
+	return 0;
+}
+
+static inline int kgd2kfd_stop_sched_all_nodes(struct kfd_dev *kfd)
+{
+	return 0;
+}
+
+static inline bool kgd2kfd_compute_active(struct kfd_dev *kfd, uint32_t node_id)
+{
+	return false;
+}
+
+static inline bool kgd2kfd_vmfault_fast_path(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry,
+				      bool retry_fault)
+{
+	return false;
+}
+
 #endif
 #endif /* AMDGPU_AMDKFD_H_INCLUDED */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
index 4485bb29bec9..7e9f7a280c1b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
@@ -23,6 +23,145 @@
 #include "amdgpu_amdkfd.h"
 #include "amdgpu_amdkfd_arcturus.h"
 #include "amdgpu_amdkfd_gfx_v9.h"
+#include "amdgpu_amdkfd_aldebaran.h"
+#include "gc/gc_9_4_2_offset.h"
+#include "gc/gc_9_4_2_sh_mask.h"
+#include <uapi/linux/kfd_ioctl.h>
+
+/*
+ * Returns TRAP_EN, EXCP_EN and EXCP_REPLACE.
+ *
+ * restore_dbg_registers is ignored here but is a general interface requirement
+ * for devices that support GFXOFF and where the RLC save/restore list
+ * does not support hw registers for debugging i.e. the driver has to manually
+ * initialize the debug mode registers after it has disabled GFX off during the
+ * debug session.
+ */
+uint32_t kgd_aldebaran_enable_debug_trap(struct amdgpu_device *adev,
+					    bool restore_dbg_registers,
+					    uint32_t vmid)
+{
+	uint32_t data = 0;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, 0);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, 0);
+
+	return data;
+}
+
+/* returns TRAP_EN, EXCP_EN and EXCP_REPLACE. */
+static uint32_t kgd_aldebaran_disable_debug_trap(struct amdgpu_device *adev,
+						bool keep_trap_enabled,
+						uint32_t vmid)
+{
+	uint32_t data = 0;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, keep_trap_enabled);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, 0);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, 0);
+
+	return data;
+}
+
+static int kgd_aldebaran_validate_trap_override_request(struct amdgpu_device *adev,
+							uint32_t trap_override,
+							uint32_t *trap_mask_supported)
+{
+	*trap_mask_supported &= KFD_DBG_TRAP_MASK_FP_INVALID |
+				KFD_DBG_TRAP_MASK_FP_INPUT_DENORMAL |
+				KFD_DBG_TRAP_MASK_FP_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_FP_OVERFLOW |
+				KFD_DBG_TRAP_MASK_FP_UNDERFLOW |
+				KFD_DBG_TRAP_MASK_FP_INEXACT |
+				KFD_DBG_TRAP_MASK_INT_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH |
+				KFD_DBG_TRAP_MASK_DBG_MEMORY_VIOLATION;
+
+	if (trap_override != KFD_DBG_TRAP_OVERRIDE_OR &&
+			trap_override != KFD_DBG_TRAP_OVERRIDE_REPLACE)
+		return -EPERM;
+
+	return 0;
+}
+
+/* returns TRAP_EN, EXCP_EN and EXCP_RPLACE. */
+static uint32_t kgd_aldebaran_set_wave_launch_trap_override(struct amdgpu_device *adev,
+					uint32_t vmid,
+					uint32_t trap_override,
+					uint32_t trap_mask_bits,
+					uint32_t trap_mask_request,
+					uint32_t *trap_mask_prev,
+					uint32_t kfd_dbg_trap_cntl_prev)
+
+{
+	uint32_t data = 0;
+
+	*trap_mask_prev = REG_GET_FIELD(kfd_dbg_trap_cntl_prev, SPI_GDBG_PER_VMID_CNTL, EXCP_EN);
+	trap_mask_bits = (trap_mask_bits & trap_mask_request) |
+		(*trap_mask_prev & ~trap_mask_request);
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, trap_mask_bits);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, trap_override);
+
+	return data;
+}
+
+uint32_t kgd_aldebaran_set_wave_launch_mode(struct amdgpu_device *adev,
+					uint8_t wave_launch_mode,
+					uint32_t vmid)
+{
+	uint32_t data = 0;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, LAUNCH_MODE, wave_launch_mode);
+
+	return data;
+}
+
+#define TCP_WATCH_STRIDE (regTCP_WATCH1_ADDR_H - regTCP_WATCH0_ADDR_H)
+static uint32_t kgd_gfx_aldebaran_set_address_watch(
+					struct amdgpu_device *adev,
+					uint64_t watch_address,
+					uint32_t watch_address_mask,
+					uint32_t watch_id,
+					uint32_t watch_mode,
+					uint32_t debug_vmid,
+					uint32_t inst)
+{
+	uint32_t watch_address_high;
+	uint32_t watch_address_low;
+	uint32_t watch_address_cntl;
+
+	watch_address_cntl = 0;
+	watch_address_low = lower_32_bits(watch_address);
+	watch_address_high = upper_32_bits(watch_address) & 0xffff;
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MODE,
+			watch_mode);
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MASK,
+			watch_address_mask >> 6);
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VALID,
+			1);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_H) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_high);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_L) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_low);
+
+	return watch_address_cntl;
+}
 
 const struct kfd2kgd_calls aldebaran_kfd2kgd = {
 	.program_sh_mem_settings = kgd_gfx_v9_program_sh_mem_settings,
@@ -42,5 +181,17 @@ const struct kfd2kgd_calls aldebaran_kfd2kgd = {
 				kgd_gfx_v9_get_atc_vmid_pasid_mapping_info,
 	.set_vm_context_page_table_base = kgd_gfx_v9_set_vm_context_page_table_base,
 	.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
-	.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings
+	.enable_debug_trap = kgd_aldebaran_enable_debug_trap,
+	.disable_debug_trap = kgd_aldebaran_disable_debug_trap,
+	.validate_trap_override_request = kgd_aldebaran_validate_trap_override_request,
+	.set_wave_launch_trap_override = kgd_aldebaran_set_wave_launch_trap_override,
+	.set_wave_launch_mode = kgd_aldebaran_set_wave_launch_mode,
+	.set_address_watch = kgd_gfx_aldebaran_set_address_watch,
+	.clear_address_watch = kgd_gfx_v9_clear_address_watch,
+	.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
+	.build_dequeue_wait_counts_packet_info = kgd_gfx_v9_build_dequeue_wait_counts_packet_info,
+	.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings,
+	.hqd_get_pq_addr = kgd_gfx_v9_hqd_get_pq_addr,
+	.hqd_reset = kgd_gfx_v9_hqd_reset,
+	.hqd_sdma_get_doorbell = kgd_gfx_v9_hqd_sdma_get_doorbell
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.h
new file mode 100644
index 000000000000..a7bdaf8d82dd
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.h
@@ -0,0 +1,27 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+uint32_t kgd_aldebaran_enable_debug_trap(struct amdgpu_device *adev,
+					bool restore_dbg_registers,
+					uint32_t vmid);
+uint32_t kgd_aldebaran_set_wave_launch_mode(struct amdgpu_device *adev,
+					uint8_t wave_launch_mode,
+					uint32_t vmid);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
index 4191af5a3f13..1105a09e55dc 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
@@ -20,12 +20,12 @@
  * OTHER DEALINGS IN THE SOFTWARE.
  */
 #include <linux/module.h>
-#include <linux/fdtable.h>
 #include <linux/uaccess.h>
 #include <linux/firmware.h>
 #include "amdgpu.h"
 #include "amdgpu_amdkfd.h"
 #include "amdgpu_amdkfd_arcturus.h"
+#include "amdgpu_reset.h"
 #include "sdma0/sdma0_4_2_2_offset.h"
 #include "sdma0/sdma0_4_2_2_sh_mask.h"
 #include "sdma1/sdma1_4_2_2_offset.h"
@@ -48,6 +48,8 @@
 #include "amdgpu_amdkfd_gfx_v9.h"
 #include "gfxhub_v1_0.h"
 #include "mmhub_v9_4.h"
+#include "gc/gc_9_0_offset.h"
+#include "gc/gc_9_0_sh_mask.h"
 
 #define HQD_N_REGS 56
 #define DUMP_REG(addr) do {				\
@@ -197,7 +199,7 @@ int kgd_arcturus_hqd_sdma_dump(struct amdgpu_device *adev,
 #undef HQD_N_REGS
 #define HQD_N_REGS (19+6+7+10)
 
-	*dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -276,6 +278,117 @@ int kgd_arcturus_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd,
 	return 0;
 }
 
+/*
+ * Helper used to suspend/resume gfx pipe for image post process work to set
+ * barrier behaviour.
+ */
+static int suspend_resume_compute_scheduler(struct amdgpu_device *adev, bool suspend)
+{
+	int i, r = 0;
+
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		struct amdgpu_ring *ring = &adev->gfx.compute_ring[i];
+
+		if (!amdgpu_ring_sched_ready(ring))
+			continue;
+
+		/* stop secheduler and drain ring. */
+		if (suspend) {
+			drm_sched_stop(&ring->sched, NULL);
+			r = amdgpu_fence_wait_empty(ring);
+			if (r)
+				goto out;
+		} else {
+			drm_sched_start(&ring->sched, 0);
+		}
+	}
+
+out:
+	/* return on resume or failure to drain rings. */
+	if (!suspend || r)
+		return r;
+
+	return amdgpu_device_ip_wait_for_idle(adev, AMD_IP_BLOCK_TYPE_GFX);
+}
+
+static void set_barrier_auto_waitcnt(struct amdgpu_device *adev, bool enable_waitcnt)
+{
+	uint32_t data;
+
+	WRITE_ONCE(adev->barrier_has_auto_waitcnt, enable_waitcnt);
+
+	if (!down_read_trylock(&adev->reset_domain->sem))
+		return;
+
+	amdgpu_amdkfd_suspend(adev, true);
+
+	if (suspend_resume_compute_scheduler(adev, true))
+		goto out;
+
+	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_CONFIG));
+	data = REG_SET_FIELD(data, SQ_CONFIG, DISABLE_BARRIER_WAITCNT,
+						!enable_waitcnt);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_CONFIG), data);
+
+out:
+	suspend_resume_compute_scheduler(adev, false);
+
+	amdgpu_amdkfd_resume(adev, true);
+
+	up_read(&adev->reset_domain->sem);
+}
+
+/*
+ * restore_dbg_registers is ignored here but is a general interface requirement
+ * for devices that support GFXOFF and where the RLC save/restore list
+ * does not support hw registers for debugging i.e. the driver has to manually
+ * initialize the debug mode registers after it has disabled GFX off during the
+ * debug session.
+ */
+static uint32_t kgd_arcturus_enable_debug_trap(struct amdgpu_device *adev,
+				bool restore_dbg_registers,
+				uint32_t vmid)
+{
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+	set_barrier_auto_waitcnt(adev, true);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+/*
+ * keep_trap_enabled is ignored here but is a general interface requirement
+ * for devices that support multi-process debugging where the performance
+ * overhead from trap temporary setup needs to be bypassed when the debug
+ * session has ended.
+ */
+static uint32_t kgd_arcturus_disable_debug_trap(struct amdgpu_device *adev,
+					bool keep_trap_enabled,
+					uint32_t vmid)
+{
+
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+	set_barrier_auto_waitcnt(adev, false);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
 const struct kfd2kgd_calls arcturus_kfd2kgd = {
 	.program_sh_mem_settings = kgd_gfx_v9_program_sh_mem_settings,
 	.set_pasid_vmid_mapping = kgd_gfx_v9_set_pasid_vmid_mapping,
@@ -294,6 +407,18 @@ const struct kfd2kgd_calls arcturus_kfd2kgd = {
 				kgd_gfx_v9_get_atc_vmid_pasid_mapping_info,
 	.set_vm_context_page_table_base =
 				kgd_gfx_v9_set_vm_context_page_table_base,
+	.enable_debug_trap = kgd_arcturus_enable_debug_trap,
+	.disable_debug_trap = kgd_arcturus_disable_debug_trap,
+	.validate_trap_override_request = kgd_gfx_v9_validate_trap_override_request,
+	.set_wave_launch_trap_override = kgd_gfx_v9_set_wave_launch_trap_override,
+	.set_wave_launch_mode = kgd_gfx_v9_set_wave_launch_mode,
+	.set_address_watch = kgd_gfx_v9_set_address_watch,
+	.clear_address_watch = kgd_gfx_v9_clear_address_watch,
+	.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
+	.build_dequeue_wait_counts_packet_info = kgd_gfx_v9_build_dequeue_wait_counts_packet_info,
 	.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
-	.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings
+	.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings,
+	.hqd_get_pq_addr = kgd_gfx_v9_hqd_get_pq_addr,
+	.hqd_reset = kgd_gfx_v9_hqd_reset,
+	.hqd_sdma_get_doorbell = kgd_gfx_v9_hqd_sdma_get_doorbell
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c
index 469785d33791..1ef758ac5076 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c
@@ -90,7 +90,7 @@ struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f)
 		return NULL;
 
 	fence = container_of(f, struct amdgpu_amdkfd_fence, base);
-	if (fence && f->ops == &amdkfd_fence_ops)
+	if (f->ops == &amdkfd_fence_ops)
 		return fence;
 
 	return NULL;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c
new file mode 100644
index 000000000000..89a45a9218f3
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c
@@ -0,0 +1,559 @@
+/*
+ * Copyright 2021 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+#include "amdgpu.h"
+#include "amdgpu_amdkfd.h"
+#include "amdgpu_amdkfd_gfx_v9.h"
+#include "amdgpu_amdkfd_aldebaran.h"
+#include "gc/gc_9_4_3_offset.h"
+#include "gc/gc_9_4_3_sh_mask.h"
+#include "athub/athub_1_8_0_offset.h"
+#include "athub/athub_1_8_0_sh_mask.h"
+#include "oss/osssys_4_4_2_offset.h"
+#include "oss/osssys_4_4_2_sh_mask.h"
+#include "v9_structs.h"
+#include "soc15.h"
+#include "sdma/sdma_4_4_2_offset.h"
+#include "sdma/sdma_4_4_2_sh_mask.h"
+#include <uapi/linux/kfd_ioctl.h>
+
+static inline struct v9_sdma_mqd *get_sdma_mqd(void *mqd)
+{
+	return (struct v9_sdma_mqd *)mqd;
+}
+
+static uint32_t get_sdma_rlc_reg_offset(struct amdgpu_device *adev,
+					unsigned int engine_id,
+					unsigned int queue_id)
+{
+	uint32_t sdma_engine_reg_base =
+		SOC15_REG_OFFSET(SDMA0, GET_INST(SDMA0, engine_id),
+				 regSDMA_RLC0_RB_CNTL) -
+		regSDMA_RLC0_RB_CNTL;
+	uint32_t retval = sdma_engine_reg_base +
+		  queue_id * (regSDMA_RLC1_RB_CNTL - regSDMA_RLC0_RB_CNTL);
+
+	pr_debug("RLC register offset for SDMA%d RLC%d: 0x%x\n", engine_id,
+							queue_id, retval);
+	return retval;
+}
+
+static int kgd_gfx_v9_4_3_hqd_sdma_load(struct amdgpu_device *adev, void *mqd,
+				 uint32_t __user *wptr, struct mm_struct *mm)
+{
+	struct v9_sdma_mqd *m;
+	uint32_t sdma_rlc_reg_offset;
+	unsigned long end_jiffies;
+	uint32_t data;
+	uint64_t data64;
+	uint64_t __user *wptr64 = (uint64_t __user *)wptr;
+
+	m = get_sdma_mqd(mqd);
+	sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev, m->sdma_engine_id,
+							m->sdma_queue_id);
+
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_CNTL,
+		m->sdmax_rlcx_rb_cntl & (~SDMA_RLC0_RB_CNTL__RB_ENABLE_MASK));
+
+	end_jiffies = msecs_to_jiffies(2000) + jiffies;
+	while (true) {
+		data = RREG32(sdma_rlc_reg_offset + regSDMA_RLC0_CONTEXT_STATUS);
+		if (data & SDMA_RLC0_CONTEXT_STATUS__IDLE_MASK)
+			break;
+		if (time_after(jiffies, end_jiffies)) {
+			pr_err("SDMA RLC not idle in %s\n", __func__);
+			return -ETIME;
+		}
+		usleep_range(500, 1000);
+	}
+
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_DOORBELL_OFFSET,
+		m->sdmax_rlcx_doorbell_offset);
+
+	data = REG_SET_FIELD(m->sdmax_rlcx_doorbell, SDMA_RLC0_DOORBELL,
+				ENABLE, 1);
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_DOORBELL, data);
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_RPTR,
+					m->sdmax_rlcx_rb_rptr);
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_RPTR_HI,
+					m->sdmax_rlcx_rb_rptr_hi);
+
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_MINOR_PTR_UPDATE, 1);
+	if (read_user_wptr(mm, wptr64, data64)) {
+		WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_WPTR,
+			lower_32_bits(data64));
+		WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_WPTR_HI,
+			upper_32_bits(data64));
+	} else {
+		WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_WPTR,
+			m->sdmax_rlcx_rb_rptr);
+		WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_WPTR_HI,
+			m->sdmax_rlcx_rb_rptr_hi);
+	}
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_MINOR_PTR_UPDATE, 0);
+
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_BASE, m->sdmax_rlcx_rb_base);
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_BASE_HI,
+			m->sdmax_rlcx_rb_base_hi);
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_RPTR_ADDR_LO,
+			m->sdmax_rlcx_rb_rptr_addr_lo);
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_RPTR_ADDR_HI,
+			m->sdmax_rlcx_rb_rptr_addr_hi);
+
+	data = REG_SET_FIELD(m->sdmax_rlcx_rb_cntl, SDMA_RLC0_RB_CNTL,
+				RB_ENABLE, 1);
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_CNTL, data);
+
+	return 0;
+}
+
+static int kgd_gfx_v9_4_3_hqd_sdma_dump(struct amdgpu_device *adev,
+				 uint32_t engine_id, uint32_t queue_id,
+				 uint32_t (**dump)[2], uint32_t *n_regs)
+{
+	uint32_t sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev,
+							engine_id, queue_id);
+	uint32_t i = 0, reg;
+#undef HQD_N_REGS
+#define HQD_N_REGS (19+6+7+12)
+#define DUMP_REG(addr) do {				\
+		if (WARN_ON_ONCE(i >= HQD_N_REGS))      \
+			break;				\
+		(*dump)[i][0] = (addr) << 2;            \
+		(*dump)[i++][1] = RREG32(addr);         \
+	} while (0)
+
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
+	if (*dump == NULL)
+		return -ENOMEM;
+
+	for (reg = regSDMA_RLC0_RB_CNTL; reg <= regSDMA_RLC0_DOORBELL; reg++)
+		DUMP_REG(sdma_rlc_reg_offset + reg);
+	for (reg = regSDMA_RLC0_STATUS; reg <= regSDMA_RLC0_CSA_ADDR_HI; reg++)
+		DUMP_REG(sdma_rlc_reg_offset + reg);
+	for (reg = regSDMA_RLC0_IB_SUB_REMAIN;
+	     reg <= regSDMA_RLC0_MINOR_PTR_UPDATE; reg++)
+		DUMP_REG(sdma_rlc_reg_offset + reg);
+	for (reg = regSDMA_RLC0_MIDCMD_DATA0;
+	     reg <= regSDMA_RLC0_MIDCMD_CNTL; reg++)
+		DUMP_REG(sdma_rlc_reg_offset + reg);
+
+	WARN_ON_ONCE(i != HQD_N_REGS);
+	*n_regs = i;
+
+	return 0;
+}
+
+static bool kgd_gfx_v9_4_3_hqd_sdma_is_occupied(struct amdgpu_device *adev, void *mqd)
+{
+	struct v9_sdma_mqd *m;
+	uint32_t sdma_rlc_reg_offset;
+	uint32_t sdma_rlc_rb_cntl;
+
+	m = get_sdma_mqd(mqd);
+	sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev, m->sdma_engine_id,
+							m->sdma_queue_id);
+
+	sdma_rlc_rb_cntl = RREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_CNTL);
+
+	if (sdma_rlc_rb_cntl & SDMA_RLC0_RB_CNTL__RB_ENABLE_MASK)
+		return true;
+
+	return false;
+}
+
+static int kgd_gfx_v9_4_3_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd,
+				    unsigned int utimeout)
+{
+	struct v9_sdma_mqd *m;
+	uint32_t sdma_rlc_reg_offset;
+	uint32_t temp;
+	unsigned long end_jiffies = (utimeout * HZ / 1000) + jiffies;
+
+	m = get_sdma_mqd(mqd);
+	sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev, m->sdma_engine_id,
+							m->sdma_queue_id);
+
+	temp = RREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_CNTL);
+	temp = temp & ~SDMA_RLC0_RB_CNTL__RB_ENABLE_MASK;
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_CNTL, temp);
+
+	while (true) {
+		temp = RREG32(sdma_rlc_reg_offset + regSDMA_RLC0_CONTEXT_STATUS);
+		if (temp & SDMA_RLC0_CONTEXT_STATUS__IDLE_MASK)
+			break;
+		if (time_after(jiffies, end_jiffies)) {
+			pr_err("SDMA RLC not idle in %s\n", __func__);
+			return -ETIME;
+		}
+		usleep_range(500, 1000);
+	}
+
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_DOORBELL, 0);
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_CNTL,
+		RREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_CNTL) |
+		SDMA_RLC0_RB_CNTL__RB_ENABLE_MASK);
+
+	m->sdmax_rlcx_rb_rptr =
+			RREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_RPTR);
+	m->sdmax_rlcx_rb_rptr_hi =
+			RREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_RPTR_HI);
+
+	return 0;
+}
+
+static int kgd_gfx_v9_4_3_set_pasid_vmid_mapping(struct amdgpu_device *adev,
+			u32 pasid, unsigned int vmid, uint32_t xcc_inst)
+{
+	unsigned long timeout;
+	unsigned int reg;
+	unsigned int phy_inst = GET_INST(GC, xcc_inst);
+	/* Every two XCCs share one AID */
+	unsigned int aid = phy_inst / 2;
+
+	/*
+	 * We have to assume that there is no outstanding mapping.
+	 * The ATC_VMID_PASID_MAPPING_UPDATE_STATUS bit could be 0 because
+	 * a mapping is in progress or because a mapping finished
+	 * and the SW cleared it.
+	 * So the protocol is to always wait & clear.
+	 */
+	uint32_t pasid_mapping = (pasid == 0) ? 0 : (uint32_t)pasid |
+			ATC_VMID0_PASID_MAPPING__VALID_MASK;
+
+	WREG32(SOC15_REG_OFFSET(ATHUB, 0,
+		regATC_VMID0_PASID_MAPPING) + vmid, pasid_mapping);
+
+	timeout = jiffies + msecs_to_jiffies(10);
+	while (!(RREG32(SOC15_REG_OFFSET(ATHUB, 0,
+			regATC_VMID_PASID_MAPPING_UPDATE_STATUS)) &
+			(1U << vmid))) {
+		if (time_after(jiffies, timeout)) {
+			pr_err("Fail to program VMID-PASID mapping\n");
+			return -ETIME;
+		}
+		cpu_relax();
+	}
+
+	WREG32(SOC15_REG_OFFSET(ATHUB, 0,
+		regATC_VMID_PASID_MAPPING_UPDATE_STATUS),
+		1U << vmid);
+
+	reg = RREG32(SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_LUT_INDEX));
+	/* Every 4 numbers is a cycle. 1st is AID, 2nd and 3rd are XCDs,
+	 * and the 4th is reserved. Therefore "aid * 4 + (xcc_inst % 2) + 1"
+	 * programs _LUT for XCC and "aid * 4" for AID where the XCC connects
+	 * to.
+	 */
+	WREG32(SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_LUT_INDEX),
+		aid * 4 + (phy_inst % 2) + 1);
+	WREG32(SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT) + vmid,
+		pasid_mapping);
+	WREG32(SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_LUT_INDEX),
+		aid * 4);
+	WREG32(SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT_MM) + vmid,
+		pasid_mapping);
+	WREG32(SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_LUT_INDEX), reg);
+
+	return 0;
+}
+
+static inline struct v9_mqd *get_mqd(void *mqd)
+{
+	return (struct v9_mqd *)mqd;
+}
+
+static int kgd_gfx_v9_4_3_hqd_load(struct amdgpu_device *adev, void *mqd,
+			uint32_t pipe_id, uint32_t queue_id,
+			uint32_t __user *wptr, uint32_t wptr_shift,
+			uint32_t wptr_mask, struct mm_struct *mm, uint32_t inst)
+{
+	struct v9_mqd *m;
+	uint32_t *mqd_hqd;
+	uint32_t reg, hqd_base, hqd_end, data;
+
+	m = get_mqd(mqd);
+
+	kgd_gfx_v9_acquire_queue(adev, pipe_id, queue_id, inst);
+
+	/* HQD registers extend to CP_HQD_AQL_DISPATCH_ID_HI */
+	mqd_hqd = &m->cp_mqd_base_addr_lo;
+	hqd_base = SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_MQD_BASE_ADDR);
+	hqd_end = SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_HQD_AQL_DISPATCH_ID_HI);
+
+	for (reg = hqd_base; reg <= hqd_end; reg++)
+		WREG32_XCC(reg, mqd_hqd[reg - hqd_base], inst);
+
+
+	/* Activate doorbell logic before triggering WPTR poll. */
+	data = REG_SET_FIELD(m->cp_hqd_pq_doorbell_control,
+			     CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, inst), regCP_HQD_PQ_DOORBELL_CONTROL, data);
+
+	if (wptr) {
+		/* Don't read wptr with get_user because the user
+		 * context may not be accessible (if this function
+		 * runs in a work queue). Instead trigger a one-shot
+		 * polling read from memory in the CP. This assumes
+		 * that wptr is GPU-accessible in the queue's VMID via
+		 * ATC or SVM. WPTR==RPTR before starting the poll so
+		 * the CP starts fetching new commands from the right
+		 * place.
+		 *
+		 * Guessing a 64-bit WPTR from a 32-bit RPTR is a bit
+		 * tricky. Assume that the queue didn't overflow. The
+		 * number of valid bits in the 32-bit RPTR depends on
+		 * the queue size. The remaining bits are taken from
+		 * the saved 64-bit WPTR. If the WPTR wrapped, add the
+		 * queue size.
+		 */
+		uint32_t queue_size =
+			2 << REG_GET_FIELD(m->cp_hqd_pq_control,
+					   CP_HQD_PQ_CONTROL, QUEUE_SIZE);
+		uint64_t guessed_wptr = m->cp_hqd_pq_rptr & (queue_size - 1);
+
+		if ((m->cp_hqd_pq_wptr_lo & (queue_size - 1)) < guessed_wptr)
+			guessed_wptr += queue_size;
+		guessed_wptr += m->cp_hqd_pq_wptr_lo & ~(queue_size - 1);
+		guessed_wptr += (uint64_t)m->cp_hqd_pq_wptr_hi << 32;
+
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), regCP_HQD_PQ_WPTR_LO,
+			lower_32_bits(guessed_wptr));
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), regCP_HQD_PQ_WPTR_HI,
+			upper_32_bits(guessed_wptr));
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), regCP_HQD_PQ_WPTR_POLL_ADDR,
+			lower_32_bits((uintptr_t)wptr));
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), regCP_HQD_PQ_WPTR_POLL_ADDR_HI,
+			upper_32_bits((uintptr_t)wptr));
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), regCP_PQ_WPTR_POLL_CNTL1,
+			(uint32_t)kgd_gfx_v9_get_queue_mask(adev, pipe_id, queue_id));
+	}
+
+	/* Start the EOP fetcher */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, inst), regCP_HQD_EOP_RPTR,
+	       REG_SET_FIELD(m->cp_hqd_eop_rptr, CP_HQD_EOP_RPTR, INIT_FETCHER, 1));
+
+	data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, inst), regCP_HQD_ACTIVE, data);
+
+	kgd_gfx_v9_release_queue(adev, inst);
+
+	return 0;
+}
+
+/* returns TRAP_EN, EXCP_EN and EXCP_REPLACE. */
+static uint32_t kgd_gfx_v9_4_3_disable_debug_trap(struct amdgpu_device *adev,
+						bool keep_trap_enabled,
+						uint32_t vmid)
+{
+	uint32_t data = 0;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, 0);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, 0);
+
+	return data;
+}
+
+static int kgd_gfx_v9_4_3_validate_trap_override_request(
+				struct amdgpu_device *adev,
+				uint32_t trap_override,
+				uint32_t *trap_mask_supported)
+{
+	*trap_mask_supported &= KFD_DBG_TRAP_MASK_FP_INVALID |
+				KFD_DBG_TRAP_MASK_FP_INPUT_DENORMAL |
+				KFD_DBG_TRAP_MASK_FP_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_FP_OVERFLOW |
+				KFD_DBG_TRAP_MASK_FP_UNDERFLOW |
+				KFD_DBG_TRAP_MASK_FP_INEXACT |
+				KFD_DBG_TRAP_MASK_INT_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH |
+				KFD_DBG_TRAP_MASK_DBG_MEMORY_VIOLATION |
+				KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START |
+				KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END;
+
+	if (trap_override != KFD_DBG_TRAP_OVERRIDE_OR &&
+			trap_override != KFD_DBG_TRAP_OVERRIDE_REPLACE)
+		return -EPERM;
+
+	return 0;
+}
+
+static uint32_t trap_mask_map_sw_to_hw(uint32_t mask)
+{
+	uint32_t trap_on_start = (mask & KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START) ? 1 : 0;
+	uint32_t trap_on_end = (mask & KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END) ? 1 : 0;
+	uint32_t excp_en = mask & (KFD_DBG_TRAP_MASK_FP_INVALID |
+				KFD_DBG_TRAP_MASK_FP_INPUT_DENORMAL |
+				KFD_DBG_TRAP_MASK_FP_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_FP_OVERFLOW |
+				KFD_DBG_TRAP_MASK_FP_UNDERFLOW |
+				KFD_DBG_TRAP_MASK_FP_INEXACT |
+				KFD_DBG_TRAP_MASK_INT_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH |
+				KFD_DBG_TRAP_MASK_DBG_MEMORY_VIOLATION);
+	uint32_t ret;
+
+	ret = REG_SET_FIELD(0, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, excp_en);
+	ret = REG_SET_FIELD(ret, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_START, trap_on_start);
+	ret = REG_SET_FIELD(ret, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_END, trap_on_end);
+
+	return ret;
+}
+
+static uint32_t trap_mask_map_hw_to_sw(uint32_t mask)
+{
+	uint32_t ret = REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, EXCP_EN);
+
+	if (REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_START))
+		ret |= KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START;
+
+	if (REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_END))
+		ret |= KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END;
+
+	return ret;
+}
+
+/* returns TRAP_EN, EXCP_EN and EXCP_REPLACE. */
+static uint32_t kgd_gfx_v9_4_3_set_wave_launch_trap_override(
+				struct amdgpu_device *adev,
+				uint32_t vmid,
+				uint32_t trap_override,
+				uint32_t trap_mask_bits,
+				uint32_t trap_mask_request,
+				uint32_t *trap_mask_prev,
+				uint32_t kfd_dbg_trap_cntl_prev)
+
+{
+	uint32_t data = 0;
+
+	*trap_mask_prev = trap_mask_map_hw_to_sw(kfd_dbg_trap_cntl_prev);
+
+	data = (trap_mask_bits & trap_mask_request) |
+	       (*trap_mask_prev & ~trap_mask_request);
+	data = trap_mask_map_sw_to_hw(data);
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, trap_override);
+
+	return data;
+}
+
+#define TCP_WATCH_STRIDE (regTCP_WATCH1_ADDR_H - regTCP_WATCH0_ADDR_H)
+static uint32_t kgd_gfx_v9_4_3_set_address_watch(
+				struct amdgpu_device *adev,
+				uint64_t watch_address,
+				uint32_t watch_address_mask,
+				uint32_t watch_id,
+				uint32_t watch_mode,
+				uint32_t debug_vmid,
+				uint32_t inst)
+{
+	uint32_t watch_address_high;
+	uint32_t watch_address_low;
+	uint32_t watch_address_cntl;
+
+	watch_address_cntl = 0;
+	watch_address_low = lower_32_bits(watch_address);
+	watch_address_high = upper_32_bits(watch_address) & 0xffff;
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MODE,
+			watch_mode);
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MASK,
+			watch_address_mask >> 7);
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VALID,
+			1);
+
+	WREG32_XCC((SOC15_REG_OFFSET(GC, GET_INST(GC, inst),
+			regTCP_WATCH0_ADDR_H) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_high, inst);
+
+	WREG32_XCC((SOC15_REG_OFFSET(GC, GET_INST(GC, inst),
+			regTCP_WATCH0_ADDR_L) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_low, inst);
+
+	return watch_address_cntl;
+}
+
+static uint32_t kgd_gfx_v9_4_3_clear_address_watch(struct amdgpu_device *adev,
+				uint32_t watch_id)
+{
+	return 0;
+}
+
+static uint32_t kgd_gfx_v9_4_3_hqd_sdma_get_doorbell(struct amdgpu_device *adev,
+						     int engine, int queue)
+{
+	uint32_t reg_offset = get_sdma_rlc_reg_offset(adev, engine, queue);
+	uint32_t status = RREG32(regSDMA_RLC0_CONTEXT_STATUS + reg_offset);
+	uint32_t doorbell_off = RREG32(regSDMA_RLC0_DOORBELL_OFFSET + reg_offset);
+	bool is_active = !!REG_GET_FIELD(status, SDMA_RLC0_CONTEXT_STATUS, SELECTED);
+
+	return is_active ? doorbell_off >> 2 : 0;
+}
+
+const struct kfd2kgd_calls gc_9_4_3_kfd2kgd = {
+	.program_sh_mem_settings = kgd_gfx_v9_program_sh_mem_settings,
+	.set_pasid_vmid_mapping = kgd_gfx_v9_4_3_set_pasid_vmid_mapping,
+	.init_interrupts = kgd_gfx_v9_init_interrupts,
+	.hqd_load = kgd_gfx_v9_4_3_hqd_load,
+	.hiq_mqd_load = kgd_gfx_v9_hiq_mqd_load,
+	.hqd_sdma_load = kgd_gfx_v9_4_3_hqd_sdma_load,
+	.hqd_dump = kgd_gfx_v9_hqd_dump,
+	.hqd_sdma_dump = kgd_gfx_v9_4_3_hqd_sdma_dump,
+	.hqd_is_occupied = kgd_gfx_v9_hqd_is_occupied,
+	.hqd_sdma_is_occupied = kgd_gfx_v9_4_3_hqd_sdma_is_occupied,
+	.hqd_destroy = kgd_gfx_v9_hqd_destroy,
+	.hqd_sdma_destroy = kgd_gfx_v9_4_3_hqd_sdma_destroy,
+	.wave_control_execute = kgd_gfx_v9_wave_control_execute,
+	.get_atc_vmid_pasid_mapping_info =
+				kgd_gfx_v9_get_atc_vmid_pasid_mapping_info,
+	.set_vm_context_page_table_base =
+				kgd_gfx_v9_set_vm_context_page_table_base,
+	.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
+	.program_trap_handler_settings =
+				kgd_gfx_v9_program_trap_handler_settings,
+	.build_dequeue_wait_counts_packet_info =
+				kgd_gfx_v9_build_dequeue_wait_counts_packet_info,
+	.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
+	.enable_debug_trap = kgd_aldebaran_enable_debug_trap,
+	.disable_debug_trap = kgd_gfx_v9_4_3_disable_debug_trap,
+	.validate_trap_override_request =
+			kgd_gfx_v9_4_3_validate_trap_override_request,
+	.set_wave_launch_trap_override =
+			kgd_gfx_v9_4_3_set_wave_launch_trap_override,
+	.set_wave_launch_mode = kgd_aldebaran_set_wave_launch_mode,
+	.set_address_watch = kgd_gfx_v9_4_3_set_address_watch,
+	.clear_address_watch = kgd_gfx_v9_4_3_clear_address_watch,
+	.hqd_get_pq_addr = kgd_gfx_v9_hqd_get_pq_addr,
+	.hqd_reset = kgd_gfx_v9_hqd_reset,
+	.hqd_sdma_get_doorbell = kgd_gfx_v9_4_3_hqd_sdma_get_doorbell
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
index 9378fc79e9ea..0239114fb6c4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
@@ -21,6 +21,7 @@
  */
 #include "amdgpu.h"
 #include "amdgpu_amdkfd.h"
+#include "amdgpu_amdkfd_gfx_v10.h"
 #include "gc/gc_10_1_0_offset.h"
 #include "gc/gc_10_1_0_sh_mask.h"
 #include "athub/athub_2_0_0_offset.h"
@@ -31,6 +32,7 @@
 #include "v10_structs.h"
 #include "nv.h"
 #include "nvd.h"
+#include <uapi/linux/kfd_ioctl.h>
 
 enum hqd_dequeue_request_type {
 	NO_ACTION = 0,
@@ -79,7 +81,7 @@ static void kgd_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmi
 					uint32_t sh_mem_config,
 					uint32_t sh_mem_ape1_base,
 					uint32_t sh_mem_ape1_limit,
-					uint32_t sh_mem_bases)
+					uint32_t sh_mem_bases, uint32_t inst)
 {
 	lock_srbm(adev, 0, 0, 0, vmid);
 
@@ -91,7 +93,7 @@ static void kgd_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmi
 }
 
 static int kgd_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
-					unsigned int vmid)
+					unsigned int vmid, uint32_t inst)
 {
 	/*
 	 * We have to assume that there is no outstanding mapping.
@@ -135,7 +137,8 @@ static int kgd_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
  * but still works
  */
 
-static int kgd_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id)
+static int kgd_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t inst)
 {
 	uint32_t mec;
 	uint32_t pipe;
@@ -205,7 +208,7 @@ static inline struct v10_sdma_mqd *get_sdma_mqd(void *mqd)
 static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd,
 			uint32_t pipe_id, uint32_t queue_id,
 			uint32_t __user *wptr, uint32_t wptr_shift,
-			uint32_t wptr_mask, struct mm_struct *mm)
+			uint32_t wptr_mask, struct mm_struct *mm, uint32_t inst)
 {
 	struct v10_compute_mqd *m;
 	uint32_t *mqd_hqd;
@@ -286,9 +289,9 @@ static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd,
 
 static int kgd_hiq_mqd_load(struct amdgpu_device *adev, void *mqd,
 			    uint32_t pipe_id, uint32_t queue_id,
-			    uint32_t doorbell_off)
+			    uint32_t doorbell_off, uint32_t inst)
 {
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
+	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq[0].ring;
 	struct v10_compute_mqd *m;
 	uint32_t mec, pipe;
 	int r;
@@ -303,7 +306,7 @@ static int kgd_hiq_mqd_load(struct amdgpu_device *adev, void *mqd,
 	pr_debug("kfd: set HIQ, mec:%d, pipe:%d, queue:%d.\n",
 		 mec, pipe, queue_id);
 
-	spin_lock(&adev->gfx.kiq.ring_lock);
+	spin_lock(&adev->gfx.kiq[0].ring_lock);
 	r = amdgpu_ring_alloc(kiq_ring, 7);
 	if (r) {
 		pr_err("Failed to alloc KIQ (%d).\n", r);
@@ -330,7 +333,7 @@ static int kgd_hiq_mqd_load(struct amdgpu_device *adev, void *mqd,
 	amdgpu_ring_commit(kiq_ring);
 
 out_unlock:
-	spin_unlock(&adev->gfx.kiq.ring_lock);
+	spin_unlock(&adev->gfx.kiq[0].ring_lock);
 	release_queue(adev);
 
 	return r;
@@ -338,7 +341,7 @@ out_unlock:
 
 static int kgd_hqd_dump(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
-			uint32_t (**dump)[2], uint32_t *n_regs)
+			uint32_t (**dump)[2], uint32_t *n_regs, uint32_t inst)
 {
 	uint32_t i = 0, reg;
 #define HQD_N_REGS 56
@@ -349,7 +352,7 @@ static int kgd_hqd_dump(struct amdgpu_device *adev,
 		(*dump)[i++][1] = RREG32_SOC15_IP(GC, addr);		\
 	} while (0)
 
-	*dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -446,7 +449,7 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 #undef HQD_N_REGS
 #define HQD_N_REGS (19+6+7+10)
 
-	*dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -469,7 +472,7 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 
 static bool kgd_hqd_is_occupied(struct amdgpu_device *adev,
 				uint64_t queue_address, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	uint32_t act;
 	bool retval = false;
@@ -510,7 +513,7 @@ static bool kgd_hqd_sdma_is_occupied(struct amdgpu_device *adev, void *mqd)
 static int kgd_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	enum hqd_dequeue_request_type type;
 	unsigned long end_jiffies;
@@ -673,7 +676,7 @@ static bool get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev,
 
 static int kgd_wave_control_execute(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
-					uint32_t sq_cmd)
+					uint32_t sq_cmd, uint32_t inst)
 {
 	uint32_t data = 0;
 
@@ -708,8 +711,342 @@ static void set_vm_context_page_table_base(struct amdgpu_device *adev,
 	adev->gfxhub.funcs->setup_vm_pt_regs(adev, vmid, page_table_base);
 }
 
+/*
+ * GFX10 helper for wave launch stall requirements on debug trap setting.
+ *
+ * vmid:
+ *   Target VMID to stall/unstall.
+ *
+ * stall:
+ *   0-unstall wave launch (enable), 1-stall wave launch (disable).
+ *   After wavefront launch has been stalled, allocated waves must drain from
+ *   SPI in order for debug trap settings to take effect on those waves.
+ *   This is roughly a ~3500 clock cycle wait on SPI where a read on
+ *   SPI_GDBG_WAVE_CNTL translates to ~32 clock cycles.
+ *   KGD_GFX_V10_WAVE_LAUNCH_SPI_DRAIN_LATENCY indicates the number of reads required.
+ *
+ *   NOTE: We can afford to clear the entire STALL_VMID field on unstall
+ *   because current GFX10 chips cannot support multi-process debugging due to
+ *   trap configuration and masking being limited to global scope.  Always
+ *   assume single process conditions.
+ *
+ */
+
+#define KGD_GFX_V10_WAVE_LAUNCH_SPI_DRAIN_LATENCY	110
+static void kgd_gfx_v10_set_wave_launch_stall(struct amdgpu_device *adev, uint32_t vmid, bool stall)
+{
+	uint32_t data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+	int i;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_VMID,
+							stall ? 1 << vmid : 0);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);
+
+	if (!stall)
+		return;
+
+	for (i = 0; i < KGD_GFX_V10_WAVE_LAUNCH_SPI_DRAIN_LATENCY; i++)
+		RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+}
+
+uint32_t kgd_gfx_v10_enable_debug_trap(struct amdgpu_device *adev,
+				bool restore_dbg_registers,
+				uint32_t vmid)
+{
+
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	kgd_gfx_v10_set_wave_launch_stall(adev, vmid, true);
+
+	/* assume gfx off is disabled for the debug session if rlc restore not supported. */
+	if (restore_dbg_registers) {
+		uint32_t data = 0;
+
+		data = REG_SET_FIELD(data, SPI_GDBG_TRAP_CONFIG,
+				VMID_SEL, 1 << vmid);
+		data = REG_SET_FIELD(data, SPI_GDBG_TRAP_CONFIG,
+				TRAP_EN, 1);
+		WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_CONFIG), data);
+		WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_DATA0), 0);
+		WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_DATA1), 0);
+
+		kgd_gfx_v10_set_wave_launch_stall(adev, vmid, false);
+
+		mutex_unlock(&adev->grbm_idx_mutex);
+
+		return 0;
+	}
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+	kgd_gfx_v10_set_wave_launch_stall(adev, vmid, false);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+uint32_t kgd_gfx_v10_disable_debug_trap(struct amdgpu_device *adev,
+					bool keep_trap_enabled,
+					uint32_t vmid)
+{
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	kgd_gfx_v10_set_wave_launch_stall(adev, vmid, true);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+	kgd_gfx_v10_set_wave_launch_stall(adev, vmid, false);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+int kgd_gfx_v10_validate_trap_override_request(struct amdgpu_device *adev,
+					      uint32_t trap_override,
+					      uint32_t *trap_mask_supported)
+{
+	*trap_mask_supported &= KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH;
+
+	/* The SPI_GDBG_TRAP_MASK register is global and affects all
+	 * processes. Only allow OR-ing the address-watch bit, since
+	 * this only affects processes under the debugger. Other bits
+	 * should stay 0 to avoid the debugger interfering with other
+	 * processes.
+	 */
+	if (trap_override != KFD_DBG_TRAP_OVERRIDE_OR)
+		return -EINVAL;
+
+	return 0;
+}
+
+uint32_t kgd_gfx_v10_set_wave_launch_trap_override(struct amdgpu_device *adev,
+					      uint32_t vmid,
+					      uint32_t trap_override,
+					      uint32_t trap_mask_bits,
+					      uint32_t trap_mask_request,
+					      uint32_t *trap_mask_prev,
+					      uint32_t kfd_dbg_trap_cntl_prev)
+{
+	uint32_t data, wave_cntl_prev;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	wave_cntl_prev = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+
+	kgd_gfx_v10_set_wave_launch_stall(adev, vmid, true);
+
+	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK));
+	*trap_mask_prev = REG_GET_FIELD(data, SPI_GDBG_TRAP_MASK, EXCP_EN);
+
+	trap_mask_bits = (trap_mask_bits & trap_mask_request) |
+		(*trap_mask_prev & ~trap_mask_request);
+
+	data = REG_SET_FIELD(data, SPI_GDBG_TRAP_MASK, EXCP_EN, trap_mask_bits);
+	data = REG_SET_FIELD(data, SPI_GDBG_TRAP_MASK, REPLACE, trap_override);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), data);
+
+	/* We need to preserve wave launch mode stall settings. */
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), wave_cntl_prev);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+uint32_t kgd_gfx_v10_set_wave_launch_mode(struct amdgpu_device *adev,
+					uint8_t wave_launch_mode,
+					uint32_t vmid)
+{
+	uint32_t data = 0;
+	bool is_mode_set = !!wave_launch_mode;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	kgd_gfx_v10_set_wave_launch_stall(adev, vmid, true);
+
+	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL2,
+			VMID_MASK, is_mode_set ? 1 << vmid : 0);
+	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL2,
+			MODE, is_mode_set ? wave_launch_mode : 0);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL2), data);
+
+	kgd_gfx_v10_set_wave_launch_stall(adev, vmid, false);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+#define TCP_WATCH_STRIDE (mmTCP_WATCH1_ADDR_H - mmTCP_WATCH0_ADDR_H)
+#define SQ_WATCH_STRIDE (mmSQ_WATCH1_ADDR_H - mmSQ_WATCH0_ADDR_H)
+uint32_t kgd_gfx_v10_set_address_watch(struct amdgpu_device *adev,
+					uint64_t watch_address,
+					uint32_t watch_address_mask,
+					uint32_t watch_id,
+					uint32_t watch_mode,
+					uint32_t debug_vmid,
+					uint32_t inst)
+{
+	/* SQ_WATCH?_ADDR_* and TCP_WATCH?_ADDR_* are programmed with the
+	 * same values.
+	 */
+	uint32_t watch_address_high;
+	uint32_t watch_address_low;
+	uint32_t tcp_watch_address_cntl;
+	uint32_t sq_watch_address_cntl;
+
+	watch_address_low = lower_32_bits(watch_address);
+	watch_address_high = upper_32_bits(watch_address) & 0xffff;
+
+	tcp_watch_address_cntl = 0;
+	tcp_watch_address_cntl = REG_SET_FIELD(tcp_watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VMID,
+			debug_vmid);
+	tcp_watch_address_cntl = REG_SET_FIELD(tcp_watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MODE,
+			watch_mode);
+	tcp_watch_address_cntl = REG_SET_FIELD(tcp_watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MASK,
+			watch_address_mask >> 7);
+
+	sq_watch_address_cntl = 0;
+	sq_watch_address_cntl = REG_SET_FIELD(sq_watch_address_cntl,
+			SQ_WATCH0_CNTL,
+			VMID,
+			debug_vmid);
+	sq_watch_address_cntl = REG_SET_FIELD(sq_watch_address_cntl,
+			SQ_WATCH0_CNTL,
+			MODE,
+			watch_mode);
+	sq_watch_address_cntl = REG_SET_FIELD(sq_watch_address_cntl,
+			SQ_WATCH0_CNTL,
+			MASK,
+			watch_address_mask >> 6);
+
+	/* Turning off this watch point until we set all the registers */
+	tcp_watch_address_cntl = REG_SET_FIELD(tcp_watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VALID,
+			0);
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			tcp_watch_address_cntl);
+
+	sq_watch_address_cntl = REG_SET_FIELD(sq_watch_address_cntl,
+			SQ_WATCH0_CNTL,
+			VALID,
+			0);
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmSQ_WATCH0_CNTL) +
+			(watch_id * SQ_WATCH_STRIDE)),
+			sq_watch_address_cntl);
+
+	/* Program {TCP,SQ}_WATCH?_ADDR* */
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_ADDR_H) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_high);
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_ADDR_L) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_low);
+
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmSQ_WATCH0_ADDR_H) +
+			(watch_id * SQ_WATCH_STRIDE)),
+			watch_address_high);
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmSQ_WATCH0_ADDR_L) +
+			(watch_id * SQ_WATCH_STRIDE)),
+			watch_address_low);
+
+	/* Enable the watch point */
+	tcp_watch_address_cntl = REG_SET_FIELD(tcp_watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VALID,
+			1);
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			tcp_watch_address_cntl);
+
+	sq_watch_address_cntl = REG_SET_FIELD(sq_watch_address_cntl,
+			SQ_WATCH0_CNTL,
+			VALID,
+			1);
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmSQ_WATCH0_CNTL) +
+			(watch_id * SQ_WATCH_STRIDE)),
+			sq_watch_address_cntl);
+
+	return 0;
+}
+
+uint32_t kgd_gfx_v10_clear_address_watch(struct amdgpu_device *adev,
+					uint32_t watch_id)
+{
+	uint32_t watch_address_cntl;
+
+	watch_address_cntl = 0;
+
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_cntl);
+
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmSQ_WATCH0_CNTL) +
+			(watch_id * SQ_WATCH_STRIDE)),
+			watch_address_cntl);
+
+	return 0;
+}
+#undef TCP_WATCH_STRIDE
+#undef SQ_WATCH_STRIDE
+
+
+/* kgd_gfx_v10_get_iq_wait_times: Returns the mmCP_IQ_WAIT_TIME1/2 values
+ * The values read are:
+ *     ib_offload_wait_time     -- Wait Count for Indirect Buffer Offloads.
+ *     atomic_offload_wait_time -- Wait Count for L2 and GDS Atomics Offloads.
+ *     wrm_offload_wait_time    -- Wait Count for WAIT_REG_MEM Offloads.
+ *     gws_wait_time            -- Wait Count for Global Wave Syncs.
+ *     que_sleep_wait_time      -- Wait Count for Dequeue Retry.
+ *     sch_wave_wait_time       -- Wait Count for Scheduling Wave Message.
+ *     sem_rearm_wait_time      -- Wait Count for Semaphore re-arm.
+ *     deq_retry_wait_time      -- Wait Count for Global Wave Syncs.
+ */
+void kgd_gfx_v10_get_iq_wait_times(struct amdgpu_device *adev,
+					uint32_t *wait_times,
+					uint32_t inst)
+
+{
+	*wait_times = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2));
+}
+
+void kgd_gfx_v10_build_dequeue_wait_counts_packet_info(struct amdgpu_device *adev,
+						uint32_t wait_times,
+						uint32_t sch_wave,
+						uint32_t que_sleep,
+						uint32_t *reg_offset,
+						uint32_t *reg_data)
+{
+	*reg_data = wait_times;
+
+	if (sch_wave)
+		*reg_data = REG_SET_FIELD(*reg_data,
+				CP_IQ_WAIT_TIME2,
+				SCH_WAVE,
+				sch_wave);
+	if (que_sleep)
+		*reg_data = REG_SET_FIELD(*reg_data,
+				CP_IQ_WAIT_TIME2,
+				QUE_SLEEP,
+				que_sleep);
+
+	*reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2);
+}
+
 static void program_trap_handler_settings(struct amdgpu_device *adev,
-		uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr)
+		uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr,
+		uint32_t inst)
 {
 	lock_srbm(adev, 0, 0, 0, vmid);
 
@@ -733,6 +1070,26 @@ static void program_trap_handler_settings(struct amdgpu_device *adev,
 	unlock_srbm(adev);
 }
 
+uint64_t kgd_gfx_v10_hqd_get_pq_addr(struct amdgpu_device *adev,
+				     uint32_t pipe_id, uint32_t queue_id,
+				     uint32_t inst)
+{
+	return 0;
+}
+
+uint64_t kgd_gfx_v10_hqd_reset(struct amdgpu_device *adev,
+			       uint32_t pipe_id, uint32_t queue_id,
+			       uint32_t inst, unsigned int utimeout)
+{
+	return 0;
+}
+
+uint32_t kgd_gfx_v10_hqd_sdma_get_doorbell(struct amdgpu_device *adev,
+					   int engine, int queue)
+{
+	return 0;
+}
+
 const struct kfd2kgd_calls gfx_v10_kfd2kgd = {
 	.program_sh_mem_settings = kgd_program_sh_mem_settings,
 	.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
@@ -750,5 +1107,17 @@ const struct kfd2kgd_calls gfx_v10_kfd2kgd = {
 	.get_atc_vmid_pasid_mapping_info =
 			get_atc_vmid_pasid_mapping_info,
 	.set_vm_context_page_table_base = set_vm_context_page_table_base,
+	.enable_debug_trap = kgd_gfx_v10_enable_debug_trap,
+	.disable_debug_trap = kgd_gfx_v10_disable_debug_trap,
+	.validate_trap_override_request = kgd_gfx_v10_validate_trap_override_request,
+	.set_wave_launch_trap_override = kgd_gfx_v10_set_wave_launch_trap_override,
+	.set_wave_launch_mode = kgd_gfx_v10_set_wave_launch_mode,
+	.set_address_watch = kgd_gfx_v10_set_address_watch,
+	.clear_address_watch = kgd_gfx_v10_clear_address_watch,
+	.get_iq_wait_times = kgd_gfx_v10_get_iq_wait_times,
+	.build_dequeue_wait_counts_packet_info = kgd_gfx_v10_build_dequeue_wait_counts_packet_info,
 	.program_trap_handler_settings = program_trap_handler_settings,
+	.hqd_get_pq_addr = kgd_gfx_v10_hqd_get_pq_addr,
+	.hqd_reset = kgd_gfx_v10_hqd_reset,
+	.hqd_sdma_get_doorbell = kgd_gfx_v10_hqd_sdma_get_doorbell
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
new file mode 100644
index 000000000000..a4c607c88178
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
@@ -0,0 +1,70 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+uint32_t kgd_gfx_v10_enable_debug_trap(struct amdgpu_device *adev,
+				      bool restore_dbg_registers,
+				      uint32_t vmid);
+uint32_t kgd_gfx_v10_disable_debug_trap(struct amdgpu_device *adev,
+					bool keep_trap_enabled,
+					uint32_t vmid);
+int kgd_gfx_v10_validate_trap_override_request(struct amdgpu_device *adev,
+					     uint32_t trap_override,
+					     uint32_t *trap_mask_supported);
+uint32_t kgd_gfx_v10_set_wave_launch_trap_override(struct amdgpu_device *adev,
+					     uint32_t vmid,
+					     uint32_t trap_override,
+					     uint32_t trap_mask_bits,
+					     uint32_t trap_mask_request,
+					     uint32_t *trap_mask_prev,
+					     uint32_t kfd_dbg_trap_cntl_prev);
+uint32_t kgd_gfx_v10_set_wave_launch_mode(struct amdgpu_device *adev,
+					 uint8_t wave_launch_mode,
+					 uint32_t vmid);
+uint32_t kgd_gfx_v10_set_address_watch(struct amdgpu_device *adev,
+					uint64_t watch_address,
+					uint32_t watch_address_mask,
+					uint32_t watch_id,
+					uint32_t watch_mode,
+					uint32_t debug_vmid,
+					uint32_t inst);
+uint32_t kgd_gfx_v10_clear_address_watch(struct amdgpu_device *adev,
+					uint32_t watch_id);
+void kgd_gfx_v10_get_iq_wait_times(struct amdgpu_device *adev,
+				uint32_t *wait_times,
+				uint32_t inst);
+void kgd_gfx_v10_build_dequeue_wait_counts_packet_info(struct amdgpu_device *adev,
+					       uint32_t wait_times,
+					       uint32_t sch_wave,
+					       uint32_t que_sleep,
+					       uint32_t *reg_offset,
+					       uint32_t *reg_data);
+uint64_t kgd_gfx_v10_hqd_get_pq_addr(struct amdgpu_device *adev,
+				    uint32_t pipe_id,
+				    uint32_t queue_id,
+				    uint32_t inst);
+uint64_t kgd_gfx_v10_hqd_reset(struct amdgpu_device *adev,
+			      uint32_t pipe_id,
+			      uint32_t queue_id,
+			      uint32_t inst,
+			      unsigned int utimeout);
+uint32_t kgd_gfx_v10_hqd_sdma_get_doorbell(struct amdgpu_device *adev,
+					   int engine, int queue);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
index ba21ec6b35e0..f2278a0937ff 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
@@ -22,6 +22,7 @@
 #include <linux/mmu_context.h>
 #include "amdgpu.h"
 #include "amdgpu_amdkfd.h"
+#include "amdgpu_amdkfd_gfx_v10.h"
 #include "gc/gc_10_3_0_offset.h"
 #include "gc/gc_10_3_0_sh_mask.h"
 #include "oss/osssys_5_0_0_offset.h"
@@ -80,7 +81,7 @@ static void program_sh_mem_settings_v10_3(struct amdgpu_device *adev, uint32_t v
 					uint32_t sh_mem_config,
 					uint32_t sh_mem_ape1_base,
 					uint32_t sh_mem_ape1_limit,
-					uint32_t sh_mem_bases)
+					uint32_t sh_mem_bases, uint32_t inst)
 {
 	lock_srbm(adev, 0, 0, 0, vmid);
 
@@ -93,7 +94,7 @@ static void program_sh_mem_settings_v10_3(struct amdgpu_device *adev, uint32_t v
 
 /* ATC is defeatured on Sienna_Cichlid */
 static int set_pasid_vmid_mapping_v10_3(struct amdgpu_device *adev, unsigned int pasid,
-					unsigned int vmid)
+					unsigned int vmid, uint32_t inst)
 {
 	uint32_t value = pasid << IH_VMID_0_LUT__PASID__SHIFT;
 
@@ -105,7 +106,8 @@ static int set_pasid_vmid_mapping_v10_3(struct amdgpu_device *adev, unsigned int
 	return 0;
 }
 
-static int init_interrupts_v10_3(struct amdgpu_device *adev, uint32_t pipe_id)
+static int init_interrupts_v10_3(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t inst)
 {
 	uint32_t mec;
 	uint32_t pipe;
@@ -177,7 +179,7 @@ static inline struct v10_sdma_mqd *get_sdma_mqd(void *mqd)
 static int hqd_load_v10_3(struct amdgpu_device *adev, void *mqd,
 			uint32_t pipe_id, uint32_t queue_id,
 			uint32_t __user *wptr, uint32_t wptr_shift,
-			uint32_t wptr_mask, struct mm_struct *mm)
+			uint32_t wptr_mask, struct mm_struct *mm, uint32_t inst)
 {
 	struct v10_compute_mqd *m;
 	uint32_t *mqd_hqd;
@@ -273,9 +275,9 @@ static int hqd_load_v10_3(struct amdgpu_device *adev, void *mqd,
 
 static int hiq_mqd_load_v10_3(struct amdgpu_device *adev, void *mqd,
 			    uint32_t pipe_id, uint32_t queue_id,
-			    uint32_t doorbell_off)
+			    uint32_t doorbell_off, uint32_t inst)
 {
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
+	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq[0].ring;
 	struct v10_compute_mqd *m;
 	uint32_t mec, pipe;
 	int r;
@@ -290,7 +292,7 @@ static int hiq_mqd_load_v10_3(struct amdgpu_device *adev, void *mqd,
 	pr_debug("kfd: set HIQ, mec:%d, pipe:%d, queue:%d.\n",
 		 mec, pipe, queue_id);
 
-	spin_lock(&adev->gfx.kiq.ring_lock);
+	spin_lock(&adev->gfx.kiq[0].ring_lock);
 	r = amdgpu_ring_alloc(kiq_ring, 7);
 	if (r) {
 		pr_err("Failed to alloc KIQ (%d).\n", r);
@@ -317,7 +319,7 @@ static int hiq_mqd_load_v10_3(struct amdgpu_device *adev, void *mqd,
 	amdgpu_ring_commit(kiq_ring);
 
 out_unlock:
-	spin_unlock(&adev->gfx.kiq.ring_lock);
+	spin_unlock(&adev->gfx.kiq[0].ring_lock);
 	release_queue(adev);
 
 	return r;
@@ -325,7 +327,7 @@ out_unlock:
 
 static int hqd_dump_v10_3(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
-			uint32_t (**dump)[2], uint32_t *n_regs)
+			uint32_t (**dump)[2], uint32_t *n_regs, uint32_t inst)
 {
 	uint32_t i = 0, reg;
 #define HQD_N_REGS 56
@@ -336,7 +338,7 @@ static int hqd_dump_v10_3(struct amdgpu_device *adev,
 		(*dump)[i++][1] = RREG32_SOC15_IP(GC, addr);		\
 	} while (0)
 
-	*dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -433,7 +435,7 @@ static int hqd_sdma_dump_v10_3(struct amdgpu_device *adev,
 #undef HQD_N_REGS
 #define HQD_N_REGS (19+6+7+12)
 
-	*dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -456,7 +458,7 @@ static int hqd_sdma_dump_v10_3(struct amdgpu_device *adev,
 
 static bool hqd_is_occupied_v10_3(struct amdgpu_device *adev,
 				uint64_t queue_address, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	uint32_t act;
 	bool retval = false;
@@ -498,7 +500,7 @@ static bool hqd_sdma_is_occupied_v10_3(struct amdgpu_device *adev,
 static int hqd_destroy_v10_3(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	enum hqd_dequeue_request_type type;
 	unsigned long end_jiffies;
@@ -586,7 +588,7 @@ static int hqd_sdma_destroy_v10_3(struct amdgpu_device *adev, void *mqd,
 
 static int wave_control_execute_v10_3(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
-					uint32_t sq_cmd)
+					uint32_t sq_cmd, uint32_t inst)
 {
 	uint32_t data = 0;
 
@@ -628,7 +630,8 @@ static void set_vm_context_page_table_base_v10_3(struct amdgpu_device *adev,
 }
 
 static void program_trap_handler_settings_v10_3(struct amdgpu_device *adev,
-			uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr)
+			uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr,
+			uint32_t inst)
 {
 	lock_srbm(adev, 0, 0, 0, vmid);
 
@@ -652,142 +655,6 @@ static void program_trap_handler_settings_v10_3(struct amdgpu_device *adev,
 	unlock_srbm(adev);
 }
 
-#if 0
-uint32_t enable_debug_trap_v10_3(struct amdgpu_device *adev,
-				uint32_t trap_debug_wave_launch_mode,
-				uint32_t vmid)
-{
-	uint32_t data = 0;
-	uint32_t orig_wave_cntl_value;
-	uint32_t orig_stall_vmid;
-
-	mutex_lock(&adev->grbm_idx_mutex);
-
-	orig_wave_cntl_value = RREG32(SOC15_REG_OFFSET(GC,
-				0,
-				mmSPI_GDBG_WAVE_CNTL));
-	orig_stall_vmid = REG_GET_FIELD(orig_wave_cntl_value,
-			SPI_GDBG_WAVE_CNTL,
-			STALL_VMID);
-
-	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_RA, 1);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);
-
-	data = 0;
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), data);
-
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), orig_stall_vmid);
-
-	mutex_unlock(&adev->grbm_idx_mutex);
-
-	return 0;
-}
-
-uint32_t disable_debug_trap_v10_3(struct amdgpu_device *adev)
-{
-	mutex_lock(&adev->grbm_idx_mutex);
-
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
-
-	mutex_unlock(&adev->grbm_idx_mutex);
-
-	return 0;
-}
-
-uint32_t set_wave_launch_trap_override_v10_3(struct amdgpu_device *adev,
-						uint32_t trap_override,
-						uint32_t trap_mask)
-{
-	uint32_t data = 0;
-
-	mutex_lock(&adev->grbm_idx_mutex);
-
-	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
-	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_RA, 1);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);
-
-	data = 0;
-	data = REG_SET_FIELD(data, SPI_GDBG_TRAP_MASK,
-			EXCP_EN, trap_mask);
-	data = REG_SET_FIELD(data, SPI_GDBG_TRAP_MASK,
-			REPLACE, trap_override);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), data);
-
-	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
-	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_RA, 0);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);
-
-	mutex_unlock(&adev->grbm_idx_mutex);
-
-	return 0;
-}
-
-uint32_t set_wave_launch_mode_v10_3(struct amdgpu_device *adev,
-					uint8_t wave_launch_mode,
-					uint32_t vmid)
-{
-	uint32_t data = 0;
-	bool is_stall_mode;
-	bool is_mode_set;
-
-	is_stall_mode = (wave_launch_mode == 4);
-	is_mode_set = (wave_launch_mode != 0 && wave_launch_mode != 4);
-
-	mutex_lock(&adev->grbm_idx_mutex);
-
-	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL2,
-			VMID_MASK, is_mode_set ? 1 << vmid : 0);
-	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL2,
-			MODE, is_mode_set ? wave_launch_mode : 0);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL2), data);
-
-	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
-	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL,
-			STALL_VMID, is_stall_mode ? 1 << vmid : 0);
-	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL,
-			STALL_RA, is_stall_mode ? 1 : 0);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);
-
-	mutex_unlock(&adev->grbm_idx_mutex);
-
-	return 0;
-}
-
-/* kgd_get_iq_wait_times: Returns the mmCP_IQ_WAIT_TIME1/2 values
- * The values read are:
- *	ib_offload_wait_time     -- Wait Count for Indirect Buffer Offloads.
- *	atomic_offload_wait_time -- Wait Count for L2 and GDS Atomics Offloads.
- *	wrm_offload_wait_time    -- Wait Count for WAIT_REG_MEM Offloads.
- *	gws_wait_time            -- Wait Count for Global Wave Syncs.
- *	que_sleep_wait_time      -- Wait Count for Dequeue Retry.
- *	sch_wave_wait_time       -- Wait Count for Scheduling Wave Message.
- *	sem_rearm_wait_time      -- Wait Count for Semaphore re-arm.
- *	deq_retry_wait_time      -- Wait Count for Global Wave Syncs.
- */
-void get_iq_wait_times_v10_3(struct amdgpu_device *adev,
-					uint32_t *wait_times)
-
-{
-	*wait_times = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2));
-}
-
-void build_grace_period_packet_info_v10_3(struct amdgpu_device *adev,
-						uint32_t wait_times,
-						uint32_t grace_period,
-						uint32_t *reg_offset,
-						uint32_t *reg_data)
-{
-	*reg_data = wait_times;
-
-	*reg_data = REG_SET_FIELD(*reg_data,
-			CP_IQ_WAIT_TIME2,
-			SCH_WAVE,
-			grace_period);
-
-	*reg_offset = mmCP_IQ_WAIT_TIME2;
-}
-#endif
-
 const struct kfd2kgd_calls gfx_v10_3_kfd2kgd = {
 	.program_sh_mem_settings = program_sh_mem_settings_v10_3,
 	.set_pasid_vmid_mapping = set_pasid_vmid_mapping_v10_3,
@@ -805,12 +672,16 @@ const struct kfd2kgd_calls gfx_v10_3_kfd2kgd = {
 	.get_atc_vmid_pasid_mapping_info = get_atc_vmid_pasid_mapping_info_v10_3,
 	.set_vm_context_page_table_base = set_vm_context_page_table_base_v10_3,
 	.program_trap_handler_settings = program_trap_handler_settings_v10_3,
-#if 0
-	.enable_debug_trap = enable_debug_trap_v10_3,
-	.disable_debug_trap = disable_debug_trap_v10_3,
-	.set_wave_launch_trap_override = set_wave_launch_trap_override_v10_3,
-	.set_wave_launch_mode = set_wave_launch_mode_v10_3,
-	.get_iq_wait_times = get_iq_wait_times_v10_3,
-	.build_grace_period_packet_info = build_grace_period_packet_info_v10_3,
-#endif
+	.get_iq_wait_times = kgd_gfx_v10_get_iq_wait_times,
+	.build_dequeue_wait_counts_packet_info = kgd_gfx_v10_build_dequeue_wait_counts_packet_info,
+	.enable_debug_trap = kgd_gfx_v10_enable_debug_trap,
+	.disable_debug_trap = kgd_gfx_v10_disable_debug_trap,
+	.validate_trap_override_request = kgd_gfx_v10_validate_trap_override_request,
+	.set_wave_launch_trap_override = kgd_gfx_v10_set_wave_launch_trap_override,
+	.set_wave_launch_mode = kgd_gfx_v10_set_wave_launch_mode,
+	.set_address_watch = kgd_gfx_v10_set_address_watch,
+	.clear_address_watch = kgd_gfx_v10_clear_address_watch,
+	.hqd_get_pq_addr = kgd_gfx_v10_hqd_get_pq_addr,
+	.hqd_reset = kgd_gfx_v10_hqd_reset,
+	.hqd_sdma_get_doorbell = kgd_gfx_v10_hqd_sdma_get_doorbell
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c
index 7e80caa05060..aaccf0b9947d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c
@@ -30,6 +30,7 @@
 #include "soc15d.h"
 #include "v11_structs.h"
 #include "soc21.h"
+#include <uapi/linux/kfd_ioctl.h>
 
 enum hqd_dequeue_request_type {
 	NO_ACTION = 0,
@@ -78,7 +79,7 @@ static void program_sh_mem_settings_v11(struct amdgpu_device *adev, uint32_t vmi
 					uint32_t sh_mem_config,
 					uint32_t sh_mem_ape1_base,
 					uint32_t sh_mem_ape1_limit,
-					uint32_t sh_mem_bases)
+					uint32_t sh_mem_bases, uint32_t inst)
 {
 	lock_srbm(adev, 0, 0, 0, vmid);
 
@@ -89,7 +90,7 @@ static void program_sh_mem_settings_v11(struct amdgpu_device *adev, uint32_t vmi
 }
 
 static int set_pasid_vmid_mapping_v11(struct amdgpu_device *adev, unsigned int pasid,
-					unsigned int vmid)
+					unsigned int vmid, uint32_t inst)
 {
 	uint32_t value = pasid << IH_VMID_0_LUT__PASID__SHIFT;
 
@@ -101,7 +102,8 @@ static int set_pasid_vmid_mapping_v11(struct amdgpu_device *adev, unsigned int p
 	return 0;
 }
 
-static int init_interrupts_v11(struct amdgpu_device *adev, uint32_t pipe_id)
+static int init_interrupts_v11(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t inst)
 {
 	uint32_t mec;
 	uint32_t pipe;
@@ -162,7 +164,7 @@ static inline struct v11_sdma_mqd *get_sdma_mqd(void *mqd)
 static int hqd_load_v11(struct amdgpu_device *adev, void *mqd, uint32_t pipe_id,
 			uint32_t queue_id, uint32_t __user *wptr,
 			uint32_t wptr_shift, uint32_t wptr_mask,
-			struct mm_struct *mm)
+			struct mm_struct *mm, uint32_t inst)
 {
 	struct v11_compute_mqd *m;
 	uint32_t *mqd_hqd;
@@ -258,9 +260,9 @@ static int hqd_load_v11(struct amdgpu_device *adev, void *mqd, uint32_t pipe_id,
 
 static int hiq_mqd_load_v11(struct amdgpu_device *adev, void *mqd,
 			      uint32_t pipe_id, uint32_t queue_id,
-			      uint32_t doorbell_off)
+			      uint32_t doorbell_off, uint32_t inst)
 {
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
+	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq[0].ring;
 	struct v11_compute_mqd *m;
 	uint32_t mec, pipe;
 	int r;
@@ -275,7 +277,7 @@ static int hiq_mqd_load_v11(struct amdgpu_device *adev, void *mqd,
 	pr_debug("kfd: set HIQ, mec:%d, pipe:%d, queue:%d.\n",
 		 mec, pipe, queue_id);
 
-	spin_lock(&adev->gfx.kiq.ring_lock);
+	spin_lock(&adev->gfx.kiq[0].ring_lock);
 	r = amdgpu_ring_alloc(kiq_ring, 7);
 	if (r) {
 		pr_err("Failed to alloc KIQ (%d).\n", r);
@@ -302,7 +304,7 @@ static int hiq_mqd_load_v11(struct amdgpu_device *adev, void *mqd,
 	amdgpu_ring_commit(kiq_ring);
 
 out_unlock:
-	spin_unlock(&adev->gfx.kiq.ring_lock);
+	spin_unlock(&adev->gfx.kiq[0].ring_lock);
 	release_queue(adev);
 
 	return r;
@@ -310,7 +312,7 @@ out_unlock:
 
 static int hqd_dump_v11(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
-			uint32_t (**dump)[2], uint32_t *n_regs)
+			uint32_t (**dump)[2], uint32_t *n_regs, uint32_t inst)
 {
 	uint32_t i = 0, reg;
 #define HQD_N_REGS 56
@@ -321,7 +323,7 @@ static int hqd_dump_v11(struct amdgpu_device *adev,
 		(*dump)[i++][1] = RREG32(addr);		\
 	} while (0)
 
-	*dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -418,7 +420,7 @@ static int hqd_sdma_dump_v11(struct amdgpu_device *adev,
 #undef HQD_N_REGS
 #define HQD_N_REGS (7+11+1+12+12)
 
-	*dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -445,7 +447,7 @@ static int hqd_sdma_dump_v11(struct amdgpu_device *adev,
 }
 
 static bool hqd_is_occupied_v11(struct amdgpu_device *adev, uint64_t queue_address,
-				uint32_t pipe_id, uint32_t queue_id)
+				uint32_t pipe_id, uint32_t queue_id, uint32_t inst)
 {
 	uint32_t act;
 	bool retval = false;
@@ -486,7 +488,7 @@ static bool hqd_sdma_is_occupied_v11(struct amdgpu_device *adev, void *mqd)
 static int hqd_destroy_v11(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	enum hqd_dequeue_request_type type;
 	unsigned long end_jiffies;
@@ -571,7 +573,7 @@ static int hqd_sdma_destroy_v11(struct amdgpu_device *adev, void *mqd,
 
 static int wave_control_execute_v11(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
-					uint32_t sq_cmd)
+					uint32_t sq_cmd, uint32_t inst)
 {
 	uint32_t data = 0;
 
@@ -606,6 +608,204 @@ static void set_vm_context_page_table_base_v11(struct amdgpu_device *adev,
 	adev->gfxhub.funcs->setup_vm_pt_regs(adev, vmid, page_table_base);
 }
 
+/*
+ * Returns TRAP_EN, EXCP_EN and EXCP_REPLACE.
+ *
+ * restore_dbg_registers is ignored here but is a general interface requirement
+ * for devices that support GFXOFF and where the RLC save/restore list
+ * does not support hw registers for debugging i.e. the driver has to manually
+ * initialize the debug mode registers after it has disabled GFX off during the
+ * debug session.
+ */
+static uint32_t kgd_gfx_v11_enable_debug_trap(struct amdgpu_device *adev,
+					    bool restore_dbg_registers,
+					    uint32_t vmid)
+{
+	uint32_t data = 0;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, 0);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, 0);
+
+	return data;
+}
+
+/* Returns TRAP_EN, EXCP_EN and EXCP_REPLACE. */
+static uint32_t kgd_gfx_v11_disable_debug_trap(struct amdgpu_device *adev,
+						bool keep_trap_enabled,
+						uint32_t vmid)
+{
+	uint32_t data = 0;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, 0);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, 0);
+
+	return data;
+}
+
+static int kgd_gfx_v11_validate_trap_override_request(struct amdgpu_device *adev,
+							uint32_t trap_override,
+							uint32_t *trap_mask_supported)
+{
+	*trap_mask_supported &= KFD_DBG_TRAP_MASK_FP_INVALID |
+				KFD_DBG_TRAP_MASK_FP_INPUT_DENORMAL |
+				KFD_DBG_TRAP_MASK_FP_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_FP_OVERFLOW |
+				KFD_DBG_TRAP_MASK_FP_UNDERFLOW |
+				KFD_DBG_TRAP_MASK_FP_INEXACT |
+				KFD_DBG_TRAP_MASK_INT_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH |
+				KFD_DBG_TRAP_MASK_DBG_MEMORY_VIOLATION;
+
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(11, 0, 4))
+		*trap_mask_supported |= KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START |
+					KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END;
+
+	if (trap_override != KFD_DBG_TRAP_OVERRIDE_OR &&
+			trap_override != KFD_DBG_TRAP_OVERRIDE_REPLACE)
+		return -EPERM;
+
+	return 0;
+}
+
+static uint32_t trap_mask_map_sw_to_hw(uint32_t mask)
+{
+	uint32_t trap_on_start = (mask & KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START) ? 1 : 0;
+	uint32_t trap_on_end = (mask & KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END) ? 1 : 0;
+	uint32_t excp_en = mask & (KFD_DBG_TRAP_MASK_FP_INVALID |
+			KFD_DBG_TRAP_MASK_FP_INPUT_DENORMAL |
+			KFD_DBG_TRAP_MASK_FP_DIVIDE_BY_ZERO |
+			KFD_DBG_TRAP_MASK_FP_OVERFLOW |
+			KFD_DBG_TRAP_MASK_FP_UNDERFLOW |
+			KFD_DBG_TRAP_MASK_FP_INEXACT |
+			KFD_DBG_TRAP_MASK_INT_DIVIDE_BY_ZERO |
+			KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH |
+			KFD_DBG_TRAP_MASK_DBG_MEMORY_VIOLATION);
+	uint32_t ret;
+
+	ret = REG_SET_FIELD(0, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, excp_en);
+	ret = REG_SET_FIELD(ret, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_START, trap_on_start);
+	ret = REG_SET_FIELD(ret, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_END, trap_on_end);
+
+	return ret;
+}
+
+static uint32_t trap_mask_map_hw_to_sw(uint32_t mask)
+{
+	uint32_t ret = REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, EXCP_EN);
+
+	if (REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_START))
+		ret |= KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START;
+
+	if (REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_END))
+		ret |= KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END;
+
+	return ret;
+}
+
+/* Returns TRAP_EN, EXCP_EN and EXCP_REPLACE. */
+static uint32_t kgd_gfx_v11_set_wave_launch_trap_override(struct amdgpu_device *adev,
+					uint32_t vmid,
+					uint32_t trap_override,
+					uint32_t trap_mask_bits,
+					uint32_t trap_mask_request,
+					uint32_t *trap_mask_prev,
+					uint32_t kfd_dbg_trap_cntl_prev)
+{
+	uint32_t data = 0;
+
+	*trap_mask_prev = trap_mask_map_hw_to_sw(kfd_dbg_trap_cntl_prev);
+
+	data = (trap_mask_bits & trap_mask_request) | (*trap_mask_prev & ~trap_mask_request);
+	data = trap_mask_map_sw_to_hw(data);
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, trap_override);
+
+	return data;
+}
+
+static uint32_t kgd_gfx_v11_set_wave_launch_mode(struct amdgpu_device *adev,
+					uint8_t wave_launch_mode,
+					uint32_t vmid)
+{
+	uint32_t data = 0;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, LAUNCH_MODE, wave_launch_mode);
+
+	return data;
+}
+
+#define TCP_WATCH_STRIDE (regTCP_WATCH1_ADDR_H - regTCP_WATCH0_ADDR_H)
+static uint32_t kgd_gfx_v11_set_address_watch(struct amdgpu_device *adev,
+					uint64_t watch_address,
+					uint32_t watch_address_mask,
+					uint32_t watch_id,
+					uint32_t watch_mode,
+					uint32_t debug_vmid,
+					uint32_t inst)
+{
+	uint32_t watch_address_high;
+	uint32_t watch_address_low;
+	uint32_t watch_address_cntl;
+
+	watch_address_cntl = 0;
+	watch_address_low = lower_32_bits(watch_address);
+	watch_address_high = upper_32_bits(watch_address) & 0xffff;
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MODE,
+			watch_mode);
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MASK,
+			watch_address_mask >> 7);
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VALID,
+			1);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_H) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_high);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_L) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_low);
+
+	return watch_address_cntl;
+}
+
+static uint32_t kgd_gfx_v11_clear_address_watch(struct amdgpu_device *adev,
+						uint32_t watch_id)
+{
+	return 0;
+}
+
+static uint64_t kgd_gfx_v11_hqd_get_pq_addr(struct amdgpu_device *adev,
+					    uint32_t pipe_id, uint32_t queue_id,
+					    uint32_t inst)
+{
+	return 0;
+}
+
+static uint64_t kgd_gfx_v11_hqd_reset(struct amdgpu_device *adev,
+				      uint32_t pipe_id, uint32_t queue_id,
+				      uint32_t inst, unsigned int utimeout)
+{
+	return 0;
+}
+
+static uint32_t kgd_gfx_v11_hqd_sdma_get_doorbell(struct amdgpu_device *adev,
+						  int engine, int queue)
+{
+	return 0;
+}
+
 const struct kfd2kgd_calls gfx_v11_kfd2kgd = {
 	.program_sh_mem_settings = program_sh_mem_settings_v11,
 	.set_pasid_vmid_mapping = set_pasid_vmid_mapping_v11,
@@ -622,4 +822,14 @@ const struct kfd2kgd_calls gfx_v11_kfd2kgd = {
 	.wave_control_execute = wave_control_execute_v11,
 	.get_atc_vmid_pasid_mapping_info = NULL,
 	.set_vm_context_page_table_base = set_vm_context_page_table_base_v11,
+	.enable_debug_trap = kgd_gfx_v11_enable_debug_trap,
+	.disable_debug_trap = kgd_gfx_v11_disable_debug_trap,
+	.validate_trap_override_request = kgd_gfx_v11_validate_trap_override_request,
+	.set_wave_launch_trap_override = kgd_gfx_v11_set_wave_launch_trap_override,
+	.set_wave_launch_mode = kgd_gfx_v11_set_wave_launch_mode,
+	.set_address_watch = kgd_gfx_v11_set_address_watch,
+	.clear_address_watch = kgd_gfx_v11_clear_address_watch,
+	.hqd_get_pq_addr = kgd_gfx_v11_hqd_get_pq_addr,
+	.hqd_reset = kgd_gfx_v11_hqd_reset,
+	.hqd_sdma_get_doorbell = kgd_gfx_v11_hqd_sdma_get_doorbell
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v12.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v12.c
new file mode 100644
index 000000000000..e0ceab400b2d
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v12.c
@@ -0,0 +1,384 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "amdgpu.h"
+#include "amdgpu_amdkfd.h"
+#include "gc/gc_12_0_0_offset.h"
+#include "gc/gc_12_0_0_sh_mask.h"
+#include "soc24.h"
+#include <uapi/linux/kfd_ioctl.h>
+
+static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
+			uint32_t queue, uint32_t vmid)
+{
+	mutex_lock(&adev->srbm_mutex);
+	soc24_grbm_select(adev, mec, pipe, queue, vmid);
+}
+
+static void unlock_srbm(struct amdgpu_device *adev)
+{
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+}
+
+static void acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t queue_id)
+{
+	uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
+	uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
+
+	lock_srbm(adev, mec, pipe, queue_id, 0);
+}
+
+static void release_queue(struct amdgpu_device *adev)
+{
+	unlock_srbm(adev);
+}
+
+static int init_interrupts_v12(struct amdgpu_device *adev, uint32_t pipe_id, uint32_t inst)
+{
+	uint32_t mec;
+	uint32_t pipe;
+
+	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
+	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
+
+	lock_srbm(adev, mec, pipe, 0, 0);
+
+	WREG32_SOC15(GC, 0, regCPC_INT_CNTL,
+		CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK |
+		CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK);
+
+	unlock_srbm(adev);
+
+	return 0;
+}
+
+static uint32_t get_sdma_rlc_reg_offset(struct amdgpu_device *adev,
+				unsigned int engine_id,
+				unsigned int queue_id)
+{
+	uint32_t sdma_engine_reg_base = 0;
+	uint32_t sdma_rlc_reg_offset;
+
+	switch (engine_id) {
+	case 0:
+		sdma_engine_reg_base = SOC15_REG_OFFSET(SDMA0, 0,
+				regSDMA0_QUEUE0_RB_CNTL) - regSDMA0_QUEUE0_RB_CNTL;
+		break;
+	case 1:
+		sdma_engine_reg_base = SOC15_REG_OFFSET(SDMA1, 0,
+				regSDMA1_QUEUE0_RB_CNTL) - regSDMA0_QUEUE0_RB_CNTL;
+		break;
+	default:
+		BUG();
+	}
+
+	sdma_rlc_reg_offset = sdma_engine_reg_base
+		+ queue_id * (regSDMA0_QUEUE1_RB_CNTL - regSDMA0_QUEUE0_RB_CNTL);
+
+	pr_debug("RLC register offset for SDMA%d RLC%d: 0x%x\n", engine_id,
+			queue_id, sdma_rlc_reg_offset);
+
+	return sdma_rlc_reg_offset;
+}
+
+static int hqd_dump_v12(struct amdgpu_device *adev,
+			uint32_t pipe_id, uint32_t queue_id,
+			uint32_t (**dump)[2], uint32_t *n_regs, uint32_t inst)
+{
+	uint32_t i = 0, reg;
+#define HQD_N_REGS 56
+#define DUMP_REG(addr) do {				\
+		if (WARN_ON_ONCE(i >= HQD_N_REGS))	\
+			break;				\
+		(*dump)[i][0] = (addr) << 2;		\
+		(*dump)[i++][1] = RREG32(addr);		\
+	} while (0)
+
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
+	if (*dump == NULL)
+		return -ENOMEM;
+
+	acquire_queue(adev, pipe_id, queue_id);
+
+	for (reg = SOC15_REG_OFFSET(GC, 0, regCP_MQD_BASE_ADDR);
+	     reg <= SOC15_REG_OFFSET(GC, 0, regCP_HQD_PQ_WPTR_HI); reg++)
+		DUMP_REG(reg);
+
+	release_queue(adev);
+
+	WARN_ON_ONCE(i != HQD_N_REGS);
+	*n_regs = i;
+
+	return 0;
+}
+
+static int hqd_sdma_dump_v12(struct amdgpu_device *adev,
+			     uint32_t engine_id, uint32_t queue_id,
+			     uint32_t (**dump)[2], uint32_t *n_regs)
+{
+	uint32_t sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev,
+			engine_id, queue_id);
+	uint32_t i = 0, reg;
+
+	const uint32_t first_reg = regSDMA0_QUEUE0_RB_CNTL;
+	const uint32_t last_reg = regSDMA0_QUEUE0_CONTEXT_STATUS;
+#undef HQD_N_REGS
+#define HQD_N_REGS (last_reg - first_reg + 1)
+
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
+	if (*dump == NULL)
+		return -ENOMEM;
+
+	for (reg = first_reg;
+	     reg <= last_reg; reg++)
+		DUMP_REG(sdma_rlc_reg_offset + reg);
+
+	WARN_ON_ONCE(i != HQD_N_REGS);
+	*n_regs = i;
+
+	return 0;
+}
+
+static int wave_control_execute_v12(struct amdgpu_device *adev,
+					uint32_t gfx_index_val,
+					uint32_t sq_cmd, uint32_t inst)
+{
+	uint32_t data = 0;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, regGRBM_GFX_INDEX), gfx_index_val);
+	WREG32(SOC15_REG_OFFSET(GC, 0, regSQ_CMD), sq_cmd);
+
+	data = REG_SET_FIELD(data, GRBM_GFX_INDEX,
+		INSTANCE_BROADCAST_WRITES, 1);
+	data = REG_SET_FIELD(data, GRBM_GFX_INDEX,
+		SA_BROADCAST_WRITES, 1);
+	data = REG_SET_FIELD(data, GRBM_GFX_INDEX,
+		SE_BROADCAST_WRITES, 1);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, regGRBM_GFX_INDEX), data);
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+/* returns TRAP_EN, EXCP_EN and EXCP_REPLACE. */
+static uint32_t kgd_gfx_v12_enable_debug_trap(struct amdgpu_device *adev,
+					    bool restore_dbg_registers,
+					    uint32_t vmid)
+{
+	uint32_t data = 0;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, 0);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, 0);
+
+	return data;
+}
+
+/* returns TRAP_EN, EXCP_EN and EXCP_REPLACE. */
+static uint32_t kgd_gfx_v12_disable_debug_trap(struct amdgpu_device *adev,
+						bool keep_trap_enabled,
+						uint32_t vmid)
+{
+	uint32_t data = 0;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, 0);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, 0);
+
+	return data;
+}
+
+static int kgd_gfx_v12_validate_trap_override_request(struct amdgpu_device *adev,
+							uint32_t trap_override,
+							uint32_t *trap_mask_supported)
+{
+	*trap_mask_supported &= KFD_DBG_TRAP_MASK_FP_INVALID |
+				KFD_DBG_TRAP_MASK_FP_INPUT_DENORMAL |
+				KFD_DBG_TRAP_MASK_FP_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_FP_OVERFLOW |
+				KFD_DBG_TRAP_MASK_FP_UNDERFLOW |
+				KFD_DBG_TRAP_MASK_FP_INEXACT |
+				KFD_DBG_TRAP_MASK_INT_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH |
+				KFD_DBG_TRAP_MASK_DBG_MEMORY_VIOLATION |
+				KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START |
+				KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END;
+
+
+	if (trap_override != KFD_DBG_TRAP_OVERRIDE_OR &&
+			trap_override != KFD_DBG_TRAP_OVERRIDE_REPLACE)
+		return -EPERM;
+
+	return 0;
+}
+
+static uint32_t trap_mask_map_sw_to_hw(uint32_t mask)
+{
+	uint32_t trap_on_start = (mask & KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START) ? 1 : 0;
+	uint32_t trap_on_end = (mask & KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END) ? 1 : 0;
+	uint32_t excp_en = mask & (KFD_DBG_TRAP_MASK_FP_INVALID |
+			KFD_DBG_TRAP_MASK_FP_INPUT_DENORMAL |
+			KFD_DBG_TRAP_MASK_FP_DIVIDE_BY_ZERO |
+			KFD_DBG_TRAP_MASK_FP_OVERFLOW |
+			KFD_DBG_TRAP_MASK_FP_UNDERFLOW |
+			KFD_DBG_TRAP_MASK_FP_INEXACT |
+			KFD_DBG_TRAP_MASK_INT_DIVIDE_BY_ZERO |
+			KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH |
+			KFD_DBG_TRAP_MASK_DBG_MEMORY_VIOLATION);
+	uint32_t ret;
+
+	ret = REG_SET_FIELD(0, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, excp_en);
+	ret = REG_SET_FIELD(ret, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_START, trap_on_start);
+	ret = REG_SET_FIELD(ret, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_END, trap_on_end);
+
+	return ret;
+}
+
+static uint32_t trap_mask_map_hw_to_sw(uint32_t mask)
+{
+	uint32_t ret = REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, EXCP_EN);
+
+	if (REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_START))
+		ret |= KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START;
+
+	if (REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_END))
+		ret |= KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END;
+
+	return ret;
+}
+
+/* returns TRAP_EN, EXCP_EN and EXCP_REPLACE. */
+static uint32_t kgd_gfx_v12_set_wave_launch_trap_override(struct amdgpu_device *adev,
+					uint32_t vmid,
+					uint32_t trap_override,
+					uint32_t trap_mask_bits,
+					uint32_t trap_mask_request,
+					uint32_t *trap_mask_prev,
+					uint32_t kfd_dbg_trap_cntl_prev)
+
+{
+	uint32_t data = 0;
+
+	*trap_mask_prev = trap_mask_map_hw_to_sw(kfd_dbg_trap_cntl_prev);
+
+	data = (trap_mask_bits & trap_mask_request) | (*trap_mask_prev & ~trap_mask_request);
+	data = trap_mask_map_sw_to_hw(data);
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, trap_override);
+
+	return data;
+}
+
+/* returns STALL_VMID or LAUNCH_MODE. */
+static uint32_t kgd_gfx_v12_set_wave_launch_mode(struct amdgpu_device *adev,
+					uint8_t wave_launch_mode,
+					uint32_t vmid)
+{
+	uint32_t data = 0;
+	bool is_stall_mode = wave_launch_mode == 4;
+
+	if (is_stall_mode)
+		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, STALL_VMID,
+									1);
+	else
+		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, LAUNCH_MODE,
+							wave_launch_mode);
+
+	return data;
+}
+
+#define TCP_WATCH_STRIDE (regTCP_WATCH1_ADDR_H - regTCP_WATCH0_ADDR_H)
+static uint32_t kgd_gfx_v12_set_address_watch(struct amdgpu_device *adev,
+					uint64_t watch_address,
+					uint32_t watch_address_mask,
+					uint32_t watch_id,
+					uint32_t watch_mode,
+					uint32_t debug_vmid,
+					uint32_t inst)
+{
+	uint32_t watch_address_high;
+	uint32_t watch_address_low;
+	uint32_t watch_address_cntl;
+
+	watch_address_cntl = 0;
+	watch_address_low = lower_32_bits(watch_address);
+	watch_address_high = upper_32_bits(watch_address) & 0xffff;
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MODE,
+			watch_mode);
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MASK,
+			watch_address_mask >> 7);
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VALID,
+			1);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_H) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_high);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_L) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_low);
+
+	return watch_address_cntl;
+}
+
+static uint32_t kgd_gfx_v12_clear_address_watch(struct amdgpu_device *adev,
+					uint32_t watch_id)
+{
+	return 0;
+}
+
+static uint32_t kgd_gfx_v12_hqd_sdma_get_doorbell(struct amdgpu_device *adev,
+						 int engine, int queue)
+{
+	return 0;
+}
+
+const struct kfd2kgd_calls gfx_v12_kfd2kgd = {
+	.init_interrupts = init_interrupts_v12,
+	.hqd_dump = hqd_dump_v12,
+	.hqd_sdma_dump = hqd_sdma_dump_v12,
+	.wave_control_execute = wave_control_execute_v12,
+	.get_atc_vmid_pasid_mapping_info = NULL,
+	.enable_debug_trap = kgd_gfx_v12_enable_debug_trap,
+	.disable_debug_trap = kgd_gfx_v12_disable_debug_trap,
+	.validate_trap_override_request = kgd_gfx_v12_validate_trap_override_request,
+	.set_wave_launch_trap_override = kgd_gfx_v12_set_wave_launch_trap_override,
+	.set_wave_launch_mode = kgd_gfx_v12_set_wave_launch_mode,
+	.set_address_watch = kgd_gfx_v12_set_address_watch,
+	.clear_address_watch = kgd_gfx_v12_clear_address_watch,
+	.hqd_sdma_get_doorbell = kgd_gfx_v12_hqd_sdma_get_doorbell
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
index e83cb1c09610..df77558e03ef 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
@@ -78,7 +78,7 @@ static void kgd_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmi
 					uint32_t sh_mem_config,
 					uint32_t sh_mem_ape1_base,
 					uint32_t sh_mem_ape1_limit,
-					uint32_t sh_mem_bases)
+					uint32_t sh_mem_bases, uint32_t inst)
 {
 	lock_srbm(adev, 0, 0, 0, vmid);
 
@@ -91,7 +91,7 @@ static void kgd_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmi
 }
 
 static int kgd_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
-					unsigned int vmid)
+					unsigned int vmid, uint32_t inst)
 {
 	/*
 	 * We have to assume that there is no outstanding mapping.
@@ -114,7 +114,8 @@ static int kgd_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
 	return 0;
 }
 
-static int kgd_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id)
+static int kgd_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t inst)
 {
 	uint32_t mec;
 	uint32_t pipe;
@@ -158,7 +159,7 @@ static inline struct cik_sdma_rlc_registers *get_sdma_mqd(void *mqd)
 static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd,
 			uint32_t pipe_id, uint32_t queue_id,
 			uint32_t __user *wptr, uint32_t wptr_shift,
-			uint32_t wptr_mask, struct mm_struct *mm)
+			uint32_t wptr_mask, struct mm_struct *mm, uint32_t inst)
 {
 	struct cik_mqd *m;
 	uint32_t *mqd_hqd;
@@ -202,7 +203,7 @@ static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd,
 
 static int kgd_hqd_dump(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
-			uint32_t (**dump)[2], uint32_t *n_regs)
+			uint32_t (**dump)[2], uint32_t *n_regs, uint32_t inst)
 {
 	uint32_t i = 0, reg;
 #define HQD_N_REGS (35+4)
@@ -213,7 +214,7 @@ static int kgd_hqd_dump(struct amdgpu_device *adev,
 		(*dump)[i++][1] = RREG32(addr);		\
 	} while (0)
 
-	*dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -300,7 +301,7 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 #undef HQD_N_REGS
 #define HQD_N_REGS (19+4)
 
-	*dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -318,7 +319,7 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 
 static bool kgd_hqd_is_occupied(struct amdgpu_device *adev,
 				uint64_t queue_address, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	uint32_t act;
 	bool retval = false;
@@ -358,7 +359,7 @@ static bool kgd_hqd_sdma_is_occupied(struct amdgpu_device *adev, void *mqd)
 static int kgd_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	uint32_t temp;
 	enum hqd_dequeue_request_type type;
@@ -494,7 +495,7 @@ static int kgd_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd,
 
 static int kgd_wave_control_execute(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
-					uint32_t sq_cmd)
+					uint32_t sq_cmd, uint32_t inst)
 {
 	uint32_t data;
 
@@ -560,6 +561,13 @@ static uint32_t read_vmid_from_vmfault_reg(struct amdgpu_device *adev)
 	return REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS, VMID);
 }
 
+static uint32_t kgd_hqd_sdma_get_doorbell(struct amdgpu_device *adev,
+					  int engine, int queue)
+
+{
+	return 0;
+}
+
 const struct kfd2kgd_calls gfx_v7_kfd2kgd = {
 	.program_sh_mem_settings = kgd_program_sh_mem_settings,
 	.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
@@ -577,4 +585,5 @@ const struct kfd2kgd_calls gfx_v7_kfd2kgd = {
 	.set_scratch_backing_va = set_scratch_backing_va,
 	.set_vm_context_page_table_base = set_vm_context_page_table_base,
 	.read_vmid_from_vmfault_reg = read_vmid_from_vmfault_reg,
+	.hqd_sdma_get_doorbell = kgd_hqd_sdma_get_doorbell,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
index 870f352837fc..e68c0fa8d751 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
@@ -72,7 +72,7 @@ static void kgd_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmi
 					uint32_t sh_mem_config,
 					uint32_t sh_mem_ape1_base,
 					uint32_t sh_mem_ape1_limit,
-					uint32_t sh_mem_bases)
+					uint32_t sh_mem_bases, uint32_t inst)
 {
 	lock_srbm(adev, 0, 0, 0, vmid);
 
@@ -85,7 +85,7 @@ static void kgd_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmi
 }
 
 static int kgd_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
-					unsigned int vmid)
+					unsigned int vmid, uint32_t inst)
 {
 	/*
 	 * We have to assume that there is no outstanding mapping.
@@ -109,7 +109,8 @@ static int kgd_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
 	return 0;
 }
 
-static int kgd_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id)
+static int kgd_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t inst)
 {
 	uint32_t mec;
 	uint32_t pipe;
@@ -153,7 +154,7 @@ static inline struct vi_sdma_mqd *get_sdma_mqd(void *mqd)
 static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd,
 			uint32_t pipe_id, uint32_t queue_id,
 			uint32_t __user *wptr, uint32_t wptr_shift,
-			uint32_t wptr_mask, struct mm_struct *mm)
+			uint32_t wptr_mask, struct mm_struct *mm, uint32_t inst)
 {
 	struct vi_mqd *m;
 	uint32_t *mqd_hqd;
@@ -226,7 +227,7 @@ static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd,
 
 static int kgd_hqd_dump(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
-			uint32_t (**dump)[2], uint32_t *n_regs)
+			uint32_t (**dump)[2], uint32_t *n_regs, uint32_t inst)
 {
 	uint32_t i = 0, reg;
 #define HQD_N_REGS (54+4)
@@ -237,7 +238,7 @@ static int kgd_hqd_dump(struct amdgpu_device *adev,
 		(*dump)[i++][1] = RREG32(addr);		\
 	} while (0)
 
-	*dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -323,7 +324,7 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 #undef HQD_N_REGS
 #define HQD_N_REGS (19+4+2+3+7)
 
-	*dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -350,7 +351,7 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 
 static bool kgd_hqd_is_occupied(struct amdgpu_device *adev,
 				uint64_t queue_address, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	uint32_t act;
 	bool retval = false;
@@ -390,7 +391,7 @@ static bool kgd_hqd_sdma_is_occupied(struct amdgpu_device *adev, void *mqd)
 static int kgd_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	uint32_t temp;
 	enum hqd_dequeue_request_type type;
@@ -540,7 +541,7 @@ static bool get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev,
 
 static int kgd_wave_control_execute(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
-					uint32_t sq_cmd)
+					uint32_t sq_cmd, uint32_t inst)
 {
 	uint32_t data = 0;
 
@@ -581,6 +582,13 @@ static void set_vm_context_page_table_base(struct amdgpu_device *adev,
 			lower_32_bits(page_table_base));
 }
 
+static uint32_t kgd_hqd_sdma_get_doorbell(struct amdgpu_device *adev,
+					  int engine, int queue)
+
+{
+	return 0;
+}
+
 const struct kfd2kgd_calls gfx_v8_kfd2kgd = {
 	.program_sh_mem_settings = kgd_program_sh_mem_settings,
 	.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
@@ -598,4 +606,5 @@ const struct kfd2kgd_calls gfx_v8_kfd2kgd = {
 			get_atc_vmid_pasid_mapping_info,
 	.set_scratch_backing_va = set_scratch_backing_va,
 	.set_vm_context_page_table_base = set_vm_context_page_table_base,
+	.hqd_sdma_get_doorbell = kgd_hqd_sdma_get_doorbell,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
index e92b93557c13..088d09cc7a72 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -38,6 +38,7 @@
 #include "soc15d.h"
 #include "gfx_v9_0.h"
 #include "amdgpu_amdkfd_gfx_v9.h"
+#include <uapi/linux/kfd_ioctl.h>
 
 enum hqd_dequeue_request_type {
 	NO_ACTION = 0,
@@ -46,29 +47,29 @@ enum hqd_dequeue_request_type {
 	SAVE_WAVES
 };
 
-static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
-			uint32_t queue, uint32_t vmid)
+static void kgd_gfx_v9_lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
+			uint32_t queue, uint32_t vmid, uint32_t inst)
 {
 	mutex_lock(&adev->srbm_mutex);
-	soc15_grbm_select(adev, mec, pipe, queue, vmid);
+	soc15_grbm_select(adev, mec, pipe, queue, vmid, GET_INST(GC, inst));
 }
 
-static void unlock_srbm(struct amdgpu_device *adev)
+static void kgd_gfx_v9_unlock_srbm(struct amdgpu_device *adev, uint32_t inst)
 {
-	soc15_grbm_select(adev, 0, 0, 0, 0);
+	soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, inst));
 	mutex_unlock(&adev->srbm_mutex);
 }
 
-static void acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id,
-				uint32_t queue_id)
+void kgd_gfx_v9_acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t queue_id, uint32_t inst)
 {
 	uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
 
-	lock_srbm(adev, mec, pipe, queue_id, 0);
+	kgd_gfx_v9_lock_srbm(adev, mec, pipe, queue_id, 0, inst);
 }
 
-static uint64_t get_queue_mask(struct amdgpu_device *adev,
+uint64_t kgd_gfx_v9_get_queue_mask(struct amdgpu_device *adev,
 			       uint32_t pipe_id, uint32_t queue_id)
 {
 	unsigned int bit = pipe_id * adev->gfx.mec.num_queue_per_pipe +
@@ -77,28 +78,28 @@ static uint64_t get_queue_mask(struct amdgpu_device *adev,
 	return 1ull << bit;
 }
 
-static void release_queue(struct amdgpu_device *adev)
+void kgd_gfx_v9_release_queue(struct amdgpu_device *adev, uint32_t inst)
 {
-	unlock_srbm(adev);
+	kgd_gfx_v9_unlock_srbm(adev, inst);
 }
 
 void kgd_gfx_v9_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmid,
 					uint32_t sh_mem_config,
 					uint32_t sh_mem_ape1_base,
 					uint32_t sh_mem_ape1_limit,
-					uint32_t sh_mem_bases)
+					uint32_t sh_mem_bases, uint32_t inst)
 {
-	lock_srbm(adev, 0, 0, 0, vmid);
+	kgd_gfx_v9_lock_srbm(adev, 0, 0, 0, vmid, inst);
 
-	WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_CONFIG), sh_mem_config);
-	WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_BASES), sh_mem_bases);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmSH_MEM_CONFIG, sh_mem_config);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmSH_MEM_BASES, sh_mem_bases);
 	/* APE1 no longer exists on GFX9 */
 
-	unlock_srbm(adev);
+	kgd_gfx_v9_unlock_srbm(adev, inst);
 }
 
 int kgd_gfx_v9_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
-					unsigned int vmid)
+					unsigned int vmid, uint32_t inst)
 {
 	/*
 	 * We have to assume that there is no outstanding mapping.
@@ -156,7 +157,8 @@ int kgd_gfx_v9_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
  * but still works
  */
 
-int kgd_gfx_v9_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id)
+int kgd_gfx_v9_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t inst)
 {
 	uint32_t mec;
 	uint32_t pipe;
@@ -164,13 +166,13 @@ int kgd_gfx_v9_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id)
 	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
 
-	lock_srbm(adev, mec, pipe, 0, 0);
+	kgd_gfx_v9_lock_srbm(adev, mec, pipe, 0, 0, inst);
 
-	WREG32_SOC15(GC, 0, mmCPC_INT_CNTL,
+	WREG32_SOC15(GC, GET_INST(GC, inst), mmCPC_INT_CNTL,
 		CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK |
 		CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK);
 
-	unlock_srbm(adev);
+	kgd_gfx_v9_unlock_srbm(adev, inst);
 
 	return 0;
 }
@@ -220,7 +222,8 @@ static inline struct v9_sdma_mqd *get_sdma_mqd(void *mqd)
 int kgd_gfx_v9_hqd_load(struct amdgpu_device *adev, void *mqd,
 			uint32_t pipe_id, uint32_t queue_id,
 			uint32_t __user *wptr, uint32_t wptr_shift,
-			uint32_t wptr_mask, struct mm_struct *mm)
+			uint32_t wptr_mask, struct mm_struct *mm,
+			uint32_t inst)
 {
 	struct v9_mqd *m;
 	uint32_t *mqd_hqd;
@@ -228,21 +231,21 @@ int kgd_gfx_v9_hqd_load(struct amdgpu_device *adev, void *mqd,
 
 	m = get_mqd(mqd);
 
-	acquire_queue(adev, pipe_id, queue_id);
+	kgd_gfx_v9_acquire_queue(adev, pipe_id, queue_id, inst);
 
 	/* HQD registers extend from CP_MQD_BASE_ADDR to CP_HQD_EOP_WPTR_MEM. */
 	mqd_hqd = &m->cp_mqd_base_addr_lo;
-	hqd_base = SOC15_REG_OFFSET(GC, 0, mmCP_MQD_BASE_ADDR);
+	hqd_base = SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_MQD_BASE_ADDR);
 
 	for (reg = hqd_base;
-	     reg <= SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI); reg++)
-		WREG32_RLC(reg, mqd_hqd[reg - hqd_base]);
+	     reg <= SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_HQD_PQ_WPTR_HI); reg++)
+		WREG32_XCC(reg, mqd_hqd[reg - hqd_base], inst);
 
 
 	/* Activate doorbell logic before triggering WPTR poll. */
 	data = REG_SET_FIELD(m->cp_hqd_pq_doorbell_control,
 			     CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1);
-	WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL), data);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmCP_HQD_PQ_DOORBELL_CONTROL, data);
 
 	if (wptr) {
 		/* Don't read wptr with get_user because the user
@@ -271,43 +274,42 @@ int kgd_gfx_v9_hqd_load(struct amdgpu_device *adev, void *mqd,
 		guessed_wptr += m->cp_hqd_pq_wptr_lo & ~(queue_size - 1);
 		guessed_wptr += (uint64_t)m->cp_hqd_pq_wptr_hi << 32;
 
-		WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_LO),
-		       lower_32_bits(guessed_wptr));
-		WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI),
-		       upper_32_bits(guessed_wptr));
-		WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR),
-		       lower_32_bits((uintptr_t)wptr));
-		WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR_HI),
-		       upper_32_bits((uintptr_t)wptr));
-		WREG32_SOC15(GC, 0, mmCP_PQ_WPTR_POLL_CNTL1,
-		       (uint32_t)get_queue_mask(adev, pipe_id, queue_id));
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmCP_HQD_PQ_WPTR_LO,
+			lower_32_bits(guessed_wptr));
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmCP_HQD_PQ_WPTR_HI,
+			upper_32_bits(guessed_wptr));
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmCP_HQD_PQ_WPTR_POLL_ADDR,
+			lower_32_bits((uintptr_t)wptr));
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmCP_HQD_PQ_WPTR_POLL_ADDR_HI,
+			upper_32_bits((uintptr_t)wptr));
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmCP_PQ_WPTR_POLL_CNTL1,
+			(uint32_t)kgd_gfx_v9_get_queue_mask(adev, pipe_id, queue_id));
 	}
 
 	/* Start the EOP fetcher */
-	WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_EOP_RPTR),
-	       REG_SET_FIELD(m->cp_hqd_eop_rptr,
-			     CP_HQD_EOP_RPTR, INIT_FETCHER, 1));
+	WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmCP_HQD_EOP_RPTR,
+	       REG_SET_FIELD(m->cp_hqd_eop_rptr, CP_HQD_EOP_RPTR, INIT_FETCHER, 1));
 
 	data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);
-	WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_ACTIVE), data);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmCP_HQD_ACTIVE, data);
 
-	release_queue(adev);
+	kgd_gfx_v9_release_queue(adev, inst);
 
 	return 0;
 }
 
 int kgd_gfx_v9_hiq_mqd_load(struct amdgpu_device *adev, void *mqd,
 			    uint32_t pipe_id, uint32_t queue_id,
-			    uint32_t doorbell_off)
+			    uint32_t doorbell_off, uint32_t inst)
 {
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
+	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq[inst].ring;
 	struct v9_mqd *m;
 	uint32_t mec, pipe;
 	int r;
 
 	m = get_mqd(mqd);
 
-	acquire_queue(adev, pipe_id, queue_id);
+	kgd_gfx_v9_acquire_queue(adev, pipe_id, queue_id, inst);
 
 	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
@@ -315,7 +317,7 @@ int kgd_gfx_v9_hiq_mqd_load(struct amdgpu_device *adev, void *mqd,
 	pr_debug("kfd: set HIQ, mec:%d, pipe:%d, queue:%d.\n",
 		 mec, pipe, queue_id);
 
-	spin_lock(&adev->gfx.kiq.ring_lock);
+	spin_lock(&adev->gfx.kiq[inst].ring_lock);
 	r = amdgpu_ring_alloc(kiq_ring, 7);
 	if (r) {
 		pr_err("Failed to alloc KIQ (%d).\n", r);
@@ -342,15 +344,15 @@ int kgd_gfx_v9_hiq_mqd_load(struct amdgpu_device *adev, void *mqd,
 	amdgpu_ring_commit(kiq_ring);
 
 out_unlock:
-	spin_unlock(&adev->gfx.kiq.ring_lock);
-	release_queue(adev);
+	spin_unlock(&adev->gfx.kiq[inst].ring_lock);
+	kgd_gfx_v9_release_queue(adev, inst);
 
 	return r;
 }
 
 int kgd_gfx_v9_hqd_dump(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
-			uint32_t (**dump)[2], uint32_t *n_regs)
+			uint32_t (**dump)[2], uint32_t *n_regs, uint32_t inst)
 {
 	uint32_t i = 0, reg;
 #define HQD_N_REGS 56
@@ -361,17 +363,17 @@ int kgd_gfx_v9_hqd_dump(struct amdgpu_device *adev,
 		(*dump)[i++][1] = RREG32(addr);		\
 	} while (0)
 
-	*dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
-	acquire_queue(adev, pipe_id, queue_id);
+	kgd_gfx_v9_acquire_queue(adev, pipe_id, queue_id, inst);
 
-	for (reg = SOC15_REG_OFFSET(GC, 0, mmCP_MQD_BASE_ADDR);
-	     reg <= SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI); reg++)
+	for (reg = SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_MQD_BASE_ADDR);
+	     reg <= SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_HQD_PQ_WPTR_HI); reg++)
 		DUMP_REG(reg);
 
-	release_queue(adev);
+	kgd_gfx_v9_release_queue(adev, inst);
 
 	WARN_ON_ONCE(i != HQD_N_REGS);
 	*n_regs = i;
@@ -458,7 +460,7 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 #undef HQD_N_REGS
 #define HQD_N_REGS (19+6+7+10)
 
-	*dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -481,23 +483,23 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 
 bool kgd_gfx_v9_hqd_is_occupied(struct amdgpu_device *adev,
 				uint64_t queue_address, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	uint32_t act;
 	bool retval = false;
 	uint32_t low, high;
 
-	acquire_queue(adev, pipe_id, queue_id);
-	act = RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE);
+	kgd_gfx_v9_acquire_queue(adev, pipe_id, queue_id, inst);
+	act = RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_ACTIVE);
 	if (act) {
 		low = lower_32_bits(queue_address >> 8);
 		high = upper_32_bits(queue_address >> 8);
 
-		if (low == RREG32_SOC15(GC, 0, mmCP_HQD_PQ_BASE) &&
-		   high == RREG32_SOC15(GC, 0, mmCP_HQD_PQ_BASE_HI))
+		if (low == RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_PQ_BASE) &&
+		   high == RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_PQ_BASE_HI))
 			retval = true;
 	}
-	release_queue(adev);
+	kgd_gfx_v9_release_queue(adev, inst);
 	return retval;
 }
 
@@ -522,7 +524,7 @@ static bool kgd_hqd_sdma_is_occupied(struct amdgpu_device *adev, void *mqd)
 int kgd_gfx_v9_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	enum hqd_dequeue_request_type type;
 	unsigned long end_jiffies;
@@ -532,10 +534,10 @@ int kgd_gfx_v9_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 	if (amdgpu_in_reset(adev))
 		return -EIO;
 
-	acquire_queue(adev, pipe_id, queue_id);
+	kgd_gfx_v9_acquire_queue(adev, pipe_id, queue_id, inst);
 
 	if (m->cp_hqd_vmid == 0)
-		WREG32_FIELD15_RLC(GC, 0, RLC_CP_SCHEDULERS, scheduler1, 0);
+		WREG32_FIELD15_RLC(GC, GET_INST(GC, inst), RLC_CP_SCHEDULERS, scheduler1, 0);
 
 	switch (reset_type) {
 	case KFD_PREEMPT_TYPE_WAVEFRONT_DRAIN:
@@ -552,22 +554,22 @@ int kgd_gfx_v9_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 		break;
 	}
 
-	WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_DEQUEUE_REQUEST), type);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmCP_HQD_DEQUEUE_REQUEST, type);
 
 	end_jiffies = (utimeout * HZ / 1000) + jiffies;
 	while (true) {
-		temp = RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE);
+		temp = RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_ACTIVE);
 		if (!(temp & CP_HQD_ACTIVE__ACTIVE_MASK))
 			break;
 		if (time_after(jiffies, end_jiffies)) {
 			pr_err("cp queue preemption time out.\n");
-			release_queue(adev);
+			kgd_gfx_v9_release_queue(adev, inst);
 			return -ETIME;
 		}
 		usleep_range(500, 1000);
 	}
 
-	release_queue(adev);
+	kgd_gfx_v9_release_queue(adev, inst);
 	return 0;
 }
 
@@ -624,14 +626,14 @@ bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev,
 
 int kgd_gfx_v9_wave_control_execute(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
-					uint32_t sq_cmd)
+					uint32_t sq_cmd, uint32_t inst)
 {
 	uint32_t data = 0;
 
 	mutex_lock(&adev->grbm_idx_mutex);
 
-	WREG32_SOC15_RLC_SHADOW(GC, 0, mmGRBM_GFX_INDEX, gfx_index_val);
-	WREG32_SOC15(GC, 0, mmSQ_CMD, sq_cmd);
+	WREG32_SOC15_RLC_SHADOW(GC, GET_INST(GC, inst), mmGRBM_GFX_INDEX, gfx_index_val);
+	WREG32_SOC15(GC, GET_INST(GC, inst), mmSQ_CMD, sq_cmd);
 
 	data = REG_SET_FIELD(data, GRBM_GFX_INDEX,
 		INSTANCE_BROADCAST_WRITES, 1);
@@ -640,12 +642,274 @@ int kgd_gfx_v9_wave_control_execute(struct amdgpu_device *adev,
 	data = REG_SET_FIELD(data, GRBM_GFX_INDEX,
 		SE_BROADCAST_WRITES, 1);
 
-	WREG32_SOC15_RLC_SHADOW(GC, 0, mmGRBM_GFX_INDEX, data);
+	WREG32_SOC15_RLC_SHADOW(GC, GET_INST(GC, inst), mmGRBM_GFX_INDEX, data);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	return 0;
 }
 
+/*
+ * GFX9 helper for wave launch stall requirements on debug trap setting.
+ *
+ * vmid:
+ *   Target VMID to stall/unstall.
+ *
+ * stall:
+ *   0-unstall wave launch (enable), 1-stall wave launch (disable).
+ *   After wavefront launch has been stalled, allocated waves must drain from
+ *   SPI in order for debug trap settings to take effect on those waves.
+ *   This is roughly a ~96 clock cycle wait on SPI where a read on
+ *   SPI_GDBG_WAVE_CNTL translates to ~32 clock cycles.
+ *   KGD_GFX_V9_WAVE_LAUNCH_SPI_DRAIN_LATENCY indicates the number of reads required.
+ *
+ *   NOTE: We can afford to clear the entire STALL_VMID field on unstall
+ *   because GFX9.4.1 cannot support multi-process debugging due to trap
+ *   configuration and masking being limited to global scope.  Always assume
+ *   single process conditions.
+ */
+#define KGD_GFX_V9_WAVE_LAUNCH_SPI_DRAIN_LATENCY	3
+void kgd_gfx_v9_set_wave_launch_stall(struct amdgpu_device *adev,
+					uint32_t vmid,
+					bool stall)
+{
+	int i;
+	uint32_t data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1))
+		data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_VMID,
+							stall ? 1 << vmid : 0);
+	else
+		data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_RA,
+							stall ? 1 : 0);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);
+
+	if (!stall)
+		return;
+
+	for (i = 0; i < KGD_GFX_V9_WAVE_LAUNCH_SPI_DRAIN_LATENCY; i++)
+		RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+}
+
+/*
+ * restore_dbg_registers is ignored here but is a general interface requirement
+ * for devices that support GFXOFF and where the RLC save/restore list
+ * does not support hw registers for debugging i.e. the driver has to manually
+ * initialize the debug mode registers after it has disabled GFX off during the
+ * debug session.
+ */
+uint32_t kgd_gfx_v9_enable_debug_trap(struct amdgpu_device *adev,
+				bool restore_dbg_registers,
+				uint32_t vmid)
+{
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+/*
+ * keep_trap_enabled is ignored here but is a general interface requirement
+ * for devices that support multi-process debugging where the performance
+ * overhead from trap temporary setup needs to be bypassed when the debug
+ * session has ended.
+ */
+uint32_t kgd_gfx_v9_disable_debug_trap(struct amdgpu_device *adev,
+					bool keep_trap_enabled,
+					uint32_t vmid)
+{
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+int kgd_gfx_v9_validate_trap_override_request(struct amdgpu_device *adev,
+					uint32_t trap_override,
+					uint32_t *trap_mask_supported)
+{
+	*trap_mask_supported &= KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH;
+
+	/* The SPI_GDBG_TRAP_MASK register is global and affects all
+	 * processes. Only allow OR-ing the address-watch bit, since
+	 * this only affects processes under the debugger. Other bits
+	 * should stay 0 to avoid the debugger interfering with other
+	 * processes.
+	 */
+	if (trap_override != KFD_DBG_TRAP_OVERRIDE_OR)
+		return -EINVAL;
+
+	return 0;
+}
+
+uint32_t kgd_gfx_v9_set_wave_launch_trap_override(struct amdgpu_device *adev,
+					     uint32_t vmid,
+					     uint32_t trap_override,
+					     uint32_t trap_mask_bits,
+					     uint32_t trap_mask_request,
+					     uint32_t *trap_mask_prev,
+					     uint32_t kfd_dbg_cntl_prev)
+{
+	uint32_t data, wave_cntl_prev;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	wave_cntl_prev = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK));
+	*trap_mask_prev = REG_GET_FIELD(data, SPI_GDBG_TRAP_MASK, EXCP_EN);
+
+	trap_mask_bits = (trap_mask_bits & trap_mask_request) |
+		(*trap_mask_prev & ~trap_mask_request);
+
+	data = REG_SET_FIELD(data, SPI_GDBG_TRAP_MASK, EXCP_EN, trap_mask_bits);
+	data = REG_SET_FIELD(data, SPI_GDBG_TRAP_MASK, REPLACE, trap_override);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), data);
+
+	/* We need to preserve wave launch mode stall settings. */
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), wave_cntl_prev);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+uint32_t kgd_gfx_v9_set_wave_launch_mode(struct amdgpu_device *adev,
+					uint8_t wave_launch_mode,
+					uint32_t vmid)
+{
+	uint32_t data = 0;
+	bool is_mode_set = !!wave_launch_mode;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL2,
+		VMID_MASK, is_mode_set ? 1 << vmid : 0);
+	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL2,
+		MODE, is_mode_set ? wave_launch_mode : 0);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL2), data);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+#define TCP_WATCH_STRIDE (mmTCP_WATCH1_ADDR_H - mmTCP_WATCH0_ADDR_H)
+uint32_t kgd_gfx_v9_set_address_watch(struct amdgpu_device *adev,
+					uint64_t watch_address,
+					uint32_t watch_address_mask,
+					uint32_t watch_id,
+					uint32_t watch_mode,
+					uint32_t debug_vmid,
+					uint32_t inst)
+{
+	uint32_t watch_address_high;
+	uint32_t watch_address_low;
+	uint32_t watch_address_cntl;
+
+	watch_address_cntl = 0;
+
+	watch_address_low = lower_32_bits(watch_address);
+	watch_address_high = upper_32_bits(watch_address) & 0xffff;
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VMID,
+			debug_vmid);
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MODE,
+			watch_mode);
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MASK,
+			watch_address_mask >> 6);
+
+	/* Turning off this watch point until we set all the registers */
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VALID,
+			0);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_cntl);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_ADDR_H) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_high);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_ADDR_L) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_low);
+
+	/* Enable the watch point */
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VALID,
+			1);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_cntl);
+
+	return 0;
+}
+
+uint32_t kgd_gfx_v9_clear_address_watch(struct amdgpu_device *adev,
+					uint32_t watch_id)
+{
+	uint32_t watch_address_cntl;
+
+	watch_address_cntl = 0;
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_cntl);
+
+	return 0;
+}
+
+/* kgd_gfx_v9_get_iq_wait_times: Returns the mmCP_IQ_WAIT_TIME1/2 values
+ * The values read are:
+ *     ib_offload_wait_time     -- Wait Count for Indirect Buffer Offloads.
+ *     atomic_offload_wait_time -- Wait Count for L2 and GDS Atomics Offloads.
+ *     wrm_offload_wait_time    -- Wait Count for WAIT_REG_MEM Offloads.
+ *     gws_wait_time            -- Wait Count for Global Wave Syncs.
+ *     que_sleep_wait_time      -- Wait Count for Dequeue Retry.
+ *     sch_wave_wait_time       -- Wait Count for Scheduling Wave Message.
+ *     sem_rearm_wait_time      -- Wait Count for Semaphore re-arm.
+ *     deq_retry_wait_time      -- Wait Count for Global Wave Syncs.
+ */
+void kgd_gfx_v9_get_iq_wait_times(struct amdgpu_device *adev,
+					uint32_t *wait_times,
+					uint32_t inst)
+
+{
+	*wait_times = RREG32_SOC15_RLC(GC, GET_INST(GC, inst),
+			mmCP_IQ_WAIT_TIME2);
+}
+
 void kgd_gfx_v9_set_vm_context_page_table_base(struct amdgpu_device *adev,
 			uint32_t vmid, uint64_t page_table_base)
 {
@@ -680,33 +944,34 @@ static void unlock_spi_csq_mutexes(struct amdgpu_device *adev)
  *
  * @adev: Handle of device whose registers are to be read
  * @queue_idx: Index of queue in the queue-map bit-field
- * @wave_cnt: Output parameter updated with number of waves in flight
- * @vmid: Output parameter updated with VMID of queue whose wave count
- * is being collected
+ * @queue_cnt: Stores the wave count and doorbell offset for an active queue
+ * @inst: xcc's instance number on a multi-XCC setup
  */
 static void get_wave_count(struct amdgpu_device *adev, int queue_idx,
-		int *wave_cnt, int *vmid)
+		struct kfd_cu_occupancy *queue_cnt, uint32_t inst)
 {
 	int pipe_idx;
 	int queue_slot;
 	unsigned int reg_val;
-
+	unsigned int wave_cnt;
 	/*
 	 * Program GRBM with appropriate MEID, PIPEID, QUEUEID and VMID
 	 * parameters to read out waves in flight. Get VMID if there are
 	 * non-zero waves in flight.
 	 */
-	*vmid = 0xFF;
-	*wave_cnt = 0;
 	pipe_idx = queue_idx / adev->gfx.mec.num_queue_per_pipe;
 	queue_slot = queue_idx % adev->gfx.mec.num_queue_per_pipe;
-	soc15_grbm_select(adev, 1, pipe_idx, queue_slot, 0);
-	reg_val = RREG32_SOC15_IP(GC, SOC15_REG_OFFSET(GC, 0, mmSPI_CSQ_WF_ACTIVE_COUNT_0) +
-			 queue_slot);
-	*wave_cnt = reg_val & SPI_CSQ_WF_ACTIVE_COUNT_0__COUNT_MASK;
-	if (*wave_cnt != 0)
-		*vmid = (RREG32_SOC15(GC, 0, mmCP_HQD_VMID) &
-			 CP_HQD_VMID__VMID_MASK) >> CP_HQD_VMID__VMID__SHIFT;
+	soc15_grbm_select(adev, 1, pipe_idx, queue_slot, 0, GET_INST(GC, inst));
+	reg_val = RREG32_SOC15_IP(GC, SOC15_REG_OFFSET(GC, GET_INST(GC, inst),
+				  mmSPI_CSQ_WF_ACTIVE_COUNT_0) + queue_slot);
+	wave_cnt = reg_val & SPI_CSQ_WF_ACTIVE_COUNT_0__COUNT_MASK;
+	if (wave_cnt != 0) {
+		queue_cnt->wave_cnt += wave_cnt;
+		queue_cnt->doorbell_off =
+			(RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_PQ_DOORBELL_CONTROL) &
+			 CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET_MASK) >>
+			 CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET__SHIFT;
+	}
 }
 
 /**
@@ -716,11 +981,11 @@ static void get_wave_count(struct amdgpu_device *adev, int queue_idx,
  * or more queues running and submitting waves to compute units.
  *
  * @adev: Handle of device from which to get number of waves in flight
- * @pasid: Identifies the process for which this query call is invoked
- * @pasid_wave_cnt: Output parameter updated with number of waves in flight that
- * belong to process with given pasid
+ * @cu_occupancy: Array that gets filled with wave_cnt and doorbell offset
+ *		  for comparison later.
  * @max_waves_per_cu: Output parameter updated with maximum number of waves
- * possible per Compute Unit
+ *                    possible per Compute Unit
+ * @inst: xcc's instance number on a multi-XCC setup
  *
  * Note: It's possible that the device has too many queues (oversubscription)
  * in which case a VMID could be remapped to a different PASID. This could lead
@@ -745,113 +1010,224 @@ static void get_wave_count(struct amdgpu_device *adev, int queue_idx,
  *    number of waves that are in flight for the queue at specified index. The
  *    index ranges from 0 to 7.
  *
- *    If non-zero waves are in flight, read CP_HQD_VMID register to obtain VMID
- *    of the wave(s).
+ *    If non-zero waves are in flight, store the corresponding doorbell offset
+ *    of the queue, along with the wave count.
  *
- *    Determine if VMID from above step maps to pasid provided as parameter. If
- *    it matches agrregate the wave count. That the VMID will not match pasid is
- *    a normal condition i.e. a device is expected to support multiple queues
- *    from multiple proceses.
+ *    Determine if the queue belongs to the process by comparing the doorbell
+ *    offset against the process's queues. If it matches, aggregate the wave
+ *    count for the process.
  *
  *  Reading registers referenced above involves programming GRBM appropriately
  */
-void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev, int pasid,
-		int *pasid_wave_cnt, int *max_waves_per_cu)
+void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev,
+				 struct kfd_cu_occupancy *cu_occupancy,
+				 int *max_waves_per_cu, uint32_t inst)
 {
 	int qidx;
-	int vmid;
 	int se_idx;
-	int sh_idx;
 	int se_cnt;
-	int sh_cnt;
-	int wave_cnt;
 	int queue_map;
-	int pasid_tmp;
 	int max_queue_cnt;
-	int vmid_wave_cnt = 0;
-	DECLARE_BITMAP(cp_queue_bitmap, KGD_MAX_QUEUES);
+	DECLARE_BITMAP(cp_queue_bitmap, AMDGPU_MAX_QUEUES);
 
 	lock_spi_csq_mutexes(adev);
-	soc15_grbm_select(adev, 1, 0, 0, 0);
+	soc15_grbm_select(adev, 1, 0, 0, 0, GET_INST(GC, inst));
 
 	/*
 	 * Iterate through the shader engines and arrays of the device
 	 * to get number of waves in flight
 	 */
-	bitmap_complement(cp_queue_bitmap, adev->gfx.mec.queue_bitmap,
-			  KGD_MAX_QUEUES);
+	bitmap_complement(cp_queue_bitmap, adev->gfx.mec_bitmap[0].queue_bitmap,
+			  AMDGPU_MAX_QUEUES);
 	max_queue_cnt = adev->gfx.mec.num_pipe_per_mec *
 			adev->gfx.mec.num_queue_per_pipe;
-	sh_cnt = adev->gfx.config.max_sh_per_se;
 	se_cnt = adev->gfx.config.max_shader_engines;
 	for (se_idx = 0; se_idx < se_cnt; se_idx++) {
-		for (sh_idx = 0; sh_idx < sh_cnt; sh_idx++) {
+		amdgpu_gfx_select_se_sh(adev, se_idx, 0, 0xffffffff, inst);
+		queue_map = RREG32_SOC15(GC, GET_INST(GC, inst), mmSPI_CSQ_WF_ACTIVE_STATUS);
+
+		/*
+		 * Assumption: queue map encodes following schema: four
+		 * pipes per each micro-engine, with each pipe mapping
+		 * eight queues. This schema is true for GFX9 devices
+		 * and must be verified for newer device families
+		 */
+		for (qidx = 0; qidx < max_queue_cnt; qidx++) {
+			/* Skip qeueus that are not associated with
+			 * compute functions
+			 */
+			if (!test_bit(qidx, cp_queue_bitmap))
+				continue;
 
-			amdgpu_gfx_select_se_sh(adev, se_idx, sh_idx, 0xffffffff);
-			queue_map = RREG32_SOC15(GC, 0, mmSPI_CSQ_WF_ACTIVE_STATUS);
+			if (!(queue_map & (1 << qidx)))
+				continue;
 
-			/*
-			 * Assumption: queue map encodes following schema: four
-			 * pipes per each micro-engine, with each pipe mapping
-			 * eight queues. This schema is true for GFX9 devices
-			 * and must be verified for newer device families
-			 */
-			for (qidx = 0; qidx < max_queue_cnt; qidx++) {
-
-				/* Skip qeueus that are not associated with
-				 * compute functions
-				 */
-				if (!test_bit(qidx, cp_queue_bitmap))
-					continue;
-
-				if (!(queue_map & (1 << qidx)))
-					continue;
-
-				/* Get number of waves in flight and aggregate them */
-				get_wave_count(adev, qidx, &wave_cnt, &vmid);
-				if (wave_cnt != 0) {
-					pasid_tmp =
-					  RREG32(SOC15_REG_OFFSET(OSSSYS, 0,
-						 mmIH_VMID_0_LUT) + vmid);
-					if (pasid_tmp == pasid)
-						vmid_wave_cnt += wave_cnt;
-				}
-			}
+			/* Get number of waves in flight and aggregate them */
+			get_wave_count(adev, qidx, &cu_occupancy[qidx],
+					inst);
 		}
 	}
 
-	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
-	soc15_grbm_select(adev, 0, 0, 0, 0);
+	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, inst);
+	soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, inst));
 	unlock_spi_csq_mutexes(adev);
 
 	/* Update the output parameters and return */
-	*pasid_wave_cnt = vmid_wave_cnt;
 	*max_waves_per_cu = adev->gfx.cu_info.simd_per_cu *
 				adev->gfx.cu_info.max_waves_per_simd;
 }
 
+void kgd_gfx_v9_build_dequeue_wait_counts_packet_info(struct amdgpu_device *adev,
+		uint32_t wait_times,
+		uint32_t sch_wave,
+		uint32_t que_sleep,
+		uint32_t *reg_offset,
+		uint32_t *reg_data)
+{
+	*reg_data = wait_times;
+
+	if (sch_wave)
+		*reg_data = REG_SET_FIELD(*reg_data,
+				CP_IQ_WAIT_TIME2,
+				SCH_WAVE,
+				sch_wave);
+	if (que_sleep)
+		*reg_data = REG_SET_FIELD(*reg_data,
+				CP_IQ_WAIT_TIME2,
+				QUE_SLEEP,
+				que_sleep);
+
+	*reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2);
+}
+
 void kgd_gfx_v9_program_trap_handler_settings(struct amdgpu_device *adev,
-                        uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr)
+		uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr, uint32_t inst)
 {
-	lock_srbm(adev, 0, 0, 0, vmid);
+	kgd_gfx_v9_lock_srbm(adev, 0, 0, 0, vmid, inst);
 
 	/*
 	 * Program TBA registers
 	 */
-	WREG32_SOC15(GC, 0, mmSQ_SHADER_TBA_LO,
-                        lower_32_bits(tba_addr >> 8));
-	WREG32_SOC15(GC, 0, mmSQ_SHADER_TBA_HI,
-                        upper_32_bits(tba_addr >> 8));
+	WREG32_SOC15(GC, GET_INST(GC, inst), mmSQ_SHADER_TBA_LO,
+			lower_32_bits(tba_addr >> 8));
+	WREG32_SOC15(GC, GET_INST(GC, inst), mmSQ_SHADER_TBA_HI,
+			upper_32_bits(tba_addr >> 8));
 
 	/*
 	 * Program TMA registers
 	 */
-	WREG32_SOC15(GC, 0, mmSQ_SHADER_TMA_LO,
+	WREG32_SOC15(GC, GET_INST(GC, inst), mmSQ_SHADER_TMA_LO,
 			lower_32_bits(tma_addr >> 8));
-	WREG32_SOC15(GC, 0, mmSQ_SHADER_TMA_HI,
+	WREG32_SOC15(GC, GET_INST(GC, inst), mmSQ_SHADER_TMA_HI,
 			upper_32_bits(tma_addr >> 8));
 
-	unlock_srbm(adev);
+	kgd_gfx_v9_unlock_srbm(adev, inst);
+}
+
+uint64_t kgd_gfx_v9_hqd_get_pq_addr(struct amdgpu_device *adev,
+				    uint32_t pipe_id, uint32_t queue_id,
+				    uint32_t inst)
+{
+	uint32_t low, high;
+	uint64_t queue_addr = 0;
+
+	kgd_gfx_v9_acquire_queue(adev, pipe_id, queue_id, inst);
+	amdgpu_gfx_rlc_enter_safe_mode(adev, inst);
+
+	if (!RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_ACTIVE))
+		goto unlock_out;
+
+	low = RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_PQ_BASE);
+	high = RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_PQ_BASE_HI);
+
+	/* only concerned with user queues. */
+	if (!high)
+		goto unlock_out;
+
+	queue_addr = (((queue_addr | high) << 32) | low) << 8;
+
+unlock_out:
+	amdgpu_gfx_rlc_exit_safe_mode(adev, inst);
+	kgd_gfx_v9_release_queue(adev, inst);
+
+	return queue_addr;
+}
+
+/* assume queue acquired  */
+static int kgd_gfx_v9_hqd_dequeue_wait(struct amdgpu_device *adev, uint32_t inst,
+				       unsigned int utimeout)
+{
+	unsigned long end_jiffies = (utimeout * HZ / 1000) + jiffies;
+
+	while (true) {
+		uint32_t temp = RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_ACTIVE);
+
+		if (!(temp & CP_HQD_ACTIVE__ACTIVE_MASK))
+			return 0;
+
+		if (time_after(jiffies, end_jiffies))
+			return -ETIME;
+
+		usleep_range(500, 1000);
+	}
+}
+
+uint64_t kgd_gfx_v9_hqd_reset(struct amdgpu_device *adev,
+			      uint32_t pipe_id, uint32_t queue_id,
+			      uint32_t inst, unsigned int utimeout)
+{
+	uint32_t low, high, pipe_reset_data = 0;
+	uint64_t queue_addr = 0;
+
+	kgd_gfx_v9_acquire_queue(adev, pipe_id, queue_id, inst);
+	amdgpu_gfx_rlc_enter_safe_mode(adev, inst);
+
+	if (!RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_ACTIVE))
+		goto unlock_out;
+
+	low = RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_PQ_BASE);
+	high = RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_PQ_BASE_HI);
+
+	/* only concerned with user queues. */
+	if (!high)
+		goto unlock_out;
+
+	queue_addr = (((queue_addr | high) << 32) | low) << 8;
+
+	pr_debug("Attempting queue reset on XCC %i pipe id %i queue id %i\n",
+		 inst, pipe_id, queue_id);
+
+	/* assume previous dequeue request issued will take affect after reset */
+	WREG32_SOC15(GC, GET_INST(GC, inst), mmSPI_COMPUTE_QUEUE_RESET, 0x1);
+
+	if (!kgd_gfx_v9_hqd_dequeue_wait(adev, inst, utimeout))
+		goto unlock_out;
+
+	pr_debug("Attempting pipe reset on XCC %i pipe id %i\n", inst, pipe_id);
+
+	pipe_reset_data = REG_SET_FIELD(pipe_reset_data, CP_MEC_CNTL, MEC_ME1_PIPE0_RESET, 1);
+	pipe_reset_data = pipe_reset_data << pipe_id;
+
+	WREG32_SOC15(GC, GET_INST(GC, inst), mmCP_MEC_CNTL, pipe_reset_data);
+	WREG32_SOC15(GC, GET_INST(GC, inst), mmCP_MEC_CNTL, 0);
+
+	if (kgd_gfx_v9_hqd_dequeue_wait(adev, inst, utimeout))
+		queue_addr = 0;
+
+unlock_out:
+	pr_debug("queue reset on XCC %i pipe id %i queue id %i %s\n",
+		 inst, pipe_id, queue_id, !!queue_addr ? "succeeded!" : "failed!");
+	amdgpu_gfx_rlc_exit_safe_mode(adev, inst);
+	kgd_gfx_v9_release_queue(adev, inst);
+
+	return queue_addr;
+}
+
+uint32_t kgd_gfx_v9_hqd_sdma_get_doorbell(struct amdgpu_device *adev,
+					  int engine, int queue)
+
+{
+	return 0;
 }
 
 const struct kfd2kgd_calls gfx_v9_kfd2kgd = {
@@ -871,6 +1247,18 @@ const struct kfd2kgd_calls gfx_v9_kfd2kgd = {
 	.get_atc_vmid_pasid_mapping_info =
 			kgd_gfx_v9_get_atc_vmid_pasid_mapping_info,
 	.set_vm_context_page_table_base = kgd_gfx_v9_set_vm_context_page_table_base,
+	.enable_debug_trap = kgd_gfx_v9_enable_debug_trap,
+	.disable_debug_trap = kgd_gfx_v9_disable_debug_trap,
+	.validate_trap_override_request = kgd_gfx_v9_validate_trap_override_request,
+	.set_wave_launch_trap_override = kgd_gfx_v9_set_wave_launch_trap_override,
+	.set_wave_launch_mode = kgd_gfx_v9_set_wave_launch_mode,
+	.set_address_watch = kgd_gfx_v9_set_address_watch,
+	.clear_address_watch = kgd_gfx_v9_clear_address_watch,
+	.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
+	.build_dequeue_wait_counts_packet_info = kgd_gfx_v9_build_dequeue_wait_counts_packet_info,
 	.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
 	.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings,
+	.hqd_get_pq_addr = kgd_gfx_v9_hqd_get_pq_addr,
+	.hqd_reset = kgd_gfx_v9_hqd_reset,
+	.hqd_sdma_get_doorbell = kgd_gfx_v9_hqd_sdma_get_doorbell
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
index c7ed3bc9053c..704452ca62f8 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
@@ -20,41 +20,97 @@
  * OTHER DEALINGS IN THE SOFTWARE.
  */
 
-
-
 void kgd_gfx_v9_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmid,
 		uint32_t sh_mem_config,
 		uint32_t sh_mem_ape1_base, uint32_t sh_mem_ape1_limit,
-		uint32_t sh_mem_bases);
+		uint32_t sh_mem_bases, uint32_t inst);
 int kgd_gfx_v9_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
-		unsigned int vmid);
-int kgd_gfx_v9_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id);
+		unsigned int vmid, uint32_t inst);
+int kgd_gfx_v9_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t inst);
 int kgd_gfx_v9_hqd_load(struct amdgpu_device *adev, void *mqd, uint32_t pipe_id,
 			uint32_t queue_id, uint32_t __user *wptr,
 			uint32_t wptr_shift, uint32_t wptr_mask,
-			struct mm_struct *mm);
+			struct mm_struct *mm, uint32_t inst);
 int kgd_gfx_v9_hiq_mqd_load(struct amdgpu_device *adev, void *mqd,
 			    uint32_t pipe_id, uint32_t queue_id,
-			    uint32_t doorbell_off);
+			    uint32_t doorbell_off, uint32_t inst);
 int kgd_gfx_v9_hqd_dump(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
-			uint32_t (**dump)[2], uint32_t *n_regs);
+			uint32_t (**dump)[2], uint32_t *n_regs, uint32_t inst);
 bool kgd_gfx_v9_hqd_is_occupied(struct amdgpu_device *adev,
 			uint64_t queue_address, uint32_t pipe_id,
-			uint32_t queue_id);
+			uint32_t queue_id, uint32_t inst);
 int kgd_gfx_v9_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
-				uint32_t queue_id);
+				uint32_t queue_id, uint32_t inst);
 int kgd_gfx_v9_wave_control_execute(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
-					uint32_t sq_cmd);
+					uint32_t sq_cmd, uint32_t inst);
 bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev,
 					uint8_t vmid, uint16_t *p_pasid);
-
 void kgd_gfx_v9_set_vm_context_page_table_base(struct amdgpu_device *adev,
 			uint32_t vmid, uint64_t page_table_base);
-void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev, int pasid,
-		int *pasid_wave_cnt, int *max_waves_per_cu);
+void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev,
+				 struct kfd_cu_occupancy *cu_occupancy,
+				 int *max_waves_per_cu, uint32_t inst);
 void kgd_gfx_v9_program_trap_handler_settings(struct amdgpu_device *adev,
-		uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr);
+		uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr,
+		uint32_t inst);
+void kgd_gfx_v9_acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t queue_id, uint32_t inst);
+uint64_t kgd_gfx_v9_get_queue_mask(struct amdgpu_device *adev,
+				uint32_t pipe_id, uint32_t queue_id);
+void kgd_gfx_v9_release_queue(struct amdgpu_device *adev, uint32_t inst);
+void kgd_gfx_v9_set_wave_launch_stall(struct amdgpu_device *adev,
+					uint32_t vmid,
+					bool stall);
+uint32_t kgd_gfx_v9_enable_debug_trap(struct amdgpu_device *adev,
+				      bool restore_dbg_registers,
+				      uint32_t vmid);
+uint32_t kgd_gfx_v9_disable_debug_trap(struct amdgpu_device *adev,
+					bool keep_trap_enabled,
+					uint32_t vmid);
+int kgd_gfx_v9_validate_trap_override_request(struct amdgpu_device *adev,
+					     uint32_t trap_override,
+					     uint32_t *trap_mask_supported);
+uint32_t kgd_gfx_v9_set_wave_launch_mode(struct amdgpu_device *adev,
+					uint8_t wave_launch_mode,
+					uint32_t vmid);
+uint32_t kgd_gfx_v9_set_wave_launch_trap_override(struct amdgpu_device *adev,
+					     uint32_t vmid,
+					     uint32_t trap_override,
+					     uint32_t trap_mask_bits,
+					     uint32_t trap_mask_request,
+					     uint32_t *trap_mask_prev,
+					     uint32_t kfd_dbg_trap_cntl_prev);
+uint32_t kgd_gfx_v9_set_address_watch(struct amdgpu_device *adev,
+					uint64_t watch_address,
+					uint32_t watch_address_mask,
+					uint32_t watch_id,
+					uint32_t watch_mode,
+					uint32_t debug_vmid,
+					uint32_t inst);
+uint32_t kgd_gfx_v9_clear_address_watch(struct amdgpu_device *adev,
+					uint32_t watch_id);
+void kgd_gfx_v9_get_iq_wait_times(struct amdgpu_device *adev,
+				uint32_t *wait_times,
+				uint32_t inst);
+void kgd_gfx_v9_build_dequeue_wait_counts_packet_info(struct amdgpu_device *adev,
+					       uint32_t wait_times,
+					       uint32_t sch_wave,
+					       uint32_t que_sleep,
+					       uint32_t *reg_offset,
+					       uint32_t *reg_data);
+uint64_t kgd_gfx_v9_hqd_get_pq_addr(struct amdgpu_device *adev,
+				    uint32_t pipe_id,
+				    uint32_t queue_id,
+				    uint32_t inst);
+uint64_t kgd_gfx_v9_hqd_reset(struct amdgpu_device *adev,
+			      uint32_t pipe_id,
+			      uint32_t queue_id,
+			      uint32_t inst,
+			      unsigned int utimeout);
+uint32_t kgd_gfx_v9_hqd_sdma_get_doorbell(struct amdgpu_device *adev,
+					  int engine, int queue);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
index 83a83ced2439..b1c24c8fa686 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
@@ -27,6 +27,8 @@
 #include <linux/sched/task.h>
 #include <drm/ttm/ttm_tt.h>
 
+#include <drm/drm_exec.h>
+
 #include "amdgpu_object.h"
 #include "amdgpu_gem.h"
 #include "amdgpu_vm.h"
@@ -35,12 +37,14 @@
 #include "amdgpu_dma_buf.h"
 #include <uapi/linux/kfd_ioctl.h>
 #include "amdgpu_xgmi.h"
+#include "kfd_priv.h"
 #include "kfd_smi_events.h"
 
 /* Userptr restore delay, just long enough to allow consecutive VM
  * changes to accumulate
  */
 #define AMDGPU_USERPTR_RESTORE_DELAY_MS 1
+#define AMDGPU_RESERVE_MEM_LIMIT			(3UL << 29)
 
 /*
  * Align VRAM availability to 2MB to avoid fragmentation caused by 4K allocations in the tail 2MB
@@ -110,13 +114,21 @@ void amdgpu_amdkfd_gpuvm_init_mem_limits(void)
 	struct sysinfo si;
 	uint64_t mem;
 
+	if (kfd_mem_limit.max_system_mem_limit)
+		return;
+
 	si_meminfo(&si);
-	mem = si.freeram - si.freehigh;
+	mem = si.totalram - si.totalhigh;
 	mem *= si.mem_unit;
 
 	spin_lock_init(&kfd_mem_limit.mem_limit_lock);
-	kfd_mem_limit.max_system_mem_limit = mem - (mem >> 4);
-	kfd_mem_limit.max_ttm_mem_limit = (mem >> 1) - (mem >> 3);
+	kfd_mem_limit.max_system_mem_limit = mem - (mem >> 6);
+	if (kfd_mem_limit.max_system_mem_limit < 2 * AMDGPU_RESERVE_MEM_LIMIT)
+		kfd_mem_limit.max_system_mem_limit >>= 1;
+	else
+		kfd_mem_limit.max_system_mem_limit -= AMDGPU_RESERVE_MEM_LIMIT;
+
+	kfd_mem_limit.max_ttm_mem_limit = ttm_tt_pages_limit() << PAGE_SHIFT;
 	pr_debug("Kernel memory limit %lluM, TTM limit %lluM\n",
 		(kfd_mem_limit.max_system_mem_limit >> 20),
 		(kfd_mem_limit.max_ttm_mem_limit >> 20));
@@ -148,16 +160,22 @@ void amdgpu_amdkfd_reserve_system_mem(uint64_t size)
  * @size: Size of buffer, in bytes, encapsulated by B0. This should be
  * equivalent to amdgpu_bo_size(BO)
  * @alloc_flag: Flag used in allocating a BO as noted above
+ * @xcp_id: xcp_id is used to get xcp from xcp manager, one xcp is
+ * managed as one compute node in driver for app
  *
- * Return: returns -ENOMEM in case of error, ZERO otherwise
+ * Return:
+ *	returns -ENOMEM in case of error, ZERO otherwise
  */
 int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
-		uint64_t size, u32 alloc_flag)
+		uint64_t size, u32 alloc_flag, int8_t xcp_id)
 {
 	uint64_t reserved_for_pt =
 		ESTIMATE_PT_SIZE(amdgpu_amdkfd_total_mem_size);
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	uint64_t reserved_for_ras = (con ? con->reserved_pages_in_bytes : 0);
 	size_t system_mem_needed, ttm_mem_needed, vram_needed;
 	int ret = 0;
+	uint64_t vram_size = 0;
 
 	system_mem_needed = 0;
 	ttm_mem_needed = 0;
@@ -172,6 +190,17 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
 		 * 2M BO chunk.
 		 */
 		vram_needed = size;
+		/*
+		 * For GFX 9.4.3, get the VRAM size from XCP structs
+		 */
+		if (WARN_ONCE(xcp_id < 0, "invalid XCP ID %d", xcp_id))
+			return -EINVAL;
+
+		vram_size = KFD_XCP_MEMORY_SIZE(adev, xcp_id);
+		if (adev->apu_prefer_gtt) {
+			system_mem_needed = size;
+			ttm_mem_needed = size;
+		}
 	} else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR) {
 		system_mem_needed = size;
 	} else if (!(alloc_flag &
@@ -184,27 +213,46 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
 	spin_lock(&kfd_mem_limit.mem_limit_lock);
 
 	if (kfd_mem_limit.system_mem_used + system_mem_needed >
-	    kfd_mem_limit.max_system_mem_limit)
+	    kfd_mem_limit.max_system_mem_limit) {
 		pr_debug("Set no_system_mem_limit=1 if using shared memory\n");
+		if (!no_system_mem_limit) {
+			ret = -ENOMEM;
+			goto release;
+		}
+	}
 
-	if ((kfd_mem_limit.system_mem_used + system_mem_needed >
-	     kfd_mem_limit.max_system_mem_limit && !no_system_mem_limit) ||
-	    (kfd_mem_limit.ttm_mem_used + ttm_mem_needed >
-	     kfd_mem_limit.max_ttm_mem_limit) ||
-	    (adev && adev->kfd.vram_used + vram_needed >
-	     adev->gmc.real_vram_size - reserved_for_pt)) {
+	if (kfd_mem_limit.ttm_mem_used + ttm_mem_needed >
+		kfd_mem_limit.max_ttm_mem_limit) {
 		ret = -ENOMEM;
 		goto release;
 	}
 
+	/*if is_app_apu is false and apu_prefer_gtt is true, it is an APU with
+	 * carve out < gtt. In that case, VRAM allocation will go to gtt domain, skip
+	 * VRAM check since ttm_mem_limit check already cover this allocation
+	 */
+
+	if (adev && xcp_id >= 0 && (!adev->apu_prefer_gtt || adev->gmc.is_app_apu)) {
+		uint64_t vram_available =
+			vram_size - reserved_for_pt - reserved_for_ras -
+			atomic64_read(&adev->vram_pin_size);
+		if (adev->kfd.vram_used[xcp_id] + vram_needed > vram_available) {
+			ret = -ENOMEM;
+			goto release;
+		}
+	}
+
 	/* Update memory accounting by decreasing available system
 	 * memory, TTM memory and GPU memory as computed above
 	 */
 	WARN_ONCE(vram_needed && !adev,
 		  "adev reference can't be null when vram is used");
-	if (adev) {
-		adev->kfd.vram_used += vram_needed;
-		adev->kfd.vram_used_aligned += ALIGN(vram_needed, VRAM_AVAILABLITY_ALIGN);
+	if (adev && xcp_id >= 0) {
+		adev->kfd.vram_used[xcp_id] += vram_needed;
+		adev->kfd.vram_used_aligned[xcp_id] +=
+				adev->apu_prefer_gtt ?
+				vram_needed :
+				ALIGN(vram_needed, VRAM_AVAILABLITY_ALIGN);
 	}
 	kfd_mem_limit.system_mem_used += system_mem_needed;
 	kfd_mem_limit.ttm_mem_used += ttm_mem_needed;
@@ -215,7 +263,7 @@ release:
 }
 
 void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
-		uint64_t size, u32 alloc_flag)
+		uint64_t size, u32 alloc_flag, int8_t xcp_id)
 {
 	spin_lock(&kfd_mem_limit.mem_limit_lock);
 
@@ -225,9 +273,19 @@ void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
 	} else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) {
 		WARN_ONCE(!adev,
 			  "adev reference can't be null when alloc mem flags vram is set");
+		if (WARN_ONCE(xcp_id < 0, "invalid XCP ID %d", xcp_id))
+			goto release;
+
 		if (adev) {
-			adev->kfd.vram_used -= size;
-			adev->kfd.vram_used_aligned -= ALIGN(size, VRAM_AVAILABLITY_ALIGN);
+			adev->kfd.vram_used[xcp_id] -= size;
+			if (adev->apu_prefer_gtt) {
+				adev->kfd.vram_used_aligned[xcp_id] -= size;
+				kfd_mem_limit.system_mem_used -= size;
+				kfd_mem_limit.ttm_mem_used -= size;
+			} else {
+				adev->kfd.vram_used_aligned[xcp_id] -=
+					ALIGN(size, VRAM_AVAILABLITY_ALIGN);
+			}
 		}
 	} else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR) {
 		kfd_mem_limit.system_mem_used -= size;
@@ -237,8 +295,8 @@ void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
 		pr_err("%s: Invalid BO type %#x\n", __func__, alloc_flag);
 		goto release;
 	}
-	WARN_ONCE(adev && adev->kfd.vram_used < 0,
-		  "KFD VRAM memory accounting unbalanced");
+	WARN_ONCE(adev && xcp_id >= 0 && adev->kfd.vram_used[xcp_id] < 0,
+		  "KFD VRAM memory accounting unbalanced for xcp: %d", xcp_id);
 	WARN_ONCE(kfd_mem_limit.ttm_mem_used < 0,
 		  "KFD TTM memory accounting unbalanced");
 	WARN_ONCE(kfd_mem_limit.system_mem_used < 0,
@@ -254,14 +312,16 @@ void amdgpu_amdkfd_release_notify(struct amdgpu_bo *bo)
 	u32 alloc_flags = bo->kfd_bo->alloc_flags;
 	u64 size = amdgpu_bo_size(bo);
 
-	amdgpu_amdkfd_unreserve_mem_limit(adev, size, alloc_flags);
+	amdgpu_amdkfd_unreserve_mem_limit(adev, size, alloc_flags,
+					  bo->xcp_id);
 
 	kfree(bo->kfd_bo);
 }
 
 /**
- * @create_dmamap_sg_bo: Creates a amdgpu_bo object to reflect information
+ * create_dmamap_sg_bo() - Creates a amdgpu_bo object to reflect information
  * about USERPTR or DOOREBELL or MMIO BO.
+ *
  * @adev: Device for which dmamap BO is being created
  * @mem: BO of peer device that is being DMA mapped. Provides parameters
  *	 in building the dmamap BO
@@ -285,7 +345,7 @@ create_dmamap_sg_bo(struct amdgpu_device *adev,
 
 	ret = amdgpu_gem_object_create(adev, mem->bo->tbo.base.size, 1,
 			AMDGPU_GEM_DOMAIN_CPU, AMDGPU_GEM_CREATE_PREEMPTIBLE | flags,
-			ttm_bo_type_sg, mem->bo->tbo.base.resv, &gem_obj);
+			ttm_bo_type_sg, mem->bo->tbo.base.resv, &gem_obj, 0);
 
 	amdgpu_bo_unreserve(mem->bo);
 
@@ -326,40 +386,32 @@ static int amdgpu_amdkfd_remove_eviction_fence(struct amdgpu_bo *bo,
 	return 0;
 }
 
-int amdgpu_amdkfd_remove_fence_on_pt_pd_bos(struct amdgpu_bo *bo)
+/**
+ * amdgpu_amdkfd_remove_all_eviction_fences - Remove all eviction fences
+ * @bo: the BO where to remove the evictions fences from.
+ *
+ * This functions should only be used on release when all references to the BO
+ * are already dropped. We remove the eviction fence from the private copy of
+ * the dma_resv object here since that is what is used during release to
+ * determine of the BO is idle or not.
+ */
+void amdgpu_amdkfd_remove_all_eviction_fences(struct amdgpu_bo *bo)
 {
-	struct amdgpu_bo *root = bo;
-	struct amdgpu_vm_bo_base *vm_bo;
-	struct amdgpu_vm *vm;
-	struct amdkfd_process_info *info;
-	struct amdgpu_amdkfd_fence *ef;
-	int ret;
-
-	/* we can always get vm_bo from root PD bo.*/
-	while (root->parent)
-		root = root->parent;
+	struct dma_resv *resv = &bo->tbo.base._resv;
+	struct dma_fence *fence, *stub;
+	struct dma_resv_iter cursor;
 
-	vm_bo = root->vm_bo;
-	if (!vm_bo)
-		return 0;
-
-	vm = vm_bo->vm;
-	if (!vm)
-		return 0;
-
-	info = vm->process_info;
-	if (!info || !info->eviction_fence)
-		return 0;
+	dma_resv_assert_held(resv);
 
-	ef = container_of(dma_fence_get(&info->eviction_fence->base),
-			struct amdgpu_amdkfd_fence, base);
-
-	BUG_ON(!dma_resv_trylock(bo->tbo.base.resv));
-	ret = amdgpu_amdkfd_remove_eviction_fence(bo, ef);
-	dma_resv_unlock(bo->tbo.base.resv);
+	stub = dma_fence_get_stub();
+	dma_resv_for_each_fence(&cursor, resv, DMA_RESV_USAGE_BOOKKEEP, fence) {
+		if (!to_amdgpu_amdkfd_fence(fence))
+			continue;
 
-	dma_fence_put(&ef->base);
-	return ret;
+		dma_resv_replace_fences(resv, fence->context, stub,
+					DMA_RESV_USAGE_BOOKKEEP);
+	}
+	dma_fence_put(stub);
 }
 
 static int amdgpu_amdkfd_bo_validate(struct amdgpu_bo *bo, uint32_t domain,
@@ -372,6 +424,10 @@ static int amdgpu_amdkfd_bo_validate(struct amdgpu_bo *bo, uint32_t domain,
 		 "Called with userptr BO"))
 		return -EINVAL;
 
+	/* bo has been pinned, not need validate it */
+	if (bo->tbo.pin_count)
+		return 0;
+
 	amdgpu_bo_placement_from_domain(bo, domain);
 
 	ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
@@ -384,6 +440,32 @@ validate_fail:
 	return ret;
 }
 
+int amdgpu_amdkfd_bo_validate_and_fence(struct amdgpu_bo *bo,
+					uint32_t domain,
+					struct dma_fence *fence)
+{
+	int ret = amdgpu_bo_reserve(bo, false);
+
+	if (ret)
+		return ret;
+
+	ret = amdgpu_amdkfd_bo_validate(bo, domain, true);
+	if (ret)
+		goto unreserve_out;
+
+	ret = dma_resv_reserve_fences(bo->tbo.base.resv, 1);
+	if (ret)
+		goto unreserve_out;
+
+	dma_resv_add_fence(bo->tbo.base.resv, fence,
+			   DMA_RESV_USAGE_BOOKKEEP);
+
+unreserve_out:
+	amdgpu_bo_unreserve(bo);
+
+	return ret;
+}
+
 static int amdgpu_amdkfd_validate_vm_bo(void *_unused, struct amdgpu_bo *bo)
 {
 	return amdgpu_amdkfd_bo_validate(bo, bo->allowed_domains, false);
@@ -396,13 +478,15 @@ static int amdgpu_amdkfd_validate_vm_bo(void *_unused, struct amdgpu_bo *bo)
  * again. Page directories are only updated after updating page
  * tables.
  */
-static int vm_validate_pt_pd_bos(struct amdgpu_vm *vm)
+static int vm_validate_pt_pd_bos(struct amdgpu_vm *vm,
+				 struct ww_acquire_ctx *ticket)
 {
 	struct amdgpu_bo *pd = vm->root.bo;
 	struct amdgpu_device *adev = amdgpu_ttm_adev(pd->tbo.bdev);
 	int ret;
 
-	ret = amdgpu_vm_validate_pt_bos(adev, vm, amdgpu_amdkfd_validate_vm_bo, NULL);
+	ret = amdgpu_vm_validate(adev, vm, ticket,
+				 amdgpu_amdkfd_validate_vm_bo, NULL);
 	if (ret) {
 		pr_err("failed to validate PT BOs\n");
 		return ret;
@@ -423,10 +507,11 @@ static int vm_update_pds(struct amdgpu_vm *vm, struct amdgpu_sync *sync)
 	if (ret)
 		return ret;
 
-	return amdgpu_sync_fence(sync, vm->last_update);
+	return amdgpu_sync_fence(sync, vm->last_update, GFP_KERNEL);
 }
 
-static uint64_t get_pte_flags(struct amdgpu_device *adev, struct kgd_mem *mem)
+static uint64_t get_pte_flags(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+			      struct kgd_mem *mem)
 {
 	uint32_t mapping_flags = AMDGPU_VM_PAGE_READABLE |
 				 AMDGPU_VM_MTYPE_DEFAULT;
@@ -436,7 +521,7 @@ static uint64_t get_pte_flags(struct amdgpu_device *adev, struct kgd_mem *mem)
 	if (mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_EXECUTABLE)
 		mapping_flags |= AMDGPU_VM_PAGE_EXECUTABLE;
 
-	return amdgpu_gem_va_map_flags(adev, mapping_flags);
+	return mapping_flags;
 }
 
 /**
@@ -648,7 +733,7 @@ kfd_mem_dmaunmap_userptr(struct kgd_mem *mem,
 		return;
 
 	amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU);
-	ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+	(void)ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 
 	dma_unmap_sgtable(adev->dev, ttm->sg, direction, 0);
 	sg_free_table(ttm->sg);
@@ -659,11 +744,10 @@ kfd_mem_dmaunmap_userptr(struct kgd_mem *mem,
 static void
 kfd_mem_dmaunmap_dmabuf(struct kfd_mem_attachment *attachment)
 {
-	struct ttm_operation_ctx ctx = {.interruptible = true};
-	struct amdgpu_bo *bo = attachment->bo_va->base.bo;
-
-	amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU);
-	ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+	/* This is a no-op. We don't want to trigger eviction fences when
+	 * unmapping DMABufs. Therefore the invalidation (moving to system
+	 * domain) is done in kfd_mem_dmamap_dmabuf.
+	 */
 }
 
 /**
@@ -693,12 +777,12 @@ kfd_mem_dmaunmap_sg_bo(struct kgd_mem *mem,
 	enum dma_data_direction dir;
 
 	if (unlikely(!ttm->sg)) {
-		pr_err("SG Table of BO is UNEXPECTEDLY NULL");
+		pr_debug("SG Table of BO is NULL");
 		return;
 	}
 
 	amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU);
-	ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+	(void)ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 
 	dir = mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_WRITABLE ?
 				DMA_BIDIRECTIONAL : DMA_TO_DEVICE;
@@ -734,13 +818,17 @@ kfd_mem_dmaunmap_attachment(struct kgd_mem *mem,
 static int kfd_mem_export_dmabuf(struct kgd_mem *mem)
 {
 	if (!mem->dmabuf) {
-		struct dma_buf *ret = amdgpu_gem_prime_export(
-			&mem->bo->tbo.base,
+		struct amdgpu_device *bo_adev;
+		struct dma_buf *dmabuf;
+
+		bo_adev = amdgpu_ttm_adev(mem->bo->tbo.bdev);
+		dmabuf = drm_gem_prime_handle_to_dmabuf(&bo_adev->ddev, bo_adev->kfd.client.file,
+					       mem->gem_handle,
 			mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_WRITABLE ?
-				DRM_RDWR : 0);
-		if (IS_ERR(ret))
-			return PTR_ERR(ret);
-		mem->dmabuf = ret;
+					       DRM_RDWR : 0);
+		if (IS_ERR(dmabuf))
+			return PTR_ERR(dmabuf);
+		mem->dmabuf = dmabuf;
 	}
 
 	return 0;
@@ -788,6 +876,7 @@ static int kfd_mem_attach(struct amdgpu_device *adev, struct kgd_mem *mem,
 	uint64_t va = mem->va;
 	struct kfd_mem_attachment *attachment[2] = {NULL, NULL};
 	struct amdgpu_bo *bo[2] = {NULL, NULL};
+	struct amdgpu_bo_va *bo_va;
 	bool same_hive = false;
 	int i, ret;
 
@@ -804,7 +893,7 @@ static int kfd_mem_attach(struct amdgpu_device *adev, struct kgd_mem *mem,
 	 * if peer device has large BAR. In contrast, access over xGMI is
 	 * allowed for both small and large BAR configurations of peer device
 	 */
-	if ((adev != bo_adev) &&
+	if ((adev != bo_adev && !adev->apu_prefer_gtt) &&
 	    ((mem->domain == AMDGPU_GEM_DOMAIN_VRAM) ||
 	     (mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL) ||
 	     (mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP))) {
@@ -826,9 +915,10 @@ static int kfd_mem_attach(struct amdgpu_device *adev, struct kgd_mem *mem,
 
 		if ((adev == bo_adev && !(mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP)) ||
 		    (amdgpu_ttm_tt_get_usermm(mem->bo->tbo.ttm) && reuse_dmamap(adev, bo_adev)) ||
-			same_hive) {
+		    (mem->domain == AMDGPU_GEM_DOMAIN_GTT && reuse_dmamap(adev, bo_adev)) ||
+		    same_hive) {
 			/* Mappings on the local GPU, or VRAM mappings in the
-			 * local hive, or userptr mapping can reuse dma map
+			 * local hive, or userptr, or GTT mapping can reuse dma map
 			 * address space share the original BO
 			 */
 			attachment[i]->type = KFD_MEM_ATT_SHARED;
@@ -874,7 +964,12 @@ static int kfd_mem_attach(struct amdgpu_device *adev, struct kgd_mem *mem,
 			pr_debug("Unable to reserve BO during memory attach");
 			goto unwind;
 		}
-		attachment[i]->bo_va = amdgpu_vm_bo_add(adev, vm, bo[i]);
+		bo_va = amdgpu_vm_bo_find(vm, bo[i]);
+		if (!bo_va)
+			bo_va = amdgpu_vm_bo_add(adev, vm, bo[i]);
+		else
+			++bo_va->ref_count;
+		attachment[i]->bo_va = bo_va;
 		amdgpu_bo_unreserve(bo[i]);
 		if (unlikely(!attachment[i]->bo_va)) {
 			ret = -ENOMEM;
@@ -883,7 +978,7 @@ static int kfd_mem_attach(struct amdgpu_device *adev, struct kgd_mem *mem,
 			goto unwind;
 		}
 		attachment[i]->va = va;
-		attachment[i]->pte_flags = get_pte_flags(adev, mem);
+		attachment[i]->pte_flags = get_pte_flags(adev, vm, mem);
 		attachment[i]->adev = adev;
 		list_add(&attachment[i]->list, &mem->attachments);
 
@@ -897,8 +992,9 @@ unwind:
 		if (!attachment[i])
 			continue;
 		if (attachment[i]->bo_va) {
-			amdgpu_bo_reserve(bo[i], true);
-			amdgpu_vm_bo_del(adev, attachment[i]->bo_va);
+			(void)amdgpu_bo_reserve(bo[i], true);
+			if (--attachment[i]->bo_va->ref_count == 0)
+				amdgpu_vm_bo_del(adev, attachment[i]->bo_va);
 			amdgpu_bo_unreserve(bo[i]);
 			list_del(&attachment[i]->list);
 		}
@@ -915,7 +1011,8 @@ static void kfd_mem_detach(struct kfd_mem_attachment *attachment)
 
 	pr_debug("\t remove VA 0x%llx in entry %p\n",
 			attachment->va, attachment);
-	amdgpu_vm_bo_del(attachment->adev, attachment->bo_va);
+	if (--attachment->bo_va->ref_count == 0)
+		amdgpu_vm_bo_del(attachment->adev, attachment->bo_va);
 	drm_gem_object_put(&bo->tbo.base);
 	list_del(&attachment->list);
 	kfree(attachment);
@@ -925,28 +1022,20 @@ static void add_kgd_mem_to_kfd_bo_list(struct kgd_mem *mem,
 				struct amdkfd_process_info *process_info,
 				bool userptr)
 {
-	struct ttm_validate_buffer *entry = &mem->validate_list;
-	struct amdgpu_bo *bo = mem->bo;
-
-	INIT_LIST_HEAD(&entry->head);
-	entry->num_shared = 1;
-	entry->bo = &bo->tbo;
 	mutex_lock(&process_info->lock);
 	if (userptr)
-		list_add_tail(&entry->head, &process_info->userptr_valid_list);
+		list_add_tail(&mem->validate_list,
+			      &process_info->userptr_valid_list);
 	else
-		list_add_tail(&entry->head, &process_info->kfd_bo_list);
+		list_add_tail(&mem->validate_list, &process_info->kfd_bo_list);
 	mutex_unlock(&process_info->lock);
 }
 
 static void remove_kgd_mem_from_kfd_bo_list(struct kgd_mem *mem,
 		struct amdkfd_process_info *process_info)
 {
-	struct ttm_validate_buffer *bo_list_entry;
-
-	bo_list_entry = &mem->validate_list;
 	mutex_lock(&process_info->lock);
-	list_del(&bo_list_entry->head);
+	list_del(&mem->validate_list);
 	mutex_unlock(&process_info->lock);
 }
 
@@ -968,7 +1057,7 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr,
 	struct amdkfd_process_info *process_info = mem->process_info;
 	struct amdgpu_bo *bo = mem->bo;
 	struct ttm_operation_ctx ctx = { true, false };
-	struct hmm_range *range;
+	struct amdgpu_hmm_range *range;
 	int ret = 0;
 
 	mutex_lock(&process_info->lock);
@@ -1000,9 +1089,19 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr,
 		return 0;
 	}
 
-	ret = amdgpu_ttm_tt_get_user_pages(bo, bo->tbo.ttm->pages, &range);
+	range = amdgpu_hmm_range_alloc(NULL);
+	if (unlikely(!range)) {
+		ret = -ENOMEM;
+		goto unregister_out;
+	}
+
+	ret = amdgpu_ttm_tt_get_user_pages(bo, range);
 	if (ret) {
-		pr_err("%s: Failed to get user pages: %d\n", __func__, ret);
+		amdgpu_hmm_range_free(range);
+		if (ret == -EAGAIN)
+			pr_debug("Failed to get user pages, try again\n");
+		else
+			pr_err("%s: Failed to get user pages: %d\n", __func__, ret);
 		goto unregister_out;
 	}
 
@@ -1011,6 +1110,9 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr,
 		pr_err("%s: Failed to reserve BO\n", __func__);
 		goto release_out;
 	}
+
+	amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, range);
+
 	amdgpu_bo_placement_from_domain(bo, mem->domain);
 	ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 	if (ret)
@@ -1018,7 +1120,7 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr,
 	amdgpu_bo_unreserve(bo);
 
 release_out:
-	amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm, range);
+	amdgpu_hmm_range_free(range);
 unregister_out:
 	if (ret)
 		amdgpu_hmm_unregister(bo);
@@ -1033,13 +1135,12 @@ out:
  * object can track VM updates.
  */
 struct bo_vm_reservation_context {
-	struct amdgpu_bo_list_entry kfd_bo; /* BO list entry for the KFD BO */
-	unsigned int n_vms;		    /* Number of VMs reserved	    */
-	struct amdgpu_bo_list_entry *vm_pd; /* Array of VM BO list entries  */
-	struct ww_acquire_ctx ticket;	    /* Reservation ticket	    */
-	struct list_head list, duplicates;  /* BO lists			    */
-	struct amdgpu_sync *sync;	    /* Pointer to sync object	    */
-	bool reserved;			    /* Whether BOs are reserved	    */
+	/* DRM execution context for the reservation */
+	struct drm_exec exec;
+	/* Number of VMs reserved */
+	unsigned int n_vms;
+	/* Pointer to sync object */
+	struct amdgpu_sync *sync;
 };
 
 enum bo_vm_match {
@@ -1063,35 +1164,26 @@ static int reserve_bo_and_vm(struct kgd_mem *mem,
 
 	WARN_ON(!vm);
 
-	ctx->reserved = false;
 	ctx->n_vms = 1;
 	ctx->sync = &mem->sync;
-
-	INIT_LIST_HEAD(&ctx->list);
-	INIT_LIST_HEAD(&ctx->duplicates);
-
-	ctx->vm_pd = kcalloc(ctx->n_vms, sizeof(*ctx->vm_pd), GFP_KERNEL);
-	if (!ctx->vm_pd)
-		return -ENOMEM;
-
-	ctx->kfd_bo.priority = 0;
-	ctx->kfd_bo.tv.bo = &bo->tbo;
-	ctx->kfd_bo.tv.num_shared = 1;
-	list_add(&ctx->kfd_bo.tv.head, &ctx->list);
-
-	amdgpu_vm_get_pd_bo(vm, &ctx->list, &ctx->vm_pd[0]);
-
-	ret = ttm_eu_reserve_buffers(&ctx->ticket, &ctx->list,
-				     false, &ctx->duplicates);
-	if (ret) {
-		pr_err("Failed to reserve buffers in ttm.\n");
-		kfree(ctx->vm_pd);
-		ctx->vm_pd = NULL;
-		return ret;
+	drm_exec_init(&ctx->exec, DRM_EXEC_INTERRUPTIBLE_WAIT, 0);
+	drm_exec_until_all_locked(&ctx->exec) {
+		ret = amdgpu_vm_lock_pd(vm, &ctx->exec, 2);
+		drm_exec_retry_on_contention(&ctx->exec);
+		if (unlikely(ret))
+			goto error;
+
+		ret = drm_exec_prepare_obj(&ctx->exec, &bo->tbo.base, 1);
+		drm_exec_retry_on_contention(&ctx->exec);
+		if (unlikely(ret))
+			goto error;
 	}
-
-	ctx->reserved = true;
 	return 0;
+
+error:
+	pr_err("Failed to reserve buffers in ttm.\n");
+	drm_exec_fini(&ctx->exec);
+	return ret;
 }
 
 /**
@@ -1108,63 +1200,40 @@ static int reserve_bo_and_cond_vms(struct kgd_mem *mem,
 				struct amdgpu_vm *vm, enum bo_vm_match map_type,
 				struct bo_vm_reservation_context *ctx)
 {
-	struct amdgpu_bo *bo = mem->bo;
 	struct kfd_mem_attachment *entry;
-	unsigned int i;
+	struct amdgpu_bo *bo = mem->bo;
 	int ret;
 
-	ctx->reserved = false;
-	ctx->n_vms = 0;
-	ctx->vm_pd = NULL;
 	ctx->sync = &mem->sync;
+	drm_exec_init(&ctx->exec, DRM_EXEC_INTERRUPTIBLE_WAIT |
+		      DRM_EXEC_IGNORE_DUPLICATES, 0);
+	drm_exec_until_all_locked(&ctx->exec) {
+		ctx->n_vms = 0;
+		list_for_each_entry(entry, &mem->attachments, list) {
+			if ((vm && vm != entry->bo_va->base.vm) ||
+				(entry->is_mapped != map_type
+				&& map_type != BO_VM_ALL))
+				continue;
 
-	INIT_LIST_HEAD(&ctx->list);
-	INIT_LIST_HEAD(&ctx->duplicates);
-
-	list_for_each_entry(entry, &mem->attachments, list) {
-		if ((vm && vm != entry->bo_va->base.vm) ||
-			(entry->is_mapped != map_type
-			&& map_type != BO_VM_ALL))
-			continue;
-
-		ctx->n_vms++;
-	}
-
-	if (ctx->n_vms != 0) {
-		ctx->vm_pd = kcalloc(ctx->n_vms, sizeof(*ctx->vm_pd),
-				     GFP_KERNEL);
-		if (!ctx->vm_pd)
-			return -ENOMEM;
-	}
-
-	ctx->kfd_bo.priority = 0;
-	ctx->kfd_bo.tv.bo = &bo->tbo;
-	ctx->kfd_bo.tv.num_shared = 1;
-	list_add(&ctx->kfd_bo.tv.head, &ctx->list);
-
-	i = 0;
-	list_for_each_entry(entry, &mem->attachments, list) {
-		if ((vm && vm != entry->bo_va->base.vm) ||
-			(entry->is_mapped != map_type
-			&& map_type != BO_VM_ALL))
-			continue;
-
-		amdgpu_vm_get_pd_bo(entry->bo_va->base.vm, &ctx->list,
-				&ctx->vm_pd[i]);
-		i++;
-	}
+			ret = amdgpu_vm_lock_pd(entry->bo_va->base.vm,
+						&ctx->exec, 2);
+			drm_exec_retry_on_contention(&ctx->exec);
+			if (unlikely(ret))
+				goto error;
+			++ctx->n_vms;
+		}
 
-	ret = ttm_eu_reserve_buffers(&ctx->ticket, &ctx->list,
-				     false, &ctx->duplicates);
-	if (ret) {
-		pr_err("Failed to reserve buffers in ttm.\n");
-		kfree(ctx->vm_pd);
-		ctx->vm_pd = NULL;
-		return ret;
+		ret = drm_exec_prepare_obj(&ctx->exec, &bo->tbo.base, 1);
+		drm_exec_retry_on_contention(&ctx->exec);
+		if (unlikely(ret))
+			goto error;
 	}
-
-	ctx->reserved = true;
 	return 0;
+
+error:
+	pr_err("Failed to reserve buffers in ttm.\n");
+	drm_exec_fini(&ctx->exec);
+	return ret;
 }
 
 /**
@@ -1185,19 +1254,12 @@ static int unreserve_bo_and_vms(struct bo_vm_reservation_context *ctx,
 	if (wait)
 		ret = amdgpu_sync_wait(ctx->sync, intr);
 
-	if (ctx->reserved)
-		ttm_eu_backoff_reservation(&ctx->ticket, &ctx->list);
-	kfree(ctx->vm_pd);
-
+	drm_exec_fini(&ctx->exec);
 	ctx->sync = NULL;
-
-	ctx->reserved = false;
-	ctx->vm_pd = NULL;
-
 	return ret;
 }
 
-static void unmap_bo_from_gpuvm(struct kgd_mem *mem,
+static int unmap_bo_from_gpuvm(struct kgd_mem *mem,
 				struct kfd_mem_attachment *entry,
 				struct amdgpu_sync *sync)
 {
@@ -1205,13 +1267,22 @@ static void unmap_bo_from_gpuvm(struct kgd_mem *mem,
 	struct amdgpu_device *adev = entry->adev;
 	struct amdgpu_vm *vm = bo_va->base.vm;
 
-	amdgpu_vm_bo_unmap(adev, bo_va, entry->va);
+	if (bo_va->queue_refcount) {
+		pr_debug("bo_va->queue_refcount %d\n", bo_va->queue_refcount);
+		return -EBUSY;
+	}
 
-	amdgpu_vm_clear_freed(adev, vm, &bo_va->last_pt_update);
+	(void)amdgpu_vm_bo_unmap(adev, bo_va, entry->va);
 
-	amdgpu_sync_fence(sync, bo_va->last_pt_update);
+	/* VM entity stopped if process killed, don't clear freed pt bo */
+	if (!amdgpu_vm_ready(vm))
+		return 0;
 
-	kfd_mem_dmaunmap_attachment(mem, entry);
+	(void)amdgpu_vm_clear_freed(adev, vm, &bo_va->last_pt_update);
+
+	(void)amdgpu_sync_fence(sync, bo_va->last_pt_update, GFP_KERNEL);
+
+	return 0;
 }
 
 static int update_gpuvm_pte(struct kgd_mem *mem,
@@ -1233,7 +1304,7 @@ static int update_gpuvm_pte(struct kgd_mem *mem,
 		return ret;
 	}
 
-	return amdgpu_sync_fence(sync, bo_va->last_pt_update);
+	return amdgpu_sync_fence(sync, bo_va->last_pt_update, GFP_KERNEL);
 }
 
 static int map_bo_to_gpuvm(struct kgd_mem *mem,
@@ -1266,17 +1337,19 @@ static int map_bo_to_gpuvm(struct kgd_mem *mem,
 
 update_gpuvm_pte_failed:
 	unmap_bo_from_gpuvm(mem, entry, sync);
+	kfd_mem_dmaunmap_attachment(mem, entry);
 	return ret;
 }
 
-static int process_validate_vms(struct amdkfd_process_info *process_info)
+static int process_validate_vms(struct amdkfd_process_info *process_info,
+				struct ww_acquire_ctx *ticket)
 {
 	struct amdgpu_vm *peer_vm;
 	int ret;
 
 	list_for_each_entry(peer_vm, &process_info->vm_list_head,
 			    vm_list_node) {
-		ret = vm_validate_pt_pd_bos(peer_vm);
+		ret = vm_validate_pt_pd_bos(peer_vm, ticket);
 		if (ret)
 			return ret;
 	}
@@ -1353,7 +1426,6 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info,
 				  amdgpu_amdkfd_restore_userptr_worker);
 
 		*process_info = info;
-		*ef = dma_fence_get(&info->eviction_fence->base);
 	}
 
 	vm->process_info = *process_info;
@@ -1362,7 +1434,7 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info,
 	ret = amdgpu_bo_reserve(vm->root.bo, true);
 	if (ret)
 		goto reserve_pd_fail;
-	ret = vm_validate_pt_pd_bos(vm);
+	ret = vm_validate_pt_pd_bos(vm, NULL);
 	if (ret) {
 		pr_err("validate_pt_pd_bos() failed\n");
 		goto validate_pd_fail;
@@ -1384,6 +1456,8 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info,
 	list_add_tail(&vm->vm_list_node,
 			&(vm->process_info->vm_list_head));
 	vm->process_info->n_vms++;
+	if (ef)
+		*ef = dma_fence_get(&vm->process_info->eviction_fence->base);
 	mutex_unlock(&vm->process_info->lock);
 
 	return 0;
@@ -1395,10 +1469,7 @@ validate_pd_fail:
 reserve_pd_fail:
 	vm->process_info = NULL;
 	if (info) {
-		/* Two fence references: one in info and one in *ef */
 		dma_fence_put(&info->eviction_fence->base);
-		dma_fence_put(*ef);
-		*ef = NULL;
 		*process_info = NULL;
 		put_pid(info->pid);
 create_evict_fence_fail:
@@ -1428,13 +1499,30 @@ static int amdgpu_amdkfd_gpuvm_pin_bo(struct amdgpu_bo *bo, u32 domain)
 	if (unlikely(ret))
 		return ret;
 
-	ret = amdgpu_bo_pin_restricted(bo, domain, 0, 0);
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS) {
+		/*
+		 * If bo is not contiguous on VRAM, move to system memory first to ensure
+		 * we can get contiguous VRAM space after evicting other BOs.
+		 */
+		if (!(bo->tbo.resource->placement & TTM_PL_FLAG_CONTIGUOUS)) {
+			struct ttm_operation_ctx ctx = { true, false };
+
+			amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT);
+			ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+			if (unlikely(ret)) {
+				pr_debug("validate bo 0x%p to GTT failed %d\n", &bo->tbo, ret);
+				goto out;
+			}
+		}
+	}
+
+	ret = amdgpu_bo_pin(bo, domain);
 	if (ret)
 		pr_err("Error in Pinning BO to domain: %d\n", domain);
 
 	amdgpu_bo_sync_wait(bo, AMDGPU_FENCE_OWNER_KFD, false);
+out:
 	amdgpu_bo_unreserve(bo);
-
 	return ret;
 }
 
@@ -1458,27 +1546,6 @@ static void amdgpu_amdkfd_gpuvm_unpin_bo(struct amdgpu_bo *bo)
 	amdgpu_bo_unreserve(bo);
 }
 
-int amdgpu_amdkfd_gpuvm_set_vm_pasid(struct amdgpu_device *adev,
-				     struct amdgpu_vm *avm, u32 pasid)
-
-{
-	int ret;
-
-	/* Free the original amdgpu allocated pasid,
-	 * will be replaced with kfd allocated pasid.
-	 */
-	if (avm->pasid) {
-		amdgpu_pasid_free(avm->pasid);
-		amdgpu_vm_set_pasid(adev, avm, 0);
-	}
-
-	ret = amdgpu_vm_set_pasid(adev, avm, pasid);
-	if (ret)
-		return ret;
-
-	return 0;
-}
-
 int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev,
 					   struct amdgpu_vm *avm,
 					   void **process_info,
@@ -1536,27 +1603,6 @@ void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev,
 	}
 }
 
-void amdgpu_amdkfd_gpuvm_release_process_vm(struct amdgpu_device *adev,
-					    void *drm_priv)
-{
-	struct amdgpu_vm *avm;
-
-	if (WARN_ON(!adev || !drm_priv))
-		return;
-
-	avm = drm_priv_to_vm(drm_priv);
-
-	pr_debug("Releasing process vm %p\n", avm);
-
-	/* The original pasid of amdgpu vm has already been
-	 * released during making a amdgpu vm to a compute vm
-	 * The current pasid is managed by kfd and will be
-	 * released on kfd process destroy. Set amdgpu pasid
-	 * to 0 to avoid duplicate release.
-	 */
-	amdgpu_vm_release_compute(adev, avm);
-}
-
 uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *drm_priv)
 {
 	struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv);
@@ -1592,30 +1638,57 @@ int amdgpu_amdkfd_criu_resume(void *p)
 		goto out_unlock;
 	}
 	WRITE_ONCE(pinfo->block_mmu_notifications, false);
-	schedule_delayed_work(&pinfo->restore_userptr_work, 0);
+	queue_delayed_work(system_freezable_wq,
+			   &pinfo->restore_userptr_work, 0);
 
 out_unlock:
 	mutex_unlock(&pinfo->lock);
 	return ret;
 }
 
-size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev)
+size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev,
+					  uint8_t xcp_id)
 {
 	uint64_t reserved_for_pt =
 		ESTIMATE_PT_SIZE(amdgpu_amdkfd_total_mem_size);
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	uint64_t reserved_for_ras = (con ? con->reserved_pages_in_bytes : 0);
 	ssize_t available;
+	uint64_t vram_available, system_mem_available, ttm_mem_available;
 
 	spin_lock(&kfd_mem_limit.mem_limit_lock);
-	available = adev->gmc.real_vram_size
-		- adev->kfd.vram_used_aligned
-		- atomic64_read(&adev->vram_pin_size)
-		- reserved_for_pt;
+	if (adev->apu_prefer_gtt && !adev->gmc.is_app_apu)
+		vram_available = KFD_XCP_MEMORY_SIZE(adev, xcp_id)
+			- adev->kfd.vram_used_aligned[xcp_id];
+	else
+		vram_available = KFD_XCP_MEMORY_SIZE(adev, xcp_id)
+			- adev->kfd.vram_used_aligned[xcp_id]
+			- atomic64_read(&adev->vram_pin_size)
+			- reserved_for_pt
+			- reserved_for_ras;
+
+	if (adev->apu_prefer_gtt) {
+		system_mem_available = no_system_mem_limit ?
+					kfd_mem_limit.max_system_mem_limit :
+					kfd_mem_limit.max_system_mem_limit -
+					kfd_mem_limit.system_mem_used;
+
+		ttm_mem_available = kfd_mem_limit.max_ttm_mem_limit -
+				kfd_mem_limit.ttm_mem_used;
+
+		available = min3(system_mem_available, ttm_mem_available,
+				 vram_available);
+		available = ALIGN_DOWN(available, PAGE_SIZE);
+	} else {
+		available = ALIGN_DOWN(vram_available, VRAM_AVAILABLITY_ALIGN);
+	}
+
 	spin_unlock(&kfd_mem_limit.mem_limit_lock);
 
 	if (available < 0)
 		available = 0;
 
-	return ALIGN_DOWN(available, VRAM_AVAILABLITY_ALIGN);
+	return available;
 }
 
 int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
@@ -1624,6 +1697,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 		uint64_t *offset, uint32_t flags, bool criu_resume)
 {
 	struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv);
+	struct amdgpu_fpriv *fpriv = container_of(avm, struct amdgpu_fpriv, vm);
 	enum ttm_bo_type bo_type = ttm_bo_type_device;
 	struct sg_table *sg = NULL;
 	uint64_t user_addr = 0;
@@ -1631,6 +1705,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 	struct drm_gem_object *gobj = NULL;
 	u32 domain, alloc_domain;
 	uint64_t aligned_size;
+	int8_t xcp_id = -1;
 	u64 alloc_flags;
 	int ret;
 
@@ -1639,9 +1714,22 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 	 */
 	if (flags & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) {
 		domain = alloc_domain = AMDGPU_GEM_DOMAIN_VRAM;
-		alloc_flags = AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE;
-		alloc_flags |= (flags & KFD_IOC_ALLOC_MEM_FLAGS_PUBLIC) ?
+
+		if (adev->apu_prefer_gtt) {
+			domain = AMDGPU_GEM_DOMAIN_GTT;
+			alloc_domain = AMDGPU_GEM_DOMAIN_GTT;
+			alloc_flags = 0;
+		} else {
+			alloc_flags = AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE;
+			alloc_flags |= (flags & KFD_IOC_ALLOC_MEM_FLAGS_PUBLIC) ?
 			AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED : 0;
+
+			/* For contiguous VRAM allocation */
+			if (flags & KFD_IOC_ALLOC_MEM_FLAGS_CONTIGUOUS)
+				alloc_flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
+		}
+		xcp_id = fpriv->xcp_id == AMDGPU_XCP_NO_PARTITION ?
+					0 : fpriv->xcp_id;
 	} else if (flags & KFD_IOC_ALLOC_MEM_FLAGS_GTT) {
 		domain = alloc_domain = AMDGPU_GEM_DOMAIN_GTT;
 		alloc_flags = 0;
@@ -1669,6 +1757,8 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 
 	if (flags & KFD_IOC_ALLOC_MEM_FLAGS_COHERENT)
 		alloc_flags |= AMDGPU_GEM_CREATE_COHERENT;
+	if (flags & KFD_IOC_ALLOC_MEM_FLAGS_EXT_COHERENT)
+		alloc_flags |= AMDGPU_GEM_CREATE_EXT_COHERENT;
 	if (flags & KFD_IOC_ALLOC_MEM_FLAGS_UNCACHED)
 		alloc_flags |= AMDGPU_GEM_CREATE_UNCACHED;
 
@@ -1693,17 +1783,19 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 
 	amdgpu_sync_create(&(*mem)->sync);
 
-	ret = amdgpu_amdkfd_reserve_mem_limit(adev, aligned_size, flags);
+	ret = amdgpu_amdkfd_reserve_mem_limit(adev, aligned_size, flags,
+					      xcp_id);
 	if (ret) {
 		pr_debug("Insufficient memory\n");
 		goto err_reserve_limit;
 	}
 
-	pr_debug("\tcreate BO VA 0x%llx size 0x%llx domain %s\n",
-			va, (*mem)->aql_queue ? size << 1 : size, domain_string(alloc_domain));
+	pr_debug("\tcreate BO VA 0x%llx size 0x%llx domain %s xcp_id %d\n",
+		 va, (*mem)->aql_queue ? size << 1 : size,
+		 domain_string(alloc_domain), xcp_id);
 
 	ret = amdgpu_gem_object_create(adev, aligned_size, 1, alloc_domain, alloc_flags,
-				       bo_type, NULL, &gobj);
+				       bo_type, NULL, &gobj, xcp_id + 1);
 	if (ret) {
 		pr_debug("Failed to create BO on domain %s. ret %d\n",
 			 domain_string(alloc_domain), ret);
@@ -1714,6 +1806,9 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 		pr_debug("Failed to allow vma node access. ret %d\n", ret);
 		goto err_node_allow;
 	}
+	ret = drm_gem_handle_create(adev->kfd.client.file, gobj, &(*mem)->gem_handle);
+	if (ret)
+		goto err_gem_handle_create;
 	bo = gem_to_amdgpu_bo(gobj);
 	if (bo_type == ttm_bo_type_sg) {
 		bo->tbo.sg = sg;
@@ -1728,6 +1823,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 	(*mem)->domain = domain;
 	(*mem)->mapped_to_gpu_memory = 0;
 	(*mem)->process_info = avm->process_info;
+
 	add_kgd_mem_to_kfd_bo_list(*mem, avm->process_info, user_addr);
 
 	if (user_addr) {
@@ -1744,6 +1840,15 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 		}
 		bo->allowed_domains = AMDGPU_GEM_DOMAIN_GTT;
 		bo->preferred_domains = AMDGPU_GEM_DOMAIN_GTT;
+	} else {
+		mutex_lock(&avm->process_info->lock);
+		if (avm->process_info->eviction_fence &&
+		    !dma_fence_is_signaled(&avm->process_info->eviction_fence->base))
+			ret = amdgpu_amdkfd_bo_validate_and_fence(bo, domain,
+				&avm->process_info->eviction_fence->base);
+		mutex_unlock(&avm->process_info->lock);
+		if (ret)
+			goto err_validate_bo;
 	}
 
 	if (offset)
@@ -1753,14 +1858,18 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 
 allocate_init_user_pages_failed:
 err_pin_bo:
+err_validate_bo:
 	remove_kgd_mem_from_kfd_bo_list(*mem, avm->process_info);
+	drm_gem_handle_delete(adev->kfd.client.file, (*mem)->gem_handle);
+err_gem_handle_create:
 	drm_vma_node_revoke(&gobj->vma_node, drm_priv);
 err_node_allow:
 	/* Don't unreserve system mem limit twice */
 	goto err_reserve_limit;
 err_bo_create:
-	amdgpu_amdkfd_unreserve_mem_limit(adev, aligned_size, flags);
+	amdgpu_amdkfd_unreserve_mem_limit(adev, aligned_size, flags, xcp_id);
 err_reserve_limit:
+	amdgpu_sync_free(&(*mem)->sync);
 	mutex_destroy(&(*mem)->lock);
 	if (gobj)
 		drm_gem_object_put(gobj);
@@ -1783,7 +1892,6 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
 	bool use_release_notifier = (mem->bo->kfd_bo == mem);
 	struct kfd_mem_attachment *entry, *tmp;
 	struct bo_vm_reservation_context ctx;
-	struct ttm_validate_buffer *bo_list_entry;
 	unsigned int mapped_to_gpu_memory;
 	int ret;
 	bool is_imported = false;
@@ -1811,16 +1919,15 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
 	}
 
 	/* Make sure restore workers don't access the BO any more */
-	bo_list_entry = &mem->validate_list;
 	mutex_lock(&process_info->lock);
-	list_del(&bo_list_entry->head);
+	list_del(&mem->validate_list);
 	mutex_unlock(&process_info->lock);
 
 	/* Cleanup user pages and MMU notifiers */
 	if (amdgpu_ttm_tt_get_usermm(mem->bo->tbo.ttm)) {
 		amdgpu_hmm_unregister(mem->bo);
 		mutex_lock(&process_info->notifier_lock);
-		amdgpu_ttm_tt_discard_user_pages(mem->bo->tbo.ttm, mem->range);
+		amdgpu_hmm_range_free(mem->range);
 		mutex_unlock(&process_info->notifier_lock);
 	}
 
@@ -1828,18 +1935,16 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
 	if (unlikely(ret))
 		return ret;
 
-	/* The eviction fence should be removed by the last unmap.
-	 * TODO: Log an error condition if the bo still has the eviction fence
-	 * attached
-	 */
 	amdgpu_amdkfd_remove_eviction_fence(mem->bo,
 					process_info->eviction_fence);
 	pr_debug("Release VA 0x%llx - 0x%llx\n", mem->va,
 		mem->va + bo_size * (1 + mem->aql_queue));
 
 	/* Remove from VM internal data structures */
-	list_for_each_entry_safe(entry, tmp, &mem->attachments, list)
+	list_for_each_entry_safe(entry, tmp, &mem->attachments, list) {
+		kfd_mem_dmaunmap_attachment(mem, entry);
 		kfd_mem_detach(entry);
+	}
 
 	ret = unreserve_bo_and_vms(&ctx, false, false);
 
@@ -1855,11 +1960,12 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
 	}
 
 	/* Update the size of the BO being freed if it was allocated from
-	 * VRAM and is not imported.
+	 * VRAM and is not imported. For APP APU VRAM allocations are done
+	 * in GTT domain
 	 */
 	if (size) {
-		if ((mem->bo->preferred_domains == AMDGPU_GEM_DOMAIN_VRAM) &&
-		    (!is_imported))
+		if (!is_imported &&
+		   mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_VRAM)
 			*size = bo_size;
 		else
 			*size = 0;
@@ -1867,8 +1973,11 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
 
 	/* Free the BO*/
 	drm_vma_node_revoke(&mem->bo->tbo.base.vma_node, drm_priv);
-	if (mem->dmabuf)
+	drm_gem_handle_delete(adev->kfd.client.file, mem->gem_handle);
+	if (mem->dmabuf) {
 		dma_buf_put(mem->dmabuf);
+		mem->dmabuf = NULL;
+	}
 	mutex_destroy(&mem->lock);
 
 	/* If this releases the last reference, it will end up calling
@@ -1951,23 +2060,10 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu(
 	    bo->tbo.resource->mem_type == TTM_PL_SYSTEM)
 		is_invalid_userptr = true;
 
-	ret = vm_validate_pt_pd_bos(avm);
+	ret = vm_validate_pt_pd_bos(avm, NULL);
 	if (unlikely(ret))
 		goto out_unreserve;
 
-	if (mem->mapped_to_gpu_memory == 0 &&
-	    !amdgpu_ttm_tt_get_usermm(bo->tbo.ttm)) {
-		/* Validate BO only once. The eviction fence gets added to BO
-		 * the first time it is mapped. Validate will wait for all
-		 * background evictions to complete.
-		 */
-		ret = amdgpu_amdkfd_bo_validate(bo, domain, true);
-		if (ret) {
-			pr_debug("Validate failed\n");
-			goto out_unreserve;
-		}
-	}
-
 	list_for_each_entry(entry, &mem->attachments, list) {
 		if (entry->bo_va->base.vm != avm || entry->is_mapped)
 			continue;
@@ -1994,10 +2090,6 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu(
 			 mem->mapped_to_gpu_memory);
 	}
 
-	if (!amdgpu_ttm_tt_get_usermm(bo->tbo.ttm) && !bo->tbo.pin_count)
-		dma_resv_add_fence(bo->tbo.base.resv,
-				   &avm->process_info->eviction_fence->base,
-				   DMA_RESV_USAGE_BOOKKEEP);
 	ret = unreserve_bo_and_vms(&ctx, false, false);
 
 	goto out;
@@ -2010,11 +2102,41 @@ out:
 	return ret;
 }
 
+int amdgpu_amdkfd_gpuvm_dmaunmap_mem(struct kgd_mem *mem, void *drm_priv)
+{
+	struct kfd_mem_attachment *entry;
+	struct amdgpu_vm *vm;
+	int ret;
+
+	vm = drm_priv_to_vm(drm_priv);
+
+	mutex_lock(&mem->lock);
+
+	ret = amdgpu_bo_reserve(mem->bo, true);
+	if (ret)
+		goto out;
+
+	list_for_each_entry(entry, &mem->attachments, list) {
+		if (entry->bo_va->base.vm != vm)
+			continue;
+		if (entry->bo_va->base.bo->tbo.ttm &&
+		    !entry->bo_va->base.bo->tbo.ttm->sg)
+			continue;
+
+		kfd_mem_dmaunmap_attachment(mem, entry);
+	}
+
+	amdgpu_bo_unreserve(mem->bo);
+out:
+	mutex_unlock(&mem->lock);
+
+	return ret;
+}
+
 int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
 		struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv)
 {
 	struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv);
-	struct amdkfd_process_info *process_info = avm->process_info;
 	unsigned long bo_size = mem->bo->tbo.base.size;
 	struct kfd_mem_attachment *entry;
 	struct bo_vm_reservation_context ctx;
@@ -2031,7 +2153,7 @@ int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
 		goto unreserve_out;
 	}
 
-	ret = vm_validate_pt_pd_bos(avm);
+	ret = vm_validate_pt_pd_bos(avm, NULL);
 	if (unlikely(ret))
 		goto unreserve_out;
 
@@ -2047,7 +2169,10 @@ int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
 		pr_debug("\t unmap VA 0x%llx - 0x%llx from entry %p\n",
 			 entry->va, entry->va + bo_size, entry);
 
-		unmap_bo_from_gpuvm(mem, entry, ctx.sync);
+		ret = unmap_bo_from_gpuvm(mem, entry, ctx.sync);
+		if (ret)
+			goto unreserve_out;
+
 		entry->is_mapped = false;
 
 		mem->mapped_to_gpu_memory--;
@@ -2055,15 +2180,6 @@ int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
 			 mem->mapped_to_gpu_memory);
 	}
 
-	/* If BO is unmapped from all VMs, unfence it. It can be evicted if
-	 * required.
-	 */
-	if (mem->mapped_to_gpu_memory == 0 &&
-	    !amdgpu_ttm_tt_get_usermm(mem->bo->tbo.ttm) &&
-	    !mem->bo->tbo.pin_count)
-		amdgpu_amdkfd_remove_eviction_fence(mem->bo,
-						process_info->eviction_fence);
-
 unreserve_out:
 	unreserve_bo_and_vms(&ctx, false, false);
 out:
@@ -2090,13 +2206,13 @@ int amdgpu_amdkfd_gpuvm_sync_memory(
 
 /**
  * amdgpu_amdkfd_map_gtt_bo_to_gart - Map BO to GART and increment reference count
- * @adev: Device to which allocated BO belongs
  * @bo: Buffer object to be mapped
+ * @bo_gart: Return bo reference
  *
  * Before return, bo reference count is incremented. To release the reference and unpin/
  * unmap the BO, call amdgpu_amdkfd_free_gtt_mem.
  */
-int amdgpu_amdkfd_map_gtt_bo_to_gart(struct amdgpu_device *adev, struct amdgpu_bo *bo)
+int amdgpu_amdkfd_map_gtt_bo_to_gart(struct amdgpu_bo *bo, struct amdgpu_bo **bo_gart)
 {
 	int ret;
 
@@ -2123,7 +2239,7 @@ int amdgpu_amdkfd_map_gtt_bo_to_gart(struct amdgpu_device *adev, struct amdgpu_b
 
 	amdgpu_bo_unreserve(bo);
 
-	bo = amdgpu_bo_ref(bo);
+	*bo_gart = amdgpu_bo_ref(bo);
 
 	return 0;
 
@@ -2213,7 +2329,7 @@ void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct kgd_mem *mem)
 {
 	struct amdgpu_bo *bo = mem->bo;
 
-	amdgpu_bo_reserve(bo, true);
+	(void)amdgpu_bo_reserve(bo, true);
 	amdgpu_bo_kunmap(bo);
 	amdgpu_bo_unpin(bo);
 	amdgpu_bo_unreserve(bo);
@@ -2222,42 +2338,33 @@ void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct kgd_mem *mem)
 int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct amdgpu_device *adev,
 					  struct kfd_vm_fault_info *mem)
 {
-	if (atomic_read(&adev->gmc.vm_fault_info_updated) == 1) {
+	if (atomic_read_acquire(&adev->gmc.vm_fault_info_updated) == 1) {
 		*mem = *adev->gmc.vm_fault_info;
-		mb(); /* make sure read happened */
-		atomic_set(&adev->gmc.vm_fault_info_updated, 0);
+		atomic_set_release(&adev->gmc.vm_fault_info_updated, 0);
 	}
 	return 0;
 }
 
-int amdgpu_amdkfd_gpuvm_import_dmabuf(struct amdgpu_device *adev,
-				      struct dma_buf *dma_buf,
-				      uint64_t va, void *drm_priv,
-				      struct kgd_mem **mem, uint64_t *size,
-				      uint64_t *mmap_offset)
+static int import_obj_create(struct amdgpu_device *adev,
+			     struct dma_buf *dma_buf,
+			     struct drm_gem_object *obj,
+			     uint64_t va, void *drm_priv,
+			     struct kgd_mem **mem, uint64_t *size,
+			     uint64_t *mmap_offset)
 {
 	struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv);
-	struct drm_gem_object *obj;
 	struct amdgpu_bo *bo;
 	int ret;
 
-	obj = amdgpu_gem_prime_import(adev_to_drm(adev), dma_buf);
-	if (IS_ERR(obj))
-		return PTR_ERR(obj);
-
 	bo = gem_to_amdgpu_bo(obj);
 	if (!(bo->preferred_domains & (AMDGPU_GEM_DOMAIN_VRAM |
-				    AMDGPU_GEM_DOMAIN_GTT))) {
+				    AMDGPU_GEM_DOMAIN_GTT)))
 		/* Only VRAM and GTT BOs are supported */
-		ret = -EINVAL;
-		goto err_put_obj;
-	}
+		return -EINVAL;
 
 	*mem = kzalloc(sizeof(struct kgd_mem), GFP_KERNEL);
-	if (!*mem) {
-		ret = -ENOMEM;
-		goto err_put_obj;
-	}
+	if (!*mem)
+		return -ENOMEM;
 
 	ret = drm_vma_node_allow(&obj->vma_node, drm_priv);
 	if (ret)
@@ -2282,20 +2389,67 @@ int amdgpu_amdkfd_gpuvm_import_dmabuf(struct amdgpu_device *adev,
 	(*mem)->dmabuf = dma_buf;
 	(*mem)->bo = bo;
 	(*mem)->va = va;
-	(*mem)->domain = (bo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM) ?
-		AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT;
+	(*mem)->domain = (bo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM) &&
+			 !adev->apu_prefer_gtt ?
+			 AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT;
+
 	(*mem)->mapped_to_gpu_memory = 0;
 	(*mem)->process_info = avm->process_info;
 	add_kgd_mem_to_kfd_bo_list(*mem, avm->process_info, false);
 	amdgpu_sync_create(&(*mem)->sync);
 	(*mem)->is_imported = true;
 
+	mutex_lock(&avm->process_info->lock);
+	if (avm->process_info->eviction_fence &&
+	    !dma_fence_is_signaled(&avm->process_info->eviction_fence->base))
+		ret = amdgpu_amdkfd_bo_validate_and_fence(bo, (*mem)->domain,
+				&avm->process_info->eviction_fence->base);
+	mutex_unlock(&avm->process_info->lock);
+	if (ret)
+		goto err_remove_mem;
+
 	return 0;
 
+err_remove_mem:
+	remove_kgd_mem_from_kfd_bo_list(*mem, avm->process_info);
+	drm_vma_node_revoke(&obj->vma_node, drm_priv);
 err_free_mem:
 	kfree(*mem);
+	return ret;
+}
+
+int amdgpu_amdkfd_gpuvm_import_dmabuf_fd(struct amdgpu_device *adev, int fd,
+					 uint64_t va, void *drm_priv,
+					 struct kgd_mem **mem, uint64_t *size,
+					 uint64_t *mmap_offset)
+{
+	struct drm_gem_object *obj;
+	uint32_t handle;
+	int ret;
+
+	ret = drm_gem_prime_fd_to_handle(&adev->ddev, adev->kfd.client.file, fd,
+					 &handle);
+	if (ret)
+		return ret;
+	obj = drm_gem_object_lookup(adev->kfd.client.file, handle);
+	if (!obj) {
+		ret = -EINVAL;
+		goto err_release_handle;
+	}
+
+	ret = import_obj_create(adev, obj->dma_buf, obj, va, drm_priv, mem, size,
+				mmap_offset);
+	if (ret)
+		goto err_put_obj;
+
+	(*mem)->gem_handle = handle;
+
+	return 0;
+
 err_put_obj:
 	drm_gem_object_put(obj);
+err_release_handle:
+	drm_gem_handle_delete(adev->kfd.client.file, handle);
 	return ret;
 }
 
@@ -2346,10 +2500,14 @@ int amdgpu_amdkfd_evict_userptr(struct mmu_interval_notifier *mni,
 		/* First eviction, stop the queues */
 		r = kgd2kfd_quiesce_mm(mni->mm,
 				       KFD_QUEUE_EVICTION_TRIGGER_USERPTR);
-		if (r)
+
+		if (r && r != -ESRCH)
 			pr_err("Failed to quiesce KFD\n");
-		schedule_delayed_work(&process_info->restore_userptr_work,
-			msecs_to_jiffies(AMDGPU_USERPTR_RESTORE_DELAY_MS));
+
+		if (r != -ESRCH)
+			queue_delayed_work(system_freezable_wq,
+				&process_info->restore_userptr_work,
+				msecs_to_jiffies(AMDGPU_USERPTR_RESTORE_DELAY_MS));
 	}
 	mutex_unlock(&process_info->notifier_lock);
 
@@ -2376,14 +2534,14 @@ static int update_invalid_user_pages(struct amdkfd_process_info *process_info,
 	/* Move all invalidated BOs to the userptr_inval_list */
 	list_for_each_entry_safe(mem, tmp_mem,
 				 &process_info->userptr_valid_list,
-				 validate_list.head)
+				 validate_list)
 		if (mem->invalid)
-			list_move_tail(&mem->validate_list.head,
+			list_move_tail(&mem->validate_list,
 				       &process_info->userptr_inval_list);
 
 	/* Go through userptr_inval_list and update any invalid user_pages */
 	list_for_each_entry(mem, &process_info->userptr_inval_list,
-			    validate_list.head) {
+			    validate_list) {
 		invalid = mem->invalid;
 		if (!invalid)
 			/* BO hasn't been invalidated since the last
@@ -2393,7 +2551,7 @@ static int update_invalid_user_pages(struct amdkfd_process_info *process_info,
 
 		bo = mem->bo;
 
-		amdgpu_ttm_tt_discard_user_pages(bo->tbo.ttm, mem->range);
+		amdgpu_hmm_range_free(mem->range);
 		mem->range = NULL;
 
 		/* BO reservations and getting user pages (hmm_range_fault)
@@ -2417,10 +2575,14 @@ static int update_invalid_user_pages(struct amdkfd_process_info *process_info,
 			}
 		}
 
+		mem->range = amdgpu_hmm_range_alloc(NULL);
+		if (unlikely(!mem->range))
+			return -ENOMEM;
 		/* Get updated user pages */
-		ret = amdgpu_ttm_tt_get_user_pages(bo, bo->tbo.ttm->pages,
-						   &mem->range);
+		ret = amdgpu_ttm_tt_get_user_pages(bo, mem->range);
 		if (ret) {
+			amdgpu_hmm_range_free(mem->range);
+			mem->range = NULL;
 			pr_debug("Failed %d to get user pages\n", ret);
 
 			/* Return -EFAULT bad address error as success. It will
@@ -2433,9 +2595,28 @@ static int update_invalid_user_pages(struct amdkfd_process_info *process_info,
 			if (ret != -EFAULT)
 				return ret;
 
+			/* If applications unmap memory before destroying the userptr
+			 * from the KFD, trigger a segmentation fault in VM debug mode.
+			 */
+			if (amdgpu_ttm_adev(bo->tbo.bdev)->debug_vm_userptr) {
+				struct kfd_process *p;
+
+				pr_err("Pid %d unmapped memory before destroying userptr at GPU addr 0x%llx\n",
+								pid_nr(process_info->pid), mem->va);
+
+				// Send GPU VM fault to user space
+				p = kfd_lookup_process_by_pid(process_info->pid);
+				if (p) {
+					kfd_signal_vm_fault_event_with_userptr(p, mem->va);
+					kfd_unref_process(p);
+				}
+			}
+
 			ret = 0;
 		}
 
+		amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, mem->range);
+
 		mutex_lock(&process_info->notifier_lock);
 
 		/* Mark the BO as valid unless it was invalidated
@@ -2445,7 +2626,9 @@ static int update_invalid_user_pages(struct amdkfd_process_info *process_info,
 			ret = -EAGAIN;
 			goto unlock_out;
 		}
-		mem->invalid = 0;
+		 /* set mem valid if mem has hmm range associated */
+		if (mem->range)
+			mem->invalid = 0;
 	}
 
 unlock_out:
@@ -2461,59 +2644,50 @@ unlock_out:
  */
 static int validate_invalid_user_pages(struct amdkfd_process_info *process_info)
 {
-	struct amdgpu_bo_list_entry *pd_bo_list_entries;
-	struct list_head resv_list, duplicates;
-	struct ww_acquire_ctx ticket;
+	struct ttm_operation_ctx ctx = { false, false };
 	struct amdgpu_sync sync;
+	struct drm_exec exec;
 
 	struct amdgpu_vm *peer_vm;
 	struct kgd_mem *mem, *tmp_mem;
 	struct amdgpu_bo *bo;
-	struct ttm_operation_ctx ctx = { false, false };
-	int i, ret;
-
-	pd_bo_list_entries = kcalloc(process_info->n_vms,
-				     sizeof(struct amdgpu_bo_list_entry),
-				     GFP_KERNEL);
-	if (!pd_bo_list_entries) {
-		pr_err("%s: Failed to allocate PD BO list entries\n", __func__);
-		ret = -ENOMEM;
-		goto out_no_mem;
-	}
-
-	INIT_LIST_HEAD(&resv_list);
-	INIT_LIST_HEAD(&duplicates);
+	int ret;
 
-	/* Get all the page directory BOs that need to be reserved */
-	i = 0;
-	list_for_each_entry(peer_vm, &process_info->vm_list_head,
-			    vm_list_node)
-		amdgpu_vm_get_pd_bo(peer_vm, &resv_list,
-				    &pd_bo_list_entries[i++]);
-	/* Add the userptr_inval_list entries to resv_list */
-	list_for_each_entry(mem, &process_info->userptr_inval_list,
-			    validate_list.head) {
-		list_add_tail(&mem->resv_list.head, &resv_list);
-		mem->resv_list.bo = mem->validate_list.bo;
-		mem->resv_list.num_shared = mem->validate_list.num_shared;
-	}
+	amdgpu_sync_create(&sync);
 
+	drm_exec_init(&exec, 0, 0);
 	/* Reserve all BOs and page tables for validation */
-	ret = ttm_eu_reserve_buffers(&ticket, &resv_list, false, &duplicates);
-	WARN(!list_empty(&duplicates), "Duplicates should be empty");
-	if (ret)
-		goto out_free;
+	drm_exec_until_all_locked(&exec) {
+		/* Reserve all the page directories */
+		list_for_each_entry(peer_vm, &process_info->vm_list_head,
+				    vm_list_node) {
+			ret = amdgpu_vm_lock_pd(peer_vm, &exec, 2);
+			drm_exec_retry_on_contention(&exec);
+			if (unlikely(ret))
+				goto unreserve_out;
+		}
 
-	amdgpu_sync_create(&sync);
+		/* Reserve the userptr_inval_list entries to resv_list */
+		list_for_each_entry(mem, &process_info->userptr_inval_list,
+				    validate_list) {
+			struct drm_gem_object *gobj;
 
-	ret = process_validate_vms(process_info);
+			gobj = &mem->bo->tbo.base;
+			ret = drm_exec_prepare_obj(&exec, gobj, 1);
+			drm_exec_retry_on_contention(&exec);
+			if (unlikely(ret))
+				goto unreserve_out;
+		}
+	}
+
+	ret = process_validate_vms(process_info, NULL);
 	if (ret)
 		goto unreserve_out;
 
 	/* Validate BOs and update GPUVM page tables */
 	list_for_each_entry_safe(mem, tmp_mem,
 				 &process_info->userptr_inval_list,
-				 validate_list.head) {
+				 validate_list) {
 		struct kfd_mem_attachment *attachment;
 
 		bo = mem->bo;
@@ -2555,12 +2729,9 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info)
 	ret = process_update_pds(process_info, &sync);
 
 unreserve_out:
-	ttm_eu_backoff_reservation(&ticket, &resv_list);
+	drm_exec_fini(&exec);
 	amdgpu_sync_wait(&sync, false);
 	amdgpu_sync_free(&sync);
-out_free:
-	kfree(pd_bo_list_entries);
-out_no_mem:
 
 	return ret;
 }
@@ -2576,9 +2747,16 @@ static int confirm_valid_user_pages_locked(struct amdkfd_process_info *process_i
 
 	list_for_each_entry_safe(mem, tmp_mem,
 				 &process_info->userptr_inval_list,
-				 validate_list.head) {
-		bool valid = amdgpu_ttm_tt_get_user_pages_done(
-				mem->bo->tbo.ttm, mem->range);
+				 validate_list) {
+		bool valid;
+
+		/* keep mem without hmm range at userptr_inval_list */
+		if (!mem->range)
+			continue;
+
+		/* Only check mem with hmm range associated */
+		valid = amdgpu_hmm_range_valid(mem->range);
+		amdgpu_hmm_range_free(mem->range);
 
 		mem->range = NULL;
 		if (!valid) {
@@ -2586,9 +2764,14 @@ static int confirm_valid_user_pages_locked(struct amdkfd_process_info *process_i
 			ret = -EAGAIN;
 			continue;
 		}
-		WARN(mem->invalid, "Valid BO is marked invalid");
 
-		list_move_tail(&mem->validate_list.head,
+		if (mem->invalid) {
+			WARN(1, "Valid BO is marked invalid");
+			ret = -EAGAIN;
+			continue;
+		}
+
+		list_move_tail(&mem->validate_list,
 			       &process_info->userptr_valid_list);
 	}
 
@@ -2669,7 +2852,8 @@ unlock_out:
 
 	/* If validation failed, reschedule another attempt */
 	if (evicted_bos) {
-		schedule_delayed_work(&process_info->restore_userptr_work,
+		queue_delayed_work(system_freezable_wq,
+			&process_info->restore_userptr_work,
 			msecs_to_jiffies(AMDGPU_USERPTR_RESTORE_DELAY_MS));
 
 		kfd_smi_event_queue_restore_rescheduled(mm);
@@ -2678,6 +2862,23 @@ unlock_out:
 	put_task_struct(usertask);
 }
 
+static void replace_eviction_fence(struct dma_fence __rcu **ef,
+				   struct dma_fence *new_ef)
+{
+	struct dma_fence *old_ef = rcu_replace_pointer(*ef, new_ef, true
+		/* protected by process_info->lock */);
+
+	/* If we're replacing an unsignaled eviction fence, that fence will
+	 * never be signaled, and if anyone is still waiting on that fence,
+	 * they will hang forever. This should never happen. We should only
+	 * replace the fence in restore_work that only gets scheduled after
+	 * eviction work signaled the fence.
+	 */
+	WARN_ONCE(!dma_fence_is_signaled(old_ef),
+		  "Replacing unsignaled eviction fence");
+	dma_fence_put(old_ef);
+}
+
 /** amdgpu_amdkfd_gpuvm_restore_process_bos - Restore all BOs for the given
  *   KFD process identified by process_info
  *
@@ -2696,74 +2897,59 @@ unlock_out:
  * 7.  Add fence to all PD and PT BOs.
  * 8.  Unreserve all BOs
  */
-int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef)
+int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence __rcu **ef)
 {
-	struct amdgpu_bo_list_entry *pd_bo_list;
 	struct amdkfd_process_info *process_info = info;
 	struct amdgpu_vm *peer_vm;
 	struct kgd_mem *mem;
-	struct bo_vm_reservation_context ctx;
-	struct amdgpu_amdkfd_fence *new_fence;
-	int ret = 0, i;
 	struct list_head duplicate_save;
 	struct amdgpu_sync sync_obj;
 	unsigned long failed_size = 0;
 	unsigned long total_size = 0;
+	struct drm_exec exec;
+	int ret;
 
 	INIT_LIST_HEAD(&duplicate_save);
-	INIT_LIST_HEAD(&ctx.list);
-	INIT_LIST_HEAD(&ctx.duplicates);
 
-	pd_bo_list = kcalloc(process_info->n_vms,
-			     sizeof(struct amdgpu_bo_list_entry),
-			     GFP_KERNEL);
-	if (!pd_bo_list)
-		return -ENOMEM;
-
-	i = 0;
 	mutex_lock(&process_info->lock);
-	list_for_each_entry(peer_vm, &process_info->vm_list_head,
-			vm_list_node)
-		amdgpu_vm_get_pd_bo(peer_vm, &ctx.list, &pd_bo_list[i++]);
-
-	/* Reserve all BOs and page tables/directory. Add all BOs from
-	 * kfd_bo_list to ctx.list
-	 */
-	list_for_each_entry(mem, &process_info->kfd_bo_list,
-			    validate_list.head) {
 
-		list_add_tail(&mem->resv_list.head, &ctx.list);
-		mem->resv_list.bo = mem->validate_list.bo;
-		mem->resv_list.num_shared = mem->validate_list.num_shared;
-	}
+	drm_exec_init(&exec, DRM_EXEC_IGNORE_DUPLICATES, 0);
+	drm_exec_until_all_locked(&exec) {
+		list_for_each_entry(peer_vm, &process_info->vm_list_head,
+				    vm_list_node) {
+			ret = amdgpu_vm_lock_pd(peer_vm, &exec, 2);
+			drm_exec_retry_on_contention(&exec);
+			if (unlikely(ret)) {
+				pr_err("Locking VM PD failed, ret: %d\n", ret);
+				goto ttm_reserve_fail;
+			}
+		}
 
-	ret = ttm_eu_reserve_buffers(&ctx.ticket, &ctx.list,
-				     false, &duplicate_save);
-	if (ret) {
-		pr_debug("Memory eviction: TTM Reserve Failed. Try again\n");
-		goto ttm_reserve_fail;
+		/* Reserve all BOs and page tables/directory. Add all BOs from
+		 * kfd_bo_list to ctx.list
+		 */
+		list_for_each_entry(mem, &process_info->kfd_bo_list,
+				    validate_list) {
+			struct drm_gem_object *gobj;
+
+			gobj = &mem->bo->tbo.base;
+			ret = drm_exec_prepare_obj(&exec, gobj, 1);
+			drm_exec_retry_on_contention(&exec);
+			if (unlikely(ret)) {
+				pr_err("drm_exec_prepare_obj failed, ret: %d\n", ret);
+				goto ttm_reserve_fail;
+			}
+		}
 	}
 
 	amdgpu_sync_create(&sync_obj);
 
-	/* Validate PDs and PTs */
-	ret = process_validate_vms(process_info);
-	if (ret)
-		goto validate_map_fail;
-
-	ret = process_sync_pds_resv(process_info, &sync_obj);
-	if (ret) {
-		pr_debug("Memory eviction: Failed to sync to PD BO moving fence. Try again\n");
-		goto validate_map_fail;
-	}
-
-	/* Validate BOs and map them to GPUVM (update VM page tables). */
+	/* Validate BOs managed by KFD */
 	list_for_each_entry(mem, &process_info->kfd_bo_list,
-			    validate_list.head) {
+			    validate_list) {
 
 		struct amdgpu_bo *bo = mem->bo;
 		uint32_t domain = mem->domain;
-		struct kfd_mem_attachment *attachment;
 		struct dma_resv_iter cursor;
 		struct dma_fence *fence;
 
@@ -2782,12 +2968,31 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef)
 		}
 		dma_resv_for_each_fence(&cursor, bo->tbo.base.resv,
 					DMA_RESV_USAGE_KERNEL, fence) {
-			ret = amdgpu_sync_fence(&sync_obj, fence);
+			ret = amdgpu_sync_fence(&sync_obj, fence, GFP_KERNEL);
 			if (ret) {
 				pr_debug("Memory eviction: Sync BO fence failed. Try again\n");
 				goto validate_map_fail;
 			}
 		}
+	}
+
+	if (failed_size)
+		pr_debug("0x%lx/0x%lx in system\n", failed_size, total_size);
+
+	/* Validate PDs, PTs and evicted DMABuf imports last. Otherwise BO
+	 * validations above would invalidate DMABuf imports again.
+	 */
+	ret = process_validate_vms(process_info, &exec.ticket);
+	if (ret) {
+		pr_debug("Validating VMs failed, ret: %d\n", ret);
+		goto validate_map_fail;
+	}
+
+	/* Update mappings managed by KFD. */
+	list_for_each_entry(mem, &process_info->kfd_bo_list,
+			    validate_list) {
+		struct kfd_mem_attachment *attachment;
+
 		list_for_each_entry(attachment, &mem->attachments, list) {
 			if (!attachment->is_mapped)
 				continue;
@@ -2801,8 +3006,31 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef)
 		}
 	}
 
-	if (failed_size)
-		pr_debug("0x%lx/0x%lx in system\n", failed_size, total_size);
+	/* Update mappings not managed by KFD */
+	list_for_each_entry(peer_vm, &process_info->vm_list_head,
+			vm_list_node) {
+		struct amdgpu_device *adev = amdgpu_ttm_adev(
+			peer_vm->root.bo->tbo.bdev);
+
+		struct amdgpu_fpriv *fpriv =
+			container_of(peer_vm, struct amdgpu_fpriv, vm);
+
+		ret = amdgpu_vm_bo_update(adev, fpriv->prt_va, false);
+		if (ret) {
+			dev_dbg(adev->dev,
+				"Memory eviction: handle PRT moved failed, pid %8d. Try again.\n",
+				pid_nr(process_info->pid));
+			goto validate_map_fail;
+		}
+
+		ret = amdgpu_vm_handle_moved(adev, peer_vm, &exec.ticket);
+		if (ret) {
+			dev_dbg(adev->dev,
+				"Memory eviction: handle moved failed, pid %8d. Try again.\n",
+				pid_nr(process_info->pid));
+			goto validate_map_fail;
+		}
+	}
 
 	/* Update page directories */
 	ret = process_update_pds(process_info, &sync_obj);
@@ -2811,29 +3039,50 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef)
 		goto validate_map_fail;
 	}
 
+	/* Sync with fences on all the page tables. They implicitly depend on any
+	 * move fences from amdgpu_vm_handle_moved above.
+	 */
+	ret = process_sync_pds_resv(process_info, &sync_obj);
+	if (ret) {
+		pr_debug("Memory eviction: Failed to sync to PD BO moving fence. Try again\n");
+		goto validate_map_fail;
+	}
+
 	/* Wait for validate and PT updates to finish */
 	amdgpu_sync_wait(&sync_obj, false);
 
-	/* Release old eviction fence and create new one, because fence only
-	 * goes from unsignaled to signaled, fence cannot be reused.
-	 * Use context and mm from the old fence.
+	/* The old eviction fence may be unsignaled if restore happens
+	 * after a GPU reset or suspend/resume. Keep the old fence in that
+	 * case. Otherwise release the old eviction fence and create new
+	 * one, because fence only goes from unsignaled to signaled once
+	 * and cannot be reused. Use context and mm from the old fence.
+	 *
+	 * If an old eviction fence signals after this check, that's OK.
+	 * Anyone signaling an eviction fence must stop the queues first
+	 * and schedule another restore worker.
 	 */
-	new_fence = amdgpu_amdkfd_fence_create(
+	if (dma_fence_is_signaled(&process_info->eviction_fence->base)) {
+		struct amdgpu_amdkfd_fence *new_fence =
+			amdgpu_amdkfd_fence_create(
 				process_info->eviction_fence->base.context,
 				process_info->eviction_fence->mm,
 				NULL);
-	if (!new_fence) {
-		pr_err("Failed to create eviction fence\n");
-		ret = -ENOMEM;
-		goto validate_map_fail;
+
+		if (!new_fence) {
+			pr_err("Failed to create eviction fence\n");
+			ret = -ENOMEM;
+			goto validate_map_fail;
+		}
+		dma_fence_put(&process_info->eviction_fence->base);
+		process_info->eviction_fence = new_fence;
+		replace_eviction_fence(ef, dma_fence_get(&new_fence->base));
+	} else {
+		WARN_ONCE(*ef != &process_info->eviction_fence->base,
+			  "KFD eviction fence doesn't match KGD process_info");
 	}
-	dma_fence_put(&process_info->eviction_fence->base);
-	process_info->eviction_fence = new_fence;
-	*ef = dma_fence_get(&new_fence->base);
 
 	/* Attach new eviction fence to all BOs except pinned ones */
-	list_for_each_entry(mem, &process_info->kfd_bo_list,
-		validate_list.head) {
+	list_for_each_entry(mem, &process_info->kfd_bo_list, validate_list) {
 		if (mem->bo->tbo.pin_count)
 			continue;
 
@@ -2841,7 +3090,7 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef)
 				   &process_info->eviction_fence->base,
 				   DMA_RESV_USAGE_BOOKKEEP);
 	}
-	/* Attach eviction fence to PD / PT BOs */
+	/* Attach eviction fence to PD / PT BOs and DMABuf imports */
 	list_for_each_entry(peer_vm, &process_info->vm_list_head,
 			    vm_list_node) {
 		struct amdgpu_bo *bo = peer_vm->root.bo;
@@ -2852,11 +3101,10 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef)
 	}
 
 validate_map_fail:
-	ttm_eu_backoff_reservation(&ctx.ticket, &ctx.list);
 	amdgpu_sync_free(&sync_obj);
 ttm_reserve_fail:
+	drm_exec_fini(&exec);
 	mutex_unlock(&process_info->lock);
-	kfree(pd_bo_list);
 	return ret;
 }
 
@@ -2972,12 +3220,13 @@ int amdgpu_amdkfd_get_tile_config(struct amdgpu_device *adev,
 	return 0;
 }
 
-bool amdgpu_amdkfd_bo_mapped_to_dev(struct amdgpu_device *adev, struct kgd_mem *mem)
+bool amdgpu_amdkfd_bo_mapped_to_dev(void *drm_priv, struct kgd_mem *mem)
 {
+	struct amdgpu_vm *vm = drm_priv_to_vm(drm_priv);
 	struct kfd_mem_attachment *entry;
 
 	list_for_each_entry(entry, &mem->attachments, list) {
-		if (entry->is_mapped && entry->adev == adev)
+		if (entry->is_mapped && entry->bo_va->base.vm == vm)
 			return true;
 	}
 	return false;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c
index 9ba4817a9148..763f2b8dcf13 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c
@@ -36,13 +36,6 @@
 #include "atombios_encoders.h"
 #include "bif/bif_4_1_d.h"
 
-static void amdgpu_atombios_lookup_i2c_gpio_quirks(struct amdgpu_device *adev,
-					  ATOM_GPIO_I2C_ASSIGMENT *gpio,
-					  u8 index)
-{
-
-}
-
 static struct amdgpu_i2c_bus_rec amdgpu_atombios_get_bus_rec_for_i2c_gpio(ATOM_GPIO_I2C_ASSIGMENT *gpio)
 {
 	struct amdgpu_i2c_bus_rec i2c;
@@ -108,9 +101,6 @@ struct amdgpu_i2c_bus_rec amdgpu_atombios_lookup_i2c_gpio(struct amdgpu_device *
 
 		gpio = &i2c_info->asGPIO_Info[0];
 		for (i = 0; i < num_indices; i++) {
-
-			amdgpu_atombios_lookup_i2c_gpio_quirks(adev, gpio, i);
-
 			if (gpio->sucI2cId.ucAccess == id) {
 				i2c = amdgpu_atombios_get_bus_rec_for_i2c_gpio(gpio);
 				break;
@@ -142,8 +132,6 @@ void amdgpu_atombios_i2c_init(struct amdgpu_device *adev)
 
 		gpio = &i2c_info->asGPIO_Info[0];
 		for (i = 0; i < num_indices; i++) {
-			amdgpu_atombios_lookup_i2c_gpio_quirks(adev, gpio, i);
-
 			i2c = amdgpu_atombios_get_bus_rec_for_i2c_gpio(gpio);
 
 			if (i2c.valid) {
@@ -156,6 +144,38 @@ void amdgpu_atombios_i2c_init(struct amdgpu_device *adev)
 	}
 }
 
+void amdgpu_atombios_oem_i2c_init(struct amdgpu_device *adev, u8 i2c_id)
+{
+	struct atom_context *ctx = adev->mode_info.atom_context;
+	ATOM_GPIO_I2C_ASSIGMENT *gpio;
+	struct amdgpu_i2c_bus_rec i2c;
+	int index = GetIndexIntoMasterTable(DATA, GPIO_I2C_Info);
+	struct _ATOM_GPIO_I2C_INFO *i2c_info;
+	uint16_t data_offset, size;
+	int i, num_indices;
+	char stmp[32];
+
+	if (amdgpu_atom_parse_data_header(ctx, index, &size, NULL, NULL, &data_offset)) {
+		i2c_info = (struct _ATOM_GPIO_I2C_INFO *)(ctx->bios + data_offset);
+
+		num_indices = (size - sizeof(ATOM_COMMON_TABLE_HEADER)) /
+			sizeof(ATOM_GPIO_I2C_ASSIGMENT);
+
+		gpio = &i2c_info->asGPIO_Info[0];
+		for (i = 0; i < num_indices; i++) {
+			i2c = amdgpu_atombios_get_bus_rec_for_i2c_gpio(gpio);
+
+			if (i2c.valid && i2c.i2c_id == i2c_id) {
+				sprintf(stmp, "OEM 0x%x", i2c.i2c_id);
+				adev->i2c_bus[i] = amdgpu_i2c_create(adev_to_drm(adev), &i2c, stmp);
+				break;
+			}
+			gpio = (ATOM_GPIO_I2C_ASSIGMENT *)
+				((u8 *)gpio + sizeof(ATOM_GPIO_I2C_ASSIGMENT));
+		}
+	}
+}
+
 struct amdgpu_gpio_rec
 amdgpu_atombios_lookup_gpio(struct amdgpu_device *adev,
 			    u8 id)
@@ -686,7 +706,6 @@ int amdgpu_atombios_get_clock_info(struct amdgpu_device *adev)
 		}
 		adev->clock.dp_extclk =
 			le16_to_cpu(firmware_info->info_21.usUniphyDPModeExtClkFreq);
-		adev->clock.current_dispclk = adev->clock.default_dispclk;
 
 		adev->clock.max_pixel_clock = le16_to_cpu(firmware_info->info.usMaxPixelClock);
 		if (adev->clock.max_pixel_clock == 0)
@@ -1018,7 +1037,9 @@ int amdgpu_atombios_get_clock_dividers(struct amdgpu_device *adev,
 		if (clock_type == COMPUTE_ENGINE_PLL_PARAM) {
 			args.v3.ulClockParams = cpu_to_le32((clock_type << 24) | clock);
 
-			amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+			if (amdgpu_atom_execute_table(adev->mode_info.atom_context,
+			    index, (uint32_t *)&args, sizeof(args)))
+				return -EINVAL;
 
 			dividers->post_div = args.v3.ucPostDiv;
 			dividers->enable_post_div = (args.v3.ucCntlFlag &
@@ -1038,7 +1059,9 @@ int amdgpu_atombios_get_clock_dividers(struct amdgpu_device *adev,
 			if (strobe_mode)
 				args.v5.ucInputFlag = ATOM_PLL_INPUT_FLAG_PLL_STROBE_MODE_EN;
 
-			amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+			if (amdgpu_atom_execute_table(adev->mode_info.atom_context,
+			    index, (uint32_t *)&args, sizeof(args)))
+				return -EINVAL;
 
 			dividers->post_div = args.v5.ucPostDiv;
 			dividers->enable_post_div = (args.v5.ucCntlFlag &
@@ -1056,7 +1079,9 @@ int amdgpu_atombios_get_clock_dividers(struct amdgpu_device *adev,
 		/* fusion */
 		args.v4.ulClock = cpu_to_le32(clock);	/* 10 khz */
 
-		amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+		if (amdgpu_atom_execute_table(adev->mode_info.atom_context,
+		    index, (uint32_t *)&args, sizeof(args)))
+			return -EINVAL;
 
 		dividers->post_divider = dividers->post_div = args.v4.ucPostDiv;
 		dividers->real_clock = le32_to_cpu(args.v4.ulClock);
@@ -1067,7 +1092,9 @@ int amdgpu_atombios_get_clock_dividers(struct amdgpu_device *adev,
 		args.v6_in.ulClock.ulComputeClockFlag = clock_type;
 		args.v6_in.ulClock.ulClockFreq = cpu_to_le32(clock);	/* 10 khz */
 
-		amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+		if (amdgpu_atom_execute_table(adev->mode_info.atom_context,
+		    index, (uint32_t *)&args, sizeof(args)))
+			return -EINVAL;
 
 		dividers->whole_fb_div = le16_to_cpu(args.v6_out.ulFbDiv.usFbDiv);
 		dividers->frac_fb_div = le16_to_cpu(args.v6_out.ulFbDiv.usFbDivFrac);
@@ -1109,7 +1136,9 @@ int amdgpu_atombios_get_memory_pll_dividers(struct amdgpu_device *adev,
 			if (strobe_mode)
 				args.ucInputFlag |= MPLL_INPUT_FLAG_STROBE_MODE_EN;
 
-			amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+			if (amdgpu_atom_execute_table(adev->mode_info.atom_context,
+			    index, (uint32_t *)&args, sizeof(args)))
+				return -EINVAL;
 
 			mpll_param->clkfrac = le16_to_cpu(args.ulFbDiv.usFbDivFrac);
 			mpll_param->clkf = le16_to_cpu(args.ulFbDiv.usFbDiv);
@@ -1135,8 +1164,8 @@ int amdgpu_atombios_get_memory_pll_dividers(struct amdgpu_device *adev,
 	return 0;
 }
 
-void amdgpu_atombios_set_engine_dram_timings(struct amdgpu_device *adev,
-					     u32 eng_clock, u32 mem_clock)
+int amdgpu_atombios_set_engine_dram_timings(struct amdgpu_device *adev,
+					    u32 eng_clock, u32 mem_clock)
 {
 	SET_ENGINE_CLOCK_PS_ALLOCATION args;
 	int index = GetIndexIntoMasterTable(COMMAND, DynamicMemorySettings);
@@ -1151,7 +1180,8 @@ void amdgpu_atombios_set_engine_dram_timings(struct amdgpu_device *adev,
 	if (mem_clock)
 		args.sReserved.ulClock = cpu_to_le32(mem_clock & SET_CLOCK_FREQ_MASK);
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	return amdgpu_atom_execute_table(adev->mode_info.atom_context, index,
+					 (uint32_t *)&args, sizeof(args));
 }
 
 void amdgpu_atombios_get_default_voltages(struct amdgpu_device *adev,
@@ -1205,7 +1235,9 @@ int amdgpu_atombios_get_max_vddc(struct amdgpu_device *adev, u8 voltage_type,
 		args.v2.ucVoltageMode = 0;
 		args.v2.usVoltageLevel = 0;
 
-		amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+		if (amdgpu_atom_execute_table(adev->mode_info.atom_context,
+		    index, (uint32_t *)&args, sizeof(args)))
+			return -EINVAL;
 
 		*voltage = le16_to_cpu(args.v2.usVoltageLevel);
 		break;
@@ -1214,7 +1246,9 @@ int amdgpu_atombios_get_max_vddc(struct amdgpu_device *adev, u8 voltage_type,
 		args.v3.ucVoltageMode = ATOM_GET_VOLTAGE_LEVEL;
 		args.v3.usVoltageLevel = cpu_to_le16(voltage_id);
 
-		amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+		if (amdgpu_atom_execute_table(adev->mode_info.atom_context,
+		    index, (uint32_t *)&args, sizeof(args)))
+			return -EINVAL;
 
 		*voltage = le16_to_cpu(args.v3.usVoltageLevel);
 		break;
@@ -1476,6 +1510,8 @@ int amdgpu_atombios_init_mc_reg_table(struct amdgpu_device *adev,
 										(u32)le32_to_cpu(*((u32 *)reg_data + j));
 									j++;
 								} else if ((reg_table->mc_reg_address[i].pre_reg_data & LOW_NIBBLE_MASK) == DATA_EQU_PREV) {
+									if (i == 0)
+										continue;
 									reg_table->mc_reg_table_entry[num_ranges].mc_data[i] =
 										reg_table->mc_reg_table_entry[num_ranges].mc_data[i - 1];
 								}
@@ -1776,21 +1812,57 @@ static ssize_t amdgpu_atombios_get_vbios_version(struct device *dev,
 	struct amdgpu_device *adev = drm_to_adev(ddev);
 	struct atom_context *ctx = adev->mode_info.atom_context;
 
-	return sysfs_emit(buf, "%s\n", ctx->vbios_version);
+	return sysfs_emit(buf, "%s\n", ctx->vbios_pn);
+}
+
+static ssize_t amdgpu_atombios_get_vbios_build(struct device *dev,
+					       struct device_attribute *attr,
+					       char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	struct atom_context *ctx = adev->mode_info.atom_context;
+
+	return sysfs_emit(buf, "%s\n", ctx->build_num);
 }
 
 static DEVICE_ATTR(vbios_version, 0444, amdgpu_atombios_get_vbios_version,
 		   NULL);
+static DEVICE_ATTR(vbios_build, 0444, amdgpu_atombios_get_vbios_build, NULL);
 
 static struct attribute *amdgpu_vbios_version_attrs[] = {
-	&dev_attr_vbios_version.attr,
-	NULL
+	&dev_attr_vbios_version.attr, &dev_attr_vbios_build.attr, NULL
 };
 
+static umode_t amdgpu_vbios_version_attrs_is_visible(struct kobject *kobj,
+						     struct attribute *attr,
+						     int index)
+{
+	struct device *dev = kobj_to_dev(kobj);
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	struct atom_context *ctx = adev->mode_info.atom_context;
+
+	if (attr == &dev_attr_vbios_build.attr && !strlen(ctx->build_num))
+		return 0;
+
+	return attr->mode;
+}
+
 const struct attribute_group amdgpu_vbios_version_attr_group = {
-	.attrs = amdgpu_vbios_version_attrs
+	.attrs = amdgpu_vbios_version_attrs,
+	.is_visible = amdgpu_vbios_version_attrs_is_visible,
 };
 
+int amdgpu_atombios_sysfs_init(struct amdgpu_device *adev)
+{
+	if (adev->mode_info.atom_context)
+		return devm_device_add_group(adev->dev,
+					     &amdgpu_vbios_version_attr_group);
+
+	return 0;
+}
+
 /**
  * amdgpu_atombios_fini - free the driver info and callbacks for atombios
  *
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h
index 4153d520e2a3..867bc5c5ce67 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h
@@ -89,8 +89,7 @@ struct atom_memory_info {
 
 #define MAX_AC_TIMING_ENTRIES 16
 
-struct atom_memory_clock_range_table
-{
+struct atom_memory_clock_range_table {
 	u8 num_entries;
 	u8 rsv[3];
 	u32 mclk[MAX_AC_TIMING_ENTRIES];
@@ -118,14 +117,12 @@ struct atom_mc_reg_table {
 
 #define MAX_VOLTAGE_ENTRIES 32
 
-struct atom_voltage_table_entry
-{
+struct atom_voltage_table_entry {
 	u16 value;
 	u32 smio_low;
 };
 
-struct atom_voltage_table
-{
+struct atom_voltage_table {
 	u32 count;
 	u32 mask_low;
 	u32 phase_delay;
@@ -139,6 +136,7 @@ amdgpu_atombios_lookup_gpio(struct amdgpu_device *adev,
 struct amdgpu_i2c_bus_rec amdgpu_atombios_lookup_i2c_gpio(struct amdgpu_device *adev,
 							  uint8_t id);
 void amdgpu_atombios_i2c_init(struct amdgpu_device *adev);
+void amdgpu_atombios_oem_i2c_init(struct amdgpu_device *adev, u8 i2c_id);
 
 bool amdgpu_atombios_has_dce_engine_info(struct amdgpu_device *adev);
 
@@ -166,8 +164,8 @@ int amdgpu_atombios_get_memory_pll_dividers(struct amdgpu_device *adev,
 					    bool strobe_mode,
 					    struct atom_mpll_param *mpll_param);
 
-void amdgpu_atombios_set_engine_dram_timings(struct amdgpu_device *adev,
-					     u32 eng_clock, u32 mem_clock);
+int amdgpu_atombios_set_engine_dram_timings(struct amdgpu_device *adev,
+					    u32 eng_clock, u32 mem_clock);
 
 bool
 amdgpu_atombios_is_voltage_gpio(struct amdgpu_device *adev,
@@ -217,5 +215,6 @@ int amdgpu_atombios_get_data_table(struct amdgpu_device *adev,
 
 void amdgpu_atombios_fini(struct amdgpu_device *adev);
 int amdgpu_atombios_init(struct amdgpu_device *adev);
+int amdgpu_atombios_sysfs_init(struct amdgpu_device *adev);
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c
index ac6fe0ae4609..636385c80f64 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c
@@ -34,6 +34,7 @@ union firmware_info {
 	struct atom_firmware_info_v3_2 v32;
 	struct atom_firmware_info_v3_3 v33;
 	struct atom_firmware_info_v3_4 v34;
+	struct atom_firmware_info_v3_5 v35;
 };
 
 /*
@@ -58,7 +59,7 @@ uint32_t amdgpu_atomfirmware_query_firmware_capability(struct amdgpu_device *ade
 	if (amdgpu_atom_parse_data_header(adev->mode_info.atom_context,
 				index, &size, &frev, &crev, &data_offset)) {
 		/* support firmware_info 3.1 + */
-		if ((frev == 3 && crev >=1) || (frev > 3)) {
+		if ((frev == 3 && crev >= 1) || (frev > 3)) {
 			firmware_info = (union firmware_info *)
 				(mode_info->atom_context->bios + data_offset);
 			fw_cap = le32_to_cpu(firmware_info->v31.firmware_capability);
@@ -180,19 +181,22 @@ int amdgpu_atomfirmware_allocate_fb_scratch(struct amdgpu_device *adev)
 	u8 frev, crev;
 	int usage_bytes = 0;
 
-	if (amdgpu_atom_parse_data_header(ctx, index, NULL, &frev, &crev, &data_offset)) {
-		if (frev == 2 && crev == 1) {
-			fw_usage_v2_1 =
-				(struct vram_usagebyfirmware_v2_1 *)(ctx->bios + data_offset);
-			amdgpu_atomfirmware_allocate_fb_v2_1(adev,
-					fw_usage_v2_1,
-					&usage_bytes);
-		} else if (frev >= 2 && crev >= 2) {
-			fw_usage_v2_2 =
-				(struct vram_usagebyfirmware_v2_2 *)(ctx->bios + data_offset);
-			amdgpu_atomfirmware_allocate_fb_v2_2(adev,
-					fw_usage_v2_2,
-					&usage_bytes);
+	/* Skip atomfirmware allocation for SRIOV VFs when dynamic crit regn is enabled */
+	if (!(amdgpu_sriov_vf(adev) && adev->virt.is_dynamic_crit_regn_enabled)) {
+		if (amdgpu_atom_parse_data_header(ctx, index, NULL, &frev, &crev, &data_offset)) {
+			if (frev == 2 && crev == 1) {
+				fw_usage_v2_1 =
+					(struct vram_usagebyfirmware_v2_1 *)(ctx->bios + data_offset);
+				amdgpu_atomfirmware_allocate_fb_v2_1(adev,
+						fw_usage_v2_1,
+						&usage_bytes);
+			} else if (frev >= 2 && crev >= 2) {
+				fw_usage_v2_2 =
+					(struct vram_usagebyfirmware_v2_2 *)(ctx->bios + data_offset);
+				amdgpu_atomfirmware_allocate_fb_v2_2(adev,
+						fw_usage_v2_2,
+						&usage_bytes);
+			}
 		}
 	}
 
@@ -211,12 +215,14 @@ union igp_info {
 	struct atom_integrated_system_info_v1_11 v11;
 	struct atom_integrated_system_info_v1_12 v12;
 	struct atom_integrated_system_info_v2_1 v21;
+	struct atom_integrated_system_info_v2_3 v23;
 };
 
 union umc_info {
 	struct atom_umc_info_v3_1 v31;
 	struct atom_umc_info_v3_2 v32;
 	struct atom_umc_info_v3_3 v33;
+	struct atom_umc_info_v4_0 v40;
 };
 
 union vram_info {
@@ -272,11 +278,15 @@ static int convert_atom_mem_type_to_vram_type(struct amdgpu_device *adev,
 			break;
 		case ATOM_DGPU_VRAM_TYPE_HBM2:
 		case ATOM_DGPU_VRAM_TYPE_HBM2E:
+		case ATOM_DGPU_VRAM_TYPE_HBM3:
 			vram_type = AMDGPU_VRAM_TYPE_HBM;
 			break;
 		case ATOM_DGPU_VRAM_TYPE_GDDR6:
 			vram_type = AMDGPU_VRAM_TYPE_GDDR6;
 			break;
+		case ATOM_DGPU_VRAM_TYPE_HBM3E:
+			vram_type = AMDGPU_VRAM_TYPE_HBM3E;
+			break;
 		default:
 			vram_type = AMDGPU_VRAM_TYPE_UNKNOWN;
 			break;
@@ -286,7 +296,6 @@ static int convert_atom_mem_type_to_vram_type(struct amdgpu_device *adev,
 	return vram_type;
 }
 
-
 int
 amdgpu_atomfirmware_get_vram_info(struct amdgpu_device *adev,
 				  int *vram_width, int *vram_type,
@@ -297,6 +306,7 @@ amdgpu_atomfirmware_get_vram_info(struct amdgpu_device *adev,
 	u16 data_offset, size;
 	union igp_info *igp_info;
 	union vram_info *vram_info;
+	union umc_info *umc_info;
 	union vram_module *vram_module;
 	u8 frev, crev;
 	u8 mem_type;
@@ -308,10 +318,16 @@ amdgpu_atomfirmware_get_vram_info(struct amdgpu_device *adev,
 	if (adev->flags & AMD_IS_APU)
 		index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1,
 						    integratedsysteminfo);
-	else
-		index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1,
-						    vram_info);
-
+	else {
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+		case IP_VERSION(12, 0, 0):
+		case IP_VERSION(12, 0, 1):
+			index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, umc_info);
+			break;
+		default:
+			index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, vram_info);
+		}
+	}
 	if (amdgpu_atom_parse_data_header(mode_info->atom_context,
 					  index, &size,
 					  &frev, &crev, &data_offset)) {
@@ -326,10 +342,13 @@ amdgpu_atomfirmware_get_vram_info(struct amdgpu_device *adev,
 					mem_channel_number = igp_info->v11.umachannelnumber;
 					if (!mem_channel_number)
 						mem_channel_number = 1;
-					/* channel width is 64 */
-					if (vram_width)
-						*vram_width = mem_channel_number * 64;
 					mem_type = igp_info->v11.memorytype;
+					if (mem_type == LpDdr5MemType)
+						mem_channel_width = 32;
+					else
+						mem_channel_width = 64;
+					if (vram_width)
+						*vram_width = mem_channel_number * mem_channel_width;
 					if (vram_type)
 						*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
 					break;
@@ -344,10 +363,27 @@ amdgpu_atomfirmware_get_vram_info(struct amdgpu_device *adev,
 					mem_channel_number = igp_info->v21.umachannelnumber;
 					if (!mem_channel_number)
 						mem_channel_number = 1;
-					/* channel width is 64 */
-					if (vram_width)
-						*vram_width = mem_channel_number * 64;
 					mem_type = igp_info->v21.memorytype;
+					if (mem_type == LpDdr5MemType)
+						mem_channel_width = 32;
+					else
+						mem_channel_width = 64;
+					if (vram_width)
+						*vram_width = mem_channel_number * mem_channel_width;
+					if (vram_type)
+						*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
+					break;
+				case 3:
+					mem_channel_number = igp_info->v23.umachannelnumber;
+					if (!mem_channel_number)
+						mem_channel_number = 1;
+					mem_type = igp_info->v23.memorytype;
+					if (mem_type == LpDdr5MemType)
+						mem_channel_width = 32;
+					else
+						mem_channel_width = 64;
+					if (vram_width)
+						*vram_width = mem_channel_number * mem_channel_width;
 					if (vram_type)
 						*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
 					break;
@@ -359,123 +395,150 @@ amdgpu_atomfirmware_get_vram_info(struct amdgpu_device *adev,
 				return -EINVAL;
 			}
 		} else {
-			vram_info = (union vram_info *)
-				(mode_info->atom_context->bios + data_offset);
-			module_id = (RREG32(adev->bios_scratch_reg_offset + 4) & 0x00ff0000) >> 16;
-			if (frev == 3) {
-				switch (crev) {
-				/* v30 */
-				case 0:
-					vram_module = (union vram_module *)vram_info->v30.vram_module;
-					mem_vendor = (vram_module->v30.dram_vendor_id) & 0xF;
-					if (vram_vendor)
-						*vram_vendor = mem_vendor;
-					mem_type = vram_info->v30.memory_type;
-					if (vram_type)
-						*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
-					mem_channel_number = vram_info->v30.channel_num;
-					mem_channel_width = vram_info->v30.channel_width;
-					if (vram_width)
-						*vram_width = mem_channel_number * (1 << mem_channel_width);
-					break;
-				default:
-					return -EINVAL;
-				}
-			} else if (frev == 2) {
-				switch (crev) {
-				/* v23 */
-				case 3:
-					if (module_id > vram_info->v23.vram_module_num)
-						module_id = 0;
-					vram_module = (union vram_module *)vram_info->v23.vram_module;
-					while (i < module_id) {
-						vram_module = (union vram_module *)
-							((u8 *)vram_module + vram_module->v9.vram_module_size);
-						i++;
+			switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+			case IP_VERSION(12, 0, 0):
+			case IP_VERSION(12, 0, 1):
+				umc_info = (union umc_info *)(mode_info->atom_context->bios + data_offset);
+
+				if (frev == 4) {
+					switch (crev) {
+					case 0:
+						mem_channel_number = le32_to_cpu(umc_info->v40.channel_num);
+						mem_type = le32_to_cpu(umc_info->v40.vram_type);
+						mem_channel_width = le32_to_cpu(umc_info->v40.channel_width);
+						mem_vendor = RREG32(adev->bios_scratch_reg_offset + 4) & 0xF;
+						if (vram_vendor)
+							*vram_vendor = mem_vendor;
+						if (vram_type)
+							*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
+						if (vram_width)
+							*vram_width = mem_channel_number * (1 << mem_channel_width);
+						break;
+					default:
+						return -EINVAL;
 					}
-					mem_type = vram_module->v9.memory_type;
-					if (vram_type)
-						*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
-					mem_channel_number = vram_module->v9.channel_num;
-					mem_channel_width = vram_module->v9.channel_width;
-					if (vram_width)
-						*vram_width = mem_channel_number * (1 << mem_channel_width);
-					mem_vendor = (vram_module->v9.vender_rev_id) & 0xF;
-					if (vram_vendor)
-						*vram_vendor = mem_vendor;
-					break;
-				/* v24 */
-				case 4:
-					if (module_id > vram_info->v24.vram_module_num)
-						module_id = 0;
-					vram_module = (union vram_module *)vram_info->v24.vram_module;
-					while (i < module_id) {
-						vram_module = (union vram_module *)
-							((u8 *)vram_module + vram_module->v10.vram_module_size);
-						i++;
+				} else
+					return -EINVAL;
+				break;
+			default:
+				vram_info = (union vram_info *)
+					(mode_info->atom_context->bios + data_offset);
+
+				module_id = (RREG32(adev->bios_scratch_reg_offset + 4) & 0x00ff0000) >> 16;
+				if (frev == 3) {
+					switch (crev) {
+					/* v30 */
+					case 0:
+						vram_module = (union vram_module *)vram_info->v30.vram_module;
+						mem_vendor = (vram_module->v30.dram_vendor_id) & 0xF;
+						if (vram_vendor)
+							*vram_vendor = mem_vendor;
+						mem_type = vram_info->v30.memory_type;
+						if (vram_type)
+							*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
+						mem_channel_number = vram_info->v30.channel_num;
+						mem_channel_width = vram_info->v30.channel_width;
+						if (vram_width)
+							*vram_width = mem_channel_number * 16;
+						break;
+					default:
+						return -EINVAL;
 					}
-					mem_type = vram_module->v10.memory_type;
-					if (vram_type)
-						*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
-					mem_channel_number = vram_module->v10.channel_num;
-					mem_channel_width = vram_module->v10.channel_width;
-					if (vram_width)
-						*vram_width = mem_channel_number * (1 << mem_channel_width);
-					mem_vendor = (vram_module->v10.vender_rev_id) & 0xF;
-					if (vram_vendor)
-						*vram_vendor = mem_vendor;
-					break;
-				/* v25 */
-				case 5:
-					if (module_id > vram_info->v25.vram_module_num)
-						module_id = 0;
-					vram_module = (union vram_module *)vram_info->v25.vram_module;
-					while (i < module_id) {
-						vram_module = (union vram_module *)
-							((u8 *)vram_module + vram_module->v11.vram_module_size);
-						i++;
+				} else if (frev == 2) {
+					switch (crev) {
+					/* v23 */
+					case 3:
+						if (module_id > vram_info->v23.vram_module_num)
+							module_id = 0;
+						vram_module = (union vram_module *)vram_info->v23.vram_module;
+						while (i < module_id) {
+							vram_module = (union vram_module *)
+								((u8 *)vram_module + vram_module->v9.vram_module_size);
+							i++;
+						}
+						mem_type = vram_module->v9.memory_type;
+						if (vram_type)
+							*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
+						mem_channel_number = vram_module->v9.channel_num;
+						mem_channel_width = vram_module->v9.channel_width;
+						if (vram_width)
+							*vram_width = mem_channel_number * (1 << mem_channel_width);
+						mem_vendor = (vram_module->v9.vender_rev_id) & 0xF;
+						if (vram_vendor)
+							*vram_vendor = mem_vendor;
+						break;
+					/* v24 */
+					case 4:
+						if (module_id > vram_info->v24.vram_module_num)
+							module_id = 0;
+						vram_module = (union vram_module *)vram_info->v24.vram_module;
+						while (i < module_id) {
+							vram_module = (union vram_module *)
+								((u8 *)vram_module + vram_module->v10.vram_module_size);
+							i++;
+						}
+						mem_type = vram_module->v10.memory_type;
+						if (vram_type)
+							*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
+						mem_channel_number = vram_module->v10.channel_num;
+						mem_channel_width = vram_module->v10.channel_width;
+						if (vram_width)
+							*vram_width = mem_channel_number * (1 << mem_channel_width);
+						mem_vendor = (vram_module->v10.vender_rev_id) & 0xF;
+						if (vram_vendor)
+							*vram_vendor = mem_vendor;
+						break;
+					/* v25 */
+					case 5:
+						if (module_id > vram_info->v25.vram_module_num)
+							module_id = 0;
+						vram_module = (union vram_module *)vram_info->v25.vram_module;
+						while (i < module_id) {
+							vram_module = (union vram_module *)
+								((u8 *)vram_module + vram_module->v11.vram_module_size);
+							i++;
+						}
+						mem_type = vram_module->v11.memory_type;
+						if (vram_type)
+							*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
+						mem_channel_number = vram_module->v11.channel_num;
+						mem_channel_width = vram_module->v11.channel_width;
+						if (vram_width)
+							*vram_width = mem_channel_number * (1 << mem_channel_width);
+						mem_vendor = (vram_module->v11.vender_rev_id) & 0xF;
+						if (vram_vendor)
+							*vram_vendor = mem_vendor;
+						break;
+					/* v26 */
+					case 6:
+						if (module_id > vram_info->v26.vram_module_num)
+							module_id = 0;
+						vram_module = (union vram_module *)vram_info->v26.vram_module;
+						while (i < module_id) {
+							vram_module = (union vram_module *)
+								((u8 *)vram_module + vram_module->v9.vram_module_size);
+							i++;
+						}
+						mem_type = vram_module->v9.memory_type;
+						if (vram_type)
+							*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
+						mem_channel_number = vram_module->v9.channel_num;
+						mem_channel_width = vram_module->v9.channel_width;
+						if (vram_width)
+							*vram_width = mem_channel_number * (1 << mem_channel_width);
+						mem_vendor = (vram_module->v9.vender_rev_id) & 0xF;
+						if (vram_vendor)
+							*vram_vendor = mem_vendor;
+						break;
+					default:
+						return -EINVAL;
 					}
-					mem_type = vram_module->v11.memory_type;
-					if (vram_type)
-						*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
-					mem_channel_number = vram_module->v11.channel_num;
-					mem_channel_width = vram_module->v11.channel_width;
-					if (vram_width)
-						*vram_width = mem_channel_number * (1 << mem_channel_width);
-					mem_vendor = (vram_module->v11.vender_rev_id) & 0xF;
-					if (vram_vendor)
-						*vram_vendor = mem_vendor;
-					break;
-				/* v26 */
-				case 6:
-					if (module_id > vram_info->v26.vram_module_num)
-						module_id = 0;
-					vram_module = (union vram_module *)vram_info->v26.vram_module;
-					while (i < module_id) {
-						vram_module = (union vram_module *)
-							((u8 *)vram_module + vram_module->v9.vram_module_size);
-						i++;
-					}
-					mem_type = vram_module->v9.memory_type;
-					if (vram_type)
-						*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
-					mem_channel_number = vram_module->v9.channel_num;
-					mem_channel_width = vram_module->v9.channel_width;
-					if (vram_width)
-						*vram_width = mem_channel_number * (1 << mem_channel_width);
-					mem_vendor = (vram_module->v9.vender_rev_id) & 0xF;
-					if (vram_vendor)
-						*vram_vendor = mem_vendor;
-					break;
-				default:
+				} else {
+					/* invalid frev */
 					return -EINVAL;
 				}
-			} else {
-				/* invalid frev */
-				return -EINVAL;
 			}
 		}
-
 	}
 
 	return 0;
@@ -492,44 +555,63 @@ bool amdgpu_atomfirmware_mem_ecc_supported(struct amdgpu_device *adev)
 	u16 data_offset, size;
 	union umc_info *umc_info;
 	u8 frev, crev;
-	bool ecc_default_enabled = false;
+	bool mem_ecc_enabled = false;
 	u8 umc_config;
 	u32 umc_config1;
+	adev->ras_default_ecc_enabled = false;
 
 	index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1,
 			umc_info);
 
 	if (amdgpu_atom_parse_data_header(mode_info->atom_context,
 				index, &size, &frev, &crev, &data_offset)) {
+		umc_info = (union umc_info *)(mode_info->atom_context->bios + data_offset);
 		if (frev == 3) {
-			umc_info = (union umc_info *)
-				(mode_info->atom_context->bios + data_offset);
 			switch (crev) {
 			case 1:
 				umc_config = le32_to_cpu(umc_info->v31.umc_config);
-				ecc_default_enabled =
+				mem_ecc_enabled =
 					(umc_config & UMC_CONFIG__DEFAULT_MEM_ECC_ENABLE) ? true : false;
 				break;
 			case 2:
 				umc_config = le32_to_cpu(umc_info->v32.umc_config);
-				ecc_default_enabled =
+				mem_ecc_enabled =
 					(umc_config & UMC_CONFIG__DEFAULT_MEM_ECC_ENABLE) ? true : false;
 				break;
 			case 3:
 				umc_config = le32_to_cpu(umc_info->v33.umc_config);
 				umc_config1 = le32_to_cpu(umc_info->v33.umc_config1);
-				ecc_default_enabled =
+				mem_ecc_enabled =
 					((umc_config & UMC_CONFIG__DEFAULT_MEM_ECC_ENABLE) ||
 					 (umc_config1 & UMC_CONFIG1__ENABLE_ECC_CAPABLE)) ? true : false;
+				adev->ras_default_ecc_enabled =
+					(umc_config & UMC_CONFIG__DEFAULT_MEM_ECC_ENABLE) ? true : false;
 				break;
 			default:
 				/* unsupported crev */
 				return false;
 			}
+		} else if (frev == 4) {
+			switch (crev) {
+			case 0:
+				umc_config = le32_to_cpu(umc_info->v40.umc_config);
+				umc_config1 = le32_to_cpu(umc_info->v40.umc_config1);
+				mem_ecc_enabled =
+					(umc_config1 & UMC_CONFIG1__ENABLE_ECC_CAPABLE) ? true : false;
+				adev->ras_default_ecc_enabled =
+					(umc_config & UMC_CONFIG__DEFAULT_MEM_ECC_ENABLE) ? true : false;
+				break;
+			default:
+				/* unsupported crev */
+				return false;
+			}
+		} else {
+			/* unsupported frev */
+			return false;
 		}
 	}
 
-	return ecc_default_enabled;
+	return mem_ecc_enabled;
 }
 
 /*
@@ -590,7 +672,7 @@ bool amdgpu_atomfirmware_ras_rom_addr(struct amdgpu_device *adev,
 					  index, &size, &frev, &crev,
 					  &data_offset)) {
 		/* support firmware_info 3.4 + */
-		if ((frev == 3 && crev >=4) || (frev > 3)) {
+		if ((frev == 3 && crev >= 4) || (frev > 3)) {
 			firmware_info = (union firmware_info *)
 				(mode_info->atom_context->bios + data_offset);
 			/* The ras_rom_i2c_slave_addr should ideally
@@ -843,7 +925,7 @@ int amdgpu_atomfirmware_get_fw_reserved_fb_size(struct amdgpu_device *adev)
 
 	firmware_info = (union firmware_info *)(ctx->bios + data_offset);
 
-	if (frev !=3)
+	if (frev != 3)
 		return -EINVAL;
 
 	switch (crev) {
@@ -851,6 +933,10 @@ int amdgpu_atomfirmware_get_fw_reserved_fb_size(struct amdgpu_device *adev)
 		fw_reserved_fb_size =
 			(firmware_info->v34.fw_reserved_size_in_kb << 10);
 		break;
+	case 5:
+		fw_reserved_fb_size =
+			(firmware_info->v35.fw_reserved_size_in_kb << 10);
+		break;
 	default:
 		fw_reserved_fb_size = 0;
 		break;
@@ -902,7 +988,7 @@ int amdgpu_atomfirmware_asic_init(struct amdgpu_device *adev, bool fb_reset)
 	}
 
 	index = get_index_into_master_table(atom_master_list_of_command_functions_v2_1,
-                                            asic_init);
+					asic_init);
 	if (amdgpu_atom_parse_cmd_header(mode_info->atom_context, index, &frev, &crev)) {
 		if (frev == 2 && crev >= 1) {
 			memset(&asic_init_ps_v2_1, 0, sizeof(asic_init_ps_v2_1));
@@ -920,5 +1006,6 @@ int amdgpu_atomfirmware_asic_init(struct amdgpu_device *adev, bool fb_reset)
 		return -EINVAL;
 	}
 
-	return amdgpu_atom_execute_table(ctx, ATOM_CMD_INIT, (uint32_t *)&asic_init_ps_v2_1);
+	return amdgpu_atom_execute_table(ctx, ATOM_CMD_INIT, (uint32_t *)&asic_init_ps_v2_1,
+		sizeof(asic_init_ps_v2_1));
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h
index c7eb2caec65a..649b5530d8ae 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h
@@ -36,7 +36,7 @@ int amdgpu_atomfirmware_get_clock_info(struct amdgpu_device *adev);
 int amdgpu_atomfirmware_get_gfx_info(struct amdgpu_device *adev);
 bool amdgpu_atomfirmware_mem_ecc_supported(struct amdgpu_device *adev);
 bool amdgpu_atomfirmware_sram_ecc_supported(struct amdgpu_device *adev);
-bool amdgpu_atomfirmware_ras_rom_addr(struct amdgpu_device *adev, uint8_t* i2c_address);
+bool amdgpu_atomfirmware_ras_rom_addr(struct amdgpu_device *adev, uint8_t *i2c_address);
 bool amdgpu_atomfirmware_mem_training_supported(struct amdgpu_device *adev);
 bool amdgpu_atomfirmware_dynamic_boot_config_supported(struct amdgpu_device *adev);
 int amdgpu_atomfirmware_get_fw_reserved_fb_size(struct amdgpu_device *adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c
index d6d986be906a..3893e6fc2f03 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c
@@ -74,28 +74,21 @@ struct atpx_mux {
 	u16 mux;
 } __packed;
 
-bool amdgpu_has_atpx(void) {
+bool amdgpu_has_atpx(void)
+{
 	return amdgpu_atpx_priv.atpx_detected;
 }
 
-bool amdgpu_has_atpx_dgpu_power_cntl(void) {
+bool amdgpu_has_atpx_dgpu_power_cntl(void)
+{
 	return amdgpu_atpx_priv.atpx.functions.power_cntl;
 }
 
-bool amdgpu_is_atpx_hybrid(void) {
+bool amdgpu_is_atpx_hybrid(void)
+{
 	return amdgpu_atpx_priv.atpx.is_hybrid;
 }
 
-bool amdgpu_atpx_dgpu_req_power_for_displays(void) {
-	return amdgpu_atpx_priv.atpx.dgpu_req_power_for_displays;
-}
-
-#if defined(CONFIG_ACPI)
-void *amdgpu_atpx_get_dhandle(void) {
-	return amdgpu_atpx_priv.dhandle;
-}
-#endif
-
 /**
  * amdgpu_atpx_call - call an ATPX method
  *
@@ -134,7 +127,7 @@ static union acpi_object *amdgpu_atpx_call(acpi_handle handle, int function,
 
 	/* Fail only if calling the method fails and ATPX is supported */
 	if (ACPI_FAILURE(status) && status != AE_NOT_FOUND) {
-		printk("failed to evaluate ATPX got %s\n",
+		pr_err("failed to evaluate ATPX got %s\n",
 		       acpi_format_exception(status));
 		kfree(buffer.pointer);
 		return NULL;
@@ -190,7 +183,7 @@ static int amdgpu_atpx_validate(struct amdgpu_atpx *atpx)
 
 		size = *(u16 *) info->buffer.pointer;
 		if (size < 10) {
-			printk("ATPX buffer is too small: %zu\n", size);
+			pr_err("ATPX buffer is too small: %zu\n", size);
 			kfree(info);
 			return -EINVAL;
 		}
@@ -223,11 +216,11 @@ static int amdgpu_atpx_validate(struct amdgpu_atpx *atpx)
 	atpx->is_hybrid = false;
 	if (valid_bits & ATPX_MS_HYBRID_GFX_SUPPORTED) {
 		if (amdgpu_atpx_priv.quirks & AMDGPU_PX_QUIRK_FORCE_ATPX) {
-			printk("ATPX Hybrid Graphics, forcing to ATPX\n");
+			pr_warn("ATPX Hybrid Graphics, forcing to ATPX\n");
 			atpx->functions.power_cntl = true;
 			atpx->is_hybrid = false;
 		} else {
-			printk("ATPX Hybrid Graphics\n");
+			pr_notice("ATPX Hybrid Graphics\n");
 			/*
 			 * Disable legacy PM methods only when pcie port PM is usable,
 			 * otherwise the device might fail to power off or power on.
@@ -269,7 +262,7 @@ static int amdgpu_atpx_verify_interface(struct amdgpu_atpx *atpx)
 
 	size = *(u16 *) info->buffer.pointer;
 	if (size < 8) {
-		printk("ATPX buffer is too small: %zu\n", size);
+		pr_err("ATPX buffer is too small: %zu\n", size);
 		err = -EINVAL;
 		goto out;
 	}
@@ -278,8 +271,8 @@ static int amdgpu_atpx_verify_interface(struct amdgpu_atpx *atpx)
 	memcpy(&output, info->buffer.pointer, size);
 
 	/* TODO: check version? */
-	printk("ATPX version %u, functions 0x%08x\n",
-	       output.version, output.function_bits);
+	pr_notice("ATPX version %u, functions 0x%08x\n",
+		  output.version, output.function_bits);
 
 	amdgpu_atpx_parse_functions(&atpx->functions, output.function_bits);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
index edc6377ec5ff..199693369c7c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
@@ -39,7 +39,7 @@ static int amdgpu_benchmark_do_move(struct amdgpu_device *adev, unsigned size,
 	for (i = 0; i < n; i++) {
 		struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
 		r = amdgpu_copy_buffer(ring, saddr, daddr, size, NULL, &fence,
-				       false, false, false);
+				       false, false, 0);
 		if (r)
 			goto exit_do_move;
 		r = dma_fence_wait(fence, false);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c
index 30c28a69e847..35d04e69aec0 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c
@@ -29,6 +29,7 @@
 #include "amdgpu.h"
 #include "atom.h"
 
+#include <linux/device.h>
 #include <linux/pci.h>
 #include <linux/slab.h>
 #include <linux/acpi.h>
@@ -46,52 +47,63 @@
 /* Check if current bios is an ATOM BIOS.
  * Return true if it is ATOM BIOS. Otherwise, return false.
  */
-static bool check_atom_bios(uint8_t *bios, size_t size)
+static bool check_atom_bios(struct amdgpu_device *adev, size_t size)
 {
 	uint16_t tmp, bios_header_start;
+	uint8_t *bios = adev->bios;
 
 	if (!bios || size < 0x49) {
-		DRM_INFO("vbios mem is null or mem size is wrong\n");
+		dev_dbg(adev->dev, "VBIOS mem is null or mem size is wrong\n");
 		return false;
 	}
 
 	if (!AMD_IS_VALID_VBIOS(bios)) {
-		DRM_INFO("BIOS signature incorrect %x %x\n", bios[0], bios[1]);
+		dev_dbg(adev->dev, "VBIOS signature incorrect %x %x\n", bios[0],
+			bios[1]);
 		return false;
 	}
 
 	bios_header_start = bios[0x48] | (bios[0x49] << 8);
 	if (!bios_header_start) {
-		DRM_INFO("Can't locate bios header\n");
+		dev_dbg(adev->dev, "Can't locate VBIOS header\n");
 		return false;
 	}
 
 	tmp = bios_header_start + 4;
 	if (size < tmp) {
-		DRM_INFO("BIOS header is broken\n");
+		dev_dbg(adev->dev, "VBIOS header is broken\n");
 		return false;
 	}
 
 	if (!memcmp(bios + tmp, "ATOM", 4) ||
 	    !memcmp(bios + tmp, "MOTA", 4)) {
-		DRM_DEBUG("ATOMBIOS detected\n");
+		dev_dbg(adev->dev, "ATOMBIOS detected\n");
 		return true;
 	}
 
 	return false;
 }
 
+void amdgpu_bios_release(struct amdgpu_device *adev)
+{
+	kfree(adev->bios);
+	adev->bios = NULL;
+	adev->bios_size = 0;
+}
+
 /* If you boot an IGP board with a discrete card as the primary,
  * the IGP rom is not accessible via the rom bar as the IGP rom is
  * part of the system bios.  On boot, the system bios puts a
  * copy of the igp rom at the start of vram if a discrete card is
  * present.
+ * For SR-IOV, if dynamic critical region is not enabled,
+ * the vbios image is also put at the start of VRAM in the VF.
  */
-static bool igp_read_bios_from_vram(struct amdgpu_device *adev)
+static bool amdgpu_read_bios_from_vram(struct amdgpu_device *adev)
 {
-	uint8_t __iomem *bios;
+	uint8_t __iomem *bios = NULL;
 	resource_size_t vram_base;
-	resource_size_t size = 256 * 1024; /* ??? */
+	u32 size = 256U * 1024U; /* ??? */
 
 	if (!(adev->flags & AMD_IS_APU))
 		if (amdgpu_device_need_post(adev))
@@ -103,22 +115,36 @@ static bool igp_read_bios_from_vram(struct amdgpu_device *adev)
 
 	adev->bios = NULL;
 	vram_base = pci_resource_start(adev->pdev, 0);
-	bios = ioremap_wc(vram_base, size);
-	if (!bios) {
-		return false;
-	}
 
 	adev->bios = kmalloc(size, GFP_KERNEL);
-	if (!adev->bios) {
-		iounmap(bios);
+	if (!adev->bios)
 		return false;
+
+	/* For SRIOV with dynamic critical region is enabled,
+	 * the vbios image is put at a dynamic offset of VRAM in the VF.
+	 * If dynamic critical region is disabled, follow the existing logic as on baremetal.
+	 */
+	if (amdgpu_sriov_vf(adev) && adev->virt.is_dynamic_crit_regn_enabled) {
+		if (amdgpu_virt_get_dynamic_data_info(adev,
+				AMD_SRIOV_MSG_VBIOS_IMG_TABLE_ID, adev->bios, &size)) {
+			amdgpu_bios_release(adev);
+			return false;
+		}
+	} else {
+		bios = ioremap_wc(vram_base, size);
+		if (!bios) {
+			amdgpu_bios_release(adev);
+			return false;
+		}
+
+		memcpy_fromio(adev->bios, bios, size);
+		iounmap(bios);
 	}
+
 	adev->bios_size = size;
-	memcpy_fromio(adev->bios, bios, size);
-	iounmap(bios);
 
-	if (!check_atom_bios(adev->bios, size)) {
-		kfree(adev->bios);
+	if (!check_atom_bios(adev, size)) {
+		amdgpu_bios_release(adev);
 		return false;
 	}
 
@@ -133,9 +159,8 @@ bool amdgpu_read_bios(struct amdgpu_device *adev)
 	adev->bios = NULL;
 	/* XXX: some cards may return 0 for rom size? ddx has a workaround */
 	bios = pci_map_rom(adev->pdev, &size);
-	if (!bios) {
+	if (!bios)
 		return false;
-	}
 
 	adev->bios = kzalloc(size, GFP_KERNEL);
 	if (adev->bios == NULL) {
@@ -146,8 +171,8 @@ bool amdgpu_read_bios(struct amdgpu_device *adev)
 	memcpy_fromio(adev->bios, bios, size);
 	pci_unmap_rom(adev->pdev, bios);
 
-	if (!check_atom_bios(adev->bios, size)) {
-		kfree(adev->bios);
+	if (!check_atom_bios(adev, size)) {
+		amdgpu_bios_release(adev);
 		return false;
 	}
 
@@ -168,9 +193,9 @@ static bool amdgpu_read_bios_from_rom(struct amdgpu_device *adev)
 	header[AMD_VBIOS_SIGNATURE_END] = 0;
 
 	if ((!AMD_IS_VALID_VBIOS(header)) ||
-	    0 != memcmp((char *)&header[AMD_VBIOS_SIGNATURE_OFFSET],
-			AMD_VBIOS_SIGNATURE,
-			strlen(AMD_VBIOS_SIGNATURE)))
+		memcmp((char *)&header[AMD_VBIOS_SIGNATURE_OFFSET],
+		       AMD_VBIOS_SIGNATURE,
+		       strlen(AMD_VBIOS_SIGNATURE)) != 0)
 		return false;
 
 	/* valid vbios, go on */
@@ -186,8 +211,8 @@ static bool amdgpu_read_bios_from_rom(struct amdgpu_device *adev)
 	/* read complete BIOS */
 	amdgpu_asic_read_bios_from_rom(adev, adev->bios, len);
 
-	if (!check_atom_bios(adev->bios, len)) {
-		kfree(adev->bios);
+	if (!check_atom_bios(adev, len)) {
+		amdgpu_bios_release(adev);
 		return false;
 	}
 
@@ -216,14 +241,15 @@ static bool amdgpu_read_platform_bios(struct amdgpu_device *adev)
 	memcpy_fromio(adev->bios, bios, romlen);
 	iounmap(bios);
 
-	if (!check_atom_bios(adev->bios, romlen))
+	if (!check_atom_bios(adev, romlen))
 		goto free_bios;
 
 	adev->bios_size = romlen;
 
 	return true;
 free_bios:
-	kfree(adev->bios);
+	amdgpu_bios_release(adev);
+
 	return false;
 }
 
@@ -264,7 +290,7 @@ static int amdgpu_atrm_call(acpi_handle atrm_handle, uint8_t *bios,
 
 	status = acpi_evaluate_object(atrm_handle, NULL, &atrm_arg, &buffer);
 	if (ACPI_FAILURE(status)) {
-		printk("failed to evaluate ATRM got %s\n", acpi_format_exception(status));
+		DRM_ERROR("failed to evaluate ATRM got %s\n", acpi_format_exception(status));
 		return -ENODEV;
 	}
 
@@ -285,11 +311,15 @@ static bool amdgpu_atrm_get_bios(struct amdgpu_device *adev)
 	acpi_status status;
 	bool found = false;
 
-	/* ATRM is for the discrete card only */
-	if (adev->flags & AMD_IS_APU)
+	/* ATRM is for on-platform devices only */
+	if (dev_is_removable(&adev->pdev->dev))
 		return false;
 
-	while ((pdev = pci_get_class(PCI_CLASS_DISPLAY_VGA << 8, pdev)) != NULL) {
+	while ((pdev = pci_get_base_class(PCI_BASE_CLASS_DISPLAY, pdev))) {
+		if ((pdev->class != PCI_CLASS_DISPLAY_VGA << 8) &&
+		    (pdev->class != PCI_CLASS_DISPLAY_OTHER << 8))
+			continue;
+
 		dhandle = ACPI_HANDLE(&pdev->dev);
 		if (!dhandle)
 			continue;
@@ -301,20 +331,6 @@ static bool amdgpu_atrm_get_bios(struct amdgpu_device *adev)
 		}
 	}
 
-	if (!found) {
-		while ((pdev = pci_get_class(PCI_CLASS_DISPLAY_OTHER << 8, pdev)) != NULL) {
-			dhandle = ACPI_HANDLE(&pdev->dev);
-			if (!dhandle)
-				continue;
-
-			status = acpi_get_handle(dhandle, "ATRM", &atrm_handle);
-			if (ACPI_SUCCESS(status)) {
-				found = true;
-				break;
-			}
-		}
-	}
-
 	if (!found)
 		return false;
 	pci_dev_put(pdev);
@@ -334,8 +350,8 @@ static bool amdgpu_atrm_get_bios(struct amdgpu_device *adev)
 			break;
 	}
 
-	if (!check_atom_bios(adev->bios, size)) {
-		kfree(adev->bios);
+	if (!check_atom_bios(adev, size)) {
+		amdgpu_bios_release(adev);
 		return false;
 	}
 	adev->bios_size = size;
@@ -350,11 +366,8 @@ static inline bool amdgpu_atrm_get_bios(struct amdgpu_device *adev)
 
 static bool amdgpu_read_disabled_bios(struct amdgpu_device *adev)
 {
-	if (adev->flags & AMD_IS_APU)
-		return igp_read_bios_from_vram(adev);
-	else
-		return (!adev->asic_funcs || !adev->asic_funcs->read_disabled_bios) ?
-			false : amdgpu_asic_read_disabled_bios(adev);
+	return (!adev->asic_funcs || !adev->asic_funcs->read_disabled_bios) ?
+		false : amdgpu_asic_read_disabled_bios(adev);
 }
 
 #ifdef CONFIG_ACPI
@@ -363,7 +376,7 @@ static bool amdgpu_acpi_vfct_bios(struct amdgpu_device *adev)
 	struct acpi_table_header *hdr;
 	acpi_size tbl_size;
 	UEFI_ACPI_VFCT *vfct;
-	unsigned offset;
+	unsigned int offset;
 
 	if (!ACPI_SUCCESS(acpi_get_table("VFCT", 1, &hdr)))
 		return false;
@@ -402,8 +415,8 @@ static bool amdgpu_acpi_vfct_bios(struct amdgpu_device *adev)
 					     vhdr->ImageLength,
 					     GFP_KERNEL);
 
-			if (!check_atom_bios(adev->bios, vhdr->ImageLength)) {
-				kfree(adev->bios);
+			if (!check_atom_bios(adev, vhdr->ImageLength)) {
+				amdgpu_bios_release(adev);
 				return false;
 			}
 			adev->bios_size = vhdr->ImageLength;
@@ -421,7 +434,43 @@ static inline bool amdgpu_acpi_vfct_bios(struct amdgpu_device *adev)
 }
 #endif
 
-bool amdgpu_get_bios(struct amdgpu_device *adev)
+static bool amdgpu_get_bios_apu(struct amdgpu_device *adev)
+{
+	if (amdgpu_acpi_vfct_bios(adev)) {
+		dev_info(adev->dev, "Fetched VBIOS from VFCT\n");
+		goto success;
+	}
+
+	if (amdgpu_read_bios_from_vram(adev)) {
+		dev_info(adev->dev, "Fetched VBIOS from VRAM BAR\n");
+		goto success;
+	}
+
+	if (amdgpu_read_bios(adev)) {
+		dev_info(adev->dev, "Fetched VBIOS from ROM BAR\n");
+		goto success;
+	}
+
+	if (amdgpu_read_platform_bios(adev)) {
+		dev_info(adev->dev, "Fetched VBIOS from platform\n");
+		goto success;
+	}
+
+	dev_err(adev->dev, "Unable to locate a BIOS ROM\n");
+	return false;
+
+success:
+	return true;
+}
+
+static bool amdgpu_prefer_rom_resource(struct amdgpu_device *adev)
+{
+	struct resource *res = &adev->pdev->resource[PCI_ROM_RESOURCE];
+
+	return (res->flags & IORESOURCE_ROM_SHADOW);
+}
+
+static bool amdgpu_get_bios_dgpu(struct amdgpu_device *adev)
 {
 	if (amdgpu_atrm_get_bios(adev)) {
 		dev_info(adev->dev, "Fetched VBIOS from ATRM\n");
@@ -433,14 +482,33 @@ bool amdgpu_get_bios(struct amdgpu_device *adev)
 		goto success;
 	}
 
-	if (igp_read_bios_from_vram(adev)) {
+	/* this is required for SR-IOV */
+	if (amdgpu_read_bios_from_vram(adev)) {
 		dev_info(adev->dev, "Fetched VBIOS from VRAM BAR\n");
 		goto success;
 	}
 
-	if (amdgpu_read_bios(adev)) {
-		dev_info(adev->dev, "Fetched VBIOS from ROM BAR\n");
-		goto success;
+	if (amdgpu_prefer_rom_resource(adev)) {
+		if (amdgpu_read_bios(adev)) {
+			dev_info(adev->dev, "Fetched VBIOS from ROM BAR\n");
+			goto success;
+		}
+
+		if (amdgpu_read_platform_bios(adev)) {
+			dev_info(adev->dev, "Fetched VBIOS from platform\n");
+			goto success;
+		}
+
+	} else {
+		if (amdgpu_read_platform_bios(adev)) {
+			dev_info(adev->dev, "Fetched VBIOS from platform\n");
+			goto success;
+		}
+
+		if (amdgpu_read_bios(adev)) {
+			dev_info(adev->dev, "Fetched VBIOS from ROM BAR\n");
+			goto success;
+		}
 	}
 
 	if (amdgpu_read_bios_from_rom(adev)) {
@@ -453,19 +521,28 @@ bool amdgpu_get_bios(struct amdgpu_device *adev)
 		goto success;
 	}
 
-	if (amdgpu_read_platform_bios(adev)) {
-		dev_info(adev->dev, "Fetched VBIOS from platform\n");
-		goto success;
-	}
-
 	dev_err(adev->dev, "Unable to locate a BIOS ROM\n");
 	return false;
 
 success:
-	adev->is_atom_fw = (adev->asic_type >= CHIP_VEGA10) ? true : false;
 	return true;
 }
 
+bool amdgpu_get_bios(struct amdgpu_device *adev)
+{
+	bool found;
+
+	if (adev->flags & AMD_IS_APU)
+		found = amdgpu_get_bios_apu(adev);
+	else
+		found = amdgpu_get_bios_dgpu(adev);
+
+	if (found)
+		adev->is_atom_fw = adev->asic_type >= CHIP_VEGA10;
+
+	return found;
+}
+
 /* helper function for soc15 and onwards to read bios from rom */
 bool amdgpu_soc15_read_bios_from_rom(struct amdgpu_device *adev,
 				     u8 *bios, u32 length_bytes)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c
index 252a876b0725..66fb37b64388 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c
@@ -28,6 +28,7 @@
  *    Christian König <deathsimple@vodafone.de>
  */
 
+#include <linux/sort.h>
 #include <linux/uaccess.h>
 
 #include "amdgpu.h"
@@ -50,13 +51,20 @@ static void amdgpu_bo_list_free(struct kref *ref)
 						   refcount);
 	struct amdgpu_bo_list_entry *e;
 
-	amdgpu_bo_list_for_each_entry(e, list) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
+	amdgpu_bo_list_for_each_entry(e, list)
+		amdgpu_bo_unref(&e->bo);
+	call_rcu(&list->rhead, amdgpu_bo_list_free_rcu);
+}
 
-		amdgpu_bo_unref(&bo);
-	}
+static int amdgpu_bo_list_entry_cmp(const void *_a, const void *_b)
+{
+	const struct amdgpu_bo_list_entry *a = _a, *b = _b;
 
-	call_rcu(&list->rhead, amdgpu_bo_list_free_rcu);
+	if (a->priority > b->priority)
+		return 1;
+	if (a->priority < b->priority)
+		return -1;
+	return 0;
 }
 
 int amdgpu_bo_list_create(struct amdgpu_device *adev, struct drm_file *filp,
@@ -67,27 +75,17 @@ int amdgpu_bo_list_create(struct amdgpu_device *adev, struct drm_file *filp,
 	struct amdgpu_bo_list_entry *array;
 	struct amdgpu_bo_list *list;
 	uint64_t total_size = 0;
-	size_t size;
 	unsigned i;
 	int r;
 
-	if (num_entries > (SIZE_MAX - sizeof(struct amdgpu_bo_list))
-				/ sizeof(struct amdgpu_bo_list_entry))
-		return -EINVAL;
-
-	size = sizeof(struct amdgpu_bo_list);
-	size += num_entries * sizeof(struct amdgpu_bo_list_entry);
-	list = kvmalloc(size, GFP_KERNEL);
+	list = kvzalloc(struct_size(list, entries, num_entries), GFP_KERNEL);
 	if (!list)
 		return -ENOMEM;
 
 	kref_init(&list->refcount);
-	list->gds_obj = NULL;
-	list->gws_obj = NULL;
-	list->oa_obj = NULL;
 
-	array = amdgpu_bo_list_array_entry(list, 0);
-	memset(array, 0, num_entries * sizeof(struct amdgpu_bo_list_entry));
+	list->num_entries = num_entries;
+	array = list->entries;
 
 	for (i = 0; i < num_entries; ++i) {
 		struct amdgpu_bo_list_entry *entry;
@@ -118,7 +116,7 @@ int amdgpu_bo_list_create(struct amdgpu_device *adev, struct drm_file *filp,
 
 		entry->priority = min(info[i].bo_priority,
 				      AMDGPU_BO_LIST_MAX_PRIORITY);
-		entry->tv.bo = &bo->tbo;
+		entry->bo = bo;
 
 		if (bo->preferred_domains == AMDGPU_GEM_DOMAIN_GDS)
 			list->gds_obj = bo;
@@ -132,7 +130,8 @@ int amdgpu_bo_list_create(struct amdgpu_device *adev, struct drm_file *filp,
 	}
 
 	list->first_userptr = first_userptr;
-	list->num_entries = num_entries;
+	sort(array, last_entry, sizeof(struct amdgpu_bo_list_entry),
+	     amdgpu_bo_list_entry_cmp, NULL);
 
 	trace_amdgpu_cs_bo_status(list->num_entries, total_size);
 
@@ -141,16 +140,10 @@ int amdgpu_bo_list_create(struct amdgpu_device *adev, struct drm_file *filp,
 	return 0;
 
 error_free:
-	for (i = 0; i < last_entry; ++i) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(array[i].tv.bo);
-
-		amdgpu_bo_unref(&bo);
-	}
-	for (i = first_userptr; i < num_entries; ++i) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(array[i].tv.bo);
-
-		amdgpu_bo_unref(&bo);
-	}
+	for (i = 0; i < last_entry; ++i)
+		amdgpu_bo_unref(&array[i].bo);
+	for (i = first_userptr; i < num_entries; ++i)
+		amdgpu_bo_unref(&array[i].bo);
 	kvfree(list);
 	return r;
 
@@ -179,44 +172,10 @@ int amdgpu_bo_list_get(struct amdgpu_fpriv *fpriv, int id,
 	}
 
 	rcu_read_unlock();
+	*result = NULL;
 	return -ENOENT;
 }
 
-void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list,
-			     struct list_head *validated)
-{
-	/* This is based on the bucket sort with O(n) time complexity.
-	 * An item with priority "i" is added to bucket[i]. The lists are then
-	 * concatenated in descending order.
-	 */
-	struct list_head bucket[AMDGPU_BO_LIST_NUM_BUCKETS];
-	struct amdgpu_bo_list_entry *e;
-	unsigned i;
-
-	for (i = 0; i < AMDGPU_BO_LIST_NUM_BUCKETS; i++)
-		INIT_LIST_HEAD(&bucket[i]);
-
-	/* Since buffers which appear sooner in the relocation list are
-	 * likely to be used more often than buffers which appear later
-	 * in the list, the sort mustn't change the ordering of buffers
-	 * with the same priority, i.e. it must be stable.
-	 */
-	amdgpu_bo_list_for_each_entry(e, list) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
-		unsigned priority = e->priority;
-
-		if (!bo->parent)
-			list_add_tail(&e->tv.head, &bucket[priority]);
-
-		e->user_pages = NULL;
-		e->range = NULL;
-	}
-
-	/* Connect the sorted buckets in the output list. */
-	for (i = 0; i < AMDGPU_BO_LIST_NUM_BUCKETS; i++)
-		list_splice(&bucket[i], validated);
-}
-
 void amdgpu_bo_list_put(struct amdgpu_bo_list *list)
 {
 	kref_put(&list->refcount, amdgpu_bo_list_free);
@@ -225,43 +184,36 @@ void amdgpu_bo_list_put(struct amdgpu_bo_list *list)
 int amdgpu_bo_create_list_entry_array(struct drm_amdgpu_bo_list_in *in,
 				      struct drm_amdgpu_bo_list_entry **info_param)
 {
-	const void __user *uptr = u64_to_user_ptr(in->bo_info_ptr);
 	const uint32_t info_size = sizeof(struct drm_amdgpu_bo_list_entry);
+	const void __user *uptr = u64_to_user_ptr(in->bo_info_ptr);
+	const uint32_t bo_info_size = in->bo_info_size;
+	const uint32_t bo_number = in->bo_number;
 	struct drm_amdgpu_bo_list_entry *info;
-	int r;
-
-	info = kvmalloc_array(in->bo_number, info_size, GFP_KERNEL);
-	if (!info)
-		return -ENOMEM;
 
 	/* copy the handle array from userspace to a kernel buffer */
-	r = -EFAULT;
-	if (likely(info_size == in->bo_info_size)) {
-		unsigned long bytes = in->bo_number *
-			in->bo_info_size;
-
-		if (copy_from_user(info, uptr, bytes))
-			goto error_free;
-
+	if (likely(info_size == bo_info_size)) {
+		info = vmemdup_array_user(uptr, bo_number, info_size);
+		if (IS_ERR(info))
+			return PTR_ERR(info);
 	} else {
-		unsigned long bytes = min(in->bo_info_size, info_size);
+		const uint32_t bytes = min(bo_info_size, info_size);
 		unsigned i;
 
-		memset(info, 0, in->bo_number * info_size);
-		for (i = 0; i < in->bo_number; ++i) {
-			if (copy_from_user(&info[i], uptr, bytes))
-				goto error_free;
+		info = kvmalloc_array(bo_number, info_size, GFP_KERNEL);
+		if (!info)
+			return -ENOMEM;
 
-			uptr += in->bo_info_size;
+		memset(info, 0, bo_number * info_size);
+		for (i = 0; i < bo_number; ++i, uptr += bo_info_size) {
+			if (copy_from_user(&info[i], uptr, bytes)) {
+				kvfree(info);
+				return -EFAULT;
+			}
 		}
 	}
 
 	*info_param = info;
 	return 0;
-
-error_free:
-	kvfree(info);
-	return r;
 }
 
 int amdgpu_bo_list_ioctl(struct drm_device *dev, void *data,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h
index ededdc01ca28..2b5e7c46a39d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h
@@ -23,7 +23,6 @@
 #ifndef __AMDGPU_BO_LIST_H__
 #define __AMDGPU_BO_LIST_H__
 
-#include <drm/ttm/ttm_execbuf_util.h>
 #include <drm/amdgpu_drm.h>
 
 struct hmm_range;
@@ -36,11 +35,10 @@ struct amdgpu_bo_va;
 struct amdgpu_fpriv;
 
 struct amdgpu_bo_list_entry {
-	struct ttm_validate_buffer	tv;
+	struct amdgpu_bo		*bo;
 	struct amdgpu_bo_va		*bo_va;
 	uint32_t			priority;
-	struct page			**user_pages;
-	struct hmm_range		*range;
+	struct amdgpu_hmm_range		*range;
 	bool				user_invalidated;
 };
 
@@ -56,12 +54,12 @@ struct amdgpu_bo_list {
 	/* Protect access during command submission.
 	 */
 	struct mutex bo_list_mutex;
+
+	struct amdgpu_bo_list_entry entries[] __counted_by(num_entries);
 };
 
 int amdgpu_bo_list_get(struct amdgpu_fpriv *fpriv, int id,
 		       struct amdgpu_bo_list **result);
-void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list,
-			     struct list_head *validated);
 void amdgpu_bo_list_put(struct amdgpu_bo_list *list);
 int amdgpu_bo_create_list_entry_array(struct drm_amdgpu_bo_list_in *in,
 				      struct drm_amdgpu_bo_list_entry **info_param);
@@ -72,22 +70,14 @@ int amdgpu_bo_list_create(struct amdgpu_device *adev,
 				 size_t num_entries,
 				 struct amdgpu_bo_list **list);
 
-static inline struct amdgpu_bo_list_entry *
-amdgpu_bo_list_array_entry(struct amdgpu_bo_list *list, unsigned index)
-{
-	struct amdgpu_bo_list_entry *array = (void *)&list[1];
-
-	return &array[index];
-}
-
 #define amdgpu_bo_list_for_each_entry(e, list) \
-	for (e = amdgpu_bo_list_array_entry(list, 0); \
-	     e != amdgpu_bo_list_array_entry(list, (list)->num_entries); \
+	for (e = list->entries; \
+	     e != &list->entries[list->num_entries]; \
 	     ++e)
 
 #define amdgpu_bo_list_for_each_userptr_entry(e, list) \
-	for (e = amdgpu_bo_list_array_entry(list, (list)->first_userptr); \
-	     e != amdgpu_bo_list_array_entry(list, (list)->num_entries); \
+	for (e = &list->entries[list->first_userptr]; \
+	     e != &list->entries[list->num_entries]; \
 	     ++e)
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c
index 456e385333b6..004a6a9d6b9f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c
@@ -41,13 +41,13 @@ struct amdgpu_cgs_device {
 		((struct amdgpu_cgs_device *)cgs_device)->adev
 
 
-static uint32_t amdgpu_cgs_read_register(struct cgs_device *cgs_device, unsigned offset)
+static uint32_t amdgpu_cgs_read_register(struct cgs_device *cgs_device, unsigned int offset)
 {
 	CGS_FUNC_ADEV;
 	return RREG32(offset);
 }
 
-static void amdgpu_cgs_write_register(struct cgs_device *cgs_device, unsigned offset,
+static void amdgpu_cgs_write_register(struct cgs_device *cgs_device, unsigned int offset,
 				      uint32_t value)
 {
 	CGS_FUNC_ADEV;
@@ -56,7 +56,7 @@ static void amdgpu_cgs_write_register(struct cgs_device *cgs_device, unsigned of
 
 static uint32_t amdgpu_cgs_read_ind_register(struct cgs_device *cgs_device,
 					     enum cgs_ind_reg space,
-					     unsigned index)
+					     unsigned int index)
 {
 	CGS_FUNC_ADEV;
 	switch (space) {
@@ -84,7 +84,7 @@ static uint32_t amdgpu_cgs_read_ind_register(struct cgs_device *cgs_device,
 
 static void amdgpu_cgs_write_ind_register(struct cgs_device *cgs_device,
 					  enum cgs_ind_reg space,
-					  unsigned index, uint32_t value)
+					  unsigned int index, uint32_t value)
 {
 	CGS_FUNC_ADEV;
 	switch (space) {
@@ -163,38 +163,38 @@ static uint16_t amdgpu_get_firmware_version(struct cgs_device *cgs_device,
 	uint16_t fw_version = 0;
 
 	switch (type) {
-		case CGS_UCODE_ID_SDMA0:
-			fw_version = adev->sdma.instance[0].fw_version;
-			break;
-		case CGS_UCODE_ID_SDMA1:
-			fw_version = adev->sdma.instance[1].fw_version;
-			break;
-		case CGS_UCODE_ID_CP_CE:
-			fw_version = adev->gfx.ce_fw_version;
-			break;
-		case CGS_UCODE_ID_CP_PFP:
-			fw_version = adev->gfx.pfp_fw_version;
-			break;
-		case CGS_UCODE_ID_CP_ME:
-			fw_version = adev->gfx.me_fw_version;
-			break;
-		case CGS_UCODE_ID_CP_MEC:
-			fw_version = adev->gfx.mec_fw_version;
-			break;
-		case CGS_UCODE_ID_CP_MEC_JT1:
-			fw_version = adev->gfx.mec_fw_version;
-			break;
-		case CGS_UCODE_ID_CP_MEC_JT2:
-			fw_version = adev->gfx.mec_fw_version;
-			break;
-		case CGS_UCODE_ID_RLC_G:
-			fw_version = adev->gfx.rlc_fw_version;
-			break;
-		case CGS_UCODE_ID_STORAGE:
-			break;
-		default:
-			DRM_ERROR("firmware type %d do not have version\n", type);
-			break;
+	case CGS_UCODE_ID_SDMA0:
+		fw_version = adev->sdma.instance[0].fw_version;
+		break;
+	case CGS_UCODE_ID_SDMA1:
+		fw_version = adev->sdma.instance[1].fw_version;
+		break;
+	case CGS_UCODE_ID_CP_CE:
+		fw_version = adev->gfx.ce_fw_version;
+		break;
+	case CGS_UCODE_ID_CP_PFP:
+		fw_version = adev->gfx.pfp_fw_version;
+		break;
+	case CGS_UCODE_ID_CP_ME:
+		fw_version = adev->gfx.me_fw_version;
+		break;
+	case CGS_UCODE_ID_CP_MEC:
+		fw_version = adev->gfx.mec_fw_version;
+		break;
+	case CGS_UCODE_ID_CP_MEC_JT1:
+		fw_version = adev->gfx.mec_fw_version;
+		break;
+	case CGS_UCODE_ID_CP_MEC_JT2:
+		fw_version = adev->gfx.mec_fw_version;
+		break;
+	case CGS_UCODE_ID_RLC_G:
+		fw_version = adev->gfx.rlc_fw_version;
+		break;
+	case CGS_UCODE_ID_STORAGE:
+		break;
+	default:
+		DRM_ERROR("firmware type %d do not have version\n", type);
+		break;
 	}
 	return fw_version;
 }
@@ -205,7 +205,7 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device,
 {
 	CGS_FUNC_ADEV;
 
-	if ((CGS_UCODE_ID_SMU != type) && (CGS_UCODE_ID_SMU_SK != type)) {
+	if (type != CGS_UCODE_ID_SMU && type != CGS_UCODE_ID_SMU_SK) {
 		uint64_t gpu_addr;
 		uint32_t data_size;
 		const struct gfx_firmware_header_v1_0 *header;
@@ -213,6 +213,9 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device,
 		struct amdgpu_firmware_info *ucode;
 
 		id = fw_type_convert(cgs_device, type);
+		if (id >= AMDGPU_UCODE_ID_MAXIMUM)
+			return -EINVAL;
+
 		ucode = &adev->firmware.ucode[id];
 		if (ucode->fw == NULL)
 			return -EINVAL;
@@ -232,7 +235,7 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device,
 		info->mc_addr = gpu_addr;
 		info->version = (uint16_t)le32_to_cpu(header->header.ucode_version);
 
-		if (CGS_UCODE_ID_CP_MEC == type)
+		if (type == CGS_UCODE_ID_CP_MEC)
 			info->image_size = le32_to_cpu(header->jt_offset) << 2;
 
 		info->fw_version = amdgpu_get_firmware_version(cgs_device, type);
@@ -249,83 +252,22 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device,
 
 		if (!adev->pm.fw) {
 			switch (adev->asic_type) {
-			case CHIP_TAHITI:
-				strcpy(fw_name, "radeon/tahiti_smc.bin");
-				break;
-			case CHIP_PITCAIRN:
-				if ((adev->pdev->revision == 0x81) &&
-				    ((adev->pdev->device == 0x6810) ||
-				    (adev->pdev->device == 0x6811))) {
-					info->is_kicker = true;
-					strcpy(fw_name, "radeon/pitcairn_k_smc.bin");
-				} else {
-					strcpy(fw_name, "radeon/pitcairn_smc.bin");
-				}
-				break;
-			case CHIP_VERDE:
-				if (((adev->pdev->device == 0x6820) &&
-					((adev->pdev->revision == 0x81) ||
-					(adev->pdev->revision == 0x83))) ||
-				    ((adev->pdev->device == 0x6821) &&
-					((adev->pdev->revision == 0x83) ||
-					(adev->pdev->revision == 0x87))) ||
-				    ((adev->pdev->revision == 0x87) &&
-					((adev->pdev->device == 0x6823) ||
-					(adev->pdev->device == 0x682b)))) {
-					info->is_kicker = true;
-					strcpy(fw_name, "radeon/verde_k_smc.bin");
-				} else {
-					strcpy(fw_name, "radeon/verde_smc.bin");
-				}
-				break;
-			case CHIP_OLAND:
-				if (((adev->pdev->revision == 0x81) &&
-					((adev->pdev->device == 0x6600) ||
-					(adev->pdev->device == 0x6604) ||
-					(adev->pdev->device == 0x6605) ||
-					(adev->pdev->device == 0x6610))) ||
-				    ((adev->pdev->revision == 0x83) &&
-					(adev->pdev->device == 0x6610))) {
-					info->is_kicker = true;
-					strcpy(fw_name, "radeon/oland_k_smc.bin");
-				} else {
-					strcpy(fw_name, "radeon/oland_smc.bin");
-				}
-				break;
-			case CHIP_HAINAN:
-				if (((adev->pdev->revision == 0x81) &&
-					(adev->pdev->device == 0x6660)) ||
-				    ((adev->pdev->revision == 0x83) &&
-					((adev->pdev->device == 0x6660) ||
-					(adev->pdev->device == 0x6663) ||
-					(adev->pdev->device == 0x6665) ||
-					 (adev->pdev->device == 0x6667)))) {
-					info->is_kicker = true;
-					strcpy(fw_name, "radeon/hainan_k_smc.bin");
-				} else if ((adev->pdev->revision == 0xc3) &&
-					 (adev->pdev->device == 0x6665)) {
-					info->is_kicker = true;
-					strcpy(fw_name, "radeon/banks_k_2_smc.bin");
-				} else {
-					strcpy(fw_name, "radeon/hainan_smc.bin");
-				}
-				break;
 			case CHIP_BONAIRE:
 				if ((adev->pdev->revision == 0x80) ||
 					(adev->pdev->revision == 0x81) ||
 					(adev->pdev->device == 0x665f)) {
 					info->is_kicker = true;
-					strcpy(fw_name, "amdgpu/bonaire_k_smc.bin");
+					strscpy(fw_name, "amdgpu/bonaire_k_smc.bin");
 				} else {
-					strcpy(fw_name, "amdgpu/bonaire_smc.bin");
+					strscpy(fw_name, "amdgpu/bonaire_smc.bin");
 				}
 				break;
 			case CHIP_HAWAII:
 				if (adev->pdev->revision == 0x80) {
 					info->is_kicker = true;
-					strcpy(fw_name, "amdgpu/hawaii_k_smc.bin");
+					strscpy(fw_name, "amdgpu/hawaii_k_smc.bin");
 				} else {
-					strcpy(fw_name, "amdgpu/hawaii_smc.bin");
+					strscpy(fw_name, "amdgpu/hawaii_smc.bin");
 				}
 				break;
 			case CHIP_TOPAZ:
@@ -335,83 +277,85 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device,
 				    ((adev->pdev->device == 0x6900) && (adev->pdev->revision == 0xD1)) ||
 				    ((adev->pdev->device == 0x6900) && (adev->pdev->revision == 0xD3))) {
 					info->is_kicker = true;
-					strcpy(fw_name, "amdgpu/topaz_k_smc.bin");
+					strscpy(fw_name, "amdgpu/topaz_k_smc.bin");
 				} else
-					strcpy(fw_name, "amdgpu/topaz_smc.bin");
+					strscpy(fw_name, "amdgpu/topaz_smc.bin");
 				break;
 			case CHIP_TONGA:
 				if (((adev->pdev->device == 0x6939) && (adev->pdev->revision == 0xf1)) ||
 				    ((adev->pdev->device == 0x6938) && (adev->pdev->revision == 0xf1))) {
 					info->is_kicker = true;
-					strcpy(fw_name, "amdgpu/tonga_k_smc.bin");
+					strscpy(fw_name, "amdgpu/tonga_k_smc.bin");
 				} else
-					strcpy(fw_name, "amdgpu/tonga_smc.bin");
+					strscpy(fw_name, "amdgpu/tonga_smc.bin");
 				break;
 			case CHIP_FIJI:
-				strcpy(fw_name, "amdgpu/fiji_smc.bin");
+				strscpy(fw_name, "amdgpu/fiji_smc.bin");
 				break;
 			case CHIP_POLARIS11:
 				if (type == CGS_UCODE_ID_SMU) {
 					if (ASICID_IS_P21(adev->pdev->device, adev->pdev->revision)) {
 						info->is_kicker = true;
-						strcpy(fw_name, "amdgpu/polaris11_k_smc.bin");
+						strscpy(fw_name, "amdgpu/polaris11_k_smc.bin");
 					} else if (ASICID_IS_P31(adev->pdev->device, adev->pdev->revision)) {
 						info->is_kicker = true;
-						strcpy(fw_name, "amdgpu/polaris11_k2_smc.bin");
+						strscpy(fw_name, "amdgpu/polaris11_k2_smc.bin");
 					} else {
-						strcpy(fw_name, "amdgpu/polaris11_smc.bin");
+						strscpy(fw_name, "amdgpu/polaris11_smc.bin");
 					}
 				} else if (type == CGS_UCODE_ID_SMU_SK) {
-					strcpy(fw_name, "amdgpu/polaris11_smc_sk.bin");
+					strscpy(fw_name, "amdgpu/polaris11_smc_sk.bin");
 				}
 				break;
 			case CHIP_POLARIS10:
 				if (type == CGS_UCODE_ID_SMU) {
 					if (ASICID_IS_P20(adev->pdev->device, adev->pdev->revision)) {
 						info->is_kicker = true;
-						strcpy(fw_name, "amdgpu/polaris10_k_smc.bin");
+						strscpy(fw_name, "amdgpu/polaris10_k_smc.bin");
 					} else if (ASICID_IS_P30(adev->pdev->device, adev->pdev->revision)) {
 						info->is_kicker = true;
-						strcpy(fw_name, "amdgpu/polaris10_k2_smc.bin");
+						strscpy(fw_name, "amdgpu/polaris10_k2_smc.bin");
 					} else {
-						strcpy(fw_name, "amdgpu/polaris10_smc.bin");
+						strscpy(fw_name, "amdgpu/polaris10_smc.bin");
 					}
 				} else if (type == CGS_UCODE_ID_SMU_SK) {
-					strcpy(fw_name, "amdgpu/polaris10_smc_sk.bin");
+					strscpy(fw_name, "amdgpu/polaris10_smc_sk.bin");
 				}
 				break;
 			case CHIP_POLARIS12:
 				if (ASICID_IS_P23(adev->pdev->device, adev->pdev->revision)) {
 					info->is_kicker = true;
-					strcpy(fw_name, "amdgpu/polaris12_k_smc.bin");
+					strscpy(fw_name, "amdgpu/polaris12_k_smc.bin");
 				} else {
-					strcpy(fw_name, "amdgpu/polaris12_smc.bin");
+					strscpy(fw_name, "amdgpu/polaris12_smc.bin");
 				}
 				break;
 			case CHIP_VEGAM:
-				strcpy(fw_name, "amdgpu/vegam_smc.bin");
+				strscpy(fw_name, "amdgpu/vegam_smc.bin");
 				break;
 			case CHIP_VEGA10:
 				if ((adev->pdev->device == 0x687f) &&
 					((adev->pdev->revision == 0xc0) ||
 					(adev->pdev->revision == 0xc1) ||
 					(adev->pdev->revision == 0xc3)))
-					strcpy(fw_name, "amdgpu/vega10_acg_smc.bin");
+					strscpy(fw_name, "amdgpu/vega10_acg_smc.bin");
 				else
-					strcpy(fw_name, "amdgpu/vega10_smc.bin");
+					strscpy(fw_name, "amdgpu/vega10_smc.bin");
 				break;
 			case CHIP_VEGA12:
-				strcpy(fw_name, "amdgpu/vega12_smc.bin");
+				strscpy(fw_name, "amdgpu/vega12_smc.bin");
 				break;
 			case CHIP_VEGA20:
-				strcpy(fw_name, "amdgpu/vega20_smc.bin");
+				strscpy(fw_name, "amdgpu/vega20_smc.bin");
 				break;
 			default:
 				DRM_ERROR("SMC firmware not supported\n");
 				return -EINVAL;
 			}
 
-			err = amdgpu_ucode_request(adev, &adev->pm.fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->pm.fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "%s", fw_name);
 			if (err) {
 				DRM_ERROR("Failed to load firmware \"%s\"", fw_name);
 				amdgpu_ucode_release(&adev->pm.fw);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c
index 6be30dcb029d..9f96d568acf2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c
@@ -103,7 +103,7 @@ int amdgpu_connector_get_monitor_bpc(struct drm_connector *connector)
 	struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
 	struct amdgpu_connector_atom_dig *dig_connector;
 	int bpc = 8;
-	unsigned mode_clock, max_tmds_clock;
+	unsigned int mode_clock, max_tmds_clock;
 
 	switch (connector->connector_type) {
 	case DRM_MODE_CONNECTOR_DVII:
@@ -246,36 +246,10 @@ amdgpu_connector_find_encoder(struct drm_connector *connector,
 	return NULL;
 }
 
-struct edid *amdgpu_connector_edid(struct drm_connector *connector)
-{
-	struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
-	struct drm_property_blob *edid_blob = connector->edid_blob_ptr;
-
-	if (amdgpu_connector->edid) {
-		return amdgpu_connector->edid;
-	} else if (edid_blob) {
-		struct edid *edid = kmemdup(edid_blob->data, edid_blob->length, GFP_KERNEL);
-		if (edid)
-			amdgpu_connector->edid = edid;
-	}
-	return amdgpu_connector->edid;
-}
-
 static struct edid *
 amdgpu_connector_get_hardcoded_edid(struct amdgpu_device *adev)
 {
-	struct edid *edid;
-
-	if (adev->mode_info.bios_hardcoded_edid) {
-		edid = kmalloc(adev->mode_info.bios_hardcoded_edid_size, GFP_KERNEL);
-		if (edid) {
-			memcpy((unsigned char *)edid,
-			       (unsigned char *)adev->mode_info.bios_hardcoded_edid,
-			       adev->mode_info.bios_hardcoded_edid_size);
-			return edid;
-		}
-	}
-	return NULL;
+	return drm_edid_duplicate(drm_edid_raw(adev->mode_info.bios_hardcoded_edid));
 }
 
 static void amdgpu_connector_get_edid(struct drm_connector *connector)
@@ -424,30 +398,28 @@ static void amdgpu_connector_add_common_modes(struct drm_encoder *encoder,
 	struct drm_display_mode *mode = NULL;
 	struct drm_display_mode *native_mode = &amdgpu_encoder->native_mode;
 	int i;
-	static const struct mode_size {
+	int n;
+	struct mode_size {
+		char name[DRM_DISPLAY_MODE_LEN];
 		int w;
 		int h;
-	} common_modes[17] = {
-		{ 640,  480},
-		{ 720,  480},
-		{ 800,  600},
-		{ 848,  480},
-		{1024,  768},
-		{1152,  768},
-		{1280,  720},
-		{1280,  800},
-		{1280,  854},
-		{1280,  960},
-		{1280, 1024},
-		{1440,  900},
-		{1400, 1050},
-		{1680, 1050},
-		{1600, 1200},
-		{1920, 1080},
-		{1920, 1200}
+	} common_modes[] = {
+		{  "640x480",  640,  480},
+		{  "800x600",  800,  600},
+		{ "1024x768", 1024,  768},
+		{ "1280x720", 1280,  720},
+		{ "1280x800", 1280,  800},
+		{"1280x1024", 1280, 1024},
+		{ "1440x900", 1440,  900},
+		{"1680x1050", 1680, 1050},
+		{"1600x1200", 1600, 1200},
+		{"1920x1080", 1920, 1080},
+		{"1920x1200", 1920, 1200}
 	};
 
-	for (i = 0; i < 17; i++) {
+	n = ARRAY_SIZE(common_modes);
+
+	for (i = 0; i < n; i++) {
 		if (amdgpu_encoder->devices & (ATOM_DEVICE_TV_SUPPORT)) {
 			if (common_modes[i].w > 1024 ||
 			    common_modes[i].h > 768)
@@ -460,10 +432,12 @@ static void amdgpu_connector_add_common_modes(struct drm_encoder *encoder,
 			     common_modes[i].h == native_mode->vdisplay))
 				continue;
 		}
-		if (common_modes[i].w < 320 || common_modes[i].h < 200)
-			continue;
 
 		mode = drm_cvt_mode(dev, common_modes[i].w, common_modes[i].h, 60, false, false, false);
+		if (!mode)
+			return;
+		strscpy(mode->name, common_modes[i].name, DRM_DISPLAY_MODE_LEN);
+
 		drm_mode_probed_add(connector, mode);
 	}
 }
@@ -588,16 +562,26 @@ static int amdgpu_connector_set_property(struct drm_connector *connector,
 			amdgpu_encoder = to_amdgpu_encoder(connector->encoder);
 		} else {
 			const struct drm_connector_helper_funcs *connector_funcs = connector->helper_private;
+
 			amdgpu_encoder = to_amdgpu_encoder(connector_funcs->best_encoder(connector));
 		}
 
 		switch (val) {
 		default:
-		case DRM_MODE_SCALE_NONE: rmx_type = RMX_OFF; break;
-		case DRM_MODE_SCALE_CENTER: rmx_type = RMX_CENTER; break;
-		case DRM_MODE_SCALE_ASPECT: rmx_type = RMX_ASPECT; break;
-		case DRM_MODE_SCALE_FULLSCREEN: rmx_type = RMX_FULL; break;
+		case DRM_MODE_SCALE_NONE:
+			rmx_type = RMX_OFF;
+			break;
+		case DRM_MODE_SCALE_CENTER:
+			rmx_type = RMX_CENTER;
+			break;
+		case DRM_MODE_SCALE_ASPECT:
+			rmx_type = RMX_ASPECT;
+			break;
+		case DRM_MODE_SCALE_FULLSCREEN:
+			rmx_type = RMX_FULL;
+			break;
 		}
+
 		if (amdgpu_encoder->rmx_type == rmx_type)
 			return 0;
 
@@ -687,7 +671,7 @@ static int amdgpu_connector_lvds_get_modes(struct drm_connector *connector)
 }
 
 static enum drm_mode_status amdgpu_connector_lvds_mode_valid(struct drm_connector *connector,
-					     struct drm_display_mode *mode)
+					     const struct drm_display_mode *mode)
 {
 	struct drm_encoder *encoder = amdgpu_connector_best_single_encoder(connector);
 
@@ -750,10 +734,8 @@ amdgpu_connector_lvds_detect(struct drm_connector *connector, bool force)
 
 	amdgpu_connector_update_scratch_regs(connector, ret);
 
-	if (!drm_kms_helper_is_poll_worker()) {
-		pm_runtime_mark_last_busy(connector->dev->dev);
+	if (!drm_kms_helper_is_poll_worker())
 		pm_runtime_put_autosuspend(connector->dev->dev);
-	}
 
 	return ret;
 }
@@ -795,16 +777,26 @@ static int amdgpu_connector_set_lcd_property(struct drm_connector *connector,
 		amdgpu_encoder = to_amdgpu_encoder(connector->encoder);
 	else {
 		const struct drm_connector_helper_funcs *connector_funcs = connector->helper_private;
+
 		amdgpu_encoder = to_amdgpu_encoder(connector_funcs->best_encoder(connector));
 	}
 
 	switch (value) {
-	case DRM_MODE_SCALE_NONE: rmx_type = RMX_OFF; break;
-	case DRM_MODE_SCALE_CENTER: rmx_type = RMX_CENTER; break;
-	case DRM_MODE_SCALE_ASPECT: rmx_type = RMX_ASPECT; break;
+	case DRM_MODE_SCALE_NONE:
+		rmx_type = RMX_OFF;
+		break;
+	case DRM_MODE_SCALE_CENTER:
+		rmx_type = RMX_CENTER;
+		break;
+	case DRM_MODE_SCALE_ASPECT:
+		rmx_type = RMX_ASPECT;
+		break;
 	default:
-	case DRM_MODE_SCALE_FULLSCREEN: rmx_type = RMX_FULL; break;
+	case DRM_MODE_SCALE_FULLSCREEN:
+		rmx_type = RMX_FULL;
+		break;
 	}
+
 	if (amdgpu_encoder->rmx_type == rmx_type)
 		return 0;
 
@@ -842,7 +834,7 @@ static int amdgpu_connector_vga_get_modes(struct drm_connector *connector)
 }
 
 static enum drm_mode_status amdgpu_connector_vga_mode_valid(struct drm_connector *connector,
-					    struct drm_display_mode *mode)
+					    const struct drm_display_mode *mode)
 {
 	struct drm_device *dev = connector->dev;
 	struct amdgpu_device *adev = drm_to_adev(dev);
@@ -925,10 +917,8 @@ amdgpu_connector_vga_detect(struct drm_connector *connector, bool force)
 	amdgpu_connector_update_scratch_regs(connector, ret);
 
 out:
-	if (!drm_kms_helper_is_poll_worker()) {
-		pm_runtime_mark_last_busy(connector->dev->dev);
+	if (!drm_kms_helper_is_poll_worker())
 		pm_runtime_put_autosuspend(connector->dev->dev);
-	}
 
 	return ret;
 }
@@ -968,6 +958,41 @@ amdgpu_connector_check_hpd_status_unchanged(struct drm_connector *connector)
 	return false;
 }
 
+static void amdgpu_connector_shared_ddc(enum drm_connector_status *status,
+					struct drm_connector *connector,
+					struct amdgpu_connector *amdgpu_connector)
+{
+	struct drm_connector *list_connector;
+	struct drm_connector_list_iter iter;
+	struct amdgpu_connector *list_amdgpu_connector;
+	struct drm_device *dev = connector->dev;
+	struct amdgpu_device *adev = drm_to_adev(dev);
+
+	if (amdgpu_connector->shared_ddc && *status == connector_status_connected) {
+		drm_connector_list_iter_begin(dev, &iter);
+		drm_for_each_connector_iter(list_connector,
+					    &iter) {
+			if (connector == list_connector)
+				continue;
+			list_amdgpu_connector = to_amdgpu_connector(list_connector);
+			if (list_amdgpu_connector->shared_ddc &&
+			    list_amdgpu_connector->ddc_bus->rec.i2c_id ==
+			     amdgpu_connector->ddc_bus->rec.i2c_id) {
+				/* cases where both connectors are digital */
+				if (list_connector->connector_type != DRM_MODE_CONNECTOR_VGA) {
+					/* hpd is our only option in this case */
+					if (!amdgpu_display_hpd_sense(adev,
+								      amdgpu_connector->hpd.hpd)) {
+						amdgpu_connector_free_edid(connector);
+						*status = connector_status_disconnected;
+					}
+				}
+			}
+		}
+		drm_connector_list_iter_end(&iter);
+	}
+}
+
 /*
  * DVI is complicated
  * Do a DDC probe, if DDC probe passes, get the full EDID so
@@ -1054,32 +1079,7 @@ amdgpu_connector_dvi_detect(struct drm_connector *connector, bool force)
 			 * DDC line.  The latter is more complex because with DVI<->HDMI adapters
 			 * you don't really know what's connected to which port as both are digital.
 			 */
-			if (amdgpu_connector->shared_ddc && (ret == connector_status_connected)) {
-				struct drm_connector *list_connector;
-				struct drm_connector_list_iter iter;
-				struct amdgpu_connector *list_amdgpu_connector;
-
-				drm_connector_list_iter_begin(dev, &iter);
-				drm_for_each_connector_iter(list_connector,
-							    &iter) {
-					if (connector == list_connector)
-						continue;
-					list_amdgpu_connector = to_amdgpu_connector(list_connector);
-					if (list_amdgpu_connector->shared_ddc &&
-					    (list_amdgpu_connector->ddc_bus->rec.i2c_id ==
-					     amdgpu_connector->ddc_bus->rec.i2c_id)) {
-						/* cases where both connectors are digital */
-						if (list_connector->connector_type != DRM_MODE_CONNECTOR_VGA) {
-							/* hpd is our only option in this case */
-							if (!amdgpu_display_hpd_sense(adev, amdgpu_connector->hpd.hpd)) {
-								amdgpu_connector_free_edid(connector);
-								ret = connector_status_disconnected;
-							}
-						}
-					}
-				}
-				drm_connector_list_iter_end(&iter);
-			}
+			amdgpu_connector_shared_ddc(&ret, connector, amdgpu_connector);
 		}
 	}
 
@@ -1127,7 +1127,8 @@ amdgpu_connector_dvi_detect(struct drm_connector *connector, bool force)
 					/* assume digital unless load detected otherwise */
 					amdgpu_connector->use_digital = true;
 					lret = encoder_funcs->detect(encoder, connector);
-					DRM_DEBUG_KMS("load_detect %x returned: %x\n",encoder->encoder_type,lret);
+					DRM_DEBUG_KMS("load_detect %x returned: %x\n",
+						      encoder->encoder_type, lret);
 					if (lret == connector_status_connected)
 						amdgpu_connector->use_digital = false;
 				}
@@ -1141,10 +1142,8 @@ out:
 	amdgpu_connector_update_scratch_regs(connector, ret);
 
 exit:
-	if (!drm_kms_helper_is_poll_worker()) {
-		pm_runtime_mark_last_busy(connector->dev->dev);
+	if (!drm_kms_helper_is_poll_worker())
 		pm_runtime_put_autosuspend(connector->dev->dev);
-	}
 
 	return ret;
 }
@@ -1180,35 +1179,76 @@ amdgpu_connector_dvi_encoder(struct drm_connector *connector)
 static void amdgpu_connector_dvi_force(struct drm_connector *connector)
 {
 	struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
+
 	if (connector->force == DRM_FORCE_ON)
 		amdgpu_connector->use_digital = false;
 	if (connector->force == DRM_FORCE_ON_DIGITAL)
 		amdgpu_connector->use_digital = true;
 }
 
+/**
+ * amdgpu_max_hdmi_pixel_clock - Return max supported HDMI (TMDS) pixel clock
+ * @adev: pointer to amdgpu_device
+ *
+ * Return: maximum supported HDMI (TMDS) pixel clock in KHz.
+ */
+static int amdgpu_max_hdmi_pixel_clock(const struct amdgpu_device *adev)
+{
+	if (adev->asic_type >= CHIP_POLARIS10)
+		return 600000;
+	else if (adev->asic_type >= CHIP_TONGA)
+		return 300000;
+	else
+		return 297000;
+}
+
+/**
+ * amdgpu_connector_dvi_mode_valid - Validate a mode on DVI/HDMI connectors
+ * @connector: DRM connector to validate the mode on
+ * @mode: display mode to validate
+ *
+ * Validate the given display mode on DVI and HDMI connectors, including
+ * analog signals on DVI-I.
+ *
+ * Return: drm_mode_status indicating whether the mode is valid.
+ */
 static enum drm_mode_status amdgpu_connector_dvi_mode_valid(struct drm_connector *connector,
-					    struct drm_display_mode *mode)
+					    const struct drm_display_mode *mode)
 {
 	struct drm_device *dev = connector->dev;
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
+	const int max_hdmi_pixel_clock = amdgpu_max_hdmi_pixel_clock(adev);
+	const int max_dvi_single_link_pixel_clock = 165000;
+	int max_digital_pixel_clock_khz;
 
 	/* XXX check mode bandwidth */
 
-	if (amdgpu_connector->use_digital && (mode->clock > 165000)) {
-		if ((amdgpu_connector->connector_object_id == CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_I) ||
-		    (amdgpu_connector->connector_object_id == CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_D) ||
-		    (amdgpu_connector->connector_object_id == CONNECTOR_OBJECT_ID_HDMI_TYPE_B)) {
-			return MODE_OK;
-		} else if (connector->display_info.is_hdmi) {
-			/* HDMI 1.3+ supports max clock of 340 Mhz */
-			if (mode->clock > 340000)
-				return MODE_CLOCK_HIGH;
-			else
-				return MODE_OK;
-		} else {
-			return MODE_CLOCK_HIGH;
+	if (amdgpu_connector->use_digital) {
+		switch (amdgpu_connector->connector_object_id) {
+		case CONNECTOR_OBJECT_ID_HDMI_TYPE_A:
+			max_digital_pixel_clock_khz = max_hdmi_pixel_clock;
+			break;
+		case CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I:
+		case CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_D:
+			max_digital_pixel_clock_khz = max_dvi_single_link_pixel_clock;
+			break;
+		case CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_I:
+		case CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_D:
+		case CONNECTOR_OBJECT_ID_HDMI_TYPE_B:
+			max_digital_pixel_clock_khz = max_dvi_single_link_pixel_clock * 2;
+			break;
 		}
+
+		/* When the display EDID claims that it's an HDMI display,
+		 * we use the HDMI encoder mode of the display HW,
+		 * so we should verify against the max HDMI clock here.
+		 */
+		if (connector->display_info.is_hdmi)
+			max_digital_pixel_clock_khz = max_hdmi_pixel_clock;
+
+		if (mode->clock > max_digital_pixel_clock_khz)
+			return MODE_CLOCK_HIGH;
 	}
 
 	/* check against the max pixel clock */
@@ -1414,6 +1454,7 @@ amdgpu_connector_dp_detect(struct drm_connector *connector, bool force)
 				ret = connector_status_connected;
 			else if (amdgpu_connector->dac_load_detect) { /* try load detection */
 				const struct drm_encoder_helper_funcs *encoder_funcs = encoder->helper_private;
+
 				ret = encoder_funcs->detect(encoder, connector);
 			}
 		}
@@ -1439,10 +1480,8 @@ amdgpu_connector_dp_detect(struct drm_connector *connector, bool force)
 
 	amdgpu_connector_update_scratch_regs(connector, ret);
 out:
-	if (!drm_kms_helper_is_poll_worker()) {
-		pm_runtime_mark_last_busy(connector->dev->dev);
+	if (!drm_kms_helper_is_poll_worker())
 		pm_runtime_put_autosuspend(connector->dev->dev);
-	}
 
 	if (connector->connector_type == DRM_MODE_CONNECTOR_DisplayPort ||
 	    connector->connector_type == DRM_MODE_CONNECTOR_eDP)
@@ -1454,7 +1493,7 @@ out:
 }
 
 static enum drm_mode_status amdgpu_connector_dp_mode_valid(struct drm_connector *connector,
-					   struct drm_display_mode *mode)
+					   const struct drm_display_mode *mode)
 {
 	struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
 	struct amdgpu_connector_atom_dig *amdgpu_dig_connector = amdgpu_connector->con_priv;
@@ -1991,7 +2030,7 @@ amdgpu_connector_add(struct amdgpu_device *adev,
 	if (amdgpu_connector->hpd.hpd == AMDGPU_HPD_NONE) {
 		if (i2c_bus->valid) {
 			connector->polled = DRM_CONNECTOR_POLL_CONNECT |
-			                    DRM_CONNECTOR_POLL_DISCONNECT;
+						DRM_CONNECTOR_POLL_DISCONNECT;
 		}
 	} else
 		connector->polled = DRM_CONNECTOR_POLL_HPD;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.h
index 61fcef15ad72..eff833b6ed31 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.h
@@ -24,7 +24,6 @@
 #ifndef __AMDGPU_CONNECTORS_H__
 #define __AMDGPU_CONNECTORS_H__
 
-struct edid *amdgpu_connector_edid(struct drm_connector *connector);
 void amdgpu_connector_hotplug(struct drm_connector *connector);
 int amdgpu_connector_get_monitor_bpc(struct drm_connector *connector);
 u16 amdgpu_connector_encoder_get_dp_bridge_encoder_id(struct drm_connector *connector);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c
new file mode 100644
index 000000000000..425a3e564360
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c
@@ -0,0 +1,591 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include <linux/list.h>
+#include "amdgpu.h"
+
+static const guid_t MCE			= CPER_NOTIFY_MCE;
+static const guid_t CMC			= CPER_NOTIFY_CMC;
+static const guid_t BOOT		= BOOT_TYPE;
+
+static const guid_t CRASHDUMP		= AMD_CRASHDUMP;
+static const guid_t RUNTIME		= AMD_GPU_NONSTANDARD_ERROR;
+
+static void __inc_entry_length(struct cper_hdr *hdr, uint32_t size)
+{
+	hdr->record_length += size;
+}
+
+static void amdgpu_cper_get_timestamp(struct cper_timestamp *timestamp)
+{
+	struct tm tm;
+	time64_t now = ktime_get_real_seconds();
+
+	time64_to_tm(now, 0, &tm);
+	timestamp->seconds = tm.tm_sec;
+	timestamp->minutes = tm.tm_min;
+	timestamp->hours = tm.tm_hour;
+	timestamp->flag = 0;
+	timestamp->day = tm.tm_mday;
+	timestamp->month = 1 + tm.tm_mon;
+	timestamp->year = (1900 + tm.tm_year) % 100;
+	timestamp->century = (1900 + tm.tm_year) / 100;
+}
+
+void amdgpu_cper_entry_fill_hdr(struct amdgpu_device *adev,
+				struct cper_hdr *hdr,
+				enum amdgpu_cper_type type,
+				enum cper_error_severity sev)
+{
+	char record_id[16];
+
+	hdr->signature[0]		= 'C';
+	hdr->signature[1]		= 'P';
+	hdr->signature[2]		= 'E';
+	hdr->signature[3]		= 'R';
+	hdr->revision			= CPER_HDR_REV_1;
+	hdr->signature_end		= 0xFFFFFFFF;
+	hdr->error_severity		= sev;
+
+	hdr->valid_bits.platform_id	= 1;
+	hdr->valid_bits.timestamp	= 1;
+
+	amdgpu_cper_get_timestamp(&hdr->timestamp);
+
+	snprintf(record_id, 9, "%d:%X",
+		 (adev->smuio.funcs && adev->smuio.funcs->get_socket_id) ?
+			 adev->smuio.funcs->get_socket_id(adev) :
+			 0,
+		 atomic_inc_return(&adev->cper.unique_id));
+	memcpy(hdr->record_id, record_id, 8);
+
+	snprintf(hdr->platform_id, 16, "0x%04X:0x%04X",
+		 adev->pdev->vendor, adev->pdev->device);
+	/* pmfw version should be part of creator_id according to CPER spec */
+	snprintf(hdr->creator_id, 16, "%s", CPER_CREATOR_ID_AMDGPU);
+
+	switch (type) {
+	case AMDGPU_CPER_TYPE_BOOT:
+		hdr->notify_type = BOOT;
+		break;
+	case AMDGPU_CPER_TYPE_FATAL:
+	case AMDGPU_CPER_TYPE_BP_THRESHOLD:
+		hdr->notify_type = MCE;
+		break;
+	case AMDGPU_CPER_TYPE_RUNTIME:
+		if (sev == CPER_SEV_NON_FATAL_CORRECTED)
+			hdr->notify_type = CMC;
+		else
+			hdr->notify_type = MCE;
+		break;
+	default:
+		dev_err(adev->dev, "Unknown CPER Type\n");
+		break;
+	}
+
+	__inc_entry_length(hdr, HDR_LEN);
+}
+
+static int amdgpu_cper_entry_fill_section_desc(struct amdgpu_device *adev,
+					       struct cper_sec_desc *section_desc,
+					       bool bp_threshold,
+					       bool poison,
+					       enum cper_error_severity sev,
+					       guid_t sec_type,
+					       uint32_t section_length,
+					       uint32_t section_offset)
+{
+	section_desc->revision_minor		= CPER_SEC_MINOR_REV_1;
+	section_desc->revision_major		= CPER_SEC_MAJOR_REV_22;
+	section_desc->sec_offset		= section_offset;
+	section_desc->sec_length		= section_length;
+	section_desc->valid_bits.fru_text	= 1;
+	section_desc->flag_bits.primary		= 1;
+	section_desc->severity			= sev;
+	section_desc->sec_type			= sec_type;
+
+	snprintf(section_desc->fru_text, 20, "OAM%d",
+		 (adev->smuio.funcs && adev->smuio.funcs->get_socket_id) ?
+			 adev->smuio.funcs->get_socket_id(adev) :
+			 0);
+
+	if (bp_threshold)
+		section_desc->flag_bits.exceed_err_threshold = 1;
+	if (poison)
+		section_desc->flag_bits.latent_err = 1;
+
+	return 0;
+}
+
+int amdgpu_cper_entry_fill_fatal_section(struct amdgpu_device *adev,
+					 struct cper_hdr *hdr,
+					 uint32_t idx,
+					 struct cper_sec_crashdump_reg_data reg_data)
+{
+	struct cper_sec_desc *section_desc;
+	struct cper_sec_crashdump_fatal *section;
+
+	section_desc = (struct cper_sec_desc *)((uint8_t *)hdr + SEC_DESC_OFFSET(idx));
+	section = (struct cper_sec_crashdump_fatal *)((uint8_t *)hdr +
+		   FATAL_SEC_OFFSET(hdr->sec_cnt, idx));
+
+	amdgpu_cper_entry_fill_section_desc(adev, section_desc, false, false,
+					    CPER_SEV_FATAL, CRASHDUMP, FATAL_SEC_LEN,
+					    FATAL_SEC_OFFSET(hdr->sec_cnt, idx));
+
+	section->body.reg_ctx_type = CPER_CTX_TYPE_CRASH;
+	section->body.reg_arr_size = sizeof(reg_data);
+	section->body.data = reg_data;
+
+	__inc_entry_length(hdr, SEC_DESC_LEN + FATAL_SEC_LEN);
+
+	return 0;
+}
+
+int amdgpu_cper_entry_fill_runtime_section(struct amdgpu_device *adev,
+					   struct cper_hdr *hdr,
+					   uint32_t idx,
+					   enum cper_error_severity sev,
+					   uint32_t *reg_dump,
+					   uint32_t reg_count)
+{
+	struct cper_sec_desc *section_desc;
+	struct cper_sec_nonstd_err *section;
+	bool poison;
+
+	poison = sev != CPER_SEV_NON_FATAL_CORRECTED;
+	section_desc = (struct cper_sec_desc *)((uint8_t *)hdr + SEC_DESC_OFFSET(idx));
+	section = (struct cper_sec_nonstd_err *)((uint8_t *)hdr +
+		   NONSTD_SEC_OFFSET(hdr->sec_cnt, idx));
+
+	amdgpu_cper_entry_fill_section_desc(adev, section_desc, false, poison,
+					    sev, RUNTIME, NONSTD_SEC_LEN,
+					    NONSTD_SEC_OFFSET(hdr->sec_cnt, idx));
+
+	reg_count = umin(reg_count, CPER_ACA_REG_COUNT);
+
+	section->hdr.valid_bits.err_info_cnt = 1;
+	section->hdr.valid_bits.err_context_cnt = 1;
+
+	section->info.error_type = RUNTIME;
+	section->info.ms_chk_bits.err_type_valid = 1;
+	section->ctx.reg_ctx_type = CPER_CTX_TYPE_CRASH;
+	section->ctx.reg_arr_size = sizeof(section->ctx.reg_dump);
+
+	memcpy(section->ctx.reg_dump, reg_dump, reg_count * sizeof(uint32_t));
+
+	__inc_entry_length(hdr, SEC_DESC_LEN + NONSTD_SEC_LEN);
+
+	return 0;
+}
+
+int amdgpu_cper_entry_fill_bad_page_threshold_section(struct amdgpu_device *adev,
+						      struct cper_hdr *hdr,
+						      uint32_t idx)
+{
+	struct cper_sec_desc *section_desc;
+	struct cper_sec_nonstd_err *section;
+	uint32_t socket_id;
+
+	section_desc = (struct cper_sec_desc *)((uint8_t *)hdr + SEC_DESC_OFFSET(idx));
+	section = (struct cper_sec_nonstd_err *)((uint8_t *)hdr +
+		   NONSTD_SEC_OFFSET(hdr->sec_cnt, idx));
+
+	amdgpu_cper_entry_fill_section_desc(adev, section_desc, true, false,
+					    CPER_SEV_FATAL, RUNTIME, NONSTD_SEC_LEN,
+					    NONSTD_SEC_OFFSET(hdr->sec_cnt, idx));
+
+	section->hdr.valid_bits.err_info_cnt = 1;
+	section->hdr.valid_bits.err_context_cnt = 1;
+
+	section->info.error_type = RUNTIME;
+	section->info.valid_bits.ms_chk = 1;
+	section->info.ms_chk_bits.err_type_valid = 1;
+	section->info.ms_chk_bits.err_type = 1;
+	section->info.ms_chk_bits.pcc = 1;
+	section->ctx.reg_ctx_type = CPER_CTX_TYPE_CRASH;
+	section->ctx.reg_arr_size = sizeof(section->ctx.reg_dump);
+
+	/* Hardcoded Reg dump for bad page threshold CPER */
+	socket_id = (adev->smuio.funcs && adev->smuio.funcs->get_socket_id) ?
+				adev->smuio.funcs->get_socket_id(adev) :
+				0;
+	section->ctx.reg_dump[CPER_ACA_REG_CTL_LO]    = 0x1;
+	section->ctx.reg_dump[CPER_ACA_REG_CTL_HI]    = 0x0;
+	section->ctx.reg_dump[CPER_ACA_REG_STATUS_LO] = 0x137;
+	section->ctx.reg_dump[CPER_ACA_REG_STATUS_HI] = 0xB0000000;
+	section->ctx.reg_dump[CPER_ACA_REG_ADDR_LO]   = 0x0;
+	section->ctx.reg_dump[CPER_ACA_REG_ADDR_HI]   = 0x0;
+	section->ctx.reg_dump[CPER_ACA_REG_MISC0_LO]  = 0x0;
+	section->ctx.reg_dump[CPER_ACA_REG_MISC0_HI]  = 0x0;
+	section->ctx.reg_dump[CPER_ACA_REG_CONFIG_LO] = 0x2;
+	section->ctx.reg_dump[CPER_ACA_REG_CONFIG_HI] = 0x1ff;
+	section->ctx.reg_dump[CPER_ACA_REG_IPID_LO]   = (socket_id / 4) & 0x01;
+	section->ctx.reg_dump[CPER_ACA_REG_IPID_HI]   = 0x096 | (((socket_id % 4) & 0x3) << 12);
+	section->ctx.reg_dump[CPER_ACA_REG_SYND_LO]   = 0x0;
+	section->ctx.reg_dump[CPER_ACA_REG_SYND_HI]   = 0x0;
+
+	__inc_entry_length(hdr, SEC_DESC_LEN + NONSTD_SEC_LEN);
+
+	return 0;
+}
+
+struct cper_hdr *amdgpu_cper_alloc_entry(struct amdgpu_device *adev,
+					 enum amdgpu_cper_type type,
+					 uint16_t section_count)
+{
+	struct cper_hdr *hdr;
+	uint32_t size = 0;
+
+	size += HDR_LEN;
+	size += (SEC_DESC_LEN * section_count);
+
+	switch (type) {
+	case AMDGPU_CPER_TYPE_RUNTIME:
+	case AMDGPU_CPER_TYPE_BP_THRESHOLD:
+		size += (NONSTD_SEC_LEN * section_count);
+		break;
+	case AMDGPU_CPER_TYPE_FATAL:
+		size += (FATAL_SEC_LEN * section_count);
+		break;
+	case AMDGPU_CPER_TYPE_BOOT:
+		size += (BOOT_SEC_LEN * section_count);
+		break;
+	default:
+		dev_err(adev->dev, "Unknown CPER Type!\n");
+		return NULL;
+	}
+
+	hdr = kzalloc(size, GFP_KERNEL);
+	if (!hdr)
+		return NULL;
+
+	/* Save this early */
+	hdr->sec_cnt = section_count;
+
+	return hdr;
+}
+
+int amdgpu_cper_generate_ue_record(struct amdgpu_device *adev,
+				   struct aca_bank *bank)
+{
+	struct cper_hdr *fatal = NULL;
+	struct cper_sec_crashdump_reg_data reg_data = { 0 };
+	struct amdgpu_ring *ring = &adev->cper.ring_buf;
+	int ret;
+
+	fatal = amdgpu_cper_alloc_entry(adev, AMDGPU_CPER_TYPE_FATAL, 1);
+	if (!fatal) {
+		dev_err(adev->dev, "fail to alloc cper entry for ue record\n");
+		return -ENOMEM;
+	}
+
+	reg_data.status_lo = lower_32_bits(bank->regs[ACA_REG_IDX_STATUS]);
+	reg_data.status_hi = upper_32_bits(bank->regs[ACA_REG_IDX_STATUS]);
+	reg_data.addr_lo   = lower_32_bits(bank->regs[ACA_REG_IDX_ADDR]);
+	reg_data.addr_hi   = upper_32_bits(bank->regs[ACA_REG_IDX_ADDR]);
+	reg_data.ipid_lo   = lower_32_bits(bank->regs[ACA_REG_IDX_IPID]);
+	reg_data.ipid_hi   = upper_32_bits(bank->regs[ACA_REG_IDX_IPID]);
+	reg_data.synd_lo   = lower_32_bits(bank->regs[ACA_REG_IDX_SYND]);
+	reg_data.synd_hi   = upper_32_bits(bank->regs[ACA_REG_IDX_SYND]);
+
+	amdgpu_cper_entry_fill_hdr(adev, fatal, AMDGPU_CPER_TYPE_FATAL, CPER_SEV_FATAL);
+	ret = amdgpu_cper_entry_fill_fatal_section(adev, fatal, 0, reg_data);
+	if (ret)
+		return ret;
+
+	amdgpu_cper_ring_write(ring, fatal, fatal->record_length);
+	kfree(fatal);
+
+	return 0;
+}
+
+int amdgpu_cper_generate_bp_threshold_record(struct amdgpu_device *adev)
+{
+	struct cper_hdr *bp_threshold = NULL;
+	struct amdgpu_ring *ring = &adev->cper.ring_buf;
+	int ret;
+
+	bp_threshold = amdgpu_cper_alloc_entry(adev, AMDGPU_CPER_TYPE_BP_THRESHOLD, 1);
+	if (!bp_threshold) {
+		dev_err(adev->dev, "fail to alloc cper entry for bad page threshold record\n");
+		return -ENOMEM;
+	}
+
+	amdgpu_cper_entry_fill_hdr(adev, bp_threshold,
+				   AMDGPU_CPER_TYPE_BP_THRESHOLD,
+				   CPER_SEV_FATAL);
+	ret = amdgpu_cper_entry_fill_bad_page_threshold_section(adev, bp_threshold, 0);
+	if (ret)
+		return ret;
+
+	amdgpu_cper_ring_write(ring, bp_threshold, bp_threshold->record_length);
+	kfree(bp_threshold);
+
+	return 0;
+}
+
+static enum cper_error_severity amdgpu_aca_err_type_to_cper_sev(struct amdgpu_device *adev,
+								enum aca_error_type aca_err_type)
+{
+	switch (aca_err_type) {
+	case ACA_ERROR_TYPE_UE:
+		return CPER_SEV_FATAL;
+	case ACA_ERROR_TYPE_CE:
+		return CPER_SEV_NON_FATAL_CORRECTED;
+	case ACA_ERROR_TYPE_DEFERRED:
+		return CPER_SEV_NON_FATAL_UNCORRECTED;
+	default:
+		dev_err(adev->dev, "Unknown ACA error type!\n");
+		return CPER_SEV_FATAL;
+	}
+}
+
+int amdgpu_cper_generate_ce_records(struct amdgpu_device *adev,
+				    struct aca_banks *banks,
+				    uint16_t bank_count)
+{
+	struct cper_hdr *corrected = NULL;
+	enum cper_error_severity sev = CPER_SEV_NON_FATAL_CORRECTED;
+	struct amdgpu_ring *ring = &adev->cper.ring_buf;
+	uint32_t reg_data[CPER_ACA_REG_COUNT] = { 0 };
+	struct aca_bank_node *node;
+	struct aca_bank *bank;
+	uint32_t i = 0;
+	int ret;
+
+	corrected = amdgpu_cper_alloc_entry(adev, AMDGPU_CPER_TYPE_RUNTIME, bank_count);
+	if (!corrected) {
+		dev_err(adev->dev, "fail to allocate cper entry for ce records\n");
+		return -ENOMEM;
+	}
+
+	/* Raise severity if any DE is detected in the ACA bank list */
+	list_for_each_entry(node, &banks->list, node) {
+		bank = &node->bank;
+		if (bank->aca_err_type == ACA_ERROR_TYPE_DEFERRED) {
+			sev = CPER_SEV_NON_FATAL_UNCORRECTED;
+			break;
+		}
+	}
+
+	amdgpu_cper_entry_fill_hdr(adev, corrected, AMDGPU_CPER_TYPE_RUNTIME, sev);
+
+	/* Combine CE and DE in cper record */
+	list_for_each_entry(node, &banks->list, node) {
+		bank = &node->bank;
+		reg_data[CPER_ACA_REG_CTL_LO]    = lower_32_bits(bank->regs[ACA_REG_IDX_CTL]);
+		reg_data[CPER_ACA_REG_CTL_HI]    = upper_32_bits(bank->regs[ACA_REG_IDX_CTL]);
+		reg_data[CPER_ACA_REG_STATUS_LO] = lower_32_bits(bank->regs[ACA_REG_IDX_STATUS]);
+		reg_data[CPER_ACA_REG_STATUS_HI] = upper_32_bits(bank->regs[ACA_REG_IDX_STATUS]);
+		reg_data[CPER_ACA_REG_ADDR_LO]   = lower_32_bits(bank->regs[ACA_REG_IDX_ADDR]);
+		reg_data[CPER_ACA_REG_ADDR_HI]   = upper_32_bits(bank->regs[ACA_REG_IDX_ADDR]);
+		reg_data[CPER_ACA_REG_MISC0_LO]  = lower_32_bits(bank->regs[ACA_REG_IDX_MISC0]);
+		reg_data[CPER_ACA_REG_MISC0_HI]  = upper_32_bits(bank->regs[ACA_REG_IDX_MISC0]);
+		reg_data[CPER_ACA_REG_CONFIG_LO] = lower_32_bits(bank->regs[ACA_REG_IDX_CONFIG]);
+		reg_data[CPER_ACA_REG_CONFIG_HI] = upper_32_bits(bank->regs[ACA_REG_IDX_CONFIG]);
+		reg_data[CPER_ACA_REG_IPID_LO]   = lower_32_bits(bank->regs[ACA_REG_IDX_IPID]);
+		reg_data[CPER_ACA_REG_IPID_HI]   = upper_32_bits(bank->regs[ACA_REG_IDX_IPID]);
+		reg_data[CPER_ACA_REG_SYND_LO]   = lower_32_bits(bank->regs[ACA_REG_IDX_SYND]);
+		reg_data[CPER_ACA_REG_SYND_HI]   = upper_32_bits(bank->regs[ACA_REG_IDX_SYND]);
+
+		ret = amdgpu_cper_entry_fill_runtime_section(adev, corrected, i++,
+				amdgpu_aca_err_type_to_cper_sev(adev, bank->aca_err_type),
+				reg_data, CPER_ACA_REG_COUNT);
+		if (ret)
+			return ret;
+	}
+
+	amdgpu_cper_ring_write(ring, corrected, corrected->record_length);
+	kfree(corrected);
+
+	return 0;
+}
+
+static bool amdgpu_cper_is_hdr(struct amdgpu_ring *ring, u64 pos)
+{
+	struct cper_hdr *chdr;
+
+	chdr = (struct cper_hdr *)&(ring->ring[pos]);
+	return strcmp(chdr->signature, "CPER") ? false : true;
+}
+
+static u32 amdgpu_cper_ring_get_ent_sz(struct amdgpu_ring *ring, u64 pos)
+{
+	struct cper_hdr *chdr;
+	u64 p;
+	u32 chunk, rec_len = 0;
+
+	chdr = (struct cper_hdr *)&(ring->ring[pos]);
+	chunk = ring->ring_size - (pos << 2);
+
+	if (!strcmp(chdr->signature, "CPER")) {
+		rec_len = chdr->record_length;
+		goto calc;
+	}
+
+	/* ring buffer is not full, no cper data after ring->wptr */
+	if (ring->count_dw)
+		goto calc;
+
+	for (p = pos + 1; p <= ring->buf_mask; p++) {
+		chdr = (struct cper_hdr *)&(ring->ring[p]);
+		if (!strcmp(chdr->signature, "CPER")) {
+			rec_len = (p - pos) << 2;
+			goto calc;
+		}
+	}
+
+calc:
+	if (!rec_len)
+		return chunk;
+	else
+		return umin(rec_len, chunk);
+}
+
+void amdgpu_cper_ring_write(struct amdgpu_ring *ring, void *src, int count)
+{
+	u64 pos, wptr_old, rptr;
+	int rec_cnt_dw = count >> 2;
+	u32 chunk, ent_sz;
+	u8 *s = (u8 *)src;
+
+	if (count >= ring->ring_size - 4) {
+		dev_err(ring->adev->dev,
+			"CPER data size(%d) is larger than ring size(%d)\n",
+			count, ring->ring_size - 4);
+
+		return;
+	}
+
+	mutex_lock(&ring->adev->cper.ring_lock);
+
+	wptr_old = ring->wptr;
+	rptr = *ring->rptr_cpu_addr & ring->ptr_mask;
+
+	while (count) {
+		ent_sz = amdgpu_cper_ring_get_ent_sz(ring, ring->wptr);
+		chunk = umin(ent_sz, count);
+
+		memcpy(&ring->ring[ring->wptr], s, chunk);
+
+		ring->wptr += (chunk >> 2);
+		ring->wptr &= ring->ptr_mask;
+		count -= chunk;
+		s += chunk;
+	}
+
+	if (ring->count_dw < rec_cnt_dw)
+		ring->count_dw = 0;
+
+	/* the buffer is overflow, adjust rptr */
+	if (((wptr_old < rptr) && (rptr <= ring->wptr)) ||
+	    ((ring->wptr < wptr_old) && (wptr_old < rptr)) ||
+	    ((rptr <= ring->wptr) && (ring->wptr < wptr_old))) {
+		pos = (ring->wptr + 1) & ring->ptr_mask;
+
+		do {
+			ent_sz = amdgpu_cper_ring_get_ent_sz(ring, pos);
+
+			rptr += (ent_sz >> 2);
+			rptr &= ring->ptr_mask;
+			*ring->rptr_cpu_addr = rptr;
+
+			pos = rptr;
+		} while (!amdgpu_cper_is_hdr(ring, rptr));
+	}
+
+	if (ring->count_dw >= rec_cnt_dw)
+		ring->count_dw -= rec_cnt_dw;
+	mutex_unlock(&ring->adev->cper.ring_lock);
+}
+
+static u64 amdgpu_cper_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	return *(ring->rptr_cpu_addr);
+}
+
+static u64 amdgpu_cper_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	return ring->wptr;
+}
+
+static const struct amdgpu_ring_funcs cper_ring_funcs = {
+	.type = AMDGPU_RING_TYPE_CPER,
+	.align_mask = 0xff,
+	.support_64bit_ptrs = false,
+	.get_rptr = amdgpu_cper_ring_get_rptr,
+	.get_wptr = amdgpu_cper_ring_get_wptr,
+};
+
+static int amdgpu_cper_ring_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring = &(adev->cper.ring_buf);
+
+	mutex_init(&adev->cper.ring_lock);
+
+	ring->adev = NULL;
+	ring->ring_obj = NULL;
+	ring->use_doorbell = false;
+	ring->no_scheduler = true;
+	ring->funcs = &cper_ring_funcs;
+
+	sprintf(ring->name, "cper");
+	return amdgpu_ring_init(adev, ring, CPER_MAX_RING_SIZE, NULL, 0,
+				AMDGPU_RING_PRIO_DEFAULT, NULL);
+}
+
+int amdgpu_cper_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	if (!amdgpu_aca_is_enabled(adev) && !amdgpu_sriov_ras_cper_en(adev))
+		return 0;
+
+	r = amdgpu_cper_ring_init(adev);
+	if (r) {
+		dev_err(adev->dev, "failed to initialize cper ring, r = %d\n", r);
+		return r;
+	}
+
+	mutex_init(&adev->cper.cper_lock);
+
+	adev->cper.enabled = true;
+	adev->cper.max_count = CPER_MAX_ALLOWED_COUNT;
+
+	return 0;
+}
+
+int amdgpu_cper_fini(struct amdgpu_device *adev)
+{
+	if (!amdgpu_aca_is_enabled(adev) && !amdgpu_sriov_ras_cper_en(adev))
+		return 0;
+
+	adev->cper.enabled = false;
+
+	amdgpu_ring_fini(&(adev->cper.ring_buf));
+	adev->cper.count = 0;
+	adev->cper.wptr = 0;
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h
new file mode 100644
index 000000000000..353421807387
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h
@@ -0,0 +1,105 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __AMDGPU_CPER_H__
+#define __AMDGPU_CPER_H__
+
+#include "amd_cper.h"
+#include "amdgpu_aca.h"
+
+#define CPER_MAX_ALLOWED_COUNT		0x1000
+#define CPER_MAX_RING_SIZE		0X100000
+#define HDR_LEN				(sizeof(struct cper_hdr))
+#define SEC_DESC_LEN			(sizeof(struct cper_sec_desc))
+
+#define BOOT_SEC_LEN			(sizeof(struct cper_sec_crashdump_boot))
+#define FATAL_SEC_LEN			(sizeof(struct cper_sec_crashdump_fatal))
+#define NONSTD_SEC_LEN			(sizeof(struct cper_sec_nonstd_err))
+
+#define SEC_DESC_OFFSET(idx)		(HDR_LEN + (SEC_DESC_LEN * idx))
+
+#define BOOT_SEC_OFFSET(count, idx)	(HDR_LEN + (SEC_DESC_LEN * count) + (BOOT_SEC_LEN * idx))
+#define FATAL_SEC_OFFSET(count, idx)	(HDR_LEN + (SEC_DESC_LEN * count) + (FATAL_SEC_LEN * idx))
+#define NONSTD_SEC_OFFSET(count, idx)	(HDR_LEN + (SEC_DESC_LEN * count) + (NONSTD_SEC_LEN * idx))
+
+enum amdgpu_cper_type {
+	AMDGPU_CPER_TYPE_RUNTIME,
+	AMDGPU_CPER_TYPE_FATAL,
+	AMDGPU_CPER_TYPE_BOOT,
+	AMDGPU_CPER_TYPE_BP_THRESHOLD,
+};
+
+struct amdgpu_cper {
+	bool enabled;
+
+	atomic_t unique_id;
+	struct mutex cper_lock;
+
+	/* Lifetime CPERs generated */
+	uint32_t count;
+	uint32_t max_count;
+
+	uint32_t wptr;
+
+	void *ring[CPER_MAX_ALLOWED_COUNT];
+	struct amdgpu_ring ring_buf;
+	struct mutex ring_lock;
+};
+
+void amdgpu_cper_entry_fill_hdr(struct amdgpu_device *adev,
+				struct cper_hdr *hdr,
+				enum amdgpu_cper_type type,
+				enum cper_error_severity sev);
+int amdgpu_cper_entry_fill_fatal_section(struct amdgpu_device *adev,
+					 struct cper_hdr *hdr,
+					 uint32_t idx,
+					 struct cper_sec_crashdump_reg_data reg_data);
+int amdgpu_cper_entry_fill_runtime_section(struct amdgpu_device *adev,
+					   struct cper_hdr *hdr,
+					   uint32_t idx,
+					   enum cper_error_severity sev,
+					   uint32_t *reg_dump,
+					   uint32_t reg_count);
+int amdgpu_cper_entry_fill_bad_page_threshold_section(struct amdgpu_device *adev,
+						      struct cper_hdr *hdr,
+						      uint32_t section_idx);
+
+struct cper_hdr *amdgpu_cper_alloc_entry(struct amdgpu_device *adev,
+					 enum amdgpu_cper_type type,
+					 uint16_t section_count);
+/* UE must be encoded into separated cper entries, 1 UE 1 cper */
+int amdgpu_cper_generate_ue_record(struct amdgpu_device *adev,
+				   struct aca_bank *bank);
+/* CEs and DEs are combined into 1 cper entry */
+int amdgpu_cper_generate_ce_records(struct amdgpu_device *adev,
+				    struct aca_banks *banks,
+				    uint16_t bank_count);
+/* Bad page threshold is encoded into separated cper entry */
+int amdgpu_cper_generate_bp_threshold_record(struct amdgpu_device *adev);
+void amdgpu_cper_ring_write(struct amdgpu_ring *ring,
+			void *src, int count);
+int amdgpu_cper_init(struct amdgpu_device *adev);
+int amdgpu_cper_fini(struct amdgpu_device *adev);
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
index 08eced097bd8..ecdfe6cb36cc 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -40,6 +40,7 @@
 #include "amdgpu_gmc.h"
 #include "amdgpu_gem.h"
 #include "amdgpu_ras.h"
+#include "amdgpu_hmm.h"
 
 static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p,
 				 struct amdgpu_device *adev,
@@ -65,6 +66,8 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p,
 	}
 
 	amdgpu_sync_create(&p->sync);
+	drm_exec_init(&p->exec, DRM_EXEC_INTERRUPTIBLE_WAIT |
+		      DRM_EXEC_IGNORE_DUPLICATES, 0);
 	return 0;
 }
 
@@ -112,6 +115,9 @@ static int amdgpu_cs_p1_ib(struct amdgpu_cs_parser *p,
 	if (r < 0)
 		return r;
 
+	if (num_ibs[r] >= amdgpu_ring_max_ibs(chunk_ib->ip_type))
+		return -EINVAL;
+
 	++(num_ibs[r]);
 	p->gang_leader_idx = r;
 	return 0;
@@ -122,40 +128,24 @@ static int amdgpu_cs_p1_user_fence(struct amdgpu_cs_parser *p,
 				   uint32_t *offset)
 {
 	struct drm_gem_object *gobj;
-	struct amdgpu_bo *bo;
 	unsigned long size;
-	int r;
 
 	gobj = drm_gem_object_lookup(p->filp, data->handle);
 	if (gobj == NULL)
 		return -EINVAL;
 
-	bo = amdgpu_bo_ref(gem_to_amdgpu_bo(gobj));
-	p->uf_entry.priority = 0;
-	p->uf_entry.tv.bo = &bo->tbo;
-	/* One for TTM and two for the CS job */
-	p->uf_entry.tv.num_shared = 3;
-
+	p->uf_bo = amdgpu_bo_ref(gem_to_amdgpu_bo(gobj));
 	drm_gem_object_put(gobj);
 
-	size = amdgpu_bo_size(bo);
-	if (size != PAGE_SIZE || (data->offset + 8) > size) {
-		r = -EINVAL;
-		goto error_unref;
-	}
+	size = amdgpu_bo_size(p->uf_bo);
+	if (size != PAGE_SIZE || data->offset > (size - 8))
+		return -EINVAL;
 
-	if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm)) {
-		r = -EINVAL;
-		goto error_unref;
-	}
+	if (amdgpu_ttm_tt_get_usermm(p->uf_bo->tbo.ttm))
+		return -EINVAL;
 
 	*offset = data->offset;
-
 	return 0;
-
-error_unref:
-	amdgpu_bo_unref(&bo);
-	return r;
 }
 
 static int amdgpu_cs_p1_bo_handles(struct amdgpu_cs_parser *p,
@@ -189,25 +179,17 @@ static int amdgpu_cs_pass1(struct amdgpu_cs_parser *p,
 	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
 	unsigned int num_ibs[AMDGPU_CS_GANG_SIZE] = { };
 	struct amdgpu_vm *vm = &fpriv->vm;
-	uint64_t *chunk_array_user;
 	uint64_t *chunk_array;
 	uint32_t uf_offset = 0;
-	unsigned int size;
+	size_t size;
 	int ret;
 	int i;
 
-	chunk_array = kvmalloc_array(cs->in.num_chunks, sizeof(uint64_t),
-				     GFP_KERNEL);
-	if (!chunk_array)
-		return -ENOMEM;
-
-	/* get chunks */
-	chunk_array_user = u64_to_user_ptr(cs->in.chunks);
-	if (copy_from_user(chunk_array, chunk_array_user,
-			   sizeof(uint64_t)*cs->in.num_chunks)) {
-		ret = -EFAULT;
-		goto free_chunk;
-	}
+	chunk_array = memdup_array_user(u64_to_user_ptr(cs->in.chunks),
+					cs->in.num_chunks,
+					sizeof(uint64_t));
+	if (IS_ERR(chunk_array))
+		return PTR_ERR(chunk_array);
 
 	p->nchunks = cs->in.num_chunks;
 	p->chunks = kvmalloc_array(p->nchunks, sizeof(struct amdgpu_cs_chunk),
@@ -218,9 +200,8 @@ static int amdgpu_cs_pass1(struct amdgpu_cs_parser *p,
 	}
 
 	for (i = 0; i < p->nchunks; i++) {
-		struct drm_amdgpu_cs_chunk __user **chunk_ptr = NULL;
+		struct drm_amdgpu_cs_chunk __user *chunk_ptr = NULL;
 		struct drm_amdgpu_cs_chunk user_chunk;
-		uint32_t __user *cdata;
 
 		chunk_ptr = u64_to_user_ptr(chunk_array[i]);
 		if (copy_from_user(&user_chunk, chunk_ptr,
@@ -233,20 +214,16 @@ static int amdgpu_cs_pass1(struct amdgpu_cs_parser *p,
 		p->chunks[i].length_dw = user_chunk.length_dw;
 
 		size = p->chunks[i].length_dw;
-		cdata = u64_to_user_ptr(user_chunk.chunk_data);
 
-		p->chunks[i].kdata = kvmalloc_array(size, sizeof(uint32_t),
-						    GFP_KERNEL);
-		if (p->chunks[i].kdata == NULL) {
-			ret = -ENOMEM;
+		p->chunks[i].kdata = vmemdup_array_user(u64_to_user_ptr(user_chunk.chunk_data),
+							size,
+							sizeof(uint32_t));
+		if (IS_ERR(p->chunks[i].kdata)) {
+			ret = PTR_ERR(p->chunks[i].kdata);
 			i--;
 			goto free_partial_kdata;
 		}
 		size *= sizeof(uint32_t);
-		if (copy_from_user(p->chunks[i].kdata, cdata, size)) {
-			ret = -EFAULT;
-			goto free_partial_kdata;
-		}
 
 		/* Assume the worst on the following checks */
 		ret = -EINVAL;
@@ -274,6 +251,10 @@ static int amdgpu_cs_pass1(struct amdgpu_cs_parser *p,
 			if (size < sizeof(struct drm_amdgpu_bo_list_in))
 				goto free_partial_kdata;
 
+			/* Only a single BO list is allowed to simplify handling. */
+			if (p->bo_list)
+				goto free_partial_kdata;
+
 			ret = amdgpu_cs_p1_bo_handles(p, p->chunks[i].kdata);
 			if (ret)
 				goto free_partial_kdata;
@@ -285,6 +266,7 @@ static int amdgpu_cs_pass1(struct amdgpu_cs_parser *p,
 		case AMDGPU_CHUNK_ID_SCHEDULED_DEPENDENCIES:
 		case AMDGPU_CHUNK_ID_SYNCOBJ_TIMELINE_WAIT:
 		case AMDGPU_CHUNK_ID_SYNCOBJ_TIMELINE_SIGNAL:
+		case AMDGPU_CHUNK_ID_CP_GFX_SHADOW:
 			break;
 
 		default:
@@ -292,25 +274,45 @@ static int amdgpu_cs_pass1(struct amdgpu_cs_parser *p,
 		}
 	}
 
-	if (!p->gang_size) {
+	if (!p->gang_size || (amdgpu_sriov_vf(p->adev) && p->gang_size > 1)) {
 		ret = -EINVAL;
-		goto free_partial_kdata;
+		goto free_all_kdata;
 	}
 
 	for (i = 0; i < p->gang_size; ++i) {
 		ret = amdgpu_job_alloc(p->adev, vm, p->entities[i], vm,
-				       num_ibs[i], &p->jobs[i]);
+				       num_ibs[i], &p->jobs[i],
+				       p->filp->client_id);
 		if (ret)
 			goto free_all_kdata;
+		switch (p->adev->enforce_isolation[fpriv->xcp_id]) {
+		case AMDGPU_ENFORCE_ISOLATION_DISABLE:
+		default:
+			p->jobs[i]->enforce_isolation = false;
+			p->jobs[i]->run_cleaner_shader = false;
+			break;
+		case AMDGPU_ENFORCE_ISOLATION_ENABLE:
+			p->jobs[i]->enforce_isolation = true;
+			p->jobs[i]->run_cleaner_shader = true;
+			break;
+		case AMDGPU_ENFORCE_ISOLATION_ENABLE_LEGACY:
+			p->jobs[i]->enforce_isolation = true;
+			p->jobs[i]->run_cleaner_shader = false;
+			break;
+		case AMDGPU_ENFORCE_ISOLATION_NO_CLEANER_SHADER:
+			p->jobs[i]->enforce_isolation = true;
+			p->jobs[i]->run_cleaner_shader = false;
+			break;
+		}
 	}
 	p->gang_leader = p->jobs[p->gang_leader_idx];
 
-	if (p->ctx->vram_lost_counter != p->gang_leader->vram_lost_counter) {
+	if (p->ctx->generation != p->gang_leader->generation) {
 		ret = -ECANCELED;
 		goto free_all_kdata;
 	}
 
-	if (p->uf_entry.tv.bo)
+	if (p->uf_bo)
 		p->gang_leader->uf_addr = uf_offset;
 	kvfree(chunk_array);
 
@@ -354,10 +356,20 @@ static int amdgpu_cs_p2_ib(struct amdgpu_cs_parser *p,
 	ring = amdgpu_job_ring(job);
 	ib = &job->ibs[job->num_ibs++];
 
+	/* submissions to kernel queues are disabled */
+	if (ring->no_user_submission)
+		return -EINVAL;
+
 	/* MM engine doesn't support user fences */
-	if (p->uf_entry.tv.bo && ring->funcs->no_user_fence)
+	if (p->uf_bo && ring->funcs->no_user_fence)
 		return -EINVAL;
 
+	if (!p->adev->debug_enable_ce_cs &&
+	    chunk_ib->flags & AMDGPU_IB_FLAG_CE) {
+		dev_err_ratelimited(p->adev->dev, "CE CS is blocked, use debug=0x400 to override\n");
+		return -EINVAL;
+	}
+
 	if (chunk_ib->ip_type == AMDGPU_HW_IP_GFX &&
 	    chunk_ib->flags & AMDGPU_IB_FLAG_PREEMPT) {
 		if (chunk_ib->flags & AMDGPU_IB_FLAG_CE)
@@ -378,7 +390,7 @@ static int amdgpu_cs_p2_ib(struct amdgpu_cs_parser *p,
 			   chunk_ib->ib_bytes : 0,
 			   AMDGPU_IB_POOL_DELAYED, ib);
 	if (r) {
-		DRM_ERROR("Failed to get ib !\n");
+		drm_err(adev_to_drm(p->adev), "Failed to get ib !\n");
 		return r;
 	}
 
@@ -393,7 +405,7 @@ static int amdgpu_cs_p2_dependencies(struct amdgpu_cs_parser *p,
 {
 	struct drm_amdgpu_cs_chunk_dep *deps = chunk->kdata;
 	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
-	unsigned num_deps;
+	unsigned int num_deps;
 	int i, r;
 
 	num_deps = chunk->length_dw * 4 /
@@ -433,7 +445,7 @@ static int amdgpu_cs_p2_dependencies(struct amdgpu_cs_parser *p,
 			dma_fence_put(old);
 		}
 
-		r = amdgpu_sync_fence(&p->sync, fence);
+		r = amdgpu_sync_fence(&p->sync, fence, GFP_KERNEL);
 		dma_fence_put(fence);
 		if (r)
 			return r;
@@ -450,12 +462,12 @@ static int amdgpu_syncobj_lookup_and_add(struct amdgpu_cs_parser *p,
 
 	r = drm_syncobj_find_fence(p->filp, handle, point, flags, &fence);
 	if (r) {
-		DRM_ERROR("syncobj %u failed to find fence @ %llu (%d)!\n",
+		drm_err(adev_to_drm(p->adev), "syncobj %u failed to find fence @ %llu (%d)!\n",
 			  handle, point, r);
 		return r;
 	}
 
-	r = amdgpu_sync_fence(&p->sync, fence);
+	r = amdgpu_sync_fence(&p->sync, fence, GFP_KERNEL);
 	dma_fence_put(fence);
 	return r;
 }
@@ -464,7 +476,7 @@ static int amdgpu_cs_p2_syncobj_in(struct amdgpu_cs_parser *p,
 				   struct amdgpu_cs_chunk *chunk)
 {
 	struct drm_amdgpu_cs_chunk_sem *deps = chunk->kdata;
-	unsigned num_deps;
+	unsigned int num_deps;
 	int i, r;
 
 	num_deps = chunk->length_dw * 4 /
@@ -482,7 +494,7 @@ static int amdgpu_cs_p2_syncobj_timeline_wait(struct amdgpu_cs_parser *p,
 					      struct amdgpu_cs_chunk *chunk)
 {
 	struct drm_amdgpu_cs_chunk_syncobj *syncobj_deps = chunk->kdata;
-	unsigned num_deps;
+	unsigned int num_deps;
 	int i, r;
 
 	num_deps = chunk->length_dw * 4 /
@@ -502,7 +514,7 @@ static int amdgpu_cs_p2_syncobj_out(struct amdgpu_cs_parser *p,
 				    struct amdgpu_cs_chunk *chunk)
 {
 	struct drm_amdgpu_cs_chunk_sem *deps = chunk->kdata;
-	unsigned num_deps;
+	unsigned int num_deps;
 	int i;
 
 	num_deps = chunk->length_dw * 4 /
@@ -536,7 +548,7 @@ static int amdgpu_cs_p2_syncobj_timeline_signal(struct amdgpu_cs_parser *p,
 						struct amdgpu_cs_chunk *chunk)
 {
 	struct drm_amdgpu_cs_chunk_syncobj *syncobj_deps = chunk->kdata;
-	unsigned num_deps;
+	unsigned int num_deps;
 	int i;
 
 	num_deps = chunk->length_dw * 4 /
@@ -575,6 +587,26 @@ static int amdgpu_cs_p2_syncobj_timeline_signal(struct amdgpu_cs_parser *p,
 	return 0;
 }
 
+static int amdgpu_cs_p2_shadow(struct amdgpu_cs_parser *p,
+			       struct amdgpu_cs_chunk *chunk)
+{
+	struct drm_amdgpu_cs_chunk_cp_gfx_shadow *shadow = chunk->kdata;
+	int i;
+
+	if (shadow->flags & ~AMDGPU_CS_CHUNK_CP_GFX_SHADOW_FLAGS_INIT_SHADOW)
+		return -EINVAL;
+
+	for (i = 0; i < p->gang_size; ++i) {
+		p->jobs[i]->shadow_va = shadow->shadow_va;
+		p->jobs[i]->csa_va = shadow->csa_va;
+		p->jobs[i]->gds_va = shadow->gds_va;
+		p->jobs[i]->init_shadow =
+			shadow->flags & AMDGPU_CS_CHUNK_CP_GFX_SHADOW_FLAGS_INIT_SHADOW;
+	}
+
+	return 0;
+}
+
 static int amdgpu_cs_pass2(struct amdgpu_cs_parser *p)
 {
 	unsigned int ce_preempt = 0, de_preempt = 0;
@@ -617,6 +649,11 @@ static int amdgpu_cs_pass2(struct amdgpu_cs_parser *p)
 			if (r)
 				return r;
 			break;
+		case AMDGPU_CHUNK_ID_CP_GFX_SHADOW:
+			r = amdgpu_cs_p2_shadow(p, chunk);
+			if (r)
+				return r;
+			break;
 		}
 	}
 
@@ -671,7 +708,7 @@ static void amdgpu_cs_get_threshold_for_moves(struct amdgpu_device *adev,
 	 */
 	const s64 us_upper_bound = 200000;
 
-	if (!adev->mm_stats.log2_max_MBps) {
+	if ((!adev->mm_stats.log2_max_MBps) || !ttm_resource_manager_used(&adev->mman.vram_mgr.manager)) {
 		*max_bytes = 0;
 		*max_vis_bytes = 0;
 		return;
@@ -729,6 +766,7 @@ static void amdgpu_cs_get_threshold_for_moves(struct amdgpu_device *adev,
 
 		if (used_vis_vram < total_vis_vram) {
 			u64 free_vis_vram = total_vis_vram - used_vis_vram;
+
 			adev->mm_stats.accum_us_vis = min(adev->mm_stats.accum_us_vis +
 							  increment_us, us_upper_bound);
 
@@ -803,7 +841,7 @@ retry:
 
 	p->bytes_moved += ctx.bytes_moved;
 	if (!amdgpu_gmc_vram_full_visible(&adev->gmc) &&
-	    amdgpu_bo_in_cpu_visible_vram(bo))
+	    amdgpu_res_cpu_visible(adev, bo->tbo.resource))
 		p->bytes_moved_vis += ctx.bytes_moved;
 
 	if (unlikely(r == -ENOMEM) && domain != bo->allowed_domains) {
@@ -814,55 +852,18 @@ retry:
 	return r;
 }
 
-static int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p,
-			    struct list_head *validated)
-{
-	struct ttm_operation_ctx ctx = { true, false };
-	struct amdgpu_bo_list_entry *lobj;
-	int r;
-
-	list_for_each_entry(lobj, validated, tv.head) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(lobj->tv.bo);
-		struct mm_struct *usermm;
-
-		usermm = amdgpu_ttm_tt_get_usermm(bo->tbo.ttm);
-		if (usermm && usermm != current->mm)
-			return -EPERM;
-
-		if (amdgpu_ttm_tt_is_userptr(bo->tbo.ttm) &&
-		    lobj->user_invalidated && lobj->user_pages) {
-			amdgpu_bo_placement_from_domain(bo,
-							AMDGPU_GEM_DOMAIN_CPU);
-			r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
-			if (r)
-				return r;
-
-			amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm,
-						     lobj->user_pages);
-		}
-
-		r = amdgpu_cs_bo_validate(p, bo);
-		if (r)
-			return r;
-
-		kvfree(lobj->user_pages);
-		lobj->user_pages = NULL;
-	}
-	return 0;
-}
-
 static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 				union drm_amdgpu_cs *cs)
 {
 	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
+	struct ttm_operation_ctx ctx = { true, false };
 	struct amdgpu_vm *vm = &fpriv->vm;
 	struct amdgpu_bo_list_entry *e;
-	struct list_head duplicates;
+	struct drm_gem_object *obj;
+	unsigned long index;
 	unsigned int i;
 	int r;
 
-	INIT_LIST_HEAD(&p->validated);
-
 	/* p->bo_list could already be assigned if AMDGPU_CHUNK_ID_BO_HANDLES is present */
 	if (cs->in.bo_list_handle) {
 		if (p->bo_list)
@@ -882,45 +883,25 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 
 	mutex_lock(&p->bo_list->bo_list_mutex);
 
-	/* One for TTM and one for the CS job */
-	amdgpu_bo_list_for_each_entry(e, p->bo_list)
-		e->tv.num_shared = 2;
-
-	amdgpu_bo_list_get_list(p->bo_list, &p->validated);
-
-	INIT_LIST_HEAD(&duplicates);
-	amdgpu_vm_get_pd_bo(&fpriv->vm, &p->validated, &p->vm_pd);
-
-	if (p->uf_entry.tv.bo && !ttm_to_amdgpu_bo(p->uf_entry.tv.bo)->parent)
-		list_add(&p->uf_entry.tv.head, &p->validated);
-
 	/* Get userptr backing pages. If pages are updated after registered
 	 * in amdgpu_gem_userptr_ioctl(), amdgpu_cs_list_validate() will do
 	 * amdgpu_ttm_backend_bind() to flush and invalidate new pages
 	 */
 	amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
 		bool userpage_invalidated = false;
-		int i;
-
-		e->user_pages = kvmalloc_array(bo->tbo.ttm->num_pages,
-					sizeof(struct page *),
-					GFP_KERNEL | __GFP_ZERO);
-		if (!e->user_pages) {
-			DRM_ERROR("kvmalloc_array failure\n");
-			r = -ENOMEM;
-			goto out_free_user_pages;
-		}
+		struct amdgpu_bo *bo = e->bo;
 
-		r = amdgpu_ttm_tt_get_user_pages(bo, e->user_pages, &e->range);
-		if (r) {
-			kvfree(e->user_pages);
-			e->user_pages = NULL;
+		e->range = amdgpu_hmm_range_alloc(NULL);
+		if (unlikely(!e->range))
+			return -ENOMEM;
+
+		r = amdgpu_ttm_tt_get_user_pages(bo, e->range);
+		if (r)
 			goto out_free_user_pages;
-		}
 
 		for (i = 0; i < bo->tbo.ttm->num_pages; i++) {
-			if (bo->tbo.ttm->pages[i] != e->user_pages[i]) {
+			if (bo->tbo.ttm->pages[i] !=
+				hmm_pfn_to_page(e->range->hmm_range.hmm_pfns[i])) {
 				userpage_invalidated = true;
 				break;
 			}
@@ -928,18 +909,53 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 		e->user_invalidated = userpage_invalidated;
 	}
 
-	r = ttm_eu_reserve_buffers(&p->ticket, &p->validated, true,
-				   &duplicates);
-	if (unlikely(r != 0)) {
-		if (r != -ERESTARTSYS)
-			DRM_ERROR("ttm_eu_reserve_buffers failed.\n");
-		goto out_free_user_pages;
+	drm_exec_until_all_locked(&p->exec) {
+		r = amdgpu_vm_lock_pd(&fpriv->vm, &p->exec, 1 + p->gang_size);
+		drm_exec_retry_on_contention(&p->exec);
+		if (unlikely(r))
+			goto out_free_user_pages;
+
+		amdgpu_bo_list_for_each_entry(e, p->bo_list) {
+			/* One fence for TTM and one for each CS job */
+			r = drm_exec_prepare_obj(&p->exec, &e->bo->tbo.base,
+						 1 + p->gang_size);
+			drm_exec_retry_on_contention(&p->exec);
+			if (unlikely(r))
+				goto out_free_user_pages;
+
+			e->bo_va = amdgpu_vm_bo_find(vm, e->bo);
+		}
+
+		if (p->uf_bo) {
+			r = drm_exec_prepare_obj(&p->exec, &p->uf_bo->tbo.base,
+						 1 + p->gang_size);
+			drm_exec_retry_on_contention(&p->exec);
+			if (unlikely(r))
+				goto out_free_user_pages;
+		}
 	}
 
-	amdgpu_bo_list_for_each_entry(e, p->bo_list) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
+	amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) {
+		struct mm_struct *usermm;
 
-		e->bo_va = amdgpu_vm_bo_find(vm, bo);
+		usermm = amdgpu_ttm_tt_get_usermm(e->bo->tbo.ttm);
+		if (usermm && usermm != current->mm) {
+			r = -EPERM;
+			goto out_free_user_pages;
+		}
+
+		if (amdgpu_ttm_tt_is_userptr(e->bo->tbo.ttm) &&
+		    e->user_invalidated) {
+			amdgpu_bo_placement_from_domain(e->bo,
+							AMDGPU_GEM_DOMAIN_CPU);
+			r = ttm_bo_validate(&e->bo->tbo, &e->bo->placement,
+					    &ctx);
+			if (r)
+				goto out_free_user_pages;
+
+			amdgpu_ttm_tt_set_user_pages(e->bo->tbo.ttm,
+						     e->range);
+		}
 	}
 
 	amdgpu_cs_get_threshold_for_moves(p->adev, &p->bytes_moved_threshold,
@@ -947,29 +963,25 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 	p->bytes_moved = 0;
 	p->bytes_moved_vis = 0;
 
-	r = amdgpu_vm_validate_pt_bos(p->adev, &fpriv->vm,
-				      amdgpu_cs_bo_validate, p);
+	r = amdgpu_vm_validate(p->adev, &fpriv->vm, NULL,
+			       amdgpu_cs_bo_validate, p);
 	if (r) {
-		DRM_ERROR("amdgpu_vm_validate_pt_bos() failed.\n");
-		goto error_validate;
+		drm_err(adev_to_drm(p->adev), "amdgpu_vm_validate() failed.\n");
+		goto out_free_user_pages;
 	}
 
-	r = amdgpu_cs_list_validate(p, &duplicates);
-	if (r)
-		goto error_validate;
-
-	r = amdgpu_cs_list_validate(p, &p->validated);
-	if (r)
-		goto error_validate;
-
-	if (p->uf_entry.tv.bo) {
-		struct amdgpu_bo *uf = ttm_to_amdgpu_bo(p->uf_entry.tv.bo);
+	drm_exec_for_each_locked_object(&p->exec, index, obj) {
+		r = amdgpu_cs_bo_validate(p, gem_to_amdgpu_bo(obj));
+		if (unlikely(r))
+			goto out_free_user_pages;
+	}
 
-		r = amdgpu_ttm_alloc_gart(&uf->tbo);
-		if (r)
-			goto error_validate;
+	if (p->uf_bo) {
+		r = amdgpu_ttm_alloc_gart(&p->uf_bo->tbo);
+		if (unlikely(r))
+			goto out_free_user_pages;
 
-		p->gang_leader->uf_addr += amdgpu_bo_gpu_offset(uf);
+		p->gang_leader->uf_addr += amdgpu_bo_gpu_offset(p->uf_bo);
 	}
 
 	amdgpu_cs_report_moved_bytes(p->adev, p->bytes_moved,
@@ -981,18 +993,9 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 					 p->bo_list->oa_obj);
 	return 0;
 
-error_validate:
-	ttm_eu_backoff_reservation(&p->ticket, &p->validated);
-
 out_free_user_pages:
 	amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
-
-		if (!e->user_pages)
-			continue;
-		amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm, e->range);
-		kvfree(e->user_pages);
-		e->user_pages = NULL;
+		amdgpu_hmm_range_free(e->range);
 		e->range = NULL;
 	}
 	mutex_unlock(&p->bo_list->bo_list_mutex);
@@ -1035,21 +1038,20 @@ static int amdgpu_cs_patch_ibs(struct amdgpu_cs_parser *p,
 		va_start = ib->gpu_addr & AMDGPU_GMC_HOLE_MASK;
 		r = amdgpu_cs_find_mapping(p, va_start, &aobj, &m);
 		if (r) {
-			DRM_ERROR("IB va_start is invalid\n");
+			drm_err(adev_to_drm(p->adev), "IB va_start is invalid\n");
 			return r;
 		}
 
 		if ((va_start + ib->length_dw * 4) >
 		    (m->last + 1) * AMDGPU_GPU_PAGE_SIZE) {
-			DRM_ERROR("IB va_start+ib_bytes is invalid\n");
+			drm_err(adev_to_drm(p->adev), "IB va_start+ib_bytes is invalid\n");
 			return -EINVAL;
 		}
 
 		/* the IB should be reserved at this point */
 		r = amdgpu_bo_kmap(aobj, (void **)&kptr);
-		if (r) {
+		if (r)
 			return r;
-		}
 
 		kptr += va_start - (m->start * AMDGPU_GPU_PAGE_SIZE);
 
@@ -1060,6 +1062,9 @@ static int amdgpu_cs_patch_ibs(struct amdgpu_cs_parser *p,
 			r = amdgpu_ring_parse_cs(ring, p, job, ib);
 			if (r)
 				return r;
+
+			if (ib->sa_bo)
+				ib->gpu_addr =  amdgpu_sa_bo_gpu_addr(ib->sa_bo);
 		} else {
 			ib->ptr = (uint32_t *)kptr;
 			r = amdgpu_ring_patch_cs_in_place(ring, p, job, ib);
@@ -1093,10 +1098,27 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p)
 	struct amdgpu_vm *vm = &fpriv->vm;
 	struct amdgpu_bo_list_entry *e;
 	struct amdgpu_bo_va *bo_va;
-	struct amdgpu_bo *bo;
 	unsigned int i;
 	int r;
 
+	/*
+	 * We can't use gang submit on with reserved VMIDs when the VM changes
+	 * can't be invalidated by more than one engine at the same time.
+	 */
+	if (p->gang_size > 1 && !adev->vm_manager.concurrent_flush) {
+		for (i = 0; i < p->gang_size; ++i) {
+			struct drm_sched_entity *entity = p->entities[i];
+			struct drm_gpu_scheduler *sched = entity->rq->sched;
+			struct amdgpu_ring *ring = to_amdgpu_ring(sched);
+
+			if (amdgpu_vmid_uses_reserved(vm, ring->vm_hub))
+				return -EINVAL;
+		}
+	}
+
+	if (!amdgpu_vm_ready(vm))
+		return -EINVAL;
+
 	r = amdgpu_vm_clear_freed(adev, vm, NULL);
 	if (r)
 		return r;
@@ -1105,7 +1127,8 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p)
 	if (r)
 		return r;
 
-	r = amdgpu_sync_fence(&p->sync, fpriv->prt_va->last_pt_update);
+	r = amdgpu_sync_fence(&p->sync, fpriv->prt_va->last_pt_update,
+			      GFP_KERNEL);
 	if (r)
 		return r;
 
@@ -1116,17 +1139,18 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p)
 		if (r)
 			return r;
 
-		r = amdgpu_sync_fence(&p->sync, bo_va->last_pt_update);
+		r = amdgpu_sync_fence(&p->sync, bo_va->last_pt_update,
+				      GFP_KERNEL);
 		if (r)
 			return r;
 	}
 
+	/* FIXME: In theory this loop shouldn't be needed any more when
+	 * amdgpu_vm_handle_moved handles all moved BOs that are reserved
+	 * with p->ticket. But removing it caused test regressions, so I'm
+	 * leaving it here for now.
+	 */
 	amdgpu_bo_list_for_each_entry(e, p->bo_list) {
-		/* ignore duplicates */
-		bo = ttm_to_amdgpu_bo(e->tv.bo);
-		if (!bo)
-			continue;
-
 		bo_va = e->bo_va;
 		if (bo_va == NULL)
 			continue;
@@ -1135,12 +1159,13 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p)
 		if (r)
 			return r;
 
-		r = amdgpu_sync_fence(&p->sync, bo_va->last_pt_update);
+		r = amdgpu_sync_fence(&p->sync, bo_va->last_pt_update,
+				      GFP_KERNEL);
 		if (r)
 			return r;
 	}
 
-	r = amdgpu_vm_handle_moved(adev, vm);
+	r = amdgpu_vm_handle_moved(adev, vm, &p->exec.ticket);
 	if (r)
 		return r;
 
@@ -1148,7 +1173,7 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p)
 	if (r)
 		return r;
 
-	r = amdgpu_sync_fence(&p->sync, vm->last_update);
+	r = amdgpu_sync_fence(&p->sync, vm->last_update, GFP_KERNEL);
 	if (r)
 		return r;
 
@@ -1161,16 +1186,16 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p)
 		job->vm_pd_addr = amdgpu_gmc_pd_addr(vm->root.bo);
 	}
 
-	if (amdgpu_vm_debug) {
+	if (adev->debug_vm) {
 		/* Invalidate all BOs to test for userspace bugs */
 		amdgpu_bo_list_for_each_entry(e, p->bo_list) {
-			struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
+			struct amdgpu_bo *bo = e->bo;
 
 			/* ignore duplicates */
 			if (!bo)
 				continue;
 
-			amdgpu_vm_bo_invalidate(adev, bo, false);
+			amdgpu_vm_bo_invalidate(bo, false);
 		}
 	}
 
@@ -1181,20 +1206,22 @@ static int amdgpu_cs_sync_rings(struct amdgpu_cs_parser *p)
 {
 	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
 	struct drm_gpu_scheduler *sched;
-	struct amdgpu_bo_list_entry *e;
+	struct drm_gem_object *obj;
 	struct dma_fence *fence;
+	unsigned long index;
 	unsigned int i;
 	int r;
 
 	r = amdgpu_ctx_wait_prev_fence(p->ctx, p->entities[p->gang_leader_idx]);
 	if (r) {
 		if (r != -ERESTARTSYS)
-			DRM_ERROR("amdgpu_ctx_wait_prev_fence failed.\n");
+			drm_err(adev_to_drm(p->adev), "amdgpu_ctx_wait_prev_fence failed.\n");
 		return r;
 	}
 
-	list_for_each_entry(e, &p->validated, tv.head) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
+	drm_exec_for_each_locked_object(&p->exec, index, obj) {
+		struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
+
 		struct dma_resv *resv = bo->tbo.base.resv;
 		enum amdgpu_sync_mode sync_mode;
 
@@ -1227,7 +1254,8 @@ static int amdgpu_cs_sync_rings(struct amdgpu_cs_parser *p)
 			continue;
 		}
 
-		r = amdgpu_sync_fence(&p->gang_leader->explicit_sync, fence);
+		r = amdgpu_sync_fence(&p->gang_leader->explicit_sync, fence,
+				      GFP_KERNEL);
 		dma_fence_put(fence);
 		if (r)
 			return r;
@@ -1258,6 +1286,8 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
 	struct amdgpu_job *leader = p->gang_leader;
 	struct amdgpu_bo_list_entry *e;
+	struct drm_gem_object *gobj;
+	unsigned long index;
 	unsigned int i;
 	uint64_t seq;
 	int r;
@@ -1276,7 +1306,7 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 		r = drm_sched_job_add_dependency(&leader->base, fence);
 		if (r) {
 			dma_fence_put(fence);
-			goto error_cleanup;
+			return r;
 		}
 	}
 
@@ -1296,31 +1326,33 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	 */
 	r = 0;
 	amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
-
-		r |= !amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm, e->range);
+		r |= !amdgpu_hmm_range_valid(e->range);
+		amdgpu_hmm_range_free(e->range);
 		e->range = NULL;
 	}
 	if (r) {
 		r = -EAGAIN;
-		goto error_unlock;
+		mutex_unlock(&p->adev->notifier_lock);
+		return r;
 	}
 
 	p->fence = dma_fence_get(&leader->base.s_fence->finished);
-	list_for_each_entry(e, &p->validated, tv.head) {
+	drm_exec_for_each_locked_object(&p->exec, index, gobj) {
+
+		ttm_bo_move_to_lru_tail_unlocked(&gem_to_amdgpu_bo(gobj)->tbo);
 
 		/* Everybody except for the gang leader uses READ */
 		for (i = 0; i < p->gang_size; ++i) {
 			if (p->jobs[i] == leader)
 				continue;
 
-			dma_resv_add_fence(e->tv.bo->base.resv,
+			dma_resv_add_fence(gobj->resv,
 					   &p->jobs[i]->base.s_fence->finished,
 					   DMA_RESV_USAGE_READ);
 		}
 
-		/* The gang leader is remembered as writer */
-		e->tv.num_shared = 0;
+		/* The gang leader as remembered as writer */
+		dma_resv_add_fence(gobj->resv, p->fence, DMA_RESV_USAGE_WRITE);
 	}
 
 	seq = amdgpu_ctx_add_fence(p->ctx, p->entities[p->gang_leader_idx],
@@ -1336,7 +1368,7 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	cs->out.handle = seq;
 	leader->uf_sequence = seq;
 
-	amdgpu_vm_bo_trace_cs(&fpriv->vm, &p->ticket);
+	amdgpu_vm_bo_trace_cs(&fpriv->vm, &p->exec.ticket);
 	for (i = 0; i < p->gang_size; ++i) {
 		amdgpu_job_free_resources(p->jobs[i]);
 		trace_amdgpu_cs_ioctl(p->jobs[i]);
@@ -1345,27 +1377,20 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	}
 
 	amdgpu_vm_move_to_lru_tail(p->adev, &fpriv->vm);
-	ttm_eu_fence_buffer_objects(&p->ticket, &p->validated, p->fence);
 
 	mutex_unlock(&p->adev->notifier_lock);
 	mutex_unlock(&p->bo_list->bo_list_mutex);
 	return 0;
-
-error_unlock:
-	mutex_unlock(&p->adev->notifier_lock);
-
-error_cleanup:
-	for (i = 0; i < p->gang_size; ++i)
-		drm_sched_job_cleanup(&p->jobs[i]->base);
-	return r;
 }
 
 /* Cleanup the parser structure */
 static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser)
 {
-	unsigned i;
+	unsigned int i;
 
 	amdgpu_sync_free(&parser->sync);
+	drm_exec_fini(&parser->exec);
+
 	for (i = 0; i < parser->num_post_deps; i++) {
 		drm_syncobj_put(parser->post_deps[i].syncobj);
 		kfree(parser->post_deps[i].chain);
@@ -1386,11 +1411,7 @@ static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser)
 		if (parser->jobs[i])
 			amdgpu_job_free(parser->jobs[i]);
 	}
-	if (parser->uf_entry.tv.bo) {
-		struct amdgpu_bo *uf = ttm_to_amdgpu_bo(parser->uf_entry.tv.bo);
-
-		amdgpu_bo_unref(&uf);
-	}
+	amdgpu_bo_unref(&parser->uf_bo);
 }
 
 int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
@@ -1407,8 +1428,7 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 
 	r = amdgpu_cs_parser_init(&parser, adev, filp, data);
 	if (r) {
-		if (printk_ratelimit())
-			DRM_ERROR("Failed to initialize parser %d!\n", r);
+		drm_err_ratelimited(dev, "Failed to initialize parser %d!\n", r);
 		return r;
 	}
 
@@ -1423,9 +1443,9 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	r = amdgpu_cs_parser_bos(&parser, data);
 	if (r) {
 		if (r == -ENOMEM)
-			DRM_ERROR("Not enough memory for command submission!\n");
+			drm_err(dev, "Not enough memory for command submission!\n");
 		else if (r != -ERESTARTSYS && r != -EAGAIN)
-			DRM_ERROR("Failed to process the buffer list %d!\n", r);
+			drm_dbg(dev, "Failed to process the buffer list %d!\n", r);
 		goto error_fini;
 	}
 
@@ -1451,7 +1471,6 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	return 0;
 
 error_backoff:
-	ttm_eu_backoff_reservation(&parser.ticket, &parser.validated);
 	mutex_unlock(&parser.bo_list->bo_list_mutex);
 
 error_fini:
@@ -1631,15 +1650,15 @@ static int amdgpu_cs_wait_all_fences(struct amdgpu_device *adev,
 			continue;
 
 		r = dma_fence_wait_timeout(fence, true, timeout);
+		if (r > 0 && fence->error)
+			r = fence->error;
+
 		dma_fence_put(fence);
 		if (r < 0)
 			return r;
 
 		if (r == 0)
 			break;
-
-		if (fence->error)
-			return fence->error;
 	}
 
 	memset(wait, 0, sizeof(*wait));
@@ -1725,30 +1744,21 @@ int amdgpu_cs_wait_fences_ioctl(struct drm_device *dev, void *data,
 {
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	union drm_amdgpu_wait_fences *wait = data;
-	uint32_t fence_count = wait->in.fence_count;
-	struct drm_amdgpu_fence *fences_user;
 	struct drm_amdgpu_fence *fences;
 	int r;
 
 	/* Get the fences from userspace */
-	fences = kmalloc_array(fence_count, sizeof(struct drm_amdgpu_fence),
-			GFP_KERNEL);
-	if (fences == NULL)
-		return -ENOMEM;
-
-	fences_user = u64_to_user_ptr(wait->in.fences);
-	if (copy_from_user(fences, fences_user,
-		sizeof(struct drm_amdgpu_fence) * fence_count)) {
-		r = -EFAULT;
-		goto err_free_fences;
-	}
+	fences = memdup_array_user(u64_to_user_ptr(wait->in.fences),
+				   wait->in.fence_count,
+				   sizeof(struct drm_amdgpu_fence));
+	if (IS_ERR(fences))
+		return PTR_ERR(fences);
 
 	if (wait->in.wait_all)
 		r = amdgpu_cs_wait_all_fences(adev, filp, wait, fences);
 	else
 		r = amdgpu_cs_wait_any_fence(adev, filp, wait, fences);
 
-err_free_fences:
 	kfree(fences);
 
 	return r;
@@ -1774,7 +1784,7 @@ int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser,
 	struct ttm_operation_ctx ctx = { false, false };
 	struct amdgpu_vm *vm = &fpriv->vm;
 	struct amdgpu_bo_va_mapping *mapping;
-	int r;
+	int i, r;
 
 	addr /= AMDGPU_GPU_PAGE_SIZE;
 
@@ -1786,12 +1796,17 @@ int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser,
 	*map = mapping;
 
 	/* Double check that the BO is reserved by this CS */
-	if (dma_resv_locking_ctx((*bo)->tbo.base.resv) != &parser->ticket)
+	if (dma_resv_locking_ctx((*bo)->tbo.base.resv) != &parser->exec.ticket)
 		return -EINVAL;
 
-	if (!((*bo)->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)) {
-		(*bo)->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
+	/* Make sure VRAM is allocated contigiously */
+	(*bo)->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
+	if ((*bo)->tbo.resource->mem_type == TTM_PL_VRAM &&
+	    !((*bo)->tbo.resource->placement & TTM_PL_FLAG_CONTIGUOUS)) {
+
 		amdgpu_bo_placement_from_domain(*bo, (*bo)->allowed_domains);
+		for (i = 0; i < (*bo)->placement.num_placement; i++)
+			(*bo)->placements[i].flags |= TTM_PL_FLAG_CONTIGUOUS;
 		r = ttm_bo_validate(&(*bo)->tbo, &(*bo)->placement, &ctx);
 		if (r)
 			return r;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.h
index fb3e3d56d427..39c33ad100cb 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.h
@@ -24,6 +24,7 @@
 #define __AMDGPU_CS_H__
 
 #include <linux/ww_mutex.h>
+#include <drm/drm_exec.h>
 
 #include "amdgpu_job.h"
 #include "amdgpu_bo_list.h"
@@ -62,11 +63,9 @@ struct amdgpu_cs_parser {
 	struct amdgpu_job	*gang_leader;
 
 	/* buffer objects */
-	struct ww_acquire_ctx		ticket;
+	struct drm_exec			exec;
 	struct amdgpu_bo_list		*bo_list;
 	struct amdgpu_mn		*mn;
-	struct amdgpu_bo_list_entry	vm_pd;
-	struct list_head		validated;
 	struct dma_fence		*fence;
 	uint64_t			bytes_moved_threshold;
 	uint64_t			bytes_moved_vis_threshold;
@@ -74,7 +73,7 @@ struct amdgpu_cs_parser {
 	uint64_t			bytes_moved_vis;
 
 	/* user fence */
-	struct amdgpu_bo_list_entry	uf_entry;
+	struct amdgpu_bo		*uf_bo;
 
 	unsigned			num_post_deps;
 	struct amdgpu_cs_post_dep	*post_deps;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c
index c6d4d41c4393..02138aa55793 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c
@@ -22,13 +22,14 @@
  * * Author: Monk.liu@amd.com
  */
 
+#include <drm/drm_exec.h>
+
 #include "amdgpu.h"
 
 uint64_t amdgpu_csa_vaddr(struct amdgpu_device *adev)
 {
-	uint64_t addr = adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT;
+	uint64_t addr = AMDGPU_VA_RESERVED_CSA_START(adev);
 
-	addr -= AMDGPU_VA_RESERVED_SIZE;
 	addr = amdgpu_gmc_sign_extend(addr);
 
 	return addr;
@@ -65,31 +66,25 @@ int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			  struct amdgpu_bo *bo, struct amdgpu_bo_va **bo_va,
 			  uint64_t csa_addr, uint32_t size)
 {
-	struct ww_acquire_ctx ticket;
-	struct list_head list;
-	struct amdgpu_bo_list_entry pd;
-	struct ttm_validate_buffer csa_tv;
+	struct drm_exec exec;
 	int r;
 
-	INIT_LIST_HEAD(&list);
-	INIT_LIST_HEAD(&csa_tv.head);
-	csa_tv.bo = &bo->tbo;
-	csa_tv.num_shared = 1;
-
-	list_add(&csa_tv.head, &list);
-	amdgpu_vm_get_pd_bo(vm, &list, &pd);
-
-	r = ttm_eu_reserve_buffers(&ticket, &list, true, NULL);
-	if (r) {
-		DRM_ERROR("failed to reserve CSA,PD BOs: err=%d\n", r);
-		return r;
+	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT, 0);
+	drm_exec_until_all_locked(&exec) {
+		r = amdgpu_vm_lock_pd(vm, &exec, 0);
+		if (likely(!r))
+			r = drm_exec_lock_obj(&exec, &bo->tbo.base);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(r)) {
+			DRM_ERROR("failed to reserve CSA,PD BOs: err=%d\n", r);
+			goto error;
+		}
 	}
 
 	*bo_va = amdgpu_vm_bo_add(adev, vm, bo);
 	if (!*bo_va) {
-		ttm_eu_backoff_reservation(&ticket, &list);
-		DRM_ERROR("failed to create bo_va for static CSA\n");
-		return -ENOMEM;
+		r = -ENOMEM;
+		goto error;
 	}
 
 	r = amdgpu_vm_bo_map(adev, *bo_va, csa_addr, 0, size,
@@ -99,10 +94,42 @@ int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	if (r) {
 		DRM_ERROR("failed to do bo_map on static CSA, err=%d\n", r);
 		amdgpu_vm_bo_del(adev, *bo_va);
-		ttm_eu_backoff_reservation(&ticket, &list);
-		return r;
+		goto error;
 	}
 
-	ttm_eu_backoff_reservation(&ticket, &list);
-	return 0;
+error:
+	drm_exec_fini(&exec);
+	return r;
+}
+
+int amdgpu_unmap_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+			    struct amdgpu_bo *bo, struct amdgpu_bo_va *bo_va,
+			    uint64_t csa_addr)
+{
+	struct drm_exec exec;
+	int r;
+
+	drm_exec_init(&exec, 0, 0);
+	drm_exec_until_all_locked(&exec) {
+		r = amdgpu_vm_lock_pd(vm, &exec, 0);
+		if (likely(!r))
+			r = drm_exec_lock_obj(&exec, &bo->tbo.base);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(r)) {
+			DRM_ERROR("failed to reserve CSA,PD BOs: err=%d\n", r);
+			goto error;
+		}
+	}
+
+	r = amdgpu_vm_bo_unmap(adev, bo_va, csa_addr);
+	if (r) {
+		DRM_ERROR("failed to do bo_unmap on static CSA, err=%d\n", r);
+		goto error;
+	}
+
+	amdgpu_vm_bo_del(adev, bo_va);
+
+error:
+	drm_exec_fini(&exec);
+	return r;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.h
index 524b4437a021..7dfc1f2012eb 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.h
@@ -34,6 +34,9 @@ int amdgpu_allocate_static_csa(struct amdgpu_device *adev, struct amdgpu_bo **bo
 int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			  struct amdgpu_bo *bo, struct amdgpu_bo_va **bo_va,
 			  uint64_t csa_addr, uint32_t size);
+int amdgpu_unmap_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+			    struct amdgpu_bo *bo, struct amdgpu_bo_va *bo_va,
+			    uint64_t csa_addr);
 void amdgpu_free_static_csa(struct amdgpu_bo **bo);
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
index d2139ac12159..afedea02188d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
@@ -42,12 +42,12 @@ const unsigned int amdgpu_ctx_num_entities[AMDGPU_HW_IP_NUM] = {
 	[AMDGPU_HW_IP_VCN_DEC]	=	1,
 	[AMDGPU_HW_IP_VCN_ENC]	=	1,
 	[AMDGPU_HW_IP_VCN_JPEG]	=	1,
+	[AMDGPU_HW_IP_VPE]	=	1,
 };
 
 bool amdgpu_ctx_priority_is_valid(int32_t ctx_prio)
 {
 	switch (ctx_prio) {
-	case AMDGPU_CTX_PRIORITY_UNSET:
 	case AMDGPU_CTX_PRIORITY_VERY_LOW:
 	case AMDGPU_CTX_PRIORITY_LOW:
 	case AMDGPU_CTX_PRIORITY_NORMAL:
@@ -55,6 +55,11 @@ bool amdgpu_ctx_priority_is_valid(int32_t ctx_prio)
 	case AMDGPU_CTX_PRIORITY_VERY_HIGH:
 		return true;
 	default:
+	case AMDGPU_CTX_PRIORITY_UNSET:
+		/* UNSET priority is not valid and we don't carry that
+		 * around, but set it to NORMAL in the only place this
+		 * function is called, amdgpu_ctx_ioctl().
+		 */
 		return false;
 	}
 }
@@ -64,13 +69,14 @@ amdgpu_ctx_to_drm_sched_prio(int32_t ctx_prio)
 {
 	switch (ctx_prio) {
 	case AMDGPU_CTX_PRIORITY_UNSET:
-		return DRM_SCHED_PRIORITY_UNSET;
+		pr_warn_once("AMD-->DRM context priority value UNSET-->NORMAL");
+		return DRM_SCHED_PRIORITY_NORMAL;
 
 	case AMDGPU_CTX_PRIORITY_VERY_LOW:
-		return DRM_SCHED_PRIORITY_MIN;
+		return DRM_SCHED_PRIORITY_LOW;
 
 	case AMDGPU_CTX_PRIORITY_LOW:
-		return DRM_SCHED_PRIORITY_MIN;
+		return DRM_SCHED_PRIORITY_LOW;
 
 	case AMDGPU_CTX_PRIORITY_NORMAL:
 		return DRM_SCHED_PRIORITY_NORMAL;
@@ -94,9 +100,6 @@ amdgpu_ctx_to_drm_sched_prio(int32_t ctx_prio)
 static int amdgpu_ctx_priority_permit(struct drm_file *filp,
 				      int32_t priority)
 {
-	if (!amdgpu_ctx_priority_is_valid(priority))
-		return -EINVAL;
-
 	/* NORMAL and below are accessible by everyone */
 	if (priority <= AMDGPU_CTX_PRIORITY_NORMAL)
 		return 0;
@@ -222,8 +225,19 @@ static int amdgpu_ctx_init_entity(struct amdgpu_ctx *ctx, u32 hw_ip,
 	drm_prio = amdgpu_ctx_to_drm_sched_prio(ctx_prio);
 
 	hw_ip = array_index_nospec(hw_ip, AMDGPU_HW_IP_NUM);
-	scheds = adev->gpu_sched[hw_ip][hw_prio].sched;
-	num_scheds = adev->gpu_sched[hw_ip][hw_prio].num_scheds;
+
+	if (!(adev)->xcp_mgr) {
+		scheds = adev->gpu_sched[hw_ip][hw_prio].sched;
+		num_scheds = adev->gpu_sched[hw_ip][hw_prio].num_scheds;
+	} else {
+		struct amdgpu_fpriv *fpriv;
+
+		fpriv = container_of(ctx->ctx_mgr, struct amdgpu_fpriv, ctx_mgr);
+		r = amdgpu_xcp_select_scheds(adev, hw_ip, hw_prio, fpriv,
+						&num_scheds, &scheds);
+		if (r)
+			goto error_free_entity;
+	}
 
 	/* disable load balance if the hw engine retains context among dependent jobs */
 	if (hw_ip == AMDGPU_HW_IP_VCN_ENC ||
@@ -255,7 +269,8 @@ error_free_entity:
 	return r;
 }
 
-static ktime_t amdgpu_ctx_fini_entity(struct amdgpu_ctx_entity *entity)
+static ktime_t amdgpu_ctx_fini_entity(struct amdgpu_device *adev,
+				  struct amdgpu_ctx_entity *entity)
 {
 	ktime_t res = ns_to_ktime(0);
 	int i;
@@ -268,6 +283,8 @@ static ktime_t amdgpu_ctx_fini_entity(struct amdgpu_ctx_entity *entity)
 		dma_fence_put(entity->fences[i]);
 	}
 
+	amdgpu_xcp_release_sched(adev, entity);
+
 	kfree(entity);
 	return res;
 }
@@ -303,6 +320,7 @@ static int amdgpu_ctx_get_stable_pstate(struct amdgpu_ctx *ctx,
 static int amdgpu_ctx_init(struct amdgpu_ctx_mgr *mgr, int32_t priority,
 			   struct drm_file *filp, struct amdgpu_ctx *ctx)
 {
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
 	u32 current_stable_pstate;
 	int r;
 
@@ -318,7 +336,7 @@ static int amdgpu_ctx_init(struct amdgpu_ctx_mgr *mgr, int32_t priority,
 
 	ctx->reset_counter = atomic_read(&mgr->adev->gpu_reset_counter);
 	ctx->reset_counter_query = ctx->reset_counter;
-	ctx->vram_lost_counter = atomic_read(&mgr->adev->vram_lost_counter);
+	ctx->generation = amdgpu_vm_generation(mgr->adev, &fpriv->vm);
 	ctx->init_priority = priority;
 	ctx->override_priority = AMDGPU_CTX_PRIORITY_UNSET;
 
@@ -331,6 +349,7 @@ static int amdgpu_ctx_init(struct amdgpu_ctx_mgr *mgr, int32_t priority,
 	else
 		ctx->stable_pstate = current_stable_pstate;
 
+	ctx->ctx_mgr = &(fpriv->ctx_mgr);
 	return 0;
 }
 
@@ -399,7 +418,7 @@ static void amdgpu_ctx_fini(struct kref *ref)
 		for (j = 0; j < AMDGPU_MAX_ENTITY_NUM; ++j) {
 			ktime_t spend;
 
-			spend = amdgpu_ctx_fini_entity(ctx->entities[i][j]);
+			spend = amdgpu_ctx_fini_entity(adev, ctx->entities[i][j]);
 			atomic64_add(ktime_to_ns(spend), &mgr->time_spend[i]);
 		}
 	}
@@ -416,6 +435,7 @@ int amdgpu_ctx_get_entity(struct amdgpu_ctx *ctx, u32 hw_ip, u32 instance,
 			  u32 ring, struct drm_sched_entity **entity)
 {
 	int r;
+	struct drm_sched_entity *ctx_entity;
 
 	if (hw_ip >= AMDGPU_HW_IP_NUM) {
 		DRM_ERROR("unknown HW IP type: %d\n", hw_ip);
@@ -439,7 +459,14 @@ int amdgpu_ctx_get_entity(struct amdgpu_ctx *ctx, u32 hw_ip, u32 instance,
 			return r;
 	}
 
-	*entity = &ctx->entities[hw_ip][ring]->entity;
+	ctx_entity = &ctx->entities[hw_ip][ring]->entity;
+	r = drm_sched_entity_error(ctx_entity);
+	if (r) {
+		DRM_DEBUG("error entity %p\n", ctx_entity);
+		return r;
+	}
+
+	*entity = ctx_entity;
 	return 0;
 }
 
@@ -570,12 +597,15 @@ static int amdgpu_ctx_query2(struct amdgpu_device *adev,
 	if (ctx->reset_counter != atomic_read(&adev->gpu_reset_counter))
 		out->state.flags |= AMDGPU_CTX_QUERY2_FLAGS_RESET;
 
-	if (ctx->vram_lost_counter != atomic_read(&adev->vram_lost_counter))
+	if (ctx->generation != amdgpu_vm_generation(adev, &fpriv->vm))
 		out->state.flags |= AMDGPU_CTX_QUERY2_FLAGS_VRAMLOST;
 
 	if (atomic_read(&ctx->guilty))
 		out->state.flags |= AMDGPU_CTX_QUERY2_FLAGS_GUILTY;
 
+	if (amdgpu_in_reset(adev))
+		out->state.flags |= AMDGPU_CTX_QUERY2_FLAGS_RESET_IN_PROGRESS;
+
 	if (adev->ras_enabled && con) {
 		/* Return the cached values in O(1),
 		 * and schedule delayed work to cache
@@ -604,8 +634,6 @@ static int amdgpu_ctx_query2(struct amdgpu_device *adev,
 	return 0;
 }
 
-
-
 static int amdgpu_ctx_stable_pstate(struct amdgpu_device *adev,
 				    struct amdgpu_fpriv *fpriv, uint32_t id,
 				    bool set, u32 *stable_pstate)
@@ -648,23 +676,33 @@ int amdgpu_ctx_ioctl(struct drm_device *dev, void *data,
 	id = args->in.ctx_id;
 	priority = args->in.priority;
 
-	/* For backwards compatibility reasons, we need to accept
-	 * ioctls with garbage in the priority field */
+	/* For backwards compatibility, we need to accept ioctls with garbage
+	 * in the priority field. Garbage values in the priority field, result
+	 * in the priority being set to NORMAL.
+	 */
 	if (!amdgpu_ctx_priority_is_valid(priority))
 		priority = AMDGPU_CTX_PRIORITY_NORMAL;
 
 	switch (args->in.op) {
 	case AMDGPU_CTX_OP_ALLOC_CTX:
+		if (args->in.flags)
+			return -EINVAL;
 		r = amdgpu_ctx_alloc(adev, fpriv, filp, priority, &id);
 		args->out.alloc.ctx_id = id;
 		break;
 	case AMDGPU_CTX_OP_FREE_CTX:
+		if (args->in.flags)
+			return -EINVAL;
 		r = amdgpu_ctx_free(fpriv, id);
 		break;
 	case AMDGPU_CTX_OP_QUERY_STATE:
+		if (args->in.flags)
+			return -EINVAL;
 		r = amdgpu_ctx_query(adev, fpriv, id, &args->out);
 		break;
 	case AMDGPU_CTX_OP_QUERY_STATE2:
+		if (args->in.flags)
+			return -EINVAL;
 		r = amdgpu_ctx_query2(adev, fpriv, id, &args->out);
 		break;
 	case AMDGPU_CTX_OP_GET_STABLE_PSTATE:
@@ -881,7 +919,7 @@ long amdgpu_ctx_mgr_entity_flush(struct amdgpu_ctx_mgr *mgr, long timeout)
 	return timeout;
 }
 
-void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr)
+static void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr)
 {
 	struct amdgpu_ctx *ctx;
 	struct idr *idp;
@@ -906,24 +944,13 @@ void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr)
 				drm_sched_entity_fini(entity);
 			}
 		}
+		kref_put(&ctx->refcount, amdgpu_ctx_fini);
 	}
 }
 
 void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr)
 {
-	struct amdgpu_ctx *ctx;
-	struct idr *idp;
-	uint32_t id;
-
 	amdgpu_ctx_mgr_entity_fini(mgr);
-
-	idp = &mgr->ctx_handles;
-
-	idr_for_each_entry(idp, ctx, id) {
-		if (kref_put(&ctx->refcount, amdgpu_ctx_fini) != 1)
-			DRM_ERROR("ctx %p is still alive\n", ctx);
-	}
-
 	idr_destroy(&mgr->ctx_handles);
 	mutex_destroy(&mgr->lock);
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
index 0fa0e56daf67..090dfe86f75b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
@@ -47,7 +47,7 @@ struct amdgpu_ctx {
 	struct amdgpu_ctx_mgr		*mgr;
 	unsigned			reset_counter;
 	unsigned			reset_counter_query;
-	uint32_t			vram_lost_counter;
+	uint64_t			generation;
 	spinlock_t			ring_lock;
 	struct amdgpu_ctx_entity	*entities[AMDGPU_HW_IP_NUM][AMDGPU_MAX_ENTITY_NUM];
 	bool				preamble_presented;
@@ -57,6 +57,7 @@ struct amdgpu_ctx {
 	unsigned long			ras_counter_ce;
 	unsigned long			ras_counter_ue;
 	uint32_t			stable_pstate;
+	struct amdgpu_ctx_mgr		*ctx_mgr;
 };
 
 struct amdgpu_ctx_mgr {
@@ -91,7 +92,6 @@ int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx,
 
 void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr,
 			 struct amdgpu_device *adev);
-void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr);
 long amdgpu_ctx_mgr_entity_flush(struct amdgpu_ctx_mgr *mgr, long timeout);
 void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr);
 void amdgpu_ctx_mgr_usage(struct amdgpu_ctx_mgr *mgr,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
index f60753f97ac5..62d43b8cbe58 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
@@ -56,14 +56,14 @@
  *
  * Bit 62:  Indicates a GRBM bank switch is needed
  * Bit 61:  Indicates a SRBM bank switch is needed (implies bit 62 is
- * 	    zero)
+ *	    zero)
  * Bits 24..33: The SE or ME selector if needed
  * Bits 34..43: The SH (or SA) or PIPE selector if needed
  * Bits 44..53: The INSTANCE (or CU/WGP) or QUEUE selector if needed
  *
  * Bit 23:  Indicates that the PM power gating lock should be held
- * 	    This is necessary to read registers that might be
- * 	    unreliable during a power gating transistion.
+ *	    This is necessary to read registers that might be
+ *	    unreliable during a power gating transistion.
  *
  * The lower bits are the BYTE offset of the register to read.  This
  * allows reading multiple registers in a single call and having
@@ -76,7 +76,7 @@ static int  amdgpu_debugfs_process_reg_op(bool read, struct file *f,
 	ssize_t result = 0;
 	int r;
 	bool pm_pg_lock, use_bank, use_ring;
-	unsigned instance_bank, sh_bank, se_bank, me, pipe, queue, vmid;
+	unsigned int instance_bank, sh_bank, se_bank, me, pipe, queue, vmid;
 
 	pm_pg_lock = use_bank = use_ring = false;
 	instance_bank = sh_bank = se_bank = me = pipe = queue = vmid = 0;
@@ -129,17 +129,16 @@ static int  amdgpu_debugfs_process_reg_op(bool read, struct file *f,
 	if (use_bank) {
 		if ((sh_bank != 0xFFFFFFFF && sh_bank >= adev->gfx.config.max_sh_per_se) ||
 		    (se_bank != 0xFFFFFFFF && se_bank >= adev->gfx.config.max_shader_engines)) {
-			pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 			pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 			amdgpu_virt_disable_access_debugfs(adev);
 			return -EINVAL;
 		}
 		mutex_lock(&adev->grbm_idx_mutex);
 		amdgpu_gfx_select_se_sh(adev, se_bank,
-					sh_bank, instance_bank);
+					sh_bank, instance_bank, 0);
 	} else if (use_ring) {
 		mutex_lock(&adev->srbm_mutex);
-		amdgpu_gfx_select_me_pipe_q(adev, me, pipe, queue, vmid);
+		amdgpu_gfx_select_me_pipe_q(adev, me, pipe, queue, vmid, 0);
 	}
 
 	if (pm_pg_lock)
@@ -154,7 +153,7 @@ static int  amdgpu_debugfs_process_reg_op(bool read, struct file *f,
 		} else {
 			r = get_user(value, (uint32_t *)buf);
 			if (!r)
-				amdgpu_mm_wreg_mmio_rlc(adev, *pos >> 2, value);
+				amdgpu_mm_wreg_mmio_rlc(adev, *pos >> 2, value, 0);
 		}
 		if (r) {
 			result = r;
@@ -169,17 +168,16 @@ static int  amdgpu_debugfs_process_reg_op(bool read, struct file *f,
 
 end:
 	if (use_bank) {
-		amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+		amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 		mutex_unlock(&adev->grbm_idx_mutex);
 	} else if (use_ring) {
-		amdgpu_gfx_select_me_pipe_q(adev, 0, 0, 0, 0);
+		amdgpu_gfx_select_me_pipe_q(adev, 0, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 	}
 
 	if (pm_pg_lock)
 		mutex_unlock(&adev->pm.mutex);
 
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	amdgpu_virt_disable_access_debugfs(adev);
@@ -208,7 +206,7 @@ static int amdgpu_debugfs_regs2_open(struct inode *inode, struct file *file)
 {
 	struct amdgpu_debugfs_regs2_data *rd;
 
-	rd = kzalloc(sizeof *rd, GFP_KERNEL);
+	rd = kzalloc(sizeof(*rd), GFP_KERNEL);
 	if (!rd)
 		return -ENOMEM;
 	rd->adev = file_inode(file)->i_private;
@@ -221,6 +219,7 @@ static int amdgpu_debugfs_regs2_open(struct inode *inode, struct file *file)
 static int amdgpu_debugfs_regs2_release(struct inode *inode, struct file *file)
 {
 	struct amdgpu_debugfs_regs2_data *rd = file->private_data;
+
 	mutex_destroy(&rd->lock);
 	kfree(file->private_data);
 	return 0;
@@ -254,7 +253,6 @@ static ssize_t amdgpu_debugfs_regs2_op(struct file *f, char __user *buf, u32 off
 	if (rd->id.use_grbm) {
 		if ((rd->id.grbm.sh != 0xFFFFFFFF && rd->id.grbm.sh >= adev->gfx.config.max_sh_per_se) ||
 		    (rd->id.grbm.se != 0xFFFFFFFF && rd->id.grbm.se >= adev->gfx.config.max_shader_engines)) {
-			pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 			pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 			amdgpu_virt_disable_access_debugfs(adev);
 			mutex_unlock(&rd->lock);
@@ -262,14 +260,14 @@ static ssize_t amdgpu_debugfs_regs2_op(struct file *f, char __user *buf, u32 off
 		}
 		mutex_lock(&adev->grbm_idx_mutex);
 		amdgpu_gfx_select_se_sh(adev, rd->id.grbm.se,
-								rd->id.grbm.sh,
-								rd->id.grbm.instance);
+						  rd->id.grbm.sh,
+						  rd->id.grbm.instance, rd->id.xcc_id);
 	}
 
 	if (rd->id.use_srbm) {
 		mutex_lock(&adev->srbm_mutex);
 		amdgpu_gfx_select_me_pipe_q(adev, rd->id.srbm.me, rd->id.srbm.pipe,
-									rd->id.srbm.queue, rd->id.srbm.vmid);
+					    rd->id.srbm.queue, rd->id.srbm.vmid, rd->id.xcc_id);
 	}
 
 	if (rd->id.pg_lock)
@@ -282,7 +280,7 @@ static ssize_t amdgpu_debugfs_regs2_op(struct file *f, char __user *buf, u32 off
 		} else {
 			r = get_user(value, (uint32_t *)buf);
 			if (!r)
-				amdgpu_mm_wreg_mmio_rlc(adev, offset >> 2, value);
+				amdgpu_mm_wreg_mmio_rlc(adev, offset >> 2, value, rd->id.xcc_id);
 		}
 		if (r) {
 			result = r;
@@ -295,12 +293,12 @@ static ssize_t amdgpu_debugfs_regs2_op(struct file *f, char __user *buf, u32 off
 	}
 end:
 	if (rd->id.use_grbm) {
-		amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+		amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, rd->id.xcc_id);
 		mutex_unlock(&adev->grbm_idx_mutex);
 	}
 
 	if (rd->id.use_srbm) {
-		amdgpu_gfx_select_me_pipe_q(adev, 0, 0, 0, 0);
+		amdgpu_gfx_select_me_pipe_q(adev, 0, 0, 0, 0, rd->id.xcc_id);
 		mutex_unlock(&adev->srbm_mutex);
 	}
 
@@ -309,7 +307,6 @@ end:
 
 	mutex_unlock(&rd->lock);
 
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	amdgpu_virt_disable_access_debugfs(adev);
@@ -319,18 +316,45 @@ end:
 static long amdgpu_debugfs_regs2_ioctl(struct file *f, unsigned int cmd, unsigned long data)
 {
 	struct amdgpu_debugfs_regs2_data *rd = f->private_data;
+	struct amdgpu_debugfs_regs2_iocdata v1_data;
 	int r;
 
+	mutex_lock(&rd->lock);
+
 	switch (cmd) {
+	case AMDGPU_DEBUGFS_REGS2_IOC_SET_STATE_V2:
+		r = copy_from_user(&rd->id, (struct amdgpu_debugfs_regs2_iocdata_v2 *)data,
+				   sizeof(rd->id));
+		if (r)
+			r = -EINVAL;
+		goto done;
 	case AMDGPU_DEBUGFS_REGS2_IOC_SET_STATE:
-		mutex_lock(&rd->lock);
-		r = copy_from_user(&rd->id, (struct amdgpu_debugfs_regs2_iocdata *)data, sizeof rd->id);
-		mutex_unlock(&rd->lock);
-		return r ? -EINVAL : 0;
+		r = copy_from_user(&v1_data, (struct amdgpu_debugfs_regs2_iocdata *)data,
+				   sizeof(v1_data));
+		if (r) {
+			r = -EINVAL;
+			goto done;
+		}
+		goto v1_copy;
 	default:
-		return -EINVAL;
-	}
-	return 0;
+		r = -EINVAL;
+		goto done;
+	}
+
+v1_copy:
+	rd->id.use_srbm = v1_data.use_srbm;
+	rd->id.use_grbm = v1_data.use_grbm;
+	rd->id.pg_lock = v1_data.pg_lock;
+	rd->id.grbm.se = v1_data.grbm.se;
+	rd->id.grbm.sh = v1_data.grbm.sh;
+	rd->id.grbm.instance = v1_data.grbm.instance;
+	rd->id.srbm.me = v1_data.srbm.me;
+	rd->id.srbm.pipe = v1_data.srbm.pipe;
+	rd->id.srbm.queue = v1_data.srbm.queue;
+	rd->id.xcc_id = 0;
+done:
+	mutex_unlock(&rd->lock);
+	return r;
 }
 
 static ssize_t amdgpu_debugfs_regs2_read(struct file *f, char __user *buf, size_t size, loff_t *pos)
@@ -343,6 +367,136 @@ static ssize_t amdgpu_debugfs_regs2_write(struct file *f, const char __user *buf
 	return amdgpu_debugfs_regs2_op(f, (char __user *)buf, *pos, size, 1);
 }
 
+static int amdgpu_debugfs_gprwave_open(struct inode *inode, struct file *file)
+{
+	struct amdgpu_debugfs_gprwave_data *rd;
+
+	rd = kzalloc(sizeof(*rd), GFP_KERNEL);
+	if (!rd)
+		return -ENOMEM;
+	rd->adev = file_inode(file)->i_private;
+	file->private_data = rd;
+	mutex_init(&rd->lock);
+
+	return 0;
+}
+
+static int amdgpu_debugfs_gprwave_release(struct inode *inode, struct file *file)
+{
+	struct amdgpu_debugfs_gprwave_data *rd = file->private_data;
+
+	mutex_destroy(&rd->lock);
+	kfree(file->private_data);
+	return 0;
+}
+
+static ssize_t amdgpu_debugfs_gprwave_read(struct file *f, char __user *buf, size_t size, loff_t *pos)
+{
+	struct amdgpu_debugfs_gprwave_data *rd = f->private_data;
+	struct amdgpu_device *adev = rd->adev;
+	ssize_t result = 0;
+	int r;
+	uint32_t *data, x;
+
+	if (size > 4096 || size & 0x3 || *pos & 0x3)
+		return -EINVAL;
+
+	r = pm_runtime_get_sync(adev_to_drm(adev)->dev);
+	if (r < 0) {
+		pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
+		return r;
+	}
+
+	r = amdgpu_virt_enable_access_debugfs(adev);
+	if (r < 0) {
+		pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
+		return r;
+	}
+
+	data = kcalloc(1024, sizeof(*data), GFP_KERNEL);
+	if (!data) {
+		pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
+		amdgpu_virt_disable_access_debugfs(adev);
+		return -ENOMEM;
+	}
+
+	/* switch to the specific se/sh/cu */
+	mutex_lock(&adev->grbm_idx_mutex);
+	amdgpu_gfx_select_se_sh(adev, rd->id.se, rd->id.sh, rd->id.cu, rd->id.xcc_id);
+
+	if (!rd->id.gpr_or_wave) {
+		x = 0;
+		if (adev->gfx.funcs->read_wave_data)
+			adev->gfx.funcs->read_wave_data(adev, rd->id.xcc_id, rd->id.simd, rd->id.wave, data, &x);
+	} else {
+		x = size >> 2;
+		if (rd->id.gpr.vpgr_or_sgpr) {
+			if (adev->gfx.funcs->read_wave_vgprs)
+				adev->gfx.funcs->read_wave_vgprs(adev, rd->id.xcc_id, rd->id.simd, rd->id.wave, rd->id.gpr.thread, *pos, size>>2, data);
+		} else {
+			if (adev->gfx.funcs->read_wave_sgprs)
+				adev->gfx.funcs->read_wave_sgprs(adev, rd->id.xcc_id, rd->id.simd, rd->id.wave, *pos, size>>2, data);
+		}
+	}
+
+	amdgpu_gfx_select_se_sh(adev, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, rd->id.xcc_id);
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
+
+	if (!x) {
+		result = -EINVAL;
+		goto done;
+	}
+
+	while (size && (*pos < x * 4)) {
+		uint32_t value;
+
+		value = data[*pos >> 2];
+		r = put_user(value, (uint32_t *)buf);
+		if (r) {
+			result = r;
+			goto done;
+		}
+
+		result += 4;
+		buf += 4;
+		*pos += 4;
+		size -= 4;
+	}
+
+done:
+	amdgpu_virt_disable_access_debugfs(adev);
+	kfree(data);
+	return result;
+}
+
+static long amdgpu_debugfs_gprwave_ioctl(struct file *f, unsigned int cmd, unsigned long data)
+{
+	struct amdgpu_debugfs_gprwave_data *rd = f->private_data;
+	int r = 0;
+
+	mutex_lock(&rd->lock);
+
+	switch (cmd) {
+	case AMDGPU_DEBUGFS_GPRWAVE_IOC_SET_STATE:
+		if (copy_from_user(&rd->id,
+				   (struct amdgpu_debugfs_gprwave_iocdata *)data,
+				   sizeof(rd->id)))
+			r = -EFAULT;
+		goto done;
+	default:
+		r = -EINVAL;
+		goto done;
+	}
+
+done:
+	mutex_unlock(&rd->lock);
+	return r;
+}
+
+
+
 
 /**
  * amdgpu_debugfs_regs_pcie_read - Read from a PCIE register
@@ -381,7 +535,11 @@ static ssize_t amdgpu_debugfs_regs_pcie_read(struct file *f, char __user *buf,
 	while (size) {
 		uint32_t value;
 
-		value = RREG32_PCIE(*pos);
+		if (upper_32_bits(*pos))
+			value = RREG32_PCIE_EXT(*pos);
+		else
+			value = RREG32_PCIE(*pos);
+
 		r = put_user(value, (uint32_t *)buf);
 		if (r)
 			goto out;
@@ -394,7 +552,6 @@ static ssize_t amdgpu_debugfs_regs_pcie_read(struct file *f, char __user *buf,
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 	amdgpu_virt_disable_access_debugfs(adev);
 	return r;
@@ -441,7 +598,10 @@ static ssize_t amdgpu_debugfs_regs_pcie_write(struct file *f, const char __user
 		if (r)
 			goto out;
 
-		WREG32_PCIE(*pos, value);
+		if (upper_32_bits(*pos))
+			WREG32_PCIE_EXT(*pos, value);
+		else
+			WREG32_PCIE(*pos, value);
 
 		result += 4;
 		buf += 4;
@@ -451,7 +611,6 @@ static ssize_t amdgpu_debugfs_regs_pcie_write(struct file *f, const char __user
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 	amdgpu_virt_disable_access_debugfs(adev);
 	return r;
@@ -479,6 +638,9 @@ static ssize_t amdgpu_debugfs_regs_didt_read(struct file *f, char __user *buf,
 	if (size & 0x3 || *pos & 0x3)
 		return -EINVAL;
 
+	if (!adev->didt_rreg)
+		return -EOPNOTSUPP;
+
 	r = pm_runtime_get_sync(adev_to_drm(adev)->dev);
 	if (r < 0) {
 		pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
@@ -507,7 +669,6 @@ static ssize_t amdgpu_debugfs_regs_didt_read(struct file *f, char __user *buf,
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 	amdgpu_virt_disable_access_debugfs(adev);
 	return r;
@@ -535,6 +696,9 @@ static ssize_t amdgpu_debugfs_regs_didt_write(struct file *f, const char __user
 	if (size & 0x3 || *pos & 0x3)
 		return -EINVAL;
 
+	if (!adev->didt_wreg)
+		return -EOPNOTSUPP;
+
 	r = pm_runtime_get_sync(adev_to_drm(adev)->dev);
 	if (r < 0) {
 		pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
@@ -564,7 +728,6 @@ static ssize_t amdgpu_debugfs_regs_didt_write(struct file *f, const char __user
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 	amdgpu_virt_disable_access_debugfs(adev);
 	return r;
@@ -589,6 +752,9 @@ static ssize_t amdgpu_debugfs_regs_smc_read(struct file *f, char __user *buf,
 	ssize_t result = 0;
 	int r;
 
+	if (!adev->smc_rreg)
+		return -EOPNOTSUPP;
+
 	if (size & 0x3 || *pos & 0x3)
 		return -EINVAL;
 
@@ -620,7 +786,6 @@ static ssize_t amdgpu_debugfs_regs_smc_read(struct file *f, char __user *buf,
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 	amdgpu_virt_disable_access_debugfs(adev);
 	return r;
@@ -645,6 +810,9 @@ static ssize_t amdgpu_debugfs_regs_smc_write(struct file *f, const char __user *
 	ssize_t result = 0;
 	int r;
 
+	if (!adev->smc_wreg)
+		return -EOPNOTSUPP;
+
 	if (size & 0x3 || *pos & 0x3)
 		return -EINVAL;
 
@@ -677,7 +845,6 @@ static ssize_t amdgpu_debugfs_regs_smc_write(struct file *f, const char __user *
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 	amdgpu_virt_disable_access_debugfs(adev);
 	return r;
@@ -740,7 +907,7 @@ static ssize_t amdgpu_debugfs_gca_config_read(struct file *f, char __user *buf,
 
 	/* rev==1 */
 	config[no_regs++] = adev->rev_id;
-	config[no_regs++] = lower_32_bits(adev->pg_flags);
+	config[no_regs++] = adev->pg_flags;
 	config[no_regs++] = lower_32_bits(adev->cg_flags);
 
 	/* rev==2 */
@@ -757,7 +924,7 @@ static ssize_t amdgpu_debugfs_gca_config_read(struct file *f, char __user *buf,
 	config[no_regs++] = adev->flags & AMD_IS_APU ? 1 : 0;
 
 	/* rev==5 PG/CG flag upper 32bit */
-	config[no_regs++] = upper_32_bits(adev->pg_flags);
+	config[no_regs++] = 0;
 	config[no_regs++] = upper_32_bits(adev->cg_flags);
 
 	while (size && (*pos < no_regs * 4)) {
@@ -825,7 +992,6 @@ static ssize_t amdgpu_debugfs_sensor_read(struct file *f, char __user *buf,
 
 	r = amdgpu_dpm_read_sensor(adev, idx, &values[0], &valuesize);
 
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	if (r) {
@@ -863,7 +1029,7 @@ static ssize_t amdgpu_debugfs_sensor_read(struct file *f, char __user *buf,
  * The offset being sought changes which wave that the status data
  * will be returned for.  The bits are used as follows:
  *
- * Bits 0..6: 	Byte offset into data
+ * Bits 0..6:	Byte offset into data
  * Bits 7..14:	SE selector
  * Bits 15..22:	SH/SA selector
  * Bits 23..30: CU/{WGP+SIMD} selector
@@ -907,16 +1073,15 @@ static ssize_t amdgpu_debugfs_wave_read(struct file *f, char __user *buf,
 
 	/* switch to the specific se/sh/cu */
 	mutex_lock(&adev->grbm_idx_mutex);
-	amdgpu_gfx_select_se_sh(adev, se, sh, cu);
+	amdgpu_gfx_select_se_sh(adev, se, sh, cu, 0);
 
 	x = 0;
 	if (adev->gfx.funcs->read_wave_data)
-		adev->gfx.funcs->read_wave_data(adev, simd, wave, data, &x);
+		adev->gfx.funcs->read_wave_data(adev, 0, simd, wave, data, &x);
 
-	amdgpu_gfx_select_se_sh(adev, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF);
+	amdgpu_gfx_select_se_sh(adev, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	if (!x) {
@@ -1001,20 +1166,19 @@ static ssize_t amdgpu_debugfs_gpr_read(struct file *f, char __user *buf,
 
 	/* switch to the specific se/sh/cu */
 	mutex_lock(&adev->grbm_idx_mutex);
-	amdgpu_gfx_select_se_sh(adev, se, sh, cu);
+	amdgpu_gfx_select_se_sh(adev, se, sh, cu, 0);
 
 	if (bank == 0) {
 		if (adev->gfx.funcs->read_wave_vgprs)
-			adev->gfx.funcs->read_wave_vgprs(adev, simd, wave, thread, offset, size>>2, data);
+			adev->gfx.funcs->read_wave_vgprs(adev, 0, simd, wave, thread, offset, size>>2, data);
 	} else {
 		if (adev->gfx.funcs->read_wave_sgprs)
-			adev->gfx.funcs->read_wave_sgprs(adev, simd, wave, offset, size>>2, data);
+			adev->gfx.funcs->read_wave_sgprs(adev, 0, simd, wave, offset, size>>2, data);
 	}
 
-	amdgpu_gfx_select_se_sh(adev, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF);
+	amdgpu_gfx_select_se_sh(adev, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	while (size) {
@@ -1088,7 +1252,6 @@ static ssize_t amdgpu_debugfs_gfxoff_residency_read(struct file *f, char __user
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	return r;
@@ -1137,7 +1300,6 @@ static ssize_t amdgpu_debugfs_gfxoff_residency_write(struct file *f, const char
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	return r;
@@ -1187,7 +1349,6 @@ static ssize_t amdgpu_debugfs_gfxoff_count_read(struct file *f, char __user *buf
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	return r;
@@ -1236,7 +1397,6 @@ static ssize_t amdgpu_debugfs_gfxoff_write(struct file *f, const char __user *bu
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	return r;
@@ -1282,7 +1442,6 @@ static ssize_t amdgpu_debugfs_gfxoff_read(struct file *f, char __user *buf,
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	return r;
@@ -1323,7 +1482,6 @@ static ssize_t amdgpu_debugfs_gfxoff_status_read(struct file *f, char __user *bu
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	return r;
@@ -1339,6 +1497,15 @@ static const struct file_operations amdgpu_debugfs_regs2_fops = {
 	.llseek = default_llseek
 };
 
+static const struct file_operations amdgpu_debugfs_gprwave_fops = {
+	.owner = THIS_MODULE,
+	.unlocked_ioctl = amdgpu_debugfs_gprwave_ioctl,
+	.read = amdgpu_debugfs_gprwave_read,
+	.open = amdgpu_debugfs_gprwave_open,
+	.release = amdgpu_debugfs_gprwave_release,
+	.llseek = default_llseek
+};
+
 static const struct file_operations amdgpu_debugfs_regs_fops = {
 	.owner = THIS_MODULE,
 	.read = amdgpu_debugfs_regs_read,
@@ -1416,6 +1583,7 @@ static const struct file_operations amdgpu_debugfs_gfxoff_residency_fops = {
 static const struct file_operations *debugfs_regs[] = {
 	&amdgpu_debugfs_regs_fops,
 	&amdgpu_debugfs_regs2_fops,
+	&amdgpu_debugfs_gprwave_fops,
 	&amdgpu_debugfs_regs_didt_fops,
 	&amdgpu_debugfs_regs_pcie_fops,
 	&amdgpu_debugfs_regs_smc_fops,
@@ -1429,9 +1597,10 @@ static const struct file_operations *debugfs_regs[] = {
 	&amdgpu_debugfs_gfxoff_residency_fops,
 };
 
-static const char *debugfs_regs_names[] = {
+static const char * const debugfs_regs_names[] = {
 	"amdgpu_regs",
 	"amdgpu_regs2",
+	"amdgpu_gprwave",
 	"amdgpu_regs_didt",
 	"amdgpu_regs_pcie",
 	"amdgpu_regs_smc",
@@ -1447,7 +1616,7 @@ static const char *debugfs_regs_names[] = {
 
 /**
  * amdgpu_debugfs_regs_init -	Initialize debugfs entries that provide
- * 				register access.
+ *				register access.
  *
  * @adev: The device to attach the debugfs entries to
  */
@@ -1459,7 +1628,7 @@ int amdgpu_debugfs_regs_init(struct amdgpu_device *adev)
 
 	for (i = 0; i < ARRAY_SIZE(debugfs_regs); i++) {
 		ent = debugfs_create_file(debugfs_regs_names[i],
-					  S_IFREG | S_IRUGO, root,
+					  S_IFREG | 0400, root,
 					  adev, debugfs_regs[i]);
 		if (!i && !IS_ERR_OR_NULL(ent))
 			i_size_write(ent->d_inode, adev->rmmio_size);
@@ -1470,7 +1639,7 @@ int amdgpu_debugfs_regs_init(struct amdgpu_device *adev)
 
 static int amdgpu_debugfs_test_ib_show(struct seq_file *m, void *unused)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct amdgpu_device *adev = m->private;
 	struct drm_device *dev = adev_to_drm(adev);
 	int r = 0, i;
 
@@ -1489,30 +1658,29 @@ static int amdgpu_debugfs_test_ib_show(struct seq_file *m, void *unused)
 	for (i = 0; i < AMDGPU_MAX_RINGS; i++) {
 		struct amdgpu_ring *ring = adev->rings[i];
 
-		if (!ring || !ring->sched.thread)
+		if (!amdgpu_ring_sched_ready(ring))
 			continue;
-		kthread_park(ring->sched.thread);
+		drm_sched_wqueue_stop(&ring->sched);
 	}
 
-	seq_printf(m, "run ib test:\n");
+	seq_puts(m, "run ib test:\n");
 	r = amdgpu_ib_ring_tests(adev);
 	if (r)
 		seq_printf(m, "ib ring tests failed (%d).\n", r);
 	else
-		seq_printf(m, "ib ring tests passed.\n");
+		seq_puts(m, "ib ring tests passed.\n");
 
 	/* go on the scheduler */
 	for (i = 0; i < AMDGPU_MAX_RINGS; i++) {
 		struct amdgpu_ring *ring = adev->rings[i];
 
-		if (!ring || !ring->sched.thread)
+		if (!amdgpu_ring_sched_ready(ring))
 			continue;
-		kthread_unpark(ring->sched.thread);
+		drm_sched_wqueue_start(&ring->sched);
 	}
 
 	up_write(&adev->reset_domain->sem);
 
-	pm_runtime_mark_last_busy(dev->dev);
 	pm_runtime_put_autosuspend(dev->dev);
 
 	return 0;
@@ -1532,7 +1700,6 @@ static int amdgpu_debugfs_evict_vram(void *data, u64 *val)
 
 	*val = amdgpu_ttm_evict_resources(adev, TTM_PL_VRAM);
 
-	pm_runtime_mark_last_busy(dev->dev);
 	pm_runtime_put_autosuspend(dev->dev);
 
 	return 0;
@@ -1553,7 +1720,6 @@ static int amdgpu_debugfs_evict_gtt(void *data, u64 *val)
 
 	*val = amdgpu_ttm_evict_resources(adev, TTM_PL_TT);
 
-	pm_runtime_mark_last_busy(dev->dev);
 	pm_runtime_put_autosuspend(dev->dev);
 
 	return 0;
@@ -1573,7 +1739,6 @@ static int amdgpu_debugfs_benchmark(void *data, u64 val)
 
 	r = amdgpu_benchmark(adev, val);
 
-	pm_runtime_mark_last_busy(dev->dev);
 	pm_runtime_put_autosuspend(dev->dev);
 
 	return r;
@@ -1581,7 +1746,7 @@ static int amdgpu_debugfs_benchmark(void *data, u64 val)
 
 static int amdgpu_debugfs_vm_info_show(struct seq_file *m, void *unused)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct amdgpu_device *adev = m->private;
 	struct drm_device *dev = adev_to_drm(adev);
 	struct drm_file *file;
 	int r;
@@ -1593,9 +1758,14 @@ static int amdgpu_debugfs_vm_info_show(struct seq_file *m, void *unused)
 	list_for_each_entry(file, &dev->filelist, lhead) {
 		struct amdgpu_fpriv *fpriv = file->driver_priv;
 		struct amdgpu_vm *vm = &fpriv->vm;
+		struct amdgpu_task_info *ti;
+
+		ti = amdgpu_vm_get_task_info_vm(vm);
+		if (ti) {
+			seq_printf(m, "pid:%d\tProcess:%s ----------\n", ti->task.pid, ti->process_name);
+			amdgpu_vm_put_task_info(ti);
+		}
 
-		seq_printf(m, "pid:%d\tProcess:%s ----------\n",
-				vm->task_info.pid, vm->task_info.process_name);
 		r = amdgpu_bo_reserve(vm->root.bo, true);
 		if (r)
 			break;
@@ -1708,7 +1878,7 @@ no_preempt:
 			continue;
 		}
 		job = to_amdgpu_job(s_job);
-		if (preempted && (&job->hw_fence) == fence)
+		if (preempted && (&job->hw_fence->base) == fence)
 			/* mark the job as preempted */
 			job->preemption_status |= AMDGPU_IB_PREEMPTED;
 	}
@@ -1727,7 +1897,8 @@ static int amdgpu_debugfs_ib_preempt(void *data, u64 val)
 
 	ring = adev->rings[val];
 
-	if (!ring || !ring->funcs->preempt_ib || !ring->sched.thread)
+	if (!amdgpu_ring_sched_ready(ring) ||
+	    !ring->funcs->preempt_ib)
 		return -EINVAL;
 
 	/* the last preemption failed */
@@ -1745,7 +1916,7 @@ static int amdgpu_debugfs_ib_preempt(void *data, u64 val)
 		goto pro_end;
 
 	/* stop the scheduler */
-	kthread_park(ring->sched.thread);
+	drm_sched_wqueue_stop(&ring->sched);
 
 	/* preempt the IB */
 	r = amdgpu_ring_preempt_ib(ring);
@@ -1779,7 +1950,7 @@ static int amdgpu_debugfs_ib_preempt(void *data, u64 val)
 
 failure:
 	/* restart the scheduler */
-	kthread_unpark(ring->sched.thread);
+	drm_sched_wqueue_start(&ring->sched);
 
 	up_read(&adev->reset_domain->sem);
 
@@ -1795,7 +1966,7 @@ static int amdgpu_debugfs_sclk_set(void *data, u64 val)
 	uint32_t max_freq, min_freq;
 	struct amdgpu_device *adev = (struct amdgpu_device *)data;
 
-	if (amdgpu_sriov_vf(adev) && !amdgpu_sriov_is_pp_one_vf(adev))
+	if (amdgpu_sriov_multi_vf_mode(adev))
 		return -EINVAL;
 
 	ret = pm_runtime_get_sync(adev_to_drm(adev)->dev);
@@ -1819,7 +1990,6 @@ static int amdgpu_debugfs_sclk_set(void *data, u64 val)
 		ret = -EINVAL;
 
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	return ret;
@@ -1831,99 +2001,6 @@ DEFINE_DEBUGFS_ATTRIBUTE(fops_ib_preempt, NULL,
 DEFINE_DEBUGFS_ATTRIBUTE(fops_sclk_set, NULL,
 			amdgpu_debugfs_sclk_set, "%llu\n");
 
-static ssize_t amdgpu_reset_dump_register_list_read(struct file *f,
-				char __user *buf, size_t size, loff_t *pos)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(f)->i_private;
-	char reg_offset[12];
-	int i, ret, len = 0;
-
-	if (*pos)
-		return 0;
-
-	memset(reg_offset, 0, 12);
-	ret = down_read_killable(&adev->reset_domain->sem);
-	if (ret)
-		return ret;
-
-	for (i = 0; i < adev->num_regs; i++) {
-		sprintf(reg_offset, "0x%x\n", adev->reset_dump_reg_list[i]);
-		up_read(&adev->reset_domain->sem);
-		if (copy_to_user(buf + len, reg_offset, strlen(reg_offset)))
-			return -EFAULT;
-
-		len += strlen(reg_offset);
-		ret = down_read_killable(&adev->reset_domain->sem);
-		if (ret)
-			return ret;
-	}
-
-	up_read(&adev->reset_domain->sem);
-	*pos += len;
-
-	return len;
-}
-
-static ssize_t amdgpu_reset_dump_register_list_write(struct file *f,
-			const char __user *buf, size_t size, loff_t *pos)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(f)->i_private;
-	char reg_offset[11];
-	uint32_t *new = NULL, *tmp = NULL;
-	int ret, i = 0, len = 0;
-
-	do {
-		memset(reg_offset, 0, 11);
-		if (copy_from_user(reg_offset, buf + len,
-					min(10, ((int)size-len)))) {
-			ret = -EFAULT;
-			goto error_free;
-		}
-
-		new = krealloc_array(tmp, i + 1, sizeof(uint32_t), GFP_KERNEL);
-		if (!new) {
-			ret = -ENOMEM;
-			goto error_free;
-		}
-		tmp = new;
-		if (sscanf(reg_offset, "%X %n", &tmp[i], &ret) != 1) {
-			ret = -EINVAL;
-			goto error_free;
-		}
-
-		len += ret;
-		i++;
-	} while (len < size);
-
-	new = kmalloc_array(i, sizeof(uint32_t), GFP_KERNEL);
-	if (!new) {
-		ret = -ENOMEM;
-		goto error_free;
-	}
-	ret = down_write_killable(&adev->reset_domain->sem);
-	if (ret)
-		goto error_free;
-
-	swap(adev->reset_dump_reg_list, tmp);
-	swap(adev->reset_dump_reg_value, new);
-	adev->num_regs = i;
-	up_write(&adev->reset_domain->sem);
-	ret = size;
-
-error_free:
-	if (tmp != new)
-		kfree(tmp);
-	kfree(new);
-	return ret;
-}
-
-static const struct file_operations amdgpu_reset_dump_register_list = {
-	.owner = THIS_MODULE,
-	.read = amdgpu_reset_dump_register_list_read,
-	.write = amdgpu_reset_dump_register_list_write,
-	.llseek = default_llseek
-};
-
 int amdgpu_debugfs_init(struct amdgpu_device *adev)
 {
 	struct dentry *root = adev_to_drm(adev)->primary->debugfs_root;
@@ -1964,6 +2041,8 @@ int amdgpu_debugfs_init(struct amdgpu_device *adev)
 	amdgpu_debugfs_firmware_init(adev);
 	amdgpu_ta_if_debugfs_init(adev);
 
+	amdgpu_debugfs_mes_event_log_init(adev);
+
 #if defined(CONFIG_DRM_AMD_DC)
 	if (adev->dc_enabled)
 		dtn_debugfs_init(adev);
@@ -1978,7 +2057,7 @@ int amdgpu_debugfs_init(struct amdgpu_device *adev)
 		amdgpu_debugfs_ring_init(adev, ring);
 	}
 
-	for ( i = 0; i < adev->vcn.num_vcn_inst; i++) {
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
 		if (!amdgpu_vcnfw_log)
 			break;
 
@@ -1988,37 +2067,99 @@ int amdgpu_debugfs_init(struct amdgpu_device *adev)
 		amdgpu_debugfs_vcn_fwlog_init(adev, i, &adev->vcn.inst[i]);
 	}
 
+	if (amdgpu_umsch_mm & amdgpu_umsch_mm_fwlog)
+		amdgpu_debugfs_umsch_fwlog_init(adev, &adev->umsch_mm);
+
+	amdgpu_debugfs_vcn_sched_mask_init(adev);
+	amdgpu_debugfs_jpeg_sched_mask_init(adev);
+	amdgpu_debugfs_gfx_sched_mask_init(adev);
+	amdgpu_debugfs_compute_sched_mask_init(adev);
+	amdgpu_debugfs_sdma_sched_mask_init(adev);
+
 	amdgpu_ras_debugfs_create_all(adev);
 	amdgpu_rap_debugfs_init(adev);
 	amdgpu_securedisplay_debugfs_init(adev);
 	amdgpu_fw_attestation_debugfs_init(adev);
+	amdgpu_psp_debugfs_init(adev);
 
-	debugfs_create_file("amdgpu_evict_vram", 0444, root, adev,
+	debugfs_create_file("amdgpu_evict_vram", 0400, root, adev,
 			    &amdgpu_evict_vram_fops);
-	debugfs_create_file("amdgpu_evict_gtt", 0444, root, adev,
+	debugfs_create_file("amdgpu_evict_gtt", 0400, root, adev,
 			    &amdgpu_evict_gtt_fops);
-	debugfs_create_file("amdgpu_test_ib", 0444, root, adev,
+	debugfs_create_file("amdgpu_test_ib", 0400, root, adev,
 			    &amdgpu_debugfs_test_ib_fops);
 	debugfs_create_file("amdgpu_vm_info", 0444, root, adev,
 			    &amdgpu_debugfs_vm_info_fops);
 	debugfs_create_file("amdgpu_benchmark", 0200, root, adev,
 			    &amdgpu_benchmark_fops);
-	debugfs_create_file("amdgpu_reset_dump_register_list", 0644, root, adev,
-			    &amdgpu_reset_dump_register_list);
 
 	adev->debugfs_vbios_blob.data = adev->bios;
 	adev->debugfs_vbios_blob.size = adev->bios_size;
 	debugfs_create_blob("amdgpu_vbios", 0444, root,
 			    &adev->debugfs_vbios_blob);
 
-	adev->debugfs_discovery_blob.data = adev->mman.discovery_bin;
-	adev->debugfs_discovery_blob.size = adev->mman.discovery_tmr_size;
-	debugfs_create_blob("amdgpu_discovery", 0444, root,
-			    &adev->debugfs_discovery_blob);
+	if (adev->discovery.debugfs_blob.size)
+		debugfs_create_blob("amdgpu_discovery", 0444, root,
+				    &adev->discovery.debugfs_blob);
+
+	return 0;
+}
+
+static int amdgpu_pt_info_read(struct seq_file *m, void *unused)
+{
+	struct drm_file *file;
+	struct amdgpu_fpriv *fpriv;
+	struct amdgpu_bo *root_bo;
+	struct amdgpu_device *adev;
+	int r;
+
+	file = m->private;
+	if (!file)
+		return -EINVAL;
+
+	adev = drm_to_adev(file->minor->dev);
+	fpriv = file->driver_priv;
+	if (!fpriv || !fpriv->vm.root.bo)
+		return -ENODEV;
+
+	root_bo = amdgpu_bo_ref(fpriv->vm.root.bo);
+	r = amdgpu_bo_reserve(root_bo, true);
+	if (r) {
+		amdgpu_bo_unref(&root_bo);
+		return -EINVAL;
+	}
+
+	seq_printf(m, "pd_address: 0x%llx\n", amdgpu_gmc_pd_addr(fpriv->vm.root.bo));
+	seq_printf(m, "max_pfn: 0x%llx\n", adev->vm_manager.max_pfn);
+	seq_printf(m, "num_level: 0x%x\n", adev->vm_manager.num_level);
+	seq_printf(m, "block_size: 0x%x\n", adev->vm_manager.block_size);
+	seq_printf(m, "fragment_size: 0x%x\n", adev->vm_manager.fragment_size);
+
+	amdgpu_bo_unreserve(root_bo);
+	amdgpu_bo_unref(&root_bo);
 
 	return 0;
 }
 
+static int amdgpu_pt_info_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, amdgpu_pt_info_read, inode->i_private);
+}
+
+static const struct file_operations amdgpu_pt_info_fops = {
+	.owner = THIS_MODULE,
+	.open = amdgpu_pt_info_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = single_release,
+};
+
+void amdgpu_debugfs_vm_init(struct drm_file *file)
+{
+	debugfs_create_file("vm_pagetable_info", 0444, file->debugfs_client, file,
+			    &amdgpu_pt_info_fops);
+}
+
 #else
 int amdgpu_debugfs_init(struct amdgpu_device *adev)
 {
@@ -2028,4 +2169,7 @@ int amdgpu_debugfs_regs_init(struct amdgpu_device *adev)
 {
 	return 0;
 }
+void amdgpu_debugfs_vm_init(struct drm_file *file)
+{
+}
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h
index 371a6f0deb29..e7b3c38e5186 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h
@@ -32,3 +32,6 @@ void amdgpu_debugfs_fini(struct amdgpu_device *adev);
 void amdgpu_debugfs_fence_init(struct amdgpu_device *adev);
 void amdgpu_debugfs_firmware_init(struct amdgpu_device *adev);
 void amdgpu_debugfs_gem_init(struct amdgpu_device *adev);
+void amdgpu_debugfs_mes_event_log_init(struct amdgpu_device *adev);
+void amdgpu_debugfs_vm_init(struct drm_file *file);
+
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c
new file mode 100644
index 000000000000..4e2fe6674db8
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c
@@ -0,0 +1,371 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <generated/utsrelease.h>
+#include <linux/devcoredump.h>
+#include "amdgpu_dev_coredump.h"
+#include "atom.h"
+
+#ifndef CONFIG_DEV_COREDUMP
+void amdgpu_coredump(struct amdgpu_device *adev, bool skip_vram_check,
+		     bool vram_lost, struct amdgpu_job *job)
+{
+}
+#else
+
+const char *hw_ip_names[MAX_HWIP] = {
+	[GC_HWIP]		= "GC",
+	[HDP_HWIP]		= "HDP",
+	[SDMA0_HWIP]		= "SDMA0",
+	[SDMA1_HWIP]		= "SDMA1",
+	[SDMA2_HWIP]		= "SDMA2",
+	[SDMA3_HWIP]		= "SDMA3",
+	[SDMA4_HWIP]		= "SDMA4",
+	[SDMA5_HWIP]		= "SDMA5",
+	[SDMA6_HWIP]		= "SDMA6",
+	[SDMA7_HWIP]		= "SDMA7",
+	[LSDMA_HWIP]		= "LSDMA",
+	[MMHUB_HWIP]		= "MMHUB",
+	[ATHUB_HWIP]		= "ATHUB",
+	[NBIO_HWIP]		= "NBIO",
+	[MP0_HWIP]		= "MP0",
+	[MP1_HWIP]		= "MP1",
+	[UVD_HWIP]		= "UVD/JPEG/VCN",
+	[VCN1_HWIP]		= "VCN1",
+	[VCE_HWIP]		= "VCE",
+	[VPE_HWIP]		= "VPE",
+	[DF_HWIP]		= "DF",
+	[DCE_HWIP]		= "DCE",
+	[OSSSYS_HWIP]		= "OSSSYS",
+	[SMUIO_HWIP]		= "SMUIO",
+	[PWR_HWIP]		= "PWR",
+	[NBIF_HWIP]		= "NBIF",
+	[THM_HWIP]		= "THM",
+	[CLK_HWIP]		= "CLK",
+	[UMC_HWIP]		= "UMC",
+	[RSMU_HWIP]		= "RSMU",
+	[XGMI_HWIP]		= "XGMI",
+	[DCI_HWIP]		= "DCI",
+	[PCIE_HWIP]		= "PCIE",
+};
+
+static void amdgpu_devcoredump_fw_info(struct amdgpu_device *adev,
+				       struct drm_printer *p)
+{
+	uint32_t version;
+	uint32_t feature;
+	uint8_t smu_program, smu_major, smu_minor, smu_debug;
+	struct atom_context *ctx = adev->mode_info.atom_context;
+
+	drm_printf(p, "VCE feature version: %u, fw version: 0x%08x\n",
+		   adev->vce.fb_version, adev->vce.fw_version);
+	drm_printf(p, "UVD feature version: %u, fw version: 0x%08x\n", 0,
+		   adev->uvd.fw_version);
+	drm_printf(p, "GMC feature version: %u, fw version: 0x%08x\n", 0,
+		   adev->gmc.fw_version);
+	drm_printf(p, "ME feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.me_feature_version, adev->gfx.me_fw_version);
+	drm_printf(p, "PFP feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.pfp_feature_version, adev->gfx.pfp_fw_version);
+	drm_printf(p, "CE feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.ce_feature_version, adev->gfx.ce_fw_version);
+	drm_printf(p, "RLC feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.rlc_feature_version, adev->gfx.rlc_fw_version);
+
+	drm_printf(p, "RLC SRLC feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.rlc_srlc_feature_version,
+		   adev->gfx.rlc_srlc_fw_version);
+	drm_printf(p, "RLC SRLG feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.rlc_srlg_feature_version,
+		   adev->gfx.rlc_srlg_fw_version);
+	drm_printf(p, "RLC SRLS feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.rlc_srls_feature_version,
+		   adev->gfx.rlc_srls_fw_version);
+	drm_printf(p, "RLCP feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.rlcp_ucode_feature_version,
+		   adev->gfx.rlcp_ucode_version);
+	drm_printf(p, "RLCV feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.rlcv_ucode_feature_version,
+		   adev->gfx.rlcv_ucode_version);
+	drm_printf(p, "MEC feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.mec_feature_version, adev->gfx.mec_fw_version);
+
+	if (adev->gfx.mec2_fw)
+		drm_printf(p, "MEC2 feature version: %u, fw version: 0x%08x\n",
+			   adev->gfx.mec2_feature_version,
+			   adev->gfx.mec2_fw_version);
+
+	drm_printf(p, "IMU feature version: %u, fw version: 0x%08x\n", 0,
+		   adev->gfx.imu_fw_version);
+	drm_printf(p, "PSP SOS feature version: %u, fw version: 0x%08x\n",
+		   adev->psp.sos.feature_version, adev->psp.sos.fw_version);
+	drm_printf(p, "PSP ASD feature version: %u, fw version: 0x%08x\n",
+		   adev->psp.asd_context.bin_desc.feature_version,
+		   adev->psp.asd_context.bin_desc.fw_version);
+
+	drm_printf(p, "TA XGMI feature version: 0x%08x, fw version: 0x%08x\n",
+		   adev->psp.xgmi_context.context.bin_desc.feature_version,
+		   adev->psp.xgmi_context.context.bin_desc.fw_version);
+	drm_printf(p, "TA RAS feature version: 0x%08x, fw version: 0x%08x\n",
+		   adev->psp.ras_context.context.bin_desc.feature_version,
+		   adev->psp.ras_context.context.bin_desc.fw_version);
+	drm_printf(p, "TA HDCP feature version: 0x%08x, fw version: 0x%08x\n",
+		   adev->psp.hdcp_context.context.bin_desc.feature_version,
+		   adev->psp.hdcp_context.context.bin_desc.fw_version);
+	drm_printf(p, "TA DTM feature version: 0x%08x, fw version: 0x%08x\n",
+		   adev->psp.dtm_context.context.bin_desc.feature_version,
+		   adev->psp.dtm_context.context.bin_desc.fw_version);
+	drm_printf(p, "TA RAP feature version: 0x%08x, fw version: 0x%08x\n",
+		   adev->psp.rap_context.context.bin_desc.feature_version,
+		   adev->psp.rap_context.context.bin_desc.fw_version);
+	drm_printf(p,
+		   "TA SECURE DISPLAY feature version: 0x%08x, fw version: 0x%08x\n",
+		   adev->psp.securedisplay_context.context.bin_desc.feature_version,
+		   adev->psp.securedisplay_context.context.bin_desc.fw_version);
+
+	/* SMC firmware */
+	version = adev->pm.fw_version;
+
+	smu_program = (version >> 24) & 0xff;
+	smu_major = (version >> 16) & 0xff;
+	smu_minor = (version >> 8) & 0xff;
+	smu_debug = (version >> 0) & 0xff;
+	drm_printf(p,
+		   "SMC feature version: %u, program: %d, fw version: 0x%08x (%d.%d.%d)\n",
+		   0, smu_program, version, smu_major, smu_minor, smu_debug);
+
+	/* SDMA firmware */
+	for (int i = 0; i < adev->sdma.num_instances; i++) {
+		drm_printf(p,
+			   "SDMA%d feature version: %u, firmware version: 0x%08x\n",
+			   i, adev->sdma.instance[i].feature_version,
+			   adev->sdma.instance[i].fw_version);
+	}
+
+	drm_printf(p, "VCN feature version: %u, fw version: 0x%08x\n", 0,
+		   adev->vcn.fw_version);
+	drm_printf(p, "DMCU feature version: %u, fw version: 0x%08x\n", 0,
+		   adev->dm.dmcu_fw_version);
+	drm_printf(p, "DMCUB feature version: %u, fw version: 0x%08x\n", 0,
+		   adev->dm.dmcub_fw_version);
+	drm_printf(p, "PSP TOC feature version: %u, fw version: 0x%08x\n",
+		   adev->psp.toc.feature_version, adev->psp.toc.fw_version);
+
+	version = adev->mes.kiq_version & AMDGPU_MES_VERSION_MASK;
+	feature = (adev->mes.kiq_version & AMDGPU_MES_FEAT_VERSION_MASK) >>
+		  AMDGPU_MES_FEAT_VERSION_SHIFT;
+	drm_printf(p, "MES_KIQ feature version: %u, fw version: 0x%08x\n",
+		   feature, version);
+
+	version = adev->mes.sched_version & AMDGPU_MES_VERSION_MASK;
+	feature = (adev->mes.sched_version & AMDGPU_MES_FEAT_VERSION_MASK) >>
+		  AMDGPU_MES_FEAT_VERSION_SHIFT;
+	drm_printf(p, "MES feature version: %u, fw version: 0x%08x\n", feature,
+		   version);
+
+	drm_printf(p, "VPE feature version: %u, fw version: 0x%08x\n",
+		   adev->vpe.feature_version, adev->vpe.fw_version);
+
+	drm_printf(p, "\nVBIOS Information\n");
+	drm_printf(p, "vbios name       : %s\n", ctx->name);
+	drm_printf(p, "vbios pn         : %s\n", ctx->vbios_pn);
+	drm_printf(p, "vbios version    : %d\n", ctx->version);
+	drm_printf(p, "vbios ver_str    : %s\n", ctx->vbios_ver_str);
+	drm_printf(p, "vbios date       : %s\n", ctx->date);
+}
+
+static ssize_t
+amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count,
+			void *data, size_t datalen)
+{
+	struct drm_printer p;
+	struct amdgpu_coredump_info *coredump = data;
+	struct drm_print_iterator iter;
+	struct amdgpu_vm_fault_info *fault_info;
+	struct amdgpu_ip_block *ip_block;
+	int ver;
+
+	iter.data = buffer;
+	iter.offset = 0;
+	iter.start = offset;
+	iter.remain = count;
+
+	p = drm_coredump_printer(&iter);
+
+	drm_printf(&p, "**** AMDGPU Device Coredump ****\n");
+	drm_printf(&p, "version: " AMDGPU_COREDUMP_VERSION "\n");
+	drm_printf(&p, "kernel: " UTS_RELEASE "\n");
+	drm_printf(&p, "module: " KBUILD_MODNAME "\n");
+	drm_printf(&p, "time: %ptSp\n", &coredump->reset_time);
+
+	if (coredump->reset_task_info.task.pid)
+		drm_printf(&p, "process_name: %s PID: %d\n",
+			   coredump->reset_task_info.process_name,
+			   coredump->reset_task_info.task.pid);
+
+	/* SOC Information */
+	drm_printf(&p, "\nSOC Information\n");
+	drm_printf(&p, "SOC Device id: %d\n", coredump->adev->pdev->device);
+	drm_printf(&p, "SOC PCI Revision id: %d\n", coredump->adev->pdev->revision);
+	drm_printf(&p, "SOC Family: %d\n", coredump->adev->family);
+	drm_printf(&p, "SOC Revision id: %d\n", coredump->adev->rev_id);
+	drm_printf(&p, "SOC External Revision id: %d\n", coredump->adev->external_rev_id);
+
+	/* Memory Information */
+	drm_printf(&p, "\nSOC Memory Information\n");
+	drm_printf(&p, "real vram size: %llu\n", coredump->adev->gmc.real_vram_size);
+	drm_printf(&p, "visible vram size: %llu\n", coredump->adev->gmc.visible_vram_size);
+	drm_printf(&p, "gtt size: %llu\n", coredump->adev->mman.gtt_mgr.manager.size);
+
+	/* GDS Config */
+	drm_printf(&p, "\nGDS Config\n");
+	drm_printf(&p, "gds: total size: %d\n", coredump->adev->gds.gds_size);
+	drm_printf(&p, "gds: compute partition size: %d\n", coredump->adev->gds.gds_size);
+	drm_printf(&p, "gds: gws per compute partition: %d\n", coredump->adev->gds.gws_size);
+	drm_printf(&p, "gds: os per compute partition: %d\n", coredump->adev->gds.oa_size);
+
+	/* HWIP Version Information */
+	drm_printf(&p, "\nHW IP Version Information\n");
+	for (int i = 1; i < MAX_HWIP; i++) {
+		for (int j = 0; j < HWIP_MAX_INSTANCE; j++) {
+			ver = coredump->adev->ip_versions[i][j];
+			if (ver)
+				drm_printf(&p, "HWIP: %s[%d][%d]: v%d.%d.%d.%d.%d\n",
+					   hw_ip_names[i], i, j,
+					   IP_VERSION_MAJ(ver),
+					   IP_VERSION_MIN(ver),
+					   IP_VERSION_REV(ver),
+					   IP_VERSION_VARIANT(ver),
+					   IP_VERSION_SUBREV(ver));
+		}
+	}
+
+	/* IP firmware information */
+	drm_printf(&p, "\nIP Firmwares\n");
+	amdgpu_devcoredump_fw_info(coredump->adev, &p);
+
+	if (coredump->ring) {
+		drm_printf(&p, "\nRing timed out details\n");
+		drm_printf(&p, "IP Type: %d Ring Name: %s\n",
+			   coredump->ring->funcs->type,
+			   coredump->ring->name);
+	}
+
+	/* Add page fault information */
+	fault_info = &coredump->adev->vm_manager.fault_info;
+	drm_printf(&p, "\n[%s] Page fault observed\n",
+		   fault_info->vmhub ? "mmhub" : "gfxhub");
+	drm_printf(&p, "Faulty page starting at address: 0x%016llx\n", fault_info->addr);
+	drm_printf(&p, "Protection fault status register: 0x%x\n\n", fault_info->status);
+
+	/* dump the ip state for each ip */
+	drm_printf(&p, "IP Dump\n");
+	for (int i = 0; i < coredump->adev->num_ip_blocks; i++) {
+		ip_block = &coredump->adev->ip_blocks[i];
+		if (ip_block->version->funcs->print_ip_state) {
+			drm_printf(&p, "IP: %s\n", ip_block->version->funcs->name);
+			ip_block->version->funcs->print_ip_state(ip_block, &p);
+			drm_printf(&p, "\n");
+		}
+	}
+
+	/* Add ring buffer information */
+	drm_printf(&p, "Ring buffer information\n");
+	for (int i = 0; i < coredump->adev->num_rings; i++) {
+		int j = 0;
+		struct amdgpu_ring *ring = coredump->adev->rings[i];
+
+		drm_printf(&p, "ring name: %s\n", ring->name);
+		drm_printf(&p, "Rptr: 0x%llx Wptr: 0x%llx RB mask: %x\n",
+			   amdgpu_ring_get_rptr(ring),
+			   amdgpu_ring_get_wptr(ring),
+			   ring->buf_mask);
+		drm_printf(&p, "Ring size in dwords: %d\n",
+			   ring->ring_size / 4);
+		drm_printf(&p, "Ring contents\n");
+		drm_printf(&p, "Offset \t Value\n");
+
+		while (j < ring->ring_size) {
+			drm_printf(&p, "0x%x \t 0x%x\n", j, ring->ring[j / 4]);
+			j += 4;
+		}
+	}
+
+	if (coredump->skip_vram_check)
+		drm_printf(&p, "VRAM lost check is skipped!\n");
+	else if (coredump->reset_vram_lost)
+		drm_printf(&p, "VRAM is lost due to GPU reset!\n");
+
+	return count - iter.remain;
+}
+
+static void amdgpu_devcoredump_free(void *data)
+{
+	kfree(data);
+}
+
+void amdgpu_coredump(struct amdgpu_device *adev, bool skip_vram_check,
+		     bool vram_lost, struct amdgpu_job *job)
+{
+	struct drm_device *dev = adev_to_drm(adev);
+	struct amdgpu_coredump_info *coredump;
+	struct drm_sched_job *s_job;
+
+	coredump = kzalloc(sizeof(*coredump), GFP_NOWAIT);
+
+	if (!coredump) {
+		DRM_ERROR("%s: failed to allocate memory for coredump\n", __func__);
+		return;
+	}
+
+	coredump->skip_vram_check = skip_vram_check;
+	coredump->reset_vram_lost = vram_lost;
+
+	if (job && job->pasid) {
+		struct amdgpu_task_info *ti;
+
+		ti = amdgpu_vm_get_task_info_pasid(adev, job->pasid);
+		if (ti) {
+			coredump->reset_task_info = *ti;
+			amdgpu_vm_put_task_info(ti);
+		}
+	}
+
+	if (job) {
+		s_job = &job->base;
+		coredump->ring = to_amdgpu_ring(s_job->sched);
+	}
+
+	coredump->adev = adev;
+
+	ktime_get_ts64(&coredump->reset_time);
+
+	dev_coredumpm(dev->dev, THIS_MODULE, coredump, 0, GFP_NOWAIT,
+		      amdgpu_devcoredump_read, amdgpu_devcoredump_free);
+
+	drm_info(dev, "AMDGPU device coredump file has been created\n");
+	drm_info(dev, "Check your /sys/class/drm/card%d/device/devcoredump/data\n",
+		 dev->primary->index);
+}
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.h
new file mode 100644
index 000000000000..ef9772c6bcc9
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.h
@@ -0,0 +1,46 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __AMDGPU_DEV_COREDUMP_H__
+#define __AMDGPU_DEV_COREDUMP_H__
+
+#include "amdgpu.h"
+
+#ifdef CONFIG_DEV_COREDUMP
+
+#define AMDGPU_COREDUMP_VERSION "1"
+
+struct amdgpu_coredump_info {
+	struct amdgpu_device            *adev;
+	struct amdgpu_task_info         reset_task_info;
+	struct timespec64               reset_time;
+	bool                            skip_vram_check;
+	bool                            reset_vram_lost;
+	struct amdgpu_ring              *ring;
+};
+#endif
+
+void amdgpu_coredump(struct amdgpu_device *adev, bool skip_vram_check,
+		     bool vram_lost, struct amdgpu_job *job);
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index a2292acf06d0..58c3ffe707d1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -25,6 +25,8 @@
  *          Alex Deucher
  *          Jerome Glisse
  */
+
+#include <linux/aperture.h>
 #include <linux/power_supply.h>
 #include <linux/kthread.h>
 #include <linux/module.h>
@@ -32,17 +34,15 @@
 #include <linux/slab.h>
 #include <linux/iommu.h>
 #include <linux/pci.h>
-#include <linux/devcoredump.h>
-#include <generated/utsrelease.h>
 #include <linux/pci-p2pdma.h>
 #include <linux/apple-gmux.h>
 
-#include <drm/drm_aperture.h>
 #include <drm/drm_atomic_helper.h>
+#include <drm/drm_client_event.h>
 #include <drm/drm_crtc_helper.h>
-#include <drm/drm_fb_helper.h>
 #include <drm/drm_probe_helper.h>
 #include <drm/amdgpu_drm.h>
+#include <linux/device.h>
 #include <linux/vgaarb.h>
 #include <linux/vga_switcheroo.h>
 #include <linux/efi.h>
@@ -71,9 +71,12 @@
 
 #include "amdgpu_xgmi.h"
 #include "amdgpu_ras.h"
+#include "amdgpu_ras_mgr.h"
 #include "amdgpu_pmu.h"
 #include "amdgpu_fru_eeprom.h"
 #include "amdgpu_reset.h"
+#include "amdgpu_virt.h"
+#include "amdgpu_dev_coredump.h"
 
 #include <linux/suspend.h>
 #include <drm/task_barrier.h>
@@ -83,6 +86,7 @@
 
 #if IS_ENABLED(CONFIG_X86)
 #include <asm/intel-family.h>
+#include <asm/cpu_device_id.h>
 #endif
 
 MODULE_FIRMWARE("amdgpu/vega10_gpu_info.bin");
@@ -92,10 +96,17 @@ MODULE_FIRMWARE("amdgpu/picasso_gpu_info.bin");
 MODULE_FIRMWARE("amdgpu/raven2_gpu_info.bin");
 MODULE_FIRMWARE("amdgpu/arcturus_gpu_info.bin");
 MODULE_FIRMWARE("amdgpu/navi12_gpu_info.bin");
+MODULE_FIRMWARE("amdgpu/cyan_skillfish_gpu_info.bin");
 
 #define AMDGPU_RESUME_MS		2000
 #define AMDGPU_MAX_RETRY_LIMIT		2
 #define AMDGPU_RETRY_SRIOV_RESET(r) ((r) == -EBUSY || (r) == -ETIMEDOUT || (r) == -EINVAL)
+#define AMDGPU_PCIE_INDEX_FALLBACK (0x38 >> 2)
+#define AMDGPU_PCIE_INDEX_HI_FALLBACK (0x44 >> 2)
+#define AMDGPU_PCIE_DATA_FALLBACK (0x3C >> 2)
+
+#define AMDGPU_VBIOS_SKIP (1U << 0)
+#define AMDGPU_VBIOS_OPTIONAL (1U << 1)
 
 static const struct drm_driver amdgpu_kms_driver;
 
@@ -140,13 +151,76 @@ const char *amdgpu_asic_name[] = {
 	"LAST",
 };
 
+#define AMDGPU_IP_BLK_MASK_ALL GENMASK(AMD_IP_BLOCK_TYPE_NUM  - 1, 0)
+/*
+ * Default init level where all blocks are expected to be initialized. This is
+ * the level of initialization expected by default and also after a full reset
+ * of the device.
+ */
+struct amdgpu_init_level amdgpu_init_default = {
+	.level = AMDGPU_INIT_LEVEL_DEFAULT,
+	.hwini_ip_block_mask = AMDGPU_IP_BLK_MASK_ALL,
+};
+
+struct amdgpu_init_level amdgpu_init_recovery = {
+	.level = AMDGPU_INIT_LEVEL_RESET_RECOVERY,
+	.hwini_ip_block_mask = AMDGPU_IP_BLK_MASK_ALL,
+};
+
+/*
+ * Minimal blocks needed to be initialized before a XGMI hive can be reset. This
+ * is used for cases like reset on initialization where the entire hive needs to
+ * be reset before first use.
+ */
+struct amdgpu_init_level amdgpu_init_minimal_xgmi = {
+	.level = AMDGPU_INIT_LEVEL_MINIMAL_XGMI,
+	.hwini_ip_block_mask =
+		BIT(AMD_IP_BLOCK_TYPE_GMC) | BIT(AMD_IP_BLOCK_TYPE_SMC) |
+		BIT(AMD_IP_BLOCK_TYPE_COMMON) | BIT(AMD_IP_BLOCK_TYPE_IH) |
+		BIT(AMD_IP_BLOCK_TYPE_PSP)
+};
+
+static int amdgpu_device_ip_resume_phase1(struct amdgpu_device *adev);
+static int amdgpu_device_ip_resume_phase2(struct amdgpu_device *adev);
+static int amdgpu_device_ip_resume_phase3(struct amdgpu_device *adev);
+
+static void amdgpu_device_load_switch_state(struct amdgpu_device *adev);
+
+static inline bool amdgpu_ip_member_of_hwini(struct amdgpu_device *adev,
+					     enum amd_ip_block_type block)
+{
+	return (adev->init_lvl->hwini_ip_block_mask & (1U << block)) != 0;
+}
+
+void amdgpu_set_init_level(struct amdgpu_device *adev,
+			   enum amdgpu_init_lvl_id lvl)
+{
+	switch (lvl) {
+	case AMDGPU_INIT_LEVEL_MINIMAL_XGMI:
+		adev->init_lvl = &amdgpu_init_minimal_xgmi;
+		break;
+	case AMDGPU_INIT_LEVEL_RESET_RECOVERY:
+		adev->init_lvl = &amdgpu_init_recovery;
+		break;
+	case AMDGPU_INIT_LEVEL_DEFAULT:
+		fallthrough;
+	default:
+		adev->init_lvl = &amdgpu_init_default;
+		break;
+	}
+}
+
+static inline void amdgpu_device_stop_pending_resets(struct amdgpu_device *adev);
+static int amdgpu_device_pm_notifier(struct notifier_block *nb, unsigned long mode,
+				     void *data);
+
 /**
  * DOC: pcie_replay_count
  *
  * The amdgpu driver provides a sysfs API for reporting the total number
- * of PCIe replays (NAKs)
+ * of PCIe replays (NAKs).
  * The file pcie_replay_count is used for this and returns the total
- * number of replays as a sum of the NAKs generated and NAKs received
+ * number of replays as a sum of the NAKs generated and NAKs received.
  */
 
 static ssize_t amdgpu_device_get_pcie_replay_count(struct device *dev,
@@ -159,89 +233,202 @@ static ssize_t amdgpu_device_get_pcie_replay_count(struct device *dev,
 	return sysfs_emit(buf, "%llu\n", cnt);
 }
 
-static DEVICE_ATTR(pcie_replay_count, S_IRUGO,
+static DEVICE_ATTR(pcie_replay_count, 0444,
 		amdgpu_device_get_pcie_replay_count, NULL);
 
-static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev);
+static int amdgpu_device_attr_sysfs_init(struct amdgpu_device *adev)
+{
+	int ret = 0;
 
-/**
- * DOC: product_name
- *
- * The amdgpu driver provides a sysfs API for reporting the product name
- * for the device
- * The file product_name is used for this and returns the product name
- * as returned from the FRU.
- * NOTE: This is only available for certain server cards
- */
+	if (amdgpu_nbio_is_replay_cnt_supported(adev))
+		ret = sysfs_create_file(&adev->dev->kobj,
+					&dev_attr_pcie_replay_count.attr);
 
-static ssize_t amdgpu_device_get_product_name(struct device *dev,
-		struct device_attribute *attr, char *buf)
+	return ret;
+}
+
+static void amdgpu_device_attr_sysfs_fini(struct amdgpu_device *adev)
 {
+	if (amdgpu_nbio_is_replay_cnt_supported(adev))
+		sysfs_remove_file(&adev->dev->kobj,
+				  &dev_attr_pcie_replay_count.attr);
+}
+
+static ssize_t amdgpu_sysfs_reg_state_get(struct file *f, struct kobject *kobj,
+					  const struct bin_attribute *attr, char *buf,
+					  loff_t ppos, size_t count)
+{
+	struct device *dev = kobj_to_dev(kobj);
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = drm_to_adev(ddev);
+	ssize_t bytes_read;
+
+	switch (ppos) {
+	case AMDGPU_SYS_REG_STATE_XGMI:
+		bytes_read = amdgpu_asic_get_reg_state(
+			adev, AMDGPU_REG_STATE_TYPE_XGMI, buf, count);
+		break;
+	case AMDGPU_SYS_REG_STATE_WAFL:
+		bytes_read = amdgpu_asic_get_reg_state(
+			adev, AMDGPU_REG_STATE_TYPE_WAFL, buf, count);
+		break;
+	case AMDGPU_SYS_REG_STATE_PCIE:
+		bytes_read = amdgpu_asic_get_reg_state(
+			adev, AMDGPU_REG_STATE_TYPE_PCIE, buf, count);
+		break;
+	case AMDGPU_SYS_REG_STATE_USR:
+		bytes_read = amdgpu_asic_get_reg_state(
+			adev, AMDGPU_REG_STATE_TYPE_USR, buf, count);
+		break;
+	case AMDGPU_SYS_REG_STATE_USR_1:
+		bytes_read = amdgpu_asic_get_reg_state(
+			adev, AMDGPU_REG_STATE_TYPE_USR_1, buf, count);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return bytes_read;
+}
+
+static const BIN_ATTR(reg_state, 0444, amdgpu_sysfs_reg_state_get, NULL,
+		      AMDGPU_SYS_REG_STATE_END);
+
+int amdgpu_reg_state_sysfs_init(struct amdgpu_device *adev)
+{
+	int ret;
+
+	if (!amdgpu_asic_get_reg_state_supported(adev))
+		return 0;
+
+	ret = sysfs_create_bin_file(&adev->dev->kobj, &bin_attr_reg_state);
 
-	return sysfs_emit(buf, "%s\n", adev->product_name);
+	return ret;
+}
+
+void amdgpu_reg_state_sysfs_fini(struct amdgpu_device *adev)
+{
+	if (!amdgpu_asic_get_reg_state_supported(adev))
+		return;
+	sysfs_remove_bin_file(&adev->dev->kobj, &bin_attr_reg_state);
 }
 
-static DEVICE_ATTR(product_name, S_IRUGO,
-		amdgpu_device_get_product_name, NULL);
+int amdgpu_ip_block_suspend(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+
+	if (ip_block->version->funcs->suspend) {
+		r = ip_block->version->funcs->suspend(ip_block);
+		if (r) {
+			dev_err(ip_block->adev->dev,
+				"suspend of IP block <%s> failed %d\n",
+				ip_block->version->funcs->name, r);
+			return r;
+		}
+	}
+
+	ip_block->status.hw = false;
+	return 0;
+}
+
+int amdgpu_ip_block_resume(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+
+	if (ip_block->version->funcs->resume) {
+		r = ip_block->version->funcs->resume(ip_block);
+		if (r) {
+			dev_err(ip_block->adev->dev,
+				"resume of IP block <%s> failed %d\n",
+				ip_block->version->funcs->name, r);
+			return r;
+		}
+	}
+
+	ip_block->status.hw = true;
+	return 0;
+}
 
 /**
- * DOC: product_number
+ * DOC: board_info
+ *
+ * The amdgpu driver provides a sysfs API for giving board related information.
+ * It provides the form factor information in the format
+ *
+ *   type : form factor
+ *
+ * Possible form factor values
+ *
+ * - "cem"		- PCIE CEM card
+ * - "oam"		- Open Compute Accelerator Module
+ * - "unknown"	- Not known
  *
- * The amdgpu driver provides a sysfs API for reporting the part number
- * for the device
- * The file product_number is used for this and returns the part number
- * as returned from the FRU.
- * NOTE: This is only available for certain server cards
  */
 
-static ssize_t amdgpu_device_get_product_number(struct device *dev,
-		struct device_attribute *attr, char *buf)
+static ssize_t amdgpu_device_get_board_info(struct device *dev,
+					    struct device_attribute *attr,
+					    char *buf)
 {
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = drm_to_adev(ddev);
+	enum amdgpu_pkg_type pkg_type = AMDGPU_PKG_TYPE_CEM;
+	const char *pkg;
+
+	if (adev->smuio.funcs && adev->smuio.funcs->get_pkg_type)
+		pkg_type = adev->smuio.funcs->get_pkg_type(adev);
+
+	switch (pkg_type) {
+	case AMDGPU_PKG_TYPE_CEM:
+		pkg = "cem";
+		break;
+	case AMDGPU_PKG_TYPE_OAM:
+		pkg = "oam";
+		break;
+	default:
+		pkg = "unknown";
+		break;
+	}
 
-	return sysfs_emit(buf, "%s\n", adev->product_number);
+	return sysfs_emit(buf, "%s : %s\n", "type", pkg);
 }
 
-static DEVICE_ATTR(product_number, S_IRUGO,
-		amdgpu_device_get_product_number, NULL);
+static DEVICE_ATTR(board_info, 0444, amdgpu_device_get_board_info, NULL);
 
-/**
- * DOC: serial_number
- *
- * The amdgpu driver provides a sysfs API for reporting the serial number
- * for the device
- * The file serial_number is used for this and returns the serial number
- * as returned from the FRU.
- * NOTE: This is only available for certain server cards
- */
+static struct attribute *amdgpu_board_attrs[] = {
+	&dev_attr_board_info.attr,
+	NULL,
+};
 
-static ssize_t amdgpu_device_get_serial_number(struct device *dev,
-		struct device_attribute *attr, char *buf)
+static umode_t amdgpu_board_attrs_is_visible(struct kobject *kobj,
+					     struct attribute *attr, int n)
 {
+	struct device *dev = kobj_to_dev(kobj);
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = drm_to_adev(ddev);
 
-	return sysfs_emit(buf, "%s\n", adev->serial);
+	if (adev->flags & AMD_IS_APU)
+		return 0;
+
+	return attr->mode;
 }
 
-static DEVICE_ATTR(serial_number, S_IRUGO,
-		amdgpu_device_get_serial_number, NULL);
+static const struct attribute_group amdgpu_board_attrs_group = {
+	.attrs = amdgpu_board_attrs,
+	.is_visible = amdgpu_board_attrs_is_visible
+};
+
+static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev);
 
 /**
  * amdgpu_device_supports_px - Is the device a dGPU with ATPX power control
  *
- * @dev: drm_device pointer
+ * @adev: amdgpu device pointer
  *
  * Returns true if the device is a dGPU with ATPX power control,
  * otherwise return false.
  */
-bool amdgpu_device_supports_px(struct drm_device *dev)
+bool amdgpu_device_supports_px(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = drm_to_adev(dev);
-
 	if ((adev->flags & AMD_IS_PX) && !amdgpu_is_atpx_hybrid())
 		return true;
 	return false;
@@ -250,14 +437,15 @@ bool amdgpu_device_supports_px(struct drm_device *dev)
 /**
  * amdgpu_device_supports_boco - Is the device a dGPU with ACPI power resources
  *
- * @dev: drm_device pointer
+ * @adev: amdgpu device pointer
  *
  * Returns true if the device is a dGPU with ACPI power control,
  * otherwise return false.
  */
-bool amdgpu_device_supports_boco(struct drm_device *dev)
+bool amdgpu_device_supports_boco(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = drm_to_adev(dev);
+	if (!IS_ENABLED(CONFIG_HOTPLUG_PCI_PCIE))
+		return false;
 
 	if (adev->has_pr3 ||
 	    ((adev->flags & AMD_IS_PX) && amdgpu_is_atpx_hybrid()))
@@ -268,30 +456,105 @@ bool amdgpu_device_supports_boco(struct drm_device *dev)
 /**
  * amdgpu_device_supports_baco - Does the device support BACO
  *
- * @dev: drm_device pointer
+ * @adev: amdgpu device pointer
  *
- * Returns true if the device supporte BACO,
- * otherwise return false.
+ * Return:
+ * 1 if the device supports BACO;
+ * 3 if the device supports MACO (only works if BACO is supported)
+ * otherwise return 0.
  */
-bool amdgpu_device_supports_baco(struct drm_device *dev)
+int amdgpu_device_supports_baco(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = drm_to_adev(dev);
-
 	return amdgpu_asic_supports_baco(adev);
 }
 
+void amdgpu_device_detect_runtime_pm_mode(struct amdgpu_device *adev)
+{
+	int bamaco_support;
+
+	adev->pm.rpm_mode = AMDGPU_RUNPM_NONE;
+	bamaco_support = amdgpu_device_supports_baco(adev);
+
+	switch (amdgpu_runtime_pm) {
+	case 2:
+		if (bamaco_support & MACO_SUPPORT) {
+			adev->pm.rpm_mode = AMDGPU_RUNPM_BAMACO;
+			dev_info(adev->dev, "Forcing BAMACO for runtime pm\n");
+		} else if (bamaco_support == BACO_SUPPORT) {
+			adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
+			dev_info(adev->dev, "Requested mode BAMACO not available,fallback to use BACO\n");
+		}
+		break;
+	case 1:
+		if (bamaco_support & BACO_SUPPORT) {
+			adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
+			dev_info(adev->dev, "Forcing BACO for runtime pm\n");
+		}
+		break;
+	case -1:
+	case -2:
+		if (amdgpu_device_supports_px(adev)) {
+			/* enable PX as runtime mode */
+			adev->pm.rpm_mode = AMDGPU_RUNPM_PX;
+			dev_info(adev->dev, "Using ATPX for runtime pm\n");
+		} else if (amdgpu_device_supports_boco(adev)) {
+			/* enable boco as runtime mode */
+			adev->pm.rpm_mode = AMDGPU_RUNPM_BOCO;
+			dev_info(adev->dev, "Using BOCO for runtime pm\n");
+		} else {
+			if (!bamaco_support)
+				goto no_runtime_pm;
+
+			switch (adev->asic_type) {
+			case CHIP_VEGA20:
+			case CHIP_ARCTURUS:
+				/* BACO are not supported on vega20 and arctrus */
+				break;
+			case CHIP_VEGA10:
+				/* enable BACO as runpm mode if noretry=0 */
+				if (!adev->gmc.noretry && !amdgpu_passthrough(adev))
+					adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
+				break;
+			default:
+				/* enable BACO as runpm mode on CI+ */
+				if (!amdgpu_passthrough(adev))
+					adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
+				break;
+			}
+
+			if (adev->pm.rpm_mode == AMDGPU_RUNPM_BACO) {
+				if (bamaco_support & MACO_SUPPORT) {
+					adev->pm.rpm_mode = AMDGPU_RUNPM_BAMACO;
+					dev_info(adev->dev, "Using BAMACO for runtime pm\n");
+				} else {
+					dev_info(adev->dev, "Using BACO for runtime pm\n");
+				}
+			}
+		}
+		break;
+	case 0:
+		dev_info(adev->dev, "runtime pm is manually disabled\n");
+		break;
+	default:
+		break;
+	}
+
+no_runtime_pm:
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_NONE)
+		dev_info(adev->dev, "Runtime PM not available\n");
+}
 /**
  * amdgpu_device_supports_smart_shift - Is the device dGPU with
  * smart shift support
  *
- * @dev: drm_device pointer
+ * @adev: amdgpu device pointer
  *
  * Returns true if the device is a dGPU with Smart Shift support,
  * otherwise returns false.
  */
-bool amdgpu_device_supports_smart_shift(struct drm_device *dev)
+bool amdgpu_device_supports_smart_shift(struct amdgpu_device *adev)
 {
-	return (amdgpu_device_supports_boco(dev) &&
+	return (amdgpu_device_supports_boco(adev) &&
 		amdgpu_acpi_is_power_shift_control_supported());
 }
 
@@ -342,7 +605,7 @@ void amdgpu_device_mm_access(struct amdgpu_device *adev, loff_t pos,
 }
 
 /**
- * amdgpu_device_aper_access - access vram by vram aperature
+ * amdgpu_device_aper_access - access vram by vram aperture
  *
  * @adev: amdgpu_device pointer
  * @pos: offset of the buffer in vram
@@ -370,10 +633,16 @@ size_t amdgpu_device_aper_access(struct amdgpu_device *adev, loff_t pos,
 
 		if (write) {
 			memcpy_toio(addr, buf, count);
+			/* Make sure HDP write cache flush happens without any reordering
+			 * after the system memory contents are sent over PCIe device
+			 */
 			mb();
 			amdgpu_device_flush_hdp(adev, NULL);
 		} else {
 			amdgpu_device_invalidate_hdp(adev, NULL);
+			/* Make sure HDP read cache is invalidated before issuing a read
+			 * to the PCIe device
+			 */
 			mb();
 			memcpy_fromio(buf, addr, count);
 		}
@@ -427,7 +696,7 @@ bool amdgpu_device_skip_hw_access(struct amdgpu_device *adev)
 	 * here is that the GPU reset is not running on another thread in parallel.
 	 *
 	 * For this we trylock the read side of the reset semaphore, if that succeeds
-	 * we know that the reset is not running in paralell.
+	 * we know that the reset is not running in parallel.
 	 *
 	 * If the trylock fails we assert that we are either already holding the read
 	 * side of the lock or are the reset thread itself and hold the write side of
@@ -464,7 +733,7 @@ uint32_t amdgpu_device_rreg(struct amdgpu_device *adev,
 		if (!(acc_flags & AMDGPU_REGS_NO_KIQ) &&
 		    amdgpu_sriov_runtime(adev) &&
 		    down_read_trylock(&adev->reset_domain->sem)) {
-			ret = amdgpu_kiq_rreg(adev, reg);
+			ret = amdgpu_kiq_rreg(adev, reg, 0);
 			up_read(&adev->reset_domain->sem);
 		} else {
 			ret = readl(((void __iomem *)adev->rmmio) + (reg * 4));
@@ -481,8 +750,7 @@ uint32_t amdgpu_device_rreg(struct amdgpu_device *adev,
 /*
  * MMIO register read with bytes helper functions
  * @offset:bytes offset from MMIO start
- *
-*/
+ */
 
 /**
  * amdgpu_mm_rreg8 - read a memory mapped IO register
@@ -502,12 +770,55 @@ uint8_t amdgpu_mm_rreg8(struct amdgpu_device *adev, uint32_t offset)
 	BUG();
 }
 
+
+/**
+ * amdgpu_device_xcc_rreg - read a memory mapped IO or indirect register with specific XCC
+ *
+ * @adev: amdgpu_device pointer
+ * @reg: dword aligned register offset
+ * @acc_flags: access flags which require special behavior
+ * @xcc_id: xcc accelerated compute core id
+ *
+ * Returns the 32 bit value from the offset specified.
+ */
+uint32_t amdgpu_device_xcc_rreg(struct amdgpu_device *adev,
+				uint32_t reg, uint32_t acc_flags,
+				uint32_t xcc_id)
+{
+	uint32_t ret, rlcg_flag;
+
+	if (amdgpu_device_skip_hw_access(adev))
+		return 0;
+
+	if ((reg * 4) < adev->rmmio_size) {
+		if (amdgpu_sriov_vf(adev) &&
+		    !amdgpu_sriov_runtime(adev) &&
+		    adev->gfx.rlc.rlcg_reg_access_supported &&
+		    amdgpu_virt_get_rlcg_reg_access_flag(adev, acc_flags,
+							 GC_HWIP, false,
+							 &rlcg_flag)) {
+			ret = amdgpu_virt_rlcg_reg_rw(adev, reg, 0, rlcg_flag, GET_INST(GC, xcc_id));
+		} else if (!(acc_flags & AMDGPU_REGS_NO_KIQ) &&
+		    amdgpu_sriov_runtime(adev) &&
+		    down_read_trylock(&adev->reset_domain->sem)) {
+			ret = amdgpu_kiq_rreg(adev, reg, xcc_id);
+			up_read(&adev->reset_domain->sem);
+		} else {
+			ret = readl(((void __iomem *)adev->rmmio) + (reg * 4));
+		}
+	} else {
+		ret = adev->pcie_rreg(adev, reg * 4);
+	}
+
+	return ret;
+}
+
 /*
  * MMIO register write with bytes helper functions
  * @offset:bytes offset from MMIO start
  * @value: the value want to be written to the register
- *
-*/
+ */
+
 /**
  * amdgpu_mm_wreg8 - read a memory mapped IO register
  *
@@ -549,7 +860,7 @@ void amdgpu_device_wreg(struct amdgpu_device *adev,
 		if (!(acc_flags & AMDGPU_REGS_NO_KIQ) &&
 		    amdgpu_sriov_runtime(adev) &&
 		    down_read_trylock(&adev->reset_domain->sem)) {
-			amdgpu_kiq_wreg(adev, reg, v);
+			amdgpu_kiq_wreg(adev, reg, v, 0);
 			up_read(&adev->reset_domain->sem);
 		} else {
 			writel(v, ((void __iomem *)adev->rmmio) + (reg * 4));
@@ -567,11 +878,13 @@ void amdgpu_device_wreg(struct amdgpu_device *adev,
  * @adev: amdgpu_device pointer
  * @reg: mmio/rlc register
  * @v: value to write
+ * @xcc_id: xcc accelerated compute core id
  *
  * this function is invoked only for the debugfs register access
  */
 void amdgpu_mm_wreg_mmio_rlc(struct amdgpu_device *adev,
-			     uint32_t reg, uint32_t v)
+			     uint32_t reg, uint32_t v,
+			     uint32_t xcc_id)
 {
 	if (amdgpu_device_skip_hw_access(adev))
 		return;
@@ -580,7 +893,7 @@ void amdgpu_mm_wreg_mmio_rlc(struct amdgpu_device *adev,
 	    adev->gfx.rlc.funcs &&
 	    adev->gfx.rlc.funcs->is_rlcg_access_range) {
 		if (adev->gfx.rlc.funcs->is_rlcg_access_range(adev, reg))
-			return amdgpu_sriov_wreg(adev, reg, v, 0, 0);
+			return amdgpu_sriov_wreg(adev, reg, v, 0, 0, xcc_id);
 	} else if ((reg * 4) >= adev->rmmio_size) {
 		adev->pcie_wreg(adev, reg * 4, v);
 	} else {
@@ -589,108 +902,144 @@ void amdgpu_mm_wreg_mmio_rlc(struct amdgpu_device *adev,
 }
 
 /**
- * amdgpu_mm_rdoorbell - read a doorbell dword
+ * amdgpu_device_xcc_wreg - write to a memory mapped IO or indirect register with specific XCC
  *
  * @adev: amdgpu_device pointer
- * @index: doorbell index
+ * @reg: dword aligned register offset
+ * @v: 32 bit value to write to the register
+ * @acc_flags: access flags which require special behavior
+ * @xcc_id: xcc accelerated compute core id
  *
- * Returns the value in the doorbell aperture at the
- * requested doorbell index (CIK).
+ * Writes the value specified to the offset specified.
  */
-u32 amdgpu_mm_rdoorbell(struct amdgpu_device *adev, u32 index)
+void amdgpu_device_xcc_wreg(struct amdgpu_device *adev,
+			uint32_t reg, uint32_t v,
+			uint32_t acc_flags, uint32_t xcc_id)
 {
+	uint32_t rlcg_flag;
+
 	if (amdgpu_device_skip_hw_access(adev))
-		return 0;
+		return;
 
-	if (index < adev->doorbell.num_kernel_doorbells) {
-		return readl(adev->doorbell.ptr + index);
+	if ((reg * 4) < adev->rmmio_size) {
+		if (amdgpu_sriov_vf(adev) &&
+		    !amdgpu_sriov_runtime(adev) &&
+		    adev->gfx.rlc.rlcg_reg_access_supported &&
+		    amdgpu_virt_get_rlcg_reg_access_flag(adev, acc_flags,
+							 GC_HWIP, true,
+							 &rlcg_flag)) {
+			amdgpu_virt_rlcg_reg_rw(adev, reg, v, rlcg_flag, GET_INST(GC, xcc_id));
+		} else if (!(acc_flags & AMDGPU_REGS_NO_KIQ) &&
+		    amdgpu_sriov_runtime(adev) &&
+		    down_read_trylock(&adev->reset_domain->sem)) {
+			amdgpu_kiq_wreg(adev, reg, v, xcc_id);
+			up_read(&adev->reset_domain->sem);
+		} else {
+			writel(v, ((void __iomem *)adev->rmmio) + (reg * 4));
+		}
 	} else {
-		DRM_ERROR("reading beyond doorbell aperture: 0x%08x!\n", index);
-		return 0;
+		adev->pcie_wreg(adev, reg * 4, v);
 	}
 }
 
 /**
- * amdgpu_mm_wdoorbell - write a doorbell dword
+ * amdgpu_device_indirect_rreg - read an indirect register
  *
  * @adev: amdgpu_device pointer
- * @index: doorbell index
- * @v: value to write
+ * @reg_addr: indirect register address to read from
  *
- * Writes @v to the doorbell aperture at the
- * requested doorbell index (CIK).
+ * Returns the value of indirect register @reg_addr
  */
-void amdgpu_mm_wdoorbell(struct amdgpu_device *adev, u32 index, u32 v)
+u32 amdgpu_device_indirect_rreg(struct amdgpu_device *adev,
+				u32 reg_addr)
 {
-	if (amdgpu_device_skip_hw_access(adev))
-		return;
+	unsigned long flags, pcie_index, pcie_data;
+	void __iomem *pcie_index_offset;
+	void __iomem *pcie_data_offset;
+	u32 r;
 
-	if (index < adev->doorbell.num_kernel_doorbells) {
-		writel(v, adev->doorbell.ptr + index);
-	} else {
-		DRM_ERROR("writing beyond doorbell aperture: 0x%08x!\n", index);
-	}
+	pcie_index = adev->nbio.funcs->get_pcie_index_offset(adev);
+	pcie_data = adev->nbio.funcs->get_pcie_data_offset(adev);
+
+	spin_lock_irqsave(&adev->pcie_idx_lock, flags);
+	pcie_index_offset = (void __iomem *)adev->rmmio + pcie_index * 4;
+	pcie_data_offset = (void __iomem *)adev->rmmio + pcie_data * 4;
+
+	writel(reg_addr, pcie_index_offset);
+	readl(pcie_index_offset);
+	r = readl(pcie_data_offset);
+	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
+
+	return r;
 }
 
-/**
- * amdgpu_mm_rdoorbell64 - read a doorbell Qword
- *
- * @adev: amdgpu_device pointer
- * @index: doorbell index
- *
- * Returns the value in the doorbell aperture at the
- * requested doorbell index (VEGA10+).
- */
-u64 amdgpu_mm_rdoorbell64(struct amdgpu_device *adev, u32 index)
+u32 amdgpu_device_indirect_rreg_ext(struct amdgpu_device *adev,
+				    u64 reg_addr)
 {
-	if (amdgpu_device_skip_hw_access(adev))
-		return 0;
+	unsigned long flags, pcie_index, pcie_index_hi, pcie_data;
+	u32 r;
+	void __iomem *pcie_index_offset;
+	void __iomem *pcie_index_hi_offset;
+	void __iomem *pcie_data_offset;
 
-	if (index < adev->doorbell.num_kernel_doorbells) {
-		return atomic64_read((atomic64_t *)(adev->doorbell.ptr + index));
+	if (unlikely(!adev->nbio.funcs)) {
+		pcie_index = AMDGPU_PCIE_INDEX_FALLBACK;
+		pcie_data = AMDGPU_PCIE_DATA_FALLBACK;
 	} else {
-		DRM_ERROR("reading beyond doorbell aperture: 0x%08x!\n", index);
-		return 0;
+		pcie_index = adev->nbio.funcs->get_pcie_index_offset(adev);
+		pcie_data = adev->nbio.funcs->get_pcie_data_offset(adev);
 	}
-}
 
-/**
- * amdgpu_mm_wdoorbell64 - write a doorbell Qword
- *
- * @adev: amdgpu_device pointer
- * @index: doorbell index
- * @v: value to write
- *
- * Writes @v to the doorbell aperture at the
- * requested doorbell index (VEGA10+).
- */
-void amdgpu_mm_wdoorbell64(struct amdgpu_device *adev, u32 index, u64 v)
-{
-	if (amdgpu_device_skip_hw_access(adev))
-		return;
-
-	if (index < adev->doorbell.num_kernel_doorbells) {
-		atomic64_set((atomic64_t *)(adev->doorbell.ptr + index), v);
+	if (reg_addr >> 32) {
+		if (unlikely(!adev->nbio.funcs))
+			pcie_index_hi = AMDGPU_PCIE_INDEX_HI_FALLBACK;
+		else
+			pcie_index_hi = adev->nbio.funcs->get_pcie_index_hi_offset(adev);
 	} else {
-		DRM_ERROR("writing beyond doorbell aperture: 0x%08x!\n", index);
+		pcie_index_hi = 0;
+	}
+
+	spin_lock_irqsave(&adev->pcie_idx_lock, flags);
+	pcie_index_offset = (void __iomem *)adev->rmmio + pcie_index * 4;
+	pcie_data_offset = (void __iomem *)adev->rmmio + pcie_data * 4;
+	if (pcie_index_hi != 0)
+		pcie_index_hi_offset = (void __iomem *)adev->rmmio +
+				pcie_index_hi * 4;
+
+	writel(reg_addr, pcie_index_offset);
+	readl(pcie_index_offset);
+	if (pcie_index_hi != 0) {
+		writel((reg_addr >> 32) & 0xff, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
 	}
+	r = readl(pcie_data_offset);
+
+	/* clear the high bits */
+	if (pcie_index_hi != 0) {
+		writel(0, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
+	}
+
+	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
+
+	return r;
 }
 
 /**
- * amdgpu_device_indirect_rreg - read an indirect register
+ * amdgpu_device_indirect_rreg64 - read a 64bits indirect register
  *
  * @adev: amdgpu_device pointer
  * @reg_addr: indirect register address to read from
  *
  * Returns the value of indirect register @reg_addr
  */
-u32 amdgpu_device_indirect_rreg(struct amdgpu_device *adev,
-				u32 reg_addr)
+u64 amdgpu_device_indirect_rreg64(struct amdgpu_device *adev,
+				  u32 reg_addr)
 {
 	unsigned long flags, pcie_index, pcie_data;
 	void __iomem *pcie_index_offset;
 	void __iomem *pcie_data_offset;
-	u32 r;
+	u64 r;
 
 	pcie_index = adev->nbio.funcs->get_pcie_index_offset(adev);
 	pcie_data = adev->nbio.funcs->get_pcie_data_offset(adev);
@@ -699,45 +1048,64 @@ u32 amdgpu_device_indirect_rreg(struct amdgpu_device *adev,
 	pcie_index_offset = (void __iomem *)adev->rmmio + pcie_index * 4;
 	pcie_data_offset = (void __iomem *)adev->rmmio + pcie_data * 4;
 
+	/* read low 32 bits */
 	writel(reg_addr, pcie_index_offset);
 	readl(pcie_index_offset);
 	r = readl(pcie_data_offset);
+	/* read high 32 bits */
+	writel(reg_addr + 4, pcie_index_offset);
+	readl(pcie_index_offset);
+	r |= ((u64)readl(pcie_data_offset) << 32);
 	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
 
 	return r;
 }
 
-/**
- * amdgpu_device_indirect_rreg64 - read a 64bits indirect register
- *
- * @adev: amdgpu_device pointer
- * @reg_addr: indirect register address to read from
- *
- * Returns the value of indirect register @reg_addr
- */
-u64 amdgpu_device_indirect_rreg64(struct amdgpu_device *adev,
-				  u32 reg_addr)
+u64 amdgpu_device_indirect_rreg64_ext(struct amdgpu_device *adev,
+				  u64 reg_addr)
 {
 	unsigned long flags, pcie_index, pcie_data;
+	unsigned long pcie_index_hi = 0;
 	void __iomem *pcie_index_offset;
+	void __iomem *pcie_index_hi_offset;
 	void __iomem *pcie_data_offset;
 	u64 r;
 
 	pcie_index = adev->nbio.funcs->get_pcie_index_offset(adev);
 	pcie_data = adev->nbio.funcs->get_pcie_data_offset(adev);
+	if ((reg_addr >> 32) && (adev->nbio.funcs->get_pcie_index_hi_offset))
+		pcie_index_hi = adev->nbio.funcs->get_pcie_index_hi_offset(adev);
 
 	spin_lock_irqsave(&adev->pcie_idx_lock, flags);
 	pcie_index_offset = (void __iomem *)adev->rmmio + pcie_index * 4;
 	pcie_data_offset = (void __iomem *)adev->rmmio + pcie_data * 4;
+	if (pcie_index_hi != 0)
+		pcie_index_hi_offset = (void __iomem *)adev->rmmio +
+			pcie_index_hi * 4;
 
 	/* read low 32 bits */
 	writel(reg_addr, pcie_index_offset);
 	readl(pcie_index_offset);
+	if (pcie_index_hi != 0) {
+		writel((reg_addr >> 32) & 0xff, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
+	}
 	r = readl(pcie_data_offset);
 	/* read high 32 bits */
 	writel(reg_addr + 4, pcie_index_offset);
 	readl(pcie_index_offset);
+	if (pcie_index_hi != 0) {
+		writel((reg_addr >> 32) & 0xff, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
+	}
 	r |= ((u64)readl(pcie_data_offset) << 32);
+
+	/* clear the high bits */
+	if (pcie_index_hi != 0) {
+		writel(0, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
+	}
+
 	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
 
 	return r;
@@ -747,8 +1115,6 @@ u64 amdgpu_device_indirect_rreg64(struct amdgpu_device *adev,
  * amdgpu_device_indirect_wreg - write an indirect register address
  *
  * @adev: amdgpu_device pointer
- * @pcie_index: mmio register offset
- * @pcie_data: mmio register offset
  * @reg_addr: indirect register offset
  * @reg_data: indirect register data
  *
@@ -774,12 +1140,50 @@ void amdgpu_device_indirect_wreg(struct amdgpu_device *adev,
 	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
 }
 
+void amdgpu_device_indirect_wreg_ext(struct amdgpu_device *adev,
+				     u64 reg_addr, u32 reg_data)
+{
+	unsigned long flags, pcie_index, pcie_index_hi, pcie_data;
+	void __iomem *pcie_index_offset;
+	void __iomem *pcie_index_hi_offset;
+	void __iomem *pcie_data_offset;
+
+	pcie_index = adev->nbio.funcs->get_pcie_index_offset(adev);
+	pcie_data = adev->nbio.funcs->get_pcie_data_offset(adev);
+	if ((reg_addr >> 32) && (adev->nbio.funcs->get_pcie_index_hi_offset))
+		pcie_index_hi = adev->nbio.funcs->get_pcie_index_hi_offset(adev);
+	else
+		pcie_index_hi = 0;
+
+	spin_lock_irqsave(&adev->pcie_idx_lock, flags);
+	pcie_index_offset = (void __iomem *)adev->rmmio + pcie_index * 4;
+	pcie_data_offset = (void __iomem *)adev->rmmio + pcie_data * 4;
+	if (pcie_index_hi != 0)
+		pcie_index_hi_offset = (void __iomem *)adev->rmmio +
+				pcie_index_hi * 4;
+
+	writel(reg_addr, pcie_index_offset);
+	readl(pcie_index_offset);
+	if (pcie_index_hi != 0) {
+		writel((reg_addr >> 32) & 0xff, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
+	}
+	writel(reg_data, pcie_data_offset);
+	readl(pcie_data_offset);
+
+	/* clear the high bits */
+	if (pcie_index_hi != 0) {
+		writel(0, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
+	}
+
+	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
+}
+
 /**
  * amdgpu_device_indirect_wreg64 - write a 64bits indirect register address
  *
  * @adev: amdgpu_device pointer
- * @pcie_index: mmio register offset
- * @pcie_data: mmio register offset
  * @reg_addr: indirect register offset
  * @reg_data: indirect register data
  *
@@ -811,6 +1215,55 @@ void amdgpu_device_indirect_wreg64(struct amdgpu_device *adev,
 	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
 }
 
+void amdgpu_device_indirect_wreg64_ext(struct amdgpu_device *adev,
+				   u64 reg_addr, u64 reg_data)
+{
+	unsigned long flags, pcie_index, pcie_data;
+	unsigned long pcie_index_hi = 0;
+	void __iomem *pcie_index_offset;
+	void __iomem *pcie_index_hi_offset;
+	void __iomem *pcie_data_offset;
+
+	pcie_index = adev->nbio.funcs->get_pcie_index_offset(adev);
+	pcie_data = adev->nbio.funcs->get_pcie_data_offset(adev);
+	if ((reg_addr >> 32) && (adev->nbio.funcs->get_pcie_index_hi_offset))
+		pcie_index_hi = adev->nbio.funcs->get_pcie_index_hi_offset(adev);
+
+	spin_lock_irqsave(&adev->pcie_idx_lock, flags);
+	pcie_index_offset = (void __iomem *)adev->rmmio + pcie_index * 4;
+	pcie_data_offset = (void __iomem *)adev->rmmio + pcie_data * 4;
+	if (pcie_index_hi != 0)
+		pcie_index_hi_offset = (void __iomem *)adev->rmmio +
+				pcie_index_hi * 4;
+
+	/* write low 32 bits */
+	writel(reg_addr, pcie_index_offset);
+	readl(pcie_index_offset);
+	if (pcie_index_hi != 0) {
+		writel((reg_addr >> 32) & 0xff, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
+	}
+	writel((u32)(reg_data & 0xffffffffULL), pcie_data_offset);
+	readl(pcie_data_offset);
+	/* write high 32 bits */
+	writel(reg_addr + 4, pcie_index_offset);
+	readl(pcie_index_offset);
+	if (pcie_index_hi != 0) {
+		writel((reg_addr >> 32) & 0xff, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
+	}
+	writel((u32)(reg_data >> 32), pcie_data_offset);
+	readl(pcie_data_offset);
+
+	/* clear the high bits */
+	if (pcie_index_hi != 0) {
+		writel(0, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
+	}
+
+	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
+}
+
 /**
  * amdgpu_device_get_rev_id - query device rev_id
  *
@@ -835,7 +1288,14 @@ u32 amdgpu_device_get_rev_id(struct amdgpu_device *adev)
  */
 static uint32_t amdgpu_invalid_rreg(struct amdgpu_device *adev, uint32_t reg)
 {
-	DRM_ERROR("Invalid callback to read register 0x%04X\n", reg);
+	dev_err(adev->dev, "Invalid callback to read register 0x%04X\n", reg);
+	BUG();
+	return 0;
+}
+
+static uint32_t amdgpu_invalid_rreg_ext(struct amdgpu_device *adev, uint64_t reg)
+{
+	dev_err(adev->dev, "Invalid callback to read register 0x%llX\n", reg);
 	BUG();
 	return 0;
 }
@@ -852,8 +1312,17 @@ static uint32_t amdgpu_invalid_rreg(struct amdgpu_device *adev, uint32_t reg)
  */
 static void amdgpu_invalid_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v)
 {
-	DRM_ERROR("Invalid callback to write register 0x%04X with 0x%08X\n",
-		  reg, v);
+	dev_err(adev->dev,
+		"Invalid callback to write register 0x%04X with 0x%08X\n", reg,
+		v);
+	BUG();
+}
+
+static void amdgpu_invalid_wreg_ext(struct amdgpu_device *adev, uint64_t reg, uint32_t v)
+{
+	dev_err(adev->dev,
+		"Invalid callback to write register 0x%llX with 0x%08X\n", reg,
+		v);
 	BUG();
 }
 
@@ -869,7 +1338,15 @@ static void amdgpu_invalid_wreg(struct amdgpu_device *adev, uint32_t reg, uint32
  */
 static uint64_t amdgpu_invalid_rreg64(struct amdgpu_device *adev, uint32_t reg)
 {
-	DRM_ERROR("Invalid callback to read 64 bit register 0x%04X\n", reg);
+	dev_err(adev->dev, "Invalid callback to read 64 bit register 0x%04X\n",
+		reg);
+	BUG();
+	return 0;
+}
+
+static uint64_t amdgpu_invalid_rreg64_ext(struct amdgpu_device *adev, uint64_t reg)
+{
+	dev_err(adev->dev, "Invalid callback to read register 0x%llX\n", reg);
 	BUG();
 	return 0;
 }
@@ -886,8 +1363,17 @@ static uint64_t amdgpu_invalid_rreg64(struct amdgpu_device *adev, uint32_t reg)
  */
 static void amdgpu_invalid_wreg64(struct amdgpu_device *adev, uint32_t reg, uint64_t v)
 {
-	DRM_ERROR("Invalid callback to write 64 bit register 0x%04X with 0x%08llX\n",
-		  reg, v);
+	dev_err(adev->dev,
+		"Invalid callback to write 64 bit register 0x%04X with 0x%08llX\n",
+		reg, v);
+	BUG();
+}
+
+static void amdgpu_invalid_wreg64_ext(struct amdgpu_device *adev, uint64_t reg, uint64_t v)
+{
+	dev_err(adev->dev,
+		"Invalid callback to write 64 bit register 0x%llX with 0x%08llX\n",
+		reg, v);
 	BUG();
 }
 
@@ -905,8 +1391,9 @@ static void amdgpu_invalid_wreg64(struct amdgpu_device *adev, uint32_t reg, uint
 static uint32_t amdgpu_block_invalid_rreg(struct amdgpu_device *adev,
 					  uint32_t block, uint32_t reg)
 {
-	DRM_ERROR("Invalid callback to read register 0x%04X in block 0x%04X\n",
-		  reg, block);
+	dev_err(adev->dev,
+		"Invalid callback to read register 0x%04X in block 0x%04X\n",
+		reg, block);
 	BUG();
 	return 0;
 }
@@ -926,11 +1413,23 @@ static void amdgpu_block_invalid_wreg(struct amdgpu_device *adev,
 				      uint32_t block,
 				      uint32_t reg, uint32_t v)
 {
-	DRM_ERROR("Invalid block callback to write register 0x%04X in block 0x%04X with 0x%08X\n",
-		  reg, block, v);
+	dev_err(adev->dev,
+		"Invalid block callback to write register 0x%04X in block 0x%04X with 0x%08X\n",
+		reg, block, v);
 	BUG();
 }
 
+static uint32_t amdgpu_device_get_vbios_flags(struct amdgpu_device *adev)
+{
+	if (hweight32(adev->aid_mask) && (adev->flags & AMD_IS_APU))
+		return AMDGPU_VBIOS_SKIP;
+
+	if (hweight32(adev->aid_mask) && amdgpu_passthrough(adev))
+		return AMDGPU_VBIOS_OPTIONAL;
+
+	return 0;
+}
+
 /**
  * amdgpu_device_asic_init - Wrapper for atom asic_init
  *
@@ -940,12 +1439,32 @@ static void amdgpu_block_invalid_wreg(struct amdgpu_device *adev,
  */
 static int amdgpu_device_asic_init(struct amdgpu_device *adev)
 {
+	uint32_t flags;
+	bool optional;
+	int ret;
+
 	amdgpu_asic_pre_asic_init(adev);
+	flags = amdgpu_device_get_vbios_flags(adev);
+	optional = !!(flags & (AMDGPU_VBIOS_OPTIONAL | AMDGPU_VBIOS_SKIP));
+
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 4) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 5, 0) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(11, 0, 0)) {
+		amdgpu_psp_wait_for_bootloader(adev);
+		if (optional && !adev->bios)
+			return 0;
+
+		ret = amdgpu_atomfirmware_asic_init(adev, true);
+		return ret;
+	} else {
+		if (optional && !adev->bios)
+			return 0;
 
-	if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(11, 0, 0))
-		return amdgpu_atomfirmware_asic_init(adev, true);
-	else
 		return amdgpu_atom_asic_init(adev->mode_info.atom_context);
+	}
+
+	return 0;
 }
 
 /**
@@ -985,7 +1504,7 @@ static void amdgpu_device_mem_scratch_fini(struct amdgpu_device *adev)
  * @registers: pointer to the register array
  * @array_size: size of the register array
  *
- * Programs an array or registers with and and or masks.
+ * Programs an array or registers with and or masks.
  * This is a helper for setting golden registers.
  */
 void amdgpu_device_program_register_sequence(struct amdgpu_device *adev,
@@ -998,7 +1517,7 @@ void amdgpu_device_program_register_sequence(struct amdgpu_device *adev,
 	if (array_size % 3)
 		return;
 
-	for (i = 0; i < array_size; i +=3) {
+	for (i = 0; i < array_size; i += 3) {
 		reg = registers[i + 0];
 		and_mask = registers[i + 1];
 		or_mask = registers[i + 2];
@@ -1043,82 +1562,6 @@ int amdgpu_device_pci_reset(struct amdgpu_device *adev)
 }
 
 /*
- * GPU doorbell aperture helpers function.
- */
-/**
- * amdgpu_device_doorbell_init - Init doorbell driver information.
- *
- * @adev: amdgpu_device pointer
- *
- * Init doorbell driver information (CIK)
- * Returns 0 on success, error on failure.
- */
-static int amdgpu_device_doorbell_init(struct amdgpu_device *adev)
-{
-
-	/* No doorbell on SI hardware generation */
-	if (adev->asic_type < CHIP_BONAIRE) {
-		adev->doorbell.base = 0;
-		adev->doorbell.size = 0;
-		adev->doorbell.num_kernel_doorbells = 0;
-		adev->doorbell.ptr = NULL;
-		return 0;
-	}
-
-	if (pci_resource_flags(adev->pdev, 2) & IORESOURCE_UNSET)
-		return -EINVAL;
-
-	amdgpu_asic_init_doorbell_index(adev);
-
-	/* doorbell bar mapping */
-	adev->doorbell.base = pci_resource_start(adev->pdev, 2);
-	adev->doorbell.size = pci_resource_len(adev->pdev, 2);
-
-	if (adev->enable_mes) {
-		adev->doorbell.num_kernel_doorbells =
-			adev->doorbell.size / sizeof(u32);
-	} else {
-		adev->doorbell.num_kernel_doorbells =
-			min_t(u32, adev->doorbell.size / sizeof(u32),
-			      adev->doorbell_index.max_assignment+1);
-		if (adev->doorbell.num_kernel_doorbells == 0)
-			return -EINVAL;
-
-		/* For Vega, reserve and map two pages on doorbell BAR since SDMA
-		 * paging queue doorbell use the second page. The
-		 * AMDGPU_DOORBELL64_MAX_ASSIGNMENT definition assumes all the
-		 * doorbells are in the first page. So with paging queue enabled,
-		 * the max num_kernel_doorbells should + 1 page (0x400 in dword)
-		 */
-		if (adev->asic_type >= CHIP_VEGA10)
-			adev->doorbell.num_kernel_doorbells += 0x400;
-	}
-
-	adev->doorbell.ptr = ioremap(adev->doorbell.base,
-				     adev->doorbell.num_kernel_doorbells *
-				     sizeof(u32));
-	if (adev->doorbell.ptr == NULL)
-		return -ENOMEM;
-
-	return 0;
-}
-
-/**
- * amdgpu_device_doorbell_fini - Tear down doorbell driver information.
- *
- * @adev: amdgpu_device pointer
- *
- * Tear down doorbell driver information (CIK)
- */
-static void amdgpu_device_doorbell_fini(struct amdgpu_device *adev)
-{
-	iounmap(adev->doorbell.ptr);
-	adev->doorbell.ptr = NULL;
-}
-
-
-
-/*
  * amdgpu_device_wb_*()
  * Writeback is the method by which the GPU updates special pages in memory
  * with the status of certain GPU events (fences, ring pointers,etc.).
@@ -1187,13 +1630,17 @@ static int amdgpu_device_wb_init(struct amdgpu_device *adev)
  */
 int amdgpu_device_wb_get(struct amdgpu_device *adev, u32 *wb)
 {
-	unsigned long offset = find_first_zero_bit(adev->wb.used, adev->wb.num_wb);
+	unsigned long flags, offset;
 
+	spin_lock_irqsave(&adev->wb.lock, flags);
+	offset = find_first_zero_bit(adev->wb.used, adev->wb.num_wb);
 	if (offset < adev->wb.num_wb) {
 		__set_bit(offset, adev->wb.used);
+		spin_unlock_irqrestore(&adev->wb.lock, flags);
 		*wb = offset << 3; /* convert to dw offset */
 		return 0;
 	} else {
+		spin_unlock_irqrestore(&adev->wb.lock, flags);
 		return -EINVAL;
 	}
 }
@@ -1208,9 +1655,13 @@ int amdgpu_device_wb_get(struct amdgpu_device *adev, u32 *wb)
  */
 void amdgpu_device_wb_free(struct amdgpu_device *adev, u32 wb)
 {
+	unsigned long flags;
+
 	wb >>= 3;
+	spin_lock_irqsave(&adev->wb.lock, flags);
 	if (wb < adev->wb.num_wb)
 		__clear_bit(wb, adev->wb.used);
+	spin_unlock_irqrestore(&adev->wb.lock, flags);
 }
 
 /**
@@ -1227,14 +1678,33 @@ int amdgpu_device_resize_fb_bar(struct amdgpu_device *adev)
 	int rbar_size = pci_rebar_bytes_to_size(adev->gmc.real_vram_size);
 	struct pci_bus *root;
 	struct resource *res;
-	unsigned i;
+	int max_size, r;
+	unsigned int i;
 	u16 cmd;
-	int r;
+
+	if (!IS_ENABLED(CONFIG_PHYS_ADDR_T_64BIT))
+		return 0;
 
 	/* Bypass for VF */
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
+	if (!amdgpu_rebar)
+		return 0;
+
+	/* resizing on Dell G5 SE platforms causes problems with runtime pm */
+	if ((amdgpu_runtime_pm != 0) &&
+	    adev->pdev->vendor == PCI_VENDOR_ID_ATI &&
+	    adev->pdev->device == 0x731f &&
+	    adev->pdev->subsystem_vendor == PCI_VENDOR_ID_DELL)
+		return 0;
+
+	/* PCI_EXT_CAP_ID_VNDR extended capability is located at 0x100 */
+	if (!pci_find_ext_capability(adev->pdev, PCI_EXT_CAP_ID_VNDR))
+		dev_warn(
+			adev->dev,
+			"System can't access extended configuration space, please check!!\n");
+
 	/* skip if the bios has already enabled large BAR */
 	if (adev->gmc.real_vram_size &&
 	    (pci_resource_len(adev->pdev, 0) >= adev->gmc.real_vram_size))
@@ -1256,33 +1726,32 @@ int amdgpu_device_resize_fb_bar(struct amdgpu_device *adev)
 		return 0;
 
 	/* Limit the BAR size to what is available */
-	rbar_size = min(fls(pci_rebar_get_possible_sizes(adev->pdev, 0)) - 1,
-			rbar_size);
+	max_size = pci_rebar_get_max_size(adev->pdev, 0);
+	if (max_size < 0)
+		return 0;
+	rbar_size = min(max_size, rbar_size);
 
 	/* Disable memory decoding while we change the BAR addresses and size */
 	pci_read_config_word(adev->pdev, PCI_COMMAND, &cmd);
 	pci_write_config_word(adev->pdev, PCI_COMMAND,
 			      cmd & ~PCI_COMMAND_MEMORY);
 
-	/* Free the VRAM and doorbell BAR, we most likely need to move both. */
-	amdgpu_device_doorbell_fini(adev);
-	if (adev->asic_type >= CHIP_BONAIRE)
-		pci_release_resource(adev->pdev, 2);
-
-	pci_release_resource(adev->pdev, 0);
+	/* Tear down doorbell as resizing will release BARs */
+	amdgpu_doorbell_fini(adev);
 
-	r = pci_resize_resource(adev->pdev, 0, rbar_size);
+	r = pci_resize_resource(adev->pdev, 0, rbar_size,
+				(adev->asic_type >= CHIP_BONAIRE) ? 1 << 5
+								  : 1 << 2);
 	if (r == -ENOSPC)
-		DRM_INFO("Not enough PCI address space for a large BAR.");
+		dev_info(adev->dev,
+			 "Not enough PCI address space for a large BAR.");
 	else if (r && r != -ENOTSUPP)
-		DRM_ERROR("Problem resizing BAR0 (%d).", r);
-
-	pci_assign_unassigned_bus_resources(adev->pdev->bus);
+		dev_err(adev->dev, "Problem resizing BAR0 (%d).", r);
 
 	/* When the doorbell or fb BAR isn't available we have no chance of
 	 * using the device.
 	 */
-	r = amdgpu_device_doorbell_init(adev);
+	r = amdgpu_doorbell_init(adev);
 	if (r || (pci_resource_flags(adev->pdev, 0) & IORESOURCE_UNSET))
 		return -ENODEV;
 
@@ -1305,11 +1774,17 @@ int amdgpu_device_resize_fb_bar(struct amdgpu_device *adev)
  */
 bool amdgpu_device_need_post(struct amdgpu_device *adev)
 {
-	uint32_t reg;
+	uint32_t reg, flags;
 
 	if (amdgpu_sriov_vf(adev))
 		return false;
 
+	flags = amdgpu_device_get_vbios_flags(adev);
+	if (flags & AMDGPU_VBIOS_SKIP)
+		return false;
+	if ((flags & AMDGPU_VBIOS_OPTIONAL) && !adev->bios)
+		return false;
+
 	if (amdgpu_passthrough(adev)) {
 		/* for FIJI: In whole GPU pass-through virtualization case, after VM reboot
 		 * some old smc fw still need driver do vPost otherwise gpu hang, while
@@ -1319,19 +1794,21 @@ bool amdgpu_device_need_post(struct amdgpu_device *adev)
 		if (adev->asic_type == CHIP_FIJI) {
 			int err;
 			uint32_t fw_ver;
+
 			err = request_firmware(&adev->pm.fw, "amdgpu/fiji_smc.bin", adev->dev);
-			/* force vPost if error occured */
+			/* force vPost if error occurred */
 			if (err)
 				return true;
 
 			fw_ver = *((uint32_t *)adev->pm.fw->data + 69);
+			release_firmware(adev->pm.fw);
 			if (fw_ver < 0x00160e00)
 				return true;
 		}
 	}
 
 	/* Don't post if we need to reset whole hive on init */
-	if (adev->gmc.xgmi.pending_reset)
+	if (adev->init_lvl->level == AMDGPU_INIT_LEVEL_MINIMAL_XGMI)
 		return false;
 
 	if (adev->has_hw_reset) {
@@ -1352,6 +1829,96 @@ bool amdgpu_device_need_post(struct amdgpu_device *adev)
 	return true;
 }
 
+/*
+ * Check whether seamless boot is supported.
+ *
+ * So far we only support seamless boot on DCE 3.0 or later.
+ * If users report that it works on older ASICS as well, we may
+ * loosen this.
+ */
+bool amdgpu_device_seamless_boot_supported(struct amdgpu_device *adev)
+{
+	switch (amdgpu_seamless) {
+	case -1:
+		break;
+	case 1:
+		return true;
+	case 0:
+		return false;
+	default:
+		dev_err(adev->dev, "Invalid value for amdgpu.seamless: %d\n",
+			amdgpu_seamless);
+		return false;
+	}
+
+	if (!(adev->flags & AMD_IS_APU))
+		return false;
+
+	if (adev->mman.keep_stolen_vga_memory)
+		return false;
+
+	return amdgpu_ip_version(adev, DCE_HWIP, 0) >= IP_VERSION(3, 0, 0);
+}
+
+/*
+ * Intel hosts such as Rocket Lake, Alder Lake, Raptor Lake and Sapphire Rapids
+ * don't support dynamic speed switching. Until we have confirmation from Intel
+ * that a specific host supports it, it's safer that we keep it disabled for all.
+ *
+ * https://edc.intel.com/content/www/us/en/design/products/platforms/details/raptor-lake-s/13th-generation-core-processors-datasheet-volume-1-of-2/005/pci-express-support/
+ * https://gitlab.freedesktop.org/drm/amd/-/issues/2663
+ */
+static bool amdgpu_device_pcie_dynamic_switching_supported(struct amdgpu_device *adev)
+{
+#if IS_ENABLED(CONFIG_X86)
+	struct cpuinfo_x86 *c = &cpu_data(0);
+
+	/* eGPU change speeds based on USB4 fabric conditions */
+	if (dev_is_removable(adev->dev))
+		return true;
+
+	if (c->x86_vendor == X86_VENDOR_INTEL)
+		return false;
+#endif
+	return true;
+}
+
+static bool amdgpu_device_aspm_support_quirk(struct amdgpu_device *adev)
+{
+	/* Enabling ASPM causes randoms hangs on Tahiti and Oland on Zen4.
+	 * It's unclear if this is a platform-specific or GPU-specific issue.
+	 * Disable ASPM on SI for the time being.
+	 */
+	if (adev->family == AMDGPU_FAMILY_SI)
+		return true;
+
+#if IS_ENABLED(CONFIG_X86)
+	struct cpuinfo_x86 *c = &cpu_data(0);
+
+	if (!(amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(12, 0, 0) ||
+		  amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(12, 0, 1)))
+		return false;
+
+	if (c->x86 == 6 &&
+		adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_GEN5) {
+		switch (c->x86_model) {
+		case VFM_MODEL(INTEL_ALDERLAKE):
+		case VFM_MODEL(INTEL_ALDERLAKE_L):
+		case VFM_MODEL(INTEL_RAPTORLAKE):
+		case VFM_MODEL(INTEL_RAPTORLAKE_P):
+		case VFM_MODEL(INTEL_RAPTORLAKE_S):
+			return true;
+		default:
+			return false;
+		}
+	} else {
+		return false;
+	}
+#else
+	return false;
+#endif
+}
+
 /**
  * amdgpu_device_should_use_aspm - check if the device should program ASPM
  *
@@ -1374,20 +1941,13 @@ bool amdgpu_device_should_use_aspm(struct amdgpu_device *adev)
 	default:
 		return false;
 	}
+	if (adev->flags & AMD_IS_APU)
+		return false;
+	if (amdgpu_device_aspm_support_quirk(adev))
+		return false;
 	return pcie_aspm_enabled(adev->pdev);
 }
 
-bool amdgpu_device_aspm_support_quirk(void)
-{
-#if IS_ENABLED(CONFIG_X86)
-	struct cpuinfo_x86 *c = &cpu_data(0);
-
-	return !(c->x86 == 6 && c->x86_model == INTEL_FAM6_ALDERLAKE);
-#else
-	return true;
-#endif
-}
-
 /* if we get transitioned to only one device, take VGA back */
 /**
  * amdgpu_device_vga_set_decode - enable/disable vga decode
@@ -1402,6 +1962,7 @@ static unsigned int amdgpu_device_vga_set_decode(struct pci_dev *pdev,
 		bool state)
 {
 	struct amdgpu_device *adev = drm_to_adev(pci_get_drvdata(pdev));
+
 	amdgpu_asic_set_vga_state(adev, state);
 	if (state)
 		return VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM |
@@ -1424,7 +1985,8 @@ static void amdgpu_device_check_block_size(struct amdgpu_device *adev)
 {
 	/* defines number of bits in page table versus page directory,
 	 * a page is 4KB so we have 12 bits offset, minimum 9 bits in the
-	 * page table and the remaining bits are in the page directory */
+	 * page table and the remaining bits are in the page directory
+	 */
 	if (amdgpu_vm_block_size == -1)
 		return;
 
@@ -1468,7 +2030,7 @@ static void amdgpu_device_check_smu_prv_buffer_size(struct amdgpu_device *adev)
 		return;
 
 	if (!is_os_64) {
-		DRM_WARN("Not 64-bit OS, feature not supported\n");
+		dev_warn(adev->dev, "Not 64-bit OS, feature not supported\n");
 		goto def_value;
 	}
 	si_meminfo(&si);
@@ -1483,7 +2045,7 @@ static void amdgpu_device_check_smu_prv_buffer_size(struct amdgpu_device *adev)
 		if (total_memory < dram_size_seven_GB)
 			goto def_value1;
 	} else {
-		DRM_WARN("Smu memory pool size not supported\n");
+		dev_warn(adev->dev, "Smu memory pool size not supported\n");
 		goto def_value;
 	}
 	adev->pm.smu_prv_buffer_size = amdgpu_smu_memory_pool_size << 28;
@@ -1491,7 +2053,7 @@ static void amdgpu_device_check_smu_prv_buffer_size(struct amdgpu_device *adev)
 	return;
 
 def_value1:
-	DRM_WARN("No enough system memory\n");
+	dev_warn(adev->dev, "No enough system memory\n");
 def_value:
 	adev->pm.smu_prv_buffer_size = 0;
 }
@@ -1543,11 +2105,13 @@ static int amdgpu_device_init_apu_flags(struct amdgpu_device *adev)
  */
 static int amdgpu_device_check_arguments(struct amdgpu_device *adev)
 {
+	int i;
+
 	if (amdgpu_sched_jobs < 4) {
 		dev_warn(adev->dev, "sched jobs (%d) must be at least 4\n",
 			 amdgpu_sched_jobs);
 		amdgpu_sched_jobs = 4;
-	} else if (!is_power_of_2(amdgpu_sched_jobs)){
+	} else if (!is_power_of_2(amdgpu_sched_jobs)) {
 		dev_warn(adev->dev, "sched jobs (%d) must be a power of 2\n",
 			 amdgpu_sched_jobs);
 		amdgpu_sched_jobs = roundup_pow_of_two(amdgpu_sched_jobs);
@@ -1597,6 +2161,32 @@ static int amdgpu_device_check_arguments(struct amdgpu_device *adev)
 
 	adev->firmware.load_type = amdgpu_ucode_get_load_type(adev, amdgpu_fw_load_type);
 
+	for (i = 0; i < MAX_XCP; i++) {
+		switch (amdgpu_enforce_isolation) {
+		case -1:
+		case 0:
+		default:
+			/* disable */
+			adev->enforce_isolation[i] = AMDGPU_ENFORCE_ISOLATION_DISABLE;
+			break;
+		case 1:
+			/* enable */
+			adev->enforce_isolation[i] =
+				AMDGPU_ENFORCE_ISOLATION_ENABLE;
+			break;
+		case 2:
+			/* enable legacy mode */
+			adev->enforce_isolation[i] =
+				AMDGPU_ENFORCE_ISOLATION_ENABLE_LEGACY;
+			break;
+		case 3:
+			/* enable only process isolation without submitting cleaner shader */
+			adev->enforce_isolation[i] =
+				AMDGPU_ENFORCE_ISOLATION_NO_CLEANER_SHADER;
+			break;
+		}
+	}
+
 	return 0;
 }
 
@@ -1615,7 +2205,8 @@ static void amdgpu_switcheroo_set_state(struct pci_dev *pdev,
 	struct drm_device *dev = pci_get_drvdata(pdev);
 	int r;
 
-	if (amdgpu_device_supports_px(dev) && state == VGA_SWITCHEROO_OFF)
+	if (amdgpu_device_supports_px(drm_to_adev(dev)) &&
+	    state == VGA_SWITCHEROO_OFF)
 		return;
 
 	if (state == VGA_SWITCHEROO_ON) {
@@ -1627,13 +2218,15 @@ static void amdgpu_switcheroo_set_state(struct pci_dev *pdev,
 		amdgpu_device_load_pci_state(pdev);
 		r = pci_enable_device(pdev);
 		if (r)
-			DRM_WARN("pci_enable_device failed (%d)\n", r);
+			dev_warn(&pdev->dev, "pci_enable_device failed (%d)\n",
+				 r);
 		amdgpu_device_resume(dev, true);
 
 		dev->switch_power_state = DRM_SWITCH_POWER_ON;
 	} else {
-		pr_info("switched off\n");
+		dev_info(&pdev->dev, "switched off\n");
 		dev->switch_power_state = DRM_SWITCH_POWER_CHANGING;
+		amdgpu_device_prepare(dev);
 		amdgpu_device_suspend(dev, true);
 		amdgpu_device_cache_pci_state(pdev);
 		/* Shut down the device */
@@ -1656,7 +2249,7 @@ static bool amdgpu_switcheroo_can_switch(struct pci_dev *pdev)
 {
 	struct drm_device *dev = pci_get_drvdata(pdev);
 
-	/*
+       /*
 	* FIXME: open_count is protected by drm_global_mutex but that would lead to
 	* locking inversion with the driver load path. And the access here is
 	* completely racy anyway. So don't bother with locking for now.
@@ -1696,10 +2289,11 @@ int amdgpu_device_ip_set_clockgating_state(void *dev,
 		if (!adev->ip_blocks[i].version->funcs->set_clockgating_state)
 			continue;
 		r = adev->ip_blocks[i].version->funcs->set_clockgating_state(
-			(void *)adev, state);
+			&adev->ip_blocks[i], state);
 		if (r)
-			DRM_ERROR("set_clockgating_state of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
+			dev_err(adev->dev,
+				"set_clockgating_state of IP block <%s> failed %d\n",
+				adev->ip_blocks[i].version->funcs->name, r);
 	}
 	return r;
 }
@@ -1730,10 +2324,11 @@ int amdgpu_device_ip_set_powergating_state(void *dev,
 		if (!adev->ip_blocks[i].version->funcs->set_powergating_state)
 			continue;
 		r = adev->ip_blocks[i].version->funcs->set_powergating_state(
-			(void *)adev, state);
+			&adev->ip_blocks[i], state);
 		if (r)
-			DRM_ERROR("set_powergating_state of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
+			dev_err(adev->dev,
+				"set_powergating_state of IP block <%s> failed %d\n",
+				adev->ip_blocks[i].version->funcs->name, r);
 	}
 	return r;
 }
@@ -1758,7 +2353,8 @@ void amdgpu_device_ip_get_clockgating_state(struct amdgpu_device *adev,
 		if (!adev->ip_blocks[i].status.valid)
 			continue;
 		if (adev->ip_blocks[i].version->funcs->get_clockgating_state)
-			adev->ip_blocks[i].version->funcs->get_clockgating_state((void *)adev, flags);
+			adev->ip_blocks[i].version->funcs->get_clockgating_state(
+				&adev->ip_blocks[i], flags);
 	}
 }
 
@@ -1780,9 +2376,12 @@ int amdgpu_device_ip_wait_for_idle(struct amdgpu_device *adev,
 		if (!adev->ip_blocks[i].status.valid)
 			continue;
 		if (adev->ip_blocks[i].version->type == block_type) {
-			r = adev->ip_blocks[i].version->funcs->wait_for_idle((void *)adev);
-			if (r)
-				return r;
+			if (adev->ip_blocks[i].version->funcs->wait_for_idle) {
+				r = adev->ip_blocks[i].version->funcs->wait_for_idle(
+								&adev->ip_blocks[i]);
+				if (r)
+					return r;
+			}
 			break;
 		}
 	}
@@ -1791,26 +2390,45 @@ int amdgpu_device_ip_wait_for_idle(struct amdgpu_device *adev,
 }
 
 /**
- * amdgpu_device_ip_is_idle - is the hardware IP idle
+ * amdgpu_device_ip_is_hw - is the hardware IP enabled
  *
  * @adev: amdgpu_device pointer
  * @block_type: Type of hardware IP (SMU, GFX, UVD, etc.)
  *
- * Check if the hardware IP is idle or not.
- * Returns true if it the IP is idle, false if not.
+ * Check if the hardware IP is enable or not.
+ * Returns true if it the IP is enable, false if not.
  */
-bool amdgpu_device_ip_is_idle(struct amdgpu_device *adev,
-			      enum amd_ip_block_type block_type)
+bool amdgpu_device_ip_is_hw(struct amdgpu_device *adev,
+			    enum amd_ip_block_type block_type)
 {
 	int i;
 
 	for (i = 0; i < adev->num_ip_blocks; i++) {
-		if (!adev->ip_blocks[i].status.valid)
-			continue;
 		if (adev->ip_blocks[i].version->type == block_type)
-			return adev->ip_blocks[i].version->funcs->is_idle((void *)adev);
+			return adev->ip_blocks[i].status.hw;
 	}
-	return true;
+	return false;
+}
+
+/**
+ * amdgpu_device_ip_is_valid - is the hardware IP valid
+ *
+ * @adev: amdgpu_device pointer
+ * @block_type: Type of hardware IP (SMU, GFX, UVD, etc.)
+ *
+ * Check if the hardware IP is valid or not.
+ * Returns true if it the IP is valid, false if not.
+ */
+bool amdgpu_device_ip_is_valid(struct amdgpu_device *adev,
+			       enum amd_ip_block_type block_type)
+{
+	int i;
+
+	for (i = 0; i < adev->num_ip_blocks; i++) {
+		if (adev->ip_blocks[i].version->type == block_type)
+			return adev->ip_blocks[i].status.valid;
+	}
+	return false;
 
 }
 
@@ -1861,6 +2479,34 @@ int amdgpu_device_ip_block_version_cmp(struct amdgpu_device *adev,
 	return 1;
 }
 
+static const char *ip_block_names[] = {
+	[AMD_IP_BLOCK_TYPE_COMMON] = "common",
+	[AMD_IP_BLOCK_TYPE_GMC] = "gmc",
+	[AMD_IP_BLOCK_TYPE_IH] = "ih",
+	[AMD_IP_BLOCK_TYPE_SMC] = "smu",
+	[AMD_IP_BLOCK_TYPE_PSP] = "psp",
+	[AMD_IP_BLOCK_TYPE_DCE] = "dce",
+	[AMD_IP_BLOCK_TYPE_GFX] = "gfx",
+	[AMD_IP_BLOCK_TYPE_SDMA] = "sdma",
+	[AMD_IP_BLOCK_TYPE_UVD] = "uvd",
+	[AMD_IP_BLOCK_TYPE_VCE] = "vce",
+	[AMD_IP_BLOCK_TYPE_ACP] = "acp",
+	[AMD_IP_BLOCK_TYPE_VCN] = "vcn",
+	[AMD_IP_BLOCK_TYPE_MES] = "mes",
+	[AMD_IP_BLOCK_TYPE_JPEG] = "jpeg",
+	[AMD_IP_BLOCK_TYPE_VPE] = "vpe",
+	[AMD_IP_BLOCK_TYPE_UMSCH_MM] = "umsch_mm",
+	[AMD_IP_BLOCK_TYPE_ISP] = "isp",
+	[AMD_IP_BLOCK_TYPE_RAS] = "ras",
+};
+
+static const char *ip_block_name(struct amdgpu_device *adev, enum amd_ip_block_type type)
+{
+	int idx = (int)type;
+
+	return idx < ARRAY_SIZE(ip_block_names) ? ip_block_names[idx] : "unknown";
+}
+
 /**
  * amdgpu_device_ip_block_add
  *
@@ -1889,8 +2535,15 @@ int amdgpu_device_ip_block_add(struct amdgpu_device *adev,
 		break;
 	}
 
-	DRM_INFO("add ip block number %d <%s>\n", adev->num_ip_blocks,
-		  ip_block_version->funcs->name);
+	dev_info(adev->dev, "detected ip block number %d <%s_v%d_%d_%d> (%s)\n",
+		 adev->num_ip_blocks,
+		 ip_block_name(adev, ip_block_version->type),
+		 ip_block_version->major,
+		 ip_block_version->minor,
+		 ip_block_version->rev,
+		 ip_block_version->funcs->name);
+
+	adev->ip_blocks[adev->num_ip_blocks].adev = adev;
 
 	adev->ip_blocks[adev->num_ip_blocks++].version = ip_block_version;
 
@@ -1906,7 +2559,7 @@ int amdgpu_device_ip_block_add(struct amdgpu_device *adev,
  * the module parameter virtual_display.  This feature provides a virtual
  * display hardware on headless boards or in virtualized environments.
  * This function parses and validates the configuration string specified by
- * the user and configues the virtual display configuration (number of
+ * the user and configures the virtual display configuration (number of
  * virtual connectors, crtcs, etc.) specified.
  */
 static void amdgpu_device_enable_virtual_display(struct amdgpu_device *adev)
@@ -1945,9 +2598,11 @@ static void amdgpu_device_enable_virtual_display(struct amdgpu_device *adev)
 			}
 		}
 
-		DRM_INFO("virtual display string:%s, %s:virtual_display:%d, num_crtc:%d\n",
-			 amdgpu_virtual_display, pci_address_name,
-			 adev->enable_virtual_display, adev->mode_info.num_crtc);
+		dev_info(
+			adev->dev,
+			"virtual display string:%s, %s:virtual_display:%d, num_crtc:%d\n",
+			amdgpu_virtual_display, pci_address_name,
+			adev->enable_virtual_display, adev->mode_info.num_crtc);
 
 		kfree(pciaddstr);
 	}
@@ -1958,8 +2613,9 @@ void amdgpu_device_set_sriov_virtual_display(struct amdgpu_device *adev)
 	if (amdgpu_sriov_vf(adev) && !adev->enable_virtual_display) {
 		adev->mode_info.num_crtc = 1;
 		adev->enable_virtual_display = true;
-		DRM_INFO("virtual_display:%d, num_crtc:%d\n",
-			 adev->enable_virtual_display, adev->mode_info.num_crtc);
+		dev_info(adev->dev, "virtual_display:%d, num_crtc:%d\n",
+			 adev->enable_virtual_display,
+			 adev->mode_info.num_crtc);
 	}
 }
 
@@ -1969,29 +2625,18 @@ void amdgpu_device_set_sriov_virtual_display(struct amdgpu_device *adev)
  * @adev: amdgpu_device pointer
  *
  * Parses the asic configuration parameters specified in the gpu info
- * firmware and makes them availale to the driver for use in configuring
+ * firmware and makes them available to the driver for use in configuring
  * the asic.
  * Returns 0 on success, -EINVAL on failure.
  */
 static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[40];
 	int err;
 	const struct gpu_info_firmware_header_v1_0 *hdr;
 
 	adev->firmware.gpu_info_fw = NULL;
 
-	if (adev->mman.discovery_bin) {
-		/*
-		 * FIXME: The bounding box is still needed by Navi12, so
-		 * temporarily read it from gpu_info firmware. Should be dropped
-		 * when DAL no longer needs it.
-		 */
-		if (adev->asic_type != CHIP_NAVI12)
-			return 0;
-	}
-
 	switch (adev->asic_type) {
 	default:
 		return 0;
@@ -2013,16 +2658,24 @@ static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev)
 		chip_name = "arcturus";
 		break;
 	case CHIP_NAVI12:
+		if (adev->discovery.bin)
+			return 0;
 		chip_name = "navi12";
 		break;
+	case CHIP_CYAN_SKILLFISH:
+		if (adev->discovery.bin)
+			return 0;
+		chip_name = "cyan_skillfish";
+		break;
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_gpu_info.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->firmware.gpu_info_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->firmware.gpu_info_fw,
+				   AMDGPU_UCODE_OPTIONAL,
+				   "amdgpu/%s_gpu_info.bin", chip_name);
 	if (err) {
 		dev_err(adev->dev,
-			"Failed to get gpu_info firmware \"%s\"\n",
-			fw_name);
+			"Failed to get gpu_info firmware \"%s_gpu_info.bin\"\n",
+			chip_name);
 		goto out;
 	}
 
@@ -2037,7 +2690,7 @@ static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev)
 								le32_to_cpu(hdr->header.ucode_array_offset_bytes));
 
 		/*
-		 * Should be droped when DAL no longer needs it.
+		 * Should be dropped when DAL no longer needs it.
 		 */
 		if (adev->asic_type == CHIP_NAVI12)
 			goto parse_soc_bounding_box;
@@ -2093,6 +2746,24 @@ out:
 	return err;
 }
 
+static void amdgpu_uid_init(struct amdgpu_device *adev)
+{
+	/* Initialize the UID for the device */
+	adev->uid_info = kzalloc(sizeof(struct amdgpu_uid), GFP_KERNEL);
+	if (!adev->uid_info) {
+		dev_warn(adev->dev, "Failed to allocate memory for UID\n");
+		return;
+	}
+	adev->uid_info->adev = adev;
+}
+
+static void amdgpu_uid_fini(struct amdgpu_device *adev)
+{
+	/* Free the UID memory */
+	kfree(adev->uid_info);
+	adev->uid_info = NULL;
+}
+
 /**
  * amdgpu_device_ip_early_init - run early init for hardware IPs
  *
@@ -2105,10 +2776,11 @@ out:
  */
 static int amdgpu_device_ip_early_init(struct amdgpu_device *adev)
 {
-	struct drm_device *dev = adev_to_drm(adev);
+	struct amdgpu_ip_block *ip_block;
 	struct pci_dev *parent;
+	bool total, skip_bios;
+	uint32_t bios_flags;
 	int i, r;
-	bool total;
 
 	amdgpu_device_enable_virtual_display(adev);
 
@@ -2116,6 +2788,10 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev)
 		r = amdgpu_virt_request_full_gpu(adev, true);
 		if (r)
 			return r;
+
+		r = amdgpu_virt_init_critical_region(adev);
+		if (r)
+			return r;
 	}
 
 	switch (adev->asic_type) {
@@ -2172,46 +2848,62 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev)
 		break;
 	}
 
+	/* Check for IP version 9.4.3 with A0 hardware */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3) &&
+	    !amdgpu_device_get_rev_id(adev)) {
+		dev_err(adev->dev, "Unsupported A0 hardware\n");
+		return -ENODEV;	/* device unsupported - no device error */
+	}
+
 	if (amdgpu_has_atpx() &&
 	    (amdgpu_is_atpx_hybrid() ||
 	     amdgpu_has_atpx_dgpu_power_cntl()) &&
 	    ((adev->flags & AMD_IS_APU) == 0) &&
-	    !pci_is_thunderbolt_attached(to_pci_dev(dev->dev)))
+	    !dev_is_removable(&adev->pdev->dev))
 		adev->flags |= AMD_IS_PX;
 
 	if (!(adev->flags & AMD_IS_APU)) {
-		parent = pci_upstream_bridge(adev->pdev);
+		parent = pcie_find_root_port(adev->pdev);
 		adev->has_pr3 = parent ? pci_pr3_present(parent) : false;
 	}
 
-
 	adev->pm.pp_feature = amdgpu_pp_feature_mask;
 	if (amdgpu_sriov_vf(adev) || sched_policy == KFD_SCHED_POLICY_NO_HWS)
 		adev->pm.pp_feature &= ~PP_GFXOFF_MASK;
 	if (amdgpu_sriov_vf(adev) && adev->asic_type == CHIP_SIENNA_CICHLID)
 		adev->pm.pp_feature &= ~PP_OVERDRIVE_MASK;
+	if (!amdgpu_device_pcie_dynamic_switching_supported(adev))
+		adev->pm.pp_feature &= ~PP_PCIE_DPM_MASK;
+
+	adev->virt.is_xgmi_node_migrate_enabled = false;
+	if (amdgpu_sriov_vf(adev)) {
+		adev->virt.is_xgmi_node_migrate_enabled =
+			amdgpu_ip_version((adev), GC_HWIP, 0) == IP_VERSION(9, 4, 4);
+	}
 
 	total = true;
 	for (i = 0; i < adev->num_ip_blocks; i++) {
+		ip_block = &adev->ip_blocks[i];
+
 		if ((amdgpu_ip_block_mask & (1 << i)) == 0) {
-			DRM_ERROR("disabled ip block: %d <%s>\n",
-				  i, adev->ip_blocks[i].version->funcs->name);
+			dev_warn(adev->dev, "disabled ip block: %d <%s>\n", i,
+				 adev->ip_blocks[i].version->funcs->name);
 			adev->ip_blocks[i].status.valid = false;
-		} else {
-			if (adev->ip_blocks[i].version->funcs->early_init) {
-				r = adev->ip_blocks[i].version->funcs->early_init((void *)adev);
-				if (r == -ENOENT) {
-					adev->ip_blocks[i].status.valid = false;
-				} else if (r) {
-					DRM_ERROR("early_init of IP block <%s> failed %d\n",
-						  adev->ip_blocks[i].version->funcs->name, r);
-					total = false;
-				} else {
-					adev->ip_blocks[i].status.valid = true;
-				}
+		} else if (ip_block->version->funcs->early_init) {
+			r = ip_block->version->funcs->early_init(ip_block);
+			if (r == -ENOENT) {
+				adev->ip_blocks[i].status.valid = false;
+			} else if (r) {
+				dev_err(adev->dev,
+					"early_init of IP block <%s> failed %d\n",
+					adev->ip_blocks[i].version->funcs->name,
+					r);
+				total = false;
 			} else {
 				adev->ip_blocks[i].status.valid = true;
 			}
+		} else {
+			adev->ip_blocks[i].status.valid = true;
 		}
 		/* get the vbios after the asic_funcs are set up */
 		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON) {
@@ -2219,15 +2911,32 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev)
 			if (r)
 				return r;
 
+			bios_flags = amdgpu_device_get_vbios_flags(adev);
+			skip_bios = !!(bios_flags & AMDGPU_VBIOS_SKIP);
 			/* Read BIOS */
-			if (!amdgpu_get_bios(adev))
-				return -EINVAL;
-
-			r = amdgpu_atombios_init(adev);
-			if (r) {
-				dev_err(adev->dev, "amdgpu_atombios_init failed\n");
-				amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_INIT_FAIL, 0, 0);
-				return r;
+			if (!skip_bios) {
+				bool optional =
+					!!(bios_flags & AMDGPU_VBIOS_OPTIONAL);
+				if (!amdgpu_get_bios(adev) && !optional)
+					return -EINVAL;
+
+				if (optional && !adev->bios)
+					dev_info(
+						adev->dev,
+						"VBIOS image optional, proceeding without VBIOS image");
+
+				if (adev->bios) {
+					r = amdgpu_atombios_init(adev);
+					if (r) {
+						dev_err(adev->dev,
+							"amdgpu_atombios_init failed\n");
+						amdgpu_vf_error_put(
+							adev,
+							AMDGIM_ERROR_VF_ATOMBIOS_INIT_FAIL,
+							0, 0);
+						return r;
+					}
+				}
 			}
 
 			/*get pf2vf msg info at it's earliest time*/
@@ -2239,7 +2948,15 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev)
 	if (!total)
 		return -ENODEV;
 
-	amdgpu_amdkfd_device_probe(adev);
+	if (adev->gmc.xgmi.supported)
+		amdgpu_xgmi_early_init(adev);
+
+	if (amdgpu_is_multi_aid(adev))
+		amdgpu_uid_init(adev);
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GFX);
+	if (ip_block->status.valid != false)
+		amdgpu_amdkfd_device_probe(adev);
+
 	adev->cg_flags &= amdgpu_cg_mask;
 	adev->pg_flags &= amdgpu_pg_mask;
 
@@ -2255,13 +2972,18 @@ static int amdgpu_device_ip_hw_init_phase1(struct amdgpu_device *adev)
 			continue;
 		if (adev->ip_blocks[i].status.hw)
 			continue;
+		if (!amdgpu_ip_member_of_hwini(
+			    adev, adev->ip_blocks[i].version->type))
+			continue;
 		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON ||
 		    (amdgpu_sriov_vf(adev) && (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP)) ||
 		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH) {
-			r = adev->ip_blocks[i].version->funcs->hw_init(adev);
+			r = adev->ip_blocks[i].version->funcs->hw_init(&adev->ip_blocks[i]);
 			if (r) {
-				DRM_ERROR("hw_init of IP block <%s> failed %d\n",
-					  adev->ip_blocks[i].version->funcs->name, r);
+				dev_err(adev->dev,
+					"hw_init of IP block <%s> failed %d\n",
+					adev->ip_blocks[i].version->funcs->name,
+					r);
 				return r;
 			}
 			adev->ip_blocks[i].status.hw = true;
@@ -2280,10 +3002,14 @@ static int amdgpu_device_ip_hw_init_phase2(struct amdgpu_device *adev)
 			continue;
 		if (adev->ip_blocks[i].status.hw)
 			continue;
-		r = adev->ip_blocks[i].version->funcs->hw_init(adev);
+		if (!amdgpu_ip_member_of_hwini(
+			    adev, adev->ip_blocks[i].version->type))
+			continue;
+		r = adev->ip_blocks[i].version->funcs->hw_init(&adev->ip_blocks[i]);
 		if (r) {
-			DRM_ERROR("hw_init of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
+			dev_err(adev->dev,
+				"hw_init of IP block <%s> failed %d\n",
+				adev->ip_blocks[i].version->funcs->name, r);
 			return r;
 		}
 		adev->ip_blocks[i].status.hw = true;
@@ -2303,6 +3029,10 @@ static int amdgpu_device_fw_loading(struct amdgpu_device *adev)
 			if (adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_PSP)
 				continue;
 
+			if (!amdgpu_ip_member_of_hwini(adev,
+						       AMD_IP_BLOCK_TYPE_PSP))
+				break;
+
 			if (!adev->ip_blocks[i].status.sw)
 				continue;
 
@@ -2311,22 +3041,21 @@ static int amdgpu_device_fw_loading(struct amdgpu_device *adev)
 				break;
 
 			if (amdgpu_in_reset(adev) || adev->in_suspend) {
-				r = adev->ip_blocks[i].version->funcs->resume(adev);
-				if (r) {
-					DRM_ERROR("resume of IP block <%s> failed %d\n",
-							  adev->ip_blocks[i].version->funcs->name, r);
+				r = amdgpu_ip_block_resume(&adev->ip_blocks[i]);
+				if (r)
 					return r;
-				}
 			} else {
-				r = adev->ip_blocks[i].version->funcs->hw_init(adev);
+				r = adev->ip_blocks[i].version->funcs->hw_init(&adev->ip_blocks[i]);
 				if (r) {
-					DRM_ERROR("hw_init of IP block <%s> failed %d\n",
-							  adev->ip_blocks[i].version->funcs->name, r);
+					dev_err(adev->dev,
+						"hw_init of IP block <%s> failed %d\n",
+						adev->ip_blocks[i]
+							.version->funcs->name,
+						r);
 					return r;
 				}
+				adev->ip_blocks[i].status.hw = true;
 			}
-
-			adev->ip_blocks[i].status.hw = true;
 			break;
 		}
 	}
@@ -2339,6 +3068,12 @@ static int amdgpu_device_fw_loading(struct amdgpu_device *adev)
 
 static int amdgpu_device_init_schedulers(struct amdgpu_device *adev)
 {
+	struct drm_sched_init_args args = {
+		.ops = &amdgpu_sched_ops,
+		.num_rqs = DRM_SCHED_PRIORITY_COUNT,
+		.timeout_wq = adev->reset_domain->wq,
+		.dev = adev->dev,
+	};
 	long timeout;
 	int r, i;
 
@@ -2364,18 +3099,37 @@ static int amdgpu_device_init_schedulers(struct amdgpu_device *adev)
 			break;
 		}
 
-		r = drm_sched_init(&ring->sched, &amdgpu_sched_ops,
-				   ring->num_hw_submission, 0,
-				   timeout, adev->reset_domain->wq,
-				   ring->sched_score, ring->name,
-				   adev->dev);
+		args.timeout = timeout;
+		args.credit_limit = ring->num_hw_submission;
+		args.score = ring->sched_score;
+		args.name = ring->name;
+
+		r = drm_sched_init(&ring->sched, &args);
+		if (r) {
+			dev_err(adev->dev,
+				"Failed to create scheduler on ring %s.\n",
+				ring->name);
+			return r;
+		}
+		r = amdgpu_uvd_entity_init(adev, ring);
+		if (r) {
+			dev_err(adev->dev,
+				"Failed to create UVD scheduling entity on ring %s.\n",
+				ring->name);
+			return r;
+		}
+		r = amdgpu_vce_entity_init(adev, ring);
 		if (r) {
-			DRM_ERROR("Failed to create scheduler on ring %s.\n",
-				  ring->name);
+			dev_err(adev->dev,
+				"Failed to create VCE scheduling entity on ring %s.\n",
+				ring->name);
 			return r;
 		}
 	}
 
+	if (adev->xcp_mgr)
+		amdgpu_xcp_update_partition_sched_list(adev);
+
 	return 0;
 }
 
@@ -2393,6 +3147,7 @@ static int amdgpu_device_init_schedulers(struct amdgpu_device *adev)
  */
 static int amdgpu_device_ip_init(struct amdgpu_device *adev)
 {
+	bool init_badpage;
 	int i, r;
 
 	r = amdgpu_ras_init(adev);
@@ -2402,19 +3157,28 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)
 	for (i = 0; i < adev->num_ip_blocks; i++) {
 		if (!adev->ip_blocks[i].status.valid)
 			continue;
-		r = adev->ip_blocks[i].version->funcs->sw_init((void *)adev);
-		if (r) {
-			DRM_ERROR("sw_init of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
-			goto init_failed;
+		if (adev->ip_blocks[i].version->funcs->sw_init) {
+			r = adev->ip_blocks[i].version->funcs->sw_init(&adev->ip_blocks[i]);
+			if (r) {
+				dev_err(adev->dev,
+					"sw_init of IP block <%s> failed %d\n",
+					adev->ip_blocks[i].version->funcs->name,
+					r);
+				goto init_failed;
+			}
 		}
 		adev->ip_blocks[i].status.sw = true;
 
+		if (!amdgpu_ip_member_of_hwini(
+			    adev, adev->ip_blocks[i].version->type))
+			continue;
+
 		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON) {
 			/* need to do common hw init early so everything is set up for gmc */
-			r = adev->ip_blocks[i].version->funcs->hw_init((void *)adev);
+			r = adev->ip_blocks[i].version->funcs->hw_init(&adev->ip_blocks[i]);
 			if (r) {
-				DRM_ERROR("hw_init %d failed %d\n", i, r);
+				dev_err(adev->dev, "hw_init %d failed %d\n", i,
+					r);
 				goto init_failed;
 			}
 			adev->ip_blocks[i].status.hw = true;
@@ -2426,32 +3190,44 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)
 
 			r = amdgpu_device_mem_scratch_init(adev);
 			if (r) {
-				DRM_ERROR("amdgpu_mem_scratch_init failed %d\n", r);
+				dev_err(adev->dev,
+					"amdgpu_mem_scratch_init failed %d\n",
+					r);
 				goto init_failed;
 			}
-			r = adev->ip_blocks[i].version->funcs->hw_init((void *)adev);
+			r = adev->ip_blocks[i].version->funcs->hw_init(&adev->ip_blocks[i]);
 			if (r) {
-				DRM_ERROR("hw_init %d failed %d\n", i, r);
+				dev_err(adev->dev, "hw_init %d failed %d\n", i,
+					r);
 				goto init_failed;
 			}
 			r = amdgpu_device_wb_init(adev);
 			if (r) {
-				DRM_ERROR("amdgpu_device_wb_init failed %d\n", r);
+				dev_err(adev->dev,
+					"amdgpu_device_wb_init failed %d\n", r);
 				goto init_failed;
 			}
 			adev->ip_blocks[i].status.hw = true;
 
 			/* right after GMC hw init, we create CSA */
-			if (amdgpu_mcbp) {
+			if (adev->gfx.mcbp) {
 				r = amdgpu_allocate_static_csa(adev, &adev->virt.csa_obj,
 							       AMDGPU_GEM_DOMAIN_VRAM |
 							       AMDGPU_GEM_DOMAIN_GTT,
 							       AMDGPU_CSA_SIZE);
 				if (r) {
-					DRM_ERROR("allocate CSA failed %d\n", r);
+					dev_err(adev->dev,
+						"allocate CSA failed %d\n", r);
 					goto init_failed;
 				}
 			}
+
+			r = amdgpu_seq64_init(adev);
+			if (r) {
+				dev_err(adev->dev, "allocate seq64 failed %d\n",
+					r);
+				goto init_failed;
+			}
 		}
 	}
 
@@ -2496,7 +3272,8 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)
 	 * Note: theoretically, this should be called before all vram allocations
 	 * to protect retired page from abusing
 	 */
-	r = amdgpu_ras_recovery_init(adev);
+	init_badpage = (adev->init_lvl->level != AMDGPU_INIT_LEVEL_MINIMAL_XGMI);
+	r = amdgpu_ras_recovery_init(adev, init_badpage);
 	if (r)
 		goto init_failed;
 
@@ -2532,15 +3309,21 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)
 	if (r)
 		goto init_failed;
 
+	if (adev->mman.buffer_funcs_ring->sched.ready)
+		amdgpu_ttm_set_buffer_funcs_status(adev, true);
+
 	/* Don't init kfd if whole hive need to be reset during init */
-	if (!adev->gmc.xgmi.pending_reset)
+	if (adev->init_lvl->level != AMDGPU_INIT_LEVEL_MINIMAL_XGMI) {
+		kgd2kfd_init_zone_device(adev);
 		amdgpu_amdkfd_device_init(adev);
+	}
 
 	amdgpu_fru_get_product_info(adev);
 
+	if (!amdgpu_sriov_vf(adev) || amdgpu_sriov_ras_cper_en(adev))
+		r = amdgpu_cper_init(adev);
+
 init_failed:
-	if (amdgpu_sriov_vf(adev))
-		amdgpu_virt_release_full_gpu(adev, true);
 
 	return r;
 }
@@ -2552,7 +3335,7 @@ init_failed:
  *
  * Writes a reset magic value to the gart pointer in VRAM.  The driver calls
  * this function before a GPU reset.  If the value is retained after a
- * GPU reset, VRAM has not been lost.  Some GPU resets may destry VRAM contents.
+ * GPU reset, VRAM has not been lost. Some GPU resets may destroy VRAM contents.
  */
 static void amdgpu_device_fill_reset_magic(struct amdgpu_device *adev)
 {
@@ -2583,6 +3366,8 @@ static bool amdgpu_device_check_vram_lost(struct amdgpu_device *adev)
 	 * always assumed to be lost.
 	 */
 	switch (amdgpu_asic_reset_method(adev)) {
+	case AMD_RESET_METHOD_LEGACY:
+	case AMD_RESET_METHOD_LINK:
 	case AMD_RESET_METHOD_BACO:
 	case AMD_RESET_METHOD_MODE1:
 		return true;
@@ -2628,11 +3413,13 @@ int amdgpu_device_set_cg_state(struct amdgpu_device *adev,
 		    adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_JPEG &&
 		    adev->ip_blocks[i].version->funcs->set_clockgating_state) {
 			/* enable clockgating to save power */
-			r = adev->ip_blocks[i].version->funcs->set_clockgating_state((void *)adev,
+			r = adev->ip_blocks[i].version->funcs->set_clockgating_state(&adev->ip_blocks[i],
 										     state);
 			if (r) {
-				DRM_ERROR("set_clockgating_state(gate) of IP block <%s> failed %d\n",
-					  adev->ip_blocks[i].version->funcs->name, r);
+				dev_err(adev->dev,
+					"set_clockgating_state(gate) of IP block <%s> failed %d\n",
+					adev->ip_blocks[i].version->funcs->name,
+					r);
 				return r;
 			}
 		}
@@ -2658,18 +3445,21 @@ int amdgpu_device_set_pg_state(struct amdgpu_device *adev,
 		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GFX ||
 		     adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SDMA))
 			continue;
-		/* skip CG for VCE/UVD, it's handled specially */
+		/* skip CG for VCE/UVD/VPE, it's handled specially */
 		if (adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_UVD &&
 		    adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_VCE &&
 		    adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_VCN &&
+		    adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_VPE &&
 		    adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_JPEG &&
 		    adev->ip_blocks[i].version->funcs->set_powergating_state) {
 			/* enable powergating to save power */
-			r = adev->ip_blocks[i].version->funcs->set_powergating_state((void *)adev,
+			r = adev->ip_blocks[i].version->funcs->set_powergating_state(&adev->ip_blocks[i],
 											state);
 			if (r) {
-				DRM_ERROR("set_powergating_state(gate) of IP block <%s> failed %d\n",
-					  adev->ip_blocks[i].version->funcs->name, r);
+				dev_err(adev->dev,
+					"set_powergating_state(gate) of IP block <%s> failed %d\n",
+					adev->ip_blocks[i].version->funcs->name,
+					r);
 				return r;
 			}
 		}
@@ -2696,7 +3486,7 @@ static int amdgpu_device_enable_mgpu_fan_boost(void)
 	for (i = 0; i < mgpu_info.num_dgpu; i++) {
 		gpu_ins = &(mgpu_info.gpu_ins[i]);
 		adev = gpu_ins->adev;
-		if (!(adev->flags & AMD_IS_APU) &&
+		if (!(adev->flags & AMD_IS_APU || amdgpu_sriov_multi_vf_mode(adev)) &&
 		    !gpu_ins->mgpu_fan_enabled) {
 			ret = amdgpu_dpm_enable_mgpu_fan_boost(adev);
 			if (ret)
@@ -2733,10 +3523,12 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev)
 		if (!adev->ip_blocks[i].status.hw)
 			continue;
 		if (adev->ip_blocks[i].version->funcs->late_init) {
-			r = adev->ip_blocks[i].version->funcs->late_init((void *)adev);
+			r = adev->ip_blocks[i].version->funcs->late_init(&adev->ip_blocks[i]);
 			if (r) {
-				DRM_ERROR("late_init of IP block <%s> failed %d\n",
-					  adev->ip_blocks[i].version->funcs->name, r);
+				dev_err(adev->dev,
+					"late_init of IP block <%s> failed %d\n",
+					adev->ip_blocks[i].version->funcs->name,
+					r);
 				return r;
 			}
 		}
@@ -2745,11 +3537,12 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev)
 
 	r = amdgpu_ras_late_init(adev);
 	if (r) {
-		DRM_ERROR("amdgpu_ras_late_init failed %d", r);
+		dev_err(adev->dev, "amdgpu_ras_late_init failed %d", r);
 		return r;
 	}
 
-	amdgpu_ras_set_error_query_ready(adev, true);
+	if (!amdgpu_reset_in_recovery(adev))
+		amdgpu_ras_set_error_query_ready(adev, true);
 
 	amdgpu_device_set_cg_state(adev, AMD_CG_STATE_GATE);
 	amdgpu_device_set_pg_state(adev, AMD_PG_STATE_GATE);
@@ -2758,11 +3551,12 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev)
 
 	r = amdgpu_device_enable_mgpu_fan_boost();
 	if (r)
-		DRM_ERROR("enable mgpu fan boost failed (%d).\n", r);
+		dev_err(adev->dev, "enable mgpu fan boost failed (%d).\n", r);
 
 	/* For passthrough configuration on arcturus and aldebaran, enable special handling SBR */
-	if (amdgpu_passthrough(adev) && ((adev->asic_type == CHIP_ARCTURUS && adev->gmc.xgmi.num_physical_nodes > 1)||
-			       adev->asic_type == CHIP_ALDEBARAN ))
+	if (amdgpu_passthrough(adev) &&
+	    ((adev->asic_type == CHIP_ARCTURUS && adev->gmc.xgmi.num_physical_nodes > 1) ||
+	     adev->asic_type == CHIP_ALDEBARAN))
 		amdgpu_dpm_handle_passthrough_sbr(adev, true);
 
 	if (adev->gmc.xgmi.num_physical_nodes > 1) {
@@ -2790,7 +3584,9 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev)
 				r = amdgpu_xgmi_set_pstate(gpu_instance->adev,
 						AMDGPU_XGMI_PSTATE_MIN);
 				if (r) {
-					DRM_ERROR("pstate setting failed (%d).\n", r);
+					dev_err(adev->dev,
+						"pstate setting failed (%d).\n",
+						r);
 					break;
 				}
 			}
@@ -2802,6 +3598,27 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev)
 	return 0;
 }
 
+static void amdgpu_ip_block_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	if (!ip_block->version->funcs->hw_fini) {
+		dev_err(adev->dev, "hw_fini of IP block <%s> not defined\n",
+			ip_block->version->funcs->name);
+	} else {
+		r = ip_block->version->funcs->hw_fini(ip_block);
+		/* XXX handle errors */
+		if (r) {
+			dev_dbg(adev->dev,
+				"hw_fini of IP block <%s> failed %d\n",
+				ip_block->version->funcs->name, r);
+		}
+	}
+
+	ip_block->status.hw = false;
+}
+
 /**
  * amdgpu_device_smu_fini_early - smu hw_fini wrapper
  *
@@ -2811,22 +3628,16 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev)
  */
 static void amdgpu_device_smu_fini_early(struct amdgpu_device *adev)
 {
-	int i, r;
+	int i;
 
-	if (adev->ip_versions[GC_HWIP][0] > IP_VERSION(9, 0, 0))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) > IP_VERSION(9, 0, 0))
 		return;
 
 	for (i = 0; i < adev->num_ip_blocks; i++) {
 		if (!adev->ip_blocks[i].status.hw)
 			continue;
 		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC) {
-			r = adev->ip_blocks[i].version->funcs->hw_fini((void *)adev);
-			/* XXX handle errors */
-			if (r) {
-				DRM_DEBUG("hw_fini of IP block <%s> failed %d\n",
-					  adev->ip_blocks[i].version->funcs->name, r);
-			}
-			adev->ip_blocks[i].status.hw = false;
+			amdgpu_ip_block_hw_fini(&adev->ip_blocks[i]);
 			break;
 		}
 	}
@@ -2840,38 +3651,48 @@ static int amdgpu_device_ip_fini_early(struct amdgpu_device *adev)
 		if (!adev->ip_blocks[i].version->funcs->early_fini)
 			continue;
 
-		r = adev->ip_blocks[i].version->funcs->early_fini((void *)adev);
+		r = adev->ip_blocks[i].version->funcs->early_fini(&adev->ip_blocks[i]);
 		if (r) {
-			DRM_DEBUG("early_fini of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
+			dev_dbg(adev->dev,
+				"early_fini of IP block <%s> failed %d\n",
+				adev->ip_blocks[i].version->funcs->name, r);
 		}
 	}
 
 	amdgpu_device_set_pg_state(adev, AMD_PG_STATE_UNGATE);
 	amdgpu_device_set_cg_state(adev, AMD_CG_STATE_UNGATE);
 
-	amdgpu_amdkfd_suspend(adev, false);
+	amdgpu_amdkfd_suspend(adev, true);
+	amdgpu_userq_suspend(adev);
 
-	/* Workaroud for ASICs need to disable SMC first */
+	/* Workaround for ASICs need to disable SMC first */
 	amdgpu_device_smu_fini_early(adev);
 
 	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
 		if (!adev->ip_blocks[i].status.hw)
 			continue;
 
-		r = adev->ip_blocks[i].version->funcs->hw_fini((void *)adev);
-		/* XXX handle errors */
-		if (r) {
-			DRM_DEBUG("hw_fini of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
-		}
-
-		adev->ip_blocks[i].status.hw = false;
+		amdgpu_ip_block_hw_fini(&adev->ip_blocks[i]);
 	}
 
 	if (amdgpu_sriov_vf(adev)) {
 		if (amdgpu_virt_release_full_gpu(adev, false))
-			DRM_ERROR("failed to release exclusive mode on fini\n");
+			dev_err(adev->dev,
+				"failed to release exclusive mode on fini\n");
+	}
+
+	/*
+	 * Driver reload on the APU can fail due to firmware validation because
+	 * the PSP is always running, as it is shared across the whole SoC.
+	 * This same issue does not occur on dGPU because it has a mechanism
+	 * that checks whether the PSP is running. A solution for those issues
+	 * in the APU is to trigger a GPU reset, but this should be done during
+	 * the unload phase to avoid adding boot latency and screen flicker.
+	 */
+	if ((adev->flags & AMD_IS_APU) && !adev->gmc.is_app_apu) {
+		r = amdgpu_asic_reset(adev);
+		if (r)
+			dev_err(adev->dev, "asic reset on %s failed\n", __func__);
 	}
 
 	return 0;
@@ -2892,6 +3713,8 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev)
 {
 	int i, r;
 
+	amdgpu_cper_fini(adev);
+
 	if (amdgpu_sriov_vf(adev) && adev->virt.ras_init_done)
 		amdgpu_virt_release_ras_err_handler_data(adev);
 
@@ -2910,13 +3733,18 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev)
 			amdgpu_device_wb_fini(adev);
 			amdgpu_device_mem_scratch_fini(adev);
 			amdgpu_ib_pool_fini(adev);
+			amdgpu_seq64_fini(adev);
+			amdgpu_doorbell_fini(adev);
 		}
-
-		r = adev->ip_blocks[i].version->funcs->sw_fini((void *)adev);
-		/* XXX handle errors */
-		if (r) {
-			DRM_DEBUG("sw_fini of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
+		if (adev->ip_blocks[i].version->funcs->sw_fini) {
+			r = adev->ip_blocks[i].version->funcs->sw_fini(&adev->ip_blocks[i]);
+			/* XXX handle errors */
+			if (r) {
+				dev_dbg(adev->dev,
+					"sw_fini of IP block <%s> failed %d\n",
+					adev->ip_blocks[i].version->funcs->name,
+					r);
+			}
 		}
 		adev->ip_blocks[i].status.sw = false;
 		adev->ip_blocks[i].status.valid = false;
@@ -2926,11 +3754,12 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev)
 		if (!adev->ip_blocks[i].status.late_initialized)
 			continue;
 		if (adev->ip_blocks[i].version->funcs->late_fini)
-			adev->ip_blocks[i].version->funcs->late_fini((void *)adev);
+			adev->ip_blocks[i].version->funcs->late_fini(&adev->ip_blocks[i]);
 		adev->ip_blocks[i].status.late_initialized = false;
 	}
 
 	amdgpu_ras_fini(adev);
+	amdgpu_uid_fini(adev);
 
 	return 0;
 }
@@ -2948,7 +3777,7 @@ static void amdgpu_device_delayed_init_work_handler(struct work_struct *work)
 
 	r = amdgpu_ib_ring_tests(adev);
 	if (r)
-		DRM_ERROR("ib ring test failed (%d).\n", r);
+		dev_err(adev->dev, "ib ring test failed (%d).\n", r);
 }
 
 static void amdgpu_device_delay_enable_gfx_off(struct work_struct *work)
@@ -2959,7 +3788,7 @@ static void amdgpu_device_delay_enable_gfx_off(struct work_struct *work)
 	WARN_ON_ONCE(adev->gfx.gfx_off_state);
 	WARN_ON_ONCE(adev->gfx.gfx_off_req_count);
 
-	if (!amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, true))
+	if (!amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, true, 0))
 		adev->gfx.gfx_off_state = true;
 }
 
@@ -2976,7 +3805,7 @@ static void amdgpu_device_delay_enable_gfx_off(struct work_struct *work)
  */
 static int amdgpu_device_ip_suspend_phase1(struct amdgpu_device *adev)
 {
-	int i, r;
+	int i, r, rec;
 
 	amdgpu_device_set_pg_state(adev, AMD_PG_STATE_UNGATE);
 	amdgpu_device_set_cg_state(adev, AMD_CG_STATE_UNGATE);
@@ -2997,19 +3826,25 @@ static int amdgpu_device_ip_suspend_phase1(struct amdgpu_device *adev)
 		if (adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_DCE)
 			continue;
 
-		/* XXX handle errors */
-		r = adev->ip_blocks[i].version->funcs->suspend(adev);
-		/* XXX handle errors */
-		if (r) {
-			DRM_ERROR("suspend of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
-			return r;
-		}
-
-		adev->ip_blocks[i].status.hw = false;
+		r = amdgpu_ip_block_suspend(&adev->ip_blocks[i]);
+		if (r)
+			goto unwind;
 	}
 
 	return 0;
+unwind:
+	rec = amdgpu_device_ip_resume_phase3(adev);
+	if (rec)
+		dev_err(adev->dev,
+			"amdgpu_device_ip_resume_phase3 failed during unwind: %d\n",
+			rec);
+
+	amdgpu_dpm_set_df_cstate(adev, DF_CSTATE_ALLOW);
+
+	amdgpu_device_set_pg_state(adev, AMD_PG_STATE_GATE);
+	amdgpu_device_set_cg_state(adev, AMD_CG_STATE_GATE);
+
+	return r;
 }
 
 /**
@@ -3025,7 +3860,7 @@ static int amdgpu_device_ip_suspend_phase1(struct amdgpu_device *adev)
  */
 static int amdgpu_device_ip_suspend_phase2(struct amdgpu_device *adev)
 {
-	int i, r;
+	int i, r, rec;
 
 	if (adev->in_s0ix)
 		amdgpu_dpm_gfx_state_change(adev, sGpuChangeState_D3Entry);
@@ -3044,15 +3879,17 @@ static int amdgpu_device_ip_suspend_phase2(struct amdgpu_device *adev)
 		}
 
 		/* skip unnecessary suspend if we do not initialize them yet */
-		if (adev->gmc.xgmi.pending_reset &&
-		    !(adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC ||
-		      adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC ||
-		      adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON ||
-		      adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH)) {
-			adev->ip_blocks[i].status.hw = false;
+		if (!amdgpu_ip_member_of_hwini(
+			    adev, adev->ip_blocks[i].version->type))
 			continue;
-		}
 
+		/* Since we skip suspend for S0i3, we need to cancel the delayed
+		 * idle work here as the suspend callback never gets called.
+		 */
+		if (adev->in_s0ix &&
+		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GFX &&
+		    amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 0, 0))
+			cancel_delayed_work_sync(&adev->gfx.idle_work);
 		/* skip suspend of gfx/mes and psp for S0ix
 		 * gfx is in gfxoff state, so on resume it will exit gfxoff just
 		 * like at runtime. PSP is also part of the always on hardware
@@ -3066,8 +3903,10 @@ static int amdgpu_device_ip_suspend_phase2(struct amdgpu_device *adev)
 
 		/* SDMA 5.x+ is part of GFX power domain so it's covered by GFXOFF */
 		if (adev->in_s0ix &&
-		    (adev->ip_versions[SDMA0_HWIP][0] >= IP_VERSION(5, 0, 0)) &&
-		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SDMA))
+		    (amdgpu_ip_version(adev, SDMA0_HWIP, 0) >=
+		     IP_VERSION(5, 0, 0)) &&
+		    (adev->ip_blocks[i].version->type ==
+		     AMD_IP_BLOCK_TYPE_SDMA))
 			continue;
 
 		/* Once swPSP provides the IMU, RLC FW binaries to TOS during cold-boot.
@@ -3082,28 +3921,52 @@ static int amdgpu_device_ip_suspend_phase2(struct amdgpu_device *adev)
 		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP)
 			continue;
 
-		/* XXX handle errors */
-		r = adev->ip_blocks[i].version->funcs->suspend(adev);
-		/* XXX handle errors */
-		if (r) {
-			DRM_ERROR("suspend of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
-		}
-		adev->ip_blocks[i].status.hw = false;
+		r = amdgpu_ip_block_suspend(&adev->ip_blocks[i]);
+		if (r)
+			goto unwind;
+
 		/* handle putting the SMC in the appropriate state */
-		if(!amdgpu_sriov_vf(adev)){
+		if (!amdgpu_sriov_vf(adev)) {
 			if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC) {
 				r = amdgpu_dpm_set_mp1_state(adev, adev->mp1_state);
 				if (r) {
-					DRM_ERROR("SMC failed to set mp1 state %d, %d\n",
-							adev->mp1_state, r);
-					return r;
+					dev_err(adev->dev,
+						"SMC failed to set mp1 state %d, %d\n",
+						adev->mp1_state, r);
+					goto unwind;
 				}
 			}
 		}
 	}
 
 	return 0;
+unwind:
+	/* suspend phase 2 = resume phase 1 + resume phase 2 */
+	rec = amdgpu_device_ip_resume_phase1(adev);
+	if (rec) {
+		dev_err(adev->dev,
+			"amdgpu_device_ip_resume_phase1 failed during unwind: %d\n",
+			rec);
+		return r;
+	}
+
+	rec = amdgpu_device_fw_loading(adev);
+	if (rec) {
+		dev_err(adev->dev,
+			"amdgpu_device_fw_loading failed during unwind: %d\n",
+			rec);
+		return r;
+	}
+
+	rec = amdgpu_device_ip_resume_phase2(adev);
+	if (rec) {
+		dev_err(adev->dev,
+			"amdgpu_device_ip_resume_phase2 failed during unwind: %d\n",
+			rec);
+		return r;
+	}
+
+	return r;
 }
 
 /**
@@ -3117,7 +3980,7 @@ static int amdgpu_device_ip_suspend_phase2(struct amdgpu_device *adev)
  * in each IP into a state suitable for suspend.
  * Returns 0 on success, negative error code on failure.
  */
-int amdgpu_device_ip_suspend(struct amdgpu_device *adev)
+static int amdgpu_device_ip_suspend(struct amdgpu_device *adev)
 {
 	int r;
 
@@ -3126,6 +3989,8 @@ int amdgpu_device_ip_suspend(struct amdgpu_device *adev)
 		amdgpu_virt_request_full_gpu(adev, false);
 	}
 
+	amdgpu_ttm_set_buffer_funcs_status(adev, false);
+
 	r = amdgpu_device_ip_suspend_phase1(adev);
 	if (r)
 		return r;
@@ -3161,10 +4026,12 @@ static int amdgpu_device_ip_reinit_early_sriov(struct amdgpu_device *adev)
 				!block->status.valid)
 				continue;
 
-			r = block->version->funcs->hw_init(adev);
-			DRM_INFO("RE-INIT-early: %s %s\n", block->version->funcs->name, r?"failed":"succeeded");
-			if (r)
+			r = block->version->funcs->hw_init(&adev->ip_blocks[i]);
+			if (r) {
+				dev_err(adev->dev, "RE-INIT-early: %s failed\n",
+					 block->version->funcs->name);
 				return r;
+			}
 			block->status.hw = true;
 		}
 	}
@@ -3174,7 +4041,8 @@ static int amdgpu_device_ip_reinit_early_sriov(struct amdgpu_device *adev)
 
 static int amdgpu_device_ip_reinit_late_sriov(struct amdgpu_device *adev)
 {
-	int i, r;
+	struct amdgpu_ip_block *block;
+	int i, r = 0;
 
 	static enum amd_ip_block_type ip_order[] = {
 		AMD_IP_BLOCK_TYPE_SMC,
@@ -3189,30 +4057,28 @@ static int amdgpu_device_ip_reinit_late_sriov(struct amdgpu_device *adev)
 	};
 
 	for (i = 0; i < ARRAY_SIZE(ip_order); i++) {
-		int j;
-		struct amdgpu_ip_block *block;
-
-		for (j = 0; j < adev->num_ip_blocks; j++) {
-			block = &adev->ip_blocks[j];
+		block = amdgpu_device_ip_get_ip_block(adev, ip_order[i]);
 
-			if (block->version->type != ip_order[i] ||
-				!block->status.valid ||
-				block->status.hw)
-				continue;
+		if (!block)
+			continue;
 
-			if (block->version->type == AMD_IP_BLOCK_TYPE_SMC)
-				r = block->version->funcs->resume(adev);
-			else
-				r = block->version->funcs->hw_init(adev);
+		if (block->status.valid && !block->status.hw) {
+			if (block->version->type == AMD_IP_BLOCK_TYPE_SMC) {
+				r = amdgpu_ip_block_resume(block);
+			} else {
+				r = block->version->funcs->hw_init(block);
+			}
 
-			DRM_INFO("RE-INIT-late: %s %s\n", block->version->funcs->name, r?"failed":"succeeded");
-			if (r)
-				return r;
+			if (r) {
+				dev_err(adev->dev, "RE-INIT-late: %s failed\n",
+					 block->version->funcs->name);
+				break;
+			}
 			block->status.hw = true;
 		}
 	}
 
-	return 0;
+	return r;
 }
 
 /**
@@ -3239,13 +4105,9 @@ static int amdgpu_device_ip_resume_phase1(struct amdgpu_device *adev)
 		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH ||
 		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP && amdgpu_sriov_vf(adev))) {
 
-			r = adev->ip_blocks[i].version->funcs->resume(adev);
-			if (r) {
-				DRM_ERROR("resume of IP block <%s> failed %d\n",
-					  adev->ip_blocks[i].version->funcs->name, r);
+			r = amdgpu_ip_block_resume(&adev->ip_blocks[i]);
+			if (r)
 				return r;
-			}
-			adev->ip_blocks[i].status.hw = true;
 		}
 	}
 
@@ -3257,7 +4119,7 @@ static int amdgpu_device_ip_resume_phase1(struct amdgpu_device *adev)
  *
  * @adev: amdgpu_device pointer
  *
- * First resume function for hardware IPs.  The list of all the hardware
+ * Second resume function for hardware IPs.  The list of all the hardware
  * IPs that make up the asic is walked and the resume callbacks are run for
  * all blocks except COMMON, GMC, and IH.  resume puts the hardware into a
  * functional state after a suspend and updates the software state as
@@ -3275,15 +4137,42 @@ static int amdgpu_device_ip_resume_phase2(struct amdgpu_device *adev)
 		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON ||
 		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC ||
 		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH ||
+		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE ||
 		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP)
 			continue;
-		r = adev->ip_blocks[i].version->funcs->resume(adev);
-		if (r) {
-			DRM_ERROR("resume of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
+		r = amdgpu_ip_block_resume(&adev->ip_blocks[i]);
+		if (r)
 			return r;
+	}
+
+	return 0;
+}
+
+/**
+ * amdgpu_device_ip_resume_phase3 - run resume for hardware IPs
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Third resume function for hardware IPs.  The list of all the hardware
+ * IPs that make up the asic is walked and the resume callbacks are run for
+ * all DCE.  resume puts the hardware into a functional state after a suspend
+ * and updates the software state as necessary.  This function is also used
+ * for restoring the GPU after a GPU reset.
+ *
+ * Returns 0 on success, negative error code on failure.
+ */
+static int amdgpu_device_ip_resume_phase3(struct amdgpu_device *adev)
+{
+	int i, r;
+
+	for (i = 0; i < adev->num_ip_blocks; i++) {
+		if (!adev->ip_blocks[i].status.valid || adev->ip_blocks[i].status.hw)
+			continue;
+		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE) {
+			r = amdgpu_ip_block_resume(&adev->ip_blocks[i]);
+			if (r)
+				return r;
 		}
-		adev->ip_blocks[i].status.hw = true;
 	}
 
 	return 0;
@@ -3296,7 +4185,7 @@ static int amdgpu_device_ip_resume_phase2(struct amdgpu_device *adev)
  *
  * Main resume function for hardware IPs.  The hardware IPs
  * are split into two resume functions because they are
- * are also used in in recovering from a GPU reset and some additional
+ * also used in recovering from a GPU reset and some additional
  * steps need to be take between them.  In this case (S3/S4) they are
  * run sequentially.
  * Returns 0 on success, negative error code on failure.
@@ -3305,12 +4194,6 @@ static int amdgpu_device_ip_resume(struct amdgpu_device *adev)
 {
 	int r;
 
-	if (!adev->in_s0ix) {
-		r = amdgpu_amdkfd_resume_iommu(adev);
-		if (r)
-			return r;
-	}
-
 	r = amdgpu_device_ip_resume_phase1(adev);
 	if (r)
 		return r;
@@ -3321,6 +4204,16 @@ static int amdgpu_device_ip_resume(struct amdgpu_device *adev)
 
 	r = amdgpu_device_ip_resume_phase2(adev);
 
+	if (adev->mman.buffer_funcs_ring->sched.ready)
+		amdgpu_ttm_set_buffer_funcs_status(adev, true);
+
+	if (r)
+		return r;
+
+	amdgpu_fence_driver_hw_init(adev);
+
+	r = amdgpu_device_ip_resume_phase3(adev);
+
 	return r;
 }
 
@@ -3350,12 +4243,14 @@ static void amdgpu_device_detect_sriov_bios(struct amdgpu_device *adev)
 /**
  * amdgpu_device_asic_has_dc_support - determine if DC supports the asic
  *
+ * @pdev : pci device context
  * @asic_type: AMD asic type
  *
  * Check if there is DC (new modesetting infrastructre) support for an asic.
  * returns true if DC has support, false if not.
  */
-bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type)
+bool amdgpu_device_asic_has_dc_support(struct pci_dev *pdev,
+				       enum amd_asic_type asic_type)
 {
 	switch (asic_type) {
 #ifdef CONFIG_DRM_AMDGPU_SI
@@ -3369,25 +4264,13 @@ bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type)
 	case CHIP_PITCAIRN:
 	case CHIP_VERDE:
 	case CHIP_OLAND:
-		/*
-		 * We have systems in the wild with these ASICs that require
-		 * LVDS and VGA support which is not supported with DC.
-		 *
-		 * Fallback to the non-DC driver here by default so as not to
-		 * cause regressions.
-		 */
-#if defined(CONFIG_DRM_AMD_DC_SI)
-		return amdgpu_dc > 0;
-#else
-		return false;
-#endif
-	case CHIP_BONAIRE:
+		return amdgpu_dc != 0 && IS_ENABLED(CONFIG_DRM_AMD_DC_SI);
 	case CHIP_KAVERI:
 	case CHIP_KABINI:
 	case CHIP_MULLINS:
 		/*
 		 * We have systems in the wild with these ASICs that require
-		 * VGA support which is not supported with DC.
+		 * TRAVIS and NUTMEG support which is not supported with DC.
 		 *
 		 * Fallback to the non-DC driver here by default so as not to
 		 * cause regressions.
@@ -3398,8 +4281,9 @@ bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type)
 #else
 	default:
 		if (amdgpu_dc > 0)
-			DRM_INFO_ONCE("Display Core has been requested via kernel parameter "
-					 "but isn't supported by ASIC, ignoring\n");
+			dev_info_once(
+				&pdev->dev,
+				"Display Core has been requested via kernel parameter but isn't supported by ASIC, ignoring\n");
 		return false;
 #endif
 	}
@@ -3418,7 +4302,7 @@ bool amdgpu_device_has_dc_support(struct amdgpu_device *adev)
 	    (adev->harvest_ip_mask & AMD_HARVEST_IP_DMU_MASK))
 		return false;
 
-	return amdgpu_device_asic_has_dc_support(adev->asic_type);
+	return amdgpu_device_asic_has_dc_support(adev->pdev, adev->asic_type);
 }
 
 static void amdgpu_device_xgmi_reset_func(struct work_struct *__work)
@@ -3440,20 +4324,18 @@ static void amdgpu_device_xgmi_reset_func(struct work_struct *__work)
 	if (amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) {
 
 		task_barrier_enter(&hive->tb);
-		adev->asic_reset_res = amdgpu_device_baco_enter(adev_to_drm(adev));
+		adev->asic_reset_res = amdgpu_device_baco_enter(adev);
 
 		if (adev->asic_reset_res)
 			goto fail;
 
 		task_barrier_exit(&hive->tb);
-		adev->asic_reset_res = amdgpu_device_baco_exit(adev_to_drm(adev));
+		adev->asic_reset_res = amdgpu_device_baco_exit(adev);
 
 		if (adev->asic_reset_res)
 			goto fail;
 
-		if (adev->mmhub.ras && adev->mmhub.ras->ras_block.hw_ops &&
-		    adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count)
-			adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count(adev);
+		amdgpu_ras_reset_error_count(adev, AMDGPU_RAS_BLOCK__MMHUB);
 	} else {
 
 		task_barrier_full(&hive->tb);
@@ -3462,7 +4344,8 @@ static void amdgpu_device_xgmi_reset_func(struct work_struct *__work)
 
 fail:
 	if (adev->asic_reset_res)
-		DRM_WARN("ASIC reset failed with error, %d for drm dev, %s",
+		dev_warn(adev->dev,
+			 "ASIC reset failed with error, %d for drm dev, %s",
 			 adev->asic_reset_res, adev_to_drm(adev)->unique);
 	amdgpu_put_xgmi_hive(hive);
 }
@@ -3475,66 +4358,53 @@ static int amdgpu_device_get_job_timeout_settings(struct amdgpu_device *adev)
 	long timeout;
 	int ret = 0;
 
-	/*
-	 * By default timeout for non compute jobs is 10000
-	 * and 60000 for compute jobs.
-	 * In SR-IOV or passthrough mode, timeout for compute
-	 * jobs are 60000 by default.
-	 */
-	adev->gfx_timeout = msecs_to_jiffies(10000);
-	adev->sdma_timeout = adev->video_timeout = adev->gfx_timeout;
-	if (amdgpu_sriov_vf(adev))
-		adev->compute_timeout = amdgpu_sriov_is_pp_one_vf(adev) ?
-					msecs_to_jiffies(60000) : msecs_to_jiffies(10000);
-	else
-		adev->compute_timeout =  msecs_to_jiffies(60000);
+	/* By default timeout for all queues is 2 sec */
+	adev->gfx_timeout = adev->compute_timeout = adev->sdma_timeout =
+		adev->video_timeout = msecs_to_jiffies(2000);
 
-	if (strnlen(input, AMDGPU_MAX_TIMEOUT_PARAM_LENGTH)) {
-		while ((timeout_setting = strsep(&input, ",")) &&
-				strnlen(timeout_setting, AMDGPU_MAX_TIMEOUT_PARAM_LENGTH)) {
-			ret = kstrtol(timeout_setting, 0, &timeout);
-			if (ret)
-				return ret;
+	if (!strnlen(input, AMDGPU_MAX_TIMEOUT_PARAM_LENGTH))
+		return 0;
 
-			if (timeout == 0) {
-				index++;
-				continue;
-			} else if (timeout < 0) {
-				timeout = MAX_SCHEDULE_TIMEOUT;
-				dev_warn(adev->dev, "lockup timeout disabled");
-				add_taint(TAINT_SOFTLOCKUP, LOCKDEP_STILL_OK);
-			} else {
-				timeout = msecs_to_jiffies(timeout);
-			}
+	while ((timeout_setting = strsep(&input, ",")) &&
+	       strnlen(timeout_setting, AMDGPU_MAX_TIMEOUT_PARAM_LENGTH)) {
+		ret = kstrtol(timeout_setting, 0, &timeout);
+		if (ret)
+			return ret;
 
-			switch (index++) {
-			case 0:
-				adev->gfx_timeout = timeout;
-				break;
-			case 1:
-				adev->compute_timeout = timeout;
-				break;
-			case 2:
-				adev->sdma_timeout = timeout;
-				break;
-			case 3:
-				adev->video_timeout = timeout;
-				break;
-			default:
-				break;
-			}
+		if (timeout == 0) {
+			index++;
+			continue;
+		} else if (timeout < 0) {
+			timeout = MAX_SCHEDULE_TIMEOUT;
+			dev_warn(adev->dev, "lockup timeout disabled");
+			add_taint(TAINT_SOFTLOCKUP, LOCKDEP_STILL_OK);
+		} else {
+			timeout = msecs_to_jiffies(timeout);
 		}
-		/*
-		 * There is only one value specified and
-		 * it should apply to all non-compute jobs.
-		 */
-		if (index == 1) {
-			adev->sdma_timeout = adev->video_timeout = adev->gfx_timeout;
-			if (amdgpu_sriov_vf(adev) || amdgpu_passthrough(adev))
-				adev->compute_timeout = adev->gfx_timeout;
+
+		switch (index++) {
+		case 0:
+			adev->gfx_timeout = timeout;
+			break;
+		case 1:
+			adev->compute_timeout = timeout;
+			break;
+		case 2:
+			adev->sdma_timeout = timeout;
+			break;
+		case 3:
+			adev->video_timeout = timeout;
+			break;
+		default:
+			break;
 		}
 	}
 
+	/* When only one value specified apply it to all queues. */
+	if (index == 1)
+		adev->gfx_timeout = adev->compute_timeout = adev->sdma_timeout =
+			adev->video_timeout = timeout;
+
 	return ret;
 }
 
@@ -3554,13 +4424,89 @@ static void amdgpu_device_check_iommu_direct_map(struct amdgpu_device *adev)
 		adev->ram_is_direct_mapped = true;
 }
 
-static const struct attribute *amdgpu_dev_attributes[] = {
-	&dev_attr_product_name.attr,
-	&dev_attr_product_number.attr,
-	&dev_attr_serial_number.attr,
-	&dev_attr_pcie_replay_count.attr,
-	NULL
-};
+#if defined(CONFIG_HSA_AMD_P2P)
+/**
+ * amdgpu_device_check_iommu_remap - Check if DMA remapping is enabled.
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * return if IOMMU remapping bar address
+ */
+static bool amdgpu_device_check_iommu_remap(struct amdgpu_device *adev)
+{
+	struct iommu_domain *domain;
+
+	domain = iommu_get_domain_for_dev(adev->dev);
+	if (domain && (domain->type == IOMMU_DOMAIN_DMA ||
+		domain->type ==	IOMMU_DOMAIN_DMA_FQ))
+		return true;
+
+	return false;
+}
+#endif
+
+static void amdgpu_device_set_mcbp(struct amdgpu_device *adev)
+{
+	if (amdgpu_mcbp == 1)
+		adev->gfx.mcbp = true;
+	else if (amdgpu_mcbp == 0)
+		adev->gfx.mcbp = false;
+
+	if (amdgpu_sriov_vf(adev))
+		adev->gfx.mcbp = true;
+
+	if (adev->gfx.mcbp)
+		dev_info(adev->dev, "MCBP is enabled\n");
+}
+
+static int amdgpu_device_sys_interface_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = amdgpu_atombios_sysfs_init(adev);
+	if (r)
+		drm_err(&adev->ddev,
+			"registering atombios sysfs failed (%d).\n", r);
+
+	r = amdgpu_pm_sysfs_init(adev);
+	if (r)
+		dev_err(adev->dev, "registering pm sysfs failed (%d).\n", r);
+
+	r = amdgpu_ucode_sysfs_init(adev);
+	if (r) {
+		adev->ucode_sysfs_en = false;
+		dev_err(adev->dev, "Creating firmware sysfs failed (%d).\n", r);
+	} else
+		adev->ucode_sysfs_en = true;
+
+	r = amdgpu_device_attr_sysfs_init(adev);
+	if (r)
+		dev_err(adev->dev, "Could not create amdgpu device attr\n");
+
+	r = devm_device_add_group(adev->dev, &amdgpu_board_attrs_group);
+	if (r)
+		dev_err(adev->dev,
+			"Could not create amdgpu board attributes\n");
+
+	amdgpu_fru_sysfs_init(adev);
+	amdgpu_reg_state_sysfs_init(adev);
+	amdgpu_xcp_sysfs_init(adev);
+
+	return r;
+}
+
+static void amdgpu_device_sys_interface_fini(struct amdgpu_device *adev)
+{
+	if (adev->pm.sysfs_initialized)
+		amdgpu_pm_sysfs_fini(adev);
+	if (adev->ucode_sysfs_en)
+		amdgpu_ucode_sysfs_fini(adev);
+	amdgpu_device_attr_sysfs_fini(adev);
+	amdgpu_fru_sysfs_fini(adev);
+
+	amdgpu_reg_state_sysfs_fini(adev);
+	amdgpu_xcp_sysfs_fini(adev);
+}
 
 /**
  * amdgpu_device_init - initialize the driver
@@ -3575,11 +4521,11 @@ static const struct attribute *amdgpu_dev_attributes[] = {
 int amdgpu_device_init(struct amdgpu_device *adev,
 		       uint32_t flags)
 {
-	struct drm_device *ddev = adev_to_drm(adev);
 	struct pci_dev *pdev = adev->pdev;
 	int r, i;
 	bool px = false;
 	u32 max_MBps;
+	int tmp;
 
 	adev->shutdown = false;
 	adev->flags = flags;
@@ -3609,10 +4555,14 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	adev->smc_wreg = &amdgpu_invalid_wreg;
 	adev->pcie_rreg = &amdgpu_invalid_rreg;
 	adev->pcie_wreg = &amdgpu_invalid_wreg;
+	adev->pcie_rreg_ext = &amdgpu_invalid_rreg_ext;
+	adev->pcie_wreg_ext = &amdgpu_invalid_wreg_ext;
 	adev->pciep_rreg = &amdgpu_invalid_rreg;
 	adev->pciep_wreg = &amdgpu_invalid_wreg;
 	adev->pcie_rreg64 = &amdgpu_invalid_rreg64;
 	adev->pcie_wreg64 = &amdgpu_invalid_wreg64;
+	adev->pcie_rreg64_ext = &amdgpu_invalid_rreg64_ext;
+	adev->pcie_wreg64_ext = &amdgpu_invalid_wreg64_ext;
 	adev->uvd_ctx_rreg = &amdgpu_invalid_rreg;
 	adev->uvd_ctx_wreg = &amdgpu_invalid_wreg;
 	adev->didt_rreg = &amdgpu_invalid_rreg;
@@ -3622,18 +4572,22 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	adev->audio_endpt_rreg = &amdgpu_block_invalid_rreg;
 	adev->audio_endpt_wreg = &amdgpu_block_invalid_wreg;
 
-	DRM_INFO("initializing kernel modesetting (%s 0x%04X:0x%04X 0x%04X:0x%04X 0x%02X).\n",
-		 amdgpu_asic_name[adev->asic_type], pdev->vendor, pdev->device,
-		 pdev->subsystem_vendor, pdev->subsystem_device, pdev->revision);
+	dev_info(
+		adev->dev,
+		"initializing kernel modesetting (%s 0x%04X:0x%04X 0x%04X:0x%04X 0x%02X).\n",
+		amdgpu_asic_name[adev->asic_type], pdev->vendor, pdev->device,
+		pdev->subsystem_vendor, pdev->subsystem_device, pdev->revision);
 
 	/* mutex initialization are all done here so we
-	 * can recall function without having locking issues */
+	 * can recall function without having locking issues
+	 */
 	mutex_init(&adev->firmware.mutex);
 	mutex_init(&adev->pm.mutex);
 	mutex_init(&adev->gfx.gpu_clock_mutex);
 	mutex_init(&adev->srbm_mutex);
 	mutex_init(&adev->gfx.pipe_reserve_mutex);
 	mutex_init(&adev->gfx.gfx_off_mutex);
+	mutex_init(&adev->gfx.partition_mutex);
 	mutex_init(&adev->grbm_idx_mutex);
 	mutex_init(&adev->mn_lock);
 	mutex_init(&adev->virt.vf_errors.lock);
@@ -3642,6 +4596,17 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	mutex_init(&adev->notifier_lock);
 	mutex_init(&adev->pm.stable_pstate_ctx_lock);
 	mutex_init(&adev->benchmark_mutex);
+	mutex_init(&adev->gfx.reset_sem_mutex);
+	/* Initialize the mutex for cleaner shader isolation between GFX and compute processes */
+	mutex_init(&adev->enforce_isolation_mutex);
+	for (i = 0; i < MAX_XCP; ++i) {
+		adev->isolation[i].spearhead = dma_fence_get_stub();
+		amdgpu_sync_create(&adev->isolation[i].active);
+		amdgpu_sync_create(&adev->isolation[i].prev);
+	}
+	mutex_init(&adev->gfx.userq_sch_mutex);
+	mutex_init(&adev->gfx.workload_profile_mutex);
+	mutex_init(&adev->vcn.workload_profile_mutex);
 
 	amdgpu_device_init_apu_flags(adev);
 
@@ -3658,20 +4623,41 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	spin_lock_init(&adev->se_cac_idx_lock);
 	spin_lock_init(&adev->audio_endpt_idx_lock);
 	spin_lock_init(&adev->mm_stats.lock);
+	spin_lock_init(&adev->virt.rlcg_reg_lock);
+	spin_lock_init(&adev->wb.lock);
 
-	INIT_LIST_HEAD(&adev->shadow_list);
-	mutex_init(&adev->shadow_list_lock);
+	xa_init_flags(&adev->userq_xa, XA_FLAGS_LOCK_IRQ);
 
 	INIT_LIST_HEAD(&adev->reset_list);
 
 	INIT_LIST_HEAD(&adev->ras_list);
 
+	INIT_LIST_HEAD(&adev->pm.od_kobj_list);
+
+	xa_init(&adev->userq_doorbell_xa);
+
 	INIT_DELAYED_WORK(&adev->delayed_init_work,
 			  amdgpu_device_delayed_init_work_handler);
 	INIT_DELAYED_WORK(&adev->gfx.gfx_off_delay_work,
 			  amdgpu_device_delay_enable_gfx_off);
+	/*
+	 * Initialize the enforce_isolation work structures for each XCP
+	 * partition.  This work handler is responsible for enforcing shader
+	 * isolation on AMD GPUs.  It counts the number of emitted fences for
+	 * each GFX and compute ring.  If there are any fences, it schedules
+	 * the `enforce_isolation_work` to be run after a delay.  If there are
+	 * no fences, it signals the Kernel Fusion Driver (KFD) to resume the
+	 * runqueue.
+	 */
+	for (i = 0; i < MAX_XCP; i++) {
+		INIT_DELAYED_WORK(&adev->gfx.enforce_isolation[i].work,
+				  amdgpu_gfx_enforce_isolation_handler);
+		adev->gfx.enforce_isolation[i].adev = adev;
+		adev->gfx.enforce_isolation[i].xcp_id = i;
+	}
 
 	INIT_WORK(&adev->xgmi_reset_work, amdgpu_device_xgmi_reset_func);
+	INIT_WORK(&adev->userq_reset_work, amdgpu_userq_reset_work);
 
 	adev->gfx.gfx_off_req_count = 1;
 	adev->gfx.gfx_off_residency = 0;
@@ -3687,6 +4673,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	 * for throttling interrupt) = 60 seconds.
 	 */
 	ratelimit_state_init(&adev->throttling_logging_rs, (60 - 1) * HZ, 1);
+
 	ratelimit_set_flags(&adev->throttling_logging_rs, RATELIMIT_MSG_ON_RELEASE);
 
 	/* Registers mapping */
@@ -3703,20 +4690,17 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 		atomic_set(&adev->pm.pwr_state[i], POWER_STATE_UNKNOWN);
 
 	adev->rmmio = ioremap(adev->rmmio_base, adev->rmmio_size);
-	if (adev->rmmio == NULL) {
+	if (!adev->rmmio)
 		return -ENOMEM;
-	}
-	DRM_INFO("register mmio base: 0x%08X\n", (uint32_t)adev->rmmio_base);
-	DRM_INFO("register mmio size: %u\n", (unsigned)adev->rmmio_size);
-
-	amdgpu_device_get_pcie_info(adev);
 
-	if (amdgpu_mcbp)
-		DRM_INFO("MCBP is enabled\n");
+	dev_info(adev->dev, "register mmio base: 0x%08X\n",
+		 (uint32_t)adev->rmmio_base);
+	dev_info(adev->dev, "register mmio size: %u\n",
+		 (unsigned int)adev->rmmio_size);
 
 	/*
 	 * Reset domain needs to be present early, before XGMI hive discovered
-	 * (if any) and intitialized to use reset sem and in_gpu reset flag
+	 * (if any) and initialized to use reset sem and in_gpu reset flag
 	 * early on during init and before calling to RREG32.
 	 */
 	adev->reset_domain = amdgpu_reset_create_reset_domain(SINGLE_DEVICE, "amdgpu-reset-dev");
@@ -3724,7 +4708,9 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 		return -ENOMEM;
 
 	/* detect hw virtualization here */
-	amdgpu_detect_virtualization(adev);
+	amdgpu_virt_init(adev);
+
+	amdgpu_device_get_pcie_info(adev);
 
 	r = amdgpu_device_get_job_timeout_settings(adev);
 	if (r) {
@@ -3732,19 +4718,41 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 		return r;
 	}
 
+	amdgpu_device_set_mcbp(adev);
+
+	/*
+	 * By default, use default mode where all blocks are expected to be
+	 * initialized. At present a 'swinit' of blocks is required to be
+	 * completed before the need for a different level is detected.
+	 */
+	amdgpu_set_init_level(adev, AMDGPU_INIT_LEVEL_DEFAULT);
 	/* early init functions */
 	r = amdgpu_device_ip_early_init(adev);
 	if (r)
 		return r;
 
-	/* Get rid of things like offb */
-	r = drm_aperture_remove_conflicting_pci_framebuffers(adev->pdev, &amdgpu_kms_driver);
-	if (r)
-		return r;
+	/*
+	 * No need to remove conflicting FBs for non-display class devices.
+	 * This prevents the sysfb from being freed accidently.
+	 */
+	if ((pdev->class >> 8) == PCI_CLASS_DISPLAY_VGA ||
+	    (pdev->class >> 8) == PCI_CLASS_DISPLAY_OTHER) {
+		/* Get rid of things like offb */
+		r = aperture_remove_conflicting_pci_devices(adev->pdev, amdgpu_kms_driver.name);
+		if (r)
+			return r;
+	}
 
 	/* Enable TMZ based on IP_VERSION */
 	amdgpu_gmc_tmz_set(adev);
 
+	if (amdgpu_sriov_vf(adev) &&
+	    amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 3, 0))
+		/* VF MMIO access (except mailbox range) from CPU
+		 * will be blocked during sriov runtime
+		 */
+		adev->virt.caps |= AMDGPU_VF_MMIO_ACCESS_PROTECT;
+
 	amdgpu_gmc_noretry_set(adev);
 	/* Need to get xgmi info early to decide the reset behavior*/
 	if (adev->gmc.xgmi.supported) {
@@ -3754,20 +4762,30 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	}
 
 	/* enable PCIE atomic ops */
-	if (amdgpu_sriov_vf(adev))
-		adev->have_atomics_support = ((struct amd_sriov_msg_pf2vf_info *)
-			adev->virt.fw_reserve.p_pf2vf)->pcie_atomic_ops_support_flags ==
-			(PCI_EXP_DEVCAP2_ATOMIC_COMP32 | PCI_EXP_DEVCAP2_ATOMIC_COMP64);
-	else
+	if (amdgpu_sriov_vf(adev)) {
+		if (adev->virt.fw_reserve.p_pf2vf)
+			adev->have_atomics_support = ((struct amd_sriov_msg_pf2vf_info *)
+						      adev->virt.fw_reserve.p_pf2vf)->pcie_atomic_ops_support_flags ==
+				(PCI_EXP_DEVCAP2_ATOMIC_COMP32 | PCI_EXP_DEVCAP2_ATOMIC_COMP64);
+	/* APUs w/ gfx9 onwards doesn't reply on PCIe atomics, rather it is a
+	 * internal path natively support atomics, set have_atomics_support to true.
+	 */
+	} else if ((adev->flags & AMD_IS_APU) &&
+		   (amdgpu_ip_version(adev, GC_HWIP, 0) >
+		    IP_VERSION(9, 0, 0))) {
+		adev->have_atomics_support = true;
+	} else {
 		adev->have_atomics_support =
 			!pci_enable_atomic_ops_to_root(adev->pdev,
 					  PCI_EXP_DEVCAP2_ATOMIC_COMP32 |
 					  PCI_EXP_DEVCAP2_ATOMIC_COMP64);
+	}
+
 	if (!adev->have_atomics_support)
 		dev_info(adev->dev, "PCIE atomic ops is not supported\n");
 
 	/* doorbell bar mapping and doorbell index init*/
-	amdgpu_device_doorbell_init(adev);
+	amdgpu_doorbell_init(adev);
 
 	if (amdgpu_emu_mode == 1) {
 		/* post the asic on emulation mode */
@@ -3778,7 +4796,8 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	amdgpu_reset_init(adev);
 
 	/* detect if we are with an SRIOV vbios */
-	amdgpu_device_detect_sriov_bios(adev);
+	if (adev->bios)
+		amdgpu_device_detect_sriov_bios(adev);
 
 	/* check if we need to reset the asic
 	 *  E.g., driver was not cleanly unloaded previously, etc.
@@ -3786,26 +4805,24 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	if (!amdgpu_sriov_vf(adev) && amdgpu_asic_need_reset_on_init(adev)) {
 		if (adev->gmc.xgmi.num_physical_nodes) {
 			dev_info(adev->dev, "Pending hive reset.\n");
-			adev->gmc.xgmi.pending_reset = true;
-			/* Only need to init necessary block for SMU to handle the reset */
-			for (i = 0; i < adev->num_ip_blocks; i++) {
-				if (!adev->ip_blocks[i].status.valid)
-					continue;
-				if (!(adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC ||
-				      adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON ||
-				      adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH ||
-				      adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC)) {
-					DRM_DEBUG("IP %s disabled for hw_init.\n",
-						adev->ip_blocks[i].version->funcs->name);
-					adev->ip_blocks[i].status.hw = true;
-				}
-			}
+			amdgpu_set_init_level(adev,
+					      AMDGPU_INIT_LEVEL_MINIMAL_XGMI);
+		} else if (amdgpu_ip_version(adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 10) &&
+				   !amdgpu_device_has_display_hardware(adev)) {
+					r = psp_gpu_reset(adev);
 		} else {
-			r = amdgpu_asic_reset(adev);
-			if (r) {
-				dev_err(adev->dev, "asic reset on init failed\n");
-				goto failed;
-			}
+				tmp = amdgpu_reset_method;
+				/* It should do a default reset when loading or reloading the driver,
+				 * regardless of the module parameter reset_method.
+				 */
+				amdgpu_reset_method = AMD_RESET_METHOD_NONE;
+				r = amdgpu_asic_reset(adev);
+				amdgpu_reset_method = tmp;
+		}
+
+		if (r) {
+		  dev_err(adev->dev, "asic reset on init failed\n");
+		  goto failed;
 		}
 	}
 
@@ -3816,7 +4833,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 			r = -EINVAL;
 			goto failed;
 		}
-		DRM_INFO("GPU posting now...\n");
+		dev_info(adev->dev, "GPU posting now...\n");
 		r = amdgpu_device_asic_init(adev);
 		if (r) {
 			dev_err(adev->dev, "gpu post error!\n");
@@ -3824,25 +4841,26 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 		}
 	}
 
-	if (adev->is_atom_fw) {
-		/* Initialize clocks */
-		r = amdgpu_atomfirmware_get_clock_info(adev);
-		if (r) {
-			dev_err(adev->dev, "amdgpu_atomfirmware_get_clock_info failed\n");
-			amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
-			goto failed;
-		}
-	} else {
-		/* Initialize clocks */
-		r = amdgpu_atombios_get_clock_info(adev);
-		if (r) {
-			dev_err(adev->dev, "amdgpu_atombios_get_clock_info failed\n");
-			amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
-			goto failed;
+	if (adev->bios) {
+		if (adev->is_atom_fw) {
+			/* Initialize clocks */
+			r = amdgpu_atomfirmware_get_clock_info(adev);
+			if (r) {
+				dev_err(adev->dev, "amdgpu_atomfirmware_get_clock_info failed\n");
+				amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
+				goto failed;
+			}
+		} else {
+			/* Initialize clocks */
+			r = amdgpu_atombios_get_clock_info(adev);
+			if (r) {
+				dev_err(adev->dev, "amdgpu_atombios_get_clock_info failed\n");
+				amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
+				goto failed;
+			}
+			/* init i2c buses */
+			amdgpu_i2c_init(adev);
 		}
-		/* init i2c buses */
-		if (!amdgpu_device_has_dc_support(adev))
-			amdgpu_atombios_i2c_init(adev);
 	}
 
 fence_driver_init:
@@ -3859,18 +4877,6 @@ fence_driver_init:
 
 	r = amdgpu_device_ip_init(adev);
 	if (r) {
-		/* failed in exclusive mode due to timeout */
-		if (amdgpu_sriov_vf(adev) &&
-		    !amdgpu_sriov_runtime(adev) &&
-		    amdgpu_virt_mmio_blocked(adev) &&
-		    !amdgpu_virt_wait_reset(adev)) {
-			dev_err(adev->dev, "VF exclusive mode timeout\n");
-			/* Don't send request since VF is inactive. */
-			adev->virt.caps &= ~AMDGPU_SRIOV_CAPS_RUNTIME;
-			adev->virt.ops = NULL;
-			r = -EAGAIN;
-			goto release_ras_con;
-		}
 		dev_err(adev->dev, "amdgpu_device_ip_init failed\n");
 		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_AMDGPU_INIT_FAIL, 0, 0);
 		goto release_ras_con;
@@ -3897,25 +4903,6 @@ fence_driver_init:
 	/* Get a log2 for easy divisions. */
 	adev->mm_stats.log2_max_MBps = ilog2(max(1u, max_MBps));
 
-	r = amdgpu_pm_sysfs_init(adev);
-	if (r)
-		DRM_ERROR("registering pm sysfs failed (%d).\n", r);
-
-	r = amdgpu_ucode_sysfs_init(adev);
-	if (r) {
-		adev->ucode_sysfs_en = false;
-		DRM_ERROR("Creating firmware sysfs failed (%d).\n", r);
-	} else
-		adev->ucode_sysfs_en = true;
-
-	r = amdgpu_psp_sysfs_init(adev);
-	if (r) {
-		adev->psp_sysfs_en = false;
-		if (!amdgpu_sriov_vf(adev))
-			DRM_ERROR("Creating psp sysfs failed\n");
-	} else
-		adev->psp_sysfs_en = true;
-
 	/*
 	 * Register gpu instance before amdgpu_device_enable_mgpu_fan_boost.
 	 * Otherwise the mgpu fan boost feature will be skipped due to the
@@ -3926,7 +4913,7 @@ fence_driver_init:
 	/* enable clockgating, etc. after ib tests, etc. since some blocks require
 	 * explicit gating rather than handling it automatically.
 	 */
-	if (!adev->gmc.xgmi.pending_reset) {
+	if (adev->init_lvl->level != AMDGPU_INIT_LEVEL_MINIMAL_XGMI) {
 		r = amdgpu_device_ip_late_init(adev);
 		if (r) {
 			dev_err(adev->dev, "amdgpu_device_ip_late_init failed\n");
@@ -3939,12 +4926,19 @@ fence_driver_init:
 				   msecs_to_jiffies(AMDGPU_RESUME_MS));
 	}
 
-	if (amdgpu_sriov_vf(adev))
+	if (amdgpu_sriov_vf(adev)) {
+		amdgpu_virt_release_full_gpu(adev, true);
 		flush_delayed_work(&adev->delayed_init_work);
+	}
 
-	r = sysfs_create_files(&adev->dev->kobj, amdgpu_dev_attributes);
-	if (r)
-		dev_err(adev->dev, "Could not create amdgpu device attr\n");
+	if (adev->init_lvl->level == AMDGPU_INIT_LEVEL_MINIMAL_XGMI)
+		amdgpu_xgmi_reset_on_init(adev);
+	/*
+	 * Place those sysfs registering after `late_init`. As some of those
+	 * operations performed in `late_init` might affect the sysfs
+	 * interfaces creating.
+	 */
+	r = amdgpu_device_sys_interface_init(adev);
 
 	if (IS_ENABLED(CONFIG_PERF_EVENTS))
 		r = amdgpu_pmu_init(adev);
@@ -3957,13 +4951,14 @@ fence_driver_init:
 
 	/* if we have > 1 VGA cards, then disable the amdgpu VGA resources */
 	/* this will fail for cards that aren't VGA class devices, just
-	 * ignore it */
+	 * ignore it
+	 */
 	if ((adev->pdev->class >> 8) == PCI_CLASS_DISPLAY_VGA)
 		vga_client_register(adev->pdev, amdgpu_device_vga_set_decode);
 
-	px = amdgpu_device_supports_px(ddev);
+	px = amdgpu_device_supports_px(adev);
 
-	if (px || (!pci_is_thunderbolt_attached(adev->pdev) &&
+	if (px || (!dev_is_removable(&adev->pdev->dev) &&
 				apple_gmux_detect(NULL, NULL)))
 		vga_switcheroo_register_client(adev->pdev,
 					       &amdgpu_switcheroo_ops, px);
@@ -3971,15 +4966,30 @@ fence_driver_init:
 	if (px)
 		vga_switcheroo_init_domain_pm_ops(adev->dev, &adev->vga_pm_domain);
 
-	if (adev->gmc.xgmi.pending_reset)
-		queue_delayed_work(system_wq, &mgpu_info.delayed_reset_work,
-				   msecs_to_jiffies(AMDGPU_RESUME_MS));
-
 	amdgpu_device_check_iommu_direct_map(adev);
 
+	adev->pm_nb.notifier_call = amdgpu_device_pm_notifier;
+	r = register_pm_notifier(&adev->pm_nb);
+	if (r)
+		goto failed;
+
 	return 0;
 
 release_ras_con:
+	if (amdgpu_sriov_vf(adev))
+		amdgpu_virt_release_full_gpu(adev, true);
+
+	/* failed in exclusive mode due to timeout */
+	if (amdgpu_sriov_vf(adev) &&
+		!amdgpu_sriov_runtime(adev) &&
+		amdgpu_virt_mmio_blocked(adev) &&
+		!amdgpu_virt_wait_reset(adev)) {
+		dev_err(adev->dev, "VF exclusive mode timeout\n");
+		/* Don't send request since VF is inactive. */
+		adev->virt.caps &= ~AMDGPU_SRIOV_CAPS_RUNTIME;
+		adev->virt.ops = NULL;
+		r = -EAGAIN;
+	}
 	amdgpu_release_ras_context(adev);
 
 failed:
@@ -3995,7 +5005,7 @@ static void amdgpu_device_unmap_mmio(struct amdgpu_device *adev)
 	unmap_mapping_range(adev->ddev.anon_inode->i_mapping, 0, 0, 1);
 
 	/* Unmap all mapped bars - Doorbell, registers and VRAM */
-	amdgpu_device_doorbell_fini(adev);
+	amdgpu_doorbell_fini(adev);
 
 	iounmap(adev->rmmio);
 	adev->rmmio = NULL;
@@ -4004,7 +5014,7 @@ static void amdgpu_device_unmap_mmio(struct amdgpu_device *adev)
 	adev->mman.aper_base_kaddr = NULL;
 
 	/* Memory manager related */
-	if (!adev->gmc.xgmi.connected_to_cpu) {
+	if (!adev->gmc.xgmi.connected_to_cpu && !adev->gmc.is_app_apu) {
 		arch_phys_wc_del(adev->gmc.vram_mtrr);
 		arch_io_free_memtype_wc(adev->gmc.aper_base, adev->gmc.aper_size);
 	}
@@ -4022,11 +5032,16 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev)
 {
 	dev_info(adev->dev, "amdgpu: finishing device.\n");
 	flush_delayed_work(&adev->delayed_init_work);
+
+	if (adev->mman.initialized)
+		drain_workqueue(adev->mman.bdev.wq);
 	adev->shutdown = true;
 
+	unregister_pm_notifier(&adev->pm_nb);
+
 	/* make sure IB test finished before entering exclusive mode
 	 * to avoid preemption on IB test
-	 * */
+	 */
 	if (amdgpu_sriov_vf(adev)) {
 		amdgpu_virt_request_full_gpu(adev, false);
 		amdgpu_virt_fini_data_exchange(adev);
@@ -4034,7 +5049,7 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev)
 
 	/* disable all interrupts */
 	amdgpu_irq_disable_all(adev);
-	if (adev->mode_info.mode_config_initialized){
+	if (adev->mode_info.mode_config_initialized) {
 		if (!drm_drv_uses_atomic_modeset(adev_to_drm(adev)))
 			drm_helper_force_disable_all(adev_to_drm(adev));
 		else
@@ -4042,20 +5057,13 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev)
 	}
 	amdgpu_fence_driver_hw_fini(adev);
 
-	if (adev->mman.initialized)
-		drain_workqueue(adev->mman.bdev.wq);
-
-	if (adev->pm.sysfs_initialized)
-		amdgpu_pm_sysfs_fini(adev);
-	if (adev->ucode_sysfs_en)
-		amdgpu_ucode_sysfs_fini(adev);
-	if (adev->psp_sysfs_en)
-		amdgpu_psp_sysfs_fini(adev);
-	sysfs_remove_files(&adev->dev->kobj, amdgpu_dev_attributes);
+	amdgpu_device_sys_interface_fini(adev);
 
 	/* disable ras feature must before hw fini */
 	amdgpu_ras_pre_fini(adev);
 
+	amdgpu_ttm_set_buffer_funcs_status(adev, false);
+
 	amdgpu_device_ip_fini_early(adev);
 
 	amdgpu_irq_fini_hw(adev);
@@ -4072,30 +5080,40 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev)
 
 void amdgpu_device_fini_sw(struct amdgpu_device *adev)
 {
-	int idx;
+	int i, idx;
 	bool px;
 
-	amdgpu_fence_driver_sw_fini(adev);
 	amdgpu_device_ip_fini(adev);
+	amdgpu_fence_driver_sw_fini(adev);
 	amdgpu_ucode_release(&adev->firmware.gpu_info_fw);
 	adev->accel_working = false;
 	dma_fence_put(rcu_dereference_protected(adev->gang_submit, true));
+	for (i = 0; i < MAX_XCP; ++i) {
+		dma_fence_put(adev->isolation[i].spearhead);
+		amdgpu_sync_free(&adev->isolation[i].active);
+		amdgpu_sync_free(&adev->isolation[i].prev);
+	}
 
 	amdgpu_reset_fini(adev);
 
 	/* free i2c buses */
-	if (!amdgpu_device_has_dc_support(adev))
-		amdgpu_i2c_fini(adev);
+	amdgpu_i2c_fini(adev);
+
+	if (adev->bios) {
+		if (amdgpu_emu_mode != 1)
+			amdgpu_atombios_fini(adev);
+		amdgpu_bios_release(adev);
+	}
 
-	if (amdgpu_emu_mode != 1)
-		amdgpu_atombios_fini(adev);
+	kfree(adev->fru_info);
+	adev->fru_info = NULL;
 
-	kfree(adev->bios);
-	adev->bios = NULL;
+	kfree(adev->xcp_mgr);
+	adev->xcp_mgr = NULL;
 
-	px = amdgpu_device_supports_px(adev_to_drm(adev));
+	px = amdgpu_device_supports_px(adev);
 
-	if (px || (!pci_is_thunderbolt_attached(adev->pdev) &&
+	if (px || (!dev_is_removable(&adev->pdev->dev) &&
 				apple_gmux_detect(NULL, NULL)))
 		vga_switcheroo_unregister_client(adev->pdev);
 
@@ -4109,20 +5127,20 @@ void amdgpu_device_fini_sw(struct amdgpu_device *adev)
 
 		iounmap(adev->rmmio);
 		adev->rmmio = NULL;
-		amdgpu_device_doorbell_fini(adev);
 		drm_dev_exit(idx);
 	}
 
 	if (IS_ENABLED(CONFIG_PERF_EVENTS))
 		amdgpu_pmu_fini(adev);
-	if (adev->mman.discovery_bin)
+	if (adev->discovery.bin)
 		amdgpu_discovery_fini(adev);
 
 	amdgpu_reset_put_reset_domain(adev->reset_domain);
 	adev->reset_domain = NULL;
 
 	kfree(adev->pci_state);
-
+	kfree(adev->pcie_reset_ctx.swds_pcistate);
+	kfree(adev->pcie_reset_ctx.swus_pcistate);
 }
 
 /**
@@ -4138,13 +5156,25 @@ static int amdgpu_device_evict_resources(struct amdgpu_device *adev)
 {
 	int ret;
 
-	/* No need to evict vram on APUs for suspend to ram or s2idle */
-	if ((adev->in_s3 || adev->in_s0ix) && (adev->flags & AMD_IS_APU))
+	/* No need to evict vram on APUs unless going to S4 */
+	if (!adev->in_s4 && (adev->flags & AMD_IS_APU))
+		return 0;
+
+	/* No need to evict when going to S5 through S4 callbacks */
+	if (system_state == SYSTEM_POWER_OFF)
 		return 0;
 
 	ret = amdgpu_ttm_evict_resources(adev, TTM_PL_VRAM);
-	if (ret)
-		DRM_WARN("evicting device resources failed\n");
+	if (ret) {
+		dev_warn(adev->dev, "evicting device resources failed\n");
+		return ret;
+	}
+
+	if (adev->in_s4) {
+		ret = ttm_device_prepare_hibernation(&adev->mman.bdev);
+		if (ret)
+			dev_err(adev->dev, "prepare hibernation failed, %d\n", ret);
+	}
 	return ret;
 }
 
@@ -4152,77 +5182,245 @@ static int amdgpu_device_evict_resources(struct amdgpu_device *adev)
  * Suspend & resume.
  */
 /**
- * amdgpu_device_suspend - initiate device suspend
+ * amdgpu_device_pm_notifier - Notification block for Suspend/Hibernate events
+ * @nb: notifier block
+ * @mode: suspend mode
+ * @data: data
+ *
+ * This function is called when the system is about to suspend or hibernate.
+ * It is used to set the appropriate flags so that eviction can be optimized
+ * in the pm prepare callback.
+ */
+static int amdgpu_device_pm_notifier(struct notifier_block *nb, unsigned long mode,
+				     void *data)
+{
+	struct amdgpu_device *adev = container_of(nb, struct amdgpu_device, pm_nb);
+
+	switch (mode) {
+	case PM_HIBERNATION_PREPARE:
+		adev->in_s4 = true;
+		break;
+	case PM_POST_HIBERNATION:
+		adev->in_s4 = false;
+		break;
+	}
+
+	return NOTIFY_DONE;
+}
+
+/**
+ * amdgpu_device_prepare - prepare for device suspend
  *
  * @dev: drm dev pointer
- * @fbcon : notify the fbdev of suspend
  *
- * Puts the hw in the suspend state (all asics).
+ * Prepare to put the hw in the suspend state (all asics).
  * Returns 0 for success or an error on failure.
  * Called at driver suspend.
  */
-int amdgpu_device_suspend(struct drm_device *dev, bool fbcon)
+int amdgpu_device_prepare(struct drm_device *dev)
 {
 	struct amdgpu_device *adev = drm_to_adev(dev);
-	int r = 0;
+	int i, r;
 
 	if (dev->switch_power_state == DRM_SWITCH_POWER_OFF)
 		return 0;
 
-	adev->in_suspend = true;
-
-	/* Evict the majority of BOs before grabbing the full access */
+	/* Evict the majority of BOs before starting suspend sequence */
 	r = amdgpu_device_evict_resources(adev);
 	if (r)
 		return r;
 
+	flush_delayed_work(&adev->gfx.gfx_off_delay_work);
+
+	for (i = 0; i < adev->num_ip_blocks; i++) {
+		if (!adev->ip_blocks[i].status.valid)
+			continue;
+		if (!adev->ip_blocks[i].version->funcs->prepare_suspend)
+			continue;
+		r = adev->ip_blocks[i].version->funcs->prepare_suspend(&adev->ip_blocks[i]);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+/**
+ * amdgpu_device_complete - complete power state transition
+ *
+ * @dev: drm dev pointer
+ *
+ * Undo the changes from amdgpu_device_prepare. This will be
+ * called on all resume transitions, including those that failed.
+ */
+void amdgpu_device_complete(struct drm_device *dev)
+{
+	struct amdgpu_device *adev = drm_to_adev(dev);
+	int i;
+
+	for (i = 0; i < adev->num_ip_blocks; i++) {
+		if (!adev->ip_blocks[i].status.valid)
+			continue;
+		if (!adev->ip_blocks[i].version->funcs->complete)
+			continue;
+		adev->ip_blocks[i].version->funcs->complete(&adev->ip_blocks[i]);
+	}
+}
+
+/**
+ * amdgpu_device_suspend - initiate device suspend
+ *
+ * @dev: drm dev pointer
+ * @notify_clients: notify in-kernel DRM clients
+ *
+ * Puts the hw in the suspend state (all asics).
+ * Returns 0 for success or an error on failure.
+ * Called at driver suspend.
+ */
+int amdgpu_device_suspend(struct drm_device *dev, bool notify_clients)
+{
+	struct amdgpu_device *adev = drm_to_adev(dev);
+	int r, rec;
+
+	if (dev->switch_power_state == DRM_SWITCH_POWER_OFF)
+		return 0;
+
+	adev->in_suspend = true;
+
 	if (amdgpu_sriov_vf(adev)) {
+		if (!adev->in_runpm)
+			amdgpu_amdkfd_suspend_process(adev);
 		amdgpu_virt_fini_data_exchange(adev);
 		r = amdgpu_virt_request_full_gpu(adev, false);
 		if (r)
 			return r;
 	}
 
-	if (amdgpu_acpi_smart_shift_update(dev, AMDGPU_SS_DEV_D3))
-		DRM_WARN("smart shift update failed\n");
+	r = amdgpu_acpi_smart_shift_update(adev, AMDGPU_SS_DEV_D3);
+	if (r)
+		goto unwind_sriov;
 
-	if (fbcon)
-		drm_fb_helper_set_suspend_unlocked(adev_to_drm(adev)->fb_helper, true);
+	if (notify_clients)
+		drm_client_dev_suspend(adev_to_drm(adev));
 
 	cancel_delayed_work_sync(&adev->delayed_init_work);
 
 	amdgpu_ras_suspend(adev);
 
-	amdgpu_device_ip_suspend_phase1(adev);
+	r = amdgpu_device_ip_suspend_phase1(adev);
+	if (r)
+		goto unwind_smartshift;
 
-	if (!adev->in_s0ix)
-		amdgpu_amdkfd_suspend(adev, adev->in_runpm);
+	amdgpu_amdkfd_suspend(adev, !amdgpu_sriov_vf(adev) && !adev->in_runpm);
+	r = amdgpu_userq_suspend(adev);
+	if (r)
+		goto unwind_ip_phase1;
 
 	r = amdgpu_device_evict_resources(adev);
 	if (r)
-		return r;
+		goto unwind_userq;
+
+	amdgpu_ttm_set_buffer_funcs_status(adev, false);
 
 	amdgpu_fence_driver_hw_fini(adev);
 
-	amdgpu_device_ip_suspend_phase2(adev);
+	r = amdgpu_device_ip_suspend_phase2(adev);
+	if (r)
+		goto unwind_evict;
 
 	if (amdgpu_sriov_vf(adev))
 		amdgpu_virt_release_full_gpu(adev, false);
 
 	return 0;
+
+unwind_evict:
+	if (adev->mman.buffer_funcs_ring->sched.ready)
+		amdgpu_ttm_set_buffer_funcs_status(adev, true);
+	amdgpu_fence_driver_hw_init(adev);
+
+unwind_userq:
+	rec = amdgpu_userq_resume(adev);
+	if (rec) {
+		dev_warn(adev->dev, "failed to re-initialize user queues: %d\n", rec);
+		return r;
+	}
+	rec = amdgpu_amdkfd_resume(adev, !amdgpu_sriov_vf(adev) && !adev->in_runpm);
+	if (rec) {
+		dev_warn(adev->dev, "failed to re-initialize kfd: %d\n", rec);
+		return r;
+	}
+
+unwind_ip_phase1:
+	/* suspend phase 1 = resume phase 3 */
+	rec = amdgpu_device_ip_resume_phase3(adev);
+	if (rec) {
+		dev_warn(adev->dev, "failed to re-initialize IPs phase1: %d\n", rec);
+		return r;
+	}
+
+unwind_smartshift:
+	rec = amdgpu_acpi_smart_shift_update(adev, AMDGPU_SS_DEV_D0);
+	if (rec) {
+		dev_warn(adev->dev, "failed to re-update smart shift: %d\n", rec);
+		return r;
+	}
+
+	if (notify_clients)
+		drm_client_dev_resume(adev_to_drm(adev));
+
+	amdgpu_ras_resume(adev);
+
+unwind_sriov:
+	if (amdgpu_sriov_vf(adev)) {
+		rec = amdgpu_virt_request_full_gpu(adev, true);
+		if (rec) {
+			dev_warn(adev->dev, "failed to reinitialize sriov: %d\n", rec);
+			return r;
+		}
+	}
+
+	adev->in_suspend = adev->in_s0ix = adev->in_s3 = false;
+
+	return r;
+}
+
+static inline int amdgpu_virt_resume(struct amdgpu_device *adev)
+{
+	int r;
+	unsigned int prev_physical_node_id = adev->gmc.xgmi.physical_node_id;
+
+	/* During VM resume, QEMU programming of VF MSIX table (register GFXMSIX_VECT0_ADDR_LO)
+	 * may not work. The access could be blocked by nBIF protection as VF isn't in
+	 * exclusive access mode. Exclusive access is enabled now, disable/enable MSIX
+	 * so that QEMU reprograms MSIX table.
+	 */
+	amdgpu_restore_msix(adev);
+
+	r = adev->gfxhub.funcs->get_xgmi_info(adev);
+	if (r)
+		return r;
+
+	dev_info(adev->dev, "xgmi node, old id %d, new id %d\n",
+		prev_physical_node_id, adev->gmc.xgmi.physical_node_id);
+
+	adev->vm_manager.vram_base_offset = adev->gfxhub.funcs->get_mc_fb_offset(adev);
+	adev->vm_manager.vram_base_offset +=
+		adev->gmc.xgmi.physical_node_id * adev->gmc.xgmi.node_segment_size;
+
+	return 0;
 }
 
 /**
  * amdgpu_device_resume - initiate device resume
  *
  * @dev: drm dev pointer
- * @fbcon : notify the fbdev of resume
+ * @notify_clients: notify in-kernel DRM clients
  *
  * Bring the hw back to operating state (all asics).
  * Returns 0 for success or an error on failure.
  * Called at driver resume.
  */
-int amdgpu_device_resume(struct drm_device *dev, bool fbcon)
+int amdgpu_device_resume(struct drm_device *dev, bool notify_clients)
 {
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	int r = 0;
@@ -4233,6 +5431,12 @@ int amdgpu_device_resume(struct drm_device *dev, bool fbcon)
 			return r;
 	}
 
+	if (amdgpu_virt_xgmi_migrate_enabled(adev)) {
+		r = amdgpu_virt_resume(adev);
+		if (r)
+			goto exit;
+	}
+
 	if (dev->switch_power_state == DRM_SWITCH_POWER_OFF)
 		return 0;
 
@@ -4252,7 +5456,14 @@ int amdgpu_device_resume(struct drm_device *dev, bool fbcon)
 		dev_err(adev->dev, "amdgpu_device_ip_resume failed (%d).\n", r);
 		goto exit;
 	}
-	amdgpu_fence_driver_hw_init(adev);
+
+	r = amdgpu_amdkfd_resume(adev, !amdgpu_sriov_vf(adev) && !adev->in_runpm);
+	if (r)
+		goto exit;
+
+	r = amdgpu_userq_resume(adev);
+	if (r)
+		goto exit;
 
 	r = amdgpu_device_ip_late_init(adev);
 	if (r)
@@ -4260,17 +5471,13 @@ int amdgpu_device_resume(struct drm_device *dev, bool fbcon)
 
 	queue_delayed_work(system_wq, &adev->delayed_init_work,
 			   msecs_to_jiffies(AMDGPU_RESUME_MS));
-
-	if (!adev->in_s0ix) {
-		r = amdgpu_amdkfd_resume(adev, adev->in_runpm);
-		if (r)
-			goto exit;
-	}
-
 exit:
 	if (amdgpu_sriov_vf(adev)) {
 		amdgpu_virt_init_data_exchange(adev);
 		amdgpu_virt_release_full_gpu(adev, true);
+
+		if (!r && !adev->in_runpm)
+			r = amdgpu_amdkfd_resume_process(adev);
 	}
 
 	if (r)
@@ -4279,8 +5486,8 @@ exit:
 	/* Make sure IB tests flushed */
 	flush_delayed_work(&adev->delayed_init_work);
 
-	if (fbcon)
-		drm_fb_helper_set_suspend_unlocked(adev_to_drm(adev)->fb_helper, false);
+	if (notify_clients)
+		drm_client_dev_resume(adev_to_drm(adev));
 
 	amdgpu_ras_resume(adev);
 
@@ -4305,13 +5512,12 @@ exit:
 		dev->dev->power.disable_depth--;
 #endif
 	}
-	adev->in_suspend = false;
 
-	if (adev->enable_mes)
-		amdgpu_mes_self_test(adev);
+	amdgpu_vram_mgr_clear_reset_blocks(adev);
+	adev->in_suspend = false;
 
-	if (amdgpu_acpi_smart_shift_update(dev, AMDGPU_SS_DEV_D0))
-		DRM_WARN("smart shift update failed\n");
+	if (amdgpu_acpi_smart_shift_update(adev, AMDGPU_SS_DEV_D0))
+		dev_warn(adev->dev, "smart shift update failed\n");
 
 	return 0;
 }
@@ -4342,7 +5548,8 @@ static bool amdgpu_device_ip_check_soft_reset(struct amdgpu_device *adev)
 			continue;
 		if (adev->ip_blocks[i].version->funcs->check_soft_reset)
 			adev->ip_blocks[i].status.hang =
-				adev->ip_blocks[i].version->funcs->check_soft_reset(adev);
+				adev->ip_blocks[i].version->funcs->check_soft_reset(
+					&adev->ip_blocks[i]);
 		if (adev->ip_blocks[i].status.hang) {
 			dev_info(adev->dev, "IP block:%s is hung!\n", adev->ip_blocks[i].version->funcs->name);
 			asic_hang = true;
@@ -4371,7 +5578,7 @@ static int amdgpu_device_ip_pre_soft_reset(struct amdgpu_device *adev)
 			continue;
 		if (adev->ip_blocks[i].status.hang &&
 		    adev->ip_blocks[i].version->funcs->pre_soft_reset) {
-			r = adev->ip_blocks[i].version->funcs->pre_soft_reset(adev);
+			r = adev->ip_blocks[i].version->funcs->pre_soft_reset(&adev->ip_blocks[i]);
 			if (r)
 				return r;
 		}
@@ -4433,7 +5640,7 @@ static int amdgpu_device_ip_soft_reset(struct amdgpu_device *adev)
 			continue;
 		if (adev->ip_blocks[i].status.hang &&
 		    adev->ip_blocks[i].version->funcs->soft_reset) {
-			r = adev->ip_blocks[i].version->funcs->soft_reset(adev);
+			r = adev->ip_blocks[i].version->funcs->soft_reset(&adev->ip_blocks[i]);
 			if (r)
 				return r;
 		}
@@ -4462,7 +5669,7 @@ static int amdgpu_device_ip_post_soft_reset(struct amdgpu_device *adev)
 			continue;
 		if (adev->ip_blocks[i].status.hang &&
 		    adev->ip_blocks[i].version->funcs->post_soft_reset)
-			r = adev->ip_blocks[i].version->funcs->post_soft_reset(adev);
+			r = adev->ip_blocks[i].version->funcs->post_soft_reset(&adev->ip_blocks[i]);
 		if (r)
 			return r;
 	}
@@ -4471,104 +5678,42 @@ static int amdgpu_device_ip_post_soft_reset(struct amdgpu_device *adev)
 }
 
 /**
- * amdgpu_device_recover_vram - Recover some VRAM contents
- *
- * @adev: amdgpu_device pointer
- *
- * Restores the contents of VRAM buffers from the shadows in GTT.  Used to
- * restore things like GPUVM page tables after a GPU reset where
- * the contents of VRAM might be lost.
- *
- * Returns:
- * 0 on success, negative error code on failure.
- */
-static int amdgpu_device_recover_vram(struct amdgpu_device *adev)
-{
-	struct dma_fence *fence = NULL, *next = NULL;
-	struct amdgpu_bo *shadow;
-	struct amdgpu_bo_vm *vmbo;
-	long r = 1, tmo;
-
-	if (amdgpu_sriov_runtime(adev))
-		tmo = msecs_to_jiffies(8000);
-	else
-		tmo = msecs_to_jiffies(100);
-
-	dev_info(adev->dev, "recover vram bo from shadow start\n");
-	mutex_lock(&adev->shadow_list_lock);
-	list_for_each_entry(vmbo, &adev->shadow_list, shadow_list) {
-		shadow = &vmbo->bo;
-		/* No need to recover an evicted BO */
-		if (shadow->tbo.resource->mem_type != TTM_PL_TT ||
-		    shadow->tbo.resource->start == AMDGPU_BO_INVALID_OFFSET ||
-		    shadow->parent->tbo.resource->mem_type != TTM_PL_VRAM)
-			continue;
-
-		r = amdgpu_bo_restore_shadow(shadow, &next);
-		if (r)
-			break;
-
-		if (fence) {
-			tmo = dma_fence_wait_timeout(fence, false, tmo);
-			dma_fence_put(fence);
-			fence = next;
-			if (tmo == 0) {
-				r = -ETIMEDOUT;
-				break;
-			} else if (tmo < 0) {
-				r = tmo;
-				break;
-			}
-		} else {
-			fence = next;
-		}
-	}
-	mutex_unlock(&adev->shadow_list_lock);
-
-	if (fence)
-		tmo = dma_fence_wait_timeout(fence, false, tmo);
-	dma_fence_put(fence);
-
-	if (r < 0 || tmo <= 0) {
-		dev_err(adev->dev, "recover vram bo from shadow failed, r is %ld, tmo is %ld\n", r, tmo);
-		return -EIO;
-	}
-
-	dev_info(adev->dev, "recover vram bo from shadow done\n");
-	return 0;
-}
-
-
-/**
  * amdgpu_device_reset_sriov - reset ASIC for SR-IOV vf
  *
  * @adev: amdgpu_device pointer
- * @from_hypervisor: request from hypervisor
+ * @reset_context: amdgpu reset context pointer
  *
  * do VF FLR and reinitialize Asic
  * return 0 means succeeded otherwise failed
  */
 static int amdgpu_device_reset_sriov(struct amdgpu_device *adev,
-				     bool from_hypervisor)
+				     struct amdgpu_reset_context *reset_context)
 {
 	int r;
 	struct amdgpu_hive_info *hive = NULL;
-	int retry_limit = 0;
-
-retry:
-	amdgpu_amdkfd_pre_reset(adev);
 
-	if (from_hypervisor)
+	if (test_bit(AMDGPU_HOST_FLR, &reset_context->flags)) {
+		if (!amdgpu_ras_get_fed_status(adev))
+			amdgpu_virt_ready_to_reset(adev);
+		amdgpu_virt_wait_reset(adev);
+		clear_bit(AMDGPU_HOST_FLR, &reset_context->flags);
 		r = amdgpu_virt_request_full_gpu(adev, true);
-	else
+	} else {
 		r = amdgpu_virt_reset_gpu(adev);
+	}
 	if (r)
 		return r;
 
+	amdgpu_ras_clear_err_state(adev);
+	amdgpu_irq_gpu_reset_resume_helper(adev);
+
+	/* some sw clean up VF needs to do before recover */
+	amdgpu_virt_post_reset(adev);
+
 	/* Resume IP prior to SMC */
 	r = amdgpu_device_ip_reinit_early_sriov(adev);
 	if (r)
-		goto error;
+		return r;
 
 	amdgpu_virt_init_data_exchange(adev);
 
@@ -4579,64 +5724,64 @@ retry:
 	/* now we are okay to resume SMC/CP/SDMA */
 	r = amdgpu_device_ip_reinit_late_sriov(adev);
 	if (r)
-		goto error;
+		return r;
 
 	hive = amdgpu_get_xgmi_hive(adev);
 	/* Update PSP FW topology after reset */
 	if (hive && adev->gmc.xgmi.num_physical_nodes > 1)
 		r = amdgpu_xgmi_update_topology(hive, adev);
-
 	if (hive)
 		amdgpu_put_xgmi_hive(hive);
+	if (r)
+		return r;
 
-	if (!r) {
-		amdgpu_irq_gpu_reset_resume_helper(adev);
-		r = amdgpu_ib_ring_tests(adev);
-
-		amdgpu_amdkfd_post_reset(adev);
-	}
+	r = amdgpu_ib_ring_tests(adev);
+	if (r)
+		return r;
 
-error:
-	if (!r && adev->virt.gim_feature & AMDGIM_FEATURE_GIM_FLR_VRAMLOST) {
+	if (adev->virt.gim_feature & AMDGIM_FEATURE_GIM_FLR_VRAMLOST)
 		amdgpu_inc_vram_lost(adev);
-		r = amdgpu_device_recover_vram(adev);
-	}
+
+	/* need to be called during full access so we can't do it later like
+	 * bare-metal does.
+	 */
+	amdgpu_amdkfd_post_reset(adev);
 	amdgpu_virt_release_full_gpu(adev, true);
 
-	if (AMDGPU_RETRY_SRIOV_RESET(r)) {
-		if (retry_limit < AMDGPU_MAX_RETRY_LIMIT) {
-			retry_limit++;
-			goto retry;
-		} else
-			DRM_ERROR("GPU reset retry is beyond the retry limit\n");
-	}
+	/* Aldebaran and gfx_11_0_3 support ras in SRIOV, so need resume ras during reset */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 4) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 5, 0) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(11, 0, 3))
+		amdgpu_ras_resume(adev);
 
-	return r;
+	amdgpu_virt_ras_telemetry_post_reset(adev);
+
+	return 0;
 }
 
 /**
- * amdgpu_device_has_job_running - check if there is any job in mirror list
+ * amdgpu_device_has_job_running - check if there is any unfinished job
  *
  * @adev: amdgpu_device pointer
  *
- * check if there is any job in mirror list
+ * check if there is any job running on the device when guest driver receives
+ * FLR notification from host driver. If there are still jobs running, then
+ * the guest driver will not respond the FLR reset. Instead, let the job hit
+ * the timeout and guest driver then issue the reset request.
  */
 bool amdgpu_device_has_job_running(struct amdgpu_device *adev)
 {
 	int i;
-	struct drm_sched_job *job;
 
 	for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 		struct amdgpu_ring *ring = adev->rings[i];
 
-		if (!ring || !ring->sched.thread)
+		if (!amdgpu_ring_sched_ready(ring))
 			continue;
 
-		spin_lock(&ring->sched.job_list_lock);
-		job = list_first_entry_or_null(&ring->sched.pending_list,
-					       struct drm_sched_job, list);
-		spin_unlock(&ring->sched.job_list_lock);
-		if (job)
+		if (amdgpu_fence_count_emitted(ring))
 			return true;
 	}
 	return false;
@@ -4695,42 +5840,83 @@ disabled:
 
 int amdgpu_device_mode1_reset(struct amdgpu_device *adev)
 {
-        u32 i;
-        int ret = 0;
+	u32 i;
+	int ret = 0;
 
-        amdgpu_atombios_scratch_regs_engine_hung(adev, true);
+	if (adev->bios)
+		amdgpu_atombios_scratch_regs_engine_hung(adev, true);
 
-        dev_info(adev->dev, "GPU mode1 reset\n");
+	dev_info(adev->dev, "GPU mode1 reset\n");
 
-        /* disable BM */
-        pci_clear_master(adev->pdev);
+	/* Cache the state before bus master disable. The saved config space
+	 * values are used in other cases like restore after mode-2 reset.
+	 */
+	amdgpu_device_cache_pci_state(adev->pdev);
 
-        amdgpu_device_cache_pci_state(adev->pdev);
+	/* disable BM */
+	pci_clear_master(adev->pdev);
 
-        if (amdgpu_dpm_is_mode1_reset_supported(adev)) {
-                dev_info(adev->dev, "GPU smu mode1 reset\n");
-                ret = amdgpu_dpm_mode1_reset(adev);
-        } else {
-                dev_info(adev->dev, "GPU psp mode1 reset\n");
-                ret = psp_gpu_reset(adev);
-        }
+	if (amdgpu_dpm_is_mode1_reset_supported(adev)) {
+		dev_info(adev->dev, "GPU smu mode1 reset\n");
+		ret = amdgpu_dpm_mode1_reset(adev);
+	} else {
+		dev_info(adev->dev, "GPU psp mode1 reset\n");
+		ret = psp_gpu_reset(adev);
+	}
 
-        if (ret)
-                dev_err(adev->dev, "GPU mode1 reset failed\n");
+	if (ret)
+		goto mode1_reset_failed;
 
-        amdgpu_device_load_pci_state(adev->pdev);
+	amdgpu_device_load_pci_state(adev->pdev);
+	ret = amdgpu_psp_wait_for_bootloader(adev);
+	if (ret)
+		goto mode1_reset_failed;
 
-        /* wait for asic to come out of reset */
-        for (i = 0; i < adev->usec_timeout; i++) {
-                u32 memsize = adev->nbio.funcs->get_memsize(adev);
+	/* wait for asic to come out of reset */
+	for (i = 0; i < adev->usec_timeout; i++) {
+		u32 memsize = adev->nbio.funcs->get_memsize(adev);
 
-                if (memsize != 0xffffffff)
-                        break;
-                udelay(1);
-        }
+		if (memsize != 0xffffffff)
+			break;
+		udelay(1);
+	}
+
+	if (i >= adev->usec_timeout) {
+		ret = -ETIMEDOUT;
+		goto mode1_reset_failed;
+	}
+
+	if (adev->bios)
+		amdgpu_atombios_scratch_regs_engine_hung(adev, false);
+
+	return 0;
 
-        amdgpu_atombios_scratch_regs_engine_hung(adev, false);
-        return ret;
+mode1_reset_failed:
+	dev_err(adev->dev, "GPU mode1 reset failed\n");
+	return ret;
+}
+
+int amdgpu_device_link_reset(struct amdgpu_device *adev)
+{
+	int ret = 0;
+
+	dev_info(adev->dev, "GPU link reset\n");
+
+	if (!amdgpu_reset_in_dpc(adev))
+		ret = amdgpu_dpm_link_reset(adev);
+
+	if (ret)
+		goto link_reset_failed;
+
+	ret = amdgpu_psp_wait_for_bootloader(adev);
+	if (ret)
+		goto link_reset_failed;
+
+	return 0;
+
+link_reset_failed:
+	dev_err(adev->dev, "GPU link reset failed\n");
+	return ret;
 }
 
 int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
@@ -4738,16 +5924,15 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
 {
 	int i, r = 0;
 	struct amdgpu_job *job = NULL;
+	struct amdgpu_device *tmp_adev = reset_context->reset_req_dev;
 	bool need_full_reset =
 		test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
 
 	if (reset_context->reset_req_dev == adev)
 		job = reset_context->job;
 
-	if (amdgpu_sriov_vf(adev)) {
-		/* stop the data exchange thread */
-		amdgpu_virt_fini_data_exchange(adev);
-	}
+	if (amdgpu_sriov_vf(adev))
+		amdgpu_virt_pre_reset(adev);
 
 	amdgpu_fence_driver_isr_toggle(adev, true);
 
@@ -4755,13 +5940,9 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
 	for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 		struct amdgpu_ring *ring = adev->rings[i];
 
-		if (!ring || !ring->sched.thread)
+		if (!amdgpu_ring_sched_ready(ring))
 			continue;
 
-		/*clear job fence from fence drv to avoid force_completion
-		 *leave NULL and vm flush fence in fence drv */
-		amdgpu_fence_driver_clear_job_fences(ring);
-
 		/* after all hw jobs are reset, hw fence is meaningless, so force_completion */
 		amdgpu_fence_driver_force_completion(ring);
 	}
@@ -4773,7 +5954,7 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
 
 	r = amdgpu_reset_prepare_hwcontext(adev, reset_context);
 	/* If reset handler not implemented, continue; otherwise return */
-	if (r == -ENOSYS)
+	if (r == -EOPNOTSUPP)
 		r = 0;
 	else
 		return r;
@@ -4795,6 +5976,16 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
 			}
 		}
 
+		if (!test_bit(AMDGPU_SKIP_COREDUMP, &reset_context->flags)) {
+			dev_info(tmp_adev->dev, "Dumping IP State\n");
+			/* Trigger ip dump before we reset the asic */
+			for (i = 0; i < tmp_adev->num_ip_blocks; i++)
+				if (tmp_adev->ip_blocks[i].version->funcs->dump_ip_state)
+					tmp_adev->ip_blocks[i].version->funcs
+						->dump_ip_state((void *)&tmp_adev->ip_blocks[i]);
+			dev_info(tmp_adev->dev, "Dumping IP State Completed\n");
+		}
+
 		if (need_full_reset)
 			r = amdgpu_device_ip_suspend(adev);
 		if (need_full_reset)
@@ -4807,188 +5998,55 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
 	return r;
 }
 
-static int amdgpu_reset_reg_dumps(struct amdgpu_device *adev)
-{
-	int i;
-
-	lockdep_assert_held(&adev->reset_domain->sem);
-
-	for (i = 0; i < adev->num_regs; i++) {
-		adev->reset_dump_reg_value[i] = RREG32(adev->reset_dump_reg_list[i]);
-		trace_amdgpu_reset_reg_dumps(adev->reset_dump_reg_list[i],
-					     adev->reset_dump_reg_value[i]);
-	}
-
-	return 0;
-}
-
-#ifdef CONFIG_DEV_COREDUMP
-static ssize_t amdgpu_devcoredump_read(char *buffer, loff_t offset,
-		size_t count, void *data, size_t datalen)
-{
-	struct drm_printer p;
-	struct amdgpu_device *adev = data;
-	struct drm_print_iterator iter;
-	int i;
-
-	iter.data = buffer;
-	iter.offset = 0;
-	iter.start = offset;
-	iter.remain = count;
-
-	p = drm_coredump_printer(&iter);
-
-	drm_printf(&p, "**** AMDGPU Device Coredump ****\n");
-	drm_printf(&p, "kernel: " UTS_RELEASE "\n");
-	drm_printf(&p, "module: " KBUILD_MODNAME "\n");
-	drm_printf(&p, "time: %lld.%09ld\n", adev->reset_time.tv_sec, adev->reset_time.tv_nsec);
-	if (adev->reset_task_info.pid)
-		drm_printf(&p, "process_name: %s PID: %d\n",
-			   adev->reset_task_info.process_name,
-			   adev->reset_task_info.pid);
-
-	if (adev->reset_vram_lost)
-		drm_printf(&p, "VRAM is lost due to GPU reset!\n");
-	if (adev->num_regs) {
-		drm_printf(&p, "AMDGPU register dumps:\nOffset:     Value:\n");
-
-		for (i = 0; i < adev->num_regs; i++)
-			drm_printf(&p, "0x%08x: 0x%08x\n",
-				   adev->reset_dump_reg_list[i],
-				   adev->reset_dump_reg_value[i]);
-	}
-
-	return count - iter.remain;
-}
-
-static void amdgpu_devcoredump_free(void *data)
+int amdgpu_device_reinit_after_reset(struct amdgpu_reset_context *reset_context)
 {
-}
-
-static void amdgpu_reset_capture_coredumpm(struct amdgpu_device *adev)
-{
-	struct drm_device *dev = adev_to_drm(adev);
-
-	ktime_get_ts64(&adev->reset_time);
-	dev_coredumpm(dev->dev, THIS_MODULE, adev, 0, GFP_KERNEL,
-		      amdgpu_devcoredump_read, amdgpu_devcoredump_free);
-}
-#endif
-
-int amdgpu_do_asic_reset(struct list_head *device_list_handle,
-			 struct amdgpu_reset_context *reset_context)
-{
-	struct amdgpu_device *tmp_adev = NULL;
-	bool need_full_reset, skip_hw_reset, vram_lost = false;
-	int r = 0;
-	bool gpu_reset_for_dev_remove = 0;
-
-	/* Try reset handler method first */
-	tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device,
-				    reset_list);
-	amdgpu_reset_reg_dumps(tmp_adev);
-
-	reset_context->reset_device_list = device_list_handle;
-	r = amdgpu_reset_perform_reset(tmp_adev, reset_context);
-	/* If reset handler not implemented, continue; otherwise return */
-	if (r == -ENOSYS)
-		r = 0;
-	else
-		return r;
-
-	/* Reset handler not implemented, use the default method */
-	need_full_reset =
-		test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
-	skip_hw_reset = test_bit(AMDGPU_SKIP_HW_RESET, &reset_context->flags);
-
-	gpu_reset_for_dev_remove =
-		test_bit(AMDGPU_RESET_FOR_DEVICE_REMOVE, &reset_context->flags) &&
-			test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
+	struct list_head *device_list_handle;
+	bool full_reset, vram_lost = false;
+	struct amdgpu_device *tmp_adev;
+	int r, init_level;
 
-	/*
-	 * ASIC reset has to be done on all XGMI hive nodes ASAP
-	 * to allow proper links negotiation in FW (within 1 sec)
-	 */
-	if (!skip_hw_reset && need_full_reset) {
-		list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
-			/* For XGMI run all resets in parallel to speed up the process */
-			if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) {
-				tmp_adev->gmc.xgmi.pending_reset = false;
-				if (!queue_work(system_unbound_wq, &tmp_adev->xgmi_reset_work))
-					r = -EALREADY;
-			} else
-				r = amdgpu_asic_reset(tmp_adev);
-
-			if (r) {
-				dev_err(tmp_adev->dev, "ASIC reset failed with error, %d for drm dev, %s",
-					 r, adev_to_drm(tmp_adev)->unique);
-				break;
-			}
-		}
-
-		/* For XGMI wait for all resets to complete before proceed */
-		if (!r) {
-			list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
-				if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) {
-					flush_work(&tmp_adev->xgmi_reset_work);
-					r = tmp_adev->asic_reset_res;
-					if (r)
-						break;
-				}
-			}
-		}
-	}
+	device_list_handle = reset_context->reset_device_list;
 
-	if (!r && amdgpu_ras_intr_triggered()) {
-		list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
-			if (tmp_adev->mmhub.ras && tmp_adev->mmhub.ras->ras_block.hw_ops &&
-			    tmp_adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count)
-				tmp_adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count(tmp_adev);
-		}
+	if (!device_list_handle)
+		return -EINVAL;
 
-		amdgpu_ras_intr_cleared();
-	}
+	full_reset = test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
 
-	/* Since the mode1 reset affects base ip blocks, the
-	 * phase1 ip blocks need to be resumed. Otherwise there
-	 * will be a BIOS signature error and the psp bootloader
-	 * can't load kdb on the next amdgpu install.
+	/**
+	 * If it's reset on init, it's default init level, otherwise keep level
+	 * as recovery level.
 	 */
-	if (gpu_reset_for_dev_remove) {
-		list_for_each_entry(tmp_adev, device_list_handle, reset_list)
-			amdgpu_device_ip_resume_phase1(tmp_adev);
-
-		goto end;
-	}
+	if (reset_context->method == AMD_RESET_METHOD_ON_INIT)
+			init_level = AMDGPU_INIT_LEVEL_DEFAULT;
+	else
+			init_level = AMDGPU_INIT_LEVEL_RESET_RECOVERY;
 
+	r = 0;
 	list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
-		if (need_full_reset) {
+		amdgpu_set_init_level(tmp_adev, init_level);
+		if (full_reset) {
 			/* post card */
+			amdgpu_reset_set_dpc_status(tmp_adev, false);
+			amdgpu_ras_clear_err_state(tmp_adev);
 			r = amdgpu_device_asic_init(tmp_adev);
 			if (r) {
 				dev_warn(tmp_adev->dev, "asic atom init failed!");
 			} else {
 				dev_info(tmp_adev->dev, "GPU reset succeeded, trying to resume\n");
-				r = amdgpu_amdkfd_resume_iommu(tmp_adev);
-				if (r)
-					goto out;
 
 				r = amdgpu_device_ip_resume_phase1(tmp_adev);
 				if (r)
 					goto out;
 
 				vram_lost = amdgpu_device_check_vram_lost(tmp_adev);
-#ifdef CONFIG_DEV_COREDUMP
-				tmp_adev->reset_vram_lost = vram_lost;
-				memset(&tmp_adev->reset_task_info, 0,
-						sizeof(tmp_adev->reset_task_info));
-				if (reset_context->job && reset_context->job->vm)
-					tmp_adev->reset_task_info =
-						reset_context->job->vm->task_info;
-				amdgpu_reset_capture_coredumpm(tmp_adev);
-#endif
+
+				if (!test_bit(AMDGPU_SKIP_COREDUMP, &reset_context->flags))
+					amdgpu_coredump(tmp_adev, false, vram_lost, reset_context->job);
+
 				if (vram_lost) {
-					DRM_INFO("VRAM is lost due to GPU reset!\n");
+					dev_info(
+						tmp_adev->dev,
+						"VRAM is lost due to GPU reset!\n");
 					amdgpu_inc_vram_lost(tmp_adev);
 				}
 
@@ -4996,10 +6054,22 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 				if (r)
 					return r;
 
+				r = amdgpu_xcp_restore_partition_mode(
+					tmp_adev->xcp_mgr);
+				if (r)
+					goto out;
+
 				r = amdgpu_device_ip_resume_phase2(tmp_adev);
 				if (r)
 					goto out;
 
+				if (tmp_adev->mman.buffer_funcs_ring->sched.ready)
+					amdgpu_ttm_set_buffer_funcs_status(tmp_adev, true);
+
+				r = amdgpu_device_ip_resume_phase3(tmp_adev);
+				if (r)
+					goto out;
+
 				if (vram_lost)
 					amdgpu_device_fill_reset_magic(tmp_adev);
 
@@ -5017,7 +6087,11 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 				if (r)
 					goto out;
 
-				drm_fb_helper_set_suspend_unlocked(adev_to_drm(tmp_adev)->fb_helper, false);
+				r = amdgpu_userq_post_reset(tmp_adev, vram_lost);
+				if (r)
+					goto out;
+
+				drm_client_dev_resume(adev_to_drm(tmp_adev));
 
 				/*
 				 * The GPU enters bad state once faulty pages
@@ -5029,7 +6103,7 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 				 * bad_page_threshold value to fix this once
 				 * probing driver again.
 				 */
-				if (!amdgpu_ras_eeprom_check_err_threshold(tmp_adev)) {
+				if (!amdgpu_ras_is_rma(tmp_adev)) {
 					/* must succeed. */
 					amdgpu_ras_resume(tmp_adev);
 				} else {
@@ -5047,27 +6121,102 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 
 out:
 		if (!r) {
+			/* IP init is complete now, set level as default */
+			amdgpu_set_init_level(tmp_adev,
+					      AMDGPU_INIT_LEVEL_DEFAULT);
 			amdgpu_irq_gpu_reset_resume_helper(tmp_adev);
 			r = amdgpu_ib_ring_tests(tmp_adev);
 			if (r) {
 				dev_err(tmp_adev->dev, "ib ring test failed (%d).\n", r);
-				need_full_reset = true;
 				r = -EAGAIN;
 				goto end;
 			}
 		}
 
-		if (!r)
-			r = amdgpu_device_recover_vram(tmp_adev);
-		else
+		if (r)
 			tmp_adev->asic_reset_res = r;
 	}
 
 end:
-	if (need_full_reset)
+	return r;
+}
+
+int amdgpu_do_asic_reset(struct list_head *device_list_handle,
+			 struct amdgpu_reset_context *reset_context)
+{
+	struct amdgpu_device *tmp_adev = NULL;
+	bool need_full_reset, skip_hw_reset;
+	int r = 0;
+
+	/* Try reset handler method first */
+	tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device,
+				    reset_list);
+
+	reset_context->reset_device_list = device_list_handle;
+	r = amdgpu_reset_perform_reset(tmp_adev, reset_context);
+	/* If reset handler not implemented, continue; otherwise return */
+	if (r == -EOPNOTSUPP)
+		r = 0;
+	else
+		return r;
+
+	/* Reset handler not implemented, use the default method */
+	need_full_reset =
+		test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
+	skip_hw_reset = test_bit(AMDGPU_SKIP_HW_RESET, &reset_context->flags);
+
+	/*
+	 * ASIC reset has to be done on all XGMI hive nodes ASAP
+	 * to allow proper links negotiation in FW (within 1 sec)
+	 */
+	if (!skip_hw_reset && need_full_reset) {
+		list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
+			/* For XGMI run all resets in parallel to speed up the process */
+			if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) {
+				if (!queue_work(system_unbound_wq,
+						&tmp_adev->xgmi_reset_work))
+					r = -EALREADY;
+			} else
+				r = amdgpu_asic_reset(tmp_adev);
+
+			if (r) {
+				dev_err(tmp_adev->dev,
+					"ASIC reset failed with error, %d for drm dev, %s",
+					r, adev_to_drm(tmp_adev)->unique);
+				goto out;
+			}
+		}
+
+		/* For XGMI wait for all resets to complete before proceed */
+		if (!r) {
+			list_for_each_entry(tmp_adev, device_list_handle,
+					    reset_list) {
+				if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) {
+					flush_work(&tmp_adev->xgmi_reset_work);
+					r = tmp_adev->asic_reset_res;
+					if (r)
+						break;
+				}
+			}
+		}
+	}
+
+	if (!r && amdgpu_ras_intr_triggered()) {
+		list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
+			amdgpu_ras_reset_error_count(tmp_adev,
+						     AMDGPU_RAS_BLOCK__MMHUB);
+		}
+
+		amdgpu_ras_intr_cleared();
+	}
+
+	r = amdgpu_device_reinit_after_reset(reset_context);
+	if (r == -EAGAIN)
 		set_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
 	else
 		clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
+
+out:
 	return r;
 }
 
@@ -5076,6 +6225,7 @@ static void amdgpu_device_set_mp1_state(struct amdgpu_device *adev)
 
 	switch (amdgpu_asic_reset_method(adev)) {
 	case AMD_RESET_METHOD_MODE1:
+	case AMD_RESET_METHOD_LINK:
 		adev->mp1_state = PP_MP1_STATE_SHUTDOWN;
 		break;
 	case AMD_RESET_METHOD_MODE2:
@@ -5163,6 +6313,7 @@ static inline void amdgpu_device_stop_pending_resets(struct amdgpu_device *adev)
 	if (!amdgpu_sriov_vf(adev))
 		cancel_work(&adev->reset_work);
 #endif
+	cancel_work(&adev->userq_reset_work);
 
 	if (adev->kfd.dev)
 		cancel_work(&adev->kfd.reset_work);
@@ -5175,89 +6326,80 @@ static inline void amdgpu_device_stop_pending_resets(struct amdgpu_device *adev)
 
 }
 
-/**
- * amdgpu_device_gpu_recover - reset the asic and recover scheduler
- *
- * @adev: amdgpu_device pointer
- * @job: which job trigger hang
- * @reset_context: amdgpu reset context pointer
- *
- * Attempt to reset the GPU if it has hung (all asics).
- * Attempt to do soft-reset or full-reset and reinitialize Asic
- * Returns 0 for success or an error on failure.
- */
-
-int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
-			      struct amdgpu_job *job,
-			      struct amdgpu_reset_context *reset_context)
+static int amdgpu_device_health_check(struct list_head *device_list_handle)
 {
-	struct list_head device_list, *device_list_handle =  NULL;
-	bool job_signaled = false;
-	struct amdgpu_hive_info *hive = NULL;
-	struct amdgpu_device *tmp_adev = NULL;
-	int i, r = 0;
-	bool need_emergency_restart = false;
-	bool audio_suspended = false;
-	bool gpu_reset_for_dev_remove = false;
-
-	gpu_reset_for_dev_remove =
-			test_bit(AMDGPU_RESET_FOR_DEVICE_REMOVE, &reset_context->flags) &&
-				test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
-
-	/*
-	 * Special case: RAS triggered and full reset isn't supported
-	 */
-	need_emergency_restart = amdgpu_ras_need_emergency_restart(adev);
-
-	/*
-	 * Flush RAM to disk so that after reboot
-	 * the user can read log and see why the system rebooted.
-	 */
-	if (need_emergency_restart && amdgpu_ras_get_context(adev)->reboot) {
-		DRM_WARN("Emergency reboot.");
+	struct amdgpu_device *tmp_adev;
+	int ret = 0;
 
-		ksys_sync_helper();
-		emergency_restart();
+	list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
+		ret |= amdgpu_device_bus_status_check(tmp_adev);
 	}
 
-	dev_info(adev->dev, "GPU %s begin!\n",
-		need_emergency_restart ? "jobs stop":"reset");
+	return ret;
+}
 
-	if (!amdgpu_sriov_vf(adev))
-		hive = amdgpu_get_xgmi_hive(adev);
-	if (hive)
-		mutex_lock(&hive->hive_lock);
+static void amdgpu_device_recovery_prepare(struct amdgpu_device *adev,
+					  struct list_head *device_list,
+					  struct amdgpu_hive_info *hive)
+{
+	struct amdgpu_device *tmp_adev = NULL;
 
-	reset_context->job = job;
-	reset_context->hive = hive;
 	/*
 	 * Build list of devices to reset.
 	 * In case we are in XGMI hive mode, resort the device list
 	 * to put adev in the 1st position.
 	 */
-	INIT_LIST_HEAD(&device_list);
-	if (!amdgpu_sriov_vf(adev) && (adev->gmc.xgmi.num_physical_nodes > 1)) {
+	if (!amdgpu_sriov_vf(adev) && (adev->gmc.xgmi.num_physical_nodes > 1) && hive) {
 		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
-			list_add_tail(&tmp_adev->reset_list, &device_list);
-			if (gpu_reset_for_dev_remove && adev->shutdown)
+			list_add_tail(&tmp_adev->reset_list, device_list);
+			if (adev->shutdown)
 				tmp_adev->shutdown = true;
+			if (amdgpu_reset_in_dpc(adev))
+				tmp_adev->pcie_reset_ctx.in_link_reset = true;
 		}
-		if (!list_is_first(&adev->reset_list, &device_list))
-			list_rotate_to_front(&adev->reset_list, &device_list);
-		device_list_handle = &device_list;
+		if (!list_is_first(&adev->reset_list, device_list))
+			list_rotate_to_front(&adev->reset_list, device_list);
 	} else {
-		list_add_tail(&adev->reset_list, &device_list);
-		device_list_handle = &device_list;
+		list_add_tail(&adev->reset_list, device_list);
 	}
+}
 
-	/* We need to lock reset domain only once both for XGMI and single device */
-	tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device,
-				    reset_list);
+static void amdgpu_device_recovery_get_reset_lock(struct amdgpu_device *adev,
+						  struct list_head *device_list)
+{
+	struct amdgpu_device *tmp_adev = NULL;
+
+	if (list_empty(device_list))
+		return;
+	tmp_adev =
+		list_first_entry(device_list, struct amdgpu_device, reset_list);
 	amdgpu_device_lock_reset_domain(tmp_adev->reset_domain);
+}
 
-	/* block all schedulers and reset given job's ring */
-	list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
+static void amdgpu_device_recovery_put_reset_lock(struct amdgpu_device *adev,
+						  struct list_head *device_list)
+{
+	struct amdgpu_device *tmp_adev = NULL;
+
+	if (list_empty(device_list))
+		return;
+	tmp_adev =
+		list_first_entry(device_list, struct amdgpu_device, reset_list);
+	amdgpu_device_unlock_reset_domain(tmp_adev->reset_domain);
+}
+
+static void amdgpu_device_halt_activities(struct amdgpu_device *adev,
+					  struct amdgpu_job *job,
+					  struct amdgpu_reset_context *reset_context,
+					  struct list_head *device_list,
+					  struct amdgpu_hive_info *hive,
+					  bool need_emergency_restart)
+{
+	struct amdgpu_device *tmp_adev = NULL;
+	int i;
 
+	/* block all schedulers and reset given job's ring */
+	list_for_each_entry(tmp_adev, device_list, reset_list) {
 		amdgpu_device_set_mp1_state(tmp_adev);
 
 		/*
@@ -5271,32 +6413,33 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 		 * some audio codec errors.
 		 */
 		if (!amdgpu_device_suspend_display_audio(tmp_adev))
-			audio_suspended = true;
+			tmp_adev->pcie_reset_ctx.audio_suspended = true;
 
 		amdgpu_ras_set_error_query_ready(tmp_adev, false);
 
 		cancel_delayed_work_sync(&tmp_adev->delayed_init_work);
 
-		if (!amdgpu_sriov_vf(tmp_adev))
-			amdgpu_amdkfd_pre_reset(tmp_adev);
+		amdgpu_amdkfd_pre_reset(tmp_adev, reset_context);
 
 		/*
-		 * Mark these ASICs to be reseted as untracked first
+		 * Mark these ASICs to be reset as untracked first
 		 * And add them back after reset completed
 		 */
 		amdgpu_unregister_gpu_instance(tmp_adev);
 
-		drm_fb_helper_set_suspend_unlocked(adev_to_drm(tmp_adev)->fb_helper, true);
+		drm_client_dev_suspend(adev_to_drm(tmp_adev));
 
 		/* disable ras on ALL IPs */
-		if (!need_emergency_restart &&
-		      amdgpu_device_ip_need_full_reset(tmp_adev))
+		if (!need_emergency_restart && !amdgpu_reset_in_dpc(adev) &&
+		    amdgpu_device_ip_need_full_reset(tmp_adev))
 			amdgpu_ras_suspend(tmp_adev);
 
+		amdgpu_userq_pre_reset(tmp_adev);
+
 		for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 			struct amdgpu_ring *ring = tmp_adev->rings[i];
 
-			if (!ring || !ring->sched.thread)
+			if (!amdgpu_ring_sched_ready(ring))
 				continue;
 
 			drm_sched_stop(&ring->sched, job ? &job->base : NULL);
@@ -5306,28 +6449,18 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 		}
 		atomic_inc(&tmp_adev->gpu_reset_counter);
 	}
+}
 
-	if (need_emergency_restart)
-		goto skip_sched_resume;
-
-	/*
-	 * Must check guilty signal here since after this point all old
-	 * HW fences are force signaled.
-	 *
-	 * job->base holds a reference to parent fence
-	 */
-	if (job && dma_fence_is_signaled(&job->hw_fence)) {
-		job_signaled = true;
-		dev_info(adev->dev, "Guilty job already signaled, skipping HW reset");
-		goto skip_hw_reset;
-	}
+static int amdgpu_device_asic_reset(struct amdgpu_device *adev,
+			      struct list_head *device_list,
+			      struct amdgpu_reset_context *reset_context)
+{
+	struct amdgpu_device *tmp_adev = NULL;
+	int retry_limit = AMDGPU_MAX_RETRY_LIMIT;
+	int r = 0;
 
 retry:	/* Rest of adevs pre asic reset from XGMI hive. */
-	list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
-		if (gpu_reset_for_dev_remove) {
-			/* Workaroud for ASICs need to disable SMC first */
-			amdgpu_device_smu_fini_early(tmp_adev);
-		}
+	list_for_each_entry(tmp_adev, device_list, reset_list) {
 		r = amdgpu_device_pre_asic_reset(tmp_adev, reset_context);
 		/*TODO Should we stop ?*/
 		if (r) {
@@ -5335,73 +6468,109 @@ retry:	/* Rest of adevs pre asic reset from XGMI hive. */
 				  r, adev_to_drm(tmp_adev)->unique);
 			tmp_adev->asic_reset_res = r;
 		}
-
-		/*
-		 * Drop all pending non scheduler resets. Scheduler resets
-		 * were already dropped during drm_sched_stop
-		 */
-		amdgpu_device_stop_pending_resets(tmp_adev);
 	}
 
 	/* Actual ASIC resets if needed.*/
 	/* Host driver will handle XGMI hive reset for SRIOV */
 	if (amdgpu_sriov_vf(adev)) {
-		r = amdgpu_device_reset_sriov(adev, job ? false : true);
+
+		/* Bail out of reset early */
+		if (amdgpu_ras_is_rma(adev))
+			return -ENODEV;
+
+		if (amdgpu_ras_get_fed_status(adev) || amdgpu_virt_rcvd_ras_interrupt(adev)) {
+			dev_dbg(adev->dev, "Detected RAS error, wait for FLR completion\n");
+			amdgpu_ras_set_fed(adev, true);
+			set_bit(AMDGPU_HOST_FLR, &reset_context->flags);
+		}
+
+		r = amdgpu_device_reset_sriov(adev, reset_context);
+		if (AMDGPU_RETRY_SRIOV_RESET(r) && (retry_limit--) > 0) {
+			amdgpu_virt_release_full_gpu(adev, true);
+			goto retry;
+		}
 		if (r)
 			adev->asic_reset_res = r;
-
-		/* Aldebaran and gfx_11_0_3 support ras in SRIOV, so need resume ras during reset */
-		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2) ||
-		    adev->ip_versions[GC_HWIP][0] == IP_VERSION(11, 0, 3))
-			amdgpu_ras_resume(adev);
 	} else {
-		r = amdgpu_do_asic_reset(device_list_handle, reset_context);
+		r = amdgpu_do_asic_reset(device_list, reset_context);
 		if (r && r == -EAGAIN)
 			goto retry;
+	}
 
-		if (!r && gpu_reset_for_dev_remove)
-			goto recover_end;
+	list_for_each_entry(tmp_adev, device_list, reset_list) {
+		/*
+		 * Drop any pending non scheduler resets queued before reset is done.
+		 * Any reset scheduled after this point would be valid. Scheduler resets
+		 * were already dropped during drm_sched_stop and no new ones can come
+		 * in before drm_sched_start.
+		 */
+		amdgpu_device_stop_pending_resets(tmp_adev);
 	}
 
-skip_hw_reset:
+	return r;
+}
+
+static int amdgpu_device_sched_resume(struct list_head *device_list,
+			      struct amdgpu_reset_context *reset_context,
+			      bool   job_signaled)
+{
+	struct amdgpu_device *tmp_adev = NULL;
+	int i, r = 0;
 
 	/* Post ASIC reset for all devs .*/
-	list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
+	list_for_each_entry(tmp_adev, device_list, reset_list) {
 
 		for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 			struct amdgpu_ring *ring = tmp_adev->rings[i];
 
-			if (!ring || !ring->sched.thread)
+			if (!amdgpu_ring_sched_ready(ring))
 				continue;
 
-			drm_sched_start(&ring->sched, true);
+			drm_sched_start(&ring->sched, 0);
 		}
 
-		if (adev->enable_mes && adev->ip_versions[GC_HWIP][0] != IP_VERSION(11, 0, 3))
-			amdgpu_mes_self_test(tmp_adev);
-
-		if (!drm_drv_uses_atomic_modeset(adev_to_drm(tmp_adev)) && !job_signaled) {
+		if (!drm_drv_uses_atomic_modeset(adev_to_drm(tmp_adev)) && !job_signaled)
 			drm_helper_resume_force_mode(adev_to_drm(tmp_adev));
-		}
 
-		if (tmp_adev->asic_reset_res)
-			r = tmp_adev->asic_reset_res;
-
-		tmp_adev->asic_reset_res = 0;
-
-		if (r) {
-			/* bad news, how to tell it to userspace ? */
-			dev_info(tmp_adev->dev, "GPU reset(%d) failed\n", atomic_read(&tmp_adev->gpu_reset_counter));
-			amdgpu_vf_error_put(tmp_adev, AMDGIM_ERROR_VF_GPU_RESET_FAIL, 0, r);
+		if (tmp_adev->asic_reset_res) {
+			/* bad news, how to tell it to userspace ?
+			 * for ras error, we should report GPU bad status instead of
+			 * reset failure
+			 */
+			if (reset_context->src != AMDGPU_RESET_SRC_RAS ||
+			    !amdgpu_ras_eeprom_check_err_threshold(tmp_adev))
+				dev_info(
+					tmp_adev->dev,
+					"GPU reset(%d) failed with error %d \n",
+					atomic_read(
+						&tmp_adev->gpu_reset_counter),
+					tmp_adev->asic_reset_res);
+			amdgpu_vf_error_put(tmp_adev,
+					    AMDGIM_ERROR_VF_GPU_RESET_FAIL, 0,
+					    tmp_adev->asic_reset_res);
+			if (!r)
+				r = tmp_adev->asic_reset_res;
+			tmp_adev->asic_reset_res = 0;
 		} else {
-			dev_info(tmp_adev->dev, "GPU reset(%d) succeeded!\n", atomic_read(&tmp_adev->gpu_reset_counter));
-			if (amdgpu_acpi_smart_shift_update(adev_to_drm(tmp_adev), AMDGPU_SS_DEV_D0))
-				DRM_WARN("smart shift update failed\n");
+			dev_info(tmp_adev->dev, "GPU reset(%d) succeeded!\n",
+				 atomic_read(&tmp_adev->gpu_reset_counter));
+			if (amdgpu_acpi_smart_shift_update(tmp_adev,
+							   AMDGPU_SS_DEV_D0))
+				dev_warn(tmp_adev->dev,
+					 "smart shift update failed\n");
 		}
 	}
 
-skip_sched_resume:
-	list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
+	return r;
+}
+
+static void amdgpu_device_gpu_resume(struct amdgpu_device *adev,
+			      struct list_head *device_list,
+			      bool   need_emergency_restart)
+{
+	struct amdgpu_device *tmp_adev = NULL;
+
+	list_for_each_entry(tmp_adev, device_list, reset_list) {
 		/* unlock kfd: SRIOV would do it separately */
 		if (!need_emergency_restart && !amdgpu_sriov_vf(tmp_adev))
 			amdgpu_amdkfd_post_reset(tmp_adev);
@@ -5412,19 +6581,125 @@ skip_sched_resume:
 		if (!adev->kfd.init_complete)
 			amdgpu_amdkfd_device_init(adev);
 
-		if (audio_suspended)
+		if (tmp_adev->pcie_reset_ctx.audio_suspended)
 			amdgpu_device_resume_display_audio(tmp_adev);
 
 		amdgpu_device_unset_mp1_state(tmp_adev);
 
 		amdgpu_ras_set_error_query_ready(tmp_adev, true);
+
 	}
+}
 
-recover_end:
-	tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device,
-					    reset_list);
-	amdgpu_device_unlock_reset_domain(tmp_adev->reset_domain);
 
+/**
+ * amdgpu_device_gpu_recover - reset the asic and recover scheduler
+ *
+ * @adev: amdgpu_device pointer
+ * @job: which job trigger hang
+ * @reset_context: amdgpu reset context pointer
+ *
+ * Attempt to reset the GPU if it has hung (all asics).
+ * Attempt to do soft-reset or full-reset and reinitialize Asic
+ * Returns 0 for success or an error on failure.
+ */
+
+int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
+			      struct amdgpu_job *job,
+			      struct amdgpu_reset_context *reset_context)
+{
+	struct list_head device_list;
+	bool job_signaled = false;
+	struct amdgpu_hive_info *hive = NULL;
+	int r = 0;
+	bool need_emergency_restart = false;
+
+	/*
+	 * If it reaches here because of hang/timeout and a RAS error is
+	 * detected at the same time, let RAS recovery take care of it.
+	 */
+	if (amdgpu_ras_is_err_state(adev, AMDGPU_RAS_BLOCK__ANY) &&
+	    !amdgpu_sriov_vf(adev) &&
+	    reset_context->src != AMDGPU_RESET_SRC_RAS) {
+		dev_dbg(adev->dev,
+			"Gpu recovery from source: %d yielding to RAS error recovery handling",
+			reset_context->src);
+		return 0;
+	}
+
+	/*
+	 * Special case: RAS triggered and full reset isn't supported
+	 */
+	need_emergency_restart = amdgpu_ras_need_emergency_restart(adev);
+
+	/*
+	 * Flush RAM to disk so that after reboot
+	 * the user can read log and see why the system rebooted.
+	 */
+	if (need_emergency_restart && amdgpu_ras_get_context(adev) &&
+		amdgpu_ras_get_context(adev)->reboot) {
+		dev_warn(adev->dev, "Emergency reboot.");
+
+		ksys_sync_helper();
+		emergency_restart();
+	}
+
+	dev_info(adev->dev, "GPU %s begin!. Source:  %d\n",
+		 need_emergency_restart ? "jobs stop" : "reset",
+		 reset_context->src);
+
+	if (!amdgpu_sriov_vf(adev))
+		hive = amdgpu_get_xgmi_hive(adev);
+	if (hive)
+		mutex_lock(&hive->hive_lock);
+
+	reset_context->job = job;
+	reset_context->hive = hive;
+	INIT_LIST_HEAD(&device_list);
+
+	amdgpu_device_recovery_prepare(adev, &device_list, hive);
+
+	if (!amdgpu_sriov_vf(adev)) {
+		r = amdgpu_device_health_check(&device_list);
+		if (r)
+			goto end_reset;
+	}
+
+	/* Cannot be called after locking reset domain */
+	amdgpu_ras_pre_reset(adev, &device_list);
+
+	/* We need to lock reset domain only once both for XGMI and single device */
+	amdgpu_device_recovery_get_reset_lock(adev, &device_list);
+
+	amdgpu_device_halt_activities(adev, job, reset_context, &device_list,
+				      hive, need_emergency_restart);
+	if (need_emergency_restart)
+		goto skip_sched_resume;
+	/*
+	 * Must check guilty signal here since after this point all old
+	 * HW fences are force signaled.
+	 *
+	 * job->base holds a reference to parent fence
+	 */
+	if (job && dma_fence_is_signaled(&job->hw_fence->base)) {
+		job_signaled = true;
+		dev_info(adev->dev, "Guilty job already signaled, skipping HW reset");
+		goto skip_hw_reset;
+	}
+
+	r = amdgpu_device_asic_reset(adev, &device_list, reset_context);
+	if (r)
+		goto reset_unlock;
+skip_hw_reset:
+	r = amdgpu_device_sched_resume(&device_list, reset_context, job_signaled);
+	if (r)
+		goto reset_unlock;
+skip_sched_resume:
+	amdgpu_device_gpu_resume(adev, &device_list, need_emergency_restart);
+reset_unlock:
+	amdgpu_device_recovery_put_reset_lock(adev, &device_list);
+	amdgpu_ras_post_reset(adev, &device_list);
+end_reset:
 	if (hive) {
 		mutex_unlock(&hive->hive_lock);
 		amdgpu_put_xgmi_hive(hive);
@@ -5434,23 +6709,111 @@ recover_end:
 		dev_info(adev->dev, "GPU reset end with ret = %d\n", r);
 
 	atomic_set(&adev->reset_domain->reset_res, r);
+
+	if (!r) {
+		struct amdgpu_task_info *ti = NULL;
+
+		if (job)
+			ti = amdgpu_vm_get_task_info_pasid(adev, job->pasid);
+
+		drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE,
+				     ti ? &ti->task : NULL);
+
+		amdgpu_vm_put_task_info(ti);
+	}
+
 	return r;
 }
 
 /**
+ * amdgpu_device_partner_bandwidth - find the bandwidth of appropriate partner
+ *
+ * @adev: amdgpu_device pointer
+ * @speed: pointer to the speed of the link
+ * @width: pointer to the width of the link
+ *
+ * Evaluate the hierarchy to find the speed and bandwidth capabilities of the
+ * first physical partner to an AMD dGPU.
+ * This will exclude any virtual switches and links.
+ */
+static void amdgpu_device_partner_bandwidth(struct amdgpu_device *adev,
+					    enum pci_bus_speed *speed,
+					    enum pcie_link_width *width)
+{
+	struct pci_dev *parent = adev->pdev;
+
+	if (!speed || !width)
+		return;
+
+	*speed = PCI_SPEED_UNKNOWN;
+	*width = PCIE_LNK_WIDTH_UNKNOWN;
+
+	if (amdgpu_device_pcie_dynamic_switching_supported(adev)) {
+		while ((parent = pci_upstream_bridge(parent))) {
+			/* skip upstream/downstream switches internal to dGPU*/
+			if (parent->vendor == PCI_VENDOR_ID_ATI)
+				continue;
+			*speed = pcie_get_speed_cap(parent);
+			*width = pcie_get_width_cap(parent);
+			break;
+		}
+	} else {
+		/* use the current speeds rather than max if switching is not supported */
+		pcie_bandwidth_available(adev->pdev, NULL, speed, width);
+	}
+}
+
+/**
+ * amdgpu_device_gpu_bandwidth - find the bandwidth of the GPU
+ *
+ * @adev: amdgpu_device pointer
+ * @speed: pointer to the speed of the link
+ * @width: pointer to the width of the link
+ *
+ * Evaluate the hierarchy to find the speed and bandwidth capabilities of the
+ * AMD dGPU which may be a virtual upstream bridge.
+ */
+static void amdgpu_device_gpu_bandwidth(struct amdgpu_device *adev,
+					enum pci_bus_speed *speed,
+					enum pcie_link_width *width)
+{
+	struct pci_dev *parent = adev->pdev;
+
+	if (!speed || !width)
+		return;
+
+	parent = pci_upstream_bridge(parent);
+	if (parent && parent->vendor == PCI_VENDOR_ID_ATI) {
+		/* use the upstream/downstream switches internal to dGPU */
+		*speed = pcie_get_speed_cap(parent);
+		*width = pcie_get_width_cap(parent);
+		while ((parent = pci_upstream_bridge(parent))) {
+			if (parent->vendor == PCI_VENDOR_ID_ATI) {
+				/* use the upstream/downstream switches internal to dGPU */
+				*speed = pcie_get_speed_cap(parent);
+				*width = pcie_get_width_cap(parent);
+			}
+		}
+	} else {
+		/* use the device itself */
+		*speed = pcie_get_speed_cap(adev->pdev);
+		*width = pcie_get_width_cap(adev->pdev);
+	}
+}
+
+/**
  * amdgpu_device_get_pcie_info - fence pcie info about the PCIE slot
  *
  * @adev: amdgpu_device pointer
  *
- * Fetchs and stores in the driver the PCIE capabilities (gen speed
+ * Fetches and stores in the driver the PCIE capabilities (gen speed
  * and lanes) of the slot the device is in. Handles APUs and
  * virtualized environments where PCIE config space may not be available.
  */
 static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev)
 {
-	struct pci_dev *pdev;
 	enum pci_bus_speed speed_cap, platform_speed_cap;
-	enum pcie_link_width platform_link_width;
+	enum pcie_link_width platform_link_width, link_width;
 
 	if (amdgpu_pcie_gen_cap)
 		adev->pm.pcie_gen_mask = amdgpu_pcie_gen_cap;
@@ -5459,7 +6822,7 @@ static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev)
 		adev->pm.pcie_mlw_mask = amdgpu_pcie_lane_cap;
 
 	/* covers APUs as well */
-	if (pci_is_root_bus(adev->pdev->bus)) {
+	if (pci_is_root_bus(adev->pdev->bus) && !amdgpu_passthrough(adev)) {
 		if (adev->pm.pcie_gen_mask == 0)
 			adev->pm.pcie_gen_mask = AMDGPU_DEFAULT_PCIE_GEN_MASK;
 		if (adev->pm.pcie_mlw_mask == 0)
@@ -5470,13 +6833,12 @@ static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev)
 	if (adev->pm.pcie_gen_mask && adev->pm.pcie_mlw_mask)
 		return;
 
-	pcie_bandwidth_available(adev->pdev, NULL,
-				 &platform_speed_cap, &platform_link_width);
+	amdgpu_device_partner_bandwidth(adev, &platform_speed_cap,
+					&platform_link_width);
+	amdgpu_device_gpu_bandwidth(adev, &speed_cap, &link_width);
 
 	if (adev->pm.pcie_gen_mask == 0) {
 		/* asic caps */
-		pdev = adev->pdev;
-		speed_cap = pcie_get_speed_cap(pdev);
 		if (speed_cap == PCI_SPEED_UNKNOWN) {
 			adev->pm.pcie_gen_mask |= (CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN1 |
 						  CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN2 |
@@ -5532,51 +6894,103 @@ static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev)
 		}
 	}
 	if (adev->pm.pcie_mlw_mask == 0) {
+		/* asic caps */
+		if (link_width == PCIE_LNK_WIDTH_UNKNOWN) {
+			adev->pm.pcie_mlw_mask |= AMDGPU_DEFAULT_ASIC_PCIE_MLW_MASK;
+		} else {
+			switch (link_width) {
+			case PCIE_LNK_X32:
+				adev->pm.pcie_mlw_mask |= (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X32 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X16 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X12 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1);
+				break;
+			case PCIE_LNK_X16:
+				adev->pm.pcie_mlw_mask |= (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X16 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X12 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1);
+				break;
+			case PCIE_LNK_X12:
+				adev->pm.pcie_mlw_mask |= (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X12 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1);
+				break;
+			case PCIE_LNK_X8:
+				adev->pm.pcie_mlw_mask |= (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1);
+				break;
+			case PCIE_LNK_X4:
+				adev->pm.pcie_mlw_mask |= (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1);
+				break;
+			case PCIE_LNK_X2:
+				adev->pm.pcie_mlw_mask |= (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1);
+				break;
+			case PCIE_LNK_X1:
+				adev->pm.pcie_mlw_mask |= CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1;
+				break;
+			default:
+				break;
+			}
+		}
+		/* platform caps */
 		if (platform_link_width == PCIE_LNK_WIDTH_UNKNOWN) {
 			adev->pm.pcie_mlw_mask |= AMDGPU_DEFAULT_PCIE_MLW_MASK;
 		} else {
 			switch (platform_link_width) {
 			case PCIE_LNK_X32:
-				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X32 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
+				adev->pm.pcie_mlw_mask |= (CAIL_PCIE_LINK_WIDTH_SUPPORT_X32 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
 				break;
 			case PCIE_LNK_X16:
-				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
+				adev->pm.pcie_mlw_mask |= (CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
 				break;
 			case PCIE_LNK_X12:
-				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
+				adev->pm.pcie_mlw_mask |= (CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
 				break;
 			case PCIE_LNK_X8:
-				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
+				adev->pm.pcie_mlw_mask |= (CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
 				break;
 			case PCIE_LNK_X4:
-				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
+				adev->pm.pcie_mlw_mask |= (CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
 				break;
 			case PCIE_LNK_X2:
-				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
+				adev->pm.pcie_mlw_mask |= (CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
 				break;
 			case PCIE_LNK_X1:
-				adev->pm.pcie_mlw_mask = CAIL_PCIE_LINK_WIDTH_SUPPORT_X1;
+				adev->pm.pcie_mlw_mask |= CAIL_PCIE_LINK_WIDTH_SUPPORT_X1;
 				break;
 			default:
 				break;
@@ -5599,29 +7013,37 @@ bool amdgpu_device_is_peer_accessible(struct amdgpu_device *adev,
 				      struct amdgpu_device *peer_adev)
 {
 #ifdef CONFIG_HSA_AMD_P2P
-	uint64_t address_mask = peer_adev->dev->dma_mask ?
-		~*peer_adev->dev->dma_mask : ~((1ULL << 32) - 1);
-	resource_size_t aper_limit =
-		adev->gmc.aper_base + adev->gmc.aper_size - 1;
 	bool p2p_access =
 		!adev->gmc.xgmi.connected_to_cpu &&
 		!(pci_p2pdma_distance(adev->pdev, peer_adev->dev, false) < 0);
+	if (!p2p_access)
+		dev_info(adev->dev, "PCIe P2P access from peer device %s is not supported by the chipset\n",
+			pci_name(peer_adev->pdev));
 
-	return pcie_p2p && p2p_access && (adev->gmc.visible_vram_size &&
-		adev->gmc.real_vram_size == adev->gmc.visible_vram_size &&
-		!(adev->gmc.aper_base & address_mask ||
-		  aper_limit & address_mask));
+	bool is_large_bar = adev->gmc.visible_vram_size &&
+		adev->gmc.real_vram_size == adev->gmc.visible_vram_size;
+	bool p2p_addressable = amdgpu_device_check_iommu_remap(peer_adev);
+
+	if (!p2p_addressable) {
+		uint64_t address_mask = peer_adev->dev->dma_mask ?
+			~*peer_adev->dev->dma_mask : ~((1ULL << 32) - 1);
+		resource_size_t aper_limit =
+			adev->gmc.aper_base + adev->gmc.aper_size - 1;
+
+		p2p_addressable = !(adev->gmc.aper_base & address_mask ||
+				     aper_limit & address_mask);
+	}
+	return pcie_p2p && is_large_bar && p2p_access && p2p_addressable;
 #else
 	return false;
 #endif
 }
 
-int amdgpu_device_baco_enter(struct drm_device *dev)
+int amdgpu_device_baco_enter(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 
-	if (!amdgpu_device_supports_baco(dev))
+	if (!amdgpu_device_supports_baco(adev))
 		return -ENOTSUPP;
 
 	if (ras && adev->ras_enabled &&
@@ -5631,13 +7053,12 @@ int amdgpu_device_baco_enter(struct drm_device *dev)
 	return amdgpu_dpm_baco_enter(adev);
 }
 
-int amdgpu_device_baco_exit(struct drm_device *dev)
+int amdgpu_device_baco_exit(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 	int ret = 0;
 
-	if (!amdgpu_device_supports_baco(dev))
+	if (!amdgpu_device_supports_baco(adev))
 		return -ENOTSUPP;
 
 	ret = amdgpu_dpm_baco_exit(adev);
@@ -5648,7 +7069,7 @@ int amdgpu_device_baco_exit(struct drm_device *dev)
 	    adev->nbio.funcs->enable_doorbell_interrupt)
 		adev->nbio.funcs->enable_doorbell_interrupt(adev, true);
 
-	if (amdgpu_passthrough(adev) &&
+	if (amdgpu_passthrough(adev) && adev->nbio.funcs &&
 	    adev->nbio.funcs->clear_doorbell_interrupt)
 		adev->nbio.funcs->clear_doorbell_interrupt(adev);
 
@@ -5668,45 +7089,52 @@ pci_ers_result_t amdgpu_pci_error_detected(struct pci_dev *pdev, pci_channel_sta
 {
 	struct drm_device *dev = pci_get_drvdata(pdev);
 	struct amdgpu_device *adev = drm_to_adev(dev);
-	int i;
+	struct amdgpu_hive_info *hive __free(xgmi_put_hive) =
+		amdgpu_get_xgmi_hive(adev);
+	struct amdgpu_reset_context reset_context;
+	struct list_head device_list;
 
-	DRM_INFO("PCI error: detected callback, state(%d)!!\n", state);
-
-	if (adev->gmc.xgmi.num_physical_nodes > 1) {
-		DRM_WARN("No support for XGMI hive yet...");
-		return PCI_ERS_RESULT_DISCONNECT;
-	}
+	dev_info(adev->dev, "PCI error: detected callback!!\n");
 
 	adev->pci_channel_state = state;
 
 	switch (state) {
 	case pci_channel_io_normal:
+		dev_info(adev->dev, "pci_channel_io_normal: state(%d)!!\n", state);
 		return PCI_ERS_RESULT_CAN_RECOVER;
-	/* Fatal error, prepare for slot reset */
 	case pci_channel_io_frozen:
-		/*
-		 * Locking adev->reset_domain->sem will prevent any external access
-		 * to GPU during PCI error recovery
-		 */
-		amdgpu_device_lock_reset_domain(adev->reset_domain);
-		amdgpu_device_set_mp1_state(adev);
-
-		/*
-		 * Block any work scheduling as we do for regular GPU reset
-		 * for the duration of the recovery
-		 */
-		for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
-			struct amdgpu_ring *ring = adev->rings[i];
-
-			if (!ring || !ring->sched.thread)
-				continue;
+		/* Fatal error, prepare for slot reset */
+		dev_info(adev->dev, "pci_channel_io_frozen: state(%d)!!\n", state);
+		if (hive) {
+			/* Hive devices should be able to support FW based
+			 * link reset on other devices, if not return.
+			 */
+			if (!amdgpu_dpm_is_link_reset_supported(adev)) {
+				dev_warn(adev->dev,
+					 "No support for XGMI hive yet...\n");
+				return PCI_ERS_RESULT_DISCONNECT;
+			}
+			/* Set dpc status only if device is part of hive
+			 * Non-hive devices should be able to recover after
+			 * link reset.
+			 */
+			amdgpu_reset_set_dpc_status(adev, true);
 
-			drm_sched_stop(&ring->sched, NULL);
+			mutex_lock(&hive->hive_lock);
 		}
-		atomic_inc(&adev->gpu_reset_counter);
+		memset(&reset_context, 0, sizeof(reset_context));
+		INIT_LIST_HEAD(&device_list);
+
+		amdgpu_device_recovery_prepare(adev, &device_list, hive);
+		amdgpu_device_recovery_get_reset_lock(adev, &device_list);
+		amdgpu_device_halt_activities(adev, NULL, &reset_context, &device_list,
+					      hive, false);
+		if (hive)
+			mutex_unlock(&hive->hive_lock);
 		return PCI_ERS_RESULT_NEED_RESET;
 	case pci_channel_io_perm_failure:
 		/* Permanent error, prepare for device removal */
+		dev_info(adev->dev, "pci_channel_io_perm_failure: state(%d)!!\n", state);
 		return PCI_ERS_RESULT_DISCONNECT;
 	}
 
@@ -5719,8 +7147,10 @@ pci_ers_result_t amdgpu_pci_error_detected(struct pci_dev *pdev, pci_channel_sta
  */
 pci_ers_result_t amdgpu_pci_mmio_enabled(struct pci_dev *pdev)
 {
+	struct drm_device *dev = pci_get_drvdata(pdev);
+	struct amdgpu_device *adev = drm_to_adev(dev);
 
-	DRM_INFO("PCI error: mmio enabled callback!!\n");
+	dev_info(adev->dev, "PCI error: mmio enabled callback!!\n");
 
 	/* TODO - dump whatever for debugging purposes */
 
@@ -5744,21 +7174,38 @@ pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev)
 {
 	struct drm_device *dev = pci_get_drvdata(pdev);
 	struct amdgpu_device *adev = drm_to_adev(dev);
-	int r, i;
 	struct amdgpu_reset_context reset_context;
-	u32 memsize;
+	struct amdgpu_device *tmp_adev;
+	struct amdgpu_hive_info *hive;
 	struct list_head device_list;
+	struct pci_dev *link_dev;
+	int r = 0, i, timeout;
+	u32 memsize;
+	u16 status;
 
-	DRM_INFO("PCI error: slot reset callback!!\n");
+	dev_info(adev->dev, "PCI error: slot reset callback!!\n");
 
 	memset(&reset_context, 0, sizeof(reset_context));
 
-	INIT_LIST_HEAD(&device_list);
-	list_add_tail(&adev->reset_list, &device_list);
+	if (adev->pcie_reset_ctx.swus)
+		link_dev = adev->pcie_reset_ctx.swus;
+	else
+		link_dev = adev->pdev;
+	/* wait for asic to come out of reset, timeout = 10s */
+	timeout = 10000;
+	do {
+		usleep_range(10000, 10500);
+		r = pci_read_config_word(link_dev, PCI_VENDOR_ID, &status);
+		timeout -= 10;
+	} while (timeout > 0 && (status != PCI_VENDOR_ID_ATI) &&
+		 (status != PCI_VENDOR_ID_AMD));
 
-	/* wait for asic to come out of reset */
-	msleep(500);
+	if ((status != PCI_VENDOR_ID_ATI) && (status != PCI_VENDOR_ID_AMD)) {
+		r = -ETIME;
+		goto out;
+	}
 
+	amdgpu_device_load_switch_state(adev);
 	/* Restore PCI confspace */
 	amdgpu_device_load_pci_state(pdev);
 
@@ -5778,26 +7225,40 @@ pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev)
 	reset_context.method = AMD_RESET_METHOD_NONE;
 	reset_context.reset_req_dev = adev;
 	set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
-	set_bit(AMDGPU_SKIP_HW_RESET, &reset_context.flags);
-
-	adev->no_hw_access = true;
-	r = amdgpu_device_pre_asic_reset(adev, &reset_context);
-	adev->no_hw_access = false;
-	if (r)
-		goto out;
+	set_bit(AMDGPU_SKIP_COREDUMP, &reset_context.flags);
+	INIT_LIST_HEAD(&device_list);
 
-	r = amdgpu_do_asic_reset(&device_list, &reset_context);
+	hive = amdgpu_get_xgmi_hive(adev);
+	if (hive) {
+		mutex_lock(&hive->hive_lock);
+		reset_context.hive = hive;
+		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
+			tmp_adev->pcie_reset_ctx.in_link_reset = true;
+			list_add_tail(&tmp_adev->reset_list, &device_list);
+		}
+	} else {
+		set_bit(AMDGPU_SKIP_HW_RESET, &reset_context.flags);
+		list_add_tail(&adev->reset_list, &device_list);
+	}
 
+	r = amdgpu_device_asic_reset(adev, &device_list, &reset_context);
 out:
 	if (!r) {
 		if (amdgpu_device_cache_pci_state(adev->pdev))
 			pci_restore_state(adev->pdev);
-
-		DRM_INFO("PCIe error recovery succeeded\n");
+		dev_info(adev->dev, "PCIe error recovery succeeded\n");
 	} else {
-		DRM_ERROR("PCIe error recovery failed, err:%d", r);
-		amdgpu_device_unset_mp1_state(adev);
-		amdgpu_device_unlock_reset_domain(adev->reset_domain);
+		dev_err(adev->dev, "PCIe error recovery failed, err:%d\n", r);
+		if (hive) {
+			list_for_each_entry(tmp_adev, &device_list, reset_list)
+				amdgpu_device_unset_mp1_state(tmp_adev);
+		}
+		amdgpu_device_recovery_put_reset_lock(adev, &device_list);
+	}
+
+	if (hive) {
+		mutex_unlock(&hive->hive_lock);
+		amdgpu_put_xgmi_hive(hive);
 	}
 
 	return r ? PCI_ERS_RESULT_DISCONNECT : PCI_ERS_RESULT_RECOVERED;
@@ -5814,26 +7275,95 @@ void amdgpu_pci_resume(struct pci_dev *pdev)
 {
 	struct drm_device *dev = pci_get_drvdata(pdev);
 	struct amdgpu_device *adev = drm_to_adev(dev);
-	int i;
-
+	struct list_head device_list;
+	struct amdgpu_hive_info *hive = NULL;
+	struct amdgpu_device *tmp_adev = NULL;
 
-	DRM_INFO("PCI error: resume callback!!\n");
+	dev_info(adev->dev, "PCI error: resume callback!!\n");
 
 	/* Only continue execution for the case of pci_channel_io_frozen */
 	if (adev->pci_channel_state != pci_channel_io_frozen)
 		return;
 
-	for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
-		struct amdgpu_ring *ring = adev->rings[i];
+	INIT_LIST_HEAD(&device_list);
 
-		if (!ring || !ring->sched.thread)
-			continue;
+	hive = amdgpu_get_xgmi_hive(adev);
+	if (hive) {
+		mutex_lock(&hive->hive_lock);
+		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
+			tmp_adev->pcie_reset_ctx.in_link_reset = false;
+			list_add_tail(&tmp_adev->reset_list, &device_list);
+		}
+	} else
+		list_add_tail(&adev->reset_list, &device_list);
+
+	amdgpu_device_sched_resume(&device_list, NULL, NULL);
+	amdgpu_device_gpu_resume(adev, &device_list, false);
+	amdgpu_device_recovery_put_reset_lock(adev, &device_list);
 
-		drm_sched_start(&ring->sched, true);
+	if (hive) {
+		mutex_unlock(&hive->hive_lock);
+		amdgpu_put_xgmi_hive(hive);
 	}
+}
+
+static void amdgpu_device_cache_switch_state(struct amdgpu_device *adev)
+{
+	struct pci_dev *swus, *swds;
+	int r;
+
+	swds = pci_upstream_bridge(adev->pdev);
+	if (!swds || swds->vendor != PCI_VENDOR_ID_ATI ||
+	    pci_pcie_type(swds) != PCI_EXP_TYPE_DOWNSTREAM)
+		return;
+	swus = pci_upstream_bridge(swds);
+	if (!swus ||
+	    (swus->vendor != PCI_VENDOR_ID_ATI &&
+	     swus->vendor != PCI_VENDOR_ID_AMD) ||
+	    pci_pcie_type(swus) != PCI_EXP_TYPE_UPSTREAM)
+		return;
+
+	/* If already saved, return */
+	if (adev->pcie_reset_ctx.swus)
+		return;
+	/* Upstream bridge is ATI, assume it's SWUS/DS architecture */
+	r = pci_save_state(swds);
+	if (r)
+		return;
+	adev->pcie_reset_ctx.swds_pcistate = pci_store_saved_state(swds);
+
+	r = pci_save_state(swus);
+	if (r)
+		return;
+	adev->pcie_reset_ctx.swus_pcistate = pci_store_saved_state(swus);
 
-	amdgpu_device_unset_mp1_state(adev);
-	amdgpu_device_unlock_reset_domain(adev->reset_domain);
+	adev->pcie_reset_ctx.swus = swus;
+}
+
+static void amdgpu_device_load_switch_state(struct amdgpu_device *adev)
+{
+	struct pci_dev *pdev;
+	int r;
+
+	if (!adev->pcie_reset_ctx.swds_pcistate ||
+	    !adev->pcie_reset_ctx.swus_pcistate)
+		return;
+
+	pdev = adev->pcie_reset_ctx.swus;
+	r = pci_load_saved_state(pdev, adev->pcie_reset_ctx.swus_pcistate);
+	if (!r) {
+		pci_restore_state(pdev);
+	} else {
+		dev_warn(adev->dev, "Failed to load SWUS state, err:%d\n", r);
+		return;
+	}
+
+	pdev = pci_upstream_bridge(adev->pdev);
+	r = pci_load_saved_state(pdev, adev->pcie_reset_ctx.swds_pcistate);
+	if (!r)
+		pci_restore_state(pdev);
+	else
+		dev_warn(adev->dev, "Failed to load SWDS state, err:%d\n", r);
 }
 
 bool amdgpu_device_cache_pci_state(struct pci_dev *pdev)
@@ -5842,6 +7372,9 @@ bool amdgpu_device_cache_pci_state(struct pci_dev *pdev)
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	int r;
 
+	if (amdgpu_sriov_vf(adev))
+		return false;
+
 	r = pci_save_state(pdev);
 	if (!r) {
 		kfree(adev->pci_state);
@@ -5849,14 +7382,16 @@ bool amdgpu_device_cache_pci_state(struct pci_dev *pdev)
 		adev->pci_state = pci_store_saved_state(pdev);
 
 		if (!adev->pci_state) {
-			DRM_ERROR("Failed to store PCI saved state");
+			dev_err(adev->dev, "Failed to store PCI saved state");
 			return false;
 		}
 	} else {
-		DRM_WARN("Failed to save PCI state, err:%d\n", r);
+		dev_warn(adev->dev, "Failed to save PCI state, err:%d\n", r);
 		return false;
 	}
 
+	amdgpu_device_cache_switch_state(adev);
+
 	return true;
 }
 
@@ -5874,7 +7409,7 @@ bool amdgpu_device_load_pci_state(struct pci_dev *pdev)
 	if (!r) {
 		pci_restore_state(pdev);
 	} else {
-		DRM_WARN("Failed to load PCI state, err:%d\n", r);
+		dev_warn(adev->dev, "Failed to load PCI state, err:%d\n", r);
 		return false;
 	}
 
@@ -5891,10 +7426,17 @@ void amdgpu_device_flush_hdp(struct amdgpu_device *adev,
 	if (adev->gmc.xgmi.connected_to_cpu)
 		return;
 
-	if (ring && ring->funcs->emit_hdp_flush)
+	if (ring && ring->funcs->emit_hdp_flush) {
 		amdgpu_ring_emit_hdp_flush(ring);
-	else
-		amdgpu_asic_flush_hdp(adev, ring);
+		return;
+	}
+
+	if (!ring && amdgpu_sriov_runtime(adev)) {
+		if (!amdgpu_kiq_hdp_flush(adev))
+			return;
+	}
+
+	amdgpu_hdp_flush(adev, ring);
 }
 
 void amdgpu_device_invalidate_hdp(struct amdgpu_device *adev,
@@ -5907,7 +7449,7 @@ void amdgpu_device_invalidate_hdp(struct amdgpu_device *adev,
 	if (adev->gmc.xgmi.connected_to_cpu)
 		return;
 
-	amdgpu_asic_invalidate_hdp(adev, ring);
+	amdgpu_hdp_invalidate(adev, ring);
 }
 
 int amdgpu_in_reset(struct amdgpu_device *adev)
@@ -5940,6 +7482,7 @@ void amdgpu_device_halt(struct amdgpu_device *adev)
 	struct pci_dev *pdev = adev->pdev;
 	struct drm_device *ddev = adev_to_drm(adev);
 
+	amdgpu_xcp_dev_unplug(adev);
 	drm_dev_unplug(ddev);
 
 	amdgpu_irq_disable_all(adev);
@@ -5988,6 +7531,22 @@ void amdgpu_device_pcie_port_wreg(struct amdgpu_device *adev,
 }
 
 /**
+ * amdgpu_device_get_gang - return a reference to the current gang
+ * @adev: amdgpu_device pointer
+ *
+ * Returns: A new reference to the current gang leader.
+ */
+struct dma_fence *amdgpu_device_get_gang(struct amdgpu_device *adev)
+{
+	struct dma_fence *fence;
+
+	rcu_read_lock();
+	fence = dma_fence_get_rcu_safe(&adev->gang_submit);
+	rcu_read_unlock();
+	return fence;
+}
+
+/**
  * amdgpu_device_switch_gang - switch to a new gang
  * @adev: amdgpu_device pointer
  * @gang: the gang to switch to
@@ -6001,25 +7560,117 @@ struct dma_fence *amdgpu_device_switch_gang(struct amdgpu_device *adev,
 {
 	struct dma_fence *old = NULL;
 
+	dma_fence_get(gang);
 	do {
 		dma_fence_put(old);
-		rcu_read_lock();
-		old = dma_fence_get_rcu_safe(&adev->gang_submit);
-		rcu_read_unlock();
-
+		old = amdgpu_device_get_gang(adev);
 		if (old == gang)
 			break;
 
-		if (!dma_fence_is_signaled(old))
+		if (!dma_fence_is_signaled(old)) {
+			dma_fence_put(gang);
 			return old;
+		}
 
 	} while (cmpxchg((struct dma_fence __force **)&adev->gang_submit,
 			 old, gang) != old);
 
+	/*
+	 * Drop it once for the exchanged reference in adev and once for the
+	 * thread local reference acquired in amdgpu_device_get_gang().
+	 */
+	dma_fence_put(old);
 	dma_fence_put(old);
 	return NULL;
 }
 
+/**
+ * amdgpu_device_enforce_isolation - enforce HW isolation
+ * @adev: the amdgpu device pointer
+ * @ring: the HW ring the job is supposed to run on
+ * @job: the job which is about to be pushed to the HW ring
+ *
+ * Makes sure that only one client at a time can use the GFX block.
+ * Returns: The dependency to wait on before the job can be pushed to the HW.
+ * The function is called multiple times until NULL is returned.
+ */
+struct dma_fence *amdgpu_device_enforce_isolation(struct amdgpu_device *adev,
+						  struct amdgpu_ring *ring,
+						  struct amdgpu_job *job)
+{
+	struct amdgpu_isolation *isolation = &adev->isolation[ring->xcp_id];
+	struct drm_sched_fence *f = job->base.s_fence;
+	struct dma_fence *dep;
+	void *owner;
+	int r;
+
+	/*
+	 * For now enforce isolation only for the GFX block since we only need
+	 * the cleaner shader on those rings.
+	 */
+	if (ring->funcs->type != AMDGPU_RING_TYPE_GFX &&
+	    ring->funcs->type != AMDGPU_RING_TYPE_COMPUTE)
+		return NULL;
+
+	/*
+	 * All submissions where enforce isolation is false are handled as if
+	 * they come from a single client. Use ~0l as the owner to distinct it
+	 * from kernel submissions where the owner is NULL.
+	 */
+	owner = job->enforce_isolation ? f->owner : (void *)~0l;
+
+	mutex_lock(&adev->enforce_isolation_mutex);
+
+	/*
+	 * The "spearhead" submission is the first one which changes the
+	 * ownership to its client. We always need to wait for it to be
+	 * pushed to the HW before proceeding with anything.
+	 */
+	if (&f->scheduled != isolation->spearhead &&
+	    !dma_fence_is_signaled(isolation->spearhead)) {
+		dep = isolation->spearhead;
+		goto out_grab_ref;
+	}
+
+	if (isolation->owner != owner) {
+
+		/*
+		 * Wait for any gang to be assembled before switching to a
+		 * different owner or otherwise we could deadlock the
+		 * submissions.
+		 */
+		if (!job->gang_submit) {
+			dep = amdgpu_device_get_gang(adev);
+			if (!dma_fence_is_signaled(dep))
+				goto out_return_dep;
+			dma_fence_put(dep);
+		}
+
+		dma_fence_put(isolation->spearhead);
+		isolation->spearhead = dma_fence_get(&f->scheduled);
+		amdgpu_sync_move(&isolation->active, &isolation->prev);
+		trace_amdgpu_isolation(isolation->owner, owner);
+		isolation->owner = owner;
+	}
+
+	/*
+	 * Specifying the ring here helps to pipeline submissions even when
+	 * isolation is enabled. If that is not desired for testing NULL can be
+	 * used instead of the ring to enforce a CPU round trip while switching
+	 * between clients.
+	 */
+	dep = amdgpu_sync_peek_fence(&isolation->prev, ring);
+	r = amdgpu_sync_fence(&isolation->active, &f->finished, GFP_NOWAIT);
+	if (r)
+		dev_warn(adev->dev, "OOM tracking isolation\n");
+
+out_grab_ref:
+	dma_fence_get(dep);
+out_return_dep:
+	mutex_unlock(&adev->enforce_isolation_mutex);
+	return dep;
+}
+
 bool amdgpu_device_has_display_hardware(struct amdgpu_device *adev)
 {
 	switch (adev->asic_type) {
@@ -6054,9 +7705,133 @@ bool amdgpu_device_has_display_hardware(struct amdgpu_device *adev)
 		return true;
 	default:
 		/* IP discovery */
-		if (!adev->ip_versions[DCE_HWIP][0] ||
+		if (!amdgpu_ip_version(adev, DCE_HWIP, 0) ||
 		    (adev->harvest_ip_mask & AMD_HARVEST_IP_DMU_MASK))
 			return false;
 		return true;
 	}
 }
+
+uint32_t amdgpu_device_wait_on_rreg(struct amdgpu_device *adev,
+		uint32_t inst, uint32_t reg_addr, char reg_name[],
+		uint32_t expected_value, uint32_t mask)
+{
+	uint32_t ret = 0;
+	uint32_t old_ = 0;
+	uint32_t tmp_ = RREG32(reg_addr);
+	uint32_t loop = adev->usec_timeout;
+
+	while ((tmp_ & (mask)) != (expected_value)) {
+		if (old_ != tmp_) {
+			loop = adev->usec_timeout;
+			old_ = tmp_;
+		} else
+			udelay(1);
+		tmp_ = RREG32(reg_addr);
+		loop--;
+		if (!loop) {
+			dev_warn(
+				adev->dev,
+				"Register(%d) [%s] failed to reach value 0x%08x != 0x%08xn",
+				inst, reg_name, (uint32_t)expected_value,
+				(uint32_t)(tmp_ & (mask)));
+			ret = -ETIMEDOUT;
+			break;
+		}
+	}
+	return ret;
+}
+
+ssize_t amdgpu_get_soft_full_reset_mask(struct amdgpu_ring *ring)
+{
+	ssize_t size = 0;
+
+	if (!ring || !ring->adev)
+		return size;
+
+	if (amdgpu_device_should_recover_gpu(ring->adev))
+		size |= AMDGPU_RESET_TYPE_FULL;
+
+	if (unlikely(!ring->adev->debug_disable_soft_recovery) &&
+	    !amdgpu_sriov_vf(ring->adev) && ring->funcs->soft_recovery)
+		size |= AMDGPU_RESET_TYPE_SOFT_RESET;
+
+	return size;
+}
+
+ssize_t amdgpu_show_reset_mask(char *buf, uint32_t supported_reset)
+{
+	ssize_t size = 0;
+
+	if (supported_reset == 0) {
+		size += sysfs_emit_at(buf, size, "unsupported");
+		size += sysfs_emit_at(buf, size, "\n");
+		return size;
+
+	}
+
+	if (supported_reset & AMDGPU_RESET_TYPE_SOFT_RESET)
+		size += sysfs_emit_at(buf, size, "soft ");
+
+	if (supported_reset & AMDGPU_RESET_TYPE_PER_QUEUE)
+		size += sysfs_emit_at(buf, size, "queue ");
+
+	if (supported_reset & AMDGPU_RESET_TYPE_PER_PIPE)
+		size += sysfs_emit_at(buf, size, "pipe ");
+
+	if (supported_reset & AMDGPU_RESET_TYPE_FULL)
+		size += sysfs_emit_at(buf, size, "full ");
+
+	size += sysfs_emit_at(buf, size, "\n");
+	return size;
+}
+
+void amdgpu_device_set_uid(struct amdgpu_uid *uid_info,
+			   enum amdgpu_uid_type type, uint8_t inst,
+			   uint64_t uid)
+{
+	if (!uid_info)
+		return;
+
+	if (type >= AMDGPU_UID_TYPE_MAX) {
+		dev_err_once(uid_info->adev->dev, "Invalid UID type %d\n",
+			     type);
+		return;
+	}
+
+	if (inst >= AMDGPU_UID_INST_MAX) {
+		dev_err_once(uid_info->adev->dev, "Invalid UID instance %d\n",
+			     inst);
+		return;
+	}
+
+	if (uid_info->uid[type][inst] != 0) {
+		dev_warn_once(
+			uid_info->adev->dev,
+			"Overwriting existing UID %llu for type %d instance %d\n",
+			uid_info->uid[type][inst], type, inst);
+	}
+
+	uid_info->uid[type][inst] = uid;
+}
+
+u64 amdgpu_device_get_uid(struct amdgpu_uid *uid_info,
+			  enum amdgpu_uid_type type, uint8_t inst)
+{
+	if (!uid_info)
+		return 0;
+
+	if (type >= AMDGPU_UID_TYPE_MAX) {
+		dev_err_once(uid_info->adev->dev, "Invalid UID type %d\n",
+			     type);
+		return 0;
+	}
+
+	if (inst >= AMDGPU_UID_INST_MAX) {
+		dev_err_once(uid_info->adev->dev, "Invalid UID instance %d\n",
+			     inst);
+		return 0;
+	}
+
+	return uid_info->uid[type][inst];
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_df.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_df.h
index 1538b2dbfff1..eb605e79ae0e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_df.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_df.h
@@ -33,6 +33,7 @@ struct amdgpu_df_hash_status {
 struct amdgpu_df_funcs {
 	void (*sw_init)(struct amdgpu_device *adev);
 	void (*sw_fini)(struct amdgpu_device *adev);
+	void (*hw_init)(struct amdgpu_device *adev);
 	void (*enable_broadcast_mode)(struct amdgpu_device *adev,
 				      bool enable);
 	u32 (*get_fb_channel_number)(struct amdgpu_device *adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c
index 0ecce0b92b82..fa2a22dfa048 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2018 Advanced Micro Devices, Inc.
+ * Copyright 2018-2024 Advanced Micro Devices, Inc. All rights reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -27,17 +27,22 @@
 #include "amdgpu_discovery.h"
 #include "soc15_hw_ip.h"
 #include "discovery.h"
+#include "amdgpu_ras.h"
 
 #include "soc15.h"
 #include "gfx_v9_0.h"
+#include "gfx_v9_4_3.h"
 #include "gmc_v9_0.h"
 #include "df_v1_7.h"
 #include "df_v3_6.h"
 #include "df_v4_3.h"
+#include "df_v4_6_2.h"
+#include "df_v4_15.h"
 #include "nbio_v6_1.h"
 #include "nbio_v7_0.h"
 #include "nbio_v7_4.h"
 #include "nbio_v7_9.h"
+#include "nbio_v7_11.h"
 #include "hdp_v4_0.h"
 #include "vega10_ih.h"
 #include "vega20_ih.h"
@@ -51,43 +56,77 @@
 #include "smuio_v9_0.h"
 #include "gmc_v10_0.h"
 #include "gmc_v11_0.h"
+#include "gmc_v12_0.h"
 #include "gfxhub_v2_0.h"
 #include "mmhub_v2_0.h"
 #include "nbio_v2_3.h"
 #include "nbio_v4_3.h"
 #include "nbio_v7_2.h"
 #include "nbio_v7_7.h"
+#include "nbif_v6_3_1.h"
 #include "hdp_v5_0.h"
 #include "hdp_v5_2.h"
 #include "hdp_v6_0.h"
+#include "hdp_v7_0.h"
 #include "nv.h"
 #include "soc21.h"
+#include "soc24.h"
 #include "navi10_ih.h"
 #include "ih_v6_0.h"
+#include "ih_v6_1.h"
+#include "ih_v7_0.h"
 #include "gfx_v10_0.h"
 #include "gfx_v11_0.h"
+#include "gfx_v12_0.h"
 #include "sdma_v5_0.h"
 #include "sdma_v5_2.h"
 #include "sdma_v6_0.h"
+#include "sdma_v7_0.h"
 #include "lsdma_v6_0.h"
+#include "lsdma_v7_0.h"
 #include "vcn_v2_0.h"
 #include "jpeg_v2_0.h"
 #include "vcn_v3_0.h"
 #include "jpeg_v3_0.h"
 #include "vcn_v4_0.h"
 #include "jpeg_v4_0.h"
+#include "vcn_v4_0_3.h"
+#include "jpeg_v4_0_3.h"
+#include "vcn_v4_0_5.h"
+#include "jpeg_v4_0_5.h"
 #include "amdgpu_vkms.h"
-#include "mes_v10_1.h"
 #include "mes_v11_0.h"
+#include "mes_v12_0.h"
 #include "smuio_v11_0.h"
 #include "smuio_v11_0_6.h"
 #include "smuio_v13_0.h"
+#include "smuio_v13_0_3.h"
 #include "smuio_v13_0_6.h"
+#include "smuio_v14_0_2.h"
+#include "vcn_v5_0_0.h"
+#include "vcn_v5_0_1.h"
+#include "jpeg_v5_0_0.h"
+#include "jpeg_v5_0_1.h"
+#include "amdgpu_ras_mgr.h"
+
+#include "amdgpu_vpe.h"
+#if defined(CONFIG_DRM_AMD_ISP)
+#include "amdgpu_isp.h"
+#endif
 
-#define FIRMWARE_IP_DISCOVERY "amdgpu/ip_discovery.bin"
-MODULE_FIRMWARE(FIRMWARE_IP_DISCOVERY);
-
+MODULE_FIRMWARE("amdgpu/ip_discovery.bin");
+MODULE_FIRMWARE("amdgpu/vega10_ip_discovery.bin");
+MODULE_FIRMWARE("amdgpu/vega12_ip_discovery.bin");
+MODULE_FIRMWARE("amdgpu/vega20_ip_discovery.bin");
+MODULE_FIRMWARE("amdgpu/raven_ip_discovery.bin");
+MODULE_FIRMWARE("amdgpu/raven2_ip_discovery.bin");
+MODULE_FIRMWARE("amdgpu/picasso_ip_discovery.bin");
+MODULE_FIRMWARE("amdgpu/arcturus_ip_discovery.bin");
+MODULE_FIRMWARE("amdgpu/aldebaran_ip_discovery.bin");
+
+#define mmIP_DISCOVERY_VERSION  0x16A00
 #define mmRCC_CONFIG_MEMSIZE	0xde3
+#define mmMP0_SMN_C2PMSG_33	0x16061
 #define mmMM_INDEX		0x0
 #define mmMM_INDEX_HI		0x6
 #define mmMM_DATA		0x1
@@ -169,6 +208,7 @@ static const char *hw_id_names[HW_ID_MAX] = {
 	[XGMI_HWID]		= "XGMI",
 	[XGBE_HWID]		= "XGBE",
 	[MP0_HWID]		= "MP0",
+	[VPE_HWID]		= "VPE",
 };
 
 static int hw_id_map[MAX_HWIP] = {
@@ -198,37 +238,117 @@ static int hw_id_map[MAX_HWIP] = {
 	[XGMI_HWIP]	= XGMI_HWID,
 	[DCI_HWIP]	= DCI_HWID,
 	[PCIE_HWIP]	= PCIE_HWID,
+	[VPE_HWIP]	= VPE_HWID,
+	[ISP_HWIP]	= ISP_HWID,
 };
 
-static int amdgpu_discovery_read_binary_from_vram(struct amdgpu_device *adev, uint8_t *binary)
+static int amdgpu_discovery_read_binary_from_sysmem(struct amdgpu_device *adev, uint8_t *binary)
 {
-	uint64_t vram_size = (uint64_t)RREG32(mmRCC_CONFIG_MEMSIZE) << 20;
-	uint64_t pos = vram_size - DISCOVERY_TMR_OFFSET;
+	u64 tmr_offset, tmr_size, pos;
+	void *discv_regn;
+	int ret;
 
-	amdgpu_device_vram_access(adev, pos, (uint32_t *)binary,
-				  adev->mman.discovery_tmr_size, false);
-	return 0;
+	ret = amdgpu_acpi_get_tmr_info(adev, &tmr_offset, &tmr_size);
+	if (ret)
+		return ret;
+
+	pos = tmr_offset + tmr_size - DISCOVERY_TMR_OFFSET;
+
+	/* This region is read-only and reserved from system use */
+	discv_regn = memremap(pos, adev->discovery.size, MEMREMAP_WC);
+	if (discv_regn) {
+		memcpy(binary, discv_regn, adev->discovery.size);
+		memunmap(discv_regn);
+		return 0;
+	}
+
+	return -ENOENT;
 }
 
-static int amdgpu_discovery_read_binary_from_file(struct amdgpu_device *adev, uint8_t *binary)
+#define IP_DISCOVERY_V2		2
+#define IP_DISCOVERY_V4		4
+
+static int amdgpu_discovery_read_binary_from_mem(struct amdgpu_device *adev,
+						 uint8_t *binary)
 {
-	const struct firmware *fw;
-	const char *fw_name;
-	int r;
+	bool sz_valid = true;
+	uint64_t vram_size;
+	int i, ret = 0;
+	u32 msg;
+
+	if (!amdgpu_sriov_vf(adev)) {
+		/* It can take up to two second for IFWI init to complete on some dGPUs,
+		 * but generally it should be in the 60-100ms range.  Normally this starts
+		 * as soon as the device gets power so by the time the OS loads this has long
+		 * completed.  However, when a card is hotplugged via e.g., USB4, we need to
+		 * wait for this to complete.  Once the C2PMSG is updated, we can
+		 * continue.
+		 */
 
-	switch (amdgpu_discovery) {
-	case 2:
-		fw_name = FIRMWARE_IP_DISCOVERY;
-		break;
-	default:
-		dev_warn(adev->dev, "amdgpu_discovery is not set properly\n");
-		return -EINVAL;
+		for (i = 0; i < 2000; i++) {
+			msg = RREG32(mmMP0_SMN_C2PMSG_33);
+			if (msg & 0x80000000)
+				break;
+			msleep(1);
+		}
+	}
+
+	vram_size = RREG32(mmRCC_CONFIG_MEMSIZE);
+	if (!vram_size || vram_size == U32_MAX)
+		sz_valid = false;
+	else
+		vram_size <<= 20;
+
+	/*
+	 * If in VRAM, discovery TMR is marked for reservation. If it is in system mem,
+	 * then it is not required to be reserved.
+	 */
+	if (sz_valid) {
+		if (amdgpu_sriov_vf(adev) && adev->virt.is_dynamic_crit_regn_enabled) {
+			/* For SRIOV VFs with dynamic critical region enabled,
+			 * we will get the IPD binary via below call.
+			 * If dynamic critical is disabled, fall through to normal seq.
+			 */
+			if (amdgpu_virt_get_dynamic_data_info(adev,
+						AMD_SRIOV_MSG_IPD_TABLE_ID, binary,
+						&adev->discovery.size)) {
+				dev_err(adev->dev,
+						"failed to read discovery info from dynamic critical region.");
+				ret = -EINVAL;
+				goto exit;
+			}
+		} else {
+			uint64_t pos = vram_size - DISCOVERY_TMR_OFFSET;
+
+			amdgpu_device_vram_access(adev, pos, (uint32_t *)binary,
+					adev->discovery.size, false);
+			adev->discovery.reserve_tmr = true;
+		}
+	} else {
+		ret = amdgpu_discovery_read_binary_from_sysmem(adev, binary);
 	}
 
-	r = request_firmware(&fw, fw_name, adev->dev);
+	if (ret)
+		dev_err(adev->dev,
+			"failed to read discovery info from memory, vram size read: %llx",
+			vram_size);
+exit:
+	return ret;
+}
+
+static int amdgpu_discovery_read_binary_from_file(struct amdgpu_device *adev,
+							uint8_t *binary,
+							const char *fw_name)
+{
+	const struct firmware *fw;
+	int r;
+
+	r = firmware_request_nowarn(&fw, fw_name, adev->dev);
 	if (r) {
-		dev_err(adev->dev, "can't load firmware \"%s\"\n",
-			fw_name);
+		if (amdgpu_discovery == 2)
+			dev_err(adev->dev, "can't load firmware \"%s\"\n", fw_name);
+		else
+			drm_info(&adev->ddev, "Optional firmware \"%s\" was not found\n", fw_name);
 		return r;
 	}
 
@@ -269,8 +389,8 @@ static void amdgpu_discovery_harvest_config_quirk(struct amdgpu_device *adev)
 	 * So far, apply this quirk only on those Navy Flounder boards which
 	 * have a bad harvest table of VCN config.
 	 */
-	if ((adev->ip_versions[UVD_HWIP][1] == IP_VERSION(3, 0, 1)) &&
-		(adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 2))) {
+	if ((amdgpu_ip_version(adev, UVD_HWIP, 1) == IP_VERSION(3, 0, 1)) &&
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 3, 2))) {
 		switch (adev->pdev->revision) {
 		case 0xC1:
 		case 0xC2:
@@ -280,6 +400,7 @@ static void amdgpu_discovery_harvest_config_quirk(struct amdgpu_device *adev)
 		case 0xCF:
 		case 0xDF:
 			adev->vcn.harvest_config |= AMDGPU_VCN_HARVEST_VCN1;
+			adev->vcn.inst_mask &= ~AMDGPU_VCN_HARVEST_VCN1;
 			break;
 		default:
 			break;
@@ -287,58 +408,118 @@ static void amdgpu_discovery_harvest_config_quirk(struct amdgpu_device *adev)
 	}
 }
 
+static int amdgpu_discovery_verify_npsinfo(struct amdgpu_device *adev,
+					   struct binary_header *bhdr)
+{
+	uint8_t *discovery_bin = adev->discovery.bin;
+	struct table_info *info;
+	uint16_t checksum;
+	uint16_t offset;
+
+	info = &bhdr->table_list[NPS_INFO];
+	offset = le16_to_cpu(info->offset);
+	checksum = le16_to_cpu(info->checksum);
+
+	struct nps_info_header *nhdr =
+		(struct nps_info_header *)(discovery_bin + offset);
+
+	if (le32_to_cpu(nhdr->table_id) != NPS_INFO_TABLE_ID) {
+		dev_dbg(adev->dev, "invalid ip discovery nps info table id\n");
+		return -EINVAL;
+	}
+
+	if (!amdgpu_discovery_verify_checksum(discovery_bin + offset,
+					      le32_to_cpu(nhdr->size_bytes),
+					      checksum)) {
+		dev_dbg(adev->dev, "invalid nps info data table checksum\n");
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static const char *amdgpu_discovery_get_fw_name(struct amdgpu_device *adev)
+{
+	if (amdgpu_discovery == 2) {
+		/* Assume there is valid discovery TMR in VRAM even if binary is sideloaded */
+		adev->discovery.reserve_tmr = true;
+		return "amdgpu/ip_discovery.bin";
+	}
+
+	switch (adev->asic_type) {
+	case CHIP_VEGA10:
+		return "amdgpu/vega10_ip_discovery.bin";
+	case CHIP_VEGA12:
+		return "amdgpu/vega12_ip_discovery.bin";
+	case CHIP_RAVEN:
+		if (adev->apu_flags & AMD_APU_IS_RAVEN2)
+			return "amdgpu/raven2_ip_discovery.bin";
+		else if (adev->apu_flags & AMD_APU_IS_PICASSO)
+			return "amdgpu/picasso_ip_discovery.bin";
+		else
+			return "amdgpu/raven_ip_discovery.bin";
+	case CHIP_VEGA20:
+		return "amdgpu/vega20_ip_discovery.bin";
+	case CHIP_ARCTURUS:
+		return "amdgpu/arcturus_ip_discovery.bin";
+	case CHIP_ALDEBARAN:
+		return "amdgpu/aldebaran_ip_discovery.bin";
+	default:
+		return NULL;
+	}
+}
+
 static int amdgpu_discovery_init(struct amdgpu_device *adev)
 {
 	struct table_info *info;
 	struct binary_header *bhdr;
+	uint8_t *discovery_bin;
+	const char *fw_name;
 	uint16_t offset;
 	uint16_t size;
 	uint16_t checksum;
 	int r;
 
-	adev->mman.discovery_tmr_size = DISCOVERY_TMR_SIZE;
-	adev->mman.discovery_bin = kzalloc(adev->mman.discovery_tmr_size, GFP_KERNEL);
-	if (!adev->mman.discovery_bin)
+	adev->discovery.bin = kzalloc(DISCOVERY_TMR_SIZE, GFP_KERNEL);
+	if (!adev->discovery.bin)
 		return -ENOMEM;
+	adev->discovery.size = DISCOVERY_TMR_SIZE;
+	adev->discovery.debugfs_blob.data = adev->discovery.bin;
+	adev->discovery.debugfs_blob.size = adev->discovery.size;
+
+	discovery_bin = adev->discovery.bin;
+	/* Read from file if it is the preferred option */
+	fw_name = amdgpu_discovery_get_fw_name(adev);
+	if (fw_name != NULL) {
+		drm_dbg(&adev->ddev, "use ip discovery information from file");
+		r = amdgpu_discovery_read_binary_from_file(adev, discovery_bin,
+							   fw_name);
+		if (r)
+			goto out;
+	} else {
+		drm_dbg(&adev->ddev, "use ip discovery information from memory");
+		r = amdgpu_discovery_read_binary_from_mem(adev, discovery_bin);
+		if (r)
+			goto out;
+	}
 
-	r = amdgpu_discovery_read_binary_from_vram(adev, adev->mman.discovery_bin);
-	if (r) {
-		dev_err(adev->dev, "failed to read ip discovery binary from vram\n");
+	/* check the ip discovery binary signature */
+	if (!amdgpu_discovery_verify_binary_signature(discovery_bin)) {
+		dev_err(adev->dev,
+			"get invalid ip discovery binary signature\n");
 		r = -EINVAL;
 		goto out;
 	}
 
-	if (!amdgpu_discovery_verify_binary_signature(adev->mman.discovery_bin) || amdgpu_discovery == 2) {
-		/* ignore the discovery binary from vram if discovery=2 in kernel module parameter */
-		if (amdgpu_discovery == 2)
-			dev_info(adev->dev,"force read ip discovery binary from file");
-		else
-			dev_warn(adev->dev, "get invalid ip discovery binary signature from vram\n");
-
-		/* retry read ip discovery binary from file */
-		r = amdgpu_discovery_read_binary_from_file(adev, adev->mman.discovery_bin);
-		if (r) {
-			dev_err(adev->dev, "failed to read ip discovery binary from file\n");
-			r = -EINVAL;
-			goto out;
-		}
-		/* check the ip discovery binary signature */
-		if(!amdgpu_discovery_verify_binary_signature(adev->mman.discovery_bin)) {
-			dev_warn(adev->dev, "get invalid ip discovery binary signature from file\n");
-			r = -EINVAL;
-			goto out;
-		}
-	}
-
-	bhdr = (struct binary_header *)adev->mman.discovery_bin;
+	bhdr = (struct binary_header *)discovery_bin;
 
 	offset = offsetof(struct binary_header, binary_checksum) +
 		sizeof(bhdr->binary_checksum);
 	size = le16_to_cpu(bhdr->binary_size) - offset;
 	checksum = le16_to_cpu(bhdr->binary_checksum);
 
-	if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset,
-					      size, checksum)) {
+	if (!amdgpu_discovery_verify_checksum(discovery_bin + offset, size,
+					      checksum)) {
 		dev_err(adev->dev, "invalid ip discovery binary checksum\n");
 		r = -EINVAL;
 		goto out;
@@ -350,15 +531,16 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 
 	if (offset) {
 		struct ip_discovery_header *ihdr =
-			(struct ip_discovery_header *)(adev->mman.discovery_bin + offset);
+			(struct ip_discovery_header *)(discovery_bin + offset);
 		if (le32_to_cpu(ihdr->signature) != DISCOVERY_TABLE_SIGNATURE) {
 			dev_err(adev->dev, "invalid ip discovery data table signature\n");
 			r = -EINVAL;
 			goto out;
 		}
 
-		if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset,
-						      le16_to_cpu(ihdr->size), checksum)) {
+		if (!amdgpu_discovery_verify_checksum(discovery_bin + offset,
+						      le16_to_cpu(ihdr->size),
+						      checksum)) {
 			dev_err(adev->dev, "invalid ip discovery data table checksum\n");
 			r = -EINVAL;
 			goto out;
@@ -371,7 +553,7 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 
 	if (offset) {
 		struct gpu_info_header *ghdr =
-			(struct gpu_info_header *)(adev->mman.discovery_bin + offset);
+			(struct gpu_info_header *)(discovery_bin + offset);
 
 		if (le32_to_cpu(ghdr->table_id) != GC_TABLE_ID) {
 			dev_err(adev->dev, "invalid ip discovery gc table id\n");
@@ -379,8 +561,9 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 			goto out;
 		}
 
-		if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset,
-						      le32_to_cpu(ghdr->size), checksum)) {
+		if (!amdgpu_discovery_verify_checksum(discovery_bin + offset,
+						      le32_to_cpu(ghdr->size),
+						      checksum)) {
 			dev_err(adev->dev, "invalid gc data table checksum\n");
 			r = -EINVAL;
 			goto out;
@@ -393,7 +576,7 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 
 	if (offset) {
 		struct harvest_info_header *hhdr =
-			(struct harvest_info_header *)(adev->mman.discovery_bin + offset);
+			(struct harvest_info_header *)(discovery_bin + offset);
 
 		if (le32_to_cpu(hhdr->signature) != HARVEST_TABLE_SIGNATURE) {
 			dev_err(adev->dev, "invalid ip discovery harvest table signature\n");
@@ -401,8 +584,9 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 			goto out;
 		}
 
-		if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset,
-						      sizeof(struct harvest_table), checksum)) {
+		if (!amdgpu_discovery_verify_checksum(
+			    discovery_bin + offset,
+			    sizeof(struct harvest_table), checksum)) {
 			dev_err(adev->dev, "invalid harvest data table checksum\n");
 			r = -EINVAL;
 			goto out;
@@ -415,7 +599,7 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 
 	if (offset) {
 		struct vcn_info_header *vhdr =
-			(struct vcn_info_header *)(adev->mman.discovery_bin + offset);
+			(struct vcn_info_header *)(discovery_bin + offset);
 
 		if (le32_to_cpu(vhdr->table_id) != VCN_INFO_TABLE_ID) {
 			dev_err(adev->dev, "invalid ip discovery vcn table id\n");
@@ -423,8 +607,9 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 			goto out;
 		}
 
-		if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset,
-						      le32_to_cpu(vhdr->size_bytes), checksum)) {
+		if (!amdgpu_discovery_verify_checksum(
+			    discovery_bin + offset,
+			    le32_to_cpu(vhdr->size_bytes), checksum)) {
 			dev_err(adev->dev, "invalid vcn data table checksum\n");
 			r = -EINVAL;
 			goto out;
@@ -437,7 +622,7 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 
 	if (0 && offset) {
 		struct mall_info_header *mhdr =
-			(struct mall_info_header *)(adev->mman.discovery_bin + offset);
+			(struct mall_info_header *)(discovery_bin + offset);
 
 		if (le32_to_cpu(mhdr->table_id) != MALL_INFO_TABLE_ID) {
 			dev_err(adev->dev, "invalid ip discovery mall table id\n");
@@ -445,8 +630,9 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 			goto out;
 		}
 
-		if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset,
-						      le32_to_cpu(mhdr->size_bytes), checksum)) {
+		if (!amdgpu_discovery_verify_checksum(
+			    discovery_bin + offset,
+			    le32_to_cpu(mhdr->size_bytes), checksum)) {
 			dev_err(adev->dev, "invalid mall data table checksum\n");
 			r = -EINVAL;
 			goto out;
@@ -456,9 +642,11 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 	return 0;
 
 out:
-	kfree(adev->mman.discovery_bin);
-	adev->mman.discovery_bin = NULL;
-
+	kfree(adev->discovery.bin);
+	adev->discovery.bin = NULL;
+	if ((amdgpu_discovery != 2) &&
+	    (RREG32(mmIP_DISCOVERY_VERSION) == 4))
+		amdgpu_ras_query_boot_status(adev, 4);
 	return r;
 }
 
@@ -467,20 +655,23 @@ static void amdgpu_discovery_sysfs_fini(struct amdgpu_device *adev);
 void amdgpu_discovery_fini(struct amdgpu_device *adev)
 {
 	amdgpu_discovery_sysfs_fini(adev);
-	kfree(adev->mman.discovery_bin);
-	adev->mman.discovery_bin = NULL;
+	kfree(adev->discovery.bin);
+	adev->discovery.bin = NULL;
 }
 
-static int amdgpu_discovery_validate_ip(const struct ip *ip)
+static int amdgpu_discovery_validate_ip(struct amdgpu_device *adev,
+					uint8_t instance, uint16_t hw_id)
 {
-	if (ip->number_instance >= HWIP_MAX_INSTANCE) {
-		DRM_ERROR("Unexpected number_instance (%d) from ip discovery blob\n",
-			  ip->number_instance);
+	if (instance >= HWIP_MAX_INSTANCE) {
+		dev_err(adev->dev,
+			"Unexpected instance_number (%d) from ip discovery blob\n",
+			instance);
 		return -EINVAL;
 	}
-	if (le16_to_cpu(ip->hw_id) >= HW_ID_MAX) {
-		DRM_ERROR("Unexpected hw_id (%d) from ip discovery blob\n",
-			  le16_to_cpu(ip->hw_id));
+	if (hw_id >= HW_ID_MAX) {
+		dev_err(adev->dev,
+			"Unexpected hw_id (%d) from ip discovery blob\n",
+			hw_id);
 		return -EINVAL;
 	}
 
@@ -490,49 +681,64 @@ static int amdgpu_discovery_validate_ip(const struct ip *ip)
 static void amdgpu_discovery_read_harvest_bit_per_ip(struct amdgpu_device *adev,
 						uint32_t *vcn_harvest_count)
 {
+	uint8_t *discovery_bin = adev->discovery.bin;
 	struct binary_header *bhdr;
 	struct ip_discovery_header *ihdr;
 	struct die_header *dhdr;
 	struct ip *ip;
 	uint16_t die_offset, ip_offset, num_dies, num_ips;
+	uint16_t hw_id;
+	uint8_t inst;
 	int i, j;
 
-	bhdr = (struct binary_header *)adev->mman.discovery_bin;
-	ihdr = (struct ip_discovery_header *)(adev->mman.discovery_bin +
-			le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset));
+	bhdr = (struct binary_header *)discovery_bin;
+	ihdr = (struct ip_discovery_header
+			*)(discovery_bin +
+			   le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset));
 	num_dies = le16_to_cpu(ihdr->num_dies);
 
 	/* scan harvest bit of all IP data structures */
 	for (i = 0; i < num_dies; i++) {
 		die_offset = le16_to_cpu(ihdr->die_info[i].die_offset);
-		dhdr = (struct die_header *)(adev->mman.discovery_bin + die_offset);
+		dhdr = (struct die_header *)(discovery_bin + die_offset);
 		num_ips = le16_to_cpu(dhdr->num_ips);
 		ip_offset = die_offset + sizeof(*dhdr);
 
 		for (j = 0; j < num_ips; j++) {
-			ip = (struct ip *)(adev->mman.discovery_bin + ip_offset);
-
-			if (amdgpu_discovery_validate_ip(ip))
+			ip = (struct ip *)(discovery_bin + ip_offset);
+			inst = ip->number_instance;
+			hw_id = le16_to_cpu(ip->hw_id);
+			if (amdgpu_discovery_validate_ip(adev, inst, hw_id))
 				goto next_ip;
 
-			if (le16_to_cpu(ip->harvest) == 1) {
-				switch (le16_to_cpu(ip->hw_id)) {
+			if (ip->harvest == 1) {
+				switch (hw_id) {
 				case VCN_HWID:
 					(*vcn_harvest_count)++;
-					if (ip->number_instance == 0)
+					if (inst == 0) {
 						adev->vcn.harvest_config |= AMDGPU_VCN_HARVEST_VCN0;
-					else
+						adev->vcn.inst_mask &=
+							~AMDGPU_VCN_HARVEST_VCN0;
+						adev->jpeg.inst_mask &=
+							~AMDGPU_VCN_HARVEST_VCN0;
+					} else {
 						adev->vcn.harvest_config |= AMDGPU_VCN_HARVEST_VCN1;
+						adev->vcn.inst_mask &=
+							~AMDGPU_VCN_HARVEST_VCN1;
+						adev->jpeg.inst_mask &=
+							~AMDGPU_VCN_HARVEST_VCN1;
+					}
 					break;
 				case DMU_HWID:
 					adev->harvest_ip_mask |= AMD_HARVEST_IP_DMU_MASK;
 					break;
 				default:
 					break;
-                                }
-                        }
+				}
+			}
 next_ip:
-			ip_offset += struct_size(ip, base_address, ip->num_base_address);
+			ip_offset += struct_size(ip, base_address,
+						 ip->num_base_address);
 		}
 	}
 }
@@ -541,13 +747,14 @@ static void amdgpu_discovery_read_from_harvest_table(struct amdgpu_device *adev,
 						     uint32_t *vcn_harvest_count,
 						     uint32_t *umc_harvest_count)
 {
+	uint8_t *discovery_bin = adev->discovery.bin;
 	struct binary_header *bhdr;
 	struct harvest_table *harvest_info;
 	u16 offset;
 	int i;
 	uint32_t umc_harvest_config = 0;
 
-	bhdr = (struct binary_header *)adev->mman.discovery_bin;
+	bhdr = (struct binary_header *)discovery_bin;
 	offset = le16_to_cpu(bhdr->table_list[HARVEST_INFO].offset);
 
 	if (!offset) {
@@ -555,7 +762,7 @@ static void amdgpu_discovery_read_from_harvest_table(struct amdgpu_device *adev,
 		return;
 	}
 
-	harvest_info = (struct harvest_table *)(adev->mman.discovery_bin + offset);
+	harvest_info = (struct harvest_table *)(discovery_bin + offset);
 
 	for (i = 0; i < 32; i++) {
 		if (le16_to_cpu(harvest_info->list[i].hw_id) == 0)
@@ -564,10 +771,15 @@ static void amdgpu_discovery_read_from_harvest_table(struct amdgpu_device *adev,
 		switch (le16_to_cpu(harvest_info->list[i].hw_id)) {
 		case VCN_HWID:
 			(*vcn_harvest_count)++;
-			if (harvest_info->list[i].number_instance == 0)
-				adev->vcn.harvest_config |= AMDGPU_VCN_HARVEST_VCN0;
-			else
-				adev->vcn.harvest_config |= AMDGPU_VCN_HARVEST_VCN1;
+			adev->vcn.harvest_config |=
+				(1 << harvest_info->list[i].number_instance);
+			adev->jpeg.harvest_config |=
+				(1 << harvest_info->list[i].number_instance);
+
+			adev->vcn.inst_mask &=
+				~(1U << harvest_info->list[i].number_instance);
+			adev->jpeg.inst_mask &=
+				~(1U << harvest_info->list[i].number_instance);
 			break;
 		case DMU_HWID:
 			adev->harvest_ip_mask |= AMD_HARVEST_IP_DMU_MASK;
@@ -577,6 +789,20 @@ static void amdgpu_discovery_read_from_harvest_table(struct amdgpu_device *adev,
 				1 << (le16_to_cpu(harvest_info->list[i].number_instance));
 			(*umc_harvest_count)++;
 			break;
+		case GC_HWID:
+			adev->gfx.xcc_mask &=
+				~(1U << harvest_info->list[i].number_instance);
+			break;
+		case SDMA0_HWID:
+			adev->sdma.sdma_mask &=
+				~(1U << harvest_info->list[i].number_instance);
+			break;
+#if defined(CONFIG_DRM_AMD_ISP)
+		case ISP_HWID:
+			adev->isp.harvest_config |=
+				~(1U << harvest_info->list[i].number_instance);
+			break;
+#endif
 		default:
 			break;
 		}
@@ -597,7 +823,7 @@ struct ip_hw_instance {
 	u8  harvest;
 
 	int num_base_addresses;
-	u32 base_addr[];
+	u32 base_addr[] __counted_by(num_base_addresses);
 };
 
 struct ip_hw_id {
@@ -832,15 +1058,51 @@ static void ip_disc_release(struct kobject *kobj)
 						       kobj);
 	struct amdgpu_device *adev = ip_top->adev;
 
-	adev->ip_top = NULL;
 	kfree(ip_top);
+	adev->discovery.ip_top = NULL;
+}
+
+static uint8_t amdgpu_discovery_get_harvest_info(struct amdgpu_device *adev,
+						 uint16_t hw_id, uint8_t inst)
+{
+	uint8_t harvest = 0;
+
+	/* Until a uniform way is figured, get mask based on hwid */
+	switch (hw_id) {
+	case VCN_HWID:
+		/* VCN vs UVD+VCE */
+		if (!amdgpu_ip_version(adev, VCE_HWIP, 0))
+			harvest = ((1 << inst) & adev->vcn.inst_mask) == 0;
+		break;
+	case DMU_HWID:
+		if (adev->harvest_ip_mask & AMD_HARVEST_IP_DMU_MASK)
+			harvest = 0x1;
+		break;
+	case UMC_HWID:
+		/* TODO: It needs another parsing; for now, ignore.*/
+		break;
+	case GC_HWID:
+		harvest = ((1 << inst) & adev->gfx.xcc_mask) == 0;
+		break;
+	case SDMA0_HWID:
+		harvest = ((1 << inst) & adev->sdma.sdma_mask) == 0;
+		break;
+	default:
+		break;
+	}
+
+	return harvest;
 }
 
 static int amdgpu_discovery_sysfs_ips(struct amdgpu_device *adev,
 				      struct ip_die_entry *ip_die_entry,
-				      const size_t _ip_offset, const int num_ips)
+				      const size_t _ip_offset, const int num_ips,
+				      bool reg_base_64)
 {
+	uint8_t *discovery_bin = adev->discovery.bin;
 	int ii, jj, kk, res;
+	uint16_t hw_id;
+	uint8_t inst;
 
 	DRM_DEBUG("num_ips:%d", num_ips);
 
@@ -852,12 +1114,14 @@ static int amdgpu_discovery_sysfs_ips(struct amdgpu_device *adev,
 		size_t ip_offset = _ip_offset;
 
 		for (jj = 0; jj < num_ips; jj++) {
-			struct ip *ip;
+			struct ip_v4 *ip;
 			struct ip_hw_instance *ip_hw_instance;
 
-			ip = (struct ip *)(adev->mman.discovery_bin + ip_offset);
-			if (amdgpu_discovery_validate_ip(ip) ||
-			    le16_to_cpu(ip->hw_id) != ii)
+			ip = (struct ip_v4 *)(discovery_bin + ip_offset);
+			inst = ip->instance_number;
+			hw_id = le16_to_cpu(ip->hw_id);
+			if (amdgpu_discovery_validate_ip(adev, inst, hw_id) ||
+			    hw_id != ii)
 				goto next_ip;
 
 			DRM_DEBUG("match:%d @ ip_offset:%zu", ii, ip_offset);
@@ -903,22 +1167,35 @@ static int amdgpu_discovery_sysfs_ips(struct amdgpu_device *adev,
 				return -ENOMEM;
 			}
 			ip_hw_instance->hw_id = le16_to_cpu(ip->hw_id); /* == ii */
-			ip_hw_instance->num_instance = ip->number_instance;
+			ip_hw_instance->num_instance = ip->instance_number;
 			ip_hw_instance->major = ip->major;
 			ip_hw_instance->minor = ip->minor;
 			ip_hw_instance->revision = ip->revision;
-			ip_hw_instance->harvest = ip->harvest;
+			ip_hw_instance->harvest =
+				amdgpu_discovery_get_harvest_info(
+					adev, ip_hw_instance->hw_id,
+					ip_hw_instance->num_instance);
 			ip_hw_instance->num_base_addresses = ip->num_base_address;
 
-			for (kk = 0; kk < ip_hw_instance->num_base_addresses; kk++)
-				ip_hw_instance->base_addr[kk] = ip->base_address[kk];
+			for (kk = 0; kk < ip_hw_instance->num_base_addresses; kk++) {
+				if (reg_base_64)
+					ip_hw_instance->base_addr[kk] =
+						lower_32_bits(le64_to_cpu(ip->base_address_64[kk])) & 0x3FFFFFFF;
+				else
+					ip_hw_instance->base_addr[kk] = ip->base_address[kk];
+			}
 
 			kobject_init(&ip_hw_instance->kobj, &ip_hw_instance_ktype);
 			ip_hw_instance->kobj.kset = &ip_hw_id->hw_id_kset;
 			res = kobject_add(&ip_hw_instance->kobj, NULL,
 					  "%d", ip_hw_instance->num_instance);
 next_ip:
-			ip_offset += struct_size(ip, base_address, ip->num_base_address);
+			if (reg_base_64)
+				ip_offset += struct_size(ip, base_address_64,
+							 ip->num_base_address);
+			else
+				ip_offset += struct_size(ip, base_address,
+							 ip->num_base_address);
 		}
 	}
 
@@ -927,17 +1204,20 @@ next_ip:
 
 static int amdgpu_discovery_sysfs_recurse(struct amdgpu_device *adev)
 {
+	struct ip_discovery_top *ip_top = adev->discovery.ip_top;
+	uint8_t *discovery_bin = adev->discovery.bin;
 	struct binary_header *bhdr;
 	struct ip_discovery_header *ihdr;
 	struct die_header *dhdr;
-	struct kset *die_kset = &adev->ip_top->die_kset;
+	struct kset *die_kset = &ip_top->die_kset;
 	u16 num_dies, die_offset, num_ips;
 	size_t ip_offset;
 	int ii, res;
 
-	bhdr = (struct binary_header *)adev->mman.discovery_bin;
-	ihdr = (struct ip_discovery_header *)(adev->mman.discovery_bin +
-					      le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset));
+	bhdr = (struct binary_header *)discovery_bin;
+	ihdr = (struct ip_discovery_header
+			*)(discovery_bin +
+			   le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset));
 	num_dies = le16_to_cpu(ihdr->num_dies);
 
 	DRM_DEBUG("number of dies: %d\n", num_dies);
@@ -946,7 +1226,7 @@ static int amdgpu_discovery_sysfs_recurse(struct amdgpu_device *adev)
 		struct ip_die_entry *ip_die_entry;
 
 		die_offset = le16_to_cpu(ihdr->die_info[ii].die_offset);
-		dhdr = (struct die_header *)(adev->mman.discovery_bin + die_offset);
+		dhdr = (struct die_header *)(discovery_bin + die_offset);
 		num_ips = le16_to_cpu(dhdr->num_ips);
 		ip_offset = die_offset + sizeof(*dhdr);
 
@@ -972,7 +1252,7 @@ static int amdgpu_discovery_sysfs_recurse(struct amdgpu_device *adev)
 			return res;
 		}
 
-		amdgpu_discovery_sysfs_ips(adev, ip_die_entry, ip_offset, num_ips);
+		amdgpu_discovery_sysfs_ips(adev, ip_die_entry, ip_offset, num_ips, !!ihdr->base_addr_64_bit);
 	}
 
 	return 0;
@@ -980,27 +1260,32 @@ static int amdgpu_discovery_sysfs_recurse(struct amdgpu_device *adev)
 
 static int amdgpu_discovery_sysfs_init(struct amdgpu_device *adev)
 {
+	uint8_t *discovery_bin = adev->discovery.bin;
+	struct ip_discovery_top *ip_top;
 	struct kset *die_kset;
 	int res, ii;
 
-	adev->ip_top = kzalloc(sizeof(*adev->ip_top), GFP_KERNEL);
-	if (!adev->ip_top)
-		return -ENOMEM;
+	if (!discovery_bin)
+		return -EINVAL;
 
-	adev->ip_top->adev = adev;
+	ip_top = kzalloc(sizeof(*ip_top), GFP_KERNEL);
+	if (!ip_top)
+		return -ENOMEM;
 
-	res = kobject_init_and_add(&adev->ip_top->kobj, &ip_discovery_ktype,
+	ip_top->adev = adev;
+	adev->discovery.ip_top = ip_top;
+	res = kobject_init_and_add(&ip_top->kobj, &ip_discovery_ktype,
 				   &adev->dev->kobj, "ip_discovery");
 	if (res) {
 		DRM_ERROR("Couldn't init and add ip_discovery/");
 		goto Err;
 	}
 
-	die_kset = &adev->ip_top->die_kset;
+	die_kset = &ip_top->die_kset;
 	kobject_set_name(&die_kset->kobj, "%s", "die");
-	die_kset->kobj.parent = &adev->ip_top->kobj;
+	die_kset->kobj.parent = &ip_top->kobj;
 	die_kset->kobj.ktype = &die_kobj_ktype;
-	res = kset_register(&adev->ip_top->die_kset);
+	res = kset_register(&ip_top->die_kset);
 	if (res) {
 		DRM_ERROR("Couldn't register die_kset");
 		goto Err;
@@ -1014,7 +1299,7 @@ static int amdgpu_discovery_sysfs_init(struct amdgpu_device *adev)
 
 	return res;
 Err:
-	kobject_put(&adev->ip_top->kobj);
+	kobject_put(&ip_top->kobj);
 	return res;
 }
 
@@ -1059,10 +1344,11 @@ static void amdgpu_discovery_sysfs_die_free(struct ip_die_entry *ip_die_entry)
 
 static void amdgpu_discovery_sysfs_fini(struct amdgpu_device *adev)
 {
+	struct ip_discovery_top *ip_top = adev->discovery.ip_top;
 	struct list_head *el, *tmp;
 	struct kset *die_kset;
 
-	die_kset = &adev->ip_top->die_kset;
+	die_kset = &ip_top->die_kset;
 	spin_lock(&die_kset->list_lock);
 	list_for_each_prev_safe(el, tmp, &die_kset->list) {
 		list_del_init(el);
@@ -1071,43 +1357,51 @@ static void amdgpu_discovery_sysfs_fini(struct amdgpu_device *adev)
 		spin_lock(&die_kset->list_lock);
 	}
 	spin_unlock(&die_kset->list_lock);
-	kobject_put(&adev->ip_top->die_kset.kobj);
-	kobject_put(&adev->ip_top->kobj);
+	kobject_put(&ip_top->die_kset.kobj);
+	kobject_put(&ip_top->kobj);
 }
 
 /* ================================================== */
 
 static int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev)
 {
+	uint8_t num_base_address, subrev, variant;
 	struct binary_header *bhdr;
 	struct ip_discovery_header *ihdr;
 	struct die_header *dhdr;
-	struct ip *ip;
+	uint8_t *discovery_bin;
+	struct ip_v4 *ip;
 	uint16_t die_offset;
 	uint16_t ip_offset;
 	uint16_t num_dies;
+	uint32_t wafl_ver;
 	uint16_t num_ips;
-	uint8_t num_base_address;
+	uint16_t hw_id;
+	uint8_t inst;
 	int hw_ip;
 	int i, j, k;
 	int r;
 
 	r = amdgpu_discovery_init(adev);
-	if (r) {
-		DRM_ERROR("amdgpu_discovery_init failed\n");
+	if (r)
 		return r;
-	}
-
-	bhdr = (struct binary_header *)adev->mman.discovery_bin;
-	ihdr = (struct ip_discovery_header *)(adev->mman.discovery_bin +
-			le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset));
+	discovery_bin = adev->discovery.bin;
+	wafl_ver = 0;
+	adev->gfx.xcc_mask = 0;
+	adev->sdma.sdma_mask = 0;
+	adev->vcn.inst_mask = 0;
+	adev->jpeg.inst_mask = 0;
+	bhdr = (struct binary_header *)discovery_bin;
+	ihdr = (struct ip_discovery_header
+			*)(discovery_bin +
+			   le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset));
 	num_dies = le16_to_cpu(ihdr->num_dies);
 
 	DRM_DEBUG("number of dies: %d\n", num_dies);
 
 	for (i = 0; i < num_dies; i++) {
 		die_offset = le16_to_cpu(ihdr->die_info[i].die_offset);
-		dhdr = (struct die_header *)(adev->mman.discovery_bin + die_offset);
+		dhdr = (struct die_header *)(discovery_bin + die_offset);
 		num_ips = le16_to_cpu(dhdr->num_ips);
 		ip_offset = die_offset + sizeof(*dhdr);
 
@@ -1121,9 +1415,11 @@ static int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev)
 				le16_to_cpu(dhdr->die_id), num_ips);
 
 		for (j = 0; j < num_ips; j++) {
-			ip = (struct ip *)(adev->mman.discovery_bin + ip_offset);
+			ip = (struct ip_v4 *)(discovery_bin + ip_offset);
 
-			if (amdgpu_discovery_validate_ip(ip))
+			inst = ip->instance_number;
+			hw_id = le16_to_cpu(ip->hw_id);
+			if (amdgpu_discovery_validate_ip(adev, inst, hw_id))
 				goto next_ip;
 
 			num_base_address = ip->num_base_address;
@@ -1131,7 +1427,7 @@ static int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev)
 			DRM_DEBUG("%s(%d) #%d v%d.%d.%d:\n",
 				  hw_id_names[le16_to_cpu(ip->hw_id)],
 				  le16_to_cpu(ip->hw_id),
-				  ip->number_instance,
+				  ip->instance_number,
 				  ip->major, ip->minor,
 				  ip->revision);
 
@@ -1142,26 +1438,45 @@ static int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev)
 				 *     0b10 : encode is disabled
 				 *     0b01 : decode is disabled
 				 */
-				adev->vcn.vcn_config[adev->vcn.num_vcn_inst] =
-					ip->revision & 0xc0;
-				ip->revision &= ~0xc0;
-				if (adev->vcn.num_vcn_inst < AMDGPU_MAX_VCN_INSTANCES)
+				if (adev->vcn.num_vcn_inst <
+				    AMDGPU_MAX_VCN_INSTANCES) {
+					adev->vcn.inst[adev->vcn.num_vcn_inst].vcn_config =
+						ip->revision & 0xc0;
 					adev->vcn.num_vcn_inst++;
-				else
+					adev->vcn.inst_mask |=
+						(1U << ip->instance_number);
+					adev->jpeg.inst_mask |=
+						(1U << ip->instance_number);
+				} else {
 					dev_err(adev->dev, "Too many VCN instances: %d vs %d\n",
 						adev->vcn.num_vcn_inst + 1,
 						AMDGPU_MAX_VCN_INSTANCES);
+				}
+				ip->revision &= ~0xc0;
 			}
 			if (le16_to_cpu(ip->hw_id) == SDMA0_HWID ||
 			    le16_to_cpu(ip->hw_id) == SDMA1_HWID ||
 			    le16_to_cpu(ip->hw_id) == SDMA2_HWID ||
 			    le16_to_cpu(ip->hw_id) == SDMA3_HWID) {
-				if (adev->sdma.num_instances < AMDGPU_MAX_SDMA_INSTANCES)
+				if (adev->sdma.num_instances <
+				    AMDGPU_MAX_SDMA_INSTANCES) {
 					adev->sdma.num_instances++;
-				else
+					adev->sdma.sdma_mask |=
+						(1U << ip->instance_number);
+				} else {
 					dev_err(adev->dev, "Too many SDMA instances: %d vs %d\n",
 						adev->sdma.num_instances + 1,
 						AMDGPU_MAX_SDMA_INSTANCES);
+				}
+			}
+
+			if (le16_to_cpu(ip->hw_id) == VPE_HWID) {
+				if (adev->vpe.num_instances < AMDGPU_MAX_VPE_INSTANCES)
+					adev->vpe.num_instances++;
+				else
+					dev_err(adev->dev, "Too many VPE instances: %d vs %d\n",
+						adev->vpe.num_instances + 1,
+						AMDGPU_MAX_VPE_INSTANCES);
 			}
 
 			if (le16_to_cpu(ip->hw_id) == UMC_HWID) {
@@ -1169,20 +1484,42 @@ static int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev)
 				adev->umc.node_inst_num++;
 			}
 
+			if (le16_to_cpu(ip->hw_id) == GC_HWID)
+				adev->gfx.xcc_mask |=
+					(1U << ip->instance_number);
+
+			if (!wafl_ver && le16_to_cpu(ip->hw_id) == WAFLC_HWID)
+				wafl_ver = IP_VERSION_FULL(ip->major, ip->minor,
+							   ip->revision, 0, 0);
+
 			for (k = 0; k < num_base_address; k++) {
 				/*
 				 * convert the endianness of base addresses in place,
 				 * so that we don't need to convert them when accessing adev->reg_offset.
 				 */
-				ip->base_address[k] = le32_to_cpu(ip->base_address[k]);
+				if (ihdr->base_addr_64_bit)
+					/* Truncate the 64bit base address from ip discovery
+					 * and only store lower 32bit ip base in reg_offset[].
+					 * Bits > 32 follows ASIC specific format, thus just
+					 * discard them and handle it within specific ASIC.
+					 * By this way reg_offset[] and related helpers can
+					 * stay unchanged.
+					 * The base address is in dwords, thus clear the
+					 * highest 2 bits to store.
+					 */
+					ip->base_address[k] =
+						lower_32_bits(le64_to_cpu(ip->base_address_64[k])) & 0x3FFFFFFF;
+				else
+					ip->base_address[k] = le32_to_cpu(ip->base_address[k]);
 				DRM_DEBUG("\t0x%08x\n", ip->base_address[k]);
 			}
 
 			for (hw_ip = 0; hw_ip < MAX_HWIP; hw_ip++) {
-				if (hw_id_map[hw_ip] == le16_to_cpu(ip->hw_id)) {
+				if (hw_id_map[hw_ip] == le16_to_cpu(ip->hw_id) &&
+				    hw_id_map[hw_ip] != 0) {
 					DRM_DEBUG("set register base offset for %s\n",
 							hw_id_names[le16_to_cpu(ip->hw_id)]);
-					adev->reg_offset[hw_ip][ip->number_instance] =
+					adev->reg_offset[hw_ip][ip->instance_number] =
 						ip->base_address;
 					/* Instance support is somewhat inconsistent.
 					 * SDMA is a good example.  Sienna cichlid has 4 total
@@ -1193,80 +1530,59 @@ static int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev)
 					 * example.  On most chips there are multiple instances
 					 * with the same HWID.
 					 */
-					adev->ip_versions[hw_ip][ip->number_instance] =
-						IP_VERSION(ip->major, ip->minor, ip->revision);
+
+					if (ihdr->version < 3) {
+						subrev = 0;
+						variant = 0;
+					} else {
+						subrev = ip->sub_revision;
+						variant = ip->variant;
+					}
+
+					adev->ip_versions[hw_ip]
+							 [ip->instance_number] =
+						IP_VERSION_FULL(ip->major,
+								ip->minor,
+								ip->revision,
+								variant,
+								subrev);
 				}
 			}
 
 next_ip:
-			ip_offset += struct_size(ip, base_address, ip->num_base_address);
+			if (ihdr->base_addr_64_bit)
+				ip_offset += struct_size(ip, base_address_64, ip->num_base_address);
+			else
+				ip_offset += struct_size(ip, base_address, ip->num_base_address);
 		}
 	}
 
-	amdgpu_discovery_sysfs_init(adev);
+	if (wafl_ver && !adev->ip_versions[XGMI_HWIP][0])
+		adev->ip_versions[XGMI_HWIP][0] = wafl_ver;
 
 	return 0;
 }
 
-int amdgpu_discovery_get_ip_version(struct amdgpu_device *adev, int hw_id, int number_instance,
-				    int *major, int *minor, int *revision)
-{
-	struct binary_header *bhdr;
-	struct ip_discovery_header *ihdr;
-	struct die_header *dhdr;
-	struct ip *ip;
-	uint16_t die_offset;
-	uint16_t ip_offset;
-	uint16_t num_dies;
-	uint16_t num_ips;
-	int i, j;
-
-	if (!adev->mman.discovery_bin) {
-		DRM_ERROR("ip discovery uninitialized\n");
-		return -EINVAL;
-	}
-
-	bhdr = (struct binary_header *)adev->mman.discovery_bin;
-	ihdr = (struct ip_discovery_header *)(adev->mman.discovery_bin +
-			le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset));
-	num_dies = le16_to_cpu(ihdr->num_dies);
-
-	for (i = 0; i < num_dies; i++) {
-		die_offset = le16_to_cpu(ihdr->die_info[i].die_offset);
-		dhdr = (struct die_header *)(adev->mman.discovery_bin + die_offset);
-		num_ips = le16_to_cpu(dhdr->num_ips);
-		ip_offset = die_offset + sizeof(*dhdr);
-
-		for (j = 0; j < num_ips; j++) {
-			ip = (struct ip *)(adev->mman.discovery_bin + ip_offset);
-
-			if ((le16_to_cpu(ip->hw_id) == hw_id) && (ip->number_instance == number_instance)) {
-				if (major)
-					*major = ip->major;
-				if (minor)
-					*minor = ip->minor;
-				if (revision)
-					*revision = ip->revision;
-				return 0;
-			}
-			ip_offset += struct_size(ip, base_address, ip->num_base_address);
-		}
-	}
-
-	return -EINVAL;
-}
-
 static void amdgpu_discovery_harvest_ip(struct amdgpu_device *adev)
 {
+	uint8_t *discovery_bin = adev->discovery.bin;
+	struct ip_discovery_header *ihdr;
+	struct binary_header *bhdr;
 	int vcn_harvest_count = 0;
 	int umc_harvest_count = 0;
+	uint16_t offset, ihdr_ver;
 
+	bhdr = (struct binary_header *)discovery_bin;
+	offset = le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset);
+	ihdr = (struct ip_discovery_header *)(discovery_bin + offset);
+	ihdr_ver = le16_to_cpu(ihdr->version);
 	/*
 	 * Harvest table does not fit Navi1x and legacy GPUs,
 	 * so read harvest bit per IP data structure to set
 	 * harvest configuration.
 	 */
-	if (adev->ip_versions[GC_HWIP][0] < IP_VERSION(10, 2, 0)) {
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(10, 2, 0) &&
+	    ihdr_ver <= 2) {
 		if ((adev->pdev->device == 0x731E &&
 			(adev->pdev->revision == 0xC6 ||
 			 adev->pdev->revision == 0xC7)) ||
@@ -1298,27 +1614,30 @@ union gc_info {
 	struct gc_info_v1_0 v1;
 	struct gc_info_v1_1 v1_1;
 	struct gc_info_v1_2 v1_2;
+	struct gc_info_v1_3 v1_3;
 	struct gc_info_v2_0 v2;
+	struct gc_info_v2_1 v2_1;
 };
 
 static int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev)
 {
+	uint8_t *discovery_bin = adev->discovery.bin;
 	struct binary_header *bhdr;
 	union gc_info *gc_info;
 	u16 offset;
 
-	if (!adev->mman.discovery_bin) {
+	if (!discovery_bin) {
 		DRM_ERROR("ip discovery uninitialized\n");
 		return -EINVAL;
 	}
 
-	bhdr = (struct binary_header *)adev->mman.discovery_bin;
+	bhdr = (struct binary_header *)discovery_bin;
 	offset = le16_to_cpu(bhdr->table_list[GC].offset);
 
 	if (!offset)
 		return 0;
 
-	gc_info = (union gc_info *)(adev->mman.discovery_bin + offset);
+	gc_info = (union gc_info *)(discovery_bin + offset);
 
 	switch (le16_to_cpu(gc_info->v1.header.version_major)) {
 	case 1:
@@ -1340,12 +1659,12 @@ static int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev)
 		adev->gfx.config.num_sc_per_sh = le32_to_cpu(gc_info->v1.gc_num_sc_per_se) /
 			le32_to_cpu(gc_info->v1.gc_num_sa_per_se);
 		adev->gfx.config.num_packer_per_sc = le32_to_cpu(gc_info->v1.gc_num_packer_per_sc);
-		if (gc_info->v1.header.version_minor >= 1) {
+		if (le16_to_cpu(gc_info->v1.header.version_minor) >= 1) {
 			adev->gfx.config.gc_num_tcp_per_sa = le32_to_cpu(gc_info->v1_1.gc_num_tcp_per_sa);
 			adev->gfx.config.gc_num_sdp_interface = le32_to_cpu(gc_info->v1_1.gc_num_sdp_interface);
 			adev->gfx.config.gc_num_tcps = le32_to_cpu(gc_info->v1_1.gc_num_tcps);
 		}
-		if (gc_info->v1.header.version_minor >= 2) {
+		if (le16_to_cpu(gc_info->v1.header.version_minor) >= 2) {
 			adev->gfx.config.gc_num_tcp_per_wpg = le32_to_cpu(gc_info->v1_2.gc_num_tcp_per_wpg);
 			adev->gfx.config.gc_tcp_l1_size = le32_to_cpu(gc_info->v1_2.gc_tcp_l1_size);
 			adev->gfx.config.gc_num_sqc_per_wgp = le32_to_cpu(gc_info->v1_2.gc_num_sqc_per_wgp);
@@ -1355,6 +1674,16 @@ static int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev)
 			adev->gfx.config.gc_gl1c_size_per_instance = le32_to_cpu(gc_info->v1_2.gc_gl1c_size_per_instance);
 			adev->gfx.config.gc_gl2c_per_gpu = le32_to_cpu(gc_info->v1_2.gc_gl2c_per_gpu);
 		}
+		if (le16_to_cpu(gc_info->v1.header.version_minor) >= 3) {
+			adev->gfx.config.gc_tcp_size_per_cu = le32_to_cpu(gc_info->v1_3.gc_tcp_size_per_cu);
+			adev->gfx.config.gc_tcp_cache_line_size = le32_to_cpu(gc_info->v1_3.gc_tcp_cache_line_size);
+			adev->gfx.config.gc_instruction_cache_size_per_sqc = le32_to_cpu(gc_info->v1_3.gc_instruction_cache_size_per_sqc);
+			adev->gfx.config.gc_instruction_cache_line_size = le32_to_cpu(gc_info->v1_3.gc_instruction_cache_line_size);
+			adev->gfx.config.gc_scalar_data_cache_size_per_sqc = le32_to_cpu(gc_info->v1_3.gc_scalar_data_cache_size_per_sqc);
+			adev->gfx.config.gc_scalar_data_cache_line_size = le32_to_cpu(gc_info->v1_3.gc_scalar_data_cache_line_size);
+			adev->gfx.config.gc_tcc_size = le32_to_cpu(gc_info->v1_3.gc_tcc_size);
+			adev->gfx.config.gc_tcc_cache_line_size = le32_to_cpu(gc_info->v1_3.gc_tcc_cache_line_size);
+		}
 		break;
 	case 2:
 		adev->gfx.config.max_shader_engines = le32_to_cpu(gc_info->v2.gc_num_se);
@@ -1374,6 +1703,15 @@ static int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev)
 		adev->gfx.config.num_sc_per_sh = le32_to_cpu(gc_info->v2.gc_num_sc_per_se) /
 			le32_to_cpu(gc_info->v2.gc_num_sh_per_se);
 		adev->gfx.config.num_packer_per_sc = le32_to_cpu(gc_info->v2.gc_num_packer_per_sc);
+		if (le16_to_cpu(gc_info->v2.header.version_minor) == 1) {
+			adev->gfx.config.gc_num_tcp_per_sa = le32_to_cpu(gc_info->v2_1.gc_num_tcp_per_sh);
+			adev->gfx.config.gc_tcp_size_per_cu = le32_to_cpu(gc_info->v2_1.gc_tcp_size_per_cu);
+			adev->gfx.config.gc_num_sdp_interface = le32_to_cpu(gc_info->v2_1.gc_num_sdp_interface); /* per XCD */
+			adev->gfx.config.gc_num_cu_per_sqc = le32_to_cpu(gc_info->v2_1.gc_num_cu_per_sqc);
+			adev->gfx.config.gc_l1_instruction_cache_size_per_sqc = le32_to_cpu(gc_info->v2_1.gc_instruction_cache_size_per_sqc);
+			adev->gfx.config.gc_l1_data_cache_size_per_sqc = le32_to_cpu(gc_info->v2_1.gc_scalar_data_cache_size_per_sqc);
+			adev->gfx.config.gc_tcc_size = le32_to_cpu(gc_info->v2_1.gc_tcc_size); /* per XCD */
+		}
 		break;
 	default:
 		dev_err(adev->dev,
@@ -1387,28 +1725,30 @@ static int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev)
 
 union mall_info {
 	struct mall_info_v1_0 v1;
+	struct mall_info_v2_0 v2;
 };
 
 static int amdgpu_discovery_get_mall_info(struct amdgpu_device *adev)
 {
+	uint8_t *discovery_bin = adev->discovery.bin;
 	struct binary_header *bhdr;
 	union mall_info *mall_info;
 	u32 u, mall_size_per_umc, m_s_present, half_use;
 	u64 mall_size;
 	u16 offset;
 
-	if (!adev->mman.discovery_bin) {
+	if (!discovery_bin) {
 		DRM_ERROR("ip discovery uninitialized\n");
 		return -EINVAL;
 	}
 
-	bhdr = (struct binary_header *)adev->mman.discovery_bin;
+	bhdr = (struct binary_header *)discovery_bin;
 	offset = le16_to_cpu(bhdr->table_list[MALL_INFO].offset);
 
 	if (!offset)
 		return 0;
 
-	mall_info = (union mall_info *)(adev->mman.discovery_bin + offset);
+	mall_info = (union mall_info *)(discovery_bin + offset);
 
 	switch (le16_to_cpu(mall_info->v1.header.version_major)) {
 	case 1:
@@ -1425,6 +1765,11 @@ static int amdgpu_discovery_get_mall_info(struct amdgpu_device *adev)
 				mall_size += mall_size_per_umc;
 		}
 		adev->gmc.mall_size = mall_size;
+		adev->gmc.m_half_use = half_use;
+		break;
+	case 2:
+		mall_size_per_umc = le32_to_cpu(mall_info->v2.mall_size_per_umc);
+		adev->gmc.mall_size = (uint64_t)mall_size_per_umc * adev->gmc.num_umc;
 		break;
 	default:
 		dev_err(adev->dev,
@@ -1442,12 +1787,13 @@ union vcn_info {
 
 static int amdgpu_discovery_get_vcn_info(struct amdgpu_device *adev)
 {
+	uint8_t *discovery_bin = adev->discovery.bin;
 	struct binary_header *bhdr;
 	union vcn_info *vcn_info;
 	u16 offset;
 	int v;
 
-	if (!adev->mman.discovery_bin) {
+	if (!discovery_bin) {
 		DRM_ERROR("ip discovery uninitialized\n");
 		return -EINVAL;
 	}
@@ -1462,13 +1808,13 @@ static int amdgpu_discovery_get_vcn_info(struct amdgpu_device *adev)
 		return -EINVAL;
 	}
 
-	bhdr = (struct binary_header *)adev->mman.discovery_bin;
+	bhdr = (struct binary_header *)discovery_bin;
 	offset = le16_to_cpu(bhdr->table_list[VCN_INFO].offset);
 
 	if (!offset)
 		return 0;
 
-	vcn_info = (union vcn_info *)(adev->mman.discovery_bin + offset);
+	vcn_info = (union vcn_info *)(discovery_bin + offset);
 
 	switch (le16_to_cpu(vcn_info->v1.header.version_major)) {
 	case 1:
@@ -1476,7 +1822,7 @@ static int amdgpu_discovery_get_vcn_info(struct amdgpu_device *adev)
 		 * so this won't overflow.
 		 */
 		for (v = 0; v < adev->vcn.num_vcn_inst; v++) {
-			adev->vcn.vcn_codec_disable_mask[v] =
+			adev->vcn.inst[v].vcn_codec_disable_mask =
 				le32_to_cpu(vcn_info->v1.instance_info[v].fuse_data.all_bits);
 		}
 		break;
@@ -1490,10 +1836,113 @@ static int amdgpu_discovery_get_vcn_info(struct amdgpu_device *adev)
 	return 0;
 }
 
+union nps_info {
+	struct nps_info_v1_0 v1;
+};
+
+static int amdgpu_discovery_refresh_nps_info(struct amdgpu_device *adev,
+					     union nps_info *nps_data)
+{
+	uint64_t vram_size, pos, offset;
+	struct nps_info_header *nhdr;
+	struct binary_header bhdr;
+	uint16_t checksum;
+
+	vram_size = (uint64_t)RREG32(mmRCC_CONFIG_MEMSIZE) << 20;
+	pos = vram_size - DISCOVERY_TMR_OFFSET;
+	amdgpu_device_vram_access(adev, pos, &bhdr, sizeof(bhdr), false);
+
+	offset = le16_to_cpu(bhdr.table_list[NPS_INFO].offset);
+	checksum = le16_to_cpu(bhdr.table_list[NPS_INFO].checksum);
+
+	amdgpu_device_vram_access(adev, (pos + offset), nps_data,
+				  sizeof(*nps_data), false);
+
+	nhdr = (struct nps_info_header *)(nps_data);
+	if (!amdgpu_discovery_verify_checksum((uint8_t *)nps_data,
+					      le32_to_cpu(nhdr->size_bytes),
+					      checksum)) {
+		dev_err(adev->dev, "nps data refresh, checksum mismatch\n");
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+int amdgpu_discovery_get_nps_info(struct amdgpu_device *adev,
+				  uint32_t *nps_type,
+				  struct amdgpu_gmc_memrange **ranges,
+				  int *range_cnt, bool refresh)
+{
+	uint8_t *discovery_bin = adev->discovery.bin;
+	struct amdgpu_gmc_memrange *mem_ranges;
+	struct binary_header *bhdr;
+	union nps_info *nps_info;
+	union nps_info nps_data;
+	u16 offset;
+	int i, r;
+
+	if (!nps_type || !range_cnt || !ranges)
+		return -EINVAL;
+
+	if (refresh) {
+		r = amdgpu_discovery_refresh_nps_info(adev, &nps_data);
+		if (r)
+			return r;
+		nps_info = &nps_data;
+	} else {
+		if (!discovery_bin) {
+			dev_err(adev->dev,
+				"fetch mem range failed, ip discovery uninitialized\n");
+			return -EINVAL;
+		}
+
+		bhdr = (struct binary_header *)discovery_bin;
+		offset = le16_to_cpu(bhdr->table_list[NPS_INFO].offset);
+
+		if (!offset)
+			return -ENOENT;
+
+		/* If verification fails, return as if NPS table doesn't exist */
+		if (amdgpu_discovery_verify_npsinfo(adev, bhdr))
+			return -ENOENT;
+
+		nps_info = (union nps_info *)(discovery_bin + offset);
+	}
+
+	switch (le16_to_cpu(nps_info->v1.header.version_major)) {
+	case 1:
+		mem_ranges = kvcalloc(nps_info->v1.count,
+				      sizeof(*mem_ranges),
+				      GFP_KERNEL);
+		if (!mem_ranges)
+			return -ENOMEM;
+		*nps_type = nps_info->v1.nps_type;
+		*range_cnt = nps_info->v1.count;
+		for (i = 0; i < *range_cnt; i++) {
+			mem_ranges[i].base_address =
+				nps_info->v1.instance_info[i].base_address;
+			mem_ranges[i].limit_address =
+				nps_info->v1.instance_info[i].limit_address;
+			mem_ranges[i].nid_mask = -1;
+			mem_ranges[i].flags = 0;
+		}
+		*ranges = mem_ranges;
+		break;
+	default:
+		dev_err(adev->dev, "Unhandled NPS info table %d.%d\n",
+			le16_to_cpu(nps_info->v1.header.version_major),
+			le16_to_cpu(nps_info->v1.header.version_minor));
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
 static int amdgpu_discovery_set_common_ip_blocks(struct amdgpu_device *adev)
 {
 	/* what IP to use for this? */
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 1, 0):
 	case IP_VERSION(9, 2, 1):
@@ -1503,6 +1952,8 @@ static int amdgpu_discovery_set_common_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(9, 4, 1):
 	case IP_VERSION(9, 4, 2):
 	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+	case IP_VERSION(9, 5, 0):
 		amdgpu_device_ip_block_add(adev, &vega10_common_ip_block);
 		break;
 	case IP_VERSION(10, 1, 10):
@@ -1525,12 +1976,20 @@ static int amdgpu_discovery_set_common_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 3):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
 		amdgpu_device_ip_block_add(adev, &soc21_common_ip_block);
 		break;
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		amdgpu_device_ip_block_add(adev, &soc24_common_ip_block);
+		break;
 	default:
 		dev_err(adev->dev,
 			"Failed to add common ip block(GC_HWIP:0x%x)\n",
-			adev->ip_versions[GC_HWIP][0]);
+			amdgpu_ip_version(adev, GC_HWIP, 0));
 		return -EINVAL;
 	}
 	return 0;
@@ -1539,7 +1998,7 @@ static int amdgpu_discovery_set_common_ip_blocks(struct amdgpu_device *adev)
 static int amdgpu_discovery_set_gmc_ip_blocks(struct amdgpu_device *adev)
 {
 	/* use GC or MMHUB IP version */
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 1, 0):
 	case IP_VERSION(9, 2, 1):
@@ -1549,6 +2008,8 @@ static int amdgpu_discovery_set_gmc_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(9, 4, 1):
 	case IP_VERSION(9, 4, 2):
 	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+	case IP_VERSION(9, 5, 0):
 		amdgpu_device_ip_block_add(adev, &gmc_v9_0_ip_block);
 		break;
 	case IP_VERSION(10, 1, 10):
@@ -1571,12 +2032,19 @@ static int amdgpu_discovery_set_gmc_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 3):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
 		amdgpu_device_ip_block_add(adev, &gmc_v11_0_ip_block);
 		break;
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		amdgpu_device_ip_block_add(adev, &gmc_v12_0_ip_block);
+		break;
 	default:
-		dev_err(adev->dev,
-			"Failed to add gmc ip block(GC_HWIP:0x%x)\n",
-			adev->ip_versions[GC_HWIP][0]);
+		dev_err(adev->dev, "Failed to add gmc ip block(GC_HWIP:0x%x)\n",
+			amdgpu_ip_version(adev, GC_HWIP, 0));
 		return -EINVAL;
 	}
 	return 0;
@@ -1584,7 +2052,7 @@ static int amdgpu_discovery_set_gmc_ip_blocks(struct amdgpu_device *adev)
 
 static int amdgpu_discovery_set_ih_ip_blocks(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[OSSSYS_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, OSSSYS_HWIP, 0)) {
 	case IP_VERSION(4, 0, 0):
 	case IP_VERSION(4, 0, 1):
 	case IP_VERSION(4, 1, 0):
@@ -1596,6 +2064,7 @@ static int amdgpu_discovery_set_ih_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(4, 2, 1):
 	case IP_VERSION(4, 4, 0):
 	case IP_VERSION(4, 4, 2):
+	case IP_VERSION(4, 4, 5):
 		amdgpu_device_ip_block_add(adev, &vega20_ih_ip_block);
 		break;
 	case IP_VERSION(5, 0, 0):
@@ -1611,10 +2080,16 @@ static int amdgpu_discovery_set_ih_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(6, 0, 2):
 		amdgpu_device_ip_block_add(adev, &ih_v6_0_ip_block);
 		break;
+	case IP_VERSION(6, 1, 0):
+		amdgpu_device_ip_block_add(adev, &ih_v6_1_ip_block);
+		break;
+	case IP_VERSION(7, 0, 0):
+		amdgpu_device_ip_block_add(adev, &ih_v7_0_ip_block);
+		break;
 	default:
 		dev_err(adev->dev,
 			"Failed to add ih ip block(OSSSYS_HWIP:0x%x)\n",
-			adev->ip_versions[OSSSYS_HWIP][0]);
+			amdgpu_ip_version(adev, OSSSYS_HWIP, 0));
 		return -EINVAL;
 	}
 	return 0;
@@ -1622,7 +2097,7 @@ static int amdgpu_discovery_set_ih_ip_blocks(struct amdgpu_device *adev)
 
 static int amdgpu_discovery_set_psp_ip_blocks(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[MP0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 	case IP_VERSION(9, 0, 0):
 		amdgpu_device_ip_block_add(adev, &psp_v3_1_ip_block);
 		break;
@@ -1640,6 +2115,7 @@ static int amdgpu_discovery_set_psp_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(11, 0, 12):
 	case IP_VERSION(11, 0, 13):
 	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 2):
 		amdgpu_device_ip_block_add(adev, &psp_v11_0_ip_block);
 		break;
 	case IP_VERSION(11, 0, 8):
@@ -1659,15 +2135,25 @@ static int amdgpu_discovery_set_psp_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(13, 0, 8):
 	case IP_VERSION(13, 0, 10):
 	case IP_VERSION(13, 0, 11):
+	case IP_VERSION(13, 0, 12):
+	case IP_VERSION(13, 0, 14):
+	case IP_VERSION(14, 0, 0):
+	case IP_VERSION(14, 0, 1):
+	case IP_VERSION(14, 0, 4):
 		amdgpu_device_ip_block_add(adev, &psp_v13_0_ip_block);
 		break;
 	case IP_VERSION(13, 0, 4):
 		amdgpu_device_ip_block_add(adev, &psp_v13_0_4_ip_block);
 		break;
+	case IP_VERSION(14, 0, 2):
+	case IP_VERSION(14, 0, 3):
+	case IP_VERSION(14, 0, 5):
+		amdgpu_device_ip_block_add(adev, &psp_v14_0_ip_block);
+		break;
 	default:
 		dev_err(adev->dev,
 			"Failed to add psp ip block(MP0_HWIP:0x%x)\n",
-			adev->ip_versions[MP0_HWIP][0]);
+			amdgpu_ip_version(adev, MP0_HWIP, 0));
 		return -EINVAL;
 	}
 	return 0;
@@ -1675,7 +2161,7 @@ static int amdgpu_discovery_set_psp_ip_blocks(struct amdgpu_device *adev)
 
 static int amdgpu_discovery_set_smu_ip_blocks(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[MP1_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
 	case IP_VERSION(9, 0, 0):
 	case IP_VERSION(10, 0, 0):
 	case IP_VERSION(10, 0, 1):
@@ -1689,13 +2175,17 @@ static int amdgpu_discovery_set_smu_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(11, 0, 5):
 	case IP_VERSION(11, 0, 9):
 	case IP_VERSION(11, 0, 7):
-	case IP_VERSION(11, 0, 8):
 	case IP_VERSION(11, 0, 11):
 	case IP_VERSION(11, 0, 12):
 	case IP_VERSION(11, 0, 13):
 	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 2):
 		amdgpu_device_ip_block_add(adev, &smu_v11_0_ip_block);
 		break;
+	case IP_VERSION(11, 0, 8):
+		if (adev->apu_flags & AMD_APU_IS_CYAN_SKILLFISH2)
+			amdgpu_device_ip_block_add(adev, &smu_v11_0_ip_block);
+		break;
 	case IP_VERSION(12, 0, 0):
 	case IP_VERSION(12, 0, 1):
 		amdgpu_device_ip_block_add(adev, &smu_v12_0_ip_block);
@@ -1706,16 +2196,27 @@ static int amdgpu_discovery_set_smu_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(13, 0, 3):
 	case IP_VERSION(13, 0, 4):
 	case IP_VERSION(13, 0, 5):
+	case IP_VERSION(13, 0, 6):
 	case IP_VERSION(13, 0, 7):
 	case IP_VERSION(13, 0, 8):
 	case IP_VERSION(13, 0, 10):
 	case IP_VERSION(13, 0, 11):
+	case IP_VERSION(13, 0, 14):
+	case IP_VERSION(13, 0, 12):
 		amdgpu_device_ip_block_add(adev, &smu_v13_0_ip_block);
 		break;
+	case IP_VERSION(14, 0, 0):
+	case IP_VERSION(14, 0, 1):
+	case IP_VERSION(14, 0, 2):
+	case IP_VERSION(14, 0, 3):
+	case IP_VERSION(14, 0, 4):
+	case IP_VERSION(14, 0, 5):
+		amdgpu_device_ip_block_add(adev, &smu_v14_0_ip_block);
+		break;
 	default:
 		dev_err(adev->dev,
 			"Failed to add smu ip block(MP1_HWIP:0x%x)\n",
-			adev->ip_versions[MP1_HWIP][0]);
+			amdgpu_ip_version(adev, MP1_HWIP, 0));
 		return -EINVAL;
 	}
 	return 0;
@@ -1740,8 +2241,8 @@ static int amdgpu_discovery_set_display_ip_blocks(struct amdgpu_device *adev)
 		return 0;
 
 #if defined(CONFIG_DRM_AMD_DC)
-	if (adev->ip_versions[DCE_HWIP][0]) {
-		switch (adev->ip_versions[DCE_HWIP][0]) {
+	if (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
+		switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
 		case IP_VERSION(1, 0, 0):
 		case IP_VERSION(1, 0, 1):
 		case IP_VERSION(2, 0, 2):
@@ -1759,6 +2260,14 @@ static int amdgpu_discovery_set_display_ip_blocks(struct amdgpu_device *adev)
 		case IP_VERSION(3, 1, 6):
 		case IP_VERSION(3, 2, 0):
 		case IP_VERSION(3, 2, 1):
+		case IP_VERSION(3, 5, 0):
+		case IP_VERSION(3, 5, 1):
+		case IP_VERSION(3, 6, 0):
+		case IP_VERSION(4, 1, 0):
+			/* TODO: Fix IP version. DC code expects version 4.0.1 */
+			if (adev->ip_versions[DCE_HWIP][0] == IP_VERSION(4, 1, 0))
+				adev->ip_versions[DCE_HWIP][0] = IP_VERSION(4, 0, 1);
+
 			if (amdgpu_sriov_vf(adev))
 				amdgpu_discovery_set_sriov_display(adev);
 			else
@@ -1767,11 +2276,11 @@ static int amdgpu_discovery_set_display_ip_blocks(struct amdgpu_device *adev)
 		default:
 			dev_err(adev->dev,
 				"Failed to add dm ip block(DCE_HWIP:0x%x)\n",
-				adev->ip_versions[DCE_HWIP][0]);
+				amdgpu_ip_version(adev, DCE_HWIP, 0));
 			return -EINVAL;
 		}
-	} else if (adev->ip_versions[DCI_HWIP][0]) {
-		switch (adev->ip_versions[DCI_HWIP][0]) {
+	} else if (amdgpu_ip_version(adev, DCI_HWIP, 0)) {
+		switch (amdgpu_ip_version(adev, DCI_HWIP, 0)) {
 		case IP_VERSION(12, 0, 0):
 		case IP_VERSION(12, 0, 1):
 		case IP_VERSION(12, 1, 0):
@@ -1783,7 +2292,7 @@ static int amdgpu_discovery_set_display_ip_blocks(struct amdgpu_device *adev)
 		default:
 			dev_err(adev->dev,
 				"Failed to add dm ip block(DCI_HWIP:0x%x)\n",
-				adev->ip_versions[DCI_HWIP][0]);
+				amdgpu_ip_version(adev, DCI_HWIP, 0));
 			return -EINVAL;
 		}
 	}
@@ -1793,7 +2302,7 @@ static int amdgpu_discovery_set_display_ip_blocks(struct amdgpu_device *adev)
 
 static int amdgpu_discovery_set_gc_ip_blocks(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 1, 0):
 	case IP_VERSION(9, 2, 1):
@@ -1804,6 +2313,11 @@ static int amdgpu_discovery_set_gc_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(9, 4, 2):
 		amdgpu_device_ip_block_add(adev, &gfx_v9_0_ip_block);
 		break;
+	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+	case IP_VERSION(9, 5, 0):
+		amdgpu_device_ip_block_add(adev, &gfx_v9_4_3_ip_block);
+		break;
 	case IP_VERSION(10, 1, 10):
 	case IP_VERSION(10, 1, 2):
 	case IP_VERSION(10, 1, 1):
@@ -1824,12 +2338,19 @@ static int amdgpu_discovery_set_gc_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 3):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
 		amdgpu_device_ip_block_add(adev, &gfx_v11_0_ip_block);
 		break;
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		amdgpu_device_ip_block_add(adev, &gfx_v12_0_ip_block);
+		break;
 	default:
-		dev_err(adev->dev,
-			"Failed to add gfx ip block(GC_HWIP:0x%x)\n",
-			adev->ip_versions[GC_HWIP][0]);
+		dev_err(adev->dev, "Failed to add gfx ip block(GC_HWIP:0x%x)\n",
+			amdgpu_ip_version(adev, GC_HWIP, 0));
 		return -EINVAL;
 	}
 	return 0;
@@ -1837,7 +2358,7 @@ static int amdgpu_discovery_set_gc_ip_blocks(struct amdgpu_device *adev)
 
 static int amdgpu_discovery_set_sdma_ip_blocks(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[SDMA0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
 	case IP_VERSION(4, 0, 0):
 	case IP_VERSION(4, 0, 1):
 	case IP_VERSION(4, 1, 0):
@@ -1849,6 +2370,8 @@ static int amdgpu_discovery_set_sdma_ip_blocks(struct amdgpu_device *adev)
 		amdgpu_device_ip_block_add(adev, &sdma_v4_0_ip_block);
 		break;
 	case IP_VERSION(4, 4, 2):
+	case IP_VERSION(4, 4, 5):
+	case IP_VERSION(4, 4, 4):
 		amdgpu_device_ip_block_add(adev, &sdma_v4_4_2_ip_block);
 		break;
 	case IP_VERSION(5, 0, 0):
@@ -1871,21 +2394,44 @@ static int amdgpu_discovery_set_sdma_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(6, 0, 1):
 	case IP_VERSION(6, 0, 2):
 	case IP_VERSION(6, 0, 3):
+	case IP_VERSION(6, 1, 0):
+	case IP_VERSION(6, 1, 1):
+	case IP_VERSION(6, 1, 2):
+	case IP_VERSION(6, 1, 3):
 		amdgpu_device_ip_block_add(adev, &sdma_v6_0_ip_block);
 		break;
+	case IP_VERSION(7, 0, 0):
+	case IP_VERSION(7, 0, 1):
+		amdgpu_device_ip_block_add(adev, &sdma_v7_0_ip_block);
+		break;
 	default:
 		dev_err(adev->dev,
 			"Failed to add sdma ip block(SDMA0_HWIP:0x%x)\n",
-			adev->ip_versions[SDMA0_HWIP][0]);
+			amdgpu_ip_version(adev, SDMA0_HWIP, 0));
 		return -EINVAL;
 	}
+
+	return 0;
+}
+
+static int amdgpu_discovery_set_ras_ip_blocks(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 12):
+	case IP_VERSION(13, 0, 14):
+		amdgpu_device_ip_block_add(adev, &ras_v1_0_ip_block);
+		break;
+	default:
+		break;
+	}
 	return 0;
 }
 
 static int amdgpu_discovery_set_mm_ip_blocks(struct amdgpu_device *adev)
 {
-	if (adev->ip_versions[VCE_HWIP][0]) {
-		switch (adev->ip_versions[UVD_HWIP][0]) {
+	if (amdgpu_ip_version(adev, VCE_HWIP, 0)) {
+		switch (amdgpu_ip_version(adev, UVD_HWIP, 0)) {
 		case IP_VERSION(7, 0, 0):
 		case IP_VERSION(7, 2, 0):
 			/* UVD is not supported on vega20 SR-IOV */
@@ -1895,10 +2441,10 @@ static int amdgpu_discovery_set_mm_ip_blocks(struct amdgpu_device *adev)
 		default:
 			dev_err(adev->dev,
 				"Failed to add uvd v7 ip block(UVD_HWIP:0x%x)\n",
-				adev->ip_versions[UVD_HWIP][0]);
+				amdgpu_ip_version(adev, UVD_HWIP, 0));
 			return -EINVAL;
 		}
-		switch (adev->ip_versions[VCE_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, VCE_HWIP, 0)) {
 		case IP_VERSION(4, 0, 0):
 		case IP_VERSION(4, 1, 0):
 			/* VCE is not supported on vega20 SR-IOV */
@@ -1908,11 +2454,11 @@ static int amdgpu_discovery_set_mm_ip_blocks(struct amdgpu_device *adev)
 		default:
 			dev_err(adev->dev,
 				"Failed to add VCE v4 ip block(VCE_HWIP:0x%x)\n",
-				adev->ip_versions[VCE_HWIP][0]);
+				amdgpu_ip_version(adev, VCE_HWIP, 0));
 			return -EINVAL;
 		}
 	} else {
-		switch (adev->ip_versions[UVD_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, UVD_HWIP, 0)) {
 		case IP_VERSION(1, 0, 0):
 		case IP_VERSION(1, 0, 1):
 			amdgpu_device_ip_block_add(adev, &vcn_v1_0_ip_block);
@@ -1939,7 +2485,6 @@ static int amdgpu_discovery_set_mm_ip_blocks(struct amdgpu_device *adev)
 		case IP_VERSION(3, 1, 1):
 		case IP_VERSION(3, 1, 2):
 		case IP_VERSION(3, 0, 2):
-		case IP_VERSION(3, 0, 192):
 			amdgpu_device_ip_block_add(adev, &vcn_v3_0_ip_block);
 			if (!amdgpu_sriov_vf(adev))
 				amdgpu_device_ip_block_add(adev, &jpeg_v3_0_ip_block);
@@ -1952,11 +2497,28 @@ static int amdgpu_discovery_set_mm_ip_blocks(struct amdgpu_device *adev)
 		case IP_VERSION(4, 0, 4):
 			amdgpu_device_ip_block_add(adev, &vcn_v4_0_ip_block);
 			amdgpu_device_ip_block_add(adev, &jpeg_v4_0_ip_block);
-			return 0;
+			break;
+		case IP_VERSION(4, 0, 3):
+			amdgpu_device_ip_block_add(adev, &vcn_v4_0_3_ip_block);
+			amdgpu_device_ip_block_add(adev, &jpeg_v4_0_3_ip_block);
+			break;
+		case IP_VERSION(4, 0, 5):
+		case IP_VERSION(4, 0, 6):
+			amdgpu_device_ip_block_add(adev, &vcn_v4_0_5_ip_block);
+			amdgpu_device_ip_block_add(adev, &jpeg_v4_0_5_ip_block);
+			break;
+		case IP_VERSION(5, 0, 0):
+			amdgpu_device_ip_block_add(adev, &vcn_v5_0_0_ip_block);
+			amdgpu_device_ip_block_add(adev, &jpeg_v5_0_0_ip_block);
+			break;
+		case IP_VERSION(5, 0, 1):
+			amdgpu_device_ip_block_add(adev, &vcn_v5_0_1_ip_block);
+			amdgpu_device_ip_block_add(adev, &jpeg_v5_0_1_ip_block);
+			break;
 		default:
 			dev_err(adev->dev,
 				"Failed to add vcn/jpeg ip block(UVD_HWIP:0x%x)\n",
-				adev->ip_versions[UVD_HWIP][0]);
+				amdgpu_ip_version(adev, UVD_HWIP, 0));
 			return -EINVAL;
 		}
 	}
@@ -1965,38 +2527,94 @@ static int amdgpu_discovery_set_mm_ip_blocks(struct amdgpu_device *adev)
 
 static int amdgpu_discovery_set_mes_ip_blocks(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
-	case IP_VERSION(10, 1, 10):
-	case IP_VERSION(10, 1, 1):
-	case IP_VERSION(10, 1, 2):
-	case IP_VERSION(10, 1, 3):
-	case IP_VERSION(10, 1, 4):
-	case IP_VERSION(10, 3, 0):
-	case IP_VERSION(10, 3, 1):
-	case IP_VERSION(10, 3, 2):
-	case IP_VERSION(10, 3, 3):
-	case IP_VERSION(10, 3, 4):
-	case IP_VERSION(10, 3, 5):
-	case IP_VERSION(10, 3, 6):
-		if (amdgpu_mes) {
-			amdgpu_device_ip_block_add(adev, &mes_v10_1_ip_block);
-			adev->enable_mes = true;
-			if (amdgpu_mes_kiq)
-				adev->enable_mes_kiq = true;
-		}
-		break;
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 0):
 	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 3):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
 		amdgpu_device_ip_block_add(adev, &mes_v11_0_ip_block);
 		adev->enable_mes = true;
 		adev->enable_mes_kiq = true;
 		break;
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		amdgpu_device_ip_block_add(adev, &mes_v12_0_ip_block);
+		adev->enable_mes = true;
+		adev->enable_mes_kiq = true;
+		if (amdgpu_uni_mes)
+			adev->enable_uni_mes = true;
+		break;
+	default:
+		break;
+	}
+	return 0;
+}
+
+static void amdgpu_discovery_init_soc_config(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+	case IP_VERSION(9, 5, 0):
+		aqua_vanjaram_init_soc_config(adev);
+		break;
+	default:
+		break;
+	}
+}
+
+static int amdgpu_discovery_set_vpe_ip_blocks(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, VPE_HWIP, 0)) {
+	case IP_VERSION(6, 1, 0):
+	case IP_VERSION(6, 1, 1):
+	case IP_VERSION(6, 1, 3):
+		amdgpu_device_ip_block_add(adev, &vpe_v6_1_ip_block);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int amdgpu_discovery_set_umsch_mm_ip_blocks(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, VCN_HWIP, 0)) {
+	case IP_VERSION(4, 0, 5):
+	case IP_VERSION(4, 0, 6):
+		if (amdgpu_umsch_mm & 0x1) {
+			amdgpu_device_ip_block_add(adev, &umsch_mm_v4_0_ip_block);
+			adev->enable_umsch_mm = true;
+		}
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int amdgpu_discovery_set_isp_ip_blocks(struct amdgpu_device *adev)
+{
+#if defined(CONFIG_DRM_AMD_ISP)
+	switch (amdgpu_ip_version(adev, ISP_HWIP, 0)) {
+	case IP_VERSION(4, 1, 0):
+		amdgpu_device_ip_block_add(adev, &isp_v4_1_0_ip_block);
+		break;
+	case IP_VERSION(4, 1, 1):
+		amdgpu_device_ip_block_add(adev, &isp_v4_1_1_ip_block);
+		break;
 	default:
 		break;
 	}
+#endif
+
 	return 0;
 }
 
@@ -2006,9 +2624,16 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 
 	switch (adev->asic_type) {
 	case CHIP_VEGA10:
+		/* This is not fatal.  We only need the discovery
+		 * binary for sysfs.  We don't need it for a
+		 * functional system.
+		 */
+		amdgpu_discovery_init(adev);
 		vega10_reg_base_init(adev);
 		adev->sdma.num_instances = 2;
+		adev->sdma.sdma_mask = 3;
 		adev->gmc.num_umc = 4;
+		adev->gfx.xcc_mask = 1;
 		adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 0, 0);
 		adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 0, 0);
 		adev->ip_versions[OSSSYS_HWIP][0] = IP_VERSION(4, 0, 0);
@@ -2028,9 +2653,16 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		adev->ip_versions[DCI_HWIP][0] = IP_VERSION(12, 0, 0);
 		break;
 	case CHIP_VEGA12:
+		/* This is not fatal.  We only need the discovery
+		 * binary for sysfs.  We don't need it for a
+		 * functional system.
+		 */
+		amdgpu_discovery_init(adev);
 		vega10_reg_base_init(adev);
 		adev->sdma.num_instances = 2;
+		adev->sdma.sdma_mask = 3;
 		adev->gmc.num_umc = 4;
+		adev->gfx.xcc_mask = 1;
 		adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 3, 0);
 		adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 3, 0);
 		adev->ip_versions[OSSSYS_HWIP][0] = IP_VERSION(4, 0, 1);
@@ -2050,10 +2682,17 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		adev->ip_versions[DCI_HWIP][0] = IP_VERSION(12, 0, 1);
 		break;
 	case CHIP_RAVEN:
+		/* This is not fatal.  We only need the discovery
+		 * binary for sysfs.  We don't need it for a
+		 * functional system.
+		 */
+		amdgpu_discovery_init(adev);
 		vega10_reg_base_init(adev);
 		adev->sdma.num_instances = 1;
+		adev->sdma.sdma_mask = 1;
 		adev->vcn.num_vcn_inst = 1;
 		adev->gmc.num_umc = 2;
+		adev->gfx.xcc_mask = 1;
 		if (adev->apu_flags & AMD_APU_IS_RAVEN2) {
 			adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 2, 0);
 			adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 2, 0);
@@ -2070,6 +2709,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 			adev->ip_versions[GC_HWIP][0] = IP_VERSION(9, 2, 2);
 			adev->ip_versions[UVD_HWIP][0] = IP_VERSION(1, 0, 1);
 			adev->ip_versions[DCE_HWIP][0] = IP_VERSION(1, 0, 1);
+			adev->ip_versions[ISP_HWIP][0] = IP_VERSION(2, 0, 0);
 		} else {
 			adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 1, 0);
 			adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 1, 0);
@@ -2086,12 +2726,20 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 			adev->ip_versions[GC_HWIP][0] = IP_VERSION(9, 1, 0);
 			adev->ip_versions[UVD_HWIP][0] = IP_VERSION(1, 0, 0);
 			adev->ip_versions[DCE_HWIP][0] = IP_VERSION(1, 0, 0);
+			adev->ip_versions[ISP_HWIP][0] = IP_VERSION(2, 0, 0);
 		}
 		break;
 	case CHIP_VEGA20:
+		/* This is not fatal.  We only need the discovery
+		 * binary for sysfs.  We don't need it for a
+		 * functional system.
+		 */
+		amdgpu_discovery_init(adev);
 		vega20_reg_base_init(adev);
 		adev->sdma.num_instances = 2;
+		adev->sdma.sdma_mask = 3;
 		adev->gmc.num_umc = 8;
+		adev->gfx.xcc_mask = 1;
 		adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 4, 0);
 		adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 4, 0);
 		adev->ip_versions[OSSSYS_HWIP][0] = IP_VERSION(4, 2, 0);
@@ -2112,10 +2760,17 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		adev->ip_versions[DCI_HWIP][0] = IP_VERSION(12, 1, 0);
 		break;
 	case CHIP_ARCTURUS:
+		/* This is not fatal.  We only need the discovery
+		 * binary for sysfs.  We don't need it for a
+		 * functional system.
+		 */
+		amdgpu_discovery_init(adev);
 		arct_reg_base_init(adev);
 		adev->sdma.num_instances = 8;
+		adev->sdma.sdma_mask = 0xff;
 		adev->vcn.num_vcn_inst = 2;
 		adev->gmc.num_umc = 8;
+		adev->gfx.xcc_mask = 1;
 		adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 4, 1);
 		adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 4, 1);
 		adev->ip_versions[OSSSYS_HWIP][0] = IP_VERSION(4, 2, 1);
@@ -2140,10 +2795,17 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		adev->ip_versions[UVD_HWIP][1] = IP_VERSION(2, 5, 0);
 		break;
 	case CHIP_ALDEBARAN:
+		/* This is not fatal.  We only need the discovery
+		 * binary for sysfs.  We don't need it for a
+		 * functional system.
+		 */
+		amdgpu_discovery_init(adev);
 		aldebaran_reg_base_init(adev);
 		adev->sdma.num_instances = 5;
+		adev->sdma.sdma_mask = 0x1f;
 		adev->vcn.num_vcn_inst = 2;
 		adev->gmc.num_umc = 4;
+		adev->gfx.xcc_mask = 1;
 		adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 4, 2);
 		adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 4, 2);
 		adev->ip_versions[OSSSYS_HWIP][0] = IP_VERSION(4, 4, 0);
@@ -2165,10 +2827,44 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		adev->ip_versions[UVD_HWIP][1] = IP_VERSION(2, 6, 0);
 		adev->ip_versions[XGMI_HWIP][0] = IP_VERSION(6, 1, 0);
 		break;
+	case CHIP_CYAN_SKILLFISH:
+		if (adev->apu_flags & AMD_APU_IS_CYAN_SKILLFISH2) {
+			r = amdgpu_discovery_reg_base_init(adev);
+			if (r)
+				return -EINVAL;
+
+			amdgpu_discovery_harvest_ip(adev);
+			amdgpu_discovery_get_gfx_info(adev);
+			amdgpu_discovery_get_mall_info(adev);
+			amdgpu_discovery_get_vcn_info(adev);
+		} else {
+			cyan_skillfish_reg_base_init(adev);
+			adev->sdma.num_instances = 2;
+			adev->sdma.sdma_mask = 3;
+			adev->gfx.xcc_mask = 1;
+			adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(2, 0, 3);
+			adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(2, 0, 3);
+			adev->ip_versions[OSSSYS_HWIP][0] = IP_VERSION(5, 0, 1);
+			adev->ip_versions[HDP_HWIP][0] = IP_VERSION(5, 0, 1);
+			adev->ip_versions[SDMA0_HWIP][0] = IP_VERSION(5, 0, 1);
+			adev->ip_versions[SDMA1_HWIP][1] = IP_VERSION(5, 0, 1);
+			adev->ip_versions[DF_HWIP][0] = IP_VERSION(3, 5, 0);
+			adev->ip_versions[NBIO_HWIP][0] = IP_VERSION(2, 1, 1);
+			adev->ip_versions[UMC_HWIP][0] = IP_VERSION(8, 1, 1);
+			adev->ip_versions[MP0_HWIP][0] = IP_VERSION(11, 0, 8);
+			adev->ip_versions[MP1_HWIP][0] = IP_VERSION(11, 0, 8);
+			adev->ip_versions[THM_HWIP][0] = IP_VERSION(11, 0, 1);
+			adev->ip_versions[SMUIO_HWIP][0] = IP_VERSION(11, 0, 8);
+			adev->ip_versions[GC_HWIP][0] = IP_VERSION(10, 1, 3);
+			adev->ip_versions[UVD_HWIP][0] = IP_VERSION(2, 0, 3);
+		}
+		break;
 	default:
 		r = amdgpu_discovery_reg_base_init(adev);
-		if (r)
-			return -EINVAL;
+		if (r) {
+			drm_err(&adev->ddev, "discovery failed: %d\n", r);
+			return r;
+		}
 
 		amdgpu_discovery_harvest_ip(adev);
 		amdgpu_discovery_get_gfx_info(adev);
@@ -2177,13 +2873,18 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		break;
 	}
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	amdgpu_discovery_init_soc_config(adev);
+	amdgpu_discovery_sysfs_init(adev);
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 2, 1):
 	case IP_VERSION(9, 4, 0):
 	case IP_VERSION(9, 4, 1):
 	case IP_VERSION(9, 4, 2):
 	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+	case IP_VERSION(9, 5, 0):
 		adev->family = AMDGPU_FAMILY_AI;
 		break;
 	case IP_VERSION(9, 1, 0):
@@ -2224,11 +2925,21 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(11, 0, 4):
 		adev->family = AMDGPU_FAMILY_GC_11_0_1;
 		break;
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
+		adev->family = AMDGPU_FAMILY_GC_11_5_0;
+		break;
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		adev->family = AMDGPU_FAMILY_GC_12_0_0;
+		break;
 	default:
 		return -EINVAL;
 	}
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 1, 0):
 	case IP_VERSION(9, 2, 2):
 	case IP_VERSION(9, 3, 0):
@@ -2240,17 +2951,18 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(10, 3, 7):
 	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
 		adev->flags |= AMD_IS_APU;
 		break;
 	default:
 		break;
 	}
 
-	if (adev->ip_versions[XGMI_HWIP][0] == IP_VERSION(4, 8, 0))
-		adev->gmc.xgmi.supported = true;
-
 	/* set NBIO version */
-	switch (adev->ip_versions[NBIO_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, NBIO_HWIP, 0)) {
 	case IP_VERSION(6, 1, 0):
 	case IP_VERSION(6, 2, 0):
 		adev->nbio.funcs = &nbio_v6_1_funcs;
@@ -2269,9 +2981,17 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		adev->nbio.hdp_flush_reg = &nbio_v7_4_hdp_flush_reg;
 		break;
 	case IP_VERSION(7, 9, 0):
+	case IP_VERSION(7, 9, 1):
 		adev->nbio.funcs = &nbio_v7_9_funcs;
 		adev->nbio.hdp_flush_reg = &nbio_v7_9_hdp_flush_reg;
 		break;
+	case IP_VERSION(7, 11, 0):
+	case IP_VERSION(7, 11, 1):
+	case IP_VERSION(7, 11, 2):
+	case IP_VERSION(7, 11, 3):
+		adev->nbio.funcs = &nbio_v7_11_funcs;
+		adev->nbio.hdp_flush_reg = &nbio_v7_11_hdp_flush_reg;
+		break;
 	case IP_VERSION(7, 2, 0):
 	case IP_VERSION(7, 2, 1):
 	case IP_VERSION(7, 3, 0):
@@ -2304,11 +3024,15 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		adev->nbio.funcs = &nbio_v7_7_funcs;
 		adev->nbio.hdp_flush_reg = &nbio_v7_7_hdp_flush_reg;
 		break;
+	case IP_VERSION(6, 3, 1):
+		adev->nbio.funcs = &nbif_v6_3_1_funcs;
+		adev->nbio.hdp_flush_reg = &nbif_v6_3_1_hdp_flush_reg;
+		break;
 	default:
 		break;
 	}
 
-	switch (adev->ip_versions[HDP_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, HDP_HWIP, 0)) {
 	case IP_VERSION(4, 0, 0):
 	case IP_VERSION(4, 0, 1):
 	case IP_VERSION(4, 1, 0):
@@ -2318,6 +3042,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(4, 2, 1):
 	case IP_VERSION(4, 4, 0):
 	case IP_VERSION(4, 4, 2):
+	case IP_VERSION(4, 4, 5):
 		adev->hdp.funcs = &hdp_v4_0_funcs;
 		break;
 	case IP_VERSION(5, 0, 0):
@@ -2333,13 +3058,17 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		break;
 	case IP_VERSION(6, 0, 0):
 	case IP_VERSION(6, 0, 1):
+	case IP_VERSION(6, 1, 0):
 		adev->hdp.funcs = &hdp_v6_0_funcs;
 		break;
+	case IP_VERSION(7, 0, 0):
+		adev->hdp.funcs = &hdp_v7_0_funcs;
+		break;
 	default:
 		break;
 	}
 
-	switch (adev->ip_versions[DF_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, DF_HWIP, 0)) {
 	case IP_VERSION(3, 6, 0):
 	case IP_VERSION(3, 6, 1):
 	case IP_VERSION(3, 6, 2):
@@ -2355,11 +3084,18 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(4, 3, 0):
 		adev->df.funcs = &df_v4_3_funcs;
 		break;
+	case IP_VERSION(4, 6, 2):
+		adev->df.funcs = &df_v4_6_2_funcs;
+		break;
+	case IP_VERSION(4, 15, 0):
+	case IP_VERSION(4, 15, 1):
+		adev->df.funcs = &df_v4_15_funcs;
+		break;
 	default:
 		break;
 	}
 
-	switch (adev->ip_versions[SMUIO_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, SMUIO_HWIP, 0)) {
 	case IP_VERSION(9, 0, 0):
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(10, 0, 0):
@@ -2379,6 +3115,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(11, 0, 10):
 	case IP_VERSION(11, 0, 11):
 	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 2):
 	case IP_VERSION(13, 0, 1):
 	case IP_VERSION(13, 0, 9):
 	case IP_VERSION(13, 0, 10):
@@ -2387,21 +3124,37 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(13, 0, 2):
 		adev->smuio.funcs = &smuio_v13_0_funcs;
 		break;
+	case IP_VERSION(13, 0, 3):
+	case IP_VERSION(13, 0, 11):
+		adev->smuio.funcs = &smuio_v13_0_3_funcs;
+		if (adev->smuio.funcs->get_pkg_type(adev) == AMDGPU_PKG_TYPE_APU) {
+			adev->flags |= AMD_IS_APU;
+		}
+		break;
 	case IP_VERSION(13, 0, 6):
 	case IP_VERSION(13, 0, 8):
+	case IP_VERSION(14, 0, 0):
+	case IP_VERSION(14, 0, 1):
 		adev->smuio.funcs = &smuio_v13_0_6_funcs;
 		break;
+	case IP_VERSION(14, 0, 2):
+		adev->smuio.funcs = &smuio_v14_0_2_funcs;
+		break;
 	default:
 		break;
 	}
 
-	switch (adev->ip_versions[LSDMA_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, LSDMA_HWIP, 0)) {
 	case IP_VERSION(6, 0, 0):
 	case IP_VERSION(6, 0, 1):
 	case IP_VERSION(6, 0, 2):
 	case IP_VERSION(6, 0, 3):
 		adev->lsdma.funcs = &lsdma_v6_0_funcs;
 		break;
+	case IP_VERSION(7, 0, 0):
+	case IP_VERSION(7, 0, 1):
+		adev->lsdma.funcs = &lsdma_v7_0_funcs;
+		break;
 	default:
 		break;
 	}
@@ -2452,6 +3205,10 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 	if (r)
 		return r;
 
+	r = amdgpu_discovery_set_ras_ip_blocks(adev);
+	if (r)
+		return r;
+
 	if ((adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT &&
 	     !amdgpu_sriov_vf(adev)) ||
 	    (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO && amdgpu_dpm == 1)) {
@@ -2468,6 +3225,17 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 	if (r)
 		return r;
 
+	r = amdgpu_discovery_set_vpe_ip_blocks(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_discovery_set_umsch_mm_ip_blocks(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_discovery_set_isp_ip_blocks(adev);
+	if (r)
+		return r;
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h
index 8563dd4a7dc2..4ce04486cc31 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h
@@ -24,12 +24,27 @@
 #ifndef __AMDGPU_DISCOVERY__
 #define __AMDGPU_DISCOVERY__
 
-#define DISCOVERY_TMR_SIZE      (4 << 10)
+#include <linux/debugfs.h>
+
+#define DISCOVERY_TMR_SIZE      (10 << 10)
 #define DISCOVERY_TMR_OFFSET    (64 << 10)
 
+struct ip_discovery_top;
+
+struct amdgpu_discovery_info {
+	struct debugfs_blob_wrapper debugfs_blob;
+	struct ip_discovery_top *ip_top;
+	uint32_t size;
+	uint8_t *bin;
+	bool reserve_tmr;
+};
+
 void amdgpu_discovery_fini(struct amdgpu_device *adev);
-int amdgpu_discovery_get_ip_version(struct amdgpu_device *adev, int hw_id, int number_instance,
-                                    int *major, int *minor, int *revision);
 int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev);
 
+int amdgpu_discovery_get_nps_info(struct amdgpu_device *adev,
+				  uint32_t *nps_type,
+				  struct amdgpu_gmc_memrange **ranges,
+				  int *range_cnt, bool refresh);
+
 #endif /* __AMDGPU_DISCOVERY__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c
index d60fe7eb5579..b5d34797d606 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c
@@ -33,11 +33,14 @@
 #include "soc15_common.h"
 #include "gc/gc_11_0_0_offset.h"
 #include "gc/gc_11_0_0_sh_mask.h"
+#include "bif/bif_4_1_d.h"
 #include <asm/div64.h>
 
 #include <linux/pci.h>
 #include <linux/pm_runtime.h>
 #include <drm/drm_crtc_helper.h>
+#include <drm/drm_damage_helper.h>
+#include <drm/drm_drv.h>
 #include <drm/drm_edid.h>
 #include <drm/drm_fb_helper.h>
 #include <drm/drm_gem_framebuffer_helper.h>
@@ -98,7 +101,7 @@ static void amdgpu_display_flip_callback(struct dma_fence *f,
 static bool amdgpu_display_flip_handle_fence(struct amdgpu_flip_work *work,
 					     struct dma_fence **f)
 {
-	struct dma_fence *fence= *f;
+	struct dma_fence *fence = *f;
 
 	if (fence == NULL)
 		return false;
@@ -124,7 +127,7 @@ static void amdgpu_display_flip_work_func(struct work_struct *__work)
 
 	struct drm_crtc *crtc = &amdgpu_crtc->base;
 	unsigned long flags;
-	unsigned i;
+	unsigned int i;
 	int vpos, hpos;
 
 	for (i = 0; i < work->shared_count; ++i)
@@ -201,7 +204,7 @@ int amdgpu_display_crtc_page_flip_target(struct drm_crtc *crtc,
 	u64 tiling_flags;
 	int i, r;
 
-	work = kzalloc(sizeof *work, GFP_KERNEL);
+	work = kzalloc(sizeof(*work), GFP_KERNEL);
 	if (work == NULL)
 		return -ENOMEM;
 
@@ -231,6 +234,7 @@ int amdgpu_display_crtc_page_flip_target(struct drm_crtc *crtc,
 	}
 
 	if (!adev->enable_virtual_display) {
+		new_abo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 		r = amdgpu_bo_pin(new_abo,
 				  amdgpu_display_supported_domains(adev, new_abo->flags));
 		if (unlikely(r != 0)) {
@@ -328,22 +332,19 @@ int amdgpu_display_crtc_set_config(struct drm_mode_set *set,
 		if (crtc->enabled)
 			active = true;
 
-	pm_runtime_mark_last_busy(dev->dev);
-
 	adev = drm_to_adev(dev);
 	/* if we have active crtcs and we don't have a power ref,
-	   take the current one */
+	 * take the current one
+	 */
 	if (active && !adev->have_disp_power_ref) {
 		adev->have_disp_power_ref = true;
 		return ret;
 	}
-	/* if we have no active crtcs, then drop the power ref
-	   we got before */
-	if (!active && adev->have_disp_power_ref) {
-		pm_runtime_put_autosuspend(dev->dev);
+	/* if we have no active crtcs, then go to
+	 * drop the power ref we got before
+	 */
+	if (!active && adev->have_disp_power_ref)
 		adev->have_disp_power_ref = false;
-	}
-
 out:
 	/* drop the power reference we got coming in here */
 	pm_runtime_put_autosuspend(dev->dev);
@@ -507,11 +508,10 @@ bool amdgpu_display_ddc_probe(struct amdgpu_connector *amdgpu_connector,
 	if (amdgpu_connector->router.ddc_valid)
 		amdgpu_i2c_router_select_ddc_port(amdgpu_connector);
 
-	if (use_aux) {
+	if (use_aux)
 		ret = i2c_transfer(&amdgpu_connector->ddc_bus->aux.ddc, msgs, 2);
-	} else {
+	else
 		ret = i2c_transfer(&amdgpu_connector->ddc_bus->adapter, msgs, 2);
-	}
 
 	if (ret != 2)
 		/* Couldn't find an accessible DDC on this connector */
@@ -520,20 +520,40 @@ bool amdgpu_display_ddc_probe(struct amdgpu_connector *amdgpu_connector,
 	 * EDID header starts with:
 	 * 0x00,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0x00.
 	 * Only the first 6 bytes must be valid as
-	 * drm_edid_block_valid() can fix the last 2 bytes */
+	 * drm_edid_block_valid() can fix the last 2 bytes
+	 */
 	if (drm_edid_header_is_valid(buf) < 6) {
 		/* Couldn't find an accessible EDID on this
-		 * connector */
+		 * connector
+		 */
 		return false;
 	}
 	return true;
 }
 
+static int amdgpu_dirtyfb(struct drm_framebuffer *fb, struct drm_file *file,
+			  unsigned int flags, unsigned int color,
+			  struct drm_clip_rect *clips, unsigned int num_clips)
+{
+
+	if (file)
+		return -ENOSYS;
+
+	return drm_atomic_helper_dirtyfb(fb, file, flags, color, clips,
+					 num_clips);
+}
+
 static const struct drm_framebuffer_funcs amdgpu_fb_funcs = {
 	.destroy = drm_gem_fb_destroy,
 	.create_handle = drm_gem_fb_create_handle,
 };
 
+static const struct drm_framebuffer_funcs amdgpu_fb_funcs_atomic = {
+	.destroy = drm_gem_fb_destroy,
+	.create_handle = drm_gem_fb_create_handle,
+	.dirty = amdgpu_dirtyfb
+};
+
 uint32_t amdgpu_display_supported_domains(struct amdgpu_device *adev,
 					  uint64_t bo_flags)
 {
@@ -634,6 +654,10 @@ amdgpu_lookup_format_info(u32 format, uint64_t modifier)
 	if (!IS_AMD_FMT_MOD(modifier))
 		return NULL;
 
+	if (AMD_FMT_MOD_GET(TILE_VERSION, modifier) < AMD_FMT_MOD_TILE_VER_GFX9 ||
+	    AMD_FMT_MOD_GET(TILE_VERSION, modifier) >= AMD_FMT_MOD_TILE_VER_GFX12)
+		return NULL;
+
 	if (AMD_FMT_MOD_GET(DCC_RETILE, modifier))
 		return lookup_format_info(dcc_retile_formats,
 					  ARRAY_SIZE(dcc_retile_formats),
@@ -698,6 +722,30 @@ extract_render_dcc_offset(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int convert_tiling_flags_to_modifier_gfx12(struct amdgpu_framebuffer *afb)
+{
+	u64 modifier = 0;
+	int swizzle_mode = AMDGPU_TILING_GET(afb->tiling_flags, GFX12_SWIZZLE_MODE);
+
+	if (!swizzle_mode) {
+		modifier = DRM_FORMAT_MOD_LINEAR;
+	} else {
+		int max_comp_block =
+			AMDGPU_TILING_GET(afb->tiling_flags, GFX12_DCC_MAX_COMPRESSED_BLOCK);
+
+		modifier =
+			AMD_FMT_MOD |
+			AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX12) |
+			AMD_FMT_MOD_SET(TILE, swizzle_mode) |
+			AMD_FMT_MOD_SET(DCC, afb->gfx12_dcc) |
+			AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, max_comp_block);
+	}
+
+	afb->base.modifier = modifier;
+	afb->base.flags |= DRM_MODE_FB_MODIFIERS;
+	return 0;
+}
+
 static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb)
 {
 	struct amdgpu_device *adev = drm_to_adev(afb->base.dev);
@@ -743,11 +791,13 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb)
 			return -EINVAL;
 		}
 
-		if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(11, 0, 0))
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(11, 0, 0))
 			version = AMD_FMT_MOD_TILE_VER_GFX11;
-		else if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0))
+		else if (amdgpu_ip_version(adev, GC_HWIP, 0) >=
+			 IP_VERSION(10, 3, 0))
 			version = AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS;
-		else if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 0, 0))
+		else if (amdgpu_ip_version(adev, GC_HWIP, 0) >=
+			 IP_VERSION(10, 0, 0))
 			version = AMD_FMT_MOD_TILE_VER_GFX10;
 		else
 			version = AMD_FMT_MOD_TILE_VER_GFX9;
@@ -756,13 +806,15 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb)
 		case 0: /* Z microtiling */
 			return -EINVAL;
 		case 1: /* S microtiling */
-			if (adev->ip_versions[GC_HWIP][0] < IP_VERSION(11, 0, 0)) {
+			if (amdgpu_ip_version(adev, GC_HWIP, 0) <
+			    IP_VERSION(11, 0, 0)) {
 				if (!has_xor)
 					version = AMD_FMT_MOD_TILE_VER_GFX9;
 			}
 			break;
 		case 2:
-			if (adev->ip_versions[GC_HWIP][0] < IP_VERSION(11, 0, 0)) {
+			if (amdgpu_ip_version(adev, GC_HWIP, 0) <
+			    IP_VERSION(11, 0, 0)) {
 				if (!has_xor && afb->base.format->cpp[0] != 4)
 					version = AMD_FMT_MOD_TILE_VER_GFX9;
 			}
@@ -815,10 +867,12 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb)
 			u64 render_dcc_offset;
 
 			/* Enable constant encode on RAVEN2 and later. */
-			bool dcc_constant_encode = (adev->asic_type > CHIP_RAVEN ||
-						   (adev->asic_type == CHIP_RAVEN &&
-						    adev->external_rev_id >= 0x81)) &&
-						    adev->ip_versions[GC_HWIP][0] < IP_VERSION(11, 0, 0);
+			bool dcc_constant_encode =
+				(adev->asic_type > CHIP_RAVEN ||
+				 (adev->asic_type == CHIP_RAVEN &&
+				  adev->external_rev_id >= 0x81)) &&
+				amdgpu_ip_version(adev, GC_HWIP, 0) <
+					IP_VERSION(11, 0, 0);
 
 			int max_cblock_size = dcc_i64b ? AMD_FMT_MOD_DCC_BLOCK_64B :
 					      dcc_i128b ? AMD_FMT_MOD_DCC_BLOCK_128B :
@@ -855,7 +909,9 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb)
 				if (adev->family >= AMDGPU_FAMILY_NV) {
 					int extra_pipe = 0;
 
-					if ((adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0)) &&
+					if ((amdgpu_ip_version(adev, GC_HWIP,
+							       0) >=
+					     IP_VERSION(10, 3, 0)) &&
 					    pipes == packers && pipes > 1)
 						extra_pipe = 1;
 
@@ -889,8 +945,7 @@ static int check_tiling_flags_gfx6(struct amdgpu_framebuffer *afb)
 {
 	u64 micro_tile_mode;
 
-	/* Zero swizzle mode means linear */
-	if (AMDGPU_TILING_GET(afb->tiling_flags, SWIZZLE_MODE) == 0)
+	if (AMDGPU_TILING_GET(afb->tiling_flags, ARRAY_MODE) == 1) /* LINEAR_ALIGNED */
 		return 0;
 
 	micro_tile_mode = AMDGPU_TILING_GET(afb->tiling_flags, MICRO_TILE_MODE);
@@ -1014,6 +1069,30 @@ static int amdgpu_display_verify_sizes(struct amdgpu_framebuffer *rfb)
 			block_width = 256 / format_info->cpp[i];
 			block_height = 1;
 			block_size_log2 = 8;
+		} else if (AMD_FMT_MOD_GET(TILE_VERSION, modifier) >= AMD_FMT_MOD_TILE_VER_GFX12) {
+			int swizzle = AMD_FMT_MOD_GET(TILE, modifier);
+
+			switch (swizzle) {
+			case AMD_FMT_MOD_TILE_GFX12_256B_2D:
+				block_size_log2 = 8;
+				break;
+			case AMD_FMT_MOD_TILE_GFX12_4K_2D:
+				block_size_log2 = 12;
+				break;
+			case AMD_FMT_MOD_TILE_GFX12_64K_2D:
+				block_size_log2 = 16;
+				break;
+			case AMD_FMT_MOD_TILE_GFX12_256K_2D:
+				block_size_log2 = 18;
+				break;
+			default:
+				drm_dbg_kms(rfb->base.dev,
+					    "Gfx12 swizzle mode with unknown block size: %d\n", swizzle);
+				return -EINVAL;
+			}
+
+			get_block_dimensions(block_size_log2, format_info->cpp[i],
+					     &block_width, &block_height);
 		} else {
 			int swizzle = AMD_FMT_MOD_GET(TILE, modifier);
 
@@ -1049,7 +1128,8 @@ static int amdgpu_display_verify_sizes(struct amdgpu_framebuffer *rfb)
 			return ret;
 	}
 
-	if (AMD_FMT_MOD_GET(DCC, modifier)) {
+	if (AMD_FMT_MOD_GET(TILE_VERSION, modifier) <= AMD_FMT_MOD_TILE_VER_GFX11 &&
+	    AMD_FMT_MOD_GET(DCC, modifier)) {
 		if (AMD_FMT_MOD_GET(DCC_RETILE, modifier)) {
 			block_size_log2 = get_dcc_block_size(modifier, false, false);
 			get_block_dimensions(block_size_log2 + 8, format_info->cpp[0],
@@ -1079,7 +1159,8 @@ static int amdgpu_display_verify_sizes(struct amdgpu_framebuffer *rfb)
 }
 
 static int amdgpu_display_get_fb_info(const struct amdgpu_framebuffer *amdgpu_fb,
-				      uint64_t *tiling_flags, bool *tmz_surface)
+				      uint64_t *tiling_flags, bool *tmz_surface,
+				      bool *gfx12_dcc)
 {
 	struct amdgpu_bo *rbo;
 	int r;
@@ -1087,6 +1168,7 @@ static int amdgpu_display_get_fb_info(const struct amdgpu_framebuffer *amdgpu_fb
 	if (!amdgpu_fb) {
 		*tiling_flags = 0;
 		*tmz_surface = false;
+		*gfx12_dcc = false;
 		return 0;
 	}
 
@@ -1100,11 +1182,9 @@ static int amdgpu_display_get_fb_info(const struct amdgpu_framebuffer *amdgpu_fb
 		return r;
 	}
 
-	if (tiling_flags)
-		amdgpu_bo_get_tiling_flags(rbo, tiling_flags);
-
-	if (tmz_surface)
-		*tmz_surface = amdgpu_bo_encrypted(rbo);
+	amdgpu_bo_get_tiling_flags(rbo, tiling_flags);
+	*tmz_surface = amdgpu_bo_encrypted(rbo);
+	*gfx12_dcc = rbo->flags & AMDGPU_GEM_CREATE_GFX12_DCC;
 
 	amdgpu_bo_unreserve(rbo);
 
@@ -1114,13 +1194,14 @@ static int amdgpu_display_get_fb_info(const struct amdgpu_framebuffer *amdgpu_fb
 static int amdgpu_display_gem_fb_verify_and_init(struct drm_device *dev,
 						 struct amdgpu_framebuffer *rfb,
 						 struct drm_file *file_priv,
+						 const struct drm_format_info *info,
 						 const struct drm_mode_fb_cmd2 *mode_cmd,
 						 struct drm_gem_object *obj)
 {
 	int ret;
 
 	rfb->base.obj[0] = obj;
-	drm_helper_mode_fill_fb_struct(dev, &rfb->base, mode_cmd);
+	drm_helper_mode_fill_fb_struct(dev, &rfb->base, info, mode_cmd);
 	/* Verify that the modifier is supported. */
 	if (!drm_any_plane_has_format(dev, mode_cmd->pixel_format,
 				      mode_cmd->modifier[0])) {
@@ -1136,7 +1217,11 @@ static int amdgpu_display_gem_fb_verify_and_init(struct drm_device *dev,
 	if (ret)
 		goto err;
 
-	ret = drm_framebuffer_init(dev, &rfb->base, &amdgpu_fb_funcs);
+	if (drm_drv_uses_atomic_modeset(dev))
+		ret = drm_framebuffer_init(dev, &rfb->base,
+					   &amdgpu_fb_funcs_atomic);
+	else
+		ret = drm_framebuffer_init(dev, &rfb->base, &amdgpu_fb_funcs);
 
 	if (ret)
 		goto err;
@@ -1169,7 +1254,8 @@ static int amdgpu_display_framebuffer_init(struct drm_device *dev,
 		}
 	}
 
-	ret = amdgpu_display_get_fb_info(rfb, &rfb->tiling_flags, &rfb->tmz_surface);
+	ret = amdgpu_display_get_fb_info(rfb, &rfb->tiling_flags, &rfb->tmz_surface,
+					 &rfb->gfx12_dcc);
 	if (ret)
 		return ret;
 
@@ -1183,7 +1269,11 @@ static int amdgpu_display_framebuffer_init(struct drm_device *dev,
 
 	if (!dev->mode_config.fb_modifiers_not_supported &&
 	    !(rfb->base.flags & DRM_MODE_FB_MODIFIERS)) {
-		ret = convert_tiling_flags_to_modifier(rfb);
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(12, 0, 0))
+			ret = convert_tiling_flags_to_modifier_gfx12(rfb);
+		else
+			ret = convert_tiling_flags_to_modifier(rfb);
+
 		if (ret) {
 			drm_dbg_kms(dev, "Failed to convert tiling flags 0x%llX to a modifier",
 				    rfb->tiling_flags);
@@ -1206,6 +1296,7 @@ static int amdgpu_display_framebuffer_init(struct drm_device *dev,
 struct drm_framebuffer *
 amdgpu_display_user_framebuffer_create(struct drm_device *dev,
 				       struct drm_file *file_priv,
+				       const struct drm_format_info *info,
 				       const struct drm_mode_fb_cmd2 *mode_cmd)
 {
 	struct amdgpu_framebuffer *amdgpu_fb;
@@ -1216,15 +1307,17 @@ amdgpu_display_user_framebuffer_create(struct drm_device *dev,
 
 	obj = drm_gem_object_lookup(file_priv, mode_cmd->handles[0]);
 	if (obj ==  NULL) {
-		drm_dbg_kms(dev, "No GEM object associated to handle 0x%08X, "
-			    "can't create framebuffer\n", mode_cmd->handles[0]);
+		drm_dbg_kms(dev,
+			    "No GEM object associated to handle 0x%08X, can't create framebuffer\n",
+			    mode_cmd->handles[0]);
+
 		return ERR_PTR(-ENOENT);
 	}
 
 	/* Handle is imported dma-buf, so cannot be migrated to VRAM for scanout */
 	bo = gem_to_amdgpu_bo(obj);
 	domains = amdgpu_display_supported_domains(drm_to_adev(dev), bo->flags);
-	if (obj->import_attach && !(domains & AMDGPU_GEM_DOMAIN_GTT)) {
+	if (drm_gem_is_imported(obj) && !(domains & AMDGPU_GEM_DOMAIN_GTT)) {
 		drm_dbg_kms(dev, "Cannot create framebuffer from imported dma_buf\n");
 		drm_gem_object_put(obj);
 		return ERR_PTR(-EINVAL);
@@ -1237,7 +1330,7 @@ amdgpu_display_user_framebuffer_create(struct drm_device *dev,
 	}
 
 	ret = amdgpu_display_gem_fb_verify_and_init(dev, amdgpu_fb, file_priv,
-						    mode_cmd, obj);
+						    info, mode_cmd, obj);
 	if (ret) {
 		kfree(amdgpu_fb);
 		drm_gem_object_put(obj);
@@ -1252,24 +1345,82 @@ const struct drm_mode_config_funcs amdgpu_mode_funcs = {
 	.fb_create = amdgpu_display_user_framebuffer_create,
 };
 
-static const struct drm_prop_enum_list amdgpu_underscan_enum_list[] =
-{	{ UNDERSCAN_OFF, "off" },
+static const struct drm_prop_enum_list amdgpu_underscan_enum_list[] = {
+	{ UNDERSCAN_OFF, "off" },
 	{ UNDERSCAN_ON, "on" },
 	{ UNDERSCAN_AUTO, "auto" },
 };
 
-static const struct drm_prop_enum_list amdgpu_audio_enum_list[] =
-{	{ AMDGPU_AUDIO_DISABLE, "off" },
+static const struct drm_prop_enum_list amdgpu_audio_enum_list[] = {
+	{ AMDGPU_AUDIO_DISABLE, "off" },
 	{ AMDGPU_AUDIO_ENABLE, "on" },
 	{ AMDGPU_AUDIO_AUTO, "auto" },
 };
 
 /* XXX support different dither options? spatial, temporal, both, etc. */
-static const struct drm_prop_enum_list amdgpu_dither_enum_list[] =
-{	{ AMDGPU_FMT_DITHER_DISABLE, "off" },
+static const struct drm_prop_enum_list amdgpu_dither_enum_list[] = {
+	{ AMDGPU_FMT_DITHER_DISABLE, "off" },
 	{ AMDGPU_FMT_DITHER_ENABLE, "on" },
 };
 
+/**
+ * DOC: property for adaptive backlight modulation
+ *
+ * The 'adaptive backlight modulation' property is used for the compositor to
+ * directly control the adaptive backlight modulation power savings feature
+ * that is part of DCN hardware.
+ *
+ * The property will be attached specifically to eDP panels that support it.
+ *
+ * The property is by default set to 'sysfs' to allow the sysfs file 'panel_power_savings'
+ * to be able to control it.
+ * If set to 'off' the compositor will ensure it stays off.
+ * The other values 'min', 'bias min', 'bias max', and 'max' will control the
+ * intensity of the power savings.
+ *
+ * Modifying this value can have implications on color accuracy, so tread
+ * carefully.
+ */
+static int amdgpu_display_setup_abm_prop(struct amdgpu_device *adev)
+{
+	const struct drm_prop_enum_list props[] = {
+		{ ABM_SYSFS_CONTROL, "sysfs" },
+		{ ABM_LEVEL_OFF, "off" },
+		{ ABM_LEVEL_MIN, "min" },
+		{ ABM_LEVEL_BIAS_MIN, "bias min" },
+		{ ABM_LEVEL_BIAS_MAX, "bias max" },
+		{ ABM_LEVEL_MAX, "max" },
+	};
+	struct drm_property *prop;
+	int i;
+
+	if (!adev->dc_enabled)
+		return 0;
+
+	prop = drm_property_create(adev_to_drm(adev), DRM_MODE_PROP_ENUM,
+				"adaptive backlight modulation",
+				6);
+	if (!prop)
+		return -ENOMEM;
+
+	for (i = 0; i < ARRAY_SIZE(props); i++) {
+		int ret;
+
+		ret = drm_property_add_enum(prop, props[i].type,
+						props[i].name);
+
+		if (ret) {
+			drm_property_destroy(adev_to_drm(adev), prop);
+
+			return ret;
+		}
+	}
+
+	adev->mode_info.abm_level_property = prop;
+
+	return 0;
+}
+
 int amdgpu_display_modeset_create_props(struct amdgpu_device *adev)
 {
 	int sz;
@@ -1316,15 +1467,7 @@ int amdgpu_display_modeset_create_props(struct amdgpu_device *adev)
 					 "dither",
 					 amdgpu_dither_enum_list, sz);
 
-	if (adev->dc_enabled) {
-		adev->mode_info.abm_level_property =
-			drm_property_create_range(adev_to_drm(adev), 0,
-						  "abm level", 0, 4);
-		if (!adev->mode_info.abm_level_property)
-			return -ENOMEM;
-	}
-
-	return 0;
+	return amdgpu_display_setup_abm_prop(adev);
 }
 
 void amdgpu_display_update_priority(struct amdgpu_device *adev)
@@ -1391,7 +1534,7 @@ bool amdgpu_display_crtc_scaling_mode_fixup(struct drm_crtc *crtc,
 		if ((!(mode->flags & DRM_MODE_FLAG_INTERLACE)) &&
 		    ((amdgpu_encoder->underscan_type == UNDERSCAN_ON) ||
 		     ((amdgpu_encoder->underscan_type == UNDERSCAN_AUTO) &&
-		      connector->display_info.is_hdmi &&
+		      connector && connector->display_info.is_hdmi &&
 		      amdgpu_display_is_hdtv_mode(mode)))) {
 			if (amdgpu_encoder->underscan_hborder != 0)
 				amdgpu_crtc->h_border = amdgpu_encoder->underscan_hborder;
@@ -1410,6 +1553,7 @@ bool amdgpu_display_crtc_scaling_mode_fixup(struct drm_crtc *crtc,
 	}
 	if (amdgpu_crtc->rmx_type != RMX_OFF) {
 		fixed20_12 a, b;
+
 		a.full = dfixed_const(src_v);
 		b.full = dfixed_const(dst_v);
 		amdgpu_crtc->vsc.full = dfixed_div(a, b);
@@ -1429,7 +1573,7 @@ bool amdgpu_display_crtc_scaling_mode_fixup(struct drm_crtc *crtc,
  *
  * \param dev Device to query.
  * \param pipe Crtc to query.
- * \param flags Flags from caller (DRM_CALLED_FROM_VBLIRQ or 0).
+ * \param flags from caller (DRM_CALLED_FROM_VBLIRQ or 0).
  *              For driver internal use only also supports these flags:
  *
  *              USE_REAL_VBLANKSTART to use the real start of vblank instead
@@ -1496,8 +1640,7 @@ int amdgpu_display_get_crtc_scanoutpos(struct drm_device *dev,
 		ret |= DRM_SCANOUTPOS_ACCURATE;
 		vbl_start = vbl & 0x1fff;
 		vbl_end = (vbl >> 16) & 0x1fff;
-	}
-	else {
+	} else {
 		/* No: Fake something reasonable which gives at least ok results. */
 		vbl_start = mode->crtc_vdisplay;
 		vbl_end = 0;
@@ -1505,8 +1648,8 @@ int amdgpu_display_get_crtc_scanoutpos(struct drm_device *dev,
 
 	/* Called from driver internal vblank counter query code? */
 	if (flags & GET_DISTANCE_TO_VBLANKSTART) {
-	    /* Caller wants distance from real vbl_start in *hpos */
-	    *hpos = *vpos - vbl_start;
+		/* Caller wants distance from real vbl_start in *hpos */
+		*hpos = *vpos - vbl_start;
 	}
 
 	/* Fudge vblank to start a few scanlines earlier to handle the
@@ -1528,7 +1671,7 @@ int amdgpu_display_get_crtc_scanoutpos(struct drm_device *dev,
 
 	/* In vblank? */
 	if (in_vbl)
-	    ret |= DRM_SCANOUTPOS_IN_VBLANK;
+		ret |= DRM_SCANOUTPOS_IN_VBLANK;
 
 	/* Called from driver internal vblank counter query code? */
 	if (flags & GET_DISTANCE_TO_VBLANKSTART) {
@@ -1636,6 +1779,7 @@ int amdgpu_display_suspend_helper(struct amdgpu_device *adev)
 
 		if (amdgpu_crtc->cursor_bo && !adev->enable_virtual_display) {
 			struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo);
+
 			r = amdgpu_bo_reserve(aobj, true);
 			if (r == 0) {
 				amdgpu_bo_unpin(aobj);
@@ -1643,9 +1787,9 @@ int amdgpu_display_suspend_helper(struct amdgpu_device *adev)
 			}
 		}
 
-		if (fb == NULL || fb->obj[0] == NULL) {
+		if (!fb || !fb->obj[0])
 			continue;
-		}
+
 		robj = gem_to_amdgpu_bo(fb->obj[0]);
 		if (!amdgpu_display_robj_is_fb(adev, robj)) {
 			r = amdgpu_bo_reserve(robj, true);
@@ -1672,8 +1816,10 @@ int amdgpu_display_resume_helper(struct amdgpu_device *adev)
 
 		if (amdgpu_crtc->cursor_bo && !adev->enable_virtual_display) {
 			struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo);
+
 			r = amdgpu_bo_reserve(aobj, true);
 			if (r == 0) {
+				aobj->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 				r = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM);
 				if (r != 0)
 					dev_err(adev->dev, "Failed to pin cursor BO (%d)\n", r);
@@ -1701,3 +1847,82 @@ int amdgpu_display_resume_helper(struct amdgpu_device *adev)
 	return 0;
 }
 
+/* panic_bo is set in amdgpu_dm_plane_get_scanout_buffer() and only used in amdgpu_dm_set_pixel()
+ * they are called from the panic handler, and protected by the drm_panic spinlock.
+ */
+static struct amdgpu_bo *panic_abo;
+
+/* Use the indirect MMIO to write each pixel to the GPU VRAM,
+ * This is a simplified version of amdgpu_device_mm_access()
+ */
+static void amdgpu_display_set_pixel(struct drm_scanout_buffer *sb,
+				     unsigned int x,
+				     unsigned int y,
+				     u32 color)
+{
+	struct amdgpu_res_cursor cursor;
+	unsigned long offset;
+	struct amdgpu_bo *abo = panic_abo;
+	struct amdgpu_device *adev = amdgpu_ttm_adev(abo->tbo.bdev);
+	uint32_t tmp;
+
+	offset = x * 4 + y * sb->pitch[0];
+	amdgpu_res_first(abo->tbo.resource, offset, 4, &cursor);
+
+	tmp = cursor.start >> 31;
+	WREG32_NO_KIQ(mmMM_INDEX, ((uint32_t) cursor.start) | 0x80000000);
+	if (tmp != 0xffffffff)
+		WREG32_NO_KIQ(mmMM_INDEX_HI, tmp);
+	WREG32_NO_KIQ(mmMM_DATA, color);
+}
+
+int amdgpu_display_get_scanout_buffer(struct drm_plane *plane,
+				      struct drm_scanout_buffer *sb)
+{
+	struct amdgpu_bo *abo;
+	struct drm_framebuffer *fb = plane->state->fb;
+
+	if (!fb)
+		return -EINVAL;
+
+	DRM_DEBUG_KMS("Framebuffer %dx%d %p4cc\n", fb->width, fb->height, &fb->format->format);
+
+	abo = gem_to_amdgpu_bo(fb->obj[0]);
+	if (!abo)
+		return -EINVAL;
+
+	sb->width = fb->width;
+	sb->height = fb->height;
+	/* Use the generic linear format, because tiling will be disabled in panic_flush() */
+	sb->format = drm_format_info(fb->format->format);
+	if (!sb->format)
+		return -EINVAL;
+
+	sb->pitch[0] = fb->pitches[0];
+
+	if (abo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) {
+		if (abo->tbo.resource->mem_type != TTM_PL_VRAM) {
+			drm_warn(plane->dev, "amdgpu panic, framebuffer not in VRAM\n");
+			return -EINVAL;
+		}
+		/* Only handle 32bits format, to simplify mmio access */
+		if (fb->format->cpp[0] != 4) {
+			drm_warn(plane->dev, "amdgpu panic, pixel format is not 32bits\n");
+			return -EINVAL;
+		}
+		sb->set_pixel = amdgpu_display_set_pixel;
+		panic_abo = abo;
+		return 0;
+	}
+	if (!abo->kmap.virtual &&
+	    ttm_bo_kmap(&abo->tbo, 0, PFN_UP(abo->tbo.base.size), &abo->kmap)) {
+		drm_warn(plane->dev, "amdgpu bo map failed, panic won't be displayed\n");
+		return -ENOMEM;
+	}
+	if (abo->kmap.bo_kmap_type & TTM_BO_MAP_IOMEM_MASK)
+		iosys_map_set_vaddr_iomem(&sb->map[0], abo->kmap.virtual);
+	else
+		iosys_map_set_vaddr(&sb->map[0], abo->kmap.virtual);
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h
index 9d19940f73c8..49a29bf47a37 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h
@@ -23,6 +23,8 @@
 #ifndef __AMDGPU_DISPLAY_H__
 #define __AMDGPU_DISPLAY_H__
 
+#include <drm/drm_panic.h>
+
 #define amdgpu_display_vblank_get_counter(adev, crtc) (adev)->mode_info.funcs->vblank_get_counter((adev), (crtc))
 #define amdgpu_display_backlight_set_level(adev, e, l) (adev)->mode_info.funcs->backlight_set_level((e), (l))
 #define amdgpu_display_backlight_get_level(adev, e) (adev)->mode_info.funcs->backlight_get_level((e))
@@ -42,6 +44,7 @@ uint32_t amdgpu_display_supported_domains(struct amdgpu_device *adev,
 struct drm_framebuffer *
 amdgpu_display_user_framebuffer_create(struct drm_device *dev,
 				       struct drm_file *file_priv,
+				       const struct drm_format_info *info,
 				       const struct drm_mode_fb_cmd2 *mode_cmd);
 const struct drm_format_info *
 amdgpu_lookup_format_info(u32 format, uint64_t modifier);
@@ -49,4 +52,14 @@ amdgpu_lookup_format_info(u32 format, uint64_t modifier);
 int amdgpu_display_suspend_helper(struct amdgpu_device *adev);
 int amdgpu_display_resume_helper(struct amdgpu_device *adev);
 
+int amdgpu_display_get_scanout_buffer(struct drm_plane *plane,
+				      struct drm_scanout_buffer *sb);
+
+#define ABM_SYSFS_CONTROL	-1
+#define ABM_LEVEL_OFF		0
+#define ABM_LEVEL_MIN		1
+#define ABM_LEVEL_BIAS_MIN	2
+#define ABM_LEVEL_BIAS_MAX	3
+#define ABM_LEVEL_MAX		4
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c
index 0c001bb8fc2b..e22cfa7c6d32 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c
@@ -36,12 +36,35 @@
 #include "amdgpu_gem.h"
 #include "amdgpu_dma_buf.h"
 #include "amdgpu_xgmi.h"
+#include "amdgpu_vm.h"
 #include <drm/amdgpu_drm.h>
 #include <drm/ttm/ttm_tt.h>
 #include <linux/dma-buf.h>
 #include <linux/dma-fence-array.h>
 #include <linux/pci-p2pdma.h>
-#include <linux/pm_runtime.h>
+
+static const struct dma_buf_attach_ops amdgpu_dma_buf_attach_ops;
+
+/**
+ * dma_buf_attach_adev - Helper to get adev of an attachment
+ *
+ * @attach: attachment
+ *
+ * Returns:
+ * A struct amdgpu_device * if the attaching device is an amdgpu device or
+ * partition, NULL otherwise.
+ */
+static struct amdgpu_device *dma_buf_attach_adev(struct dma_buf_attachment *attach)
+{
+	if (attach->importer_ops == &amdgpu_dma_buf_attach_ops) {
+		struct drm_gem_object *obj = attach->importer_priv;
+		struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
+
+		return amdgpu_ttm_adev(bo->tbo.bdev);
+	}
+
+	return NULL;
+}
 
 /**
  * amdgpu_dma_buf_attach - &dma_buf_ops.attach implementation
@@ -54,42 +77,49 @@
 static int amdgpu_dma_buf_attach(struct dma_buf *dmabuf,
 				 struct dma_buf_attachment *attach)
 {
+	struct amdgpu_device *attach_adev = dma_buf_attach_adev(attach);
 	struct drm_gem_object *obj = dmabuf->priv;
 	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
 	int r;
 
-	if (pci_p2pdma_distance(adev->pdev, attach->dev, false) < 0)
+	/*
+	 * Disable peer-to-peer access for DCC-enabled VRAM surfaces on GFX12+.
+	 * Such buffers cannot be safely accessed over P2P due to device-local
+	 * compression metadata. Fallback to system-memory path instead.
+	 * Device supports GFX12 (GC 12.x or newer)
+	 * BO was created with the AMDGPU_GEM_CREATE_GFX12_DCC flag
+	 *
+	 */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(12, 0, 0) &&
+	    bo->flags & AMDGPU_GEM_CREATE_GFX12_DCC)
 		attach->peer2peer = false;
 
-	r = pm_runtime_get_sync(adev_to_drm(adev)->dev);
-	if (r < 0)
-		goto out;
+	/*
+	 * Disable peer-to-peer access for DCC-enabled VRAM surfaces on GFX12+.
+	 * Such buffers cannot be safely accessed over P2P due to device-local
+	 * compression metadata. Fallback to system-memory path instead.
+	 * Device supports GFX12 (GC 12.x or newer)
+	 * BO was created with the AMDGPU_GEM_CREATE_GFX12_DCC flag
+	 *
+	 */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(12, 0, 0) &&
+	    bo->flags & AMDGPU_GEM_CREATE_GFX12_DCC)
+		attach->peer2peer = false;
 
-	return 0;
+	if (!amdgpu_dmabuf_is_xgmi_accessible(attach_adev, bo) &&
+	    pci_p2pdma_distance(adev->pdev, attach->dev, false) < 0)
+		attach->peer2peer = false;
 
-out:
-	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
-	return r;
-}
+	r = dma_resv_lock(bo->tbo.base.resv, NULL);
+	if (r)
+		return r;
 
-/**
- * amdgpu_dma_buf_detach - &dma_buf_ops.detach implementation
- *
- * @dmabuf: DMA-buf where we remove the attachment from
- * @attach: the attachment to remove
- *
- * Called when an attachment is removed from the DMA-buf.
- */
-static void amdgpu_dma_buf_detach(struct dma_buf *dmabuf,
-				  struct dma_buf_attachment *attach)
-{
-	struct drm_gem_object *obj = dmabuf->priv;
-	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
-	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
+	amdgpu_vm_bo_update_shared(bo);
 
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
-	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
+	dma_resv_unlock(bo->tbo.base.resv);
+
+	return 0;
 }
 
 /**
@@ -101,11 +131,35 @@ static void amdgpu_dma_buf_detach(struct dma_buf *dmabuf,
  */
 static int amdgpu_dma_buf_pin(struct dma_buf_attachment *attach)
 {
-	struct drm_gem_object *obj = attach->dmabuf->priv;
-	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
+	struct dma_buf *dmabuf = attach->dmabuf;
+	struct amdgpu_bo *bo = gem_to_amdgpu_bo(dmabuf->priv);
+	u32 domains = bo->allowed_domains;
+
+	dma_resv_assert_held(dmabuf->resv);
+
+	/* Try pinning into VRAM to allow P2P with RDMA NICs without ODP
+	 * support if all attachments can do P2P. If any attachment can't do
+	 * P2P just pin into GTT instead.
+	 *
+	 * To avoid with conflicting pinnings between GPUs and RDMA when move
+	 * notifiers are disabled, only allow pinning in VRAM when move
+	 * notiers are enabled.
+	 */
+	if (!IS_ENABLED(CONFIG_DMABUF_MOVE_NOTIFY)) {
+		domains &= ~AMDGPU_GEM_DOMAIN_VRAM;
+	} else {
+		list_for_each_entry(attach, &dmabuf->attachments, node)
+			if (!attach->peer2peer)
+				domains &= ~AMDGPU_GEM_DOMAIN_VRAM;
+	}
+
+	if (domains & AMDGPU_GEM_DOMAIN_VRAM)
+		bo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
 
-	/* pin buffer into GTT */
-	return amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT);
+	if (WARN_ON(!domains))
+		return -EINVAL;
+
+	return amdgpu_bo_pin(bo, domains);
 }
 
 /**
@@ -149,7 +203,7 @@ static struct sg_table *amdgpu_dma_buf_map(struct dma_buf_attachment *attach,
 	if (!bo->tbo.pin_count) {
 		/* move buffer into GTT or VRAM */
 		struct ttm_operation_ctx ctx = { false, false };
-		unsigned domains = AMDGPU_GEM_DOMAIN_GTT;
+		unsigned int domains = AMDGPU_GEM_DOMAIN_GTT;
 
 		if (bo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM &&
 		    attach->peer2peer) {
@@ -160,10 +214,6 @@ static struct sg_table *amdgpu_dma_buf_map(struct dma_buf_attachment *attach,
 		r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 		if (r)
 			return ERR_PTR(r);
-
-	} else if (!(amdgpu_mem_type_to_domain(bo->tbo.resource->mem_type) &
-		     AMDGPU_GEM_DOMAIN_GTT)) {
-		return ERR_PTR(-EBUSY);
 	}
 
 	switch (bo->tbo.resource->mem_type) {
@@ -180,6 +230,11 @@ static struct sg_table *amdgpu_dma_buf_map(struct dma_buf_attachment *attach,
 		break;
 
 	case TTM_PL_VRAM:
+		/* XGMI-accessible memory should never be DMA-mapped */
+		if (WARN_ON(amdgpu_dmabuf_is_xgmi_accessible(
+				dma_buf_attach_adev(attach), bo)))
+			return ERR_PTR(-EINVAL);
+
 		r = amdgpu_vram_mgr_alloc_sgt(adev, bo->tbo.resource, 0,
 					      bo->tbo.base.size, attach->dev,
 					      dir, &sgt);
@@ -211,7 +266,7 @@ static void amdgpu_dma_buf_unmap(struct dma_buf_attachment *attach,
 				 struct sg_table *sgt,
 				 enum dma_data_direction dir)
 {
-	if (sgt->sgl->page_link) {
+	if (sg_page(sgt->sgl)) {
 		dma_unmap_sgtable(attach->dev, sgt, dir, 0);
 		sg_free_table(sgt);
 		kfree(sgt);
@@ -261,9 +316,38 @@ static int amdgpu_dma_buf_begin_cpu_access(struct dma_buf *dma_buf,
 	return ret;
 }
 
+static int amdgpu_dma_buf_vmap(struct dma_buf *dma_buf, struct iosys_map *map)
+{
+	struct drm_gem_object *obj = dma_buf->priv;
+	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
+	int ret;
+
+	/*
+	 * Pin to keep buffer in place while it's vmap'ed. The actual
+	 * domain is not that important as long as it's mapable. Using
+	 * GTT and VRAM should be compatible with most use cases.
+	 */
+	ret = amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT | AMDGPU_GEM_DOMAIN_VRAM);
+	if (ret)
+		return ret;
+	ret = drm_gem_dmabuf_vmap(dma_buf, map);
+	if (ret)
+		amdgpu_bo_unpin(bo);
+
+	return ret;
+}
+
+static void amdgpu_dma_buf_vunmap(struct dma_buf *dma_buf, struct iosys_map *map)
+{
+	struct drm_gem_object *obj = dma_buf->priv;
+	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
+
+	drm_gem_dmabuf_vunmap(dma_buf, map);
+	amdgpu_bo_unpin(bo);
+}
+
 const struct dma_buf_ops amdgpu_dmabuf_ops = {
 	.attach = amdgpu_dma_buf_attach,
-	.detach = amdgpu_dma_buf_detach,
 	.pin = amdgpu_dma_buf_pin,
 	.unpin = amdgpu_dma_buf_unpin,
 	.map_dma_buf = amdgpu_dma_buf_map,
@@ -271,8 +355,8 @@ const struct dma_buf_ops amdgpu_dmabuf_ops = {
 	.release = drm_gem_dmabuf_release,
 	.begin_cpu_access = amdgpu_dma_buf_begin_cpu_access,
 	.mmap = drm_gem_dmabuf_mmap,
-	.vmap = drm_gem_dmabuf_vmap,
-	.vunmap = drm_gem_dmabuf_vunmap,
+	.vmap = amdgpu_dma_buf_vmap,
+	.vunmap = amdgpu_dma_buf_vunmap,
 };
 
 /**
@@ -290,11 +374,23 @@ struct dma_buf *amdgpu_gem_prime_export(struct drm_gem_object *gobj,
 {
 	struct amdgpu_bo *bo = gem_to_amdgpu_bo(gobj);
 	struct dma_buf *buf;
+	struct ttm_operation_ctx ctx = {
+		.interruptible = true,
+		.no_wait_gpu = true,
+		/* We opt to avoid OOM on system pages allocations */
+		.gfp_retry_mayfail = true,
+		.allow_res_evict = false,
+	};
+	int ret;
 
 	if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm) ||
 	    bo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID)
 		return ERR_PTR(-EPERM);
 
+	ret = ttm_bo_setup_export(&bo->tbo, &ctx);
+	if (ret)
+		return ERR_PTR(ret);
+
 	buf = drm_gem_prime_export(gobj, flags);
 	if (!IS_ERR(buf))
 		buf->ops = &amdgpu_dmabuf_ops;
@@ -331,12 +427,13 @@ amdgpu_dma_buf_create_obj(struct drm_device *dev, struct dma_buf *dma_buf)
 
 		flags |= other->flags & (AMDGPU_GEM_CREATE_CPU_GTT_USWC |
 					 AMDGPU_GEM_CREATE_COHERENT |
+					 AMDGPU_GEM_CREATE_EXT_COHERENT |
 					 AMDGPU_GEM_CREATE_UNCACHED);
 	}
 
 	ret = amdgpu_gem_object_create(adev, dma_buf->size, PAGE_SIZE,
 				       AMDGPU_GEM_DOMAIN_CPU, flags,
-				       ttm_bo_type_sg, resv, &gobj);
+				       ttm_bo_type_sg, resv, &gobj, 0);
 	if (ret)
 		goto error;
 
@@ -372,6 +469,10 @@ amdgpu_dma_buf_move_notify(struct dma_buf_attachment *attach)
 	struct amdgpu_vm_bo_base *bo_base;
 	int r;
 
+	/* FIXME: This should be after the "if", but needs a fix to make sure
+	 * DMABuf imports are initialized in the right VM list.
+	 */
+	amdgpu_vm_bo_invalidate(bo, false);
 	if (!bo->tbo.resource || bo->tbo.resource->mem_type == TTM_PL_SYSTEM)
 		return;
 
@@ -403,9 +504,12 @@ amdgpu_dma_buf_move_notify(struct dma_buf_attachment *attach)
 				continue;
 		}
 
-		r = amdgpu_vm_clear_freed(adev, vm, NULL);
+		/* Reserve fences for two SDMA page table updates */
+		r = dma_resv_reserve_fences(resv, 2);
+		if (!r)
+			r = amdgpu_vm_clear_freed(adev, vm, NULL);
 		if (!r)
-			r = amdgpu_vm_handle_moved(adev, vm);
+			r = amdgpu_vm_handle_moved(adev, vm, ticket);
 
 		if (r && r != -EBUSY)
 			DRM_ERROR("Failed to invalidate VM page tables (%d))\n",
@@ -479,7 +583,10 @@ bool amdgpu_dmabuf_is_xgmi_accessible(struct amdgpu_device *adev,
 	struct drm_gem_object *obj = &bo->tbo.base;
 	struct drm_gem_object *gobj;
 
-	if (obj->import_attach) {
+	if (!adev)
+		return false;
+
+	if (drm_gem_is_imported(obj)) {
 		struct dma_buf *dma_buf = obj->import_attach->dmabuf;
 
 		if (dma_buf->ops != &amdgpu_dmabuf_ops)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h
index 8fd11497faba..2675689ef70f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h
@@ -31,10 +31,15 @@ struct amdgpu_doorbell {
 	/* doorbell mmio */
 	resource_size_t		base;
 	resource_size_t		size;
-	u32 __iomem		*ptr;
 
 	/* Number of doorbells reserved for amdgpu kernel driver */
 	u32 num_kernel_doorbells;
+
+	/* Kernel doorbells */
+	struct amdgpu_bo *kernel_doorbells;
+
+	/* For CPU access of doorbells */
+	uint32_t *cpu_addr;
 };
 
 /* Reserved doorbells for amdgpu (including multimedia).
@@ -59,7 +64,7 @@ struct amdgpu_doorbell_index {
 	uint32_t gfx_ring1;
 	uint32_t gfx_userqueue_start;
 	uint32_t gfx_userqueue_end;
-	uint32_t sdma_engine[8];
+	uint32_t sdma_engine[16];
 	uint32_t mes_ring0;
 	uint32_t mes_ring1;
 	uint32_t ih;
@@ -81,15 +86,17 @@ struct amdgpu_doorbell_index {
 			uint32_t vce_ring6_7;
 		} uvd_vce;
 	};
+	uint32_t vpe_ring;
 	uint32_t first_non_cp;
 	uint32_t last_non_cp;
 	uint32_t max_assignment;
 	/* Per engine SDMA doorbell size in dword */
 	uint32_t sdma_doorbell_range;
+	/* Per xcc doorbell size for KIQ/KCQ */
+	uint32_t xcc_doorbell_range;
 };
 
-typedef enum _AMDGPU_DOORBELL_ASSIGNMENT
-{
+enum AMDGPU_DOORBELL_ASSIGNMENT {
 	AMDGPU_DOORBELL_KIQ                     = 0x000,
 	AMDGPU_DOORBELL_HIQ                     = 0x001,
 	AMDGPU_DOORBELL_DIQ                     = 0x002,
@@ -107,10 +114,10 @@ typedef enum _AMDGPU_DOORBELL_ASSIGNMENT
 	AMDGPU_DOORBELL_IH                      = 0x1E8,
 	AMDGPU_DOORBELL_MAX_ASSIGNMENT          = 0x3FF,
 	AMDGPU_DOORBELL_INVALID                 = 0xFFFF
-} AMDGPU_DOORBELL_ASSIGNMENT;
+};
+
+enum AMDGPU_VEGA20_DOORBELL_ASSIGNMENT {
 
-typedef enum _AMDGPU_VEGA20_DOORBELL_ASSIGNMENT
-{
 	/* Compute + GFX: 0~255 */
 	AMDGPU_VEGA20_DOORBELL_KIQ                     = 0x000,
 	AMDGPU_VEGA20_DOORBELL_HIQ                     = 0x001,
@@ -164,12 +171,20 @@ typedef enum _AMDGPU_VEGA20_DOORBELL_ASSIGNMENT
 	AMDGPU_VEGA20_DOORBELL64_FIRST_NON_CP            = AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE0,
 	AMDGPU_VEGA20_DOORBELL64_LAST_NON_CP             = AMDGPU_VEGA20_DOORBELL64_VCE_RING6_7,
 
-	AMDGPU_VEGA20_DOORBELL_MAX_ASSIGNMENT            = 0x18F,
+	/* kiq/kcq from second XCD. Max 8 XCDs */
+	AMDGPU_VEGA20_DOORBELL_XCC1_KIQ_START             = 0x190,
+	/* 8 compute rings per GC. Max to 0x1CE */
+	AMDGPU_VEGA20_DOORBELL_XCC1_MEC_RING0_START       = 0x197,
+
+	/* AID1 SDMA: 0x1D0 ~ 0x1F7 */
+	AMDGPU_VEGA20_DOORBELL_AID1_sDMA_START           = 0x1D0,
+
+	AMDGPU_VEGA20_DOORBELL_MAX_ASSIGNMENT            = 0x1F7,
 	AMDGPU_VEGA20_DOORBELL_INVALID                   = 0xFFFF
-} AMDGPU_VEGA20_DOORBELL_ASSIGNMENT;
+};
+
+enum AMDGPU_NAVI10_DOORBELL_ASSIGNMENT {
 
-typedef enum _AMDGPU_NAVI10_DOORBELL_ASSIGNMENT
-{
 	/* Compute + GFX: 0~255 */
 	AMDGPU_NAVI10_DOORBELL_KIQ			= 0x000,
 	AMDGPU_NAVI10_DOORBELL_HIQ			= 0x001,
@@ -212,18 +227,19 @@ typedef enum _AMDGPU_NAVI10_DOORBELL_ASSIGNMENT
 	AMDGPU_NAVI10_DOORBELL64_VCNc_d			= 0x18E,
 	AMDGPU_NAVI10_DOORBELL64_VCNe_f			= 0x18F,
 
+	AMDGPU_NAVI10_DOORBELL64_VPE			= 0x190,
+
 	AMDGPU_NAVI10_DOORBELL64_FIRST_NON_CP		= AMDGPU_NAVI10_DOORBELL_sDMA_ENGINE0,
-	AMDGPU_NAVI10_DOORBELL64_LAST_NON_CP		= AMDGPU_NAVI10_DOORBELL64_VCNe_f,
+	AMDGPU_NAVI10_DOORBELL64_LAST_NON_CP		= AMDGPU_NAVI10_DOORBELL64_VPE,
 
-	AMDGPU_NAVI10_DOORBELL_MAX_ASSIGNMENT		= 0x18F,
+	AMDGPU_NAVI10_DOORBELL_MAX_ASSIGNMENT		= AMDGPU_NAVI10_DOORBELL64_VPE,
 	AMDGPU_NAVI10_DOORBELL_INVALID			= 0xFFFF
-} AMDGPU_NAVI10_DOORBELL_ASSIGNMENT;
+};
 
 /*
  * 64bit doorbell, offset are in QWORD, occupy 2KB doorbell space
  */
-typedef enum _AMDGPU_DOORBELL64_ASSIGNMENT
-{
+enum AMDGPU_DOORBELL64_ASSIGNMENT {
 	/*
 	 * All compute related doorbells: kiq, hiq, diq, traditional compute queue, user queue, should locate in
 	 * a continues range so that programming CP_MEC_DOORBELL_RANGE_LOWER/UPPER can cover this range.
@@ -299,13 +315,55 @@ typedef enum _AMDGPU_DOORBELL64_ASSIGNMENT
 
 	AMDGPU_DOORBELL64_MAX_ASSIGNMENT          = 0xFF,
 	AMDGPU_DOORBELL64_INVALID                 = 0xFFFF
-} AMDGPU_DOORBELL64_ASSIGNMENT;
+};
+
+enum AMDGPU_DOORBELL_ASSIGNMENT_LAYOUT1 {
+
+	/* XCC0: 0x00 ~20, XCC1: 20 ~ 2F ... */
+
+	/* KIQ/HIQ/DIQ */
+	AMDGPU_DOORBELL_LAYOUT1_KIQ_START		= 0x000,
+	AMDGPU_DOORBELL_LAYOUT1_HIQ			= 0x001,
+	AMDGPU_DOORBELL_LAYOUT1_DIQ			= 0x002,
+	/* Compute: 0x08 ~ 0x20  */
+	AMDGPU_DOORBELL_LAYOUT1_MEC_RING_START		= 0x008,
+	AMDGPU_DOORBELL_LAYOUT1_MEC_RING_END		= 0x00F,
+	AMDGPU_DOORBELL_LAYOUT1_USERQUEUE_START		= 0x010,
+	AMDGPU_DOORBELL_LAYOUT1_USERQUEUE_END		= 0x01F,
+	AMDGPU_DOORBELL_LAYOUT1_XCC_RANGE		= 0x020,
+
+	/* SDMA: 0x100 ~ 0x19F */
+	AMDGPU_DOORBELL_LAYOUT1_sDMA_ENGINE_START	= 0x100,
+	AMDGPU_DOORBELL_LAYOUT1_sDMA_ENGINE_END		= 0x19F,
+	/* IH: 0x1A0 ~ 0x1AF */
+	AMDGPU_DOORBELL_LAYOUT1_IH                      = 0x1A0,
+	/* VCN: 0x1B0 ~ 0x1E8 */
+	AMDGPU_DOORBELL_LAYOUT1_VCN_START               = 0x1B0,
+	AMDGPU_DOORBELL_LAYOUT1_VCN_END                 = 0x1E8,
+
+	AMDGPU_DOORBELL_LAYOUT1_FIRST_NON_CP		= AMDGPU_DOORBELL_LAYOUT1_sDMA_ENGINE_START,
+	AMDGPU_DOORBELL_LAYOUT1_LAST_NON_CP		= AMDGPU_DOORBELL_LAYOUT1_VCN_END,
+
+	AMDGPU_DOORBELL_LAYOUT1_MAX_ASSIGNMENT          = 0x1E8,
+	AMDGPU_DOORBELL_LAYOUT1_INVALID                 = 0xFFFF
+};
 
 u32 amdgpu_mm_rdoorbell(struct amdgpu_device *adev, u32 index);
 void amdgpu_mm_wdoorbell(struct amdgpu_device *adev, u32 index, u32 v);
 u64 amdgpu_mm_rdoorbell64(struct amdgpu_device *adev, u32 index);
 void amdgpu_mm_wdoorbell64(struct amdgpu_device *adev, u32 index, u64 v);
 
+/*
+ * GPU doorbell aperture helpers function.
+ */
+int amdgpu_doorbell_init(struct amdgpu_device *adev);
+void amdgpu_doorbell_fini(struct amdgpu_device *adev);
+int amdgpu_doorbell_create_kernel_doorbells(struct amdgpu_device *adev);
+uint32_t amdgpu_doorbell_index_on_bar(struct amdgpu_device *adev,
+				      struct amdgpu_bo *db_bo,
+				      uint32_t doorbell_index,
+				      uint32_t db_size);
+
 #define RDOORBELL32(index) amdgpu_mm_rdoorbell(adev, (index))
 #define WDOORBELL32(index, v) amdgpu_mm_wdoorbell(adev, (index), (v))
 #define RDOORBELL64(index) amdgpu_mm_rdoorbell64(adev, (index))
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell_mgr.c
new file mode 100644
index 000000000000..3040437d99c2
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell_mgr.c
@@ -0,0 +1,244 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "amdgpu.h"
+
+/**
+ * amdgpu_mm_rdoorbell - read a doorbell dword
+ *
+ * @adev: amdgpu_device pointer
+ * @index: doorbell index
+ *
+ * Returns the value in the doorbell aperture at the
+ * requested doorbell index (CIK).
+ */
+u32 amdgpu_mm_rdoorbell(struct amdgpu_device *adev, u32 index)
+{
+	if (amdgpu_device_skip_hw_access(adev))
+		return 0;
+
+	if (index < adev->doorbell.num_kernel_doorbells)
+		return readl(adev->doorbell.cpu_addr + index);
+
+	dev_err(adev->dev, "reading beyond doorbell aperture: 0x%08x!\n",
+		index);
+	return 0;
+}
+
+/**
+ * amdgpu_mm_wdoorbell - write a doorbell dword
+ *
+ * @adev: amdgpu_device pointer
+ * @index: doorbell index
+ * @v: value to write
+ *
+ * Writes @v to the doorbell aperture at the
+ * requested doorbell index (CIK).
+ */
+void amdgpu_mm_wdoorbell(struct amdgpu_device *adev, u32 index, u32 v)
+{
+	if (amdgpu_device_skip_hw_access(adev))
+		return;
+
+	if (index < adev->doorbell.num_kernel_doorbells)
+		writel(v, adev->doorbell.cpu_addr + index);
+	else
+		dev_err(adev->dev,
+			"writing beyond doorbell aperture: 0x%08x!\n", index);
+}
+
+/**
+ * amdgpu_mm_rdoorbell64 - read a doorbell Qword
+ *
+ * @adev: amdgpu_device pointer
+ * @index: doorbell index
+ *
+ * Returns the value in the doorbell aperture at the
+ * requested doorbell index (VEGA10+).
+ */
+u64 amdgpu_mm_rdoorbell64(struct amdgpu_device *adev, u32 index)
+{
+	if (amdgpu_device_skip_hw_access(adev))
+		return 0;
+
+	if (index < adev->doorbell.num_kernel_doorbells)
+		return atomic64_read((atomic64_t *)(adev->doorbell.cpu_addr + index));
+
+	dev_err(adev->dev, "reading beyond doorbell aperture: 0x%08x!\n",
+		index);
+	return 0;
+}
+
+/**
+ * amdgpu_mm_wdoorbell64 - write a doorbell Qword
+ *
+ * @adev: amdgpu_device pointer
+ * @index: doorbell index
+ * @v: value to write
+ *
+ * Writes @v to the doorbell aperture at the
+ * requested doorbell index (VEGA10+).
+ */
+void amdgpu_mm_wdoorbell64(struct amdgpu_device *adev, u32 index, u64 v)
+{
+	if (amdgpu_device_skip_hw_access(adev))
+		return;
+
+	if (index < adev->doorbell.num_kernel_doorbells)
+		atomic64_set((atomic64_t *)(adev->doorbell.cpu_addr + index), v);
+	else
+		dev_err(adev->dev,
+			"writing beyond doorbell aperture: 0x%08x!\n", index);
+}
+
+/**
+ * amdgpu_doorbell_index_on_bar - Find doorbell's absolute offset in BAR
+ *
+ * @adev: amdgpu_device pointer
+ * @db_bo: doorbell object's bo
+ * @doorbell_index: doorbell relative index in this doorbell object
+ * @db_size: doorbell size is in byte
+ *
+ * returns doorbell's absolute index in BAR
+ */
+uint32_t amdgpu_doorbell_index_on_bar(struct amdgpu_device *adev,
+				      struct amdgpu_bo *db_bo,
+				      uint32_t doorbell_index,
+				      uint32_t db_size)
+{
+	int db_bo_offset;
+
+	db_bo_offset = amdgpu_bo_gpu_offset_no_check(db_bo);
+
+	/* doorbell index is 32 bit but doorbell's size can be 32 bit
+	 * or 64 bit, so *db_size(in byte)/4 for alignment.
+	 */
+	return db_bo_offset / sizeof(u32) + doorbell_index *
+	       DIV_ROUND_UP(db_size, 4);
+}
+
+/**
+ * amdgpu_doorbell_create_kernel_doorbells - Create kernel doorbells for graphics
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Creates doorbells for graphics driver usages.
+ * returns 0 on success, error otherwise.
+ */
+int amdgpu_doorbell_create_kernel_doorbells(struct amdgpu_device *adev)
+{
+	int r;
+	int size;
+
+	/* SI HW does not have doorbells, skip allocation */
+	if (adev->doorbell.num_kernel_doorbells == 0)
+		return 0;
+
+	/* Reserve first num_kernel_doorbells (page-aligned) for kernel ops */
+	size = ALIGN(adev->doorbell.num_kernel_doorbells * sizeof(u32), PAGE_SIZE);
+
+	/* Allocate an extra page for MES kernel usages (ring test) */
+	adev->mes.db_start_dw_offset = size / sizeof(u32);
+	size += PAGE_SIZE;
+
+	r = amdgpu_bo_create_kernel(adev,
+				    size,
+				    PAGE_SIZE,
+				    AMDGPU_GEM_DOMAIN_DOORBELL,
+				    &adev->doorbell.kernel_doorbells,
+				    NULL,
+				    (void **)&adev->doorbell.cpu_addr);
+	if (r) {
+		dev_err(adev->dev,
+			"Failed to allocate kernel doorbells, err=%d\n", r);
+		return r;
+	}
+
+	adev->doorbell.num_kernel_doorbells = size / sizeof(u32);
+	return 0;
+}
+
+/*
+ * GPU doorbell aperture helpers function.
+ */
+/**
+ * amdgpu_doorbell_init - Init doorbell driver information.
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Init doorbell driver information (CIK)
+ * Returns 0 on success, error on failure.
+ */
+int amdgpu_doorbell_init(struct amdgpu_device *adev)
+{
+
+	/* No doorbell on SI hardware generation */
+	if (adev->asic_type < CHIP_BONAIRE) {
+		adev->doorbell.base = 0;
+		adev->doorbell.size = 0;
+		adev->doorbell.num_kernel_doorbells = 0;
+		return 0;
+	}
+
+	if (pci_resource_flags(adev->pdev, 2) & IORESOURCE_UNSET)
+		return -EINVAL;
+
+	amdgpu_asic_init_doorbell_index(adev);
+
+	/* doorbell bar mapping */
+	adev->doorbell.base = pci_resource_start(adev->pdev, 2);
+	adev->doorbell.size = pci_resource_len(adev->pdev, 2);
+
+	adev->doorbell.num_kernel_doorbells =
+		min_t(u32, adev->doorbell.size / sizeof(u32),
+		      adev->doorbell_index.max_assignment + 1);
+	if (adev->doorbell.num_kernel_doorbells == 0)
+		return -EINVAL;
+
+	/*
+	 * For Vega, reserve and map two pages on doorbell BAR since SDMA
+	 * paging queue doorbell use the second page. The
+	 * AMDGPU_DOORBELL64_MAX_ASSIGNMENT definition assumes all the
+	 * doorbells are in the first page. So with paging queue enabled,
+	 * the max num_kernel_doorbells should + 1 page (0x400 in dword)
+	 */
+	if (adev->asic_type >= CHIP_VEGA10)
+		adev->doorbell.num_kernel_doorbells += 0x400;
+
+	return 0;
+}
+
+/**
+ * amdgpu_doorbell_fini - Tear down doorbell driver information.
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Tear down doorbell driver information (CIK)
+ */
+void amdgpu_doorbell_fini(struct amdgpu_device *adev)
+{
+	amdgpu_bo_free_kernel(&adev->doorbell.kernel_doorbells,
+			      NULL,
+			      (void **)&adev->doorbell.cpu_addr);
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
index b1ca1ab6d6ad..2dfbddcef9ab 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
@@ -23,33 +23,37 @@
  */
 
 #include <drm/amdgpu_drm.h>
+#include <drm/clients/drm_client_setup.h>
 #include <drm/drm_drv.h>
-#include <drm/drm_fbdev_generic.h>
+#include <drm/drm_fbdev_ttm.h>
 #include <drm/drm_gem.h>
-#include <drm/drm_vblank.h>
 #include <drm/drm_managed.h>
-#include "amdgpu_drv.h"
-
 #include <drm/drm_pciids.h>
-#include <linux/module.h>
-#include <linux/pm_runtime.h>
-#include <linux/vga_switcheroo.h>
 #include <drm/drm_probe_helper.h>
-#include <linux/mmu_notifier.h>
-#include <linux/suspend.h>
+#include <drm/drm_vblank.h>
+
 #include <linux/cc_platform.h>
 #include <linux/dynamic_debug.h>
+#include <linux/module.h>
+#include <linux/mmu_notifier.h>
+#include <linux/pm_runtime.h>
+#include <linux/suspend.h>
+#include <linux/vga_switcheroo.h>
 
 #include "amdgpu.h"
-#include "amdgpu_irq.h"
+#include "amdgpu_amdkfd.h"
 #include "amdgpu_dma_buf.h"
-#include "amdgpu_sched.h"
+#include "amdgpu_drv.h"
 #include "amdgpu_fdinfo.h"
-#include "amdgpu_amdkfd.h"
-
+#include "amdgpu_irq.h"
+#include "amdgpu_psp.h"
 #include "amdgpu_ras.h"
-#include "amdgpu_xgmi.h"
 #include "amdgpu_reset.h"
+#include "amdgpu_sched.h"
+#include "amdgpu_xgmi.h"
+#include "amdgpu_userq.h"
+#include "amdgpu_userq_fence.h"
+#include "../amdxcp/amdgpu_xcp_drv.h"
 
 /*
  * KMS wrapper.
@@ -110,11 +114,40 @@
  *   3.52.0 - Add AMDGPU_IDS_FLAGS_CONFORMANT_TRUNC_COORD, add device_info fields:
  *            tcp_cache_size, num_sqc_per_wgp, sqc_data_cache_size, sqc_inst_cache_size,
  *            gl1c_cache_size, gl2c_cache_size, mall_size, enabled_rb_pipes_mask_hi
+ *   3.53.0 - Support for GFX11 CP GFX shadowing
+ *   3.54.0 - Add AMDGPU_CTX_QUERY2_FLAGS_RESET_IN_PROGRESS support
+ * - 3.55.0 - Add AMDGPU_INFO_GPUVM_FAULT query
+ * - 3.56.0 - Update IB start address and size alignment for decode and encode
+ * - 3.57.0 - Compute tunneling on GFX10+
+ * - 3.58.0 - Add GFX12 DCC support
+ * - 3.59.0 - Cleared VRAM
+ * - 3.60.0 - Add AMDGPU_TILING_GFX12_DCC_WRITE_COMPRESS_DISABLE (Vulkan requirement)
+ * - 3.61.0 - Contains fix for RV/PCO compute queues
+ * - 3.62.0 - Add AMDGPU_IDS_FLAGS_MODE_PF, AMDGPU_IDS_FLAGS_MODE_VF & AMDGPU_IDS_FLAGS_MODE_PT
+ * - 3.63.0 - GFX12 display DCC supports 256B max compressed block size
+ * - 3.64.0 - Userq IP support query
  */
 #define KMS_DRIVER_MAJOR	3
-#define KMS_DRIVER_MINOR	52
+#define KMS_DRIVER_MINOR	64
 #define KMS_DRIVER_PATCHLEVEL	0
 
+/*
+ * amdgpu.debug module options. Are all disabled by default
+ */
+enum AMDGPU_DEBUG_MASK {
+	AMDGPU_DEBUG_VM = BIT(0),
+	AMDGPU_DEBUG_LARGEBAR = BIT(1),
+	AMDGPU_DEBUG_DISABLE_GPU_SOFT_RECOVERY = BIT(2),
+	AMDGPU_DEBUG_USE_VRAM_FW_BUF = BIT(3),
+	AMDGPU_DEBUG_ENABLE_RAS_ACA = BIT(4),
+	AMDGPU_DEBUG_ENABLE_EXP_RESETS = BIT(5),
+	AMDGPU_DEBUG_DISABLE_GPU_RING_RESET = BIT(6),
+	AMDGPU_DEBUG_SMU_POOL = BIT(7),
+	AMDGPU_DEBUG_VM_USERPTR = BIT(8),
+	AMDGPU_DEBUG_DISABLE_RAS_CE_LOG = BIT(9),
+	AMDGPU_DEBUG_ENABLE_CE_CS = BIT(10)
+};
+
 unsigned int amdgpu_vram_limit = UINT_MAX;
 int amdgpu_vis_vram_limit;
 int amdgpu_gart_size = -1; /* auto */
@@ -137,7 +170,6 @@ int amdgpu_vm_size = -1;
 int amdgpu_vm_fragment_size = -1;
 int amdgpu_vm_block_size = -1;
 int amdgpu_vm_fault_stop;
-int amdgpu_vm_debug;
 int amdgpu_vm_update_mode = -1;
 int amdgpu_exp_hw_support;
 int amdgpu_dc = -1;
@@ -150,6 +182,17 @@ uint amdgpu_pg_mask = 0xffffffff;
 uint amdgpu_sdma_phase_quantum = 32;
 char *amdgpu_disable_cu;
 char *amdgpu_virtual_display;
+int amdgpu_enforce_isolation = -1;
+int amdgpu_modeset = -1;
+
+/* Specifies the default granularity for SVM, used in buffer
+ * migration and restoration of backing memory when handling
+ * recoverable page faults.
+ *
+ * The value is given as log(numPages(buffer)); for a 2 MiB
+ * buffer it computes to be 9
+ */
+uint amdgpu_svm_default_granularity = 9;
 
 /*
  * OverDrive(bit 14) disabled by default
@@ -177,10 +220,12 @@ uint amdgpu_dc_feature_mask = 2;
 uint amdgpu_dc_debug_mask;
 uint amdgpu_dc_visual_confirm;
 int amdgpu_async_gfx_ring = 1;
-int amdgpu_mcbp;
+int amdgpu_mcbp = -1;
 int amdgpu_discovery = -1;
 int amdgpu_mes;
+int amdgpu_mes_log_enable = 0;
 int amdgpu_mes_kiq;
+int amdgpu_uni_mes = 1;
 int amdgpu_noretry = -1;
 int amdgpu_force_asic_type = -1;
 int amdgpu_tmz = -1; /* auto */
@@ -191,8 +236,16 @@ int amdgpu_smartshift_bias;
 int amdgpu_use_xgmi_p2p = 1;
 int amdgpu_vcnfw_log;
 int amdgpu_sg_display = -1; /* auto */
-
-static void amdgpu_drv_delayed_reset_work_handler(struct work_struct *work);
+int amdgpu_user_partt_mode = AMDGPU_AUTO_COMPUTE_PARTITION_MODE;
+int amdgpu_umsch_mm;
+int amdgpu_seamless = -1; /* auto */
+uint amdgpu_debug_mask;
+int amdgpu_agp = -1; /* auto */
+int amdgpu_wbrf = -1;
+int amdgpu_damage_clips = -1; /* auto */
+int amdgpu_umsch_mm_fwlog;
+int amdgpu_rebar = -1; /* auto */
+int amdgpu_user_queue = -1;
 
 DECLARE_DYNDBG_CLASSMAP(drm_debug_classes, DD_CLASS_TYPE_DISJOINT_BITS, 0,
 			"DRM_UT_CORE",
@@ -208,9 +261,6 @@ DECLARE_DYNDBG_CLASSMAP(drm_debug_classes, DD_CLASS_TYPE_DISJOINT_BITS, 0,
 
 struct amdgpu_mgpu_info mgpu_info = {
 	.mutex = __MUTEX_INITIALIZER(mgpu_info.mutex),
-	.delayed_reset_work = __DELAYED_WORK_INITIALIZER(
-			mgpu_info.delayed_reset_work,
-			amdgpu_drv_delayed_reset_work_handler, 0),
 };
 int amdgpu_ras_enable = -1;
 uint amdgpu_ras_mask = 0xffffffff;
@@ -245,7 +295,8 @@ module_param_named(gartsize, amdgpu_gart_size, uint, 0600);
 /**
  * DOC: gttsize (int)
  * Restrict the size of GTT domain (for userspace use) in MiB for testing.
- * The default is -1 (Use 1/2 RAM, minimum value is 3GB).
+ * The default is -1 (Use value specified by TTM).
+ * This parameter is deprecated and will be removed in the future.
  */
 MODULE_PARM_DESC(gttsize, "Size of the GTT userspace domain in megabytes (-1 = auto)");
 module_param_named(gttsize, amdgpu_gtt_size, int, 0600);
@@ -261,7 +312,7 @@ module_param_named(moverate, amdgpu_moverate, int, 0600);
  * DOC: audio (int)
  * Set HDMI/DPAudio. Only affects non-DC display handling. The default is -1 (Enabled), set 0 to disabled it.
  */
-MODULE_PARM_DESC(audio, "Audio enable (-1 = auto, 0 = disable, 1 = enable)");
+MODULE_PARM_DESC(audio, "HDMI/DP Audio enable for non DC displays (-1 = auto, 0 = disable, 1 = enable)");
 module_param_named(audio, amdgpu_audio, int, 0444);
 
 /**
@@ -293,27 +344,26 @@ MODULE_PARM_DESC(msi, "MSI support (1 = enable, 0 = disable, -1 = auto)");
 module_param_named(msi, amdgpu_msi, int, 0444);
 
 /**
+ * DOC: svm_default_granularity (uint)
+ * Used in buffer migration and handling of recoverable page faults
+ */
+MODULE_PARM_DESC(svm_default_granularity, "SVM's default granularity in log(2^Pages), default 9 = 2^9 = 2 MiB");
+module_param_named(svm_default_granularity, amdgpu_svm_default_granularity, uint, 0644);
+
+/**
  * DOC: lockup_timeout (string)
  * Set GPU scheduler timeout value in ms.
  *
- * The format can be [Non-Compute] or [GFX,Compute,SDMA,Video]. That is there can be one or
- * multiple values specified. 0 and negative values are invalidated. They will be adjusted
- * to the default timeout.
- *
- * - With one value specified, the setting will apply to all non-compute jobs.
- * - With multiple values specified, the first one will be for GFX.
- *   The second one is for Compute. The third and fourth ones are
- *   for SDMA and Video.
+ * The format can be [single value] for setting all timeouts at once or
+ * [GFX,Compute,SDMA,Video] to set individual timeouts.
+ * Negative values mean infinity.
  *
- * By default(with no lockup_timeout settings), the timeout for all non-compute(GFX, SDMA and Video)
- * jobs is 10000. The timeout for compute is 60000.
+ * By default(with no lockup_timeout settings), the timeout for all queues is 2000.
  */
-MODULE_PARM_DESC(lockup_timeout, "GPU lockup timeout in ms (default: for bare metal 10000 for non-compute jobs and 60000 for compute jobs; "
-		"for passthrough or sriov, 10000 for all jobs."
-		" 0: keep default value. negative: infinity timeout), "
-		"format: for bare metal [Non-Compute] or [GFX,Compute,SDMA,Video]; "
-		"for passthrough or sriov [all jobs] or [GFX,Compute,SDMA,Video].");
-module_param_string(lockup_timeout, amdgpu_lockup_timeout, sizeof(amdgpu_lockup_timeout), 0444);
+MODULE_PARM_DESC(lockup_timeout,
+		 "GPU lockup timeout in ms (default: 2000. 0: keep default value. negative: infinity timeout), format: [single value for all] or [GFX,Compute,SDMA,Video].");
+module_param_string(lockup_timeout, amdgpu_lockup_timeout,
+		    sizeof(amdgpu_lockup_timeout), 0444);
 
 /**
  * DOC: dpm (int)
@@ -346,8 +396,9 @@ module_param_named(aspm, amdgpu_aspm, int, 0444);
  * Override for runtime power management control for dGPUs. The amdgpu driver can dynamically power down
  * the dGPUs when they are idle if supported. The default is -1 (auto enable).
  * Setting the value to 0 disables this functionality.
+ * Setting the value to -2 is auto enabled with power down when displays are attached.
  */
-MODULE_PARM_DESC(runpm, "PX runtime pm (2 = force enable with BAMACO, 1 = force enable with BACO, 0 = disable, -1 = auto)");
+MODULE_PARM_DESC(runpm, "PX runtime pm (2 = force enable with BAMACO, 1 = force enable with BACO, 0 = disable, -1 = auto, -2 = auto with displays)");
 module_param_named(runpm, amdgpu_runtime_pm, int, 0444);
 
 /**
@@ -358,7 +409,7 @@ module_param_named(runpm, amdgpu_runtime_pm, int, 0444);
  * the kernel log for the list of IPs on the asic. The default is 0xffffffff (enable all blocks on a device).
  */
 MODULE_PARM_DESC(ip_block_mask, "IP Block Mask (all blocks enabled (default))");
-module_param_named(ip_block_mask, amdgpu_ip_block_mask, uint, 0444);
+module_param_named_unsafe(ip_block_mask, amdgpu_ip_block_mask, uint, 0444);
 
 /**
  * DOC: bapm (int)
@@ -404,13 +455,6 @@ MODULE_PARM_DESC(vm_fault_stop, "Stop on VM fault (0 = never (default), 1 = prin
 module_param_named(vm_fault_stop, amdgpu_vm_fault_stop, int, 0444);
 
 /**
- * DOC: vm_debug (int)
- * Debug VM handling (0 = disabled, 1 = enabled). The default is 0 (Disabled).
- */
-MODULE_PARM_DESC(vm_debug, "Debug VM handling (0 = disabled (default), 1 = enabled)");
-module_param_named(vm_debug, amdgpu_vm_debug, int, 0644);
-
-/**
  * DOC: vm_update_mode (int)
  * Override VM update mode. VM updated by using CPU (0 = never, 1 = Graphics only, 2 = Compute only, 3 = Both). The default
  * is -1 (Only in large BAR(LB) systems Compute VM tables will be updated by CPU, otherwise 0, never).
@@ -423,7 +467,7 @@ module_param_named(vm_update_mode, amdgpu_vm_update_mode, int, 0444);
  * Enable experimental hw support (1 = enable). The default is 0 (disabled).
  */
 MODULE_PARM_DESC(exp_hw_support, "experimental hw support (1 = enable, 0 = disable (default))");
-module_param_named(exp_hw_support, amdgpu_exp_hw_support, int, 0444);
+module_param_named_unsafe(exp_hw_support, amdgpu_exp_hw_support, int, 0444);
 
 /**
  * DOC: dc (int)
@@ -534,14 +578,14 @@ module_param_named(compute_multipipe, amdgpu_compute_multipipe, int, 0444);
  * Set to enable GPU recovery mechanism (1 = enable, 0 = disable). The default is -1 (auto, disabled except SRIOV).
  */
 MODULE_PARM_DESC(gpu_recovery, "Enable GPU recovery mechanism, (1 = enable, 0 = disable, -1 = auto)");
-module_param_named(gpu_recovery, amdgpu_gpu_recovery, int, 0444);
+module_param_named_unsafe(gpu_recovery, amdgpu_gpu_recovery, int, 0444);
 
 /**
  * DOC: emu_mode (int)
  * Set value 1 to enable emulation mode. This is only needed when running on an emulator. The default is 0 (disabled).
  */
 MODULE_PARM_DESC(emu_mode, "Emulation mode, (1 = enable, 0 = disable)");
-module_param_named(emu_mode, amdgpu_emu_mode, int, 0444);
+module_param_named_unsafe(emu_mode, amdgpu_emu_mode, int, 0444);
 
 /**
  * DOC: ras_enable (int)
@@ -574,39 +618,39 @@ module_param_named(timeout_period, amdgpu_watchdog_timer.period, uint, 0644);
 
 /**
  * DOC: si_support (int)
- * Set SI support driver. This parameter works after set config CONFIG_DRM_AMDGPU_SI. For SI asic, when radeon driver is enabled,
- * set value 0 to use radeon driver, while set value 1 to use amdgpu driver. The default is using radeon driver when it available,
- * otherwise using amdgpu driver.
- */
+ * 1 = enabled, 0 = disabled, -1 = default
+ *
+ * SI (Southern Islands) are first generation GCN GPUs, supported by both
+ * drivers: radeon (old) and amdgpu (new). This parameter controls whether
+ * amdgpu should support SI.
+ * By default, SI dedicated GPUs are supported by amdgpu.
+ * Only relevant when CONFIG_DRM_AMDGPU_SI is enabled to build SI support in amdgpu.
+ * See also radeon.si_support which should be disabled when amdgpu.si_support is
+ * enabled, and vice versa.
+ */
+int amdgpu_si_support = -1;
 #ifdef CONFIG_DRM_AMDGPU_SI
-
-#if defined(CONFIG_DRM_RADEON) || defined(CONFIG_DRM_RADEON_MODULE)
-int amdgpu_si_support = 0;
-MODULE_PARM_DESC(si_support, "SI support (1 = enabled, 0 = disabled (default))");
-#else
-int amdgpu_si_support = 1;
-MODULE_PARM_DESC(si_support, "SI support (1 = enabled (default), 0 = disabled)");
-#endif
-
+MODULE_PARM_DESC(si_support, "SI support (1 = enabled, 0 = disabled, -1 = default)");
 module_param_named(si_support, amdgpu_si_support, int, 0444);
 #endif
 
 /**
  * DOC: cik_support (int)
- * Set CIK support driver. This parameter works after set config CONFIG_DRM_AMDGPU_CIK. For CIK asic, when radeon driver is enabled,
- * set value 0 to use radeon driver, while set value 1 to use amdgpu driver. The default is using radeon driver when it available,
- * otherwise using amdgpu driver.
- */
+ * 1 = enabled, 0 = disabled, -1 = default
+ *
+ * CIK (Sea Islands) are second generation GCN GPUs, supported by both
+ * drivers: radeon (old) and amdgpu (new). This parameter controls whether
+ * amdgpu should support CIK.
+ * By default:
+ * - CIK dedicated GPUs are supported by amdgpu.
+ * - CIK APUs are supported by radeon (except when radeon is not built).
+ * Only relevant when CONFIG_DRM_AMDGPU_CIK is enabled to build CIK support in amdgpu.
+ * See also radeon.cik_support which should be disabled when amdgpu.cik_support is
+ * enabled, and vice versa.
+ */
+int amdgpu_cik_support = -1;
 #ifdef CONFIG_DRM_AMDGPU_CIK
-
-#if defined(CONFIG_DRM_RADEON) || defined(CONFIG_DRM_RADEON_MODULE)
-int amdgpu_cik_support = 0;
-MODULE_PARM_DESC(cik_support, "CIK support (1 = enabled, 0 = disabled (default))");
-#else
-int amdgpu_cik_support = 1;
-MODULE_PARM_DESC(cik_support, "CIK support (1 = enabled (default), 0 = disabled)");
-#endif
-
+MODULE_PARM_DESC(cik_support, "CIK support  (1 = enabled, 0 = disabled, -1 = default)");
 module_param_named(cik_support, amdgpu_cik_support, int, 0444);
 #endif
 
@@ -616,8 +660,7 @@ module_param_named(cik_support, amdgpu_cik_support, int, 0444);
  * E.g. 0x1 = 256Mbyte, 0x2 = 512Mbyte, 0x4 = 1 Gbyte, 0x8 = 2GByte. The default is 0 (disabled).
  */
 MODULE_PARM_DESC(smu_memory_pool_size,
-	"reserve gtt for smu debug usage, 0 = disable,"
-		"0x1 = 256Mbyte, 0x2 = 512Mbyte, 0x4 = 1 Gbyte, 0x8 = 2GByte");
+	"reserve gtt for smu debug usage, 0 = disable,0x1 = 256Mbyte, 0x2 = 512Mbyte, 0x4 = 1 Gbyte, 0x8 = 2GByte");
 module_param_named(smu_memory_pool_size, amdgpu_smu_memory_pool_size, uint, 0444);
 
 /**
@@ -630,10 +673,10 @@ module_param_named(async_gfx_ring, amdgpu_async_gfx_ring, int, 0444);
 
 /**
  * DOC: mcbp (int)
- * It is used to enable mid command buffer preemption. (0 = disabled (default), 1 = enabled)
+ * It is used to enable mid command buffer preemption. (0 = disabled, 1 = enabled, -1 auto (default))
  */
 MODULE_PARM_DESC(mcbp,
-	"Enable Mid-command buffer preemption (0 = disabled (default), 1 = enabled)");
+	"Enable Mid-command buffer preemption (0 = disabled, 1 = enabled), -1 = auto (default)");
 module_param_named(mcbp, amdgpu_mcbp, int, 0444);
 
 /**
@@ -655,6 +698,15 @@ MODULE_PARM_DESC(mes,
 module_param_named(mes, amdgpu_mes, int, 0444);
 
 /**
+ * DOC: mes_log_enable (int)
+ * Enable Micro Engine Scheduler log. This is used to enable/disable MES internal log.
+ * (0 = disabled (default), 1 = enabled)
+ */
+MODULE_PARM_DESC(mes_log_enable,
+	"Enable Micro Engine Scheduler log (0 = disabled (default), 1 = enabled)");
+module_param_named(mes_log_enable, amdgpu_mes_log_enable, int, 0444);
+
+/**
  * DOC: mes_kiq (int)
  * Enable Micro Engine Scheduler KIQ. This is a new engine pipe for kiq.
  * (0 = disabled (default), 1 = enabled)
@@ -664,6 +716,15 @@ MODULE_PARM_DESC(mes_kiq,
 module_param_named(mes_kiq, amdgpu_mes_kiq, int, 0444);
 
 /**
+ * DOC: uni_mes (int)
+ * Enable Unified Micro Engine Scheduler. This is a new engine pipe for unified scheduler.
+ * (0 = disabled (default), 1 = enabled)
+ */
+MODULE_PARM_DESC(uni_mes,
+	"Enable Unified Micro Engine Scheduler (0 = disabled, 1 = enabled(default)");
+module_param_named(uni_mes, amdgpu_uni_mes, int, 0444);
+
+/**
  * DOC: noretry (int)
  * Disable XNACK retry in the SQ by default on GFXv9 hardware. On ASICs that
  * do not support per-process XNACK this also disables retry page faults.
@@ -679,7 +740,7 @@ module_param_named(noretry, amdgpu_noretry, int, 0644);
  */
 MODULE_PARM_DESC(force_asic_type,
 	"A non negative value used to specify the asic type for all supported GPUs");
-module_param_named(force_asic_type, amdgpu_force_asic_type, int, 0444);
+module_param_named_unsafe(force_asic_type, amdgpu_force_asic_type, int, 0444);
 
 /**
  * DOC: use_xgmi_p2p (int)
@@ -698,7 +759,7 @@ module_param_named(use_xgmi_p2p, amdgpu_use_xgmi_p2p, int, 0444);
  * assigns queues to HQDs.
  */
 int sched_policy = KFD_SCHED_POLICY_HWS;
-module_param(sched_policy, int, 0444);
+module_param_unsafe(sched_policy, int, 0444);
 MODULE_PARM_DESC(sched_policy,
 	"Scheduling policy (0 = HWS (Default), 1 = HWS without over-subscription, 2 = Non-HWS (Used for debugging only)");
 
@@ -743,38 +804,12 @@ MODULE_PARM_DESC(send_sigterm,
 	"Send sigterm to HSA process on unhandled exception (0 = disable, 1 = enable)");
 
 /**
- * DOC: debug_largebar (int)
- * Set debug_largebar as 1 to enable simulating large-bar capability on non-large bar
- * system. This limits the VRAM size reported to ROCm applications to the visible
- * size, usually 256MB.
- * Default value is 0, diabled.
- */
-int debug_largebar;
-module_param(debug_largebar, int, 0444);
-MODULE_PARM_DESC(debug_largebar,
-	"Debug large-bar flag used to simulate large-bar capability on non-large bar machine (0 = disable, 1 = enable)");
-
-/**
- * DOC: ignore_crat (int)
- * Ignore CRAT table during KFD initialization. By default, KFD uses the ACPI CRAT
- * table to get information about AMD APUs. This option can serve as a workaround on
- * systems with a broken CRAT table.
- *
- * Default is auto (according to asic type, iommu_v2, and crat table, to decide
- * whether use CRAT)
- */
-int ignore_crat;
-module_param(ignore_crat, int, 0444);
-MODULE_PARM_DESC(ignore_crat,
-	"Ignore CRAT table during KFD initialization (0 = auto (default), 1 = ignore CRAT)");
-
-/**
  * DOC: halt_if_hws_hang (int)
  * Halt if HWS hang is detected. Default value, 0, disables the halt on hang.
  * Setting 1 enables halt on hang.
  */
 int halt_if_hws_hang;
-module_param(halt_if_hws_hang, int, 0644);
+module_param_unsafe(halt_if_hws_hang, int, 0644);
 MODULE_PARM_DESC(halt_if_hws_hang, "Halt if HWS hang is detected (0 = off (default), 1 = on)");
 
 /**
@@ -783,13 +818,13 @@ MODULE_PARM_DESC(halt_if_hws_hang, "Halt if HWS hang is detected (0 = off (defau
  * check says. Default value: false (rely on MEC2 firmware version check).
  */
 bool hws_gws_support;
-module_param(hws_gws_support, bool, 0444);
+module_param_unsafe(hws_gws_support, bool, 0444);
 MODULE_PARM_DESC(hws_gws_support, "Assume MEC2 FW supports GWS barriers (false = rely on FW version check (Default), true = force supported)");
 
 /**
-  * DOC: queue_preemption_timeout_ms (int)
-  * queue preemption timeout in ms (1 = Minimum, 9000 = default)
-  */
+ * DOC: queue_preemption_timeout_ms (int)
+ * queue preemption timeout in ms (1 = Minimum, 9000 = default)
+ */
 int queue_preemption_timeout_ms = 9000;
 module_param(queue_preemption_timeout_ms, int, 0644);
 MODULE_PARM_DESC(queue_preemption_timeout_ms, "queue preemption timeout in ms (1 = Minimum, 9000 = default)");
@@ -816,10 +851,17 @@ MODULE_PARM_DESC(no_system_mem_limit, "disable system memory limit (false = defa
  */
 int amdgpu_no_queue_eviction_on_vm_fault;
 MODULE_PARM_DESC(no_queue_eviction_on_vm_fault, "No queue eviction on VM fault (0 = queue eviction, 1 = no queue eviction)");
-module_param_named(no_queue_eviction_on_vm_fault, amdgpu_no_queue_eviction_on_vm_fault, int, 0444);
+module_param_named_unsafe(no_queue_eviction_on_vm_fault, amdgpu_no_queue_eviction_on_vm_fault, int, 0444);
 #endif
 
 /**
+ * DOC: mtype_local (int)
+ */
+int amdgpu_mtype_local;
+MODULE_PARM_DESC(mtype_local, "MTYPE for local memory (0 = MTYPE_RW (default), 1 = MTYPE_NC, 2 = MTYPE_CC)");
+module_param_named_unsafe(mtype_local, amdgpu_mtype_local, int, 0444);
+
+/**
  * DOC: pcie_p2p (bool)
  * Enable PCIe P2P (requires large-BAR). Default value: true (on)
  */
@@ -839,7 +881,7 @@ module_param_named(dcfeaturemask, amdgpu_dc_feature_mask, uint, 0444);
 
 /**
  * DOC: dcdebugmask (uint)
- * Override display features enabled. See enum DC_DEBUG_MASK in drivers/gpu/drm/amd/include/amd_shared.h.
+ * Display debug options. See enum DC_DEBUG_MASK in drivers/gpu/drm/amd/include/amd_shared.h.
  */
 MODULE_PARM_DESC(dcdebugmask, "all debug options disabled (default))");
 module_param_named(dcdebugmask, amdgpu_dc_debug_mask, uint, 0444);
@@ -856,18 +898,31 @@ module_param_named(visualconfirm, amdgpu_dc_visual_confirm, uint, 0444);
  * the ABM algorithm, with 1 being the least reduction and 4 being the most
  * reduction.
  *
- * Defaults to 0, or disabled. Userspace can still override this level later
- * after boot.
+ * Defaults to -1, or auto. Userspace can only override this level after
+ * boot if it's set to auto.
  */
-uint amdgpu_dm_abm_level;
-MODULE_PARM_DESC(abmlevel, "ABM level (0 = off (default), 1-4 = backlight reduction level) ");
-module_param_named(abmlevel, amdgpu_dm_abm_level, uint, 0444);
+int amdgpu_dm_abm_level = -1;
+MODULE_PARM_DESC(abmlevel,
+		 "ABM level (0 = off, 1-4 = backlight reduction level, -1 auto (default))");
+module_param_named(abmlevel, amdgpu_dm_abm_level, int, 0444);
 
 int amdgpu_backlight = -1;
 MODULE_PARM_DESC(backlight, "Backlight control (0 = pwm, 1 = aux, -1 auto (default))");
 module_param_named(backlight, amdgpu_backlight, bint, 0444);
 
 /**
+ * DOC: damageclips (int)
+ * Enable or disable damage clips support. If damage clips support is disabled,
+ * we will force full frame updates, irrespective of what user space sends to
+ * us.
+ *
+ * Defaults to -1 (where it is enabled unless a PSR-SU display is detected).
+ */
+MODULE_PARM_DESC(damageclips,
+		 "Damage clips support (0 = disable, 1 = enable, -1 auto (default))");
+module_param_named(damageclips, amdgpu_damage_clips, int, 0444);
+
+/**
  * DOC: tmz (int)
  * Trusted Memory Zone (TMZ) is a method to protect data being written
  * to or read from memory.
@@ -900,7 +955,7 @@ module_param_named(tmz, amdgpu_tmz, int, 0444);
  */
 MODULE_PARM_DESC(
 	freesync_video,
-	"Enable freesync modesetting optimization feature (0 = off (default), 1 = on)");
+	"Adds additional modes via VRR for refresh changes without a full modeset (0 = off (default), 1 = on)");
 module_param_named(freesync_video, amdgpu_freesync_vid_mode, uint, 0444);
 
 /**
@@ -908,7 +963,7 @@ module_param_named(freesync_video, amdgpu_freesync_vid_mode, uint, 0444);
  * GPU reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = mode1, 3 = mode2, 4 = baco)
  */
 MODULE_PARM_DESC(reset_method, "GPU reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = mode1, 3 = mode2, 4 = baco/bamaco)");
-module_param_named(reset_method, amdgpu_reset_method, int, 0444);
+module_param_named_unsafe(reset_method, amdgpu_reset_method, int, 0644);
 
 /**
  * DOC: bad_page_threshold (int) Bad page threshold is specifies the
@@ -916,7 +971,7 @@ module_param_named(reset_method, amdgpu_reset_method, int, 0444);
  * result in the GPU entering bad status when the number of total
  * faulty pages by ECC exceeds the threshold value.
  */
-MODULE_PARM_DESC(bad_page_threshold, "Bad page threshold(-1 = ignore threshold (default value), 0 = disable bad page retirement, -2 = driver sets threshold)");
+MODULE_PARM_DESC(bad_page_threshold, "Bad page threshold(-1 = ignore threshold (default value), 0 = disable bad page retirement, -2 = threshold determined by a formula, 0 < threshold < max records, user-defined threshold)");
 module_param_named(bad_page_threshold, amdgpu_bad_page_threshold, int, 0444);
 
 MODULE_PARM_DESC(num_kcq, "number of kernel compute queue user want to setup (8 if set to greater than 8 or less than 0, only affect gfx 8+)");
@@ -940,6 +995,22 @@ MODULE_PARM_DESC(sg_display, "S/G Display (-1 = auto (default), 0 = disable)");
 module_param_named(sg_display, amdgpu_sg_display, int, 0444);
 
 /**
+ * DOC: umsch_mm (int)
+ * Enable Multi Media User Mode Scheduler. This is a HW scheduling engine for VCN and VPE.
+ * (0 = disabled (default), 1 = enabled)
+ */
+MODULE_PARM_DESC(umsch_mm,
+	"Enable Multi Media User Mode Scheduler (0 = disabled (default), 1 = enabled)");
+module_param_named(umsch_mm, amdgpu_umsch_mm, int, 0444);
+
+/**
+ * DOC: umsch_mm_fwlog (int)
+ * Enable umschfw log output for debugging, the default is disabled.
+ */
+MODULE_PARM_DESC(umsch_mm_fwlog, "Enable umschfw log(0 = disable (default value), 1 = enable)");
+module_param_named(umsch_mm_fwlog, amdgpu_umsch_mm_fwlog, int, 0444);
+
+/**
  * DOC: smu_pptable_id (int)
  * Used to override pptable id. id = 0 use VBIOS pptable.
  * id > 0 use the soft pptable with specicfied id.
@@ -948,6 +1019,109 @@ MODULE_PARM_DESC(smu_pptable_id,
 	"specify pptable id to be used (-1 = auto(default) value, 0 = use pptable from vbios, > 0 = soft pptable id)");
 module_param_named(smu_pptable_id, amdgpu_smu_pptable_id, int, 0444);
 
+/**
+ * DOC: partition_mode (int)
+ * Used to override the default SPX mode.
+ */
+MODULE_PARM_DESC(
+	user_partt_mode,
+	"specify partition mode to be used (-2 = AMDGPU_AUTO_COMPUTE_PARTITION_MODE(default value) \
+						0 = AMDGPU_SPX_PARTITION_MODE, \
+						1 = AMDGPU_DPX_PARTITION_MODE, \
+						2 = AMDGPU_TPX_PARTITION_MODE, \
+						3 = AMDGPU_QPX_PARTITION_MODE, \
+						4 = AMDGPU_CPX_PARTITION_MODE)");
+module_param_named(user_partt_mode, amdgpu_user_partt_mode, uint, 0444);
+
+
+/**
+ * DOC: enforce_isolation (int)
+ * enforce process isolation between graphics and compute.
+ * (-1 = auto, 0 = disable, 1 = enable, 2 = enable legacy mode, 3 = enable without cleaner shader)
+ */
+module_param_named(enforce_isolation, amdgpu_enforce_isolation, int, 0444);
+MODULE_PARM_DESC(enforce_isolation,
+"enforce process isolation between graphics and compute. (-1 = auto, 0 = disable, 1 = enable, 2 = enable legacy mode, 3 = enable without cleaner shader)");
+
+/**
+ * DOC: modeset (int)
+ * Override nomodeset (1 = override, -1 = auto). The default is -1 (auto).
+ */
+MODULE_PARM_DESC(modeset, "Override nomodeset (1 = enable, -1 = auto)");
+module_param_named(modeset, amdgpu_modeset, int, 0444);
+
+/**
+ * DOC: seamless (int)
+ * Seamless boot will keep the image on the screen during the boot process.
+ */
+MODULE_PARM_DESC(seamless, "Seamless boot (-1 = auto (default), 0 = disable, 1 = enable)");
+module_param_named(seamless, amdgpu_seamless, int, 0444);
+
+/**
+ * DOC: debug_mask (uint)
+ * Debug options for amdgpu, work as a binary mask with the following options:
+ *
+ * - 0x1: Debug VM handling
+ * - 0x2: Enable simulating large-bar capability on non-large bar system. This
+ *   limits the VRAM size reported to ROCm applications to the visible
+ *   size, usually 256MB.
+ * - 0x4: Disable GPU soft recovery, always do a full reset
+ * - 0x8: Use VRAM for firmware loading
+ * - 0x10: Enable ACA based RAS logging
+ * - 0x20: Enable experimental resets
+ * - 0x40: Disable ring resets
+ * - 0x80: Use VRAM for SMU pool
+ */
+MODULE_PARM_DESC(debug_mask, "debug options for amdgpu, disabled by default");
+module_param_named_unsafe(debug_mask, amdgpu_debug_mask, uint, 0444);
+
+/**
+ * DOC: agp (int)
+ * Enable the AGP aperture.  This provides an aperture in the GPU's internal
+ * address space for direct access to system memory.  Note that these accesses
+ * are non-snooped, so they are only used for access to uncached memory.
+ */
+MODULE_PARM_DESC(agp, "AGP (-1 = auto (default), 0 = disable, 1 = enable)");
+module_param_named(agp, amdgpu_agp, int, 0444);
+
+/**
+ * DOC: wbrf (int)
+ * Enable Wifi RFI interference mitigation feature.
+ * Due to electrical and mechanical constraints there may be likely interference of
+ * relatively high-powered harmonics of the (G-)DDR memory clocks with local radio
+ * module frequency bands used by Wifi 6/6e/7. To mitigate the possible RFI interference,
+ * with this feature enabled, PMFW will use either “shadowed P-State” or “P-State” based
+ * on active list of frequencies in-use (to be avoided) as part of initial setting or
+ * P-state transition. However, there may be potential performance impact with this
+ * feature enabled.
+ * (0 = disabled, 1 = enabled, -1 = auto (default setting, will be enabled if supported))
+ */
+MODULE_PARM_DESC(wbrf,
+	"Enable Wifi RFI interference mitigation (0 = disabled, 1 = enabled, -1 = auto(default)");
+module_param_named(wbrf, amdgpu_wbrf, int, 0444);
+
+/**
+ * DOC: rebar (int)
+ * Allow BAR resizing.  Disable this to prevent the driver from attempting
+ * to resize the BAR if the GPU supports it and there is available MMIO space.
+ * Note that this just prevents the driver from resizing the BAR.  The BIOS
+ * may have already resized the BAR at boot time.
+ */
+MODULE_PARM_DESC(rebar, "Resizable BAR (-1 = auto (default), 0 = disable, 1 = enable)");
+module_param_named(rebar, amdgpu_rebar, int, 0444);
+
+/**
+ * DOC: user_queue (int)
+ * Enable user queues on systems that support user queues. Possible values:
+ *
+ * - -1 = auto (ASIC specific default)
+ * -  0 = user queues disabled
+ * -  1 = user queues enabled and kernel queues enabled (if supported)
+ * -  2 = user queues enabled and kernel queues disabled
+ */
+MODULE_PARM_DESC(user_queue, "Enable user queues (-1 = auto (default), 0 = disable, 1 = enable, 2 = enable UQs and disable KQs)");
+module_param_named(user_queue, amdgpu_user_queue, int, 0444);
+
 /* These devices are not supported by amdgpu.
  * They are supported by the mach64, r128, radeon drivers
  */
@@ -1615,6 +1789,7 @@ static const u16 amdgpu_unsupported_pciidlist[] = {
 	0x5874,
 	0x5940,
 	0x5941,
+	0x5b70,
 	0x5b72,
 	0x5b73,
 	0x5b74,
@@ -1660,7 +1835,6 @@ static const u16 amdgpu_unsupported_pciidlist[] = {
 };
 
 static const struct pci_device_id pciidlist[] = {
-#ifdef  CONFIG_DRM_AMDGPU_SI
 	{0x1002, 0x6780, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TAHITI},
 	{0x1002, 0x6784, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TAHITI},
 	{0x1002, 0x6788, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TAHITI},
@@ -1733,8 +1907,6 @@ static const struct pci_device_id pciidlist[] = {
 	{0x1002, 0x6665, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_HAINAN|AMD_IS_MOBILITY},
 	{0x1002, 0x6667, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_HAINAN|AMD_IS_MOBILITY},
 	{0x1002, 0x666F, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_HAINAN|AMD_IS_MOBILITY},
-#endif
-#ifdef CONFIG_DRM_AMDGPU_CIK
 	/* Kaveri */
 	{0x1002, 0x1304, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_KAVERI|AMD_IS_MOBILITY|AMD_IS_APU},
 	{0x1002, 0x1305, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_KAVERI|AMD_IS_APU},
@@ -1817,7 +1989,6 @@ static const struct pci_device_id pciidlist[] = {
 	{0x1002, 0x985D, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_MULLINS|AMD_IS_MOBILITY|AMD_IS_APU},
 	{0x1002, 0x985E, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_MULLINS|AMD_IS_MOBILITY|AMD_IS_APU},
 	{0x1002, 0x985F, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_MULLINS|AMD_IS_MOBILITY|AMD_IS_APU},
-#endif
 	/* topaz */
 	{0x1002, 0x6900, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TOPAZ},
 	{0x1002, 0x6901, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TOPAZ},
@@ -1996,6 +2167,11 @@ static const struct pci_device_id pciidlist[] = {
 	{0x1002, 0x7410, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_ALDEBARAN},
 
 	/* CYAN_SKILLFISH */
+	{0x1002, 0x13DB, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CYAN_SKILLFISH|AMD_IS_APU},
+	{0x1002, 0x13F9, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CYAN_SKILLFISH|AMD_IS_APU},
+	{0x1002, 0x13FA, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CYAN_SKILLFISH|AMD_IS_APU},
+	{0x1002, 0x13FB, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CYAN_SKILLFISH|AMD_IS_APU},
+	{0x1002, 0x13FC, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CYAN_SKILLFISH|AMD_IS_APU},
 	{0x1002, 0x13FE, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CYAN_SKILLFISH|AMD_IS_APU},
 	{0x1002, 0x143F, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CYAN_SKILLFISH|AMD_IS_APU},
 
@@ -2017,11 +2193,24 @@ static const struct pci_device_id pciidlist[] = {
 	  .class_mask = 0xffffff,
 	  .driver_data = CHIP_IP_DISCOVERY },
 
+	{ PCI_DEVICE(0x1002, PCI_ANY_ID),
+	  .class = PCI_CLASS_ACCELERATOR_PROCESSING << 8,
+	  .class_mask = 0xffffff,
+	  .driver_data = CHIP_IP_DISCOVERY },
+
 	{0, 0, 0}
 };
 
 MODULE_DEVICE_TABLE(pci, pciidlist);
 
+static const struct amdgpu_asic_type_quirk asic_type_quirks[] = {
+	/* differentiate between P10 and P11 asics with the same DID */
+	{0x67FF, 0xE3, CHIP_POLARIS10},
+	{0x67FF, 0xE7, CHIP_POLARIS10},
+	{0x67FF, 0xF3, CHIP_POLARIS10},
+	{0x67FF, 0xF7, CHIP_POLARIS10},
+};
+
 static const struct drm_driver amdgpu_kms_driver;
 
 static void amdgpu_get_secondary_funcs(struct amdgpu_device *adev)
@@ -2039,13 +2228,150 @@ static void amdgpu_get_secondary_funcs(struct amdgpu_device *adev)
 						adev->pdev->bus->number, i);
 		if (p) {
 			pm_runtime_get_sync(&p->dev);
-			pm_runtime_mark_last_busy(&p->dev);
 			pm_runtime_put_autosuspend(&p->dev);
 			pci_dev_put(p);
 		}
 	}
 }
 
+static void amdgpu_init_debug_options(struct amdgpu_device *adev)
+{
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_VM) {
+		pr_info("debug: VM handling debug enabled\n");
+		adev->debug_vm = true;
+	}
+
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_LARGEBAR) {
+		pr_info("debug: enabled simulating large-bar capability on non-large bar system\n");
+		adev->debug_largebar = true;
+	}
+
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_DISABLE_GPU_SOFT_RECOVERY) {
+		pr_info("debug: soft reset for GPU recovery disabled\n");
+		adev->debug_disable_soft_recovery = true;
+	}
+
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_USE_VRAM_FW_BUF) {
+		pr_info("debug: place fw in vram for frontdoor loading\n");
+		adev->debug_use_vram_fw_buf = true;
+	}
+
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_ENABLE_RAS_ACA) {
+		pr_info("debug: enable RAS ACA\n");
+		adev->debug_enable_ras_aca = true;
+	}
+
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_ENABLE_EXP_RESETS) {
+		pr_info("debug: enable experimental reset features\n");
+		adev->debug_exp_resets = true;
+	}
+
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_DISABLE_GPU_RING_RESET) {
+		pr_info("debug: ring reset disabled\n");
+		adev->debug_disable_gpu_ring_reset = true;
+	}
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_SMU_POOL) {
+		pr_info("debug: use vram for smu pool\n");
+		adev->pm.smu_debug_mask |= SMU_DEBUG_POOL_USE_VRAM;
+	}
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_VM_USERPTR) {
+		pr_info("debug: VM mode debug for userptr is enabled\n");
+		adev->debug_vm_userptr = true;
+	}
+
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_DISABLE_RAS_CE_LOG) {
+		pr_info("debug: disable kernel logs of correctable errors\n");
+		adev->debug_disable_ce_logs = true;
+	}
+
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_ENABLE_CE_CS) {
+		pr_info("debug: allowing command submission to CE engine\n");
+		adev->debug_enable_ce_cs = true;
+	}
+}
+
+static unsigned long amdgpu_fix_asic_type(struct pci_dev *pdev, unsigned long flags)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(asic_type_quirks); i++) {
+		if (pdev->device == asic_type_quirks[i].device &&
+			pdev->revision == asic_type_quirks[i].revision) {
+				flags &= ~AMD_ASIC_MASK;
+				flags |= asic_type_quirks[i].type;
+				break;
+			}
+	}
+
+	return flags;
+}
+
+static bool amdgpu_support_enabled(struct device *dev,
+				   const enum amd_asic_type family)
+{
+	const char *gen;
+	const char *param;
+	int module_param = -1;
+	bool radeon_support_built = IS_ENABLED(CONFIG_DRM_RADEON);
+	bool amdgpu_support_built = false;
+	bool support_by_default = false;
+
+	switch (family) {
+	case CHIP_TAHITI:
+	case CHIP_PITCAIRN:
+	case CHIP_VERDE:
+	case CHIP_OLAND:
+	case CHIP_HAINAN:
+		gen = "SI";
+		param = "si_support";
+		module_param = amdgpu_si_support;
+		amdgpu_support_built = IS_ENABLED(CONFIG_DRM_AMDGPU_SI);
+		support_by_default = true;
+		break;
+
+	case CHIP_BONAIRE:
+	case CHIP_HAWAII:
+		support_by_default = true;
+		fallthrough;
+	case CHIP_KAVERI:
+	case CHIP_KABINI:
+	case CHIP_MULLINS:
+		gen = "CIK";
+		param = "cik_support";
+		module_param = amdgpu_cik_support;
+		amdgpu_support_built = IS_ENABLED(CONFIG_DRM_AMDGPU_CIK);
+		break;
+
+	default:
+		/* All other chips are supported by amdgpu only */
+		return true;
+	}
+
+	if (!amdgpu_support_built) {
+		dev_info(dev, "amdgpu built without %s support\n", gen);
+		return false;
+	}
+
+	if ((module_param == -1 && (support_by_default || !radeon_support_built)) ||
+	    module_param == 1) {
+		if (radeon_support_built)
+			dev_info(dev, "%s support provided by amdgpu.\n"
+				 "Use radeon.%s=1 amdgpu.%s=0 to override.\n",
+				 gen, param, param);
+
+		return true;
+	}
+
+	if (radeon_support_built)
+		dev_info(dev, "%s support provided by radeon.\n"
+			 "Use radeon.%s=0 amdgpu.%s=1 to override.\n",
+			 gen, param, param);
+	else if (module_param == 0)
+		dev_info(dev, "%s support disabled by module param\n", gen);
+
+	return false;
+}
+
 static int amdgpu_pci_probe(struct pci_dev *pdev,
 			    const struct pci_device_id *ent)
 {
@@ -2055,6 +2381,12 @@ static int amdgpu_pci_probe(struct pci_dev *pdev,
 	int ret, retry = 0, i;
 	bool supports_atomic = false;
 
+	if ((pdev->class >> 8) == PCI_CLASS_DISPLAY_VGA ||
+	    (pdev->class >> 8) == PCI_CLASS_DISPLAY_OTHER) {
+		if (drm_firmware_drivers_only() && amdgpu_modeset == -1)
+			return -EINVAL;
+	}
+
 	/* skip devices which are owned by radeon */
 	for (i = 0; i < ARRAY_SIZE(amdgpu_unsupported_pciidlist); i++) {
 		if (amdgpu_unsupported_pciidlist[i] == pdev->device)
@@ -2065,7 +2397,7 @@ static int amdgpu_pci_probe(struct pci_dev *pdev,
 		amdgpu_aspm = 0;
 
 	if (amdgpu_virtual_display ||
-	    amdgpu_device_asic_has_dc_support(flags & AMD_ASIC_MASK))
+	    amdgpu_device_asic_has_dc_support(pdev, flags & AMD_ASIC_MASK))
 		supports_atomic = true;
 
 	if ((flags & AMD_EXP_HW_SUPPORT) && !amdgpu_exp_hw_support) {
@@ -2073,15 +2405,8 @@ static int amdgpu_pci_probe(struct pci_dev *pdev,
 			 "See modparam exp_hw_support\n");
 		return -ENODEV;
 	}
-	/* differentiate between P10 and P11 asics with the same DID */
-	if (pdev->device == 0x67FF &&
-	    (pdev->revision == 0xE3 ||
-	     pdev->revision == 0xE7 ||
-	     pdev->revision == 0xF3 ||
-	     pdev->revision == 0xF7)) {
-		flags &= ~AMD_ASIC_MASK;
-		flags |= CHIP_POLARIS10;
-	}
+
+	flags = amdgpu_fix_asic_type(pdev, flags);
 
 	/* Due to hardware bugs, S/G Display on raven requires a 1:1 IOMMU mapping,
 	 * however, SME requires an indirect IOMMU mapping because the encryption
@@ -2094,40 +2419,8 @@ static int amdgpu_pci_probe(struct pci_dev *pdev,
 		return -ENOTSUPP;
 	}
 
-#ifdef CONFIG_DRM_AMDGPU_SI
-	if (!amdgpu_si_support) {
-		switch (flags & AMD_ASIC_MASK) {
-		case CHIP_TAHITI:
-		case CHIP_PITCAIRN:
-		case CHIP_VERDE:
-		case CHIP_OLAND:
-		case CHIP_HAINAN:
-			dev_info(&pdev->dev,
-				 "SI support provided by radeon.\n");
-			dev_info(&pdev->dev,
-				 "Use radeon.si_support=0 amdgpu.si_support=1 to override.\n"
-				);
-			return -ENODEV;
-		}
-	}
-#endif
-#ifdef CONFIG_DRM_AMDGPU_CIK
-	if (!amdgpu_cik_support) {
-		switch (flags & AMD_ASIC_MASK) {
-		case CHIP_KAVERI:
-		case CHIP_BONAIRE:
-		case CHIP_HAWAII:
-		case CHIP_KABINI:
-		case CHIP_MULLINS:
-			dev_info(&pdev->dev,
-				 "CIK support provided by radeon.\n");
-			dev_info(&pdev->dev,
-				 "Use radeon.cik_support=0 amdgpu.cik_support=1 to override.\n"
-				);
-			return -ENODEV;
-		}
-	}
-#endif
+	if (!amdgpu_support_enabled(&pdev->dev, flags & AMD_ASIC_MASK))
+		return -ENODEV;
 
 	adev = devm_drm_dev_alloc(&pdev->dev, &amdgpu_kms_driver, typeof(*adev), ddev);
 	if (IS_ERR(adev))
@@ -2146,6 +2439,8 @@ static int amdgpu_pci_probe(struct pci_dev *pdev,
 
 	pci_set_drvdata(pdev, ddev);
 
+	amdgpu_init_debug_options(adev);
+
 	ret = amdgpu_driver_load_kms(adev, flags);
 	if (ret)
 		goto err_pci;
@@ -2161,17 +2456,29 @@ retry_init:
 		goto err_pci;
 	}
 
+	ret = amdgpu_xcp_dev_register(adev, ent);
+	if (ret)
+		goto err_pci;
+
+	ret = amdgpu_amdkfd_drm_client_create(adev);
+	if (ret)
+		goto err_pci;
+
 	/*
 	 * 1. don't init fbdev on hw without DCE
 	 * 2. don't init fbdev if there are no connectors
 	 */
 	if (adev->mode_info.mode_config_initialized &&
 	    !list_empty(&adev_to_drm(adev)->mode_config.connector_list)) {
+		const struct drm_format_info *format;
+
 		/* select 8 bpp console on low vram cards */
 		if (adev->gmc.real_vram_size <= (32*1024*1024))
-			drm_fbdev_generic_setup(adev_to_drm(adev), 8);
+			format = drm_format_info(DRM_FORMAT_C8);
 		else
-			drm_fbdev_generic_setup(adev_to_drm(adev), 32);
+			format = NULL;
+
+		drm_client_setup(adev_to_drm(adev), format);
 	}
 
 	ret = amdgpu_debugfs_init(adev);
@@ -2180,10 +2487,10 @@ retry_init:
 
 	if (adev->pm.rpm_mode != AMDGPU_RUNPM_NONE) {
 		/* only need to skip on ATPX */
-		if (amdgpu_device_supports_px(ddev))
+		if (amdgpu_device_supports_px(adev))
 			dev_pm_set_driver_flags(ddev->dev, DPM_FLAG_NO_DIRECT_COMPLETE);
 		/* we want direct complete for BOCO */
-		if (amdgpu_device_supports_boco(ddev))
+		if (amdgpu_device_supports_boco(adev))
 			dev_pm_set_driver_flags(ddev->dev, DPM_FLAG_SMART_PREPARE |
 						DPM_FLAG_SMART_SUSPEND |
 						DPM_FLAG_MAY_SKIP_RESUME);
@@ -2192,9 +2499,10 @@ retry_init:
 
 		pm_runtime_allow(ddev->dev);
 
-		pm_runtime_mark_last_busy(ddev->dev);
 		pm_runtime_put_autosuspend(ddev->dev);
 
+		pci_wake_from_d3(pdev, TRUE);
+
 		/*
 		 * For runpm implemented via BACO, PMFW will handle the
 		 * timing for BACO in and out:
@@ -2214,9 +2522,9 @@ retry_init:
 		 * into D0 state. Then there will be a PMFW-aware D-state
 		 * transition(D0->D3) on runpm suspend.
 		 */
-		if (amdgpu_device_supports_baco(ddev) &&
+		if (amdgpu_device_supports_baco(adev) &&
 		    !(adev->flags & AMD_IS_APU) &&
-		    (adev->asic_type >= CHIP_NAVI10))
+		    adev->asic_type >= CHIP_NAVI10)
 			amdgpu_get_secondary_funcs(adev);
 	}
 
@@ -2233,6 +2541,9 @@ amdgpu_pci_remove(struct pci_dev *pdev)
 	struct drm_device *dev = pci_get_drvdata(pdev);
 	struct amdgpu_device *adev = drm_to_adev(dev);
 
+	amdgpu_ras_eeprom_check_and_recover(adev);
+	amdgpu_xcp_dev_unplug(adev);
+	amdgpu_gmc_prepare_nps_mode_change(adev);
 	drm_dev_unplug(dev);
 
 	if (adev->pm.rpm_mode != AMDGPU_RUNPM_NONE) {
@@ -2240,38 +2551,6 @@ amdgpu_pci_remove(struct pci_dev *pdev)
 		pm_runtime_forbid(dev->dev);
 	}
 
-	if (adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 2) &&
-	    !amdgpu_sriov_vf(adev)) {
-		bool need_to_reset_gpu = false;
-
-		if (adev->gmc.xgmi.num_physical_nodes > 1) {
-			struct amdgpu_hive_info *hive;
-
-			hive = amdgpu_get_xgmi_hive(adev);
-			if (hive->device_remove_count == 0)
-				need_to_reset_gpu = true;
-			hive->device_remove_count++;
-			amdgpu_put_xgmi_hive(hive);
-		} else {
-			need_to_reset_gpu = true;
-		}
-
-		/* Workaround for ASICs need to reset SMU.
-		 * Called only when the first device is removed.
-		 */
-		if (need_to_reset_gpu) {
-			struct amdgpu_reset_context reset_context;
-
-			adev->shutdown = true;
-			memset(&reset_context, 0, sizeof(reset_context));
-			reset_context.method = AMD_RESET_METHOD_NONE;
-			reset_context.reset_req_dev = adev;
-			set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
-			set_bit(AMDGPU_RESET_FOR_DEVICE_REMOVE, &reset_context.flags);
-			amdgpu_device_gpu_recover(adev, NULL, &reset_context);
-		}
-	}
-
 	amdgpu_driver_unload_kms(dev);
 
 	/*
@@ -2292,6 +2571,10 @@ amdgpu_pci_shutdown(struct pci_dev *pdev)
 	if (amdgpu_ras_intr_triggered())
 		return;
 
+	/* device maybe not resumed here, return immediately in this case */
+	if (adev->in_s4 && adev->in_suspend)
+		return;
+
 	/* if we are running in a VM, make sure the device
 	 * torn down properly on reboot/shutdown.
 	 * unfortunately we can't detect certain
@@ -2299,93 +2582,25 @@ amdgpu_pci_shutdown(struct pci_dev *pdev)
 	 */
 	if (!amdgpu_passthrough(adev))
 		adev->mp1_state = PP_MP1_STATE_UNLOAD;
-	amdgpu_device_ip_suspend(adev);
+	amdgpu_device_prepare(dev);
+	amdgpu_device_suspend(dev, true);
 	adev->mp1_state = PP_MP1_STATE_NONE;
 }
 
-/**
- * amdgpu_drv_delayed_reset_work_handler - work handler for reset
- *
- * @work: work_struct.
- */
-static void amdgpu_drv_delayed_reset_work_handler(struct work_struct *work)
-{
-	struct list_head device_list;
-	struct amdgpu_device *adev;
-	int i, r;
-	struct amdgpu_reset_context reset_context;
-
-	memset(&reset_context, 0, sizeof(reset_context));
-
-	mutex_lock(&mgpu_info.mutex);
-	if (mgpu_info.pending_reset == true) {
-		mutex_unlock(&mgpu_info.mutex);
-		return;
-	}
-	mgpu_info.pending_reset = true;
-	mutex_unlock(&mgpu_info.mutex);
-
-	/* Use a common context, just need to make sure full reset is done */
-	reset_context.method = AMD_RESET_METHOD_NONE;
-	set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
-
-	for (i = 0; i < mgpu_info.num_dgpu; i++) {
-		adev = mgpu_info.gpu_ins[i].adev;
-		reset_context.reset_req_dev = adev;
-		r = amdgpu_device_pre_asic_reset(adev, &reset_context);
-		if (r) {
-			dev_err(adev->dev, "GPU pre asic reset failed with err, %d for drm dev, %s ",
-				r, adev_to_drm(adev)->unique);
-		}
-		if (!queue_work(system_unbound_wq, &adev->xgmi_reset_work))
-			r = -EALREADY;
-	}
-	for (i = 0; i < mgpu_info.num_dgpu; i++) {
-		adev = mgpu_info.gpu_ins[i].adev;
-		flush_work(&adev->xgmi_reset_work);
-		adev->gmc.xgmi.pending_reset = false;
-	}
-
-	/* reset function will rebuild the xgmi hive info , clear it now */
-	for (i = 0; i < mgpu_info.num_dgpu; i++)
-		amdgpu_xgmi_remove_device(mgpu_info.gpu_ins[i].adev);
-
-	INIT_LIST_HEAD(&device_list);
-
-	for (i = 0; i < mgpu_info.num_dgpu; i++)
-		list_add_tail(&mgpu_info.gpu_ins[i].adev->reset_list, &device_list);
-
-	/* unregister the GPU first, reset function will add them back */
-	list_for_each_entry(adev, &device_list, reset_list)
-		amdgpu_unregister_gpu_instance(adev);
-
-	/* Use a common context, just need to make sure full reset is done */
-	set_bit(AMDGPU_SKIP_HW_RESET, &reset_context.flags);
-	r = amdgpu_do_asic_reset(&device_list, &reset_context);
-
-	if (r) {
-		DRM_ERROR("reinit gpus failure");
-		return;
-	}
-	for (i = 0; i < mgpu_info.num_dgpu; i++) {
-		adev = mgpu_info.gpu_ins[i].adev;
-		if (!adev->kfd.init_complete)
-			amdgpu_amdkfd_device_init(adev);
-		amdgpu_ttm_set_buffer_funcs_status(adev, true);
-	}
-	return;
-}
-
 static int amdgpu_pmops_prepare(struct device *dev)
 {
 	struct drm_device *drm_dev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = drm_to_adev(drm_dev);
 
+	/* device maybe not resumed here, return immediately in this case */
+	if (adev->in_s4 && adev->in_suspend)
+		return 0;
+
 	/* Return a positive number here so
 	 * DPM_FLAG_SMART_SUSPEND works properly
 	 */
-	if (amdgpu_device_supports_boco(drm_dev))
-		return pm_runtime_suspended(dev);
+	if (amdgpu_device_supports_boco(adev) && pm_runtime_suspended(dev))
+		return 1;
 
 	/* if we will not support s3 or s2i for the device
 	 *  then skip suspend
@@ -2394,12 +2609,12 @@ static int amdgpu_pmops_prepare(struct device *dev)
 	    !amdgpu_acpi_is_s3_active(adev))
 		return 1;
 
-	return 0;
+	return amdgpu_device_prepare(drm_dev);
 }
 
 static void amdgpu_pmops_complete(struct device *dev)
 {
-	/* nothing to do */
+	amdgpu_device_complete(dev_get_drvdata(dev));
 }
 
 static int amdgpu_pmops_suspend(struct device *dev)
@@ -2411,8 +2626,24 @@ static int amdgpu_pmops_suspend(struct device *dev)
 		adev->in_s0ix = true;
 	else if (amdgpu_acpi_is_s3_active(adev))
 		adev->in_s3 = true;
-	if (!adev->in_s0ix && !adev->in_s3)
+	if (!adev->in_s0ix && !adev->in_s3) {
+#if IS_ENABLED(CONFIG_SUSPEND)
+		/* don't allow going deep first time followed by s2idle the next time */
+		if (adev->last_suspend_state != PM_SUSPEND_ON &&
+		    adev->last_suspend_state != pm_suspend_target_state) {
+			drm_err_once(drm_dev, "Unsupported suspend state %d\n",
+				     pm_suspend_target_state);
+			return -EINVAL;
+		}
+#endif
 		return 0;
+	}
+
+#if IS_ENABLED(CONFIG_SUSPEND)
+	/* cache the state last used for suspend */
+	adev->last_suspend_state = pm_suspend_target_state;
+#endif
+
 	return amdgpu_device_suspend(drm_dev, true);
 }
 
@@ -2420,9 +2651,14 @@ static int amdgpu_pmops_suspend_noirq(struct device *dev)
 {
 	struct drm_device *drm_dev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = drm_to_adev(drm_dev);
+	int r;
 
-	if (amdgpu_acpi_should_gpu_reset(adev))
-		return amdgpu_asic_reset(adev);
+	if (amdgpu_acpi_should_gpu_reset(adev)) {
+		amdgpu_device_lock_reset_domain(adev->reset_domain);
+		r = amdgpu_asic_reset(adev);
+		amdgpu_device_unlock_reset_domain(adev->reset_domain);
+		return r;
+	}
 
 	return 0;
 }
@@ -2454,9 +2690,7 @@ static int amdgpu_pmops_freeze(struct device *dev)
 	struct amdgpu_device *adev = drm_to_adev(drm_dev);
 	int r;
 
-	adev->in_s4 = true;
 	r = amdgpu_device_suspend(drm_dev, true);
-	adev->in_s4 = false;
 	if (r)
 		return r;
 
@@ -2469,12 +2703,21 @@ static int amdgpu_pmops_thaw(struct device *dev)
 {
 	struct drm_device *drm_dev = dev_get_drvdata(dev);
 
+	/* do not resume device if it's normal hibernation */
+	if (!pm_hibernate_is_recovering() && !pm_hibernation_mode_is_suspend())
+		return 0;
+
 	return amdgpu_device_resume(drm_dev, true);
 }
 
 static int amdgpu_pmops_poweroff(struct device *dev)
 {
 	struct drm_device *drm_dev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(drm_dev);
+
+	/* device maybe not resumed here, return immediately in this case */
+	if (adev->in_s4 && adev->in_suspend)
+		return 0;
 
 	return amdgpu_device_suspend(drm_dev, true);
 }
@@ -2497,24 +2740,26 @@ static int amdgpu_runtime_idle_check_display(struct device *dev)
 		struct drm_connector_list_iter iter;
 		int ret = 0;
 
-		/* XXX: Return busy if any displays are connected to avoid
-		 * possible display wakeups after runtime resume due to
-		 * hotplug events in case any displays were connected while
-		 * the GPU was in suspend.  Remove this once that is fixed.
-		 */
-		mutex_lock(&drm_dev->mode_config.mutex);
-		drm_connector_list_iter_begin(drm_dev, &iter);
-		drm_for_each_connector_iter(list_connector, &iter) {
-			if (list_connector->status == connector_status_connected) {
-				ret = -EBUSY;
-				break;
+		if (amdgpu_runtime_pm != -2) {
+			/* XXX: Return busy if any displays are connected to avoid
+			 * possible display wakeups after runtime resume due to
+			 * hotplug events in case any displays were connected while
+			 * the GPU was in suspend.  Remove this once that is fixed.
+			 */
+			mutex_lock(&drm_dev->mode_config.mutex);
+			drm_connector_list_iter_begin(drm_dev, &iter);
+			drm_for_each_connector_iter(list_connector, &iter) {
+				if (list_connector->status == connector_status_connected) {
+					ret = -EBUSY;
+					break;
+				}
 			}
-		}
-		drm_connector_list_iter_end(&iter);
-		mutex_unlock(&drm_dev->mode_config.mutex);
+			drm_connector_list_iter_end(&iter);
+			mutex_unlock(&drm_dev->mode_config.mutex);
 
-		if (ret)
-			return ret;
+			if (ret)
+				return ret;
+		}
 
 		if (adev->dc_enabled) {
 			struct drm_crtc *crtc;
@@ -2551,6 +2796,15 @@ static int amdgpu_runtime_idle_check_display(struct device *dev)
 	return 0;
 }
 
+static int amdgpu_runtime_idle_check_userq(struct device *dev)
+{
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct drm_device *drm_dev = pci_get_drvdata(pdev);
+	struct amdgpu_device *adev = drm_to_adev(drm_dev);
+
+	return xa_empty(&adev->userq_doorbell_xa) ? 0 : -EBUSY;
+}
+
 static int amdgpu_pmops_runtime_suspend(struct device *dev)
 {
 	struct pci_dev *pdev = to_pci_dev(dev);
@@ -2566,10 +2820,14 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev)
 	ret = amdgpu_runtime_idle_check_display(dev);
 	if (ret)
 		return ret;
+	ret = amdgpu_runtime_idle_check_userq(dev);
+	if (ret)
+		return ret;
 
 	/* wait for all rings to drain before suspending */
 	for (i = 0; i < AMDGPU_MAX_RINGS; i++) {
 		struct amdgpu_ring *ring = adev->rings[i];
+
 		if (ring && ring->sched.ready) {
 			ret = amdgpu_fence_wait_empty(ring);
 			if (ret)
@@ -2578,7 +2836,7 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev)
 	}
 
 	adev->in_runpm = true;
-	if (amdgpu_device_supports_px(drm_dev))
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX)
 		drm_dev->switch_power_state = DRM_SWITCH_POWER_CHANGING;
 
 	/*
@@ -2588,21 +2846,24 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev)
 	 * platforms.
 	 * TODO: this may be also needed for PX capable platform.
 	 */
-	if (amdgpu_device_supports_boco(drm_dev))
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO)
 		adev->mp1_state = PP_MP1_STATE_UNLOAD;
 
+	ret = amdgpu_device_prepare(drm_dev);
+	if (ret)
+		return ret;
 	ret = amdgpu_device_suspend(drm_dev, false);
 	if (ret) {
 		adev->in_runpm = false;
-		if (amdgpu_device_supports_boco(drm_dev))
+		if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO)
 			adev->mp1_state = PP_MP1_STATE_NONE;
 		return ret;
 	}
 
-	if (amdgpu_device_supports_boco(drm_dev))
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO)
 		adev->mp1_state = PP_MP1_STATE_NONE;
 
-	if (amdgpu_device_supports_px(drm_dev)) {
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX) {
 		/* Only need to handle PCI state in the driver for ATPX
 		 * PCI core handles it for _PR3.
 		 */
@@ -2611,10 +2872,11 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev)
 		pci_ignore_hotplug(pdev);
 		pci_set_power_state(pdev, PCI_D3cold);
 		drm_dev->switch_power_state = DRM_SWITCH_POWER_DYNAMIC_OFF;
-	} else if (amdgpu_device_supports_boco(drm_dev)) {
+	} else if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO) {
 		/* nothing to do */
-	} else if (amdgpu_device_supports_baco(drm_dev)) {
-		amdgpu_device_baco_enter(drm_dev);
+	} else if ((adev->pm.rpm_mode == AMDGPU_RUNPM_BACO) ||
+			(adev->pm.rpm_mode == AMDGPU_RUNPM_BAMACO)) {
+		amdgpu_device_baco_enter(adev);
 	}
 
 	dev_dbg(&pdev->dev, "asic/device is runtime suspended\n");
@@ -2636,7 +2898,7 @@ static int amdgpu_pmops_runtime_resume(struct device *dev)
 	if (!pci_device_is_present(adev->pdev))
 		adev->no_hw_access = true;
 
-	if (amdgpu_device_supports_px(drm_dev)) {
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX) {
 		drm_dev->switch_power_state = DRM_SWITCH_POWER_CHANGING;
 
 		/* Only need to handle PCI state in the driver for ATPX
@@ -2648,22 +2910,23 @@ static int amdgpu_pmops_runtime_resume(struct device *dev)
 		if (ret)
 			return ret;
 		pci_set_master(pdev);
-	} else if (amdgpu_device_supports_boco(drm_dev)) {
+	} else if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO) {
 		/* Only need to handle PCI state in the driver for ATPX
 		 * PCI core handles it for _PR3.
 		 */
 		pci_set_master(pdev);
-	} else if (amdgpu_device_supports_baco(drm_dev)) {
-		amdgpu_device_baco_exit(drm_dev);
+	} else if ((adev->pm.rpm_mode == AMDGPU_RUNPM_BACO) ||
+			(adev->pm.rpm_mode == AMDGPU_RUNPM_BAMACO)) {
+		amdgpu_device_baco_exit(adev);
 	}
 	ret = amdgpu_device_resume(drm_dev, false);
 	if (ret) {
-		if (amdgpu_device_supports_px(drm_dev))
+		if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX)
 			pci_disable_device(pdev);
 		return ret;
 	}
 
-	if (amdgpu_device_supports_px(drm_dev))
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX)
 		drm_dev->switch_power_state = DRM_SWITCH_POWER_ON;
 	adev->in_runpm = false;
 	return 0;
@@ -2673,8 +2936,7 @@ static int amdgpu_pmops_runtime_idle(struct device *dev)
 {
 	struct drm_device *drm_dev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = drm_to_adev(drm_dev);
-	/* we don't want the main rpm_idle to call suspend - we want to autosuspend */
-	int ret = 1;
+	int ret;
 
 	if (adev->pm.rpm_mode == AMDGPU_RUNPM_NONE) {
 		pm_runtime_forbid(dev);
@@ -2682,18 +2944,39 @@ static int amdgpu_pmops_runtime_idle(struct device *dev)
 	}
 
 	ret = amdgpu_runtime_idle_check_display(dev);
+	if (ret)
+		goto done;
 
-	pm_runtime_mark_last_busy(dev);
+	ret = amdgpu_runtime_idle_check_userq(dev);
+done:
 	pm_runtime_autosuspend(dev);
 	return ret;
 }
 
+static int amdgpu_drm_release(struct inode *inode, struct file *filp)
+{
+	struct drm_file *file_priv = filp->private_data;
+	struct amdgpu_fpriv *fpriv = file_priv->driver_priv;
+	struct drm_device *dev = file_priv->minor->dev;
+	int idx;
+
+	if (fpriv && drm_dev_enter(dev, &idx)) {
+		fpriv->evf_mgr.fd_closing = true;
+		amdgpu_eviction_fence_destroy(&fpriv->evf_mgr);
+		amdgpu_userq_mgr_fini(&fpriv->userq_mgr);
+		drm_dev_exit(idx);
+	}
+
+	return drm_release(inode, filp);
+}
+
 long amdgpu_drm_ioctl(struct file *filp,
 		      unsigned int cmd, unsigned long arg)
 {
 	struct drm_file *file_priv = filp->private_data;
 	struct drm_device *dev;
 	long ret;
+
 	dev = file_priv->minor->dev;
 	ret = pm_runtime_get_sync(dev->dev);
 	if (ret < 0)
@@ -2701,22 +2984,21 @@ long amdgpu_drm_ioctl(struct file *filp,
 
 	ret = drm_ioctl(filp, cmd, arg);
 
-	pm_runtime_mark_last_busy(dev->dev);
 out:
 	pm_runtime_put_autosuspend(dev->dev);
 	return ret;
 }
 
 static const struct dev_pm_ops amdgpu_pm_ops = {
-	.prepare = amdgpu_pmops_prepare,
-	.complete = amdgpu_pmops_complete,
-	.suspend = amdgpu_pmops_suspend,
-	.suspend_noirq = amdgpu_pmops_suspend_noirq,
-	.resume = amdgpu_pmops_resume,
-	.freeze = amdgpu_pmops_freeze,
-	.thaw = amdgpu_pmops_thaw,
-	.poweroff = amdgpu_pmops_poweroff,
-	.restore = amdgpu_pmops_restore,
+	.prepare = pm_sleep_ptr(amdgpu_pmops_prepare),
+	.complete = pm_sleep_ptr(amdgpu_pmops_complete),
+	.suspend = pm_sleep_ptr(amdgpu_pmops_suspend),
+	.suspend_noirq = pm_sleep_ptr(amdgpu_pmops_suspend_noirq),
+	.resume = pm_sleep_ptr(amdgpu_pmops_resume),
+	.freeze = pm_sleep_ptr(amdgpu_pmops_freeze),
+	.thaw = pm_sleep_ptr(amdgpu_pmops_thaw),
+	.poweroff = pm_sleep_ptr(amdgpu_pmops_poweroff),
+	.restore = pm_sleep_ptr(amdgpu_pmops_restore),
 	.runtime_suspend = amdgpu_pmops_runtime_suspend,
 	.runtime_resume = amdgpu_pmops_runtime_resume,
 	.runtime_idle = amdgpu_pmops_runtime_idle,
@@ -2738,7 +3020,7 @@ static const struct file_operations amdgpu_driver_kms_fops = {
 	.owner = THIS_MODULE,
 	.open = drm_open,
 	.flush = amdgpu_flush,
-	.release = drm_release,
+	.release = amdgpu_drm_release,
 	.unlocked_ioctl = amdgpu_drm_ioctl,
 	.mmap = drm_gem_mmap,
 	.poll = drm_poll,
@@ -2747,8 +3029,9 @@ static const struct file_operations amdgpu_driver_kms_fops = {
 	.compat_ioctl = amdgpu_kms_compat_ioctl,
 #endif
 #ifdef CONFIG_PROC_FS
-	.show_fdinfo = amdgpu_show_fdinfo
+	.show_fdinfo = drm_show_fdinfo,
 #endif
+	.fop_flags = FOP_UNSIGNED_OFFSET,
 };
 
 int amdgpu_file_to_fpriv(struct file *filp, struct amdgpu_fpriv **fpriv)
@@ -2758,9 +3041,8 @@ int amdgpu_file_to_fpriv(struct file *filp, struct amdgpu_fpriv **fpriv)
 	if (!filp)
 		return -EINVAL;
 
-	if (filp->f_op != &amdgpu_driver_kms_fops) {
+	if (filp->f_op != &amdgpu_driver_kms_fops)
 		return -EINVAL;
-	}
 
 	file = filp->private_data;
 	*fpriv = file->driver_priv;
@@ -2785,6 +3067,10 @@ const struct drm_ioctl_desc amdgpu_ioctls_kms[] = {
 	DRM_IOCTL_DEF_DRV(AMDGPU_GEM_VA, amdgpu_gem_va_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	DRM_IOCTL_DEF_DRV(AMDGPU_GEM_OP, amdgpu_gem_op_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	DRM_IOCTL_DEF_DRV(AMDGPU_GEM_USERPTR, amdgpu_gem_userptr_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(AMDGPU_USERQ, amdgpu_userq_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(AMDGPU_USERQ_SIGNAL, amdgpu_userq_signal_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(AMDGPU_USERQ_WAIT, amdgpu_userq_wait_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(AMDGPU_GEM_LIST_HANDLES, amdgpu_gem_list_handles_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 };
 
 static const struct drm_driver amdgpu_kms_driver = {
@@ -2795,22 +3081,44 @@ static const struct drm_driver amdgpu_kms_driver = {
 	    DRIVER_SYNCOBJ_TIMELINE,
 	.open = amdgpu_driver_open_kms,
 	.postclose = amdgpu_driver_postclose_kms,
-	.lastclose = amdgpu_driver_lastclose_kms,
 	.ioctls = amdgpu_ioctls_kms,
 	.num_ioctls = ARRAY_SIZE(amdgpu_ioctls_kms),
 	.dumb_create = amdgpu_mode_dumb_create,
 	.dumb_map_offset = amdgpu_mode_dumb_mmap,
+	DRM_FBDEV_TTM_DRIVER_OPS,
+	.fops = &amdgpu_driver_kms_fops,
+	.release = &amdgpu_driver_release_kms,
+#ifdef CONFIG_PROC_FS
+	.show_fdinfo = amdgpu_show_fdinfo,
+#endif
+
+	.gem_prime_import = amdgpu_gem_prime_import,
+
+	.name = DRIVER_NAME,
+	.desc = DRIVER_DESC,
+	.major = KMS_DRIVER_MAJOR,
+	.minor = KMS_DRIVER_MINOR,
+	.patchlevel = KMS_DRIVER_PATCHLEVEL,
+};
+
+const struct drm_driver amdgpu_partition_driver = {
+	.driver_features =
+	    DRIVER_GEM | DRIVER_RENDER | DRIVER_SYNCOBJ |
+	    DRIVER_SYNCOBJ_TIMELINE,
+	.open = amdgpu_driver_open_kms,
+	.postclose = amdgpu_driver_postclose_kms,
+	.ioctls = amdgpu_ioctls_kms,
+	.num_ioctls = ARRAY_SIZE(amdgpu_ioctls_kms),
+	.dumb_create = amdgpu_mode_dumb_create,
+	.dumb_map_offset = amdgpu_mode_dumb_mmap,
+	DRM_FBDEV_TTM_DRIVER_OPS,
 	.fops = &amdgpu_driver_kms_fops,
 	.release = &amdgpu_driver_release_kms,
 
-	.prime_handle_to_fd = drm_gem_prime_handle_to_fd,
-	.prime_fd_to_handle = drm_gem_prime_fd_to_handle,
 	.gem_prime_import = amdgpu_gem_prime_import,
-	.gem_prime_mmap = drm_gem_prime_mmap,
 
 	.name = DRIVER_NAME,
 	.desc = DRIVER_DESC,
-	.date = DRIVER_DATE,
 	.major = KMS_DRIVER_MAJOR,
 	.minor = KMS_DRIVER_MINOR,
 	.patchlevel = KMS_DRIVER_PATCHLEVEL,
@@ -2823,25 +3131,20 @@ static struct pci_error_handlers amdgpu_pci_err_handler = {
 	.resume		= amdgpu_pci_resume,
 };
 
-extern const struct attribute_group amdgpu_vram_mgr_attr_group;
-extern const struct attribute_group amdgpu_gtt_mgr_attr_group;
-extern const struct attribute_group amdgpu_vbios_version_attr_group;
-
 static const struct attribute_group *amdgpu_sysfs_groups[] = {
 	&amdgpu_vram_mgr_attr_group,
 	&amdgpu_gtt_mgr_attr_group,
-	&amdgpu_vbios_version_attr_group,
+	&amdgpu_flash_attr_group,
 	NULL,
 };
 
-
 static struct pci_driver amdgpu_kms_pci_driver = {
 	.name = DRIVER_NAME,
 	.id_table = pciidlist,
 	.probe = amdgpu_pci_probe,
 	.remove = amdgpu_pci_remove,
 	.shutdown = amdgpu_pci_shutdown,
-	.driver.pm = &amdgpu_pm_ops,
+	.driver.pm = pm_ptr(&amdgpu_pm_ops),
 	.err_handler = &amdgpu_pci_err_handler,
 	.dev_groups = amdgpu_sysfs_groups,
 };
@@ -2850,14 +3153,11 @@ static int __init amdgpu_init(void)
 {
 	int r;
 
-	if (drm_firmware_drivers_only())
-		return -EINVAL;
-
 	r = amdgpu_sync_init();
 	if (r)
 		goto error_sync;
 
-	r = amdgpu_fence_slab_init();
+	r = amdgpu_userq_fence_slab_init();
 	if (r)
 		goto error_fence;
 
@@ -2868,6 +3168,12 @@ static int __init amdgpu_init(void)
 	/* Ignore KFD init failures. Normal when CONFIG_HSA_AMD is not set. */
 	amdgpu_amdkfd_init();
 
+	if (amdgpu_pp_feature_mask & PP_OVERDRIVE_MASK) {
+		add_taint(TAINT_CPU_OUT_OF_SPEC, LOCKDEP_STILL_OK);
+		pr_crit("Overdrive is enabled, please disable it before "
+			"reporting any bugs unrelated to overdrive.\n");
+	}
+
 	/* let modprobe override vga console setting */
 	return pci_register_driver(&amdgpu_kms_pci_driver);
 
@@ -2883,9 +3189,11 @@ static void __exit amdgpu_exit(void)
 	amdgpu_amdkfd_fini();
 	pci_unregister_driver(&amdgpu_kms_pci_driver);
 	amdgpu_unregister_atpx_handler();
+	amdgpu_acpi_release();
 	amdgpu_sync_fini();
-	amdgpu_fence_slab_fini();
+	amdgpu_userq_fence_slab_fini();
 	mmu_notifier_synchronize();
+	amdgpu_xcp_drv_release();
 }
 
 module_init(amdgpu_init);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.h
index 8178323e4bef..2d86cc6f7f4d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.h
@@ -40,7 +40,8 @@
 
 #define DRIVER_NAME		"amdgpu"
 #define DRIVER_DESC		"AMD GPU"
-#define DRIVER_DATE		"20150101"
+
+extern const struct drm_driver amdgpu_partition_driver;
 
 long amdgpu_drm_ioctl(struct file *filp,
 		      unsigned int cmd, unsigned long arg);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.c
index 7d2a908438e9..8cd69836dd99 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.c
@@ -90,7 +90,7 @@
 #define MAKE_I2C_ADDR(_aa) ((0xA << 3) | (((_aa) >> 16) & 0xF))
 
 static int __amdgpu_eeprom_xfer(struct i2c_adapter *i2c_adap, u32 eeprom_addr,
-				u8 *eeprom_buf, u16 buf_size, bool read)
+				u8 *eeprom_buf, u32 buf_size, bool read)
 {
 	u8 eeprom_offset_buf[EEPROM_OFFSET_SIZE];
 	struct i2c_msg msgs[] = {
@@ -133,15 +133,15 @@ static int __amdgpu_eeprom_xfer(struct i2c_adapter *i2c_adap, u32 eeprom_addr,
 			 * cycle begins. This is implied for the
 			 * "i2c_transfer()" abstraction.
 			 */
-			len = min(EEPROM_PAGE_SIZE - (eeprom_addr &
-						      EEPROM_PAGE_MASK),
-				  (u32)buf_size);
+			len = min(EEPROM_PAGE_SIZE - (eeprom_addr & EEPROM_PAGE_MASK),
+					buf_size);
 		} else {
 			/* Reading from the EEPROM has no limitation
 			 * on the number of bytes read from the EEPROM
 			 * device--they are simply sequenced out.
+			 * Keep in mind that i2c_msg.len is u16 type.
 			 */
-			len = buf_size;
+			len = min(U16_MAX, buf_size);
 		}
 		msgs[1].len = len;
 		msgs[1].buf = eeprom_buf;
@@ -179,10 +179,12 @@ static int __amdgpu_eeprom_xfer(struct i2c_adapter *i2c_adap, u32 eeprom_addr,
  * Returns the number of bytes read/written; -errno on error.
  */
 static int amdgpu_eeprom_xfer(struct i2c_adapter *i2c_adap, u32 eeprom_addr,
-			      u8 *eeprom_buf, u16 buf_size, bool read)
+			      u8 *eeprom_buf, u32 buf_size, bool read)
 {
 	const struct i2c_adapter_quirks *quirks = i2c_adap->quirks;
 	u16 limit;
+	u16 ps; /* Partial size */
+	int res = 0, r;
 
 	if (!quirks)
 		limit = 0;
@@ -198,35 +200,32 @@ static int amdgpu_eeprom_xfer(struct i2c_adapter *i2c_adap, u32 eeprom_addr,
 		dev_err_ratelimited(&i2c_adap->dev,
 				    "maddr:0x%04X size:0x%02X:quirk max_%s_len must be > %d",
 				    eeprom_addr, buf_size,
-				    read ? "read" : "write", EEPROM_OFFSET_SIZE);
+				    str_read_write(read), EEPROM_OFFSET_SIZE);
 		return -EINVAL;
-	} else {
-		u16 ps; /* Partial size */
-		int res = 0, r;
-
-		/* The "limit" includes all data bytes sent/received,
-		 * which would include the EEPROM_OFFSET_SIZE bytes.
-		 * Account for them here.
-		 */
-		limit -= EEPROM_OFFSET_SIZE;
-		for ( ; buf_size > 0;
-		      buf_size -= ps, eeprom_addr += ps, eeprom_buf += ps) {
-			ps = min(limit, buf_size);
-
-			r = __amdgpu_eeprom_xfer(i2c_adap, eeprom_addr,
-						 eeprom_buf, ps, read);
-			if (r < 0)
-				return r;
-			res += r;
-		}
+	}
 
-		return res;
+	/* The "limit" includes all data bytes sent/received,
+	 * which would include the EEPROM_OFFSET_SIZE bytes.
+	 * Account for them here.
+	 */
+	limit -= EEPROM_OFFSET_SIZE;
+	for ( ; buf_size > 0;
+	      buf_size -= ps, eeprom_addr += ps, eeprom_buf += ps) {
+		ps = min(limit, buf_size);
+
+		r = __amdgpu_eeprom_xfer(i2c_adap, eeprom_addr,
+					 eeprom_buf, ps, read);
+		if (r < 0)
+			return r;
+		res += r;
 	}
+
+	return res;
 }
 
 int amdgpu_eeprom_read(struct i2c_adapter *i2c_adap,
 		       u32 eeprom_addr, u8 *eeprom_buf,
-		       u16 bytes)
+		       u32 bytes)
 {
 	return amdgpu_eeprom_xfer(i2c_adap, eeprom_addr, eeprom_buf, bytes,
 				  true);
@@ -234,7 +233,7 @@ int amdgpu_eeprom_read(struct i2c_adapter *i2c_adap,
 
 int amdgpu_eeprom_write(struct i2c_adapter *i2c_adap,
 			u32 eeprom_addr, u8 *eeprom_buf,
-			u16 bytes)
+			u32 bytes)
 {
 	return amdgpu_eeprom_xfer(i2c_adap, eeprom_addr, eeprom_buf, bytes,
 				  false);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.h
index 6935adb2be1f..8083b8253ef4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.h
@@ -28,10 +28,10 @@
 
 int amdgpu_eeprom_read(struct i2c_adapter *i2c_adap,
 		       u32 eeprom_addr, u8 *eeprom_buf,
-		       u16 bytes);
+		       u32 bytes);
 
 int amdgpu_eeprom_write(struct i2c_adapter *i2c_adap,
 			u32 eeprom_addr, u8 *eeprom_buf,
-			u16 bytes);
+			u32 bytes);
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_encoders.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_encoders.c
index 27a782a9dc72..3aaeed2d3562 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_encoders.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_encoders.c
@@ -70,6 +70,7 @@ void amdgpu_encoder_set_active_device(struct drm_encoder *encoder)
 	drm_for_each_connector_iter(connector, &iter) {
 		if (connector->encoder == encoder) {
 			struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
+
 			amdgpu_encoder->active_device = amdgpu_encoder->devices & amdgpu_connector->devices;
 			DRM_DEBUG_KMS("setting active device to %08x from %08x %08x for encoder %d\n",
 				  amdgpu_encoder->active_device, amdgpu_encoder->devices,
@@ -165,12 +166,12 @@ void amdgpu_panel_mode_fixup(struct drm_encoder *encoder,
 {
 	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
 	struct drm_display_mode *native_mode = &amdgpu_encoder->native_mode;
-	unsigned hblank = native_mode->htotal - native_mode->hdisplay;
-	unsigned vblank = native_mode->vtotal - native_mode->vdisplay;
-	unsigned hover = native_mode->hsync_start - native_mode->hdisplay;
-	unsigned vover = native_mode->vsync_start - native_mode->vdisplay;
-	unsigned hsync_width = native_mode->hsync_end - native_mode->hsync_start;
-	unsigned vsync_width = native_mode->vsync_end - native_mode->vsync_start;
+	unsigned int hblank = native_mode->htotal - native_mode->hdisplay;
+	unsigned int vblank = native_mode->vtotal - native_mode->vdisplay;
+	unsigned int hover = native_mode->hsync_start - native_mode->hdisplay;
+	unsigned int vover = native_mode->vsync_start - native_mode->vdisplay;
+	unsigned int hsync_width = native_mode->hsync_end - native_mode->hsync_start;
+	unsigned int vsync_width = native_mode->vsync_end - native_mode->vsync_start;
 
 	adjusted_mode->clock = native_mode->clock;
 	adjusted_mode->flags = native_mode->flags;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c
new file mode 100644
index 000000000000..23d7d0b0d625
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c
@@ -0,0 +1,241 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include <linux/sched.h>
+#include <drm/drm_exec.h>
+#include "amdgpu.h"
+
+#define work_to_evf_mgr(w, name) container_of(w, struct amdgpu_eviction_fence_mgr, name)
+#define evf_mgr_to_fpriv(e) container_of(e, struct amdgpu_fpriv, evf_mgr)
+
+static const char *
+amdgpu_eviction_fence_get_driver_name(struct dma_fence *fence)
+{
+	return "amdgpu_eviction_fence";
+}
+
+static const char *
+amdgpu_eviction_fence_get_timeline_name(struct dma_fence *f)
+{
+	struct amdgpu_eviction_fence *ef;
+
+	ef = container_of(f, struct amdgpu_eviction_fence, base);
+	return ef->timeline_name;
+}
+
+int
+amdgpu_eviction_fence_replace_fence(struct amdgpu_eviction_fence_mgr *evf_mgr,
+				    struct drm_exec *exec)
+{
+	struct amdgpu_eviction_fence *old_ef, *new_ef;
+	struct drm_gem_object *obj;
+	unsigned long index;
+	int ret;
+
+	if (evf_mgr->ev_fence &&
+	    !dma_fence_is_signaled(&evf_mgr->ev_fence->base))
+		return 0;
+	/*
+	 * Steps to replace eviction fence:
+	 * * lock all objects in exec (caller)
+	 * * create a new eviction fence
+	 * * update new eviction fence in evf_mgr
+	 * * attach the new eviction fence to BOs
+	 * * release the old fence
+	 * * unlock the objects (caller)
+	 */
+	new_ef = amdgpu_eviction_fence_create(evf_mgr);
+	if (!new_ef) {
+		DRM_ERROR("Failed to create new eviction fence\n");
+		return -ENOMEM;
+	}
+
+	/* Update the eviction fence now */
+	spin_lock(&evf_mgr->ev_fence_lock);
+	old_ef = evf_mgr->ev_fence;
+	evf_mgr->ev_fence = new_ef;
+	spin_unlock(&evf_mgr->ev_fence_lock);
+
+	/* Attach the new fence */
+	drm_exec_for_each_locked_object(exec, index, obj) {
+		struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
+
+		if (!bo)
+			continue;
+		ret = amdgpu_eviction_fence_attach(evf_mgr, bo);
+		if (ret) {
+			DRM_ERROR("Failed to attch new eviction fence\n");
+			goto free_err;
+		}
+	}
+
+	/* Free old fence */
+	if (old_ef)
+		dma_fence_put(&old_ef->base);
+	return 0;
+
+free_err:
+	kfree(new_ef);
+	return ret;
+}
+
+static void
+amdgpu_eviction_fence_suspend_worker(struct work_struct *work)
+{
+	struct amdgpu_eviction_fence_mgr *evf_mgr = work_to_evf_mgr(work, suspend_work.work);
+	struct amdgpu_fpriv *fpriv = evf_mgr_to_fpriv(evf_mgr);
+	struct amdgpu_userq_mgr *uq_mgr = &fpriv->userq_mgr;
+	struct amdgpu_eviction_fence *ev_fence;
+
+	mutex_lock(&uq_mgr->userq_mutex);
+	spin_lock(&evf_mgr->ev_fence_lock);
+	ev_fence = evf_mgr->ev_fence;
+	if (ev_fence)
+		dma_fence_get(&ev_fence->base);
+	else
+		goto unlock;
+	spin_unlock(&evf_mgr->ev_fence_lock);
+
+	amdgpu_userq_evict(uq_mgr, ev_fence);
+
+	mutex_unlock(&uq_mgr->userq_mutex);
+	dma_fence_put(&ev_fence->base);
+	return;
+
+unlock:
+	spin_unlock(&evf_mgr->ev_fence_lock);
+	mutex_unlock(&uq_mgr->userq_mutex);
+}
+
+static bool amdgpu_eviction_fence_enable_signaling(struct dma_fence *f)
+{
+	struct amdgpu_eviction_fence_mgr *evf_mgr;
+	struct amdgpu_eviction_fence *ev_fence;
+
+	if (!f)
+		return true;
+
+	ev_fence = to_ev_fence(f);
+	evf_mgr = ev_fence->evf_mgr;
+
+	schedule_delayed_work(&evf_mgr->suspend_work, 0);
+	return true;
+}
+
+static const struct dma_fence_ops amdgpu_eviction_fence_ops = {
+	.get_driver_name = amdgpu_eviction_fence_get_driver_name,
+	.get_timeline_name = amdgpu_eviction_fence_get_timeline_name,
+	.enable_signaling = amdgpu_eviction_fence_enable_signaling,
+};
+
+void amdgpu_eviction_fence_signal(struct amdgpu_eviction_fence_mgr *evf_mgr,
+				  struct amdgpu_eviction_fence *ev_fence)
+{
+	spin_lock(&evf_mgr->ev_fence_lock);
+	dma_fence_signal(&ev_fence->base);
+	spin_unlock(&evf_mgr->ev_fence_lock);
+}
+
+struct amdgpu_eviction_fence *
+amdgpu_eviction_fence_create(struct amdgpu_eviction_fence_mgr *evf_mgr)
+{
+	struct amdgpu_eviction_fence *ev_fence;
+
+	ev_fence = kzalloc(sizeof(*ev_fence), GFP_KERNEL);
+	if (!ev_fence)
+		return NULL;
+
+	ev_fence->evf_mgr = evf_mgr;
+	get_task_comm(ev_fence->timeline_name, current);
+	spin_lock_init(&ev_fence->lock);
+	dma_fence_init64(&ev_fence->base, &amdgpu_eviction_fence_ops,
+			 &ev_fence->lock, evf_mgr->ev_fence_ctx,
+			 atomic_inc_return(&evf_mgr->ev_fence_seq));
+	return ev_fence;
+}
+
+void amdgpu_eviction_fence_destroy(struct amdgpu_eviction_fence_mgr *evf_mgr)
+{
+	struct amdgpu_eviction_fence *ev_fence;
+
+	/* Wait for any pending work to execute */
+	flush_delayed_work(&evf_mgr->suspend_work);
+
+	spin_lock(&evf_mgr->ev_fence_lock);
+	ev_fence = evf_mgr->ev_fence;
+	spin_unlock(&evf_mgr->ev_fence_lock);
+
+	if (!ev_fence)
+		return;
+
+	dma_fence_wait(&ev_fence->base, false);
+
+	/* Last unref of ev_fence */
+	dma_fence_put(&ev_fence->base);
+}
+
+int amdgpu_eviction_fence_attach(struct amdgpu_eviction_fence_mgr *evf_mgr,
+				 struct amdgpu_bo *bo)
+{
+	struct amdgpu_eviction_fence *ev_fence;
+	struct dma_resv *resv = bo->tbo.base.resv;
+	int ret;
+
+	if (!resv)
+		return 0;
+
+	ret = dma_resv_reserve_fences(resv, 1);
+	if (ret) {
+		DRM_DEBUG_DRIVER("Failed to resv fence space\n");
+		return ret;
+	}
+
+	spin_lock(&evf_mgr->ev_fence_lock);
+	ev_fence = evf_mgr->ev_fence;
+	if (ev_fence)
+		dma_resv_add_fence(resv, &ev_fence->base, DMA_RESV_USAGE_BOOKKEEP);
+	spin_unlock(&evf_mgr->ev_fence_lock);
+
+	return 0;
+}
+
+void amdgpu_eviction_fence_detach(struct amdgpu_eviction_fence_mgr *evf_mgr,
+				  struct amdgpu_bo *bo)
+{
+	struct dma_fence *stub = dma_fence_get_stub();
+
+	dma_resv_replace_fences(bo->tbo.base.resv, evf_mgr->ev_fence_ctx,
+				stub, DMA_RESV_USAGE_BOOKKEEP);
+	dma_fence_put(stub);
+}
+
+int amdgpu_eviction_fence_init(struct amdgpu_eviction_fence_mgr *evf_mgr)
+{
+	/* This needs to be done one time per open */
+	atomic_set(&evf_mgr->ev_fence_seq, 0);
+	evf_mgr->ev_fence_ctx = dma_fence_context_alloc(1);
+	spin_lock_init(&evf_mgr->ev_fence_lock);
+
+	INIT_DELAYED_WORK(&evf_mgr->suspend_work, amdgpu_eviction_fence_suspend_worker);
+	return 0;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.h
new file mode 100644
index 000000000000..fcd867b7147d
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.h
@@ -0,0 +1,69 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef AMDGPU_EV_FENCE_H_
+#define AMDGPU_EV_FENCE_H_
+
+struct amdgpu_eviction_fence {
+	struct dma_fence base;
+	spinlock_t	 lock;
+	char		 timeline_name[TASK_COMM_LEN];
+	struct amdgpu_eviction_fence_mgr *evf_mgr;
+};
+
+struct amdgpu_eviction_fence_mgr {
+	u64			ev_fence_ctx;
+	atomic_t		ev_fence_seq;
+	spinlock_t		ev_fence_lock;
+	struct amdgpu_eviction_fence *ev_fence;
+	struct delayed_work	suspend_work;
+	uint8_t fd_closing;
+};
+
+/* Eviction fence helper functions */
+struct amdgpu_eviction_fence *
+amdgpu_eviction_fence_create(struct amdgpu_eviction_fence_mgr *evf_mgr);
+
+void
+amdgpu_eviction_fence_destroy(struct amdgpu_eviction_fence_mgr *evf_mgr);
+
+int
+amdgpu_eviction_fence_attach(struct amdgpu_eviction_fence_mgr *evf_mgr,
+			     struct amdgpu_bo *bo);
+
+void
+amdgpu_eviction_fence_detach(struct amdgpu_eviction_fence_mgr *evf_mgr,
+			     struct amdgpu_bo *bo);
+
+int
+amdgpu_eviction_fence_init(struct amdgpu_eviction_fence_mgr *evf_mgr);
+
+void
+amdgpu_eviction_fence_signal(struct amdgpu_eviction_fence_mgr *evf_mgr,
+			     struct amdgpu_eviction_fence *ev_fence);
+
+int
+amdgpu_eviction_fence_replace_fence(struct amdgpu_eviction_fence_mgr *evf_mgr,
+				    struct drm_exec *exec);
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c
index c57252f004e8..b349bb3676d5 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c
@@ -33,6 +33,7 @@
 #include <drm/amdgpu_drm.h>
 #include <drm/drm_debugfs.h>
 #include <drm/drm_drv.h>
+#include <drm/drm_file.h>
 
 #include "amdgpu.h"
 #include "amdgpu_vm.h"
@@ -51,34 +52,29 @@ static const char *amdgpu_ip_name[AMDGPU_HW_IP_NUM] = {
 	[AMDGPU_HW_IP_VCN_DEC]	=	"dec",
 	[AMDGPU_HW_IP_VCN_ENC]	=	"enc",
 	[AMDGPU_HW_IP_VCN_JPEG]	=	"jpeg",
+	[AMDGPU_HW_IP_VPE]	=	"vpe",
 };
 
-void amdgpu_show_fdinfo(struct seq_file *m, struct file *f)
+void amdgpu_show_fdinfo(struct drm_printer *p, struct drm_file *file)
 {
-	struct drm_file *file = f->private_data;
-	struct amdgpu_device *adev = drm_to_adev(file->minor->dev);
 	struct amdgpu_fpriv *fpriv = file->driver_priv;
 	struct amdgpu_vm *vm = &fpriv->vm;
 
-	struct amdgpu_mem_stats stats;
+	struct amdgpu_mem_stats stats[__AMDGPU_PL_NUM];
 	ktime_t usage[AMDGPU_HW_IP_NUM];
-	uint32_t bus, dev, fn, domain;
-	unsigned int hw_ip;
-	int ret;
-
-	memset(&stats, 0, sizeof(stats));
-	bus = adev->pdev->bus->number;
-	domain = pci_domain_nr(adev->pdev->bus);
-	dev = PCI_SLOT(adev->pdev->devfn);
-	fn = PCI_FUNC(adev->pdev->devfn);
-
-	ret = amdgpu_bo_reserve(vm->root.bo, false);
-	if (ret)
-		return;
-
-	amdgpu_vm_get_memory(vm, &stats);
-	amdgpu_bo_unreserve(vm->root.bo);
+	const char *pl_name[] = {
+		[TTM_PL_VRAM] = "vram",
+		[TTM_PL_TT] = "gtt",
+		[TTM_PL_SYSTEM] = "cpu",
+		[AMDGPU_PL_GDS] = "gds",
+		[AMDGPU_PL_GWS] = "gws",
+		[AMDGPU_PL_OA] = "oa",
+		[AMDGPU_PL_DOORBELL] = "doorbell",
+		[AMDGPU_PL_MMIO_REMAP] = "mmioremap",
+	};
+	unsigned int hw_ip, i;
 
+	amdgpu_vm_get_memory(vm, stats);
 	amdgpu_ctx_mgr_usage(&fpriv->ctx_mgr, usage);
 
 	/*
@@ -87,31 +83,42 @@ void amdgpu_show_fdinfo(struct seq_file *m, struct file *f)
 	 * ******************************************************************
 	 */
 
-	seq_printf(m, "pasid:\t%u\n", fpriv->vm.pasid);
-	seq_printf(m, "drm-driver:\t%s\n", file->minor->dev->driver->name);
-	seq_printf(m, "drm-pdev:\t%04x:%02x:%02x.%d\n", domain, bus, dev, fn);
-	seq_printf(m, "drm-client-id:\t%Lu\n", vm->immediate.fence_context);
-	seq_printf(m, "drm-memory-vram:\t%llu KiB\n", stats.vram/1024UL);
-	seq_printf(m, "drm-memory-gtt: \t%llu KiB\n", stats.gtt/1024UL);
-	seq_printf(m, "drm-memory-cpu: \t%llu KiB\n", stats.cpu/1024UL);
-	seq_printf(m, "amd-memory-visible-vram:\t%llu KiB\n",
-		   stats.visible_vram/1024UL);
-	seq_printf(m, "amd-evicted-vram:\t%llu KiB\n",
-		   stats.evicted_vram/1024UL);
-	seq_printf(m, "amd-evicted-visible-vram:\t%llu KiB\n",
-		   stats.evicted_visible_vram/1024UL);
-	seq_printf(m, "amd-requested-vram:\t%llu KiB\n",
-		   stats.requested_vram/1024UL);
-	seq_printf(m, "amd-requested-visible-vram:\t%llu KiB\n",
-		   stats.requested_visible_vram/1024UL);
-	seq_printf(m, "amd-requested-gtt:\t%llu KiB\n",
-		   stats.requested_gtt/1024UL);
+	drm_printf(p, "pasid:\t%u\n", fpriv->vm.pasid);
+
+	for (i = 0; i < ARRAY_SIZE(pl_name); i++) {
+		if (!pl_name[i])
+			continue;
+
+		drm_print_memory_stats(p,
+				       &stats[i].drm,
+				       DRM_GEM_OBJECT_RESIDENT |
+				       DRM_GEM_OBJECT_PURGEABLE,
+				       pl_name[i]);
+	}
+
+	/* Legacy amdgpu keys, alias to drm-resident-memory-: */
+	drm_printf(p, "drm-memory-vram:\t%llu KiB\n",
+		   stats[TTM_PL_VRAM].drm.resident/1024UL);
+	drm_printf(p, "drm-memory-gtt: \t%llu KiB\n",
+		   stats[TTM_PL_TT].drm.resident/1024UL);
+	drm_printf(p, "drm-memory-cpu: \t%llu KiB\n",
+		   stats[TTM_PL_SYSTEM].drm.resident/1024UL);
+
+	/* Amdgpu specific memory accounting keys: */
+	drm_printf(p, "amd-evicted-vram:\t%llu KiB\n",
+		   stats[TTM_PL_VRAM].evicted/1024UL);
+	drm_printf(p, "amd-requested-vram:\t%llu KiB\n",
+		   (stats[TTM_PL_VRAM].drm.shared +
+		    stats[TTM_PL_VRAM].drm.private) / 1024UL);
+	drm_printf(p, "amd-requested-gtt:\t%llu KiB\n",
+		   (stats[TTM_PL_TT].drm.shared +
+		    stats[TTM_PL_TT].drm.private) / 1024UL);
 
 	for (hw_ip = 0; hw_ip < AMDGPU_HW_IP_NUM; ++hw_ip) {
 		if (!usage[hw_ip])
 			continue;
 
-		seq_printf(m, "drm-engine-%s:\t%Ld ns\n", amdgpu_ip_name[hw_ip],
+		drm_printf(p, "drm-engine-%s:\t%lld ns\n", amdgpu_ip_name[hw_ip],
 			   ktime_to_ns(usage[hw_ip]));
 	}
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.h
index e86834bfea1d..0398f5a159ef 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.h
@@ -37,6 +37,6 @@
 #include "amdgpu_ids.h"
 
 uint32_t amdgpu_get_ip_count(struct amdgpu_device *adev, int id);
-void amdgpu_show_fdinfo(struct seq_file *m, struct file *f);
+void amdgpu_show_fdinfo(struct drm_printer *p, struct drm_file *file);
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
index f52d0ba91a77..c7843e336310 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
@@ -42,53 +42,14 @@
 #include "amdgpu_reset.h"
 
 /*
- * Fences
- * Fences mark an event in the GPUs pipeline and are used
- * for GPU/CPU synchronization.  When the fence is written,
- * it is expected that all buffers associated with that fence
- * are no longer in use by the associated ring on the GPU and
- * that the relevant GPU caches have been flushed.
- */
-
-struct amdgpu_fence {
-	struct dma_fence base;
-
-	/* RB, DMA, etc. */
-	struct amdgpu_ring		*ring;
-	ktime_t				start_timestamp;
-};
-
-static struct kmem_cache *amdgpu_fence_slab;
-
-int amdgpu_fence_slab_init(void)
-{
-	amdgpu_fence_slab = kmem_cache_create(
-		"amdgpu_fence", sizeof(struct amdgpu_fence), 0,
-		SLAB_HWCACHE_ALIGN, NULL);
-	if (!amdgpu_fence_slab)
-		return -ENOMEM;
-	return 0;
-}
-
-void amdgpu_fence_slab_fini(void)
-{
-	rcu_barrier();
-	kmem_cache_destroy(amdgpu_fence_slab);
-}
-/*
  * Cast helper
  */
 static const struct dma_fence_ops amdgpu_fence_ops;
-static const struct dma_fence_ops amdgpu_job_fence_ops;
 static inline struct amdgpu_fence *to_amdgpu_fence(struct dma_fence *f)
 {
 	struct amdgpu_fence *__f = container_of(f, struct amdgpu_fence, base);
 
-	if (__f->base.ops == &amdgpu_fence_ops ||
-	    __f->base.ops == &amdgpu_job_fence_ops)
-		return __f;
-
-	return NULL;
+	return __f;
 }
 
 /**
@@ -132,57 +93,32 @@ static u32 amdgpu_fence_read(struct amdgpu_ring *ring)
  * amdgpu_fence_emit - emit a fence on the requested ring
  *
  * @ring: ring the fence is associated with
- * @f: resulting fence object
- * @job: job the fence is embedded in
+ * @af: amdgpu fence input
  * @flags: flags to pass into the subordinate .emit_fence() call
  *
  * Emits a fence command on the requested ring (all asics).
  * Returns 0 on success, -ENOMEM on failure.
  */
-int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **f, struct amdgpu_job *job,
-		      unsigned flags)
+int amdgpu_fence_emit(struct amdgpu_ring *ring, struct amdgpu_fence *af,
+		      unsigned int flags)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct dma_fence *fence;
-	struct amdgpu_fence *am_fence;
 	struct dma_fence __rcu **ptr;
 	uint32_t seq;
 	int r;
 
-	if (job == NULL) {
-		/* create a sperate hw fence */
-		am_fence = kmem_cache_alloc(amdgpu_fence_slab, GFP_ATOMIC);
-		if (am_fence == NULL)
-			return -ENOMEM;
-		fence = &am_fence->base;
-		am_fence->ring = ring;
-	} else {
-		/* take use of job-embedded fence */
-		fence = &job->hw_fence;
-	}
+	fence = &af->base;
+	af->ring = ring;
 
 	seq = ++ring->fence_drv.sync_seq;
-	if (job && job->job_run_counter) {
-		/* reinit seq for resubmitted jobs */
-		fence->seqno = seq;
-		/* TO be inline with external fence creation and other drivers */
-		dma_fence_get(fence);
-	} else {
-		if (job) {
-			dma_fence_init(fence, &amdgpu_job_fence_ops,
-				       &ring->fence_drv.lock,
-				       adev->fence_context + ring->idx, seq);
-			/* Against remove in amdgpu_job_{free, free_cb} */
-			dma_fence_get(fence);
-		}
-		else
-			dma_fence_init(fence, &amdgpu_fence_ops,
-				       &ring->fence_drv.lock,
-				       adev->fence_context + ring->idx, seq);
-	}
+	dma_fence_init(fence, &amdgpu_fence_ops,
+		       &ring->fence_drv.lock,
+		       adev->fence_context + ring->idx, seq);
 
 	amdgpu_ring_emit_fence(ring, ring->fence_drv.gpu_addr,
 			       seq, flags | AMDGPU_FENCE_FLAG_INT);
+	amdgpu_fence_save_wptr(af);
 	pm_runtime_get_noresume(adev_to_drm(adev)->dev);
 	ptr = &ring->fence_drv.fences[seq & ring->fence_drv.num_fences_mask];
 	if (unlikely(rcu_dereference_protected(*ptr, 1))) {
@@ -207,8 +143,6 @@ int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **f, struct amd
 	 */
 	rcu_assign_pointer(*ptr, dma_fence_get(fence));
 
-	*f = fence;
-
 	return 0;
 }
 
@@ -283,7 +217,7 @@ bool amdgpu_fence_process(struct amdgpu_ring *ring)
 
 	} while (atomic_cmpxchg(&drv->last_seq, last_seq, seq) != last_seq);
 
-	if (del_timer(&ring->fence_drv.fallback_timer) &&
+	if (timer_delete(&ring->fence_drv.fallback_timer) &&
 	    seq != ring->fence_drv.sync_seq)
 		amdgpu_fence_schedule_fallback(ring);
 
@@ -295,6 +229,7 @@ bool amdgpu_fence_process(struct amdgpu_ring *ring)
 
 	do {
 		struct dma_fence *fence, **ptr;
+		struct amdgpu_fence *am_fence;
 
 		++last_seq;
 		last_seq &= drv->num_fences_mask;
@@ -307,9 +242,14 @@ bool amdgpu_fence_process(struct amdgpu_ring *ring)
 		if (!fence)
 			continue;
 
+		/* Save the wptr in the fence driver so we know what the last processed
+		 * wptr was.  This is required for re-emitting the ring state for
+		 * queues that are reset but are not guilty and thus have no guilty fence.
+		 */
+		am_fence = container_of(fence, struct amdgpu_fence, base);
+		drv->signalled_wptr = am_fence->wptr;
 		dma_fence_signal(fence);
 		dma_fence_put(fence);
-		pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 		pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 	} while (last_seq != seq);
 
@@ -325,11 +265,13 @@ bool amdgpu_fence_process(struct amdgpu_ring *ring)
  */
 static void amdgpu_fence_fallback(struct timer_list *t)
 {
-	struct amdgpu_ring *ring = from_timer(ring, t,
-					      fence_drv.fallback_timer);
+	struct amdgpu_ring *ring = timer_container_of(ring, t,
+						      fence_drv.fallback_timer);
 
 	if (amdgpu_fence_process(ring))
-		DRM_WARN("Fence fallback timer expired on ring %s\n", ring->name);
+		dev_warn(ring->adev->dev,
+			 "Fence fallback timer expired on ring %s\n",
+			 ring->name);
 }
 
 /**
@@ -377,14 +319,11 @@ signed long amdgpu_fence_wait_polling(struct amdgpu_ring *ring,
 				      uint32_t wait_seq,
 				      signed long timeout)
 {
-	uint32_t seq;
-
-	do {
-		seq = amdgpu_fence_read(ring);
-		udelay(5);
-		timeout -= 5;
-	} while ((int32_t)(wait_seq - seq) > 0 && timeout > 0);
 
+	while ((int32_t)(wait_seq - amdgpu_fence_read(ring)) > 0 && timeout > 0) {
+		udelay(2);
+		timeout -= 2;
+	}
 	return timeout > 0 ? timeout : 0;
 }
 /**
@@ -396,7 +335,7 @@ signed long amdgpu_fence_wait_polling(struct amdgpu_ring *ring,
  * Returns the number of emitted fences on the ring.  Used by the
  * dynpm code to ring track activity.
  */
-unsigned amdgpu_fence_count_emitted(struct amdgpu_ring *ring)
+unsigned int amdgpu_fence_count_emitted(struct amdgpu_ring *ring)
 {
 	uint64_t emitted;
 
@@ -475,7 +414,7 @@ void amdgpu_fence_update_start_timestamp(struct amdgpu_ring *ring, uint32_t seq,
  */
 int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring,
 				   struct amdgpu_irq_src *irq_src,
-				   unsigned irq_type)
+				   unsigned int irq_type)
 {
 	struct amdgpu_device *adev = ring->adev;
 	uint64_t index;
@@ -556,6 +495,42 @@ int amdgpu_fence_driver_sw_init(struct amdgpu_device *adev)
 }
 
 /**
+ * amdgpu_fence_need_ring_interrupt_restore - helper function to check whether
+ * fence driver interrupts need to be restored.
+ *
+ * @ring: ring that to be checked
+ *
+ * Interrupts for rings that belong to GFX IP don't need to be restored
+ * when the target power state is s0ix.
+ *
+ * Return true if need to restore interrupts, false otherwise.
+ */
+static bool amdgpu_fence_need_ring_interrupt_restore(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	bool is_gfx_power_domain = false;
+
+	switch (ring->funcs->type) {
+	case AMDGPU_RING_TYPE_SDMA:
+	/* SDMA 5.x+ is part of GFX power domain so it's covered by GFXOFF */
+		if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) >=
+		    IP_VERSION(5, 0, 0))
+			is_gfx_power_domain = true;
+		break;
+	case AMDGPU_RING_TYPE_GFX:
+	case AMDGPU_RING_TYPE_COMPUTE:
+	case AMDGPU_RING_TYPE_KIQ:
+	case AMDGPU_RING_TYPE_MES:
+		is_gfx_power_domain = true;
+		break;
+	default:
+		break;
+	}
+
+	return !(adev->in_s0ix && is_gfx_power_domain);
+}
+
+/**
  * amdgpu_fence_driver_hw_fini - tear down the fence driver
  * for all possible rings.
  *
@@ -582,11 +557,13 @@ void amdgpu_fence_driver_hw_fini(struct amdgpu_device *adev)
 		if (r)
 			amdgpu_fence_driver_force_completion(ring);
 
-		if (ring->fence_drv.irq_src)
+		if (!drm_dev_is_unplugged(adev_to_drm(adev)) &&
+		    ring->fence_drv.irq_src &&
+		    amdgpu_fence_need_ring_interrupt_restore(ring))
 			amdgpu_irq_put(adev, ring->fence_drv.irq_src,
 				       ring->fence_drv.irq_type);
 
-		del_timer_sync(&ring->fence_drv.fallback_timer);
+		timer_delete_sync(&ring->fence_drv.fallback_timer);
 	}
 }
 
@@ -653,44 +630,40 @@ void amdgpu_fence_driver_hw_init(struct amdgpu_device *adev)
 
 	for (i = 0; i < AMDGPU_MAX_RINGS; i++) {
 		struct amdgpu_ring *ring = adev->rings[i];
+
 		if (!ring || !ring->fence_drv.initialized)
 			continue;
 
 		/* enable the interrupt */
-		if (ring->fence_drv.irq_src)
+		if (ring->fence_drv.irq_src &&
+		    amdgpu_fence_need_ring_interrupt_restore(ring))
 			amdgpu_irq_get(adev, ring->fence_drv.irq_src,
 				       ring->fence_drv.irq_type);
 	}
 }
 
 /**
- * amdgpu_fence_driver_clear_job_fences - clear job embedded fences of ring
- *
- * @ring: fence of the ring to be cleared
+ * amdgpu_fence_driver_set_error - set error code on fences
+ * @ring: the ring which contains the fences
+ * @error: the error code to set
  *
+ * Set an error code to all the fences pending on the ring.
  */
-void amdgpu_fence_driver_clear_job_fences(struct amdgpu_ring *ring)
+void amdgpu_fence_driver_set_error(struct amdgpu_ring *ring, int error)
 {
-	int i;
-	struct dma_fence *old, **ptr;
+	struct amdgpu_fence_driver *drv = &ring->fence_drv;
+	unsigned long flags;
 
-	for (i = 0; i <= ring->fence_drv.num_fences_mask; i++) {
-		ptr = &ring->fence_drv.fences[i];
-		old = rcu_dereference_protected(*ptr, 1);
-		if (old && old->ops == &amdgpu_job_fence_ops) {
-			struct amdgpu_job *job;
+	spin_lock_irqsave(&drv->lock, flags);
+	for (unsigned int i = 0; i <= drv->num_fences_mask; ++i) {
+		struct dma_fence *fence;
 
-			/* For non-scheduler bad job, i.e. failed ib test, we need to signal
-			 * it right here or we won't be able to track them in fence_drv
-			 * and they will remain unsignaled during sa_bo free.
-			 */
-			job = container_of(old, struct amdgpu_job, hw_fence);
-			if (!job->base.s_fence && !dma_fence_is_signaled(old))
-				dma_fence_signal(old);
-			RCU_INIT_POINTER(*ptr, NULL);
-			dma_fence_put(old);
-		}
+		fence = rcu_dereference_protected(drv->fences[i],
+						  lockdep_is_held(&drv->lock));
+		if (fence && !dma_fence_is_signaled_locked(fence))
+			dma_fence_set_error(fence, error);
 	}
+	spin_unlock_irqrestore(&drv->lock, flags);
 }
 
 /**
@@ -701,10 +674,125 @@ void amdgpu_fence_driver_clear_job_fences(struct amdgpu_ring *ring)
  */
 void amdgpu_fence_driver_force_completion(struct amdgpu_ring *ring)
 {
+	amdgpu_fence_driver_set_error(ring, -ECANCELED);
 	amdgpu_fence_write(ring, ring->fence_drv.sync_seq);
 	amdgpu_fence_process(ring);
 }
 
+
+/*
+ * Kernel queue reset handling
+ *
+ * The driver can reset individual queues for most engines, but those queues
+ * may contain work from multiple contexts.  Resetting the queue will reset
+ * lose all of that state.  In order to minimize the collateral damage, the
+ * driver will save the ring contents which are not associated with the guilty
+ * context prior to resetting the queue.  After resetting the queue the queue
+ * contents from the other contexts is re-emitted to the rings so that it can
+ * be processed by the engine.  To handle this, we save the queue's write
+ * pointer (wptr) in the fences associated with each context.  If we get a
+ * queue timeout, we can then use the wptrs from the fences to determine
+ * which data needs to be saved out of the queue's ring buffer.
+ */
+
+/**
+ * amdgpu_fence_driver_guilty_force_completion - force signal of specified sequence
+ *
+ * @af: fence of the ring to signal
+ *
+ */
+void amdgpu_fence_driver_guilty_force_completion(struct amdgpu_fence *af)
+{
+	struct dma_fence *unprocessed;
+	struct dma_fence __rcu **ptr;
+	struct amdgpu_fence *fence;
+	struct amdgpu_ring *ring = af->ring;
+	unsigned long flags;
+	u32 seq, last_seq;
+
+	last_seq = amdgpu_fence_read(ring) & ring->fence_drv.num_fences_mask;
+	seq = ring->fence_drv.sync_seq & ring->fence_drv.num_fences_mask;
+
+	/* mark all fences from the guilty context with an error */
+	spin_lock_irqsave(&ring->fence_drv.lock, flags);
+	do {
+		last_seq++;
+		last_seq &= ring->fence_drv.num_fences_mask;
+
+		ptr = &ring->fence_drv.fences[last_seq];
+		rcu_read_lock();
+		unprocessed = rcu_dereference(*ptr);
+
+		if (unprocessed && !dma_fence_is_signaled_locked(unprocessed)) {
+			fence = container_of(unprocessed, struct amdgpu_fence, base);
+
+			if (fence == af)
+				dma_fence_set_error(&fence->base, -ETIME);
+			else if (fence->context == af->context)
+				dma_fence_set_error(&fence->base, -ECANCELED);
+		}
+		rcu_read_unlock();
+	} while (last_seq != seq);
+	spin_unlock_irqrestore(&ring->fence_drv.lock, flags);
+	/* signal the guilty fence */
+	amdgpu_fence_write(ring, (u32)af->base.seqno);
+	amdgpu_fence_process(ring);
+}
+
+void amdgpu_fence_save_wptr(struct amdgpu_fence *af)
+{
+	af->wptr = af->ring->wptr;
+}
+
+static void amdgpu_ring_backup_unprocessed_command(struct amdgpu_ring *ring,
+						   u64 start_wptr, u32 end_wptr)
+{
+	unsigned int first_idx = start_wptr & ring->buf_mask;
+	unsigned int last_idx = end_wptr & ring->buf_mask;
+	unsigned int i;
+
+	/* Backup the contents of the ring buffer. */
+	for (i = first_idx; i != last_idx; ++i, i &= ring->buf_mask)
+		ring->ring_backup[ring->ring_backup_entries_to_copy++] = ring->ring[i];
+}
+
+void amdgpu_ring_backup_unprocessed_commands(struct amdgpu_ring *ring,
+					     struct amdgpu_fence *guilty_fence)
+{
+	struct dma_fence *unprocessed;
+	struct dma_fence __rcu **ptr;
+	struct amdgpu_fence *fence;
+	u64 wptr;
+	u32 seq, last_seq;
+
+	last_seq = amdgpu_fence_read(ring) & ring->fence_drv.num_fences_mask;
+	seq = ring->fence_drv.sync_seq & ring->fence_drv.num_fences_mask;
+	wptr = ring->fence_drv.signalled_wptr;
+	ring->ring_backup_entries_to_copy = 0;
+
+	do {
+		last_seq++;
+		last_seq &= ring->fence_drv.num_fences_mask;
+
+		ptr = &ring->fence_drv.fences[last_seq];
+		rcu_read_lock();
+		unprocessed = rcu_dereference(*ptr);
+
+		if (unprocessed && !dma_fence_is_signaled(unprocessed)) {
+			fence = container_of(unprocessed, struct amdgpu_fence, base);
+
+			/* save everything if the ring is not guilty, otherwise
+			 * just save the content from other contexts.
+			 */
+			if (!guilty_fence || (fence->context != guilty_fence->context))
+				amdgpu_ring_backup_unprocessed_command(ring, wptr,
+								       fence->wptr);
+			wptr = fence->wptr;
+		}
+		rcu_read_unlock();
+	} while (last_seq != seq);
+}
+
 /*
  * Common fence implementation
  */
@@ -719,13 +807,6 @@ static const char *amdgpu_fence_get_timeline_name(struct dma_fence *f)
 	return (const char *)to_amdgpu_fence(f)->ring->name;
 }
 
-static const char *amdgpu_job_fence_get_timeline_name(struct dma_fence *f)
-{
-	struct amdgpu_job *job = container_of(f, struct amdgpu_job, hw_fence);
-
-	return (const char *)to_amdgpu_ring(job->base.sched)->name;
-}
-
 /**
  * amdgpu_fence_enable_signaling - enable signalling on fence
  * @f: fence
@@ -743,23 +824,6 @@ static bool amdgpu_fence_enable_signaling(struct dma_fence *f)
 }
 
 /**
- * amdgpu_job_fence_enable_signaling - enable signalling on job fence
- * @f: fence
- *
- * This is the simliar function with amdgpu_fence_enable_signaling above, it
- * only handles the job embedded fence.
- */
-static bool amdgpu_job_fence_enable_signaling(struct dma_fence *f)
-{
-	struct amdgpu_job *job = container_of(f, struct amdgpu_job, hw_fence);
-
-	if (!timer_pending(&to_amdgpu_ring(job->base.sched)->fence_drv.fallback_timer))
-		amdgpu_fence_schedule_fallback(to_amdgpu_ring(job->base.sched));
-
-	return true;
-}
-
-/**
  * amdgpu_fence_free - free up the fence memory
  *
  * @rcu: RCU callback head
@@ -771,22 +835,7 @@ static void amdgpu_fence_free(struct rcu_head *rcu)
 	struct dma_fence *f = container_of(rcu, struct dma_fence, rcu);
 
 	/* free fence_slab if it's separated fence*/
-	kmem_cache_free(amdgpu_fence_slab, to_amdgpu_fence(f));
-}
-
-/**
- * amdgpu_job_fence_free - free up the job with embedded fence
- *
- * @rcu: RCU callback head
- *
- * Free up the job with embedded fence after the RCU grace period.
- */
-static void amdgpu_job_fence_free(struct rcu_head *rcu)
-{
-	struct dma_fence *f = container_of(rcu, struct dma_fence, rcu);
-
-	/* free job if fence has a parent job */
-	kfree(container_of(f, struct amdgpu_job, hw_fence));
+	kfree(to_amdgpu_fence(f));
 }
 
 /**
@@ -802,19 +851,6 @@ static void amdgpu_fence_release(struct dma_fence *f)
 	call_rcu(&f->rcu, amdgpu_fence_free);
 }
 
-/**
- * amdgpu_job_fence_release - callback that job embedded fence can be freed
- *
- * @f: fence
- *
- * This is the simliar function with amdgpu_fence_release above, it
- * only handles the job embedded fence.
- */
-static void amdgpu_job_fence_release(struct dma_fence *f)
-{
-	call_rcu(&f->rcu, amdgpu_job_fence_free);
-}
-
 static const struct dma_fence_ops amdgpu_fence_ops = {
 	.get_driver_name = amdgpu_fence_get_driver_name,
 	.get_timeline_name = amdgpu_fence_get_timeline_name,
@@ -822,24 +858,18 @@ static const struct dma_fence_ops amdgpu_fence_ops = {
 	.release = amdgpu_fence_release,
 };
 
-static const struct dma_fence_ops amdgpu_job_fence_ops = {
-	.get_driver_name = amdgpu_fence_get_driver_name,
-	.get_timeline_name = amdgpu_job_fence_get_timeline_name,
-	.enable_signaling = amdgpu_job_fence_enable_signaling,
-	.release = amdgpu_job_fence_release,
-};
-
 /*
  * Fence debugfs
  */
 #if defined(CONFIG_DEBUG_FS)
 static int amdgpu_debugfs_fence_info_show(struct seq_file *m, void *unused)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct amdgpu_device *adev = m->private;
 	int i;
 
 	for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 		struct amdgpu_ring *ring = adev->rings[i];
+
 		if (!ring || !ring->fence_drv.initialized)
 			continue;
 
@@ -897,7 +927,6 @@ static int gpu_recover_get(void *data, u64 *val)
 
 	*val = atomic_read(&adev->reset_domain->reset_res);
 
-	pm_runtime_mark_last_busy(dev->dev);
 	pm_runtime_put_autosuspend(dev->dev);
 
 	return 0;
@@ -913,11 +942,14 @@ static void amdgpu_debugfs_reset_work(struct work_struct *work)
 						  reset_work);
 
 	struct amdgpu_reset_context reset_context;
+
 	memset(&reset_context, 0, sizeof(reset_context));
 
 	reset_context.method = AMD_RESET_METHOD_NONE;
 	reset_context.reset_req_dev = adev;
+	reset_context.src = AMDGPU_RESET_SRC_USER;
 	set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
+	set_bit(AMDGPU_SKIP_COREDUMP, &reset_context.flags);
 
 	amdgpu_device_gpu_recover(adev, NULL, &reset_context);
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c
index 4620c4712ce3..b0082aa7f3c6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c
@@ -31,6 +31,7 @@
 
 #define FRU_EEPROM_MADDR_6      0x60000
 #define FRU_EEPROM_MADDR_8      0x80000
+#define FRU_EEPROM_MADDR_INV    0xFFFFF
 
 static bool is_fru_eeprom_supported(struct amdgpu_device *adev, u32 *fru_addr)
 {
@@ -42,8 +43,9 @@ static bool is_fru_eeprom_supported(struct amdgpu_device *adev, u32 *fru_addr)
 
 	/* The i2c access is blocked on VF
 	 * TODO: Need other way to get the info
+	 * Also, FRU not valid for APU devices.
 	 */
-	if (amdgpu_sriov_vf(adev))
+	if (amdgpu_sriov_vf(adev) || (adev->flags & AMD_IS_APU))
 		return false;
 
 	/* The default I2C EEPROM address of the FRU.
@@ -57,40 +59,56 @@ static bool is_fru_eeprom_supported(struct amdgpu_device *adev, u32 *fru_addr)
 	 * for ease/speed/readability. For now, 2 string comparisons are
 	 * reasonable and not too expensive
 	 */
-	switch (adev->asic_type) {
-	case CHIP_VEGA20:
-		/* D161 and D163 are the VG20 server SKUs */
-		if (strnstr(atom_ctx->vbios_version, "D161",
-			    sizeof(atom_ctx->vbios_version)) ||
-		    strnstr(atom_ctx->vbios_version, "D163",
-			    sizeof(atom_ctx->vbios_version))) {
+	switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
+	case IP_VERSION(11, 0, 2):
+		switch (adev->asic_type) {
+		case CHIP_VEGA20:
+			/* D161 and D163 are the VG20 server SKUs */
+			if (atom_ctx && (strnstr(atom_ctx->vbios_pn, "D161",
+						 sizeof(atom_ctx->vbios_pn)) ||
+					 strnstr(atom_ctx->vbios_pn, "D163",
+						 sizeof(atom_ctx->vbios_pn)))) {
+				if (fru_addr)
+					*fru_addr = FRU_EEPROM_MADDR_6;
+				return true;
+			} else {
+				return false;
+			}
+		case CHIP_ARCTURUS:
+		default:
+			return false;
+		}
+	case IP_VERSION(11, 0, 7):
+		if (atom_ctx && strnstr(atom_ctx->vbios_pn, "D603",
+					sizeof(atom_ctx->vbios_pn))) {
+			if (strnstr(atom_ctx->vbios_pn, "D603GLXE",
+				    sizeof(atom_ctx->vbios_pn))) {
+				return false;
+			}
+
 			if (fru_addr)
 				*fru_addr = FRU_EEPROM_MADDR_6;
 			return true;
+
 		} else {
 			return false;
 		}
-	case CHIP_ALDEBARAN:
+	case IP_VERSION(13, 0, 2):
 		/* All Aldebaran SKUs have an FRU */
-		if (!strnstr(atom_ctx->vbios_version, "D673",
-			     sizeof(atom_ctx->vbios_version)))
+		if (atom_ctx && !strnstr(atom_ctx->vbios_pn, "D673",
+					 sizeof(atom_ctx->vbios_pn)))
 			if (fru_addr)
 				*fru_addr = FRU_EEPROM_MADDR_6;
 		return true;
-	case CHIP_SIENNA_CICHLID:
-		if (strnstr(atom_ctx->vbios_version, "D603",
-			    sizeof(atom_ctx->vbios_version))) {
-			if (strnstr(atom_ctx->vbios_version, "D603GLXE",
-				    sizeof(atom_ctx->vbios_version))) {
-				return false;
-			} else {
-				if (fru_addr)
-					*fru_addr = FRU_EEPROM_MADDR_6;
-				return true;
-			}
-		} else {
-			return false;
-		}
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 14):
+			if (fru_addr)
+				*fru_addr = FRU_EEPROM_MADDR_8;
+			return true;
+	case IP_VERSION(13, 0, 12):
+			if (fru_addr)
+				*fru_addr = FRU_EEPROM_MADDR_INV;
+			return true;
 	default:
 		return false;
 	}
@@ -98,6 +116,7 @@ static bool is_fru_eeprom_supported(struct amdgpu_device *adev, u32 *fru_addr)
 
 int amdgpu_fru_get_product_info(struct amdgpu_device *adev)
 {
+	struct amdgpu_fru_info *fru_info;
 	unsigned char buf[8], *pia;
 	u32 addr, fru_addr;
 	int size, len;
@@ -106,9 +125,27 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev)
 	if (!is_fru_eeprom_supported(adev, &fru_addr))
 		return 0;
 
+	/* FRU data avaialble, but no direct EEPROM access */
+	if (fru_addr == FRU_EEPROM_MADDR_INV)
+		return 0;
+
+	if (!adev->fru_info) {
+		adev->fru_info = kzalloc(sizeof(*adev->fru_info), GFP_KERNEL);
+		if (!adev->fru_info)
+			return -ENOMEM;
+	}
+
+	fru_info = adev->fru_info;
+	/* For Arcturus-and-later, default value of serial_number is unique_id
+	 * so convert it to a 16-digit HEX string for convenience and
+	 * backwards-compatibility.
+	 */
+	sprintf(fru_info->serial, "%llx", adev->unique_id);
+
 	/* If algo exists, it means that the i2c_adapter's initialized */
 	if (!adev->pm.fru_eeprom_i2c_bus || !adev->pm.fru_eeprom_i2c_bus->algo) {
-		DRM_WARN("Cannot access FRU, EEPROM accessor not initialized");
+		dev_warn(adev->dev,
+			 "Cannot access FRU, EEPROM accessor not initialized");
 		return -ENODEV;
 	}
 
@@ -116,19 +153,22 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev)
 	len = amdgpu_eeprom_read(adev->pm.fru_eeprom_i2c_bus, fru_addr, buf,
 				 sizeof(buf));
 	if (len != 8) {
-		DRM_ERROR("Couldn't read the IPMI Common Header: %d", len);
+		dev_err(adev->dev, "Couldn't read the IPMI Common Header: %d",
+			len);
 		return len < 0 ? len : -EIO;
 	}
 
 	if (buf[0] != 1) {
-		DRM_ERROR("Bad IPMI Common Header version: 0x%02x", buf[0]);
+		dev_err(adev->dev, "Bad IPMI Common Header version: 0x%02x",
+			buf[0]);
 		return -EIO;
 	}
 
 	for (csum = 0; len > 0; len--)
 		csum += buf[len - 1];
 	if (csum) {
-		DRM_ERROR("Bad IPMI Common Header checksum: 0x%02x", csum);
+		dev_err(adev->dev, "Bad IPMI Common Header checksum: 0x%02x",
+			csum);
 		return -EIO;
 	}
 
@@ -143,12 +183,14 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev)
 	/* Read the header of the PIA. */
 	len = amdgpu_eeprom_read(adev->pm.fru_eeprom_i2c_bus, addr, buf, 3);
 	if (len != 3) {
-		DRM_ERROR("Couldn't read the Product Info Area header: %d", len);
+		dev_err(adev->dev,
+			"Couldn't read the Product Info Area header: %d", len);
 		return len < 0 ? len : -EIO;
 	}
 
 	if (buf[0] != 1) {
-		DRM_ERROR("Bad IPMI Product Info Area version: 0x%02x", buf[0]);
+		dev_err(adev->dev, "Bad IPMI Product Info Area version: 0x%02x",
+			buf[0]);
 		return -EIO;
 	}
 
@@ -161,40 +203,49 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev)
 	len = amdgpu_eeprom_read(adev->pm.fru_eeprom_i2c_bus, addr, pia, size);
 	if (len != size) {
 		kfree(pia);
-		DRM_ERROR("Couldn't read the Product Info Area: %d", len);
+		dev_err(adev->dev, "Couldn't read the Product Info Area: %d",
+			len);
 		return len < 0 ? len : -EIO;
 	}
 
 	for (csum = 0; size > 0; size--)
 		csum += pia[size - 1];
 	if (csum) {
-		DRM_ERROR("Bad Product Info Area checksum: 0x%02x", csum);
+		dev_err(adev->dev, "Bad Product Info Area checksum: 0x%02x",
+			csum);
+		kfree(pia);
 		return -EIO;
 	}
 
 	/* Now extract useful information from the PIA.
 	 *
-	 * Skip the Manufacturer Name at [3] and go directly to
-	 * the Product Name field.
+	 * Read Manufacturer Name field whose length is [3].
 	 */
-	addr = 3 + 1 + (pia[3] & 0x3F);
+	addr = 3;
 	if (addr + 1 >= len)
 		goto Out;
-	memcpy(adev->product_name, pia + addr + 1,
-	       min_t(size_t,
-		     sizeof(adev->product_name),
+	memcpy(fru_info->manufacturer_name, pia + addr + 1,
+	       min_t(size_t, sizeof(fru_info->manufacturer_name),
 		     pia[addr] & 0x3F));
-	adev->product_name[sizeof(adev->product_name) - 1] = '\0';
+	fru_info->manufacturer_name[sizeof(fru_info->manufacturer_name) - 1] =
+		'\0';
+
+	/* Read Product Name field. */
+	addr += 1 + (pia[addr] & 0x3F);
+	if (addr + 1 >= len)
+		goto Out;
+	memcpy(fru_info->product_name, pia + addr + 1,
+	       min_t(size_t, sizeof(fru_info->product_name), pia[addr] & 0x3F));
+	fru_info->product_name[sizeof(fru_info->product_name) - 1] = '\0';
 
 	/* Go to the Product Part/Model Number field. */
 	addr += 1 + (pia[addr] & 0x3F);
 	if (addr + 1 >= len)
 		goto Out;
-	memcpy(adev->product_number, pia + addr + 1,
-	       min_t(size_t,
-		     sizeof(adev->product_number),
+	memcpy(fru_info->product_number, pia + addr + 1,
+	       min_t(size_t, sizeof(fru_info->product_number),
 		     pia[addr] & 0x3F));
-	adev->product_number[sizeof(adev->product_number) - 1] = '\0';
+	fru_info->product_number[sizeof(fru_info->product_number) - 1] = '\0';
 
 	/* Go to the Product Version field. */
 	addr += 1 + (pia[addr] & 0x3F);
@@ -203,11 +254,155 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev)
 	addr += 1 + (pia[addr] & 0x3F);
 	if (addr + 1 >= len)
 		goto Out;
-	memcpy(adev->serial, pia + addr + 1, min_t(size_t,
-						   sizeof(adev->serial),
-						   pia[addr] & 0x3F));
-	adev->serial[sizeof(adev->serial) - 1] = '\0';
+	memcpy(fru_info->serial, pia + addr + 1,
+	       min_t(size_t, sizeof(fru_info->serial), pia[addr] & 0x3F));
+	fru_info->serial[sizeof(fru_info->serial) - 1] = '\0';
+
+	/* Asset Tag field */
+	addr += 1 + (pia[addr] & 0x3F);
+
+	/* FRU File Id field. This could be 'null'. */
+	addr += 1 + (pia[addr] & 0x3F);
+	if ((addr + 1 >= len) || !(pia[addr] & 0x3F))
+		goto Out;
+	memcpy(fru_info->fru_id, pia + addr + 1,
+	       min_t(size_t, sizeof(fru_info->fru_id), pia[addr] & 0x3F));
+	fru_info->fru_id[sizeof(fru_info->fru_id) - 1] = '\0';
+
 Out:
 	kfree(pia);
 	return 0;
 }
+
+/**
+ * DOC: product_name
+ *
+ * The amdgpu driver provides a sysfs API for reporting the product name
+ * for the device
+ * The file product_name is used for this and returns the product name
+ * as returned from the FRU.
+ * NOTE: This is only available for certain server cards
+ */
+
+static ssize_t amdgpu_fru_product_name_show(struct device *dev,
+					    struct device_attribute *attr,
+					    char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	return sysfs_emit(buf, "%s\n", adev->fru_info->product_name);
+}
+
+static DEVICE_ATTR(product_name, 0444, amdgpu_fru_product_name_show, NULL);
+
+/**
+ * DOC: product_number
+ *
+ * The amdgpu driver provides a sysfs API for reporting the part number
+ * for the device
+ * The file product_number is used for this and returns the part number
+ * as returned from the FRU.
+ * NOTE: This is only available for certain server cards
+ */
+
+static ssize_t amdgpu_fru_product_number_show(struct device *dev,
+					      struct device_attribute *attr,
+					      char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	return sysfs_emit(buf, "%s\n", adev->fru_info->product_number);
+}
+
+static DEVICE_ATTR(product_number, 0444, amdgpu_fru_product_number_show, NULL);
+
+/**
+ * DOC: serial_number
+ *
+ * The amdgpu driver provides a sysfs API for reporting the serial number
+ * for the device
+ * The file serial_number is used for this and returns the serial number
+ * as returned from the FRU.
+ * NOTE: This is only available for certain server cards
+ */
+
+static ssize_t amdgpu_fru_serial_number_show(struct device *dev,
+					     struct device_attribute *attr,
+					     char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	return sysfs_emit(buf, "%s\n", adev->fru_info->serial);
+}
+
+static DEVICE_ATTR(serial_number, 0444, amdgpu_fru_serial_number_show, NULL);
+
+/**
+ * DOC: fru_id
+ *
+ * The amdgpu driver provides a sysfs API for reporting FRU File Id
+ * for the device.
+ * The file fru_id is used for this and returns the File Id value
+ * as returned from the FRU.
+ * NOTE: This is only available for certain server cards
+ */
+
+static ssize_t amdgpu_fru_id_show(struct device *dev,
+				  struct device_attribute *attr, char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	return sysfs_emit(buf, "%s\n", adev->fru_info->fru_id);
+}
+
+static DEVICE_ATTR(fru_id, 0444, amdgpu_fru_id_show, NULL);
+
+/**
+ * DOC: manufacturer
+ *
+ * The amdgpu driver provides a sysfs API for reporting manufacturer name from
+ * FRU information.
+ * The file manufacturer returns the value as returned from the FRU.
+ * NOTE: This is only available for certain server cards
+ */
+
+static ssize_t amdgpu_fru_manufacturer_name_show(struct device *dev,
+						 struct device_attribute *attr,
+						 char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	return sysfs_emit(buf, "%s\n", adev->fru_info->manufacturer_name);
+}
+
+static DEVICE_ATTR(manufacturer, 0444, amdgpu_fru_manufacturer_name_show, NULL);
+
+static const struct attribute *amdgpu_fru_attributes[] = {
+	&dev_attr_product_name.attr,
+	&dev_attr_product_number.attr,
+	&dev_attr_serial_number.attr,
+	&dev_attr_fru_id.attr,
+	&dev_attr_manufacturer.attr,
+	NULL
+};
+
+int amdgpu_fru_sysfs_init(struct amdgpu_device *adev)
+{
+	if (!is_fru_eeprom_supported(adev, NULL) || !adev->fru_info)
+		return 0;
+
+	return sysfs_create_files(&adev->dev->kobj, amdgpu_fru_attributes);
+}
+
+void amdgpu_fru_sysfs_fini(struct amdgpu_device *adev)
+{
+	if (!adev->fru_info)
+		return;
+
+	sysfs_remove_files(&adev->dev->kobj, amdgpu_fru_attributes);
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.h
index 1308d976d60e..98f3196599ef 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.h
@@ -24,6 +24,19 @@
 #ifndef __AMDGPU_FRU_EEPROM_H__
 #define __AMDGPU_FRU_EEPROM_H__
 
+#define AMDGPU_PRODUCT_NAME_LEN 64
+
+/* FRU product information */
+struct amdgpu_fru_info {
+	char				product_number[20];
+	char				product_name[AMDGPU_PRODUCT_NAME_LEN];
+	char				serial[20];
+	char				manufacturer_name[32];
+	char				fru_id[50];
+};
+
 int amdgpu_fru_get_product_info(struct amdgpu_device *adev);
+int amdgpu_fru_sysfs_init(struct amdgpu_device *adev);
+void amdgpu_fru_sysfs_fini(struct amdgpu_device *adev);
 
 #endif  // __AMDGPU_FRU_EEPROM_H__
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fw_attestation.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fw_attestation.c
index 2ca3c329de6d..328a1b963548 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fw_attestation.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fw_attestation.c
@@ -32,17 +32,15 @@
 #include "soc15_common.h"
 
 #define FW_ATTESTATION_DB_COOKIE        0x143b6a37
-#define FW_ATTESTATION_RECORD_VALID  	1
+#define FW_ATTESTATION_RECORD_VALID	1
 #define FW_ATTESTATION_MAX_SIZE		4096
 
-typedef struct FW_ATT_DB_HEADER
-{
+struct FW_ATT_DB_HEADER {
 	uint32_t AttDbVersion;           /* version of the fwar feature */
 	uint32_t AttDbCookie;            /* cookie as an extra check for corrupt data */
-} FW_ATT_DB_HEADER;
+};
 
-typedef struct FW_ATT_RECORD
-{
+struct FW_ATT_RECORD {
 	uint16_t AttFwIdV1;              /* Legacy FW Type field */
 	uint16_t AttFwIdV2;              /* V2 FW ID field */
 	uint32_t AttFWVersion;           /* FW Version */
@@ -50,7 +48,7 @@ typedef struct FW_ATT_RECORD
 	uint8_t  AttSource;              /* FW source indicator */
 	uint8_t  RecordValid;            /* Indicates whether the record is a valid entry */
 	uint32_t AttFwTaId;              /* Ta ID (only in TA Attestation Table) */
-} FW_ATT_RECORD;
+};
 
 static ssize_t amdgpu_fw_attestation_debugfs_read(struct file *f,
 						  char __user *buf,
@@ -60,15 +58,15 @@ static ssize_t amdgpu_fw_attestation_debugfs_read(struct file *f,
 	struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(f)->i_private;
 	uint64_t records_addr = 0;
 	uint64_t vram_pos = 0;
-	FW_ATT_DB_HEADER fw_att_hdr = {0};
-	FW_ATT_RECORD fw_att_record = {0};
+	struct FW_ATT_DB_HEADER fw_att_hdr = {0};
+	struct FW_ATT_RECORD fw_att_record = {0};
 
-	if (size < sizeof(FW_ATT_RECORD)) {
+	if (size < sizeof(struct FW_ATT_RECORD)) {
 		DRM_WARN("FW attestation input buffer not enough memory");
 		return -EINVAL;
 	}
 
-	if ((*pos + sizeof(FW_ATT_DB_HEADER)) >= FW_ATTESTATION_MAX_SIZE) {
+	if ((*pos + sizeof(struct FW_ATT_DB_HEADER)) >= FW_ATTESTATION_MAX_SIZE) {
 		DRM_WARN("FW attestation out of bounds");
 		return 0;
 	}
@@ -83,8 +81,8 @@ static ssize_t amdgpu_fw_attestation_debugfs_read(struct file *f,
 	if (*pos == 0) {
 		amdgpu_device_vram_access(adev,
 					  vram_pos,
-					  (uint32_t*)&fw_att_hdr,
-					  sizeof(FW_ATT_DB_HEADER),
+					  (uint32_t *)&fw_att_hdr,
+					  sizeof(struct FW_ATT_DB_HEADER),
 					  false);
 
 		if (fw_att_hdr.AttDbCookie != FW_ATTESTATION_DB_COOKIE) {
@@ -96,20 +94,20 @@ static ssize_t amdgpu_fw_attestation_debugfs_read(struct file *f,
 	}
 
 	amdgpu_device_vram_access(adev,
-				  vram_pos + sizeof(FW_ATT_DB_HEADER) + *pos,
-				  (uint32_t*)&fw_att_record,
-				  sizeof(FW_ATT_RECORD),
+				  vram_pos + sizeof(struct FW_ATT_DB_HEADER) + *pos,
+				  (uint32_t *)&fw_att_record,
+				  sizeof(struct FW_ATT_RECORD),
 				  false);
 
 	if (fw_att_record.RecordValid != FW_ATTESTATION_RECORD_VALID)
 		return 0;
 
-	if (copy_to_user(buf, (void*)&fw_att_record, sizeof(FW_ATT_RECORD)))
+	if (copy_to_user(buf, (void *)&fw_att_record, sizeof(struct FW_ATT_RECORD)))
 		return -EINVAL;
 
-	*pos += sizeof(FW_ATT_RECORD);
+	*pos += sizeof(struct FW_ATT_RECORD);
 
-	return sizeof(FW_ATT_RECORD);
+	return sizeof(struct FW_ATT_RECORD);
 }
 
 static const struct file_operations amdgpu_fw_attestation_debugfs_ops = {
@@ -124,6 +122,10 @@ static int amdgpu_is_fw_attestation_supported(struct amdgpu_device *adev)
 	if (adev->flags & AMD_IS_APU)
 		return 0;
 
+	if (amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(14, 0, 2) ||
+	    amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(14, 0, 3))
+		return 0;
+
 	if (adev->asic_type >= CHIP_SIENNA_CICHLID)
 		return 1;
 
@@ -136,7 +138,7 @@ void amdgpu_fw_attestation_debugfs_init(struct amdgpu_device *adev)
 		return;
 
 	debugfs_create_file("amdgpu_fw_attestation",
-			    S_IRUSR,
+			    0400,
 			    adev_to_drm(adev)->primary->debugfs_root,
 			    adev,
 			    &amdgpu_fw_attestation_debugfs_ops);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c
index 01cb89ffbd56..d2237ce9da70 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c
@@ -34,7 +34,9 @@
 #include <asm/set_memory.h>
 #endif
 #include "amdgpu.h"
+#include "amdgpu_reset.h"
 #include <drm/drm_drv.h>
+#include <drm/ttm/ttm_tt.h>
 
 /*
  * GART
@@ -76,8 +78,9 @@ static int amdgpu_gart_dummy_page_init(struct amdgpu_device *adev)
 
 	if (adev->dummy_page_addr)
 		return 0;
-	adev->dummy_page_addr = dma_map_page(&adev->pdev->dev, dummy_page, 0,
-					     PAGE_SIZE, DMA_BIDIRECTIONAL);
+	adev->dummy_page_addr = dma_map_page_attrs(&adev->pdev->dev, dummy_page, 0,
+							PAGE_SIZE, DMA_BIDIRECTIONAL,
+							DMA_ATTR_SKIP_CPU_SYNC);
 	if (dma_mapping_error(&adev->pdev->dev, adev->dummy_page_addr)) {
 		dev_err(&adev->pdev->dev, "Failed to DMA MAP the dummy page\n");
 		adev->dummy_page_addr = 0;
@@ -97,12 +100,157 @@ void amdgpu_gart_dummy_page_fini(struct amdgpu_device *adev)
 {
 	if (!adev->dummy_page_addr)
 		return;
-	dma_unmap_page(&adev->pdev->dev, adev->dummy_page_addr, PAGE_SIZE,
-		       DMA_BIDIRECTIONAL);
+	dma_unmap_page_attrs(&adev->pdev->dev, adev->dummy_page_addr, PAGE_SIZE,
+				DMA_BIDIRECTIONAL,
+				DMA_ATTR_SKIP_CPU_SYNC);
 	adev->dummy_page_addr = 0;
 }
 
 /**
+ * amdgpu_gart_table_ram_alloc - allocate system ram for gart page table
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Allocate system memory for GART page table for ASICs that don't have
+ * dedicated VRAM.
+ * Returns 0 for success, error for failure.
+ */
+int amdgpu_gart_table_ram_alloc(struct amdgpu_device *adev)
+{
+	unsigned int order = get_order(adev->gart.table_size);
+	gfp_t gfp_flags = GFP_KERNEL | __GFP_ZERO;
+	struct amdgpu_bo *bo = NULL;
+	struct sg_table *sg = NULL;
+	struct amdgpu_bo_param bp;
+	dma_addr_t dma_addr;
+	struct page *p;
+	unsigned long x;
+	int ret;
+
+	if (adev->gart.bo != NULL)
+		return 0;
+
+	p = alloc_pages(gfp_flags, order);
+	if (!p)
+		return -ENOMEM;
+
+	/* assign pages to this device */
+	for (x = 0; x < (1UL << order); x++)
+		p[x].mapping = adev->mman.bdev.dev_mapping;
+
+	/* If the hardware does not support UTCL2 snooping of the CPU caches
+	 * then set_memory_wc() could be used as a workaround to mark the pages
+	 * as write combine memory.
+	 */
+	dma_addr = dma_map_page(&adev->pdev->dev, p, 0, adev->gart.table_size,
+				DMA_BIDIRECTIONAL);
+	if (dma_mapping_error(&adev->pdev->dev, dma_addr)) {
+		dev_err(&adev->pdev->dev, "Failed to DMA MAP the GART BO page\n");
+		__free_pages(p, order);
+		p = NULL;
+		return -EFAULT;
+	}
+
+	dev_info(adev->dev, "%s dma_addr:%pad\n", __func__, &dma_addr);
+	/* Create SG table */
+	sg = kmalloc(sizeof(*sg), GFP_KERNEL);
+	if (!sg) {
+		ret = -ENOMEM;
+		goto error;
+	}
+	ret = sg_alloc_table(sg, 1, GFP_KERNEL);
+	if (ret)
+		goto error;
+
+	sg_dma_address(sg->sgl) = dma_addr;
+	sg->sgl->length = adev->gart.table_size;
+#ifdef CONFIG_NEED_SG_DMA_LENGTH
+	sg->sgl->dma_length = adev->gart.table_size;
+#endif
+	/* Create SG BO */
+	memset(&bp, 0, sizeof(bp));
+	bp.size = adev->gart.table_size;
+	bp.byte_align = PAGE_SIZE;
+	bp.domain = AMDGPU_GEM_DOMAIN_CPU;
+	bp.type = ttm_bo_type_sg;
+	bp.resv = NULL;
+	bp.bo_ptr_size = sizeof(struct amdgpu_bo);
+	bp.flags = 0;
+	ret = amdgpu_bo_create(adev, &bp, &bo);
+	if (ret)
+		goto error;
+
+	bo->tbo.sg = sg;
+	bo->tbo.ttm->sg = sg;
+	bo->allowed_domains = AMDGPU_GEM_DOMAIN_GTT;
+	bo->preferred_domains = AMDGPU_GEM_DOMAIN_GTT;
+
+	ret = amdgpu_bo_reserve(bo, true);
+	if (ret) {
+		dev_err(adev->dev, "(%d) failed to reserve bo for GART system bo\n", ret);
+		goto error;
+	}
+
+	ret = amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT);
+	WARN(ret, "Pinning the GART table failed");
+	if (ret)
+		goto error_resv;
+
+	adev->gart.bo = bo;
+	adev->gart.ptr = page_to_virt(p);
+	/* Make GART table accessible in VMID0 */
+	ret = amdgpu_ttm_alloc_gart(&adev->gart.bo->tbo);
+	if (ret)
+		amdgpu_gart_table_ram_free(adev);
+	amdgpu_bo_unreserve(bo);
+
+	return 0;
+
+error_resv:
+	amdgpu_bo_unreserve(bo);
+error:
+	amdgpu_bo_unref(&bo);
+	if (sg) {
+		sg_free_table(sg);
+		kfree(sg);
+	}
+	__free_pages(p, order);
+	return ret;
+}
+
+/**
+ * amdgpu_gart_table_ram_free - free gart page table system ram
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Free the system memory used for the GART page tableon ASICs that don't
+ * have dedicated VRAM.
+ */
+void amdgpu_gart_table_ram_free(struct amdgpu_device *adev)
+{
+	unsigned int order = get_order(adev->gart.table_size);
+	struct sg_table *sg = adev->gart.bo->tbo.sg;
+	struct page *p;
+	unsigned long x;
+	int ret;
+
+	ret = amdgpu_bo_reserve(adev->gart.bo, false);
+	if (!ret) {
+		amdgpu_bo_unpin(adev->gart.bo);
+		amdgpu_bo_unreserve(adev->gart.bo);
+	}
+	amdgpu_bo_unref(&adev->gart.bo);
+	sg_free_table(sg);
+	kfree(sg);
+	p = virt_to_page(adev->gart.ptr);
+	for (x = 0; x < (1UL << order); x++)
+		p[x].mapping = NULL;
+	__free_pages(p, order);
+
+	adev->gart.ptr = NULL;
+}
+
+/**
  * amdgpu_gart_table_vram_alloc - allocate vram for gart page table
  *
  * @adev: amdgpu_device pointer
@@ -154,7 +302,6 @@ void amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset,
 			int pages)
 {
 	unsigned t;
-	unsigned p;
 	int i, j;
 	u64 page_base;
 	/* Starting from VEGA10, system bit must be 0 to mean invalid. */
@@ -168,8 +315,7 @@ void amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset,
 		return;
 
 	t = offset / AMDGPU_GPU_PAGE_SIZE;
-	p = t / AMDGPU_GPU_PAGES_IN_CPU_PAGE;
-	for (i = 0; i < pages; i++, p++) {
+	for (i = 0; i < pages; i++) {
 		page_base = adev->dummy_page_addr;
 		if (!adev->gart.ptr)
 			continue;
@@ -180,10 +326,7 @@ void amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset,
 			page_base += AMDGPU_GPU_PAGE_SIZE;
 		}
 	}
-	mb();
-	amdgpu_device_flush_hdp(adev, NULL);
-	for (i = 0; i < adev->num_vmhubs; i++)
-		amdgpu_gmc_flush_gpu_tlb(adev, 0, i, 0);
+	amdgpu_gart_invalidate_tlb(adev);
 
 	drm_dev_exit(idx);
 }
@@ -225,6 +368,42 @@ void amdgpu_gart_map(struct amdgpu_device *adev, uint64_t offset,
 }
 
 /**
+ * amdgpu_gart_map_vram_range - map VRAM pages into the GART page table
+ *
+ * @adev: amdgpu_device pointer
+ * @pa: physical address of the first page to be mapped
+ * @start_page: first page to map in the GART aperture
+ * @num_pages: number of pages to be mapped
+ * @flags: page table entry flags
+ * @dst: CPU address of the GART table
+ *
+ * Binds a BO that is allocated in VRAM to the GART page table
+ * (all ASICs).
+ *
+ * Useful when a kernel BO is located in VRAM but
+ * needs to be accessed from the GART address space.
+ */
+void amdgpu_gart_map_vram_range(struct amdgpu_device *adev, uint64_t pa,
+				uint64_t start_page, uint64_t num_pages,
+				uint64_t flags, void *dst)
+{
+	u32 i, idx;
+
+	/* The SYSTEM flag indicates the pages aren't in VRAM. */
+	WARN_ON_ONCE(flags & AMDGPU_PTE_SYSTEM);
+
+	if (!drm_dev_enter(adev_to_drm(adev), &idx))
+		return;
+
+	for (i = 0; i < num_pages; ++i) {
+		amdgpu_gmc_set_pte_pde(adev, adev->gart.ptr,
+			start_page + i, pa + AMDGPU_GPU_PAGE_SIZE * i, flags);
+	}
+
+	drm_dev_exit(idx);
+}
+
+/**
  * amdgpu_gart_bind - bind pages into the gart page table
  *
  * @adev: amdgpu_device pointer
@@ -263,8 +442,11 @@ void amdgpu_gart_invalidate_tlb(struct amdgpu_device *adev)
 		return;
 
 	mb();
-	amdgpu_device_flush_hdp(adev, NULL);
-	for (i = 0; i < adev->num_vmhubs; i++)
+	if (down_read_trylock(&adev->reset_domain->sem)) {
+		amdgpu_device_flush_hdp(adev, NULL);
+		up_read(&adev->reset_domain->sem);
+	}
+	for_each_set_bit(i, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS)
 		amdgpu_gmc_flush_gpu_tlb(adev, 0, i, 0);
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h
index 8fea3e04e411..d3118275ddae 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h
@@ -51,10 +51,10 @@ struct amdgpu_gart {
 	uint64_t			gart_pte_flags;
 };
 
+int amdgpu_gart_table_ram_alloc(struct amdgpu_device *adev);
+void amdgpu_gart_table_ram_free(struct amdgpu_device *adev);
 int amdgpu_gart_table_vram_alloc(struct amdgpu_device *adev);
 void amdgpu_gart_table_vram_free(struct amdgpu_device *adev);
-int amdgpu_gart_table_vram_pin(struct amdgpu_device *adev);
-void amdgpu_gart_table_vram_unpin(struct amdgpu_device *adev);
 int amdgpu_gart_init(struct amdgpu_device *adev);
 void amdgpu_gart_dummy_page_fini(struct amdgpu_device *adev);
 void amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset,
@@ -64,5 +64,8 @@ void amdgpu_gart_map(struct amdgpu_device *adev, uint64_t offset,
 		     void *dst);
 void amdgpu_gart_bind(struct amdgpu_device *adev, uint64_t offset,
 		      int pages, dma_addr_t *dma_addr, uint64_t flags);
+void amdgpu_gart_map_vram_range(struct amdgpu_device *adev, uint64_t pa,
+				uint64_t start_page, uint64_t num_pages,
+				uint64_t flags, void *dst);
 void amdgpu_gart_invalidate_tlb(struct amdgpu_device *adev);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
index 863cb668e000..3e38c5db2987 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
@@ -33,16 +33,125 @@
 
 #include <drm/amdgpu_drm.h>
 #include <drm/drm_drv.h>
+#include <drm/drm_exec.h>
 #include <drm/drm_gem_ttm_helper.h>
 #include <drm/ttm/ttm_tt.h>
+#include <drm/drm_syncobj.h>
 
 #include "amdgpu.h"
 #include "amdgpu_display.h"
 #include "amdgpu_dma_buf.h"
 #include "amdgpu_hmm.h"
 #include "amdgpu_xgmi.h"
+#include "amdgpu_vm.h"
 
-static const struct drm_gem_object_funcs amdgpu_gem_object_funcs;
+static int
+amdgpu_gem_add_input_fence(struct drm_file *filp,
+			   uint64_t syncobj_handles_array,
+			   uint32_t num_syncobj_handles)
+{
+	struct dma_fence *fence;
+	uint32_t *syncobj_handles;
+	int ret, i;
+
+	if (!num_syncobj_handles)
+		return 0;
+
+	syncobj_handles = memdup_user(u64_to_user_ptr(syncobj_handles_array),
+				      size_mul(sizeof(uint32_t), num_syncobj_handles));
+	if (IS_ERR(syncobj_handles))
+		return PTR_ERR(syncobj_handles);
+
+	for (i = 0; i < num_syncobj_handles; i++) {
+
+		if (!syncobj_handles[i]) {
+			ret = -EINVAL;
+			goto free_memdup;
+		}
+
+		ret = drm_syncobj_find_fence(filp, syncobj_handles[i], 0, 0, &fence);
+		if (ret)
+			goto free_memdup;
+
+		dma_fence_wait(fence, false);
+
+		/* TODO: optimize async handling */
+		dma_fence_put(fence);
+	}
+
+free_memdup:
+	kfree(syncobj_handles);
+	return ret;
+}
+
+static int
+amdgpu_gem_update_timeline_node(struct drm_file *filp,
+				uint32_t syncobj_handle,
+				uint64_t point,
+				struct drm_syncobj **syncobj,
+				struct dma_fence_chain **chain)
+{
+	if (!syncobj_handle)
+		return 0;
+
+	/* Find the sync object */
+	*syncobj = drm_syncobj_find(filp, syncobj_handle);
+	if (!*syncobj)
+		return -ENOENT;
+
+	if (!point)
+		return 0;
+
+	/* Allocate the chain node */
+	*chain = dma_fence_chain_alloc();
+	if (!*chain) {
+		drm_syncobj_put(*syncobj);
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static void
+amdgpu_gem_update_bo_mapping(struct drm_file *filp,
+			     struct amdgpu_bo_va *bo_va,
+			     uint32_t operation,
+			     uint64_t point,
+			     struct dma_fence *fence,
+			     struct drm_syncobj *syncobj,
+			     struct dma_fence_chain *chain)
+{
+	struct amdgpu_bo *bo = bo_va ? bo_va->base.bo : NULL;
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
+	struct amdgpu_vm *vm = &fpriv->vm;
+	struct dma_fence *last_update;
+
+	if (!syncobj)
+		return;
+
+	/* Find the last update fence */
+	switch (operation) {
+	case AMDGPU_VA_OP_MAP:
+	case AMDGPU_VA_OP_REPLACE:
+		if (bo && (bo->tbo.base.resv == vm->root.bo->tbo.base.resv))
+			last_update = vm->last_update;
+		else
+			last_update = bo_va->last_pt_update;
+		break;
+	case AMDGPU_VA_OP_UNMAP:
+	case AMDGPU_VA_OP_CLEAR:
+		last_update = fence;
+		break;
+	default:
+		return;
+	}
+
+	/* Add fence to timeline */
+	if (!point)
+		drm_syncobj_replace_fence(syncobj, last_update);
+	else
+		drm_syncobj_add_point(syncobj, chain, last_update, point);
+}
 
 static vm_fault_t amdgpu_gem_fault(struct vm_fault *vmf)
 {
@@ -86,19 +195,17 @@ static const struct vm_operations_struct amdgpu_gem_vm_ops = {
 
 static void amdgpu_gem_object_free(struct drm_gem_object *gobj)
 {
-	struct amdgpu_bo *robj = gem_to_amdgpu_bo(gobj);
+	struct amdgpu_bo *aobj = gem_to_amdgpu_bo(gobj);
 
-	if (robj) {
-		amdgpu_hmm_unregister(robj);
-		amdgpu_bo_unref(&robj);
-	}
+	amdgpu_hmm_unregister(aobj);
+	ttm_bo_fini(&aobj->tbo);
 }
 
 int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
 			     int alignment, u32 initial_domain,
 			     u64 flags, enum ttm_bo_type type,
 			     struct dma_resv *resv,
-			     struct drm_gem_object **obj)
+			     struct drm_gem_object **obj, int8_t xcp_id_plus1)
 {
 	struct amdgpu_bo *bo;
 	struct amdgpu_bo_user *ubo;
@@ -107,6 +214,7 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
 
 	memset(&bp, 0, sizeof(bp));
 	*obj = NULL;
+	flags |= AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE;
 
 	bp.size = size;
 	bp.byte_align = alignment;
@@ -116,6 +224,7 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
 	bp.flags = flags;
 	bp.domain = initial_domain;
 	bp.bo_ptr_size = sizeof(struct amdgpu_bo);
+	bp.xcp_id_plus1 = xcp_id_plus1;
 
 	r = amdgpu_bo_create_user(adev, &bp, &ubo);
 	if (r)
@@ -123,7 +232,6 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
 
 	bo = &ubo->bo;
 	*obj = &bo->tbo.base;
-	(*obj)->funcs = &amdgpu_gem_object_funcs;
 
 	return 0;
 }
@@ -172,21 +280,63 @@ static int amdgpu_gem_object_open(struct drm_gem_object *obj,
 		return -EPERM;
 
 	if (abo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID &&
-	    abo->tbo.base.resv != vm->root.bo->tbo.base.resv)
+	    !amdgpu_vm_is_bo_always_valid(vm, abo))
 		return -EPERM;
 
 	r = amdgpu_bo_reserve(abo, false);
 	if (r)
 		return r;
 
+	amdgpu_vm_bo_update_shared(abo);
 	bo_va = amdgpu_vm_bo_find(vm, abo);
-	if (!bo_va) {
+	if (!bo_va)
 		bo_va = amdgpu_vm_bo_add(adev, vm, abo);
-	} else {
+	else
 		++bo_va->ref_count;
+
+	/* attach gfx eviction fence */
+	r = amdgpu_eviction_fence_attach(&fpriv->evf_mgr, abo);
+	if (r) {
+		DRM_DEBUG_DRIVER("Failed to attach eviction fence to BO\n");
+		amdgpu_bo_unreserve(abo);
+		return r;
 	}
+
 	amdgpu_bo_unreserve(abo);
-	return 0;
+
+	/* Validate and add eviction fence to DMABuf imports with dynamic
+	 * attachment in compute VMs. Re-validation will be done by
+	 * amdgpu_vm_validate. Fences are on the reservation shared with the
+	 * export, which is currently required to be validated and fenced
+	 * already by amdgpu_amdkfd_gpuvm_restore_process_bos.
+	 *
+	 * Nested locking below for the case that a GEM object is opened in
+	 * kfd_mem_export_dmabuf. Since the lock below is only taken for imports,
+	 * but not for export, this is a different lock class that cannot lead to
+	 * circular lock dependencies.
+	 */
+	if (!vm->is_compute_context || !vm->process_info)
+		return 0;
+	if (!drm_gem_is_imported(obj) ||
+	    !dma_buf_is_dynamic(obj->import_attach->dmabuf))
+		return 0;
+	mutex_lock_nested(&vm->process_info->lock, 1);
+	if (!WARN_ON(!vm->process_info->eviction_fence)) {
+		r = amdgpu_amdkfd_bo_validate_and_fence(abo, AMDGPU_GEM_DOMAIN_GTT,
+							&vm->process_info->eviction_fence->base);
+		if (r) {
+			struct amdgpu_task_info *ti = amdgpu_vm_get_task_info_vm(vm);
+
+			dev_warn(adev->dev, "validate_and_fence failed: %d\n", r);
+			if (ti) {
+				dev_warn(adev->dev, "pid %d\n", ti->task.pid);
+				amdgpu_vm_put_task_info(ti);
+			}
+		}
+	}
+	mutex_unlock(&vm->process_info->lock);
+
+	return r;
 }
 
 static void amdgpu_gem_object_close(struct drm_gem_object *obj,
@@ -197,38 +347,40 @@ static void amdgpu_gem_object_close(struct drm_gem_object *obj,
 	struct amdgpu_fpriv *fpriv = file_priv->driver_priv;
 	struct amdgpu_vm *vm = &fpriv->vm;
 
-	struct amdgpu_bo_list_entry vm_pd;
-	struct list_head list, duplicates;
 	struct dma_fence *fence = NULL;
-	struct ttm_validate_buffer tv;
-	struct ww_acquire_ctx ticket;
 	struct amdgpu_bo_va *bo_va;
+	struct drm_exec exec;
 	long r;
 
-	INIT_LIST_HEAD(&list);
-	INIT_LIST_HEAD(&duplicates);
-
-	tv.bo = &bo->tbo;
-	tv.num_shared = 2;
-	list_add(&tv.head, &list);
+	drm_exec_init(&exec, DRM_EXEC_IGNORE_DUPLICATES, 0);
+	drm_exec_until_all_locked(&exec) {
+		r = drm_exec_prepare_obj(&exec, &bo->tbo.base, 1);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(r))
+			goto out_unlock;
+
+		r = amdgpu_vm_lock_pd(vm, &exec, 0);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(r))
+			goto out_unlock;
+	}
 
-	amdgpu_vm_get_pd_bo(vm, &list, &vm_pd);
+	if (!amdgpu_vm_is_bo_always_valid(vm, bo))
+		amdgpu_eviction_fence_detach(&fpriv->evf_mgr, bo);
 
-	r = ttm_eu_reserve_buffers(&ticket, &list, false, &duplicates);
-	if (r) {
-		dev_err(adev->dev, "leaking bo va because "
-			"we fail to reserve bo (%ld)\n", r);
-		return;
-	}
 	bo_va = amdgpu_vm_bo_find(vm, bo);
 	if (!bo_va || --bo_va->ref_count)
 		goto out_unlock;
 
 	amdgpu_vm_bo_del(adev, bo_va);
+	amdgpu_vm_bo_update_shared(bo);
 	if (!amdgpu_vm_ready(vm))
 		goto out_unlock;
 
 	r = amdgpu_vm_clear_freed(adev, vm, &fence);
+	if (unlikely(r < 0))
+		dev_err(adev->dev, "failed to clear page "
+			"tables on GEM object close (%ld)\n", r);
 	if (r || !fence)
 		goto out_unlock;
 
@@ -236,10 +388,9 @@ static void amdgpu_gem_object_close(struct drm_gem_object *obj,
 	dma_fence_put(fence);
 
 out_unlock:
-	if (unlikely(r < 0))
-		dev_err(adev->dev, "failed to clear page "
-			"tables on GEM object close (%ld)\n", r);
-	ttm_eu_backoff_reservation(&ticket, &list);
+	if (r)
+		dev_err(adev->dev, "leaking bo va (%ld)\n", r);
+	drm_exec_fini(&exec);
 }
 
 static int amdgpu_gem_object_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)
@@ -263,7 +414,7 @@ static int amdgpu_gem_object_mmap(struct drm_gem_object *obj, struct vm_area_str
 	return drm_gem_ttm_mmap(obj, vma);
 }
 
-static const struct drm_gem_object_funcs amdgpu_gem_object_funcs = {
+const struct drm_gem_object_funcs amdgpu_gem_object_funcs = {
 	.free = amdgpu_gem_object_free,
 	.open = amdgpu_gem_object_open,
 	.close = amdgpu_gem_object_close,
@@ -292,14 +443,7 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
 	int r;
 
 	/* reject invalid gem flags */
-	if (flags & ~(AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
-		      AMDGPU_GEM_CREATE_NO_CPU_ACCESS |
-		      AMDGPU_GEM_CREATE_CPU_GTT_USWC |
-		      AMDGPU_GEM_CREATE_VRAM_CLEARED |
-		      AMDGPU_GEM_CREATE_VM_ALWAYS_VALID |
-		      AMDGPU_GEM_CREATE_EXPLICIT_SYNC |
-		      AMDGPU_GEM_CREATE_ENCRYPTED |
-		      AMDGPU_GEM_CREATE_DISCARDABLE))
+	if (flags & ~AMDGPU_GEM_CREATE_SETTABLE_MASK)
 		return -EINVAL;
 
 	/* reject invalid gem domains */
@@ -311,6 +455,12 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
 		return -EINVAL;
 	}
 
+	/* always clear VRAM */
+	flags |= AMDGPU_GEM_CREATE_VRAM_CLEARED;
+
+	if (args->in.domains & AMDGPU_GEM_DOMAIN_MMIO_REMAP)
+		return -EINVAL;
+
 	/* create a gem object to contain this object in */
 	if (args->in.domains & (AMDGPU_GEM_DOMAIN_GDS |
 	    AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA)) {
@@ -336,7 +486,7 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
 retry:
 	r = amdgpu_gem_object_create(adev, size, args->in.alignment,
 				     initial_domain,
-				     flags, ttm_bo_type_device, resv, &gobj);
+				     flags, ttm_bo_type_device, resv, &gobj, fpriv->xcp_id + 1);
 	if (r && r != -ERESTARTSYS) {
 		if (flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) {
 			flags &= ~AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
@@ -379,8 +529,9 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
 	struct ttm_operation_ctx ctx = { true, false };
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct drm_amdgpu_gem_userptr *args = data;
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
 	struct drm_gem_object *gobj;
-	struct hmm_range *range;
+	struct amdgpu_hmm_range *range;
 	struct amdgpu_bo *bo;
 	uint32_t handle;
 	int r;
@@ -405,7 +556,7 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
 
 	/* create a gem object to contain this object in */
 	r = amdgpu_gem_object_create(adev, args->size, 0, AMDGPU_GEM_DOMAIN_CPU,
-				     0, ttm_bo_type_device, NULL, &gobj);
+				     0, ttm_bo_type_device, NULL, &gobj, fpriv->xcp_id + 1);
 	if (r)
 		return r;
 
@@ -421,15 +572,20 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
 		goto release_object;
 
 	if (args->flags & AMDGPU_GEM_USERPTR_VALIDATE) {
-		r = amdgpu_ttm_tt_get_user_pages(bo, bo->tbo.ttm->pages,
-						 &range);
-		if (r)
+		range = amdgpu_hmm_range_alloc(NULL);
+		if (unlikely(!range))
+			return -ENOMEM;
+		r = amdgpu_ttm_tt_get_user_pages(bo, range);
+		if (r) {
+			amdgpu_hmm_range_free(range);
 			goto release_object;
-
+		}
 		r = amdgpu_bo_reserve(bo, true);
 		if (r)
 			goto user_pages_done;
 
+		amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, range);
+
 		amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT);
 		r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 		amdgpu_bo_unreserve(bo);
@@ -445,8 +601,7 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
 
 user_pages_done:
 	if (args->flags & AMDGPU_GEM_USERPTR_VALIDATE)
-		amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm, range);
-
+		amdgpu_hmm_range_free(range);
 release_object:
 	drm_gem_object_put(gobj);
 
@@ -461,9 +616,9 @@ int amdgpu_mode_dumb_mmap(struct drm_file *filp,
 	struct amdgpu_bo *robj;
 
 	gobj = drm_gem_object_lookup(filp, handle);
-	if (gobj == NULL) {
+	if (!gobj)
 		return -ENOENT;
-	}
+
 	robj = gem_to_amdgpu_bo(gobj);
 	if (amdgpu_ttm_tt_get_usermm(robj->tbo.ttm) ||
 	    (robj->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS)) {
@@ -480,6 +635,7 @@ int amdgpu_gem_mmap_ioctl(struct drm_device *dev, void *data,
 {
 	union drm_amdgpu_gem_mmap *args = data;
 	uint32_t handle = args->in.handle;
+
 	memset(args, 0, sizeof(*args));
 	return amdgpu_mode_dumb_mmap(filp, dev, handle, &args->out.addr_ptr);
 }
@@ -506,7 +662,7 @@ unsigned long amdgpu_gem_timeout(uint64_t timeout_ns)
 
 	timeout_jiffies = nsecs_to_jiffies(ktime_to_ns(timeout));
 	/*  clamp timeout to avoid unsigned-> signed overflow */
-	if (timeout_jiffies > MAX_SCHEDULE_TIMEOUT )
+	if (timeout_jiffies > MAX_SCHEDULE_TIMEOUT)
 		return MAX_SCHEDULE_TIMEOUT - 1;
 
 	return timeout_jiffies;
@@ -524,9 +680,9 @@ int amdgpu_gem_wait_idle_ioctl(struct drm_device *dev, void *data,
 	long ret;
 
 	gobj = drm_gem_object_lookup(filp, handle);
-	if (gobj == NULL) {
+	if (!gobj)
 		return -ENOENT;
-	}
+
 	robj = gem_to_amdgpu_bo(gobj);
 	ret = dma_resv_wait_timeout(robj->tbo.base.resv, DMA_RESV_USAGE_READ,
 				    true, timeout);
@@ -553,7 +709,7 @@ int amdgpu_gem_metadata_ioctl(struct drm_device *dev, void *data,
 	struct amdgpu_bo *robj;
 	int r = -1;
 
-	DRM_DEBUG("%d \n", args->handle);
+	DRM_DEBUG("%d\n", args->handle);
 	gobj = drm_gem_object_lookup(filp, args->handle);
 	if (gobj == NULL)
 		return -ENOENT;
@@ -598,18 +754,23 @@ out:
  *
  * Update the bo_va directly after setting its address. Errors are not
  * vital here, so they are not reported back to userspace.
+ *
+ * Returns resulting fence if freed BO(s) got cleared from the PT.
+ * otherwise stub fence in case of error.
  */
-static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
-				    struct amdgpu_vm *vm,
-				    struct amdgpu_bo_va *bo_va,
-				    uint32_t operation)
+static struct dma_fence *
+amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
+			struct amdgpu_vm *vm,
+			struct amdgpu_bo_va *bo_va,
+			uint32_t operation)
 {
+	struct dma_fence *fence = dma_fence_get_stub();
 	int r;
 
 	if (!amdgpu_vm_ready(vm))
-		return;
+		return fence;
 
-	r = amdgpu_vm_clear_freed(adev, vm, NULL);
+	r = amdgpu_vm_clear_freed(adev, vm, &fence);
 	if (r)
 		goto error;
 
@@ -625,36 +786,8 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
 error:
 	if (r && r != -ERESTARTSYS)
 		DRM_ERROR("Couldn't update BO_VA (%d)\n", r);
-}
 
-/**
- * amdgpu_gem_va_map_flags - map GEM UAPI flags into hardware flags
- *
- * @adev: amdgpu_device pointer
- * @flags: GEM UAPI flags
- *
- * Returns the GEM UAPI flags mapped into hardware for the ASIC.
- */
-uint64_t amdgpu_gem_va_map_flags(struct amdgpu_device *adev, uint32_t flags)
-{
-	uint64_t pte_flag = 0;
-
-	if (flags & AMDGPU_VM_PAGE_EXECUTABLE)
-		pte_flag |= AMDGPU_PTE_EXECUTABLE;
-	if (flags & AMDGPU_VM_PAGE_READABLE)
-		pte_flag |= AMDGPU_PTE_READABLE;
-	if (flags & AMDGPU_VM_PAGE_WRITEABLE)
-		pte_flag |= AMDGPU_PTE_WRITEABLE;
-	if (flags & AMDGPU_VM_PAGE_PRT)
-		pte_flag |= AMDGPU_PTE_PRT;
-	if (flags & AMDGPU_VM_PAGE_NOALLOC)
-		pte_flag |= AMDGPU_PTE_NOALLOC;
-
-	if (adev->gmc.gmc_funcs->map_mtype)
-		pte_flag |= amdgpu_gmc_map_mtype(adev,
-						 flags & AMDGPU_VM_MTYPE_MASK);
-
-	return pte_flag;
+	return fence;
 }
 
 int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
@@ -673,25 +806,24 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 	struct amdgpu_fpriv *fpriv = filp->driver_priv;
 	struct amdgpu_bo *abo;
 	struct amdgpu_bo_va *bo_va;
-	struct amdgpu_bo_list_entry vm_pd;
-	struct ttm_validate_buffer tv;
-	struct ww_acquire_ctx ticket;
-	struct list_head list, duplicates;
-	uint64_t va_flags;
+	struct drm_syncobj *timeline_syncobj = NULL;
+	struct dma_fence_chain *timeline_chain = NULL;
+	struct dma_fence *fence;
+	struct drm_exec exec;
 	uint64_t vm_size;
 	int r = 0;
 
-	if (args->va_address < AMDGPU_VA_RESERVED_SIZE) {
+	if (args->va_address < AMDGPU_VA_RESERVED_BOTTOM) {
 		dev_dbg(dev->dev,
-			"va_address 0x%LX is in reserved area 0x%LX\n",
-			args->va_address, AMDGPU_VA_RESERVED_SIZE);
+			"va_address 0x%llx is in reserved area 0x%llx\n",
+			args->va_address, AMDGPU_VA_RESERVED_BOTTOM);
 		return -EINVAL;
 	}
 
 	if (args->va_address >= AMDGPU_GMC_HOLE_START &&
 	    args->va_address < AMDGPU_GMC_HOLE_END) {
 		dev_dbg(dev->dev,
-			"va_address 0x%LX is in VA hole 0x%LX-0x%LX\n",
+			"va_address 0x%llx is in VA hole 0x%llx-0x%llx\n",
 			args->va_address, AMDGPU_GMC_HOLE_START,
 			AMDGPU_GMC_HOLE_END);
 		return -EINVAL;
@@ -700,7 +832,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 	args->va_address &= AMDGPU_GMC_HOLE_MASK;
 
 	vm_size = adev->vm_manager.max_pfn * AMDGPU_GPU_PAGE_SIZE;
-	vm_size -= AMDGPU_VA_RESERVED_SIZE;
+	vm_size -= AMDGPU_VA_RESERVED_TOP;
 	if (args->va_address + args->map_size > vm_size) {
 		dev_dbg(dev->dev,
 			"va_address 0x%llx is in top reserved area 0x%llx\n",
@@ -726,36 +858,44 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 		return -EINVAL;
 	}
 
-	INIT_LIST_HEAD(&list);
-	INIT_LIST_HEAD(&duplicates);
 	if ((args->operation != AMDGPU_VA_OP_CLEAR) &&
 	    !(args->flags & AMDGPU_VM_PAGE_PRT)) {
 		gobj = drm_gem_object_lookup(filp, args->handle);
 		if (gobj == NULL)
 			return -ENOENT;
 		abo = gem_to_amdgpu_bo(gobj);
-		tv.bo = &abo->tbo;
-		if (abo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID)
-			tv.num_shared = 1;
-		else
-			tv.num_shared = 0;
-		list_add(&tv.head, &list);
 	} else {
 		gobj = NULL;
 		abo = NULL;
 	}
 
-	amdgpu_vm_get_pd_bo(&fpriv->vm, &list, &vm_pd);
-
-	r = ttm_eu_reserve_buffers(&ticket, &list, true, &duplicates);
+	r = amdgpu_gem_add_input_fence(filp,
+				       args->input_fence_syncobj_handles,
+				       args->num_syncobj_handles);
 	if (r)
-		goto error_unref;
+		goto error_put_gobj;
+
+	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT |
+		      DRM_EXEC_IGNORE_DUPLICATES, 0);
+	drm_exec_until_all_locked(&exec) {
+		if (gobj) {
+			r = drm_exec_lock_obj(&exec, gobj);
+			drm_exec_retry_on_contention(&exec);
+			if (unlikely(r))
+				goto error;
+		}
+
+		r = amdgpu_vm_lock_pd(&fpriv->vm, &exec, 2);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(r))
+			goto error;
+	}
 
 	if (abo) {
 		bo_va = amdgpu_vm_bo_find(&fpriv->vm, abo);
 		if (!bo_va) {
 			r = -ENOENT;
-			goto error_backoff;
+			goto error;
 		}
 	} else if (args->operation != AMDGPU_VA_OP_CLEAR) {
 		bo_va = fpriv->prt_va;
@@ -763,12 +903,19 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 		bo_va = NULL;
 	}
 
+	r = amdgpu_gem_update_timeline_node(filp,
+					    args->vm_timeline_syncobj_out,
+					    args->vm_timeline_point,
+					    &timeline_syncobj,
+					    &timeline_chain);
+	if (r)
+		goto error;
+
 	switch (args->operation) {
 	case AMDGPU_VA_OP_MAP:
-		va_flags = amdgpu_gem_va_map_flags(adev, args->flags);
 		r = amdgpu_vm_bo_map(adev, bo_va, args->va_address,
 				     args->offset_in_bo, args->map_size,
-				     va_flags);
+				     args->flags);
 		break;
 	case AMDGPU_VA_OP_UNMAP:
 		r = amdgpu_vm_bo_unmap(adev, bo_va, args->va_address);
@@ -780,22 +927,31 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 						args->map_size);
 		break;
 	case AMDGPU_VA_OP_REPLACE:
-		va_flags = amdgpu_gem_va_map_flags(adev, args->flags);
 		r = amdgpu_vm_bo_replace_map(adev, bo_va, args->va_address,
 					     args->offset_in_bo, args->map_size,
-					     va_flags);
+					     args->flags);
 		break;
 	default:
 		break;
 	}
-	if (!r && !(args->flags & AMDGPU_VM_DELAY_UPDATE) && !amdgpu_vm_debug)
-		amdgpu_gem_va_update_vm(adev, &fpriv->vm, bo_va,
-					args->operation);
+	if (!r && !(args->flags & AMDGPU_VM_DELAY_UPDATE) && !adev->debug_vm) {
+		fence = amdgpu_gem_va_update_vm(adev, &fpriv->vm, bo_va,
+						args->operation);
+
+		if (timeline_syncobj)
+			amdgpu_gem_update_bo_mapping(filp, bo_va,
+					     args->operation,
+					     args->vm_timeline_point,
+					     fence, timeline_syncobj,
+					     timeline_chain);
+		else
+			dma_fence_put(fence);
 
-error_backoff:
-	ttm_eu_backoff_reservation(&ticket, &list);
+	}
 
-error_unref:
+error:
+	drm_exec_fini(&exec);
+error_put_gobj:
 	drm_gem_object_put(gobj);
 	return r;
 }
@@ -803,22 +959,38 @@ error_unref:
 int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *filp)
 {
-	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct drm_amdgpu_gem_op *args = data;
 	struct drm_gem_object *gobj;
 	struct amdgpu_vm_bo_base *base;
 	struct amdgpu_bo *robj;
+	struct drm_exec exec;
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
 	int r;
 
+	if (args->padding)
+		return -EINVAL;
+
 	gobj = drm_gem_object_lookup(filp, args->handle);
-	if (gobj == NULL) {
+	if (!gobj)
 		return -ENOENT;
-	}
+
 	robj = gem_to_amdgpu_bo(gobj);
 
-	r = amdgpu_bo_reserve(robj, false);
-	if (unlikely(r))
-		goto out;
+	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT |
+			  DRM_EXEC_IGNORE_DUPLICATES, 0);
+	drm_exec_until_all_locked(&exec) {
+		r = drm_exec_lock_obj(&exec, gobj);
+		drm_exec_retry_on_contention(&exec);
+		if (r)
+			goto out_exec;
+
+		if (args->op == AMDGPU_GEM_OP_GET_MAPPING_INFO) {
+			r = amdgpu_vm_lock_pd(&fpriv->vm, &exec, 0);
+			drm_exec_retry_on_contention(&exec);
+			if (r)
+				goto out_exec;
+		}
+	}
 
 	switch (args->op) {
 	case AMDGPU_GEM_OP_GET_GEM_CREATE_INFO: {
@@ -829,29 +1001,26 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 		info.alignment = robj->tbo.page_alignment << PAGE_SHIFT;
 		info.domains = robj->preferred_domains;
 		info.domain_flags = robj->flags;
-		amdgpu_bo_unreserve(robj);
+		drm_exec_fini(&exec);
 		if (copy_to_user(out, &info, sizeof(info)))
 			r = -EFAULT;
 		break;
 	}
 	case AMDGPU_GEM_OP_SET_PLACEMENT:
-		if (robj->tbo.base.import_attach &&
+		if (drm_gem_is_imported(&robj->tbo.base) &&
 		    args->value & AMDGPU_GEM_DOMAIN_VRAM) {
 			r = -EINVAL;
-			amdgpu_bo_unreserve(robj);
-			break;
+			goto out_exec;
 		}
 		if (amdgpu_ttm_tt_get_usermm(robj->tbo.ttm)) {
 			r = -EPERM;
-			amdgpu_bo_unreserve(robj);
-			break;
+			goto out_exec;
 		}
 		for (base = robj->vm_bo; base; base = base->next)
 			if (amdgpu_xgmi_same_hive(amdgpu_ttm_adev(robj->tbo.bdev),
 				amdgpu_ttm_adev(base->vm->root.bo->tbo.bdev))) {
 				r = -EINVAL;
-				amdgpu_bo_unreserve(robj);
-				goto out;
+				goto out_exec;
 			}
 
 
@@ -863,18 +1032,147 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 			robj->allowed_domains |= AMDGPU_GEM_DOMAIN_GTT;
 
 		if (robj->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID)
-			amdgpu_vm_bo_invalidate(adev, robj, true);
+			amdgpu_vm_bo_invalidate(robj, true);
+		drm_exec_fini(&exec);
+		break;
+	case AMDGPU_GEM_OP_GET_MAPPING_INFO: {
+		struct amdgpu_bo_va *bo_va = amdgpu_vm_bo_find(&fpriv->vm, robj);
+		struct drm_amdgpu_gem_vm_entry *vm_entries;
+		struct amdgpu_bo_va_mapping *mapping;
+		int num_mappings = 0;
+		/*
+		 * num_entries is set as an input to the size of the user-allocated array of
+		 * drm_amdgpu_gem_vm_entry stored at args->value.
+		 * num_entries is sent back as output as the number of mappings the bo has.
+		 * If that number is larger than the size of the array, the ioctl must
+		 * be retried.
+		 */
+		vm_entries = kvcalloc(args->num_entries, sizeof(*vm_entries), GFP_KERNEL);
+		if (!vm_entries)
+			return -ENOMEM;
+
+		amdgpu_vm_bo_va_for_each_valid_mapping(bo_va, mapping) {
+			if (num_mappings < args->num_entries) {
+				vm_entries[num_mappings].addr = mapping->start * AMDGPU_GPU_PAGE_SIZE;
+				vm_entries[num_mappings].size = (mapping->last - mapping->start + 1) * AMDGPU_GPU_PAGE_SIZE;
+				vm_entries[num_mappings].offset = mapping->offset;
+				vm_entries[num_mappings].flags = mapping->flags;
+			}
+			num_mappings += 1;
+		}
+
+		amdgpu_vm_bo_va_for_each_invalid_mapping(bo_va, mapping) {
+			if (num_mappings < args->num_entries) {
+				vm_entries[num_mappings].addr = mapping->start * AMDGPU_GPU_PAGE_SIZE;
+				vm_entries[num_mappings].size = (mapping->last - mapping->start + 1) * AMDGPU_GPU_PAGE_SIZE;
+				vm_entries[num_mappings].offset = mapping->offset;
+				vm_entries[num_mappings].flags = mapping->flags;
+			}
+			num_mappings += 1;
+		}
 
-		amdgpu_bo_unreserve(robj);
+		drm_exec_fini(&exec);
+
+		if (num_mappings > 0 && num_mappings <= args->num_entries)
+			if (copy_to_user(u64_to_user_ptr(args->value), vm_entries, num_mappings * sizeof(*vm_entries)))
+				r = -EFAULT;
+
+		args->num_entries = num_mappings;
+
+		kvfree(vm_entries);
 		break;
+	}
 	default:
-		amdgpu_bo_unreserve(robj);
+		drm_exec_fini(&exec);
 		r = -EINVAL;
 	}
 
-out:
 	drm_gem_object_put(gobj);
 	return r;
+out_exec:
+	drm_exec_fini(&exec);
+	drm_gem_object_put(gobj);
+	return r;
+}
+
+/**
+ * amdgpu_gem_list_handles_ioctl - get information about a process' buffer objects
+ *
+ * @dev: drm device pointer
+ * @data: drm_amdgpu_gem_list_handles
+ * @filp: drm file pointer
+ *
+ * num_entries is set as an input to the size of the entries array.
+ * num_entries is sent back as output as the number of bos in the process.
+ * If that number is larger than the size of the array, the ioctl must
+ * be retried.
+ *
+ * Returns:
+ * 0 for success, -errno for errors.
+ */
+int amdgpu_gem_list_handles_ioctl(struct drm_device *dev, void *data,
+				  struct drm_file *filp)
+{
+	struct drm_amdgpu_gem_list_handles *args = data;
+	struct drm_amdgpu_gem_list_handles_entry *bo_entries;
+	struct drm_gem_object *gobj;
+	int id, ret = 0;
+	int bo_index = 0;
+	int num_bos = 0;
+
+	spin_lock(&filp->table_lock);
+	idr_for_each_entry(&filp->object_idr, gobj, id)
+		num_bos += 1;
+	spin_unlock(&filp->table_lock);
+
+	if (args->num_entries < num_bos) {
+		args->num_entries = num_bos;
+		return 0;
+	}
+
+	if (num_bos == 0) {
+		args->num_entries = 0;
+		return 0;
+	}
+
+	bo_entries = kvcalloc(num_bos, sizeof(*bo_entries), GFP_KERNEL);
+	if (!bo_entries)
+		return -ENOMEM;
+
+	spin_lock(&filp->table_lock);
+	idr_for_each_entry(&filp->object_idr, gobj, id) {
+		struct amdgpu_bo *bo = gem_to_amdgpu_bo(gobj);
+		struct drm_amdgpu_gem_list_handles_entry *bo_entry;
+
+		if (bo_index >= num_bos) {
+			ret = -EAGAIN;
+			break;
+		}
+
+		bo_entry = &bo_entries[bo_index];
+
+		bo_entry->size = amdgpu_bo_size(bo);
+		bo_entry->alloc_flags = bo->flags & AMDGPU_GEM_CREATE_SETTABLE_MASK;
+		bo_entry->preferred_domains = bo->preferred_domains;
+		bo_entry->gem_handle = id;
+		bo_entry->alignment = bo->tbo.page_alignment;
+
+		if (bo->tbo.base.import_attach)
+			bo_entry->flags |= AMDGPU_GEM_LIST_HANDLES_FLAG_IS_IMPORT;
+
+		bo_index += 1;
+	}
+	spin_unlock(&filp->table_lock);
+
+	args->num_entries = bo_index;
+
+	if (!ret)
+		if (copy_to_user(u64_to_user_ptr(args->entries), bo_entries, num_bos * sizeof(*bo_entries)))
+			ret = -EFAULT;
+
+	kvfree(bo_entries);
+
+	return ret;
 }
 
 static int amdgpu_gem_align_pitch(struct amdgpu_device *adev,
@@ -908,6 +1206,7 @@ int amdgpu_mode_dumb_create(struct drm_file *file_priv,
 			    struct drm_mode_create_dumb *args)
 {
 	struct amdgpu_device *adev = drm_to_adev(dev);
+	struct amdgpu_fpriv *fpriv = file_priv->driver_priv;
 	struct drm_gem_object *gobj;
 	uint32_t handle;
 	u64 flags = AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
@@ -931,16 +1230,16 @@ int amdgpu_mode_dumb_create(struct drm_file *file_priv,
 	domain = amdgpu_bo_get_preferred_domain(adev,
 				amdgpu_display_supported_domains(adev, flags));
 	r = amdgpu_gem_object_create(adev, args->size, 0, domain, flags,
-				     ttm_bo_type_device, NULL, &gobj);
+				     ttm_bo_type_device, NULL, &gobj, fpriv->xcp_id + 1);
 	if (r)
 		return -ENOMEM;
 
 	r = drm_gem_handle_create(file_priv, gobj, &handle);
 	/* drop reference from allocate - handle holds it now */
 	drm_gem_object_put(gobj);
-	if (r) {
+	if (r)
 		return r;
-	}
+
 	args->handle = handle;
 	return 0;
 }
@@ -948,7 +1247,7 @@ int amdgpu_mode_dumb_create(struct drm_file *file_priv,
 #if defined(CONFIG_DEBUG_FS)
 static int amdgpu_debugfs_gem_info_show(struct seq_file *m, void *unused)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct amdgpu_device *adev = m->private;
 	struct drm_device *dev = adev_to_drm(adev);
 	struct drm_file *file;
 	int r;
@@ -960,6 +1259,7 @@ static int amdgpu_debugfs_gem_info_show(struct seq_file *m, void *unused)
 	list_for_each_entry(file, &dev->filelist, lhead) {
 		struct task_struct *task;
 		struct drm_gem_object *gobj;
+		struct pid *pid;
 		int id;
 
 		/*
@@ -969,8 +1269,9 @@ static int amdgpu_debugfs_gem_info_show(struct seq_file *m, void *unused)
 		 * Therefore, we need to protect this ->comm access using RCU.
 		 */
 		rcu_read_lock();
-		task = pid_task(file->pid, PIDTYPE_TGID);
-		seq_printf(m, "pid %8d command %s:\n", pid_nr(file->pid),
+		pid = rcu_dereference(file->pid);
+		task = pid_task(pid, PIDTYPE_TGID);
+		seq_printf(m, "pid %8d command %s:\n", pid_nr(pid),
 			   task ? task->comm : "<unknown>");
 		rcu_read_unlock();
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h
index 637bf51dbf06..b558336bc4c6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h
@@ -33,6 +33,8 @@
 #define AMDGPU_GEM_DOMAIN_MAX		0x3
 #define gem_to_amdgpu_bo(gobj) container_of((gobj), struct amdgpu_bo, tbo.base)
 
+extern const struct drm_gem_object_funcs amdgpu_gem_object_funcs;
+
 unsigned long amdgpu_gem_timeout(uint64_t timeout_ns);
 
 /*
@@ -43,8 +45,7 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
 			     int alignment, u32 initial_domain,
 			     u64 flags, enum ttm_bo_type type,
 			     struct dma_resv *resv,
-			     struct drm_gem_object **obj);
-
+			     struct drm_gem_object **obj, int8_t xcp_id_plus1);
 int amdgpu_mode_dumb_create(struct drm_file *file_priv,
 			    struct drm_device *dev,
 			    struct drm_mode_create_dumb *args);
@@ -62,13 +63,28 @@ int amdgpu_gem_mmap_ioctl(struct drm_device *dev, void *data,
 			  struct drm_file *filp);
 int amdgpu_gem_wait_idle_ioctl(struct drm_device *dev, void *data,
 			      struct drm_file *filp);
-uint64_t amdgpu_gem_va_map_flags(struct amdgpu_device *adev, uint32_t flags);
 int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 			  struct drm_file *filp);
 int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *filp);
+int amdgpu_gem_list_handles_ioctl(struct drm_device *dev, void *data,
+				  struct drm_file *filp);
 
 int amdgpu_gem_metadata_ioctl(struct drm_device *dev, void *data,
 				struct drm_file *filp);
 
+#define AMDGPU_GEM_CREATE_SETTABLE_MASK	(AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | \
+	AMDGPU_GEM_CREATE_NO_CPU_ACCESS | \
+	AMDGPU_GEM_CREATE_CPU_GTT_USWC | \
+	AMDGPU_GEM_CREATE_VRAM_CLEARED | \
+	AMDGPU_GEM_CREATE_VM_ALWAYS_VALID | \
+	AMDGPU_GEM_CREATE_EXPLICIT_SYNC | \
+	AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE | \
+	AMDGPU_GEM_CREATE_ENCRYPTED | \
+	AMDGPU_GEM_CREATE_GFX12_DCC | \
+	AMDGPU_GEM_CREATE_DISCARDABLE | \
+	AMDGPU_GEM_CREATE_COHERENT | \
+	AMDGPU_GEM_CREATE_UNCACHED | \
+	AMDGPU_GEM_CREATE_EXT_COHERENT)
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
index 9d3a0542c996..8b118c53f351 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
@@ -24,10 +24,17 @@
  */
 
 #include <linux/firmware.h>
+#include <linux/pm_runtime.h>
+
 #include "amdgpu.h"
 #include "amdgpu_gfx.h"
 #include "amdgpu_rlc.h"
 #include "amdgpu_ras.h"
+#include "amdgpu_reset.h"
+#include "amdgpu_xcp.h"
+#include "amdgpu_xgmi.h"
+#include "amdgpu_mes.h"
+#include "nvd.h"
 
 /* delay 0.1 second to enable gfx off feature */
 #define GFX_OFF_DELAY_ENABLE         msecs_to_jiffies(100)
@@ -63,35 +70,26 @@ void amdgpu_queue_mask_bit_to_mec_queue(struct amdgpu_device *adev, int bit,
 }
 
 bool amdgpu_gfx_is_mec_queue_enabled(struct amdgpu_device *adev,
-				     int mec, int pipe, int queue)
+				     int xcc_id, int mec, int pipe, int queue)
 {
 	return test_bit(amdgpu_gfx_mec_queue_to_bit(adev, mec, pipe, queue),
-			adev->gfx.mec.queue_bitmap);
+			adev->gfx.mec_bitmap[xcc_id].queue_bitmap);
 }
 
-int amdgpu_gfx_me_queue_to_bit(struct amdgpu_device *adev,
-			       int me, int pipe, int queue)
+static int amdgpu_gfx_me_queue_to_bit(struct amdgpu_device *adev,
+				      int me, int pipe, int queue)
 {
+	int num_queue_per_pipe = 1; /* we only enable 1 KGQ per pipe */
 	int bit = 0;
 
 	bit += me * adev->gfx.me.num_pipe_per_me
-		* adev->gfx.me.num_queue_per_pipe;
-	bit += pipe * adev->gfx.me.num_queue_per_pipe;
+		* num_queue_per_pipe;
+	bit += pipe * num_queue_per_pipe;
 	bit += queue;
 
 	return bit;
 }
 
-void amdgpu_gfx_bit_to_me_queue(struct amdgpu_device *adev, int bit,
-				int *me, int *pipe, int *queue)
-{
-	*queue = bit % adev->gfx.me.num_queue_per_pipe;
-	*pipe = (bit / adev->gfx.me.num_queue_per_pipe)
-		% adev->gfx.me.num_pipe_per_me;
-	*me = (bit / adev->gfx.me.num_queue_per_pipe)
-		/ adev->gfx.me.num_pipe_per_me;
-}
-
 bool amdgpu_gfx_is_me_queue_enabled(struct amdgpu_device *adev,
 				    int me, int pipe, int queue)
 {
@@ -109,9 +107,9 @@ bool amdgpu_gfx_is_me_queue_enabled(struct amdgpu_device *adev,
  * The bitmask of CUs to be disabled in the shader array determined by se and
  * sh is stored in mask[se * max_sh + sh].
  */
-void amdgpu_gfx_parse_disable_cu(unsigned *mask, unsigned max_se, unsigned max_sh)
+void amdgpu_gfx_parse_disable_cu(unsigned int *mask, unsigned int max_se, unsigned int max_sh)
 {
-	unsigned se, sh, cu;
+	unsigned int se, sh, cu;
 	const char *p;
 
 	memset(mask, 0, sizeof(*mask) * max_se * max_sh);
@@ -123,6 +121,7 @@ void amdgpu_gfx_parse_disable_cu(unsigned *mask, unsigned max_se, unsigned max_s
 	for (;;) {
 		char *next;
 		int ret = sscanf(p, "%u.%u.%u", &se, &sh, &cu);
+
 		if (ret < 3) {
 			DRM_ERROR("amdgpu: could not parse disable_cu\n");
 			return;
@@ -151,12 +150,12 @@ static bool amdgpu_gfx_is_graphics_multipipe_capable(struct amdgpu_device *adev)
 static bool amdgpu_gfx_is_compute_multipipe_capable(struct amdgpu_device *adev)
 {
 	if (amdgpu_compute_multipipe != -1) {
-		DRM_INFO("amdgpu: forcing compute pipe policy %d\n",
+		dev_info(adev->dev, "amdgpu: forcing compute pipe policy %d\n",
 			 amdgpu_compute_multipipe);
 		return amdgpu_compute_multipipe == 1;
 	}
 
-	if (adev->ip_versions[GC_HWIP][0] > IP_VERSION(9, 0, 0))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) > IP_VERSION(9, 0, 0))
 		return true;
 
 	/* FIXME: spreading the queues across pipes causes perf regressions
@@ -204,37 +203,46 @@ bool amdgpu_gfx_is_high_priority_compute_queue(struct amdgpu_device *adev,
 
 void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev)
 {
-	int i, queue, pipe;
+	int i, j, queue, pipe;
 	bool multipipe_policy = amdgpu_gfx_is_compute_multipipe_capable(adev);
 	int max_queues_per_mec = min(adev->gfx.mec.num_pipe_per_mec *
 				     adev->gfx.mec.num_queue_per_pipe,
 				     adev->gfx.num_compute_rings);
+	int num_xcc = adev->gfx.xcc_mask ? NUM_XCC(adev->gfx.xcc_mask) : 1;
 
 	if (multipipe_policy) {
-		/* policy: make queues evenly cross all pipes on MEC1 only */
-		for (i = 0; i < max_queues_per_mec; i++) {
-			pipe = i % adev->gfx.mec.num_pipe_per_mec;
-			queue = (i / adev->gfx.mec.num_pipe_per_mec) %
-				adev->gfx.mec.num_queue_per_pipe;
-
-			set_bit(pipe * adev->gfx.mec.num_queue_per_pipe + queue,
-					adev->gfx.mec.queue_bitmap);
+		/* policy: make queues evenly cross all pipes on MEC1 only
+		 * for multiple xcc, just use the original policy for simplicity */
+		for (j = 0; j < num_xcc; j++) {
+			for (i = 0; i < max_queues_per_mec; i++) {
+				pipe = i % adev->gfx.mec.num_pipe_per_mec;
+				queue = (i / adev->gfx.mec.num_pipe_per_mec) %
+					 adev->gfx.mec.num_queue_per_pipe;
+
+				set_bit(pipe * adev->gfx.mec.num_queue_per_pipe + queue,
+					adev->gfx.mec_bitmap[j].queue_bitmap);
+			}
 		}
 	} else {
 		/* policy: amdgpu owns all queues in the given pipe */
-		for (i = 0; i < max_queues_per_mec; ++i)
-			set_bit(i, adev->gfx.mec.queue_bitmap);
+		for (j = 0; j < num_xcc; j++) {
+			for (i = 0; i < max_queues_per_mec; ++i)
+				set_bit(i, adev->gfx.mec_bitmap[j].queue_bitmap);
+		}
 	}
 
-	dev_dbg(adev->dev, "mec queue bitmap weight=%d\n", bitmap_weight(adev->gfx.mec.queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES));
+	for (j = 0; j < num_xcc; j++) {
+		dev_dbg(adev->dev, "mec queue bitmap weight=%d\n",
+			bitmap_weight(adev->gfx.mec_bitmap[j].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES));
+	}
 }
 
 void amdgpu_gfx_graphics_queue_acquire(struct amdgpu_device *adev)
 {
 	int i, queue, pipe;
 	bool multipipe_policy = amdgpu_gfx_is_graphics_multipipe_capable(adev);
-	int max_queues_per_me = adev->gfx.me.num_pipe_per_me *
-					adev->gfx.me.num_queue_per_pipe;
+	int num_queue_per_pipe = 1; /* we only enable 1 KGQ per pipe */
+	int max_queues_per_me = adev->gfx.me.num_pipe_per_me * num_queue_per_pipe;
 
 	if (multipipe_policy) {
 		/* policy: amdgpu owns the first queue per pipe at this stage
@@ -242,9 +250,9 @@ void amdgpu_gfx_graphics_queue_acquire(struct amdgpu_device *adev)
 		for (i = 0; i < max_queues_per_me; i++) {
 			pipe = i % adev->gfx.me.num_pipe_per_me;
 			queue = (i / adev->gfx.me.num_pipe_per_me) %
-				adev->gfx.me.num_queue_per_pipe;
+				num_queue_per_pipe;
 
-			set_bit(pipe * adev->gfx.me.num_queue_per_pipe + queue,
+			set_bit(pipe * num_queue_per_pipe + queue,
 				adev->gfx.me.queue_bitmap);
 		}
 	} else {
@@ -253,12 +261,13 @@ void amdgpu_gfx_graphics_queue_acquire(struct amdgpu_device *adev)
 	}
 
 	/* update the number of active graphics rings */
-	adev->gfx.num_gfx_rings =
-		bitmap_weight(adev->gfx.me.queue_bitmap, AMDGPU_MAX_GFX_QUEUES);
+	if (adev->gfx.num_gfx_rings)
+		adev->gfx.num_gfx_rings =
+			bitmap_weight(adev->gfx.me.queue_bitmap, AMDGPU_MAX_GFX_QUEUES);
 }
 
 static int amdgpu_gfx_kiq_acquire(struct amdgpu_device *adev,
-				  struct amdgpu_ring *ring)
+				  struct amdgpu_ring *ring, int xcc_id)
 {
 	int queue_bit;
 	int mec, pipe, queue;
@@ -268,7 +277,7 @@ static int amdgpu_gfx_kiq_acquire(struct amdgpu_device *adev,
 		    * adev->gfx.mec.num_queue_per_pipe;
 
 	while (--queue_bit >= 0) {
-		if (test_bit(queue_bit, adev->gfx.mec.queue_bitmap))
+		if (test_bit(queue_bit, adev->gfx.mec_bitmap[xcc_id].queue_bitmap))
 			continue;
 
 		amdgpu_queue_mask_bit_to_mec_queue(adev, queue_bit, &mec, &pipe, &queue);
@@ -292,11 +301,11 @@ static int amdgpu_gfx_kiq_acquire(struct amdgpu_device *adev,
 	return -EINVAL;
 }
 
-int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,
-			     struct amdgpu_ring *ring,
-			     struct amdgpu_irq_src *irq)
+int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, int xcc_id)
 {
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
+	struct amdgpu_irq_src *irq = &kiq->irq;
+	struct amdgpu_ring *ring = &kiq->ring;
 	int r = 0;
 
 	spin_lock_init(&kiq->ring_lock);
@@ -304,16 +313,22 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,
 	ring->adev = NULL;
 	ring->ring_obj = NULL;
 	ring->use_doorbell = true;
-	ring->doorbell_index = adev->doorbell_index.kiq;
-	ring->vm_hub = AMDGPU_GFXHUB_0;
-
-	r = amdgpu_gfx_kiq_acquire(adev, ring);
+	ring->xcc_id = xcc_id;
+	ring->vm_hub = AMDGPU_GFXHUB(xcc_id);
+	ring->doorbell_index =
+		(adev->doorbell_index.kiq +
+		 xcc_id * adev->doorbell_index.xcc_doorbell_range)
+		<< 1;
+
+	r = amdgpu_gfx_kiq_acquire(adev, ring, xcc_id);
 	if (r)
 		return r;
 
 	ring->eop_gpu_addr = kiq->eop_gpu_addr;
 	ring->no_scheduler = true;
-	sprintf(ring->name, "kiq_%d.%d.%d", ring->me, ring->pipe, ring->queue);
+	snprintf(ring->name, sizeof(ring->name), "kiq_%hhu.%hhu.%hhu.%hhu",
+		 (unsigned char)xcc_id, (unsigned char)ring->me,
+		 (unsigned char)ring->pipe, (unsigned char)ring->queue);
 	r = amdgpu_ring_init(adev, ring, 1024, irq, AMDGPU_CP_KIQ_IRQ_DRIVER0,
 			     AMDGPU_RING_PRIO_DEFAULT, NULL);
 	if (r)
@@ -327,19 +342,19 @@ void amdgpu_gfx_kiq_free_ring(struct amdgpu_ring *ring)
 	amdgpu_ring_fini(ring);
 }
 
-void amdgpu_gfx_kiq_fini(struct amdgpu_device *adev)
+void amdgpu_gfx_kiq_fini(struct amdgpu_device *adev, int xcc_id)
 {
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
 
 	amdgpu_bo_free_kernel(&kiq->eop_obj, &kiq->eop_gpu_addr, NULL);
 }
 
 int amdgpu_gfx_kiq_init(struct amdgpu_device *adev,
-			unsigned hpd_size)
+			unsigned int hpd_size, int xcc_id)
 {
 	int r;
 	u32 *hpd;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
 
 	r = amdgpu_bo_create_kernel(adev, hpd_size, PAGE_SIZE,
 				    AMDGPU_GEM_DOMAIN_GTT, &kiq->eop_obj,
@@ -362,13 +377,20 @@ int amdgpu_gfx_kiq_init(struct amdgpu_device *adev,
 
 /* create MQD for each compute/gfx queue */
 int amdgpu_gfx_mqd_sw_init(struct amdgpu_device *adev,
-			   unsigned mqd_size)
+			   unsigned int mqd_size, int xcc_id)
 {
-	struct amdgpu_ring *ring = NULL;
-	int r, i;
+	int r, i, j;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
+	struct amdgpu_ring *ring = &kiq->ring;
+	u32 domain = AMDGPU_GEM_DOMAIN_GTT;
+
+#if !defined(CONFIG_ARM) && !defined(CONFIG_ARM64)
+	/* Only enable on gfx10 and 11 for now to avoid changing behavior on older chips */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 0, 0))
+		domain |= AMDGPU_GEM_DOMAIN_VRAM;
+#endif
 
 	/* create MQD for KIQ */
-	ring = &adev->gfx.kiq.ring;
 	if (!adev->enable_mes_kiq && !ring->mqd_obj) {
 		/* originaly the KIQ MQD is put in GTT domain, but for SRIOV VRAM domain is a must
 		 * otherwise hypervisor trigger SAVE_VF fail after driver unloaded which mean MQD
@@ -387,9 +409,12 @@ int amdgpu_gfx_mqd_sw_init(struct amdgpu_device *adev,
 		}
 
 		/* prepare MQD backup */
-		adev->gfx.mec.mqd_backup[AMDGPU_MAX_COMPUTE_RINGS] = kmalloc(mqd_size, GFP_KERNEL);
-		if (!adev->gfx.mec.mqd_backup[AMDGPU_MAX_COMPUTE_RINGS])
-				dev_warn(adev->dev, "no memory to create MQD backup for ring %s\n", ring->name);
+		kiq->mqd_backup = kzalloc(mqd_size, GFP_KERNEL);
+		if (!kiq->mqd_backup) {
+			dev_warn(adev->dev,
+				 "no memory to create MQD backup for ring %s\n", ring->name);
+			return -ENOMEM;
+		}
 	}
 
 	if (adev->asic_type >= CHIP_NAVI10 && amdgpu_async_gfx_ring) {
@@ -398,47 +423,55 @@ int amdgpu_gfx_mqd_sw_init(struct amdgpu_device *adev,
 			ring = &adev->gfx.gfx_ring[i];
 			if (!ring->mqd_obj) {
 				r = amdgpu_bo_create_kernel(adev, mqd_size, PAGE_SIZE,
-							    AMDGPU_GEM_DOMAIN_GTT, &ring->mqd_obj,
+							    domain, &ring->mqd_obj,
 							    &ring->mqd_gpu_addr, &ring->mqd_ptr);
 				if (r) {
 					dev_warn(adev->dev, "failed to create ring mqd bo (%d)", r);
 					return r;
 				}
 
+				ring->mqd_size = mqd_size;
 				/* prepare MQD backup */
-				adev->gfx.me.mqd_backup[i] = kmalloc(mqd_size, GFP_KERNEL);
-				if (!adev->gfx.me.mqd_backup[i])
+				adev->gfx.me.mqd_backup[i] = kzalloc(mqd_size, GFP_KERNEL);
+				if (!adev->gfx.me.mqd_backup[i]) {
 					dev_warn(adev->dev, "no memory to create MQD backup for ring %s\n", ring->name);
+					return -ENOMEM;
+				}
 			}
 		}
 	}
 
 	/* create MQD for each KCQ */
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-		ring = &adev->gfx.compute_ring[i];
+		j = i + xcc_id * adev->gfx.num_compute_rings;
+		ring = &adev->gfx.compute_ring[j];
 		if (!ring->mqd_obj) {
 			r = amdgpu_bo_create_kernel(adev, mqd_size, PAGE_SIZE,
-						    AMDGPU_GEM_DOMAIN_GTT, &ring->mqd_obj,
+						    domain, &ring->mqd_obj,
 						    &ring->mqd_gpu_addr, &ring->mqd_ptr);
 			if (r) {
 				dev_warn(adev->dev, "failed to create ring mqd bo (%d)", r);
 				return r;
 			}
 
+			ring->mqd_size = mqd_size;
 			/* prepare MQD backup */
-			adev->gfx.mec.mqd_backup[i] = kmalloc(mqd_size, GFP_KERNEL);
-			if (!adev->gfx.mec.mqd_backup[i])
+			adev->gfx.mec.mqd_backup[j] = kzalloc(mqd_size, GFP_KERNEL);
+			if (!adev->gfx.mec.mqd_backup[j]) {
 				dev_warn(adev->dev, "no memory to create MQD backup for ring %s\n", ring->name);
+				return -ENOMEM;
+			}
 		}
 	}
 
 	return 0;
 }
 
-void amdgpu_gfx_mqd_sw_fini(struct amdgpu_device *adev)
+void amdgpu_gfx_mqd_sw_fini(struct amdgpu_device *adev, int xcc_id)
 {
 	struct amdgpu_ring *ring = NULL;
-	int i;
+	int i, j;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
 
 	if (adev->asic_type >= CHIP_NAVI10 && amdgpu_async_gfx_ring) {
 		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
@@ -451,43 +484,121 @@ void amdgpu_gfx_mqd_sw_fini(struct amdgpu_device *adev)
 	}
 
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-		ring = &adev->gfx.compute_ring[i];
-		kfree(adev->gfx.mec.mqd_backup[i]);
+		j = i + xcc_id * adev->gfx.num_compute_rings;
+		ring = &adev->gfx.compute_ring[j];
+		kfree(adev->gfx.mec.mqd_backup[j]);
 		amdgpu_bo_free_kernel(&ring->mqd_obj,
 				      &ring->mqd_gpu_addr,
 				      &ring->mqd_ptr);
 	}
 
-	ring = &adev->gfx.kiq.ring;
-	kfree(adev->gfx.mec.mqd_backup[AMDGPU_MAX_COMPUTE_RINGS]);
+	ring = &kiq->ring;
+	kfree(kiq->mqd_backup);
 	amdgpu_bo_free_kernel(&ring->mqd_obj,
 			      &ring->mqd_gpu_addr,
 			      &ring->mqd_ptr);
 }
 
-int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev)
+int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev, int xcc_id)
 {
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
 	struct amdgpu_ring *kiq_ring = &kiq->ring;
 	int i, r = 0;
+	int j;
+
+	if (adev->enable_mes) {
+		for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+			j = i + xcc_id * adev->gfx.num_compute_rings;
+			amdgpu_mes_unmap_legacy_queue(adev,
+						   &adev->gfx.compute_ring[j],
+						   RESET_QUEUES, 0, 0);
+		}
+		return 0;
+	}
 
 	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
 		return -EINVAL;
 
-	spin_lock(&adev->gfx.kiq.ring_lock);
+	if (!kiq_ring->sched.ready || amdgpu_in_reset(adev))
+		return 0;
+
+	spin_lock(&kiq->ring_lock);
 	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size *
 					adev->gfx.num_compute_rings)) {
-		spin_unlock(&adev->gfx.kiq.ring_lock);
+		spin_unlock(&kiq->ring_lock);
 		return -ENOMEM;
 	}
 
-	for (i = 0; i < adev->gfx.num_compute_rings; i++)
-		kiq->pmf->kiq_unmap_queues(kiq_ring, &adev->gfx.compute_ring[i],
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		j = i + xcc_id * adev->gfx.num_compute_rings;
+		kiq->pmf->kiq_unmap_queues(kiq_ring,
+					   &adev->gfx.compute_ring[j],
 					   RESET_QUEUES, 0, 0);
+	}
+	/* Submit unmap queue packet */
+	amdgpu_ring_commit(kiq_ring);
+	/*
+	 * Ring test will do a basic scratch register change check. Just run
+	 * this to ensure that unmap queues that is submitted before got
+	 * processed successfully before returning.
+	 */
+	r = amdgpu_ring_test_helper(kiq_ring);
+
+	spin_unlock(&kiq->ring_lock);
+
+	return r;
+}
+
+int amdgpu_gfx_disable_kgq(struct amdgpu_device *adev, int xcc_id)
+{
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
+	int i, r = 0;
+	int j;
+
+	if (adev->enable_mes) {
+		if (amdgpu_gfx_is_master_xcc(adev, xcc_id)) {
+			for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+				j = i + xcc_id * adev->gfx.num_gfx_rings;
+				amdgpu_mes_unmap_legacy_queue(adev,
+						      &adev->gfx.gfx_ring[j],
+						      PREEMPT_QUEUES, 0, 0);
+			}
+		}
+		return 0;
+	}
+
+	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
+		return -EINVAL;
+
+	if (!adev->gfx.kiq[0].ring.sched.ready || amdgpu_in_reset(adev))
+		return 0;
+
+	if (amdgpu_gfx_is_master_xcc(adev, xcc_id)) {
+		spin_lock(&kiq->ring_lock);
+		if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size *
+						adev->gfx.num_gfx_rings)) {
+			spin_unlock(&kiq->ring_lock);
+			return -ENOMEM;
+		}
+
+		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+			j = i + xcc_id * adev->gfx.num_gfx_rings;
+			kiq->pmf->kiq_unmap_queues(kiq_ring,
+						   &adev->gfx.gfx_ring[j],
+						   PREEMPT_QUEUES, 0, 0);
+		}
+		/* Submit unmap queue packet */
+		amdgpu_ring_commit(kiq_ring);
 
-	if (adev->gfx.kiq.ring.sched.ready && !adev->job_hang)
+		/*
+		 * Ring test will do a basic scratch register change check.
+		 * Just run this to ensure that unmap queues that is submitted
+		 * before got processed successfully before returning.
+		 */
 		r = amdgpu_ring_test_helper(kiq_ring);
-	spin_unlock(&adev->gfx.kiq.ring_lock);
+		spin_unlock(&kiq->ring_lock);
+	}
 
 	return r;
 }
@@ -505,70 +616,167 @@ int amdgpu_queue_mask_bit_to_set_resource_bit(struct amdgpu_device *adev,
 	return set_resource_bit;
 }
 
-int amdgpu_gfx_enable_kcq(struct amdgpu_device *adev)
+static int amdgpu_gfx_mes_enable_kcq(struct amdgpu_device *adev, int xcc_id)
 {
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
+	uint64_t queue_mask = ~0ULL;
+	int r, i, j;
+
+	amdgpu_device_flush_hdp(adev, NULL);
+
+	if (!adev->enable_uni_mes) {
+		spin_lock(&kiq->ring_lock);
+		r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->set_resources_size);
+		if (r) {
+			dev_err(adev->dev, "Failed to lock KIQ (%d).\n", r);
+			spin_unlock(&kiq->ring_lock);
+			return r;
+		}
+
+		kiq->pmf->kiq_set_resources(kiq_ring, queue_mask);
+		r = amdgpu_ring_test_helper(kiq_ring);
+		spin_unlock(&kiq->ring_lock);
+		if (r)
+			dev_err(adev->dev, "KIQ failed to set resources\n");
+	}
+
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		j = i + xcc_id * adev->gfx.num_compute_rings;
+		r = amdgpu_mes_map_legacy_queue(adev,
+						&adev->gfx.compute_ring[j]);
+		if (r) {
+			dev_err(adev->dev, "failed to map compute queue\n");
+			return r;
+		}
+	}
+
+	return 0;
+}
+
+int amdgpu_gfx_enable_kcq(struct amdgpu_device *adev, int xcc_id)
+{
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
 	uint64_t queue_mask = 0;
-	int r, i;
+	int r, i, j;
+
+	if (adev->mes.enable_legacy_queue_map)
+		return amdgpu_gfx_mes_enable_kcq(adev, xcc_id);
 
 	if (!kiq->pmf || !kiq->pmf->kiq_map_queues || !kiq->pmf->kiq_set_resources)
 		return -EINVAL;
 
 	for (i = 0; i < AMDGPU_MAX_COMPUTE_QUEUES; ++i) {
-		if (!test_bit(i, adev->gfx.mec.queue_bitmap))
+		if (!test_bit(i, adev->gfx.mec_bitmap[xcc_id].queue_bitmap))
 			continue;
 
 		/* This situation may be hit in the future if a new HW
 		 * generation exposes more than 64 queues. If so, the
 		 * definition of queue_mask needs updating */
 		if (WARN_ON(i > (sizeof(queue_mask)*8))) {
-			DRM_ERROR("Invalid KCQ enabled: %d\n", i);
+			dev_err(adev->dev, "Invalid KCQ enabled: %d\n", i);
 			break;
 		}
 
 		queue_mask |= (1ull << amdgpu_queue_mask_bit_to_set_resource_bit(adev, i));
 	}
 
-	DRM_INFO("kiq ring mec %d pipe %d q %d\n", kiq_ring->me, kiq_ring->pipe,
-							kiq_ring->queue);
-	spin_lock(&adev->gfx.kiq.ring_lock);
+	amdgpu_device_flush_hdp(adev, NULL);
+
+	dev_info(adev->dev, "kiq ring mec %d pipe %d q %d\n", kiq_ring->me,
+		 kiq_ring->pipe, kiq_ring->queue);
+
+	spin_lock(&kiq->ring_lock);
 	r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size *
 					adev->gfx.num_compute_rings +
 					kiq->pmf->set_resources_size);
 	if (r) {
-		DRM_ERROR("Failed to lock KIQ (%d).\n", r);
-		spin_unlock(&adev->gfx.kiq.ring_lock);
+		dev_err(adev->dev, "Failed to lock KIQ (%d).\n", r);
+		spin_unlock(&kiq->ring_lock);
 		return r;
 	}
 
-	if (adev->enable_mes)
-		queue_mask = ~0ULL;
-
 	kiq->pmf->kiq_set_resources(kiq_ring, queue_mask);
-	for (i = 0; i < adev->gfx.num_compute_rings; i++)
-		kiq->pmf->kiq_map_queues(kiq_ring, &adev->gfx.compute_ring[i]);
-
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		j = i + xcc_id * adev->gfx.num_compute_rings;
+		kiq->pmf->kiq_map_queues(kiq_ring,
+					 &adev->gfx.compute_ring[j]);
+	}
+	/* Submit map queue packet */
+	amdgpu_ring_commit(kiq_ring);
+	/*
+	 * Ring test will do a basic scratch register change check. Just run
+	 * this to ensure that map queues that is submitted before got
+	 * processed successfully before returning.
+	 */
 	r = amdgpu_ring_test_helper(kiq_ring);
-	spin_unlock(&adev->gfx.kiq.ring_lock);
+	spin_unlock(&kiq->ring_lock);
 	if (r)
-		DRM_ERROR("KCQ enable failed\n");
+		dev_err(adev->dev, "KCQ enable failed\n");
 
 	return r;
 }
 
-/* amdgpu_gfx_off_ctrl - Handle gfx off feature enable/disable
- *
- * @adev: amdgpu_device pointer
- * @bool enable true: enable gfx off feature, false: disable gfx off feature
- *
- * 1. gfx off feature will be enabled by gfx ip after gfx cg gp enabled.
- * 2. other client can send request to disable gfx off feature, the request should be honored.
- * 3. other client can cancel their request of disable gfx off feature
- * 4. other client should not send request to enable gfx off feature before disable gfx off feature.
- */
+int amdgpu_gfx_enable_kgq(struct amdgpu_device *adev, int xcc_id)
+{
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
+	int r, i, j;
 
-void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable)
+	if (!kiq->pmf || !kiq->pmf->kiq_map_queues)
+		return -EINVAL;
+
+	amdgpu_device_flush_hdp(adev, NULL);
+
+	if (adev->mes.enable_legacy_queue_map) {
+		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+			j = i + xcc_id * adev->gfx.num_gfx_rings;
+			r = amdgpu_mes_map_legacy_queue(adev,
+							&adev->gfx.gfx_ring[j]);
+			if (r) {
+				dev_err(adev->dev, "failed to map gfx queue\n");
+				return r;
+			}
+		}
+
+		return 0;
+	}
+
+	spin_lock(&kiq->ring_lock);
+	/* No need to map kcq on the slave */
+	if (amdgpu_gfx_is_master_xcc(adev, xcc_id)) {
+		r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size *
+						adev->gfx.num_gfx_rings);
+		if (r) {
+			dev_err(adev->dev, "Failed to lock KIQ (%d).\n", r);
+			spin_unlock(&kiq->ring_lock);
+			return r;
+		}
+
+		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+			j = i + xcc_id * adev->gfx.num_gfx_rings;
+			kiq->pmf->kiq_map_queues(kiq_ring,
+						 &adev->gfx.gfx_ring[j]);
+		}
+	}
+	/* Submit map queue packet */
+	amdgpu_ring_commit(kiq_ring);
+	/*
+	 * Ring test will do a basic scratch register change check. Just run
+	 * this to ensure that map queues that is submitted before got
+	 * processed successfully before returning.
+	 */
+	r = amdgpu_ring_test_helper(kiq_ring);
+	spin_unlock(&kiq->ring_lock);
+	if (r)
+		dev_err(adev->dev, "KGQ enable failed\n");
+
+	return r;
+}
+
+static void amdgpu_gfx_do_off_ctrl(struct amdgpu_device *adev, bool enable,
+				   bool no_delay)
 {
 	unsigned long delay = GFX_OFF_DELAY_ENABLE;
 
@@ -590,9 +798,9 @@ void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable)
 		if (adev->gfx.gfx_off_req_count == 0 &&
 		    !adev->gfx.gfx_off_state) {
 			/* If going to s2idle, no need to wait */
-			if (adev->in_s0ix) {
+			if (no_delay) {
 				if (!amdgpu_dpm_set_powergating_by_smu(adev,
-						AMD_IP_BLOCK_TYPE_GFX, true))
+						AMD_IP_BLOCK_TYPE_GFX, true, 0))
 					adev->gfx.gfx_off_state = true;
 			} else {
 				schedule_delayed_work(&adev->gfx.gfx_off_delay_work,
@@ -604,7 +812,7 @@ void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable)
 			cancel_delayed_work_sync(&adev->gfx.gfx_off_delay_work);
 
 			if (adev->gfx.gfx_off_state &&
-			    !amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, false)) {
+			    !amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, false, 0)) {
 				adev->gfx.gfx_off_state = false;
 
 				if (adev->gfx.funcs->init_spm_golden) {
@@ -622,6 +830,43 @@ unlock:
 	mutex_unlock(&adev->gfx.gfx_off_mutex);
 }
 
+/* amdgpu_gfx_off_ctrl - Handle gfx off feature enable/disable
+ *
+ * @adev: amdgpu_device pointer
+ * @bool enable true: enable gfx off feature, false: disable gfx off feature
+ *
+ * 1. gfx off feature will be enabled by gfx ip after gfx cg pg enabled.
+ * 2. other client can send request to disable gfx off feature, the request should be honored.
+ * 3. other client can cancel their request of disable gfx off feature
+ * 4. other client should not send request to enable gfx off feature before disable gfx off feature.
+ *
+ * gfx off allow will be delayed by GFX_OFF_DELAY_ENABLE ms.
+ */
+void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable)
+{
+	/* If going to s2idle, no need to wait */
+	bool no_delay = adev->in_s0ix ? true : false;
+
+	amdgpu_gfx_do_off_ctrl(adev, enable, no_delay);
+}
+
+/* amdgpu_gfx_off_ctrl_immediate - Handle gfx off feature enable/disable
+ *
+ * @adev: amdgpu_device pointer
+ * @bool enable true: enable gfx off feature, false: disable gfx off feature
+ *
+ * 1. gfx off feature will be enabled by gfx ip after gfx cg pg enabled.
+ * 2. other client can send request to disable gfx off feature, the request should be honored.
+ * 3. other client can cancel their request of disable gfx off feature
+ * 4. other client should not send request to enable gfx off feature before disable gfx off feature.
+ *
+ * gfx off allow will be issued immediately.
+ */
+void amdgpu_gfx_off_ctrl_immediate(struct amdgpu_device *adev, bool enable)
+{
+	amdgpu_gfx_do_off_ctrl(adev, enable, true);
+}
+
 int amdgpu_set_gfx_off_residency(struct amdgpu_device *adev, bool value)
 {
 	int r = 0;
@@ -680,16 +925,24 @@ int amdgpu_gfx_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *r
 	int r;
 
 	if (amdgpu_ras_is_supported(adev, ras_block->block)) {
-		if (!amdgpu_persistent_edc_harvesting_supported(adev))
-			amdgpu_ras_reset_error_status(adev, AMDGPU_RAS_BLOCK__GFX);
+		if (!amdgpu_persistent_edc_harvesting_supported(adev)) {
+			r = amdgpu_ras_reset_error_status(adev, AMDGPU_RAS_BLOCK__GFX);
+			if (r)
+				return r;
+		}
 
 		r = amdgpu_ras_block_late_init(adev, ras_block);
 		if (r)
 			return r;
 
-		r = amdgpu_irq_get(adev, &adev->gfx.cp_ecc_error_irq, 0);
-		if (r)
-			goto late_fini;
+		if (amdgpu_sriov_vf(adev))
+			return r;
+
+		if (adev->gfx.cp_ecc_error_irq.funcs) {
+			r = amdgpu_irq_get(adev, &adev->gfx.cp_ecc_error_irq, 0);
+			if (r)
+				goto late_fini;
+		}
 	} else {
 		amdgpu_ras_feature_enable_on_boot(adev, ras_block, 0);
 	}
@@ -778,23 +1031,42 @@ int amdgpu_gfx_cp_ecc_error_irq(struct amdgpu_device *adev,
 
 	ih_data.head = *ras_if;
 
-	DRM_ERROR("CP ECC ERROR IRQ\n");
+	dev_err(adev->dev, "CP ECC ERROR IRQ\n");
 	amdgpu_ras_interrupt_dispatch(adev, &ih_data);
 	return 0;
 }
 
-uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
+void amdgpu_gfx_ras_error_func(struct amdgpu_device *adev,
+		void *ras_error_status,
+		void (*func)(struct amdgpu_device *adev, void *ras_error_status,
+				int xcc_id))
+{
+	int i;
+	int num_xcc = adev->gfx.xcc_mask ? NUM_XCC(adev->gfx.xcc_mask) : 1;
+	uint32_t xcc_mask = GENMASK(num_xcc - 1, 0);
+	struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
+
+	if (err_data) {
+		err_data->ue_count = 0;
+		err_data->ce_count = 0;
+	}
+
+	for_each_inst(i, xcc_mask)
+		func(adev, ras_error_status, i);
+}
+
+uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg, uint32_t xcc_id)
 {
 	signed long r, cnt = 0;
 	unsigned long flags;
 	uint32_t seq, reg_val_offs = 0, value = 0;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
 	struct amdgpu_ring *ring = &kiq->ring;
 
 	if (amdgpu_device_skip_hw_access(adev))
 		return 0;
 
-	if (adev->mes.ring.sched.ready)
+	if (adev->mes.ring[0].sched.ready)
 		return amdgpu_mes_rreg(adev, reg);
 
 	BUG_ON(!ring->funcs->emit_rreg);
@@ -804,7 +1076,10 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
 		pr_err("critical bug! too many kiq readers\n");
 		goto failed_unlock;
 	}
-	amdgpu_ring_alloc(ring, 32);
+	r = amdgpu_ring_alloc(ring, 32);
+	if (r)
+		goto failed_unlock;
+
 	amdgpu_ring_emit_rreg(ring, reg, reg_val_offs);
 	r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
 	if (r)
@@ -828,6 +1103,9 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
 
 	might_sleep();
 	while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) {
+		if (amdgpu_in_reset(adev))
+			goto failed_kiq_read;
+
 		msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
 		r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
 	}
@@ -851,12 +1129,12 @@ failed_kiq_read:
 	return ~0;
 }
 
-void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v)
+void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v, uint32_t xcc_id)
 {
 	signed long r, cnt = 0;
 	unsigned long flags;
 	uint32_t seq;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
 	struct amdgpu_ring *ring = &kiq->ring;
 
 	BUG_ON(!ring->funcs->emit_wreg);
@@ -864,13 +1142,16 @@ void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v)
 	if (amdgpu_device_skip_hw_access(adev))
 		return;
 
-	if (adev->mes.ring.sched.ready) {
+	if (adev->mes.ring[0].sched.ready) {
 		amdgpu_mes_wreg(adev, reg, v);
 		return;
 	}
 
 	spin_lock_irqsave(&kiq->ring_lock, flags);
-	amdgpu_ring_alloc(ring, 32);
+	r = amdgpu_ring_alloc(ring, 32);
+	if (r)
+		goto failed_unlock;
+
 	amdgpu_ring_emit_wreg(ring, reg, v);
 	r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
 	if (r)
@@ -894,6 +1175,8 @@ void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v)
 
 	might_sleep();
 	while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) {
+		if (amdgpu_in_reset(adev))
+			goto failed_kiq_write;
 
 		msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
 		r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
@@ -906,11 +1189,81 @@ void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v)
 
 failed_undo:
 	amdgpu_ring_undo(ring);
+failed_unlock:
 	spin_unlock_irqrestore(&kiq->ring_lock, flags);
 failed_kiq_write:
 	dev_err(adev->dev, "failed to write reg:%x\n", reg);
 }
 
+int amdgpu_kiq_hdp_flush(struct amdgpu_device *adev)
+{
+	signed long r, cnt = 0;
+	unsigned long flags;
+	uint32_t seq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
+	struct amdgpu_ring *ring = &kiq->ring;
+
+	if (amdgpu_device_skip_hw_access(adev))
+		return 0;
+
+	if (adev->enable_mes_kiq && adev->mes.ring[0].sched.ready)
+		return amdgpu_mes_hdp_flush(adev);
+
+	if (!ring->funcs->emit_hdp_flush) {
+		return -EOPNOTSUPP;
+	}
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+	r = amdgpu_ring_alloc(ring, 32);
+	if (r)
+		goto failed_unlock;
+
+	amdgpu_ring_emit_hdp_flush(ring);
+	r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
+	if (r)
+		goto failed_undo;
+
+	amdgpu_ring_commit(ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+
+	r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
+
+	/* don't wait anymore for gpu reset case because this way may
+	 * block gpu_recover() routine forever, e.g. this virt_kiq_rreg
+	 * is triggered in TTM and ttm_bo_lock_delayed_workqueue() will
+	 * never return if we keep waiting in virt_kiq_rreg, which cause
+	 * gpu_recover() hang there.
+	 *
+	 * also don't wait anymore for IRQ context
+	 * */
+	if (r < 1 && (amdgpu_in_reset(adev) || in_interrupt()))
+		goto failed_kiq_hdp_flush;
+
+	might_sleep();
+	while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) {
+		if (amdgpu_in_reset(adev))
+			goto failed_kiq_hdp_flush;
+
+		msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
+		r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
+	}
+
+	if (cnt > MAX_KIQ_REG_TRY) {
+		dev_err(adev->dev, "failed to flush HDP via KIQ timeout\n");
+		return -ETIMEDOUT;
+	}
+
+	return 0;
+
+failed_undo:
+	amdgpu_ring_undo(ring);
+failed_unlock:
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+failed_kiq_hdp_flush:
+	dev_err(adev->dev, "failed to flush HDP via KIQ\n");
+	return r < 0 ? r : -EIO;
+}
+
 int amdgpu_gfx_get_num_kcq(struct amdgpu_device *adev)
 {
 	if (amdgpu_num_kcq == -1) {
@@ -1050,7 +1403,8 @@ void amdgpu_gfx_cp_init_microcode(struct amdgpu_device *adev,
 		fw_size = le32_to_cpu(cp_hdr_v2_0->data_size_bytes);
 		break;
 	default:
-		break;
+		dev_err(adev->dev, "Invalid ucode id %u\n", ucode_id);
+		return;
 	}
 
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
@@ -1060,3 +1414,1144 @@ void amdgpu_gfx_cp_init_microcode(struct amdgpu_device *adev,
 		adev->firmware.fw_size += ALIGN(fw_size, PAGE_SIZE);
 	}
 }
+
+bool amdgpu_gfx_is_master_xcc(struct amdgpu_device *adev, int xcc_id)
+{
+	return !(xcc_id % (adev->gfx.num_xcc_per_xcp ?
+			adev->gfx.num_xcc_per_xcp : 1));
+}
+
+static ssize_t amdgpu_gfx_get_current_compute_partition(struct device *dev,
+						struct device_attribute *addr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	int mode;
+
+	/* Only minimal precaution taken to reject requests while in reset.*/
+	if (amdgpu_in_reset(adev))
+		return -EPERM;
+
+	mode = amdgpu_xcp_query_partition_mode(adev->xcp_mgr,
+					       AMDGPU_XCP_FL_NONE);
+
+	return sysfs_emit(buf, "%s\n", amdgpu_gfx_compute_mode_desc(mode));
+}
+
+static ssize_t amdgpu_gfx_set_compute_partition(struct device *dev,
+						struct device_attribute *addr,
+						const char *buf, size_t count)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	enum amdgpu_gfx_partition mode;
+	int ret = 0, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	if (num_xcc % 2 != 0)
+		return -EINVAL;
+
+	if (!strncasecmp("SPX", buf, strlen("SPX"))) {
+		mode = AMDGPU_SPX_PARTITION_MODE;
+	} else if (!strncasecmp("DPX", buf, strlen("DPX"))) {
+		/*
+		 * DPX mode needs AIDs to be in multiple of 2.
+		 * Each AID connects 2 XCCs.
+		 */
+		if (num_xcc%4)
+			return -EINVAL;
+		mode = AMDGPU_DPX_PARTITION_MODE;
+	} else if (!strncasecmp("TPX", buf, strlen("TPX"))) {
+		if (num_xcc != 6)
+			return -EINVAL;
+		mode = AMDGPU_TPX_PARTITION_MODE;
+	} else if (!strncasecmp("QPX", buf, strlen("QPX"))) {
+		if (num_xcc != 8)
+			return -EINVAL;
+		mode = AMDGPU_QPX_PARTITION_MODE;
+	} else if (!strncasecmp("CPX", buf, strlen("CPX"))) {
+		mode = AMDGPU_CPX_PARTITION_MODE;
+	} else {
+		return -EINVAL;
+	}
+
+	/* Don't allow a switch while under reset */
+	if (!down_read_trylock(&adev->reset_domain->sem))
+		return -EPERM;
+
+	ret = amdgpu_xcp_switch_partition_mode(adev->xcp_mgr, mode);
+
+	up_read(&adev->reset_domain->sem);
+
+	if (ret)
+		return ret;
+
+	return count;
+}
+
+static const char *xcp_desc[] = {
+	[AMDGPU_SPX_PARTITION_MODE] = "SPX",
+	[AMDGPU_DPX_PARTITION_MODE] = "DPX",
+	[AMDGPU_TPX_PARTITION_MODE] = "TPX",
+	[AMDGPU_QPX_PARTITION_MODE] = "QPX",
+	[AMDGPU_CPX_PARTITION_MODE] = "CPX",
+};
+
+static ssize_t amdgpu_gfx_get_available_compute_partition(struct device *dev,
+						struct device_attribute *addr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	struct amdgpu_xcp_mgr *xcp_mgr = adev->xcp_mgr;
+	int size = 0, mode;
+	char *sep = "";
+
+	if (!xcp_mgr || !xcp_mgr->avail_xcp_modes)
+		return sysfs_emit(buf, "Not supported\n");
+
+	for_each_inst(mode, xcp_mgr->avail_xcp_modes) {
+		size += sysfs_emit_at(buf, size, "%s%s", sep, xcp_desc[mode]);
+		sep = ", ";
+	}
+
+	size += sysfs_emit_at(buf, size, "\n");
+
+	return size;
+}
+
+static int amdgpu_gfx_run_cleaner_shader_job(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct drm_gpu_scheduler *sched = &ring->sched;
+	struct drm_sched_entity entity;
+	static atomic_t counter;
+	struct dma_fence *f;
+	struct amdgpu_job *job;
+	struct amdgpu_ib *ib;
+	void *owner;
+	int i, r;
+
+	/* Initialize the scheduler entity */
+	r = drm_sched_entity_init(&entity, DRM_SCHED_PRIORITY_NORMAL,
+				  &sched, 1, NULL);
+	if (r) {
+		dev_err(adev->dev, "Failed setting up GFX kernel entity.\n");
+		goto err;
+	}
+
+	/*
+	 * Use some unique dummy value as the owner to make sure we execute
+	 * the cleaner shader on each submission. The value just need to change
+	 * for each submission and is otherwise meaningless.
+	 */
+	owner = (void *)(unsigned long)atomic_inc_return(&counter);
+
+	r = amdgpu_job_alloc_with_ib(ring->adev, &entity, owner,
+				     64, 0, &job,
+				     AMDGPU_KERNEL_JOB_ID_CLEANER_SHADER);
+	if (r)
+		goto err;
+
+	job->enforce_isolation = true;
+	/* always run the cleaner shader */
+	job->run_cleaner_shader = true;
+
+	ib = &job->ibs[0];
+	for (i = 0; i <= ring->funcs->align_mask; ++i)
+		ib->ptr[i] = ring->funcs->nop;
+	ib->length_dw = ring->funcs->align_mask + 1;
+
+	f = amdgpu_job_submit(job);
+
+	r = dma_fence_wait(f, false);
+	if (r)
+		goto err;
+
+	dma_fence_put(f);
+
+	/* Clean up the scheduler entity */
+	drm_sched_entity_destroy(&entity);
+	return 0;
+
+err:
+	return r;
+}
+
+static int amdgpu_gfx_run_cleaner_shader(struct amdgpu_device *adev, int xcp_id)
+{
+	int num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	struct amdgpu_ring *ring;
+	int num_xcc_to_clear;
+	int i, r, xcc_id;
+
+	if (adev->gfx.num_xcc_per_xcp)
+		num_xcc_to_clear = adev->gfx.num_xcc_per_xcp;
+	else
+		num_xcc_to_clear = 1;
+
+	for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) {
+		for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+			ring = &adev->gfx.compute_ring[i + xcc_id * adev->gfx.num_compute_rings];
+			if ((ring->xcp_id == xcp_id) && ring->sched.ready) {
+				r = amdgpu_gfx_run_cleaner_shader_job(ring);
+				if (r)
+					return r;
+				num_xcc_to_clear--;
+				break;
+			}
+		}
+	}
+
+	if (num_xcc_to_clear)
+		return -ENOENT;
+
+	return 0;
+}
+
+/**
+ * amdgpu_gfx_set_run_cleaner_shader - Execute the AMDGPU GFX Cleaner Shader
+ * @dev: The device structure
+ * @attr: The device attribute structure
+ * @buf: The buffer containing the input data
+ * @count: The size of the input data
+ *
+ * Provides the sysfs interface to manually run a cleaner shader, which is
+ * used to clear the GPU state between different tasks. Writing a value to the
+ * 'run_cleaner_shader' sysfs file triggers the cleaner shader execution.
+ * The value written corresponds to the partition index on multi-partition
+ * devices. On single-partition devices, the value should be '0'.
+ *
+ * The cleaner shader clears the Local Data Store (LDS) and General Purpose
+ * Registers (GPRs) to ensure data isolation between GPU workloads.
+ *
+ * Return: The number of bytes written to the sysfs file.
+ */
+static ssize_t amdgpu_gfx_set_run_cleaner_shader(struct device *dev,
+						 struct device_attribute *attr,
+						 const char *buf,
+						 size_t count)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	int ret;
+	long value;
+
+	if (amdgpu_in_reset(adev))
+		return -EPERM;
+	if (adev->in_suspend && !adev->in_runpm)
+		return -EPERM;
+
+	if (adev->gfx.disable_kq)
+		return -EPERM;
+
+	ret = kstrtol(buf, 0, &value);
+
+	if (ret)
+		return -EINVAL;
+
+	if (value < 0)
+		return -EINVAL;
+
+	if (adev->xcp_mgr) {
+		if (value >= adev->xcp_mgr->num_xcps)
+			return -EINVAL;
+	} else {
+		if (value > 1)
+			return -EINVAL;
+	}
+
+	ret = pm_runtime_get_sync(ddev->dev);
+	if (ret < 0) {
+		pm_runtime_put_autosuspend(ddev->dev);
+		return ret;
+	}
+
+	ret = amdgpu_gfx_run_cleaner_shader(adev, value);
+
+	pm_runtime_put_autosuspend(ddev->dev);
+
+	if (ret)
+		return ret;
+
+	return count;
+}
+
+/**
+ * amdgpu_gfx_get_enforce_isolation - Query AMDGPU GFX Enforce Isolation Settings
+ * @dev: The device structure
+ * @attr: The device attribute structure
+ * @buf: The buffer to store the output data
+ *
+ * Provides the sysfs read interface to get the current settings of the 'enforce_isolation'
+ * feature for each GPU partition. Reading from the 'enforce_isolation'
+ * sysfs file returns the isolation settings for all partitions, where '0'
+ * indicates disabled, '1' indicates enabled, and '2' indicates enabled in legacy mode,
+ * and '3' indicates enabled without cleaner shader.
+ *
+ * Return: The number of bytes read from the sysfs file.
+ */
+static ssize_t amdgpu_gfx_get_enforce_isolation(struct device *dev,
+						struct device_attribute *attr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	int i;
+	ssize_t size = 0;
+
+	if (adev->xcp_mgr) {
+		for (i = 0; i < adev->xcp_mgr->num_xcps; i++) {
+			size += sysfs_emit_at(buf, size, "%u", adev->enforce_isolation[i]);
+			if (i < (adev->xcp_mgr->num_xcps - 1))
+				size += sysfs_emit_at(buf, size, " ");
+		}
+		buf[size++] = '\n';
+	} else {
+		size = sysfs_emit_at(buf, 0, "%u\n", adev->enforce_isolation[0]);
+	}
+
+	return size;
+}
+
+/**
+ * amdgpu_gfx_set_enforce_isolation - Control AMDGPU GFX Enforce Isolation
+ * @dev: The device structure
+ * @attr: The device attribute structure
+ * @buf: The buffer containing the input data
+ * @count: The size of the input data
+ *
+ * This function allows control over the 'enforce_isolation' feature, which
+ * serializes access to the graphics engine. Writing '0' to disable, '1' to
+ * enable isolation with cleaner shader, '2' to enable legacy isolation without
+ * cleaner shader, or '3' to enable process isolation without submitting the
+ * cleaner shader to the 'enforce_isolation' sysfs file sets the isolation mode
+ * for each partition. The input should specify the setting for all
+ * partitions.
+ *
+ * Return: The number of bytes written to the sysfs file.
+ */
+static ssize_t amdgpu_gfx_set_enforce_isolation(struct device *dev,
+						struct device_attribute *attr,
+						const char *buf, size_t count)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	long partition_values[MAX_XCP] = {0};
+	int ret, i, num_partitions;
+	const char *input_buf = buf;
+
+	for (i = 0; i < (adev->xcp_mgr ? adev->xcp_mgr->num_xcps : 1); i++) {
+		ret = sscanf(input_buf, "%ld", &partition_values[i]);
+		if (ret <= 0)
+			break;
+
+		/* Move the pointer to the next value in the string */
+		input_buf = strchr(input_buf, ' ');
+		if (input_buf) {
+			input_buf++;
+		} else {
+			i++;
+			break;
+		}
+	}
+	num_partitions = i;
+
+	if (adev->xcp_mgr && num_partitions != adev->xcp_mgr->num_xcps)
+		return -EINVAL;
+
+	if (!adev->xcp_mgr && num_partitions != 1)
+		return -EINVAL;
+
+	for (i = 0; i < num_partitions; i++) {
+		if (partition_values[i] != 0 &&
+		    partition_values[i] != 1 &&
+		    partition_values[i] != 2 &&
+		    partition_values[i] != 3)
+			return -EINVAL;
+	}
+
+	mutex_lock(&adev->enforce_isolation_mutex);
+	for (i = 0; i < num_partitions; i++) {
+		switch (partition_values[i]) {
+		case 0:
+		default:
+			adev->enforce_isolation[i] = AMDGPU_ENFORCE_ISOLATION_DISABLE;
+			break;
+		case 1:
+			adev->enforce_isolation[i] =
+				AMDGPU_ENFORCE_ISOLATION_ENABLE;
+			break;
+		case 2:
+			adev->enforce_isolation[i] =
+				AMDGPU_ENFORCE_ISOLATION_ENABLE_LEGACY;
+			break;
+		case 3:
+			adev->enforce_isolation[i] =
+				AMDGPU_ENFORCE_ISOLATION_NO_CLEANER_SHADER;
+			break;
+		}
+	}
+	mutex_unlock(&adev->enforce_isolation_mutex);
+
+	amdgpu_mes_update_enforce_isolation(adev);
+
+	return count;
+}
+
+static ssize_t amdgpu_gfx_get_gfx_reset_mask(struct device *dev,
+						struct device_attribute *attr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	if (!adev)
+		return -ENODEV;
+
+	return amdgpu_show_reset_mask(buf, adev->gfx.gfx_supported_reset);
+}
+
+static ssize_t amdgpu_gfx_get_compute_reset_mask(struct device *dev,
+						struct device_attribute *attr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	if (!adev)
+		return -ENODEV;
+
+	return amdgpu_show_reset_mask(buf, adev->gfx.compute_supported_reset);
+}
+
+static DEVICE_ATTR(run_cleaner_shader, 0200,
+		   NULL, amdgpu_gfx_set_run_cleaner_shader);
+
+static DEVICE_ATTR(enforce_isolation, 0644,
+		   amdgpu_gfx_get_enforce_isolation,
+		   amdgpu_gfx_set_enforce_isolation);
+
+static DEVICE_ATTR(current_compute_partition, 0644,
+		   amdgpu_gfx_get_current_compute_partition,
+		   amdgpu_gfx_set_compute_partition);
+
+static DEVICE_ATTR(available_compute_partition, 0444,
+		   amdgpu_gfx_get_available_compute_partition, NULL);
+static DEVICE_ATTR(gfx_reset_mask, 0444,
+		   amdgpu_gfx_get_gfx_reset_mask, NULL);
+
+static DEVICE_ATTR(compute_reset_mask, 0444,
+		   amdgpu_gfx_get_compute_reset_mask, NULL);
+
+static int amdgpu_gfx_sysfs_xcp_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_xcp_mgr *xcp_mgr = adev->xcp_mgr;
+	bool xcp_switch_supported;
+	int r;
+
+	if (!xcp_mgr)
+		return 0;
+
+	xcp_switch_supported =
+		(xcp_mgr->funcs && xcp_mgr->funcs->switch_partition_mode);
+
+	if (!xcp_switch_supported)
+		dev_attr_current_compute_partition.attr.mode &=
+			~(S_IWUSR | S_IWGRP | S_IWOTH);
+
+	r = device_create_file(adev->dev, &dev_attr_current_compute_partition);
+	if (r)
+		return r;
+
+	if (xcp_switch_supported)
+		r = device_create_file(adev->dev,
+				       &dev_attr_available_compute_partition);
+
+	return r;
+}
+
+static void amdgpu_gfx_sysfs_xcp_fini(struct amdgpu_device *adev)
+{
+	struct amdgpu_xcp_mgr *xcp_mgr = adev->xcp_mgr;
+	bool xcp_switch_supported;
+
+	if (!xcp_mgr)
+		return;
+
+	xcp_switch_supported =
+		(xcp_mgr->funcs && xcp_mgr->funcs->switch_partition_mode);
+	device_remove_file(adev->dev, &dev_attr_current_compute_partition);
+
+	if (xcp_switch_supported)
+		device_remove_file(adev->dev,
+				   &dev_attr_available_compute_partition);
+}
+
+static int amdgpu_gfx_sysfs_isolation_shader_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = device_create_file(adev->dev, &dev_attr_enforce_isolation);
+	if (r)
+		return r;
+	if (adev->gfx.enable_cleaner_shader)
+		r = device_create_file(adev->dev, &dev_attr_run_cleaner_shader);
+
+	return r;
+}
+
+static void amdgpu_gfx_sysfs_isolation_shader_fini(struct amdgpu_device *adev)
+{
+	device_remove_file(adev->dev, &dev_attr_enforce_isolation);
+	if (adev->gfx.enable_cleaner_shader)
+		device_remove_file(adev->dev, &dev_attr_run_cleaner_shader);
+}
+
+static int amdgpu_gfx_sysfs_reset_mask_init(struct amdgpu_device *adev)
+{
+	int r = 0;
+
+	if (!amdgpu_gpu_recovery)
+		return r;
+
+	if (adev->gfx.num_gfx_rings) {
+		r = device_create_file(adev->dev, &dev_attr_gfx_reset_mask);
+		if (r)
+			return r;
+	}
+
+	if (adev->gfx.num_compute_rings) {
+		r = device_create_file(adev->dev, &dev_attr_compute_reset_mask);
+		if (r)
+			return r;
+	}
+
+	return r;
+}
+
+static void amdgpu_gfx_sysfs_reset_mask_fini(struct amdgpu_device *adev)
+{
+	if (!amdgpu_gpu_recovery)
+		return;
+
+	if (adev->gfx.num_gfx_rings)
+		device_remove_file(adev->dev, &dev_attr_gfx_reset_mask);
+
+	if (adev->gfx.num_compute_rings)
+		device_remove_file(adev->dev, &dev_attr_compute_reset_mask);
+}
+
+int amdgpu_gfx_sysfs_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = amdgpu_gfx_sysfs_xcp_init(adev);
+	if (r) {
+		dev_err(adev->dev, "failed to create xcp sysfs files");
+		return r;
+	}
+
+	r = amdgpu_gfx_sysfs_isolation_shader_init(adev);
+	if (r)
+		dev_err(adev->dev, "failed to create isolation sysfs files");
+
+	r = amdgpu_gfx_sysfs_reset_mask_init(adev);
+	if (r)
+		dev_err(adev->dev, "failed to create reset mask sysfs files");
+
+	return r;
+}
+
+void amdgpu_gfx_sysfs_fini(struct amdgpu_device *adev)
+{
+	if (adev->dev->kobj.sd) {
+		amdgpu_gfx_sysfs_xcp_fini(adev);
+		amdgpu_gfx_sysfs_isolation_shader_fini(adev);
+		amdgpu_gfx_sysfs_reset_mask_fini(adev);
+	}
+}
+
+int amdgpu_gfx_cleaner_shader_sw_init(struct amdgpu_device *adev,
+				      unsigned int cleaner_shader_size)
+{
+	if (!adev->gfx.enable_cleaner_shader)
+		return -EOPNOTSUPP;
+
+	return amdgpu_bo_create_kernel(adev, cleaner_shader_size, PAGE_SIZE,
+				       AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT,
+				       &adev->gfx.cleaner_shader_obj,
+				       &adev->gfx.cleaner_shader_gpu_addr,
+				       (void **)&adev->gfx.cleaner_shader_cpu_ptr);
+}
+
+void amdgpu_gfx_cleaner_shader_sw_fini(struct amdgpu_device *adev)
+{
+	if (!adev->gfx.enable_cleaner_shader)
+		return;
+
+	amdgpu_bo_free_kernel(&adev->gfx.cleaner_shader_obj,
+			      &adev->gfx.cleaner_shader_gpu_addr,
+			      (void **)&adev->gfx.cleaner_shader_cpu_ptr);
+}
+
+void amdgpu_gfx_cleaner_shader_init(struct amdgpu_device *adev,
+				    unsigned int cleaner_shader_size,
+				    const void *cleaner_shader_ptr)
+{
+	if (!adev->gfx.enable_cleaner_shader)
+		return;
+
+	if (adev->gfx.cleaner_shader_cpu_ptr && cleaner_shader_ptr)
+		memcpy_toio(adev->gfx.cleaner_shader_cpu_ptr, cleaner_shader_ptr,
+			    cleaner_shader_size);
+}
+
+/**
+ * amdgpu_gfx_kfd_sch_ctrl - Control the KFD scheduler from the KGD (Graphics Driver)
+ * @adev: amdgpu_device pointer
+ * @idx: Index of the scheduler to control
+ * @enable: Whether to enable or disable the KFD scheduler
+ *
+ * This function is used to control the KFD (Kernel Fusion Driver) scheduler
+ * from the KGD. It is part of the cleaner shader feature. This function plays
+ * a key role in enforcing process isolation on the GPU.
+ *
+ * The function uses a reference count mechanism (kfd_sch_req_count) to keep
+ * track of the number of requests to enable the KFD scheduler. When a request
+ * to enable the KFD scheduler is made, the reference count is decremented.
+ * When the reference count reaches zero, a delayed work is scheduled to
+ * enforce isolation after a delay of GFX_SLICE_PERIOD.
+ *
+ * When a request to disable the KFD scheduler is made, the function first
+ * checks if the reference count is zero. If it is, it cancels the delayed work
+ * for enforcing isolation and checks if the KFD scheduler is active. If the
+ * KFD scheduler is active, it sends a request to stop the KFD scheduler and
+ * sets the KFD scheduler state to inactive. Then, it increments the reference
+ * count.
+ *
+ * The function is synchronized using the kfd_sch_mutex to ensure that the KFD
+ * scheduler state and reference count are updated atomically.
+ *
+ * Note: If the reference count is already zero when a request to enable the
+ * KFD scheduler is made, it means there's an imbalance bug somewhere. The
+ * function triggers a warning in this case.
+ */
+static void amdgpu_gfx_kfd_sch_ctrl(struct amdgpu_device *adev, u32 idx,
+				    bool enable)
+{
+	mutex_lock(&adev->gfx.userq_sch_mutex);
+
+	if (enable) {
+		/* If the count is already 0, it means there's an imbalance bug somewhere.
+		 * Note that the bug may be in a different caller than the one which triggers the
+		 * WARN_ON_ONCE.
+		 */
+		if (WARN_ON_ONCE(adev->gfx.userq_sch_req_count[idx] == 0)) {
+			dev_err(adev->dev, "Attempted to enable KFD scheduler when reference count is already zero\n");
+			goto unlock;
+		}
+
+		adev->gfx.userq_sch_req_count[idx]--;
+
+		if (adev->gfx.userq_sch_req_count[idx] == 0 &&
+		    adev->gfx.userq_sch_inactive[idx]) {
+			schedule_delayed_work(&adev->gfx.enforce_isolation[idx].work,
+					      msecs_to_jiffies(adev->gfx.enforce_isolation_time[idx]));
+		}
+	} else {
+		if (adev->gfx.userq_sch_req_count[idx] == 0) {
+			cancel_delayed_work_sync(&adev->gfx.enforce_isolation[idx].work);
+			if (!adev->gfx.userq_sch_inactive[idx]) {
+				amdgpu_userq_stop_sched_for_enforce_isolation(adev, idx);
+				if (adev->kfd.init_complete)
+					amdgpu_amdkfd_stop_sched(adev, idx);
+				adev->gfx.userq_sch_inactive[idx] = true;
+			}
+		}
+
+		adev->gfx.userq_sch_req_count[idx]++;
+	}
+
+unlock:
+	mutex_unlock(&adev->gfx.userq_sch_mutex);
+}
+
+/**
+ * amdgpu_gfx_enforce_isolation_handler - work handler for enforcing shader isolation
+ *
+ * @work: work_struct.
+ *
+ * This function is the work handler for enforcing shader isolation on AMD GPUs.
+ * It counts the number of emitted fences for each GFX and compute ring. If there
+ * are any fences, it schedules the `enforce_isolation_work` to be run after a
+ * delay of `GFX_SLICE_PERIOD`. If there are no fences, it signals the Kernel Fusion
+ * Driver (KFD) to resume the runqueue. The function is synchronized using the
+ * `enforce_isolation_mutex`.
+ */
+void amdgpu_gfx_enforce_isolation_handler(struct work_struct *work)
+{
+	struct amdgpu_isolation_work *isolation_work =
+		container_of(work, struct amdgpu_isolation_work, work.work);
+	struct amdgpu_device *adev = isolation_work->adev;
+	u32 i, idx, fences = 0;
+
+	if (isolation_work->xcp_id == AMDGPU_XCP_NO_PARTITION)
+		idx = 0;
+	else
+		idx = isolation_work->xcp_id;
+
+	if (idx >= MAX_XCP)
+		return;
+
+	mutex_lock(&adev->enforce_isolation_mutex);
+	for (i = 0; i < AMDGPU_MAX_GFX_RINGS; ++i) {
+		if (isolation_work->xcp_id == adev->gfx.gfx_ring[i].xcp_id)
+			fences += amdgpu_fence_count_emitted(&adev->gfx.gfx_ring[i]);
+	}
+	for (i = 0; i < (AMDGPU_MAX_COMPUTE_RINGS * AMDGPU_MAX_GC_INSTANCES); ++i) {
+		if (isolation_work->xcp_id == adev->gfx.compute_ring[i].xcp_id)
+			fences += amdgpu_fence_count_emitted(&adev->gfx.compute_ring[i]);
+	}
+	if (fences) {
+		/* we've already had our timeslice, so let's wrap this up */
+		schedule_delayed_work(&adev->gfx.enforce_isolation[idx].work,
+				      msecs_to_jiffies(1));
+	} else {
+		/* Tell KFD to resume the runqueue */
+		WARN_ON_ONCE(!adev->gfx.userq_sch_inactive[idx]);
+		WARN_ON_ONCE(adev->gfx.userq_sch_req_count[idx]);
+
+		amdgpu_userq_start_sched_for_enforce_isolation(adev, idx);
+		if (adev->kfd.init_complete)
+			amdgpu_amdkfd_start_sched(adev, idx);
+		adev->gfx.userq_sch_inactive[idx] = false;
+	}
+	mutex_unlock(&adev->enforce_isolation_mutex);
+}
+
+/**
+ * amdgpu_gfx_enforce_isolation_wait_for_kfd - Manage KFD wait period for process isolation
+ * @adev: amdgpu_device pointer
+ * @idx: Index of the GPU partition
+ *
+ * When kernel submissions come in, the jobs are given a time slice and once
+ * that time slice is up, if there are KFD user queues active, kernel
+ * submissions are blocked until KFD has had its time slice. Once the KFD time
+ * slice is up, KFD user queues are preempted and kernel submissions are
+ * unblocked and allowed to run again.
+ */
+static void
+amdgpu_gfx_enforce_isolation_wait_for_kfd(struct amdgpu_device *adev,
+					  u32 idx)
+{
+	unsigned long cjiffies;
+	bool wait = false;
+
+	mutex_lock(&adev->enforce_isolation_mutex);
+	if (adev->enforce_isolation[idx] == AMDGPU_ENFORCE_ISOLATION_ENABLE) {
+		/* set the initial values if nothing is set */
+		if (!adev->gfx.enforce_isolation_jiffies[idx]) {
+			adev->gfx.enforce_isolation_jiffies[idx] = jiffies;
+			adev->gfx.enforce_isolation_time[idx] =	GFX_SLICE_PERIOD_MS;
+		}
+		/* Make sure KFD gets a chance to run */
+		if (amdgpu_amdkfd_compute_active(adev, idx)) {
+			cjiffies = jiffies;
+			if (time_after(cjiffies, adev->gfx.enforce_isolation_jiffies[idx])) {
+				cjiffies -= adev->gfx.enforce_isolation_jiffies[idx];
+				if ((jiffies_to_msecs(cjiffies) >= GFX_SLICE_PERIOD_MS)) {
+					/* if our time is up, let KGD work drain before scheduling more */
+					wait = true;
+					/* reset the timer period */
+					adev->gfx.enforce_isolation_time[idx] =	GFX_SLICE_PERIOD_MS;
+				} else {
+					/* set the timer period to what's left in our time slice */
+					adev->gfx.enforce_isolation_time[idx] =
+						GFX_SLICE_PERIOD_MS - jiffies_to_msecs(cjiffies);
+				}
+			} else {
+				/* if jiffies wrap around we will just wait a little longer */
+				adev->gfx.enforce_isolation_jiffies[idx] = jiffies;
+			}
+		} else {
+			/* if there is no KFD work, then set the full slice period */
+			adev->gfx.enforce_isolation_jiffies[idx] = jiffies;
+			adev->gfx.enforce_isolation_time[idx] = GFX_SLICE_PERIOD_MS;
+		}
+	}
+	mutex_unlock(&adev->enforce_isolation_mutex);
+
+	if (wait)
+		msleep(GFX_SLICE_PERIOD_MS);
+}
+
+/**
+ * amdgpu_gfx_enforce_isolation_ring_begin_use - Begin use of a ring with enforced isolation
+ * @ring: Pointer to the amdgpu_ring structure
+ *
+ * Ring begin_use helper implementation for gfx which serializes access to the
+ * gfx IP between kernel submission IOCTLs and KFD user queues when isolation
+ * enforcement is enabled. The kernel submission IOCTLs and KFD user queues
+ * each get a time slice when both are active.
+ */
+void amdgpu_gfx_enforce_isolation_ring_begin_use(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 idx;
+	bool sched_work = false;
+
+	if (!adev->gfx.enable_cleaner_shader)
+		return;
+
+	if (ring->xcp_id == AMDGPU_XCP_NO_PARTITION)
+		idx = 0;
+	else
+		idx = ring->xcp_id;
+
+	if (idx >= MAX_XCP)
+		return;
+
+	/* Don't submit more work until KFD has had some time */
+	amdgpu_gfx_enforce_isolation_wait_for_kfd(adev, idx);
+
+	mutex_lock(&adev->enforce_isolation_mutex);
+	if (adev->enforce_isolation[idx] == AMDGPU_ENFORCE_ISOLATION_ENABLE) {
+		if (adev->kfd.init_complete)
+			sched_work = true;
+	}
+	mutex_unlock(&adev->enforce_isolation_mutex);
+
+	if (sched_work)
+		amdgpu_gfx_kfd_sch_ctrl(adev, idx, false);
+}
+
+/**
+ * amdgpu_gfx_enforce_isolation_ring_end_use - End use of a ring with enforced isolation
+ * @ring: Pointer to the amdgpu_ring structure
+ *
+ * Ring end_use helper implementation for gfx which serializes access to the
+ * gfx IP between kernel submission IOCTLs and KFD user queues when isolation
+ * enforcement is enabled. The kernel submission IOCTLs and KFD user queues
+ * each get a time slice when both are active.
+ */
+void amdgpu_gfx_enforce_isolation_ring_end_use(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 idx;
+	bool sched_work = false;
+
+	if (!adev->gfx.enable_cleaner_shader)
+		return;
+
+	if (ring->xcp_id == AMDGPU_XCP_NO_PARTITION)
+		idx = 0;
+	else
+		idx = ring->xcp_id;
+
+	if (idx >= MAX_XCP)
+		return;
+
+	mutex_lock(&adev->enforce_isolation_mutex);
+	if (adev->enforce_isolation[idx] == AMDGPU_ENFORCE_ISOLATION_ENABLE) {
+		if (adev->kfd.init_complete)
+			sched_work = true;
+	}
+	mutex_unlock(&adev->enforce_isolation_mutex);
+
+	if (sched_work)
+		amdgpu_gfx_kfd_sch_ctrl(adev, idx, true);
+}
+
+void amdgpu_gfx_profile_idle_work_handler(struct work_struct *work)
+{
+	struct amdgpu_device *adev =
+		container_of(work, struct amdgpu_device, gfx.idle_work.work);
+	enum PP_SMC_POWER_PROFILE profile;
+	u32 i, fences = 0;
+	int r;
+
+	if (adev->gfx.num_gfx_rings)
+		profile = PP_SMC_POWER_PROFILE_FULLSCREEN3D;
+	else
+		profile = PP_SMC_POWER_PROFILE_COMPUTE;
+
+	for (i = 0; i < AMDGPU_MAX_GFX_RINGS; ++i)
+		fences += amdgpu_fence_count_emitted(&adev->gfx.gfx_ring[i]);
+	for (i = 0; i < (AMDGPU_MAX_COMPUTE_RINGS * AMDGPU_MAX_GC_INSTANCES); ++i)
+		fences += amdgpu_fence_count_emitted(&adev->gfx.compute_ring[i]);
+	if (!fences && !atomic_read(&adev->gfx.total_submission_cnt)) {
+		mutex_lock(&adev->gfx.workload_profile_mutex);
+		if (adev->gfx.workload_profile_active) {
+			r = amdgpu_dpm_switch_power_profile(adev, profile, false);
+			if (r)
+				dev_warn(adev->dev, "(%d) failed to disable %s power profile mode\n", r,
+					 profile == PP_SMC_POWER_PROFILE_FULLSCREEN3D ?
+					 "fullscreen 3D" : "compute");
+			adev->gfx.workload_profile_active = false;
+		}
+		mutex_unlock(&adev->gfx.workload_profile_mutex);
+	} else {
+		schedule_delayed_work(&adev->gfx.idle_work, GFX_PROFILE_IDLE_TIMEOUT);
+	}
+}
+
+void amdgpu_gfx_profile_ring_begin_use(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	enum PP_SMC_POWER_PROFILE profile;
+	int r;
+
+	if (amdgpu_dpm_is_overdrive_enabled(adev))
+		return;
+
+	if (adev->gfx.num_gfx_rings)
+		profile = PP_SMC_POWER_PROFILE_FULLSCREEN3D;
+	else
+		profile = PP_SMC_POWER_PROFILE_COMPUTE;
+
+	atomic_inc(&adev->gfx.total_submission_cnt);
+
+	cancel_delayed_work_sync(&adev->gfx.idle_work);
+
+	/* We can safely return early here because we've cancelled the
+	 * the delayed work so there is no one else to set it to false
+	 * and we don't care if someone else sets it to true.
+	 */
+	if (adev->gfx.workload_profile_active)
+		return;
+
+	mutex_lock(&adev->gfx.workload_profile_mutex);
+	if (!adev->gfx.workload_profile_active) {
+		r = amdgpu_dpm_switch_power_profile(adev, profile, true);
+		if (r)
+			dev_warn(adev->dev, "(%d) failed to disable %s power profile mode\n", r,
+				 profile == PP_SMC_POWER_PROFILE_FULLSCREEN3D ?
+				 "fullscreen 3D" : "compute");
+		adev->gfx.workload_profile_active = true;
+	}
+	mutex_unlock(&adev->gfx.workload_profile_mutex);
+}
+
+void amdgpu_gfx_profile_ring_end_use(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (amdgpu_dpm_is_overdrive_enabled(adev))
+		return;
+
+	atomic_dec(&ring->adev->gfx.total_submission_cnt);
+
+	schedule_delayed_work(&ring->adev->gfx.idle_work, GFX_PROFILE_IDLE_TIMEOUT);
+}
+
+/**
+ * amdgpu_gfx_csb_preamble_start - Set CSB preamble start
+ *
+ * @buffer: This is an output variable that gets the PACKET3 preamble setup.
+ *
+ * Return:
+ * return the latest index.
+ */
+u32 amdgpu_gfx_csb_preamble_start(u32 *buffer)
+{
+	u32 count = 0;
+
+	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
+	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
+
+	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1));
+	buffer[count++] = cpu_to_le32(0x80000000);
+	buffer[count++] = cpu_to_le32(0x80000000);
+
+	return count;
+}
+
+/**
+ * amdgpu_gfx_csb_data_parser - Parser CS data
+ *
+ * @adev: amdgpu_device pointer used to get the CS data and other gfx info.
+ * @buffer: This is an output variable that gets the PACKET3 preamble end.
+ * @count: Index to start set the preemble end.
+ *
+ * Return:
+ * return the latest index.
+ */
+u32 amdgpu_gfx_csb_data_parser(struct amdgpu_device *adev, u32 *buffer, u32 count)
+{
+	const struct cs_section_def *sect = NULL;
+	const struct cs_extent_def *ext = NULL;
+	u32 i;
+
+	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
+		for (ext = sect->section; ext->extent != NULL; ++ext) {
+			if (sect->id == SECT_CONTEXT) {
+				buffer[count++] = cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count));
+				buffer[count++] = cpu_to_le32(ext->reg_index - PACKET3_SET_CONTEXT_REG_START);
+
+				for (i = 0; i < ext->reg_count; i++)
+					buffer[count++] = cpu_to_le32(ext->extent[i]);
+			}
+		}
+	}
+
+	return count;
+}
+
+/**
+ * amdgpu_gfx_csb_preamble_end - Set CSB preamble end
+ *
+ * @buffer: This is an output variable that gets the PACKET3 preamble end.
+ * @count: Index to start set the preemble end.
+ */
+void amdgpu_gfx_csb_preamble_end(u32 *buffer, u32 count)
+{
+	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
+	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE);
+
+	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0));
+	buffer[count++] = cpu_to_le32(0);
+}
+
+/*
+ * debugfs for to enable/disable gfx job submission to specific core.
+ */
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_debugfs_gfx_sched_mask_set(void *data, u64 val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i;
+	u64 mask = 0;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+
+	mask = (1ULL << adev->gfx.num_gfx_rings) - 1;
+	if ((val & mask) == 0)
+		return -EINVAL;
+
+	for (i = 0; i < adev->gfx.num_gfx_rings; ++i) {
+		ring = &adev->gfx.gfx_ring[i];
+		if (val & (1 << i))
+			ring->sched.ready = true;
+		else
+			ring->sched.ready = false;
+	}
+	/* publish sched.ready flag update effective immediately across smp */
+	smp_rmb();
+	return 0;
+}
+
+static int amdgpu_debugfs_gfx_sched_mask_get(void *data, u64 *val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i;
+	u64 mask = 0;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+	for (i = 0; i < adev->gfx.num_gfx_rings; ++i) {
+		ring = &adev->gfx.gfx_ring[i];
+		if (ring->sched.ready)
+			mask |= 1ULL << i;
+	}
+
+	*val = mask;
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(amdgpu_debugfs_gfx_sched_mask_fops,
+			 amdgpu_debugfs_gfx_sched_mask_get,
+			 amdgpu_debugfs_gfx_sched_mask_set, "%llx\n");
+
+#endif
+
+void amdgpu_debugfs_gfx_sched_mask_init(struct amdgpu_device *adev)
+{
+#if defined(CONFIG_DEBUG_FS)
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+	struct dentry *root = minor->debugfs_root;
+	char name[32];
+
+	if (!(adev->gfx.num_gfx_rings > 1))
+		return;
+	sprintf(name, "amdgpu_gfx_sched_mask");
+	debugfs_create_file(name, 0600, root, adev,
+			    &amdgpu_debugfs_gfx_sched_mask_fops);
+#endif
+}
+
+/*
+ * debugfs for to enable/disable compute job submission to specific core.
+ */
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_debugfs_compute_sched_mask_set(void *data, u64 val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i;
+	u64 mask = 0;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+
+	mask = (1ULL << adev->gfx.num_compute_rings) - 1;
+	if ((val & mask) == 0)
+		return -EINVAL;
+
+	for (i = 0; i < adev->gfx.num_compute_rings; ++i) {
+		ring = &adev->gfx.compute_ring[i];
+		if (val & (1 << i))
+			ring->sched.ready = true;
+		else
+			ring->sched.ready = false;
+	}
+
+	/* publish sched.ready flag update effective immediately across smp */
+	smp_rmb();
+	return 0;
+}
+
+static int amdgpu_debugfs_compute_sched_mask_get(void *data, u64 *val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i;
+	u64 mask = 0;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+	for (i = 0; i < adev->gfx.num_compute_rings; ++i) {
+		ring = &adev->gfx.compute_ring[i];
+		if (ring->sched.ready)
+			mask |= 1ULL << i;
+	}
+
+	*val = mask;
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(amdgpu_debugfs_compute_sched_mask_fops,
+			 amdgpu_debugfs_compute_sched_mask_get,
+			 amdgpu_debugfs_compute_sched_mask_set, "%llx\n");
+
+#endif
+
+void amdgpu_debugfs_compute_sched_mask_init(struct amdgpu_device *adev)
+{
+#if defined(CONFIG_DEBUG_FS)
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+	struct dentry *root = minor->debugfs_root;
+	char name[32];
+
+	if (!(adev->gfx.num_compute_rings > 1))
+		return;
+	sprintf(name, "amdgpu_compute_sched_mask");
+	debugfs_create_file(name, 0600, root, adev,
+			    &amdgpu_debugfs_compute_sched_mask_fops);
+#endif
+}
+
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index bfabea76d166..efd61a1ccc66 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
@@ -34,6 +34,7 @@
 #include "soc15.h"
 #include "amdgpu_ras.h"
 #include "amdgpu_ring_mux.h"
+#include "amdgpu_xcp.h"
 
 /* GFX current status */
 #define AMDGPU_GFX_NORMAL_MODE			0x00000000L
@@ -43,9 +44,10 @@
 #define AMDGPU_GFX_LBPW_DISABLED_MODE		0x00000008L
 
 #define AMDGPU_MAX_GC_INSTANCES		8
+#define AMDGPU_MAX_QUEUES		128
 
-#define AMDGPU_MAX_GFX_QUEUES KGD_MAX_QUEUES
-#define AMDGPU_MAX_COMPUTE_QUEUES KGD_MAX_QUEUES
+#define AMDGPU_MAX_GFX_QUEUES AMDGPU_MAX_QUEUES
+#define AMDGPU_MAX_COMPUTE_QUEUES AMDGPU_MAX_QUEUES
 
 enum amdgpu_gfx_pipe_priority {
 	AMDGPU_GFX_PIPE_PRIO_NORMAL = AMDGPU_RING_PRIO_1,
@@ -55,13 +57,46 @@ enum amdgpu_gfx_pipe_priority {
 #define AMDGPU_GFX_QUEUE_PRIORITY_MINIMUM  0
 #define AMDGPU_GFX_QUEUE_PRIORITY_MAXIMUM  15
 
+/* 1 second timeout */
+#define GFX_PROFILE_IDLE_TIMEOUT	msecs_to_jiffies(1000)
+
 enum amdgpu_gfx_partition {
 	AMDGPU_SPX_PARTITION_MODE = 0,
 	AMDGPU_DPX_PARTITION_MODE = 1,
 	AMDGPU_TPX_PARTITION_MODE = 2,
 	AMDGPU_QPX_PARTITION_MODE = 3,
 	AMDGPU_CPX_PARTITION_MODE = 4,
-	AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE,
+	AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE = -1,
+	/* Automatically choose the right mode */
+	AMDGPU_AUTO_COMPUTE_PARTITION_MODE = -2,
+};
+
+#define NUM_XCC(x) hweight16(x)
+
+enum amdgpu_gfx_ras_mem_id_type {
+	AMDGPU_GFX_CP_MEM = 0,
+	AMDGPU_GFX_GCEA_MEM,
+	AMDGPU_GFX_GC_CANE_MEM,
+	AMDGPU_GFX_GCUTCL2_MEM,
+	AMDGPU_GFX_GDS_MEM,
+	AMDGPU_GFX_LDS_MEM,
+	AMDGPU_GFX_RLC_MEM,
+	AMDGPU_GFX_SP_MEM,
+	AMDGPU_GFX_SPI_MEM,
+	AMDGPU_GFX_SQC_MEM,
+	AMDGPU_GFX_SQ_MEM,
+	AMDGPU_GFX_TA_MEM,
+	AMDGPU_GFX_TCC_MEM,
+	AMDGPU_GFX_TCA_MEM,
+	AMDGPU_GFX_TCI_MEM,
+	AMDGPU_GFX_TCP_MEM,
+	AMDGPU_GFX_TD_MEM,
+	AMDGPU_GFX_TCX_MEM,
+	AMDGPU_GFX_ATC_L2_MEM,
+	AMDGPU_GFX_UTCL2_MEM,
+	AMDGPU_GFX_VML2_MEM,
+	AMDGPU_GFX_VML2_WALKER_MEM,
+	AMDGPU_GFX_MEM_TYPE_NUM
 };
 
 struct amdgpu_mec {
@@ -75,8 +110,10 @@ struct amdgpu_mec {
 	u32 num_mec;
 	u32 num_pipe_per_mec;
 	u32 num_queue_per_pipe;
-	void			*mqd_backup[AMDGPU_MAX_COMPUTE_RINGS + 1];
+	void			*mqd_backup[AMDGPU_MAX_COMPUTE_RINGS * AMDGPU_MAX_GC_INSTANCES];
+};
 
+struct amdgpu_mec_bitmap {
 	/* These are the resources for which amdgpu takes ownership */
 	DECLARE_BITMAP(queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
 };
@@ -105,6 +142,10 @@ struct kiq_pm4_funcs {
 	void (*kiq_invalidate_tlbs)(struct amdgpu_ring *kiq_ring,
 				uint16_t pasid, uint32_t flush_type,
 				bool all_hub);
+	void (*kiq_reset_hw_queue)(struct amdgpu_ring *kiq_ring,
+				   uint32_t queue_type, uint32_t me_id,
+				   uint32_t pipe_id, uint32_t queue_id,
+				   uint32_t xcc_id, uint32_t vmid);
 	/* Packet sizes */
 	int set_resources_size;
 	int map_queues_size;
@@ -120,6 +161,7 @@ struct amdgpu_kiq {
 	struct amdgpu_ring	ring;
 	struct amdgpu_irq_src	irq;
 	const struct kiq_pm4_funcs *pmf;
+	void			*mqd_backup;
 };
 
 /*
@@ -128,10 +170,46 @@ struct amdgpu_kiq {
 #define AMDGPU_GFX_MAX_SE 4
 #define AMDGPU_GFX_MAX_SH_PER_SE 2
 
+/**
+ * amdgpu_rb_config - Configure a single Render Backend (RB)
+ *
+ * Bad RBs are fused off and there is a harvest register the driver reads to
+ * determine which RB(s) are fused off so that the driver can configure the
+ * hardware state so that nothing gets sent to them. There are also user
+ * harvest registers that the driver can program to disable additional RBs,
+ * etc., for testing purposes.
+ */
 struct amdgpu_rb_config {
+	/**
+	 * @rb_backend_disable:
+	 *
+	 * The value captured from register RB_BACKEND_DISABLE indicates if the
+	 * RB backend is disabled or not.
+	 */
 	uint32_t rb_backend_disable;
+
+	/**
+	 * @user_rb_backend_disable:
+	 *
+	 * The value captured from register USER_RB_BACKEND_DISABLE indicates
+	 * if the User RB backend is disabled or not.
+	 */
 	uint32_t user_rb_backend_disable;
+
+	/**
+	 * @raster_config:
+	 *
+	 * To set up all of the states, it is necessary to have two registers
+	 * to keep all of the states. This field holds the first register.
+	 */
 	uint32_t raster_config;
+
+	/**
+	 * @raster_config_1:
+	 *
+	 * To set up all of the states, it is necessary to have two registers
+	 * to keep all of the states. This field holds the second register.
+	 */
 	uint32_t raster_config_1;
 };
 
@@ -179,6 +257,13 @@ struct amdgpu_gfx_config {
 	uint32_t macrotile_mode_array[16];
 
 	struct gb_addr_config gb_addr_config_fields;
+
+	/**
+	 * @rb_config:
+	 *
+	 * Matrix that keeps all the Render Backend (color and depth buffer
+	 * handling) configuration on the 3D engine.
+	 */
 	struct amdgpu_rb_config rb_config[AMDGPU_GFX_MAX_SE][AMDGPU_GFX_MAX_SH_PER_SE];
 
 	/* gfx configure feature */
@@ -203,6 +288,15 @@ struct amdgpu_gfx_config {
 	uint32_t gc_gl1c_per_sa;
 	uint32_t gc_gl1c_size_per_instance;
 	uint32_t gc_gl2c_per_gpu;
+	uint32_t gc_tcp_size_per_cu;
+	uint32_t gc_num_cu_per_sqc;
+	uint32_t gc_tcc_size;
+	uint32_t gc_tcp_cache_line_size;
+	uint32_t gc_instruction_cache_size_per_sqc;
+	uint32_t gc_instruction_cache_line_size;
+	uint32_t gc_scalar_data_cache_size_per_sqc;
+	uint32_t gc_scalar_data_cache_line_size;
+	uint32_t gc_tcc_cache_line_size;
 };
 
 struct amdgpu_cu_info {
@@ -216,13 +310,12 @@ struct amdgpu_cu_info {
 	uint32_t number;
 	uint32_t ao_cu_mask;
 	uint32_t ao_cu_bitmap[4][4];
-	uint32_t bitmap[4][4];
+	uint32_t bitmap[AMDGPU_MAX_GC_INSTANCES][4][4];
 };
 
 struct amdgpu_gfx_ras {
 	struct amdgpu_ras_block_object  ras_block;
 	void (*enable_watchdog_timer)(struct amdgpu_device *adev);
-	bool (*query_utcl2_poison_status)(struct amdgpu_device *adev);
 	int (*rlc_gc_fed_irq)(struct amdgpu_device *adev,
 				struct amdgpu_irq_src *source,
 				struct amdgpu_iv_entry *entry);
@@ -230,23 +323,39 @@ struct amdgpu_gfx_ras {
 						struct amdgpu_iv_entry *entry);
 };
 
+struct amdgpu_gfx_shadow_info {
+	u32 shadow_size;
+	u32 shadow_alignment;
+	u32 csa_size;
+	u32 csa_alignment;
+};
+
 struct amdgpu_gfx_funcs {
 	/* get the gpu clock counter */
 	uint64_t (*get_gpu_clock_counter)(struct amdgpu_device *adev);
 	void (*select_se_sh)(struct amdgpu_device *adev, u32 se_num,
-			     u32 sh_num, u32 instance);
-	void (*read_wave_data)(struct amdgpu_device *adev, uint32_t simd,
+			     u32 sh_num, u32 instance, int xcc_id);
+	void (*read_wave_data)(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 			       uint32_t wave, uint32_t *dst, int *no_fields);
-	void (*read_wave_vgprs)(struct amdgpu_device *adev, uint32_t simd,
+	void (*read_wave_vgprs)(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				uint32_t wave, uint32_t thread, uint32_t start,
 				uint32_t size, uint32_t *dst);
-	void (*read_wave_sgprs)(struct amdgpu_device *adev, uint32_t simd,
+	void (*read_wave_sgprs)(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				uint32_t wave, uint32_t start, uint32_t size,
 				uint32_t *dst);
 	void (*select_me_pipe_q)(struct amdgpu_device *adev, u32 me, u32 pipe,
-				 u32 queue, u32 vmid);
+				 u32 queue, u32 vmid, u32 xcc_id);
 	void (*init_spm_golden)(struct amdgpu_device *adev);
 	void (*update_perfmon_mgcg)(struct amdgpu_device *adev, bool enable);
+	int (*get_gfx_shadow_info)(struct amdgpu_device *adev,
+				   struct amdgpu_gfx_shadow_info *shadow_info,
+				   bool skip_check);
+	enum amdgpu_gfx_partition
+			(*query_partition_mode)(struct amdgpu_device *adev);
+	int (*switch_partition_mode)(struct amdgpu_device *adev,
+				     int num_xccs_per_xcp);
+	int (*ih_node_to_logical_xcc)(struct amdgpu_device *adev, int ih_node);
+	int (*get_xccs_per_xcp)(struct amdgpu_device *adev);
 };
 
 struct sq_work {
@@ -288,6 +397,12 @@ struct amdgpu_me {
 	DECLARE_BITMAP(queue_bitmap, AMDGPU_MAX_GFX_QUEUES);
 };
 
+struct amdgpu_isolation_work {
+	struct amdgpu_device		*adev;
+	u32				xcp_id;
+	struct delayed_work		work;
+};
+
 struct amdgpu_gfx {
 	struct mutex			gpu_clock_mutex;
 	struct amdgpu_gfx_config	config;
@@ -296,7 +411,8 @@ struct amdgpu_gfx {
 	struct amdgpu_ce		ce;
 	struct amdgpu_me		me;
 	struct amdgpu_mec		mec;
-	struct amdgpu_kiq		kiq;
+	struct amdgpu_mec_bitmap	mec_bitmap[AMDGPU_MAX_GC_INSTANCES];
+	struct amdgpu_kiq		kiq[AMDGPU_MAX_GC_INSTANCES];
 	struct amdgpu_imu		imu;
 	bool				rs64_enable; /* firmware format */
 	const struct firmware		*me_fw;	/* ME firmware */
@@ -339,6 +455,7 @@ struct amdgpu_gfx {
 	struct amdgpu_irq_src		eop_irq;
 	struct amdgpu_irq_src		priv_reg_irq;
 	struct amdgpu_irq_src		priv_inst_irq;
+	struct amdgpu_irq_src		bad_op_irq;
 	struct amdgpu_irq_src		cp_ecc_error_irq;
 	struct amdgpu_irq_src		sq_irq;
 	struct amdgpu_irq_src		rlc_gc_fed_irq;
@@ -354,6 +471,8 @@ struct amdgpu_gfx {
 	/* reset mask */
 	uint32_t                        grbm_soft_reset;
 	uint32_t                        srbm_soft_reset;
+	uint32_t 			gfx_supported_reset;
+	uint32_t 			compute_supported_reset;
 
 	/* gfx off */
 	bool                            gfx_off_state;      /* true: enabled, false: disabled */
@@ -376,15 +495,62 @@ struct amdgpu_gfx {
 	struct amdgpu_ring		sw_gfx_ring[AMDGPU_MAX_SW_GFX_RINGS];
 	struct amdgpu_ring_mux          muxer;
 
-	enum amdgpu_gfx_partition	partition_mode;
-	uint32_t			num_xcd;
+	bool				cp_gfx_shadow; /* for gfx11 */
+
+	uint16_t 			xcc_mask;
 	uint32_t			num_xcc_per_xcp;
+	struct mutex			partition_mutex;
+	bool				mcbp; /* mid command buffer preemption */
+
+	/* IP reg dump */
+	uint32_t			*ip_dump_core;
+	uint32_t			*ip_dump_compute_queues;
+	uint32_t			*ip_dump_gfx_queues;
+
+	struct mutex			reset_sem_mutex;
+
+	/* cleaner shader */
+	struct amdgpu_bo		*cleaner_shader_obj;
+	unsigned int                    cleaner_shader_size;
+	u64				cleaner_shader_gpu_addr;
+	void				*cleaner_shader_cpu_ptr;
+	const void			*cleaner_shader_ptr;
+	bool				enable_cleaner_shader;
+	struct amdgpu_isolation_work	enforce_isolation[MAX_XCP];
+	/* Mutex for synchronizing KFD scheduler operations */
+	struct mutex                    userq_sch_mutex;
+	u64				userq_sch_req_count[MAX_XCP];
+	bool				userq_sch_inactive[MAX_XCP];
+	unsigned long			enforce_isolation_jiffies[MAX_XCP];
+	unsigned long			enforce_isolation_time[MAX_XCP];
+
+	atomic_t			total_submission_cnt;
+	struct delayed_work		idle_work;
+	bool				workload_profile_active;
+	struct mutex                    workload_profile_mutex;
+
+	bool				disable_kq;
+	bool				disable_uq;
+};
+
+struct amdgpu_gfx_ras_reg_entry {
+	struct amdgpu_ras_err_status_reg_entry reg_entry;
+	enum amdgpu_gfx_ras_mem_id_type mem_id_type;
+	uint32_t se_num;
+};
+
+struct amdgpu_gfx_ras_mem_id_entry {
+	const struct amdgpu_ras_memory_id_entry *mem_id_ent;
+	uint32_t size;
 };
 
+#define AMDGPU_GFX_MEMID_ENT(x) {(x), ARRAY_SIZE(x)},
+
 #define amdgpu_gfx_get_gpu_clock_counter(adev) (adev)->gfx.funcs->get_gpu_clock_counter((adev))
-#define amdgpu_gfx_select_se_sh(adev, se, sh, instance) (adev)->gfx.funcs->select_se_sh((adev), (se), (sh), (instance))
-#define amdgpu_gfx_select_me_pipe_q(adev, me, pipe, q, vmid) (adev)->gfx.funcs->select_me_pipe_q((adev), (me), (pipe), (q), (vmid))
+#define amdgpu_gfx_select_se_sh(adev, se, sh, instance, xcc_id) ((adev)->gfx.funcs->select_se_sh((adev), (se), (sh), (instance), (xcc_id)))
+#define amdgpu_gfx_select_me_pipe_q(adev, me, pipe, q, vmid, xcc_id) ((adev)->gfx.funcs->select_me_pipe_q((adev), (me), (pipe), (q), (vmid), (xcc_id)))
 #define amdgpu_gfx_init_spm_golden(adev) (adev)->gfx.funcs->init_spm_golden((adev))
+#define amdgpu_gfx_get_gfx_shadow_info(adev, si) ((adev)->gfx.funcs->get_gfx_shadow_info((adev), (si), false))
 
 /**
  * amdgpu_gfx_create_bitmask - create a bitmask
@@ -402,21 +568,21 @@ static inline u32 amdgpu_gfx_create_bitmask(u32 bit_width)
 void amdgpu_gfx_parse_disable_cu(unsigned *mask, unsigned max_se,
 				 unsigned max_sh);
 
-int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,
-			     struct amdgpu_ring *ring,
-			     struct amdgpu_irq_src *irq);
+int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, int xcc_id);
 
 void amdgpu_gfx_kiq_free_ring(struct amdgpu_ring *ring);
 
-void amdgpu_gfx_kiq_fini(struct amdgpu_device *adev);
+void amdgpu_gfx_kiq_fini(struct amdgpu_device *adev, int xcc_id);
 int amdgpu_gfx_kiq_init(struct amdgpu_device *adev,
-			unsigned hpd_size);
+			unsigned hpd_size, int xcc_id);
 
 int amdgpu_gfx_mqd_sw_init(struct amdgpu_device *adev,
-			   unsigned mqd_size);
-void amdgpu_gfx_mqd_sw_fini(struct amdgpu_device *adev);
-int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev);
-int amdgpu_gfx_enable_kcq(struct amdgpu_device *adev);
+			   unsigned mqd_size, int xcc_id);
+void amdgpu_gfx_mqd_sw_fini(struct amdgpu_device *adev, int xcc_id);
+int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev, int xcc_id);
+int amdgpu_gfx_enable_kcq(struct amdgpu_device *adev, int xcc_id);
+int amdgpu_gfx_disable_kgq(struct amdgpu_device *adev, int xcc_id);
+int amdgpu_gfx_enable_kgq(struct amdgpu_device *adev, int xcc_id);
 
 void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev);
 void amdgpu_gfx_graphics_queue_acquire(struct amdgpu_device *adev);
@@ -425,19 +591,16 @@ int amdgpu_gfx_mec_queue_to_bit(struct amdgpu_device *adev, int mec,
 				int pipe, int queue);
 void amdgpu_queue_mask_bit_to_mec_queue(struct amdgpu_device *adev, int bit,
 				 int *mec, int *pipe, int *queue);
-bool amdgpu_gfx_is_mec_queue_enabled(struct amdgpu_device *adev, int mec,
-				     int pipe, int queue);
+bool amdgpu_gfx_is_mec_queue_enabled(struct amdgpu_device *adev, int xcc_id,
+				     int mec, int pipe, int queue);
 bool amdgpu_gfx_is_high_priority_compute_queue(struct amdgpu_device *adev,
 					       struct amdgpu_ring *ring);
 bool amdgpu_gfx_is_high_priority_graphics_queue(struct amdgpu_device *adev,
 						struct amdgpu_ring *ring);
-int amdgpu_gfx_me_queue_to_bit(struct amdgpu_device *adev, int me,
-			       int pipe, int queue);
-void amdgpu_gfx_bit_to_me_queue(struct amdgpu_device *adev, int bit,
-				int *me, int *pipe, int *queue);
 bool amdgpu_gfx_is_me_queue_enabled(struct amdgpu_device *adev, int me,
 				    int pipe, int queue);
 void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable);
+void amdgpu_gfx_off_ctrl_immediate(struct amdgpu_device *adev, bool enable);
 int amdgpu_get_gfx_off_status(struct amdgpu_device *adev, uint32_t *value);
 int amdgpu_gfx_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block);
 void amdgpu_gfx_ras_fini(struct amdgpu_device *adev);
@@ -450,12 +613,59 @@ int amdgpu_gfx_process_ras_data_cb(struct amdgpu_device *adev,
 int amdgpu_gfx_cp_ecc_error_irq(struct amdgpu_device *adev,
 				  struct amdgpu_irq_src *source,
 				  struct amdgpu_iv_entry *entry);
-uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg);
-void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v);
+uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg, uint32_t xcc_id);
+void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v, uint32_t xcc_id);
+int amdgpu_kiq_hdp_flush(struct amdgpu_device *adev);
 int amdgpu_gfx_get_num_kcq(struct amdgpu_device *adev);
 void amdgpu_gfx_cp_init_microcode(struct amdgpu_device *adev, uint32_t ucode_id);
 
 int amdgpu_gfx_ras_sw_init(struct amdgpu_device *adev);
 int amdgpu_gfx_poison_consumption_handler(struct amdgpu_device *adev,
 						struct amdgpu_iv_entry *entry);
+
+bool amdgpu_gfx_is_master_xcc(struct amdgpu_device *adev, int xcc_id);
+int amdgpu_gfx_sysfs_init(struct amdgpu_device *adev);
+void amdgpu_gfx_sysfs_fini(struct amdgpu_device *adev);
+void amdgpu_gfx_ras_error_func(struct amdgpu_device *adev,
+		void *ras_error_status,
+		void (*func)(struct amdgpu_device *adev, void *ras_error_status,
+				int xcc_id));
+int amdgpu_gfx_cleaner_shader_sw_init(struct amdgpu_device *adev,
+				      unsigned int cleaner_shader_size);
+void amdgpu_gfx_cleaner_shader_sw_fini(struct amdgpu_device *adev);
+void amdgpu_gfx_cleaner_shader_init(struct amdgpu_device *adev,
+				    unsigned int cleaner_shader_size,
+				    const void *cleaner_shader_ptr);
+void amdgpu_gfx_enforce_isolation_handler(struct work_struct *work);
+void amdgpu_gfx_enforce_isolation_ring_begin_use(struct amdgpu_ring *ring);
+void amdgpu_gfx_enforce_isolation_ring_end_use(struct amdgpu_ring *ring);
+
+void amdgpu_gfx_profile_idle_work_handler(struct work_struct *work);
+void amdgpu_gfx_profile_ring_begin_use(struct amdgpu_ring *ring);
+void amdgpu_gfx_profile_ring_end_use(struct amdgpu_ring *ring);
+u32 amdgpu_gfx_csb_preamble_start(u32 *buffer);
+u32 amdgpu_gfx_csb_data_parser(struct amdgpu_device *adev, u32 *buffer, u32 count);
+void amdgpu_gfx_csb_preamble_end(u32 *buffer, u32 count);
+
+void amdgpu_debugfs_gfx_sched_mask_init(struct amdgpu_device *adev);
+void amdgpu_debugfs_compute_sched_mask_init(struct amdgpu_device *adev);
+
+static inline const char *amdgpu_gfx_compute_mode_desc(int mode)
+{
+	switch (mode) {
+	case AMDGPU_SPX_PARTITION_MODE:
+		return "SPX";
+	case AMDGPU_DPX_PARTITION_MODE:
+		return "DPX";
+	case AMDGPU_TPX_PARTITION_MODE:
+		return "TPX";
+	case AMDGPU_QPX_PARTITION_MODE:
+		return "QPX";
+	case AMDGPU_CPX_PARTITION_MODE:
+		return "CPX";
+	default:
+		return "UNKNOWN";
+	}
+}
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
index 4e2531758866..869bceb0fe2c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
@@ -32,11 +32,19 @@
 #include "amdgpu.h"
 #include "amdgpu_gmc.h"
 #include "amdgpu_ras.h"
+#include "amdgpu_reset.h"
 #include "amdgpu_xgmi.h"
 
 #include <drm/drm_drv.h>
 #include <drm/ttm/ttm_tt.h>
 
+static const u64 four_gb = 0x100000000ULL;
+
+bool amdgpu_gmc_is_pdb0_enabled(struct amdgpu_device *adev)
+{
+	return adev->gmc.xgmi.connected_to_cpu || amdgpu_virt_xgmi_migrate_enabled(adev);
+}
+
 /**
  * amdgpu_gmc_pdb0_alloc - allocate vram for pdb0
  *
@@ -51,7 +59,7 @@ int amdgpu_gmc_pdb0_alloc(struct amdgpu_device *adev)
 	struct amdgpu_bo_param bp;
 	u64 vram_size = adev->gmc.xgmi.node_segment_size * adev->gmc.xgmi.num_physical_nodes;
 	uint32_t pde0_page_shift = adev->gmc.vmid0_page_table_block_size + 21;
-	uint32_t npdes = (vram_size + (1ULL << pde0_page_shift) -1) >> pde0_page_shift;
+	uint32_t npdes = (vram_size + (1ULL << pde0_page_shift) - 1) >> pde0_page_shift;
 
 	memset(&bp, 0, sizeof(bp));
 	bp.size = PAGE_ALIGN((npdes + 1) * 8);
@@ -180,6 +188,9 @@ uint64_t amdgpu_gmc_agp_addr(struct ttm_buffer_object *bo)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev);
 
+	if (!bo->ttm)
+		return AMDGPU_BO_INVALID_OFFSET;
+
 	if (bo->ttm->num_pages != 1 || bo->ttm->caching == ttm_cached)
 		return AMDGPU_BO_INVALID_OFFSET;
 
@@ -247,10 +258,20 @@ void amdgpu_gmc_sysvm_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc
 	u64 hive_vram_end = mc->xgmi.node_segment_size * mc->xgmi.num_physical_nodes - 1;
 	mc->vram_start = mc->xgmi.node_segment_size * mc->xgmi.physical_node_id;
 	mc->vram_end = mc->vram_start + mc->xgmi.node_segment_size - 1;
-	mc->gart_start = hive_vram_end + 1;
+	/* node_segment_size may not 4GB aligned on SRIOV, align up is needed. */
+	mc->gart_start = ALIGN(hive_vram_end + 1, four_gb);
 	mc->gart_end = mc->gart_start + mc->gart_size - 1;
-	mc->fb_start = hive_vram_start;
-	mc->fb_end = hive_vram_end;
+	if (amdgpu_virt_xgmi_migrate_enabled(adev)) {
+		/* set mc->vram_start to 0 to switch the returned GPU address of
+		 * amdgpu_bo_create_reserved() from FB aperture to GART aperture.
+		 */
+		mc->vram_start = 0;
+		mc->vram_end = mc->vram_start + mc->mc_vram_size - 1;
+		mc->visible_vram_size = min(mc->visible_vram_size, mc->real_vram_size);
+	} else {
+		mc->fb_start = hive_vram_start;
+		mc->fb_end = hive_vram_end;
+	}
 	dev_info(adev->dev, "VRAM: %lluM 0x%016llX - 0x%016llX (%lluM used)\n",
 			mc->mc_vram_size >> 20, mc->vram_start,
 			mc->vram_end, mc->real_vram_size >> 20);
@@ -263,14 +284,15 @@ void amdgpu_gmc_sysvm_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc
  *
  * @adev: amdgpu device structure holding all necessary information
  * @mc: memory controller structure holding memory information
+ * @gart_placement: GART placement policy with respect to VRAM
  *
- * Function will place try to place GART before or after VRAM.
+ * Function will try to place GART before or after VRAM.
  * If GART size is bigger than space left then we ajust GART size.
  * Thus function will never fails.
  */
-void amdgpu_gmc_gart_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc)
+void amdgpu_gmc_gart_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc,
+			      enum amdgpu_gart_placement gart_placement)
 {
-	const uint64_t four_gb = 0x100000000ULL;
 	u64 size_af, size_bf;
 	/*To avoid the hole, limit the max mc address to AMDGPU_GMC_HOLE_START*/
 	u64 max_mc_address = min(adev->gmc.mc_mask, AMDGPU_GMC_HOLE_START - 1);
@@ -286,11 +308,22 @@ void amdgpu_gmc_gart_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc)
 		mc->gart_size = max(size_bf, size_af);
 	}
 
-	if ((size_bf >= mc->gart_size && size_bf < size_af) ||
-	    (size_af < mc->gart_size))
-		mc->gart_start = 0;
-	else
+	switch (gart_placement) {
+	case AMDGPU_GART_PLACEMENT_HIGH:
 		mc->gart_start = max_mc_address - mc->gart_size + 1;
+		break;
+	case AMDGPU_GART_PLACEMENT_LOW:
+		mc->gart_start = 0;
+		break;
+	case AMDGPU_GART_PLACEMENT_BEST_FIT:
+	default:
+		if ((size_bf >= mc->gart_size && size_bf < size_af) ||
+		    (size_af < mc->gart_size))
+			mc->gart_start = 0;
+		else
+			mc->gart_start = max_mc_address - mc->gart_size + 1;
+		break;
+	}
 
 	mc->gart_start &= ~(four_gb - 1);
 	mc->gart_end = mc->gart_start + mc->gart_size - 1;
@@ -315,14 +348,6 @@ void amdgpu_gmc_agp_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc)
 	const uint64_t sixteen_gb_mask = ~(sixteen_gb - 1);
 	u64 size_af, size_bf;
 
-	if (amdgpu_sriov_vf(adev)) {
-		mc->agp_start = 0xffffffffffff;
-		mc->agp_end = 0x0;
-		mc->agp_size = 0;
-
-		return;
-	}
-
 	if (mc->fb_start > mc->gart_start) {
 		size_bf = (mc->fb_start & sixteen_gb_mask) -
 			ALIGN(mc->gart_end + 1, sixteen_gb);
@@ -347,6 +372,25 @@ void amdgpu_gmc_agp_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc)
 }
 
 /**
+ * amdgpu_gmc_set_agp_default - Set the default AGP aperture value.
+ * @adev: amdgpu device structure holding all necessary information
+ * @mc: memory controller structure holding memory information
+ *
+ * To disable the AGP aperture, you need to set the start to a larger
+ * value than the end.  This function sets the default value which
+ * can then be overridden using amdgpu_gmc_agp_location() if you want
+ * to enable the AGP aperture on a specific chip.
+ *
+ */
+void amdgpu_gmc_set_agp_default(struct amdgpu_device *adev,
+				struct amdgpu_gmc *mc)
+{
+	mc->agp_start = 0xffffffffffff;
+	mc->agp_end = 0;
+	mc->agp_size = 0;
+}
+
+/**
  * amdgpu_gmc_fault_key - get hask key from vm fault address and pasid
  *
  * @addr: 48 bit physical address, page aligned (36 significant bits)
@@ -452,7 +496,10 @@ void amdgpu_gmc_filter_faults_remove(struct amdgpu_device *adev, uint64_t addr,
 	uint32_t hash;
 	uint64_t tmp;
 
-	ih = adev->irq.retry_cam_enabled ? &adev->irq.ih_soft : &adev->irq.ih1;
+	if (adev->irq.retry_cam_enabled)
+		return;
+
+	ih = &adev->irq.ih1;
 	/* Get the WPTR of the last entry in IH ring */
 	last_wptr = amdgpu_ih_get_wptr(adev, ih);
 	/* Order wptr with ring data. */
@@ -534,29 +581,42 @@ void amdgpu_gmc_ras_fini(struct amdgpu_device *adev)
 	 *                    subject to change when ring number changes
 	 * Engine 17: Gart flushes
 	 */
-#define GFXHUB_FREE_VM_INV_ENGS_BITMAP		0x1FFF3
-#define MMHUB_FREE_VM_INV_ENGS_BITMAP		0x1FFF3
+#define AMDGPU_VMHUB_INV_ENG_BITMAP		0x1FFF3
 
 int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev)
 {
 	struct amdgpu_ring *ring;
-	unsigned vm_inv_engs[AMDGPU_MAX_VMHUBS] =
-		{GFXHUB_FREE_VM_INV_ENGS_BITMAP, MMHUB_FREE_VM_INV_ENGS_BITMAP,
-		GFXHUB_FREE_VM_INV_ENGS_BITMAP};
+	unsigned vm_inv_engs[AMDGPU_MAX_VMHUBS] = {0};
 	unsigned i;
 	unsigned vmhub, inv_eng;
+	struct amdgpu_ring *shared_ring;
 
-	if (adev->enable_mes) {
+	/* init the vm inv eng for all vmhubs */
+	for_each_set_bit(i, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS) {
+		vm_inv_engs[i] = AMDGPU_VMHUB_INV_ENG_BITMAP;
 		/* reserve engine 5 for firmware */
-		for (vmhub = 0; vmhub < AMDGPU_MAX_VMHUBS; vmhub++)
-			vm_inv_engs[vmhub] &= ~(1 << 5);
+		if (adev->enable_mes)
+			vm_inv_engs[i] &= ~(1 << 5);
+		/* reserve engine 6 for uni mes */
+		if (adev->enable_uni_mes)
+			vm_inv_engs[i] &= ~(1 << 6);
+		/* reserve mmhub engine 3 for firmware */
+		if (adev->enable_umsch_mm)
+			vm_inv_engs[i] &= ~(1 << 3);
 	}
 
 	for (i = 0; i < adev->num_rings; ++i) {
 		ring = adev->rings[i];
 		vmhub = ring->vm_hub;
 
-		if (ring == &adev->mes.ring)
+		if (ring == &adev->mes.ring[0] ||
+		    ring == &adev->mes.ring[1] ||
+		    ring == &adev->umsch_mm.ring ||
+		    ring == &adev->cper.ring_buf)
+			continue;
+
+		/* Skip if the ring is a shared ring */
+		if (amdgpu_sdma_is_shared_inv_eng(adev, ring))
 			continue;
 
 		inv_eng = ffs(vm_inv_engs[vmhub]);
@@ -571,11 +631,229 @@ int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev)
 
 		dev_info(adev->dev, "ring %s uses VM inv eng %u on hub %u\n",
 			 ring->name, ring->vm_inv_eng, ring->vm_hub);
+		/* SDMA has a special packet which allows it to use the same
+		 * invalidation engine for all the rings in one instance.
+		 * Therefore, we do not allocate a separate VM invalidation engine
+		 * for SDMA page rings. Instead, they share the VM invalidation
+		 * engine with the SDMA gfx ring. This change ensures efficient
+		 * resource management and avoids the issue of insufficient VM
+		 * invalidation engines.
+		 */
+		shared_ring = amdgpu_sdma_get_shared_ring(adev, ring);
+		if (shared_ring) {
+			shared_ring->vm_inv_eng = ring->vm_inv_eng;
+			dev_info(adev->dev, "ring %s shares VM invalidation engine %u with ring %s on hub %u\n",
+					ring->name, ring->vm_inv_eng, shared_ring->name, ring->vm_hub);
+			continue;
+		}
 	}
 
 	return 0;
 }
 
+void amdgpu_gmc_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
+			      uint32_t vmhub, uint32_t flush_type)
+{
+	struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
+	struct amdgpu_vmhub *hub = &adev->vmhub[vmhub];
+	struct dma_fence *fence;
+	struct amdgpu_job *job;
+	int r;
+
+	if (!hub->sdma_invalidation_workaround || vmid ||
+	    !adev->mman.buffer_funcs_enabled || !adev->ib_pool_ready ||
+	    !ring->sched.ready) {
+		/*
+		 * A GPU reset should flush all TLBs anyway, so no need to do
+		 * this while one is ongoing.
+		 */
+		if (!down_read_trylock(&adev->reset_domain->sem))
+			return;
+
+		if (adev->gmc.flush_tlb_needs_extra_type_2)
+			adev->gmc.gmc_funcs->flush_gpu_tlb(adev, vmid,
+							   vmhub, 2);
+
+		if (adev->gmc.flush_tlb_needs_extra_type_0 && flush_type == 2)
+			adev->gmc.gmc_funcs->flush_gpu_tlb(adev, vmid,
+							   vmhub, 0);
+
+		adev->gmc.gmc_funcs->flush_gpu_tlb(adev, vmid, vmhub,
+						   flush_type);
+		up_read(&adev->reset_domain->sem);
+		return;
+	}
+
+	/* The SDMA on Navi 1x has a bug which can theoretically result in memory
+	 * corruption if an invalidation happens at the same time as an VA
+	 * translation. Avoid this by doing the invalidation from the SDMA
+	 * itself at least for GART.
+	 */
+	mutex_lock(&adev->mman.gtt_window_lock);
+	r = amdgpu_job_alloc_with_ib(ring->adev, &adev->mman.high_pr,
+				     AMDGPU_FENCE_OWNER_UNDEFINED,
+				     16 * 4, AMDGPU_IB_POOL_IMMEDIATE,
+				     &job, AMDGPU_KERNEL_JOB_ID_FLUSH_GPU_TLB);
+	if (r)
+		goto error_alloc;
+
+	job->vm_pd_addr = amdgpu_gmc_pd_addr(adev->gart.bo);
+	job->vm_needs_flush = true;
+	job->ibs->ptr[job->ibs->length_dw++] = ring->funcs->nop;
+	amdgpu_ring_pad_ib(ring, &job->ibs[0]);
+	fence = amdgpu_job_submit(job);
+	mutex_unlock(&adev->mman.gtt_window_lock);
+
+	dma_fence_wait(fence, false);
+	dma_fence_put(fence);
+
+	return;
+
+error_alloc:
+	mutex_unlock(&adev->mman.gtt_window_lock);
+	dev_err(adev->dev, "Error flushing GPU TLB using the SDMA (%d)!\n", r);
+}
+
+int amdgpu_gmc_flush_gpu_tlb_pasid(struct amdgpu_device *adev, uint16_t pasid,
+				   uint32_t flush_type, bool all_hub,
+				   uint32_t inst)
+{
+	struct amdgpu_ring *ring = &adev->gfx.kiq[inst].ring;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[inst];
+	unsigned int ndw;
+	int r, cnt = 0;
+	uint32_t seq;
+
+	/*
+	 * A GPU reset should flush all TLBs anyway, so no need to do
+	 * this while one is ongoing.
+	 */
+	if (!down_read_trylock(&adev->reset_domain->sem))
+		return 0;
+
+	if (!adev->gmc.flush_pasid_uses_kiq || !ring->sched.ready) {
+		if (adev->gmc.flush_tlb_needs_extra_type_2)
+			adev->gmc.gmc_funcs->flush_gpu_tlb_pasid(adev, pasid,
+								 2, all_hub,
+								 inst);
+
+		if (adev->gmc.flush_tlb_needs_extra_type_0 && flush_type == 2)
+			adev->gmc.gmc_funcs->flush_gpu_tlb_pasid(adev, pasid,
+								 0, all_hub,
+								 inst);
+
+		adev->gmc.gmc_funcs->flush_gpu_tlb_pasid(adev, pasid,
+							 flush_type, all_hub,
+							 inst);
+		r = 0;
+	} else {
+		/* 2 dwords flush + 8 dwords fence */
+		ndw = kiq->pmf->invalidate_tlbs_size + 8;
+
+		if (adev->gmc.flush_tlb_needs_extra_type_2)
+			ndw += kiq->pmf->invalidate_tlbs_size;
+
+		if (adev->gmc.flush_tlb_needs_extra_type_0)
+			ndw += kiq->pmf->invalidate_tlbs_size;
+
+		spin_lock(&adev->gfx.kiq[inst].ring_lock);
+		r = amdgpu_ring_alloc(ring, ndw);
+		if (r) {
+			spin_unlock(&adev->gfx.kiq[inst].ring_lock);
+			goto error_unlock_reset;
+		}
+		if (adev->gmc.flush_tlb_needs_extra_type_2)
+			kiq->pmf->kiq_invalidate_tlbs(ring, pasid, 2, all_hub);
+
+		if (flush_type == 2 && adev->gmc.flush_tlb_needs_extra_type_0)
+			kiq->pmf->kiq_invalidate_tlbs(ring, pasid, 0, all_hub);
+
+		kiq->pmf->kiq_invalidate_tlbs(ring, pasid, flush_type, all_hub);
+		r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
+		if (r) {
+			amdgpu_ring_undo(ring);
+			spin_unlock(&adev->gfx.kiq[inst].ring_lock);
+			goto error_unlock_reset;
+		}
+
+		amdgpu_ring_commit(ring);
+		spin_unlock(&adev->gfx.kiq[inst].ring_lock);
+
+		r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
+
+		might_sleep();
+		while (r < 1 && cnt++ < MAX_KIQ_REG_TRY &&
+		       !amdgpu_reset_pending(adev->reset_domain)) {
+			msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
+			r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
+		}
+
+		if (cnt > MAX_KIQ_REG_TRY) {
+			dev_err(adev->dev, "timeout waiting for kiq fence\n");
+			r = -ETIME;
+		} else
+			r = 0;
+	}
+
+error_unlock_reset:
+	up_read(&adev->reset_domain->sem);
+	return r;
+}
+
+void amdgpu_gmc_fw_reg_write_reg_wait(struct amdgpu_device *adev,
+				      uint32_t reg0, uint32_t reg1,
+				      uint32_t ref, uint32_t mask,
+				      uint32_t xcc_inst)
+{
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_inst];
+	struct amdgpu_ring *ring = &kiq->ring;
+	signed long r, cnt = 0;
+	unsigned long flags;
+	uint32_t seq;
+
+	if (adev->mes.ring[0].sched.ready) {
+		amdgpu_mes_reg_write_reg_wait(adev, reg0, reg1,
+					      ref, mask);
+		return;
+	}
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+	amdgpu_ring_alloc(ring, 32);
+	amdgpu_ring_emit_reg_write_reg_wait(ring, reg0, reg1,
+					    ref, mask);
+	r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
+	if (r)
+		goto failed_undo;
+
+	amdgpu_ring_commit(ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+
+	r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
+
+	/* don't wait anymore for IRQ context */
+	if (r < 1 && in_interrupt())
+		goto failed_kiq;
+
+	might_sleep();
+	while (r < 1 && cnt++ < MAX_KIQ_REG_TRY &&
+	       !amdgpu_reset_pending(adev->reset_domain)) {
+
+		msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
+		r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
+	}
+
+	if (cnt > MAX_KIQ_REG_TRY)
+		goto failed_kiq;
+
+	return;
+
+failed_undo:
+	amdgpu_ring_undo(ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+failed_kiq:
+	dev_err(adev->dev, "failed to write reg %x wait reg %x\n", reg0, reg1);
+}
+
 /**
  * amdgpu_gmc_tmz_set -- check and set if a device supports TMZ
  * @adev: amdgpu_device pointer
@@ -585,7 +863,7 @@ int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev)
  */
 void amdgpu_gmc_tmz_set(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	/* RAVEN */
 	case IP_VERSION(9, 2, 2):
 	case IP_VERSION(9, 1, 0):
@@ -593,6 +871,8 @@ void amdgpu_gmc_tmz_set(struct amdgpu_device *adev)
 	case IP_VERSION(9, 3, 0):
 	/* GC 10.3.7 */
 	case IP_VERSION(10, 3, 7):
+	/* GC 11.0.1 */
+	case IP_VERSION(11, 0, 1):
 		if (amdgpu_tmz == 0) {
 			adev->gmc.tmz_enabled = false;
 			dev_info(adev->dev,
@@ -616,8 +896,11 @@ void amdgpu_gmc_tmz_set(struct amdgpu_device *adev)
 	case IP_VERSION(10, 3, 1):
 	/* YELLOW_CARP*/
 	case IP_VERSION(10, 3, 3):
-	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
 		/* Don't enable it by default yet.
 		 */
 		if (amdgpu_tmz < 1) {
@@ -648,16 +931,20 @@ void amdgpu_gmc_tmz_set(struct amdgpu_device *adev)
 void amdgpu_gmc_noretry_set(struct amdgpu_device *adev)
 {
 	struct amdgpu_gmc *gmc = &adev->gmc;
-	uint32_t gc_ver = adev->ip_versions[GC_HWIP][0];
+	uint32_t gc_ver = amdgpu_ip_version(adev, GC_HWIP, 0);
 	bool noretry_default = (gc_ver == IP_VERSION(9, 0, 1) ||
-				gc_ver == IP_VERSION(9, 3, 0) ||
 				gc_ver == IP_VERSION(9, 4, 0) ||
 				gc_ver == IP_VERSION(9, 4, 1) ||
 				gc_ver == IP_VERSION(9, 4, 2) ||
 				gc_ver == IP_VERSION(9, 4, 3) ||
+				gc_ver == IP_VERSION(9, 4, 4) ||
+				gc_ver == IP_VERSION(9, 5, 0) ||
 				gc_ver >= IP_VERSION(10, 3, 0));
 
-	gmc->noretry = (amdgpu_noretry == -1) ? noretry_default : amdgpu_noretry;
+	if (!amdgpu_sriov_xnack_support(adev))
+		gmc->noretry = 1;
+	else
+		gmc->noretry = (amdgpu_noretry == -1) ? noretry_default : amdgpu_noretry;
 }
 
 void amdgpu_gmc_set_vm_fault_masks(struct amdgpu_device *adev, int hub_type,
@@ -670,7 +957,7 @@ void amdgpu_gmc_set_vm_fault_masks(struct amdgpu_device *adev, int hub_type,
 	for (i = 0; i < 16; i++) {
 		reg = hub->vm_context0_cntl + hub->ctx_distance * i;
 
-		tmp = (hub_type == AMDGPU_GFXHUB_0) ?
+		tmp = (hub_type == AMDGPU_GFXHUB(0)) ?
 			RREG32_SOC15_IP(GC, reg) :
 			RREG32_SOC15_IP(MMHUB, reg);
 
@@ -679,7 +966,7 @@ void amdgpu_gmc_set_vm_fault_masks(struct amdgpu_device *adev, int hub_type,
 		else
 			tmp &= ~hub->vm_cntx_cntl_vm_fault;
 
-		(hub_type == AMDGPU_GFXHUB_0) ?
+		(hub_type == AMDGPU_GFXHUB(0)) ?
 			WREG32_SOC15_IP(GC, reg, tmp) :
 			WREG32_SOC15_IP(MMHUB, reg, tmp);
 	}
@@ -721,12 +1008,6 @@ void amdgpu_gmc_get_vbios_allocations(struct amdgpu_device *adev)
 	case CHIP_RENOIR:
 		adev->mman.keep_stolen_vga_memory = true;
 		break;
-	case CHIP_YELLOW_CARP:
-		if (amdgpu_discovery == 0) {
-			adev->mman.stolen_reserved_offset = 0x1ffb0000;
-			adev->mman.stolen_reserved_size = 64 * PAGE_SIZE;
-		}
-		break;
 	default:
 		adev->mman.keep_stolen_vga_memory = false;
 		break;
@@ -779,9 +1060,7 @@ void amdgpu_gmc_init_pdb0(struct amdgpu_device *adev)
 	 */
 	u64 vram_size = adev->gmc.xgmi.node_segment_size * adev->gmc.xgmi.num_physical_nodes;
 	u64 pde0_page_size = (1ULL<<adev->gmc.vmid0_page_table_block_size)<<21;
-	u64 vram_addr = adev->vm_manager.vram_base_offset -
-		adev->gmc.xgmi.physical_node_id * adev->gmc.xgmi.node_segment_size;
-	u64 vram_end = vram_addr + vram_size;
+	u64 vram_addr, vram_end;
 	u64 gart_ptb_gpu_pa = amdgpu_gmc_vram_pa(adev, adev->gart.bo);
 	int idx;
 
@@ -792,7 +1071,12 @@ void amdgpu_gmc_init_pdb0(struct amdgpu_device *adev)
 	flags |= AMDGPU_PTE_WRITEABLE;
 	flags |= AMDGPU_PTE_SNOOPED;
 	flags |= AMDGPU_PTE_FRAG((adev->gmc.vmid0_page_table_block_size + 9*1));
-	flags |= AMDGPU_PDE_PTE;
+	flags |= AMDGPU_PDE_PTE_FLAG(adev);
+
+	vram_addr = adev->vm_manager.vram_base_offset;
+	if (!amdgpu_virt_xgmi_migrate_enabled(adev))
+		vram_addr -= adev->gmc.xgmi.physical_node_id * adev->gmc.xgmi.node_segment_size;
+	vram_end = vram_addr + vram_size;
 
 	/* The first n PDE0 entries are used as PTE,
 	 * pointing to vram
@@ -805,7 +1089,7 @@ void amdgpu_gmc_init_pdb0(struct amdgpu_device *adev)
 	 * pointing to a 4K system page
 	 */
 	flags = AMDGPU_PTE_VALID;
-	flags |= AMDGPU_PDE_BFS(0) | AMDGPU_PTE_SNOOPED;
+	flags |= AMDGPU_PTE_SNOOPED | AMDGPU_PDE_BFS_FLAG(adev, 0);
 	/* Requires gart_ptb_gpu_pa to be 4K aligned */
 	amdgpu_gmc_set_pte_pde(adev, adev->gmc.ptr_pdb0, i, gart_ptb_gpu_pa, flags);
 	drm_dev_exit(idx);
@@ -835,18 +1119,6 @@ uint64_t amdgpu_gmc_vram_pa(struct amdgpu_device *adev, struct amdgpu_bo *bo)
 	return amdgpu_gmc_vram_mc2pa(adev, amdgpu_bo_gpu_offset(bo));
 }
 
-/**
- * amdgpu_gmc_vram_cpu_pa - calculate vram buffer object's physical address
- * from CPU's view
- *
- * @adev: amdgpu_device pointer
- * @bo: amdgpu buffer object
- */
-uint64_t amdgpu_gmc_vram_cpu_pa(struct amdgpu_device *adev, struct amdgpu_bo *bo)
-{
-	return amdgpu_bo_gpu_offset(bo) - adev->gmc.vram_start + adev->gmc.aper_base;
-}
-
 int amdgpu_gmc_vram_checking(struct amdgpu_device *adev)
 {
 	struct amdgpu_bo *vram_bo = NULL;
@@ -876,19 +1148,535 @@ int amdgpu_gmc_vram_checking(struct amdgpu_device *adev)
 	 * seconds, so here, we just pick up three parts for emulation.
 	 */
 	ret = memcmp(vram_ptr, cptr, 10);
-	if (ret)
-		return ret;
+	if (ret) {
+		ret = -EIO;
+		goto release_buffer;
+	}
 
 	ret = memcmp(vram_ptr + (size / 2), cptr, 10);
-	if (ret)
-		return ret;
+	if (ret) {
+		ret = -EIO;
+		goto release_buffer;
+	}
 
 	ret = memcmp(vram_ptr + size - 10, cptr, 10);
-	if (ret)
-		return ret;
+	if (ret) {
+		ret = -EIO;
+		goto release_buffer;
+	}
 
+release_buffer:
 	amdgpu_bo_free_kernel(&vram_bo, &vram_gpu,
 			&vram_ptr);
 
+	return ret;
+}
+
+static const char *nps_desc[] = {
+	[AMDGPU_NPS1_PARTITION_MODE] = "NPS1",
+	[AMDGPU_NPS2_PARTITION_MODE] = "NPS2",
+	[AMDGPU_NPS3_PARTITION_MODE] = "NPS3",
+	[AMDGPU_NPS4_PARTITION_MODE] = "NPS4",
+	[AMDGPU_NPS6_PARTITION_MODE] = "NPS6",
+	[AMDGPU_NPS8_PARTITION_MODE] = "NPS8",
+};
+
+static ssize_t available_memory_partition_show(struct device *dev,
+					       struct device_attribute *addr,
+					       char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	int size = 0, mode;
+	char *sep = "";
+
+	for_each_inst(mode, adev->gmc.supported_nps_modes) {
+		size += sysfs_emit_at(buf, size, "%s%s", sep, nps_desc[mode]);
+		sep = ", ";
+	}
+	size += sysfs_emit_at(buf, size, "\n");
+
+	return size;
+}
+
+static ssize_t current_memory_partition_store(struct device *dev,
+					      struct device_attribute *attr,
+					      const char *buf, size_t count)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	enum amdgpu_memory_partition mode;
+	struct amdgpu_hive_info *hive;
+	int i;
+
+	mode = UNKNOWN_MEMORY_PARTITION_MODE;
+	for_each_inst(i, adev->gmc.supported_nps_modes) {
+		if (!strncasecmp(nps_desc[i], buf, strlen(nps_desc[i]))) {
+			mode = i;
+			break;
+		}
+	}
+
+	if (mode == UNKNOWN_MEMORY_PARTITION_MODE)
+		return -EINVAL;
+
+	if (mode == adev->gmc.gmc_funcs->query_mem_partition_mode(adev)) {
+		dev_info(
+			adev->dev,
+			"requested NPS mode is same as current NPS mode, skipping\n");
+		return count;
+	}
+
+	/* If device is part of hive, all devices in the hive should request the
+	 * same mode. Hence store the requested mode in hive.
+	 */
+	hive = amdgpu_get_xgmi_hive(adev);
+	if (hive) {
+		atomic_set(&hive->requested_nps_mode, mode);
+		amdgpu_put_xgmi_hive(hive);
+	} else {
+		adev->gmc.requested_nps_mode = mode;
+	}
+
+	dev_info(
+		adev->dev,
+		"NPS mode change requested, please remove and reload the driver\n");
+
+	return count;
+}
+
+static ssize_t current_memory_partition_show(
+	struct device *dev, struct device_attribute *addr, char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	enum amdgpu_memory_partition mode;
+
+	/* Only minimal precaution taken to reject requests while in reset */
+	if (amdgpu_in_reset(adev))
+		return -EPERM;
+
+	mode = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
+	if ((mode >= ARRAY_SIZE(nps_desc)) ||
+	    (BIT(mode) & AMDGPU_ALL_NPS_MASK) != BIT(mode))
+		return sysfs_emit(buf, "UNKNOWN\n");
+
+	return sysfs_emit(buf, "%s\n", nps_desc[mode]);
+}
+
+static DEVICE_ATTR_RW(current_memory_partition);
+static DEVICE_ATTR_RO(available_memory_partition);
+
+int amdgpu_gmc_sysfs_init(struct amdgpu_device *adev)
+{
+	bool nps_switch_support;
+	int r = 0;
+
+	if (!adev->gmc.gmc_funcs->query_mem_partition_mode)
+		return 0;
+
+	nps_switch_support = (hweight32(adev->gmc.supported_nps_modes &
+					AMDGPU_ALL_NPS_MASK) > 1);
+	if (!nps_switch_support)
+		dev_attr_current_memory_partition.attr.mode &=
+			~(S_IWUSR | S_IWGRP | S_IWOTH);
+	else
+		r = device_create_file(adev->dev,
+				       &dev_attr_available_memory_partition);
+
+	if (r)
+		return r;
+
+	return device_create_file(adev->dev,
+				  &dev_attr_current_memory_partition);
+}
+
+void amdgpu_gmc_sysfs_fini(struct amdgpu_device *adev)
+{
+	if (!adev->gmc.gmc_funcs->query_mem_partition_mode)
+		return;
+
+	device_remove_file(adev->dev, &dev_attr_current_memory_partition);
+	device_remove_file(adev->dev, &dev_attr_available_memory_partition);
+}
+
+int amdgpu_gmc_get_nps_memranges(struct amdgpu_device *adev,
+				 struct amdgpu_mem_partition_info *mem_ranges,
+				 uint8_t *exp_ranges)
+{
+	struct amdgpu_gmc_memrange *ranges;
+	int range_cnt, ret, i, j;
+	uint32_t nps_type;
+	bool refresh;
+
+	if (!mem_ranges || !exp_ranges)
+		return -EINVAL;
+
+	refresh = (adev->init_lvl->level != AMDGPU_INIT_LEVEL_MINIMAL_XGMI) &&
+		  (adev->gmc.reset_flags & AMDGPU_GMC_INIT_RESET_NPS);
+	ret = amdgpu_discovery_get_nps_info(adev, &nps_type, &ranges,
+					    &range_cnt, refresh);
+
+	if (ret)
+		return ret;
+
+	/* TODO: For now, expect ranges and partition count to be the same.
+	 * Adjust if there are holes expected in any NPS domain.
+	 */
+	if (*exp_ranges && (range_cnt != *exp_ranges)) {
+		dev_warn(
+			adev->dev,
+			"NPS config mismatch - expected ranges: %d discovery - nps mode: %d, nps ranges: %d",
+			*exp_ranges, nps_type, range_cnt);
+		ret = -EINVAL;
+		goto err;
+	}
+
+	for (i = 0; i < range_cnt; ++i) {
+		if (ranges[i].base_address >= ranges[i].limit_address) {
+			dev_warn(
+				adev->dev,
+				"Invalid NPS range - nps mode: %d, range[%d]: base: %llx limit: %llx",
+				nps_type, i, ranges[i].base_address,
+				ranges[i].limit_address);
+			ret = -EINVAL;
+			goto err;
+		}
+
+		/* Check for overlaps, not expecting any now */
+		for (j = i - 1; j >= 0; j--) {
+			if (max(ranges[j].base_address,
+				ranges[i].base_address) <=
+			    min(ranges[j].limit_address,
+				ranges[i].limit_address)) {
+				dev_warn(
+					adev->dev,
+					"overlapping ranges detected [ %llx - %llx ] | [%llx - %llx]",
+					ranges[j].base_address,
+					ranges[j].limit_address,
+					ranges[i].base_address,
+					ranges[i].limit_address);
+				ret = -EINVAL;
+				goto err;
+			}
+		}
+
+		mem_ranges[i].range.fpfn =
+			(ranges[i].base_address -
+			 adev->vm_manager.vram_base_offset) >>
+			AMDGPU_GPU_PAGE_SHIFT;
+		mem_ranges[i].range.lpfn =
+			(ranges[i].limit_address -
+			 adev->vm_manager.vram_base_offset) >>
+			AMDGPU_GPU_PAGE_SHIFT;
+		mem_ranges[i].size =
+			ranges[i].limit_address - ranges[i].base_address + 1;
+	}
+
+	if (!*exp_ranges)
+		*exp_ranges = range_cnt;
+err:
+	kfree(ranges);
+
+	return ret;
+}
+
+int amdgpu_gmc_request_memory_partition(struct amdgpu_device *adev,
+					int nps_mode)
+{
+	/* Not supported on VF devices and APUs */
+	if (amdgpu_sriov_vf(adev) || (adev->flags & AMD_IS_APU))
+		return -EOPNOTSUPP;
+
+	if (!adev->psp.funcs) {
+		dev_err(adev->dev,
+			"PSP interface not available for nps mode change request");
+		return -EINVAL;
+	}
+
+	return psp_memory_partition(&adev->psp, nps_mode);
+}
+
+static inline bool amdgpu_gmc_need_nps_switch_req(struct amdgpu_device *adev,
+						  int req_nps_mode,
+						  int cur_nps_mode)
+{
+	return (((BIT(req_nps_mode) & adev->gmc.supported_nps_modes) ==
+			BIT(req_nps_mode)) &&
+		req_nps_mode != cur_nps_mode);
+}
+
+void amdgpu_gmc_prepare_nps_mode_change(struct amdgpu_device *adev)
+{
+	int req_nps_mode, cur_nps_mode, r;
+	struct amdgpu_hive_info *hive;
+
+	if (amdgpu_sriov_vf(adev) || !adev->gmc.supported_nps_modes ||
+	    !adev->gmc.gmc_funcs->request_mem_partition_mode)
+		return;
+
+	cur_nps_mode = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
+	hive = amdgpu_get_xgmi_hive(adev);
+	if (hive) {
+		req_nps_mode = atomic_read(&hive->requested_nps_mode);
+		if (!amdgpu_gmc_need_nps_switch_req(adev, req_nps_mode,
+						    cur_nps_mode)) {
+			amdgpu_put_xgmi_hive(hive);
+			return;
+		}
+		r = amdgpu_xgmi_request_nps_change(adev, hive, req_nps_mode);
+		amdgpu_put_xgmi_hive(hive);
+		goto out;
+	}
+
+	req_nps_mode = adev->gmc.requested_nps_mode;
+	if (!amdgpu_gmc_need_nps_switch_req(adev, req_nps_mode, cur_nps_mode))
+		return;
+
+	/* even if this fails, we should let driver unload w/o blocking */
+	r = adev->gmc.gmc_funcs->request_mem_partition_mode(adev, req_nps_mode);
+out:
+	if (r)
+		dev_err(adev->dev, "NPS mode change request failed\n");
+	else
+		dev_info(
+			adev->dev,
+			"NPS mode change request done, reload driver to complete the change\n");
+}
+
+bool amdgpu_gmc_need_reset_on_init(struct amdgpu_device *adev)
+{
+	if (adev->gmc.gmc_funcs->need_reset_on_init)
+		return adev->gmc.gmc_funcs->need_reset_on_init(adev);
+
+	return false;
+}
+
+enum amdgpu_memory_partition
+amdgpu_gmc_get_vf_memory_partition(struct amdgpu_device *adev)
+{
+	switch (adev->gmc.num_mem_partitions) {
+	case 0:
+		return UNKNOWN_MEMORY_PARTITION_MODE;
+	case 1:
+		return AMDGPU_NPS1_PARTITION_MODE;
+	case 2:
+		return AMDGPU_NPS2_PARTITION_MODE;
+	case 4:
+		return AMDGPU_NPS4_PARTITION_MODE;
+	case 8:
+		return AMDGPU_NPS8_PARTITION_MODE;
+	default:
+		return AMDGPU_NPS1_PARTITION_MODE;
+	}
+}
+
+enum amdgpu_memory_partition
+amdgpu_gmc_get_memory_partition(struct amdgpu_device *adev, u32 *supp_modes)
+{
+	enum amdgpu_memory_partition mode = UNKNOWN_MEMORY_PARTITION_MODE;
+
+	if (adev->nbio.funcs &&
+	    adev->nbio.funcs->get_memory_partition_mode)
+		mode = adev->nbio.funcs->get_memory_partition_mode(adev,
+								   supp_modes);
+	else
+		dev_warn(adev->dev, "memory partition mode query is not supported\n");
+
+	return mode;
+}
+
+enum amdgpu_memory_partition
+amdgpu_gmc_query_memory_partition(struct amdgpu_device *adev)
+{
+	if (amdgpu_sriov_vf(adev))
+		return amdgpu_gmc_get_vf_memory_partition(adev);
+	else
+		return amdgpu_gmc_get_memory_partition(adev, NULL);
+}
+
+static bool amdgpu_gmc_validate_partition_info(struct amdgpu_device *adev)
+{
+	enum amdgpu_memory_partition mode;
+	u32 supp_modes;
+	bool valid;
+
+	mode = amdgpu_gmc_get_memory_partition(adev, &supp_modes);
+
+	/* Mode detected by hardware not present in supported modes */
+	if ((mode != UNKNOWN_MEMORY_PARTITION_MODE) &&
+	    !(BIT(mode - 1) & supp_modes))
+		return false;
+
+	switch (mode) {
+	case UNKNOWN_MEMORY_PARTITION_MODE:
+	case AMDGPU_NPS1_PARTITION_MODE:
+		valid = (adev->gmc.num_mem_partitions == 1);
+		break;
+	case AMDGPU_NPS2_PARTITION_MODE:
+		valid = (adev->gmc.num_mem_partitions == 2);
+		break;
+	case AMDGPU_NPS4_PARTITION_MODE:
+		valid = (adev->gmc.num_mem_partitions == 3 ||
+			 adev->gmc.num_mem_partitions == 4);
+		break;
+	case AMDGPU_NPS8_PARTITION_MODE:
+		valid = (adev->gmc.num_mem_partitions == 8);
+		break;
+	default:
+		valid = false;
+	}
+
+	return valid;
+}
+
+static bool amdgpu_gmc_is_node_present(int *node_ids, int num_ids, int nid)
+{
+	int i;
+
+	/* Check if node with id 'nid' is present in 'node_ids' array */
+	for (i = 0; i < num_ids; ++i)
+		if (node_ids[i] == nid)
+			return true;
+
+	return false;
+}
+
+static void
+amdgpu_gmc_init_acpi_mem_ranges(struct amdgpu_device *adev,
+				struct amdgpu_mem_partition_info *mem_ranges)
+{
+	struct amdgpu_numa_info numa_info;
+	int node_ids[AMDGPU_MAX_MEM_RANGES];
+	int num_ranges = 0, ret;
+	int num_xcc, xcc_id;
+	uint32_t xcc_mask;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	xcc_mask = (1U << num_xcc) - 1;
+
+	for_each_inst(xcc_id, xcc_mask)	{
+		ret = amdgpu_acpi_get_mem_info(adev, xcc_id, &numa_info);
+		if (ret)
+			continue;
+
+		if (numa_info.nid == NUMA_NO_NODE) {
+			mem_ranges[0].size = numa_info.size;
+			mem_ranges[0].numa.node = numa_info.nid;
+			num_ranges = 1;
+			break;
+		}
+
+		if (amdgpu_gmc_is_node_present(node_ids, num_ranges,
+					     numa_info.nid))
+			continue;
+
+		node_ids[num_ranges] = numa_info.nid;
+		mem_ranges[num_ranges].numa.node = numa_info.nid;
+		mem_ranges[num_ranges].size = numa_info.size;
+		++num_ranges;
+	}
+
+	adev->gmc.num_mem_partitions = num_ranges;
+}
+
+void amdgpu_gmc_init_sw_mem_ranges(struct amdgpu_device *adev,
+				   struct amdgpu_mem_partition_info *mem_ranges)
+{
+	enum amdgpu_memory_partition mode;
+	u32 start_addr = 0, size;
+	int i, r, l;
+
+	mode = amdgpu_gmc_query_memory_partition(adev);
+
+	switch (mode) {
+	case UNKNOWN_MEMORY_PARTITION_MODE:
+		adev->gmc.num_mem_partitions = 0;
+		break;
+	case AMDGPU_NPS1_PARTITION_MODE:
+		adev->gmc.num_mem_partitions = 1;
+		break;
+	case AMDGPU_NPS2_PARTITION_MODE:
+		adev->gmc.num_mem_partitions = 2;
+		break;
+	case AMDGPU_NPS4_PARTITION_MODE:
+		if (adev->flags & AMD_IS_APU)
+			adev->gmc.num_mem_partitions = 3;
+		else
+			adev->gmc.num_mem_partitions = 4;
+		break;
+	case AMDGPU_NPS8_PARTITION_MODE:
+		adev->gmc.num_mem_partitions = 8;
+		break;
+	default:
+		adev->gmc.num_mem_partitions = 1;
+		break;
+	}
+
+	/* Use NPS range info, if populated */
+	r = amdgpu_gmc_get_nps_memranges(adev, mem_ranges,
+					 &adev->gmc.num_mem_partitions);
+	if (!r) {
+		l = 0;
+		for (i = 1; i < adev->gmc.num_mem_partitions; ++i) {
+			if (mem_ranges[i].range.lpfn >
+			    mem_ranges[i - 1].range.lpfn)
+				l = i;
+		}
+
+	} else {
+		if (!adev->gmc.num_mem_partitions) {
+			dev_warn(adev->dev,
+				 "Not able to detect NPS mode, fall back to NPS1\n");
+			adev->gmc.num_mem_partitions = 1;
+		}
+		/* Fallback to sw based calculation */
+		size = (adev->gmc.real_vram_size + SZ_16M) >> AMDGPU_GPU_PAGE_SHIFT;
+		size /= adev->gmc.num_mem_partitions;
+
+		for (i = 0; i < adev->gmc.num_mem_partitions; ++i) {
+			mem_ranges[i].range.fpfn = start_addr;
+			mem_ranges[i].size =
+				((u64)size << AMDGPU_GPU_PAGE_SHIFT);
+			mem_ranges[i].range.lpfn = start_addr + size - 1;
+			start_addr += size;
+		}
+
+		l = adev->gmc.num_mem_partitions - 1;
+	}
+
+	/* Adjust the last one */
+	mem_ranges[l].range.lpfn =
+		(adev->gmc.real_vram_size >> AMDGPU_GPU_PAGE_SHIFT) - 1;
+	mem_ranges[l].size =
+		adev->gmc.real_vram_size -
+		((u64)mem_ranges[l].range.fpfn << AMDGPU_GPU_PAGE_SHIFT);
+}
+
+int amdgpu_gmc_init_mem_ranges(struct amdgpu_device *adev)
+{
+	bool valid;
+
+	adev->gmc.mem_partitions = kcalloc(AMDGPU_MAX_MEM_RANGES,
+					   sizeof(struct amdgpu_mem_partition_info),
+					   GFP_KERNEL);
+	if (!adev->gmc.mem_partitions)
+		return -ENOMEM;
+
+	if (adev->gmc.is_app_apu)
+		amdgpu_gmc_init_acpi_mem_ranges(adev, adev->gmc.mem_partitions);
+	else
+		amdgpu_gmc_init_sw_mem_ranges(adev, adev->gmc.mem_partitions);
+
+	if (amdgpu_sriov_vf(adev))
+		valid = true;
+	else
+		valid = amdgpu_gmc_validate_partition_info(adev);
+	if (!valid) {
+		/* TODO: handle invalid case */
+		dev_warn(adev->dev,
+			 "Mem ranges not matching with hardware config\n");
+	}
+
 	return 0;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h
index 6d105d7fb98b..727342689d4b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h
@@ -29,6 +29,7 @@
 #include <linux/types.h>
 
 #include "amdgpu_irq.h"
+#include "amdgpu_xgmi.h"
 #include "amdgpu_ras.h"
 
 /* VA hole for 48bit addresses on Vega10 */
@@ -61,8 +62,35 @@
  */
 #define AMDGPU_GMC_FAULT_TIMEOUT	5000ULL
 
+/* XNACK flags */
+#define AMDGPU_GMC_XNACK_FLAG_CHAIN BIT(0)
+
 struct firmware;
 
+enum amdgpu_memory_partition {
+	UNKNOWN_MEMORY_PARTITION_MODE = 0,
+	AMDGPU_NPS1_PARTITION_MODE = 1,
+	AMDGPU_NPS2_PARTITION_MODE = 2,
+	AMDGPU_NPS3_PARTITION_MODE = 3,
+	AMDGPU_NPS4_PARTITION_MODE = 4,
+	AMDGPU_NPS6_PARTITION_MODE = 6,
+	AMDGPU_NPS8_PARTITION_MODE = 8,
+};
+
+#define AMDGPU_ALL_NPS_MASK                                                  \
+	(BIT(AMDGPU_NPS1_PARTITION_MODE) | BIT(AMDGPU_NPS2_PARTITION_MODE) | \
+	 BIT(AMDGPU_NPS3_PARTITION_MODE) | BIT(AMDGPU_NPS4_PARTITION_MODE) | \
+	 BIT(AMDGPU_NPS6_PARTITION_MODE) | BIT(AMDGPU_NPS8_PARTITION_MODE))
+
+#define AMDGPU_GMC_INIT_RESET_NPS  BIT(0)
+
+#define AMDGPU_MAX_MEM_RANGES 8
+
+#define AMDGPU_GMC9_FAULT_SOURCE_DATA_RETRY 0x80
+#define AMDGPU_GMC9_FAULT_SOURCE_DATA_READ  0x40
+#define AMDGPU_GMC9_FAULT_SOURCE_DATA_WRITE 0x20
+#define AMDGPU_GMC9_FAULT_SOURCE_DATA_EXE   0x10
+
 /*
  * GMC page fault information
  */
@@ -107,6 +135,8 @@ struct amdgpu_vmhub {
 
 	uint32_t	vm_contexts_disable;
 
+	bool		sdma_invalidation_workaround;
+
 	const struct amdgpu_vmhub_funcs *vmhub_funcs;
 };
 
@@ -118,8 +148,9 @@ struct amdgpu_gmc_funcs {
 	void (*flush_gpu_tlb)(struct amdgpu_device *adev, uint32_t vmid,
 				uint32_t vmhub, uint32_t flush_type);
 	/* flush the vm tlb via pasid */
-	int (*flush_gpu_tlb_pasid)(struct amdgpu_device *adev, uint16_t pasid,
-					uint32_t flush_type, bool all_hub);
+	void (*flush_gpu_tlb_pasid)(struct amdgpu_device *adev, uint16_t pasid,
+				    uint32_t flush_type, bool all_hub,
+				    uint32_t inst);
 	/* flush the vm tlb via ring */
 	uint64_t (*emit_flush_gpu_tlb)(struct amdgpu_ring *ring, unsigned vmid,
 				       uint64_t pd_addr);
@@ -128,40 +159,58 @@ struct amdgpu_gmc_funcs {
 				   unsigned pasid);
 	/* enable/disable PRT support */
 	void (*set_prt)(struct amdgpu_device *adev, bool enable);
-	/* map mtype to hardware flags */
-	uint64_t (*map_mtype)(struct amdgpu_device *adev, uint32_t flags);
 	/* get the pde for a given mc addr */
 	void (*get_vm_pde)(struct amdgpu_device *adev, int level,
 			   u64 *dst, u64 *flags);
-	/* get the pte flags to use for a BO VA mapping */
+	/* get the pte flags to use for PTEs */
 	void (*get_vm_pte)(struct amdgpu_device *adev,
-			   struct amdgpu_bo_va_mapping *mapping,
-			   uint64_t *flags);
+			   struct amdgpu_vm *vm,
+			   struct amdgpu_bo *bo,
+			   uint32_t vm_flags,
+			   uint64_t *pte_flags);
+	/* override per-page pte flags */
+	void (*override_vm_pte_flags)(struct amdgpu_device *dev,
+				      struct amdgpu_vm *vm,
+				      uint64_t addr, uint64_t *flags);
 	/* get the amount of memory used by the vbios for pre-OS console */
 	unsigned int (*get_vbios_fb_size)(struct amdgpu_device *adev);
+	/* get the DCC buffer alignment */
+	unsigned int (*get_dcc_alignment)(struct amdgpu_device *adev);
+
+	enum amdgpu_memory_partition (*query_mem_partition_mode)(
+		struct amdgpu_device *adev);
+	/* Request NPS mode */
+	int (*request_mem_partition_mode)(struct amdgpu_device *adev,
+					  int nps_mode);
+	bool (*need_reset_on_init)(struct amdgpu_device *adev);
+};
+
+struct amdgpu_mem_partition_info {
+	union {
+		struct {
+			uint32_t fpfn;
+			uint32_t lpfn;
+		} range;
+		struct {
+			int node;
+		} numa;
+	};
+	uint64_t size;
 };
 
-struct amdgpu_xgmi_ras {
-	struct amdgpu_ras_block_object ras_block;
+#define INVALID_PFN    -1
+
+struct amdgpu_gmc_memrange {
+	uint64_t base_address;
+	uint64_t limit_address;
+	uint32_t flags;
+	int nid_mask;
 };
 
-struct amdgpu_xgmi {
-	/* from psp */
-	u64 node_id;
-	u64 hive_id;
-	/* fixed per family */
-	u64 node_segment_size;
-	/* physical node (0-3) */
-	unsigned physical_node_id;
-	/* number of nodes (0-4) */
-	unsigned num_physical_nodes;
-	/* gpu list in the same hive */
-	struct list_head head;
-	bool supported;
-	struct ras_common_if *ras_if;
-	bool connected_to_cpu;
-	bool pending_reset;
-	struct amdgpu_xgmi_ras *ras;
+enum amdgpu_gart_placement {
+	AMDGPU_GART_PLACEMENT_BEST_FIT = 0,
+	AMDGPU_GART_PLACEMENT_HIGH,
+	AMDGPU_GART_PLACEMENT_LOW,
 };
 
 struct amdgpu_gmc {
@@ -250,12 +299,19 @@ struct amdgpu_gmc {
 	uint64_t		last_fault:AMDGPU_GMC_FAULT_RING_ORDER;
 
 	bool tmz_enabled;
+	bool is_app_apu;
 
+	struct amdgpu_mem_partition_info *mem_partitions;
+	uint8_t num_mem_partitions;
 	const struct amdgpu_gmc_funcs	*gmc_funcs;
+	enum amdgpu_memory_partition	requested_nps_mode;
+	uint32_t supported_nps_modes;
+	uint32_t reset_flags;
 
 	struct amdgpu_xgmi xgmi;
 	struct amdgpu_irq_src	ecc_irq;
 	int noretry;
+	uint32_t xnack_flags;
 
 	uint32_t	vmid0_page_table_block_size;
 	uint32_t	vmid0_page_table_depth;
@@ -265,6 +321,8 @@ struct amdgpu_gmc {
 
 	/* MALL size */
 	u64 mall_size;
+	uint32_t m_half_use;
+
 	/* number of UMC instances */
 	int num_umc;
 	/* mode2 save restore */
@@ -293,18 +351,28 @@ struct amdgpu_gmc {
 	u64 VM_CONTEXT_PAGE_TABLE_END_ADDR_LO32[16];
 	u64 VM_CONTEXT_PAGE_TABLE_END_ADDR_HI32[16];
 	u64 MC_VM_MX_L1_TLB_CNTL;
+
+	u64 noretry_flags;
+
+	bool flush_tlb_needs_extra_type_0;
+	bool flush_tlb_needs_extra_type_2;
+	bool flush_pasid_uses_kiq;
 };
 
-#define amdgpu_gmc_flush_gpu_tlb(adev, vmid, vmhub, type) ((adev)->gmc.gmc_funcs->flush_gpu_tlb((adev), (vmid), (vmhub), (type)))
-#define amdgpu_gmc_flush_gpu_tlb_pasid(adev, pasid, type, allhub) \
-	((adev)->gmc.gmc_funcs->flush_gpu_tlb_pasid \
-	((adev), (pasid), (type), (allhub)))
 #define amdgpu_gmc_emit_flush_gpu_tlb(r, vmid, addr) (r)->adev->gmc.gmc_funcs->emit_flush_gpu_tlb((r), (vmid), (addr))
 #define amdgpu_gmc_emit_pasid_mapping(r, vmid, pasid) (r)->adev->gmc.gmc_funcs->emit_pasid_mapping((r), (vmid), (pasid))
-#define amdgpu_gmc_map_mtype(adev, flags) (adev)->gmc.gmc_funcs->map_mtype((adev),(flags))
 #define amdgpu_gmc_get_vm_pde(adev, level, dst, flags) (adev)->gmc.gmc_funcs->get_vm_pde((adev), (level), (dst), (flags))
-#define amdgpu_gmc_get_vm_pte(adev, mapping, flags) (adev)->gmc.gmc_funcs->get_vm_pte((adev), (mapping), (flags))
+#define amdgpu_gmc_get_vm_pte(adev, vm, bo, vm_flags, pte_flags) \
+	((adev)->gmc.gmc_funcs->get_vm_pte((adev), (vm), (bo), (vm_flags), \
+					   (pte_flags)))
+#define amdgpu_gmc_override_vm_pte_flags(adev, vm, addr, pte_flags)	\
+	(adev)->gmc.gmc_funcs->override_vm_pte_flags			\
+		((adev), (vm), (addr), (pte_flags))
 #define amdgpu_gmc_get_vbios_fb_size(adev) (adev)->gmc.gmc_funcs->get_vbios_fb_size((adev))
+#define amdgpu_gmc_get_dcc_alignment(adev) ({			\
+	typeof(adev) _adev = (adev);				\
+	_adev->gmc.gmc_funcs->get_dcc_alignment(_adev);		\
+})
 
 /**
  * amdgpu_gmc_vram_full_visible - Check if full VRAM is visible through the BAR
@@ -334,6 +402,7 @@ static inline uint64_t amdgpu_gmc_sign_extend(uint64_t addr)
 	return addr;
 }
 
+bool amdgpu_gmc_is_pdb0_enabled(struct amdgpu_device *adev);
 int amdgpu_gmc_pdb0_alloc(struct amdgpu_device *adev);
 void amdgpu_gmc_get_pde_for_bo(struct amdgpu_bo *bo, int level,
 			       uint64_t *addr, uint64_t *flags);
@@ -346,9 +415,12 @@ void amdgpu_gmc_sysvm_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc
 void amdgpu_gmc_vram_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc,
 			      u64 base);
 void amdgpu_gmc_gart_location(struct amdgpu_device *adev,
-			      struct amdgpu_gmc *mc);
+			      struct amdgpu_gmc *mc,
+			      enum amdgpu_gart_placement gart_placement);
 void amdgpu_gmc_agp_location(struct amdgpu_device *adev,
 			     struct amdgpu_gmc *mc);
+void amdgpu_gmc_set_agp_default(struct amdgpu_device *adev,
+				struct amdgpu_gmc *mc);
 bool amdgpu_gmc_filter_faults(struct amdgpu_device *adev,
 			      struct amdgpu_ih_ring *ih, uint64_t addr,
 			      uint16_t pasid, uint64_t timestamp);
@@ -358,6 +430,15 @@ int amdgpu_gmc_ras_sw_init(struct amdgpu_device *adev);
 int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev);
 void amdgpu_gmc_ras_fini(struct amdgpu_device *adev);
 int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev);
+void amdgpu_gmc_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
+			      uint32_t vmhub, uint32_t flush_type);
+int amdgpu_gmc_flush_gpu_tlb_pasid(struct amdgpu_device *adev, uint16_t pasid,
+				   uint32_t flush_type, bool all_hub,
+				   uint32_t inst);
+void amdgpu_gmc_fw_reg_write_reg_wait(struct amdgpu_device *adev,
+				      uint32_t reg0, uint32_t reg1,
+				      uint32_t ref, uint32_t mask,
+				      uint32_t xcc_inst);
 
 extern void amdgpu_gmc_tmz_set(struct amdgpu_device *adev);
 extern void amdgpu_gmc_noretry_set(struct amdgpu_device *adev);
@@ -371,6 +452,25 @@ void amdgpu_gmc_get_vbios_allocations(struct amdgpu_device *adev);
 void amdgpu_gmc_init_pdb0(struct amdgpu_device *adev);
 uint64_t amdgpu_gmc_vram_mc2pa(struct amdgpu_device *adev, uint64_t mc_addr);
 uint64_t amdgpu_gmc_vram_pa(struct amdgpu_device *adev, struct amdgpu_bo *bo);
-uint64_t amdgpu_gmc_vram_cpu_pa(struct amdgpu_device *adev, struct amdgpu_bo *bo);
 int amdgpu_gmc_vram_checking(struct amdgpu_device *adev);
+int amdgpu_gmc_sysfs_init(struct amdgpu_device *adev);
+void amdgpu_gmc_sysfs_fini(struct amdgpu_device *adev);
+
+int amdgpu_gmc_get_nps_memranges(struct amdgpu_device *adev,
+				 struct amdgpu_mem_partition_info *mem_ranges,
+				 uint8_t *exp_ranges);
+
+int amdgpu_gmc_request_memory_partition(struct amdgpu_device *adev,
+					int nps_mode);
+void amdgpu_gmc_prepare_nps_mode_change(struct amdgpu_device *adev);
+bool amdgpu_gmc_need_reset_on_init(struct amdgpu_device *adev);
+enum amdgpu_memory_partition
+amdgpu_gmc_get_vf_memory_partition(struct amdgpu_device *adev);
+enum amdgpu_memory_partition
+amdgpu_gmc_get_memory_partition(struct amdgpu_device *adev, u32 *supp_modes);
+enum amdgpu_memory_partition
+amdgpu_gmc_query_memory_partition(struct amdgpu_device *adev);
+int amdgpu_gmc_init_mem_ranges(struct amdgpu_device *adev);
+void amdgpu_gmc_init_sw_mem_ranges(struct amdgpu_device *adev,
+				   struct amdgpu_mem_partition_info *mem_ranges);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c
index 44367f03316f..895c1e4c6747 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c
@@ -200,8 +200,6 @@ void amdgpu_gtt_mgr_recover(struct amdgpu_gtt_mgr *mgr)
 		amdgpu_ttm_recover_gart(node->base.bo);
 	}
 	spin_unlock(&mgr->lock);
-
-	amdgpu_gart_invalidate_tlb(adev);
 }
 
 /**
@@ -286,6 +284,7 @@ int amdgpu_gtt_mgr_init(struct amdgpu_device *adev, uint64_t gtt_size)
 	ttm_resource_manager_init(man, &adev->mman.bdev, gtt_size);
 
 	start = AMDGPU_GTT_MAX_TRANSFER_SIZE * AMDGPU_GTT_NUM_TRANSFER_WINDOWS;
+	start += amdgpu_vce_required_gart_pages(adev);
 	size = (adev->gmc.gart_size >> PAGE_SHIFT) - start;
 	drm_mm_init(&mgr->mm, start, size);
 	spin_lock_init(&mgr->lock);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c
index b6cf801939aa..5a60d69a3e1f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c
@@ -22,6 +22,7 @@
  */
 #include "amdgpu.h"
 #include "amdgpu_ras.h"
+#include <uapi/linux/kfd_ioctl.h>
 
 int amdgpu_hdp_ras_sw_init(struct amdgpu_device *adev)
 {
@@ -46,3 +47,38 @@ int amdgpu_hdp_ras_sw_init(struct amdgpu_device *adev)
 	/* hdp ras follows amdgpu_ras_block_late_init_default for late init */
 	return 0;
 }
+
+void amdgpu_hdp_generic_flush(struct amdgpu_device *adev,
+			      struct amdgpu_ring *ring)
+{
+	if (!ring || !ring->funcs->emit_wreg) {
+		WREG32((adev->rmmio_remap.reg_offset +
+			KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >>
+			       2,
+		       0);
+		if (adev->nbio.funcs->get_memsize)
+			adev->nbio.funcs->get_memsize(adev);
+	} else {
+		amdgpu_ring_emit_wreg(ring,
+				      (adev->rmmio_remap.reg_offset +
+				       KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >>
+					      2,
+				      0);
+	}
+}
+
+void amdgpu_hdp_invalidate(struct amdgpu_device *adev, struct amdgpu_ring *ring)
+{
+	if (adev->asic_funcs && adev->asic_funcs->invalidate_hdp)
+		adev->asic_funcs->invalidate_hdp(adev, ring);
+	else if (adev->hdp.funcs && adev->hdp.funcs->invalidate_hdp)
+		adev->hdp.funcs->invalidate_hdp(adev, ring);
+}
+
+void amdgpu_hdp_flush(struct amdgpu_device *adev, struct amdgpu_ring *ring)
+{
+	if (adev->asic_funcs && adev->asic_funcs->flush_hdp)
+		adev->asic_funcs->flush_hdp(adev, ring);
+	else if (adev->hdp.funcs && adev->hdp.funcs->flush_hdp)
+		adev->hdp.funcs->flush_hdp(adev, ring);
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h
index 7b8a6152dc8d..d9f488fa76b9 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h
@@ -44,4 +44,10 @@ struct amdgpu_hdp {
 };
 
 int amdgpu_hdp_ras_sw_init(struct amdgpu_device *adev);
+void amdgpu_hdp_generic_flush(struct amdgpu_device *adev,
+			      struct amdgpu_ring *ring);
+void amdgpu_hdp_invalidate(struct amdgpu_device *adev,
+			   struct amdgpu_ring *ring);
+void amdgpu_hdp_flush(struct amdgpu_device *adev,
+		      struct amdgpu_ring *ring);
 #endif /* __AMDGPU_HDP_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c
index 2dadcfe43d03..90d26d820bac 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c
@@ -129,13 +129,25 @@ static const struct mmu_interval_notifier_ops amdgpu_hmm_hsa_ops = {
  */
 int amdgpu_hmm_register(struct amdgpu_bo *bo, unsigned long addr)
 {
+	int r;
+
 	if (bo->kfd_bo)
-		return mmu_interval_notifier_insert(&bo->notifier, current->mm,
+		r = mmu_interval_notifier_insert(&bo->notifier, current->mm,
 						    addr, amdgpu_bo_size(bo),
 						    &amdgpu_hmm_hsa_ops);
-	return mmu_interval_notifier_insert(&bo->notifier, current->mm, addr,
-					    amdgpu_bo_size(bo),
-					    &amdgpu_hmm_gfx_ops);
+	else
+		r = mmu_interval_notifier_insert(&bo->notifier, current->mm, addr,
+							amdgpu_bo_size(bo),
+							&amdgpu_hmm_gfx_ops);
+	if (r)
+		/*
+		 * Make sure amdgpu_hmm_unregister() doesn't call
+		 * mmu_interval_notifier_remove() when the notifier isn't properly
+		 * initialized.
+		 */
+		bo->notifier.mm = NULL;
+
+	return r;
 }
 
 /**
@@ -155,19 +167,14 @@ void amdgpu_hmm_unregister(struct amdgpu_bo *bo)
 
 int amdgpu_hmm_range_get_pages(struct mmu_interval_notifier *notifier,
 			       uint64_t start, uint64_t npages, bool readonly,
-			       void *owner, struct page **pages,
-			       struct hmm_range **phmm_range)
+			       void *owner,
+			       struct amdgpu_hmm_range *range)
 {
-	struct hmm_range *hmm_range;
 	unsigned long end;
 	unsigned long timeout;
-	unsigned long i;
 	unsigned long *pfns;
 	int r = 0;
-
-	hmm_range = kzalloc(sizeof(*hmm_range), GFP_KERNEL);
-	if (unlikely(!hmm_range))
-		return -ENOMEM;
+	struct hmm_range *hmm_range = &range->hmm_range;
 
 	pfns = kvmalloc_array(npages, sizeof(*pfns), GFP_KERNEL);
 	if (unlikely(!pfns)) {
@@ -190,19 +197,12 @@ int amdgpu_hmm_range_get_pages(struct mmu_interval_notifier *notifier,
 		pr_debug("hmm range: start = 0x%lx, end = 0x%lx",
 			hmm_range->start, hmm_range->end);
 
-		/* Assuming 512MB takes maxmium 1 second to fault page address */
-		timeout = max((hmm_range->end - hmm_range->start) >> 29, 1UL);
-		timeout *= HMM_RANGE_DEFAULT_TIMEOUT;
-		timeout = jiffies + msecs_to_jiffies(timeout);
+		timeout = jiffies + msecs_to_jiffies(HMM_RANGE_DEFAULT_TIMEOUT);
 
 retry:
 		hmm_range->notifier_seq = mmu_interval_read_begin(notifier);
 		r = hmm_range_fault(hmm_range);
 		if (unlikely(r)) {
-			/*
-			 * FIXME: This timeout should encompass the retry from
-			 * mmu_interval_read_retry() as well.
-			 */
 			if (r == -EBUSY && !time_after(jiffies, timeout))
 				goto retry;
 			goto out_free_pfns;
@@ -212,40 +212,82 @@ retry:
 			break;
 		hmm_range->hmm_pfns += MAX_WALK_BYTE >> PAGE_SHIFT;
 		hmm_range->start = hmm_range->end;
-		schedule();
 	} while (hmm_range->end < end);
 
 	hmm_range->start = start;
 	hmm_range->hmm_pfns = pfns;
 
-	/*
-	 * Due to default_flags, all pages are HMM_PFN_VALID or
-	 * hmm_range_fault() fails. FIXME: The pages cannot be touched outside
-	 * the notifier_lock, and mmu_interval_read_retry() must be done first.
-	 */
-	for (i = 0; pages && i < npages; i++)
-		pages[i] = hmm_pfn_to_page(pfns[i]);
-
-	*phmm_range = hmm_range;
-
 	return 0;
 
 out_free_pfns:
 	kvfree(pfns);
+	hmm_range->hmm_pfns = NULL;
 out_free_range:
-	kfree(hmm_range);
-
+	if (r == -EBUSY)
+		r = -EAGAIN;
 	return r;
 }
 
-bool amdgpu_hmm_range_get_pages_done(struct hmm_range *hmm_range)
+/**
+ * amdgpu_hmm_range_valid - check if an HMM range is still valid
+ * @range: pointer to the &struct amdgpu_hmm_range to validate
+ *
+ * Determines whether the given HMM range @range is still valid by
+ * checking for invalidations via the MMU notifier sequence. This is
+ * typically used to verify that the range has not been invalidated
+ * by concurrent address space updates before it is accessed.
+ *
+ * Return:
+ * * true if @range is valid and can be used safely
+ * * false if @range is NULL or has been invalidated
+ */
+bool amdgpu_hmm_range_valid(struct amdgpu_hmm_range *range)
 {
-	bool r;
+	if (!range)
+		return false;
 
-	r = mmu_interval_read_retry(hmm_range->notifier,
-				    hmm_range->notifier_seq);
-	kvfree(hmm_range->hmm_pfns);
-	kfree(hmm_range);
+	return !mmu_interval_read_retry(range->hmm_range.notifier,
+					range->hmm_range.notifier_seq);
+}
 
-	return r;
+/**
+ * amdgpu_hmm_range_alloc - allocate and initialize an AMDGPU HMM range
+ * @bo: optional buffer object to associate with this HMM range
+ *
+ * Allocates memory for amdgpu_hmm_range and associates it with the @bo passed.
+ * The reference count of the @bo is incremented.
+ *
+ * Return:
+ * Pointer to a newly allocated struct amdgpu_hmm_range on success,
+ * or NULL if memory allocation fails.
+ */
+struct amdgpu_hmm_range *amdgpu_hmm_range_alloc(struct amdgpu_bo *bo)
+{
+	struct amdgpu_hmm_range *range;
+
+	range = kzalloc(sizeof(*range), GFP_KERNEL);
+	if (!range)
+		return NULL;
+
+	range->bo = amdgpu_bo_ref(bo);
+	return range;
+}
+
+/**
+ * amdgpu_hmm_range_free - release an AMDGPU HMM range
+ * @range: pointer to the range object to free
+ *
+ * Releases all resources held by @range, including the associated
+ * hmm_pfns and the dropping reference of associated bo if any.
+ *
+ * Return: void
+ */
+void amdgpu_hmm_range_free(struct amdgpu_hmm_range *range)
+{
+	if (!range)
+		return;
+
+	kvfree(range->hmm_range.hmm_pfns);
+	amdgpu_bo_unref(&range->bo);
+	kfree(range);
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.h
index e2edcd010ccc..140bc9cd57b4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.h
@@ -31,13 +31,20 @@
 #include <linux/interval_tree.h>
 #include <linux/mmu_notifier.h>
 
+struct amdgpu_hmm_range {
+	struct hmm_range hmm_range;
+	struct amdgpu_bo *bo;
+};
+
 int amdgpu_hmm_range_get_pages(struct mmu_interval_notifier *notifier,
 			       uint64_t start, uint64_t npages, bool readonly,
-			       void *owner, struct page **pages,
-			       struct hmm_range **phmm_range);
-bool amdgpu_hmm_range_get_pages_done(struct hmm_range *hmm_range);
+			       void *owner,
+			       struct amdgpu_hmm_range *range);
 
 #if defined(CONFIG_HMM_MIRROR)
+bool amdgpu_hmm_range_valid(struct amdgpu_hmm_range *range);
+struct amdgpu_hmm_range *amdgpu_hmm_range_alloc(struct amdgpu_bo *bo);
+void amdgpu_hmm_range_free(struct amdgpu_hmm_range *range);
 int amdgpu_hmm_register(struct amdgpu_bo *bo, unsigned long addr);
 void amdgpu_hmm_unregister(struct amdgpu_bo *bo);
 #else
@@ -47,7 +54,20 @@ static inline int amdgpu_hmm_register(struct amdgpu_bo *bo, unsigned long addr)
 		      "add CONFIG_ZONE_DEVICE=y in config file to fix this\n");
 	return -ENODEV;
 }
+
 static inline void amdgpu_hmm_unregister(struct amdgpu_bo *bo) {}
+
+static inline bool amdgpu_hmm_range_valid(struct amdgpu_hmm_range *range)
+{
+	return false;
+}
+
+static inline struct amdgpu_hmm_range *amdgpu_hmm_range_alloc(struct amdgpu_bo *bo)
+{
+	return NULL;
+}
+
+static inline void amdgpu_hmm_range_free(struct amdgpu_hmm_range *range) {}
 #endif
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c
index 82608df43396..9cb72f0c5277 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c
@@ -24,7 +24,6 @@
  *          Alex Deucher
  */
 
-#include <linux/export.h>
 #include <linux/pci.h>
 
 #include <drm/drm_edid.h>
@@ -175,7 +174,6 @@ struct amdgpu_i2c_chan *amdgpu_i2c_create(struct drm_device *dev,
 
 	i2c->rec = *rec;
 	i2c->adapter.owner = THIS_MODULE;
-	i2c->adapter.class = I2C_CLASS_DDC;
 	i2c->adapter.dev.parent = dev->dev;
 	i2c->dev = dev;
 	i2c_set_adapdata(&i2c->adapter, i2c);
@@ -186,7 +184,7 @@ struct amdgpu_i2c_chan *amdgpu_i2c_create(struct drm_device *dev,
 		snprintf(i2c->adapter.name, sizeof(i2c->adapter.name),
 			 "AMDGPU i2c hw bus %s", name);
 		i2c->adapter.algo = &amdgpu_atombios_i2c_algo;
-		ret = i2c_add_adapter(&i2c->adapter);
+		ret = devm_i2c_add_adapter(dev->dev, &i2c->adapter);
 		if (ret)
 			goto out_free;
 	} else {
@@ -217,22 +215,23 @@ out_free:
 
 }
 
-void amdgpu_i2c_destroy(struct amdgpu_i2c_chan *i2c)
-{
-	if (!i2c)
-		return;
-	WARN_ON(i2c->has_aux);
-	i2c_del_adapter(&i2c->adapter);
-	kfree(i2c);
-}
-
-/* Add the default buses */
 void amdgpu_i2c_init(struct amdgpu_device *adev)
 {
-	if (amdgpu_hw_i2c)
-		DRM_INFO("hw_i2c forced on, you may experience display detection problems!\n");
-
-	amdgpu_atombios_i2c_init(adev);
+	if (!adev->is_atom_fw) {
+		if (!amdgpu_device_has_dc_support(adev)) {
+			amdgpu_atombios_i2c_init(adev);
+		} else {
+			switch (adev->asic_type) {
+			case CHIP_POLARIS10:
+			case CHIP_POLARIS11:
+			case CHIP_POLARIS12:
+				amdgpu_atombios_oem_i2c_init(adev, 0x97);
+				break;
+			default:
+				break;
+			}
+		}
+	}
 }
 
 /* remove all the buses */
@@ -240,28 +239,9 @@ void amdgpu_i2c_fini(struct amdgpu_device *adev)
 {
 	int i;
 
-	for (i = 0; i < AMDGPU_MAX_I2C_BUS; i++) {
-		if (adev->i2c_bus[i]) {
-			amdgpu_i2c_destroy(adev->i2c_bus[i]);
+	for (i = 0; i < AMDGPU_MAX_I2C_BUS; i++)
+		if (adev->i2c_bus[i])
 			adev->i2c_bus[i] = NULL;
-		}
-	}
-}
-
-/* Add additional buses */
-void amdgpu_i2c_add(struct amdgpu_device *adev,
-		    const struct amdgpu_i2c_bus_rec *rec,
-		    const char *name)
-{
-	struct drm_device *dev = adev_to_drm(adev);
-	int i;
-
-	for (i = 0; i < AMDGPU_MAX_I2C_BUS; i++) {
-		if (!adev->i2c_bus[i]) {
-			adev->i2c_bus[i] = amdgpu_i2c_create(dev, rec, name);
-			return;
-		}
-	}
 }
 
 /* looks up bus based on id */
@@ -280,7 +260,7 @@ amdgpu_i2c_lookup(struct amdgpu_device *adev,
 	return NULL;
 }
 
-static void amdgpu_i2c_get_byte(struct amdgpu_i2c_chan *i2c_bus,
+static int amdgpu_i2c_get_byte(struct amdgpu_i2c_chan *i2c_bus,
 				 u8 slave_addr,
 				 u8 addr,
 				 u8 *val)
@@ -305,16 +285,18 @@ static void amdgpu_i2c_get_byte(struct amdgpu_i2c_chan *i2c_bus,
 	out_buf[0] = addr;
 	out_buf[1] = 0;
 
-	if (i2c_transfer(&i2c_bus->adapter, msgs, 2) == 2) {
-		*val = in_buf[0];
-		DRM_DEBUG("val = 0x%02x\n", *val);
-	} else {
-		DRM_DEBUG("i2c 0x%02x 0x%02x read failed\n",
-			  addr, *val);
+	if (i2c_transfer(&i2c_bus->adapter, msgs, 2) != 2) {
+		DRM_DEBUG("i2c 0x%02x read failed\n", addr);
+		return -EIO;
 	}
+
+	*val = in_buf[0];
+	DRM_DEBUG("val = 0x%02x\n", *val);
+
+	return 0;
 }
 
-static void amdgpu_i2c_put_byte(struct amdgpu_i2c_chan *i2c_bus,
+static int amdgpu_i2c_put_byte(struct amdgpu_i2c_chan *i2c_bus,
 				 u8 slave_addr,
 				 u8 addr,
 				 u8 val)
@@ -330,9 +312,12 @@ static void amdgpu_i2c_put_byte(struct amdgpu_i2c_chan *i2c_bus,
 	out_buf[0] = addr;
 	out_buf[1] = val;
 
-	if (i2c_transfer(&i2c_bus->adapter, &msg, 1) != 1)
-		DRM_DEBUG("i2c 0x%02x 0x%02x write failed\n",
-			  addr, val);
+	if (i2c_transfer(&i2c_bus->adapter, &msg, 1) != 1) {
+		DRM_DEBUG("i2c 0x%02x 0x%02x write failed\n", addr, val);
+		return -EIO;
+	}
+
+	return 0;
 }
 
 /* ddc router switching */
@@ -347,16 +332,18 @@ amdgpu_i2c_router_select_ddc_port(const struct amdgpu_connector *amdgpu_connecto
 	if (!amdgpu_connector->router_bus)
 		return;
 
-	amdgpu_i2c_get_byte(amdgpu_connector->router_bus,
+	if (amdgpu_i2c_get_byte(amdgpu_connector->router_bus,
 			    amdgpu_connector->router.i2c_addr,
-			    0x3, &val);
+			    0x3, &val))
+		return;
 	val &= ~amdgpu_connector->router.ddc_mux_control_pin;
 	amdgpu_i2c_put_byte(amdgpu_connector->router_bus,
 			    amdgpu_connector->router.i2c_addr,
 			    0x3, val);
-	amdgpu_i2c_get_byte(amdgpu_connector->router_bus,
+	if (amdgpu_i2c_get_byte(amdgpu_connector->router_bus,
 			    amdgpu_connector->router.i2c_addr,
-			    0x1, &val);
+			    0x1, &val))
+		return;
 	val &= ~amdgpu_connector->router.ddc_mux_control_pin;
 	val |= amdgpu_connector->router.ddc_mux_state;
 	amdgpu_i2c_put_byte(amdgpu_connector->router_bus,
@@ -376,16 +363,18 @@ amdgpu_i2c_router_select_cd_port(const struct amdgpu_connector *amdgpu_connector
 	if (!amdgpu_connector->router_bus)
 		return;
 
-	amdgpu_i2c_get_byte(amdgpu_connector->router_bus,
+	if (amdgpu_i2c_get_byte(amdgpu_connector->router_bus,
 			    amdgpu_connector->router.i2c_addr,
-			    0x3, &val);
+			    0x3, &val))
+		return;
 	val &= ~amdgpu_connector->router.cd_mux_control_pin;
 	amdgpu_i2c_put_byte(amdgpu_connector->router_bus,
 			    amdgpu_connector->router.i2c_addr,
 			    0x3, val);
-	amdgpu_i2c_get_byte(amdgpu_connector->router_bus,
+	if (amdgpu_i2c_get_byte(amdgpu_connector->router_bus,
 			    amdgpu_connector->router.i2c_addr,
-			    0x1, &val);
+			    0x1, &val))
+		return;
 	val &= ~amdgpu_connector->router.cd_mux_control_pin;
 	val |= amdgpu_connector->router.cd_mux_state;
 	amdgpu_i2c_put_byte(amdgpu_connector->router_bus,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.h
index 63c2ff7499e1..1d3d3806e0dd 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.h
@@ -30,9 +30,6 @@ struct amdgpu_i2c_chan *amdgpu_i2c_create(struct drm_device *dev,
 void amdgpu_i2c_destroy(struct amdgpu_i2c_chan *i2c);
 void amdgpu_i2c_init(struct amdgpu_device *adev);
 void amdgpu_i2c_fini(struct amdgpu_device *adev);
-void amdgpu_i2c_add(struct amdgpu_device *adev,
-		    const struct amdgpu_i2c_bus_rec *rec,
-		    const char *name);
 struct amdgpu_i2c_chan *
 amdgpu_i2c_lookup(struct amdgpu_device *adev,
 		  const struct amdgpu_i2c_bus_rec *i2c_bus);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
index 4ff348e10e4d..586a58facca1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
@@ -62,7 +62,7 @@
  * Returns 0 on success, error on failure.
  */
 int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-		  unsigned size, enum amdgpu_ib_pool_type pool_type,
+		  unsigned int size, enum amdgpu_ib_pool_type pool_type,
 		  struct amdgpu_ib *ib)
 {
 	int r;
@@ -89,16 +89,14 @@ int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 /**
  * amdgpu_ib_free - free an IB (Indirect Buffer)
  *
- * @adev: amdgpu_device pointer
  * @ib: IB object to free
  * @f: the fence SA bo need wait on for the ib alloation
  *
  * Free an IB (all asics).
  */
-void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib,
-		    struct dma_fence *f)
+void amdgpu_ib_free(struct amdgpu_ib *ib, struct dma_fence *f)
 {
-	amdgpu_sa_bo_free(adev, &ib->sa_bo, f);
+	amdgpu_sa_bo_free(&ib->sa_bo, f);
 }
 
 /**
@@ -123,24 +121,26 @@ void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib,
  * a CONST_IB), it will be put on the ring prior to the DE IB.  Prior
  * to SI there was just a DE IB.
  */
-int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
+int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned int num_ibs,
 		       struct amdgpu_ib *ibs, struct amdgpu_job *job,
 		       struct dma_fence **f)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct amdgpu_ib *ib = &ibs[0];
 	struct dma_fence *tmp = NULL;
+	struct amdgpu_fence *af;
 	bool need_ctx_switch;
-	unsigned patch_offset = ~0;
 	struct amdgpu_vm *vm;
 	uint64_t fence_ctx;
 	uint32_t status = 0, alloc_size;
-	unsigned fence_flags = 0;
-	bool secure;
-
-	unsigned i;
-	int r = 0;
+	unsigned int fence_flags = 0;
+	bool secure, init_shadow;
+	u64 shadow_va, csa_va, gds_va;
+	int vmid = AMDGPU_JOB_GET_VMID(job);
 	bool need_pipe_sync = false;
+	unsigned int cond_exec;
+	unsigned int i;
+	int r = 0;
 
 	if (num_ibs == 0)
 		return -EINVAL;
@@ -149,26 +149,48 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 	if (job) {
 		vm = job->vm;
 		fence_ctx = job->base.s_fence ?
-			job->base.s_fence->scheduled.context : 0;
+			job->base.s_fence->finished.context : 0;
+		shadow_va = job->shadow_va;
+		csa_va = job->csa_va;
+		gds_va = job->gds_va;
+		init_shadow = job->init_shadow;
+		af = job->hw_fence;
+		/* Save the context of the job for reset handling.
+		 * The driver needs this so it can skip the ring
+		 * contents for guilty contexts.
+		 */
+		af->context = fence_ctx;
+		/* the vm fence is also part of the job's context */
+		job->hw_vm_fence->context = fence_ctx;
 	} else {
 		vm = NULL;
 		fence_ctx = 0;
+		shadow_va = 0;
+		csa_va = 0;
+		gds_va = 0;
+		init_shadow = false;
+		af = kzalloc(sizeof(*af), GFP_ATOMIC);
+		if (!af)
+			return -ENOMEM;
 	}
 
-	if (!ring->sched.ready && !ring->is_mes_queue) {
+	if (!ring->sched.ready) {
 		dev_err(adev->dev, "couldn't schedule ib on ring <%s>\n", ring->name);
-		return -EINVAL;
+		r = -EINVAL;
+		goto free_fence;
 	}
 
-	if (vm && !job->vmid && !ring->is_mes_queue) {
+	if (vm && !job->vmid) {
 		dev_err(adev->dev, "VM IB without ID\n");
-		return -EINVAL;
+		r = -EINVAL;
+		goto free_fence;
 	}
 
 	if ((ib->flags & AMDGPU_IB_FLAGS_SECURE) &&
 	    (!ring->funcs->secure_submission_supported)) {
 		dev_err(adev->dev, "secure submissions not supported on ring <%s>\n", ring->name);
-		return -EINVAL;
+		r = -EINVAL;
+		goto free_fence;
 	}
 
 	alloc_size = ring->funcs->emit_frame_size + num_ibs *
@@ -177,14 +199,14 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 	r = amdgpu_ring_alloc(ring, alloc_size);
 	if (r) {
 		dev_err(adev->dev, "scheduling IB failed (%d).\n", r);
-		return r;
+		goto free_fence;
 	}
 
 	need_ctx_switch = ring->current_ctx != fence_ctx;
 	if (ring->funcs->emit_pipeline_sync && job &&
 	    ((tmp = amdgpu_sync_get_fence(&job->explicit_sync)) ||
-	     (amdgpu_sriov_vf(adev) && need_ctx_switch) ||
-	     amdgpu_vm_need_pipeline_sync(ring, job))) {
+	     need_ctx_switch || amdgpu_vm_need_pipeline_sync(ring, job))) {
+
 		need_pipe_sync = true;
 
 		if (tmp)
@@ -212,8 +234,14 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 	}
 
 	amdgpu_ring_ib_begin(ring);
-	if (job && ring->funcs->init_cond_exec)
-		patch_offset = amdgpu_ring_init_cond_exec(ring);
+
+	if (ring->funcs->emit_gfx_shadow)
+		amdgpu_ring_emit_gfx_shadow(ring, shadow_va, csa_va, gds_va,
+					    init_shadow, vmid);
+
+	if (ring->funcs->init_cond_exec)
+		cond_exec = amdgpu_ring_init_cond_exec(ring,
+						       ring->cond_exe_gpu_addr);
 
 	amdgpu_device_flush_hdp(adev, ring);
 
@@ -263,7 +291,12 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 				       fence_flags | AMDGPU_FENCE_FLAG_64BIT);
 	}
 
-	r = amdgpu_fence_emit(ring, f, job, fence_flags);
+	if (ring->funcs->emit_gfx_shadow && ring->funcs->init_cond_exec) {
+		amdgpu_ring_emit_gfx_shadow(ring, 0, 0, 0, false, 0);
+		amdgpu_ring_init_cond_exec(ring, ring->cond_exe_gpu_addr);
+	}
+
+	r = amdgpu_fence_emit(ring, af, fence_flags);
 	if (r) {
 		dev_err(adev->dev, "failed to emit fence (%d)\n", r);
 		if (job && job->vmid)
@@ -271,24 +304,41 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 		amdgpu_ring_undo(ring);
 		return r;
 	}
+	*f = &af->base;
+	/* get a ref for the job */
+	if (job)
+		dma_fence_get(*f);
 
 	if (ring->funcs->insert_end)
 		ring->funcs->insert_end(ring);
 
-	if (patch_offset != ~0 && ring->funcs->patch_cond_exec)
-		amdgpu_ring_patch_cond_exec(ring, patch_offset);
+	amdgpu_ring_patch_cond_exec(ring, cond_exec);
 
 	ring->current_ctx = fence_ctx;
-	if (vm && ring->funcs->emit_switch_buffer)
+	if (job && ring->funcs->emit_switch_buffer)
 		amdgpu_ring_emit_switch_buffer(ring);
 
 	if (ring->funcs->emit_wave_limit &&
 	    ring->hw_prio == AMDGPU_GFX_PIPE_PRIO_HIGH)
 		ring->funcs->emit_wave_limit(ring, false);
 
+	/* Save the wptr associated with this fence.
+	 * This must be last for resets to work properly
+	 * as we need to save the wptr associated with this
+	 * fence so we know what rings contents to backup
+	 * after we reset the queue.
+	 */
+	amdgpu_fence_save_wptr(af);
+
 	amdgpu_ring_ib_end(ring);
 	amdgpu_ring_commit(ring);
+
 	return 0;
+
+free_fence:
+	if (!job)
+		kfree(af);
+	return r;
 }
 
 /**
@@ -358,7 +408,7 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev)
 {
 	long tmo_gfx, tmo_mm;
 	int r, ret = 0;
-	unsigned i;
+	unsigned int i;
 
 	tmo_mm = tmo_gfx = AMDGPU_IB_TEST_TIMEOUT;
 	if (amdgpu_sriov_vf(adev)) {
@@ -375,7 +425,7 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev)
 		/* for CP & SDMA engines since they are scheduled together so
 		 * need to make the timeout width enough to cover the time
 		 * cost waiting for it coming back under RUNTIME only
-		*/
+		 */
 		tmo_gfx = 8 * AMDGPU_IB_TEST_TIMEOUT;
 	} else if (adev->gmc.xgmi.hive_id) {
 		tmo_gfx = AMDGPU_IB_TEST_GFX_XGMI_TIMEOUT;
@@ -436,15 +486,15 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev)
 
 static int amdgpu_debugfs_sa_info_show(struct seq_file *m, void *unused)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct amdgpu_device *adev = m->private;
 
-	seq_printf(m, "--------------------- DELAYED --------------------- \n");
+	seq_puts(m, "--------------------- DELAYED ---------------------\n");
 	amdgpu_sa_bo_dump_debug_info(&adev->ib_pools[AMDGPU_IB_POOL_DELAYED],
 				     m);
-	seq_printf(m, "-------------------- IMMEDIATE -------------------- \n");
+	seq_puts(m, "-------------------- IMMEDIATE --------------------\n");
 	amdgpu_sa_bo_dump_debug_info(&adev->ib_pools[AMDGPU_IB_POOL_IMMEDIATE],
 				     m);
-	seq_printf(m, "--------------------- DIRECT ---------------------- \n");
+	seq_puts(m, "--------------------- DIRECT ----------------------\n");
 	amdgpu_sa_bo_dump_debug_info(&adev->ib_pools[AMDGPU_IB_POOL_DIRECT], m);
 
 	return 0;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
index c991ca0b7a1c..9cab36322c16 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
@@ -62,9 +62,8 @@ int amdgpu_pasid_alloc(unsigned int bits)
 	int pasid = -EINVAL;
 
 	for (bits = min(bits, 31U); bits > 0; bits--) {
-		pasid = ida_simple_get(&amdgpu_pasid_ida,
-				       1U << (bits - 1), 1U << bits,
-				       GFP_KERNEL);
+		pasid = ida_alloc_range(&amdgpu_pasid_ida, 1U << (bits - 1),
+					(1U << bits) - 1, GFP_KERNEL);
 		if (pasid != -ENOSPC)
 			break;
 	}
@@ -82,7 +81,7 @@ int amdgpu_pasid_alloc(unsigned int bits)
 void amdgpu_pasid_free(u32 pasid)
 {
 	trace_amdgpu_pasid_freed(pasid);
-	ida_simple_remove(&amdgpu_pasid_ida, pasid);
+	ida_free(&amdgpu_pasid_ida, pasid);
 }
 
 static void amdgpu_pasid_free_cb(struct dma_fence *fence,
@@ -188,7 +187,6 @@ static bool amdgpu_vmid_compatible(struct amdgpu_vmid *id,
 /**
  * amdgpu_vmid_grab_idle - grab idle VMID
  *
- * @vm: vm to allocate id for
  * @ring: ring we want to submit job to
  * @idle: resulting idle VMID
  * @fence: fence to wait for if no id could be grabbed
@@ -196,66 +194,41 @@ static bool amdgpu_vmid_compatible(struct amdgpu_vmid *id,
  * Try to find an idle VMID, if none is idle add a fence to wait to the sync
  * object. Returns -ENOMEM when we are out of memory.
  */
-static int amdgpu_vmid_grab_idle(struct amdgpu_vm *vm,
-				 struct amdgpu_ring *ring,
+static int amdgpu_vmid_grab_idle(struct amdgpu_ring *ring,
 				 struct amdgpu_vmid **idle,
 				 struct dma_fence **fence)
 {
 	struct amdgpu_device *adev = ring->adev;
 	unsigned vmhub = ring->vm_hub;
 	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
-	struct dma_fence **fences;
-	unsigned i;
 
+	/* If anybody is waiting for a VMID let everybody wait for fairness */
 	if (!dma_fence_is_signaled(ring->vmid_wait)) {
 		*fence = dma_fence_get(ring->vmid_wait);
 		return 0;
 	}
 
-	fences = kmalloc_array(id_mgr->num_ids, sizeof(void *), GFP_KERNEL);
-	if (!fences)
-		return -ENOMEM;
-
 	/* Check if we have an idle VMID */
-	i = 0;
-	list_for_each_entry((*idle), &id_mgr->ids_lru, list) {
+	list_for_each_entry_reverse((*idle), &id_mgr->ids_lru, list) {
 		/* Don't use per engine and per process VMID at the same time */
 		struct amdgpu_ring *r = adev->vm_manager.concurrent_flush ?
 			NULL : ring;
 
-		fences[i] = amdgpu_sync_peek_fence(&(*idle)->active, r);
-		if (!fences[i])
-			break;
-		++i;
+		*fence = amdgpu_sync_peek_fence(&(*idle)->active, r);
+		if (!(*fence))
+			return 0;
 	}
 
-	/* If we can't find a idle VMID to use, wait till one becomes available */
-	if (&(*idle)->list == &id_mgr->ids_lru) {
-		u64 fence_context = adev->vm_manager.fence_context + ring->idx;
-		unsigned seqno = ++adev->vm_manager.seqno[ring->idx];
-		struct dma_fence_array *array;
-		unsigned j;
-
-		*idle = NULL;
-		for (j = 0; j < i; ++j)
-			dma_fence_get(fences[j]);
-
-		array = dma_fence_array_create(i, fences, fence_context,
-					       seqno, true);
-		if (!array) {
-			for (j = 0; j < i; ++j)
-				dma_fence_put(fences[j]);
-			kfree(fences);
-			return -ENOMEM;
-		}
-
-		*fence = dma_fence_get(&array->base);
-		dma_fence_put(ring->vmid_wait);
-		ring->vmid_wait = &array->base;
-		return 0;
-	}
-	kfree(fences);
+	/*
+	 * If we can't find a idle VMID to use, wait on a fence from the least
+	 * recently used in the hope that it will be available soon.
+	 */
+	*idle = NULL;
+	dma_fence_put(ring->vmid_wait);
+	ring->vmid_wait = dma_fence_get(*fence);
 
+	/* This is the reference we return */
+	dma_fence_get(*fence);
 	return 0;
 }
 
@@ -278,22 +251,27 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm,
 {
 	struct amdgpu_device *adev = ring->adev;
 	unsigned vmhub = ring->vm_hub;
-	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
 	uint64_t fence_context = adev->fence_context + ring->idx;
 	bool needs_flush = vm->use_cpu_for_update;
 	uint64_t updates = amdgpu_vm_tlb_seq(vm);
 	int r;
 
-	*id = id_mgr->reserved;
+	*id = vm->reserved_vmid[vmhub];
 	if ((*id)->owner != vm->immediate.fence_context ||
 	    !amdgpu_vmid_compatible(*id, job) ||
 	    (*id)->flushed_updates < updates ||
 	    !(*id)->last_flush ||
 	    ((*id)->last_flush->context != fence_context &&
-	     !dma_fence_is_signaled((*id)->last_flush))) {
+	     !dma_fence_is_signaled((*id)->last_flush)))
+		needs_flush = true;
+
+	if ((*id)->owner != vm->immediate.fence_context ||
+	    (!adev->vm_manager.concurrent_flush && needs_flush)) {
 		struct dma_fence *tmp;
 
-		/* Don't use per engine and per process VMID at the same time */
+		/* Don't use per engine and per process VMID at the
+		 * same time
+		 */
 		if (adev->vm_manager.concurrent_flush)
 			ring = NULL;
 
@@ -305,13 +283,13 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm,
 			*fence = dma_fence_get(tmp);
 			return 0;
 		}
-		needs_flush = true;
 	}
 
 	/* Good we can use this VMID. Remember this submission as
 	* user of the VMID.
 	*/
-	r = amdgpu_sync_fence(&(*id)->active, &job->base.s_fence->finished);
+	r = amdgpu_sync_fence(&(*id)->active, &job->base.s_fence->finished,
+			      GFP_ATOMIC);
 	if (r)
 		return r;
 
@@ -327,15 +305,13 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm,
  * @ring: ring we want to submit job to
  * @job: job who wants to use the VMID
  * @id: resulting VMID
- * @fence: fence to wait for if no id could be grabbed
  *
  * Try to reuse a VMID for this submission.
  */
 static int amdgpu_vmid_grab_used(struct amdgpu_vm *vm,
 				 struct amdgpu_ring *ring,
 				 struct amdgpu_job *job,
-				 struct amdgpu_vmid **id,
-				 struct dma_fence **fence)
+				 struct amdgpu_vmid **id)
 {
 	struct amdgpu_device *adev = ring->adev;
 	unsigned vmhub = ring->vm_hub;
@@ -372,7 +348,8 @@ static int amdgpu_vmid_grab_used(struct amdgpu_vm *vm,
 		 * user of the VMID.
 		 */
 		r = amdgpu_sync_fence(&(*id)->active,
-				      &job->base.s_fence->finished);
+				      &job->base.s_fence->finished,
+				      GFP_ATOMIC);
 		if (r)
 			return r;
 
@@ -405,16 +382,16 @@ int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
 	int r = 0;
 
 	mutex_lock(&id_mgr->lock);
-	r = amdgpu_vmid_grab_idle(vm, ring, &idle, fence);
+	r = amdgpu_vmid_grab_idle(ring, &idle, fence);
 	if (r || !idle)
 		goto error;
 
-	if (vm->reserved_vmid[vmhub]) {
+	if (amdgpu_vmid_uses_reserved(vm, vmhub)) {
 		r = amdgpu_vmid_grab_reserved(vm, ring, job, &id, fence);
 		if (r || !id)
 			goto error;
 	} else {
-		r = amdgpu_vmid_grab_used(vm, ring, job, &id, fence);
+		r = amdgpu_vmid_grab_used(vm, ring, job, &id);
 		if (r)
 			goto error;
 
@@ -424,7 +401,8 @@ int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
 
 			/* Remember this submission as user of the VMID */
 			r = amdgpu_sync_fence(&id->active,
-					      &job->base.s_fence->finished);
+					      &job->base.s_fence->finished,
+					      GFP_ATOMIC);
 			if (r)
 				goto error;
 
@@ -459,47 +437,73 @@ error:
 	return r;
 }
 
-int amdgpu_vmid_alloc_reserved(struct amdgpu_device *adev,
-			       struct amdgpu_vm *vm,
+/*
+ * amdgpu_vmid_uses_reserved - check if a VM will use a reserved VMID
+ * @vm: the VM to check
+ * @vmhub: the VMHUB which will be used
+ *
+ * Returns: True if the VM will use a reserved VMID.
+ */
+bool amdgpu_vmid_uses_reserved(struct amdgpu_vm *vm, unsigned int vmhub)
+{
+	return vm->reserved_vmid[vmhub];
+}
+
+/*
+ * amdgpu_vmid_alloc_reserved - reserve a specific VMID for this vm
+ * @adev: amdgpu device structure
+ * @vm: the VM to reserve an ID for
+ * @vmhub: the VMHUB which should be used
+ *
+ * Mostly used to have a reserved VMID for debugging and SPM.
+ *
+ * Returns: 0 for success, -ENOENT if an ID is already reserved.
+ */
+int amdgpu_vmid_alloc_reserved(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			       unsigned vmhub)
 {
 	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
+	struct amdgpu_vmid *id;
+	int r = 0;
 
 	mutex_lock(&id_mgr->lock);
 	if (vm->reserved_vmid[vmhub])
 		goto unlock;
-
-	++id_mgr->reserved_use_count;
-	if (!id_mgr->reserved) {
-		struct amdgpu_vmid *id;
-
-		id = list_first_entry(&id_mgr->ids_lru, struct amdgpu_vmid,
-				      list);
-		/* Remove from normal round robin handling */
-		list_del_init(&id->list);
-		id_mgr->reserved = id;
+	if (id_mgr->reserved_vmid) {
+		r = -ENOENT;
+		goto unlock;
 	}
-	vm->reserved_vmid[vmhub] = true;
+	/* Remove from normal round robin handling */
+	id = list_first_entry(&id_mgr->ids_lru, struct amdgpu_vmid, list);
+	list_del_init(&id->list);
+	vm->reserved_vmid[vmhub] = id;
+	id_mgr->reserved_vmid = true;
+	mutex_unlock(&id_mgr->lock);
 
+	return 0;
 unlock:
 	mutex_unlock(&id_mgr->lock);
-	return 0;
+	return r;
 }
 
-void amdgpu_vmid_free_reserved(struct amdgpu_device *adev,
-			       struct amdgpu_vm *vm,
+/*
+ * amdgpu_vmid_free_reserved - free up a reserved VMID again
+ * @adev: amdgpu device structure
+ * @vm: the VM with the reserved ID
+ * @vmhub: the VMHUB which should be used
+ */
+void amdgpu_vmid_free_reserved(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			       unsigned vmhub)
 {
 	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
 
 	mutex_lock(&id_mgr->lock);
-	if (vm->reserved_vmid[vmhub] &&
-	    !--id_mgr->reserved_use_count) {
-		/* give the reserved ID back to normal round robin */
-		list_add(&id_mgr->reserved->list, &id_mgr->ids_lru);
-		id_mgr->reserved = NULL;
+	if (vm->reserved_vmid[vmhub]) {
+		list_add(&vm->reserved_vmid[vmhub]->list,
+			&id_mgr->ids_lru);
+		vm->reserved_vmid[vmhub] = NULL;
+		id_mgr->reserved_vmid = false;
 	}
-	vm->reserved_vmid[vmhub] = false;
 	mutex_unlock(&id_mgr->lock);
 }
 
@@ -566,10 +570,17 @@ void amdgpu_vmid_mgr_init(struct amdgpu_device *adev)
 
 		mutex_init(&id_mgr->lock);
 		INIT_LIST_HEAD(&id_mgr->ids_lru);
-		id_mgr->reserved_use_count = 0;
 
-		/* manage only VMIDs not used by KFD */
-		id_mgr->num_ids = adev->vm_manager.first_kfd_vmid;
+		/* for GC <10, SDMA uses MMHUB so use first_kfd_vmid for both GC and MM */
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(10, 0, 0))
+			/* manage only VMIDs not used by KFD */
+			id_mgr->num_ids = adev->vm_manager.first_kfd_vmid;
+		else if (AMDGPU_IS_MMHUB0(i) ||
+			 AMDGPU_IS_MMHUB1(i))
+			id_mgr->num_ids = 16;
+		else
+			/* manage only VMIDs not used by KFD */
+			id_mgr->num_ids = adev->vm_manager.first_kfd_vmid;
 
 		/* skip over VMID 0, since it is the system VM */
 		for (j = 1; j < id_mgr->num_ids; ++j) {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h
index d1cc09b45da4..b3649cd3af56 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h
@@ -67,8 +67,7 @@ struct amdgpu_vmid_mgr {
 	unsigned		num_ids;
 	struct list_head	ids_lru;
 	struct amdgpu_vmid	ids[AMDGPU_NUM_VMID];
-	struct amdgpu_vmid	*reserved;
-	unsigned int		reserved_use_count;
+	bool			reserved_vmid;
 };
 
 int amdgpu_pasid_alloc(unsigned int bits);
@@ -78,11 +77,10 @@ void amdgpu_pasid_free_delayed(struct dma_resv *resv,
 
 bool amdgpu_vmid_had_gpu_reset(struct amdgpu_device *adev,
 			       struct amdgpu_vmid *id);
-int amdgpu_vmid_alloc_reserved(struct amdgpu_device *adev,
-			       struct amdgpu_vm *vm,
+bool amdgpu_vmid_uses_reserved(struct amdgpu_vm *vm, unsigned int vmhub);
+int amdgpu_vmid_alloc_reserved(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			       unsigned vmhub);
-void amdgpu_vmid_free_reserved(struct amdgpu_device *adev,
-			       struct amdgpu_vm *vm,
+void amdgpu_vmid_free_reserved(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			       unsigned vmhub);
 int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
 		     struct amdgpu_job *job, struct dma_fence **fence);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c
index 1d5af50331e4..a6419246e9c2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c
@@ -25,6 +25,7 @@
 
 #include "amdgpu.h"
 #include "amdgpu_ih.h"
+#include "amdgpu_reset.h"
 
 /**
  * amdgpu_ih_ring_init - initialize the IH state
@@ -138,6 +139,7 @@ void amdgpu_ih_ring_fini(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih)
 /**
  * amdgpu_ih_ring_write - write IV to the ring buffer
  *
+ * @adev: amdgpu_device pointer
  * @ih: ih ring to write to
  * @iv: the iv to write
  * @num_dw: size of the iv in dw
@@ -145,8 +147,8 @@ void amdgpu_ih_ring_fini(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih)
  * Writes an IV to the ring buffer using the CPU and increment the wptr.
  * Used for testing and delegating IVs to a software ring.
  */
-void amdgpu_ih_ring_write(struct amdgpu_ih_ring *ih, const uint32_t *iv,
-			  unsigned int num_dw)
+void amdgpu_ih_ring_write(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih,
+			  const uint32_t *iv, unsigned int num_dw)
 {
 	uint32_t wptr = le32_to_cpu(*ih->wptr_cpu) >> 2;
 	unsigned int i;
@@ -161,6 +163,9 @@ void amdgpu_ih_ring_write(struct amdgpu_ih_ring *ih, const uint32_t *iv,
 	if (wptr != READ_ONCE(ih->rptr)) {
 		wmb();
 		WRITE_ONCE(*ih->wptr_cpu, cpu_to_le32(wptr));
+	} else if (adev->irq.retry_cam_enabled) {
+		dev_warn_once(adev->dev, "IH soft ring buffer overflow 0x%X, 0x%X\n",
+			      wptr, ih->rptr);
 	}
 }
 
@@ -213,7 +218,7 @@ int amdgpu_ih_process(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih)
 
 restart_ih:
 	count  = AMDGPU_IH_MAX_NUM_IVS;
-	DRM_DEBUG("%s: rptr %d, wptr %d\n", __func__, ih->rptr, wptr);
+	dev_dbg(adev->dev, "%s: rptr %d, wptr %d\n", __func__, ih->rptr, wptr);
 
 	/* Order reading of wptr vs. reading of IH ring data */
 	rmb();
@@ -223,13 +228,23 @@ restart_ih:
 		ih->rptr &= ih->ptr_mask;
 	}
 
-	amdgpu_ih_set_rptr(adev, ih);
+	if (!ih->overflow)
+		amdgpu_ih_set_rptr(adev, ih);
+
 	wake_up_all(&ih->wait_process);
 
 	/* make sure wptr hasn't changed while processing */
 	wptr = amdgpu_ih_get_wptr(adev, ih);
 	if (wptr != ih->rptr)
-		goto restart_ih;
+		if (!ih->overflow)
+			goto restart_ih;
+
+	if (ih->overflow)
+		if (amdgpu_sriov_runtime(adev))
+			WARN_ONCE(!amdgpu_reset_domain_schedule(adev->reset_domain,
+				   &adev->virt.flr_work),
+				  "Failed to queue work! at %s",
+				  __func__);
 
 	return IRQ_HANDLED;
 }
@@ -270,7 +285,7 @@ void amdgpu_ih_decode_iv_helper(struct amdgpu_device *adev,
 	entry->timestamp = dw[1] | ((u64)(dw[2] & 0xffff) << 32);
 	entry->timestamp_src = dw[2] >> 31;
 	entry->pasid = dw[3] & 0xffff;
-	entry->pasid_src = dw[3] >> 31;
+	entry->node_id = (dw[3] >> 16) & 0xff;
 	entry->src_data[0] = dw[4];
 	entry->src_data[1] = dw[5];
 	entry->src_data[2] = dw[6];
@@ -294,3 +309,9 @@ uint64_t amdgpu_ih_decode_iv_ts_helper(struct amdgpu_ih_ring *ih, u32 rptr,
 	dw2 = le32_to_cpu(ih->ring[ring_index + 2]);
 	return dw1 | ((u64)(dw2 & 0xffff) << 32);
 }
+
+const char *amdgpu_ih_ring_name(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih)
+{
+	return ih == &adev->irq.ih ? "ih" : ih == &adev->irq.ih_soft ? "sw ih" :
+	       ih == &adev->irq.ih1 ? "ih1" : ih == &adev->irq.ih2 ? "ih2" : "unknown";
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h
index dd1c2eded6b9..f58b6be7fccc 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h
@@ -27,6 +27,9 @@
 /* Maximum number of IVs processed at once */
 #define AMDGPU_IH_MAX_NUM_IVS	32
 
+#define IH_RING_SIZE	(256 * 1024)
+#define IH_SW_RING_SIZE	(16 * 1024)	/* enough for 512 CAM entries */
+
 struct amdgpu_device;
 struct amdgpu_iv_entry;
 
@@ -53,14 +56,14 @@ struct amdgpu_ih_ring {
 	bool			use_bus_addr;
 
 	struct amdgpu_bo	*ring_obj;
-	volatile uint32_t	*ring;
+	uint32_t		*ring;
 	uint64_t		gpu_addr;
 
 	uint64_t		wptr_addr;
-	volatile uint32_t	*wptr_cpu;
+	uint32_t		*wptr_cpu;
 
 	uint64_t		rptr_addr;
-	volatile uint32_t	*rptr_cpu;
+	uint32_t		*rptr_cpu;
 
 	bool                    enabled;
 	unsigned		rptr;
@@ -69,12 +72,16 @@ struct amdgpu_ih_ring {
 	/* For waiting on IH processing at checkpoint. */
 	wait_queue_head_t wait_process;
 	uint64_t		processed_timestamp;
+	bool overflow;
 };
 
 /* return true if time stamp t2 is after t1 with 48bit wrap around */
 #define amdgpu_ih_ts_after(t1, t2) \
 		(((int64_t)((t2) << 16) - (int64_t)((t1) << 16)) > 0LL)
 
+#define amdgpu_ih_ts_after_or_equal(t1, t2) \
+		(((int64_t)((t2) << 16) - (int64_t)((t1) << 16)) >= 0LL)
+
 /* provided by the ih block */
 struct amdgpu_ih_funcs {
 	/* ring read/write ptr handling, called from interrupt context */
@@ -97,8 +104,8 @@ struct amdgpu_ih_funcs {
 int amdgpu_ih_ring_init(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih,
 			unsigned ring_size, bool use_bus_addr);
 void amdgpu_ih_ring_fini(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih);
-void amdgpu_ih_ring_write(struct amdgpu_ih_ring *ih, const uint32_t *iv,
-			  unsigned int num_dw);
+void amdgpu_ih_ring_write(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih,
+			  const uint32_t *iv, unsigned int num_dw);
 int amdgpu_ih_wait_on_checkpoint_process_ts(struct amdgpu_device *adev,
 					    struct amdgpu_ih_ring *ih);
 int amdgpu_ih_process(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih);
@@ -107,4 +114,5 @@ void amdgpu_ih_decode_iv_helper(struct amdgpu_device *adev,
 				struct amdgpu_iv_entry *entry);
 uint64_t amdgpu_ih_decode_iv_ts_helper(struct amdgpu_ih_ring *ih, u32 rptr,
 				       signed int offset);
+const char *amdgpu_ih_ring_name(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ip.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ip.c
new file mode 100644
index 000000000000..99e1cf4fc955
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ip.c
@@ -0,0 +1,96 @@
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "amdgpu.h"
+#include "amdgpu_ip.h"
+
+static int8_t amdgpu_logical_to_dev_inst(struct amdgpu_device *adev,
+					 enum amd_hw_ip_block_type block,
+					 int8_t inst)
+{
+	int8_t dev_inst;
+
+	switch (block) {
+	case GC_HWIP:
+	case SDMA0_HWIP:
+	/* Both JPEG and VCN as JPEG is only alias of VCN */
+	case VCN_HWIP:
+		dev_inst = adev->ip_map.dev_inst[block][inst];
+		break;
+	default:
+		/* For rest of the IPs, no look up required.
+		 * Assume 'logical instance == physical instance' for all configs. */
+		dev_inst = inst;
+		break;
+	}
+
+	return dev_inst;
+}
+
+static uint32_t amdgpu_logical_to_dev_mask(struct amdgpu_device *adev,
+					   enum amd_hw_ip_block_type block,
+					   uint32_t mask)
+{
+	uint32_t dev_mask = 0;
+	int8_t log_inst, dev_inst;
+
+	while (mask) {
+		log_inst = ffs(mask) - 1;
+		dev_inst = amdgpu_logical_to_dev_inst(adev, block, log_inst);
+		dev_mask |= (1 << dev_inst);
+		mask &= ~(1 << log_inst);
+	}
+
+	return dev_mask;
+}
+
+static void amdgpu_populate_ip_map(struct amdgpu_device *adev,
+				   enum amd_hw_ip_block_type ip_block,
+				   uint32_t inst_mask)
+{
+	int l = 0, i;
+
+	while (inst_mask) {
+		i = ffs(inst_mask) - 1;
+		adev->ip_map.dev_inst[ip_block][l++] = i;
+		inst_mask &= ~(1 << i);
+	}
+	for (; l < HWIP_MAX_INSTANCE; l++)
+		adev->ip_map.dev_inst[ip_block][l] = -1;
+}
+
+void amdgpu_ip_map_init(struct amdgpu_device *adev)
+{
+	u32 ip_map[][2] = {
+		{ GC_HWIP, adev->gfx.xcc_mask },
+		{ SDMA0_HWIP, adev->sdma.sdma_mask },
+		{ VCN_HWIP, adev->vcn.inst_mask },
+	};
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(ip_map); ++i)
+		amdgpu_populate_ip_map(adev, ip_map[i][0], ip_map[i][1]);
+
+	adev->ip_map.logical_to_dev_inst = amdgpu_logical_to_dev_inst;
+	adev->ip_map.logical_to_dev_mask = amdgpu_logical_to_dev_mask;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ip.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ip.h
new file mode 100644
index 000000000000..2490fd322aec
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ip.h
@@ -0,0 +1,29 @@
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __AMDGPU_IP_H__
+#define __AMDGPU_IP_H__
+
+void amdgpu_ip_map_init(struct amdgpu_device *adev);
+
+#endif /* __AMDGPU_IP_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
index fafebec5b7b6..8112ffc85995 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
@@ -99,6 +99,21 @@ const char *soc15_ih_clientid_name[] = {
 	"MP1"
 };
 
+const int node_id_to_phys_map[NODEID_MAX] = {
+	[AID0_NODEID] = 0,
+	[XCD0_NODEID] = 0,
+	[XCD1_NODEID] = 1,
+	[AID1_NODEID] = 1,
+	[XCD2_NODEID] = 2,
+	[XCD3_NODEID] = 3,
+	[AID2_NODEID] = 2,
+	[XCD4_NODEID] = 4,
+	[XCD5_NODEID] = 5,
+	[AID3_NODEID] = 3,
+	[XCD6_NODEID] = 6,
+	[XCD7_NODEID] = 7,
+};
+
 /**
  * amdgpu_irq_disable_all - disable *all* interrupts
  *
@@ -109,7 +124,7 @@ const char *soc15_ih_clientid_name[] = {
 void amdgpu_irq_disable_all(struct amdgpu_device *adev)
 {
 	unsigned long irqflags;
-	unsigned i, j, k;
+	unsigned int i, j, k;
 	int r;
 
 	spin_lock_irqsave(&adev->irq.lock, irqflags);
@@ -124,12 +139,12 @@ void amdgpu_irq_disable_all(struct amdgpu_device *adev)
 				continue;
 
 			for (k = 0; k < src->num_types; ++k) {
-				atomic_set(&src->enabled_types[k], 0);
 				r = src->funcs->set(adev, src, k,
 						    AMDGPU_IRQ_STATE_DISABLE);
 				if (r)
-					DRM_ERROR("error disabling interrupt (%d)\n",
-						  r);
+					dev_err(adev->dev,
+						"error disabling interrupt (%d)\n",
+						r);
 			}
 		}
 	}
@@ -228,7 +243,7 @@ static bool amdgpu_msi_ok(struct amdgpu_device *adev)
 	return true;
 }
 
-static void amdgpu_restore_msix(struct amdgpu_device *adev)
+void amdgpu_restore_msix(struct amdgpu_device *adev)
 {
 	u16 ctrl;
 
@@ -256,29 +271,29 @@ static void amdgpu_restore_msix(struct amdgpu_device *adev)
  */
 int amdgpu_irq_init(struct amdgpu_device *adev)
 {
-	int r = 0;
-	unsigned int irq;
+	unsigned int irq, flags;
+	int r;
 
 	spin_lock_init(&adev->irq.lock);
 
 	/* Enable MSI if not disabled by module parameter */
 	adev->irq.msi_enabled = false;
 
-	if (amdgpu_msi_ok(adev)) {
-		int nvec = pci_msix_vec_count(adev->pdev);
-		unsigned int flags;
+	if (!amdgpu_msi_ok(adev))
+		flags = PCI_IRQ_INTX;
+	else
+		flags = PCI_IRQ_ALL_TYPES;
 
-		if (nvec <= 0) {
-			flags = PCI_IRQ_MSI;
-		} else {
-			flags = PCI_IRQ_MSI | PCI_IRQ_MSIX;
-		}
-		/* we only need one vector */
-		nvec = pci_alloc_irq_vectors(adev->pdev, 1, 1, flags);
-		if (nvec > 0) {
-			adev->irq.msi_enabled = true;
-			dev_dbg(adev->dev, "using MSI/MSI-X.\n");
-		}
+	/* we only need one vector */
+	r = pci_alloc_irq_vectors(adev->pdev, 1, 1, flags);
+	if (r < 0) {
+		dev_err(adev->dev, "Failed to alloc msi vectors\n");
+		return r;
+	}
+
+	if (amdgpu_msi_ok(adev)) {
+		adev->irq.msi_enabled = true;
+		dev_dbg(adev->dev, "using MSI/MSI-X.\n");
 	}
 
 	INIT_WORK(&adev->irq.ih1_work, amdgpu_irq_handle_ih1);
@@ -288,22 +303,29 @@ int amdgpu_irq_init(struct amdgpu_device *adev)
 	/* Use vector 0 for MSI-X. */
 	r = pci_irq_vector(adev->pdev, 0);
 	if (r < 0)
-		return r;
+		goto free_vectors;
 	irq = r;
 
 	/* PCI devices require shared interrupts. */
 	r = request_irq(irq, amdgpu_irq_handler, IRQF_SHARED, adev_to_drm(adev)->driver->name,
 			adev_to_drm(adev));
 	if (r)
-		return r;
+		goto free_vectors;
+
 	adev->irq.installed = true;
 	adev->irq.irq = irq;
 	adev_to_drm(adev)->max_vblank_count = 0x00ffffff;
 
-	DRM_DEBUG("amdgpu: irq initialized.\n");
+	dev_dbg(adev->dev, "amdgpu: irq initialized.\n");
 	return 0;
-}
 
+free_vectors:
+	if (adev->irq.msi_enabled)
+		pci_free_irq_vectors(adev->pdev);
+
+	adev->irq.msi_enabled = false;
+	return r;
+}
 
 void amdgpu_irq_fini_hw(struct amdgpu_device *adev)
 {
@@ -331,7 +353,7 @@ void amdgpu_irq_fini_hw(struct amdgpu_device *adev)
  */
 void amdgpu_irq_fini_sw(struct amdgpu_device *adev)
 {
-	unsigned i, j;
+	unsigned int i, j;
 
 	for (i = 0; i < AMDGPU_IRQ_CLIENTID_MAX; ++i) {
 		if (!adev->irq.client[i].sources)
@@ -365,7 +387,7 @@ void amdgpu_irq_fini_sw(struct amdgpu_device *adev)
  * 0 on success or error code otherwise
  */
 int amdgpu_irq_add_id(struct amdgpu_device *adev,
-		      unsigned client_id, unsigned src_id,
+		      unsigned int client_id, unsigned int src_id,
 		      struct amdgpu_irq_src *source)
 {
 	if (client_id >= AMDGPU_IRQ_CLIENTID_MAX)
@@ -417,13 +439,21 @@ void amdgpu_irq_dispatch(struct amdgpu_device *adev,
 {
 	u32 ring_index = ih->rptr >> 2;
 	struct amdgpu_iv_entry entry;
-	unsigned client_id, src_id;
+	unsigned int client_id, src_id;
 	struct amdgpu_irq_src *src;
 	bool handled = false;
 	int r;
 
 	entry.ih = ih;
 	entry.iv_entry = (const uint32_t *)&ih->ring[ring_index];
+
+	/*
+	 * timestamp is not supported on some legacy SOCs (cik, cz, iceland,
+	 * si and tonga), so initialize timestamp and timestamp_src to 0
+	 */
+	entry.timestamp = 0;
+	entry.timestamp_src = 0;
+
 	amdgpu_ih_decode_iv(adev, &entry);
 
 	trace_amdgpu_iv(ih - &adev->irq.ih, &entry);
@@ -432,28 +462,33 @@ void amdgpu_irq_dispatch(struct amdgpu_device *adev,
 	src_id = entry.src_id;
 
 	if (client_id >= AMDGPU_IRQ_CLIENTID_MAX) {
-		DRM_DEBUG("Invalid client_id in IV: %d\n", client_id);
+		dev_dbg(adev->dev, "Invalid client_id in IV: %d\n", client_id);
 
 	} else	if (src_id >= AMDGPU_MAX_IRQ_SRC_ID) {
-		DRM_DEBUG("Invalid src_id in IV: %d\n", src_id);
+		dev_dbg(adev->dev, "Invalid src_id in IV: %d\n", src_id);
 
-	} else if ((client_id == AMDGPU_IRQ_CLIENTID_LEGACY) &&
+	} else if (((client_id == AMDGPU_IRQ_CLIENTID_LEGACY) ||
+		    (client_id == SOC15_IH_CLIENTID_ISP)) &&
 		   adev->irq.virq[src_id]) {
 		generic_handle_domain_irq(adev->irq.domain, src_id);
 
 	} else if (!adev->irq.client[client_id].sources) {
-		DRM_DEBUG("Unregistered interrupt client_id: %d src_id: %d\n",
-			  client_id, src_id);
+		dev_dbg(adev->dev,
+			"Unregistered interrupt client_id: %d src_id: %d\n",
+			client_id, src_id);
 
 	} else if ((src = adev->irq.client[client_id].sources[src_id])) {
 		r = src->funcs->process(adev, src, &entry);
 		if (r < 0)
-			DRM_ERROR("error processing interrupt (%d)\n", r);
+			dev_err(adev->dev, "error processing interrupt (%d)\n",
+				r);
 		else if (r)
 			handled = true;
 
 	} else {
-		DRM_DEBUG("Unhandled interrupt src_id: %d\n", src_id);
+		dev_dbg(adev->dev,
+			"Unregistered interrupt src_id: %d of client_id:%d\n",
+			src_id, client_id);
 	}
 
 	/* Send it to amdkfd as well if it isn't already handled */
@@ -478,7 +513,7 @@ void amdgpu_irq_delegate(struct amdgpu_device *adev,
 			 struct amdgpu_iv_entry *entry,
 			 unsigned int num_dw)
 {
-	amdgpu_ih_ring_write(&adev->irq.ih_soft, entry->iv_entry, num_dw);
+	amdgpu_ih_ring_write(adev, &adev->irq.ih_soft, entry->iv_entry, num_dw);
 	schedule_work(&adev->irq.ih_soft_work);
 }
 
@@ -492,7 +527,7 @@ void amdgpu_irq_delegate(struct amdgpu_device *adev,
  * Updates interrupt state for the specific source (all ASICs).
  */
 int amdgpu_irq_update(struct amdgpu_device *adev,
-			     struct amdgpu_irq_src *src, unsigned type)
+			     struct amdgpu_irq_src *src, unsigned int type)
 {
 	unsigned long irqflags;
 	enum amdgpu_interrupt_state state;
@@ -501,7 +536,8 @@ int amdgpu_irq_update(struct amdgpu_device *adev,
 	spin_lock_irqsave(&adev->irq.lock, irqflags);
 
 	/* We need to determine after taking the lock, otherwise
-	   we might disable just enabled interrupts again */
+	 * we might disable just enabled interrupts again
+	 */
 	if (amdgpu_irq_enabled(adev, src, type))
 		state = AMDGPU_IRQ_STATE_ENABLE;
 	else
@@ -555,7 +591,7 @@ void amdgpu_irq_gpu_reset_resume_helper(struct amdgpu_device *adev)
  * 0 on success or error code otherwise
  */
 int amdgpu_irq_get(struct amdgpu_device *adev, struct amdgpu_irq_src *src,
-		   unsigned type)
+		   unsigned int type)
 {
 	if (!adev->irq.installed)
 		return -ENOENT;
@@ -585,8 +621,12 @@ int amdgpu_irq_get(struct amdgpu_device *adev, struct amdgpu_irq_src *src,
  * 0 on success or error code otherwise
  */
 int amdgpu_irq_put(struct amdgpu_device *adev, struct amdgpu_irq_src *src,
-		   unsigned type)
+		   unsigned int type)
 {
+	/* When the threshold is reached,the interrupt source may not be enabled.return -EINVAL */
+	if (amdgpu_ras_is_rma(adev) && !amdgpu_irq_enabled(adev, src, type))
+		return -EINVAL;
+
 	if (!adev->irq.installed)
 		return -ENOENT;
 
@@ -619,7 +659,7 @@ int amdgpu_irq_put(struct amdgpu_device *adev, struct amdgpu_irq_src *src,
  * invalid parameters
  */
 bool amdgpu_irq_enabled(struct amdgpu_device *adev, struct amdgpu_irq_src *src,
-			unsigned type)
+			unsigned int type)
 {
 	if (!adev->irq.installed)
 		return false;
@@ -693,10 +733,10 @@ static const struct irq_domain_ops amdgpu_hw_irqdomain_ops = {
  */
 int amdgpu_irq_add_domain(struct amdgpu_device *adev)
 {
-	adev->irq.domain = irq_domain_add_linear(NULL, AMDGPU_MAX_IRQ_SRC_ID,
-						 &amdgpu_hw_irqdomain_ops, adev);
+	adev->irq.domain = irq_domain_create_linear(NULL, AMDGPU_MAX_IRQ_SRC_ID,
+						    &amdgpu_hw_irqdomain_ops, adev);
 	if (!adev->irq.domain) {
-		DRM_ERROR("GPU irq add domain failed\n");
+		dev_err(adev->dev, "GPU irq add domain failed\n");
 		return -ENODEV;
 	}
 
@@ -732,7 +772,7 @@ void amdgpu_irq_remove_domain(struct amdgpu_device *adev)
  * Returns:
  * Linux IRQ
  */
-unsigned amdgpu_irq_create_mapping(struct amdgpu_device *adev, unsigned src_id)
+unsigned int amdgpu_irq_create_mapping(struct amdgpu_device *adev, unsigned int src_id)
 {
 	adev->irq.virq[src_id] = irq_create_mapping(adev->irq.domain, src_id);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h
index be243adf3e65..9f0417456abd 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h
@@ -53,7 +53,7 @@ struct amdgpu_iv_entry {
 	uint64_t timestamp;
 	unsigned timestamp_src;
 	unsigned pasid;
-	unsigned pasid_src;
+	unsigned node_id;
 	unsigned src_data[AMDGPU_IRQ_SRC_DATA_MAX_SIZE_DW];
 	const uint32_t *iv_entry;
 };
@@ -102,6 +102,24 @@ struct amdgpu_irq {
 	bool                            retry_cam_enabled;
 };
 
+enum interrupt_node_id_per_aid {
+	AID0_NODEID = 0,
+	XCD0_NODEID = 1,
+	XCD1_NODEID = 2,
+	AID1_NODEID = 4,
+	XCD2_NODEID = 5,
+	XCD3_NODEID = 6,
+	AID2_NODEID = 8,
+	XCD4_NODEID = 9,
+	XCD5_NODEID = 10,
+	AID3_NODEID = 12,
+	XCD6_NODEID = 13,
+	XCD7_NODEID = 14,
+	NODEID_MAX,
+};
+
+extern const int node_id_to_phys_map[NODEID_MAX];
+
 void amdgpu_irq_disable_all(struct amdgpu_device *adev);
 
 int amdgpu_irq_init(struct amdgpu_device *adev);
@@ -128,5 +146,6 @@ void amdgpu_irq_gpu_reset_resume_helper(struct amdgpu_device *adev);
 int amdgpu_irq_add_domain(struct amdgpu_device *adev);
 void amdgpu_irq_remove_domain(struct amdgpu_device *adev);
 unsigned amdgpu_irq_create_mapping(struct amdgpu_device *adev, unsigned src_id);
+void amdgpu_restore_msix(struct amdgpu_device *adev);
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_isp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_isp.c
new file mode 100644
index 000000000000..37270c4dab8d
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_isp.c
@@ -0,0 +1,345 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright (C) 2024 Advanced Micro Devices, Inc. All rights reserved.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ */
+
+#include <linux/firmware.h>
+#include <linux/mfd/core.h>
+
+#include "amdgpu.h"
+#include "amdgpu_isp.h"
+#include "isp_v4_1_0.h"
+#include "isp_v4_1_1.h"
+
+#define ISP_MC_ADDR_ALIGN (1024 * 32)
+
+/**
+ * isp_hw_init - start and test isp block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ */
+static int isp_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_isp *isp = &adev->isp;
+
+	if (isp->funcs->hw_init != NULL)
+		return isp->funcs->hw_init(isp);
+
+	return -ENODEV;
+}
+
+/**
+ * isp_hw_fini - stop the hardware block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ */
+static int isp_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_isp *isp = &ip_block->adev->isp;
+
+	if (isp->funcs->hw_fini != NULL)
+		return isp->funcs->hw_fini(isp);
+
+	return -ENODEV;
+}
+
+static int isp_load_fw_by_psp(struct amdgpu_device *adev)
+{
+	const struct common_firmware_header *hdr;
+	char ucode_prefix[10];
+	int r = 0;
+
+	/* get isp fw binary name and path */
+	amdgpu_ucode_ip_version_decode(adev, ISP_HWIP, ucode_prefix,
+				       sizeof(ucode_prefix));
+
+	/* read isp fw */
+	r = amdgpu_ucode_request(adev, &adev->isp.fw, AMDGPU_UCODE_OPTIONAL,
+				"amdgpu/%s.bin", ucode_prefix);
+	if (r) {
+		amdgpu_ucode_release(&adev->isp.fw);
+		return r;
+	}
+
+	hdr = (const struct common_firmware_header *)adev->isp.fw->data;
+
+	adev->firmware.ucode[AMDGPU_UCODE_ID_ISP].ucode_id =
+		AMDGPU_UCODE_ID_ISP;
+	adev->firmware.ucode[AMDGPU_UCODE_ID_ISP].fw = adev->isp.fw;
+
+	adev->firmware.fw_size +=
+		ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE);
+
+	return r;
+}
+
+static int isp_early_init(struct amdgpu_ip_block *ip_block)
+{
+
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_isp *isp = &adev->isp;
+
+	switch (amdgpu_ip_version(adev, ISP_HWIP, 0)) {
+	case IP_VERSION(4, 1, 0):
+		isp_v4_1_0_set_isp_funcs(isp);
+		break;
+	case IP_VERSION(4, 1, 1):
+		isp_v4_1_1_set_isp_funcs(isp);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	isp->adev = adev;
+	isp->parent = adev->dev;
+
+	if (isp_load_fw_by_psp(adev)) {
+		DRM_DEBUG_DRIVER("%s: isp fw load failed\n", __func__);
+		return -ENOENT;
+	}
+
+	return 0;
+}
+
+static bool isp_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	return true;
+}
+
+static int isp_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+				     enum amd_clockgating_state state)
+{
+	return 0;
+}
+
+static int isp_set_powergating_state(struct amdgpu_ip_block *ip_block,
+				     enum amd_powergating_state state)
+{
+	return 0;
+}
+
+static int is_valid_isp_device(struct device *isp_parent, struct device *amdgpu_dev)
+{
+	if (isp_parent != amdgpu_dev)
+		return -EINVAL;
+
+	return 0;
+}
+
+/**
+ * isp_user_buffer_alloc - create user buffer object (BO) for isp
+ *
+ * @dev: isp device handle
+ * @dmabuf: DMABUF handle for isp buffer allocated in system memory
+ * @buf_obj: GPU buffer object handle to initialize
+ * @buf_addr: GPU addr of the pinned BO to initialize
+ *
+ * Imports isp DMABUF to allocate and pin a user BO for isp internal use. It does
+ * GART alloc to generate GPU addr for BO to make it accessible through the
+ * GART aperture for ISP HW.
+ *
+ * This function is exported to allow the V4L2 isp device external to drm device
+ * to create and access the isp user BO.
+ *
+ * Returns:
+ * 0 on success, negative error code otherwise.
+ */
+int isp_user_buffer_alloc(struct device *dev, void *dmabuf,
+			  void **buf_obj, u64 *buf_addr)
+{
+	struct platform_device *ispdev = to_platform_device(dev);
+	const struct isp_platform_data *isp_pdata;
+	struct amdgpu_device *adev;
+	struct mfd_cell *mfd_cell;
+	struct amdgpu_bo *bo;
+	u64 gpu_addr;
+	int ret;
+
+	if (WARN_ON(!ispdev))
+		return -ENODEV;
+
+	if (WARN_ON(!buf_obj))
+		return -EINVAL;
+
+	if (WARN_ON(!buf_addr))
+		return -EINVAL;
+
+	mfd_cell = &ispdev->mfd_cell[0];
+	if (!mfd_cell)
+		return -ENODEV;
+
+	isp_pdata = mfd_cell->platform_data;
+	adev = isp_pdata->adev;
+
+	ret = is_valid_isp_device(ispdev->dev.parent, adev->dev);
+	if (ret)
+		return ret;
+
+	ret = amdgpu_bo_create_isp_user(adev, dmabuf,
+					AMDGPU_GEM_DOMAIN_GTT, &bo, &gpu_addr);
+	if (ret) {
+		drm_err(&adev->ddev, "failed to alloc gart user buffer (%d)", ret);
+		return ret;
+	}
+
+	*buf_obj = (void *)bo;
+	*buf_addr = gpu_addr;
+
+	return 0;
+}
+EXPORT_SYMBOL(isp_user_buffer_alloc);
+
+/**
+ * isp_user_buffer_free - free isp user buffer object (BO)
+ *
+ * @buf_obj: amdgpu isp user BO to free
+ *
+ * unpin and unref BO for isp internal use.
+ *
+ * This function is exported to allow the V4L2 isp device
+ * external to drm device to free the isp user BO.
+ */
+void isp_user_buffer_free(void *buf_obj)
+{
+	amdgpu_bo_free_isp_user(buf_obj);
+}
+EXPORT_SYMBOL(isp_user_buffer_free);
+
+/**
+ * isp_kernel_buffer_alloc - create kernel buffer object (BO) for isp
+ *
+ * @dev: isp device handle
+ * @size: size for the new BO
+ * @buf_obj: GPU BO handle to initialize
+ * @gpu_addr: GPU addr of the pinned BO
+ * @cpu_addr: CPU address mapping of BO
+ *
+ * Allocates and pins a kernel BO for internal isp firmware use.
+ *
+ * This function is exported to allow the V4L2 isp device
+ * external to drm device to create and access the kernel BO.
+ *
+ * Returns:
+ * 0 on success, negative error code otherwise.
+ */
+int isp_kernel_buffer_alloc(struct device *dev, u64 size,
+			    void **buf_obj, u64 *gpu_addr, void **cpu_addr)
+{
+	struct platform_device *ispdev = to_platform_device(dev);
+	struct amdgpu_bo **bo = (struct amdgpu_bo **)buf_obj;
+	const struct isp_platform_data *isp_pdata;
+	struct amdgpu_device *adev;
+	struct mfd_cell *mfd_cell;
+	int ret;
+
+	if (WARN_ON(!ispdev))
+		return -ENODEV;
+
+	if (WARN_ON(!buf_obj))
+		return -EINVAL;
+
+	if (WARN_ON(!gpu_addr))
+		return -EINVAL;
+
+	if (WARN_ON(!cpu_addr))
+		return -EINVAL;
+
+	mfd_cell = &ispdev->mfd_cell[0];
+	if (!mfd_cell)
+		return -ENODEV;
+
+	isp_pdata = mfd_cell->platform_data;
+	adev = isp_pdata->adev;
+
+	ret = is_valid_isp_device(ispdev->dev.parent, adev->dev);
+	if (ret)
+		return ret;
+
+	/* Ensure *bo is NULL so a new BO will be created */
+	*bo = NULL;
+	ret = amdgpu_bo_create_kernel(adev,
+				      size,
+				      ISP_MC_ADDR_ALIGN,
+				      AMDGPU_GEM_DOMAIN_GTT,
+				      bo,
+				      gpu_addr,
+				      cpu_addr);
+	if (!cpu_addr || ret) {
+		drm_err(&adev->ddev, "failed to alloc gart kernel buffer (%d)", ret);
+		return ret;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL(isp_kernel_buffer_alloc);
+
+/**
+ * isp_kernel_buffer_free - free isp kernel buffer object (BO)
+ *
+ * @buf_obj: amdgpu isp user BO to free
+ * @gpu_addr: GPU addr of isp kernel BO
+ * @cpu_addr: CPU addr of isp kernel BO
+ *
+ * unmaps and unpin a isp kernel BO.
+ *
+ * This function is exported to allow the V4L2 isp device
+ * external to drm device to free the kernel BO.
+ */
+void isp_kernel_buffer_free(void **buf_obj, u64 *gpu_addr, void **cpu_addr)
+{
+	struct amdgpu_bo **bo = (struct amdgpu_bo **)buf_obj;
+
+	amdgpu_bo_free_kernel(bo, gpu_addr, cpu_addr);
+}
+EXPORT_SYMBOL(isp_kernel_buffer_free);
+
+static const struct amd_ip_funcs isp_ip_funcs = {
+	.name = "isp_ip",
+	.early_init = isp_early_init,
+	.hw_init = isp_hw_init,
+	.hw_fini = isp_hw_fini,
+	.is_idle = isp_is_idle,
+	.set_clockgating_state = isp_set_clockgating_state,
+	.set_powergating_state = isp_set_powergating_state,
+};
+
+const struct amdgpu_ip_block_version isp_v4_1_0_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_ISP,
+	.major = 4,
+	.minor = 1,
+	.rev = 0,
+	.funcs = &isp_ip_funcs,
+};
+
+const struct amdgpu_ip_block_version isp_v4_1_1_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_ISP,
+	.major = 4,
+	.minor = 1,
+	.rev = 1,
+	.funcs = &isp_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_isp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_isp.h
new file mode 100644
index 000000000000..d6f4ffa4c97c
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_isp.h
@@ -0,0 +1,60 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright (C) 2024 Advanced Micro Devices, Inc. All rights reserved.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ */
+
+#ifndef __AMDGPU_ISP_H__
+#define __AMDGPU_ISP_H__
+
+#include <drm/amd/isp.h>
+#include <linux/pm_domain.h>
+
+#define ISP_REGS_OFFSET_END 0x629A4
+
+struct amdgpu_isp;
+
+struct isp_funcs {
+	int (*hw_init)(struct amdgpu_isp *isp);
+	int (*hw_fini)(struct amdgpu_isp *isp);
+};
+
+struct amdgpu_isp {
+	struct device *parent;
+	struct amdgpu_device	*adev;
+	const struct isp_funcs	*funcs;
+	struct mfd_cell *isp_cell;
+	struct resource *isp_res;
+	struct resource *isp_i2c_res;
+	struct resource *isp_gpio_res;
+	struct isp_platform_data *isp_pdata;
+	unsigned int harvest_config;
+	const struct firmware	*fw;
+	struct generic_pm_domain ispgpd;
+};
+
+extern const struct amdgpu_ip_block_version isp_v4_1_0_ip_block;
+extern const struct amdgpu_ip_block_version isp_v4_1_1_ip_block;
+
+#endif /* __AMDGPU_ISP_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
index c3d9d75143f4..0a0dcbf0798d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
@@ -30,40 +30,124 @@
 #include "amdgpu.h"
 #include "amdgpu_trace.h"
 #include "amdgpu_reset.h"
+#include "amdgpu_dev_coredump.h"
+#include "amdgpu_xgmi.h"
+
+static void amdgpu_job_do_core_dump(struct amdgpu_device *adev,
+				    struct amdgpu_job *job)
+{
+	int i;
+
+	dev_info(adev->dev, "Dumping IP State\n");
+	for (i = 0; i < adev->num_ip_blocks; i++)
+		if (adev->ip_blocks[i].version->funcs->dump_ip_state)
+			adev->ip_blocks[i].version->funcs
+				->dump_ip_state((void *)&adev->ip_blocks[i]);
+	dev_info(adev->dev, "Dumping IP State Completed\n");
+
+	amdgpu_coredump(adev, true, false, job);
+}
+
+static void amdgpu_job_core_dump(struct amdgpu_device *adev,
+				 struct amdgpu_job *job)
+{
+	struct list_head device_list, *device_list_handle =  NULL;
+	struct amdgpu_device *tmp_adev = NULL;
+	struct amdgpu_hive_info *hive = NULL;
+
+	if (!amdgpu_sriov_vf(adev))
+		hive = amdgpu_get_xgmi_hive(adev);
+	if (hive)
+		mutex_lock(&hive->hive_lock);
+	/*
+	 * Reuse the logic in amdgpu_device_gpu_recover() to build list of
+	 * devices for code dump
+	 */
+	INIT_LIST_HEAD(&device_list);
+	if (!amdgpu_sriov_vf(adev) && (adev->gmc.xgmi.num_physical_nodes > 1) && hive) {
+		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head)
+			list_add_tail(&tmp_adev->reset_list, &device_list);
+		if (!list_is_first(&adev->reset_list, &device_list))
+			list_rotate_to_front(&adev->reset_list, &device_list);
+		device_list_handle = &device_list;
+	} else {
+		list_add_tail(&adev->reset_list, &device_list);
+		device_list_handle = &device_list;
+	}
+
+	/* Do the coredump for each device */
+	list_for_each_entry(tmp_adev, device_list_handle, reset_list)
+		amdgpu_job_do_core_dump(tmp_adev, job);
+
+	if (hive) {
+		mutex_unlock(&hive->hive_lock);
+		amdgpu_put_xgmi_hive(hive);
+	}
+}
 
 static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
 {
 	struct amdgpu_ring *ring = to_amdgpu_ring(s_job->sched);
 	struct amdgpu_job *job = to_amdgpu_job(s_job);
-	struct amdgpu_task_info ti;
+	struct drm_wedge_task_info *info = NULL;
+	struct amdgpu_task_info *ti = NULL;
 	struct amdgpu_device *adev = ring->adev;
-	int idx;
-	int r;
+	int idx, r;
 
 	if (!drm_dev_enter(adev_to_drm(adev), &idx)) {
-		DRM_INFO("%s - device unplugged skipping recovery on scheduler:%s",
+		dev_info(adev->dev, "%s - device unplugged skipping recovery on scheduler:%s",
 			 __func__, s_job->sched->name);
 
 		/* Effectively the job is aborted as the device is gone */
 		return DRM_GPU_SCHED_STAT_ENODEV;
 	}
 
-	memset(&ti, 0, sizeof(struct amdgpu_task_info));
-	adev->job_hang = true;
+	/*
+	 * Do the coredump immediately after a job timeout to get a very
+	 * close dump/snapshot/representation of GPU's current error status
+	 * Skip it for SRIOV, since VF FLR will be triggered by host driver
+	 * before job timeout
+	 */
+	if (!amdgpu_sriov_vf(adev))
+		amdgpu_job_core_dump(adev, job);
 
 	if (amdgpu_gpu_recovery &&
+	    amdgpu_ring_is_reset_type_supported(ring, AMDGPU_RESET_TYPE_SOFT_RESET) &&
 	    amdgpu_ring_soft_recovery(ring, job->vmid, s_job->s_fence->parent)) {
-		DRM_ERROR("ring %s timeout, but soft recovered\n",
-			  s_job->sched->name);
+		dev_err(adev->dev, "ring %s timeout, but soft recovered\n",
+			s_job->sched->name);
 		goto exit;
 	}
 
-	amdgpu_vm_get_task_info(ring->adev, job->pasid, &ti);
-	DRM_ERROR("ring %s timeout, signaled seq=%u, emitted seq=%u\n",
-		  job->base.sched->name, atomic_read(&ring->fence_drv.last_seq),
-		  ring->fence_drv.sync_seq);
-	DRM_ERROR("Process information: process %s pid %d thread %s pid %d\n",
-		  ti.process_name, ti.tgid, ti.task_name, ti.pid);
+	dev_err(adev->dev, "ring %s timeout, signaled seq=%u, emitted seq=%u\n",
+		job->base.sched->name, atomic_read(&ring->fence_drv.last_seq),
+		ring->fence_drv.sync_seq);
+
+	ti = amdgpu_vm_get_task_info_pasid(ring->adev, job->pasid);
+	if (ti) {
+		amdgpu_vm_print_task_info(adev, ti);
+		info = &ti->task;
+	}
+
+	/* attempt a per ring reset */
+	if (amdgpu_gpu_recovery &&
+	    amdgpu_ring_is_reset_type_supported(ring, AMDGPU_RESET_TYPE_PER_QUEUE) &&
+	    ring->funcs->reset) {
+		dev_err(adev->dev, "Starting %s ring reset\n",
+			s_job->sched->name);
+		r = amdgpu_ring_reset(ring, job->vmid, job->hw_fence);
+		if (!r) {
+			atomic_inc(&ring->adev->gpu_reset_counter);
+			dev_err(adev->dev, "Ring %s reset succeeded\n",
+				ring->sched.name);
+			drm_dev_wedged_event(adev_to_drm(adev),
+					     DRM_WEDGE_RECOVERY_NONE, info);
+			goto exit;
+		}
+		dev_err(adev->dev, "Ring %s reset failed\n", ring->sched.name);
+	}
+
+	dma_fence_set_error(&s_job->s_fence->finished, -ETIME);
 
 	if (amdgpu_device_should_recover_gpu(ring->adev)) {
 		struct amdgpu_reset_context reset_context;
@@ -71,11 +155,18 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
 
 		reset_context.method = AMD_RESET_METHOD_NONE;
 		reset_context.reset_req_dev = adev;
+		reset_context.src = AMDGPU_RESET_SRC_JOB;
 		clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
 
+		/*
+		 * To avoid an unnecessary extra coredump, as we have already
+		 * got the very close representation of GPU's error status
+		 */
+		set_bit(AMDGPU_SKIP_COREDUMP, &reset_context.flags);
+
 		r = amdgpu_device_gpu_recover(ring->adev, job, &reset_context);
 		if (r)
-			DRM_ERROR("GPU Recovery Failed: %d\n", r);
+			dev_err(adev->dev, "GPU Recovery Failed: %d\n", r);
 	} else {
 		drm_sched_suspend_timeout(&ring->sched);
 		if (amdgpu_sriov_vf(adev))
@@ -83,15 +174,19 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
 	}
 
 exit:
-	adev->job_hang = false;
+	amdgpu_vm_put_task_info(ti);
 	drm_dev_exit(idx);
-	return DRM_GPU_SCHED_STAT_NOMINAL;
+	return DRM_GPU_SCHED_STAT_RESET;
 }
 
 int amdgpu_job_alloc(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		     struct drm_sched_entity *entity, void *owner,
-		     unsigned int num_ibs, struct amdgpu_job **job)
+		     unsigned int num_ibs, struct amdgpu_job **job,
+		     u64 drm_client_id)
 {
+	struct amdgpu_fence *af;
+	int r;
+
 	if (num_ibs == 0)
 		return -EINVAL;
 
@@ -99,31 +194,50 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	if (!*job)
 		return -ENOMEM;
 
-	/*
-	 * Initialize the scheduler to at least some ring so that we always
-	 * have a pointer to adev.
-	 */
-	(*job)->base.sched = &adev->rings[0]->sched;
+	af = kzalloc(sizeof(struct amdgpu_fence), GFP_KERNEL);
+	if (!af) {
+		r = -ENOMEM;
+		goto err_job;
+	}
+	(*job)->hw_fence = af;
+
+	af = kzalloc(sizeof(struct amdgpu_fence), GFP_KERNEL);
+	if (!af) {
+		r = -ENOMEM;
+		goto err_fence;
+	}
+	(*job)->hw_vm_fence = af;
+
 	(*job)->vm = vm;
 
 	amdgpu_sync_create(&(*job)->explicit_sync);
-	(*job)->vram_lost_counter = atomic_read(&adev->vram_lost_counter);
+	(*job)->generation = amdgpu_vm_generation(adev, vm);
 	(*job)->vm_pd_addr = AMDGPU_BO_INVALID_OFFSET;
 
 	if (!entity)
 		return 0;
 
-	return drm_sched_job_init(&(*job)->base, entity, owner);
+	return drm_sched_job_init(&(*job)->base, entity, 1, owner,
+				  drm_client_id);
+
+err_fence:
+	kfree((*job)->hw_fence);
+err_job:
+	kfree(*job);
+	*job = NULL;
+
+	return r;
 }
 
 int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev,
 			     struct drm_sched_entity *entity, void *owner,
 			     size_t size, enum amdgpu_ib_pool_type pool_type,
-			     struct amdgpu_job **job)
+			     struct amdgpu_job **job, u64 k_job_id)
 {
 	int r;
 
-	r = amdgpu_job_alloc(adev, NULL, entity, owner, 1, job);
+	r = amdgpu_job_alloc(adev, NULL, entity, owner, 1, job,
+			     k_job_id);
 	if (r)
 		return r;
 
@@ -132,7 +246,10 @@ int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev,
 	if (r) {
 		if (entity)
 			drm_sched_job_cleanup(&(*job)->base);
+		kfree((*job)->hw_vm_fence);
+		kfree((*job)->hw_fence);
 		kfree(*job);
+		*job = NULL;
 	}
 
 	return r;
@@ -157,20 +274,19 @@ void amdgpu_job_set_resources(struct amdgpu_job *job, struct amdgpu_bo *gds,
 
 void amdgpu_job_free_resources(struct amdgpu_job *job)
 {
-	struct amdgpu_ring *ring = to_amdgpu_ring(job->base.sched);
 	struct dma_fence *f;
 	unsigned i;
 
-	/* Check if any fences where initialized */
+	/* Check if any fences were initialized */
 	if (job->base.s_fence && job->base.s_fence->finished.ops)
 		f = &job->base.s_fence->finished;
-	else if (job->hw_fence.ops)
-		f = &job->hw_fence;
+	else if (job->hw_fence && job->hw_fence->base.ops)
+		f = &job->hw_fence->base;
 	else
 		f = NULL;
 
 	for (i = 0; i < job->num_ibs; ++i)
-		amdgpu_ib_free(ring->adev, &job->ibs[i], f);
+		amdgpu_ib_free(&job->ibs[i], f);
 }
 
 static void amdgpu_job_free_cb(struct drm_sched_job *s_job)
@@ -181,11 +297,16 @@ static void amdgpu_job_free_cb(struct drm_sched_job *s_job)
 
 	amdgpu_sync_free(&job->explicit_sync);
 
-	/* only put the hw fence if has embedded fence */
-	if (!job->hw_fence.ops)
-		kfree(job);
+	if (job->hw_fence->base.ops)
+		dma_fence_put(&job->hw_fence->base);
 	else
-		dma_fence_put(&job->hw_fence);
+		kfree(job->hw_fence);
+	if (job->hw_vm_fence->base.ops)
+		dma_fence_put(&job->hw_vm_fence->base);
+	else
+		kfree(job->hw_vm_fence);
+
+	kfree(job);
 }
 
 void amdgpu_job_set_gang_leader(struct amdgpu_job *job,
@@ -214,10 +335,16 @@ void amdgpu_job_free(struct amdgpu_job *job)
 	if (job->gang_submit != &job->base.s_fence->scheduled)
 		dma_fence_put(job->gang_submit);
 
-	if (!job->hw_fence.ops)
-		kfree(job);
+	if (job->hw_fence->base.ops)
+		dma_fence_put(&job->hw_fence->base);
 	else
-		dma_fence_put(&job->hw_fence);
+		kfree(job->hw_fence);
+	if (job->hw_vm_fence->base.ops)
+		dma_fence_put(&job->hw_vm_fence->base);
+	else
+		kfree(job->hw_vm_fence);
+
+	kfree(job);
 }
 
 struct dma_fence *amdgpu_job_submit(struct amdgpu_job *job)
@@ -253,19 +380,37 @@ amdgpu_job_prepare_job(struct drm_sched_job *sched_job,
 {
 	struct amdgpu_ring *ring = to_amdgpu_ring(s_entity->rq->sched);
 	struct amdgpu_job *job = to_amdgpu_job(sched_job);
-	struct dma_fence *fence = NULL;
+	struct dma_fence *fence;
 	int r;
 
-	if (!fence && job->gang_submit)
+	r = drm_sched_entity_error(s_entity);
+	if (r)
+		goto error;
+
+	if (job->gang_submit) {
 		fence = amdgpu_device_switch_gang(ring->adev, job->gang_submit);
+		if (fence)
+			return fence;
+	}
 
-	while (!fence && job->vm && !job->vmid) {
+	fence = amdgpu_device_enforce_isolation(ring->adev, ring, job);
+	if (fence)
+		return fence;
+
+	if (job->vm && !job->vmid) {
 		r = amdgpu_vmid_grab(job->vm, ring, job, &fence);
-		if (r)
-			DRM_ERROR("Error getting VM ID (%d)\n", r);
+		if (r) {
+			dev_err(ring->adev->dev, "Error getting VM ID (%d)\n", r);
+			goto error;
+		}
+		return fence;
 	}
 
-	return fence;
+	return NULL;
+
+error:
+	dma_fence_set_error(&job->base.s_fence->finished, r);
+	return NULL;
 }
 
 static struct dma_fence *amdgpu_job_run(struct drm_sched_job *sched_job)
@@ -282,17 +427,20 @@ static struct dma_fence *amdgpu_job_run(struct drm_sched_job *sched_job)
 	trace_amdgpu_sched_run_job(job);
 
 	/* Skip job if VRAM is lost and never resubmit gangs */
-	if (job->vram_lost_counter != atomic_read(&adev->vram_lost_counter) ||
+	if (job->generation != amdgpu_vm_generation(adev, job->vm) ||
 	    (job->job_run_counter && job->gang_submit))
 		dma_fence_set_error(finished, -ECANCELED);
 
 	if (finished->error < 0) {
-		DRM_INFO("Skip scheduling IBs!\n");
+		dev_dbg(adev->dev, "Skip scheduling IBs in ring(%s)",
+			ring->name);
 	} else {
 		r = amdgpu_ib_schedule(ring, job->num_ibs, job->ibs, job,
 				       &fence);
 		if (r)
-			DRM_ERROR("Error scheduling IBs (%d)\n", r);
+			dev_err(adev->dev,
+				"Error scheduling IBs (%d) in ring(%s)", r,
+				ring->name);
 	}
 
 	job->job_run_counter++;
@@ -302,8 +450,24 @@ static struct dma_fence *amdgpu_job_run(struct drm_sched_job *sched_job)
 	return fence;
 }
 
-#define to_drm_sched_job(sched_job)		\
-		container_of((sched_job), struct drm_sched_job, queue_node)
+/*
+ * This is a duplicate function from DRM scheduler sched_internal.h.
+ * Plan is to remove it when amdgpu_job_stop_all_jobs_on_sched is removed, due
+ * latter being incorrect and racy.
+ *
+ * See https://lore.kernel.org/amd-gfx/44edde63-7181-44fb-a4f7-94e50514f539@amd.com/
+ */
+static struct drm_sched_job *
+drm_sched_entity_queue_pop(struct drm_sched_entity *entity)
+{
+	struct spsc_node *node;
+
+	node = spsc_queue_pop(&entity->job_queue);
+	if (!node)
+		return NULL;
+
+	return container_of(node, struct drm_sched_job, queue_node);
+}
 
 void amdgpu_job_stop_all_jobs_on_sched(struct drm_gpu_scheduler *sched)
 {
@@ -312,11 +476,11 @@ void amdgpu_job_stop_all_jobs_on_sched(struct drm_gpu_scheduler *sched)
 	int i;
 
 	/* Signal all jobs not yet scheduled */
-	for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
-		struct drm_sched_rq *rq = &sched->sched_rq[i];
+	for (i = DRM_SCHED_PRIORITY_KERNEL; i < sched->num_rqs; i++) {
+		struct drm_sched_rq *rq = sched->sched_rq[i];
 		spin_lock(&rq->lock);
 		list_for_each_entry(s_entity, &rq->entities, list) {
-			while ((s_job = to_drm_sched_job(spsc_queue_pop(&s_entity->job_queue)))) {
+			while ((s_job = drm_sched_entity_queue_pop(s_entity))) {
 				struct drm_sched_fence *s_fence = s_job->s_fence;
 
 				dma_fence_signal(&s_fence->scheduled);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h
index 52f2e313ea17..7abf069d17d4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h
@@ -44,11 +44,28 @@
 struct amdgpu_fence;
 enum amdgpu_ib_pool_type;
 
+/* Internal kernel job ids. (decreasing values, starting from U64_MAX). */
+#define AMDGPU_KERNEL_JOB_ID_VM_UPDATE              (18446744073709551615ULL)
+#define AMDGPU_KERNEL_JOB_ID_VM_UPDATE_PDES         (18446744073709551614ULL)
+#define AMDGPU_KERNEL_JOB_ID_VM_UPDATE_RANGE        (18446744073709551613ULL)
+#define AMDGPU_KERNEL_JOB_ID_VM_PT_CLEAR            (18446744073709551612ULL)
+#define AMDGPU_KERNEL_JOB_ID_TTM_MAP_BUFFER         (18446744073709551611ULL)
+#define AMDGPU_KERNEL_JOB_ID_TTM_ACCESS_MEMORY_SDMA (18446744073709551610ULL)
+#define AMDGPU_KERNEL_JOB_ID_TTM_COPY_BUFFER        (18446744073709551609ULL)
+#define AMDGPU_KERNEL_JOB_ID_CLEAR_ON_RELEASE       (18446744073709551608ULL)
+#define AMDGPU_KERNEL_JOB_ID_MOVE_BLIT              (18446744073709551607ULL)
+#define AMDGPU_KERNEL_JOB_ID_TTM_CLEAR_BUFFER       (18446744073709551606ULL)
+#define AMDGPU_KERNEL_JOB_ID_CLEANER_SHADER         (18446744073709551605ULL)
+#define AMDGPU_KERNEL_JOB_ID_FLUSH_GPU_TLB          (18446744073709551604ULL)
+#define AMDGPU_KERNEL_JOB_ID_KFD_GART_MAP           (18446744073709551603ULL)
+#define AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST          (18446744073709551602ULL)
+
 struct amdgpu_job {
 	struct drm_sched_job    base;
 	struct amdgpu_vm	*vm;
 	struct amdgpu_sync	explicit_sync;
-	struct dma_fence	hw_fence;
+	struct amdgpu_fence	*hw_fence;
+	struct amdgpu_fence	*hw_vm_fence;
 	struct dma_fence	*gang_submit;
 	uint32_t		preamble_status;
 	uint32_t                preemption_status;
@@ -61,15 +78,25 @@ struct amdgpu_job {
 	uint32_t		gds_base, gds_size;
 	uint32_t		gws_base, gws_size;
 	uint32_t		oa_base, oa_size;
-	uint32_t		vram_lost_counter;
+	uint64_t		generation;
 
 	/* user fence handling */
 	uint64_t		uf_addr;
 	uint64_t		uf_sequence;
 
+	/* virtual addresses for shadow/GDS/CSA */
+	uint64_t		shadow_va;
+	uint64_t		csa_va;
+	uint64_t		gds_va;
+	bool			init_shadow;
+
 	/* job_run_counter >= 1 means a resubmit job */
 	uint32_t		job_run_counter;
 
+	/* enforce isolation */
+	bool			enforce_isolation;
+	bool			run_cleaner_shader;
+
 	uint32_t		num_ibs;
 	struct amdgpu_ib	ibs[];
 };
@@ -81,11 +108,13 @@ static inline struct amdgpu_ring *amdgpu_job_ring(struct amdgpu_job *job)
 
 int amdgpu_job_alloc(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		     struct drm_sched_entity *entity, void *owner,
-		     unsigned int num_ibs, struct amdgpu_job **job);
+		     unsigned int num_ibs, struct amdgpu_job **job,
+		     u64 drm_client_id);
 int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev,
 			     struct drm_sched_entity *entity, void *owner,
 			     size_t size, enum amdgpu_ib_pool_type pool_type,
-			     struct amdgpu_job **job);
+			     struct amdgpu_job **job,
+			     u64 k_job_id);
 void amdgpu_job_set_resources(struct amdgpu_job *job, struct amdgpu_bo *gds,
 			      struct amdgpu_bo *gws, struct amdgpu_bo *oa);
 void amdgpu_job_free_resources(struct amdgpu_job *job);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.c
index b07c000fc8ba..63ee6ba6a931 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.c
@@ -33,27 +33,62 @@
 #define JPEG_IDLE_TIMEOUT	msecs_to_jiffies(1000)
 
 static void amdgpu_jpeg_idle_work_handler(struct work_struct *work);
+static void amdgpu_jpeg_reg_dump_fini(struct amdgpu_device *adev);
 
 int amdgpu_jpeg_sw_init(struct amdgpu_device *adev)
 {
+	int i, r;
+
 	INIT_DELAYED_WORK(&adev->jpeg.idle_work, amdgpu_jpeg_idle_work_handler);
 	mutex_init(&adev->jpeg.jpeg_pg_lock);
 	atomic_set(&adev->jpeg.total_submission_cnt, 0);
 
+	if ((adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) &&
+	    (adev->pg_flags & AMD_PG_SUPPORT_JPEG_DPG))
+		adev->jpeg.indirect_sram = true;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; i++) {
+		if (adev->jpeg.harvest_config & (1U << i))
+			continue;
+
+		if (adev->jpeg.indirect_sram) {
+			r = amdgpu_bo_create_kernel(adev, 64 * 2 * 4, PAGE_SIZE,
+					AMDGPU_GEM_DOMAIN_VRAM |
+					AMDGPU_GEM_DOMAIN_GTT,
+					&adev->jpeg.inst[i].dpg_sram_bo,
+					&adev->jpeg.inst[i].dpg_sram_gpu_addr,
+					&adev->jpeg.inst[i].dpg_sram_cpu_addr);
+			if (r) {
+				dev_err(adev->dev,
+				"JPEG %d (%d) failed to allocate DPG bo\n", i, r);
+				return r;
+			}
+		}
+	}
+
 	return 0;
 }
 
 int amdgpu_jpeg_sw_fini(struct amdgpu_device *adev)
 {
-	int i;
+	int i, j;
 
 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
-		if (adev->jpeg.harvest_config & (1 << i))
+		if (adev->jpeg.harvest_config & (1U << i))
 			continue;
 
-		amdgpu_ring_fini(&adev->jpeg.inst[i].ring_dec);
+		amdgpu_bo_free_kernel(
+			&adev->jpeg.inst[i].dpg_sram_bo,
+			&adev->jpeg.inst[i].dpg_sram_gpu_addr,
+			(void **)&adev->jpeg.inst[i].dpg_sram_cpu_addr);
+
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j)
+			amdgpu_ring_fini(&adev->jpeg.inst[i].ring_dec[j]);
 	}
 
+	if (adev->jpeg.reg_list)
+		amdgpu_jpeg_reg_dump_fini(adev);
+
 	mutex_destroy(&adev->jpeg.jpeg_pg_lock);
 
 	return 0;
@@ -76,19 +111,22 @@ static void amdgpu_jpeg_idle_work_handler(struct work_struct *work)
 	struct amdgpu_device *adev =
 		container_of(work, struct amdgpu_device, jpeg.idle_work.work);
 	unsigned int fences = 0;
-	unsigned int i;
+	unsigned int i, j;
 
 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
-		if (adev->jpeg.harvest_config & (1 << i))
+		if (adev->jpeg.harvest_config & (1U << i))
 			continue;
 
-		fences += amdgpu_fence_count_emitted(&adev->jpeg.inst[i].ring_dec);
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j)
+			fences += amdgpu_fence_count_emitted(&adev->jpeg.inst[i].ring_dec[j]);
 	}
 
-	if (!fences && !atomic_read(&adev->jpeg.total_submission_cnt))
+	if (!fences && !atomic_read(&adev->jpeg.total_submission_cnt)) {
+		mutex_lock(&adev->jpeg.jpeg_pg_lock);
 		amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_JPEG,
 						       AMD_PG_STATE_GATE);
-	else
+		mutex_unlock(&adev->jpeg.jpeg_pg_lock);
+	} else
 		schedule_delayed_work(&adev->jpeg.idle_work, JPEG_IDLE_TIMEOUT);
 }
 
@@ -122,18 +160,21 @@ int amdgpu_jpeg_dec_ring_test_ring(struct amdgpu_ring *ring)
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	WREG32(adev->jpeg.inst[ring->me].external.jpeg_pitch, 0xCAFEDEAD);
 	r = amdgpu_ring_alloc(ring, 3);
 	if (r)
 		return r;
 
-	amdgpu_ring_write(ring, PACKET0(adev->jpeg.internal.jpeg_pitch, 0));
-	amdgpu_ring_write(ring, 0xDEADBEEF);
+	WREG32(adev->jpeg.inst[ring->me].external.jpeg_pitch[ring->pipe], 0xCAFEDEAD);
+	/* Add a read register to make sure the write register is executed. */
+	RREG32(adev->jpeg.inst[ring->me].external.jpeg_pitch[ring->pipe]);
+
+	amdgpu_ring_write(ring, PACKET0(adev->jpeg.internal.jpeg_pitch[ring->pipe], 0));
+	amdgpu_ring_write(ring, 0xABADCAFE);
 	amdgpu_ring_commit(ring);
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(adev->jpeg.inst[ring->me].external.jpeg_pitch);
-		if (tmp == 0xDEADBEEF)
+		tmp = RREG32(adev->jpeg.inst[ring->me].external.jpeg_pitch[ring->pipe]);
+		if (tmp == 0xABADCAFE)
 			break;
 		udelay(1);
 	}
@@ -155,14 +196,14 @@ static int amdgpu_jpeg_dec_set_reg(struct amdgpu_ring *ring, uint32_t handle,
 	int i, r;
 
 	r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, ib_size_dw * 4,
-				     AMDGPU_IB_POOL_DIRECT, &job);
+				     AMDGPU_IB_POOL_DIRECT, &job,
+				     AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		return r;
 
 	ib = &job->ibs[0];
 
-	ib->ptr[0] = PACKETJ(adev->jpeg.internal.jpeg_pitch, 0, 0,
-			     PACKETJ_TYPE0);
+	ib->ptr[0] = PACKETJ(adev->jpeg.internal.jpeg_pitch[ring->pipe], 0, 0, PACKETJ_TYPE0);
 	ib->ptr[1] = 0xDEADBEEF;
 	for (i = 2; i < 16; i += 2) {
 		ib->ptr[i] = PACKETJ(0, 0, 0, PACKETJ_TYPE6);
@@ -206,12 +247,15 @@ int amdgpu_jpeg_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	} else {
 		r = 0;
 	}
+
 	if (!amdgpu_sriov_vf(adev)) {
 		for (i = 0; i < adev->usec_timeout; i++) {
-			tmp = RREG32(adev->jpeg.inst[ring->me].external.jpeg_pitch);
+			tmp = RREG32(adev->jpeg.inst[ring->me].external.jpeg_pitch[ring->pipe]);
 			if (tmp == 0xDEADBEEF)
 				break;
 			udelay(1);
+			if (amdgpu_emu_mode == 1)
+				udelay(10);
 		}
 
 		if (i >= adev->usec_timeout)
@@ -241,6 +285,32 @@ int amdgpu_jpeg_process_poison_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
+int amdgpu_jpeg_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
+{
+	int r, i;
+
+	r = amdgpu_ras_block_late_init(adev, ras_block);
+	if (r)
+		return r;
+
+	if (amdgpu_ras_is_supported(adev, ras_block->block)) {
+		for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+			if (adev->jpeg.harvest_config & (1 << i) ||
+			    !adev->jpeg.inst[i].ras_poison_irq.funcs)
+				continue;
+
+			r = amdgpu_irq_get(adev, &adev->jpeg.inst[i].ras_poison_irq, 0);
+			if (r)
+				goto late_fini;
+		}
+	}
+	return 0;
+
+late_fini:
+	amdgpu_ras_block_late_fini(adev, ras_block);
+	return r;
+}
+
 int amdgpu_jpeg_ras_sw_init(struct amdgpu_device *adev)
 {
 	int err;
@@ -262,7 +332,276 @@ int amdgpu_jpeg_ras_sw_init(struct amdgpu_device *adev)
 	adev->jpeg.ras_if = &ras->ras_block.ras_comm;
 
 	if (!ras->ras_block.ras_late_init)
-		ras->ras_block.ras_late_init = amdgpu_ras_block_late_init;
+		ras->ras_block.ras_late_init = amdgpu_jpeg_ras_late_init;
+
+	return 0;
+}
+
+int amdgpu_jpeg_psp_update_sram(struct amdgpu_device *adev, int inst_idx,
+			       enum AMDGPU_UCODE_ID ucode_id)
+{
+	struct amdgpu_firmware_info ucode = {
+		.ucode_id = AMDGPU_UCODE_ID_JPEG_RAM,
+		.mc_addr = adev->jpeg.inst[inst_idx].dpg_sram_gpu_addr,
+		.ucode_size = ((uintptr_t)adev->jpeg.inst[inst_idx].dpg_sram_curr_addr -
+			      (uintptr_t)adev->jpeg.inst[inst_idx].dpg_sram_cpu_addr),
+	};
+
+	return psp_execute_ip_fw_load(&adev->psp, &ucode);
+}
+
+/*
+ * debugfs for to enable/disable jpeg job submission to specific core.
+ */
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_debugfs_jpeg_sched_mask_set(void *data, u64 val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i, j;
+	u64 mask = 0;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+
+	mask = (1ULL << (adev->jpeg.num_jpeg_inst * adev->jpeg.num_jpeg_rings)) - 1;
+	if ((val & mask) == 0)
+		return -EINVAL;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+			ring = &adev->jpeg.inst[i].ring_dec[j];
+			if (val & (BIT_ULL((i * adev->jpeg.num_jpeg_rings) + j)))
+				ring->sched.ready = true;
+			else
+				ring->sched.ready = false;
+		}
+	}
+	/* publish sched.ready flag update effective immediately across smp */
+	smp_rmb();
+	return 0;
+}
+
+static int amdgpu_debugfs_jpeg_sched_mask_get(void *data, u64 *val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i, j;
+	u64 mask = 0;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+			ring = &adev->jpeg.inst[i].ring_dec[j];
+			if (ring->sched.ready)
+				mask |= 1ULL << ((i * adev->jpeg.num_jpeg_rings) + j);
+		}
+	}
+	*val = mask;
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(amdgpu_debugfs_jpeg_sched_mask_fops,
+			 amdgpu_debugfs_jpeg_sched_mask_get,
+			 amdgpu_debugfs_jpeg_sched_mask_set, "%llx\n");
+
+#endif
+
+void amdgpu_debugfs_jpeg_sched_mask_init(struct amdgpu_device *adev)
+{
+#if defined(CONFIG_DEBUG_FS)
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+	struct dentry *root = minor->debugfs_root;
+	char name[32];
+
+	if (!(adev->jpeg.num_jpeg_inst > 1) && !(adev->jpeg.num_jpeg_rings > 1))
+		return;
+	sprintf(name, "amdgpu_jpeg_sched_mask");
+	debugfs_create_file(name, 0600, root, adev,
+			    &amdgpu_debugfs_jpeg_sched_mask_fops);
+#endif
+}
+
+static ssize_t amdgpu_get_jpeg_reset_mask(struct device *dev,
+						struct device_attribute *attr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	if (!adev)
+		return -ENODEV;
+
+	return amdgpu_show_reset_mask(buf, adev->jpeg.supported_reset);
+}
+
+static DEVICE_ATTR(jpeg_reset_mask, 0444,
+		   amdgpu_get_jpeg_reset_mask, NULL);
+
+int amdgpu_jpeg_sysfs_reset_mask_init(struct amdgpu_device *adev)
+{
+	int r = 0;
+
+	if (adev->jpeg.num_jpeg_inst) {
+		r = device_create_file(adev->dev, &dev_attr_jpeg_reset_mask);
+		if (r)
+			return r;
+	}
+
+	return r;
+}
+
+void amdgpu_jpeg_sysfs_reset_mask_fini(struct amdgpu_device *adev)
+{
+	if (adev->dev->kobj.sd) {
+		if (adev->jpeg.num_jpeg_inst)
+			device_remove_file(adev->dev, &dev_attr_jpeg_reset_mask);
+	}
+}
+
+int amdgpu_jpeg_reg_dump_init(struct amdgpu_device *adev,
+			       const struct amdgpu_hwip_reg_entry *reg, u32 count)
+{
+	adev->jpeg.ip_dump = kcalloc(adev->jpeg.num_jpeg_inst * count,
+				     sizeof(uint32_t), GFP_KERNEL);
+	if (!adev->jpeg.ip_dump) {
+		dev_err(adev->dev,
+			"Failed to allocate memory for JPEG IP Dump\n");
+		return -ENOMEM;
+	}
+	adev->jpeg.reg_list = reg;
+	adev->jpeg.reg_count = count;
+
+	return 0;
+}
+
+static void amdgpu_jpeg_reg_dump_fini(struct amdgpu_device *adev)
+{
+	kfree(adev->jpeg.ip_dump);
+	adev->jpeg.reg_list = NULL;
+	adev->jpeg.reg_count = 0;
+}
+
+void amdgpu_jpeg_dump_ip_state(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	u32 inst_off, inst_id, is_powered;
+	int i, j;
+
+	if (!adev->jpeg.ip_dump)
+		return;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; i++) {
+		if (adev->jpeg.harvest_config & (1 << i))
+			continue;
+
+		inst_id = GET_INST(JPEG, i);
+		inst_off = i * adev->jpeg.reg_count;
+		/* check power status from UVD_JPEG_POWER_STATUS */
+		adev->jpeg.ip_dump[inst_off] =
+			RREG32(SOC15_REG_ENTRY_OFFSET_INST(adev->jpeg.reg_list[0],
+							   inst_id));
+		is_powered = ((adev->jpeg.ip_dump[inst_off] & 0x1) != 1);
+
+		if (is_powered)
+			for (j = 1; j < adev->jpeg.reg_count; j++)
+				adev->jpeg.ip_dump[inst_off + j] =
+					RREG32(SOC15_REG_ENTRY_OFFSET_INST(adev->jpeg.reg_list[j],
+									   inst_id));
+	}
+}
+
+void amdgpu_jpeg_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	u32 inst_off, is_powered;
+	int i, j;
+
+	if (!adev->jpeg.ip_dump)
+		return;
+
+	drm_printf(p, "num_instances:%d\n", adev->jpeg.num_jpeg_inst);
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; i++) {
+		if (adev->jpeg.harvest_config & (1 << i)) {
+			drm_printf(p, "\nHarvested Instance:JPEG%d Skipping dump\n", i);
+			continue;
+		}
+
+		inst_off = i * adev->jpeg.reg_count;
+		is_powered = ((adev->jpeg.ip_dump[inst_off] & 0x1) != 1);
+
+		if (is_powered) {
+			drm_printf(p, "Active Instance:JPEG%d\n", i);
+			for (j = 0; j < adev->jpeg.reg_count; j++)
+				drm_printf(p, "%-50s \t 0x%08x\n", adev->jpeg.reg_list[j].reg_name,
+					   adev->jpeg.ip_dump[inst_off + j]);
+		} else
+			drm_printf(p, "\nInactive Instance:JPEG%d\n", i);
+	}
+}
+
+static inline bool amdgpu_jpeg_reg_valid(u32 reg)
+{
+	if (reg < JPEG_REG_RANGE_START || reg > JPEG_REG_RANGE_END ||
+	    (reg >= JPEG_ATOMIC_RANGE_START && reg <= JPEG_ATOMIC_RANGE_END))
+		return false;
+	else
+		return true;
+}
+
+/**
+ * amdgpu_jpeg_dec_parse_cs - command submission parser
+ *
+ * @parser: Command submission parser context
+ * @job: the job to parse
+ * @ib: the IB to parse
+ *
+ * Parse the command stream, return -EINVAL for invalid packet,
+ * 0 otherwise
+ */
+
+int amdgpu_jpeg_dec_parse_cs(struct amdgpu_cs_parser *parser,
+			      struct amdgpu_job *job,
+			      struct amdgpu_ib *ib)
+{
+	u32 i, reg, res, cond, type;
+	struct amdgpu_device *adev = parser->adev;
+
+	for (i = 0; i < ib->length_dw ; i += 2) {
+		reg  = CP_PACKETJ_GET_REG(ib->ptr[i]);
+		res  = CP_PACKETJ_GET_RES(ib->ptr[i]);
+		cond = CP_PACKETJ_GET_COND(ib->ptr[i]);
+		type = CP_PACKETJ_GET_TYPE(ib->ptr[i]);
+
+		if (res) /* only support 0 at the moment */
+			return -EINVAL;
+
+		switch (type) {
+		case PACKETJ_TYPE0:
+			if (cond != PACKETJ_CONDITION_CHECK0 ||
+			    !amdgpu_jpeg_reg_valid(reg)) {
+				dev_err(adev->dev, "Invalid packet [0x%08x]!\n", ib->ptr[i]);
+				return -EINVAL;
+			}
+			break;
+		case PACKETJ_TYPE3:
+			if (cond != PACKETJ_CONDITION_CHECK3 ||
+			    !amdgpu_jpeg_reg_valid(reg)) {
+				dev_err(adev->dev, "Invalid packet [0x%08x]!\n", ib->ptr[i]);
+				return -EINVAL;
+			}
+			break;
+		case PACKETJ_TYPE6:
+			if (ib->ptr[i] == CP_PACKETJ_NOP)
+				continue;
+			dev_err(adev->dev, "Invalid packet [0x%08x]!\n", ib->ptr[i]);
+			return -EINVAL;
+		default:
+			dev_err(adev->dev, "Unknown packet type %d !\n", type);
+			return -EINVAL;
+		}
+	}
 
 	return 0;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h
index 0ca76f0f23e9..346ae0ab09d3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h
@@ -25,20 +25,103 @@
 #define __AMDGPU_JPEG_H__
 
 #include "amdgpu_ras.h"
+#include "amdgpu_cs.h"
+
+#define AMDGPU_MAX_JPEG_INSTANCES	4
+#define AMDGPU_MAX_JPEG_RINGS           10
+#define AMDGPU_MAX_JPEG_RINGS_4_0_3     8
+
+#define JPEG_REG_RANGE_START            0x4000
+#define JPEG_REG_RANGE_END              0x41c2
+#define JPEG_ATOMIC_RANGE_START         0x4120
+#define JPEG_ATOMIC_RANGE_END           0x412A
 
-#define AMDGPU_MAX_JPEG_INSTANCES	2
 
 #define AMDGPU_JPEG_HARVEST_JPEG0 (1 << 0)
 #define AMDGPU_JPEG_HARVEST_JPEG1 (1 << 1)
 
+#define WREG32_SOC15_JPEG_DPG_MODE(inst_idx, offset, value, indirect)			\
+	do {										\
+		if (!indirect) {							\
+			WREG32_SOC15(JPEG, GET_INST(JPEG, inst_idx),			\
+				     mmUVD_DPG_LMA_DATA, value);			\
+			WREG32_SOC15(							\
+				JPEG, GET_INST(JPEG, inst_idx),				\
+				mmUVD_DPG_LMA_CTL,					\
+				(UVD_DPG_LMA_CTL__READ_WRITE_MASK |			\
+				 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT |	\
+				 indirect << UVD_DPG_LMA_CTL__SRAM_SEL__SHIFT));	\
+		} else {								\
+			*adev->jpeg.inst[inst_idx].dpg_sram_curr_addr++ =		\
+				offset;							\
+			*adev->jpeg.inst[inst_idx].dpg_sram_curr_addr++ =		\
+				value;							\
+		}									\
+	} while (0)
+
+#define RREG32_SOC15_JPEG_DPG_MODE(inst_idx, offset, mask_en)					\
+	({											\
+		WREG32_SOC15(JPEG, inst_idx, mmUVD_DPG_LMA_CTL,					\
+			(0x0 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT |				\
+			mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT |				\
+			offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT));			\
+		RREG32_SOC15(JPEG, inst_idx, mmUVD_DPG_LMA_DATA);				\
+	})
+
+#define WREG32_SOC24_JPEG_DPG_MODE(inst_idx, offset, value, indirect)		\
+	do {									\
+		WREG32_SOC15(JPEG, GET_INST(JPEG, inst_idx),			\
+			     regUVD_DPG_LMA_DATA, value);			\
+		WREG32_SOC15(JPEG, GET_INST(JPEG, inst_idx),			\
+			     regUVD_DPG_LMA_MASK, 0xFFFFFFFF);			\
+		WREG32_SOC15(							\
+			JPEG, GET_INST(JPEG, inst_idx),				\
+			regUVD_DPG_LMA_CTL,					\
+			(UVD_DPG_LMA_CTL__READ_WRITE_MASK |			\
+			 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT |	\
+			 indirect << UVD_DPG_LMA_CTL__SRAM_SEL__SHIFT));	\
+	} while (0)
+
+#define RREG32_SOC24_JPEG_DPG_MODE(inst_idx, offset, mask_en)			\
+	do {									\
+		WREG32_SOC15(JPEG, GET_INST(JPEG, inst_idx),			\
+			regUVD_DPG_LMA_MASK, 0xFFFFFFFF);			\
+		WREG32_SOC15(JPEG, GET_INST(JPEG, inst_idx),			\
+			regUVD_DPG_LMA_CTL,					\
+			(UVD_DPG_LMA_CTL__MASK_EN_MASK |			\
+			offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT));	\
+		RREG32_SOC15(JPEG, inst_idx, regUVD_DPG_LMA_DATA);		\
+	} while (0)
+
+#define ADD_SOC24_JPEG_TO_DPG_SRAM(inst_idx, offset, value, indirect)		\
+	do {									\
+		*adev->jpeg.inst[inst_idx].dpg_sram_curr_addr++ = offset;	\
+		*adev->jpeg.inst[inst_idx].dpg_sram_curr_addr++ = value;	\
+	} while (0)
+
+struct amdgpu_hwip_reg_entry;
+
+enum amdgpu_jpeg_caps {
+	AMDGPU_JPEG_RRMT_ENABLED,
+};
+
+#define AMDGPU_JPEG_CAPS(caps) BIT(AMDGPU_JPEG_##caps)
+
 struct amdgpu_jpeg_reg{
-	unsigned jpeg_pitch;
+	unsigned jpeg_pitch[AMDGPU_MAX_JPEG_RINGS];
 };
 
 struct amdgpu_jpeg_inst {
-	struct amdgpu_ring ring_dec;
+	struct amdgpu_ring ring_dec[AMDGPU_MAX_JPEG_RINGS];
 	struct amdgpu_irq_src irq;
+	struct amdgpu_irq_src ras_poison_irq;
 	struct amdgpu_jpeg_reg external;
+	struct amdgpu_bo	*dpg_sram_bo;
+	struct dpg_pause_state	pause_state;
+	void			*dpg_sram_cpu_addr;
+	uint64_t		dpg_sram_gpu_addr;
+	uint32_t		*dpg_sram_curr_addr;
+	uint8_t aid_id;
 };
 
 struct amdgpu_jpeg_ras {
@@ -48,6 +131,7 @@ struct amdgpu_jpeg_ras {
 struct amdgpu_jpeg {
 	uint8_t	num_jpeg_inst;
 	struct amdgpu_jpeg_inst inst[AMDGPU_MAX_JPEG_INSTANCES];
+	unsigned num_jpeg_rings;
 	struct amdgpu_jpeg_reg internal;
 	unsigned harvest_config;
 	struct delayed_work idle_work;
@@ -56,6 +140,15 @@ struct amdgpu_jpeg {
 	atomic_t total_submission_cnt;
 	struct ras_common_if	*ras_if;
 	struct amdgpu_jpeg_ras	*ras;
+
+	uint16_t inst_mask;
+	uint8_t num_inst_per_aid;
+	bool	indirect_sram;
+	uint32_t supported_reset;
+	uint32_t caps;
+	u32 *ip_dump;
+	u32 reg_count;
+	const struct amdgpu_hwip_reg_entry *reg_list;
 };
 
 int amdgpu_jpeg_sw_init(struct amdgpu_device *adev);
@@ -72,6 +165,20 @@ int amdgpu_jpeg_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout);
 int amdgpu_jpeg_process_poison_irq(struct amdgpu_device *adev,
 				struct amdgpu_irq_src *source,
 				struct amdgpu_iv_entry *entry);
+int amdgpu_jpeg_ras_late_init(struct amdgpu_device *adev,
+				struct ras_common_if *ras_block);
 int amdgpu_jpeg_ras_sw_init(struct amdgpu_device *adev);
+int amdgpu_jpeg_psp_update_sram(struct amdgpu_device *adev, int inst_idx,
+			       enum AMDGPU_UCODE_ID ucode_id);
+void amdgpu_debugfs_jpeg_sched_mask_init(struct amdgpu_device *adev);
+int amdgpu_jpeg_sysfs_reset_mask_init(struct amdgpu_device *adev);
+void amdgpu_jpeg_sysfs_reset_mask_fini(struct amdgpu_device *adev);
+int amdgpu_jpeg_reg_dump_init(struct amdgpu_device *adev,
+			       const struct amdgpu_hwip_reg_entry *reg, u32 count);
+void amdgpu_jpeg_dump_ip_state(struct amdgpu_ip_block *ip_block);
+void amdgpu_jpeg_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p);
+int amdgpu_jpeg_dec_parse_cs(struct amdgpu_cs_parser *parser,
+			     struct amdgpu_job *job,
+			     struct amdgpu_ib *ib);
 
 #endif /*__AMDGPU_JPEG_H__*/
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
index 0efb38539d70..6ee77f431d56 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
@@ -43,7 +43,9 @@
 #include "amdgpu_gem.h"
 #include "amdgpu_display.h"
 #include "amdgpu_ras.h"
+#include "amdgpu_reset.h"
 #include "amd_pcie.h"
+#include "amdgpu_userq.h"
 
 void amdgpu_unregister_gpu_instance(struct amdgpu_device *adev)
 {
@@ -89,7 +91,7 @@ void amdgpu_driver_unload_kms(struct drm_device *dev)
 	if (adev->rmmio == NULL)
 		return;
 
-	if (amdgpu_acpi_smart_shift_update(dev, AMDGPU_SS_DRV_UNLOAD))
+	if (amdgpu_acpi_smart_shift_update(adev, AMDGPU_SS_DRV_UNLOAD))
 		DRM_WARN("smart shift update failed\n");
 
 	amdgpu_acpi_fini(adev);
@@ -149,38 +151,7 @@ int amdgpu_driver_load_kms(struct amdgpu_device *adev, unsigned long flags)
 		goto out;
 	}
 
-	adev->pm.rpm_mode = AMDGPU_RUNPM_NONE;
-	if (amdgpu_device_supports_px(dev) &&
-	    (amdgpu_runtime_pm != 0)) { /* enable PX as runtime mode */
-		adev->pm.rpm_mode = AMDGPU_RUNPM_PX;
-		dev_info(adev->dev, "Using ATPX for runtime pm\n");
-	} else if (amdgpu_device_supports_boco(dev) &&
-		   (amdgpu_runtime_pm != 0)) { /* enable boco as runtime mode */
-		adev->pm.rpm_mode = AMDGPU_RUNPM_BOCO;
-		dev_info(adev->dev, "Using BOCO for runtime pm\n");
-	} else if (amdgpu_device_supports_baco(dev) &&
-		   (amdgpu_runtime_pm != 0)) {
-		switch (adev->asic_type) {
-		case CHIP_VEGA20:
-		case CHIP_ARCTURUS:
-			/* enable BACO as runpm mode if runpm=1 */
-			if (amdgpu_runtime_pm > 0)
-				adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
-			break;
-		case CHIP_VEGA10:
-			/* enable BACO as runpm mode if noretry=0 */
-			if (!adev->gmc.noretry)
-				adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
-			break;
-		default:
-			/* enable BACO as runpm mode on CI+ */
-			adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
-			break;
-		}
-
-		if (adev->pm.rpm_mode == AMDGPU_RUNPM_BACO)
-			dev_info(adev->dev, "Using BACO for runtime pm\n");
-	}
+	amdgpu_device_detect_runtime_pm_mode(adev);
 
 	/* Call ACPI methods: require modeset init
 	 * but failure is not fatal
@@ -190,7 +161,7 @@ int amdgpu_driver_load_kms(struct amdgpu_device *adev, unsigned long flags)
 	if (acpi_status)
 		dev_dbg(dev->dev, "Error during ACPI methods call\n");
 
-	if (amdgpu_acpi_smart_shift_update(dev, AMDGPU_SS_DRV_LOAD))
+	if (amdgpu_acpi_smart_shift_update(adev, AMDGPU_SS_DRV_LOAD))
 		DRM_WARN("smart shift update failed\n");
 
 out:
@@ -200,6 +171,44 @@ out:
 	return r;
 }
 
+static enum amd_ip_block_type
+	amdgpu_ip_get_block_type(struct amdgpu_device *adev, uint32_t ip)
+{
+	enum amd_ip_block_type type;
+
+	switch (ip) {
+	case AMDGPU_HW_IP_GFX:
+		type = AMD_IP_BLOCK_TYPE_GFX;
+		break;
+	case AMDGPU_HW_IP_COMPUTE:
+		type = AMD_IP_BLOCK_TYPE_GFX;
+		break;
+	case AMDGPU_HW_IP_DMA:
+		type = AMD_IP_BLOCK_TYPE_SDMA;
+		break;
+	case AMDGPU_HW_IP_UVD:
+	case AMDGPU_HW_IP_UVD_ENC:
+		type = AMD_IP_BLOCK_TYPE_UVD;
+		break;
+	case AMDGPU_HW_IP_VCE:
+		type = AMD_IP_BLOCK_TYPE_VCE;
+		break;
+	case AMDGPU_HW_IP_VCN_DEC:
+	case AMDGPU_HW_IP_VCN_ENC:
+		type = AMD_IP_BLOCK_TYPE_VCN;
+		break;
+	case AMDGPU_HW_IP_VCN_JPEG:
+		type = (amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_JPEG)) ?
+				   AMD_IP_BLOCK_TYPE_JPEG : AMD_IP_BLOCK_TYPE_VCN;
+		break;
+	default:
+		type = AMD_IP_BLOCK_TYPE_NUM;
+		break;
+	}
+
+	return type;
+}
+
 static int amdgpu_firmware_info(struct drm_amdgpu_info_firmware *fw_info,
 				struct drm_amdgpu_query_fw *query_fw,
 				struct amdgpu_device *adev)
@@ -352,12 +361,36 @@ static int amdgpu_firmware_info(struct drm_amdgpu_info_firmware *fw_info,
 		fw_info->ver = adev->gfx.imu_fw_version;
 		fw_info->feature = 0;
 		break;
+	case AMDGPU_INFO_FW_VPE:
+		fw_info->ver = adev->vpe.fw_version;
+		fw_info->feature = adev->vpe.feature_version;
+		break;
 	default:
 		return -EINVAL;
 	}
 	return 0;
 }
 
+static int amdgpu_userq_metadata_info_gfx(struct amdgpu_device *adev,
+					  struct drm_amdgpu_info *info,
+					  struct drm_amdgpu_info_uq_metadata_gfx *meta)
+{
+	int ret = -EOPNOTSUPP;
+
+	if (adev->gfx.funcs->get_gfx_shadow_info) {
+		struct amdgpu_gfx_shadow_info shadow = {};
+
+		adev->gfx.funcs->get_gfx_shadow_info(adev, &shadow, true);
+		meta->shadow_size = shadow.shadow_size;
+		meta->shadow_alignment = shadow.shadow_alignment;
+		meta->csa_size = shadow.csa_size;
+		meta->csa_alignment = shadow.csa_alignment;
+		ret = 0;
+	}
+
+	return ret;
+}
+
 static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 			     struct drm_amdgpu_info *info,
 			     struct drm_amdgpu_info_hw_ip *result)
@@ -366,6 +399,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 	uint32_t ib_size_alignment = 0;
 	enum amd_ip_block_type type;
 	unsigned int num_rings = 0;
+	uint32_t num_slots = 0;
 	unsigned int i, j;
 
 	if (info->query_hw_ip.ip_instance >= AMDGPU_HW_IP_INSTANCE_MAX_COUNT)
@@ -375,24 +409,45 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 	case AMDGPU_HW_IP_GFX:
 		type = AMD_IP_BLOCK_TYPE_GFX;
 		for (i = 0; i < adev->gfx.num_gfx_rings; i++)
-			if (adev->gfx.gfx_ring[i].sched.ready)
+			if (adev->gfx.gfx_ring[i].sched.ready &&
+			    !adev->gfx.gfx_ring[i].no_user_submission)
 				++num_rings;
+
+		if (!adev->gfx.disable_uq) {
+			for (i = 0; i < AMDGPU_MES_MAX_GFX_PIPES; i++)
+				num_slots += hweight32(adev->mes.gfx_hqd_mask[i]);
+		}
+
 		ib_start_alignment = 32;
 		ib_size_alignment = 32;
 		break;
 	case AMDGPU_HW_IP_COMPUTE:
 		type = AMD_IP_BLOCK_TYPE_GFX;
 		for (i = 0; i < adev->gfx.num_compute_rings; i++)
-			if (adev->gfx.compute_ring[i].sched.ready)
+			if (adev->gfx.compute_ring[i].sched.ready &&
+			    !adev->gfx.compute_ring[i].no_user_submission)
 				++num_rings;
+
+		if (!adev->sdma.disable_uq) {
+			for (i = 0; i < AMDGPU_MES_MAX_COMPUTE_PIPES; i++)
+				num_slots += hweight32(adev->mes.compute_hqd_mask[i]);
+		}
+
 		ib_start_alignment = 32;
 		ib_size_alignment = 32;
 		break;
 	case AMDGPU_HW_IP_DMA:
 		type = AMD_IP_BLOCK_TYPE_SDMA;
 		for (i = 0; i < adev->sdma.num_instances; i++)
-			if (adev->sdma.instance[i].ring.sched.ready)
+			if (adev->sdma.instance[i].ring.sched.ready &&
+			    !adev->sdma.instance[i].ring.no_user_submission)
 				++num_rings;
+
+		if (!adev->gfx.disable_uq) {
+			for (i = 0; i < AMDGPU_MES_MAX_SDMA_PIPES; i++)
+				num_slots += hweight32(adev->mes.sdma_hqd_mask[i]);
+		}
+
 		ib_start_alignment = 256;
 		ib_size_alignment = 4;
 		break;
@@ -402,19 +457,21 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 			if (adev->uvd.harvest_config & (1 << i))
 				continue;
 
-			if (adev->uvd.inst[i].ring.sched.ready)
+			if (adev->uvd.inst[i].ring.sched.ready &&
+			    !adev->uvd.inst[i].ring.no_user_submission)
 				++num_rings;
 		}
-		ib_start_alignment = 64;
+		ib_start_alignment = 256;
 		ib_size_alignment = 64;
 		break;
 	case AMDGPU_HW_IP_VCE:
 		type = AMD_IP_BLOCK_TYPE_VCE;
 		for (i = 0; i < adev->vce.num_rings; i++)
-			if (adev->vce.ring[i].sched.ready)
+			if (adev->vce.ring[i].sched.ready &&
+			    !adev->vce.ring[i].no_user_submission)
 				++num_rings;
-		ib_start_alignment = 4;
-		ib_size_alignment = 1;
+		ib_start_alignment = 256;
+		ib_size_alignment = 4;
 		break;
 	case AMDGPU_HW_IP_UVD_ENC:
 		type = AMD_IP_BLOCK_TYPE_UVD;
@@ -423,11 +480,12 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 				continue;
 
 			for (j = 0; j < adev->uvd.num_enc_rings; j++)
-				if (adev->uvd.inst[i].ring_enc[j].sched.ready)
+				if (adev->uvd.inst[i].ring_enc[j].sched.ready &&
+				    !adev->uvd.inst[i].ring_enc[j].no_user_submission)
 					++num_rings;
 		}
-		ib_start_alignment = 64;
-		ib_size_alignment = 64;
+		ib_start_alignment = 256;
+		ib_size_alignment = 4;
 		break;
 	case AMDGPU_HW_IP_VCN_DEC:
 		type = AMD_IP_BLOCK_TYPE_VCN;
@@ -435,11 +493,12 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 			if (adev->vcn.harvest_config & (1 << i))
 				continue;
 
-			if (adev->vcn.inst[i].ring_dec.sched.ready)
+			if (adev->vcn.inst[i].ring_dec.sched.ready &&
+			    !adev->vcn.inst[i].ring_dec.no_user_submission)
 				++num_rings;
 		}
-		ib_start_alignment = 16;
-		ib_size_alignment = 16;
+		ib_start_alignment = 256;
+		ib_size_alignment = 64;
 		break;
 	case AMDGPU_HW_IP_VCN_ENC:
 		type = AMD_IP_BLOCK_TYPE_VCN;
@@ -447,12 +506,13 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 			if (adev->vcn.harvest_config & (1 << i))
 				continue;
 
-			for (j = 0; j < adev->vcn.num_enc_rings; j++)
-				if (adev->vcn.inst[i].ring_enc[j].sched.ready)
+			for (j = 0; j < adev->vcn.inst[i].num_enc_rings; j++)
+				if (adev->vcn.inst[i].ring_enc[j].sched.ready &&
+				    !adev->vcn.inst[i].ring_enc[j].no_user_submission)
 					++num_rings;
 		}
-		ib_start_alignment = 64;
-		ib_size_alignment = 1;
+		ib_start_alignment = 256;
+		ib_size_alignment = 4;
 		break;
 	case AMDGPU_HW_IP_VCN_JPEG:
 		type = (amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_JPEG)) ?
@@ -462,11 +522,21 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 			if (adev->jpeg.harvest_config & (1 << i))
 				continue;
 
-			if (adev->jpeg.inst[i].ring_dec.sched.ready)
-				++num_rings;
+			for (j = 0; j < adev->jpeg.num_jpeg_rings; j++)
+				if (adev->jpeg.inst[i].ring_dec[j].sched.ready &&
+				    !adev->jpeg.inst[i].ring_dec[j].no_user_submission)
+					++num_rings;
 		}
-		ib_start_alignment = 16;
-		ib_size_alignment = 16;
+		ib_start_alignment = 256;
+		ib_size_alignment = 64;
+		break;
+	case AMDGPU_HW_IP_VPE:
+		type = AMD_IP_BLOCK_TYPE_VPE;
+		if (adev->vpe.ring.sched.ready &&
+		    !adev->vpe.ring.no_user_submission)
+			++num_rings;
+		ib_start_alignment = 256;
+		ib_size_alignment = 4;
 		break;
 	default:
 		return -EINVAL;
@@ -489,18 +559,26 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 	if (adev->asic_type >= CHIP_VEGA10) {
 		switch (type) {
 		case AMD_IP_BLOCK_TYPE_GFX:
-			result->ip_discovery_version = adev->ip_versions[GC_HWIP][0];
+			result->ip_discovery_version =
+				IP_VERSION_MAJ_MIN_REV(amdgpu_ip_version(adev, GC_HWIP, 0));
 			break;
 		case AMD_IP_BLOCK_TYPE_SDMA:
-			result->ip_discovery_version = adev->ip_versions[SDMA0_HWIP][0];
+			result->ip_discovery_version =
+				IP_VERSION_MAJ_MIN_REV(amdgpu_ip_version(adev, SDMA0_HWIP, 0));
 			break;
 		case AMD_IP_BLOCK_TYPE_UVD:
 		case AMD_IP_BLOCK_TYPE_VCN:
 		case AMD_IP_BLOCK_TYPE_JPEG:
-			result->ip_discovery_version = adev->ip_versions[UVD_HWIP][0];
+			result->ip_discovery_version =
+				IP_VERSION_MAJ_MIN_REV(amdgpu_ip_version(adev, UVD_HWIP, 0));
 			break;
 		case AMD_IP_BLOCK_TYPE_VCE:
-			result->ip_discovery_version = adev->ip_versions[VCE_HWIP][0];
+			result->ip_discovery_version =
+				IP_VERSION_MAJ_MIN_REV(amdgpu_ip_version(adev, VCE_HWIP, 0));
+			break;
+		case AMD_IP_BLOCK_TYPE_VPE:
+			result->ip_discovery_version =
+				IP_VERSION_MAJ_MIN_REV(amdgpu_ip_version(adev, VPE_HWIP, 0));
 			break;
 		default:
 			result->ip_discovery_version = 0;
@@ -511,6 +589,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 	}
 	result->capabilities_flags = 0;
 	result->available_rings = (1 << num_rings) - 1;
+	result->userq_num_slots = num_slots;
 	result->ib_start_alignment = ib_start_alignment;
 	result->ib_size_alignment = ib_size_alignment;
 	return 0;
@@ -537,11 +616,16 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	struct drm_amdgpu_info *info = data;
 	struct amdgpu_mode_info *minfo = &adev->mode_info;
 	void __user *out = (void __user *)(uintptr_t)info->return_pointer;
+	struct amdgpu_fpriv *fpriv;
+	struct amdgpu_ip_block *ip_block;
+	enum amd_ip_block_type type;
+	struct amdgpu_xcp *xcp;
+	u32 count, inst_mask;
 	uint32_t size = info->return_size;
 	struct drm_crtc *crtc;
 	uint32_t ui32 = 0;
 	uint64_t ui64 = 0;
-	int i, found;
+	int i, found, ret;
 	int ui32_size = sizeof(ui32);
 
 	if (!info->return_size || !info->return_pointer)
@@ -556,6 +640,7 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 			crtc = (struct drm_crtc *)minfo->crtcs[i];
 			if (crtc && crtc->base.id == info->mode_crtc.id) {
 				struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+
 				ui32 = amdgpu_crtc->crtc_id;
 				found = 1;
 				break;
@@ -568,56 +653,82 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		return copy_to_user(out, &ui32, min(size, 4u)) ? -EFAULT : 0;
 	case AMDGPU_INFO_HW_IP_INFO: {
 		struct drm_amdgpu_info_hw_ip ip = {};
-		int ret;
 
 		ret = amdgpu_hw_ip_info(adev, info, &ip);
 		if (ret)
 			return ret;
 
-		ret = copy_to_user(out, &ip, min((size_t)size, sizeof(ip)));
+		ret = copy_to_user(out, &ip, min_t(size_t, size, sizeof(ip)));
 		return ret ? -EFAULT : 0;
 	}
 	case AMDGPU_INFO_HW_IP_COUNT: {
-		enum amd_ip_block_type type;
-		uint32_t count = 0;
+		fpriv = (struct amdgpu_fpriv *)filp->driver_priv;
+		type = amdgpu_ip_get_block_type(adev, info->query_hw_ip.type);
+		ip_block = amdgpu_device_ip_get_ip_block(adev, type);
 
-		switch (info->query_hw_ip.type) {
-		case AMDGPU_HW_IP_GFX:
-			type = AMD_IP_BLOCK_TYPE_GFX;
-			break;
-		case AMDGPU_HW_IP_COMPUTE:
-			type = AMD_IP_BLOCK_TYPE_GFX;
-			break;
-		case AMDGPU_HW_IP_DMA:
-			type = AMD_IP_BLOCK_TYPE_SDMA;
-			break;
-		case AMDGPU_HW_IP_UVD:
-			type = AMD_IP_BLOCK_TYPE_UVD;
+		if (!ip_block || !ip_block->status.valid)
+			return -EINVAL;
+
+		if (adev->xcp_mgr && adev->xcp_mgr->num_xcps > 0 &&
+		    fpriv->xcp_id < adev->xcp_mgr->num_xcps) {
+			xcp = &adev->xcp_mgr->xcp[fpriv->xcp_id];
+			switch (type) {
+			case AMD_IP_BLOCK_TYPE_GFX:
+				ret = amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_GFX, &inst_mask);
+				if (ret)
+					return ret;
+				count = hweight32(inst_mask);
+				break;
+			case AMD_IP_BLOCK_TYPE_SDMA:
+				ret = amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_SDMA, &inst_mask);
+				if (ret)
+					return ret;
+				count = hweight32(inst_mask);
+				break;
+			case AMD_IP_BLOCK_TYPE_JPEG:
+				ret = amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_VCN, &inst_mask);
+				if (ret)
+					return ret;
+				count = hweight32(inst_mask) * adev->jpeg.num_jpeg_rings;
+				break;
+			case AMD_IP_BLOCK_TYPE_VCN:
+				ret = amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_VCN, &inst_mask);
+				if (ret)
+					return ret;
+				count = hweight32(inst_mask);
+				break;
+			default:
+				return -EINVAL;
+			}
+
+			return copy_to_user(out, &count, min(size, 4u)) ? -EFAULT : 0;
+		}
+
+		switch (type) {
+		case AMD_IP_BLOCK_TYPE_GFX:
+		case AMD_IP_BLOCK_TYPE_VCE:
+			count = 1;
 			break;
-		case AMDGPU_HW_IP_VCE:
-			type = AMD_IP_BLOCK_TYPE_VCE;
+		case AMD_IP_BLOCK_TYPE_SDMA:
+			count = adev->sdma.num_instances;
 			break;
-		case AMDGPU_HW_IP_UVD_ENC:
-			type = AMD_IP_BLOCK_TYPE_UVD;
+		case AMD_IP_BLOCK_TYPE_JPEG:
+			count = adev->jpeg.num_jpeg_inst * adev->jpeg.num_jpeg_rings;
 			break;
-		case AMDGPU_HW_IP_VCN_DEC:
-		case AMDGPU_HW_IP_VCN_ENC:
-			type = AMD_IP_BLOCK_TYPE_VCN;
+		case AMD_IP_BLOCK_TYPE_VCN:
+			count = adev->vcn.num_vcn_inst;
 			break;
-		case AMDGPU_HW_IP_VCN_JPEG:
-			type = (amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_JPEG)) ?
-				AMD_IP_BLOCK_TYPE_JPEG : AMD_IP_BLOCK_TYPE_VCN;
+		case AMD_IP_BLOCK_TYPE_UVD:
+			count = adev->uvd.num_uvd_inst;
 			break;
+		/* For all other IP block types not listed in the switch statement
+		 * the ip status is valid here and the instance count is one.
+		 */
 		default:
-			return -EINVAL;
+			count = 1;
+			break;
 		}
 
-		for (i = 0; i < adev->num_ip_blocks; i++)
-			if (adev->ip_blocks[i].version->type == type &&
-			    adev->ip_blocks[i].status.valid &&
-			    count < AMDGPU_HW_IP_INSTANCE_MAX_COUNT)
-				count++;
-
 		return copy_to_user(out, &count, min(size, 4u)) ? -EFAULT : 0;
 	}
 	case AMDGPU_INFO_TIMESTAMP:
@@ -625,7 +736,6 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0;
 	case AMDGPU_INFO_FW_VERSION: {
 		struct drm_amdgpu_info_firmware fw_info;
-		int ret;
 
 		/* We only support one instance of each IP block right now. */
 		if (info->query_fw.ip_instance != 0)
@@ -648,7 +758,8 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		ui64 = atomic64_read(&adev->num_vram_cpu_page_faults);
 		return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0;
 	case AMDGPU_INFO_VRAM_USAGE:
-		ui64 = ttm_resource_manager_usage(&adev->mman.vram_mgr.manager);
+		ui64 = ttm_resource_manager_used(&adev->mman.vram_mgr.manager) ?
+			ttm_resource_manager_usage(&adev->mman.vram_mgr.manager) : 0;
 		return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0;
 	case AMDGPU_INFO_VIS_VRAM_USAGE:
 		ui64 = amdgpu_vram_mgr_vis_usage(&adev->mman.vram_mgr);
@@ -694,8 +805,8 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		mem.vram.usable_heap_size = adev->gmc.real_vram_size -
 			atomic64_read(&adev->vram_pin_size) -
 			AMDGPU_VM_RESERVED_VRAM;
-		mem.vram.heap_usage =
-			ttm_resource_manager_usage(vram_man);
+		mem.vram.heap_usage = ttm_resource_manager_used(&adev->mman.vram_mgr.manager) ?
+				ttm_resource_manager_usage(vram_man) : 0;
 		mem.vram.max_allocation = mem.vram.usable_heap_size * 3 / 4;
 
 		mem.cpu_accessible_vram.total_heap_size =
@@ -720,32 +831,47 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 				    ? -EFAULT : 0;
 	}
 	case AMDGPU_INFO_READ_MMR_REG: {
-		unsigned n, alloc_size;
+		int ret = 0;
+		unsigned int n, alloc_size;
 		uint32_t *regs;
-		unsigned se_num = (info->read_mmr_reg.instance >>
+		unsigned int se_num = (info->read_mmr_reg.instance >>
 				   AMDGPU_INFO_MMR_SE_INDEX_SHIFT) &
 				  AMDGPU_INFO_MMR_SE_INDEX_MASK;
-		unsigned sh_num = (info->read_mmr_reg.instance >>
+		unsigned int sh_num = (info->read_mmr_reg.instance >>
 				   AMDGPU_INFO_MMR_SH_INDEX_SHIFT) &
 				  AMDGPU_INFO_MMR_SH_INDEX_MASK;
 
+		if (!down_read_trylock(&adev->reset_domain->sem))
+			return -ENOENT;
+
 		/* set full masks if the userspace set all bits
-		 * in the bitfields */
-		if (se_num == AMDGPU_INFO_MMR_SE_INDEX_MASK)
+		 * in the bitfields
+		 */
+		if (se_num == AMDGPU_INFO_MMR_SE_INDEX_MASK) {
 			se_num = 0xffffffff;
-		else if (se_num >= AMDGPU_GFX_MAX_SE)
-			return -EINVAL;
-		if (sh_num == AMDGPU_INFO_MMR_SH_INDEX_MASK)
+		} else if (se_num >= AMDGPU_GFX_MAX_SE) {
+			ret = -EINVAL;
+			goto out;
+		}
+
+		if (sh_num == AMDGPU_INFO_MMR_SH_INDEX_MASK) {
 			sh_num = 0xffffffff;
-		else if (sh_num >= AMDGPU_GFX_MAX_SH_PER_SE)
-			return -EINVAL;
+		} else if (sh_num >= AMDGPU_GFX_MAX_SH_PER_SE) {
+			ret = -EINVAL;
+			goto out;
+		}
 
-		if (info->read_mmr_reg.count > 128)
-			return -EINVAL;
+		if (info->read_mmr_reg.count > 128) {
+			ret = -EINVAL;
+			goto out;
+		}
 
 		regs = kmalloc_array(info->read_mmr_reg.count, sizeof(*regs), GFP_KERNEL);
-		if (!regs)
-			return -ENOMEM;
+		if (!regs) {
+			ret = -ENOMEM;
+			goto out;
+		}
+
 		alloc_size = info->read_mmr_reg.count * sizeof(*regs);
 
 		amdgpu_gfx_off_ctrl(adev, false);
@@ -757,19 +883,22 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 					      info->read_mmr_reg.dword_offset + i);
 				kfree(regs);
 				amdgpu_gfx_off_ctrl(adev, true);
-				return -EFAULT;
+				ret = -EFAULT;
+				goto out;
 			}
 		}
 		amdgpu_gfx_off_ctrl(adev, true);
 		n = copy_to_user(out, regs, min(size, alloc_size));
 		kfree(regs);
-		return n ? -EFAULT : 0;
+		ret = (n ? -EFAULT : 0);
+out:
+		up_read(&adev->reset_domain->sem);
+		return ret;
 	}
 	case AMDGPU_INFO_DEV_INFO: {
 		struct drm_amdgpu_info_device *dev_info;
 		uint64_t vm_size;
-		uint32_t pcie_gen_mask;
-		int ret;
+		uint32_t pcie_gen_mask, pcie_width_mask;
 
 		dev_info = kzalloc(sizeof(*dev_info), GFP_KERNEL);
 		if (!dev_info)
@@ -804,22 +933,35 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		dev_info->ids_flags = 0;
 		if (adev->flags & AMD_IS_APU)
 			dev_info->ids_flags |= AMDGPU_IDS_FLAGS_FUSION;
-		if (amdgpu_mcbp)
+		if (adev->gfx.mcbp)
 			dev_info->ids_flags |= AMDGPU_IDS_FLAGS_PREEMPTION;
 		if (amdgpu_is_tmz(adev))
 			dev_info->ids_flags |= AMDGPU_IDS_FLAGS_TMZ;
 		if (adev->gfx.config.ta_cntl2_truncate_coord_mode)
 			dev_info->ids_flags |= AMDGPU_IDS_FLAGS_CONFORMANT_TRUNC_COORD;
 
+		/* Gang submit is not supported under SRIOV currently */
+		if (!amdgpu_sriov_vf(adev))
+			dev_info->ids_flags |= AMDGPU_IDS_FLAGS_GANG_SUBMIT;
+
+		if (amdgpu_passthrough(adev))
+			dev_info->ids_flags |= (AMDGPU_IDS_FLAGS_MODE_PT <<
+						AMDGPU_IDS_FLAGS_MODE_SHIFT) &
+						AMDGPU_IDS_FLAGS_MODE_MASK;
+		else if (amdgpu_sriov_vf(adev))
+			dev_info->ids_flags |= (AMDGPU_IDS_FLAGS_MODE_VF <<
+						AMDGPU_IDS_FLAGS_MODE_SHIFT) &
+						AMDGPU_IDS_FLAGS_MODE_MASK;
+
 		vm_size = adev->vm_manager.max_pfn * AMDGPU_GPU_PAGE_SIZE;
-		vm_size -= AMDGPU_VA_RESERVED_SIZE;
+		vm_size -= AMDGPU_VA_RESERVED_TOP;
 
 		/* Older VCE FW versions are buggy and can handle only 40bits */
 		if (adev->vce.fw_version &&
 		    adev->vce.fw_version < AMDGPU_VCE_FW_53_45)
 			vm_size = min(vm_size, 1ULL << 40);
 
-		dev_info->virtual_address_offset = AMDGPU_VA_RESERVED_SIZE;
+		dev_info->virtual_address_offset = AMDGPU_VA_RESERVED_BOTTOM;
 		dev_info->virtual_address_max =
 			min(vm_size, AMDGPU_GMC_HOLE_START);
 
@@ -836,7 +978,7 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		memcpy(&dev_info->cu_ao_bitmap[0], &adev->gfx.cu_info.ao_cu_bitmap[0],
 		       sizeof(adev->gfx.cu_info.ao_cu_bitmap));
 		memcpy(&dev_info->cu_bitmap[0], &adev->gfx.cu_info.bitmap[0],
-		       sizeof(adev->gfx.cu_info.bitmap));
+		       sizeof(dev_info->cu_bitmap));
 		dev_info->vram_type = adev->gmc.vram_type;
 		dev_info->vram_bit_width = adev->gmc.vram_width;
 		dev_info->vce_harvest_config = adev->vce.harvest_config;
@@ -857,15 +999,18 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		dev_info->tcc_disabled_mask = adev->gfx.config.tcc_disabled_mask;
 
 		/* Combine the chip gen mask with the platform (CPU/mobo) mask. */
-		pcie_gen_mask = adev->pm.pcie_gen_mask & (adev->pm.pcie_gen_mask >> 16);
+		pcie_gen_mask = adev->pm.pcie_gen_mask &
+			(adev->pm.pcie_gen_mask >> CAIL_PCIE_LINK_SPEED_SUPPORT_SHIFT);
+		pcie_width_mask = adev->pm.pcie_mlw_mask &
+			(adev->pm.pcie_mlw_mask >> CAIL_PCIE_LINK_WIDTH_SUPPORT_SHIFT);
 		dev_info->pcie_gen = fls(pcie_gen_mask);
 		dev_info->pcie_num_lanes =
-			adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X32 ? 32 :
-			adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 ? 16 :
-			adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 ? 12 :
-			adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 ? 8 :
-			adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 ? 4 :
-			adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 ? 2 : 1;
+			pcie_width_mask & CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X32 ? 32 :
+			pcie_width_mask & CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X16 ? 16 :
+			pcie_width_mask & CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X12 ? 12 :
+			pcie_width_mask & CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X8 ? 8 :
+			pcie_width_mask & CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 ? 4 :
+			pcie_width_mask & CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 ? 2 : 1;
 
 		dev_info->tcp_cache_size = adev->gfx.config.gc_tcp_l1_size;
 		dev_info->num_sqc_per_wgp = adev->gfx.config.gc_num_sqc_per_wgp;
@@ -876,13 +1021,28 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		dev_info->gl2c_cache_size = adev->gfx.config.gc_gl2c_per_gpu;
 		dev_info->mall_size = adev->gmc.mall_size;
 
+
+		if (adev->gfx.funcs->get_gfx_shadow_info) {
+			struct amdgpu_gfx_shadow_info shadow_info;
+
+			ret = amdgpu_gfx_get_gfx_shadow_info(adev, &shadow_info);
+			if (!ret) {
+				dev_info->shadow_size = shadow_info.shadow_size;
+				dev_info->shadow_alignment = shadow_info.shadow_alignment;
+				dev_info->csa_size = shadow_info.csa_size;
+				dev_info->csa_alignment = shadow_info.csa_alignment;
+			}
+		}
+
+		dev_info->userq_ip_mask = amdgpu_userq_get_supported_ip_mask(adev);
+
 		ret = copy_to_user(out, dev_info,
 				   min((size_t)size, sizeof(*dev_info))) ? -EFAULT : 0;
 		kfree(dev_info);
 		return ret;
 	}
 	case AMDGPU_INFO_VCE_CLOCK_TABLE: {
-		unsigned i;
+		unsigned int i;
 		struct drm_amdgpu_info_vce_clock_table vce_clk_table = {};
 		struct amd_vce_state *vce_state;
 
@@ -924,12 +1084,17 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 			struct atom_context *atom_context;
 
 			atom_context = adev->mode_info.atom_context;
-			memcpy(vbios_info.name, atom_context->name, sizeof(atom_context->name));
-			memcpy(vbios_info.vbios_pn, atom_context->vbios_pn, sizeof(atom_context->vbios_pn));
-			vbios_info.version = atom_context->version;
-			memcpy(vbios_info.vbios_ver_str, atom_context->vbios_ver_str,
-						sizeof(atom_context->vbios_ver_str));
-			memcpy(vbios_info.date, atom_context->date, sizeof(atom_context->date));
+			if (atom_context) {
+				memcpy(vbios_info.name, atom_context->name,
+				       sizeof(atom_context->name));
+				memcpy(vbios_info.vbios_pn, atom_context->vbios_pn,
+				       sizeof(atom_context->vbios_pn));
+				vbios_info.version = atom_context->version;
+				memcpy(vbios_info.vbios_ver_str, atom_context->vbios_ver_str,
+				       sizeof(atom_context->vbios_ver_str));
+				memcpy(vbios_info.date, atom_context->date,
+				       sizeof(atom_context->date));
+			}
 
 			return copy_to_user(out, &vbios_info,
 						min((size_t)size, sizeof(vbios_info))) ? -EFAULT : 0;
@@ -1003,7 +1168,21 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		case AMDGPU_INFO_SENSOR_GPU_AVG_POWER:
 			/* get average GPU power */
 			if (amdgpu_dpm_read_sensor(adev,
-						   AMDGPU_PP_SENSOR_GPU_POWER,
+						   AMDGPU_PP_SENSOR_GPU_AVG_POWER,
+						   (void *)&ui32, &ui32_size)) {
+				/* fall back to input power for backwards compat */
+				if (amdgpu_dpm_read_sensor(adev,
+							   AMDGPU_PP_SENSOR_GPU_INPUT_POWER,
+							   (void *)&ui32, &ui32_size)) {
+					return -EINVAL;
+				}
+			}
+			ui32 >>= 8;
+			break;
+		case AMDGPU_INFO_SENSOR_GPU_INPUT_POWER:
+			/* get input GPU power */
+			if (amdgpu_dpm_read_sensor(adev,
+						   AMDGPU_PP_SENSOR_GPU_INPUT_POWER,
 						   (void *)&ui32, &ui32_size)) {
 				return -EINVAL;
 			}
@@ -1088,6 +1267,9 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		struct drm_amdgpu_info_video_caps *caps;
 		int r;
 
+		if (!adev->asic_funcs->query_video_codecs)
+			return -EINVAL;
+
 		switch (info->video_cap.type) {
 		case AMDGPU_INFO_VIDEO_CAPS_DECODE:
 			r = amdgpu_asic_query_video_codecs(adev, false, &codecs);
@@ -1140,6 +1322,51 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		kfree(caps);
 		return r;
 	}
+	case AMDGPU_INFO_MAX_IBS: {
+		uint32_t max_ibs[AMDGPU_HW_IP_NUM];
+
+		for (i = 0; i < AMDGPU_HW_IP_NUM; ++i)
+			max_ibs[i] = amdgpu_ring_max_ibs(i);
+
+		return copy_to_user(out, max_ibs,
+				    min((size_t)size, sizeof(max_ibs))) ? -EFAULT : 0;
+	}
+	case AMDGPU_INFO_GPUVM_FAULT: {
+		struct amdgpu_fpriv *fpriv = filp->driver_priv;
+		struct amdgpu_vm *vm = &fpriv->vm;
+		struct drm_amdgpu_info_gpuvm_fault gpuvm_fault;
+		unsigned long flags;
+
+		if (!vm)
+			return -EINVAL;
+
+		memset(&gpuvm_fault, 0, sizeof(gpuvm_fault));
+
+		xa_lock_irqsave(&adev->vm_manager.pasids, flags);
+		gpuvm_fault.addr = vm->fault_info.addr;
+		gpuvm_fault.status = vm->fault_info.status;
+		gpuvm_fault.vmhub = vm->fault_info.vmhub;
+		xa_unlock_irqrestore(&adev->vm_manager.pasids, flags);
+
+		return copy_to_user(out, &gpuvm_fault,
+				    min((size_t)size, sizeof(gpuvm_fault))) ? -EFAULT : 0;
+	}
+	case AMDGPU_INFO_UQ_FW_AREAS: {
+		struct drm_amdgpu_info_uq_metadata meta_info = {};
+
+		switch (info->query_hw_ip.type) {
+		case AMDGPU_HW_IP_GFX:
+			ret = amdgpu_userq_metadata_info_gfx(adev, info, &meta_info.gfx);
+			if (ret)
+				return ret;
+
+			ret = copy_to_user(out, &meta_info,
+						min((size_t)size, sizeof(meta_info))) ? -EFAULT : 0;
+			return 0;
+		default:
+			return -EINVAL;
+		}
+	}
 	default:
 		DRM_DEBUG_KMS("Invalid request %d\n", info->query);
 		return -EINVAL;
@@ -1147,23 +1374,6 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	return 0;
 }
 
-
-/*
- * Outdated mess for old drm with Xorg being in charge (void function now).
- */
-/**
- * amdgpu_driver_lastclose_kms - drm callback for last close
- *
- * @dev: drm dev pointer
- *
- * Switch vga_switcheroo state after last close (all asics).
- */
-void amdgpu_driver_lastclose_kms(struct drm_device *dev)
-{
-	drm_fb_helper_lastclose(dev);
-	vga_switcheroo_process_delayed_switch();
-}
-
 /**
  * amdgpu_driver_open_kms - drm callback for open
  *
@@ -1206,13 +1416,15 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv)
 		pasid = 0;
 	}
 
-	r = amdgpu_vm_init(adev, &fpriv->vm);
+	r = amdgpu_xcp_open_device(adev, fpriv, file_priv);
 	if (r)
 		goto error_pasid;
 
-	r = amdgpu_vm_set_pasid(adev, &fpriv->vm, pasid);
+	amdgpu_debugfs_vm_init(file_priv);
+
+	r = amdgpu_vm_init(adev, &fpriv->vm, fpriv->xcp_id, pasid);
 	if (r)
-		goto error_vm;
+		goto error_pasid;
 
 	fpriv->prt_va = amdgpu_vm_bo_add(adev, &fpriv->vm, NULL);
 	if (!fpriv->prt_va) {
@@ -1220,7 +1432,7 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv)
 		goto error_vm;
 	}
 
-	if (amdgpu_mcbp) {
+	if (adev->gfx.mcbp) {
 		uint64_t csa_addr = amdgpu_csa_vaddr(adev) & AMDGPU_GMC_HOLE_MASK;
 
 		r = amdgpu_map_static_csa(adev, &fpriv->vm, adev->virt.csa_obj,
@@ -1229,9 +1441,21 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv)
 			goto error_vm;
 	}
 
+	r = amdgpu_seq64_map(adev, &fpriv->vm, &fpriv->seq64_va);
+	if (r)
+		goto error_vm;
+
 	mutex_init(&fpriv->bo_list_lock);
 	idr_init_base(&fpriv->bo_list_handles, 1);
 
+	r = amdgpu_userq_mgr_init(&fpriv->userq_mgr, file_priv, adev);
+	if (r)
+		DRM_WARN("Can't setup usermode queues, use legacy workload submission only\n");
+
+	r = amdgpu_eviction_fence_init(&fpriv->evf_mgr);
+	if (r)
+		goto error_vm;
+
 	amdgpu_ctx_mgr_init(&fpriv->ctx_mgr, adev);
 
 	file_priv->driver_priv = fpriv;
@@ -1241,15 +1465,12 @@ error_vm:
 	amdgpu_vm_fini(adev, &fpriv->vm);
 
 error_pasid:
-	if (pasid) {
+	if (pasid)
 		amdgpu_pasid_free(pasid);
-		amdgpu_vm_set_pasid(adev, &fpriv->vm, 0);
-	}
 
 	kfree(fpriv);
 
 out_suspend:
-	pm_runtime_mark_last_busy(dev->dev);
 pm_put:
 	pm_runtime_put_autosuspend(dev->dev);
 
@@ -1284,14 +1505,16 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev,
 	if (amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_VCE) != NULL)
 		amdgpu_vce_free_handles(adev, file_priv);
 
-	if (amdgpu_mcbp) {
-		/* TODO: how to handle reserve failure */
-		BUG_ON(amdgpu_bo_reserve(adev->virt.csa_obj, true));
-		amdgpu_vm_bo_del(adev, fpriv->csa_va);
+	if (fpriv->csa_va) {
+		uint64_t csa_addr = amdgpu_csa_vaddr(adev) & AMDGPU_GMC_HOLE_MASK;
+
+		WARN_ON(amdgpu_unmap_static_csa(adev, &fpriv->vm, adev->virt.csa_obj,
+						fpriv->csa_va, csa_addr));
 		fpriv->csa_va = NULL;
-		amdgpu_bo_unreserve(adev->virt.csa_obj);
 	}
 
+	amdgpu_seq64_unmap(adev, fpriv);
+
 	pasid = fpriv->vm.pasid;
 	pd = amdgpu_bo_ref(fpriv->vm.root.bo);
 	if (!WARN_ON(amdgpu_bo_reserve(pd, true))) {
@@ -1315,7 +1538,6 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev,
 	kfree(fpriv);
 	file_priv->driver_priv = NULL;
 
-	pm_runtime_mark_last_busy(dev->dev);
 	pm_runtime_put_autosuspend(dev->dev);
 }
 
@@ -1441,7 +1663,7 @@ void amdgpu_disable_vblank_kms(struct drm_crtc *crtc)
 
 static int amdgpu_debugfs_firmware_info_show(struct seq_file *m, void *unused)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct amdgpu_device *adev = m->private;
 	struct drm_amdgpu_info_firmware fw_info;
 	struct drm_amdgpu_query_fw query_fw;
 	struct atom_context *ctx = adev->mode_info.atom_context;
@@ -1449,7 +1671,7 @@ static int amdgpu_debugfs_firmware_info_show(struct seq_file *m, void *unused)
 	int ret, i;
 
 	static const char *ta_fw_name[TA_FW_TYPE_MAX_INDEX] = {
-#define TA_FW_NAME(type) [TA_FW_TYPE_PSP_##type] = #type
+#define TA_FW_NAME(type)[TA_FW_TYPE_PSP_##type] = #type
 		TA_FW_NAME(XGMI),
 		TA_FW_NAME(RAS),
 		TA_FW_NAME(HDCP),
@@ -1548,7 +1770,7 @@ static int amdgpu_debugfs_firmware_info_show(struct seq_file *m, void *unused)
 		   fw_info.feature, fw_info.ver);
 
 	/* RLCV */
-        query_fw.fw_type = AMDGPU_INFO_FW_GFX_RLCV;
+	query_fw.fw_type = AMDGPU_INFO_FW_GFX_RLCV;
 	ret = amdgpu_firmware_info(&fw_info, &query_fw, adev);
 	if (ret)
 		return ret;
@@ -1692,7 +1914,15 @@ static int amdgpu_debugfs_firmware_info_show(struct seq_file *m, void *unused)
 	seq_printf(m, "MES feature version: %u, firmware version: 0x%08x\n",
 		   fw_info.feature, fw_info.ver);
 
-	seq_printf(m, "VBIOS version: %s\n", ctx->vbios_version);
+	/* VPE */
+	query_fw.fw_type = AMDGPU_INFO_FW_VPE;
+	ret = amdgpu_firmware_info(&fw_info, &query_fw, adev);
+	if (ret)
+		return ret;
+	seq_printf(m, "VPE feature version: %u, firmware version: 0x%08x\n",
+		   fw_info.feature, fw_info.ver);
+
+	seq_printf(m, "VBIOS version: %s\n", ctx->vbios_pn);
 
 	return 0;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c
index 8d9ff9e151de..3ca03b5e0f91 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c
@@ -27,6 +27,16 @@
 #include "umc/umc_6_7_0_offset.h"
 #include "umc/umc_6_7_0_sh_mask.h"
 
+static bool amdgpu_mca_is_deferred_error(struct amdgpu_device *adev,
+					uint64_t mc_status)
+{
+	if (adev->umc.ras->check_ecc_err_status)
+		return adev->umc.ras->check_ecc_err_status(adev,
+				AMDGPU_MCA_ERROR_TYPE_DE, &mc_status);
+
+	return false;
+}
+
 void amdgpu_mca_query_correctable_error_count(struct amdgpu_device *adev,
 					      uint64_t mc_status_addr,
 					      unsigned long *error_count)
@@ -142,3 +152,479 @@ int amdgpu_mca_mpio_ras_sw_init(struct amdgpu_device *adev)
 
 	return 0;
 }
+
+static void amdgpu_mca_bank_set_init(struct mca_bank_set *mca_set)
+{
+	if (!mca_set)
+		return;
+
+	memset(mca_set, 0, sizeof(*mca_set));
+	INIT_LIST_HEAD(&mca_set->list);
+}
+
+static int amdgpu_mca_bank_set_add_entry(struct mca_bank_set *mca_set, struct mca_bank_entry *entry)
+{
+	struct mca_bank_node *node;
+
+	if (!entry)
+		return -EINVAL;
+
+	node = kvzalloc(sizeof(*node), GFP_KERNEL);
+	if (!node)
+		return -ENOMEM;
+
+	memcpy(&node->entry, entry, sizeof(*entry));
+
+	INIT_LIST_HEAD(&node->node);
+	list_add_tail(&node->node, &mca_set->list);
+
+	mca_set->nr_entries++;
+
+	return 0;
+}
+
+static int amdgpu_mca_bank_set_merge(struct mca_bank_set *mca_set, struct mca_bank_set *new)
+{
+	struct mca_bank_node *node;
+
+	list_for_each_entry(node, &new->list, node)
+		amdgpu_mca_bank_set_add_entry(mca_set, &node->entry);
+
+	return 0;
+}
+
+static void amdgpu_mca_bank_set_remove_node(struct mca_bank_set *mca_set, struct mca_bank_node *node)
+{
+	if (!node)
+		return;
+
+	list_del(&node->node);
+	kvfree(node);
+
+	mca_set->nr_entries--;
+}
+
+static void amdgpu_mca_bank_set_release(struct mca_bank_set *mca_set)
+{
+	struct mca_bank_node *node, *tmp;
+
+	if (list_empty(&mca_set->list))
+		return;
+
+	list_for_each_entry_safe(node, tmp, &mca_set->list, node)
+		amdgpu_mca_bank_set_remove_node(mca_set, node);
+}
+
+void amdgpu_mca_smu_init_funcs(struct amdgpu_device *adev, const struct amdgpu_mca_smu_funcs *mca_funcs)
+{
+	struct amdgpu_mca *mca = &adev->mca;
+
+	mca->mca_funcs = mca_funcs;
+}
+
+int amdgpu_mca_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_mca *mca = &adev->mca;
+	struct mca_bank_cache *mca_cache;
+	int i;
+
+	atomic_set(&mca->ue_update_flag, 0);
+
+	for (i = 0; i < ARRAY_SIZE(mca->mca_caches); i++) {
+		mca_cache = &mca->mca_caches[i];
+		mutex_init(&mca_cache->lock);
+		amdgpu_mca_bank_set_init(&mca_cache->mca_set);
+	}
+
+	return 0;
+}
+
+void amdgpu_mca_fini(struct amdgpu_device *adev)
+{
+	struct amdgpu_mca *mca = &adev->mca;
+	struct mca_bank_cache *mca_cache;
+	int i;
+
+	atomic_set(&mca->ue_update_flag, 0);
+
+	for (i = 0; i < ARRAY_SIZE(mca->mca_caches); i++) {
+		mca_cache = &mca->mca_caches[i];
+		amdgpu_mca_bank_set_release(&mca_cache->mca_set);
+		mutex_destroy(&mca_cache->lock);
+	}
+}
+
+int amdgpu_mca_reset(struct amdgpu_device *adev)
+{
+	amdgpu_mca_fini(adev);
+
+	return amdgpu_mca_init(adev);
+}
+
+int amdgpu_mca_smu_set_debug_mode(struct amdgpu_device *adev, bool enable)
+{
+	const struct amdgpu_mca_smu_funcs *mca_funcs = adev->mca.mca_funcs;
+
+	if (mca_funcs && mca_funcs->mca_set_debug_mode)
+		return mca_funcs->mca_set_debug_mode(adev, enable);
+
+	return -EOPNOTSUPP;
+}
+
+static void amdgpu_mca_smu_mca_bank_dump(struct amdgpu_device *adev, int idx, struct mca_bank_entry *entry,
+					 struct ras_query_context *qctx)
+{
+	u64 event_id = qctx ? qctx->evid.event_id : RAS_EVENT_INVALID_ID;
+
+	RAS_EVENT_LOG(adev, event_id, HW_ERR "Accelerator Check Architecture events logged\n");
+	RAS_EVENT_LOG(adev, event_id, HW_ERR "aca entry[%02d].STATUS=0x%016llx\n",
+		      idx, entry->regs[MCA_REG_IDX_STATUS]);
+	RAS_EVENT_LOG(adev, event_id, HW_ERR "aca entry[%02d].ADDR=0x%016llx\n",
+		      idx, entry->regs[MCA_REG_IDX_ADDR]);
+	RAS_EVENT_LOG(adev, event_id, HW_ERR "aca entry[%02d].MISC0=0x%016llx\n",
+		      idx, entry->regs[MCA_REG_IDX_MISC0]);
+	RAS_EVENT_LOG(adev, event_id, HW_ERR "aca entry[%02d].IPID=0x%016llx\n",
+		      idx, entry->regs[MCA_REG_IDX_IPID]);
+	RAS_EVENT_LOG(adev, event_id, HW_ERR "aca entry[%02d].SYND=0x%016llx\n",
+		      idx, entry->regs[MCA_REG_IDX_SYND]);
+}
+
+static int amdgpu_mca_smu_get_valid_mca_count(struct amdgpu_device *adev, enum amdgpu_mca_error_type type, uint32_t *count)
+{
+	const struct amdgpu_mca_smu_funcs *mca_funcs = adev->mca.mca_funcs;
+
+	if (!count)
+		return -EINVAL;
+
+	if (mca_funcs && mca_funcs->mca_get_valid_mca_count)
+		return mca_funcs->mca_get_valid_mca_count(adev, type, count);
+
+	return -EOPNOTSUPP;
+}
+
+static int amdgpu_mca_smu_get_mca_entry(struct amdgpu_device *adev, enum amdgpu_mca_error_type type,
+					int idx, struct mca_bank_entry *entry)
+{
+	const struct amdgpu_mca_smu_funcs *mca_funcs = adev->mca.mca_funcs;
+	int count;
+
+	if (!mca_funcs || !mca_funcs->mca_get_mca_entry)
+		return -EOPNOTSUPP;
+
+	switch (type) {
+	case AMDGPU_MCA_ERROR_TYPE_UE:
+		count = mca_funcs->max_ue_count;
+		break;
+	case AMDGPU_MCA_ERROR_TYPE_CE:
+		count = mca_funcs->max_ce_count;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if (idx >= count)
+		return -EINVAL;
+
+	return mca_funcs->mca_get_mca_entry(adev, type, idx, entry);
+}
+
+static bool amdgpu_mca_bank_should_update(struct amdgpu_device *adev, enum amdgpu_mca_error_type type)
+{
+	struct amdgpu_mca *mca = &adev->mca;
+	bool ret = true;
+
+	/*
+	 * Because the UE Valid MCA count will only be cleared after reset,
+	 * in order to avoid repeated counting of the error count,
+	 * the aca bank is only updated once during the gpu recovery stage.
+	 */
+	if (type == AMDGPU_MCA_ERROR_TYPE_UE) {
+		if (amdgpu_ras_intr_triggered())
+			ret = atomic_cmpxchg(&mca->ue_update_flag, 0, 1) == 0;
+		else
+			atomic_set(&mca->ue_update_flag, 0);
+	}
+
+	return ret;
+}
+
+static bool amdgpu_mca_bank_should_dump(struct amdgpu_device *adev, enum amdgpu_mca_error_type type,
+					struct mca_bank_entry *entry)
+{
+	bool ret;
+
+	switch (type) {
+	case AMDGPU_MCA_ERROR_TYPE_CE:
+		ret = amdgpu_mca_is_deferred_error(adev, entry->regs[MCA_REG_IDX_STATUS]);
+		break;
+	case AMDGPU_MCA_ERROR_TYPE_UE:
+	default:
+		ret = true;
+		break;
+	}
+
+	return ret;
+}
+
+static int amdgpu_mca_smu_get_mca_set(struct amdgpu_device *adev, enum amdgpu_mca_error_type type, struct mca_bank_set *mca_set,
+				      struct ras_query_context *qctx)
+{
+	struct mca_bank_entry entry;
+	uint32_t count = 0, i;
+	int ret;
+
+	if (!mca_set)
+		return -EINVAL;
+
+	if (!amdgpu_mca_bank_should_update(adev, type))
+		return 0;
+
+	ret = amdgpu_mca_smu_get_valid_mca_count(adev, type, &count);
+	if (ret)
+		return ret;
+
+	for (i = 0; i < count; i++) {
+		memset(&entry, 0, sizeof(entry));
+		ret = amdgpu_mca_smu_get_mca_entry(adev, type, i, &entry);
+		if (ret)
+			return ret;
+
+		amdgpu_mca_bank_set_add_entry(mca_set, &entry);
+
+		if (amdgpu_mca_bank_should_dump(adev, type, &entry))
+			amdgpu_mca_smu_mca_bank_dump(adev, i, &entry, qctx);
+	}
+
+	return 0;
+}
+
+static int amdgpu_mca_smu_parse_mca_error_count(struct amdgpu_device *adev, enum amdgpu_ras_block blk,
+						enum amdgpu_mca_error_type type, struct mca_bank_entry *entry, uint32_t *count)
+{
+	const struct amdgpu_mca_smu_funcs *mca_funcs = adev->mca.mca_funcs;
+
+	if (!count || !entry)
+		return -EINVAL;
+
+	if (!mca_funcs || !mca_funcs->mca_parse_mca_error_count)
+		return -EOPNOTSUPP;
+
+	return mca_funcs->mca_parse_mca_error_count(adev, blk, type, entry, count);
+}
+
+static int amdgpu_mca_dispatch_mca_set(struct amdgpu_device *adev, enum amdgpu_ras_block blk, enum amdgpu_mca_error_type type,
+				       struct mca_bank_set *mca_set, struct ras_err_data *err_data)
+{
+	struct amdgpu_smuio_mcm_config_info mcm_info;
+	struct mca_bank_node *node, *tmp;
+	struct mca_bank_entry *entry;
+	uint32_t count;
+	int ret;
+
+	if (!mca_set)
+		return -EINVAL;
+
+	if (!mca_set->nr_entries)
+		return 0;
+
+	list_for_each_entry_safe(node, tmp, &mca_set->list, node) {
+		entry = &node->entry;
+
+		count = 0;
+		ret = amdgpu_mca_smu_parse_mca_error_count(adev, blk, type, entry, &count);
+		if (ret && ret != -EOPNOTSUPP)
+			return ret;
+
+		if (!count)
+			continue;
+
+		memset(&mcm_info, 0, sizeof(mcm_info));
+
+		mcm_info.socket_id = entry->info.socket_id;
+		mcm_info.die_id = entry->info.aid;
+
+		if (type == AMDGPU_MCA_ERROR_TYPE_UE) {
+			amdgpu_ras_error_statistic_ue_count(err_data,
+							    &mcm_info, (uint64_t)count);
+		} else {
+			if (amdgpu_mca_is_deferred_error(adev, entry->regs[MCA_REG_IDX_STATUS]))
+				amdgpu_ras_error_statistic_de_count(err_data,
+								    &mcm_info, (uint64_t)count);
+			else
+				amdgpu_ras_error_statistic_ce_count(err_data,
+								    &mcm_info, (uint64_t)count);
+		}
+
+		amdgpu_mca_bank_set_remove_node(mca_set, node);
+	}
+
+	return 0;
+}
+
+static int amdgpu_mca_add_mca_set_to_cache(struct amdgpu_device *adev, enum amdgpu_mca_error_type type, struct mca_bank_set *new)
+{
+	struct mca_bank_cache *mca_cache = &adev->mca.mca_caches[type];
+	int ret;
+
+	mutex_lock(&mca_cache->lock);
+	ret = amdgpu_mca_bank_set_merge(&mca_cache->mca_set, new);
+	mutex_unlock(&mca_cache->lock);
+
+	return ret;
+}
+
+int amdgpu_mca_smu_log_ras_error(struct amdgpu_device *adev, enum amdgpu_ras_block blk, enum amdgpu_mca_error_type type,
+				 struct ras_err_data *err_data, struct ras_query_context *qctx)
+{
+	struct mca_bank_set mca_set;
+	struct mca_bank_cache *mca_cache = &adev->mca.mca_caches[type];
+	int ret;
+
+	amdgpu_mca_bank_set_init(&mca_set);
+
+	ret = amdgpu_mca_smu_get_mca_set(adev, type, &mca_set, qctx);
+	if (ret)
+		goto out_mca_release;
+
+	ret = amdgpu_mca_dispatch_mca_set(adev, blk, type, &mca_set, err_data);
+	if (ret)
+		goto out_mca_release;
+
+	/* add remain mca bank to mca cache */
+	if (mca_set.nr_entries) {
+		ret = amdgpu_mca_add_mca_set_to_cache(adev, type, &mca_set);
+		if (ret)
+			goto out_mca_release;
+	}
+
+	/* dispatch mca set again if mca cache has valid data */
+	mutex_lock(&mca_cache->lock);
+	if (mca_cache->mca_set.nr_entries)
+		ret = amdgpu_mca_dispatch_mca_set(adev, blk, type, &mca_cache->mca_set, err_data);
+	mutex_unlock(&mca_cache->lock);
+
+out_mca_release:
+	amdgpu_mca_bank_set_release(&mca_set);
+
+	return ret;
+}
+
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_mca_smu_debug_mode_set(void *data, u64 val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	int ret;
+
+	ret = amdgpu_ras_set_mca_debug_mode(adev, val ? true : false);
+	if (ret)
+		return ret;
+
+	dev_info(adev->dev, "amdgpu set smu mca debug mode %s success\n", val ? "on" : "off");
+
+	return 0;
+}
+
+static void mca_dump_entry(struct seq_file *m, struct mca_bank_entry *entry)
+{
+	int i, idx = entry->idx;
+	int reg_idx_array[] = {
+		MCA_REG_IDX_STATUS,
+		MCA_REG_IDX_ADDR,
+		MCA_REG_IDX_MISC0,
+		MCA_REG_IDX_IPID,
+		MCA_REG_IDX_SYND,
+	};
+
+	seq_printf(m, "mca entry[%d].type: %s\n", idx, entry->type == AMDGPU_MCA_ERROR_TYPE_UE ? "UE" : "CE");
+	seq_printf(m, "mca entry[%d].ip: %d\n", idx, entry->ip);
+	seq_printf(m, "mca entry[%d].info: socketid:%d aid:%d hwid:0x%03x mcatype:0x%04x\n",
+		   idx, entry->info.socket_id, entry->info.aid, entry->info.hwid, entry->info.mcatype);
+
+	for (i = 0; i < ARRAY_SIZE(reg_idx_array); i++)
+		seq_printf(m, "mca entry[%d].regs[%d]: 0x%016llx\n", idx, reg_idx_array[i], entry->regs[reg_idx_array[i]]);
+}
+
+static int mca_dump_show(struct seq_file *m, enum amdgpu_mca_error_type type)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct mca_bank_node *node;
+	struct mca_bank_set mca_set;
+	struct ras_query_context qctx;
+	int ret;
+
+	amdgpu_mca_bank_set_init(&mca_set);
+
+	qctx.evid.event_id = RAS_EVENT_INVALID_ID;
+	ret = amdgpu_mca_smu_get_mca_set(adev, type, &mca_set, &qctx);
+	if (ret)
+		goto err_free_mca_set;
+
+	seq_printf(m, "amdgpu smu %s valid mca count: %d\n",
+		   type == AMDGPU_MCA_ERROR_TYPE_UE ? "UE" : "CE", mca_set.nr_entries);
+
+	if (!mca_set.nr_entries)
+		goto err_free_mca_set;
+
+	list_for_each_entry(node, &mca_set.list, node)
+		mca_dump_entry(m, &node->entry);
+
+	/* add mca bank to mca bank cache */
+	ret = amdgpu_mca_add_mca_set_to_cache(adev, type, &mca_set);
+
+err_free_mca_set:
+	amdgpu_mca_bank_set_release(&mca_set);
+
+	return ret;
+}
+
+static int mca_dump_ce_show(struct seq_file *m, void *unused)
+{
+	return mca_dump_show(m, AMDGPU_MCA_ERROR_TYPE_CE);
+}
+
+static int mca_dump_ce_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, mca_dump_ce_show, inode->i_private);
+}
+
+static const struct file_operations mca_ce_dump_debug_fops = {
+	.owner = THIS_MODULE,
+	.open = mca_dump_ce_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = single_release,
+};
+
+static int mca_dump_ue_show(struct seq_file *m, void *unused)
+{
+	return mca_dump_show(m, AMDGPU_MCA_ERROR_TYPE_UE);
+}
+
+static int mca_dump_ue_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, mca_dump_ue_show, inode->i_private);
+}
+
+static const struct file_operations mca_ue_dump_debug_fops = {
+	.owner = THIS_MODULE,
+	.open = mca_dump_ue_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = single_release,
+};
+
+DEFINE_DEBUGFS_ATTRIBUTE(mca_debug_mode_fops, NULL, amdgpu_mca_smu_debug_mode_set, "%llu\n");
+#endif
+
+void amdgpu_mca_smu_debugfs_init(struct amdgpu_device *adev, struct dentry *root)
+{
+#if defined(CONFIG_DEBUG_FS)
+	if (!root)
+		return;
+
+	debugfs_create_file("mca_debug_mode", 0200, root, adev, &mca_debug_mode_fops);
+	debugfs_create_file("mca_ue_dump", 0400, root, adev, &mca_ue_dump_debug_fops);
+	debugfs_create_file("mca_ce_dump", 0400, root, adev, &mca_ce_dump_debug_fops);
+#endif
+}
+
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h
index 997a073e2409..e80323ff90c1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h
@@ -21,6 +21,53 @@
 #ifndef __AMDGPU_MCA_H__
 #define __AMDGPU_MCA_H__
 
+#include "amdgpu_ras.h"
+
+#define MCA_MAX_REGS_COUNT	(16)
+
+#define MCA_REG_FIELD(x, h, l)			(((x) & GENMASK_ULL(h, l)) >> l)
+#define MCA_REG__STATUS__VAL(x)			MCA_REG_FIELD(x, 63, 63)
+#define MCA_REG__STATUS__OVERFLOW(x)		MCA_REG_FIELD(x, 62, 62)
+#define MCA_REG__STATUS__UC(x)			MCA_REG_FIELD(x, 61, 61)
+#define MCA_REG__STATUS__EN(x)			MCA_REG_FIELD(x, 60, 60)
+#define MCA_REG__STATUS__MISCV(x)		MCA_REG_FIELD(x, 59, 59)
+#define MCA_REG__STATUS__ADDRV(x)		MCA_REG_FIELD(x, 58, 58)
+#define MCA_REG__STATUS__PCC(x)			MCA_REG_FIELD(x, 57, 57)
+#define MCA_REG__STATUS__ERRCOREIDVAL(x)	MCA_REG_FIELD(x, 56, 56)
+#define MCA_REG__STATUS__TCC(x)			MCA_REG_FIELD(x, 55, 55)
+#define MCA_REG__STATUS__SYNDV(x)		MCA_REG_FIELD(x, 53, 53)
+#define MCA_REG__STATUS__CECC(x)		MCA_REG_FIELD(x, 46, 46)
+#define MCA_REG__STATUS__UECC(x)		MCA_REG_FIELD(x, 45, 45)
+#define MCA_REG__STATUS__DEFERRED(x)		MCA_REG_FIELD(x, 44, 44)
+#define MCA_REG__STATUS__POISON(x)		MCA_REG_FIELD(x, 43, 43)
+#define MCA_REG__STATUS__SCRUB(x)		MCA_REG_FIELD(x, 40, 40)
+#define MCA_REG__STATUS__ERRCOREID(x)		MCA_REG_FIELD(x, 37, 32)
+#define MCA_REG__STATUS__ADDRLSB(x)		MCA_REG_FIELD(x, 29, 24)
+#define MCA_REG__STATUS__ERRORCODEEXT(x)	MCA_REG_FIELD(x, 21, 16)
+#define MCA_REG__STATUS__ERRORCODE(x)		MCA_REG_FIELD(x, 15, 0)
+
+#define MCA_REG__MISC0__ERRCNT(x)		MCA_REG_FIELD(x, 43, 32)
+
+#define MCA_REG__SYND__ERRORINFORMATION(x)	MCA_REG_FIELD(x, 17, 0)
+
+enum amdgpu_mca_ip {
+	AMDGPU_MCA_IP_UNKNOW = -1,
+	AMDGPU_MCA_IP_PSP = 0,
+	AMDGPU_MCA_IP_SDMA,
+	AMDGPU_MCA_IP_GC,
+	AMDGPU_MCA_IP_SMU,
+	AMDGPU_MCA_IP_MP5,
+	AMDGPU_MCA_IP_UMC,
+	AMDGPU_MCA_IP_PCS_XGMI,
+	AMDGPU_MCA_IP_COUNT,
+};
+
+enum amdgpu_mca_error_type {
+	AMDGPU_MCA_ERROR_TYPE_UE = 0,
+	AMDGPU_MCA_ERROR_TYPE_CE,
+	AMDGPU_MCA_ERROR_TYPE_DE,
+};
+
 struct amdgpu_mca_ras_block {
 	struct amdgpu_ras_block_object ras_block;
 };
@@ -30,10 +77,64 @@ struct amdgpu_mca_ras {
 	struct amdgpu_mca_ras_block *ras;
 };
 
+struct mca_bank_set {
+	int nr_entries;
+	struct list_head list;
+};
+
+struct mca_bank_cache {
+	struct mca_bank_set mca_set;
+	struct mutex lock;
+};
+
 struct amdgpu_mca {
 	struct amdgpu_mca_ras mp0;
 	struct amdgpu_mca_ras mp1;
 	struct amdgpu_mca_ras mpio;
+	const struct amdgpu_mca_smu_funcs *mca_funcs;
+	struct mca_bank_cache mca_caches[AMDGPU_MCA_ERROR_TYPE_DE];
+	atomic_t ue_update_flag;
+};
+
+enum mca_reg_idx {
+	MCA_REG_IDX_STATUS		= 1,
+	MCA_REG_IDX_ADDR		= 2,
+	MCA_REG_IDX_MISC0		= 3,
+	MCA_REG_IDX_IPID		= 5,
+	MCA_REG_IDX_SYND		= 6,
+	MCA_REG_IDX_COUNT		= 16,
+};
+
+struct mca_bank_info {
+	int socket_id;
+	int aid;
+	int hwid;
+	int mcatype;
+};
+
+struct mca_bank_entry {
+	int idx;
+	enum amdgpu_mca_error_type type;
+	enum amdgpu_mca_ip ip;
+	struct mca_bank_info info;
+	uint64_t regs[MCA_MAX_REGS_COUNT];
+};
+
+struct mca_bank_node {
+	struct mca_bank_entry entry;
+	struct list_head node;
+};
+
+struct amdgpu_mca_smu_funcs {
+	int max_ue_count;
+	int max_ce_count;
+	int (*mca_set_debug_mode)(struct amdgpu_device *adev, bool enable);
+	int (*mca_parse_mca_error_count)(struct amdgpu_device *adev, enum amdgpu_ras_block blk, enum amdgpu_mca_error_type type,
+					 struct mca_bank_entry *entry, uint32_t *count);
+	int (*mca_get_valid_mca_count)(struct amdgpu_device *adev, enum amdgpu_mca_error_type type,
+				       uint32_t *count);
+	int (*mca_get_mca_entry)(struct amdgpu_device *adev, enum amdgpu_mca_error_type type,
+				 int idx, struct mca_bank_entry *entry);
 };
 
 void amdgpu_mca_query_correctable_error_count(struct amdgpu_device *adev,
@@ -53,4 +154,16 @@ void amdgpu_mca_query_ras_error_count(struct amdgpu_device *adev,
 int amdgpu_mca_mp0_ras_sw_init(struct amdgpu_device *adev);
 int amdgpu_mca_mp1_ras_sw_init(struct amdgpu_device *adev);
 int amdgpu_mca_mpio_ras_sw_init(struct amdgpu_device *adev);
+
+void amdgpu_mca_smu_init_funcs(struct amdgpu_device *adev, const struct amdgpu_mca_smu_funcs *mca_funcs);
+int amdgpu_mca_init(struct amdgpu_device *adev);
+void amdgpu_mca_fini(struct amdgpu_device *adev);
+int amdgpu_mca_reset(struct amdgpu_device *adev);
+int amdgpu_mca_smu_set_debug_mode(struct amdgpu_device *adev, bool enable);
+int amdgpu_mca_smu_get_mca_set_error_count(struct amdgpu_device *adev, enum amdgpu_ras_block blk,
+					   enum amdgpu_mca_error_type type, uint32_t *total);
+void amdgpu_mca_smu_debugfs_init(struct amdgpu_device *adev, struct dentry *root);
+int amdgpu_mca_smu_log_ras_error(struct amdgpu_device *adev, enum amdgpu_ras_block blk, enum amdgpu_mca_error_type type,
+				 struct ras_err_data *err_data, struct ras_query_context *qctx);
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
index f0f00466b59f..9c182ce501af 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
@@ -22,6 +22,7 @@
  */
 
 #include <linux/firmware.h>
+#include <drm/drm_exec.h>
 
 #include "amdgpu_mes.h"
 #include "amdgpu.h"
@@ -38,123 +39,58 @@ int amdgpu_mes_doorbell_process_slice(struct amdgpu_device *adev)
 		       PAGE_SIZE);
 }
 
-int amdgpu_mes_alloc_process_doorbells(struct amdgpu_device *adev,
-				      unsigned int *doorbell_index)
-{
-	int r = ida_simple_get(&adev->mes.doorbell_ida, 2,
-			       adev->mes.max_doorbell_slices,
-			       GFP_KERNEL);
-	if (r > 0)
-		*doorbell_index = r;
-
-	return r;
-}
-
-void amdgpu_mes_free_process_doorbells(struct amdgpu_device *adev,
-				      unsigned int doorbell_index)
-{
-	if (doorbell_index)
-		ida_simple_remove(&adev->mes.doorbell_ida, doorbell_index);
-}
-
-unsigned int amdgpu_mes_get_doorbell_dw_offset_in_bar(
-					struct amdgpu_device *adev,
-					uint32_t doorbell_index,
-					unsigned int doorbell_id)
-{
-	return ((doorbell_index *
-		amdgpu_mes_doorbell_process_slice(adev)) / sizeof(u32) +
-		doorbell_id * 2);
-}
-
-static int amdgpu_mes_queue_doorbell_get(struct amdgpu_device *adev,
-					 struct amdgpu_mes_process *process,
-					 int ip_type, uint64_t *doorbell_index)
+static int amdgpu_mes_doorbell_init(struct amdgpu_device *adev)
 {
-	unsigned int offset, found;
+	int i;
+	struct amdgpu_mes *mes = &adev->mes;
 
-	if (ip_type == AMDGPU_RING_TYPE_SDMA) {
-		offset = adev->doorbell_index.sdma_engine[0];
-		found = find_next_zero_bit(process->doorbell_bitmap,
-					   AMDGPU_MES_MAX_NUM_OF_QUEUES_PER_PROCESS,
-					   offset);
-	} else {
-		found = find_first_zero_bit(process->doorbell_bitmap,
-					    AMDGPU_MES_MAX_NUM_OF_QUEUES_PER_PROCESS);
+	/* Bitmap for dynamic allocation of kernel doorbells */
+	mes->doorbell_bitmap = bitmap_zalloc(PAGE_SIZE / sizeof(u32), GFP_KERNEL);
+	if (!mes->doorbell_bitmap) {
+		dev_err(adev->dev, "Failed to allocate MES doorbell bitmap\n");
+		return -ENOMEM;
 	}
 
-	if (found >= AMDGPU_MES_MAX_NUM_OF_QUEUES_PER_PROCESS) {
-		DRM_WARN("No doorbell available\n");
-		return -ENOSPC;
+	mes->num_mes_dbs = PAGE_SIZE / AMDGPU_ONE_DOORBELL_SIZE;
+	for (i = 0; i < AMDGPU_MES_PRIORITY_NUM_LEVELS; i++) {
+		adev->mes.aggregated_doorbells[i] = mes->db_start_dw_offset + i * 2;
+		set_bit(i, mes->doorbell_bitmap);
 	}
 
-	set_bit(found, process->doorbell_bitmap);
-
-	*doorbell_index = amdgpu_mes_get_doorbell_dw_offset_in_bar(adev,
-				process->doorbell_index, found);
-
 	return 0;
 }
 
-static void amdgpu_mes_queue_doorbell_free(struct amdgpu_device *adev,
-					   struct amdgpu_mes_process *process,
-					   uint32_t doorbell_index)
-{
-	unsigned int old, doorbell_id;
-
-	doorbell_id = doorbell_index -
-		(process->doorbell_index *
-		 amdgpu_mes_doorbell_process_slice(adev)) / sizeof(u32);
-	doorbell_id /= 2;
-
-	old = test_and_clear_bit(doorbell_id, process->doorbell_bitmap);
-	WARN_ON(!old);
-}
-
-static int amdgpu_mes_doorbell_init(struct amdgpu_device *adev)
+static int amdgpu_mes_event_log_init(struct amdgpu_device *adev)
 {
-	size_t doorbell_start_offset;
-	size_t doorbell_aperture_size;
-	size_t doorbell_process_limit;
-	size_t aggregated_doorbell_start;
-	int i;
-
-	aggregated_doorbell_start = (adev->doorbell_index.max_assignment + 1) * sizeof(u32);
-	aggregated_doorbell_start =
-		roundup(aggregated_doorbell_start, PAGE_SIZE);
+	int r;
 
-	doorbell_start_offset = aggregated_doorbell_start + PAGE_SIZE;
-	doorbell_start_offset =
-		roundup(doorbell_start_offset,
-			amdgpu_mes_doorbell_process_slice(adev));
+	if (!amdgpu_mes_log_enable)
+		return 0;
 
-	doorbell_aperture_size = adev->doorbell.size;
-	doorbell_aperture_size =
-			rounddown(doorbell_aperture_size,
-				  amdgpu_mes_doorbell_process_slice(adev));
+	r = amdgpu_bo_create_kernel(adev, adev->mes.event_log_size, PAGE_SIZE,
+				    AMDGPU_GEM_DOMAIN_VRAM,
+				    &adev->mes.event_log_gpu_obj,
+				    &adev->mes.event_log_gpu_addr,
+				    &adev->mes.event_log_cpu_addr);
+	if (r) {
+		dev_warn(adev->dev, "failed to create MES event log buffer (%d)", r);
+		return r;
+	}
 
-	if (doorbell_aperture_size > doorbell_start_offset)
-		doorbell_process_limit =
-			(doorbell_aperture_size - doorbell_start_offset) /
-			amdgpu_mes_doorbell_process_slice(adev);
-	else
-		return -ENOSPC;
+	memset(adev->mes.event_log_cpu_addr, 0, adev->mes.event_log_size);
 
-	adev->mes.doorbell_id_offset = doorbell_start_offset / sizeof(u32);
-	adev->mes.max_doorbell_slices = doorbell_process_limit;
+	return  0;
 
-	/* allocate Qword range for aggregated doorbell */
-	for (i = 0; i < AMDGPU_MES_PRIORITY_NUM_LEVELS; i++)
-		adev->mes.aggregated_doorbells[i] =
-			aggregated_doorbell_start / sizeof(u32) + i * 2;
+}
 
-	DRM_INFO("max_doorbell_slices=%zu\n", doorbell_process_limit);
-	return 0;
+static void amdgpu_mes_doorbell_free(struct amdgpu_device *adev)
+{
+	bitmap_free(adev->mes.doorbell_bitmap);
 }
 
 int amdgpu_mes_init(struct amdgpu_device *adev)
 {
-	int i, r;
+	int i, r, num_pipes;
 
 	adev->mes.adev = adev;
 
@@ -163,80 +99,125 @@ int amdgpu_mes_init(struct amdgpu_device *adev)
 	idr_init(&adev->mes.queue_id_idr);
 	ida_init(&adev->mes.doorbell_ida);
 	spin_lock_init(&adev->mes.queue_id_lock);
-	spin_lock_init(&adev->mes.ring_lock);
 	mutex_init(&adev->mes.mutex_hidden);
 
-	adev->mes.total_max_queue = AMDGPU_FENCE_MES_QUEUE_ID_MASK;
-	adev->mes.vmid_mask_mmhub = 0xffffff00;
-	adev->mes.vmid_mask_gfxhub = 0xffffff00;
+	for (i = 0; i < AMDGPU_MAX_MES_PIPES; i++)
+		spin_lock_init(&adev->mes.ring_lock[i]);
 
-	for (i = 0; i < AMDGPU_MES_MAX_COMPUTE_PIPES; i++) {
-		/* use only 1st MEC pipes */
-		if (i >= 4)
-			continue;
-		adev->mes.compute_hqd_mask[i] = 0xc;
-	}
+	adev->mes.total_max_queue = AMDGPU_FENCE_MES_QUEUE_ID_MASK;
+	adev->mes.vmid_mask_mmhub = 0xFF00;
+	adev->mes.vmid_mask_gfxhub = adev->gfx.disable_kq ? 0xFFFE : 0xFF00;
 
-	for (i = 0; i < AMDGPU_MES_MAX_GFX_PIPES; i++)
-		adev->mes.gfx_hqd_mask[i] = i ? 0 : 0xfffffffe;
+	num_pipes = adev->gfx.me.num_pipe_per_me * adev->gfx.me.num_me;
+	if (num_pipes > AMDGPU_MES_MAX_GFX_PIPES)
+		dev_warn(adev->dev, "more gfx pipes than supported by MES! (%d vs %d)\n",
+			 num_pipes, AMDGPU_MES_MAX_GFX_PIPES);
 
-	for (i = 0; i < AMDGPU_MES_MAX_SDMA_PIPES; i++) {
-		if (adev->ip_versions[SDMA0_HWIP][0] < IP_VERSION(6, 0, 0))
-			adev->mes.sdma_hqd_mask[i] = i ? 0 : 0x3fc;
-		/* zero sdma_hqd_mask for non-existent engine */
-		else if (adev->sdma.num_instances == 1)
-			adev->mes.sdma_hqd_mask[i] = i ? 0 : 0xfc;
+	for (i = 0; i < AMDGPU_MES_MAX_GFX_PIPES; i++) {
+		if (i >= num_pipes)
+			break;
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) >=
+		    IP_VERSION(12, 0, 0))
+			/*
+			 * GFX V12 has only one GFX pipe, but 8 queues in it.
+			 * GFX pipe 0 queue 0 is being used by Kernel queue.
+			 * Set GFX pipe 0 queue 1-7 for MES scheduling
+			 * mask = 1111 1110b
+			 */
+			adev->mes.gfx_hqd_mask[i] = adev->gfx.disable_kq ? 0xFF : 0xFE;
 		else
-			adev->mes.sdma_hqd_mask[i] = 0xfc;
+			/*
+			 * GFX pipe 0 queue 0 is being used by Kernel queue.
+			 * Set GFX pipe 0 queue 1 for MES scheduling
+			 * mask = 10b
+			 */
+			adev->mes.gfx_hqd_mask[i] = adev->gfx.disable_kq ? 0x3 : 0x2;
 	}
 
-	r = amdgpu_device_wb_get(adev, &adev->mes.sch_ctx_offs);
-	if (r) {
-		dev_err(adev->dev,
-			"(%d) ring trail_fence_offs wb alloc failed\n", r);
-		goto error_ids;
+	num_pipes = adev->gfx.mec.num_pipe_per_mec * adev->gfx.mec.num_mec;
+	if (num_pipes > AMDGPU_MES_MAX_COMPUTE_PIPES)
+		dev_warn(adev->dev, "more compute pipes than supported by MES! (%d vs %d)\n",
+			 num_pipes, AMDGPU_MES_MAX_COMPUTE_PIPES);
+
+	for (i = 0; i < AMDGPU_MES_MAX_COMPUTE_PIPES; i++) {
+		if (i >= num_pipes)
+			break;
+		adev->mes.compute_hqd_mask[i] = adev->gfx.disable_kq ? 0xF : 0xC;
 	}
-	adev->mes.sch_ctx_gpu_addr =
-		adev->wb.gpu_addr + (adev->mes.sch_ctx_offs * 4);
-	adev->mes.sch_ctx_ptr =
-		(uint64_t *)&adev->wb.wb[adev->mes.sch_ctx_offs];
 
-	r = amdgpu_device_wb_get(adev, &adev->mes.query_status_fence_offs);
-	if (r) {
-		amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs);
-		dev_err(adev->dev,
-			"(%d) query_status_fence_offs wb alloc failed\n", r);
-		goto error_ids;
+	num_pipes = adev->sdma.num_instances;
+	if (num_pipes > AMDGPU_MES_MAX_SDMA_PIPES)
+		dev_warn(adev->dev, "more SDMA pipes than supported by MES! (%d vs %d)\n",
+			 num_pipes, AMDGPU_MES_MAX_SDMA_PIPES);
+
+	for (i = 0; i < AMDGPU_MES_MAX_SDMA_PIPES; i++) {
+		if (i >= num_pipes)
+			break;
+		adev->mes.sdma_hqd_mask[i] = 0xfc;
 	}
-	adev->mes.query_status_fence_gpu_addr =
-		adev->wb.gpu_addr + (adev->mes.query_status_fence_offs * 4);
-	adev->mes.query_status_fence_ptr =
-		(uint64_t *)&adev->wb.wb[adev->mes.query_status_fence_offs];
 
-	r = amdgpu_device_wb_get(adev, &adev->mes.read_val_offs);
-	if (r) {
-		amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs);
-		amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs);
-		dev_err(adev->dev,
-			"(%d) read_val_offs alloc failed\n", r);
-		goto error_ids;
+	for (i = 0; i < AMDGPU_MAX_MES_PIPES; i++) {
+		r = amdgpu_device_wb_get(adev, &adev->mes.sch_ctx_offs[i]);
+		if (r) {
+			dev_err(adev->dev,
+				"(%d) ring trail_fence_offs wb alloc failed\n",
+				r);
+			goto error;
+		}
+		adev->mes.sch_ctx_gpu_addr[i] =
+			adev->wb.gpu_addr + (adev->mes.sch_ctx_offs[i] * 4);
+		adev->mes.sch_ctx_ptr[i] =
+			(uint64_t *)&adev->wb.wb[adev->mes.sch_ctx_offs[i]];
+
+		r = amdgpu_device_wb_get(adev,
+				 &adev->mes.query_status_fence_offs[i]);
+		if (r) {
+			dev_err(adev->dev,
+			      "(%d) query_status_fence_offs wb alloc failed\n",
+			      r);
+			goto error;
+		}
+		adev->mes.query_status_fence_gpu_addr[i] = adev->wb.gpu_addr +
+			(adev->mes.query_status_fence_offs[i] * 4);
+		adev->mes.query_status_fence_ptr[i] =
+			(uint64_t *)&adev->wb.wb[adev->mes.query_status_fence_offs[i]];
 	}
-	adev->mes.read_val_gpu_addr =
-		adev->wb.gpu_addr + (adev->mes.read_val_offs * 4);
-	adev->mes.read_val_ptr =
-		(uint32_t *)&adev->wb.wb[adev->mes.read_val_offs];
 
 	r = amdgpu_mes_doorbell_init(adev);
 	if (r)
 		goto error;
 
+	r = amdgpu_mes_event_log_init(adev);
+	if (r)
+		goto error_doorbell;
+
+	if (adev->mes.hung_queue_db_array_size) {
+		r = amdgpu_bo_create_kernel(adev,
+					    adev->mes.hung_queue_db_array_size * sizeof(u32),
+					    PAGE_SIZE,
+					    AMDGPU_GEM_DOMAIN_GTT,
+					    &adev->mes.hung_queue_db_array_gpu_obj,
+					    &adev->mes.hung_queue_db_array_gpu_addr,
+					    &adev->mes.hung_queue_db_array_cpu_addr);
+		if (r) {
+			dev_warn(adev->dev, "failed to create MES hung db array buffer (%d)", r);
+			goto error_doorbell;
+		}
+	}
+
 	return 0;
 
+error_doorbell:
+	amdgpu_mes_doorbell_free(adev);
 error:
-	amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs);
-	amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs);
-	amdgpu_device_wb_free(adev, adev->mes.read_val_offs);
-error_ids:
+	for (i = 0; i < AMDGPU_MAX_MES_PIPES; i++) {
+		if (adev->mes.sch_ctx_ptr[i])
+			amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs[i]);
+		if (adev->mes.query_status_fence_ptr[i])
+			amdgpu_device_wb_free(adev,
+				      adev->mes.query_status_fence_offs[i]);
+	}
+
 	idr_destroy(&adev->mes.pasid_idr);
 	idr_destroy(&adev->mes.gang_id_idr);
 	idr_destroy(&adev->mes.queue_id_idr);
@@ -247,9 +228,25 @@ error_ids:
 
 void amdgpu_mes_fini(struct amdgpu_device *adev)
 {
-	amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs);
-	amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs);
-	amdgpu_device_wb_free(adev, adev->mes.read_val_offs);
+	int i;
+
+	amdgpu_bo_free_kernel(&adev->mes.hung_queue_db_array_gpu_obj,
+			      &adev->mes.hung_queue_db_array_gpu_addr,
+			      &adev->mes.hung_queue_db_array_cpu_addr);
+
+	amdgpu_bo_free_kernel(&adev->mes.event_log_gpu_obj,
+			      &adev->mes.event_log_gpu_addr,
+			      &adev->mes.event_log_cpu_addr);
+
+	for (i = 0; i < AMDGPU_MAX_MES_PIPES; i++) {
+		if (adev->mes.sch_ctx_ptr[i])
+			amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs[i]);
+		if (adev->mes.query_status_fence_ptr[i])
+			amdgpu_device_wb_free(adev,
+				      adev->mes.query_status_fence_offs[i]);
+	}
+
+	amdgpu_mes_doorbell_free(adev);
 
 	idr_destroy(&adev->mes.pasid_idr);
 	idr_destroy(&adev->mes.gang_id_idr);
@@ -258,568 +255,183 @@ void amdgpu_mes_fini(struct amdgpu_device *adev)
 	mutex_destroy(&adev->mes.mutex_hidden);
 }
 
-static void amdgpu_mes_queue_free_mqd(struct amdgpu_mes_queue *q)
-{
-	amdgpu_bo_free_kernel(&q->mqd_obj,
-			      &q->mqd_gpu_addr,
-			      &q->mqd_cpu_ptr);
-}
-
-int amdgpu_mes_create_process(struct amdgpu_device *adev, int pasid,
-			      struct amdgpu_vm *vm)
+int amdgpu_mes_suspend(struct amdgpu_device *adev)
 {
-	struct amdgpu_mes_process *process;
+	struct mes_suspend_gang_input input;
 	int r;
 
-	/* allocate the mes process buffer */
-	process = kzalloc(sizeof(struct amdgpu_mes_process), GFP_KERNEL);
-	if (!process) {
-		DRM_ERROR("no more memory to create mes process\n");
-		return -ENOMEM;
-	}
+	if (!amdgpu_mes_suspend_resume_all_supported(adev))
+		return 0;
 
-	process->doorbell_bitmap =
-		kzalloc(DIV_ROUND_UP(AMDGPU_MES_MAX_NUM_OF_QUEUES_PER_PROCESS,
-				     BITS_PER_BYTE), GFP_KERNEL);
-	if (!process->doorbell_bitmap) {
-		DRM_ERROR("failed to allocate doorbell bitmap\n");
-		kfree(process);
-		return -ENOMEM;
-	}
-
-	/* allocate the process context bo and map it */
-	r = amdgpu_bo_create_kernel(adev, AMDGPU_MES_PROC_CTX_SIZE, PAGE_SIZE,
-				    AMDGPU_GEM_DOMAIN_GTT,
-				    &process->proc_ctx_bo,
-				    &process->proc_ctx_gpu_addr,
-				    &process->proc_ctx_cpu_ptr);
-	if (r) {
-		DRM_ERROR("failed to allocate process context bo\n");
-		goto clean_up_memory;
-	}
-	memset(process->proc_ctx_cpu_ptr, 0, AMDGPU_MES_PROC_CTX_SIZE);
+	memset(&input, 0x0, sizeof(struct mes_suspend_gang_input));
+	input.suspend_all_gangs = 1;
 
 	/*
 	 * Avoid taking any other locks under MES lock to avoid circular
 	 * lock dependencies.
 	 */
 	amdgpu_mes_lock(&adev->mes);
-
-	/* add the mes process to idr list */
-	r = idr_alloc(&adev->mes.pasid_idr, process, pasid, pasid + 1,
-		      GFP_KERNEL);
-	if (r < 0) {
-		DRM_ERROR("failed to lock pasid=%d\n", pasid);
-		goto clean_up_ctx;
-	}
-
-	/* allocate the starting doorbell index of the process */
-	r = amdgpu_mes_alloc_process_doorbells(adev, &process->doorbell_index);
-	if (r < 0) {
-		DRM_ERROR("failed to allocate doorbell for process\n");
-		goto clean_up_pasid;
-	}
-
-	DRM_DEBUG("process doorbell index = %d\n", process->doorbell_index);
-
-	INIT_LIST_HEAD(&process->gang_list);
-	process->vm = vm;
-	process->pasid = pasid;
-	process->process_quantum = adev->mes.default_process_quantum;
-	process->pd_gpu_addr = amdgpu_bo_gpu_offset(vm->root.bo);
-
+	r = adev->mes.funcs->suspend_gang(&adev->mes, &input);
 	amdgpu_mes_unlock(&adev->mes);
-	return 0;
+	if (r)
+		dev_err(adev->dev, "failed to suspend all gangs");
 
-clean_up_pasid:
-	idr_remove(&adev->mes.pasid_idr, pasid);
-	amdgpu_mes_unlock(&adev->mes);
-clean_up_ctx:
-	amdgpu_bo_free_kernel(&process->proc_ctx_bo,
-			      &process->proc_ctx_gpu_addr,
-			      &process->proc_ctx_cpu_ptr);
-clean_up_memory:
-	kfree(process->doorbell_bitmap);
-	kfree(process);
 	return r;
 }
 
-void amdgpu_mes_destroy_process(struct amdgpu_device *adev, int pasid)
-{
-	struct amdgpu_mes_process *process;
-	struct amdgpu_mes_gang *gang, *tmp1;
-	struct amdgpu_mes_queue *queue, *tmp2;
-	struct mes_remove_queue_input queue_input;
-	unsigned long flags;
-	int r;
-
-	/*
-	 * Avoid taking any other locks under MES lock to avoid circular
-	 * lock dependencies.
-	 */
-	amdgpu_mes_lock(&adev->mes);
-
-	process = idr_find(&adev->mes.pasid_idr, pasid);
-	if (!process) {
-		DRM_WARN("pasid %d doesn't exist\n", pasid);
-		amdgpu_mes_unlock(&adev->mes);
-		return;
-	}
-
-	/* Remove all queues from hardware */
-	list_for_each_entry_safe(gang, tmp1, &process->gang_list, list) {
-		list_for_each_entry_safe(queue, tmp2, &gang->queue_list, list) {
-			spin_lock_irqsave(&adev->mes.queue_id_lock, flags);
-			idr_remove(&adev->mes.queue_id_idr, queue->queue_id);
-			spin_unlock_irqrestore(&adev->mes.queue_id_lock, flags);
-
-			queue_input.doorbell_offset = queue->doorbell_off;
-			queue_input.gang_context_addr = gang->gang_ctx_gpu_addr;
-
-			r = adev->mes.funcs->remove_hw_queue(&adev->mes,
-							     &queue_input);
-			if (r)
-				DRM_WARN("failed to remove hardware queue\n");
-		}
-
-		idr_remove(&adev->mes.gang_id_idr, gang->gang_id);
-	}
-
-	amdgpu_mes_free_process_doorbells(adev, process->doorbell_index);
-	idr_remove(&adev->mes.pasid_idr, pasid);
-	amdgpu_mes_unlock(&adev->mes);
-
-	/* free all memory allocated by the process */
-	list_for_each_entry_safe(gang, tmp1, &process->gang_list, list) {
-		/* free all queues in the gang */
-		list_for_each_entry_safe(queue, tmp2, &gang->queue_list, list) {
-			amdgpu_mes_queue_free_mqd(queue);
-			list_del(&queue->list);
-			kfree(queue);
-		}
-		amdgpu_bo_free_kernel(&gang->gang_ctx_bo,
-				      &gang->gang_ctx_gpu_addr,
-				      &gang->gang_ctx_cpu_ptr);
-		list_del(&gang->list);
-		kfree(gang);
-
-	}
-	amdgpu_bo_free_kernel(&process->proc_ctx_bo,
-			      &process->proc_ctx_gpu_addr,
-			      &process->proc_ctx_cpu_ptr);
-	kfree(process->doorbell_bitmap);
-	kfree(process);
-}
-
-int amdgpu_mes_add_gang(struct amdgpu_device *adev, int pasid,
-			struct amdgpu_mes_gang_properties *gprops,
-			int *gang_id)
+int amdgpu_mes_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_mes_process *process;
-	struct amdgpu_mes_gang *gang;
+	struct mes_resume_gang_input input;
 	int r;
 
-	/* allocate the mes gang buffer */
-	gang = kzalloc(sizeof(struct amdgpu_mes_gang), GFP_KERNEL);
-	if (!gang) {
-		return -ENOMEM;
-	}
+	if (!amdgpu_mes_suspend_resume_all_supported(adev))
+		return 0;
 
-	/* allocate the gang context bo and map it to cpu space */
-	r = amdgpu_bo_create_kernel(adev, AMDGPU_MES_GANG_CTX_SIZE, PAGE_SIZE,
-				    AMDGPU_GEM_DOMAIN_GTT,
-				    &gang->gang_ctx_bo,
-				    &gang->gang_ctx_gpu_addr,
-				    &gang->gang_ctx_cpu_ptr);
-	if (r) {
-		DRM_ERROR("failed to allocate process context bo\n");
-		goto clean_up_mem;
-	}
-	memset(gang->gang_ctx_cpu_ptr, 0, AMDGPU_MES_GANG_CTX_SIZE);
+	memset(&input, 0x0, sizeof(struct mes_resume_gang_input));
+	input.resume_all_gangs = 1;
 
 	/*
 	 * Avoid taking any other locks under MES lock to avoid circular
 	 * lock dependencies.
 	 */
 	amdgpu_mes_lock(&adev->mes);
-
-	process = idr_find(&adev->mes.pasid_idr, pasid);
-	if (!process) {
-		DRM_ERROR("pasid %d doesn't exist\n", pasid);
-		r = -EINVAL;
-		goto clean_up_ctx;
-	}
-
-	/* add the mes gang to idr list */
-	r = idr_alloc(&adev->mes.gang_id_idr, gang, 1, 0,
-		      GFP_KERNEL);
-	if (r < 0) {
-		DRM_ERROR("failed to allocate idr for gang\n");
-		goto clean_up_ctx;
-	}
-
-	gang->gang_id = r;
-	*gang_id = r;
-
-	INIT_LIST_HEAD(&gang->queue_list);
-	gang->process = process;
-	gang->priority = gprops->priority;
-	gang->gang_quantum = gprops->gang_quantum ?
-		gprops->gang_quantum : adev->mes.default_gang_quantum;
-	gang->global_priority_level = gprops->global_priority_level;
-	gang->inprocess_gang_priority = gprops->inprocess_gang_priority;
-	list_add_tail(&gang->list, &process->gang_list);
-
+	r = adev->mes.funcs->resume_gang(&adev->mes, &input);
 	amdgpu_mes_unlock(&adev->mes);
-	return 0;
+	if (r)
+		dev_err(adev->dev, "failed to resume all gangs");
 
-clean_up_ctx:
-	amdgpu_mes_unlock(&adev->mes);
-	amdgpu_bo_free_kernel(&gang->gang_ctx_bo,
-			      &gang->gang_ctx_gpu_addr,
-			      &gang->gang_ctx_cpu_ptr);
-clean_up_mem:
-	kfree(gang);
 	return r;
 }
 
-int amdgpu_mes_remove_gang(struct amdgpu_device *adev, int gang_id)
+int amdgpu_mes_map_legacy_queue(struct amdgpu_device *adev,
+				struct amdgpu_ring *ring)
 {
-	struct amdgpu_mes_gang *gang;
-
-	/*
-	 * Avoid taking any other locks under MES lock to avoid circular
-	 * lock dependencies.
-	 */
-	amdgpu_mes_lock(&adev->mes);
-
-	gang = idr_find(&adev->mes.gang_id_idr, gang_id);
-	if (!gang) {
-		DRM_ERROR("gang id %d doesn't exist\n", gang_id);
-		amdgpu_mes_unlock(&adev->mes);
-		return -EINVAL;
-	}
-
-	if (!list_empty(&gang->queue_list)) {
-		DRM_ERROR("queue list is not empty\n");
-		amdgpu_mes_unlock(&adev->mes);
-		return -EBUSY;
-	}
-
-	idr_remove(&adev->mes.gang_id_idr, gang->gang_id);
-	list_del(&gang->list);
-	amdgpu_mes_unlock(&adev->mes);
-
-	amdgpu_bo_free_kernel(&gang->gang_ctx_bo,
-			      &gang->gang_ctx_gpu_addr,
-			      &gang->gang_ctx_cpu_ptr);
-
-	kfree(gang);
+	struct mes_map_legacy_queue_input queue_input;
+	int r;
 
-	return 0;
-}
+	memset(&queue_input, 0, sizeof(queue_input));
 
-int amdgpu_mes_suspend(struct amdgpu_device *adev)
-{
-	struct idr *idp;
-	struct amdgpu_mes_process *process;
-	struct amdgpu_mes_gang *gang;
-	struct mes_suspend_gang_input input;
-	int r, pasid;
+	queue_input.queue_type = ring->funcs->type;
+	queue_input.doorbell_offset = ring->doorbell_index;
+	queue_input.pipe_id = ring->pipe;
+	queue_input.queue_id = ring->queue;
+	queue_input.mqd_addr = amdgpu_bo_gpu_offset(ring->mqd_obj);
+	queue_input.wptr_addr = ring->wptr_gpu_addr;
 
-	/*
-	 * Avoid taking any other locks under MES lock to avoid circular
-	 * lock dependencies.
-	 */
 	amdgpu_mes_lock(&adev->mes);
-
-	idp = &adev->mes.pasid_idr;
-
-	idr_for_each_entry(idp, process, pasid) {
-		list_for_each_entry(gang, &process->gang_list, list) {
-			r = adev->mes.funcs->suspend_gang(&adev->mes, &input);
-			if (r)
-				DRM_ERROR("failed to suspend pasid %d gangid %d",
-					 pasid, gang->gang_id);
-		}
-	}
-
+	r = adev->mes.funcs->map_legacy_queue(&adev->mes, &queue_input);
 	amdgpu_mes_unlock(&adev->mes);
-	return 0;
-}
-
-int amdgpu_mes_resume(struct amdgpu_device *adev)
-{
-	struct idr *idp;
-	struct amdgpu_mes_process *process;
-	struct amdgpu_mes_gang *gang;
-	struct mes_resume_gang_input input;
-	int r, pasid;
-
-	/*
-	 * Avoid taking any other locks under MES lock to avoid circular
-	 * lock dependencies.
-	 */
-	amdgpu_mes_lock(&adev->mes);
-
-	idp = &adev->mes.pasid_idr;
-
-	idr_for_each_entry(idp, process, pasid) {
-		list_for_each_entry(gang, &process->gang_list, list) {
-			r = adev->mes.funcs->resume_gang(&adev->mes, &input);
-			if (r)
-				DRM_ERROR("failed to resume pasid %d gangid %d",
-					 pasid, gang->gang_id);
-		}
-	}
+	if (r)
+		dev_err(adev->dev, "failed to map legacy queue\n");
 
-	amdgpu_mes_unlock(&adev->mes);
-	return 0;
+	return r;
 }
 
-static int amdgpu_mes_queue_alloc_mqd(struct amdgpu_device *adev,
-				     struct amdgpu_mes_queue *q,
-				     struct amdgpu_mes_queue_properties *p)
+int amdgpu_mes_unmap_legacy_queue(struct amdgpu_device *adev,
+				  struct amdgpu_ring *ring,
+				  enum amdgpu_unmap_queues_action action,
+				  u64 gpu_addr, u64 seq)
 {
-	struct amdgpu_mqd *mqd_mgr = &adev->mqds[p->queue_type];
-	u32 mqd_size = mqd_mgr->mqd_size;
+	struct mes_unmap_legacy_queue_input queue_input;
 	int r;
 
-	r = amdgpu_bo_create_kernel(adev, mqd_size, PAGE_SIZE,
-				    AMDGPU_GEM_DOMAIN_GTT,
-				    &q->mqd_obj,
-				    &q->mqd_gpu_addr, &q->mqd_cpu_ptr);
-	if (r) {
-		dev_warn(adev->dev, "failed to create queue mqd bo (%d)", r);
-		return r;
-	}
-	memset(q->mqd_cpu_ptr, 0, mqd_size);
-
-	r = amdgpu_bo_reserve(q->mqd_obj, false);
-	if (unlikely(r != 0))
-		goto clean_up;
+	queue_input.action = action;
+	queue_input.queue_type = ring->funcs->type;
+	queue_input.doorbell_offset = ring->doorbell_index;
+	queue_input.pipe_id = ring->pipe;
+	queue_input.queue_id = ring->queue;
+	queue_input.trail_fence_addr = gpu_addr;
+	queue_input.trail_fence_data = seq;
 
-	return 0;
+	amdgpu_mes_lock(&adev->mes);
+	r = adev->mes.funcs->unmap_legacy_queue(&adev->mes, &queue_input);
+	amdgpu_mes_unlock(&adev->mes);
+	if (r)
+		dev_err(adev->dev, "failed to unmap legacy queue\n");
 
-clean_up:
-	amdgpu_bo_free_kernel(&q->mqd_obj,
-			      &q->mqd_gpu_addr,
-			      &q->mqd_cpu_ptr);
 	return r;
 }
 
-static void amdgpu_mes_queue_init_mqd(struct amdgpu_device *adev,
-				     struct amdgpu_mes_queue *q,
-				     struct amdgpu_mes_queue_properties *p)
-{
-	struct amdgpu_mqd *mqd_mgr = &adev->mqds[p->queue_type];
-	struct amdgpu_mqd_prop mqd_prop = {0};
-
-	mqd_prop.mqd_gpu_addr = q->mqd_gpu_addr;
-	mqd_prop.hqd_base_gpu_addr = p->hqd_base_gpu_addr;
-	mqd_prop.rptr_gpu_addr = p->rptr_gpu_addr;
-	mqd_prop.wptr_gpu_addr = p->wptr_gpu_addr;
-	mqd_prop.queue_size = p->queue_size;
-	mqd_prop.use_doorbell = true;
-	mqd_prop.doorbell_index = p->doorbell_off;
-	mqd_prop.eop_gpu_addr = p->eop_gpu_addr;
-	mqd_prop.hqd_pipe_priority = p->hqd_pipe_priority;
-	mqd_prop.hqd_queue_priority = p->hqd_queue_priority;
-	mqd_prop.hqd_active = false;
-
-	mqd_mgr->init_mqd(adev, q->mqd_cpu_ptr, &mqd_prop);
-
-	amdgpu_bo_unreserve(q->mqd_obj);
-}
-
-int amdgpu_mes_add_hw_queue(struct amdgpu_device *adev, int gang_id,
-			    struct amdgpu_mes_queue_properties *qprops,
-			    int *queue_id)
+int amdgpu_mes_reset_legacy_queue(struct amdgpu_device *adev,
+				  struct amdgpu_ring *ring,
+				  unsigned int vmid,
+				  bool use_mmio)
 {
-	struct amdgpu_mes_queue *queue;
-	struct amdgpu_mes_gang *gang;
-	struct mes_add_queue_input queue_input;
-	unsigned long flags;
+	struct mes_reset_queue_input queue_input;
 	int r;
 
-	/* allocate the mes queue buffer */
-	queue = kzalloc(sizeof(struct amdgpu_mes_queue), GFP_KERNEL);
-	if (!queue) {
-		DRM_ERROR("Failed to allocate memory for queue\n");
-		return -ENOMEM;
-	}
+	memset(&queue_input, 0, sizeof(queue_input));
 
-	/* Allocate the queue mqd */
-	r = amdgpu_mes_queue_alloc_mqd(adev, queue, qprops);
-	if (r)
-		goto clean_up_memory;
+	queue_input.queue_type = ring->funcs->type;
+	queue_input.doorbell_offset = ring->doorbell_index;
+	queue_input.me_id = ring->me;
+	queue_input.pipe_id = ring->pipe;
+	queue_input.queue_id = ring->queue;
+	queue_input.mqd_addr = ring->mqd_obj ? amdgpu_bo_gpu_offset(ring->mqd_obj) : 0;
+	queue_input.wptr_addr = ring->wptr_gpu_addr;
+	queue_input.vmid = vmid;
+	queue_input.use_mmio = use_mmio;
+	queue_input.is_kq = true;
+	if (ring->funcs->type == AMDGPU_RING_TYPE_GFX)
+		queue_input.legacy_gfx = true;
 
-	/*
-	 * Avoid taking any other locks under MES lock to avoid circular
-	 * lock dependencies.
-	 */
 	amdgpu_mes_lock(&adev->mes);
-
-	gang = idr_find(&adev->mes.gang_id_idr, gang_id);
-	if (!gang) {
-		DRM_ERROR("gang id %d doesn't exist\n", gang_id);
-		r = -EINVAL;
-		goto clean_up_mqd;
-	}
-
-	/* add the mes gang to idr list */
-	spin_lock_irqsave(&adev->mes.queue_id_lock, flags);
-	r = idr_alloc(&adev->mes.queue_id_idr, queue, 1, 0,
-		      GFP_ATOMIC);
-	if (r < 0) {
-		spin_unlock_irqrestore(&adev->mes.queue_id_lock, flags);
-		goto clean_up_mqd;
-	}
-	spin_unlock_irqrestore(&adev->mes.queue_id_lock, flags);
-	*queue_id = queue->queue_id = r;
-
-	/* allocate a doorbell index for the queue */
-	r = amdgpu_mes_queue_doorbell_get(adev, gang->process,
-					  qprops->queue_type,
-					  &qprops->doorbell_off);
-	if (r)
-		goto clean_up_queue_id;
-
-	/* initialize the queue mqd */
-	amdgpu_mes_queue_init_mqd(adev, queue, qprops);
-
-	/* add hw queue to mes */
-	queue_input.process_id = gang->process->pasid;
-
-	queue_input.page_table_base_addr =
-		adev->vm_manager.vram_base_offset + gang->process->pd_gpu_addr -
-		adev->gmc.vram_start;
-
-	queue_input.process_va_start = 0;
-	queue_input.process_va_end =
-		(adev->vm_manager.max_pfn - 1) << AMDGPU_GPU_PAGE_SHIFT;
-	queue_input.process_quantum = gang->process->process_quantum;
-	queue_input.process_context_addr = gang->process->proc_ctx_gpu_addr;
-	queue_input.gang_quantum = gang->gang_quantum;
-	queue_input.gang_context_addr = gang->gang_ctx_gpu_addr;
-	queue_input.inprocess_gang_priority = gang->inprocess_gang_priority;
-	queue_input.gang_global_priority_level = gang->global_priority_level;
-	queue_input.doorbell_offset = qprops->doorbell_off;
-	queue_input.mqd_addr = queue->mqd_gpu_addr;
-	queue_input.wptr_addr = qprops->wptr_gpu_addr;
-	queue_input.wptr_mc_addr = qprops->wptr_mc_addr;
-	queue_input.queue_type = qprops->queue_type;
-	queue_input.paging = qprops->paging;
-	queue_input.is_kfd_process = 0;
-
-	r = adev->mes.funcs->add_hw_queue(&adev->mes, &queue_input);
-	if (r) {
-		DRM_ERROR("failed to add hardware queue to MES, doorbell=0x%llx\n",
-			  qprops->doorbell_off);
-		goto clean_up_doorbell;
-	}
-
-	DRM_DEBUG("MES hw queue was added, pasid=%d, gang id=%d, "
-		  "queue type=%d, doorbell=0x%llx\n",
-		  gang->process->pasid, gang_id, qprops->queue_type,
-		  qprops->doorbell_off);
-
-	queue->ring = qprops->ring;
-	queue->doorbell_off = qprops->doorbell_off;
-	queue->wptr_gpu_addr = qprops->wptr_gpu_addr;
-	queue->queue_type = qprops->queue_type;
-	queue->paging = qprops->paging;
-	queue->gang = gang;
-	queue->ring->mqd_ptr = queue->mqd_cpu_ptr;
-	list_add_tail(&queue->list, &gang->queue_list);
-
+	r = adev->mes.funcs->reset_hw_queue(&adev->mes, &queue_input);
 	amdgpu_mes_unlock(&adev->mes);
-	return 0;
+	if (r)
+		dev_err(adev->dev, "failed to reset legacy queue\n");
 
-clean_up_doorbell:
-	amdgpu_mes_queue_doorbell_free(adev, gang->process,
-				       qprops->doorbell_off);
-clean_up_queue_id:
-	spin_lock_irqsave(&adev->mes.queue_id_lock, flags);
-	idr_remove(&adev->mes.queue_id_idr, queue->queue_id);
-	spin_unlock_irqrestore(&adev->mes.queue_id_lock, flags);
-clean_up_mqd:
-	amdgpu_mes_unlock(&adev->mes);
-	amdgpu_mes_queue_free_mqd(queue);
-clean_up_memory:
-	kfree(queue);
 	return r;
 }
 
-int amdgpu_mes_remove_hw_queue(struct amdgpu_device *adev, int queue_id)
+int amdgpu_mes_get_hung_queue_db_array_size(struct amdgpu_device *adev)
 {
-	unsigned long flags;
-	struct amdgpu_mes_queue *queue;
-	struct amdgpu_mes_gang *gang;
-	struct mes_remove_queue_input queue_input;
-	int r;
+	return adev->mes.hung_queue_db_array_size;
+}
 
-	/*
-	 * Avoid taking any other locks under MES lock to avoid circular
-	 * lock dependencies.
-	 */
-	amdgpu_mes_lock(&adev->mes);
+int amdgpu_mes_detect_and_reset_hung_queues(struct amdgpu_device *adev,
+					    int queue_type,
+					    bool detect_only,
+					    unsigned int *hung_db_num,
+					    u32 *hung_db_array)
 
-	/* remove the mes gang from idr list */
-	spin_lock_irqsave(&adev->mes.queue_id_lock, flags);
+{
+	struct mes_detect_and_reset_queue_input input;
+	u32 *db_array = adev->mes.hung_queue_db_array_cpu_addr;
+	int r, i;
 
-	queue = idr_find(&adev->mes.queue_id_idr, queue_id);
-	if (!queue) {
-		spin_unlock_irqrestore(&adev->mes.queue_id_lock, flags);
-		amdgpu_mes_unlock(&adev->mes);
-		DRM_ERROR("queue id %d doesn't exist\n", queue_id);
+	if (!hung_db_num || !hung_db_array)
 		return -EINVAL;
-	}
-
-	idr_remove(&adev->mes.queue_id_idr, queue_id);
-	spin_unlock_irqrestore(&adev->mes.queue_id_lock, flags);
-
-	DRM_DEBUG("try to remove queue, doorbell off = 0x%llx\n",
-		  queue->doorbell_off);
-
-	gang = queue->gang;
-	queue_input.doorbell_offset = queue->doorbell_off;
-	queue_input.gang_context_addr = gang->gang_ctx_gpu_addr;
 
-	r = adev->mes.funcs->remove_hw_queue(&adev->mes, &queue_input);
-	if (r)
-		DRM_ERROR("failed to remove hardware queue, queue id = %d\n",
-			  queue_id);
-
-	list_del(&queue->list);
-	amdgpu_mes_queue_doorbell_free(adev, gang->process,
-				       queue->doorbell_off);
-	amdgpu_mes_unlock(&adev->mes);
+	if ((queue_type != AMDGPU_RING_TYPE_GFX) &&
+	    (queue_type != AMDGPU_RING_TYPE_COMPUTE) &&
+	    (queue_type != AMDGPU_RING_TYPE_SDMA))
+		return -EINVAL;
 
-	amdgpu_mes_queue_free_mqd(queue);
-	kfree(queue);
-	return 0;
-}
+	/* Clear the doorbell array before detection */
+	memset(adev->mes.hung_queue_db_array_cpu_addr, AMDGPU_MES_INVALID_DB_OFFSET,
+		adev->mes.hung_queue_db_array_size * sizeof(u32));
+	input.queue_type = queue_type;
+	input.detect_only = detect_only;
 
-int amdgpu_mes_unmap_legacy_queue(struct amdgpu_device *adev,
-				  struct amdgpu_ring *ring,
-				  enum amdgpu_unmap_queues_action action,
-				  u64 gpu_addr, u64 seq)
-{
-	struct mes_unmap_legacy_queue_input queue_input;
-	int r;
-
-	queue_input.action = action;
-	queue_input.queue_type = ring->funcs->type;
-	queue_input.doorbell_offset = ring->doorbell_index;
-	queue_input.pipe_id = ring->pipe;
-	queue_input.queue_id = ring->queue;
-	queue_input.trail_fence_addr = gpu_addr;
-	queue_input.trail_fence_data = seq;
+	r = adev->mes.funcs->detect_and_reset_hung_queues(&adev->mes,
+							  &input);
+	if (r) {
+		dev_err(adev->dev, "failed to detect and reset\n");
+	} else {
+		*hung_db_num = 0;
+		for (i = 0; i < adev->mes.hung_queue_hqd_info_offset; i++) {
+			if (db_array[i] != AMDGPU_MES_INVALID_DB_OFFSET) {
+				hung_db_array[i] = db_array[i];
+				*hung_db_num += 1;
+			}
+		}
 
-	r = adev->mes.funcs->unmap_legacy_queue(&adev->mes, &queue_input);
-	if (r)
-		DRM_ERROR("failed to unmap legacy queue\n");
+		/*
+		 * TODO: return HQD info for MES scheduled user compute queue reset cases
+		 * stored in hung_db_array hqd info offset to full array size
+		 */
+	}
 
 	return r;
 }
@@ -828,23 +440,36 @@ uint32_t amdgpu_mes_rreg(struct amdgpu_device *adev, uint32_t reg)
 {
 	struct mes_misc_op_input op_input;
 	int r, val = 0;
+	uint32_t addr_offset = 0;
+	uint64_t read_val_gpu_addr;
+	uint32_t *read_val_ptr;
 
+	if (amdgpu_device_wb_get(adev, &addr_offset)) {
+		dev_err(adev->dev, "critical bug! too many mes readers\n");
+		goto error;
+	}
+	read_val_gpu_addr = adev->wb.gpu_addr + (addr_offset * 4);
+	read_val_ptr = (uint32_t *)&adev->wb.wb[addr_offset];
 	op_input.op = MES_MISC_OP_READ_REG;
 	op_input.read_reg.reg_offset = reg;
-	op_input.read_reg.buffer_addr = adev->mes.read_val_gpu_addr;
+	op_input.read_reg.buffer_addr = read_val_gpu_addr;
 
 	if (!adev->mes.funcs->misc_op) {
-		DRM_ERROR("mes rreg is not supported!\n");
+		dev_err(adev->dev, "mes rreg is not supported!\n");
 		goto error;
 	}
 
+	amdgpu_mes_lock(&adev->mes);
 	r = adev->mes.funcs->misc_op(&adev->mes, &op_input);
+	amdgpu_mes_unlock(&adev->mes);
 	if (r)
-		DRM_ERROR("failed to read reg (0x%x)\n", reg);
+		dev_err(adev->dev, "failed to read reg (0x%x)\n", reg);
 	else
-		val = *(adev->mes.read_val_ptr);
+		val = *(read_val_ptr);
 
 error:
+	if (addr_offset)
+		amdgpu_device_wb_free(adev, addr_offset);
 	return val;
 }
 
@@ -859,14 +484,16 @@ int amdgpu_mes_wreg(struct amdgpu_device *adev,
 	op_input.write_reg.reg_value = val;
 
 	if (!adev->mes.funcs->misc_op) {
-		DRM_ERROR("mes wreg is not supported!\n");
+		dev_err(adev->dev, "mes wreg is not supported!\n");
 		r = -EINVAL;
 		goto error;
 	}
 
+	amdgpu_mes_lock(&adev->mes);
 	r = adev->mes.funcs->misc_op(&adev->mes, &op_input);
+	amdgpu_mes_unlock(&adev->mes);
 	if (r)
-		DRM_ERROR("failed to write reg (0x%x)\n", reg);
+		dev_err(adev->dev, "failed to write reg (0x%x)\n", reg);
 
 error:
 	return r;
@@ -886,546 +513,107 @@ int amdgpu_mes_reg_write_reg_wait(struct amdgpu_device *adev,
 	op_input.wrm_reg.mask = mask;
 
 	if (!adev->mes.funcs->misc_op) {
-		DRM_ERROR("mes reg_write_reg_wait is not supported!\n");
+		dev_err(adev->dev, "mes reg_write_reg_wait is not supported!\n");
 		r = -EINVAL;
 		goto error;
 	}
 
+	amdgpu_mes_lock(&adev->mes);
 	r = adev->mes.funcs->misc_op(&adev->mes, &op_input);
+	amdgpu_mes_unlock(&adev->mes);
 	if (r)
-		DRM_ERROR("failed to reg_write_reg_wait\n");
+		dev_err(adev->dev, "failed to reg_write_reg_wait\n");
 
 error:
 	return r;
 }
 
-int amdgpu_mes_reg_wait(struct amdgpu_device *adev, uint32_t reg,
-			uint32_t val, uint32_t mask)
+int amdgpu_mes_hdp_flush(struct amdgpu_device *adev)
 {
-	struct mes_misc_op_input op_input;
-	int r;
-
-	op_input.op = MES_MISC_OP_WRM_REG_WAIT;
-	op_input.wrm_reg.reg0 = reg;
-	op_input.wrm_reg.ref = val;
-	op_input.wrm_reg.mask = mask;
+	uint32_t hdp_flush_req_offset, hdp_flush_done_offset, ref_and_mask;
 
-	if (!adev->mes.funcs->misc_op) {
-		DRM_ERROR("mes reg wait is not supported!\n");
-		r = -EINVAL;
-		goto error;
-	}
+	hdp_flush_req_offset = adev->nbio.funcs->get_hdp_flush_req_offset(adev);
+	hdp_flush_done_offset = adev->nbio.funcs->get_hdp_flush_done_offset(adev);
+	ref_and_mask = adev->nbio.hdp_flush_reg->ref_and_mask_cp0;
 
-	r = adev->mes.funcs->misc_op(&adev->mes, &op_input);
-	if (r)
-		DRM_ERROR("failed to reg_write_reg_wait\n");
-
-error:
-	return r;
+	return amdgpu_mes_reg_write_reg_wait(adev, hdp_flush_req_offset, hdp_flush_done_offset,
+					     ref_and_mask, ref_and_mask);
 }
 
-static void
-amdgpu_mes_ring_to_queue_props(struct amdgpu_device *adev,
-			       struct amdgpu_ring *ring,
-			       struct amdgpu_mes_queue_properties *props)
+int amdgpu_mes_set_shader_debugger(struct amdgpu_device *adev,
+				uint64_t process_context_addr,
+				uint32_t spi_gdbg_per_vmid_cntl,
+				const uint32_t *tcp_watch_cntl,
+				uint32_t flags,
+				bool trap_en)
 {
-	props->queue_type = ring->funcs->type;
-	props->hqd_base_gpu_addr = ring->gpu_addr;
-	props->rptr_gpu_addr = ring->rptr_gpu_addr;
-	props->wptr_gpu_addr = ring->wptr_gpu_addr;
-	props->wptr_mc_addr =
-		ring->mes_ctx->meta_data_mc_addr + ring->wptr_offs;
-	props->queue_size = ring->ring_size;
-	props->eop_gpu_addr = ring->eop_gpu_addr;
-	props->hqd_pipe_priority = AMDGPU_GFX_PIPE_PRIO_NORMAL;
-	props->hqd_queue_priority = AMDGPU_GFX_QUEUE_PRIORITY_MINIMUM;
-	props->paging = false;
-	props->ring = ring;
-}
+	struct mes_misc_op_input op_input = {0};
+	int r;
 
-#define DEFINE_AMDGPU_MES_CTX_GET_OFFS_ENG(_eng)			\
-do {									\
-       if (id_offs < AMDGPU_MES_CTX_MAX_OFFS)				\
-		return offsetof(struct amdgpu_mes_ctx_meta_data,	\
-				_eng[ring->idx].slots[id_offs]);        \
-       else if (id_offs == AMDGPU_MES_CTX_RING_OFFS)			\
-		return offsetof(struct amdgpu_mes_ctx_meta_data,        \
-				_eng[ring->idx].ring);                  \
-       else if (id_offs == AMDGPU_MES_CTX_IB_OFFS)			\
-		return offsetof(struct amdgpu_mes_ctx_meta_data,        \
-				_eng[ring->idx].ib);                    \
-       else if (id_offs == AMDGPU_MES_CTX_PADDING_OFFS)			\
-		return offsetof(struct amdgpu_mes_ctx_meta_data,        \
-				_eng[ring->idx].padding);               \
-} while(0)
-
-int amdgpu_mes_ctx_get_offs(struct amdgpu_ring *ring, unsigned int id_offs)
-{
-	switch (ring->funcs->type) {
-	case AMDGPU_RING_TYPE_GFX:
-		DEFINE_AMDGPU_MES_CTX_GET_OFFS_ENG(gfx);
-		break;
-	case AMDGPU_RING_TYPE_COMPUTE:
-		DEFINE_AMDGPU_MES_CTX_GET_OFFS_ENG(compute);
-		break;
-	case AMDGPU_RING_TYPE_SDMA:
-		DEFINE_AMDGPU_MES_CTX_GET_OFFS_ENG(sdma);
-		break;
-	default:
-		break;
+	if (!adev->mes.funcs->misc_op) {
+		dev_err(adev->dev,
+			"mes set shader debugger is not supported!\n");
+		return -EINVAL;
 	}
 
-	WARN_ON(1);
-	return -EINVAL;
-}
+	op_input.op = MES_MISC_OP_SET_SHADER_DEBUGGER;
+	op_input.set_shader_debugger.process_context_addr = process_context_addr;
+	op_input.set_shader_debugger.flags.u32all = flags;
 
-int amdgpu_mes_add_ring(struct amdgpu_device *adev, int gang_id,
-			int queue_type, int idx,
-			struct amdgpu_mes_ctx_data *ctx_data,
-			struct amdgpu_ring **out)
-{
-	struct amdgpu_ring *ring;
-	struct amdgpu_mes_gang *gang;
-	struct amdgpu_mes_queue_properties qprops = {0};
-	int r, queue_id, pasid;
-
-	/*
-	 * Avoid taking any other locks under MES lock to avoid circular
-	 * lock dependencies.
-	 */
-	amdgpu_mes_lock(&adev->mes);
-	gang = idr_find(&adev->mes.gang_id_idr, gang_id);
-	if (!gang) {
-		DRM_ERROR("gang id %d doesn't exist\n", gang_id);
-		amdgpu_mes_unlock(&adev->mes);
+	/* use amdgpu mes_flush_shader_debugger instead */
+	if (op_input.set_shader_debugger.flags.process_ctx_flush)
 		return -EINVAL;
-	}
-	pasid = gang->process->pasid;
 
-	ring = kzalloc(sizeof(struct amdgpu_ring), GFP_KERNEL);
-	if (!ring) {
-		amdgpu_mes_unlock(&adev->mes);
-		return -ENOMEM;
-	}
+	op_input.set_shader_debugger.spi_gdbg_per_vmid_cntl = spi_gdbg_per_vmid_cntl;
+	memcpy(op_input.set_shader_debugger.tcp_watch_cntl, tcp_watch_cntl,
+			sizeof(op_input.set_shader_debugger.tcp_watch_cntl));
 
-	ring->ring_obj = NULL;
-	ring->use_doorbell = true;
-	ring->is_mes_queue = true;
-	ring->mes_ctx = ctx_data;
-	ring->idx = idx;
-	ring->no_scheduler = true;
-
-	if (queue_type == AMDGPU_RING_TYPE_COMPUTE) {
-		int offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				      compute[ring->idx].mec_hpd);
-		ring->eop_gpu_addr =
-			amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-	}
+	if (((adev->mes.sched_version & AMDGPU_MES_API_VERSION_MASK) >>
+			AMDGPU_MES_API_VERSION_SHIFT) >= 14)
+		op_input.set_shader_debugger.trap_en = trap_en;
 
-	switch (queue_type) {
-	case AMDGPU_RING_TYPE_GFX:
-		ring->funcs = adev->gfx.gfx_ring[0].funcs;
-		break;
-	case AMDGPU_RING_TYPE_COMPUTE:
-		ring->funcs = adev->gfx.compute_ring[0].funcs;
-		break;
-	case AMDGPU_RING_TYPE_SDMA:
-		ring->funcs = adev->sdma.instance[0].ring.funcs;
-		break;
-	default:
-		BUG();
-	}
+	amdgpu_mes_lock(&adev->mes);
 
-	r = amdgpu_ring_init(adev, ring, 1024, NULL, 0,
-			     AMDGPU_RING_PRIO_DEFAULT, NULL);
+	r = adev->mes.funcs->misc_op(&adev->mes, &op_input);
 	if (r)
-		goto clean_up_memory;
-
-	amdgpu_mes_ring_to_queue_props(adev, ring, &qprops);
+		dev_err(adev->dev, "failed to set_shader_debugger\n");
 
-	dma_fence_wait(gang->process->vm->last_update, false);
-	dma_fence_wait(ctx_data->meta_data_va->last_pt_update, false);
 	amdgpu_mes_unlock(&adev->mes);
 
-	r = amdgpu_mes_add_hw_queue(adev, gang_id, &qprops, &queue_id);
-	if (r)
-		goto clean_up_ring;
-
-	ring->hw_queue_id = queue_id;
-	ring->doorbell_index = qprops.doorbell_off;
-
-	if (queue_type == AMDGPU_RING_TYPE_GFX)
-		sprintf(ring->name, "gfx_%d.%d.%d", pasid, gang_id, queue_id);
-	else if (queue_type == AMDGPU_RING_TYPE_COMPUTE)
-		sprintf(ring->name, "compute_%d.%d.%d", pasid, gang_id,
-			queue_id);
-	else if (queue_type == AMDGPU_RING_TYPE_SDMA)
-		sprintf(ring->name, "sdma_%d.%d.%d", pasid, gang_id,
-			queue_id);
-	else
-		BUG();
-
-	*out = ring;
-	return 0;
-
-clean_up_ring:
-	amdgpu_ring_fini(ring);
-clean_up_memory:
-	kfree(ring);
-	amdgpu_mes_unlock(&adev->mes);
 	return r;
 }
 
-void amdgpu_mes_remove_ring(struct amdgpu_device *adev,
-			    struct amdgpu_ring *ring)
-{
-	if (!ring)
-		return;
-
-	amdgpu_mes_remove_hw_queue(adev, ring->hw_queue_id);
-	amdgpu_ring_fini(ring);
-	kfree(ring);
-}
-
-uint32_t amdgpu_mes_get_aggregated_doorbell_index(struct amdgpu_device *adev,
-						   enum amdgpu_mes_priority_level prio)
-{
-	return adev->mes.aggregated_doorbells[prio];
-}
-
-int amdgpu_mes_ctx_alloc_meta_data(struct amdgpu_device *adev,
-				   struct amdgpu_mes_ctx_data *ctx_data)
+int amdgpu_mes_flush_shader_debugger(struct amdgpu_device *adev,
+				     uint64_t process_context_addr)
 {
+	struct mes_misc_op_input op_input = {0};
 	int r;
 
-	r = amdgpu_bo_create_kernel(adev,
-			    sizeof(struct amdgpu_mes_ctx_meta_data),
-			    PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT,
-			    &ctx_data->meta_data_obj,
-			    &ctx_data->meta_data_mc_addr,
-			    &ctx_data->meta_data_ptr);
-	if (r) {
-		dev_warn(adev->dev, "(%d) create CTX bo failed\n", r);
-		return r;
-	}
-
-	if (!ctx_data->meta_data_obj)
-		return -ENOMEM;
-
-	memset(ctx_data->meta_data_ptr, 0,
-	       sizeof(struct amdgpu_mes_ctx_meta_data));
-
-	return 0;
-}
-
-void amdgpu_mes_ctx_free_meta_data(struct amdgpu_mes_ctx_data *ctx_data)
-{
-	if (ctx_data->meta_data_obj)
-		amdgpu_bo_free_kernel(&ctx_data->meta_data_obj,
-				      &ctx_data->meta_data_mc_addr,
-				      &ctx_data->meta_data_ptr);
-}
-
-int amdgpu_mes_ctx_map_meta_data(struct amdgpu_device *adev,
-				 struct amdgpu_vm *vm,
-				 struct amdgpu_mes_ctx_data *ctx_data)
-{
-	struct amdgpu_bo_va *bo_va;
-	struct ww_acquire_ctx ticket;
-	struct list_head list;
-	struct amdgpu_bo_list_entry pd;
-	struct ttm_validate_buffer csa_tv;
-	struct amdgpu_sync sync;
-	int r;
-
-	amdgpu_sync_create(&sync);
-	INIT_LIST_HEAD(&list);
-	INIT_LIST_HEAD(&csa_tv.head);
-
-	csa_tv.bo = &ctx_data->meta_data_obj->tbo;
-	csa_tv.num_shared = 1;
-
-	list_add(&csa_tv.head, &list);
-	amdgpu_vm_get_pd_bo(vm, &list, &pd);
-
-	r = ttm_eu_reserve_buffers(&ticket, &list, true, NULL);
-	if (r) {
-		DRM_ERROR("failed to reserve meta data BO: err=%d\n", r);
-		return r;
-	}
-
-	bo_va = amdgpu_vm_bo_add(adev, vm, ctx_data->meta_data_obj);
-	if (!bo_va) {
-		ttm_eu_backoff_reservation(&ticket, &list);
-		DRM_ERROR("failed to create bo_va for meta data BO\n");
-		return -ENOMEM;
-	}
-
-	r = amdgpu_vm_bo_map(adev, bo_va, ctx_data->meta_data_gpu_addr, 0,
-			     sizeof(struct amdgpu_mes_ctx_meta_data),
-			     AMDGPU_PTE_READABLE | AMDGPU_PTE_WRITEABLE |
-			     AMDGPU_PTE_EXECUTABLE);
-
-	if (r) {
-		DRM_ERROR("failed to do bo_map on meta data, err=%d\n", r);
-		goto error;
-	}
-
-	r = amdgpu_vm_bo_update(adev, bo_va, false);
-	if (r) {
-		DRM_ERROR("failed to do vm_bo_update on meta data\n");
-		goto error;
-	}
-	amdgpu_sync_fence(&sync, bo_va->last_pt_update);
-
-	r = amdgpu_vm_update_pdes(adev, vm, false);
-	if (r) {
-		DRM_ERROR("failed to update pdes on meta data\n");
-		goto error;
+	if (!adev->mes.funcs->misc_op) {
+		dev_err(adev->dev,
+			"mes flush shader debugger is not supported!\n");
+		return -EINVAL;
 	}
-	amdgpu_sync_fence(&sync, vm->last_update);
-
-	amdgpu_sync_wait(&sync, false);
-	ttm_eu_backoff_reservation(&ticket, &list);
 
-	amdgpu_sync_free(&sync);
-	ctx_data->meta_data_va = bo_va;
-	return 0;
-
-error:
-	amdgpu_vm_bo_del(adev, bo_va);
-	ttm_eu_backoff_reservation(&ticket, &list);
-	amdgpu_sync_free(&sync);
-	return r;
-}
+	op_input.op = MES_MISC_OP_SET_SHADER_DEBUGGER;
+	op_input.set_shader_debugger.process_context_addr = process_context_addr;
+	op_input.set_shader_debugger.flags.process_ctx_flush = true;
 
-int amdgpu_mes_ctx_unmap_meta_data(struct amdgpu_device *adev,
-				   struct amdgpu_mes_ctx_data *ctx_data)
-{
-	struct amdgpu_bo_va *bo_va = ctx_data->meta_data_va;
-	struct amdgpu_bo *bo = ctx_data->meta_data_obj;
-	struct amdgpu_vm *vm = bo_va->base.vm;
-	struct amdgpu_bo_list_entry vm_pd;
-	struct list_head list, duplicates;
-	struct dma_fence *fence = NULL;
-	struct ttm_validate_buffer tv;
-	struct ww_acquire_ctx ticket;
-	long r = 0;
-
-	INIT_LIST_HEAD(&list);
-	INIT_LIST_HEAD(&duplicates);
-
-	tv.bo = &bo->tbo;
-	tv.num_shared = 2;
-	list_add(&tv.head, &list);
-
-	amdgpu_vm_get_pd_bo(vm, &list, &vm_pd);
-
-	r = ttm_eu_reserve_buffers(&ticket, &list, false, &duplicates);
-	if (r) {
-		dev_err(adev->dev, "leaking bo va because "
-			"we fail to reserve bo (%ld)\n", r);
-		return r;
-	}
-
-	amdgpu_vm_bo_del(adev, bo_va);
-	if (!amdgpu_vm_ready(vm))
-		goto out_unlock;
+	amdgpu_mes_lock(&adev->mes);
 
-	r = dma_resv_get_singleton(bo->tbo.base.resv, DMA_RESV_USAGE_BOOKKEEP, &fence);
+	r = adev->mes.funcs->misc_op(&adev->mes, &op_input);
 	if (r)
-		goto out_unlock;
-	if (fence) {
-		amdgpu_bo_fence(bo, fence, true);
-		fence = NULL;
-	}
-
-	r = amdgpu_vm_clear_freed(adev, vm, &fence);
-	if (r || !fence)
-		goto out_unlock;
+		dev_err(adev->dev, "failed to set_shader_debugger\n");
 
-	dma_fence_wait(fence, false);
-	amdgpu_bo_fence(bo, fence, true);
-	dma_fence_put(fence);
-
-out_unlock:
-	if (unlikely(r < 0))
-		dev_err(adev->dev, "failed to clear page tables (%ld)\n", r);
-	ttm_eu_backoff_reservation(&ticket, &list);
+	amdgpu_mes_unlock(&adev->mes);
 
 	return r;
 }
 
-static int amdgpu_mes_test_create_gang_and_queues(struct amdgpu_device *adev,
-					  int pasid, int *gang_id,
-					  int queue_type, int num_queue,
-					  struct amdgpu_ring **added_rings,
-					  struct amdgpu_mes_ctx_data *ctx_data)
-{
-	struct amdgpu_ring *ring;
-	struct amdgpu_mes_gang_properties gprops = {0};
-	int r, j;
-
-	/* create a gang for the process */
-	gprops.priority = AMDGPU_MES_PRIORITY_LEVEL_NORMAL;
-	gprops.gang_quantum = adev->mes.default_gang_quantum;
-	gprops.inprocess_gang_priority = AMDGPU_MES_PRIORITY_LEVEL_NORMAL;
-	gprops.priority_level = AMDGPU_MES_PRIORITY_LEVEL_NORMAL;
-	gprops.global_priority_level = AMDGPU_MES_PRIORITY_LEVEL_NORMAL;
-
-	r = amdgpu_mes_add_gang(adev, pasid, &gprops, gang_id);
-	if (r) {
-		DRM_ERROR("failed to add gang\n");
-		return r;
-	}
-
-	/* create queues for the gang */
-	for (j = 0; j < num_queue; j++) {
-		r = amdgpu_mes_add_ring(adev, *gang_id, queue_type, j,
-					ctx_data, &ring);
-		if (r) {
-			DRM_ERROR("failed to add ring\n");
-			break;
-		}
-
-		DRM_INFO("ring %s was added\n", ring->name);
-		added_rings[j] = ring;
-	}
-
-	return 0;
-}
-
-static int amdgpu_mes_test_queues(struct amdgpu_ring **added_rings)
-{
-	struct amdgpu_ring *ring;
-	int i, r;
-
-	for (i = 0; i < AMDGPU_MES_CTX_MAX_RINGS; i++) {
-		ring = added_rings[i];
-		if (!ring)
-			continue;
-
-		r = amdgpu_ring_test_ring(ring);
-		if (r) {
-			DRM_DEV_ERROR(ring->adev->dev,
-				      "ring %s test failed (%d)\n",
-				      ring->name, r);
-			return r;
-		} else
-			DRM_INFO("ring %s test pass\n", ring->name);
-
-		r = amdgpu_ring_test_ib(ring, 1000 * 10);
-		if (r) {
-			DRM_DEV_ERROR(ring->adev->dev,
-				      "ring %s ib test failed (%d)\n",
-				      ring->name, r);
-			return r;
-		} else
-			DRM_INFO("ring %s ib test pass\n", ring->name);
-	}
-
-	return 0;
-}
-
-int amdgpu_mes_self_test(struct amdgpu_device *adev)
+uint32_t amdgpu_mes_get_aggregated_doorbell_index(struct amdgpu_device *adev,
+						   enum amdgpu_mes_priority_level prio)
 {
-	struct amdgpu_vm *vm = NULL;
-	struct amdgpu_mes_ctx_data ctx_data = {0};
-	struct amdgpu_ring *added_rings[AMDGPU_MES_CTX_MAX_RINGS] = { NULL };
-	int gang_ids[3] = {0};
-	int queue_types[][2] = { { AMDGPU_RING_TYPE_GFX, 1 },
-				 { AMDGPU_RING_TYPE_COMPUTE, 1 },
-				 { AMDGPU_RING_TYPE_SDMA, 1} };
-	int i, r, pasid, k = 0;
-
-	pasid = amdgpu_pasid_alloc(16);
-	if (pasid < 0) {
-		dev_warn(adev->dev, "No more PASIDs available!");
-		pasid = 0;
-	}
-
-	vm = kzalloc(sizeof(*vm), GFP_KERNEL);
-	if (!vm) {
-		r = -ENOMEM;
-		goto error_pasid;
-	}
-
-	r = amdgpu_vm_init(adev, vm);
-	if (r) {
-		DRM_ERROR("failed to initialize vm\n");
-		goto error_pasid;
-	}
-
-	r = amdgpu_mes_ctx_alloc_meta_data(adev, &ctx_data);
-	if (r) {
-		DRM_ERROR("failed to alloc ctx meta data\n");
-		goto error_fini;
-	}
-
-	ctx_data.meta_data_gpu_addr = AMDGPU_VA_RESERVED_SIZE;
-	r = amdgpu_mes_ctx_map_meta_data(adev, vm, &ctx_data);
-	if (r) {
-		DRM_ERROR("failed to map ctx meta data\n");
-		goto error_vm;
-	}
-
-	r = amdgpu_mes_create_process(adev, pasid, vm);
-	if (r) {
-		DRM_ERROR("failed to create MES process\n");
-		goto error_vm;
-	}
-
-	for (i = 0; i < ARRAY_SIZE(queue_types); i++) {
-		/* On GFX v10.3, fw hasn't supported to map sdma queue. */
-		if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0) &&
-		    adev->ip_versions[GC_HWIP][0] < IP_VERSION(11, 0, 0) &&
-		    queue_types[i][0] == AMDGPU_RING_TYPE_SDMA)
-			continue;
-
-		r = amdgpu_mes_test_create_gang_and_queues(adev, pasid,
-							   &gang_ids[i],
-							   queue_types[i][0],
-							   queue_types[i][1],
-							   &added_rings[k],
-							   &ctx_data);
-		if (r)
-			goto error_queues;
-
-		k += queue_types[i][1];
-	}
-
-	/* start ring test and ib test for MES queues */
-	amdgpu_mes_test_queues(added_rings);
-
-error_queues:
-	/* remove all queues */
-	for (i = 0; i < ARRAY_SIZE(added_rings); i++) {
-		if (!added_rings[i])
-			continue;
-		amdgpu_mes_remove_ring(adev, added_rings[i]);
-	}
-
-	for (i = 0; i < ARRAY_SIZE(gang_ids); i++) {
-		if (!gang_ids[i])
-			continue;
-		amdgpu_mes_remove_gang(adev, gang_ids[i]);
-	}
-
-	amdgpu_mes_destroy_process(adev, pasid);
-
-error_vm:
-	amdgpu_mes_ctx_unmap_meta_data(adev, &ctx_data);
-
-error_fini:
-	amdgpu_vm_fini(adev, vm);
-
-error_pasid:
-	if (pasid)
-		amdgpu_pasid_free(pasid);
-
-	amdgpu_mes_ctx_free_meta_data(&ctx_data);
-	kfree(vm);
-	return 0;
+	return adev->mes.aggregated_doorbells[prio];
 }
 
 int amdgpu_mes_init_microcode(struct amdgpu_device *adev, int pipe)
@@ -1433,13 +621,18 @@ int amdgpu_mes_init_microcode(struct amdgpu_device *adev, int pipe)
 	const struct mes_firmware_header_v1_0 *mes_hdr;
 	struct amdgpu_firmware_info *info;
 	char ucode_prefix[30];
-	char fw_name[40];
+	char fw_name[50];
 	bool need_retry = false;
+	u32 *ucode_ptr;
 	int r;
 
 	amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix,
 				       sizeof(ucode_prefix));
-	if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(11, 0, 0)) {
+	if (adev->enable_uni_mes) {
+		snprintf(fw_name, sizeof(fw_name),
+			 "amdgpu/%s_uni_mes.bin", ucode_prefix);
+	} else if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(11, 0, 0) &&
+	    amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(12, 0, 0)) {
 		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mes%s.bin",
 			 ucode_prefix,
 			 pipe == AMDGPU_MES_SCHED_PIPE ? "_2" : "1");
@@ -1450,13 +643,13 @@ int amdgpu_mes_init_microcode(struct amdgpu_device *adev, int pipe)
 			 pipe == AMDGPU_MES_SCHED_PIPE ? "" : "1");
 	}
 
-	r = amdgpu_ucode_request(adev, &adev->mes.fw[pipe], fw_name);
+	r = amdgpu_ucode_request(adev, &adev->mes.fw[pipe], AMDGPU_UCODE_REQUIRED,
+				 "%s", fw_name);
 	if (r && need_retry && pipe == AMDGPU_MES_SCHED_PIPE) {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mes.bin",
-			 ucode_prefix);
-		DRM_INFO("try to fall back to %s\n", fw_name);
+		dev_info(adev->dev, "try to fall back to %s_mes.bin\n", ucode_prefix);
 		r = amdgpu_ucode_request(adev, &adev->mes.fw[pipe],
-					 fw_name);
+					 AMDGPU_UCODE_REQUIRED,
+					 "amdgpu/%s_mes.bin", ucode_prefix);
 	}
 
 	if (r)
@@ -1470,6 +663,10 @@ int amdgpu_mes_init_microcode(struct amdgpu_device *adev, int pipe)
 	adev->mes.data_start_addr[pipe] =
 		le32_to_cpu(mes_hdr->mes_data_start_addr_lo) |
 		((uint64_t)(le32_to_cpu(mes_hdr->mes_data_start_addr_hi)) << 32);
+	ucode_ptr = (u32 *)(adev->mes.fw[pipe]->data +
+			  sizeof(union amdgpu_firmware_header));
+	adev->mes.fw_version[pipe] =
+		le32_to_cpu(ucode_ptr[24]) & AMDGPU_MES_VERSION_MASK;
 
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 		int ucode, ucode_data;
@@ -1502,3 +699,86 @@ out:
 	amdgpu_ucode_release(&adev->mes.fw[pipe]);
 	return r;
 }
+
+bool amdgpu_mes_suspend_resume_all_supported(struct amdgpu_device *adev)
+{
+	uint32_t mes_rev = adev->mes.sched_version & AMDGPU_MES_VERSION_MASK;
+
+	return ((amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(11, 0, 0) &&
+		 amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(12, 0, 0) &&
+		 mes_rev >= 0x63) ||
+		amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(12, 0, 0));
+}
+
+/* Fix me -- node_id is used to identify the correct MES instances in the future */
+static int amdgpu_mes_set_enforce_isolation(struct amdgpu_device *adev,
+					    uint32_t node_id, bool enable)
+{
+	struct mes_misc_op_input op_input = {0};
+	int r;
+
+	op_input.op = MES_MISC_OP_CHANGE_CONFIG;
+	op_input.change_config.option.limit_single_process = enable ? 1 : 0;
+
+	if (!adev->mes.funcs->misc_op) {
+		dev_err(adev->dev, "mes change config is not supported!\n");
+		r = -EINVAL;
+		goto error;
+	}
+
+	amdgpu_mes_lock(&adev->mes);
+	r = adev->mes.funcs->misc_op(&adev->mes, &op_input);
+	amdgpu_mes_unlock(&adev->mes);
+	if (r)
+		dev_err(adev->dev, "failed to change_config.\n");
+
+error:
+	return r;
+}
+
+int amdgpu_mes_update_enforce_isolation(struct amdgpu_device *adev)
+{
+	int i, r = 0;
+
+	if (adev->enable_mes && adev->gfx.enable_cleaner_shader) {
+		mutex_lock(&adev->enforce_isolation_mutex);
+		for (i = 0; i < (adev->xcp_mgr ? adev->xcp_mgr->num_xcps : 1); i++) {
+			if (adev->enforce_isolation[i] == AMDGPU_ENFORCE_ISOLATION_ENABLE)
+				r |= amdgpu_mes_set_enforce_isolation(adev, i, true);
+			else
+				r |= amdgpu_mes_set_enforce_isolation(adev, i, false);
+		}
+		mutex_unlock(&adev->enforce_isolation_mutex);
+	}
+	return r;
+}
+
+#if defined(CONFIG_DEBUG_FS)
+
+static int amdgpu_debugfs_mes_event_log_show(struct seq_file *m, void *unused)
+{
+	struct amdgpu_device *adev = m->private;
+	uint32_t *mem = (uint32_t *)(adev->mes.event_log_cpu_addr);
+
+	seq_hex_dump(m, "", DUMP_PREFIX_OFFSET, 32, 4,
+		     mem, adev->mes.event_log_size, false);
+
+	return 0;
+}
+
+DEFINE_SHOW_ATTRIBUTE(amdgpu_debugfs_mes_event_log);
+
+#endif
+
+void amdgpu_debugfs_mes_event_log_init(struct amdgpu_device *adev)
+{
+
+#if defined(CONFIG_DEBUG_FS)
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+	struct dentry *root = minor->debugfs_root;
+	if (adev->enable_mes && amdgpu_mes_log_enable)
+		debugfs_create_file("amdgpu_mes_event_log", 0444, root,
+				    adev, &amdgpu_debugfs_mes_event_log_fops);
+
+#endif
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
index 547ec35691fa..e989225b354b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
@@ -27,6 +27,7 @@
 #include "amdgpu_irq.h"
 #include "kgd_kfd_interface.h"
 #include "amdgpu_gfx.h"
+#include "amdgpu_doorbell.h"
 #include <linux/sched/mm.h>
 
 #define AMDGPU_MES_MAX_COMPUTE_PIPES        8
@@ -39,6 +40,8 @@
 #define AMDGPU_MES_VERSION_MASK		0x00000fff
 #define AMDGPU_MES_API_VERSION_MASK	0x00fff000
 #define AMDGPU_MES_FEAT_VERSION_MASK	0xff000000
+#define AMDGPU_MES_MSCRATCH_SIZE	0x40000
+#define AMDGPU_MES_INVALID_DB_OFFSET	0xffffffff
 
 enum amdgpu_mes_priority_level {
 	AMDGPU_MES_PRIORITY_LEVEL_LOW       = 0,
@@ -54,7 +57,7 @@ enum amdgpu_mes_priority_level {
 
 struct amdgpu_mes_funcs;
 
-enum admgpu_mes_pipe {
+enum amdgpu_mes_pipe {
 	AMDGPU_MES_SCHED_PIPE = 0,
 	AMDGPU_MES_KIQ_PIPE,
 	AMDGPU_MAX_MES_PIPES = 2,
@@ -74,16 +77,17 @@ struct amdgpu_mes {
 
 	uint32_t			sched_version;
 	uint32_t			kiq_version;
+	uint32_t			fw_version[AMDGPU_MAX_MES_PIPES];
+	bool                            enable_legacy_queue_map;
 
 	uint32_t                        total_max_queue;
-	uint32_t                        doorbell_id_offset;
 	uint32_t                        max_doorbell_slices;
 
 	uint64_t                        default_process_quantum;
 	uint64_t                        default_gang_quantum;
 
-	struct amdgpu_ring              ring;
-	spinlock_t                      ring_lock;
+	struct amdgpu_ring              ring[AMDGPU_MAX_MES_PIPES];
+	spinlock_t                      ring_lock[AMDGPU_MAX_MES_PIPES];
 
 	const struct firmware           *fw[AMDGPU_MAX_MES_PIPES];
 
@@ -108,19 +112,16 @@ struct amdgpu_mes {
 
 	uint32_t                        vmid_mask_gfxhub;
 	uint32_t                        vmid_mask_mmhub;
-	uint32_t                        compute_hqd_mask[AMDGPU_MES_MAX_COMPUTE_PIPES];
 	uint32_t                        gfx_hqd_mask[AMDGPU_MES_MAX_GFX_PIPES];
+	uint32_t                        compute_hqd_mask[AMDGPU_MES_MAX_COMPUTE_PIPES];
 	uint32_t                        sdma_hqd_mask[AMDGPU_MES_MAX_SDMA_PIPES];
 	uint32_t                        aggregated_doorbells[AMDGPU_MES_PRIORITY_NUM_LEVELS];
-	uint32_t                        sch_ctx_offs;
-	uint64_t			sch_ctx_gpu_addr;
-	uint64_t			*sch_ctx_ptr;
-	uint32_t			query_status_fence_offs;
-	uint64_t			query_status_fence_gpu_addr;
-	uint64_t			*query_status_fence_ptr;
-	uint32_t                        read_val_offs;
-	uint64_t			read_val_gpu_addr;
-	uint32_t			*read_val_ptr;
+	uint32_t                        sch_ctx_offs[AMDGPU_MAX_MES_PIPES];
+	uint64_t			sch_ctx_gpu_addr[AMDGPU_MAX_MES_PIPES];
+	uint64_t			*sch_ctx_ptr[AMDGPU_MAX_MES_PIPES];
+	uint32_t			query_status_fence_offs[AMDGPU_MAX_MES_PIPES];
+	uint64_t			query_status_fence_gpu_addr[AMDGPU_MAX_MES_PIPES];
+	uint64_t			*query_status_fence_ptr[AMDGPU_MAX_MES_PIPES];
 
 	uint32_t			saved_flags;
 
@@ -128,22 +129,30 @@ struct amdgpu_mes {
 	int                             (*kiq_hw_init)(struct amdgpu_device *adev);
 	int                             (*kiq_hw_fini)(struct amdgpu_device *adev);
 
+	/* MES doorbells */
+	uint32_t			db_start_dw_offset;
+	uint32_t			num_mes_dbs;
+	unsigned long			*doorbell_bitmap;
+
+	/* MES event log buffer */
+	uint32_t			event_log_size;
+	struct amdgpu_bo	*event_log_gpu_obj;
+	uint64_t			event_log_gpu_addr;
+	void				*event_log_cpu_addr;
+
 	/* ip specific functions */
 	const struct amdgpu_mes_funcs   *funcs;
-};
 
-struct amdgpu_mes_process {
-	int			pasid;
-	struct			amdgpu_vm *vm;
-	uint64_t		pd_gpu_addr;
-	struct amdgpu_bo 	*proc_ctx_bo;
-	uint64_t 		proc_ctx_gpu_addr;
-	void 			*proc_ctx_cpu_ptr;
-	uint64_t 		process_quantum;
-	struct 			list_head gang_list;
-	uint32_t 		doorbell_index;
-	unsigned long 		*doorbell_bitmap;
-	struct mutex		doorbell_lock;
+	/* mes resource_1 bo*/
+	struct amdgpu_bo    *resource_1[AMDGPU_MAX_MES_PIPES];
+	uint64_t            resource_1_gpu_addr[AMDGPU_MAX_MES_PIPES];
+	void                *resource_1_addr[AMDGPU_MAX_MES_PIPES];
+
+	int				hung_queue_db_array_size;
+	int				hung_queue_hqd_info_offset;
+	struct amdgpu_bo		*hung_queue_db_array_gpu_obj;
+	uint64_t			hung_queue_db_array_gpu_addr;
+	void				*hung_queue_db_array_cpu_addr;
 };
 
 struct amdgpu_mes_gang {
@@ -219,14 +228,27 @@ struct mes_add_queue_input {
 	uint32_t        gws_size;
 	uint64_t	tba_addr;
 	uint64_t	tma_addr;
+	uint32_t	trap_en;
+	uint32_t	skip_process_ctx_clear;
 	uint32_t	is_kfd_process;
 	uint32_t	is_aql_queue;
 	uint32_t	queue_size;
+	uint32_t	exclusively_scheduled;
 };
 
 struct mes_remove_queue_input {
 	uint32_t	doorbell_offset;
 	uint64_t	gang_context_addr;
+	bool		remove_queue_after_reset;
+};
+
+struct mes_map_legacy_queue_input {
+	uint32_t                           queue_type;
+	uint32_t                           doorbell_offset;
+	uint32_t                           pipe_id;
+	uint32_t                           queue_id;
+	uint64_t                           mqd_addr;
+	uint64_t                           wptr_addr;
 };
 
 struct mes_unmap_legacy_queue_input {
@@ -251,11 +273,39 @@ struct mes_resume_gang_input {
 	uint64_t	gang_context_addr;
 };
 
+struct mes_reset_queue_input {
+	uint32_t                           queue_type;
+	uint32_t                           doorbell_offset;
+	bool                               use_mmio;
+	uint32_t                           me_id;
+	uint32_t                           pipe_id;
+	uint32_t                           queue_id;
+	uint64_t                           mqd_addr;
+	uint64_t                           wptr_addr;
+	uint32_t                           vmid;
+	bool                               legacy_gfx;
+	bool                               is_kq;
+};
+
+struct mes_detect_and_reset_queue_input {
+	uint32_t                           queue_type;
+	bool                               detect_only;
+};
+
+struct mes_inv_tlbs_pasid_input {
+	uint32_t        xcc_id;
+	uint16_t        pasid;
+	uint8_t         hub_id;
+	uint8_t         flush_type;
+};
+
 enum mes_misc_opcode {
 	MES_MISC_OP_WRITE_REG,
 	MES_MISC_OP_READ_REG,
 	MES_MISC_OP_WRM_REG_WAIT,
 	MES_MISC_OP_WRM_REG_WR_WAIT,
+	MES_MISC_OP_SET_SHADER_DEBUGGER,
+	MES_MISC_OP_CHANGE_CONFIG,
 };
 
 struct mes_misc_op_input {
@@ -278,6 +328,37 @@ struct mes_misc_op_input {
 			uint32_t                   reg0;
 			uint32_t                   reg1;
 		} wrm_reg;
+
+		struct {
+			uint64_t process_context_addr;
+			union {
+				struct {
+					uint32_t single_memop : 1;
+					uint32_t single_alu_op : 1;
+					uint32_t reserved: 29;
+					uint32_t process_ctx_flush: 1;
+				};
+				uint32_t u32all;
+			} flags;
+			uint32_t spi_gdbg_per_vmid_cntl;
+			uint32_t tcp_watch_cntl[4];
+			uint32_t trap_en;
+		} set_shader_debugger;
+
+		struct {
+			union {
+				struct {
+					uint32_t limit_single_process : 1;
+					uint32_t enable_hws_logging_buffer : 1;
+					uint32_t reserved : 30;
+				};
+				uint32_t all;
+			} option;
+			struct {
+				uint32_t tdr_level;
+				uint32_t tdr_delay;
+			} tdr_config;
+		} change_config;
 	};
 };
 
@@ -288,6 +369,9 @@ struct amdgpu_mes_funcs {
 	int (*remove_hw_queue)(struct amdgpu_mes *mes,
 			       struct mes_remove_queue_input *input);
 
+	int (*map_legacy_queue)(struct amdgpu_mes *mes,
+				struct mes_map_legacy_queue_input *input);
+
 	int (*unmap_legacy_queue)(struct amdgpu_mes *mes,
 				  struct mes_unmap_legacy_queue_input *input);
 
@@ -299,77 +383,65 @@ struct amdgpu_mes_funcs {
 
 	int (*misc_op)(struct amdgpu_mes *mes,
 		       struct mes_misc_op_input *input);
+
+	int (*reset_hw_queue)(struct amdgpu_mes *mes,
+			      struct mes_reset_queue_input *input);
+
+	int (*detect_and_reset_hung_queues)(struct amdgpu_mes *mes,
+			      struct mes_detect_and_reset_queue_input *input);
+
+
+	int (*invalidate_tlbs_pasid)(struct amdgpu_mes *mes,
+			      struct mes_inv_tlbs_pasid_input *input);
 };
 
 #define amdgpu_mes_kiq_hw_init(adev) (adev)->mes.kiq_hw_init((adev))
 #define amdgpu_mes_kiq_hw_fini(adev) (adev)->mes.kiq_hw_fini((adev))
 
-int amdgpu_mes_ctx_get_offs(struct amdgpu_ring *ring, unsigned int id_offs);
-
 int amdgpu_mes_init_microcode(struct amdgpu_device *adev, int pipe);
 int amdgpu_mes_init(struct amdgpu_device *adev);
 void amdgpu_mes_fini(struct amdgpu_device *adev);
 
-int amdgpu_mes_create_process(struct amdgpu_device *adev, int pasid,
-			      struct amdgpu_vm *vm);
-void amdgpu_mes_destroy_process(struct amdgpu_device *adev, int pasid);
-
-int amdgpu_mes_add_gang(struct amdgpu_device *adev, int pasid,
-			struct amdgpu_mes_gang_properties *gprops,
-			int *gang_id);
-int amdgpu_mes_remove_gang(struct amdgpu_device *adev, int gang_id);
-
 int amdgpu_mes_suspend(struct amdgpu_device *adev);
 int amdgpu_mes_resume(struct amdgpu_device *adev);
 
-int amdgpu_mes_add_hw_queue(struct amdgpu_device *adev, int gang_id,
-			    struct amdgpu_mes_queue_properties *qprops,
-			    int *queue_id);
-int amdgpu_mes_remove_hw_queue(struct amdgpu_device *adev, int queue_id);
-
+int amdgpu_mes_map_legacy_queue(struct amdgpu_device *adev,
+				struct amdgpu_ring *ring);
 int amdgpu_mes_unmap_legacy_queue(struct amdgpu_device *adev,
 				  struct amdgpu_ring *ring,
 				  enum amdgpu_unmap_queues_action action,
 				  u64 gpu_addr, u64 seq);
+int amdgpu_mes_reset_legacy_queue(struct amdgpu_device *adev,
+				  struct amdgpu_ring *ring,
+				  unsigned int vmid,
+				  bool use_mmio);
+
+int amdgpu_mes_get_hung_queue_db_array_size(struct amdgpu_device *adev);
+int amdgpu_mes_detect_and_reset_hung_queues(struct amdgpu_device *adev,
+					    int queue_type,
+					    bool detect_only,
+					    unsigned int *hung_db_num,
+					    u32 *hung_db_array);
 
 uint32_t amdgpu_mes_rreg(struct amdgpu_device *adev, uint32_t reg);
 int amdgpu_mes_wreg(struct amdgpu_device *adev,
 		    uint32_t reg, uint32_t val);
-int amdgpu_mes_reg_wait(struct amdgpu_device *adev, uint32_t reg,
-			uint32_t val, uint32_t mask);
 int amdgpu_mes_reg_write_reg_wait(struct amdgpu_device *adev,
 				  uint32_t reg0, uint32_t reg1,
 				  uint32_t ref, uint32_t mask);
-
-int amdgpu_mes_add_ring(struct amdgpu_device *adev, int gang_id,
-			int queue_type, int idx,
-			struct amdgpu_mes_ctx_data *ctx_data,
-			struct amdgpu_ring **out);
-void amdgpu_mes_remove_ring(struct amdgpu_device *adev,
-			    struct amdgpu_ring *ring);
+int amdgpu_mes_hdp_flush(struct amdgpu_device *adev);
+int amdgpu_mes_set_shader_debugger(struct amdgpu_device *adev,
+				uint64_t process_context_addr,
+				uint32_t spi_gdbg_per_vmid_cntl,
+				const uint32_t *tcp_watch_cntl,
+				uint32_t flags,
+				bool trap_en);
+int amdgpu_mes_flush_shader_debugger(struct amdgpu_device *adev,
+				uint64_t process_context_addr);
 
 uint32_t amdgpu_mes_get_aggregated_doorbell_index(struct amdgpu_device *adev,
 						   enum amdgpu_mes_priority_level prio);
 
-int amdgpu_mes_ctx_alloc_meta_data(struct amdgpu_device *adev,
-				   struct amdgpu_mes_ctx_data *ctx_data);
-void amdgpu_mes_ctx_free_meta_data(struct amdgpu_mes_ctx_data *ctx_data);
-int amdgpu_mes_ctx_map_meta_data(struct amdgpu_device *adev,
-				 struct amdgpu_vm *vm,
-				 struct amdgpu_mes_ctx_data *ctx_data);
-int amdgpu_mes_ctx_unmap_meta_data(struct amdgpu_device *adev,
-				   struct amdgpu_mes_ctx_data *ctx_data);
-
-int amdgpu_mes_self_test(struct amdgpu_device *adev);
-
-int amdgpu_mes_alloc_process_doorbells(struct amdgpu_device *adev,
-					unsigned int *doorbell_index);
-void amdgpu_mes_free_process_doorbells(struct amdgpu_device *adev,
-					unsigned int doorbell_index);
-unsigned int amdgpu_mes_get_doorbell_dw_offset_in_bar(
-					struct amdgpu_device *adev,
-					uint32_t doorbell_index,
-					unsigned int doorbell_id);
 int amdgpu_mes_doorbell_process_slice(struct amdgpu_device *adev);
 
 /*
@@ -429,4 +501,9 @@ static inline void amdgpu_mes_unlock(struct amdgpu_mes *mes)
 	memalloc_noreclaim_restore(mes->saved_flags);
 	mutex_unlock(&mes->mutex_hidden);
 }
+
+bool amdgpu_mes_suspend_resume_all_supported(struct amdgpu_device *adev);
+
+int amdgpu_mes_update_enforce_isolation(struct amdgpu_device *adev);
+
 #endif /* __AMDGPU_MES_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h
index d21bb6dae56e..1ca9d4ed8063 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h
@@ -21,6 +21,29 @@
 #ifndef __AMDGPU_MMHUB_H__
 #define __AMDGPU_MMHUB_H__
 
+enum amdgpu_mmhub_ras_memory_id {
+	AMDGPU_MMHUB_WGMI_PAGEMEM = 0,
+	AMDGPU_MMHUB_RGMI_PAGEMEM = 1,
+	AMDGPU_MMHUB_WDRAM_PAGEMEM = 2,
+	AMDGPU_MMHUB_RDRAM_PAGEMEM = 3,
+	AMDGPU_MMHUB_WIO_CMDMEM = 4,
+	AMDGPU_MMHUB_RIO_CMDMEM = 5,
+	AMDGPU_MMHUB_WGMI_CMDMEM = 6,
+	AMDGPU_MMHUB_RGMI_CMDMEM = 7,
+	AMDGPU_MMHUB_WDRAM_CMDMEM = 8,
+	AMDGPU_MMHUB_RDRAM_CMDMEM = 9,
+	AMDGPU_MMHUB_MAM_DMEM0 = 10,
+	AMDGPU_MMHUB_MAM_DMEM1 = 11,
+	AMDGPU_MMHUB_MAM_DMEM2 = 12,
+	AMDGPU_MMHUB_MAM_DMEM3 = 13,
+	AMDGPU_MMHUB_WRET_TAGMEM = 19,
+	AMDGPU_MMHUB_RRET_TAGMEM = 20,
+	AMDGPU_MMHUB_WIO_DATAMEM = 21,
+	AMDGPU_MMHUB_WGMI_DATAMEM = 22,
+	AMDGPU_MMHUB_WDRAM_DATAMEM = 23,
+	AMDGPU_MMHUB_MEMORY_BLOCK_LAST,
+};
+
 struct amdgpu_mmhub_ras {
 	struct amdgpu_ras_block_object ras_block;
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h
index 32fe05c810c6..dc8d2f52c7d6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h
@@ -32,7 +32,6 @@
 
 #include <drm/display/drm_dp_helper.h>
 #include <drm/drm_crtc.h>
-#include <drm/drm_edid.h>
 #include <drm/drm_encoder.h>
 #include <drm/drm_fixed.h>
 #include <drm/drm_framebuffer.h>
@@ -51,6 +50,8 @@ struct amdgpu_device;
 struct amdgpu_encoder;
 struct amdgpu_router;
 struct amdgpu_hpd;
+struct edid;
+struct drm_edid;
 
 #define to_amdgpu_crtc(x) container_of(x, struct amdgpu_crtc, base)
 #define to_amdgpu_connector(x) container_of(x, struct amdgpu_connector, base)
@@ -300,6 +301,7 @@ struct amdgpu_framebuffer {
 
 	uint64_t tiling_flags;
 	bool tmz_surface;
+	bool gfx12_dcc;
 
 	/* caching for later use */
 	uint64_t address;
@@ -327,8 +329,7 @@ struct amdgpu_mode_info {
 	/* Adaptive Backlight Modulation (power feature) */
 	struct drm_property *abm_level_property;
 	/* hardcoded DFP edid from BIOS */
-	struct edid *bios_hardcoded_edid;
-	int bios_hardcoded_edid_size;
+	const struct drm_edid *bios_hardcoded_edid;
 
 	/* firmware flags */
 	u32 firmware_flags;
@@ -343,6 +344,97 @@ struct amdgpu_mode_info {
 	int			disp_priority;
 	const struct amdgpu_display_funcs *funcs;
 	const enum drm_plane_type *plane_type;
+
+	/* Driver-private color mgmt props */
+
+	/* @plane_degamma_lut_property: Plane property to set a degamma LUT to
+	 * convert encoded values to light linear values before sampling or
+	 * blending.
+	 */
+	struct drm_property *plane_degamma_lut_property;
+	/* @plane_degamma_lut_size_property: Plane property to define the max
+	 * size of degamma LUT as supported by the driver (read-only).
+	 */
+	struct drm_property *plane_degamma_lut_size_property;
+	/**
+	 * @plane_degamma_tf_property: Plane pre-defined transfer function to
+	 * to go from scanout/encoded values to linear values.
+	 */
+	struct drm_property *plane_degamma_tf_property;
+	/**
+	 * @plane_hdr_mult_property:
+	 */
+	struct drm_property *plane_hdr_mult_property;
+
+	struct drm_property *plane_ctm_property;
+	/**
+	 * @shaper_lut_property: Plane property to set pre-blending shaper LUT
+	 * that converts color content before 3D LUT. If
+	 * plane_shaper_tf_property != Identity TF, AMD color module will
+	 * combine the user LUT values with pre-defined TF into the LUT
+	 * parameters to be programmed.
+	 */
+	struct drm_property *plane_shaper_lut_property;
+	/**
+	 * @shaper_lut_size_property: Plane property for the size of
+	 * pre-blending shaper LUT as supported by the driver (read-only).
+	 */
+	struct drm_property *plane_shaper_lut_size_property;
+	/**
+	 * @plane_shaper_tf_property: Plane property to set a predefined
+	 * transfer function for pre-blending shaper (before applying 3D LUT)
+	 * with or without LUT. There is no shaper ROM, but we can use AMD
+	 * color modules to program LUT parameters from predefined TF (or
+	 * from a combination of pre-defined TF and the custom 1D LUT).
+	 */
+	struct drm_property *plane_shaper_tf_property;
+	/**
+	 * @plane_lut3d_property: Plane property for color transformation using
+	 * a 3D LUT (pre-blending), a three-dimensional array where each
+	 * element is an RGB triplet. Each dimension has the size of
+	 * lut3d_size. The array contains samples from the approximated
+	 * function. On AMD, values between samples are estimated by
+	 * tetrahedral interpolation. The array is accessed with three indices,
+	 * one for each input dimension (color channel), blue being the
+	 * outermost dimension, red the innermost.
+	 */
+	struct drm_property *plane_lut3d_property;
+	/**
+	 * @plane_degamma_lut_size_property: Plane property to define the max
+	 * size of 3D LUT as supported by the driver (read-only). The max size
+	 * is the max size of one dimension and, therefore, the max number of
+	 * entries for 3D LUT array is the 3D LUT size cubed;
+	 */
+	struct drm_property *plane_lut3d_size_property;
+	/**
+	 * @plane_blend_lut_property: Plane property for output gamma before
+	 * blending. Userspace set a blend LUT to convert colors after 3D LUT
+	 * conversion. It works as a post-3DLUT 1D LUT. With shaper LUT, they
+	 * are sandwiching 3D LUT with two 1D LUT. If plane_blend_tf_property
+	 * != Identity TF, AMD color module will combine the user LUT values
+	 * with pre-defined TF into the LUT parameters to be programmed.
+	 */
+	struct drm_property *plane_blend_lut_property;
+	/**
+	 * @plane_blend_lut_size_property: Plane property to define the max
+	 * size of blend LUT as supported by the driver (read-only).
+	 */
+	struct drm_property *plane_blend_lut_size_property;
+	/**
+	 * @plane_blend_tf_property: Plane property to set a predefined
+	 * transfer function for pre-blending blend/out_gamma (after applying
+	 * 3D LUT) with or without LUT. There is no blend ROM, but we can use
+	 * AMD color modules to program LUT parameters from predefined TF (or
+	 * from a combination of pre-defined TF and the custom 1D LUT).
+	 */
+	struct drm_property *plane_blend_tf_property;
+	/* @regamma_tf_property: Transfer function for CRTC regamma
+	 * (post-blending). Possible values are defined by `enum
+	 * amdgpu_transfer_function`. There is no regamma ROM, but we can use
+	 * AMD color modules to program LUT parameters from predefined TF (or
+	 * from a combination of pre-defined TF and the custom 1D LUT).
+	 */
+	struct drm_property *regamma_tf_property;
 };
 
 #define AMDGPU_MAX_BL_LEVEL 0xFF
@@ -406,8 +498,6 @@ struct amdgpu_crtc {
 	struct drm_connector *connector;
 	/* for dpm */
 	u32 line_time;
-	u32 wm_low;
-	u32 wm_high;
 	u32 lb_vblank_lead_lines;
 	struct drm_display_mode hw_mode;
 	/* for virtual dce */
@@ -416,6 +506,10 @@ struct amdgpu_crtc {
 
 	int otg_inst;
 	struct drm_pending_vblank_event *event;
+
+	bool wb_pending;
+	bool wb_enabled;
+	struct drm_writeback_connector *wb_conn;
 };
 
 struct amdgpu_encoder_atom_dig {
@@ -515,6 +609,7 @@ struct amdgpu_i2c_adapter {
 	struct i2c_adapter base;
 
 	struct ddc_service *ddc_service;
+	bool oem;
 };
 
 #define TO_DM_AUX(x) container_of((x), struct amdgpu_dm_dp_aux, aux)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c
index a3bc00577a7c..a974265837f0 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c
@@ -45,6 +45,24 @@ int amdgpu_nbio_ras_sw_init(struct amdgpu_device *adev)
 	return 0;
 }
 
+u64 amdgpu_nbio_get_pcie_replay_count(struct amdgpu_device *adev)
+{
+	if (adev->nbio.funcs && adev->nbio.funcs->get_pcie_replay_count)
+		return adev->nbio.funcs->get_pcie_replay_count(adev);
+
+	return 0;
+}
+
+bool amdgpu_nbio_is_replay_cnt_supported(struct amdgpu_device *adev)
+{
+	if (amdgpu_sriov_vf(adev) || !adev->asic_funcs ||
+	    !adev->asic_funcs->get_pcie_replay_count ||
+	    (!adev->nbio.funcs || !adev->nbio.funcs->get_pcie_replay_count))
+		return false;
+
+	return true;
+}
+
 int amdgpu_nbio_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
 {
 	int r;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.h
index c686ff4bcc39..b528de6a01f6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.h
@@ -61,6 +61,7 @@ struct amdgpu_nbio_funcs {
 	u32 (*get_hdp_flush_done_offset)(struct amdgpu_device *adev);
 	u32 (*get_pcie_index_offset)(struct amdgpu_device *adev);
 	u32 (*get_pcie_data_offset)(struct amdgpu_device *adev);
+	u32 (*get_pcie_index_hi_offset)(struct amdgpu_device *adev);
 	u32 (*get_pcie_port_index_offset)(struct amdgpu_device *adev);
 	u32 (*get_pcie_port_data_offset)(struct amdgpu_device *adev);
 	u32 (*get_rev_id)(struct amdgpu_device *adev);
@@ -68,6 +69,8 @@ struct amdgpu_nbio_funcs {
 	u32 (*get_memsize)(struct amdgpu_device *adev);
 	void (*sdma_doorbell_range)(struct amdgpu_device *adev, int instance,
 			bool use_doorbell, int doorbell_index, int doorbell_size);
+	void (*vpe_doorbell_range)(struct amdgpu_device *adev, int instance,
+			bool use_doorbell, int doorbell_index, int doorbell_size);
 	void (*vcn_doorbell_range)(struct amdgpu_device *adev, bool use_doorbell,
 				   int doorbell_index, int instance);
 	void (*gc_doorbell_init)(struct amdgpu_device *adev);
@@ -95,6 +98,12 @@ struct amdgpu_nbio_funcs {
 	void (*apply_l1_link_width_reconfig_wa)(struct amdgpu_device *adev);
 	void (*clear_doorbell_interrupt)(struct amdgpu_device *adev);
 	u32 (*get_rom_offset)(struct amdgpu_device *adev);
+	int (*get_compute_partition_mode)(struct amdgpu_device *adev);
+	u32 (*get_memory_partition_mode)(struct amdgpu_device *adev,
+					 u32 *supp_modes);
+	bool (*is_nps_switch_requested)(struct amdgpu_device *adev);
+	u64 (*get_pcie_replay_count)(struct amdgpu_device *adev);
+	void (*set_reg_remap)(struct amdgpu_device *adev);
 };
 
 struct amdgpu_nbio {
@@ -108,4 +117,8 @@ struct amdgpu_nbio {
 
 int amdgpu_nbio_ras_sw_init(struct amdgpu_device *adev);
 int amdgpu_nbio_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block);
+u64 amdgpu_nbio_get_pcie_replay_count(struct amdgpu_device *adev);
+
+bool amdgpu_nbio_is_replay_cnt_supported(struct amdgpu_device *adev);
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
index 2bd1a54ee866..e08f58de4b17 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
@@ -32,6 +32,7 @@
 #include <linux/list.h>
 #include <linux/slab.h>
 #include <linux/dma-buf.h>
+#include <linux/export.h>
 
 #include <drm/drm_drv.h>
 #include <drm/amdgpu_drm.h>
@@ -39,6 +40,9 @@
 #include "amdgpu.h"
 #include "amdgpu_trace.h"
 #include "amdgpu_amdkfd.h"
+#include "amdgpu_vram_mgr.h"
+#include "amdgpu_vm.h"
+#include "amdgpu_dma_buf.h"
 
 /**
  * DOC: amdgpu_object
@@ -59,7 +63,7 @@ static void amdgpu_bo_destroy(struct ttm_buffer_object *tbo)
 
 	amdgpu_bo_kunmap(bo);
 
-	if (bo->tbo.base.import_attach)
+	if (drm_gem_is_imported(&bo->tbo.base))
 		drm_prime_gem_destroy(&bo->tbo.base, bo->tbo.sg);
 	drm_gem_object_release(&bo->tbo.base);
 	amdgpu_bo_unref(&bo->parent);
@@ -76,23 +80,6 @@ static void amdgpu_bo_user_destroy(struct ttm_buffer_object *tbo)
 	amdgpu_bo_destroy(tbo);
 }
 
-static void amdgpu_bo_vm_destroy(struct ttm_buffer_object *tbo)
-{
-	struct amdgpu_device *adev = amdgpu_ttm_adev(tbo->bdev);
-	struct amdgpu_bo *bo = ttm_to_amdgpu_bo(tbo);
-	struct amdgpu_bo_vm *vmbo;
-
-	vmbo = to_amdgpu_bo_vm(bo);
-	/* in case amdgpu_device_recover_vram got NULL of bo->parent */
-	if (!list_empty(&vmbo->shadow_list)) {
-		mutex_lock(&adev->shadow_list_lock);
-		list_del_init(&vmbo->shadow_list);
-		mutex_unlock(&adev->shadow_list_lock);
-	}
-
-	amdgpu_bo_destroy(tbo);
-}
-
 /**
  * amdgpu_bo_is_amdgpu_bo - check if the buffer object is an &amdgpu_bo
  * @bo: buffer object to be checked
@@ -106,8 +93,7 @@ static void amdgpu_bo_vm_destroy(struct ttm_buffer_object *tbo)
 bool amdgpu_bo_is_amdgpu_bo(struct ttm_buffer_object *bo)
 {
 	if (bo->destroy == &amdgpu_bo_destroy ||
-	    bo->destroy == &amdgpu_bo_user_destroy ||
-	    bo->destroy == &amdgpu_bo_vm_destroy)
+	    bo->destroy == &amdgpu_bo_user_destroy)
 		return true;
 
 	return false;
@@ -130,20 +116,48 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
 	u32 c = 0;
 
 	if (domain & AMDGPU_GEM_DOMAIN_VRAM) {
-		unsigned visible_pfn = adev->gmc.visible_vram_size >> PAGE_SHIFT;
-
-		places[c].fpfn = 0;
-		places[c].lpfn = 0;
+		unsigned int visible_pfn = adev->gmc.visible_vram_size >> PAGE_SHIFT;
+		int8_t mem_id = KFD_XCP_MEM_ID(adev, abo->xcp_id);
+
+		if (adev->gmc.mem_partitions && mem_id >= 0) {
+			places[c].fpfn = adev->gmc.mem_partitions[mem_id].range.fpfn;
+			/*
+			 * memory partition range lpfn is inclusive start + size - 1
+			 * TTM place lpfn is exclusive start + size
+			 */
+			places[c].lpfn = adev->gmc.mem_partitions[mem_id].range.lpfn + 1;
+		} else {
+			places[c].fpfn = 0;
+			places[c].lpfn = 0;
+		}
 		places[c].mem_type = TTM_PL_VRAM;
 		places[c].flags = 0;
 
 		if (flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)
-			places[c].lpfn = visible_pfn;
-		else if (adev->gmc.real_vram_size != adev->gmc.visible_vram_size)
+			places[c].lpfn = min_not_zero(places[c].lpfn, visible_pfn);
+		else
 			places[c].flags |= TTM_PL_FLAG_TOPDOWN;
 
-		if (flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)
+		if (abo->tbo.type == ttm_bo_type_kernel &&
+		    flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)
 			places[c].flags |= TTM_PL_FLAG_CONTIGUOUS;
+
+		c++;
+	}
+
+	if (domain & AMDGPU_GEM_DOMAIN_DOORBELL) {
+		places[c].fpfn = 0;
+		places[c].lpfn = 0;
+		places[c].mem_type = AMDGPU_PL_DOORBELL;
+		places[c].flags = 0;
+		c++;
+	}
+
+	if (domain & AMDGPU_GEM_DOMAIN_MMIO_REMAP) {
+		places[c].fpfn = 0;
+		places[c].lpfn = 0;
+		places[c].mem_type = AMDGPU_PL_MMIO_REMAP;
+		places[c].flags = 0;
 		c++;
 	}
 
@@ -154,6 +168,13 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
 			abo->flags & AMDGPU_GEM_CREATE_PREEMPTIBLE ?
 			AMDGPU_PL_PREEMPT : TTM_PL_TT;
 		places[c].flags = 0;
+		/*
+		 * When GTT is just an alternative to VRAM make sure that we
+		 * only use it as fallback and still try to fill up VRAM first.
+		 */
+		if (abo->tbo.resource && !(adev->flags & AMD_IS_APU) &&
+		    domain & abo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM)
+			places[c].flags |= TTM_PL_FLAG_FALLBACK;
 		c++;
 	}
 
@@ -201,9 +222,6 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
 
 	placement->num_placement = c;
 	placement->placement = places;
-
-	placement->num_busy_placement = c;
-	placement->busy_placement = places;
 }
 
 /**
@@ -316,6 +334,9 @@ error_free:
  *
  * Allocates and pins a BO for kernel internal use.
  *
+ * This function is exported to allow the V4L2 isp device
+ * external to drm device to create and access the kernel BO.
+ *
  * Note: For bo_ptr new BO is only created if bo_ptr points to NULL.
  *
  * Returns:
@@ -341,6 +362,74 @@ int amdgpu_bo_create_kernel(struct amdgpu_device *adev,
 }
 
 /**
+ * amdgpu_bo_create_isp_user - create user BO for isp
+ *
+ * @adev: amdgpu device object
+ * @dma_buf: DMABUF handle for isp buffer
+ * @domain: where to place it
+ * @bo:  used to initialize BOs in structures
+ * @gpu_addr: GPU addr of the pinned BO
+ *
+ * Imports isp DMABUF to allocate and pin a user BO for isp internal use. It does
+ * GART alloc to generate gpu_addr for BO to make it accessible through the
+ * GART aperture for ISP HW.
+ *
+ * This function is exported to allow the V4L2 isp device external to drm device
+ * to create and access the isp user BO.
+ *
+ * Returns:
+ * 0 on success, negative error code otherwise.
+ */
+int amdgpu_bo_create_isp_user(struct amdgpu_device *adev,
+			   struct dma_buf *dma_buf, u32 domain, struct amdgpu_bo **bo,
+			   u64 *gpu_addr)
+
+{
+	struct drm_gem_object *gem_obj;
+	int r;
+
+	gem_obj = amdgpu_gem_prime_import(&adev->ddev, dma_buf);
+	*bo = gem_to_amdgpu_bo(gem_obj);
+	if (!(*bo)) {
+		dev_err(adev->dev, "failed to get valid isp user bo\n");
+		return -EINVAL;
+	}
+
+	r = amdgpu_bo_reserve(*bo, false);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to reserve isp user bo\n", r);
+		return r;
+	}
+
+	r = amdgpu_bo_pin(*bo, domain);
+	if (r) {
+		dev_err(adev->dev, "(%d) isp user bo pin failed\n", r);
+		goto error_unreserve;
+	}
+
+	r = amdgpu_ttm_alloc_gart(&(*bo)->tbo);
+	if (r) {
+		dev_err(adev->dev, "%p bind failed\n", *bo);
+		goto error_unpin;
+	}
+
+	if (!WARN_ON(!gpu_addr))
+		*gpu_addr = amdgpu_bo_gpu_offset(*bo);
+
+	amdgpu_bo_unreserve(*bo);
+
+	return 0;
+
+error_unpin:
+	amdgpu_bo_unpin(*bo);
+error_unreserve:
+	amdgpu_bo_unreserve(*bo);
+	amdgpu_bo_unref(bo);
+
+	return r;
+}
+
+/**
  * amdgpu_bo_create_kernel_at - create BO for kernel use at specific location
  *
  * @adev: amdgpu device object
@@ -415,6 +504,9 @@ error:
  * @cpu_addr: pointer to where the BO's CPU memory space address was stored
  *
  * unmaps and unpin a BO for kernel internal use.
+ *
+ * This function is exported to allow the V4L2 isp device
+ * external to drm device to free the kernel BO.
  */
 void amdgpu_bo_free_kernel(struct amdgpu_bo **bo, u64 *gpu_addr,
 			   void **cpu_addr)
@@ -440,7 +532,29 @@ void amdgpu_bo_free_kernel(struct amdgpu_bo **bo, u64 *gpu_addr,
 		*cpu_addr = NULL;
 }
 
-/* Validate bo size is bit bigger then the request domain */
+/**
+ * amdgpu_bo_free_isp_user - free BO for isp use
+ *
+ * @bo: amdgpu isp user BO to free
+ *
+ * unpin and unref BO for isp internal use.
+ *
+ * This function is exported to allow the V4L2 isp device
+ * external to drm device to free the isp user BO.
+ */
+void amdgpu_bo_free_isp_user(struct amdgpu_bo *bo)
+{
+	if (bo == NULL)
+		return;
+
+	if (amdgpu_bo_reserve(bo, true) == 0) {
+		amdgpu_bo_unpin(bo);
+		amdgpu_bo_unreserve(bo);
+	}
+	amdgpu_bo_unref(&bo);
+}
+
+/* Validate bo size is bit bigger than the request domain */
 static bool amdgpu_bo_validate_size(struct amdgpu_device *adev,
 					  unsigned long size, u32 domain)
 {
@@ -450,29 +564,24 @@ static bool amdgpu_bo_validate_size(struct amdgpu_device *adev,
 	 * If GTT is part of requested domains the check must succeed to
 	 * allow fall back to GTT.
 	 */
-	if (domain & AMDGPU_GEM_DOMAIN_GTT) {
+	if (domain & AMDGPU_GEM_DOMAIN_GTT)
 		man = ttm_manager_type(&adev->mman.bdev, TTM_PL_TT);
-
-		if (man && size < man->size)
-			return true;
-		else if (!man)
-			WARN_ON_ONCE("GTT domain requested but GTT mem manager uninitialized");
-		goto fail;
-	} else if (domain & AMDGPU_GEM_DOMAIN_VRAM) {
+	else if (domain & AMDGPU_GEM_DOMAIN_VRAM)
 		man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM);
+	else
+		return true;
 
-		if (man && size < man->size)
-			return true;
-		goto fail;
+	if (!man) {
+		if (domain & AMDGPU_GEM_DOMAIN_GTT)
+			WARN_ON_ONCE("GTT domain requested but GTT mem manager uninitialized");
+		return false;
 	}
 
-	/* TODO add more domains checks, such as AMDGPU_GEM_DOMAIN_CPU */
-	return true;
+	/* TODO add more domains checks, such as AMDGPU_GEM_DOMAIN_CPU, _DOMAIN_DOORBELL */
+	if (size < man->size)
+		return true;
 
-fail:
-	if (man)
-		DRM_DEBUG("BO size %lu > total memory in domain: %llu\n", size,
-			  man->size);
+	DRM_DEBUG("BO size %lu > total memory in domain: %llu\n", size, man->size);
 	return false;
 }
 
@@ -563,6 +672,7 @@ int amdgpu_bo_create(struct amdgpu_device *adev,
 	if (bo == NULL)
 		return -ENOMEM;
 	drm_gem_private_object_init(adev_to_drm(adev), &bo->tbo.base, size);
+	bo->tbo.base.funcs = &amdgpu_gem_object_funcs;
 	bo->vm_bo = NULL;
 	bo->preferred_domains = bp->preferred_domain ? bp->preferred_domain :
 		bp->domain;
@@ -574,12 +684,16 @@ int amdgpu_bo_create(struct amdgpu_device *adev,
 
 	bo->flags = bp->flags;
 
+	if (adev->gmc.mem_partitions)
+		/* For GPUs with spatial partitioning, bo->xcp_id=-1 means any partition */
+		bo->xcp_id = bp->xcp_id_plus1 - 1;
+	else
+		/* For GPUs without spatial partitioning */
+		bo->xcp_id = 0;
+
 	if (!amdgpu_bo_support_uswc(bo->flags))
 		bo->flags &= ~AMDGPU_GEM_CREATE_CPU_GTT_USWC;
 
-	if (adev->ras_enabled)
-		bo->flags |= AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE;
-
 	bo->tbo.bdev = &adev->mman.bdev;
 	if (bp->domain & (AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA |
 			  AMDGPU_GEM_DOMAIN_GDS))
@@ -587,6 +701,8 @@ int amdgpu_bo_create(struct amdgpu_device *adev,
 	else
 		amdgpu_bo_placement_from_domain(bo, bp->domain);
 	if (bp->type == ttm_bo_type_kernel)
+		bo->tbo.priority = 2;
+	else if (!(bp->flags & AMDGPU_GEM_CREATE_DISCARDABLE))
 		bo->tbo.priority = 1;
 
 	if (!bp->destroy)
@@ -599,8 +715,7 @@ int amdgpu_bo_create(struct amdgpu_device *adev,
 		return r;
 
 	if (!amdgpu_gmc_vram_full_visible(&adev->gmc) &&
-	    bo->tbo.resource->mem_type == TTM_PL_VRAM &&
-	    amdgpu_bo_in_cpu_visible_vram(bo))
+	    amdgpu_res_cpu_visible(adev, bo->tbo.resource))
 		amdgpu_cs_report_moved_bytes(adev, ctx.bytes_moved,
 					     ctx.bytes_moved);
 	else
@@ -610,7 +725,7 @@ int amdgpu_bo_create(struct amdgpu_device *adev,
 	    bo->tbo.resource->mem_type == TTM_PL_VRAM) {
 		struct dma_fence *fence;
 
-		r = amdgpu_fill_buffer(bo, 0, bo->tbo.base.resv, &fence);
+		r = amdgpu_ttm_clear_buffer(bo, bo->tbo.base.resv, &fence);
 		if (unlikely(r))
 			goto fail_unreserve;
 
@@ -694,59 +809,10 @@ int amdgpu_bo_create_vm(struct amdgpu_device *adev,
 		return r;
 
 	*vmbo_ptr = to_amdgpu_bo_vm(bo_ptr);
-	INIT_LIST_HEAD(&(*vmbo_ptr)->shadow_list);
-	/* Set destroy callback to amdgpu_bo_vm_destroy after vmbo->shadow_list
-	 * is initialized.
-	 */
-	bo_ptr->tbo.destroy = &amdgpu_bo_vm_destroy;
 	return r;
 }
 
 /**
- * amdgpu_bo_add_to_shadow_list - add a BO to the shadow list
- *
- * @vmbo: BO that will be inserted into the shadow list
- *
- * Insert a BO to the shadow list.
- */
-void amdgpu_bo_add_to_shadow_list(struct amdgpu_bo_vm *vmbo)
-{
-	struct amdgpu_device *adev = amdgpu_ttm_adev(vmbo->bo.tbo.bdev);
-
-	mutex_lock(&adev->shadow_list_lock);
-	list_add_tail(&vmbo->shadow_list, &adev->shadow_list);
-	mutex_unlock(&adev->shadow_list_lock);
-}
-
-/**
- * amdgpu_bo_restore_shadow - restore an &amdgpu_bo shadow
- *
- * @shadow: &amdgpu_bo shadow to be restored
- * @fence: dma_fence associated with the operation
- *
- * Copies a buffer object's shadow content back to the object.
- * This is used for recovering a buffer from its shadow in case of a gpu
- * reset where vram context may be lost.
- *
- * Returns:
- * 0 for success or a negative error code on failure.
- */
-int amdgpu_bo_restore_shadow(struct amdgpu_bo *shadow, struct dma_fence **fence)
-
-{
-	struct amdgpu_device *adev = amdgpu_ttm_adev(shadow->tbo.bdev);
-	struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
-	uint64_t shadow_addr, parent_addr;
-
-	shadow_addr = amdgpu_bo_gpu_offset(shadow);
-	parent_addr = amdgpu_bo_gpu_offset(shadow->parent);
-
-	return amdgpu_copy_buffer(ring, shadow_addr, parent_addr,
-				  amdgpu_bo_size(shadow), NULL, fence,
-				  true, false, false);
-}
-
-/**
  * amdgpu_bo_kmap - map an &amdgpu_bo buffer object
  * @bo: &amdgpu_bo buffer object to be mapped
  * @ptr: kernel virtual address to be returned
@@ -829,7 +895,7 @@ struct amdgpu_bo *amdgpu_bo_ref(struct amdgpu_bo *bo)
 	if (bo == NULL)
 		return NULL;
 
-	ttm_bo_get(&bo->tbo);
+	drm_gem_object_get(&bo->tbo.base);
 	return bo;
 }
 
@@ -841,40 +907,30 @@ struct amdgpu_bo *amdgpu_bo_ref(struct amdgpu_bo *bo)
  */
 void amdgpu_bo_unref(struct amdgpu_bo **bo)
 {
-	struct ttm_buffer_object *tbo;
-
 	if ((*bo) == NULL)
 		return;
 
-	tbo = &((*bo)->tbo);
-	ttm_bo_put(tbo);
+	drm_gem_object_put(&(*bo)->tbo.base);
 	*bo = NULL;
 }
 
 /**
- * amdgpu_bo_pin_restricted - pin an &amdgpu_bo buffer object
+ * amdgpu_bo_pin - pin an &amdgpu_bo buffer object
  * @bo: &amdgpu_bo buffer object to be pinned
  * @domain: domain to be pinned to
- * @min_offset: the start of requested address range
- * @max_offset: the end of requested address range
  *
- * Pins the buffer object according to requested domain and address range. If
- * the memory is unbound gart memory, binds the pages into gart table. Adjusts
- * pin_count and pin_size accordingly.
+ * Pins the buffer object according to requested domain. If the memory is
+ * unbound gart memory, binds the pages into gart table. Adjusts pin_count and
+ * pin_size accordingly.
  *
  * Pinning means to lock pages in memory along with keeping them at a fixed
  * offset. It is required when a buffer can not be moved, for example, when
  * a display buffer is being scanned out.
  *
- * Compared with amdgpu_bo_pin(), this function gives more flexibility on
- * where to pin a buffer if there are specific restrictions on where a buffer
- * must be located.
- *
  * Returns:
  * 0 for success or a negative error code on failure.
  */
-int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
-			     u64 min_offset, u64 max_offset)
+int amdgpu_bo_pin(struct amdgpu_bo *bo, u32 domain)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
 	struct ttm_operation_ctx ctx = { false, false };
@@ -883,15 +939,12 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
 	if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm))
 		return -EPERM;
 
-	if (WARN_ON_ONCE(min_offset > max_offset))
-		return -EINVAL;
-
 	/* Check domain to be pinned to against preferred domains */
 	if (bo->preferred_domains & domain)
 		domain = bo->preferred_domains & domain;
 
 	/* A shared bo cannot be migrated to VRAM */
-	if (bo->tbo.base.import_attach) {
+	if (drm_gem_is_imported(&bo->tbo.base)) {
 		if (domain & AMDGPU_GEM_DOMAIN_GTT)
 			domain = AMDGPU_GEM_DOMAIN_GTT;
 		else
@@ -911,14 +964,6 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
 			return -EINVAL;
 
 		ttm_bo_pin(&bo->tbo);
-
-		if (max_offset != 0) {
-			u64 domain_start = amdgpu_ttm_domain_start(adev,
-								   mem_type);
-			WARN_ON_ONCE(max_offset <
-				     (amdgpu_bo_gpu_offset(bo) - domain_start));
-		}
-
 		return 0;
 	}
 
@@ -927,7 +972,7 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
 	 */
 	domain = amdgpu_bo_get_preferred_domain(adev, domain);
 
-	if (bo->tbo.base.import_attach)
+	if (drm_gem_is_imported(&bo->tbo.base))
 		dma_buf_pin(bo->tbo.base.import_attach);
 
 	/* force to pin into visible video ram */
@@ -935,16 +980,9 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
 		bo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
 	amdgpu_bo_placement_from_domain(bo, domain);
 	for (i = 0; i < bo->placement.num_placement; i++) {
-		unsigned fpfn, lpfn;
-
-		fpfn = min_offset >> PAGE_SHIFT;
-		lpfn = max_offset >> PAGE_SHIFT;
-
-		if (fpfn > bo->placements[i].fpfn)
-			bo->placements[i].fpfn = fpfn;
-		if (!bo->placements[i].lpfn ||
-		    (lpfn && lpfn < bo->placements[i].lpfn))
-			bo->placements[i].lpfn = lpfn;
+		if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS &&
+		    bo->placements[i].mem_type == TTM_PL_VRAM)
+			bo->placements[i].flags |= TTM_PL_FLAG_CONTIGUOUS;
 	}
 
 	r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
@@ -955,12 +993,11 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
 
 	ttm_bo_pin(&bo->tbo);
 
-	domain = amdgpu_mem_type_to_domain(bo->tbo.resource->mem_type);
-	if (domain == AMDGPU_GEM_DOMAIN_VRAM) {
+	if (bo->tbo.resource->mem_type == TTM_PL_VRAM) {
 		atomic64_add(amdgpu_bo_size(bo), &adev->vram_pin_size);
 		atomic64_add(amdgpu_vram_mgr_bo_visible_size(bo),
 			     &adev->visible_pin_size);
-	} else if (domain == AMDGPU_GEM_DOMAIN_GTT) {
+	} else if (bo->tbo.resource->mem_type == TTM_PL_TT) {
 		atomic64_add(amdgpu_bo_size(bo), &adev->gart_pin_size);
 	}
 
@@ -969,24 +1006,6 @@ error:
 }
 
 /**
- * amdgpu_bo_pin - pin an &amdgpu_bo buffer object
- * @bo: &amdgpu_bo buffer object to be pinned
- * @domain: domain to be pinned to
- *
- * A simple wrapper to amdgpu_bo_pin_restricted().
- * Provides a simpler API for buffers that do not have any strict restrictions
- * on where a buffer must be located.
- *
- * Returns:
- * 0 for success or a negative error code on failure.
- */
-int amdgpu_bo_pin(struct amdgpu_bo *bo, u32 domain)
-{
-	bo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
-	return amdgpu_bo_pin_restricted(bo, domain, 0, 0);
-}
-
-/**
  * amdgpu_bo_unpin - unpin an &amdgpu_bo buffer object
  * @bo: &amdgpu_bo buffer object to be unpinned
  *
@@ -1004,7 +1023,7 @@ void amdgpu_bo_unpin(struct amdgpu_bo *bo)
 	if (bo->tbo.pin_count)
 		return;
 
-	if (bo->tbo.base.import_attach)
+	if (drm_gem_is_imported(&bo->tbo.base))
 		dma_buf_unpin(bo->tbo.base.import_attach);
 
 	if (bo->tbo.resource->mem_type == TTM_PL_VRAM) {
@@ -1014,9 +1033,10 @@ void amdgpu_bo_unpin(struct amdgpu_bo *bo)
 	} else if (bo->tbo.resource->mem_type == TTM_PL_TT) {
 		atomic64_sub(amdgpu_bo_size(bo), &adev->gart_pin_size);
 	}
+
 }
 
-static const char *amdgpu_vram_names[] = {
+static const char * const amdgpu_vram_names[] = {
 	"UNKNOWN",
 	"GDDR1",
 	"DDR2",
@@ -1029,7 +1049,8 @@ static const char *amdgpu_vram_names[] = {
 	"GDDR6",
 	"DDR5",
 	"LPDDR4",
-	"LPDDR5"
+	"LPDDR5",
+	"HBM3E"
 };
 
 /**
@@ -1044,7 +1065,7 @@ static const char *amdgpu_vram_names[] = {
 int amdgpu_bo_init(struct amdgpu_device *adev)
 {
 	/* On A+A platform, VRAM can be mapped as WB */
-	if (!adev->gmc.xgmi.connected_to_cpu) {
+	if (!adev->gmc.xgmi.connected_to_cpu && !adev->gmc.is_app_apu) {
 		/* reserve PAT memory space to WC for VRAM */
 		int r = arch_io_reserve_memtype_wc(adev->gmc.aper_base,
 				adev->gmc.aper_size);
@@ -1080,8 +1101,7 @@ void amdgpu_bo_fini(struct amdgpu_device *adev)
 	amdgpu_ttm_fini(adev);
 
 	if (drm_dev_enter(adev_to_drm(adev), &idx)) {
-
-		if (!adev->gmc.xgmi.connected_to_cpu) {
+		if (!adev->gmc.xgmi.connected_to_cpu && !adev->gmc.is_app_apu) {
 			arch_phys_wc_del(adev->gmc.vram_mtrr);
 			arch_io_free_memtype_wc(adev->gmc.aper_base, adev->gmc.aper_size);
 		}
@@ -1148,8 +1168,8 @@ void amdgpu_bo_get_tiling_flags(struct amdgpu_bo *bo, u64 *tiling_flags)
  * Returns:
  * 0 for success or a negative error code on failure.
  */
-int amdgpu_bo_set_metadata (struct amdgpu_bo *bo, void *metadata,
-			    uint32_t metadata_size, uint64_t flags)
+int amdgpu_bo_set_metadata(struct amdgpu_bo *bo, void *metadata,
+			   u32 metadata_size, uint64_t flags)
 {
 	struct amdgpu_bo_user *ubo;
 	void *buffer;
@@ -1227,7 +1247,7 @@ int amdgpu_bo_get_metadata(struct amdgpu_bo *bo, void *buffer,
  * amdgpu_bo_move_notify - notification about a memory move
  * @bo: pointer to a buffer object
  * @evict: if this move is evicting the buffer from the graphics address space
- * @new_mem: new information of the bufer object
+ * @new_mem: new resource for backing the BO
  *
  * Marks the corresponding &amdgpu_bo buffer object as invalid, also performs
  * bookkeeping.
@@ -1237,69 +1257,24 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo,
 			   bool evict,
 			   struct ttm_resource *new_mem)
 {
-	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev);
-	struct amdgpu_bo *abo;
 	struct ttm_resource *old_mem = bo->resource;
+	struct amdgpu_bo *abo;
 
 	if (!amdgpu_bo_is_amdgpu_bo(bo))
 		return;
 
 	abo = ttm_to_amdgpu_bo(bo);
-	amdgpu_vm_bo_invalidate(adev, abo, evict);
+	amdgpu_vm_bo_move(abo, new_mem, evict);
 
 	amdgpu_bo_kunmap(abo);
 
-	if (abo->tbo.base.dma_buf && !abo->tbo.base.import_attach &&
-	    bo->resource->mem_type != TTM_PL_SYSTEM)
+	if (abo->tbo.base.dma_buf && !drm_gem_is_imported(&abo->tbo.base) &&
+	    old_mem && old_mem->mem_type != TTM_PL_SYSTEM)
 		dma_buf_move_notify(abo->tbo.base.dma_buf);
 
-	/* remember the eviction */
-	if (evict)
-		atomic64_inc(&adev->num_evictions);
-
-	/* update statistics */
-	if (!new_mem)
-		return;
-
 	/* move_notify is called before move happens */
-	trace_amdgpu_bo_move(abo, new_mem->mem_type, old_mem->mem_type);
-}
-
-void amdgpu_bo_get_memory(struct amdgpu_bo *bo,
-			  struct amdgpu_mem_stats *stats)
-{
-	unsigned int domain;
-	uint64_t size = amdgpu_bo_size(bo);
-
-	domain = amdgpu_mem_type_to_domain(bo->tbo.resource->mem_type);
-	switch (domain) {
-	case AMDGPU_GEM_DOMAIN_VRAM:
-		stats->vram += size;
-		if (amdgpu_bo_in_cpu_visible_vram(bo))
-			stats->visible_vram += size;
-		break;
-	case AMDGPU_GEM_DOMAIN_GTT:
-		stats->gtt += size;
-		break;
-	case AMDGPU_GEM_DOMAIN_CPU:
-	default:
-		stats->cpu += size;
-		break;
-	}
-
-	if (bo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM) {
-		stats->requested_vram += size;
-		if (bo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)
-			stats->requested_visible_vram += size;
-
-		if (domain != AMDGPU_GEM_DOMAIN_VRAM) {
-			stats->evicted_vram += size;
-			if (bo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)
-				stats->evicted_visible_vram += size;
-		}
-	} else if (bo->preferred_domains & AMDGPU_GEM_DOMAIN_GTT) {
-		stats->requested_gtt += size;
-	}
+	trace_amdgpu_bo_move(abo, new_mem ? new_mem->mem_type : -1,
+			     old_mem ? old_mem->mem_type : -1);
 }
 
 /**
@@ -1321,30 +1296,42 @@ void amdgpu_bo_release_notify(struct ttm_buffer_object *bo)
 
 	abo = ttm_to_amdgpu_bo(bo);
 
+	WARN_ON(abo->vm_bo);
+
 	if (abo->kfd_bo)
 		amdgpu_amdkfd_release_notify(abo);
 
-	/* We only remove the fence if the resv has individualized. */
-	WARN_ON_ONCE(bo->type == ttm_bo_type_kernel
-			&& bo->base.resv != &bo->base._resv);
-	if (bo->base.resv == &bo->base._resv)
-		amdgpu_amdkfd_remove_fence_on_pt_pd_bos(abo);
+	/*
+	 * We lock the private dma_resv object here and since the BO is about to
+	 * be released nobody else should have a pointer to it.
+	 * So when this locking here fails something is wrong with the reference
+	 * counting.
+	 */
+	if (WARN_ON_ONCE(!dma_resv_trylock(&bo->base._resv)))
+		return;
+
+	amdgpu_amdkfd_remove_all_eviction_fences(abo);
 
 	if (!bo->resource || bo->resource->mem_type != TTM_PL_VRAM ||
 	    !(abo->flags & AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE) ||
 	    adev->in_suspend || drm_dev_is_unplugged(adev_to_drm(adev)))
-		return;
+		goto out;
 
-	if (WARN_ON_ONCE(!dma_resv_trylock(bo->base.resv)))
-		return;
+	r = dma_resv_reserve_fences(&bo->base._resv, 1);
+	if (r)
+		goto out;
 
-	r = amdgpu_fill_buffer(abo, AMDGPU_POISON, bo->base.resv, &fence);
-	if (!WARN_ON(r)) {
-		amdgpu_bo_fence(abo, fence, false);
-		dma_fence_put(fence);
-	}
+	r = amdgpu_fill_buffer(abo, 0, &bo->base._resv, &fence, true,
+			       AMDGPU_KERNEL_JOB_ID_CLEAR_ON_RELEASE);
+	if (WARN_ON(r))
+		goto out;
+
+	amdgpu_vram_mgr_set_cleared(bo->resource);
+	dma_resv_add_fence(&bo->base._resv, fence, DMA_RESV_USAGE_KERNEL);
+	dma_fence_put(fence);
 
-	dma_resv_unlock(bo->base.resv);
+out:
+	dma_resv_unlock(&bo->base._resv);
 }
 
 /**
@@ -1368,10 +1355,7 @@ vm_fault_t amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo)
 	/* Remember that this BO was accessed by the CPU */
 	abo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
 
-	if (bo->resource->mem_type != TTM_PL_VRAM)
-		return 0;
-
-	if (amdgpu_bo_in_cpu_visible_vram(abo))
+	if (amdgpu_res_cpu_visible(adev, bo->resource))
 		return 0;
 
 	/* Can't move a pinned BO to visible VRAM */
@@ -1384,8 +1368,7 @@ vm_fault_t amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo)
 					AMDGPU_GEM_DOMAIN_GTT);
 
 	/* Avoid costly evictions; only set GTT as a busy placement */
-	abo->placement.num_busy_placement = 1;
-	abo->placement.busy_placement = &abo->placements[1];
+	abo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
 
 	r = ttm_bo_validate(bo, &abo->placement, &ctx);
 	if (unlikely(r == -EBUSY || r == -ERESTARTSYS))
@@ -1395,7 +1378,7 @@ vm_fault_t amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo)
 
 	/* this should never happen */
 	if (bo->resource->mem_type == TTM_PL_VRAM &&
-	    !amdgpu_bo_in_cpu_visible_vram(abo))
+	    !amdgpu_res_cpu_visible(adev, bo->resource))
 		return VM_FAULT_SIGBUS;
 
 	ttm_bo_move_to_lru_tail_unlocked(bo);
@@ -1496,6 +1479,26 @@ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo)
 }
 
 /**
+ * amdgpu_bo_fb_aper_addr - return FB aperture GPU offset of the VRAM bo
+ * @bo:	amdgpu VRAM buffer object for which we query the offset
+ *
+ * Returns:
+ * current FB aperture GPU offset of the object.
+ */
+u64 amdgpu_bo_fb_aper_addr(struct amdgpu_bo *bo)
+{
+	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
+	uint64_t offset, fb_base;
+
+	WARN_ON_ONCE(bo->tbo.resource->mem_type != TTM_PL_VRAM);
+
+	fb_base = adev->gmc.fb_start;
+	fb_base += adev->gmc.xgmi.physical_node_id * adev->gmc.xgmi.node_segment_size;
+	offset = (bo->tbo.resource->start << PAGE_SHIFT) + fb_base;
+	return amdgpu_gmc_sign_extend(offset);
+}
+
+/**
  * amdgpu_bo_gpu_offset_no_check - return GPU offset of bo
  * @bo:	amdgpu object for which we query the offset
  *
@@ -1505,15 +1508,60 @@ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo)
 u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
-	uint64_t offset;
+	uint64_t offset = AMDGPU_BO_INVALID_OFFSET;
 
-	offset = (bo->tbo.resource->start << PAGE_SHIFT) +
-		 amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
+	if (bo->tbo.resource->mem_type == TTM_PL_TT)
+		offset = amdgpu_gmc_agp_addr(&bo->tbo);
+
+	if (offset == AMDGPU_BO_INVALID_OFFSET)
+		offset = (bo->tbo.resource->start << PAGE_SHIFT) +
+			amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
 
 	return amdgpu_gmc_sign_extend(offset);
 }
 
 /**
+ * amdgpu_bo_mem_stats_placement - bo placement for memory accounting
+ * @bo:	the buffer object we should look at
+ *
+ * BO can have multiple preferred placements, to avoid double counting we want
+ * to file it under a single placement for memory stats.
+ * Luckily, if we take the highest set bit in preferred_domains the result is
+ * quite sensible.
+ *
+ * Returns:
+ * Which of the placements should the BO be accounted under.
+ */
+uint32_t amdgpu_bo_mem_stats_placement(struct amdgpu_bo *bo)
+{
+	uint32_t domain = bo->preferred_domains & AMDGPU_GEM_DOMAIN_MASK;
+
+	if (!domain)
+		return TTM_PL_SYSTEM;
+
+	switch (rounddown_pow_of_two(domain)) {
+	case AMDGPU_GEM_DOMAIN_CPU:
+		return TTM_PL_SYSTEM;
+	case AMDGPU_GEM_DOMAIN_GTT:
+		return TTM_PL_TT;
+	case AMDGPU_GEM_DOMAIN_VRAM:
+		return TTM_PL_VRAM;
+	case AMDGPU_GEM_DOMAIN_GDS:
+		return AMDGPU_PL_GDS;
+	case AMDGPU_GEM_DOMAIN_GWS:
+		return AMDGPU_PL_GWS;
+	case AMDGPU_GEM_DOMAIN_OA:
+		return AMDGPU_PL_OA;
+	case AMDGPU_GEM_DOMAIN_DOORBELL:
+		return AMDGPU_PL_DOORBELL;
+	case AMDGPU_GEM_DOMAIN_MMIO_REMAP:
+		return AMDGPU_PL_MMIO_REMAP;
+	default:
+		return TTM_PL_SYSTEM;
+	}
+}
+
+/**
  * amdgpu_bo_get_preferred_domain - get preferred domain
  * @adev: amdgpu device object
  * @domain: allowed :ref:`memory domains <amdgpu_memory_domains>`
@@ -1555,25 +1603,54 @@ uint32_t amdgpu_bo_get_preferred_domain(struct amdgpu_device *adev,
  */
 u64 amdgpu_bo_print_info(int id, struct amdgpu_bo *bo, struct seq_file *m)
 {
+	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
 	struct dma_buf_attachment *attachment;
 	struct dma_buf *dma_buf;
-	unsigned int domain;
 	const char *placement;
 	unsigned int pin_count;
 	u64 size;
 
-	domain = amdgpu_mem_type_to_domain(bo->tbo.resource->mem_type);
-	switch (domain) {
-	case AMDGPU_GEM_DOMAIN_VRAM:
-		placement = "VRAM";
-		break;
-	case AMDGPU_GEM_DOMAIN_GTT:
-		placement = " GTT";
-		break;
-	case AMDGPU_GEM_DOMAIN_CPU:
-	default:
-		placement = " CPU";
-		break;
+	if (dma_resv_trylock(bo->tbo.base.resv)) {
+		if (!bo->tbo.resource) {
+			placement = "NONE";
+		} else {
+			switch (bo->tbo.resource->mem_type) {
+			case TTM_PL_VRAM:
+				if (amdgpu_res_cpu_visible(adev, bo->tbo.resource))
+					placement = "VRAM VISIBLE";
+				else
+					placement = "VRAM";
+				break;
+			case TTM_PL_TT:
+				placement = "GTT";
+				break;
+			case AMDGPU_PL_GDS:
+				placement = "GDS";
+				break;
+			case AMDGPU_PL_GWS:
+				placement = "GWS";
+				break;
+			case AMDGPU_PL_OA:
+				placement = "OA";
+				break;
+			case AMDGPU_PL_PREEMPT:
+				placement = "PREEMPTIBLE";
+				break;
+			case AMDGPU_PL_DOORBELL:
+				placement = "DOORBELL";
+				break;
+			case AMDGPU_PL_MMIO_REMAP:
+				placement = "MMIO REMAP";
+				break;
+			case TTM_PL_SYSTEM:
+			default:
+				placement = "CPU";
+				break;
+			}
+		}
+		dma_resv_unlock(bo->tbo.base.resv);
+	} else {
+		placement = "UNKNOWN";
 	}
 
 	size = amdgpu_bo_size(bo);
@@ -1599,7 +1676,11 @@ u64 amdgpu_bo_print_info(int id, struct amdgpu_bo *bo, struct seq_file *m)
 	amdgpu_bo_print_flag(m, bo, VRAM_CONTIGUOUS);
 	amdgpu_bo_print_flag(m, bo, VM_ALWAYS_VALID);
 	amdgpu_bo_print_flag(m, bo, EXPLICIT_SYNC);
-
+	/* Add the gem obj resv fence dump*/
+	if (dma_resv_trylock(bo->tbo.base.resv)) {
+		dma_resv_describe(bo->tbo.base.resv, m);
+		dma_resv_unlock(bo->tbo.base.resv);
+	}
 	seq_puts(m, "\n");
 
 	return size;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
index 35b8106816a1..52c2d1731aab 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
@@ -56,6 +56,8 @@ struct amdgpu_bo_param {
 	bool				no_wait_gpu;
 	struct dma_resv			*resv;
 	void				(*destroy)(struct ttm_buffer_object *bo);
+	/* xcp partition number plus 1, 0 means any partition */
+	int8_t				xcp_id_plus1;
 };
 
 /* bo virtual addresses in a vm */
@@ -67,7 +69,7 @@ struct amdgpu_bo_va_mapping {
 	uint64_t			last;
 	uint64_t			__subtree_last;
 	uint64_t			offset;
-	uint64_t			flags;
+	uint32_t			flags;
 };
 
 /* User space allocated BO in a VM */
@@ -88,6 +90,13 @@ struct amdgpu_bo_va {
 	bool				cleared;
 
 	bool				is_xgmi;
+
+	/*
+	 * protected by vm reservation lock
+	 * if non-zero, cannot unmap from GPU because user queues may still access it
+	 */
+	unsigned int			queue_refcount;
+	atomic_t			userq_va_mapped;
 };
 
 struct amdgpu_bo {
@@ -108,6 +117,13 @@ struct amdgpu_bo {
 	struct mmu_interval_notifier	notifier;
 #endif
 	struct kgd_mem                  *kfd_bo;
+
+	/*
+	 * For GPUs with spatial partitioning, xcp partition number, -1 means
+	 * any partition. For other ASICs without spatial partition, always 0
+	 * for memory accounting.
+	 */
+	int8_t				xcp_id;
 };
 
 struct amdgpu_bo_user {
@@ -121,32 +137,9 @@ struct amdgpu_bo_user {
 
 struct amdgpu_bo_vm {
 	struct amdgpu_bo		bo;
-	struct amdgpu_bo		*shadow;
-	struct list_head		shadow_list;
 	struct amdgpu_vm_bo_base        entries[];
 };
 
-struct amdgpu_mem_stats {
-	/* current VRAM usage, includes visible VRAM */
-	uint64_t vram;
-	/* current visible VRAM usage */
-	uint64_t visible_vram;
-	/* current GTT usage */
-	uint64_t gtt;
-	/* current system memory usage */
-	uint64_t cpu;
-	/* sum of evicted buffers, includes visible VRAM */
-	uint64_t evicted_vram;
-	/* sum of evicted buffers due to CPU access */
-	uint64_t evicted_visible_vram;
-	/* how much userspace asked for, includes vis.VRAM */
-	uint64_t requested_vram;
-	/* how much userspace asked for */
-	uint64_t requested_visible_vram;
-	/* how much userspace asked for */
-	uint64_t requested_gtt;
-};
-
 static inline struct amdgpu_bo *ttm_to_amdgpu_bo(struct ttm_buffer_object *tbo)
 {
 	return container_of(tbo, struct amdgpu_bo, tbo);
@@ -173,6 +166,10 @@ static inline unsigned amdgpu_mem_type_to_domain(u32 mem_type)
 		return AMDGPU_GEM_DOMAIN_GWS;
 	case AMDGPU_PL_OA:
 		return AMDGPU_GEM_DOMAIN_OA;
+	case AMDGPU_PL_DOORBELL:
+		return AMDGPU_GEM_DOMAIN_DOORBELL;
+	case AMDGPU_PL_MMIO_REMAP:
+		return AMDGPU_GEM_DOMAIN_MMIO_REMAP;
 	default:
 		break;
 	}
@@ -234,28 +231,6 @@ static inline u64 amdgpu_bo_mmap_offset(struct amdgpu_bo *bo)
 }
 
 /**
- * amdgpu_bo_in_cpu_visible_vram - check if BO is (partly) in visible VRAM
- */
-static inline bool amdgpu_bo_in_cpu_visible_vram(struct amdgpu_bo *bo)
-{
-	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
-	struct amdgpu_res_cursor cursor;
-
-	if (bo->tbo.resource->mem_type != TTM_PL_VRAM)
-		return false;
-
-	amdgpu_res_first(bo->tbo.resource, 0, amdgpu_bo_size(bo), &cursor);
-	while (cursor.remaining) {
-		if (cursor.start < adev->gmc.visible_vram_size)
-			return true;
-
-		amdgpu_res_next(&cursor, cursor.size);
-	}
-
-	return false;
-}
-
-/**
  * amdgpu_bo_explicit_sync - return whether the bo is explicitly synced
  */
 static inline bool amdgpu_bo_explicit_sync(struct amdgpu_bo *bo)
@@ -274,22 +249,6 @@ static inline bool amdgpu_bo_encrypted(struct amdgpu_bo *bo)
 	return bo->flags & AMDGPU_GEM_CREATE_ENCRYPTED;
 }
 
-/**
- * amdgpu_bo_shadowed - check if the BO is shadowed
- *
- * @bo: BO to be tested.
- *
- * Returns:
- * NULL if not shadowed or else return a BO pointer.
- */
-static inline struct amdgpu_bo *amdgpu_bo_shadowed(struct amdgpu_bo *bo)
-{
-	if (bo->tbo.type == ttm_bo_type_kernel)
-		return to_amdgpu_bo_vm(bo)->shadow;
-
-	return NULL;
-}
-
 bool amdgpu_bo_is_amdgpu_bo(struct ttm_buffer_object *bo);
 void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain);
 
@@ -304,6 +263,10 @@ int amdgpu_bo_create_kernel(struct amdgpu_device *adev,
 			    unsigned long size, int align,
 			    u32 domain, struct amdgpu_bo **bo_ptr,
 			    u64 *gpu_addr, void **cpu_addr);
+int amdgpu_bo_create_isp_user(struct amdgpu_device *adev,
+			   struct dma_buf *dbuf, u32 domain,
+			   struct amdgpu_bo **bo,
+			   u64 *gpu_addr);
 int amdgpu_bo_create_kernel_at(struct amdgpu_device *adev,
 			       uint64_t offset, uint64_t size,
 			       struct amdgpu_bo **bo_ptr, void **cpu_addr);
@@ -315,14 +278,13 @@ int amdgpu_bo_create_vm(struct amdgpu_device *adev,
 			struct amdgpu_bo_vm **ubo_ptr);
 void amdgpu_bo_free_kernel(struct amdgpu_bo **bo, u64 *gpu_addr,
 			   void **cpu_addr);
+void amdgpu_bo_free_isp_user(struct amdgpu_bo *bo);
 int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr);
 void *amdgpu_bo_kptr(struct amdgpu_bo *bo);
 void amdgpu_bo_kunmap(struct amdgpu_bo *bo);
 struct amdgpu_bo *amdgpu_bo_ref(struct amdgpu_bo *bo);
 void amdgpu_bo_unref(struct amdgpu_bo **bo);
 int amdgpu_bo_pin(struct amdgpu_bo *bo, u32 domain);
-int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
-			     u64 min_offset, u64 max_offset);
 void amdgpu_bo_unpin(struct amdgpu_bo *bo);
 int amdgpu_bo_init(struct amdgpu_device *adev);
 void amdgpu_bo_fini(struct amdgpu_device *adev);
@@ -345,12 +307,9 @@ int amdgpu_bo_sync_wait_resv(struct amdgpu_device *adev, struct dma_resv *resv,
 			     bool intr);
 int amdgpu_bo_sync_wait(struct amdgpu_bo *bo, void *owner, bool intr);
 u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo);
+u64 amdgpu_bo_fb_aper_addr(struct amdgpu_bo *bo);
 u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo);
-void amdgpu_bo_get_memory(struct amdgpu_bo *bo,
-			  struct amdgpu_mem_stats *stats);
-void amdgpu_bo_add_to_shadow_list(struct amdgpu_bo_vm *vmbo);
-int amdgpu_bo_restore_shadow(struct amdgpu_bo *shadow,
-			     struct dma_fence **fence);
+uint32_t amdgpu_bo_mem_stats_placement(struct amdgpu_bo *bo);
 uint32_t amdgpu_bo_get_preferred_domain(struct amdgpu_device *adev,
 					    uint32_t domain);
 
@@ -385,8 +344,7 @@ int amdgpu_sa_bo_manager_start(struct amdgpu_device *adev,
 int amdgpu_sa_bo_new(struct amdgpu_sa_manager *sa_manager,
 		     struct drm_suballoc **sa_bo,
 		     unsigned int size);
-void amdgpu_sa_bo_free(struct amdgpu_device *adev,
-		       struct drm_suballoc **sa_bo,
+void amdgpu_sa_bo_free(struct drm_suballoc **sa_bo,
 		       struct dma_fence *fence);
 #if defined(CONFIG_DEBUG_FS)
 void amdgpu_sa_bo_dump_debug_info(struct amdgpu_sa_manager *sa_manager,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pll.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pll.c
index 0bb2466d539a..675aa138ea11 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pll.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pll.c
@@ -94,7 +94,7 @@ static void amdgpu_pll_get_fb_ref_div(struct amdgpu_device *adev, unsigned int n
 		ref_div_max = min(128 / post_div, ref_div_max);
 
 	/* get matching reference and feedback divider */
-	*ref_div = min(max(DIV_ROUND_CLOSEST(den, post_div), 1u), ref_div_max);
+	*ref_div = clamp(DIV_ROUND_CLOSEST(den, post_div), 1u, ref_div_max);
 	*fb_div = DIV_ROUND_CLOSEST(nom * *ref_div * post_div, den);
 
 	/* limit fb divider to its maximum */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c
index 71ee361d0972..6e91ea1de5aa 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c
@@ -276,9 +276,8 @@ static void amdgpu_perf_read(struct perf_event *event)
 	    (!pe->adev->df.funcs->pmc_get_count))
 		return;
 
+	prev = local64_read(&hwc->prev_count);
 	do {
-		prev = local64_read(&hwc->prev_count);
-
 		switch (hwc->config_base) {
 		case AMDGPU_PMU_EVENT_CONFIG_TYPE_DF:
 		case AMDGPU_PMU_EVENT_CONFIG_TYPE_XGMI:
@@ -289,7 +288,7 @@ static void amdgpu_perf_read(struct perf_event *event)
 			count = 0;
 			break;
 		}
-	} while (local64_cmpxchg(&hwc->prev_count, prev, count) != prev);
+	} while (!local64_try_cmpxchg(&hwc->prev_count, &prev, count));
 
 	local64_add(count - prev, &event->count);
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_preempt_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_preempt_mgr.c
index e8adfd0a570a..34b5e22b44e5 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_preempt_mgr.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_preempt_mgr.c
@@ -137,7 +137,8 @@ void amdgpu_preempt_mgr_fini(struct amdgpu_device *adev)
 	if (ret)
 		return;
 
-	device_remove_file(adev->dev, &dev_attr_mem_info_preempt_used);
+	if (adev->dev->kobj.sd)
+		device_remove_file(adev->dev, &dev_attr_mem_info_preempt_used);
 
 	ttm_resource_manager_cleanup(man);
 	ttm_set_driver_manager(&adev->mman.bdev, AMDGPU_PL_PREEMPT, NULL);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
index 9d7e6e0e73ed..0b10497d487c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
@@ -38,15 +38,13 @@
 #include "psp_v12_0.h"
 #include "psp_v13_0.h"
 #include "psp_v13_0_4.h"
+#include "psp_v14_0.h"
 
 #include "amdgpu_ras.h"
 #include "amdgpu_securedisplay.h"
 #include "amdgpu_atomfirmware.h"
 
-#define AMD_VBIOS_FILE_MAX_SIZE_B      (1024*1024*3)
-
-static int psp_sysfs_init(struct amdgpu_device *adev);
-static void psp_sysfs_fini(struct amdgpu_device *adev);
+#define AMD_VBIOS_FILE_MAX_SIZE_B      (1024*1024*16)
 
 static int psp_load_smu_fw(struct psp_context *psp);
 static int psp_rap_terminate(struct psp_context *psp);
@@ -103,7 +101,7 @@ static void psp_check_pmfw_centralized_cstate_management(struct psp_context *psp
 		return;
 	}
 
-	switch (adev->ip_versions[MP0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 	case IP_VERSION(11, 0, 0):
 	case IP_VERSION(11, 0, 4):
 	case IP_VERSION(11, 0, 5):
@@ -131,7 +129,7 @@ static int psp_init_sriov_microcode(struct psp_context *psp)
 
 	amdgpu_ucode_ip_version_decode(adev, MP0_HWIP, ucode_prefix, sizeof(ucode_prefix));
 
-	switch (adev->ip_versions[MP0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 	case IP_VERSION(9, 0, 0):
 	case IP_VERSION(11, 0, 7):
 	case IP_VERSION(11, 0, 9):
@@ -146,78 +144,120 @@ static int psp_init_sriov_microcode(struct psp_context *psp)
 	case IP_VERSION(13, 0, 0):
 		adev->virt.autoload_ucode_id = 0;
 		break;
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 14):
+		ret = psp_init_cap_microcode(psp, ucode_prefix);
+		ret &= psp_init_ta_microcode(psp, ucode_prefix);
+		break;
 	case IP_VERSION(13, 0, 10):
 		adev->virt.autoload_ucode_id = AMDGPU_UCODE_ID_CP_MES1_DATA;
 		ret = psp_init_cap_microcode(psp, ucode_prefix);
 		break;
+	case IP_VERSION(13, 0, 12):
+		ret = psp_init_ta_microcode(psp, ucode_prefix);
+		break;
 	default:
 		return -EINVAL;
 	}
 	return ret;
 }
 
-static int psp_early_init(void *handle)
+static int psp_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct psp_context *psp = &adev->psp;
 
-	switch (adev->ip_versions[MP0_HWIP][0]) {
+	psp->autoload_supported = true;
+	psp->boot_time_tmr = true;
+
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 	case IP_VERSION(9, 0, 0):
 		psp_v3_1_set_psp_funcs(psp);
 		psp->autoload_supported = false;
+		psp->boot_time_tmr = false;
 		break;
 	case IP_VERSION(10, 0, 0):
 	case IP_VERSION(10, 0, 1):
 		psp_v10_0_set_psp_funcs(psp);
 		psp->autoload_supported = false;
+		psp->boot_time_tmr = false;
 		break;
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 4):
 		psp_v11_0_set_psp_funcs(psp);
 		psp->autoload_supported = false;
+		psp->boot_time_tmr = false;
 		break;
 	case IP_VERSION(11, 0, 0):
+	case IP_VERSION(11, 0, 7):
+		adev->psp.sup_pd_fw_up = !amdgpu_sriov_vf(adev);
+		fallthrough;
 	case IP_VERSION(11, 0, 5):
 	case IP_VERSION(11, 0, 9):
-	case IP_VERSION(11, 0, 7):
 	case IP_VERSION(11, 0, 11):
 	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 2):
 	case IP_VERSION(11, 0, 12):
 	case IP_VERSION(11, 0, 13):
 		psp_v11_0_set_psp_funcs(psp);
-		psp->autoload_supported = true;
+		psp->boot_time_tmr = false;
 		break;
 	case IP_VERSION(11, 0, 3):
 	case IP_VERSION(12, 0, 1):
 		psp_v12_0_set_psp_funcs(psp);
+		psp->autoload_supported = false;
+		psp->boot_time_tmr = false;
 		break;
 	case IP_VERSION(13, 0, 2):
+		psp->boot_time_tmr = false;
+		fallthrough;
 	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 14):
+		psp_v13_0_set_psp_funcs(psp);
+		psp->autoload_supported = false;
+		break;
+	case IP_VERSION(13, 0, 12):
 		psp_v13_0_set_psp_funcs(psp);
+		psp->autoload_supported = false;
+		adev->psp.sup_ifwi_up = !amdgpu_sriov_vf(adev);
 		break;
 	case IP_VERSION(13, 0, 1):
 	case IP_VERSION(13, 0, 3):
 	case IP_VERSION(13, 0, 5):
 	case IP_VERSION(13, 0, 8):
-	case IP_VERSION(13, 0, 10):
 	case IP_VERSION(13, 0, 11):
+	case IP_VERSION(14, 0, 0):
+	case IP_VERSION(14, 0, 1):
+	case IP_VERSION(14, 0, 4):
 		psp_v13_0_set_psp_funcs(psp);
-		psp->autoload_supported = true;
+		psp->boot_time_tmr = false;
 		break;
 	case IP_VERSION(11, 0, 8):
 		if (adev->apu_flags & AMD_APU_IS_CYAN_SKILLFISH2) {
 			psp_v11_0_8_set_psp_funcs(psp);
-			psp->autoload_supported = false;
 		}
+		psp->autoload_supported = false;
+		psp->boot_time_tmr = false;
 		break;
 	case IP_VERSION(13, 0, 0):
 	case IP_VERSION(13, 0, 7):
+	case IP_VERSION(13, 0, 10):
 		psp_v13_0_set_psp_funcs(psp);
-		psp->autoload_supported = true;
+		adev->psp.sup_ifwi_up = !amdgpu_sriov_vf(adev);
+		psp->boot_time_tmr = false;
 		break;
 	case IP_VERSION(13, 0, 4):
 		psp_v13_0_4_set_psp_funcs(psp);
-		psp->autoload_supported = true;
+		psp->boot_time_tmr = false;
+		break;
+	case IP_VERSION(14, 0, 2):
+	case IP_VERSION(14, 0, 3):
+		adev->psp.sup_ifwi_up = !amdgpu_sriov_vf(adev);
+		psp_v14_0_set_psp_funcs(psp);
+		break;
+	case IP_VERSION(14, 0, 5):
+		psp_v14_0_set_psp_funcs(psp);
+		psp->boot_time_tmr = false;
 		break;
 	default:
 		return -EINVAL;
@@ -225,6 +265,8 @@ static int psp_early_init(void *handle)
 
 	psp->adev = adev;
 
+	adev->psp_timeout = 20000;
+
 	psp_check_pmfw_centralized_cstate_management(psp);
 
 	if (amdgpu_sriov_vf(adev))
@@ -286,21 +328,22 @@ static int psp_memory_training_init(struct psp_context *psp)
 	struct psp_memory_training_context *ctx = &psp->mem_train_ctx;
 
 	if (ctx->init != PSP_MEM_TRAIN_RESERVE_SUCCESS) {
-		DRM_DEBUG("memory training is not supported!\n");
+		dev_dbg(psp->adev->dev, "memory training is not supported!\n");
 		return 0;
 	}
 
 	ctx->sys_cache = kzalloc(ctx->train_data_size, GFP_KERNEL);
 	if (ctx->sys_cache == NULL) {
-		DRM_ERROR("alloc mem_train_ctx.sys_cache failed!\n");
+		dev_err(psp->adev->dev, "alloc mem_train_ctx.sys_cache failed!\n");
 		ret = -ENOMEM;
 		goto Err_out;
 	}
 
-	DRM_DEBUG("train_data_size:%llx,p2c_train_data_offset:%llx,c2p_train_data_offset:%llx.\n",
-		  ctx->train_data_size,
-		  ctx->p2c_train_data_offset,
-		  ctx->c2p_train_data_offset);
+	dev_dbg(psp->adev->dev,
+		"train_data_size:%llx,p2c_train_data_offset:%llx,c2p_train_data_offset:%llx.\n",
+		ctx->train_data_size,
+		ctx->p2c_train_data_offset,
+		ctx->c2p_train_data_offset);
 	ctx->init = PSP_MEM_TRAIN_INIT_SUCCESS;
 	return 0;
 
@@ -329,6 +372,11 @@ static bool psp_get_runtime_db_entry(struct amdgpu_device *adev,
 	bool ret = false;
 	int i;
 
+	if (amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 6) ||
+	    amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 12) ||
+	    amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 14))
+		return false;
+
 	db_header_pos = adev->gmc.mc_vram_size - PSP_RUNTIME_DB_OFFSET;
 	db_dir_pos = db_header_pos + sizeof(struct psp_runtime_data_header);
 
@@ -388,9 +436,9 @@ static bool psp_get_runtime_db_entry(struct amdgpu_device *adev,
 	return ret;
 }
 
-static int psp_sw_init(void *handle)
+static int psp_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct psp_context *psp = &adev->psp;
 	int ret;
 	struct psp_runtime_boot_cfg_entry boot_cfg_entry;
@@ -399,19 +447,19 @@ static int psp_sw_init(void *handle)
 
 	psp->cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL);
 	if (!psp->cmd) {
-		DRM_ERROR("Failed to allocate memory to command buffer!\n");
-		ret = -ENOMEM;
+		dev_err(adev->dev, "Failed to allocate memory to command buffer!\n");
+		return -ENOMEM;
 	}
 
 	adev->psp.xgmi_context.supports_extended_data =
 		!adev->gmc.xgmi.connected_to_cpu &&
-			adev->ip_versions[MP0_HWIP][0] == IP_VERSION(13, 0, 2);
+		amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 2);
 
 	memset(&scpm_entry, 0, sizeof(scpm_entry));
 	if ((psp_get_runtime_db_entry(adev,
 				PSP_RUNTIME_ENTRY_TYPE_PPTABLE_ERR_STATUS,
 				&scpm_entry)) &&
-	    (SCPM_DISABLE != scpm_entry.scpm_status)) {
+	    (scpm_entry.scpm_status != SCPM_DISABLE)) {
 		adev->scpm_enabled = true;
 		adev->scpm_status = scpm_entry.scpm_status;
 	} else {
@@ -431,41 +479,34 @@ static int psp_sw_init(void *handle)
 			/* If psp runtime database exists, then
 			 * only enable two stage memory training
 			 * when TWO_STAGE_DRAM_TRAINING bit is set
-			 * in runtime database */
+			 * in runtime database
+			 */
 			mem_training_ctx->enable_mem_training = true;
 		}
 
 	} else {
-		/* If psp runtime database doesn't exist or
-		 * is invalid, force enable two stage memory
-		 * training */
+		/* If psp runtime database doesn't exist or is
+		 * invalid, force enable two stage memory training
+		 */
 		mem_training_ctx->enable_mem_training = true;
 	}
 
 	if (mem_training_ctx->enable_mem_training) {
 		ret = psp_memory_training_init(psp);
 		if (ret) {
-			DRM_ERROR("Failed to initialize memory training!\n");
+			dev_err(adev->dev, "Failed to initialize memory training!\n");
 			return ret;
 		}
 
 		ret = psp_mem_training(psp, PSP_MEM_TRAIN_COLD_BOOT);
 		if (ret) {
-			DRM_ERROR("Failed to process memory training!\n");
-			return ret;
-		}
-	}
-
-	if (adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 0) ||
-	    adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 7)) {
-		ret= psp_sysfs_init(adev);
-		if (ret) {
+			dev_err(adev->dev, "Failed to process memory training!\n");
 			return ret;
 		}
 	}
 
 	ret = amdgpu_bo_create_kernel(adev, PSP_1_MEG, PSP_1_MEG,
-				      amdgpu_sriov_vf(adev) ?
+				      (amdgpu_sriov_vf(adev) || adev->debug_use_vram_fw_buf) ?
 				      AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT,
 				      &psp->fw_pri_bo,
 				      &psp->fw_pri_mc_addr,
@@ -474,7 +515,8 @@ static int psp_sw_init(void *handle)
 		return ret;
 
 	ret = amdgpu_bo_create_kernel(adev, PSP_FENCE_BUFFER_SIZE, PAGE_SIZE,
-				      AMDGPU_GEM_DOMAIN_VRAM,
+				      AMDGPU_GEM_DOMAIN_VRAM |
+				      AMDGPU_GEM_DOMAIN_GTT,
 				      &psp->fence_buf_bo,
 				      &psp->fence_buf_mc_addr,
 				      &psp->fence_buf);
@@ -482,7 +524,8 @@ static int psp_sw_init(void *handle)
 		goto failed1;
 
 	ret = amdgpu_bo_create_kernel(adev, PSP_CMD_BUFFER_SIZE, PAGE_SIZE,
-				      AMDGPU_GEM_DOMAIN_VRAM,
+				      AMDGPU_GEM_DOMAIN_VRAM |
+				      AMDGPU_GEM_DOMAIN_GTT,
 				      &psp->cmd_buf_bo, &psp->cmd_buf_mc_addr,
 				      (void **)&psp->cmd_buf_mem);
 	if (ret)
@@ -491,19 +534,18 @@ static int psp_sw_init(void *handle)
 	return 0;
 
 failed2:
-	amdgpu_bo_free_kernel(&psp->fw_pri_bo,
-			      &psp->fw_pri_mc_addr, &psp->fw_pri_buf);
-failed1:
 	amdgpu_bo_free_kernel(&psp->fence_buf_bo,
 			      &psp->fence_buf_mc_addr, &psp->fence_buf);
+failed1:
+	amdgpu_bo_free_kernel(&psp->fw_pri_bo,
+			      &psp->fw_pri_mc_addr, &psp->fw_pri_buf);
 	return ret;
 }
 
-static int psp_sw_fini(void *handle)
+static int psp_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct psp_context *psp = &adev->psp;
-	struct psp_gfx_cmd_resp *cmd = psp->cmd;
 
 	psp_memory_training_fini(psp);
 
@@ -513,12 +555,10 @@ static int psp_sw_fini(void *handle)
 	amdgpu_ucode_release(&psp->cap_fw);
 	amdgpu_ucode_release(&psp->toc_fw);
 
-	if (adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 0) ||
-	    adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 7))
-		psp_sysfs_fini(adev);
+	kfree(psp->cmd);
+	psp->cmd = NULL;
 
-	kfree(cmd);
-	cmd = NULL;
+	psp_free_shared_bufs(psp);
 
 	if (psp->km_ring.ring_mem)
 		amdgpu_bo_free_kernel(&adev->firmware.rbuf,
@@ -535,9 +575,11 @@ static int psp_sw_fini(void *handle)
 	return 0;
 }
 
-int psp_wait_for(struct psp_context *psp, uint32_t reg_index,
-		 uint32_t reg_val, uint32_t mask, bool check_changed)
+int psp_wait_for(struct psp_context *psp, uint32_t reg_index, uint32_t reg_val,
+		 uint32_t mask, uint32_t flags)
 {
+	bool check_changed = flags & PSP_WAITREG_CHANGED;
+	bool verbose = !(flags & PSP_WAITREG_NOVERBOSE);
 	uint32_t val;
 	int i;
 	struct amdgpu_device *adev = psp->adev;
@@ -557,6 +599,31 @@ int psp_wait_for(struct psp_context *psp, uint32_t reg_index,
 		udelay(1);
 	}
 
+	if (verbose)
+		dev_err(adev->dev,
+			"psp reg (0x%x) wait timed out, mask: %x, read: %x exp: %x",
+			reg_index, mask, val, reg_val);
+
+	return -ETIME;
+}
+
+int psp_wait_for_spirom_update(struct psp_context *psp, uint32_t reg_index,
+			       uint32_t reg_val, uint32_t mask, uint32_t msec_timeout)
+{
+	uint32_t val;
+	int i;
+	struct amdgpu_device *adev = psp->adev;
+
+	if (psp->adev->no_hw_access)
+		return 0;
+
+	for (i = 0; i < msec_timeout; i++) {
+		val = RREG32(reg_index);
+		if ((val & mask) == reg_val)
+			return 0;
+		msleep(1);
+	}
+
 	return -ETIME;
 }
 
@@ -593,11 +660,35 @@ static const char *psp_gfx_cmd_name(enum psp_gfx_cmd_id cmd_id)
 		return "AUTOLOAD_RLC";
 	case GFX_CMD_ID_BOOT_CFG:
 		return "BOOT_CFG";
+	case GFX_CMD_ID_CONFIG_SQ_PERFMON:
+		return "CONFIG_SQ_PERFMON";
+	case GFX_CMD_ID_FB_FW_RESERV_ADDR:
+		return "FB_FW_RESERV_ADDR";
+	case GFX_CMD_ID_FB_FW_RESERV_EXT_ADDR:
+		return "FB_FW_RESERV_EXT_ADDR";
+	case GFX_CMD_ID_SRIOV_SPATIAL_PART:
+		return "SPATIAL_PARTITION";
+	case GFX_CMD_ID_FB_NPS_MODE:
+		return "NPS_MODE_CHANGE";
 	default:
 		return "UNKNOWN CMD";
 	}
 }
 
+static bool psp_err_warn(struct psp_context *psp)
+{
+	struct psp_gfx_cmd_resp *cmd = psp->cmd_buf_mem;
+
+	/* This response indicates reg list is already loaded */
+	if (amdgpu_ip_version(psp->adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 2) &&
+	    cmd->cmd_id == GFX_CMD_ID_LOAD_IP_FW &&
+	    cmd->cmd.cmd_load_ip_fw.fw_type == GFX_FW_TYPE_REG_LIST &&
+	    cmd->resp.status == TEE_ERROR_CANCEL)
+		return false;
+
+	return true;
+}
+
 static int
 psp_cmd_submit_buf(struct psp_context *psp,
 		   struct amdgpu_firmware_info *ucode,
@@ -605,7 +696,7 @@ psp_cmd_submit_buf(struct psp_context *psp,
 {
 	int ret;
 	int index;
-	int timeout = 20000;
+	int timeout = psp->adev->psp_timeout;
 	bool ras_intr = false;
 	bool skip_unsupport = false;
 
@@ -643,7 +734,7 @@ psp_cmd_submit_buf(struct psp_context *psp,
 	skip_unsupport = (psp->cmd_buf_mem->resp.status == TEE_ERROR_NOT_SUPPORTED ||
 		psp->cmd_buf_mem->resp.status == PSP_ERR_UNKNOWN_COMMAND) && amdgpu_sriov_vf(psp->adev);
 
-	memcpy((void*)&cmd->resp, (void*)&psp->cmd_buf_mem->resp, sizeof(struct psp_gfx_resp));
+	memcpy(&cmd->resp, &psp->cmd_buf_mem->resp, sizeof(struct psp_gfx_resp));
 
 	/* In some cases, psp response status is not 0 even there is no
 	 * problem while the command is submitted. Some version of PSP FW
@@ -654,11 +745,16 @@ psp_cmd_submit_buf(struct psp_context *psp,
 	 */
 	if (!skip_unsupport && (psp->cmd_buf_mem->resp.status || !timeout) && !ras_intr) {
 		if (ucode)
-			DRM_WARN("failed to load ucode %s(0x%X) ",
-				  amdgpu_ucode_name(ucode->ucode_id), ucode->ucode_id);
-		DRM_WARN("psp gfx command %s(0x%X) failed and response status is (0x%X)\n",
-			 psp_gfx_cmd_name(psp->cmd_buf_mem->cmd_id), psp->cmd_buf_mem->cmd_id,
-			 psp->cmd_buf_mem->resp.status);
+			dev_warn(psp->adev->dev,
+				 "failed to load ucode %s(0x%X) ",
+				 amdgpu_ucode_name(ucode->ucode_id), ucode->ucode_id);
+		if (psp_err_warn(psp))
+			dev_warn(
+				psp->adev->dev,
+				"psp gfx command %s(0x%X) failed and response status is (0x%X)\n",
+				psp_gfx_cmd_name(psp->cmd_buf_mem->cmd_id),
+				psp->cmd_buf_mem->cmd_id,
+				psp->cmd_buf_mem->resp.status);
 		/* If any firmware (including CAP) load fails under SRIOV, it should
 		 * return failure to stop the VF from initializing.
 		 * Also return failure in case of timeout
@@ -699,8 +795,13 @@ static void psp_prep_tmr_cmd_buf(struct psp_context *psp,
 				 uint64_t tmr_mc, struct amdgpu_bo *tmr_bo)
 {
 	struct amdgpu_device *adev = psp->adev;
-	uint32_t size = amdgpu_bo_size(tmr_bo);
-	uint64_t tmr_pa = amdgpu_gmc_vram_pa(adev, tmr_bo);
+	uint32_t size = 0;
+	uint64_t tmr_pa = 0;
+
+	if (tmr_bo) {
+		size = amdgpu_bo_size(tmr_bo);
+		tmr_pa = amdgpu_gmc_vram_pa(adev, tmr_bo);
+	}
 
 	if (amdgpu_sriov_vf(psp->adev))
 		cmd->cmd_id = GFX_CMD_ID_SETUP_VMR;
@@ -763,39 +864,43 @@ static int psp_tmr_init(struct psp_context *psp)
 	tmr_size = PSP_TMR_SIZE(psp->adev);
 
 	/* For ASICs support RLC autoload, psp will parse the toc
-	 * and calculate the total size of TMR needed */
+	 * and calculate the total size of TMR needed
+	 */
 	if (!amdgpu_sriov_vf(psp->adev) &&
 	    psp->toc.start_addr &&
 	    psp->toc.size_bytes &&
 	    psp->fw_pri_buf) {
 		ret = psp_load_toc(psp, &tmr_size);
 		if (ret) {
-			DRM_ERROR("Failed to load toc\n");
+			dev_err(psp->adev->dev, "Failed to load toc\n");
 			return ret;
 		}
 	}
 
-	if (!psp->tmr_bo) {
+	if (!psp->tmr_bo && !psp->boot_time_tmr) {
 		pptr = amdgpu_sriov_vf(psp->adev) ? &tmr_buf : NULL;
 		ret = amdgpu_bo_create_kernel(psp->adev, tmr_size,
 					      PSP_TMR_ALIGNMENT,
-					      AMDGPU_HAS_VRAM(psp->adev) ?
-					      AMDGPU_GEM_DOMAIN_VRAM :
-					      AMDGPU_GEM_DOMAIN_GTT,
+					      AMDGPU_GEM_DOMAIN_GTT | AMDGPU_GEM_DOMAIN_VRAM,
 					      &psp->tmr_bo, &psp->tmr_mc_addr,
 					      pptr);
 	}
+	if (amdgpu_virt_xgmi_migrate_enabled(psp->adev) && psp->tmr_bo)
+		psp->tmr_mc_addr = amdgpu_bo_fb_aper_addr(psp->tmr_bo);
 
 	return ret;
 }
 
 static bool psp_skip_tmr(struct psp_context *psp)
 {
-	switch (psp->adev->ip_versions[MP0_HWIP][0]) {
+	switch (amdgpu_ip_version(psp->adev, MP0_HWIP, 0)) {
 	case IP_VERSION(11, 0, 9):
 	case IP_VERSION(11, 0, 7):
 	case IP_VERSION(13, 0, 2):
+	case IP_VERSION(13, 0, 6):
 	case IP_VERSION(13, 0, 10):
+	case IP_VERSION(13, 0, 12):
+	case IP_VERSION(13, 0, 14):
 		return true;
 	default:
 		return false;
@@ -816,8 +921,9 @@ static int psp_tmr_load(struct psp_context *psp)
 	cmd = acquire_psp_cmd_buf(psp);
 
 	psp_prep_tmr_cmd_buf(psp, cmd, psp->tmr_mc_addr, psp->tmr_bo);
-	DRM_INFO("reserve 0x%lx from 0x%llx for PSP TMR\n",
-		 amdgpu_bo_size(psp->tmr_bo), psp->tmr_mc_addr);
+	if (psp->tmr_bo)
+		dev_info(psp->adev->dev, "reserve 0x%lx from 0x%llx for PSP TMR\n",
+			 amdgpu_bo_size(psp->tmr_bo), psp->tmr_mc_addr);
 
 	ret = psp_cmd_submit_buf(psp, NULL, cmd,
 				 psp->fence_buf_mc_addr);
@@ -828,7 +934,7 @@ static int psp_tmr_load(struct psp_context *psp)
 }
 
 static void psp_prep_tmr_unload_cmd_buf(struct psp_context *psp,
-				        struct psp_gfx_cmd_resp *cmd)
+					struct psp_gfx_cmd_resp *cmd)
 {
 	if (amdgpu_sriov_vf(psp->adev))
 		cmd->cmd_id = GFX_CMD_ID_DESTROY_VMR;
@@ -894,6 +1000,106 @@ int psp_get_fw_attestation_records_addr(struct psp_context *psp,
 	return ret;
 }
 
+static int psp_get_fw_reservation_info(struct psp_context *psp,
+						   uint32_t cmd_id,
+						   uint64_t *addr,
+						   uint32_t *size)
+{
+	int ret;
+	uint32_t status;
+	struct psp_gfx_cmd_resp *cmd;
+
+	cmd = acquire_psp_cmd_buf(psp);
+
+	cmd->cmd_id = cmd_id;
+
+	ret = psp_cmd_submit_buf(psp, NULL, cmd,
+				 psp->fence_buf_mc_addr);
+	if (ret) {
+		release_psp_cmd_buf(psp);
+		return ret;
+	}
+
+	status = cmd->resp.status;
+	if (status == PSP_ERR_UNKNOWN_COMMAND) {
+		release_psp_cmd_buf(psp);
+		*addr = 0;
+		*size = 0;
+		return 0;
+	}
+
+	*addr = (uint64_t)cmd->resp.uresp.fw_reserve_info.reserve_base_address_hi << 32 |
+		cmd->resp.uresp.fw_reserve_info.reserve_base_address_lo;
+	*size = cmd->resp.uresp.fw_reserve_info.reserve_size;
+
+	release_psp_cmd_buf(psp);
+
+	return 0;
+}
+
+int psp_update_fw_reservation(struct psp_context *psp)
+{
+	int ret;
+	uint64_t reserv_addr, reserv_addr_ext;
+	uint32_t reserv_size, reserv_size_ext, mp0_ip_ver;
+	struct amdgpu_device *adev = psp->adev;
+
+	mp0_ip_ver = amdgpu_ip_version(adev, MP0_HWIP, 0);
+
+	if (amdgpu_sriov_vf(psp->adev))
+		return 0;
+
+	switch (mp0_ip_ver) {
+	case IP_VERSION(14, 0, 2):
+		if (adev->psp.sos.fw_version < 0x3b0e0d)
+			return 0;
+		break;
+
+	case IP_VERSION(14, 0, 3):
+		if (adev->psp.sos.fw_version < 0x3a0e14)
+			return 0;
+		break;
+
+	default:
+		return 0;
+	}
+
+	ret = psp_get_fw_reservation_info(psp, GFX_CMD_ID_FB_FW_RESERV_ADDR, &reserv_addr, &reserv_size);
+	if (ret)
+		return ret;
+	ret = psp_get_fw_reservation_info(psp, GFX_CMD_ID_FB_FW_RESERV_EXT_ADDR, &reserv_addr_ext, &reserv_size_ext);
+	if (ret)
+		return ret;
+
+	if (reserv_addr != adev->gmc.real_vram_size - reserv_size) {
+		dev_warn(adev->dev, "reserve fw region is not valid!\n");
+		return 0;
+	}
+
+	amdgpu_bo_free_kernel(&adev->mman.fw_reserved_memory, NULL, NULL);
+
+	reserv_size = roundup(reserv_size, SZ_1M);
+
+	ret = amdgpu_bo_create_kernel_at(adev, reserv_addr, reserv_size, &adev->mman.fw_reserved_memory, NULL);
+	if (ret) {
+		dev_err(adev->dev, "reserve fw region failed(%d)!\n", ret);
+		amdgpu_bo_free_kernel(&adev->mman.fw_reserved_memory, NULL, NULL);
+		return ret;
+	}
+
+	reserv_size_ext = roundup(reserv_size_ext, SZ_1M);
+
+	ret = amdgpu_bo_create_kernel_at(adev, reserv_addr_ext, reserv_size_ext,
+					 &adev->mman.fw_reserved_memory_extend, NULL);
+	if (ret) {
+		dev_err(adev->dev, "reserve extend fw region failed(%d)!\n", ret);
+		amdgpu_bo_free_kernel(&adev->mman.fw_reserved_memory_extend, NULL, NULL);
+		return ret;
+	}
+
+	return 0;
+}
+
 static int psp_boot_config_get(struct amdgpu_device *adev, uint32_t *boot_cfg)
 {
 	struct psp_context *psp = &adev->psp;
@@ -969,6 +1175,52 @@ static int psp_rl_load(struct amdgpu_device *adev)
 	return ret;
 }
 
+int psp_memory_partition(struct psp_context *psp, int mode)
+{
+	struct psp_gfx_cmd_resp *cmd;
+	int ret;
+
+	if (amdgpu_sriov_vf(psp->adev))
+		return 0;
+
+	cmd = acquire_psp_cmd_buf(psp);
+
+	cmd->cmd_id = GFX_CMD_ID_FB_NPS_MODE;
+	cmd->cmd.cmd_memory_part.mode = mode;
+
+	dev_info(psp->adev->dev,
+		 "Requesting %d memory partition change through PSP", mode);
+	ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr);
+	if (ret)
+		dev_err(psp->adev->dev,
+			"PSP request failed to change to NPS%d mode\n", mode);
+
+	release_psp_cmd_buf(psp);
+
+	return ret;
+}
+
+int psp_spatial_partition(struct psp_context *psp, int mode)
+{
+	struct psp_gfx_cmd_resp *cmd;
+	int ret;
+
+	if (amdgpu_sriov_vf(psp->adev))
+		return 0;
+
+	cmd = acquire_psp_cmd_buf(psp);
+
+	cmd->cmd_id = GFX_CMD_ID_SRIOV_SPATIAL_PART;
+	cmd->cmd.cmd_spatial_part.mode = mode;
+
+	dev_info(psp->adev->dev, "Requesting %d partitions through PSP", mode);
+	ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr);
+
+	release_psp_cmd_buf(psp);
+
+	return ret;
+}
+
 static int psp_asd_initialize(struct psp_context *psp)
 {
 	int ret;
@@ -980,6 +1232,11 @@ static int psp_asd_initialize(struct psp_context *psp)
 	if (amdgpu_sriov_vf(psp->adev) || !psp->asd_context.bin_desc.size_bytes)
 		return 0;
 
+	/* bypass asd if display hardware is not available */
+	if (!amdgpu_device_has_display_hardware(psp->adev) &&
+	    amdgpu_ip_version(psp->adev, MP0_HWIP, 0) >= IP_VERSION(13, 0, 10))
+		return 0;
+
 	psp->asd_context.mem_context.shared_mc_addr  = 0;
 	psp->asd_context.mem_context.shared_mem_size = PSP_ASD_SHARED_MEM_SIZE;
 	psp->asd_context.ta_load_type                = GFX_CMD_ID_LOAD_ASD;
@@ -1053,7 +1310,7 @@ int psp_reg_program(struct psp_context *psp, enum psp_reg_prog_id reg,
 	psp_prep_reg_prog_cmd_buf(cmd, reg, value);
 	ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr);
 	if (ret)
-		DRM_ERROR("PSP failed to program reg id %d", reg);
+		dev_err(psp->adev->dev, "PSP failed to program reg id %d\n", reg);
 
 	release_psp_cmd_buf(psp);
 
@@ -1065,7 +1322,7 @@ static void psp_prep_ta_load_cmd_buf(struct psp_gfx_cmd_resp *cmd,
 				     struct ta_context *context)
 {
 	cmd->cmd_id				= context->ta_load_type;
-	cmd->cmd.cmd_load_ta.app_phy_addr_lo 	= lower_32_bits(ta_bin_mc);
+	cmd->cmd.cmd_load_ta.app_phy_addr_lo	= lower_32_bits(ta_bin_mc);
 	cmd->cmd.cmd_load_ta.app_phy_addr_hi	= upper_32_bits(ta_bin_mc);
 	cmd->cmd.cmd_load_ta.app_len		= context->bin_desc.size_bytes;
 
@@ -1080,9 +1337,9 @@ int psp_ta_init_shared_buf(struct psp_context *psp,
 				  struct ta_mem_context *mem_ctx)
 {
 	/*
-	* Allocate 16k memory aligned to 4k from Frame Buffer (local
-	* physical) for ta to host memory
-	*/
+	 * Allocate 16k memory aligned to 4k from Frame Buffer (local
+	 * physical) for ta to host memory
+	 */
 	return amdgpu_bo_create_kernel(psp->adev, mem_ctx->shared_mem_size,
 				      PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM |
 				      AMDGPU_GEM_DOMAIN_GTT,
@@ -1129,6 +1386,11 @@ int psp_ta_load(struct psp_context *psp, struct ta_context *context)
 	psp_copy_fw(psp, context->bin_desc.start_addr,
 		    context->bin_desc.size_bytes);
 
+	if (amdgpu_virt_xgmi_migrate_enabled(psp->adev) &&
+		context->mem_context.shared_bo)
+		context->mem_context.shared_mc_addr =
+			amdgpu_bo_fb_aper_addr(context->mem_context.shared_bo);
+
 	psp_prep_ta_load_cmd_buf(cmd, psp->fw_pri_mc_addr, context);
 
 	ret = psp_cmd_submit_buf(psp, NULL, cmd,
@@ -1136,9 +1398,8 @@ int psp_ta_load(struct psp_context *psp, struct ta_context *context)
 
 	context->resp_status = cmd->resp.status;
 
-	if (!ret) {
+	if (!ret)
 		context->session_id = cmd->resp.session_id;
-	}
 
 	release_psp_cmd_buf(psp);
 
@@ -1156,8 +1417,8 @@ int psp_xgmi_terminate(struct psp_context *psp)
 	struct amdgpu_device *adev = psp->adev;
 
 	/* XGMI TA unload currently is not supported on Arcturus/Aldebaran A+A */
-	if (adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 4) ||
-	    (adev->ip_versions[MP0_HWIP][0] == IP_VERSION(13, 0, 2) &&
+	if (amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(11, 0, 4) ||
+	    (amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 2) &&
 	     adev->gmc.xgmi.connected_to_cpu))
 		return 0;
 
@@ -1208,6 +1469,8 @@ invoke:
 	xgmi_cmd->cmd_id = TA_COMMAND_XGMI__INITIALIZE;
 
 	ret = psp_xgmi_invoke(psp, xgmi_cmd->cmd_id);
+	/* note down the capbility flag for XGMI TA */
+	psp->xgmi_context.xgmi_ta_caps = xgmi_cmd->caps_flag;
 
 	return ret;
 }
@@ -1254,8 +1517,11 @@ int psp_xgmi_get_node_id(struct psp_context *psp, uint64_t *node_id)
 
 static bool psp_xgmi_peer_link_info_supported(struct psp_context *psp)
 {
-	return psp->adev->ip_versions[MP0_HWIP][0] == IP_VERSION(13, 0, 2) &&
-		psp->xgmi_context.context.bin_desc.fw_version >= 0x2000000b;
+	return (amdgpu_ip_version(psp->adev, MP0_HWIP, 0) ==
+			IP_VERSION(13, 0, 2) &&
+		psp->xgmi_context.context.bin_desc.fw_version >= 0x2000000b) ||
+	       amdgpu_ip_version(psp->adev, MP0_HWIP, 0) >=
+		       IP_VERSION(13, 0, 6);
 }
 
 /*
@@ -1273,9 +1539,13 @@ static void psp_xgmi_reflect_topology_info(struct psp_context *psp,
 	uint64_t src_node_id = psp->adev->gmc.xgmi.node_id;
 	uint64_t dst_node_id = node_info.node_id;
 	uint8_t dst_num_hops = node_info.num_hops;
+	uint8_t dst_is_sharing_enabled = node_info.is_sharing_enabled;
 	uint8_t dst_num_links = node_info.num_links;
 
 	hive = amdgpu_get_xgmi_hive(psp->adev);
+	if (WARN_ON(!hive))
+		return;
+
 	list_for_each_entry(mirror_adev, &hive->device_list, gmc.xgmi.head) {
 		struct psp_xgmi_topology_info *mirror_top_info;
 		int j;
@@ -1289,13 +1559,20 @@ static void psp_xgmi_reflect_topology_info(struct psp_context *psp,
 				continue;
 
 			mirror_top_info->nodes[j].num_hops = dst_num_hops;
-			/*
-			 * prevent 0 num_links value re-reflection since reflection
+			mirror_top_info->nodes[j].is_sharing_enabled = dst_is_sharing_enabled;
+			/* prevent 0 num_links value re-reflection since reflection
 			 * criteria is based on num_hops (direct or indirect).
-			 *
 			 */
-			if (dst_num_links)
+			if (dst_num_links) {
 				mirror_top_info->nodes[j].num_links = dst_num_links;
+				/* swap src and dst due to frame of reference */
+				for (int k = 0; k < dst_num_links; k++) {
+					mirror_top_info->nodes[j].port_num[k].src_xgmi_port_num =
+						node_info.port_num[k].dst_xgmi_port_num;
+					mirror_top_info->nodes[j].port_num[k].dst_xgmi_port_num =
+						node_info.port_num[k].src_xgmi_port_num;
+				}
+			}
 
 			break;
 		}
@@ -1326,7 +1603,7 @@ int psp_xgmi_get_topology_info(struct psp_context *psp,
 
 	/* Fill in the shared memory with topology information as input */
 	topology_info_input = &xgmi_cmd->xgmi_in_message.get_topology_info;
-	xgmi_cmd->cmd_id = TA_COMMAND_XGMI__GET_GET_TOPOLOGY_INFO;
+	xgmi_cmd->cmd_id = TA_COMMAND_XGMI__GET_TOPOLOGY_INFO;
 	topology_info_input->num_nodes = number_devices;
 
 	for (i = 0; i < topology_info_input->num_nodes; i++) {
@@ -1337,7 +1614,7 @@ int psp_xgmi_get_topology_info(struct psp_context *psp,
 	}
 
 	/* Invoke xgmi ta to get the topology information */
-	ret = psp_xgmi_invoke(psp, TA_COMMAND_XGMI__GET_GET_TOPOLOGY_INFO);
+	ret = psp_xgmi_invoke(psp, TA_COMMAND_XGMI__GET_TOPOLOGY_INFO);
 	if (ret)
 		return ret;
 
@@ -1362,26 +1639,64 @@ int psp_xgmi_get_topology_info(struct psp_context *psp,
 
 	/* Invoke xgmi ta again to get the link information */
 	if (psp_xgmi_peer_link_info_supported(psp)) {
-		struct ta_xgmi_cmd_get_peer_link_info_output *link_info_output;
+		struct ta_xgmi_cmd_get_peer_link_info *link_info_output;
+		struct ta_xgmi_cmd_get_extend_peer_link_info *link_extend_info_output;
+		bool requires_reflection =
+			(psp->xgmi_context.supports_extended_data &&
+			 get_extended_data) ||
+			amdgpu_ip_version(psp->adev, MP0_HWIP, 0) ==
+				IP_VERSION(13, 0, 6) ||
+			amdgpu_ip_version(psp->adev, MP0_HWIP, 0) ==
+				IP_VERSION(13, 0, 14) ||
+			amdgpu_sriov_vf(psp->adev);
+		bool ta_port_num_support = psp->xgmi_context.xgmi_ta_caps & EXTEND_PEER_LINK_INFO_CMD_FLAG ||
+			amdgpu_sriov_xgmi_ta_ext_peer_link_en(psp->adev);
+
+		/* popluate the shared output buffer rather than the cmd input buffer
+		 * with node_ids as the input for GET_PEER_LINKS command execution.
+		 * This is required for GET_PEER_LINKS per xgmi ta implementation.
+		 * The same requirement for GET_EXTEND_PEER_LINKS command.
+		 */
+		if (ta_port_num_support) {
+			link_extend_info_output = &xgmi_cmd->xgmi_out_message.get_extend_link_info;
+
+			for (i = 0; i < topology->num_nodes; i++)
+				link_extend_info_output->nodes[i].node_id = topology->nodes[i].node_id;
 
-		xgmi_cmd->cmd_id = TA_COMMAND_XGMI__GET_PEER_LINKS;
+			link_extend_info_output->num_nodes = topology->num_nodes;
+			xgmi_cmd->cmd_id = TA_COMMAND_XGMI__GET_EXTEND_PEER_LINKS;
+		} else {
+			link_info_output = &xgmi_cmd->xgmi_out_message.get_link_info;
+
+			for (i = 0; i < topology->num_nodes; i++)
+				link_info_output->nodes[i].node_id = topology->nodes[i].node_id;
 
-		ret = psp_xgmi_invoke(psp, TA_COMMAND_XGMI__GET_PEER_LINKS);
+			link_info_output->num_nodes = topology->num_nodes;
+			xgmi_cmd->cmd_id = TA_COMMAND_XGMI__GET_PEER_LINKS;
+		}
 
+		ret = psp_xgmi_invoke(psp, xgmi_cmd->cmd_id);
 		if (ret)
 			return ret;
 
-		link_info_output = &xgmi_cmd->xgmi_out_message.get_link_info;
 		for (i = 0; i < topology->num_nodes; i++) {
+			uint8_t node_num_links = ta_port_num_support ?
+				link_extend_info_output->nodes[i].num_links : link_info_output->nodes[i].num_links;
 			/* accumulate num_links on extended data */
-			topology->nodes[i].num_links = get_extended_data ?
-					topology->nodes[i].num_links +
-							link_info_output->nodes[i].num_links :
-					link_info_output->nodes[i].num_links;
+			if (get_extended_data) {
+				topology->nodes[i].num_links = topology->nodes[i].num_links + node_num_links;
+			} else {
+				topology->nodes[i].num_links = (requires_reflection && topology->nodes[i].num_links) ?
+								topology->nodes[i].num_links : node_num_links;
+			}
+			/* popluate the connected port num info if supported and available */
+			if (ta_port_num_support && topology->nodes[i].num_links) {
+				memcpy(topology->nodes[i].port_num, link_extend_info_output->nodes[i].port_num,
+				       sizeof(struct xgmi_connected_port_num) * TA_XGMI__MAX_PORT_NUM);
+			}
 
 			/* reflect the topology information for bi-directionality */
-			if (psp->xgmi_context.supports_extended_data &&
-					get_extended_data && topology->nodes[i].num_hops)
+			if (requires_reflection && topology->nodes[i].num_hops)
 				psp_xgmi_reflect_topology_info(psp, topology->nodes[i]);
 		}
 	}
@@ -1427,24 +1742,84 @@ static void psp_ras_ta_check_status(struct psp_context *psp)
 	switch (ras_cmd->ras_status) {
 	case TA_RAS_STATUS__ERROR_UNSUPPORTED_IP:
 		dev_warn(psp->adev->dev,
-				"RAS WARNING: cmd failed due to unsupported ip\n");
+			 "RAS WARNING: cmd failed due to unsupported ip\n");
 		break;
 	case TA_RAS_STATUS__ERROR_UNSUPPORTED_ERROR_INJ:
 		dev_warn(psp->adev->dev,
-				"RAS WARNING: cmd failed due to unsupported error injection\n");
+			 "RAS WARNING: cmd failed due to unsupported error injection\n");
 		break;
 	case TA_RAS_STATUS__SUCCESS:
 		break;
 	case TA_RAS_STATUS__TEE_ERROR_ACCESS_DENIED:
 		if (ras_cmd->cmd_id == TA_RAS_COMMAND__TRIGGER_ERROR)
 			dev_warn(psp->adev->dev,
-					"RAS WARNING: Inject error to critical region is not allowed\n");
+				 "RAS WARNING: Inject error to critical region is not allowed\n");
 		break;
 	default:
 		dev_warn(psp->adev->dev,
-				"RAS WARNING: ras status = 0x%X\n", ras_cmd->ras_status);
+			 "RAS WARNING: ras status = 0x%X\n", ras_cmd->ras_status);
+		break;
+	}
+}
+
+static int psp_ras_send_cmd(struct psp_context *psp,
+		enum ras_command cmd_id, void *in, void *out)
+{
+	struct ta_ras_shared_memory *ras_cmd;
+	uint32_t cmd = cmd_id;
+	int ret = 0;
+
+	if (!in)
+		return -EINVAL;
+
+	mutex_lock(&psp->ras_context.mutex);
+	ras_cmd = (struct ta_ras_shared_memory *)psp->ras_context.context.mem_context.shared_buf;
+	memset(ras_cmd, 0, sizeof(struct ta_ras_shared_memory));
+
+	switch (cmd) {
+	case TA_RAS_COMMAND__ENABLE_FEATURES:
+	case TA_RAS_COMMAND__DISABLE_FEATURES:
+		memcpy(&ras_cmd->ras_in_message,
+			in, sizeof(ras_cmd->ras_in_message));
+		break;
+	case TA_RAS_COMMAND__TRIGGER_ERROR:
+		memcpy(&ras_cmd->ras_in_message.trigger_error,
+			in, sizeof(ras_cmd->ras_in_message.trigger_error));
 		break;
+	case TA_RAS_COMMAND__QUERY_ADDRESS:
+		memcpy(&ras_cmd->ras_in_message.address,
+			in, sizeof(ras_cmd->ras_in_message.address));
+		break;
+	default:
+		dev_err(psp->adev->dev, "Invalid ras cmd id: %u\n", cmd);
+		ret = -EINVAL;
+		goto err_out;
 	}
+
+	ras_cmd->cmd_id = cmd;
+	ret = psp_ras_invoke(psp, ras_cmd->cmd_id);
+
+	switch (cmd) {
+	case TA_RAS_COMMAND__TRIGGER_ERROR:
+		if (!ret && out)
+			memcpy(out, &ras_cmd->ras_status, sizeof(ras_cmd->ras_status));
+		break;
+	case TA_RAS_COMMAND__QUERY_ADDRESS:
+		if (ret || ras_cmd->ras_status || psp->cmd_buf_mem->resp.status)
+			ret = -EINVAL;
+		else if (out)
+			memcpy(out,
+				&ras_cmd->ras_out_message.address,
+				sizeof(ras_cmd->ras_out_message.address));
+		break;
+	default:
+		break;
+	}
+
+err_out:
+	mutex_unlock(&psp->ras_context.mutex);
+
+	return ret;
 }
 
 int psp_ras_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
@@ -1465,9 +1840,8 @@ int psp_ras_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
 	if (amdgpu_ras_intr_triggered())
 		return ret;
 
-	if (ras_cmd->if_version > RAS_TA_HOST_IF_VER)
-	{
-		DRM_WARN("RAS: Unsupported Interface");
+	if (ras_cmd->if_version > RAS_TA_HOST_IF_VER) {
+		dev_warn(psp->adev->dev, "RAS: Unsupported Interface\n");
 		return -EINVAL;
 	}
 
@@ -1476,8 +1850,7 @@ int psp_ras_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
 			dev_warn(psp->adev->dev, "ECC switch disabled\n");
 
 			ras_cmd->ras_status = TA_RAS_STATUS__ERROR_RAS_NOT_AVAILABLE;
-		}
-		else if (ras_cmd->ras_out_message.flags.reg_access_failure_flag)
+		} else if (ras_cmd->ras_out_message.flags.reg_access_failure_flag)
 			dev_warn(psp->adev->dev,
 				 "RAS internal register access blocked\n");
 
@@ -1490,23 +1863,15 @@ int psp_ras_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
 int psp_ras_enable_features(struct psp_context *psp,
 		union ta_ras_cmd_input *info, bool enable)
 {
-	struct ta_ras_shared_memory *ras_cmd;
+	enum ras_command cmd_id;
 	int ret;
 
-	if (!psp->ras_context.context.initialized)
+	if (!psp->ras_context.context.initialized || !info)
 		return -EINVAL;
 
-	ras_cmd = (struct ta_ras_shared_memory *)psp->ras_context.context.mem_context.shared_buf;
-	memset(ras_cmd, 0, sizeof(struct ta_ras_shared_memory));
-
-	if (enable)
-		ras_cmd->cmd_id = TA_RAS_COMMAND__ENABLE_FEATURES;
-	else
-		ras_cmd->cmd_id = TA_RAS_COMMAND__DISABLE_FEATURES;
-
-	ras_cmd->ras_in_message = *info;
-
-	ret = psp_ras_invoke(psp, ras_cmd->cmd_id);
+	cmd_id = enable ?
+		TA_RAS_COMMAND__ENABLE_FEATURES : TA_RAS_COMMAND__DISABLE_FEATURES;
+	ret = psp_ras_send_cmd(psp, cmd_id, info, NULL);
 	if (ret)
 		return -EINVAL;
 
@@ -1530,6 +1895,8 @@ int psp_ras_terminate(struct psp_context *psp)
 
 	psp->ras_context.context.initialized = false;
 
+	mutex_destroy(&psp->ras_context.mutex);
+
 	return ret;
 }
 
@@ -1560,35 +1927,47 @@ int psp_ras_initialize(struct psp_context *psp)
 		if (ret)
 			dev_warn(adev->dev, "PSP get boot config failed\n");
 
-		if (!amdgpu_ras_is_supported(psp->adev, AMDGPU_RAS_BLOCK__UMC)) {
-			if (!boot_cfg) {
-				dev_info(adev->dev, "GECC is disabled\n");
-			} else {
-				/* disable GECC in next boot cycle if ras is
-				 * disabled by module parameter amdgpu_ras_enable
-				 * and/or amdgpu_ras_mask, or boot_config_get call
-				 * is failed
-				 */
-				ret = psp_boot_config_set(adev, 0);
-				if (ret)
-					dev_warn(adev->dev, "PSP set boot config failed\n");
-				else
-					dev_warn(adev->dev, "GECC will be disabled in next boot cycle "
-						 "if set amdgpu_ras_enable and/or amdgpu_ras_mask to 0x0\n");
-			}
+		if (boot_cfg == 1 && !adev->ras_default_ecc_enabled &&
+		    amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC)) {
+			dev_warn(adev->dev, "GECC is currently enabled, which may affect performance\n");
+			dev_warn(adev->dev,
+				"To disable GECC, please reboot the system and load the amdgpu driver with the parameter amdgpu_ras_enable=0\n");
 		} else {
-			if (1 == boot_cfg) {
-				dev_info(adev->dev, "GECC is enabled\n");
+			if ((adev->ras_default_ecc_enabled || amdgpu_ras_enable == 1) &&
+				amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC)) {
+				if (boot_cfg == 1) {
+					dev_info(adev->dev, "GECC is enabled\n");
+				} else {
+					/* enable GECC in next boot cycle if it is disabled
+					 * in boot config, or force enable GECC if failed to
+					 * get boot configuration
+					 */
+					ret = psp_boot_config_set(adev, BOOT_CONFIG_GECC);
+					if (ret)
+						dev_warn(adev->dev, "PSP set boot config failed\n");
+					else
+						dev_warn(adev->dev, "GECC will be enabled in next boot cycle\n");
+				}
 			} else {
-				/* enable GECC in next boot cycle if it is disabled
-				 * in boot config, or force enable GECC if failed to
-				 * get boot configuration
-				 */
-				ret = psp_boot_config_set(adev, BOOT_CONFIG_GECC);
-				if (ret)
-					dev_warn(adev->dev, "PSP set boot config failed\n");
-				else
-					dev_warn(adev->dev, "GECC will be enabled in next boot cycle\n");
+				if (!boot_cfg) {
+					if (!adev->ras_default_ecc_enabled &&
+					    amdgpu_ras_enable != 1 &&
+					    amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC))
+						dev_warn(adev->dev, "GECC is disabled, set amdgpu_ras_enable=1 to enable GECC in next boot cycle if needed\n");
+					else
+						dev_info(adev->dev, "GECC is disabled\n");
+				} else {
+					/* disable GECC in next boot cycle if ras is
+					 * disabled by module parameter amdgpu_ras_enable
+					 * and/or amdgpu_ras_mask, or boot_config_get call
+					 * is failed
+					 */
+					ret = psp_boot_config_set(adev, 0);
+					if (ret)
+						dev_warn(adev->dev, "PSP set boot config failed\n");
+					else
+						dev_warn(adev->dev, "GECC will be disabled in next boot cycle if set amdgpu_ras_enable and/or amdgpu_ras_mask to 0x0\n");
+				}
 			}
 		}
 	}
@@ -1607,16 +1986,24 @@ int psp_ras_initialize(struct psp_context *psp)
 
 	if (amdgpu_ras_is_poison_mode_supported(adev))
 		ras_cmd->ras_in_message.init_flags.poison_mode_en = 1;
-	if (!adev->gmc.xgmi.connected_to_cpu)
+	if (!adev->gmc.xgmi.connected_to_cpu && !adev->gmc.is_app_apu)
 		ras_cmd->ras_in_message.init_flags.dgpu_mode = 1;
+	ras_cmd->ras_in_message.init_flags.xcc_mask =
+		adev->gfx.xcc_mask;
+	ras_cmd->ras_in_message.init_flags.channel_dis_num = hweight32(adev->gmc.m_half_use) * 2;
+	if (adev->gmc.gmc_funcs->query_mem_partition_mode)
+		ras_cmd->ras_in_message.init_flags.nps_mode =
+			adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
+	ras_cmd->ras_in_message.init_flags.active_umc_mask = adev->umc.active_mask;
 
 	ret = psp_ta_load(psp, &psp->ras_context.context);
 
-	if (!ret && !ras_cmd->ras_status)
+	if (!ret && !ras_cmd->ras_status) {
 		psp->ras_context.context.initialized = true;
-	else {
+		mutex_init(&psp->ras_context.mutex);
+	} else {
 		if (ras_cmd->ras_status)
-			dev_warn(psp->adev->dev, "RAS Init Status: 0x%X\n", ras_cmd->ras_status);
+			dev_warn(adev->dev, "RAS Init Status: 0x%X\n", ras_cmd->ras_status);
 
 		/* fail to load RAS TA */
 		psp->ras_context.context.initialized = false;
@@ -1626,36 +2013,71 @@ int psp_ras_initialize(struct psp_context *psp)
 }
 
 int psp_ras_trigger_error(struct psp_context *psp,
-			  struct ta_ras_trigger_error_input *info)
+			  struct ta_ras_trigger_error_input *info, uint32_t instance_mask)
 {
-	struct ta_ras_shared_memory *ras_cmd;
+	struct amdgpu_device *adev = psp->adev;
 	int ret;
+	uint32_t dev_mask;
+	uint32_t ras_status = 0;
 
-	if (!psp->ras_context.context.initialized)
+	if (!psp->ras_context.context.initialized || !info)
 		return -EINVAL;
 
-	ras_cmd = (struct ta_ras_shared_memory *)psp->ras_context.context.mem_context.shared_buf;
-	memset(ras_cmd, 0, sizeof(struct ta_ras_shared_memory));
+	switch (info->block_id) {
+	case TA_RAS_BLOCK__GFX:
+		dev_mask = GET_MASK(GC, instance_mask);
+		break;
+	case TA_RAS_BLOCK__SDMA:
+		dev_mask = GET_MASK(SDMA0, instance_mask);
+		break;
+	case TA_RAS_BLOCK__VCN:
+	case TA_RAS_BLOCK__JPEG:
+		dev_mask = GET_MASK(VCN, instance_mask);
+		break;
+	default:
+		dev_mask = instance_mask;
+		break;
+	}
 
-	ras_cmd->cmd_id = TA_RAS_COMMAND__TRIGGER_ERROR;
-	ras_cmd->ras_in_message.trigger_error = *info;
+	/* reuse sub_block_index for backward compatibility */
+	dev_mask <<= AMDGPU_RAS_INST_SHIFT;
+	dev_mask &= AMDGPU_RAS_INST_MASK;
+	info->sub_block_index |= dev_mask;
 
-	ret = psp_ras_invoke(psp, ras_cmd->cmd_id);
+	ret = psp_ras_send_cmd(psp,
+			TA_RAS_COMMAND__TRIGGER_ERROR, info, &ras_status);
 	if (ret)
 		return -EINVAL;
 
 	/* If err_event_athub occurs error inject was successful, however
-	   return status from TA is no long reliable */
+	 *  return status from TA is no long reliable
+	 */
 	if (amdgpu_ras_intr_triggered())
 		return 0;
 
-	if (ras_cmd->ras_status == TA_RAS_STATUS__TEE_ERROR_ACCESS_DENIED)
+	if (ras_status == TA_RAS_STATUS__TEE_ERROR_ACCESS_DENIED)
 		return -EACCES;
-	else if (ras_cmd->ras_status)
+	else if (ras_status)
 		return -EINVAL;
 
 	return 0;
 }
+
+int psp_ras_query_address(struct psp_context *psp,
+			  struct ta_ras_query_address_input *addr_in,
+			  struct ta_ras_query_address_output *addr_out)
+{
+	int ret;
+
+	if (!psp->ras_context.context.initialized ||
+		!addr_in || !addr_out)
+		return -EINVAL;
+
+	ret = psp_ras_send_cmd(psp,
+			TA_RAS_COMMAND__QUERY_ADDRESS, addr_in, addr_out);
+
+	return ret;
+}
 // ras end
 
 // HDCP start
@@ -1669,6 +2091,10 @@ static int psp_hdcp_initialize(struct psp_context *psp)
 	if (amdgpu_sriov_vf(psp->adev))
 		return 0;
 
+	/* bypass hdcp initialization if dmu is harvested */
+	if (!amdgpu_device_has_display_hardware(psp->adev))
+		return 0;
+
 	if (!psp->hdcp_context.context.bin_desc.size_bytes ||
 	    !psp->hdcp_context.context.bin_desc.start_addr) {
 		dev_info(psp->adev->dev, "HDCP: optional hdcp ta ucode is not available\n");
@@ -1701,6 +2127,9 @@ int psp_hdcp_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
 	if (amdgpu_sriov_vf(psp->adev))
 		return 0;
 
+	if (!psp->hdcp_context.context.initialized)
+		return 0;
+
 	return psp_ta_invoke(psp, ta_cmd_id, &psp->hdcp_context.context);
 }
 
@@ -1736,6 +2165,10 @@ static int psp_dtm_initialize(struct psp_context *psp)
 	if (amdgpu_sriov_vf(psp->adev))
 		return 0;
 
+	/* bypass dtm initialization if dmu is harvested */
+	if (!amdgpu_device_has_display_hardware(psp->adev))
+		return 0;
+
 	if (!psp->dtm_context.context.bin_desc.size_bytes ||
 	    !psp->dtm_context.context.bin_desc.start_addr) {
 		dev_info(psp->adev->dev, "DTM: optional dtm ta ucode is not available\n");
@@ -1768,6 +2201,9 @@ int psp_dtm_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
 	if (amdgpu_sriov_vf(psp->adev))
 		return 0;
 
+	if (!psp->dtm_context.context.initialized)
+		return 0;
+
 	return psp_ta_invoke(psp, ta_cmd_id, &psp->dtm_context.context);
 }
 
@@ -1902,9 +2338,14 @@ static int psp_securedisplay_initialize(struct psp_context *psp)
 	if (amdgpu_sriov_vf(psp->adev))
 		return 0;
 
+	/* bypass securedisplay initialization if dmu is harvested */
+	if (!amdgpu_device_has_display_hardware(psp->adev))
+		return 0;
+
 	if (!psp->securedisplay_context.context.bin_desc.size_bytes ||
 	    !psp->securedisplay_context.context.bin_desc.start_addr) {
-		dev_info(psp->adev->dev, "SECUREDISPLAY: securedisplay ta ucode is not available\n");
+		dev_info(psp->adev->dev,
+			 "SECUREDISPLAY: optional securedisplay ta ucode is not available\n");
 		return 0;
 	}
 
@@ -1920,11 +2361,14 @@ static int psp_securedisplay_initialize(struct psp_context *psp)
 	}
 
 	ret = psp_ta_load(psp, &psp->securedisplay_context.context);
-	if (!ret) {
+	if (!ret && !psp->securedisplay_context.context.resp_status) {
 		psp->securedisplay_context.context.initialized = true;
 		mutex_init(&psp->securedisplay_context.mutex);
-	} else
+	} else {
+		/* don't try again */
+		psp->securedisplay_context.context.bin_desc.size_bytes = 0;
 		return ret;
+	}
 
 	mutex_lock(&psp->securedisplay_context.mutex);
 
@@ -1947,6 +2391,8 @@ static int psp_securedisplay_initialize(struct psp_context *psp)
 		psp_securedisplay_parse_resp_status(psp, securedisplay_cmd->status);
 		dev_err(psp->adev->dev, "SECUREDISPLAY: query securedisplay TA failed. ret 0x%x\n",
 			securedisplay_cmd->securedisplay_out_message.query_ta.query_cmd_ret);
+		/* don't try again */
+		psp->securedisplay_context.context.bin_desc.size_bytes = 0;
 	}
 
 	return 0;
@@ -1980,7 +2426,8 @@ int psp_securedisplay_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
 		return -EINVAL;
 
 	if (ta_cmd_id != TA_SECUREDISPLAY_COMMAND__QUERY_TA &&
-	    ta_cmd_id != TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC)
+	    ta_cmd_id != TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC &&
+	    ta_cmd_id != TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC_V2)
 		return -EINVAL;
 
 	ret = psp_ta_invoke(psp, ta_cmd_id, &psp->securedisplay_context.context);
@@ -1989,17 +2436,67 @@ int psp_securedisplay_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
 }
 /* SECUREDISPLAY end */
 
+int amdgpu_psp_wait_for_bootloader(struct amdgpu_device *adev)
+{
+	struct psp_context *psp = &adev->psp;
+	int ret = 0;
+
+	if (!amdgpu_sriov_vf(adev) && psp->funcs && psp->funcs->wait_for_bootloader != NULL)
+		ret = psp->funcs->wait_for_bootloader(psp);
+
+	return ret;
+}
+
+bool amdgpu_psp_get_ras_capability(struct psp_context *psp)
+{
+	if (psp->funcs &&
+	    psp->funcs->get_ras_capability) {
+		return psp->funcs->get_ras_capability(psp);
+	} else {
+		return false;
+	}
+}
+
+bool amdgpu_psp_tos_reload_needed(struct amdgpu_device *adev)
+{
+	struct psp_context *psp = &adev->psp;
+
+	if (amdgpu_sriov_vf(adev) || (adev->flags & AMD_IS_APU))
+		return false;
+
+	if (psp->funcs && psp->funcs->is_reload_needed)
+		return psp->funcs->is_reload_needed(psp);
+
+	return false;
+}
+
+static void psp_update_gpu_addresses(struct amdgpu_device *adev)
+{
+	struct psp_context *psp = &adev->psp;
+
+	if (psp->cmd_buf_bo && psp->cmd_buf_mem) {
+		psp->fw_pri_mc_addr = amdgpu_bo_fb_aper_addr(psp->fw_pri_bo);
+		psp->fence_buf_mc_addr = amdgpu_bo_fb_aper_addr(psp->fence_buf_bo);
+		psp->cmd_buf_mc_addr = amdgpu_bo_fb_aper_addr(psp->cmd_buf_bo);
+	}
+	if (adev->firmware.rbuf && psp->km_ring.ring_mem)
+		psp->km_ring.ring_mem_mc_addr = amdgpu_bo_fb_aper_addr(adev->firmware.rbuf);
+}
+
 static int psp_hw_start(struct psp_context *psp)
 {
 	struct amdgpu_device *adev = psp->adev;
 	int ret;
 
+	if (amdgpu_virt_xgmi_migrate_enabled(adev))
+		psp_update_gpu_addresses(adev);
+
 	if (!amdgpu_sriov_vf(adev)) {
 		if ((is_psp_fw_valid(psp->kdb)) &&
 		    (psp->funcs->bootloader_load_kdb != NULL)) {
 			ret = psp_bootloader_load_kdb(psp);
 			if (ret) {
-				DRM_ERROR("PSP load kdb failed!\n");
+				dev_err(adev->dev, "PSP load kdb failed!\n");
 				return ret;
 			}
 		}
@@ -2008,7 +2505,7 @@ static int psp_hw_start(struct psp_context *psp)
 		    (psp->funcs->bootloader_load_spl != NULL)) {
 			ret = psp_bootloader_load_spl(psp);
 			if (ret) {
-				DRM_ERROR("PSP load spl failed!\n");
+				dev_err(adev->dev, "PSP load spl failed!\n");
 				return ret;
 			}
 		}
@@ -2017,7 +2514,7 @@ static int psp_hw_start(struct psp_context *psp)
 		    (psp->funcs->bootloader_load_sysdrv != NULL)) {
 			ret = psp_bootloader_load_sysdrv(psp);
 			if (ret) {
-				DRM_ERROR("PSP load sys drv failed!\n");
+				dev_err(adev->dev, "PSP load sys drv failed!\n");
 				return ret;
 			}
 		}
@@ -2026,7 +2523,7 @@ static int psp_hw_start(struct psp_context *psp)
 		    (psp->funcs->bootloader_load_soc_drv != NULL)) {
 			ret = psp_bootloader_load_soc_drv(psp);
 			if (ret) {
-				DRM_ERROR("PSP load soc drv failed!\n");
+				dev_err(adev->dev, "PSP load soc drv failed!\n");
 				return ret;
 			}
 		}
@@ -2035,7 +2532,7 @@ static int psp_hw_start(struct psp_context *psp)
 		    (psp->funcs->bootloader_load_intf_drv != NULL)) {
 			ret = psp_bootloader_load_intf_drv(psp);
 			if (ret) {
-				DRM_ERROR("PSP load intf drv failed!\n");
+				dev_err(adev->dev, "PSP load intf drv failed!\n");
 				return ret;
 			}
 		}
@@ -2044,7 +2541,7 @@ static int psp_hw_start(struct psp_context *psp)
 		    (psp->funcs->bootloader_load_dbg_drv != NULL)) {
 			ret = psp_bootloader_load_dbg_drv(psp);
 			if (ret) {
-				DRM_ERROR("PSP load dbg drv failed!\n");
+				dev_err(adev->dev, "PSP load dbg drv failed!\n");
 				return ret;
 			}
 		}
@@ -2053,7 +2550,25 @@ static int psp_hw_start(struct psp_context *psp)
 		    (psp->funcs->bootloader_load_ras_drv != NULL)) {
 			ret = psp_bootloader_load_ras_drv(psp);
 			if (ret) {
-				DRM_ERROR("PSP load ras_drv failed!\n");
+				dev_err(adev->dev, "PSP load ras_drv failed!\n");
+				return ret;
+			}
+		}
+
+		if ((is_psp_fw_valid(psp->ipkeymgr_drv)) &&
+		    (psp->funcs->bootloader_load_ipkeymgr_drv != NULL)) {
+			ret = psp_bootloader_load_ipkeymgr_drv(psp);
+			if (ret) {
+				dev_err(adev->dev, "PSP load ipkeymgr_drv failed!\n");
+				return ret;
+			}
+		}
+
+		if ((is_psp_fw_valid(psp->spdm_drv)) &&
+		    (psp->funcs->bootloader_load_spdm_drv != NULL)) {
+			ret = psp_bootloader_load_spdm_drv(psp);
+			if (ret) {
+				dev_err(adev->dev, "PSP load spdm_drv failed!\n");
 				return ret;
 			}
 		}
@@ -2062,7 +2577,7 @@ static int psp_hw_start(struct psp_context *psp)
 		    (psp->funcs->bootloader_load_sos != NULL)) {
 			ret = psp_bootloader_load_sos(psp);
 			if (ret) {
-				DRM_ERROR("PSP load sos failed!\n");
+				dev_err(adev->dev, "PSP load sos failed!\n");
 				return ret;
 			}
 		}
@@ -2070,17 +2585,27 @@ static int psp_hw_start(struct psp_context *psp)
 
 	ret = psp_ring_create(psp, PSP_RING_TYPE__KM);
 	if (ret) {
-		DRM_ERROR("PSP create ring failed!\n");
+		dev_err(adev->dev, "PSP create ring failed!\n");
 		return ret;
 	}
 
+	if (!amdgpu_in_reset(adev) && !adev->in_suspend) {
+		ret = psp_update_fw_reservation(psp);
+		if (ret) {
+			dev_err(adev->dev, "update fw reservation failed!\n");
+			return ret;
+		}
+	}
+
 	if (amdgpu_sriov_vf(adev) && amdgpu_in_reset(adev))
 		goto skip_pin_bo;
 
-	ret = psp_tmr_init(psp);
-	if (ret) {
-		DRM_ERROR("PSP tmr init failed!\n");
-		return ret;
+	if (!psp->boot_time_tmr || psp->autoload_supported) {
+		ret = psp_tmr_init(psp);
+		if (ret) {
+			dev_err(adev->dev, "PSP tmr init failed!\n");
+			return ret;
+		}
 	}
 
 skip_pin_bo:
@@ -2095,10 +2620,12 @@ skip_pin_bo:
 			return ret;
 	}
 
-	ret = psp_tmr_load(psp);
-	if (ret) {
-		DRM_ERROR("PSP load tmr failed!\n");
-		return ret;
+	if (!psp->boot_time_tmr || !psp->autoload_supported) {
+		ret = psp_tmr_load(psp);
+		if (ret) {
+			dev_err(adev->dev, "PSP load tmr failed!\n");
+			return ret;
+		}
 	}
 
 	return 0;
@@ -2244,6 +2771,7 @@ static int psp_get_fw_type(struct amdgpu_firmware_info *ucode,
 		*type = GFX_FW_TYPE_DMUB;
 		break;
 	case AMDGPU_UCODE_ID_SDMA_UCODE_TH0:
+	case AMDGPU_UCODE_ID_SDMA_RS64:
 		*type = GFX_FW_TYPE_SDMA_UCODE_TH0;
 		break;
 	case AMDGPU_UCODE_ID_SDMA_UCODE_TH1:
@@ -2288,6 +2816,33 @@ static int psp_get_fw_type(struct amdgpu_firmware_info *ucode,
 	case AMDGPU_UCODE_ID_CP_RS64_MEC_P3_STACK:
 		*type = GFX_FW_TYPE_RS64_MEC_P3_STACK;
 		break;
+	case AMDGPU_UCODE_ID_VPE_CTX:
+		*type = GFX_FW_TYPE_VPEC_FW1;
+		break;
+	case AMDGPU_UCODE_ID_VPE_CTL:
+		*type = GFX_FW_TYPE_VPEC_FW2;
+		break;
+	case AMDGPU_UCODE_ID_VPE:
+		*type = GFX_FW_TYPE_VPE;
+		break;
+	case AMDGPU_UCODE_ID_UMSCH_MM_UCODE:
+		*type = GFX_FW_TYPE_UMSCH_UCODE;
+		break;
+	case AMDGPU_UCODE_ID_UMSCH_MM_DATA:
+		*type = GFX_FW_TYPE_UMSCH_DATA;
+		break;
+	case AMDGPU_UCODE_ID_UMSCH_MM_CMD_BUFFER:
+		*type = GFX_FW_TYPE_UMSCH_CMD_BUFFER;
+		break;
+	case AMDGPU_UCODE_ID_P2S_TABLE:
+		*type = GFX_FW_TYPE_P2S_TABLE;
+		break;
+	case AMDGPU_UCODE_ID_JPEG_RAM:
+		*type = GFX_FW_TYPE_JPEG_RAM;
+		break;
+	case AMDGPU_UCODE_ID_ISP:
+		*type = GFX_FW_TYPE_ISP;
+		break;
 	case AMDGPU_UCODE_ID_MAXIMUM:
 	default:
 		return -EINVAL;
@@ -2344,7 +2899,8 @@ static void psp_print_fw_hdr(struct psp_context *psp,
 	}
 }
 
-static int psp_prep_load_ip_fw_cmd_buf(struct amdgpu_firmware_info *ucode,
+static int psp_prep_load_ip_fw_cmd_buf(struct psp_context *psp,
+				       struct amdgpu_firmware_info *ucode,
 				       struct psp_gfx_cmd_resp *cmd)
 {
 	int ret;
@@ -2357,18 +2913,18 @@ static int psp_prep_load_ip_fw_cmd_buf(struct amdgpu_firmware_info *ucode,
 
 	ret = psp_get_fw_type(ucode, &cmd->cmd.cmd_load_ip_fw.fw_type);
 	if (ret)
-		DRM_ERROR("Unknown firmware type\n");
+		dev_err(psp->adev->dev, "Unknown firmware type\n");
 
 	return ret;
 }
 
-static int psp_execute_non_psp_fw_load(struct psp_context *psp,
-			          struct amdgpu_firmware_info *ucode)
+int psp_execute_ip_fw_load(struct psp_context *psp,
+			   struct amdgpu_firmware_info *ucode)
 {
 	int ret = 0;
 	struct psp_gfx_cmd_resp *cmd = acquire_psp_cmd_buf(psp);
 
-	ret = psp_prep_load_ip_fw_cmd_buf(ucode, cmd);
+	ret = psp_prep_load_ip_fw_cmd_buf(psp, ucode, cmd);
 	if (!ret) {
 		ret = psp_cmd_submit_buf(psp, ucode, cmd,
 					 psp->fence_buf_mc_addr);
@@ -2379,6 +2935,33 @@ static int psp_execute_non_psp_fw_load(struct psp_context *psp,
 	return ret;
 }
 
+static int psp_load_p2s_table(struct psp_context *psp)
+{
+	int ret;
+	struct amdgpu_device *adev = psp->adev;
+	struct amdgpu_firmware_info *ucode =
+		&adev->firmware.ucode[AMDGPU_UCODE_ID_P2S_TABLE];
+
+	if (adev->in_runpm && ((adev->pm.rpm_mode == AMDGPU_RUNPM_BACO) ||
+				(adev->pm.rpm_mode == AMDGPU_RUNPM_BAMACO)))
+		return 0;
+
+	if (amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 6) ||
+	    amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 14)) {
+		uint32_t supp_vers = adev->flags & AMD_IS_APU ? 0x0036013D :
+								0x0036003C;
+		if (psp->sos.fw_version < supp_vers)
+			return 0;
+	}
+
+	if (!ucode->fw || amdgpu_sriov_vf(psp->adev))
+		return 0;
+
+	ret = psp_execute_ip_fw_load(psp, ucode);
+
+	return ret;
+}
+
 static int psp_load_smu_fw(struct psp_context *psp)
 {
 	int ret;
@@ -2391,26 +2974,25 @@ static int psp_load_smu_fw(struct psp_context *psp)
 	 * Skip SMU FW reloading in case of using BACO for runpm only,
 	 * as SMU is always alive.
 	 */
-	if (adev->in_runpm && (adev->pm.rpm_mode == AMDGPU_RUNPM_BACO))
+	if (adev->in_runpm && ((adev->pm.rpm_mode == AMDGPU_RUNPM_BACO) ||
+				(adev->pm.rpm_mode == AMDGPU_RUNPM_BAMACO)))
 		return 0;
 
 	if (!ucode->fw || amdgpu_sriov_vf(psp->adev))
 		return 0;
 
-	if ((amdgpu_in_reset(adev) &&
-	     ras && adev->ras_enabled &&
-	     (adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 4) ||
-	      adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 2)))) {
+	if ((amdgpu_in_reset(adev) && ras && adev->ras_enabled &&
+	     (amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(11, 0, 4) ||
+	      amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(11, 0, 2)))) {
 		ret = amdgpu_dpm_set_mp1_state(adev, PP_MP1_STATE_UNLOAD);
-		if (ret) {
-			DRM_WARN("Failed to set MP1 state prepare for reload\n");
-		}
+		if (ret)
+			dev_err(adev->dev, "Failed to set MP1 state prepare for reload\n");
 	}
 
-	ret = psp_execute_non_psp_fw_load(psp, ucode);
+	ret = psp_execute_ip_fw_load(psp, ucode);
 
 	if (ret)
-		DRM_ERROR("PSP load smu failed!\n");
+		dev_err(adev->dev, "PSP load smu failed!\n");
 
 	return ret;
 }
@@ -2421,6 +3003,9 @@ static bool fw_load_skip_check(struct psp_context *psp,
 	if (!ucode->fw || !ucode->ucode_size)
 		return true;
 
+	if (ucode->ucode_id == AMDGPU_UCODE_ID_P2S_TABLE)
+		return true;
+
 	if (ucode->ucode_id == AMDGPU_UCODE_ID_SMC &&
 	    (psp_smu_reload_quirk(psp) ||
 	     psp->autoload_supported ||
@@ -2449,7 +3034,7 @@ int psp_load_fw_list(struct psp_context *psp,
 	for (i = 0; i < ucode_count; ++i) {
 		ucode = ucode_list[i];
 		psp_print_fw_hdr(psp, ucode);
-		ret = psp_execute_non_psp_fw_load(psp, ucode);
+		ret = psp_execute_ip_fw_load(psp, ucode);
 		if (ret)
 			return ret;
 	}
@@ -2469,6 +3054,9 @@ static int psp_load_non_psp_fw(struct psp_context *psp)
 			return ret;
 	}
 
+	/* Load P2S table first if it's available */
+	psp_load_p2s_table(psp);
+
 	for (i = 0; i < adev->firmware.max_ucodes; i++) {
 		ucode = &adev->firmware.ucode[i];
 
@@ -2484,28 +3072,32 @@ static int psp_load_non_psp_fw(struct psp_context *psp)
 			continue;
 
 		if (psp->autoload_supported &&
-		    (adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 7) ||
-		     adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 11) ||
-		     adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 12)) &&
+		    (amdgpu_ip_version(adev, MP0_HWIP, 0) ==
+			     IP_VERSION(11, 0, 7) ||
+		     amdgpu_ip_version(adev, MP0_HWIP, 0) ==
+			     IP_VERSION(11, 0, 11) ||
+		     amdgpu_ip_version(adev, MP0_HWIP, 0) ==
+			     IP_VERSION(11, 0, 12)) &&
 		    (ucode->ucode_id == AMDGPU_UCODE_ID_SDMA1 ||
 		     ucode->ucode_id == AMDGPU_UCODE_ID_SDMA2 ||
 		     ucode->ucode_id == AMDGPU_UCODE_ID_SDMA3))
 			/* PSP only receive one SDMA fw for sienna_cichlid,
-			 * as all four sdma fw are same */
+			 * as all four sdma fw are same
+			 */
 			continue;
 
 		psp_print_fw_hdr(psp, ucode);
 
-		ret = psp_execute_non_psp_fw_load(psp, ucode);
+		ret = psp_execute_ip_fw_load(psp, ucode);
 		if (ret)
 			return ret;
 
-		/* Start rlc autoload after psp recieved all the gfx firmware */
+		/* Start rlc autoload after psp received all the gfx firmware */
 		if (psp->autoload_supported && ucode->ucode_id == (amdgpu_sriov_vf(adev) ?
 		    adev->virt.autoload_ucode_id : AMDGPU_UCODE_ID_RLC_G)) {
 			ret = psp_rlc_autoload_start(psp);
 			if (ret) {
-				DRM_ERROR("Failed to start rlc autoload\n");
+				dev_err(adev->dev, "Failed to start rlc autoload\n");
 				return ret;
 			}
 		}
@@ -2527,7 +3119,7 @@ static int psp_load_fw(struct amdgpu_device *adev)
 
 		ret = psp_ring_init(psp, PSP_RING_TYPE__KM);
 		if (ret) {
-			DRM_ERROR("PSP ring init failed!\n");
+			dev_err(adev->dev, "PSP ring init failed!\n");
 			goto failed;
 		}
 	}
@@ -2542,13 +3134,13 @@ static int psp_load_fw(struct amdgpu_device *adev)
 
 	ret = psp_asd_initialize(psp);
 	if (ret) {
-		DRM_ERROR("PSP load asd failed!\n");
+		dev_err(adev->dev, "PSP load asd failed!\n");
 		goto failed1;
 	}
 
 	ret = psp_rl_load(adev);
 	if (ret) {
-		DRM_ERROR("PSP load RL failed!\n");
+		dev_err(adev->dev, "PSP load RL failed!\n");
 		goto failed1;
 	}
 
@@ -2556,8 +3148,8 @@ static int psp_load_fw(struct amdgpu_device *adev)
 		if (adev->gmc.xgmi.num_physical_nodes > 1) {
 			ret = psp_xgmi_initialize(psp, false, true);
 			/* Warning the XGMI seesion initialize failure
-			* Instead of stop driver initialization
-			*/
+			 * Instead of stop driver initialization
+			 */
 			if (ret)
 				dev_err(psp->adev->dev,
 					"XGMI: Failed to initialize XGMI session\n");
@@ -2568,7 +3160,7 @@ static int psp_load_fw(struct amdgpu_device *adev)
 		ret = psp_ras_initialize(psp);
 		if (ret)
 			dev_err(psp->adev->dev,
-					"RAS: Failed to initialize RAS\n");
+				"RAS: Failed to initialize RAS\n");
 
 		ret = psp_hdcp_initialize(psp);
 		if (ret)
@@ -2605,23 +3197,20 @@ failed:
 	return ret;
 }
 
-static int psp_hw_init(void *handle)
+static int psp_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int ret;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	mutex_lock(&adev->firmware.mutex);
-	/*
-	 * This sequence is just used on hw_init only once, no need on
-	 * resume.
-	 */
+
 	ret = amdgpu_ucode_init_bo(adev);
 	if (ret)
 		goto failed;
 
 	ret = psp_load_fw(adev);
 	if (ret) {
-		DRM_ERROR("PSP firmware loading failed\n");
+		dev_err(adev->dev, "PSP firmware loading failed\n");
 		goto failed;
 	}
 
@@ -2634,9 +3223,9 @@ failed:
 	return -EINVAL;
 }
 
-static int psp_hw_fini(void *handle)
+static int psp_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct psp_context *psp = &adev->psp;
 
 	if (psp->ta_fw) {
@@ -2655,22 +3244,20 @@ static int psp_hw_fini(void *handle)
 
 	psp_ring_destroy(psp, PSP_RING_TYPE__KM);
 
-	psp_free_shared_bufs(psp);
-
 	return 0;
 }
 
-static int psp_suspend(void *handle)
+static int psp_suspend(struct amdgpu_ip_block *ip_block)
 {
 	int ret = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct psp_context *psp = &adev->psp;
 
 	if (adev->gmc.xgmi.num_physical_nodes > 1 &&
 	    psp->xgmi_context.context.initialized) {
 		ret = psp_xgmi_terminate(psp);
 		if (ret) {
-			DRM_ERROR("Failed to terminate xgmi ta\n");
+			dev_err(adev->dev, "Failed to terminate xgmi ta\n");
 			goto out;
 		}
 	}
@@ -2678,70 +3265,73 @@ static int psp_suspend(void *handle)
 	if (psp->ta_fw) {
 		ret = psp_ras_terminate(psp);
 		if (ret) {
-			DRM_ERROR("Failed to terminate ras ta\n");
+			dev_err(adev->dev, "Failed to terminate ras ta\n");
 			goto out;
 		}
 		ret = psp_hdcp_terminate(psp);
 		if (ret) {
-			DRM_ERROR("Failed to terminate hdcp ta\n");
+			dev_err(adev->dev, "Failed to terminate hdcp ta\n");
 			goto out;
 		}
 		ret = psp_dtm_terminate(psp);
 		if (ret) {
-			DRM_ERROR("Failed to terminate dtm ta\n");
+			dev_err(adev->dev, "Failed to terminate dtm ta\n");
 			goto out;
 		}
 		ret = psp_rap_terminate(psp);
 		if (ret) {
-			DRM_ERROR("Failed to terminate rap ta\n");
+			dev_err(adev->dev, "Failed to terminate rap ta\n");
 			goto out;
 		}
 		ret = psp_securedisplay_terminate(psp);
 		if (ret) {
-			DRM_ERROR("Failed to terminate securedisplay ta\n");
+			dev_err(adev->dev, "Failed to terminate securedisplay ta\n");
 			goto out;
 		}
 	}
 
 	ret = psp_asd_terminate(psp);
 	if (ret) {
-		DRM_ERROR("Failed to terminate asd\n");
+		dev_err(adev->dev, "Failed to terminate asd\n");
 		goto out;
 	}
 
 	ret = psp_tmr_terminate(psp);
 	if (ret) {
-		DRM_ERROR("Failed to terminate tmr\n");
+		dev_err(adev->dev, "Failed to terminate tmr\n");
 		goto out;
 	}
 
 	ret = psp_ring_stop(psp, PSP_RING_TYPE__KM);
-	if (ret) {
-		DRM_ERROR("PSP ring stop failed\n");
-	}
+	if (ret)
+		dev_err(adev->dev, "PSP ring stop failed\n");
 
 out:
 	return ret;
 }
 
-static int psp_resume(void *handle)
+static int psp_resume(struct amdgpu_ip_block *ip_block)
 {
 	int ret;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct psp_context *psp = &adev->psp;
 
-	DRM_INFO("PSP is resuming...\n");
+	dev_info(adev->dev, "PSP is resuming...\n");
 
 	if (psp->mem_train_ctx.enable_mem_training) {
 		ret = psp_mem_training(psp, PSP_MEM_TRAIN_RESUME);
 		if (ret) {
-			DRM_ERROR("Failed to process memory training!\n");
+			dev_err(adev->dev, "Failed to process memory training!\n");
 			return ret;
 		}
 	}
 
 	mutex_lock(&adev->firmware.mutex);
 
+	ret = amdgpu_ucode_init_bo(adev);
+	if (ret)
+		goto failed;
+
 	ret = psp_hw_start(psp);
 	if (ret)
 		goto failed;
@@ -2752,7 +3342,7 @@ static int psp_resume(void *handle)
 
 	ret = psp_asd_initialize(psp);
 	if (ret) {
-		DRM_ERROR("PSP load asd failed!\n");
+		dev_err(adev->dev, "PSP load asd failed!\n");
 		goto failed;
 	}
 
@@ -2776,7 +3366,7 @@ static int psp_resume(void *handle)
 		ret = psp_ras_initialize(psp);
 		if (ret)
 			dev_err(psp->adev->dev,
-					"RAS: Failed to initialize RAS\n");
+				"RAS: Failed to initialize RAS\n");
 
 		ret = psp_hdcp_initialize(psp);
 		if (ret)
@@ -2804,7 +3394,7 @@ static int psp_resume(void *handle)
 	return 0;
 
 failed:
-	DRM_ERROR("PSP resume failed\n");
+	dev_err(adev->dev, "PSP resume failed\n");
 	mutex_unlock(&adev->firmware.mutex);
 	return ret;
 }
@@ -2838,19 +3428,6 @@ int psp_rlc_autoload_start(struct psp_context *psp)
 	return ret;
 }
 
-int psp_update_vcn_sram(struct amdgpu_device *adev, int inst_idx,
-			uint64_t cmd_gpu_addr, int cmd_size)
-{
-	struct amdgpu_firmware_info ucode = {0};
-
-	ucode.ucode_id = inst_idx ? AMDGPU_UCODE_ID_VCN1_RAM :
-		AMDGPU_UCODE_ID_VCN0_RAM;
-	ucode.mc_addr = cmd_gpu_addr;
-	ucode.ucode_size = cmd_size;
-
-	return psp_execute_non_psp_fw_load(&adev->psp, &ucode);
-}
-
 int psp_ring_cmd_submit(struct psp_context *psp,
 			uint64_t cmd_buf_mc_addr,
 			uint64_t fence_mc_addr,
@@ -2878,9 +3455,11 @@ int psp_ring_cmd_submit(struct psp_context *psp,
 		write_frame = ring_buffer_start + (psp_write_ptr_reg / rb_frame_size_dw);
 	/* Check invalid write_frame ptr address */
 	if ((write_frame < ring_buffer_start) || (ring_buffer_end < write_frame)) {
-		DRM_ERROR("ring_buffer_start = %p; ring_buffer_end = %p; write_frame = %p\n",
-			  ring_buffer_start, ring_buffer_end, write_frame);
-		DRM_ERROR("write_frame is pointing to address out of bounds\n");
+		dev_err(adev->dev,
+			"ring_buffer_start = %p; ring_buffer_end = %p; write_frame = %p\n",
+			ring_buffer_start, ring_buffer_end, write_frame);
+		dev_err(adev->dev,
+			"write_frame is pointing to address out of bounds\n");
 		return -EINVAL;
 	}
 
@@ -2904,12 +3483,11 @@ int psp_ring_cmd_submit(struct psp_context *psp,
 int psp_init_asd_microcode(struct psp_context *psp, const char *chip_name)
 {
 	struct amdgpu_device *adev = psp->adev;
-	char fw_name[PSP_FW_NAME_LEN];
 	const struct psp_firmware_header_v1_0 *asd_hdr;
 	int err = 0;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_asd.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->psp.asd_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->psp.asd_fw, AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_asd.bin", chip_name);
 	if (err)
 		goto out;
 
@@ -2928,12 +3506,11 @@ out:
 int psp_init_toc_microcode(struct psp_context *psp, const char *chip_name)
 {
 	struct amdgpu_device *adev = psp->adev;
-	char fw_name[PSP_FW_NAME_LEN];
 	const struct psp_firmware_header_v1_0 *toc_hdr;
 	int err = 0;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_toc.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->psp.toc_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->psp.toc_fw, AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_toc.bin", chip_name);
 	if (err)
 		goto out;
 
@@ -2967,7 +3544,7 @@ static int parse_sos_bin_descriptor(struct psp_context *psp,
 		psp->sos.fw_version        = le32_to_cpu(desc->fw_version);
 		psp->sos.feature_version   = le32_to_cpu(desc->fw_version);
 		psp->sos.size_bytes        = le32_to_cpu(desc->size_bytes);
-		psp->sos.start_addr 	   = ucode_start_addr;
+		psp->sos.start_addr	   = ucode_start_addr;
 		break;
 	case PSP_FW_TYPE_PSP_SYS_DRV:
 		psp->sys.fw_version        = le32_to_cpu(desc->fw_version);
@@ -3023,6 +3600,18 @@ static int parse_sos_bin_descriptor(struct psp_context *psp,
 		psp->ras_drv.size_bytes         = le32_to_cpu(desc->size_bytes);
 		psp->ras_drv.start_addr         = ucode_start_addr;
 		break;
+	case PSP_FW_TYPE_PSP_IPKEYMGR_DRV:
+		psp->ipkeymgr_drv.fw_version         = le32_to_cpu(desc->fw_version);
+		psp->ipkeymgr_drv.feature_version    = le32_to_cpu(desc->fw_version);
+		psp->ipkeymgr_drv.size_bytes         = le32_to_cpu(desc->size_bytes);
+		psp->ipkeymgr_drv.start_addr         = ucode_start_addr;
+		break;
+	case PSP_FW_TYPE_PSP_SPDM_DRV:
+		psp->spdm_drv.fw_version	= le32_to_cpu(desc->fw_version);
+		psp->spdm_drv.feature_version	= le32_to_cpu(desc->fw_version);
+		psp->spdm_drv.size_bytes	= le32_to_cpu(desc->size_bytes);
+		psp->spdm_drv.start_addr	= ucode_start_addr;
+		break;
 	default:
 		dev_warn(psp->adev->dev, "Unsupported PSP FW type: %d\n", desc->fw_type);
 		break;
@@ -3042,7 +3631,7 @@ static int psp_init_sos_base_fw(struct amdgpu_device *adev)
 		le32_to_cpu(sos_hdr->header.ucode_array_offset_bytes);
 
 	if (adev->gmc.xgmi.connected_to_cpu ||
-	    (adev->ip_versions[MP0_HWIP][0] != IP_VERSION(13, 0, 2))) {
+	    (amdgpu_ip_version(adev, MP0_HWIP, 0) != IP_VERSION(13, 0, 2))) {
 		adev->psp.sos.fw_version = le32_to_cpu(sos_hdr->header.ucode_version);
 		adev->psp.sos.feature_version = le32_to_cpu(sos_hdr->sos.fw_version);
 
@@ -3079,18 +3668,23 @@ static int psp_init_sos_base_fw(struct amdgpu_device *adev)
 int psp_init_sos_microcode(struct psp_context *psp, const char *chip_name)
 {
 	struct amdgpu_device *adev = psp->adev;
-	char fw_name[PSP_FW_NAME_LEN];
 	const struct psp_firmware_header_v1_0 *sos_hdr;
 	const struct psp_firmware_header_v1_1 *sos_hdr_v1_1;
 	const struct psp_firmware_header_v1_2 *sos_hdr_v1_2;
 	const struct psp_firmware_header_v1_3 *sos_hdr_v1_3;
 	const struct psp_firmware_header_v2_0 *sos_hdr_v2_0;
-	int err = 0;
+	const struct psp_firmware_header_v2_1 *sos_hdr_v2_1;
+	int fw_index, fw_bin_count, start_index = 0;
+	const struct psp_fw_bin_desc *fw_bin;
 	uint8_t *ucode_array_start_addr;
-	int fw_index = 0;
+	int err = 0;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sos.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->psp.sos_fw, fw_name);
+	if (amdgpu_is_kicker_fw(adev))
+		err = amdgpu_ucode_request(adev, &adev->psp.sos_fw, AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_sos_kicker.bin", chip_name);
+	else
+		err = amdgpu_ucode_request(adev, &adev->psp.sos_fw, AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_sos.bin", chip_name);
 	if (err)
 		goto out;
 
@@ -3139,15 +3733,30 @@ int psp_init_sos_microcode(struct psp_context *psp, const char *chip_name)
 	case 2:
 		sos_hdr_v2_0 = (const struct psp_firmware_header_v2_0 *)adev->psp.sos_fw->data;
 
-		if (le32_to_cpu(sos_hdr_v2_0->psp_fw_bin_count) >= UCODE_MAX_PSP_PACKAGING) {
+		fw_bin_count = le32_to_cpu(sos_hdr_v2_0->psp_fw_bin_count);
+
+		if (fw_bin_count >= UCODE_MAX_PSP_PACKAGING) {
 			dev_err(adev->dev, "packed SOS count exceeds maximum limit\n");
 			err = -EINVAL;
 			goto out;
 		}
 
-		for (fw_index = 0; fw_index < le32_to_cpu(sos_hdr_v2_0->psp_fw_bin_count); fw_index++) {
-			err = parse_sos_bin_descriptor(psp,
-						       &sos_hdr_v2_0->psp_fw_bin[fw_index],
+		if (sos_hdr_v2_0->header.header_version_minor == 1) {
+			sos_hdr_v2_1 = (const struct psp_firmware_header_v2_1 *)adev->psp.sos_fw->data;
+
+			fw_bin = sos_hdr_v2_1->psp_fw_bin;
+
+			if (psp_is_aux_sos_load_required(psp))
+				start_index = le32_to_cpu(sos_hdr_v2_1->psp_aux_fw_bin_index);
+			else
+				fw_bin_count -= le32_to_cpu(sos_hdr_v2_1->psp_aux_fw_bin_index);
+
+		} else {
+			fw_bin = sos_hdr_v2_0->psp_fw_bin;
+		}
+
+		for (fw_index = start_index; fw_index < fw_bin_count; fw_index++) {
+			err = parse_sos_bin_descriptor(psp, fw_bin + fw_index,
 						       sos_hdr_v2_0);
 			if (err)
 				goto out;
@@ -3167,6 +3776,36 @@ out:
 	return err;
 }
 
+static bool is_ta_fw_applicable(struct psp_context *psp,
+			     const struct psp_fw_bin_desc *desc)
+{
+	struct amdgpu_device *adev = psp->adev;
+	uint32_t fw_version;
+
+	switch (desc->fw_type) {
+	case TA_FW_TYPE_PSP_XGMI:
+	case TA_FW_TYPE_PSP_XGMI_AUX:
+		/* for now, AUX TA only exists on 13.0.6 ta bin,
+		 * from v20.00.0x.14
+		 */
+		if (amdgpu_ip_version(adev, MP0_HWIP, 0) ==
+		    IP_VERSION(13, 0, 6)) {
+			fw_version = le32_to_cpu(desc->fw_version);
+
+			if (adev->flags & AMD_IS_APU &&
+			    (fw_version & 0xff) >= 0x14)
+				return desc->fw_type == TA_FW_TYPE_PSP_XGMI_AUX;
+			else
+				return desc->fw_type == TA_FW_TYPE_PSP_XGMI;
+		}
+		break;
+	default:
+		break;
+	}
+
+	return true;
+}
+
 static int parse_ta_bin_descriptor(struct psp_context *psp,
 				   const struct psp_fw_bin_desc *desc,
 				   const struct ta_firmware_header_v2_0 *ta_hdr)
@@ -3176,6 +3815,9 @@ static int parse_ta_bin_descriptor(struct psp_context *psp,
 	if (!psp || !desc || !ta_hdr)
 		return -EINVAL;
 
+	if (!is_ta_fw_applicable(psp, desc))
+		return 0;
+
 	ucode_start_addr  = (uint8_t *)ta_hdr +
 			    le32_to_cpu(desc->offset_bytes) +
 			    le32_to_cpu(ta_hdr->header.ucode_array_offset_bytes);
@@ -3188,6 +3830,7 @@ static int parse_ta_bin_descriptor(struct psp_context *psp,
 		psp->asd_context.bin_desc.start_addr        = ucode_start_addr;
 		break;
 	case TA_FW_TYPE_PSP_XGMI:
+	case TA_FW_TYPE_PSP_XGMI_AUX:
 		psp->xgmi_context.context.bin_desc.fw_version       = le32_to_cpu(desc->fw_version);
 		psp->xgmi_context.context.bin_desc.size_bytes       = le32_to_cpu(desc->size_bytes);
 		psp->xgmi_context.context.bin_desc.start_addr       = ucode_start_addr;
@@ -3315,11 +3958,14 @@ int psp_init_ta_microcode(struct psp_context *psp, const char *chip_name)
 {
 	const struct common_firmware_header *hdr;
 	struct amdgpu_device *adev = psp->adev;
-	char fw_name[PSP_FW_NAME_LEN];
 	int err;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ta.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->psp.ta_fw, fw_name);
+	if (amdgpu_is_kicker_fw(adev))
+		err = amdgpu_ucode_request(adev, &adev->psp.ta_fw, AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_ta_kicker.bin", chip_name);
+	else
+		err = amdgpu_ucode_request(adev, &adev->psp.ta_fw, AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_ta.bin", chip_name);
 	if (err)
 		return err;
 
@@ -3345,7 +3991,6 @@ int psp_init_ta_microcode(struct psp_context *psp, const char *chip_name)
 int psp_init_cap_microcode(struct psp_context *psp, const char *chip_name)
 {
 	struct amdgpu_device *adev = psp->adev;
-	char fw_name[PSP_FW_NAME_LEN];
 	const struct psp_firmware_header_v1_0 *cap_hdr_v1_0;
 	struct amdgpu_firmware_info *info = NULL;
 	int err = 0;
@@ -3355,15 +4000,16 @@ int psp_init_cap_microcode(struct psp_context *psp, const char *chip_name)
 		return -EINVAL;
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_cap.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->psp.cap_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->psp.cap_fw, AMDGPU_UCODE_OPTIONAL,
+				   "amdgpu/%s_cap.bin", chip_name);
 	if (err) {
 		if (err == -ENODEV) {
 			dev_warn(adev->dev, "cap microcode does not exist, skip\n");
 			err = 0;
-			goto out;
+		} else {
+			dev_err(adev->dev, "fail to initialize cap microcode\n");
 		}
-		dev_err(adev->dev, "fail to initialize cap microcode\n");
+		goto out;
 	}
 
 	info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CAP];
@@ -3384,13 +4030,49 @@ out:
 	return err;
 }
 
-static int psp_set_clockgating_state(void *handle,
-				     enum amd_clockgating_state state)
+int psp_config_sq_perfmon(struct psp_context *psp,
+		uint32_t xcp_id, bool core_override_enable,
+		bool reg_override_enable, bool perfmon_override_enable)
+{
+	int ret;
+
+	if (amdgpu_sriov_vf(psp->adev))
+		return 0;
+
+	if (xcp_id > MAX_XCP) {
+		dev_err(psp->adev->dev, "invalid xcp_id %d\n", xcp_id);
+		return -EINVAL;
+	}
+
+	if (amdgpu_ip_version(psp->adev, MP0_HWIP, 0) != IP_VERSION(13, 0, 6)) {
+		dev_err(psp->adev->dev, "Unsupported MP0 version 0x%x for CONFIG_SQ_PERFMON command\n",
+			amdgpu_ip_version(psp->adev, MP0_HWIP, 0));
+		return -EINVAL;
+	}
+	struct psp_gfx_cmd_resp *cmd = acquire_psp_cmd_buf(psp);
+
+	cmd->cmd_id	=	GFX_CMD_ID_CONFIG_SQ_PERFMON;
+	cmd->cmd.config_sq_perfmon.gfx_xcp_mask	=	BIT_MASK(xcp_id);
+	cmd->cmd.config_sq_perfmon.core_override	=	core_override_enable;
+	cmd->cmd.config_sq_perfmon.reg_override	=	reg_override_enable;
+	cmd->cmd.config_sq_perfmon.perfmon_override = perfmon_override_enable;
+
+	ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr);
+	if (ret)
+		dev_warn(psp->adev->dev, "PSP failed to config sq: xcp%d core%d reg%d perfmon%d\n",
+			xcp_id, core_override_enable, reg_override_enable, perfmon_override_enable);
+
+	release_psp_cmd_buf(psp);
+	return ret;
+}
+
+static int psp_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int psp_set_powergating_state(void *handle,
+static int psp_set_powergating_state(struct amdgpu_ip_block *ip_block,
 				     enum amd_powergating_state state)
 {
 	return 0;
@@ -3402,11 +4084,13 @@ static ssize_t psp_usbc_pd_fw_sysfs_read(struct device *dev,
 {
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = drm_to_adev(ddev);
+	struct amdgpu_ip_block *ip_block;
 	uint32_t fw_ver;
 	int ret;
 
-	if (!adev->ip_blocks[AMD_IP_BLOCK_TYPE_PSP].status.late_initialized) {
-		DRM_INFO("PSP block is not ready yet.");
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_PSP);
+	if (!ip_block || !ip_block->status.late_initialized) {
+		dev_info(adev->dev, "PSP block is not ready yet\n.");
 		return -EBUSY;
 	}
 
@@ -3415,7 +4099,7 @@ static ssize_t psp_usbc_pd_fw_sysfs_read(struct device *dev,
 	mutex_unlock(&adev->psp.mutex);
 
 	if (ret) {
-		DRM_ERROR("Failed to read USBC PD FW, err = %d", ret);
+		dev_err(adev->dev, "Failed to read USBC PD FW, err = %d\n", ret);
 		return ret;
 	}
 
@@ -3430,22 +4114,23 @@ static ssize_t psp_usbc_pd_fw_sysfs_write(struct device *dev,
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = drm_to_adev(ddev);
 	int ret, idx;
-	char fw_name[100];
 	const struct firmware *usbc_pd_fw;
 	struct amdgpu_bo *fw_buf_bo = NULL;
 	uint64_t fw_pri_mc_addr;
 	void *fw_pri_cpu_addr;
+	struct amdgpu_ip_block *ip_block;
 
-	if (!adev->ip_blocks[AMD_IP_BLOCK_TYPE_PSP].status.late_initialized) {
-		DRM_INFO("PSP block is not ready yet.");
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_PSP);
+	if (!ip_block || !ip_block->status.late_initialized) {
+		dev_err(adev->dev, "PSP block is not ready yet.");
 		return -EBUSY;
 	}
 
 	if (!drm_dev_enter(ddev, &idx))
 		return -ENODEV;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s", buf);
-	ret = request_firmware(&usbc_pd_fw, fw_name, adev->dev);
+	ret = amdgpu_ucode_request(adev, &usbc_pd_fw, AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s", buf);
 	if (ret)
 		goto fail;
 
@@ -3467,10 +4152,10 @@ static ssize_t psp_usbc_pd_fw_sysfs_write(struct device *dev,
 	amdgpu_bo_free_kernel(&fw_buf_bo, &fw_pri_mc_addr, &fw_pri_cpu_addr);
 
 rel_buf:
-	release_firmware(usbc_pd_fw);
+	amdgpu_ucode_release(&usbc_pd_fw);
 fail:
 	if (ret) {
-		DRM_ERROR("Failed to load USBC PD FW, err = %d", ret);
+		dev_err(adev->dev, "Failed to load USBC PD FW, err = %d", ret);
 		count = ret;
 	}
 
@@ -3491,7 +4176,12 @@ void psp_copy_fw(struct psp_context *psp, uint8_t *start_addr, uint32_t bin_size
 	drm_dev_exit(idx);
 }
 
-static DEVICE_ATTR(usbc_pd_fw, S_IRUGO | S_IWUSR,
+/**
+ * DOC: usbc_pd_fw
+ * Reading from this file will retrieve the USB-C PD firmware version. Writing to
+ * this file will trigger the update process.
+ */
+static DEVICE_ATTR(usbc_pd_fw, 0644,
 		   psp_usbc_pd_fw_sysfs_read,
 		   psp_usbc_pd_fw_sysfs_write);
 
@@ -3501,7 +4191,7 @@ int is_psp_fw_valid(struct psp_bin_desc bin)
 }
 
 static ssize_t amdgpu_psp_vbflash_write(struct file *filp, struct kobject *kobj,
-					struct bin_attribute *bin_attr,
+					const struct bin_attribute *bin_attr,
 					char *buffer, loff_t pos, size_t count)
 {
 	struct device *dev = kobj_to_dev(kobj);
@@ -3512,7 +4202,7 @@ static ssize_t amdgpu_psp_vbflash_write(struct file *filp, struct kobject *kobj,
 
 	/* Safeguard against memory drain */
 	if (adev->psp.vbflash_image_size > AMD_VBIOS_FILE_MAX_SIZE_B) {
-		dev_err(adev->dev, "File size cannot exceed %u", AMD_VBIOS_FILE_MAX_SIZE_B);
+		dev_err(adev->dev, "File size cannot exceed %u\n", AMD_VBIOS_FILE_MAX_SIZE_B);
 		kvfree(adev->psp.vbflash_tmp_buf);
 		adev->psp.vbflash_tmp_buf = NULL;
 		adev->psp.vbflash_image_size = 0;
@@ -3531,13 +4221,13 @@ static ssize_t amdgpu_psp_vbflash_write(struct file *filp, struct kobject *kobj,
 	adev->psp.vbflash_image_size += count;
 	mutex_unlock(&adev->psp.mutex);
 
-	dev_info(adev->dev, "VBIOS flash write PSP done");
+	dev_dbg(adev->dev, "IFWI staged for update\n");
 
 	return count;
 }
 
 static ssize_t amdgpu_psp_vbflash_read(struct file *filp, struct kobject *kobj,
-				       struct bin_attribute *bin_attr, char *buffer,
+				       const struct bin_attribute *bin_attr, char *buffer,
 				       loff_t pos, size_t count)
 {
 	struct device *dev = kobj_to_dev(kobj);
@@ -3548,7 +4238,10 @@ static ssize_t amdgpu_psp_vbflash_read(struct file *filp, struct kobject *kobj,
 	void *fw_pri_cpu_addr;
 	int ret;
 
-	dev_info(adev->dev, "VBIOS flash to PSP started");
+	if (adev->psp.vbflash_image_size == 0)
+		return -EINVAL;
+
+	dev_dbg(adev->dev, "PSP IFWI flash process initiated\n");
 
 	ret = amdgpu_bo_create_kernel(adev, adev->psp.vbflash_image_size,
 					AMDGPU_GPU_PAGE_SIZE,
@@ -3573,14 +4266,32 @@ rel_buf:
 	adev->psp.vbflash_image_size = 0;
 
 	if (ret) {
-		dev_err(adev->dev, "Failed to load VBIOS FW, err = %d", ret);
+		dev_err(adev->dev, "Failed to load IFWI, err = %d\n", ret);
 		return ret;
 	}
 
-	dev_info(adev->dev, "VBIOS flash to PSP done");
+	dev_dbg(adev->dev, "PSP IFWI flash process done\n");
 	return 0;
 }
 
+/**
+ * DOC: psp_vbflash
+ * Writing to this file will stage an IFWI for update. Reading from this file
+ * will trigger the update process.
+ */
+static const struct bin_attribute psp_vbflash_bin_attr = {
+	.attr = {.name = "psp_vbflash", .mode = 0660},
+	.size = 0,
+	.write = amdgpu_psp_vbflash_write,
+	.read = amdgpu_psp_vbflash_read,
+};
+
+/**
+ * DOC: psp_vbflash_status
+ * The status of the flash process.
+ * 0: IFWI flash not complete.
+ * 1: IFWI flash complete.
+ */
 static ssize_t amdgpu_psp_vbflash_status(struct device *dev,
 					 struct device_attribute *attr,
 					 char *buf)
@@ -3597,84 +4308,167 @@ static ssize_t amdgpu_psp_vbflash_status(struct device *dev,
 
 	return sysfs_emit(buf, "0x%x\n", vbflash_status);
 }
+static DEVICE_ATTR(psp_vbflash_status, 0440, amdgpu_psp_vbflash_status, NULL);
 
-static const struct bin_attribute psp_vbflash_bin_attr = {
-	.attr = {.name = "psp_vbflash", .mode = 0664},
-	.size = 0,
-	.write = amdgpu_psp_vbflash_write,
-	.read = amdgpu_psp_vbflash_read,
+static const struct bin_attribute *const bin_flash_attrs[] = {
+	&psp_vbflash_bin_attr,
+	NULL
 };
 
-static DEVICE_ATTR(psp_vbflash_status, 0444, amdgpu_psp_vbflash_status, NULL);
+static struct attribute *flash_attrs[] = {
+	&dev_attr_psp_vbflash_status.attr,
+	&dev_attr_usbc_pd_fw.attr,
+	NULL
+};
 
-int amdgpu_psp_sysfs_init(struct amdgpu_device *adev)
+static umode_t amdgpu_flash_attr_is_visible(struct kobject *kobj, struct attribute *attr, int idx)
 {
-	int ret = 0;
-	struct psp_context *psp = &adev->psp;
+	struct device *dev = kobj_to_dev(kobj);
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
 
-	if (amdgpu_sriov_vf(adev))
-		return -EINVAL;
+	if (attr == &dev_attr_usbc_pd_fw.attr)
+		return adev->psp.sup_pd_fw_up ? 0660 : 0;
 
-	switch (adev->ip_versions[MP0_HWIP][0]) {
-	case IP_VERSION(13, 0, 0):
-	case IP_VERSION(13, 0, 7):
-		if (!psp->adev) {
-			psp->adev = adev;
-			psp_v13_0_set_psp_funcs(psp);
-		}
-		ret = sysfs_create_bin_file(&adev->dev->kobj, &psp_vbflash_bin_attr);
-		if (ret)
-			dev_err(adev->dev, "Failed to create device file psp_vbflash");
-		ret = device_create_file(adev->dev, &dev_attr_psp_vbflash_status);
-		if (ret)
-			dev_err(adev->dev, "Failed to create device file psp_vbflash_status");
-		return ret;
-	default:
-		return 0;
-	}
+	return adev->psp.sup_ifwi_up ? 0440 : 0;
 }
 
-const struct amd_ip_funcs psp_ip_funcs = {
-	.name = "psp",
-	.early_init = psp_early_init,
-	.late_init = NULL,
-	.sw_init = psp_sw_init,
-	.sw_fini = psp_sw_fini,
-	.hw_init = psp_hw_init,
-	.hw_fini = psp_hw_fini,
-	.suspend = psp_suspend,
-	.resume = psp_resume,
-	.is_idle = NULL,
-	.check_soft_reset = NULL,
-	.wait_for_idle = NULL,
-	.soft_reset = NULL,
-	.set_clockgating_state = psp_set_clockgating_state,
-	.set_powergating_state = psp_set_powergating_state,
+static umode_t amdgpu_bin_flash_attr_is_visible(struct kobject *kobj,
+						const struct bin_attribute *attr,
+						int idx)
+{
+	struct device *dev = kobj_to_dev(kobj);
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	return adev->psp.sup_ifwi_up ? 0660 : 0;
+}
+
+const struct attribute_group amdgpu_flash_attr_group = {
+	.attrs = flash_attrs,
+	.bin_attrs = bin_flash_attrs,
+	.is_bin_visible = amdgpu_bin_flash_attr_is_visible,
+	.is_visible = amdgpu_flash_attr_is_visible,
 };
 
-static int psp_sysfs_init(struct amdgpu_device *adev)
+#if defined(CONFIG_DEBUG_FS)
+static int psp_read_spirom_debugfs_open(struct inode *inode, struct file *filp)
 {
-	int ret = device_create_file(adev->dev, &dev_attr_usbc_pd_fw);
+	struct amdgpu_device *adev = filp->f_inode->i_private;
+	struct spirom_bo *bo_triplet;
+	int ret;
 
+	/* serialize the open() file calling */
+	if (!mutex_trylock(&adev->psp.mutex))
+		return -EBUSY;
+
+	/*
+	 * make sure only one userpace process is alive for dumping so that
+	 * only one memory buffer of AMD_VBIOS_FILE_MAX_SIZE * 2 is consumed.
+	 * let's say the case where one process try opening the file while
+	 * another one has proceeded to read or release. In this way, eliminate
+	 * the use of mutex for read() or release() callback as well.
+	 */
+	if (adev->psp.spirom_dump_trip) {
+		mutex_unlock(&adev->psp.mutex);
+		return -EBUSY;
+	}
+
+	bo_triplet = kzalloc(sizeof(struct spirom_bo), GFP_KERNEL);
+	if (!bo_triplet) {
+		mutex_unlock(&adev->psp.mutex);
+		return -ENOMEM;
+	}
+
+	ret = amdgpu_bo_create_kernel(adev, AMD_VBIOS_FILE_MAX_SIZE_B * 2,
+				      AMDGPU_GPU_PAGE_SIZE,
+				      AMDGPU_GEM_DOMAIN_GTT,
+				      &bo_triplet->bo,
+				      &bo_triplet->mc_addr,
+				      &bo_triplet->cpu_addr);
+	if (ret)
+		goto rel_trip;
+
+	ret = psp_dump_spirom(&adev->psp, bo_triplet->mc_addr);
 	if (ret)
-		DRM_ERROR("Failed to create USBC PD FW control file!");
+		goto rel_bo;
 
+	adev->psp.spirom_dump_trip = bo_triplet;
+	mutex_unlock(&adev->psp.mutex);
+	return 0;
+rel_bo:
+	amdgpu_bo_free_kernel(&bo_triplet->bo, &bo_triplet->mc_addr,
+			      &bo_triplet->cpu_addr);
+rel_trip:
+	kfree(bo_triplet);
+	mutex_unlock(&adev->psp.mutex);
+	dev_err(adev->dev, "Trying IFWI dump fails, err = %d\n", ret);
 	return ret;
 }
 
-void amdgpu_psp_sysfs_fini(struct amdgpu_device *adev)
+static ssize_t psp_read_spirom_debugfs_read(struct file *filp, char __user *buf, size_t size,
+					    loff_t *pos)
 {
-	sysfs_remove_bin_file(&adev->dev->kobj, &psp_vbflash_bin_attr);
-	device_remove_file(adev->dev, &dev_attr_psp_vbflash_status);
+	struct amdgpu_device *adev = filp->f_inode->i_private;
+	struct spirom_bo *bo_triplet = adev->psp.spirom_dump_trip;
+
+	if (!bo_triplet)
+		return -EINVAL;
+
+	return simple_read_from_buffer(buf,
+				       size,
+				       pos, bo_triplet->cpu_addr,
+				       AMD_VBIOS_FILE_MAX_SIZE_B * 2);
 }
 
-static void psp_sysfs_fini(struct amdgpu_device *adev)
+static int psp_read_spirom_debugfs_release(struct inode *inode, struct file *filp)
 {
-	device_remove_file(adev->dev, &dev_attr_usbc_pd_fw);
+	struct amdgpu_device *adev = filp->f_inode->i_private;
+	struct spirom_bo *bo_triplet = adev->psp.spirom_dump_trip;
+
+	if (bo_triplet) {
+		amdgpu_bo_free_kernel(&bo_triplet->bo, &bo_triplet->mc_addr,
+				      &bo_triplet->cpu_addr);
+		kfree(bo_triplet);
+	}
+
+	adev->psp.spirom_dump_trip = NULL;
+	return 0;
 }
 
-const struct amdgpu_ip_block_version psp_v3_1_ip_block =
+static const struct file_operations psp_dump_spirom_debugfs_ops = {
+	.owner = THIS_MODULE,
+	.open = psp_read_spirom_debugfs_open,
+	.read = psp_read_spirom_debugfs_read,
+	.release = psp_read_spirom_debugfs_release,
+	.llseek = default_llseek,
+};
+#endif
+
+void amdgpu_psp_debugfs_init(struct amdgpu_device *adev)
 {
+#if defined(CONFIG_DEBUG_FS)
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+
+	debugfs_create_file_size("psp_spirom_dump", 0444, minor->debugfs_root,
+				 adev, &psp_dump_spirom_debugfs_ops, AMD_VBIOS_FILE_MAX_SIZE_B * 2);
+#endif
+}
+
+const struct amd_ip_funcs psp_ip_funcs = {
+	.name = "psp",
+	.early_init = psp_early_init,
+	.sw_init = psp_sw_init,
+	.sw_fini = psp_sw_fini,
+	.hw_init = psp_hw_init,
+	.hw_fini = psp_hw_fini,
+	.suspend = psp_suspend,
+	.resume = psp_resume,
+	.set_clockgating_state = psp_set_clockgating_state,
+	.set_powergating_state = psp_set_powergating_state,
+};
+
+const struct amdgpu_ip_block_version psp_v3_1_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_PSP,
 	.major = 3,
 	.minor = 1,
@@ -3682,8 +4476,7 @@ const struct amdgpu_ip_block_version psp_v3_1_ip_block =
 	.funcs = &psp_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version psp_v10_0_ip_block =
-{
+const struct amdgpu_ip_block_version psp_v10_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_PSP,
 	.major = 10,
 	.minor = 0,
@@ -3691,8 +4484,7 @@ const struct amdgpu_ip_block_version psp_v10_0_ip_block =
 	.funcs = &psp_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version psp_v11_0_ip_block =
-{
+const struct amdgpu_ip_block_version psp_v11_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_PSP,
 	.major = 11,
 	.minor = 0,
@@ -3708,8 +4500,7 @@ const struct amdgpu_ip_block_version psp_v11_0_8_ip_block = {
 	.funcs = &psp_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version psp_v12_0_ip_block =
-{
+const struct amdgpu_ip_block_version psp_v12_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_PSP,
 	.major = 12,
 	.minor = 0,
@@ -3732,3 +4523,11 @@ const struct amdgpu_ip_block_version psp_v13_0_4_ip_block = {
 	.rev = 4,
 	.funcs = &psp_ip_funcs,
 };
+
+const struct amdgpu_ip_block_version psp_v14_0_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_PSP,
+	.major = 14,
+	.minor = 0,
+	.rev = 0,
+	.funcs = &psp_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
index cf4f60c66122..237b624aa51c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
@@ -39,6 +39,31 @@
 #define PSP_TMR_ALIGNMENT	0x100000
 #define PSP_FW_NAME_LEN		0x24
 
+/* VBIOS gfl defines */
+#define MBOX_READY_MASK 0x80000000
+#define MBOX_STATUS_MASK 0x0000FFFF
+#define MBOX_COMMAND_MASK 0x00FF0000
+#define MBOX_READY_FLAG 0x80000000
+#define C2PMSG_CMD_SPI_UPDATE_ROM_IMAGE_ADDR_LO 0x2
+#define C2PMSG_CMD_SPI_UPDATE_ROM_IMAGE_ADDR_HI 0x3
+#define C2PMSG_CMD_SPI_UPDATE_FLASH_IMAGE 0x4
+#define C2PMSG_CMD_SPI_GET_ROM_IMAGE_ADDR_LO 0xf
+#define C2PMSG_CMD_SPI_GET_ROM_IMAGE_ADDR_HI 0x10
+#define C2PMSG_CMD_SPI_GET_FLASH_IMAGE 0x11
+
+/* Command register bit 31 set to indicate readiness */
+#define MBOX_TOS_READY_FLAG (GFX_FLAG_RESPONSE)
+#define MBOX_TOS_READY_MASK (GFX_CMD_RESPONSE_MASK | GFX_CMD_STATUS_MASK)
+
+/* Values to check for a successful GFX_CMD response wait. Check against
+ * both status bits and response state - helps to detect a command failure
+ * or other unexpected cases like a device drop reading all 0xFFs
+ */
+#define MBOX_TOS_RESP_FLAG (GFX_FLAG_RESPONSE)
+#define MBOX_TOS_RESP_MASK (GFX_CMD_RESPONSE_MASK | GFX_CMD_STATUS_MASK)
+
+extern const struct attribute_group amdgpu_flash_attr_group;
+
 enum psp_shared_mem_size {
 	PSP_ASD_SHARED_MEM_SIZE				= 0x0,
 	PSP_XGMI_SHARED_MEM_SIZE			= 0x4000,
@@ -71,15 +96,17 @@ enum psp_bootloader_cmd {
 	PSP_BL__LOAD_KEY_DATABASE	= 0x80000,
 	PSP_BL__LOAD_SOCDRV             = 0xB0000,
 	PSP_BL__LOAD_DBGDRV             = 0xC0000,
+	PSP_BL__LOAD_HADDRV		= PSP_BL__LOAD_DBGDRV,
 	PSP_BL__LOAD_INTFDRV		= 0xD0000,
-	PSP_BL__LOAD_RASDRV		    = 0xE0000,
+	PSP_BL__LOAD_RASDRV		= 0xE0000,
+	PSP_BL__LOAD_IPKEYMGRDRV	= 0xF0000,
 	PSP_BL__DRAM_LONG_TRAIN		= 0x100000,
 	PSP_BL__DRAM_SHORT_TRAIN	= 0x200000,
 	PSP_BL__LOAD_TOS_SPL_TABLE	= 0x10000000,
+	PSP_BL__LOAD_SPDMDRV		= 0x20000000,
 };
 
-enum psp_ring_type
-{
+enum psp_ring_type {
 	PSP_RING_TYPE__INVALID = 0,
 	/*
 	 * These values map to the way the PSP kernel identifies the
@@ -89,8 +116,7 @@ enum psp_ring_type
 	PSP_RING_TYPE__KM = 2  /* Kernel mode ring (formerly called GPCOM) */
 };
 
-struct psp_ring
-{
+struct psp_ring {
 	enum psp_ring_type		ring_type;
 	struct psp_gfx_rb_frame		*ring_mem;
 	uint64_t			ring_mem_mc_addr;
@@ -104,12 +130,16 @@ enum psp_reg_prog_id {
 	PSP_REG_IH_RB_CNTL        = 0,  /* register IH_RB_CNTL */
 	PSP_REG_IH_RB_CNTL_RING1  = 1,  /* register IH_RB_CNTL_RING1 */
 	PSP_REG_IH_RB_CNTL_RING2  = 2,  /* register IH_RB_CNTL_RING2 */
+	PSP_REG_MMHUB_L1_TLB_CNTL = 25,
 	PSP_REG_LAST
 };
 
-struct psp_funcs
-{
+#define PSP_WAITREG_CHANGED BIT(0) /* check if the value has changed */
+#define PSP_WAITREG_NOVERBOSE BIT(1) /* No error verbose */
+
+struct psp_funcs {
 	int (*init_microcode)(struct psp_context *psp);
+	int (*wait_for_bootloader)(struct psp_context *psp);
 	int (*bootloader_load_kdb)(struct psp_context *psp);
 	int (*bootloader_load_spl)(struct psp_context *psp);
 	int (*bootloader_load_sysdrv)(struct psp_context *psp);
@@ -117,6 +147,8 @@ struct psp_funcs
 	int (*bootloader_load_intf_drv)(struct psp_context *psp);
 	int (*bootloader_load_dbg_drv)(struct psp_context *psp);
 	int (*bootloader_load_ras_drv)(struct psp_context *psp);
+	int (*bootloader_load_ipkeymgr_drv)(struct psp_context *psp);
+	int (*bootloader_load_spdm_drv)(struct psp_context *psp);
 	int (*bootloader_load_sos)(struct psp_context *psp);
 	int (*ring_create)(struct psp_context *psp,
 			   enum psp_ring_type ring_type);
@@ -132,7 +164,14 @@ struct psp_funcs
 	int (*load_usbc_pd_fw)(struct psp_context *psp, uint64_t fw_pri_mc_addr);
 	int (*read_usbc_pd_fw)(struct psp_context *psp, uint32_t *fw_ver);
 	int (*update_spirom)(struct psp_context *psp, uint64_t fw_pri_mc_addr);
+	int (*dump_spirom)(struct psp_context *psp, uint64_t fw_pri_mc_addr);
 	int (*vbflash_stat)(struct psp_context *psp);
+	int (*fatal_error_recovery_quirk)(struct psp_context *psp);
+	bool (*get_ras_capability)(struct psp_context *psp);
+	bool (*is_aux_sos_load_required)(struct psp_context *psp);
+	bool (*is_reload_needed)(struct psp_context *psp);
+	int (*reg_program_no_ring)(struct psp_context *psp, uint32_t val,
+				   enum psp_reg_prog_id id);
 };
 
 struct ta_funcs {
@@ -148,6 +187,7 @@ struct psp_xgmi_node_info {
 	uint8_t					is_sharing_enabled;
 	enum ta_xgmi_assigned_sdma_engine	sdma_engine;
 	uint8_t					num_links;
+	struct xgmi_connected_port_num		port_num[TA_XGMI__MAX_PORT_NUM];
 };
 
 struct psp_xgmi_topology_info {
@@ -188,18 +228,20 @@ struct psp_xgmi_context {
 	struct ta_context		context;
 	struct psp_xgmi_topology_info	top_info;
 	bool				supports_extended_data;
+	uint8_t				xgmi_ta_caps;
 };
 
 struct psp_ras_context {
 	struct ta_context		context;
 	struct amdgpu_ras		*ras;
+	struct mutex			mutex;
 };
 
 #define MEM_TRAIN_SYSTEM_SIGNATURE		0x54534942
 #define GDDR6_MEM_TRAINING_DATA_SIZE_IN_BYTES	0x1000
 #define GDDR6_MEM_TRAINING_OFFSET		0x8000
 /*Define the VRAM size that will be encroached by BIST training.*/
-#define GDDR6_MEM_TRAINING_ENCROACHED_SIZE	0x2000000
+#define BIST_MEM_TRAINING_ENCROACHED_SIZE	0x2000000
 
 enum psp_memory_training_init_flag {
 	PSP_MEM_TRAIN_NOT_SUPPORT	= 0x0,
@@ -307,10 +349,17 @@ struct psp_runtime_scpm_entry {
 	enum psp_runtime_scpm_authentication scpm_status;
 };
 
-struct psp_context
-{
-	struct amdgpu_device            *adev;
-	struct psp_ring                 km_ring;
+#if defined(CONFIG_DEBUG_FS)
+struct spirom_bo {
+	struct amdgpu_bo *bo;
+	uint64_t mc_addr;
+	void *cpu_addr;
+};
+#endif
+
+struct psp_context {
+	struct amdgpu_device		*adev;
+	struct psp_ring			km_ring;
 	struct psp_gfx_cmd_resp		*cmd;
 
 	const struct psp_funcs		*funcs;
@@ -333,13 +382,15 @@ struct psp_context
 	struct psp_bin_desc		intf_drv;
 	struct psp_bin_desc		dbg_drv;
 	struct psp_bin_desc		ras_drv;
+	struct psp_bin_desc		ipkeymgr_drv;
+	struct psp_bin_desc		spdm_drv;
 
 	/* tmr buffer */
 	struct amdgpu_bo		*tmr_bo;
 	uint64_t			tmr_mc_addr;
 
 	/* asd firmware */
-	const struct firmware	*asd_fw;
+	const struct firmware		*asd_fw;
 
 	/* toc firmware */
 	const struct firmware		*toc_fw;
@@ -361,6 +412,8 @@ struct psp_context
 	atomic_t			fence_value;
 	/* flag to mark whether gfx fw autoload is supported or not */
 	bool				autoload_supported;
+	/* flag to mark whether psp use runtime TMR or boottime TMR */
+	bool				boot_time_tmr;
 	/* flag to mark whether df cstate management centralized to PMFW */
 	bool				pmfw_centralized_cstate_management;
 
@@ -384,9 +437,16 @@ struct psp_context
 
 	uint32_t			boot_cfg_bitmask;
 
-	char *vbflash_tmp_buf;
-	size_t vbflash_image_size;
-	bool vbflash_done;
+	/* firmware upgrades supported */
+	bool				sup_pd_fw_up;
+	bool				sup_ifwi_up;
+
+	char				*vbflash_tmp_buf;
+	size_t				vbflash_image_size;
+	bool				vbflash_done;
+#if defined(CONFIG_DEBUG_FS)
+	struct spirom_bo *spirom_dump_trip;
+#endif
 };
 
 struct amdgpu_psp_funcs {
@@ -415,6 +475,12 @@ struct amdgpu_psp_funcs {
 #define psp_bootloader_load_ras_drv(psp) \
 		((psp)->funcs->bootloader_load_ras_drv ? \
 		(psp)->funcs->bootloader_load_ras_drv((psp)) : 0)
+#define psp_bootloader_load_ipkeymgr_drv(psp) \
+		((psp)->funcs->bootloader_load_ipkeymgr_drv ? \
+		 (psp)->funcs->bootloader_load_ipkeymgr_drv((psp)) : 0)
+#define psp_bootloader_load_spdm_drv(psp) \
+		((psp)->funcs->bootloader_load_spdm_drv ? \
+		 (psp)->funcs->bootloader_load_spdm_drv((psp)) : 0)
 #define psp_bootloader_load_sos(psp) \
 		((psp)->funcs->bootloader_load_sos ? (psp)->funcs->bootloader_load_sos((psp)) : 0)
 #define psp_smu_reload_quirk(psp) \
@@ -439,10 +505,25 @@ struct amdgpu_psp_funcs {
 	((psp)->funcs->update_spirom ? \
 	(psp)->funcs->update_spirom((psp), fw_pri_mc_addr) : -EINVAL)
 
+#define psp_dump_spirom(psp, fw_pri_mc_addr) \
+	((psp)->funcs->dump_spirom ? \
+	(psp)->funcs->dump_spirom((psp), fw_pri_mc_addr) : -EINVAL)
+
 #define psp_vbflash_status(psp) \
 	((psp)->funcs->vbflash_stat ? \
 	(psp)->funcs->vbflash_stat((psp)) : -EINVAL)
 
+#define psp_fatal_error_recovery_quirk(psp) \
+	((psp)->funcs->fatal_error_recovery_quirk ? \
+	(psp)->funcs->fatal_error_recovery_quirk((psp)) : 0)
+
+#define psp_is_aux_sos_load_required(psp) \
+	((psp)->funcs->is_aux_sos_load_required ? (psp)->funcs->is_aux_sos_load_required((psp)) : 0)
+
+#define psp_reg_program_no_ring(psp, val, id) \
+	((psp)->funcs->reg_program_no_ring ? \
+	(psp)->funcs->reg_program_no_ring((psp), val, id) : -EINVAL)
+
 extern const struct amd_ip_funcs psp_ip_funcs;
 
 extern const struct amdgpu_ip_block_version psp_v3_1_ip_block;
@@ -452,13 +533,17 @@ extern const struct amdgpu_ip_block_version psp_v11_0_8_ip_block;
 extern const struct amdgpu_ip_block_version psp_v12_0_ip_block;
 extern const struct amdgpu_ip_block_version psp_v13_0_ip_block;
 extern const struct amdgpu_ip_block_version psp_v13_0_4_ip_block;
+extern const struct amdgpu_ip_block_version psp_v14_0_ip_block;
+
+int psp_wait_for(struct psp_context *psp, uint32_t reg_index,
+		 uint32_t field_val, uint32_t mask, uint32_t flags);
+extern int psp_wait_for_spirom_update(struct psp_context *psp, uint32_t reg_index,
+			uint32_t field_val, uint32_t mask, uint32_t msec_timeout);
 
-extern int psp_wait_for(struct psp_context *psp, uint32_t reg_index,
-			uint32_t field_val, uint32_t mask, bool check_changed);
+int psp_execute_ip_fw_load(struct psp_context *psp,
+			   struct amdgpu_firmware_info *ucode);
 
 int psp_gpu_reset(struct amdgpu_device *adev);
-int psp_update_vcn_sram(struct amdgpu_device *adev, int inst_idx,
-			uint64_t cmd_gpu_addr, int cmd_size);
 
 int psp_ta_init_shared_buf(struct psp_context *psp,
 				  struct ta_mem_context *mem_ctx);
@@ -486,8 +571,11 @@ int psp_ras_invoke(struct psp_context *psp, uint32_t ta_cmd_id);
 int psp_ras_enable_features(struct psp_context *psp,
 		union ta_ras_cmd_input *info, bool enable);
 int psp_ras_trigger_error(struct psp_context *psp,
-			  struct ta_ras_trigger_error_input *info);
+			  struct ta_ras_trigger_error_input *info, uint32_t instance_mask);
 int psp_ras_terminate(struct psp_context *psp);
+int psp_ras_query_address(struct psp_context *psp,
+			  struct ta_ras_query_address_input *addr_in,
+			  struct ta_ras_query_address_output *addr_out);
 
 int psp_hdcp_invoke(struct psp_context *psp, uint32_t ta_cmd_id);
 int psp_dtm_invoke(struct psp_context *psp, uint32_t ta_cmd_id);
@@ -514,13 +602,25 @@ int psp_init_cap_microcode(struct psp_context *psp,
 			  const char *chip_name);
 int psp_get_fw_attestation_records_addr(struct psp_context *psp,
 					uint64_t *output_ptr);
-
+int psp_update_fw_reservation(struct psp_context *psp);
 int psp_load_fw_list(struct psp_context *psp,
 		     struct amdgpu_firmware_info **ucode_list, int ucode_count);
 void psp_copy_fw(struct psp_context *psp, uint8_t *start_addr, uint32_t bin_size);
 
+int psp_spatial_partition(struct psp_context *psp, int mode);
+int psp_memory_partition(struct psp_context *psp, int mode);
+
 int is_psp_fw_valid(struct psp_bin_desc bin);
 
-int amdgpu_psp_sysfs_init(struct amdgpu_device *adev);
-void amdgpu_psp_sysfs_fini(struct amdgpu_device *adev);
+int amdgpu_psp_wait_for_bootloader(struct amdgpu_device *adev);
+bool amdgpu_psp_get_ras_capability(struct psp_context *psp);
+
+int psp_config_sq_perfmon(struct psp_context *psp, uint32_t xcp_id,
+	bool core_override_enable, bool reg_override_enable, bool perfmon_override_enable);
+bool amdgpu_psp_tos_reload_needed(struct amdgpu_device *adev);
+int amdgpu_psp_reg_program_no_ring(struct psp_context *psp, uint32_t val,
+				   enum psp_reg_prog_id id);
+void amdgpu_psp_debugfs_init(struct amdgpu_device *adev);
+
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c
index 468a67b302d4..6e8aad91bcd3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c
@@ -166,15 +166,14 @@ static ssize_t ta_if_load_debugfs_write(struct file *fp, const char *buf, size_t
 	if (ret)
 		return -EFAULT;
 
+	if (ta_bin_len > PSP_1_MEG)
+		return -EINVAL;
+
 	copy_pos += sizeof(uint32_t);
 
-	ta_bin = kzalloc(ta_bin_len, GFP_KERNEL);
-	if (!ta_bin)
-		return -ENOMEM;
-	if (copy_from_user((void *)ta_bin, &buf[copy_pos], ta_bin_len)) {
-		ret = -EFAULT;
-		goto err_free_bin;
-	}
+	ta_bin = memdup_user(&buf[copy_pos], ta_bin_len);
+	if (IS_ERR(ta_bin))
+		return PTR_ERR(ta_bin);
 
 	/* Set TA context and functions */
 	set_ta_context_funcs(psp, ta_type, &context);
@@ -324,17 +323,13 @@ static ssize_t ta_if_invoke_debugfs_write(struct file *fp, const char *buf, size
 		return -EFAULT;
 	copy_pos += sizeof(uint32_t);
 
-	shared_buf = kzalloc(shared_buf_len, GFP_KERNEL);
-	if (!shared_buf)
-		return -ENOMEM;
-	if (copy_from_user((void *)shared_buf, &buf[copy_pos], shared_buf_len)) {
-		ret = -EFAULT;
-		goto err_free_shared_buf;
-	}
+	shared_buf = memdup_user(&buf[copy_pos], shared_buf_len);
+	if (IS_ERR(shared_buf))
+		return PTR_ERR(shared_buf);
 
 	set_ta_context_funcs(psp, ta_type, &context);
 
-	if (!context->initialized) {
+	if (!context || !context->initialized) {
 		dev_err(adev->dev, "TA is not initialized\n");
 		ret = -EINVAL;
 		goto err_free_shared_buf;
@@ -348,6 +343,7 @@ static ssize_t ta_if_invoke_debugfs_write(struct file *fp, const char *buf, size
 
 	context->session_id = ta_id;
 
+	mutex_lock(&psp->ras_context.mutex);
 	ret = prep_ta_mem_context(&context->mem_context, shared_buf, shared_buf_len);
 	if (ret)
 		goto err_free_shared_buf;
@@ -362,10 +358,11 @@ static ssize_t ta_if_invoke_debugfs_write(struct file *fp, const char *buf, size
 		}
 	}
 
-	if (copy_to_user((char *)buf, context->mem_context.shared_buf, shared_buf_len))
+	if (copy_to_user((char *)&buf[copy_pos], context->mem_context.shared_buf, shared_buf_len))
 		ret = -EFAULT;
 
 err_free_shared_buf:
+	mutex_unlock(&psp->ras_context.mutex);
 	kfree(shared_buf);
 
 	return ret;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_rap.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_rap.c
index 12010c988c8b..bacf888735db 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_rap.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_rap.c
@@ -101,7 +101,6 @@ static ssize_t amdgpu_rap_debugfs_write(struct file *f, const char __user *buf,
 	}
 
 	amdgpu_gfx_off_ctrl(adev, true);
-	pm_runtime_mark_last_busy(dev->dev);
 	pm_runtime_put_autosuspend(dev->dev);
 
 	return size;
@@ -116,7 +115,6 @@ static const struct file_operations amdgpu_rap_debugfs_ops = {
 
 void amdgpu_rap_debugfs_init(struct amdgpu_device *adev)
 {
-#if defined(CONFIG_DEBUG_FS)
 	struct drm_minor *minor = adev_to_drm(adev)->primary;
 
 	if (!adev->psp.rap_context.context.initialized)
@@ -124,5 +122,4 @@ void amdgpu_rap_debugfs_init(struct amdgpu_device *adev)
 
 	debugfs_create_file("rap_test", S_IWUSR, minor->debugfs_root,
 				adev, &amdgpu_rap_debugfs_ops);
-#endif
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
index 3ab8a88789c8..2a6cf7963dde 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
@@ -28,6 +28,7 @@
 #include <linux/reboot.h>
 #include <linux/syscalls.h>
 #include <linux/pm_runtime.h>
+#include <linux/list_sort.h>
 
 #include "amdgpu.h"
 #include "amdgpu_ras.h"
@@ -35,8 +36,12 @@
 #include "amdgpu_xgmi.h"
 #include "ivsrcid/nbio/irqsrcs_nbif_7_4.h"
 #include "nbio_v4_3.h"
+#include "nbif_v6_3_1.h"
+#include "nbio_v7_9.h"
 #include "atom.h"
 #include "amdgpu_reset.h"
+#include "amdgpu_psp.h"
+#include "amdgpu_ras_mgr.h"
 
 #ifdef CONFIG_X86_MCE_AMD
 #include <asm/mce.h>
@@ -71,6 +76,9 @@ const char *ras_block_string[] = {
 	"mca",
 	"vcn",
 	"jpeg",
+	"ih",
+	"mpio",
+	"mmsch",
 };
 
 const char *ras_mca_block_string[] = {
@@ -92,7 +100,8 @@ const char *get_ras_block_str(struct ras_common_if *ras_block)
 	if (!ras_block)
 		return "NULL";
 
-	if (ras_block->block >= AMDGPU_RAS_BLOCK_COUNT)
+	if (ras_block->block >= AMDGPU_RAS_BLOCK_COUNT ||
+	    ras_block->block >= ARRAY_SIZE(ras_block_string))
 		return "OUT OF RANGE";
 
 	if (ras_block->block == AMDGPU_RAS_BLOCK__MCA)
@@ -114,6 +123,15 @@ const char *get_ras_block_str(struct ras_common_if *ras_block)
 /* typical ECC bad page rate is 1 bad page per 100MB VRAM */
 #define RAS_BAD_PAGE_COVER              (100 * 1024 * 1024ULL)
 
+#define MAX_UMC_POISON_POLLING_TIME_ASYNC  10
+
+#define AMDGPU_RAS_RETIRE_PAGE_INTERVAL 100  //ms
+
+#define MAX_FLUSH_RETIRE_DWORK_TIMES  100
+
+#define BYPASS_ALLOCATED_ADDRESS        0x0
+#define BYPASS_INITIALIZATION_ADDRESS   0x1
+
 enum amdgpu_ras_retire_page_reservation {
 	AMDGPU_RAS_RETIRE_PAGE_RESERVED,
 	AMDGPU_RAS_RETIRE_PAGE_PENDING,
@@ -122,12 +140,18 @@ enum amdgpu_ras_retire_page_reservation {
 
 atomic_t amdgpu_ras_in_intr = ATOMIC_INIT(0);
 
-static bool amdgpu_ras_check_bad_page_unlock(struct amdgpu_ras *con,
+static int amdgpu_ras_check_bad_page_unlock(struct amdgpu_ras *con,
 				uint64_t addr);
-static bool amdgpu_ras_check_bad_page(struct amdgpu_device *adev,
+static int amdgpu_ras_check_bad_page(struct amdgpu_device *adev,
 				uint64_t addr);
+
+static void amdgpu_ras_critical_region_init(struct amdgpu_device *adev);
+static void amdgpu_ras_critical_region_fini(struct amdgpu_device *adev);
+
 #ifdef CONFIG_X86_MCE_AMD
 static void amdgpu_register_bad_pages_mca_notifier(struct amdgpu_device *adev);
+static void
+amdgpu_unregister_bad_pages_mca_notifier(struct amdgpu_device *adev);
 struct mce_notifier_adev_list {
 	struct amdgpu_device *devs[MAX_GPU_INSTANCE];
 	int num_gpu;
@@ -151,35 +175,39 @@ static bool amdgpu_ras_get_error_query_ready(struct amdgpu_device *adev)
 
 static int amdgpu_reserve_page_direct(struct amdgpu_device *adev, uint64_t address)
 {
-	struct ras_err_data err_data = {0, 0, 0, NULL};
+	struct ras_err_data err_data;
 	struct eeprom_table_record err_rec;
+	int ret;
 
-	if ((address >= adev->gmc.mc_vram_size) ||
-	    (address >= RAS_UMC_INJECT_ADDR_LIMIT)) {
+	ret = amdgpu_ras_check_bad_page(adev, address);
+	if (ret == -EINVAL) {
 		dev_warn(adev->dev,
-		         "RAS WARN: input address 0x%llx is invalid.\n",
-		         address);
+			"RAS WARN: input address 0x%llx is invalid.\n",
+			address);
 		return -EINVAL;
-	}
-
-	if (amdgpu_ras_check_bad_page(adev, address)) {
+	} else if (ret == 1) {
 		dev_warn(adev->dev,
-			 "RAS WARN: 0x%llx has already been marked as bad page!\n",
-			 address);
+			"RAS WARN: 0x%llx has already been marked as bad page!\n",
+			address);
 		return 0;
 	}
 
+	ret = amdgpu_ras_error_data_init(&err_data);
+	if (ret)
+		return ret;
+
 	memset(&err_rec, 0x0, sizeof(struct eeprom_table_record));
 	err_data.err_addr = &err_rec;
-	amdgpu_umc_fill_error_record(&err_data, address,
-			(address >> AMDGPU_GPU_PAGE_SHIFT), 0, 0);
+	amdgpu_umc_fill_error_record(&err_data, address, address, 0, 0);
 
 	if (amdgpu_bad_page_threshold != 0) {
 		amdgpu_ras_add_bad_pages(adev, err_data.err_addr,
-					 err_data.err_addr_cnt);
+					 err_data.err_addr_cnt, false);
 		amdgpu_ras_save_bad_pages(adev, NULL);
 	}
 
+	amdgpu_ras_error_data_fini(&err_data);
+
 	dev_warn(adev->dev, "WARNING: THIS IS ONLY FOR TEST PURPOSES AND WILL CORRUPT RAS EEPROM\n");
 	dev_warn(adev->dev, "Clear EEPROM:\n");
 	dev_warn(adev->dev, "    echo 1 > /sys/kernel/debug/dri/0/ras/ras_eeprom_reset\n");
@@ -187,6 +215,56 @@ static int amdgpu_reserve_page_direct(struct amdgpu_device *adev, uint64_t addre
 	return 0;
 }
 
+static int amdgpu_check_address_validity(struct amdgpu_device *adev,
+			uint64_t address, uint64_t flags)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct amdgpu_vram_block_info blk_info;
+	uint64_t page_pfns[32] = {0};
+	int i, ret, count;
+	bool hit = false;
+
+	if (amdgpu_ip_version(adev, UMC_HWIP, 0) < IP_VERSION(12, 0, 0))
+		return 0;
+
+	if (amdgpu_sriov_vf(adev)) {
+		if (amdgpu_virt_check_vf_critical_region(adev, address, &hit))
+			return -EPERM;
+		return hit ? -EACCES : 0;
+	}
+
+	if ((address >= adev->gmc.mc_vram_size) ||
+	    (address >= RAS_UMC_INJECT_ADDR_LIMIT))
+		return -EFAULT;
+
+	count = amdgpu_umc_lookup_bad_pages_in_a_row(adev,
+				address, page_pfns, ARRAY_SIZE(page_pfns));
+	if (count <= 0)
+		return -EPERM;
+
+	for (i = 0; i < count; i++) {
+		memset(&blk_info, 0, sizeof(blk_info));
+		ret = amdgpu_vram_mgr_query_address_block_info(&adev->mman.vram_mgr,
+					page_pfns[i] << AMDGPU_GPU_PAGE_SHIFT, &blk_info);
+		if (!ret) {
+			/* The input address that needs to be checked is allocated by
+			 * current calling process, so it is necessary to exclude
+			 * the calling process.
+			 */
+			if ((flags == BYPASS_ALLOCATED_ADDRESS) &&
+			    ((blk_info.task.pid != task_pid_nr(current)) ||
+				strncmp(blk_info.task.comm, current->comm, TASK_COMM_LEN)))
+				return -EACCES;
+			else if ((flags == BYPASS_INITIALIZATION_ADDRESS) &&
+				(blk_info.task.pid == con->init_task_pid) &&
+				!strncmp(blk_info.task.comm, con->init_task_comm, TASK_COMM_LEN))
+				return -EACCES;
+		}
+	}
+
+	return 0;
+}
+
 static ssize_t amdgpu_ras_debugfs_read(struct file *f, char __user *buf,
 					size_t size, loff_t *pos)
 {
@@ -201,8 +279,8 @@ static ssize_t amdgpu_ras_debugfs_read(struct file *f, char __user *buf,
 		return -EINVAL;
 
 	/* Hardware counter will be reset automatically after the query on Vega20 and Arcturus */
-	if (obj->adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 2) &&
-	    obj->adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 4)) {
+	if (amdgpu_ip_version(obj->adev, MP0_HWIP, 0) != IP_VERSION(11, 0, 2) &&
+	    amdgpu_ip_version(obj->adev, MP0_HWIP, 0) != IP_VERSION(11, 0, 4)) {
 		if (amdgpu_ras_reset_error_status(obj->adev, info.head.block))
 			dev_warn(obj->adev->dev, "Failed to reset error counter and error status");
 	}
@@ -256,6 +334,8 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f,
 	int block_id;
 	uint32_t sub_block;
 	u64 address, value;
+	/* default value is 0 if the mask is not set by user */
+	u32 instance_mask = 0;
 
 	if (*pos)
 		return -EINVAL;
@@ -275,6 +355,8 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f,
 		op = 2;
 	else if (strstr(str, "retire_page") != NULL)
 		op = 3;
+	else if (strstr(str, "check_address") != NULL)
+		op = 4;
 	else if (str[0] && str[1] && str[2] && str[3])
 		/* ascii string, but commands are not matched. */
 		return -EINVAL;
@@ -289,24 +371,39 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f,
 			data->inject.address = address;
 
 			return 0;
+		} else if (op == 4) {
+			if (sscanf(str, "%*s 0x%llx 0x%llx", &address, &value) != 2 &&
+			    sscanf(str, "%*s %llu %llu", &address, &value) != 2)
+				return -EINVAL;
+
+			data->op = op;
+			data->inject.address = address;
+			data->inject.value = value;
+			return 0;
 		}
 
 		if (amdgpu_ras_find_block_id_by_name(block_name, &block_id))
 			return -EINVAL;
 
 		data->head.block = block_id;
-		/* only ue and ce errors are supported */
+		/* only ue, ce and poison errors are supported */
 		if (!memcmp("ue", err, 2))
 			data->head.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE;
 		else if (!memcmp("ce", err, 2))
 			data->head.type = AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE;
+		else if (!memcmp("poison", err, 6))
+			data->head.type = AMDGPU_RAS_ERROR__POISON;
 		else
 			return -EINVAL;
 
 		data->op = op;
 
 		if (op == 2) {
-			if (sscanf(str, "%*s %*s %*s 0x%x 0x%llx 0x%llx",
+			if (sscanf(str, "%*s %*s %*s 0x%x 0x%llx 0x%llx 0x%x",
+				   &sub_block, &address, &value, &instance_mask) != 4 &&
+			    sscanf(str, "%*s %*s %*s %u %llu %llu %u",
+				   &sub_block, &address, &value, &instance_mask) != 4 &&
+				sscanf(str, "%*s %*s %*s 0x%x 0x%llx 0x%llx",
 				   &sub_block, &address, &value) != 3 &&
 			    sscanf(str, "%*s %*s %*s %u %llu %llu",
 				   &sub_block, &address, &value) != 3)
@@ -314,6 +411,7 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f,
 			data->head.sub_block_index = sub_block;
 			data->inject.address = address;
 			data->inject.value = value;
+			data->inject.instance_mask = instance_mask;
 		}
 	} else {
 		if (size < sizeof(*data))
@@ -326,6 +424,46 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f,
 	return 0;
 }
 
+static void amdgpu_ras_instance_mask_check(struct amdgpu_device *adev,
+				struct ras_debug_if *data)
+{
+	int num_xcc = adev->gfx.xcc_mask ? NUM_XCC(adev->gfx.xcc_mask) : 1;
+	uint32_t mask, inst_mask = data->inject.instance_mask;
+
+	/* no need to set instance mask if there is only one instance */
+	if (num_xcc <= 1 && inst_mask) {
+		data->inject.instance_mask = 0;
+		dev_dbg(adev->dev,
+			"RAS inject mask(0x%x) isn't supported and force it to 0.\n",
+			inst_mask);
+
+		return;
+	}
+
+	switch (data->head.block) {
+	case AMDGPU_RAS_BLOCK__GFX:
+		mask = GENMASK(num_xcc - 1, 0);
+		break;
+	case AMDGPU_RAS_BLOCK__SDMA:
+		mask = GENMASK(adev->sdma.num_instances - 1, 0);
+		break;
+	case AMDGPU_RAS_BLOCK__VCN:
+	case AMDGPU_RAS_BLOCK__JPEG:
+		mask = GENMASK(adev->vcn.num_vcn_inst - 1, 0);
+		break;
+	default:
+		mask = inst_mask;
+		break;
+	}
+
+	/* remove invalid bits in instance mask */
+	data->inject.instance_mask &= mask;
+	if (inst_mask != data->inject.instance_mask)
+		dev_dbg(adev->dev,
+			"Adjust RAS inject mask 0x%x to 0x%x\n",
+			inst_mask, data->inject.instance_mask);
+}
+
 /**
  * DOC: AMDGPU RAS debugfs control interface
  *
@@ -341,7 +479,7 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f,
  * sub_block_index: some IPs have subcomponets. say, GFX, sDMA.
  * name: the name of IP.
  *
- * inject has two more members than head, they are address, value.
+ * inject has three more members than head, they are address, value and mask.
  * As their names indicate, inject operation will write the
  * value to the address.
  *
@@ -365,7 +503,7 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f,
  *
  *	echo "disable <block>" > /sys/kernel/debug/dri/<N>/ras/ras_ctrl
  *	echo "enable  <block> <error>" > /sys/kernel/debug/dri/<N>/ras/ras_ctrl
- *	echo "inject  <block> <error> <sub-block> <address> <value> > /sys/kernel/debug/dri/<N>/ras/ras_ctrl
+ *	echo "inject  <block> <error> <sub-block> <address> <value> <mask>" > /sys/kernel/debug/dri/<N>/ras/ras_ctrl
  *
  * Where N, is the card which you want to affect.
  *
@@ -376,19 +514,21 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f,
  * The block is one of: umc, sdma, gfx, etc.
  *	see ras_block_string[] for details
  *
- * The error type is one of: ue, ce, where,
+ * The error type is one of: ue, ce and poison where,
  *	ue is multi-uncorrectable
  *	ce is single-correctable
+ *	poison is poison
  *
  * The sub-block is a the sub-block index, pass 0 if there is no sub-block.
  * The address and value are hexadecimal numbers, leading 0x is optional.
+ * The mask means instance mask, is optional, default value is 0x1.
  *
  * For instance,
  *
  * .. code-block:: bash
  *
  *	echo inject umc ue 0x0 0x0 0x0 > /sys/kernel/debug/dri/0/ras/ras_ctrl
- *	echo inject umc ce 0 0 0 > /sys/kernel/debug/dri/0/ras/ras_ctrl
+ *	echo inject umc ce 0 0 0 3 > /sys/kernel/debug/dri/0/ras/ras_ctrl
  *	echo disable umc > /sys/kernel/debug/dri/0/ras/ras_ctrl
  *
  * How to check the result of the operation?
@@ -429,6 +569,9 @@ static ssize_t amdgpu_ras_debugfs_ctrl_write(struct file *f,
 			return size;
 		else
 			return ret;
+	} else if (data.op == 4) {
+		ret = amdgpu_check_address_validity(adev, data.inject.address, data.inject.value);
+		return ret ? ret : size;
 	}
 
 	if (!amdgpu_ras_is_supported(adev, data.head.block))
@@ -442,24 +585,21 @@ static ssize_t amdgpu_ras_debugfs_ctrl_write(struct file *f,
 		ret = amdgpu_ras_feature_enable(adev, &data.head, 1);
 		break;
 	case 2:
-		if ((data.inject.address >= adev->gmc.mc_vram_size) ||
-		    (data.inject.address >= RAS_UMC_INJECT_ADDR_LIMIT)) {
-			dev_warn(adev->dev, "RAS WARN: input address "
-					"0x%llx is invalid.",
+		/* umc ce/ue error injection for a bad page is not allowed */
+		if (data.head.block == AMDGPU_RAS_BLOCK__UMC)
+			ret = amdgpu_ras_check_bad_page(adev, data.inject.address);
+		if (ret == -EINVAL) {
+			dev_warn(adev->dev, "RAS WARN: input address 0x%llx is invalid.",
 					data.inject.address);
-			ret = -EINVAL;
 			break;
-		}
-
-		/* umc ce/ue error injection for a bad page is not allowed */
-		if ((data.head.block == AMDGPU_RAS_BLOCK__UMC) &&
-		    amdgpu_ras_check_bad_page(adev, data.inject.address)) {
-			dev_warn(adev->dev, "RAS WARN: inject: 0x%llx has "
-				 "already been marked as bad!\n",
-				 data.inject.address);
+		} else if (ret == 1) {
+			dev_warn(adev->dev, "RAS WARN: inject: 0x%llx has already been marked as bad!\n",
+					data.inject.address);
 			break;
 		}
 
+		amdgpu_ras_instance_mask_check(adev, &data);
+
 		/* data.inject.address is offset instead of absolute gpu address */
 		ret = amdgpu_ras_error_inject(adev, &data.inject);
 		break;
@@ -474,6 +614,8 @@ static ssize_t amdgpu_ras_debugfs_ctrl_write(struct file *f,
 	return size;
 }
 
+static int amdgpu_uniras_clear_badpages_info(struct amdgpu_device *adev);
+
 /**
  * DOC: AMDGPU RAS debugfs EEPROM table reset interface
  *
@@ -498,6 +640,11 @@ static ssize_t amdgpu_ras_debugfs_eeprom_write(struct file *f,
 		(struct amdgpu_device *)file_inode(f)->i_private;
 	int ret;
 
+	if (amdgpu_uniras_enabled(adev)) {
+		ret = amdgpu_uniras_clear_badpages_info(adev);
+		return ret ? ret : size;
+	}
+
 	ret = amdgpu_ras_eeprom_reset_table(
 		&(amdgpu_ras_get_context(adev)->eeprom_control));
 
@@ -560,14 +707,18 @@ static ssize_t amdgpu_ras_sysfs_read(struct device *dev,
 	if (amdgpu_ras_query_error_status(obj->adev, &info))
 		return -EINVAL;
 
-	if (obj->adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 2) &&
-	    obj->adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 4)) {
+	if (amdgpu_ip_version(obj->adev, MP0_HWIP, 0) != IP_VERSION(11, 0, 2) &&
+	    amdgpu_ip_version(obj->adev, MP0_HWIP, 0) != IP_VERSION(11, 0, 4)) {
 		if (amdgpu_ras_reset_error_status(obj->adev, info.head.block))
 			dev_warn(obj->adev->dev, "Failed to reset error counter and error status");
 	}
 
-	return sysfs_emit(buf, "%s: %lu\n%s: %lu\n", "ue", info.ue_count,
-			  "ce", info.ce_count);
+	if (info.head.block == AMDGPU_RAS_BLOCK__UMC)
+		return sysfs_emit(buf, "%s: %lu\n%s: %lu\n%s: %lu\n", "ue", info.ue_count,
+				"ce", info.ce_count, "de", info.de_count);
+	else
+		return sysfs_emit(buf, "%s: %lu\n%s: %lu\n", "ue", info.ue_count,
+				"ce", info.ce_count);
 }
 
 /* obj begin */
@@ -577,8 +728,11 @@ static ssize_t amdgpu_ras_sysfs_read(struct device *dev,
 
 static inline void put_obj(struct ras_manager *obj)
 {
-	if (obj && (--obj->use == 0))
+	if (obj && (--obj->use == 0)) {
 		list_del(&obj->node);
+		amdgpu_ras_error_data_fini(&obj->err_data);
+	}
+
 	if (obj && (obj->use < 0))
 		DRM_ERROR("RAS ERROR: Unbalance obj(%s) use\n", get_ras_block_str(&obj->head));
 }
@@ -608,6 +762,9 @@ static struct ras_manager *amdgpu_ras_create_obj(struct amdgpu_device *adev,
 	if (alive_obj(obj))
 		return NULL;
 
+	if (amdgpu_ras_error_data_init(&obj->err_data))
+		return NULL;
+
 	obj->head = *head;
 	obj->adev = adev;
 	list_add(&obj->node, &con->head);
@@ -707,28 +864,28 @@ static int __amdgpu_ras_feature_enable(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int amdgpu_ras_check_feature_allowed(struct amdgpu_device *adev,
-		struct ras_common_if *head)
-{
-	if (amdgpu_ras_is_feature_allowed(adev, head) ||
-		amdgpu_ras_is_poison_mode_supported(adev))
-		return 1;
-	else
-		return 0;
-}
-
 /* wrapper of psp_ras_enable_features */
 int amdgpu_ras_feature_enable(struct amdgpu_device *adev,
 		struct ras_common_if *head, bool enable)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 	union ta_ras_cmd_input *info;
-	int ret = 0;
+	int ret;
 
 	if (!con)
 		return -EINVAL;
 
-	if (head->block == AMDGPU_RAS_BLOCK__GFX) {
+	/* For non-gfx ip, do not enable ras feature if it is not allowed */
+	/* For gfx ip, regardless of feature support status, */
+	/* Force issue enable or disable ras feature commands */
+	if (head->block != AMDGPU_RAS_BLOCK__GFX &&
+	    !amdgpu_ras_is_feature_allowed(adev, head))
+		return 0;
+
+	/* Only enable gfx ras feature from host side */
+	if (head->block == AMDGPU_RAS_BLOCK__GFX &&
+	    !amdgpu_sriov_vf(adev) &&
+	    !amdgpu_ras_intr_triggered()) {
 		info = kzalloc(sizeof(union ta_ras_cmd_input), GFP_KERNEL);
 		if (!info)
 			return -ENOMEM;
@@ -744,32 +901,24 @@ int amdgpu_ras_feature_enable(struct amdgpu_device *adev,
 				.error_type = amdgpu_ras_error_to_ta(head->type),
 			};
 		}
-	}
 
-	/* Do not enable if it is not allowed. */
-	if (enable && !amdgpu_ras_check_feature_allowed(adev, head))
-		goto out;
-
-	/* Only enable ras feature operation handle on host side */
-	if (head->block == AMDGPU_RAS_BLOCK__GFX &&
-		!amdgpu_sriov_vf(adev) &&
-		!amdgpu_ras_intr_triggered()) {
 		ret = psp_ras_enable_features(&adev->psp, info, enable);
 		if (ret) {
 			dev_err(adev->dev, "ras %s %s failed poison:%d ret:%d\n",
 				enable ? "enable":"disable",
 				get_ras_block_str(head),
 				amdgpu_ras_is_poison_mode_supported(adev), ret);
-			goto out;
+			kfree(info);
+			return ret;
 		}
+
+		kfree(info);
 	}
 
 	/* setup the obj */
 	__amdgpu_ras_feature_enable(adev, head, enable);
-out:
-	if (head->block == AMDGPU_RAS_BLOCK__GFX)
-		kfree(info);
-	return ret;
+
+	return 0;
 }
 
 /* Only used in device probe stage and called only once. */
@@ -808,7 +957,7 @@ int amdgpu_ras_feature_enable_on_boot(struct amdgpu_device *adev,
 			if (ret)
 				return ret;
 
-			/* gfx block ras dsiable cmd must send to ras-ta */
+			/* gfx block ras disable cmd must send to ras-ta */
 			if (head->block == AMDGPU_RAS_BLOCK__GFX)
 				con->features |= BIT(head->block);
 
@@ -972,103 +1121,524 @@ static void amdgpu_ras_get_ecc_info(struct amdgpu_device *adev, struct ras_err_d
 	}
 }
 
-/* query/inject/cure begin */
-int amdgpu_ras_query_error_status(struct amdgpu_device *adev,
-				  struct ras_query_if *info)
+static void amdgpu_ras_error_print_error_data(struct amdgpu_device *adev,
+					      struct ras_manager *ras_mgr,
+					      struct ras_err_data *err_data,
+					      struct ras_query_context *qctx,
+					      const char *blk_name,
+					      bool is_ue,
+					      bool is_de)
 {
-	struct amdgpu_ras_block_object *block_obj = NULL;
-	struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head);
-	struct ras_err_data err_data = {0, 0, 0, NULL};
+	struct amdgpu_smuio_mcm_config_info *mcm_info;
+	struct ras_err_node *err_node;
+	struct ras_err_info *err_info;
+	u64 event_id = qctx->evid.event_id;
+
+	if (is_ue) {
+		for_each_ras_error(err_node, err_data) {
+			err_info = &err_node->err_info;
+			mcm_info = &err_info->mcm_info;
+			if (err_info->ue_count) {
+				RAS_EVENT_LOG(adev, event_id, "socket: %d, die: %d, "
+					      "%lld new uncorrectable hardware errors detected in %s block\n",
+					      mcm_info->socket_id,
+					      mcm_info->die_id,
+					      err_info->ue_count,
+					      blk_name);
+			}
+		}
 
-	if (!obj)
-		return -EINVAL;
+		for_each_ras_error(err_node, &ras_mgr->err_data) {
+			err_info = &err_node->err_info;
+			mcm_info = &err_info->mcm_info;
+			RAS_EVENT_LOG(adev, event_id, "socket: %d, die: %d, "
+				      "%lld uncorrectable hardware errors detected in total in %s block\n",
+				      mcm_info->socket_id, mcm_info->die_id, err_info->ue_count, blk_name);
+		}
 
-	if (info->head.block == AMDGPU_RAS_BLOCK__UMC) {
-		amdgpu_ras_get_ecc_info(adev, &err_data);
 	} else {
-		block_obj = amdgpu_ras_get_ras_block(adev, info->head.block, 0);
-		if (!block_obj || !block_obj->hw_ops)   {
-			dev_dbg_once(adev->dev, "%s doesn't config RAS function\n",
-				     get_ras_block_str(&info->head));
-			return -EINVAL;
+		if (is_de) {
+			for_each_ras_error(err_node, err_data) {
+				err_info = &err_node->err_info;
+				mcm_info = &err_info->mcm_info;
+				if (err_info->de_count) {
+					RAS_EVENT_LOG(adev, event_id, "socket: %d, die: %d, "
+						      "%lld new deferred hardware errors detected in %s block\n",
+						      mcm_info->socket_id,
+						      mcm_info->die_id,
+						      err_info->de_count,
+						      blk_name);
+				}
+			}
+
+			for_each_ras_error(err_node, &ras_mgr->err_data) {
+				err_info = &err_node->err_info;
+				mcm_info = &err_info->mcm_info;
+				RAS_EVENT_LOG(adev, event_id, "socket: %d, die: %d, "
+					      "%lld deferred hardware errors detected in total in %s block\n",
+					      mcm_info->socket_id, mcm_info->die_id,
+					      err_info->de_count, blk_name);
+			}
+		} else {
+			if (adev->debug_disable_ce_logs)
+				return;
+
+			for_each_ras_error(err_node, err_data) {
+				err_info = &err_node->err_info;
+				mcm_info = &err_info->mcm_info;
+				if (err_info->ce_count) {
+					RAS_EVENT_LOG(adev, event_id, "socket: %d, die: %d, "
+						      "%lld new correctable hardware errors detected in %s block\n",
+						      mcm_info->socket_id,
+						      mcm_info->die_id,
+						      err_info->ce_count,
+						      blk_name);
+				}
+			}
+
+			for_each_ras_error(err_node, &ras_mgr->err_data) {
+				err_info = &err_node->err_info;
+				mcm_info = &err_info->mcm_info;
+				RAS_EVENT_LOG(adev, event_id, "socket: %d, die: %d, "
+					      "%lld correctable hardware errors detected in total in %s block\n",
+					      mcm_info->socket_id, mcm_info->die_id,
+					      err_info->ce_count, blk_name);
+			}
+		}
+	}
+}
+
+static inline bool err_data_has_source_info(struct ras_err_data *data)
+{
+	return !list_empty(&data->err_node_list);
+}
+
+static void amdgpu_ras_error_generate_report(struct amdgpu_device *adev,
+					     struct ras_query_if *query_if,
+					     struct ras_err_data *err_data,
+					     struct ras_query_context *qctx)
+{
+	struct ras_manager *ras_mgr = amdgpu_ras_find_obj(adev, &query_if->head);
+	const char *blk_name = get_ras_block_str(&query_if->head);
+	u64 event_id = qctx->evid.event_id;
+
+	if (err_data->ce_count) {
+		if (err_data_has_source_info(err_data)) {
+			amdgpu_ras_error_print_error_data(adev, ras_mgr, err_data, qctx,
+							  blk_name, false, false);
+		} else if (!adev->aid_mask &&
+			   adev->smuio.funcs &&
+			   adev->smuio.funcs->get_socket_id &&
+			   adev->smuio.funcs->get_die_id) {
+			RAS_EVENT_LOG(adev, event_id, "socket: %d, die: %d "
+				      "%ld correctable hardware errors "
+				      "detected in %s block\n",
+				      adev->smuio.funcs->get_socket_id(adev),
+				      adev->smuio.funcs->get_die_id(adev),
+				      ras_mgr->err_data.ce_count,
+				      blk_name);
+		} else {
+			RAS_EVENT_LOG(adev, event_id, "%ld correctable hardware errors "
+				      "detected in %s block\n",
+				      ras_mgr->err_data.ce_count,
+				      blk_name);
+		}
+	}
+
+	if (err_data->ue_count) {
+		if (err_data_has_source_info(err_data)) {
+			amdgpu_ras_error_print_error_data(adev, ras_mgr, err_data, qctx,
+							  blk_name, true, false);
+		} else if (!adev->aid_mask &&
+			   adev->smuio.funcs &&
+			   adev->smuio.funcs->get_socket_id &&
+			   adev->smuio.funcs->get_die_id) {
+			RAS_EVENT_LOG(adev, event_id, "socket: %d, die: %d "
+				      "%ld uncorrectable hardware errors "
+				      "detected in %s block\n",
+				      adev->smuio.funcs->get_socket_id(adev),
+				      adev->smuio.funcs->get_die_id(adev),
+				      ras_mgr->err_data.ue_count,
+				      blk_name);
+		} else {
+			RAS_EVENT_LOG(adev, event_id, "%ld uncorrectable hardware errors "
+				      "detected in %s block\n",
+				      ras_mgr->err_data.ue_count,
+				      blk_name);
+		}
+	}
+
+	if (err_data->de_count) {
+		if (err_data_has_source_info(err_data)) {
+			amdgpu_ras_error_print_error_data(adev, ras_mgr, err_data, qctx,
+							  blk_name, false, true);
+		} else if (!adev->aid_mask &&
+			   adev->smuio.funcs &&
+			   adev->smuio.funcs->get_socket_id &&
+			   adev->smuio.funcs->get_die_id) {
+			RAS_EVENT_LOG(adev, event_id, "socket: %d, die: %d "
+				      "%ld deferred hardware errors "
+				      "detected in %s block\n",
+				      adev->smuio.funcs->get_socket_id(adev),
+				      adev->smuio.funcs->get_die_id(adev),
+				      ras_mgr->err_data.de_count,
+				      blk_name);
+		} else {
+			RAS_EVENT_LOG(adev, event_id, "%ld deferred hardware errors "
+				      "detected in %s block\n",
+				      ras_mgr->err_data.de_count,
+				      blk_name);
+		}
+	}
+}
+
+static void amdgpu_ras_virt_error_generate_report(struct amdgpu_device *adev,
+						  struct ras_query_if *query_if,
+						  struct ras_err_data *err_data,
+						  struct ras_query_context *qctx)
+{
+	unsigned long new_ue, new_ce, new_de;
+	struct ras_manager *obj = amdgpu_ras_find_obj(adev, &query_if->head);
+	const char *blk_name = get_ras_block_str(&query_if->head);
+	u64 event_id = qctx->evid.event_id;
+
+	new_ce = err_data->ce_count - obj->err_data.ce_count;
+	new_ue = err_data->ue_count - obj->err_data.ue_count;
+	new_de = err_data->de_count - obj->err_data.de_count;
+
+	if (new_ce) {
+		RAS_EVENT_LOG(adev, event_id, "%lu correctable hardware errors "
+			      "detected in %s block\n",
+			      new_ce,
+			      blk_name);
+	}
+
+	if (new_ue) {
+		RAS_EVENT_LOG(adev, event_id, "%lu uncorrectable hardware errors "
+			      "detected in %s block\n",
+			      new_ue,
+			      blk_name);
+	}
+
+	if (new_de) {
+		RAS_EVENT_LOG(adev, event_id, "%lu deferred hardware errors "
+			      "detected in %s block\n",
+			      new_de,
+			      blk_name);
+	}
+}
+
+static void amdgpu_rasmgr_error_data_statistic_update(struct ras_manager *obj, struct ras_err_data *err_data)
+{
+	struct ras_err_node *err_node;
+	struct ras_err_info *err_info;
+
+	if (err_data_has_source_info(err_data)) {
+		for_each_ras_error(err_node, err_data) {
+			err_info = &err_node->err_info;
+			amdgpu_ras_error_statistic_de_count(&obj->err_data,
+					&err_info->mcm_info, err_info->de_count);
+			amdgpu_ras_error_statistic_ce_count(&obj->err_data,
+					&err_info->mcm_info, err_info->ce_count);
+			amdgpu_ras_error_statistic_ue_count(&obj->err_data,
+					&err_info->mcm_info, err_info->ue_count);
 		}
+	} else {
+		/* for legacy asic path which doesn't has error source info */
+		obj->err_data.ue_count += err_data->ue_count;
+		obj->err_data.ce_count += err_data->ce_count;
+		obj->err_data.de_count += err_data->de_count;
+	}
+}
+
+static void amdgpu_ras_mgr_virt_error_data_statistics_update(struct ras_manager *obj,
+							     struct ras_err_data *err_data)
+{
+	/* Host reports absolute counts */
+	obj->err_data.ue_count = err_data->ue_count;
+	obj->err_data.ce_count = err_data->ce_count;
+	obj->err_data.de_count = err_data->de_count;
+}
+
+static struct ras_manager *get_ras_manager(struct amdgpu_device *adev, enum amdgpu_ras_block blk)
+{
+	struct ras_common_if head;
+
+	memset(&head, 0, sizeof(head));
+	head.block = blk;
+
+	return amdgpu_ras_find_obj(adev, &head);
+}
+
+int amdgpu_ras_bind_aca(struct amdgpu_device *adev, enum amdgpu_ras_block blk,
+			const struct aca_info *aca_info, void *data)
+{
+	struct ras_manager *obj;
+
+	/* in resume phase, no need to create aca fs node */
+	if (adev->in_suspend || amdgpu_reset_in_recovery(adev))
+		return 0;
+
+	obj = get_ras_manager(adev, blk);
+	if (!obj)
+		return -EINVAL;
+
+	return amdgpu_aca_add_handle(adev, &obj->aca_handle, ras_block_str(blk), aca_info, data);
+}
+
+int amdgpu_ras_unbind_aca(struct amdgpu_device *adev, enum amdgpu_ras_block blk)
+{
+	struct ras_manager *obj;
+
+	obj = get_ras_manager(adev, blk);
+	if (!obj)
+		return -EINVAL;
+
+	amdgpu_aca_remove_handle(&obj->aca_handle);
+
+	return 0;
+}
+
+static int amdgpu_aca_log_ras_error_data(struct amdgpu_device *adev, enum amdgpu_ras_block blk,
+					 enum aca_error_type type, struct ras_err_data *err_data,
+					 struct ras_query_context *qctx)
+{
+	struct ras_manager *obj;
+
+	obj = get_ras_manager(adev, blk);
+	if (!obj)
+		return -EINVAL;
+
+	return amdgpu_aca_get_error_data(adev, &obj->aca_handle, type, err_data, qctx);
+}
+
+ssize_t amdgpu_ras_aca_sysfs_read(struct device *dev, struct device_attribute *attr,
+				  struct aca_handle *handle, char *buf, void *data)
+{
+	struct ras_manager *obj = container_of(handle, struct ras_manager, aca_handle);
+	struct ras_query_if info = {
+		.head = obj->head,
+	};
+
+	if (!amdgpu_ras_get_error_query_ready(obj->adev))
+		return sysfs_emit(buf, "Query currently inaccessible\n");
+
+	if (amdgpu_ras_query_error_status(obj->adev, &info))
+		return -EINVAL;
 
-		if (block_obj->hw_ops->query_ras_error_count)
-			block_obj->hw_ops->query_ras_error_count(adev, &err_data);
+	return sysfs_emit(buf, "%s: %lu\n%s: %lu\n%s: %lu\n", "ue", info.ue_count,
+			  "ce", info.ce_count, "de", info.de_count);
+}
+
+static int amdgpu_ras_query_error_status_helper(struct amdgpu_device *adev,
+						struct ras_query_if *info,
+						struct ras_err_data *err_data,
+						struct ras_query_context *qctx,
+						unsigned int error_query_mode)
+{
+	enum amdgpu_ras_block blk = info ? info->head.block : AMDGPU_RAS_BLOCK_COUNT;
+	struct amdgpu_ras_block_object *block_obj = NULL;
+	int ret;
 
-		if ((info->head.block == AMDGPU_RAS_BLOCK__SDMA) ||
-		    (info->head.block == AMDGPU_RAS_BLOCK__GFX) ||
-		    (info->head.block == AMDGPU_RAS_BLOCK__MMHUB)) {
+	if (blk == AMDGPU_RAS_BLOCK_COUNT)
+		return -EINVAL;
+
+	if (error_query_mode == AMDGPU_RAS_INVALID_ERROR_QUERY)
+		return -EINVAL;
+
+	if (error_query_mode == AMDGPU_RAS_VIRT_ERROR_COUNT_QUERY) {
+		return amdgpu_virt_req_ras_err_count(adev, blk, err_data);
+	} else if (error_query_mode == AMDGPU_RAS_DIRECT_ERROR_QUERY) {
+		if (info->head.block == AMDGPU_RAS_BLOCK__UMC) {
+			amdgpu_ras_get_ecc_info(adev, err_data);
+		} else {
+			block_obj = amdgpu_ras_get_ras_block(adev, info->head.block, 0);
+			if (!block_obj || !block_obj->hw_ops) {
+				dev_dbg_once(adev->dev, "%s doesn't config RAS function\n",
+					     get_ras_block_str(&info->head));
+				return -EINVAL;
+			}
+
+			if (block_obj->hw_ops->query_ras_error_count)
+				block_obj->hw_ops->query_ras_error_count(adev, err_data);
+
+			if ((info->head.block == AMDGPU_RAS_BLOCK__SDMA) ||
+			    (info->head.block == AMDGPU_RAS_BLOCK__GFX) ||
+			    (info->head.block == AMDGPU_RAS_BLOCK__MMHUB)) {
 				if (block_obj->hw_ops->query_ras_error_status)
 					block_obj->hw_ops->query_ras_error_status(adev);
 			}
+		}
+	} else {
+		if (amdgpu_aca_is_enabled(adev)) {
+			ret = amdgpu_aca_log_ras_error_data(adev, blk, ACA_ERROR_TYPE_UE, err_data, qctx);
+			if (ret)
+				return ret;
+
+			ret = amdgpu_aca_log_ras_error_data(adev, blk, ACA_ERROR_TYPE_CE, err_data, qctx);
+			if (ret)
+				return ret;
+
+			ret = amdgpu_aca_log_ras_error_data(adev, blk, ACA_ERROR_TYPE_DEFERRED, err_data, qctx);
+			if (ret)
+				return ret;
+		} else {
+			/* FIXME: add code to check return value later */
+			amdgpu_mca_smu_log_ras_error(adev, blk, AMDGPU_MCA_ERROR_TYPE_UE, err_data, qctx);
+			amdgpu_mca_smu_log_ras_error(adev, blk, AMDGPU_MCA_ERROR_TYPE_CE, err_data, qctx);
+		}
 	}
 
-	obj->err_data.ue_count += err_data.ue_count;
-	obj->err_data.ce_count += err_data.ce_count;
+	return 0;
+}
+
+/* query/inject/cure begin */
+static int amdgpu_ras_query_error_status_with_event(struct amdgpu_device *adev,
+						    struct ras_query_if *info,
+						    enum ras_event_type type)
+{
+	struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head);
+	struct ras_err_data err_data;
+	struct ras_query_context qctx;
+	unsigned int error_query_mode;
+	int ret;
+
+	if (!obj)
+		return -EINVAL;
+
+	ret = amdgpu_ras_error_data_init(&err_data);
+	if (ret)
+		return ret;
+
+	if (!amdgpu_ras_get_error_query_mode(adev, &error_query_mode))
+		return -EINVAL;
+
+	memset(&qctx, 0, sizeof(qctx));
+	qctx.evid.type = type;
+	qctx.evid.event_id = amdgpu_ras_acquire_event_id(adev, type);
+
+	if (!down_read_trylock(&adev->reset_domain->sem)) {
+		ret = -EIO;
+		goto out_fini_err_data;
+	}
+
+	ret = amdgpu_ras_query_error_status_helper(adev, info,
+						   &err_data,
+						   &qctx,
+						   error_query_mode);
+	up_read(&adev->reset_domain->sem);
+	if (ret)
+		goto out_fini_err_data;
+
+	if (error_query_mode != AMDGPU_RAS_VIRT_ERROR_COUNT_QUERY) {
+		amdgpu_rasmgr_error_data_statistic_update(obj, &err_data);
+		amdgpu_ras_error_generate_report(adev, info, &err_data, &qctx);
+	} else {
+		/* Host provides absolute error counts. First generate the report
+		 * using the previous VF internal count against new host count.
+		 * Then Update VF internal count.
+		 */
+		amdgpu_ras_virt_error_generate_report(adev, info, &err_data, &qctx);
+		amdgpu_ras_mgr_virt_error_data_statistics_update(obj, &err_data);
+	}
 
 	info->ue_count = obj->err_data.ue_count;
 	info->ce_count = obj->err_data.ce_count;
+	info->de_count = obj->err_data.de_count;
 
-	if (err_data.ce_count) {
-		if (adev->smuio.funcs &&
-		    adev->smuio.funcs->get_socket_id &&
-		    adev->smuio.funcs->get_die_id) {
-			dev_info(adev->dev, "socket: %d, die: %d "
-					"%ld correctable hardware errors "
-					"detected in %s block, no user "
-					"action is needed.\n",
-					adev->smuio.funcs->get_socket_id(adev),
-					adev->smuio.funcs->get_die_id(adev),
-					obj->err_data.ce_count,
-					get_ras_block_str(&info->head));
-		} else {
-			dev_info(adev->dev, "%ld correctable hardware errors "
-					"detected in %s block, no user "
-					"action is needed.\n",
-					obj->err_data.ce_count,
-					get_ras_block_str(&info->head));
-		}
-	}
-	if (err_data.ue_count) {
-		if (adev->smuio.funcs &&
-		    adev->smuio.funcs->get_socket_id &&
-		    adev->smuio.funcs->get_die_id) {
-			dev_info(adev->dev, "socket: %d, die: %d "
-					"%ld uncorrectable hardware errors "
-					"detected in %s block\n",
-					adev->smuio.funcs->get_socket_id(adev),
-					adev->smuio.funcs->get_die_id(adev),
-					obj->err_data.ue_count,
-					get_ras_block_str(&info->head));
-		} else {
-			dev_info(adev->dev, "%ld uncorrectable hardware errors "
-					"detected in %s block\n",
-					obj->err_data.ue_count,
-					get_ras_block_str(&info->head));
-		}
+out_fini_err_data:
+	amdgpu_ras_error_data_fini(&err_data);
+
+	return ret;
+}
+
+static int amdgpu_uniras_clear_badpages_info(struct amdgpu_device *adev)
+{
+	struct ras_cmd_dev_handle req = {0};
+	int ret;
+
+	ret = amdgpu_ras_mgr_handle_ras_cmd(adev, RAS_CMD__CLEAR_BAD_PAGE_INFO,
+				&req, sizeof(req), NULL, 0);
+	if (ret) {
+		dev_err(adev->dev, "Failed to clear bad pages info, ret: %d\n", ret);
+		return ret;
 	}
 
 	return 0;
 }
 
-int amdgpu_ras_reset_error_status(struct amdgpu_device *adev,
-		enum amdgpu_ras_block block)
+static int amdgpu_uniras_query_block_ecc(struct amdgpu_device *adev,
+			struct ras_query_if *info)
 {
-	struct amdgpu_ras_block_object *block_obj = amdgpu_ras_get_ras_block(adev, block, 0);
+	struct ras_cmd_block_ecc_info_req req = {0};
+	struct ras_cmd_block_ecc_info_rsp rsp = {0};
+	int ret;
 
-	if (!amdgpu_ras_is_supported(adev, block))
+	if (!info)
 		return -EINVAL;
 
-	if (!block_obj || !block_obj->hw_ops)   {
+	req.block_id = info->head.block;
+	req.subblock_id = info->head.sub_block_index;
+
+	ret = amdgpu_ras_mgr_handle_ras_cmd(adev, RAS_CMD__GET_BLOCK_ECC_STATUS,
+				&req, sizeof(req), &rsp, sizeof(rsp));
+	if (!ret) {
+		info->ce_count = rsp.ce_count;
+		info->ue_count = rsp.ue_count;
+		info->de_count = rsp.de_count;
+	}
+
+	return ret;
+}
+
+int amdgpu_ras_query_error_status(struct amdgpu_device *adev, struct ras_query_if *info)
+{
+	if (amdgpu_uniras_enabled(adev))
+		return amdgpu_uniras_query_block_ecc(adev, info);
+	else
+		return amdgpu_ras_query_error_status_with_event(adev, info, RAS_EVENT_TYPE_INVALID);
+}
+
+int amdgpu_ras_reset_error_count(struct amdgpu_device *adev,
+		enum amdgpu_ras_block block)
+{
+	struct amdgpu_ras_block_object *block_obj = amdgpu_ras_get_ras_block(adev, block, 0);
+	const struct amdgpu_mca_smu_funcs *mca_funcs = adev->mca.mca_funcs;
+	const struct aca_smu_funcs *smu_funcs = adev->aca.smu_funcs;
+
+	if (!block_obj || !block_obj->hw_ops) {
 		dev_dbg_once(adev->dev, "%s doesn't config RAS function\n",
-			     ras_block_str(block));
-		return -EINVAL;
+				ras_block_str(block));
+		return -EOPNOTSUPP;
 	}
 
+	if (!amdgpu_ras_is_supported(adev, block) ||
+	    !amdgpu_ras_get_aca_debug_mode(adev))
+		return -EOPNOTSUPP;
+
+	if (amdgpu_sriov_vf(adev))
+		return -EOPNOTSUPP;
+
+	/* skip ras error reset in gpu reset */
+	if ((amdgpu_in_reset(adev) || amdgpu_ras_in_recovery(adev)) &&
+	    ((smu_funcs && smu_funcs->set_debug_mode) ||
+	     (mca_funcs && mca_funcs->mca_set_debug_mode)))
+		return -EOPNOTSUPP;
+
 	if (block_obj->hw_ops->reset_ras_error_count)
 		block_obj->hw_ops->reset_ras_error_count(adev);
 
+	return 0;
+}
+
+int amdgpu_ras_reset_error_status(struct amdgpu_device *adev,
+		enum amdgpu_ras_block block)
+{
+	struct amdgpu_ras_block_object *block_obj = amdgpu_ras_get_ras_block(adev, block, 0);
+
+	if (amdgpu_ras_reset_error_count(adev, block) == -EOPNOTSUPP)
+		return 0;
+
 	if ((block == AMDGPU_RAS_BLOCK__GFX) ||
 	    (block == AMDGPU_RAS_BLOCK__MMHUB)) {
 		if (block_obj->hw_ops->reset_ras_error_status)
@@ -1078,6 +1648,27 @@ int amdgpu_ras_reset_error_status(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int amdgpu_uniras_error_inject(struct amdgpu_device *adev,
+		struct ras_inject_if *info)
+{
+	struct ras_cmd_inject_error_req inject_req;
+	struct ras_cmd_inject_error_rsp rsp;
+
+	if (!info)
+		return -EINVAL;
+
+	memset(&inject_req, 0, sizeof(inject_req));
+	inject_req.block_id = info->head.block;
+	inject_req.subblock_id = info->head.sub_block_index;
+	inject_req.address = info->address;
+	inject_req.error_type = info->head.type;
+	inject_req.instance_mask = info->instance_mask;
+	inject_req.method = info->value;
+
+	return amdgpu_ras_mgr_handle_ras_cmd(adev, RAS_CMD__INJECT_ERROR,
+			&inject_req, sizeof(inject_req), &rsp, sizeof(rsp));
+}
+
 /* wrapper of psp_ras_trigger_error */
 int amdgpu_ras_error_inject(struct amdgpu_device *adev,
 		struct ras_inject_if *info)
@@ -1095,6 +1686,9 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev,
 							info->head.block,
 							info->head.sub_block_index);
 
+	if (amdgpu_uniras_enabled(adev))
+		return amdgpu_uniras_error_inject(adev, info);
+
 	/* inject on guest isn't allowed, return success directly */
 	if (amdgpu_sriov_vf(adev))
 		return 0;
@@ -1109,21 +1703,22 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev,
 	}
 
 	/* Calculate XGMI relative offset */
-	if (adev->gmc.xgmi.num_physical_nodes > 1) {
+	if (adev->gmc.xgmi.num_physical_nodes > 1 &&
+	    info->head.block != AMDGPU_RAS_BLOCK__GFX) {
 		block_info.address =
 			amdgpu_xgmi_get_relative_phy_addr(adev,
 							  block_info.address);
 	}
 
-	if (info->head.block == AMDGPU_RAS_BLOCK__GFX) {
-		if (block_obj->hw_ops->ras_error_inject)
-			ret = block_obj->hw_ops->ras_error_inject(adev, info);
+	if (block_obj->hw_ops->ras_error_inject) {
+		if (info->head.block == AMDGPU_RAS_BLOCK__GFX)
+			ret = block_obj->hw_ops->ras_error_inject(adev, info, info->instance_mask);
+		else /* Special ras_error_inject is defined (e.g: xgmi) */
+			ret = block_obj->hw_ops->ras_error_inject(adev, &block_info,
+						info->instance_mask);
 	} else {
-		/* If defined special ras_error_inject(e.g: xgmi), implement special ras_error_inject */
-		if (block_obj->hw_ops->ras_error_inject)
-			ret = block_obj->hw_ops->ras_error_inject(adev, &block_info);
-		else  /*If not defined .ras_error_inject, use default ras_error_inject*/
-			ret = psp_ras_trigger_error(&adev->psp, &block_info);
+		/* default path */
+		ret = psp_ras_trigger_error(&adev->psp, &block_info, info->instance_mask);
 	}
 
 	if (ret)
@@ -1163,8 +1758,8 @@ static int amdgpu_ras_query_error_count_helper(struct amdgpu_device *adev,
 
 	/* some hardware/IP supports read to clear
 	 * no need to explictly reset the err status after the query call */
-	if (adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 2) &&
-	    adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 4)) {
+	if (amdgpu_ip_version(adev, MP0_HWIP, 0) != IP_VERSION(11, 0, 2) &&
+	    amdgpu_ip_version(adev, MP0_HWIP, 0) != IP_VERSION(11, 0, 4)) {
 		if (amdgpu_ras_reset_error_status(adev, query_info->head.block))
 			dev_warn(adev->dev,
 				 "Failed to reset error counter and error status\n");
@@ -1238,7 +1833,9 @@ int amdgpu_ras_query_error_count(struct amdgpu_device *adev,
 /* sysfs begin */
 
 static int amdgpu_ras_badpages_read(struct amdgpu_device *adev,
-		struct ras_badpage **bps, unsigned int *count);
+		struct ras_badpage *bps, uint32_t count, uint32_t start);
+static int amdgpu_uniras_badpages_read(struct amdgpu_device *adev,
+		struct ras_badpage *bps, uint32_t count, uint32_t start);
 
 static char *amdgpu_ras_badpage_flags_str(unsigned int flags)
 {
@@ -1284,7 +1881,7 @@ static char *amdgpu_ras_badpage_flags_str(unsigned int flags)
  */
 
 static ssize_t amdgpu_ras_sysfs_badpages_read(struct file *f,
-		struct kobject *kobj, struct bin_attribute *attr,
+		struct kobject *kobj, const struct bin_attribute *attr,
 		char *buf, loff_t ppos, size_t count)
 {
 	struct amdgpu_ras *con =
@@ -1296,19 +1893,50 @@ static ssize_t amdgpu_ras_sysfs_badpages_read(struct file *f,
 	unsigned int end = div64_ul(ppos + count - 1, element_size);
 	ssize_t s = 0;
 	struct ras_badpage *bps = NULL;
-	unsigned int bps_count = 0;
+	int bps_count = 0, i, status;
+	uint64_t address;
 
 	memset(buf, 0, count);
 
-	if (amdgpu_ras_badpages_read(adev, &bps, &bps_count))
+	bps_count = end - start;
+	bps = kmalloc_array(bps_count, sizeof(*bps), GFP_KERNEL);
+	if (!bps)
 		return 0;
 
-	for (; start < end && start < bps_count; start++)
+	memset(bps, 0, sizeof(*bps) * bps_count);
+
+	if (amdgpu_uniras_enabled(adev))
+		bps_count = amdgpu_uniras_badpages_read(adev, bps, bps_count, start);
+	else
+		bps_count = amdgpu_ras_badpages_read(adev, bps, bps_count, start);
+
+	if (bps_count <= 0) {
+		kfree(bps);
+		return 0;
+	}
+
+	for (i = 0; i < bps_count; i++) {
+		address = ((uint64_t)bps[i].bp) << AMDGPU_GPU_PAGE_SHIFT;
+		if (amdgpu_ras_check_critical_address(adev, address))
+			continue;
+
+		bps[i].size = AMDGPU_GPU_PAGE_SIZE;
+
+		status = amdgpu_vram_mgr_query_page_status(&adev->mman.vram_mgr,
+					address);
+		if (status == -EBUSY)
+			bps[i].flags = AMDGPU_RAS_RETIRE_PAGE_PENDING;
+		else if (status == -ENOENT)
+			bps[i].flags = AMDGPU_RAS_RETIRE_PAGE_FAULT;
+		else
+			bps[i].flags = AMDGPU_RAS_RETIRE_PAGE_RESERVED;
+
 		s += scnprintf(&buf[s], element_size + 1,
 				"0x%08x : 0x%08x : %1s\n",
-				bps[start].bp,
-				bps[start].size,
-				amdgpu_ras_badpage_flags_str(bps[start].flags));
+				bps[i].bp,
+				bps[i].size,
+				amdgpu_ras_badpage_flags_str(bps[i].flags));
+	}
 
 	kfree(bps);
 
@@ -1324,20 +1952,103 @@ static ssize_t amdgpu_ras_sysfs_features_read(struct device *dev,
 	return sysfs_emit(buf, "feature mask: 0x%x\n", con->features);
 }
 
+static bool amdgpu_ras_get_version_info(struct amdgpu_device *adev, u32 *major,
+			u32 *minor, u32 *rev)
+{
+	int i;
+
+	if (!adev || !major || !minor || !rev || !amdgpu_uniras_enabled(adev))
+		return false;
+
+	for (i = 0; i < adev->num_ip_blocks; i++) {
+		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_RAS) {
+			*major = adev->ip_blocks[i].version->major;
+			*minor = adev->ip_blocks[i].version->minor;
+			*rev = adev->ip_blocks[i].version->rev;
+			return true;
+		}
+	}
+
+	return false;
+}
+
+static ssize_t amdgpu_ras_sysfs_version_show(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct amdgpu_ras *con =
+		container_of(attr, struct amdgpu_ras, version_attr);
+	u32 major, minor, rev;
+	ssize_t size = 0;
+
+	size += sysfs_emit_at(buf, size, "table version: 0x%x\n",
+			con->eeprom_control.tbl_hdr.version);
+
+	if (amdgpu_ras_get_version_info(con->adev, &major, &minor, &rev))
+		size += sysfs_emit_at(buf, size, "ras version: %u.%u.%u\n",
+			major, minor, rev);
+
+	return size;
+}
+
+static ssize_t amdgpu_ras_sysfs_schema_show(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct amdgpu_ras *con =
+		container_of(attr, struct amdgpu_ras, schema_attr);
+	return sysfs_emit(buf, "schema: 0x%x\n", con->schema);
+}
+
+static struct {
+	enum ras_event_type type;
+	const char *name;
+} dump_event[] = {
+	{RAS_EVENT_TYPE_FATAL, "Fatal Error"},
+	{RAS_EVENT_TYPE_POISON_CREATION, "Poison Creation"},
+	{RAS_EVENT_TYPE_POISON_CONSUMPTION, "Poison Consumption"},
+};
+
+static ssize_t amdgpu_ras_sysfs_event_state_show(struct device *dev,
+						 struct device_attribute *attr, char *buf)
+{
+	struct amdgpu_ras *con =
+		container_of(attr, struct amdgpu_ras, event_state_attr);
+	struct ras_event_manager *event_mgr = con->event_mgr;
+	struct ras_event_state *event_state;
+	int i, size = 0;
+
+	if (!event_mgr)
+		return -EINVAL;
+
+	size += sysfs_emit_at(buf, size, "current seqno: %llu\n", atomic64_read(&event_mgr->seqno));
+	for (i = 0; i < ARRAY_SIZE(dump_event); i++) {
+		event_state = &event_mgr->event_state[dump_event[i].type];
+		size += sysfs_emit_at(buf, size, "%s: count:%llu, last_seqno:%llu\n",
+				      dump_event[i].name,
+				      atomic64_read(&event_state->count),
+				      event_state->last_seqno);
+	}
+
+	return (ssize_t)size;
+}
+
 static void amdgpu_ras_sysfs_remove_bad_page_node(struct amdgpu_device *adev)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 
-	sysfs_remove_file_from_group(&adev->dev->kobj,
+	if (adev->dev->kobj.sd)
+		sysfs_remove_file_from_group(&adev->dev->kobj,
 				&con->badpages_attr.attr,
 				RAS_FS_NAME);
 }
 
-static int amdgpu_ras_sysfs_remove_feature_node(struct amdgpu_device *adev)
+static int amdgpu_ras_sysfs_remove_dev_attr_node(struct amdgpu_device *adev)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 	struct attribute *attrs[] = {
 		&con->features_attr.attr,
+		&con->version_attr.attr,
+		&con->schema_attr.attr,
+		&con->event_state_attr.attr,
 		NULL
 	};
 	struct attribute_group group = {
@@ -1345,7 +2056,8 @@ static int amdgpu_ras_sysfs_remove_feature_node(struct amdgpu_device *adev)
 		.attrs = attrs,
 	};
 
-	sysfs_remove_group(&adev->dev->kobj, &group);
+	if (adev->dev->kobj.sd)
+		sysfs_remove_group(&adev->dev->kobj, &group);
 
 	return 0;
 }
@@ -1355,9 +2067,15 @@ int amdgpu_ras_sysfs_create(struct amdgpu_device *adev,
 {
 	struct ras_manager *obj = amdgpu_ras_find_obj(adev, head);
 
+	if (amdgpu_aca_is_enabled(adev))
+		return 0;
+
 	if (!obj || obj->attr_inuse)
 		return -EINVAL;
 
+	if (amdgpu_sriov_vf(adev) && !amdgpu_virt_ras_telemetry_block_en(adev, head->block))
+		return 0;
+
 	get_obj(obj);
 
 	snprintf(obj->fs_data.sysfs_name, sizeof(obj->fs_data.sysfs_name),
@@ -1389,10 +2107,14 @@ int amdgpu_ras_sysfs_remove(struct amdgpu_device *adev,
 {
 	struct ras_manager *obj = amdgpu_ras_find_obj(adev, head);
 
+	if (amdgpu_aca_is_enabled(adev))
+		return 0;
+
 	if (!obj || !obj->attr_inuse)
 		return -EINVAL;
 
-	sysfs_remove_file_from_group(&adev->dev->kobj,
+	if (adev->dev->kobj.sd)
+		sysfs_remove_file_from_group(&adev->dev->kobj,
 				&obj->sysfs_attr.attr,
 				RAS_FS_NAME);
 	obj->attr_inuse = 0;
@@ -1413,7 +2135,7 @@ static int amdgpu_ras_sysfs_remove_all(struct amdgpu_device *adev)
 	if (amdgpu_bad_page_threshold != 0)
 		amdgpu_ras_sysfs_remove_bad_page_node(adev);
 
-	amdgpu_ras_sysfs_remove_feature_node(adev);
+	amdgpu_ras_sysfs_remove_dev_attr_node(adev);
 
 	return 0;
 }
@@ -1441,6 +2163,7 @@ static int amdgpu_ras_sysfs_remove_all(struct amdgpu_device *adev)
 static struct dentry *amdgpu_ras_debugfs_create_ctrl_node(struct amdgpu_device *adev)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct amdgpu_ras_eeprom_control *eeprom = &con->eeprom_control;
 	struct drm_minor  *minor = adev_to_drm(adev)->primary;
 	struct dentry     *dir;
 
@@ -1451,6 +2174,7 @@ static struct dentry *amdgpu_ras_debugfs_create_ctrl_node(struct amdgpu_device *
 			    &amdgpu_ras_debugfs_eeprom_ops);
 	debugfs_create_u32("bad_page_cnt_threshold", 0444, dir,
 			   &con->bad_page_cnt_threshold);
+	debugfs_create_u32("ras_num_recs", 0444, dir, &eeprom->ras_num_recs);
 	debugfs_create_x32("ras_hw_enabled", 0444, dir, &adev->ras_hw_enabled);
 	debugfs_create_x32("ras_enabled", 0444, dir, &adev->ras_enabled);
 	debugfs_create_file("ras_eeprom_size", S_IRUGO, dir, adev,
@@ -1498,6 +2222,24 @@ static void amdgpu_ras_debugfs_create(struct amdgpu_device *adev,
 			    obj, &amdgpu_ras_debugfs_ops);
 }
 
+static bool amdgpu_ras_aca_is_supported(struct amdgpu_device *adev)
+{
+	bool ret;
+
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 12):
+	case IP_VERSION(13, 0, 14):
+		ret = true;
+		break;
+	default:
+		ret = false;
+		break;
+	}
+
+	return ret;
+}
+
 void amdgpu_ras_debugfs_create_all(struct amdgpu_device *adev)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
@@ -1523,15 +2265,28 @@ void amdgpu_ras_debugfs_create_all(struct amdgpu_device *adev)
 			amdgpu_ras_debugfs_create(adev, &fs_info, dir);
 		}
 	}
+
+	if (amdgpu_ras_aca_is_supported(adev)) {
+		if (amdgpu_aca_is_enabled(adev))
+			amdgpu_aca_smu_debugfs_init(adev, dir);
+		else
+			amdgpu_mca_smu_debugfs_init(adev, dir);
+	}
 }
 
 /* debugfs end */
 
 /* ras fs */
-static BIN_ATTR(gpu_vram_bad_pages, S_IRUGO,
-		amdgpu_ras_sysfs_badpages_read, NULL, 0);
+static const BIN_ATTR(gpu_vram_bad_pages, S_IRUGO,
+		      amdgpu_ras_sysfs_badpages_read, NULL, 0);
 static DEVICE_ATTR(features, S_IRUGO,
 		amdgpu_ras_sysfs_features_read, NULL);
+static DEVICE_ATTR(version, 0444,
+		amdgpu_ras_sysfs_version_show, NULL);
+static DEVICE_ATTR(schema, 0444,
+		amdgpu_ras_sysfs_schema_show, NULL);
+static DEVICE_ATTR(event_state, 0444,
+		   amdgpu_ras_sysfs_event_state_show, NULL);
 static int amdgpu_ras_fs_init(struct amdgpu_device *adev)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
@@ -1540,26 +2295,41 @@ static int amdgpu_ras_fs_init(struct amdgpu_device *adev)
 	};
 	struct attribute *attrs[] = {
 		&con->features_attr.attr,
+		&con->version_attr.attr,
+		&con->schema_attr.attr,
+		&con->event_state_attr.attr,
 		NULL
 	};
-	struct bin_attribute *bin_attrs[] = {
+	const struct bin_attribute *bin_attrs[] = {
 		NULL,
 		NULL,
 	};
 	int r;
 
+	group.attrs = attrs;
+
 	/* add features entry */
 	con->features_attr = dev_attr_features;
-	group.attrs = attrs;
 	sysfs_attr_init(attrs[0]);
 
+	/* add version entry */
+	con->version_attr = dev_attr_version;
+	sysfs_attr_init(attrs[1]);
+
+	/* add schema entry */
+	con->schema_attr = dev_attr_schema;
+	sysfs_attr_init(attrs[2]);
+
+	/* add event_state entry */
+	con->event_state_attr = dev_attr_event_state;
+	sysfs_attr_init(attrs[3]);
+
 	if (amdgpu_bad_page_threshold != 0) {
 		/* add bad_page_features entry */
-		bin_attr_gpu_vram_bad_pages.private = NULL;
 		con->badpages_attr = bin_attr_gpu_vram_bad_pages;
+		sysfs_bin_attr_init(&con->badpages_attr);
 		bin_attrs[0] = &con->badpages_attr;
 		group.bin_attrs = bin_attrs;
-		sysfs_bin_attr_init(bin_attrs[0]);
 	}
 
 	r = sysfs_create_group(&adev->dev->kobj, &group);
@@ -1597,10 +2367,24 @@ static int amdgpu_ras_fs_fini(struct amdgpu_device *adev)
 void amdgpu_ras_interrupt_fatal_error_handler(struct amdgpu_device *adev)
 {
 	/* Fatal error events are handled on host side */
-	if (amdgpu_sriov_vf(adev) ||
-		!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__PCIE_BIF))
+	if (amdgpu_sriov_vf(adev))
+		return;
+	/*
+	 * If the current interrupt is caused by a non-fatal RAS error, skip
+	 * check for fatal error. For fatal errors, FED status of all devices
+	 * in XGMI hive gets set when the first device gets fatal error
+	 * interrupt. The error gets propagated to other devices as well, so
+	 * make sure to ack the interrupt regardless of FED status.
+	 */
+	if (!amdgpu_ras_get_fed_status(adev) &&
+	    amdgpu_ras_is_err_state(adev, AMDGPU_RAS_BLOCK__ANY))
 		return;
 
+	if (amdgpu_uniras_enabled(adev)) {
+		amdgpu_ras_mgr_handle_fatal_interrupt(adev, NULL);
+		return;
+	}
+
 	if (adev->nbio.ras &&
 	    adev->nbio.ras->handle_ras_controller_intr_no_bifring)
 		adev->nbio.ras->handle_ras_controller_intr_no_bifring(adev);
@@ -1617,10 +2401,19 @@ static void amdgpu_ras_interrupt_poison_consumption_handler(struct ras_manager *
 	struct amdgpu_device *adev = obj->adev;
 	struct amdgpu_ras_block_object *block_obj =
 		amdgpu_ras_get_ras_block(adev, obj->head.block, 0);
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	enum ras_event_type type = RAS_EVENT_TYPE_POISON_CONSUMPTION;
+	u64 event_id;
+	int ret;
 
-	if (!block_obj)
+	if (!block_obj || !con)
 		return;
 
+	ret = amdgpu_ras_mark_ras_event(adev, type);
+	if (ret)
+		return;
+
+	amdgpu_ras_set_err_poison(adev, block_obj->ras_comm.block);
 	/* both query_poison_status and handle_poison_consumption are optional,
 	 * but at least one of them should be implemented if we need poison
 	 * consumption handler
@@ -1636,42 +2429,69 @@ static void amdgpu_ras_interrupt_poison_consumption_handler(struct ras_manager *
 		}
 	}
 
-	if (!adev->gmc.xgmi.connected_to_cpu)
-		amdgpu_umc_poison_handler(adev, false);
+	amdgpu_umc_poison_handler(adev, obj->head.block, 0);
 
 	if (block_obj->hw_ops && block_obj->hw_ops->handle_poison_consumption)
 		poison_stat = block_obj->hw_ops->handle_poison_consumption(adev);
 
-	/* gpu reset is fallback for failed and default cases */
-	if (poison_stat) {
-		dev_info(adev->dev, "GPU reset for %s RAS poison consumption is issued!\n",
-				block_obj->ras_comm.name);
+	/* gpu reset is fallback for failed and default cases.
+	 * For RMA case, amdgpu_umc_poison_handler will handle gpu reset.
+	 */
+	if (poison_stat && !amdgpu_ras_is_rma(adev)) {
+		event_id = amdgpu_ras_acquire_event_id(adev, type);
+		RAS_EVENT_LOG(adev, event_id,
+			      "GPU reset for %s RAS poison consumption is issued!\n",
+			      block_obj->ras_comm.name);
 		amdgpu_ras_reset_gpu(adev);
-	} else {
-		amdgpu_gfx_poison_consumption_handler(adev, entry);
 	}
+
+	if (!poison_stat)
+		amdgpu_gfx_poison_consumption_handler(adev, entry);
 }
 
 static void amdgpu_ras_interrupt_poison_creation_handler(struct ras_manager *obj,
 				struct amdgpu_iv_entry *entry)
 {
-	dev_info(obj->adev->dev,
-		"Poison is created, no user action is needed.\n");
+	struct amdgpu_device *adev = obj->adev;
+	enum ras_event_type type = RAS_EVENT_TYPE_POISON_CREATION;
+	u64 event_id;
+	int ret;
+
+	ret = amdgpu_ras_mark_ras_event(adev, type);
+	if (ret)
+		return;
+
+	event_id = amdgpu_ras_acquire_event_id(adev, type);
+	RAS_EVENT_LOG(adev, event_id, "Poison is created\n");
+
+	if (amdgpu_ip_version(obj->adev, UMC_HWIP, 0) >= IP_VERSION(12, 0, 0)) {
+		struct amdgpu_ras *con = amdgpu_ras_get_context(obj->adev);
+
+		atomic_inc(&con->page_retirement_req_cnt);
+		atomic_inc(&con->poison_creation_count);
+
+		wake_up(&con->page_retirement_wq);
+	}
 }
 
 static void amdgpu_ras_interrupt_umc_handler(struct ras_manager *obj,
 				struct amdgpu_iv_entry *entry)
 {
 	struct ras_ih_data *data = &obj->ih_data;
-	struct ras_err_data err_data = {0, 0, 0, NULL};
+	struct ras_err_data err_data;
 	int ret;
 
 	if (!data->cb)
 		return;
 
+	ret = amdgpu_ras_error_data_init(&err_data);
+	if (ret)
+		return;
+
 	/* Let IP handle its data, maybe we need get the output
 	 * from the callback to update the error type/count, etc
 	 */
+	amdgpu_ras_set_fed(obj->adev, true);
 	ret = data->cb(obj->adev, &err_data, entry);
 	/* ue will trigger an interrupt, and in that case
 	 * we need do a reset to recovery the whole system.
@@ -1684,7 +2504,10 @@ static void amdgpu_ras_interrupt_umc_handler(struct ras_manager *obj,
 		 */
 		obj->err_data.ue_count += err_data.ue_count;
 		obj->err_data.ce_count += err_data.ce_count;
+		obj->err_data.de_count += err_data.de_count;
 	}
+
+	amdgpu_ras_error_data_fini(&err_data);
 }
 
 static void amdgpu_ras_interrupt_handler(struct ras_manager *obj)
@@ -1729,12 +2552,25 @@ static void amdgpu_ras_interrupt_process_handler(struct work_struct *work)
 int amdgpu_ras_interrupt_dispatch(struct amdgpu_device *adev,
 		struct ras_dispatch_if *info)
 {
-	struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head);
-	struct ras_ih_data *data = &obj->ih_data;
+	struct ras_manager *obj;
+	struct ras_ih_data *data;
 
+	if (amdgpu_uniras_enabled(adev)) {
+		struct ras_ih_info ih_info;
+
+		memset(&ih_info, 0, sizeof(ih_info));
+		ih_info.block = info->head.block;
+		memcpy(&ih_info.iv_entry, info->entry, sizeof(struct amdgpu_iv_entry));
+
+		return amdgpu_ras_mgr_handle_controller_interrupt(adev, &ih_info);
+	}
+
+	obj = amdgpu_ras_find_obj(adev, &info->head);
 	if (!obj)
 		return -EINVAL;
 
+	data = &obj->ih_data;
+
 	if (data->inuse == 0)
 		return 0;
 
@@ -1831,7 +2667,7 @@ static int amdgpu_ras_interrupt_remove_all(struct amdgpu_device *adev)
 /* ih end */
 
 /* traversal all IPs except NBIO to query error counter */
-static void amdgpu_ras_log_on_err_counter(struct amdgpu_device *adev)
+static void amdgpu_ras_log_on_err_counter(struct amdgpu_device *adev, enum ras_event_type type)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 	struct ras_manager *obj;
@@ -1860,14 +2696,18 @@ static void amdgpu_ras_log_on_err_counter(struct amdgpu_device *adev)
 		 * should be removed until smu fix handle ecc_info table.
 		 */
 		if ((info.head.block == AMDGPU_RAS_BLOCK__UMC) &&
-			(adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 2)))
+		    (amdgpu_ip_version(adev, MP1_HWIP, 0) ==
+		     IP_VERSION(13, 0, 2)))
 			continue;
 
-		amdgpu_ras_query_error_status(adev, &info);
+		amdgpu_ras_query_error_status_with_event(adev, &info, type);
 
-		if (adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 2) &&
-		    adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 4) &&
-		    adev->ip_versions[MP0_HWIP][0] != IP_VERSION(13, 0, 0)) {
+		if (amdgpu_ip_version(adev, MP0_HWIP, 0) !=
+			    IP_VERSION(11, 0, 2) &&
+		    amdgpu_ip_version(adev, MP0_HWIP, 0) !=
+			    IP_VERSION(11, 0, 4) &&
+		    amdgpu_ip_version(adev, MP0_HWIP, 0) !=
+			    IP_VERSION(13, 0, 0)) {
 			if (amdgpu_ras_reset_error_status(adev, info.head.block))
 				dev_warn(adev->dev, "Failed to reset error counter and error status");
 		}
@@ -1919,54 +2759,121 @@ static void amdgpu_ras_query_err_status(struct amdgpu_device *adev)
 	}
 }
 
-/* recovery begin */
-
-/* return 0 on success.
- * caller need free bps.
- */
 static int amdgpu_ras_badpages_read(struct amdgpu_device *adev,
-		struct ras_badpage **bps, unsigned int *count)
+		struct ras_badpage *bps, uint32_t count, uint32_t start)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 	struct ras_err_handler_data *data;
-	int i = 0;
-	int ret = 0, status;
+	int r = 0;
+	uint32_t i;
 
 	if (!con || !con->eh_data || !bps || !count)
 		return -EINVAL;
 
 	mutex_lock(&con->recovery_lock);
 	data = con->eh_data;
-	if (!data || data->count == 0) {
-		*bps = NULL;
-		ret = -EINVAL;
-		goto out;
+	if (start < data->count) {
+		for (i = start; i < data->count; i++) {
+			if (!data->bps[i].ts)
+				continue;
+
+			bps[r].bp = data->bps[i].retired_page;
+			r++;
+			if (r >= count)
+				break;
+		}
 	}
+	mutex_unlock(&con->recovery_lock);
 
-	*bps = kmalloc(sizeof(struct ras_badpage) * data->count, GFP_KERNEL);
-	if (!*bps) {
-		ret = -ENOMEM;
-		goto out;
-	}
+	return r;
+}
 
-	for (; i < data->count; i++) {
-		(*bps)[i] = (struct ras_badpage){
-			.bp = data->bps[i].retired_page,
-			.size = AMDGPU_GPU_PAGE_SIZE,
-			.flags = AMDGPU_RAS_RETIRE_PAGE_RESERVED,
-		};
-		status = amdgpu_vram_mgr_query_page_status(&adev->mman.vram_mgr,
-				data->bps[i].retired_page);
-		if (status == -EBUSY)
-			(*bps)[i].flags = AMDGPU_RAS_RETIRE_PAGE_PENDING;
-		else if (status == -ENOENT)
-			(*bps)[i].flags = AMDGPU_RAS_RETIRE_PAGE_FAULT;
+static int amdgpu_uniras_badpages_read(struct amdgpu_device *adev,
+		struct ras_badpage *bps, uint32_t count, uint32_t start)
+{
+	struct ras_cmd_bad_pages_info_req cmd_input;
+	struct ras_cmd_bad_pages_info_rsp *output;
+	uint32_t group, start_group, end_group;
+	uint32_t pos, pos_in_group;
+	int r = 0, i;
+
+	if (!bps || !count)
+		return -EINVAL;
+
+	output = kmalloc(sizeof(*output), GFP_KERNEL);
+	if (!output)
+		return -ENOMEM;
+
+	memset(&cmd_input, 0, sizeof(cmd_input));
+
+	start_group = start / RAS_CMD_MAX_BAD_PAGES_PER_GROUP;
+	end_group = (start + count + RAS_CMD_MAX_BAD_PAGES_PER_GROUP - 1) /
+				RAS_CMD_MAX_BAD_PAGES_PER_GROUP;
+
+	pos = start;
+	for (group = start_group; group < end_group; group++) {
+		memset(output, 0, sizeof(*output));
+		cmd_input.group_index = group;
+		if (amdgpu_ras_mgr_handle_ras_cmd(adev, RAS_CMD__GET_BAD_PAGES,
+			&cmd_input, sizeof(cmd_input), output, sizeof(*output)))
+			goto out;
+
+		if (pos >= output->bp_total_cnt)
+			goto out;
+
+		pos_in_group = pos - group * RAS_CMD_MAX_BAD_PAGES_PER_GROUP;
+		for (i = pos_in_group; i < output->bp_in_group; i++, pos++) {
+			if (!output->records[i].ts)
+				continue;
+
+			bps[r].bp = output->records[i].retired_page;
+			r++;
+			if (r >= count)
+				goto out;
+		}
 	}
 
-	*count = data->count;
 out:
-	mutex_unlock(&con->recovery_lock);
-	return ret;
+	kfree(output);
+	return r;
+}
+
+static void amdgpu_ras_set_fed_all(struct amdgpu_device *adev,
+				   struct amdgpu_hive_info *hive, bool status)
+{
+	struct amdgpu_device *tmp_adev;
+
+	if (hive) {
+		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head)
+			amdgpu_ras_set_fed(tmp_adev, status);
+	} else {
+		amdgpu_ras_set_fed(adev, status);
+	}
+}
+
+bool amdgpu_ras_in_recovery(struct amdgpu_device *adev)
+{
+	struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev);
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+	int hive_ras_recovery = 0;
+
+	if (hive) {
+		hive_ras_recovery = atomic_read(&hive->ras_recovery);
+		amdgpu_put_xgmi_hive(hive);
+	}
+
+	if (ras && (atomic_read(&ras->in_recovery) || hive_ras_recovery))
+		return true;
+
+	return false;
+}
+
+static enum ras_event_type amdgpu_ras_get_fatal_error_event(struct amdgpu_device *adev)
+{
+	if (amdgpu_ras_intr_triggered())
+		return RAS_EVENT_TYPE_FATAL;
+	else
+		return RAS_EVENT_TYPE_POISON_CONSUMPTION;
 }
 
 static void amdgpu_ras_do_recovery(struct work_struct *work)
@@ -1976,9 +2883,26 @@ static void amdgpu_ras_do_recovery(struct work_struct *work)
 	struct amdgpu_device *remote_adev = NULL;
 	struct amdgpu_device *adev = ras->adev;
 	struct list_head device_list, *device_list_handle =  NULL;
+	struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev);
+	unsigned int error_query_mode;
+	enum ras_event_type type;
+
+	if (hive) {
+		atomic_set(&hive->ras_recovery, 1);
 
+		/* If any device which is part of the hive received RAS fatal
+		 * error interrupt, set fatal error status on all. This
+		 * condition will need a recovery, and flag will be cleared
+		 * as part of recovery.
+		 */
+		list_for_each_entry(remote_adev, &hive->device_list,
+				    gmc.xgmi.head)
+			if (amdgpu_ras_get_fed_status(remote_adev)) {
+				amdgpu_ras_set_fed_all(adev, hive, true);
+				break;
+			}
+	}
 	if (!ras->disable_ras_err_cnt_harvest) {
-		struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev);
 
 		/* Build list of devices to query RAS related errors */
 		if  (hive && adev->gmc.xgmi.num_physical_nodes > 1) {
@@ -1989,13 +2913,24 @@ static void amdgpu_ras_do_recovery(struct work_struct *work)
 			device_list_handle = &device_list;
 		}
 
+		if (amdgpu_ras_get_error_query_mode(adev, &error_query_mode)) {
+			if (error_query_mode == AMDGPU_RAS_FIRMWARE_ERROR_QUERY) {
+				/* wait 500ms to ensure pmfw polling mca bank info done */
+				msleep(500);
+			}
+		}
+
+		type = amdgpu_ras_get_fatal_error_event(adev);
 		list_for_each_entry(remote_adev,
 				device_list_handle, gmc.xgmi.head) {
-			amdgpu_ras_query_err_status(remote_adev);
-			amdgpu_ras_log_on_err_counter(remote_adev);
+			if (amdgpu_uniras_enabled(remote_adev)) {
+				amdgpu_ras_mgr_update_ras_ecc(remote_adev);
+			} else {
+				amdgpu_ras_query_err_status(remote_adev);
+				amdgpu_ras_log_on_err_counter(remote_adev, type);
+			}
 		}
 
-		amdgpu_put_xgmi_hive(hive);
 	}
 
 	if (amdgpu_device_should_recover_gpu(ras->adev)) {
@@ -2004,16 +2939,38 @@ static void amdgpu_ras_do_recovery(struct work_struct *work)
 
 		reset_context.method = AMD_RESET_METHOD_NONE;
 		reset_context.reset_req_dev = adev;
+		reset_context.src = AMDGPU_RESET_SRC_RAS;
+		set_bit(AMDGPU_SKIP_COREDUMP, &reset_context.flags);
 
 		/* Perform full reset in fatal error mode */
 		if (!amdgpu_ras_is_poison_mode_supported(ras->adev))
 			set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
-		else
+		else {
 			clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
 
+			if (ras->gpu_reset_flags & AMDGPU_RAS_GPU_RESET_MODE2_RESET) {
+				ras->gpu_reset_flags &= ~AMDGPU_RAS_GPU_RESET_MODE2_RESET;
+				reset_context.method = AMD_RESET_METHOD_MODE2;
+			}
+
+			/* Fatal error occurs in poison mode, mode1 reset is used to
+			 * recover gpu.
+			 */
+			if (ras->gpu_reset_flags & AMDGPU_RAS_GPU_RESET_MODE1_RESET) {
+				ras->gpu_reset_flags &= ~AMDGPU_RAS_GPU_RESET_MODE1_RESET;
+				set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
+
+				psp_fatal_error_recovery_quirk(&adev->psp);
+			}
+		}
+
 		amdgpu_device_gpu_recover(ras->adev, NULL, &reset_context);
 	}
 	atomic_set(&ras->in_recovery, 0);
+	if (hive) {
+		atomic_set(&hive->ras_recovery, 0);
+		amdgpu_put_xgmi_hive(hive);
+	}
 }
 
 /* alloc/realloc bps array */
@@ -2023,7 +2980,7 @@ static int amdgpu_ras_realloc_eh_data_space(struct amdgpu_device *adev,
 	unsigned int old_space = data->count + data->space_left;
 	unsigned int new_space = old_space + pages;
 	unsigned int align_space = ALIGN(new_space, 512);
-	void *bps = kmalloc(align_space * sizeof(*data->bps), GFP_KERNEL);
+	void *bps = kmalloc_array(align_space, sizeof(*data->bps), GFP_KERNEL);
 
 	if (!bps) {
 		return -ENOMEM;
@@ -2040,43 +2997,283 @@ static int amdgpu_ras_realloc_eh_data_space(struct amdgpu_device *adev,
 	return 0;
 }
 
-/* it deal with vram only. */
-int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev,
-		struct eeprom_table_record *bps, int pages)
+static int amdgpu_ras_mca2pa_by_idx(struct amdgpu_device *adev,
+			struct eeprom_table_record *bps,
+			struct ras_err_data *err_data)
 {
-	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
-	struct ras_err_handler_data *data;
+	struct ta_ras_query_address_input addr_in;
+	uint32_t socket = 0;
 	int ret = 0;
-	uint32_t i;
 
-	if (!con || !con->eh_data || !bps || pages <= 0)
-		return 0;
+	if (adev->smuio.funcs && adev->smuio.funcs->get_socket_id)
+		socket = adev->smuio.funcs->get_socket_id(adev);
 
-	mutex_lock(&con->recovery_lock);
-	data = con->eh_data;
-	if (!data)
-		goto out;
+	/* reinit err_data */
+	err_data->err_addr_cnt = 0;
+	err_data->err_addr_len = adev->umc.retire_unit;
+
+	memset(&addr_in, 0, sizeof(addr_in));
+	addr_in.ma.err_addr = bps->address;
+	addr_in.ma.socket_id = socket;
+	addr_in.ma.ch_inst = bps->mem_channel;
+	if (!amdgpu_ras_smu_eeprom_supported(adev)) {
+		/* tell RAS TA the node instance is not used */
+		addr_in.ma.node_inst = TA_RAS_INV_NODE;
+	} else {
+		addr_in.ma.umc_inst = bps->mcumc_id;
+		addr_in.ma.node_inst = bps->cu;
+	}
+
+	if (adev->umc.ras && adev->umc.ras->convert_ras_err_addr)
+		ret = adev->umc.ras->convert_ras_err_addr(adev, err_data,
+				&addr_in, NULL, false);
+
+	return ret;
+}
+
+static int amdgpu_ras_mca2pa(struct amdgpu_device *adev,
+			struct eeprom_table_record *bps,
+			struct ras_err_data *err_data)
+{
+	struct ta_ras_query_address_input addr_in;
+	uint32_t die_id, socket = 0;
+
+	if (adev->smuio.funcs && adev->smuio.funcs->get_socket_id)
+		socket = adev->smuio.funcs->get_socket_id(adev);
+
+	/* although die id is gotten from PA in nps1 mode, the id is
+	 * fitable for any nps mode
+	 */
+	if (adev->umc.ras && adev->umc.ras->get_die_id_from_pa)
+		die_id = adev->umc.ras->get_die_id_from_pa(adev, bps->address,
+					bps->retired_page << AMDGPU_GPU_PAGE_SHIFT);
+	else
+		return -EINVAL;
+
+	/* reinit err_data */
+	err_data->err_addr_cnt = 0;
+	err_data->err_addr_len = adev->umc.retire_unit;
 
-	for (i = 0; i < pages; i++) {
+	memset(&addr_in, 0, sizeof(addr_in));
+	addr_in.ma.err_addr = bps->address;
+	addr_in.ma.ch_inst = bps->mem_channel;
+	addr_in.ma.umc_inst = bps->mcumc_id;
+	addr_in.ma.node_inst = die_id;
+	addr_in.ma.socket_id = socket;
+
+	if (adev->umc.ras && adev->umc.ras->convert_ras_err_addr)
+		return adev->umc.ras->convert_ras_err_addr(adev, err_data,
+					&addr_in, NULL, false);
+	else
+		return  -EINVAL;
+}
+
+static int __amdgpu_ras_restore_bad_pages(struct amdgpu_device *adev,
+					struct eeprom_table_record *bps, int count)
+{
+	int j;
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct ras_err_handler_data *data = con->eh_data;
+
+	for (j = 0; j < count; j++) {
 		if (amdgpu_ras_check_bad_page_unlock(con,
-			bps[i].retired_page << AMDGPU_GPU_PAGE_SHIFT))
+			bps[j].retired_page << AMDGPU_GPU_PAGE_SHIFT)) {
+			data->count++;
+			data->space_left--;
 			continue;
+		}
 
 		if (!data->space_left &&
-			amdgpu_ras_realloc_eh_data_space(adev, data, 256)) {
-			ret = -ENOMEM;
-			goto out;
+		    amdgpu_ras_realloc_eh_data_space(adev, data, 256)) {
+			return -ENOMEM;
 		}
 
-		amdgpu_vram_mgr_reserve_range(&adev->mman.vram_mgr,
-			bps[i].retired_page << AMDGPU_GPU_PAGE_SHIFT,
-			AMDGPU_GPU_PAGE_SIZE);
+		amdgpu_ras_reserve_page(adev, bps[j].retired_page);
 
-		memcpy(&data->bps[data->count], &bps[i], sizeof(*data->bps));
+		memcpy(&data->bps[data->count], &(bps[j]),
+				sizeof(struct eeprom_table_record));
 		data->count++;
 		data->space_left--;
+		con->bad_page_num++;
+	}
+
+	return 0;
+}
+
+static int __amdgpu_ras_convert_rec_array_from_rom(struct amdgpu_device *adev,
+				struct eeprom_table_record *bps, struct ras_err_data *err_data,
+				enum amdgpu_memory_partition nps)
+{
+	int i = 0;
+	enum amdgpu_memory_partition save_nps;
+
+	save_nps = (bps[0].retired_page >> UMC_NPS_SHIFT) & UMC_NPS_MASK;
+
+	/*old asics just have pa in eeprom*/
+	if (IP_VERSION_MAJ(amdgpu_ip_version(adev, UMC_HWIP, 0)) < 12) {
+		memcpy(err_data->err_addr, bps,
+			sizeof(struct eeprom_table_record) * adev->umc.retire_unit);
+		goto out;
 	}
+
+	for (i = 0; i < adev->umc.retire_unit; i++)
+		bps[i].retired_page &= ~(UMC_NPS_MASK << UMC_NPS_SHIFT);
+
+	if (save_nps) {
+		if (save_nps == nps) {
+			if (amdgpu_umc_pages_in_a_row(adev, err_data,
+					bps[0].retired_page << AMDGPU_GPU_PAGE_SHIFT))
+				return -EINVAL;
+			for (i = 0; i < adev->umc.retire_unit; i++) {
+				err_data->err_addr[i].address = bps[0].address;
+				err_data->err_addr[i].mem_channel = bps[0].mem_channel;
+				err_data->err_addr[i].bank = bps[0].bank;
+				err_data->err_addr[i].err_type = bps[0].err_type;
+				err_data->err_addr[i].mcumc_id = bps[0].mcumc_id;
+			}
+		} else {
+			if (amdgpu_ras_mca2pa_by_idx(adev, &bps[0], err_data))
+				return -EINVAL;
+		}
+	} else {
+		if (bps[0].address == 0) {
+			/* for specific old eeprom data, mca address is not stored,
+			 * calc it from pa
+			 */
+			if (amdgpu_umc_pa2mca(adev, bps[0].retired_page << AMDGPU_GPU_PAGE_SHIFT,
+				&(bps[0].address), AMDGPU_NPS1_PARTITION_MODE))
+				return -EINVAL;
+		}
+
+		if (amdgpu_ras_mca2pa(adev, &bps[0], err_data)) {
+			if (nps == AMDGPU_NPS1_PARTITION_MODE)
+				memcpy(err_data->err_addr, bps,
+					sizeof(struct eeprom_table_record) * adev->umc.retire_unit);
+			else
+				return -EOPNOTSUPP;
+		}
+	}
+
 out:
+	return __amdgpu_ras_restore_bad_pages(adev, err_data->err_addr, adev->umc.retire_unit);
+}
+
+static int __amdgpu_ras_convert_rec_from_rom(struct amdgpu_device *adev,
+				struct eeprom_table_record *bps, struct ras_err_data *err_data,
+				enum amdgpu_memory_partition nps)
+{
+	int i = 0;
+	enum amdgpu_memory_partition save_nps;
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev)) {
+		save_nps = (bps->retired_page >> UMC_NPS_SHIFT) & UMC_NPS_MASK;
+		bps->retired_page &= ~(UMC_NPS_MASK << UMC_NPS_SHIFT);
+	} else {
+		/* if pmfw manages eeprom, save_nps is not stored on eeprom,
+		 * we should always convert mca address into physical address,
+		 * make save_nps different from nps
+		 */
+		save_nps = nps + 1;
+	}
+
+	if (save_nps == nps) {
+		if (amdgpu_umc_pages_in_a_row(adev, err_data,
+				bps->retired_page << AMDGPU_GPU_PAGE_SHIFT))
+			return -EINVAL;
+		for (i = 0; i < adev->umc.retire_unit; i++) {
+			err_data->err_addr[i].address = bps->address;
+			err_data->err_addr[i].mem_channel = bps->mem_channel;
+			err_data->err_addr[i].bank = bps->bank;
+			err_data->err_addr[i].err_type = bps->err_type;
+			err_data->err_addr[i].mcumc_id = bps->mcumc_id;
+		}
+	} else {
+		if (bps->address) {
+			if (amdgpu_ras_mca2pa_by_idx(adev, bps, err_data))
+				return -EINVAL;
+		} else {
+			/* for specific old eeprom data, mca address is not stored,
+			 * calc it from pa
+			 */
+			if (amdgpu_umc_pa2mca(adev, bps->retired_page << AMDGPU_GPU_PAGE_SHIFT,
+				&(bps->address), AMDGPU_NPS1_PARTITION_MODE))
+				return -EINVAL;
+
+			if (amdgpu_ras_mca2pa(adev, bps, err_data))
+				return -EOPNOTSUPP;
+		}
+	}
+
+	return __amdgpu_ras_restore_bad_pages(adev, err_data->err_addr,
+									adev->umc.retire_unit);
+}
+
+/* it deal with vram only. */
+int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev,
+		struct eeprom_table_record *bps, int pages, bool from_rom)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct ras_err_data err_data;
+	struct amdgpu_ras_eeprom_control *control =
+			&adev->psp.ras_context.ras->eeprom_control;
+	enum amdgpu_memory_partition nps = AMDGPU_NPS1_PARTITION_MODE;
+	int ret = 0;
+	uint32_t i = 0;
+
+	if (!con || !con->eh_data || !bps || pages <= 0)
+		return 0;
+
+	if (from_rom) {
+		err_data.err_addr =
+			kcalloc(adev->umc.retire_unit,
+				sizeof(struct eeprom_table_record), GFP_KERNEL);
+		if (!err_data.err_addr) {
+			dev_warn(adev->dev, "Failed to alloc UMC error address record in mca2pa conversion!\n");
+			return -ENOMEM;
+		}
+
+		if (adev->gmc.gmc_funcs->query_mem_partition_mode)
+			nps = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
+	}
+
+	mutex_lock(&con->recovery_lock);
+
+	if (from_rom) {
+		/* there is no pa recs in V3, so skip pa recs processing */
+		if ((control->tbl_hdr.version < RAS_TABLE_VER_V3) &&
+		    !amdgpu_ras_smu_eeprom_supported(adev)) {
+			for (i = 0; i < pages; i++) {
+				if (control->ras_num_recs - i >= adev->umc.retire_unit) {
+					if ((bps[i].address == bps[i + 1].address) &&
+						(bps[i].mem_channel == bps[i + 1].mem_channel)) {
+						/* deal with retire_unit records a time */
+						ret = __amdgpu_ras_convert_rec_array_from_rom(adev,
+										&bps[i], &err_data, nps);
+						if (ret)
+							con->bad_page_num -= adev->umc.retire_unit;
+						i += (adev->umc.retire_unit - 1);
+					} else {
+						break;
+					}
+				} else {
+					break;
+				}
+			}
+		}
+		for (; i < pages; i++) {
+			ret = __amdgpu_ras_convert_rec_from_rom(adev,
+				&bps[i], &err_data, nps);
+			if (ret)
+				con->bad_page_num -= adev->umc.retire_unit;
+		}
+
+		con->eh_data->count_saved = con->eh_data->count;
+	} else {
+		ret = __amdgpu_ras_restore_bad_pages(adev, bps, pages);
+	}
+
+	if (from_rom)
+		kfree(err_data.err_addr);
 	mutex_unlock(&con->recovery_lock);
 
 	return ret;
@@ -2093,7 +3290,7 @@ int amdgpu_ras_save_bad_pages(struct amdgpu_device *adev,
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 	struct ras_err_handler_data *data;
 	struct amdgpu_ras_eeprom_control *control;
-	int save_count;
+	int save_count, unit_num, i;
 
 	if (!con || !con->eh_data) {
 		if (new_cnt)
@@ -2102,25 +3299,53 @@ int amdgpu_ras_save_bad_pages(struct amdgpu_device *adev,
 		return 0;
 	}
 
+	if (!con->eeprom_control.is_eeprom_valid) {
+		dev_warn(adev->dev,
+			"Failed to save EEPROM table data because of EEPROM data corruption!");
+		if (new_cnt)
+			*new_cnt = 0;
+
+		return 0;
+	}
+
 	mutex_lock(&con->recovery_lock);
 	control = &con->eeprom_control;
 	data = con->eh_data;
-	save_count = data->count - control->ras_num_recs;
+	if (amdgpu_ras_smu_eeprom_supported(adev))
+		unit_num = control->ras_num_recs -
+			control->ras_num_recs_old;
+	else
+		unit_num = data->count / adev->umc.retire_unit -
+			control->ras_num_recs;
+
+	save_count = con->bad_page_num - control->ras_num_bad_pages;
 	mutex_unlock(&con->recovery_lock);
 
 	if (new_cnt)
-		*new_cnt = save_count / adev->umc.retire_unit;
+		*new_cnt = unit_num;
 
 	/* only new entries are saved */
-	if (save_count > 0) {
-		if (amdgpu_ras_eeprom_append(control,
-					     &data->bps[control->ras_num_recs],
-					     save_count)) {
-			dev_err(adev->dev, "Failed to save EEPROM table data!");
-			return -EIO;
+	if (unit_num && save_count) {
+		/*old asics only save pa to eeprom like before*/
+		if (IP_VERSION_MAJ(amdgpu_ip_version(adev, UMC_HWIP, 0)) < 12) {
+			if (amdgpu_ras_eeprom_append(control,
+					&data->bps[data->count_saved], unit_num)) {
+				dev_err(adev->dev, "Failed to save EEPROM table data!");
+				return -EIO;
+			}
+		} else {
+			for (i = 0; i < unit_num; i++) {
+				if (amdgpu_ras_eeprom_append(control,
+						&data->bps[data->count_saved +
+						i * adev->umc.retire_unit], 1)) {
+					dev_err(adev->dev, "Failed to save EEPROM table data!");
+					return -EIO;
+				}
+			}
 		}
 
 		dev_info(adev->dev, "Saved %d pages to EEPROM table.\n", save_count);
+		data->count_saved = data->count;
 	}
 
 	return 0;
@@ -2135,7 +3360,7 @@ static int amdgpu_ras_load_bad_pages(struct amdgpu_device *adev)
 	struct amdgpu_ras_eeprom_control *control =
 		&adev->psp.ras_context.ras->eeprom_control;
 	struct eeprom_table_record *bps;
-	int ret;
+	int ret, i = 0;
 
 	/* no bad page record, skip eeprom access */
 	if (control->ras_num_recs == 0 || amdgpu_bad_page_threshold == 0)
@@ -2146,27 +3371,72 @@ static int amdgpu_ras_load_bad_pages(struct amdgpu_device *adev)
 		return -ENOMEM;
 
 	ret = amdgpu_ras_eeprom_read(control, bps, control->ras_num_recs);
-	if (ret)
+	if (ret) {
 		dev_err(adev->dev, "Failed to load EEPROM table records!");
-	else
-		ret = amdgpu_ras_add_bad_pages(adev, bps, control->ras_num_recs);
+	} else {
+		if (adev->umc.ras && adev->umc.ras->convert_ras_err_addr) {
+			/*In V3, there is no pa recs, and some cases(when address==0) may be parsed
+			as pa recs, so add verion check to avoid it.
+			*/
+			if ((control->tbl_hdr.version < RAS_TABLE_VER_V3) &&
+			    !amdgpu_ras_smu_eeprom_supported(adev)) {
+				for (i = 0; i < control->ras_num_recs; i++) {
+					if ((control->ras_num_recs - i) >= adev->umc.retire_unit) {
+						if ((bps[i].address == bps[i + 1].address) &&
+							(bps[i].mem_channel == bps[i + 1].mem_channel)) {
+							control->ras_num_pa_recs += adev->umc.retire_unit;
+							i += (adev->umc.retire_unit - 1);
+						} else {
+							control->ras_num_mca_recs +=
+										(control->ras_num_recs - i);
+							break;
+						}
+					} else {
+						control->ras_num_mca_recs += (control->ras_num_recs - i);
+						break;
+					}
+				}
+			} else {
+				control->ras_num_mca_recs = control->ras_num_recs;
+			}
+		}
+
+		ret = amdgpu_ras_add_bad_pages(adev, bps, control->ras_num_recs, true);
+		if (ret)
+			goto out;
 
+		ret = amdgpu_ras_eeprom_check(control);
+		if (ret)
+			goto out;
+
+		/* HW not usable */
+		if (amdgpu_ras_is_rma(adev))
+			ret = -EHWPOISON;
+	}
+
+out:
 	kfree(bps);
 	return ret;
 }
 
-static bool amdgpu_ras_check_bad_page_unlock(struct amdgpu_ras *con,
+static int amdgpu_ras_check_bad_page_unlock(struct amdgpu_ras *con,
 				uint64_t addr)
 {
 	struct ras_err_handler_data *data = con->eh_data;
+	struct amdgpu_device *adev = con->adev;
 	int i;
 
+	if ((addr >= adev->gmc.mc_vram_size &&
+	    adev->gmc.mc_vram_size) ||
+	    (addr >= RAS_UMC_INJECT_ADDR_LIMIT))
+		return -EINVAL;
+
 	addr >>= AMDGPU_GPU_PAGE_SHIFT;
 	for (i = 0; i < data->count; i++)
 		if (addr == data->bps[i].retired_page)
-			return true;
+			return 1;
 
-	return false;
+	return 0;
 }
 
 /*
@@ -2174,11 +3444,11 @@ static bool amdgpu_ras_check_bad_page_unlock(struct amdgpu_ras *con,
  *
  * Note: this check is only for umc block
  */
-static bool amdgpu_ras_check_bad_page(struct amdgpu_device *adev,
+static int amdgpu_ras_check_bad_page(struct amdgpu_device *adev,
 				uint64_t addr)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
-	bool ret = false;
+	int ret = 0;
 
 	if (!con || !con->eh_data)
 		return ret;
@@ -2195,43 +3465,379 @@ static void amdgpu_ras_validate_threshold(struct amdgpu_device *adev,
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 
 	/*
-	 * Justification of value bad_page_cnt_threshold in ras structure
-	 *
-	 * Generally, 0 <= amdgpu_bad_page_threshold <= max record length
-	 * in eeprom or amdgpu_bad_page_threshold == -2, introduce two
-	 * scenarios accordingly.
-	 *
-	 * Bad page retirement enablement:
-	 *    - If amdgpu_bad_page_threshold = -2,
-	 *      bad_page_cnt_threshold = typical value by formula.
-	 *
-	 *    - When the value from user is 0 < amdgpu_bad_page_threshold <
-	 *      max record length in eeprom, use it directly.
-	 *
-	 * Bad page retirement disablement:
-	 *    - If amdgpu_bad_page_threshold = 0, bad page retirement
-	 *      functionality is disabled, and bad_page_cnt_threshold will
-	 *      take no effect.
+	 * amdgpu_bad_page_threshold is used to config
+	 * the threshold for the number of bad pages.
+	 * -1:  Threshold is set to default value
+	 *      Driver will issue a warning message when threshold is reached
+	 *      and continue runtime services.
+	 * 0:   Disable bad page retirement
+	 *      Driver will not retire bad pages
+	 *      which is intended for debugging purpose.
+	 * -2:  Threshold is determined by a formula
+	 *      that assumes 1 bad page per 100M of local memory.
+	 *      Driver will continue runtime services when threhold is reached.
+	 * 0 < threshold < max number of bad page records in EEPROM,
+	 *      A user-defined threshold is set
+	 *      Driver will halt runtime services when this custom threshold is reached.
 	 */
-
-	if (amdgpu_bad_page_threshold < 0) {
+	if (amdgpu_bad_page_threshold == -2) {
 		u64 val = adev->gmc.mc_vram_size;
 
 		do_div(val, RAS_BAD_PAGE_COVER);
 		con->bad_page_cnt_threshold = min(lower_32_bits(val),
 						  max_count);
+	} else if (amdgpu_bad_page_threshold == -1) {
+		con->bad_page_cnt_threshold = ((con->reserved_pages_in_bytes) >> 21) << 4;
 	} else {
 		con->bad_page_cnt_threshold = min_t(int, max_count,
 						    amdgpu_bad_page_threshold);
 	}
 }
 
-int amdgpu_ras_recovery_init(struct amdgpu_device *adev)
+int amdgpu_ras_put_poison_req(struct amdgpu_device *adev,
+		enum amdgpu_ras_block block, uint16_t pasid,
+		pasid_notify pasid_fn, void *data, uint32_t reset)
+{
+	int ret = 0;
+	struct ras_poison_msg poison_msg;
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+
+	memset(&poison_msg, 0, sizeof(poison_msg));
+	poison_msg.block = block;
+	poison_msg.pasid = pasid;
+	poison_msg.reset = reset;
+	poison_msg.pasid_fn = pasid_fn;
+	poison_msg.data = data;
+
+	ret = kfifo_put(&con->poison_fifo, poison_msg);
+	if (!ret) {
+		dev_err(adev->dev, "Poison message fifo is full!\n");
+		return -ENOSPC;
+	}
+
+	return 0;
+}
+
+static int amdgpu_ras_get_poison_req(struct amdgpu_device *adev,
+		struct ras_poison_msg *poison_msg)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+
+	return kfifo_get(&con->poison_fifo, poison_msg);
+}
+
+static void amdgpu_ras_ecc_log_init(struct ras_ecc_log_info *ecc_log)
+{
+	mutex_init(&ecc_log->lock);
+
+	INIT_RADIX_TREE(&ecc_log->de_page_tree, GFP_KERNEL);
+	ecc_log->de_queried_count = 0;
+	ecc_log->consumption_q_count = 0;
+}
+
+static void amdgpu_ras_ecc_log_fini(struct ras_ecc_log_info *ecc_log)
+{
+	struct radix_tree_iter iter;
+	void __rcu **slot;
+	struct ras_ecc_err *ecc_err;
+
+	mutex_lock(&ecc_log->lock);
+	radix_tree_for_each_slot(slot, &ecc_log->de_page_tree, &iter, 0) {
+		ecc_err = radix_tree_deref_slot(slot);
+		kfree(ecc_err->err_pages.pfn);
+		kfree(ecc_err);
+		radix_tree_iter_delete(&ecc_log->de_page_tree, &iter, slot);
+	}
+	mutex_unlock(&ecc_log->lock);
+
+	mutex_destroy(&ecc_log->lock);
+	ecc_log->de_queried_count = 0;
+	ecc_log->consumption_q_count = 0;
+}
+
+static bool amdgpu_ras_schedule_retirement_dwork(struct amdgpu_ras *con,
+				uint32_t delayed_ms)
+{
+	int ret;
+
+	mutex_lock(&con->umc_ecc_log.lock);
+	ret = radix_tree_tagged(&con->umc_ecc_log.de_page_tree,
+			UMC_ECC_NEW_DETECTED_TAG);
+	mutex_unlock(&con->umc_ecc_log.lock);
+
+	if (ret)
+		schedule_delayed_work(&con->page_retirement_dwork,
+			msecs_to_jiffies(delayed_ms));
+
+	return ret ? true : false;
+}
+
+static void amdgpu_ras_do_page_retirement(struct work_struct *work)
+{
+	struct amdgpu_ras *con = container_of(work, struct amdgpu_ras,
+					      page_retirement_dwork.work);
+	struct amdgpu_device *adev = con->adev;
+	struct ras_err_data err_data;
+
+	/* If gpu reset is ongoing, delay retiring the bad pages */
+	if (amdgpu_in_reset(adev) || amdgpu_ras_in_recovery(adev)) {
+		amdgpu_ras_schedule_retirement_dwork(con,
+				AMDGPU_RAS_RETIRE_PAGE_INTERVAL * 3);
+		return;
+	}
+
+	amdgpu_ras_error_data_init(&err_data);
+
+	amdgpu_umc_handle_bad_pages(adev, &err_data);
+
+	amdgpu_ras_error_data_fini(&err_data);
+
+	amdgpu_ras_schedule_retirement_dwork(con,
+			AMDGPU_RAS_RETIRE_PAGE_INTERVAL);
+}
+
+static int amdgpu_ras_poison_creation_handler(struct amdgpu_device *adev,
+				uint32_t poison_creation_count)
+{
+	int ret = 0;
+	struct ras_ecc_log_info *ecc_log;
+	struct ras_query_if info;
+	u32 timeout = MAX_UMC_POISON_POLLING_TIME_ASYNC;
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+	u64 de_queried_count;
+	u64 consumption_q_count;
+	enum ras_event_type type = RAS_EVENT_TYPE_POISON_CREATION;
+
+	memset(&info, 0, sizeof(info));
+	info.head.block = AMDGPU_RAS_BLOCK__UMC;
+
+	ecc_log = &ras->umc_ecc_log;
+	ecc_log->de_queried_count = 0;
+	ecc_log->consumption_q_count = 0;
+
+	do {
+		ret = amdgpu_ras_query_error_status_with_event(adev, &info, type);
+		if (ret)
+			return ret;
+
+		de_queried_count = ecc_log->de_queried_count;
+		consumption_q_count = ecc_log->consumption_q_count;
+
+		if (de_queried_count && consumption_q_count)
+			break;
+
+		msleep(100);
+	} while (--timeout);
+
+	if (de_queried_count)
+		schedule_delayed_work(&ras->page_retirement_dwork, 0);
+
+	if (amdgpu_ras_is_rma(adev) && atomic_cmpxchg(&ras->rma_in_recovery, 0, 1) == 0)
+		amdgpu_ras_reset_gpu(adev);
+
+	return 0;
+}
+
+static void amdgpu_ras_clear_poison_fifo(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct ras_poison_msg msg;
+	int ret;
+
+	do {
+		ret = kfifo_get(&con->poison_fifo, &msg);
+	} while (ret);
+}
+
+static int amdgpu_ras_poison_consumption_handler(struct amdgpu_device *adev,
+			uint32_t msg_count, uint32_t *gpu_reset)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	uint32_t reset_flags = 0, reset = 0;
+	struct ras_poison_msg msg;
+	int ret, i;
+
+	kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
+
+	for (i = 0; i < msg_count; i++) {
+		ret = amdgpu_ras_get_poison_req(adev, &msg);
+		if (!ret)
+			continue;
+
+		if (msg.pasid_fn)
+			msg.pasid_fn(adev, msg.pasid, msg.data);
+
+		reset_flags |= msg.reset;
+	}
+
+	/*
+	 * Try to ensure poison creation handler is completed first
+	 * to set rma if bad page exceed threshold.
+	 */
+	flush_delayed_work(&con->page_retirement_dwork);
+
+	/* for RMA, amdgpu_ras_poison_creation_handler will trigger gpu reset */
+	if (reset_flags && !amdgpu_ras_is_rma(adev)) {
+		if (reset_flags & AMDGPU_RAS_GPU_RESET_MODE1_RESET)
+			reset = AMDGPU_RAS_GPU_RESET_MODE1_RESET;
+		else if (reset_flags & AMDGPU_RAS_GPU_RESET_MODE2_RESET)
+			reset = AMDGPU_RAS_GPU_RESET_MODE2_RESET;
+		else
+			reset = reset_flags;
+
+		con->gpu_reset_flags |= reset;
+		amdgpu_ras_reset_gpu(adev);
+
+		*gpu_reset = reset;
+
+		/* Wait for gpu recovery to complete */
+		flush_work(&con->recovery_work);
+	}
+
+	return 0;
+}
+
+static int amdgpu_ras_page_retirement_thread(void *param)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)param;
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	uint32_t poison_creation_count, msg_count;
+	uint32_t gpu_reset;
+	int ret;
+
+	while (!kthread_should_stop()) {
+
+		wait_event_interruptible(con->page_retirement_wq,
+				kthread_should_stop() ||
+				atomic_read(&con->page_retirement_req_cnt));
+
+		if (kthread_should_stop())
+			break;
+
+		mutex_lock(&con->poison_lock);
+		gpu_reset = 0;
+
+		do {
+			poison_creation_count = atomic_read(&con->poison_creation_count);
+			ret = amdgpu_ras_poison_creation_handler(adev, poison_creation_count);
+			if (ret == -EIO)
+				break;
+
+			if (poison_creation_count) {
+				atomic_sub(poison_creation_count, &con->poison_creation_count);
+				atomic_sub(poison_creation_count, &con->page_retirement_req_cnt);
+			}
+		} while (atomic_read(&con->poison_creation_count) &&
+			!atomic_read(&con->poison_consumption_count));
+
+		if (ret != -EIO) {
+			msg_count = kfifo_len(&con->poison_fifo);
+			if (msg_count) {
+				ret = amdgpu_ras_poison_consumption_handler(adev,
+						msg_count, &gpu_reset);
+				if ((ret != -EIO) &&
+				    (gpu_reset != AMDGPU_RAS_GPU_RESET_MODE1_RESET))
+					atomic_sub(msg_count, &con->page_retirement_req_cnt);
+			}
+		}
+
+		if ((ret == -EIO) || (gpu_reset == AMDGPU_RAS_GPU_RESET_MODE1_RESET)) {
+			/* gpu mode-1 reset is ongoing or just completed ras mode-1 reset */
+			/* Clear poison creation request */
+			atomic_set(&con->poison_creation_count, 0);
+			atomic_set(&con->poison_consumption_count, 0);
+
+			/* Clear poison fifo */
+			amdgpu_ras_clear_poison_fifo(adev);
+
+			/* Clear all poison requests */
+			atomic_set(&con->page_retirement_req_cnt, 0);
+
+			if (ret == -EIO) {
+				/* Wait for mode-1 reset to complete */
+				down_read(&adev->reset_domain->sem);
+				up_read(&adev->reset_domain->sem);
+			}
+
+			/* Wake up work to save bad pages to eeprom */
+			schedule_delayed_work(&con->page_retirement_dwork, 0);
+		} else if (gpu_reset) {
+			/* gpu just completed mode-2 reset or other reset */
+			/* Clear poison consumption messages cached in fifo */
+			msg_count = kfifo_len(&con->poison_fifo);
+			if (msg_count) {
+				amdgpu_ras_clear_poison_fifo(adev);
+				atomic_sub(msg_count, &con->page_retirement_req_cnt);
+			}
+
+			atomic_set(&con->poison_consumption_count, 0);
+
+			/* Wake up work to save bad pages to eeprom */
+			schedule_delayed_work(&con->page_retirement_dwork, 0);
+		}
+		mutex_unlock(&con->poison_lock);
+	}
+
+	return 0;
+}
+
+int amdgpu_ras_init_badpage_info(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct amdgpu_ras_eeprom_control *control;
+	int ret;
+
+	if (!con || amdgpu_sriov_vf(adev))
+		return 0;
+
+	if (amdgpu_uniras_enabled(adev))
+		return 0;
+
+	control = &con->eeprom_control;
+	con->ras_smu_drv = amdgpu_dpm_get_ras_smu_driver(adev);
+
+	ret = amdgpu_ras_eeprom_init(control);
+	control->is_eeprom_valid = !ret;
+
+	if (!adev->umc.ras || !adev->umc.ras->convert_ras_err_addr)
+		control->ras_num_pa_recs = control->ras_num_recs;
+
+	if (adev->umc.ras &&
+	    adev->umc.ras->get_retire_flip_bits)
+		adev->umc.ras->get_retire_flip_bits(adev);
+
+	if (control->ras_num_recs && control->is_eeprom_valid) {
+		ret = amdgpu_ras_load_bad_pages(adev);
+		if (ret) {
+			control->is_eeprom_valid = false;
+			return 0;
+		}
+
+		amdgpu_dpm_send_hbm_bad_pages_num(
+			adev, control->ras_num_bad_pages);
+
+		if (con->update_channel_flag == true) {
+			amdgpu_dpm_send_hbm_bad_channel_flag(
+				adev, control->bad_channel_bitmap);
+			con->update_channel_flag = false;
+		}
+
+		/* The format action is only applied to new ASICs */
+		if (IP_VERSION_MAJ(amdgpu_ip_version(adev, UMC_HWIP, 0)) >= 12 &&
+		    control->tbl_hdr.version < RAS_TABLE_VER_V3)
+			if (!amdgpu_ras_eeprom_reset_table(control))
+				if (amdgpu_ras_save_bad_pages(adev, NULL))
+					dev_warn(adev->dev, "Failed to format RAS EEPROM data in V3 version!\n");
+	}
+
+	return 0;
+}
+
+int amdgpu_ras_recovery_init(struct amdgpu_device *adev, bool init_bp_info)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 	struct ras_err_handler_data **data;
 	u32  max_eeprom_records_count = 0;
-	bool exc_err_limit = false;
 	int ret;
 
 	if (!con || amdgpu_sriov_vf(adev))
@@ -2248,47 +3854,44 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev)
 		return 0;
 
 	data = &con->eh_data;
-	*data = kmalloc(sizeof(**data), GFP_KERNEL | __GFP_ZERO);
+	*data = kzalloc(sizeof(**data), GFP_KERNEL);
 	if (!*data) {
 		ret = -ENOMEM;
 		goto out;
 	}
 
 	mutex_init(&con->recovery_lock);
+	mutex_init(&con->poison_lock);
 	INIT_WORK(&con->recovery_work, amdgpu_ras_do_recovery);
 	atomic_set(&con->in_recovery, 0);
+	atomic_set(&con->rma_in_recovery, 0);
 	con->eeprom_control.bad_channel_bitmap = 0;
 
-	max_eeprom_records_count = amdgpu_ras_eeprom_max_record_count();
+	max_eeprom_records_count = amdgpu_ras_eeprom_max_record_count(&con->eeprom_control);
 	amdgpu_ras_validate_threshold(adev, max_eeprom_records_count);
 
-	/* Todo: During test the SMU might fail to read the eeprom through I2C
-	 * when the GPU is pending on XGMI reset during probe time
-	 * (Mostly after second bus reset), skip it now
-	 */
-	if (adev->gmc.xgmi.pending_reset)
-		return 0;
-	ret = amdgpu_ras_eeprom_init(&con->eeprom_control, &exc_err_limit);
-	/*
-	 * This calling fails when exc_err_limit is true or
-	 * ret != 0.
-	 */
-	if (exc_err_limit || ret)
-		goto free;
-
-	if (con->eeprom_control.ras_num_recs) {
-		ret = amdgpu_ras_load_bad_pages(adev);
+	if (init_bp_info) {
+		ret = amdgpu_ras_init_badpage_info(adev);
 		if (ret)
 			goto free;
+	}
 
-		amdgpu_dpm_send_hbm_bad_pages_num(adev, con->eeprom_control.ras_num_recs);
-
-		if (con->update_channel_flag == true) {
-			amdgpu_dpm_send_hbm_bad_channel_flag(adev, con->eeprom_control.bad_channel_bitmap);
-			con->update_channel_flag = false;
-		}
+	mutex_init(&con->page_rsv_lock);
+	INIT_KFIFO(con->poison_fifo);
+	mutex_init(&con->page_retirement_lock);
+	init_waitqueue_head(&con->page_retirement_wq);
+	atomic_set(&con->page_retirement_req_cnt, 0);
+	atomic_set(&con->poison_creation_count, 0);
+	atomic_set(&con->poison_consumption_count, 0);
+	con->page_retirement_thread =
+		kthread_run(amdgpu_ras_page_retirement_thread, adev, "umc_page_retirement");
+	if (IS_ERR(con->page_retirement_thread)) {
+		con->page_retirement_thread = NULL;
+		dev_warn(adev->dev, "Failed to create umc_page_retirement thread!!!\n");
 	}
 
+	INIT_DELAYED_WORK(&con->page_retirement_dwork, amdgpu_ras_do_page_retirement);
+	amdgpu_ras_ecc_log_init(&con->umc_ecc_log);
 #ifdef CONFIG_X86_MCE_AMD
 	if ((adev->asic_type == CHIP_ALDEBARAN) &&
 	    (adev->gmc.xgmi.connected_to_cpu))
@@ -2307,7 +3910,7 @@ out:
 	 * Except error threshold exceeding case, other failure cases in this
 	 * function would not fail amdgpu driver init.
 	 */
-	if (!exc_err_limit)
+	if (!amdgpu_ras_is_rma(adev))
 		ret = 0;
 	else
 		ret = -EINVAL;
@@ -2319,19 +3922,43 @@ static int amdgpu_ras_recovery_fini(struct amdgpu_device *adev)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 	struct ras_err_handler_data *data = con->eh_data;
+	int max_flush_timeout = MAX_FLUSH_RETIRE_DWORK_TIMES;
+	bool ret;
 
 	/* recovery_init failed to init it, fini is useless */
 	if (!data)
 		return 0;
 
+	/* Save all cached bad pages to eeprom */
+	do {
+		flush_delayed_work(&con->page_retirement_dwork);
+		ret = amdgpu_ras_schedule_retirement_dwork(con, 0);
+	} while (ret && max_flush_timeout--);
+
+	if (con->page_retirement_thread)
+		kthread_stop(con->page_retirement_thread);
+
+	atomic_set(&con->page_retirement_req_cnt, 0);
+	atomic_set(&con->poison_creation_count, 0);
+
+	mutex_destroy(&con->page_rsv_lock);
+
 	cancel_work_sync(&con->recovery_work);
 
+	cancel_delayed_work_sync(&con->page_retirement_dwork);
+
+	amdgpu_ras_ecc_log_fini(&con->umc_ecc_log);
+
 	mutex_lock(&con->recovery_lock);
 	con->eh_data = NULL;
 	kfree(data->bps);
 	kfree(data);
 	mutex_unlock(&con->recovery_lock);
 
+	amdgpu_ras_critical_region_init(adev);
+#ifdef CONFIG_X86_MCE_AMD
+	amdgpu_unregister_bad_pages_mca_notifier(adev);
+#endif
 	return 0;
 }
 /* recovery end */
@@ -2339,8 +3966,11 @@ static int amdgpu_ras_recovery_fini(struct amdgpu_device *adev)
 static bool amdgpu_ras_asic_supported(struct amdgpu_device *adev)
 {
 	if (amdgpu_sriov_vf(adev)) {
-		switch (adev->ip_versions[MP0_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 		case IP_VERSION(13, 0, 2):
+		case IP_VERSION(13, 0, 6):
+		case IP_VERSION(13, 0, 12):
+		case IP_VERSION(13, 0, 14):
 			return true;
 		default:
 			return false;
@@ -2348,9 +3978,13 @@ static bool amdgpu_ras_asic_supported(struct amdgpu_device *adev)
 	}
 
 	if (adev->asic_type == CHIP_IP_DISCOVERY) {
-		switch (adev->ip_versions[MP0_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 		case IP_VERSION(13, 0, 0):
+		case IP_VERSION(13, 0, 6):
 		case IP_VERSION(13, 0, 10):
+		case IP_VERSION(13, 0, 12):
+		case IP_VERSION(13, 0, 14):
+		case IP_VERSION(14, 0, 3):
 			return true;
 		default:
 			return false;
@@ -2376,13 +4010,95 @@ static void amdgpu_ras_get_quirks(struct amdgpu_device *adev)
 	if (!ctx)
 		return;
 
-	if (strnstr(ctx->vbios_version, "D16406",
-		    sizeof(ctx->vbios_version)) ||
-		strnstr(ctx->vbios_version, "D36002",
-			sizeof(ctx->vbios_version)))
+	if (strnstr(ctx->vbios_pn, "D16406",
+		    sizeof(ctx->vbios_pn)) ||
+		strnstr(ctx->vbios_pn, "D36002",
+			sizeof(ctx->vbios_pn)))
 		adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__GFX);
 }
 
+/* Query ras capablity via atomfirmware interface */
+static void amdgpu_ras_query_ras_capablity_from_vbios(struct amdgpu_device *adev)
+{
+	/* mem_ecc cap */
+	if (amdgpu_atomfirmware_mem_ecc_supported(adev)) {
+		dev_info(adev->dev, "MEM ECC is active.\n");
+		adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__UMC |
+					 1 << AMDGPU_RAS_BLOCK__DF);
+	} else {
+		dev_info(adev->dev, "MEM ECC is not presented.\n");
+	}
+
+	/* sram_ecc cap */
+	if (amdgpu_atomfirmware_sram_ecc_supported(adev)) {
+		dev_info(adev->dev, "SRAM ECC is active.\n");
+		if (!amdgpu_sriov_vf(adev))
+			adev->ras_hw_enabled |= ~(1 << AMDGPU_RAS_BLOCK__UMC |
+						  1 << AMDGPU_RAS_BLOCK__DF);
+		else
+			adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__PCIE_BIF |
+						 1 << AMDGPU_RAS_BLOCK__SDMA |
+						 1 << AMDGPU_RAS_BLOCK__GFX);
+
+		/*
+		 * VCN/JPEG RAS can be supported on both bare metal and
+		 * SRIOV environment
+		 */
+		if (amdgpu_ip_version(adev, VCN_HWIP, 0) == IP_VERSION(2, 6, 0) ||
+		    amdgpu_ip_version(adev, VCN_HWIP, 0) == IP_VERSION(4, 0, 0) ||
+		    amdgpu_ip_version(adev, VCN_HWIP, 0) == IP_VERSION(4, 0, 3) ||
+		    amdgpu_ip_version(adev, VCN_HWIP, 0) == IP_VERSION(5, 0, 1))
+			adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__VCN |
+						 1 << AMDGPU_RAS_BLOCK__JPEG);
+		else
+			adev->ras_hw_enabled &= ~(1 << AMDGPU_RAS_BLOCK__VCN |
+						  1 << AMDGPU_RAS_BLOCK__JPEG);
+
+		/*
+		 * XGMI RAS is not supported if xgmi num physical nodes
+		 * is zero
+		 */
+		if (!adev->gmc.xgmi.num_physical_nodes)
+			adev->ras_hw_enabled &= ~(1 << AMDGPU_RAS_BLOCK__XGMI_WAFL);
+	} else {
+		dev_info(adev->dev, "SRAM ECC is not presented.\n");
+	}
+}
+
+/* Query poison mode from umc/df IP callbacks */
+static void amdgpu_ras_query_poison_mode(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	bool df_poison, umc_poison;
+
+	/* poison setting is useless on SRIOV guest */
+	if (amdgpu_sriov_vf(adev) || !con)
+		return;
+
+	/* Init poison supported flag, the default value is false */
+	if (adev->gmc.xgmi.connected_to_cpu ||
+	    adev->gmc.is_app_apu) {
+		/* enabled by default when GPU is connected to CPU */
+		con->poison_supported = true;
+	} else if (adev->df.funcs &&
+	    adev->df.funcs->query_ras_poison_mode &&
+	    adev->umc.ras &&
+	    adev->umc.ras->query_ras_poison_mode) {
+		df_poison =
+			adev->df.funcs->query_ras_poison_mode(adev);
+		umc_poison =
+			adev->umc.ras->query_ras_poison_mode(adev);
+
+		/* Only poison is set in both DF and UMC, we can support it */
+		if (df_poison && umc_poison)
+			con->poison_supported = true;
+		else if (df_poison != umc_poison)
+			dev_warn(adev->dev,
+				"Poison setting is inconsistent in DF/UMC(%d:%d)!\n",
+				df_poison, umc_poison);
+	}
+}
+
 /*
  * check hardware's ras ability which will be saved in hw_supported.
  * if hardware does not support ras, we can skip some ras initializtion and
@@ -2396,49 +4112,21 @@ static void amdgpu_ras_check_supported(struct amdgpu_device *adev)
 {
 	adev->ras_hw_enabled = adev->ras_enabled = 0;
 
-	if (!adev->is_atom_fw ||
-	    !amdgpu_ras_asic_supported(adev))
+	if (!amdgpu_ras_asic_supported(adev))
 		return;
 
-	if (!adev->gmc.xgmi.connected_to_cpu) {
-		if (amdgpu_atomfirmware_mem_ecc_supported(adev)) {
-			dev_info(adev->dev, "MEM ECC is active.\n");
-			adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__UMC |
-						   1 << AMDGPU_RAS_BLOCK__DF);
-		} else {
-			dev_info(adev->dev, "MEM ECC is not presented.\n");
-		}
-
-		if (amdgpu_atomfirmware_sram_ecc_supported(adev)) {
-			dev_info(adev->dev, "SRAM ECC is active.\n");
-			if (!amdgpu_sriov_vf(adev))
-				adev->ras_hw_enabled |= ~(1 << AMDGPU_RAS_BLOCK__UMC |
-							    1 << AMDGPU_RAS_BLOCK__DF);
-			else
-				adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__PCIE_BIF |
-								1 << AMDGPU_RAS_BLOCK__SDMA |
-								1 << AMDGPU_RAS_BLOCK__GFX);
+	if (amdgpu_sriov_vf(adev)) {
+		if (amdgpu_virt_get_ras_capability(adev))
+			goto init_ras_enabled_flag;
+	}
 
-			/* VCN/JPEG RAS can be supported on both bare metal and
-			 * SRIOV environment
-			 */
-			if (adev->ip_versions[VCN_HWIP][0] == IP_VERSION(2, 6, 0) ||
-			    adev->ip_versions[VCN_HWIP][0] == IP_VERSION(4, 0, 0))
-				adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__VCN |
-							1 << AMDGPU_RAS_BLOCK__JPEG);
-			else
-				adev->ras_hw_enabled &= ~(1 << AMDGPU_RAS_BLOCK__VCN |
-							1 << AMDGPU_RAS_BLOCK__JPEG);
+	/* query ras capability from psp */
+	if (amdgpu_psp_get_ras_capability(&adev->psp))
+		goto init_ras_enabled_flag;
 
-			/*
-			 * XGMI RAS is not supported if xgmi num physical nodes
-			 * is zero
-			 */
-			if (!adev->gmc.xgmi.num_physical_nodes)
-				adev->ras_hw_enabled &= ~(1 << AMDGPU_RAS_BLOCK__XGMI_WAFL);
-		} else {
-			dev_info(adev->dev, "SRAM ECC is not presented.\n");
-		}
+	/* query ras capablity from bios */
+	if (!adev->gmc.xgmi.connected_to_cpu && !adev->gmc.is_app_apu) {
+		amdgpu_ras_query_ras_capablity_from_vbios(adev);
 	} else {
 		/* driver only manages a few IP blocks RAS feature
 		 * when GPU is connected cpu through XGMI */
@@ -2447,13 +4135,31 @@ static void amdgpu_ras_check_supported(struct amdgpu_device *adev)
 					   1 << AMDGPU_RAS_BLOCK__MMHUB);
 	}
 
+	/* apply asic specific settings (vega20 only for now) */
 	amdgpu_ras_get_quirks(adev);
 
+	/* query poison mode from umc/df ip callback */
+	amdgpu_ras_query_poison_mode(adev);
+
+init_ras_enabled_flag:
 	/* hw_supported needs to be aligned with RAS block mask. */
 	adev->ras_hw_enabled &= AMDGPU_RAS_BLOCK_MASK;
 
 	adev->ras_enabled = amdgpu_ras_enable == 0 ? 0 :
 		adev->ras_hw_enabled & amdgpu_ras_mask;
+
+	/* aca is disabled by default except for psp v13_0_6/v13_0_12/v13_0_14 */
+	if (!amdgpu_sriov_vf(adev)) {
+		adev->aca.is_enabled =
+			(amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 6) ||
+			amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 12) ||
+			amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 14));
+	}
+
+	/* bad page feature is not applicable to specific app platform */
+	if (adev->gmc.is_app_apu &&
+	    amdgpu_ip_version(adev, UMC_HWIP, 0) == IP_VERSION(12, 0, 0))
+		amdgpu_bad_page_threshold = 0;
 }
 
 static void amdgpu_ras_counte_dw(struct work_struct *work)
@@ -2476,40 +4182,72 @@ static void amdgpu_ras_counte_dw(struct work_struct *work)
 		atomic_set(&con->ras_ue_count, ue_count);
 	}
 
-	pm_runtime_mark_last_busy(dev->dev);
 Out:
 	pm_runtime_put_autosuspend(dev->dev);
 }
 
-static void amdgpu_ras_query_poison_mode(struct amdgpu_device *adev)
+static int amdgpu_get_ras_schema(struct amdgpu_device *adev)
 {
-	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
-	bool df_poison, umc_poison;
+	return  amdgpu_ras_is_poison_mode_supported(adev) ? AMDGPU_RAS_ERROR__POISON : 0 |
+			AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE |
+			AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE |
+			AMDGPU_RAS_ERROR__PARITY;
+}
 
-	/* poison setting is useless on SRIOV guest */
-	if (amdgpu_sriov_vf(adev) || !con)
+static void ras_event_mgr_init(struct ras_event_manager *mgr)
+{
+	struct ras_event_state *event_state;
+	int i;
+
+	memset(mgr, 0, sizeof(*mgr));
+	atomic64_set(&mgr->seqno, 0);
+
+	for (i = 0; i < ARRAY_SIZE(mgr->event_state); i++) {
+		event_state = &mgr->event_state[i];
+		event_state->last_seqno = RAS_EVENT_INVALID_ID;
+		atomic64_set(&event_state->count, 0);
+	}
+}
+
+static void amdgpu_ras_event_mgr_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+	struct amdgpu_hive_info *hive;
+
+	if (!ras)
 		return;
 
-	/* Init poison supported flag, the default value is false */
-	if (adev->gmc.xgmi.connected_to_cpu) {
-		/* enabled by default when GPU is connected to CPU */
-		con->poison_supported = true;
-	} else if (adev->df.funcs &&
-	    adev->df.funcs->query_ras_poison_mode &&
-	    adev->umc.ras &&
-	    adev->umc.ras->query_ras_poison_mode) {
-		df_poison =
-			adev->df.funcs->query_ras_poison_mode(adev);
-		umc_poison =
-			adev->umc.ras->query_ras_poison_mode(adev);
+	hive = amdgpu_get_xgmi_hive(adev);
+	ras->event_mgr = hive ? &hive->event_mgr : &ras->__event_mgr;
 
-		/* Only poison is set in both DF and UMC, we can support it */
-		if (df_poison && umc_poison)
-			con->poison_supported = true;
-		else if (df_poison != umc_poison)
-			dev_warn(adev->dev,
-				"Poison setting is inconsistent in DF/UMC(%d:%d)!\n",
-				df_poison, umc_poison);
+	/* init event manager with node 0 on xgmi system */
+	if (!amdgpu_reset_in_recovery(adev)) {
+		if (!hive || adev->gmc.xgmi.node_id == 0)
+			ras_event_mgr_init(ras->event_mgr);
+	}
+
+	if (hive)
+		amdgpu_put_xgmi_hive(hive);
+}
+
+static void amdgpu_ras_init_reserved_vram_size(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+
+	if (!con || (adev->flags & AMD_IS_APU))
+		return;
+
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
+	case IP_VERSION(13, 0, 2):
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 12):
+		con->reserved_pages_in_bytes = AMDGPU_RAS_RESERVED_VRAM_SIZE_DEFAULT;
+		break;
+	case IP_VERSION(13, 0, 14):
+		con->reserved_pages_in_bytes = (AMDGPU_RAS_RESERVED_VRAM_SIZE_DEFAULT << 1);
+		break;
+	default:
+		break;
 	}
 }
 
@@ -2521,10 +4259,10 @@ int amdgpu_ras_init(struct amdgpu_device *adev)
 	if (con)
 		return 0;
 
-	con = kmalloc(sizeof(struct amdgpu_ras) +
+	con = kzalloc(sizeof(*con) +
 			sizeof(struct ras_manager) * AMDGPU_RAS_BLOCK_COUNT +
 			sizeof(struct ras_manager) * AMDGPU_RAS_MCA_BLOCK_COUNT,
-			GFP_KERNEL|__GFP_ZERO);
+			GFP_KERNEL);
 	if (!con)
 		return -ENOMEM;
 
@@ -2555,6 +4293,7 @@ int amdgpu_ras_init(struct amdgpu_device *adev)
 
 	con->update_channel_flag = false;
 	con->features = 0;
+	con->schema = 0;
 	INIT_LIST_HEAD(&con->head);
 	/* Might need get this flag from vbios. */
 	con->flags = RAS_DEFAULT_FLAGS;
@@ -2562,7 +4301,7 @@ int amdgpu_ras_init(struct amdgpu_device *adev)
 	/* initialize nbio ras function ahead of any other
 	 * ras functions so hardware fatal error interrupt
 	 * can be enabled as early as possible */
-	switch (adev->ip_versions[NBIO_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, NBIO_HWIP, 0)) {
 	case IP_VERSION(7, 4, 0):
 	case IP_VERSION(7, 4, 1):
 	case IP_VERSION(7, 4, 4):
@@ -2579,6 +4318,22 @@ int amdgpu_ras_init(struct amdgpu_device *adev)
 			 * check DF RAS */
 			adev->nbio.ras = &nbio_v4_3_ras;
 		break;
+	case IP_VERSION(6, 3, 1):
+		if (adev->ras_hw_enabled & (1 << AMDGPU_RAS_BLOCK__DF))
+			/* unlike other generation of nbio ras,
+			 * nbif v6_3_1 only support fatal error interrupt
+			 * to inform software that DF is freezed due to
+			 * system fatal error event. driver should not
+			 * enable nbio ras in such case. Instead,
+			 * check DF RAS
+			 */
+			adev->nbio.ras = &nbif_v6_3_1_ras;
+		break;
+	case IP_VERSION(7, 9, 0):
+	case IP_VERSION(7, 9, 1):
+		if (!adev->gmc.is_app_apu)
+			adev->nbio.ras = &nbio_v7_9_ras;
+		break;
 	default:
 		/* nbio ras is not available */
 		break;
@@ -2604,13 +4359,37 @@ int amdgpu_ras_init(struct amdgpu_device *adev)
 			goto release_con;
 	}
 
-	amdgpu_ras_query_poison_mode(adev);
+	/* Packed socket_id to ras feature mask bits[31:29] */
+	if (adev->smuio.funcs &&
+	    adev->smuio.funcs->get_socket_id)
+		con->features |= ((adev->smuio.funcs->get_socket_id(adev)) <<
+					AMDGPU_RAS_FEATURES_SOCKETID_SHIFT);
+
+	/* Get RAS schema for particular SOC */
+	con->schema = amdgpu_get_ras_schema(adev);
+
+	amdgpu_ras_init_reserved_vram_size(adev);
 
 	if (amdgpu_ras_fs_init(adev)) {
 		r = -EINVAL;
 		goto release_con;
 	}
 
+	if (amdgpu_ras_aca_is_supported(adev)) {
+		if (amdgpu_aca_is_enabled(adev))
+			r = amdgpu_aca_init(adev);
+		else
+			r = amdgpu_mca_init(adev);
+		if (r)
+			goto release_con;
+	}
+
+	con->init_task_pid = task_pid_nr(current);
+	get_task_comm(con->init_task_comm, current);
+
+	mutex_init(&con->critical_region_lock);
+	INIT_LIST_HEAD(&con->critical_region_head);
+
 	dev_info(adev->dev, "RAS INFO: ras initialized successfully, "
 		 "hardware ability[%x] ras_mask[%x]\n",
 		 adev->ras_hw_enabled, adev->ras_enabled);
@@ -2625,7 +4404,8 @@ release_con:
 
 int amdgpu_persistent_edc_harvesting_supported(struct amdgpu_device *adev)
 {
-	if (adev->gmc.xgmi.connected_to_cpu)
+	if (adev->gmc.xgmi.connected_to_cpu ||
+	    adev->gmc.is_app_apu)
 		return 1;
 	return 0;
 }
@@ -2677,7 +4457,7 @@ int amdgpu_ras_block_late_init(struct amdgpu_device *adev,
 
 	r = amdgpu_ras_feature_enable_on_boot(adev, ras_block, 1);
 	if (r) {
-		if (adev->in_suspend || amdgpu_in_reset(adev)) {
+		if (adev->in_suspend || amdgpu_reset_in_recovery(adev)) {
 			/* in resume phase, if fail to enable ras,
 			 * clean up all ras fs nodes, and disable ras */
 			goto cleanup;
@@ -2689,7 +4469,7 @@ int amdgpu_ras_block_late_init(struct amdgpu_device *adev,
 	amdgpu_persistent_edc_harvesting(adev, ras_block);
 
 	/* in resume phase, no need to create ras fs node */
-	if (adev->in_suspend || amdgpu_in_reset(adev))
+	if (adev->in_suspend || amdgpu_reset_in_recovery(adev))
 		return 0;
 
 	ras_obj = container_of(ras_block, struct amdgpu_ras_block_object, ras_comm);
@@ -2701,23 +4481,28 @@ int amdgpu_ras_block_late_init(struct amdgpu_device *adev,
 			goto cleanup;
 	}
 
-	r = amdgpu_ras_sysfs_create(adev, ras_block);
-	if (r)
-		goto interrupt;
+	if (ras_obj->hw_ops &&
+	    (ras_obj->hw_ops->query_ras_error_count ||
+	     ras_obj->hw_ops->query_ras_error_status)) {
+		r = amdgpu_ras_sysfs_create(adev, ras_block);
+		if (r)
+			goto interrupt;
 
-	/* Those are the cached values at init.
-	 */
-	query_info = kzalloc(sizeof(struct ras_query_if), GFP_KERNEL);
-	if (!query_info)
-		return -ENOMEM;
-	memcpy(&query_info->head, ras_block, sizeof(struct ras_common_if));
+		/* Those are the cached values at init.
+		 */
+		query_info = kzalloc(sizeof(*query_info), GFP_KERNEL);
+		if (!query_info)
+			return -ENOMEM;
+		memcpy(&query_info->head, ras_block, sizeof(struct ras_common_if));
 
-	if (amdgpu_ras_query_error_count(adev, &ce_count, &ue_count, query_info) == 0) {
-		atomic_set(&con->ras_ce_count, ce_count);
-		atomic_set(&con->ras_ue_count, ue_count);
+		if (amdgpu_ras_query_error_count(adev, &ce_count, &ue_count, query_info) == 0) {
+			atomic_set(&con->ras_ce_count, ce_count);
+			atomic_set(&con->ras_ue_count, ue_count);
+		}
+
+		kfree(query_info);
 	}
 
-	kfree(query_info);
 	return 0;
 
 interrupt:
@@ -2801,7 +4586,7 @@ void amdgpu_ras_suspend(struct amdgpu_device *adev)
 
 	amdgpu_ras_disable_all_features(adev, 0);
 	/* Make sure all ras objects are disabled. */
-	if (con->features)
+	if (AMDGPU_RAS_GET_FEATURES(con->features))
 		amdgpu_ras_disable_all_features(adev, 1);
 }
 
@@ -2811,17 +4596,40 @@ int amdgpu_ras_late_init(struct amdgpu_device *adev)
 	struct amdgpu_ras_block_object *obj;
 	int r;
 
+	amdgpu_ras_event_mgr_init(adev);
+
+	if (amdgpu_ras_aca_is_supported(adev)) {
+		if (amdgpu_reset_in_recovery(adev)) {
+			if (amdgpu_aca_is_enabled(adev))
+				r = amdgpu_aca_reset(adev);
+			else
+				r = amdgpu_mca_reset(adev);
+			if (r)
+				return r;
+		}
+
+		if (!amdgpu_sriov_vf(adev)) {
+			if (amdgpu_aca_is_enabled(adev))
+				amdgpu_ras_set_aca_debug_mode(adev, false);
+			else
+				amdgpu_ras_set_mca_debug_mode(adev, false);
+		}
+	}
+
 	/* Guest side doesn't need init ras feature */
-	if (amdgpu_sriov_vf(adev))
+	if (amdgpu_sriov_vf(adev) && !amdgpu_sriov_ras_telemetry_en(adev))
 		return 0;
 
 	list_for_each_entry_safe(node, tmp, &adev->ras_list, node) {
-		if (!node->ras_obj) {
+		obj = node->ras_obj;
+		if (!obj) {
 			dev_warn(adev->dev, "Warning: abnormal ras list node.\n");
 			continue;
 		}
 
-		obj = node->ras_obj;
+		if (!amdgpu_ras_is_supported(adev, obj->ras_comm.block))
+			continue;
+
 		if (obj->ras_late_init) {
 			r = obj->ras_late_init(adev, &obj->ras_comm);
 			if (r) {
@@ -2846,7 +4654,7 @@ int amdgpu_ras_pre_fini(struct amdgpu_device *adev)
 
 
 	/* Need disable ras on all IPs here before ip [hw/sw]fini */
-	if (con->features)
+	if (AMDGPU_RAS_GET_FEATURES(con->features))
 		amdgpu_ras_disable_all_features(adev, 0);
 	amdgpu_ras_recovery_fini(adev);
 	return 0;
@@ -2861,6 +4669,9 @@ int amdgpu_ras_fini(struct amdgpu_device *adev)
 	if (!adev->ras_enabled || !con)
 		return 0;
 
+	amdgpu_ras_critical_region_fini(adev);
+	mutex_destroy(&con->critical_region_lock);
+
 	list_for_each_entry_safe(ras_node, tmp, &adev->ras_list, node) {
 		if (ras_node->ras_obj) {
 			obj = ras_node->ras_obj;
@@ -2879,10 +4690,17 @@ int amdgpu_ras_fini(struct amdgpu_device *adev)
 	amdgpu_ras_fs_fini(adev);
 	amdgpu_ras_interrupt_remove_all(adev);
 
-	WARN(con->features, "Feature mask is not cleared");
+	if (amdgpu_ras_aca_is_supported(adev)) {
+		if (amdgpu_aca_is_enabled(adev))
+			amdgpu_aca_fini(adev);
+		else
+			amdgpu_mca_fini(adev);
+	}
 
-	if (con->features)
-		amdgpu_ras_disable_all_features(adev, 1);
+	WARN(AMDGPU_RAS_GET_FEATURES(con->features), "Feature mask is not cleared");
+
+	if (AMDGPU_RAS_GET_FEATURES(con->features))
+		amdgpu_ras_disable_all_features(adev, 0);
 
 	cancel_delayed_work_sync(&con->ras_counte_delay_work);
 
@@ -2892,18 +4710,161 @@ int amdgpu_ras_fini(struct amdgpu_device *adev)
 	return 0;
 }
 
-void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev)
+bool amdgpu_ras_get_fed_status(struct amdgpu_device *adev)
 {
-	amdgpu_ras_check_supported(adev);
-	if (!adev->ras_hw_enabled)
-		return;
+	struct amdgpu_ras *ras;
+
+	ras = amdgpu_ras_get_context(adev);
+	if (!ras)
+		return false;
+
+	return test_bit(AMDGPU_RAS_BLOCK__LAST, &ras->ras_err_state);
+}
+
+void amdgpu_ras_set_fed(struct amdgpu_device *adev, bool status)
+{
+	struct amdgpu_ras *ras;
+
+	ras = amdgpu_ras_get_context(adev);
+	if (ras) {
+		if (status)
+			set_bit(AMDGPU_RAS_BLOCK__LAST, &ras->ras_err_state);
+		else
+			clear_bit(AMDGPU_RAS_BLOCK__LAST, &ras->ras_err_state);
+	}
+}
+
+void amdgpu_ras_clear_err_state(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *ras;
+
+	ras = amdgpu_ras_get_context(adev);
+	if (ras) {
+		ras->ras_err_state = 0;
+		ras->gpu_reset_flags = 0;
+	}
+}
+
+void amdgpu_ras_set_err_poison(struct amdgpu_device *adev,
+			       enum amdgpu_ras_block block)
+{
+	struct amdgpu_ras *ras;
 
+	ras = amdgpu_ras_get_context(adev);
+	if (ras)
+		set_bit(block, &ras->ras_err_state);
+}
+
+bool amdgpu_ras_is_err_state(struct amdgpu_device *adev, int block)
+{
+	struct amdgpu_ras *ras;
+
+	ras = amdgpu_ras_get_context(adev);
+	if (ras) {
+		if (block == AMDGPU_RAS_BLOCK__ANY)
+			return (ras->ras_err_state != 0);
+		else
+			return test_bit(block, &ras->ras_err_state) ||
+			       test_bit(AMDGPU_RAS_BLOCK__LAST,
+					&ras->ras_err_state);
+	}
+
+	return false;
+}
+
+static struct ras_event_manager *__get_ras_event_mgr(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *ras;
+
+	ras = amdgpu_ras_get_context(adev);
+	if (!ras)
+		return NULL;
+
+	return ras->event_mgr;
+}
+
+int amdgpu_ras_mark_ras_event_caller(struct amdgpu_device *adev, enum ras_event_type type,
+				     const void *caller)
+{
+	struct ras_event_manager *event_mgr;
+	struct ras_event_state *event_state;
+	int ret = 0;
+
+	if (amdgpu_uniras_enabled(adev))
+		return 0;
+
+	if (type >= RAS_EVENT_TYPE_COUNT) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	event_mgr = __get_ras_event_mgr(adev);
+	if (!event_mgr) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	event_state = &event_mgr->event_state[type];
+	event_state->last_seqno = atomic64_inc_return(&event_mgr->seqno);
+	atomic64_inc(&event_state->count);
+
+out:
+	if (ret && caller)
+		dev_warn(adev->dev, "failed mark ras event (%d) in %ps, ret:%d\n",
+			 (int)type, caller, ret);
+
+	return ret;
+}
+
+u64 amdgpu_ras_acquire_event_id(struct amdgpu_device *adev, enum ras_event_type type)
+{
+	struct ras_event_manager *event_mgr;
+	u64 id;
+
+	if (type >= RAS_EVENT_TYPE_COUNT)
+		return RAS_EVENT_INVALID_ID;
+
+	switch (type) {
+	case RAS_EVENT_TYPE_FATAL:
+	case RAS_EVENT_TYPE_POISON_CREATION:
+	case RAS_EVENT_TYPE_POISON_CONSUMPTION:
+		event_mgr = __get_ras_event_mgr(adev);
+		if (!event_mgr)
+			return RAS_EVENT_INVALID_ID;
+
+		id = event_mgr->event_state[type].last_seqno;
+		break;
+	case RAS_EVENT_TYPE_INVALID:
+	default:
+		id = RAS_EVENT_INVALID_ID;
+		break;
+	}
+
+	return id;
+}
+
+int amdgpu_ras_global_ras_isr(struct amdgpu_device *adev)
+{
 	if (atomic_cmpxchg(&amdgpu_ras_in_intr, 0, 1) == 0) {
-		dev_info(adev->dev, "uncorrectable hardware error"
-			"(ERREVENT_ATHUB_INTERRUPT) detected!\n");
+		struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+		enum ras_event_type type = RAS_EVENT_TYPE_FATAL;
+		u64 event_id = RAS_EVENT_INVALID_ID;
 
+		if (amdgpu_uniras_enabled(adev))
+			return 0;
+
+		if (!amdgpu_ras_mark_ras_event(adev, type))
+			event_id = amdgpu_ras_acquire_event_id(adev, type);
+
+		RAS_EVENT_LOG(adev, event_id, "uncorrectable hardware error"
+			      "(ERREVENT_ATHUB_INTERRUPT) detected!\n");
+
+		amdgpu_ras_set_fed(adev, true);
+		ras->gpu_reset_flags |= AMDGPU_RAS_GPU_RESET_MODE1_RESET;
 		amdgpu_ras_reset_gpu(adev);
 	}
+
+	return -EBUSY;
 }
 
 bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev)
@@ -3031,6 +4992,28 @@ static void amdgpu_register_bad_pages_mca_notifier(struct amdgpu_device *adev)
 		notifier_registered = true;
 	}
 }
+static void amdgpu_unregister_bad_pages_mca_notifier(struct amdgpu_device *adev)
+{
+	int i, j;
+
+	if (!notifier_registered && !mce_adev_list.num_gpu)
+		return;
+	for (i = 0, j = 0; i < mce_adev_list.num_gpu; i++) {
+		if (mce_adev_list.devs[i] == adev)
+			mce_adev_list.devs[i] = NULL;
+		if (!mce_adev_list.devs[i])
+			++j;
+	}
+
+	if (j == mce_adev_list.num_gpu) {
+		mce_adev_list.num_gpu = 0;
+		/* Unregister x86 notifier with MCE subsystem. */
+		if (notifier_registered) {
+			mce_unregister_decode_chain(&amdgpu_bad_page_nb);
+			notifier_registered = false;
+		}
+	}
+}
 #endif
 
 struct amdgpu_ras *amdgpu_ras_get_context(struct amdgpu_device *adev)
@@ -3069,6 +5052,11 @@ int amdgpu_ras_is_supported(struct amdgpu_device *adev,
 	 * that the ras block supports ras function.
 	 */
 	if (!ret &&
+	    (block == AMDGPU_RAS_BLOCK__GFX ||
+	     block == AMDGPU_RAS_BLOCK__SDMA ||
+	     block == AMDGPU_RAS_BLOCK__VCN ||
+	     block == AMDGPU_RAS_BLOCK__JPEG) &&
+		(amdgpu_ras_mask & (1 << block)) &&
 	    amdgpu_ras_is_poison_mode_supported(adev) &&
 	    amdgpu_ras_get_ras_block(adev, block, 0))
 		ret = 1;
@@ -3080,11 +5068,106 @@ int amdgpu_ras_reset_gpu(struct amdgpu_device *adev)
 {
 	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 
-	if (atomic_cmpxchg(&ras->in_recovery, 0, 1) == 0)
+	/* mode1 is the only selection for RMA status */
+	if (amdgpu_ras_is_rma(adev)) {
+		ras->gpu_reset_flags = 0;
+		ras->gpu_reset_flags |= AMDGPU_RAS_GPU_RESET_MODE1_RESET;
+	}
+
+	if (atomic_cmpxchg(&ras->in_recovery, 0, 1) == 0) {
+		struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev);
+		int hive_ras_recovery = 0;
+
+		if (hive) {
+			hive_ras_recovery = atomic_read(&hive->ras_recovery);
+			amdgpu_put_xgmi_hive(hive);
+		}
+		/* In the case of multiple GPUs, after a GPU has started
+		 * resetting all GPUs on hive, other GPUs do not need to
+		 * trigger GPU reset again.
+		 */
+		if (!hive_ras_recovery)
+			amdgpu_reset_domain_schedule(ras->adev->reset_domain, &ras->recovery_work);
+		else
+			atomic_set(&ras->in_recovery, 0);
+	} else {
+		flush_work(&ras->recovery_work);
 		amdgpu_reset_domain_schedule(ras->adev->reset_domain, &ras->recovery_work);
+	}
+
 	return 0;
 }
 
+int amdgpu_ras_set_mca_debug_mode(struct amdgpu_device *adev, bool enable)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	int ret = 0;
+
+	if (con) {
+		ret = amdgpu_mca_smu_set_debug_mode(adev, enable);
+		if (!ret)
+			con->is_aca_debug_mode = enable;
+	}
+
+	return ret;
+}
+
+int amdgpu_ras_set_aca_debug_mode(struct amdgpu_device *adev, bool enable)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	int ret = 0;
+
+	if (con) {
+		if (amdgpu_aca_is_enabled(adev))
+			ret = amdgpu_aca_smu_set_debug_mode(adev, enable);
+		else
+			ret = amdgpu_mca_smu_set_debug_mode(adev, enable);
+		if (!ret)
+			con->is_aca_debug_mode = enable;
+	}
+
+	return ret;
+}
+
+bool amdgpu_ras_get_aca_debug_mode(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	const struct aca_smu_funcs *smu_funcs = adev->aca.smu_funcs;
+	const struct amdgpu_mca_smu_funcs *mca_funcs = adev->mca.mca_funcs;
+
+	if (!con)
+		return false;
+
+	if ((amdgpu_aca_is_enabled(adev) && smu_funcs && smu_funcs->set_debug_mode) ||
+	    (!amdgpu_aca_is_enabled(adev) && mca_funcs && mca_funcs->mca_set_debug_mode))
+		return con->is_aca_debug_mode;
+	else
+		return true;
+}
+
+bool amdgpu_ras_get_error_query_mode(struct amdgpu_device *adev,
+				     unsigned int *error_query_mode)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	const struct amdgpu_mca_smu_funcs *mca_funcs = adev->mca.mca_funcs;
+	const struct aca_smu_funcs *smu_funcs = adev->aca.smu_funcs;
+
+	if (!con) {
+		*error_query_mode = AMDGPU_RAS_INVALID_ERROR_QUERY;
+		return false;
+	}
+
+	if (amdgpu_sriov_vf(adev)) {
+		*error_query_mode = AMDGPU_RAS_VIRT_ERROR_COUNT_QUERY;
+	} else if ((smu_funcs && smu_funcs->set_debug_mode) || (mca_funcs && mca_funcs->mca_set_debug_mode)) {
+		*error_query_mode =
+			(con->is_aca_debug_mode) ? AMDGPU_RAS_DIRECT_ERROR_QUERY : AMDGPU_RAS_FIRMWARE_ERROR_QUERY;
+	} else {
+		*error_query_mode = AMDGPU_RAS_DIRECT_ERROR_QUERY;
+	}
+
+	return true;
+}
 
 /* Register each ip ras block into amdgpu ras */
 int amdgpu_ras_register_ras_block(struct amdgpu_device *adev,
@@ -3104,3 +5187,562 @@ int amdgpu_ras_register_ras_block(struct amdgpu_device *adev,
 
 	return 0;
 }
+
+void amdgpu_ras_get_error_type_name(uint32_t err_type, char *err_type_name)
+{
+	if (!err_type_name)
+		return;
+
+	switch (err_type) {
+	case AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE:
+		sprintf(err_type_name, "correctable");
+		break;
+	case AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE:
+		sprintf(err_type_name, "uncorrectable");
+		break;
+	default:
+		sprintf(err_type_name, "unknown");
+		break;
+	}
+}
+
+bool amdgpu_ras_inst_get_memory_id_field(struct amdgpu_device *adev,
+					 const struct amdgpu_ras_err_status_reg_entry *reg_entry,
+					 uint32_t instance,
+					 uint32_t *memory_id)
+{
+	uint32_t err_status_lo_data, err_status_lo_offset;
+
+	if (!reg_entry)
+		return false;
+
+	err_status_lo_offset =
+		AMDGPU_RAS_REG_ENTRY_OFFSET(reg_entry->hwip, instance,
+					    reg_entry->seg_lo, reg_entry->reg_lo);
+	err_status_lo_data = RREG32(err_status_lo_offset);
+
+	if ((reg_entry->flags & AMDGPU_RAS_ERR_STATUS_VALID) &&
+	    !REG_GET_FIELD(err_status_lo_data, ERR_STATUS_LO, ERR_STATUS_VALID_FLAG))
+		return false;
+
+	*memory_id = REG_GET_FIELD(err_status_lo_data, ERR_STATUS_LO, MEMORY_ID);
+
+	return true;
+}
+
+bool amdgpu_ras_inst_get_err_cnt_field(struct amdgpu_device *adev,
+				       const struct amdgpu_ras_err_status_reg_entry *reg_entry,
+				       uint32_t instance,
+				       unsigned long *err_cnt)
+{
+	uint32_t err_status_hi_data, err_status_hi_offset;
+
+	if (!reg_entry)
+		return false;
+
+	err_status_hi_offset =
+		AMDGPU_RAS_REG_ENTRY_OFFSET(reg_entry->hwip, instance,
+					    reg_entry->seg_hi, reg_entry->reg_hi);
+	err_status_hi_data = RREG32(err_status_hi_offset);
+
+	if ((reg_entry->flags & AMDGPU_RAS_ERR_INFO_VALID) &&
+	    !REG_GET_FIELD(err_status_hi_data, ERR_STATUS_HI, ERR_INFO_VALID_FLAG))
+		/* keep the check here in case we need to refer to the result later */
+		dev_dbg(adev->dev, "Invalid err_info field\n");
+
+	/* read err count */
+	*err_cnt = REG_GET_FIELD(err_status_hi_data, ERR_STATUS, ERR_CNT);
+
+	return true;
+}
+
+void amdgpu_ras_inst_query_ras_error_count(struct amdgpu_device *adev,
+					   const struct amdgpu_ras_err_status_reg_entry *reg_list,
+					   uint32_t reg_list_size,
+					   const struct amdgpu_ras_memory_id_entry *mem_list,
+					   uint32_t mem_list_size,
+					   uint32_t instance,
+					   uint32_t err_type,
+					   unsigned long *err_count)
+{
+	uint32_t memory_id;
+	unsigned long err_cnt;
+	char err_type_name[16];
+	uint32_t i, j;
+
+	for (i = 0; i < reg_list_size; i++) {
+		/* query memory_id from err_status_lo */
+		if (!amdgpu_ras_inst_get_memory_id_field(adev, &reg_list[i],
+							 instance, &memory_id))
+			continue;
+
+		/* query err_cnt from err_status_hi */
+		if (!amdgpu_ras_inst_get_err_cnt_field(adev, &reg_list[i],
+						       instance, &err_cnt) ||
+		    !err_cnt)
+			continue;
+
+		*err_count += err_cnt;
+
+		/* log the errors */
+		amdgpu_ras_get_error_type_name(err_type, err_type_name);
+		if (!mem_list) {
+			/* memory_list is not supported */
+			dev_info(adev->dev,
+				 "%ld %s hardware errors detected in %s, instance: %d, memory_id: %d\n",
+				 err_cnt, err_type_name,
+				 reg_list[i].block_name,
+				 instance, memory_id);
+		} else {
+			for (j = 0; j < mem_list_size; j++) {
+				if (memory_id == mem_list[j].memory_id) {
+					dev_info(adev->dev,
+						 "%ld %s hardware errors detected in %s, instance: %d, memory block: %s\n",
+						 err_cnt, err_type_name,
+						 reg_list[i].block_name,
+						 instance, mem_list[j].name);
+					break;
+				}
+			}
+		}
+	}
+}
+
+void amdgpu_ras_inst_reset_ras_error_count(struct amdgpu_device *adev,
+					   const struct amdgpu_ras_err_status_reg_entry *reg_list,
+					   uint32_t reg_list_size,
+					   uint32_t instance)
+{
+	uint32_t err_status_lo_offset, err_status_hi_offset;
+	uint32_t i;
+
+	for (i = 0; i < reg_list_size; i++) {
+		err_status_lo_offset =
+			AMDGPU_RAS_REG_ENTRY_OFFSET(reg_list[i].hwip, instance,
+						    reg_list[i].seg_lo, reg_list[i].reg_lo);
+		err_status_hi_offset =
+			AMDGPU_RAS_REG_ENTRY_OFFSET(reg_list[i].hwip, instance,
+						    reg_list[i].seg_hi, reg_list[i].reg_hi);
+		WREG32(err_status_lo_offset, 0);
+		WREG32(err_status_hi_offset, 0);
+	}
+}
+
+int amdgpu_ras_error_data_init(struct ras_err_data *err_data)
+{
+	memset(err_data, 0, sizeof(*err_data));
+
+	INIT_LIST_HEAD(&err_data->err_node_list);
+
+	return 0;
+}
+
+static void amdgpu_ras_error_node_release(struct ras_err_node *err_node)
+{
+	if (!err_node)
+		return;
+
+	list_del(&err_node->node);
+	kvfree(err_node);
+}
+
+void amdgpu_ras_error_data_fini(struct ras_err_data *err_data)
+{
+	struct ras_err_node *err_node, *tmp;
+
+	list_for_each_entry_safe(err_node, tmp, &err_data->err_node_list, node)
+		amdgpu_ras_error_node_release(err_node);
+}
+
+static struct ras_err_node *amdgpu_ras_error_find_node_by_id(struct ras_err_data *err_data,
+							     struct amdgpu_smuio_mcm_config_info *mcm_info)
+{
+	struct ras_err_node *err_node;
+	struct amdgpu_smuio_mcm_config_info *ref_id;
+
+	if (!err_data || !mcm_info)
+		return NULL;
+
+	for_each_ras_error(err_node, err_data) {
+		ref_id = &err_node->err_info.mcm_info;
+
+		if (mcm_info->socket_id == ref_id->socket_id &&
+		    mcm_info->die_id == ref_id->die_id)
+			return err_node;
+	}
+
+	return NULL;
+}
+
+static struct ras_err_node *amdgpu_ras_error_node_new(void)
+{
+	struct ras_err_node *err_node;
+
+	err_node = kvzalloc(sizeof(*err_node), GFP_KERNEL);
+	if (!err_node)
+		return NULL;
+
+	INIT_LIST_HEAD(&err_node->node);
+
+	return err_node;
+}
+
+static int ras_err_info_cmp(void *priv, const struct list_head *a, const struct list_head *b)
+{
+	struct ras_err_node *nodea = container_of(a, struct ras_err_node, node);
+	struct ras_err_node *nodeb = container_of(b, struct ras_err_node, node);
+	struct amdgpu_smuio_mcm_config_info *infoa = &nodea->err_info.mcm_info;
+	struct amdgpu_smuio_mcm_config_info *infob = &nodeb->err_info.mcm_info;
+
+	if (unlikely(infoa->socket_id != infob->socket_id))
+		return infoa->socket_id - infob->socket_id;
+	else
+		return infoa->die_id - infob->die_id;
+
+	return 0;
+}
+
+static struct ras_err_info *amdgpu_ras_error_get_info(struct ras_err_data *err_data,
+				struct amdgpu_smuio_mcm_config_info *mcm_info)
+{
+	struct ras_err_node *err_node;
+
+	err_node = amdgpu_ras_error_find_node_by_id(err_data, mcm_info);
+	if (err_node)
+		return &err_node->err_info;
+
+	err_node = amdgpu_ras_error_node_new();
+	if (!err_node)
+		return NULL;
+
+	memcpy(&err_node->err_info.mcm_info, mcm_info, sizeof(*mcm_info));
+
+	err_data->err_list_count++;
+	list_add_tail(&err_node->node, &err_data->err_node_list);
+	list_sort(NULL, &err_data->err_node_list, ras_err_info_cmp);
+
+	return &err_node->err_info;
+}
+
+int amdgpu_ras_error_statistic_ue_count(struct ras_err_data *err_data,
+					struct amdgpu_smuio_mcm_config_info *mcm_info,
+					u64 count)
+{
+	struct ras_err_info *err_info;
+
+	if (!err_data || !mcm_info)
+		return -EINVAL;
+
+	if (!count)
+		return 0;
+
+	err_info = amdgpu_ras_error_get_info(err_data, mcm_info);
+	if (!err_info)
+		return -EINVAL;
+
+	err_info->ue_count += count;
+	err_data->ue_count += count;
+
+	return 0;
+}
+
+int amdgpu_ras_error_statistic_ce_count(struct ras_err_data *err_data,
+					struct amdgpu_smuio_mcm_config_info *mcm_info,
+					u64 count)
+{
+	struct ras_err_info *err_info;
+
+	if (!err_data || !mcm_info)
+		return -EINVAL;
+
+	if (!count)
+		return 0;
+
+	err_info = amdgpu_ras_error_get_info(err_data, mcm_info);
+	if (!err_info)
+		return -EINVAL;
+
+	err_info->ce_count += count;
+	err_data->ce_count += count;
+
+	return 0;
+}
+
+int amdgpu_ras_error_statistic_de_count(struct ras_err_data *err_data,
+					struct amdgpu_smuio_mcm_config_info *mcm_info,
+					u64 count)
+{
+	struct ras_err_info *err_info;
+
+	if (!err_data || !mcm_info)
+		return -EINVAL;
+
+	if (!count)
+		return 0;
+
+	err_info = amdgpu_ras_error_get_info(err_data, mcm_info);
+	if (!err_info)
+		return -EINVAL;
+
+	err_info->de_count += count;
+	err_data->de_count += count;
+
+	return 0;
+}
+
+#define mmMP0_SMN_C2PMSG_92	0x1609C
+#define mmMP0_SMN_C2PMSG_126	0x160BE
+static void amdgpu_ras_boot_time_error_reporting(struct amdgpu_device *adev,
+						 u32 instance)
+{
+	u32 socket_id, aid_id, hbm_id;
+	u32 fw_status;
+	u32 boot_error;
+	u64 reg_addr;
+
+	/* The pattern for smn addressing in other SOC could be different from
+	 * the one for aqua_vanjaram. We should revisit the code if the pattern
+	 * is changed. In such case, replace the aqua_vanjaram implementation
+	 * with more common helper */
+	reg_addr = (mmMP0_SMN_C2PMSG_92 << 2) +
+		   aqua_vanjaram_encode_ext_smn_addressing(instance);
+	fw_status = amdgpu_device_indirect_rreg_ext(adev, reg_addr);
+
+	reg_addr = (mmMP0_SMN_C2PMSG_126 << 2) +
+		   aqua_vanjaram_encode_ext_smn_addressing(instance);
+	boot_error = amdgpu_device_indirect_rreg_ext(adev, reg_addr);
+
+	socket_id = AMDGPU_RAS_GPU_ERR_SOCKET_ID(boot_error);
+	aid_id = AMDGPU_RAS_GPU_ERR_AID_ID(boot_error);
+	hbm_id = ((1 == AMDGPU_RAS_GPU_ERR_HBM_ID(boot_error)) ? 0 : 1);
+
+	if (AMDGPU_RAS_GPU_ERR_MEM_TRAINING(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, hbm: %d, fw_status: 0x%x, memory training failed\n",
+			 socket_id, aid_id, hbm_id, fw_status);
+
+	if (AMDGPU_RAS_GPU_ERR_FW_LOAD(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, fw_status: 0x%x, firmware load failed at boot time\n",
+			 socket_id, aid_id, fw_status);
+
+	if (AMDGPU_RAS_GPU_ERR_WAFL_LINK_TRAINING(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, fw_status: 0x%x, wafl link training failed\n",
+			 socket_id, aid_id, fw_status);
+
+	if (AMDGPU_RAS_GPU_ERR_XGMI_LINK_TRAINING(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, fw_status: 0x%x, xgmi link training failed\n",
+			 socket_id, aid_id, fw_status);
+
+	if (AMDGPU_RAS_GPU_ERR_USR_CP_LINK_TRAINING(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, fw_status: 0x%x, usr cp link training failed\n",
+			 socket_id, aid_id, fw_status);
+
+	if (AMDGPU_RAS_GPU_ERR_USR_DP_LINK_TRAINING(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, fw_status: 0x%x, usr dp link training failed\n",
+			 socket_id, aid_id, fw_status);
+
+	if (AMDGPU_RAS_GPU_ERR_HBM_MEM_TEST(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, hbm: %d, fw_status: 0x%x, hbm memory test failed\n",
+			 socket_id, aid_id, hbm_id, fw_status);
+
+	if (AMDGPU_RAS_GPU_ERR_HBM_BIST_TEST(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, hbm: %d, fw_status: 0x%x, hbm bist test failed\n",
+			 socket_id, aid_id, hbm_id, fw_status);
+
+	if (AMDGPU_RAS_GPU_ERR_DATA_ABORT(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, fw_status: 0x%x, data abort exception\n",
+			 socket_id, aid_id, fw_status);
+
+	if (AMDGPU_RAS_GPU_ERR_GENERIC(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, fw_status: 0x%x, Boot Controller Generic Error\n",
+			 socket_id, aid_id, fw_status);
+}
+
+static bool amdgpu_ras_boot_error_detected(struct amdgpu_device *adev,
+					   u32 instance)
+{
+	u64 reg_addr;
+	u32 reg_data;
+	int retry_loop;
+
+	reg_addr = (mmMP0_SMN_C2PMSG_92 << 2) +
+		   aqua_vanjaram_encode_ext_smn_addressing(instance);
+
+	for (retry_loop = 0; retry_loop < AMDGPU_RAS_BOOT_STATUS_POLLING_LIMIT; retry_loop++) {
+		reg_data = amdgpu_device_indirect_rreg_ext(adev, reg_addr);
+		if ((reg_data & AMDGPU_RAS_BOOT_STATUS_MASK) == AMDGPU_RAS_BOOT_STEADY_STATUS)
+			return false;
+		else
+			msleep(1);
+	}
+
+	return true;
+}
+
+void amdgpu_ras_query_boot_status(struct amdgpu_device *adev, u32 num_instances)
+{
+	u32 i;
+
+	for (i = 0; i < num_instances; i++) {
+		if (amdgpu_ras_boot_error_detected(adev, i))
+			amdgpu_ras_boot_time_error_reporting(adev, i);
+	}
+}
+
+int amdgpu_ras_reserve_page(struct amdgpu_device *adev, uint64_t pfn)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct amdgpu_vram_mgr *mgr = &adev->mman.vram_mgr;
+	uint64_t start = pfn << AMDGPU_GPU_PAGE_SHIFT;
+	int ret = 0;
+
+	if (amdgpu_ras_check_critical_address(adev, start))
+		return 0;
+
+	mutex_lock(&con->page_rsv_lock);
+	ret = amdgpu_vram_mgr_query_page_status(mgr, start);
+	if (ret == -ENOENT)
+		ret = amdgpu_vram_mgr_reserve_range(mgr, start, AMDGPU_GPU_PAGE_SIZE);
+	mutex_unlock(&con->page_rsv_lock);
+
+	return ret;
+}
+
+void amdgpu_ras_event_log_print(struct amdgpu_device *adev, u64 event_id,
+				const char *fmt, ...)
+{
+	struct va_format vaf;
+	va_list args;
+
+	va_start(args, fmt);
+	vaf.fmt = fmt;
+	vaf.va = &args;
+
+	if (RAS_EVENT_ID_IS_VALID(event_id))
+		dev_printk(KERN_INFO, adev->dev, "{%llu}%pV", event_id, &vaf);
+	else
+		dev_printk(KERN_INFO, adev->dev, "%pV", &vaf);
+
+	va_end(args);
+}
+
+bool amdgpu_ras_is_rma(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+
+	if (amdgpu_uniras_enabled(adev))
+		return amdgpu_ras_mgr_is_rma(adev);
+
+	if (!con)
+		return false;
+
+	return con->is_rma;
+}
+
+int amdgpu_ras_add_critical_region(struct amdgpu_device *adev,
+			struct amdgpu_bo *bo)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct amdgpu_vram_mgr_resource *vres;
+	struct ras_critical_region *region;
+	struct drm_buddy_block *block;
+	int ret = 0;
+
+	if (!bo || !bo->tbo.resource)
+		return -EINVAL;
+
+	vres = to_amdgpu_vram_mgr_resource(bo->tbo.resource);
+
+	mutex_lock(&con->critical_region_lock);
+
+	/* Check if the bo had been recorded */
+	list_for_each_entry(region, &con->critical_region_head, node)
+		if (region->bo == bo)
+			goto out;
+
+	/* Record new critical amdgpu bo */
+	list_for_each_entry(block, &vres->blocks, link) {
+		region = kzalloc(sizeof(*region), GFP_KERNEL);
+		if (!region) {
+			ret = -ENOMEM;
+			goto out;
+		}
+		region->bo = bo;
+		region->start = amdgpu_vram_mgr_block_start(block);
+		region->size = amdgpu_vram_mgr_block_size(block);
+		list_add_tail(&region->node, &con->critical_region_head);
+	}
+
+out:
+	mutex_unlock(&con->critical_region_lock);
+
+	return ret;
+}
+
+static void amdgpu_ras_critical_region_init(struct amdgpu_device *adev)
+{
+	amdgpu_ras_add_critical_region(adev, adev->mman.fw_reserved_memory);
+}
+
+static void amdgpu_ras_critical_region_fini(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct ras_critical_region *region, *tmp;
+
+	mutex_lock(&con->critical_region_lock);
+	list_for_each_entry_safe(region, tmp, &con->critical_region_head, node) {
+		list_del(&region->node);
+		kfree(region);
+	}
+	mutex_unlock(&con->critical_region_lock);
+}
+
+bool amdgpu_ras_check_critical_address(struct amdgpu_device *adev, uint64_t addr)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct ras_critical_region *region;
+	bool ret = false;
+
+	mutex_lock(&con->critical_region_lock);
+	list_for_each_entry(region, &con->critical_region_head, node) {
+		if ((region->start <= addr) &&
+		    (addr < (region->start + region->size))) {
+			ret = true;
+			break;
+		}
+	}
+	mutex_unlock(&con->critical_region_lock);
+
+	return ret;
+}
+
+void amdgpu_ras_pre_reset(struct amdgpu_device *adev,
+					  struct list_head *device_list)
+{
+	struct amdgpu_device *tmp_adev = NULL;
+
+	list_for_each_entry(tmp_adev, device_list, reset_list) {
+		if (amdgpu_uniras_enabled(tmp_adev))
+			amdgpu_ras_mgr_pre_reset(tmp_adev);
+	}
+}
+
+void amdgpu_ras_post_reset(struct amdgpu_device *adev,
+					  struct list_head *device_list)
+{
+	struct amdgpu_device *tmp_adev = NULL;
+
+	list_for_each_entry(tmp_adev, device_list, reset_list) {
+		if (amdgpu_uniras_enabled(tmp_adev))
+			amdgpu_ras_mgr_post_reset(tmp_adev);
+	}
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
index 17b3d1992e80..ff44190d7d98 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
@@ -26,12 +26,57 @@
 
 #include <linux/debugfs.h>
 #include <linux/list.h>
+#include <linux/kfifo.h>
+#include <linux/radix-tree.h>
 #include "ta_ras_if.h"
 #include "amdgpu_ras_eeprom.h"
+#include "amdgpu_smuio.h"
+#include "amdgpu_aca.h"
 
 struct amdgpu_iv_entry;
 
+#define AMDGPU_RAS_GPU_ERR_MEM_TRAINING(x)		AMDGPU_GET_REG_FIELD(x, 0, 0)
+#define AMDGPU_RAS_GPU_ERR_FW_LOAD(x)			AMDGPU_GET_REG_FIELD(x, 1, 1)
+#define AMDGPU_RAS_GPU_ERR_WAFL_LINK_TRAINING(x)	AMDGPU_GET_REG_FIELD(x, 2, 2)
+#define AMDGPU_RAS_GPU_ERR_XGMI_LINK_TRAINING(x)	AMDGPU_GET_REG_FIELD(x, 3, 3)
+#define AMDGPU_RAS_GPU_ERR_USR_CP_LINK_TRAINING(x)	AMDGPU_GET_REG_FIELD(x, 4, 4)
+#define AMDGPU_RAS_GPU_ERR_USR_DP_LINK_TRAINING(x)	AMDGPU_GET_REG_FIELD(x, 5, 5)
+#define AMDGPU_RAS_GPU_ERR_HBM_MEM_TEST(x)		AMDGPU_GET_REG_FIELD(x, 6, 6)
+#define AMDGPU_RAS_GPU_ERR_HBM_BIST_TEST(x)		AMDGPU_GET_REG_FIELD(x, 7, 7)
+#define AMDGPU_RAS_GPU_ERR_SOCKET_ID(x)			AMDGPU_GET_REG_FIELD(x, 10, 8)
+#define AMDGPU_RAS_GPU_ERR_AID_ID(x)			AMDGPU_GET_REG_FIELD(x, 12, 11)
+#define AMDGPU_RAS_GPU_ERR_HBM_ID(x)			AMDGPU_GET_REG_FIELD(x, 14, 13)
+#define AMDGPU_RAS_GPU_ERR_DATA_ABORT(x)		AMDGPU_GET_REG_FIELD(x, 29, 29)
+#define AMDGPU_RAS_GPU_ERR_GENERIC(x)			AMDGPU_GET_REG_FIELD(x, 30, 30)
+
+#define AMDGPU_RAS_BOOT_STATUS_POLLING_LIMIT	100
+#define AMDGPU_RAS_BOOT_STEADY_STATUS		0xBA
+#define AMDGPU_RAS_BOOT_STATUS_MASK		0xFF
+
 #define AMDGPU_RAS_FLAG_INIT_BY_VBIOS		(0x1 << 0)
+/* position of instance value in sub_block_index of
+ * ta_ras_trigger_error_input, the sub block uses lower 12 bits
+ */
+#define AMDGPU_RAS_INST_MASK 0xfffff000
+#define AMDGPU_RAS_INST_SHIFT 0xc
+
+#define AMDGPU_RAS_FEATURES_SOCKETID_SHIFT 29
+#define AMDGPU_RAS_FEATURES_SOCKETID_MASK 0xe0000000
+
+/* Reserve 8 physical dram row for possible retirement.
+ * In worst cases, it will lose 8 * 2MB memory in vram domain */
+#define AMDGPU_RAS_RESERVED_VRAM_SIZE_DEFAULT	(16ULL << 20)
+/* The high three bits indicates socketid */
+#define AMDGPU_RAS_GET_FEATURES(val)  ((val) & ~AMDGPU_RAS_FEATURES_SOCKETID_MASK)
+
+#define RAS_EVENT_INVALID_ID		(BIT_ULL(63))
+#define RAS_EVENT_ID_IS_VALID(x)	(!((x) & BIT_ULL(63)))
+
+#define RAS_EVENT_LOG(adev, id, fmt, ...)	\
+	amdgpu_ras_event_log_print((adev), (id), (fmt), ##__VA_ARGS__)
+
+#define amdgpu_ras_mark_ras_event(adev, type)	\
+	(amdgpu_ras_mark_ras_event_caller((adev), (type), __builtin_return_address(0)))
 
 enum amdgpu_ras_block {
 	AMDGPU_RAS_BLOCK__UMC = 0,
@@ -51,8 +96,12 @@ enum amdgpu_ras_block {
 	AMDGPU_RAS_BLOCK__MCA,
 	AMDGPU_RAS_BLOCK__VCN,
 	AMDGPU_RAS_BLOCK__JPEG,
+	AMDGPU_RAS_BLOCK__IH,
+	AMDGPU_RAS_BLOCK__MPIO,
+	AMDGPU_RAS_BLOCK__MMSCH,
 
-	AMDGPU_RAS_BLOCK__LAST
+	AMDGPU_RAS_BLOCK__LAST,
+	AMDGPU_RAS_BLOCK__ANY = -1
 };
 
 enum amdgpu_ras_mca_block {
@@ -314,6 +363,53 @@ enum amdgpu_ras_ret {
 	AMDGPU_RAS_PT,
 };
 
+enum amdgpu_ras_error_query_mode {
+	AMDGPU_RAS_INVALID_ERROR_QUERY		= 0,
+	AMDGPU_RAS_DIRECT_ERROR_QUERY		= 1,
+	AMDGPU_RAS_FIRMWARE_ERROR_QUERY		= 2,
+	AMDGPU_RAS_VIRT_ERROR_COUNT_QUERY	= 3,
+};
+
+/* ras error status reisger fields */
+#define ERR_STATUS_LO__ERR_STATUS_VALID_FLAG__SHIFT	0x0
+#define ERR_STATUS_LO__ERR_STATUS_VALID_FLAG_MASK	0x00000001L
+#define ERR_STATUS_LO__MEMORY_ID__SHIFT			0x18
+#define ERR_STATUS_LO__MEMORY_ID_MASK			0xFF000000L
+#define ERR_STATUS_HI__ERR_INFO_VALID_FLAG__SHIFT	0x2
+#define ERR_STATUS_HI__ERR_INFO_VALID_FLAG_MASK		0x00000004L
+#define ERR_STATUS__ERR_CNT__SHIFT			0x17
+#define ERR_STATUS__ERR_CNT_MASK			0x03800000L
+
+#define AMDGPU_RAS_REG_ENTRY(ip, inst, reg_lo, reg_hi) \
+	ip##_HWIP, inst, reg_lo##_BASE_IDX, reg_lo, reg_hi##_BASE_IDX, reg_hi
+
+#define AMDGPU_RAS_REG_ENTRY_OFFSET(hwip, ip_inst, segment, reg) \
+	(adev->reg_offset[hwip][ip_inst][segment] + (reg))
+
+#define AMDGPU_RAS_ERR_INFO_VALID	(1 << 0)
+#define AMDGPU_RAS_ERR_STATUS_VALID	(1 << 1)
+#define AMDGPU_RAS_ERR_ADDRESS_VALID	(1 << 2)
+
+#define AMDGPU_RAS_GPU_RESET_MODE2_RESET  (0x1 << 0)
+#define AMDGPU_RAS_GPU_RESET_MODE1_RESET  (0x1 << 1)
+
+struct amdgpu_ras_err_status_reg_entry {
+	uint32_t hwip;
+	uint32_t ip_inst;
+	uint32_t seg_lo;
+	uint32_t reg_lo;
+	uint32_t seg_hi;
+	uint32_t reg_hi;
+	uint32_t reg_inst;
+	uint32_t flags;
+	const char *block_name;
+};
+
+struct amdgpu_ras_memory_id_entry {
+	uint32_t memory_id;
+	const char *name;
+};
+
 struct ras_common_if {
 	enum amdgpu_ras_block block;
 	enum amdgpu_ras_error_type type;
@@ -340,13 +436,111 @@ struct umc_ecc_info {
 	int record_ce_addr_supported;
 };
 
+enum ras_event_type {
+	RAS_EVENT_TYPE_INVALID = 0,
+	RAS_EVENT_TYPE_FATAL,
+	RAS_EVENT_TYPE_POISON_CREATION,
+	RAS_EVENT_TYPE_POISON_CONSUMPTION,
+	RAS_EVENT_TYPE_COUNT,
+};
+
+struct ras_event_state {
+	u64 last_seqno;
+	atomic64_t count;
+};
+
+struct ras_event_manager {
+	atomic64_t seqno;
+	struct ras_event_state event_state[RAS_EVENT_TYPE_COUNT];
+};
+
+struct ras_event_id {
+	enum ras_event_type type;
+	u64 event_id;
+};
+
+struct ras_query_context {
+	struct ras_event_id evid;
+};
+
+typedef int (*pasid_notify)(struct amdgpu_device *adev,
+		uint16_t pasid, void *data);
+
+struct ras_poison_msg {
+	enum amdgpu_ras_block block;
+	uint16_t pasid;
+	uint32_t reset;
+	pasid_notify pasid_fn;
+	void *data;
+};
+
+struct ras_err_pages {
+	uint32_t count;
+	uint64_t *pfn;
+};
+
+struct ras_ecc_err {
+	uint64_t status;
+	uint64_t ipid;
+	uint64_t addr;
+	uint64_t pa_pfn;
+	/* save global channel index across all UMC instances */
+	uint32_t channel_idx;
+	struct ras_err_pages err_pages;
+};
+
+struct ras_ecc_log_info {
+	struct mutex lock;
+	struct radix_tree_root de_page_tree;
+	uint64_t de_queried_count;
+	uint64_t consumption_q_count;
+};
+
+struct ras_critical_region {
+	struct list_head node;
+	struct amdgpu_bo *bo;
+	uint64_t start;
+	uint64_t size;
+};
+
+struct ras_eeprom_table_version {
+	uint32_t minor    : 16;
+	uint32_t major    : 16;
+};
+
+struct ras_eeprom_smu_funcs {
+	int (*get_ras_table_version)(struct amdgpu_device *adev,
+							uint32_t *table_version);
+	int (*get_badpage_count)(struct amdgpu_device *adev, uint32_t *count, uint32_t timeout);
+	int (*get_badpage_mca_addr)(struct amdgpu_device *adev, uint16_t index, uint64_t *mca_addr);
+	int (*set_timestamp)(struct amdgpu_device *adev, uint64_t timestamp);
+	int (*get_timestamp)(struct amdgpu_device *adev,
+							uint16_t index, uint64_t *timestamp);
+	int (*get_badpage_ipid)(struct amdgpu_device *adev, uint16_t index, uint64_t *ipid);
+	int (*erase_ras_table)(struct amdgpu_device *adev, uint32_t *result);
+};
+
+enum ras_smu_feature_flags {
+	RAS_SMU_FEATURE_BIT__RAS_EEPROM = BIT_ULL(0),
+};
+
+struct ras_smu_drv {
+	const struct ras_eeprom_smu_funcs *smu_eeprom_funcs;
+	void (*ras_smu_feature_flags)(struct amdgpu_device *adev, uint64_t *flags);
+};
+
 struct amdgpu_ras {
+	void *ras_mgr;
 	/* ras infrastructure */
 	/* for ras itself. */
 	uint32_t features;
+	uint32_t schema;
 	struct list_head head;
 	/* sysfs */
 	struct device_attribute features_attr;
+	struct device_attribute version_attr;
+	struct device_attribute schema_attr;
+	struct device_attribute event_state_attr;
 	struct bin_attribute badpages_attr;
 	struct dentry *de_ras_eeprom_table;
 	/* block array */
@@ -355,6 +549,7 @@ struct amdgpu_ras {
 	/* gpu recovery */
 	struct work_struct recovery_work;
 	atomic_t in_recovery;
+	atomic_t rma_in_recovery;
 	struct amdgpu_device *adev;
 	/* error handler data */
 	struct ras_err_handler_data *eh_data;
@@ -385,25 +580,86 @@ struct amdgpu_ras {
 
 	/* Indicates smu whether need update bad channel info */
 	bool update_channel_flag;
+	/* Record status of smu mca debug mode */
+	bool is_aca_debug_mode;
+	bool is_rma;
+
+	/* Record special requirements of gpu reset caller */
+	uint32_t  gpu_reset_flags;
+
+	struct task_struct *page_retirement_thread;
+	wait_queue_head_t page_retirement_wq;
+	struct mutex page_retirement_lock;
+	atomic_t page_retirement_req_cnt;
+	atomic_t poison_creation_count;
+	atomic_t poison_consumption_count;
+	struct mutex page_rsv_lock;
+	DECLARE_KFIFO(poison_fifo, struct ras_poison_msg, 128);
+	struct ras_ecc_log_info  umc_ecc_log;
+	struct delayed_work page_retirement_dwork;
+
+	/* ras errors detected */
+	unsigned long ras_err_state;
+
+	/* RAS event manager */
+	struct ras_event_manager __event_mgr;
+	struct ras_event_manager *event_mgr;
+
+	uint64_t reserved_pages_in_bytes;
+
+	pid_t init_task_pid;
+	char init_task_comm[TASK_COMM_LEN];
+
+	int bad_page_num;
+
+	struct list_head critical_region_head;
+	struct mutex critical_region_lock;
+
+	/* Protect poison injection */
+	struct mutex poison_lock;
+
+	/* Disable/Enable uniras switch */
+	bool uniras_enabled;
+	const struct ras_smu_drv *ras_smu_drv;
 };
 
 struct ras_fs_data {
-	char sysfs_name[32];
+	char sysfs_name[48];
 	char debugfs_name[32];
 };
 
+struct ras_err_info {
+	struct amdgpu_smuio_mcm_config_info mcm_info;
+	u64 ce_count;
+	u64 ue_count;
+	u64 de_count;
+};
+
+struct ras_err_node {
+	struct list_head node;
+	struct ras_err_info err_info;
+};
+
 struct ras_err_data {
 	unsigned long ue_count;
 	unsigned long ce_count;
+	unsigned long de_count;
 	unsigned long err_addr_cnt;
 	struct eeprom_table_record *err_addr;
+	unsigned long err_addr_len;
+	u32 err_list_count;
+	struct list_head err_node_list;
 };
 
+#define for_each_ras_error(err_node, err_data) \
+	list_for_each_entry(err_node, &(err_data)->err_node_list, node)
+
 struct ras_err_handler_data {
 	/* point to bad page records array */
 	struct eeprom_table_record *bps;
 	/* the count of entries */
 	int count;
+	int count_saved;
 	/* the space can place new entries */
 	int space_left;
 };
@@ -446,6 +702,8 @@ struct ras_manager {
 	struct ras_ih_data ih_data;
 
 	struct ras_err_data err_data;
+
+	struct aca_handle aca_handle;
 };
 
 struct ras_badpage {
@@ -465,12 +723,14 @@ struct ras_query_if {
 	struct ras_common_if head;
 	unsigned long ue_count;
 	unsigned long ce_count;
+	unsigned long de_count;
 };
 
 struct ras_inject_if {
 	struct ras_common_if head;
 	uint64_t address;
 	uint64_t value;
+	uint32_t instance_mask;
 };
 
 struct ras_cure_if {
@@ -508,7 +768,8 @@ struct amdgpu_ras_block_object {
 };
 
 struct amdgpu_ras_block_hw_ops {
-	int  (*ras_error_inject)(struct amdgpu_device *adev, void *inject_if);
+	int  (*ras_error_inject)(struct amdgpu_device *adev,
+			void *inject_if, uint32_t instance_mask);
 	void (*query_ras_error_count)(struct amdgpu_device *adev, void *ras_error_status);
 	void (*query_ras_error_status)(struct amdgpu_device *adev);
 	void (*query_ras_error_address)(struct amdgpu_device *adev, void *ras_error_status);
@@ -532,8 +793,8 @@ struct amdgpu_ras_block_hw_ops {
  * 8: feature disable
  */
 
-
-int amdgpu_ras_recovery_init(struct amdgpu_device *adev);
+int amdgpu_ras_init_badpage_info(struct amdgpu_device *adev);
+int amdgpu_ras_recovery_init(struct amdgpu_device *adev, bool init_bp_info);
 
 void amdgpu_ras_resume(struct amdgpu_device *adev);
 void amdgpu_ras_suspend(struct amdgpu_device *adev);
@@ -545,7 +806,7 @@ int amdgpu_ras_query_error_count(struct amdgpu_device *adev,
 
 /* error handling functions */
 int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev,
-		struct eeprom_table_record *bps, int pages);
+		struct eeprom_table_record *bps, int pages, bool from_rom);
 
 int amdgpu_ras_save_bad_pages(struct amdgpu_device *adev,
 		unsigned long *new_cnt);
@@ -587,6 +848,12 @@ amdgpu_ras_block_to_ta(enum amdgpu_ras_block block) {
 		return TA_RAS_BLOCK__VCN;
 	case AMDGPU_RAS_BLOCK__JPEG:
 		return TA_RAS_BLOCK__JPEG;
+	case AMDGPU_RAS_BLOCK__IH:
+		return TA_RAS_BLOCK__IH;
+	case AMDGPU_RAS_BLOCK__MPIO:
+		return TA_RAS_BLOCK__MPIO;
+	case AMDGPU_RAS_BLOCK__MMSCH:
+		return TA_RAS_BLOCK__MMSCH;
 	default:
 		WARN_ONCE(1, "RAS ERROR: unexpected block id %d\n", block);
 		return TA_RAS_BLOCK__UMC;
@@ -641,6 +908,8 @@ void amdgpu_ras_debugfs_create_all(struct amdgpu_device *adev);
 int amdgpu_ras_query_error_status(struct amdgpu_device *adev,
 		struct ras_query_if *info);
 
+int amdgpu_ras_reset_error_count(struct amdgpu_device *adev,
+		enum amdgpu_ras_block block);
 int amdgpu_ras_reset_error_status(struct amdgpu_device *adev,
 		enum amdgpu_ras_block block);
 
@@ -671,7 +940,7 @@ static inline void amdgpu_ras_intr_cleared(void)
 	atomic_set(&amdgpu_ras_in_intr, 0);
 }
 
-void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev);
+int amdgpu_ras_global_ras_isr(struct amdgpu_device *adev);
 
 void amdgpu_ras_set_error_query_ready(struct amdgpu_device *adev, bool ready);
 
@@ -693,7 +962,86 @@ struct amdgpu_ras* amdgpu_ras_get_context(struct amdgpu_device *adev);
 
 int amdgpu_ras_set_context(struct amdgpu_device *adev, struct amdgpu_ras *ras_con);
 
+int amdgpu_ras_set_mca_debug_mode(struct amdgpu_device *adev, bool enable);
+int amdgpu_ras_set_aca_debug_mode(struct amdgpu_device *adev, bool enable);
+bool amdgpu_ras_get_aca_debug_mode(struct amdgpu_device *adev);
+bool amdgpu_ras_get_error_query_mode(struct amdgpu_device *adev,
+				     unsigned int *mode);
+
 int amdgpu_ras_register_ras_block(struct amdgpu_device *adev,
 				struct amdgpu_ras_block_object *ras_block_obj);
 void amdgpu_ras_interrupt_fatal_error_handler(struct amdgpu_device *adev);
+void amdgpu_ras_get_error_type_name(uint32_t err_type, char *err_type_name);
+bool amdgpu_ras_inst_get_memory_id_field(struct amdgpu_device *adev,
+					 const struct amdgpu_ras_err_status_reg_entry *reg_entry,
+					 uint32_t instance,
+					 uint32_t *memory_id);
+bool amdgpu_ras_inst_get_err_cnt_field(struct amdgpu_device *adev,
+				       const struct amdgpu_ras_err_status_reg_entry *reg_entry,
+				       uint32_t instance,
+				       unsigned long *err_cnt);
+void amdgpu_ras_inst_query_ras_error_count(struct amdgpu_device *adev,
+					   const struct amdgpu_ras_err_status_reg_entry *reg_list,
+					   uint32_t reg_list_size,
+					   const struct amdgpu_ras_memory_id_entry *mem_list,
+					   uint32_t mem_list_size,
+					   uint32_t instance,
+					   uint32_t err_type,
+					   unsigned long *err_count);
+void amdgpu_ras_inst_reset_ras_error_count(struct amdgpu_device *adev,
+					   const struct amdgpu_ras_err_status_reg_entry *reg_list,
+					   uint32_t reg_list_size,
+					   uint32_t instance);
+
+int amdgpu_ras_error_data_init(struct ras_err_data *err_data);
+void amdgpu_ras_error_data_fini(struct ras_err_data *err_data);
+int amdgpu_ras_error_statistic_ce_count(struct ras_err_data *err_data,
+					struct amdgpu_smuio_mcm_config_info *mcm_info,
+					u64 count);
+int amdgpu_ras_error_statistic_ue_count(struct ras_err_data *err_data,
+					struct amdgpu_smuio_mcm_config_info *mcm_info,
+					u64 count);
+int amdgpu_ras_error_statistic_de_count(struct ras_err_data *err_data,
+					struct amdgpu_smuio_mcm_config_info *mcm_info,
+					u64 count);
+void amdgpu_ras_query_boot_status(struct amdgpu_device *adev, u32 num_instances);
+int amdgpu_ras_bind_aca(struct amdgpu_device *adev, enum amdgpu_ras_block blk,
+			       const struct aca_info *aca_info, void *data);
+int amdgpu_ras_unbind_aca(struct amdgpu_device *adev, enum amdgpu_ras_block blk);
+
+ssize_t amdgpu_ras_aca_sysfs_read(struct device *dev, struct device_attribute *attr,
+				  struct aca_handle *handle, char *buf, void *data);
+
+void amdgpu_ras_set_fed(struct amdgpu_device *adev, bool status);
+bool amdgpu_ras_get_fed_status(struct amdgpu_device *adev);
+void amdgpu_ras_set_err_poison(struct amdgpu_device *adev,
+			       enum amdgpu_ras_block block);
+void amdgpu_ras_clear_err_state(struct amdgpu_device *adev);
+bool amdgpu_ras_is_err_state(struct amdgpu_device *adev, int block);
+
+u64 amdgpu_ras_acquire_event_id(struct amdgpu_device *adev, enum ras_event_type type);
+int amdgpu_ras_mark_ras_event_caller(struct amdgpu_device *adev, enum ras_event_type type,
+				     const void *caller);
+
+int amdgpu_ras_reserve_page(struct amdgpu_device *adev, uint64_t pfn);
+
+int amdgpu_ras_add_critical_region(struct amdgpu_device *adev, struct amdgpu_bo *bo);
+bool amdgpu_ras_check_critical_address(struct amdgpu_device *adev, uint64_t addr);
+
+int amdgpu_ras_put_poison_req(struct amdgpu_device *adev,
+		enum amdgpu_ras_block block, uint16_t pasid,
+		pasid_notify pasid_fn, void *data, uint32_t reset);
+
+bool amdgpu_ras_in_recovery(struct amdgpu_device *adev);
+
+__printf(3, 4)
+void amdgpu_ras_event_log_print(struct amdgpu_device *adev, u64 event_id,
+				const char *fmt, ...);
+
+bool amdgpu_ras_is_rma(struct amdgpu_device *adev);
+
+void amdgpu_ras_pre_reset(struct amdgpu_device *adev,
+					  struct list_head *device_list);
+void amdgpu_ras_post_reset(struct amdgpu_device *adev,
+					  struct list_head *device_list);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c
index c2c2a7718613..64dd7a81bff5 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c
@@ -32,6 +32,7 @@
 #include <linux/uaccess.h>
 
 #include "amdgpu_reset.h"
+#include "amdgpu_ras_mgr.h"
 
 /* These are memory addresses as would be seen by one or more EEPROM
  * chips strung on the I2C bus, usually by manipulating pins 1-3 of a
@@ -58,7 +59,7 @@
 #define EEPROM_I2C_MADDR_4      0x40000
 
 /*
- * The 2 macros bellow represent the actual size in bytes that
+ * The 2 macros below represent the actual size in bytes that
  * those entities occupy in the EEPROM memory.
  * RAS_TABLE_RECORD_SIZE is different than sizeof(eeprom_table_record) which
  * uses uint64 to store 6b fields such as retired_page.
@@ -68,11 +69,24 @@
 
 /* Table hdr is 'AMDR' */
 #define RAS_TABLE_HDR_VAL       0x414d4452
-#define RAS_TABLE_VER           0x00010000
 
 /* Bad GPU tag ‘BADG’ */
 #define RAS_TABLE_HDR_BAD       0x42414447
 
+/*
+ * EEPROM Table structure v1
+ * ---------------------------------
+ * |                               |
+ * |     EEPROM TABLE HEADER       |
+ * |      ( size 20 Bytes )        |
+ * |                               |
+ * ---------------------------------
+ * |                               |
+ * |    BAD PAGE RECORD AREA       |
+ * |                               |
+ * ---------------------------------
+ */
+
 /* Assume 2-Mbit size EEPROM and take up the whole space. */
 #define RAS_TBL_SIZE_BYTES      (256 * 1024)
 #define RAS_TABLE_START         0
@@ -81,6 +95,37 @@
 #define RAS_MAX_RECORD_COUNT    ((RAS_TBL_SIZE_BYTES - RAS_TABLE_HEADER_SIZE) \
 				 / RAS_TABLE_RECORD_SIZE)
 
+/*
+ * EEPROM Table structrue v2.1
+ * ---------------------------------
+ * |                               |
+ * |     EEPROM TABLE HEADER       |
+ * |      ( size 20 Bytes )        |
+ * |                               |
+ * ---------------------------------
+ * |                               |
+ * |     EEPROM TABLE RAS INFO     |
+ * | (available info size 4 Bytes) |
+ * |  ( reserved size 252 Bytes )  |
+ * |                               |
+ * ---------------------------------
+ * |                               |
+ * |     BAD PAGE RECORD AREA      |
+ * |                               |
+ * ---------------------------------
+ */
+
+/* EEPROM Table V2_1 */
+#define RAS_TABLE_V2_1_INFO_SIZE       256
+#define RAS_TABLE_V2_1_INFO_START      RAS_TABLE_HEADER_SIZE
+#define RAS_RECORD_START_V2_1          (RAS_HDR_START + RAS_TABLE_HEADER_SIZE + \
+					RAS_TABLE_V2_1_INFO_SIZE)
+#define RAS_MAX_RECORD_COUNT_V2_1      ((RAS_TBL_SIZE_BYTES - RAS_TABLE_HEADER_SIZE - \
+					RAS_TABLE_V2_1_INFO_SIZE) \
+					/ RAS_TABLE_RECORD_SIZE)
+
+#define RAS_SMU_MESSAGE_TIMEOUT_MS 1000 /* 1s */
+
 /* Given a zero-based index of an EEPROM RAS record, yields the EEPROM
  * offset off of RAS_TABLE_START.  That is, this is something you can
  * add to control->i2c_address, and then tell I2C layer to read
@@ -103,17 +148,25 @@
 #define RAS_NUM_RECS(_tbl_hdr)  (((_tbl_hdr)->tbl_size - \
 				  RAS_TABLE_HEADER_SIZE) / RAS_TABLE_RECORD_SIZE)
 
-#define to_amdgpu_device(x) (container_of(x, struct amdgpu_ras, eeprom_control))->adev
+#define RAS_NUM_RECS_V2_1(_tbl_hdr)  (((_tbl_hdr)->tbl_size - \
+				       RAS_TABLE_HEADER_SIZE - \
+				       RAS_TABLE_V2_1_INFO_SIZE) / RAS_TABLE_RECORD_SIZE)
+
+#define to_amdgpu_device(x) ((container_of(x, struct amdgpu_ras, eeprom_control))->adev)
 
 static bool __is_ras_eeprom_supported(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[MP1_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
 	case IP_VERSION(11, 0, 2): /* VEGA20 and ARCTURUS */
 	case IP_VERSION(11, 0, 7): /* Sienna cichlid */
 	case IP_VERSION(13, 0, 0):
 	case IP_VERSION(13, 0, 2): /* Aldebaran */
 	case IP_VERSION(13, 0, 10):
 		return true;
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 12):
+	case IP_VERSION(13, 0, 14):
+		return (adev->gmc.is_app_apu) ? false : true;
 	default:
 		return false;
 	}
@@ -128,7 +181,7 @@ static bool __get_eeprom_i2c_addr(struct amdgpu_device *adev,
 	if (!control)
 		return false;
 
-	if (amdgpu_atomfirmware_ras_rom_addr(adev, &i2c_addr)) {
+	if (adev->bios && amdgpu_atomfirmware_ras_rom_addr(adev, &i2c_addr)) {
 		/* The address given by VBIOS is an 8-bit, wire-format
 		 * address, i.e. the most significant byte.
 		 *
@@ -143,14 +196,14 @@ static bool __get_eeprom_i2c_addr(struct amdgpu_device *adev,
 		return true;
 	}
 
-	switch (adev->ip_versions[MP1_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
 	case IP_VERSION(11, 0, 2):
 		/* VEGA20 and ARCTURUS */
 		if (adev->asic_type == CHIP_VEGA20)
 			control->i2c_address = EEPROM_I2C_MADDR_0;
-		else if (strnstr(atom_ctx->vbios_version,
+		else if (strnstr(atom_ctx->vbios_pn,
 				 "D342",
-				 sizeof(atom_ctx->vbios_version)))
+				 sizeof(atom_ctx->vbios_pn)))
 			control->i2c_address = EEPROM_I2C_MADDR_0;
 		else
 			control->i2c_address = EEPROM_I2C_MADDR_4;
@@ -159,14 +212,23 @@ static bool __get_eeprom_i2c_addr(struct amdgpu_device *adev,
 		control->i2c_address = EEPROM_I2C_MADDR_0;
 		return true;
 	case IP_VERSION(13, 0, 2):
-		if (strnstr(atom_ctx->vbios_version, "D673",
-			    sizeof(atom_ctx->vbios_version)))
+		if (strnstr(atom_ctx->vbios_pn, "D673",
+			    sizeof(atom_ctx->vbios_pn)))
 			control->i2c_address = EEPROM_I2C_MADDR_4;
 		else
 			control->i2c_address = EEPROM_I2C_MADDR_0;
 		return true;
 	case IP_VERSION(13, 0, 0):
+		if (strnstr(atom_ctx->vbios_pn, "D707",
+			    sizeof(atom_ctx->vbios_pn)))
+			control->i2c_address = EEPROM_I2C_MADDR_0;
+		else
+			control->i2c_address = EEPROM_I2C_MADDR_4;
+		return true;
+	case IP_VERSION(13, 0, 6):
 	case IP_VERSION(13, 0, 10):
+	case IP_VERSION(13, 0, 12):
+	case IP_VERSION(13, 0, 14):
 		control->i2c_address = EEPROM_I2C_MADDR_4;
 		return true;
 	default:
@@ -218,10 +280,11 @@ static int __write_table_header(struct amdgpu_ras_eeprom_control *control)
 	up_read(&adev->reset_domain->sem);
 
 	if (res < 0) {
-		DRM_ERROR("Failed to write EEPROM table header:%d", res);
+		dev_err(adev->dev, "Failed to write EEPROM table header:%d",
+			res);
 	} else if (res < RAS_TABLE_HEADER_SIZE) {
-		DRM_ERROR("Short write:%d out of %d\n",
-			  res, RAS_TABLE_HEADER_SIZE);
+		dev_err(adev->dev, "Short write:%d out of %d\n", res,
+			RAS_TABLE_HEADER_SIZE);
 		res = -EIO;
 	} else {
 		res = 0;
@@ -230,6 +293,71 @@ static int __write_table_header(struct amdgpu_ras_eeprom_control *control)
 	return res;
 }
 
+static void
+__encode_table_ras_info_to_buf(struct amdgpu_ras_eeprom_table_ras_info *rai,
+			       unsigned char *buf)
+{
+	u32 *pp = (uint32_t *)buf;
+	u32 tmp;
+
+	tmp = ((uint32_t)(rai->rma_status) & 0xFF) |
+	      (((uint32_t)(rai->health_percent) << 8) & 0xFF00) |
+	      (((uint32_t)(rai->ecc_page_threshold) << 16) & 0xFFFF0000);
+	pp[0] = cpu_to_le32(tmp);
+}
+
+static void
+__decode_table_ras_info_from_buf(struct amdgpu_ras_eeprom_table_ras_info *rai,
+				 unsigned char *buf)
+{
+	u32 *pp = (uint32_t *)buf;
+	u32 tmp;
+
+	tmp = le32_to_cpu(pp[0]);
+	rai->rma_status = tmp & 0xFF;
+	rai->health_percent = (tmp >> 8) & 0xFF;
+	rai->ecc_page_threshold = (tmp >> 16) & 0xFFFF;
+}
+
+static int __write_table_ras_info(struct amdgpu_ras_eeprom_control *control)
+{
+	struct amdgpu_device *adev = to_amdgpu_device(control);
+	u8 *buf;
+	int res;
+
+	buf = kzalloc(RAS_TABLE_V2_1_INFO_SIZE, GFP_KERNEL);
+	if (!buf) {
+		dev_err(adev->dev,
+			"Failed to alloc buf to write table ras info\n");
+		return -ENOMEM;
+	}
+
+	__encode_table_ras_info_to_buf(&control->tbl_rai, buf);
+
+	/* i2c may be unstable in gpu reset */
+	down_read(&adev->reset_domain->sem);
+	res = amdgpu_eeprom_write(adev->pm.ras_eeprom_i2c_bus,
+				  control->i2c_address +
+				  control->ras_info_offset,
+				  buf, RAS_TABLE_V2_1_INFO_SIZE);
+	up_read(&adev->reset_domain->sem);
+
+	if (res < 0) {
+		dev_err(adev->dev, "Failed to write EEPROM table ras info:%d",
+			res);
+	} else if (res < RAS_TABLE_V2_1_INFO_SIZE) {
+		dev_err(adev->dev, "Short write:%d out of %d\n", res,
+			RAS_TABLE_V2_1_INFO_SIZE);
+		res = -EIO;
+	} else {
+		res = 0;
+	}
+
+	kfree(buf);
+
+	return res;
+}
+
 static u8 __calc_hdr_byte_sum(const struct amdgpu_ras_eeprom_control *control)
 {
 	int ii;
@@ -246,6 +374,21 @@ static u8 __calc_hdr_byte_sum(const struct amdgpu_ras_eeprom_control *control)
 	return csum;
 }
 
+static u8 __calc_ras_info_byte_sum(const struct amdgpu_ras_eeprom_control *control)
+{
+	int ii;
+	u8  *pp, csum;
+	size_t sz;
+
+	sz = sizeof(control->tbl_rai);
+	pp = (u8 *) &control->tbl_rai;
+	csum = 0;
+	for (ii = 0; ii < sz; ii++, pp++)
+		csum += *pp;
+
+	return csum;
+}
+
 static int amdgpu_ras_eeprom_correct_header_tag(
 	struct amdgpu_ras_eeprom_control *control,
 	uint32_t header)
@@ -271,6 +414,25 @@ static int amdgpu_ras_eeprom_correct_header_tag(
 	return res;
 }
 
+static void amdgpu_ras_set_eeprom_table_version(struct amdgpu_ras_eeprom_control *control)
+{
+	struct amdgpu_device *adev = to_amdgpu_device(control);
+	struct amdgpu_ras_eeprom_table_header *hdr = &control->tbl_hdr;
+
+	switch (amdgpu_ip_version(adev, UMC_HWIP, 0)) {
+	case IP_VERSION(8, 10, 0):
+		hdr->version = RAS_TABLE_VER_V2_1;
+		return;
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 5, 0):
+		hdr->version = RAS_TABLE_VER_V3;
+		return;
+	default:
+		hdr->version = RAS_TABLE_VER_V1;
+		return;
+	}
+}
+
 /**
  * amdgpu_ras_eeprom_reset_table -- Reset the RAS EEPROM table
  * @control: pointer to control structure
@@ -282,26 +444,66 @@ int amdgpu_ras_eeprom_reset_table(struct amdgpu_ras_eeprom_control *control)
 {
 	struct amdgpu_device *adev = to_amdgpu_device(control);
 	struct amdgpu_ras_eeprom_table_header *hdr = &control->tbl_hdr;
+	struct amdgpu_ras_eeprom_table_ras_info *rai = &control->tbl_rai;
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	u32 erase_res = 0;
 	u8 csum;
 	int res;
 
 	mutex_lock(&control->ras_tbl_mutex);
 
-	hdr->header = RAS_TABLE_HDR_VAL;
-	hdr->version = RAS_TABLE_VER;
-	hdr->first_rec_offset = RAS_RECORD_START;
-	hdr->tbl_size = RAS_TABLE_HEADER_SIZE;
+	if (!amdgpu_ras_smu_eeprom_supported(adev)) {
+		hdr->header = RAS_TABLE_HDR_VAL;
+		amdgpu_ras_set_eeprom_table_version(control);
+
+		if (hdr->version >= RAS_TABLE_VER_V2_1) {
+			hdr->first_rec_offset = RAS_RECORD_START_V2_1;
+			hdr->tbl_size = RAS_TABLE_HEADER_SIZE +
+					RAS_TABLE_V2_1_INFO_SIZE;
+			rai->rma_status = GPU_HEALTH_USABLE;
+
+			control->ras_record_offset = RAS_RECORD_START_V2_1;
+			control->ras_max_record_count = RAS_MAX_RECORD_COUNT_V2_1;
+			/**
+			 * GPU health represented as a percentage.
+			 * 0 means worst health, 100 means fully health.
+			 */
+			rai->health_percent = 100;
+			/* ecc_page_threshold = 0 means disable bad page retirement */
+			rai->ecc_page_threshold = con->bad_page_cnt_threshold;
+		} else {
+			hdr->first_rec_offset = RAS_RECORD_START;
+			hdr->tbl_size = RAS_TABLE_HEADER_SIZE;
 
-	csum = __calc_hdr_byte_sum(control);
-	csum = -csum;
-	hdr->checksum = csum;
-	res = __write_table_header(control);
+			control->ras_record_offset = RAS_RECORD_START;
+			control->ras_max_record_count = RAS_MAX_RECORD_COUNT;
+		}
+
+		csum = __calc_hdr_byte_sum(control);
+		if (hdr->version >= RAS_TABLE_VER_V2_1)
+			csum += __calc_ras_info_byte_sum(control);
+		csum = -csum;
+		hdr->checksum = csum;
+		res = __write_table_header(control);
+		if (!res && hdr->version > RAS_TABLE_VER_V1)
+			res = __write_table_ras_info(control);
+	} else {
+		res = amdgpu_ras_smu_erase_ras_table(adev, &erase_res);
+		if (res || erase_res) {
+			dev_warn(adev->dev, "RAS EEPROM reset failed, res:%d result:%d",
+										res, erase_res);
+			if (!res)
+				res = -EIO;
+		}
+	}
 
 	control->ras_num_recs = 0;
+	control->ras_num_bad_pages = 0;
+	control->ras_num_mca_recs = 0;
+	control->ras_num_pa_recs = 0;
 	control->ras_fri = 0;
 
-	amdgpu_dpm_send_hbm_bad_pages_num(adev, control->ras_num_recs);
+	amdgpu_dpm_send_hbm_bad_pages_num(adev, control->ras_num_bad_pages);
 
 	control->bad_channel_bitmap = 0;
 	amdgpu_dpm_send_hbm_bad_channel_flag(adev, control->bad_channel_bitmap);
@@ -374,6 +576,9 @@ bool amdgpu_ras_eeprom_check_err_threshold(struct amdgpu_device *adev)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 
+	if (amdgpu_uniras_enabled(adev))
+		return amdgpu_ras_mgr_check_eeprom_safety_watermark(adev);
+
 	if (!__is_ras_eeprom_supported(adev) ||
 	    !amdgpu_bad_page_threshold)
 		return false;
@@ -386,16 +591,17 @@ bool amdgpu_ras_eeprom_check_err_threshold(struct amdgpu_device *adev)
 			return false;
 
 	if (con->eeprom_control.tbl_hdr.header == RAS_TABLE_HDR_BAD) {
-		if (amdgpu_bad_page_threshold == -1) {
+		if (con->eeprom_control.ras_num_bad_pages > con->bad_page_cnt_threshold)
 			dev_warn(adev->dev, "RAS records:%d exceed threshold:%d",
-				con->eeprom_control.ras_num_recs, con->bad_page_cnt_threshold);
+				 con->eeprom_control.ras_num_bad_pages, con->bad_page_cnt_threshold);
+		if ((amdgpu_bad_page_threshold == -1) ||
+		    (amdgpu_bad_page_threshold == -2)) {
 			dev_warn(adev->dev,
-				"But GPU can be operated due to bad_page_threshold = -1.\n");
+				 "Please consult AMD Service Action Guide (SAG) for appropriate service procedures.\n");
 			return false;
 		} else {
-			dev_warn(adev->dev, "This GPU is in BAD status.");
-			dev_warn(adev->dev, "Please retire it or set a larger "
-				 "threshold value when reloading driver.\n");
+			dev_warn(adev->dev,
+				 "Please consider adjusting the customized threshold.\n");
 			return true;
 		}
 	}
@@ -429,13 +635,13 @@ static int __amdgpu_ras_eeprom_write(struct amdgpu_ras_eeprom_control *control,
 				  buf, buf_size);
 	up_read(&adev->reset_domain->sem);
 	if (res < 0) {
-		DRM_ERROR("Writing %d EEPROM table records error:%d",
-			  num, res);
+		dev_err(adev->dev, "Writing %d EEPROM table records error:%d",
+			num, res);
 	} else if (res < buf_size) {
 		/* Short write, return error.
 		 */
-		DRM_ERROR("Wrote %d records out of %d",
-			  res / RAS_TABLE_RECORD_SIZE, num);
+		dev_err(adev->dev, "Wrote %d records out of %d",
+			res / RAS_TABLE_RECORD_SIZE, num);
 		res = -EIO;
 	} else {
 		res = 0;
@@ -450,6 +656,7 @@ amdgpu_ras_eeprom_append_table(struct amdgpu_ras_eeprom_control *control,
 			       const u32 num)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(to_amdgpu_device(control));
+	struct amdgpu_device *adev = to_amdgpu_device(control);
 	u32 a, b, i;
 	u8 *buf, *pp;
 	int res;
@@ -465,7 +672,8 @@ amdgpu_ras_eeprom_append_table(struct amdgpu_ras_eeprom_control *control,
 		__encode_table_record_to_buf(control, &record[i], pp);
 
 		/* update bad channel bitmap */
-		if (!(control->bad_channel_bitmap & (1 << record[i].mem_channel))) {
+		if ((record[i].mem_channel < BITS_PER_TYPE(control->bad_channel_bitmap)) &&
+		    !(control->bad_channel_bitmap & (1 << record[i].mem_channel))) {
 			control->bad_channel_bitmap |= 1 << record[i].mem_channel;
 			con->update_channel_flag = true;
 		}
@@ -551,6 +759,14 @@ amdgpu_ras_eeprom_append_table(struct amdgpu_ras_eeprom_control *control,
 	control->ras_num_recs = 1 + (control->ras_max_record_count + b
 				     - control->ras_fri)
 		% control->ras_max_record_count;
+
+	/*old asics only save pa to eeprom like before*/
+	if (IP_VERSION_MAJ(amdgpu_ip_version(adev, UMC_HWIP, 0)) < 12)
+		control->ras_num_pa_recs += num;
+	else
+		control->ras_num_mca_recs += num;
+
+	control->ras_num_bad_pages = con->bad_page_num;
 Out:
 	kfree(buf);
 	return res;
@@ -568,23 +784,44 @@ amdgpu_ras_eeprom_update_header(struct amdgpu_ras_eeprom_control *control)
 	/* Modify the header if it exceeds.
 	 */
 	if (amdgpu_bad_page_threshold != 0 &&
-	    control->ras_num_recs >= ras->bad_page_cnt_threshold) {
+	    control->ras_num_bad_pages > ras->bad_page_cnt_threshold) {
 		dev_warn(adev->dev,
 			"Saved bad pages %d reaches threshold value %d\n",
-			control->ras_num_recs, ras->bad_page_cnt_threshold);
-		control->tbl_hdr.header = RAS_TABLE_HDR_BAD;
+			control->ras_num_bad_pages, ras->bad_page_cnt_threshold);
+
+		if (adev->cper.enabled && !amdgpu_uniras_enabled(adev) &&
+		    amdgpu_cper_generate_bp_threshold_record(adev))
+			dev_warn(adev->dev, "fail to generate bad page threshold cper records\n");
+
+		if ((amdgpu_bad_page_threshold != -1) &&
+		    (amdgpu_bad_page_threshold != -2)) {
+			control->tbl_hdr.header = RAS_TABLE_HDR_BAD;
+			if (control->tbl_hdr.version >= RAS_TABLE_VER_V2_1) {
+				control->tbl_rai.rma_status = GPU_RETIRED__ECC_REACH_THRESHOLD;
+				control->tbl_rai.health_percent = 0;
+			}
+			ras->is_rma = true;
+		}
+
+		/* ignore the -ENOTSUPP return value */
+		amdgpu_dpm_send_rma_reason(adev);
 	}
 
-	control->tbl_hdr.version = RAS_TABLE_VER;
-	control->tbl_hdr.first_rec_offset = RAS_INDEX_TO_OFFSET(control, control->ras_fri);
-	control->tbl_hdr.tbl_size = RAS_TABLE_HEADER_SIZE + control->ras_num_recs * RAS_TABLE_RECORD_SIZE;
+	if (control->tbl_hdr.version >= RAS_TABLE_VER_V2_1)
+		control->tbl_hdr.tbl_size = RAS_TABLE_HEADER_SIZE +
+					    RAS_TABLE_V2_1_INFO_SIZE +
+					    control->ras_num_recs * RAS_TABLE_RECORD_SIZE;
+	else
+		control->tbl_hdr.tbl_size = RAS_TABLE_HEADER_SIZE +
+					    control->ras_num_recs * RAS_TABLE_RECORD_SIZE;
 	control->tbl_hdr.checksum = 0;
 
 	buf_size = control->ras_num_recs * RAS_TABLE_RECORD_SIZE;
 	buf = kcalloc(control->ras_num_recs, RAS_TABLE_RECORD_SIZE, GFP_KERNEL);
 	if (!buf) {
-		DRM_ERROR("allocating memory for table of size %d bytes failed\n",
-			  control->tbl_hdr.tbl_size);
+		dev_err(adev->dev,
+			"allocating memory for table of size %d bytes failed\n",
+			control->tbl_hdr.tbl_size);
 		res = -ENOMEM;
 		goto Out;
 	}
@@ -596,16 +833,26 @@ amdgpu_ras_eeprom_update_header(struct amdgpu_ras_eeprom_control *control)
 				 buf, buf_size);
 	up_read(&adev->reset_domain->sem);
 	if (res < 0) {
-		DRM_ERROR("EEPROM failed reading records:%d\n",
-			  res);
+		dev_err(adev->dev, "EEPROM failed reading records:%d\n", res);
 		goto Out;
 	} else if (res < buf_size) {
-		DRM_ERROR("EEPROM read %d out of %d bytes\n",
-			  res, buf_size);
+		dev_err(adev->dev, "EEPROM read %d out of %d bytes\n", res,
+			buf_size);
 		res = -EIO;
 		goto Out;
 	}
 
+	/**
+	 * bad page records have been stored in eeprom,
+	 * now calculate gpu health percent
+	 */
+	if (amdgpu_bad_page_threshold != 0 &&
+	    control->tbl_hdr.version >= RAS_TABLE_VER_V2_1 &&
+	    control->ras_num_bad_pages <= ras->bad_page_cnt_threshold)
+		control->tbl_rai.health_percent = ((ras->bad_page_cnt_threshold -
+						   control->ras_num_bad_pages) * 100) /
+						   ras->bad_page_cnt_threshold;
+
 	/* Recalc the checksum.
 	 */
 	csum = 0;
@@ -613,15 +860,84 @@ amdgpu_ras_eeprom_update_header(struct amdgpu_ras_eeprom_control *control)
 		csum += *pp;
 
 	csum += __calc_hdr_byte_sum(control);
+	if (control->tbl_hdr.version >= RAS_TABLE_VER_V2_1)
+		csum += __calc_ras_info_byte_sum(control);
 	/* avoid sign extension when assigning to "checksum" */
 	csum = -csum;
 	control->tbl_hdr.checksum = csum;
 	res = __write_table_header(control);
+	if (!res && control->tbl_hdr.version > RAS_TABLE_VER_V1)
+		res = __write_table_ras_info(control);
 Out:
 	kfree(buf);
 	return res;
 }
 
+int amdgpu_ras_eeprom_update_record_num(struct amdgpu_ras_eeprom_control *control)
+{
+	struct amdgpu_device *adev = to_amdgpu_device(control);
+	int ret, retry = 20;
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev))
+		return 0;
+
+	control->ras_num_recs_old = control->ras_num_recs;
+
+	do {
+		/* 1000ms timeout is long enough, smu_get_badpage_count won't
+		 * return -EBUSY before timeout.
+		 */
+		ret = amdgpu_ras_smu_get_badpage_count(adev,
+			&(control->ras_num_recs), RAS_SMU_MESSAGE_TIMEOUT_MS);
+		if (!ret &&
+		    (control->ras_num_recs_old == control->ras_num_recs)) {
+			/* record number update in PMFW needs some time,
+			 * smu_get_badpage_count may return immediately without
+			 * count update, sleep for a while and retry again.
+			 */
+			msleep(50);
+			retry--;
+		} else {
+			break;
+		}
+	} while (retry);
+
+	/* no update of record number is not a real failure,
+	 * don't print warning here
+	 */
+	if (!ret && (control->ras_num_recs_old == control->ras_num_recs))
+		ret = -EINVAL;
+
+	return ret;
+}
+
+static int amdgpu_ras_smu_eeprom_append(struct amdgpu_ras_eeprom_control *control)
+{
+	struct amdgpu_device *adev = to_amdgpu_device(control);
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev) || !con)
+		return 0;
+
+	control->ras_num_bad_pages = con->bad_page_num;
+
+	if (amdgpu_bad_page_threshold != 0 &&
+	    control->ras_num_bad_pages > con->bad_page_cnt_threshold) {
+		dev_warn(adev->dev,
+			"Saved bad pages %d reaches threshold value %d\n",
+			control->ras_num_bad_pages, con->bad_page_cnt_threshold);
+
+		if (adev->cper.enabled && amdgpu_cper_generate_bp_threshold_record(adev))
+			dev_warn(adev->dev, "fail to generate bad page threshold cper records\n");
+
+		if ((amdgpu_bad_page_threshold != -1) &&
+		    (amdgpu_bad_page_threshold != -2))
+			con->is_rma = true;
+	}
+
+	return 0;
+}
+
 /**
  * amdgpu_ras_eeprom_append -- append records to the EEPROM RAS table
  * @control: pointer to control structure
@@ -640,20 +956,32 @@ int amdgpu_ras_eeprom_append(struct amdgpu_ras_eeprom_control *control,
 			     const u32 num)
 {
 	struct amdgpu_device *adev = to_amdgpu_device(control);
-	int res;
+	int res, i;
+	uint64_t nps = AMDGPU_NPS1_PARTITION_MODE;
 
 	if (!__is_ras_eeprom_supported(adev))
 		return 0;
 
+	if (amdgpu_ras_smu_eeprom_supported(adev))
+		return amdgpu_ras_smu_eeprom_append(control);
+
 	if (num == 0) {
-		DRM_ERROR("will not append 0 records\n");
+		dev_err(adev->dev, "will not append 0 records\n");
 		return -EINVAL;
 	} else if (num > control->ras_max_record_count) {
-		DRM_ERROR("cannot append %d records than the size of table %d\n",
-			  num, control->ras_max_record_count);
+		dev_err(adev->dev,
+			"cannot append %d records than the size of table %d\n",
+			num, control->ras_max_record_count);
 		return -EINVAL;
 	}
 
+	if (adev->gmc.gmc_funcs->query_mem_partition_mode)
+		nps = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
+
+	/* set the new channel index flag */
+	for (i = 0; i < num; i++)
+		record[i].retired_page |= (nps << UMC_NPS_SHIFT);
+
 	mutex_lock(&control->ras_tbl_mutex);
 
 	res = amdgpu_ras_eeprom_append_table(control, record, num);
@@ -663,6 +991,11 @@ int amdgpu_ras_eeprom_append(struct amdgpu_ras_eeprom_control *control,
 		amdgpu_ras_debugfs_set_ret_size(control);
 
 	mutex_unlock(&control->ras_tbl_mutex);
+
+	/* clear channel index flag, the flag is only saved on eeprom */
+	for (i = 0; i < num; i++)
+		record[i].retired_page &= ~(nps << UMC_NPS_SHIFT);
+
 	return res;
 }
 
@@ -692,13 +1025,13 @@ static int __amdgpu_ras_eeprom_read(struct amdgpu_ras_eeprom_control *control,
 				 buf, buf_size);
 	up_read(&adev->reset_domain->sem);
 	if (res < 0) {
-		DRM_ERROR("Reading %d EEPROM table records error:%d",
-			  num, res);
+		dev_err(adev->dev, "Reading %d EEPROM table records error:%d",
+			num, res);
 	} else if (res < buf_size) {
 		/* Short read, return error.
 		 */
-		DRM_ERROR("Read %d records out of %d",
-			  res / RAS_TABLE_RECORD_SIZE, num);
+		dev_err(adev->dev, "Read %d records out of %d",
+			res / RAS_TABLE_RECORD_SIZE, num);
 		res = -EIO;
 	} else {
 		res = 0;
@@ -707,6 +1040,50 @@ static int __amdgpu_ras_eeprom_read(struct amdgpu_ras_eeprom_control *control,
 	return res;
 }
 
+int amdgpu_ras_eeprom_read_idx(struct amdgpu_ras_eeprom_control *control,
+			struct eeprom_table_record *record, u32 rec_idx,
+			const u32 num)
+{
+	struct amdgpu_device *adev = to_amdgpu_device(control);
+	uint64_t ts, end_idx;
+	int i, ret;
+	u64 mca, ipid;
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev))
+		return 0;
+
+	if (!adev->umc.ras || !adev->umc.ras->mca_ipid_parse)
+		return -EOPNOTSUPP;
+
+	end_idx = rec_idx + num;
+	for (i = rec_idx; i < end_idx; i++) {
+		ret = amdgpu_ras_smu_get_badpage_mca_addr(adev, i, &mca);
+		if (ret)
+			return ret;
+
+		ret = amdgpu_ras_smu_get_badpage_ipid(adev, i, &ipid);
+		if (ret)
+			return ret;
+
+		ret = amdgpu_ras_smu_get_timestamp(adev, i, &ts);
+		if (ret)
+			return ret;
+
+		record[i - rec_idx].address = mca;
+		/* retired_page (pa) is unused now */
+		record[i - rec_idx].retired_page = 0x1ULL;
+		record[i - rec_idx].ts = ts;
+		record[i - rec_idx].err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE;
+
+		adev->umc.ras->mca_ipid_parse(adev, ipid,
+			(uint32_t *)&(record[i - rec_idx].cu),
+			(uint32_t *)&(record[i - rec_idx].mem_channel),
+			(uint32_t *)&(record[i - rec_idx].mcumc_id), NULL);
+	}
+
+	return 0;
+}
+
 /**
  * amdgpu_ras_eeprom_read -- read EEPROM
  * @control: pointer to control structure
@@ -728,15 +1105,18 @@ int amdgpu_ras_eeprom_read(struct amdgpu_ras_eeprom_control *control,
 	u8 *buf, *pp;
 	u32 g0, g1;
 
+	if (amdgpu_ras_smu_eeprom_supported(adev))
+		return amdgpu_ras_eeprom_read_idx(control, record, 0, num);
+
 	if (!__is_ras_eeprom_supported(adev))
 		return 0;
 
 	if (num == 0) {
-		DRM_ERROR("will not read 0 records\n");
+		dev_err(adev->dev, "will not read 0 records\n");
 		return -EINVAL;
 	} else if (num > control->ras_num_recs) {
-		DRM_ERROR("too many records to read:%d available:%d\n",
-			  num, control->ras_num_recs);
+		dev_err(adev->dev, "too many records to read:%d available:%d\n",
+			num, control->ras_num_recs);
 		return -EINVAL;
 	}
 
@@ -795,7 +1175,8 @@ int amdgpu_ras_eeprom_read(struct amdgpu_ras_eeprom_control *control,
 		__decode_table_record_from_buf(control, &record[i], pp);
 
 		/* update bad channel bitmap */
-		if (!(control->bad_channel_bitmap & (1 << record[i].mem_channel))) {
+		if ((record[i].mem_channel < BITS_PER_TYPE(control->bad_channel_bitmap)) &&
+		    !(control->bad_channel_bitmap & (1 << record[i].mem_channel))) {
 			control->bad_channel_bitmap |= 1 << record[i].mem_channel;
 			con->update_channel_flag = true;
 		}
@@ -807,9 +1188,15 @@ Out:
 	return res;
 }
 
-uint32_t amdgpu_ras_eeprom_max_record_count(void)
+uint32_t amdgpu_ras_eeprom_max_record_count(struct amdgpu_ras_eeprom_control *control)
 {
-	return RAS_MAX_RECORD_COUNT;
+	/* get available eeprom table version first before eeprom table init */
+	amdgpu_ras_set_eeprom_table_version(control);
+
+	if (control->tbl_hdr.version >= RAS_TABLE_VER_V2_1)
+		return RAS_MAX_RECORD_COUNT_V2_1;
+	else
+		return RAS_MAX_RECORD_COUNT;
 }
 
 static ssize_t
@@ -892,6 +1279,10 @@ static ssize_t amdgpu_ras_debugfs_table_read(struct file *f, char __user *buf,
 	int res = -EFAULT;
 	size_t data_len;
 
+	/* pmfw manages eeprom data by itself */
+	if (amdgpu_ras_smu_eeprom_supported(adev))
+		return 0;
+
 	mutex_lock(&control->ras_tbl_mutex);
 
 	/* We want *pos - data_len > 0, which means there's
@@ -1051,11 +1442,18 @@ static int __verify_ras_table_checksum(struct amdgpu_ras_eeprom_control *control
 	int buf_size, res;
 	u8  csum, *buf, *pp;
 
-	buf_size = RAS_TABLE_HEADER_SIZE +
-		control->ras_num_recs * RAS_TABLE_RECORD_SIZE;
+	if (control->tbl_hdr.version >= RAS_TABLE_VER_V2_1)
+		buf_size = RAS_TABLE_HEADER_SIZE +
+			   RAS_TABLE_V2_1_INFO_SIZE +
+			   control->ras_num_recs * RAS_TABLE_RECORD_SIZE;
+	else
+		buf_size = RAS_TABLE_HEADER_SIZE +
+			   control->ras_num_recs * RAS_TABLE_RECORD_SIZE;
+
 	buf = kzalloc(buf_size, GFP_KERNEL);
 	if (!buf) {
-		DRM_ERROR("Out of memory checking RAS table checksum.\n");
+		dev_err(adev->dev,
+			"Out of memory checking RAS table checksum.\n");
 		return -ENOMEM;
 	}
 
@@ -1064,7 +1462,7 @@ static int __verify_ras_table_checksum(struct amdgpu_ras_eeprom_control *control
 				 control->ras_header_offset,
 				 buf, buf_size);
 	if (res < buf_size) {
-		DRM_ERROR("Partial read for checksum, res:%d\n", res);
+		dev_err(adev->dev, "Partial read for checksum, res:%d\n", res);
 		/* On partial reads, return -EIO.
 		 */
 		if (res >= 0)
@@ -1080,8 +1478,78 @@ Out:
 	return res < 0 ? res : csum;
 }
 
-int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control,
-			   bool *exceed_err_limit)
+static int __read_table_ras_info(struct amdgpu_ras_eeprom_control *control)
+{
+	struct amdgpu_ras_eeprom_table_ras_info *rai = &control->tbl_rai;
+	struct amdgpu_device *adev = to_amdgpu_device(control);
+	unsigned char *buf;
+	int res;
+
+	buf = kzalloc(RAS_TABLE_V2_1_INFO_SIZE, GFP_KERNEL);
+	if (!buf) {
+		dev_err(adev->dev,
+			"Failed to alloc buf to read EEPROM table ras info\n");
+		return -ENOMEM;
+	}
+
+	/**
+	 * EEPROM table V2_1 supports ras info,
+	 * read EEPROM table ras info
+	 */
+	res = amdgpu_eeprom_read(adev->pm.ras_eeprom_i2c_bus,
+				 control->i2c_address + control->ras_info_offset,
+				 buf, RAS_TABLE_V2_1_INFO_SIZE);
+	if (res < RAS_TABLE_V2_1_INFO_SIZE) {
+		dev_err(adev->dev,
+			"Failed to read EEPROM table ras info, res:%d", res);
+		res = res >= 0 ? -EIO : res;
+		goto Out;
+	}
+
+	__decode_table_ras_info_from_buf(rai, buf);
+
+Out:
+	kfree(buf);
+	return res == RAS_TABLE_V2_1_INFO_SIZE ? 0 : res;
+}
+
+static int amdgpu_ras_smu_eeprom_init(struct amdgpu_ras_eeprom_control *control)
+{
+	struct amdgpu_device *adev = to_amdgpu_device(control);
+	struct amdgpu_ras_eeprom_table_header *hdr = &control->tbl_hdr;
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+	uint64_t local_time;
+	int res;
+
+	ras->is_rma = false;
+
+	if (!__is_ras_eeprom_supported(adev))
+		return 0;
+	mutex_init(&control->ras_tbl_mutex);
+
+	res = amdgpu_ras_smu_get_table_version(adev, &(hdr->version));
+	if (res)
+		return res;
+
+	res = amdgpu_ras_smu_get_badpage_count(adev,
+								&(control->ras_num_recs), 100);
+	if (res)
+		return res;
+
+	local_time = (uint64_t)ktime_get_real_seconds();
+	res = amdgpu_ras_smu_set_timestamp(adev, local_time);
+	if (res)
+		return res;
+
+	control->ras_max_record_count = 4000;
+
+	control->ras_num_mca_recs = 0;
+	control->ras_num_pa_recs = 0;
+
+	return 0;
+}
+
+int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control)
 {
 	struct amdgpu_device *adev = to_amdgpu_device(control);
 	unsigned char buf[RAS_TABLE_HEADER_SIZE] = { 0 };
@@ -1089,7 +1557,10 @@ int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control,
 	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 	int res;
 
-	*exceed_err_limit = false;
+	if (amdgpu_ras_smu_eeprom_supported(adev))
+		return amdgpu_ras_smu_eeprom_init(control);
+
+	ras->is_rma = false;
 
 	if (!__is_ras_eeprom_supported(adev))
 		return 0;
@@ -1102,8 +1573,7 @@ int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control,
 		return -EINVAL;
 
 	control->ras_header_offset = RAS_HDR_START;
-	control->ras_record_offset = RAS_RECORD_START;
-	control->ras_max_record_count  = RAS_MAX_RECORD_COUNT;
+	control->ras_info_offset = RAS_TABLE_V2_1_INFO_START;
 	mutex_init(&control->ras_tbl_mutex);
 
 	/* Read the table header from EEPROM address */
@@ -1111,36 +1581,153 @@ int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control,
 				 control->i2c_address + control->ras_header_offset,
 				 buf, RAS_TABLE_HEADER_SIZE);
 	if (res < RAS_TABLE_HEADER_SIZE) {
-		DRM_ERROR("Failed to read EEPROM table header, res:%d", res);
+		dev_err(adev->dev, "Failed to read EEPROM table header, res:%d",
+			res);
 		return res >= 0 ? -EIO : res;
 	}
 
 	__decode_table_header_from_buf(hdr, buf);
 
-	control->ras_num_recs = RAS_NUM_RECS(hdr);
+	if (hdr->header != RAS_TABLE_HDR_VAL &&
+	    hdr->header != RAS_TABLE_HDR_BAD) {
+		dev_info(adev->dev, "Creating a new EEPROM table");
+		return amdgpu_ras_eeprom_reset_table(control);
+	}
+
+	switch (hdr->version) {
+	case RAS_TABLE_VER_V2_1:
+	case RAS_TABLE_VER_V3:
+		control->ras_num_recs = RAS_NUM_RECS_V2_1(hdr);
+		control->ras_record_offset = RAS_RECORD_START_V2_1;
+		control->ras_max_record_count = RAS_MAX_RECORD_COUNT_V2_1;
+		break;
+	case RAS_TABLE_VER_V1:
+		control->ras_num_recs = RAS_NUM_RECS(hdr);
+		control->ras_record_offset = RAS_RECORD_START;
+		control->ras_max_record_count = RAS_MAX_RECORD_COUNT;
+		break;
+	default:
+		dev_err(adev->dev,
+			"RAS header invalid, unsupported version: %u",
+			hdr->version);
+		return -EINVAL;
+	}
+
+	if (control->ras_num_recs > control->ras_max_record_count) {
+		dev_err(adev->dev,
+			"RAS header invalid, records in header: %u max allowed :%u",
+			control->ras_num_recs, control->ras_max_record_count);
+		return -EINVAL;
+	}
+
 	control->ras_fri = RAS_OFFSET_TO_INDEX(control, hdr->first_rec_offset);
+	control->ras_num_mca_recs = 0;
+	control->ras_num_pa_recs = 0;
+	return 0;
+}
+
+static int amdgpu_ras_smu_eeprom_check(struct amdgpu_ras_eeprom_control *control)
+{
+	struct amdgpu_device *adev = to_amdgpu_device(control);
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+
+	if (!__is_ras_eeprom_supported(adev))
+		return 0;
+
+	control->ras_num_bad_pages = ras->bad_page_num;
+
+	if ((ras->bad_page_cnt_threshold < control->ras_num_bad_pages) &&
+	    amdgpu_bad_page_threshold != 0) {
+		dev_warn(adev->dev,
+			"RAS records:%d exceed threshold:%d\n",
+			control->ras_num_bad_pages, ras->bad_page_cnt_threshold);
+		if ((amdgpu_bad_page_threshold == -1) ||
+			(amdgpu_bad_page_threshold == -2)) {
+			dev_warn(adev->dev,
+				 "Please consult AMD Service Action Guide (SAG) for appropriate service procedures\n");
+		} else {
+			ras->is_rma = true;
+			dev_warn(adev->dev,
+				 "User defined threshold is set, runtime service will be halt when threshold is reached\n");
+		}
+
+		return 0;
+	}
+
+	dev_dbg(adev->dev,
+		"Found existing EEPROM table with %d records",
+		control->ras_num_bad_pages);
+
+	/* Warn if we are at 90% of the threshold or above
+	 */
+	if (10 * control->ras_num_bad_pages >= 9 * ras->bad_page_cnt_threshold)
+		dev_warn(adev->dev, "RAS records:%u exceeds 90%% of threshold:%d",
+				control->ras_num_bad_pages,
+				ras->bad_page_cnt_threshold);
+	return 0;
+}
+
+int amdgpu_ras_eeprom_check(struct amdgpu_ras_eeprom_control *control)
+{
+	struct amdgpu_device *adev = to_amdgpu_device(control);
+	struct amdgpu_ras_eeprom_table_header *hdr = &control->tbl_hdr;
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+	int res = 0;
+
+	if (amdgpu_ras_smu_eeprom_supported(adev))
+		return amdgpu_ras_smu_eeprom_check(control);
+
+	if (!__is_ras_eeprom_supported(adev))
+		return 0;
+
+	/* Verify i2c adapter is initialized */
+	if (!adev->pm.ras_eeprom_i2c_bus || !adev->pm.ras_eeprom_i2c_bus->algo)
+		return -ENOENT;
+
+	if (!__get_eeprom_i2c_addr(adev, control))
+		return -EINVAL;
+
+	control->ras_num_bad_pages = ras->bad_page_num;
 
 	if (hdr->header == RAS_TABLE_HDR_VAL) {
-		DRM_DEBUG_DRIVER("Found existing EEPROM table with %d records",
-				 control->ras_num_recs);
+		dev_dbg(adev->dev,
+			"Found existing EEPROM table with %d records",
+			control->ras_num_bad_pages);
+
+		if (hdr->version >= RAS_TABLE_VER_V2_1) {
+			res = __read_table_ras_info(control);
+			if (res)
+				return res;
+		}
+
 		res = __verify_ras_table_checksum(control);
 		if (res)
-			DRM_ERROR("RAS table incorrect checksum or error:%d\n",
-				  res);
+			dev_err(adev->dev,
+				"RAS table incorrect checksum or error:%d\n",
+				res);
 
 		/* Warn if we are at 90% of the threshold or above
 		 */
-		if (10 * control->ras_num_recs >= 9 * ras->bad_page_cnt_threshold)
+		if (10 * control->ras_num_bad_pages >= 9 * ras->bad_page_cnt_threshold)
 			dev_warn(adev->dev, "RAS records:%u exceeds 90%% of threshold:%d",
-					control->ras_num_recs,
+					control->ras_num_bad_pages,
 					ras->bad_page_cnt_threshold);
 	} else if (hdr->header == RAS_TABLE_HDR_BAD &&
 		   amdgpu_bad_page_threshold != 0) {
+		if (hdr->version >= RAS_TABLE_VER_V2_1) {
+			res = __read_table_ras_info(control);
+			if (res)
+				return res;
+		}
+
 		res = __verify_ras_table_checksum(control);
-		if (res)
-			DRM_ERROR("RAS Table incorrect checksum or error:%d\n",
-				  res);
-		if (ras->bad_page_cnt_threshold > control->ras_num_recs) {
+		if (res) {
+			dev_err(adev->dev,
+				"RAS Table incorrect checksum or error:%d\n",
+				res);
+			return -EINVAL;
+		}
+		if (ras->bad_page_cnt_threshold >= control->ras_num_bad_pages) {
 			/* This means that, the threshold was increased since
 			 * the last time the system was booted, and now,
 			 * ras->bad_page_cnt_threshold - control->num_recs > 0,
@@ -1150,29 +1737,194 @@ int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control,
 			dev_info(adev->dev,
 				 "records:%d threshold:%d, resetting "
 				 "RAS table header signature",
-				 control->ras_num_recs,
+				 control->ras_num_bad_pages,
 				 ras->bad_page_cnt_threshold);
 			res = amdgpu_ras_eeprom_correct_header_tag(control,
 								   RAS_TABLE_HDR_VAL);
 		} else {
-			dev_err(adev->dev, "RAS records:%d exceed threshold:%d",
-				control->ras_num_recs, ras->bad_page_cnt_threshold);
-			if (amdgpu_bad_page_threshold == -1) {
-				dev_warn(adev->dev, "GPU will be initialized due to bad_page_threshold = -1.");
+			dev_warn(adev->dev,
+				"RAS records:%d exceed threshold:%d\n",
+				control->ras_num_bad_pages, ras->bad_page_cnt_threshold);
+			if ((amdgpu_bad_page_threshold == -1) ||
+			    (amdgpu_bad_page_threshold == -2)) {
 				res = 0;
+				dev_warn(adev->dev,
+					 "Please consult AMD Service Action Guide (SAG) for appropriate service procedures\n");
 			} else {
-				*exceed_err_limit = true;
-				dev_err(adev->dev,
-					"RAS records:%d exceed threshold:%d, "
-					"GPU will not be initialized. Replace this GPU or increase the threshold",
-					control->ras_num_recs, ras->bad_page_cnt_threshold);
+				ras->is_rma = true;
+				dev_warn(adev->dev,
+					 "User defined threshold is set, runtime service will be halt when threshold is reached\n");
 			}
 		}
-	} else {
-		DRM_INFO("Creating a new EEPROM table");
-
-		res = amdgpu_ras_eeprom_reset_table(control);
 	}
 
 	return res < 0 ? res : 0;
 }
+
+void amdgpu_ras_eeprom_check_and_recover(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+	struct amdgpu_ras_eeprom_control *control;
+	int res;
+
+	if (!__is_ras_eeprom_supported(adev) || !ras ||
+	    amdgpu_ras_smu_eeprom_supported(adev))
+		return;
+	control = &ras->eeprom_control;
+	if (!control->is_eeprom_valid)
+		return;
+	res = __verify_ras_table_checksum(control);
+	if (res) {
+		dev_warn(adev->dev,
+			"RAS table incorrect checksum or error:%d, try to recover\n",
+			res);
+		if (!amdgpu_ras_eeprom_reset_table(control))
+			if (!amdgpu_ras_save_bad_pages(adev, NULL))
+				if (!__verify_ras_table_checksum(control)) {
+					dev_info(adev->dev, "RAS table recovery succeed\n");
+					return;
+				}
+		dev_err(adev->dev, "RAS table recovery failed\n");
+		control->is_eeprom_valid = false;
+	}
+	return;
+}
+
+static const struct ras_smu_drv *amdgpu_ras_get_smu_ras_drv(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+
+	if (!ras)
+		return NULL;
+
+	return ras->ras_smu_drv;
+}
+
+static uint64_t amdgpu_ras_smu_get_feature_flags(struct amdgpu_device *adev)
+{
+	const struct ras_smu_drv *ras_smu_drv = amdgpu_ras_get_smu_ras_drv(adev);
+	uint64_t flags = 0ULL;
+
+	if (!ras_smu_drv)
+		goto out;
+
+	if (ras_smu_drv->ras_smu_feature_flags)
+		ras_smu_drv->ras_smu_feature_flags(adev, &flags);
+
+out:
+	return flags;
+}
+
+bool amdgpu_ras_smu_eeprom_supported(struct amdgpu_device *adev)
+{
+	const struct ras_smu_drv *smu_ras_drv = amdgpu_ras_get_smu_ras_drv(adev);
+	uint64_t flags = 0ULL;
+
+	if (!__is_ras_eeprom_supported(adev) || !smu_ras_drv)
+		return false;
+
+	if (!smu_ras_drv->smu_eeprom_funcs)
+		return false;
+
+	flags = amdgpu_ras_smu_get_feature_flags(adev);
+
+	return !!(flags & RAS_SMU_FEATURE_BIT__RAS_EEPROM);
+}
+
+int amdgpu_ras_smu_get_table_version(struct amdgpu_device *adev,
+				     uint32_t *table_version)
+{
+	const struct ras_smu_drv *smu_ras_drv = amdgpu_ras_get_smu_ras_drv(adev);
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev))
+		return -EOPNOTSUPP;
+
+	if (smu_ras_drv->smu_eeprom_funcs->get_ras_table_version)
+		return smu_ras_drv->smu_eeprom_funcs->get_ras_table_version(adev,
+										 table_version);
+	return -EOPNOTSUPP;
+}
+
+int amdgpu_ras_smu_get_badpage_count(struct amdgpu_device *adev,
+				     uint32_t *count, uint32_t timeout)
+{
+	const struct ras_smu_drv *smu_ras_drv = amdgpu_ras_get_smu_ras_drv(adev);
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev))
+		return -EOPNOTSUPP;
+
+	if (smu_ras_drv->smu_eeprom_funcs->get_badpage_count)
+		return smu_ras_drv->smu_eeprom_funcs->get_badpage_count(adev,
+									     count, timeout);
+	return -EOPNOTSUPP;
+}
+
+int amdgpu_ras_smu_get_badpage_mca_addr(struct amdgpu_device *adev,
+					uint16_t index, uint64_t *mca_addr)
+{
+	const struct ras_smu_drv *smu_ras_drv = amdgpu_ras_get_smu_ras_drv(adev);
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev))
+		return -EOPNOTSUPP;
+
+	if (smu_ras_drv->smu_eeprom_funcs->get_badpage_mca_addr)
+		return smu_ras_drv->smu_eeprom_funcs->get_badpage_mca_addr(adev,
+										index, mca_addr);
+	return -EOPNOTSUPP;
+}
+
+int amdgpu_ras_smu_set_timestamp(struct amdgpu_device *adev,
+				 uint64_t timestamp)
+{
+	const struct ras_smu_drv *smu_ras_drv = amdgpu_ras_get_smu_ras_drv(adev);
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev))
+		return -EOPNOTSUPP;
+
+	if (smu_ras_drv->smu_eeprom_funcs->set_timestamp)
+		return smu_ras_drv->smu_eeprom_funcs->set_timestamp(adev,
+									 timestamp);
+	return -EOPNOTSUPP;
+}
+
+int amdgpu_ras_smu_get_timestamp(struct amdgpu_device *adev,
+				 uint16_t index, uint64_t *timestamp)
+{
+	const struct ras_smu_drv *smu_ras_drv = amdgpu_ras_get_smu_ras_drv(adev);
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev))
+		return -EOPNOTSUPP;
+
+	if (smu_ras_drv->smu_eeprom_funcs->get_timestamp)
+		return smu_ras_drv->smu_eeprom_funcs->get_timestamp(adev,
+									 index, timestamp);
+	return -EOPNOTSUPP;
+}
+
+int amdgpu_ras_smu_get_badpage_ipid(struct amdgpu_device *adev,
+				    uint16_t index, uint64_t *ipid)
+{
+	const struct ras_smu_drv *smu_ras_drv = amdgpu_ras_get_smu_ras_drv(adev);
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev))
+		return -EOPNOTSUPP;
+
+	if (smu_ras_drv->smu_eeprom_funcs->get_badpage_ipid)
+		return smu_ras_drv->smu_eeprom_funcs->get_badpage_ipid(adev,
+									    index, ipid);
+	return -EOPNOTSUPP;
+}
+
+int amdgpu_ras_smu_erase_ras_table(struct amdgpu_device *adev,
+				   uint32_t *result)
+{
+	const struct ras_smu_drv *smu_ras_drv = amdgpu_ras_get_smu_ras_drv(adev);
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev))
+		return -EOPNOTSUPP;
+
+	if (smu_ras_drv->smu_eeprom_funcs->erase_ras_table)
+		return smu_ras_drv->smu_eeprom_funcs->erase_ras_table(adev,
+									   result);
+	return -EOPNOTSUPP;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.h
index 54d9bfe0881d..2e5d63957e71 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.h
@@ -26,8 +26,17 @@
 
 #include <linux/i2c.h>
 
+#define RAS_TABLE_VER_V1           0x00010000
+#define RAS_TABLE_VER_V2_1         0x00021000
+#define RAS_TABLE_VER_V3           0x00030000
+
 struct amdgpu_device;
 
+enum amdgpu_ras_gpu_health_status {
+	GPU_HEALTH_USABLE = 0,
+	GPU_RETIRED__ECC_REACH_THRESHOLD = 2,
+};
+
 enum amdgpu_ras_eeprom_err_type {
 	AMDGPU_RAS_EEPROM_ERR_NA,
 	AMDGPU_RAS_EEPROM_ERR_RECOVERABLE,
@@ -43,9 +52,18 @@ struct amdgpu_ras_eeprom_table_header {
 	uint32_t checksum;
 } __packed;
 
+struct amdgpu_ras_eeprom_table_ras_info {
+	u8  rma_status;
+	u8  health_percent;
+	u16 ecc_page_threshold;
+	u32 padding[64 - 1];
+} __packed;
+
 struct amdgpu_ras_eeprom_control {
 	struct amdgpu_ras_eeprom_table_header tbl_hdr;
 
+	struct amdgpu_ras_eeprom_table_ras_info tbl_rai;
+
 	/* Base I2C EEPPROM 19-bit memory address,
 	 * where the table is located. For more information,
 	 * see top of amdgpu_eeprom.c.
@@ -58,11 +76,24 @@ struct amdgpu_ras_eeprom_control {
 	 * right after the header.
 	 */
 	u32 ras_header_offset;
+	u32 ras_info_offset;
 	u32 ras_record_offset;
 
 	/* Number of records in the table.
 	 */
 	u32 ras_num_recs;
+	u32 ras_num_recs_old;
+
+	/* the bad page number is ras_num_recs or
+	 * ras_num_recs * umc.retire_unit
+	 */
+	u32 ras_num_bad_pages;
+
+	/* Number of records store mca address */
+	u32 ras_num_mca_recs;
+
+	/* Number of records store physical address */
+	u32 ras_num_pa_recs;
 
 	/* First record index to read, 0-based.
 	 * Range is [0, num_recs-1]. This is
@@ -84,6 +115,8 @@ struct amdgpu_ras_eeprom_control {
 	/* Record channel info which occurred bad pages
 	 */
 	u32 bad_channel_bitmap;
+
+	bool is_eeprom_valid;
 };
 
 /*
@@ -111,8 +144,7 @@ struct eeprom_table_record {
 	unsigned char mcumc_id;
 } __packed;
 
-int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control,
-			   bool *exceed_err_limit);
+int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control);
 
 int amdgpu_ras_eeprom_reset_table(struct amdgpu_ras_eeprom_control *control);
 
@@ -124,10 +156,43 @@ int amdgpu_ras_eeprom_read(struct amdgpu_ras_eeprom_control *control,
 int amdgpu_ras_eeprom_append(struct amdgpu_ras_eeprom_control *control,
 			     struct eeprom_table_record *records, const u32 num);
 
-uint32_t amdgpu_ras_eeprom_max_record_count(void);
+uint32_t amdgpu_ras_eeprom_max_record_count(struct amdgpu_ras_eeprom_control *control);
 
 void amdgpu_ras_debugfs_set_ret_size(struct amdgpu_ras_eeprom_control *control);
 
+int amdgpu_ras_eeprom_check(struct amdgpu_ras_eeprom_control *control);
+
+void amdgpu_ras_eeprom_check_and_recover(struct amdgpu_device *adev);
+
+bool amdgpu_ras_smu_eeprom_supported(struct amdgpu_device *adev);
+
+int amdgpu_ras_smu_get_table_version(struct amdgpu_device *adev,
+							uint32_t *table_version);
+
+int amdgpu_ras_smu_get_badpage_count(struct amdgpu_device *adev,
+								uint32_t *count, uint32_t timeout);
+
+int amdgpu_ras_smu_get_badpage_mca_addr(struct amdgpu_device *adev,
+								uint16_t index, uint64_t *mca_addr);
+
+int amdgpu_ras_smu_set_timestamp(struct amdgpu_device *adev,
+										uint64_t timestamp);
+
+int amdgpu_ras_smu_get_timestamp(struct amdgpu_device *adev,
+							uint16_t index, uint64_t *timestamp);
+
+int amdgpu_ras_smu_get_badpage_ipid(struct amdgpu_device *adev,
+								uint16_t index, uint64_t *ipid);
+
+int amdgpu_ras_smu_erase_ras_table(struct amdgpu_device *adev,
+									uint32_t *result);
+
+int amdgpu_ras_eeprom_read_idx(struct amdgpu_ras_eeprom_control *control,
+			struct eeprom_table_record *record, u32 rec_idx,
+			const u32 num);
+
+int amdgpu_ras_eeprom_update_record_num(struct amdgpu_ras_eeprom_control *control);
+
 extern const struct file_operations amdgpu_ras_debugfs_eeprom_size_ops;
 extern const struct file_operations amdgpu_ras_debugfs_eeprom_table_ops;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_res_cursor.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_res_cursor.h
index 5c4f93ee0c57..be2e56ce1355 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_res_cursor.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_res_cursor.h
@@ -90,6 +90,8 @@ static inline void amdgpu_res_first(struct ttm_resource *res,
 		cur->node = block;
 		break;
 	case TTM_PL_TT:
+	case AMDGPU_PL_DOORBELL:
+	case AMDGPU_PL_MMIO_REMAP:
 		node = to_ttm_range_mgr_node(res)->mm_nodes;
 		while (start >= node->size << PAGE_SHIFT)
 			start -= node++->size << PAGE_SHIFT;
@@ -111,7 +113,6 @@ fallback:
 	cur->remaining = size;
 	cur->node = NULL;
 	WARN_ON(res && start + size > res->size);
-	return;
 }
 
 /**
@@ -152,6 +153,8 @@ static inline void amdgpu_res_next(struct amdgpu_res_cursor *cur, uint64_t size)
 		cur->size = min(amdgpu_vram_mgr_block_size(block), cur->remaining);
 		break;
 	case TTM_PL_TT:
+	case AMDGPU_PL_DOORBELL:
+	case AMDGPU_PL_MMIO_REMAP:
 		node = cur->node;
 
 		cur->node = ++node;
@@ -163,4 +166,29 @@ static inline void amdgpu_res_next(struct amdgpu_res_cursor *cur, uint64_t size)
 	}
 }
 
+/**
+ * amdgpu_res_cleared - check if blocks are cleared
+ *
+ * @cur: the cursor to extract the block
+ *
+ * Check if the @cur block is cleared
+ */
+static inline bool amdgpu_res_cleared(struct amdgpu_res_cursor *cur)
+{
+	struct drm_buddy_block *block;
+
+	switch (cur->mem_type) {
+	case TTM_PL_VRAM:
+		block = cur->node;
+
+		if (!amdgpu_vram_mgr_is_cleared(block))
+			return false;
+		break;
+	default:
+		return false;
+	}
+
+	return true;
+}
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c
index 6437ead87e5f..28c4ad62f50e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c
@@ -26,20 +26,165 @@
 #include "sienna_cichlid.h"
 #include "smu_v13_0_10.h"
 
-int amdgpu_reset_add_handler(struct amdgpu_reset_control *reset_ctl,
-			     struct amdgpu_reset_handler *handler)
+static int amdgpu_reset_xgmi_reset_on_init_suspend(struct amdgpu_device *adev)
 {
-	/* TODO: Check if handler exists? */
-	list_add_tail(&handler->handler_list, &reset_ctl->reset_handlers);
+	int i;
+
+	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
+		if (!adev->ip_blocks[i].status.valid)
+			continue;
+		if (!adev->ip_blocks[i].status.hw)
+			continue;
+		/* displays are handled in phase1 */
+		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE)
+			continue;
+
+		/* XXX handle errors */
+		amdgpu_ip_block_suspend(&adev->ip_blocks[i]);
+		adev->ip_blocks[i].status.hw = false;
+	}
+
+	/* VCN FW shared region is in frambuffer, there are some flags
+	 * initialized in that region during sw_init. Make sure the region is
+	 * backed up.
+	 */
+	amdgpu_vcn_save_vcpu_bo(adev);
+
 	return 0;
 }
 
+static int amdgpu_reset_xgmi_reset_on_init_prep_hwctxt(
+	struct amdgpu_reset_control *reset_ctl,
+	struct amdgpu_reset_context *reset_context)
+{
+	struct list_head *reset_device_list = reset_context->reset_device_list;
+	struct amdgpu_device *tmp_adev;
+	int r;
+
+	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
+		amdgpu_unregister_gpu_instance(tmp_adev);
+		r = amdgpu_reset_xgmi_reset_on_init_suspend(tmp_adev);
+		if (r) {
+			dev_err(tmp_adev->dev,
+				"xgmi reset on init: prepare for reset failed");
+			return r;
+		}
+	}
+
+	return r;
+}
+
+static int amdgpu_reset_xgmi_reset_on_init_restore_hwctxt(
+	struct amdgpu_reset_control *reset_ctl,
+	struct amdgpu_reset_context *reset_context)
+{
+	struct list_head *reset_device_list = reset_context->reset_device_list;
+	struct amdgpu_device *tmp_adev = NULL;
+	int r;
+
+	r = amdgpu_device_reinit_after_reset(reset_context);
+	if (r)
+		return r;
+	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
+		if (!tmp_adev->kfd.init_complete) {
+			kgd2kfd_init_zone_device(tmp_adev);
+			amdgpu_amdkfd_device_init(tmp_adev);
+			amdgpu_amdkfd_drm_client_create(tmp_adev);
+		}
+	}
+
+	return r;
+}
+
+static int amdgpu_reset_xgmi_reset_on_init_perform_reset(
+	struct amdgpu_reset_control *reset_ctl,
+	struct amdgpu_reset_context *reset_context)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle;
+	struct list_head *reset_device_list = reset_context->reset_device_list;
+	struct amdgpu_device *tmp_adev = NULL;
+	int r;
+
+	dev_dbg(adev->dev, "xgmi roi - hw reset\n");
+
+	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
+		mutex_lock(&tmp_adev->reset_cntl->reset_lock);
+		tmp_adev->reset_cntl->active_reset =
+			amdgpu_asic_reset_method(adev);
+	}
+	r = 0;
+	/* Mode1 reset needs to be triggered on all devices together */
+	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
+		/* For XGMI run all resets in parallel to speed up the process */
+		if (!queue_work(system_unbound_wq, &tmp_adev->xgmi_reset_work))
+			r = -EALREADY;
+		if (r) {
+			dev_err(tmp_adev->dev,
+				"xgmi reset on init: reset failed with error, %d",
+				r);
+			break;
+		}
+	}
+
+	/* For XGMI wait for all resets to complete before proceed */
+	if (!r) {
+		list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
+			flush_work(&tmp_adev->xgmi_reset_work);
+			r = tmp_adev->asic_reset_res;
+			if (r)
+				break;
+		}
+	}
+
+	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
+		mutex_unlock(&tmp_adev->reset_cntl->reset_lock);
+		tmp_adev->reset_cntl->active_reset = AMD_RESET_METHOD_NONE;
+	}
+
+	return r;
+}
+
+int amdgpu_reset_do_xgmi_reset_on_init(
+	struct amdgpu_reset_context *reset_context)
+{
+	struct list_head *reset_device_list = reset_context->reset_device_list;
+	struct amdgpu_device *adev;
+	int r;
+
+	if (!reset_device_list || list_empty(reset_device_list) ||
+	    list_is_singular(reset_device_list))
+		return -EINVAL;
+
+	adev = list_first_entry(reset_device_list, struct amdgpu_device,
+				reset_list);
+	r = amdgpu_reset_prepare_hwcontext(adev, reset_context);
+	if (r)
+		return r;
+
+	r = amdgpu_reset_perform_reset(adev, reset_context);
+
+	return r;
+}
+
+struct amdgpu_reset_handler xgmi_reset_on_init_handler = {
+	.reset_method = AMD_RESET_METHOD_ON_INIT,
+	.prepare_env = NULL,
+	.prepare_hwcontext = amdgpu_reset_xgmi_reset_on_init_prep_hwctxt,
+	.perform_reset = amdgpu_reset_xgmi_reset_on_init_perform_reset,
+	.restore_hwcontext = amdgpu_reset_xgmi_reset_on_init_restore_hwctxt,
+	.restore_env = NULL,
+	.do_reset = NULL,
+};
+
 int amdgpu_reset_init(struct amdgpu_device *adev)
 {
 	int ret = 0;
 
-	switch (adev->ip_versions[MP1_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
 	case IP_VERSION(13, 0, 2):
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 12):
+	case IP_VERSION(13, 0, 14):
 		ret = aldebaran_reset_init(adev);
 		break;
 	case IP_VERSION(11, 0, 7):
@@ -59,8 +204,11 @@ int amdgpu_reset_fini(struct amdgpu_device *adev)
 {
 	int ret = 0;
 
-	switch (adev->ip_versions[MP1_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
 	case IP_VERSION(13, 0, 2):
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 12):
+	case IP_VERSION(13, 0, 14):
 		ret = aldebaran_reset_fini(adev);
 		break;
 	case IP_VERSION(11, 0, 7):
@@ -85,7 +233,7 @@ int amdgpu_reset_prepare_hwcontext(struct amdgpu_device *adev,
 		reset_handler = adev->reset_cntl->get_reset_handler(
 			adev->reset_cntl, reset_context);
 	if (!reset_handler)
-		return -ENOSYS;
+		return -EOPNOTSUPP;
 
 	return reset_handler->prepare_hwcontext(adev->reset_cntl,
 						reset_context);
@@ -101,7 +249,7 @@ int amdgpu_reset_perform_reset(struct amdgpu_device *adev,
 		reset_handler = adev->reset_cntl->get_reset_handler(
 			adev->reset_cntl, reset_context);
 	if (!reset_handler)
-		return -ENOSYS;
+		return -EOPNOTSUPP;
 
 	ret = reset_handler->perform_reset(adev->reset_cntl, reset_context);
 	if (ret)
@@ -165,5 +313,42 @@ void amdgpu_device_unlock_reset_domain(struct amdgpu_reset_domain *reset_domain)
 	up_write(&reset_domain->sem);
 }
 
+void amdgpu_reset_get_desc(struct amdgpu_reset_context *rst_ctxt, char *buf,
+			   size_t len)
+{
+	if (!buf || !len)
+		return;
 
+	switch (rst_ctxt->src) {
+	case AMDGPU_RESET_SRC_JOB:
+		if (rst_ctxt->job) {
+			snprintf(buf, len, "job hang on ring:%s",
+				 rst_ctxt->job->base.sched->name);
+		} else {
+			strscpy(buf, "job hang", len);
+		}
+		break;
+	case AMDGPU_RESET_SRC_RAS:
+		strscpy(buf, "RAS error", len);
+		break;
+	case AMDGPU_RESET_SRC_MES:
+		strscpy(buf, "MES hang", len);
+		break;
+	case AMDGPU_RESET_SRC_HWS:
+		strscpy(buf, "HWS hang", len);
+		break;
+	case AMDGPU_RESET_SRC_USER:
+		strscpy(buf, "user trigger", len);
+		break;
+	case AMDGPU_RESET_SRC_USERQ:
+		strscpy(buf, "user queue trigger", len);
+		break;
+	default:
+		strscpy(buf, "unknown", len);
+	}
+}
 
+bool amdgpu_reset_in_recovery(struct amdgpu_device *adev)
+{
+	return (adev->init_lvl->level == AMDGPU_INIT_LEVEL_RESET_RECOVERY);
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h
index f4a501ff87d9..07b4d37f1db6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h
@@ -26,11 +26,24 @@
 
 #include "amdgpu.h"
 
+#define AMDGPU_RESET_MAX_HANDLERS 5
+
 enum AMDGPU_RESET_FLAGS {
 
 	AMDGPU_NEED_FULL_RESET = 0,
 	AMDGPU_SKIP_HW_RESET = 1,
-	AMDGPU_RESET_FOR_DEVICE_REMOVE = 2,
+	AMDGPU_SKIP_COREDUMP = 2,
+	AMDGPU_HOST_FLR = 3,
+};
+
+enum AMDGPU_RESET_SRCS {
+	AMDGPU_RESET_SRC_UNKNOWN,
+	AMDGPU_RESET_SRC_JOB,
+	AMDGPU_RESET_SRC_RAS,
+	AMDGPU_RESET_SRC_MES,
+	AMDGPU_RESET_SRC_HWS,
+	AMDGPU_RESET_SRC_USER,
+	AMDGPU_RESET_SRC_USERQ,
 };
 
 struct amdgpu_reset_context {
@@ -40,11 +53,11 @@ struct amdgpu_reset_context {
 	struct amdgpu_hive_info *hive;
 	struct list_head *reset_device_list;
 	unsigned long flags;
+	enum AMDGPU_RESET_SRCS src;
 };
 
 struct amdgpu_reset_handler {
 	enum amd_reset_method reset_method;
-	struct list_head handler_list;
 	int (*prepare_env)(struct amdgpu_reset_control *reset_ctl,
 			   struct amdgpu_reset_context *context);
 	int (*prepare_hwcontext)(struct amdgpu_reset_control *reset_ctl,
@@ -63,7 +76,8 @@ struct amdgpu_reset_control {
 	void *handle;
 	struct work_struct reset_work;
 	struct mutex reset_lock;
-	struct list_head reset_handlers;
+	struct amdgpu_reset_handler *(
+		*reset_handlers)[AMDGPU_RESET_MAX_HANDLERS];
 	atomic_t in_reset;
 	enum amd_reset_method active_reset;
 	struct amdgpu_reset_handler *(*get_reset_handler)(
@@ -87,7 +101,6 @@ struct amdgpu_reset_domain {
 	atomic_t reset_res;
 };
 
-
 int amdgpu_reset_init(struct amdgpu_device *adev);
 int amdgpu_reset_fini(struct amdgpu_device *adev);
 
@@ -97,8 +110,10 @@ int amdgpu_reset_prepare_hwcontext(struct amdgpu_device *adev,
 int amdgpu_reset_perform_reset(struct amdgpu_device *adev,
 			       struct amdgpu_reset_context *reset_context);
 
-int amdgpu_reset_add_handler(struct amdgpu_reset_control *reset_ctl,
-			     struct amdgpu_reset_handler *handler);
+int amdgpu_reset_prepare_env(struct amdgpu_device *adev,
+			     struct amdgpu_reset_context *reset_context);
+int amdgpu_reset_restore_env(struct amdgpu_device *adev,
+			     struct amdgpu_reset_context *reset_context);
 
 struct amdgpu_reset_domain *amdgpu_reset_create_reset_domain(enum amdgpu_reset_domain_type type,
 							     char *wq_name);
@@ -122,8 +137,40 @@ static inline bool amdgpu_reset_domain_schedule(struct amdgpu_reset_domain *doma
 	return queue_work(domain->wq, work);
 }
 
+static inline bool amdgpu_reset_pending(struct amdgpu_reset_domain *domain)
+{
+	lockdep_assert_held(&domain->sem);
+	return rwsem_is_contended(&domain->sem);
+}
+
 void amdgpu_device_lock_reset_domain(struct amdgpu_reset_domain *reset_domain);
 
 void amdgpu_device_unlock_reset_domain(struct amdgpu_reset_domain *reset_domain);
 
+void amdgpu_reset_get_desc(struct amdgpu_reset_context *rst_ctxt, char *buf,
+			   size_t len);
+
+#define for_each_handler(i, handler, reset_ctl)                  \
+	for (i = 0; (i < AMDGPU_RESET_MAX_HANDLERS) &&           \
+		    (handler = (*reset_ctl->reset_handlers)[i]); \
+	     ++i)
+
+extern struct amdgpu_reset_handler xgmi_reset_on_init_handler;
+int amdgpu_reset_do_xgmi_reset_on_init(
+	struct amdgpu_reset_context *reset_context);
+
+bool amdgpu_reset_in_recovery(struct amdgpu_device *adev);
+
+static inline void amdgpu_reset_set_dpc_status(struct amdgpu_device *adev,
+					       bool status)
+{
+	adev->pcie_reset_ctx.occurs_dpc = status;
+	adev->no_hw_access = status;
+}
+
+static inline bool amdgpu_reset_in_dpc(struct amdgpu_device *adev)
+{
+	return adev->pcie_reset_ctx.occurs_dpc;
+}
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c
index dc474b809604..c596b6df2e2d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c
@@ -33,6 +33,7 @@
 
 #include <drm/amdgpu_drm.h>
 #include "amdgpu.h"
+#include "amdgpu_ras_mgr.h"
 #include "atom.h"
 
 /*
@@ -50,6 +51,26 @@
  */
 
 /**
+ * amdgpu_ring_max_ibs - Return max IBs that fit in a single submission.
+ *
+ * @type: ring type for which to return the limit.
+ */
+unsigned int amdgpu_ring_max_ibs(enum amdgpu_ring_type type)
+{
+	switch (type) {
+	case AMDGPU_RING_TYPE_GFX:
+		/* Need to keep at least 192 on GFX7+ for old radv. */
+		return 192;
+	case AMDGPU_RING_TYPE_COMPUTE:
+		return 125;
+	case AMDGPU_RING_TYPE_VCN_JPEG:
+		return 16;
+	default:
+		return 49;
+	}
+}
+
+/**
  * amdgpu_ring_alloc - allocate space on the ring buffer
  *
  * @ring: amdgpu_ring structure holding ring information
@@ -58,7 +79,7 @@
  * Allocate @ndw dwords in the ring buffer (all asics).
  * Returns 0 on success, error on failure.
  */
-int amdgpu_ring_alloc(struct amdgpu_ring *ring, unsigned ndw)
+int amdgpu_ring_alloc(struct amdgpu_ring *ring, unsigned int ndw)
 {
 	/* Align requested size with padding so unlock_commit can
 	 * pad safely */
@@ -79,6 +100,29 @@ int amdgpu_ring_alloc(struct amdgpu_ring *ring, unsigned ndw)
 	return 0;
 }
 
+/**
+ * amdgpu_ring_alloc_reemit - allocate space on the ring buffer for reemit
+ *
+ * @ring: amdgpu_ring structure holding ring information
+ * @ndw: number of dwords to allocate in the ring buffer
+ *
+ * Allocate @ndw dwords in the ring buffer (all asics).
+ * doesn't check the max_dw limit as we may be reemitting
+ * several submissions.
+ */
+static void amdgpu_ring_alloc_reemit(struct amdgpu_ring *ring, unsigned int ndw)
+{
+	/* Align requested size with padding so unlock_commit can
+	 * pad safely */
+	ndw = (ndw + ring->funcs->align_mask) & ~ring->funcs->align_mask;
+
+	ring->count_dw = ndw;
+	ring->wptr_old = ring->wptr;
+
+	if (ring->funcs->begin_use)
+		ring->funcs->begin_use(ring);
+}
+
 /** amdgpu_ring_insert_nop - insert NOP packets
  *
  * @ring: amdgpu_ring structure holding ring information
@@ -88,10 +132,22 @@ int amdgpu_ring_alloc(struct amdgpu_ring *ring, unsigned ndw)
  */
 void amdgpu_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
 {
-	int i;
+	uint32_t occupied, chunk1, chunk2;
+
+	occupied = ring->wptr & ring->buf_mask;
+	chunk1 = ring->buf_mask + 1 - occupied;
+	chunk1 = (chunk1 >= count) ? count : chunk1;
+	chunk2 = count - chunk1;
+
+	if (chunk1)
+		memset32(&ring->ring[occupied], ring->funcs->nop, chunk1);
+
+	if (chunk2)
+		memset32(ring->ring, ring->funcs->nop, chunk2);
 
-	for (i = 0; i < count; i++)
-		amdgpu_ring_write(ring, ring->funcs->nop);
+	ring->wptr += count;
+	ring->wptr &= ring->ptr_mask;
+	ring->count_dw -= count;
 }
 
 /**
@@ -104,8 +160,16 @@ void amdgpu_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
  */
 void amdgpu_ring_generic_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib)
 {
-	while (ib->length_dw & ring->funcs->align_mask)
-		ib->ptr[ib->length_dw++] = ring->funcs->nop;
+	u32 align_mask = ring->funcs->align_mask;
+	u32 count = ib->length_dw & align_mask;
+
+	if (count) {
+		count = align_mask + 1 - count;
+
+		memset32(&ib->ptr[ib->length_dw], ring->funcs->nop, count);
+
+		ib->length_dw += count;
+	}
 }
 
 /**
@@ -121,11 +185,16 @@ void amdgpu_ring_commit(struct amdgpu_ring *ring)
 {
 	uint32_t count;
 
+	if (ring->count_dw < 0)
+		DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n");
+
 	/* We pad to match fetch size */
 	count = ring->funcs->align_mask + 1 -
 		(ring->wptr & ring->funcs->align_mask);
-	count %= ring->funcs->align_mask + 1;
-	ring->funcs->insert_nop(ring, count);
+	count &= ring->funcs->align_mask;
+
+	if (count != 0)
+		ring->funcs->insert_nop(ring, count);
 
 	mb();
 	amdgpu_ring_set_wptr(ring);
@@ -150,14 +219,10 @@ void amdgpu_ring_undo(struct amdgpu_ring *ring)
 }
 
 #define amdgpu_ring_get_gpu_addr(ring, offset)				\
-	(ring->is_mes_queue ?						\
-	 (ring->mes_ctx->meta_data_gpu_addr + offset) :			\
-	 (ring->adev->wb.gpu_addr + offset * 4))
+	 (ring->adev->wb.gpu_addr + offset * 4)
 
 #define amdgpu_ring_get_cpu_addr(ring, offset)				\
-	(ring->is_mes_queue ?						\
-	 (void *)((uint8_t *)(ring->mes_ctx->meta_data_ptr) + offset) : \
-	 (&ring->adev->wb.wb[offset]))
+	 (&ring->adev->wb.wb[offset])
 
 /**
  * amdgpu_ring_init - init driver ring struct.
@@ -182,6 +247,7 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 	int sched_hw_submission = amdgpu_sched_hw_submission;
 	u32 *num_sched;
 	u32 hw_ip;
+	unsigned int max_ibs_dw;
 
 	/* Set the hw submission limit higher for KIQ because
 	 * it's used for a number of gfx/compute tasks by both
@@ -191,6 +257,8 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 	 */
 	if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ)
 		sched_hw_submission = max(sched_hw_submission, 256);
+	if (ring->funcs->type == AMDGPU_RING_TYPE_MES)
+		sched_hw_submission = 8;
 	else if (ring == &adev->sdma.instance[0].page)
 		sched_hw_submission = 256;
 
@@ -203,57 +271,42 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 		ring->sched_score = sched_score;
 		ring->vmid_wait = dma_fence_get_stub();
 
-		if (!ring->is_mes_queue) {
-			ring->idx = adev->num_rings++;
-			adev->rings[ring->idx] = ring;
-		}
+		ring->idx = adev->num_rings++;
+		adev->rings[ring->idx] = ring;
 
 		r = amdgpu_fence_driver_init_ring(ring);
 		if (r)
 			return r;
 	}
 
-	if (ring->is_mes_queue) {
-		ring->rptr_offs = amdgpu_mes_ctx_get_offs(ring,
-				AMDGPU_MES_CTX_RPTR_OFFS);
-		ring->wptr_offs = amdgpu_mes_ctx_get_offs(ring,
-				AMDGPU_MES_CTX_WPTR_OFFS);
-		ring->fence_offs = amdgpu_mes_ctx_get_offs(ring,
-				AMDGPU_MES_CTX_FENCE_OFFS);
-		ring->trail_fence_offs = amdgpu_mes_ctx_get_offs(ring,
-				AMDGPU_MES_CTX_TRAIL_FENCE_OFFS);
-		ring->cond_exe_offs = amdgpu_mes_ctx_get_offs(ring,
-				AMDGPU_MES_CTX_COND_EXE_OFFS);
-	} else {
-		r = amdgpu_device_wb_get(adev, &ring->rptr_offs);
-		if (r) {
-			dev_err(adev->dev, "(%d) ring rptr_offs wb alloc failed\n", r);
-			return r;
-		}
+	r = amdgpu_device_wb_get(adev, &ring->rptr_offs);
+	if (r) {
+		dev_err(adev->dev, "(%d) ring rptr_offs wb alloc failed\n", r);
+		return r;
+	}
 
-		r = amdgpu_device_wb_get(adev, &ring->wptr_offs);
-		if (r) {
-			dev_err(adev->dev, "(%d) ring wptr_offs wb alloc failed\n", r);
-			return r;
-		}
+	r = amdgpu_device_wb_get(adev, &ring->wptr_offs);
+	if (r) {
+		dev_err(adev->dev, "(%d) ring wptr_offs wb alloc failed\n", r);
+		return r;
+	}
 
-		r = amdgpu_device_wb_get(adev, &ring->fence_offs);
-		if (r) {
-			dev_err(adev->dev, "(%d) ring fence_offs wb alloc failed\n", r);
-			return r;
-		}
+	r = amdgpu_device_wb_get(adev, &ring->fence_offs);
+	if (r) {
+		dev_err(adev->dev, "(%d) ring fence_offs wb alloc failed\n", r);
+		return r;
+	}
 
-		r = amdgpu_device_wb_get(adev, &ring->trail_fence_offs);
-		if (r) {
-			dev_err(adev->dev, "(%d) ring trail_fence_offs wb alloc failed\n", r);
-			return r;
-		}
+	r = amdgpu_device_wb_get(adev, &ring->trail_fence_offs);
+	if (r) {
+		dev_err(adev->dev, "(%d) ring trail_fence_offs wb alloc failed\n", r);
+		return r;
+	}
 
-		r = amdgpu_device_wb_get(adev, &ring->cond_exe_offs);
-		if (r) {
-			dev_err(adev->dev, "(%d) ring cond_exec_polling wb alloc failed\n", r);
-			return r;
-		}
+	r = amdgpu_device_wb_get(adev, &ring->cond_exe_offs);
+	if (r) {
+		dev_err(adev->dev, "(%d) ring cond_exec_polling wb alloc failed\n", r);
+		return r;
 	}
 
 	ring->fence_gpu_addr =
@@ -284,38 +337,51 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 	/* always set cond_exec_polling to CONTINUE */
 	*ring->cond_exe_cpu_addr = 1;
 
-	r = amdgpu_fence_driver_start_ring(ring, irq_src, irq_type);
-	if (r) {
-		dev_err(adev->dev, "failed initializing fences (%d).\n", r);
-		return r;
-	}
+	if (ring->funcs->type != AMDGPU_RING_TYPE_CPER) {
+		r = amdgpu_fence_driver_start_ring(ring, irq_src, irq_type);
+		if (r) {
+			dev_err(adev->dev, "failed initializing fences (%d).\n", r);
+			return r;
+		}
 
-	ring->ring_size = roundup_pow_of_two(max_dw * 4 * sched_hw_submission);
+		max_ibs_dw = ring->funcs->emit_frame_size +
+			     amdgpu_ring_max_ibs(ring->funcs->type) * ring->funcs->emit_ib_size;
+		max_ibs_dw = (max_ibs_dw + ring->funcs->align_mask) & ~ring->funcs->align_mask;
+
+		if (WARN_ON(max_ibs_dw > max_dw))
+			max_dw = max_ibs_dw;
+
+		ring->ring_size = roundup_pow_of_two(max_dw * 4 * sched_hw_submission);
+	} else {
+		ring->ring_size = roundup_pow_of_two(max_dw * 4);
+		ring->count_dw = (ring->ring_size - 4) >> 2;
+		/* ring buffer is empty now */
+		ring->wptr = *ring->rptr_cpu_addr = 0;
+	}
 
 	ring->buf_mask = (ring->ring_size / 4) - 1;
 	ring->ptr_mask = ring->funcs->support_64bit_ptrs ?
 		0xffffffffffffffff : ring->buf_mask;
+	/*  Initialize cached_rptr to 0 */
+	ring->cached_rptr = 0;
 
-	/* Allocate ring buffer */
-	if (ring->is_mes_queue) {
-		int offset = 0;
-
-		BUG_ON(ring->ring_size > PAGE_SIZE*4);
-
-		offset = amdgpu_mes_ctx_get_offs(ring,
-					 AMDGPU_MES_CTX_RING_OFFS);
-		ring->gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		ring->ring = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-		amdgpu_ring_clear_ring(ring);
+	if (!ring->ring_backup) {
+		ring->ring_backup = kvzalloc(ring->ring_size, GFP_KERNEL);
+		if (!ring->ring_backup)
+			return -ENOMEM;
+	}
 
-	} else if (ring->ring_obj == NULL) {
-		r = amdgpu_bo_create_kernel(adev, ring->ring_size + ring->funcs->extra_dw, PAGE_SIZE,
+	/* Allocate ring buffer */
+	if (ring->ring_obj == NULL) {
+		r = amdgpu_bo_create_kernel(adev, ring->ring_size + ring->funcs->extra_bytes,
+					    PAGE_SIZE,
 					    AMDGPU_GEM_DOMAIN_GTT,
 					    &ring->ring_obj,
 					    &ring->gpu_addr,
 					    (void **)&ring->ring);
 		if (r) {
 			dev_err(adev->dev, "(%d) ring create failed\n", r);
+			kvfree(ring->ring_backup);
 			return r;
 		}
 		amdgpu_ring_clear_ring(ring);
@@ -324,7 +390,7 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 	ring->max_dw = max_dw;
 	ring->hw_prio = hw_prio;
 
-	if (!ring->no_scheduler) {
+	if (!ring->no_scheduler && ring->funcs->type < AMDGPU_HW_IP_NUM) {
 		hw_ip = ring->funcs->type;
 		num_sched = &adev->gpu_sched[hw_ip][hw_prio].num_scheds;
 		adev->gpu_sched[hw_ip][hw_prio].sched[(*num_sched)++] =
@@ -345,30 +411,26 @@ void amdgpu_ring_fini(struct amdgpu_ring *ring)
 {
 
 	/* Not to finish a ring which is not initialized */
-	if (!(ring->adev) ||
-	    (!ring->is_mes_queue && !(ring->adev->rings[ring->idx])))
+	if (!(ring->adev) || !(ring->adev->rings[ring->idx]))
 		return;
 
 	ring->sched.ready = false;
 
-	if (!ring->is_mes_queue) {
-		amdgpu_device_wb_free(ring->adev, ring->rptr_offs);
-		amdgpu_device_wb_free(ring->adev, ring->wptr_offs);
+	amdgpu_device_wb_free(ring->adev, ring->rptr_offs);
+	amdgpu_device_wb_free(ring->adev, ring->wptr_offs);
 
-		amdgpu_device_wb_free(ring->adev, ring->cond_exe_offs);
-		amdgpu_device_wb_free(ring->adev, ring->fence_offs);
+	amdgpu_device_wb_free(ring->adev, ring->cond_exe_offs);
+	amdgpu_device_wb_free(ring->adev, ring->fence_offs);
 
-		amdgpu_bo_free_kernel(&ring->ring_obj,
-				      &ring->gpu_addr,
-				      (void **)&ring->ring);
-	}
+	amdgpu_bo_free_kernel(&ring->ring_obj,
+			      &ring->gpu_addr,
+			      (void **)&ring->ring);
+	kvfree(ring->ring_backup);
+	ring->ring_backup = NULL;
 
 	dma_fence_put(ring->vmid_wait);
 	ring->vmid_wait = NULL;
 	ring->me = 0;
-
-	if (!ring->is_mes_queue)
-		ring->adev->rings[ring->idx] = NULL;
 }
 
 /**
@@ -403,17 +465,30 @@ void amdgpu_ring_emit_reg_write_reg_wait_helper(struct amdgpu_ring *ring,
 bool amdgpu_ring_soft_recovery(struct amdgpu_ring *ring, unsigned int vmid,
 			       struct dma_fence *fence)
 {
-	ktime_t deadline = ktime_add_us(ktime_get(), 10000);
+	unsigned long flags;
+	ktime_t deadline;
+	bool ret;
+
+	deadline = ktime_add_us(ktime_get(), 10000);
 
 	if (amdgpu_sriov_vf(ring->adev) || !ring->funcs->soft_recovery || !fence)
 		return false;
 
-	atomic_inc(&ring->adev->gpu_reset_counter);
+	spin_lock_irqsave(fence->lock, flags);
+	if (!dma_fence_is_signaled_locked(fence))
+		dma_fence_set_error(fence, -ENODATA);
+	spin_unlock_irqrestore(fence->lock, flags);
+
 	while (!dma_fence_is_signaled(fence) &&
 	       ktime_to_ns(ktime_sub(deadline, ktime_get())) > 0)
 		ring->funcs->soft_recovery(ring, vmid);
 
-	return dma_fence_is_signaled(fence);
+	ret = dma_fence_is_signaled(fence);
+	/* increment the counter only if soft reset worked */
+	if (ret)
+		atomic_inc(&ring->adev->gpu_reset_counter);
+
+	return ret;
 }
 
 /*
@@ -421,6 +496,66 @@ bool amdgpu_ring_soft_recovery(struct amdgpu_ring *ring, unsigned int vmid,
  */
 #if defined(CONFIG_DEBUG_FS)
 
+static ssize_t amdgpu_ras_cper_debugfs_read(struct file *f, char __user *buf,
+					    size_t size, loff_t *offset)
+{
+	const uint8_t ring_header_size = 12;
+	struct amdgpu_ring *ring = file_inode(f)->i_private;
+	struct ras_cmd_cper_snapshot_req *snapshot_req __free(kfree) =
+		kzalloc(sizeof(struct ras_cmd_cper_snapshot_req), GFP_KERNEL);
+	struct ras_cmd_cper_snapshot_rsp *snapshot_rsp __free(kfree) =
+		kzalloc(sizeof(struct ras_cmd_cper_snapshot_rsp), GFP_KERNEL);
+	struct ras_cmd_cper_record_req *record_req __free(kfree) =
+		kzalloc(sizeof(struct ras_cmd_cper_record_req), GFP_KERNEL);
+	struct ras_cmd_cper_record_rsp *record_rsp __free(kfree) =
+		kzalloc(sizeof(struct ras_cmd_cper_record_rsp), GFP_KERNEL);
+	uint8_t *ring_header __free(kfree) =
+		kzalloc(ring_header_size, GFP_KERNEL);
+	uint32_t total_cper_num;
+	uint64_t start_cper_id;
+	int r;
+
+	if (!snapshot_req || !snapshot_rsp || !record_req || !record_rsp ||
+	    !ring_header)
+		return -ENOMEM;
+
+	if (!(*offset)) {
+		/* Need at least 12 bytes for the header on the first read */
+		if (size < ring_header_size)
+			return -EINVAL;
+
+		if (copy_to_user(buf, ring_header, ring_header_size))
+			return -EFAULT;
+		buf += ring_header_size;
+		size -= ring_header_size;
+	}
+
+	r = amdgpu_ras_mgr_handle_ras_cmd(ring->adev,
+					  RAS_CMD__GET_CPER_SNAPSHOT,
+					  snapshot_req, sizeof(struct ras_cmd_cper_snapshot_req),
+					  snapshot_rsp, sizeof(struct ras_cmd_cper_snapshot_rsp));
+	if (r || !snapshot_rsp->total_cper_num)
+		return r;
+
+	start_cper_id = snapshot_rsp->start_cper_id;
+	total_cper_num = snapshot_rsp->total_cper_num;
+
+	record_req->buf_ptr = (uint64_t)(uintptr_t)buf;
+	record_req->buf_size = size;
+	record_req->cper_start_id = start_cper_id + *offset;
+	record_req->cper_num = total_cper_num;
+	r = amdgpu_ras_mgr_handle_ras_cmd(ring->adev, RAS_CMD__GET_CPER_RECORD,
+					  record_req, sizeof(struct ras_cmd_cper_record_req),
+					  record_rsp, sizeof(struct ras_cmd_cper_record_rsp));
+	if (r)
+		return r;
+
+	r = *offset ? record_rsp->real_data_size : record_rsp->real_data_size + ring_header_size;
+	(*offset) += record_rsp->real_cper_num;
+
+	return r;
+}
+
 /* Layout of file is 12 bytes consisting of
  * - rptr
  * - wptr
@@ -432,8 +567,13 @@ static ssize_t amdgpu_debugfs_ring_read(struct file *f, char __user *buf,
 					size_t size, loff_t *pos)
 {
 	struct amdgpu_ring *ring = file_inode(f)->i_private;
-	int r, i;
 	uint32_t value, result, early[3];
+	uint64_t p;
+	loff_t i;
+	int r;
+
+	if (ring->funcs->type == AMDGPU_RING_TYPE_CPER && amdgpu_uniras_enabled(ring->adev))
+		return amdgpu_ras_cper_debugfs_read(f, buf, size, pos);
 
 	if (*pos & 3 || size & 3)
 		return -EINVAL;
@@ -441,13 +581,18 @@ static ssize_t amdgpu_debugfs_ring_read(struct file *f, char __user *buf,
 	result = 0;
 
 	if (*pos < 12) {
+		if (ring->funcs->type == AMDGPU_RING_TYPE_CPER)
+			mutex_lock(&ring->adev->cper.ring_lock);
+
 		early[0] = amdgpu_ring_get_rptr(ring) & ring->buf_mask;
 		early[1] = amdgpu_ring_get_wptr(ring) & ring->buf_mask;
 		early[2] = ring->wptr & ring->buf_mask;
 		for (i = *pos / 4; i < 3 && size; i++) {
 			r = put_user(early[i], (uint32_t *)buf);
-			if (r)
-				return r;
+			if (r) {
+				result = r;
+				goto out;
+			}
 			buf += 4;
 			result += 4;
 			size -= 4;
@@ -455,29 +600,113 @@ static ssize_t amdgpu_debugfs_ring_read(struct file *f, char __user *buf,
 		}
 	}
 
-	while (size) {
-		if (*pos >= (ring->ring_size + 12))
-			return result;
+	if (ring->funcs->type != AMDGPU_RING_TYPE_CPER) {
+		while (size) {
+			if (*pos >= (ring->ring_size + 12))
+				return result;
 
-		value = ring->ring[(*pos - 12)/4];
-		r = put_user(value, (uint32_t *)buf);
-		if (r)
-			return r;
-		buf += 4;
-		result += 4;
-		size -= 4;
-		*pos += 4;
+			value = ring->ring[(*pos - 12)/4];
+			r = put_user(value, (uint32_t *)buf);
+			if (r)
+				return r;
+			buf += 4;
+			result += 4;
+			size -= 4;
+			*pos += 4;
+		}
+	} else {
+		p = early[0];
+		if (early[0] <= early[1])
+			size = (early[1] - early[0]);
+		else
+			size = ring->ring_size - (early[0] - early[1]);
+
+		while (size) {
+			if (p == early[1])
+				goto out;
+
+			value = ring->ring[p];
+			r = put_user(value, (uint32_t *)buf);
+			if (r) {
+				result = r;
+				goto out;
+			}
+
+			buf += 4;
+			result += 4;
+			size--;
+			p++;
+			p &= ring->ptr_mask;
+		}
 	}
 
+out:
+	if (ring->funcs->type == AMDGPU_RING_TYPE_CPER)
+		mutex_unlock(&ring->adev->cper.ring_lock);
+
 	return result;
 }
 
+static ssize_t amdgpu_debugfs_virt_ring_read(struct file *f, char __user *buf,
+	size_t size, loff_t *pos)
+{
+	struct amdgpu_ring *ring = file_inode(f)->i_private;
+
+	if (*pos & 3 || size & 3)
+		return -EINVAL;
+
+	if (ring->funcs->type == AMDGPU_RING_TYPE_CPER)
+		amdgpu_virt_req_ras_cper_dump(ring->adev, false);
+
+	return amdgpu_debugfs_ring_read(f, buf, size, pos);
+}
+
 static const struct file_operations amdgpu_debugfs_ring_fops = {
 	.owner = THIS_MODULE,
 	.read = amdgpu_debugfs_ring_read,
 	.llseek = default_llseek
 };
 
+static const struct file_operations amdgpu_debugfs_virt_ring_fops = {
+	.owner = THIS_MODULE,
+	.read = amdgpu_debugfs_virt_ring_read,
+	.llseek = default_llseek
+};
+
+static ssize_t amdgpu_debugfs_mqd_read(struct file *f, char __user *buf,
+				       size_t size, loff_t *pos)
+{
+	struct amdgpu_ring *ring = file_inode(f)->i_private;
+	ssize_t bytes = min_t(ssize_t, ring->mqd_size - *pos, size);
+	void *from = ((u8 *)ring->mqd_ptr) + *pos;
+
+	if (*pos > ring->mqd_size)
+		return 0;
+
+	if (copy_to_user(buf, from, bytes))
+		return -EFAULT;
+
+	*pos += bytes;
+	return bytes;
+}
+
+static const struct file_operations amdgpu_debugfs_mqd_fops = {
+	.owner = THIS_MODULE,
+	.read = amdgpu_debugfs_mqd_read,
+	.llseek = default_llseek
+};
+
+static int amdgpu_debugfs_ring_error(void *data, u64 val)
+{
+	struct amdgpu_ring *ring = data;
+
+	amdgpu_fence_driver_set_error(ring, val);
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE_SIGNED(amdgpu_debugfs_error_fops, NULL,
+				amdgpu_debugfs_ring_error, "%lld\n");
+
 #endif
 
 void amdgpu_debugfs_ring_init(struct amdgpu_device *adev,
@@ -489,9 +718,25 @@ void amdgpu_debugfs_ring_init(struct amdgpu_device *adev,
 	char name[32];
 
 	sprintf(name, "amdgpu_ring_%s", ring->name);
-	debugfs_create_file_size(name, S_IFREG | S_IRUGO, root, ring,
-				 &amdgpu_debugfs_ring_fops,
-				 ring->ring_size + 12);
+	if (amdgpu_sriov_vf(adev))
+		debugfs_create_file_size(name, S_IFREG | 0444, root, ring,
+					 &amdgpu_debugfs_virt_ring_fops,
+					 ring->ring_size + 12);
+	else
+		debugfs_create_file_size(name, S_IFREG | 0444, root, ring,
+					 &amdgpu_debugfs_ring_fops,
+					 ring->ring_size + 12);
+
+	if (ring->mqd_obj) {
+		sprintf(name, "amdgpu_mqd_%s", ring->name);
+		debugfs_create_file_size(name, S_IFREG | 0444, root, ring,
+					 &amdgpu_debugfs_mqd_fops,
+					 ring->mqd_size);
+	}
+
+	sprintf(name, "amdgpu_error_%s", ring->name);
+	debugfs_create_file(name, 0200, root, ring,
+			    &amdgpu_debugfs_error_fops);
 
 #endif
 }
@@ -519,6 +764,7 @@ int amdgpu_ring_test_helper(struct amdgpu_ring *ring)
 			      ring->name);
 
 	ring->sched.ready = !r;
+
 	return r;
 }
 
@@ -526,6 +772,10 @@ static void amdgpu_ring_to_mqd_prop(struct amdgpu_ring *ring,
 				    struct amdgpu_mqd_prop *prop)
 {
 	struct amdgpu_device *adev = ring->adev;
+	bool is_high_prio_compute = ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE &&
+				    amdgpu_gfx_is_high_priority_compute_queue(adev, ring);
+	bool is_high_prio_gfx = ring->funcs->type == AMDGPU_RING_TYPE_GFX &&
+				amdgpu_gfx_is_high_priority_graphics_queue(adev, ring);
 
 	memset(prop, 0, sizeof(*prop));
 
@@ -537,16 +787,15 @@ static void amdgpu_ring_to_mqd_prop(struct amdgpu_ring *ring,
 	prop->eop_gpu_addr = ring->eop_gpu_addr;
 	prop->use_doorbell = ring->use_doorbell;
 	prop->doorbell_index = ring->doorbell_index;
+	prop->kernel_queue = true;
 
 	/* map_queues packet doesn't need activate the queue,
 	 * so only kiq need set this field.
 	 */
 	prop->hqd_active = ring->funcs->type == AMDGPU_RING_TYPE_KIQ;
 
-	if ((ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE &&
-	     amdgpu_gfx_is_high_priority_compute_queue(adev, ring)) ||
-	    (ring->funcs->type == AMDGPU_RING_TYPE_GFX &&
-	     amdgpu_gfx_is_high_priority_graphics_queue(adev, ring))) {
+	prop->allow_tunneling = is_high_prio_compute;
+	if (is_high_prio_compute || is_high_prio_gfx) {
 		prop->hqd_pipe_priority = AMDGPU_GFX_PIPE_PRIO_HIGH;
 		prop->hqd_queue_priority = AMDGPU_GFX_QUEUE_PRIORITY_MAXIMUM;
 	}
@@ -581,3 +830,98 @@ void amdgpu_ring_ib_end(struct amdgpu_ring *ring)
 	if (ring->is_sw_ring)
 		amdgpu_sw_ring_ib_end(ring);
 }
+
+void amdgpu_ring_ib_on_emit_cntl(struct amdgpu_ring *ring)
+{
+	if (ring->is_sw_ring)
+		amdgpu_sw_ring_ib_mark_offset(ring, AMDGPU_MUX_OFFSET_TYPE_CONTROL);
+}
+
+void amdgpu_ring_ib_on_emit_ce(struct amdgpu_ring *ring)
+{
+	if (ring->is_sw_ring)
+		amdgpu_sw_ring_ib_mark_offset(ring, AMDGPU_MUX_OFFSET_TYPE_CE);
+}
+
+void amdgpu_ring_ib_on_emit_de(struct amdgpu_ring *ring)
+{
+	if (ring->is_sw_ring)
+		amdgpu_sw_ring_ib_mark_offset(ring, AMDGPU_MUX_OFFSET_TYPE_DE);
+}
+
+bool amdgpu_ring_sched_ready(struct amdgpu_ring *ring)
+{
+	if (!ring)
+		return false;
+
+	if (ring->no_scheduler || !drm_sched_wqueue_ready(&ring->sched))
+		return false;
+
+	return true;
+}
+
+void amdgpu_ring_reset_helper_begin(struct amdgpu_ring *ring,
+				    struct amdgpu_fence *guilty_fence)
+{
+	/* Stop the scheduler to prevent anybody else from touching the ring buffer. */
+	drm_sched_wqueue_stop(&ring->sched);
+	/* back up the non-guilty commands */
+	amdgpu_ring_backup_unprocessed_commands(ring, guilty_fence);
+}
+
+int amdgpu_ring_reset_helper_end(struct amdgpu_ring *ring,
+				 struct amdgpu_fence *guilty_fence)
+{
+	unsigned int i;
+	int r;
+
+	/* verify that the ring is functional */
+	r = amdgpu_ring_test_ring(ring);
+	if (r)
+		return r;
+
+	/* signal the guilty fence and set an error on all fences from the context */
+	if (guilty_fence)
+		amdgpu_fence_driver_guilty_force_completion(guilty_fence);
+	/* Re-emit the non-guilty commands */
+	if (ring->ring_backup_entries_to_copy) {
+		amdgpu_ring_alloc_reemit(ring, ring->ring_backup_entries_to_copy);
+		for (i = 0; i < ring->ring_backup_entries_to_copy; i++)
+			amdgpu_ring_write(ring, ring->ring_backup[i]);
+		amdgpu_ring_commit(ring);
+	}
+	/* Start the scheduler again */
+	drm_sched_wqueue_start(&ring->sched);
+	return 0;
+}
+
+bool amdgpu_ring_is_reset_type_supported(struct amdgpu_ring *ring,
+					 u32 reset_type)
+{
+	switch (ring->funcs->type) {
+	case AMDGPU_RING_TYPE_GFX:
+		if (ring->adev->gfx.gfx_supported_reset & reset_type)
+			return true;
+		break;
+	case AMDGPU_RING_TYPE_COMPUTE:
+		if (ring->adev->gfx.compute_supported_reset & reset_type)
+			return true;
+		break;
+	case AMDGPU_RING_TYPE_SDMA:
+		if (ring->adev->sdma.supported_reset & reset_type)
+			return true;
+		break;
+	case AMDGPU_RING_TYPE_VCN_DEC:
+	case AMDGPU_RING_TYPE_VCN_ENC:
+		if (ring->adev->vcn.supported_reset & reset_type)
+			return true;
+		break;
+	case AMDGPU_RING_TYPE_VCN_JPEG:
+		if (ring->adev->jpeg.supported_reset & reset_type)
+			return true;
+		break;
+	default:
+		break;
+	}
+	return false;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
index d8749444b689..7a27c6c4bb44 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
@@ -37,13 +37,14 @@ struct amdgpu_job;
 struct amdgpu_vm;
 
 /* max number of rings */
-#define AMDGPU_MAX_RINGS		28
-#define AMDGPU_MAX_HWIP_RINGS		8
+#define AMDGPU_MAX_RINGS		149
+#define AMDGPU_MAX_HWIP_RINGS		64
 #define AMDGPU_MAX_GFX_RINGS		2
 #define AMDGPU_MAX_SW_GFX_RINGS         2
 #define AMDGPU_MAX_COMPUTE_RINGS	8
 #define AMDGPU_MAX_VCE_RINGS		3
 #define AMDGPU_MAX_UVD_ENC_RINGS	2
+#define AMDGPU_MAX_VPE_RINGS		2
 
 enum amdgpu_ring_priority_level {
 	AMDGPU_RING_PRIO_0,
@@ -77,8 +78,12 @@ enum amdgpu_ring_type {
 	AMDGPU_RING_TYPE_VCN_DEC	= AMDGPU_HW_IP_VCN_DEC,
 	AMDGPU_RING_TYPE_VCN_ENC	= AMDGPU_HW_IP_VCN_ENC,
 	AMDGPU_RING_TYPE_VCN_JPEG	= AMDGPU_HW_IP_VCN_JPEG,
+	AMDGPU_RING_TYPE_VPE		= AMDGPU_HW_IP_VPE,
 	AMDGPU_RING_TYPE_KIQ,
-	AMDGPU_RING_TYPE_MES
+	AMDGPU_RING_TYPE_MES,
+	AMDGPU_RING_TYPE_UMSCH_MM,
+	AMDGPU_RING_TYPE_CPER,
+	AMDGPU_RING_TYPE_MAX,
 };
 
 enum amdgpu_ib_pool_type {
@@ -110,10 +115,11 @@ struct amdgpu_sched {
  */
 struct amdgpu_fence_driver {
 	uint64_t			gpu_addr;
-	volatile uint32_t		*cpu_addr;
+	uint32_t			*cpu_addr;
 	/* sync_seq is protected by ring emission lock */
 	uint32_t			sync_seq;
 	atomic_t			last_seq;
+	u64				signalled_wptr;
 	bool				initialized;
 	struct amdgpu_irq_src		*irq_src;
 	unsigned			irq_type;
@@ -123,10 +129,33 @@ struct amdgpu_fence_driver {
 	struct dma_fence		**fences;
 };
 
+/*
+ * Fences mark an event in the GPUs pipeline and are used
+ * for GPU/CPU synchronization.  When the fence is written,
+ * it is expected that all buffers associated with that fence
+ * are no longer in use by the associated ring on the GPU and
+ * that the relevant GPU caches have been flushed.
+ */
+
+struct amdgpu_fence {
+	struct dma_fence base;
+
+	/* RB, DMA, etc. */
+	struct amdgpu_ring		*ring;
+	ktime_t				start_timestamp;
+
+	/* wptr for the fence for resets */
+	u64				wptr;
+	/* fence context for resets */
+	u64				context;
+};
+
 extern const struct drm_sched_backend_ops amdgpu_sched_ops;
 
-void amdgpu_fence_driver_clear_job_fences(struct amdgpu_ring *ring);
+void amdgpu_fence_driver_set_error(struct amdgpu_ring *ring, int error);
 void amdgpu_fence_driver_force_completion(struct amdgpu_ring *ring);
+void amdgpu_fence_driver_guilty_force_completion(struct amdgpu_fence *af);
+void amdgpu_fence_save_wptr(struct amdgpu_fence *af);
 
 int amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring);
 int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring,
@@ -136,8 +165,8 @@ void amdgpu_fence_driver_hw_init(struct amdgpu_device *adev);
 void amdgpu_fence_driver_hw_fini(struct amdgpu_device *adev);
 int amdgpu_fence_driver_sw_init(struct amdgpu_device *adev);
 void amdgpu_fence_driver_sw_fini(struct amdgpu_device *adev);
-int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **fence, struct amdgpu_job *job,
-		      unsigned flags);
+int amdgpu_fence_emit(struct amdgpu_ring *ring, struct amdgpu_fence *af,
+		      unsigned int flags);
 int amdgpu_fence_emit_polling(struct amdgpu_ring *ring, uint32_t *s,
 			      uint32_t timeout);
 bool amdgpu_fence_process(struct amdgpu_ring *ring);
@@ -159,13 +188,40 @@ void amdgpu_fence_update_start_timestamp(struct amdgpu_ring *ring, uint32_t seq,
 
 /* provided by hw blocks that expose a ring buffer for commands */
 struct amdgpu_ring_funcs {
+	/**
+	 * @type:
+	 *
+	 * GFX, Compute, SDMA, UVD, VCE, VCN, VPE, KIQ, MES, UMSCH, and CPER
+	 * use ring buffers. The type field just identifies which component the
+	 * ring buffer is associated with.
+	 */
 	enum amdgpu_ring_type	type;
 	uint32_t		align_mask;
+
+	/**
+	 * @nop:
+	 *
+	 * Every block in the amdgpu has no-op instructions (e.g., GFX 10
+	 * uses PACKET3(PACKET3_NOP, 0x3FFF), VCN 5 uses VCN_ENC_CMD_NO_OP,
+	 * etc). This field receives the specific no-op for the component
+	 * that initializes the ring.
+	 */
 	u32			nop;
 	bool			support_64bit_ptrs;
 	bool			no_user_fence;
 	bool			secure_submission_supported;
-	unsigned		extra_dw;
+
+	/**
+	 * @extra_bytes:
+	 *
+	 * Optional extra space in bytes that is added to the ring size
+	 * when allocating the BO that holds the contents of the ring.
+	 * This space isn't used for command submission to the ring,
+	 * but is just there to satisfy some hardware requirements or
+	 * implement workarounds. It's up to the implementation of each
+	 * specific ring to initialize this space.
+	 */
+	unsigned		extra_bytes;
 
 	/* ring read/write ptr handling */
 	u64 (*get_rptr)(struct amdgpu_ring *ring);
@@ -205,13 +261,14 @@ struct amdgpu_ring_funcs {
 	void (*insert_end)(struct amdgpu_ring *ring);
 	/* pad the indirect buffer to the necessary number of dw */
 	void (*pad_ib)(struct amdgpu_ring *ring, struct amdgpu_ib *ib);
-	unsigned (*init_cond_exec)(struct amdgpu_ring *ring);
-	void (*patch_cond_exec)(struct amdgpu_ring *ring, unsigned offset);
+	unsigned (*init_cond_exec)(struct amdgpu_ring *ring, uint64_t addr);
 	/* note usage for clock and power gating */
 	void (*begin_use)(struct amdgpu_ring *ring);
 	void (*end_use)(struct amdgpu_ring *ring);
 	void (*emit_switch_buffer) (struct amdgpu_ring *ring);
 	void (*emit_cntxcntl) (struct amdgpu_ring *ring, uint32_t flags);
+	void (*emit_gfx_shadow)(struct amdgpu_ring *ring, u64 shadow_va, u64 csa_va,
+				u64 gds_va, bool init_shadow, int vmid);
 	void (*emit_rreg)(struct amdgpu_ring *ring, uint32_t reg,
 			  uint32_t reg_val_offs);
 	void (*emit_wreg)(struct amdgpu_ring *ring, uint32_t reg, uint32_t val);
@@ -227,8 +284,17 @@ struct amdgpu_ring_funcs {
 	int (*preempt_ib)(struct amdgpu_ring *ring);
 	void (*emit_mem_sync)(struct amdgpu_ring *ring);
 	void (*emit_wave_limit)(struct amdgpu_ring *ring, bool enable);
+	void (*patch_cntl)(struct amdgpu_ring *ring, unsigned offset);
+	void (*patch_ce)(struct amdgpu_ring *ring, unsigned offset);
+	void (*patch_de)(struct amdgpu_ring *ring, unsigned offset);
+	int (*reset)(struct amdgpu_ring *ring, unsigned int vmid,
+		     struct amdgpu_fence *timedout_fence);
+	void (*emit_cleaner_shader)(struct amdgpu_ring *ring);
 };
 
+/**
+ * amdgpu_ring - Holds ring information
+ */
 struct amdgpu_ring {
 	struct amdgpu_device		*adev;
 	const struct amdgpu_ring_funcs	*funcs;
@@ -236,62 +302,122 @@ struct amdgpu_ring {
 	struct drm_gpu_scheduler	sched;
 
 	struct amdgpu_bo	*ring_obj;
-	volatile uint32_t	*ring;
+	uint32_t		*ring;
+	/* backups for resets */
+	uint32_t		*ring_backup;
+	unsigned int		ring_backup_entries_to_copy;
 	unsigned		rptr_offs;
 	u64			rptr_gpu_addr;
-	volatile u32		*rptr_cpu_addr;
+	u32			*rptr_cpu_addr;
+
+	/**
+	 * @wptr:
+	 *
+	 * This is part of the Ring buffer implementation and represents the
+	 * write pointer. The wptr determines where the host has written.
+	 */
 	u64			wptr;
+
+	/**
+	 * @wptr_old:
+	 *
+	 * Before update wptr with the new value, usually the old value is
+	 * stored in the wptr_old.
+	 */
 	u64			wptr_old;
 	unsigned		ring_size;
+
+	/**
+	 * @max_dw:
+	 *
+	 * Maximum number of DWords for ring allocation. This information is
+	 * provided at the ring initialization time, and each IP block can
+	 * specify a specific value. Check places that invoke
+	 * amdgpu_ring_init() to see the maximum size per block.
+	 */
 	unsigned		max_dw;
+
+	/**
+	 * @count_dw:
+	 *
+	 * This value starts with the maximum amount of DWords supported by the
+	 * ring. This value is updated based on the ring manipulation.
+	 */
 	int			count_dw;
 	uint64_t		gpu_addr;
+
+	/**
+	 * @ptr_mask:
+	 *
+	 * Some IPs provide support for 64-bit pointers and others for 32-bit
+	 * only; this behavior is component-specific and defined by the field
+	 * support_64bit_ptr. If the IP block supports 64-bits, the mask
+	 * 0xffffffffffffffff is set; otherwise, this value assumes buf_mask.
+	 * Notice that this field is used to keep wptr under a valid range.
+	 */
 	uint64_t		ptr_mask;
+
+	/**
+	 * @buf_mask:
+	 *
+	 * Buffer mask is a value used to keep wptr count under its
+	 * thresholding. Buffer mask initialized during the ring buffer
+	 * initialization time, and it is defined as (ring_size / 4) -1.
+	 */
 	uint32_t		buf_mask;
 	u32			idx;
 	u32			xcc_id;
+	u32			xcp_id;
 	u32			me;
 	u32			pipe;
 	u32			queue;
 	struct amdgpu_bo	*mqd_obj;
 	uint64_t                mqd_gpu_addr;
 	void                    *mqd_ptr;
+	unsigned                mqd_size;
 	uint64_t                eop_gpu_addr;
 	u32			doorbell_index;
 	bool			use_doorbell;
 	bool			use_pollmem;
 	unsigned		wptr_offs;
 	u64			wptr_gpu_addr;
-	volatile u32		*wptr_cpu_addr;
+
+	/**
+	 * @wptr_cpu_addr:
+	 *
+	 * This is the CPU address pointer in the writeback slot. This is used
+	 * to commit changes to the GPU.
+	 */
+	u32			*wptr_cpu_addr;
 	unsigned		fence_offs;
 	u64			fence_gpu_addr;
-	volatile u32		*fence_cpu_addr;
+	u32			*fence_cpu_addr;
 	uint64_t		current_ctx;
 	char			name[16];
 	u32                     trail_seq;
 	unsigned		trail_fence_offs;
 	u64			trail_fence_gpu_addr;
-	volatile u32		*trail_fence_cpu_addr;
+	u32			*trail_fence_cpu_addr;
 	unsigned		cond_exe_offs;
 	u64			cond_exe_gpu_addr;
-	volatile u32		*cond_exe_cpu_addr;
+	u32			*cond_exe_cpu_addr;
+	unsigned int		set_q_mode_offs;
+	u32			*set_q_mode_ptr;
+	u64			set_q_mode_token;
 	unsigned		vm_hub;
 	unsigned		vm_inv_eng;
 	struct dma_fence	*vmid_wait;
 	bool			has_compute_vm_bug;
 	bool			no_scheduler;
+	bool			no_user_submission;
 	int			hw_prio;
 	unsigned 		num_hw_submission;
 	atomic_t		*sched_score;
 
-	/* used for mes */
-	bool			is_mes_queue;
-	uint32_t		hw_queue_id;
-	struct amdgpu_mes_ctx_data *mes_ctx;
-
 	bool            is_sw_ring;
 	unsigned int    entry_index;
-
+	/* store the cached rptr to restore after reset */
+	uint64_t cached_rptr;
 };
 
 #define amdgpu_ring_parse_cs(r, p, job, ib) ((r)->funcs->parse_cs((p), (job), (ib)))
@@ -309,19 +435,27 @@ struct amdgpu_ring {
 #define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r))
 #define amdgpu_ring_emit_switch_buffer(r) (r)->funcs->emit_switch_buffer((r))
 #define amdgpu_ring_emit_cntxcntl(r, d) (r)->funcs->emit_cntxcntl((r), (d))
+#define amdgpu_ring_emit_gfx_shadow(r, s, c, g, i, v) ((r)->funcs->emit_gfx_shadow((r), (s), (c), (g), (i), (v)))
 #define amdgpu_ring_emit_rreg(r, d, o) (r)->funcs->emit_rreg((r), (d), (o))
 #define amdgpu_ring_emit_wreg(r, d, v) (r)->funcs->emit_wreg((r), (d), (v))
 #define amdgpu_ring_emit_reg_wait(r, d, v, m) (r)->funcs->emit_reg_wait((r), (d), (v), (m))
 #define amdgpu_ring_emit_reg_write_reg_wait(r, d0, d1, v, m) (r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m))
 #define amdgpu_ring_emit_frame_cntl(r, b, s) (r)->funcs->emit_frame_cntl((r), (b), (s))
 #define amdgpu_ring_pad_ib(r, ib) ((r)->funcs->pad_ib((r), (ib)))
-#define amdgpu_ring_init_cond_exec(r) (r)->funcs->init_cond_exec((r))
-#define amdgpu_ring_patch_cond_exec(r,o) (r)->funcs->patch_cond_exec((r),(o))
+#define amdgpu_ring_init_cond_exec(r, a) (r)->funcs->init_cond_exec((r), (a))
 #define amdgpu_ring_preempt_ib(r) (r)->funcs->preempt_ib(r)
+#define amdgpu_ring_patch_cntl(r, o) ((r)->funcs->patch_cntl((r), (o)))
+#define amdgpu_ring_patch_ce(r, o) ((r)->funcs->patch_ce((r), (o)))
+#define amdgpu_ring_patch_de(r, o) ((r)->funcs->patch_de((r), (o)))
+#define amdgpu_ring_reset(r, v, f) (r)->funcs->reset((r), (v), (f))
 
+unsigned int amdgpu_ring_max_ibs(enum amdgpu_ring_type type);
 int amdgpu_ring_alloc(struct amdgpu_ring *ring, unsigned ndw);
 void amdgpu_ring_ib_begin(struct amdgpu_ring *ring);
 void amdgpu_ring_ib_end(struct amdgpu_ring *ring);
+void amdgpu_ring_ib_on_emit_cntl(struct amdgpu_ring *ring);
+void amdgpu_ring_ib_on_emit_ce(struct amdgpu_ring *ring);
+void amdgpu_ring_ib_on_emit_de(struct amdgpu_ring *ring);
 
 void amdgpu_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count);
 void amdgpu_ring_generic_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib);
@@ -346,16 +480,11 @@ static inline void amdgpu_ring_set_preempt_cond_exec(struct amdgpu_ring *ring,
 
 static inline void amdgpu_ring_clear_ring(struct amdgpu_ring *ring)
 {
-	int i = 0;
-	while (i <= ring->buf_mask)
-		ring->ring[i++] = ring->funcs->nop;
-
+	memset32(ring->ring, ring->funcs->nop, ring->buf_mask + 1);
 }
 
 static inline void amdgpu_ring_write(struct amdgpu_ring *ring, uint32_t v)
 {
-	if (ring->count_dw <= 0)
-		DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n");
 	ring->ring[ring->wptr++ & ring->buf_mask] = v;
 	ring->wptr &= ring->ptr_mask;
 	ring->count_dw--;
@@ -365,26 +494,20 @@ static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring,
 					      void *src, int count_dw)
 {
 	unsigned occupied, chunk1, chunk2;
-	void *dst;
-
-	if (unlikely(ring->count_dw < count_dw))
-		DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n");
 
 	occupied = ring->wptr & ring->buf_mask;
-	dst = (void *)&ring->ring[occupied];
 	chunk1 = ring->buf_mask + 1 - occupied;
-	chunk1 = (chunk1 >= count_dw) ? count_dw: chunk1;
+	chunk1 = (chunk1 >= count_dw) ? count_dw : chunk1;
 	chunk2 = count_dw - chunk1;
 	chunk1 <<= 2;
 	chunk2 <<= 2;
 
 	if (chunk1)
-		memcpy(dst, src, chunk1);
+		memcpy(&ring->ring[occupied], src, chunk1);
 
 	if (chunk2) {
 		src += chunk1;
-		dst = (void *)ring->ring;
-		memcpy(dst, src, chunk2);
+		memcpy(ring->ring, src, chunk2);
 	}
 
 	ring->wptr += count_dw;
@@ -392,14 +515,29 @@ static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring,
 	ring->count_dw -= count_dw;
 }
 
-#define amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset)			\
-	(ring->is_mes_queue && ring->mes_ctx ?				\
-	 (ring->mes_ctx->meta_data_gpu_addr + offset) : 0)
+/**
+ * amdgpu_ring_patch_cond_exec - patch dw count of conditional execute
+ * @ring: amdgpu_ring structure
+ * @offset: offset returned by amdgpu_ring_init_cond_exec
+ *
+ * Calculate the dw count and patch it into a cond_exec command.
+ */
+static inline void amdgpu_ring_patch_cond_exec(struct amdgpu_ring *ring,
+					       unsigned int offset)
+{
+	unsigned cur;
+
+	if (!ring->funcs->init_cond_exec)
+		return;
 
-#define amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset)			\
-	(ring->is_mes_queue && ring->mes_ctx ?				\
-	 (void *)((uint8_t *)(ring->mes_ctx->meta_data_ptr) + offset) : \
-	 NULL)
+	WARN_ON(offset > ring->buf_mask);
+	WARN_ON(ring->ring[offset] != 0);
+
+	cur = (ring->wptr - 1) & ring->buf_mask;
+	if (cur < offset)
+		cur += ring->ring_size >> 2;
+	ring->ring[offset] = cur - offset;
+}
 
 int amdgpu_ring_test_helper(struct amdgpu_ring *ring);
 
@@ -423,13 +561,20 @@ int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		  unsigned size,
 		  enum amdgpu_ib_pool_type pool,
 		  struct amdgpu_ib *ib);
-void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib,
-		    struct dma_fence *f);
+void amdgpu_ib_free(struct amdgpu_ib *ib, struct dma_fence *f);
 int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 		       struct amdgpu_ib *ibs, struct amdgpu_job *job,
 		       struct dma_fence **f);
 int amdgpu_ib_pool_init(struct amdgpu_device *adev);
 void amdgpu_ib_pool_fini(struct amdgpu_device *adev);
 int amdgpu_ib_ring_tests(struct amdgpu_device *adev);
-
+bool amdgpu_ring_sched_ready(struct amdgpu_ring *ring);
+void amdgpu_ring_backup_unprocessed_commands(struct amdgpu_ring *ring,
+					     struct amdgpu_fence *guilty_fence);
+void amdgpu_ring_reset_helper_begin(struct amdgpu_ring *ring,
+				    struct amdgpu_fence *guilty_fence);
+int amdgpu_ring_reset_helper_end(struct amdgpu_ring *ring,
+				 struct amdgpu_fence *guilty_fence);
+bool amdgpu_ring_is_reset_type_supported(struct amdgpu_ring *ring,
+					 u32 reset_type);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.c
index 62079f0e3ee8..7e7d6c3865bc 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.c
@@ -105,6 +105,16 @@ static void amdgpu_mux_resubmit_chunks(struct amdgpu_ring_mux *mux)
 				amdgpu_fence_update_start_timestamp(e->ring,
 								    chunk->sync_seq,
 								    ktime_get());
+				if (chunk->sync_seq ==
+					le32_to_cpu(*(e->ring->fence_drv.cpu_addr + 2))) {
+					if (chunk->cntl_offset <= e->ring->buf_mask)
+						amdgpu_ring_patch_cntl(e->ring,
+								       chunk->cntl_offset);
+					if (chunk->ce_offset <= e->ring->buf_mask)
+						amdgpu_ring_patch_ce(e->ring, chunk->ce_offset);
+					if (chunk->de_offset <= e->ring->buf_mask)
+						amdgpu_ring_patch_de(e->ring, chunk->de_offset);
+				}
 				amdgpu_ring_mux_copy_pkt_from_sw_ring(mux, e->ring,
 								      chunk->start,
 								      chunk->end);
@@ -114,7 +124,7 @@ static void amdgpu_mux_resubmit_chunks(struct amdgpu_ring_mux *mux)
 		}
 	}
 
-	del_timer(&mux->resubmit_timer);
+	timer_delete(&mux->resubmit_timer);
 	mux->s_resubmit = false;
 }
 
@@ -125,7 +135,8 @@ static void amdgpu_ring_mux_schedule_resubmit(struct amdgpu_ring_mux *mux)
 
 static void amdgpu_mux_resubmit_fallback(struct timer_list *t)
 {
-	struct amdgpu_ring_mux *mux = from_timer(mux, t, resubmit_timer);
+	struct amdgpu_ring_mux *mux = timer_container_of(mux, t,
+							 resubmit_timer);
 
 	if (!spin_trylock(&mux->lock)) {
 		amdgpu_ring_mux_schedule_resubmit(mux);
@@ -149,9 +160,7 @@ int amdgpu_ring_mux_init(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring,
 	mux->ring_entry_size = entry_size;
 	mux->s_resubmit = false;
 
-	amdgpu_mux_chunk_slab = kmem_cache_create("amdgpu_mux_chunk",
-						  sizeof(struct amdgpu_mux_chunk), 0,
-						  SLAB_HWCACHE_ALIGN, NULL);
+	amdgpu_mux_chunk_slab = KMEM_CACHE(amdgpu_mux_chunk, SLAB_HWCACHE_ALIGN);
 	if (!amdgpu_mux_chunk_slab) {
 		DRM_ERROR("create amdgpu_mux_chunk cache failed\n");
 		return -ENOMEM;
@@ -387,7 +396,7 @@ void amdgpu_sw_ring_ib_begin(struct amdgpu_ring *ring)
 	struct amdgpu_ring_mux *mux = &adev->gfx.muxer;
 
 	WARN_ON(!ring->is_sw_ring);
-	if (ring->hw_prio > AMDGPU_RING_PRIO_DEFAULT) {
+	if (adev->gfx.mcbp && ring->hw_prio > AMDGPU_RING_PRIO_DEFAULT) {
 		if (amdgpu_mcbp_scan(mux) > 0)
 			amdgpu_mcbp_trigger_preempt(mux);
 		return;
@@ -402,11 +411,25 @@ void amdgpu_sw_ring_ib_end(struct amdgpu_ring *ring)
 	struct amdgpu_ring_mux *mux = &adev->gfx.muxer;
 
 	WARN_ON(!ring->is_sw_ring);
-	if (ring->hw_prio > AMDGPU_RING_PRIO_DEFAULT)
+	if (adev->gfx.mcbp && ring->hw_prio > AMDGPU_RING_PRIO_DEFAULT)
 		return;
 	amdgpu_ring_mux_end_ib(mux, ring);
 }
 
+void amdgpu_sw_ring_ib_mark_offset(struct amdgpu_ring *ring, enum amdgpu_ring_mux_offset_type type)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_ring_mux *mux = &adev->gfx.muxer;
+	unsigned offset;
+
+	if (ring->hw_prio > AMDGPU_RING_PRIO_DEFAULT)
+		return;
+
+	offset = ring->wptr & ring->buf_mask;
+
+	amdgpu_ring_mux_ib_mark_offset(mux, ring, offset, type);
+}
+
 void amdgpu_ring_mux_start_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring)
 {
 	struct amdgpu_mux_entry *e;
@@ -429,6 +452,10 @@ void amdgpu_ring_mux_start_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *r
 	}
 
 	chunk->start = ring->wptr;
+	/* the initialized value used to check if they are set by the ib submission*/
+	chunk->cntl_offset = ring->buf_mask + 1;
+	chunk->de_offset = ring->buf_mask + 1;
+	chunk->ce_offset = ring->buf_mask + 1;
 	list_add_tail(&chunk->entry, &e->list);
 }
 
@@ -454,6 +481,41 @@ static void scan_and_remove_signaled_chunk(struct amdgpu_ring_mux *mux, struct a
 	}
 }
 
+void amdgpu_ring_mux_ib_mark_offset(struct amdgpu_ring_mux *mux,
+				    struct amdgpu_ring *ring, u64 offset,
+				    enum amdgpu_ring_mux_offset_type type)
+{
+	struct amdgpu_mux_entry *e;
+	struct amdgpu_mux_chunk *chunk;
+
+	e = amdgpu_ring_mux_sw_entry(mux, ring);
+	if (!e) {
+		DRM_ERROR("cannot find entry!\n");
+		return;
+	}
+
+	chunk = list_last_entry(&e->list, struct amdgpu_mux_chunk, entry);
+	if (!chunk) {
+		DRM_ERROR("cannot find chunk!\n");
+		return;
+	}
+
+	switch (type) {
+	case AMDGPU_MUX_OFFSET_TYPE_CONTROL:
+		chunk->cntl_offset = offset;
+		break;
+	case AMDGPU_MUX_OFFSET_TYPE_DE:
+		chunk->de_offset = offset;
+		break;
+	case AMDGPU_MUX_OFFSET_TYPE_CE:
+		chunk->ce_offset = offset;
+		break;
+	default:
+		DRM_ERROR("invalid type (%d)\n", type);
+		break;
+	}
+}
+
 void amdgpu_ring_mux_end_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring)
 {
 	struct amdgpu_mux_entry *e;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.h
index 4be45fc14954..d3186b570b82 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.h
@@ -50,6 +50,21 @@ struct amdgpu_mux_entry {
 	struct list_head        list;
 };
 
+enum amdgpu_ring_mux_offset_type {
+	AMDGPU_MUX_OFFSET_TYPE_CONTROL,
+	AMDGPU_MUX_OFFSET_TYPE_DE,
+	AMDGPU_MUX_OFFSET_TYPE_CE,
+};
+
+enum ib_complete_status {
+	/* IB not started/reset value, default value. */
+	IB_COMPLETION_STATUS_DEFAULT = 0,
+	/* IB preempted, started but not completed. */
+	IB_COMPLETION_STATUS_PREEMPTED = 1,
+	/* IB completed. */
+	IB_COMPLETION_STATUS_COMPLETED = 2,
+};
+
 struct amdgpu_ring_mux {
 	struct amdgpu_ring      *real_ring;
 
@@ -72,12 +87,18 @@ struct amdgpu_ring_mux {
  * @sync_seq: the fence seqno related with the saved IB.
  * @start:- start location on the software ring.
  * @end:- end location on the software ring.
+ * @control_offset:- the PRE_RESUME bit position used for resubmission.
+ * @de_offset:- the anchor in write_data for de meta of resubmission.
+ * @ce_offset:- the anchor in write_data for ce meta of resubmission.
  */
 struct amdgpu_mux_chunk {
 	struct list_head        entry;
 	uint32_t                sync_seq;
 	u64                     start;
 	u64                     end;
+	u64                     cntl_offset;
+	u64                     de_offset;
+	u64                     ce_offset;
 };
 
 int amdgpu_ring_mux_init(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring,
@@ -89,6 +110,8 @@ u64 amdgpu_ring_mux_get_wptr(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ri
 u64 amdgpu_ring_mux_get_rptr(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring);
 void amdgpu_ring_mux_start_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring);
 void amdgpu_ring_mux_end_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring);
+void amdgpu_ring_mux_ib_mark_offset(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring,
+				    u64 offset, enum amdgpu_ring_mux_offset_type type);
 bool amdgpu_mcbp_handle_trailing_fence_irq(struct amdgpu_ring_mux *mux);
 
 u64 amdgpu_sw_ring_get_rptr_gfx(struct amdgpu_ring *ring);
@@ -97,6 +120,7 @@ void amdgpu_sw_ring_set_wptr_gfx(struct amdgpu_ring *ring);
 void amdgpu_sw_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count);
 void amdgpu_sw_ring_ib_begin(struct amdgpu_ring *ring);
 void amdgpu_sw_ring_ib_end(struct amdgpu_ring *ring);
+void amdgpu_sw_ring_ib_mark_offset(struct amdgpu_ring *ring, enum amdgpu_ring_mux_offset_type type);
 const char *amdgpu_sw_ring_name(int idx);
 unsigned int amdgpu_sw_ring_priority(int idx);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c
index 85fb730d9fc8..5aa830a02d80 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c
@@ -31,12 +31,13 @@
  * amdgpu_gfx_rlc_enter_safe_mode - Set RLC into safe mode
  *
  * @adev: amdgpu_device pointer
+ * @xcc_id: xcc accelerated compute core id
  *
  * Set RLC enter into safe mode if RLC is enabled and haven't in safe mode.
  */
-void amdgpu_gfx_rlc_enter_safe_mode(struct amdgpu_device *adev)
+void amdgpu_gfx_rlc_enter_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
-	if (adev->gfx.rlc.in_safe_mode)
+	if (adev->gfx.rlc.in_safe_mode[xcc_id])
 		return;
 
 	/* if RLC is not enabled, do nothing */
@@ -46,8 +47,8 @@ void amdgpu_gfx_rlc_enter_safe_mode(struct amdgpu_device *adev)
 	if (adev->cg_flags &
 	    (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_MGCG |
 	     AMD_CG_SUPPORT_GFX_3D_CGCG)) {
-		adev->gfx.rlc.funcs->set_safe_mode(adev);
-		adev->gfx.rlc.in_safe_mode = true;
+		adev->gfx.rlc.funcs->set_safe_mode(adev, xcc_id);
+		adev->gfx.rlc.in_safe_mode[xcc_id] = true;
 	}
 }
 
@@ -55,12 +56,13 @@ void amdgpu_gfx_rlc_enter_safe_mode(struct amdgpu_device *adev)
  * amdgpu_gfx_rlc_exit_safe_mode - Set RLC out of safe mode
  *
  * @adev: amdgpu_device pointer
+ * @xcc_id: xcc accelerated compute core id
  *
  * Set RLC exit safe mode if RLC is enabled and have entered into safe mode.
  */
-void amdgpu_gfx_rlc_exit_safe_mode(struct amdgpu_device *adev)
+void amdgpu_gfx_rlc_exit_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
-	if (!(adev->gfx.rlc.in_safe_mode))
+	if (!(adev->gfx.rlc.in_safe_mode[xcc_id]))
 		return;
 
 	/* if RLC is not enabled, do nothing */
@@ -70,8 +72,8 @@ void amdgpu_gfx_rlc_exit_safe_mode(struct amdgpu_device *adev)
 	if (adev->cg_flags &
 	    (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_MGCG |
 	     AMD_CG_SUPPORT_GFX_3D_CGCG)) {
-		adev->gfx.rlc.funcs->unset_safe_mode(adev);
-		adev->gfx.rlc.in_safe_mode = false;
+		adev->gfx.rlc.funcs->unset_safe_mode(adev, xcc_id);
+		adev->gfx.rlc.in_safe_mode[xcc_id] = false;
 	}
 }
 
@@ -87,7 +89,7 @@ void amdgpu_gfx_rlc_exit_safe_mode(struct amdgpu_device *adev)
 int amdgpu_gfx_rlc_init_sr(struct amdgpu_device *adev, u32 dws)
 {
 	const u32 *src_ptr;
-	volatile u32 *dst_ptr;
+	u32 *dst_ptr;
 	u32 i;
 	int r;
 
@@ -187,7 +189,7 @@ int amdgpu_gfx_rlc_init_cpt(struct amdgpu_device *adev)
 void amdgpu_gfx_rlc_setup_cp_table(struct amdgpu_device *adev)
 {
 	const __le32 *fw_data;
-	volatile u32 *dst_ptr;
+	u32 *dst_ptr;
 	int me, i, max_me;
 	u32 bo_offset = 0;
 	u32 table_offset, table_size;
@@ -239,7 +241,7 @@ void amdgpu_gfx_rlc_setup_cp_table(struct amdgpu_device *adev)
 			table_size = le32_to_cpu(hdr->jt_size);
 		}
 
-		for (i = 0; i < table_size; i ++) {
+		for (i = 0; i < table_size; i++) {
 			dst_ptr[bo_offset + i] =
 				cpu_to_le32(le32_to_cpu(fw_data[table_offset + i]));
 		}
@@ -529,13 +531,12 @@ int amdgpu_gfx_rlc_init_microcode(struct amdgpu_device *adev,
 	if (version_major == 2 && version_minor == 1)
 		adev->gfx.rlc.is_rlc_v2_1 = true;
 
-	if (version_minor >= 0) {
-		err = amdgpu_gfx_rlc_init_microcode_v2_0(adev);
-		if (err) {
-			dev_err(adev->dev, "fail to init rlc v2_0 microcode\n");
-			return err;
-		}
+	err = amdgpu_gfx_rlc_init_microcode_v2_0(adev);
+	if (err) {
+		dev_err(adev->dev, "fail to init rlc v2_0 microcode\n");
+		return err;
 	}
+
 	if (version_minor >= 1)
 		amdgpu_gfx_rlc_init_microcode_v2_1(adev);
 	if (version_minor >= 2)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.h
index 23f060db9255..2ce310b31942 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.h
@@ -26,6 +26,8 @@
 
 #include "clearstate_defs.h"
 
+#define AMDGPU_MAX_RLC_INSTANCES	8
+
 /* firmware ID used in rlc toc */
 typedef enum _FIRMWARE_ID_ {
 	FIRMWARE_ID_INVALID					= 0,
@@ -110,6 +112,53 @@ typedef enum _SOC21_FIRMWARE_ID_ {
     SOC21_FIRMWARE_ID_MAX                         = 37
 } SOC21_FIRMWARE_ID;
 
+typedef enum _SOC24_FIRMWARE_ID_ {
+    SOC24_FIRMWARE_ID_INVALID                     = 0,
+    SOC24_FIRMWARE_ID_RLC_G_UCODE                 = 1,
+    SOC24_FIRMWARE_ID_RLC_TOC                     = 2,
+    SOC24_FIRMWARE_ID_RLCG_SCRATCH                = 3,
+    SOC24_FIRMWARE_ID_RLC_SRM_ARAM                = 4,
+    SOC24_FIRMWARE_ID_RLC_P_UCODE                 = 5,
+    SOC24_FIRMWARE_ID_RLC_V_UCODE                 = 6,
+    SOC24_FIRMWARE_ID_RLX6_UCODE                  = 7,
+    SOC24_FIRMWARE_ID_RLX6_UCODE_CORE1            = 8,
+    SOC24_FIRMWARE_ID_RLX6_DRAM_BOOT              = 9,
+    SOC24_FIRMWARE_ID_RLX6_DRAM_BOOT_CORE1        = 10,
+    SOC24_FIRMWARE_ID_SDMA_UCODE_TH0              = 11,
+    SOC24_FIRMWARE_ID_SDMA_UCODE_TH1              = 12,
+    SOC24_FIRMWARE_ID_CP_PFP                      = 13,
+    SOC24_FIRMWARE_ID_CP_ME                       = 14,
+    SOC24_FIRMWARE_ID_CP_MEC                      = 15,
+    SOC24_FIRMWARE_ID_RS64_MES_P0                 = 16,
+    SOC24_FIRMWARE_ID_RS64_MES_P1                 = 17,
+    SOC24_FIRMWARE_ID_RS64_PFP                    = 18,
+    SOC24_FIRMWARE_ID_RS64_ME                     = 19,
+    SOC24_FIRMWARE_ID_RS64_MEC                    = 20,
+    SOC24_FIRMWARE_ID_RS64_MES_P0_STACK           = 21,
+    SOC24_FIRMWARE_ID_RS64_MES_P1_STACK           = 22,
+    SOC24_FIRMWARE_ID_RS64_PFP_P0_STACK           = 23,
+    SOC24_FIRMWARE_ID_RS64_PFP_P1_STACK           = 24,
+    SOC24_FIRMWARE_ID_RS64_ME_P0_STACK            = 25,
+    SOC24_FIRMWARE_ID_RS64_ME_P1_STACK            = 26,
+    SOC24_FIRMWARE_ID_RS64_MEC_P0_STACK           = 27,
+    SOC24_FIRMWARE_ID_RS64_MEC_P1_STACK           = 28,
+    SOC24_FIRMWARE_ID_RS64_MEC_P2_STACK           = 29,
+    SOC24_FIRMWARE_ID_RS64_MEC_P3_STACK           = 30,
+    SOC24_FIRMWARE_ID_RLC_SRM_DRAM_SR             = 31,
+    SOC24_FIRMWARE_ID_RLCG_SCRATCH_SR             = 32,
+    SOC24_FIRMWARE_ID_RLCP_SCRATCH_SR             = 33,
+    SOC24_FIRMWARE_ID_RLCV_SCRATCH_SR             = 34,
+    SOC24_FIRMWARE_ID_RLX6_DRAM_SR                = 35,
+    SOC24_FIRMWARE_ID_RLX6_DRAM_SR_CORE1          = 36,
+    SOC24_FIRMWARE_ID_RLCDEBUGLOG                 = 37,
+    SOC24_FIRMWARE_ID_SRIOV_DEBUG                 = 38,
+    SOC24_FIRMWARE_ID_SRIOV_CSA_RLC               = 39,
+    SOC24_FIRMWARE_ID_SRIOV_CSA_SDMA              = 40,
+    SOC24_FIRMWARE_ID_SRIOV_CSA_CP                = 41,
+    SOC24_FIRMWARE_ID_UMF_ZONE_PAD                = 42,
+    SOC24_FIRMWARE_ID_MAX                         = 43
+} SOC24_FIRMWARE_ID;
+
 typedef struct _RLC_TABLE_OF_CONTENT {
 	union {
 		unsigned int	DW0;
@@ -153,21 +202,62 @@ typedef struct _RLC_TABLE_OF_CONTENT {
 	};
 } RLC_TABLE_OF_CONTENT;
 
+typedef struct _RLC_TABLE_OF_CONTENT_V2 {
+	union {
+		unsigned int    DW0;
+		struct {
+			uint32_t offset         : 25;
+			uint32_t id             : 7;
+		};
+	};
+
+	union {
+		unsigned int    DW1;
+		struct {
+			uint32_t reserved0              : 1;
+			uint32_t reserved1              : 1;
+			uint32_t reserved2              : 1;
+			uint32_t memory_destination     : 2;
+			uint32_t vfflr_image_code       : 4;
+			uint32_t reserved9              : 1;
+			uint32_t reserved10             : 1;
+			uint32_t reserved11             : 1;
+			uint32_t size_x16               : 1;
+			uint32_t reserved13             : 1;
+			uint32_t size                   : 18;
+		};
+	};
+} RLC_TABLE_OF_CONTENT_V2;
+
 #define RLC_TOC_MAX_SIZE		64
 
 struct amdgpu_rlc_funcs {
 	bool (*is_rlc_enabled)(struct amdgpu_device *adev);
-	void (*set_safe_mode)(struct amdgpu_device *adev);
-	void (*unset_safe_mode)(struct amdgpu_device *adev);
+	void (*set_safe_mode)(struct amdgpu_device *adev, int xcc_id);
+	void (*unset_safe_mode)(struct amdgpu_device *adev, int xcc_id);
 	int  (*init)(struct amdgpu_device *adev);
 	u32  (*get_csb_size)(struct amdgpu_device *adev);
-	void (*get_csb_buffer)(struct amdgpu_device *adev, volatile u32 *buffer);
+
+	/**
+	 * @get_csb_buffer: Get the clear state to be put into the hardware.
+	 *
+	 * The parameter adev is used to get the CS data and other gfx info,
+	 * and buffer is the RLC CS pointer
+	 *
+	 * Sometimes, the user space puts a request to clear the state in the
+	 * command buffer; this function provides the clear state that gets put
+	 * into the hardware. Note that the driver programs Clear State
+	 * Indirect Buffer (CSB) explicitly when it sets up the kernel rings,
+	 * and it also provides a pointer to it which is used by the firmware
+	 * to load the clear state in some cases.
+	 */
+	void (*get_csb_buffer)(struct amdgpu_device *adev, u32 *buffer);
 	int  (*get_cp_table_num)(struct amdgpu_device *adev);
 	int  (*resume)(struct amdgpu_device *adev);
 	void (*stop)(struct amdgpu_device *adev);
 	void (*reset)(struct amdgpu_device *adev);
 	void (*start)(struct amdgpu_device *adev);
-	void (*update_spm_vmid)(struct amdgpu_device *adev, unsigned vmid);
+	void (*update_spm_vmid)(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned vmid);
 	bool (*is_rlcg_access_range)(struct amdgpu_device *adev, uint32_t reg);
 };
 
@@ -185,23 +275,23 @@ struct amdgpu_rlc {
 	/* for power gating */
 	struct amdgpu_bo        *save_restore_obj;
 	uint64_t                save_restore_gpu_addr;
-	volatile uint32_t       *sr_ptr;
+	uint32_t		*sr_ptr;
 	const u32               *reg_list;
 	u32                     reg_list_size;
 	/* for clear state */
 	struct amdgpu_bo        *clear_state_obj;
 	uint64_t                clear_state_gpu_addr;
-	volatile uint32_t       *cs_ptr;
+	uint32_t		*cs_ptr;
 	const struct cs_section_def   *cs_data;
 	u32                     clear_state_size;
 	/* for cp tables */
 	struct amdgpu_bo        *cp_table_obj;
 	uint64_t                cp_table_gpu_addr;
-	volatile uint32_t       *cp_table_ptr;
+	uint32_t		*cp_table_ptr;
 	u32                     cp_table_size;
 
 	/* safe mode for updating CG/PG state */
-	bool in_safe_mode;
+	bool in_safe_mode[AMDGPU_MAX_RLC_INSTANCES];
 	const struct amdgpu_rlc_funcs *funcs;
 
 	/* for firmware data */
@@ -257,11 +347,11 @@ struct amdgpu_rlc {
 
 	bool rlcg_reg_access_supported;
 	/* registers for rlcg indirect reg access */
-	struct amdgpu_rlcg_reg_access_ctrl reg_access_ctrl;
+	struct amdgpu_rlcg_reg_access_ctrl reg_access_ctrl[AMDGPU_MAX_RLC_INSTANCES];
 };
 
-void amdgpu_gfx_rlc_enter_safe_mode(struct amdgpu_device *adev);
-void amdgpu_gfx_rlc_exit_safe_mode(struct amdgpu_device *adev);
+void amdgpu_gfx_rlc_enter_safe_mode(struct amdgpu_device *adev, int xcc_id);
+void amdgpu_gfx_rlc_exit_safe_mode(struct amdgpu_device *adev, int xcc_id);
 int amdgpu_gfx_rlc_init_sr(struct amdgpu_device *adev, u32 dws);
 int amdgpu_gfx_rlc_init_csb(struct amdgpu_device *adev);
 int amdgpu_gfx_rlc_init_cpt(struct amdgpu_device *adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c
index c6b4337eb20c..39070b2a4c04 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c
@@ -81,7 +81,7 @@ int amdgpu_sa_bo_new(struct amdgpu_sa_manager *sa_manager,
 		     unsigned int size)
 {
 	struct drm_suballoc *sa = drm_suballoc_new(&sa_manager->base, size,
-						   GFP_KERNEL, true, 0);
+						   GFP_KERNEL, false, 0);
 
 	if (IS_ERR(sa)) {
 		*sa_bo = NULL;
@@ -93,8 +93,7 @@ int amdgpu_sa_bo_new(struct amdgpu_sa_manager *sa_manager,
 	return 0;
 }
 
-void amdgpu_sa_bo_free(struct amdgpu_device *adev, struct drm_suballoc **sa_bo,
-		       struct dma_fence *fence)
+void amdgpu_sa_bo_free(struct drm_suballoc **sa_bo, struct dma_fence *fence)
 {
 	if (sa_bo == NULL || *sa_bo == NULL) {
 		return;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c
index e9b45089a28a..341beec59537 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c
@@ -22,7 +22,6 @@
  * Authors: Andres Rodriguez <andresx7@gmail.com>
  */
 
-#include <linux/fdtable.h>
 #include <linux/file.h>
 #include <linux/pid.h>
 
@@ -36,25 +35,26 @@ static int amdgpu_sched_process_priority_override(struct amdgpu_device *adev,
 						  int fd,
 						  int32_t priority)
 {
-	struct fd f = fdget(fd);
+	CLASS(fd, f)(fd);
 	struct amdgpu_fpriv *fpriv;
+	struct amdgpu_ctx_mgr *mgr;
 	struct amdgpu_ctx *ctx;
 	uint32_t id;
 	int r;
 
-	if (!f.file)
+	if (fd_empty(f))
 		return -EINVAL;
 
-	r = amdgpu_file_to_fpriv(f.file, &fpriv);
-	if (r) {
-		fdput(f);
+	r = amdgpu_file_to_fpriv(fd_file(f), &fpriv);
+	if (r)
 		return r;
-	}
 
-	idr_for_each_entry(&fpriv->ctx_mgr.ctx_handles, ctx, id)
+	mgr = &fpriv->ctx_mgr;
+	mutex_lock(&mgr->lock);
+	idr_for_each_entry(&mgr->ctx_handles, ctx, id)
 		amdgpu_ctx_priority_override(ctx, priority);
+	mutex_unlock(&mgr->lock);
 
-	fdput(f);
 	return 0;
 }
 
@@ -63,31 +63,25 @@ static int amdgpu_sched_context_priority_override(struct amdgpu_device *adev,
 						  unsigned ctx_id,
 						  int32_t priority)
 {
-	struct fd f = fdget(fd);
+	CLASS(fd, f)(fd);
 	struct amdgpu_fpriv *fpriv;
 	struct amdgpu_ctx *ctx;
 	int r;
 
-	if (!f.file)
+	if (fd_empty(f))
 		return -EINVAL;
 
-	r = amdgpu_file_to_fpriv(f.file, &fpriv);
-	if (r) {
-		fdput(f);
+	r = amdgpu_file_to_fpriv(fd_file(f), &fpriv);
+	if (r)
 		return r;
-	}
 
 	ctx = amdgpu_ctx_get(fpriv, ctx_id);
 
-	if (!ctx) {
-		fdput(f);
+	if (!ctx)
 		return -EINVAL;
-	}
 
 	amdgpu_ctx_priority_override(ctx, priority);
 	amdgpu_ctx_put(ctx);
-	fdput(f);
-
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c
index 231ca06bc9c7..8b8a04138711 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c
@@ -25,6 +25,9 @@
 #include "amdgpu.h"
 #include "amdgpu_sdma.h"
 #include "amdgpu_ras.h"
+#include "amdgpu_reset.h"
+#include "gc/gc_10_1_0_offset.h"
+#include "gc/gc_10_3_0_sh_mask.h"
 
 #define AMDGPU_CSA_SDMA_SIZE 64
 /* SDMA CSA reside in the 3rd page of CSA */
@@ -64,7 +67,7 @@ int amdgpu_sdma_get_index_from_ring(struct amdgpu_ring *ring, uint32_t *index)
 }
 
 uint64_t amdgpu_sdma_get_csa_mc_addr(struct amdgpu_ring *ring,
-				     unsigned vmid)
+				     unsigned int vmid)
 {
 	struct amdgpu_device *adev = ring->adev;
 	uint64_t csa_mc_addr;
@@ -72,25 +75,17 @@ uint64_t amdgpu_sdma_get_csa_mc_addr(struct amdgpu_ring *ring,
 	int r;
 
 	/* don't enable OS preemption on SDMA under SRIOV */
-	if (amdgpu_sriov_vf(adev) || vmid == 0 || !amdgpu_mcbp)
+	if (amdgpu_sriov_vf(adev) || vmid == 0 || !adev->gfx.mcbp)
 		return 0;
 
-	if (ring->is_mes_queue) {
-		uint32_t offset = 0;
+	r = amdgpu_sdma_get_index_from_ring(ring, &index);
 
-		offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				  sdma[ring->idx].sdma_meta_data);
-		csa_mc_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-	} else {
-		r = amdgpu_sdma_get_index_from_ring(ring, &index);
-
-		if (r || index > 31)
-			csa_mc_addr = 0;
-		else
-			csa_mc_addr = amdgpu_csa_vaddr(adev) +
-				AMDGPU_CSA_SDMA_OFFSET +
-				index * AMDGPU_CSA_SDMA_SIZE;
-	}
+	if (r || index > 31)
+		csa_mc_addr = 0;
+	else
+		csa_mc_addr = amdgpu_csa_vaddr(adev) +
+			AMDGPU_CSA_SDMA_OFFSET +
+			index * AMDGPU_CSA_SDMA_SIZE;
 
 	return csa_mc_addr;
 }
@@ -158,6 +153,7 @@ static int amdgpu_sdma_init_inst_ctx(struct amdgpu_sdma_instance *sdma_inst)
 	const struct common_firmware_header *header = NULL;
 	const struct sdma_firmware_header_v1_0 *hdr;
 	const struct sdma_firmware_header_v2_0 *hdr_v2;
+	const struct sdma_firmware_header_v3_0 *hdr_v3;
 
 	header = (const struct common_firmware_header *)
 		sdma_inst->fw->data;
@@ -174,6 +170,11 @@ static int amdgpu_sdma_init_inst_ctx(struct amdgpu_sdma_instance *sdma_inst)
 		sdma_inst->fw_version = le32_to_cpu(hdr_v2->header.ucode_version);
 		sdma_inst->feature_version = le32_to_cpu(hdr_v2->ucode_feature_version);
 		break;
+	case 3:
+		hdr_v3 = (const struct sdma_firmware_header_v3_0 *)sdma_inst->fw->data;
+		sdma_inst->fw_version = le32_to_cpu(hdr_v3->header.ucode_version);
+		sdma_inst->feature_version = le32_to_cpu(hdr_v3->ucode_feature_version);
+		break;
 	default:
 		return -EINVAL;
 	}
@@ -206,16 +207,19 @@ int amdgpu_sdma_init_microcode(struct amdgpu_device *adev,
 	const struct common_firmware_header *header = NULL;
 	int err, i;
 	const struct sdma_firmware_header_v2_0 *sdma_hdr;
+	const struct sdma_firmware_header_v3_0 *sdma_hv3;
 	uint16_t version_major;
 	char ucode_prefix[30];
-	char fw_name[40];
 
 	amdgpu_ucode_ip_version_decode(adev, SDMA0_HWIP, ucode_prefix, sizeof(ucode_prefix));
 	if (instance == 0)
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s.bin", ucode_prefix);
+		err = amdgpu_ucode_request(adev, &adev->sdma.instance[instance].fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s.bin", ucode_prefix);
 	else
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s%d.bin", ucode_prefix, instance);
-	err = amdgpu_ucode_request(adev, &adev->sdma.instance[instance].fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->sdma.instance[instance].fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s%d.bin", ucode_prefix, instance);
 	if (err)
 		goto out;
 
@@ -239,9 +243,6 @@ int amdgpu_sdma_init_microcode(struct amdgpu_device *adev,
 			       sizeof(struct amdgpu_sdma_instance));
 	}
 
-	if (amdgpu_sriov_vf(adev))
-		return 0;
-
 	DRM_DEBUG("psp_load == '%s'\n",
 		  adev->firmware.load_type == AMDGPU_FW_LOAD_PSP ? "true" : "false");
 
@@ -252,6 +253,19 @@ int amdgpu_sdma_init_microcode(struct amdgpu_device *adev,
 				if (!duplicate && (instance != i))
 					continue;
 				else {
+					/* Use a single copy per SDMA firmware type. PSP uses the same instance for all
+					 * groups of SDMAs */
+					if ((amdgpu_ip_version(adev, SDMA0_HWIP, 0) ==
+						IP_VERSION(4, 4, 2) ||
+					     amdgpu_ip_version(adev, SDMA0_HWIP, 0) ==
+						IP_VERSION(4, 4, 4) ||
+					     amdgpu_ip_version(adev, SDMA0_HWIP, 0) ==
+						IP_VERSION(4, 4, 5)) &&
+					    adev->firmware.load_type ==
+						AMDGPU_FW_LOAD_PSP &&
+					    adev->sdma.num_inst_per_aid == i) {
+						break;
+					}
 					info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SDMA0 + i];
 					info->ucode_id = AMDGPU_UCODE_ID_SDMA0 + i;
 					info->fw = adev->sdma.instance[i].fw;
@@ -274,6 +288,15 @@ int amdgpu_sdma_init_microcode(struct amdgpu_device *adev,
 			adev->firmware.fw_size +=
 				ALIGN(le32_to_cpu(sdma_hdr->ctl_ucode_size_bytes), PAGE_SIZE);
 			break;
+		case 3:
+			sdma_hv3 = (const struct sdma_firmware_header_v3_0 *)
+				adev->sdma.instance[0].fw->data;
+			info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SDMA_RS64];
+			info->ucode_id = AMDGPU_UCODE_ID_SDMA_RS64;
+			info->fw = adev->sdma.instance[0].fw;
+			adev->firmware.fw_size +=
+				ALIGN(le32_to_cpu(sdma_hv3->ucode_size_bytes), PAGE_SIZE);
+			break;
 		default:
 			err = -EINVAL;
 		}
@@ -285,27 +308,6 @@ out:
 	return err;
 }
 
-void amdgpu_sdma_unset_buffer_funcs_helper(struct amdgpu_device *adev)
-{
-	struct amdgpu_ring *sdma;
-	int i;
-
-	for (i = 0; i < adev->sdma.num_instances; i++) {
-		if (adev->sdma.has_page_queue) {
-			sdma = &adev->sdma.instance[i].page;
-			if (adev->mman.buffer_funcs_ring == sdma) {
-				amdgpu_ttm_set_buffer_funcs_status(adev, false);
-				break;
-			}
-		}
-		sdma = &adev->sdma.instance[i].ring;
-		if (adev->mman.buffer_funcs_ring == sdma) {
-			amdgpu_ttm_set_buffer_funcs_status(adev, false);
-			break;
-		}
-	}
-}
-
 int amdgpu_sdma_ras_sw_init(struct amdgpu_device *adev)
 {
 	int err = 0;
@@ -340,3 +342,270 @@ int amdgpu_sdma_ras_sw_init(struct amdgpu_device *adev)
 
 	return 0;
 }
+
+/*
+ * debugfs for to enable/disable sdma job submission to specific core.
+ */
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_debugfs_sdma_sched_mask_set(void *data, u64 val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u64 i, num_ring;
+	u64 mask = 0;
+	struct amdgpu_ring *ring, *page = NULL;
+
+	if (!adev)
+		return -ENODEV;
+
+	/* Determine the number of rings per SDMA instance
+	 * (1 for sdma gfx ring, 2 if page queue exists)
+	 */
+	if (adev->sdma.has_page_queue)
+		num_ring = 2;
+	else
+		num_ring = 1;
+
+	/* Calculate the maximum possible mask value
+	 * based on the number of SDMA instances and rings
+	*/
+	mask = BIT_ULL(adev->sdma.num_instances * num_ring) - 1;
+
+	if ((val & mask) == 0)
+		return -EINVAL;
+
+	for (i = 0; i < adev->sdma.num_instances; ++i) {
+		ring = &adev->sdma.instance[i].ring;
+		if (adev->sdma.has_page_queue)
+			page = &adev->sdma.instance[i].page;
+		if (val & BIT_ULL(i * num_ring))
+			ring->sched.ready = true;
+		else
+			ring->sched.ready = false;
+
+		if (page) {
+			if (val & BIT_ULL(i * num_ring + 1))
+				page->sched.ready = true;
+			else
+				page->sched.ready = false;
+		}
+	}
+	/* publish sched.ready flag update effective immediately across smp */
+	smp_rmb();
+	return 0;
+}
+
+static int amdgpu_debugfs_sdma_sched_mask_get(void *data, u64 *val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u64 i, num_ring;
+	u64 mask = 0;
+	struct amdgpu_ring *ring, *page = NULL;
+
+	if (!adev)
+		return -ENODEV;
+
+	/* Determine the number of rings per SDMA instance
+	 * (1 for sdma gfx ring, 2 if page queue exists)
+	 */
+	if (adev->sdma.has_page_queue)
+		num_ring = 2;
+	else
+		num_ring = 1;
+
+	for (i = 0; i < adev->sdma.num_instances; ++i) {
+		ring = &adev->sdma.instance[i].ring;
+		if (adev->sdma.has_page_queue)
+			page = &adev->sdma.instance[i].page;
+
+		if (ring->sched.ready)
+			mask |= BIT_ULL(i * num_ring);
+		else
+			mask &= ~BIT_ULL(i * num_ring);
+
+		if (page) {
+			if (page->sched.ready)
+				mask |= BIT_ULL(i * num_ring + 1);
+			else
+				mask &= ~BIT_ULL(i * num_ring + 1);
+		}
+	}
+
+	*val = mask;
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(amdgpu_debugfs_sdma_sched_mask_fops,
+			 amdgpu_debugfs_sdma_sched_mask_get,
+			 amdgpu_debugfs_sdma_sched_mask_set, "%llx\n");
+
+#endif
+
+void amdgpu_debugfs_sdma_sched_mask_init(struct amdgpu_device *adev)
+{
+#if defined(CONFIG_DEBUG_FS)
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+	struct dentry *root = minor->debugfs_root;
+	char name[32];
+
+	if (!(adev->sdma.num_instances > 1))
+		return;
+	sprintf(name, "amdgpu_sdma_sched_mask");
+	debugfs_create_file(name, 0600, root, adev,
+			    &amdgpu_debugfs_sdma_sched_mask_fops);
+#endif
+}
+
+static ssize_t amdgpu_get_sdma_reset_mask(struct device *dev,
+						struct device_attribute *attr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	if (!adev)
+		return -ENODEV;
+
+	return amdgpu_show_reset_mask(buf, adev->sdma.supported_reset);
+}
+
+static DEVICE_ATTR(sdma_reset_mask, 0444,
+		   amdgpu_get_sdma_reset_mask, NULL);
+
+int amdgpu_sdma_sysfs_reset_mask_init(struct amdgpu_device *adev)
+{
+	int r = 0;
+
+	if (!amdgpu_gpu_recovery)
+		return r;
+
+	if (adev->sdma.num_instances) {
+		r = device_create_file(adev->dev, &dev_attr_sdma_reset_mask);
+		if (r)
+			return r;
+	}
+
+	return r;
+}
+
+void amdgpu_sdma_sysfs_reset_mask_fini(struct amdgpu_device *adev)
+{
+	if (!amdgpu_gpu_recovery)
+		return;
+
+	if (adev->dev->kobj.sd) {
+		if (adev->sdma.num_instances)
+			device_remove_file(adev->dev, &dev_attr_sdma_reset_mask);
+	}
+}
+
+struct amdgpu_ring *amdgpu_sdma_get_shared_ring(struct amdgpu_device *adev, struct amdgpu_ring *ring)
+{
+	if (adev->sdma.has_page_queue &&
+	    (ring->me < adev->sdma.num_instances) &&
+	    (ring == &adev->sdma.instance[ring->me].ring))
+		return &adev->sdma.instance[ring->me].page;
+	else
+		return NULL;
+}
+
+/**
+* amdgpu_sdma_is_shared_inv_eng - Check if a ring is an SDMA ring that shares a VM invalidation engine
+* @adev: Pointer to the AMDGPU device structure
+* @ring: Pointer to the ring structure to check
+*
+* This function checks if the given ring is an SDMA ring that shares a VM invalidation engine.
+* It returns true if the ring is such an SDMA ring, false otherwise.
+*/
+bool amdgpu_sdma_is_shared_inv_eng(struct amdgpu_device *adev, struct amdgpu_ring *ring)
+{
+	int i = ring->me;
+
+	if (!adev->sdma.has_page_queue || i >= adev->sdma.num_instances)
+		return false;
+
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 4) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 5, 0))
+		return (ring == &adev->sdma.instance[i].page);
+	else
+		return false;
+}
+
+static int amdgpu_sdma_soft_reset(struct amdgpu_device *adev, u32 instance_id)
+{
+	struct amdgpu_sdma_instance *sdma_instance = &adev->sdma.instance[instance_id];
+
+	if (sdma_instance->funcs->soft_reset_kernel_queue)
+		return sdma_instance->funcs->soft_reset_kernel_queue(adev, instance_id);
+
+	return -EOPNOTSUPP;
+}
+
+/**
+ * amdgpu_sdma_reset_engine - Reset a specific SDMA engine
+ * @adev: Pointer to the AMDGPU device
+ * @instance_id: Logical ID of the SDMA engine instance to reset
+ * @caller_handles_kernel_queues: Skip kernel queue processing. Caller
+ * will handle it.
+ *
+ * Returns: 0 on success, or a negative error code on failure.
+ */
+int amdgpu_sdma_reset_engine(struct amdgpu_device *adev, uint32_t instance_id,
+			     bool caller_handles_kernel_queues)
+{
+	int ret = 0;
+	struct amdgpu_sdma_instance *sdma_instance = &adev->sdma.instance[instance_id];
+	struct amdgpu_ring *gfx_ring = &sdma_instance->ring;
+	struct amdgpu_ring *page_ring = &sdma_instance->page;
+
+	mutex_lock(&sdma_instance->engine_reset_mutex);
+
+	if (!caller_handles_kernel_queues) {
+		/* Stop the scheduler's work queue for the GFX and page rings if they are running.
+		 * This ensures that no new tasks are submitted to the queues while
+		 * the reset is in progress.
+		 */
+		drm_sched_wqueue_stop(&gfx_ring->sched);
+
+		if (adev->sdma.has_page_queue)
+			drm_sched_wqueue_stop(&page_ring->sched);
+	}
+
+	if (sdma_instance->funcs->stop_kernel_queue) {
+		sdma_instance->funcs->stop_kernel_queue(gfx_ring);
+		if (adev->sdma.has_page_queue)
+			sdma_instance->funcs->stop_kernel_queue(page_ring);
+	}
+
+	/* Perform the SDMA reset for the specified instance */
+	ret = amdgpu_sdma_soft_reset(adev, instance_id);
+	if (ret) {
+		dev_err(adev->dev, "Failed to reset SDMA logical instance %u\n", instance_id);
+		goto exit;
+	}
+
+	if (sdma_instance->funcs->start_kernel_queue) {
+		sdma_instance->funcs->start_kernel_queue(gfx_ring);
+		if (adev->sdma.has_page_queue)
+			sdma_instance->funcs->start_kernel_queue(page_ring);
+	}
+
+exit:
+	if (!caller_handles_kernel_queues) {
+		/* Restart the scheduler's work queue for the GFX and page rings
+		 * if they were stopped by this function. This allows new tasks
+		 * to be submitted to the queues after the reset is complete.
+		 */
+		if (!ret) {
+			amdgpu_fence_driver_force_completion(gfx_ring);
+			drm_sched_wqueue_start(&gfx_ring->sched);
+			if (adev->sdma.has_page_queue) {
+				amdgpu_fence_driver_force_completion(page_ring);
+				drm_sched_wqueue_start(&page_ring->sched);
+			}
+		}
+	}
+	mutex_unlock(&sdma_instance->engine_reset_mutex);
+
+	return ret;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h
index fc8528812598..34311f32be4c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h
@@ -26,7 +26,7 @@
 #include "amdgpu_ras.h"
 
 /* max number of IP instances */
-#define AMDGPU_MAX_SDMA_INSTANCES		8
+#define AMDGPU_MAX_SDMA_INSTANCES		16
 
 enum amdgpu_sdma_irq {
 	AMDGPU_SDMA_IRQ_INSTANCE0  = 0,
@@ -37,9 +37,25 @@ enum amdgpu_sdma_irq {
 	AMDGPU_SDMA_IRQ_INSTANCE5,
 	AMDGPU_SDMA_IRQ_INSTANCE6,
 	AMDGPU_SDMA_IRQ_INSTANCE7,
+	AMDGPU_SDMA_IRQ_INSTANCE8,
+	AMDGPU_SDMA_IRQ_INSTANCE9,
+	AMDGPU_SDMA_IRQ_INSTANCE10,
+	AMDGPU_SDMA_IRQ_INSTANCE11,
+	AMDGPU_SDMA_IRQ_INSTANCE12,
+	AMDGPU_SDMA_IRQ_INSTANCE13,
+	AMDGPU_SDMA_IRQ_INSTANCE14,
+	AMDGPU_SDMA_IRQ_INSTANCE15,
 	AMDGPU_SDMA_IRQ_LAST
 };
 
+#define NUM_SDMA(x) hweight32(x)
+
+struct amdgpu_sdma_funcs {
+	int (*stop_kernel_queue)(struct amdgpu_ring *ring);
+	int (*start_kernel_queue)(struct amdgpu_ring *ring);
+	int (*soft_reset_kernel_queue)(struct amdgpu_device *adev, u32 instance_id);
+};
+
 struct amdgpu_sdma_instance {
 	/* SDMA firmware */
 	const struct firmware	*fw;
@@ -49,6 +65,44 @@ struct amdgpu_sdma_instance {
 	struct amdgpu_ring	ring;
 	struct amdgpu_ring	page;
 	bool			burst_nop;
+	uint32_t		aid_id;
+
+	struct amdgpu_bo	*sdma_fw_obj;
+	uint64_t		sdma_fw_gpu_addr;
+	uint32_t		*sdma_fw_ptr;
+	struct mutex		engine_reset_mutex;
+	/* track guilty state of GFX and PAGE queues */
+	bool			gfx_guilty;
+	bool			page_guilty;
+	const struct amdgpu_sdma_funcs   *funcs;
+};
+
+enum amdgpu_sdma_ras_memory_id {
+	AMDGPU_SDMA_MBANK_DATA_BUF0 = 1,
+	AMDGPU_SDMA_MBANK_DATA_BUF1 = 2,
+	AMDGPU_SDMA_MBANK_DATA_BUF2 = 3,
+	AMDGPU_SDMA_MBANK_DATA_BUF3 = 4,
+	AMDGPU_SDMA_MBANK_DATA_BUF4 = 5,
+	AMDGPU_SDMA_MBANK_DATA_BUF5 = 6,
+	AMDGPU_SDMA_MBANK_DATA_BUF6 = 7,
+	AMDGPU_SDMA_MBANK_DATA_BUF7 = 8,
+	AMDGPU_SDMA_MBANK_DATA_BUF8 = 9,
+	AMDGPU_SDMA_MBANK_DATA_BUF9 = 10,
+	AMDGPU_SDMA_MBANK_DATA_BUF10 = 11,
+	AMDGPU_SDMA_MBANK_DATA_BUF11 = 12,
+	AMDGPU_SDMA_MBANK_DATA_BUF12 = 13,
+	AMDGPU_SDMA_MBANK_DATA_BUF13 = 14,
+	AMDGPU_SDMA_MBANK_DATA_BUF14 = 15,
+	AMDGPU_SDMA_MBANK_DATA_BUF15 = 16,
+	AMDGPU_SDMA_UCODE_BUF = 17,
+	AMDGPU_SDMA_RB_CMD_BUF = 18,
+	AMDGPU_SDMA_IB_CMD_BUF = 19,
+	AMDGPU_SDMA_UTCL1_RD_FIFO = 20,
+	AMDGPU_SDMA_UTCL1_RDBST_FIFO = 21,
+	AMDGPU_SDMA_UTCL1_WR_FIFO = 22,
+	AMDGPU_SDMA_DATA_LUT_FIFO = 23,
+	AMDGPU_SDMA_SPLIT_DAT_BUF = 24,
+	AMDGPU_SDMA_MEMORY_BLOCK_LAST,
 };
 
 struct amdgpu_sdma_ras {
@@ -59,17 +113,26 @@ struct amdgpu_sdma {
 	struct amdgpu_sdma_instance instance[AMDGPU_MAX_SDMA_INSTANCES];
 	struct amdgpu_irq_src	trap_irq;
 	struct amdgpu_irq_src	illegal_inst_irq;
+	struct amdgpu_irq_src	fence_irq;
 	struct amdgpu_irq_src	ecc_irq;
 	struct amdgpu_irq_src	vm_hole_irq;
 	struct amdgpu_irq_src	doorbell_invalid_irq;
 	struct amdgpu_irq_src	pool_timeout_irq;
 	struct amdgpu_irq_src	srbm_write_irq;
+	struct amdgpu_irq_src	ctxt_empty_irq;
 
 	int			num_instances;
+	uint32_t 		sdma_mask;
+	int			num_inst_per_aid;
 	uint32_t                    srbm_soft_reset;
 	bool			has_page_queue;
 	struct ras_common_if	*ras_if;
 	struct amdgpu_sdma_ras	*ras;
+	uint32_t		*ip_dump;
+	uint32_t 		supported_reset;
+	struct list_head	reset_callback_list;
+	bool			no_user_submission;
+	bool			disable_uq;
 };
 
 /*
@@ -91,7 +154,7 @@ struct amdgpu_buffer_funcs {
 				 uint64_t dst_offset,
 				 /* number of byte to transfer */
 				 uint32_t byte_count,
-				 bool tmz);
+				 uint32_t copy_flags);
 
 	/* maximum bytes in a single operation */
 	uint32_t	fill_max_bytes;
@@ -109,6 +172,9 @@ struct amdgpu_buffer_funcs {
 				 uint32_t byte_count);
 };
 
+int amdgpu_sdma_reset_engine(struct amdgpu_device *adev, uint32_t instance_id,
+			     bool caller_handles_kernel_queues);
+
 #define amdgpu_emit_copy_buffer(adev, ib, s, d, b, t) (adev)->mman.buffer_funcs->emit_copy_buffer((ib),  (s), (d), (b), (t))
 #define amdgpu_emit_fill_buffer(adev, ib, s, d, b) (adev)->mman.buffer_funcs->emit_fill_buffer((ib), (s), (d), (b))
 
@@ -128,7 +194,11 @@ int amdgpu_sdma_init_microcode(struct amdgpu_device *adev, u32 instance,
 			       bool duplicate);
 void amdgpu_sdma_destroy_inst_ctx(struct amdgpu_device *adev,
         bool duplicate);
-void amdgpu_sdma_unset_buffer_funcs_helper(struct amdgpu_device *adev);
 int amdgpu_sdma_ras_sw_init(struct amdgpu_device *adev);
-
+void amdgpu_debugfs_sdma_sched_mask_init(struct amdgpu_device *adev);
+int amdgpu_sdma_sysfs_reset_mask_init(struct amdgpu_device *adev);
+void amdgpu_sdma_sysfs_reset_mask_fini(struct amdgpu_device *adev);
+bool amdgpu_sdma_is_shared_inv_eng(struct amdgpu_device *adev, struct amdgpu_ring *ring);
+struct amdgpu_ring *amdgpu_sdma_get_shared_ring(struct amdgpu_device *adev,
+	struct amdgpu_ring *ring);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_securedisplay.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_securedisplay.c
index 8ed0e073656f..3739be1b71e0 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_securedisplay.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_securedisplay.c
@@ -135,6 +135,10 @@ static ssize_t amdgpu_securedisplay_debugfs_write(struct file *f, const char __u
 		mutex_unlock(&psp->securedisplay_context.mutex);
 		break;
 	case 2:
+		if (size < 3 || phy_id >= TA_SECUREDISPLAY_MAX_PHY) {
+			dev_err(adev->dev, "Invalid input: %s\n", str);
+			return -EINVAL;
+		}
 		mutex_lock(&psp->securedisplay_context.mutex);
 		psp_prep_securedisplay_cmd_buf(psp, &securedisplay_cmd,
 			TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC);
@@ -155,7 +159,6 @@ static ssize_t amdgpu_securedisplay_debugfs_write(struct file *f, const char __u
 		dev_err(adev->dev, "Invalid input: %s\n", str);
 	}
 
-	pm_runtime_mark_last_busy(dev->dev);
 	pm_runtime_put_autosuspend(dev->dev);
 
 	return size;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c
new file mode 100644
index 000000000000..a0b479d5fff1
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c
@@ -0,0 +1,262 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "amdgpu.h"
+#include "amdgpu_seq64.h"
+
+#include <drm/drm_exec.h>
+
+/**
+ * DOC: amdgpu_seq64
+ *
+ * amdgpu_seq64 allocates a 64bit memory on each request in sequence order.
+ * seq64 driver is required for user queue fence memory allocation, TLB
+ * counters and VM updates. It has maximum count of 32768 64 bit slots.
+ */
+
+/**
+ * amdgpu_seq64_get_va_base - Get the seq64 va base address
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Returns:
+ * va base address on success
+ */
+static inline u64 amdgpu_seq64_get_va_base(struct amdgpu_device *adev)
+{
+	u64 addr = AMDGPU_VA_RESERVED_SEQ64_START(adev);
+
+	addr = amdgpu_gmc_sign_extend(addr);
+
+	return addr;
+}
+
+/**
+ * amdgpu_seq64_map - Map the seq64 memory to VM
+ *
+ * @adev: amdgpu_device pointer
+ * @vm: vm pointer
+ * @bo_va: bo_va pointer
+ *
+ * Map the seq64 memory to the given VM.
+ *
+ * Returns:
+ * 0 on success or a negative error code on failure
+ */
+int amdgpu_seq64_map(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+		     struct amdgpu_bo_va **bo_va)
+{
+	struct amdgpu_bo *bo;
+	struct drm_exec exec;
+	u64 seq64_addr;
+	int r;
+
+	bo = adev->seq64.sbo;
+	if (!bo)
+		return -EINVAL;
+
+	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT, 0);
+	drm_exec_until_all_locked(&exec) {
+		r = amdgpu_vm_lock_pd(vm, &exec, 0);
+		if (likely(!r))
+			r = drm_exec_lock_obj(&exec, &bo->tbo.base);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(r))
+			goto error;
+	}
+
+	*bo_va = amdgpu_vm_bo_add(adev, vm, bo);
+	if (!*bo_va) {
+		r = -ENOMEM;
+		goto error;
+	}
+
+	seq64_addr = amdgpu_seq64_get_va_base(adev) & AMDGPU_GMC_HOLE_MASK;
+
+	r = amdgpu_vm_bo_map(adev, *bo_va, seq64_addr, 0,
+			     AMDGPU_VA_RESERVED_SEQ64_SIZE,
+			     AMDGPU_VM_PAGE_READABLE | AMDGPU_VM_MTYPE_UC);
+	if (r) {
+		DRM_ERROR("failed to do bo_map on userq sem, err=%d\n", r);
+		amdgpu_vm_bo_del(adev, *bo_va);
+		goto error;
+	}
+
+	r = amdgpu_vm_bo_update(adev, *bo_va, false);
+	if (r) {
+		DRM_ERROR("failed to do vm_bo_update on userq sem\n");
+		amdgpu_vm_bo_del(adev, *bo_va);
+		goto error;
+	}
+
+error:
+	drm_exec_fini(&exec);
+	return r;
+}
+
+/**
+ * amdgpu_seq64_unmap - Unmap the seq64 memory
+ *
+ * @adev: amdgpu_device pointer
+ * @fpriv: DRM file private
+ *
+ * Unmap the seq64 memory from the given VM.
+ */
+void amdgpu_seq64_unmap(struct amdgpu_device *adev, struct amdgpu_fpriv *fpriv)
+{
+	struct amdgpu_vm *vm;
+	struct amdgpu_bo *bo;
+	struct drm_exec exec;
+	int r;
+
+	if (!fpriv->seq64_va)
+		return;
+
+	bo = adev->seq64.sbo;
+	if (!bo)
+		return;
+
+	vm = &fpriv->vm;
+
+	drm_exec_init(&exec, 0, 0);
+	drm_exec_until_all_locked(&exec) {
+		r = amdgpu_vm_lock_pd(vm, &exec, 0);
+		if (likely(!r))
+			r = drm_exec_lock_obj(&exec, &bo->tbo.base);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(r))
+			goto error;
+	}
+
+	amdgpu_vm_bo_del(adev, fpriv->seq64_va);
+
+	fpriv->seq64_va = NULL;
+
+error:
+	drm_exec_fini(&exec);
+}
+
+/**
+ * amdgpu_seq64_alloc - Allocate a 64 bit memory
+ *
+ * @adev: amdgpu_device pointer
+ * @va: VA to access the seq in process address space
+ * @gpu_addr: GPU address to access the seq
+ * @cpu_addr: CPU address to access the seq
+ *
+ * Alloc a 64 bit memory from seq64 pool.
+ *
+ * Returns:
+ * 0 on success or a negative error code on failure
+ */
+int amdgpu_seq64_alloc(struct amdgpu_device *adev, u64 *va,
+		       u64 *gpu_addr, u64 **cpu_addr)
+{
+	unsigned long bit_pos;
+
+	bit_pos = find_first_zero_bit(adev->seq64.used, adev->seq64.num_sem);
+	if (bit_pos >= adev->seq64.num_sem)
+		return -ENOSPC;
+
+	__set_bit(bit_pos, adev->seq64.used);
+
+	*va = bit_pos * sizeof(u64) + amdgpu_seq64_get_va_base(adev);
+
+	if (gpu_addr)
+		*gpu_addr = bit_pos * sizeof(u64) + adev->seq64.gpu_addr;
+
+	*cpu_addr = bit_pos + adev->seq64.cpu_base_addr;
+
+	return 0;
+}
+
+/**
+ * amdgpu_seq64_free - Free the given 64 bit memory
+ *
+ * @adev: amdgpu_device pointer
+ * @va: gpu start address to be freed
+ *
+ * Free the given 64 bit memory from seq64 pool.
+ */
+void amdgpu_seq64_free(struct amdgpu_device *adev, u64 va)
+{
+	unsigned long bit_pos;
+
+	bit_pos = (va - amdgpu_seq64_get_va_base(adev)) / sizeof(u64);
+	if (bit_pos < adev->seq64.num_sem)
+		__clear_bit(bit_pos, adev->seq64.used);
+}
+
+/**
+ * amdgpu_seq64_fini - Cleanup seq64 driver
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Free the memory space allocated for seq64.
+ *
+ */
+void amdgpu_seq64_fini(struct amdgpu_device *adev)
+{
+	amdgpu_bo_free_kernel(&adev->seq64.sbo,
+			      NULL,
+			      (void **)&adev->seq64.cpu_base_addr);
+}
+
+/**
+ * amdgpu_seq64_init - Initialize seq64 driver
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Allocate the required memory space for seq64.
+ *
+ * Returns:
+ * 0 on success or a negative error code on failure
+ */
+int amdgpu_seq64_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	if (adev->seq64.sbo)
+		return 0;
+
+	/*
+	 * AMDGPU_MAX_SEQ64_SLOTS * sizeof(u64) * 8 = AMDGPU_MAX_SEQ64_SLOTS
+	 * 64bit slots
+	 */
+	r = amdgpu_bo_create_kernel(adev, AMDGPU_VA_RESERVED_SEQ64_SIZE,
+				    PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT,
+				    &adev->seq64.sbo, &adev->seq64.gpu_addr,
+				    (void **)&adev->seq64.cpu_base_addr);
+	if (r) {
+		dev_warn(adev->dev, "(%d) create seq64 failed\n", r);
+		return r;
+	}
+
+	memset(adev->seq64.cpu_base_addr, 0, AMDGPU_VA_RESERVED_SEQ64_SIZE);
+
+	adev->seq64.num_sem = AMDGPU_MAX_SEQ64_SLOTS;
+	memset(&adev->seq64.used, 0, sizeof(adev->seq64.used));
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.h
new file mode 100644
index 000000000000..26a249aaaee1
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.h
@@ -0,0 +1,49 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __AMDGPU_SEQ64_H__
+#define __AMDGPU_SEQ64_H__
+
+#include "amdgpu_vm.h"
+
+#define AMDGPU_MAX_SEQ64_SLOTS         (AMDGPU_VA_RESERVED_SEQ64_SIZE / sizeof(u64))
+
+struct amdgpu_seq64 {
+	struct amdgpu_bo *sbo;
+	u32 num_sem;
+	u64 gpu_addr;
+	u64 *cpu_base_addr;
+	DECLARE_BITMAP(used, AMDGPU_MAX_SEQ64_SLOTS);
+};
+
+void amdgpu_seq64_fini(struct amdgpu_device *adev);
+int amdgpu_seq64_init(struct amdgpu_device *adev);
+int amdgpu_seq64_alloc(struct amdgpu_device *adev, u64 *va, u64 *gpu_addr, u64 **cpu_addr);
+void amdgpu_seq64_free(struct amdgpu_device *adev, u64 gpu_addr);
+int amdgpu_seq64_map(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+		     struct amdgpu_bo_va **bo_va);
+void amdgpu_seq64_unmap(struct amdgpu_device *adev, struct amdgpu_fpriv *fpriv);
+
+#endif
+
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h
index c7a823f3f2c5..ec9d12f85f39 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h
@@ -23,6 +23,18 @@
 #ifndef __AMDGPU_SMUIO_H__
 #define __AMDGPU_SMUIO_H__
 
+enum amdgpu_pkg_type {
+	AMDGPU_PKG_TYPE_APU = 2,
+	AMDGPU_PKG_TYPE_CEM = 3,
+	AMDGPU_PKG_TYPE_OAM = 4,
+	AMDGPU_PKG_TYPE_UNKNOWN,
+};
+
+struct amdgpu_smuio_mcm_config_info {
+	int socket_id;
+	int die_id;
+};
+
 struct amdgpu_smuio_funcs {
 	u32 (*get_rom_index_offset)(struct amdgpu_device *adev);
 	u32 (*get_rom_data_offset)(struct amdgpu_device *adev);
@@ -30,7 +42,9 @@ struct amdgpu_smuio_funcs {
 	void (*get_clock_gating_state)(struct amdgpu_device *adev, u64 *flags);
 	u32 (*get_die_id)(struct amdgpu_device *adev);
 	u32 (*get_socket_id)(struct amdgpu_device *adev);
+	enum amdgpu_pkg_type (*get_pkg_type)(struct amdgpu_device *adev);
 	bool (*is_host_gpu_xgmi_supported)(struct amdgpu_device *adev);
+	u64 (*get_gpu_clock_counter)(struct amdgpu_device *adev);
 };
 
 struct amdgpu_smuio {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
index dcd8c066bc1f..d6ae9974c952 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
@@ -135,11 +135,16 @@ static bool amdgpu_sync_add_later(struct amdgpu_sync *sync, struct dma_fence *f)
 	struct amdgpu_sync_entry *e;
 
 	hash_for_each_possible(sync->fences, e, node, f->context) {
-		if (unlikely(e->fence->context != f->context))
-			continue;
+		if (dma_fence_is_signaled(e->fence)) {
+			dma_fence_put(e->fence);
+			e->fence = dma_fence_get(f);
+			return true;
+		}
 
-		amdgpu_sync_keep_later(&e->fence, f);
-		return true;
+		if (likely(e->fence->context == f->context)) {
+			amdgpu_sync_keep_later(&e->fence, f);
+			return true;
+		}
 	}
 	return false;
 }
@@ -149,10 +154,12 @@ static bool amdgpu_sync_add_later(struct amdgpu_sync *sync, struct dma_fence *f)
  *
  * @sync: sync object to add fence to
  * @f: fence to sync to
+ * @flags: memory allocation flags to use when allocating sync entry
  *
  * Add the fence to the sync object.
  */
-int amdgpu_sync_fence(struct amdgpu_sync *sync, struct dma_fence *f)
+int amdgpu_sync_fence(struct amdgpu_sync *sync, struct dma_fence *f,
+		      gfp_t flags)
 {
 	struct amdgpu_sync_entry *e;
 
@@ -162,7 +169,7 @@ int amdgpu_sync_fence(struct amdgpu_sync *sync, struct dma_fence *f)
 	if (amdgpu_sync_add_later(sync, f))
 		return 0;
 
-	e = kmem_cache_alloc(amdgpu_sync_slab, GFP_KERNEL);
+	e = kmem_cache_alloc(amdgpu_sync_slab, flags);
 	if (!e)
 		return -ENOMEM;
 
@@ -191,7 +198,8 @@ static bool amdgpu_sync_test_fence(struct amdgpu_device *adev,
 
 	/* Never sync to VM updates either. */
 	if (fence_owner == AMDGPU_FENCE_OWNER_VM &&
-	    owner != AMDGPU_FENCE_OWNER_UNDEFINED)
+	    owner != AMDGPU_FENCE_OWNER_UNDEFINED &&
+	    owner != AMDGPU_FENCE_OWNER_KFD)
 		return false;
 
 	/* Ignore fences depending on the sync mode */
@@ -241,14 +249,13 @@ int amdgpu_sync_resv(struct amdgpu_device *adev, struct amdgpu_sync *sync,
 
 	if (resv == NULL)
 		return -EINVAL;
-
-	/* TODO: Use DMA_RESV_USAGE_READ here */
-	dma_resv_for_each_fence(&cursor, resv, DMA_RESV_USAGE_BOOKKEEP, f) {
+	/* Implicitly sync only to KERNEL, WRITE and READ */
+	dma_resv_for_each_fence(&cursor, resv, DMA_RESV_USAGE_READ, f) {
 		dma_fence_chain_for_each(f, f) {
 			struct dma_fence *tmp = dma_fence_chain_contained(f);
 
 			if (amdgpu_sync_test_fence(adev, mode, owner, tmp)) {
-				r = amdgpu_sync_fence(sync, f);
+				r = amdgpu_sync_fence(sync, f, GFP_KERNEL);
 				dma_fence_put(f);
 				if (r)
 					return r;
@@ -259,6 +266,36 @@ int amdgpu_sync_resv(struct amdgpu_device *adev, struct amdgpu_sync *sync,
 	return 0;
 }
 
+/**
+ * amdgpu_sync_kfd - sync to KFD fences
+ *
+ * @sync: sync object to add KFD fences to
+ * @resv: reservation object with KFD fences
+ *
+ * Extract all KFD fences and add them to the sync object.
+ */
+int amdgpu_sync_kfd(struct amdgpu_sync *sync, struct dma_resv *resv)
+{
+	struct dma_resv_iter cursor;
+	struct dma_fence *f;
+	int r = 0;
+
+	dma_resv_iter_begin(&cursor, resv, DMA_RESV_USAGE_BOOKKEEP);
+	dma_resv_for_each_fence_unlocked(&cursor, f) {
+		void *fence_owner = amdgpu_sync_get_owner(f);
+
+		if (fence_owner != AMDGPU_FENCE_OWNER_KFD)
+			continue;
+
+		r = amdgpu_sync_fence(sync, f, GFP_KERNEL);
+		if (r)
+			break;
+	}
+	dma_resv_iter_end(&cursor);
+
+	return r;
+}
+
 /* Free the entry back to the slab */
 static void amdgpu_sync_entry_free(struct amdgpu_sync_entry *e)
 {
@@ -357,7 +394,7 @@ int amdgpu_sync_clone(struct amdgpu_sync *source, struct amdgpu_sync *clone)
 	hash_for_each_safe(source->fences, i, tmp, e, node) {
 		f = e->fence;
 		if (!dma_fence_is_signaled(f)) {
-			r = amdgpu_sync_fence(clone, f);
+			r = amdgpu_sync_fence(clone, f, GFP_KERNEL);
 			if (r)
 				return r;
 		} else {
@@ -369,6 +406,25 @@ int amdgpu_sync_clone(struct amdgpu_sync *source, struct amdgpu_sync *clone)
 }
 
 /**
+ * amdgpu_sync_move - move all fences from src to dst
+ *
+ * @src: source of the fences, empty after function
+ * @dst: destination for the fences
+ *
+ * Moves all fences from source to destination. All fences in destination are
+ * freed and source is empty after the function call.
+ */
+void amdgpu_sync_move(struct amdgpu_sync *src, struct amdgpu_sync *dst)
+{
+	unsigned int i;
+
+	amdgpu_sync_free(dst);
+
+	for (i = 0; i < HASH_SIZE(src->fences); ++i)
+		hlist_move_list(&src->fences[i], &dst->fences[i]);
+}
+
+/**
  * amdgpu_sync_push_to_job - push fences into job
  * @sync: sync object to get the fences from
  * @job: job to push the fences into
@@ -440,9 +496,7 @@ void amdgpu_sync_free(struct amdgpu_sync *sync)
  */
 int amdgpu_sync_init(void)
 {
-	amdgpu_sync_slab = kmem_cache_create(
-		"amdgpu_sync", sizeof(struct amdgpu_sync_entry), 0,
-		SLAB_HWCACHE_ALIGN, NULL);
+	amdgpu_sync_slab = KMEM_CACHE(amdgpu_sync_entry, SLAB_HWCACHE_ALIGN);
 	if (!amdgpu_sync_slab)
 		return -ENOMEM;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h
index cf1e9e858efd..51eb4382c91e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h
@@ -47,14 +47,17 @@ struct amdgpu_sync {
 };
 
 void amdgpu_sync_create(struct amdgpu_sync *sync);
-int amdgpu_sync_fence(struct amdgpu_sync *sync, struct dma_fence *f);
+int amdgpu_sync_fence(struct amdgpu_sync *sync, struct dma_fence *f,
+		      gfp_t flags);
 int amdgpu_sync_resv(struct amdgpu_device *adev, struct amdgpu_sync *sync,
 		     struct dma_resv *resv, enum amdgpu_sync_mode mode,
 		     void *owner);
+int amdgpu_sync_kfd(struct amdgpu_sync *sync, struct dma_resv *resv);
 struct dma_fence *amdgpu_sync_peek_fence(struct amdgpu_sync *sync,
 				     struct amdgpu_ring *ring);
 struct dma_fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync);
 int amdgpu_sync_clone(struct amdgpu_sync *source, struct amdgpu_sync *clone);
+void amdgpu_sync_move(struct amdgpu_sync *src, struct amdgpu_sync *dst);
 int amdgpu_sync_push_to_job(struct amdgpu_sync *sync, struct amdgpu_job *job);
 int amdgpu_sync_wait(struct amdgpu_sync *sync, bool intr);
 void amdgpu_sync_free(struct amdgpu_sync *sync);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h
index 525dffbe046a..d13e64a69e25 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h
@@ -167,25 +167,23 @@ TRACE_EVENT(amdgpu_cs_ioctl,
 	    TP_PROTO(struct amdgpu_job *job),
 	    TP_ARGS(job),
 	    TP_STRUCT__entry(
-			     __field(uint64_t, sched_job_id)
 			     __string(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job))
-			     __field(unsigned int, context)
-			     __field(unsigned int, seqno)
+			     __field(u64, context)
+			     __field(u64, seqno)
 			     __field(struct dma_fence *, fence)
 			     __string(ring, to_amdgpu_ring(job->base.sched)->name)
 			     __field(u32, num_ibs)
 			     ),
 
 	    TP_fast_assign(
-			   __entry->sched_job_id = job->base.id;
-			   __assign_str(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job));
+			   __assign_str(timeline);
 			   __entry->context = job->base.s_fence->finished.context;
 			   __entry->seqno = job->base.s_fence->finished.seqno;
-			   __assign_str(ring, to_amdgpu_ring(job->base.sched)->name);
+			   __assign_str(ring);
 			   __entry->num_ibs = job->num_ibs;
 			   ),
-	    TP_printk("sched_job=%llu, timeline=%s, context=%u, seqno=%u, ring_name=%s, num_ibs=%u",
-		      __entry->sched_job_id, __get_str(timeline), __entry->context,
+	    TP_printk("timeline=%s, fence=%llu:%llu, ring_name=%s, num_ibs=%u",
+		      __get_str(timeline), __entry->context,
 		      __entry->seqno, __get_str(ring), __entry->num_ibs)
 );
 
@@ -193,24 +191,22 @@ TRACE_EVENT(amdgpu_sched_run_job,
 	    TP_PROTO(struct amdgpu_job *job),
 	    TP_ARGS(job),
 	    TP_STRUCT__entry(
-			     __field(uint64_t, sched_job_id)
 			     __string(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job))
-			     __field(unsigned int, context)
-			     __field(unsigned int, seqno)
+			     __field(u64, context)
+			     __field(u64, seqno)
 			     __string(ring, to_amdgpu_ring(job->base.sched)->name)
 			     __field(u32, num_ibs)
 			     ),
 
 	    TP_fast_assign(
-			   __entry->sched_job_id = job->base.id;
-			   __assign_str(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job));
+			   __assign_str(timeline);
 			   __entry->context = job->base.s_fence->finished.context;
 			   __entry->seqno = job->base.s_fence->finished.seqno;
-			   __assign_str(ring, to_amdgpu_ring(job->base.sched)->name);
+			   __assign_str(ring);
 			   __entry->num_ibs = job->num_ibs;
 			   ),
-	    TP_printk("sched_job=%llu, timeline=%s, context=%u, seqno=%u, ring_name=%s, num_ibs=%u",
-		      __entry->sched_job_id, __get_str(timeline), __entry->context,
+	    TP_printk("timeline=%s, fence=%llu:%llu, ring_name=%s, num_ibs=%u",
+		      __get_str(timeline), __entry->context,
 		      __entry->seqno, __get_str(ring), __entry->num_ibs)
 );
 
@@ -231,7 +227,7 @@ TRACE_EVENT(amdgpu_vm_grab_id,
 
 	    TP_fast_assign(
 			   __entry->pasid = vm->pasid;
-			   __assign_str(ring, ring->name);
+			   __assign_str(ring);
 			   __entry->vmid = job->vmid;
 			   __entry->vm_hub = ring->vm_hub,
 			   __entry->pd_addr = job->vm_pd_addr;
@@ -425,14 +421,14 @@ TRACE_EVENT(amdgpu_vm_flush,
 			     ),
 
 	    TP_fast_assign(
-			   __assign_str(ring, ring->name);
+			   __assign_str(ring);
 			   __entry->vmid = vmid;
 			   __entry->vm_hub = ring->vm_hub;
 			   __entry->pd_addr = pd_addr;
 			   ),
 	    TP_printk("ring=%s, id=%u, hub=%u, pd_addr=%010Lx",
 		      __get_str(ring), __entry->vmid,
-		      __entry->vm_hub,__entry->pd_addr)
+		      __entry->vm_hub, __entry->pd_addr)
 );
 
 DECLARE_EVENT_CLASS(amdgpu_pasid,
@@ -457,6 +453,38 @@ DEFINE_EVENT(amdgpu_pasid, amdgpu_pasid_freed,
 	    TP_ARGS(pasid)
 );
 
+TRACE_EVENT(amdgpu_isolation,
+	    TP_PROTO(void *prev, void *next),
+	    TP_ARGS(prev, next),
+	    TP_STRUCT__entry(
+			     __field(void *, prev)
+			     __field(void *, next)
+			     ),
+
+	    TP_fast_assign(
+			   __entry->prev = prev;
+			   __entry->next = next;
+			   ),
+	    TP_printk("prev=%p, next=%p",
+		      __entry->prev,
+		      __entry->next)
+);
+
+TRACE_EVENT(amdgpu_cleaner_shader,
+	    TP_PROTO(struct amdgpu_ring *ring, struct dma_fence *fence),
+	    TP_ARGS(ring, fence),
+	    TP_STRUCT__entry(
+			     __string(ring, ring->name)
+			     __field(u64, seqno)
+			     ),
+
+	    TP_fast_assign(
+			   __assign_str(ring);
+			   __entry->seqno = fence->seqno;
+			   ),
+	    TP_printk("ring=%s, seqno=%Lu", __get_str(ring), __entry->seqno)
+);
+
 TRACE_EVENT(amdgpu_bo_list_set,
 	    TP_PROTO(struct amdgpu_bo_list *list, struct amdgpu_bo *bo),
 	    TP_ARGS(list, bo),
@@ -494,7 +522,7 @@ TRACE_EVENT(amdgpu_cs_bo_status,
 );
 
 TRACE_EVENT(amdgpu_bo_move,
-	    TP_PROTO(struct amdgpu_bo* bo, uint32_t new_placement, uint32_t old_placement),
+	    TP_PROTO(struct amdgpu_bo *bo, uint32_t new_placement, uint32_t old_placement),
 	    TP_ARGS(bo, new_placement, old_placement),
 	    TP_STRUCT__entry(
 			__field(struct amdgpu_bo *, bo)
@@ -519,23 +547,19 @@ TRACE_EVENT(amdgpu_ib_pipe_sync,
 	    TP_ARGS(sched_job, fence),
 	    TP_STRUCT__entry(
 			     __string(ring, sched_job->base.sched->name)
-			     __field(uint64_t, id)
 			     __field(struct dma_fence *, fence)
-			     __field(uint64_t, ctx)
-			     __field(unsigned, seqno)
+			     __field(u64, ctx)
+			     __field(u64, seqno)
 			     ),
 
 	    TP_fast_assign(
-			   __assign_str(ring, sched_job->base.sched->name);
-			   __entry->id = sched_job->base.id;
+			   __assign_str(ring);
 			   __entry->fence = fence;
 			   __entry->ctx = fence->context;
 			   __entry->seqno = fence->seqno;
 			   ),
-	    TP_printk("job ring=%s, id=%llu, need pipe sync to fence=%p, context=%llu, seq=%u",
-		      __get_str(ring), __entry->id,
-		      __entry->fence, __entry->ctx,
-		      __entry->seqno)
+	    TP_printk("job ring=%s need pipe sync to fence=%llu:%llu",
+		      __get_str(ring), __entry->ctx, __entry->seqno)
 );
 
 TRACE_EVENT(amdgpu_reset_reg_dumps,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 2cd081cbf706..2b931e855abd 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -38,7 +38,6 @@
 #include <linux/seq_file.h>
 #include <linux/slab.h>
 #include <linux/swap.h>
-#include <linux/swiotlb.h>
 #include <linux/dma-buf.h>
 #include <linux/sizes.h>
 #include <linux/module.h>
@@ -50,7 +49,6 @@
 #include <drm/ttm/ttm_tt.h>
 
 #include <drm/amdgpu_drm.h>
-#include <drm/drm_drv.h>
 
 #include "amdgpu.h"
 #include "amdgpu_object.h"
@@ -63,9 +61,9 @@
 #include "amdgpu_res_cursor.h"
 #include "bif/bif_4_1_d.h"
 
-MODULE_IMPORT_NS(DMA_BUF);
+MODULE_IMPORT_NS("DMA_BUF");
 
-#define AMDGPU_TTM_VRAM_MAX_DW_READ	(size_t)128
+#define AMDGPU_TTM_VRAM_MAX_DW_READ	((size_t)128)
 
 static int amdgpu_ttm_backend_bind(struct ttm_device *bdev,
 				   struct ttm_tt *ttm,
@@ -104,23 +102,19 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
 	/* Don't handle scatter gather BOs */
 	if (bo->type == ttm_bo_type_sg) {
 		placement->num_placement = 0;
-		placement->num_busy_placement = 0;
 		return;
 	}
 
 	/* Object isn't an AMDGPU object so ignore */
 	if (!amdgpu_bo_is_amdgpu_bo(bo)) {
 		placement->placement = &placements;
-		placement->busy_placement = &placements;
 		placement->num_placement = 1;
-		placement->num_busy_placement = 1;
 		return;
 	}
 
 	abo = ttm_to_amdgpu_bo(bo);
 	if (abo->flags & AMDGPU_GEM_CREATE_DISCARDABLE) {
 		placement->num_placement = 0;
-		placement->num_busy_placement = 0;
 		return;
 	}
 
@@ -128,17 +122,19 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
 	case AMDGPU_PL_GDS:
 	case AMDGPU_PL_GWS:
 	case AMDGPU_PL_OA:
+	case AMDGPU_PL_DOORBELL:
+	case AMDGPU_PL_MMIO_REMAP:
 		placement->num_placement = 0;
-		placement->num_busy_placement = 0;
 		return;
 
 	case TTM_PL_VRAM:
 		if (!adev->mman.buffer_funcs_enabled) {
 			/* Move to system memory */
 			amdgpu_bo_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_CPU);
+
 		} else if (!amdgpu_gmc_vram_full_visible(&adev->gmc) &&
 			   !(abo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) &&
-			   amdgpu_bo_in_cpu_visible_vram(abo)) {
+			   amdgpu_res_cpu_visible(adev, bo->resource)) {
 
 			/* Try evicting to the CPU inaccessible part of VRAM
 			 * first, but only set GTT as busy placement, so this
@@ -150,8 +146,7 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
 							AMDGPU_GEM_DOMAIN_CPU);
 			abo->placements[0].fpfn = adev->gmc.visible_vram_size >> PAGE_SHIFT;
 			abo->placements[0].lpfn = 0;
-			abo->placement.busy_placement = &abo->placements[1];
-			abo->placement.num_busy_placement = 1;
+			abo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
 		} else {
 			/* Move to GTT memory */
 			amdgpu_bo_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_GTT |
@@ -184,16 +179,15 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
 static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo,
 				 struct ttm_resource *mem,
 				 struct amdgpu_res_cursor *mm_cur,
-				 unsigned window, struct amdgpu_ring *ring,
+				 unsigned int window, struct amdgpu_ring *ring,
 				 bool tmz, uint64_t *size, uint64_t *addr)
 {
 	struct amdgpu_device *adev = ring->adev;
-	unsigned offset, num_pages, num_dw, num_bytes;
+	unsigned int offset, num_pages, num_dw, num_bytes;
 	uint64_t src_addr, dst_addr;
 	struct amdgpu_job *job;
 	void *cpu_addr;
 	uint64_t flags;
-	unsigned int i;
 	int r;
 
 	BUG_ON(adev->mman.buffer_funcs->copy_max_bytes <
@@ -229,10 +223,11 @@ static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo,
 	num_dw = ALIGN(adev->mman.buffer_funcs->copy_num_dw, 8);
 	num_bytes = num_pages * 8 * AMDGPU_GPU_PAGES_IN_CPU_PAGE;
 
-	r = amdgpu_job_alloc_with_ib(adev, &adev->mman.entity,
+	r = amdgpu_job_alloc_with_ib(adev, &adev->mman.high_pr,
 				     AMDGPU_FENCE_OWNER_UNDEFINED,
 				     num_dw * 4 + num_bytes,
-				     AMDGPU_IB_POOL_DELAYED, &job);
+				     AMDGPU_IB_POOL_DELAYED, &job,
+				     AMDGPU_KERNEL_JOB_ID_TTM_MAP_BUFFER);
 	if (r)
 		return r;
 
@@ -242,7 +237,7 @@ static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo,
 	dst_addr = amdgpu_bo_gpu_offset(adev->gart.bo);
 	dst_addr += window * AMDGPU_GTT_MAX_TRANSFER_SIZE * 8;
 	amdgpu_emit_copy_buffer(adev, &job->ibs[0], src_addr,
-				dst_addr, num_bytes, false);
+				dst_addr, num_bytes, 0);
 
 	amdgpu_ring_pad_ib(ring, &job->ibs[0]);
 	WARN_ON(job->ibs[0].length_dw > num_dw);
@@ -259,16 +254,9 @@ static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo,
 		dma_addr = &bo->ttm->dma_address[mm_cur->start >> PAGE_SHIFT];
 		amdgpu_gart_map(adev, 0, num_pages, dma_addr, flags, cpu_addr);
 	} else {
-		dma_addr_t dma_address;
-
-		dma_address = mm_cur->start;
-		dma_address += adev->vm_manager.vram_base_offset;
+		u64 pa = mm_cur->start + adev->vm_manager.vram_base_offset;
 
-		for (i = 0; i < num_pages; ++i) {
-			amdgpu_gart_map(adev, i << PAGE_SHIFT, 1, &dma_address,
-					flags, cpu_addr);
-			dma_address += PAGE_SIZE;
-		}
+		amdgpu_gart_map_vram_range(adev, pa, 0, num_pages, flags, cpu_addr);
 	}
 
 	dma_fence_put(amdgpu_job_submit(job));
@@ -290,20 +278,24 @@ static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo,
  * move and different for a BO to BO copy.
  *
  */
-int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev,
-			       const struct amdgpu_copy_mem *src,
-			       const struct amdgpu_copy_mem *dst,
-			       uint64_t size, bool tmz,
-			       struct dma_resv *resv,
-			       struct dma_fence **f)
+__attribute__((nonnull))
+static int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev,
+				      const struct amdgpu_copy_mem *src,
+				      const struct amdgpu_copy_mem *dst,
+				      uint64_t size, bool tmz,
+				      struct dma_resv *resv,
+				      struct dma_fence **f)
 {
 	struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
 	struct amdgpu_res_cursor src_mm, dst_mm;
 	struct dma_fence *fence = NULL;
 	int r = 0;
+	uint32_t copy_flags = 0;
+	struct amdgpu_bo *abo_src, *abo_dst;
 
 	if (!adev->mman.buffer_funcs_enabled) {
-		DRM_ERROR("Trying to move memory with ring turned off.\n");
+		dev_err(adev->dev,
+			"Trying to move memory with ring turned off.\n");
 		return -EINVAL;
 	}
 
@@ -312,7 +304,8 @@ int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev,
 
 	mutex_lock(&adev->mman.gtt_window_lock);
 	while (src_mm.remaining) {
-		uint64_t from, to, cur_size;
+		uint64_t from, to, cur_size, tiling_flags;
+		uint32_t num_type, data_format, max_com, write_compress_disable;
 		struct dma_fence *next;
 
 		/* Never copy more than 256MiB at once to avoid a timeout */
@@ -329,8 +322,31 @@ int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev,
 		if (r)
 			goto error;
 
-		r = amdgpu_copy_buffer(ring, from, to, cur_size,
-				       resv, &next, false, true, tmz);
+		abo_src = ttm_to_amdgpu_bo(src->bo);
+		abo_dst = ttm_to_amdgpu_bo(dst->bo);
+		if (tmz)
+			copy_flags |= AMDGPU_COPY_FLAGS_TMZ;
+		if ((abo_src->flags & AMDGPU_GEM_CREATE_GFX12_DCC) &&
+		    (abo_src->tbo.resource->mem_type == TTM_PL_VRAM))
+			copy_flags |= AMDGPU_COPY_FLAGS_READ_DECOMPRESSED;
+		if ((abo_dst->flags & AMDGPU_GEM_CREATE_GFX12_DCC) &&
+		    (dst->mem->mem_type == TTM_PL_VRAM)) {
+			copy_flags |= AMDGPU_COPY_FLAGS_WRITE_COMPRESSED;
+			amdgpu_bo_get_tiling_flags(abo_dst, &tiling_flags);
+			max_com = AMDGPU_TILING_GET(tiling_flags, GFX12_DCC_MAX_COMPRESSED_BLOCK);
+			num_type = AMDGPU_TILING_GET(tiling_flags, GFX12_DCC_NUMBER_TYPE);
+			data_format = AMDGPU_TILING_GET(tiling_flags, GFX12_DCC_DATA_FORMAT);
+			write_compress_disable =
+				AMDGPU_TILING_GET(tiling_flags, GFX12_DCC_WRITE_COMPRESS_DISABLE);
+			copy_flags |= (AMDGPU_COPY_FLAGS_SET(MAX_COMPRESSED, max_com) |
+				       AMDGPU_COPY_FLAGS_SET(NUMBER_TYPE, num_type) |
+				       AMDGPU_COPY_FLAGS_SET(DATA_FORMAT, data_format) |
+				       AMDGPU_COPY_FLAGS_SET(WRITE_COMPRESS_DISABLE,
+							     write_compress_disable));
+		}
+
+		r = amdgpu_copy_buffer(ring, from, to, cur_size, resv,
+				       &next, false, true, copy_flags);
 		if (r)
 			goto error;
 
@@ -342,9 +358,7 @@ int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev,
 	}
 error:
 	mutex_unlock(&adev->mman.gtt_window_lock);
-	if (f)
-		*f = dma_fence_get(fence);
-	dma_fence_put(fence);
+	*f = fence;
 	return r;
 }
 
@@ -384,10 +398,12 @@ static int amdgpu_move_blit(struct ttm_buffer_object *bo,
 	    (abo->flags & AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE)) {
 		struct dma_fence *wipe_fence = NULL;
 
-		r = amdgpu_fill_buffer(abo, AMDGPU_POISON, NULL, &wipe_fence);
+		r = amdgpu_fill_buffer(abo, 0, NULL, &wipe_fence,
+				       false, AMDGPU_KERNEL_JOB_ID_MOVE_BLIT);
 		if (r) {
 			goto error;
 		} else if (wipe_fence) {
+			amdgpu_vram_mgr_set_cleared(bo->resource);
 			dma_fence_put(fence);
 			fence = wipe_fence;
 		}
@@ -408,40 +424,56 @@ error:
 	return r;
 }
 
-/*
- * amdgpu_mem_visible - Check that memory can be accessed by ttm_bo_move_memcpy
+/**
+ * amdgpu_res_cpu_visible - Check that resource can be accessed by CPU
+ * @adev: amdgpu device
+ * @res: the resource to check
  *
- * Called by amdgpu_bo_move()
+ * Returns: true if the full resource is CPU visible, false otherwise.
  */
-static bool amdgpu_mem_visible(struct amdgpu_device *adev,
-			       struct ttm_resource *mem)
+bool amdgpu_res_cpu_visible(struct amdgpu_device *adev,
+			    struct ttm_resource *res)
 {
-	u64 mem_size = (u64)mem->size;
 	struct amdgpu_res_cursor cursor;
-	u64 end;
 
-	if (mem->mem_type == TTM_PL_SYSTEM ||
-	    mem->mem_type == TTM_PL_TT)
+	if (!res)
+		return false;
+
+	if (res->mem_type == TTM_PL_SYSTEM || res->mem_type == TTM_PL_TT ||
+	    res->mem_type == AMDGPU_PL_PREEMPT || res->mem_type == AMDGPU_PL_DOORBELL ||
+	    res->mem_type == AMDGPU_PL_MMIO_REMAP)
 		return true;
-	if (mem->mem_type != TTM_PL_VRAM)
+
+	if (res->mem_type != TTM_PL_VRAM)
 		return false;
 
-	amdgpu_res_first(mem, 0, mem_size, &cursor);
-	end = cursor.start + cursor.size;
+	amdgpu_res_first(res, 0, res->size, &cursor);
 	while (cursor.remaining) {
+		if ((cursor.start + cursor.size) > adev->gmc.visible_vram_size)
+			return false;
 		amdgpu_res_next(&cursor, cursor.size);
+	}
 
-		if (!cursor.remaining)
-			break;
+	return true;
+}
 
-		/* ttm_resource_ioremap only supports contiguous memory */
-		if (end != cursor.start)
-			return false;
+/*
+ * amdgpu_res_copyable - Check that memory can be accessed by ttm_bo_move_memcpy
+ *
+ * Called by amdgpu_bo_move()
+ */
+static bool amdgpu_res_copyable(struct amdgpu_device *adev,
+				struct ttm_resource *mem)
+{
+	if (!amdgpu_res_cpu_visible(adev, mem))
+		return false;
 
-		end = cursor.start + cursor.size;
-	}
+	/* ttm_resource_ioremap only supports contiguous memory */
+	if (mem->mem_type == TTM_PL_VRAM &&
+	    !(mem->placement & TTM_PL_FLAG_CONTIGUOUS))
+		return false;
 
-	return end <= adev->gmc.visible_vram_size;
+	return true;
 }
 
 /*
@@ -471,14 +503,16 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict,
 
 	if (!old_mem || (old_mem->mem_type == TTM_PL_SYSTEM &&
 			 bo->ttm == NULL)) {
+		amdgpu_bo_move_notify(bo, evict, new_mem);
 		ttm_bo_move_null(bo, new_mem);
-		goto out;
+		return 0;
 	}
 	if (old_mem->mem_type == TTM_PL_SYSTEM &&
 	    (new_mem->mem_type == TTM_PL_TT ||
 	     new_mem->mem_type == AMDGPU_PL_PREEMPT)) {
+		amdgpu_bo_move_notify(bo, evict, new_mem);
 		ttm_bo_move_null(bo, new_mem);
-		goto out;
+		return 0;
 	}
 	if ((old_mem->mem_type == TTM_PL_TT ||
 	     old_mem->mem_type == AMDGPU_PL_PREEMPT) &&
@@ -488,20 +522,26 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict,
 			return r;
 
 		amdgpu_ttm_backend_unbind(bo->bdev, bo->ttm);
+		amdgpu_bo_move_notify(bo, evict, new_mem);
 		ttm_resource_free(bo, &bo->resource);
 		ttm_bo_assign_mem(bo, new_mem);
-		goto out;
+		return 0;
 	}
 
 	if (old_mem->mem_type == AMDGPU_PL_GDS ||
 	    old_mem->mem_type == AMDGPU_PL_GWS ||
 	    old_mem->mem_type == AMDGPU_PL_OA ||
+	    old_mem->mem_type == AMDGPU_PL_DOORBELL ||
+	    old_mem->mem_type == AMDGPU_PL_MMIO_REMAP ||
 	    new_mem->mem_type == AMDGPU_PL_GDS ||
 	    new_mem->mem_type == AMDGPU_PL_GWS ||
-	    new_mem->mem_type == AMDGPU_PL_OA) {
+	    new_mem->mem_type == AMDGPU_PL_OA ||
+	    new_mem->mem_type == AMDGPU_PL_DOORBELL ||
+	    new_mem->mem_type == AMDGPU_PL_MMIO_REMAP) {
 		/* Nothing to save here */
+		amdgpu_bo_move_notify(bo, evict, new_mem);
 		ttm_bo_move_null(bo, new_mem);
-		goto out;
+		return 0;
 	}
 
 	if (bo->type == ttm_bo_type_device &&
@@ -513,27 +553,28 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict,
 		abo->flags &= ~AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
 	}
 
-	if (adev->mman.buffer_funcs_enabled) {
-		if (((old_mem->mem_type == TTM_PL_SYSTEM &&
-		      new_mem->mem_type == TTM_PL_VRAM) ||
-		     (old_mem->mem_type == TTM_PL_VRAM &&
-		      new_mem->mem_type == TTM_PL_SYSTEM))) {
-			hop->fpfn = 0;
-			hop->lpfn = 0;
-			hop->mem_type = TTM_PL_TT;
-			hop->flags = TTM_PL_FLAG_TEMPORARY;
-			return -EMULTIHOP;
-		}
+	if (adev->mman.buffer_funcs_enabled &&
+	    ((old_mem->mem_type == TTM_PL_SYSTEM &&
+	      new_mem->mem_type == TTM_PL_VRAM) ||
+	     (old_mem->mem_type == TTM_PL_VRAM &&
+	      new_mem->mem_type == TTM_PL_SYSTEM))) {
+		hop->fpfn = 0;
+		hop->lpfn = 0;
+		hop->mem_type = TTM_PL_TT;
+		hop->flags = TTM_PL_FLAG_TEMPORARY;
+		return -EMULTIHOP;
+	}
 
+	amdgpu_bo_move_notify(bo, evict, new_mem);
+	if (adev->mman.buffer_funcs_enabled)
 		r = amdgpu_move_blit(bo, evict, new_mem, old_mem);
-	} else {
+	else
 		r = -ENODEV;
-	}
 
 	if (r) {
 		/* Check that all memory is CPU accessible */
-		if (!amdgpu_mem_visible(adev, old_mem) ||
-		    !amdgpu_mem_visible(adev, new_mem)) {
+		if (!amdgpu_res_copyable(adev, old_mem) ||
+		    !amdgpu_res_copyable(adev, new_mem)) {
 			pr_err("Move buffer fallback to memcpy unavailable\n");
 			return r;
 		}
@@ -543,10 +584,10 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict,
 			return r;
 	}
 
-out:
-	/* update statistics */
+	/* update statistics after the move */
+	if (evict)
+		atomic64_inc(&adev->num_evictions);
 	atomic64_add(bo->base.size, &adev->num_bytes_moved);
-	amdgpu_bo_move_notify(bo, evict, new_mem);
 	return 0;
 }
 
@@ -559,7 +600,6 @@ static int amdgpu_ttm_io_mem_reserve(struct ttm_device *bdev,
 				     struct ttm_resource *mem)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bdev);
-	size_t bus_size = (size_t)mem->size;
 
 	switch (mem->mem_type) {
 	case TTM_PL_SYSTEM:
@@ -570,9 +610,6 @@ static int amdgpu_ttm_io_mem_reserve(struct ttm_device *bdev,
 		break;
 	case TTM_PL_VRAM:
 		mem->bus.offset = mem->start << PAGE_SHIFT;
-		/* check if it's visible */
-		if ((mem->bus.offset + bus_size) > adev->gmc.visible_vram_size)
-			return -EINVAL;
 
 		if (adev->mman.aper_base_kaddr &&
 		    mem->placement & TTM_PL_FLAG_CONTIGUOUS)
@@ -582,6 +619,18 @@ static int amdgpu_ttm_io_mem_reserve(struct ttm_device *bdev,
 		mem->bus.offset += adev->gmc.aper_base;
 		mem->bus.is_iomem = true;
 		break;
+	case AMDGPU_PL_DOORBELL:
+		mem->bus.offset = mem->start << PAGE_SHIFT;
+		mem->bus.offset += adev->doorbell.base;
+		mem->bus.is_iomem = true;
+		mem->bus.caching = ttm_uncached;
+		break;
+	case AMDGPU_PL_MMIO_REMAP:
+		mem->bus.offset = mem->start << PAGE_SHIFT;
+		mem->bus.offset += adev->rmmio_remap.bus_addr;
+		mem->bus.is_iomem = true;
+		mem->bus.caching = ttm_uncached;
+		break;
 	default:
 		return -EINVAL;
 	}
@@ -596,6 +645,12 @@ static unsigned long amdgpu_ttm_io_mem_pfn(struct ttm_buffer_object *bo,
 
 	amdgpu_res_first(bo->resource, (u64)page_offset << PAGE_SHIFT, 0,
 			 &cursor);
+
+	if (bo->resource->mem_type == AMDGPU_PL_DOORBELL)
+		return ((uint64_t)(adev->doorbell.base + cursor.start)) >> PAGE_SHIFT;
+	else if (bo->resource->mem_type == AMDGPU_PL_MMIO_REMAP)
+		return ((uint64_t)(adev->rmmio_remap.bus_addr + cursor.start)) >> PAGE_SHIFT;
+
 	return (adev->gmc.aper_base + cursor.start) >> PAGE_SHIFT;
 }
 
@@ -631,6 +686,7 @@ struct amdgpu_ttm_tt {
 	struct task_struct	*usertask;
 	uint32_t		userflags;
 	bool			bound;
+	int32_t			pool_id;
 };
 
 #define ttm_to_amdgpu_ttm_tt(ptr)	container_of(ptr, struct amdgpu_ttm_tt, ttm)
@@ -641,10 +697,11 @@ struct amdgpu_ttm_tt {
  * memory and start HMM tracking CPU page table update
  *
  * Calling function must call amdgpu_ttm_tt_userptr_range_done() once and only
- * once afterwards to stop HMM tracking
+ * once afterwards to stop HMM tracking. Its the caller responsibility to ensure
+ * that range is a valid memory and it is freed too.
  */
-int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages,
-				 struct hmm_range **range)
+int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo,
+				 struct amdgpu_hmm_range *range)
 {
 	struct ttm_tt *ttm = bo->tbo.ttm;
 	struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm);
@@ -654,9 +711,6 @@ int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages,
 	bool readonly;
 	int r = 0;
 
-	/* Make sure get_user_pages_done() can cleanup gracefully */
-	*range = NULL;
-
 	mm = bo->notifier.mm;
 	if (unlikely(!mm)) {
 		DRM_DEBUG_DRIVER("BO is not registered?\n");
@@ -680,7 +734,7 @@ int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages,
 
 	readonly = amdgpu_ttm_tt_is_readonly(ttm);
 	r = amdgpu_hmm_range_get_pages(&bo->notifier, start, ttm->num_pages,
-				       readonly, NULL, pages, range);
+				       readonly, NULL, range);
 out_unlock:
 	mmap_read_unlock(mm);
 	if (r)
@@ -691,38 +745,6 @@ out_unlock:
 	return r;
 }
 
-/* amdgpu_ttm_tt_discard_user_pages - Discard range and pfn array allocations
- */
-void amdgpu_ttm_tt_discard_user_pages(struct ttm_tt *ttm,
-				      struct hmm_range *range)
-{
-	struct amdgpu_ttm_tt *gtt = (void *)ttm;
-
-	if (gtt && gtt->userptr && range)
-		amdgpu_hmm_range_get_pages_done(range);
-}
-
-/*
- * amdgpu_ttm_tt_get_user_pages_done - stop HMM track the CPU page table change
- * Check if the pages backing this ttm range have been invalidated
- *
- * Returns: true if pages are still valid
- */
-bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm,
-				       struct hmm_range *range)
-{
-	struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm);
-
-	if (!gtt || !gtt->userptr || !range)
-		return false;
-
-	DRM_DEBUG_DRIVER("user_pages_done 0x%llx pages 0x%x\n",
-		gtt->userptr, ttm->num_pages);
-
-	WARN_ONCE(!range->hmm_pfns, "No user pages to check\n");
-
-	return !amdgpu_hmm_range_get_pages_done(range);
-}
 #endif
 
 /*
@@ -732,12 +754,12 @@ bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm,
  * that backs user memory and will ultimately be mapped into the device
  * address space.
  */
-void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages)
+void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct amdgpu_hmm_range *range)
 {
 	unsigned long i;
 
 	for (i = 0; i < ttm->num_pages; ++i)
-		ttm->pages[i] = pages ? pages[i] : NULL;
+		ttm->pages[i] = range ? hmm_pfn_to_page(range->hmm_range.hmm_pfns[i]) : NULL;
 }
 
 /*
@@ -765,7 +787,7 @@ static int amdgpu_ttm_tt_pin_userptr(struct ttm_device *bdev,
 	/* Map SG to device */
 	r = dma_map_sgtable(adev->dev, ttm->sg, direction, 0);
 	if (r)
-		goto release_sg;
+		goto release_sg_table;
 
 	/* convert SG to linear array of pages and dma addresses */
 	drm_prime_sg_to_dma_addr_array(ttm->sg, gtt->ttm.dma_address,
@@ -773,6 +795,8 @@ static int amdgpu_ttm_tt_pin_userptr(struct ttm_device *bdev,
 
 	return 0;
 
+release_sg_table:
+	sg_free_table(ttm->sg);
 release_sg:
 	kfree(ttm->sg);
 	ttm->sg = NULL;
@@ -800,6 +824,43 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_device *bdev,
 	sg_free_table(ttm->sg);
 }
 
+/*
+ * total_pages is constructed as MQD0+CtrlStack0 + MQD1+CtrlStack1 + ...
+ * MQDn+CtrlStackn where n is the number of XCCs per partition.
+ * pages_per_xcc is the size of one MQD+CtrlStack. The first page is MQD
+ * and uses memory type default, UC. The rest of pages_per_xcc are
+ * Ctrl stack and modify their memory type to NC.
+ */
+static void amdgpu_ttm_gart_bind_gfx9_mqd(struct amdgpu_device *adev,
+				struct ttm_tt *ttm, uint64_t flags)
+{
+	struct amdgpu_ttm_tt *gtt = (void *)ttm;
+	uint64_t total_pages = ttm->num_pages;
+	int num_xcc = max(1U, adev->gfx.num_xcc_per_xcp);
+	uint64_t page_idx, pages_per_xcc;
+	int i;
+	uint64_t ctrl_flags = AMDGPU_PTE_MTYPE_VG10(flags, AMDGPU_MTYPE_NC);
+
+	pages_per_xcc = total_pages;
+	do_div(pages_per_xcc, num_xcc);
+
+	for (i = 0, page_idx = 0; i < num_xcc; i++, page_idx += pages_per_xcc) {
+		/* MQD page: use default flags */
+		amdgpu_gart_bind(adev,
+				gtt->offset + (page_idx << PAGE_SHIFT),
+				1, &gtt->ttm.dma_address[page_idx], flags);
+		/*
+		 * Ctrl pages - modify the memory type to NC (ctrl_flags) from
+		 * the second page of the BO onward.
+		 */
+		amdgpu_gart_bind(adev,
+				gtt->offset + ((page_idx + 1) << PAGE_SHIFT),
+				pages_per_xcc - 1,
+				&gtt->ttm.dma_address[page_idx + 1],
+				ctrl_flags);
+	}
+}
+
 static void amdgpu_ttm_gart_bind(struct amdgpu_device *adev,
 				 struct ttm_buffer_object *tbo,
 				 uint64_t flags)
@@ -812,25 +873,12 @@ static void amdgpu_ttm_gart_bind(struct amdgpu_device *adev,
 		flags |= AMDGPU_PTE_TMZ;
 
 	if (abo->flags & AMDGPU_GEM_CREATE_CP_MQD_GFX9) {
-		uint64_t page_idx = 1;
-
-		amdgpu_gart_bind(adev, gtt->offset, page_idx,
-				 gtt->ttm.dma_address, flags);
-
-		/* The memory type of the first page defaults to UC. Now
-		 * modify the memory type to NC from the second page of
-		 * the BO onward.
-		 */
-		flags &= ~AMDGPU_PTE_MTYPE_VG10_MASK;
-		flags |= AMDGPU_PTE_MTYPE_VG10(AMDGPU_MTYPE_NC);
-
-		amdgpu_gart_bind(adev, gtt->offset + (page_idx << PAGE_SHIFT),
-				 ttm->num_pages - page_idx,
-				 &(gtt->ttm.dma_address[page_idx]), flags);
+		amdgpu_ttm_gart_bind_gfx9_mqd(adev, ttm, flags);
 	} else {
 		amdgpu_gart_bind(adev, gtt->offset, ttm->num_pages,
 				 gtt->ttm.dma_address, flags);
 	}
+	gtt->bound = true;
 }
 
 /*
@@ -857,7 +905,7 @@ static int amdgpu_ttm_backend_bind(struct ttm_device *bdev,
 	if (gtt->userptr) {
 		r = amdgpu_ttm_tt_pin_userptr(bdev, ttm);
 		if (r) {
-			DRM_ERROR("failed to pin userptr\n");
+			dev_err(adev->dev, "failed to pin userptr\n");
 			return r;
 		}
 	} else if (ttm->page_flags & TTM_TT_FLAG_EXTERNAL) {
@@ -922,16 +970,12 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
 		return 0;
 
 	addr = amdgpu_gmc_agp_addr(bo);
-	if (addr != AMDGPU_BO_INVALID_OFFSET) {
-		bo->resource->start = addr >> PAGE_SHIFT;
+	if (addr != AMDGPU_BO_INVALID_OFFSET)
 		return 0;
-	}
 
 	/* allocate GART space */
 	placement.num_placement = 1;
 	placement.placement = &placements;
-	placement.num_busy_placement = 1;
-	placement.busy_placement = &placements;
 	placements.fpfn = 0;
 	placements.lpfn = adev->gmc.gart_size >> PAGE_SHIFT;
 	placements.mem_type = TTM_PL_TT;
@@ -987,7 +1031,7 @@ static void amdgpu_ttm_backend_unbind(struct ttm_device *bdev,
 	/* if the pages have userptr pinning then clear that first */
 	if (gtt->userptr) {
 		amdgpu_ttm_tt_unpin_userptr(bdev, ttm);
-	} else if (ttm->sg && gtt->gobj->import_attach) {
+	} else if (ttm->sg && drm_gem_is_imported(gtt->gobj)) {
 		struct dma_buf_attachment *attach;
 
 		attach = gtt->gobj->import_attach;
@@ -1029,15 +1073,20 @@ static void amdgpu_ttm_backend_destroy(struct ttm_device *bdev,
 static struct ttm_tt *amdgpu_ttm_tt_create(struct ttm_buffer_object *bo,
 					   uint32_t page_flags)
 {
+	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev);
 	struct amdgpu_bo *abo = ttm_to_amdgpu_bo(bo);
 	struct amdgpu_ttm_tt *gtt;
 	enum ttm_caching caching;
 
 	gtt = kzalloc(sizeof(struct amdgpu_ttm_tt), GFP_KERNEL);
-	if (gtt == NULL) {
+	if (!gtt)
 		return NULL;
-	}
+
 	gtt->gobj = &bo->base;
+	if (adev->gmc.mem_partitions && abo->xcp_id >= 0)
+		gtt->pool_id = KFD_XCP_MEM_ID(adev, abo->xcp_id);
+	else
+		gtt->pool_id = abo->xcp_id;
 
 	if (abo->flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC)
 		caching = ttm_write_combined;
@@ -1064,6 +1113,7 @@ static int amdgpu_ttm_tt_populate(struct ttm_device *bdev,
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bdev);
 	struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm);
+	struct ttm_pool *pool;
 	pgoff_t i;
 	int ret;
 
@@ -1078,7 +1128,11 @@ static int amdgpu_ttm_tt_populate(struct ttm_device *bdev,
 	if (ttm->page_flags & TTM_TT_FLAG_EXTERNAL)
 		return 0;
 
-	ret = ttm_pool_alloc(&adev->mman.bdev.pool, ttm, ctx);
+	if (adev->mman.ttm_pools && gtt->pool_id >= 0)
+		pool = &adev->mman.ttm_pools[gtt->pool_id];
+	else
+		pool = &adev->mman.bdev.pool;
+	ret = ttm_pool_alloc(pool, ttm, ctx);
 	if (ret)
 		return ret;
 
@@ -1099,6 +1153,7 @@ static void amdgpu_ttm_tt_unpopulate(struct ttm_device *bdev,
 {
 	struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm);
 	struct amdgpu_device *adev;
+	struct ttm_pool *pool;
 	pgoff_t i;
 
 	amdgpu_ttm_backend_unbind(bdev, ttm);
@@ -1117,7 +1172,13 @@ static void amdgpu_ttm_tt_unpopulate(struct ttm_device *bdev,
 		ttm->pages[i]->mapping = NULL;
 
 	adev = amdgpu_ttm_adev(bdev);
-	return ttm_pool_free(&adev->mman.bdev.pool, ttm);
+
+	if (adev->mman.ttm_pools && gtt->pool_id >= 0)
+		pool = &adev->mman.ttm_pools[gtt->pool_id];
+	else
+		pool = &adev->mman.bdev.pool;
+
+	return ttm_pool_free(pool, ttm);
 }
 
 /**
@@ -1263,10 +1324,12 @@ uint64_t amdgpu_ttm_tt_pde_flags(struct ttm_tt *ttm, struct ttm_resource *mem)
 		flags |= AMDGPU_PTE_VALID;
 
 	if (mem && (mem->mem_type == TTM_PL_TT ||
-		    mem->mem_type == AMDGPU_PL_PREEMPT)) {
+		    mem->mem_type == AMDGPU_PL_DOORBELL ||
+		    mem->mem_type == AMDGPU_PL_PREEMPT ||
+		    mem->mem_type == AMDGPU_PL_MMIO_REMAP)) {
 		flags |= AMDGPU_PTE_SYSTEM;
 
-		if (ttm->caching == ttm_cached)
+		if (ttm && ttm->caching == ttm_cached)
 			flags |= AMDGPU_PTE_SNOOPED;
 	}
 
@@ -1332,7 +1395,8 @@ static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo,
 	 */
 	dma_resv_for_each_fence(&resv_cursor, bo->base.resv,
 				DMA_RESV_USAGE_BOOKKEEP, f) {
-		if (amdkfd_fence_check_mm(f, current->mm))
+		if (amdkfd_fence_check_mm(f, current->mm) &&
+		    !(place->flags & TTM_PL_FLAG_CONTIGUOUS))
 			return false;
 	}
 
@@ -1414,13 +1478,15 @@ static int amdgpu_ttm_access_memory_sdma(struct ttm_buffer_object *bo,
 		memcpy(adev->mman.sdma_access_ptr, buf, len);
 
 	num_dw = ALIGN(adev->mman.buffer_funcs->copy_num_dw, 8);
-	r = amdgpu_job_alloc_with_ib(adev, &adev->mman.entity,
+	r = amdgpu_job_alloc_with_ib(adev, &adev->mman.high_pr,
 				     AMDGPU_FENCE_OWNER_UNDEFINED,
 				     num_dw * 4, AMDGPU_IB_POOL_DELAYED,
-				     &job);
+				     &job,
+				     AMDGPU_KERNEL_JOB_ID_TTM_ACCESS_MEMORY_SDMA);
 	if (r)
 		goto out;
 
+	mutex_lock(&adev->mman.gtt_window_lock);
 	amdgpu_res_first(abo->tbo.resource, offset, len, &src_mm);
 	src_addr = amdgpu_ttm_domain_start(adev, bo->resource->mem_type) +
 		src_mm.start;
@@ -1429,12 +1495,13 @@ static int amdgpu_ttm_access_memory_sdma(struct ttm_buffer_object *bo,
 		swap(src_addr, dst_addr);
 
 	amdgpu_emit_copy_buffer(adev, &job->ibs[0], src_addr, dst_addr,
-				PAGE_SIZE, false);
+				PAGE_SIZE, 0);
 
 	amdgpu_ring_pad_ib(adev->mman.buffer_funcs_ring, &job->ibs[0]);
 	WARN_ON(job->ibs[0].length_dw > num_dw);
 
 	fence = amdgpu_job_submit(job);
+	mutex_unlock(&adev->mman.gtt_window_lock);
 
 	if (!dma_fence_wait_timeout(fence, false, adev->sdma_timeout))
 		r = -ETIMEDOUT;
@@ -1623,14 +1690,15 @@ static int amdgpu_ttm_training_reserve_vram_fini(struct amdgpu_device *adev)
 	return 0;
 }
 
-static void amdgpu_ttm_training_data_block_init(struct amdgpu_device *adev)
+static void amdgpu_ttm_training_data_block_init(struct amdgpu_device *adev,
+						uint32_t reserve_size)
 {
 	struct psp_memory_training_context *ctx = &adev->psp.mem_train_ctx;
 
 	memset(ctx, 0, sizeof(*ctx));
 
 	ctx->c2p_train_data_offset =
-		ALIGN((adev->gmc.mc_vram_size - adev->mman.discovery_tmr_size - SZ_1M), SZ_1M);
+		ALIGN((adev->gmc.mc_vram_size - reserve_size - SZ_1M), SZ_1M);
 	ctx->p2c_train_data_offset =
 		(adev->gmc.mc_vram_size - GDDR6_MEM_TRAINING_OFFSET);
 	ctx->train_data_size =
@@ -1648,11 +1716,12 @@ static void amdgpu_ttm_training_data_block_init(struct amdgpu_device *adev)
  */
 static int amdgpu_ttm_reserve_tmr(struct amdgpu_device *adev)
 {
-	int ret;
 	struct psp_memory_training_context *ctx = &adev->psp.mem_train_ctx;
 	bool mem_train_support = false;
+	uint32_t reserve_size = 0;
+	int ret;
 
-	if (!amdgpu_sriov_vf(adev)) {
+	if (adev->bios && !amdgpu_sriov_vf(adev)) {
 		if (amdgpu_atomfirmware_mem_training_supported(adev))
 			mem_train_support = true;
 		else
@@ -1666,41 +1735,135 @@ static int amdgpu_ttm_reserve_tmr(struct amdgpu_device *adev)
 	 * Otherwise, fallback to legacy approach to check and reserve tmr block for ip
 	 * discovery data and G6 memory training data respectively
 	 */
-	adev->mman.discovery_tmr_size =
-		amdgpu_atomfirmware_get_fw_reserved_fb_size(adev);
-	if (!adev->mman.discovery_tmr_size)
-		adev->mman.discovery_tmr_size = DISCOVERY_TMR_OFFSET;
+	if (adev->bios)
+		reserve_size =
+			amdgpu_atomfirmware_get_fw_reserved_fb_size(adev);
+
+	if (!adev->bios &&
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3) ||
+	     amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 4) ||
+	     amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 5, 0)))
+		reserve_size = max(reserve_size, (uint32_t)280 << 20);
+	else if (!reserve_size)
+		reserve_size = DISCOVERY_TMR_OFFSET;
 
 	if (mem_train_support) {
 		/* reserve vram for mem train according to TMR location */
-		amdgpu_ttm_training_data_block_init(adev);
+		amdgpu_ttm_training_data_block_init(adev, reserve_size);
 		ret = amdgpu_bo_create_kernel_at(adev,
 						 ctx->c2p_train_data_offset,
 						 ctx->train_data_size,
 						 &ctx->c2p_bo,
 						 NULL);
 		if (ret) {
-			DRM_ERROR("alloc c2p_bo failed(%d)!\n", ret);
+			dev_err(adev->dev, "alloc c2p_bo failed(%d)!\n", ret);
 			amdgpu_ttm_training_reserve_vram_fini(adev);
 			return ret;
 		}
 		ctx->init = PSP_MEM_TRAIN_RESERVE_SUCCESS;
 	}
 
-	ret = amdgpu_bo_create_kernel_at(adev,
-					 adev->gmc.real_vram_size - adev->mman.discovery_tmr_size,
-					 adev->mman.discovery_tmr_size,
-					 &adev->mman.discovery_memory,
-					 NULL);
+	ret = amdgpu_bo_create_kernel_at(
+		adev, adev->gmc.real_vram_size - reserve_size, reserve_size,
+		&adev->mman.fw_reserved_memory, NULL);
 	if (ret) {
-		DRM_ERROR("alloc tmr failed(%d)!\n", ret);
-		amdgpu_bo_free_kernel(&adev->mman.discovery_memory, NULL, NULL);
+		dev_err(adev->dev, "alloc tmr failed(%d)!\n", ret);
+		amdgpu_bo_free_kernel(&adev->mman.fw_reserved_memory, NULL,
+				      NULL);
 		return ret;
 	}
 
 	return 0;
 }
 
+static int amdgpu_ttm_pools_init(struct amdgpu_device *adev)
+{
+	int i;
+
+	if (!adev->gmc.is_app_apu || !adev->gmc.num_mem_partitions)
+		return 0;
+
+	adev->mman.ttm_pools = kcalloc(adev->gmc.num_mem_partitions,
+				       sizeof(*adev->mman.ttm_pools),
+				       GFP_KERNEL);
+	if (!adev->mman.ttm_pools)
+		return -ENOMEM;
+
+	for (i = 0; i < adev->gmc.num_mem_partitions; i++) {
+		ttm_pool_init(&adev->mman.ttm_pools[i], adev->dev,
+			      adev->gmc.mem_partitions[i].numa.node,
+			      TTM_ALLOCATION_POOL_BENEFICIAL_ORDER(get_order(SZ_2M)));
+	}
+	return 0;
+}
+
+static void amdgpu_ttm_pools_fini(struct amdgpu_device *adev)
+{
+	int i;
+
+	if (!adev->gmc.is_app_apu || !adev->mman.ttm_pools)
+		return;
+
+	for (i = 0; i < adev->gmc.num_mem_partitions; i++)
+		ttm_pool_fini(&adev->mman.ttm_pools[i]);
+
+	kfree(adev->mman.ttm_pools);
+	adev->mman.ttm_pools = NULL;
+}
+
+/**
+ * amdgpu_ttm_mmio_remap_bo_init - Allocate the singleton 4K MMIO_REMAP BO
+ * @adev: amdgpu device
+ *
+ * Allocates a one-page (4K) GEM BO in AMDGPU_GEM_DOMAIN_MMIO_REMAP when the
+ * hardware exposes a remap base (adev->rmmio_remap.bus_addr) and the host
+ * PAGE_SIZE is <= AMDGPU_GPU_PAGE_SIZE (4K). The BO is created as a regular
+ * GEM object (amdgpu_bo_create).
+ *
+ * Return:
+ *  * 0 on success or intentional skip (feature not present/unsupported)
+ *  * negative errno on allocation failure
+ */
+static int amdgpu_ttm_mmio_remap_bo_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_bo_param bp;
+	int r;
+
+	/* Skip if HW doesn't expose remap, or if PAGE_SIZE > AMDGPU_GPU_PAGE_SIZE (4K). */
+	if (!adev->rmmio_remap.bus_addr || PAGE_SIZE > AMDGPU_GPU_PAGE_SIZE)
+		return 0;
+
+	memset(&bp, 0, sizeof(bp));
+
+	/* Create exactly one GEM BO in the MMIO_REMAP domain. */
+	bp.type        = ttm_bo_type_device;          /* userspace-mappable GEM */
+	bp.size        = AMDGPU_GPU_PAGE_SIZE;        /* 4K */
+	bp.byte_align  = AMDGPU_GPU_PAGE_SIZE;
+	bp.domain      = AMDGPU_GEM_DOMAIN_MMIO_REMAP;
+	bp.flags       = 0;
+	bp.resv        = NULL;
+	bp.bo_ptr_size = sizeof(struct amdgpu_bo);
+
+	r = amdgpu_bo_create(adev, &bp, &adev->rmmio_remap.bo);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+/**
+ * amdgpu_ttm_mmio_remap_bo_fini - Free the singleton MMIO_REMAP BO
+ * @adev: amdgpu device
+ *
+ * Frees the kernel-owned MMIO_REMAP BO if it was allocated by
+ * amdgpu_ttm_mmio_remap_bo_init().
+ */
+static void amdgpu_ttm_mmio_remap_bo_fini(struct amdgpu_device *adev)
+{
+	amdgpu_bo_unref(&adev->rmmio_remap.bo);
+	adev->rmmio_remap.bo = NULL;
+}
+
 /*
  * amdgpu_ttm_init - Init the memory management (ttm) as well as various
  * gtt/vram related fields.
@@ -1717,24 +1880,37 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
 
 	mutex_init(&adev->mman.gtt_window_lock);
 
+	dma_set_max_seg_size(adev->dev, UINT_MAX);
 	/* No others user of address space so set it to 0 */
 	r = ttm_device_init(&adev->mman.bdev, &amdgpu_bo_driver, adev->dev,
 			       adev_to_drm(adev)->anon_inode->i_mapping,
 			       adev_to_drm(adev)->vma_offset_manager,
-			       adev->need_swiotlb,
-			       dma_addressing_limited(adev->dev));
+			       (adev->need_swiotlb ?
+				TTM_ALLOCATION_POOL_USE_DMA_ALLOC : 0) |
+			       (dma_addressing_limited(adev->dev) ?
+				TTM_ALLOCATION_POOL_USE_DMA32 : 0) |
+			       TTM_ALLOCATION_POOL_BENEFICIAL_ORDER(get_order(SZ_2M)));
 	if (r) {
-		DRM_ERROR("failed initializing buffer object driver(%d).\n", r);
+		dev_err(adev->dev,
+			"failed initializing buffer object driver(%d).\n", r);
 		return r;
 	}
-	adev->mman.initialized = true;
 
-	/* Initialize VRAM pool with all of VRAM divided into pages */
-	r = amdgpu_vram_mgr_init(adev);
+	r = amdgpu_ttm_pools_init(adev);
 	if (r) {
-		DRM_ERROR("Failed initializing VRAM heap.\n");
+		dev_err(adev->dev, "failed to init ttm pools(%d).\n", r);
 		return r;
 	}
+	adev->mman.initialized = true;
+
+	if (!adev->gmc.is_app_apu) {
+		/* Initialize VRAM pool with all of VRAM divided into pages */
+		r = amdgpu_vram_mgr_init(adev);
+		if (r) {
+			dev_err(adev->dev, "Failed initializing VRAM heap.\n");
+			return r;
+		}
+	}
 
 	/* Change the size here instead of the init above so only lpfn is affected */
 	amdgpu_ttm_set_buffer_funcs_status(adev, false);
@@ -1744,6 +1920,9 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
 		adev->mman.aper_base_kaddr = ioremap_cache(adev->gmc.aper_base,
 				adev->gmc.visible_vram_size);
 
+	else if (adev->gmc.is_app_apu)
+		DRM_DEBUG_DRIVER(
+			"No need to ioremap when real vram size is 0\n");
 	else
 #endif
 		adev->mman.aper_base_kaddr = ioremap_wc(adev->gmc.aper_base,
@@ -1755,24 +1934,23 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
 	 *place on the VRAM, so reserve it early.
 	 */
 	r = amdgpu_ttm_fw_reserve_vram_init(adev);
-	if (r) {
+	if (r)
 		return r;
-	}
 
 	/*
-	 *The reserved vram for driver must be pinned to the specified
-	 *place on the VRAM, so reserve it early.
+	 * The reserved VRAM for the driver must be pinned to a specific
+	 * location in VRAM, so reserve it early.
 	 */
 	r = amdgpu_ttm_drv_reserve_vram_init(adev);
 	if (r)
 		return r;
 
 	/*
-	 * only NAVI10 and onwards ASIC support for IP discovery.
-	 * If IP discovery enabled, a block of memory should be
-	 * reserved for IP discovey.
+	 * only NAVI10 and later ASICs support IP discovery.
+	 * If IP discovery is enabled, a block of memory should be
+	 * reserved for it.
 	 */
-	if (adev->mman.discovery_bin) {
+	if (adev->discovery.reserve_tmr) {
 		r = amdgpu_ttm_reserve_tmr(adev);
 		if (r)
 			return r;
@@ -1781,84 +1959,120 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
 	/* allocate memory as required for VGA
 	 * This is used for VGA emulation and pre-OS scanout buffers to
 	 * avoid display artifacts while transitioning between pre-OS
-	 * and driver.  */
-	r = amdgpu_bo_create_kernel_at(adev, 0, adev->mman.stolen_vga_size,
-				       &adev->mman.stolen_vga_memory,
-				       NULL);
-	if (r)
-		return r;
-	r = amdgpu_bo_create_kernel_at(adev, adev->mman.stolen_vga_size,
-				       adev->mman.stolen_extended_size,
-				       &adev->mman.stolen_extended_memory,
-				       NULL);
-	if (r)
-		return r;
-	r = amdgpu_bo_create_kernel_at(adev, adev->mman.stolen_reserved_offset,
-				       adev->mman.stolen_reserved_size,
-				       &adev->mman.stolen_reserved_memory,
-				       NULL);
-	if (r)
-		return r;
+	 * and driver.
+	 */
+	if (!adev->gmc.is_app_apu) {
+		r = amdgpu_bo_create_kernel_at(adev, 0,
+					       adev->mman.stolen_vga_size,
+					       &adev->mman.stolen_vga_memory,
+					       NULL);
+		if (r)
+			return r;
 
-	DRM_INFO("amdgpu: %uM of VRAM memory ready\n",
-		 (unsigned) (adev->gmc.real_vram_size / (1024 * 1024)));
-
-	/* Compute GTT size, either based on 1/2 the size of RAM size
-	 * or whatever the user passed on module init */
-	if (amdgpu_gtt_size == -1) {
-		struct sysinfo si;
-
-		si_meminfo(&si);
-		/* Certain GL unit tests for large textures can cause problems
-		 * with the OOM killer since there is no way to link this memory
-		 * to a process.  This was originally mitigated (but not necessarily
-		 * eliminated) by limiting the GTT size.  The problem is this limit
-		 * is often too low for many modern games so just make the limit 1/2
-		 * of system memory which aligns with TTM. The OOM accounting needs
-		 * to be addressed, but we shouldn't prevent common 3D applications
-		 * from being usable just to potentially mitigate that corner case.
-		 */
-		gtt_size = max((AMDGPU_DEFAULT_GTT_SIZE_MB << 20),
-			       (u64)si.totalram * si.mem_unit / 2);
+		r = amdgpu_bo_create_kernel_at(adev, adev->mman.stolen_vga_size,
+					       adev->mman.stolen_extended_size,
+					       &adev->mman.stolen_extended_memory,
+					       NULL);
+
+		if (r)
+			return r;
+
+		r = amdgpu_bo_create_kernel_at(adev,
+					       adev->mman.stolen_reserved_offset,
+					       adev->mman.stolen_reserved_size,
+					       &adev->mman.stolen_reserved_memory,
+					       NULL);
+		if (r)
+			return r;
 	} else {
-		gtt_size = (uint64_t)amdgpu_gtt_size << 20;
+		DRM_DEBUG_DRIVER("Skipped stolen memory reservation\n");
+	}
+
+	dev_info(adev->dev, "amdgpu: %uM of VRAM memory ready\n",
+		 (unsigned int)(adev->gmc.real_vram_size / (1024 * 1024)));
+
+	/* Compute GTT size, either based on TTM limit
+	 * or whatever the user passed on module init.
+	 */
+	gtt_size = ttm_tt_pages_limit() << PAGE_SHIFT;
+	if (amdgpu_gtt_size != -1) {
+		uint64_t configured_size = (uint64_t)amdgpu_gtt_size << 20;
+
+		drm_warn(&adev->ddev,
+			"Configuring gttsize via module parameter is deprecated, please use ttm.pages_limit\n");
+		if (gtt_size != configured_size)
+			drm_warn(&adev->ddev,
+				"GTT size has been set as %llu but TTM size has been set as %llu, this is unusual\n",
+				configured_size, gtt_size);
+
+		gtt_size = configured_size;
 	}
 
 	/* Initialize GTT memory pool */
 	r = amdgpu_gtt_mgr_init(adev, gtt_size);
 	if (r) {
-		DRM_ERROR("Failed initializing GTT heap.\n");
+		dev_err(adev->dev, "Failed initializing GTT heap.\n");
 		return r;
 	}
-	DRM_INFO("amdgpu: %uM of GTT memory ready.\n",
-		 (unsigned)(gtt_size / (1024 * 1024)));
+	dev_info(adev->dev, "amdgpu: %uM of GTT memory ready.\n",
+		 (unsigned int)(gtt_size / (1024 * 1024)));
+
+	if (adev->flags & AMD_IS_APU) {
+		if (adev->gmc.real_vram_size < gtt_size)
+			adev->apu_prefer_gtt = true;
+	}
+
+	/* Initialize doorbell pool on PCI BAR */
+	r = amdgpu_ttm_init_on_chip(adev, AMDGPU_PL_DOORBELL, adev->doorbell.size / PAGE_SIZE);
+	if (r) {
+		dev_err(adev->dev, "Failed initializing doorbell heap.\n");
+		return r;
+	}
+
+	/* Create a boorbell page for kernel usages */
+	r = amdgpu_doorbell_create_kernel_doorbells(adev);
+	if (r) {
+		dev_err(adev->dev, "Failed to initialize kernel doorbells.\n");
+		return r;
+	}
+
+	/* Initialize MMIO-remap pool (single page 4K) */
+	r = amdgpu_ttm_init_on_chip(adev, AMDGPU_PL_MMIO_REMAP, 1);
+	if (r) {
+		dev_err(adev->dev, "Failed initializing MMIO-remap heap.\n");
+		return r;
+	}
+
+	/* Allocate the singleton MMIO_REMAP BO (4K) if supported */
+	r = amdgpu_ttm_mmio_remap_bo_init(adev);
+	if (r)
+		return r;
 
 	/* Initialize preemptible memory pool */
 	r = amdgpu_preempt_mgr_init(adev);
 	if (r) {
-		DRM_ERROR("Failed initializing PREEMPT heap.\n");
+		dev_err(adev->dev, "Failed initializing PREEMPT heap.\n");
 		return r;
 	}
 
 	/* Initialize various on-chip memory pools */
 	r = amdgpu_ttm_init_on_chip(adev, AMDGPU_PL_GDS, adev->gds.gds_size);
 	if (r) {
-		DRM_ERROR("Failed initializing GDS heap.\n");
+		dev_err(adev->dev, "Failed initializing GDS heap.\n");
 		return r;
 	}
 
 	r = amdgpu_ttm_init_on_chip(adev, AMDGPU_PL_GWS, adev->gds.gws_size);
 	if (r) {
-		DRM_ERROR("Failed initializing gws heap.\n");
+		dev_err(adev->dev, "Failed initializing gws heap.\n");
 		return r;
 	}
 
 	r = amdgpu_ttm_init_on_chip(adev, AMDGPU_PL_OA, adev->gds.oa_size);
 	if (r) {
-		DRM_ERROR("Failed initializing oa heap.\n");
+		dev_err(adev->dev, "Failed initializing oa heap.\n");
 		return r;
 	}
-
 	if (amdgpu_bo_create_kernel(adev, PAGE_SIZE, PAGE_SIZE,
 				AMDGPU_GEM_DOMAIN_GTT,
 				&adev->mman.sdma_access_bo, NULL,
@@ -1874,20 +2088,30 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
 void amdgpu_ttm_fini(struct amdgpu_device *adev)
 {
 	int idx;
+
 	if (!adev->mman.initialized)
 		return;
 
+	amdgpu_ttm_pools_fini(adev);
+
 	amdgpu_ttm_training_reserve_vram_fini(adev);
 	/* return the stolen vga memory back to VRAM */
-	amdgpu_bo_free_kernel(&adev->mman.stolen_vga_memory, NULL, NULL);
-	amdgpu_bo_free_kernel(&adev->mman.stolen_extended_memory, NULL, NULL);
-	/* return the IP Discovery TMR memory back to VRAM */
-	amdgpu_bo_free_kernel(&adev->mman.discovery_memory, NULL, NULL);
-	if (adev->mman.stolen_reserved_size)
-		amdgpu_bo_free_kernel(&adev->mman.stolen_reserved_memory,
-				      NULL, NULL);
+	if (!adev->gmc.is_app_apu) {
+		amdgpu_bo_free_kernel(&adev->mman.stolen_vga_memory, NULL, NULL);
+		amdgpu_bo_free_kernel(&adev->mman.stolen_extended_memory, NULL, NULL);
+		/* return the FW reserved memory back to VRAM */
+		amdgpu_bo_free_kernel(&adev->mman.fw_reserved_memory, NULL,
+				      NULL);
+		amdgpu_bo_free_kernel(&adev->mman.fw_reserved_memory_extend, NULL,
+				      NULL);
+		if (adev->mman.stolen_reserved_size)
+			amdgpu_bo_free_kernel(&adev->mman.stolen_reserved_memory,
+					      NULL, NULL);
+	}
 	amdgpu_bo_free_kernel(&adev->mman.sdma_access_bo, NULL,
 					&adev->mman.sdma_access_ptr);
+
+	amdgpu_ttm_mmio_remap_bo_fini(adev);
 	amdgpu_ttm_fw_reserve_vram_fini(adev);
 	amdgpu_ttm_drv_reserve_vram_fini(adev);
 
@@ -1900,15 +2124,20 @@ void amdgpu_ttm_fini(struct amdgpu_device *adev)
 		drm_dev_exit(idx);
 	}
 
-	amdgpu_vram_mgr_fini(adev);
+	if (!adev->gmc.is_app_apu)
+		amdgpu_vram_mgr_fini(adev);
 	amdgpu_gtt_mgr_fini(adev);
 	amdgpu_preempt_mgr_fini(adev);
+	amdgpu_doorbell_fini(adev);
+
 	ttm_range_man_fini(&adev->mman.bdev, AMDGPU_PL_GDS);
 	ttm_range_man_fini(&adev->mman.bdev, AMDGPU_PL_GWS);
 	ttm_range_man_fini(&adev->mman.bdev, AMDGPU_PL_OA);
+	ttm_range_man_fini(&adev->mman.bdev, AMDGPU_PL_DOORBELL);
+	ttm_range_man_fini(&adev->mman.bdev, AMDGPU_PL_MMIO_REMAP);
 	ttm_device_fini(&adev->mman.bdev);
 	adev->mman.initialized = false;
-	DRM_INFO("amdgpu: ttm finalized\n");
+	dev_info(adev->dev, "amdgpu: ttm finalized\n");
 }
 
 /**
@@ -1927,7 +2156,7 @@ void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, bool enable)
 	int r;
 
 	if (!adev->mman.initialized || amdgpu_in_reset(adev) ||
-	    adev->mman.buffer_funcs_enabled == enable)
+	    adev->mman.buffer_funcs_enabled == enable || adev->gmc.is_app_apu)
 		return;
 
 	if (enable) {
@@ -1936,18 +2165,32 @@ void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, bool enable)
 
 		ring = adev->mman.buffer_funcs_ring;
 		sched = &ring->sched;
-		r = drm_sched_entity_init(&adev->mman.entity,
+		r = drm_sched_entity_init(&adev->mman.high_pr,
 					  DRM_SCHED_PRIORITY_KERNEL, &sched,
 					  1, NULL);
 		if (r) {
-			DRM_ERROR("Failed setting up TTM BO move entity (%d)\n",
-				  r);
+			dev_err(adev->dev,
+				"Failed setting up TTM BO move entity (%d)\n",
+				r);
 			return;
 		}
+
+		r = drm_sched_entity_init(&adev->mman.low_pr,
+					  DRM_SCHED_PRIORITY_NORMAL, &sched,
+					  1, NULL);
+		if (r) {
+			dev_err(adev->dev,
+				"Failed setting up TTM BO move entity (%d)\n",
+				r);
+			goto error_free_entity;
+		}
 	} else {
-		drm_sched_entity_destroy(&adev->mman.entity);
-		dma_fence_put(man->move);
-		man->move = NULL;
+		drm_sched_entity_destroy(&adev->mman.high_pr);
+		drm_sched_entity_destroy(&adev->mman.low_pr);
+		/* Drop all the old fences since re-creating the scheduler entities
+		 * will allocate new contexts.
+		 */
+		ttm_resource_manager_cleanup(man);
 	}
 
 	/* this just adjusts TTM size idea, which sets lpfn to the correct value */
@@ -1957,6 +2200,11 @@ void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, bool enable)
 		size = adev->gmc.visible_vram_size;
 	man->size = size;
 	adev->mman.buffer_funcs_enabled = enable;
+
+	return;
+
+error_free_entity:
+	drm_sched_entity_destroy(&adev->mman.high_pr);
 }
 
 static int amdgpu_ttm_prepare_job(struct amdgpu_device *adev,
@@ -1964,16 +2212,18 @@ static int amdgpu_ttm_prepare_job(struct amdgpu_device *adev,
 				  unsigned int num_dw,
 				  struct dma_resv *resv,
 				  bool vm_needs_flush,
-				  struct amdgpu_job **job)
+				  struct amdgpu_job **job,
+				  bool delayed, u64 k_job_id)
 {
 	enum amdgpu_ib_pool_type pool = direct_submit ?
 		AMDGPU_IB_POOL_DIRECT :
 		AMDGPU_IB_POOL_DELAYED;
 	int r;
-
-	r = amdgpu_job_alloc_with_ib(adev, &adev->mman.entity,
+	struct drm_sched_entity *entity = delayed ? &adev->mman.low_pr :
+						    &adev->mman.high_pr;
+	r = amdgpu_job_alloc_with_ib(adev, entity,
 				     AMDGPU_FENCE_OWNER_UNDEFINED,
-				     num_dw * 4, pool, job);
+				     num_dw * 4, pool, job, k_job_id);
 	if (r)
 		return r;
 
@@ -1994,17 +2244,18 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset,
 		       uint64_t dst_offset, uint32_t byte_count,
 		       struct dma_resv *resv,
 		       struct dma_fence **fence, bool direct_submit,
-		       bool vm_needs_flush, bool tmz)
+		       bool vm_needs_flush, uint32_t copy_flags)
 {
 	struct amdgpu_device *adev = ring->adev;
-	unsigned num_loops, num_dw;
+	unsigned int num_loops, num_dw;
 	struct amdgpu_job *job;
 	uint32_t max_bytes;
-	unsigned i;
+	unsigned int i;
 	int r;
 
 	if (!direct_submit && !ring->sched.ready) {
-		DRM_ERROR("Trying to move memory with ring turned off.\n");
+		dev_err(adev->dev,
+			"Trying to move memory with ring turned off.\n");
 		return -EINVAL;
 	}
 
@@ -2012,7 +2263,8 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset,
 	num_loops = DIV_ROUND_UP(byte_count, max_bytes);
 	num_dw = ALIGN(num_loops * adev->mman.buffer_funcs->copy_num_dw, 8);
 	r = amdgpu_ttm_prepare_job(adev, direct_submit, num_dw,
-				   resv, vm_needs_flush, &job);
+				   resv, vm_needs_flush, &job, false,
+				   AMDGPU_KERNEL_JOB_ID_TTM_COPY_BUFFER);
 	if (r)
 		return r;
 
@@ -2020,8 +2272,7 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset,
 		uint32_t cur_size_in_bytes = min(byte_count, max_bytes);
 
 		amdgpu_emit_copy_buffer(adev, &job->ibs[0], src_offset,
-					dst_offset, cur_size_in_bytes, tmz);
-
+					dst_offset, cur_size_in_bytes, copy_flags);
 		src_offset += cur_size_in_bytes;
 		dst_offset += cur_size_in_bytes;
 		byte_count -= cur_size_in_bytes;
@@ -2040,7 +2291,7 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset,
 
 error_free:
 	amdgpu_job_free(job);
-	DRM_ERROR("Error scheduling IBs (%d)\n", r);
+	dev_err(adev->dev, "Error scheduling IBs (%d)\n", r);
 	return r;
 }
 
@@ -2048,7 +2299,8 @@ static int amdgpu_ttm_fill_mem(struct amdgpu_ring *ring, uint32_t src_data,
 			       uint64_t dst_addr, uint32_t byte_count,
 			       struct dma_resv *resv,
 			       struct dma_fence **fence,
-			       bool vm_needs_flush)
+			       bool vm_needs_flush, bool delayed,
+			       u64 k_job_id)
 {
 	struct amdgpu_device *adev = ring->adev;
 	unsigned int num_loops, num_dw;
@@ -2061,7 +2313,7 @@ static int amdgpu_ttm_fill_mem(struct amdgpu_ring *ring, uint32_t src_data,
 	num_loops = DIV_ROUND_UP_ULL(byte_count, max_bytes);
 	num_dw = ALIGN(num_loops * adev->mman.buffer_funcs->fill_num_dw, 8);
 	r = amdgpu_ttm_prepare_job(adev, false, num_dw, resv, vm_needs_flush,
-				   &job);
+				   &job, delayed, k_job_id);
 	if (r)
 		return r;
 
@@ -2081,10 +2333,78 @@ static int amdgpu_ttm_fill_mem(struct amdgpu_ring *ring, uint32_t src_data,
 	return 0;
 }
 
+/**
+ * amdgpu_ttm_clear_buffer - clear memory buffers
+ * @bo: amdgpu buffer object
+ * @resv: reservation object
+ * @fence: dma_fence associated with the operation
+ *
+ * Clear the memory buffer resource.
+ *
+ * Returns:
+ * 0 for success or a negative error code on failure.
+ */
+int amdgpu_ttm_clear_buffer(struct amdgpu_bo *bo,
+			    struct dma_resv *resv,
+			    struct dma_fence **fence)
+{
+	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
+	struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
+	struct amdgpu_res_cursor cursor;
+	u64 addr;
+	int r = 0;
+
+	if (!adev->mman.buffer_funcs_enabled)
+		return -EINVAL;
+
+	if (!fence)
+		return -EINVAL;
+
+	*fence = dma_fence_get_stub();
+
+	amdgpu_res_first(bo->tbo.resource, 0, amdgpu_bo_size(bo), &cursor);
+
+	mutex_lock(&adev->mman.gtt_window_lock);
+	while (cursor.remaining) {
+		struct dma_fence *next = NULL;
+		u64 size;
+
+		if (amdgpu_res_cleared(&cursor)) {
+			amdgpu_res_next(&cursor, cursor.size);
+			continue;
+		}
+
+		/* Never clear more than 256MiB at once to avoid timeouts */
+		size = min(cursor.size, 256ULL << 20);
+
+		r = amdgpu_ttm_map_buffer(&bo->tbo, bo->tbo.resource, &cursor,
+					  1, ring, false, &size, &addr);
+		if (r)
+			goto err;
+
+		r = amdgpu_ttm_fill_mem(ring, 0, addr, size, resv,
+					&next, true, true,
+					AMDGPU_KERNEL_JOB_ID_TTM_CLEAR_BUFFER);
+		if (r)
+			goto err;
+
+		dma_fence_put(*fence);
+		*fence = next;
+
+		amdgpu_res_next(&cursor, size);
+	}
+err:
+	mutex_unlock(&adev->mman.gtt_window_lock);
+
+	return r;
+}
+
 int amdgpu_fill_buffer(struct amdgpu_bo *bo,
 			uint32_t src_data,
 			struct dma_resv *resv,
-			struct dma_fence **f)
+			struct dma_fence **f,
+			bool delayed,
+			u64 k_job_id)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
 	struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
@@ -2093,7 +2413,8 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo,
 	int r;
 
 	if (!adev->mman.buffer_funcs_enabled) {
-		DRM_ERROR("Trying to clear memory with ring turned off.\n");
+		dev_err(adev->dev,
+			"Trying to clear memory with ring turned off.\n");
 		return -EINVAL;
 	}
 
@@ -2113,7 +2434,7 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo,
 			goto error;
 
 		r = amdgpu_ttm_fill_mem(ring, src_data, to, cur_size, resv,
-					&next, true);
+					&next, true, delayed, k_job_id);
 		if (r)
 			goto error;
 
@@ -2153,7 +2474,7 @@ int amdgpu_ttm_evict_resources(struct amdgpu_device *adev, int mem_type)
 		man = ttm_manager_type(&adev->mman.bdev, mem_type);
 		break;
 	default:
-		DRM_ERROR("Trying to evict invalid memory type\n");
+		dev_err(adev->dev, "Trying to evict invalid memory type\n");
 		return -EINVAL;
 	}
 
@@ -2164,7 +2485,7 @@ int amdgpu_ttm_evict_resources(struct amdgpu_device *adev, int mem_type)
 
 static int amdgpu_ttm_page_pool_show(struct seq_file *m, void *unused)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct amdgpu_device *adev = m->private;
 
 	return ttm_pool_debugfs(&adev->mman.bdev.pool, m);
 }
@@ -2278,7 +2599,7 @@ static ssize_t amdgpu_iomem_read(struct file *f, char __user *buf,
 		struct page *p;
 		void *ptr;
 
-		bytes = bytes < size ? bytes : size;
+		bytes = min(bytes, size);
 
 		/* Translate the bus address to a physical address.  If
 		 * the domain is NULL it means there is no IOMMU active
@@ -2333,7 +2654,7 @@ static ssize_t amdgpu_iomem_write(struct file *f, const char __user *buf,
 		struct page *p;
 		void *ptr;
 
-		bytes = bytes < size ? bytes : size;
+		bytes = min(bytes, size);
 
 		addr = dom ? iommu_iova_to_phys(dom, addr) : addr;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
index e2cd5894afc9..577ee04ce0bf 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
@@ -26,18 +26,23 @@
 
 #include <linux/dma-direction.h>
 #include <drm/gpu_scheduler.h>
+#include <drm/ttm/ttm_placement.h>
 #include "amdgpu_vram_mgr.h"
-#include "amdgpu.h"
+#include "amdgpu_hmm.h"
 
 #define AMDGPU_PL_GDS		(TTM_PL_PRIV + 0)
 #define AMDGPU_PL_GWS		(TTM_PL_PRIV + 1)
 #define AMDGPU_PL_OA		(TTM_PL_PRIV + 2)
 #define AMDGPU_PL_PREEMPT	(TTM_PL_PRIV + 3)
+#define AMDGPU_PL_DOORBELL	(TTM_PL_PRIV + 4)
+#define AMDGPU_PL_MMIO_REMAP	(TTM_PL_PRIV + 5)
+#define __AMDGPU_PL_NUM	(TTM_PL_PRIV + 6)
 
 #define AMDGPU_GTT_MAX_TRANSFER_SIZE	512
 #define AMDGPU_GTT_NUM_TRANSFER_WINDOWS	2
 
-#define AMDGPU_POISON	0xd0bed0be
+extern const struct attribute_group amdgpu_vram_mgr_attr_group;
+extern const struct attribute_group amdgpu_gtt_mgr_attr_group;
 
 struct hmm_range;
 
@@ -49,6 +54,7 @@ struct amdgpu_gtt_mgr {
 
 struct amdgpu_mman {
 	struct ttm_device		bdev;
+	struct ttm_pool			*ttm_pools;
 	bool				initialized;
 	void __iomem			*aper_base_kaddr;
 
@@ -58,8 +64,10 @@ struct amdgpu_mman {
 	bool					buffer_funcs_enabled;
 
 	struct mutex				gtt_window_lock;
-	/* Scheduler entity for buffer moves */
-	struct drm_sched_entity			entity;
+	/* High priority scheduler entity for buffer moves */
+	struct drm_sched_entity			high_pr;
+	/* Low priority scheduler entity for VRAM clearing */
+	struct drm_sched_entity			low_pr;
 
 	struct amdgpu_vram_mgr vram_mgr;
 	struct amdgpu_gtt_mgr gtt_mgr;
@@ -75,10 +83,9 @@ struct amdgpu_mman {
 	uint64_t		stolen_reserved_offset;
 	uint64_t		stolen_reserved_size;
 
-	/* discovery */
-	uint8_t				*discovery_bin;
-	uint32_t			discovery_tmr_size;
-	struct amdgpu_bo		*discovery_memory;
+	/* fw reserved memory */
+	struct amdgpu_bo		*fw_reserved_memory;
+	struct amdgpu_bo		*fw_reserved_memory_extend;
 
 	/* firmware VRAM reservation */
 	u64		fw_vram_usage_start_offset;
@@ -103,6 +110,23 @@ struct amdgpu_copy_mem {
 	unsigned long			offset;
 };
 
+#define AMDGPU_COPY_FLAGS_TMZ		(1 << 0)
+#define AMDGPU_COPY_FLAGS_READ_DECOMPRESSED	(1 << 1)
+#define AMDGPU_COPY_FLAGS_WRITE_COMPRESSED	(1 << 2)
+#define AMDGPU_COPY_FLAGS_MAX_COMPRESSED_SHIFT		3
+#define AMDGPU_COPY_FLAGS_MAX_COMPRESSED_MASK		0x03
+#define AMDGPU_COPY_FLAGS_NUMBER_TYPE_SHIFT		5
+#define AMDGPU_COPY_FLAGS_NUMBER_TYPE_MASK		0x07
+#define AMDGPU_COPY_FLAGS_DATA_FORMAT_SHIFT		8
+#define AMDGPU_COPY_FLAGS_DATA_FORMAT_MASK		0x3f
+#define AMDGPU_COPY_FLAGS_WRITE_COMPRESS_DISABLE_SHIFT	14
+#define AMDGPU_COPY_FLAGS_WRITE_COMPRESS_DISABLE_MASK	0x1
+
+#define AMDGPU_COPY_FLAGS_SET(field, value) \
+	(((__u32)(value) & AMDGPU_COPY_FLAGS_##field##_MASK) << AMDGPU_COPY_FLAGS_##field##_SHIFT)
+#define AMDGPU_COPY_FLAGS_GET(value, field) \
+	(((__u32)(value) >> AMDGPU_COPY_FLAGS_##field##_SHIFT) & AMDGPU_COPY_FLAGS_##field##_MASK)
+
 int amdgpu_gtt_mgr_init(struct amdgpu_device *adev, uint64_t gtt_size);
 void amdgpu_gtt_mgr_fini(struct amdgpu_device *adev);
 int amdgpu_preempt_mgr_init(struct amdgpu_device *adev);
@@ -130,58 +154,46 @@ int amdgpu_vram_mgr_reserve_range(struct amdgpu_vram_mgr *mgr,
 				  uint64_t start, uint64_t size);
 int amdgpu_vram_mgr_query_page_status(struct amdgpu_vram_mgr *mgr,
 				      uint64_t start);
+void amdgpu_vram_mgr_clear_reset_blocks(struct amdgpu_device *adev);
+
+bool amdgpu_res_cpu_visible(struct amdgpu_device *adev,
+			    struct ttm_resource *res);
 
 int amdgpu_ttm_init(struct amdgpu_device *adev);
 void amdgpu_ttm_fini(struct amdgpu_device *adev);
 void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev,
 					bool enable);
-
 int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset,
 		       uint64_t dst_offset, uint32_t byte_count,
 		       struct dma_resv *resv,
 		       struct dma_fence **fence, bool direct_submit,
-		       bool vm_needs_flush, bool tmz);
-int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev,
-			       const struct amdgpu_copy_mem *src,
-			       const struct amdgpu_copy_mem *dst,
-			       uint64_t size, bool tmz,
-			       struct dma_resv *resv,
-			       struct dma_fence **f);
+		       bool vm_needs_flush, uint32_t copy_flags);
+int amdgpu_ttm_clear_buffer(struct amdgpu_bo *bo,
+			    struct dma_resv *resv,
+			    struct dma_fence **fence);
 int amdgpu_fill_buffer(struct amdgpu_bo *bo,
 			uint32_t src_data,
 			struct dma_resv *resv,
-			struct dma_fence **fence);
+			struct dma_fence **fence,
+			bool delayed,
+			u64 k_job_id);
 
 int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo);
 void amdgpu_ttm_recover_gart(struct ttm_buffer_object *tbo);
 uint64_t amdgpu_ttm_domain_start(struct amdgpu_device *adev, uint32_t type);
 
 #if IS_ENABLED(CONFIG_DRM_AMDGPU_USERPTR)
-int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages,
-				 struct hmm_range **range);
-void amdgpu_ttm_tt_discard_user_pages(struct ttm_tt *ttm,
-				      struct hmm_range *range);
-bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm,
-				       struct hmm_range *range);
+int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo,
+				 struct amdgpu_hmm_range *range);
 #else
 static inline int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo,
-					       struct page **pages,
-					       struct hmm_range **range)
+					       struct amdgpu_hmm_range *range)
 {
 	return -EPERM;
 }
-static inline void amdgpu_ttm_tt_discard_user_pages(struct ttm_tt *ttm,
-						    struct hmm_range *range)
-{
-}
-static inline bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm,
-						     struct hmm_range *range)
-{
-	return false;
-}
 #endif
 
-void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages);
+void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct amdgpu_hmm_range *range);
 int amdgpu_ttm_tt_get_userptr(const struct ttm_buffer_object *tbo,
 			      uint64_t *user_addr);
 int amdgpu_ttm_tt_set_userptr(struct ttm_buffer_object *bo,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c
index f76b1cb8baf8..e96f24e9ad57 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c
@@ -28,6 +28,13 @@
 #include "amdgpu.h"
 #include "amdgpu_ucode.h"
 
+#define AMDGPU_UCODE_NAME_MAX		(128)
+
+static const struct kicker_device kicker_device_list[] = {
+	{0x744B, 0x00},
+	{0x7551, 0xC8}
+};
+
 static void amdgpu_ucode_print_common_hdr(const struct common_firmware_header *hdr)
 {
 	DRM_DEBUG("size_bytes: %u\n", le32_to_cpu(hdr->size_bytes));
@@ -323,6 +330,12 @@ void amdgpu_ucode_print_sdma_hdr(const struct common_firmware_header *hdr)
 		DRM_DEBUG("ctl_ucode_offset: %u\n", le32_to_cpu(sdma_hdr->ctl_ucode_offset));
 		DRM_DEBUG("ctl_jt_offset: %u\n", le32_to_cpu(sdma_hdr->ctl_jt_offset));
 		DRM_DEBUG("ctl_jt_size: %u\n", le32_to_cpu(sdma_hdr->ctl_jt_size));
+	} else if (version_major == 3) {
+		const struct sdma_firmware_header_v3_0 *sdma_hdr =
+			container_of(hdr, struct sdma_firmware_header_v3_0, header);
+
+		DRM_DEBUG("ucode_reversion: %u\n",
+			  le32_to_cpu(sdma_hdr->ucode_feature_version));
 	} else {
 		DRM_ERROR("Unknown SDMA ucode version: %u.%u\n",
 			  version_major, version_minor);
@@ -556,6 +569,8 @@ amdgpu_ucode_get_load_type(struct amdgpu_device *adev, int load_type)
 	default:
 		if (!load_type)
 			return AMDGPU_FW_LOAD_DIRECT;
+		else if (load_type == 3)
+			return AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO;
 		else
 			return AMDGPU_FW_LOAD_PSP;
 	}
@@ -642,6 +657,8 @@ const char *amdgpu_ucode_name(enum AMDGPU_UCODE_ID ucode_id)
 		return "SMC";
 	case AMDGPU_UCODE_ID_PPTABLE:
 		return "PPTABLE";
+	case AMDGPU_UCODE_ID_P2S_TABLE:
+		return "P2STABLE";
 	case AMDGPU_UCODE_ID_UVD:
 		return "UVD";
 	case AMDGPU_UCODE_ID_UVD1:
@@ -664,20 +681,70 @@ const char *amdgpu_ucode_name(enum AMDGPU_UCODE_ID ucode_id)
 		return "DMCUB";
 	case AMDGPU_UCODE_ID_CAP:
 		return "CAP";
+	case AMDGPU_UCODE_ID_VPE_CTX:
+		return "VPE_CTX";
+	case AMDGPU_UCODE_ID_VPE_CTL:
+		return "VPE_CTL";
+	case AMDGPU_UCODE_ID_VPE:
+		return "VPE";
+	case AMDGPU_UCODE_ID_UMSCH_MM_UCODE:
+		return "UMSCH_MM_UCODE";
+	case AMDGPU_UCODE_ID_UMSCH_MM_DATA:
+		return "UMSCH_MM_DATA";
+	case AMDGPU_UCODE_ID_UMSCH_MM_CMD_BUFFER:
+		return "UMSCH_MM_CMD_BUFFER";
+	case AMDGPU_UCODE_ID_JPEG_RAM:
+		return "JPEG";
+	case AMDGPU_UCODE_ID_SDMA_RS64:
+		return "RS64_SDMA";
+	case AMDGPU_UCODE_ID_CP_RS64_PFP:
+		return "RS64_PFP";
+	case AMDGPU_UCODE_ID_CP_RS64_ME:
+		return "RS64_ME";
+	case AMDGPU_UCODE_ID_CP_RS64_MEC:
+		return "RS64_MEC";
+	case AMDGPU_UCODE_ID_CP_RS64_PFP_P0_STACK:
+		return "RS64_PFP_P0_STACK";
+	case AMDGPU_UCODE_ID_CP_RS64_PFP_P1_STACK:
+		return "RS64_PFP_P1_STACK";
+	case AMDGPU_UCODE_ID_CP_RS64_ME_P0_STACK:
+		return "RS64_ME_P0_STACK";
+	case AMDGPU_UCODE_ID_CP_RS64_ME_P1_STACK:
+		return "RS64_ME_P1_STACK";
+	case AMDGPU_UCODE_ID_CP_RS64_MEC_P0_STACK:
+		return "RS64_MEC_P0_STACK";
+	case AMDGPU_UCODE_ID_CP_RS64_MEC_P1_STACK:
+		return "RS64_MEC_P1_STACK";
+	case AMDGPU_UCODE_ID_CP_RS64_MEC_P2_STACK:
+		return "RS64_MEC_P2_STACK";
+	case AMDGPU_UCODE_ID_CP_RS64_MEC_P3_STACK:
+		return "RS64_MEC_P3_STACK";
+	case AMDGPU_UCODE_ID_ISP:
+		return "ISP";
 	default:
 		return "UNKNOWN UCODE";
 	}
 }
 
+static inline int amdgpu_ucode_is_valid(uint32_t fw_version)
+{
+	if (!fw_version)
+		return -EINVAL;
+
+	return 0;
+}
+
 #define FW_VERSION_ATTR(name, mode, field)				\
 static ssize_t show_##name(struct device *dev,				\
-			  struct device_attribute *attr,		\
-			  char *buf)					\
+			   struct device_attribute *attr, char *buf)	\
 {									\
 	struct drm_device *ddev = dev_get_drvdata(dev);			\
 	struct amdgpu_device *adev = drm_to_adev(ddev);			\
 									\
-	return sysfs_emit(buf, "0x%08x\n", adev->field);	\
+	if (!buf)							\
+		return amdgpu_ucode_is_valid(adev->field);		\
+									\
+	return sysfs_emit(buf, "0x%08x\n", adev->field);		\
 }									\
 static DEVICE_ATTR(name, mode, show_##name, NULL)
 
@@ -703,6 +770,10 @@ FW_VERSION_ATTR(sdma_fw_version, 0444, sdma.instance[0].fw_version);
 FW_VERSION_ATTR(sdma2_fw_version, 0444, sdma.instance[1].fw_version);
 FW_VERSION_ATTR(vcn_fw_version, 0444, vcn.fw_version);
 FW_VERSION_ATTR(dmcu_fw_version, 0444, dm.dmcu_fw_version);
+FW_VERSION_ATTR(dmcub_fw_version, 0444, dm.dmcub_fw_version);
+FW_VERSION_ATTR(mes_fw_version, 0444, mes.sched_version & AMDGPU_MES_VERSION_MASK);
+FW_VERSION_ATTR(mes_kiq_fw_version, 0444, mes.kiq_version & AMDGPU_MES_VERSION_MASK);
+FW_VERSION_ATTR(pldm_fw_version, 0444, firmware.pldm_version);
 
 static struct attribute *fw_attrs[] = {
 	&dev_attr_vce_fw_version.attr, &dev_attr_uvd_fw_version.attr,
@@ -715,13 +786,30 @@ static struct attribute *fw_attrs[] = {
 	&dev_attr_ta_ras_fw_version.attr, &dev_attr_ta_xgmi_fw_version.attr,
 	&dev_attr_smc_fw_version.attr, &dev_attr_sdma_fw_version.attr,
 	&dev_attr_sdma2_fw_version.attr, &dev_attr_vcn_fw_version.attr,
-	&dev_attr_dmcu_fw_version.attr, &dev_attr_imu_fw_version.attr,
+	&dev_attr_dmcu_fw_version.attr, &dev_attr_dmcub_fw_version.attr,
+	&dev_attr_imu_fw_version.attr, &dev_attr_mes_fw_version.attr,
+	&dev_attr_mes_kiq_fw_version.attr, &dev_attr_pldm_fw_version.attr,
 	NULL
 };
 
+#define to_dev_attr(x) container_of(x, struct device_attribute, attr)
+
+static umode_t amdgpu_ucode_sys_visible(struct kobject *kobj,
+					struct attribute *attr, int idx)
+{
+	struct device_attribute *dev_attr = to_dev_attr(attr);
+	struct device *dev = kobj_to_dev(kobj);
+
+	if (dev_attr->show(dev, dev_attr, NULL) == -EINVAL)
+		return 0;
+
+	return attr->mode;
+}
+
 static const struct attribute_group fw_attr_group = {
 	.name = "fw_version",
-	.attrs = fw_attrs
+	.attrs = fw_attrs,
+	.is_visible = amdgpu_ucode_sys_visible
 };
 
 int amdgpu_ucode_sysfs_init(struct amdgpu_device *adev)
@@ -745,10 +833,13 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev,
 	const struct dmcub_firmware_header_v1_0 *dmcub_hdr = NULL;
 	const struct mes_firmware_header_v1_0 *mes_hdr = NULL;
 	const struct sdma_firmware_header_v2_0 *sdma_hdr = NULL;
+	const struct sdma_firmware_header_v3_0 *sdmav3_hdr = NULL;
 	const struct imu_firmware_header_v1_0 *imu_hdr = NULL;
+	const struct vpe_firmware_header_v1_0 *vpe_hdr = NULL;
+	const struct umsch_mm_firmware_header_v1_0 *umsch_mm_hdr = NULL;
 	u8 *ucode_addr;
 
-	if (NULL == ucode->fw)
+	if (!ucode->fw)
 		return 0;
 
 	ucode->mc_addr = mc_addr;
@@ -764,7 +855,10 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev,
 	dmcub_hdr = (const struct dmcub_firmware_header_v1_0 *)ucode->fw->data;
 	mes_hdr = (const struct mes_firmware_header_v1_0 *)ucode->fw->data;
 	sdma_hdr = (const struct sdma_firmware_header_v2_0 *)ucode->fw->data;
+	sdmav3_hdr = (const struct sdma_firmware_header_v3_0 *)ucode->fw->data;
 	imu_hdr = (const struct imu_firmware_header_v1_0 *)ucode->fw->data;
+	vpe_hdr = (const struct vpe_firmware_header_v1_0 *)ucode->fw->data;
+	umsch_mm_hdr = (const struct umsch_mm_firmware_header_v1_0 *)ucode->fw->data;
 
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 		switch (ucode->ucode_id) {
@@ -778,6 +872,11 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev,
 			ucode_addr = (u8 *)ucode->fw->data +
 				le32_to_cpu(sdma_hdr->ctl_ucode_offset);
 			break;
+		case AMDGPU_UCODE_ID_SDMA_RS64:
+			ucode->ucode_size = le32_to_cpu(sdmav3_hdr->ucode_size_bytes);
+			ucode_addr = (u8 *)ucode->fw->data +
+				le32_to_cpu(sdmav3_hdr->header.ucode_array_offset_bytes);
+			break;
 		case AMDGPU_UCODE_ID_CP_MEC1:
 		case AMDGPU_UCODE_ID_CP_MEC2:
 			ucode->ucode_size = le32_to_cpu(header->ucode_size_bytes) -
@@ -881,6 +980,10 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev,
 			ucode->ucode_size = ucode->fw->size;
 			ucode_addr = (u8 *)ucode->fw->data;
 			break;
+		case AMDGPU_UCODE_ID_P2S_TABLE:
+			ucode->ucode_size = ucode->fw->size;
+			ucode_addr = (u8 *)ucode->fw->data;
+			break;
 		case AMDGPU_UCODE_ID_IMU_I:
 			ucode->ucode_size = le32_to_cpu(imu_hdr->imu_iram_ucode_size_bytes);
 			ucode_addr = (u8 *)ucode->fw->data +
@@ -947,6 +1050,26 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev,
 			ucode_addr = (u8 *)ucode->fw->data +
 				le32_to_cpu(cpv2_hdr->data_offset_bytes);
 			break;
+		case AMDGPU_UCODE_ID_VPE_CTX:
+			ucode->ucode_size = le32_to_cpu(vpe_hdr->ctx_ucode_size_bytes);
+			ucode_addr = (u8 *)ucode->fw->data +
+				le32_to_cpu(vpe_hdr->header.ucode_array_offset_bytes);
+			break;
+		case AMDGPU_UCODE_ID_VPE_CTL:
+			ucode->ucode_size = le32_to_cpu(vpe_hdr->ctl_ucode_size_bytes);
+			ucode_addr = (u8 *)ucode->fw->data +
+				le32_to_cpu(vpe_hdr->ctl_ucode_offset);
+			break;
+		case AMDGPU_UCODE_ID_UMSCH_MM_UCODE:
+			ucode->ucode_size = le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_size_bytes);
+			ucode_addr = (u8 *)ucode->fw->data +
+				le32_to_cpu(umsch_mm_hdr->header.ucode_array_offset_bytes);
+			break;
+		case AMDGPU_UCODE_ID_UMSCH_MM_DATA:
+			ucode->ucode_size = le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_data_size_bytes);
+			ucode_addr = (u8 *)ucode->fw->data +
+				le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_data_offset_bytes);
+			break;
 		default:
 			ucode->ucode_size = le32_to_cpu(header->ucode_size_bytes);
 			ucode_addr = (u8 *)ucode->fw->data +
@@ -972,7 +1095,7 @@ static int amdgpu_ucode_patch_jt(struct amdgpu_firmware_info *ucode,
 	uint8_t *src_addr = NULL;
 	uint8_t *dst_addr = NULL;
 
-	if (NULL == ucode->fw)
+	if (!ucode->fw)
 		return 0;
 
 	comm_hdr = (const struct common_firmware_header *)ucode->fw->data;
@@ -990,9 +1113,11 @@ static int amdgpu_ucode_patch_jt(struct amdgpu_firmware_info *ucode,
 
 int amdgpu_ucode_create_bo(struct amdgpu_device *adev)
 {
-	if (adev->firmware.load_type != AMDGPU_FW_LOAD_DIRECT) {
+	if ((adev->firmware.load_type != AMDGPU_FW_LOAD_DIRECT) &&
+	    (adev->firmware.load_type != AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO)) {
 		amdgpu_bo_create_kernel(adev, adev->firmware.fw_size, PAGE_SIZE,
-			amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT,
+			(amdgpu_sriov_vf(adev) || adev->debug_use_vram_fw_buf) ?
+			AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT,
 			&adev->firmware.fw_buf,
 			&adev->firmware.fw_buf_mc,
 			&adev->firmware.fw_buf_ptr);
@@ -1035,6 +1160,9 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev)
 		adev->firmware.max_ucodes = AMDGPU_UCODE_ID_MAXIMUM;
 	}
 
+	if (amdgpu_virt_xgmi_migrate_enabled(adev) && adev->firmware.fw_buf)
+		adev->firmware.fw_buf_mc = amdgpu_bo_fb_aper_addr(adev->firmware.fw_buf);
+
 	for (i = 0; i < adev->firmware.max_ucodes; i++) {
 		ucode = &adev->firmware.ucode[i];
 		if (ucode->fw) {
@@ -1043,6 +1171,7 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev)
 			if (i == AMDGPU_UCODE_ID_CP_MEC1 &&
 			    adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
 				const struct gfx_firmware_header_v1_0 *cp_hdr;
+
 				cp_hdr = (const struct gfx_firmware_header_v1_0 *)ucode->fw->data;
 				amdgpu_ucode_patch_jt(ucode,  adev->firmware.fw_buf_mc + fw_offset,
 						    adev->firmware.fw_buf_ptr + fw_offset);
@@ -1057,7 +1186,7 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev)
 static const char *amdgpu_ucode_legacy_naming(struct amdgpu_device *adev, int block_type)
 {
 	if (block_type == MP0_HWIP) {
-		switch (adev->ip_versions[MP0_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 		case IP_VERSION(9, 0, 0):
 			switch (adev->asic_type) {
 			case CHIP_VEGA10:
@@ -1098,6 +1227,7 @@ static const char *amdgpu_ucode_legacy_naming(struct amdgpu_device *adev, int bl
 		case IP_VERSION(11, 0, 13):
 			return "beige_goby";
 		case IP_VERSION(11, 5, 0):
+		case IP_VERSION(11, 5, 2):
 			return "vangogh";
 		case IP_VERSION(12, 0, 1):
 			return "green_sardine";
@@ -1108,7 +1238,7 @@ static const char *amdgpu_ucode_legacy_naming(struct amdgpu_device *adev, int bl
 			return "yellow_carp";
 		}
 	} else if (block_type == MP1_HWIP) {
-		switch (adev->ip_versions[MP1_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
 		case IP_VERSION(9, 0, 0):
 		case IP_VERSION(10, 0, 0):
 		case IP_VERSION(10, 0, 1):
@@ -1134,7 +1264,7 @@ static const char *amdgpu_ucode_legacy_naming(struct amdgpu_device *adev, int bl
 			return "aldebaran_smc";
 		}
 	} else if (block_type == SDMA0_HWIP) {
-		switch (adev->ip_versions[SDMA0_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
 		case IP_VERSION(4, 0, 0):
 			return "vega10_sdma";
 		case IP_VERSION(4, 0, 1):
@@ -1178,7 +1308,7 @@ static const char *amdgpu_ucode_legacy_naming(struct amdgpu_device *adev, int bl
 			return "vangogh_sdma";
 		}
 	} else if (block_type == UVD_HWIP) {
-		switch (adev->ip_versions[UVD_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, UVD_HWIP, 0)) {
 		case IP_VERSION(1, 0, 0):
 		case IP_VERSION(1, 0, 1):
 			if (adev->apu_flags & AMD_APU_IS_RAVEN2)
@@ -1203,7 +1333,8 @@ static const char *amdgpu_ucode_legacy_naming(struct amdgpu_device *adev, int bl
 		case IP_VERSION(3, 0, 0):
 		case IP_VERSION(3, 0, 64):
 		case IP_VERSION(3, 0, 192):
-			if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 0))
+			if (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			    IP_VERSION(10, 3, 0))
 				return "sienna_cichlid_vcn";
 			return "navy_flounder_vcn";
 		case IP_VERSION(3, 0, 2):
@@ -1216,7 +1347,7 @@ static const char *amdgpu_ucode_legacy_naming(struct amdgpu_device *adev, int bl
 			return "yellow_carp_vcn";
 		}
 	} else if (block_type == GC_HWIP) {
-		switch (adev->ip_versions[GC_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 		case IP_VERSION(9, 0, 1):
 			return "vega10";
 		case IP_VERSION(9, 2, 1):
@@ -1264,12 +1395,25 @@ static const char *amdgpu_ucode_legacy_naming(struct amdgpu_device *adev, int bl
 	return NULL;
 }
 
+bool amdgpu_is_kicker_fw(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(kicker_device_list); i++) {
+		if (adev->pdev->device == kicker_device_list[i].device &&
+		    adev->pdev->revision == kicker_device_list[i].revision)
+			return true;
+	}
+
+	return false;
+}
+
 void amdgpu_ucode_ip_version_decode(struct amdgpu_device *adev, int block_type, char *ucode_prefix, int len)
 {
 	int maj, min, rev;
 	char *ip_name;
 	const char *legacy;
-	uint32_t version = adev->ip_versions[block_type][0];
+	uint32_t version = amdgpu_ip_version(adev, block_type, 0);
 
 	legacy = amdgpu_ucode_legacy_naming(adev, block_type);
 	if (legacy) {
@@ -1293,6 +1437,12 @@ void amdgpu_ucode_ip_version_decode(struct amdgpu_device *adev, int block_type,
 	case UVD_HWIP:
 		ip_name = "vcn";
 		break;
+	case VPE_HWIP:
+		ip_name = "vpe";
+		break;
+	case ISP_HWIP:
+		ip_name = "isp";
+		break;
 	default:
 		BUG();
 	}
@@ -1309,24 +1459,49 @@ void amdgpu_ucode_ip_version_decode(struct amdgpu_device *adev, int block_type,
  *
  * @adev: amdgpu device
  * @fw: pointer to load firmware to
- * @fw_name: firmware to load
+ * @required: whether the firmware is required
+ * @fmt: firmware name format string
+ * @...: variable arguments
  *
  * This is a helper that will use request_firmware and amdgpu_ucode_validate
  * to load and run basic validation on firmware. If the load fails, remap
  * the error code to -ENODEV, so that early_init functions will fail to load.
  */
 int amdgpu_ucode_request(struct amdgpu_device *adev, const struct firmware **fw,
-			 const char *fw_name)
+			 enum amdgpu_ucode_required required, const char *fmt, ...)
 {
-	int err = request_firmware(fw, fw_name, adev->dev);
+	char fname[AMDGPU_UCODE_NAME_MAX];
+	va_list ap;
+	int r;
+
+	va_start(ap, fmt);
+	r = vsnprintf(fname, sizeof(fname), fmt, ap);
+	va_end(ap);
+	if (r == sizeof(fname)) {
+		dev_warn(adev->dev, "amdgpu firmware name buffer overflow\n");
+		return -EOVERFLOW;
+	}
 
-	if (err)
+	if (required == AMDGPU_UCODE_REQUIRED)
+		r = request_firmware(fw, fname, adev->dev);
+	else {
+		r = firmware_request_nowarn(fw, fname, adev->dev);
+		if (r)
+			drm_info(&adev->ddev, "Optional firmware \"%s\" was not found\n", fname);
+	}
+	if (r)
 		return -ENODEV;
-	err = amdgpu_ucode_validate(*fw);
-	if (err)
-		dev_dbg(adev->dev, "\"%s\" failed to validate\n", fw_name);
 
-	return err;
+	r = amdgpu_ucode_validate(*fw);
+	if (r)
+		/*
+		 * The amdgpu_ucode_request() should be paired with amdgpu_ucode_release()
+		 * regardless of success/failure, and the amdgpu_ucode_release() takes care of
+		 * firmware release and need to avoid redundant release FW operation here.
+		 */
+		dev_dbg(adev->dev, "\"%s\" failed to validate\n", fname);
+
+	return r;
 }
 
 /*
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h
index b03321e7d2d8..6349aad6da35 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h
@@ -25,6 +25,8 @@
 
 #include "amdgpu_socbb.h"
 
+#define RS64_FW_UC_START_ADDR_LO 0x3000
+
 struct common_firmware_header {
 	uint32_t size_bytes; /* size of the entire header+image(s) in bytes */
 	uint32_t header_size_bytes; /* size of just the header in bytes */
@@ -125,6 +127,8 @@ enum psp_fw_type {
 	PSP_FW_TYPE_PSP_INTF_DRV,
 	PSP_FW_TYPE_PSP_DBG_DRV,
 	PSP_FW_TYPE_PSP_RAS_DRV,
+	PSP_FW_TYPE_PSP_IPKEYMGR_DRV,
+	PSP_FW_TYPE_PSP_SPDM_DRV,
 	PSP_FW_TYPE_MAX_INDEX,
 };
 
@@ -135,6 +139,14 @@ struct psp_firmware_header_v2_0 {
 	struct psp_fw_bin_desc psp_fw_bin[];
 };
 
+/* version_major=2, version_minor=1 */
+struct psp_firmware_header_v2_1 {
+	struct common_firmware_header header;
+	uint32_t psp_fw_bin_count;
+	uint32_t psp_aux_fw_bin_index;
+	struct psp_fw_bin_desc psp_fw_bin[];
+};
+
 /* version_major=1, version_minor=0 */
 struct ta_firmware_header_v1_0 {
 	struct common_firmware_header header;
@@ -154,6 +166,7 @@ enum ta_fw_type {
 	TA_FW_TYPE_PSP_DTM,
 	TA_FW_TYPE_PSP_RAP,
 	TA_FW_TYPE_PSP_SECUREDISPLAY,
+	TA_FW_TYPE_PSP_XGMI_AUX,
 	TA_FW_TYPE_MAX_INDEX,
 };
 
@@ -315,6 +328,44 @@ struct sdma_firmware_header_v2_0 {
 	uint32_t ctl_jt_size; /* control thread size of jt */
 };
 
+/* version_major=1, version_minor=0 */
+struct vpe_firmware_header_v1_0 {
+	struct common_firmware_header header;
+	uint32_t ucode_feature_version;
+	uint32_t ctx_ucode_size_bytes; /* context thread ucode size */
+	uint32_t ctx_jt_offset; /* context thread jt location */
+	uint32_t ctx_jt_size; /* context thread size of jt */
+	uint32_t ctl_ucode_offset;
+	uint32_t ctl_ucode_size_bytes; /* control thread ucode size */
+	uint32_t ctl_jt_offset; /* control thread jt location */
+	uint32_t ctl_jt_size; /* control thread size of jt */
+};
+
+/* version_major=1, version_minor=0 */
+struct umsch_mm_firmware_header_v1_0 {
+	struct common_firmware_header header;
+	uint32_t umsch_mm_ucode_version;
+	uint32_t umsch_mm_ucode_size_bytes;
+	uint32_t umsch_mm_ucode_offset_bytes;
+	uint32_t umsch_mm_ucode_data_version;
+	uint32_t umsch_mm_ucode_data_size_bytes;
+	uint32_t umsch_mm_ucode_data_offset_bytes;
+	uint32_t umsch_mm_irq_start_addr_lo;
+	uint32_t umsch_mm_irq_start_addr_hi;
+	uint32_t umsch_mm_uc_start_addr_lo;
+	uint32_t umsch_mm_uc_start_addr_hi;
+	uint32_t umsch_mm_data_start_addr_lo;
+	uint32_t umsch_mm_data_start_addr_hi;
+};
+
+/* version_major=3, version_minor=0 */
+struct sdma_firmware_header_v3_0 {
+    struct common_firmware_header header;
+    uint32_t ucode_feature_version;
+    uint32_t ucode_offset_bytes;
+    uint32_t ucode_size_bytes;
+};
+
 /* gpu info payload */
 struct gpu_info_firmware_v1_0 {
 	uint32_t gc_num_se;
@@ -387,6 +438,7 @@ union amdgpu_firmware_header {
 	struct psp_firmware_header_v1_1 psp_v1_1;
 	struct psp_firmware_header_v1_3 psp_v1_3;
 	struct psp_firmware_header_v2_0 psp_v2_0;
+	struct psp_firmware_header_v2_0 psp_v2_1;
 	struct ta_firmware_header_v1_0 ta;
 	struct ta_firmware_header_v2_0 ta_v2_0;
 	struct gfx_firmware_header_v1_0 gfx;
@@ -400,6 +452,7 @@ union amdgpu_firmware_header {
 	struct sdma_firmware_header_v1_0 sdma;
 	struct sdma_firmware_header_v1_1 sdma_v1_1;
 	struct sdma_firmware_header_v2_0 sdma_v2_0;
+	struct sdma_firmware_header_v3_0 sdma_v3_0;
 	struct gpu_info_firmware_header_v1_0 gpu_info;
 	struct dmcu_firmware_header_v1_0 dmcu;
 	struct dmcub_firmware_header_v1_0 dmcub;
@@ -407,7 +460,7 @@ union amdgpu_firmware_header {
 	uint8_t raw[0x100];
 };
 
-#define UCODE_MAX_PSP_PACKAGING ((sizeof(union amdgpu_firmware_header) - sizeof(struct common_firmware_header) - 4) / sizeof(struct psp_fw_bin_desc))
+#define UCODE_MAX_PSP_PACKAGING (((sizeof(union amdgpu_firmware_header) - sizeof(struct common_firmware_header) - 4) / sizeof(struct psp_fw_bin_desc)) * 2)
 
 /*
  * fw loading support
@@ -424,6 +477,7 @@ enum AMDGPU_UCODE_ID {
 	AMDGPU_UCODE_ID_SDMA7,
 	AMDGPU_UCODE_ID_SDMA_UCODE_TH0,
 	AMDGPU_UCODE_ID_SDMA_UCODE_TH1,
+	AMDGPU_UCODE_ID_SDMA_RS64,
 	AMDGPU_UCODE_ID_CP_CE,
 	AMDGPU_UCODE_ID_CP_PFP,
 	AMDGPU_UCODE_ID_CP_ME,
@@ -474,6 +528,15 @@ enum AMDGPU_UCODE_ID {
 	AMDGPU_UCODE_ID_VCN0_RAM,
 	AMDGPU_UCODE_ID_VCN1_RAM,
 	AMDGPU_UCODE_ID_DMCUB,
+	AMDGPU_UCODE_ID_VPE_CTX,
+	AMDGPU_UCODE_ID_VPE_CTL,
+	AMDGPU_UCODE_ID_VPE,
+	AMDGPU_UCODE_ID_UMSCH_MM_UCODE,
+	AMDGPU_UCODE_ID_UMSCH_MM_DATA,
+	AMDGPU_UCODE_ID_UMSCH_MM_CMD_BUFFER,
+	AMDGPU_UCODE_ID_P2S_TABLE,
+	AMDGPU_UCODE_ID_JPEG_RAM,
+	AMDGPU_UCODE_ID_ISP,
 	AMDGPU_UCODE_ID_MAXIMUM,
 };
 
@@ -491,6 +554,11 @@ enum amdgpu_firmware_load_type {
 	AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO,
 };
 
+enum amdgpu_ucode_required {
+	AMDGPU_UCODE_OPTIONAL,
+	AMDGPU_UCODE_REQUIRED,
+};
+
 /* conform to smu_ucode_xfer_cz.h */
 #define AMDGPU_SDMA0_UCODE_LOADED	0x00000001
 #define AMDGPU_SDMA1_UCODE_LOADED	0x00000002
@@ -534,6 +602,12 @@ struct amdgpu_firmware {
 
 	void *fw_buf_ptr;
 	uint64_t fw_buf_mc;
+	uint32_t pldm_version;
+};
+
+struct kicker_device{
+	unsigned short device;
+	u8 revision;
 };
 
 void amdgpu_ucode_print_mc_hdr(const struct common_firmware_header *hdr);
@@ -544,8 +618,9 @@ void amdgpu_ucode_print_rlc_hdr(const struct common_firmware_header *hdr);
 void amdgpu_ucode_print_sdma_hdr(const struct common_firmware_header *hdr);
 void amdgpu_ucode_print_psp_hdr(const struct common_firmware_header *hdr);
 void amdgpu_ucode_print_gpu_info_hdr(const struct common_firmware_header *hdr);
+__printf(4, 5)
 int amdgpu_ucode_request(struct amdgpu_device *adev, const struct firmware **fw,
-			 const char *fw_name);
+			 enum amdgpu_ucode_required required, const char *fmt, ...);
 void amdgpu_ucode_release(const struct firmware **fw);
 bool amdgpu_ucode_hdr_version(union amdgpu_firmware_header *hdr,
 				uint16_t hdr_major, uint16_t hdr_minor);
@@ -562,5 +637,6 @@ amdgpu_ucode_get_load_type(struct amdgpu_device *adev, int load_type);
 const char *amdgpu_ucode_name(enum AMDGPU_UCODE_ID ucode_id);
 
 void amdgpu_ucode_ip_version_decode(struct amdgpu_device *adev, int block_type, char *ucode_prefix, int len);
+bool amdgpu_is_kicker_fw(struct amdgpu_device *adev);
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c
index 1edf8e6aeb16..3f0b0e9af4f3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c
@@ -21,14 +21,19 @@
  *
  */
 
+#include <linux/sort.h>
 #include "amdgpu.h"
 #include "umc_v6_7.h"
+#include "amdgpu_ras_mgr.h"
+#define MAX_UMC_POISON_POLLING_TIME_SYNC   20  //ms
+
+#define MAX_UMC_HASH_STRING_SIZE  256
 
 static int amdgpu_umc_convert_error_address(struct amdgpu_device *adev,
 				    struct ras_err_data *err_data, uint64_t err_addr,
 				    uint32_t ch_inst, uint32_t umc_inst)
 {
-	switch (adev->ip_versions[UMC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, UMC_HWIP, 0)) {
 	case IP_VERSION(6, 7, 0):
 		umc_v6_7_convert_error_address(adev,
 				err_data, err_addr, ch_inst, umc_inst);
@@ -45,8 +50,12 @@ static int amdgpu_umc_convert_error_address(struct amdgpu_device *adev,
 int amdgpu_umc_page_retirement_mca(struct amdgpu_device *adev,
 			uint64_t err_addr, uint32_t ch_inst, uint32_t umc_inst)
 {
-	struct ras_err_data err_data = {0, 0, 0, NULL};
-	int ret = AMDGPU_RAS_FAIL;
+	struct ras_err_data err_data;
+	int ret;
+
+	ret = amdgpu_ras_error_data_init(&err_data);
+	if (ret)
+		return ret;
 
 	err_data.err_addr =
 		kcalloc(adev->umc.max_ras_err_cnt_per_query,
@@ -54,145 +63,241 @@ int amdgpu_umc_page_retirement_mca(struct amdgpu_device *adev,
 	if (!err_data.err_addr) {
 		dev_warn(adev->dev,
 			"Failed to alloc memory for umc error record in MCA notifier!\n");
-		return AMDGPU_RAS_FAIL;
+		ret = AMDGPU_RAS_FAIL;
+		goto out_fini_err_data;
 	}
 
+	err_data.err_addr_len = adev->umc.max_ras_err_cnt_per_query;
+
 	/*
 	 * Translate UMC channel address to Physical address
 	 */
 	ret = amdgpu_umc_convert_error_address(adev, &err_data, err_addr,
 					ch_inst, umc_inst);
 	if (ret)
-		goto out;
+		goto out_free_err_addr;
 
 	if (amdgpu_bad_page_threshold != 0) {
 		amdgpu_ras_add_bad_pages(adev, err_data.err_addr,
-						err_data.err_addr_cnt);
+						err_data.err_addr_cnt, false);
 		amdgpu_ras_save_bad_pages(adev, NULL);
 	}
 
-out:
+out_free_err_addr:
 	kfree(err_data.err_addr);
+
+out_fini_err_data:
+	amdgpu_ras_error_data_fini(&err_data);
+
 	return ret;
 }
 
-static int amdgpu_umc_do_page_retirement(struct amdgpu_device *adev,
-		void *ras_error_status,
-		struct amdgpu_iv_entry *entry,
-		bool reset)
+void amdgpu_umc_handle_bad_pages(struct amdgpu_device *adev,
+			void *ras_error_status)
 {
 	struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct amdgpu_ras_eeprom_control *control = &con->eeprom_control;
+	unsigned int error_query_mode;
 	int ret = 0;
+	unsigned long err_count;
 
-	kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
-	ret = amdgpu_dpm_get_ecc_info(adev, (void *)&(con->umc_ecc));
-	if (ret == -EOPNOTSUPP) {
-		if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops &&
-		    adev->umc.ras->ras_block.hw_ops->query_ras_error_count)
-		    adev->umc.ras->ras_block.hw_ops->query_ras_error_count(adev, ras_error_status);
-
-		if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops &&
-		    adev->umc.ras->ras_block.hw_ops->query_ras_error_address &&
-		    adev->umc.max_ras_err_cnt_per_query) {
-			err_data->err_addr =
-				kcalloc(adev->umc.max_ras_err_cnt_per_query,
-					sizeof(struct eeprom_table_record), GFP_KERNEL);
-
-			/* still call query_ras_error_address to clear error status
-			 * even NOMEM error is encountered
-			 */
-			if(!err_data->err_addr)
-				dev_warn(adev->dev, "Failed to alloc memory for "
-						"umc error address record!\n");
+	amdgpu_ras_get_error_query_mode(adev, &error_query_mode);
 
-			/* umc query_ras_error_address is also responsible for clearing
-			 * error status
-			 */
-			adev->umc.ras->ras_block.hw_ops->query_ras_error_address(adev, ras_error_status);
-		}
-	} else if (!ret) {
-		if (adev->umc.ras &&
-		    adev->umc.ras->ecc_info_query_ras_error_count)
-		    adev->umc.ras->ecc_info_query_ras_error_count(adev, ras_error_status);
-
-		if (adev->umc.ras &&
-		    adev->umc.ras->ecc_info_query_ras_error_address &&
-		    adev->umc.max_ras_err_cnt_per_query) {
-			err_data->err_addr =
-				kcalloc(adev->umc.max_ras_err_cnt_per_query,
-					sizeof(struct eeprom_table_record), GFP_KERNEL);
-
-			/* still call query_ras_error_address to clear error status
-			 * even NOMEM error is encountered
-			 */
-			if(!err_data->err_addr)
-				dev_warn(adev->dev, "Failed to alloc memory for "
-						"umc error address record!\n");
+	err_data->err_addr =
+		kcalloc(adev->umc.max_ras_err_cnt_per_query,
+			sizeof(struct eeprom_table_record), GFP_KERNEL);
 
-			/* umc query_ras_error_address is also responsible for clearing
-			 * error status
-			 */
-			adev->umc.ras->ecc_info_query_ras_error_address(adev, ras_error_status);
+	/* still call query_ras_error_address to clear error status
+	 * even NOMEM error is encountered
+	 */
+	if (!err_data->err_addr)
+		dev_warn(adev->dev,
+			"Failed to alloc memory for umc error address record!\n");
+	else
+		err_data->err_addr_len = adev->umc.max_ras_err_cnt_per_query;
+
+	mutex_lock(&con->page_retirement_lock);
+	if (!amdgpu_ras_smu_eeprom_supported(adev)) {
+		ret = amdgpu_dpm_get_ecc_info(adev, (void *)&(con->umc_ecc));
+		if (ret == -EOPNOTSUPP &&
+		    error_query_mode == AMDGPU_RAS_DIRECT_ERROR_QUERY) {
+			if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops &&
+			    adev->umc.ras->ras_block.hw_ops->query_ras_error_count)
+				adev->umc.ras->ras_block.hw_ops->query_ras_error_count(adev,
+								ras_error_status);
+
+			if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops &&
+			    adev->umc.ras->ras_block.hw_ops->query_ras_error_address &&
+			    adev->umc.max_ras_err_cnt_per_query) {
+				err_data->err_addr =
+					kcalloc(adev->umc.max_ras_err_cnt_per_query,
+						sizeof(struct eeprom_table_record), GFP_KERNEL);
+
+				/* still call query_ras_error_address to clear error status
+				 * even NOMEM error is encountered
+				 */
+				if (!err_data->err_addr)
+					dev_warn(adev->dev,
+						"Failed to alloc memory for umc error address record!\n");
+				else
+					err_data->err_addr_len =
+						adev->umc.max_ras_err_cnt_per_query;
+
+				/* umc query_ras_error_address is also responsible for clearing
+				 * error status
+				 */
+				adev->umc.ras->ras_block.hw_ops->query_ras_error_address(adev,
+								ras_error_status);
+			}
+		} else if (error_query_mode == AMDGPU_RAS_FIRMWARE_ERROR_QUERY ||
+		    (!ret && error_query_mode == AMDGPU_RAS_DIRECT_ERROR_QUERY)) {
+			if (adev->umc.ras &&
+			    adev->umc.ras->ecc_info_query_ras_error_count)
+				adev->umc.ras->ecc_info_query_ras_error_count(adev,
+								ras_error_status);
+
+			if (adev->umc.ras &&
+			    adev->umc.ras->ecc_info_query_ras_error_address &&
+			    adev->umc.max_ras_err_cnt_per_query) {
+				err_data->err_addr =
+					kcalloc(adev->umc.max_ras_err_cnt_per_query,
+						sizeof(struct eeprom_table_record), GFP_KERNEL);
+
+				/* still call query_ras_error_address to clear error status
+				 * even NOMEM error is encountered
+				 */
+				if (!err_data->err_addr)
+					dev_warn(adev->dev,
+						"Failed to alloc memory for umc error address record!\n");
+				else
+					err_data->err_addr_len =
+						adev->umc.max_ras_err_cnt_per_query;
+
+				/* umc query_ras_error_address is also responsible for clearing
+				 * error status
+				 */
+				adev->umc.ras->ecc_info_query_ras_error_address(adev,
+								ras_error_status);
+			}
+		}
+	} else {
+		if (!amdgpu_ras_eeprom_update_record_num(control)) {
+			err_data->err_addr_cnt = err_data->de_count =
+				control->ras_num_recs -	control->ras_num_recs_old;
+			amdgpu_ras_eeprom_read_idx(control, err_data->err_addr,
+				control->ras_num_recs_old, err_data->de_count);
 		}
 	}
 
 	/* only uncorrectable error needs gpu reset */
-	if (err_data->ue_count) {
-		dev_info(adev->dev, "%ld uncorrectable hardware errors "
-				"detected in UMC block\n",
-				err_data->ue_count);
-
+	if (err_data->ue_count || err_data->de_count) {
+		err_count = err_data->ue_count + err_data->de_count;
 		if ((amdgpu_bad_page_threshold != 0) &&
 			err_data->err_addr_cnt) {
 			amdgpu_ras_add_bad_pages(adev, err_data->err_addr,
-						err_data->err_addr_cnt);
-			amdgpu_ras_save_bad_pages(adev, &(err_data->ue_count));
+				err_data->err_addr_cnt, amdgpu_ras_smu_eeprom_supported(adev));
+			amdgpu_ras_save_bad_pages(adev, &err_count);
 
-			amdgpu_dpm_send_hbm_bad_pages_num(adev, con->eeprom_control.ras_num_recs);
+			amdgpu_dpm_send_hbm_bad_pages_num(adev,
+					con->eeprom_control.ras_num_bad_pages);
 
 			if (con->update_channel_flag == true) {
 				amdgpu_dpm_send_hbm_bad_channel_flag(adev, con->eeprom_control.bad_channel_bitmap);
 				con->update_channel_flag = false;
 			}
 		}
-
-		if (reset)
-			amdgpu_ras_reset_gpu(adev);
 	}
 
 	kfree(err_data->err_addr);
+	err_data->err_addr = NULL;
+
+	mutex_unlock(&con->page_retirement_lock);
+}
+
+static int amdgpu_umc_do_page_retirement(struct amdgpu_device *adev,
+		void *ras_error_status,
+		struct amdgpu_iv_entry *entry,
+		uint32_t reset)
+{
+	struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+
+	kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
+	amdgpu_umc_handle_bad_pages(adev, ras_error_status);
+
+	if ((err_data->ue_count || err_data->de_count) &&
+	    (reset || amdgpu_ras_is_rma(adev))) {
+		con->gpu_reset_flags |= reset;
+		amdgpu_ras_reset_gpu(adev);
+	}
+
 	return AMDGPU_RAS_SUCCESS;
 }
 
-int amdgpu_umc_poison_handler(struct amdgpu_device *adev, bool reset)
+int amdgpu_umc_pasid_poison_handler(struct amdgpu_device *adev,
+			enum amdgpu_ras_block block, uint16_t pasid,
+			pasid_notify pasid_fn, void *data, uint32_t reset)
 {
 	int ret = AMDGPU_RAS_SUCCESS;
 
+	if (adev->gmc.xgmi.connected_to_cpu ||
+		adev->gmc.is_app_apu) {
+		if (reset) {
+			/* MCA poison handler is only responsible for GPU reset,
+			 * let MCA notifier do page retirement.
+			 */
+			kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
+			amdgpu_ras_reset_gpu(adev);
+		}
+		return ret;
+	}
+
 	if (!amdgpu_sriov_vf(adev)) {
-		if (!adev->gmc.xgmi.connected_to_cpu) {
-			struct ras_err_data err_data = {0, 0, 0, NULL};
+		if (amdgpu_ip_version(adev, UMC_HWIP, 0) < IP_VERSION(12, 0, 0)) {
+			struct ras_err_data err_data;
 			struct ras_common_if head = {
 				.block = AMDGPU_RAS_BLOCK__UMC,
 			};
 			struct ras_manager *obj = amdgpu_ras_find_obj(adev, &head);
 
+			ret = amdgpu_ras_error_data_init(&err_data);
+			if (ret)
+				return ret;
+
 			ret = amdgpu_umc_do_page_retirement(adev, &err_data, NULL, reset);
 
 			if (ret == AMDGPU_RAS_SUCCESS && obj) {
 				obj->err_data.ue_count += err_data.ue_count;
 				obj->err_data.ce_count += err_data.ce_count;
+				obj->err_data.de_count += err_data.de_count;
+			}
+
+			amdgpu_ras_error_data_fini(&err_data);
+		} else if (amdgpu_uniras_enabled(adev)) {
+			struct ras_ih_info ih_info = {0};
+
+			ih_info.block = block;
+			ih_info.pasid = pasid;
+			ih_info.reset = reset;
+			ih_info.pasid_fn = pasid_fn;
+			ih_info.data = data;
+			amdgpu_ras_mgr_handle_consumer_interrupt(adev, &ih_info);
+		} else {
+			struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+			int ret;
+
+			ret = amdgpu_ras_put_poison_req(adev,
+				block, pasid, pasid_fn, data, reset);
+			if (!ret) {
+				atomic_inc(&con->page_retirement_req_cnt);
+				atomic_inc(&con->poison_consumption_count);
+				wake_up(&con->page_retirement_wq);
 			}
-		} else if (reset) {
-			/* MCA poison handler is only responsible for GPU reset,
-			 * let MCA notifier do page retirement.
-			 */
-			kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
-			amdgpu_ras_reset_gpu(adev);
 		}
 	} else {
 		if (adev->virt.ops && adev->virt.ops->ras_poison_handler)
-			adev->virt.ops->ras_poison_handler(adev);
+			adev->virt.ops->ras_poison_handler(adev, block);
 		else
 			dev_warn(adev->dev,
 				"No ras_poison_handler interface in SRIOV!\n");
@@ -201,11 +306,19 @@ int amdgpu_umc_poison_handler(struct amdgpu_device *adev, bool reset)
 	return ret;
 }
 
+int amdgpu_umc_poison_handler(struct amdgpu_device *adev,
+			enum amdgpu_ras_block block, uint32_t reset)
+{
+	return amdgpu_umc_pasid_poison_handler(adev,
+				block, 0, NULL, NULL, reset);
+}
+
 int amdgpu_umc_process_ras_data_cb(struct amdgpu_device *adev,
 		void *ras_error_status,
 		struct amdgpu_iv_entry *entry)
 {
-	return amdgpu_umc_do_page_retirement(adev, ras_error_status, entry, true);
+	return amdgpu_umc_do_page_retirement(adev, ras_error_status, entry,
+				AMDGPU_RAS_GPU_RESET_MODE1_RESET);
 }
 
 int amdgpu_umc_ras_sw_init(struct amdgpu_device *adev)
@@ -246,6 +359,9 @@ int amdgpu_umc_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *r
 	if (r)
 		return r;
 
+	if (amdgpu_sriov_vf(adev))
+		return r;
+
 	if (amdgpu_ras_is_supported(adev, ras_block->block)) {
 		r = amdgpu_irq_get(adev, &adev->gmc.ecc_irq, 0);
 		if (r)
@@ -282,14 +398,20 @@ int amdgpu_umc_process_ecc_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-void amdgpu_umc_fill_error_record(struct ras_err_data *err_data,
+int amdgpu_umc_fill_error_record(struct ras_err_data *err_data,
 		uint64_t err_addr,
 		uint64_t retired_page,
 		uint32_t channel_index,
 		uint32_t umc_inst)
 {
-	struct eeprom_table_record *err_rec =
-		&err_data->err_addr[err_data->err_addr_cnt];
+	struct eeprom_table_record *err_rec;
+
+	if (!err_data ||
+	    !err_data->err_addr ||
+	    (err_data->err_addr_cnt >= err_data->err_addr_len))
+		return -EINVAL;
+
+	err_rec = &err_data->err_addr[err_data->err_addr_cnt];
 
 	err_rec->address = err_addr;
 	/* page frame address is saved */
@@ -301,6 +423,47 @@ void amdgpu_umc_fill_error_record(struct ras_err_data *err_data,
 	err_rec->mcumc_id = umc_inst;
 
 	err_data->err_addr_cnt++;
+
+	return 0;
+}
+
+static int amdgpu_umc_loop_all_aid(struct amdgpu_device *adev, umc_func func,
+				   void *data)
+{
+	uint32_t umc_node_inst;
+	uint32_t node_inst;
+	uint32_t umc_inst;
+	uint32_t ch_inst;
+	int ret;
+
+	/*
+	 * This loop is done based on the following -
+	 * umc.active mask = mask of active umc instances across all nodes
+	 * umc.umc_inst_num = maximum number of umc instancess per node
+	 * umc.node_inst_num = maximum number of node instances
+	 * Channel instances are not assumed to be harvested.
+	 */
+	dev_dbg(adev->dev, "active umcs :%lx umc_inst per node: %d",
+		adev->umc.active_mask, adev->umc.umc_inst_num);
+	for_each_set_bit(umc_node_inst, &(adev->umc.active_mask),
+			 adev->umc.node_inst_num * adev->umc.umc_inst_num) {
+		node_inst = umc_node_inst / adev->umc.umc_inst_num;
+		umc_inst = umc_node_inst % adev->umc.umc_inst_num;
+		LOOP_UMC_CH_INST(ch_inst) {
+			dev_dbg(adev->dev,
+				"node_inst :%d umc_inst: %d ch_inst: %d",
+				node_inst, umc_inst, ch_inst);
+			ret = func(adev, node_inst, umc_inst, ch_inst, data);
+			if (ret) {
+				dev_err(adev->dev,
+					"Node %d umc %d ch %d func returns %d\n",
+					node_inst, umc_inst, ch_inst, ret);
+				return ret;
+			}
+		}
+	}
+
+	return 0;
 }
 
 int amdgpu_umc_loop_channels(struct amdgpu_device *adev,
@@ -311,6 +474,9 @@ int amdgpu_umc_loop_channels(struct amdgpu_device *adev,
 	uint32_t ch_inst         = 0;
 	int ret = 0;
 
+	if (adev->aid_mask)
+		return amdgpu_umc_loop_all_aid(adev, func, data);
+
 	if (adev->umc.node_inst_num) {
 		LOOP_UMC_EACH_NODE_INST_AND_CH(node_inst, umc_inst, ch_inst) {
 			ret = func(adev, node_inst, umc_inst, ch_inst, data);
@@ -333,3 +499,129 @@ int amdgpu_umc_loop_channels(struct amdgpu_device *adev,
 
 	return 0;
 }
+
+int amdgpu_umc_update_ecc_status(struct amdgpu_device *adev,
+				uint64_t status, uint64_t ipid, uint64_t addr)
+{
+	if (adev->umc.ras->update_ecc_status)
+		return adev->umc.ras->update_ecc_status(adev,
+					status, ipid, addr);
+	return 0;
+}
+
+int amdgpu_umc_logs_ecc_err(struct amdgpu_device *adev,
+		struct radix_tree_root *ecc_tree, struct ras_ecc_err *ecc_err)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct ras_ecc_log_info *ecc_log;
+	int ret;
+
+	ecc_log = &con->umc_ecc_log;
+
+	mutex_lock(&ecc_log->lock);
+	ret = radix_tree_insert(ecc_tree, ecc_err->pa_pfn, ecc_err);
+	if (!ret)
+		radix_tree_tag_set(ecc_tree,
+			ecc_err->pa_pfn, UMC_ECC_NEW_DETECTED_TAG);
+	mutex_unlock(&ecc_log->lock);
+
+	return ret;
+}
+
+int amdgpu_umc_pages_in_a_row(struct amdgpu_device *adev,
+			struct ras_err_data *err_data, uint64_t pa_addr)
+{
+	struct ta_ras_query_address_output addr_out;
+
+	/* reinit err_data */
+	err_data->err_addr_cnt = 0;
+	err_data->err_addr_len = adev->umc.retire_unit;
+
+	addr_out.pa.pa = pa_addr;
+	if (adev->umc.ras && adev->umc.ras->convert_ras_err_addr)
+		return adev->umc.ras->convert_ras_err_addr(adev, err_data, NULL,
+				&addr_out, false);
+	else
+		return -EINVAL;
+}
+
+int amdgpu_umc_lookup_bad_pages_in_a_row(struct amdgpu_device *adev,
+			uint64_t pa_addr, uint64_t *pfns, int len)
+{
+	int i, ret;
+	struct ras_err_data err_data;
+
+	err_data.err_addr = kcalloc(adev->umc.retire_unit,
+				sizeof(struct eeprom_table_record), GFP_KERNEL);
+	if (!err_data.err_addr) {
+		dev_warn(adev->dev, "Failed to alloc memory in bad page lookup!\n");
+		return 0;
+	}
+
+	ret = amdgpu_umc_pages_in_a_row(adev, &err_data, pa_addr);
+	if (ret)
+		goto out;
+
+	for (i = 0; i < adev->umc.retire_unit; i++) {
+		if (i >= len)
+			goto out;
+
+		pfns[i] = err_data.err_addr[i].retired_page;
+	}
+	ret = i;
+	adev->umc.err_addr_cnt = err_data.err_addr_cnt;
+
+out:
+	kfree(err_data.err_addr);
+	return ret;
+}
+
+int amdgpu_umc_mca_to_addr(struct amdgpu_device *adev,
+			uint64_t err_addr, uint32_t ch, uint32_t umc,
+			uint32_t node, uint32_t socket,
+			struct ta_ras_query_address_output *addr_out, bool dump_addr)
+{
+	struct ta_ras_query_address_input addr_in;
+	int ret;
+
+	memset(&addr_in, 0, sizeof(addr_in));
+	addr_in.ma.err_addr = err_addr;
+	addr_in.ma.ch_inst = ch;
+	addr_in.ma.umc_inst = umc;
+	addr_in.ma.node_inst = node;
+	addr_in.ma.socket_id = socket;
+
+	if (adev->umc.ras && adev->umc.ras->convert_ras_err_addr) {
+		ret = adev->umc.ras->convert_ras_err_addr(adev, NULL, &addr_in,
+				addr_out, dump_addr);
+		if (ret)
+			return ret;
+	} else {
+		return 0;
+	}
+
+	return 0;
+}
+
+int amdgpu_umc_pa2mca(struct amdgpu_device *adev,
+		uint64_t pa, uint64_t *mca, enum amdgpu_memory_partition nps)
+{
+	struct ta_ras_query_address_input addr_in;
+	struct ta_ras_query_address_output addr_out;
+	int ret;
+
+	/* nps: the pa belongs to */
+	addr_in.pa.pa = pa | ((uint64_t)nps << 58);
+	addr_in.addr_type = TA_RAS_PA_TO_MCA;
+	ret = psp_ras_query_address(&adev->psp, &addr_in, &addr_out);
+	if (ret) {
+		dev_warn(adev->dev, "Failed to query RAS MCA address for 0x%llx",
+			pa);
+
+		return ret;
+	}
+
+	*mca = addr_out.ma.err_addr;
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h
index 86133f77a9a4..28dff750c47e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h
@@ -21,7 +21,7 @@
 #ifndef __AMDGPU_UMC_H__
 #define __AMDGPU_UMC_H__
 #include "amdgpu_ras.h"
-
+#include "amdgpu_mca.h"
 /*
  * (addr / 256) * 4096, the higher 26 bits in ErrorAddr
  * is the index of 4KB block
@@ -32,6 +32,11 @@
  * is the index of 8KB block
  */
 #define ADDR_OF_8KB_BLOCK(addr)			(((addr) & ~0xffULL) << 5)
+/*
+ * (addr / 256) * 32768, the higher 26 bits in ErrorAddr
+ * is the index of 8KB block
+ */
+#define ADDR_OF_32KB_BLOCK(addr)			(((addr) & ~0xffULL) << 7)
 /* channel index is the index of 256B block */
 #define ADDR_OF_256B_BLOCK(channel_index)	((channel_index) << 8)
 /* offset in 256B block */
@@ -47,6 +52,43 @@
 #define LOOP_UMC_EACH_NODE_INST_AND_CH(node_inst, umc_inst, ch_inst) \
 		LOOP_UMC_NODE_INST((node_inst)) LOOP_UMC_INST_AND_CH((umc_inst), (ch_inst))
 
+/* Page retirement tag */
+#define UMC_ECC_NEW_DETECTED_TAG       0x1
+/*
+ * a flag to indicate v2 of channel index stored in eeprom
+ *
+ * v1 (legacy way): store channel index within a umc instance in eeprom
+ *    range in UMC v12: 0 ~ 7
+ * v2: store global channel index in eeprom
+ *    range in UMC v12: 0 ~ 127
+ *
+ * NOTE: it's better to store it in eeprom_table_record.mem_channel,
+ * but there is only 8 bits in mem_channel, and the channel number may
+ * increase in the future, we decide to save it in
+ * eeprom_table_record.retired_page. retired_page is useless in v2,
+ * we depend on eeprom_table_record.address instead of retired_page in v2.
+ * Only 48 bits are saved on eeprom, use bit 47 here.
+ */
+#define UMC_CHANNEL_IDX_V2	BIT_ULL(47)
+
+/*
+ * save nps value to eeprom_table_record.retired_page[47:40],
+ * the channel index flag above will be retired.
+ */
+#define UMC_NPS_SHIFT 40
+#define UMC_NPS_MASK 0xffULL
+
+/* three column bits and one row bit in MCA address flip
+ * in bad page retirement
+ */
+#define RETIRE_FLIP_BITS_NUM 4
+
+struct amdgpu_umc_flip_bits {
+	uint32_t flip_bits_in_pa[RETIRE_FLIP_BITS_NUM];
+	uint32_t flip_row_bit;
+	uint32_t r13_in_pa;
+	uint32_t bit_num;
+};
 
 typedef int (*umc_func)(struct amdgpu_device *adev, uint32_t node_inst,
 			uint32_t umc_inst, uint32_t ch_inst, void *data);
@@ -59,6 +101,20 @@ struct amdgpu_umc_ras {
 				      void *ras_error_status);
 	void (*ecc_info_query_ras_error_address)(struct amdgpu_device *adev,
 					void *ras_error_status);
+	bool (*check_ecc_err_status)(struct amdgpu_device *adev,
+			enum amdgpu_mca_error_type type, void *ras_error_status);
+	int (*update_ecc_status)(struct amdgpu_device *adev,
+			uint64_t status, uint64_t ipid, uint64_t addr);
+	int (*convert_ras_err_addr)(struct amdgpu_device *adev,
+			struct ras_err_data *err_data,
+			struct ta_ras_query_address_input *addr_in,
+			struct ta_ras_query_address_output *addr_out,
+			bool dump_addr);
+	uint32_t (*get_die_id_from_pa)(struct amdgpu_device *adev,
+			uint64_t mca_addr, uint64_t retired_page);
+	void (*get_retire_flip_bits)(struct amdgpu_device *adev);
+	void (*mca_ipid_parse)(struct amdgpu_device *adev, uint64_t ipid,
+			uint32_t *did, uint32_t *ch, uint32_t *umc_inst, uint32_t *sid);
 };
 
 struct amdgpu_umc_funcs {
@@ -89,15 +145,23 @@ struct amdgpu_umc {
 
 	/* active mask for umc node instance */
 	unsigned long active_mask;
+
+	struct amdgpu_umc_flip_bits flip_bits;
+
+	unsigned long err_addr_cnt;
 };
 
 int amdgpu_umc_ras_sw_init(struct amdgpu_device *adev);
 int amdgpu_umc_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block);
-int amdgpu_umc_poison_handler(struct amdgpu_device *adev, bool reset);
+int amdgpu_umc_poison_handler(struct amdgpu_device *adev,
+			enum amdgpu_ras_block block, uint32_t reset);
+int amdgpu_umc_pasid_poison_handler(struct amdgpu_device *adev,
+			enum amdgpu_ras_block block, uint16_t pasid,
+			pasid_notify pasid_fn, void *data, uint32_t reset);
 int amdgpu_umc_process_ecc_irq(struct amdgpu_device *adev,
 		struct amdgpu_irq_src *source,
 		struct amdgpu_iv_entry *entry);
-void amdgpu_umc_fill_error_record(struct ras_err_data *err_data,
+int amdgpu_umc_fill_error_record(struct ras_err_data *err_data,
 		uint64_t err_addr,
 		uint64_t retired_page,
 		uint32_t channel_index,
@@ -111,4 +175,22 @@ int amdgpu_umc_page_retirement_mca(struct amdgpu_device *adev,
 
 int amdgpu_umc_loop_channels(struct amdgpu_device *adev,
 			umc_func func, void *data);
+
+int amdgpu_umc_update_ecc_status(struct amdgpu_device *adev,
+				uint64_t status, uint64_t ipid, uint64_t addr);
+int amdgpu_umc_logs_ecc_err(struct amdgpu_device *adev,
+		struct radix_tree_root *ecc_tree, struct ras_ecc_err *ecc_err);
+
+void amdgpu_umc_handle_bad_pages(struct amdgpu_device *adev,
+			void *ras_error_status);
+int amdgpu_umc_pages_in_a_row(struct amdgpu_device *adev,
+			struct ras_err_data *err_data, uint64_t pa_addr);
+int amdgpu_umc_lookup_bad_pages_in_a_row(struct amdgpu_device *adev,
+			uint64_t pa_addr, uint64_t *pfns, int len);
+int amdgpu_umc_mca_to_addr(struct amdgpu_device *adev,
+			uint64_t err_addr, uint32_t ch, uint32_t umc,
+			uint32_t node, uint32_t socket,
+			struct ta_ras_query_address_output *addr_out, bool dump_addr);
+int amdgpu_umc_pa2mca(struct amdgpu_device *adev,
+		uint64_t pa, uint64_t *mca, enum amdgpu_memory_partition nps);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umr.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_umr.h
index 919d9d401750..5b27fc41ffbf 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umr.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umr.h
@@ -35,17 +35,51 @@ struct amdgpu_debugfs_regs2_iocdata {
 	} srbm;
 };
 
+struct amdgpu_debugfs_regs2_iocdata_v2 {
+	__u32 use_srbm, use_grbm, pg_lock;
+	struct {
+		__u32 se, sh, instance;
+	} grbm;
+	struct {
+		__u32 me, pipe, queue, vmid;
+	} srbm;
+	u32 xcc_id;
+};
+
+struct amdgpu_debugfs_gprwave_iocdata {
+	u32 gpr_or_wave, se, sh, cu, wave, simd, xcc_id;
+	struct {
+		u32 thread, vpgr_or_sgpr;
+	} gpr;
+};
+
 /*
  * MMIO debugfs state data (per file* handle)
  */
 struct amdgpu_debugfs_regs2_data {
 	struct amdgpu_device *adev;
 	struct mutex lock;
-	struct amdgpu_debugfs_regs2_iocdata id;
+	struct amdgpu_debugfs_regs2_iocdata_v2 id;
+};
+
+struct amdgpu_debugfs_gprwave_data {
+	struct amdgpu_device *adev;
+	struct mutex lock;
+	struct amdgpu_debugfs_gprwave_iocdata id;
 };
 
 enum AMDGPU_DEBUGFS_REGS2_CMDS {
-	AMDGPU_DEBUGFS_REGS2_CMD_SET_STATE=0,
+	AMDGPU_DEBUGFS_REGS2_CMD_SET_STATE = 0,
+	AMDGPU_DEBUGFS_REGS2_CMD_SET_STATE_V2,
+};
+
+enum AMDGPU_DEBUGFS_GPRWAVE_CMDS {
+	AMDGPU_DEBUGFS_GPRWAVE_CMD_SET_STATE = 0,
 };
 
+//reg2 interface
 #define AMDGPU_DEBUGFS_REGS2_IOC_SET_STATE _IOWR(0x20, AMDGPU_DEBUGFS_REGS2_CMD_SET_STATE, struct amdgpu_debugfs_regs2_iocdata)
+#define AMDGPU_DEBUGFS_REGS2_IOC_SET_STATE_V2 _IOWR(0x20, AMDGPU_DEBUGFS_REGS2_CMD_SET_STATE_V2, struct amdgpu_debugfs_regs2_iocdata_v2)
+
+//gprwave interface
+#define AMDGPU_DEBUGFS_GPRWAVE_IOC_SET_STATE _IOWR(0x20, AMDGPU_DEBUGFS_GPRWAVE_CMD_SET_STATE, struct amdgpu_debugfs_gprwave_iocdata)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c
new file mode 100644
index 000000000000..cd707d70a0bf
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c
@@ -0,0 +1,550 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/firmware.h>
+#include <linux/module.h>
+#include <linux/debugfs.h>
+#include <drm/drm_exec.h>
+#include <drm/drm_drv.h>
+
+#include "amdgpu.h"
+#include "amdgpu_umsch_mm.h"
+#include "umsch_mm_v4_0.h"
+
+MODULE_FIRMWARE("amdgpu/umsch_mm_4_0_0.bin");
+
+int amdgpu_umsch_mm_submit_pkt(struct amdgpu_umsch_mm *umsch, void *pkt, int ndws)
+{
+	struct amdgpu_ring *ring = &umsch->ring;
+
+	if (amdgpu_ring_alloc(ring, ndws))
+		return -ENOMEM;
+
+	amdgpu_ring_write_multiple(ring, pkt, ndws);
+	amdgpu_ring_commit(ring);
+
+	return 0;
+}
+
+int amdgpu_umsch_mm_query_fence(struct amdgpu_umsch_mm *umsch)
+{
+	struct amdgpu_ring *ring = &umsch->ring;
+	struct amdgpu_device *adev = ring->adev;
+	int r;
+
+	r = amdgpu_fence_wait_polling(ring, ring->fence_drv.sync_seq, adev->usec_timeout);
+	if (r < 1) {
+		dev_err(adev->dev, "ring umsch timeout, emitted fence %u\n",
+			ring->fence_drv.sync_seq);
+		return -ETIMEDOUT;
+	}
+
+	return 0;
+}
+
+static void umsch_mm_ring_set_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_umsch_mm *umsch = (struct amdgpu_umsch_mm *)ring;
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->use_doorbell)
+		WDOORBELL32(ring->doorbell_index, ring->wptr << 2);
+	else
+		WREG32(umsch->rb_wptr, ring->wptr << 2);
+}
+
+static u64 umsch_mm_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_umsch_mm *umsch = (struct amdgpu_umsch_mm *)ring;
+	struct amdgpu_device *adev = ring->adev;
+
+	return RREG32(umsch->rb_rptr);
+}
+
+static u64 umsch_mm_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_umsch_mm *umsch = (struct amdgpu_umsch_mm *)ring;
+	struct amdgpu_device *adev = ring->adev;
+
+	return RREG32(umsch->rb_wptr);
+}
+
+static const struct amdgpu_ring_funcs umsch_v4_0_ring_funcs = {
+	.type = AMDGPU_RING_TYPE_UMSCH_MM,
+	.align_mask = 0,
+	.nop = 0,
+	.support_64bit_ptrs = false,
+	.get_rptr = umsch_mm_ring_get_rptr,
+	.get_wptr = umsch_mm_ring_get_wptr,
+	.set_wptr = umsch_mm_ring_set_wptr,
+	.insert_nop = amdgpu_ring_insert_nop,
+};
+
+int amdgpu_umsch_mm_ring_init(struct amdgpu_umsch_mm *umsch)
+{
+	struct amdgpu_device *adev = container_of(umsch, struct amdgpu_device, umsch_mm);
+	struct amdgpu_ring *ring = &umsch->ring;
+
+	ring->vm_hub = AMDGPU_MMHUB0(0);
+	ring->use_doorbell = true;
+	ring->no_scheduler = true;
+	ring->doorbell_index = (AMDGPU_NAVI10_DOORBELL64_VCN0_1 << 1) + 6;
+
+	snprintf(ring->name, sizeof(ring->name), "umsch");
+
+	return amdgpu_ring_init(adev, ring, 1024, NULL, 0, AMDGPU_RING_PRIO_DEFAULT, NULL);
+}
+
+int amdgpu_umsch_mm_init_microcode(struct amdgpu_umsch_mm *umsch)
+{
+	const struct umsch_mm_firmware_header_v1_0 *umsch_mm_hdr;
+	struct amdgpu_device *adev = umsch->ring.adev;
+	const char *fw_name = NULL;
+	int r;
+
+	switch (amdgpu_ip_version(adev, VCN_HWIP, 0)) {
+	case IP_VERSION(4, 0, 5):
+	case IP_VERSION(4, 0, 6):
+		fw_name = "4_0_0";
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	r = amdgpu_ucode_request(adev, &adev->umsch_mm.fw, AMDGPU_UCODE_REQUIRED,
+				 "amdgpu/umsch_mm_%s.bin", fw_name);
+	if (r) {
+		release_firmware(adev->umsch_mm.fw);
+		adev->umsch_mm.fw = NULL;
+		return r;
+	}
+
+	umsch_mm_hdr = (const struct umsch_mm_firmware_header_v1_0 *)adev->umsch_mm.fw->data;
+
+	adev->umsch_mm.ucode_size = le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_size_bytes);
+	adev->umsch_mm.data_size = le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_data_size_bytes);
+
+	adev->umsch_mm.irq_start_addr =
+		le32_to_cpu(umsch_mm_hdr->umsch_mm_irq_start_addr_lo) |
+		((uint64_t)(le32_to_cpu(umsch_mm_hdr->umsch_mm_irq_start_addr_hi)) << 32);
+	adev->umsch_mm.uc_start_addr =
+		le32_to_cpu(umsch_mm_hdr->umsch_mm_uc_start_addr_lo) |
+		((uint64_t)(le32_to_cpu(umsch_mm_hdr->umsch_mm_uc_start_addr_hi)) << 32);
+	adev->umsch_mm.data_start_addr =
+		le32_to_cpu(umsch_mm_hdr->umsch_mm_data_start_addr_lo) |
+		((uint64_t)(le32_to_cpu(umsch_mm_hdr->umsch_mm_data_start_addr_hi)) << 32);
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		struct amdgpu_firmware_info *info;
+
+		info = &adev->firmware.ucode[AMDGPU_UCODE_ID_UMSCH_MM_UCODE];
+		info->ucode_id = AMDGPU_UCODE_ID_UMSCH_MM_UCODE;
+		info->fw = adev->umsch_mm.fw;
+		adev->firmware.fw_size +=
+			ALIGN(le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_size_bytes), PAGE_SIZE);
+
+		info = &adev->firmware.ucode[AMDGPU_UCODE_ID_UMSCH_MM_DATA];
+		info->ucode_id = AMDGPU_UCODE_ID_UMSCH_MM_DATA;
+		info->fw = adev->umsch_mm.fw;
+		adev->firmware.fw_size +=
+			ALIGN(le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_data_size_bytes), PAGE_SIZE);
+	}
+
+	return 0;
+}
+
+int amdgpu_umsch_mm_allocate_ucode_buffer(struct amdgpu_umsch_mm *umsch)
+{
+	const struct umsch_mm_firmware_header_v1_0 *umsch_mm_hdr;
+	struct amdgpu_device *adev = umsch->ring.adev;
+	const __le32 *fw_data;
+	uint32_t fw_size;
+	int r;
+
+	umsch_mm_hdr = (const struct umsch_mm_firmware_header_v1_0 *)
+		       adev->umsch_mm.fw->data;
+
+	fw_data = (const __le32 *)(adev->umsch_mm.fw->data +
+		  le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_offset_bytes));
+	fw_size = le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_size_bytes);
+
+	r = amdgpu_bo_create_reserved(adev, fw_size,
+				      4 * 1024, AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->umsch_mm.ucode_fw_obj,
+				      &adev->umsch_mm.ucode_fw_gpu_addr,
+				      (void **)&adev->umsch_mm.ucode_fw_ptr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create umsch_mm fw ucode bo\n", r);
+		return r;
+	}
+
+	memcpy(adev->umsch_mm.ucode_fw_ptr, fw_data, fw_size);
+
+	amdgpu_bo_kunmap(adev->umsch_mm.ucode_fw_obj);
+	amdgpu_bo_unreserve(adev->umsch_mm.ucode_fw_obj);
+	return 0;
+}
+
+int amdgpu_umsch_mm_allocate_ucode_data_buffer(struct amdgpu_umsch_mm *umsch)
+{
+	const struct umsch_mm_firmware_header_v1_0 *umsch_mm_hdr;
+	struct amdgpu_device *adev = umsch->ring.adev;
+	const __le32 *fw_data;
+	uint32_t fw_size;
+	int r;
+
+	umsch_mm_hdr = (const struct umsch_mm_firmware_header_v1_0 *)
+		       adev->umsch_mm.fw->data;
+
+	fw_data = (const __le32 *)(adev->umsch_mm.fw->data +
+		  le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_data_offset_bytes));
+	fw_size = le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_data_size_bytes);
+
+	r = amdgpu_bo_create_reserved(adev, fw_size,
+				      64 * 1024, AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->umsch_mm.data_fw_obj,
+				      &adev->umsch_mm.data_fw_gpu_addr,
+				      (void **)&adev->umsch_mm.data_fw_ptr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create umsch_mm fw data bo\n", r);
+		return r;
+	}
+
+	memcpy(adev->umsch_mm.data_fw_ptr, fw_data, fw_size);
+
+	amdgpu_bo_kunmap(adev->umsch_mm.data_fw_obj);
+	amdgpu_bo_unreserve(adev->umsch_mm.data_fw_obj);
+	return 0;
+}
+
+int amdgpu_umsch_mm_psp_execute_cmd_buf(struct amdgpu_umsch_mm *umsch)
+{
+	struct amdgpu_device *adev = umsch->ring.adev;
+	struct amdgpu_firmware_info ucode = {
+		.ucode_id = AMDGPU_UCODE_ID_UMSCH_MM_CMD_BUFFER,
+		.mc_addr = adev->umsch_mm.cmd_buf_gpu_addr,
+		.ucode_size = ((uintptr_t)adev->umsch_mm.cmd_buf_curr_ptr -
+			      (uintptr_t)adev->umsch_mm.cmd_buf_ptr),
+	};
+
+	return psp_execute_ip_fw_load(&adev->psp, &ucode);
+}
+
+static void umsch_mm_agdb_index_init(struct amdgpu_device *adev)
+{
+	uint32_t umsch_mm_agdb_start;
+	int i;
+
+	umsch_mm_agdb_start = adev->doorbell_index.max_assignment + 1;
+	umsch_mm_agdb_start = roundup(umsch_mm_agdb_start, 1024);
+	umsch_mm_agdb_start += (AMDGPU_NAVI10_DOORBELL64_VCN0_1 << 1);
+
+	for (i = 0; i < CONTEXT_PRIORITY_NUM_LEVELS; i++)
+		adev->umsch_mm.agdb_index[i] = umsch_mm_agdb_start + i;
+}
+
+static int umsch_mm_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	adev->umsch_mm.vmid_mask_mm_vpe = 0xf00;
+	adev->umsch_mm.engine_mask = (1 << UMSCH_SWIP_ENGINE_TYPE_VPE);
+	adev->umsch_mm.vpe_hqd_mask = 0xfe;
+
+	r = amdgpu_device_wb_get(adev, &adev->umsch_mm.wb_index);
+	if (r) {
+		dev_err(adev->dev, "failed to alloc wb for umsch: %d\n", r);
+		return r;
+	}
+
+	adev->umsch_mm.sch_ctx_gpu_addr = adev->wb.gpu_addr +
+					  (adev->umsch_mm.wb_index * 4);
+
+	r = amdgpu_bo_create_kernel(adev, PAGE_SIZE, PAGE_SIZE,
+				    AMDGPU_GEM_DOMAIN_GTT,
+				    &adev->umsch_mm.cmd_buf_obj,
+				    &adev->umsch_mm.cmd_buf_gpu_addr,
+				    (void **)&adev->umsch_mm.cmd_buf_ptr);
+	if (r) {
+		dev_err(adev->dev, "failed to allocate cmdbuf bo %d\n", r);
+		amdgpu_device_wb_free(adev, adev->umsch_mm.wb_index);
+		return r;
+	}
+
+	r = amdgpu_bo_create_kernel(adev, AMDGPU_UMSCHFW_LOG_SIZE, PAGE_SIZE,
+				    AMDGPU_GEM_DOMAIN_VRAM |
+				    AMDGPU_GEM_DOMAIN_GTT,
+				    &adev->umsch_mm.dbglog_bo,
+				    &adev->umsch_mm.log_gpu_addr,
+				    &adev->umsch_mm.log_cpu_addr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to allocate umsch debug bo\n", r);
+		return r;
+	}
+
+	mutex_init(&adev->umsch_mm.mutex_hidden);
+
+	umsch_mm_agdb_index_init(adev);
+
+	return 0;
+}
+
+
+static int umsch_mm_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	switch (amdgpu_ip_version(adev, VCN_HWIP, 0)) {
+	case IP_VERSION(4, 0, 5):
+	case IP_VERSION(4, 0, 6):
+		umsch_mm_v4_0_set_funcs(&adev->umsch_mm);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	adev->umsch_mm.ring.funcs = &umsch_v4_0_ring_funcs;
+	umsch_mm_set_regs(&adev->umsch_mm);
+
+	return 0;
+}
+
+static int umsch_mm_late_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (amdgpu_in_reset(adev) || adev->in_s0ix || adev->in_suspend)
+		return 0;
+
+	return 0;
+}
+
+static int umsch_mm_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = umsch_mm_init(adev);
+	if (r)
+		return r;
+
+	amdgpu_umsch_fwlog_init(&adev->umsch_mm);
+	r = umsch_mm_ring_init(&adev->umsch_mm);
+	if (r)
+		return r;
+
+	r = umsch_mm_init_microcode(&adev->umsch_mm);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static int umsch_mm_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	release_firmware(adev->umsch_mm.fw);
+	adev->umsch_mm.fw = NULL;
+
+	amdgpu_ring_fini(&adev->umsch_mm.ring);
+
+	mutex_destroy(&adev->umsch_mm.mutex_hidden);
+
+	amdgpu_bo_free_kernel(&adev->umsch_mm.cmd_buf_obj,
+			      &adev->umsch_mm.cmd_buf_gpu_addr,
+			      (void **)&adev->umsch_mm.cmd_buf_ptr);
+
+	amdgpu_bo_free_kernel(&adev->umsch_mm.dbglog_bo,
+				    &adev->umsch_mm.log_gpu_addr,
+				    (void **)&adev->umsch_mm.log_cpu_addr);
+
+	amdgpu_device_wb_free(adev, adev->umsch_mm.wb_index);
+
+	return 0;
+}
+
+static int umsch_mm_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = umsch_mm_load_microcode(&adev->umsch_mm);
+	if (r)
+		return r;
+
+	umsch_mm_ring_start(&adev->umsch_mm);
+
+	r = umsch_mm_set_hw_resources(&adev->umsch_mm);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static int umsch_mm_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	umsch_mm_ring_stop(&adev->umsch_mm);
+
+	amdgpu_bo_free_kernel(&adev->umsch_mm.data_fw_obj,
+			      &adev->umsch_mm.data_fw_gpu_addr,
+			      (void **)&adev->umsch_mm.data_fw_ptr);
+
+	amdgpu_bo_free_kernel(&adev->umsch_mm.ucode_fw_obj,
+			      &adev->umsch_mm.ucode_fw_gpu_addr,
+			      (void **)&adev->umsch_mm.ucode_fw_ptr);
+	return 0;
+}
+
+static int umsch_mm_suspend(struct amdgpu_ip_block *ip_block)
+{
+	return umsch_mm_hw_fini(ip_block);
+}
+
+static int umsch_mm_resume(struct amdgpu_ip_block *ip_block)
+{
+	return umsch_mm_hw_init(ip_block);
+}
+
+void amdgpu_umsch_fwlog_init(struct amdgpu_umsch_mm *umsch_mm)
+{
+#if defined(CONFIG_DEBUG_FS)
+	void *fw_log_cpu_addr = umsch_mm->log_cpu_addr;
+	volatile struct amdgpu_umsch_fwlog *log_buf = fw_log_cpu_addr;
+
+	log_buf->header_size = sizeof(struct amdgpu_umsch_fwlog);
+	log_buf->buffer_size = AMDGPU_UMSCHFW_LOG_SIZE;
+	log_buf->rptr = log_buf->header_size;
+	log_buf->wptr = log_buf->header_size;
+	log_buf->wrapped = 0;
+#endif
+}
+
+/*
+ * debugfs for mapping umsch firmware log buffer.
+ */
+#if defined(CONFIG_DEBUG_FS)
+static ssize_t amdgpu_debugfs_umsch_fwlog_read(struct file *f, char __user *buf,
+					     size_t size, loff_t *pos)
+{
+	struct amdgpu_umsch_mm *umsch_mm;
+	void *log_buf;
+	volatile struct amdgpu_umsch_fwlog *plog;
+	unsigned int read_pos, write_pos, available, i, read_bytes = 0;
+	unsigned int read_num[2] = {0};
+
+	umsch_mm = file_inode(f)->i_private;
+	if (!umsch_mm)
+		return -ENODEV;
+
+	if (!umsch_mm->log_cpu_addr)
+		return -EFAULT;
+
+	log_buf = umsch_mm->log_cpu_addr;
+
+	plog = (volatile struct amdgpu_umsch_fwlog *)log_buf;
+	read_pos = plog->rptr;
+	write_pos = plog->wptr;
+
+	if (read_pos > AMDGPU_UMSCHFW_LOG_SIZE || write_pos > AMDGPU_UMSCHFW_LOG_SIZE)
+		return -EFAULT;
+
+	if (!size || (read_pos == write_pos))
+		return 0;
+
+	if (write_pos > read_pos) {
+		available = write_pos - read_pos;
+		read_num[0] = min_t(size_t, size, available);
+	} else {
+		read_num[0] = AMDGPU_UMSCHFW_LOG_SIZE - read_pos;
+		available = read_num[0] + write_pos - plog->header_size;
+		if (size > available)
+			read_num[1] = write_pos - plog->header_size;
+		else if (size > read_num[0])
+			read_num[1] = size - read_num[0];
+		else
+			read_num[0] = size;
+	}
+
+	for (i = 0; i < 2; i++) {
+		if (read_num[i]) {
+			if (read_pos == AMDGPU_UMSCHFW_LOG_SIZE)
+				read_pos = plog->header_size;
+			if (read_num[i] == copy_to_user((buf + read_bytes),
+							(log_buf + read_pos), read_num[i]))
+				return -EFAULT;
+
+			read_bytes += read_num[i];
+			read_pos += read_num[i];
+		}
+	}
+
+	plog->rptr = read_pos;
+	*pos += read_bytes;
+	return read_bytes;
+}
+
+static const struct file_operations amdgpu_debugfs_umschfwlog_fops = {
+	.owner = THIS_MODULE,
+	.read = amdgpu_debugfs_umsch_fwlog_read,
+	.llseek = default_llseek
+};
+#endif
+
+void amdgpu_debugfs_umsch_fwlog_init(struct amdgpu_device *adev,
+			struct amdgpu_umsch_mm *umsch_mm)
+{
+#if defined(CONFIG_DEBUG_FS)
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+	struct dentry *root = minor->debugfs_root;
+	char name[32];
+
+	sprintf(name, "amdgpu_umsch_fwlog");
+	debugfs_create_file_size(name, S_IFREG | 0444, root, umsch_mm,
+				 &amdgpu_debugfs_umschfwlog_fops,
+				 AMDGPU_UMSCHFW_LOG_SIZE);
+#endif
+}
+
+static const struct amd_ip_funcs umsch_mm_v4_0_ip_funcs = {
+	.name = "umsch_mm_v4_0",
+	.early_init = umsch_mm_early_init,
+	.late_init = umsch_mm_late_init,
+	.sw_init = umsch_mm_sw_init,
+	.sw_fini = umsch_mm_sw_fini,
+	.hw_init = umsch_mm_hw_init,
+	.hw_fini = umsch_mm_hw_fini,
+	.suspend = umsch_mm_suspend,
+	.resume = umsch_mm_resume,
+};
+
+const struct amdgpu_ip_block_version umsch_mm_v4_0_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_UMSCH_MM,
+	.major = 4,
+	.minor = 0,
+	.rev = 0,
+	.funcs = &umsch_mm_v4_0_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.h
new file mode 100644
index 000000000000..2c771a753778
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.h
@@ -0,0 +1,246 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __AMDGPU_UMSCH_MM_H__
+#define __AMDGPU_UMSCH_MM_H__
+
+enum UMSCH_SWIP_ENGINE_TYPE {
+	UMSCH_SWIP_ENGINE_TYPE_VCN0 = 0,
+	UMSCH_SWIP_ENGINE_TYPE_VCN1 = 1,
+	UMSCH_SWIP_ENGINE_TYPE_VCN = 2,
+	UMSCH_SWIP_ENGINE_TYPE_VPE = 3,
+	UMSCH_SWIP_ENGINE_TYPE_MAX
+};
+
+enum UMSCH_CONTEXT_PRIORITY_LEVEL {
+	CONTEXT_PRIORITY_LEVEL_IDLE = 0,
+	CONTEXT_PRIORITY_LEVEL_NORMAL = 1,
+	CONTEXT_PRIORITY_LEVEL_FOCUS = 2,
+	CONTEXT_PRIORITY_LEVEL_REALTIME = 3,
+	CONTEXT_PRIORITY_NUM_LEVELS
+};
+
+struct umsch_mm_set_resource_input {
+	uint32_t vmid_mask_mm_vcn;
+	uint32_t vmid_mask_mm_vpe;
+	uint32_t collaboration_mask_vpe;
+	uint32_t logging_vmid;
+	uint32_t engine_mask;
+	union {
+		struct {
+			uint32_t disable_reset : 1;
+			uint32_t disable_umsch_mm_log : 1;
+			uint32_t use_rs64mem_for_proc_ctx_csa : 1;
+			uint32_t reserved : 29;
+		};
+		uint32_t uint32_all;
+	};
+};
+
+struct amdgpu_umsch_fwlog {
+	uint32_t rptr;
+	uint32_t wptr;
+	uint32_t buffer_size;
+	uint32_t header_size;
+	uint32_t wrapped;
+};
+
+struct umsch_mm_add_queue_input {
+	uint32_t process_id;
+	uint64_t page_table_base_addr;
+	uint64_t process_va_start;
+	uint64_t process_va_end;
+	uint64_t process_quantum;
+	uint64_t process_csa_addr;
+	uint64_t context_quantum;
+	uint64_t context_csa_addr;
+	uint32_t inprocess_context_priority;
+	enum UMSCH_CONTEXT_PRIORITY_LEVEL context_global_priority_level;
+	uint32_t doorbell_offset_0;
+	uint32_t doorbell_offset_1;
+	enum UMSCH_SWIP_ENGINE_TYPE engine_type;
+	uint32_t affinity;
+	uint64_t mqd_addr;
+	uint64_t h_context;
+	uint64_t h_queue;
+	uint32_t vm_context_cntl;
+
+	uint32_t process_csa_array_index;
+	uint32_t context_csa_array_index;
+
+	struct {
+		uint32_t is_context_suspended : 1;
+		uint32_t collaboration_mode : 1;
+		uint32_t reserved : 30;
+	};
+};
+
+struct umsch_mm_remove_queue_input {
+	uint32_t doorbell_offset_0;
+	uint32_t doorbell_offset_1;
+	uint64_t context_csa_addr;
+	uint32_t context_csa_array_index;
+};
+
+struct MQD_INFO {
+	uint32_t rb_base_hi;
+	uint32_t rb_base_lo;
+	uint32_t rb_size;
+	uint32_t wptr_val;
+	uint32_t rptr_val;
+	uint32_t unmapped;
+	uint32_t vmid;
+};
+
+struct amdgpu_umsch_mm;
+
+struct umsch_mm_funcs {
+	int (*set_hw_resources)(struct amdgpu_umsch_mm *umsch);
+	int (*add_queue)(struct amdgpu_umsch_mm *umsch,
+			 struct umsch_mm_add_queue_input *input);
+	int (*remove_queue)(struct amdgpu_umsch_mm *umsch,
+			    struct umsch_mm_remove_queue_input *input);
+	int (*set_regs)(struct amdgpu_umsch_mm *umsch);
+	int (*init_microcode)(struct amdgpu_umsch_mm *umsch);
+	int (*load_microcode)(struct amdgpu_umsch_mm *umsch);
+	int (*ring_init)(struct amdgpu_umsch_mm *umsch);
+	int (*ring_start)(struct amdgpu_umsch_mm *umsch);
+	int (*ring_stop)(struct amdgpu_umsch_mm *umsch);
+	int (*ring_fini)(struct amdgpu_umsch_mm *umsch);
+};
+
+struct amdgpu_umsch_mm {
+	struct amdgpu_ring		ring;
+
+	uint32_t			rb_wptr;
+	uint32_t			rb_rptr;
+
+	const struct umsch_mm_funcs	*funcs;
+
+	const struct firmware		*fw;
+	uint32_t			fw_version;
+	uint32_t			feature_version;
+
+	struct amdgpu_bo		*ucode_fw_obj;
+	uint64_t			ucode_fw_gpu_addr;
+	uint32_t			*ucode_fw_ptr;
+	uint64_t			irq_start_addr;
+	uint64_t			uc_start_addr;
+	uint32_t			ucode_size;
+
+	struct amdgpu_bo		*data_fw_obj;
+	uint64_t			data_fw_gpu_addr;
+	uint32_t			*data_fw_ptr;
+	uint64_t			data_start_addr;
+	uint32_t			data_size;
+
+	struct amdgpu_bo		*cmd_buf_obj;
+	uint64_t			cmd_buf_gpu_addr;
+	uint32_t			*cmd_buf_ptr;
+	uint32_t			*cmd_buf_curr_ptr;
+
+	uint32_t			wb_index;
+	uint64_t			sch_ctx_gpu_addr;
+	uint32_t			*sch_ctx_cpu_addr;
+
+	uint32_t			vmid_mask_mm_vcn;
+	uint32_t			vmid_mask_mm_vpe;
+	uint32_t			engine_mask;
+	uint32_t			vcn0_hqd_mask;
+	uint32_t			vcn1_hqd_mask;
+	uint32_t			vcn_hqd_mask[2];
+	uint32_t			vpe_hqd_mask;
+	uint32_t			agdb_index[CONTEXT_PRIORITY_NUM_LEVELS];
+
+	struct mutex			mutex_hidden;
+	struct amdgpu_bo		*dbglog_bo;
+	void				*log_cpu_addr;
+	uint64_t			log_gpu_addr;
+	uint32_t			mem_size;
+	uint32_t			log_offset;
+};
+
+int amdgpu_umsch_mm_submit_pkt(struct amdgpu_umsch_mm *umsch, void *pkt, int ndws);
+int amdgpu_umsch_mm_query_fence(struct amdgpu_umsch_mm *umsch);
+
+int amdgpu_umsch_mm_init_microcode(struct amdgpu_umsch_mm *umsch);
+int amdgpu_umsch_mm_allocate_ucode_buffer(struct amdgpu_umsch_mm *umsch);
+int amdgpu_umsch_mm_allocate_ucode_data_buffer(struct amdgpu_umsch_mm *umsch);
+
+int amdgpu_umsch_mm_psp_execute_cmd_buf(struct amdgpu_umsch_mm *umsch);
+
+int amdgpu_umsch_mm_ring_init(struct amdgpu_umsch_mm *umsch);
+
+void amdgpu_debugfs_umsch_fwlog_init(struct amdgpu_device *adev,
+			struct amdgpu_umsch_mm *umsch);
+
+void amdgpu_umsch_fwlog_init(struct amdgpu_umsch_mm *umsch_mm);
+
+#define WREG32_SOC15_UMSCH(reg, value)								\
+	do {											\
+		uint32_t reg_offset = adev->reg_offset[VCN_HWIP][0][reg##_BASE_IDX] + reg;	\
+		if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {				\
+			*adev->umsch_mm.cmd_buf_curr_ptr++ = (reg_offset << 2);			\
+			*adev->umsch_mm.cmd_buf_curr_ptr++ = value;				\
+		} else	{									\
+			WREG32(reg_offset, value);						\
+		}										\
+	} while (0)
+
+#define umsch_mm_set_hw_resources(umsch) \
+	((umsch)->funcs->set_hw_resources ? (umsch)->funcs->set_hw_resources((umsch)) : 0)
+#define umsch_mm_add_queue(umsch, input) \
+	((umsch)->funcs->add_queue ? (umsch)->funcs->add_queue((umsch), (input)) : 0)
+#define umsch_mm_remove_queue(umsch, input) \
+	((umsch)->funcs->remove_queue ? (umsch)->funcs->remove_queue((umsch), (input)) : 0)
+
+#define umsch_mm_set_regs(umsch) \
+	((umsch)->funcs->set_regs ? (umsch)->funcs->set_regs((umsch)) : 0)
+#define umsch_mm_init_microcode(umsch) \
+	((umsch)->funcs->init_microcode ? (umsch)->funcs->init_microcode((umsch)) : 0)
+#define umsch_mm_load_microcode(umsch) \
+	((umsch)->funcs->load_microcode ? (umsch)->funcs->load_microcode((umsch)) : 0)
+
+#define umsch_mm_ring_init(umsch) \
+	((umsch)->funcs->ring_init ? (umsch)->funcs->ring_init((umsch)) : 0)
+#define umsch_mm_ring_start(umsch) \
+	((umsch)->funcs->ring_start ? (umsch)->funcs->ring_start((umsch)) : 0)
+#define umsch_mm_ring_stop(umsch) \
+	((umsch)->funcs->ring_stop ? (umsch)->funcs->ring_stop((umsch)) : 0)
+#define umsch_mm_ring_fini(umsch) \
+	((umsch)->funcs->ring_fini ? (umsch)->funcs->ring_fini((umsch)) : 0)
+
+static inline void amdgpu_umsch_mm_lock(struct amdgpu_umsch_mm *umsch)
+{
+	mutex_lock(&umsch->mutex_hidden);
+}
+
+static inline void amdgpu_umsch_mm_unlock(struct amdgpu_umsch_mm *umsch)
+{
+	mutex_unlock(&umsch->mutex_hidden);
+}
+
+extern const struct amdgpu_ip_block_version umsch_mm_v4_0_ip_block;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
new file mode 100644
index 000000000000..9a969175900e
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
@@ -0,0 +1,1482 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <drm/drm_auth.h>
+#include <drm/drm_exec.h>
+#include <linux/pm_runtime.h>
+#include <drm/drm_drv.h>
+
+#include "amdgpu.h"
+#include "amdgpu_reset.h"
+#include "amdgpu_vm.h"
+#include "amdgpu_userq.h"
+#include "amdgpu_hmm.h"
+#include "amdgpu_userq_fence.h"
+
+u32 amdgpu_userq_get_supported_ip_mask(struct amdgpu_device *adev)
+{
+	int i;
+	u32 userq_ip_mask = 0;
+
+	for (i = 0; i < AMDGPU_HW_IP_NUM; i++) {
+		if (adev->userq_funcs[i])
+			userq_ip_mask |= (1 << i);
+	}
+
+	return userq_ip_mask;
+}
+
+static bool amdgpu_userq_is_reset_type_supported(struct amdgpu_device *adev,
+				enum amdgpu_ring_type ring_type, int reset_type)
+{
+
+	if (ring_type < 0 || ring_type >= AMDGPU_RING_TYPE_MAX)
+		return false;
+
+	switch (ring_type) {
+	case AMDGPU_RING_TYPE_GFX:
+		if (adev->gfx.gfx_supported_reset & reset_type)
+			return true;
+		break;
+	case AMDGPU_RING_TYPE_COMPUTE:
+		if (adev->gfx.compute_supported_reset & reset_type)
+			return true;
+		break;
+	case AMDGPU_RING_TYPE_SDMA:
+		if (adev->sdma.supported_reset & reset_type)
+			return true;
+		break;
+	case AMDGPU_RING_TYPE_VCN_DEC:
+	case AMDGPU_RING_TYPE_VCN_ENC:
+		if (adev->vcn.supported_reset & reset_type)
+			return true;
+		break;
+	case AMDGPU_RING_TYPE_VCN_JPEG:
+		if (adev->jpeg.supported_reset & reset_type)
+			return true;
+		break;
+	default:
+		break;
+	}
+	return false;
+}
+
+static void amdgpu_userq_gpu_reset(struct amdgpu_device *adev)
+{
+	if (amdgpu_device_should_recover_gpu(adev)) {
+		amdgpu_reset_domain_schedule(adev->reset_domain,
+					     &adev->userq_reset_work);
+		/* Wait for the reset job to complete */
+		flush_work(&adev->userq_reset_work);
+	}
+}
+
+static int
+amdgpu_userq_detect_and_reset_queues(struct amdgpu_userq_mgr *uq_mgr)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	const int queue_types[] = {
+		AMDGPU_RING_TYPE_COMPUTE,
+		AMDGPU_RING_TYPE_GFX,
+		AMDGPU_RING_TYPE_SDMA
+	};
+	const int num_queue_types = ARRAY_SIZE(queue_types);
+	bool gpu_reset = false;
+	int r = 0;
+	int i;
+
+	/* Warning if current process mutex is not held */
+	WARN_ON(!mutex_is_locked(&uq_mgr->userq_mutex));
+
+	if (unlikely(adev->debug_disable_gpu_ring_reset)) {
+		dev_err(adev->dev, "userq reset disabled by debug mask\n");
+		return 0;
+	}
+
+	/*
+	 * If GPU recovery feature is disabled system-wide,
+	 * skip all reset detection logic
+	 */
+	if (!amdgpu_gpu_recovery)
+		return 0;
+
+	/*
+	 * Iterate through all queue types to detect and reset problematic queues
+	 * Process each queue type in the defined order
+	 */
+	for (i = 0; i < num_queue_types; i++) {
+		int ring_type = queue_types[i];
+		const struct amdgpu_userq_funcs *funcs = adev->userq_funcs[ring_type];
+
+		if (!amdgpu_userq_is_reset_type_supported(adev, ring_type, AMDGPU_RESET_TYPE_PER_QUEUE))
+				continue;
+
+		if (atomic_read(&uq_mgr->userq_count[ring_type]) > 0 &&
+		    funcs && funcs->detect_and_reset) {
+			r = funcs->detect_and_reset(adev, ring_type);
+			if (r) {
+				gpu_reset = true;
+				break;
+			}
+		}
+	}
+
+	if (gpu_reset)
+		amdgpu_userq_gpu_reset(adev);
+
+	return r;
+}
+
+static int amdgpu_userq_buffer_va_list_add(struct amdgpu_usermode_queue *queue,
+					   struct amdgpu_bo_va_mapping *va_map, u64 addr)
+{
+	struct amdgpu_userq_va_cursor *va_cursor;
+	struct userq_va_list;
+
+	va_cursor = kzalloc(sizeof(*va_cursor), GFP_KERNEL);
+	if (!va_cursor)
+		return -ENOMEM;
+
+	INIT_LIST_HEAD(&va_cursor->list);
+	va_cursor->gpu_addr = addr;
+	atomic_set(&va_map->bo_va->userq_va_mapped, 1);
+	list_add(&va_cursor->list, &queue->userq_va_list);
+
+	return 0;
+}
+
+int amdgpu_userq_input_va_validate(struct amdgpu_usermode_queue *queue,
+				   u64 addr, u64 expected_size)
+{
+	struct amdgpu_bo_va_mapping *va_map;
+	struct amdgpu_vm *vm = queue->vm;
+	u64 user_addr;
+	u64 size;
+	int r = 0;
+
+	user_addr = (addr & AMDGPU_GMC_HOLE_MASK) >> AMDGPU_GPU_PAGE_SHIFT;
+	size = expected_size >> AMDGPU_GPU_PAGE_SHIFT;
+
+	r = amdgpu_bo_reserve(vm->root.bo, false);
+	if (r)
+		return r;
+
+	va_map = amdgpu_vm_bo_lookup_mapping(vm, user_addr);
+	if (!va_map) {
+		r = -EINVAL;
+		goto out_err;
+	}
+	/* Only validate the userq whether resident in the VM mapping range */
+	if (user_addr >= va_map->start  &&
+	    va_map->last - user_addr + 1 >= size) {
+		amdgpu_userq_buffer_va_list_add(queue, va_map, user_addr);
+		amdgpu_bo_unreserve(vm->root.bo);
+		return 0;
+	}
+
+	r = -EINVAL;
+out_err:
+	amdgpu_bo_unreserve(vm->root.bo);
+	return r;
+}
+
+static bool amdgpu_userq_buffer_va_mapped(struct amdgpu_vm *vm, u64 addr)
+{
+	struct amdgpu_bo_va_mapping *mapping;
+	bool r;
+
+	if (amdgpu_bo_reserve(vm->root.bo, false))
+		return false;
+
+	mapping = amdgpu_vm_bo_lookup_mapping(vm, addr);
+	if (!IS_ERR_OR_NULL(mapping) && atomic_read(&mapping->bo_va->userq_va_mapped))
+		r = true;
+	else
+		r = false;
+	amdgpu_bo_unreserve(vm->root.bo);
+
+	return r;
+}
+
+static bool amdgpu_userq_buffer_vas_mapped(struct amdgpu_usermode_queue *queue)
+{
+	struct amdgpu_userq_va_cursor *va_cursor, *tmp;
+	int r = 0;
+
+	list_for_each_entry_safe(va_cursor, tmp, &queue->userq_va_list, list) {
+		r += amdgpu_userq_buffer_va_mapped(queue->vm, va_cursor->gpu_addr);
+		dev_dbg(queue->userq_mgr->adev->dev,
+			"validate the userq mapping:%p va:%llx r:%d\n",
+			queue, va_cursor->gpu_addr, r);
+	}
+
+	if (r != 0)
+		return true;
+
+	return false;
+}
+
+static void amdgpu_userq_buffer_va_list_del(struct amdgpu_bo_va_mapping *mapping,
+					    struct amdgpu_userq_va_cursor *va_cursor)
+{
+	atomic_set(&mapping->bo_va->userq_va_mapped, 0);
+	list_del(&va_cursor->list);
+	kfree(va_cursor);
+}
+
+static int amdgpu_userq_buffer_vas_list_cleanup(struct amdgpu_device *adev,
+						struct amdgpu_usermode_queue *queue)
+{
+	struct amdgpu_userq_va_cursor *va_cursor, *tmp;
+	struct amdgpu_bo_va_mapping *mapping;
+	int r;
+
+	r = amdgpu_bo_reserve(queue->vm->root.bo, false);
+	if (r)
+		return r;
+
+	list_for_each_entry_safe(va_cursor, tmp, &queue->userq_va_list, list) {
+		mapping = amdgpu_vm_bo_lookup_mapping(queue->vm, va_cursor->gpu_addr);
+		if (!mapping) {
+			r = -EINVAL;
+			goto err;
+		}
+		dev_dbg(adev->dev, "delete the userq:%p va:%llx\n",
+			queue, va_cursor->gpu_addr);
+		amdgpu_userq_buffer_va_list_del(mapping, va_cursor);
+	}
+err:
+	amdgpu_bo_unreserve(queue->vm->root.bo);
+	return r;
+}
+
+static int
+amdgpu_userq_preempt_helper(struct amdgpu_userq_mgr *uq_mgr,
+			  struct amdgpu_usermode_queue *queue)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	const struct amdgpu_userq_funcs *userq_funcs =
+		adev->userq_funcs[queue->queue_type];
+	bool found_hung_queue = false;
+	int r = 0;
+
+	if (queue->state == AMDGPU_USERQ_STATE_MAPPED) {
+		r = userq_funcs->preempt(uq_mgr, queue);
+		if (r) {
+			queue->state = AMDGPU_USERQ_STATE_HUNG;
+			found_hung_queue = true;
+		} else {
+			queue->state = AMDGPU_USERQ_STATE_PREEMPTED;
+		}
+	}
+
+	if (found_hung_queue)
+		amdgpu_userq_detect_and_reset_queues(uq_mgr);
+
+	return r;
+}
+
+static int
+amdgpu_userq_restore_helper(struct amdgpu_userq_mgr *uq_mgr,
+			struct amdgpu_usermode_queue *queue)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	const struct amdgpu_userq_funcs *userq_funcs =
+		adev->userq_funcs[queue->queue_type];
+	int r = 0;
+
+	if (queue->state == AMDGPU_USERQ_STATE_PREEMPTED) {
+		r = userq_funcs->restore(uq_mgr, queue);
+		if (r) {
+			queue->state = AMDGPU_USERQ_STATE_HUNG;
+		} else {
+			queue->state = AMDGPU_USERQ_STATE_MAPPED;
+		}
+	}
+
+	return r;
+}
+
+static int
+amdgpu_userq_unmap_helper(struct amdgpu_userq_mgr *uq_mgr,
+			  struct amdgpu_usermode_queue *queue)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	const struct amdgpu_userq_funcs *userq_funcs =
+		adev->userq_funcs[queue->queue_type];
+	bool found_hung_queue = false;
+	int r = 0;
+
+	if ((queue->state == AMDGPU_USERQ_STATE_MAPPED) ||
+		(queue->state == AMDGPU_USERQ_STATE_PREEMPTED)) {
+		r = userq_funcs->unmap(uq_mgr, queue);
+		if (r) {
+			queue->state = AMDGPU_USERQ_STATE_HUNG;
+			found_hung_queue = true;
+		} else {
+			queue->state = AMDGPU_USERQ_STATE_UNMAPPED;
+		}
+	}
+
+	if (found_hung_queue)
+		amdgpu_userq_detect_and_reset_queues(uq_mgr);
+
+	return r;
+}
+
+static int
+amdgpu_userq_map_helper(struct amdgpu_userq_mgr *uq_mgr,
+			struct amdgpu_usermode_queue *queue)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	const struct amdgpu_userq_funcs *userq_funcs =
+		adev->userq_funcs[queue->queue_type];
+	int r = 0;
+
+	if (queue->state == AMDGPU_USERQ_STATE_UNMAPPED) {
+		r = userq_funcs->map(uq_mgr, queue);
+		if (r) {
+			queue->state = AMDGPU_USERQ_STATE_HUNG;
+			amdgpu_userq_detect_and_reset_queues(uq_mgr);
+		} else {
+			queue->state = AMDGPU_USERQ_STATE_MAPPED;
+		}
+	}
+
+	return r;
+}
+
+static int
+amdgpu_userq_wait_for_last_fence(struct amdgpu_userq_mgr *uq_mgr,
+				 struct amdgpu_usermode_queue *queue)
+{
+	struct dma_fence *f = queue->last_fence;
+	int ret = 0;
+
+	if (f && !dma_fence_is_signaled(f)) {
+		ret = dma_fence_wait_timeout(f, true, MAX_SCHEDULE_TIMEOUT);
+		if (ret <= 0) {
+			drm_file_err(uq_mgr->file, "Timed out waiting for fence=%llu:%llu\n",
+				     f->context, f->seqno);
+			queue->state = AMDGPU_USERQ_STATE_HUNG;
+			return -ETIME;
+		}
+	}
+
+	return ret;
+}
+
+static void
+amdgpu_userq_cleanup(struct amdgpu_userq_mgr *uq_mgr,
+		     struct amdgpu_usermode_queue *queue,
+		     int queue_id)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	const struct amdgpu_userq_funcs *uq_funcs = adev->userq_funcs[queue->queue_type];
+
+	/* Wait for mode-1 reset to complete */
+	down_read(&adev->reset_domain->sem);
+
+	/* Drop the userq reference. */
+	amdgpu_userq_buffer_vas_list_cleanup(adev, queue);
+	uq_funcs->mqd_destroy(uq_mgr, queue);
+	amdgpu_userq_fence_driver_free(queue);
+	/* Use interrupt-safe locking since IRQ handlers may access these XArrays */
+	xa_erase_irq(&uq_mgr->userq_mgr_xa, (unsigned long)queue_id);
+	xa_erase_irq(&adev->userq_doorbell_xa, queue->doorbell_index);
+	queue->userq_mgr = NULL;
+	list_del(&queue->userq_va_list);
+	kfree(queue);
+
+	up_read(&adev->reset_domain->sem);
+}
+
+static struct amdgpu_usermode_queue *
+amdgpu_userq_find(struct amdgpu_userq_mgr *uq_mgr, int qid)
+{
+	return xa_load(&uq_mgr->userq_mgr_xa, qid);
+}
+
+void
+amdgpu_userq_ensure_ev_fence(struct amdgpu_userq_mgr *uq_mgr,
+			     struct amdgpu_eviction_fence_mgr *evf_mgr)
+{
+	struct amdgpu_eviction_fence *ev_fence;
+
+retry:
+	/* Flush any pending resume work to create ev_fence */
+	flush_delayed_work(&uq_mgr->resume_work);
+
+	mutex_lock(&uq_mgr->userq_mutex);
+	spin_lock(&evf_mgr->ev_fence_lock);
+	ev_fence = evf_mgr->ev_fence;
+	spin_unlock(&evf_mgr->ev_fence_lock);
+	if (!ev_fence || dma_fence_is_signaled(&ev_fence->base)) {
+		mutex_unlock(&uq_mgr->userq_mutex);
+		/*
+		 * Looks like there was no pending resume work,
+		 * add one now to create a valid eviction fence
+		 */
+		schedule_delayed_work(&uq_mgr->resume_work, 0);
+		goto retry;
+	}
+}
+
+int amdgpu_userq_create_object(struct amdgpu_userq_mgr *uq_mgr,
+			       struct amdgpu_userq_obj *userq_obj,
+			       int size)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	struct amdgpu_bo_param bp;
+	int r;
+
+	memset(&bp, 0, sizeof(bp));
+	bp.byte_align = PAGE_SIZE;
+	bp.domain = AMDGPU_GEM_DOMAIN_GTT;
+	bp.flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS |
+		   AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
+	bp.type = ttm_bo_type_kernel;
+	bp.size = size;
+	bp.resv = NULL;
+	bp.bo_ptr_size = sizeof(struct amdgpu_bo);
+
+	r = amdgpu_bo_create(adev, &bp, &userq_obj->obj);
+	if (r) {
+		drm_file_err(uq_mgr->file, "Failed to allocate BO for userqueue (%d)", r);
+		return r;
+	}
+
+	r = amdgpu_bo_reserve(userq_obj->obj, true);
+	if (r) {
+		drm_file_err(uq_mgr->file, "Failed to reserve BO to map (%d)", r);
+		goto free_obj;
+	}
+
+	r = amdgpu_ttm_alloc_gart(&(userq_obj->obj)->tbo);
+	if (r) {
+		drm_file_err(uq_mgr->file, "Failed to alloc GART for userqueue object (%d)", r);
+		goto unresv;
+	}
+
+	r = amdgpu_bo_kmap(userq_obj->obj, &userq_obj->cpu_ptr);
+	if (r) {
+		drm_file_err(uq_mgr->file, "Failed to map BO for userqueue (%d)", r);
+		goto unresv;
+	}
+
+	userq_obj->gpu_addr = amdgpu_bo_gpu_offset(userq_obj->obj);
+	amdgpu_bo_unreserve(userq_obj->obj);
+	memset(userq_obj->cpu_ptr, 0, size);
+	return 0;
+
+unresv:
+	amdgpu_bo_unreserve(userq_obj->obj);
+
+free_obj:
+	amdgpu_bo_unref(&userq_obj->obj);
+	return r;
+}
+
+void amdgpu_userq_destroy_object(struct amdgpu_userq_mgr *uq_mgr,
+				 struct amdgpu_userq_obj *userq_obj)
+{
+	amdgpu_bo_kunmap(userq_obj->obj);
+	amdgpu_bo_unref(&userq_obj->obj);
+}
+
+uint64_t
+amdgpu_userq_get_doorbell_index(struct amdgpu_userq_mgr *uq_mgr,
+				struct amdgpu_db_info *db_info,
+				struct drm_file *filp)
+{
+	uint64_t index;
+	struct drm_gem_object *gobj;
+	struct amdgpu_userq_obj *db_obj = db_info->db_obj;
+	int r, db_size;
+
+	gobj = drm_gem_object_lookup(filp, db_info->doorbell_handle);
+	if (gobj == NULL) {
+		drm_file_err(uq_mgr->file, "Can't find GEM object for doorbell\n");
+		return -EINVAL;
+	}
+
+	db_obj->obj = amdgpu_bo_ref(gem_to_amdgpu_bo(gobj));
+	drm_gem_object_put(gobj);
+
+	r = amdgpu_bo_reserve(db_obj->obj, true);
+	if (r) {
+		drm_file_err(uq_mgr->file, "[Usermode queues] Failed to pin doorbell object\n");
+		goto unref_bo;
+	}
+
+	/* Pin the BO before generating the index, unpin in queue destroy */
+	r = amdgpu_bo_pin(db_obj->obj, AMDGPU_GEM_DOMAIN_DOORBELL);
+	if (r) {
+		drm_file_err(uq_mgr->file, "[Usermode queues] Failed to pin doorbell object\n");
+		goto unresv_bo;
+	}
+
+	switch (db_info->queue_type) {
+	case AMDGPU_HW_IP_GFX:
+	case AMDGPU_HW_IP_COMPUTE:
+	case AMDGPU_HW_IP_DMA:
+		db_size = sizeof(u64);
+		break;
+	default:
+		drm_file_err(uq_mgr->file, "[Usermode queues] IP %d not support\n",
+			     db_info->queue_type);
+		r = -EINVAL;
+		goto unpin_bo;
+	}
+
+	index = amdgpu_doorbell_index_on_bar(uq_mgr->adev, db_obj->obj,
+					     db_info->doorbell_offset, db_size);
+	drm_dbg_driver(adev_to_drm(uq_mgr->adev),
+		       "[Usermode queues] doorbell index=%lld\n", index);
+	amdgpu_bo_unreserve(db_obj->obj);
+	return index;
+
+unpin_bo:
+	amdgpu_bo_unpin(db_obj->obj);
+unresv_bo:
+	amdgpu_bo_unreserve(db_obj->obj);
+unref_bo:
+	amdgpu_bo_unref(&db_obj->obj);
+	return r;
+}
+
+static int
+amdgpu_userq_destroy(struct drm_file *filp, int queue_id)
+{
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
+	struct amdgpu_userq_mgr *uq_mgr = &fpriv->userq_mgr;
+	struct amdgpu_device *adev = uq_mgr->adev;
+	struct amdgpu_usermode_queue *queue;
+	int r = 0;
+
+	cancel_delayed_work_sync(&uq_mgr->resume_work);
+	mutex_lock(&uq_mgr->userq_mutex);
+
+	queue = amdgpu_userq_find(uq_mgr, queue_id);
+	if (!queue) {
+		drm_dbg_driver(adev_to_drm(uq_mgr->adev), "Invalid queue id to destroy\n");
+		mutex_unlock(&uq_mgr->userq_mutex);
+		return -EINVAL;
+	}
+	amdgpu_userq_wait_for_last_fence(uq_mgr, queue);
+	r = amdgpu_bo_reserve(queue->db_obj.obj, true);
+	if (!r) {
+		amdgpu_bo_unpin(queue->db_obj.obj);
+		amdgpu_bo_unreserve(queue->db_obj.obj);
+	}
+	amdgpu_bo_unref(&queue->db_obj.obj);
+	atomic_dec(&uq_mgr->userq_count[queue->queue_type]);
+#if defined(CONFIG_DEBUG_FS)
+	debugfs_remove_recursive(queue->debugfs_queue);
+#endif
+	amdgpu_userq_detect_and_reset_queues(uq_mgr);
+	r = amdgpu_userq_unmap_helper(uq_mgr, queue);
+	/*TODO: It requires a reset for userq hw unmap error*/
+	if (unlikely(r != AMDGPU_USERQ_STATE_UNMAPPED)) {
+		drm_warn(adev_to_drm(uq_mgr->adev), "trying to destroy a HW mapping userq\n");
+		queue->state = AMDGPU_USERQ_STATE_HUNG;
+	}
+	amdgpu_userq_cleanup(uq_mgr, queue, queue_id);
+	mutex_unlock(&uq_mgr->userq_mutex);
+
+	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
+
+	return r;
+}
+
+static int amdgpu_userq_priority_permit(struct drm_file *filp,
+					int priority)
+{
+	if (priority < AMDGPU_USERQ_CREATE_FLAGS_QUEUE_PRIORITY_HIGH)
+		return 0;
+
+	if (capable(CAP_SYS_NICE))
+		return 0;
+
+	if (drm_is_current_master(filp))
+		return 0;
+
+	return -EACCES;
+}
+
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_mqd_info_read(struct seq_file *m, void *unused)
+{
+	struct amdgpu_usermode_queue *queue = m->private;
+	struct amdgpu_bo *bo;
+	int r;
+
+	if (!queue || !queue->mqd.obj)
+		return -EINVAL;
+
+	bo = amdgpu_bo_ref(queue->mqd.obj);
+	r = amdgpu_bo_reserve(bo, true);
+	if (r) {
+		amdgpu_bo_unref(&bo);
+		return -EINVAL;
+	}
+
+	seq_printf(m, "queue_type: %d\n", queue->queue_type);
+	seq_printf(m, "mqd_gpu_address: 0x%llx\n", amdgpu_bo_gpu_offset(queue->mqd.obj));
+
+	amdgpu_bo_unreserve(bo);
+	amdgpu_bo_unref(&bo);
+
+	return 0;
+}
+
+static int amdgpu_mqd_info_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, amdgpu_mqd_info_read, inode->i_private);
+}
+
+static const struct file_operations amdgpu_mqd_info_fops = {
+	.owner = THIS_MODULE,
+	.open = amdgpu_mqd_info_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = single_release,
+};
+#endif
+
+static int
+amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args)
+{
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
+	struct amdgpu_userq_mgr *uq_mgr = &fpriv->userq_mgr;
+	struct amdgpu_device *adev = uq_mgr->adev;
+	const struct amdgpu_userq_funcs *uq_funcs;
+	struct amdgpu_usermode_queue *queue;
+	struct amdgpu_db_info db_info;
+	char *queue_name;
+	bool skip_map_queue;
+	u32 qid;
+	uint64_t index;
+	int r = 0;
+	int priority =
+		(args->in.flags & AMDGPU_USERQ_CREATE_FLAGS_QUEUE_PRIORITY_MASK) >>
+		AMDGPU_USERQ_CREATE_FLAGS_QUEUE_PRIORITY_SHIFT;
+
+	r = amdgpu_userq_priority_permit(filp, priority);
+	if (r)
+		return r;
+
+	r = pm_runtime_get_sync(adev_to_drm(adev)->dev);
+	if (r < 0) {
+		drm_file_err(uq_mgr->file, "pm_runtime_get_sync() failed for userqueue create\n");
+		pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
+		return r;
+	}
+
+	/*
+	 * There could be a situation that we are creating a new queue while
+	 * the other queues under this UQ_mgr are suspended. So if there is any
+	 * resume work pending, wait for it to get done.
+	 *
+	 * This will also make sure we have a valid eviction fence ready to be used.
+	 */
+	amdgpu_userq_ensure_ev_fence(&fpriv->userq_mgr, &fpriv->evf_mgr);
+
+	uq_funcs = adev->userq_funcs[args->in.ip_type];
+	if (!uq_funcs) {
+		drm_file_err(uq_mgr->file, "Usermode queue is not supported for this IP (%u)\n",
+			     args->in.ip_type);
+		r = -EINVAL;
+		goto unlock;
+	}
+
+	queue = kzalloc(sizeof(struct amdgpu_usermode_queue), GFP_KERNEL);
+	if (!queue) {
+		drm_file_err(uq_mgr->file, "Failed to allocate memory for queue\n");
+		r = -ENOMEM;
+		goto unlock;
+	}
+
+	INIT_LIST_HEAD(&queue->userq_va_list);
+	queue->doorbell_handle = args->in.doorbell_handle;
+	queue->queue_type = args->in.ip_type;
+	queue->vm = &fpriv->vm;
+	queue->priority = priority;
+
+	db_info.queue_type = queue->queue_type;
+	db_info.doorbell_handle = queue->doorbell_handle;
+	db_info.db_obj = &queue->db_obj;
+	db_info.doorbell_offset = args->in.doorbell_offset;
+
+	/* Validate the userq virtual address.*/
+	if (amdgpu_userq_input_va_validate(queue, args->in.queue_va, args->in.queue_size) ||
+	    amdgpu_userq_input_va_validate(queue, args->in.rptr_va, AMDGPU_GPU_PAGE_SIZE) ||
+	    amdgpu_userq_input_va_validate(queue, args->in.wptr_va, AMDGPU_GPU_PAGE_SIZE)) {
+		r = -EINVAL;
+		kfree(queue);
+		goto unlock;
+	}
+
+	/* Convert relative doorbell offset into absolute doorbell index */
+	index = amdgpu_userq_get_doorbell_index(uq_mgr, &db_info, filp);
+	if (index == (uint64_t)-EINVAL) {
+		drm_file_err(uq_mgr->file, "Failed to get doorbell for queue\n");
+		kfree(queue);
+		r = -EINVAL;
+		goto unlock;
+	}
+
+	queue->doorbell_index = index;
+	xa_init_flags(&queue->fence_drv_xa, XA_FLAGS_ALLOC);
+	r = amdgpu_userq_fence_driver_alloc(adev, queue);
+	if (r) {
+		drm_file_err(uq_mgr->file, "Failed to alloc fence driver\n");
+		goto unlock;
+	}
+
+	r = uq_funcs->mqd_create(uq_mgr, &args->in, queue);
+	if (r) {
+		drm_file_err(uq_mgr->file, "Failed to create Queue\n");
+		amdgpu_userq_fence_driver_free(queue);
+		kfree(queue);
+		goto unlock;
+	}
+
+	/* Wait for mode-1 reset to complete */
+	down_read(&adev->reset_domain->sem);
+	r = xa_err(xa_store_irq(&adev->userq_doorbell_xa, index, queue, GFP_KERNEL));
+	if (r) {
+		kfree(queue);
+		up_read(&adev->reset_domain->sem);
+		goto unlock;
+	}
+
+	r = xa_alloc(&uq_mgr->userq_mgr_xa, &qid, queue, XA_LIMIT(1, AMDGPU_MAX_USERQ_COUNT), GFP_KERNEL);
+	if (r) {
+		drm_file_err(uq_mgr->file, "Failed to allocate a queue id\n");
+		amdgpu_userq_fence_driver_free(queue);
+		uq_funcs->mqd_destroy(uq_mgr, queue);
+		kfree(queue);
+		r = -ENOMEM;
+		up_read(&adev->reset_domain->sem);
+		goto unlock;
+	}
+	up_read(&adev->reset_domain->sem);
+	queue->userq_mgr = uq_mgr;
+
+	/* don't map the queue if scheduling is halted */
+	if (adev->userq_halt_for_enforce_isolation &&
+	    ((queue->queue_type == AMDGPU_HW_IP_GFX) ||
+	     (queue->queue_type == AMDGPU_HW_IP_COMPUTE)))
+		skip_map_queue = true;
+	else
+		skip_map_queue = false;
+	if (!skip_map_queue) {
+		r = amdgpu_userq_map_helper(uq_mgr, queue);
+		if (r) {
+			drm_file_err(uq_mgr->file, "Failed to map Queue\n");
+			xa_erase(&uq_mgr->userq_mgr_xa, qid);
+			amdgpu_userq_fence_driver_free(queue);
+			uq_funcs->mqd_destroy(uq_mgr, queue);
+			kfree(queue);
+			goto unlock;
+		}
+	}
+
+	queue_name = kasprintf(GFP_KERNEL, "queue-%d", qid);
+	if (!queue_name) {
+		r = -ENOMEM;
+		goto unlock;
+	}
+
+#if defined(CONFIG_DEBUG_FS)
+	/* Queue dentry per client to hold MQD information   */
+	queue->debugfs_queue = debugfs_create_dir(queue_name, filp->debugfs_client);
+	debugfs_create_file("mqd_info", 0444, queue->debugfs_queue, queue, &amdgpu_mqd_info_fops);
+#endif
+	kfree(queue_name);
+
+	args->out.queue_id = qid;
+	atomic_inc(&uq_mgr->userq_count[queue->queue_type]);
+
+unlock:
+	mutex_unlock(&uq_mgr->userq_mutex);
+
+	return r;
+}
+
+static int amdgpu_userq_input_args_validate(struct drm_device *dev,
+					union drm_amdgpu_userq *args,
+					struct drm_file *filp)
+{
+	struct amdgpu_device *adev = drm_to_adev(dev);
+
+	switch (args->in.op) {
+	case AMDGPU_USERQ_OP_CREATE:
+		if (args->in.flags & ~(AMDGPU_USERQ_CREATE_FLAGS_QUEUE_PRIORITY_MASK |
+				       AMDGPU_USERQ_CREATE_FLAGS_QUEUE_SECURE))
+			return -EINVAL;
+		/* Usermode queues are only supported for GFX IP as of now */
+		if (args->in.ip_type != AMDGPU_HW_IP_GFX &&
+		    args->in.ip_type != AMDGPU_HW_IP_DMA &&
+		    args->in.ip_type != AMDGPU_HW_IP_COMPUTE) {
+			drm_file_err(filp, "Usermode queue doesn't support IP type %u\n",
+				     args->in.ip_type);
+			return -EINVAL;
+		}
+
+		if ((args->in.flags & AMDGPU_USERQ_CREATE_FLAGS_QUEUE_SECURE) &&
+		    (args->in.ip_type != AMDGPU_HW_IP_GFX) &&
+		    (args->in.ip_type != AMDGPU_HW_IP_COMPUTE) &&
+		    !amdgpu_is_tmz(adev)) {
+			drm_file_err(filp, "Secure only supported on GFX/Compute queues\n");
+			return -EINVAL;
+		}
+
+		if (args->in.queue_va == AMDGPU_BO_INVALID_OFFSET ||
+		    args->in.queue_va == 0 ||
+		    args->in.queue_size == 0) {
+			drm_file_err(filp, "invalidate userq queue va or size\n");
+			return -EINVAL;
+		}
+		if (!args->in.wptr_va || !args->in.rptr_va) {
+			drm_file_err(filp, "invalidate userq queue rptr or wptr\n");
+			return -EINVAL;
+		}
+		break;
+	case AMDGPU_USERQ_OP_FREE:
+		if (args->in.ip_type ||
+		    args->in.doorbell_handle ||
+		    args->in.doorbell_offset ||
+		    args->in.flags ||
+		    args->in.queue_va ||
+		    args->in.queue_size ||
+		    args->in.rptr_va ||
+		    args->in.wptr_va ||
+		    args->in.mqd ||
+		    args->in.mqd_size)
+			return -EINVAL;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+int amdgpu_userq_ioctl(struct drm_device *dev, void *data,
+		       struct drm_file *filp)
+{
+	union drm_amdgpu_userq *args = data;
+	int r;
+
+	if (amdgpu_userq_input_args_validate(dev, args, filp) < 0)
+		return -EINVAL;
+
+	switch (args->in.op) {
+	case AMDGPU_USERQ_OP_CREATE:
+		r = amdgpu_userq_create(filp, args);
+		if (r)
+			drm_file_err(filp, "Failed to create usermode queue\n");
+		break;
+
+	case AMDGPU_USERQ_OP_FREE:
+		r = amdgpu_userq_destroy(filp, args->in.queue_id);
+		if (r)
+			drm_file_err(filp, "Failed to destroy usermode queue\n");
+		break;
+
+	default:
+		drm_dbg_driver(dev, "Invalid user queue op specified: %d\n", args->in.op);
+		return -EINVAL;
+	}
+
+	return r;
+}
+
+static int
+amdgpu_userq_restore_all(struct amdgpu_userq_mgr *uq_mgr)
+{
+	struct amdgpu_usermode_queue *queue;
+	unsigned long queue_id;
+	int ret = 0, r;
+
+	/* Resume all the queues for this process */
+	xa_for_each(&uq_mgr->userq_mgr_xa, queue_id, queue) {
+
+		if (!amdgpu_userq_buffer_vas_mapped(queue)) {
+			drm_file_err(uq_mgr->file,
+				     "trying restore queue without va mapping\n");
+			queue->state = AMDGPU_USERQ_STATE_INVALID_VA;
+			continue;
+		}
+
+		r = amdgpu_userq_restore_helper(uq_mgr, queue);
+		if (r)
+			ret = r;
+	}
+
+	if (ret)
+		drm_file_err(uq_mgr->file, "Failed to map all the queues\n");
+	return ret;
+}
+
+static int amdgpu_userq_validate_vm(void *param, struct amdgpu_bo *bo)
+{
+	struct ttm_operation_ctx ctx = { false, false };
+
+	amdgpu_bo_placement_from_domain(bo, bo->allowed_domains);
+	return ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+}
+
+/* Handle all BOs on the invalidated list, validate them and update the PTs */
+static int
+amdgpu_userq_bo_validate(struct amdgpu_device *adev, struct drm_exec *exec,
+			 struct amdgpu_vm *vm)
+{
+	struct ttm_operation_ctx ctx = { false, false };
+	struct amdgpu_bo_va *bo_va;
+	struct amdgpu_bo *bo;
+	int ret;
+
+	spin_lock(&vm->status_lock);
+	while (!list_empty(&vm->invalidated)) {
+		bo_va = list_first_entry(&vm->invalidated,
+					 struct amdgpu_bo_va,
+					 base.vm_status);
+		spin_unlock(&vm->status_lock);
+
+		bo = bo_va->base.bo;
+		ret = drm_exec_prepare_obj(exec, &bo->tbo.base, 2);
+		if (unlikely(ret))
+			return ret;
+
+		amdgpu_bo_placement_from_domain(bo, bo->allowed_domains);
+		ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+		if (ret)
+			return ret;
+
+		/* This moves the bo_va to the done list */
+		ret = amdgpu_vm_bo_update(adev, bo_va, false);
+		if (ret)
+			return ret;
+
+		spin_lock(&vm->status_lock);
+	}
+	spin_unlock(&vm->status_lock);
+
+	return 0;
+}
+
+/* Make sure the whole VM is ready to be used */
+static int
+amdgpu_userq_vm_validate(struct amdgpu_userq_mgr *uq_mgr)
+{
+	struct amdgpu_fpriv *fpriv = uq_mgr_to_fpriv(uq_mgr);
+	bool invalidated = false, new_addition = false;
+	struct ttm_operation_ctx ctx = { true, false };
+	struct amdgpu_device *adev = uq_mgr->adev;
+	struct amdgpu_hmm_range *range;
+	struct amdgpu_vm *vm = &fpriv->vm;
+	unsigned long key, tmp_key;
+	struct amdgpu_bo_va *bo_va;
+	struct amdgpu_bo *bo;
+	struct drm_exec exec;
+	struct xarray xa;
+	int ret;
+
+	xa_init(&xa);
+
+retry_lock:
+	drm_exec_init(&exec, DRM_EXEC_IGNORE_DUPLICATES, 0);
+	drm_exec_until_all_locked(&exec) {
+		ret = amdgpu_vm_lock_pd(vm, &exec, 1);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(ret))
+			goto unlock_all;
+
+		ret = amdgpu_vm_lock_done_list(vm, &exec, 1);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(ret))
+			goto unlock_all;
+
+		/* This validates PDs, PTs and per VM BOs */
+		ret = amdgpu_vm_validate(adev, vm, NULL,
+					 amdgpu_userq_validate_vm,
+					 NULL);
+		if (unlikely(ret))
+			goto unlock_all;
+
+		/* This locks and validates the remaining evicted BOs */
+		ret = amdgpu_userq_bo_validate(adev, &exec, vm);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(ret))
+			goto unlock_all;
+	}
+
+	if (invalidated) {
+		xa_for_each(&xa, tmp_key, range) {
+			bo = range->bo;
+			amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU);
+			ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+			if (ret)
+				goto unlock_all;
+
+			amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, range);
+
+			amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT);
+			ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+			if (ret)
+				goto unlock_all;
+		}
+		invalidated = false;
+	}
+
+	ret = amdgpu_vm_handle_moved(adev, vm, NULL);
+	if (ret)
+		goto unlock_all;
+
+	key = 0;
+	/* Validate User Ptr BOs */
+	list_for_each_entry(bo_va, &vm->done, base.vm_status) {
+		bo = bo_va->base.bo;
+		if (!bo)
+			continue;
+
+		if (!amdgpu_ttm_tt_is_userptr(bo->tbo.ttm))
+			continue;
+
+		range = xa_load(&xa, key);
+		if (range && range->bo != bo) {
+			xa_erase(&xa, key);
+			amdgpu_hmm_range_free(range);
+			range = NULL;
+		}
+
+		if (!range) {
+			range = amdgpu_hmm_range_alloc(bo);
+			if (!range) {
+				ret = -ENOMEM;
+				goto unlock_all;
+			}
+
+			xa_store(&xa, key, range, GFP_KERNEL);
+			new_addition = true;
+		}
+		key++;
+	}
+
+	if (new_addition) {
+		drm_exec_fini(&exec);
+		xa_for_each(&xa, tmp_key, range) {
+			if (!range)
+				continue;
+			bo = range->bo;
+			ret = amdgpu_ttm_tt_get_user_pages(bo, range);
+			if (ret)
+				goto unlock_all;
+		}
+
+		invalidated = true;
+		new_addition = false;
+		goto retry_lock;
+	}
+
+	ret = amdgpu_vm_update_pdes(adev, vm, false);
+	if (ret)
+		goto unlock_all;
+
+	/*
+	 * We need to wait for all VM updates to finish before restarting the
+	 * queues. Using the done list like that is now ok since everything is
+	 * locked in place.
+	 */
+	list_for_each_entry(bo_va, &vm->done, base.vm_status)
+		dma_fence_wait(bo_va->last_pt_update, false);
+	dma_fence_wait(vm->last_update, false);
+
+	ret = amdgpu_eviction_fence_replace_fence(&fpriv->evf_mgr, &exec);
+	if (ret)
+		drm_file_err(uq_mgr->file, "Failed to replace eviction fence\n");
+
+unlock_all:
+	drm_exec_fini(&exec);
+	xa_for_each(&xa, tmp_key, range) {
+		if (!range)
+			continue;
+		bo = range->bo;
+		amdgpu_hmm_range_free(range);
+	}
+	xa_destroy(&xa);
+	return ret;
+}
+
+static void amdgpu_userq_restore_worker(struct work_struct *work)
+{
+	struct amdgpu_userq_mgr *uq_mgr = work_to_uq_mgr(work, resume_work.work);
+	struct amdgpu_fpriv *fpriv = uq_mgr_to_fpriv(uq_mgr);
+	int ret;
+
+	flush_delayed_work(&fpriv->evf_mgr.suspend_work);
+
+	mutex_lock(&uq_mgr->userq_mutex);
+
+	ret = amdgpu_userq_vm_validate(uq_mgr);
+	if (ret) {
+		drm_file_err(uq_mgr->file, "Failed to validate BOs to restore\n");
+		goto unlock;
+	}
+
+	ret = amdgpu_userq_restore_all(uq_mgr);
+	if (ret) {
+		drm_file_err(uq_mgr->file, "Failed to restore all queues\n");
+		goto unlock;
+	}
+
+unlock:
+	mutex_unlock(&uq_mgr->userq_mutex);
+}
+
+static int
+amdgpu_userq_evict_all(struct amdgpu_userq_mgr *uq_mgr)
+{
+	struct amdgpu_usermode_queue *queue;
+	unsigned long queue_id;
+	int ret = 0, r;
+
+	amdgpu_userq_detect_and_reset_queues(uq_mgr);
+	/* Try to unmap all the queues in this process ctx */
+	xa_for_each(&uq_mgr->userq_mgr_xa, queue_id, queue) {
+		r = amdgpu_userq_preempt_helper(uq_mgr, queue);
+		if (r)
+			ret = r;
+	}
+
+	if (ret)
+		drm_file_err(uq_mgr->file, "Couldn't unmap all the queues\n");
+	return ret;
+}
+
+void amdgpu_userq_reset_work(struct work_struct *work)
+{
+	struct amdgpu_device *adev = container_of(work, struct amdgpu_device,
+						  userq_reset_work);
+	struct amdgpu_reset_context reset_context;
+
+	memset(&reset_context, 0, sizeof(reset_context));
+
+	reset_context.method = AMD_RESET_METHOD_NONE;
+	reset_context.reset_req_dev = adev;
+	reset_context.src = AMDGPU_RESET_SRC_USERQ;
+	set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
+	/*set_bit(AMDGPU_SKIP_COREDUMP, &reset_context.flags);*/
+
+	amdgpu_device_gpu_recover(adev, NULL, &reset_context);
+}
+
+static int
+amdgpu_userq_wait_for_signal(struct amdgpu_userq_mgr *uq_mgr)
+{
+	struct amdgpu_usermode_queue *queue;
+	unsigned long queue_id;
+	int ret;
+
+	xa_for_each(&uq_mgr->userq_mgr_xa, queue_id, queue) {
+		struct dma_fence *f = queue->last_fence;
+
+		if (!f || dma_fence_is_signaled(f))
+			continue;
+		ret = dma_fence_wait_timeout(f, true, msecs_to_jiffies(100));
+		if (ret <= 0) {
+			drm_file_err(uq_mgr->file, "Timed out waiting for fence=%llu:%llu\n",
+				     f->context, f->seqno);
+			return -ETIMEDOUT;
+		}
+	}
+
+	return 0;
+}
+
+void
+amdgpu_userq_evict(struct amdgpu_userq_mgr *uq_mgr,
+		   struct amdgpu_eviction_fence *ev_fence)
+{
+	struct amdgpu_fpriv *fpriv = uq_mgr_to_fpriv(uq_mgr);
+	struct amdgpu_eviction_fence_mgr *evf_mgr = &fpriv->evf_mgr;
+	struct amdgpu_device *adev = uq_mgr->adev;
+	int ret;
+
+	/* Wait for any pending userqueue fence work to finish */
+	ret = amdgpu_userq_wait_for_signal(uq_mgr);
+	if (ret)
+		dev_err(adev->dev, "Not evicting userqueue, timeout waiting for work\n");
+
+	ret = amdgpu_userq_evict_all(uq_mgr);
+	if (ret)
+		dev_err(adev->dev, "Failed to evict userqueue\n");
+
+	/* Signal current eviction fence */
+	amdgpu_eviction_fence_signal(evf_mgr, ev_fence);
+
+	if (evf_mgr->fd_closing) {
+		cancel_delayed_work_sync(&uq_mgr->resume_work);
+		return;
+	}
+
+	/* Schedule a resume work */
+	schedule_delayed_work(&uq_mgr->resume_work, 0);
+}
+
+int amdgpu_userq_mgr_init(struct amdgpu_userq_mgr *userq_mgr, struct drm_file *file_priv,
+			  struct amdgpu_device *adev)
+{
+	mutex_init(&userq_mgr->userq_mutex);
+	xa_init_flags(&userq_mgr->userq_mgr_xa, XA_FLAGS_ALLOC);
+	userq_mgr->adev = adev;
+	userq_mgr->file = file_priv;
+
+	INIT_DELAYED_WORK(&userq_mgr->resume_work, amdgpu_userq_restore_worker);
+	return 0;
+}
+
+void amdgpu_userq_mgr_fini(struct amdgpu_userq_mgr *userq_mgr)
+{
+	struct amdgpu_usermode_queue *queue;
+	unsigned long queue_id;
+
+	cancel_delayed_work_sync(&userq_mgr->resume_work);
+
+	mutex_lock(&userq_mgr->userq_mutex);
+	amdgpu_userq_detect_and_reset_queues(userq_mgr);
+	xa_for_each(&userq_mgr->userq_mgr_xa, queue_id, queue) {
+		amdgpu_userq_wait_for_last_fence(userq_mgr, queue);
+		amdgpu_userq_unmap_helper(userq_mgr, queue);
+		amdgpu_userq_cleanup(userq_mgr, queue, queue_id);
+	}
+
+	xa_destroy(&userq_mgr->userq_mgr_xa);
+	mutex_unlock(&userq_mgr->userq_mutex);
+	mutex_destroy(&userq_mgr->userq_mutex);
+}
+
+int amdgpu_userq_suspend(struct amdgpu_device *adev)
+{
+	u32 ip_mask = amdgpu_userq_get_supported_ip_mask(adev);
+	struct amdgpu_usermode_queue *queue;
+	struct amdgpu_userq_mgr *uqm;
+	unsigned long queue_id;
+	int r;
+
+	if (!ip_mask)
+		return 0;
+
+	xa_for_each(&adev->userq_doorbell_xa, queue_id, queue) {
+		uqm = queue->userq_mgr;
+		cancel_delayed_work_sync(&uqm->resume_work);
+		guard(mutex)(&uqm->userq_mutex);
+		amdgpu_userq_detect_and_reset_queues(uqm);
+		if (adev->in_s0ix)
+			r = amdgpu_userq_preempt_helper(uqm, queue);
+		else
+			r = amdgpu_userq_unmap_helper(uqm, queue);
+		if (r)
+			return r;
+	}
+	return 0;
+}
+
+int amdgpu_userq_resume(struct amdgpu_device *adev)
+{
+	u32 ip_mask = amdgpu_userq_get_supported_ip_mask(adev);
+	struct amdgpu_usermode_queue *queue;
+	struct amdgpu_userq_mgr *uqm;
+	unsigned long queue_id;
+	int r;
+
+	if (!ip_mask)
+		return 0;
+
+	xa_for_each(&adev->userq_doorbell_xa, queue_id, queue) {
+		uqm = queue->userq_mgr;
+		guard(mutex)(&uqm->userq_mutex);
+		if (adev->in_s0ix)
+			r = amdgpu_userq_restore_helper(uqm, queue);
+		else
+			r = amdgpu_userq_map_helper(uqm, queue);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+int amdgpu_userq_stop_sched_for_enforce_isolation(struct amdgpu_device *adev,
+						  u32 idx)
+{
+	u32 ip_mask = amdgpu_userq_get_supported_ip_mask(adev);
+	struct amdgpu_usermode_queue *queue;
+	struct amdgpu_userq_mgr *uqm;
+	unsigned long queue_id;
+	int ret = 0, r;
+
+	/* only need to stop gfx/compute */
+	if (!(ip_mask & ((1 << AMDGPU_HW_IP_GFX) | (1 << AMDGPU_HW_IP_COMPUTE))))
+		return 0;
+
+	if (adev->userq_halt_for_enforce_isolation)
+		dev_warn(adev->dev, "userq scheduling already stopped!\n");
+	adev->userq_halt_for_enforce_isolation = true;
+	xa_for_each(&adev->userq_doorbell_xa, queue_id, queue) {
+		uqm = queue->userq_mgr;
+		cancel_delayed_work_sync(&uqm->resume_work);
+		mutex_lock(&uqm->userq_mutex);
+		if (((queue->queue_type == AMDGPU_HW_IP_GFX) ||
+		     (queue->queue_type == AMDGPU_HW_IP_COMPUTE)) &&
+		    (queue->xcp_id == idx)) {
+			amdgpu_userq_detect_and_reset_queues(uqm);
+			r = amdgpu_userq_preempt_helper(uqm, queue);
+			if (r)
+				ret = r;
+		}
+		mutex_unlock(&uqm->userq_mutex);
+	}
+
+	return ret;
+}
+
+int amdgpu_userq_start_sched_for_enforce_isolation(struct amdgpu_device *adev,
+						   u32 idx)
+{
+	u32 ip_mask = amdgpu_userq_get_supported_ip_mask(adev);
+	struct amdgpu_usermode_queue *queue;
+	struct amdgpu_userq_mgr *uqm;
+	unsigned long queue_id;
+	int ret = 0, r;
+
+	/* only need to stop gfx/compute */
+	if (!(ip_mask & ((1 << AMDGPU_HW_IP_GFX) | (1 << AMDGPU_HW_IP_COMPUTE))))
+		return 0;
+
+	if (!adev->userq_halt_for_enforce_isolation)
+		dev_warn(adev->dev, "userq scheduling already started!\n");
+	adev->userq_halt_for_enforce_isolation = false;
+	xa_for_each(&adev->userq_doorbell_xa, queue_id, queue) {
+		uqm = queue->userq_mgr;
+		mutex_lock(&uqm->userq_mutex);
+			if (((queue->queue_type == AMDGPU_HW_IP_GFX) ||
+			     (queue->queue_type == AMDGPU_HW_IP_COMPUTE)) &&
+			    (queue->xcp_id == idx)) {
+				r = amdgpu_userq_restore_helper(uqm, queue);
+				if (r)
+					ret = r;
+			}
+		mutex_unlock(&uqm->userq_mutex);
+	}
+
+	return ret;
+}
+
+int amdgpu_userq_gem_va_unmap_validate(struct amdgpu_device *adev,
+				       struct amdgpu_bo_va_mapping *mapping,
+				       uint64_t saddr)
+{
+	u32 ip_mask = amdgpu_userq_get_supported_ip_mask(adev);
+	struct amdgpu_bo_va *bo_va = mapping->bo_va;
+	struct dma_resv *resv = bo_va->base.bo->tbo.base.resv;
+	int ret = 0;
+
+	if (!ip_mask)
+		return 0;
+
+	dev_warn_once(adev->dev, "now unmapping a vital queue va:%llx\n", saddr);
+	/**
+	 * The userq VA mapping reservation should include the eviction fence,
+	 * if the eviction fence can't signal successfully during unmapping,
+	 * then driver will warn to flag this improper unmap of the userq VA.
+	 * Note: The eviction fence may be attached to different BOs, and this
+	 * unmap is only for one kind of userq VAs, so at this point suppose
+	 * the eviction fence is always unsignaled.
+	 */
+	if (!dma_resv_test_signaled(resv, DMA_RESV_USAGE_BOOKKEEP)) {
+		ret = dma_resv_wait_timeout(resv, DMA_RESV_USAGE_BOOKKEEP, true,
+					    MAX_SCHEDULE_TIMEOUT);
+		if (ret <= 0)
+			return -EBUSY;
+	}
+
+	return 0;
+}
+
+void amdgpu_userq_pre_reset(struct amdgpu_device *adev)
+{
+	const struct amdgpu_userq_funcs *userq_funcs;
+	struct amdgpu_usermode_queue *queue;
+	struct amdgpu_userq_mgr *uqm;
+	unsigned long queue_id;
+
+	xa_for_each(&adev->userq_doorbell_xa, queue_id, queue) {
+		uqm = queue->userq_mgr;
+		cancel_delayed_work_sync(&uqm->resume_work);
+		if (queue->state == AMDGPU_USERQ_STATE_MAPPED) {
+			amdgpu_userq_wait_for_last_fence(uqm, queue);
+			userq_funcs = adev->userq_funcs[queue->queue_type];
+			userq_funcs->unmap(uqm, queue);
+			/* just mark all queues as hung at this point.
+			 * if unmap succeeds, we could map again
+			 * in amdgpu_userq_post_reset() if vram is not lost
+			 */
+			queue->state = AMDGPU_USERQ_STATE_HUNG;
+			amdgpu_userq_fence_driver_force_completion(queue);
+		}
+	}
+}
+
+int amdgpu_userq_post_reset(struct amdgpu_device *adev, bool vram_lost)
+{
+	/* if any queue state is AMDGPU_USERQ_STATE_UNMAPPED
+	 * at this point, we should be able to map it again
+	 * and continue if vram is not lost.
+	 */
+	struct amdgpu_userq_mgr *uqm;
+	struct amdgpu_usermode_queue *queue;
+	const struct amdgpu_userq_funcs *userq_funcs;
+	unsigned long queue_id;
+	int r = 0;
+
+	xa_for_each(&adev->userq_doorbell_xa, queue_id, queue) {
+		uqm = queue->userq_mgr;
+		if (queue->state == AMDGPU_USERQ_STATE_HUNG && !vram_lost) {
+			userq_funcs = adev->userq_funcs[queue->queue_type];
+			/* Re-map queue */
+			r = userq_funcs->map(uqm, queue);
+			if (r) {
+				dev_err(adev->dev, "Failed to remap queue %ld\n", queue_id);
+				continue;
+			}
+			queue->state = AMDGPU_USERQ_STATE_MAPPED;
+		}
+	}
+
+	return r;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h
new file mode 100644
index 000000000000..c37444427a14
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h
@@ -0,0 +1,161 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef AMDGPU_USERQ_H_
+#define AMDGPU_USERQ_H_
+#include "amdgpu_eviction_fence.h"
+
+#define AMDGPU_MAX_USERQ_COUNT 512
+
+#define to_ev_fence(f) container_of(f, struct amdgpu_eviction_fence, base)
+#define uq_mgr_to_fpriv(u) container_of(u, struct amdgpu_fpriv, userq_mgr)
+#define work_to_uq_mgr(w, name) container_of(w, struct amdgpu_userq_mgr, name)
+
+enum amdgpu_userq_state {
+	AMDGPU_USERQ_STATE_UNMAPPED = 0,
+	AMDGPU_USERQ_STATE_MAPPED,
+	AMDGPU_USERQ_STATE_PREEMPTED,
+	AMDGPU_USERQ_STATE_HUNG,
+	AMDGPU_USERQ_STATE_INVALID_VA,
+};
+
+struct amdgpu_mqd_prop;
+
+struct amdgpu_userq_obj {
+	void		 *cpu_ptr;
+	uint64_t	 gpu_addr;
+	struct amdgpu_bo *obj;
+};
+
+struct amdgpu_userq_va_cursor {
+	u64			gpu_addr;
+	struct list_head	list;
+};
+
+struct amdgpu_usermode_queue {
+	int			queue_type;
+	enum amdgpu_userq_state state;
+	uint64_t		doorbell_handle;
+	uint64_t		doorbell_index;
+	uint64_t		flags;
+	struct amdgpu_mqd_prop	*userq_prop;
+	struct amdgpu_userq_mgr *userq_mgr;
+	struct amdgpu_vm	*vm;
+	struct amdgpu_userq_obj mqd;
+	struct amdgpu_userq_obj	db_obj;
+	struct amdgpu_userq_obj fw_obj;
+	struct amdgpu_userq_obj wptr_obj;
+	struct xarray		fence_drv_xa;
+	struct amdgpu_userq_fence_driver *fence_drv;
+	struct dma_fence	*last_fence;
+	u32			xcp_id;
+	int			priority;
+	struct dentry		*debugfs_queue;
+
+	struct list_head	userq_va_list;
+};
+
+struct amdgpu_userq_funcs {
+	int (*mqd_create)(struct amdgpu_userq_mgr *uq_mgr,
+			  struct drm_amdgpu_userq_in *args,
+			  struct amdgpu_usermode_queue *queue);
+	void (*mqd_destroy)(struct amdgpu_userq_mgr *uq_mgr,
+			    struct amdgpu_usermode_queue *uq);
+	int (*unmap)(struct amdgpu_userq_mgr *uq_mgr,
+		     struct amdgpu_usermode_queue *queue);
+	int (*map)(struct amdgpu_userq_mgr *uq_mgr,
+		   struct amdgpu_usermode_queue *queue);
+	int (*preempt)(struct amdgpu_userq_mgr *uq_mgr,
+		   struct amdgpu_usermode_queue *queue);
+	int (*restore)(struct amdgpu_userq_mgr *uq_mgr,
+		   struct amdgpu_usermode_queue *queue);
+	int (*detect_and_reset)(struct amdgpu_device *adev,
+		  int queue_type);
+};
+
+/* Usermode queues for gfx */
+struct amdgpu_userq_mgr {
+	/**
+	 * @userq_mgr_xa: Per-process user queue map (queue ID → queue)
+	 * Key: queue_id (unique ID within the process's userq manager)
+	 * Value: struct amdgpu_usermode_queue
+	 */
+	struct xarray			userq_mgr_xa;
+	struct mutex			userq_mutex;
+	struct amdgpu_device		*adev;
+	struct delayed_work		resume_work;
+	struct drm_file			*file;
+	atomic_t                        userq_count[AMDGPU_RING_TYPE_MAX];
+};
+
+struct amdgpu_db_info {
+	uint64_t doorbell_handle;
+	uint32_t queue_type;
+	uint32_t doorbell_offset;
+	struct amdgpu_userq_obj	*db_obj;
+};
+
+int amdgpu_userq_ioctl(struct drm_device *dev, void *data, struct drm_file *filp);
+
+int amdgpu_userq_mgr_init(struct amdgpu_userq_mgr *userq_mgr, struct drm_file *file_priv,
+			  struct amdgpu_device *adev);
+
+void amdgpu_userq_mgr_fini(struct amdgpu_userq_mgr *userq_mgr);
+
+int amdgpu_userq_create_object(struct amdgpu_userq_mgr *uq_mgr,
+			       struct amdgpu_userq_obj *userq_obj,
+			       int size);
+
+void amdgpu_userq_destroy_object(struct amdgpu_userq_mgr *uq_mgr,
+				 struct amdgpu_userq_obj *userq_obj);
+
+void amdgpu_userq_evict(struct amdgpu_userq_mgr *uq_mgr,
+			struct amdgpu_eviction_fence *ev_fence);
+
+void amdgpu_userq_ensure_ev_fence(struct amdgpu_userq_mgr *userq_mgr,
+				  struct amdgpu_eviction_fence_mgr *evf_mgr);
+
+uint64_t amdgpu_userq_get_doorbell_index(struct amdgpu_userq_mgr *uq_mgr,
+					 struct amdgpu_db_info *db_info,
+					     struct drm_file *filp);
+
+u32 amdgpu_userq_get_supported_ip_mask(struct amdgpu_device *adev);
+
+int amdgpu_userq_suspend(struct amdgpu_device *adev);
+int amdgpu_userq_resume(struct amdgpu_device *adev);
+
+int amdgpu_userq_stop_sched_for_enforce_isolation(struct amdgpu_device *adev,
+						  u32 idx);
+int amdgpu_userq_start_sched_for_enforce_isolation(struct amdgpu_device *adev,
+						   u32 idx);
+void amdgpu_userq_reset_work(struct work_struct *work);
+void amdgpu_userq_pre_reset(struct amdgpu_device *adev);
+int amdgpu_userq_post_reset(struct amdgpu_device *adev, bool vram_lost);
+
+int amdgpu_userq_input_va_validate(struct amdgpu_usermode_queue *queue,
+				   u64 addr, u64 expected_size);
+int amdgpu_userq_gem_va_unmap_validate(struct amdgpu_device *adev,
+				       struct amdgpu_bo_va_mapping *mapping,
+				       uint64_t saddr);
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c
new file mode 100644
index 000000000000..eba9fb359047
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c
@@ -0,0 +1,1011 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/kref.h>
+#include <linux/slab.h>
+#include <linux/dma-fence-unwrap.h>
+
+#include <drm/drm_exec.h>
+#include <drm/drm_syncobj.h>
+
+#include "amdgpu.h"
+#include "amdgpu_userq_fence.h"
+
+static const struct dma_fence_ops amdgpu_userq_fence_ops;
+static struct kmem_cache *amdgpu_userq_fence_slab;
+
+int amdgpu_userq_fence_slab_init(void)
+{
+	amdgpu_userq_fence_slab = kmem_cache_create("amdgpu_userq_fence",
+						    sizeof(struct amdgpu_userq_fence),
+						    0,
+						    SLAB_HWCACHE_ALIGN,
+						    NULL);
+	if (!amdgpu_userq_fence_slab)
+		return -ENOMEM;
+
+	return 0;
+}
+
+void amdgpu_userq_fence_slab_fini(void)
+{
+	rcu_barrier();
+	kmem_cache_destroy(amdgpu_userq_fence_slab);
+}
+
+static inline struct amdgpu_userq_fence *to_amdgpu_userq_fence(struct dma_fence *f)
+{
+	if (!f || f->ops != &amdgpu_userq_fence_ops)
+		return NULL;
+
+	return container_of(f, struct amdgpu_userq_fence, base);
+}
+
+static u64 amdgpu_userq_fence_read(struct amdgpu_userq_fence_driver *fence_drv)
+{
+	return le64_to_cpu(*fence_drv->cpu_addr);
+}
+
+static void
+amdgpu_userq_fence_write(struct amdgpu_userq_fence_driver *fence_drv,
+			 u64 seq)
+{
+	if (fence_drv->cpu_addr)
+		*fence_drv->cpu_addr = cpu_to_le64(seq);
+}
+
+int amdgpu_userq_fence_driver_alloc(struct amdgpu_device *adev,
+				    struct amdgpu_usermode_queue *userq)
+{
+	struct amdgpu_userq_fence_driver *fence_drv;
+	unsigned long flags;
+	int r;
+
+	fence_drv = kzalloc(sizeof(*fence_drv), GFP_KERNEL);
+	if (!fence_drv)
+		return -ENOMEM;
+
+	/* Acquire seq64 memory */
+	r = amdgpu_seq64_alloc(adev, &fence_drv->va, &fence_drv->gpu_addr,
+			       &fence_drv->cpu_addr);
+	if (r)
+		goto free_fence_drv;
+
+	memset(fence_drv->cpu_addr, 0, sizeof(u64));
+
+	kref_init(&fence_drv->refcount);
+	INIT_LIST_HEAD(&fence_drv->fences);
+	spin_lock_init(&fence_drv->fence_list_lock);
+
+	fence_drv->adev = adev;
+	fence_drv->context = dma_fence_context_alloc(1);
+	get_task_comm(fence_drv->timeline_name, current);
+
+	xa_lock_irqsave(&adev->userq_xa, flags);
+	r = xa_err(__xa_store(&adev->userq_xa, userq->doorbell_index,
+			      fence_drv, GFP_KERNEL));
+	xa_unlock_irqrestore(&adev->userq_xa, flags);
+	if (r)
+		goto free_seq64;
+
+	userq->fence_drv = fence_drv;
+
+	return 0;
+
+free_seq64:
+	amdgpu_seq64_free(adev, fence_drv->va);
+free_fence_drv:
+	kfree(fence_drv);
+
+	return r;
+}
+
+static void amdgpu_userq_walk_and_drop_fence_drv(struct xarray *xa)
+{
+	struct amdgpu_userq_fence_driver *fence_drv;
+	unsigned long index;
+
+	if (xa_empty(xa))
+		return;
+
+	xa_lock(xa);
+	xa_for_each(xa, index, fence_drv) {
+		__xa_erase(xa, index);
+		amdgpu_userq_fence_driver_put(fence_drv);
+	}
+
+	xa_unlock(xa);
+}
+
+void
+amdgpu_userq_fence_driver_free(struct amdgpu_usermode_queue *userq)
+{
+	amdgpu_userq_walk_and_drop_fence_drv(&userq->fence_drv_xa);
+	xa_destroy(&userq->fence_drv_xa);
+	/* Drop the fence_drv reference held by user queue */
+	amdgpu_userq_fence_driver_put(userq->fence_drv);
+}
+
+void amdgpu_userq_fence_driver_process(struct amdgpu_userq_fence_driver *fence_drv)
+{
+	struct amdgpu_userq_fence *userq_fence, *tmp;
+	struct dma_fence *fence;
+	unsigned long flags;
+	u64 rptr;
+	int i;
+
+	if (!fence_drv)
+		return;
+
+	spin_lock_irqsave(&fence_drv->fence_list_lock, flags);
+	rptr = amdgpu_userq_fence_read(fence_drv);
+
+	list_for_each_entry_safe(userq_fence, tmp, &fence_drv->fences, link) {
+		fence = &userq_fence->base;
+
+		if (rptr < fence->seqno)
+			break;
+
+		dma_fence_signal(fence);
+
+		for (i = 0; i < userq_fence->fence_drv_array_count; i++)
+			amdgpu_userq_fence_driver_put(userq_fence->fence_drv_array[i]);
+
+		list_del(&userq_fence->link);
+		dma_fence_put(fence);
+	}
+	spin_unlock_irqrestore(&fence_drv->fence_list_lock, flags);
+}
+
+void amdgpu_userq_fence_driver_destroy(struct kref *ref)
+{
+	struct amdgpu_userq_fence_driver *fence_drv = container_of(ref,
+					 struct amdgpu_userq_fence_driver,
+					 refcount);
+	struct amdgpu_userq_fence_driver *xa_fence_drv;
+	struct amdgpu_device *adev = fence_drv->adev;
+	struct amdgpu_userq_fence *fence, *tmp;
+	struct xarray *xa = &adev->userq_xa;
+	unsigned long index, flags;
+	struct dma_fence *f;
+
+	spin_lock_irqsave(&fence_drv->fence_list_lock, flags);
+	list_for_each_entry_safe(fence, tmp, &fence_drv->fences, link) {
+		f = &fence->base;
+
+		if (!dma_fence_is_signaled(f)) {
+			dma_fence_set_error(f, -ECANCELED);
+			dma_fence_signal(f);
+		}
+
+		list_del(&fence->link);
+		dma_fence_put(f);
+	}
+	spin_unlock_irqrestore(&fence_drv->fence_list_lock, flags);
+
+	xa_lock_irqsave(xa, flags);
+	xa_for_each(xa, index, xa_fence_drv)
+		if (xa_fence_drv == fence_drv)
+			__xa_erase(xa, index);
+	xa_unlock_irqrestore(xa, flags);
+
+	/* Free seq64 memory */
+	amdgpu_seq64_free(adev, fence_drv->va);
+	kfree(fence_drv);
+}
+
+void amdgpu_userq_fence_driver_get(struct amdgpu_userq_fence_driver *fence_drv)
+{
+	kref_get(&fence_drv->refcount);
+}
+
+void amdgpu_userq_fence_driver_put(struct amdgpu_userq_fence_driver *fence_drv)
+{
+	kref_put(&fence_drv->refcount, amdgpu_userq_fence_driver_destroy);
+}
+
+static int amdgpu_userq_fence_alloc(struct amdgpu_userq_fence **userq_fence)
+{
+	*userq_fence = kmem_cache_alloc(amdgpu_userq_fence_slab, GFP_ATOMIC);
+	return *userq_fence ? 0 : -ENOMEM;
+}
+
+static int amdgpu_userq_fence_create(struct amdgpu_usermode_queue *userq,
+				     struct amdgpu_userq_fence *userq_fence,
+				     u64 seq, struct dma_fence **f)
+{
+	struct amdgpu_userq_fence_driver *fence_drv;
+	struct dma_fence *fence;
+	unsigned long flags;
+
+	fence_drv = userq->fence_drv;
+	if (!fence_drv)
+		return -EINVAL;
+
+	spin_lock_init(&userq_fence->lock);
+	INIT_LIST_HEAD(&userq_fence->link);
+	fence = &userq_fence->base;
+	userq_fence->fence_drv = fence_drv;
+
+	dma_fence_init64(fence, &amdgpu_userq_fence_ops, &userq_fence->lock,
+			 fence_drv->context, seq);
+
+	amdgpu_userq_fence_driver_get(fence_drv);
+	dma_fence_get(fence);
+
+	if (!xa_empty(&userq->fence_drv_xa)) {
+		struct amdgpu_userq_fence_driver *stored_fence_drv;
+		unsigned long index, count = 0;
+		int i = 0;
+
+		xa_lock(&userq->fence_drv_xa);
+		xa_for_each(&userq->fence_drv_xa, index, stored_fence_drv)
+			count++;
+
+		userq_fence->fence_drv_array =
+			kvmalloc_array(count,
+				       sizeof(struct amdgpu_userq_fence_driver *),
+				       GFP_ATOMIC);
+
+		if (userq_fence->fence_drv_array) {
+			xa_for_each(&userq->fence_drv_xa, index, stored_fence_drv) {
+				userq_fence->fence_drv_array[i] = stored_fence_drv;
+				__xa_erase(&userq->fence_drv_xa, index);
+				i++;
+			}
+		}
+
+		userq_fence->fence_drv_array_count = i;
+		xa_unlock(&userq->fence_drv_xa);
+	} else {
+		userq_fence->fence_drv_array = NULL;
+		userq_fence->fence_drv_array_count = 0;
+	}
+
+	/* Check if hardware has already processed the job */
+	spin_lock_irqsave(&fence_drv->fence_list_lock, flags);
+	if (!dma_fence_is_signaled(fence))
+		list_add_tail(&userq_fence->link, &fence_drv->fences);
+	else
+		dma_fence_put(fence);
+
+	spin_unlock_irqrestore(&fence_drv->fence_list_lock, flags);
+
+	*f = fence;
+
+	return 0;
+}
+
+static const char *amdgpu_userq_fence_get_driver_name(struct dma_fence *f)
+{
+	return "amdgpu_userq_fence";
+}
+
+static const char *amdgpu_userq_fence_get_timeline_name(struct dma_fence *f)
+{
+	struct amdgpu_userq_fence *fence = to_amdgpu_userq_fence(f);
+
+	return fence->fence_drv->timeline_name;
+}
+
+static bool amdgpu_userq_fence_signaled(struct dma_fence *f)
+{
+	struct amdgpu_userq_fence *fence = to_amdgpu_userq_fence(f);
+	struct amdgpu_userq_fence_driver *fence_drv = fence->fence_drv;
+	u64 rptr, wptr;
+
+	rptr = amdgpu_userq_fence_read(fence_drv);
+	wptr = fence->base.seqno;
+
+	if (rptr >= wptr)
+		return true;
+
+	return false;
+}
+
+static void amdgpu_userq_fence_free(struct rcu_head *rcu)
+{
+	struct dma_fence *fence = container_of(rcu, struct dma_fence, rcu);
+	struct amdgpu_userq_fence *userq_fence = to_amdgpu_userq_fence(fence);
+	struct amdgpu_userq_fence_driver *fence_drv = userq_fence->fence_drv;
+
+	/* Release the fence driver reference */
+	amdgpu_userq_fence_driver_put(fence_drv);
+
+	kvfree(userq_fence->fence_drv_array);
+	kmem_cache_free(amdgpu_userq_fence_slab, userq_fence);
+}
+
+static void amdgpu_userq_fence_release(struct dma_fence *f)
+{
+	call_rcu(&f->rcu, amdgpu_userq_fence_free);
+}
+
+static const struct dma_fence_ops amdgpu_userq_fence_ops = {
+	.get_driver_name = amdgpu_userq_fence_get_driver_name,
+	.get_timeline_name = amdgpu_userq_fence_get_timeline_name,
+	.signaled = amdgpu_userq_fence_signaled,
+	.release = amdgpu_userq_fence_release,
+};
+
+/**
+ * amdgpu_userq_fence_read_wptr - Read the userq wptr value
+ *
+ * @queue: user mode queue structure pointer
+ * @wptr: write pointer value
+ *
+ * Read the wptr value from userq's MQD. The userq signal IOCTL
+ * creates a dma_fence for the shared buffers that expects the
+ * RPTR value written to seq64 memory >= WPTR.
+ *
+ * Returns wptr value on success, error on failure.
+ */
+static int amdgpu_userq_fence_read_wptr(struct amdgpu_usermode_queue *queue,
+					u64 *wptr)
+{
+	struct amdgpu_bo_va_mapping *mapping;
+	struct amdgpu_bo *bo;
+	u64 addr, *ptr;
+	int r;
+
+	r = amdgpu_bo_reserve(queue->vm->root.bo, false);
+	if (r)
+		return r;
+
+	addr = queue->userq_prop->wptr_gpu_addr;
+	addr &= AMDGPU_GMC_HOLE_MASK;
+
+	mapping = amdgpu_vm_bo_lookup_mapping(queue->vm, addr >> PAGE_SHIFT);
+	if (!mapping) {
+		amdgpu_bo_unreserve(queue->vm->root.bo);
+		DRM_ERROR("Failed to lookup amdgpu_bo_va_mapping\n");
+		return -EINVAL;
+	}
+
+	bo = amdgpu_bo_ref(mapping->bo_va->base.bo);
+	amdgpu_bo_unreserve(queue->vm->root.bo);
+	r = amdgpu_bo_reserve(bo, true);
+	if (r) {
+		amdgpu_bo_unref(&bo);
+		DRM_ERROR("Failed to reserve userqueue wptr bo");
+		return r;
+	}
+
+	r = amdgpu_bo_kmap(bo, (void **)&ptr);
+	if (r) {
+		DRM_ERROR("Failed mapping the userqueue wptr bo");
+		goto map_error;
+	}
+
+	*wptr = le64_to_cpu(*ptr);
+
+	amdgpu_bo_kunmap(bo);
+	amdgpu_bo_unreserve(bo);
+	amdgpu_bo_unref(&bo);
+
+	return 0;
+
+map_error:
+	amdgpu_bo_unreserve(bo);
+	amdgpu_bo_unref(&bo);
+
+	return r;
+}
+
+static void amdgpu_userq_fence_cleanup(struct dma_fence *fence)
+{
+	dma_fence_put(fence);
+}
+
+static void
+amdgpu_userq_fence_driver_set_error(struct amdgpu_userq_fence *fence,
+				    int error)
+{
+	struct amdgpu_userq_fence_driver *fence_drv = fence->fence_drv;
+	unsigned long flags;
+	struct dma_fence *f;
+
+	spin_lock_irqsave(&fence_drv->fence_list_lock, flags);
+
+	f = rcu_dereference_protected(&fence->base,
+				      lockdep_is_held(&fence_drv->fence_list_lock));
+	if (f && !dma_fence_is_signaled_locked(f))
+		dma_fence_set_error(f, error);
+	spin_unlock_irqrestore(&fence_drv->fence_list_lock, flags);
+}
+
+void
+amdgpu_userq_fence_driver_force_completion(struct amdgpu_usermode_queue *userq)
+{
+	struct dma_fence *f = userq->last_fence;
+
+	if (f) {
+		struct amdgpu_userq_fence *fence = to_amdgpu_userq_fence(f);
+		struct amdgpu_userq_fence_driver *fence_drv = fence->fence_drv;
+		u64 wptr = fence->base.seqno;
+
+		amdgpu_userq_fence_driver_set_error(fence, -ECANCELED);
+		amdgpu_userq_fence_write(fence_drv, wptr);
+		amdgpu_userq_fence_driver_process(fence_drv);
+
+	}
+}
+
+int amdgpu_userq_signal_ioctl(struct drm_device *dev, void *data,
+			      struct drm_file *filp)
+{
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
+	struct amdgpu_userq_mgr *userq_mgr = &fpriv->userq_mgr;
+	struct drm_amdgpu_userq_signal *args = data;
+	struct drm_gem_object **gobj_write = NULL;
+	struct drm_gem_object **gobj_read = NULL;
+	struct amdgpu_usermode_queue *queue;
+	struct amdgpu_userq_fence *userq_fence;
+	struct drm_syncobj **syncobj = NULL;
+	u32 *bo_handles_write, num_write_bo_handles;
+	u32 *syncobj_handles, num_syncobj_handles;
+	u32 *bo_handles_read, num_read_bo_handles;
+	int r, i, entry, rentry, wentry;
+	struct dma_fence *fence;
+	struct drm_exec exec;
+	u64 wptr;
+
+	num_syncobj_handles = args->num_syncobj_handles;
+	syncobj_handles = memdup_user(u64_to_user_ptr(args->syncobj_handles),
+				      size_mul(sizeof(u32), num_syncobj_handles));
+	if (IS_ERR(syncobj_handles))
+		return PTR_ERR(syncobj_handles);
+
+	/* Array of pointers to the looked up syncobjs */
+	syncobj = kmalloc_array(num_syncobj_handles, sizeof(*syncobj), GFP_KERNEL);
+	if (!syncobj) {
+		r = -ENOMEM;
+		goto free_syncobj_handles;
+	}
+
+	for (entry = 0; entry < num_syncobj_handles; entry++) {
+		syncobj[entry] = drm_syncobj_find(filp, syncobj_handles[entry]);
+		if (!syncobj[entry]) {
+			r = -ENOENT;
+			goto free_syncobj;
+		}
+	}
+
+	num_read_bo_handles = args->num_bo_read_handles;
+	bo_handles_read = memdup_user(u64_to_user_ptr(args->bo_read_handles),
+				      sizeof(u32) * num_read_bo_handles);
+	if (IS_ERR(bo_handles_read)) {
+		r = PTR_ERR(bo_handles_read);
+		goto free_syncobj;
+	}
+
+	/* Array of pointers to the GEM read objects */
+	gobj_read = kmalloc_array(num_read_bo_handles, sizeof(*gobj_read), GFP_KERNEL);
+	if (!gobj_read) {
+		r = -ENOMEM;
+		goto free_bo_handles_read;
+	}
+
+	for (rentry = 0; rentry < num_read_bo_handles; rentry++) {
+		gobj_read[rentry] = drm_gem_object_lookup(filp, bo_handles_read[rentry]);
+		if (!gobj_read[rentry]) {
+			r = -ENOENT;
+			goto put_gobj_read;
+		}
+	}
+
+	num_write_bo_handles = args->num_bo_write_handles;
+	bo_handles_write = memdup_user(u64_to_user_ptr(args->bo_write_handles),
+				       sizeof(u32) * num_write_bo_handles);
+	if (IS_ERR(bo_handles_write)) {
+		r = PTR_ERR(bo_handles_write);
+		goto put_gobj_read;
+	}
+
+	/* Array of pointers to the GEM write objects */
+	gobj_write = kmalloc_array(num_write_bo_handles, sizeof(*gobj_write), GFP_KERNEL);
+	if (!gobj_write) {
+		r = -ENOMEM;
+		goto free_bo_handles_write;
+	}
+
+	for (wentry = 0; wentry < num_write_bo_handles; wentry++) {
+		gobj_write[wentry] = drm_gem_object_lookup(filp, bo_handles_write[wentry]);
+		if (!gobj_write[wentry]) {
+			r = -ENOENT;
+			goto put_gobj_write;
+		}
+	}
+
+	/* Retrieve the user queue */
+	queue = xa_load(&userq_mgr->userq_mgr_xa, args->queue_id);
+	if (!queue) {
+		r = -ENOENT;
+		goto put_gobj_write;
+	}
+
+	r = amdgpu_userq_fence_read_wptr(queue, &wptr);
+	if (r)
+		goto put_gobj_write;
+
+	r = amdgpu_userq_fence_alloc(&userq_fence);
+	if (r)
+		goto put_gobj_write;
+
+	/* We are here means UQ is active, make sure the eviction fence is valid */
+	amdgpu_userq_ensure_ev_fence(&fpriv->userq_mgr, &fpriv->evf_mgr);
+
+	/* Create a new fence */
+	r = amdgpu_userq_fence_create(queue, userq_fence, wptr, &fence);
+	if (r) {
+		mutex_unlock(&userq_mgr->userq_mutex);
+		kmem_cache_free(amdgpu_userq_fence_slab, userq_fence);
+		goto put_gobj_write;
+	}
+
+	dma_fence_put(queue->last_fence);
+	queue->last_fence = dma_fence_get(fence);
+	mutex_unlock(&userq_mgr->userq_mutex);
+
+	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT,
+		      (num_read_bo_handles + num_write_bo_handles));
+
+	/* Lock all BOs with retry handling */
+	drm_exec_until_all_locked(&exec) {
+		r = drm_exec_prepare_array(&exec, gobj_read, num_read_bo_handles, 1);
+		drm_exec_retry_on_contention(&exec);
+		if (r) {
+			amdgpu_userq_fence_cleanup(fence);
+			goto exec_fini;
+		}
+
+		r = drm_exec_prepare_array(&exec, gobj_write, num_write_bo_handles, 1);
+		drm_exec_retry_on_contention(&exec);
+		if (r) {
+			amdgpu_userq_fence_cleanup(fence);
+			goto exec_fini;
+		}
+	}
+
+	for (i = 0; i < num_read_bo_handles; i++) {
+		if (!gobj_read || !gobj_read[i]->resv)
+			continue;
+
+		dma_resv_add_fence(gobj_read[i]->resv, fence,
+				   DMA_RESV_USAGE_READ);
+	}
+
+	for (i = 0; i < num_write_bo_handles; i++) {
+		if (!gobj_write || !gobj_write[i]->resv)
+			continue;
+
+		dma_resv_add_fence(gobj_write[i]->resv, fence,
+				   DMA_RESV_USAGE_WRITE);
+	}
+
+	/* Add the created fence to syncobj/BO's */
+	for (i = 0; i < num_syncobj_handles; i++)
+		drm_syncobj_replace_fence(syncobj[i], fence);
+
+	/* drop the reference acquired in fence creation function */
+	dma_fence_put(fence);
+
+exec_fini:
+	drm_exec_fini(&exec);
+put_gobj_write:
+	while (wentry-- > 0)
+		drm_gem_object_put(gobj_write[wentry]);
+	kfree(gobj_write);
+free_bo_handles_write:
+	kfree(bo_handles_write);
+put_gobj_read:
+	while (rentry-- > 0)
+		drm_gem_object_put(gobj_read[rentry]);
+	kfree(gobj_read);
+free_bo_handles_read:
+	kfree(bo_handles_read);
+free_syncobj:
+	while (entry-- > 0)
+		if (syncobj[entry])
+			drm_syncobj_put(syncobj[entry]);
+	kfree(syncobj);
+free_syncobj_handles:
+	kfree(syncobj_handles);
+
+	return r;
+}
+
+int amdgpu_userq_wait_ioctl(struct drm_device *dev, void *data,
+			    struct drm_file *filp)
+{
+	u32 *syncobj_handles, *timeline_points, *timeline_handles, *bo_handles_read, *bo_handles_write;
+	u32 num_syncobj, num_read_bo_handles, num_write_bo_handles;
+	struct drm_amdgpu_userq_fence_info *fence_info = NULL;
+	struct drm_amdgpu_userq_wait *wait_info = data;
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
+	struct amdgpu_userq_mgr *userq_mgr = &fpriv->userq_mgr;
+	struct amdgpu_usermode_queue *waitq;
+	struct drm_gem_object **gobj_write;
+	struct drm_gem_object **gobj_read;
+	struct dma_fence **fences = NULL;
+	u16 num_points, num_fences = 0;
+	int r, i, rentry, wentry, cnt;
+	struct drm_exec exec;
+
+	num_read_bo_handles = wait_info->num_bo_read_handles;
+	bo_handles_read = memdup_user(u64_to_user_ptr(wait_info->bo_read_handles),
+				      size_mul(sizeof(u32), num_read_bo_handles));
+	if (IS_ERR(bo_handles_read))
+		return PTR_ERR(bo_handles_read);
+
+	num_write_bo_handles = wait_info->num_bo_write_handles;
+	bo_handles_write = memdup_user(u64_to_user_ptr(wait_info->bo_write_handles),
+				       size_mul(sizeof(u32), num_write_bo_handles));
+	if (IS_ERR(bo_handles_write)) {
+		r = PTR_ERR(bo_handles_write);
+		goto free_bo_handles_read;
+	}
+
+	num_syncobj = wait_info->num_syncobj_handles;
+	syncobj_handles = memdup_user(u64_to_user_ptr(wait_info->syncobj_handles),
+				      size_mul(sizeof(u32), num_syncobj));
+	if (IS_ERR(syncobj_handles)) {
+		r = PTR_ERR(syncobj_handles);
+		goto free_bo_handles_write;
+	}
+
+	num_points = wait_info->num_syncobj_timeline_handles;
+	timeline_handles = memdup_user(u64_to_user_ptr(wait_info->syncobj_timeline_handles),
+				       sizeof(u32) * num_points);
+	if (IS_ERR(timeline_handles)) {
+		r = PTR_ERR(timeline_handles);
+		goto free_syncobj_handles;
+	}
+
+	timeline_points = memdup_user(u64_to_user_ptr(wait_info->syncobj_timeline_points),
+				      sizeof(u32) * num_points);
+	if (IS_ERR(timeline_points)) {
+		r = PTR_ERR(timeline_points);
+		goto free_timeline_handles;
+	}
+
+	gobj_read = kmalloc_array(num_read_bo_handles, sizeof(*gobj_read), GFP_KERNEL);
+	if (!gobj_read) {
+		r = -ENOMEM;
+		goto free_timeline_points;
+	}
+
+	for (rentry = 0; rentry < num_read_bo_handles; rentry++) {
+		gobj_read[rentry] = drm_gem_object_lookup(filp, bo_handles_read[rentry]);
+		if (!gobj_read[rentry]) {
+			r = -ENOENT;
+			goto put_gobj_read;
+		}
+	}
+
+	gobj_write = kmalloc_array(num_write_bo_handles, sizeof(*gobj_write), GFP_KERNEL);
+	if (!gobj_write) {
+		r = -ENOMEM;
+		goto put_gobj_read;
+	}
+
+	for (wentry = 0; wentry < num_write_bo_handles; wentry++) {
+		gobj_write[wentry] = drm_gem_object_lookup(filp, bo_handles_write[wentry]);
+		if (!gobj_write[wentry]) {
+			r = -ENOENT;
+			goto put_gobj_write;
+		}
+	}
+
+	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT,
+		      (num_read_bo_handles + num_write_bo_handles));
+
+	/* Lock all BOs with retry handling */
+	drm_exec_until_all_locked(&exec) {
+		r = drm_exec_prepare_array(&exec, gobj_read, num_read_bo_handles, 1);
+		drm_exec_retry_on_contention(&exec);
+		if (r) {
+			drm_exec_fini(&exec);
+			goto put_gobj_write;
+		}
+
+		r = drm_exec_prepare_array(&exec, gobj_write, num_write_bo_handles, 1);
+		drm_exec_retry_on_contention(&exec);
+		if (r) {
+			drm_exec_fini(&exec);
+			goto put_gobj_write;
+		}
+	}
+
+	if (!wait_info->num_fences) {
+		if (num_points) {
+			struct dma_fence_unwrap iter;
+			struct dma_fence *fence;
+			struct dma_fence *f;
+
+			for (i = 0; i < num_points; i++) {
+				r = drm_syncobj_find_fence(filp, timeline_handles[i],
+							   timeline_points[i],
+							   DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT,
+							   &fence);
+				if (r)
+					goto exec_fini;
+
+				dma_fence_unwrap_for_each(f, &iter, fence)
+					num_fences++;
+
+				dma_fence_put(fence);
+			}
+		}
+
+		/* Count syncobj's fence */
+		for (i = 0; i < num_syncobj; i++) {
+			struct dma_fence *fence;
+
+			r = drm_syncobj_find_fence(filp, syncobj_handles[i],
+						   0,
+						   DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT,
+						   &fence);
+			if (r)
+				goto exec_fini;
+
+			num_fences++;
+			dma_fence_put(fence);
+		}
+
+		/* Count GEM objects fence */
+		for (i = 0; i < num_read_bo_handles; i++) {
+			struct dma_resv_iter resv_cursor;
+			struct dma_fence *fence;
+
+			dma_resv_for_each_fence(&resv_cursor, gobj_read[i]->resv,
+						DMA_RESV_USAGE_READ, fence)
+				num_fences++;
+		}
+
+		for (i = 0; i < num_write_bo_handles; i++) {
+			struct dma_resv_iter resv_cursor;
+			struct dma_fence *fence;
+
+			dma_resv_for_each_fence(&resv_cursor, gobj_write[i]->resv,
+						DMA_RESV_USAGE_WRITE, fence)
+				num_fences++;
+		}
+
+		/*
+		 * Passing num_fences = 0 means that userspace doesn't want to
+		 * retrieve userq_fence_info. If num_fences = 0 we skip filling
+		 * userq_fence_info and return the actual number of fences on
+		 * args->num_fences.
+		 */
+		wait_info->num_fences = num_fences;
+	} else {
+		/* Array of fence info */
+		fence_info = kmalloc_array(wait_info->num_fences, sizeof(*fence_info), GFP_KERNEL);
+		if (!fence_info) {
+			r = -ENOMEM;
+			goto exec_fini;
+		}
+
+		/* Array of fences */
+		fences = kmalloc_array(wait_info->num_fences, sizeof(*fences), GFP_KERNEL);
+		if (!fences) {
+			r = -ENOMEM;
+			goto free_fence_info;
+		}
+
+		/* Retrieve GEM read objects fence */
+		for (i = 0; i < num_read_bo_handles; i++) {
+			struct dma_resv_iter resv_cursor;
+			struct dma_fence *fence;
+
+			dma_resv_for_each_fence(&resv_cursor, gobj_read[i]->resv,
+						DMA_RESV_USAGE_READ, fence) {
+				if (WARN_ON_ONCE(num_fences >= wait_info->num_fences)) {
+					r = -EINVAL;
+					goto free_fences;
+				}
+
+				fences[num_fences++] = fence;
+				dma_fence_get(fence);
+			}
+		}
+
+		/* Retrieve GEM write objects fence */
+		for (i = 0; i < num_write_bo_handles; i++) {
+			struct dma_resv_iter resv_cursor;
+			struct dma_fence *fence;
+
+			dma_resv_for_each_fence(&resv_cursor, gobj_write[i]->resv,
+						DMA_RESV_USAGE_WRITE, fence) {
+				if (WARN_ON_ONCE(num_fences >= wait_info->num_fences)) {
+					r = -EINVAL;
+					goto free_fences;
+				}
+
+				fences[num_fences++] = fence;
+				dma_fence_get(fence);
+			}
+		}
+
+		if (num_points) {
+			struct dma_fence_unwrap iter;
+			struct dma_fence *fence;
+			struct dma_fence *f;
+
+			for (i = 0; i < num_points; i++) {
+				r = drm_syncobj_find_fence(filp, timeline_handles[i],
+							   timeline_points[i],
+							   DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT,
+							   &fence);
+				if (r)
+					goto free_fences;
+
+				dma_fence_unwrap_for_each(f, &iter, fence) {
+					if (WARN_ON_ONCE(num_fences >= wait_info->num_fences)) {
+						r = -EINVAL;
+						goto free_fences;
+					}
+
+					dma_fence_get(f);
+					fences[num_fences++] = f;
+				}
+
+				dma_fence_put(fence);
+			}
+		}
+
+		/* Retrieve syncobj's fence */
+		for (i = 0; i < num_syncobj; i++) {
+			struct dma_fence *fence;
+
+			r = drm_syncobj_find_fence(filp, syncobj_handles[i],
+						   0,
+						   DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT,
+						   &fence);
+			if (r)
+				goto free_fences;
+
+			if (WARN_ON_ONCE(num_fences >= wait_info->num_fences)) {
+				r = -EINVAL;
+				goto free_fences;
+			}
+
+			fences[num_fences++] = fence;
+		}
+
+		/*
+		 * Keep only the latest fences to reduce the number of values
+		 * given back to userspace.
+		 */
+		num_fences = dma_fence_dedup_array(fences, num_fences);
+
+		waitq = xa_load(&userq_mgr->userq_mgr_xa, wait_info->waitq_id);
+		if (!waitq) {
+			r = -EINVAL;
+			goto free_fences;
+		}
+
+		for (i = 0, cnt = 0; i < num_fences; i++) {
+			struct amdgpu_userq_fence_driver *fence_drv;
+			struct amdgpu_userq_fence *userq_fence;
+			u32 index;
+
+			userq_fence = to_amdgpu_userq_fence(fences[i]);
+			if (!userq_fence) {
+				/*
+				 * Just waiting on other driver fences should
+				 * be good for now
+				 */
+				r = dma_fence_wait(fences[i], true);
+				if (r) {
+					dma_fence_put(fences[i]);
+					goto free_fences;
+				}
+
+				dma_fence_put(fences[i]);
+				continue;
+			}
+
+			fence_drv = userq_fence->fence_drv;
+			/*
+			 * We need to make sure the user queue release their reference
+			 * to the fence drivers at some point before queue destruction.
+			 * Otherwise, we would gather those references until we don't
+			 * have any more space left and crash.
+			 */
+			r = xa_alloc(&waitq->fence_drv_xa, &index, fence_drv,
+				     xa_limit_32b, GFP_KERNEL);
+			if (r)
+				goto free_fences;
+
+			amdgpu_userq_fence_driver_get(fence_drv);
+
+			/* Store drm syncobj's gpu va address and value */
+			fence_info[cnt].va = fence_drv->va;
+			fence_info[cnt].value = fences[i]->seqno;
+
+			dma_fence_put(fences[i]);
+			/* Increment the actual userq fence count */
+			cnt++;
+		}
+
+		wait_info->num_fences = cnt;
+		/* Copy userq fence info to user space */
+		if (copy_to_user(u64_to_user_ptr(wait_info->out_fences),
+				 fence_info, wait_info->num_fences * sizeof(*fence_info))) {
+			r = -EFAULT;
+			goto free_fences;
+		}
+
+		kfree(fences);
+		kfree(fence_info);
+	}
+
+	drm_exec_fini(&exec);
+	for (i = 0; i < num_read_bo_handles; i++)
+		drm_gem_object_put(gobj_read[i]);
+	kfree(gobj_read);
+
+	for (i = 0; i < num_write_bo_handles; i++)
+		drm_gem_object_put(gobj_write[i]);
+	kfree(gobj_write);
+
+	kfree(timeline_points);
+	kfree(timeline_handles);
+	kfree(syncobj_handles);
+	kfree(bo_handles_write);
+	kfree(bo_handles_read);
+
+	return 0;
+
+free_fences:
+	while (num_fences-- > 0)
+		dma_fence_put(fences[num_fences]);
+	kfree(fences);
+free_fence_info:
+	kfree(fence_info);
+exec_fini:
+	drm_exec_fini(&exec);
+put_gobj_write:
+	while (wentry-- > 0)
+		drm_gem_object_put(gobj_write[wentry]);
+	kfree(gobj_write);
+put_gobj_read:
+	while (rentry-- > 0)
+		drm_gem_object_put(gobj_read[rentry]);
+	kfree(gobj_read);
+free_timeline_points:
+	kfree(timeline_points);
+free_timeline_handles:
+	kfree(timeline_handles);
+free_syncobj_handles:
+	kfree(syncobj_handles);
+free_bo_handles_write:
+	kfree(bo_handles_write);
+free_bo_handles_read:
+	kfree(bo_handles_read);
+
+	return r;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.h
new file mode 100644
index 000000000000..d76add2afc77
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.h
@@ -0,0 +1,77 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __AMDGPU_USERQ_FENCE_H__
+#define __AMDGPU_USERQ_FENCE_H__
+
+#include <linux/types.h>
+
+#include "amdgpu_userq.h"
+
+struct amdgpu_userq_fence {
+	struct dma_fence base;
+	/*
+	 * This lock is necessary to synchronize the
+	 * userqueue dma fence operations.
+	 */
+	spinlock_t lock;
+	struct list_head link;
+	unsigned long fence_drv_array_count;
+	struct amdgpu_userq_fence_driver *fence_drv;
+	struct amdgpu_userq_fence_driver **fence_drv_array;
+};
+
+struct amdgpu_userq_fence_driver {
+	struct kref refcount;
+	u64 va;
+	u64 gpu_addr;
+	u64 *cpu_addr;
+	u64 context;
+	/*
+	 * This lock is necesaary to synchronize the access
+	 * to the fences list by the fence driver.
+	 */
+	spinlock_t fence_list_lock;
+	struct list_head fences;
+	struct amdgpu_device *adev;
+	char timeline_name[TASK_COMM_LEN];
+};
+
+int amdgpu_userq_fence_slab_init(void);
+void amdgpu_userq_fence_slab_fini(void);
+
+void amdgpu_userq_fence_driver_get(struct amdgpu_userq_fence_driver *fence_drv);
+void amdgpu_userq_fence_driver_put(struct amdgpu_userq_fence_driver *fence_drv);
+int amdgpu_userq_fence_driver_alloc(struct amdgpu_device *adev,
+				    struct amdgpu_usermode_queue *userq);
+void amdgpu_userq_fence_driver_free(struct amdgpu_usermode_queue *userq);
+void amdgpu_userq_fence_driver_process(struct amdgpu_userq_fence_driver *fence_drv);
+void amdgpu_userq_fence_driver_force_completion(struct amdgpu_usermode_queue *userq);
+void amdgpu_userq_fence_driver_destroy(struct kref *ref);
+int amdgpu_userq_signal_ioctl(struct drm_device *dev, void *data,
+			      struct drm_file *filp);
+int amdgpu_userq_wait_ioctl(struct drm_device *dev, void *data,
+			    struct drm_file *filp);
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_utils.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_utils.h
new file mode 100644
index 000000000000..1e40ca3b1584
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_utils.h
@@ -0,0 +1,91 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef AMDGPU_UTILS_H_
+#define AMDGPU_UTILS_H_
+
+/* ---------- Generic 2‑bit capability attribute encoding ----------
+ * 00 INVALID, 01 RO, 10 WO, 11 RW
+ */
+enum amdgpu_cap_attr {
+	AMDGPU_CAP_ATTR_INVALID = 0,
+	AMDGPU_CAP_ATTR_RO      = 1 << 0,
+	AMDGPU_CAP_ATTR_WO      = 1 << 1,
+	AMDGPU_CAP_ATTR_RW      = (AMDGPU_CAP_ATTR_RO | AMDGPU_CAP_ATTR_WO),
+};
+
+#define AMDGPU_CAP_ATTR_BITS 2
+#define AMDGPU_CAP_ATTR_MAX  ((1U << AMDGPU_CAP_ATTR_BITS) - 1)
+
+/* Internal helper to build helpers for a given enum NAME */
+#define DECLARE_ATTR_CAP_CLASS_HELPERS(NAME)							\
+enum { NAME##_BITMAP_BITS = NAME##_COUNT * AMDGPU_CAP_ATTR_BITS };				\
+struct NAME##_caps {										\
+	DECLARE_BITMAP(bmap, NAME##_BITMAP_BITS);						\
+};												\
+static inline unsigned int NAME##_ATTR_START(enum NAME##_cap_id cap)				\
+{ return (unsigned int)cap * AMDGPU_CAP_ATTR_BITS; }						\
+static inline void NAME##_attr_init(struct NAME##_caps *c)					\
+{ if (c) bitmap_zero(c->bmap, NAME##_BITMAP_BITS); }						\
+static inline int NAME##_attr_set(struct NAME##_caps *c,					\
+				  enum NAME##_cap_id cap, enum amdgpu_cap_attr attr)		\
+{												\
+	if (!c)											\
+		return -EINVAL;									\
+	if (cap >= NAME##_COUNT)								\
+		return -EINVAL;									\
+	if ((unsigned int)attr > AMDGPU_CAP_ATTR_MAX)						\
+		return -EINVAL;									\
+	bitmap_write(c->bmap, (unsigned long)attr,						\
+			NAME##_ATTR_START(cap), AMDGPU_CAP_ATTR_BITS);				\
+	return 0;										\
+}												\
+static inline int NAME##_attr_get(const struct NAME##_caps *c,					\
+				  enum NAME##_cap_id cap, enum amdgpu_cap_attr *out)		\
+{												\
+	unsigned long v;									\
+	if (!c || !out)										\
+		return -EINVAL;									\
+	if (cap >= NAME##_COUNT)								\
+		return -EINVAL;									\
+	v = bitmap_read(c->bmap, NAME##_ATTR_START(cap), AMDGPU_CAP_ATTR_BITS);			\
+	*out = (enum amdgpu_cap_attr)v;								\
+	return 0;										\
+}												\
+static inline bool NAME##_cap_is_ro(const struct NAME##_caps *c, enum NAME##_cap_id id)		\
+{ enum amdgpu_cap_attr a; return !NAME##_attr_get(c, id, &a) && a == AMDGPU_CAP_ATTR_RO; }	\
+static inline bool NAME##_cap_is_wo(const struct NAME##_caps *c, enum NAME##_cap_id id)		\
+{ enum amdgpu_cap_attr a; return !NAME##_attr_get(c, id, &a) && a == AMDGPU_CAP_ATTR_WO; }	\
+static inline bool NAME##_cap_is_rw(const struct NAME##_caps *c, enum NAME##_cap_id id)		\
+{ enum amdgpu_cap_attr a; return !NAME##_attr_get(c, id, &a) && a == AMDGPU_CAP_ATTR_RW; }
+
+/* Element expander for enum creation */
+#define _CAP_ENUM_ELEM(x) x,
+
+/* Public macro: declare enum + helpers from an X‑macro list */
+#define DECLARE_ATTR_CAP_CLASS(NAME, LIST_MACRO)						\
+	enum NAME##_cap_id { LIST_MACRO(_CAP_ENUM_ELEM) NAME##_COUNT };				\
+	DECLARE_ATTR_CAP_CLASS_HELPERS(NAME)
+
+#endif /* AMDGPU_UTILS_H_ */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
index 6887109abb13..5c38f0d30c87 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
@@ -96,16 +96,16 @@
  */
 struct amdgpu_uvd_cs_ctx {
 	struct amdgpu_cs_parser *parser;
-	unsigned reg, count;
-	unsigned data0, data1;
-	unsigned idx;
+	unsigned int reg, count;
+	unsigned int data0, data1;
+	unsigned int idx;
 	struct amdgpu_ib *ib;
 
 	/* does the IB has a msg command */
 	bool has_msg_cmd;
 
 	/* minimum buffer sizes */
-	unsigned *buf_sizes;
+	unsigned int *buf_sizes;
 };
 
 #ifdef CONFIG_DRM_AMDGPU_SI
@@ -186,7 +186,7 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev)
 	unsigned long bo_size;
 	const char *fw_name;
 	const struct common_firmware_header *hdr;
-	unsigned family_id;
+	unsigned int family_id;
 	int i, j, r;
 
 	INIT_DELAYED_WORK(&adev->uvd.idle_work, amdgpu_uvd_idle_work_handler);
@@ -260,7 +260,7 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev)
 		return -EINVAL;
 	}
 
-	r = amdgpu_ucode_request(adev, &adev->uvd.fw, fw_name);
+	r = amdgpu_ucode_request(adev, &adev->uvd.fw, AMDGPU_UCODE_REQUIRED, "%s", fw_name);
 	if (r) {
 		dev_err(adev->dev, "amdgpu_uvd: Can't validate firmware \"%s\"\n",
 			fw_name);
@@ -275,7 +275,7 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev)
 	family_id = le32_to_cpu(hdr->ucode_version) & 0xff;
 
 	if (adev->asic_type < CHIP_VEGA20) {
-		unsigned version_major, version_minor;
+		unsigned int version_major, version_minor;
 
 		version_major = (le32_to_cpu(hdr->ucode_version) >> 24) & 0xff;
 		version_minor = (le32_to_cpu(hdr->ucode_version) >> 8) & 0xff;
@@ -398,32 +398,32 @@ int amdgpu_uvd_sw_fini(struct amdgpu_device *adev)
  * amdgpu_uvd_entity_init - init entity
  *
  * @adev: amdgpu_device pointer
+ * @ring: amdgpu_ring pointer to check
  *
+ * Initialize the entity used for handle management in the kernel driver.
  */
-int amdgpu_uvd_entity_init(struct amdgpu_device *adev)
+int amdgpu_uvd_entity_init(struct amdgpu_device *adev, struct amdgpu_ring *ring)
 {
-	struct amdgpu_ring *ring;
-	struct drm_gpu_scheduler *sched;
-	int r;
+	if (ring == &adev->uvd.inst[0].ring) {
+		struct drm_gpu_scheduler *sched = &ring->sched;
+		int r;
 
-	ring = &adev->uvd.inst[0].ring;
-	sched = &ring->sched;
-	r = drm_sched_entity_init(&adev->uvd.entity, DRM_SCHED_PRIORITY_NORMAL,
-				  &sched, 1, NULL);
-	if (r) {
-		DRM_ERROR("Failed setting up UVD kernel entity.\n");
-		return r;
+		r = drm_sched_entity_init(&adev->uvd.entity, DRM_SCHED_PRIORITY_NORMAL,
+					  &sched, 1, NULL);
+		if (r) {
+			DRM_ERROR("Failed setting up UVD kernel entity.\n");
+			return r;
+		}
 	}
 
 	return 0;
 }
 
-int amdgpu_uvd_suspend(struct amdgpu_device *adev)
+int amdgpu_uvd_prepare_suspend(struct amdgpu_device *adev)
 {
-	unsigned size;
+	unsigned int size;
 	void *ptr;
 	int i, j, idx;
-	bool in_ras_intr = amdgpu_ras_intr_triggered();
 
 	cancel_delayed_work_sync(&adev->uvd.idle_work);
 
@@ -452,7 +452,7 @@ int amdgpu_uvd_suspend(struct amdgpu_device *adev)
 
 		if (drm_dev_enter(adev_to_drm(adev), &idx)) {
 			/* re-write 0 since err_event_athub will corrupt VCPU buffer */
-			if (in_ras_intr)
+			if (amdgpu_ras_intr_triggered())
 				memset(adev->uvd.inst[j].saved_bo, 0, size);
 			else
 				memcpy_fromio(adev->uvd.inst[j].saved_bo, ptr, size);
@@ -461,7 +461,12 @@ int amdgpu_uvd_suspend(struct amdgpu_device *adev)
 		}
 	}
 
-	if (in_ras_intr)
+	return 0;
+}
+
+int amdgpu_uvd_suspend(struct amdgpu_device *adev)
+{
+	if (amdgpu_ras_intr_triggered())
 		DRM_WARN("UVD VCPU state may lost due to RAS ERREVENT_ATHUB_INTERRUPT\n");
 
 	return 0;
@@ -469,7 +474,7 @@ int amdgpu_uvd_suspend(struct amdgpu_device *adev)
 
 int amdgpu_uvd_resume(struct amdgpu_device *adev)
 {
-	unsigned size;
+	unsigned int size;
 	void *ptr;
 	int i, idx;
 
@@ -491,7 +496,7 @@ int amdgpu_uvd_resume(struct amdgpu_device *adev)
 			adev->uvd.inst[i].saved_bo = NULL;
 		} else {
 			const struct common_firmware_header *hdr;
-			unsigned offset;
+			unsigned int offset;
 
 			hdr = (const struct common_firmware_header *)adev->uvd.fw->data;
 			if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
@@ -542,9 +547,12 @@ void amdgpu_uvd_free_handles(struct amdgpu_device *adev, struct drm_file *filp)
 static void amdgpu_uvd_force_into_uvd_segment(struct amdgpu_bo *abo)
 {
 	int i;
+
 	for (i = 0; i < abo->placement.num_placement; ++i) {
 		abo->placements[i].fpfn = 0 >> PAGE_SHIFT;
 		abo->placements[i].lpfn = (256 * 1024 * 1024) >> PAGE_SHIFT;
+		if (abo->placements[i].mem_type == TTM_PL_VRAM)
+			abo->placements[i].flags |= TTM_PL_FLAG_CONTIGUOUS;
 	}
 }
 
@@ -579,7 +587,7 @@ static int amdgpu_uvd_cs_pass1(struct amdgpu_uvd_cs_ctx *ctx)
 
 	r = amdgpu_cs_find_mapping(ctx->parser, addr, &bo, &mapping);
 	if (r) {
-		DRM_ERROR("Can't find BO for addr 0x%08Lx\n", addr);
+		DRM_ERROR("Can't find BO for addr 0x%08llx\n", addr);
 		return r;
 	}
 
@@ -589,6 +597,7 @@ static int amdgpu_uvd_cs_pass1(struct amdgpu_uvd_cs_ctx *ctx)
 		if (cmd == 0x0 || cmd == 0x3) {
 			/* yes, force it into VRAM */
 			uint32_t domain = AMDGPU_GEM_DOMAIN_VRAM;
+
 			amdgpu_bo_placement_from_domain(bo, domain);
 		}
 		amdgpu_uvd_force_into_uvd_segment(bo);
@@ -609,21 +618,21 @@ static int amdgpu_uvd_cs_pass1(struct amdgpu_uvd_cs_ctx *ctx)
  * Peek into the decode message and calculate the necessary buffer sizes.
  */
 static int amdgpu_uvd_cs_msg_decode(struct amdgpu_device *adev, uint32_t *msg,
-	unsigned buf_sizes[])
+	unsigned int buf_sizes[])
 {
-	unsigned stream_type = msg[4];
-	unsigned width = msg[6];
-	unsigned height = msg[7];
-	unsigned dpb_size = msg[9];
-	unsigned pitch = msg[28];
-	unsigned level = msg[57];
+	unsigned int stream_type = msg[4];
+	unsigned int width = msg[6];
+	unsigned int height = msg[7];
+	unsigned int dpb_size = msg[9];
+	unsigned int pitch = msg[28];
+	unsigned int level = msg[57];
 
-	unsigned width_in_mb = width / 16;
-	unsigned height_in_mb = ALIGN(height / 16, 2);
-	unsigned fs_in_mb = width_in_mb * height_in_mb;
+	unsigned int width_in_mb = width / 16;
+	unsigned int height_in_mb = ALIGN(height / 16, 2);
+	unsigned int fs_in_mb = width_in_mb * height_in_mb;
 
-	unsigned image_size, tmp, min_dpb_size, num_dpb_buffer;
-	unsigned min_ctx_size = ~0;
+	unsigned int image_size, tmp, min_dpb_size, num_dpb_buffer;
+	unsigned int min_ctx_size = ~0;
 
 	image_size = width * height;
 	image_size += image_size / 2;
@@ -631,7 +640,7 @@ static int amdgpu_uvd_cs_msg_decode(struct amdgpu_device *adev, uint32_t *msg,
 
 	switch (stream_type) {
 	case 0: /* H264 */
-		switch(level) {
+		switch (level) {
 		case 30:
 			num_dpb_buffer = 8100 / fs_in_mb;
 			break;
@@ -709,7 +718,7 @@ static int amdgpu_uvd_cs_msg_decode(struct amdgpu_device *adev, uint32_t *msg,
 		break;
 
 	case 7: /* H264 Perf */
-		switch(level) {
+		switch (level) {
 		case 30:
 			num_dpb_buffer = 8100 / fs_in_mb;
 			break;
@@ -742,7 +751,7 @@ static int amdgpu_uvd_cs_msg_decode(struct amdgpu_device *adev, uint32_t *msg,
 		/* reference picture buffer */
 		min_dpb_size = image_size * num_dpb_buffer;
 
-		if (!adev->uvd.use_ctx_buf){
+		if (!adev->uvd.use_ctx_buf) {
 			/* macroblock context buffer */
 			min_dpb_size +=
 				width_in_mb * height_in_mb * num_dpb_buffer * 192;
@@ -805,7 +814,7 @@ static int amdgpu_uvd_cs_msg_decode(struct amdgpu_device *adev, uint32_t *msg,
  * Make sure that we don't open up to many sessions.
  */
 static int amdgpu_uvd_cs_msg(struct amdgpu_uvd_cs_ctx *ctx,
-			     struct amdgpu_bo *bo, unsigned offset)
+			     struct amdgpu_bo *bo, unsigned int offset)
 {
 	struct amdgpu_device *adev = ctx->parser->adev;
 	int32_t *msg, msg_type, handle;
@@ -911,7 +920,7 @@ static int amdgpu_uvd_cs_pass2(struct amdgpu_uvd_cs_ctx *ctx)
 
 	r = amdgpu_cs_find_mapping(ctx->parser, addr, &bo, &mapping);
 	if (r) {
-		DRM_ERROR("Can't find BO for addr 0x%08Lx\n", addr);
+		DRM_ERROR("Can't find BO for addr 0x%08llx\n", addr);
 		return r;
 	}
 
@@ -930,7 +939,7 @@ static int amdgpu_uvd_cs_pass2(struct amdgpu_uvd_cs_ctx *ctx)
 	if (cmd < 0x4) {
 		if ((end - start) < ctx->buf_sizes[cmd]) {
 			DRM_ERROR("buffer (%d) to small (%d / %d)!\n", cmd,
-				  (unsigned)(end - start),
+				  (unsigned int)(end - start),
 				  ctx->buf_sizes[cmd]);
 			return -EINVAL;
 		}
@@ -938,7 +947,7 @@ static int amdgpu_uvd_cs_pass2(struct amdgpu_uvd_cs_ctx *ctx)
 	} else if (cmd == 0x206) {
 		if ((end - start) < ctx->buf_sizes[4]) {
 			DRM_ERROR("buffer (%d) to small (%d / %d)!\n", cmd,
-					  (unsigned)(end - start),
+					  (unsigned int)(end - start),
 					  ctx->buf_sizes[4]);
 			return -EINVAL;
 		}
@@ -949,14 +958,14 @@ static int amdgpu_uvd_cs_pass2(struct amdgpu_uvd_cs_ctx *ctx)
 
 	if (!ctx->parser->adev->uvd.address_64_bit) {
 		if ((start >> 28) != ((end - 1) >> 28)) {
-			DRM_ERROR("reloc %LX-%LX crossing 256MB boundary!\n",
+			DRM_ERROR("reloc %llx-%llx crossing 256MB boundary!\n",
 				  start, end);
 			return -EINVAL;
 		}
 
 		if ((cmd == 0 || cmd == 0x3) &&
 		    (start >> 28) != (ctx->parser->adev->uvd.inst->gpu_addr >> 28)) {
-			DRM_ERROR("msg/fb buffer %LX-%LX out of 256MB segment!\n",
+			DRM_ERROR("msg/fb buffer %llx-%llx out of 256MB segment!\n",
 				  start, end);
 			return -EINVAL;
 		}
@@ -990,7 +999,7 @@ static int amdgpu_uvd_cs_reg(struct amdgpu_uvd_cs_ctx *ctx,
 
 	ctx->idx++;
 	for (i = 0; i <= ctx->count; ++i) {
-		unsigned reg = ctx->reg + i;
+		unsigned int reg = ctx->reg + i;
 
 		if (ctx->idx >= ctx->ib->length_dw) {
 			DRM_ERROR("Register command after end of CS!\n");
@@ -1036,7 +1045,8 @@ static int amdgpu_uvd_cs_packets(struct amdgpu_uvd_cs_ctx *ctx,
 
 	for (ctx->idx = 0 ; ctx->idx < ctx->ib->length_dw; ) {
 		uint32_t cmd = amdgpu_ib_get_value(ctx->ib, ctx->idx);
-		unsigned type = CP_PACKET_GET_TYPE(cmd);
+		unsigned int type = CP_PACKET_GET_TYPE(cmd);
+
 		switch (type) {
 		case PACKET_TYPE0:
 			ctx->reg = CP_PACKET0_GET_REG(cmd);
@@ -1070,7 +1080,7 @@ int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser,
 			     struct amdgpu_ib *ib)
 {
 	struct amdgpu_uvd_cs_ctx ctx = {};
-	unsigned buf_sizes[] = {
+	unsigned int buf_sizes[] = {
 		[0x00000000]	=	2048,
 		[0x00000001]	=	0xFFFFFFFF,
 		[0x00000002]	=	0xFFFFFFFF,
@@ -1080,7 +1090,6 @@ int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser,
 	int r;
 
 	job->vm = NULL;
-	ib->gpu_addr = amdgpu_sa_bo_gpu_addr(ib->sa_bo);
 
 	if (ib->length_dw % 16) {
 		DRM_ERROR("UVD IB length (%d) not 16 dwords aligned!\n",
@@ -1127,7 +1136,8 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo,
 	r = amdgpu_job_alloc_with_ib(ring->adev, &adev->uvd.entity,
 				     AMDGPU_FENCE_OWNER_UNDEFINED,
 				     64, direct ? AMDGPU_IB_POOL_DIRECT :
-				     AMDGPU_IB_POOL_DELAYED, &job);
+				     AMDGPU_IB_POOL_DELAYED, &job,
+				     AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		return r;
 
@@ -1185,8 +1195,9 @@ err_free:
 }
 
 /* multiple fence commands without any stream commands in between can
-   crash the vcpu so just try to emmit a dummy create/destroy msg to
-   avoid this */
+ * crash the vcpu so just try to emmit a dummy create/destroy msg to
+ * avoid this
+ */
 int amdgpu_uvd_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
 			      struct dma_fence **fence)
 {
@@ -1252,15 +1263,14 @@ static void amdgpu_uvd_idle_work_handler(struct work_struct *work)
 {
 	struct amdgpu_device *adev =
 		container_of(work, struct amdgpu_device, uvd.idle_work.work);
-	unsigned fences = 0, i, j;
+	unsigned int fences = 0, i, j;
 
 	for (i = 0; i < adev->uvd.num_uvd_inst; ++i) {
 		if (adev->uvd.harvest_config & (1 << i))
 			continue;
 		fences += amdgpu_fence_count_emitted(&adev->uvd.inst[i].ring);
-		for (j = 0; j < adev->uvd.num_enc_rings; ++j) {
+		for (j = 0; j < adev->uvd.num_enc_rings; ++j)
 			fences += amdgpu_fence_count_emitted(&adev->uvd.inst[i].ring_enc[j]);
-		}
 	}
 
 	if (fences == 0) {
@@ -1356,7 +1366,7 @@ error:
  */
 uint32_t amdgpu_uvd_used_handles(struct amdgpu_device *adev)
 {
-	unsigned i;
+	unsigned int i;
 	uint32_t used_handles = 0;
 
 	for (i = 0; i < adev->uvd.max_handles; ++i) {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h
index 9f89bb7cd60b..9dfad2f48ef4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h
@@ -73,7 +73,8 @@ struct amdgpu_uvd {
 
 int amdgpu_uvd_sw_init(struct amdgpu_device *adev);
 int amdgpu_uvd_sw_fini(struct amdgpu_device *adev);
-int amdgpu_uvd_entity_init(struct amdgpu_device *adev);
+int amdgpu_uvd_entity_init(struct amdgpu_device *adev, struct amdgpu_ring *ring);
+int amdgpu_uvd_prepare_suspend(struct amdgpu_device *adev);
 int amdgpu_uvd_suspend(struct amdgpu_device *adev);
 int amdgpu_uvd_resume(struct amdgpu_device *adev);
 int amdgpu_uvd_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
index e2b7324a70cb..a7d8f1ce6ac2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
@@ -41,6 +41,9 @@
 #define VCE_IDLE_TIMEOUT	msecs_to_jiffies(1000)
 
 /* Firmware Names */
+#ifdef CONFIG_DRM_AMDGPU_SI
+#define FIRMWARE_VCE_V1_0	"amdgpu/vce_1_0_0.bin"
+#endif
 #ifdef CONFIG_DRM_AMDGPU_CIK
 #define FIRMWARE_BONAIRE	"amdgpu/bonaire_vce.bin"
 #define FIRMWARE_KABINI	"amdgpu/kabini_vce.bin"
@@ -61,6 +64,9 @@
 #define FIRMWARE_VEGA12		"amdgpu/vega12_vce.bin"
 #define FIRMWARE_VEGA20		"amdgpu/vega20_vce.bin"
 
+#ifdef CONFIG_DRM_AMDGPU_SI
+MODULE_FIRMWARE(FIRMWARE_VCE_V1_0);
+#endif
 #ifdef CONFIG_DRM_AMDGPU_CIK
 MODULE_FIRMWARE(FIRMWARE_BONAIRE);
 MODULE_FIRMWARE(FIRMWARE_KABINI);
@@ -88,82 +94,93 @@ static int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle,
 				      bool direct, struct dma_fence **fence);
 
 /**
- * amdgpu_vce_sw_init - allocate memory, load vce firmware
+ * amdgpu_vce_firmware_name() - determine the firmware file name for VCE
  *
  * @adev: amdgpu_device pointer
- * @size: size for the new BO
  *
- * First step to get VCE online, allocate memory and load the firmware
+ * Each chip that has VCE IP may need a different firmware.
+ * This function returns the name of the VCE firmware file
+ * appropriate for the current chip.
  */
-int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size)
+static const char *amdgpu_vce_firmware_name(struct amdgpu_device *adev)
 {
-	const char *fw_name;
-	const struct common_firmware_header *hdr;
-	unsigned ucode_version, version_major, version_minor, binary_id;
-	int i, r;
-
 	switch (adev->asic_type) {
+#ifdef CONFIG_DRM_AMDGPU_SI
+	case CHIP_PITCAIRN:
+	case CHIP_TAHITI:
+	case CHIP_VERDE:
+		return FIRMWARE_VCE_V1_0;
+#endif
 #ifdef CONFIG_DRM_AMDGPU_CIK
 	case CHIP_BONAIRE:
-		fw_name = FIRMWARE_BONAIRE;
-		break;
+		return FIRMWARE_BONAIRE;
 	case CHIP_KAVERI:
-		fw_name = FIRMWARE_KAVERI;
-		break;
+		return FIRMWARE_KAVERI;
 	case CHIP_KABINI:
-		fw_name = FIRMWARE_KABINI;
-		break;
+		return FIRMWARE_KABINI;
 	case CHIP_HAWAII:
-		fw_name = FIRMWARE_HAWAII;
-		break;
+		return FIRMWARE_HAWAII;
 	case CHIP_MULLINS:
-		fw_name = FIRMWARE_MULLINS;
-		break;
+		return FIRMWARE_MULLINS;
 #endif
 	case CHIP_TONGA:
-		fw_name = FIRMWARE_TONGA;
-		break;
+		return  FIRMWARE_TONGA;
 	case CHIP_CARRIZO:
-		fw_name = FIRMWARE_CARRIZO;
-		break;
+		return  FIRMWARE_CARRIZO;
 	case CHIP_FIJI:
-		fw_name = FIRMWARE_FIJI;
-		break;
+		return  FIRMWARE_FIJI;
 	case CHIP_STONEY:
-		fw_name = FIRMWARE_STONEY;
-		break;
+		return  FIRMWARE_STONEY;
 	case CHIP_POLARIS10:
-		fw_name = FIRMWARE_POLARIS10;
-		break;
+		return  FIRMWARE_POLARIS10;
 	case CHIP_POLARIS11:
-		fw_name = FIRMWARE_POLARIS11;
-		break;
+		return  FIRMWARE_POLARIS11;
 	case CHIP_POLARIS12:
-		fw_name = FIRMWARE_POLARIS12;
-		break;
+		return  FIRMWARE_POLARIS12;
 	case CHIP_VEGAM:
-		fw_name = FIRMWARE_VEGAM;
-		break;
+		return  FIRMWARE_VEGAM;
 	case CHIP_VEGA10:
-		fw_name = FIRMWARE_VEGA10;
-		break;
+		return  FIRMWARE_VEGA10;
 	case CHIP_VEGA12:
-		fw_name = FIRMWARE_VEGA12;
-		break;
+		return  FIRMWARE_VEGA12;
 	case CHIP_VEGA20:
-		fw_name = FIRMWARE_VEGA20;
-		break;
+		return  FIRMWARE_VEGA20;
 
 	default:
-		return -EINVAL;
+		return NULL;
 	}
+}
 
-	r = amdgpu_ucode_request(adev, &adev->vce.fw, fw_name);
+/**
+ * amdgpu_vce_early_init() - try to load VCE firmware
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Tries to load the VCE firmware.
+ *
+ * When not found, returns ENOENT so that the driver can
+ * still load and initialize the rest of the IP blocks.
+ * The GPU can function just fine without VCE, they will just
+ * not support video encoding.
+ */
+int amdgpu_vce_early_init(struct amdgpu_device *adev)
+{
+	const char *fw_name = amdgpu_vce_firmware_name(adev);
+	const struct common_firmware_header *hdr;
+	unsigned int ucode_version, version_major, version_minor, binary_id;
+	int r;
+
+	if (!fw_name)
+		return -ENOENT;
+
+	r = amdgpu_ucode_request(adev, &adev->vce.fw, AMDGPU_UCODE_REQUIRED, "%s", fw_name);
 	if (r) {
-		dev_err(adev->dev, "amdgpu_vce: Can't validate firmware \"%s\"\n",
-			fw_name);
+		dev_err(adev->dev,
+			"amdgpu_vce: Firmware \"%s\" not found or failed to validate (%d)\n",
+			fw_name, r);
+
 		amdgpu_ucode_release(&adev->vce.fw);
-		return r;
+		return -ENOENT;
 	}
 
 	hdr = (const struct common_firmware_header *)adev->vce.fw->data;
@@ -172,11 +189,35 @@ int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size)
 	version_major = (ucode_version >> 20) & 0xfff;
 	version_minor = (ucode_version >> 8) & 0xfff;
 	binary_id = ucode_version & 0xff;
-	DRM_INFO("Found VCE firmware Version: %d.%d Binary ID: %d\n",
+	dev_info(adev->dev, "Found VCE firmware Version: %d.%d Binary ID: %d\n",
 		version_major, version_minor, binary_id);
 	adev->vce.fw_version = ((version_major << 24) | (version_minor << 16) |
 				(binary_id << 8));
 
+	return 0;
+}
+
+/**
+ * amdgpu_vce_sw_init() - allocate memory for VCE BO
+ *
+ * @adev: amdgpu_device pointer
+ * @size: size for the new BO
+ *
+ * First step to get VCE online: allocate memory for VCE BO.
+ * The VCE firmware binary is copied into the VCE BO later,
+ * in amdgpu_vce_resume. The VCE executes its code from the
+ * VCE BO and also uses the space in this BO for its stack and data.
+ *
+ * Ideally this BO should be placed in VRAM for optimal performance,
+ * although technically it also runs from system RAM (albeit slowly).
+ */
+int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size)
+{
+	int i, r;
+
+	if (!adev->vce.fw)
+		return -ENOENT;
+
 	r = amdgpu_bo_create_kernel(adev, size, PAGE_SIZE,
 				    AMDGPU_GEM_DOMAIN_VRAM |
 				    AMDGPU_GEM_DOMAIN_GTT,
@@ -207,22 +248,22 @@ int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size)
  */
 int amdgpu_vce_sw_fini(struct amdgpu_device *adev)
 {
-	unsigned i;
+	unsigned int i;
 
 	if (adev->vce.vcpu_bo == NULL)
 		return 0;
 
 	drm_sched_entity_destroy(&adev->vce.entity);
 
-	amdgpu_bo_free_kernel(&adev->vce.vcpu_bo, &adev->vce.gpu_addr,
-		(void **)&adev->vce.cpu_addr);
-
 	for (i = 0; i < adev->vce.num_rings; i++)
 		amdgpu_ring_fini(&adev->vce.ring[i]);
 
 	amdgpu_ucode_release(&adev->vce.fw);
 	mutex_destroy(&adev->vce.idle_mutex);
 
+	amdgpu_bo_free_kernel(&adev->vce.vcpu_bo, &adev->vce.gpu_addr,
+		(void **)&adev->vce.cpu_addr);
+
 	return 0;
 }
 
@@ -230,21 +271,22 @@ int amdgpu_vce_sw_fini(struct amdgpu_device *adev)
  * amdgpu_vce_entity_init - init entity
  *
  * @adev: amdgpu_device pointer
+ * @ring: amdgpu_ring pointer to check
  *
+ * Initialize the entity used for handle management in the kernel driver.
  */
-int amdgpu_vce_entity_init(struct amdgpu_device *adev)
+int amdgpu_vce_entity_init(struct amdgpu_device *adev, struct amdgpu_ring *ring)
 {
-	struct amdgpu_ring *ring;
-	struct drm_gpu_scheduler *sched;
-	int r;
-
-	ring = &adev->vce.ring[0];
-	sched = &ring->sched;
-	r = drm_sched_entity_init(&adev->vce.entity, DRM_SCHED_PRIORITY_NORMAL,
-				  &sched, 1, NULL);
-	if (r != 0) {
-		DRM_ERROR("Failed setting up VCE run queue.\n");
-		return r;
+	if (ring == &adev->vce.ring[0]) {
+		struct drm_gpu_scheduler *sched = &ring->sched;
+		int r;
+
+		r = drm_sched_entity_init(&adev->vce.entity, DRM_SCHED_PRIORITY_NORMAL,
+					  &sched, 1, NULL);
+		if (r != 0) {
+			DRM_ERROR("Failed setting up VCE run queue.\n");
+			return r;
+		}
 	}
 
 	return 0;
@@ -284,40 +326,23 @@ int amdgpu_vce_suspend(struct amdgpu_device *adev)
  */
 int amdgpu_vce_resume(struct amdgpu_device *adev)
 {
-	void *cpu_addr;
 	const struct common_firmware_header *hdr;
-	unsigned offset;
-	int r, idx;
+	unsigned int offset;
+	int idx;
 
 	if (adev->vce.vcpu_bo == NULL)
 		return -EINVAL;
 
-	r = amdgpu_bo_reserve(adev->vce.vcpu_bo, false);
-	if (r) {
-		dev_err(adev->dev, "(%d) failed to reserve VCE bo\n", r);
-		return r;
-	}
-
-	r = amdgpu_bo_kmap(adev->vce.vcpu_bo, &cpu_addr);
-	if (r) {
-		amdgpu_bo_unreserve(adev->vce.vcpu_bo);
-		dev_err(adev->dev, "(%d) VCE map failed\n", r);
-		return r;
-	}
-
 	hdr = (const struct common_firmware_header *)adev->vce.fw->data;
 	offset = le32_to_cpu(hdr->ucode_array_offset_bytes);
 
 	if (drm_dev_enter(adev_to_drm(adev), &idx)) {
-		memcpy_toio(cpu_addr, adev->vce.fw->data + offset,
+		memset_io(adev->vce.cpu_addr, 0, amdgpu_bo_size(adev->vce.vcpu_bo));
+		memcpy_toio(adev->vce.cpu_addr, adev->vce.fw->data + offset,
 			    adev->vce.fw->size - offset);
 		drm_dev_exit(idx);
 	}
 
-	amdgpu_bo_kunmap(adev->vce.vcpu_bo);
-
-	amdgpu_bo_unreserve(adev->vce.vcpu_bo);
-
 	return 0;
 }
 
@@ -332,7 +357,7 @@ static void amdgpu_vce_idle_work_handler(struct work_struct *work)
 {
 	struct amdgpu_device *adev =
 		container_of(work, struct amdgpu_device, vce.idle_work.work);
-	unsigned i, count = 0;
+	unsigned int i, count = 0;
 
 	for (i = 0; i < adev->vce.num_rings; i++)
 		count += amdgpu_fence_count_emitted(&adev->vce.ring[i]);
@@ -409,6 +434,7 @@ void amdgpu_vce_free_handles(struct amdgpu_device *adev, struct drm_file *filp)
 {
 	struct amdgpu_ring *ring = &adev->vce.ring[0];
 	int i, r;
+
 	for (i = 0; i < AMDGPU_MAX_VCE_HANDLES; ++i) {
 		uint32_t handle = atomic_read(&adev->vce.handles[i]);
 
@@ -425,6 +451,24 @@ void amdgpu_vce_free_handles(struct amdgpu_device *adev, struct drm_file *filp)
 }
 
 /**
+ * amdgpu_vce_required_gart_pages() - gets number of GART pages required by VCE
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Returns how many GART pages we need before GTT for the VCE IP block.
+ * For VCE1, see vce_v1_0_ensure_vcpu_bo_32bit_addr for details.
+ * For VCE2+, this is not needed so return zero.
+ */
+u32 amdgpu_vce_required_gart_pages(struct amdgpu_device *adev)
+{
+	/* VCE IP block not added yet, so can't use amdgpu_ip_version */
+	if (adev->family == AMDGPU_FAMILY_SI)
+		return 512;
+
+	return 0;
+}
+
+/**
  * amdgpu_vce_get_create_msg - generate a VCE create msg
  *
  * @ring: ring we should submit the msg to
@@ -436,7 +480,7 @@ void amdgpu_vce_free_handles(struct amdgpu_device *adev, struct drm_file *filp)
 static int amdgpu_vce_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
 				     struct dma_fence **fence)
 {
-	const unsigned ib_size_dw = 1024;
+	const unsigned int ib_size_dw = 1024;
 	struct amdgpu_job *job;
 	struct amdgpu_ib *ib;
 	struct amdgpu_ib ib_msg;
@@ -447,7 +491,7 @@ static int amdgpu_vce_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
 	r = amdgpu_job_alloc_with_ib(ring->adev, &ring->adev->vce.entity,
 				     AMDGPU_FENCE_OWNER_UNDEFINED,
 				     ib_size_dw * 4, AMDGPU_IB_POOL_DIRECT,
-				     &job);
+				     &job, AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		return r;
 
@@ -501,7 +545,7 @@ static int amdgpu_vce_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
 		ib->ptr[i] = 0x0;
 
 	r = amdgpu_job_submit_direct(job, ring, &f);
-	amdgpu_ib_free(ring->adev, &ib_msg, f);
+	amdgpu_ib_free(&ib_msg, f);
 	if (r)
 		goto err;
 
@@ -528,7 +572,7 @@ err:
 static int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle,
 				      bool direct, struct dma_fence **fence)
 {
-	const unsigned ib_size_dw = 1024;
+	const unsigned int ib_size_dw = 1024;
 	struct amdgpu_job *job;
 	struct amdgpu_ib *ib;
 	struct dma_fence *f = NULL;
@@ -538,7 +582,8 @@ static int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle,
 				     AMDGPU_FENCE_OWNER_UNDEFINED,
 				     ib_size_dw * 4,
 				     direct ? AMDGPU_IB_POOL_DIRECT :
-				     AMDGPU_IB_POOL_DELAYED, &job);
+				     AMDGPU_IB_POOL_DELAYED, &job,
+				     AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		return r;
 
@@ -596,12 +641,12 @@ err:
  */
 static int amdgpu_vce_validate_bo(struct amdgpu_cs_parser *p,
 				  struct amdgpu_ib *ib, int lo, int hi,
-				  unsigned size, int32_t index)
+				  unsigned int size, int32_t index)
 {
 	int64_t offset = ((uint64_t)size) * ((int64_t)index);
 	struct ttm_operation_ctx ctx = { false, false };
 	struct amdgpu_bo_va_mapping *mapping;
-	unsigned i, fpfn, lpfn;
+	unsigned int i, fpfn, lpfn;
 	struct amdgpu_bo *bo;
 	uint64_t addr;
 	int r;
@@ -619,7 +664,7 @@ static int amdgpu_vce_validate_bo(struct amdgpu_cs_parser *p,
 
 	r = amdgpu_cs_find_mapping(p, addr, &bo, &mapping);
 	if (r) {
-		DRM_ERROR("Can't find BO for addr 0x%010Lx %d %d %d %d\n",
+		DRM_ERROR("Can't find BO for addr 0x%010llx %d %d %d %d\n",
 			  addr, lo, hi, size, index);
 		return r;
 	}
@@ -646,7 +691,7 @@ static int amdgpu_vce_validate_bo(struct amdgpu_cs_parser *p,
  * Patch relocation inside command stream with real buffer address
  */
 static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, struct amdgpu_ib *ib,
-			       int lo, int hi, unsigned size, uint32_t index)
+			       int lo, int hi, unsigned int size, uint32_t index)
 {
 	struct amdgpu_bo_va_mapping *mapping;
 	struct amdgpu_bo *bo;
@@ -662,14 +707,14 @@ static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, struct amdgpu_ib *ib,
 
 	r = amdgpu_cs_find_mapping(p, addr, &bo, &mapping);
 	if (r) {
-		DRM_ERROR("Can't find BO for addr 0x%010Lx %d %d %d %d\n",
+		DRM_ERROR("Can't find BO for addr 0x%010llx %d %d %d %d\n",
 			  addr, lo, hi, size, index);
 		return r;
 	}
 
 	if ((addr + (uint64_t)size) >
 	    (mapping->last + 1) * AMDGPU_GPU_PAGE_SIZE) {
-		DRM_ERROR("BO too small for addr 0x%010Lx %d %d\n",
+		DRM_ERROR("BO too small for addr 0x%010llx %d %d\n",
 			  addr, lo, hi);
 		return -EINVAL;
 	}
@@ -692,12 +737,12 @@ static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, struct amdgpu_ib *ib,
  * @allocated: allocated a new handle?
  *
  * Validates the handle and return the found session index or -EINVAL
- * we we don't have another free session index.
+ * we don't have another free session index.
  */
 static int amdgpu_vce_validate_handle(struct amdgpu_cs_parser *p,
 				      uint32_t handle, uint32_t *allocated)
 {
-	unsigned i;
+	unsigned int i;
 
 	/* validate the handle */
 	for (i = 0; i < AMDGPU_MAX_VCE_HANDLES; ++i) {
@@ -735,18 +780,18 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p,
 			     struct amdgpu_job *job,
 			     struct amdgpu_ib *ib)
 {
-	unsigned fb_idx = 0, bs_idx = 0;
+	unsigned int fb_idx = 0, bs_idx = 0;
 	int session_idx = -1;
 	uint32_t destroyed = 0;
 	uint32_t created = 0;
 	uint32_t allocated = 0;
 	uint32_t tmp, handle = 0;
-	uint32_t *size = &tmp;
-	unsigned idx;
+	uint32_t dummy = 0xffffffff;
+	uint32_t *size = &dummy;
+	unsigned int idx;
 	int i, r = 0;
 
 	job->vm = NULL;
-	ib->gpu_addr = amdgpu_sa_bo_gpu_addr(ib->sa_bo);
 
 	for (idx = 0; idx < ib->length_dw;) {
 		uint32_t len = amdgpu_ib_get_value(ib, idx);
@@ -1041,7 +1086,6 @@ out:
 	if (!r) {
 		/* No error, free all destroyed handle slots */
 		tmp = destroyed;
-		amdgpu_ib_free(p->adev, ib, NULL);
 	} else {
 		/* Error during parsing, free all allocated handle slots */
 		tmp = allocated;
@@ -1084,7 +1128,7 @@ void amdgpu_vce_ring_emit_ib(struct amdgpu_ring *ring,
  *
  */
 void amdgpu_vce_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq,
-				unsigned flags)
+				unsigned int flags)
 {
 	WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
 
@@ -1106,7 +1150,7 @@ int amdgpu_vce_ring_test_ring(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
 	uint32_t rptr;
-	unsigned i;
+	unsigned int i;
 	int r, timeout = adev->usec_timeout;
 
 	/* skip ring test for sriov*/
@@ -1171,7 +1215,7 @@ error:
 
 enum amdgpu_ring_priority_level amdgpu_vce_get_ring_prio(int ring)
 {
-	switch(ring) {
+	switch (ring) {
 	case 0:
 		return AMDGPU_RING_PRIO_0;
 	case 1:
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h
index ea680fc9a6c3..1c3464ce5037 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h
@@ -51,14 +51,17 @@ struct amdgpu_vce {
 	struct drm_sched_entity	entity;
 	uint32_t                srbm_soft_reset;
 	unsigned		num_rings;
+	uint32_t		keyselect;
 };
 
+int amdgpu_vce_early_init(struct amdgpu_device *adev);
 int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size);
 int amdgpu_vce_sw_fini(struct amdgpu_device *adev);
-int amdgpu_vce_entity_init(struct amdgpu_device *adev);
+int amdgpu_vce_entity_init(struct amdgpu_device *adev, struct amdgpu_ring *ring);
 int amdgpu_vce_suspend(struct amdgpu_device *adev);
 int amdgpu_vce_resume(struct amdgpu_device *adev);
 void amdgpu_vce_free_handles(struct amdgpu_device *adev, struct drm_file *filp);
+u32 amdgpu_vce_required_gart_pages(struct amdgpu_device *adev);
 int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, struct amdgpu_job *job,
 			     struct amdgpu_ib *ib);
 int amdgpu_vce_ring_parse_cs_vm(struct amdgpu_cs_parser *p,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
index e63fcc58e8e0..5e0786ea911b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2016 Advanced Micro Devices, Inc.
+ * Copyright 2016-2024 Advanced Micro Devices, Inc.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -56,7 +56,13 @@
 #define FIRMWARE_VCN_3_1_2		"amdgpu/vcn_3_1_2.bin"
 #define FIRMWARE_VCN4_0_0		"amdgpu/vcn_4_0_0.bin"
 #define FIRMWARE_VCN4_0_2		"amdgpu/vcn_4_0_2.bin"
+#define FIRMWARE_VCN4_0_3		"amdgpu/vcn_4_0_3.bin"
 #define FIRMWARE_VCN4_0_4		"amdgpu/vcn_4_0_4.bin"
+#define FIRMWARE_VCN4_0_5		"amdgpu/vcn_4_0_5.bin"
+#define FIRMWARE_VCN4_0_6		"amdgpu/vcn_4_0_6.bin"
+#define FIRMWARE_VCN4_0_6_1		"amdgpu/vcn_4_0_6_1.bin"
+#define FIRMWARE_VCN5_0_0		"amdgpu/vcn_5_0_0.bin"
+#define FIRMWARE_VCN5_0_1		"amdgpu/vcn_5_0_1.bin"
 
 MODULE_FIRMWARE(FIRMWARE_RAVEN);
 MODULE_FIRMWARE(FIRMWARE_PICASSO);
@@ -77,43 +83,65 @@ MODULE_FIRMWARE(FIRMWARE_YELLOW_CARP);
 MODULE_FIRMWARE(FIRMWARE_VCN_3_1_2);
 MODULE_FIRMWARE(FIRMWARE_VCN4_0_0);
 MODULE_FIRMWARE(FIRMWARE_VCN4_0_2);
+MODULE_FIRMWARE(FIRMWARE_VCN4_0_3);
 MODULE_FIRMWARE(FIRMWARE_VCN4_0_4);
+MODULE_FIRMWARE(FIRMWARE_VCN4_0_5);
+MODULE_FIRMWARE(FIRMWARE_VCN4_0_6);
+MODULE_FIRMWARE(FIRMWARE_VCN4_0_6_1);
+MODULE_FIRMWARE(FIRMWARE_VCN5_0_0);
+MODULE_FIRMWARE(FIRMWARE_VCN5_0_1);
 
 static void amdgpu_vcn_idle_work_handler(struct work_struct *work);
+static void amdgpu_vcn_reg_dump_fini(struct amdgpu_device *adev);
 
-int amdgpu_vcn_early_init(struct amdgpu_device *adev)
+int amdgpu_vcn_early_init(struct amdgpu_device *adev, int i)
 {
-	char ucode_prefix[30];
-	char fw_name[40];
+	char ucode_prefix[25];
 	int r;
 
+	adev->vcn.inst[i].adev = adev;
+	adev->vcn.inst[i].inst = i;
 	amdgpu_ucode_ip_version_decode(adev, UVD_HWIP, ucode_prefix, sizeof(ucode_prefix));
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s.bin", ucode_prefix);
-	r = amdgpu_ucode_request(adev, &adev->vcn.fw, fw_name);
-	if (r)
-		amdgpu_ucode_release(&adev->vcn.fw);
+
+	if (i != 0 && adev->vcn.per_inst_fw) {
+		r = amdgpu_ucode_request(adev, &adev->vcn.inst[i].fw,
+					 AMDGPU_UCODE_REQUIRED,
+					 "amdgpu/%s_%d.bin", ucode_prefix, i);
+		if (r)
+			amdgpu_ucode_release(&adev->vcn.inst[i].fw);
+	} else {
+		if (!adev->vcn.inst[0].fw) {
+			r = amdgpu_ucode_request(adev, &adev->vcn.inst[0].fw,
+						 AMDGPU_UCODE_REQUIRED,
+						 "amdgpu/%s.bin", ucode_prefix);
+			if (r)
+				amdgpu_ucode_release(&adev->vcn.inst[0].fw);
+		} else {
+			r = 0;
+		}
+		adev->vcn.inst[i].fw = adev->vcn.inst[0].fw;
+	}
 
 	return r;
 }
 
-int amdgpu_vcn_sw_init(struct amdgpu_device *adev)
+int amdgpu_vcn_sw_init(struct amdgpu_device *adev, int i)
 {
 	unsigned long bo_size;
 	const struct common_firmware_header *hdr;
 	unsigned char fw_check;
 	unsigned int fw_shared_size, log_offset;
-	int i, r;
-
-	INIT_DELAYED_WORK(&adev->vcn.idle_work, amdgpu_vcn_idle_work_handler);
-	mutex_init(&adev->vcn.vcn_pg_lock);
-	mutex_init(&adev->vcn.vcn1_jpeg1_workaround);
-	atomic_set(&adev->vcn.total_submission_cnt, 0);
-	for (i = 0; i < adev->vcn.num_vcn_inst; i++)
-		atomic_set(&adev->vcn.inst[i].dpg_enc_submission_cnt, 0);
+	int r;
 
+	mutex_init(&adev->vcn.inst[i].vcn1_jpeg1_workaround);
+	mutex_init(&adev->vcn.inst[i].vcn_pg_lock);
+	mutex_init(&adev->vcn.inst[i].engine_reset_mutex);
+	atomic_set(&adev->vcn.inst[i].total_submission_cnt, 0);
+	INIT_DELAYED_WORK(&adev->vcn.inst[i].idle_work, amdgpu_vcn_idle_work_handler);
+	atomic_set(&adev->vcn.inst[i].dpg_enc_submission_cnt, 0);
 	if ((adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) &&
 	    (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG))
-		adev->vcn.indirect_sram = true;
+		adev->vcn.inst[i].indirect_sram = true;
 
 	/*
 	 * Some Steam Deck's BIOS versions are incompatible with the
@@ -122,18 +150,23 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev)
 	 * Hence, check for these versions here - notice this is
 	 * restricted to Vangogh (Deck's APU).
 	 */
-	if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(3, 0, 2)) {
+	if (amdgpu_ip_version(adev, UVD_HWIP, 0) == IP_VERSION(3, 0, 2)) {
 		const char *bios_ver = dmi_get_system_info(DMI_BIOS_VERSION);
 
 		if (bios_ver && (!strncmp("F7A0113", bios_ver, 7) ||
-		     !strncmp("F7A0114", bios_ver, 7))) {
-			adev->vcn.indirect_sram = false;
+				 !strncmp("F7A0114", bios_ver, 7))) {
+			adev->vcn.inst[i].indirect_sram = false;
 			dev_info(adev->dev,
-				"Steam Deck quirk: indirect SRAM disabled on BIOS %s\n", bios_ver);
+				 "Steam Deck quirk: indirect SRAM disabled on BIOS %s\n", bios_ver);
 		}
 	}
 
-	hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
+	/* from vcn4 and above, only unified queue is used */
+	adev->vcn.inst[i].using_unified_queue =
+		amdgpu_ip_version(adev, UVD_HWIP, 0) >= IP_VERSION(4, 0, 0);
+
+	hdr = (const struct common_firmware_header *)adev->vcn.inst[i].fw->data;
+	adev->vcn.inst[i].fw_version = le32_to_cpu(hdr->ucode_version);
 	adev->vcn.fw_version = le32_to_cpu(hdr->ucode_version);
 
 	/* Bit 20-23, it is encode major and non-zero for new naming convention.
@@ -151,23 +184,27 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev)
 		enc_major = fw_check;
 		dec_ver = (le32_to_cpu(hdr->ucode_version) >> 24) & 0xf;
 		vep = (le32_to_cpu(hdr->ucode_version) >> 28) & 0xf;
-		DRM_INFO("Found VCN firmware Version ENC: %u.%u DEC: %u VEP: %u Revision: %u\n",
-			enc_major, enc_minor, dec_ver, vep, fw_rev);
+		dev_info(adev->dev,
+			 "[VCN instance %d] Found VCN firmware Version ENC: %u.%u DEC: %u VEP: %u Revision: %u\n",
+			 i, enc_major, enc_minor, dec_ver, vep, fw_rev);
 	} else {
 		unsigned int version_major, version_minor, family_id;
 
 		family_id = le32_to_cpu(hdr->ucode_version) & 0xff;
 		version_major = (le32_to_cpu(hdr->ucode_version) >> 24) & 0xff;
 		version_minor = (le32_to_cpu(hdr->ucode_version) >> 8) & 0xff;
-		DRM_INFO("Found VCN firmware Version: %u.%u Family ID: %u\n",
-			version_major, version_minor, family_id);
+		dev_info(adev->dev, "[VCN instance %d] Found VCN firmware Version: %u.%u Family ID: %u\n",
+			 i, version_major, version_minor, family_id);
 	}
 
 	bo_size = AMDGPU_VCN_STACK_SIZE + AMDGPU_VCN_CONTEXT_SIZE;
 	if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP)
 		bo_size += AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
 
-	if (adev->ip_versions[UVD_HWIP][0] >= IP_VERSION(4, 0, 0)){
+	if (amdgpu_ip_version(adev, UVD_HWIP, 0) >= IP_VERSION(5, 0, 0)) {
+		fw_shared_size = AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn5_fw_shared));
+		log_offset = offsetof(struct amdgpu_vcn5_fw_shared, fw_log);
+	} else if (amdgpu_ip_version(adev, UVD_HWIP, 0) >= IP_VERSION(4, 0, 0)) {
 		fw_shared_size = AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn4_fw_shared));
 		log_offset = offsetof(struct amdgpu_vcn4_fw_shared, fw_log);
 	} else {
@@ -180,285 +217,345 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev)
 	if (amdgpu_vcnfw_log)
 		bo_size += AMDGPU_VCNFW_LOG_SIZE;
 
-	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
-		if (adev->vcn.harvest_config & (1 << i))
-			continue;
+	r = amdgpu_bo_create_kernel(adev, bo_size, PAGE_SIZE,
+				    AMDGPU_GEM_DOMAIN_VRAM |
+				    AMDGPU_GEM_DOMAIN_GTT,
+				    &adev->vcn.inst[i].vcpu_bo,
+				    &adev->vcn.inst[i].gpu_addr,
+				    &adev->vcn.inst[i].cpu_addr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to allocate vcn bo\n", r);
+		return r;
+	}
+
+	adev->vcn.inst[i].fw_shared.cpu_addr = adev->vcn.inst[i].cpu_addr +
+		bo_size - fw_shared_size;
+	adev->vcn.inst[i].fw_shared.gpu_addr = adev->vcn.inst[i].gpu_addr +
+		bo_size - fw_shared_size;
+
+	adev->vcn.inst[i].fw_shared.mem_size = fw_shared_size;
 
-		r = amdgpu_bo_create_kernel(adev, bo_size, PAGE_SIZE,
+	if (amdgpu_vcnfw_log) {
+		adev->vcn.inst[i].fw_shared.cpu_addr -= AMDGPU_VCNFW_LOG_SIZE;
+		adev->vcn.inst[i].fw_shared.gpu_addr -= AMDGPU_VCNFW_LOG_SIZE;
+		adev->vcn.inst[i].fw_shared.log_offset = log_offset;
+	}
+
+	if (adev->vcn.inst[i].indirect_sram) {
+		r = amdgpu_bo_create_kernel(adev, 64 * 2 * 4, PAGE_SIZE,
 					    AMDGPU_GEM_DOMAIN_VRAM |
 					    AMDGPU_GEM_DOMAIN_GTT,
-					    &adev->vcn.inst[i].vcpu_bo,
-					    &adev->vcn.inst[i].gpu_addr,
-					    &adev->vcn.inst[i].cpu_addr);
+					    &adev->vcn.inst[i].dpg_sram_bo,
+					    &adev->vcn.inst[i].dpg_sram_gpu_addr,
+					    &adev->vcn.inst[i].dpg_sram_cpu_addr);
 		if (r) {
-			dev_err(adev->dev, "(%d) failed to allocate vcn bo\n", r);
+			dev_err(adev->dev, "VCN %d (%d) failed to allocate DPG bo\n", i, r);
 			return r;
 		}
-
-		adev->vcn.inst[i].fw_shared.cpu_addr = adev->vcn.inst[i].cpu_addr +
-				bo_size - fw_shared_size;
-		adev->vcn.inst[i].fw_shared.gpu_addr = adev->vcn.inst[i].gpu_addr +
-				bo_size - fw_shared_size;
-
-		adev->vcn.inst[i].fw_shared.mem_size = fw_shared_size;
-
-		if (amdgpu_vcnfw_log) {
-			adev->vcn.inst[i].fw_shared.cpu_addr -= AMDGPU_VCNFW_LOG_SIZE;
-			adev->vcn.inst[i].fw_shared.gpu_addr -= AMDGPU_VCNFW_LOG_SIZE;
-			adev->vcn.inst[i].fw_shared.log_offset = log_offset;
-		}
-
-		if (adev->vcn.indirect_sram) {
-			r = amdgpu_bo_create_kernel(adev, 64 * 2 * 4, PAGE_SIZE,
-					AMDGPU_GEM_DOMAIN_VRAM |
-					AMDGPU_GEM_DOMAIN_GTT,
-					&adev->vcn.inst[i].dpg_sram_bo,
-					&adev->vcn.inst[i].dpg_sram_gpu_addr,
-					&adev->vcn.inst[i].dpg_sram_cpu_addr);
-			if (r) {
-				dev_err(adev->dev, "VCN %d (%d) failed to allocate DPG bo\n", i, r);
-				return r;
-			}
-		}
 	}
 
 	return 0;
 }
 
-int amdgpu_vcn_sw_fini(struct amdgpu_device *adev)
+void amdgpu_vcn_sw_fini(struct amdgpu_device *adev, int i)
 {
-	int i, j;
-
-	for (j = 0; j < adev->vcn.num_vcn_inst; ++j) {
-		if (adev->vcn.harvest_config & (1 << j))
-			continue;
+	int j;
 
-		if (adev->vcn.indirect_sram) {
-			amdgpu_bo_free_kernel(&adev->vcn.inst[j].dpg_sram_bo,
-						  &adev->vcn.inst[j].dpg_sram_gpu_addr,
-						  (void **)&adev->vcn.inst[j].dpg_sram_cpu_addr);
-		}
-		kvfree(adev->vcn.inst[j].saved_bo);
+	if (adev->vcn.harvest_config & (1 << i))
+		return;
 
-		amdgpu_bo_free_kernel(&adev->vcn.inst[j].vcpu_bo,
-					  &adev->vcn.inst[j].gpu_addr,
-					  (void **)&adev->vcn.inst[j].cpu_addr);
+	amdgpu_bo_free_kernel(
+		&adev->vcn.inst[i].dpg_sram_bo,
+		&adev->vcn.inst[i].dpg_sram_gpu_addr,
+		(void **)&adev->vcn.inst[i].dpg_sram_cpu_addr);
 
-		amdgpu_ring_fini(&adev->vcn.inst[j].ring_dec);
+	kvfree(adev->vcn.inst[i].saved_bo);
 
-		for (i = 0; i < adev->vcn.num_enc_rings; ++i)
-			amdgpu_ring_fini(&adev->vcn.inst[j].ring_enc[i]);
-	}
+	amdgpu_bo_free_kernel(&adev->vcn.inst[i].vcpu_bo,
+			      &adev->vcn.inst[i].gpu_addr,
+			      (void **)&adev->vcn.inst[i].cpu_addr);
 
-	amdgpu_ucode_release(&adev->vcn.fw);
-	mutex_destroy(&adev->vcn.vcn1_jpeg1_workaround);
-	mutex_destroy(&adev->vcn.vcn_pg_lock);
+	amdgpu_ring_fini(&adev->vcn.inst[i].ring_dec);
 
-	return 0;
-}
+	for (j = 0; j < adev->vcn.inst[i].num_enc_rings; ++j)
+		amdgpu_ring_fini(&adev->vcn.inst[i].ring_enc[j]);
 
-/* from vcn4 and above, only unified queue is used */
-static bool amdgpu_vcn_using_unified_queue(struct amdgpu_ring *ring)
-{
-	struct amdgpu_device *adev = ring->adev;
-	bool ret = false;
+	if (adev->vcn.per_inst_fw) {
+		amdgpu_ucode_release(&adev->vcn.inst[i].fw);
+	} else {
+		amdgpu_ucode_release(&adev->vcn.inst[0].fw);
+		adev->vcn.inst[i].fw = NULL;
+	}
 
-	if (adev->ip_versions[UVD_HWIP][0] >= IP_VERSION(4, 0, 0))
-		ret = true;
+	if (adev->vcn.reg_list)
+		amdgpu_vcn_reg_dump_fini(adev);
 
-	return ret;
+	mutex_destroy(&adev->vcn.inst[i].vcn_pg_lock);
+	mutex_destroy(&adev->vcn.inst[i].vcn1_jpeg1_workaround);
 }
 
 bool amdgpu_vcn_is_disabled_vcn(struct amdgpu_device *adev, enum vcn_ring_type type, uint32_t vcn_instance)
 {
 	bool ret = false;
-	int vcn_config = adev->vcn.vcn_config[vcn_instance];
+	int vcn_config = adev->vcn.inst[vcn_instance].vcn_config;
 
-	if ((type == VCN_ENCODE_RING) && (vcn_config & VCN_BLOCK_ENCODE_DISABLE_MASK)) {
+	if ((type == VCN_ENCODE_RING) && (vcn_config & VCN_BLOCK_ENCODE_DISABLE_MASK))
 		ret = true;
-	} else if ((type == VCN_DECODE_RING) && (vcn_config & VCN_BLOCK_DECODE_DISABLE_MASK)) {
+	else if ((type == VCN_DECODE_RING) && (vcn_config & VCN_BLOCK_DECODE_DISABLE_MASK))
 		ret = true;
-	} else if ((type == VCN_UNIFIED_RING) && (vcn_config & VCN_BLOCK_QUEUE_DISABLE_MASK)) {
+	else if ((type == VCN_UNIFIED_RING) && (vcn_config & VCN_BLOCK_QUEUE_DISABLE_MASK))
 		ret = true;
-	}
 
 	return ret;
 }
 
-int amdgpu_vcn_suspend(struct amdgpu_device *adev)
+static int amdgpu_vcn_save_vcpu_bo_inst(struct amdgpu_device *adev, int i)
 {
-	unsigned size;
+	unsigned int size;
 	void *ptr;
-	int i, idx;
+	int idx;
 
-	cancel_delayed_work_sync(&adev->vcn.idle_work);
+	if (adev->vcn.harvest_config & (1 << i))
+		return 0;
+	if (adev->vcn.inst[i].vcpu_bo == NULL)
+		return 0;
 
-	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
-		if (adev->vcn.harvest_config & (1 << i))
-			continue;
-		if (adev->vcn.inst[i].vcpu_bo == NULL)
-			return 0;
+	size = amdgpu_bo_size(adev->vcn.inst[i].vcpu_bo);
+	ptr = adev->vcn.inst[i].cpu_addr;
 
-		size = amdgpu_bo_size(adev->vcn.inst[i].vcpu_bo);
-		ptr = adev->vcn.inst[i].cpu_addr;
+	adev->vcn.inst[i].saved_bo = kvmalloc(size, GFP_KERNEL);
+	if (!adev->vcn.inst[i].saved_bo)
+		return -ENOMEM;
 
-		adev->vcn.inst[i].saved_bo = kvmalloc(size, GFP_KERNEL);
-		if (!adev->vcn.inst[i].saved_bo)
-			return -ENOMEM;
+	if (drm_dev_enter(adev_to_drm(adev), &idx)) {
+		memcpy_fromio(adev->vcn.inst[i].saved_bo, ptr, size);
+		drm_dev_exit(idx);
+	}
 
-		if (drm_dev_enter(adev_to_drm(adev), &idx)) {
-			memcpy_fromio(adev->vcn.inst[i].saved_bo, ptr, size);
-			drm_dev_exit(idx);
-		}
+	return 0;
+}
+
+int amdgpu_vcn_save_vcpu_bo(struct amdgpu_device *adev)
+{
+	int ret, i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		ret = amdgpu_vcn_save_vcpu_bo_inst(adev, i);
+		if (ret)
+			return ret;
 	}
+
 	return 0;
 }
 
-int amdgpu_vcn_resume(struct amdgpu_device *adev)
+int amdgpu_vcn_suspend(struct amdgpu_device *adev, int i)
 {
-	unsigned size;
+	bool in_ras_intr = amdgpu_ras_intr_triggered();
+
+	if (adev->vcn.harvest_config & (1 << i))
+		return 0;
+
+	/* err_event_athub and dpc recovery will corrupt VCPU buffer, so we need to
+	 * restore fw data and clear buffer in amdgpu_vcn_resume() */
+	if (in_ras_intr || adev->pcie_reset_ctx.in_link_reset)
+		return 0;
+
+	return amdgpu_vcn_save_vcpu_bo_inst(adev, i);
+}
+
+int amdgpu_vcn_resume(struct amdgpu_device *adev, int i)
+{
+	unsigned int size;
 	void *ptr;
-	int i, idx;
+	int idx;
 
-	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
-		if (adev->vcn.harvest_config & (1 << i))
-			continue;
-		if (adev->vcn.inst[i].vcpu_bo == NULL)
-			return -EINVAL;
+	if (adev->vcn.harvest_config & (1 << i))
+		return 0;
+	if (adev->vcn.inst[i].vcpu_bo == NULL)
+		return -EINVAL;
 
-		size = amdgpu_bo_size(adev->vcn.inst[i].vcpu_bo);
-		ptr = adev->vcn.inst[i].cpu_addr;
+	size = amdgpu_bo_size(adev->vcn.inst[i].vcpu_bo);
+	ptr = adev->vcn.inst[i].cpu_addr;
 
-		if (adev->vcn.inst[i].saved_bo != NULL) {
+	if (adev->vcn.inst[i].saved_bo != NULL) {
+		if (drm_dev_enter(adev_to_drm(adev), &idx)) {
+			memcpy_toio(ptr, adev->vcn.inst[i].saved_bo, size);
+			drm_dev_exit(idx);
+		}
+		kvfree(adev->vcn.inst[i].saved_bo);
+		adev->vcn.inst[i].saved_bo = NULL;
+	} else {
+		const struct common_firmware_header *hdr;
+		unsigned int offset;
+
+		hdr = (const struct common_firmware_header *)adev->vcn.inst[i].fw->data;
+		if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
+			offset = le32_to_cpu(hdr->ucode_array_offset_bytes);
 			if (drm_dev_enter(adev_to_drm(adev), &idx)) {
-				memcpy_toio(ptr, adev->vcn.inst[i].saved_bo, size);
+				memcpy_toio(adev->vcn.inst[i].cpu_addr,
+					    adev->vcn.inst[i].fw->data + offset,
+					    le32_to_cpu(hdr->ucode_size_bytes));
 				drm_dev_exit(idx);
 			}
-			kvfree(adev->vcn.inst[i].saved_bo);
-			adev->vcn.inst[i].saved_bo = NULL;
-		} else {
-			const struct common_firmware_header *hdr;
-			unsigned offset;
-
-			hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
-			if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
-				offset = le32_to_cpu(hdr->ucode_array_offset_bytes);
-				if (drm_dev_enter(adev_to_drm(adev), &idx)) {
-					memcpy_toio(adev->vcn.inst[i].cpu_addr, adev->vcn.fw->data + offset,
-						    le32_to_cpu(hdr->ucode_size_bytes));
-					drm_dev_exit(idx);
-				}
-				size -= le32_to_cpu(hdr->ucode_size_bytes);
-				ptr += le32_to_cpu(hdr->ucode_size_bytes);
-			}
-			memset_io(ptr, 0, size);
+			size -= le32_to_cpu(hdr->ucode_size_bytes);
+			ptr += le32_to_cpu(hdr->ucode_size_bytes);
 		}
+		memset_io(ptr, 0, size);
 	}
+
 	return 0;
 }
 
-static void amdgpu_vcn_idle_work_handler(struct work_struct *work)
+void amdgpu_vcn_get_profile(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev =
-		container_of(work, struct amdgpu_device, vcn.idle_work.work);
-	unsigned int fences = 0, fence[AMDGPU_MAX_VCN_INSTANCES] = {0};
-	unsigned int i, j;
-	int r = 0;
+	int r;
 
-	for (j = 0; j < adev->vcn.num_vcn_inst; ++j) {
-		if (adev->vcn.harvest_config & (1 << j))
-			continue;
+	mutex_lock(&adev->vcn.workload_profile_mutex);
+
+	if (adev->vcn.workload_profile_active) {
+		mutex_unlock(&adev->vcn.workload_profile_mutex);
+		return;
+	}
+	r = amdgpu_dpm_switch_power_profile(adev, PP_SMC_POWER_PROFILE_VIDEO,
+					    true);
+	if (r)
+		dev_warn(adev->dev,
+			 "(%d) failed to enable video power profile mode\n", r);
+	else
+		adev->vcn.workload_profile_active = true;
+	mutex_unlock(&adev->vcn.workload_profile_mutex);
+}
 
-		for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
-			fence[j] += amdgpu_fence_count_emitted(&adev->vcn.inst[j].ring_enc[i]);
+void amdgpu_vcn_put_profile(struct amdgpu_device *adev)
+{
+	bool pg = true;
+	int r, i;
+
+	mutex_lock(&adev->vcn.workload_profile_mutex);
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		if (adev->vcn.inst[i].cur_state != AMD_PG_STATE_GATE) {
+			pg = false;
+			break;
 		}
+	}
 
-		if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)	{
-			struct dpg_pause_state new_state;
+	if (pg) {
+		r = amdgpu_dpm_switch_power_profile(
+			adev, PP_SMC_POWER_PROFILE_VIDEO, false);
+		if (r)
+			dev_warn(
+				adev->dev,
+				"(%d) failed to disable video power profile mode\n",
+				r);
+		else
+			adev->vcn.workload_profile_active = false;
+	}
 
-			if (fence[j] ||
-				unlikely(atomic_read(&adev->vcn.inst[j].dpg_enc_submission_cnt)))
-				new_state.fw_based = VCN_DPG_STATE__PAUSE;
-			else
-				new_state.fw_based = VCN_DPG_STATE__UNPAUSE;
+	mutex_unlock(&adev->vcn.workload_profile_mutex);
+}
 
-			adev->vcn.pause_dpg_mode(adev, j, &new_state);
-		}
+static void amdgpu_vcn_idle_work_handler(struct work_struct *work)
+{
+	struct amdgpu_vcn_inst *vcn_inst =
+		container_of(work, struct amdgpu_vcn_inst, idle_work.work);
+	struct amdgpu_device *adev = vcn_inst->adev;
+	unsigned int fences = 0, fence[AMDGPU_MAX_VCN_INSTANCES] = {0};
+	unsigned int i = vcn_inst->inst, j;
 
-		fence[j] += amdgpu_fence_count_emitted(&adev->vcn.inst[j].ring_dec);
-		fences += fence[j];
+	if (adev->vcn.harvest_config & (1 << i))
+		return;
+
+	for (j = 0; j < adev->vcn.inst[i].num_enc_rings; ++j)
+		fence[i] += amdgpu_fence_count_emitted(&vcn_inst->ring_enc[j]);
+
+	/* Only set DPG pause for VCN3 or below, VCN4 and above will be handled by FW */
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG &&
+	    !adev->vcn.inst[i].using_unified_queue) {
+		struct dpg_pause_state new_state;
+
+		if (fence[i] ||
+		    unlikely(atomic_read(&vcn_inst->dpg_enc_submission_cnt)))
+			new_state.fw_based = VCN_DPG_STATE__PAUSE;
+		else
+			new_state.fw_based = VCN_DPG_STATE__UNPAUSE;
+
+		adev->vcn.inst[i].pause_dpg_mode(vcn_inst, &new_state);
 	}
 
-	if (!fences && !atomic_read(&adev->vcn.total_submission_cnt)) {
-		amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN,
-		       AMD_PG_STATE_GATE);
-		r = amdgpu_dpm_switch_power_profile(adev, PP_SMC_POWER_PROFILE_VIDEO,
-				false);
-		if (r)
-			dev_warn(adev->dev, "(%d) failed to disable video power profile mode\n", r);
+	fence[i] += amdgpu_fence_count_emitted(&vcn_inst->ring_dec);
+	fences += fence[i];
+
+	if (!fences && !atomic_read(&vcn_inst->total_submission_cnt)) {
+		mutex_lock(&vcn_inst->vcn_pg_lock);
+		vcn_inst->set_pg_state(vcn_inst, AMD_PG_STATE_GATE);
+		mutex_unlock(&vcn_inst->vcn_pg_lock);
+		amdgpu_vcn_put_profile(adev);
+
 	} else {
-		schedule_delayed_work(&adev->vcn.idle_work, VCN_IDLE_TIMEOUT);
+		schedule_delayed_work(&vcn_inst->idle_work, VCN_IDLE_TIMEOUT);
 	}
 }
 
 void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	int r = 0;
+	struct amdgpu_vcn_inst *vcn_inst = &adev->vcn.inst[ring->me];
 
-	atomic_inc(&adev->vcn.total_submission_cnt);
+	atomic_inc(&vcn_inst->total_submission_cnt);
 
-	if (!cancel_delayed_work_sync(&adev->vcn.idle_work)) {
-		r = amdgpu_dpm_switch_power_profile(adev, PP_SMC_POWER_PROFILE_VIDEO,
-				true);
-		if (r)
-			dev_warn(adev->dev, "(%d) failed to switch to video power profile mode\n", r);
-	}
+	cancel_delayed_work_sync(&vcn_inst->idle_work);
 
-	mutex_lock(&adev->vcn.vcn_pg_lock);
-	amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN,
-	       AMD_PG_STATE_UNGATE);
+	mutex_lock(&vcn_inst->vcn_pg_lock);
+	vcn_inst->set_pg_state(vcn_inst, AMD_PG_STATE_UNGATE);
 
-	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)	{
+	/* Only set DPG pause for VCN3 or below, VCN4 and above will be handled by FW */
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG &&
+	    !vcn_inst->using_unified_queue) {
 		struct dpg_pause_state new_state;
 
 		if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC) {
-			atomic_inc(&adev->vcn.inst[ring->me].dpg_enc_submission_cnt);
+			atomic_inc(&vcn_inst->dpg_enc_submission_cnt);
 			new_state.fw_based = VCN_DPG_STATE__PAUSE;
 		} else {
 			unsigned int fences = 0;
 			unsigned int i;
 
-			for (i = 0; i < adev->vcn.num_enc_rings; ++i)
-				fences += amdgpu_fence_count_emitted(&adev->vcn.inst[ring->me].ring_enc[i]);
+			for (i = 0; i < vcn_inst->num_enc_rings; ++i)
+				fences += amdgpu_fence_count_emitted(&vcn_inst->ring_enc[i]);
 
-			if (fences || atomic_read(&adev->vcn.inst[ring->me].dpg_enc_submission_cnt))
+			if (fences || atomic_read(&vcn_inst->dpg_enc_submission_cnt))
 				new_state.fw_based = VCN_DPG_STATE__PAUSE;
 			else
 				new_state.fw_based = VCN_DPG_STATE__UNPAUSE;
 		}
 
-		adev->vcn.pause_dpg_mode(adev, ring->me, &new_state);
+		vcn_inst->pause_dpg_mode(vcn_inst, &new_state);
 	}
-	mutex_unlock(&adev->vcn.vcn_pg_lock);
+	mutex_unlock(&vcn_inst->vcn_pg_lock);
+	amdgpu_vcn_get_profile(adev);
 }
 
 void amdgpu_vcn_ring_end_use(struct amdgpu_ring *ring)
 {
+	struct amdgpu_device *adev = ring->adev;
+
+	/* Only set DPG pause for VCN3 or below, VCN4 and above will be handled by FW */
 	if (ring->adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG &&
-		ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC)
+	    ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC &&
+	    !adev->vcn.inst[ring->me].using_unified_queue)
 		atomic_dec(&ring->adev->vcn.inst[ring->me].dpg_enc_submission_cnt);
 
-	atomic_dec(&ring->adev->vcn.total_submission_cnt);
+	atomic_dec(&ring->adev->vcn.inst[ring->me].total_submission_cnt);
 
-	schedule_delayed_work(&ring->adev->vcn.idle_work, VCN_IDLE_TIMEOUT);
+	schedule_delayed_work(&ring->adev->vcn.inst[ring->me].idle_work,
+			      VCN_IDLE_TIMEOUT);
 }
 
 int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
 	uint32_t tmp = 0;
-	unsigned i;
+	unsigned int i;
 	int r;
 
 	/* VCN in SRIOV does not support direct register read/write */
@@ -469,7 +566,7 @@ int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring)
 	r = amdgpu_ring_alloc(ring, 3);
 	if (r)
 		return r;
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.scratch9, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.scratch9, 0));
 	amdgpu_ring_write(ring, 0xDEADBEEF);
 	amdgpu_ring_commit(ring);
 	for (i = 0; i < adev->usec_timeout; i++) {
@@ -529,19 +626,19 @@ static int amdgpu_vcn_dec_send_msg(struct amdgpu_ring *ring,
 
 	r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL,
 				     64, AMDGPU_IB_POOL_DIRECT,
-				     &job);
+				     &job, AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		goto err;
 
 	ib = &job->ibs[0];
-	ib->ptr[0] = PACKET0(adev->vcn.internal.data0, 0);
+	ib->ptr[0] = PACKET0(adev->vcn.inst[ring->me].internal.data0, 0);
 	ib->ptr[1] = addr;
-	ib->ptr[2] = PACKET0(adev->vcn.internal.data1, 0);
+	ib->ptr[2] = PACKET0(adev->vcn.inst[ring->me].internal.data1, 0);
 	ib->ptr[3] = addr >> 32;
-	ib->ptr[4] = PACKET0(adev->vcn.internal.cmd, 0);
+	ib->ptr[4] = PACKET0(adev->vcn.inst[ring->me].internal.cmd, 0);
 	ib->ptr[5] = 0;
 	for (i = 6; i < 16; i += 2) {
-		ib->ptr[i] = PACKET0(adev->vcn.internal.nop, 0);
+		ib->ptr[i] = PACKET0(adev->vcn.inst[ring->me].internal.nop, 0);
 		ib->ptr[i+1] = 0;
 	}
 	ib->length_dw = 16;
@@ -550,7 +647,7 @@ static int amdgpu_vcn_dec_send_msg(struct amdgpu_ring *ring,
 	if (r)
 		goto err_free;
 
-	amdgpu_ib_free(adev, ib_msg, f);
+	amdgpu_ib_free(ib_msg, f);
 
 	if (fence)
 		*fence = dma_fence_get(f);
@@ -561,7 +658,7 @@ static int amdgpu_vcn_dec_send_msg(struct amdgpu_ring *ring,
 err_free:
 	amdgpu_job_free(job);
 err:
-	amdgpu_ib_free(adev, ib_msg, f);
+	amdgpu_ib_free(ib_msg, f);
 	return r;
 }
 
@@ -700,17 +797,16 @@ static int amdgpu_vcn_dec_sw_send_msg(struct amdgpu_ring *ring,
 	struct amdgpu_job *job;
 	struct amdgpu_ib *ib;
 	uint64_t addr = AMDGPU_GPU_PAGE_ALIGN(ib_msg->gpu_addr);
-	bool sq = amdgpu_vcn_using_unified_queue(ring);
 	uint32_t *ib_checksum;
 	uint32_t ib_pack_in_dw;
 	int i, r;
 
-	if (sq)
+	if (adev->vcn.inst[ring->me].using_unified_queue)
 		ib_size_dw += 8;
 
 	r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL,
 				     ib_size_dw * 4, AMDGPU_IB_POOL_DIRECT,
-				     &job);
+				     &job, AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		goto err;
 
@@ -718,7 +814,7 @@ static int amdgpu_vcn_dec_sw_send_msg(struct amdgpu_ring *ring,
 	ib->length_dw = 0;
 
 	/* single queue headers */
-	if (sq) {
+	if (adev->vcn.inst[ring->me].using_unified_queue) {
 		ib_pack_in_dw = sizeof(struct amdgpu_vcn_decode_buffer) / sizeof(uint32_t)
 						+ 4 + 2; /* engine info + decoding ib in dw */
 		ib_checksum = amdgpu_vcn_unified_ring_ib_header(ib, ib_pack_in_dw, false);
@@ -737,14 +833,14 @@ static int amdgpu_vcn_dec_sw_send_msg(struct amdgpu_ring *ring,
 	for (i = ib->length_dw; i < ib_size_dw; ++i)
 		ib->ptr[i] = 0x0;
 
-	if (sq)
+	if (adev->vcn.inst[ring->me].using_unified_queue)
 		amdgpu_vcn_unified_ring_ib_checksum(&ib_checksum, ib_pack_in_dw);
 
 	r = amdgpu_job_submit_direct(job, ring, &f);
 	if (r)
 		goto err_free;
 
-	amdgpu_ib_free(adev, ib_msg, f);
+	amdgpu_ib_free(ib_msg, f);
 
 	if (fence)
 		*fence = dma_fence_get(f);
@@ -755,7 +851,7 @@ static int amdgpu_vcn_dec_sw_send_msg(struct amdgpu_ring *ring,
 err_free:
 	amdgpu_job_free(job);
 err:
-	amdgpu_ib_free(adev, ib_msg, f);
+	amdgpu_ib_free(ib_msg, f);
 	return r;
 }
 
@@ -795,7 +891,7 @@ int amdgpu_vcn_enc_ring_test_ring(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
 	uint32_t rptr;
-	unsigned i;
+	unsigned int i;
 	int r;
 
 	if (amdgpu_sriov_vf(adev))
@@ -827,20 +923,20 @@ static int amdgpu_vcn_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t hand
 					 struct dma_fence **fence)
 {
 	unsigned int ib_size_dw = 16;
+	struct amdgpu_device *adev = ring->adev;
 	struct amdgpu_job *job;
 	struct amdgpu_ib *ib;
 	struct dma_fence *f = NULL;
 	uint32_t *ib_checksum = NULL;
 	uint64_t addr;
-	bool sq = amdgpu_vcn_using_unified_queue(ring);
 	int i, r;
 
-	if (sq)
+	if (adev->vcn.inst[ring->me].using_unified_queue)
 		ib_size_dw += 8;
 
 	r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL,
 				     ib_size_dw * 4, AMDGPU_IB_POOL_DIRECT,
-				     &job);
+				     &job, AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		return r;
 
@@ -849,7 +945,7 @@ static int amdgpu_vcn_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t hand
 
 	ib->length_dw = 0;
 
-	if (sq)
+	if (adev->vcn.inst[ring->me].using_unified_queue)
 		ib_checksum = amdgpu_vcn_unified_ring_ib_header(ib, 0x11, true);
 
 	ib->ptr[ib->length_dw++] = 0x00000018;
@@ -857,7 +953,7 @@ static int amdgpu_vcn_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t hand
 	ib->ptr[ib->length_dw++] = handle;
 	ib->ptr[ib->length_dw++] = upper_32_bits(addr);
 	ib->ptr[ib->length_dw++] = addr;
-	ib->ptr[ib->length_dw++] = 0x0000000b;
+	ib->ptr[ib->length_dw++] = 0x00000000;
 
 	ib->ptr[ib->length_dw++] = 0x00000014;
 	ib->ptr[ib->length_dw++] = 0x00000002; /* task info */
@@ -871,7 +967,7 @@ static int amdgpu_vcn_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t hand
 	for (i = ib->length_dw; i < ib_size_dw; ++i)
 		ib->ptr[i] = 0x0;
 
-	if (sq)
+	if (adev->vcn.inst[ring->me].using_unified_queue)
 		amdgpu_vcn_unified_ring_ib_checksum(&ib_checksum, 0x11);
 
 	r = amdgpu_job_submit_direct(job, ring, &f);
@@ -894,20 +990,20 @@ static int amdgpu_vcn_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han
 					  struct dma_fence **fence)
 {
 	unsigned int ib_size_dw = 16;
+	struct amdgpu_device *adev = ring->adev;
 	struct amdgpu_job *job;
 	struct amdgpu_ib *ib;
 	struct dma_fence *f = NULL;
 	uint32_t *ib_checksum = NULL;
 	uint64_t addr;
-	bool sq = amdgpu_vcn_using_unified_queue(ring);
 	int i, r;
 
-	if (sq)
+	if (adev->vcn.inst[ring->me].using_unified_queue)
 		ib_size_dw += 8;
 
 	r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL,
 				     ib_size_dw * 4, AMDGPU_IB_POOL_DIRECT,
-				     &job);
+				     &job, AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		return r;
 
@@ -916,7 +1012,7 @@ static int amdgpu_vcn_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han
 
 	ib->length_dw = 0;
 
-	if (sq)
+	if (adev->vcn.inst[ring->me].using_unified_queue)
 		ib_checksum = amdgpu_vcn_unified_ring_ib_header(ib, 0x11, true);
 
 	ib->ptr[ib->length_dw++] = 0x00000018;
@@ -924,7 +1020,7 @@ static int amdgpu_vcn_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han
 	ib->ptr[ib->length_dw++] = handle;
 	ib->ptr[ib->length_dw++] = upper_32_bits(addr);
 	ib->ptr[ib->length_dw++] = addr;
-	ib->ptr[ib->length_dw++] = 0x0000000b;
+	ib->ptr[ib->length_dw++] = 0x00000000;
 
 	ib->ptr[ib->length_dw++] = 0x00000014;
 	ib->ptr[ib->length_dw++] = 0x00000002;
@@ -938,7 +1034,7 @@ static int amdgpu_vcn_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han
 	for (i = ib->length_dw; i < ib_size_dw; ++i)
 		ib->ptr[i] = 0x0;
 
-	if (sq)
+	if (adev->vcn.inst[ring->me].using_unified_queue)
 		amdgpu_vcn_unified_ring_ib_checksum(&ib_checksum, 0x11);
 
 	r = amdgpu_job_submit_direct(job, ring, &f);
@@ -985,7 +1081,7 @@ int amdgpu_vcn_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = 0;
 
 error:
-	amdgpu_ib_free(adev, &ib, fence);
+	amdgpu_ib_free(&ib, fence);
 	dma_fence_put(fence);
 
 	return r;
@@ -993,11 +1089,15 @@ error:
 
 int amdgpu_vcn_unified_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 {
+	struct amdgpu_device *adev = ring->adev;
 	long r;
 
-	r = amdgpu_vcn_enc_ring_test_ib(ring, timeout);
-	if (r)
-		goto error;
+	if ((amdgpu_ip_version(adev, UVD_HWIP, 0) != IP_VERSION(4, 0, 3)) &&
+	    (amdgpu_ip_version(adev, UVD_HWIP, 0) != IP_VERSION(5, 0, 1))) {
+		r = amdgpu_vcn_enc_ring_test_ib(ring, timeout);
+		if (r)
+			goto error;
+	}
 
 	r =  amdgpu_vcn_dec_sw_ring_test_ib(ring, timeout);
 
@@ -1007,7 +1107,7 @@ error:
 
 enum amdgpu_ring_priority_level amdgpu_vcn_get_enc_ring_prio(int ring)
 {
-	switch(ring) {
+	switch (ring) {
 	case 0:
 		return AMDGPU_RING_PRIO_0;
 	case 1:
@@ -1019,30 +1119,32 @@ enum amdgpu_ring_priority_level amdgpu_vcn_get_enc_ring_prio(int ring)
 	}
 }
 
-void amdgpu_vcn_setup_ucode(struct amdgpu_device *adev)
+void amdgpu_vcn_setup_ucode(struct amdgpu_device *adev, int i)
 {
-	int i;
 	unsigned int idx;
 
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 		const struct common_firmware_header *hdr;
-		hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
 
-		for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
-			if (adev->vcn.harvest_config & (1 << i))
-				continue;
-			/* currently only support 2 FW instances */
-			if (i >= 2) {
-				dev_info(adev->dev, "More then 2 VCN FW instances!\n");
-				break;
-			}
-			idx = AMDGPU_UCODE_ID_VCN + i;
-			adev->firmware.ucode[idx].ucode_id = idx;
-			adev->firmware.ucode[idx].fw = adev->vcn.fw;
-			adev->firmware.fw_size +=
-				ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE);
+		if (adev->vcn.harvest_config & (1 << i))
+			return;
+
+		if ((amdgpu_ip_version(adev, UVD_HWIP, 0) == IP_VERSION(4, 0, 3) ||
+		     amdgpu_ip_version(adev, UVD_HWIP, 0) == IP_VERSION(5, 0, 1))
+		    && (i > 0))
+			return;
+
+		hdr = (const struct common_firmware_header *)adev->vcn.inst[i].fw->data;
+		/* currently only support 2 FW instances */
+		if (i >= 2) {
+			dev_info(adev->dev, "More then 2 VCN FW instances!\n");
+			return;
 		}
-		dev_info(adev->dev, "Will use PSP to load VCN firmware\n");
+		idx = AMDGPU_UCODE_ID_VCN + i;
+		adev->firmware.ucode[idx].ucode_id = idx;
+		adev->firmware.ucode[idx].fw = adev->vcn.inst[i].fw;
+		adev->firmware.fw_size +=
+			ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE);
 	}
 }
 
@@ -1051,11 +1153,11 @@ void amdgpu_vcn_setup_ucode(struct amdgpu_device *adev)
  */
 #if defined(CONFIG_DEBUG_FS)
 static ssize_t amdgpu_debugfs_vcn_fwlog_read(struct file *f, char __user *buf,
-                                             size_t size, loff_t *pos)
+					     size_t size, loff_t *pos)
 {
 	struct amdgpu_vcn_inst *vcn;
 	void *log_buf;
-	volatile struct amdgpu_vcn_fwlog *plog;
+	struct amdgpu_vcn_fwlog *plog;
 	unsigned int read_pos, write_pos, available, i, read_bytes = 0;
 	unsigned int read_num[2] = {0};
 
@@ -1068,7 +1170,7 @@ static ssize_t amdgpu_debugfs_vcn_fwlog_read(struct file *f, char __user *buf,
 
 	log_buf = vcn->fw_shared.cpu_addr + vcn->fw_shared.mem_size;
 
-	plog = (volatile struct amdgpu_vcn_fwlog *)log_buf;
+	plog = (struct amdgpu_vcn_fwlog *)log_buf;
 	read_pos = plog->rptr;
 	write_pos = plog->wptr;
 
@@ -1080,7 +1182,7 @@ static ssize_t amdgpu_debugfs_vcn_fwlog_read(struct file *f, char __user *buf,
 
 	if (write_pos > read_pos) {
 		available = write_pos - read_pos;
-		read_num[0] = min(size, (size_t)available);
+		read_num[0] = min_t(size_t, size, available);
 	} else {
 		read_num[0] = AMDGPU_VCNFW_LOG_SIZE - read_pos;
 		available = read_num[0] + write_pos - plog->header_size;
@@ -1097,7 +1199,7 @@ static ssize_t amdgpu_debugfs_vcn_fwlog_read(struct file *f, char __user *buf,
 			if (read_pos == AMDGPU_VCNFW_LOG_SIZE)
 				read_pos = plog->header_size;
 			if (read_num[i] == copy_to_user((buf + read_bytes),
-			                                (log_buf + read_pos), read_num[i]))
+							(log_buf + read_pos), read_num[i]))
 				return -EFAULT;
 
 			read_bytes += read_num[i];
@@ -1118,7 +1220,7 @@ static const struct file_operations amdgpu_debugfs_vcnfwlog_fops = {
 #endif
 
 void amdgpu_debugfs_vcn_fwlog_init(struct amdgpu_device *adev, uint8_t i,
-                                   struct amdgpu_vcn_inst *vcn)
+				   struct amdgpu_vcn_inst *vcn)
 {
 #if defined(CONFIG_DEBUG_FS)
 	struct drm_minor *minor = adev_to_drm(adev)->primary;
@@ -1126,7 +1228,7 @@ void amdgpu_debugfs_vcn_fwlog_init(struct amdgpu_device *adev, uint8_t i,
 	char name[32];
 
 	sprintf(name, "amdgpu_vcn_%d_fwlog", i);
-	debugfs_create_file_size(name, S_IFREG | S_IRUGO, root, vcn,
+	debugfs_create_file_size(name, S_IFREG | 0444, root, vcn,
 				 &amdgpu_debugfs_vcnfwlog_fops,
 				 AMDGPU_VCNFW_LOG_SIZE);
 #endif
@@ -1135,12 +1237,12 @@ void amdgpu_debugfs_vcn_fwlog_init(struct amdgpu_device *adev, uint8_t i,
 void amdgpu_vcn_fwlog_init(struct amdgpu_vcn_inst *vcn)
 {
 #if defined(CONFIG_DEBUG_FS)
-	volatile uint32_t *flag = vcn->fw_shared.cpu_addr;
+	uint32_t *flag = vcn->fw_shared.cpu_addr;
 	void *fw_log_cpu_addr = vcn->fw_shared.cpu_addr + vcn->fw_shared.mem_size;
 	uint64_t fw_log_gpu_addr = vcn->fw_shared.gpu_addr + vcn->fw_shared.mem_size;
-	volatile struct amdgpu_vcn_fwlog *log_buf = fw_log_cpu_addr;
-	volatile struct amdgpu_fw_shared_fw_logging *fw_log = vcn->fw_shared.cpu_addr
-                                                         + vcn->fw_shared.log_offset;
+	struct amdgpu_vcn_fwlog *log_buf = fw_log_cpu_addr;
+	struct amdgpu_fw_shared_fw_logging *fw_log = vcn->fw_shared.cpu_addr
+							 + vcn->fw_shared.log_offset;
 	*flag |= cpu_to_le32(AMDGPU_VCN_FW_LOGGING_FLAG);
 	fw_log->is_enabled = 1;
 	fw_log->addr_lo = cpu_to_le32(fw_log_gpu_addr & 0xFFFFFFFF);
@@ -1172,7 +1274,7 @@ int amdgpu_vcn_process_poison_irq(struct amdgpu_device *adev,
 		amdgpu_ras_interrupt_dispatch(adev, &ih_data);
 	} else {
 		if (adev->virt.ops && adev->virt.ops->ras_poison_handler)
-			adev->virt.ops->ras_poison_handler(adev);
+			adev->virt.ops->ras_poison_handler(adev, ras_if->block);
 		else
 			dev_warn(adev->dev,
 				"No ras_poison_handler interface in SRIOV for VCN!\n");
@@ -1181,6 +1283,32 @@ int amdgpu_vcn_process_poison_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
+int amdgpu_vcn_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
+{
+	int r, i;
+
+	r = amdgpu_ras_block_late_init(adev, ras_block);
+	if (r)
+		return r;
+
+	if (amdgpu_ras_is_supported(adev, ras_block->block)) {
+		for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+			if (adev->vcn.harvest_config & (1 << i) ||
+			    !adev->vcn.inst[i].ras_poison_irq.funcs)
+				continue;
+
+			r = amdgpu_irq_get(adev, &adev->vcn.inst[i].ras_poison_irq, 0);
+			if (r)
+				goto late_fini;
+		}
+	}
+	return 0;
+
+late_fini:
+	amdgpu_ras_block_late_fini(adev, ras_block);
+	return r;
+}
+
 int amdgpu_vcn_ras_sw_init(struct amdgpu_device *adev)
 {
 	int err;
@@ -1202,7 +1330,308 @@ int amdgpu_vcn_ras_sw_init(struct amdgpu_device *adev)
 	adev->vcn.ras_if = &ras->ras_block.ras_comm;
 
 	if (!ras->ras_block.ras_late_init)
-		ras->ras_block.ras_late_init = amdgpu_ras_block_late_init;
+		ras->ras_block.ras_late_init = amdgpu_vcn_ras_late_init;
+
+	return 0;
+}
+
+int amdgpu_vcn_psp_update_sram(struct amdgpu_device *adev, int inst_idx,
+			       enum AMDGPU_UCODE_ID ucode_id)
+{
+	struct amdgpu_firmware_info ucode = {
+		.ucode_id = (ucode_id ? ucode_id :
+			    (inst_idx ? AMDGPU_UCODE_ID_VCN1_RAM :
+					AMDGPU_UCODE_ID_VCN0_RAM)),
+		.mc_addr = adev->vcn.inst[inst_idx].dpg_sram_gpu_addr,
+		.ucode_size = ((uintptr_t)adev->vcn.inst[inst_idx].dpg_sram_curr_addr -
+			      (uintptr_t)adev->vcn.inst[inst_idx].dpg_sram_cpu_addr),
+	};
+
+	return psp_execute_ip_fw_load(&adev->psp, &ucode);
+}
+
+static ssize_t amdgpu_get_vcn_reset_mask(struct device *dev,
+						struct device_attribute *attr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	if (!adev)
+		return -ENODEV;
+
+	return amdgpu_show_reset_mask(buf, adev->vcn.supported_reset);
+}
+
+static DEVICE_ATTR(vcn_reset_mask, 0444,
+		   amdgpu_get_vcn_reset_mask, NULL);
+
+int amdgpu_vcn_sysfs_reset_mask_init(struct amdgpu_device *adev)
+{
+	int r = 0;
+
+	if (adev->vcn.num_vcn_inst) {
+		r = device_create_file(adev->dev, &dev_attr_vcn_reset_mask);
+		if (r)
+			return r;
+	}
+
+	return r;
+}
+
+void amdgpu_vcn_sysfs_reset_mask_fini(struct amdgpu_device *adev)
+{
+	if (adev->dev->kobj.sd) {
+		if (adev->vcn.num_vcn_inst)
+			device_remove_file(adev->dev, &dev_attr_vcn_reset_mask);
+	}
+}
+
+/*
+ * debugfs to enable/disable vcn job submission to specific core or
+ * instance. It is created only if the queue type is unified.
+ */
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_debugfs_vcn_sched_mask_set(void *data, u64 val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i;
+	u64 mask;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+
+	mask = (1ULL << adev->vcn.num_vcn_inst) - 1;
+	if ((val & mask) == 0)
+		return -EINVAL;
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		ring = &adev->vcn.inst[i].ring_enc[0];
+		if (val & (1ULL << i))
+			ring->sched.ready = true;
+		else
+			ring->sched.ready = false;
+	}
+	/* publish sched.ready flag update effective immediately across smp */
+	smp_rmb();
+	return 0;
+}
+
+static int amdgpu_debugfs_vcn_sched_mask_get(void *data, u64 *val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i;
+	u64 mask = 0;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		ring = &adev->vcn.inst[i].ring_enc[0];
+		if (ring->sched.ready)
+			mask |= 1ULL << i;
+		}
+	*val = mask;
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(amdgpu_debugfs_vcn_sched_mask_fops,
+			 amdgpu_debugfs_vcn_sched_mask_get,
+			 amdgpu_debugfs_vcn_sched_mask_set, "%llx\n");
+#endif
+
+void amdgpu_debugfs_vcn_sched_mask_init(struct amdgpu_device *adev)
+{
+#if defined(CONFIG_DEBUG_FS)
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+	struct dentry *root = minor->debugfs_root;
+	char name[32];
+
+	if (adev->vcn.num_vcn_inst <= 1 || !adev->vcn.inst[0].using_unified_queue)
+		return;
+	sprintf(name, "amdgpu_vcn_sched_mask");
+	debugfs_create_file(name, 0600, root, adev,
+			    &amdgpu_debugfs_vcn_sched_mask_fops);
+#endif
+}
+
+/**
+ * vcn_set_powergating_state - set VCN block powergating state
+ *
+ * @ip_block: amdgpu_ip_block pointer
+ * @state: power gating state
+ *
+ * Set VCN block powergating state
+ */
+int vcn_set_powergating_state(struct amdgpu_ip_block *ip_block,
+			      enum amd_powergating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int ret = 0, i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[i];
+
+		ret |= vinst->set_pg_state(vinst, state);
+	}
+
+	return ret;
+}
+
+/**
+ * amdgpu_vcn_reset_engine - Reset a specific VCN engine
+ * @adev: Pointer to the AMDGPU device
+ * @instance_id: VCN engine instance to reset
+ *
+ * Returns: 0 on success, or a negative error code on failure.
+ */
+static int amdgpu_vcn_reset_engine(struct amdgpu_device *adev,
+				   uint32_t instance_id)
+{
+	struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[instance_id];
+	int r, i;
+
+	mutex_lock(&vinst->engine_reset_mutex);
+	/* Stop the scheduler's work queue for the dec and enc rings if they are running.
+	 * This ensures that no new tasks are submitted to the queues while
+	 * the reset is in progress.
+	 */
+	drm_sched_wqueue_stop(&vinst->ring_dec.sched);
+	for (i = 0; i < vinst->num_enc_rings; i++)
+		drm_sched_wqueue_stop(&vinst->ring_enc[i].sched);
+
+	/* Perform the VCN reset for the specified instance */
+	r = vinst->reset(vinst);
+	if (r)
+		goto unlock;
+	r = amdgpu_ring_test_ring(&vinst->ring_dec);
+	if (r)
+		goto unlock;
+	for (i = 0; i < vinst->num_enc_rings; i++) {
+		r = amdgpu_ring_test_ring(&vinst->ring_enc[i]);
+		if (r)
+			goto unlock;
+	}
+	amdgpu_fence_driver_force_completion(&vinst->ring_dec);
+	for (i = 0; i < vinst->num_enc_rings; i++)
+		amdgpu_fence_driver_force_completion(&vinst->ring_enc[i]);
+
+	/* Restart the scheduler's work queue for the dec and enc rings
+	 * if they were stopped by this function. This allows new tasks
+	 * to be submitted to the queues after the reset is complete.
+	 */
+	drm_sched_wqueue_start(&vinst->ring_dec.sched);
+	for (i = 0; i < vinst->num_enc_rings; i++)
+		drm_sched_wqueue_start(&vinst->ring_enc[i].sched);
+
+unlock:
+	mutex_unlock(&vinst->engine_reset_mutex);
+
+	return r;
+}
+
+/**
+ * amdgpu_vcn_ring_reset - Reset a VCN ring
+ * @ring: ring to reset
+ * @vmid: vmid of guilty job
+ * @timedout_fence: fence of timed out job
+ *
+ * This helper is for VCN blocks without unified queues because
+ * resetting the engine resets all queues in that case.  With
+ * unified queues we have one queue per engine.
+ * Returns: 0 on success, or a negative error code on failure.
+ */
+int amdgpu_vcn_ring_reset(struct amdgpu_ring *ring,
+			  unsigned int vmid,
+			  struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (adev->vcn.inst[ring->me].using_unified_queue)
+		return -EINVAL;
+
+	return amdgpu_vcn_reset_engine(adev, ring->me);
+}
+
+int amdgpu_vcn_reg_dump_init(struct amdgpu_device *adev,
+			     const struct amdgpu_hwip_reg_entry *reg, u32 count)
+{
+	adev->vcn.ip_dump = kcalloc(adev->vcn.num_vcn_inst * count,
+				     sizeof(uint32_t), GFP_KERNEL);
+	if (!adev->vcn.ip_dump)
+		return -ENOMEM;
+	adev->vcn.reg_list = reg;
+	adev->vcn.reg_count = count;
 
 	return 0;
 }
+
+static void amdgpu_vcn_reg_dump_fini(struct amdgpu_device *adev)
+{
+	kfree(adev->vcn.ip_dump);
+	adev->vcn.ip_dump = NULL;
+	adev->vcn.reg_list = NULL;
+	adev->vcn.reg_count = 0;
+}
+
+void amdgpu_vcn_dump_ip_state(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, j;
+	bool is_powered;
+	u32 inst_off;
+
+	if (!adev->vcn.ip_dump)
+		return;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		inst_off = i * adev->vcn.reg_count;
+		/* mmUVD_POWER_STATUS is always readable and is the first in reg_list */
+		adev->vcn.ip_dump[inst_off] =
+			RREG32(SOC15_REG_ENTRY_OFFSET_INST(adev->vcn.reg_list[0], i));
+		is_powered = (adev->vcn.ip_dump[inst_off] &
+			      UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF) !=
+			      UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF;
+
+		if (is_powered)
+			for (j = 1; j < adev->vcn.reg_count; j++)
+				adev->vcn.ip_dump[inst_off + j] =
+				RREG32(SOC15_REG_ENTRY_OFFSET_INST(adev->vcn.reg_list[j], i));
+	}
+}
+
+void amdgpu_vcn_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, j;
+	bool is_powered;
+	u32 inst_off;
+
+	if (!adev->vcn.ip_dump)
+		return;
+
+	drm_printf(p, "num_instances:%d\n", adev->vcn.num_vcn_inst);
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		if (adev->vcn.harvest_config & (1 << i)) {
+			drm_printf(p, "\nHarvested Instance:VCN%d Skipping dump\n", i);
+			continue;
+		}
+
+		inst_off = i * adev->vcn.reg_count;
+		is_powered = (adev->vcn.ip_dump[inst_off] &
+			      UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF) !=
+			      UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF;
+
+		if (is_powered) {
+			drm_printf(p, "\nActive Instance:VCN%d\n", i);
+			for (j = 0; j < adev->vcn.reg_count; j++)
+				drm_printf(p, "%-50s \t 0x%08x\n", adev->vcn.reg_list[j].reg_name,
+					   adev->vcn.ip_dump[inst_off + j]);
+		} else {
+			drm_printf(p, "\nInactive Instance:VCN%d\n", i);
+		}
+	}
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
index c730949ece7d..82624b44e661 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
@@ -1,5 +1,5 @@
 /*
- * Copyright 2016 Advanced Micro Devices, Inc.
+ * Copyright 2016-2024 Advanced Micro Devices, Inc. All rights reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -32,8 +32,8 @@
 #define AMDGPU_VCN_FIRMWARE_OFFSET	256
 #define AMDGPU_VCN_MAX_ENC_RINGS	3
 
-#define AMDGPU_MAX_VCN_INSTANCES	2
-#define AMDGPU_MAX_VCN_ENC_RINGS  AMDGPU_VCN_MAX_ENC_RINGS * AMDGPU_MAX_VCN_INSTANCES
+#define AMDGPU_MAX_VCN_INSTANCES	4
+#define AMDGPU_MAX_VCN_ENC_RINGS  (AMDGPU_VCN_MAX_ENC_RINGS * AMDGPU_MAX_VCN_INSTANCES)
 
 #define AMDGPU_VCN_HARVEST_VCN0 (1 << 0)
 #define AMDGPU_VCN_HARVEST_VCN1 (1 << 1)
@@ -66,7 +66,6 @@
 #define VCN_ENC_CMD_REG_WAIT		0x0000000c
 
 #define VCN_AON_SOC_ADDRESS_2_0 	0x1f800
-#define VCN1_AON_SOC_ADDRESS_3_0 	0x48000
 #define VCN_VID_IP_ADDRESS_2_0		0x0
 #define VCN_AON_IP_ADDRESS_2_0		0x30000
 
@@ -101,7 +100,8 @@
 
 #define SOC15_DPG_MODE_OFFSET(ip, inst_idx, reg) 						\
 	({											\
-		uint32_t internal_reg_offset, addr;						\
+		/* To avoid a -Wunused-but-set-variable warning. */				\
+		uint32_t internal_reg_offset __maybe_unused, addr;				\
 		bool video_range, video1_range, aon_range, aon1_range;				\
 												\
 		addr = (adev->reg_offset[ip##_HWIP][inst_idx][reg##_BASE_IDX] + reg);		\
@@ -141,21 +141,80 @@
 		RREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_DATA);				\
 	})
 
-#define WREG32_SOC15_DPG_MODE(inst_idx, offset, value, mask_en, indirect)			\
-	do {											\
-		if (!indirect) {								\
-			WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_DATA, value);			\
-			WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_CTL, 				\
-				(0x1 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT |			\
-				 mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT |			\
-				 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT));		\
-		} else {									\
-			*adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ = offset;		\
-			*adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ = value;			\
-		}										\
+#define WREG32_SOC15_DPG_MODE(inst_idx, offset, value, mask_en, indirect)             \
+	do {                                                                          \
+		if (!indirect) {                                                      \
+			WREG32_SOC15(VCN, GET_INST(VCN, inst_idx),                    \
+				     mmUVD_DPG_LMA_DATA, value);                      \
+			WREG32_SOC15(                                                 \
+				VCN, GET_INST(VCN, inst_idx),                         \
+				mmUVD_DPG_LMA_CTL,                                    \
+				(0x1 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT |          \
+				 mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT |         \
+				 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT)); \
+		} else {                                                              \
+			*adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ =              \
+				offset;                                               \
+			*adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ =              \
+				value;                                                \
+		}                                                                     \
+	} while (0)
+
+#define SOC24_DPG_MODE_OFFSET(ip, inst_idx, reg)						\
+	({											\
+		/* To avoid a -Wunused-but-set-variable warning. */				\
+		uint32_t internal_reg_offset __maybe_unused, addr;				\
+		bool video_range, video1_range, aon_range, aon1_range;				\
+												\
+		addr = (adev->reg_offset[ip##_HWIP][inst_idx][reg##_BASE_IDX] + reg);		\
+		addr <<= 2;									\
+		video_range = ((((0xFFFFF & addr) >= (VCN_VID_SOC_ADDRESS)) &&			\
+				((0xFFFFF & addr) < ((VCN_VID_SOC_ADDRESS + 0x2600)))));	\
+		video1_range = ((((0xFFFFF & addr) >= (VCN1_VID_SOC_ADDRESS)) &&		\
+				((0xFFFFF & addr) < ((VCN1_VID_SOC_ADDRESS + 0x2600)))));	\
+		aon_range   = ((((0xFFFFF & addr) >= (VCN_AON_SOC_ADDRESS)) &&			\
+				((0xFFFFF & addr) < ((VCN_AON_SOC_ADDRESS + 0x600)))));		\
+		aon1_range   = ((((0xFFFFF & addr) >= (VCN1_AON_SOC_ADDRESS)) &&		\
+				((0xFFFFF & addr) < ((VCN1_AON_SOC_ADDRESS + 0x600)))));	\
+		if (video_range)								\
+			internal_reg_offset = ((0xFFFFF & addr) - (VCN_VID_SOC_ADDRESS) +	\
+				(VCN_VID_IP_ADDRESS));						\
+		else if (aon_range)								\
+			internal_reg_offset = ((0xFFFFF & addr) - (VCN_AON_SOC_ADDRESS) +	\
+				(VCN_AON_IP_ADDRESS));						\
+		else if (video1_range)								\
+			internal_reg_offset = ((0xFFFFF & addr) - (VCN1_VID_SOC_ADDRESS) +	\
+				(VCN_VID_IP_ADDRESS));						\
+		else if (aon1_range)								\
+			internal_reg_offset = ((0xFFFFF & addr) - (VCN1_AON_SOC_ADDRESS) +	\
+				(VCN_AON_IP_ADDRESS));						\
+		else										\
+			internal_reg_offset = (0xFFFFF & addr);					\
+												\
+		internal_reg_offset >>= 2;							\
+	})
+
+#define WREG32_SOC24_DPG_MODE(inst_idx, offset, value, mask_en, indirect)		\
+	do {										\
+		if (!indirect) {							\
+			WREG32_SOC15(VCN, GET_INST(VCN, inst_idx),			\
+				     regUVD_DPG_LMA_DATA, value);			\
+			WREG32_SOC15(							\
+				VCN, GET_INST(VCN, inst_idx),				\
+				regUVD_DPG_LMA_CTL,					\
+				(0x1 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT |		\
+				 mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT |		\
+				 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT));	\
+		} else {								\
+			*adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ =		\
+				offset;							\
+			*adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ =		\
+				value;							\
+		}									\
 	} while (0)
 
 #define AMDGPU_FW_SHARED_FLAG_0_UNIFIED_QUEUE (1 << 2)
+#define AMDGPU_FW_SHARED_FLAG_0_DRM_KEY_INJECT (1 << 4)
 #define AMDGPU_VCN_FW_SHARED_FLAG_0_RB	(1 << 6)
 #define AMDGPU_VCN_MULTI_QUEUE_FLAG	(1 << 8)
 #define AMDGPU_VCN_SW_RING_FLAG		(1 << 9)
@@ -163,6 +222,9 @@
 #define AMDGPU_VCN_SMU_VERSION_INFO_FLAG (1 << 11)
 #define AMDGPU_VCN_SMU_DPM_INTERFACE_FLAG (1 << 11)
 #define AMDGPU_VCN_VF_RB_SETUP_FLAG (1 << 14)
+#define AMDGPU_VCN_VF_RB_DECOUPLE_FLAG (1 << 15)
+
+#define MAX_NUM_VCN_RB_SETUP 4
 
 #define AMDGPU_VCN_IB_FLAG_DECODE_BUFFER	0x00000001
 #define AMDGPU_VCN_CMD_FLAG_MSG_BUFFER		0x00000001
@@ -175,6 +237,16 @@
 #define AMDGPU_VCN_SMU_DPM_INTERFACE_DGPU (0)
 #define AMDGPU_VCN_SMU_DPM_INTERFACE_APU (1)
 
+#define AMDGPU_DRM_KEY_INJECT_WORKAROUND_VCNFW_ASD_HANDSHAKING 2
+
+struct amdgpu_hwip_reg_entry;
+
+enum amdgpu_vcn_caps {
+	AMDGPU_VCN_RRMT_ENABLED,
+};
+
+#define AMDGPU_VCN_CAPS(caps) BIT(AMDGPU_VCN_##caps)
+
 enum fw_queue_mode {
 	FW_QUEUE_RING_RESET = 1,
 	FW_QUEUE_DPG_HOLD_OFF = 2,
@@ -226,6 +298,8 @@ struct amdgpu_vcn_fw_shared {
 };
 
 struct amdgpu_vcn_inst {
+	struct amdgpu_device	*adev;
+	int			inst;
 	struct amdgpu_bo	*vcpu_bo;
 	void			*cpu_addr;
 	uint64_t		gpu_addr;
@@ -234,6 +308,7 @@ struct amdgpu_vcn_inst {
 	struct amdgpu_ring	ring_enc[AMDGPU_VCN_MAX_ENC_RINGS];
 	atomic_t		sched_score;
 	struct amdgpu_irq_src	irq;
+	struct amdgpu_irq_src	ras_poison_irq;
 	struct amdgpu_vcn_reg	external;
 	struct amdgpu_bo	*dpg_sram_bo;
 	struct dpg_pause_state	pause_state;
@@ -242,6 +317,26 @@ struct amdgpu_vcn_inst {
 	uint32_t		*dpg_sram_curr_addr;
 	atomic_t		dpg_enc_submission_cnt;
 	struct amdgpu_vcn_fw_shared fw_shared;
+	uint8_t			aid_id;
+	const struct firmware	*fw; /* VCN firmware */
+	uint8_t			vcn_config;
+	uint32_t		vcn_codec_disable_mask;
+	atomic_t		total_submission_cnt;
+	struct mutex		vcn_pg_lock;
+	enum amd_powergating_state cur_state;
+	struct delayed_work	idle_work;
+	unsigned		fw_version;
+	unsigned		num_enc_rings;
+	bool			indirect_sram;
+	struct amdgpu_vcn_reg	 internal;
+	struct mutex		vcn1_jpeg1_workaround;
+	int (*pause_dpg_mode)(struct amdgpu_vcn_inst *vinst,
+			      struct dpg_pause_state *new_state);
+	int (*set_pg_state)(struct amdgpu_vcn_inst *vinst,
+			    enum amd_powergating_state state);
+	int (*reset)(struct amdgpu_vcn_inst *vinst);
+	bool using_unified_queue;
+	struct mutex		engine_reset_mutex;
 };
 
 struct amdgpu_vcn_ras {
@@ -249,28 +344,30 @@ struct amdgpu_vcn_ras {
 };
 
 struct amdgpu_vcn {
-	unsigned		fw_version;
-	struct delayed_work	idle_work;
-	const struct firmware	*fw;	/* VCN firmware */
-	unsigned		num_enc_rings;
-	enum amd_powergating_state cur_state;
-	bool			indirect_sram;
-
 	uint8_t	num_vcn_inst;
 	struct amdgpu_vcn_inst	 inst[AMDGPU_MAX_VCN_INSTANCES];
-	uint8_t			 vcn_config[AMDGPU_MAX_VCN_INSTANCES];
-	uint32_t		 vcn_codec_disable_mask[AMDGPU_MAX_VCN_INSTANCES];
-	struct amdgpu_vcn_reg	 internal;
-	struct mutex		 vcn_pg_lock;
-	struct mutex		vcn1_jpeg1_workaround;
-	atomic_t		 total_submission_cnt;
 
 	unsigned	harvest_config;
-	int (*pause_dpg_mode)(struct amdgpu_device *adev,
-		int inst_idx, struct dpg_pause_state *new_state);
 
 	struct ras_common_if    *ras_if;
 	struct amdgpu_vcn_ras   *ras;
+
+	uint16_t inst_mask;
+	uint8_t	num_inst_per_aid;
+
+	/* IP reg dump */
+	uint32_t		*ip_dump;
+
+	uint32_t		supported_reset;
+	uint32_t		caps;
+
+	bool			per_inst_fw;
+	unsigned		fw_version;
+
+	bool			workload_profile_active;
+	struct mutex            workload_profile_mutex;
+	u32 reg_count;
+	const struct amdgpu_hwip_reg_entry *reg_list;
 };
 
 struct amdgpu_fw_shared_rb_ptrs_struct {
@@ -322,15 +419,40 @@ struct amdgpu_fw_shared {
 	struct amdgpu_fw_shared_smu_interface_info smu_interface_info;
 };
 
+struct amdgpu_vcn_rb_setup_info {
+	uint32_t  rb_addr_lo;
+	uint32_t  rb_addr_hi;
+	uint32_t  rb_size;
+};
+
 struct amdgpu_fw_shared_rb_setup {
 	uint32_t is_rb_enabled_flags;
-	uint32_t rb_addr_lo;
-	uint32_t rb_addr_hi;
-	uint32_t  rb_size;
-	uint32_t  rb4_addr_lo;
-	uint32_t  rb4_addr_hi;
-	uint32_t  rb4_size;
-	uint32_t  reserved[6];
+
+	union {
+		struct {
+			uint32_t rb_addr_lo;
+			uint32_t rb_addr_hi;
+			uint32_t  rb_size;
+			uint32_t  rb4_addr_lo;
+			uint32_t  rb4_addr_hi;
+			uint32_t  rb4_size;
+			uint32_t  reserved[6];
+		};
+
+		struct {
+			struct amdgpu_vcn_rb_setup_info rb_info[MAX_NUM_VCN_RB_SETUP];
+		};
+	};
+};
+
+struct amdgpu_fw_shared_drm_key_wa {
+	uint8_t  method;
+	uint8_t  reserved[3];
+};
+
+struct amdgpu_fw_shared_queue_decouple {
+	uint8_t  is_enabled;
+	uint8_t  reserved[7];
 };
 
 struct amdgpu_vcn4_fw_shared {
@@ -342,6 +464,9 @@ struct amdgpu_vcn4_fw_shared {
 	uint8_t pad2[20];
 	struct amdgpu_fw_shared_rb_setup rb_setup;
 	struct amdgpu_fw_shared_smu_interface_info smu_dpm_interface;
+	struct amdgpu_fw_shared_drm_key_wa drm_key_wa;
+	uint8_t pad3[9];
+	struct amdgpu_fw_shared_queue_decouple decouple;
 };
 
 struct amdgpu_vcn_fwlog {
@@ -359,6 +484,28 @@ struct amdgpu_vcn_decode_buffer {
 	uint32_t pad[30];
 };
 
+struct amdgpu_vcn_rb_metadata {
+	uint32_t size;
+	uint32_t present_flag_0;
+
+	uint8_t version;
+	uint8_t ring_id;
+	uint8_t pad[26];
+};
+
+struct amdgpu_vcn5_fw_shared {
+	uint32_t present_flag_0;
+	uint8_t pad[12];
+	struct amdgpu_fw_shared_unified_queue_struct sq;
+	uint8_t pad1[8];
+	struct amdgpu_fw_shared_fw_logging fw_log;
+	uint8_t pad2[20];
+	struct amdgpu_fw_shared_rb_setup rb_setup;
+	struct amdgpu_fw_shared_smu_interface_info smu_dpm_interface;
+	struct amdgpu_fw_shared_drm_key_wa drm_key_wa;
+	uint8_t pad3[404];
+};
+
 #define VCN_BLOCK_ENCODE_DISABLE_MASK 0x80
 #define VCN_BLOCK_DECODE_DISABLE_MASK 0x40
 #define VCN_BLOCK_QUEUE_DISABLE_MASK 0xC0
@@ -369,11 +516,11 @@ enum vcn_ring_type {
 	VCN_UNIFIED_RING,
 };
 
-int amdgpu_vcn_early_init(struct amdgpu_device *adev);
-int amdgpu_vcn_sw_init(struct amdgpu_device *adev);
-int amdgpu_vcn_sw_fini(struct amdgpu_device *adev);
-int amdgpu_vcn_suspend(struct amdgpu_device *adev);
-int amdgpu_vcn_resume(struct amdgpu_device *adev);
+int amdgpu_vcn_early_init(struct amdgpu_device *adev, int i);
+int amdgpu_vcn_sw_init(struct amdgpu_device *adev, int i);
+void amdgpu_vcn_sw_fini(struct amdgpu_device *adev, int i);
+int amdgpu_vcn_suspend(struct amdgpu_device *adev, int i);
+int amdgpu_vcn_resume(struct amdgpu_device *adev, int i);
 void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring);
 void amdgpu_vcn_ring_end_use(struct amdgpu_ring *ring);
 
@@ -391,7 +538,7 @@ int amdgpu_vcn_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout);
 
 enum amdgpu_ring_priority_level amdgpu_vcn_get_enc_ring_prio(int ring);
 
-void amdgpu_vcn_setup_ucode(struct amdgpu_device *adev);
+void amdgpu_vcn_setup_ucode(struct amdgpu_device *adev, int i);
 
 void amdgpu_vcn_fwlog_init(struct amdgpu_vcn_inst *vcn);
 void amdgpu_debugfs_vcn_fwlog_init(struct amdgpu_device *adev,
@@ -400,6 +547,27 @@ void amdgpu_debugfs_vcn_fwlog_init(struct amdgpu_device *adev,
 int amdgpu_vcn_process_poison_irq(struct amdgpu_device *adev,
 			struct amdgpu_irq_src *source,
 			struct amdgpu_iv_entry *entry);
+int amdgpu_vcn_ras_late_init(struct amdgpu_device *adev,
+			struct ras_common_if *ras_block);
 int amdgpu_vcn_ras_sw_init(struct amdgpu_device *adev);
 
+int amdgpu_vcn_psp_update_sram(struct amdgpu_device *adev, int inst_idx,
+			       enum AMDGPU_UCODE_ID ucode_id);
+int amdgpu_vcn_save_vcpu_bo(struct amdgpu_device *adev);
+int amdgpu_vcn_sysfs_reset_mask_init(struct amdgpu_device *adev);
+void amdgpu_vcn_sysfs_reset_mask_fini(struct amdgpu_device *adev);
+void amdgpu_debugfs_vcn_sched_mask_init(struct amdgpu_device *adev);
+
+int vcn_set_powergating_state(struct amdgpu_ip_block *ip_block,
+			      enum amd_powergating_state state);
+int amdgpu_vcn_ring_reset(struct amdgpu_ring *ring,
+			  unsigned int vmid,
+			  struct amdgpu_fence *guilty_fence);
+int amdgpu_vcn_reg_dump_init(struct amdgpu_device *adev,
+			     const struct amdgpu_hwip_reg_entry *reg, u32 count);
+void amdgpu_vcn_dump_ip_state(struct amdgpu_ip_block *ip_block);
+void amdgpu_vcn_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p);
+void amdgpu_vcn_get_profile(struct amdgpu_device *adev);
+void amdgpu_vcn_put_profile(struct amdgpu_device *adev);
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
index f2e2cbaa7fde..47a6ce4fdc74 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
@@ -32,6 +32,8 @@
 
 #include "amdgpu.h"
 #include "amdgpu_ras.h"
+#include "amdgpu_reset.h"
+#include "amdgpu_dpm.h"
 #include "vi.h"
 #include "soc15.h"
 #include "nv.h"
@@ -42,6 +44,18 @@
 		vf2pf_info->ucode_info[ucode].version = ver; \
 	} while (0)
 
+#define mmRCC_CONFIG_MEMSIZE    0xde3
+
+const char *amdgpu_virt_dynamic_crit_table_name[] = {
+	"IP DISCOVERY",
+	"VBIOS IMG",
+	"RAS TELEMETRY",
+	"DATA EXCHANGE",
+	"BAD PAGE INFO",
+	"INIT HEADER",
+	"LAST",
+};
+
 bool amdgpu_virt_mmio_blocked(struct amdgpu_device *adev)
 {
 	/* By now all MMIO pages except mailbox are blocked */
@@ -56,7 +70,8 @@ void amdgpu_virt_init_setting(struct amdgpu_device *adev)
 
 	/* enable virtual display */
 	if (adev->asic_type != CHIP_ALDEBARAN &&
-	    adev->asic_type != CHIP_ARCTURUS) {
+	    adev->asic_type != CHIP_ARCTURUS &&
+	    ((adev->pdev->class >> 8) != PCI_CLASS_ACCELERATOR_PROCESSING)) {
 		if (adev->mode_info.num_crtc == 0)
 			adev->mode_info.num_crtc = 1;
 		adev->enable_virtual_display = true;
@@ -65,61 +80,9 @@ void amdgpu_virt_init_setting(struct amdgpu_device *adev)
 	adev->cg_flags = 0;
 	adev->pg_flags = 0;
 
-	/* enable mcbp for sriov asic_type before soc21 */
-	amdgpu_mcbp = (adev->asic_type < CHIP_IP_DISCOVERY) ? 1 : 0;
-
-}
-
-void amdgpu_virt_kiq_reg_write_reg_wait(struct amdgpu_device *adev,
-					uint32_t reg0, uint32_t reg1,
-					uint32_t ref, uint32_t mask)
-{
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-	struct amdgpu_ring *ring = &kiq->ring;
-	signed long r, cnt = 0;
-	unsigned long flags;
-	uint32_t seq;
-
-	if (adev->mes.ring.sched.ready) {
-		amdgpu_mes_reg_write_reg_wait(adev, reg0, reg1,
-					      ref, mask);
-		return;
-	}
-
-	spin_lock_irqsave(&kiq->ring_lock, flags);
-	amdgpu_ring_alloc(ring, 32);
-	amdgpu_ring_emit_reg_write_reg_wait(ring, reg0, reg1,
-					    ref, mask);
-	r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
-	if (r)
-		goto failed_undo;
-
-	amdgpu_ring_commit(ring);
-	spin_unlock_irqrestore(&kiq->ring_lock, flags);
-
-	r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
-
-	/* don't wait anymore for IRQ context */
-	if (r < 1 && in_interrupt())
-		goto failed_kiq;
-
-	might_sleep();
-	while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) {
-
-		msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
-		r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
-	}
-
-	if (cnt > MAX_KIQ_REG_TRY)
-		goto failed_kiq;
-
-	return;
-
-failed_undo:
-	amdgpu_ring_undo(ring);
-	spin_unlock_irqrestore(&kiq->ring_lock, flags);
-failed_kiq:
-	dev_err(adev->dev, "failed to write reg %x wait reg %x\n", reg0, reg1);
+	/* Reduce kcq number to 2 to reduce latency */
+	if (amdgpu_num_kcq == -1)
+		amdgpu_num_kcq = 2;
 }
 
 /**
@@ -136,8 +99,10 @@ int amdgpu_virt_request_full_gpu(struct amdgpu_device *adev, bool init)
 
 	if (virt->ops && virt->ops->req_full_gpu) {
 		r = virt->ops->req_full_gpu(adev, init);
-		if (r)
+		if (r) {
+			adev->no_hw_access = true;
 			return r;
+		}
 
 		adev->virt.caps &= ~AMDGPU_SRIOV_CAPS_RUNTIME;
 	}
@@ -197,9 +162,24 @@ void amdgpu_virt_request_init_data(struct amdgpu_device *adev)
 		virt->ops->req_init_data(adev);
 
 	if (adev->virt.req_init_data_ver > 0)
-		DRM_INFO("host supports REQ_INIT_DATA handshake\n");
+		dev_info(adev->dev, "host supports REQ_INIT_DATA handshake of critical_region_version %d\n",
+				 adev->virt.req_init_data_ver);
 	else
-		DRM_WARN("host doesn't support REQ_INIT_DATA handshake\n");
+		dev_warn(adev->dev, "host doesn't support REQ_INIT_DATA handshake\n");
+}
+
+/**
+ * amdgpu_virt_ready_to_reset() - send ready to reset to host
+ * @adev:	amdgpu device.
+ * Send ready to reset message to GPU hypervisor to signal we have stopped GPU
+ * activity and is ready for host FLR
+ */
+void amdgpu_virt_ready_to_reset(struct amdgpu_device *adev)
+{
+	struct amdgpu_virt *virt = &adev->virt;
+
+	if (virt->ops && virt->ops->reset_gpu)
+		virt->ops->ready_to_reset(adev);
 }
 
 /**
@@ -238,12 +218,12 @@ int amdgpu_virt_alloc_mm_table(struct amdgpu_device *adev)
 				    &adev->virt.mm_table.gpu_addr,
 				    (void *)&adev->virt.mm_table.cpu_addr);
 	if (r) {
-		DRM_ERROR("failed to alloc mm table and error = %d.\n", r);
+		dev_err(adev->dev, "failed to alloc mm table and error = %d.\n", r);
 		return r;
 	}
 
 	memset((void *)adev->virt.mm_table.cpu_addr, 0, PAGE_SIZE);
-	DRM_INFO("MM table gpu addr = 0x%llx, cpu addr = %p.\n",
+	dev_info(adev->dev, "MM table gpu addr = 0x%llx, cpu addr = %p.\n",
 		 adev->virt.mm_table.gpu_addr,
 		 adev->virt.mm_table.cpu_addr);
 	return 0;
@@ -265,6 +245,22 @@ void amdgpu_virt_free_mm_table(struct amdgpu_device *adev)
 	adev->virt.mm_table.gpu_addr = 0;
 }
 
+/**
+ * amdgpu_virt_rcvd_ras_interrupt() - receive ras interrupt
+ * @adev:	amdgpu device.
+ * Check whether host sent RAS error message
+ * Return: true if found, otherwise false
+ */
+bool amdgpu_virt_rcvd_ras_interrupt(struct amdgpu_device *adev)
+{
+	struct amdgpu_virt *virt = &adev->virt;
+
+	if (!virt->ops || !virt->ops->rcvd_ras_intr)
+		return false;
+
+	return virt->ops->rcvd_ras_intr(adev);
+}
+
 
 unsigned int amd_sriov_msg_checksum(void *obj,
 				unsigned long obj_size,
@@ -301,11 +297,11 @@ static int amdgpu_virt_init_ras_err_handler_data(struct amdgpu_device *adev)
 	if (!*data)
 		goto data_failure;
 
-	bps = kmalloc_array(align_space, sizeof((*data)->bps), GFP_KERNEL);
+	bps = kmalloc_array(align_space, sizeof(*(*data)->bps), GFP_KERNEL);
 	if (!bps)
 		goto bps_failure;
 
-	bps_bo = kmalloc_array(align_space, sizeof((*data)->bps_bo), GFP_KERNEL);
+	bps_bo = kmalloc_array(align_space, sizeof(*(*data)->bps_bo), GFP_KERNEL);
 	if (!bps_bo)
 		goto bps_bo_failure;
 
@@ -338,8 +334,10 @@ static void amdgpu_virt_ras_release_bp(struct amdgpu_device *adev)
 
 	for (i = data->last_reserved - 1; i >= 0; i--) {
 		bo = data->bps_bo[i];
-		amdgpu_bo_free_kernel(&bo, NULL, NULL);
-		data->bps_bo[i] = bo;
+		if (bo) {
+			amdgpu_bo_free_kernel(&bo, NULL, NULL);
+			data->bps_bo[i] = bo;
+		}
 		data->last_reserved = i;
 	}
 }
@@ -379,6 +377,8 @@ static void amdgpu_virt_ras_reserve_bps(struct amdgpu_device *adev)
 {
 	struct amdgpu_virt *virt = &adev->virt;
 	struct amdgpu_virt_ras_err_handler_data *data = virt->virt_eh_data;
+	struct amdgpu_vram_mgr *mgr = &adev->mman.vram_mgr;
+	struct ttm_resource_manager *man = &mgr->manager;
 	struct amdgpu_bo *bo = NULL;
 	uint64_t bp;
 	int i;
@@ -394,12 +394,20 @@ static void amdgpu_virt_ras_reserve_bps(struct amdgpu_device *adev)
 		 * 2) a ras bad page has been reserved (duplicate error injection
 		 *    for one page);
 		 */
-		if (amdgpu_bo_create_kernel_at(adev, bp << AMDGPU_GPU_PAGE_SHIFT,
-					       AMDGPU_GPU_PAGE_SIZE,
-					       &bo, NULL))
-			DRM_DEBUG("RAS WARN: reserve vram for retired page %llx fail\n", bp);
-
-		data->bps_bo[i] = bo;
+		if  (ttm_resource_manager_used(man)) {
+			amdgpu_vram_mgr_reserve_range(&adev->mman.vram_mgr,
+				bp << AMDGPU_GPU_PAGE_SHIFT,
+				AMDGPU_GPU_PAGE_SIZE);
+			data->bps_bo[i] = NULL;
+		} else {
+			if (amdgpu_bo_create_kernel_at(adev, bp << AMDGPU_GPU_PAGE_SHIFT,
+							AMDGPU_GPU_PAGE_SIZE,
+							&bo, NULL))
+				dev_dbg(adev->dev,
+						"RAS WARN: reserve vram for retired page %llx fail\n",
+						bp);
+			data->bps_bo[i] = bo;
+		}
 		data->last_reserved = i + 1;
 		bo = NULL;
 	}
@@ -435,6 +443,8 @@ static void amdgpu_virt_add_bad_page(struct amdgpu_device *adev,
 	else
 		vram_usage_va = adev->mman.drv_vram_usage_va;
 
+	memset(&bp, 0, sizeof(bp));
+
 	if (bp_block_size) {
 		bp_cnt = bp_block_size / sizeof(uint64_t);
 		for (bp_idx = 0; bp_idx < bp_cnt; bp_idx++) {
@@ -465,7 +475,7 @@ static int amdgpu_virt_read_pf2vf_data(struct amdgpu_device *adev)
 		return -EINVAL;
 
 	if (pf2vf_info->size > 1024) {
-		DRM_ERROR("invalid pf2vf message size\n");
+		dev_err(adev->dev, "invalid pf2vf message size: 0x%x\n", pf2vf_info->size);
 		return -EINVAL;
 	}
 
@@ -476,7 +486,9 @@ static int amdgpu_virt_read_pf2vf_data(struct amdgpu_device *adev)
 			adev->virt.fw_reserve.p_pf2vf, pf2vf_info->size,
 			adev->virt.fw_reserve.checksum_key, checksum);
 		if (checksum != checkval) {
-			DRM_ERROR("invalid pf2vf message\n");
+			dev_err(adev->dev,
+				"invalid pf2vf message: header checksum=0x%x calculated checksum=0x%x\n",
+				checksum, checkval);
 			return -EINVAL;
 		}
 
@@ -490,7 +502,9 @@ static int amdgpu_virt_read_pf2vf_data(struct amdgpu_device *adev)
 			adev->virt.fw_reserve.p_pf2vf, pf2vf_info->size,
 			0, checksum);
 		if (checksum != checkval) {
-			DRM_ERROR("invalid pf2vf message\n");
+			dev_err(adev->dev,
+				"invalid pf2vf message: header checksum=0x%x calculated checksum=0x%x\n",
+				checksum, checkval);
 			return -EINVAL;
 		}
 
@@ -519,14 +533,17 @@ static int amdgpu_virt_read_pf2vf_data(struct amdgpu_device *adev)
 			tmp = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->mm_bw_management[i].encode_max_frame_pixels;
 			adev->virt.encode_max_frame_pixels = max(tmp, adev->virt.encode_max_frame_pixels);
 		}
-		if((adev->virt.decode_max_dimension_pixels > 0) || (adev->virt.encode_max_dimension_pixels > 0))
+		if ((adev->virt.decode_max_dimension_pixels > 0) || (adev->virt.encode_max_dimension_pixels > 0))
 			adev->virt.is_mm_bw_enabled = true;
 
 		adev->unique_id =
 			((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->uuid;
+		adev->virt.ras_en_caps.all = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->ras_en_caps.all;
+		adev->virt.ras_telemetry_en_caps.all =
+			((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->ras_telemetry_en_caps.all;
 		break;
 	default:
-		DRM_ERROR("invalid pf2vf version\n");
+		dev_err(adev->dev, "invalid pf2vf version: 0x%x\n", pf2vf_info->version);
 		return -EINVAL;
 	}
 
@@ -557,7 +574,6 @@ static void amdgpu_virt_populate_vf2pf_ucode_info(struct amdgpu_device *adev)
 	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_RLC_SRLS, adev->gfx.rlc_srls_fw_version);
 	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_MEC,      adev->gfx.mec_fw_version);
 	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_MEC2,     adev->gfx.mec2_fw_version);
-	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_IMU,      adev->gfx.imu_fw_version);
 	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_SOS,      adev->psp.sos.fw_version);
 	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_ASD,
 			    adev->psp.asd_context.bin_desc.fw_version);
@@ -597,8 +613,8 @@ static int amdgpu_virt_write_vf2pf_data(struct amdgpu_device *adev)
 	vf2pf_info->driver_cert = 0;
 	vf2pf_info->os_info.all = 0;
 
-	vf2pf_info->fb_usage =
-		ttm_resource_manager_usage(&adev->mman.vram_mgr.manager) >> 20;
+	vf2pf_info->fb_usage = ttm_resource_manager_used(&adev->mman.vram_mgr.manager) ?
+		 ttm_resource_manager_usage(&adev->mman.vram_mgr.manager) >> 20 : 0;
 	vf2pf_info->fb_vis_usage =
 		amdgpu_vram_mgr_vis_usage(&adev->mman.vram_mgr) >> 20;
 	vf2pf_info->fb_size = adev->gmc.real_vram_size >> 20;
@@ -613,9 +629,15 @@ static int amdgpu_virt_write_vf2pf_data(struct amdgpu_device *adev)
 	vf2pf_info->decode_usage = 0;
 
 	vf2pf_info->dummy_page_addr = (uint64_t)adev->dummy_page_addr;
+	if (amdgpu_sriov_is_mes_info_enable(adev)) {
+		vf2pf_info->mes_info_addr =
+			(uint64_t)(adev->mes.resource_1_gpu_addr[0] + AMDGPU_GPU_PAGE_SIZE);
+		vf2pf_info->mes_info_size =
+			adev->mes.resource_1[0]->tbo.base.size - AMDGPU_GPU_PAGE_SIZE;
+	}
 	vf2pf_info->checksum =
 		amd_sriov_msg_checksum(
-		vf2pf_info, vf2pf_info->header.size, 0, 0);
+		vf2pf_info, sizeof(*vf2pf_info), 0, 0);
 
 	return 0;
 }
@@ -626,18 +648,59 @@ static void amdgpu_virt_update_vf2pf_work_item(struct work_struct *work)
 	int ret;
 
 	ret = amdgpu_virt_read_pf2vf_data(adev);
-	if (ret)
+	if (ret) {
+		adev->virt.vf2pf_update_retry_cnt++;
+
+		if ((amdgpu_virt_rcvd_ras_interrupt(adev) ||
+			adev->virt.vf2pf_update_retry_cnt >= AMDGPU_VF2PF_UPDATE_MAX_RETRY_LIMIT) &&
+			amdgpu_sriov_runtime(adev)) {
+
+			amdgpu_ras_set_fed(adev, true);
+			if (amdgpu_reset_domain_schedule(adev->reset_domain,
+							&adev->kfd.reset_work))
+				return;
+			else
+				dev_err(adev->dev, "Failed to queue work! at %s", __func__);
+		}
+
 		goto out;
+	}
+
+	adev->virt.vf2pf_update_retry_cnt = 0;
 	amdgpu_virt_write_vf2pf_data(adev);
 
 out:
 	schedule_delayed_work(&(adev->virt.vf2pf_work), adev->virt.vf2pf_update_interval_ms);
 }
 
+static int amdgpu_virt_read_exchange_data_from_mem(struct amdgpu_device *adev, uint32_t *pfvf_data)
+{
+	uint32_t dataexchange_offset =
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID].offset;
+	uint32_t dataexchange_size =
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID].size_kb << 10;
+	uint64_t pos = 0;
+
+	dev_info(adev->dev,
+			"Got data exchange info from dynamic crit_region_table at offset 0x%x with size of 0x%x bytes.\n",
+			dataexchange_offset, dataexchange_size);
+
+	if (!IS_ALIGNED(dataexchange_offset, 4) || !IS_ALIGNED(dataexchange_size, 4)) {
+		dev_err(adev->dev, "Data exchange data not aligned to 4 bytes\n");
+		return -EINVAL;
+	}
+
+	pos = (uint64_t)dataexchange_offset;
+	amdgpu_device_vram_access(adev, pos, pfvf_data,
+			dataexchange_size, false);
+
+	return 0;
+}
+
 void amdgpu_virt_fini_data_exchange(struct amdgpu_device *adev)
 {
 	if (adev->virt.vf2pf_update_interval_ms != 0) {
-		DRM_INFO("clean up the vf2pf work item\n");
+		dev_info(adev->dev, "clean up the vf2pf work item\n");
 		cancel_delayed_work_sync(&adev->virt.vf2pf_work);
 		adev->virt.vf2pf_update_interval_ms = 0;
 	}
@@ -645,12 +708,15 @@ void amdgpu_virt_fini_data_exchange(struct amdgpu_device *adev)
 
 void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev)
 {
+	uint32_t *pfvf_data = NULL;
+
 	adev->virt.fw_reserve.p_pf2vf = NULL;
 	adev->virt.fw_reserve.p_vf2pf = NULL;
 	adev->virt.vf2pf_update_interval_ms = 0;
+	adev->virt.vf2pf_update_retry_cnt = 0;
 
 	if (adev->mman.fw_vram_usage_va && adev->mman.drv_vram_usage_va) {
-		DRM_WARN("Currently fw_vram and drv_vram should not have values at the same time!");
+		dev_warn(adev->dev, "Currently fw_vram and drv_vram should not have values at the same time!");
 	} else if (adev->mman.fw_vram_usage_va || adev->mman.drv_vram_usage_va) {
 		/* go through this logic in ip_init and reset to init workqueue*/
 		amdgpu_virt_exchange_data(adev);
@@ -659,11 +725,34 @@ void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev)
 		schedule_delayed_work(&(adev->virt.vf2pf_work), msecs_to_jiffies(adev->virt.vf2pf_update_interval_ms));
 	} else if (adev->bios != NULL) {
 		/* got through this logic in early init stage to get necessary flags, e.g. rlcg_acc related*/
-		adev->virt.fw_reserve.p_pf2vf =
-			(struct amd_sriov_msg_pf2vf_info_header *)
-			(adev->bios + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB << 10));
+		if (adev->virt.req_init_data_ver == GPU_CRIT_REGION_V2) {
+			pfvf_data =
+				kzalloc(adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID].size_kb << 10,
+					GFP_KERNEL);
+			if (!pfvf_data) {
+				dev_err(adev->dev, "Failed to allocate memory for pfvf_data\n");
+				return;
+			}
 
-		amdgpu_virt_read_pf2vf_data(adev);
+			if (amdgpu_virt_read_exchange_data_from_mem(adev, pfvf_data))
+				goto free_pfvf_data;
+
+			adev->virt.fw_reserve.p_pf2vf =
+				(struct amd_sriov_msg_pf2vf_info_header *)pfvf_data;
+
+			amdgpu_virt_read_pf2vf_data(adev);
+
+free_pfvf_data:
+			kfree(pfvf_data);
+			pfvf_data = NULL;
+			adev->virt.fw_reserve.p_pf2vf = NULL;
+		} else {
+			adev->virt.fw_reserve.p_pf2vf =
+				(struct amd_sriov_msg_pf2vf_info_header *)
+				(adev->bios + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB_V1 << 10));
+
+			amdgpu_virt_read_pf2vf_data(adev);
+		}
 	}
 }
 
@@ -676,19 +765,38 @@ void amdgpu_virt_exchange_data(struct amdgpu_device *adev)
 
 	if (adev->mman.fw_vram_usage_va || adev->mman.drv_vram_usage_va) {
 		if (adev->mman.fw_vram_usage_va) {
-			adev->virt.fw_reserve.p_pf2vf =
-				(struct amd_sriov_msg_pf2vf_info_header *)
-				(adev->mman.fw_vram_usage_va + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB << 10));
-			adev->virt.fw_reserve.p_vf2pf =
-				(struct amd_sriov_msg_vf2pf_info_header *)
-				(adev->mman.fw_vram_usage_va + (AMD_SRIOV_MSG_VF2PF_OFFSET_KB << 10));
+			if (adev->virt.req_init_data_ver == GPU_CRIT_REGION_V2) {
+				adev->virt.fw_reserve.p_pf2vf =
+					(struct amd_sriov_msg_pf2vf_info_header *)
+					(adev->mman.fw_vram_usage_va +
+					adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID].offset);
+				adev->virt.fw_reserve.p_vf2pf =
+					(struct amd_sriov_msg_vf2pf_info_header *)
+					(adev->mman.fw_vram_usage_va +
+					adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID].offset +
+					(AMD_SRIOV_MSG_SIZE_KB << 10));
+				adev->virt.fw_reserve.ras_telemetry =
+					(adev->mman.fw_vram_usage_va +
+					adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_RAS_TELEMETRY_TABLE_ID].offset);
+			} else {
+				adev->virt.fw_reserve.p_pf2vf =
+					(struct amd_sriov_msg_pf2vf_info_header *)
+					(adev->mman.fw_vram_usage_va + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB_V1 << 10));
+				adev->virt.fw_reserve.p_vf2pf =
+					(struct amd_sriov_msg_vf2pf_info_header *)
+					(adev->mman.fw_vram_usage_va + (AMD_SRIOV_MSG_VF2PF_OFFSET_KB_V1 << 10));
+				adev->virt.fw_reserve.ras_telemetry =
+					(adev->mman.fw_vram_usage_va + (AMD_SRIOV_MSG_RAS_TELEMETRY_OFFSET_KB_V1 << 10));
+			}
 		} else if (adev->mman.drv_vram_usage_va) {
 			adev->virt.fw_reserve.p_pf2vf =
 				(struct amd_sriov_msg_pf2vf_info_header *)
-				(adev->mman.drv_vram_usage_va + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB << 10));
+				(adev->mman.drv_vram_usage_va + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB_V1 << 10));
 			adev->virt.fw_reserve.p_vf2pf =
 				(struct amd_sriov_msg_vf2pf_info_header *)
-				(adev->mman.drv_vram_usage_va + (AMD_SRIOV_MSG_VF2PF_OFFSET_KB << 10));
+				(adev->mman.drv_vram_usage_va + (AMD_SRIOV_MSG_VF2PF_OFFSET_KB_V1 << 10));
+			adev->virt.fw_reserve.ras_telemetry =
+				(adev->mman.drv_vram_usage_va + (AMD_SRIOV_MSG_RAS_TELEMETRY_OFFSET_KB_V1 << 10));
 		}
 
 		amdgpu_virt_read_pf2vf_data(adev);
@@ -711,7 +819,7 @@ void amdgpu_virt_exchange_data(struct amdgpu_device *adev)
 	}
 }
 
-void amdgpu_detect_virtualization(struct amdgpu_device *adev)
+static u32 amdgpu_virt_init_detect_asic(struct amdgpu_device *adev)
 {
 	uint32_t reg;
 
@@ -747,14 +855,17 @@ void amdgpu_detect_virtualization(struct amdgpu_device *adev)
 			adev->virt.caps |= AMDGPU_PASSTHROUGH_MODE;
 	}
 
-	if (amdgpu_sriov_vf(adev) && adev->asic_type == CHIP_SIENNA_CICHLID)
-		/* VF MMIO access (except mailbox range) from CPU
-		 * will be blocked during sriov runtime
-		 */
-		adev->virt.caps |= AMDGPU_VF_MMIO_ACCESS_PROTECT;
+	return reg;
+}
+
+static bool amdgpu_virt_init_req_data(struct amdgpu_device *adev, u32 reg)
+{
+	bool is_sriov = false;
 
 	/* we have the ability to check now */
 	if (amdgpu_sriov_vf(adev)) {
+		is_sriov = true;
+
 		switch (adev->asic_type) {
 		case CHIP_TONGA:
 		case CHIP_FIJI:
@@ -783,10 +894,252 @@ void amdgpu_detect_virtualization(struct amdgpu_device *adev)
 			amdgpu_virt_request_init_data(adev);
 			break;
 		default: /* other chip doesn't support SRIOV */
-			DRM_ERROR("Unknown asic type: %d!\n", adev->asic_type);
+			is_sriov = false;
+			dev_err(adev->dev, "Unknown asic type: %d!\n", adev->asic_type);
 			break;
 		}
 	}
+
+	return is_sriov;
+}
+
+static void amdgpu_virt_init_ras(struct amdgpu_device *adev)
+{
+	ratelimit_state_init(&adev->virt.ras.ras_error_cnt_rs, 5 * HZ, 1);
+	ratelimit_state_init(&adev->virt.ras.ras_cper_dump_rs, 5 * HZ, 1);
+	ratelimit_state_init(&adev->virt.ras.ras_chk_criti_rs, 5 * HZ, 1);
+
+	ratelimit_set_flags(&adev->virt.ras.ras_error_cnt_rs,
+			    RATELIMIT_MSG_ON_RELEASE);
+	ratelimit_set_flags(&adev->virt.ras.ras_cper_dump_rs,
+			    RATELIMIT_MSG_ON_RELEASE);
+	ratelimit_set_flags(&adev->virt.ras.ras_chk_criti_rs,
+			    RATELIMIT_MSG_ON_RELEASE);
+
+	mutex_init(&adev->virt.ras.ras_telemetry_mutex);
+	mutex_init(&adev->virt.access_req_mutex);
+
+	adev->virt.ras.cper_rptr = 0;
+}
+
+static uint8_t amdgpu_virt_crit_region_calc_checksum(uint8_t *buf_start, uint8_t *buf_end)
+{
+	uint32_t sum = 0;
+
+	if (buf_start >= buf_end)
+		return 0;
+
+	for (; buf_start < buf_end; buf_start++)
+		sum += buf_start[0];
+
+	return 0xffffffff - sum;
+}
+
+int amdgpu_virt_init_critical_region(struct amdgpu_device *adev)
+{
+	struct amd_sriov_msg_init_data_header *init_data_hdr = NULL;
+	u64 init_hdr_offset = adev->virt.init_data_header.offset;
+	u64 init_hdr_size = (u64)adev->virt.init_data_header.size_kb << 10; /* KB → bytes */
+	u64 vram_size;
+	u64 end;
+	int r = 0;
+	uint8_t checksum = 0;
+
+	/* Skip below init if critical region version != v2 */
+	if (adev->virt.req_init_data_ver != GPU_CRIT_REGION_V2)
+		return 0;
+
+	if (init_hdr_offset < 0) {
+		dev_err(adev->dev, "Invalid init header offset\n");
+		return -EINVAL;
+	}
+
+	vram_size = RREG32(mmRCC_CONFIG_MEMSIZE);
+	if (!vram_size || vram_size == U32_MAX)
+		return -EINVAL;
+	vram_size <<= 20;
+
+	if (check_add_overflow(init_hdr_offset, init_hdr_size, &end) || end > vram_size) {
+		dev_err(adev->dev, "init_data_header exceeds VRAM size, exiting\n");
+		return -EINVAL;
+	}
+
+	/* Allocate for init_data_hdr */
+	init_data_hdr = kzalloc(sizeof(struct amd_sriov_msg_init_data_header), GFP_KERNEL);
+	if (!init_data_hdr)
+		return -ENOMEM;
+
+	amdgpu_device_vram_access(adev, (uint64_t)init_hdr_offset, (uint32_t *)init_data_hdr,
+					sizeof(struct amd_sriov_msg_init_data_header), false);
+
+	/* Table validation */
+	if (strncmp(init_data_hdr->signature,
+				AMDGPU_SRIOV_CRIT_DATA_SIGNATURE,
+				AMDGPU_SRIOV_CRIT_DATA_SIG_LEN) != 0) {
+		dev_err(adev->dev, "Invalid init data signature: %.4s\n",
+			init_data_hdr->signature);
+		r = -EINVAL;
+		goto out;
+	}
+
+	checksum = amdgpu_virt_crit_region_calc_checksum(
+			(uint8_t *)&init_data_hdr->initdata_offset,
+			(uint8_t *)init_data_hdr +
+			sizeof(struct amd_sriov_msg_init_data_header));
+	if (checksum != init_data_hdr->checksum) {
+		dev_err(adev->dev, "Found unmatching checksum from calculation 0x%x and init_data 0x%x\n",
+				checksum, init_data_hdr->checksum);
+		r = -EINVAL;
+		goto out;
+	}
+
+	memset(&adev->virt.crit_regn, 0, sizeof(adev->virt.crit_regn));
+	memset(adev->virt.crit_regn_tbl, 0, sizeof(adev->virt.crit_regn_tbl));
+
+	adev->virt.crit_regn.offset = init_data_hdr->initdata_offset;
+	adev->virt.crit_regn.size_kb = init_data_hdr->initdata_size_in_kb;
+
+	/* Validation and initialization for each table entry */
+	if (IS_SRIOV_CRIT_REGN_ENTRY_VALID(init_data_hdr, AMD_SRIOV_MSG_IPD_TABLE_ID)) {
+		if (!init_data_hdr->ip_discovery_size_in_kb ||
+				init_data_hdr->ip_discovery_size_in_kb > DISCOVERY_TMR_SIZE) {
+			dev_err(adev->dev, "Invalid %s size: 0x%x\n",
+				amdgpu_virt_dynamic_crit_table_name[AMD_SRIOV_MSG_IPD_TABLE_ID],
+				init_data_hdr->ip_discovery_size_in_kb);
+			r = -EINVAL;
+			goto out;
+		}
+
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_IPD_TABLE_ID].offset =
+			init_data_hdr->ip_discovery_offset;
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_IPD_TABLE_ID].size_kb =
+			init_data_hdr->ip_discovery_size_in_kb;
+	}
+
+	if (IS_SRIOV_CRIT_REGN_ENTRY_VALID(init_data_hdr, AMD_SRIOV_MSG_VBIOS_IMG_TABLE_ID)) {
+		if (!init_data_hdr->vbios_img_size_in_kb) {
+			dev_err(adev->dev, "Invalid %s size: 0x%x\n",
+				amdgpu_virt_dynamic_crit_table_name[AMD_SRIOV_MSG_VBIOS_IMG_TABLE_ID],
+				init_data_hdr->vbios_img_size_in_kb);
+			r = -EINVAL;
+			goto out;
+		}
+
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_VBIOS_IMG_TABLE_ID].offset =
+			init_data_hdr->vbios_img_offset;
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_VBIOS_IMG_TABLE_ID].size_kb =
+			init_data_hdr->vbios_img_size_in_kb;
+	}
+
+	if (IS_SRIOV_CRIT_REGN_ENTRY_VALID(init_data_hdr, AMD_SRIOV_MSG_RAS_TELEMETRY_TABLE_ID)) {
+		if (!init_data_hdr->ras_tele_info_size_in_kb) {
+			dev_err(adev->dev, "Invalid %s size: 0x%x\n",
+				amdgpu_virt_dynamic_crit_table_name[AMD_SRIOV_MSG_RAS_TELEMETRY_TABLE_ID],
+				init_data_hdr->ras_tele_info_size_in_kb);
+			r = -EINVAL;
+			goto out;
+		}
+
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_RAS_TELEMETRY_TABLE_ID].offset =
+			init_data_hdr->ras_tele_info_offset;
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_RAS_TELEMETRY_TABLE_ID].size_kb =
+			init_data_hdr->ras_tele_info_size_in_kb;
+	}
+
+	if (IS_SRIOV_CRIT_REGN_ENTRY_VALID(init_data_hdr, AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID)) {
+		if (!init_data_hdr->dataexchange_size_in_kb) {
+			dev_err(adev->dev, "Invalid %s size: 0x%x\n",
+				amdgpu_virt_dynamic_crit_table_name[AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID],
+				init_data_hdr->dataexchange_size_in_kb);
+			r = -EINVAL;
+			goto out;
+		}
+
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID].offset =
+			init_data_hdr->dataexchange_offset;
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID].size_kb =
+			init_data_hdr->dataexchange_size_in_kb;
+	}
+
+	if (IS_SRIOV_CRIT_REGN_ENTRY_VALID(init_data_hdr, AMD_SRIOV_MSG_BAD_PAGE_INFO_TABLE_ID)) {
+		if (!init_data_hdr->bad_page_size_in_kb) {
+			dev_err(adev->dev, "Invalid %s size: 0x%x\n",
+				amdgpu_virt_dynamic_crit_table_name[AMD_SRIOV_MSG_BAD_PAGE_INFO_TABLE_ID],
+				init_data_hdr->bad_page_size_in_kb);
+			r = -EINVAL;
+			goto out;
+		}
+
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_BAD_PAGE_INFO_TABLE_ID].offset =
+			init_data_hdr->bad_page_info_offset;
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_BAD_PAGE_INFO_TABLE_ID].size_kb =
+			init_data_hdr->bad_page_size_in_kb;
+	}
+
+	/* Validation for critical region info */
+	if (adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_IPD_TABLE_ID].size_kb > DISCOVERY_TMR_SIZE) {
+		dev_err(adev->dev, "Invalid IP discovery size: 0x%x\n",
+				adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_IPD_TABLE_ID].size_kb);
+		r = -EINVAL;
+		goto out;
+	}
+
+	/* reserved memory starts from crit region base offset with the size of 5MB */
+	adev->mman.fw_vram_usage_start_offset = adev->virt.crit_regn.offset;
+	adev->mman.fw_vram_usage_size = adev->virt.crit_regn.size_kb << 10;
+	dev_info(adev->dev,
+		"critical region v%d requested to reserve memory start at %08llx with %llu KB.\n",
+			init_data_hdr->version,
+			adev->mman.fw_vram_usage_start_offset,
+			adev->mman.fw_vram_usage_size >> 10);
+
+	adev->virt.is_dynamic_crit_regn_enabled = true;
+
+out:
+	kfree(init_data_hdr);
+	init_data_hdr = NULL;
+
+	return r;
+}
+
+int amdgpu_virt_get_dynamic_data_info(struct amdgpu_device *adev,
+	int data_id, uint8_t *binary, u32 *size)
+{
+	uint32_t data_offset = 0;
+	uint32_t data_size = 0;
+	enum amd_sriov_msg_table_id_enum data_table_id = data_id;
+
+	if (data_table_id >= AMD_SRIOV_MSG_MAX_TABLE_ID)
+		return -EINVAL;
+
+	data_offset = adev->virt.crit_regn_tbl[data_table_id].offset;
+	data_size = adev->virt.crit_regn_tbl[data_table_id].size_kb << 10;
+
+	/* Validate on input params */
+	if (!binary || !size || *size < (uint64_t)data_size)
+		return -EINVAL;
+
+	/* Proceed to copy the dynamic content */
+	amdgpu_device_vram_access(adev,
+			(uint64_t)data_offset, (uint32_t *)binary, data_size, false);
+	*size = (uint64_t)data_size;
+
+	dev_dbg(adev->dev,
+		"Got %s info from dynamic crit_region_table at offset 0x%x with size of 0x%x bytes.\n",
+		amdgpu_virt_dynamic_crit_table_name[data_id], data_offset, data_size);
+
+	return 0;
+}
+
+void amdgpu_virt_init(struct amdgpu_device *adev)
+{
+	bool is_sriov = false;
+	uint32_t reg = amdgpu_virt_init_detect_asic(adev);
+
+	is_sriov = amdgpu_virt_init_req_data(adev, reg);
+
+	if (is_sriov)
+		amdgpu_virt_init_ras(adev);
 }
 
 static bool amdgpu_virt_access_debugfs_is_mmio(struct amdgpu_device *adev)
@@ -835,9 +1188,28 @@ enum amdgpu_sriov_vf_mode amdgpu_virt_get_sriov_vf_mode(struct amdgpu_device *ad
 	return mode;
 }
 
+void amdgpu_virt_pre_reset(struct amdgpu_device *adev)
+{
+	/* stop the data exchange thread */
+	amdgpu_virt_fini_data_exchange(adev);
+	amdgpu_dpm_set_mp1_state(adev, PP_MP1_STATE_FLR);
+}
+
+void amdgpu_virt_post_reset(struct amdgpu_device *adev)
+{
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(11, 0, 3)) {
+		/* force set to GFXOFF state after reset,
+		 * to avoid some invalid operation before GC enable
+		 */
+		adev->gfx.is_poweron = false;
+	}
+
+	adev->mes.ring[0].sched.ready = false;
+}
+
 bool amdgpu_virt_fw_load_skip_check(struct amdgpu_device *adev, uint32_t ucode_id)
 {
-	switch (adev->ip_versions[MP0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 	case IP_VERSION(13, 0, 0):
 		/* no vf autoload, white list */
 		if (ucode_id == AMDGPU_UCODE_ID_VCN1 ||
@@ -845,6 +1217,17 @@ bool amdgpu_virt_fw_load_skip_check(struct amdgpu_device *adev, uint32_t ucode_i
 			return false;
 		else
 			return true;
+	case IP_VERSION(11, 0, 9):
+	case IP_VERSION(11, 0, 7):
+		/* black list for CHIP_NAVI12 and CHIP_SIENNA_CICHLID */
+		if (ucode_id == AMDGPU_UCODE_ID_RLC_G
+		    || ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL
+		    || ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM
+		    || ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM
+		    || ucode_id == AMDGPU_UCODE_ID_SMC)
+			return true;
+		else
+			return false;
 	case IP_VERSION(13, 0, 10):
 		/* white list */
 		if (ucode_id == AMDGPU_UCODE_ID_CAP
@@ -921,7 +1304,7 @@ void amdgpu_virt_update_sriov_video_codec(struct amdgpu_device *adev,
 	}
 }
 
-static bool amdgpu_virt_get_rlcg_reg_access_flag(struct amdgpu_device *adev,
+bool amdgpu_virt_get_rlcg_reg_access_flag(struct amdgpu_device *adev,
 						 u32 acc_flags, u32 hwip,
 						 bool write, u32 *rlcg_flag)
 {
@@ -954,7 +1337,7 @@ static bool amdgpu_virt_get_rlcg_reg_access_flag(struct amdgpu_device *adev,
 	return ret;
 }
 
-static u32 amdgpu_virt_rlcg_reg_rw(struct amdgpu_device *adev, u32 offset, u32 v, u32 flag)
+u32 amdgpu_virt_rlcg_reg_rw(struct amdgpu_device *adev, u32 offset, u32 v, u32 flag, u32 xcc_id)
 {
 	struct amdgpu_rlcg_reg_access_ctrl *reg_access_ctrl;
 	uint32_t timeout = 50000;
@@ -965,6 +1348,7 @@ static u32 amdgpu_virt_rlcg_reg_rw(struct amdgpu_device *adev, u32 offset, u32 v
 	void *scratch_reg2;
 	void *scratch_reg3;
 	void *spare_int;
+	unsigned long flags;
 
 	if (!adev->gfx.rlc.rlcg_reg_access_supported) {
 		dev_err(adev->dev,
@@ -972,11 +1356,22 @@ static u32 amdgpu_virt_rlcg_reg_rw(struct amdgpu_device *adev, u32 offset, u32 v
 		return 0;
 	}
 
-	reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl;
+	if (adev->gfx.xcc_mask && (((1 << xcc_id) & adev->gfx.xcc_mask) == 0)) {
+		dev_err(adev->dev, "invalid xcc\n");
+		return 0;
+	}
+
+	if (amdgpu_device_skip_hw_access(adev))
+		return 0;
+
+	reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl[xcc_id];
 	scratch_reg0 = (void __iomem *)adev->rmmio + 4 * reg_access_ctrl->scratch_reg0;
 	scratch_reg1 = (void __iomem *)adev->rmmio + 4 * reg_access_ctrl->scratch_reg1;
 	scratch_reg2 = (void __iomem *)adev->rmmio + 4 * reg_access_ctrl->scratch_reg2;
 	scratch_reg3 = (void __iomem *)adev->rmmio + 4 * reg_access_ctrl->scratch_reg3;
+
+	spin_lock_irqsave(&adev->virt.rlcg_reg_lock, flags);
+
 	if (reg_access_ctrl->spare_int)
 		spare_int = (void __iomem *)adev->rmmio + 4 * reg_access_ctrl->spare_int;
 
@@ -995,7 +1390,7 @@ static u32 amdgpu_virt_rlcg_reg_rw(struct amdgpu_device *adev, u32 offset, u32 v
 		 * SCRATCH_REG0 	= read/write value
 		 * SCRATCH_REG1[30:28]	= command
 		 * SCRATCH_REG1[19:0]	= address in dword
-		 * SCRATCH_REG1[26:24]	= Error reporting
+		 * SCRATCH_REG1[27:24]	= Error reporting
 		 */
 		writel(v, scratch_reg0);
 		writel((offset | flag), scratch_reg1);
@@ -1009,7 +1404,8 @@ static u32 amdgpu_virt_rlcg_reg_rw(struct amdgpu_device *adev, u32 offset, u32 v
 			udelay(10);
 		}
 
-		if (i >= timeout) {
+		tmp = readl(scratch_reg1);
+		if (i >= timeout || (tmp & AMDGPU_RLCG_SCRATCH1_ERROR_MASK) != 0) {
 			if (amdgpu_sriov_rlcg_error_report_enabled(adev)) {
 				if (tmp & AMDGPU_RLCG_VFGATE_DISABLED) {
 					dev_err(adev->dev,
@@ -1032,18 +1428,24 @@ static u32 amdgpu_virt_rlcg_reg_rw(struct amdgpu_device *adev, u32 offset, u32 v
 	}
 
 	ret = readl(scratch_reg0);
+
+	spin_unlock_irqrestore(&adev->virt.rlcg_reg_lock, flags);
+
 	return ret;
 }
 
 void amdgpu_sriov_wreg(struct amdgpu_device *adev,
 		       u32 offset, u32 value,
-		       u32 acc_flags, u32 hwip)
+		       u32 acc_flags, u32 hwip, u32 xcc_id)
 {
 	u32 rlcg_flag;
 
+	if (amdgpu_device_skip_hw_access(adev))
+		return;
+
 	if (!amdgpu_sriov_runtime(adev) &&
 		amdgpu_virt_get_rlcg_reg_access_flag(adev, acc_flags, hwip, true, &rlcg_flag)) {
-		amdgpu_virt_rlcg_reg_rw(adev, offset, value, rlcg_flag);
+		amdgpu_virt_rlcg_reg_rw(adev, offset, value, rlcg_flag, xcc_id);
 		return;
 	}
 
@@ -1054,16 +1456,392 @@ void amdgpu_sriov_wreg(struct amdgpu_device *adev,
 }
 
 u32 amdgpu_sriov_rreg(struct amdgpu_device *adev,
-		      u32 offset, u32 acc_flags, u32 hwip)
+		      u32 offset, u32 acc_flags, u32 hwip, u32 xcc_id)
 {
 	u32 rlcg_flag;
 
+	if (amdgpu_device_skip_hw_access(adev))
+		return 0;
+
 	if (!amdgpu_sriov_runtime(adev) &&
 		amdgpu_virt_get_rlcg_reg_access_flag(adev, acc_flags, hwip, false, &rlcg_flag))
-		return amdgpu_virt_rlcg_reg_rw(adev, offset, 0, rlcg_flag);
+		return amdgpu_virt_rlcg_reg_rw(adev, offset, 0, rlcg_flag, xcc_id);
 
 	if (acc_flags & AMDGPU_REGS_NO_KIQ)
 		return RREG32_NO_KIQ(offset);
 	else
 		return RREG32(offset);
 }
+
+bool amdgpu_sriov_xnack_support(struct amdgpu_device *adev)
+{
+	bool xnack_mode = true;
+
+	if (amdgpu_sriov_vf(adev) &&
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2))
+		xnack_mode = false;
+
+	return xnack_mode;
+}
+
+bool amdgpu_virt_get_ras_capability(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+
+	if (!amdgpu_sriov_ras_caps_en(adev))
+		return false;
+
+	if (adev->virt.ras_en_caps.bits.block_umc)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__UMC);
+	if (adev->virt.ras_en_caps.bits.block_sdma)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__SDMA);
+	if (adev->virt.ras_en_caps.bits.block_gfx)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__GFX);
+	if (adev->virt.ras_en_caps.bits.block_mmhub)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__MMHUB);
+	if (adev->virt.ras_en_caps.bits.block_athub)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__ATHUB);
+	if (adev->virt.ras_en_caps.bits.block_pcie_bif)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__PCIE_BIF);
+	if (adev->virt.ras_en_caps.bits.block_hdp)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__HDP);
+	if (adev->virt.ras_en_caps.bits.block_xgmi_wafl)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__XGMI_WAFL);
+	if (adev->virt.ras_en_caps.bits.block_df)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__DF);
+	if (adev->virt.ras_en_caps.bits.block_smn)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__SMN);
+	if (adev->virt.ras_en_caps.bits.block_sem)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__SEM);
+	if (adev->virt.ras_en_caps.bits.block_mp0)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__MP0);
+	if (adev->virt.ras_en_caps.bits.block_mp1)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__MP1);
+	if (adev->virt.ras_en_caps.bits.block_fuse)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__FUSE);
+	if (adev->virt.ras_en_caps.bits.block_mca)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__MCA);
+	if (adev->virt.ras_en_caps.bits.block_vcn)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__VCN);
+	if (adev->virt.ras_en_caps.bits.block_jpeg)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__JPEG);
+	if (adev->virt.ras_en_caps.bits.block_ih)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__IH);
+	if (adev->virt.ras_en_caps.bits.block_mpio)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__MPIO);
+
+	if (adev->virt.ras_en_caps.bits.poison_propogation_mode)
+		con->poison_supported = true; /* Poison is handled by host */
+
+	return true;
+}
+
+static inline enum amd_sriov_ras_telemetry_gpu_block
+amdgpu_ras_block_to_sriov(struct amdgpu_device *adev, enum amdgpu_ras_block block) {
+	switch (block) {
+	case AMDGPU_RAS_BLOCK__UMC:
+		return RAS_TELEMETRY_GPU_BLOCK_UMC;
+	case AMDGPU_RAS_BLOCK__SDMA:
+		return RAS_TELEMETRY_GPU_BLOCK_SDMA;
+	case AMDGPU_RAS_BLOCK__GFX:
+		return RAS_TELEMETRY_GPU_BLOCK_GFX;
+	case AMDGPU_RAS_BLOCK__MMHUB:
+		return RAS_TELEMETRY_GPU_BLOCK_MMHUB;
+	case AMDGPU_RAS_BLOCK__ATHUB:
+		return RAS_TELEMETRY_GPU_BLOCK_ATHUB;
+	case AMDGPU_RAS_BLOCK__PCIE_BIF:
+		return RAS_TELEMETRY_GPU_BLOCK_PCIE_BIF;
+	case AMDGPU_RAS_BLOCK__HDP:
+		return RAS_TELEMETRY_GPU_BLOCK_HDP;
+	case AMDGPU_RAS_BLOCK__XGMI_WAFL:
+		return RAS_TELEMETRY_GPU_BLOCK_XGMI_WAFL;
+	case AMDGPU_RAS_BLOCK__DF:
+		return RAS_TELEMETRY_GPU_BLOCK_DF;
+	case AMDGPU_RAS_BLOCK__SMN:
+		return RAS_TELEMETRY_GPU_BLOCK_SMN;
+	case AMDGPU_RAS_BLOCK__SEM:
+		return RAS_TELEMETRY_GPU_BLOCK_SEM;
+	case AMDGPU_RAS_BLOCK__MP0:
+		return RAS_TELEMETRY_GPU_BLOCK_MP0;
+	case AMDGPU_RAS_BLOCK__MP1:
+		return RAS_TELEMETRY_GPU_BLOCK_MP1;
+	case AMDGPU_RAS_BLOCK__FUSE:
+		return RAS_TELEMETRY_GPU_BLOCK_FUSE;
+	case AMDGPU_RAS_BLOCK__MCA:
+		return RAS_TELEMETRY_GPU_BLOCK_MCA;
+	case AMDGPU_RAS_BLOCK__VCN:
+		return RAS_TELEMETRY_GPU_BLOCK_VCN;
+	case AMDGPU_RAS_BLOCK__JPEG:
+		return RAS_TELEMETRY_GPU_BLOCK_JPEG;
+	case AMDGPU_RAS_BLOCK__IH:
+		return RAS_TELEMETRY_GPU_BLOCK_IH;
+	case AMDGPU_RAS_BLOCK__MPIO:
+		return RAS_TELEMETRY_GPU_BLOCK_MPIO;
+	default:
+		dev_warn(adev->dev, "Unsupported SRIOV RAS telemetry block 0x%x\n",
+			      block);
+		return RAS_TELEMETRY_GPU_BLOCK_COUNT;
+	}
+}
+
+static int amdgpu_virt_cache_host_error_counts(struct amdgpu_device *adev,
+					       struct amdsriov_ras_telemetry *host_telemetry)
+{
+	struct amd_sriov_ras_telemetry_error_count *tmp = NULL;
+	uint32_t checksum, used_size;
+
+	checksum = host_telemetry->header.checksum;
+	used_size = host_telemetry->header.used_size;
+
+	if (used_size > (AMD_SRIOV_MSG_RAS_TELEMETRY_SIZE_KB_V1 << 10))
+		return 0;
+
+	tmp = kmemdup(&host_telemetry->body.error_count, used_size, GFP_KERNEL);
+	if (!tmp)
+		return -ENOMEM;
+
+	if (checksum != amd_sriov_msg_checksum(tmp, used_size, 0, 0))
+		goto out;
+
+	memcpy(&adev->virt.count_cache, tmp,
+	       min(used_size, sizeof(adev->virt.count_cache)));
+out:
+	kfree(tmp);
+
+	return 0;
+}
+
+static int amdgpu_virt_req_ras_err_count_internal(struct amdgpu_device *adev, bool force_update)
+{
+	struct amdgpu_virt *virt = &adev->virt;
+
+	if (!virt->ops || !virt->ops->req_ras_err_count)
+		return -EOPNOTSUPP;
+
+	/* Host allows 15 ras telemetry requests per 60 seconds. Afterwhich, the Host
+	 * will ignore incoming guest messages. Ratelimit the guest messages to
+	 * prevent guest self DOS.
+	 */
+	if (__ratelimit(&virt->ras.ras_error_cnt_rs) || force_update) {
+		mutex_lock(&virt->ras.ras_telemetry_mutex);
+		if (!virt->ops->req_ras_err_count(adev))
+			amdgpu_virt_cache_host_error_counts(adev,
+				virt->fw_reserve.ras_telemetry);
+		mutex_unlock(&virt->ras.ras_telemetry_mutex);
+	}
+
+	return 0;
+}
+
+/* Bypass ACA interface and query ECC counts directly from host */
+int amdgpu_virt_req_ras_err_count(struct amdgpu_device *adev, enum amdgpu_ras_block block,
+				  struct ras_err_data *err_data)
+{
+	enum amd_sriov_ras_telemetry_gpu_block sriov_block;
+
+	sriov_block = amdgpu_ras_block_to_sriov(adev, block);
+
+	if (sriov_block >= RAS_TELEMETRY_GPU_BLOCK_COUNT ||
+	    !amdgpu_sriov_ras_telemetry_block_en(adev, sriov_block))
+		return -EOPNOTSUPP;
+
+	/* Host Access may be lost during reset, just return last cached data. */
+	if (down_read_trylock(&adev->reset_domain->sem)) {
+		amdgpu_virt_req_ras_err_count_internal(adev, false);
+		up_read(&adev->reset_domain->sem);
+	}
+
+	err_data->ue_count = adev->virt.count_cache.block[sriov_block].ue_count;
+	err_data->ce_count = adev->virt.count_cache.block[sriov_block].ce_count;
+	err_data->de_count = adev->virt.count_cache.block[sriov_block].de_count;
+
+	return 0;
+}
+
+static int
+amdgpu_virt_write_cpers_to_ring(struct amdgpu_device *adev,
+				struct amdsriov_ras_telemetry *host_telemetry,
+				u32 *more)
+{
+	struct amd_sriov_ras_cper_dump *cper_dump = NULL;
+	struct cper_hdr *entry = NULL;
+	struct amdgpu_ring *ring = &adev->cper.ring_buf;
+	uint32_t checksum, used_size, i;
+	int ret = 0;
+
+	checksum = host_telemetry->header.checksum;
+	used_size = host_telemetry->header.used_size;
+
+	if (used_size > (AMD_SRIOV_MSG_RAS_TELEMETRY_SIZE_KB_V1 << 10))
+		return -EINVAL;
+
+	cper_dump = kmemdup(&host_telemetry->body.cper_dump, used_size, GFP_KERNEL);
+	if (!cper_dump)
+		return -ENOMEM;
+
+	if (checksum != amd_sriov_msg_checksum(cper_dump, used_size, 0, 0)) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	*more = cper_dump->more;
+
+	if (cper_dump->wptr < adev->virt.ras.cper_rptr) {
+		dev_warn(
+			adev->dev,
+			"guest specified rptr that was too high! guest rptr: 0x%llx, host rptr: 0x%llx\n",
+			adev->virt.ras.cper_rptr, cper_dump->wptr);
+
+		adev->virt.ras.cper_rptr = cper_dump->wptr;
+		goto out;
+	}
+
+	entry = (struct cper_hdr *)&cper_dump->buf[0];
+
+	for (i = 0; i < cper_dump->count; i++) {
+		amdgpu_cper_ring_write(ring, entry, entry->record_length);
+		entry = (struct cper_hdr *)((char *)entry +
+					    entry->record_length);
+	}
+
+	if (cper_dump->overflow_count)
+		dev_warn(adev->dev,
+			 "host reported CPER overflow of 0x%llx entries!\n",
+			 cper_dump->overflow_count);
+
+	adev->virt.ras.cper_rptr = cper_dump->wptr;
+out:
+	kfree(cper_dump);
+
+	return ret;
+}
+
+static int amdgpu_virt_req_ras_cper_dump_internal(struct amdgpu_device *adev)
+{
+	struct amdgpu_virt *virt = &adev->virt;
+	int ret = 0;
+	uint32_t more = 0;
+
+	if (!virt->ops || !virt->ops->req_ras_cper_dump)
+		return -EOPNOTSUPP;
+
+	do {
+		if (!virt->ops->req_ras_cper_dump(adev, virt->ras.cper_rptr))
+			ret = amdgpu_virt_write_cpers_to_ring(
+				adev, virt->fw_reserve.ras_telemetry, &more);
+		else
+			ret = 0;
+	} while (more && !ret);
+
+	return ret;
+}
+
+int amdgpu_virt_req_ras_cper_dump(struct amdgpu_device *adev, bool force_update)
+{
+	struct amdgpu_virt *virt = &adev->virt;
+	int ret = 0;
+
+	if (!amdgpu_sriov_ras_cper_en(adev))
+		return -EOPNOTSUPP;
+
+	if ((__ratelimit(&virt->ras.ras_cper_dump_rs) || force_update) &&
+	    down_read_trylock(&adev->reset_domain->sem)) {
+		mutex_lock(&virt->ras.ras_telemetry_mutex);
+		ret = amdgpu_virt_req_ras_cper_dump_internal(adev);
+		mutex_unlock(&virt->ras.ras_telemetry_mutex);
+		up_read(&adev->reset_domain->sem);
+	}
+
+	return ret;
+}
+
+int amdgpu_virt_ras_telemetry_post_reset(struct amdgpu_device *adev)
+{
+	unsigned long ue_count, ce_count;
+
+	if (amdgpu_sriov_ras_telemetry_en(adev)) {
+		amdgpu_virt_req_ras_err_count_internal(adev, true);
+		amdgpu_ras_query_error_count(adev, &ce_count, &ue_count, NULL);
+	}
+
+	return 0;
+}
+
+bool amdgpu_virt_ras_telemetry_block_en(struct amdgpu_device *adev,
+					enum amdgpu_ras_block block)
+{
+	enum amd_sriov_ras_telemetry_gpu_block sriov_block;
+
+	sriov_block = amdgpu_ras_block_to_sriov(adev, block);
+
+	if (sriov_block >= RAS_TELEMETRY_GPU_BLOCK_COUNT ||
+	    !amdgpu_sriov_ras_telemetry_block_en(adev, sriov_block))
+		return false;
+
+	return true;
+}
+
+/*
+ * amdgpu_virt_request_bad_pages() - request bad pages
+ * @adev: amdgpu device.
+ * Send command to GPU hypervisor to write new bad pages into the shared PF2VF region
+ */
+void amdgpu_virt_request_bad_pages(struct amdgpu_device *adev)
+{
+	struct amdgpu_virt *virt = &adev->virt;
+
+	if (virt->ops && virt->ops->req_bad_pages)
+		virt->ops->req_bad_pages(adev);
+}
+
+static int amdgpu_virt_cache_chk_criti_hit(struct amdgpu_device *adev,
+					   struct amdsriov_ras_telemetry *host_telemetry,
+					   bool *hit)
+{
+	struct amd_sriov_ras_chk_criti *tmp = NULL;
+	uint32_t checksum, used_size;
+
+	checksum = host_telemetry->header.checksum;
+	used_size = host_telemetry->header.used_size;
+
+	if (used_size > (AMD_SRIOV_MSG_RAS_TELEMETRY_SIZE_KB_V1 << 10))
+		return 0;
+
+	tmp = kmemdup(&host_telemetry->body.chk_criti, used_size, GFP_KERNEL);
+	if (!tmp)
+		return -ENOMEM;
+
+	if (checksum != amd_sriov_msg_checksum(tmp, used_size, 0, 0))
+		goto out;
+
+	if (hit)
+		*hit = tmp->hit ? true : false;
+
+out:
+	kfree(tmp);
+
+	return 0;
+}
+
+int amdgpu_virt_check_vf_critical_region(struct amdgpu_device *adev, u64 addr, bool *hit)
+{
+	struct amdgpu_virt *virt = &adev->virt;
+	int r = -EPERM;
+
+	if (!virt->ops || !virt->ops->req_ras_chk_criti)
+		return -EOPNOTSUPP;
+
+	/* Host allows 15 ras telemetry requests per 60 seconds. Afterwhich, the Host
+	 * will ignore incoming guest messages. Ratelimit the guest messages to
+	 * prevent guest self DOS.
+	 */
+	if (__ratelimit(&virt->ras.ras_chk_criti_rs)) {
+		mutex_lock(&virt->ras.ras_telemetry_mutex);
+		if (!virt->ops->req_ras_chk_criti(adev, addr))
+			r = amdgpu_virt_cache_chk_criti_hit(
+				adev, virt->fw_reserve.ras_telemetry, hit);
+		mutex_unlock(&virt->ras.ras_telemetry_mutex);
+	}
+
+	return r;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
index 4f7bab52282a..01d5bca2dee1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
@@ -45,12 +45,21 @@
 #define AMDGPU_RLCG_REG_NOT_IN_RANGE		0x1000000
 
 #define AMDGPU_RLCG_SCRATCH1_ADDRESS_MASK	0xFFFFF
+#define AMDGPU_RLCG_SCRATCH1_ERROR_MASK	0xF000000
 
 /* all asic after AI use this offset */
 #define mmRCC_IOV_FUNC_IDENTIFIER 0xDE5
 /* tonga/fiji use this offset */
 #define mmBIF_IOV_FUNC_IDENTIFIER 0x1503
 
+#define AMDGPU_VF2PF_UPDATE_MAX_RETRY_LIMIT 2
+
+/* Signature used to validate the SR-IOV dynamic critical region init data header ("INDA") */
+#define AMDGPU_SRIOV_CRIT_DATA_SIGNATURE "INDA"
+#define AMDGPU_SRIOV_CRIT_DATA_SIG_LEN   4
+
+#define IS_SRIOV_CRIT_REGN_ENTRY_VALID(hdr, id) ((hdr)->valid_tables & (1 << (id)))
+
 enum amdgpu_sriov_vf_mode {
 	SRIOV_VF_MODE_BARE_METAL = 0,
 	SRIOV_VF_MODE_ONE_VF,
@@ -85,10 +94,17 @@ struct amdgpu_virt_ops {
 	int (*rel_full_gpu)(struct amdgpu_device *adev, bool init);
 	int (*req_init_data)(struct amdgpu_device *adev);
 	int (*reset_gpu)(struct amdgpu_device *adev);
+	void (*ready_to_reset)(struct amdgpu_device *adev);
 	int (*wait_reset)(struct amdgpu_device *adev);
 	void (*trans_msg)(struct amdgpu_device *adev, enum idh_request req,
 			  u32 data1, u32 data2, u32 data3);
-	void (*ras_poison_handler)(struct amdgpu_device *adev);
+	void (*ras_poison_handler)(struct amdgpu_device *adev,
+					enum amdgpu_ras_block block);
+	bool (*rcvd_ras_intr)(struct amdgpu_device *adev);
+	int (*req_ras_err_count)(struct amdgpu_device *adev);
+	int (*req_ras_cper_dump)(struct amdgpu_device *adev, u64 vf_rptr);
+	int (*req_bad_pages)(struct amdgpu_device *adev);
+	int (*req_ras_chk_criti)(struct amdgpu_device *adev, u64 addr);
 };
 
 /*
@@ -97,6 +113,7 @@ struct amdgpu_virt_ops {
 struct amdgpu_virt_fw_reserve {
 	struct amd_sriov_msg_pf2vf_info_header *p_pf2vf;
 	struct amd_sriov_msg_vf2pf_info_header *p_vf2pf;
+	void *ras_telemetry;
 	unsigned int checksum_key;
 };
 
@@ -126,15 +143,27 @@ enum AMDGIM_FEATURE_FLAG {
 	AMDGIM_FEATURE_INDIRECT_REG_ACCESS = (1 << 5),
 	/* AV1 Support MODE*/
 	AMDGIM_FEATURE_AV1_SUPPORT = (1 << 6),
+	/* VCN RB decouple */
+	AMDGIM_FEATURE_VCN_RB_DECOUPLE = (1 << 7),
+	/* MES info */
+	AMDGIM_FEATURE_MES_INFO_ENABLE = (1 << 8),
+	AMDGIM_FEATURE_RAS_CAPS = (1 << 9),
+	AMDGIM_FEATURE_RAS_TELEMETRY = (1 << 10),
+	AMDGIM_FEATURE_RAS_CPER = (1 << 11),
+	AMDGIM_FEATURE_XGMI_TA_EXT_PEER_LINK = (1 << 12),
 };
 
 enum AMDGIM_REG_ACCESS_FLAG {
 	/* Use PSP to program IH_RB_CNTL */
-	AMDGIM_FEATURE_IH_REG_PSP_EN     = (1 << 0),
+	AMDGIM_FEATURE_IH_REG_PSP_EN      = (1 << 0),
 	/* Use RLC to program MMHUB regs */
-	AMDGIM_FEATURE_MMHUB_REG_RLC_EN  = (1 << 1),
+	AMDGIM_FEATURE_MMHUB_REG_RLC_EN   = (1 << 1),
 	/* Use RLC to program GC regs */
-	AMDGIM_FEATURE_GC_REG_RLC_EN     = (1 << 2),
+	AMDGIM_FEATURE_GC_REG_RLC_EN      = (1 << 2),
+	/* Use PSP to program L1_TLB_CNTL */
+	AMDGIM_FEATURE_L1_TLB_CNTL_PSP_EN = (1 << 3),
+	/* Use RLCG to program SQ_CONFIG1 */
+	AMDGIM_FEATURE_REG_ACCESS_SQ_CONFIG = (1 << 4),
 };
 
 struct amdgim_pf2vf_info_v1 {
@@ -228,6 +257,23 @@ struct amdgpu_virt_ras_err_handler_data {
 	int last_reserved;
 };
 
+struct amdgpu_virt_ras {
+	struct ratelimit_state ras_error_cnt_rs;
+	struct ratelimit_state ras_cper_dump_rs;
+	struct ratelimit_state ras_chk_criti_rs;
+	struct mutex ras_telemetry_mutex;
+	uint64_t cper_rptr;
+};
+
+#define AMDGPU_VIRT_CAPS_LIST(X) X(AMDGPU_VIRT_CAP_POWER_LIMIT)
+
+DECLARE_ATTR_CAP_CLASS(amdgpu_virt, AMDGPU_VIRT_CAPS_LIST);
+
+struct amdgpu_virt_region {
+	uint32_t offset;
+	uint32_t size_kb;
+};
+
 /* GPU virtualization */
 struct amdgpu_virt {
 	uint32_t			caps;
@@ -237,11 +283,16 @@ struct amdgpu_virt {
 	uint32_t			reg_val_offs;
 	struct amdgpu_irq_src		ack_irq;
 	struct amdgpu_irq_src		rcv_irq;
+
 	struct work_struct		flr_work;
+	struct work_struct		req_bad_pages_work;
+	struct work_struct		handle_bad_pages_work;
+
 	struct amdgpu_mm_table		mm_table;
 	const struct amdgpu_virt_ops	*ops;
 	struct amdgpu_vf_error_buffer	vf_errors;
 	struct amdgpu_virt_fw_reserve	fw_reserve;
+	struct amdgpu_virt_caps virt_caps;
 	uint32_t gim_feature;
 	uint32_t reg_access_mode;
 	int req_init_data_ver;
@@ -250,9 +301,16 @@ struct amdgpu_virt {
 	bool ras_init_done;
 	uint32_t reg_access;
 
+	/* dynamic(v2) critical regions */
+	struct amdgpu_virt_region init_data_header;
+	struct amdgpu_virt_region crit_regn;
+	struct amdgpu_virt_region crit_regn_tbl[AMD_SRIOV_MSG_MAX_TABLE_ID];
+	bool is_dynamic_crit_regn_enabled;
+
 	/* vf2pf message */
 	struct delayed_work vf2pf_work;
 	uint32_t vf2pf_update_interval_ms;
+	int vf2pf_update_retry_cnt;
 
 	/* multimedia bandwidth config */
 	bool     is_mm_bw_enabled;
@@ -263,6 +321,19 @@ struct amdgpu_virt {
 
 	/* the ucode id to signal the autoload */
 	uint32_t autoload_ucode_id;
+
+	/* Spinlock to protect access to the RLCG register interface */
+	spinlock_t rlcg_reg_lock;
+
+	struct mutex access_req_mutex;
+
+	union amd_sriov_ras_caps ras_en_caps;
+	union amd_sriov_ras_caps ras_telemetry_en_caps;
+	struct amdgpu_virt_ras ras;
+	struct amd_sriov_ras_telemetry_error_count count_cache;
+
+	/* hibernate and resume with different VF feature for xgmi enabled system */
+	bool is_xgmi_node_migrate_enabled;
 };
 
 struct amdgpu_video_codec_info;
@@ -298,15 +369,38 @@ struct amdgpu_video_codec_info;
 (amdgpu_sriov_vf((adev)) && \
 	((adev)->virt.reg_access & (AMDGIM_FEATURE_GC_REG_RLC_EN)))
 
+#define amdgpu_sriov_reg_indirect_l1_tlb_cntl(adev) \
+(amdgpu_sriov_vf((adev)) && \
+	((adev)->virt.reg_access & (AMDGIM_FEATURE_L1_TLB_CNTL_PSP_EN)))
+
 #define amdgpu_sriov_rlcg_error_report_enabled(adev) \
         (amdgpu_sriov_reg_indirect_mmhub(adev) || amdgpu_sriov_reg_indirect_gc(adev))
 
+#define amdgpu_sriov_reg_access_sq_config(adev) \
+(amdgpu_sriov_vf((adev)) && \
+	((adev)->virt.reg_access & (AMDGIM_FEATURE_REG_ACCESS_SQ_CONFIG)))
+
 #define amdgpu_passthrough(adev) \
 ((adev)->virt.caps & AMDGPU_PASSTHROUGH_MODE)
 
 #define amdgpu_sriov_vf_mmio_access_protection(adev) \
 ((adev)->virt.caps & AMDGPU_VF_MMIO_ACCESS_PROTECT)
 
+#define amdgpu_sriov_ras_caps_en(adev) \
+((adev)->virt.gim_feature & AMDGIM_FEATURE_RAS_CAPS)
+
+#define amdgpu_sriov_ras_telemetry_en(adev) \
+(((adev)->virt.gim_feature & AMDGIM_FEATURE_RAS_TELEMETRY) && (adev)->virt.fw_reserve.ras_telemetry)
+
+#define amdgpu_sriov_ras_telemetry_block_en(adev, sriov_blk) \
+(amdgpu_sriov_ras_telemetry_en((adev)) && (adev)->virt.ras_telemetry_en_caps.all & BIT(sriov_blk))
+
+#define amdgpu_sriov_ras_cper_en(adev) \
+((adev)->virt.gim_feature & AMDGIM_FEATURE_RAS_CPER)
+
+#define amdgpu_sriov_xgmi_ta_ext_peer_link_en(adev) \
+((adev)->virt.gim_feature & AMDGIM_FEATURE_XGMI_TA_EXT_PEER_LINK)
+
 static inline bool is_virtual_machine(void)
 {
 #if defined(CONFIG_X86)
@@ -320,29 +414,42 @@ static inline bool is_virtual_machine(void)
 
 #define amdgpu_sriov_is_pp_one_vf(adev) \
 	((adev)->virt.gim_feature & AMDGIM_FEATURE_PP_ONE_VF)
+#define amdgpu_sriov_multi_vf_mode(adev) \
+	(amdgpu_sriov_vf(adev) && !amdgpu_sriov_is_pp_one_vf(adev))
 #define amdgpu_sriov_is_debug(adev) \
 	((!amdgpu_in_reset(adev)) && adev->virt.tdr_debug)
 #define amdgpu_sriov_is_normal(adev) \
 	((!amdgpu_in_reset(adev)) && (!adev->virt.tdr_debug))
 #define amdgpu_sriov_is_av1_support(adev) \
 	((adev)->virt.gim_feature & AMDGIM_FEATURE_AV1_SUPPORT)
+#define amdgpu_sriov_is_vcn_rb_decouple(adev) \
+	((adev)->virt.gim_feature & AMDGIM_FEATURE_VCN_RB_DECOUPLE)
+#define amdgpu_sriov_is_mes_info_enable(adev) \
+	((adev)->virt.gim_feature & AMDGIM_FEATURE_MES_INFO_ENABLE)
+
+#define amdgpu_virt_xgmi_migrate_enabled(adev) \
+	((adev)->virt.is_xgmi_node_migrate_enabled && (adev)->gmc.xgmi.node_segment_size != 0)
+
 bool amdgpu_virt_mmio_blocked(struct amdgpu_device *adev);
 void amdgpu_virt_init_setting(struct amdgpu_device *adev);
-void amdgpu_virt_kiq_reg_write_reg_wait(struct amdgpu_device *adev,
-					uint32_t reg0, uint32_t rreg1,
-					uint32_t ref, uint32_t mask);
 int amdgpu_virt_request_full_gpu(struct amdgpu_device *adev, bool init);
 int amdgpu_virt_release_full_gpu(struct amdgpu_device *adev, bool init);
 int amdgpu_virt_reset_gpu(struct amdgpu_device *adev);
 void amdgpu_virt_request_init_data(struct amdgpu_device *adev);
+void amdgpu_virt_ready_to_reset(struct amdgpu_device *adev);
 int amdgpu_virt_wait_reset(struct amdgpu_device *adev);
 int amdgpu_virt_alloc_mm_table(struct amdgpu_device *adev);
 void amdgpu_virt_free_mm_table(struct amdgpu_device *adev);
+bool amdgpu_virt_rcvd_ras_interrupt(struct amdgpu_device *adev);
 void amdgpu_virt_release_ras_err_handler_data(struct amdgpu_device *adev);
 void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev);
 void amdgpu_virt_exchange_data(struct amdgpu_device *adev);
 void amdgpu_virt_fini_data_exchange(struct amdgpu_device *adev);
-void amdgpu_detect_virtualization(struct amdgpu_device *adev);
+void amdgpu_virt_init(struct amdgpu_device *adev);
+
+int amdgpu_virt_init_critical_region(struct amdgpu_device *adev);
+int amdgpu_virt_get_dynamic_data_info(struct amdgpu_device *adev,
+	int data_id, uint8_t *binary, u32 *size);
 
 bool amdgpu_virt_can_access_debugfs(struct amdgpu_device *adev);
 int amdgpu_virt_enable_access_debugfs(struct amdgpu_device *adev);
@@ -355,9 +462,25 @@ void amdgpu_virt_update_sriov_video_codec(struct amdgpu_device *adev,
 			struct amdgpu_video_codec_info *decode, uint32_t decode_array_size);
 void amdgpu_sriov_wreg(struct amdgpu_device *adev,
 		       u32 offset, u32 value,
-		       u32 acc_flags, u32 hwip);
+		       u32 acc_flags, u32 hwip, u32 xcc_id);
 u32 amdgpu_sriov_rreg(struct amdgpu_device *adev,
-		      u32 offset, u32 acc_flags, u32 hwip);
+		      u32 offset, u32 acc_flags, u32 hwip, u32 xcc_id);
 bool amdgpu_virt_fw_load_skip_check(struct amdgpu_device *adev,
 			uint32_t ucode_id);
+void amdgpu_virt_pre_reset(struct amdgpu_device *adev);
+void amdgpu_virt_post_reset(struct amdgpu_device *adev);
+bool amdgpu_sriov_xnack_support(struct amdgpu_device *adev);
+bool amdgpu_virt_get_rlcg_reg_access_flag(struct amdgpu_device *adev,
+					  u32 acc_flags, u32 hwip,
+					  bool write, u32 *rlcg_flag);
+u32 amdgpu_virt_rlcg_reg_rw(struct amdgpu_device *adev, u32 offset, u32 v, u32 flag, u32 xcc_id);
+bool amdgpu_virt_get_ras_capability(struct amdgpu_device *adev);
+int amdgpu_virt_req_ras_err_count(struct amdgpu_device *adev, enum amdgpu_ras_block block,
+				  struct ras_err_data *err_data);
+int amdgpu_virt_req_ras_cper_dump(struct amdgpu_device *adev, bool force_update);
+int amdgpu_virt_ras_telemetry_post_reset(struct amdgpu_device *adev);
+bool amdgpu_virt_ras_telemetry_block_en(struct amdgpu_device *adev,
+					enum amdgpu_ras_block block);
+void amdgpu_virt_request_bad_pages(struct amdgpu_device *adev);
+int amdgpu_virt_check_vf_critical_region(struct amdgpu_device *adev, u64 addr, bool *hit);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c
index 53ff91fc6cf6..79bad9cbe2ab 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c
@@ -1,7 +1,9 @@
 // SPDX-License-Identifier: GPL-2.0+
 
 #include <drm/drm_atomic_helper.h>
+#include <drm/drm_edid.h>
 #include <drm/drm_simple_kms_helper.h>
+#include <drm/drm_gem_framebuffer_helper.h>
 #include <drm/drm_vblank.h>
 
 #include "amdgpu.h"
@@ -12,7 +14,6 @@
 #include "dce_v8_0.h"
 #endif
 #include "dce_v10_0.h"
-#include "dce_v11_0.h"
 #include "ivsrcid/ivsrcid_vislands30.h"
 #include "amdgpu_vkms.h"
 #include "amdgpu_display.h"
@@ -55,17 +56,16 @@ static enum hrtimer_restart amdgpu_vkms_vblank_simulate(struct hrtimer *timer)
 		DRM_WARN("%s: vblank timer overrun\n", __func__);
 
 	ret = drm_crtc_handle_vblank(crtc);
+	/* Don't queue timer again when vblank is disabled. */
 	if (!ret)
-		DRM_ERROR("amdgpu_vkms failure on handling vblank");
+		return HRTIMER_NORESTART;
 
 	return HRTIMER_RESTART;
 }
 
 static int amdgpu_vkms_enable_vblank(struct drm_crtc *crtc)
 {
-	struct drm_device *dev = crtc->dev;
-	unsigned int pipe = drm_crtc_index(crtc);
-	struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
+	struct drm_vblank_crtc *vblank = drm_crtc_vblank_crtc(crtc);
 	struct amdgpu_vkms_output *out = drm_crtc_to_amdgpu_vkms_output(crtc);
 	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
 
@@ -81,7 +81,7 @@ static void amdgpu_vkms_disable_vblank(struct drm_crtc *crtc)
 {
 	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
 
-	hrtimer_cancel(&amdgpu_crtc->vblank_timer);
+	hrtimer_try_to_cancel(&amdgpu_crtc->vblank_timer);
 }
 
 static bool amdgpu_vkms_get_vblank_timestamp(struct drm_crtc *crtc,
@@ -89,10 +89,8 @@ static bool amdgpu_vkms_get_vblank_timestamp(struct drm_crtc *crtc,
 					     ktime_t *vblank_time,
 					     bool in_vblank_irq)
 {
-	struct drm_device *dev = crtc->dev;
-	unsigned int pipe = crtc->index;
 	struct amdgpu_vkms_output *output = drm_crtc_to_amdgpu_vkms_output(crtc);
-	struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
+	struct drm_vblank_crtc *vblank = drm_crtc_vblank_crtc(crtc);
 	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
 
 	if (!READ_ONCE(vblank->enabled)) {
@@ -189,8 +187,8 @@ static int amdgpu_vkms_crtc_init(struct drm_device *dev, struct drm_crtc *crtc,
 	amdgpu_crtc->connector = NULL;
 	amdgpu_crtc->vsync_timer_enabled = AMDGPU_IRQ_STATE_DISABLE;
 
-	hrtimer_init(&amdgpu_crtc->vblank_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
-	amdgpu_crtc->vblank_timer.function = &amdgpu_vkms_vblank_simulate;
+	hrtimer_setup(&amdgpu_crtc->vblank_timer, &amdgpu_vkms_vblank_simulate, CLOCK_MONOTONIC,
+		      HRTIMER_MODE_REL);
 
 	return ret;
 }
@@ -238,6 +236,8 @@ static int amdgpu_vkms_conn_get_modes(struct drm_connector *connector)
 
 	for (i = 0; i < ARRAY_SIZE(common_modes); i++) {
 		mode = drm_cvt_mode(dev, common_modes[i].w, common_modes[i].h, 60, false, false, false);
+		if (!mode)
+			continue;
 		drm_mode_probed_add(connector, mode);
 	}
 
@@ -310,7 +310,13 @@ static int amdgpu_vkms_prepare_fb(struct drm_plane *plane,
 		return 0;
 	}
 	afb = to_amdgpu_framebuffer(new_state->fb);
-	obj = new_state->fb->obj[0];
+
+	obj = drm_gem_fb_get_obj(new_state->fb, 0);
+	if (!obj) {
+		DRM_ERROR("Failed to get obj from framebuffer\n");
+		return -EINVAL;
+	}
+
 	rbo = gem_to_amdgpu_bo(obj);
 	adev = amdgpu_ttm_adev(rbo->tbo.bdev);
 
@@ -331,6 +337,7 @@ static int amdgpu_vkms_prepare_fb(struct drm_plane *plane,
 	else
 		domain = AMDGPU_GEM_DOMAIN_VRAM;
 
+	rbo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 	r = amdgpu_bo_pin(rbo, domain);
 	if (unlikely(r != 0)) {
 		if (r != -ERESTARTSYS)
@@ -364,12 +371,19 @@ static void amdgpu_vkms_cleanup_fb(struct drm_plane *plane,
 				   struct drm_plane_state *old_state)
 {
 	struct amdgpu_bo *rbo;
+	struct drm_gem_object *obj;
 	int r;
 
 	if (!old_state->fb)
 		return;
 
-	rbo = gem_to_amdgpu_bo(old_state->fb->obj[0]);
+	obj = drm_gem_fb_get_obj(old_state->fb, 0);
+	if (!obj) {
+		DRM_ERROR("Failed to get obj from framebuffer\n");
+		return;
+	}
+
+	rbo = gem_to_amdgpu_bo(obj);
 	r = amdgpu_bo_reserve(rbo, false);
 	if (unlikely(r)) {
 		DRM_ERROR("failed to reserve rbo before unpin\n");
@@ -478,10 +492,10 @@ const struct drm_mode_config_funcs amdgpu_vkms_mode_funcs = {
 	.atomic_commit = drm_atomic_helper_commit,
 };
 
-static int amdgpu_vkms_sw_init(void *handle)
+static int amdgpu_vkms_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->amdgpu_vkms_output = kcalloc(adev->mode_info.num_crtc,
 		sizeof(struct amdgpu_vkms_output), GFP_KERNEL);
@@ -500,8 +514,6 @@ static int amdgpu_vkms_sw_init(void *handle)
 
 	adev_to_drm(adev)->mode_config.fb_modifiers_not_supported = true;
 
-	adev_to_drm(adev)->mode_config.fb_modifiers_not_supported = true;
-
 	r = amdgpu_display_modeset_create_props(adev);
 	if (r)
 		return r;
@@ -523,9 +535,9 @@ static int amdgpu_vkms_sw_init(void *handle)
 	return 0;
 }
 
-static int amdgpu_vkms_sw_fini(void *handle)
+static int amdgpu_vkms_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i = 0;
 
 	for (i = 0; i < adev->mode_info.num_crtc; i++)
@@ -537,14 +549,14 @@ static int amdgpu_vkms_sw_fini(void *handle)
 
 	adev->mode_info.mode_config_initialized = false;
 
-	kfree(adev->mode_info.bios_hardcoded_edid);
+	drm_edid_free(adev->mode_info.bios_hardcoded_edid);
 	kfree(adev->amdgpu_vkms_output);
 	return 0;
 }
 
-static int amdgpu_vkms_hw_init(void *handle)
+static int amdgpu_vkms_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	switch (adev->asic_type) {
 #ifdef CONFIG_DRM_AMDGPU_SI
@@ -568,13 +580,6 @@ static int amdgpu_vkms_hw_init(void *handle)
 	case CHIP_TONGA:
 		dce_v10_0_disable_dce(adev);
 		break;
-	case CHIP_CARRIZO:
-	case CHIP_STONEY:
-	case CHIP_POLARIS10:
-	case CHIP_POLARIS11:
-	case CHIP_VEGAM:
-		dce_v11_0_disable_dce(adev);
-		break;
 	case CHIP_TOPAZ:
 #ifdef CONFIG_DRM_AMDGPU_SI
 	case CHIP_HAINAN:
@@ -587,55 +592,45 @@ static int amdgpu_vkms_hw_init(void *handle)
 	return 0;
 }
 
-static int amdgpu_vkms_hw_fini(void *handle)
+static int amdgpu_vkms_hw_fini(struct amdgpu_ip_block *ip_block)
 {
 	return 0;
 }
 
-static int amdgpu_vkms_suspend(void *handle)
+static int amdgpu_vkms_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = drm_mode_config_helper_suspend(adev_to_drm(adev));
 	if (r)
 		return r;
-	return amdgpu_vkms_hw_fini(handle);
+
+	return 0;
 }
 
-static int amdgpu_vkms_resume(void *handle)
+static int amdgpu_vkms_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	int r;
 
-	r = amdgpu_vkms_hw_init(handle);
+	r = amdgpu_vkms_hw_init(ip_block);
 	if (r)
 		return r;
-	return drm_mode_config_helper_resume(adev_to_drm(adev));
+	return drm_mode_config_helper_resume(adev_to_drm(ip_block->adev));
 }
 
-static bool amdgpu_vkms_is_idle(void *handle)
+static bool amdgpu_vkms_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	return true;
 }
 
-static int amdgpu_vkms_wait_for_idle(void *handle)
-{
-	return 0;
-}
-
-static int amdgpu_vkms_soft_reset(void *handle)
-{
-	return 0;
-}
-
-static int amdgpu_vkms_set_clockgating_state(void *handle,
+static int amdgpu_vkms_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int amdgpu_vkms_set_powergating_state(void *handle,
+static int amdgpu_vkms_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -643,8 +638,6 @@ static int amdgpu_vkms_set_powergating_state(void *handle,
 
 static const struct amd_ip_funcs amdgpu_vkms_ip_funcs = {
 	.name = "amdgpu_vkms",
-	.early_init = NULL,
-	.late_init = NULL,
 	.sw_init = amdgpu_vkms_sw_init,
 	.sw_fini = amdgpu_vkms_sw_fini,
 	.hw_init = amdgpu_vkms_hw_init,
@@ -652,14 +645,11 @@ static const struct amd_ip_funcs amdgpu_vkms_ip_funcs = {
 	.suspend = amdgpu_vkms_suspend,
 	.resume = amdgpu_vkms_resume,
 	.is_idle = amdgpu_vkms_is_idle,
-	.wait_for_idle = amdgpu_vkms_wait_for_idle,
-	.soft_reset = amdgpu_vkms_soft_reset,
 	.set_clockgating_state = amdgpu_vkms_set_clockgating_state,
 	.set_powergating_state = amdgpu_vkms_set_powergating_state,
 };
 
-const struct amdgpu_ip_block_version amdgpu_vkms_ip_block =
-{
+const struct amdgpu_ip_block_version amdgpu_vkms_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_DCE,
 	.major = 1,
 	.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index 3c0310576b3b..a67285118c37 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -34,7 +34,9 @@
 #include <drm/amdgpu_drm.h>
 #include <drm/drm_drv.h>
 #include <drm/ttm/ttm_tt.h>
+#include <drm/drm_exec.h>
 #include "amdgpu.h"
+#include "amdgpu_vm.h"
 #include "amdgpu_trace.h"
 #include "amdgpu_amdkfd.h"
 #include "amdgpu_gmc.h"
@@ -111,9 +113,9 @@ struct amdgpu_prt_cb {
 };
 
 /**
- * struct amdgpu_vm_tlb_seq_cb - Helper to increment the TLB flush sequence
+ * struct amdgpu_vm_tlb_seq_struct - Helper to increment the TLB flush sequence
  */
-struct amdgpu_vm_tlb_seq_cb {
+struct amdgpu_vm_tlb_seq_struct {
 	/**
 	 * @vm: pointer to the amdgpu_vm structure to set the fence sequence on
 	 */
@@ -126,43 +128,14 @@ struct amdgpu_vm_tlb_seq_cb {
 };
 
 /**
- * amdgpu_vm_set_pasid - manage pasid and vm ptr mapping
- *
- * @adev: amdgpu_device pointer
- * @vm: amdgpu_vm pointer
- * @pasid: the pasid the VM is using on this GPU
- *
- * Set the pasid this VM is using on this GPU, can also be used to remove the
- * pasid by passing in zero.
+ * amdgpu_vm_assert_locked - check if VM is correctly locked
+ * @vm: the VM which schould be tested
  *
+ * Asserts that the VM root PD is locked.
  */
-int amdgpu_vm_set_pasid(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-			u32 pasid)
+static void amdgpu_vm_assert_locked(struct amdgpu_vm *vm)
 {
-	int r;
-
-	if (vm->pasid == pasid)
-		return 0;
-
-	if (vm->pasid) {
-		r = xa_err(xa_erase_irq(&adev->vm_manager.pasids, vm->pasid));
-		if (r < 0)
-			return r;
-
-		vm->pasid = 0;
-	}
-
-	if (pasid) {
-		r = xa_err(xa_store_irq(&adev->vm_manager.pasids, pasid, vm,
-					GFP_KERNEL));
-		if (r < 0)
-			return r;
-
-		vm->pasid = pasid;
-	}
-
-
-	return 0;
+	dma_resv_assert_held(vm->root.bo->tbo.base.resv);
 }
 
 /**
@@ -179,6 +152,7 @@ static void amdgpu_vm_bo_evicted(struct amdgpu_vm_bo_base *vm_bo)
 	struct amdgpu_bo *bo = vm_bo->bo;
 
 	vm_bo->moved = true;
+	amdgpu_vm_assert_locked(vm);
 	spin_lock(&vm_bo->vm->status_lock);
 	if (bo->tbo.type == ttm_bo_type_kernel)
 		list_move(&vm_bo->vm_status, &vm->evicted);
@@ -196,6 +170,7 @@ static void amdgpu_vm_bo_evicted(struct amdgpu_vm_bo_base *vm_bo)
  */
 static void amdgpu_vm_bo_moved(struct amdgpu_vm_bo_base *vm_bo)
 {
+	amdgpu_vm_assert_locked(vm_bo->vm);
 	spin_lock(&vm_bo->vm->status_lock);
 	list_move(&vm_bo->vm_status, &vm_bo->vm->moved);
 	spin_unlock(&vm_bo->vm->status_lock);
@@ -211,6 +186,7 @@ static void amdgpu_vm_bo_moved(struct amdgpu_vm_bo_base *vm_bo)
  */
 static void amdgpu_vm_bo_idle(struct amdgpu_vm_bo_base *vm_bo)
 {
+	amdgpu_vm_assert_locked(vm_bo->vm);
 	spin_lock(&vm_bo->vm->status_lock);
 	list_move(&vm_bo->vm_status, &vm_bo->vm->idle);
 	spin_unlock(&vm_bo->vm->status_lock);
@@ -233,6 +209,22 @@ static void amdgpu_vm_bo_invalidated(struct amdgpu_vm_bo_base *vm_bo)
 }
 
 /**
+ * amdgpu_vm_bo_evicted_user - vm_bo is evicted
+ *
+ * @vm_bo: vm_bo which is evicted
+ *
+ * State for BOs used by user mode queues which are not at the location they
+ * should be.
+ */
+static void amdgpu_vm_bo_evicted_user(struct amdgpu_vm_bo_base *vm_bo)
+{
+	vm_bo->moved = true;
+	spin_lock(&vm_bo->vm->status_lock);
+	list_move(&vm_bo->vm_status, &vm_bo->vm->evicted_user);
+	spin_unlock(&vm_bo->vm->status_lock);
+}
+
+/**
  * amdgpu_vm_bo_relocated - vm_bo is reloacted
  *
  * @vm_bo: vm_bo which is relocated
@@ -242,6 +234,7 @@ static void amdgpu_vm_bo_invalidated(struct amdgpu_vm_bo_base *vm_bo)
  */
 static void amdgpu_vm_bo_relocated(struct amdgpu_vm_bo_base *vm_bo)
 {
+	amdgpu_vm_assert_locked(vm_bo->vm);
 	if (vm_bo->bo->parent) {
 		spin_lock(&vm_bo->vm->status_lock);
 		list_move(&vm_bo->vm_status, &vm_bo->vm->relocated);
@@ -261,12 +254,149 @@ static void amdgpu_vm_bo_relocated(struct amdgpu_vm_bo_base *vm_bo)
  */
 static void amdgpu_vm_bo_done(struct amdgpu_vm_bo_base *vm_bo)
 {
+	amdgpu_vm_assert_locked(vm_bo->vm);
 	spin_lock(&vm_bo->vm->status_lock);
 	list_move(&vm_bo->vm_status, &vm_bo->vm->done);
 	spin_unlock(&vm_bo->vm->status_lock);
 }
 
 /**
+ * amdgpu_vm_bo_reset_state_machine - reset the vm_bo state machine
+ * @vm: the VM which state machine to reset
+ *
+ * Move all vm_bo object in the VM into a state where they will be updated
+ * again during validation.
+ */
+static void amdgpu_vm_bo_reset_state_machine(struct amdgpu_vm *vm)
+{
+	struct amdgpu_vm_bo_base *vm_bo, *tmp;
+
+	amdgpu_vm_assert_locked(vm);
+
+	spin_lock(&vm->status_lock);
+	list_splice_init(&vm->done, &vm->invalidated);
+	list_for_each_entry(vm_bo, &vm->invalidated, vm_status)
+		vm_bo->moved = true;
+
+	list_for_each_entry_safe(vm_bo, tmp, &vm->idle, vm_status) {
+		struct amdgpu_bo *bo = vm_bo->bo;
+
+		vm_bo->moved = true;
+		if (!bo || bo->tbo.type != ttm_bo_type_kernel)
+			list_move(&vm_bo->vm_status, &vm_bo->vm->moved);
+		else if (bo->parent)
+			list_move(&vm_bo->vm_status, &vm_bo->vm->relocated);
+	}
+	spin_unlock(&vm->status_lock);
+}
+
+/**
+ * amdgpu_vm_update_shared - helper to update shared memory stat
+ * @base: base structure for tracking BO usage in a VM
+ *
+ * Takes the vm status_lock and updates the shared memory stat. If the basic
+ * stat changed (e.g. buffer was moved) amdgpu_vm_update_stats need to be called
+ * as well.
+ */
+static void amdgpu_vm_update_shared(struct amdgpu_vm_bo_base *base)
+{
+	struct amdgpu_vm *vm = base->vm;
+	struct amdgpu_bo *bo = base->bo;
+	uint64_t size = amdgpu_bo_size(bo);
+	uint32_t bo_memtype = amdgpu_bo_mem_stats_placement(bo);
+	bool shared;
+
+	dma_resv_assert_held(bo->tbo.base.resv);
+	spin_lock(&vm->status_lock);
+	shared = drm_gem_object_is_shared_for_memory_stats(&bo->tbo.base);
+	if (base->shared != shared) {
+		base->shared = shared;
+		if (shared) {
+			vm->stats[bo_memtype].drm.shared += size;
+			vm->stats[bo_memtype].drm.private -= size;
+		} else {
+			vm->stats[bo_memtype].drm.shared -= size;
+			vm->stats[bo_memtype].drm.private += size;
+		}
+	}
+	spin_unlock(&vm->status_lock);
+}
+
+/**
+ * amdgpu_vm_bo_update_shared - callback when bo gets shared/unshared
+ * @bo: amdgpu buffer object
+ *
+ * Update the per VM stats for all the vm if needed from private to shared or
+ * vice versa.
+ */
+void amdgpu_vm_bo_update_shared(struct amdgpu_bo *bo)
+{
+	struct amdgpu_vm_bo_base *base;
+
+	for (base = bo->vm_bo; base; base = base->next)
+		amdgpu_vm_update_shared(base);
+}
+
+/**
+ * amdgpu_vm_update_stats_locked - helper to update normal memory stat
+ * @base: base structure for tracking BO usage in a VM
+ * @res:  the ttm_resource to use for the purpose of accounting, may or may not
+ *        be bo->tbo.resource
+ * @sign: if we should add (+1) or subtract (-1) from the stat
+ *
+ * Caller need to have the vm status_lock held. Useful for when multiple update
+ * need to happen at the same time.
+ */
+static void amdgpu_vm_update_stats_locked(struct amdgpu_vm_bo_base *base,
+			    struct ttm_resource *res, int sign)
+{
+	struct amdgpu_vm *vm = base->vm;
+	struct amdgpu_bo *bo = base->bo;
+	int64_t size = sign * amdgpu_bo_size(bo);
+	uint32_t bo_memtype = amdgpu_bo_mem_stats_placement(bo);
+
+	/* For drm-total- and drm-shared-, BO are accounted by their preferred
+	 * placement, see also amdgpu_bo_mem_stats_placement.
+	 */
+	if (base->shared)
+		vm->stats[bo_memtype].drm.shared += size;
+	else
+		vm->stats[bo_memtype].drm.private += size;
+
+	if (res && res->mem_type < __AMDGPU_PL_NUM) {
+		uint32_t res_memtype = res->mem_type;
+
+		vm->stats[res_memtype].drm.resident += size;
+		/* BO only count as purgeable if it is resident,
+		 * since otherwise there's nothing to purge.
+		 */
+		if (bo->flags & AMDGPU_GEM_CREATE_DISCARDABLE)
+			vm->stats[res_memtype].drm.purgeable += size;
+		if (!(bo->preferred_domains & amdgpu_mem_type_to_domain(res_memtype)))
+			vm->stats[bo_memtype].evicted += size;
+	}
+}
+
+/**
+ * amdgpu_vm_update_stats - helper to update normal memory stat
+ * @base: base structure for tracking BO usage in a VM
+ * @res:  the ttm_resource to use for the purpose of accounting, may or may not
+ *        be bo->tbo.resource
+ * @sign: if we should add (+1) or subtract (-1) from the stat
+ *
+ * Updates the basic memory stat when bo is added/deleted/moved.
+ */
+void amdgpu_vm_update_stats(struct amdgpu_vm_bo_base *base,
+			    struct ttm_resource *res, int sign)
+{
+	struct amdgpu_vm *vm = base->vm;
+
+	spin_lock(&vm->status_lock);
+	amdgpu_vm_update_stats_locked(base, res, sign);
+	spin_unlock(&vm->status_lock);
+}
+
+/**
  * amdgpu_vm_bo_base_init - Adds bo to the list of bos associated with the vm
  *
  * @base: base structure for tracking BO usage in a VM
@@ -289,7 +419,12 @@ void amdgpu_vm_bo_base_init(struct amdgpu_vm_bo_base *base,
 	base->next = bo->vm_bo;
 	bo->vm_bo = base;
 
-	if (bo->tbo.base.resv != vm->root.bo->tbo.base.resv)
+	spin_lock(&vm->status_lock);
+	base->shared = drm_gem_object_is_shared_for_memory_stats(&bo->tbo.base);
+	amdgpu_vm_update_stats_locked(base, bo->tbo.resource, +1);
+	spin_unlock(&vm->status_lock);
+
+	if (!amdgpu_vm_is_bo_always_valid(vm, bo))
 		return;
 
 	dma_resv_assert_held(vm->root.bo->tbo.base.resv);
@@ -313,25 +448,60 @@ void amdgpu_vm_bo_base_init(struct amdgpu_vm_bo_base *base,
 }
 
 /**
- * amdgpu_vm_get_pd_bo - add the VM PD to a validation list
+ * amdgpu_vm_lock_pd - lock PD in drm_exec
  *
  * @vm: vm providing the BOs
- * @validated: head of validation list
- * @entry: entry to add
+ * @exec: drm execution context
+ * @num_fences: number of extra fences to reserve
  *
- * Add the page directory to the list of BOs to
- * validate for command submission.
+ * Lock the VM root PD in the DRM execution context.
  */
-void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm,
-			 struct list_head *validated,
-			 struct amdgpu_bo_list_entry *entry)
+int amdgpu_vm_lock_pd(struct amdgpu_vm *vm, struct drm_exec *exec,
+		      unsigned int num_fences)
 {
-	entry->priority = 0;
-	entry->tv.bo = &vm->root.bo->tbo;
-	/* Two for VM updates, one for TTM and one for the CS job */
-	entry->tv.num_shared = 4;
-	entry->user_pages = NULL;
-	list_add(&entry->tv.head, validated);
+	/* We need at least two fences for the VM PD/PT updates */
+	return drm_exec_prepare_obj(exec, &vm->root.bo->tbo.base,
+				    2 + num_fences);
+}
+
+/**
+ * amdgpu_vm_lock_done_list - lock all BOs on the done list
+ * @vm: vm providing the BOs
+ * @exec: drm execution context
+ * @num_fences: number of extra fences to reserve
+ *
+ * Lock the BOs on the done list in the DRM execution context.
+ */
+int amdgpu_vm_lock_done_list(struct amdgpu_vm *vm, struct drm_exec *exec,
+			     unsigned int num_fences)
+{
+	struct list_head *prev = &vm->done;
+	struct amdgpu_bo_va *bo_va;
+	struct amdgpu_bo *bo;
+	int ret;
+
+	/* We can only trust prev->next while holding the lock */
+	spin_lock(&vm->status_lock);
+	while (!list_is_head(prev->next, &vm->done)) {
+		bo_va = list_entry(prev->next, typeof(*bo_va), base.vm_status);
+
+		bo = bo_va->base.bo;
+		if (bo) {
+			amdgpu_bo_ref(bo);
+			spin_unlock(&vm->status_lock);
+
+			ret = drm_exec_prepare_obj(exec, &bo->tbo.base, 1);
+			amdgpu_bo_unref(&bo);
+			if (unlikely(ret))
+				return ret;
+
+			spin_lock(&vm->status_lock);
+		}
+		prev = prev->next;
+	}
+	spin_unlock(&vm->status_lock);
+
+	return 0;
 }
 
 /**
@@ -351,28 +521,93 @@ void amdgpu_vm_move_to_lru_tail(struct amdgpu_device *adev,
 	spin_unlock(&adev->mman.bdev.lru_lock);
 }
 
+/* Create scheduler entities for page table updates */
+static int amdgpu_vm_init_entities(struct amdgpu_device *adev,
+				   struct amdgpu_vm *vm)
+{
+	int r;
+
+	r = drm_sched_entity_init(&vm->immediate, DRM_SCHED_PRIORITY_NORMAL,
+				  adev->vm_manager.vm_pte_scheds,
+				  adev->vm_manager.vm_pte_num_scheds, NULL);
+	if (r)
+		goto error;
+
+	return drm_sched_entity_init(&vm->delayed, DRM_SCHED_PRIORITY_NORMAL,
+				     adev->vm_manager.vm_pte_scheds,
+				     adev->vm_manager.vm_pte_num_scheds, NULL);
+
+error:
+	drm_sched_entity_destroy(&vm->immediate);
+	return r;
+}
+
+/* Destroy the entities for page table updates again */
+static void amdgpu_vm_fini_entities(struct amdgpu_vm *vm)
+{
+	drm_sched_entity_destroy(&vm->immediate);
+	drm_sched_entity_destroy(&vm->delayed);
+}
+
 /**
- * amdgpu_vm_validate_pt_bos - validate the page table BOs
+ * amdgpu_vm_generation - return the page table re-generation counter
+ * @adev: the amdgpu_device
+ * @vm: optional VM to check, might be NULL
+ *
+ * Returns a page table re-generation token to allow checking if submissions
+ * are still valid to use this VM. The VM parameter might be NULL in which case
+ * just the VRAM lost counter will be used.
+ */
+uint64_t amdgpu_vm_generation(struct amdgpu_device *adev, struct amdgpu_vm *vm)
+{
+	uint64_t result = (u64)atomic_read(&adev->vram_lost_counter) << 32;
+
+	if (!vm)
+		return result;
+
+	result += lower_32_bits(vm->generation);
+	/* Add one if the page tables will be re-generated on next CS */
+	if (drm_sched_entity_error(&vm->delayed))
+		++result;
+
+	return result;
+}
+
+/**
+ * amdgpu_vm_validate - validate evicted BOs tracked in the VM
  *
  * @adev: amdgpu device pointer
  * @vm: vm providing the BOs
+ * @ticket: optional reservation ticket used to reserve the VM
  * @validate: callback to do the validation
  * @param: parameter for the validation callback
  *
- * Validate the page table BOs on command submission if neccessary.
+ * Validate the page table BOs and per-VM BOs on command submission if
+ * necessary. If a ticket is given, also try to validate evicted user queue
+ * BOs. They must already be reserved with the given ticket.
  *
  * Returns:
  * Validation result.
  */
-int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-			      int (*validate)(void *p, struct amdgpu_bo *bo),
-			      void *param)
+int amdgpu_vm_validate(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+		       struct ww_acquire_ctx *ticket,
+		       int (*validate)(void *p, struct amdgpu_bo *bo),
+		       void *param)
 {
+	uint64_t new_vm_generation = amdgpu_vm_generation(adev, vm);
 	struct amdgpu_vm_bo_base *bo_base;
-	struct amdgpu_bo *shadow;
 	struct amdgpu_bo *bo;
 	int r;
 
+	if (vm->generation != new_vm_generation) {
+		vm->generation = new_vm_generation;
+		amdgpu_vm_bo_reset_state_machine(vm);
+		amdgpu_vm_fini_entities(vm);
+		r = amdgpu_vm_init_entities(adev, vm);
+		if (r)
+			return r;
+	}
+
 	spin_lock(&vm->status_lock);
 	while (!list_empty(&vm->evicted)) {
 		bo_base = list_first_entry(&vm->evicted,
@@ -381,16 +616,10 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		spin_unlock(&vm->status_lock);
 
 		bo = bo_base->bo;
-		shadow = amdgpu_bo_shadowed(bo);
 
 		r = validate(param, bo);
 		if (r)
 			return r;
-		if (shadow) {
-			r = validate(param, shadow);
-			if (r)
-				return r;
-		}
 
 		if (bo->tbo.type != ttm_bo_type_kernel) {
 			amdgpu_vm_bo_moved(bo_base);
@@ -400,6 +629,23 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		}
 		spin_lock(&vm->status_lock);
 	}
+	while (ticket && !list_empty(&vm->evicted_user)) {
+		bo_base = list_first_entry(&vm->evicted_user,
+					   struct amdgpu_vm_bo_base,
+					   vm_status);
+		spin_unlock(&vm->status_lock);
+
+		bo = bo_base->bo;
+		dma_resv_assert_held(bo->tbo.base.resv);
+
+		r = validate(param, bo);
+		if (r)
+			return r;
+
+		amdgpu_vm_bo_invalidated(bo_base);
+
+		spin_lock(&vm->status_lock);
+	}
 	spin_unlock(&vm->status_lock);
 
 	amdgpu_vm_eviction_lock(vm);
@@ -417,22 +663,31 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
  * Check if all VM PDs/PTs are ready for updates
  *
  * Returns:
- * True if VM is not evicting.
+ * True if VM is not evicting and all VM entities are not stopped
  */
 bool amdgpu_vm_ready(struct amdgpu_vm *vm)
 {
-	bool empty;
 	bool ret;
 
+	amdgpu_vm_assert_locked(vm);
+
 	amdgpu_vm_eviction_lock(vm);
 	ret = !vm->evicting;
 	amdgpu_vm_eviction_unlock(vm);
 
 	spin_lock(&vm->status_lock);
-	empty = list_empty(&vm->evicted);
+	ret &= list_empty(&vm->evicted);
 	spin_unlock(&vm->status_lock);
 
-	return ret && empty;
+	spin_lock(&vm->immediate.lock);
+	ret &= !vm->immediate.stopped;
+	spin_unlock(&vm->immediate.lock);
+
+	spin_lock(&vm->delayed.lock);
+	ret &= !vm->delayed.stopped;
+	spin_unlock(&vm->delayed.lock);
+
+	return ret;
 }
 
 /**
@@ -517,6 +772,7 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job,
 		    bool need_pipe_sync)
 {
 	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_isolation *isolation = &adev->isolation[ring->xcp_id];
 	unsigned vmhub = ring->vm_hub;
 	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
 	struct amdgpu_vmid *id = &id_mgr->ids[job->vmid];
@@ -524,9 +780,10 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job,
 	bool gds_switch_needed = ring->funcs->emit_gds_switch &&
 		job->gds_switch_needed;
 	bool vm_flush_needed = job->vm_needs_flush;
-	struct dma_fence *fence = NULL;
+	bool cleaner_shader_needed = false;
 	bool pasid_mapping_needed = false;
-	unsigned patch_offset = 0;
+	struct dma_fence *fence = NULL;
+	unsigned int patch;
 	int r;
 
 	if (amdgpu_vmid_had_gpu_reset(adev, id)) {
@@ -548,16 +805,26 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job,
 	pasid_mapping_needed &= adev->gmc.gmc_funcs->emit_pasid_mapping &&
 		ring->funcs->emit_wreg;
 
-	if (!vm_flush_needed && !gds_switch_needed && !need_pipe_sync)
+	cleaner_shader_needed = job->run_cleaner_shader &&
+		adev->gfx.enable_cleaner_shader &&
+		ring->funcs->emit_cleaner_shader && job->base.s_fence &&
+		&job->base.s_fence->scheduled == isolation->spearhead;
+
+	if (!vm_flush_needed && !gds_switch_needed && !need_pipe_sync &&
+	    !cleaner_shader_needed)
 		return 0;
 
 	amdgpu_ring_ib_begin(ring);
 	if (ring->funcs->init_cond_exec)
-		patch_offset = amdgpu_ring_init_cond_exec(ring);
+		patch = amdgpu_ring_init_cond_exec(ring,
+						   ring->cond_exe_gpu_addr);
 
 	if (need_pipe_sync)
 		amdgpu_ring_emit_pipeline_sync(ring);
 
+	if (cleaner_shader_needed)
+		ring->funcs->emit_cleaner_shader(ring);
+
 	if (vm_flush_needed) {
 		trace_amdgpu_vm_flush(ring, job->vmid, job->vm_pd_addr);
 		amdgpu_ring_emit_vm_flush(ring, job->vmid, job->vm_pd_addr);
@@ -567,9 +834,9 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job,
 		amdgpu_gmc_emit_pasid_mapping(ring, job->vmid, job->pasid);
 
 	if (spm_update_needed && adev->gfx.rlc.funcs->update_spm_vmid)
-		adev->gfx.rlc.funcs->update_spm_vmid(adev, job->vmid);
+		adev->gfx.rlc.funcs->update_spm_vmid(adev, ring, job->vmid);
 
-	if (!ring->is_mes_queue && ring->funcs->emit_gds_switch &&
+	if (ring->funcs->emit_gds_switch &&
 	    gds_switch_needed) {
 		amdgpu_ring_emit_gds_switch(ring, job->vmid, job->gds_base,
 					    job->gds_size, job->gws_base,
@@ -577,10 +844,13 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job,
 					    job->oa_size);
 	}
 
-	if (vm_flush_needed || pasid_mapping_needed) {
-		r = amdgpu_fence_emit(ring, &fence, NULL, 0);
+	if (vm_flush_needed || pasid_mapping_needed || cleaner_shader_needed) {
+		r = amdgpu_fence_emit(ring, job->hw_vm_fence, 0);
 		if (r)
 			return r;
+		fence = &job->hw_vm_fence->base;
+		/* get a ref for the job */
+		dma_fence_get(fence);
 	}
 
 	if (vm_flush_needed) {
@@ -599,16 +869,28 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job,
 		id->pasid_mapping = dma_fence_get(fence);
 		mutex_unlock(&id_mgr->lock);
 	}
+
+	/*
+	 * Make sure that all other submissions wait for the cleaner shader to
+	 * finish before we push them to the HW.
+	 */
+	if (cleaner_shader_needed) {
+		trace_amdgpu_cleaner_shader(ring, fence);
+		mutex_lock(&adev->enforce_isolation_mutex);
+		dma_fence_put(isolation->spearhead);
+		isolation->spearhead = dma_fence_get(fence);
+		mutex_unlock(&adev->enforce_isolation_mutex);
+	}
 	dma_fence_put(fence);
 
-	if (ring->funcs->patch_cond_exec)
-		amdgpu_ring_patch_cond_exec(ring, patch_offset);
+	amdgpu_ring_patch_cond_exec(ring, patch);
 
 	/* the double SWITCH_BUFFER here *cannot* be skipped by COND_EXEC */
 	if (ring->funcs->emit_switch_buffer) {
 		amdgpu_ring_emit_switch_buffer(ring);
 		amdgpu_ring_emit_switch_buffer(ring);
 	}
+
 	amdgpu_ring_ib_end(ring);
 	return 0;
 }
@@ -690,6 +972,8 @@ int amdgpu_vm_update_pdes(struct amdgpu_device *adev,
 	LIST_HEAD(relocated);
 	int r, idx;
 
+	amdgpu_vm_assert_locked(vm);
+
 	spin_lock(&vm->status_lock);
 	list_splice_init(&vm->relocated, &relocated);
 	spin_unlock(&vm->status_lock);
@@ -705,7 +989,8 @@ int amdgpu_vm_update_pdes(struct amdgpu_device *adev,
 	params.vm = vm;
 	params.immediate = immediate;
 
-	r = vm->update_funcs->prepare(&params, NULL, AMDGPU_SYNC_EXPLICIT);
+	r = vm->update_funcs->prepare(&params, NULL,
+				      AMDGPU_KERNEL_JOB_ID_VM_UPDATE_PDES);
 	if (r)
 		goto error;
 
@@ -746,7 +1031,7 @@ error:
 static void amdgpu_vm_tlb_seq_cb(struct dma_fence *fence,
 				 struct dma_fence_cb *cb)
 {
-	struct amdgpu_vm_tlb_seq_cb *tlb_cb;
+	struct amdgpu_vm_tlb_seq_struct *tlb_cb;
 
 	tlb_cb = container_of(cb, typeof(*tlb_cb), cb);
 	atomic64_inc(&tlb_cb->vm->tlb_seq);
@@ -754,6 +1039,46 @@ static void amdgpu_vm_tlb_seq_cb(struct dma_fence *fence,
 }
 
 /**
+ * amdgpu_vm_tlb_flush - prepare TLB flush
+ *
+ * @params: parameters for update
+ * @fence: input fence to sync TLB flush with
+ * @tlb_cb: the callback structure
+ *
+ * Increments the tlb sequence to make sure that future CS execute a VM flush.
+ */
+static void
+amdgpu_vm_tlb_flush(struct amdgpu_vm_update_params *params,
+		    struct dma_fence **fence,
+		    struct amdgpu_vm_tlb_seq_struct *tlb_cb)
+{
+	struct amdgpu_vm *vm = params->vm;
+
+	tlb_cb->vm = vm;
+	if (!fence || !*fence) {
+		amdgpu_vm_tlb_seq_cb(NULL, &tlb_cb->cb);
+		return;
+	}
+
+	if (!dma_fence_add_callback(*fence, &tlb_cb->cb,
+				    amdgpu_vm_tlb_seq_cb)) {
+		dma_fence_put(vm->last_tlb_flush);
+		vm->last_tlb_flush = dma_fence_get(*fence);
+	} else {
+		amdgpu_vm_tlb_seq_cb(NULL, &tlb_cb->cb);
+	}
+
+	/* Prepare a TLB flush fence to be attached to PTs */
+	if (!params->unlocked) {
+		amdgpu_vm_tlb_fence_create(params->adev, vm, fence);
+
+		/* Makes sure no PD/PT is freed before the flush */
+		dma_resv_add_fence(vm->root.bo->tbo.base.resv, *fence,
+				   DMA_RESV_USAGE_BOOKKEEP);
+	}
+}
+
+/**
  * amdgpu_vm_update_range - update a range in the vm page table
  *
  * @adev: amdgpu_device pointer to use for commands
@@ -761,7 +1086,8 @@ static void amdgpu_vm_tlb_seq_cb(struct dma_fence *fence,
  * @immediate: immediate submission in a page fault
  * @unlocked: unlocked invalidation during MM callback
  * @flush_tlb: trigger tlb invalidation after update completed
- * @resv: fences we need to sync to
+ * @allow_override: change MTYPE for local NUMA nodes
+ * @sync: fences we need to sync to
  * @start: start of mapped range
  * @last: last mapped entry
  * @flags: flags for the entries
@@ -778,15 +1104,15 @@ static void amdgpu_vm_tlb_seq_cb(struct dma_fence *fence,
  */
 int amdgpu_vm_update_range(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			   bool immediate, bool unlocked, bool flush_tlb,
-			   struct dma_resv *resv, uint64_t start, uint64_t last,
-			   uint64_t flags, uint64_t offset, uint64_t vram_base,
+			   bool allow_override, struct amdgpu_sync *sync,
+			   uint64_t start, uint64_t last, uint64_t flags,
+			   uint64_t offset, uint64_t vram_base,
 			   struct ttm_resource *res, dma_addr_t *pages_addr,
 			   struct dma_fence **fence)
 {
+	struct amdgpu_vm_tlb_seq_struct *tlb_cb;
 	struct amdgpu_vm_update_params params;
-	struct amdgpu_vm_tlb_seq_cb *tlb_cb;
 	struct amdgpu_res_cursor cursor;
-	enum amdgpu_sync_mode sync_mode;
 	int r, idx;
 
 	if (!drm_dev_enter(adev_to_drm(adev), &idx))
@@ -794,20 +1120,20 @@ int amdgpu_vm_update_range(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 
 	tlb_cb = kmalloc(sizeof(*tlb_cb), GFP_KERNEL);
 	if (!tlb_cb) {
-		r = -ENOMEM;
-		goto error_unlock;
+		drm_dev_exit(idx);
+		return -ENOMEM;
 	}
 
 	/* Vega20+XGMI where PTEs get inadvertently cached in L2 texture cache,
 	 * heavy-weight flush TLB unconditionally.
 	 */
 	flush_tlb |= adev->gmc.xgmi.num_physical_nodes &&
-		     adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 0);
+		     amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 0);
 
 	/*
 	 * On GFX8 and older any 8 PTE block with a valid bit set enters the TLB
 	 */
-	flush_tlb |= adev->ip_versions[GC_HWIP][0] < IP_VERSION(9, 0, 0);
+	flush_tlb |= amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(9, 0, 0);
 
 	memset(&params, 0, sizeof(params));
 	params.adev = adev;
@@ -815,14 +1141,9 @@ int amdgpu_vm_update_range(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	params.immediate = immediate;
 	params.pages_addr = pages_addr;
 	params.unlocked = unlocked;
-
-	/* Implicitly sync to command submissions in the same VM before
-	 * unmapping. Sync to moving fences before mapping.
-	 */
-	if (!(flags & AMDGPU_PTE_VALID))
-		sync_mode = AMDGPU_SYNC_EQ_OWNER;
-	else
-		sync_mode = AMDGPU_SYNC_EXPLICIT;
+	params.needs_flush = flush_tlb;
+	params.allow_override = allow_override;
+	INIT_LIST_HEAD(&params.tlb_flush_waitlist);
 
 	amdgpu_vm_eviction_lock(vm);
 	if (vm->evicting) {
@@ -838,7 +1159,8 @@ int amdgpu_vm_update_range(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		dma_fence_put(tmp);
 	}
 
-	r = vm->update_funcs->prepare(&params, resv, sync_mode);
+	r = vm->update_funcs->prepare(&params, sync,
+				      AMDGPU_KERNEL_JOB_ID_VM_UPDATE_RANGE);
 	if (r)
 		goto error_free;
 
@@ -881,7 +1203,7 @@ int amdgpu_vm_update_range(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 				params.pages_addr = NULL;
 			}
 
-		} else if (flags & (AMDGPU_PTE_VALID | AMDGPU_PTE_PRT)) {
+		} else if (flags & (AMDGPU_PTE_VALID | AMDGPU_PTE_PRT_FLAG(adev))) {
 			addr = vram_base + cursor.start;
 		} else {
 			addr = 0;
@@ -897,65 +1219,28 @@ int amdgpu_vm_update_range(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	}
 
 	r = vm->update_funcs->commit(&params, fence);
+	if (r)
+		goto error_free;
 
-	if (flush_tlb || params.table_freed) {
-		tlb_cb->vm = vm;
-		if (fence && *fence &&
-		    !dma_fence_add_callback(*fence, &tlb_cb->cb,
-					   amdgpu_vm_tlb_seq_cb)) {
-			dma_fence_put(vm->last_tlb_flush);
-			vm->last_tlb_flush = dma_fence_get(*fence);
-		} else {
-			amdgpu_vm_tlb_seq_cb(NULL, &tlb_cb->cb);
-		}
+	if (params.needs_flush) {
+		amdgpu_vm_tlb_flush(&params, fence, tlb_cb);
 		tlb_cb = NULL;
 	}
 
+	amdgpu_vm_pt_free_list(adev, &params);
+
 error_free:
 	kfree(tlb_cb);
-
-error_unlock:
 	amdgpu_vm_eviction_unlock(vm);
 	drm_dev_exit(idx);
 	return r;
 }
 
 void amdgpu_vm_get_memory(struct amdgpu_vm *vm,
-			  struct amdgpu_mem_stats *stats)
+			  struct amdgpu_mem_stats stats[__AMDGPU_PL_NUM])
 {
-	struct amdgpu_bo_va *bo_va, *tmp;
-
 	spin_lock(&vm->status_lock);
-	list_for_each_entry_safe(bo_va, tmp, &vm->idle, base.vm_status) {
-		if (!bo_va->base.bo)
-			continue;
-		amdgpu_bo_get_memory(bo_va->base.bo, stats);
-	}
-	list_for_each_entry_safe(bo_va, tmp, &vm->evicted, base.vm_status) {
-		if (!bo_va->base.bo)
-			continue;
-		amdgpu_bo_get_memory(bo_va->base.bo, stats);
-	}
-	list_for_each_entry_safe(bo_va, tmp, &vm->relocated, base.vm_status) {
-		if (!bo_va->base.bo)
-			continue;
-		amdgpu_bo_get_memory(bo_va->base.bo, stats);
-	}
-	list_for_each_entry_safe(bo_va, tmp, &vm->moved, base.vm_status) {
-		if (!bo_va->base.bo)
-			continue;
-		amdgpu_bo_get_memory(bo_va->base.bo, stats);
-	}
-	list_for_each_entry_safe(bo_va, tmp, &vm->invalidated, base.vm_status) {
-		if (!bo_va->base.bo)
-			continue;
-		amdgpu_bo_get_memory(bo_va->base.bo, stats);
-	}
-	list_for_each_entry_safe(bo_va, tmp, &vm->done, base.vm_status) {
-		if (!bo_va->base.bo)
-			continue;
-		amdgpu_bo_get_memory(bo_va->base.bo, stats);
-	}
+	memcpy(stats, vm->stats, sizeof(*stats) * __AMDGPU_PL_NUM);
 	spin_unlock(&vm->status_lock);
 }
 
@@ -977,34 +1262,59 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va,
 	struct amdgpu_bo *bo = bo_va->base.bo;
 	struct amdgpu_vm *vm = bo_va->base.vm;
 	struct amdgpu_bo_va_mapping *mapping;
+	struct dma_fence **last_update;
 	dma_addr_t *pages_addr = NULL;
 	struct ttm_resource *mem;
-	struct dma_fence **last_update;
+	struct amdgpu_sync sync;
 	bool flush_tlb = clear;
-	struct dma_resv *resv;
 	uint64_t vram_base;
 	uint64_t flags;
+	bool uncached;
 	int r;
 
-	if (clear || !bo) {
+	amdgpu_sync_create(&sync);
+	if (clear) {
+		mem = NULL;
+
+		/* Implicitly sync to command submissions in the same VM before
+		 * unmapping.
+		 */
+		r = amdgpu_sync_resv(adev, &sync, vm->root.bo->tbo.base.resv,
+				     AMDGPU_SYNC_EQ_OWNER, vm);
+		if (r)
+			goto error_free;
+		if (bo) {
+			r = amdgpu_sync_kfd(&sync, bo->tbo.base.resv);
+			if (r)
+				goto error_free;
+		}
+	} else if (!bo) {
 		mem = NULL;
-		resv = vm->root.bo->tbo.base.resv;
+
+		/* PRT map operations don't need to sync to anything. */
+
 	} else {
 		struct drm_gem_object *obj = &bo->tbo.base;
 
-		resv = bo->tbo.base.resv;
-		if (obj->import_attach && bo_va->is_xgmi) {
+		if (drm_gem_is_imported(obj) && bo_va->is_xgmi) {
 			struct dma_buf *dma_buf = obj->import_attach->dmabuf;
 			struct drm_gem_object *gobj = dma_buf->priv;
 			struct amdgpu_bo *abo = gem_to_amdgpu_bo(gobj);
 
-			if (abo->tbo.resource->mem_type == TTM_PL_VRAM)
+			if (abo->tbo.resource &&
+			    abo->tbo.resource->mem_type == TTM_PL_VRAM)
 				bo = gem_to_amdgpu_bo(gobj);
 		}
 		mem = bo->tbo.resource;
-		if (mem->mem_type == TTM_PL_TT ||
-		    mem->mem_type == AMDGPU_PL_PREEMPT)
+		if (mem && (mem->mem_type == TTM_PL_TT ||
+			    mem->mem_type == AMDGPU_PL_PREEMPT))
 			pages_addr = bo->tbo.ttm->dma_address;
+
+		/* Implicitly sync to moving fences before mapping anything */
+		r = amdgpu_sync_resv(adev, &sync, bo->tbo.base.resv,
+				     AMDGPU_SYNC_EXPLICIT, vm);
+		if (r)
+			goto error_free;
 	}
 
 	if (bo) {
@@ -1017,13 +1327,14 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va,
 
 		bo_adev = amdgpu_ttm_adev(bo->tbo.bdev);
 		vram_base = bo_adev->vm_manager.vram_base_offset;
+		uncached = (bo->flags & AMDGPU_GEM_CREATE_UNCACHED) != 0;
 	} else {
 		flags = 0x0;
 		vram_base = 0;
+		uncached = false;
 	}
 
-	if (clear || (bo && bo->tbo.base.resv ==
-		      vm->root.bo->tbo.base.resv))
+	if (clear || amdgpu_vm_is_bo_always_valid(vm, bo))
 		last_update = &vm->last_update;
 	else
 		last_update = &bo_va->last_pt_update;
@@ -1042,34 +1353,34 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va,
 		/* normally,bo_va->flags only contians READABLE and WIRTEABLE bit go here
 		 * but in case of something, we filter the flags in first place
 		 */
-		if (!(mapping->flags & AMDGPU_PTE_READABLE))
+		if (!(mapping->flags & AMDGPU_VM_PAGE_READABLE))
 			update_flags &= ~AMDGPU_PTE_READABLE;
-		if (!(mapping->flags & AMDGPU_PTE_WRITEABLE))
+		if (!(mapping->flags & AMDGPU_VM_PAGE_WRITEABLE))
 			update_flags &= ~AMDGPU_PTE_WRITEABLE;
 
 		/* Apply ASIC specific mapping flags */
-		amdgpu_gmc_get_vm_pte(adev, mapping, &update_flags);
+		amdgpu_gmc_get_vm_pte(adev, vm, bo, mapping->flags,
+				      &update_flags);
 
 		trace_amdgpu_vm_bo_update(mapping);
 
 		r = amdgpu_vm_update_range(adev, vm, false, false, flush_tlb,
-					   resv, mapping->start, mapping->last,
-					   update_flags, mapping->offset,
-					   vram_base, mem, pages_addr,
-					   last_update);
+					   !uncached, &sync, mapping->start,
+					   mapping->last, update_flags,
+					   mapping->offset, vram_base, mem,
+					   pages_addr, last_update);
 		if (r)
-			return r;
+			goto error_free;
 	}
 
 	/* If the BO is not in its preferred location add it back to
 	 * the evicted list so that it gets validated again on the
 	 * next command submission.
 	 */
-	if (bo && bo->tbo.base.resv == vm->root.bo->tbo.base.resv) {
-		uint32_t mem_type = bo->tbo.resource->mem_type;
-
-		if (!(bo->preferred_domains &
-		      amdgpu_mem_type_to_domain(mem_type)))
+	if (amdgpu_vm_is_bo_always_valid(vm, bo)) {
+		if (bo->tbo.resource &&
+		    !(bo->preferred_domains &
+		      amdgpu_mem_type_to_domain(bo->tbo.resource->mem_type)))
 			amdgpu_vm_bo_evicted(&bo_va->base);
 		else
 			amdgpu_vm_bo_idle(&bo_va->base);
@@ -1086,7 +1397,9 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va,
 			trace_amdgpu_vm_bo_mapping(mapping);
 	}
 
-	return 0;
+error_free:
+	amdgpu_sync_free(&sync);
+	return r;
 }
 
 /**
@@ -1188,7 +1501,7 @@ static void amdgpu_vm_free_mapping(struct amdgpu_device *adev,
 				   struct amdgpu_bo_va_mapping *mapping,
 				   struct dma_fence *fence)
 {
-	if (mapping->flags & AMDGPU_PTE_PRT)
+	if (mapping->flags & AMDGPU_VM_PAGE_PRT)
 		amdgpu_vm_add_prt_cb(adev, fence);
 	kfree(mapping);
 }
@@ -1233,29 +1546,34 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
 			  struct amdgpu_vm *vm,
 			  struct dma_fence **fence)
 {
-	struct dma_resv *resv = vm->root.bo->tbo.base.resv;
 	struct amdgpu_bo_va_mapping *mapping;
-	uint64_t init_pte_value = 0;
 	struct dma_fence *f = NULL;
+	struct amdgpu_sync sync;
 	int r;
 
+
+	/*
+	 * Implicitly sync to command submissions in the same VM before
+	 * unmapping.
+	 */
+	amdgpu_sync_create(&sync);
+	r = amdgpu_sync_resv(adev, &sync, vm->root.bo->tbo.base.resv,
+			     AMDGPU_SYNC_EQ_OWNER, vm);
+	if (r)
+		goto error_free;
+
 	while (!list_empty(&vm->freed)) {
 		mapping = list_first_entry(&vm->freed,
 			struct amdgpu_bo_va_mapping, list);
 		list_del(&mapping->list);
 
-		if (vm->pte_support_ats &&
-		    mapping->start < AMDGPU_GMC_HOLE_START)
-			init_pte_value = AMDGPU_PTE_DEFAULT_ATC;
-
-		r = amdgpu_vm_update_range(adev, vm, false, false, true, resv,
-					   mapping->start, mapping->last,
-					   init_pte_value, 0, 0, NULL, NULL,
-					   &f);
+		r = amdgpu_vm_update_range(adev, vm, false, false, true, false,
+					   &sync, mapping->start, mapping->last,
+					   0, 0, 0, NULL, NULL, &f);
 		amdgpu_vm_free_mapping(adev, vm, mapping, f);
 		if (r) {
 			dma_fence_put(f);
-			return r;
+			goto error_free;
 		}
 	}
 
@@ -1266,7 +1584,9 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
 		dma_fence_put(f);
 	}
 
-	return 0;
+error_free:
+	amdgpu_sync_free(&sync);
+	return r;
 
 }
 
@@ -1275,6 +1595,7 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
  *
  * @adev: amdgpu_device pointer
  * @vm: requested vm
+ * @ticket: optional reservation ticket used to reserve the VM
  *
  * Make sure all BOs which are moved are updated in the PTs.
  *
@@ -1284,11 +1605,12 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
  * PTs have to be reserved!
  */
 int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
-			   struct amdgpu_vm *vm)
+			   struct amdgpu_vm *vm,
+			   struct ww_acquire_ctx *ticket)
 {
 	struct amdgpu_bo_va *bo_va;
 	struct dma_resv *resv;
-	bool clear;
+	bool clear, unlock;
 	int r;
 
 	spin_lock(&vm->status_lock);
@@ -1311,18 +1633,35 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
 		spin_unlock(&vm->status_lock);
 
 		/* Try to reserve the BO to avoid clearing its ptes */
-		if (!amdgpu_vm_debug && dma_resv_trylock(resv))
+		if (!adev->debug_vm && dma_resv_trylock(resv)) {
+			clear = false;
+			unlock = true;
+		/* The caller is already holding the reservation lock */
+		} else if (ticket && dma_resv_locking_ctx(resv) == ticket) {
 			clear = false;
+			unlock = false;
 		/* Somebody else is using the BO right now */
-		else
+		} else {
 			clear = true;
+			unlock = false;
+		}
 
 		r = amdgpu_vm_bo_update(adev, bo_va, clear);
+
+		if (unlock)
+			dma_resv_unlock(resv);
 		if (r)
 			return r;
 
-		if (!clear)
-			dma_resv_unlock(resv);
+		/* Remember evicted DMABuf imports in compute VMs for later
+		 * validation
+		 */
+		if (vm->is_compute_context &&
+		    drm_gem_is_imported(&bo_va->base.bo->tbo.base) &&
+		    (!bo_va->base.bo->tbo.resource ||
+		     bo_va->base.bo->tbo.resource->mem_type == TTM_PL_SYSTEM))
+			amdgpu_vm_bo_evicted_user(&bo_va->base);
+
 		spin_lock(&vm->status_lock);
 	}
 	spin_unlock(&vm->status_lock);
@@ -1331,6 +1670,51 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
 }
 
 /**
+ * amdgpu_vm_flush_compute_tlb - Flush TLB on compute VM
+ *
+ * @adev: amdgpu_device pointer
+ * @vm: requested vm
+ * @flush_type: flush type
+ * @xcc_mask: mask of XCCs that belong to the compute partition in need of a TLB flush.
+ *
+ * Flush TLB if needed for a compute VM.
+ *
+ * Returns:
+ * 0 for success.
+ */
+int amdgpu_vm_flush_compute_tlb(struct amdgpu_device *adev,
+				struct amdgpu_vm *vm,
+				uint32_t flush_type,
+				uint32_t xcc_mask)
+{
+	uint64_t tlb_seq = amdgpu_vm_tlb_seq(vm);
+	bool all_hub = false;
+	int xcc = 0, r = 0;
+
+	WARN_ON_ONCE(!vm->is_compute_context);
+
+	/*
+	 * It can be that we race and lose here, but that is extremely unlikely
+	 * and the worst thing which could happen is that we flush the changes
+	 * into the TLB once more which is harmless.
+	 */
+	if (atomic64_xchg(&vm->kfd_last_flushed_seq, tlb_seq) == tlb_seq)
+		return 0;
+
+	if (adev->family == AMDGPU_FAMILY_AI ||
+	    adev->family == AMDGPU_FAMILY_RV)
+		all_hub = true;
+
+	for_each_inst(xcc, xcc_mask) {
+		r = amdgpu_gmc_flush_gpu_tlb_pasid(adev, vm->pasid, flush_type,
+						   all_hub, xcc);
+		if (r)
+			break;
+	}
+	return r;
+}
+
+/**
  * amdgpu_vm_bo_add - add a bo to a specific vm
  *
  * @adev: amdgpu_device pointer
@@ -1358,6 +1742,7 @@ struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev,
 	amdgpu_vm_bo_base_init(&bo_va->base, vm, bo);
 
 	bo_va->ref_count = 1;
+	bo_va->last_pt_update = dma_fence_get_stub();
 	INIT_LIST_HEAD(&bo_va->valids);
 	INIT_LIST_HEAD(&bo_va->invalids);
 
@@ -1395,16 +1780,46 @@ static void amdgpu_vm_bo_insert_map(struct amdgpu_device *adev,
 	list_add(&mapping->list, &bo_va->invalids);
 	amdgpu_vm_it_insert(mapping, &vm->va);
 
-	if (mapping->flags & AMDGPU_PTE_PRT)
+	if (mapping->flags & AMDGPU_VM_PAGE_PRT)
 		amdgpu_vm_prt_get(adev);
 
-	if (bo && bo->tbo.base.resv == vm->root.bo->tbo.base.resv &&
-	    !bo_va->base.moved) {
+	if (amdgpu_vm_is_bo_always_valid(vm, bo) && !bo_va->base.moved)
 		amdgpu_vm_bo_moved(&bo_va->base);
-	}
+
 	trace_amdgpu_vm_bo_map(bo_va, mapping);
 }
 
+/* Validate operation parameters to prevent potential abuse */
+static int amdgpu_vm_verify_parameters(struct amdgpu_device *adev,
+					  struct amdgpu_bo *bo,
+					  uint64_t saddr,
+					  uint64_t offset,
+					  uint64_t size)
+{
+	uint64_t tmp, lpfn;
+
+	if (saddr & AMDGPU_GPU_PAGE_MASK
+	    || offset & AMDGPU_GPU_PAGE_MASK
+	    || size & AMDGPU_GPU_PAGE_MASK)
+		return -EINVAL;
+
+	if (check_add_overflow(saddr, size, &tmp)
+	    || check_add_overflow(offset, size, &tmp)
+	    || size == 0 /* which also leads to end < begin */)
+		return -EINVAL;
+
+	/* make sure object fit at this offset */
+	if (bo && offset + size > amdgpu_bo_size(bo))
+		return -EINVAL;
+
+	/* Ensure last pfn not exceed max_pfn */
+	lpfn = (saddr + size - 1) >> AMDGPU_GPU_PAGE_SHIFT;
+	if (lpfn >= adev->vm_manager.max_pfn)
+		return -EINVAL;
+
+	return 0;
+}
+
 /**
  * amdgpu_vm_bo_map - map bo inside a vm
  *
@@ -1425,27 +1840,20 @@ static void amdgpu_vm_bo_insert_map(struct amdgpu_device *adev,
 int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 		     struct amdgpu_bo_va *bo_va,
 		     uint64_t saddr, uint64_t offset,
-		     uint64_t size, uint64_t flags)
+		     uint64_t size, uint32_t flags)
 {
 	struct amdgpu_bo_va_mapping *mapping, *tmp;
 	struct amdgpu_bo *bo = bo_va->base.bo;
 	struct amdgpu_vm *vm = bo_va->base.vm;
 	uint64_t eaddr;
+	int r;
 
-	/* validate the parameters */
-	if (saddr & ~PAGE_MASK || offset & ~PAGE_MASK ||
-	    size == 0 || size & ~PAGE_MASK)
-		return -EINVAL;
-
-	/* make sure object fit at this offset */
-	eaddr = saddr + size - 1;
-	if (saddr >= eaddr ||
-	    (bo && offset + size > amdgpu_bo_size(bo)) ||
-	    (eaddr >= adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT))
-		return -EINVAL;
+	r = amdgpu_vm_verify_parameters(adev, bo, saddr, offset, size);
+	if (r)
+		return r;
 
 	saddr /= AMDGPU_GPU_PAGE_SIZE;
-	eaddr /= AMDGPU_GPU_PAGE_SIZE;
+	eaddr = saddr + (size - 1) / AMDGPU_GPU_PAGE_SIZE;
 
 	tmp = amdgpu_vm_it_iter_first(&vm->va, saddr, eaddr);
 	if (tmp) {
@@ -1491,24 +1899,16 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev,
 			     struct amdgpu_bo_va *bo_va,
 			     uint64_t saddr, uint64_t offset,
-			     uint64_t size, uint64_t flags)
+			     uint64_t size, uint32_t flags)
 {
 	struct amdgpu_bo_va_mapping *mapping;
 	struct amdgpu_bo *bo = bo_va->base.bo;
 	uint64_t eaddr;
 	int r;
 
-	/* validate the parameters */
-	if (saddr & ~PAGE_MASK || offset & ~PAGE_MASK ||
-	    size == 0 || size & ~PAGE_MASK)
-		return -EINVAL;
-
-	/* make sure object fit at this offset */
-	eaddr = saddr + size - 1;
-	if (saddr >= eaddr ||
-	    (bo && offset + size > amdgpu_bo_size(bo)) ||
-	    (eaddr >= adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT))
-		return -EINVAL;
+	r = amdgpu_vm_verify_parameters(adev, bo, saddr, offset, size);
+	if (r)
+		return r;
 
 	/* Allocate all the needed memory */
 	mapping = kmalloc(sizeof(*mapping), GFP_KERNEL);
@@ -1522,7 +1922,7 @@ int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev,
 	}
 
 	saddr /= AMDGPU_GPU_PAGE_SIZE;
-	eaddr /= AMDGPU_GPU_PAGE_SIZE;
+	eaddr = saddr + (size - 1) / AMDGPU_GPU_PAGE_SIZE;
 
 	mapping->start = saddr;
 	mapping->last = eaddr;
@@ -1555,6 +1955,7 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
 	struct amdgpu_bo_va_mapping *mapping;
 	struct amdgpu_vm *vm = bo_va->base.vm;
 	bool valid = true;
+	int r;
 
 	saddr /= AMDGPU_GPU_PAGE_SIZE;
 
@@ -1575,6 +1976,17 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
 			return -ENOENT;
 	}
 
+	/* It's unlikely to happen that the mapping userq hasn't been idled
+	 * during user requests GEM unmap IOCTL except for forcing the unmap
+	 * from user space.
+	 */
+	if (unlikely(atomic_read(&bo_va->userq_va_mapped) > 0)) {
+		r = amdgpu_userq_gem_va_unmap_validate(adev, mapping, saddr);
+		if (unlikely(r == -EBUSY))
+			dev_warn_once(adev->dev,
+				      "Attempt to unmap an active userq buffer\n");
+	}
+
 	list_del(&mapping->list);
 	amdgpu_vm_it_remove(mapping, &vm->va);
 	mapping->bo_va = NULL;
@@ -1609,10 +2021,14 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev,
 	struct amdgpu_bo_va_mapping *before, *after, *tmp, *next;
 	LIST_HEAD(removed);
 	uint64_t eaddr;
+	int r;
+
+	r = amdgpu_vm_verify_parameters(adev, NULL, saddr, 0, size);
+	if (r)
+		return r;
 
-	eaddr = saddr + size - 1;
 	saddr /= AMDGPU_GPU_PAGE_SIZE;
-	eaddr /= AMDGPU_GPU_PAGE_SIZE;
+	eaddr = saddr + (size - 1) / AMDGPU_GPU_PAGE_SIZE;
 
 	/* Allocate all the needed memory */
 	before = kzalloc(sizeof(*before), GFP_KERNEL);
@@ -1674,18 +2090,30 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev,
 
 	/* Insert partial mapping before the range */
 	if (!list_empty(&before->list)) {
+		struct amdgpu_bo *bo = before->bo_va->base.bo;
+
 		amdgpu_vm_it_insert(before, &vm->va);
-		if (before->flags & AMDGPU_PTE_PRT)
+		if (before->flags & AMDGPU_VM_PAGE_PRT)
 			amdgpu_vm_prt_get(adev);
+
+		if (amdgpu_vm_is_bo_always_valid(vm, bo) &&
+		    !before->bo_va->base.moved)
+			amdgpu_vm_bo_moved(&before->bo_va->base);
 	} else {
 		kfree(before);
 	}
 
 	/* Insert partial mapping after the range */
 	if (!list_empty(&after->list)) {
+		struct amdgpu_bo *bo = after->bo_va->base.bo;
+
 		amdgpu_vm_it_insert(after, &vm->va);
-		if (after->flags & AMDGPU_PTE_PRT)
+		if (after->flags & AMDGPU_VM_PAGE_PRT)
 			amdgpu_vm_prt_get(adev);
+
+		if (amdgpu_vm_is_bo_always_valid(vm, bo) &&
+		    !after->bo_va->base.moved)
+			amdgpu_vm_bo_moved(&after->bo_va->base);
 	} else {
 		kfree(after);
 	}
@@ -1763,7 +2191,7 @@ void amdgpu_vm_bo_del(struct amdgpu_device *adev,
 
 	if (bo) {
 		dma_resv_assert_held(bo->tbo.base.resv);
-		if (bo->tbo.base.resv == vm->root.bo->tbo.base.resv)
+		if (amdgpu_vm_is_bo_always_valid(vm, bo))
 			ttm_bo_set_bulk_move(&bo->tbo, NULL);
 
 		for (base = &bo_va->base.bo->vm_bo; *base;
@@ -1771,6 +2199,7 @@ void amdgpu_vm_bo_del(struct amdgpu_device *adev,
 			if (*base != &bo_va->base)
 				continue;
 
+			amdgpu_vm_update_stats(*base, bo->tbo.resource, -1);
 			*base = bo_va->base.next;
 			break;
 		}
@@ -1839,25 +2268,19 @@ bool amdgpu_vm_evictable(struct amdgpu_bo *bo)
 /**
  * amdgpu_vm_bo_invalidate - mark the bo as invalid
  *
- * @adev: amdgpu_device pointer
  * @bo: amdgpu buffer object
  * @evicted: is the BO evicted
  *
  * Mark @bo as invalid.
  */
-void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
-			     struct amdgpu_bo *bo, bool evicted)
+void amdgpu_vm_bo_invalidate(struct amdgpu_bo *bo, bool evicted)
 {
 	struct amdgpu_vm_bo_base *bo_base;
 
-	/* shadow bo doesn't have bo base, its validation needs its parent */
-	if (bo->parent && (amdgpu_bo_shadowed(bo->parent) == bo))
-		bo = bo->parent;
-
 	for (bo_base = bo->vm_bo; bo_base; bo_base = bo_base->next) {
 		struct amdgpu_vm *vm = bo_base->vm;
 
-		if (evicted && bo->tbo.base.resv == vm->root.bo->tbo.base.resv) {
+		if (evicted && amdgpu_vm_is_bo_always_valid(vm, bo)) {
 			amdgpu_vm_bo_evicted(bo_base);
 			continue;
 		}
@@ -1868,7 +2291,7 @@ void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
 
 		if (bo->tbo.type == ttm_bo_type_kernel)
 			amdgpu_vm_bo_relocated(bo_base);
-		else if (bo->tbo.base.resv == vm->root.bo->tbo.base.resv)
+		else if (amdgpu_vm_is_bo_always_valid(vm, bo))
 			amdgpu_vm_bo_moved(bo_base);
 		else
 			amdgpu_vm_bo_invalidated(bo_base);
@@ -1876,6 +2299,32 @@ void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
 }
 
 /**
+ * amdgpu_vm_bo_move - handle BO move
+ *
+ * @bo: amdgpu buffer object
+ * @new_mem: the new placement of the BO move
+ * @evicted: is the BO evicted
+ *
+ * Update the memory stats for the new placement and mark @bo as invalid.
+ */
+void amdgpu_vm_bo_move(struct amdgpu_bo *bo, struct ttm_resource *new_mem,
+		       bool evicted)
+{
+	struct amdgpu_vm_bo_base *bo_base;
+
+	for (bo_base = bo->vm_bo; bo_base; bo_base = bo_base->next) {
+		struct amdgpu_vm *vm = bo_base->vm;
+
+		spin_lock(&vm->status_lock);
+		amdgpu_vm_update_stats_locked(bo_base, bo->tbo.resource, -1);
+		amdgpu_vm_update_stats_locked(bo_base, new_mem, +1);
+		spin_unlock(&vm->status_lock);
+	}
+
+	amdgpu_vm_bo_invalidate(bo, evicted);
+}
+
+/**
  * amdgpu_vm_get_block_size - calculate VM page table size as power of two
  *
  * @vm_size: VM size
@@ -1945,7 +2394,7 @@ void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint32_t min_vm_size,
 		phys_ram_gb = ((uint64_t)si.totalram * si.mem_unit +
 			       (1 << 30) - 1) >> 30;
 		vm_size = roundup_pow_of_two(
-			min(max(phys_ram_gb * 3, min_vm_size), max_size));
+			clamp(phys_ram_gb * 3, min_vm_size, max_size));
 	}
 
 	adev->vm_manager.max_pfn = (uint64_t)vm_size << 18;
@@ -1954,7 +2403,7 @@ void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint32_t min_vm_size,
 	if (amdgpu_vm_block_size != -1)
 		tmp >>= amdgpu_vm_block_size - 9;
 	tmp = DIV_ROUND_UP(fls64(tmp) - 1, 9) - 1;
-	adev->vm_manager.num_level = min(max_level, (unsigned)tmp);
+	adev->vm_manager.num_level = min_t(unsigned int, max_level, tmp);
 	switch (adev->vm_manager.num_level) {
 	case 3:
 		adev->vm_manager.root_level = AMDGPU_VM_PDB2;
@@ -1984,10 +2433,11 @@ void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint32_t min_vm_size,
 	else
 		adev->vm_manager.fragment_size = amdgpu_vm_fragment_size;
 
-	DRM_INFO("vm size is %u GB, %u levels, block size is %u-bit, fragment size is %u-bit\n",
-		 vm_size, adev->vm_manager.num_level + 1,
-		 adev->vm_manager.block_size,
-		 adev->vm_manager.fragment_size);
+	dev_info(
+		adev->dev,
+		"vm size is %u GB, %u levels, block size is %u-bit, fragment size is %u-bit\n",
+		vm_size, adev->vm_manager.num_level + 1,
+		adev->vm_manager.block_size, adev->vm_manager.fragment_size);
 }
 
 /**
@@ -1998,13 +2448,114 @@ void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint32_t min_vm_size,
  */
 long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout)
 {
-	timeout = dma_resv_wait_timeout(vm->root.bo->tbo.base.resv,
-					DMA_RESV_USAGE_BOOKKEEP,
-					true, timeout);
+	timeout = drm_sched_entity_flush(&vm->immediate, timeout);
 	if (timeout <= 0)
 		return timeout;
 
-	return dma_fence_wait_timeout(vm->last_unlocked, true, timeout);
+	return drm_sched_entity_flush(&vm->delayed, timeout);
+}
+
+static void amdgpu_vm_destroy_task_info(struct kref *kref)
+{
+	struct amdgpu_task_info *ti = container_of(kref, struct amdgpu_task_info, refcount);
+
+	kfree(ti);
+}
+
+static inline struct amdgpu_vm *
+amdgpu_vm_get_vm_from_pasid(struct amdgpu_device *adev, u32 pasid)
+{
+	struct amdgpu_vm *vm;
+	unsigned long flags;
+
+	xa_lock_irqsave(&adev->vm_manager.pasids, flags);
+	vm = xa_load(&adev->vm_manager.pasids, pasid);
+	xa_unlock_irqrestore(&adev->vm_manager.pasids, flags);
+
+	return vm;
+}
+
+/**
+ * amdgpu_vm_put_task_info - reference down the vm task_info ptr
+ *
+ * @task_info: task_info struct under discussion.
+ *
+ * frees the vm task_info ptr at the last put
+ */
+void amdgpu_vm_put_task_info(struct amdgpu_task_info *task_info)
+{
+	if (task_info)
+		kref_put(&task_info->refcount, amdgpu_vm_destroy_task_info);
+}
+
+/**
+ * amdgpu_vm_get_task_info_vm - Extracts task info for a vm.
+ *
+ * @vm: VM to get info from
+ *
+ * Returns the reference counted task_info structure, which must be
+ * referenced down with amdgpu_vm_put_task_info.
+ */
+struct amdgpu_task_info *
+amdgpu_vm_get_task_info_vm(struct amdgpu_vm *vm)
+{
+	struct amdgpu_task_info *ti = NULL;
+
+	if (vm) {
+		ti = vm->task_info;
+		kref_get(&vm->task_info->refcount);
+	}
+
+	return ti;
+}
+
+/**
+ * amdgpu_vm_get_task_info_pasid - Extracts task info for a PASID.
+ *
+ * @adev: drm device pointer
+ * @pasid: PASID identifier for VM
+ *
+ * Returns the reference counted task_info structure, which must be
+ * referenced down with amdgpu_vm_put_task_info.
+ */
+struct amdgpu_task_info *
+amdgpu_vm_get_task_info_pasid(struct amdgpu_device *adev, u32 pasid)
+{
+	return amdgpu_vm_get_task_info_vm(
+			amdgpu_vm_get_vm_from_pasid(adev, pasid));
+}
+
+static int amdgpu_vm_create_task_info(struct amdgpu_vm *vm)
+{
+	vm->task_info = kzalloc(sizeof(struct amdgpu_task_info), GFP_KERNEL);
+	if (!vm->task_info)
+		return -ENOMEM;
+
+	kref_init(&vm->task_info->refcount);
+	return 0;
+}
+
+/**
+ * amdgpu_vm_set_task_info - Sets VMs task info.
+ *
+ * @vm: vm for which to set the info
+ */
+void amdgpu_vm_set_task_info(struct amdgpu_vm *vm)
+{
+	if (!vm->task_info)
+		return;
+
+	if (vm->task_info->task.pid == current->pid)
+		return;
+
+	vm->task_info->task.pid = current->pid;
+	get_task_comm(vm->task_info->task.comm, current);
+
+	if (current->group_leader->mm != current->mm)
+		return;
+
+	vm->task_info->tgid = current->group_leader->pid;
+	get_task_comm(vm->task_info->process_name, current->group_leader);
 }
 
 /**
@@ -2012,13 +2563,16 @@ long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout)
  *
  * @adev: amdgpu_device pointer
  * @vm: requested vm
+ * @xcp_id: GPU partition selection id
+ * @pasid: the pasid the VM is using on this GPU
  *
  * Init @vm fields.
  *
  * Returns:
  * 0 for success, error for failure.
  */
-int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm)
+int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+		   int32_t xcp_id, uint32_t pasid)
 {
 	struct amdgpu_bo *root_bo;
 	struct amdgpu_bo_vm *root;
@@ -2028,6 +2582,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++)
 		vm->reserved_vmid[i] = NULL;
 	INIT_LIST_HEAD(&vm->evicted);
+	INIT_LIST_HEAD(&vm->evicted_user);
 	INIT_LIST_HEAD(&vm->relocated);
 	INIT_LIST_HEAD(&vm->moved);
 	INIT_LIST_HEAD(&vm->idle);
@@ -2035,30 +2590,21 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 	spin_lock_init(&vm->status_lock);
 	INIT_LIST_HEAD(&vm->freed);
 	INIT_LIST_HEAD(&vm->done);
-	INIT_LIST_HEAD(&vm->pt_freed);
-	INIT_WORK(&vm->pt_free_work, amdgpu_vm_pt_free_work);
+	INIT_KFIFO(vm->faults);
 
-	/* create scheduler entities for page table updates */
-	r = drm_sched_entity_init(&vm->immediate, DRM_SCHED_PRIORITY_NORMAL,
-				  adev->vm_manager.vm_pte_scheds,
-				  adev->vm_manager.vm_pte_num_scheds, NULL);
+	r = amdgpu_vm_init_entities(adev, vm);
 	if (r)
 		return r;
 
-	r = drm_sched_entity_init(&vm->delayed, DRM_SCHED_PRIORITY_NORMAL,
-				  adev->vm_manager.vm_pte_scheds,
-				  adev->vm_manager.vm_pte_num_scheds, NULL);
-	if (r)
-		goto error_free_immediate;
+	ttm_lru_bulk_move_init(&vm->lru_bulk_move);
 
-	vm->pte_support_ats = false;
 	vm->is_compute_context = false;
 
 	vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode &
 				    AMDGPU_VM_USE_CPU_FOR_GFX);
 
-	DRM_DEBUG_DRIVER("VM update mode is %s\n",
-			 vm->use_cpu_for_update ? "CPU" : "SDMA");
+	dev_dbg(adev->dev, "VM update mode is %s\n",
+		vm->use_cpu_for_update ? "CPU" : "SDMA");
 	WARN_ONCE((vm->use_cpu_for_update &&
 		   !amdgpu_gmc_vram_full_visible(&adev->gmc)),
 		  "CPU update of VM recommended only for large BAR system\n");
@@ -2067,53 +2613,70 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 		vm->update_funcs = &amdgpu_vm_cpu_funcs;
 	else
 		vm->update_funcs = &amdgpu_vm_sdma_funcs;
-	vm->last_update = NULL;
+
+	vm->last_update = dma_fence_get_stub();
 	vm->last_unlocked = dma_fence_get_stub();
 	vm->last_tlb_flush = dma_fence_get_stub();
+	vm->generation = amdgpu_vm_generation(adev, NULL);
 
 	mutex_init(&vm->eviction_lock);
 	vm->evicting = false;
+	vm->tlb_fence_context = dma_fence_context_alloc(1);
 
 	r = amdgpu_vm_pt_create(adev, vm, adev->vm_manager.root_level,
-				false, &root);
+				false, &root, xcp_id);
 	if (r)
 		goto error_free_delayed;
-	root_bo = &root->bo;
+
+	root_bo = amdgpu_bo_ref(&root->bo);
 	r = amdgpu_bo_reserve(root_bo, true);
-	if (r)
-		goto error_free_root;
+	if (r) {
+		amdgpu_bo_unref(&root_bo);
+		goto error_free_delayed;
+	}
 
+	amdgpu_vm_bo_base_init(&vm->root, vm, root_bo);
 	r = dma_resv_reserve_fences(root_bo->tbo.base.resv, 1);
 	if (r)
-		goto error_unreserve;
-
-	amdgpu_vm_bo_base_init(&vm->root, vm, root_bo);
+		goto error_free_root;
 
 	r = amdgpu_vm_pt_clear(adev, vm, root, false);
 	if (r)
-		goto error_unreserve;
+		goto error_free_root;
 
-	amdgpu_bo_unreserve(vm->root.bo);
+	r = amdgpu_vm_create_task_info(vm);
+	if (r)
+		dev_dbg(adev->dev, "Failed to create task info for VM\n");
 
-	INIT_KFIFO(vm->faults);
+	/* Store new PASID in XArray (if non-zero) */
+	if (pasid != 0) {
+		r = xa_err(xa_store_irq(&adev->vm_manager.pasids, pasid, vm, GFP_KERNEL));
+		if (r < 0)
+			goto error_free_root;
 
-	return 0;
+		vm->pasid = pasid;
+	}
 
-error_unreserve:
 	amdgpu_bo_unreserve(vm->root.bo);
+	amdgpu_bo_unref(&root_bo);
+
+	return 0;
 
 error_free_root:
-	amdgpu_bo_unref(&root->shadow);
+	/* If PASID was partially set, erase it from XArray before failing */
+	if (vm->pasid != 0) {
+		xa_erase_irq(&adev->vm_manager.pasids, vm->pasid);
+		vm->pasid = 0;
+	}
+	amdgpu_vm_pt_free_root(adev, vm);
+	amdgpu_bo_unreserve(vm->root.bo);
 	amdgpu_bo_unref(&root_bo);
-	vm->root.bo = NULL;
 
 error_free_delayed:
 	dma_fence_put(vm->last_tlb_flush);
 	dma_fence_put(vm->last_unlocked);
-	drm_sched_entity_destroy(&vm->delayed);
-
-error_free_immediate:
-	drm_sched_entity_destroy(&vm->immediate);
+	ttm_lru_bulk_move_fini(&adev->mman.bdev, &vm->lru_bulk_move);
+	amdgpu_vm_fini_entities(vm);
 
 	return r;
 }
@@ -2139,35 +2702,17 @@ error_free_immediate:
  */
 int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 {
-	bool pte_support_ats = (adev->asic_type == CHIP_RAVEN);
 	int r;
 
 	r = amdgpu_bo_reserve(vm->root.bo, true);
 	if (r)
 		return r;
 
-	/* Sanity checks */
-	if (!amdgpu_vm_pt_is_root_clean(adev, vm)) {
-		r = -EINVAL;
-		goto unreserve_bo;
-	}
-
-	/* Check if PD needs to be reinitialized and do it before
-	 * changing any other state, in case it fails.
-	 */
-	if (pte_support_ats != vm->pte_support_ats) {
-		vm->pte_support_ats = pte_support_ats;
-		r = amdgpu_vm_pt_clear(adev, vm, to_amdgpu_bo_vm(vm->root.bo),
-				       false);
-		if (r)
-			goto unreserve_bo;
-	}
-
 	/* Update VM state */
 	vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode &
 				    AMDGPU_VM_USE_CPU_FOR_COMPUTE);
-	DRM_DEBUG_DRIVER("VM update mode is %s\n",
-			 vm->use_cpu_for_update ? "CPU" : "SDMA");
+	dev_dbg(adev->dev, "VM update mode is %s\n",
+		vm->use_cpu_for_update ? "CPU" : "SDMA");
 	WARN_ONCE((vm->use_cpu_for_update &&
 		   !amdgpu_gmc_vram_full_visible(&adev->gmc)),
 		  "CPU update of VM recommended only for large BAR system\n");
@@ -2180,43 +2725,31 @@ int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 			goto unreserve_bo;
 
 		vm->update_funcs = &amdgpu_vm_cpu_funcs;
+		r = amdgpu_vm_pt_map_tables(adev, vm);
+		if (r)
+			goto unreserve_bo;
+
 	} else {
 		vm->update_funcs = &amdgpu_vm_sdma_funcs;
 	}
-	/*
-	 * Make sure root PD gets mapped. As vm_update_mode could be changed
-	 * when turning a GFX VM into a compute VM.
-	 */
-	r = vm->update_funcs->map_table(to_amdgpu_bo_vm(vm->root.bo));
-	if (r)
-		goto unreserve_bo;
 
 	dma_fence_put(vm->last_update);
-	vm->last_update = NULL;
+	vm->last_update = dma_fence_get_stub();
 	vm->is_compute_context = true;
 
-	/* Free the shadow bo for compute VM */
-	amdgpu_bo_unref(&to_amdgpu_bo_vm(vm->root.bo)->shadow);
-
-	goto unreserve_bo;
-
 unreserve_bo:
 	amdgpu_bo_unreserve(vm->root.bo);
 	return r;
 }
 
-/**
- * amdgpu_vm_release_compute - release a compute vm
- * @adev: amdgpu_device pointer
- * @vm: a vm turned into compute vm by calling amdgpu_vm_make_compute
- *
- * This is a correspondant of amdgpu_vm_make_compute. It decouples compute
- * pasid from vm. Compute should stop use of vm after this call.
- */
-void amdgpu_vm_release_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm)
+static int amdgpu_vm_stats_is_zero(struct amdgpu_vm *vm)
 {
-	amdgpu_vm_set_pasid(adev, vm, 0);
-	vm->is_compute_context = false;
+	for (int i = 0; i < __AMDGPU_PL_NUM; ++i) {
+		if (!(drm_memory_stats_is_zero(&vm->stats[i].drm) &&
+		      vm->stats[i].evicted == 0))
+			return false;
+	}
+	return true;
 }
 
 /**
@@ -2238,11 +2771,13 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 
 	amdgpu_amdkfd_gpuvm_destroy_cb(adev, vm);
 
-	flush_work(&vm->pt_free_work);
-
 	root = amdgpu_bo_ref(vm->root.bo);
 	amdgpu_bo_reserve(root, true);
-	amdgpu_vm_set_pasid(adev, vm, 0);
+	/* Remove PASID mapping before destroying VM */
+	if (vm->pasid != 0) {
+		xa_erase_irq(&adev->vm_manager.pasids, vm->pasid);
+		vm->pasid = 0;
+	}
 	dma_fence_wait(vm->last_unlocked, false);
 	dma_fence_put(vm->last_unlocked);
 	dma_fence_wait(vm->last_tlb_flush, false);
@@ -2252,7 +2787,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 	dma_fence_put(vm->last_tlb_flush);
 
 	list_for_each_entry_safe(mapping, tmp, &vm->freed, list) {
-		if (mapping->flags & AMDGPU_PTE_PRT && prt_fini_needed) {
+		if (mapping->flags & AMDGPU_VM_PAGE_PRT && prt_fini_needed) {
 			amdgpu_vm_prt_fini(adev, vm);
 			prt_fini_needed = false;
 		}
@@ -2266,8 +2801,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 	amdgpu_bo_unref(&root);
 	WARN_ON(vm->root.bo);
 
-	drm_sched_entity_destroy(&vm->immediate);
-	drm_sched_entity_destroy(&vm->delayed);
+	amdgpu_vm_fini_entities(vm);
 
 	if (!RB_EMPTY_ROOT(&vm->va.rb_root)) {
 		dev_err(adev->dev, "still active bo inside vm\n");
@@ -2282,8 +2816,22 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 	}
 
 	dma_fence_put(vm->last_update);
-	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++)
+
+	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++) {
 		amdgpu_vmid_free_reserved(adev, vm, i);
+	}
+
+	ttm_lru_bulk_move_fini(&adev->mman.bdev, &vm->lru_bulk_move);
+
+	if (!amdgpu_vm_stats_is_zero(vm)) {
+		struct amdgpu_task_info *ti = vm->task_info;
+
+		dev_warn(adev->dev,
+			 "VM memory stats for proc %s(%d) task %s(%d) is non-zero when fini\n",
+			 ti->process_name, ti->task.pid, ti->task.comm, ti->tgid);
+	}
+
+	amdgpu_vm_put_task_info(vm->task_info);
 }
 
 /**
@@ -2295,8 +2843,6 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
  */
 void amdgpu_vm_manager_init(struct amdgpu_device *adev)
 {
-	unsigned i;
-
 	/* Concurrent flushes are only possible starting with Vega10 and
 	 * are broken on Navi10 and Navi14.
 	 */
@@ -2305,11 +2851,6 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev)
 					      adev->asic_type == CHIP_NAVI14);
 	amdgpu_vmid_mgr_init(adev);
 
-	adev->vm_manager.fence_context =
-		dma_fence_context_alloc(AMDGPU_MAX_RINGS);
-	for (i = 0; i < AMDGPU_MAX_RINGS; ++i)
-		adev->vm_manager.seqno[i] = 0;
-
 	spin_lock_init(&adev->vm_manager.prt_lock);
 	atomic_set(&adev->vm_manager.num_prt_users, 0);
 
@@ -2366,18 +2907,18 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	union drm_amdgpu_vm *args = data;
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct amdgpu_fpriv *fpriv = filp->driver_priv;
-	int r;
+	struct amdgpu_vm *vm = &fpriv->vm;
+
+	/* No valid flags defined yet */
+	if (args->in.flags)
+		return -EINVAL;
 
 	switch (args->in.op) {
 	case AMDGPU_VM_OP_RESERVE_VMID:
 		/* We only have requirement to reserve vmid from gfxhub */
-		r = amdgpu_vmid_alloc_reserved(adev, &fpriv->vm,
-					       AMDGPU_GFXHUB_0);
-		if (r)
-			return r;
-		break;
+		return amdgpu_vmid_alloc_reserved(adev, vm, AMDGPU_GFXHUB(0));
 	case AMDGPU_VM_OP_UNRESERVE_VMID:
-		amdgpu_vmid_free_reserved(adev, &fpriv->vm, AMDGPU_GFXHUB_0);
+		amdgpu_vmid_free_reserved(adev, vm, AMDGPU_GFXHUB(0));
 		break;
 	default:
 		return -EINVAL;
@@ -2387,51 +2928,13 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 }
 
 /**
- * amdgpu_vm_get_task_info - Extracts task info for a PASID.
- *
- * @adev: drm device pointer
- * @pasid: PASID identifier for VM
- * @task_info: task_info to fill.
- */
-void amdgpu_vm_get_task_info(struct amdgpu_device *adev, u32 pasid,
-			 struct amdgpu_task_info *task_info)
-{
-	struct amdgpu_vm *vm;
-	unsigned long flags;
-
-	xa_lock_irqsave(&adev->vm_manager.pasids, flags);
-
-	vm = xa_load(&adev->vm_manager.pasids, pasid);
-	if (vm)
-		*task_info = vm->task_info;
-
-	xa_unlock_irqrestore(&adev->vm_manager.pasids, flags);
-}
-
-/**
- * amdgpu_vm_set_task_info - Sets VMs task info.
- *
- * @vm: vm for which to set the info
- */
-void amdgpu_vm_set_task_info(struct amdgpu_vm *vm)
-{
-	if (vm->task_info.pid)
-		return;
-
-	vm->task_info.pid = current->pid;
-	get_task_comm(vm->task_info.task_name, current);
-
-	if (current->group_leader->mm != current->mm)
-		return;
-
-	vm->task_info.tgid = current->group_leader->pid;
-	get_task_comm(vm->task_info.process_name, current->group_leader);
-}
-
-/**
  * amdgpu_vm_handle_fault - graceful handling of VM faults.
  * @adev: amdgpu device pointer
  * @pasid: PASID of the VM
+ * @ts: Timestamp of the fault
+ * @vmid: VMID, only used for GFX 9.4.3.
+ * @node_id: Node_id received in IH cookie. Only applicable for
+ *           GFX 9.4.3.
  * @addr: Address of the fault
  * @write_fault: true is write fault, false is read fault
  *
@@ -2439,7 +2942,8 @@ void amdgpu_vm_set_task_info(struct amdgpu_vm *vm)
  * shouldn't be reported any more.
  */
 bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid,
-			    uint64_t addr, bool write_fault)
+			    u32 vmid, u32 node_id, uint64_t addr, uint64_t ts,
+			    bool write_fault)
 {
 	bool is_compute_context = false;
 	struct amdgpu_bo *root;
@@ -2463,8 +2967,8 @@ bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid,
 
 	addr /= AMDGPU_GPU_PAGE_SIZE;
 
-	if (is_compute_context &&
-	    !svm_range_restore_pages(adev, pasid, addr, write_fault)) {
+	if (is_compute_context && !svm_range_restore_pages(adev, pasid, vmid,
+	    node_id, addr, ts, write_fault)) {
 		amdgpu_bo_unref(&root);
 		return true;
 	}
@@ -2489,7 +2993,7 @@ bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid,
 		/* Intentionally setting invalid PTE flag
 		 * combination to force a no-retry-fault
 		 */
-		flags = AMDGPU_PTE_SNOOPED | AMDGPU_PTE_PRT;
+		flags = AMDGPU_VM_NORETRY_FLAGS;
 		value = 0;
 	} else if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_NEVER) {
 		/* Redirect the access to the dummy page */
@@ -2508,8 +3012,8 @@ bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid,
 		goto error_unlock;
 	}
 
-	r = amdgpu_vm_update_range(adev, vm, true, false, false, NULL, addr,
-				   addr, flags, value, 0, NULL, NULL, NULL);
+	r = amdgpu_vm_update_range(adev, vm, true, false, false, false,
+				   NULL, addr, addr, flags, value, 0, NULL, NULL, NULL);
 	if (r)
 		goto error_unlock;
 
@@ -2518,7 +3022,7 @@ bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid,
 error_unlock:
 	amdgpu_bo_unreserve(root);
 	if (r < 0)
-		DRM_ERROR("Can't handle page fault (%d)\n", r);
+		dev_err(adev->dev, "Can't handle page fault (%d)\n", r);
 
 error_unref:
 	amdgpu_bo_unref(&root);
@@ -2552,6 +3056,8 @@ void amdgpu_debugfs_vm_bo_info(struct amdgpu_vm *vm, struct seq_file *m)
 	unsigned int total_done_objs = 0;
 	unsigned int id = 0;
 
+	amdgpu_vm_assert_locked(vm);
+
 	spin_lock(&vm->status_lock);
 	seq_puts(m, "\tIdle BOs:\n");
 	list_for_each_entry_safe(bo_va, tmp, &vm->idle, base.vm_status) {
@@ -2621,3 +3127,83 @@ void amdgpu_debugfs_vm_bo_info(struct amdgpu_vm *vm, struct seq_file *m)
 		   total_done_objs);
 }
 #endif
+
+/**
+ * amdgpu_vm_update_fault_cache - update cached fault into.
+ * @adev: amdgpu device pointer
+ * @pasid: PASID of the VM
+ * @addr: Address of the fault
+ * @status: GPUVM fault status register
+ * @vmhub: which vmhub got the fault
+ *
+ * Cache the fault info for later use by userspace in debugging.
+ */
+void amdgpu_vm_update_fault_cache(struct amdgpu_device *adev,
+				  unsigned int pasid,
+				  uint64_t addr,
+				  uint32_t status,
+				  unsigned int vmhub)
+{
+	struct amdgpu_vm *vm;
+	unsigned long flags;
+
+	xa_lock_irqsave(&adev->vm_manager.pasids, flags);
+
+	vm = xa_load(&adev->vm_manager.pasids, pasid);
+	/* Don't update the fault cache if status is 0.  In the multiple
+	 * fault case, subsequent faults will return a 0 status which is
+	 * useless for userspace and replaces the useful fault status, so
+	 * only update if status is non-0.
+	 */
+	if (vm && status) {
+		vm->fault_info.addr = addr;
+		vm->fault_info.status = status;
+		/*
+		 * Update the fault information globally for later usage
+		 * when vm could be stale or freed.
+		 */
+		adev->vm_manager.fault_info.addr = addr;
+		adev->vm_manager.fault_info.vmhub = vmhub;
+		adev->vm_manager.fault_info.status = status;
+
+		if (AMDGPU_IS_GFXHUB(vmhub)) {
+			vm->fault_info.vmhub = AMDGPU_VMHUB_TYPE_GFX;
+			vm->fault_info.vmhub |=
+				(vmhub - AMDGPU_GFXHUB_START) << AMDGPU_VMHUB_IDX_SHIFT;
+		} else if (AMDGPU_IS_MMHUB0(vmhub)) {
+			vm->fault_info.vmhub = AMDGPU_VMHUB_TYPE_MM0;
+			vm->fault_info.vmhub |=
+				(vmhub - AMDGPU_MMHUB0_START) << AMDGPU_VMHUB_IDX_SHIFT;
+		} else if (AMDGPU_IS_MMHUB1(vmhub)) {
+			vm->fault_info.vmhub = AMDGPU_VMHUB_TYPE_MM1;
+			vm->fault_info.vmhub |=
+				(vmhub - AMDGPU_MMHUB1_START) << AMDGPU_VMHUB_IDX_SHIFT;
+		} else {
+			WARN_ONCE(1, "Invalid vmhub %u\n", vmhub);
+		}
+	}
+	xa_unlock_irqrestore(&adev->vm_manager.pasids, flags);
+}
+
+/**
+ * amdgpu_vm_is_bo_always_valid - check if the BO is VM always valid
+ *
+ * @vm: VM to test against.
+ * @bo: BO to be tested.
+ *
+ * Returns true if the BO shares the dma_resv object with the root PD and is
+ * always guaranteed to be valid inside the VM.
+ */
+bool amdgpu_vm_is_bo_always_valid(struct amdgpu_vm *vm, struct amdgpu_bo *bo)
+{
+	return bo && bo->tbo.base.resv == vm->root.bo->tbo.base.resv;
+}
+
+void amdgpu_vm_print_task_info(struct amdgpu_device *adev,
+			       struct amdgpu_task_info *task_info)
+{
+	dev_err(adev->dev,
+		" Process %s pid %d thread %s pid %d\n",
+		task_info->process_name, task_info->tgid,
+		task_info->task.comm, task_info->task.pid);
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
index 6f085f0b4ef3..15d757c016cb 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
@@ -35,12 +35,14 @@
 #include "amdgpu_sync.h"
 #include "amdgpu_ring.h"
 #include "amdgpu_ids.h"
+#include "amdgpu_ttm.h"
+
+struct drm_exec;
 
 struct amdgpu_bo_va;
 struct amdgpu_job;
 struct amdgpu_bo_list_entry;
 struct amdgpu_bo_vm;
-struct amdgpu_mem_stats;
 
 /*
  * GPUVM handling
@@ -84,10 +86,19 @@ struct amdgpu_mem_stats;
 /* PDE Block Fragment Size for VEGA10 */
 #define AMDGPU_PDE_BFS(a)	((uint64_t)a << 59)
 
+/* Flag combination to set no-retry with TF disabled */
+#define AMDGPU_VM_NORETRY_FLAGS	(AMDGPU_PTE_EXECUTABLE | AMDGPU_PDE_PTE | \
+				AMDGPU_PTE_TF)
 
+/* Flag combination to set no-retry with TF enabled */
+#define AMDGPU_VM_NORETRY_FLAGS_TF (AMDGPU_PTE_VALID | AMDGPU_PTE_SYSTEM | \
+				   AMDGPU_PTE_PRT)
 /* For GFX9 */
-#define AMDGPU_PTE_MTYPE_VG10(a)	((uint64_t)(a) << 57)
-#define AMDGPU_PTE_MTYPE_VG10_MASK	AMDGPU_PTE_MTYPE_VG10(3ULL)
+#define AMDGPU_PTE_MTYPE_VG10_SHIFT(mtype)	((uint64_t)(mtype) << 57)
+#define AMDGPU_PTE_MTYPE_VG10_MASK	AMDGPU_PTE_MTYPE_VG10_SHIFT(3ULL)
+#define AMDGPU_PTE_MTYPE_VG10(flags, mtype)			\
+	(((uint64_t)(flags) & (~AMDGPU_PTE_MTYPE_VG10_MASK)) |	\
+	  AMDGPU_PTE_MTYPE_VG10_SHIFT(mtype))
 
 #define AMDGPU_MTYPE_NC 0
 #define AMDGPU_MTYPE_CC 2
@@ -100,25 +111,74 @@ struct amdgpu_mem_stats;
                                 | AMDGPU_PTE_MTYPE_VG10(AMDGPU_MTYPE_CC))
 
 /* gfx10 */
-#define AMDGPU_PTE_MTYPE_NV10(a)       ((uint64_t)(a) << 48)
-#define AMDGPU_PTE_MTYPE_NV10_MASK     AMDGPU_PTE_MTYPE_NV10(7ULL)
+#define AMDGPU_PTE_MTYPE_NV10_SHIFT(mtype)	((uint64_t)(mtype) << 48)
+#define AMDGPU_PTE_MTYPE_NV10_MASK     AMDGPU_PTE_MTYPE_NV10_SHIFT(7ULL)
+#define AMDGPU_PTE_MTYPE_NV10(flags, mtype)			\
+	(((uint64_t)(flags) & (~AMDGPU_PTE_MTYPE_NV10_MASK)) |	\
+	  AMDGPU_PTE_MTYPE_NV10_SHIFT(mtype))
+
+/* gfx12 */
+#define AMDGPU_PTE_PRT_GFX12		(1ULL << 56)
+#define AMDGPU_PTE_PRT_FLAG(adev)	\
+	((amdgpu_ip_version((adev), GC_HWIP, 0) >= IP_VERSION(12, 0, 0)) ? AMDGPU_PTE_PRT_GFX12 : AMDGPU_PTE_PRT)
+
+#define AMDGPU_PTE_MTYPE_GFX12_SHIFT(mtype)	((uint64_t)(mtype) << 54)
+#define AMDGPU_PTE_MTYPE_GFX12_MASK	AMDGPU_PTE_MTYPE_GFX12_SHIFT(3ULL)
+#define AMDGPU_PTE_MTYPE_GFX12(flags, mtype)				\
+	(((uint64_t)(flags) & (~AMDGPU_PTE_MTYPE_GFX12_MASK)) |	\
+	  AMDGPU_PTE_MTYPE_GFX12_SHIFT(mtype))
+
+#define AMDGPU_PTE_DCC			(1ULL << 58)
+#define AMDGPU_PTE_IS_PTE		(1ULL << 63)
+
+/* PDE Block Fragment Size for gfx v12 */
+#define AMDGPU_PDE_BFS_GFX12(a)		((uint64_t)((a) & 0x1fULL) << 58)
+#define AMDGPU_PDE_BFS_FLAG(adev, a)	\
+	((amdgpu_ip_version((adev), GC_HWIP, 0) >= IP_VERSION(12, 0, 0)) ? AMDGPU_PDE_BFS_GFX12(a) : AMDGPU_PDE_BFS(a))
+/* PDE is handled as PTE for gfx v12 */
+#define AMDGPU_PDE_PTE_GFX12		(1ULL << 63)
+#define AMDGPU_PDE_PTE_FLAG(adev)	\
+	((amdgpu_ip_version((adev), GC_HWIP, 0) >= IP_VERSION(12, 0, 0)) ? AMDGPU_PDE_PTE_GFX12 : AMDGPU_PDE_PTE)
 
 /* How to program VM fault handling */
 #define AMDGPU_VM_FAULT_STOP_NEVER	0
 #define AMDGPU_VM_FAULT_STOP_FIRST	1
 #define AMDGPU_VM_FAULT_STOP_ALWAYS	2
 
-/* Reserve 4MB VRAM for page tables */
+/* How much VRAM be reserved for page tables */
 #define AMDGPU_VM_RESERVED_VRAM		(8ULL << 20)
 
-/* max number of VMHUB */
-#define AMDGPU_MAX_VMHUBS			3
-#define AMDGPU_GFXHUB_0				0
-#define AMDGPU_MMHUB_0				1
-#define AMDGPU_MMHUB_1				2
-
-/* Reserve 2MB at top/bottom of address space for kernel use */
-#define AMDGPU_VA_RESERVED_SIZE			(2ULL << 20)
+/*
+ * max number of VMHUB
+ * layout: max 8 GFXHUB + 4 MMHUB0 + 1 MMHUB1
+ */
+#define AMDGPU_MAX_VMHUBS			13
+#define AMDGPU_GFXHUB_START			0
+#define AMDGPU_MMHUB0_START			8
+#define AMDGPU_MMHUB1_START			12
+#define AMDGPU_GFXHUB(x)			(AMDGPU_GFXHUB_START + (x))
+#define AMDGPU_MMHUB0(x)			(AMDGPU_MMHUB0_START + (x))
+#define AMDGPU_MMHUB1(x)			(AMDGPU_MMHUB1_START + (x))
+
+#define AMDGPU_IS_GFXHUB(x) ((x) >= AMDGPU_GFXHUB_START && (x) < AMDGPU_MMHUB0_START)
+#define AMDGPU_IS_MMHUB0(x) ((x) >= AMDGPU_MMHUB0_START && (x) < AMDGPU_MMHUB1_START)
+#define AMDGPU_IS_MMHUB1(x) ((x) >= AMDGPU_MMHUB1_START && (x) < AMDGPU_MAX_VMHUBS)
+
+/* Reserve space at top/bottom of address space for kernel use */
+#define AMDGPU_VA_RESERVED_CSA_SIZE		(2ULL << 20)
+#define AMDGPU_VA_RESERVED_CSA_START(adev)	(((adev)->vm_manager.max_pfn \
+						  << AMDGPU_GPU_PAGE_SHIFT)  \
+						 - AMDGPU_VA_RESERVED_CSA_SIZE)
+#define AMDGPU_VA_RESERVED_SEQ64_SIZE		(2ULL << 20)
+#define AMDGPU_VA_RESERVED_SEQ64_START(adev)	(AMDGPU_VA_RESERVED_CSA_START(adev) \
+						 - AMDGPU_VA_RESERVED_SEQ64_SIZE)
+#define AMDGPU_VA_RESERVED_TRAP_SIZE		(2ULL << 12)
+#define AMDGPU_VA_RESERVED_TRAP_START(adev)	(AMDGPU_VA_RESERVED_SEQ64_START(adev) \
+						 - AMDGPU_VA_RESERVED_TRAP_SIZE)
+#define AMDGPU_VA_RESERVED_BOTTOM		(1ULL << 16)
+#define AMDGPU_VA_RESERVED_TOP			(AMDGPU_VA_RESERVED_TRAP_SIZE + \
+						 AMDGPU_VA_RESERVED_SEQ64_SIZE + \
+						 AMDGPU_VA_RESERVED_CSA_SIZE)
 
 /* See vm_update_mode */
 #define AMDGPU_VM_USE_CPU_FOR_GFX (1 << 0)
@@ -143,9 +203,13 @@ struct amdgpu_vm_bo_base {
 	/* protected by bo being reserved */
 	struct amdgpu_vm_bo_base	*next;
 
-	/* protected by spinlock */
+	/* protected by vm status_lock */
 	struct list_head		vm_status;
 
+	/* if the bo is counted as shared in mem stats
+	 * protected by vm status_lock */
+	bool				shared;
+
 	/* protected by the BO being reserved */
 	bool				moved;
 };
@@ -172,10 +236,10 @@ struct amdgpu_vm_pte_funcs {
 };
 
 struct amdgpu_task_info {
-	char	process_name[TASK_COMM_LEN];
-	char	task_name[TASK_COMM_LEN];
-	pid_t	pid;
-	pid_t	tgid;
+	struct drm_wedge_task_info task;
+	char		process_name[TASK_COMM_LEN];
+	pid_t		tgid;
+	struct kref	refcount;
 };
 
 /**
@@ -225,15 +289,26 @@ struct amdgpu_vm_update_params {
 	unsigned int num_dw_left;
 
 	/**
-	 * @table_freed: return true if page table is freed when updating
+	 * @needs_flush: true whenever we need to invalidate the TLB
+	 */
+	bool needs_flush;
+
+	/**
+	 * @allow_override: true for memory that is not uncached: allows MTYPE
+	 * to be overridden for NUMA local memory.
+	 */
+	bool allow_override;
+
+	/**
+	 * @tlb_flush_waitlist: temporary storage for BOs until tlb_flush
 	 */
-	bool table_freed;
+	struct list_head tlb_flush_waitlist;
 };
 
 struct amdgpu_vm_update_funcs {
 	int (*map_table)(struct amdgpu_bo_vm *bo);
-	int (*prepare)(struct amdgpu_vm_update_params *p, struct dma_resv *resv,
-		       enum amdgpu_sync_mode sync_mode);
+	int (*prepare)(struct amdgpu_vm_update_params *p,
+		       struct amdgpu_sync *sync, u64 k_job_id);
 	int (*update)(struct amdgpu_vm_update_params *p,
 		      struct amdgpu_bo_vm *bo, uint64_t pe, uint64_t addr,
 		      unsigned count, uint32_t incr, uint64_t flags);
@@ -241,6 +316,22 @@ struct amdgpu_vm_update_funcs {
 		      struct dma_fence **fence);
 };
 
+struct amdgpu_vm_fault_info {
+	/* fault address */
+	uint64_t	addr;
+	/* fault status register */
+	uint32_t	status;
+	/* which vmhub? gfxhub, mmhub, etc. */
+	unsigned int	vmhub;
+};
+
+struct amdgpu_mem_stats {
+	struct drm_memory_stats drm;
+
+	/* buffers that requested this placement but are currently evicted */
+	uint64_t evicted;
+};
+
 struct amdgpu_vm {
 	/* tree of virtual addresses mapped */
 	struct rb_root_cached	va;
@@ -255,7 +346,17 @@ struct amdgpu_vm {
 	/* Lock to protect vm_bo add/del/move on all lists of vm */
 	spinlock_t		status_lock;
 
-	/* BOs who needs a validation */
+	/* Memory statistics for this vm, protected by status_lock */
+	struct amdgpu_mem_stats stats[__AMDGPU_PL_NUM];
+
+	/*
+	 * The following lists contain amdgpu_vm_bo_base objects for either
+	 * PDs, PTs or per VM BOs. The state transits are:
+	 *
+	 * evicted -> relocated (PDs, PTs) or moved (per VM BOs) -> idle
+	 */
+
+	/* Per-VM and PT BOs who needs a validation */
 	struct list_head	evicted;
 
 	/* PT BOs which relocated and their parent need an update */
@@ -267,18 +368,28 @@ struct amdgpu_vm {
 	/* All BOs of this VM not currently in the state machine */
 	struct list_head	idle;
 
+	/*
+	 * The following lists contain amdgpu_vm_bo_base objects for BOs which
+	 * have their own dma_resv object and not depend on the root PD. Their
+	 * state transits are:
+	 *
+	 * evicted_user or invalidated -> done
+	 */
+
+	/* BOs for user mode queues that need a validation */
+	struct list_head	evicted_user;
+
 	/* regular invalidated BOs, but not yet updated in the PT */
 	struct list_head	invalidated;
 
-	/* BO mappings freed, but not yet updated in the PT */
-	struct list_head	freed;
-
 	/* BOs which are invalidated, has been updated in the PTs */
 	struct list_head        done;
 
-	/* PT BOs scheduled to free and fill with zero if vm_resv is not hold */
-	struct list_head	pt_freed;
-	struct work_struct	pt_free_work;
+	/*
+	 * This list contains amdgpu_bo_va_mapping objects which have been freed
+	 * but not updated in the PTs
+	 */
+	struct list_head	freed;
 
 	/* contains the page directory */
 	struct amdgpu_vm_bo_base     root;
@@ -291,12 +402,17 @@ struct amdgpu_vm {
 	/* Last finished delayed update */
 	atomic64_t		tlb_seq;
 	struct dma_fence	*last_tlb_flush;
+	atomic64_t		kfd_last_flushed_seq;
+	uint64_t		tlb_fence_context;
+
+	/* How many times we had to re-generate the page tables */
+	uint64_t		generation;
 
 	/* Last unlocked submission to the scheduler entities */
 	struct dma_fence	*last_unlocked;
 
 	unsigned int		pasid;
-	bool			reserved_vmid[AMDGPU_MAX_VMHUBS];
+	struct amdgpu_vmid	*reserved_vmid[AMDGPU_MAX_VMHUBS];
 
 	/* Flag to indicate if VM tables are updated by CPU or GPU (SDMA) */
 	bool					use_cpu_for_update;
@@ -304,9 +420,6 @@ struct amdgpu_vm {
 	/* Functions to use for VM table updates */
 	const struct amdgpu_vm_update_funcs	*update_funcs;
 
-	/* Flag to indicate ATS support from PTE for GFX9 */
-	bool			pte_support_ats;
-
 	/* Up to 128 pending retry page faults */
 	DECLARE_KFIFO(faults, u64, 128);
 
@@ -320,12 +433,18 @@ struct amdgpu_vm {
 	uint64_t		pd_phys_addr;
 
 	/* Some basic info about the task */
-	struct amdgpu_task_info task_info;
+	struct amdgpu_task_info *task_info;
 
 	/* Store positions of group of BOs */
 	struct ttm_lru_bulk_move lru_bulk_move;
 	/* Flag to indicate if VM is used for compute */
 	bool			is_compute_context;
+
+	/* Memory partition number, -1 means any partition */
+	int8_t			mem_id;
+
+	/* cached fault info */
+	struct amdgpu_vm_fault_info fault_info;
 };
 
 struct amdgpu_vm_manager {
@@ -334,10 +453,6 @@ struct amdgpu_vm_manager {
 	unsigned int				first_kfd_vmid;
 	bool					concurrent_flush;
 
-	/* Handling of VM fences */
-	u64					fence_context;
-	unsigned				seqno[AMDGPU_MAX_RINGS];
-
 	uint64_t				max_pfn;
 	uint32_t				num_level;
 	uint32_t				block_size;
@@ -365,6 +480,8 @@ struct amdgpu_vm_manager {
 	 * look up VM of a page fault
 	 */
 	struct xarray				pasids;
+	/* Global registration of recent page fault information */
+	struct amdgpu_vm_fault_info	fault_info;
 };
 
 struct amdgpu_bo_va_mapping;
@@ -379,21 +496,20 @@ extern const struct amdgpu_vm_update_funcs amdgpu_vm_sdma_funcs;
 void amdgpu_vm_manager_init(struct amdgpu_device *adev);
 void amdgpu_vm_manager_fini(struct amdgpu_device *adev);
 
-int amdgpu_vm_set_pasid(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-			u32 pasid);
-
 long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout);
-int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm);
+int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, int32_t xcp_id, uint32_t pasid);
 int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm);
-void amdgpu_vm_release_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm);
 void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm);
-void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm,
-			 struct list_head *validated,
-			 struct amdgpu_bo_list_entry *entry);
+int amdgpu_vm_lock_pd(struct amdgpu_vm *vm, struct drm_exec *exec,
+		      unsigned int num_fences);
+int amdgpu_vm_lock_done_list(struct amdgpu_vm *vm, struct drm_exec *exec,
+			     unsigned int num_fences);
 bool amdgpu_vm_ready(struct amdgpu_vm *vm);
-int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-			      int (*callback)(void *p, struct amdgpu_bo *bo),
-			      void *param);
+uint64_t amdgpu_vm_generation(struct amdgpu_device *adev, struct amdgpu_vm *vm);
+int amdgpu_vm_validate(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+		       struct ww_acquire_ctx *ticket,
+		       int (*callback)(void *p, struct amdgpu_bo *bo),
+		       void *param);
 int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_pipe_sync);
 int amdgpu_vm_update_pdes(struct amdgpu_device *adev,
 			  struct amdgpu_vm *vm, bool immediate);
@@ -401,21 +517,31 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
 			  struct amdgpu_vm *vm,
 			  struct dma_fence **fence);
 int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
-			   struct amdgpu_vm *vm);
+			   struct amdgpu_vm *vm,
+			   struct ww_acquire_ctx *ticket);
+int amdgpu_vm_flush_compute_tlb(struct amdgpu_device *adev,
+				struct amdgpu_vm *vm,
+				uint32_t flush_type,
+				uint32_t xcc_mask);
 void amdgpu_vm_bo_base_init(struct amdgpu_vm_bo_base *base,
 			    struct amdgpu_vm *vm, struct amdgpu_bo *bo);
 int amdgpu_vm_update_range(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			   bool immediate, bool unlocked, bool flush_tlb,
-			   struct dma_resv *resv, uint64_t start, uint64_t last,
-			   uint64_t flags, uint64_t offset, uint64_t vram_base,
+			   bool allow_override, struct amdgpu_sync *sync,
+			   uint64_t start, uint64_t last, uint64_t flags,
+			   uint64_t offset, uint64_t vram_base,
 			   struct ttm_resource *res, dma_addr_t *pages_addr,
 			   struct dma_fence **fence);
 int amdgpu_vm_bo_update(struct amdgpu_device *adev,
 			struct amdgpu_bo_va *bo_va,
 			bool clear);
 bool amdgpu_vm_evictable(struct amdgpu_bo *bo);
-void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
-			     struct amdgpu_bo *bo, bool evicted);
+void amdgpu_vm_bo_invalidate(struct amdgpu_bo *bo, bool evicted);
+void amdgpu_vm_update_stats(struct amdgpu_vm_bo_base *base,
+			    struct ttm_resource *new_res, int sign);
+void amdgpu_vm_bo_update_shared(struct amdgpu_bo *bo);
+void amdgpu_vm_bo_move(struct amdgpu_bo *bo, struct ttm_resource *new_mem,
+		       bool evicted);
 uint64_t amdgpu_vm_map_gart(const dma_addr_t *pages_addr, uint64_t addr);
 struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm,
 				       struct amdgpu_bo *bo);
@@ -425,11 +551,11 @@ struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev,
 int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 		     struct amdgpu_bo_va *bo_va,
 		     uint64_t addr, uint64_t offset,
-		     uint64_t size, uint64_t flags);
+		     uint64_t size, uint32_t flags);
 int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev,
 			     struct amdgpu_bo_va *bo_va,
 			     uint64_t addr, uint64_t offset,
-			     uint64_t size, uint64_t flags);
+			     uint64_t size, uint32_t flags);
 int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
 		       struct amdgpu_bo_va *bo_va,
 		       uint64_t addr);
@@ -449,25 +575,31 @@ bool amdgpu_vm_need_pipeline_sync(struct amdgpu_ring *ring,
 				  struct amdgpu_job *job);
 void amdgpu_vm_check_compute_bug(struct amdgpu_device *adev);
 
-void amdgpu_vm_get_task_info(struct amdgpu_device *adev, u32 pasid,
-			     struct amdgpu_task_info *task_info);
+struct amdgpu_task_info *
+amdgpu_vm_get_task_info_pasid(struct amdgpu_device *adev, u32 pasid);
+
+struct amdgpu_task_info *
+amdgpu_vm_get_task_info_vm(struct amdgpu_vm *vm);
+
+void amdgpu_vm_put_task_info(struct amdgpu_task_info *task_info);
+
 bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid,
-			    uint64_t addr, bool write_fault);
+			    u32 vmid, u32 node_id, uint64_t addr, uint64_t ts,
+			    bool write_fault);
 
 void amdgpu_vm_set_task_info(struct amdgpu_vm *vm);
 
 void amdgpu_vm_move_to_lru_tail(struct amdgpu_device *adev,
 				struct amdgpu_vm *vm);
 void amdgpu_vm_get_memory(struct amdgpu_vm *vm,
-			  struct amdgpu_mem_stats *stats);
+			  struct amdgpu_mem_stats stats[__AMDGPU_PL_NUM]);
 
 int amdgpu_vm_pt_clear(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		       struct amdgpu_bo_vm *vmbo, bool immediate);
 int amdgpu_vm_pt_create(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-			int level, bool immediate, struct amdgpu_bo_vm **vmbo);
+			int level, bool immediate, struct amdgpu_bo_vm **vmbo,
+			int32_t xcp_id);
 void amdgpu_vm_pt_free_root(struct amdgpu_device *adev, struct amdgpu_vm *vm);
-bool amdgpu_vm_pt_is_root_clean(struct amdgpu_device *adev,
-				struct amdgpu_vm *vm);
 
 int amdgpu_vm_pde_update(struct amdgpu_vm_update_params *params,
 			 struct amdgpu_vm_bo_base *entry);
@@ -475,11 +607,17 @@ int amdgpu_vm_ptes_update(struct amdgpu_vm_update_params *params,
 			  uint64_t start, uint64_t end,
 			  uint64_t dst, uint64_t flags);
 void amdgpu_vm_pt_free_work(struct work_struct *work);
+void amdgpu_vm_pt_free_list(struct amdgpu_device *adev,
+			    struct amdgpu_vm_update_params *params);
 
 #if defined(CONFIG_DEBUG_FS)
 void amdgpu_debugfs_vm_bo_info(struct amdgpu_vm *vm, struct seq_file *m);
 #endif
 
+int amdgpu_vm_pt_map_tables(struct amdgpu_device *adev, struct amdgpu_vm *vm);
+
+bool amdgpu_vm_is_bo_always_valid(struct amdgpu_vm *vm, struct amdgpu_bo *bo);
+
 /**
  * amdgpu_vm_tlb_seq - return tlb flush sequence number
  * @vm: the amdgpu_vm structure to query
@@ -533,4 +671,21 @@ static inline void amdgpu_vm_eviction_unlock(struct amdgpu_vm *vm)
 	mutex_unlock(&vm->eviction_lock);
 }
 
+void amdgpu_vm_update_fault_cache(struct amdgpu_device *adev,
+				  unsigned int pasid,
+				  uint64_t addr,
+				  uint32_t status,
+				  unsigned int vmhub);
+void amdgpu_vm_tlb_fence_create(struct amdgpu_device *adev,
+				 struct amdgpu_vm *vm,
+				 struct dma_fence **fence);
+
+void amdgpu_vm_print_task_info(struct amdgpu_device *adev,
+			       struct amdgpu_task_info *task_info);
+
+#define amdgpu_vm_bo_va_for_each_valid_mapping(bo_va, mapping) \
+		list_for_each_entry(mapping, &(bo_va)->valids, list)
+#define amdgpu_vm_bo_va_for_each_invalid_mapping(bo_va, mapping) \
+		list_for_each_entry(mapping, &(bo_va)->invalids, list)
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_cpu.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_cpu.c
index 31913ae86de6..22e2e5b47341 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_cpu.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_cpu.c
@@ -31,6 +31,7 @@
  */
 static int amdgpu_vm_cpu_map_table(struct amdgpu_bo_vm *table)
 {
+	table->bo.flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
 	return amdgpu_bo_kmap(&table->bo, NULL);
 }
 
@@ -38,20 +39,20 @@ static int amdgpu_vm_cpu_map_table(struct amdgpu_bo_vm *table)
  * amdgpu_vm_cpu_prepare - prepare page table update with the CPU
  *
  * @p: see amdgpu_vm_update_params definition
- * @resv: reservation object with embedded fence
- * @sync_mode: synchronization mode
+ * @sync: sync obj with fences to wait on
+ * @k_job_id: the id for tracing/debug purposes
  *
  * Returns:
  * Negativ errno, 0 for success.
  */
 static int amdgpu_vm_cpu_prepare(struct amdgpu_vm_update_params *p,
-				 struct dma_resv *resv,
-				 enum amdgpu_sync_mode sync_mode)
+				 struct amdgpu_sync *sync,
+				 u64 k_job_id)
 {
-	if (!resv)
+	if (!sync)
 		return 0;
 
-	return amdgpu_bo_sync_wait_resv(p->adev, resv, sync_mode, p->vm, true);
+	return amdgpu_sync_wait(sync, true);
 }
 
 /**
@@ -107,7 +108,9 @@ static int amdgpu_vm_cpu_update(struct amdgpu_vm_update_params *p,
 static int amdgpu_vm_cpu_commit(struct amdgpu_vm_update_params *p,
 				struct dma_fence **fence)
 {
-	/* Flush HDP */
+	if (p->needs_flush)
+		atomic64_inc(&p->vm->tlb_seq);
+
 	mb();
 	amdgpu_device_flush_hdp(p->adev, NULL);
 	return 0;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c
index df63dc3bca18..f794fb1cc06e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c
@@ -26,6 +26,7 @@
 #include "amdgpu.h"
 #include "amdgpu_trace.h"
 #include "amdgpu_vm.h"
+#include "amdgpu_job.h"
 
 /*
  * amdgpu_vm_pt_cursor - state for for_each_amdgpu_vm_pt
@@ -90,22 +91,6 @@ static unsigned int amdgpu_vm_pt_num_entries(struct amdgpu_device *adev,
 }
 
 /**
- * amdgpu_vm_pt_num_ats_entries - return the number of ATS entries in the root PD
- *
- * @adev: amdgpu_device pointer
- *
- * Returns:
- * The number of entries in the root page directory which needs the ATS setting.
- */
-static unsigned int amdgpu_vm_pt_num_ats_entries(struct amdgpu_device *adev)
-{
-	unsigned int shift;
-
-	shift = amdgpu_vm_pt_level_shift(adev, adev->vm_manager.root_level);
-	return AMDGPU_GMC_HOLE_START >> (shift + AMDGPU_GPU_PAGE_SHIFT);
-}
-
-/**
  * amdgpu_vm_pt_entries_mask - the mask to get the entry number of a PD/PT
  *
  * @adev: amdgpu_device pointer
@@ -379,7 +364,7 @@ int amdgpu_vm_pt_clear(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	struct ttm_operation_ctx ctx = { true, false };
 	struct amdgpu_vm_update_params params;
 	struct amdgpu_bo *ancestor = &vmbo->bo;
-	unsigned int entries, ats_entries;
+	unsigned int entries;
 	struct amdgpu_bo *bo = &vmbo->bo;
 	uint64_t addr;
 	int r, idx;
@@ -394,40 +379,11 @@ int amdgpu_vm_pt_clear(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	}
 
 	entries = amdgpu_bo_size(bo) / 8;
-	if (!vm->pte_support_ats) {
-		ats_entries = 0;
-
-	} else if (!bo->parent) {
-		ats_entries = amdgpu_vm_pt_num_ats_entries(adev);
-		ats_entries = min(ats_entries, entries);
-		entries -= ats_entries;
-
-	} else {
-		struct amdgpu_vm_bo_base *pt;
-
-		pt = ancestor->vm_bo;
-		ats_entries = amdgpu_vm_pt_num_ats_entries(adev);
-		if ((pt - to_amdgpu_bo_vm(vm->root.bo)->entries) >=
-		    ats_entries) {
-			ats_entries = 0;
-		} else {
-			ats_entries = entries;
-			entries = 0;
-		}
-	}
 
 	r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 	if (r)
 		return r;
 
-	if (vmbo->shadow) {
-		struct amdgpu_bo *shadow = vmbo->shadow;
-
-		r = ttm_bo_validate(&shadow->tbo, &shadow->placement, &ctx);
-		if (r)
-			return r;
-	}
-
 	if (!drm_dev_enter(adev_to_drm(adev), &idx))
 		return -ENODEV;
 
@@ -440,49 +396,30 @@ int amdgpu_vm_pt_clear(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	params.vm = vm;
 	params.immediate = immediate;
 
-	r = vm->update_funcs->prepare(&params, NULL, AMDGPU_SYNC_EXPLICIT);
+	r = vm->update_funcs->prepare(&params, NULL,
+				      AMDGPU_KERNEL_JOB_ID_VM_PT_CLEAR);
 	if (r)
 		goto exit;
 
 	addr = 0;
-	if (ats_entries) {
-		uint64_t value = 0, flags;
 
-		flags = AMDGPU_PTE_DEFAULT_ATC;
+	uint64_t value = 0, flags = 0;
+	if (adev->asic_type >= CHIP_VEGA10) {
 		if (level != AMDGPU_VM_PTB) {
 			/* Handle leaf PDEs as PTEs */
-			flags |= AMDGPU_PDE_PTE;
-			amdgpu_gmc_get_vm_pde(adev, level, &value, &flags);
+			flags |= AMDGPU_PDE_PTE_FLAG(adev);
+			amdgpu_gmc_get_vm_pde(adev, level,
+					      &value, &flags);
+		} else {
+			/* Workaround for fault priority problem on GMC9 */
+			flags = AMDGPU_PTE_EXECUTABLE;
 		}
-
-		r = vm->update_funcs->update(&params, vmbo, addr, 0,
-					     ats_entries, value, flags);
-		if (r)
-			goto exit;
-
-		addr += ats_entries * 8;
 	}
 
-	if (entries) {
-		uint64_t value = 0, flags = 0;
-
-		if (adev->asic_type >= CHIP_VEGA10) {
-			if (level != AMDGPU_VM_PTB) {
-				/* Handle leaf PDEs as PTEs */
-				flags |= AMDGPU_PDE_PTE;
-				amdgpu_gmc_get_vm_pde(adev, level,
-						      &value, &flags);
-			} else {
-				/* Workaround for fault priority problem on GMC9 */
-				flags = AMDGPU_PTE_EXECUTABLE;
-			}
-		}
-
-		r = vm->update_funcs->update(&params, vmbo, addr, 0, entries,
-					     value, flags);
-		if (r)
-			goto exit;
-	}
+	r = vm->update_funcs->update(&params, vmbo, addr, 0, entries,
+				     value, flags);
+	if (r)
+		goto exit;
 
 	r = vm->update_funcs->commit(&params, NULL);
 exit:
@@ -498,21 +435,25 @@ exit:
  * @level: the page table level
  * @immediate: use a immediate update
  * @vmbo: pointer to the buffer object pointer
+ * @xcp_id: GPU partition id
  */
 int amdgpu_vm_pt_create(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-			int level, bool immediate, struct amdgpu_bo_vm **vmbo)
+			int level, bool immediate, struct amdgpu_bo_vm **vmbo,
+			int32_t xcp_id)
 {
 	struct amdgpu_bo_param bp;
-	struct amdgpu_bo *bo;
-	struct dma_resv *resv;
 	unsigned int num_entries;
-	int r;
 
 	memset(&bp, 0, sizeof(bp));
 
 	bp.size = amdgpu_vm_pt_size(adev, level);
 	bp.byte_align = AMDGPU_GPU_PAGE_SIZE;
-	bp.domain = AMDGPU_GEM_DOMAIN_VRAM;
+
+	if (!adev->gmc.is_app_apu)
+		bp.domain = AMDGPU_GEM_DOMAIN_VRAM;
+	else
+		bp.domain = AMDGPU_GEM_DOMAIN_GTT;
+
 	bp.domain = amdgpu_bo_get_preferred_domain(adev, bp.domain);
 	bp.flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS |
 		AMDGPU_GEM_CREATE_CPU_GTT_USWC;
@@ -529,45 +470,12 @@ int amdgpu_vm_pt_create(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 
 	bp.type = ttm_bo_type_kernel;
 	bp.no_wait_gpu = immediate;
+	bp.xcp_id_plus1 = xcp_id + 1;
+
 	if (vm->root.bo)
 		bp.resv = vm->root.bo->tbo.base.resv;
 
-	r = amdgpu_bo_create_vm(adev, &bp, vmbo);
-	if (r)
-		return r;
-
-	bo = &(*vmbo)->bo;
-	if (vm->is_compute_context || (adev->flags & AMD_IS_APU)) {
-		(*vmbo)->shadow = NULL;
-		return 0;
-	}
-
-	if (!bp.resv)
-		WARN_ON(dma_resv_lock(bo->tbo.base.resv,
-				      NULL));
-	resv = bp.resv;
-	memset(&bp, 0, sizeof(bp));
-	bp.size = amdgpu_vm_pt_size(adev, level);
-	bp.domain = AMDGPU_GEM_DOMAIN_GTT;
-	bp.flags = AMDGPU_GEM_CREATE_CPU_GTT_USWC;
-	bp.type = ttm_bo_type_kernel;
-	bp.resv = bo->tbo.base.resv;
-	bp.bo_ptr_size = sizeof(struct amdgpu_bo);
-
-	r = amdgpu_bo_create(adev, &bp, &(*vmbo)->shadow);
-
-	if (!resv)
-		dma_resv_unlock(bo->tbo.base.resv);
-
-	if (r) {
-		amdgpu_bo_unref(&bo);
-		return r;
-	}
-
-	(*vmbo)->shadow->parent = amdgpu_bo_ref(bo);
-	amdgpu_bo_add_to_shadow_list(*vmbo);
-
-	return 0;
+	return amdgpu_bo_create_vm(adev, &bp, vmbo);
 }
 
 /**
@@ -598,7 +506,8 @@ static int amdgpu_vm_pt_alloc(struct amdgpu_device *adev,
 		return 0;
 
 	amdgpu_vm_eviction_unlock(vm);
-	r = amdgpu_vm_pt_create(adev, vm, cursor->level, immediate, &pt);
+	r = amdgpu_vm_pt_create(adev, vm, cursor->level, immediate, &pt,
+				vm->root.bo->xcp_id);
 	amdgpu_vm_eviction_lock(vm);
 	if (r)
 		return r;
@@ -616,7 +525,6 @@ static int amdgpu_vm_pt_alloc(struct amdgpu_device *adev,
 	return 0;
 
 error_free_pt:
-	amdgpu_bo_unref(&pt->shadow);
 	amdgpu_bo_unref(&pt_bo);
 	return r;
 }
@@ -628,17 +536,12 @@ error_free_pt:
  */
 static void amdgpu_vm_pt_free(struct amdgpu_vm_bo_base *entry)
 {
-	struct amdgpu_bo *shadow;
-
 	if (!entry->bo)
 		return;
-	shadow = amdgpu_bo_shadowed(entry->bo);
-	if (shadow) {
-		ttm_bo_set_bulk_move(&shadow->tbo, NULL);
-		amdgpu_bo_unref(&shadow);
-	}
-	ttm_bo_set_bulk_move(&entry->bo->tbo, NULL);
+
+	amdgpu_vm_update_stats(entry, entry->bo->tbo.resource, -1);
 	entry->bo->vm_bo = NULL;
+	ttm_bo_set_bulk_move(&entry->bo->tbo, NULL);
 
 	spin_lock(&entry->vm->status_lock);
 	list_del(&entry->vm_status);
@@ -646,62 +549,55 @@ static void amdgpu_vm_pt_free(struct amdgpu_vm_bo_base *entry)
 	amdgpu_bo_unref(&entry->bo);
 }
 
-void amdgpu_vm_pt_free_work(struct work_struct *work)
+/**
+ * amdgpu_vm_pt_free_list - free PD/PT levels
+ *
+ * @adev: amdgpu device structure
+ * @params: see amdgpu_vm_update_params definition
+ *
+ * Free the page directory objects saved in the flush list
+ */
+void amdgpu_vm_pt_free_list(struct amdgpu_device *adev,
+			    struct amdgpu_vm_update_params *params)
 {
 	struct amdgpu_vm_bo_base *entry, *next;
-	struct amdgpu_vm *vm;
-	LIST_HEAD(pt_freed);
-
-	vm = container_of(work, struct amdgpu_vm, pt_free_work);
+	bool unlocked = params->unlocked;
 
-	spin_lock(&vm->status_lock);
-	list_splice_init(&vm->pt_freed, &pt_freed);
-	spin_unlock(&vm->status_lock);
+	if (list_empty(&params->tlb_flush_waitlist))
+		return;
 
-	/* flush_work in amdgpu_vm_fini ensure vm->root.bo is valid. */
-	amdgpu_bo_reserve(vm->root.bo, true);
+	/*
+	 * unlocked unmap clear page table leaves, warning to free the page entry.
+	 */
+	WARN_ON(unlocked);
 
-	list_for_each_entry_safe(entry, next, &pt_freed, vm_status)
+	list_for_each_entry_safe(entry, next, &params->tlb_flush_waitlist, vm_status)
 		amdgpu_vm_pt_free(entry);
-
-	amdgpu_bo_unreserve(vm->root.bo);
 }
 
 /**
- * amdgpu_vm_pt_free_dfs - free PD/PT levels
+ * amdgpu_vm_pt_add_list - add PD/PT level to the flush list
  *
- * @adev: amdgpu device structure
- * @vm: amdgpu vm structure
- * @start: optional cursor where to start freeing PDs/PTs
- * @unlocked: vm resv unlock status
+ * @params: parameters for the update
+ * @cursor: first PT entry to start DF search from, non NULL
  *
- * Free the page directory or page table level and all sub levels.
+ * This list will be freed after TLB flush.
  */
-static void amdgpu_vm_pt_free_dfs(struct amdgpu_device *adev,
-				  struct amdgpu_vm *vm,
-				  struct amdgpu_vm_pt_cursor *start,
-				  bool unlocked)
+static void amdgpu_vm_pt_add_list(struct amdgpu_vm_update_params *params,
+				  struct amdgpu_vm_pt_cursor *cursor)
 {
-	struct amdgpu_vm_pt_cursor cursor;
+	struct amdgpu_vm_pt_cursor seek;
 	struct amdgpu_vm_bo_base *entry;
 
-	if (unlocked) {
-		spin_lock(&vm->status_lock);
-		for_each_amdgpu_vm_pt_dfs_safe(adev, vm, start, cursor, entry)
-			list_move(&entry->vm_status, &vm->pt_freed);
-
-		if (start)
-			list_move(&start->entry->vm_status, &vm->pt_freed);
-		spin_unlock(&vm->status_lock);
-		schedule_work(&vm->pt_free_work);
-		return;
+	spin_lock(&params->vm->status_lock);
+	for_each_amdgpu_vm_pt_dfs_safe(params->adev, params->vm, cursor, seek, entry) {
+		if (entry && entry->bo)
+			list_move(&entry->vm_status, &params->tlb_flush_waitlist);
 	}
 
-	for_each_amdgpu_vm_pt_dfs_safe(adev, vm, start, cursor, entry)
-		amdgpu_vm_pt_free(entry);
-
-	if (start)
-		amdgpu_vm_pt_free(start->entry);
+	/* enter start node now */
+	list_move(&cursor->entry->vm_status, &params->tlb_flush_waitlist);
+	spin_unlock(&params->vm->status_lock);
 }
 
 /**
@@ -713,34 +609,13 @@ static void amdgpu_vm_pt_free_dfs(struct amdgpu_device *adev,
  */
 void amdgpu_vm_pt_free_root(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 {
-	amdgpu_vm_pt_free_dfs(adev, vm, NULL, false);
-}
-
-/**
- * amdgpu_vm_pt_is_root_clean - check if a root PD is clean
- *
- * @adev: amdgpu_device pointer
- * @vm: the VM to check
- *
- * Check all entries of the root PD, if any subsequent PDs are allocated,
- * it means there are page table creating and filling, and is no a clean
- * VM
- *
- * Returns:
- *	0 if this VM is clean
- */
-bool amdgpu_vm_pt_is_root_clean(struct amdgpu_device *adev,
-				struct amdgpu_vm *vm)
-{
-	enum amdgpu_vm_level root = adev->vm_manager.root_level;
-	unsigned int entries = amdgpu_vm_pt_num_entries(adev, root);
-	unsigned int i = 0;
+	struct amdgpu_vm_pt_cursor cursor;
+	struct amdgpu_vm_bo_base *entry;
 
-	for (i = 0; i < entries; i++) {
-		if (to_amdgpu_bo_vm(vm->root.bo)->entries[i].bo)
-			return false;
+	for_each_amdgpu_vm_pt_dfs_safe(adev, vm, NULL, cursor, entry) {
+		if (entry)
+			amdgpu_vm_pt_free(entry);
 	}
-	return true;
 }
 
 /**
@@ -755,11 +630,15 @@ int amdgpu_vm_pde_update(struct amdgpu_vm_update_params *params,
 			 struct amdgpu_vm_bo_base *entry)
 {
 	struct amdgpu_vm_bo_base *parent = amdgpu_vm_pt_parent(entry);
-	struct amdgpu_bo *bo = parent->bo, *pbo;
+	struct amdgpu_bo *bo, *pbo;
 	struct amdgpu_vm *vm = params->vm;
 	uint64_t pde, pt, flags;
 	unsigned int level;
 
+	if (WARN_ON(!parent))
+		return -EINVAL;
+
+	bo = parent->bo;
 	for (level = 0, pbo = bo->parent; pbo; ++level)
 		pbo = pbo->parent;
 
@@ -770,6 +649,27 @@ int amdgpu_vm_pde_update(struct amdgpu_vm_update_params *params,
 					1, 0, flags);
 }
 
+/**
+ * amdgpu_vm_pte_update_noretry_flags - Update PTE no-retry flags
+ *
+ * @adev: amdgpu_device pointer
+ * @flags: pointer to PTE flags
+ *
+ * Update PTE no-retry flags when TF is enabled.
+ */
+static void amdgpu_vm_pte_update_noretry_flags(struct amdgpu_device *adev,
+						uint64_t *flags)
+{
+	/*
+	 * Update no-retry flags with the corresponding TF
+	 * no-retry combination.
+	 */
+	if ((*flags & AMDGPU_VM_NORETRY_FLAGS) == AMDGPU_VM_NORETRY_FLAGS) {
+		*flags &= ~AMDGPU_VM_NORETRY_FLAGS;
+		*flags |= adev->gmc.noretry_flags;
+	}
+}
+
 /*
  * amdgpu_vm_pte_update_flags - figure out flags for PTE updates
  *
@@ -781,20 +681,40 @@ static void amdgpu_vm_pte_update_flags(struct amdgpu_vm_update_params *params,
 				       uint64_t pe, uint64_t addr,
 				       unsigned int count, uint32_t incr,
 				       uint64_t flags)
-
 {
+	struct amdgpu_device *adev = params->adev;
+
 	if (level != AMDGPU_VM_PTB) {
-		flags |= AMDGPU_PDE_PTE;
-		amdgpu_gmc_get_vm_pde(params->adev, level, &addr, &flags);
+		flags |= AMDGPU_PDE_PTE_FLAG(params->adev);
+		amdgpu_gmc_get_vm_pde(adev, level, &addr, &flags);
 
-	} else if (params->adev->asic_type >= CHIP_VEGA10 &&
+	} else if (adev->asic_type >= CHIP_VEGA10 &&
 		   !(flags & AMDGPU_PTE_VALID) &&
-		   !(flags & AMDGPU_PTE_PRT)) {
+		   !(flags & AMDGPU_PTE_PRT_FLAG(params->adev))) {
 
 		/* Workaround for fault priority problem on GMC9 */
 		flags |= AMDGPU_PTE_EXECUTABLE;
 	}
 
+	/*
+	 * Update no-retry flags to use the no-retry flag combination
+	 * with TF enabled. The AMDGPU_VM_NORETRY_FLAGS flag combination
+	 * does not work when TF is enabled. So, replace them with
+	 * AMDGPU_VM_NORETRY_FLAGS_TF flag combination which works for
+	 * all cases.
+	 */
+	if (level == AMDGPU_VM_PTB)
+		amdgpu_vm_pte_update_noretry_flags(adev, &flags);
+
+	/* APUs mapping system memory may need different MTYPEs on different
+	 * NUMA nodes. Only do this for contiguous ranges that can be assumed
+	 * to be on the same NUMA node.
+	 */
+	if ((flags & AMDGPU_PTE_SYSTEM) && (adev->flags & AMD_IS_APU) &&
+	    adev->gmc.gmc_funcs->override_vm_pte_flags &&
+	    num_possible_nodes() > 1 && !params->pages_addr && params->allow_override)
+		amdgpu_gmc_override_vm_pte_flags(adev, params->vm, addr, &flags);
+
 	params->vm->update_funcs->update(params, pt, pe, addr, count, incr,
 					 flags);
 }
@@ -957,7 +877,15 @@ int amdgpu_vm_ptes_update(struct amdgpu_vm_update_params *params,
 		incr = (uint64_t)AMDGPU_GPU_PAGE_SIZE << shift;
 		mask = amdgpu_vm_pt_entries_mask(adev, cursor.level);
 		pe_start = ((cursor.pfn >> shift) & mask) * 8;
-		entry_end = ((uint64_t)mask + 1) << shift;
+
+		if (cursor.level < AMDGPU_VM_PTB && params->unlocked)
+			/*
+			 * MMU notifier callback unlocked unmap huge page, leave is PDE entry,
+			 * only clear one entry. Next entry search again for PDE or PTE leave.
+			 */
+			entry_end = 1ULL << shift;
+		else
+			entry_end = ((uint64_t)mask + 1) << shift;
 		entry_end += cursor.pfn & ~(entry_end - 1);
 		entry_end = min(entry_end, end);
 
@@ -975,7 +903,7 @@ int amdgpu_vm_ptes_update(struct amdgpu_vm_update_params *params,
 			trace_amdgpu_vm_update_ptes(params, frag_start, upd_end,
 						    min(nptes, 32u), dst, incr,
 						    upd_flags,
-						    vm->task_info.tgid,
+						    vm->task_info ? vm->task_info->tgid : 0,
 						    vm->immediate.fence_context);
 			amdgpu_vm_pte_update_flags(params, to_amdgpu_bo_vm(pt),
 						   cursor.level, pe_start, dst,
@@ -1004,10 +932,8 @@ int amdgpu_vm_ptes_update(struct amdgpu_vm_update_params *params,
 			while (cursor.pfn < frag_start) {
 				/* Make sure previous mapping is freed */
 				if (cursor.entry->bo) {
-					params->table_freed = true;
-					amdgpu_vm_pt_free_dfs(adev, params->vm,
-							      &cursor,
-							      params->unlocked);
+					params->needs_flush = true;
+					amdgpu_vm_pt_add_list(params, &cursor);
 				}
 				amdgpu_vm_pt_next(adev, &cursor);
 			}
@@ -1020,3 +946,31 @@ int amdgpu_vm_ptes_update(struct amdgpu_vm_update_params *params,
 
 	return 0;
 }
+
+/**
+ * amdgpu_vm_pt_map_tables - have bo of root PD cpu accessible
+ * @adev: amdgpu device structure
+ * @vm: amdgpu vm structure
+ *
+ * make root page directory and everything below it cpu accessible.
+ */
+int amdgpu_vm_pt_map_tables(struct amdgpu_device *adev, struct amdgpu_vm *vm)
+{
+	struct amdgpu_vm_pt_cursor cursor;
+	struct amdgpu_vm_bo_base *entry;
+
+	for_each_amdgpu_vm_pt_dfs_safe(adev, vm, NULL, cursor, entry) {
+
+		struct amdgpu_bo_vm *bo;
+		int r;
+
+		if (entry->bo) {
+			bo = to_amdgpu_bo_vm(entry->bo);
+			r = vm->update_funcs->map_table(bo);
+			if (r)
+				return r;
+		}
+	}
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
index 349416e176a1..36805dcfa159 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
@@ -35,21 +35,12 @@
  */
 static int amdgpu_vm_sdma_map_table(struct amdgpu_bo_vm *table)
 {
-	int r;
-
-	r = amdgpu_ttm_alloc_gart(&table->bo.tbo);
-	if (r)
-		return r;
-
-	if (table->shadow)
-		r = amdgpu_ttm_alloc_gart(&table->shadow->tbo);
-
-	return r;
+	return amdgpu_ttm_alloc_gart(&table->bo.tbo);
 }
 
 /* Allocate a new job for @count PTE updates */
 static int amdgpu_vm_sdma_alloc_job(struct amdgpu_vm_update_params *p,
-				    unsigned int count)
+				    unsigned int count, u64 k_job_id)
 {
 	enum amdgpu_ib_pool_type pool = p->immediate ? AMDGPU_IB_POOL_IMMEDIATE
 		: AMDGPU_IB_POOL_DELAYED;
@@ -65,7 +56,7 @@ static int amdgpu_vm_sdma_alloc_job(struct amdgpu_vm_update_params *p,
 	ndw = min(ndw, AMDGPU_VM_SDMA_MAX_NUM_DW);
 
 	r = amdgpu_job_alloc_with_ib(p->adev, entity, AMDGPU_FENCE_OWNER_VM,
-				     ndw * 4, pool, &p->job);
+				     ndw * 4, pool, &p->job, k_job_id);
 	if (r)
 		return r;
 
@@ -77,31 +68,29 @@ static int amdgpu_vm_sdma_alloc_job(struct amdgpu_vm_update_params *p,
  * amdgpu_vm_sdma_prepare - prepare SDMA command submission
  *
  * @p: see amdgpu_vm_update_params definition
- * @resv: reservation object with embedded fence
- * @sync_mode: synchronization mode
+ * @sync: amdgpu_sync object with fences to wait for
+ * @k_job_id: identifier of the job, for tracing purpose
  *
  * Returns:
  * Negativ errno, 0 for success.
  */
 static int amdgpu_vm_sdma_prepare(struct amdgpu_vm_update_params *p,
-				  struct dma_resv *resv,
-				  enum amdgpu_sync_mode sync_mode)
+				  struct amdgpu_sync *sync, u64 k_job_id)
 {
-	struct amdgpu_sync sync;
 	int r;
 
-	r = amdgpu_vm_sdma_alloc_job(p, 0);
+	r = amdgpu_vm_sdma_alloc_job(p, 0, k_job_id);
 	if (r)
 		return r;
 
-	if (!resv)
+	if (!sync)
 		return 0;
 
-	amdgpu_sync_create(&sync);
-	r = amdgpu_sync_resv(p->adev, &sync, resv, sync_mode, p->vm);
-	if (!r)
-		r = amdgpu_sync_push_to_job(&sync, p->job);
-	amdgpu_sync_free(&sync);
+	r = amdgpu_sync_push_to_job(sync, p->job);
+	if (r) {
+		p->num_dw_left = 0;
+		amdgpu_job_free(p->job);
+	}
 	return r;
 }
 
@@ -126,6 +115,10 @@ static int amdgpu_vm_sdma_commit(struct amdgpu_vm_update_params *p,
 
 	WARN_ON(ib->length_dw == 0);
 	amdgpu_ring_pad_ib(ring, ib);
+
+	if (p->needs_flush)
+		atomic64_inc(&p->vm->tlb_seq);
+
 	WARN_ON(ib->length_dw > p->num_dw_left);
 	f = amdgpu_job_submit(p->job);
 
@@ -257,24 +250,21 @@ static int amdgpu_vm_sdma_update(struct amdgpu_vm_update_params *p,
 			if (r)
 				return r;
 
-			r = amdgpu_vm_sdma_alloc_job(p, count);
+			r = amdgpu_vm_sdma_alloc_job(p, count,
+						     AMDGPU_KERNEL_JOB_ID_VM_UPDATE);
 			if (r)
 				return r;
 		}
 
 		if (!p->pages_addr) {
 			/* set page commands needed */
-			if (vmbo->shadow)
-				amdgpu_vm_sdma_set_ptes(p, vmbo->shadow, pe, addr,
-							count, incr, flags);
 			amdgpu_vm_sdma_set_ptes(p, bo, pe, addr, count,
 						incr, flags);
 			return 0;
 		}
 
 		/* copy commands needed */
-		ndw -= p->adev->vm_manager.vm_pte_funcs->copy_pte_num_dw *
-			(vmbo->shadow ? 2 : 1);
+		ndw -= p->adev->vm_manager.vm_pte_funcs->copy_pte_num_dw;
 
 		/* for padding */
 		ndw -= 7;
@@ -289,8 +279,6 @@ static int amdgpu_vm_sdma_update(struct amdgpu_vm_update_params *p,
 			pte[i] |= flags;
 		}
 
-		if (vmbo->shadow)
-			amdgpu_vm_sdma_copy_ptes(p, vmbo->shadow, pe, nptes);
 		amdgpu_vm_sdma_copy_ptes(p, bo, pe, nptes);
 
 		pe += nptes * 8;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_tlb_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_tlb_fence.c
new file mode 100644
index 000000000000..5d26797356a3
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_tlb_fence.c
@@ -0,0 +1,111 @@
+// SPDX-License-Identifier: GPL-2.0 OR MIT
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include <linux/dma-fence.h>
+#include <linux/workqueue.h>
+
+#include "amdgpu.h"
+#include "amdgpu_vm.h"
+#include "amdgpu_gmc.h"
+
+struct amdgpu_tlb_fence {
+	struct dma_fence	base;
+	struct amdgpu_device	*adev;
+	struct dma_fence	*dependency;
+	struct work_struct	work;
+	spinlock_t		lock;
+	uint16_t		pasid;
+
+};
+
+static const char *amdgpu_tlb_fence_get_driver_name(struct dma_fence *fence)
+{
+	return "amdgpu tlb fence";
+}
+
+static const char *amdgpu_tlb_fence_get_timeline_name(struct dma_fence *f)
+{
+	return "amdgpu tlb timeline";
+}
+
+static void amdgpu_tlb_fence_work(struct work_struct *work)
+{
+	struct amdgpu_tlb_fence *f = container_of(work, typeof(*f), work);
+	int r;
+
+	if (f->dependency) {
+		dma_fence_wait(f->dependency, false);
+		dma_fence_put(f->dependency);
+		f->dependency = NULL;
+	}
+
+	r = amdgpu_gmc_flush_gpu_tlb_pasid(f->adev, f->pasid, 2, true, 0);
+	if (r) {
+		dev_err(f->adev->dev, "TLB flush failed for PASID %d.\n",
+			f->pasid);
+		dma_fence_set_error(&f->base, r);
+	}
+
+	dma_fence_signal(&f->base);
+	dma_fence_put(&f->base);
+}
+
+static const struct dma_fence_ops amdgpu_tlb_fence_ops = {
+	.get_driver_name = amdgpu_tlb_fence_get_driver_name,
+	.get_timeline_name = amdgpu_tlb_fence_get_timeline_name
+};
+
+void amdgpu_vm_tlb_fence_create(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+				struct dma_fence **fence)
+{
+	struct amdgpu_tlb_fence *f;
+
+	f = kmalloc(sizeof(*f), GFP_KERNEL);
+	if (!f) {
+		/*
+		 * We can't fail since the PDEs and PTEs are already updated, so
+		 * just block for the dependency and execute the TLB flush
+		 */
+		if (*fence)
+			dma_fence_wait(*fence, false);
+
+		amdgpu_gmc_flush_gpu_tlb_pasid(adev, vm->pasid, 2, true, 0);
+		*fence = dma_fence_get_stub();
+		return;
+	}
+
+	f->adev = adev;
+	f->dependency = *fence;
+	f->pasid = vm->pasid;
+	INIT_WORK(&f->work, amdgpu_tlb_fence_work);
+	spin_lock_init(&f->lock);
+
+	dma_fence_init64(&f->base, &amdgpu_tlb_fence_ops, &f->lock,
+			 vm->tlb_fence_context, atomic64_read(&vm->tlb_seq));
+
+	/* TODO: We probably need a separate wq here */
+	dma_fence_get(&f->base);
+	schedule_work(&f->work);
+
+	*fence = &f->base;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c
new file mode 100644
index 000000000000..aa78c2ee9e21
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c
@@ -0,0 +1,1018 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include <linux/firmware.h>
+#include <drm/drm_drv.h>
+
+#include "amdgpu.h"
+#include "amdgpu_ucode.h"
+#include "amdgpu_vpe.h"
+#include "amdgpu_smu.h"
+#include "soc15_common.h"
+#include "vpe_v6_1.h"
+
+#define AMDGPU_CSA_VPE_SIZE 	64
+/* VPE CSA resides in the 4th page of CSA */
+#define AMDGPU_CSA_VPE_OFFSET 	(4096 * 3)
+
+/* 1 second timeout */
+#define VPE_IDLE_TIMEOUT	msecs_to_jiffies(1000)
+
+#define VPE_MAX_DPM_LEVEL			4
+#define FIXED1_8_BITS_PER_FRACTIONAL_PART	8
+#define GET_PRATIO_INTEGER_PART(x)		((x) >> FIXED1_8_BITS_PER_FRACTIONAL_PART)
+
+static void vpe_set_ring_funcs(struct amdgpu_device *adev);
+
+static inline uint16_t div16_u16_rem(uint16_t dividend, uint16_t divisor, uint16_t *remainder)
+{
+	*remainder = dividend % divisor;
+	return dividend / divisor;
+}
+
+static inline uint16_t complete_integer_division_u16(
+	uint16_t dividend,
+	uint16_t divisor,
+	uint16_t *remainder)
+{
+	return div16_u16_rem(dividend, divisor, (uint16_t *)remainder);
+}
+
+static uint16_t vpe_u1_8_from_fraction(uint16_t numerator, uint16_t denominator)
+{
+	u16 arg1_value = numerator;
+	u16 arg2_value = denominator;
+
+	uint16_t remainder;
+
+	/* determine integer part */
+	uint16_t res_value = complete_integer_division_u16(
+		arg1_value, arg2_value, &remainder);
+
+	if (res_value > 127 /* CHAR_MAX */)
+		return 0;
+
+	/* determine fractional part */
+	{
+		unsigned int i = FIXED1_8_BITS_PER_FRACTIONAL_PART;
+
+		do {
+			remainder <<= 1;
+
+			res_value <<= 1;
+
+			if (remainder >= arg2_value) {
+				res_value |= 1;
+				remainder -= arg2_value;
+			}
+		} while (--i != 0);
+	}
+
+	/* round up LSB */
+	{
+		uint16_t summand = (remainder << 1) >= arg2_value;
+
+		if ((res_value + summand) > 32767 /* SHRT_MAX */)
+			return 0;
+
+		res_value += summand;
+	}
+
+	return res_value;
+}
+
+static uint16_t vpe_internal_get_pratio(uint16_t from_frequency, uint16_t to_frequency)
+{
+	uint16_t pratio = vpe_u1_8_from_fraction(from_frequency, to_frequency);
+
+	if (GET_PRATIO_INTEGER_PART(pratio) > 1)
+		pratio = 0;
+
+	return pratio;
+}
+
+/*
+ * VPE has 4 DPM levels from level 0 (lowerest) to 3 (highest),
+ * VPE FW will dynamically decide which level should be used according to current loading.
+ *
+ * Get VPE and SOC clocks from PM, and select the appropriate four clock values,
+ * calculate the ratios of adjusting from one clock to another.
+ * The VPE FW can then request the appropriate frequency from the PMFW.
+ */
+int amdgpu_vpe_configure_dpm(struct amdgpu_vpe *vpe)
+{
+	struct amdgpu_device *adev = vpe->ring.adev;
+	uint32_t dpm_ctl;
+
+	if (adev->pm.dpm_enabled) {
+		struct dpm_clocks clock_table = { 0 };
+		struct dpm_clock *VPEClks;
+		struct dpm_clock *SOCClks;
+		uint32_t idx;
+		uint32_t vpeclk_enalbled_num = 0;
+		uint32_t pratio_vmax_vnorm = 0, pratio_vnorm_vmid = 0, pratio_vmid_vmin = 0;
+		uint16_t pratio_vmin_freq = 0, pratio_vmid_freq = 0, pratio_vnorm_freq = 0, pratio_vmax_freq = 0;
+
+		dpm_ctl = RREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_enable));
+		dpm_ctl |= 1; /* DPM enablement */
+		WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_enable), dpm_ctl);
+
+		/* Get VPECLK and SOCCLK */
+		if (amdgpu_dpm_get_dpm_clock_table(adev, &clock_table)) {
+			dev_dbg(adev->dev, "%s: get clock failed!\n", __func__);
+			goto disable_dpm;
+		}
+
+		SOCClks = clock_table.SocClocks;
+		VPEClks = clock_table.VPEClocks;
+
+		/* Comfirm enabled vpe clk num
+		 * Enabled VPE clocks are ordered from low to high in VPEClks
+		 * The highest valid clock index+1 is the number of VPEClks
+		 */
+		for (idx = PP_SMU_NUM_VPECLK_DPM_LEVELS; idx && !vpeclk_enalbled_num; idx--)
+			if (VPEClks[idx-1].Freq)
+				vpeclk_enalbled_num = idx;
+
+		/* vpe dpm only cares 4 levels. */
+		for (idx = 0; idx < VPE_MAX_DPM_LEVEL; idx++) {
+			uint32_t soc_dpm_level;
+			uint32_t min_freq;
+
+			if (idx == 0)
+				soc_dpm_level = 0;
+			else
+				soc_dpm_level = (idx * 2) + 1;
+
+			/* clamp the max level */
+			if (soc_dpm_level > vpeclk_enalbled_num - 1)
+				soc_dpm_level = vpeclk_enalbled_num - 1;
+
+			min_freq = (SOCClks[soc_dpm_level].Freq < VPEClks[soc_dpm_level].Freq) ?
+				   SOCClks[soc_dpm_level].Freq : VPEClks[soc_dpm_level].Freq;
+
+			switch (idx) {
+			case 0:
+				pratio_vmin_freq = min_freq;
+				break;
+			case 1:
+				pratio_vmid_freq = min_freq;
+				break;
+			case 2:
+				pratio_vnorm_freq = min_freq;
+				break;
+			case 3:
+				pratio_vmax_freq = min_freq;
+				break;
+			default:
+				break;
+			}
+		}
+
+		if (pratio_vmin_freq && pratio_vmid_freq && pratio_vnorm_freq && pratio_vmax_freq) {
+			uint32_t pratio_ctl;
+
+			pratio_vmax_vnorm = (uint32_t)vpe_internal_get_pratio(pratio_vmax_freq, pratio_vnorm_freq);
+			pratio_vnorm_vmid = (uint32_t)vpe_internal_get_pratio(pratio_vnorm_freq, pratio_vmid_freq);
+			pratio_vmid_vmin = (uint32_t)vpe_internal_get_pratio(pratio_vmid_freq, pratio_vmin_freq);
+
+			pratio_ctl = pratio_vmax_vnorm | (pratio_vnorm_vmid << 9) | (pratio_vmid_vmin << 18);
+			WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_pratio), pratio_ctl);		/* PRatio */
+			WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_request_interval), 24000);	/* 1ms, unit=1/24MHz */
+			WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_decision_threshold), 1200000);	/* 50ms */
+			WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_busy_clamp_threshold), 1200000);/* 50ms */
+			WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_idle_clamp_threshold), 1200000);/* 50ms */
+			dev_dbg(adev->dev, "%s: configure vpe dpm pratio done!\n", __func__);
+		} else {
+			dev_dbg(adev->dev, "%s: invalid pratio parameters!\n", __func__);
+			goto disable_dpm;
+		}
+	}
+	return 0;
+
+disable_dpm:
+	dpm_ctl = RREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_enable));
+	dpm_ctl &= 0xfffffffe; /* Disable DPM */
+	WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_enable), dpm_ctl);
+	dev_dbg(adev->dev, "%s: disable vpe dpm\n", __func__);
+	return -EINVAL;
+}
+
+int amdgpu_vpe_psp_update_sram(struct amdgpu_device *adev)
+{
+	struct amdgpu_firmware_info ucode = {
+		.ucode_id = AMDGPU_UCODE_ID_VPE,
+		.mc_addr = adev->vpe.cmdbuf_gpu_addr,
+		.ucode_size = 8,
+	};
+
+	return psp_execute_ip_fw_load(&adev->psp, &ucode);
+}
+
+int amdgpu_vpe_init_microcode(struct amdgpu_vpe *vpe)
+{
+	struct amdgpu_device *adev = vpe->ring.adev;
+	const struct vpe_firmware_header_v1_0 *vpe_hdr;
+	char fw_prefix[32];
+	int ret;
+
+	amdgpu_ucode_ip_version_decode(adev, VPE_HWIP, fw_prefix, sizeof(fw_prefix));
+	ret = amdgpu_ucode_request(adev, &adev->vpe.fw, AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s.bin", fw_prefix);
+	if (ret)
+		goto out;
+
+	vpe_hdr = (const struct vpe_firmware_header_v1_0 *)adev->vpe.fw->data;
+	adev->vpe.fw_version = le32_to_cpu(vpe_hdr->header.ucode_version);
+	adev->vpe.feature_version = le32_to_cpu(vpe_hdr->ucode_feature_version);
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		struct amdgpu_firmware_info *info;
+
+		info = &adev->firmware.ucode[AMDGPU_UCODE_ID_VPE_CTX];
+		info->ucode_id = AMDGPU_UCODE_ID_VPE_CTX;
+		info->fw = adev->vpe.fw;
+		adev->firmware.fw_size +=
+			ALIGN(le32_to_cpu(vpe_hdr->ctx_ucode_size_bytes), PAGE_SIZE);
+
+		info = &adev->firmware.ucode[AMDGPU_UCODE_ID_VPE_CTL];
+		info->ucode_id = AMDGPU_UCODE_ID_VPE_CTL;
+		info->fw = adev->vpe.fw;
+		adev->firmware.fw_size +=
+			ALIGN(le32_to_cpu(vpe_hdr->ctl_ucode_size_bytes), PAGE_SIZE);
+	}
+
+	return 0;
+out:
+	dev_err(adev->dev, "fail to initialize vpe microcode\n");
+	release_firmware(adev->vpe.fw);
+	adev->vpe.fw = NULL;
+	return ret;
+}
+
+int amdgpu_vpe_ring_init(struct amdgpu_vpe *vpe)
+{
+	struct amdgpu_device *adev = container_of(vpe, struct amdgpu_device, vpe);
+	struct amdgpu_ring *ring = &vpe->ring;
+	int ret;
+
+	ring->ring_obj = NULL;
+	ring->use_doorbell = true;
+	ring->vm_hub = AMDGPU_MMHUB0(0);
+	ring->doorbell_index = (adev->doorbell_index.vpe_ring << 1);
+	snprintf(ring->name, 4, "vpe");
+
+	ret = amdgpu_ring_init(adev, ring, 1024, &vpe->trap_irq, 0,
+			     AMDGPU_RING_PRIO_DEFAULT, NULL);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+int amdgpu_vpe_ring_fini(struct amdgpu_vpe *vpe)
+{
+	amdgpu_ring_fini(&vpe->ring);
+
+	return 0;
+}
+
+static int vpe_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+
+	switch (amdgpu_ip_version(adev, VPE_HWIP, 0)) {
+	case IP_VERSION(6, 1, 0):
+	case IP_VERSION(6, 1, 3):
+		vpe_v6_1_set_funcs(vpe);
+		break;
+	case IP_VERSION(6, 1, 1):
+		vpe_v6_1_set_funcs(vpe);
+		vpe->collaborate_mode = true;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	vpe_set_ring_funcs(adev);
+	vpe_set_regs(vpe);
+
+	dev_info(adev->dev, "VPE: collaborate mode %s", vpe->collaborate_mode ? "true" : "false");
+
+	return 0;
+}
+
+static bool vpe_need_dpm0_at_power_down(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, VPE_HWIP, 0)) {
+	case IP_VERSION(6, 1, 1):
+		return adev->pm.fw_version < 0x0a640500;
+	default:
+		return false;
+	}
+}
+
+static int vpe_get_dpm_level(struct amdgpu_device *adev)
+{
+	struct amdgpu_vpe *vpe = &adev->vpe;
+
+	if (!adev->pm.dpm_enabled)
+		return 0;
+
+	return RREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_request_lv));
+}
+
+static void vpe_idle_work_handler(struct work_struct *work)
+{
+	struct amdgpu_device *adev =
+		container_of(work, struct amdgpu_device, vpe.idle_work.work);
+	unsigned int fences = 0;
+
+	fences += amdgpu_fence_count_emitted(&adev->vpe.ring);
+	if (fences)
+		goto reschedule;
+
+	if (vpe_need_dpm0_at_power_down(adev) && vpe_get_dpm_level(adev) != 0)
+		goto reschedule;
+
+	amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE, AMD_PG_STATE_GATE);
+	return;
+
+reschedule:
+	schedule_delayed_work(&adev->vpe.idle_work, VPE_IDLE_TIMEOUT);
+}
+
+static int vpe_common_init(struct amdgpu_vpe *vpe)
+{
+	struct amdgpu_device *adev = container_of(vpe, struct amdgpu_device, vpe);
+	int r;
+
+	r = amdgpu_bo_create_kernel(adev, PAGE_SIZE, PAGE_SIZE,
+				    AMDGPU_GEM_DOMAIN_GTT,
+				    &adev->vpe.cmdbuf_obj,
+				    &adev->vpe.cmdbuf_gpu_addr,
+				    (void **)&adev->vpe.cmdbuf_cpu_addr);
+	if (r) {
+		dev_err(adev->dev, "VPE: failed to allocate cmdbuf bo %d\n", r);
+		return r;
+	}
+
+	vpe->context_started = false;
+	INIT_DELAYED_WORK(&adev->vpe.idle_work, vpe_idle_work_handler);
+
+	return 0;
+}
+
+static int vpe_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+	int ret;
+
+	ret = vpe_common_init(vpe);
+	if (ret)
+		goto out;
+
+	ret = vpe_irq_init(vpe);
+	if (ret)
+		goto out;
+
+	ret = vpe_ring_init(vpe);
+	if (ret)
+		goto out;
+
+	ret = vpe_init_microcode(vpe);
+	if (ret)
+		goto out;
+
+	adev->vpe.supported_reset =
+		 amdgpu_get_soft_full_reset_mask(&adev->vpe.ring);
+	if (!amdgpu_sriov_vf(adev))
+		adev->vpe.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+	ret = amdgpu_vpe_sysfs_reset_mask_init(adev);
+	if (ret)
+		goto out;
+out:
+	return ret;
+}
+
+static int vpe_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+
+	release_firmware(vpe->fw);
+	vpe->fw = NULL;
+
+	amdgpu_vpe_sysfs_reset_mask_fini(adev);
+	vpe_ring_fini(vpe);
+
+	amdgpu_bo_free_kernel(&adev->vpe.cmdbuf_obj,
+			      &adev->vpe.cmdbuf_gpu_addr,
+			      (void **)&adev->vpe.cmdbuf_cpu_addr);
+
+	return 0;
+}
+
+static int vpe_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+	int ret;
+
+	/* Power on VPE */
+	ret = amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE,
+						     AMD_PG_STATE_UNGATE);
+	if (ret)
+		return ret;
+
+	ret = vpe_load_microcode(vpe);
+	if (ret)
+		return ret;
+
+	ret = vpe_ring_start(vpe);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static int vpe_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+
+	cancel_delayed_work_sync(&adev->vpe.idle_work);
+
+	vpe_ring_stop(vpe);
+
+	/* Power off VPE */
+	amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE, AMD_PG_STATE_GATE);
+
+	return 0;
+}
+
+static int vpe_suspend(struct amdgpu_ip_block *ip_block)
+{
+	return vpe_hw_fini(ip_block);
+}
+
+static int vpe_resume(struct amdgpu_ip_block *ip_block)
+{
+	return vpe_hw_init(ip_block);
+}
+
+static void vpe_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
+{
+	int i;
+
+	for (i = 0; i < count; i++)
+		if (i == 0)
+			amdgpu_ring_write(ring, ring->funcs->nop |
+				VPE_CMD_NOP_HEADER_COUNT(count - 1));
+		else
+			amdgpu_ring_write(ring, ring->funcs->nop);
+}
+
+static uint64_t vpe_get_csa_mc_addr(struct amdgpu_ring *ring, uint32_t vmid)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t index = 0;
+	uint64_t csa_mc_addr;
+
+	if (amdgpu_sriov_vf(adev) || vmid == 0 || !adev->gfx.mcbp)
+		return 0;
+
+	csa_mc_addr = amdgpu_csa_vaddr(adev) + AMDGPU_CSA_VPE_OFFSET +
+		      index * AMDGPU_CSA_VPE_SIZE;
+
+	return csa_mc_addr;
+}
+
+static void vpe_ring_emit_pred_exec(struct amdgpu_ring *ring,
+				    uint32_t device_select,
+				    uint32_t exec_count)
+{
+	if (!ring->adev->vpe.collaborate_mode)
+		return;
+
+	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_PRED_EXE, 0) |
+				(device_select << 16));
+	amdgpu_ring_write(ring, exec_count & 0x1fff);
+}
+
+static void vpe_ring_emit_ib(struct amdgpu_ring *ring,
+			     struct amdgpu_job *job,
+			     struct amdgpu_ib *ib,
+			     uint32_t flags)
+{
+	uint32_t vmid = AMDGPU_JOB_GET_VMID(job);
+	uint64_t csa_mc_addr = vpe_get_csa_mc_addr(ring, vmid);
+
+	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_INDIRECT, 0) |
+				VPE_CMD_INDIRECT_HEADER_VMID(vmid & 0xf));
+
+	/* base must be 32 byte aligned */
+	amdgpu_ring_write(ring, ib->gpu_addr & 0xffffffe0);
+	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
+	amdgpu_ring_write(ring, ib->length_dw);
+	amdgpu_ring_write(ring, lower_32_bits(csa_mc_addr));
+	amdgpu_ring_write(ring, upper_32_bits(csa_mc_addr));
+}
+
+static void vpe_ring_emit_fence(struct amdgpu_ring *ring, uint64_t addr,
+				uint64_t seq, unsigned int flags)
+{
+	int i = 0;
+
+	do {
+		/* write the fence */
+		amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_FENCE, 0));
+		/* zero in first two bits */
+		WARN_ON_ONCE(addr & 0x3);
+		amdgpu_ring_write(ring, lower_32_bits(addr));
+		amdgpu_ring_write(ring, upper_32_bits(addr));
+		amdgpu_ring_write(ring, i == 0 ? lower_32_bits(seq) : upper_32_bits(seq));
+		addr += 4;
+	} while ((flags & AMDGPU_FENCE_FLAG_64BIT) && (i++ < 1));
+
+	if (flags & AMDGPU_FENCE_FLAG_INT) {
+		/* generate an interrupt */
+		amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_TRAP, 0));
+		amdgpu_ring_write(ring, 0);
+	}
+
+}
+
+static void vpe_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
+{
+	uint32_t seq = ring->fence_drv.sync_seq;
+	uint64_t addr = ring->fence_drv.gpu_addr;
+
+	vpe_ring_emit_pred_exec(ring, 0, 6);
+
+	/* wait for idle */
+	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_POLL_REGMEM,
+				VPE_POLL_REGMEM_SUBOP_REGMEM) |
+				VPE_CMD_POLL_REGMEM_HEADER_FUNC(3) | /* equal */
+				VPE_CMD_POLL_REGMEM_HEADER_MEM(1));
+	amdgpu_ring_write(ring, addr & 0xfffffffc);
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	amdgpu_ring_write(ring, seq); /* reference */
+	amdgpu_ring_write(ring, 0xffffffff); /* mask */
+	amdgpu_ring_write(ring, VPE_CMD_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) |
+				VPE_CMD_POLL_REGMEM_DW5_INTERVAL(4));
+}
+
+static void vpe_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg, uint32_t val)
+{
+	vpe_ring_emit_pred_exec(ring, 0, 3);
+
+	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_REG_WRITE, 0));
+	amdgpu_ring_write(ring,	reg << 2);
+	amdgpu_ring_write(ring, val);
+}
+
+static void vpe_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
+				   uint32_t val, uint32_t mask)
+{
+	vpe_ring_emit_pred_exec(ring, 0, 6);
+
+	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_POLL_REGMEM,
+				VPE_POLL_REGMEM_SUBOP_REGMEM) |
+				VPE_CMD_POLL_REGMEM_HEADER_FUNC(3) | /* equal */
+				VPE_CMD_POLL_REGMEM_HEADER_MEM(0));
+	amdgpu_ring_write(ring, reg << 2);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, val); /* reference */
+	amdgpu_ring_write(ring, mask); /* mask */
+	amdgpu_ring_write(ring, VPE_CMD_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) |
+				VPE_CMD_POLL_REGMEM_DW5_INTERVAL(10));
+}
+
+static void vpe_ring_emit_vm_flush(struct amdgpu_ring *ring, unsigned int vmid,
+				   uint64_t pd_addr)
+{
+	amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
+}
+
+static unsigned int vpe_ring_init_cond_exec(struct amdgpu_ring *ring,
+					    uint64_t addr)
+{
+	unsigned int ret;
+
+	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_COND_EXE, 0));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	amdgpu_ring_write(ring, 1);
+	ret = ring->wptr & ring->buf_mask;
+	amdgpu_ring_write(ring, 0);
+
+	return ret;
+}
+
+static int vpe_ring_preempt_ib(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+	uint32_t preempt_reg = vpe->regs.queue0_preempt;
+	int i, r = 0;
+
+	/* assert preemption condition */
+	amdgpu_ring_set_preempt_cond_exec(ring, false);
+
+	/* emit the trailing fence */
+	ring->trail_seq += 1;
+	amdgpu_ring_alloc(ring, 10);
+	vpe_ring_emit_fence(ring, ring->trail_fence_gpu_addr, ring->trail_seq, 0);
+	amdgpu_ring_commit(ring);
+
+	/* assert IB preemption */
+	WREG32(vpe_get_reg_offset(vpe, ring->me, preempt_reg), 1);
+
+	/* poll the trailing fence */
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (ring->trail_seq ==
+		    le32_to_cpu(*(ring->trail_fence_cpu_addr)))
+			break;
+		udelay(1);
+	}
+
+	if (i >= adev->usec_timeout) {
+		r = -EINVAL;
+		dev_err(adev->dev, "ring %d failed to be preempted\n", ring->idx);
+	}
+
+	/* deassert IB preemption */
+	WREG32(vpe_get_reg_offset(vpe, ring->me, preempt_reg), 0);
+
+	/* deassert the preemption condition */
+	amdgpu_ring_set_preempt_cond_exec(ring, true);
+
+	return r;
+}
+
+static int vpe_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+				     enum amd_clockgating_state state)
+{
+	return 0;
+}
+
+static int vpe_set_powergating_state(struct amdgpu_ip_block *ip_block,
+				     enum amd_powergating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+
+	if (!adev->pm.dpm_enabled)
+		dev_err(adev->dev, "Without PM, cannot support powergating\n");
+
+	dev_dbg(adev->dev, "%s: %s!\n", __func__, (state == AMD_PG_STATE_GATE) ? "GATE":"UNGATE");
+
+	if (state == AMD_PG_STATE_GATE) {
+		amdgpu_dpm_enable_vpe(adev, false);
+		vpe->context_started = false;
+	} else {
+		amdgpu_dpm_enable_vpe(adev, true);
+	}
+
+	return 0;
+}
+
+static uint64_t vpe_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+	uint64_t rptr;
+
+	if (ring->use_doorbell) {
+		rptr = atomic64_read((atomic64_t *)ring->rptr_cpu_addr);
+		dev_dbg(adev->dev, "rptr/doorbell before shift == 0x%016llx\n", rptr);
+	} else {
+		rptr = RREG32(vpe_get_reg_offset(vpe, ring->me, vpe->regs.queue0_rb_rptr_hi));
+		rptr = rptr << 32;
+		rptr |= RREG32(vpe_get_reg_offset(vpe, ring->me, vpe->regs.queue0_rb_rptr_lo));
+		dev_dbg(adev->dev, "rptr before shift [%i] == 0x%016llx\n", ring->me, rptr);
+	}
+
+	return (rptr >> 2);
+}
+
+static uint64_t vpe_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+	uint64_t wptr;
+
+	if (ring->use_doorbell) {
+		wptr = atomic64_read((atomic64_t *)ring->wptr_cpu_addr);
+		dev_dbg(adev->dev, "wptr/doorbell before shift == 0x%016llx\n", wptr);
+	} else {
+		wptr = RREG32(vpe_get_reg_offset(vpe, ring->me, vpe->regs.queue0_rb_wptr_hi));
+		wptr = wptr << 32;
+		wptr |= RREG32(vpe_get_reg_offset(vpe, ring->me, vpe->regs.queue0_rb_wptr_lo));
+		dev_dbg(adev->dev, "wptr before shift [%i] == 0x%016llx\n", ring->me, wptr);
+	}
+
+	return (wptr >> 2);
+}
+
+static void vpe_ring_set_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+
+	if (ring->use_doorbell) {
+		dev_dbg(adev->dev, "Using doorbell, \
+			wptr_offs == 0x%08x, \
+			lower_32_bits(ring->wptr) << 2 == 0x%08x, \
+			upper_32_bits(ring->wptr) << 2 == 0x%08x\n",
+			ring->wptr_offs,
+			lower_32_bits(ring->wptr << 2),
+			upper_32_bits(ring->wptr << 2));
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, ring->wptr << 2);
+		WDOORBELL64(ring->doorbell_index, ring->wptr << 2);
+		if (vpe->collaborate_mode)
+			WDOORBELL64(ring->doorbell_index + 4, ring->wptr << 2);
+	} else {
+		int i;
+
+		for (i = 0; i < vpe->num_instances; i++) {
+			dev_dbg(adev->dev, "Not using doorbell, \
+				regVPEC_QUEUE0_RB_WPTR == 0x%08x, \
+				regVPEC_QUEUE0_RB_WPTR_HI == 0x%08x\n",
+				lower_32_bits(ring->wptr << 2),
+				upper_32_bits(ring->wptr << 2));
+			WREG32(vpe_get_reg_offset(vpe, i, vpe->regs.queue0_rb_wptr_lo),
+			       lower_32_bits(ring->wptr << 2));
+			WREG32(vpe_get_reg_offset(vpe, i, vpe->regs.queue0_rb_wptr_hi),
+			       upper_32_bits(ring->wptr << 2));
+		}
+	}
+}
+
+static int vpe_ring_test_ring(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	const uint32_t test_pattern = 0xdeadbeef;
+	uint32_t index, i;
+	uint64_t wb_addr;
+	int ret;
+
+	ret = amdgpu_device_wb_get(adev, &index);
+	if (ret) {
+		dev_err(adev->dev, "(%d) failed to allocate wb slot\n", ret);
+		return ret;
+	}
+
+	adev->wb.wb[index] = 0;
+	wb_addr = adev->wb.gpu_addr + (index * 4);
+
+	ret = amdgpu_ring_alloc(ring, 4);
+	if (ret) {
+		dev_err(adev->dev, "amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, ret);
+		goto out;
+	}
+
+	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_FENCE, 0));
+	amdgpu_ring_write(ring, lower_32_bits(wb_addr));
+	amdgpu_ring_write(ring, upper_32_bits(wb_addr));
+	amdgpu_ring_write(ring, test_pattern);
+	amdgpu_ring_commit(ring);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (le32_to_cpu(adev->wb.wb[index]) == test_pattern)
+			goto out;
+		udelay(1);
+	}
+
+	ret = -ETIMEDOUT;
+out:
+	amdgpu_device_wb_free(adev, index);
+
+	return ret;
+}
+
+static int vpe_ring_test_ib(struct amdgpu_ring *ring, long timeout)
+{
+	struct amdgpu_device *adev = ring->adev;
+	const uint32_t test_pattern = 0xdeadbeef;
+	struct amdgpu_ib ib = {};
+	struct dma_fence *f = NULL;
+	uint32_t index;
+	uint64_t wb_addr;
+	int ret;
+
+	ret = amdgpu_device_wb_get(adev, &index);
+	if (ret) {
+		dev_err(adev->dev, "(%d) failed to allocate wb slot\n", ret);
+		return ret;
+	}
+
+	adev->wb.wb[index] = 0;
+	wb_addr = adev->wb.gpu_addr + (index * 4);
+
+	ret = amdgpu_ib_get(adev, NULL, 256, AMDGPU_IB_POOL_DIRECT, &ib);
+	if (ret)
+		goto err0;
+
+	ib.ptr[0] = VPE_CMD_HEADER(VPE_CMD_OPCODE_FENCE, 0);
+	ib.ptr[1] = lower_32_bits(wb_addr);
+	ib.ptr[2] = upper_32_bits(wb_addr);
+	ib.ptr[3] = test_pattern;
+	ib.ptr[4] = VPE_CMD_HEADER(VPE_CMD_OPCODE_NOP, 0);
+	ib.ptr[5] = VPE_CMD_HEADER(VPE_CMD_OPCODE_NOP, 0);
+	ib.ptr[6] = VPE_CMD_HEADER(VPE_CMD_OPCODE_NOP, 0);
+	ib.ptr[7] = VPE_CMD_HEADER(VPE_CMD_OPCODE_NOP, 0);
+	ib.length_dw = 8;
+
+	ret = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f);
+	if (ret)
+		goto err1;
+
+	ret = dma_fence_wait_timeout(f, false, timeout);
+	if (ret <= 0) {
+		ret = ret ? : -ETIMEDOUT;
+		goto err1;
+	}
+
+	ret = (le32_to_cpu(adev->wb.wb[index]) == test_pattern) ? 0 : -EINVAL;
+
+err1:
+	amdgpu_ib_free(&ib, NULL);
+	dma_fence_put(f);
+err0:
+	amdgpu_device_wb_free(adev, index);
+
+	return ret;
+}
+
+static void vpe_ring_begin_use(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+
+	cancel_delayed_work_sync(&adev->vpe.idle_work);
+
+	/* Power on VPE and notify VPE of new context  */
+	if (!vpe->context_started) {
+		uint32_t context_notify;
+
+		/* Power on VPE */
+		amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE, AMD_PG_STATE_UNGATE);
+
+		/* Indicates that a job from a new context has been submitted. */
+		context_notify = RREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.context_indicator));
+		if ((context_notify & 0x1) == 0)
+			context_notify |= 0x1;
+		else
+			context_notify &= ~(0x1);
+		WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.context_indicator), context_notify);
+		vpe->context_started = true;
+	}
+}
+
+static void vpe_ring_end_use(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	schedule_delayed_work(&adev->vpe.idle_work, VPE_IDLE_TIMEOUT);
+}
+
+static int vpe_ring_reset(struct amdgpu_ring *ring,
+			  unsigned int vmid,
+			  struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int r;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	r = amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE,
+						   AMD_PG_STATE_GATE);
+	if (r)
+		return r;
+	r = amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE,
+						   AMD_PG_STATE_UNGATE);
+	if (r)
+		return r;
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static ssize_t amdgpu_get_vpe_reset_mask(struct device *dev,
+						struct device_attribute *attr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	if (!adev)
+		return -ENODEV;
+
+	return amdgpu_show_reset_mask(buf, adev->vpe.supported_reset);
+}
+
+static DEVICE_ATTR(vpe_reset_mask, 0444,
+		   amdgpu_get_vpe_reset_mask, NULL);
+
+int amdgpu_vpe_sysfs_reset_mask_init(struct amdgpu_device *adev)
+{
+	int r = 0;
+
+	if (adev->vpe.num_instances) {
+		r = device_create_file(adev->dev, &dev_attr_vpe_reset_mask);
+		if (r)
+			return r;
+	}
+
+	return r;
+}
+
+void amdgpu_vpe_sysfs_reset_mask_fini(struct amdgpu_device *adev)
+{
+	if (adev->dev->kobj.sd) {
+		if (adev->vpe.num_instances)
+			device_remove_file(adev->dev, &dev_attr_vpe_reset_mask);
+	}
+}
+
+static const struct amdgpu_ring_funcs vpe_ring_funcs = {
+	.type = AMDGPU_RING_TYPE_VPE,
+	.align_mask = 0xf,
+	.nop = VPE_CMD_HEADER(VPE_CMD_OPCODE_NOP, 0),
+	.support_64bit_ptrs = true,
+	.get_rptr = vpe_ring_get_rptr,
+	.get_wptr = vpe_ring_get_wptr,
+	.set_wptr = vpe_ring_set_wptr,
+	.emit_frame_size =
+		5 + /* vpe_ring_init_cond_exec */
+		6 + /* vpe_ring_emit_pipeline_sync */
+		10 + 10 + 10 + /* vpe_ring_emit_fence */
+		/* vpe_ring_emit_vm_flush */
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 6,
+	.emit_ib_size = 7 + 6,
+	.emit_ib = vpe_ring_emit_ib,
+	.emit_pipeline_sync = vpe_ring_emit_pipeline_sync,
+	.emit_fence = vpe_ring_emit_fence,
+	.emit_vm_flush = vpe_ring_emit_vm_flush,
+	.emit_wreg = vpe_ring_emit_wreg,
+	.emit_reg_wait = vpe_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.insert_nop = vpe_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.test_ring = vpe_ring_test_ring,
+	.test_ib = vpe_ring_test_ib,
+	.init_cond_exec = vpe_ring_init_cond_exec,
+	.preempt_ib = vpe_ring_preempt_ib,
+	.begin_use = vpe_ring_begin_use,
+	.end_use = vpe_ring_end_use,
+	.reset = vpe_ring_reset,
+};
+
+static void vpe_set_ring_funcs(struct amdgpu_device *adev)
+{
+	adev->vpe.ring.funcs = &vpe_ring_funcs;
+}
+
+const struct amd_ip_funcs vpe_ip_funcs = {
+	.name = "vpe_v6_1",
+	.early_init = vpe_early_init,
+	.sw_init = vpe_sw_init,
+	.sw_fini = vpe_sw_fini,
+	.hw_init = vpe_hw_init,
+	.hw_fini = vpe_hw_fini,
+	.suspend = vpe_suspend,
+	.resume = vpe_resume,
+	.set_clockgating_state = vpe_set_clockgating_state,
+	.set_powergating_state = vpe_set_powergating_state,
+};
+
+const struct amdgpu_ip_block_version vpe_v6_1_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_VPE,
+	.major = 6,
+	.minor = 1,
+	.rev = 0,
+	.funcs = &vpe_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h
new file mode 100644
index 000000000000..695da740a97e
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h
@@ -0,0 +1,111 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef __AMDGPU_VPE_H__
+#define __AMDGPU_VPE_H__
+
+#include "amdgpu_ring.h"
+#include "amdgpu_irq.h"
+#include "vpe_6_1_fw_if.h"
+
+#define AMDGPU_MAX_VPE_INSTANCES 2
+
+struct amdgpu_vpe;
+
+struct vpe_funcs {
+	uint32_t (*get_reg_offset)(struct amdgpu_vpe *vpe, uint32_t inst, uint32_t offset);
+	int (*set_regs)(struct amdgpu_vpe *vpe);
+	int (*irq_init)(struct amdgpu_vpe *vpe);
+	int (*init_microcode)(struct amdgpu_vpe *vpe);
+	int (*load_microcode)(struct amdgpu_vpe *vpe);
+	int (*ring_init)(struct amdgpu_vpe *vpe);
+	int (*ring_start)(struct amdgpu_vpe *vpe);
+	int (*ring_stop)(struct amdgpu_vpe *vpe);
+	int (*ring_fini)(struct amdgpu_vpe *vpe);
+};
+
+struct vpe_regs {
+	uint32_t queue0_rb_rptr_lo;
+	uint32_t queue0_rb_rptr_hi;
+	uint32_t queue0_rb_wptr_lo;
+	uint32_t queue0_rb_wptr_hi;
+	uint32_t queue0_preempt;
+
+	uint32_t dpm_enable;
+	uint32_t dpm_pratio;
+	uint32_t dpm_request_interval;
+	uint32_t dpm_decision_threshold;
+	uint32_t dpm_busy_clamp_threshold;
+	uint32_t dpm_idle_clamp_threshold;
+	uint32_t dpm_request_lv;
+	uint32_t context_indicator;
+};
+
+struct amdgpu_vpe {
+	struct amdgpu_ring		ring;
+	struct amdgpu_irq_src		trap_irq;
+
+	const struct vpe_funcs		*funcs;
+	struct vpe_regs			regs;
+
+	const struct firmware		*fw;
+	uint32_t			fw_version;
+	uint32_t			feature_version;
+
+	struct amdgpu_bo		*cmdbuf_obj;
+	uint64_t			cmdbuf_gpu_addr;
+	uint32_t			*cmdbuf_cpu_addr;
+	struct delayed_work		idle_work;
+	bool				context_started;
+
+	uint32_t			num_instances;
+	bool				collaborate_mode;
+	uint32_t			supported_reset;
+};
+
+int amdgpu_vpe_psp_update_sram(struct amdgpu_device *adev);
+int amdgpu_vpe_init_microcode(struct amdgpu_vpe *vpe);
+int amdgpu_vpe_ring_init(struct amdgpu_vpe *vpe);
+int amdgpu_vpe_ring_fini(struct amdgpu_vpe *vpe);
+int amdgpu_vpe_configure_dpm(struct amdgpu_vpe *vpe);
+void amdgpu_vpe_sysfs_reset_mask_fini(struct amdgpu_device *adev);
+int amdgpu_vpe_sysfs_reset_mask_init(struct amdgpu_device *adev);
+
+#define vpe_ring_init(vpe) ((vpe)->funcs->ring_init ? (vpe)->funcs->ring_init((vpe)) : 0)
+#define vpe_ring_start(vpe) ((vpe)->funcs->ring_start ? (vpe)->funcs->ring_start((vpe)) : 0)
+#define vpe_ring_stop(vpe) ((vpe)->funcs->ring_stop ? (vpe)->funcs->ring_stop((vpe)) : 0)
+#define vpe_ring_fini(vpe) ((vpe)->funcs->ring_fini ? (vpe)->funcs->ring_fini((vpe)) : 0)
+
+#define vpe_get_reg_offset(vpe, inst, offset) \
+		((vpe)->funcs->get_reg_offset ? (vpe)->funcs->get_reg_offset((vpe), (inst), (offset)) : 0)
+#define vpe_set_regs(vpe) \
+		((vpe)->funcs->set_regs ? (vpe)->funcs->set_regs((vpe)) : 0)
+#define vpe_irq_init(vpe) \
+		((vpe)->funcs->irq_init ? (vpe)->funcs->irq_init((vpe)) : 0)
+#define vpe_init_microcode(vpe) \
+		((vpe)->funcs->init_microcode ? (vpe)->funcs->init_microcode((vpe)) : 0)
+#define vpe_load_microcode(vpe) \
+		((vpe)->funcs->load_microcode ? (vpe)->funcs->load_microcode((vpe)) : 0)
+
+extern const struct amdgpu_ip_block_version vpe_v6_1_ip_block;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
index 43d6a9d6a538..9d934c07fa6b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
@@ -24,13 +24,15 @@
 
 #include <linux/dma-mapping.h>
 #include <drm/ttm/ttm_range_manager.h>
+#include <drm/drm_drv.h>
 
 #include "amdgpu.h"
 #include "amdgpu_vm.h"
 #include "amdgpu_res_cursor.h"
-#include "amdgpu_atomfirmware.h"
 #include "atom.h"
 
+#define AMDGPU_MAX_SG_SEGMENT_SIZE	(2UL << 30)
+
 struct amdgpu_vram_reservation {
 	u64 start;
 	u64 size;
@@ -77,7 +79,16 @@ static inline bool amdgpu_is_vram_mgr_blocks_contiguous(struct list_head *head)
 	return true;
 }
 
+static inline u64 amdgpu_vram_mgr_blocks_size(struct list_head *head)
+{
+	struct drm_buddy_block *block;
+	u64 size = 0;
+
+	list_for_each_entry(block, head, link)
+		size += amdgpu_vram_mgr_block_size(block);
 
+	return size;
+}
 
 /**
  * DOC: mem_info_vram_total
@@ -212,8 +223,26 @@ static struct attribute *amdgpu_vram_mgr_attributes[] = {
 	NULL
 };
 
+static umode_t amdgpu_vram_attrs_is_visible(struct kobject *kobj,
+					    struct attribute *attr, int i)
+{
+	struct device *dev = kobj_to_dev(kobj);
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	if (attr == &dev_attr_mem_info_vram_vendor.attr &&
+	    !adev->gmc.vram_vendor)
+		return 0;
+
+	if (!ttm_resource_manager_used(&adev->mman.vram_mgr.manager))
+		return 0;
+
+	return attr->mode;
+}
+
 const struct attribute_group amdgpu_vram_mgr_attr_group = {
-	.attrs = amdgpu_vram_mgr_attributes
+	.attrs = amdgpu_vram_mgr_attributes,
+	.is_visible = amdgpu_vram_attrs_is_visible
 };
 
 /**
@@ -370,6 +399,35 @@ out:
 	return ret;
 }
 
+int amdgpu_vram_mgr_query_address_block_info(struct amdgpu_vram_mgr *mgr,
+			uint64_t address, struct amdgpu_vram_block_info *info)
+{
+	struct amdgpu_vram_mgr_resource *vres;
+	struct drm_buddy_block *block;
+	u64 start, size;
+	int ret = -ENOENT;
+
+	mutex_lock(&mgr->lock);
+	list_for_each_entry(vres, &mgr->allocated_vres_list, vres_node) {
+		list_for_each_entry(block, &vres->blocks, link) {
+			start = amdgpu_vram_mgr_block_start(block);
+			size = amdgpu_vram_mgr_block_size(block);
+			if ((start <= address) && (address < (start + size))) {
+				info->start = start;
+				info->size = size;
+				memcpy(&info->task, &vres->task, sizeof(vres->task));
+				ret = 0;
+				goto out;
+			}
+		}
+	}
+
+out:
+	mutex_unlock(&mgr->lock);
+
+	return ret;
+}
+
 /**
  * amdgpu_vram_mgr_new - allocate new ranges
  *
@@ -385,18 +443,20 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 			       const struct ttm_place *place,
 			       struct ttm_resource **res)
 {
-	u64 vis_usage = 0, max_bytes, cur_size, min_block_size;
 	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
 	struct amdgpu_device *adev = to_amdgpu_device(mgr);
+	struct amdgpu_bo *bo = ttm_to_amdgpu_bo(tbo);
+	u64 vis_usage = 0, max_bytes, min_block_size;
 	struct amdgpu_vram_mgr_resource *vres;
 	u64 size, remaining_size, lpfn, fpfn;
+	unsigned int adjust_dcc_size = 0;
 	struct drm_buddy *mm = &mgr->mm;
 	struct drm_buddy_block *block;
 	unsigned long pages_per_block;
 	int r;
 
 	lpfn = (u64)place->lpfn << PAGE_SHIFT;
-	if (!lpfn)
+	if (!lpfn || lpfn > man->size)
 		lpfn = man->size;
 
 	fpfn = (u64)place->fpfn << PAGE_SHIFT;
@@ -405,7 +465,7 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 	if (tbo->type != ttm_bo_type_kernel)
 		max_bytes -= AMDGPU_VM_RESERVED_VRAM;
 
-	if (place->flags & TTM_PL_FLAG_CONTIGUOUS) {
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS) {
 		pages_per_block = ~0ul;
 	} else {
 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
@@ -414,7 +474,7 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 		/* default to 2MB */
 		pages_per_block = 2UL << (20UL - PAGE_SHIFT);
 #endif
-		pages_per_block = max_t(uint32_t, pages_per_block,
+		pages_per_block = max_t(u32, pages_per_block,
 					tbo->page_alignment);
 	}
 
@@ -435,11 +495,29 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 	if (place->flags & TTM_PL_FLAG_TOPDOWN)
 		vres->flags |= DRM_BUDDY_TOPDOWN_ALLOCATION;
 
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)
+		vres->flags |= DRM_BUDDY_CONTIGUOUS_ALLOCATION;
+
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CLEARED)
+		vres->flags |= DRM_BUDDY_CLEAR_ALLOCATION;
+
 	if (fpfn || lpfn != mgr->mm.size)
 		/* Allocate blocks in desired range */
 		vres->flags |= DRM_BUDDY_RANGE_ALLOCATION;
 
+	if (bo->flags & AMDGPU_GEM_CREATE_GFX12_DCC &&
+	    adev->gmc.gmc_funcs->get_dcc_alignment)
+		adjust_dcc_size = amdgpu_gmc_get_dcc_alignment(adev);
+
 	remaining_size = (u64)vres->base.size;
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS && adjust_dcc_size) {
+		unsigned int dcc_size;
+
+		dcc_size = roundup_pow_of_two(vres->base.size + adjust_dcc_size);
+		remaining_size = (u64)dcc_size;
+
+		vres->flags |= DRM_BUDDY_TRIM_DISABLE;
+	}
 
 	mutex_lock(&mgr->lock);
 	while (remaining_size) {
@@ -448,33 +526,15 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 		else
 			min_block_size = mgr->default_page_size;
 
-		BUG_ON(min_block_size < mm->chunk_size);
-
-		/* Limit maximum size to 2GiB due to SG table limitations */
-		size = min(remaining_size, 2ULL << 30);
+		size = remaining_size;
 
-		if ((size >= (u64)pages_per_block << PAGE_SHIFT) &&
-				!(size & (((u64)pages_per_block << PAGE_SHIFT) - 1)))
+		if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS && adjust_dcc_size)
+			min_block_size = size;
+		else if ((size >= (u64)pages_per_block << PAGE_SHIFT) &&
+			 !(size & (((u64)pages_per_block << PAGE_SHIFT) - 1)))
 			min_block_size = (u64)pages_per_block << PAGE_SHIFT;
 
-		cur_size = size;
-
-		if (fpfn + size != (u64)place->lpfn << PAGE_SHIFT) {
-			/*
-			 * Except for actual range allocation, modify the size and
-			 * min_block_size conforming to continuous flag enablement
-			 */
-			if (place->flags & TTM_PL_FLAG_CONTIGUOUS) {
-				size = roundup_pow_of_two(size);
-				min_block_size = size;
-			/*
-			 * Modify the size value if size is not
-			 * aligned with min_block_size
-			 */
-			} else if (!IS_ALIGNED(size, min_block_size)) {
-				size = round_up(size, min_block_size);
-			}
-		}
+		BUG_ON(min_block_size < mm->chunk_size);
 
 		r = drm_buddy_alloc_blocks(mm, fpfn,
 					   lpfn,
@@ -482,6 +542,16 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 					   min_block_size,
 					   &vres->blocks,
 					   vres->flags);
+
+		if (unlikely(r == -ENOSPC) && pages_per_block == ~0ul &&
+		    !(place->flags & TTM_PL_FLAG_CONTIGUOUS)) {
+			vres->flags &= ~DRM_BUDDY_CONTIGUOUS_ALLOCATION;
+			pages_per_block = max_t(u32, 2UL << (20UL - PAGE_SHIFT),
+						tbo->page_alignment);
+
+			continue;
+		}
+
 		if (unlikely(r))
 			goto error_free_blocks;
 
@@ -490,43 +560,31 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 		else
 			remaining_size -= size;
 	}
-	mutex_unlock(&mgr->lock);
-
-	if (cur_size != size) {
-		struct drm_buddy_block *block;
-		struct list_head *trim_list;
-		u64 original_size;
-		LIST_HEAD(temp);
-
-		trim_list = &vres->blocks;
-		original_size = (u64)vres->base.size;
-
-		/*
-		 * If size value is rounded up to min_block_size, trim the last
-		 * block to the required size
-		 */
-		if (!list_is_singular(&vres->blocks)) {
-			block = list_last_entry(&vres->blocks, typeof(*block), link);
-			list_move_tail(&block->link, &temp);
-			trim_list = &temp;
-			/*
-			 * Compute the original_size value by subtracting the
-			 * last block size with (aligned size - original size)
-			 */
-			original_size = amdgpu_vram_mgr_block_size(block) - (size - cur_size);
-		}
-
-		mutex_lock(&mgr->lock);
-		drm_buddy_block_trim(mm,
-				     original_size,
-				     trim_list);
-		mutex_unlock(&mgr->lock);
 
-		if (!list_empty(&temp))
-			list_splice_tail(trim_list, &vres->blocks);
+	vres->task.pid = task_pid_nr(current);
+	get_task_comm(vres->task.comm, current);
+	list_add_tail(&vres->vres_node, &mgr->allocated_vres_list);
+
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS && adjust_dcc_size) {
+		struct drm_buddy_block *dcc_block;
+		unsigned long dcc_start;
+		u64 trim_start;
+
+		dcc_block = amdgpu_vram_mgr_first_block(&vres->blocks);
+		/* Adjust the start address for DCC buffers only */
+		dcc_start =
+			roundup((unsigned long)amdgpu_vram_mgr_block_start(dcc_block),
+				adjust_dcc_size);
+		trim_start = (u64)dcc_start;
+		drm_buddy_block_trim(mm, &trim_start,
+				     (u64)vres->base.size,
+				     &vres->blocks);
 	}
+	mutex_unlock(&mgr->lock);
 
 	vres->base.start = 0;
+	size = max_t(u64, amdgpu_vram_mgr_blocks_size(&vres->blocks),
+		     vres->base.size);
 	list_for_each_entry(block, &vres->blocks, link) {
 		unsigned long start;
 
@@ -534,8 +592,8 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 			amdgpu_vram_mgr_block_size(block);
 		start >>= PAGE_SHIFT;
 
-		if (start > PFN_UP(vres->base.size))
-			start -= PFN_UP(vres->base.size);
+		if (start > PFN_UP(size))
+			start -= PFN_UP(size);
 		else
 			start = 0;
 		vres->base.start = max(vres->base.start, start);
@@ -556,7 +614,7 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 	return 0;
 
 error_free_blocks:
-	drm_buddy_free_list(mm, &vres->blocks);
+	drm_buddy_free_list(mm, &vres->blocks, 0);
 	mutex_unlock(&mgr->lock);
 error_fini:
 	ttm_resource_fini(man, &vres->base);
@@ -584,12 +642,15 @@ static void amdgpu_vram_mgr_del(struct ttm_resource_manager *man,
 	uint64_t vis_usage = 0;
 
 	mutex_lock(&mgr->lock);
+
+	list_del(&vres->vres_node);
+	memset(&vres->task, 0, sizeof(vres->task));
+
 	list_for_each_entry(block, &vres->blocks, link)
 		vis_usage += amdgpu_vram_mgr_vis_size(adev, block);
 
+	drm_buddy_free_list(mm, &vres->blocks, vres->flags);
 	amdgpu_vram_mgr_do_reserve(man);
-
-	drm_buddy_free_list(mm, &vres->blocks);
 	mutex_unlock(&mgr->lock);
 
 	atomic64_sub(vis_usage, &mgr->vis_usage);
@@ -631,7 +692,7 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev,
 	amdgpu_res_first(res, offset, length, &cursor);
 	while (cursor.remaining) {
 		num_entries++;
-		amdgpu_res_next(&cursor, cursor.size);
+		amdgpu_res_next(&cursor, min(cursor.size, AMDGPU_MAX_SG_SEGMENT_SIZE));
 	}
 
 	r = sg_alloc_table(*sgt, num_entries, GFP_KERNEL);
@@ -651,7 +712,7 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev,
 	amdgpu_res_first(res, offset, length, &cursor);
 	for_each_sgtable_sg((*sgt), sg, i) {
 		phys_addr_t phys = cursor.start + adev->gmc.aper_base;
-		size_t size = cursor.size;
+		unsigned long size = min(cursor.size, AMDGPU_MAX_SG_SEGMENT_SIZE);
 		dma_addr_t addr;
 
 		addr = dma_map_resource(dev, phys, size, dir,
@@ -664,7 +725,7 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev,
 		sg_dma_address(sg) = addr;
 		sg_dma_len(sg) = size;
 
-		amdgpu_res_next(&cursor, cursor.size);
+		amdgpu_res_next(&cursor, size);
 	}
 
 	return 0;
@@ -722,6 +783,23 @@ uint64_t amdgpu_vram_mgr_vis_usage(struct amdgpu_vram_mgr *mgr)
 }
 
 /**
+ * amdgpu_vram_mgr_clear_reset_blocks - reset clear blocks
+ *
+ * @adev: amdgpu device pointer
+ *
+ * Reset the cleared drm buddy blocks.
+ */
+void amdgpu_vram_mgr_clear_reset_blocks(struct amdgpu_device *adev)
+{
+	struct amdgpu_vram_mgr *mgr = &adev->mman.vram_mgr;
+	struct drm_buddy *mm = &mgr->mm;
+
+	mutex_lock(&mgr->lock);
+	drm_buddy_reset_clear(mm, false);
+	mutex_unlock(&mgr->lock);
+}
+
+/**
  * amdgpu_vram_mgr_intersects - test each drm buddy block for intersection
  *
  * @man: TTM memory type manager
@@ -800,7 +878,7 @@ static void amdgpu_vram_mgr_debug(struct ttm_resource_manager *man,
 {
 	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
 	struct drm_buddy *mm = &mgr->mm;
-	struct drm_buddy_block *block;
+	struct amdgpu_vram_reservation *rsv;
 
 	drm_printf(printer, "  vis usage:%llu\n",
 		   amdgpu_vram_mgr_vis_usage(mgr));
@@ -812,8 +890,9 @@ static void amdgpu_vram_mgr_debug(struct ttm_resource_manager *man,
 	drm_buddy_print(mm, printer);
 
 	drm_printf(printer, "reserved:\n");
-	list_for_each_entry(block, &mgr->reserved_pages, link)
-		drm_buddy_block_print(mm, block, printer);
+	list_for_each_entry(rsv, &mgr->reserved_pages, blocks)
+		drm_printf(printer, "%#018llx-%#018llx: %llu\n",
+			rsv->start, rsv->start + rsv->size, rsv->size);
 	mutex_unlock(&mgr->lock);
 }
 
@@ -838,20 +917,23 @@ int amdgpu_vram_mgr_init(struct amdgpu_device *adev)
 	struct ttm_resource_manager *man = &mgr->manager;
 	int err;
 
+	man->cg = drmm_cgroup_register_region(adev_to_drm(adev), "vram", adev->gmc.real_vram_size);
+	if (IS_ERR(man->cg))
+		return PTR_ERR(man->cg);
 	ttm_resource_manager_init(man, &adev->mman.bdev,
 				  adev->gmc.real_vram_size);
 
-	man->func = &amdgpu_vram_mgr_func;
-
-	err = drm_buddy_init(&mgr->mm, man->size, PAGE_SIZE);
-	if (err)
-		return err;
-
 	mutex_init(&mgr->lock);
 	INIT_LIST_HEAD(&mgr->reservations_pending);
 	INIT_LIST_HEAD(&mgr->reserved_pages);
+	INIT_LIST_HEAD(&mgr->allocated_vres_list);
 	mgr->default_page_size = PAGE_SIZE;
 
+	man->func = &amdgpu_vram_mgr_func;
+	err = drm_buddy_init(&mgr->mm, man->size, PAGE_SIZE);
+	if (err)
+		return err;
+
 	ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_VRAM, &mgr->manager);
 	ttm_resource_manager_set_used(man, true);
 	return 0;
@@ -883,10 +965,11 @@ void amdgpu_vram_mgr_fini(struct amdgpu_device *adev)
 		kfree(rsv);
 
 	list_for_each_entry_safe(rsv, temp, &mgr->reserved_pages, blocks) {
-		drm_buddy_free_list(&mgr->mm, &rsv->allocated);
+		drm_buddy_free_list(&mgr->mm, &rsv->allocated, 0);
 		kfree(rsv);
 	}
-	drm_buddy_fini(&mgr->mm);
+	if (!adev->gmc.is_app_apu)
+		drm_buddy_fini(&mgr->mm);
 	mutex_unlock(&mgr->lock);
 
 	ttm_resource_manager_cleanup(man);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.h
index 0e04e42cf809..5f5fd9a911c2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.h
@@ -35,12 +35,26 @@ struct amdgpu_vram_mgr {
 	struct list_head reserved_pages;
 	atomic64_t vis_usage;
 	u64 default_page_size;
+	struct list_head allocated_vres_list;
+};
+
+struct amdgpu_vres_task {
+	pid_t pid;
+	char  comm[TASK_COMM_LEN];
+};
+
+struct amdgpu_vram_block_info {
+	u64 start;
+	u64 size;
+	struct amdgpu_vres_task task;
 };
 
 struct amdgpu_vram_mgr_resource {
 	struct ttm_resource base;
 	struct list_head blocks;
 	unsigned long flags;
+	struct list_head vres_node;
+	struct amdgpu_vres_task task;
 };
 
 static inline u64 amdgpu_vram_mgr_block_start(struct drm_buddy_block *block)
@@ -53,10 +67,26 @@ static inline u64 amdgpu_vram_mgr_block_size(struct drm_buddy_block *block)
 	return (u64)PAGE_SIZE << drm_buddy_block_order(block);
 }
 
+static inline bool amdgpu_vram_mgr_is_cleared(struct drm_buddy_block *block)
+{
+	return drm_buddy_block_is_clear(block);
+}
+
 static inline struct amdgpu_vram_mgr_resource *
 to_amdgpu_vram_mgr_resource(struct ttm_resource *res)
 {
 	return container_of(res, struct amdgpu_vram_mgr_resource, base);
 }
 
+static inline void amdgpu_vram_mgr_set_cleared(struct ttm_resource *res)
+{
+	struct amdgpu_vram_mgr_resource *ares = to_amdgpu_vram_mgr_resource(res);
+
+	WARN_ON(ares->flags & DRM_BUDDY_CLEARED);
+	ares->flags |= DRM_BUDDY_CLEARED;
+}
+
+int amdgpu_vram_mgr_query_address_block_info(struct amdgpu_vram_mgr *mgr,
+		uint64_t address, struct amdgpu_vram_block_info *info);
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.c
new file mode 100644
index 000000000000..1083db8cea2e
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.c
@@ -0,0 +1,1107 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "amdgpu.h"
+#include "amdgpu_xcp.h"
+#include "amdgpu_drv.h"
+
+#include <drm/drm_drv.h>
+#include "../amdxcp/amdgpu_xcp_drv.h"
+
+static void amdgpu_xcp_sysfs_entries_init(struct amdgpu_xcp_mgr *xcp_mgr);
+static void amdgpu_xcp_sysfs_entries_update(struct amdgpu_xcp_mgr *xcp_mgr);
+
+static int __amdgpu_xcp_run(struct amdgpu_xcp_mgr *xcp_mgr,
+			    struct amdgpu_xcp_ip *xcp_ip, int xcp_state)
+{
+	int (*run_func)(void *handle, uint32_t inst_mask);
+	int ret = 0;
+
+	if (!xcp_ip || !xcp_ip->valid || !xcp_ip->ip_funcs)
+		return 0;
+
+	run_func = NULL;
+
+	switch (xcp_state) {
+	case AMDGPU_XCP_PREPARE_SUSPEND:
+		run_func = xcp_ip->ip_funcs->prepare_suspend;
+		break;
+	case AMDGPU_XCP_SUSPEND:
+		run_func = xcp_ip->ip_funcs->suspend;
+		break;
+	case AMDGPU_XCP_PREPARE_RESUME:
+		run_func = xcp_ip->ip_funcs->prepare_resume;
+		break;
+	case AMDGPU_XCP_RESUME:
+		run_func = xcp_ip->ip_funcs->resume;
+		break;
+	}
+
+	if (run_func)
+		ret = run_func(xcp_mgr->adev, xcp_ip->inst_mask);
+
+	return ret;
+}
+
+static int amdgpu_xcp_run_transition(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id,
+				     int state)
+{
+	struct amdgpu_xcp_ip *xcp_ip;
+	struct amdgpu_xcp *xcp;
+	int i, ret;
+
+	if (xcp_id >= MAX_XCP || !xcp_mgr->xcp[xcp_id].valid)
+		return -EINVAL;
+
+	xcp = &xcp_mgr->xcp[xcp_id];
+	for (i = 0; i < AMDGPU_XCP_MAX_BLOCKS; ++i) {
+		xcp_ip = &xcp->ip[i];
+		ret = __amdgpu_xcp_run(xcp_mgr, xcp_ip, state);
+		if (ret)
+			break;
+	}
+
+	return ret;
+}
+
+int amdgpu_xcp_prepare_suspend(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id)
+{
+	return amdgpu_xcp_run_transition(xcp_mgr, xcp_id,
+					 AMDGPU_XCP_PREPARE_SUSPEND);
+}
+
+int amdgpu_xcp_suspend(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id)
+{
+	return amdgpu_xcp_run_transition(xcp_mgr, xcp_id, AMDGPU_XCP_SUSPEND);
+}
+
+int amdgpu_xcp_prepare_resume(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id)
+{
+	return amdgpu_xcp_run_transition(xcp_mgr, xcp_id,
+					 AMDGPU_XCP_PREPARE_RESUME);
+}
+
+int amdgpu_xcp_resume(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id)
+{
+	return amdgpu_xcp_run_transition(xcp_mgr, xcp_id, AMDGPU_XCP_RESUME);
+}
+
+static void __amdgpu_xcp_add_block(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id,
+				   struct amdgpu_xcp_ip *ip)
+{
+	struct amdgpu_xcp *xcp;
+
+	if (!ip)
+		return;
+
+	xcp = &xcp_mgr->xcp[xcp_id];
+	xcp->ip[ip->ip_id] = *ip;
+	xcp->ip[ip->ip_id].valid = true;
+
+	xcp->valid = true;
+}
+
+static void __amdgpu_xcp_set_unique_id(struct amdgpu_xcp_mgr *xcp_mgr,
+				       int xcp_id)
+{
+	struct amdgpu_xcp *xcp = &xcp_mgr->xcp[xcp_id];
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	uint32_t inst_mask;
+	uint64_t uid;
+	int i;
+
+	if (!amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_GFX, &inst_mask) &&
+	    inst_mask) {
+		i = GET_INST(GC, (ffs(inst_mask) - 1));
+		uid = amdgpu_device_get_uid(xcp_mgr->adev->uid_info,
+					    AMDGPU_UID_TYPE_XCD, i);
+		if (uid)
+			xcp->unique_id = uid;
+	}
+}
+
+int amdgpu_xcp_init(struct amdgpu_xcp_mgr *xcp_mgr, int num_xcps, int mode)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	struct amdgpu_xcp_ip ip;
+	uint8_t mem_id;
+	int i, j, ret;
+
+	if (!num_xcps || num_xcps > MAX_XCP)
+		return -EINVAL;
+
+	xcp_mgr->mode = mode;
+
+	for (i = 0; i < MAX_XCP; ++i)
+		xcp_mgr->xcp[i].valid = false;
+
+	/* This is needed for figuring out memory id of xcp */
+	xcp_mgr->num_xcp_per_mem_partition = num_xcps / xcp_mgr->adev->gmc.num_mem_partitions;
+
+	for (i = 0; i < num_xcps; ++i) {
+		for (j = AMDGPU_XCP_GFXHUB; j < AMDGPU_XCP_MAX_BLOCKS; ++j) {
+			ret = xcp_mgr->funcs->get_ip_details(xcp_mgr, i, j,
+							     &ip);
+			if (ret)
+				continue;
+
+			__amdgpu_xcp_add_block(xcp_mgr, i, &ip);
+		}
+
+		xcp_mgr->xcp[i].id = i;
+
+		if (xcp_mgr->funcs->get_xcp_mem_id) {
+			ret = xcp_mgr->funcs->get_xcp_mem_id(
+				xcp_mgr, &xcp_mgr->xcp[i], &mem_id);
+			if (ret)
+				continue;
+			else
+				xcp_mgr->xcp[i].mem_id = mem_id;
+		}
+		__amdgpu_xcp_set_unique_id(xcp_mgr, i);
+	}
+
+	xcp_mgr->num_xcps = num_xcps;
+	amdgpu_xcp_update_partition_sched_list(adev);
+
+	return 0;
+}
+
+static int __amdgpu_xcp_switch_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr,
+					      int mode)
+{
+	int ret, curr_mode, num_xcps = 0;
+
+	if (!xcp_mgr->funcs || !xcp_mgr->funcs->switch_partition_mode)
+		return 0;
+
+	mutex_lock(&xcp_mgr->xcp_lock);
+
+	curr_mode = xcp_mgr->mode;
+	/* State set to transient mode */
+	xcp_mgr->mode = AMDGPU_XCP_MODE_TRANS;
+
+	ret = xcp_mgr->funcs->switch_partition_mode(xcp_mgr, mode, &num_xcps);
+
+	if (ret) {
+		/* Failed, get whatever mode it's at now */
+		if (xcp_mgr->funcs->query_partition_mode)
+			xcp_mgr->mode = amdgpu_xcp_query_partition_mode(
+				xcp_mgr, AMDGPU_XCP_FL_LOCKED);
+		else
+			xcp_mgr->mode = curr_mode;
+
+		goto out;
+	}
+	amdgpu_xcp_sysfs_entries_update(xcp_mgr);
+out:
+	mutex_unlock(&xcp_mgr->xcp_lock);
+
+	return ret;
+}
+
+int amdgpu_xcp_switch_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr, int mode)
+{
+	if (!xcp_mgr || mode == AMDGPU_XCP_MODE_NONE)
+		return -EINVAL;
+
+	if (xcp_mgr->mode == mode)
+		return 0;
+
+	return __amdgpu_xcp_switch_partition_mode(xcp_mgr, mode);
+}
+
+int amdgpu_xcp_restore_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	if (!xcp_mgr || xcp_mgr->mode == AMDGPU_XCP_MODE_NONE)
+		return 0;
+
+	return __amdgpu_xcp_switch_partition_mode(xcp_mgr, xcp_mgr->mode);
+}
+
+static bool __amdgpu_xcp_is_cached_mode_valid(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	if (!xcp_mgr->funcs || !xcp_mgr->funcs->query_partition_mode)
+		return true;
+
+	if (!amdgpu_sriov_vf(xcp_mgr->adev) &&
+	    xcp_mgr->mode == AMDGPU_XCP_MODE_NONE)
+		return true;
+
+	if (xcp_mgr->mode != AMDGPU_XCP_MODE_NONE &&
+	    xcp_mgr->mode != AMDGPU_XCP_MODE_TRANS)
+		return true;
+
+	return false;
+}
+
+int amdgpu_xcp_query_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr, u32 flags)
+{
+	int mode;
+
+	if (__amdgpu_xcp_is_cached_mode_valid(xcp_mgr))
+		return xcp_mgr->mode;
+
+	if (!(flags & AMDGPU_XCP_FL_LOCKED))
+		mutex_lock(&xcp_mgr->xcp_lock);
+	mode = xcp_mgr->funcs->query_partition_mode(xcp_mgr);
+
+	/* First time query for VF, set the mode here */
+	if (amdgpu_sriov_vf(xcp_mgr->adev) &&
+	    xcp_mgr->mode == AMDGPU_XCP_MODE_NONE)
+		xcp_mgr->mode = mode;
+
+	if (xcp_mgr->mode != AMDGPU_XCP_MODE_TRANS && mode != xcp_mgr->mode)
+		dev_WARN(
+			xcp_mgr->adev->dev,
+			"Cached partition mode %d not matching with device mode %d",
+			xcp_mgr->mode, mode);
+
+	if (!(flags & AMDGPU_XCP_FL_LOCKED))
+		mutex_unlock(&xcp_mgr->xcp_lock);
+
+	return mode;
+}
+
+static int amdgpu_xcp_dev_alloc(struct amdgpu_device *adev)
+{
+	struct drm_device *p_ddev;
+	struct drm_device *ddev;
+	int i, ret;
+
+	ddev = adev_to_drm(adev);
+
+	/* xcp #0 shares drm device setting with adev */
+	adev->xcp_mgr->xcp->ddev = ddev;
+
+	for (i = 1; i < MAX_XCP; i++) {
+		ret = amdgpu_xcp_drm_dev_alloc(&p_ddev);
+		if (ret == -ENOSPC) {
+			dev_warn(adev->dev,
+			"Skip xcp node #%d when out of drm node resource.", i);
+			ret = 0;
+			goto out;
+		} else if (ret) {
+			goto out;
+		}
+
+		/* Redirect all IOCTLs to the primary device */
+		adev->xcp_mgr->xcp[i].rdev = p_ddev->render->dev;
+		adev->xcp_mgr->xcp[i].pdev = p_ddev->primary->dev;
+		adev->xcp_mgr->xcp[i].driver = (struct drm_driver *)p_ddev->driver;
+		adev->xcp_mgr->xcp[i].vma_offset_manager = p_ddev->vma_offset_manager;
+		p_ddev->render->dev = ddev;
+		p_ddev->primary->dev = ddev;
+		p_ddev->vma_offset_manager = ddev->vma_offset_manager;
+		p_ddev->driver = &amdgpu_partition_driver;
+		adev->xcp_mgr->xcp[i].ddev = p_ddev;
+
+		dev_set_drvdata(p_ddev->dev, &adev->xcp_mgr->xcp[i]);
+	}
+	ret = 0;
+out:
+	amdgpu_xcp_sysfs_entries_init(adev->xcp_mgr);
+
+	return ret;
+}
+
+int amdgpu_xcp_mgr_init(struct amdgpu_device *adev, int init_mode,
+			int init_num_xcps,
+			struct amdgpu_xcp_mgr_funcs *xcp_funcs)
+{
+	struct amdgpu_xcp_mgr *xcp_mgr;
+	int i;
+
+	if (!xcp_funcs || !xcp_funcs->get_ip_details)
+		return -EINVAL;
+
+	xcp_mgr = kzalloc(sizeof(*xcp_mgr), GFP_KERNEL);
+
+	if (!xcp_mgr)
+		return -ENOMEM;
+
+	xcp_mgr->adev = adev;
+	xcp_mgr->funcs = xcp_funcs;
+	xcp_mgr->mode = init_mode;
+	mutex_init(&xcp_mgr->xcp_lock);
+
+	if (init_mode != AMDGPU_XCP_MODE_NONE)
+		amdgpu_xcp_init(xcp_mgr, init_num_xcps, init_mode);
+
+	adev->xcp_mgr = xcp_mgr;
+	for (i = 0; i < MAX_XCP; ++i)
+		xcp_mgr->xcp[i].xcp_mgr = xcp_mgr;
+
+	return amdgpu_xcp_dev_alloc(adev);
+}
+
+int amdgpu_xcp_get_partition(struct amdgpu_xcp_mgr *xcp_mgr,
+			     enum AMDGPU_XCP_IP_BLOCK ip, int instance)
+{
+	struct amdgpu_xcp *xcp;
+	int i, id_mask = 0;
+
+	if (ip >= AMDGPU_XCP_MAX_BLOCKS)
+		return -EINVAL;
+
+	for (i = 0; i < xcp_mgr->num_xcps; ++i) {
+		xcp = &xcp_mgr->xcp[i];
+		if ((xcp->valid) && (xcp->ip[ip].valid) &&
+		    (xcp->ip[ip].inst_mask & BIT(instance)))
+			id_mask |= BIT(i);
+	}
+
+	if (!id_mask)
+		id_mask = -ENXIO;
+
+	return id_mask;
+}
+
+int amdgpu_xcp_get_inst_details(struct amdgpu_xcp *xcp,
+				enum AMDGPU_XCP_IP_BLOCK ip,
+				uint32_t *inst_mask)
+{
+	if (!xcp->valid || !inst_mask || !(xcp->ip[ip].valid))
+		return -EINVAL;
+
+	*inst_mask = xcp->ip[ip].inst_mask;
+
+	return 0;
+}
+
+int amdgpu_xcp_dev_register(struct amdgpu_device *adev,
+			const struct pci_device_id *ent)
+{
+	int i, ret;
+
+	if (!adev->xcp_mgr)
+		return 0;
+
+	for (i = 1; i < MAX_XCP; i++) {
+		if (!adev->xcp_mgr->xcp[i].ddev)
+			break;
+
+		ret = drm_dev_register(adev->xcp_mgr->xcp[i].ddev, ent->driver_data);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+void amdgpu_xcp_dev_unplug(struct amdgpu_device *adev)
+{
+	struct drm_device *p_ddev;
+	int i;
+
+	if (!adev->xcp_mgr)
+		return;
+
+	for (i = 1; i < MAX_XCP; i++) {
+		if (!adev->xcp_mgr->xcp[i].ddev)
+			break;
+
+		p_ddev = adev->xcp_mgr->xcp[i].ddev;
+		drm_dev_unplug(p_ddev);
+		p_ddev->render->dev = adev->xcp_mgr->xcp[i].rdev;
+		p_ddev->primary->dev = adev->xcp_mgr->xcp[i].pdev;
+		p_ddev->driver =  adev->xcp_mgr->xcp[i].driver;
+		p_ddev->vma_offset_manager = adev->xcp_mgr->xcp[i].vma_offset_manager;
+		amdgpu_xcp_drm_dev_free(p_ddev);
+	}
+}
+
+int amdgpu_xcp_open_device(struct amdgpu_device *adev,
+			   struct amdgpu_fpriv *fpriv,
+			   struct drm_file *file_priv)
+{
+	int i;
+
+	if (!adev->xcp_mgr)
+		return 0;
+
+	fpriv->xcp_id = AMDGPU_XCP_NO_PARTITION;
+	for (i = 0; i < MAX_XCP; ++i) {
+		if (!adev->xcp_mgr->xcp[i].ddev)
+			break;
+
+		if (file_priv->minor == adev->xcp_mgr->xcp[i].ddev->render) {
+			if (adev->xcp_mgr->xcp[i].valid == FALSE) {
+				dev_err(adev->dev, "renderD%d partition %d not valid!",
+						file_priv->minor->index, i);
+				return -ENOENT;
+			}
+			dev_dbg(adev->dev, "renderD%d partition %d opened!",
+					file_priv->minor->index, i);
+			fpriv->xcp_id = i;
+			break;
+		}
+	}
+
+	fpriv->vm.mem_id = fpriv->xcp_id == AMDGPU_XCP_NO_PARTITION ? -1 :
+				adev->xcp_mgr->xcp[fpriv->xcp_id].mem_id;
+	return 0;
+}
+
+void amdgpu_xcp_release_sched(struct amdgpu_device *adev,
+				  struct amdgpu_ctx_entity *entity)
+{
+	struct drm_gpu_scheduler *sched;
+	struct amdgpu_ring *ring;
+
+	if (!adev->xcp_mgr)
+		return;
+
+	sched = entity->entity.rq->sched;
+	if (drm_sched_wqueue_ready(sched)) {
+		ring = to_amdgpu_ring(entity->entity.rq->sched);
+		atomic_dec(&adev->xcp_mgr->xcp[ring->xcp_id].ref_cnt);
+	}
+}
+
+int amdgpu_xcp_select_scheds(struct amdgpu_device *adev,
+			     u32 hw_ip, u32 hw_prio,
+			     struct amdgpu_fpriv *fpriv,
+			     unsigned int *num_scheds,
+			     struct drm_gpu_scheduler ***scheds)
+{
+	u32 sel_xcp_id;
+	int i;
+	struct amdgpu_xcp_mgr *xcp_mgr = adev->xcp_mgr;
+
+	if (fpriv->xcp_id == AMDGPU_XCP_NO_PARTITION) {
+		u32 least_ref_cnt = ~0;
+
+		fpriv->xcp_id = 0;
+		for (i = 0; i < xcp_mgr->num_xcps; i++) {
+			u32 total_ref_cnt;
+
+			total_ref_cnt = atomic_read(&xcp_mgr->xcp[i].ref_cnt);
+			if (total_ref_cnt < least_ref_cnt) {
+				fpriv->xcp_id = i;
+				least_ref_cnt = total_ref_cnt;
+			}
+		}
+	}
+	sel_xcp_id = fpriv->xcp_id;
+
+	if (xcp_mgr->xcp[sel_xcp_id].gpu_sched[hw_ip][hw_prio].num_scheds) {
+		*num_scheds =
+			xcp_mgr->xcp[fpriv->xcp_id].gpu_sched[hw_ip][hw_prio].num_scheds;
+		*scheds =
+			xcp_mgr->xcp[fpriv->xcp_id].gpu_sched[hw_ip][hw_prio].sched;
+		atomic_inc(&adev->xcp_mgr->xcp[sel_xcp_id].ref_cnt);
+		dev_dbg(adev->dev, "Selected partition #%d", sel_xcp_id);
+	} else {
+		dev_err(adev->dev, "Failed to schedule partition #%d.", sel_xcp_id);
+		return -ENOENT;
+	}
+
+	return 0;
+}
+
+static void amdgpu_set_xcp_id(struct amdgpu_device *adev,
+			      uint32_t inst_idx,
+			      struct amdgpu_ring *ring)
+{
+	int xcp_id;
+	enum AMDGPU_XCP_IP_BLOCK ip_blk;
+	uint32_t inst_mask;
+
+	ring->xcp_id = AMDGPU_XCP_NO_PARTITION;
+	if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE)
+		adev->gfx.enforce_isolation[0].xcp_id = ring->xcp_id;
+	if ((adev->xcp_mgr->mode == AMDGPU_XCP_MODE_NONE) ||
+	    (ring->funcs->type == AMDGPU_RING_TYPE_CPER))
+		return;
+
+	inst_mask = 1 << inst_idx;
+
+	switch (ring->funcs->type) {
+	case AMDGPU_HW_IP_GFX:
+	case AMDGPU_RING_TYPE_COMPUTE:
+	case AMDGPU_RING_TYPE_KIQ:
+		ip_blk = AMDGPU_XCP_GFX;
+		break;
+	case AMDGPU_RING_TYPE_SDMA:
+		ip_blk = AMDGPU_XCP_SDMA;
+		break;
+	case AMDGPU_RING_TYPE_VCN_ENC:
+	case AMDGPU_RING_TYPE_VCN_JPEG:
+		ip_blk = AMDGPU_XCP_VCN;
+		break;
+	default:
+		dev_err(adev->dev, "Not support ring type %d!", ring->funcs->type);
+		return;
+	}
+
+	for (xcp_id = 0; xcp_id < adev->xcp_mgr->num_xcps; xcp_id++) {
+		if (adev->xcp_mgr->xcp[xcp_id].ip[ip_blk].inst_mask & inst_mask) {
+			ring->xcp_id = xcp_id;
+			dev_dbg(adev->dev, "ring:%s xcp_id :%u", ring->name,
+				ring->xcp_id);
+			if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE)
+				adev->gfx.enforce_isolation[xcp_id].xcp_id = xcp_id;
+			break;
+		}
+	}
+}
+
+static void amdgpu_xcp_gpu_sched_update(struct amdgpu_device *adev,
+					struct amdgpu_ring *ring,
+					unsigned int sel_xcp_id)
+{
+	unsigned int *num_gpu_sched;
+
+	num_gpu_sched = &adev->xcp_mgr->xcp[sel_xcp_id]
+			.gpu_sched[ring->funcs->type][ring->hw_prio].num_scheds;
+	adev->xcp_mgr->xcp[sel_xcp_id].gpu_sched[ring->funcs->type][ring->hw_prio]
+			.sched[(*num_gpu_sched)++] = &ring->sched;
+	dev_dbg(adev->dev, "%s :[%d] gpu_sched[%d][%d] = %d",
+		ring->name, sel_xcp_id, ring->funcs->type,
+		ring->hw_prio, *num_gpu_sched);
+}
+
+static int amdgpu_xcp_sched_list_update(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	int i;
+
+	for (i = 0; i < MAX_XCP; i++) {
+		atomic_set(&adev->xcp_mgr->xcp[i].ref_cnt, 0);
+		memset(adev->xcp_mgr->xcp[i].gpu_sched, 0, sizeof(adev->xcp_mgr->xcp->gpu_sched));
+	}
+
+	if (adev->xcp_mgr->mode == AMDGPU_XCP_MODE_NONE)
+		return 0;
+
+	for (i = 0; i < AMDGPU_MAX_RINGS; i++) {
+		ring = adev->rings[i];
+		if (!ring || !ring->sched.ready || ring->no_scheduler)
+			continue;
+
+		amdgpu_xcp_gpu_sched_update(adev, ring, ring->xcp_id);
+
+		/* VCN may be shared by two partitions under CPX MODE in certain
+		 * configs.
+		 */
+		if ((ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC ||
+		     ring->funcs->type == AMDGPU_RING_TYPE_VCN_JPEG) &&
+		    (adev->xcp_mgr->num_xcps > adev->vcn.num_vcn_inst))
+			amdgpu_xcp_gpu_sched_update(adev, ring, ring->xcp_id + 1);
+	}
+
+	return 0;
+}
+
+int amdgpu_xcp_update_partition_sched_list(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->num_rings; i++) {
+		struct amdgpu_ring *ring = adev->rings[i];
+
+		if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE ||
+			ring->funcs->type == AMDGPU_RING_TYPE_KIQ)
+			amdgpu_set_xcp_id(adev, ring->xcc_id, ring);
+		else
+			amdgpu_set_xcp_id(adev, ring->me, ring);
+	}
+
+	return amdgpu_xcp_sched_list_update(adev);
+}
+
+void amdgpu_xcp_update_supported_modes(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+
+	xcp_mgr->supp_xcp_modes = 0;
+
+	switch (NUM_XCC(adev->gfx.xcc_mask)) {
+	case 8:
+		xcp_mgr->supp_xcp_modes = BIT(AMDGPU_SPX_PARTITION_MODE) |
+					  BIT(AMDGPU_DPX_PARTITION_MODE) |
+					  BIT(AMDGPU_QPX_PARTITION_MODE) |
+					  BIT(AMDGPU_CPX_PARTITION_MODE);
+		break;
+	case 6:
+		xcp_mgr->supp_xcp_modes = BIT(AMDGPU_SPX_PARTITION_MODE) |
+					  BIT(AMDGPU_TPX_PARTITION_MODE) |
+					  BIT(AMDGPU_CPX_PARTITION_MODE);
+		break;
+	case 4:
+		xcp_mgr->supp_xcp_modes = BIT(AMDGPU_SPX_PARTITION_MODE) |
+					  BIT(AMDGPU_DPX_PARTITION_MODE) |
+					  BIT(AMDGPU_CPX_PARTITION_MODE);
+		break;
+	case 2:
+		xcp_mgr->supp_xcp_modes = BIT(AMDGPU_SPX_PARTITION_MODE) |
+					  BIT(AMDGPU_CPX_PARTITION_MODE);
+		break;
+	case 1:
+		xcp_mgr->supp_xcp_modes = BIT(AMDGPU_SPX_PARTITION_MODE) |
+					  BIT(AMDGPU_CPX_PARTITION_MODE);
+		break;
+
+	default:
+		break;
+	}
+}
+
+int amdgpu_xcp_pre_partition_switch(struct amdgpu_xcp_mgr *xcp_mgr, u32 flags)
+{
+	/* TODO:
+	 * Stop user queues and threads, and make sure GPU is empty of work.
+	 */
+
+	if (flags & AMDGPU_XCP_OPS_KFD)
+		amdgpu_amdkfd_device_fini_sw(xcp_mgr->adev);
+
+	return 0;
+}
+
+int amdgpu_xcp_post_partition_switch(struct amdgpu_xcp_mgr *xcp_mgr, u32 flags)
+{
+	int ret = 0;
+
+	if (flags & AMDGPU_XCP_OPS_KFD) {
+		amdgpu_amdkfd_device_probe(xcp_mgr->adev);
+		amdgpu_amdkfd_device_init(xcp_mgr->adev);
+		/* If KFD init failed, return failure */
+		if (!xcp_mgr->adev->kfd.init_complete)
+			ret = -EIO;
+	}
+
+	return ret;
+}
+
+/*====================== xcp sysfs - configuration ======================*/
+#define XCP_CFG_SYSFS_RES_ATTR_SHOW(_name)                         \
+	static ssize_t amdgpu_xcp_res_sysfs_##_name##_show(        \
+		struct amdgpu_xcp_res_details *xcp_res, char *buf) \
+	{                                                          \
+		return sysfs_emit(buf, "%d\n", xcp_res->_name);    \
+	}
+
+struct amdgpu_xcp_res_sysfs_attribute {
+	struct attribute attr;
+	ssize_t (*show)(struct amdgpu_xcp_res_details *xcp_res, char *buf);
+};
+
+#define XCP_CFG_SYSFS_RES_ATTR(_name)                                        \
+	struct amdgpu_xcp_res_sysfs_attribute xcp_res_sysfs_attr_##_name = { \
+		.attr = { .name = __stringify(_name), .mode = 0400 },        \
+		.show = amdgpu_xcp_res_sysfs_##_name##_show,                 \
+	}
+
+XCP_CFG_SYSFS_RES_ATTR_SHOW(num_inst)
+XCP_CFG_SYSFS_RES_ATTR(num_inst);
+XCP_CFG_SYSFS_RES_ATTR_SHOW(num_shared)
+XCP_CFG_SYSFS_RES_ATTR(num_shared);
+
+#define XCP_CFG_SYSFS_RES_ATTR_PTR(_name) xcp_res_sysfs_attr_##_name.attr
+
+static struct attribute *xcp_cfg_res_sysfs_attrs[] = {
+	&XCP_CFG_SYSFS_RES_ATTR_PTR(num_inst),
+	&XCP_CFG_SYSFS_RES_ATTR_PTR(num_shared), NULL
+};
+
+static const char *xcp_desc[] = {
+	[AMDGPU_SPX_PARTITION_MODE] = "SPX",
+	[AMDGPU_DPX_PARTITION_MODE] = "DPX",
+	[AMDGPU_TPX_PARTITION_MODE] = "TPX",
+	[AMDGPU_QPX_PARTITION_MODE] = "QPX",
+	[AMDGPU_CPX_PARTITION_MODE] = "CPX",
+};
+
+static const char *nps_desc[] = {
+	[UNKNOWN_MEMORY_PARTITION_MODE] = "UNKNOWN",
+	[AMDGPU_NPS1_PARTITION_MODE] = "NPS1",
+	[AMDGPU_NPS2_PARTITION_MODE] = "NPS2",
+	[AMDGPU_NPS3_PARTITION_MODE] = "NPS3",
+	[AMDGPU_NPS4_PARTITION_MODE] = "NPS4",
+	[AMDGPU_NPS6_PARTITION_MODE] = "NPS6",
+	[AMDGPU_NPS8_PARTITION_MODE] = "NPS8",
+};
+
+ATTRIBUTE_GROUPS(xcp_cfg_res_sysfs);
+
+#define to_xcp_attr(x) \
+	container_of(x, struct amdgpu_xcp_res_sysfs_attribute, attr)
+#define to_xcp_res(x) container_of(x, struct amdgpu_xcp_res_details, kobj)
+
+static ssize_t xcp_cfg_res_sysfs_attr_show(struct kobject *kobj,
+					   struct attribute *attr, char *buf)
+{
+	struct amdgpu_xcp_res_sysfs_attribute *attribute;
+	struct amdgpu_xcp_res_details *xcp_res;
+
+	attribute = to_xcp_attr(attr);
+	xcp_res = to_xcp_res(kobj);
+
+	if (!attribute->show)
+		return -EIO;
+
+	return attribute->show(xcp_res, buf);
+}
+
+static const struct sysfs_ops xcp_cfg_res_sysfs_ops = {
+	.show = xcp_cfg_res_sysfs_attr_show,
+};
+
+static const struct kobj_type xcp_cfg_res_sysfs_ktype = {
+	.sysfs_ops = &xcp_cfg_res_sysfs_ops,
+	.default_groups = xcp_cfg_res_sysfs_groups,
+};
+
+const char *xcp_res_names[] = {
+	[AMDGPU_XCP_RES_XCC] = "xcc",
+	[AMDGPU_XCP_RES_DMA] = "dma",
+	[AMDGPU_XCP_RES_DEC] = "dec",
+	[AMDGPU_XCP_RES_JPEG] = "jpeg",
+};
+
+static int amdgpu_xcp_get_res_info(struct amdgpu_xcp_mgr *xcp_mgr,
+				   int mode,
+				   struct amdgpu_xcp_cfg *xcp_cfg)
+{
+	if (xcp_mgr->funcs && xcp_mgr->funcs->get_xcp_res_info)
+		return xcp_mgr->funcs->get_xcp_res_info(xcp_mgr, mode, xcp_cfg);
+
+	return -EOPNOTSUPP;
+}
+
+#define to_xcp_cfg(x) container_of(x, struct amdgpu_xcp_cfg, kobj)
+static ssize_t supported_xcp_configs_show(struct kobject *kobj,
+					  struct kobj_attribute *attr, char *buf)
+{
+	struct amdgpu_xcp_cfg *xcp_cfg = to_xcp_cfg(kobj);
+	struct amdgpu_xcp_mgr *xcp_mgr = xcp_cfg->xcp_mgr;
+	int size = 0, mode;
+	char *sep = "";
+
+	if (!xcp_mgr || !xcp_mgr->supp_xcp_modes)
+		return sysfs_emit(buf, "Not supported\n");
+
+	for_each_inst(mode, xcp_mgr->supp_xcp_modes) {
+		size += sysfs_emit_at(buf, size, "%s%s", sep, xcp_desc[mode]);
+		sep = ", ";
+	}
+
+	size += sysfs_emit_at(buf, size, "\n");
+
+	return size;
+}
+
+static ssize_t supported_nps_configs_show(struct kobject *kobj,
+					  struct kobj_attribute *attr, char *buf)
+{
+	struct amdgpu_xcp_cfg *xcp_cfg = to_xcp_cfg(kobj);
+	int size = 0, mode;
+	char *sep = "";
+
+	if (!xcp_cfg || !xcp_cfg->compatible_nps_modes)
+		return sysfs_emit(buf, "Not supported\n");
+
+	for_each_inst(mode, xcp_cfg->compatible_nps_modes) {
+		size += sysfs_emit_at(buf, size, "%s%s", sep, nps_desc[mode]);
+		sep = ", ";
+	}
+
+	size += sysfs_emit_at(buf, size, "\n");
+
+	return size;
+}
+
+static ssize_t xcp_config_show(struct kobject *kobj,
+			       struct kobj_attribute *attr, char *buf)
+{
+	struct amdgpu_xcp_cfg *xcp_cfg = to_xcp_cfg(kobj);
+
+	return sysfs_emit(buf, "%s\n",
+			  amdgpu_gfx_compute_mode_desc(xcp_cfg->mode));
+}
+
+static ssize_t xcp_config_store(struct kobject *kobj,
+				struct kobj_attribute *attr,
+				const char *buf, size_t size)
+{
+	struct amdgpu_xcp_cfg *xcp_cfg = to_xcp_cfg(kobj);
+	int mode, r;
+
+	if (!strncasecmp("SPX", buf, strlen("SPX")))
+		mode = AMDGPU_SPX_PARTITION_MODE;
+	else if (!strncasecmp("DPX", buf, strlen("DPX")))
+		mode = AMDGPU_DPX_PARTITION_MODE;
+	else if (!strncasecmp("TPX", buf, strlen("TPX")))
+		mode = AMDGPU_TPX_PARTITION_MODE;
+	else if (!strncasecmp("QPX", buf, strlen("QPX")))
+		mode = AMDGPU_QPX_PARTITION_MODE;
+	else if (!strncasecmp("CPX", buf, strlen("CPX")))
+		mode = AMDGPU_CPX_PARTITION_MODE;
+	else
+		return -EINVAL;
+
+	r = amdgpu_xcp_get_res_info(xcp_cfg->xcp_mgr, mode, xcp_cfg);
+
+	if (r)
+		return r;
+
+	xcp_cfg->mode = mode;
+	return size;
+}
+
+static struct kobj_attribute xcp_cfg_sysfs_mode =
+	__ATTR_RW_MODE(xcp_config, 0644);
+
+static void xcp_cfg_sysfs_release(struct kobject *kobj)
+{
+	struct amdgpu_xcp_cfg *xcp_cfg = to_xcp_cfg(kobj);
+
+	kfree(xcp_cfg);
+}
+
+static const struct kobj_type xcp_cfg_sysfs_ktype = {
+	.release = xcp_cfg_sysfs_release,
+	.sysfs_ops = &kobj_sysfs_ops,
+};
+
+static struct kobj_attribute supp_part_sysfs_mode =
+	__ATTR_RO(supported_xcp_configs);
+
+static struct kobj_attribute supp_nps_sysfs_mode =
+	__ATTR_RO(supported_nps_configs);
+
+static const struct attribute *xcp_attrs[] = {
+	&supp_part_sysfs_mode.attr,
+	&xcp_cfg_sysfs_mode.attr,
+	NULL,
+};
+
+static void amdgpu_xcp_cfg_sysfs_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_xcp_res_details *xcp_res;
+	struct amdgpu_xcp_cfg *xcp_cfg;
+	int i, r, j, rid, mode;
+
+	if (!adev->xcp_mgr)
+		return;
+
+	xcp_cfg = kzalloc(sizeof(*xcp_cfg), GFP_KERNEL);
+	if (!xcp_cfg)
+		return;
+	xcp_cfg->xcp_mgr = adev->xcp_mgr;
+
+	r = kobject_init_and_add(&xcp_cfg->kobj, &xcp_cfg_sysfs_ktype,
+				 &adev->dev->kobj, "compute_partition_config");
+	if (r)
+		goto err1;
+
+	r = sysfs_create_files(&xcp_cfg->kobj, xcp_attrs);
+	if (r)
+		goto err1;
+
+	if (adev->gmc.supported_nps_modes != 0) {
+		r = sysfs_create_file(&xcp_cfg->kobj, &supp_nps_sysfs_mode.attr);
+		if (r) {
+			sysfs_remove_files(&xcp_cfg->kobj, xcp_attrs);
+			goto err1;
+		}
+	}
+
+	mode = (xcp_cfg->xcp_mgr->mode ==
+		AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE) ?
+		       AMDGPU_SPX_PARTITION_MODE :
+		       xcp_cfg->xcp_mgr->mode;
+	r = amdgpu_xcp_get_res_info(xcp_cfg->xcp_mgr, mode, xcp_cfg);
+	if (r) {
+		sysfs_remove_file(&xcp_cfg->kobj, &supp_nps_sysfs_mode.attr);
+		sysfs_remove_files(&xcp_cfg->kobj, xcp_attrs);
+		goto err1;
+	}
+
+	xcp_cfg->mode = mode;
+	for (i = 0; i < xcp_cfg->num_res; i++) {
+		xcp_res = &xcp_cfg->xcp_res[i];
+		rid = xcp_res->id;
+		r = kobject_init_and_add(&xcp_res->kobj,
+					 &xcp_cfg_res_sysfs_ktype,
+					 &xcp_cfg->kobj, "%s",
+					 xcp_res_names[rid]);
+		if (r)
+			goto err;
+	}
+
+	adev->xcp_mgr->xcp_cfg = xcp_cfg;
+	return;
+err:
+	for (j = 0; j < i; j++) {
+		xcp_res = &xcp_cfg->xcp_res[i];
+		kobject_put(&xcp_res->kobj);
+	}
+
+	sysfs_remove_file(&xcp_cfg->kobj, &supp_nps_sysfs_mode.attr);
+	sysfs_remove_files(&xcp_cfg->kobj, xcp_attrs);
+err1:
+	kobject_put(&xcp_cfg->kobj);
+}
+
+static void amdgpu_xcp_cfg_sysfs_fini(struct amdgpu_device *adev)
+{
+	struct amdgpu_xcp_res_details *xcp_res;
+	struct amdgpu_xcp_cfg *xcp_cfg;
+	int i;
+
+	if (!adev->xcp_mgr || !adev->xcp_mgr->xcp_cfg)
+		return;
+
+	xcp_cfg = adev->xcp_mgr->xcp_cfg;
+	for (i = 0; i < xcp_cfg->num_res; i++) {
+		xcp_res = &xcp_cfg->xcp_res[i];
+		kobject_put(&xcp_res->kobj);
+	}
+
+	sysfs_remove_file(&xcp_cfg->kobj, &supp_nps_sysfs_mode.attr);
+	sysfs_remove_files(&xcp_cfg->kobj, xcp_attrs);
+	kobject_put(&xcp_cfg->kobj);
+}
+
+/*====================== xcp sysfs - data entries ======================*/
+
+#define to_xcp(x) container_of(x, struct amdgpu_xcp, kobj)
+
+static ssize_t xcp_metrics_show(struct kobject *kobj,
+				struct kobj_attribute *attr, char *buf)
+{
+	struct amdgpu_xcp *xcp = to_xcp(kobj);
+	struct amdgpu_xcp_mgr *xcp_mgr;
+	ssize_t size;
+
+	xcp_mgr = xcp->xcp_mgr;
+	size = amdgpu_dpm_get_xcp_metrics(xcp_mgr->adev, xcp->id, NULL);
+	if (size <= 0)
+		return size;
+
+	if (size > PAGE_SIZE)
+		return -ENOSPC;
+
+	return amdgpu_dpm_get_xcp_metrics(xcp_mgr->adev, xcp->id, buf);
+}
+
+static umode_t amdgpu_xcp_attrs_is_visible(struct kobject *kobj,
+					   struct attribute *attr, int n)
+{
+	struct amdgpu_xcp *xcp = to_xcp(kobj);
+
+	if (!xcp || !xcp->valid)
+		return 0;
+
+	return attr->mode;
+}
+
+static struct kobj_attribute xcp_sysfs_metrics = __ATTR_RO(xcp_metrics);
+
+static struct attribute *amdgpu_xcp_attrs[] = {
+	&xcp_sysfs_metrics.attr,
+	NULL,
+};
+
+static const struct attribute_group amdgpu_xcp_attrs_group = {
+	.attrs = amdgpu_xcp_attrs,
+	.is_visible = amdgpu_xcp_attrs_is_visible
+};
+
+static const struct kobj_type xcp_sysfs_ktype = {
+	.sysfs_ops = &kobj_sysfs_ops,
+};
+
+static void amdgpu_xcp_sysfs_entries_fini(struct amdgpu_xcp_mgr *xcp_mgr, int n)
+{
+	struct amdgpu_xcp *xcp;
+
+	for (n--; n >= 0; n--) {
+		xcp = &xcp_mgr->xcp[n];
+		if (!xcp->ddev || !xcp->valid)
+			continue;
+		sysfs_remove_group(&xcp->kobj, &amdgpu_xcp_attrs_group);
+		kobject_put(&xcp->kobj);
+	}
+}
+
+static void amdgpu_xcp_sysfs_entries_init(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	struct amdgpu_xcp *xcp;
+	int i, r;
+
+	for (i = 0; i < MAX_XCP; i++) {
+		/* Redirect all IOCTLs to the primary device */
+		xcp = &xcp_mgr->xcp[i];
+		if (!xcp->ddev)
+			break;
+		r = kobject_init_and_add(&xcp->kobj, &xcp_sysfs_ktype,
+					 &xcp->ddev->dev->kobj, "xcp");
+		if (r)
+			goto out;
+
+		r = sysfs_create_group(&xcp->kobj, &amdgpu_xcp_attrs_group);
+		if (r)
+			goto out;
+	}
+
+	return;
+out:
+	kobject_put(&xcp->kobj);
+}
+
+static void amdgpu_xcp_sysfs_entries_update(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	struct amdgpu_xcp *xcp;
+	int i;
+
+	for (i = 0; i < MAX_XCP; i++) {
+		/* Redirect all IOCTLs to the primary device */
+		xcp = &xcp_mgr->xcp[i];
+		if (!xcp->ddev)
+			continue;
+		sysfs_update_group(&xcp->kobj, &amdgpu_xcp_attrs_group);
+	}
+
+	return;
+}
+
+void amdgpu_xcp_sysfs_init(struct amdgpu_device *adev)
+{
+	if (!adev->xcp_mgr)
+		return;
+
+	amdgpu_xcp_cfg_sysfs_init(adev);
+
+	return;
+}
+
+void amdgpu_xcp_sysfs_fini(struct amdgpu_device *adev)
+{
+	if (!adev->xcp_mgr)
+		return;
+	amdgpu_xcp_sysfs_entries_fini(adev->xcp_mgr, MAX_XCP);
+	amdgpu_xcp_cfg_sysfs_fini(adev);
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.h
new file mode 100644
index 000000000000..1928d9e224fc
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.h
@@ -0,0 +1,217 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef AMDGPU_XCP_H
+#define AMDGPU_XCP_H
+
+#include <linux/pci.h>
+#include <linux/xarray.h>
+
+#include "amdgpu_ctx.h"
+
+#define MAX_XCP 8
+
+#define AMDGPU_XCP_MODE_NONE -1
+#define AMDGPU_XCP_MODE_TRANS -2
+
+#define AMDGPU_XCP_FL_NONE 0
+#define AMDGPU_XCP_FL_LOCKED (1 << 0)
+
+#define AMDGPU_XCP_NO_PARTITION (~0)
+
+#define AMDGPU_XCP_OPS_KFD	(1 << 0)
+
+struct amdgpu_fpriv;
+
+enum AMDGPU_XCP_IP_BLOCK {
+	AMDGPU_XCP_GFXHUB,
+	AMDGPU_XCP_GFX,
+	AMDGPU_XCP_SDMA,
+	AMDGPU_XCP_VCN,
+	AMDGPU_XCP_MAX_BLOCKS
+};
+
+enum AMDGPU_XCP_STATE {
+	AMDGPU_XCP_PREPARE_SUSPEND,
+	AMDGPU_XCP_SUSPEND,
+	AMDGPU_XCP_PREPARE_RESUME,
+	AMDGPU_XCP_RESUME,
+};
+
+enum amdgpu_xcp_res_id {
+	AMDGPU_XCP_RES_XCC,
+	AMDGPU_XCP_RES_DMA,
+	AMDGPU_XCP_RES_DEC,
+	AMDGPU_XCP_RES_JPEG,
+	AMDGPU_XCP_RES_MAX,
+};
+
+struct amdgpu_xcp_res_details {
+	enum amdgpu_xcp_res_id id;
+	u8 num_inst;
+	u8 num_shared;
+	struct kobject kobj;
+};
+
+struct amdgpu_xcp_cfg {
+	u8 mode;
+	struct amdgpu_xcp_res_details xcp_res[AMDGPU_XCP_RES_MAX];
+	u8 num_res;
+	struct amdgpu_xcp_mgr *xcp_mgr;
+	struct kobject kobj;
+	u16 compatible_nps_modes;
+};
+
+struct amdgpu_xcp_ip_funcs {
+	int (*prepare_suspend)(void *handle, uint32_t inst_mask);
+	int (*suspend)(void *handle, uint32_t inst_mask);
+	int (*prepare_resume)(void *handle, uint32_t inst_mask);
+	int (*resume)(void *handle, uint32_t inst_mask);
+};
+
+struct amdgpu_xcp_ip {
+	struct amdgpu_xcp_ip_funcs *ip_funcs;
+	uint32_t inst_mask;
+
+	enum AMDGPU_XCP_IP_BLOCK ip_id;
+	bool valid;
+};
+
+struct amdgpu_xcp {
+	struct amdgpu_xcp_ip ip[AMDGPU_XCP_MAX_BLOCKS];
+
+	uint8_t id;
+	uint8_t mem_id;
+	bool valid;
+	atomic_t	ref_cnt;
+	struct drm_device *ddev;
+	struct drm_device *rdev;
+	struct drm_device *pdev;
+	struct drm_driver *driver;
+	struct drm_vma_offset_manager *vma_offset_manager;
+	struct amdgpu_sched	gpu_sched[AMDGPU_HW_IP_NUM][AMDGPU_RING_PRIO_MAX];
+	struct amdgpu_xcp_mgr *xcp_mgr;
+	struct kobject kobj;
+	uint64_t unique_id;
+};
+
+struct amdgpu_xcp_mgr {
+	struct amdgpu_device *adev;
+	struct mutex xcp_lock;
+	struct amdgpu_xcp_mgr_funcs *funcs;
+
+	struct amdgpu_xcp xcp[MAX_XCP];
+	uint8_t num_xcps;
+	int8_t mode;
+
+	 /* Used to determine KFD memory size limits per XCP */
+	unsigned int num_xcp_per_mem_partition;
+	struct amdgpu_xcp_cfg *xcp_cfg;
+	uint32_t supp_xcp_modes;
+	uint32_t avail_xcp_modes;
+};
+
+struct amdgpu_xcp_mgr_funcs {
+	int (*switch_partition_mode)(struct amdgpu_xcp_mgr *xcp_mgr, int mode,
+				     int *num_xcps);
+	int (*query_partition_mode)(struct amdgpu_xcp_mgr *xcp_mgr);
+	int (*get_ip_details)(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id,
+			      enum AMDGPU_XCP_IP_BLOCK ip_id,
+			      struct amdgpu_xcp_ip *ip);
+	int (*get_xcp_mem_id)(struct amdgpu_xcp_mgr *xcp_mgr,
+			      struct amdgpu_xcp *xcp, uint8_t *mem_id);
+	int (*get_xcp_res_info)(struct amdgpu_xcp_mgr *xcp_mgr,
+				int mode,
+				struct amdgpu_xcp_cfg *xcp_cfg);
+	int (*prepare_suspend)(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id);
+	int (*suspend)(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id);
+	int (*prepare_resume)(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id);
+	int (*resume)(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id);
+};
+
+int amdgpu_xcp_prepare_suspend(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id);
+int amdgpu_xcp_suspend(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id);
+int amdgpu_xcp_prepare_resume(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id);
+int amdgpu_xcp_resume(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id);
+
+int amdgpu_xcp_mgr_init(struct amdgpu_device *adev, int init_mode,
+			int init_xcps, struct amdgpu_xcp_mgr_funcs *xcp_funcs);
+int amdgpu_xcp_init(struct amdgpu_xcp_mgr *xcp_mgr, int num_xcps, int mode);
+int amdgpu_xcp_query_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr, u32 flags);
+int amdgpu_xcp_switch_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr, int mode);
+int amdgpu_xcp_restore_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr);
+int amdgpu_xcp_get_partition(struct amdgpu_xcp_mgr *xcp_mgr,
+			     enum AMDGPU_XCP_IP_BLOCK ip, int instance);
+
+int amdgpu_xcp_get_inst_details(struct amdgpu_xcp *xcp,
+				enum AMDGPU_XCP_IP_BLOCK ip,
+				uint32_t *inst_mask);
+
+int amdgpu_xcp_dev_register(struct amdgpu_device *adev,
+				const struct pci_device_id *ent);
+void amdgpu_xcp_dev_unplug(struct amdgpu_device *adev);
+int amdgpu_xcp_open_device(struct amdgpu_device *adev,
+			   struct amdgpu_fpriv *fpriv,
+			   struct drm_file *file_priv);
+void amdgpu_xcp_release_sched(struct amdgpu_device *adev,
+			      struct amdgpu_ctx_entity *entity);
+int amdgpu_xcp_select_scheds(struct amdgpu_device *adev,
+			     u32 hw_ip, u32 hw_prio,
+			     struct amdgpu_fpriv *fpriv,
+			     unsigned int *num_scheds,
+			     struct drm_gpu_scheduler ***scheds);
+void amdgpu_xcp_update_supported_modes(struct amdgpu_xcp_mgr *xcp_mgr);
+int amdgpu_xcp_update_partition_sched_list(struct amdgpu_device *adev);
+int amdgpu_xcp_pre_partition_switch(struct amdgpu_xcp_mgr *xcp_mgr, u32 flags);
+int amdgpu_xcp_post_partition_switch(struct amdgpu_xcp_mgr *xcp_mgr, u32 flags);
+void amdgpu_xcp_sysfs_init(struct amdgpu_device *adev);
+void amdgpu_xcp_sysfs_fini(struct amdgpu_device *adev);
+
+static inline int amdgpu_xcp_get_num_xcp(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	if (!xcp_mgr)
+		return 1;
+	else
+		return xcp_mgr->num_xcps;
+}
+
+static inline struct amdgpu_xcp *
+amdgpu_get_next_xcp(struct amdgpu_xcp_mgr *xcp_mgr, int *from)
+{
+	if (!xcp_mgr)
+		return NULL;
+
+	while (*from < MAX_XCP) {
+		if (xcp_mgr->xcp[*from].valid)
+			return &xcp_mgr->xcp[*from];
+		++(*from);
+	}
+
+	return NULL;
+}
+
+#define for_each_xcp(xcp_mgr, xcp, i)                            \
+	for (i = 0, xcp = amdgpu_get_next_xcp(xcp_mgr, &i); xcp; \
+	     ++i, xcp = amdgpu_get_next_xcp(xcp_mgr, &i))
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c
index 439925477fb8..aad530c46a9f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c
@@ -40,6 +40,11 @@
 #define smnPCS_GOPX1_PCS_ERROR_STATUS    0x12200210
 #define smnPCS_GOPX1_PCS_ERROR_NONCORRECTABLE_MASK      0x12200218
 
+#define XGMI_STATE_DISABLE                      0xD1
+#define XGMI_STATE_LS0                          0x81
+#define XGMI_LINK_ACTIVE			1
+#define XGMI_LINK_INACTIVE			0
+
 static DEFINE_MUTEX(xgmi_mutex);
 
 #define AMDGPU_MAX_XGMI_DEVICE_PER_HIVE		4
@@ -103,6 +108,53 @@ static const int walf_pcs_err_noncorrectable_mask_reg_aldebaran[] = {
 	smnPCS_GOPX1_PCS_ERROR_NONCORRECTABLE_MASK + 0x100000
 };
 
+static const int xgmi3x16_pcs_err_status_reg_v6_4[] = {
+	smnPCS_XGMI3X16_PCS_ERROR_STATUS,
+	smnPCS_XGMI3X16_PCS_ERROR_STATUS + 0x100000
+};
+
+static const int xgmi3x16_pcs_err_noncorrectable_mask_reg_v6_4[] = {
+	smnPCS_XGMI3X16_PCS_ERROR_NONCORRECTABLE_MASK,
+	smnPCS_XGMI3X16_PCS_ERROR_NONCORRECTABLE_MASK + 0x100000
+};
+
+static const u64 xgmi_v6_4_0_mca_base_array[] = {
+	0x11a09200,
+	0x11b09200,
+};
+
+static const char *xgmi_v6_4_0_ras_error_code_ext[32] = {
+	[0x00] = "XGMI PCS DataLossErr",
+	[0x01] = "XGMI PCS TrainingErr",
+	[0x02] = "XGMI PCS FlowCtrlAckErr",
+	[0x03] = "XGMI PCS RxFifoUnderflowErr",
+	[0x04] = "XGMI PCS RxFifoOverflowErr",
+	[0x05] = "XGMI PCS CRCErr",
+	[0x06] = "XGMI PCS BERExceededErr",
+	[0x07] = "XGMI PCS TxMetaDataErr",
+	[0x08] = "XGMI PCS ReplayBufParityErr",
+	[0x09] = "XGMI PCS DataParityErr",
+	[0x0a] = "XGMI PCS ReplayFifoOverflowErr",
+	[0x0b] = "XGMI PCS ReplayFifoUnderflowErr",
+	[0x0c] = "XGMI PCS ElasticFifoOverflowErr",
+	[0x0d] = "XGMI PCS DeskewErr",
+	[0x0e] = "XGMI PCS FlowCtrlCRCErr",
+	[0x0f] = "XGMI PCS DataStartupLimitErr",
+	[0x10] = "XGMI PCS FCInitTimeoutErr",
+	[0x11] = "XGMI PCS RecoveryTimeoutErr",
+	[0x12] = "XGMI PCS ReadySerialTimeoutErr",
+	[0x13] = "XGMI PCS ReadySerialAttemptErr",
+	[0x14] = "XGMI PCS RecoveryAttemptErr",
+	[0x15] = "XGMI PCS RecoveryRelockAttemptErr",
+	[0x16] = "XGMI PCS ReplayAttemptErr",
+	[0x17] = "XGMI PCS SyncHdrErr",
+	[0x18] = "XGMI PCS TxReplayTimeoutErr",
+	[0x19] = "XGMI PCS RxReplayTimeoutErr",
+	[0x1a] = "XGMI PCS LinkSubTxTimeoutErr",
+	[0x1b] = "XGMI PCS LinkSubRxTimeoutErr",
+	[0x1c] = "XGMI PCS RxCMDPktErr",
+};
+
 static const struct amdgpu_pcs_ras_field xgmi_pcs_ras_fields[] = {
 	{"XGMI PCS DataLossErr",
 	 SOC15_REG_FIELD(XGMI0_PCS_GOPX16_PCS_ERROR_STATUS, DataLossErr)},
@@ -242,6 +294,82 @@ static const struct amdgpu_pcs_ras_field xgmi3x16_pcs_ras_fields[] = {
 	 SOC15_REG_FIELD(PCS_XGMI3X16_PCS_ERROR_STATUS, RxCMDPktErr)},
 };
 
+int amdgpu_xgmi_get_ext_link(struct amdgpu_device *adev, int link_num)
+{
+	int link_map_6_4_x[8] = { 0, 3, 1, 2, 7, 6, 4, 5 };
+
+	if (adev->gmc.xgmi.num_physical_nodes <= 1)
+		return -EINVAL;
+
+	switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) {
+	case IP_VERSION(6, 4, 0):
+	case IP_VERSION(6, 4, 1):
+		if (link_num < ARRAY_SIZE(link_map_6_4_x))
+			return link_map_6_4_x[link_num];
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return -EINVAL;
+}
+
+static u32 xgmi_v6_4_get_link_status(struct amdgpu_device *adev, int global_link_num)
+{
+	const u32 smn_xgmi_6_4_pcs_state_hist1[2] = { 0x11a00070, 0x11b00070 };
+	const u32 smn_xgmi_6_4_1_pcs_state_hist1[2] = { 0x12100070,
+							0x11b00070 };
+	u32 i, n;
+	u64 addr;
+
+	switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) {
+	case IP_VERSION(6, 4, 0):
+		n = ARRAY_SIZE(smn_xgmi_6_4_pcs_state_hist1);
+		addr = smn_xgmi_6_4_pcs_state_hist1[global_link_num % n];
+		break;
+	case IP_VERSION(6, 4, 1):
+		n = ARRAY_SIZE(smn_xgmi_6_4_1_pcs_state_hist1);
+		addr = smn_xgmi_6_4_1_pcs_state_hist1[global_link_num % n];
+		break;
+	default:
+		return U32_MAX;
+	}
+
+	i = global_link_num / n;
+
+	if (!(adev->aid_mask & BIT(i)))
+		return U32_MAX;
+
+	addr += adev->asic_funcs->encode_ext_smn_addressing(i);
+
+	return RREG32_PCIE_EXT(addr);
+}
+
+int amdgpu_get_xgmi_link_status(struct amdgpu_device *adev, int global_link_num)
+{
+	u32 xgmi_state_reg_val;
+
+	if (adev->gmc.xgmi.num_physical_nodes <= 1)
+		return -EINVAL;
+
+	switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) {
+	case IP_VERSION(6, 4, 0):
+	case IP_VERSION(6, 4, 1):
+		xgmi_state_reg_val = xgmi_v6_4_get_link_status(adev, global_link_num);
+		break;
+	default:
+		return -EOPNOTSUPP;
+	}
+
+	if ((xgmi_state_reg_val & 0xFF) == XGMI_STATE_DISABLE)
+		return -ENOLINK;
+
+	if ((xgmi_state_reg_val & 0xFF) == XGMI_STATE_LS0)
+		return XGMI_LINK_ACTIVE;
+
+	return XGMI_LINK_INACTIVE;
+}
+
 /**
  * DOC: AMDGPU XGMI Support
  *
@@ -325,6 +453,82 @@ static ssize_t amdgpu_xgmi_show_device_id(struct device *dev,
 
 }
 
+static ssize_t amdgpu_xgmi_show_physical_id(struct device *dev,
+				     struct device_attribute *attr,
+				     char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	return sysfs_emit(buf, "%u\n", adev->gmc.xgmi.physical_node_id);
+
+}
+
+static ssize_t amdgpu_xgmi_show_num_hops(struct device *dev,
+					struct device_attribute *attr,
+					char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	struct psp_xgmi_topology_info *top = &adev->psp.xgmi_context.top_info;
+	int i;
+
+	for (i = 0; i < top->num_nodes; i++)
+		sprintf(buf + 3 * i, "%02x ", top->nodes[i].num_hops);
+
+	return sysfs_emit(buf, "%s\n", buf);
+}
+
+static ssize_t amdgpu_xgmi_show_num_links(struct device *dev,
+					struct device_attribute *attr,
+					char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	struct psp_xgmi_topology_info *top = &adev->psp.xgmi_context.top_info;
+	int i;
+
+	for (i = 0; i < top->num_nodes; i++)
+		sprintf(buf + 3 * i, "%02x ", top->nodes[i].num_links);
+
+	return sysfs_emit(buf, "%s\n", buf);
+}
+
+static ssize_t amdgpu_xgmi_show_connected_port_num(struct device *dev,
+					struct device_attribute *attr,
+					char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	struct psp_xgmi_topology_info *top = &adev->psp.xgmi_context.top_info;
+	int i, j, size = 0;
+	int current_node;
+	/*
+	 * get the node id in the sysfs for the current socket and show
+	 * it in the port num info output in the sysfs for easy reading.
+	 * it is NOT the one retrieved from xgmi ta.
+	 */
+	for (i = 0; i < top->num_nodes; i++) {
+		if (top->nodes[i].node_id == adev->gmc.xgmi.node_id) {
+			current_node = i;
+			break;
+		}
+	}
+
+	if (i == top->num_nodes)
+		return -EINVAL;
+
+	for (i = 0; i < top->num_nodes; i++) {
+		for (j = 0; j < top->nodes[i].num_links; j++)
+			/* node id in sysfs starts from 1 rather than 0 so +1 here */
+			size += sysfs_emit_at(buf, size, "%02x:%02x ->  %02x:%02x\n", current_node + 1,
+					      top->nodes[i].port_num[j].src_xgmi_port_num, i + 1,
+					      top->nodes[i].port_num[j].dst_xgmi_port_num);
+	}
+
+	return size;
+}
+
 #define AMDGPU_XGMI_SET_FICAA(o)	((o) | 0x456801)
 static ssize_t amdgpu_xgmi_show_error(struct device *dev,
 				      struct device_attribute *attr,
@@ -360,7 +564,11 @@ static ssize_t amdgpu_xgmi_show_error(struct device *dev,
 
 
 static DEVICE_ATTR(xgmi_device_id, S_IRUGO, amdgpu_xgmi_show_device_id, NULL);
+static DEVICE_ATTR(xgmi_physical_id, 0444, amdgpu_xgmi_show_physical_id, NULL);
 static DEVICE_ATTR(xgmi_error, S_IRUGO, amdgpu_xgmi_show_error, NULL);
+static DEVICE_ATTR(xgmi_num_hops, S_IRUGO, amdgpu_xgmi_show_num_hops, NULL);
+static DEVICE_ATTR(xgmi_num_links, S_IRUGO, amdgpu_xgmi_show_num_links, NULL);
+static DEVICE_ATTR(xgmi_port_num, S_IRUGO, amdgpu_xgmi_show_connected_port_num, NULL);
 
 static int amdgpu_xgmi_sysfs_add_dev_info(struct amdgpu_device *adev,
 					 struct amdgpu_hive_info *hive)
@@ -375,11 +583,33 @@ static int amdgpu_xgmi_sysfs_add_dev_info(struct amdgpu_device *adev,
 		return ret;
 	}
 
+	ret = device_create_file(adev->dev, &dev_attr_xgmi_physical_id);
+	if (ret) {
+		dev_err(adev->dev, "XGMI: Failed to create device file xgmi_physical_id\n");
+		return ret;
+	}
+
 	/* Create xgmi error file */
 	ret = device_create_file(adev->dev, &dev_attr_xgmi_error);
 	if (ret)
 		pr_err("failed to create xgmi_error\n");
 
+	/* Create xgmi num hops file */
+	ret = device_create_file(adev->dev, &dev_attr_xgmi_num_hops);
+	if (ret)
+		pr_err("failed to create xgmi_num_hops\n");
+
+	/* Create xgmi num links file */
+	ret = device_create_file(adev->dev, &dev_attr_xgmi_num_links);
+	if (ret)
+		pr_err("failed to create xgmi_num_links\n");
+
+	/* Create xgmi port num file if supported */
+	if (adev->psp.xgmi_context.xgmi_ta_caps & EXTEND_PEER_LINK_INFO_CMD_FLAG) {
+		ret = device_create_file(adev->dev, &dev_attr_xgmi_port_num);
+		if (ret)
+			dev_err(adev->dev, "failed to create xgmi_port_num\n");
+	}
 
 	/* Create sysfs link to hive info folder on the first device */
 	if (hive->kobj.parent != (&adev->dev->kobj)) {
@@ -407,6 +637,12 @@ remove_link:
 
 remove_file:
 	device_remove_file(adev->dev, &dev_attr_xgmi_device_id);
+	device_remove_file(adev->dev, &dev_attr_xgmi_physical_id);
+	device_remove_file(adev->dev, &dev_attr_xgmi_error);
+	device_remove_file(adev->dev, &dev_attr_xgmi_num_hops);
+	device_remove_file(adev->dev, &dev_attr_xgmi_num_links);
+	if (adev->psp.xgmi_context.xgmi_ta_caps & EXTEND_PEER_LINK_INFO_CMD_FLAG)
+		device_remove_file(adev->dev, &dev_attr_xgmi_port_num);
 
 success:
 	return ret;
@@ -419,7 +655,12 @@ static void amdgpu_xgmi_sysfs_rem_dev_info(struct amdgpu_device *adev,
 	memset(node, 0, sizeof(node));
 
 	device_remove_file(adev->dev, &dev_attr_xgmi_device_id);
+	device_remove_file(adev->dev, &dev_attr_xgmi_physical_id);
 	device_remove_file(adev->dev, &dev_attr_xgmi_error);
+	device_remove_file(adev->dev, &dev_attr_xgmi_num_hops);
+	device_remove_file(adev->dev, &dev_attr_xgmi_num_links);
+	if (adev->psp.xgmi_context.xgmi_ta_caps & EXTEND_PEER_LINK_INFO_CMD_FLAG)
+		device_remove_file(adev->dev, &dev_attr_xgmi_port_num);
 
 	if (hive->kobj.parent != (&adev->dev->kobj))
 		sysfs_remove_link(&adev->dev->kobj,"xgmi_hive_info");
@@ -454,6 +695,7 @@ struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev)
 	hive = kzalloc(sizeof(*hive), GFP_KERNEL);
 	if (!hive) {
 		dev_err(adev->dev, "XGMI: allocation failed\n");
+		ret = -ENOMEM;
 		hive = NULL;
 		goto pro_end;
 	}
@@ -506,6 +748,7 @@ struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev)
 	task_barrier_init(&hive->tb);
 	hive->pstate = AMDGPU_XGMI_PSTATE_UNKNOWN;
 	hive->hi_req_gpu = NULL;
+	atomic_set(&hive->requested_nps_mode, UNKNOWN_MEMORY_PARTITION_MODE);
 
 	/*
 	 * hive pstate on boot is high in vega20 so we have to go to low
@@ -615,28 +858,88 @@ int amdgpu_xgmi_update_topology(struct amdgpu_hive_info *hive, struct amdgpu_dev
  * num_hops[2:0] = number of hops
  */
 int amdgpu_xgmi_get_hops_count(struct amdgpu_device *adev,
-		struct amdgpu_device *peer_adev)
+			       struct amdgpu_device *peer_adev)
 {
 	struct psp_xgmi_topology_info *top = &adev->psp.xgmi_context.top_info;
 	uint8_t num_hops_mask = 0x7;
 	int i;
 
+	if (!adev->gmc.xgmi.supported)
+		return 0;
+
 	for (i = 0 ; i < top->num_nodes; ++i)
 		if (top->nodes[i].node_id == peer_adev->gmc.xgmi.node_id)
 			return top->nodes[i].num_hops & num_hops_mask;
-	return	-EINVAL;
+
+	dev_err(adev->dev, "Failed to get xgmi hops count for peer %d.\n",
+		peer_adev->gmc.xgmi.physical_node_id);
+
+	return 0;
 }
 
-int amdgpu_xgmi_get_num_links(struct amdgpu_device *adev,
-		struct amdgpu_device *peer_adev)
+int amdgpu_xgmi_get_bandwidth(struct amdgpu_device *adev, struct amdgpu_device *peer_adev,
+			      enum amdgpu_xgmi_bw_mode bw_mode, enum amdgpu_xgmi_bw_unit bw_unit,
+			      uint32_t *min_bw, uint32_t *max_bw)
+{
+	bool peer_mode = bw_mode == AMDGPU_XGMI_BW_MODE_PER_PEER;
+	int unit_scale = bw_unit == AMDGPU_XGMI_BW_UNIT_MBYTES ? 1000 : 1;
+	int num_lanes = adev->gmc.xgmi.max_width;
+	int speed = adev->gmc.xgmi.max_speed;
+	int num_links = !peer_mode ? 1 : -1;
+
+	if (!(min_bw && max_bw))
+		return -EINVAL;
+
+	*min_bw = 0;
+	*max_bw = 0;
+
+	if (!adev->gmc.xgmi.supported)
+		return -ENODATA;
+
+	if (peer_mode && !peer_adev)
+		return -EINVAL;
+
+	if (peer_mode) {
+		struct psp_xgmi_topology_info *top = &adev->psp.xgmi_context.top_info;
+		int i;
+
+		for (i = 0 ; i < top->num_nodes; ++i) {
+			if (top->nodes[i].node_id != peer_adev->gmc.xgmi.node_id)
+				continue;
+
+			num_links =  top->nodes[i].num_links;
+			break;
+		}
+	}
+
+	if (num_links == -1) {
+		dev_err(adev->dev, "Failed to get number of xgmi links for peer %d.\n",
+			peer_adev->gmc.xgmi.physical_node_id);
+	} else if (num_links) {
+		int per_link_bw = (speed * num_lanes * unit_scale)/BITS_PER_BYTE;
+
+		*min_bw = per_link_bw;
+		*max_bw = num_links * per_link_bw;
+	}
+
+	return 0;
+}
+
+bool amdgpu_xgmi_get_is_sharing_enabled(struct amdgpu_device *adev,
+					struct amdgpu_device *peer_adev)
 {
 	struct psp_xgmi_topology_info *top = &adev->psp.xgmi_context.top_info;
 	int i;
 
+	/* Sharing should always be enabled for non-SRIOV. */
+	if (!amdgpu_sriov_vf(adev))
+		return true;
+
 	for (i = 0 ; i < top->num_nodes; ++i)
 		if (top->nodes[i].node_id == peer_adev->gmc.xgmi.node_id)
-			return top->nodes[i].num_links;
-	return	-EINVAL;
+			return !!top->nodes[i].is_sharing_enabled;
+
+	return false;
 }
 
 /*
@@ -677,8 +980,7 @@ int amdgpu_xgmi_add_device(struct amdgpu_device *adev)
 	if (!adev->gmc.xgmi.supported)
 		return 0;
 
-	if (!adev->gmc.xgmi.pending_reset &&
-	    amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_PSP)) {
+	if (amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_PSP)) {
 		ret = psp_xgmi_initialize(&adev->psp, false, true);
 		if (ret) {
 			dev_err(adev->dev,
@@ -724,8 +1026,7 @@ int amdgpu_xgmi_add_device(struct amdgpu_device *adev)
 
 	task_barrier_add_task(&hive->tb);
 
-	if (!adev->gmc.xgmi.pending_reset &&
-	    amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_PSP)) {
+	if (amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_PSP)) {
 		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
 			/* update node list for other device in the hive */
 			if (tmp_adev != adev) {
@@ -739,18 +1040,33 @@ int amdgpu_xgmi_add_device(struct amdgpu_device *adev)
 				goto exit_unlock;
 		}
 
-		/* get latest topology info for each device from psp */
-		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
-			ret = psp_xgmi_get_topology_info(&tmp_adev->psp, count,
-					&tmp_adev->psp.xgmi_context.top_info, false);
+		if (amdgpu_sriov_vf(adev) &&
+			adev->psp.xgmi_context.xgmi_ta_caps & EXTEND_PEER_LINK_INFO_CMD_FLAG) {
+			/* only get topology for VF being init if it can support full duplex */
+			ret = psp_xgmi_get_topology_info(&adev->psp, count,
+						&adev->psp.xgmi_context.top_info, false);
 			if (ret) {
-				dev_err(tmp_adev->dev,
+				dev_err(adev->dev,
 					"XGMI: Get topology failure on device %llx, hive %llx, ret %d",
-					tmp_adev->gmc.xgmi.node_id,
-					tmp_adev->gmc.xgmi.hive_id, ret);
-				/* To do : continue with some node failed or disable the whole hive */
+					adev->gmc.xgmi.node_id,
+					adev->gmc.xgmi.hive_id, ret);
+				/* To do: continue with some node failed or disable the whole hive*/
 				goto exit_unlock;
 			}
+		} else {
+			/* get latest topology info for each device from psp */
+			list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
+				ret = psp_xgmi_get_topology_info(&tmp_adev->psp, count,
+					&tmp_adev->psp.xgmi_context.top_info, false);
+				if (ret) {
+					dev_err(tmp_adev->dev,
+						"XGMI: Get topology failure on device %llx, hive %llx, ret %d",
+						tmp_adev->gmc.xgmi.node_id,
+						tmp_adev->gmc.xgmi.hive_id, ret);
+					/* To do : continue with some node failed or disable the whole hive */
+					goto exit_unlock;
+				}
+			}
 		}
 
 		/* get topology again for hives that support extended data */
@@ -782,7 +1098,7 @@ int amdgpu_xgmi_add_device(struct amdgpu_device *adev)
 		}
 	}
 
-	if (!ret && !adev->gmc.xgmi.pending_reset)
+	if (!ret)
 		ret = amdgpu_xgmi_sysfs_add_dev_info(adev, hive);
 
 exit_unlock:
@@ -835,15 +1151,91 @@ int amdgpu_xgmi_remove_device(struct amdgpu_device *adev)
 	return 0;
 }
 
+static int xgmi_v6_4_0_aca_bank_parser(struct aca_handle *handle, struct aca_bank *bank,
+				       enum aca_smu_type type, void *data)
+{
+	struct amdgpu_device *adev = handle->adev;
+	struct aca_bank_info info;
+	const char *error_str;
+	u64 status, count;
+	int ret, ext_error_code;
+
+	ret = aca_bank_info_decode(bank, &info);
+	if (ret)
+		return ret;
+
+	status = bank->regs[ACA_REG_IDX_STATUS];
+	ext_error_code = ACA_REG__STATUS__ERRORCODEEXT(status);
+
+	error_str = ext_error_code < ARRAY_SIZE(xgmi_v6_4_0_ras_error_code_ext) ?
+		xgmi_v6_4_0_ras_error_code_ext[ext_error_code] : NULL;
+	if (error_str)
+		dev_info(adev->dev, "%s detected\n", error_str);
+
+	count = ACA_REG__MISC0__ERRCNT(bank->regs[ACA_REG_IDX_MISC0]);
+
+	switch (type) {
+	case ACA_SMU_TYPE_UE:
+		if (ext_error_code != 0 && ext_error_code != 9)
+			count = 0ULL;
+
+		bank->aca_err_type = ACA_ERROR_TYPE_UE;
+		ret = aca_error_cache_log_bank_error(handle, &info, ACA_ERROR_TYPE_UE, count);
+		break;
+	case ACA_SMU_TYPE_CE:
+		count = ext_error_code == 6 ? count : 0ULL;
+		bank->aca_err_type = ACA_ERROR_TYPE_CE;
+		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type, count);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return ret;
+}
+
+static const struct aca_bank_ops xgmi_v6_4_0_aca_bank_ops = {
+	.aca_bank_parser = xgmi_v6_4_0_aca_bank_parser,
+};
+
+static const struct aca_info xgmi_v6_4_0_aca_info = {
+	.hwip = ACA_HWIP_TYPE_PCS_XGMI,
+	.mask = ACA_ERROR_UE_MASK | ACA_ERROR_CE_MASK,
+	.bank_ops = &xgmi_v6_4_0_aca_bank_ops,
+};
+
 static int amdgpu_xgmi_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
 {
+	int r;
+
 	if (!adev->gmc.xgmi.supported ||
 	    adev->gmc.xgmi.num_physical_nodes == 0)
 		return 0;
 
-	adev->gmc.xgmi.ras->ras_block.hw_ops->reset_ras_error_count(adev);
+	amdgpu_ras_reset_error_count(adev, AMDGPU_RAS_BLOCK__XGMI_WAFL);
+
+	r = amdgpu_ras_block_late_init(adev, ras_block);
+	if (r)
+		return r;
+
+	switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) {
+	case IP_VERSION(6, 4, 0):
+	case IP_VERSION(6, 4, 1):
+		r = amdgpu_ras_bind_aca(adev, AMDGPU_RAS_BLOCK__XGMI_WAFL,
+					&xgmi_v6_4_0_aca_info, NULL);
+		if (r)
+			goto late_fini;
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+
+late_fini:
+	amdgpu_ras_block_late_fini(adev, ras_block);
 
-	return amdgpu_ras_block_late_init(adev, ras_block);
+	return r;
 }
 
 uint64_t amdgpu_xgmi_get_relative_phy_addr(struct amdgpu_device *adev,
@@ -859,7 +1251,7 @@ static void pcs_clear_status(struct amdgpu_device *adev, uint32_t pcs_status_reg
 	WREG32_PCIE(pcs_status_reg, 0);
 }
 
-static void amdgpu_xgmi_reset_ras_error_count(struct amdgpu_device *adev)
+static void amdgpu_xgmi_legacy_reset_ras_error_count(struct amdgpu_device *adev)
 {
 	uint32_t i;
 
@@ -885,6 +1277,51 @@ static void amdgpu_xgmi_reset_ras_error_count(struct amdgpu_device *adev)
 	default:
 		break;
 	}
+
+	switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) {
+	case IP_VERSION(6, 4, 0):
+	case IP_VERSION(6, 4, 1):
+		for (i = 0; i < ARRAY_SIZE(xgmi3x16_pcs_err_status_reg_v6_4); i++)
+			pcs_clear_status(adev,
+					xgmi3x16_pcs_err_status_reg_v6_4[i]);
+		break;
+	default:
+		break;
+	}
+}
+
+static void __xgmi_v6_4_0_reset_error_count(struct amdgpu_device *adev, int xgmi_inst, u64 mca_base)
+{
+	WREG64_MCA(xgmi_inst, mca_base, ACA_REG_IDX_STATUS, 0ULL);
+}
+
+static void xgmi_v6_4_0_reset_error_count(struct amdgpu_device *adev, int xgmi_inst)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(xgmi_v6_4_0_mca_base_array); i++)
+		__xgmi_v6_4_0_reset_error_count(adev, xgmi_inst, xgmi_v6_4_0_mca_base_array[i]);
+}
+
+static void xgmi_v6_4_0_reset_ras_error_count(struct amdgpu_device *adev)
+{
+	int i;
+
+	for_each_inst(i, adev->aid_mask)
+		xgmi_v6_4_0_reset_error_count(adev, i);
+}
+
+static void amdgpu_xgmi_reset_ras_error_count(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) {
+	case IP_VERSION(6, 4, 0):
+	case IP_VERSION(6, 4, 1):
+		xgmi_v6_4_0_reset_ras_error_count(adev);
+		break;
+	default:
+		amdgpu_xgmi_legacy_reset_ras_error_count(adev);
+		break;
+	}
 }
 
 static int amdgpu_xgmi_query_pcs_error_status(struct amdgpu_device *adev,
@@ -901,7 +1338,12 @@ static int amdgpu_xgmi_query_pcs_error_status(struct amdgpu_device *adev,
 	uint32_t field_array_size = 0;
 
 	if (is_xgmi_pcs) {
-		if (adev->ip_versions[XGMI_HWIP][0] == IP_VERSION(6, 1, 0)) {
+		if (amdgpu_ip_version(adev, XGMI_HWIP, 0) ==
+		    IP_VERSION(6, 1, 0) ||
+		    amdgpu_ip_version(adev, XGMI_HWIP, 0) ==
+		    IP_VERSION(6, 4, 0) ||
+		    amdgpu_ip_version(adev, XGMI_HWIP, 0) ==
+		    IP_VERSION(6, 4, 1)) {
 			pcs_ras_fields = &xgmi3x16_pcs_ras_fields[0];
 			field_array_size = ARRAY_SIZE(xgmi3x16_pcs_ras_fields);
 		} else {
@@ -935,11 +1377,11 @@ static int amdgpu_xgmi_query_pcs_error_status(struct amdgpu_device *adev,
 	return 0;
 }
 
-static void amdgpu_xgmi_query_ras_error_count(struct amdgpu_device *adev,
-					     void *ras_error_status)
+static void amdgpu_xgmi_legacy_query_ras_error_count(struct amdgpu_device *adev,
+						     void *ras_error_status)
 {
 	struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
-	int i;
+	int i, supported = 1;
 	uint32_t data, mask_data = 0;
 	uint32_t ue_cnt = 0, ce_cnt = 0;
 
@@ -1003,41 +1445,146 @@ static void amdgpu_xgmi_query_ras_error_count(struct amdgpu_device *adev,
 		}
 		break;
 	default:
-		dev_warn(adev->dev, "XGMI RAS error query not supported");
+		supported = 0;
+		break;
+	}
+
+	switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) {
+	case IP_VERSION(6, 4, 0):
+	case IP_VERSION(6, 4, 1):
+		/* check xgmi3x16 pcs error */
+		for (i = 0; i < ARRAY_SIZE(xgmi3x16_pcs_err_status_reg_v6_4); i++) {
+			data = RREG32_PCIE(xgmi3x16_pcs_err_status_reg_v6_4[i]);
+			mask_data =
+				RREG32_PCIE(xgmi3x16_pcs_err_noncorrectable_mask_reg_v6_4[i]);
+			if (data)
+				amdgpu_xgmi_query_pcs_error_status(adev, data,
+						mask_data, &ue_cnt, &ce_cnt, true, true);
+		}
+		break;
+	default:
+		if (!supported)
+			dev_warn(adev->dev, "XGMI RAS error query not supported");
 		break;
 	}
 
-	adev->gmc.xgmi.ras->ras_block.hw_ops->reset_ras_error_count(adev);
+	amdgpu_ras_reset_error_count(adev, AMDGPU_RAS_BLOCK__XGMI_WAFL);
 
 	err_data->ue_count += ue_cnt;
 	err_data->ce_count += ce_cnt;
 }
 
+static enum aca_error_type xgmi_v6_4_0_pcs_mca_get_error_type(struct amdgpu_device *adev, u64 status)
+{
+	const char *error_str;
+	int ext_error_code;
+
+	ext_error_code = ACA_REG__STATUS__ERRORCODEEXT(status);
+
+	error_str = ext_error_code < ARRAY_SIZE(xgmi_v6_4_0_ras_error_code_ext) ?
+		xgmi_v6_4_0_ras_error_code_ext[ext_error_code] : NULL;
+	if (error_str)
+		dev_info(adev->dev, "%s detected\n", error_str);
+
+	switch (ext_error_code) {
+	case 0:
+		return ACA_ERROR_TYPE_UE;
+	case 6:
+		return ACA_ERROR_TYPE_CE;
+	default:
+		return -EINVAL;
+	}
+
+	return -EINVAL;
+}
+
+static void __xgmi_v6_4_0_query_error_count(struct amdgpu_device *adev, struct amdgpu_smuio_mcm_config_info *mcm_info,
+					    u64 mca_base, struct ras_err_data *err_data)
+{
+	int xgmi_inst = mcm_info->die_id;
+	u64 status = 0;
+
+	status = RREG64_MCA(xgmi_inst, mca_base, ACA_REG_IDX_STATUS);
+	if (!ACA_REG__STATUS__VAL(status))
+		return;
+
+	switch (xgmi_v6_4_0_pcs_mca_get_error_type(adev, status)) {
+	case ACA_ERROR_TYPE_UE:
+		amdgpu_ras_error_statistic_ue_count(err_data, mcm_info, 1ULL);
+		break;
+	case ACA_ERROR_TYPE_CE:
+		amdgpu_ras_error_statistic_ce_count(err_data, mcm_info, 1ULL);
+		break;
+	default:
+		break;
+	}
+
+	WREG64_MCA(xgmi_inst, mca_base, ACA_REG_IDX_STATUS, 0ULL);
+}
+
+static void xgmi_v6_4_0_query_error_count(struct amdgpu_device *adev, int xgmi_inst, struct ras_err_data *err_data)
+{
+	struct amdgpu_smuio_mcm_config_info mcm_info = {
+		.socket_id = adev->smuio.funcs->get_socket_id(adev),
+		.die_id = xgmi_inst,
+	};
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(xgmi_v6_4_0_mca_base_array); i++)
+		__xgmi_v6_4_0_query_error_count(adev, &mcm_info, xgmi_v6_4_0_mca_base_array[i], err_data);
+}
+
+static void xgmi_v6_4_0_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status)
+{
+	struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
+	int i;
+
+	for_each_inst(i, adev->aid_mask)
+		xgmi_v6_4_0_query_error_count(adev, i, err_data);
+}
+
+static void amdgpu_xgmi_query_ras_error_count(struct amdgpu_device *adev,
+					      void *ras_error_status)
+{
+	switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) {
+	case IP_VERSION(6, 4, 0):
+	case IP_VERSION(6, 4, 1):
+		xgmi_v6_4_0_query_ras_error_count(adev, ras_error_status);
+		break;
+	default:
+		amdgpu_xgmi_legacy_query_ras_error_count(adev, ras_error_status);
+		break;
+	}
+}
+
 /* Trigger XGMI/WAFL error */
-static int amdgpu_ras_error_inject_xgmi(struct amdgpu_device *adev,  void *inject_if)
+static int amdgpu_ras_error_inject_xgmi(struct amdgpu_device *adev,
+			void *inject_if, uint32_t instance_mask)
 {
-	int ret = 0;
+	int ret1, ret2;
 	struct ta_ras_trigger_error_input *block_info =
 				(struct ta_ras_trigger_error_input *)inject_if;
 
 	if (amdgpu_dpm_set_df_cstate(adev, DF_CSTATE_DISALLOW))
 		dev_warn(adev->dev, "Failed to disallow df cstate");
 
-	if (amdgpu_dpm_allow_xgmi_power_down(adev, false))
+	ret1 = amdgpu_dpm_set_pm_policy(adev, PP_PM_POLICY_XGMI_PLPD, XGMI_PLPD_DISALLOW);
+	if (ret1 && ret1 != -EOPNOTSUPP)
 		dev_warn(adev->dev, "Failed to disallow XGMI power down");
 
-	ret = psp_ras_trigger_error(&adev->psp, block_info);
+	ret2 = psp_ras_trigger_error(&adev->psp, block_info, instance_mask);
 
 	if (amdgpu_ras_intr_triggered())
-		return ret;
+		return ret2;
 
-	if (amdgpu_dpm_allow_xgmi_power_down(adev, true))
+	ret1 = amdgpu_dpm_set_pm_policy(adev, PP_PM_POLICY_XGMI_PLPD, XGMI_PLPD_DEFAULT);
+	if (ret1 && ret1 != -EOPNOTSUPP)
 		dev_warn(adev->dev, "Failed to allow XGMI power down");
 
 	if (amdgpu_dpm_set_df_cstate(adev, DF_CSTATE_ALLOW))
 		dev_warn(adev->dev, "Failed to allow df cstate");
 
-	return ret;
+	return ret2;
 }
 
 struct amdgpu_ras_block_hw_ops  xgmi_ras_hw_ops = {
@@ -1075,3 +1622,157 @@ int amdgpu_xgmi_ras_sw_init(struct amdgpu_device *adev)
 
 	return 0;
 }
+
+static void amdgpu_xgmi_reset_on_init_work(struct work_struct *work)
+{
+	struct amdgpu_hive_info *hive =
+		container_of(work, struct amdgpu_hive_info, reset_on_init_work);
+	struct amdgpu_reset_context reset_context;
+	struct amdgpu_device *tmp_adev;
+	struct list_head device_list;
+	int r;
+
+	mutex_lock(&hive->hive_lock);
+
+	INIT_LIST_HEAD(&device_list);
+	list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head)
+		list_add_tail(&tmp_adev->reset_list, &device_list);
+
+	tmp_adev = list_first_entry(&device_list, struct amdgpu_device,
+				    reset_list);
+	amdgpu_device_lock_reset_domain(tmp_adev->reset_domain);
+
+	reset_context.method = AMD_RESET_METHOD_ON_INIT;
+	reset_context.reset_req_dev = tmp_adev;
+	reset_context.hive = hive;
+	reset_context.reset_device_list = &device_list;
+	set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
+	set_bit(AMDGPU_SKIP_COREDUMP, &reset_context.flags);
+
+	amdgpu_reset_do_xgmi_reset_on_init(&reset_context);
+	mutex_unlock(&hive->hive_lock);
+	amdgpu_device_unlock_reset_domain(tmp_adev->reset_domain);
+
+	list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
+		r = amdgpu_ras_init_badpage_info(tmp_adev);
+		if (r && r != -EHWPOISON)
+			dev_err(tmp_adev->dev,
+				"error during bad page data initialization");
+	}
+}
+
+static void amdgpu_xgmi_schedule_reset_on_init(struct amdgpu_hive_info *hive)
+{
+	INIT_WORK(&hive->reset_on_init_work, amdgpu_xgmi_reset_on_init_work);
+	amdgpu_reset_domain_schedule(hive->reset_domain,
+				     &hive->reset_on_init_work);
+}
+
+int amdgpu_xgmi_reset_on_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_hive_info *hive;
+	bool reset_scheduled;
+	int num_devs;
+
+	hive = amdgpu_get_xgmi_hive(adev);
+	if (!hive)
+		return -EINVAL;
+
+	mutex_lock(&hive->hive_lock);
+	num_devs = atomic_read(&hive->number_devices);
+	reset_scheduled = false;
+	if (num_devs == adev->gmc.xgmi.num_physical_nodes) {
+		amdgpu_xgmi_schedule_reset_on_init(hive);
+		reset_scheduled = true;
+	}
+
+	mutex_unlock(&hive->hive_lock);
+	amdgpu_put_xgmi_hive(hive);
+
+	if (reset_scheduled)
+		flush_work(&hive->reset_on_init_work);
+
+	return 0;
+}
+
+int amdgpu_xgmi_request_nps_change(struct amdgpu_device *adev,
+				   struct amdgpu_hive_info *hive,
+				   int req_nps_mode)
+{
+	struct amdgpu_device *tmp_adev;
+	int cur_nps_mode, r;
+
+	/* This is expected to be called only during unload of driver. The
+	 * request needs to be placed only once for all devices in the hive. If
+	 * one of them fail, revert the request for previous successful devices.
+	 * After placing the request, make hive mode as UNKNOWN so that other
+	 * devices don't request anymore.
+	 */
+	mutex_lock(&hive->hive_lock);
+	if (atomic_read(&hive->requested_nps_mode) ==
+	    UNKNOWN_MEMORY_PARTITION_MODE) {
+		dev_dbg(adev->dev, "Unexpected entry for hive NPS change");
+		mutex_unlock(&hive->hive_lock);
+		return 0;
+	}
+	list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
+		r = adev->gmc.gmc_funcs->request_mem_partition_mode(
+			tmp_adev, req_nps_mode);
+		if (r)
+			break;
+	}
+	if (r) {
+		/* Request back current mode if one of the requests failed */
+		cur_nps_mode =
+			adev->gmc.gmc_funcs->query_mem_partition_mode(tmp_adev);
+		list_for_each_entry_continue_reverse(
+			tmp_adev, &hive->device_list, gmc.xgmi.head)
+			adev->gmc.gmc_funcs->request_mem_partition_mode(
+				tmp_adev, cur_nps_mode);
+	}
+	/* Set to UNKNOWN so that other devices don't request anymore */
+	atomic_set(&hive->requested_nps_mode, UNKNOWN_MEMORY_PARTITION_MODE);
+	mutex_unlock(&hive->hive_lock);
+
+	return r;
+}
+
+bool amdgpu_xgmi_same_hive(struct amdgpu_device *adev,
+			   struct amdgpu_device *bo_adev)
+{
+	return (amdgpu_use_xgmi_p2p && adev != bo_adev &&
+		adev->gmc.xgmi.hive_id &&
+		adev->gmc.xgmi.hive_id == bo_adev->gmc.xgmi.hive_id);
+}
+
+void amdgpu_xgmi_early_init(struct amdgpu_device *adev)
+{
+	if (!adev->gmc.xgmi.supported)
+		return;
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(9, 4, 0):
+	case IP_VERSION(9, 4, 1):
+	case IP_VERSION(9, 4, 2):
+		/* 25 GT/s */
+		adev->gmc.xgmi.max_speed = 25;
+		adev->gmc.xgmi.max_width = 16;
+		break;
+	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+	case IP_VERSION(9, 5, 0):
+		/* 32 GT/s */
+		adev->gmc.xgmi.max_speed = 32;
+		adev->gmc.xgmi.max_width = 16;
+		break;
+	default:
+		break;
+	}
+}
+
+void amgpu_xgmi_set_max_speed_width(struct amdgpu_device *adev,
+				    uint16_t max_speed, uint8_t max_width)
+{
+	adev->gmc.xgmi.max_speed = max_speed;
+	adev->gmc.xgmi.max_width = max_width;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h
index 86fbf56938f4..5f36aff17e79 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h
@@ -23,7 +23,6 @@
 #define __AMDGPU_XGMI_H__
 
 #include <drm/task_barrier.h>
-#include "amdgpu_psp.h"
 #include "amdgpu_ras.h"
 
 struct amdgpu_hive_info {
@@ -43,7 +42,10 @@ struct amdgpu_hive_info {
 	} pstate;
 
 	struct amdgpu_reset_domain *reset_domain;
-	uint32_t device_remove_count;
+	atomic_t ras_recovery;
+	struct ras_event_manager event_mgr;
+	struct work_struct reset_on_init_work;
+	atomic_t requested_nps_mode;
 };
 
 struct amdgpu_pcs_ras_field {
@@ -52,27 +54,80 @@ struct amdgpu_pcs_ras_field {
 	uint32_t pcs_err_shift;
 };
 
-extern struct amdgpu_xgmi_ras  xgmi_ras;
+/**
+ * Bandwidth range reporting comes in two modes.
+ *
+ * PER_LINK - range for any xgmi link
+ * PER_PEER - range of max of single xgmi link to max of multiple links based on source peer
+ */
+enum amdgpu_xgmi_bw_mode {
+	AMDGPU_XGMI_BW_MODE_PER_LINK = 0,
+	AMDGPU_XGMI_BW_MODE_PER_PEER
+};
+
+enum amdgpu_xgmi_bw_unit {
+	AMDGPU_XGMI_BW_UNIT_GBYTES = 0,
+	AMDGPU_XGMI_BW_UNIT_MBYTES
+};
+
+struct amdgpu_xgmi_ras {
+	struct amdgpu_ras_block_object ras_block;
+};
+extern struct amdgpu_xgmi_ras xgmi_ras;
+
+struct amdgpu_xgmi {
+	/* from psp */
+	u64 node_id;
+	u64 hive_id;
+	/* fixed per family */
+	u64 node_segment_size;
+	/* physical node (0-3) */
+	unsigned physical_node_id;
+	/* number of nodes (0-4) */
+	unsigned num_physical_nodes;
+	/* gpu list in the same hive */
+	struct list_head head;
+	bool supported;
+	struct ras_common_if *ras_if;
+	bool connected_to_cpu;
+	struct amdgpu_xgmi_ras *ras;
+	uint16_t max_speed;
+	uint8_t max_width;
+};
+
 struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev);
 void amdgpu_put_xgmi_hive(struct amdgpu_hive_info *hive);
 int amdgpu_xgmi_update_topology(struct amdgpu_hive_info *hive, struct amdgpu_device *adev);
 int amdgpu_xgmi_add_device(struct amdgpu_device *adev);
 int amdgpu_xgmi_remove_device(struct amdgpu_device *adev);
 int amdgpu_xgmi_set_pstate(struct amdgpu_device *adev, int pstate);
-int amdgpu_xgmi_get_hops_count(struct amdgpu_device *adev,
-		struct amdgpu_device *peer_adev);
-int amdgpu_xgmi_get_num_links(struct amdgpu_device *adev,
-		struct amdgpu_device *peer_adev);
+int amdgpu_xgmi_get_hops_count(struct amdgpu_device *adev, struct amdgpu_device *peer_adev);
+int amdgpu_xgmi_get_bandwidth(struct amdgpu_device *adev, struct amdgpu_device *peer_adev,
+			      enum amdgpu_xgmi_bw_mode bw_mode, enum amdgpu_xgmi_bw_unit bw_unit,
+			      uint32_t *min_bw, uint32_t *max_bw);
+bool amdgpu_xgmi_get_is_sharing_enabled(struct amdgpu_device *adev,
+					struct amdgpu_device *peer_adev);
 uint64_t amdgpu_xgmi_get_relative_phy_addr(struct amdgpu_device *adev,
 					   uint64_t addr);
-static inline bool amdgpu_xgmi_same_hive(struct amdgpu_device *adev,
-		struct amdgpu_device *bo_adev)
-{
-	return (amdgpu_use_xgmi_p2p &&
-		adev != bo_adev &&
-		adev->gmc.xgmi.hive_id &&
-		adev->gmc.xgmi.hive_id == bo_adev->gmc.xgmi.hive_id);
-}
+bool amdgpu_xgmi_same_hive(struct amdgpu_device *adev,
+			   struct amdgpu_device *bo_adev);
 int amdgpu_xgmi_ras_sw_init(struct amdgpu_device *adev);
+int amdgpu_xgmi_reset_on_init(struct amdgpu_device *adev);
+
+int amdgpu_xgmi_request_nps_change(struct amdgpu_device *adev,
+				   struct amdgpu_hive_info *hive,
+				   int req_nps_mode);
+int amdgpu_get_xgmi_link_status(struct amdgpu_device *adev,
+				int global_link_num);
+int amdgpu_xgmi_get_ext_link(struct amdgpu_device *adev, int link_num);
+
+void amdgpu_xgmi_early_init(struct amdgpu_device *adev);
+uint32_t amdgpu_xgmi_get_max_bandwidth(struct amdgpu_device *adev);
+
+void amgpu_xgmi_set_max_speed_width(struct amdgpu_device *adev,
+				    uint16_t max_speed, uint8_t max_width);
+
+/* Cleanup macro for use with __free(xgmi_put_hive) */
+DEFINE_FREE(xgmi_put_hive, struct amdgpu_hive_info *, if (_T) amdgpu_put_xgmi_hive(_T))
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h b/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h
index 24d42d24e6a0..3cdb1e0eca37 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h
@@ -23,22 +23,84 @@
 #ifndef AMDGV_SRIOV_MSG__H_
 #define AMDGV_SRIOV_MSG__H_
 
-/* unit in kilobytes */
-#define AMD_SRIOV_MSG_VBIOS_OFFSET	     0
-#define AMD_SRIOV_MSG_VBIOS_SIZE_KB	     64
-#define AMD_SRIOV_MSG_DATAEXCHANGE_OFFSET_KB AMD_SRIOV_MSG_VBIOS_SIZE_KB
-#define AMD_SRIOV_MSG_DATAEXCHANGE_SIZE_KB   4
+#define AMD_SRIOV_MSG_SIZE_KB                           1
 
 /*
- * layout
- * 0           64KB        65KB        66KB
- * |   VBIOS   |   PF2VF   |   VF2PF   |   Bad Page   | ...
- * |   64KB    |   1KB     |   1KB     |
+ * layout v1
+ * 0           64KB        65KB        66KB           68KB                   132KB
+ * |   VBIOS   |   PF2VF   |   VF2PF   |   Bad Page   | RAS Telemetry Region | ...
+ * |   64KB    |   1KB     |   1KB     |   2KB        | 64KB                 | ...
  */
-#define AMD_SRIOV_MSG_SIZE_KB                   1
-#define AMD_SRIOV_MSG_PF2VF_OFFSET_KB           AMD_SRIOV_MSG_DATAEXCHANGE_OFFSET_KB
-#define AMD_SRIOV_MSG_VF2PF_OFFSET_KB           (AMD_SRIOV_MSG_PF2VF_OFFSET_KB + AMD_SRIOV_MSG_SIZE_KB)
-#define AMD_SRIOV_MSG_BAD_PAGE_OFFSET_KB        (AMD_SRIOV_MSG_VF2PF_OFFSET_KB + AMD_SRIOV_MSG_SIZE_KB)
+
+/*
+ * layout v2 (offsets are dynamically allocated and the offsets below are examples)
+ * 0           1KB         64KB        65KB        66KB           68KB                   132KB
+ * |  INITD_H  |   VBIOS   |   PF2VF   |   VF2PF   |   Bad Page   | RAS Telemetry Region | ...
+ * |   1KB     |   64KB    |   1KB     |   1KB     |   2KB        | 64KB                 | ...
+ *
+ * Note: PF2VF + VF2PF + Bad Page = DataExchange region (allocated contiguously)
+ */
+
+/* v1 layout sizes */
+#define AMD_SRIOV_MSG_VBIOS_SIZE_KB_V1			64
+#define AMD_SRIOV_MSG_PF2VF_SIZE_KB_V1			1
+#define AMD_SRIOV_MSG_VF2PF_SIZE_KB_V1			1
+#define AMD_SRIOV_MSG_BAD_PAGE_SIZE_KB_V1		2
+#define AMD_SRIOV_MSG_RAS_TELEMETRY_SIZE_KB_V1		64
+#define AMD_SRIOV_MSG_DATAEXCHANGE_SIZE_KB_V1		\
+	(AMD_SRIOV_MSG_PF2VF_SIZE_KB_V1 + AMD_SRIOV_MSG_VF2PF_SIZE_KB_V1 + \
+	 AMD_SRIOV_MSG_BAD_PAGE_SIZE_KB_V1)
+
+/* v1 offsets */
+#define AMD_SRIOV_MSG_VBIOS_OFFSET_V1			0
+#define AMD_SRIOV_MSG_DATAEXCHANGE_OFFSET_KB_V1		AMD_SRIOV_MSG_VBIOS_SIZE_KB_V1
+#define AMD_SRIOV_MSG_TMR_OFFSET_KB			2048
+#define AMD_SRIOV_MSG_PF2VF_OFFSET_KB_V1		AMD_SRIOV_MSG_DATAEXCHANGE_OFFSET_KB_V1
+#define AMD_SRIOV_MSG_VF2PF_OFFSET_KB_V1		\
+	(AMD_SRIOV_MSG_PF2VF_OFFSET_KB_V1 + AMD_SRIOV_MSG_SIZE_KB)
+#define AMD_SRIOV_MSG_BAD_PAGE_OFFSET_KB_V1		\
+	(AMD_SRIOV_MSG_VF2PF_OFFSET_KB_V1 + AMD_SRIOV_MSG_SIZE_KB)
+#define AMD_SRIOV_MSG_RAS_TELEMETRY_OFFSET_KB_V1	\
+	(AMD_SRIOV_MSG_BAD_PAGE_OFFSET_KB_V1 + AMD_SRIOV_MSG_BAD_PAGE_SIZE_KB_V1)
+#define AMD_SRIOV_MSG_INIT_DATA_TOT_SIZE_KB_V1		\
+	(AMD_SRIOV_MSG_VBIOS_SIZE_KB_V1 + AMD_SRIOV_MSG_DATAEXCHANGE_SIZE_KB_V1 + \
+	 AMD_SRIOV_MSG_RAS_TELEMETRY_SIZE_KB_V1)
+
+enum amd_sriov_crit_region_version {
+	GPU_CRIT_REGION_V1 = 1,
+	GPU_CRIT_REGION_V2 = 2,
+};
+
+/* v2 layout offset enum (in order of allocation) */
+enum amd_sriov_msg_table_id_enum {
+	AMD_SRIOV_MSG_IPD_TABLE_ID = 0,
+	AMD_SRIOV_MSG_VBIOS_IMG_TABLE_ID,
+	AMD_SRIOV_MSG_RAS_TELEMETRY_TABLE_ID,
+	AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID,
+	AMD_SRIOV_MSG_BAD_PAGE_INFO_TABLE_ID,
+	AMD_SRIOV_MSG_INITD_H_TABLE_ID,
+	AMD_SRIOV_MSG_MAX_TABLE_ID,
+};
+
+struct amd_sriov_msg_init_data_header {
+	char     signature[4];  /* "INDA"  */
+	uint32_t version;
+	uint32_t checksum;
+	uint32_t initdata_offset; /* 0 */
+	uint32_t initdata_size_in_kb; /* 5MB */
+	uint32_t valid_tables;
+	uint32_t vbios_img_offset;
+	uint32_t vbios_img_size_in_kb;
+	uint32_t dataexchange_offset;
+	uint32_t dataexchange_size_in_kb;
+	uint32_t ras_tele_info_offset;
+	uint32_t ras_tele_info_size_in_kb;
+	uint32_t ip_discovery_offset;
+	uint32_t ip_discovery_size_in_kb;
+	uint32_t bad_page_info_offset;
+	uint32_t bad_page_size_in_kb;
+	uint32_t reserved[8];
+};
 
 /*
  * PF2VF history log:
@@ -70,7 +132,6 @@ enum amd_sriov_ucode_engine_id {
 	AMD_SRIOV_UCODE_ID_RLC_SRLS,
 	AMD_SRIOV_UCODE_ID_MEC,
 	AMD_SRIOV_UCODE_ID_MEC2,
-	AMD_SRIOV_UCODE_ID_IMU,
 	AMD_SRIOV_UCODE_ID_SOS,
 	AMD_SRIOV_UCODE_ID_ASD,
 	AMD_SRIOV_UCODE_ID_TA_RAS,
@@ -87,28 +148,64 @@ enum amd_sriov_ucode_engine_id {
 
 union amd_sriov_msg_feature_flags {
 	struct {
-		uint32_t error_log_collect : 1;
-		uint32_t host_load_ucodes  : 1;
-		uint32_t host_flr_vramlost : 1;
-		uint32_t mm_bw_management  : 1;
-		uint32_t pp_one_vf_mode	   : 1;
-		uint32_t reg_indirect_acc  : 1;
-		uint32_t av1_support       : 1;
-		uint32_t reserved	   : 25;
+		uint32_t error_log_collect	: 1;
+		uint32_t host_load_ucodes	: 1;
+		uint32_t host_flr_vramlost	: 1;
+		uint32_t mm_bw_management	: 1;
+		uint32_t pp_one_vf_mode		: 1;
+		uint32_t reg_indirect_acc	: 1;
+		uint32_t av1_support		: 1;
+		uint32_t vcn_rb_decouple	: 1;
+		uint32_t mes_info_dump_enable	: 1;
+		uint32_t ras_caps		: 1;
+		uint32_t ras_telemetry		: 1;
+		uint32_t ras_cper		: 1;
+		uint32_t xgmi_ta_ext_peer_link	: 1;
+		uint32_t reserved		: 19;
 	} flags;
 	uint32_t all;
 };
 
 union amd_sriov_reg_access_flags {
 	struct {
-		uint32_t vf_reg_access_ih 	 : 1;
-		uint32_t vf_reg_access_mmhub : 1;
-		uint32_t vf_reg_access_gc 	 : 1;
-		uint32_t reserved	         : 29;
+		uint32_t vf_reg_access_ih		: 1;
+		uint32_t vf_reg_access_mmhub		: 1;
+		uint32_t vf_reg_access_gc		: 1;
+		uint32_t vf_reg_access_l1_tlb_cntl	: 1;
+		uint32_t vf_reg_access_sq_config	: 1;
+		uint32_t reserved			: 27;
 	} flags;
 	uint32_t all;
 };
 
+union amd_sriov_ras_caps {
+	struct {
+		uint64_t block_umc			: 1;
+		uint64_t block_sdma			: 1;
+		uint64_t block_gfx			: 1;
+		uint64_t block_mmhub			: 1;
+		uint64_t block_athub			: 1;
+		uint64_t block_pcie_bif			: 1;
+		uint64_t block_hdp			: 1;
+		uint64_t block_xgmi_wafl		: 1;
+		uint64_t block_df			: 1;
+		uint64_t block_smn			: 1;
+		uint64_t block_sem			: 1;
+		uint64_t block_mp0			: 1;
+		uint64_t block_mp1			: 1;
+		uint64_t block_fuse			: 1;
+		uint64_t block_mca			: 1;
+		uint64_t block_vcn			: 1;
+		uint64_t block_jpeg			: 1;
+		uint64_t block_ih			: 1;
+		uint64_t block_mpio			: 1;
+		uint64_t block_mmsch			: 1;
+		uint64_t poison_propogation_mode	: 1;
+		uint64_t reserved			: 43;
+	} bits;
+	uint64_t all;
+};
+
 union amd_sriov_msg_os_info {
 	struct {
 		uint32_t windows  : 1;
@@ -157,7 +254,7 @@ struct amd_sriov_msg_pf2vf_info_header {
 	uint32_t reserved[2];
 };
 
-#define AMD_SRIOV_MSG_PF2VF_INFO_FILLED_SIZE (48)
+#define AMD_SRIOV_MSG_PF2VF_INFO_FILLED_SIZE (55)
 struct amd_sriov_msg_pf2vf_info {
 	/* header contains size and version */
 	struct amd_sriov_msg_pf2vf_info_header header;
@@ -208,9 +305,17 @@ struct amd_sriov_msg_pf2vf_info {
 	struct amd_sriov_msg_uuid_info uuid_info;
 	/* PCIE atomic ops support flag */
 	uint32_t pcie_atomic_ops_support_flags;
+	/* Portion of GPU memory occupied by VF.  MAX value is 65535, but set to uint32_t to maintain alignment with reserved size */
+	uint32_t gpu_capacity;
+	/* vf bdf on host pci tree for debug only */
+	uint32_t bdf_on_host;
+	uint32_t more_bp;	//Reserved for future use.
+	union amd_sriov_ras_caps ras_en_caps;
+	union amd_sriov_ras_caps ras_telemetry_en_caps;
+
 	/* reserved */
 	uint32_t reserved[256 - AMD_SRIOV_MSG_PF2VF_INFO_FILLED_SIZE];
-};
+} __packed;
 
 struct amd_sriov_msg_vf2pf_info_header {
 	/* the total structure size in byte */
@@ -221,7 +326,7 @@ struct amd_sriov_msg_vf2pf_info_header {
 	uint32_t reserved[2];
 };
 
-#define AMD_SRIOV_MSG_VF2PF_INFO_FILLED_SIZE (70)
+#define AMD_SRIOV_MSG_VF2PF_INFO_FILLED_SIZE (73)
 struct amd_sriov_msg_vf2pf_info {
 	/* header contains size and version */
 	struct amd_sriov_msg_vf2pf_info_header header;
@@ -265,10 +370,12 @@ struct amd_sriov_msg_vf2pf_info {
 		uint32_t version;
 	} ucode_info[AMD_SRIOV_MSG_RESERVE_UCODE];
 	uint64_t dummy_page_addr;
-
+	/* FB allocated for guest MES to record UQ info */
+	uint64_t mes_info_addr;
+	uint32_t mes_info_size;
 	/* reserved */
 	uint32_t reserved[256 - AMD_SRIOV_MSG_VF2PF_INFO_FILLED_SIZE];
-};
+} __packed;
 
 /* mailbox message send from guest to host  */
 enum amd_sriov_mailbox_request_message {
@@ -278,22 +385,98 @@ enum amd_sriov_mailbox_request_message {
 	MB_REQ_MSG_REL_GPU_FINI_ACCESS,
 	MB_REQ_MSG_REQ_GPU_RESET_ACCESS,
 	MB_REQ_MSG_REQ_GPU_INIT_DATA,
+	MB_REQ_MSG_PSP_VF_CMD_RELAY,
 
 	MB_REQ_MSG_LOG_VF_ERROR = 200,
+	MB_REQ_MSG_READY_TO_RESET = 201,
+	MB_REQ_MSG_RAS_POISON = 202,
+	MB_REQ_RAS_ERROR_COUNT = 203,
+	MB_REQ_RAS_CPER_DUMP = 204,
+	MB_REQ_RAS_BAD_PAGES = 205,
 };
 
 /* mailbox message send from host to guest  */
 enum amd_sriov_mailbox_response_message {
-	MB_RES_MSG_CLR_MSG_BUF = 0,
-	MB_RES_MSG_READY_TO_ACCESS_GPU = 1,
-	MB_RES_MSG_FLR_NOTIFICATION,
-	MB_RES_MSG_FLR_NOTIFICATION_COMPLETION,
-	MB_RES_MSG_SUCCESS,
-	MB_RES_MSG_FAIL,
-	MB_RES_MSG_QUERY_ALIVE,
-	MB_RES_MSG_GPU_INIT_DATA_READY,
-
-	MB_RES_MSG_TEXT_MESSAGE = 255
+	MB_RES_MSG_CLR_MSG_BUF			= 0,
+	MB_RES_MSG_READY_TO_ACCESS_GPU		= 1,
+	MB_RES_MSG_FLR_NOTIFICATION		= 2,
+	MB_RES_MSG_FLR_NOTIFICATION_COMPLETION  = 3,
+	MB_RES_MSG_SUCCESS			= 4,
+	MB_RES_MSG_FAIL				= 5,
+	MB_RES_MSG_QUERY_ALIVE			= 6,
+	MB_RES_MSG_GPU_INIT_DATA_READY		= 7,
+	MB_RES_MSG_RAS_POISON_READY		= 8,
+	MB_RES_MSG_PF_SOFT_FLR_NOTIFICATION	= 9,
+	MB_RES_MSG_GPU_RMA			= 10,
+	MB_RES_MSG_RAS_ERROR_COUNT_READY	= 11,
+	MB_REQ_RAS_CPER_DUMP_READY		= 14,
+	MB_RES_MSG_RAS_BAD_PAGES_READY		= 15,
+	MB_RES_MSG_RAS_BAD_PAGES_NOTIFICATION	= 16,
+	MB_RES_MSG_UNRECOV_ERR_NOTIFICATION	= 17,
+	MB_RES_MSG_TEXT_MESSAGE			= 255
+};
+
+enum amd_sriov_ras_telemetry_gpu_block {
+	RAS_TELEMETRY_GPU_BLOCK_UMC		= 0,
+	RAS_TELEMETRY_GPU_BLOCK_SDMA		= 1,
+	RAS_TELEMETRY_GPU_BLOCK_GFX		= 2,
+	RAS_TELEMETRY_GPU_BLOCK_MMHUB		= 3,
+	RAS_TELEMETRY_GPU_BLOCK_ATHUB		= 4,
+	RAS_TELEMETRY_GPU_BLOCK_PCIE_BIF	= 5,
+	RAS_TELEMETRY_GPU_BLOCK_HDP		= 6,
+	RAS_TELEMETRY_GPU_BLOCK_XGMI_WAFL	= 7,
+	RAS_TELEMETRY_GPU_BLOCK_DF		= 8,
+	RAS_TELEMETRY_GPU_BLOCK_SMN		= 9,
+	RAS_TELEMETRY_GPU_BLOCK_SEM		= 10,
+	RAS_TELEMETRY_GPU_BLOCK_MP0		= 11,
+	RAS_TELEMETRY_GPU_BLOCK_MP1		= 12,
+	RAS_TELEMETRY_GPU_BLOCK_FUSE		= 13,
+	RAS_TELEMETRY_GPU_BLOCK_MCA		= 14,
+	RAS_TELEMETRY_GPU_BLOCK_VCN		= 15,
+	RAS_TELEMETRY_GPU_BLOCK_JPEG		= 16,
+	RAS_TELEMETRY_GPU_BLOCK_IH		= 17,
+	RAS_TELEMETRY_GPU_BLOCK_MPIO		= 18,
+	RAS_TELEMETRY_GPU_BLOCK_COUNT		= 19,
+};
+
+struct amd_sriov_ras_telemetry_header {
+	uint32_t checksum;
+	uint32_t used_size;
+	uint32_t reserved[2];
+};
+
+struct amd_sriov_ras_telemetry_error_count {
+	struct {
+		uint32_t ce_count;
+		uint32_t ue_count;
+		uint32_t de_count;
+		uint32_t ce_overflow_count;
+		uint32_t ue_overflow_count;
+		uint32_t de_overflow_count;
+		uint32_t reserved[6];
+	} block[RAS_TELEMETRY_GPU_BLOCK_COUNT];
+};
+
+struct amd_sriov_ras_cper_dump {
+	uint32_t more;
+	uint64_t overflow_count;
+	uint64_t count;
+	uint64_t wptr;
+	uint32_t buf[];
+};
+
+struct amd_sriov_ras_chk_criti {
+	uint32_t hit;
+};
+
+struct amdsriov_ras_telemetry {
+	struct amd_sriov_ras_telemetry_header header;
+
+	union {
+		struct amd_sriov_ras_telemetry_error_count error_count;
+		struct amd_sriov_ras_cper_dump cper_dump;
+		struct amd_sriov_ras_chk_criti chk_criti;
+	} body;
 };
 
 /* version data stored in MAILBOX_MSGBUF_RCV_DW1 for future expansion */
diff --git a/drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c b/drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c
new file mode 100644
index 000000000000..f9e2edf5260b
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c
@@ -0,0 +1,986 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "amdgpu.h"
+#include "soc15.h"
+
+#include "soc15_common.h"
+#include "amdgpu_reg_state.h"
+#include "amdgpu_xcp.h"
+#include "gfx_v9_4_3.h"
+#include "gfxhub_v1_2.h"
+#include "sdma_v4_4_2.h"
+#include "amdgpu_ip.h"
+
+#define XCP_INST_MASK(num_inst, xcp_id)                                        \
+	(num_inst ? GENMASK(num_inst - 1, 0) << (xcp_id * num_inst) : 0)
+
+void aqua_vanjaram_doorbell_index_init(struct amdgpu_device *adev)
+{
+	int i;
+
+	adev->doorbell_index.kiq = AMDGPU_DOORBELL_LAYOUT1_KIQ_START;
+
+	adev->doorbell_index.mec_ring0 = AMDGPU_DOORBELL_LAYOUT1_MEC_RING_START;
+
+	adev->doorbell_index.userqueue_start = AMDGPU_DOORBELL_LAYOUT1_USERQUEUE_START;
+	adev->doorbell_index.userqueue_end = AMDGPU_DOORBELL_LAYOUT1_USERQUEUE_END;
+	adev->doorbell_index.xcc_doorbell_range = AMDGPU_DOORBELL_LAYOUT1_XCC_RANGE;
+
+	adev->doorbell_index.sdma_doorbell_range = 20;
+	for (i = 0; i < adev->sdma.num_instances; i++)
+		adev->doorbell_index.sdma_engine[i] =
+			AMDGPU_DOORBELL_LAYOUT1_sDMA_ENGINE_START +
+			i * (adev->doorbell_index.sdma_doorbell_range >> 1);
+
+	adev->doorbell_index.ih = AMDGPU_DOORBELL_LAYOUT1_IH;
+	adev->doorbell_index.vcn.vcn_ring0_1 = AMDGPU_DOORBELL_LAYOUT1_VCN_START;
+
+	adev->doorbell_index.first_non_cp = AMDGPU_DOORBELL_LAYOUT1_FIRST_NON_CP;
+	adev->doorbell_index.last_non_cp = AMDGPU_DOORBELL_LAYOUT1_LAST_NON_CP;
+
+	adev->doorbell_index.max_assignment = AMDGPU_DOORBELL_LAYOUT1_MAX_ASSIGNMENT << 1;
+}
+
+/* Fixed pattern for smn addressing on different AIDs:
+ *   bit[34]: indicate cross AID access
+ *   bit[33:32]: indicate target AID id
+ * AID id range is 0 ~ 3 as maximum AID number is 4.
+ */
+u64 aqua_vanjaram_encode_ext_smn_addressing(int ext_id)
+{
+	u64 ext_offset;
+
+	/* local routing and bit[34:32] will be zeros */
+	if (ext_id == 0)
+		return 0;
+
+	/* Initiated from host, accessing to all non-zero aids are cross traffic */
+	ext_offset = ((u64)(ext_id & 0x3) << 32) | (1ULL << 34);
+
+	return ext_offset;
+}
+
+static enum amdgpu_gfx_partition
+__aqua_vanjaram_calc_xcp_mode(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	int num_xcc, num_xcc_per_xcp = 0, mode = 0;
+
+	num_xcc = NUM_XCC(xcp_mgr->adev->gfx.xcc_mask);
+	if (adev->gfx.funcs->get_xccs_per_xcp)
+		num_xcc_per_xcp = adev->gfx.funcs->get_xccs_per_xcp(adev);
+	if ((num_xcc_per_xcp) && (num_xcc % num_xcc_per_xcp == 0))
+		mode = num_xcc / num_xcc_per_xcp;
+
+	if (num_xcc_per_xcp == 1)
+		return AMDGPU_CPX_PARTITION_MODE;
+
+	switch (mode) {
+	case 1:
+		return AMDGPU_SPX_PARTITION_MODE;
+	case 2:
+		return AMDGPU_DPX_PARTITION_MODE;
+	case 3:
+		return AMDGPU_TPX_PARTITION_MODE;
+	case 4:
+		return AMDGPU_QPX_PARTITION_MODE;
+	default:
+		return AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE;
+	}
+
+	return AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE;
+}
+
+static int aqua_vanjaram_query_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	enum amdgpu_gfx_partition derv_mode,
+		mode = AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE;
+	struct amdgpu_device *adev = xcp_mgr->adev;
+
+	derv_mode = __aqua_vanjaram_calc_xcp_mode(xcp_mgr);
+
+	if (amdgpu_sriov_vf(adev))
+		return derv_mode;
+
+	if (adev->nbio.funcs->get_compute_partition_mode) {
+		mode = adev->nbio.funcs->get_compute_partition_mode(adev);
+		if (mode != derv_mode) {
+			dev_warn(
+				adev->dev,
+				"Mismatch in compute partition mode - reported : %d derived : %d",
+				mode, derv_mode);
+			if (derv_mode == AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE)
+				amdgpu_device_bus_status_check(adev);
+		}
+	}
+
+	return mode;
+}
+
+static int __aqua_vanjaram_get_xcc_per_xcp(struct amdgpu_xcp_mgr *xcp_mgr, int mode)
+{
+	int num_xcc, num_xcc_per_xcp = 0;
+
+	num_xcc = NUM_XCC(xcp_mgr->adev->gfx.xcc_mask);
+
+	switch (mode) {
+	case AMDGPU_SPX_PARTITION_MODE:
+		num_xcc_per_xcp = num_xcc;
+		break;
+	case AMDGPU_DPX_PARTITION_MODE:
+		num_xcc_per_xcp = num_xcc / 2;
+		break;
+	case AMDGPU_TPX_PARTITION_MODE:
+		num_xcc_per_xcp = num_xcc / 3;
+		break;
+	case AMDGPU_QPX_PARTITION_MODE:
+		num_xcc_per_xcp = num_xcc / 4;
+		break;
+	case AMDGPU_CPX_PARTITION_MODE:
+		num_xcc_per_xcp = 1;
+		break;
+	}
+
+	return num_xcc_per_xcp;
+}
+
+static int __aqua_vanjaram_get_xcp_ip_info(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id,
+				    enum AMDGPU_XCP_IP_BLOCK ip_id,
+				    struct amdgpu_xcp_ip *ip)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	int num_sdma, num_vcn, num_shared_vcn, num_xcp;
+	int num_xcc_xcp, num_sdma_xcp, num_vcn_xcp;
+
+	num_sdma = adev->sdma.num_instances;
+	num_vcn = adev->vcn.num_vcn_inst;
+	num_shared_vcn = 1;
+
+	num_xcc_xcp = adev->gfx.num_xcc_per_xcp;
+	num_xcp = NUM_XCC(adev->gfx.xcc_mask) / num_xcc_xcp;
+
+	switch (xcp_mgr->mode) {
+	case AMDGPU_SPX_PARTITION_MODE:
+	case AMDGPU_DPX_PARTITION_MODE:
+	case AMDGPU_TPX_PARTITION_MODE:
+	case AMDGPU_QPX_PARTITION_MODE:
+	case AMDGPU_CPX_PARTITION_MODE:
+		num_sdma_xcp = DIV_ROUND_UP(num_sdma, num_xcp);
+		num_vcn_xcp = DIV_ROUND_UP(num_vcn, num_xcp);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if (num_vcn && num_xcp > num_vcn)
+		num_shared_vcn = num_xcp / num_vcn;
+
+	switch (ip_id) {
+	case AMDGPU_XCP_GFXHUB:
+		ip->inst_mask = XCP_INST_MASK(num_xcc_xcp, xcp_id);
+		ip->ip_funcs = &gfxhub_v1_2_xcp_funcs;
+		break;
+	case AMDGPU_XCP_GFX:
+		ip->inst_mask = XCP_INST_MASK(num_xcc_xcp, xcp_id);
+		ip->ip_funcs = &gfx_v9_4_3_xcp_funcs;
+		break;
+	case AMDGPU_XCP_SDMA:
+		ip->inst_mask = XCP_INST_MASK(num_sdma_xcp, xcp_id);
+		ip->ip_funcs = &sdma_v4_4_2_xcp_funcs;
+		break;
+	case AMDGPU_XCP_VCN:
+		ip->inst_mask =
+			XCP_INST_MASK(num_vcn_xcp, xcp_id / num_shared_vcn);
+		/* TODO : Assign IP funcs */
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	ip->ip_id = ip_id;
+
+	return 0;
+}
+
+static int __aqua_vanjaram_get_px_mode_info(struct amdgpu_xcp_mgr *xcp_mgr,
+					    int px_mode, int *num_xcp,
+					    uint16_t *nps_modes)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	uint32_t gc_ver = amdgpu_ip_version(adev, GC_HWIP, 0);
+
+	if (!num_xcp || !nps_modes || !(xcp_mgr->supp_xcp_modes & BIT(px_mode)))
+		return -EINVAL;
+
+	switch (px_mode) {
+	case AMDGPU_SPX_PARTITION_MODE:
+		*num_xcp = 1;
+		*nps_modes = BIT(AMDGPU_NPS1_PARTITION_MODE);
+		break;
+	case AMDGPU_DPX_PARTITION_MODE:
+		*num_xcp = 2;
+		*nps_modes = BIT(AMDGPU_NPS1_PARTITION_MODE) |
+			     BIT(AMDGPU_NPS2_PARTITION_MODE);
+		break;
+	case AMDGPU_TPX_PARTITION_MODE:
+		*num_xcp = 3;
+		*nps_modes = BIT(AMDGPU_NPS1_PARTITION_MODE) |
+			     BIT(AMDGPU_NPS4_PARTITION_MODE);
+		break;
+	case AMDGPU_QPX_PARTITION_MODE:
+		*num_xcp = 4;
+		*nps_modes = BIT(AMDGPU_NPS1_PARTITION_MODE) |
+			     BIT(AMDGPU_NPS4_PARTITION_MODE);
+		if (gc_ver == IP_VERSION(9, 5, 0))
+			*nps_modes |= BIT(AMDGPU_NPS2_PARTITION_MODE);
+		break;
+	case AMDGPU_CPX_PARTITION_MODE:
+		*num_xcp = NUM_XCC(adev->gfx.xcc_mask);
+		*nps_modes = BIT(AMDGPU_NPS1_PARTITION_MODE) |
+			     BIT(AMDGPU_NPS4_PARTITION_MODE);
+		if (gc_ver == IP_VERSION(9, 5, 0))
+			*nps_modes |= BIT(AMDGPU_NPS2_PARTITION_MODE);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int aqua_vanjaram_get_xcp_res_info(struct amdgpu_xcp_mgr *xcp_mgr,
+					  int mode,
+					  struct amdgpu_xcp_cfg *xcp_cfg)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	int max_res[AMDGPU_XCP_RES_MAX] = {};
+	bool res_lt_xcp;
+	int num_xcp, i, r;
+	u16 nps_modes;
+
+	if (!(xcp_mgr->supp_xcp_modes & BIT(mode)))
+		return -EINVAL;
+
+	max_res[AMDGPU_XCP_RES_XCC] = NUM_XCC(adev->gfx.xcc_mask);
+	max_res[AMDGPU_XCP_RES_DMA] = adev->sdma.num_instances;
+	max_res[AMDGPU_XCP_RES_DEC] = adev->vcn.num_vcn_inst;
+	max_res[AMDGPU_XCP_RES_JPEG] = adev->jpeg.num_jpeg_inst;
+
+	r = __aqua_vanjaram_get_px_mode_info(xcp_mgr, mode, &num_xcp, &nps_modes);
+	if (r)
+		return r;
+
+	xcp_cfg->compatible_nps_modes =
+		(adev->gmc.supported_nps_modes & nps_modes);
+	xcp_cfg->num_res = ARRAY_SIZE(max_res);
+
+	for (i = 0; i < xcp_cfg->num_res; i++) {
+		res_lt_xcp = max_res[i] < num_xcp;
+		xcp_cfg->xcp_res[i].id = i;
+		xcp_cfg->xcp_res[i].num_inst =
+			res_lt_xcp ? 1 : max_res[i] / num_xcp;
+		xcp_cfg->xcp_res[i].num_inst =
+			i == AMDGPU_XCP_RES_JPEG ?
+			xcp_cfg->xcp_res[i].num_inst *
+			adev->jpeg.num_jpeg_rings : xcp_cfg->xcp_res[i].num_inst;
+		xcp_cfg->xcp_res[i].num_shared =
+			res_lt_xcp ? num_xcp / max_res[i] : 1;
+	}
+
+	return 0;
+}
+
+static enum amdgpu_gfx_partition
+__aqua_vanjaram_get_auto_mode(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	int num_xcc;
+
+	num_xcc = NUM_XCC(xcp_mgr->adev->gfx.xcc_mask);
+
+	if (adev->gmc.num_mem_partitions == 1)
+		return AMDGPU_SPX_PARTITION_MODE;
+
+	if (adev->gmc.num_mem_partitions == num_xcc)
+		return AMDGPU_CPX_PARTITION_MODE;
+
+	if (adev->gmc.num_mem_partitions == num_xcc / 2)
+		return (adev->flags & AMD_IS_APU) ? AMDGPU_TPX_PARTITION_MODE :
+						    AMDGPU_CPX_PARTITION_MODE;
+
+	if (adev->gmc.num_mem_partitions == 2 && !(adev->flags & AMD_IS_APU))
+		return AMDGPU_DPX_PARTITION_MODE;
+
+	return AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE;
+}
+
+static bool __aqua_vanjaram_is_valid_mode(struct amdgpu_xcp_mgr *xcp_mgr,
+					  enum amdgpu_gfx_partition mode)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	int num_xcc, num_xccs_per_xcp, r;
+	int num_xcp, nps_mode;
+	u16 supp_nps_modes;
+	bool comp_mode;
+
+	nps_mode = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
+	r = __aqua_vanjaram_get_px_mode_info(xcp_mgr, mode, &num_xcp,
+					       &supp_nps_modes);
+	if (r)
+		return false;
+
+	comp_mode = !!(BIT(nps_mode) & supp_nps_modes);
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	switch (mode) {
+	case AMDGPU_SPX_PARTITION_MODE:
+		return comp_mode && num_xcc > 0;
+	case AMDGPU_DPX_PARTITION_MODE:
+		return comp_mode && (num_xcc % 4) == 0;
+	case AMDGPU_TPX_PARTITION_MODE:
+		return comp_mode && ((num_xcc % 3) == 0);
+	case AMDGPU_QPX_PARTITION_MODE:
+		num_xccs_per_xcp = num_xcc / 4;
+		return comp_mode && (num_xccs_per_xcp >= 2);
+	case AMDGPU_CPX_PARTITION_MODE:
+		return comp_mode && (num_xcc > 1);
+	default:
+		return false;
+	}
+
+	return false;
+}
+
+static void __aqua_vanjaram_update_available_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	int mode;
+
+	xcp_mgr->avail_xcp_modes = 0;
+
+	for_each_inst(mode, xcp_mgr->supp_xcp_modes) {
+		if (__aqua_vanjaram_is_valid_mode(xcp_mgr, mode))
+			xcp_mgr->avail_xcp_modes |= BIT(mode);
+	}
+}
+
+static int aqua_vanjaram_switch_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr,
+					       int mode, int *num_xcps)
+{
+	int num_xcc_per_xcp, num_xcc, ret;
+	struct amdgpu_device *adev;
+	u32 flags = 0;
+
+	adev = xcp_mgr->adev;
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+
+	if (mode == AMDGPU_AUTO_COMPUTE_PARTITION_MODE) {
+		mode = __aqua_vanjaram_get_auto_mode(xcp_mgr);
+		if (mode == AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE) {
+			dev_err(adev->dev,
+				"Invalid config, no compatible compute partition mode found, available memory partitions: %d",
+				adev->gmc.num_mem_partitions);
+			return -EINVAL;
+		}
+	} else if (!__aqua_vanjaram_is_valid_mode(xcp_mgr, mode)) {
+		dev_err(adev->dev,
+			"Invalid compute partition mode requested, requested: %s, available memory partitions: %d",
+			amdgpu_gfx_compute_mode_desc(mode), adev->gmc.num_mem_partitions);
+		return -EINVAL;
+	}
+
+	if (adev->kfd.init_complete && !amdgpu_in_reset(adev) &&
+		!adev->in_suspend)
+		flags |= AMDGPU_XCP_OPS_KFD;
+
+	if (flags & AMDGPU_XCP_OPS_KFD) {
+		ret = amdgpu_amdkfd_check_and_lock_kfd(adev);
+		if (ret)
+			goto out;
+	}
+
+	ret = amdgpu_xcp_pre_partition_switch(xcp_mgr, flags);
+	if (ret)
+		goto unlock;
+
+	num_xcc_per_xcp = __aqua_vanjaram_get_xcc_per_xcp(xcp_mgr, mode);
+	if (adev->gfx.funcs->switch_partition_mode)
+		adev->gfx.funcs->switch_partition_mode(xcp_mgr->adev,
+						       num_xcc_per_xcp);
+
+	/* Init info about new xcps */
+	*num_xcps = num_xcc / num_xcc_per_xcp;
+	amdgpu_xcp_init(xcp_mgr, *num_xcps, mode);
+
+	ret = amdgpu_xcp_post_partition_switch(xcp_mgr, flags);
+	if (!ret)
+		__aqua_vanjaram_update_available_partition_mode(xcp_mgr);
+unlock:
+	if (flags & AMDGPU_XCP_OPS_KFD)
+		amdgpu_amdkfd_unlock_kfd(adev);
+out:
+	return ret;
+}
+
+static int __aqua_vanjaram_get_xcp_mem_id(struct amdgpu_device *adev,
+					  int xcc_id, uint8_t *mem_id)
+{
+	/* memory/spatial modes validation check is already done */
+	*mem_id = xcc_id / adev->gfx.num_xcc_per_xcp;
+	*mem_id /= adev->xcp_mgr->num_xcp_per_mem_partition;
+
+	return 0;
+}
+
+static int aqua_vanjaram_get_xcp_mem_id(struct amdgpu_xcp_mgr *xcp_mgr,
+					struct amdgpu_xcp *xcp, uint8_t *mem_id)
+{
+	struct amdgpu_numa_info numa_info;
+	struct amdgpu_device *adev;
+	uint32_t xcc_mask;
+	int r, i, xcc_id;
+
+	adev = xcp_mgr->adev;
+	/* TODO: BIOS is not returning the right info now
+	 * Check on this later
+	 */
+	/*
+	if (adev->gmc.gmc_funcs->query_mem_partition_mode)
+		mode = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
+	*/
+	if (adev->gmc.num_mem_partitions == 1) {
+		/* Only one range */
+		*mem_id = 0;
+		return 0;
+	}
+
+	r = amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_GFX, &xcc_mask);
+	if (r || !xcc_mask)
+		return -EINVAL;
+
+	xcc_id = ffs(xcc_mask) - 1;
+	if (!adev->gmc.is_app_apu)
+		return __aqua_vanjaram_get_xcp_mem_id(adev, xcc_id, mem_id);
+
+	r = amdgpu_acpi_get_mem_info(adev, xcc_id, &numa_info);
+
+	if (r)
+		return r;
+
+	r = -EINVAL;
+	for (i = 0; i < adev->gmc.num_mem_partitions; ++i) {
+		if (adev->gmc.mem_partitions[i].numa.node == numa_info.nid) {
+			*mem_id = i;
+			r = 0;
+			break;
+		}
+	}
+
+	return r;
+}
+
+static int aqua_vanjaram_get_xcp_ip_details(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id,
+				     enum AMDGPU_XCP_IP_BLOCK ip_id,
+				     struct amdgpu_xcp_ip *ip)
+{
+	if (!ip)
+		return -EINVAL;
+
+	return __aqua_vanjaram_get_xcp_ip_info(xcp_mgr, xcp_id, ip_id, ip);
+}
+
+struct amdgpu_xcp_mgr_funcs aqua_vanjaram_xcp_funcs = {
+	.switch_partition_mode = &aqua_vanjaram_switch_partition_mode,
+	.query_partition_mode = &aqua_vanjaram_query_partition_mode,
+	.get_ip_details = &aqua_vanjaram_get_xcp_ip_details,
+	.get_xcp_res_info = &aqua_vanjaram_get_xcp_res_info,
+	.get_xcp_mem_id = &aqua_vanjaram_get_xcp_mem_id,
+};
+
+static int aqua_vanjaram_xcp_mgr_init(struct amdgpu_device *adev)
+{
+	int ret;
+
+	if (amdgpu_sriov_vf(adev))
+		aqua_vanjaram_xcp_funcs.switch_partition_mode = NULL;
+
+	ret = amdgpu_xcp_mgr_init(adev, AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE, 1,
+				  &aqua_vanjaram_xcp_funcs);
+	if (ret)
+		return ret;
+
+	amdgpu_xcp_update_supported_modes(adev->xcp_mgr);
+	/* TODO: Default memory node affinity init */
+
+	return ret;
+}
+
+int aqua_vanjaram_init_soc_config(struct amdgpu_device *adev)
+{
+	u32 mask, avail_inst, inst_mask = adev->sdma.sdma_mask;
+	int ret, i;
+
+	/* generally 1 AID supports 4 instances */
+	adev->sdma.num_inst_per_aid = 4;
+	adev->sdma.num_instances = NUM_SDMA(adev->sdma.sdma_mask);
+
+	adev->aid_mask = i = 1;
+	inst_mask >>= adev->sdma.num_inst_per_aid;
+
+	for (mask = (1 << adev->sdma.num_inst_per_aid) - 1; inst_mask;
+	     inst_mask >>= adev->sdma.num_inst_per_aid, ++i) {
+		avail_inst = inst_mask & mask;
+		if (avail_inst == mask || avail_inst == 0x3 ||
+		    avail_inst == 0xc)
+			adev->aid_mask |= (1 << i);
+	}
+
+	/* Harvest config is not used for aqua vanjaram. VCN and JPEGs will be
+	 * addressed based on logical instance ids.
+	 */
+	adev->vcn.harvest_config = 0;
+	adev->vcn.num_inst_per_aid = 1;
+	adev->vcn.num_vcn_inst = hweight32(adev->vcn.inst_mask);
+	adev->jpeg.harvest_config = 0;
+	adev->jpeg.num_inst_per_aid = 1;
+	adev->jpeg.num_jpeg_inst = hweight32(adev->jpeg.inst_mask);
+
+	ret = aqua_vanjaram_xcp_mgr_init(adev);
+	if (ret)
+		return ret;
+
+	amdgpu_ip_map_init(adev);
+
+	return 0;
+}
+
+static void aqua_read_smn(struct amdgpu_device *adev,
+			  struct amdgpu_smn_reg_data *regdata,
+			  uint64_t smn_addr)
+{
+	regdata->addr = smn_addr;
+	regdata->value = RREG32_PCIE(smn_addr);
+}
+
+struct aqua_reg_list {
+	uint64_t start_addr;
+	uint32_t num_regs;
+	uint32_t incrx;
+};
+
+#define DW_ADDR_INCR	4
+
+static void aqua_read_smn_ext(struct amdgpu_device *adev,
+			      struct amdgpu_smn_reg_data *regdata,
+			      uint64_t smn_addr, int i)
+{
+	regdata->addr =
+		smn_addr + adev->asic_funcs->encode_ext_smn_addressing(i);
+	regdata->value = RREG32_PCIE_EXT(regdata->addr);
+}
+
+#define smnreg_0x1A340218	0x1A340218
+#define smnreg_0x1A3402E4	0x1A3402E4
+#define smnreg_0x1A340294	0x1A340294
+#define smreg_0x1A380088	0x1A380088
+
+#define NUM_PCIE_SMN_REGS	14
+
+static struct aqua_reg_list pcie_reg_addrs[] = {
+	{ smnreg_0x1A340218, 1, 0 },
+	{ smnreg_0x1A3402E4, 1, 0 },
+	{ smnreg_0x1A340294, 6, DW_ADDR_INCR },
+	{ smreg_0x1A380088, 6, DW_ADDR_INCR },
+};
+
+static ssize_t aqua_vanjaram_read_pcie_state(struct amdgpu_device *adev,
+					     void *buf, size_t max_size)
+{
+	struct amdgpu_reg_state_pcie_v1_0 *pcie_reg_state;
+	uint32_t start_addr, incrx, num_regs, szbuf;
+	struct amdgpu_regs_pcie_v1_0 *pcie_regs;
+	struct amdgpu_smn_reg_data *reg_data;
+	struct pci_dev *us_pdev, *ds_pdev;
+	int aer_cap, r, n;
+
+	if (!buf || !max_size)
+		return -EINVAL;
+
+	pcie_reg_state = (struct amdgpu_reg_state_pcie_v1_0 *)buf;
+
+	szbuf = sizeof(*pcie_reg_state) +
+		amdgpu_reginst_size(1, sizeof(*pcie_regs), NUM_PCIE_SMN_REGS);
+	/* Only one instance of pcie regs */
+	if (max_size < szbuf)
+		return -EOVERFLOW;
+
+	pcie_regs = (struct amdgpu_regs_pcie_v1_0 *)((uint8_t *)buf +
+						     sizeof(*pcie_reg_state));
+	pcie_regs->inst_header.instance = 0;
+	pcie_regs->inst_header.state = AMDGPU_INST_S_OK;
+	pcie_regs->inst_header.num_smn_regs = NUM_PCIE_SMN_REGS;
+
+	reg_data = pcie_regs->smn_reg_values;
+
+	for (r = 0; r < ARRAY_SIZE(pcie_reg_addrs); r++) {
+		start_addr = pcie_reg_addrs[r].start_addr;
+		incrx = pcie_reg_addrs[r].incrx;
+		num_regs = pcie_reg_addrs[r].num_regs;
+		for (n = 0; n < num_regs; n++) {
+			aqua_read_smn(adev, reg_data, start_addr + n * incrx);
+			++reg_data;
+		}
+	}
+
+	ds_pdev = pci_upstream_bridge(adev->pdev);
+	us_pdev = pci_upstream_bridge(ds_pdev);
+
+	pcie_capability_read_word(us_pdev, PCI_EXP_DEVSTA,
+				  &pcie_regs->device_status);
+	pcie_capability_read_word(us_pdev, PCI_EXP_LNKSTA,
+				  &pcie_regs->link_status);
+
+	aer_cap = pci_find_ext_capability(us_pdev, PCI_EXT_CAP_ID_ERR);
+	if (aer_cap) {
+		pci_read_config_dword(us_pdev, aer_cap + PCI_ERR_COR_STATUS,
+				      &pcie_regs->pcie_corr_err_status);
+		pci_read_config_dword(us_pdev, aer_cap + PCI_ERR_UNCOR_STATUS,
+				      &pcie_regs->pcie_uncorr_err_status);
+	}
+
+	pci_read_config_dword(us_pdev, PCI_PRIMARY_BUS,
+			      &pcie_regs->sub_bus_number_latency);
+
+	pcie_reg_state->common_header.structure_size = szbuf;
+	pcie_reg_state->common_header.format_revision = 1;
+	pcie_reg_state->common_header.content_revision = 0;
+	pcie_reg_state->common_header.state_type = AMDGPU_REG_STATE_TYPE_PCIE;
+	pcie_reg_state->common_header.num_instances = 1;
+
+	return pcie_reg_state->common_header.structure_size;
+}
+
+#define smnreg_0x11A00050	0x11A00050
+#define smnreg_0x11A00180	0x11A00180
+#define smnreg_0x11A00070	0x11A00070
+#define smnreg_0x11A00200	0x11A00200
+#define smnreg_0x11A0020C	0x11A0020C
+#define smnreg_0x11A00210	0x11A00210
+#define smnreg_0x11A00108	0x11A00108
+
+#define XGMI_LINK_REG(smnreg, l) ((smnreg) | (l << 20))
+
+#define NUM_XGMI_SMN_REGS 25
+
+static struct aqua_reg_list xgmi_reg_addrs[] = {
+	{ smnreg_0x11A00050, 1, 0 },
+	{ smnreg_0x11A00180, 16, DW_ADDR_INCR },
+	{ smnreg_0x11A00070, 4, DW_ADDR_INCR },
+	{ smnreg_0x11A00200, 1, 0 },
+	{ smnreg_0x11A0020C, 1, 0 },
+	{ smnreg_0x11A00210, 1, 0 },
+	{ smnreg_0x11A00108, 1, 0 },
+};
+
+static ssize_t aqua_vanjaram_read_xgmi_state(struct amdgpu_device *adev,
+					     void *buf, size_t max_size)
+{
+	struct amdgpu_reg_state_xgmi_v1_0 *xgmi_reg_state;
+	uint32_t start_addr, incrx, num_regs, szbuf;
+	struct amdgpu_regs_xgmi_v1_0 *xgmi_regs;
+	struct amdgpu_smn_reg_data *reg_data;
+	const int max_xgmi_instances = 8;
+	int inst = 0, i, j, r, n;
+	const int xgmi_inst = 2;
+	void *p;
+
+	if (!buf || !max_size)
+		return -EINVAL;
+
+	xgmi_reg_state = (struct amdgpu_reg_state_xgmi_v1_0 *)buf;
+
+	szbuf = sizeof(*xgmi_reg_state) +
+		amdgpu_reginst_size(max_xgmi_instances, sizeof(*xgmi_regs),
+				    NUM_XGMI_SMN_REGS);
+	/* Only one instance of pcie regs */
+	if (max_size < szbuf)
+		return -EOVERFLOW;
+
+	p = &xgmi_reg_state->xgmi_state_regs[0];
+	for_each_inst(i, adev->aid_mask) {
+		for (j = 0; j < xgmi_inst; ++j) {
+			xgmi_regs = (struct amdgpu_regs_xgmi_v1_0 *)p;
+			xgmi_regs->inst_header.instance = inst++;
+
+			xgmi_regs->inst_header.state = AMDGPU_INST_S_OK;
+			xgmi_regs->inst_header.num_smn_regs = NUM_XGMI_SMN_REGS;
+
+			reg_data = xgmi_regs->smn_reg_values;
+
+			for (r = 0; r < ARRAY_SIZE(xgmi_reg_addrs); r++) {
+				start_addr = xgmi_reg_addrs[r].start_addr;
+				incrx = xgmi_reg_addrs[r].incrx;
+				num_regs = xgmi_reg_addrs[r].num_regs;
+
+				for (n = 0; n < num_regs; n++) {
+					aqua_read_smn_ext(
+						adev, reg_data,
+						XGMI_LINK_REG(start_addr, j) +
+							n * incrx,
+						i);
+					++reg_data;
+				}
+			}
+			p = reg_data;
+		}
+	}
+
+	xgmi_reg_state->common_header.structure_size = szbuf;
+	xgmi_reg_state->common_header.format_revision = 1;
+	xgmi_reg_state->common_header.content_revision = 0;
+	xgmi_reg_state->common_header.state_type = AMDGPU_REG_STATE_TYPE_XGMI;
+	xgmi_reg_state->common_header.num_instances = max_xgmi_instances;
+
+	return xgmi_reg_state->common_header.structure_size;
+}
+
+#define smnreg_0x11C00070	0x11C00070
+#define smnreg_0x11C00210	0x11C00210
+
+static struct aqua_reg_list wafl_reg_addrs[] = {
+	{ smnreg_0x11C00070, 4, DW_ADDR_INCR },
+	{ smnreg_0x11C00210, 1, 0 },
+};
+
+#define WAFL_LINK_REG(smnreg, l) ((smnreg) | (l << 20))
+
+#define NUM_WAFL_SMN_REGS 5
+
+static ssize_t aqua_vanjaram_read_wafl_state(struct amdgpu_device *adev,
+					     void *buf, size_t max_size)
+{
+	struct amdgpu_reg_state_wafl_v1_0 *wafl_reg_state;
+	uint32_t start_addr, incrx, num_regs, szbuf;
+	struct amdgpu_regs_wafl_v1_0 *wafl_regs;
+	struct amdgpu_smn_reg_data *reg_data;
+	const int max_wafl_instances = 8;
+	int inst = 0, i, j, r, n;
+	const int wafl_inst = 2;
+	void *p;
+
+	if (!buf || !max_size)
+		return -EINVAL;
+
+	wafl_reg_state = (struct amdgpu_reg_state_wafl_v1_0 *)buf;
+
+	szbuf = sizeof(*wafl_reg_state) +
+		amdgpu_reginst_size(max_wafl_instances, sizeof(*wafl_regs),
+				    NUM_WAFL_SMN_REGS);
+
+	if (max_size < szbuf)
+		return -EOVERFLOW;
+
+	p = &wafl_reg_state->wafl_state_regs[0];
+	for_each_inst(i, adev->aid_mask) {
+		for (j = 0; j < wafl_inst; ++j) {
+			wafl_regs = (struct amdgpu_regs_wafl_v1_0 *)p;
+			wafl_regs->inst_header.instance = inst++;
+
+			wafl_regs->inst_header.state = AMDGPU_INST_S_OK;
+			wafl_regs->inst_header.num_smn_regs = NUM_WAFL_SMN_REGS;
+
+			reg_data = wafl_regs->smn_reg_values;
+
+			for (r = 0; r < ARRAY_SIZE(wafl_reg_addrs); r++) {
+				start_addr = wafl_reg_addrs[r].start_addr;
+				incrx = wafl_reg_addrs[r].incrx;
+				num_regs = wafl_reg_addrs[r].num_regs;
+				for (n = 0; n < num_regs; n++) {
+					aqua_read_smn_ext(
+						adev, reg_data,
+						WAFL_LINK_REG(start_addr, j) +
+							n * incrx,
+						i);
+					++reg_data;
+				}
+			}
+			p = reg_data;
+		}
+	}
+
+	wafl_reg_state->common_header.structure_size = szbuf;
+	wafl_reg_state->common_header.format_revision = 1;
+	wafl_reg_state->common_header.content_revision = 0;
+	wafl_reg_state->common_header.state_type = AMDGPU_REG_STATE_TYPE_WAFL;
+	wafl_reg_state->common_header.num_instances = max_wafl_instances;
+
+	return wafl_reg_state->common_header.structure_size;
+}
+
+#define smnreg_0x1B311060 0x1B311060
+#define smnreg_0x1B411060 0x1B411060
+#define smnreg_0x1B511060 0x1B511060
+#define smnreg_0x1B611060 0x1B611060
+
+#define smnreg_0x1C307120 0x1C307120
+#define smnreg_0x1C317120 0x1C317120
+
+#define smnreg_0x1C320830 0x1C320830
+#define smnreg_0x1C380830 0x1C380830
+#define smnreg_0x1C3D0830 0x1C3D0830
+#define smnreg_0x1C420830 0x1C420830
+
+#define smnreg_0x1C320100 0x1C320100
+#define smnreg_0x1C380100 0x1C380100
+#define smnreg_0x1C3D0100 0x1C3D0100
+#define smnreg_0x1C420100 0x1C420100
+
+#define smnreg_0x1B310500 0x1B310500
+#define smnreg_0x1C300400 0x1C300400
+
+#define USR_CAKE_INCR 0x11000
+#define USR_LINK_INCR 0x100000
+#define USR_CP_INCR 0x10000
+
+#define NUM_USR_SMN_REGS	20
+
+struct aqua_reg_list usr_reg_addrs[] = {
+	{ smnreg_0x1B311060, 4, DW_ADDR_INCR },
+	{ smnreg_0x1B411060, 4, DW_ADDR_INCR },
+	{ smnreg_0x1B511060, 4, DW_ADDR_INCR },
+	{ smnreg_0x1B611060, 4, DW_ADDR_INCR },
+	{ smnreg_0x1C307120, 2, DW_ADDR_INCR },
+	{ smnreg_0x1C317120, 2, DW_ADDR_INCR },
+};
+
+#define NUM_USR1_SMN_REGS	46
+struct aqua_reg_list usr1_reg_addrs[] = {
+	{ smnreg_0x1C320830, 6, USR_CAKE_INCR },
+	{ smnreg_0x1C380830, 5, USR_CAKE_INCR },
+	{ smnreg_0x1C3D0830, 5, USR_CAKE_INCR },
+	{ smnreg_0x1C420830, 4, USR_CAKE_INCR },
+	{ smnreg_0x1C320100, 6, USR_CAKE_INCR },
+	{ smnreg_0x1C380100, 5, USR_CAKE_INCR },
+	{ smnreg_0x1C3D0100, 5, USR_CAKE_INCR },
+	{ smnreg_0x1C420100, 4, USR_CAKE_INCR },
+	{ smnreg_0x1B310500, 4, USR_LINK_INCR },
+	{ smnreg_0x1C300400, 2, USR_CP_INCR },
+};
+
+static ssize_t aqua_vanjaram_read_usr_state(struct amdgpu_device *adev,
+					    void *buf, size_t max_size,
+					    int reg_state)
+{
+	uint32_t start_addr, incrx, num_regs, szbuf, num_smn;
+	struct amdgpu_reg_state_usr_v1_0 *usr_reg_state;
+	struct amdgpu_regs_usr_v1_0 *usr_regs;
+	struct amdgpu_smn_reg_data *reg_data;
+	const int max_usr_instances = 4;
+	struct aqua_reg_list *reg_addrs;
+	int inst = 0, i, n, r, arr_size;
+	void *p;
+
+	if (!buf || !max_size)
+		return -EINVAL;
+
+	switch (reg_state) {
+	case AMDGPU_REG_STATE_TYPE_USR:
+		arr_size = ARRAY_SIZE(usr_reg_addrs);
+		reg_addrs = usr_reg_addrs;
+		num_smn = NUM_USR_SMN_REGS;
+		break;
+	case AMDGPU_REG_STATE_TYPE_USR_1:
+		arr_size = ARRAY_SIZE(usr1_reg_addrs);
+		reg_addrs = usr1_reg_addrs;
+		num_smn = NUM_USR1_SMN_REGS;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	usr_reg_state = (struct amdgpu_reg_state_usr_v1_0 *)buf;
+
+	szbuf = sizeof(*usr_reg_state) + amdgpu_reginst_size(max_usr_instances,
+							     sizeof(*usr_regs),
+							     num_smn);
+	if (max_size < szbuf)
+		return -EOVERFLOW;
+
+	p = &usr_reg_state->usr_state_regs[0];
+	for_each_inst(i, adev->aid_mask) {
+		usr_regs = (struct amdgpu_regs_usr_v1_0 *)p;
+		usr_regs->inst_header.instance = inst++;
+		usr_regs->inst_header.state = AMDGPU_INST_S_OK;
+		usr_regs->inst_header.num_smn_regs = num_smn;
+		reg_data = usr_regs->smn_reg_values;
+
+		for (r = 0; r < arr_size; r++) {
+			start_addr = reg_addrs[r].start_addr;
+			incrx = reg_addrs[r].incrx;
+			num_regs = reg_addrs[r].num_regs;
+			for (n = 0; n < num_regs; n++) {
+				aqua_read_smn_ext(adev, reg_data,
+						  start_addr + n * incrx, i);
+				reg_data++;
+			}
+		}
+		p = reg_data;
+	}
+
+	usr_reg_state->common_header.structure_size = szbuf;
+	usr_reg_state->common_header.format_revision = 1;
+	usr_reg_state->common_header.content_revision = 0;
+	usr_reg_state->common_header.state_type = AMDGPU_REG_STATE_TYPE_USR;
+	usr_reg_state->common_header.num_instances = max_usr_instances;
+
+	return usr_reg_state->common_header.structure_size;
+}
+
+ssize_t aqua_vanjaram_get_reg_state(struct amdgpu_device *adev,
+				    enum amdgpu_reg_state reg_state, void *buf,
+				    size_t max_size)
+{
+	ssize_t size;
+
+	switch (reg_state) {
+	case AMDGPU_REG_STATE_TYPE_PCIE:
+		size = aqua_vanjaram_read_pcie_state(adev, buf, max_size);
+		break;
+	case AMDGPU_REG_STATE_TYPE_XGMI:
+		size = aqua_vanjaram_read_xgmi_state(adev, buf, max_size);
+		break;
+	case AMDGPU_REG_STATE_TYPE_WAFL:
+		size = aqua_vanjaram_read_wafl_state(adev, buf, max_size);
+		break;
+	case AMDGPU_REG_STATE_TYPE_USR:
+		size = aqua_vanjaram_read_usr_state(adev, buf, max_size,
+						    AMDGPU_REG_STATE_TYPE_USR);
+		break;
+	case AMDGPU_REG_STATE_TYPE_USR_1:
+		size = aqua_vanjaram_read_usr_state(
+			adev, buf, max_size, AMDGPU_REG_STATE_TYPE_USR_1);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return size;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/athub_v1_0.c
index a13c443ea10f..42f4e163e251 100644
--- a/drivers/gpu/drm/amd/amdgpu/athub_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/athub_v1_0.c
@@ -68,7 +68,7 @@ int athub_v1_0_set_clockgating(struct amdgpu_device *adev,
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[ATHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, ATHUB_HWIP, 0)) {
 	case IP_VERSION(9, 0, 0):
 	case IP_VERSION(9, 1, 0):
 	case IP_VERSION(9, 2, 0):
diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v2_0.c b/drivers/gpu/drm/amd/amdgpu/athub_v2_0.c
index a9521c98e7f7..5a122f50a6e7 100644
--- a/drivers/gpu/drm/amd/amdgpu/athub_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/athub_v2_0.c
@@ -77,7 +77,7 @@ int athub_v2_0_set_clockgating(struct amdgpu_device *adev,
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[ATHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, ATHUB_HWIP, 0)) {
 	case IP_VERSION(1, 3, 1):
 	case IP_VERSION(2, 0, 0):
 	case IP_VERSION(2, 0, 2):
diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v2_1.c b/drivers/gpu/drm/amd/amdgpu/athub_v2_1.c
index 78508ae6a670..e143fcc46148 100644
--- a/drivers/gpu/drm/amd/amdgpu/athub_v2_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/athub_v2_1.c
@@ -70,7 +70,7 @@ int athub_v2_1_set_clockgating(struct amdgpu_device *adev,
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[ATHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, ATHUB_HWIP, 0)) {
 	case IP_VERSION(2, 1, 0):
 	case IP_VERSION(2, 1, 1):
 	case IP_VERSION(2, 1, 2):
diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v3_0.c b/drivers/gpu/drm/amd/amdgpu/athub_v3_0.c
index f0e235f98afb..d1bba9c64e16 100644
--- a/drivers/gpu/drm/amd/amdgpu/athub_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/athub_v3_0.c
@@ -30,16 +30,21 @@
 
 #define regATHUB_MISC_CNTL_V3_0_1			0x00d7
 #define regATHUB_MISC_CNTL_V3_0_1_BASE_IDX		0
+#define regATHUB_MISC_CNTL_V3_3_0			0x00d8
+#define regATHUB_MISC_CNTL_V3_3_0_BASE_IDX		0
 
 
 static uint32_t athub_v3_0_get_cg_cntl(struct amdgpu_device *adev)
 {
 	uint32_t data;
 
-	switch (adev->ip_versions[ATHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, ATHUB_HWIP, 0)) {
 	case IP_VERSION(3, 0, 1):
 		data = RREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL_V3_0_1);
 		break;
+	case IP_VERSION(3, 3, 0):
+		data = RREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL_V3_3_0);
+		break;
 	default:
 		data = RREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL);
 		break;
@@ -49,10 +54,13 @@ static uint32_t athub_v3_0_get_cg_cntl(struct amdgpu_device *adev)
 
 static void athub_v3_0_set_cg_cntl(struct amdgpu_device *adev, uint32_t data)
 {
-	switch (adev->ip_versions[ATHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, ATHUB_HWIP, 0)) {
 	case IP_VERSION(3, 0, 1):
 		WREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL_V3_0_1, data);
 		break;
+	case IP_VERSION(3, 3, 0):
+		WREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL_V3_3_0, data);
+		break;
 	default:
 		WREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL, data);
 		break;
@@ -99,10 +107,11 @@ int athub_v3_0_set_clockgating(struct amdgpu_device *adev,
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[ATHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, ATHUB_HWIP, 0)) {
 	case IP_VERSION(3, 0, 0):
 	case IP_VERSION(3, 0, 1):
 	case IP_VERSION(3, 0, 2):
+	case IP_VERSION(3, 3, 0):
 		athub_v3_0_update_medium_grain_clock_gating(adev,
 				state == AMD_CG_STATE_GATE);
 		athub_v3_0_update_medium_grain_light_sleep(adev,
diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v4_1_0.c b/drivers/gpu/drm/amd/amdgpu/athub_v4_1_0.c
new file mode 100644
index 000000000000..8a0773b80864
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/athub_v4_1_0.c
@@ -0,0 +1,122 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "amdgpu.h"
+#include "athub_v4_1_0.h"
+#include "athub/athub_4_1_0_offset.h"
+#include "athub/athub_4_1_0_sh_mask.h"
+#include "soc15_common.h"
+
+static uint32_t athub_v4_1_0_get_cg_cntl(struct amdgpu_device *adev)
+{
+	uint32_t data;
+
+	switch (amdgpu_ip_version(adev, ATHUB_HWIP, 0)) {
+	case IP_VERSION(4, 1, 0):
+		data = RREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL);
+		break;
+	default:
+		data = 0;
+		break;
+	}
+	return data;
+}
+
+static void athub_v4_1_0_set_cg_cntl(struct amdgpu_device *adev, uint32_t data)
+{
+	switch (amdgpu_ip_version(adev, ATHUB_HWIP, 0)) {
+	case IP_VERSION(4, 1, 0):
+		WREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL, data);
+		break;
+	default:
+		break;
+	}
+}
+
+static void
+athub_v4_1_0_update_medium_grain_clock_gating(struct amdgpu_device *adev,
+					      bool enable)
+{
+	uint32_t def, data;
+
+	def = data = athub_v4_1_0_get_cg_cntl(adev);
+
+	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_ATHUB_MGCG))
+		data |= ATHUB_MISC_CNTL__CG_ENABLE_MASK;
+	else
+		data &= ~ATHUB_MISC_CNTL__CG_ENABLE_MASK;
+
+	if (def != data)
+		athub_v4_1_0_set_cg_cntl(adev, data);
+}
+
+static void
+athub_v4_1_0_update_medium_grain_light_sleep(struct amdgpu_device *adev,
+					     bool enable)
+{
+	uint32_t def, data;
+
+	def = data = athub_v4_1_0_get_cg_cntl(adev);
+
+	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_ATHUB_LS))
+		data |= ATHUB_MISC_CNTL__CG_MEM_LS_ENABLE_MASK;
+	else
+		data &= ~ATHUB_MISC_CNTL__CG_MEM_LS_ENABLE_MASK;
+
+	if (def != data)
+		athub_v4_1_0_set_cg_cntl(adev, data);
+}
+
+int athub_v4_1_0_set_clockgating(struct amdgpu_device *adev,
+				 enum amd_clockgating_state state)
+{
+	if (amdgpu_sriov_vf(adev))
+		return 0;
+
+	switch (amdgpu_ip_version(adev, ATHUB_HWIP, 0)) {
+	case IP_VERSION(4, 1, 0):
+		athub_v4_1_0_update_medium_grain_clock_gating(adev,
+				state == AMD_CG_STATE_GATE);
+		athub_v4_1_0_update_medium_grain_light_sleep(adev,
+				state == AMD_CG_STATE_GATE);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+void athub_v4_1_0_get_clockgating(struct amdgpu_device *adev, u64 *flags)
+{
+	int data;
+
+	/* AMD_CG_SUPPORT_ATHUB_MGCG */
+	data = athub_v4_1_0_get_cg_cntl(adev);
+	if (data & ATHUB_MISC_CNTL__CG_ENABLE_MASK)
+		*flags |= AMD_CG_SUPPORT_ATHUB_MGCG;
+
+	/* AMD_CG_SUPPORT_ATHUB_LS */
+	if (data & ATHUB_MISC_CNTL__CG_MEM_LS_ENABLE_MASK)
+		*flags |= AMD_CG_SUPPORT_ATHUB_LS;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v4_1_0.h b/drivers/gpu/drm/amd/amdgpu/athub_v4_1_0.h
new file mode 100644
index 000000000000..4d18d0998fa8
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/athub_v4_1_0.h
@@ -0,0 +1,30 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __ATHUB_V4_1_0_H__
+#define __ATHUB_V4_1_0_H__
+
+int athub_v4_1_0_set_clockgating(struct amdgpu_device *adev,
+				 enum amd_clockgating_state state);
+void athub_v4_1_0_get_clockgating(struct amdgpu_device *adev, u64 *flags);
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/atom.c b/drivers/gpu/drm/amd/amdgpu/atom.c
index 1c5d9388ad0b..7a063e44d429 100644
--- a/drivers/gpu/drm/amd/amdgpu/atom.c
+++ b/drivers/gpu/drm/amd/amdgpu/atom.c
@@ -27,7 +27,7 @@
 #include <linux/slab.h>
 #include <linux/string_helpers.h>
 
-#include <asm/unaligned.h>
+#include <linux/unaligned.h>
 
 #include <drm/drm_util.h>
 
@@ -62,6 +62,7 @@
 typedef struct {
 	struct atom_context *ctx;
 	uint32_t *ps, *ws;
+	int ps_size, ws_size;
 	int ps_shift;
 	uint16_t start;
 	unsigned last_jump;
@@ -70,8 +71,8 @@ typedef struct {
 } atom_exec_context;
 
 int amdgpu_atom_debug;
-static int amdgpu_atom_execute_table_locked(struct atom_context *ctx, int index, uint32_t *params);
-int amdgpu_atom_execute_table(struct atom_context *ctx, int index, uint32_t *params);
+static int amdgpu_atom_execute_table_locked(struct atom_context *ctx, int index, uint32_t *params, int params_size);
+int amdgpu_atom_execute_table(struct atom_context *ctx, int index, uint32_t *params, int params_size);
 
 static uint32_t atom_arg_mask[8] =
 	{ 0xFFFFFFFF, 0xFFFF, 0xFFFF00, 0xFFFF0000, 0xFF, 0xFF00, 0xFF0000,
@@ -223,7 +224,10 @@ static uint32_t atom_get_src_int(atom_exec_context *ctx, uint8_t attr,
 		(*ptr)++;
 		/* get_unaligned_le32 avoids unaligned accesses from atombios
 		 * tables, noticed on a DEC Alpha. */
-		val = get_unaligned_le32((u32 *)&ctx->ps[idx]);
+		if (idx < ctx->ps_size)
+			val = get_unaligned_le32((u32 *)&ctx->ps[idx]);
+		else
+			pr_info("PS index out of range: %i > %i\n", idx, ctx->ps_size);
 		if (print)
 			DEBUG("PS[0x%02X,0x%04X]", idx, val);
 		break;
@@ -261,7 +265,10 @@ static uint32_t atom_get_src_int(atom_exec_context *ctx, uint8_t attr,
 			val = gctx->reg_block;
 			break;
 		default:
-			val = ctx->ws[idx];
+			if (idx < ctx->ws_size)
+				val = ctx->ws[idx];
+			else
+				pr_info("WS index out of range: %i > %i\n", idx, ctx->ws_size);
 		}
 		break;
 	case ATOM_ARG_ID:
@@ -294,7 +301,7 @@ static uint32_t atom_get_src_int(atom_exec_context *ctx, uint8_t attr,
 			(*ptr) += 4;
 			if (print)
 				DEBUG("IMM 0x%08X\n", val);
-			return val;
+			break;
 		case ATOM_SRC_WORD0:
 		case ATOM_SRC_WORD8:
 		case ATOM_SRC_WORD16:
@@ -302,7 +309,7 @@ static uint32_t atom_get_src_int(atom_exec_context *ctx, uint8_t attr,
 			(*ptr) += 2;
 			if (print)
 				DEBUG("IMM 0x%04X\n", val);
-			return val;
+			break;
 		case ATOM_SRC_BYTE0:
 		case ATOM_SRC_BYTE8:
 		case ATOM_SRC_BYTE16:
@@ -311,9 +318,9 @@ static uint32_t atom_get_src_int(atom_exec_context *ctx, uint8_t attr,
 			(*ptr)++;
 			if (print)
 				DEBUG("IMM 0x%02X\n", val);
-			return val;
+			break;
 		}
-		return 0;
+		return val;
 	case ATOM_ARG_PLL:
 		idx = U8(*ptr);
 		(*ptr)++;
@@ -395,7 +402,6 @@ static void atom_skip_src_int(atom_exec_context *ctx, uint8_t attr, int *ptr)
 			(*ptr)++;
 			return;
 		}
-		return;
 	}
 }
 
@@ -496,6 +502,10 @@ static void atom_put_dst(atom_exec_context *ctx, int arg, uint8_t attr,
 		idx = U8(*ptr);
 		(*ptr)++;
 		DEBUG("PS[0x%02X]", idx);
+		if (idx >= ctx->ps_size) {
+			pr_info("PS index out of range: %i > %i\n", idx, ctx->ps_size);
+			return;
+		}
 		ctx->ps[idx] = cpu_to_le32(val);
 		break;
 	case ATOM_ARG_WS:
@@ -528,6 +538,10 @@ static void atom_put_dst(atom_exec_context *ctx, int arg, uint8_t attr,
 			gctx->reg_block = val;
 			break;
 		default:
+			if (idx >= ctx->ws_size) {
+				pr_info("WS index out of range: %i > %i\n", idx, ctx->ws_size);
+				return;
+			}
 			ctx->ws[idx] = val;
 		}
 		break;
@@ -625,7 +639,7 @@ static void atom_op_calltable(atom_exec_context *ctx, int *ptr, int arg)
 	else
 		SDEBUG("   table: %d\n", idx);
 	if (U16(ctx->ctx->cmd_table + 4 + 2 * idx))
-		r = amdgpu_atom_execute_table_locked(ctx->ctx, idx, ctx->ps + ctx->ps_shift);
+		r = amdgpu_atom_execute_table_locked(ctx->ctx, idx, ctx->ps + ctx->ps_shift, ctx->ps_size - ctx->ps_shift);
 	if (r) {
 		ctx->abort = true;
 	}
@@ -1204,7 +1218,7 @@ static struct {
 	atom_op_div32, ATOM_ARG_WS},
 };
 
-static int amdgpu_atom_execute_table_locked(struct atom_context *ctx, int index, uint32_t *params)
+static int amdgpu_atom_execute_table_locked(struct atom_context *ctx, int index, uint32_t *params, int params_size)
 {
 	int base = CU16(ctx->cmd_table + 4 + 2 * index);
 	int len, ws, ps, ptr;
@@ -1226,12 +1240,21 @@ static int amdgpu_atom_execute_table_locked(struct atom_context *ctx, int index,
 	ectx.ps_shift = ps / 4;
 	ectx.start = base;
 	ectx.ps = params;
+	ectx.ps_size = params_size;
 	ectx.abort = false;
 	ectx.last_jump = 0;
-	if (ws)
+	ectx.last_jump_jiffies = 0;
+	if (ws) {
 		ectx.ws = kcalloc(4, ws, GFP_KERNEL);
-	else
+		if (!ectx.ws) {
+			ret = -ENOMEM;
+			goto free;
+		}
+		ectx.ws_size = ws;
+	} else {
 		ectx.ws = NULL;
+		ectx.ws_size = 0;
+	}
 
 	debug_depth++;
 	while (1) {
@@ -1265,7 +1288,7 @@ free:
 	return ret;
 }
 
-int amdgpu_atom_execute_table(struct atom_context *ctx, int index, uint32_t *params)
+int amdgpu_atom_execute_table(struct atom_context *ctx, int index, uint32_t *params, int params_size)
 {
 	int r;
 
@@ -1281,7 +1304,7 @@ int amdgpu_atom_execute_table(struct atom_context *ctx, int index, uint32_t *par
 	/* reset divmul */
 	ctx->divmul[0] = 0;
 	ctx->divmul[1] = 0;
-	r = amdgpu_atom_execute_table_locked(ctx, index, params);
+	r = amdgpu_atom_execute_table_locked(ctx, index, params, params_size);
 	mutex_unlock(&ctx->mutex);
 	return r;
 }
@@ -1425,6 +1448,7 @@ static void atom_get_vbios_pn(struct atom_context *ctx)
 		if (vbios_str == NULL)
 			vbios_str += sizeof(BIOS_ATOM_PREFIX) - 1;
 	}
+	OPTIMIZER_HIDE_VAR(vbios_str);
 	if (vbios_str != NULL && *vbios_str == 0)
 		vbios_str++;
 
@@ -1438,14 +1462,33 @@ static void atom_get_vbios_pn(struct atom_context *ctx)
 
 		ctx->vbios_pn[count] = 0;
 	}
+
+	pr_info("ATOM BIOS: %s\n", ctx->vbios_pn);
 }
 
 static void atom_get_vbios_version(struct atom_context *ctx)
 {
+	unsigned short start = 3, end;
 	unsigned char *vbios_ver;
+	unsigned char *p_rom;
+
+	p_rom = ctx->bios;
+	/* Search from strings offset if it's present */
+	start = *(unsigned short *)(p_rom +
+				    OFFSET_TO_GET_ATOMBIOS_STRING_START);
+
+	/* Search till atom rom header start point */
+	end = *(unsigned short *)(p_rom + OFFSET_TO_ATOM_ROM_HEADER_POINTER);
+
+	/* Use hardcoded offsets, if the offsets are not populated */
+	if (end <= start) {
+		start = 3;
+		end = 1024;
+	}
 
 	/* find anchor ATOMBIOSBK-AMD */
-	vbios_ver = atom_find_str_in_rom(ctx, BIOS_VERSION_PREFIX, 3, 1024, 64);
+	vbios_ver =
+		atom_find_str_in_rom(ctx, BIOS_VERSION_PREFIX, start, end, 64);
 	if (vbios_ver != NULL) {
 		/* skip ATOMBIOSBK-AMD VER */
 		vbios_ver += 18;
@@ -1455,16 +1498,36 @@ static void atom_get_vbios_version(struct atom_context *ctx)
 	}
 }
 
+static void atom_get_vbios_build(struct atom_context *ctx)
+{
+	unsigned char *atom_rom_hdr;
+	unsigned char *str;
+	uint16_t base, len;
+
+	base = CU16(ATOM_ROM_TABLE_PTR);
+	atom_rom_hdr = CSTR(base);
+
+	str = CSTR(CU16(base + ATOM_ROM_CFG_PTR));
+	/* Skip config string */
+	while (str < atom_rom_hdr && *str++)
+		;
+	/* Skip change list string */
+	while (str < atom_rom_hdr && *str++)
+		;
+
+	len = min(atom_rom_hdr - str, STRLEN_NORMAL);
+	if (len)
+		strscpy(ctx->build_num, str, len);
+}
+
 struct atom_context *amdgpu_atom_parse(struct card_info *card, void *bios)
 {
 	int base;
 	struct atom_context *ctx =
 	    kzalloc(sizeof(struct atom_context), GFP_KERNEL);
-	char *str;
 	struct _ATOM_ROM_HEADER *atom_rom_header;
 	struct _ATOM_MASTER_DATA_TABLE *master_table;
 	struct _ATOM_FIRMWARE_INFO *atom_fw_info;
-	u16 idx;
 
 	if (!ctx)
 		return NULL;
@@ -1502,16 +1565,6 @@ struct atom_context *amdgpu_atom_parse(struct card_info *card, void *bios)
 		return NULL;
 	}
 
-	idx = CU16(ATOM_ROM_PART_NUMBER_PTR);
-	if (idx == 0)
-		idx = 0x80;
-
-	str = CSTR(idx);
-	if (*str != '\0') {
-		pr_info("ATOM BIOS: %s\n", str);
-		strlcpy(ctx->vbios_version, str, sizeof(ctx->vbios_version));
-	}
-
 	atom_rom_header = (struct _ATOM_ROM_HEADER *)CSTR(base);
 	if (atom_rom_header->usMasterDataTableOffset != 0) {
 		master_table = (struct _ATOM_MASTER_DATA_TABLE *)
@@ -1527,6 +1580,7 @@ struct atom_context *amdgpu_atom_parse(struct card_info *card, void *bios)
 	atom_get_vbios_pn(ctx);
 	atom_get_vbios_date(ctx);
 	atom_get_vbios_version(ctx);
+	atom_get_vbios_build(ctx);
 
 	return ctx;
 }
@@ -1546,7 +1600,7 @@ int amdgpu_atom_asic_init(struct atom_context *ctx)
 
 	if (!CU16(ctx->cmd_table + 4 + 2 * ATOM_CMD_INIT))
 		return 1;
-	ret = amdgpu_atom_execute_table(ctx, ATOM_CMD_INIT, ps);
+	ret = amdgpu_atom_execute_table(ctx, ATOM_CMD_INIT, ps, 16);
 	if (ret)
 		return ret;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/atom.h b/drivers/gpu/drm/amd/amdgpu/atom.h
index 0c1839824520..825ff28731f5 100644
--- a/drivers/gpu/drm/amd/amdgpu/atom.h
+++ b/drivers/gpu/drm/amd/amdgpu/atom.h
@@ -33,11 +33,11 @@ struct drm_device;
 #define ATOM_ATI_MAGIC_PTR	0x30
 #define ATOM_ATI_MAGIC		" 761295520"
 #define ATOM_ROM_TABLE_PTR	0x48
-#define ATOM_ROM_PART_NUMBER_PTR	0x6E
 
 #define ATOM_ROM_MAGIC		"ATOM"
 #define ATOM_ROM_MAGIC_PTR	4
 
+#define ATOM_ROM_CFG_PTR 0xC
 #define ATOM_ROM_MSG_PTR	0x10
 #define ATOM_ROM_CMD_PTR	0x1E
 #define ATOM_ROM_DATA_PTR	0x20
@@ -118,12 +118,15 @@ struct drm_device;
 
 struct card_info {
 	struct drm_device *dev;
-	void (* reg_write)(struct card_info *, uint32_t, uint32_t);   /*  filled by driver */
-	uint32_t (* reg_read)(struct card_info *, uint32_t);          /*  filled by driver */
-	void (* mc_write)(struct card_info *, uint32_t, uint32_t);   /*  filled by driver */
-	uint32_t (* mc_read)(struct card_info *, uint32_t);          /*  filled by driver */
-	void (* pll_write)(struct card_info *, uint32_t, uint32_t);   /*  filled by driver */
-	uint32_t (* pll_read)(struct card_info *, uint32_t);          /*  filled by driver */
+	void (*reg_write)(struct card_info *info,
+			  u32 reg, uint32_t val);   /*  filled by driver */
+	uint32_t (*reg_read)(struct card_info *info, uint32_t reg);          /*  filled by driver */
+	void (*mc_write)(struct card_info *info,
+			 u32 reg, uint32_t val);   /*  filled by driver */
+	uint32_t (*mc_read)(struct card_info *info, uint32_t reg);          /*  filled by driver */
+	void (*pll_write)(struct card_info *info,
+			  u32 reg, uint32_t val);   /*  filled by driver */
+	uint32_t (*pll_read)(struct card_info *info, uint32_t reg);          /*  filled by driver */
 };
 
 struct atom_context {
@@ -143,21 +146,21 @@ struct atom_context {
 	int io_mode;
 	uint32_t *scratch;
 	int scratch_size_bytes;
-	char vbios_version[20];
 
 	uint8_t name[STRLEN_LONG];
 	uint8_t vbios_pn[STRLEN_LONG];
 	uint32_t version;
 	uint8_t vbios_ver_str[STRLEN_NORMAL];
 	uint8_t date[STRLEN_NORMAL];
+	uint8_t build_num[STRLEN_NORMAL];
 };
 
 extern int amdgpu_atom_debug;
 
-struct atom_context *amdgpu_atom_parse(struct card_info *, void *);
-int amdgpu_atom_execute_table(struct atom_context *, int, uint32_t *);
-int amdgpu_atom_asic_init(struct atom_context *);
-void amdgpu_atom_destroy(struct atom_context *);
+struct atom_context *amdgpu_atom_parse(struct card_info *card, void *bios);
+int amdgpu_atom_execute_table(struct atom_context *ctx, int index, uint32_t *params, int params_size);
+int amdgpu_atom_asic_init(struct atom_context *ctx);
+void amdgpu_atom_destroy(struct atom_context *ctx);
 bool amdgpu_atom_parse_data_header(struct atom_context *ctx, int index, uint16_t *size,
 			    uint8_t *frev, uint8_t *crev, uint16_t *data_start);
 bool amdgpu_atom_parse_cmd_header(struct atom_context *ctx, int index,
diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_crtc.c b/drivers/gpu/drm/amd/amdgpu/atombios_crtc.c
index 10098fdd33fc..3dfc28840a7d 100644
--- a/drivers/gpu/drm/amd/amdgpu/atombios_crtc.c
+++ b/drivers/gpu/drm/amd/amdgpu/atombios_crtc.c
@@ -77,7 +77,7 @@ void amdgpu_atombios_crtc_overscan_setup(struct drm_crtc *crtc,
 		args.usOverscanTop = cpu_to_le16(amdgpu_crtc->v_border);
 		break;
 	}
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 void amdgpu_atombios_crtc_scaler_setup(struct drm_crtc *crtc)
@@ -106,7 +106,7 @@ void amdgpu_atombios_crtc_scaler_setup(struct drm_crtc *crtc)
 		args.ucEnable = ATOM_SCALER_DISABLE;
 		break;
 	}
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 void amdgpu_atombios_crtc_lock(struct drm_crtc *crtc, int lock)
@@ -123,7 +123,7 @@ void amdgpu_atombios_crtc_lock(struct drm_crtc *crtc, int lock)
 	args.ucCRTC = amdgpu_crtc->crtc_id;
 	args.ucEnable = lock;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 void amdgpu_atombios_crtc_enable(struct drm_crtc *crtc, int state)
@@ -139,7 +139,7 @@ void amdgpu_atombios_crtc_enable(struct drm_crtc *crtc, int state)
 	args.ucCRTC = amdgpu_crtc->crtc_id;
 	args.ucEnable = state;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 void amdgpu_atombios_crtc_blank(struct drm_crtc *crtc, int state)
@@ -155,7 +155,7 @@ void amdgpu_atombios_crtc_blank(struct drm_crtc *crtc, int state)
 	args.ucCRTC = amdgpu_crtc->crtc_id;
 	args.ucBlanking = state;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 void amdgpu_atombios_crtc_powergate(struct drm_crtc *crtc, int state)
@@ -171,7 +171,7 @@ void amdgpu_atombios_crtc_powergate(struct drm_crtc *crtc, int state)
 	args.ucDispPipeId = amdgpu_crtc->crtc_id;
 	args.ucEnable = state;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 void amdgpu_atombios_crtc_powergate_init(struct amdgpu_device *adev)
@@ -183,7 +183,7 @@ void amdgpu_atombios_crtc_powergate_init(struct amdgpu_device *adev)
 
 	args.ucEnable = ATOM_INIT;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 void amdgpu_atombios_crtc_set_dtd_timing(struct drm_crtc *crtc,
@@ -228,7 +228,7 @@ void amdgpu_atombios_crtc_set_dtd_timing(struct drm_crtc *crtc,
 	args.susModeMiscInfo.usAccess = cpu_to_le16(misc);
 	args.ucCRTC = amdgpu_crtc->crtc_id;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 union atom_enable_ss {
@@ -293,7 +293,7 @@ static void amdgpu_atombios_crtc_program_ss(struct amdgpu_device *adev,
 	args.v3.usSpreadSpectrumStep = cpu_to_le16(ss->step);
 	args.v3.ucEnable = enable;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 union adjust_pixel_clock {
@@ -395,7 +395,7 @@ static u32 amdgpu_atombios_crtc_adjust_pll(struct drm_crtc *crtc,
 					ADJUST_DISPLAY_CONFIG_SS_ENABLE;
 
 			amdgpu_atom_execute_table(adev->mode_info.atom_context,
-					   index, (uint32_t *)&args);
+					   index, (uint32_t *)&args, sizeof(args));
 			adjusted_clock = le16_to_cpu(args.v1.usPixelClock) * 10;
 			break;
 		case 3:
@@ -428,7 +428,7 @@ static u32 amdgpu_atombios_crtc_adjust_pll(struct drm_crtc *crtc,
 				args.v3.sInput.ucExtTransmitterID = 0;
 
 			amdgpu_atom_execute_table(adev->mode_info.atom_context,
-					   index, (uint32_t *)&args);
+					   index, (uint32_t *)&args, sizeof(args));
 			adjusted_clock = le32_to_cpu(args.v3.sOutput.ulDispPllFreq) * 10;
 			if (args.v3.sOutput.ucRefDiv) {
 				amdgpu_crtc->pll_flags |= AMDGPU_PLL_USE_FRAC_FB_DIV;
@@ -514,7 +514,7 @@ void amdgpu_atombios_crtc_set_disp_eng_pll(struct amdgpu_device *adev,
 		DRM_ERROR("Unknown table version %d %d\n", frev, crev);
 		return;
 	}
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 union set_dce_clock {
@@ -544,7 +544,7 @@ u32 amdgpu_atombios_crtc_set_dce_clock(struct amdgpu_device *adev,
 			args.v2_1.asParam.ulDCEClkFreq = cpu_to_le32(freq); /* 10kHz units */
 			args.v2_1.asParam.ucDCEClkType = clk_type;
 			args.v2_1.asParam.ucDCEClkSrc = clk_src;
-			amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+			amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 			ret_freq = le32_to_cpu(args.v2_1.asParam.ulDCEClkFreq) * 10;
 			break;
 		default:
@@ -740,7 +740,7 @@ void amdgpu_atombios_crtc_program_pll(struct drm_crtc *crtc,
 		return;
 	}
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 int amdgpu_atombios_crtc_prepare_pll(struct drm_crtc *crtc,
diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_dp.c b/drivers/gpu/drm/amd/amdgpu/atombios_dp.c
index 87c41e0e9b7c..492813ab1b54 100644
--- a/drivers/gpu/drm/amd/amdgpu/atombios_dp.c
+++ b/drivers/gpu/drm/amd/amdgpu/atombios_dp.c
@@ -83,7 +83,7 @@ static int amdgpu_atombios_dp_process_aux_ch(struct amdgpu_i2c_chan *chan,
 	args.v2.ucDelay = delay / 10;
 	args.v2.ucHPD_ID = chan->rec.hpd;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 
 	*ack = args.v2.ucReplyStatus;
 
@@ -301,7 +301,7 @@ static u8 amdgpu_atombios_dp_encoder_service(struct amdgpu_device *adev,
 	args.ucLaneNum = lane_num;
 	args.ucStatus = 0;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 	return args.ucStatus;
 }
 
@@ -430,7 +430,7 @@ void amdgpu_atombios_dp_set_link_config(struct drm_connector *connector,
 }
 
 int amdgpu_atombios_dp_mode_valid_helper(struct drm_connector *connector,
-				  struct drm_display_mode *mode)
+				  const struct drm_display_mode *mode)
 {
 	struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
 	struct amdgpu_connector_atom_dig *dig_connector;
@@ -458,8 +458,8 @@ bool amdgpu_atombios_dp_needs_link_train(struct amdgpu_connector *amdgpu_connect
 	u8 link_status[DP_LINK_STATUS_SIZE];
 	struct amdgpu_connector_atom_dig *dig = amdgpu_connector->con_priv;
 
-	if (drm_dp_dpcd_read_link_status(&amdgpu_connector->ddc_bus->aux, link_status)
-	    <= 0)
+	if (drm_dp_dpcd_read_link_status(&amdgpu_connector->ddc_bus->aux,
+					 link_status) < 0)
 		return false;
 	if (drm_dp_channel_eq_ok(link_status, dig->dp_lane_count))
 		return false;
@@ -616,7 +616,7 @@ amdgpu_atombios_dp_link_train_cr(struct amdgpu_atombios_dp_link_train_info *dp_i
 		drm_dp_link_train_clock_recovery_delay(dp_info->aux, dp_info->dpcd);
 
 		if (drm_dp_dpcd_read_link_status(dp_info->aux,
-						 dp_info->link_status) <= 0) {
+						 dp_info->link_status) < 0) {
 			DRM_ERROR("displayport link status failed\n");
 			break;
 		}
@@ -681,7 +681,7 @@ amdgpu_atombios_dp_link_train_ce(struct amdgpu_atombios_dp_link_train_info *dp_i
 		drm_dp_link_train_channel_eq_delay(dp_info->aux, dp_info->dpcd);
 
 		if (drm_dp_dpcd_read_link_status(dp_info->aux,
-						 dp_info->link_status) <= 0) {
+						 dp_info->link_status) < 0) {
 			DRM_ERROR("displayport link status failed\n");
 			break;
 		}
diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_dp.h b/drivers/gpu/drm/amd/amdgpu/atombios_dp.h
index f59d85eaddf0..3e24acf8133f 100644
--- a/drivers/gpu/drm/amd/amdgpu/atombios_dp.h
+++ b/drivers/gpu/drm/amd/amdgpu/atombios_dp.h
@@ -32,7 +32,7 @@ int amdgpu_atombios_dp_get_panel_mode(struct drm_encoder *encoder,
 void amdgpu_atombios_dp_set_link_config(struct drm_connector *connector,
 				 const struct drm_display_mode *mode);
 int amdgpu_atombios_dp_mode_valid_helper(struct drm_connector *connector,
-				  struct drm_display_mode *mode);
+				  const struct drm_display_mode *mode);
 bool amdgpu_atombios_dp_needs_link_train(struct amdgpu_connector *amdgpu_connector);
 void amdgpu_atombios_dp_set_rx_power_state(struct drm_connector *connector,
 				    u8 power_state);
diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c b/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c
index d95b2dc78063..a51f3414b65d 100644
--- a/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c
+++ b/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c
@@ -28,6 +28,7 @@
 
 #include <acpi/video.h>
 
+#include <drm/drm_edid.h>
 #include <drm/amdgpu_drm.h>
 #include "amdgpu.h"
 #include "amdgpu_connectors.h"
@@ -214,7 +215,7 @@ void amdgpu_atombios_encoder_init_backlight(struct amdgpu_encoder *amdgpu_encode
 	dig->bl_dev = bd;
 
 	bd->props.brightness = amdgpu_atombios_encoder_get_backlight_brightness(bd);
-	bd->props.power = FB_BLANK_UNBLANK;
+	bd->props.power = BACKLIGHT_POWER_ON;
 	backlight_update_status(bd);
 
 	DRM_INFO("amdgpu atom DIG backlight initialized\n");
@@ -228,7 +229,6 @@ error:
 register_acpi_backlight:
 	/* Try registering an ACPI video backlight device instead. */
 	acpi_video_register_backlight();
-	return;
 }
 
 void
@@ -335,7 +335,7 @@ amdgpu_atombios_encoder_setup_dac(struct drm_encoder *encoder, int action)
 	args.ucDacStandard = ATOM_DAC1_PS2;
 	args.usPixelClock = cpu_to_le16(amdgpu_encoder->pixel_clock / 10);
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 
 }
 
@@ -432,7 +432,7 @@ amdgpu_atombios_encoder_setup_dvo(struct drm_encoder *encoder, int action)
 		break;
 	}
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 int amdgpu_atombios_encoder_get_encoder_mode(struct drm_encoder *encoder)
@@ -732,7 +732,7 @@ amdgpu_atombios_encoder_setup_dig_encoder(struct drm_encoder *encoder,
 		break;
 	}
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 
 }
 
@@ -1136,7 +1136,7 @@ amdgpu_atombios_encoder_setup_dig_transmitter(struct drm_encoder *encoder, int a
 		break;
 	}
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 bool
@@ -1164,7 +1164,7 @@ amdgpu_atombios_encoder_set_edp_panel_power(struct drm_connector *connector,
 
 	args.v1.ucAction = action;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 
 	/* wait for the panel to power up */
 	if (action == ATOM_TRANSMITTER_ACTION_POWER_ON) {
@@ -1288,7 +1288,7 @@ amdgpu_atombios_encoder_setup_external_encoder(struct drm_encoder *encoder,
 		DRM_ERROR("Unknown table version: %d, %d\n", frev, crev);
 		return;
 	}
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 static void
@@ -1633,7 +1633,7 @@ amdgpu_atombios_encoder_set_crtc_source(struct drm_encoder *encoder)
 		return;
 	}
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 /* This only needs to be called once at startup */
@@ -1706,7 +1706,7 @@ amdgpu_atombios_encoder_dac_load_detect(struct drm_encoder *encoder,
 				args.sDacload.ucMisc = DAC_LOAD_MISC_YPrPb;
 		}
 
-		amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+		amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 
 		return true;
 	} else
@@ -2064,27 +2064,25 @@ amdgpu_atombios_encoder_get_lcd_info(struct amdgpu_encoder *encoder)
 				case LCD_FAKE_EDID_PATCH_RECORD_TYPE:
 					fake_edid_record = (ATOM_FAKE_EDID_PATCH_RECORD *)record;
 					if (fake_edid_record->ucFakeEDIDLength) {
-						struct edid *edid;
-						int edid_size =
-							max((int)EDID_LENGTH, (int)fake_edid_record->ucFakeEDIDLength);
-						edid = kmalloc(edid_size, GFP_KERNEL);
-						if (edid) {
-							memcpy((u8 *)edid, (u8 *)&fake_edid_record->ucFakeEDIDString[0],
-							       fake_edid_record->ucFakeEDIDLength);
-
-							if (drm_edid_is_valid(edid)) {
-								adev->mode_info.bios_hardcoded_edid = edid;
-								adev->mode_info.bios_hardcoded_edid_size = edid_size;
-							} else
-								kfree(edid);
-						}
+						const struct drm_edid *edid;
+						int edid_size;
+
+						if (fake_edid_record->ucFakeEDIDLength == 128)
+							edid_size = fake_edid_record->ucFakeEDIDLength;
+						else
+							edid_size = fake_edid_record->ucFakeEDIDLength * 128;
+						edid = drm_edid_alloc(fake_edid_record->ucFakeEDIDString, edid_size);
+						if (drm_edid_valid(edid))
+							adev->mode_info.bios_hardcoded_edid = edid;
+						else
+							drm_edid_free(edid);
+						record += struct_size(fake_edid_record,
+								      ucFakeEDIDString,
+								      edid_size);
+					} else {
+						/* empty fake edid record must be 3 bytes long */
+						record += sizeof(ATOM_FAKE_EDID_PATCH_RECORD) + 1;
 					}
-					record += fake_edid_record->ucFakeEDIDLength ?
-						  struct_size(fake_edid_record,
-							      ucFakeEDIDString,
-							      fake_edid_record->ucFakeEDIDLength) :
-						  /* empty fake edid record must be 3 bytes long */
-						  sizeof(ATOM_FAKE_EDID_PATCH_RECORD) + 1;
 					break;
 				case LCD_PANEL_RESOLUTION_RECORD_TYPE:
 					panel_res_record = (ATOM_PANEL_RESOLUTION_PATCH_RECORD *)record;
diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_i2c.c b/drivers/gpu/drm/amd/amdgpu/atombios_i2c.c
index af0335535f82..a6501114322f 100644
--- a/drivers/gpu/drm/amd/amdgpu/atombios_i2c.c
+++ b/drivers/gpu/drm/amd/amdgpu/atombios_i2c.c
@@ -86,7 +86,7 @@ static int amdgpu_atombios_i2c_process_i2c_ch(struct amdgpu_i2c_chan *chan,
 	args.ucSlaveAddr = slave_addr << 1;
 	args.ucLineNumber = chan->rec.i2c_id;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 
 	/* error */
 	if (args.ucStatus != HW_ASSISTED_I2C_STATUS_SUCCESS) {
@@ -172,5 +172,5 @@ void amdgpu_atombios_i2c_channel_trans(struct amdgpu_device *adev, u8 slave_addr
 	args.ucSlaveAddr = slave_addr;
 	args.ucLineNumber = line_number;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/cik.c b/drivers/gpu/drm/amd/amdgpu/cik.c
index de6d10390ab2..9cd63b4177bf 100644
--- a/drivers/gpu/drm/amd/amdgpu/cik.c
+++ b/drivers/gpu/drm/amd/amdgpu/cik.c
@@ -1141,12 +1141,12 @@ static uint32_t cik_get_register_value(struct amdgpu_device *adev,
 
 		mutex_lock(&adev->grbm_idx_mutex);
 		if (se_num != 0xffffffff || sh_num != 0xffffffff)
-			amdgpu_gfx_select_se_sh(adev, se_num, sh_num, 0xffffffff);
+			amdgpu_gfx_select_se_sh(adev, se_num, sh_num, 0xffffffff, 0);
 
 		val = RREG32(reg_offset);
 
 		if (se_num != 0xffffffff || sh_num != 0xffffffff)
-			amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+			amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 		mutex_unlock(&adev->grbm_idx_mutex);
 		return val;
 	} else {
@@ -1375,14 +1375,14 @@ static int cik_asic_pci_config_reset(struct amdgpu_device *adev)
 	return r;
 }
 
-static bool cik_asic_supports_baco(struct amdgpu_device *adev)
+static int cik_asic_supports_baco(struct amdgpu_device *adev)
 {
 	switch (adev->asic_type) {
 	case CHIP_BONAIRE:
 	case CHIP_HAWAII:
 		return amdgpu_dpm_is_baco_supported(adev);
 	default:
-		return false;
+		return 0;
 	}
 }
 
@@ -1574,17 +1574,8 @@ static void cik_pcie_gen3_enable(struct amdgpu_device *adev)
 			u16 bridge_cfg2, gpu_cfg2;
 			u32 max_lw, current_lw, tmp;
 
-			pcie_capability_read_word(root, PCI_EXP_LNKCTL,
-						  &bridge_cfg);
-			pcie_capability_read_word(adev->pdev, PCI_EXP_LNKCTL,
-						  &gpu_cfg);
-
-			tmp16 = bridge_cfg | PCI_EXP_LNKCTL_HAWD;
-			pcie_capability_write_word(root, PCI_EXP_LNKCTL, tmp16);
-
-			tmp16 = gpu_cfg | PCI_EXP_LNKCTL_HAWD;
-			pcie_capability_write_word(adev->pdev, PCI_EXP_LNKCTL,
-						   tmp16);
+			pcie_capability_set_word(root, PCI_EXP_LNKCTL, PCI_EXP_LNKCTL_HAWD);
+			pcie_capability_set_word(adev->pdev, PCI_EXP_LNKCTL, PCI_EXP_LNKCTL_HAWD);
 
 			tmp = RREG32_PCIE(ixPCIE_LC_STATUS1);
 			max_lw = (tmp & PCIE_LC_STATUS1__LC_DETECTED_LINK_WIDTH_MASK) >>
@@ -1637,45 +1628,28 @@ static void cik_pcie_gen3_enable(struct amdgpu_device *adev)
 				msleep(100);
 
 				/* linkctl */
-				pcie_capability_read_word(root, PCI_EXP_LNKCTL,
-							  &tmp16);
-				tmp16 &= ~PCI_EXP_LNKCTL_HAWD;
-				tmp16 |= (bridge_cfg & PCI_EXP_LNKCTL_HAWD);
-				pcie_capability_write_word(root, PCI_EXP_LNKCTL,
-							   tmp16);
-
-				pcie_capability_read_word(adev->pdev,
-							  PCI_EXP_LNKCTL,
-							  &tmp16);
-				tmp16 &= ~PCI_EXP_LNKCTL_HAWD;
-				tmp16 |= (gpu_cfg & PCI_EXP_LNKCTL_HAWD);
-				pcie_capability_write_word(adev->pdev,
-							   PCI_EXP_LNKCTL,
-							   tmp16);
+				pcie_capability_clear_and_set_word(root, PCI_EXP_LNKCTL,
+								   PCI_EXP_LNKCTL_HAWD,
+								   bridge_cfg &
+								   PCI_EXP_LNKCTL_HAWD);
+				pcie_capability_clear_and_set_word(adev->pdev, PCI_EXP_LNKCTL,
+								   PCI_EXP_LNKCTL_HAWD,
+								   gpu_cfg &
+								   PCI_EXP_LNKCTL_HAWD);
 
 				/* linkctl2 */
-				pcie_capability_read_word(root, PCI_EXP_LNKCTL2,
-							  &tmp16);
-				tmp16 &= ~(PCI_EXP_LNKCTL2_ENTER_COMP |
-					   PCI_EXP_LNKCTL2_TX_MARGIN);
-				tmp16 |= (bridge_cfg2 &
-					  (PCI_EXP_LNKCTL2_ENTER_COMP |
-					   PCI_EXP_LNKCTL2_TX_MARGIN));
-				pcie_capability_write_word(root,
-							   PCI_EXP_LNKCTL2,
-							   tmp16);
-
-				pcie_capability_read_word(adev->pdev,
-							  PCI_EXP_LNKCTL2,
-							  &tmp16);
-				tmp16 &= ~(PCI_EXP_LNKCTL2_ENTER_COMP |
-					   PCI_EXP_LNKCTL2_TX_MARGIN);
-				tmp16 |= (gpu_cfg2 &
-					  (PCI_EXP_LNKCTL2_ENTER_COMP |
-					   PCI_EXP_LNKCTL2_TX_MARGIN));
-				pcie_capability_write_word(adev->pdev,
-							   PCI_EXP_LNKCTL2,
-							   tmp16);
+				pcie_capability_clear_and_set_word(root, PCI_EXP_LNKCTL2,
+								   PCI_EXP_LNKCTL2_ENTER_COMP |
+								   PCI_EXP_LNKCTL2_TX_MARGIN,
+								   bridge_cfg2 &
+								   (PCI_EXP_LNKCTL2_ENTER_COMP |
+								    PCI_EXP_LNKCTL2_TX_MARGIN));
+				pcie_capability_clear_and_set_word(adev->pdev, PCI_EXP_LNKCTL2,
+								   PCI_EXP_LNKCTL2_ENTER_COMP |
+								   PCI_EXP_LNKCTL2_TX_MARGIN,
+								   gpu_cfg2 &
+								   (PCI_EXP_LNKCTL2_ENTER_COMP |
+								    PCI_EXP_LNKCTL2_TX_MARGIN));
 
 				tmp = RREG32_PCIE(ixPCIE_LC_CNTL4);
 				tmp &= ~PCIE_LC_CNTL4__LC_SET_QUIESCE_MASK;
@@ -1690,16 +1664,15 @@ static void cik_pcie_gen3_enable(struct amdgpu_device *adev)
 	speed_cntl &= ~PCIE_LC_SPEED_CNTL__LC_FORCE_DIS_SW_SPEED_CHANGE_MASK;
 	WREG32_PCIE(ixPCIE_LC_SPEED_CNTL, speed_cntl);
 
-	pcie_capability_read_word(adev->pdev, PCI_EXP_LNKCTL2, &tmp16);
-	tmp16 &= ~PCI_EXP_LNKCTL2_TLS;
-
+	tmp16 = 0;
 	if (adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_GEN3)
 		tmp16 |= PCI_EXP_LNKCTL2_TLS_8_0GT; /* gen3 */
 	else if (adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_GEN2)
 		tmp16 |= PCI_EXP_LNKCTL2_TLS_5_0GT; /* gen2 */
 	else
 		tmp16 |= PCI_EXP_LNKCTL2_TLS_2_5GT; /* gen1 */
-	pcie_capability_write_word(adev->pdev, PCI_EXP_LNKCTL2, tmp16);
+	pcie_capability_clear_and_set_word(adev->pdev, PCI_EXP_LNKCTL2,
+					   PCI_EXP_LNKCTL2_TLS, tmp16);
 
 	speed_cntl = RREG32_PCIE(ixPCIE_LC_SPEED_CNTL);
 	speed_cntl |= PCIE_LC_SPEED_CNTL__LC_INITIATE_LINK_SPEED_CHANGE_MASK;
@@ -1725,10 +1698,6 @@ static void cik_program_aspm(struct amdgpu_device *adev)
 	if (pci_is_root_bus(adev->pdev->bus))
 		return;
 
-	/* XXX double check APUs */
-	if (adev->flags & AMD_IS_APU)
-		return;
-
 	orig = data = RREG32_PCIE(ixPCIE_LC_N_FTS_CNTL);
 	data &= ~PCIE_LC_N_FTS_CNTL__LC_XMIT_N_FTS_MASK;
 	data |= (0x24 << PCIE_LC_N_FTS_CNTL__LC_XMIT_N_FTS__SHIFT) |
@@ -2016,9 +1985,9 @@ static const struct amdgpu_asic_funcs cik_asic_funcs =
 	.query_video_codecs = &cik_query_video_codecs,
 };
 
-static int cik_common_early_init(void *handle)
+static int cik_common_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->smc_rreg = &cik_smc_rreg;
 	adev->smc_wreg = &cik_smc_wreg;
@@ -2155,19 +2124,9 @@ static int cik_common_early_init(void *handle)
 	return 0;
 }
 
-static int cik_common_sw_init(void *handle)
-{
-	return 0;
-}
-
-static int cik_common_sw_fini(void *handle)
-{
-	return 0;
-}
-
-static int cik_common_hw_init(void *handle)
+static int cik_common_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* move the golden regs per IP block */
 	cik_init_golden_registers(adev);
@@ -2179,48 +2138,36 @@ static int cik_common_hw_init(void *handle)
 	return 0;
 }
 
-static int cik_common_hw_fini(void *handle)
+static int cik_common_hw_fini(struct amdgpu_ip_block *ip_block)
 {
 	return 0;
 }
 
-static int cik_common_suspend(void *handle)
+static int cik_common_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return cik_common_hw_fini(adev);
-}
-
-static int cik_common_resume(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return cik_common_hw_init(adev);
+	return cik_common_hw_init(ip_block);
 }
 
-static bool cik_common_is_idle(void *handle)
+static bool cik_common_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	return true;
 }
 
-static int cik_common_wait_for_idle(void *handle)
-{
-	return 0;
-}
 
-static int cik_common_soft_reset(void *handle)
+
+static int cik_common_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	/* XXX hard reset?? */
 	return 0;
 }
 
-static int cik_common_set_clockgating_state(void *handle,
+static int cik_common_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					    enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int cik_common_set_powergating_state(void *handle,
+static int cik_common_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					    enum amd_powergating_state state)
 {
 	return 0;
@@ -2229,15 +2176,10 @@ static int cik_common_set_powergating_state(void *handle,
 static const struct amd_ip_funcs cik_common_ip_funcs = {
 	.name = "cik_common",
 	.early_init = cik_common_early_init,
-	.late_init = NULL,
-	.sw_init = cik_common_sw_init,
-	.sw_fini = cik_common_sw_fini,
 	.hw_init = cik_common_hw_init,
 	.hw_fini = cik_common_hw_fini,
-	.suspend = cik_common_suspend,
 	.resume = cik_common_resume,
 	.is_idle = cik_common_is_idle,
-	.wait_for_idle = cik_common_wait_for_idle,
 	.soft_reset = cik_common_soft_reset,
 	.set_clockgating_state = cik_common_set_clockgating_state,
 	.set_powergating_state = cik_common_set_powergating_state,
diff --git a/drivers/gpu/drm/amd/amdgpu/cik_ih.c b/drivers/gpu/drm/amd/amdgpu/cik_ih.c
index df385ffc9768..876a3256dba4 100644
--- a/drivers/gpu/drm/amd/amdgpu/cik_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/cik_ih.c
@@ -156,6 +156,9 @@ static int cik_ih_irq_init(struct amdgpu_device *adev)
 	/* enable irqs */
 	cik_ih_enable_interrupts(adev);
 
+	if (adev->irq.ih_soft.ring_size)
+		adev->irq.ih_soft.enabled = true;
+
 	return 0;
 }
 
@@ -192,6 +195,9 @@ static u32 cik_ih_get_wptr(struct amdgpu_device *adev,
 
 	wptr = le32_to_cpu(*ih->wptr_cpu);
 
+	if (ih == &adev->irq.ih_soft)
+		goto out;
+
 	if (wptr & IH_RB_WPTR__RB_OVERFLOW_MASK) {
 		wptr &= ~IH_RB_WPTR__RB_OVERFLOW_MASK;
 		/* When a ring buffer overflow happen start parsing interrupt
@@ -204,7 +210,15 @@ static u32 cik_ih_get_wptr(struct amdgpu_device *adev,
 		tmp = RREG32(mmIH_RB_CNTL);
 		tmp |= IH_RB_CNTL__WPTR_OVERFLOW_CLEAR_MASK;
 		WREG32(mmIH_RB_CNTL, tmp);
+
+		/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+		 * can be detected.
+		 */
+		tmp &= ~IH_RB_CNTL__WPTR_OVERFLOW_CLEAR_MASK;
+		WREG32(mmIH_RB_CNTL, tmp);
 	}
+
+out:
 	return (wptr & ih->ptr_mask);
 }
 
@@ -277,9 +291,9 @@ static void cik_ih_set_rptr(struct amdgpu_device *adev,
 	WREG32(mmIH_RB_RPTR, ih->rptr);
 }
 
-static int cik_ih_early_init(void *handle)
+static int cik_ih_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	ret = amdgpu_irq_add_domain(adev);
@@ -291,23 +305,27 @@ static int cik_ih_early_init(void *handle)
 	return 0;
 }
 
-static int cik_ih_sw_init(void *handle)
+static int cik_ih_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 64 * 1024, false);
 	if (r)
 		return r;
 
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, IH_SW_RING_SIZE, true);
+	if (r)
+		return r;
+
 	r = amdgpu_irq_init(adev);
 
 	return r;
 }
 
-static int cik_ih_sw_fini(void *handle)
+static int cik_ih_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_fini_sw(adev);
 	amdgpu_irq_remove_domain(adev);
@@ -315,39 +333,33 @@ static int cik_ih_sw_fini(void *handle)
 	return 0;
 }
 
-static int cik_ih_hw_init(void *handle)
+static int cik_ih_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return cik_ih_irq_init(adev);
 }
 
-static int cik_ih_hw_fini(void *handle)
+static int cik_ih_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	cik_ih_irq_disable(adev);
+	cik_ih_irq_disable(ip_block->adev);
 
 	return 0;
 }
 
-static int cik_ih_suspend(void *handle)
+static int cik_ih_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return cik_ih_hw_fini(adev);
+	return cik_ih_hw_fini(ip_block);
 }
 
-static int cik_ih_resume(void *handle)
+static int cik_ih_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return cik_ih_hw_init(adev);
+	return cik_ih_hw_init(ip_block);
 }
 
-static bool cik_ih_is_idle(void *handle)
+static bool cik_ih_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (tmp & SRBM_STATUS__IH_BUSY_MASK)
@@ -356,11 +368,11 @@ static bool cik_ih_is_idle(void *handle)
 	return true;
 }
 
-static int cik_ih_wait_for_idle(void *handle)
+static int cik_ih_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		/* read MC_STATUS */
@@ -372,9 +384,9 @@ static int cik_ih_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int cik_ih_soft_reset(void *handle)
+static int cik_ih_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	u32 srbm_soft_reset = 0;
 	u32 tmp = RREG32(mmSRBM_STATUS);
@@ -402,13 +414,13 @@ static int cik_ih_soft_reset(void *handle)
 	return 0;
 }
 
-static int cik_ih_set_clockgating_state(void *handle,
+static int cik_ih_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int cik_ih_set_powergating_state(void *handle,
+static int cik_ih_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -417,7 +429,6 @@ static int cik_ih_set_powergating_state(void *handle,
 static const struct amd_ip_funcs cik_ih_ip_funcs = {
 	.name = "cik_ih",
 	.early_init = cik_ih_early_init,
-	.late_init = NULL,
 	.sw_init = cik_ih_sw_init,
 	.sw_fini = cik_ih_sw_fini,
 	.hw_init = cik_ih_hw_init,
@@ -442,8 +453,7 @@ static void cik_ih_set_interrupt_funcs(struct amdgpu_device *adev)
 	adev->irq.ih_funcs = &cik_ih_funcs;
 }
 
-const struct amdgpu_ip_block_version cik_ih_ip_block =
-{
+const struct amdgpu_ip_block_version cik_ih_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_IH,
 	.major = 2,
 	.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
index 67d16236b216..9e8715b4739d 100644
--- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
+++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
@@ -54,7 +54,9 @@ static void cik_sdma_set_ring_funcs(struct amdgpu_device *adev);
 static void cik_sdma_set_irq_funcs(struct amdgpu_device *adev);
 static void cik_sdma_set_buffer_funcs(struct amdgpu_device *adev);
 static void cik_sdma_set_vm_pte_funcs(struct amdgpu_device *adev);
-static int cik_sdma_soft_reset(void *handle);
+static int cik_sdma_soft_reset(struct amdgpu_ip_block *ip_block);
+
+u32 amdgpu_cik_gpu_check_soft_reset(struct amdgpu_device *adev);
 
 MODULE_FIRMWARE("amdgpu/bonaire_sdma.bin");
 MODULE_FIRMWARE("amdgpu/bonaire_sdma1.bin");
@@ -67,9 +69,6 @@ MODULE_FIRMWARE("amdgpu/kabini_sdma1.bin");
 MODULE_FIRMWARE("amdgpu/mullins_sdma.bin");
 MODULE_FIRMWARE("amdgpu/mullins_sdma1.bin");
 
-u32 amdgpu_cik_gpu_check_soft_reset(struct amdgpu_device *adev);
-
-
 static void cik_sdma_free_microcode(struct amdgpu_device *adev)
 {
 	int i;
@@ -107,7 +106,6 @@ static void cik_sdma_free_microcode(struct amdgpu_device *adev)
 static int cik_sdma_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[30];
 	int err = 0, i;
 
 	DRM_DEBUG("\n");
@@ -133,16 +131,20 @@ static int cik_sdma_init_microcode(struct amdgpu_device *adev)
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		if (i == 0)
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sdma.bin", chip_name);
+			err = amdgpu_ucode_request(adev, &adev->sdma.instance[i].fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_sdma.bin", chip_name);
 		else
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sdma1.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->sdma.instance[i].fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->sdma.instance[i].fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_sdma1.bin", chip_name);
 		if (err)
 			goto out;
 	}
 out:
 	if (err) {
-		pr_err("cik_sdma: Failed to load firmware \"%s\"\n", fw_name);
+		pr_err("cik_sdma: Failed to load firmware \"%s_sdma%s.bin\"\n",
+		       chip_name, i == 0 ? "" : "1");
 		for (i = 0; i < adev->sdma.num_instances; i++)
 			amdgpu_ucode_release(&adev->sdma.instance[i].fw);
 	}
@@ -308,8 +310,6 @@ static void cik_sdma_gfx_stop(struct amdgpu_device *adev)
 	u32 rb_cntl;
 	int i;
 
-	amdgpu_sdma_unset_buffer_funcs_helper(adev);
-
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		rb_cntl = RREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i]);
 		rb_cntl &= ~SDMA0_GFX_RB_CNTL__RB_ENABLE_MASK;
@@ -489,8 +489,6 @@ static int cik_sdma_gfx_resume(struct amdgpu_device *adev)
 #endif
 		/* enable DMA IBs */
 		WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl);
-
-		ring->sched.ready = true;
 	}
 
 	cik_sdma_enable(adev, true);
@@ -500,9 +498,6 @@ static int cik_sdma_gfx_resume(struct amdgpu_device *adev)
 		r = amdgpu_ring_test_helper(ring);
 		if (r)
 			return r;
-
-		if (adev->mman.buffer_funcs_ring == ring)
-			amdgpu_ttm_set_buffer_funcs_status(adev, true);
 	}
 
 	return 0;
@@ -702,7 +697,7 @@ static int cik_sdma_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = -EINVAL;
 
 err1:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err0:
 	amdgpu_device_wb_free(adev, index);
@@ -924,12 +919,17 @@ static void cik_enable_sdma_mgls(struct amdgpu_device *adev,
 	}
 }
 
-static int cik_sdma_early_init(void *handle)
+static int cik_sdma_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
 
 	adev->sdma.num_instances = SDMA_MAX_INSTANCE;
 
+	r = cik_sdma_init_microcode(adev);
+	if (r)
+		return r;
+
 	cik_sdma_set_ring_funcs(adev);
 	cik_sdma_set_irq_funcs(adev);
 	cik_sdma_set_buffer_funcs(adev);
@@ -938,18 +938,12 @@ static int cik_sdma_early_init(void *handle)
 	return 0;
 }
 
-static int cik_sdma_sw_init(void *handle)
+static int cik_sdma_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r, i;
 
-	r = cik_sdma_init_microcode(adev);
-	if (r) {
-		DRM_ERROR("Failed to load sdma firmware!\n");
-		return r;
-	}
-
 	/* SDMA trap event */
 	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 224,
 			      &adev->sdma.trap_irq);
@@ -984,9 +978,9 @@ static int cik_sdma_sw_init(void *handle)
 	return r;
 }
 
-static int cik_sdma_sw_fini(void *handle)
+static int cik_sdma_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++)
@@ -996,21 +990,16 @@ static int cik_sdma_sw_fini(void *handle)
 	return 0;
 }
 
-static int cik_sdma_hw_init(void *handle)
+static int cik_sdma_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	r = cik_sdma_start(adev);
-	if (r)
-		return r;
-
-	return r;
+	return cik_sdma_start(adev);
 }
 
-static int cik_sdma_hw_fini(void *handle)
+static int cik_sdma_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	cik_ctx_switch_enable(adev, false);
 	cik_sdma_enable(adev, false);
@@ -1018,25 +1007,21 @@ static int cik_sdma_hw_fini(void *handle)
 	return 0;
 }
 
-static int cik_sdma_suspend(void *handle)
+static int cik_sdma_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return cik_sdma_hw_fini(adev);
+	return cik_sdma_hw_fini(ip_block);
 }
 
-static int cik_sdma_resume(void *handle)
+static int cik_sdma_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	cik_sdma_soft_reset(ip_block);
 
-	cik_sdma_soft_reset(handle);
-
-	return cik_sdma_hw_init(adev);
+	return cik_sdma_hw_init(ip_block);
 }
 
-static bool cik_sdma_is_idle(void *handle)
+static bool cik_sdma_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS2);
 
 	if (tmp & (SRBM_STATUS2__SDMA_BUSY_MASK |
@@ -1046,27 +1031,23 @@ static bool cik_sdma_is_idle(void *handle)
 	return true;
 }
 
-static int cik_sdma_wait_for_idle(void *handle)
+static int cik_sdma_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
-	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(mmSRBM_STATUS2) & (SRBM_STATUS2__SDMA_BUSY_MASK |
-				SRBM_STATUS2__SDMA1_BUSY_MASK);
-
-		if (!tmp)
+		if (cik_sdma_is_idle(ip_block))
 			return 0;
 		udelay(1);
 	}
 	return -ETIMEDOUT;
 }
 
-static int cik_sdma_soft_reset(void *handle)
+static int cik_sdma_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 srbm_soft_reset = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp;
 
 	/* sdma0 */
@@ -1200,11 +1181,11 @@ static int cik_sdma_process_illegal_inst_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int cik_sdma_set_clockgating_state(void *handle,
+static int cik_sdma_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	bool gate = false;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (state == AMD_CG_STATE_GATE)
 		gate = true;
@@ -1215,7 +1196,7 @@ static int cik_sdma_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int cik_sdma_set_powergating_state(void *handle,
+static int cik_sdma_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -1224,7 +1205,6 @@ static int cik_sdma_set_powergating_state(void *handle,
 static const struct amd_ip_funcs cik_sdma_ip_funcs = {
 	.name = "cik_sdma",
 	.early_init = cik_sdma_early_init,
-	.late_init = NULL,
 	.sw_init = cik_sdma_sw_init,
 	.sw_fini = cik_sdma_sw_fini,
 	.hw_init = cik_sdma_hw_init,
@@ -1298,7 +1278,7 @@ static void cik_sdma_set_irq_funcs(struct amdgpu_device *adev)
  * @src_offset: src GPU address
  * @dst_offset: dst GPU address
  * @byte_count: number of bytes to xfer
- * @tmz: is this a secure operation
+ * @copy_flags: unused
  *
  * Copy GPU buffers using the DMA engine (CIK).
  * Used by the amdgpu ttm implementation to move pages if
@@ -1308,7 +1288,7 @@ static void cik_sdma_emit_copy_buffer(struct amdgpu_ib *ib,
 				      uint64_t src_offset,
 				      uint64_t dst_offset,
 				      uint32_t byte_count,
-				      bool tmz)
+				      uint32_t copy_flags)
 {
 	ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_COPY, SDMA_COPY_SUB_OPCODE_LINEAR, 0);
 	ib->ptr[ib->length_dw++] = byte_count;
diff --git a/drivers/gpu/drm/amd/amdgpu/cikd.h b/drivers/gpu/drm/amd/amdgpu/cikd.h
index 55982c0064b5..8aca4f2734f2 100644
--- a/drivers/gpu/drm/amd/amdgpu/cikd.h
+++ b/drivers/gpu/drm/amd/amdgpu/cikd.h
@@ -51,8 +51,14 @@
 #define HPD4_REGISTER_OFFSET                 (0x1813 - 0x1807)
 #define HPD5_REGISTER_OFFSET                 (0x1816 - 0x1807)
 
-#define BONAIRE_GB_ADDR_CONFIG_GOLDEN        0x12010001
-#define HAWAII_GB_ADDR_CONFIG_GOLDEN         0x12011003
+/* audio endpt instance offsets */
+#define AUD0_REGISTER_OFFSET                 (0x1780 - 0x1780)
+#define AUD1_REGISTER_OFFSET                 (0x1786 - 0x1780)
+#define AUD2_REGISTER_OFFSET                 (0x178c - 0x1780)
+#define AUD3_REGISTER_OFFSET                 (0x1792 - 0x1780)
+#define AUD4_REGISTER_OFFSET                 (0x1798 - 0x1780)
+#define AUD5_REGISTER_OFFSET                 (0x179d - 0x1780)
+#define AUD6_REGISTER_OFFSET                 (0x17a4 - 0x1780)
 
 #define		PIPEID(x)					((x) << 0)
 #define		MEID(x)						((x) << 2)
@@ -364,6 +370,7 @@
 		 * 1 - Stream
 		 * 2 - Bypass
 		 */
+#define		EOP_EXEC				(1 << 28) /* For Trailing Fence */
 #define		DATA_SEL(x)                             ((x) << 29)
 		/* 0 - discard
 		 * 1 - send low 32bit data
diff --git a/drivers/gpu/drm/amd/amdgpu/clearstate_gfx12.h b/drivers/gpu/drm/amd/amdgpu/clearstate_gfx12.h
new file mode 100644
index 000000000000..2f6c9d11d5ae
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/clearstate_gfx12.h
@@ -0,0 +1,121 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __CLEARSTATE_GFX12_H_
+#define __CLEARSTATE_GFX12_H_
+
+static const unsigned int gfx12_SECT_CONTEXT_def_1[] = {
+0x00000000, //mmSC_MEM_TEMPORAL
+0x00000000, //mmSC_MEM_SPEC_READ
+0x00000000, //mmPA_SC_VPORT_0_TL
+0x00000000, //mmPA_SC_VPORT_0_BR
+0x00000000, //mmPA_SC_VPORT_1_TL
+0x00000000, //mmPA_SC_VPORT_1_BR
+0x00000000, //mmPA_SC_VPORT_2_TL
+0x00000000, //mmPA_SC_VPORT_2_BR
+0x00000000, //mmPA_SC_VPORT_3_TL
+0x00000000, //mmPA_SC_VPORT_3_BR
+0x00000000, //mmPA_SC_VPORT_4_TL
+0x00000000, //mmPA_SC_VPORT_4_BR
+0x00000000, //mmPA_SC_VPORT_5_TL
+0x00000000, //mmPA_SC_VPORT_5_BR
+0x00000000, //mmPA_SC_VPORT_6_TL
+0x00000000, //mmPA_SC_VPORT_6_BR
+0x00000000, //mmPA_SC_VPORT_7_TL
+0x00000000, //mmPA_SC_VPORT_7_BR
+0x00000000, //mmPA_SC_VPORT_8_TL
+0x00000000, //mmPA_SC_VPORT_8_BR
+0x00000000, //mmPA_SC_VPORT_9_TL
+0x00000000, //mmPA_SC_VPORT_9_BR
+0x00000000, //mmPA_SC_VPORT_10_TL
+0x00000000, //mmPA_SC_VPORT_10_BR
+0x00000000, //mmPA_SC_VPORT_11_TL
+0x00000000, //mmPA_SC_VPORT_11_BR
+0x00000000, //mmPA_SC_VPORT_12_TL
+0x00000000, //mmPA_SC_VPORT_12_BR
+0x00000000, //mmPA_SC_VPORT_13_TL
+0x00000000, //mmPA_SC_VPORT_13_BR
+0x00000000, //mmPA_SC_VPORT_14_TL
+0x00000000, //mmPA_SC_VPORT_14_BR
+0x00000000, //mmPA_SC_VPORT_15_TL
+0x00000000, //mmPA_SC_VPORT_15_BR
+};
+
+static const unsigned int gfx12_SECT_CONTEXT_def_2[] = {
+0x00000000, //mmPA_CL_PROG_NEAR_CLIP_Z
+0x00000000, //mmPA_RATE_CNTL
+};
+
+static const unsigned int gfx12_SECT_CONTEXT_def_3[] = {
+0x00000000, //mmCP_PERFMON_CNTX_CNTL
+};
+
+static const unsigned int gfx12_SECT_CONTEXT_def_4[] = {
+0x00000000, //mmCONTEXT_RESERVED_REG0
+0x00000000, //mmCONTEXT_RESERVED_REG1
+0x00000000, //mmPA_SC_CLIPRECT_0_EXT
+0x00000000, //mmPA_SC_CLIPRECT_1_EXT
+0x00000000, //mmPA_SC_CLIPRECT_2_EXT
+0x00000000, //mmPA_SC_CLIPRECT_3_EXT
+};
+
+static const unsigned int gfx12_SECT_CONTEXT_def_5[] = {
+0x00000000, //mmPA_SC_HIZ_INFO
+0x00000000, //mmPA_SC_HIS_INFO
+0x00000000, //mmPA_SC_HIZ_BASE
+0x00000000, //mmPA_SC_HIZ_BASE_EXT
+0x00000000, //mmPA_SC_HIZ_SIZE_XY
+0x00000000, //mmPA_SC_HIS_BASE
+0x00000000, //mmPA_SC_HIS_BASE_EXT
+0x00000000, //mmPA_SC_HIS_SIZE_XY
+0x00000000, //mmPA_SC_BINNER_OUTPUT_TIMEOUT_CNTL
+0x00000000, //mmPA_SC_BINNER_DYNAMIC_BATCH_LIMIT
+0x00000000, //mmPA_SC_HISZ_CONTROL
+};
+
+static const unsigned int gfx12_SECT_CONTEXT_def_6[] = {
+0x00000000, //mmCB_MEM0_INFO
+0x00000000, //mmCB_MEM1_INFO
+0x00000000, //mmCB_MEM2_INFO
+0x00000000, //mmCB_MEM3_INFO
+0x00000000, //mmCB_MEM4_INFO
+0x00000000, //mmCB_MEM5_INFO
+0x00000000, //mmCB_MEM6_INFO
+0x00000000, //mmCB_MEM7_INFO
+};
+
+static const struct cs_extent_def gfx12_SECT_CONTEXT_defs[] = {
+    {gfx12_SECT_CONTEXT_def_1, 0x0000a03e, 34 },
+    {gfx12_SECT_CONTEXT_def_2, 0x0000a0cc, 2 },
+    {gfx12_SECT_CONTEXT_def_3, 0x0000a0d8, 1 },
+    {gfx12_SECT_CONTEXT_def_4, 0x0000a0db, 6 },
+    {gfx12_SECT_CONTEXT_def_5, 0x0000a2e5, 11 },
+    {gfx12_SECT_CONTEXT_def_6, 0x0000a3c0, 8 },
+    { 0, 0, 0 }
+};
+
+static const struct cs_section_def gfx12_cs_data[] = {
+    { gfx12_SECT_CONTEXT_defs, SECT_CONTEXT },
+    { 0, SECT_NONE }
+};
+
+#endif /* __CLEARSTATE_GFX12_H_ */
diff --git a/drivers/gpu/drm/amd/amdgpu/clearstate_gfx9.h b/drivers/gpu/drm/amd/amdgpu/clearstate_gfx9.h
index 567a904804bc..9c85ca6358c1 100644
--- a/drivers/gpu/drm/amd/amdgpu/clearstate_gfx9.h
+++ b/drivers/gpu/drm/amd/amdgpu/clearstate_gfx9.h
@@ -21,8 +21,7 @@
  *
  */
 
-static const unsigned int gfx9_SECT_CONTEXT_def_1[] =
-{
+static const unsigned int gfx9_SECT_CONTEXT_def_1[] = {
     0x00000000, // DB_RENDER_CONTROL
     0x00000000, // DB_COUNT_CONTROL
     0x00000000, // DB_DEPTH_VIEW
@@ -236,8 +235,7 @@ static const unsigned int gfx9_SECT_CONTEXT_def_1[] =
     0x00000000, // PA_SC_VPORT_ZMIN_15
     0x3f800000, // PA_SC_VPORT_ZMAX_15
 };
-static const unsigned int gfx9_SECT_CONTEXT_def_2[] =
-{
+static const unsigned int gfx9_SECT_CONTEXT_def_2[] = {
     0x00000000, // PA_SC_SCREEN_EXTENT_CONTROL
     0x00000000, // PA_SC_TILE_STEERING_OVERRIDE
     0x00000000, // CP_PERFMON_CNTX_CNTL
@@ -521,15 +519,13 @@ static const unsigned int gfx9_SECT_CONTEXT_def_2[] =
     0x00000000, // CB_MRT6_EPITCH
     0x00000000, // CB_MRT7_EPITCH
 };
-static const unsigned int gfx9_SECT_CONTEXT_def_3[] =
-{
+static const unsigned int gfx9_SECT_CONTEXT_def_3[] = {
     0x00000000, // PA_CL_POINT_X_RAD
     0x00000000, // PA_CL_POINT_Y_RAD
     0x00000000, // PA_CL_POINT_SIZE
     0x00000000, // PA_CL_POINT_CULL_RAD
 };
-static const unsigned int gfx9_SECT_CONTEXT_def_4[] =
-{
+static const unsigned int gfx9_SECT_CONTEXT_def_4[] = {
     0x00000000, // DB_DEPTH_CONTROL
     0x00000000, // DB_EQAA
     0x00000000, // CB_COLOR_CONTROL
@@ -688,17 +684,14 @@ static const unsigned int gfx9_SECT_CONTEXT_def_4[] =
     0x00000000, // VGT_GS_OUT_PRIM_TYPE
     0x00000000, // IA_ENHANCE
 };
-static const unsigned int gfx9_SECT_CONTEXT_def_5[] =
-{
+static const unsigned int gfx9_SECT_CONTEXT_def_5[] = {
     0x00000000, // WD_ENHANCE
     0x00000000, // VGT_PRIMITIVEID_EN
 };
-static const unsigned int gfx9_SECT_CONTEXT_def_6[] =
-{
+static const unsigned int gfx9_SECT_CONTEXT_def_6[] = {
     0x00000000, // VGT_PRIMITIVEID_RESET
 };
-static const unsigned int gfx9_SECT_CONTEXT_def_7[] =
-{
+static const unsigned int gfx9_SECT_CONTEXT_def_7[] = {
     0x00000000, // VGT_GS_MAX_PRIMS_PER_SUBGROUP
     0x00000000, // VGT_DRAW_PAYLOAD_CNTL
     0, // HOLE
@@ -766,8 +759,7 @@ static const unsigned int gfx9_SECT_CONTEXT_def_7[] =
     0x00000000, // VGT_STRMOUT_CONFIG
     0x00000000, // VGT_STRMOUT_BUFFER_CONFIG
 };
-static const unsigned int gfx9_SECT_CONTEXT_def_8[] =
-{
+static const unsigned int gfx9_SECT_CONTEXT_def_8[] = {
     0x00000000, // PA_SC_CENTROID_PRIORITY_0
     0x00000000, // PA_SC_CENTROID_PRIORITY_1
     0x00001000, // PA_SC_LINE_CNTL
@@ -924,8 +916,7 @@ static const unsigned int gfx9_SECT_CONTEXT_def_8[] =
     0x00000000, // CB_COLOR7_DCC_BASE
     0x00000000, // CB_COLOR7_DCC_BASE_EXT
 };
-static const struct cs_extent_def gfx9_SECT_CONTEXT_defs[] =
-{
+static const struct cs_extent_def gfx9_SECT_CONTEXT_defs[] = {
     {gfx9_SECT_CONTEXT_def_1, 0x0000a000, 212 },
     {gfx9_SECT_CONTEXT_def_2, 0x0000a0d6, 282 },
     {gfx9_SECT_CONTEXT_def_3, 0x0000a1f5, 4 },
diff --git a/drivers/gpu/drm/amd/amdgpu/clearstate_si.h b/drivers/gpu/drm/amd/amdgpu/clearstate_si.h
index 66e39cdb5cb0..5fd96ddd7f0f 100644
--- a/drivers/gpu/drm/amd/amdgpu/clearstate_si.h
+++ b/drivers/gpu/drm/amd/amdgpu/clearstate_si.h
@@ -21,8 +21,7 @@
  *
  */
 
-static const u32 si_SECT_CONTEXT_def_1[] =
-{
+static const u32 si_SECT_CONTEXT_def_1[] = {
     0x00000000, // DB_RENDER_CONTROL
     0x00000000, // DB_COUNT_CONTROL
     0x00000000, // DB_DEPTH_VIEW
@@ -236,8 +235,7 @@ static const u32 si_SECT_CONTEXT_def_1[] =
     0x00000000, // PA_SC_VPORT_ZMIN_15
     0x3f800000, // PA_SC_VPORT_ZMAX_15
 };
-static const u32 si_SECT_CONTEXT_def_2[] =
-{
+static const u32 si_SECT_CONTEXT_def_2[] = {
     0x00000000, // CP_PERFMON_CNTX_CNTL
     0x00000000, // CP_RINGID
     0x00000000, // CP_VMID
@@ -511,8 +509,7 @@ static const u32 si_SECT_CONTEXT_def_2[] =
     0x00000000, // CB_BLEND6_CONTROL
     0x00000000, // CB_BLEND7_CONTROL
 };
-static const u32 si_SECT_CONTEXT_def_3[] =
-{
+static const u32 si_SECT_CONTEXT_def_3[] = {
     0x00000000, // PA_CL_POINT_X_RAD
     0x00000000, // PA_CL_POINT_Y_RAD
     0x00000000, // PA_CL_POINT_SIZE
@@ -520,8 +517,7 @@ static const u32 si_SECT_CONTEXT_def_3[] =
     0x00000000, // VGT_DMA_BASE_HI
     0x00000000, // VGT_DMA_BASE
 };
-static const u32 si_SECT_CONTEXT_def_4[] =
-{
+static const u32 si_SECT_CONTEXT_def_4[] = {
     0x00000000, // DB_DEPTH_CONTROL
     0x00000000, // DB_EQAA
     0x00000000, // CB_COLOR_CONTROL
@@ -680,16 +676,13 @@ static const u32 si_SECT_CONTEXT_def_4[] =
     0x00000000, // VGT_GS_OUT_PRIM_TYPE
     0x00000000, // IA_ENHANCE
 };
-static const u32 si_SECT_CONTEXT_def_5[] =
-{
+static const u32 si_SECT_CONTEXT_def_5[] = {
     0x00000000, // VGT_PRIMITIVEID_EN
 };
-static const u32 si_SECT_CONTEXT_def_6[] =
-{
+static const u32 si_SECT_CONTEXT_def_6[] = {
     0x00000000, // VGT_PRIMITIVEID_RESET
 };
-static const u32 si_SECT_CONTEXT_def_7[] =
-{
+static const u32 si_SECT_CONTEXT_def_7[] = {
     0x00000000, // VGT_MULTI_PRIM_IB_RESET_EN
     0, // HOLE
     0, // HOLE
@@ -924,8 +917,7 @@ static const u32 si_SECT_CONTEXT_def_7[] =
     0x00000000, // CB_COLOR7_CLEAR_WORD0
     0x00000000, // CB_COLOR7_CLEAR_WORD1
 };
-static const struct cs_extent_def si_SECT_CONTEXT_defs[] =
-{
+static const struct cs_extent_def si_SECT_CONTEXT_defs[] = {
     {si_SECT_CONTEXT_def_1, 0x0000a000, 212 },
     {si_SECT_CONTEXT_def_2, 0x0000a0d8, 272 },
     {si_SECT_CONTEXT_def_3, 0x0000a1f5, 6 },
diff --git a/drivers/gpu/drm/amd/amdgpu/cyan_skillfish_reg_init.c b/drivers/gpu/drm/amd/amdgpu/cyan_skillfish_reg_init.c
new file mode 100644
index 000000000000..ed1e25661706
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/cyan_skillfish_reg_init.c
@@ -0,0 +1,56 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2018 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "amdgpu.h"
+#include "nv.h"
+
+#include "soc15_common.h"
+#include "soc15_hw_ip.h"
+#include "cyan_skillfish_ip_offset.h"
+
+int cyan_skillfish_reg_base_init(struct amdgpu_device *adev)
+{
+	/* HW has more IP blocks,  only initialized the blocke needed by driver */
+	uint32_t i;
+
+	adev->gfx.xcc_mask = 1;
+	for (i = 0 ; i < MAX_INSTANCE ; ++i) {
+		adev->reg_offset[GC_HWIP][i] = (uint32_t *)(&(GC_BASE.instance[i]));
+		adev->reg_offset[HDP_HWIP][i] = (uint32_t *)(&(HDP_BASE.instance[i]));
+		adev->reg_offset[MMHUB_HWIP][i] = (uint32_t *)(&(MMHUB_BASE.instance[i]));
+		adev->reg_offset[ATHUB_HWIP][i] = (uint32_t *)(&(ATHUB_BASE.instance[i]));
+		adev->reg_offset[NBIO_HWIP][i] = (uint32_t *)(&(NBIO_BASE.instance[i]));
+		adev->reg_offset[MP0_HWIP][i] = (uint32_t *)(&(MP0_BASE.instance[i]));
+		adev->reg_offset[MP1_HWIP][i] = (uint32_t *)(&(MP1_BASE.instance[i]));
+		adev->reg_offset[VCN_HWIP][i] = (uint32_t *)(&(UVD0_BASE.instance[i]));
+		adev->reg_offset[DF_HWIP][i] = (uint32_t *)(&(DF_BASE.instance[i]));
+		adev->reg_offset[DCE_HWIP][i] = (uint32_t *)(&(DMU_BASE.instance[i]));
+		adev->reg_offset[OSSSYS_HWIP][i] = (uint32_t *)(&(OSSSYS_BASE.instance[i]));
+		adev->reg_offset[SDMA0_HWIP][i] = (uint32_t *)(&(GC_BASE.instance[i]));
+		adev->reg_offset[SDMA1_HWIP][i] = (uint32_t *)(&(GC_BASE.instance[i]));
+		adev->reg_offset[SMUIO_HWIP][i] = (uint32_t *)(&(SMUIO_BASE.instance[i]));
+		adev->reg_offset[THM_HWIP][i] = (uint32_t *)(&(THM_BASE.instance[i]));
+		adev->reg_offset[CLK_HWIP][i] = (uint32_t *)(&(CLK_BASE.instance[i]));
+	}
+	return 0;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/cz_ih.c b/drivers/gpu/drm/amd/amdgpu/cz_ih.c
index b8c47e0cf37a..bc7a2e06ab5f 100644
--- a/drivers/gpu/drm/amd/amdgpu/cz_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/cz_ih.c
@@ -157,6 +157,9 @@ static int cz_ih_irq_init(struct amdgpu_device *adev)
 	/* enable interrupts */
 	cz_ih_enable_interrupts(adev);
 
+	if (adev->irq.ih_soft.ring_size)
+		adev->irq.ih_soft.enabled = true;
+
 	return 0;
 }
 
@@ -194,6 +197,9 @@ static u32 cz_ih_get_wptr(struct amdgpu_device *adev,
 
 	wptr = le32_to_cpu(*ih->wptr_cpu);
 
+	if (ih == &adev->irq.ih_soft)
+		goto out;
+
 	if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
 		goto out;
 
@@ -216,6 +222,11 @@ static u32 cz_ih_get_wptr(struct amdgpu_device *adev,
 	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
 	WREG32(mmIH_RB_CNTL, tmp);
 
+	/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+	 * can be detected.
+	 */
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+	WREG32(mmIH_RB_CNTL, tmp);
 
 out:
 	return (wptr & ih->ptr_mask);
@@ -269,9 +280,9 @@ static void cz_ih_set_rptr(struct amdgpu_device *adev,
 	WREG32(mmIH_RB_RPTR, ih->rptr);
 }
 
-static int cz_ih_early_init(void *handle)
+static int cz_ih_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	ret = amdgpu_irq_add_domain(adev);
@@ -283,23 +294,27 @@ static int cz_ih_early_init(void *handle)
 	return 0;
 }
 
-static int cz_ih_sw_init(void *handle)
+static int cz_ih_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 64 * 1024, false);
 	if (r)
 		return r;
 
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, IH_SW_RING_SIZE, true);
+	if (r)
+		return r;
+
 	r = amdgpu_irq_init(adev);
 
 	return r;
 }
 
-static int cz_ih_sw_fini(void *handle)
+static int cz_ih_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_fini_sw(adev);
 	amdgpu_irq_remove_domain(adev);
@@ -307,10 +322,10 @@ static int cz_ih_sw_fini(void *handle)
 	return 0;
 }
 
-static int cz_ih_hw_init(void *handle)
+static int cz_ih_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = cz_ih_irq_init(adev);
 	if (r)
@@ -319,32 +334,26 @@ static int cz_ih_hw_init(void *handle)
 	return 0;
 }
 
-static int cz_ih_hw_fini(void *handle)
+static int cz_ih_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	cz_ih_irq_disable(adev);
+	cz_ih_irq_disable(ip_block->adev);
 
 	return 0;
 }
 
-static int cz_ih_suspend(void *handle)
+static int cz_ih_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return cz_ih_hw_fini(adev);
+	return cz_ih_hw_fini(ip_block);
 }
 
-static int cz_ih_resume(void *handle)
+static int cz_ih_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return cz_ih_hw_init(adev);
+	return cz_ih_hw_init(ip_block);
 }
 
-static bool cz_ih_is_idle(void *handle)
+static bool cz_ih_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (REG_GET_FIELD(tmp, SRBM_STATUS, IH_BUSY))
@@ -353,11 +362,11 @@ static bool cz_ih_is_idle(void *handle)
 	return true;
 }
 
-static int cz_ih_wait_for_idle(void *handle)
+static int cz_ih_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		/* read MC_STATUS */
@@ -369,10 +378,10 @@ static int cz_ih_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int cz_ih_soft_reset(void *handle)
+static int cz_ih_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 srbm_soft_reset = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (tmp & SRBM_STATUS__IH_BUSY_MASK)
@@ -399,14 +408,14 @@ static int cz_ih_soft_reset(void *handle)
 	return 0;
 }
 
-static int cz_ih_set_clockgating_state(void *handle,
+static int cz_ih_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	// TODO
 	return 0;
 }
 
-static int cz_ih_set_powergating_state(void *handle,
+static int cz_ih_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	// TODO
@@ -416,7 +425,6 @@ static int cz_ih_set_powergating_state(void *handle,
 static const struct amd_ip_funcs cz_ih_ip_funcs = {
 	.name = "cz_ih",
 	.early_init = cz_ih_early_init,
-	.late_init = NULL,
 	.sw_init = cz_ih_sw_init,
 	.sw_fini = cz_ih_sw_fini,
 	.hw_init = cz_ih_hw_init,
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c
index 9a24ed463abd..72ca6538b2e4 100644
--- a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c
@@ -21,6 +21,7 @@
  *
  */
 
+#include <drm/drm_edid.h>
 #include <drm/drm_fourcc.h>
 #include <drm/drm_modeset_helper.h>
 #include <drm/drm_modeset_helper_vtables.h>
@@ -51,9 +52,9 @@
 
 static void dce_v10_0_set_display_funcs(struct amdgpu_device *adev);
 static void dce_v10_0_set_irq_funcs(struct amdgpu_device *adev);
+static void dce_v10_0_hpd_int_ack(struct amdgpu_device *adev, int hpd);
 
-static const u32 crtc_offsets[] =
-{
+static const u32 crtc_offsets[] = {
 	CRTC0_REGISTER_OFFSET,
 	CRTC1_REGISTER_OFFSET,
 	CRTC2_REGISTER_OFFSET,
@@ -63,8 +64,7 @@ static const u32 crtc_offsets[] =
 	CRTC6_REGISTER_OFFSET
 };
 
-static const u32 hpd_offsets[] =
-{
+static const u32 hpd_offsets[] = {
 	HPD0_REGISTER_OFFSET,
 	HPD1_REGISTER_OFFSET,
 	HPD2_REGISTER_OFFSET,
@@ -121,30 +121,26 @@ static const struct {
 	.hpd = DISP_INTERRUPT_STATUS_CONTINUE5__DC_HPD6_INTERRUPT_MASK
 } };
 
-static const u32 golden_settings_tonga_a11[] =
-{
+static const u32 golden_settings_tonga_a11[] = {
 	mmDCI_CLK_CNTL, 0x00000080, 0x00000000,
 	mmFBC_DEBUG_COMP, 0x000000f0, 0x00000070,
 	mmFBC_MISC, 0x1f311fff, 0x12300000,
 	mmHDMI_CONTROL, 0x31000111, 0x00000011,
 };
 
-static const u32 tonga_mgcg_cgcg_init[] =
-{
+static const u32 tonga_mgcg_cgcg_init[] = {
 	mmXDMA_CLOCK_GATING_CNTL, 0xffffffff, 0x00000100,
 	mmXDMA_MEM_POWER_CNTL, 0x00000101, 0x00000000,
 };
 
-static const u32 golden_settings_fiji_a10[] =
-{
+static const u32 golden_settings_fiji_a10[] = {
 	mmDCI_CLK_CNTL, 0x00000080, 0x00000000,
 	mmFBC_DEBUG_COMP, 0x000000f0, 0x00000070,
 	mmFBC_MISC, 0x1f311fff, 0x12300000,
 	mmHDMI_CONTROL, 0x31000111, 0x00000011,
 };
 
-static const u32 fiji_mgcg_cgcg_init[] =
-{
+static const u32 fiji_mgcg_cgcg_init[] = {
 	mmXDMA_CLOCK_GATING_CNTL, 0xffffffff, 0x00000100,
 	mmXDMA_MEM_POWER_CNTL, 0x00000101, 0x00000000,
 };
@@ -369,6 +365,7 @@ static void dce_v10_0_hpd_init(struct amdgpu_device *adev)
 				    AMDGPU_HPD_DISCONNECT_INT_DELAY_IN_MS);
 		WREG32(mmDC_HPD_TOGGLE_FILT_CNTL + hpd_offsets[amdgpu_connector->hpd.hpd], tmp);
 
+		dce_v10_0_hpd_int_ack(adev, amdgpu_connector->hpd.hpd);
 		dce_v10_0_hpd_set_polarity(adev, amdgpu_connector->hpd.hpd);
 		amdgpu_irq_get(adev, &adev->hpd_irq,
 			       amdgpu_connector->hpd.hpd);
@@ -1042,7 +1039,7 @@ static void dce_v10_0_program_watermarks(struct amdgpu_device *adev,
 					    (u32)mode->clock);
 		line_time = (u32) div_u64((u64)mode->crtc_htotal * 1000000,
 					  (u32)mode->clock);
-		line_time = min(line_time, (u32)65535);
+		line_time = min_t(u32, line_time, 65535);
 
 		/* watermark for high clocks */
 		if (adev->pm.dpm_enabled) {
@@ -1072,7 +1069,7 @@ static void dce_v10_0_program_watermarks(struct amdgpu_device *adev,
 		wm_high.num_heads = num_heads;
 
 		/* set for high clocks */
-		latency_watermark_a = min(dce_v10_0_latency_watermark(&wm_high), (u32)65535);
+		latency_watermark_a = min_t(u32, dce_v10_0_latency_watermark(&wm_high), 65535);
 
 		/* possibly force display priority to high */
 		/* should really do this at mode validation time... */
@@ -1111,7 +1108,7 @@ static void dce_v10_0_program_watermarks(struct amdgpu_device *adev,
 		wm_low.num_heads = num_heads;
 
 		/* set for low clocks */
-		latency_watermark_b = min(dce_v10_0_latency_watermark(&wm_low), (u32)65535);
+		latency_watermark_b = min_t(u32, dce_v10_0_latency_watermark(&wm_low), 65535);
 
 		/* possibly force display priority to high */
 		/* should really do this at mode validation time... */
@@ -1144,8 +1141,7 @@ static void dce_v10_0_program_watermarks(struct amdgpu_device *adev,
 
 	/* save values for DPM */
 	amdgpu_crtc->line_time = line_time;
-	amdgpu_crtc->wm_high = latency_watermark_a;
-	amdgpu_crtc->wm_low = latency_watermark_b;
+
 	/* Save number of lines the linebuffer leads before the scanout */
 	amdgpu_crtc->lb_vblank_lead_lines = lb_vblank_lead_lines;
 }
@@ -1302,7 +1298,7 @@ static void dce_v10_0_audio_write_speaker_allocation(struct drm_encoder *encoder
 		return;
 	}
 
-	sad_count = drm_edid_to_speaker_allocation(amdgpu_connector_edid(connector), &sadb);
+	sad_count = drm_edid_to_speaker_allocation(amdgpu_connector->edid, &sadb);
 	if (sad_count < 0) {
 		DRM_ERROR("Couldn't read Speaker Allocation Data Block: %d\n", sad_count);
 		sad_count = 0;
@@ -1372,7 +1368,7 @@ static void dce_v10_0_audio_write_sad_regs(struct drm_encoder *encoder)
 		return;
 	}
 
-	sad_count = drm_edid_to_sad(amdgpu_connector_edid(connector), &sads);
+	sad_count = drm_edid_to_sad(amdgpu_connector->edid, &sads);
 	if (sad_count < 0)
 		DRM_ERROR("Couldn't read SADs: %d\n", sad_count);
 	if (sad_count <= 0)
@@ -1425,8 +1421,7 @@ static void dce_v10_0_audio_enable(struct amdgpu_device *adev,
 			   enable ? AZALIA_F0_CODEC_PIN_CONTROL_HOT_PLUG_CONTROL__AUDIO_ENABLED_MASK : 0);
 }
 
-static const u32 pin_offsets[] =
-{
+static const u32 pin_offsets[] = {
 	AUD0_REGISTER_OFFSET,
 	AUD1_REGISTER_OFFSET,
 	AUD2_REGISTER_OFFSET,
@@ -1466,17 +1461,12 @@ static int dce_v10_0_audio_init(struct amdgpu_device *adev)
 
 static void dce_v10_0_audio_fini(struct amdgpu_device *adev)
 {
-	int i;
-
 	if (!amdgpu_audio)
 		return;
 
 	if (!adev->mode_info.audio.enabled)
 		return;
 
-	for (i = 0; i < adev->mode_info.audio.num_pins; i++)
-		dce_v10_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
-
 	adev->mode_info.audio.enabled = false;
 }
 
@@ -1811,8 +1801,7 @@ static void dce_v10_0_afmt_fini(struct amdgpu_device *adev)
 	}
 }
 
-static const u32 vga_control_regs[6] =
-{
+static const u32 vga_control_regs[6] = {
 	mmD1VGA_CONTROL,
 	mmD2VGA_CONTROL,
 	mmD3VGA_CONTROL,
@@ -1886,6 +1875,7 @@ static int dce_v10_0_crtc_do_set_base(struct drm_crtc *crtc,
 		return r;
 
 	if (!atomic) {
+		abo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 		r = amdgpu_bo_pin(abo, AMDGPU_GEM_DOMAIN_VRAM);
 		if (unlikely(r != 0)) {
 			amdgpu_bo_unreserve(abo);
@@ -2406,6 +2396,7 @@ static int dce_v10_0_crtc_cursor_set2(struct drm_crtc *crtc,
 		return ret;
 	}
 
+	aobj->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 	ret = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM);
 	amdgpu_bo_unreserve(aobj);
 	if (ret) {
@@ -2690,6 +2681,32 @@ static const struct drm_crtc_helper_funcs dce_v10_0_crtc_helper_funcs = {
 	.get_scanout_position = amdgpu_crtc_get_scanout_position,
 };
 
+static void dce_v10_0_panic_flush(struct drm_plane *plane)
+{
+	struct drm_framebuffer *fb;
+	struct amdgpu_crtc *amdgpu_crtc;
+	struct amdgpu_device *adev;
+	uint32_t fb_format;
+
+	if (!plane->fb)
+		return;
+
+	fb = plane->fb;
+	amdgpu_crtc = to_amdgpu_crtc(plane->crtc);
+	adev = drm_to_adev(fb->dev);
+
+	/* Disable DC tiling */
+	fb_format = RREG32(mmGRPH_CONTROL + amdgpu_crtc->crtc_offset);
+	fb_format &= ~GRPH_CONTROL__GRPH_ARRAY_MODE_MASK;
+	WREG32(mmGRPH_CONTROL + amdgpu_crtc->crtc_offset, fb_format);
+
+}
+
+static const struct drm_plane_helper_funcs dce_v10_0_drm_primary_plane_helper_funcs = {
+	.get_scanout_buffer = amdgpu_display_get_scanout_buffer,
+	.panic_flush = dce_v10_0_panic_flush,
+};
+
 static int dce_v10_0_crtc_init(struct amdgpu_device *adev, int index)
 {
 	struct amdgpu_crtc *amdgpu_crtc;
@@ -2737,13 +2754,14 @@ static int dce_v10_0_crtc_init(struct amdgpu_device *adev, int index)
 	amdgpu_crtc->encoder = NULL;
 	amdgpu_crtc->connector = NULL;
 	drm_crtc_helper_add(&amdgpu_crtc->base, &dce_v10_0_crtc_helper_funcs);
+	drm_plane_helper_add(amdgpu_crtc->base.primary, &dce_v10_0_drm_primary_plane_helper_funcs);
 
 	return 0;
 }
 
-static int dce_v10_0_early_init(void *handle)
+static int dce_v10_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->audio_endpt_rreg = &dce_v10_0_audio_endpt_rreg;
 	adev->audio_endpt_wreg = &dce_v10_0_audio_endpt_wreg;
@@ -2768,10 +2786,10 @@ static int dce_v10_0_early_init(void *handle)
 	return 0;
 }
 
-static int dce_v10_0_sw_init(void *handle)
+static int dce_v10_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->mode_info.num_crtc; i++) {
 		r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i + 1, &adev->crtc_irq);
@@ -2847,11 +2865,11 @@ static int dce_v10_0_sw_init(void *handle)
 	return 0;
 }
 
-static int dce_v10_0_sw_fini(void *handle)
+static int dce_v10_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	kfree(adev->mode_info.bios_hardcoded_edid);
+	drm_edid_free(adev->mode_info.bios_hardcoded_edid);
 
 	drm_kms_helper_poll_fini(adev_to_drm(adev));
 
@@ -2865,10 +2883,10 @@ static int dce_v10_0_sw_fini(void *handle)
 	return 0;
 }
 
-static int dce_v10_0_hw_init(void *handle)
+static int dce_v10_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	dce_v10_0_init_golden_registers(adev);
 
@@ -2890,10 +2908,10 @@ static int dce_v10_0_hw_init(void *handle)
 	return 0;
 }
 
-static int dce_v10_0_hw_fini(void *handle)
+static int dce_v10_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	dce_v10_0_hpd_fini(adev);
 
@@ -2908,9 +2926,9 @@ static int dce_v10_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int dce_v10_0_suspend(void *handle)
+static int dce_v10_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_display_suspend_helper(adev);
@@ -2920,18 +2938,18 @@ static int dce_v10_0_suspend(void *handle)
 	adev->mode_info.bl_level =
 		amdgpu_atombios_encoder_get_backlight_level_from_reg(adev);
 
-	return dce_v10_0_hw_fini(handle);
+	return dce_v10_0_hw_fini(ip_block);
 }
 
-static int dce_v10_0_resume(void *handle)
+static int dce_v10_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	amdgpu_atombios_encoder_set_backlight_level_to_reg(adev,
 							   adev->mode_info.bl_level);
 
-	ret = dce_v10_0_hw_init(handle);
+	ret = dce_v10_0_hw_init(ip_block);
 
 	/* turn on the BL */
 	if (adev->mode_info.bl_encoder) {
@@ -2946,27 +2964,22 @@ static int dce_v10_0_resume(void *handle)
 	return amdgpu_display_resume_helper(adev);
 }
 
-static bool dce_v10_0_is_idle(void *handle)
+static bool dce_v10_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	return true;
 }
 
-static int dce_v10_0_wait_for_idle(void *handle)
+static bool dce_v10_0_check_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	return 0;
-}
-
-static bool dce_v10_0_check_soft_reset(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return dce_v10_0_is_display_hung(adev);
 }
 
-static int dce_v10_0_soft_reset(void *handle)
+static int dce_v10_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 srbm_soft_reset = 0, tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (dce_v10_0_is_display_hung(adev))
 		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_DC_MASK;
@@ -3056,7 +3069,7 @@ static int dce_v10_0_set_hpd_irq_state(struct amdgpu_device *adev,
 	u32 tmp;
 
 	if (hpd >= adev->mode_info.num_hpd) {
-		DRM_DEBUG("invalid hdp %d\n", hpd);
+		DRM_DEBUG("invalid hpd %d\n", hpd);
 		return 0;
 	}
 
@@ -3208,7 +3221,7 @@ static void dce_v10_0_hpd_int_ack(struct amdgpu_device *adev,
 	u32 tmp;
 
 	if (hpd >= adev->mode_info.num_hpd) {
-		DRM_DEBUG("invalid hdp %d\n", hpd);
+		DRM_DEBUG("invalid hpd %d\n", hpd);
 		return;
 	}
 
@@ -3310,13 +3323,13 @@ static int dce_v10_0_hpd_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int dce_v10_0_set_clockgating_state(void *handle,
+static int dce_v10_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int dce_v10_0_set_powergating_state(void *handle,
+static int dce_v10_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -3325,7 +3338,6 @@ static int dce_v10_0_set_powergating_state(void *handle,
 static const struct amd_ip_funcs dce_v10_0_ip_funcs = {
 	.name = "dce_v10_0",
 	.early_init = dce_v10_0_early_init,
-	.late_init = NULL,
 	.sw_init = dce_v10_0_sw_init,
 	.sw_fini = dce_v10_0_sw_fini,
 	.hw_init = dce_v10_0_hw_init,
@@ -3333,7 +3345,6 @@ static const struct amd_ip_funcs dce_v10_0_ip_funcs = {
 	.suspend = dce_v10_0_suspend,
 	.resume = dce_v10_0_resume,
 	.is_idle = dce_v10_0_is_idle,
-	.wait_for_idle = dce_v10_0_wait_for_idle,
 	.check_soft_reset = dce_v10_0_check_soft_reset,
 	.soft_reset = dce_v10_0_soft_reset,
 	.set_clockgating_state = dce_v10_0_set_clockgating_state,
@@ -3651,8 +3662,7 @@ static void dce_v10_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->hpd_irq.funcs = &dce_v10_0_hpd_irq_funcs;
 }
 
-const struct amdgpu_ip_block_version dce_v10_0_ip_block =
-{
+const struct amdgpu_ip_block_version dce_v10_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_DCE,
 	.major = 10,
 	.minor = 0,
@@ -3660,8 +3670,7 @@ const struct amdgpu_ip_block_version dce_v10_0_ip_block =
 	.funcs = &dce_v10_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version dce_v10_1_ip_block =
-{
+const struct amdgpu_ip_block_version dce_v10_1_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_DCE,
 	.major = 10,
 	.minor = 1,
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c
deleted file mode 100644
index c14b70350a51..000000000000
--- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c
+++ /dev/null
@@ -1,3799 +0,0 @@
-/*
- * Copyright 2014 Advanced Micro Devices, Inc.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
- * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
- * OTHER DEALINGS IN THE SOFTWARE.
- *
- */
-
-#include <drm/drm_fourcc.h>
-#include <drm/drm_modeset_helper.h>
-#include <drm/drm_modeset_helper_vtables.h>
-#include <drm/drm_vblank.h>
-
-#include "amdgpu.h"
-#include "amdgpu_pm.h"
-#include "amdgpu_i2c.h"
-#include "vid.h"
-#include "atom.h"
-#include "amdgpu_atombios.h"
-#include "atombios_crtc.h"
-#include "atombios_encoders.h"
-#include "amdgpu_pll.h"
-#include "amdgpu_connectors.h"
-#include "amdgpu_display.h"
-#include "dce_v11_0.h"
-
-#include "dce/dce_11_0_d.h"
-#include "dce/dce_11_0_sh_mask.h"
-#include "dce/dce_11_0_enum.h"
-#include "oss/oss_3_0_d.h"
-#include "oss/oss_3_0_sh_mask.h"
-#include "gmc/gmc_8_1_d.h"
-#include "gmc/gmc_8_1_sh_mask.h"
-
-#include "ivsrcid/ivsrcid_vislands30.h"
-
-static void dce_v11_0_set_display_funcs(struct amdgpu_device *adev);
-static void dce_v11_0_set_irq_funcs(struct amdgpu_device *adev);
-
-static const u32 crtc_offsets[] =
-{
-	CRTC0_REGISTER_OFFSET,
-	CRTC1_REGISTER_OFFSET,
-	CRTC2_REGISTER_OFFSET,
-	CRTC3_REGISTER_OFFSET,
-	CRTC4_REGISTER_OFFSET,
-	CRTC5_REGISTER_OFFSET,
-	CRTC6_REGISTER_OFFSET
-};
-
-static const u32 hpd_offsets[] =
-{
-	HPD0_REGISTER_OFFSET,
-	HPD1_REGISTER_OFFSET,
-	HPD2_REGISTER_OFFSET,
-	HPD3_REGISTER_OFFSET,
-	HPD4_REGISTER_OFFSET,
-	HPD5_REGISTER_OFFSET
-};
-
-static const uint32_t dig_offsets[] = {
-	DIG0_REGISTER_OFFSET,
-	DIG1_REGISTER_OFFSET,
-	DIG2_REGISTER_OFFSET,
-	DIG3_REGISTER_OFFSET,
-	DIG4_REGISTER_OFFSET,
-	DIG5_REGISTER_OFFSET,
-	DIG6_REGISTER_OFFSET,
-	DIG7_REGISTER_OFFSET,
-	DIG8_REGISTER_OFFSET
-};
-
-static const struct {
-	uint32_t        reg;
-	uint32_t        vblank;
-	uint32_t        vline;
-	uint32_t        hpd;
-
-} interrupt_status_offsets[] = { {
-	.reg = mmDISP_INTERRUPT_STATUS,
-	.vblank = DISP_INTERRUPT_STATUS__LB_D1_VBLANK_INTERRUPT_MASK,
-	.vline = DISP_INTERRUPT_STATUS__LB_D1_VLINE_INTERRUPT_MASK,
-	.hpd = DISP_INTERRUPT_STATUS__DC_HPD1_INTERRUPT_MASK
-}, {
-	.reg = mmDISP_INTERRUPT_STATUS_CONTINUE,
-	.vblank = DISP_INTERRUPT_STATUS_CONTINUE__LB_D2_VBLANK_INTERRUPT_MASK,
-	.vline = DISP_INTERRUPT_STATUS_CONTINUE__LB_D2_VLINE_INTERRUPT_MASK,
-	.hpd = DISP_INTERRUPT_STATUS_CONTINUE__DC_HPD2_INTERRUPT_MASK
-}, {
-	.reg = mmDISP_INTERRUPT_STATUS_CONTINUE2,
-	.vblank = DISP_INTERRUPT_STATUS_CONTINUE2__LB_D3_VBLANK_INTERRUPT_MASK,
-	.vline = DISP_INTERRUPT_STATUS_CONTINUE2__LB_D3_VLINE_INTERRUPT_MASK,
-	.hpd = DISP_INTERRUPT_STATUS_CONTINUE2__DC_HPD3_INTERRUPT_MASK
-}, {
-	.reg = mmDISP_INTERRUPT_STATUS_CONTINUE3,
-	.vblank = DISP_INTERRUPT_STATUS_CONTINUE3__LB_D4_VBLANK_INTERRUPT_MASK,
-	.vline = DISP_INTERRUPT_STATUS_CONTINUE3__LB_D4_VLINE_INTERRUPT_MASK,
-	.hpd = DISP_INTERRUPT_STATUS_CONTINUE3__DC_HPD4_INTERRUPT_MASK
-}, {
-	.reg = mmDISP_INTERRUPT_STATUS_CONTINUE4,
-	.vblank = DISP_INTERRUPT_STATUS_CONTINUE4__LB_D5_VBLANK_INTERRUPT_MASK,
-	.vline = DISP_INTERRUPT_STATUS_CONTINUE4__LB_D5_VLINE_INTERRUPT_MASK,
-	.hpd = DISP_INTERRUPT_STATUS_CONTINUE4__DC_HPD5_INTERRUPT_MASK
-}, {
-	.reg = mmDISP_INTERRUPT_STATUS_CONTINUE5,
-	.vblank = DISP_INTERRUPT_STATUS_CONTINUE5__LB_D6_VBLANK_INTERRUPT_MASK,
-	.vline = DISP_INTERRUPT_STATUS_CONTINUE5__LB_D6_VLINE_INTERRUPT_MASK,
-	.hpd = DISP_INTERRUPT_STATUS_CONTINUE5__DC_HPD6_INTERRUPT_MASK
-} };
-
-static const u32 cz_golden_settings_a11[] =
-{
-	mmCRTC_DOUBLE_BUFFER_CONTROL, 0x00010101, 0x00010000,
-	mmFBC_MISC, 0x1f311fff, 0x14300000,
-};
-
-static const u32 cz_mgcg_cgcg_init[] =
-{
-	mmXDMA_CLOCK_GATING_CNTL, 0xffffffff, 0x00000100,
-	mmXDMA_MEM_POWER_CNTL, 0x00000101, 0x00000000,
-};
-
-static const u32 stoney_golden_settings_a11[] =
-{
-	mmCRTC_DOUBLE_BUFFER_CONTROL, 0x00010101, 0x00010000,
-	mmFBC_MISC, 0x1f311fff, 0x14302000,
-};
-
-static const u32 polaris11_golden_settings_a11[] =
-{
-	mmDCI_CLK_CNTL, 0x00000080, 0x00000000,
-	mmFBC_DEBUG_COMP, 0x000000f0, 0x00000070,
-	mmFBC_DEBUG1, 0xffffffff, 0x00000008,
-	mmFBC_MISC, 0x9f313fff, 0x14302008,
-	mmHDMI_CONTROL, 0x313f031f, 0x00000011,
-};
-
-static const u32 polaris10_golden_settings_a11[] =
-{
-	mmDCI_CLK_CNTL, 0x00000080, 0x00000000,
-	mmFBC_DEBUG_COMP, 0x000000f0, 0x00000070,
-	mmFBC_MISC, 0x9f313fff, 0x14302008,
-	mmHDMI_CONTROL, 0x313f031f, 0x00000011,
-};
-
-static void dce_v11_0_init_golden_registers(struct amdgpu_device *adev)
-{
-	switch (adev->asic_type) {
-	case CHIP_CARRIZO:
-		amdgpu_device_program_register_sequence(adev,
-							cz_mgcg_cgcg_init,
-							ARRAY_SIZE(cz_mgcg_cgcg_init));
-		amdgpu_device_program_register_sequence(adev,
-							cz_golden_settings_a11,
-							ARRAY_SIZE(cz_golden_settings_a11));
-		break;
-	case CHIP_STONEY:
-		amdgpu_device_program_register_sequence(adev,
-							stoney_golden_settings_a11,
-							ARRAY_SIZE(stoney_golden_settings_a11));
-		break;
-	case CHIP_POLARIS11:
-	case CHIP_POLARIS12:
-		amdgpu_device_program_register_sequence(adev,
-							polaris11_golden_settings_a11,
-							ARRAY_SIZE(polaris11_golden_settings_a11));
-		break;
-	case CHIP_POLARIS10:
-	case CHIP_VEGAM:
-		amdgpu_device_program_register_sequence(adev,
-							polaris10_golden_settings_a11,
-							ARRAY_SIZE(polaris10_golden_settings_a11));
-		break;
-	default:
-		break;
-	}
-}
-
-static u32 dce_v11_0_audio_endpt_rreg(struct amdgpu_device *adev,
-				     u32 block_offset, u32 reg)
-{
-	unsigned long flags;
-	u32 r;
-
-	spin_lock_irqsave(&adev->audio_endpt_idx_lock, flags);
-	WREG32(mmAZALIA_F0_CODEC_ENDPOINT_INDEX + block_offset, reg);
-	r = RREG32(mmAZALIA_F0_CODEC_ENDPOINT_DATA + block_offset);
-	spin_unlock_irqrestore(&adev->audio_endpt_idx_lock, flags);
-
-	return r;
-}
-
-static void dce_v11_0_audio_endpt_wreg(struct amdgpu_device *adev,
-				      u32 block_offset, u32 reg, u32 v)
-{
-	unsigned long flags;
-
-	spin_lock_irqsave(&adev->audio_endpt_idx_lock, flags);
-	WREG32(mmAZALIA_F0_CODEC_ENDPOINT_INDEX + block_offset, reg);
-	WREG32(mmAZALIA_F0_CODEC_ENDPOINT_DATA + block_offset, v);
-	spin_unlock_irqrestore(&adev->audio_endpt_idx_lock, flags);
-}
-
-static u32 dce_v11_0_vblank_get_counter(struct amdgpu_device *adev, int crtc)
-{
-	if (crtc < 0 || crtc >= adev->mode_info.num_crtc)
-		return 0;
-	else
-		return RREG32(mmCRTC_STATUS_FRAME_COUNT + crtc_offsets[crtc]);
-}
-
-static void dce_v11_0_pageflip_interrupt_init(struct amdgpu_device *adev)
-{
-	unsigned i;
-
-	/* Enable pflip interrupts */
-	for (i = 0; i < adev->mode_info.num_crtc; i++)
-		amdgpu_irq_get(adev, &adev->pageflip_irq, i);
-}
-
-static void dce_v11_0_pageflip_interrupt_fini(struct amdgpu_device *adev)
-{
-	unsigned i;
-
-	/* Disable pflip interrupts */
-	for (i = 0; i < adev->mode_info.num_crtc; i++)
-		amdgpu_irq_put(adev, &adev->pageflip_irq, i);
-}
-
-/**
- * dce_v11_0_page_flip - pageflip callback.
- *
- * @adev: amdgpu_device pointer
- * @crtc_id: crtc to cleanup pageflip on
- * @crtc_base: new address of the crtc (GPU MC address)
- * @async: asynchronous flip
- *
- * Triggers the actual pageflip by updating the primary
- * surface base address.
- */
-static void dce_v11_0_page_flip(struct amdgpu_device *adev,
-				int crtc_id, u64 crtc_base, bool async)
-{
-	struct amdgpu_crtc *amdgpu_crtc = adev->mode_info.crtcs[crtc_id];
-	struct drm_framebuffer *fb = amdgpu_crtc->base.primary->fb;
-	u32 tmp;
-
-	/* flip immediate for async, default is vsync */
-	tmp = RREG32(mmGRPH_FLIP_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, GRPH_FLIP_CONTROL,
-			    GRPH_SURFACE_UPDATE_IMMEDIATE_EN, async ? 1 : 0);
-	WREG32(mmGRPH_FLIP_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-	/* update pitch */
-	WREG32(mmGRPH_PITCH + amdgpu_crtc->crtc_offset,
-	       fb->pitches[0] / fb->format->cpp[0]);
-	/* update the scanout addresses */
-	WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS_HIGH + amdgpu_crtc->crtc_offset,
-	       upper_32_bits(crtc_base));
-	/* writing to the low address triggers the update */
-	WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
-	       lower_32_bits(crtc_base));
-	/* post the write */
-	RREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset);
-}
-
-static int dce_v11_0_crtc_get_scanoutpos(struct amdgpu_device *adev, int crtc,
-					u32 *vbl, u32 *position)
-{
-	if ((crtc < 0) || (crtc >= adev->mode_info.num_crtc))
-		return -EINVAL;
-
-	*vbl = RREG32(mmCRTC_V_BLANK_START_END + crtc_offsets[crtc]);
-	*position = RREG32(mmCRTC_STATUS_POSITION + crtc_offsets[crtc]);
-
-	return 0;
-}
-
-/**
- * dce_v11_0_hpd_sense - hpd sense callback.
- *
- * @adev: amdgpu_device pointer
- * @hpd: hpd (hotplug detect) pin
- *
- * Checks if a digital monitor is connected (evergreen+).
- * Returns true if connected, false if not connected.
- */
-static bool dce_v11_0_hpd_sense(struct amdgpu_device *adev,
-			       enum amdgpu_hpd_id hpd)
-{
-	bool connected = false;
-
-	if (hpd >= adev->mode_info.num_hpd)
-		return connected;
-
-	if (RREG32(mmDC_HPD_INT_STATUS + hpd_offsets[hpd]) &
-	    DC_HPD_INT_STATUS__DC_HPD_SENSE_MASK)
-		connected = true;
-
-	return connected;
-}
-
-/**
- * dce_v11_0_hpd_set_polarity - hpd set polarity callback.
- *
- * @adev: amdgpu_device pointer
- * @hpd: hpd (hotplug detect) pin
- *
- * Set the polarity of the hpd pin (evergreen+).
- */
-static void dce_v11_0_hpd_set_polarity(struct amdgpu_device *adev,
-				      enum amdgpu_hpd_id hpd)
-{
-	u32 tmp;
-	bool connected = dce_v11_0_hpd_sense(adev, hpd);
-
-	if (hpd >= adev->mode_info.num_hpd)
-		return;
-
-	tmp = RREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[hpd]);
-	if (connected)
-		tmp = REG_SET_FIELD(tmp, DC_HPD_INT_CONTROL, DC_HPD_INT_POLARITY, 0);
-	else
-		tmp = REG_SET_FIELD(tmp, DC_HPD_INT_CONTROL, DC_HPD_INT_POLARITY, 1);
-	WREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[hpd], tmp);
-}
-
-/**
- * dce_v11_0_hpd_init - hpd setup callback.
- *
- * @adev: amdgpu_device pointer
- *
- * Setup the hpd pins used by the card (evergreen+).
- * Enable the pin, set the polarity, and enable the hpd interrupts.
- */
-static void dce_v11_0_hpd_init(struct amdgpu_device *adev)
-{
-	struct drm_device *dev = adev_to_drm(adev);
-	struct drm_connector *connector;
-	struct drm_connector_list_iter iter;
-	u32 tmp;
-
-	drm_connector_list_iter_begin(dev, &iter);
-	drm_for_each_connector_iter(connector, &iter) {
-		struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
-
-		if (amdgpu_connector->hpd.hpd >= adev->mode_info.num_hpd)
-			continue;
-
-		if (connector->connector_type == DRM_MODE_CONNECTOR_eDP ||
-		    connector->connector_type == DRM_MODE_CONNECTOR_LVDS) {
-			/* don't try to enable hpd on eDP or LVDS avoid breaking the
-			 * aux dp channel on imac and help (but not completely fix)
-			 * https://bugzilla.redhat.com/show_bug.cgi?id=726143
-			 * also avoid interrupt storms during dpms.
-			 */
-			tmp = RREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[amdgpu_connector->hpd.hpd]);
-			tmp = REG_SET_FIELD(tmp, DC_HPD_INT_CONTROL, DC_HPD_INT_EN, 0);
-			WREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[amdgpu_connector->hpd.hpd], tmp);
-			continue;
-		}
-
-		tmp = RREG32(mmDC_HPD_CONTROL + hpd_offsets[amdgpu_connector->hpd.hpd]);
-		tmp = REG_SET_FIELD(tmp, DC_HPD_CONTROL, DC_HPD_EN, 1);
-		WREG32(mmDC_HPD_CONTROL + hpd_offsets[amdgpu_connector->hpd.hpd], tmp);
-
-		tmp = RREG32(mmDC_HPD_TOGGLE_FILT_CNTL + hpd_offsets[amdgpu_connector->hpd.hpd]);
-		tmp = REG_SET_FIELD(tmp, DC_HPD_TOGGLE_FILT_CNTL,
-				    DC_HPD_CONNECT_INT_DELAY,
-				    AMDGPU_HPD_CONNECT_INT_DELAY_IN_MS);
-		tmp = REG_SET_FIELD(tmp, DC_HPD_TOGGLE_FILT_CNTL,
-				    DC_HPD_DISCONNECT_INT_DELAY,
-				    AMDGPU_HPD_DISCONNECT_INT_DELAY_IN_MS);
-		WREG32(mmDC_HPD_TOGGLE_FILT_CNTL + hpd_offsets[amdgpu_connector->hpd.hpd], tmp);
-
-		dce_v11_0_hpd_set_polarity(adev, amdgpu_connector->hpd.hpd);
-		amdgpu_irq_get(adev, &adev->hpd_irq, amdgpu_connector->hpd.hpd);
-	}
-	drm_connector_list_iter_end(&iter);
-}
-
-/**
- * dce_v11_0_hpd_fini - hpd tear down callback.
- *
- * @adev: amdgpu_device pointer
- *
- * Tear down the hpd pins used by the card (evergreen+).
- * Disable the hpd interrupts.
- */
-static void dce_v11_0_hpd_fini(struct amdgpu_device *adev)
-{
-	struct drm_device *dev = adev_to_drm(adev);
-	struct drm_connector *connector;
-	struct drm_connector_list_iter iter;
-	u32 tmp;
-
-	drm_connector_list_iter_begin(dev, &iter);
-	drm_for_each_connector_iter(connector, &iter) {
-		struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
-
-		if (amdgpu_connector->hpd.hpd >= adev->mode_info.num_hpd)
-			continue;
-
-		tmp = RREG32(mmDC_HPD_CONTROL + hpd_offsets[amdgpu_connector->hpd.hpd]);
-		tmp = REG_SET_FIELD(tmp, DC_HPD_CONTROL, DC_HPD_EN, 0);
-		WREG32(mmDC_HPD_CONTROL + hpd_offsets[amdgpu_connector->hpd.hpd], tmp);
-
-		amdgpu_irq_put(adev, &adev->hpd_irq, amdgpu_connector->hpd.hpd);
-	}
-	drm_connector_list_iter_end(&iter);
-}
-
-static u32 dce_v11_0_hpd_get_gpio_reg(struct amdgpu_device *adev)
-{
-	return mmDC_GPIO_HPD_A;
-}
-
-static bool dce_v11_0_is_display_hung(struct amdgpu_device *adev)
-{
-	u32 crtc_hung = 0;
-	u32 crtc_status[6];
-	u32 i, j, tmp;
-
-	for (i = 0; i < adev->mode_info.num_crtc; i++) {
-		tmp = RREG32(mmCRTC_CONTROL + crtc_offsets[i]);
-		if (REG_GET_FIELD(tmp, CRTC_CONTROL, CRTC_MASTER_EN)) {
-			crtc_status[i] = RREG32(mmCRTC_STATUS_HV_COUNT + crtc_offsets[i]);
-			crtc_hung |= (1 << i);
-		}
-	}
-
-	for (j = 0; j < 10; j++) {
-		for (i = 0; i < adev->mode_info.num_crtc; i++) {
-			if (crtc_hung & (1 << i)) {
-				tmp = RREG32(mmCRTC_STATUS_HV_COUNT + crtc_offsets[i]);
-				if (tmp != crtc_status[i])
-					crtc_hung &= ~(1 << i);
-			}
-		}
-		if (crtc_hung == 0)
-			return false;
-		udelay(100);
-	}
-
-	return true;
-}
-
-static void dce_v11_0_set_vga_render_state(struct amdgpu_device *adev,
-					   bool render)
-{
-	u32 tmp;
-
-	/* Lockout access through VGA aperture*/
-	tmp = RREG32(mmVGA_HDP_CONTROL);
-	if (render)
-		tmp = REG_SET_FIELD(tmp, VGA_HDP_CONTROL, VGA_MEMORY_DISABLE, 0);
-	else
-		tmp = REG_SET_FIELD(tmp, VGA_HDP_CONTROL, VGA_MEMORY_DISABLE, 1);
-	WREG32(mmVGA_HDP_CONTROL, tmp);
-
-	/* disable VGA render */
-	tmp = RREG32(mmVGA_RENDER_CONTROL);
-	if (render)
-		tmp = REG_SET_FIELD(tmp, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 1);
-	else
-		tmp = REG_SET_FIELD(tmp, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 0);
-	WREG32(mmVGA_RENDER_CONTROL, tmp);
-}
-
-static int dce_v11_0_get_num_crtc (struct amdgpu_device *adev)
-{
-	int num_crtc = 0;
-
-	switch (adev->asic_type) {
-	case CHIP_CARRIZO:
-		num_crtc = 3;
-		break;
-	case CHIP_STONEY:
-		num_crtc = 2;
-		break;
-	case CHIP_POLARIS10:
-	case CHIP_VEGAM:
-		num_crtc = 6;
-		break;
-	case CHIP_POLARIS11:
-	case CHIP_POLARIS12:
-		num_crtc = 5;
-		break;
-	default:
-		num_crtc = 0;
-	}
-	return num_crtc;
-}
-
-void dce_v11_0_disable_dce(struct amdgpu_device *adev)
-{
-	/*Disable VGA render and enabled crtc, if has DCE engine*/
-	if (amdgpu_atombios_has_dce_engine_info(adev)) {
-		u32 tmp;
-		int crtc_enabled, i;
-
-		dce_v11_0_set_vga_render_state(adev, false);
-
-		/*Disable crtc*/
-		for (i = 0; i < dce_v11_0_get_num_crtc(adev); i++) {
-			crtc_enabled = REG_GET_FIELD(RREG32(mmCRTC_CONTROL + crtc_offsets[i]),
-									 CRTC_CONTROL, CRTC_MASTER_EN);
-			if (crtc_enabled) {
-				WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 1);
-				tmp = RREG32(mmCRTC_CONTROL + crtc_offsets[i]);
-				tmp = REG_SET_FIELD(tmp, CRTC_CONTROL, CRTC_MASTER_EN, 0);
-				WREG32(mmCRTC_CONTROL + crtc_offsets[i], tmp);
-				WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 0);
-			}
-		}
-	}
-}
-
-static void dce_v11_0_program_fmt(struct drm_encoder *encoder)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(encoder->crtc);
-	struct drm_connector *connector = amdgpu_get_connector_for_encoder(encoder);
-	int bpc = 0;
-	u32 tmp = 0;
-	enum amdgpu_connector_dither dither = AMDGPU_FMT_DITHER_DISABLE;
-
-	if (connector) {
-		struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
-		bpc = amdgpu_connector_get_monitor_bpc(connector);
-		dither = amdgpu_connector->dither;
-	}
-
-	/* LVDS/eDP FMT is set up by atom */
-	if (amdgpu_encoder->devices & ATOM_DEVICE_LCD_SUPPORT)
-		return;
-
-	/* not needed for analog */
-	if ((amdgpu_encoder->encoder_id == ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC1) ||
-	    (amdgpu_encoder->encoder_id == ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC2))
-		return;
-
-	if (bpc == 0)
-		return;
-
-	switch (bpc) {
-	case 6:
-		if (dither == AMDGPU_FMT_DITHER_ENABLE) {
-			/* XXX sort out optimal dither settings */
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_FRAME_RANDOM_ENABLE, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_HIGHPASS_RANDOM_ENABLE, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_SPATIAL_DITHER_EN, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_SPATIAL_DITHER_DEPTH, 0);
-		} else {
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_TRUNCATE_EN, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_TRUNCATE_DEPTH, 0);
-		}
-		break;
-	case 8:
-		if (dither == AMDGPU_FMT_DITHER_ENABLE) {
-			/* XXX sort out optimal dither settings */
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_FRAME_RANDOM_ENABLE, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_HIGHPASS_RANDOM_ENABLE, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_RGB_RANDOM_ENABLE, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_SPATIAL_DITHER_EN, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_SPATIAL_DITHER_DEPTH, 1);
-		} else {
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_TRUNCATE_EN, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_TRUNCATE_DEPTH, 1);
-		}
-		break;
-	case 10:
-		if (dither == AMDGPU_FMT_DITHER_ENABLE) {
-			/* XXX sort out optimal dither settings */
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_FRAME_RANDOM_ENABLE, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_HIGHPASS_RANDOM_ENABLE, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_RGB_RANDOM_ENABLE, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_SPATIAL_DITHER_EN, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_SPATIAL_DITHER_DEPTH, 2);
-		} else {
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_TRUNCATE_EN, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_TRUNCATE_DEPTH, 2);
-		}
-		break;
-	default:
-		/* not needed */
-		break;
-	}
-
-	WREG32(mmFMT_BIT_DEPTH_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-}
-
-
-/* display watermark setup */
-/**
- * dce_v11_0_line_buffer_adjust - Set up the line buffer
- *
- * @adev: amdgpu_device pointer
- * @amdgpu_crtc: the selected display controller
- * @mode: the current display mode on the selected display
- * controller
- *
- * Setup up the line buffer allocation for
- * the selected display controller (CIK).
- * Returns the line buffer size in pixels.
- */
-static u32 dce_v11_0_line_buffer_adjust(struct amdgpu_device *adev,
-				       struct amdgpu_crtc *amdgpu_crtc,
-				       struct drm_display_mode *mode)
-{
-	u32 tmp, buffer_alloc, i, mem_cfg;
-	u32 pipe_offset = amdgpu_crtc->crtc_id;
-	/*
-	 * Line Buffer Setup
-	 * There are 6 line buffers, one for each display controllers.
-	 * There are 3 partitions per LB. Select the number of partitions
-	 * to enable based on the display width.  For display widths larger
-	 * than 4096, you need use to use 2 display controllers and combine
-	 * them using the stereo blender.
-	 */
-	if (amdgpu_crtc->base.enabled && mode) {
-		if (mode->crtc_hdisplay < 1920) {
-			mem_cfg = 1;
-			buffer_alloc = 2;
-		} else if (mode->crtc_hdisplay < 2560) {
-			mem_cfg = 2;
-			buffer_alloc = 2;
-		} else if (mode->crtc_hdisplay < 4096) {
-			mem_cfg = 0;
-			buffer_alloc = (adev->flags & AMD_IS_APU) ? 2 : 4;
-		} else {
-			DRM_DEBUG_KMS("Mode too big for LB!\n");
-			mem_cfg = 0;
-			buffer_alloc = (adev->flags & AMD_IS_APU) ? 2 : 4;
-		}
-	} else {
-		mem_cfg = 1;
-		buffer_alloc = 0;
-	}
-
-	tmp = RREG32(mmLB_MEMORY_CTRL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, LB_MEMORY_CTRL, LB_MEMORY_CONFIG, mem_cfg);
-	WREG32(mmLB_MEMORY_CTRL + amdgpu_crtc->crtc_offset, tmp);
-
-	tmp = RREG32(mmPIPE0_DMIF_BUFFER_CONTROL + pipe_offset);
-	tmp = REG_SET_FIELD(tmp, PIPE0_DMIF_BUFFER_CONTROL, DMIF_BUFFERS_ALLOCATED, buffer_alloc);
-	WREG32(mmPIPE0_DMIF_BUFFER_CONTROL + pipe_offset, tmp);
-
-	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(mmPIPE0_DMIF_BUFFER_CONTROL + pipe_offset);
-		if (REG_GET_FIELD(tmp, PIPE0_DMIF_BUFFER_CONTROL, DMIF_BUFFERS_ALLOCATION_COMPLETED))
-			break;
-		udelay(1);
-	}
-
-	if (amdgpu_crtc->base.enabled && mode) {
-		switch (mem_cfg) {
-		case 0:
-		default:
-			return 4096 * 2;
-		case 1:
-			return 1920 * 2;
-		case 2:
-			return 2560 * 2;
-		}
-	}
-
-	/* controller not enabled, so no lb used */
-	return 0;
-}
-
-/**
- * cik_get_number_of_dram_channels - get the number of dram channels
- *
- * @adev: amdgpu_device pointer
- *
- * Look up the number of video ram channels (CIK).
- * Used for display watermark bandwidth calculations
- * Returns the number of dram channels
- */
-static u32 cik_get_number_of_dram_channels(struct amdgpu_device *adev)
-{
-	u32 tmp = RREG32(mmMC_SHARED_CHMAP);
-
-	switch (REG_GET_FIELD(tmp, MC_SHARED_CHMAP, NOOFCHAN)) {
-	case 0:
-	default:
-		return 1;
-	case 1:
-		return 2;
-	case 2:
-		return 4;
-	case 3:
-		return 8;
-	case 4:
-		return 3;
-	case 5:
-		return 6;
-	case 6:
-		return 10;
-	case 7:
-		return 12;
-	case 8:
-		return 16;
-	}
-}
-
-struct dce10_wm_params {
-	u32 dram_channels; /* number of dram channels */
-	u32 yclk;          /* bandwidth per dram data pin in kHz */
-	u32 sclk;          /* engine clock in kHz */
-	u32 disp_clk;      /* display clock in kHz */
-	u32 src_width;     /* viewport width */
-	u32 active_time;   /* active display time in ns */
-	u32 blank_time;    /* blank time in ns */
-	bool interlaced;    /* mode is interlaced */
-	fixed20_12 vsc;    /* vertical scale ratio */
-	u32 num_heads;     /* number of active crtcs */
-	u32 bytes_per_pixel; /* bytes per pixel display + overlay */
-	u32 lb_size;       /* line buffer allocated to pipe */
-	u32 vtaps;         /* vertical scaler taps */
-};
-
-/**
- * dce_v11_0_dram_bandwidth - get the dram bandwidth
- *
- * @wm: watermark calculation data
- *
- * Calculate the raw dram bandwidth (CIK).
- * Used for display watermark bandwidth calculations
- * Returns the dram bandwidth in MBytes/s
- */
-static u32 dce_v11_0_dram_bandwidth(struct dce10_wm_params *wm)
-{
-	/* Calculate raw DRAM Bandwidth */
-	fixed20_12 dram_efficiency; /* 0.7 */
-	fixed20_12 yclk, dram_channels, bandwidth;
-	fixed20_12 a;
-
-	a.full = dfixed_const(1000);
-	yclk.full = dfixed_const(wm->yclk);
-	yclk.full = dfixed_div(yclk, a);
-	dram_channels.full = dfixed_const(wm->dram_channels * 4);
-	a.full = dfixed_const(10);
-	dram_efficiency.full = dfixed_const(7);
-	dram_efficiency.full = dfixed_div(dram_efficiency, a);
-	bandwidth.full = dfixed_mul(dram_channels, yclk);
-	bandwidth.full = dfixed_mul(bandwidth, dram_efficiency);
-
-	return dfixed_trunc(bandwidth);
-}
-
-/**
- * dce_v11_0_dram_bandwidth_for_display - get the dram bandwidth for display
- *
- * @wm: watermark calculation data
- *
- * Calculate the dram bandwidth used for display (CIK).
- * Used for display watermark bandwidth calculations
- * Returns the dram bandwidth for display in MBytes/s
- */
-static u32 dce_v11_0_dram_bandwidth_for_display(struct dce10_wm_params *wm)
-{
-	/* Calculate DRAM Bandwidth and the part allocated to display. */
-	fixed20_12 disp_dram_allocation; /* 0.3 to 0.7 */
-	fixed20_12 yclk, dram_channels, bandwidth;
-	fixed20_12 a;
-
-	a.full = dfixed_const(1000);
-	yclk.full = dfixed_const(wm->yclk);
-	yclk.full = dfixed_div(yclk, a);
-	dram_channels.full = dfixed_const(wm->dram_channels * 4);
-	a.full = dfixed_const(10);
-	disp_dram_allocation.full = dfixed_const(3); /* XXX worse case value 0.3 */
-	disp_dram_allocation.full = dfixed_div(disp_dram_allocation, a);
-	bandwidth.full = dfixed_mul(dram_channels, yclk);
-	bandwidth.full = dfixed_mul(bandwidth, disp_dram_allocation);
-
-	return dfixed_trunc(bandwidth);
-}
-
-/**
- * dce_v11_0_data_return_bandwidth - get the data return bandwidth
- *
- * @wm: watermark calculation data
- *
- * Calculate the data return bandwidth used for display (CIK).
- * Used for display watermark bandwidth calculations
- * Returns the data return bandwidth in MBytes/s
- */
-static u32 dce_v11_0_data_return_bandwidth(struct dce10_wm_params *wm)
-{
-	/* Calculate the display Data return Bandwidth */
-	fixed20_12 return_efficiency; /* 0.8 */
-	fixed20_12 sclk, bandwidth;
-	fixed20_12 a;
-
-	a.full = dfixed_const(1000);
-	sclk.full = dfixed_const(wm->sclk);
-	sclk.full = dfixed_div(sclk, a);
-	a.full = dfixed_const(10);
-	return_efficiency.full = dfixed_const(8);
-	return_efficiency.full = dfixed_div(return_efficiency, a);
-	a.full = dfixed_const(32);
-	bandwidth.full = dfixed_mul(a, sclk);
-	bandwidth.full = dfixed_mul(bandwidth, return_efficiency);
-
-	return dfixed_trunc(bandwidth);
-}
-
-/**
- * dce_v11_0_dmif_request_bandwidth - get the dmif bandwidth
- *
- * @wm: watermark calculation data
- *
- * Calculate the dmif bandwidth used for display (CIK).
- * Used for display watermark bandwidth calculations
- * Returns the dmif bandwidth in MBytes/s
- */
-static u32 dce_v11_0_dmif_request_bandwidth(struct dce10_wm_params *wm)
-{
-	/* Calculate the DMIF Request Bandwidth */
-	fixed20_12 disp_clk_request_efficiency; /* 0.8 */
-	fixed20_12 disp_clk, bandwidth;
-	fixed20_12 a, b;
-
-	a.full = dfixed_const(1000);
-	disp_clk.full = dfixed_const(wm->disp_clk);
-	disp_clk.full = dfixed_div(disp_clk, a);
-	a.full = dfixed_const(32);
-	b.full = dfixed_mul(a, disp_clk);
-
-	a.full = dfixed_const(10);
-	disp_clk_request_efficiency.full = dfixed_const(8);
-	disp_clk_request_efficiency.full = dfixed_div(disp_clk_request_efficiency, a);
-
-	bandwidth.full = dfixed_mul(b, disp_clk_request_efficiency);
-
-	return dfixed_trunc(bandwidth);
-}
-
-/**
- * dce_v11_0_available_bandwidth - get the min available bandwidth
- *
- * @wm: watermark calculation data
- *
- * Calculate the min available bandwidth used for display (CIK).
- * Used for display watermark bandwidth calculations
- * Returns the min available bandwidth in MBytes/s
- */
-static u32 dce_v11_0_available_bandwidth(struct dce10_wm_params *wm)
-{
-	/* Calculate the Available bandwidth. Display can use this temporarily but not in average. */
-	u32 dram_bandwidth = dce_v11_0_dram_bandwidth(wm);
-	u32 data_return_bandwidth = dce_v11_0_data_return_bandwidth(wm);
-	u32 dmif_req_bandwidth = dce_v11_0_dmif_request_bandwidth(wm);
-
-	return min(dram_bandwidth, min(data_return_bandwidth, dmif_req_bandwidth));
-}
-
-/**
- * dce_v11_0_average_bandwidth - get the average available bandwidth
- *
- * @wm: watermark calculation data
- *
- * Calculate the average available bandwidth used for display (CIK).
- * Used for display watermark bandwidth calculations
- * Returns the average available bandwidth in MBytes/s
- */
-static u32 dce_v11_0_average_bandwidth(struct dce10_wm_params *wm)
-{
-	/* Calculate the display mode Average Bandwidth
-	 * DisplayMode should contain the source and destination dimensions,
-	 * timing, etc.
-	 */
-	fixed20_12 bpp;
-	fixed20_12 line_time;
-	fixed20_12 src_width;
-	fixed20_12 bandwidth;
-	fixed20_12 a;
-
-	a.full = dfixed_const(1000);
-	line_time.full = dfixed_const(wm->active_time + wm->blank_time);
-	line_time.full = dfixed_div(line_time, a);
-	bpp.full = dfixed_const(wm->bytes_per_pixel);
-	src_width.full = dfixed_const(wm->src_width);
-	bandwidth.full = dfixed_mul(src_width, bpp);
-	bandwidth.full = dfixed_mul(bandwidth, wm->vsc);
-	bandwidth.full = dfixed_div(bandwidth, line_time);
-
-	return dfixed_trunc(bandwidth);
-}
-
-/**
- * dce_v11_0_latency_watermark - get the latency watermark
- *
- * @wm: watermark calculation data
- *
- * Calculate the latency watermark (CIK).
- * Used for display watermark bandwidth calculations
- * Returns the latency watermark in ns
- */
-static u32 dce_v11_0_latency_watermark(struct dce10_wm_params *wm)
-{
-	/* First calculate the latency in ns */
-	u32 mc_latency = 2000; /* 2000 ns. */
-	u32 available_bandwidth = dce_v11_0_available_bandwidth(wm);
-	u32 worst_chunk_return_time = (512 * 8 * 1000) / available_bandwidth;
-	u32 cursor_line_pair_return_time = (128 * 4 * 1000) / available_bandwidth;
-	u32 dc_latency = 40000000 / wm->disp_clk; /* dc pipe latency */
-	u32 other_heads_data_return_time = ((wm->num_heads + 1) * worst_chunk_return_time) +
-		(wm->num_heads * cursor_line_pair_return_time);
-	u32 latency = mc_latency + other_heads_data_return_time + dc_latency;
-	u32 max_src_lines_per_dst_line, lb_fill_bw, line_fill_time;
-	u32 tmp, dmif_size = 12288;
-	fixed20_12 a, b, c;
-
-	if (wm->num_heads == 0)
-		return 0;
-
-	a.full = dfixed_const(2);
-	b.full = dfixed_const(1);
-	if ((wm->vsc.full > a.full) ||
-	    ((wm->vsc.full > b.full) && (wm->vtaps >= 3)) ||
-	    (wm->vtaps >= 5) ||
-	    ((wm->vsc.full >= a.full) && wm->interlaced))
-		max_src_lines_per_dst_line = 4;
-	else
-		max_src_lines_per_dst_line = 2;
-
-	a.full = dfixed_const(available_bandwidth);
-	b.full = dfixed_const(wm->num_heads);
-	a.full = dfixed_div(a, b);
-	tmp = div_u64((u64) dmif_size * (u64) wm->disp_clk, mc_latency + 512);
-	tmp = min(dfixed_trunc(a), tmp);
-
-	lb_fill_bw = min(tmp, wm->disp_clk * wm->bytes_per_pixel / 1000);
-
-	a.full = dfixed_const(max_src_lines_per_dst_line * wm->src_width * wm->bytes_per_pixel);
-	b.full = dfixed_const(1000);
-	c.full = dfixed_const(lb_fill_bw);
-	b.full = dfixed_div(c, b);
-	a.full = dfixed_div(a, b);
-	line_fill_time = dfixed_trunc(a);
-
-	if (line_fill_time < wm->active_time)
-		return latency;
-	else
-		return latency + (line_fill_time - wm->active_time);
-
-}
-
-/**
- * dce_v11_0_average_bandwidth_vs_dram_bandwidth_for_display - check
- * average and available dram bandwidth
- *
- * @wm: watermark calculation data
- *
- * Check if the display average bandwidth fits in the display
- * dram bandwidth (CIK).
- * Used for display watermark bandwidth calculations
- * Returns true if the display fits, false if not.
- */
-static bool dce_v11_0_average_bandwidth_vs_dram_bandwidth_for_display(struct dce10_wm_params *wm)
-{
-	if (dce_v11_0_average_bandwidth(wm) <=
-	    (dce_v11_0_dram_bandwidth_for_display(wm) / wm->num_heads))
-		return true;
-	else
-		return false;
-}
-
-/**
- * dce_v11_0_average_bandwidth_vs_available_bandwidth - check
- * average and available bandwidth
- *
- * @wm: watermark calculation data
- *
- * Check if the display average bandwidth fits in the display
- * available bandwidth (CIK).
- * Used for display watermark bandwidth calculations
- * Returns true if the display fits, false if not.
- */
-static bool dce_v11_0_average_bandwidth_vs_available_bandwidth(struct dce10_wm_params *wm)
-{
-	if (dce_v11_0_average_bandwidth(wm) <=
-	    (dce_v11_0_available_bandwidth(wm) / wm->num_heads))
-		return true;
-	else
-		return false;
-}
-
-/**
- * dce_v11_0_check_latency_hiding - check latency hiding
- *
- * @wm: watermark calculation data
- *
- * Check latency hiding (CIK).
- * Used for display watermark bandwidth calculations
- * Returns true if the display fits, false if not.
- */
-static bool dce_v11_0_check_latency_hiding(struct dce10_wm_params *wm)
-{
-	u32 lb_partitions = wm->lb_size / wm->src_width;
-	u32 line_time = wm->active_time + wm->blank_time;
-	u32 latency_tolerant_lines;
-	u32 latency_hiding;
-	fixed20_12 a;
-
-	a.full = dfixed_const(1);
-	if (wm->vsc.full > a.full)
-		latency_tolerant_lines = 1;
-	else {
-		if (lb_partitions <= (wm->vtaps + 1))
-			latency_tolerant_lines = 1;
-		else
-			latency_tolerant_lines = 2;
-	}
-
-	latency_hiding = (latency_tolerant_lines * line_time + wm->blank_time);
-
-	if (dce_v11_0_latency_watermark(wm) <= latency_hiding)
-		return true;
-	else
-		return false;
-}
-
-/**
- * dce_v11_0_program_watermarks - program display watermarks
- *
- * @adev: amdgpu_device pointer
- * @amdgpu_crtc: the selected display controller
- * @lb_size: line buffer size
- * @num_heads: number of display controllers in use
- *
- * Calculate and program the display watermarks for the
- * selected display controller (CIK).
- */
-static void dce_v11_0_program_watermarks(struct amdgpu_device *adev,
-					struct amdgpu_crtc *amdgpu_crtc,
-					u32 lb_size, u32 num_heads)
-{
-	struct drm_display_mode *mode = &amdgpu_crtc->base.mode;
-	struct dce10_wm_params wm_low, wm_high;
-	u32 active_time;
-	u32 line_time = 0;
-	u32 latency_watermark_a = 0, latency_watermark_b = 0;
-	u32 tmp, wm_mask, lb_vblank_lead_lines = 0;
-
-	if (amdgpu_crtc->base.enabled && num_heads && mode) {
-		active_time = (u32) div_u64((u64)mode->crtc_hdisplay * 1000000,
-					    (u32)mode->clock);
-		line_time = (u32) div_u64((u64)mode->crtc_htotal * 1000000,
-					  (u32)mode->clock);
-		line_time = min(line_time, (u32)65535);
-
-		/* watermark for high clocks */
-		if (adev->pm.dpm_enabled) {
-			wm_high.yclk =
-				amdgpu_dpm_get_mclk(adev, false) * 10;
-			wm_high.sclk =
-				amdgpu_dpm_get_sclk(adev, false) * 10;
-		} else {
-			wm_high.yclk = adev->pm.current_mclk * 10;
-			wm_high.sclk = adev->pm.current_sclk * 10;
-		}
-
-		wm_high.disp_clk = mode->clock;
-		wm_high.src_width = mode->crtc_hdisplay;
-		wm_high.active_time = active_time;
-		wm_high.blank_time = line_time - wm_high.active_time;
-		wm_high.interlaced = false;
-		if (mode->flags & DRM_MODE_FLAG_INTERLACE)
-			wm_high.interlaced = true;
-		wm_high.vsc = amdgpu_crtc->vsc;
-		wm_high.vtaps = 1;
-		if (amdgpu_crtc->rmx_type != RMX_OFF)
-			wm_high.vtaps = 2;
-		wm_high.bytes_per_pixel = 4; /* XXX: get this from fb config */
-		wm_high.lb_size = lb_size;
-		wm_high.dram_channels = cik_get_number_of_dram_channels(adev);
-		wm_high.num_heads = num_heads;
-
-		/* set for high clocks */
-		latency_watermark_a = min(dce_v11_0_latency_watermark(&wm_high), (u32)65535);
-
-		/* possibly force display priority to high */
-		/* should really do this at mode validation time... */
-		if (!dce_v11_0_average_bandwidth_vs_dram_bandwidth_for_display(&wm_high) ||
-		    !dce_v11_0_average_bandwidth_vs_available_bandwidth(&wm_high) ||
-		    !dce_v11_0_check_latency_hiding(&wm_high) ||
-		    (adev->mode_info.disp_priority == 2)) {
-			DRM_DEBUG_KMS("force priority to high\n");
-		}
-
-		/* watermark for low clocks */
-		if (adev->pm.dpm_enabled) {
-			wm_low.yclk =
-				amdgpu_dpm_get_mclk(adev, true) * 10;
-			wm_low.sclk =
-				amdgpu_dpm_get_sclk(adev, true) * 10;
-		} else {
-			wm_low.yclk = adev->pm.current_mclk * 10;
-			wm_low.sclk = adev->pm.current_sclk * 10;
-		}
-
-		wm_low.disp_clk = mode->clock;
-		wm_low.src_width = mode->crtc_hdisplay;
-		wm_low.active_time = active_time;
-		wm_low.blank_time = line_time - wm_low.active_time;
-		wm_low.interlaced = false;
-		if (mode->flags & DRM_MODE_FLAG_INTERLACE)
-			wm_low.interlaced = true;
-		wm_low.vsc = amdgpu_crtc->vsc;
-		wm_low.vtaps = 1;
-		if (amdgpu_crtc->rmx_type != RMX_OFF)
-			wm_low.vtaps = 2;
-		wm_low.bytes_per_pixel = 4; /* XXX: get this from fb config */
-		wm_low.lb_size = lb_size;
-		wm_low.dram_channels = cik_get_number_of_dram_channels(adev);
-		wm_low.num_heads = num_heads;
-
-		/* set for low clocks */
-		latency_watermark_b = min(dce_v11_0_latency_watermark(&wm_low), (u32)65535);
-
-		/* possibly force display priority to high */
-		/* should really do this at mode validation time... */
-		if (!dce_v11_0_average_bandwidth_vs_dram_bandwidth_for_display(&wm_low) ||
-		    !dce_v11_0_average_bandwidth_vs_available_bandwidth(&wm_low) ||
-		    !dce_v11_0_check_latency_hiding(&wm_low) ||
-		    (adev->mode_info.disp_priority == 2)) {
-			DRM_DEBUG_KMS("force priority to high\n");
-		}
-		lb_vblank_lead_lines = DIV_ROUND_UP(lb_size, mode->crtc_hdisplay);
-	}
-
-	/* select wm A */
-	wm_mask = RREG32(mmDPG_WATERMARK_MASK_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(wm_mask, DPG_WATERMARK_MASK_CONTROL, URGENCY_WATERMARK_MASK, 1);
-	WREG32(mmDPG_WATERMARK_MASK_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-	tmp = RREG32(mmDPG_PIPE_URGENCY_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, DPG_PIPE_URGENCY_CONTROL, URGENCY_LOW_WATERMARK, latency_watermark_a);
-	tmp = REG_SET_FIELD(tmp, DPG_PIPE_URGENCY_CONTROL, URGENCY_HIGH_WATERMARK, line_time);
-	WREG32(mmDPG_PIPE_URGENCY_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-	/* select wm B */
-	tmp = REG_SET_FIELD(wm_mask, DPG_WATERMARK_MASK_CONTROL, URGENCY_WATERMARK_MASK, 2);
-	WREG32(mmDPG_WATERMARK_MASK_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-	tmp = RREG32(mmDPG_PIPE_URGENCY_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, DPG_PIPE_URGENCY_CONTROL, URGENCY_LOW_WATERMARK, latency_watermark_b);
-	tmp = REG_SET_FIELD(tmp, DPG_PIPE_URGENCY_CONTROL, URGENCY_HIGH_WATERMARK, line_time);
-	WREG32(mmDPG_PIPE_URGENCY_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-	/* restore original selection */
-	WREG32(mmDPG_WATERMARK_MASK_CONTROL + amdgpu_crtc->crtc_offset, wm_mask);
-
-	/* save values for DPM */
-	amdgpu_crtc->line_time = line_time;
-	amdgpu_crtc->wm_high = latency_watermark_a;
-	amdgpu_crtc->wm_low = latency_watermark_b;
-	/* Save number of lines the linebuffer leads before the scanout */
-	amdgpu_crtc->lb_vblank_lead_lines = lb_vblank_lead_lines;
-}
-
-/**
- * dce_v11_0_bandwidth_update - program display watermarks
- *
- * @adev: amdgpu_device pointer
- *
- * Calculate and program the display watermarks and line
- * buffer allocation (CIK).
- */
-static void dce_v11_0_bandwidth_update(struct amdgpu_device *adev)
-{
-	struct drm_display_mode *mode = NULL;
-	u32 num_heads = 0, lb_size;
-	int i;
-
-	amdgpu_display_update_priority(adev);
-
-	for (i = 0; i < adev->mode_info.num_crtc; i++) {
-		if (adev->mode_info.crtcs[i]->base.enabled)
-			num_heads++;
-	}
-	for (i = 0; i < adev->mode_info.num_crtc; i++) {
-		mode = &adev->mode_info.crtcs[i]->base.mode;
-		lb_size = dce_v11_0_line_buffer_adjust(adev, adev->mode_info.crtcs[i], mode);
-		dce_v11_0_program_watermarks(adev, adev->mode_info.crtcs[i],
-					    lb_size, num_heads);
-	}
-}
-
-static void dce_v11_0_audio_get_connected_pins(struct amdgpu_device *adev)
-{
-	int i;
-	u32 offset, tmp;
-
-	for (i = 0; i < adev->mode_info.audio.num_pins; i++) {
-		offset = adev->mode_info.audio.pin[i].offset;
-		tmp = RREG32_AUDIO_ENDPT(offset,
-					 ixAZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_CONFIGURATION_DEFAULT);
-		if (((tmp &
-		AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_CONFIGURATION_DEFAULT__PORT_CONNECTIVITY_MASK) >>
-		AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_CONFIGURATION_DEFAULT__PORT_CONNECTIVITY__SHIFT) == 1)
-			adev->mode_info.audio.pin[i].connected = false;
-		else
-			adev->mode_info.audio.pin[i].connected = true;
-	}
-}
-
-static struct amdgpu_audio_pin *dce_v11_0_audio_get_pin(struct amdgpu_device *adev)
-{
-	int i;
-
-	dce_v11_0_audio_get_connected_pins(adev);
-
-	for (i = 0; i < adev->mode_info.audio.num_pins; i++) {
-		if (adev->mode_info.audio.pin[i].connected)
-			return &adev->mode_info.audio.pin[i];
-	}
-	DRM_ERROR("No connected audio pins found!\n");
-	return NULL;
-}
-
-static void dce_v11_0_afmt_audio_select_pin(struct drm_encoder *encoder)
-{
-	struct amdgpu_device *adev = drm_to_adev(encoder->dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-	u32 tmp;
-
-	if (!dig || !dig->afmt || !dig->afmt->pin)
-		return;
-
-	tmp = RREG32(mmAFMT_AUDIO_SRC_CONTROL + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, AFMT_AUDIO_SRC_CONTROL, AFMT_AUDIO_SRC_SELECT, dig->afmt->pin->id);
-	WREG32(mmAFMT_AUDIO_SRC_CONTROL + dig->afmt->offset, tmp);
-}
-
-static void dce_v11_0_audio_write_latency_fields(struct drm_encoder *encoder,
-						struct drm_display_mode *mode)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-	struct drm_connector *connector;
-	struct drm_connector_list_iter iter;
-	struct amdgpu_connector *amdgpu_connector = NULL;
-	u32 tmp;
-	int interlace = 0;
-
-	if (!dig || !dig->afmt || !dig->afmt->pin)
-		return;
-
-	drm_connector_list_iter_begin(dev, &iter);
-	drm_for_each_connector_iter(connector, &iter) {
-		if (connector->encoder == encoder) {
-			amdgpu_connector = to_amdgpu_connector(connector);
-			break;
-		}
-	}
-	drm_connector_list_iter_end(&iter);
-
-	if (!amdgpu_connector) {
-		DRM_ERROR("Couldn't find encoder's connector\n");
-		return;
-	}
-
-	if (mode->flags & DRM_MODE_FLAG_INTERLACE)
-		interlace = 1;
-	if (connector->latency_present[interlace]) {
-		tmp = REG_SET_FIELD(0, AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC,
-				    VIDEO_LIPSYNC, connector->video_latency[interlace]);
-		tmp = REG_SET_FIELD(0, AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC,
-				    AUDIO_LIPSYNC, connector->audio_latency[interlace]);
-	} else {
-		tmp = REG_SET_FIELD(0, AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC,
-				    VIDEO_LIPSYNC, 0);
-		tmp = REG_SET_FIELD(0, AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC,
-				    AUDIO_LIPSYNC, 0);
-	}
-	WREG32_AUDIO_ENDPT(dig->afmt->pin->offset,
-			   ixAZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC, tmp);
-}
-
-static void dce_v11_0_audio_write_speaker_allocation(struct drm_encoder *encoder)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-	struct drm_connector *connector;
-	struct drm_connector_list_iter iter;
-	struct amdgpu_connector *amdgpu_connector = NULL;
-	u32 tmp;
-	u8 *sadb = NULL;
-	int sad_count;
-
-	if (!dig || !dig->afmt || !dig->afmt->pin)
-		return;
-
-	drm_connector_list_iter_begin(dev, &iter);
-	drm_for_each_connector_iter(connector, &iter) {
-		if (connector->encoder == encoder) {
-			amdgpu_connector = to_amdgpu_connector(connector);
-			break;
-		}
-	}
-	drm_connector_list_iter_end(&iter);
-
-	if (!amdgpu_connector) {
-		DRM_ERROR("Couldn't find encoder's connector\n");
-		return;
-	}
-
-	sad_count = drm_edid_to_speaker_allocation(amdgpu_connector_edid(connector), &sadb);
-	if (sad_count < 0) {
-		DRM_ERROR("Couldn't read Speaker Allocation Data Block: %d\n", sad_count);
-		sad_count = 0;
-	}
-
-	/* program the speaker allocation */
-	tmp = RREG32_AUDIO_ENDPT(dig->afmt->pin->offset,
-				 ixAZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER);
-	tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER,
-			    DP_CONNECTION, 0);
-	/* set HDMI mode */
-	tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER,
-			    HDMI_CONNECTION, 1);
-	if (sad_count)
-		tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER,
-				    SPEAKER_ALLOCATION, sadb[0]);
-	else
-		tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER,
-				    SPEAKER_ALLOCATION, 5); /* stereo */
-	WREG32_AUDIO_ENDPT(dig->afmt->pin->offset,
-			   ixAZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER, tmp);
-
-	kfree(sadb);
-}
-
-static void dce_v11_0_audio_write_sad_regs(struct drm_encoder *encoder)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-	struct drm_connector *connector;
-	struct drm_connector_list_iter iter;
-	struct amdgpu_connector *amdgpu_connector = NULL;
-	struct cea_sad *sads;
-	int i, sad_count;
-
-	static const u16 eld_reg_to_type[][2] = {
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0, HDMI_AUDIO_CODING_TYPE_PCM },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR1, HDMI_AUDIO_CODING_TYPE_AC3 },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR2, HDMI_AUDIO_CODING_TYPE_MPEG1 },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR3, HDMI_AUDIO_CODING_TYPE_MP3 },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR4, HDMI_AUDIO_CODING_TYPE_MPEG2 },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR5, HDMI_AUDIO_CODING_TYPE_AAC_LC },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR6, HDMI_AUDIO_CODING_TYPE_DTS },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR7, HDMI_AUDIO_CODING_TYPE_ATRAC },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR9, HDMI_AUDIO_CODING_TYPE_EAC3 },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR10, HDMI_AUDIO_CODING_TYPE_DTS_HD },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR11, HDMI_AUDIO_CODING_TYPE_MLP },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR13, HDMI_AUDIO_CODING_TYPE_WMA_PRO },
-	};
-
-	if (!dig || !dig->afmt || !dig->afmt->pin)
-		return;
-
-	drm_connector_list_iter_begin(dev, &iter);
-	drm_for_each_connector_iter(connector, &iter) {
-		if (connector->encoder == encoder) {
-			amdgpu_connector = to_amdgpu_connector(connector);
-			break;
-		}
-	}
-	drm_connector_list_iter_end(&iter);
-
-	if (!amdgpu_connector) {
-		DRM_ERROR("Couldn't find encoder's connector\n");
-		return;
-	}
-
-	sad_count = drm_edid_to_sad(amdgpu_connector_edid(connector), &sads);
-	if (sad_count < 0)
-		DRM_ERROR("Couldn't read SADs: %d\n", sad_count);
-	if (sad_count <= 0)
-		return;
-	BUG_ON(!sads);
-
-	for (i = 0; i < ARRAY_SIZE(eld_reg_to_type); i++) {
-		u32 tmp = 0;
-		u8 stereo_freqs = 0;
-		int max_channels = -1;
-		int j;
-
-		for (j = 0; j < sad_count; j++) {
-			struct cea_sad *sad = &sads[j];
-
-			if (sad->format == eld_reg_to_type[i][1]) {
-				if (sad->channels > max_channels) {
-					tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0,
-							    MAX_CHANNELS, sad->channels);
-					tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0,
-							    DESCRIPTOR_BYTE_2, sad->byte2);
-					tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0,
-							    SUPPORTED_FREQUENCIES, sad->freq);
-					max_channels = sad->channels;
-				}
-
-				if (sad->format == HDMI_AUDIO_CODING_TYPE_PCM)
-					stereo_freqs |= sad->freq;
-				else
-					break;
-			}
-		}
-
-		tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0,
-				    SUPPORTED_FREQUENCIES_STEREO, stereo_freqs);
-		WREG32_AUDIO_ENDPT(dig->afmt->pin->offset, eld_reg_to_type[i][0], tmp);
-	}
-
-	kfree(sads);
-}
-
-static void dce_v11_0_audio_enable(struct amdgpu_device *adev,
-				  struct amdgpu_audio_pin *pin,
-				  bool enable)
-{
-	if (!pin)
-		return;
-
-	WREG32_AUDIO_ENDPT(pin->offset, ixAZALIA_F0_CODEC_PIN_CONTROL_HOT_PLUG_CONTROL,
-			   enable ? AZALIA_F0_CODEC_PIN_CONTROL_HOT_PLUG_CONTROL__AUDIO_ENABLED_MASK : 0);
-}
-
-static const u32 pin_offsets[] =
-{
-	AUD0_REGISTER_OFFSET,
-	AUD1_REGISTER_OFFSET,
-	AUD2_REGISTER_OFFSET,
-	AUD3_REGISTER_OFFSET,
-	AUD4_REGISTER_OFFSET,
-	AUD5_REGISTER_OFFSET,
-	AUD6_REGISTER_OFFSET,
-	AUD7_REGISTER_OFFSET,
-};
-
-static int dce_v11_0_audio_init(struct amdgpu_device *adev)
-{
-	int i;
-
-	if (!amdgpu_audio)
-		return 0;
-
-	adev->mode_info.audio.enabled = true;
-
-	switch (adev->asic_type) {
-	case CHIP_CARRIZO:
-	case CHIP_STONEY:
-		adev->mode_info.audio.num_pins = 7;
-		break;
-	case CHIP_POLARIS10:
-	case CHIP_VEGAM:
-		adev->mode_info.audio.num_pins = 8;
-		break;
-	case CHIP_POLARIS11:
-	case CHIP_POLARIS12:
-		adev->mode_info.audio.num_pins = 6;
-		break;
-	default:
-		return -EINVAL;
-	}
-
-	for (i = 0; i < adev->mode_info.audio.num_pins; i++) {
-		adev->mode_info.audio.pin[i].channels = -1;
-		adev->mode_info.audio.pin[i].rate = -1;
-		adev->mode_info.audio.pin[i].bits_per_sample = -1;
-		adev->mode_info.audio.pin[i].status_bits = 0;
-		adev->mode_info.audio.pin[i].category_code = 0;
-		adev->mode_info.audio.pin[i].connected = false;
-		adev->mode_info.audio.pin[i].offset = pin_offsets[i];
-		adev->mode_info.audio.pin[i].id = i;
-		/* disable audio.  it will be set up later */
-		/* XXX remove once we switch to ip funcs */
-		dce_v11_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
-	}
-
-	return 0;
-}
-
-static void dce_v11_0_audio_fini(struct amdgpu_device *adev)
-{
-	int i;
-
-	if (!amdgpu_audio)
-		return;
-
-	if (!adev->mode_info.audio.enabled)
-		return;
-
-	for (i = 0; i < adev->mode_info.audio.num_pins; i++)
-		dce_v11_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
-
-	adev->mode_info.audio.enabled = false;
-}
-
-/*
- * update the N and CTS parameters for a given pixel clock rate
- */
-static void dce_v11_0_afmt_update_ACR(struct drm_encoder *encoder, uint32_t clock)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_afmt_acr acr = amdgpu_afmt_acr(clock);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-	u32 tmp;
-
-	tmp = RREG32(mmHDMI_ACR_32_0 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, HDMI_ACR_32_0, HDMI_ACR_CTS_32, acr.cts_32khz);
-	WREG32(mmHDMI_ACR_32_0 + dig->afmt->offset, tmp);
-	tmp = RREG32(mmHDMI_ACR_32_1 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, HDMI_ACR_32_1, HDMI_ACR_N_32, acr.n_32khz);
-	WREG32(mmHDMI_ACR_32_1 + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmHDMI_ACR_44_0 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, HDMI_ACR_44_0, HDMI_ACR_CTS_44, acr.cts_44_1khz);
-	WREG32(mmHDMI_ACR_44_0 + dig->afmt->offset, tmp);
-	tmp = RREG32(mmHDMI_ACR_44_1 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, HDMI_ACR_44_1, HDMI_ACR_N_44, acr.n_44_1khz);
-	WREG32(mmHDMI_ACR_44_1 + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmHDMI_ACR_48_0 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, HDMI_ACR_48_0, HDMI_ACR_CTS_48, acr.cts_48khz);
-	WREG32(mmHDMI_ACR_48_0 + dig->afmt->offset, tmp);
-	tmp = RREG32(mmHDMI_ACR_48_1 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, HDMI_ACR_48_1, HDMI_ACR_N_48, acr.n_48khz);
-	WREG32(mmHDMI_ACR_48_1 + dig->afmt->offset, tmp);
-
-}
-
-/*
- * build a HDMI Video Info Frame
- */
-static void dce_v11_0_afmt_update_avi_infoframe(struct drm_encoder *encoder,
-					       void *buffer, size_t size)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-	uint8_t *frame = buffer + 3;
-	uint8_t *header = buffer;
-
-	WREG32(mmAFMT_AVI_INFO0 + dig->afmt->offset,
-		frame[0x0] | (frame[0x1] << 8) | (frame[0x2] << 16) | (frame[0x3] << 24));
-	WREG32(mmAFMT_AVI_INFO1 + dig->afmt->offset,
-		frame[0x4] | (frame[0x5] << 8) | (frame[0x6] << 16) | (frame[0x7] << 24));
-	WREG32(mmAFMT_AVI_INFO2 + dig->afmt->offset,
-		frame[0x8] | (frame[0x9] << 8) | (frame[0xA] << 16) | (frame[0xB] << 24));
-	WREG32(mmAFMT_AVI_INFO3 + dig->afmt->offset,
-		frame[0xC] | (frame[0xD] << 8) | (header[1] << 24));
-}
-
-static void dce_v11_0_audio_set_dto(struct drm_encoder *encoder, u32 clock)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(encoder->crtc);
-	u32 dto_phase = 24 * 1000;
-	u32 dto_modulo = clock;
-	u32 tmp;
-
-	if (!dig || !dig->afmt)
-		return;
-
-	/* XXX two dtos; generally use dto0 for hdmi */
-	/* Express [24MHz / target pixel clock] as an exact rational
-	 * number (coefficient of two integer numbers.  DCCG_AUDIO_DTOx_PHASE
-	 * is the numerator, DCCG_AUDIO_DTOx_MODULE is the denominator
-	 */
-	tmp = RREG32(mmDCCG_AUDIO_DTO_SOURCE);
-	tmp = REG_SET_FIELD(tmp, DCCG_AUDIO_DTO_SOURCE, DCCG_AUDIO_DTO0_SOURCE_SEL,
-			    amdgpu_crtc->crtc_id);
-	WREG32(mmDCCG_AUDIO_DTO_SOURCE, tmp);
-	WREG32(mmDCCG_AUDIO_DTO0_PHASE, dto_phase);
-	WREG32(mmDCCG_AUDIO_DTO0_MODULE, dto_modulo);
-}
-
-/*
- * update the info frames with the data from the current display mode
- */
-static void dce_v11_0_afmt_setmode(struct drm_encoder *encoder,
-				  struct drm_display_mode *mode)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-	struct drm_connector *connector = amdgpu_get_connector_for_encoder(encoder);
-	u8 buffer[HDMI_INFOFRAME_HEADER_SIZE + HDMI_AVI_INFOFRAME_SIZE];
-	struct hdmi_avi_infoframe frame;
-	ssize_t err;
-	u32 tmp;
-	int bpc = 8;
-
-	if (!dig || !dig->afmt)
-		return;
-
-	/* Silent, r600_hdmi_enable will raise WARN for us */
-	if (!dig->afmt->enabled)
-		return;
-
-	/* hdmi deep color mode general control packets setup, if bpc > 8 */
-	if (encoder->crtc) {
-		struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(encoder->crtc);
-		bpc = amdgpu_crtc->bpc;
-	}
-
-	/* disable audio prior to setting up hw */
-	dig->afmt->pin = dce_v11_0_audio_get_pin(adev);
-	dce_v11_0_audio_enable(adev, dig->afmt->pin, false);
-
-	dce_v11_0_audio_set_dto(encoder, mode->clock);
-
-	tmp = RREG32(mmHDMI_VBI_PACKET_CONTROL + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, HDMI_VBI_PACKET_CONTROL, HDMI_NULL_SEND, 1);
-	WREG32(mmHDMI_VBI_PACKET_CONTROL + dig->afmt->offset, tmp); /* send null packets when required */
-
-	WREG32(mmAFMT_AUDIO_CRC_CONTROL + dig->afmt->offset, 0x1000);
-
-	tmp = RREG32(mmHDMI_CONTROL + dig->afmt->offset);
-	switch (bpc) {
-	case 0:
-	case 6:
-	case 8:
-	case 16:
-	default:
-		tmp = REG_SET_FIELD(tmp, HDMI_CONTROL, HDMI_DEEP_COLOR_ENABLE, 0);
-		tmp = REG_SET_FIELD(tmp, HDMI_CONTROL, HDMI_DEEP_COLOR_DEPTH, 0);
-		DRM_DEBUG("%s: Disabling hdmi deep color for %d bpc.\n",
-			  connector->name, bpc);
-		break;
-	case 10:
-		tmp = REG_SET_FIELD(tmp, HDMI_CONTROL, HDMI_DEEP_COLOR_ENABLE, 1);
-		tmp = REG_SET_FIELD(tmp, HDMI_CONTROL, HDMI_DEEP_COLOR_DEPTH, 1);
-		DRM_DEBUG("%s: Enabling hdmi deep color 30 for 10 bpc.\n",
-			  connector->name);
-		break;
-	case 12:
-		tmp = REG_SET_FIELD(tmp, HDMI_CONTROL, HDMI_DEEP_COLOR_ENABLE, 1);
-		tmp = REG_SET_FIELD(tmp, HDMI_CONTROL, HDMI_DEEP_COLOR_DEPTH, 2);
-		DRM_DEBUG("%s: Enabling hdmi deep color 36 for 12 bpc.\n",
-			  connector->name);
-		break;
-	}
-	WREG32(mmHDMI_CONTROL + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmHDMI_VBI_PACKET_CONTROL + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, HDMI_VBI_PACKET_CONTROL, HDMI_NULL_SEND, 1); /* send null packets when required */
-	tmp = REG_SET_FIELD(tmp, HDMI_VBI_PACKET_CONTROL, HDMI_GC_SEND, 1); /* send general control packets */
-	tmp = REG_SET_FIELD(tmp, HDMI_VBI_PACKET_CONTROL, HDMI_GC_CONT, 1); /* send general control packets every frame */
-	WREG32(mmHDMI_VBI_PACKET_CONTROL + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmHDMI_INFOFRAME_CONTROL0 + dig->afmt->offset);
-	/* enable audio info frames (frames won't be set until audio is enabled) */
-	tmp = REG_SET_FIELD(tmp, HDMI_INFOFRAME_CONTROL0, HDMI_AUDIO_INFO_SEND, 1);
-	/* required for audio info values to be updated */
-	tmp = REG_SET_FIELD(tmp, HDMI_INFOFRAME_CONTROL0, HDMI_AUDIO_INFO_CONT, 1);
-	WREG32(mmHDMI_INFOFRAME_CONTROL0 + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmAFMT_INFOFRAME_CONTROL0 + dig->afmt->offset);
-	/* required for audio info values to be updated */
-	tmp = REG_SET_FIELD(tmp, AFMT_INFOFRAME_CONTROL0, AFMT_AUDIO_INFO_UPDATE, 1);
-	WREG32(mmAFMT_INFOFRAME_CONTROL0 + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmHDMI_INFOFRAME_CONTROL1 + dig->afmt->offset);
-	/* anything other than 0 */
-	tmp = REG_SET_FIELD(tmp, HDMI_INFOFRAME_CONTROL1, HDMI_AUDIO_INFO_LINE, 2);
-	WREG32(mmHDMI_INFOFRAME_CONTROL1 + dig->afmt->offset, tmp);
-
-	WREG32(mmHDMI_GC + dig->afmt->offset, 0); /* unset HDMI_GC_AVMUTE */
-
-	tmp = RREG32(mmHDMI_AUDIO_PACKET_CONTROL + dig->afmt->offset);
-	/* set the default audio delay */
-	tmp = REG_SET_FIELD(tmp, HDMI_AUDIO_PACKET_CONTROL, HDMI_AUDIO_DELAY_EN, 1);
-	/* should be suffient for all audio modes and small enough for all hblanks */
-	tmp = REG_SET_FIELD(tmp, HDMI_AUDIO_PACKET_CONTROL, HDMI_AUDIO_PACKETS_PER_LINE, 3);
-	WREG32(mmHDMI_AUDIO_PACKET_CONTROL + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmAFMT_AUDIO_PACKET_CONTROL + dig->afmt->offset);
-	/* allow 60958 channel status fields to be updated */
-	tmp = REG_SET_FIELD(tmp, AFMT_AUDIO_PACKET_CONTROL, AFMT_60958_CS_UPDATE, 1);
-	WREG32(mmAFMT_AUDIO_PACKET_CONTROL + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmHDMI_ACR_PACKET_CONTROL + dig->afmt->offset);
-	if (bpc > 8)
-		/* clear SW CTS value */
-		tmp = REG_SET_FIELD(tmp, HDMI_ACR_PACKET_CONTROL, HDMI_ACR_SOURCE, 0);
-	else
-		/* select SW CTS value */
-		tmp = REG_SET_FIELD(tmp, HDMI_ACR_PACKET_CONTROL, HDMI_ACR_SOURCE, 1);
-	/* allow hw to sent ACR packets when required */
-	tmp = REG_SET_FIELD(tmp, HDMI_ACR_PACKET_CONTROL, HDMI_ACR_AUTO_SEND, 1);
-	WREG32(mmHDMI_ACR_PACKET_CONTROL + dig->afmt->offset, tmp);
-
-	dce_v11_0_afmt_update_ACR(encoder, mode->clock);
-
-	tmp = RREG32(mmAFMT_60958_0 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, AFMT_60958_0, AFMT_60958_CS_CHANNEL_NUMBER_L, 1);
-	WREG32(mmAFMT_60958_0 + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmAFMT_60958_1 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, AFMT_60958_1, AFMT_60958_CS_CHANNEL_NUMBER_R, 2);
-	WREG32(mmAFMT_60958_1 + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmAFMT_60958_2 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, AFMT_60958_2, AFMT_60958_CS_CHANNEL_NUMBER_2, 3);
-	tmp = REG_SET_FIELD(tmp, AFMT_60958_2, AFMT_60958_CS_CHANNEL_NUMBER_3, 4);
-	tmp = REG_SET_FIELD(tmp, AFMT_60958_2, AFMT_60958_CS_CHANNEL_NUMBER_4, 5);
-	tmp = REG_SET_FIELD(tmp, AFMT_60958_2, AFMT_60958_CS_CHANNEL_NUMBER_5, 6);
-	tmp = REG_SET_FIELD(tmp, AFMT_60958_2, AFMT_60958_CS_CHANNEL_NUMBER_6, 7);
-	tmp = REG_SET_FIELD(tmp, AFMT_60958_2, AFMT_60958_CS_CHANNEL_NUMBER_7, 8);
-	WREG32(mmAFMT_60958_2 + dig->afmt->offset, tmp);
-
-	dce_v11_0_audio_write_speaker_allocation(encoder);
-
-	WREG32(mmAFMT_AUDIO_PACKET_CONTROL2 + dig->afmt->offset,
-	       (0xff << AFMT_AUDIO_PACKET_CONTROL2__AFMT_AUDIO_CHANNEL_ENABLE__SHIFT));
-
-	dce_v11_0_afmt_audio_select_pin(encoder);
-	dce_v11_0_audio_write_sad_regs(encoder);
-	dce_v11_0_audio_write_latency_fields(encoder, mode);
-
-	err = drm_hdmi_avi_infoframe_from_display_mode(&frame, connector, mode);
-	if (err < 0) {
-		DRM_ERROR("failed to setup AVI infoframe: %zd\n", err);
-		return;
-	}
-
-	err = hdmi_avi_infoframe_pack(&frame, buffer, sizeof(buffer));
-	if (err < 0) {
-		DRM_ERROR("failed to pack AVI infoframe: %zd\n", err);
-		return;
-	}
-
-	dce_v11_0_afmt_update_avi_infoframe(encoder, buffer, sizeof(buffer));
-
-	tmp = RREG32(mmHDMI_INFOFRAME_CONTROL0 + dig->afmt->offset);
-	/* enable AVI info frames */
-	tmp = REG_SET_FIELD(tmp, HDMI_INFOFRAME_CONTROL0, HDMI_AVI_INFO_SEND, 1);
-	/* required for audio info values to be updated */
-	tmp = REG_SET_FIELD(tmp, HDMI_INFOFRAME_CONTROL0, HDMI_AVI_INFO_CONT, 1);
-	WREG32(mmHDMI_INFOFRAME_CONTROL0 + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmHDMI_INFOFRAME_CONTROL1 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, HDMI_INFOFRAME_CONTROL1, HDMI_AVI_INFO_LINE, 2);
-	WREG32(mmHDMI_INFOFRAME_CONTROL1 + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmAFMT_AUDIO_PACKET_CONTROL + dig->afmt->offset);
-	/* send audio packets */
-	tmp = REG_SET_FIELD(tmp, AFMT_AUDIO_PACKET_CONTROL, AFMT_AUDIO_SAMPLE_SEND, 1);
-	WREG32(mmAFMT_AUDIO_PACKET_CONTROL + dig->afmt->offset, tmp);
-
-	WREG32(mmAFMT_RAMP_CONTROL0 + dig->afmt->offset, 0x00FFFFFF);
-	WREG32(mmAFMT_RAMP_CONTROL1 + dig->afmt->offset, 0x007FFFFF);
-	WREG32(mmAFMT_RAMP_CONTROL2 + dig->afmt->offset, 0x00000001);
-	WREG32(mmAFMT_RAMP_CONTROL3 + dig->afmt->offset, 0x00000001);
-
-	/* enable audio after to setting up hw */
-	dce_v11_0_audio_enable(adev, dig->afmt->pin, true);
-}
-
-static void dce_v11_0_afmt_enable(struct drm_encoder *encoder, bool enable)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-
-	if (!dig || !dig->afmt)
-		return;
-
-	/* Silent, r600_hdmi_enable will raise WARN for us */
-	if (enable && dig->afmt->enabled)
-		return;
-	if (!enable && !dig->afmt->enabled)
-		return;
-
-	if (!enable && dig->afmt->pin) {
-		dce_v11_0_audio_enable(adev, dig->afmt->pin, false);
-		dig->afmt->pin = NULL;
-	}
-
-	dig->afmt->enabled = enable;
-
-	DRM_DEBUG("%sabling AFMT interface @ 0x%04X for encoder 0x%x\n",
-		  enable ? "En" : "Dis", dig->afmt->offset, amdgpu_encoder->encoder_id);
-}
-
-static int dce_v11_0_afmt_init(struct amdgpu_device *adev)
-{
-	int i;
-
-	for (i = 0; i < adev->mode_info.num_dig; i++)
-		adev->mode_info.afmt[i] = NULL;
-
-	/* DCE11 has audio blocks tied to DIG encoders */
-	for (i = 0; i < adev->mode_info.num_dig; i++) {
-		adev->mode_info.afmt[i] = kzalloc(sizeof(struct amdgpu_afmt), GFP_KERNEL);
-		if (adev->mode_info.afmt[i]) {
-			adev->mode_info.afmt[i]->offset = dig_offsets[i];
-			adev->mode_info.afmt[i]->id = i;
-		} else {
-			int j;
-			for (j = 0; j < i; j++) {
-				kfree(adev->mode_info.afmt[j]);
-				adev->mode_info.afmt[j] = NULL;
-			}
-			return -ENOMEM;
-		}
-	}
-	return 0;
-}
-
-static void dce_v11_0_afmt_fini(struct amdgpu_device *adev)
-{
-	int i;
-
-	for (i = 0; i < adev->mode_info.num_dig; i++) {
-		kfree(adev->mode_info.afmt[i]);
-		adev->mode_info.afmt[i] = NULL;
-	}
-}
-
-static const u32 vga_control_regs[6] =
-{
-	mmD1VGA_CONTROL,
-	mmD2VGA_CONTROL,
-	mmD3VGA_CONTROL,
-	mmD4VGA_CONTROL,
-	mmD5VGA_CONTROL,
-	mmD6VGA_CONTROL,
-};
-
-static void dce_v11_0_vga_enable(struct drm_crtc *crtc, bool enable)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct drm_device *dev = crtc->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	u32 vga_control;
-
-	vga_control = RREG32(vga_control_regs[amdgpu_crtc->crtc_id]) & ~1;
-	if (enable)
-		WREG32(vga_control_regs[amdgpu_crtc->crtc_id], vga_control | 1);
-	else
-		WREG32(vga_control_regs[amdgpu_crtc->crtc_id], vga_control);
-}
-
-static void dce_v11_0_grph_enable(struct drm_crtc *crtc, bool enable)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct drm_device *dev = crtc->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-
-	if (enable)
-		WREG32(mmGRPH_ENABLE + amdgpu_crtc->crtc_offset, 1);
-	else
-		WREG32(mmGRPH_ENABLE + amdgpu_crtc->crtc_offset, 0);
-}
-
-static int dce_v11_0_crtc_do_set_base(struct drm_crtc *crtc,
-				     struct drm_framebuffer *fb,
-				     int x, int y, int atomic)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct drm_device *dev = crtc->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct drm_framebuffer *target_fb;
-	struct drm_gem_object *obj;
-	struct amdgpu_bo *abo;
-	uint64_t fb_location, tiling_flags;
-	uint32_t fb_format, fb_pitch_pixels;
-	u32 fb_swap = REG_SET_FIELD(0, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP, ENDIAN_NONE);
-	u32 pipe_config;
-	u32 tmp, viewport_w, viewport_h;
-	int r;
-	bool bypass_lut = false;
-
-	/* no fb bound */
-	if (!atomic && !crtc->primary->fb) {
-		DRM_DEBUG_KMS("No FB bound\n");
-		return 0;
-	}
-
-	if (atomic)
-		target_fb = fb;
-	else
-		target_fb = crtc->primary->fb;
-
-	/* If atomic, assume fb object is pinned & idle & fenced and
-	 * just update base pointers
-	 */
-	obj = target_fb->obj[0];
-	abo = gem_to_amdgpu_bo(obj);
-	r = amdgpu_bo_reserve(abo, false);
-	if (unlikely(r != 0))
-		return r;
-
-	if (!atomic) {
-		r = amdgpu_bo_pin(abo, AMDGPU_GEM_DOMAIN_VRAM);
-		if (unlikely(r != 0)) {
-			amdgpu_bo_unreserve(abo);
-			return -EINVAL;
-		}
-	}
-	fb_location = amdgpu_bo_gpu_offset(abo);
-
-	amdgpu_bo_get_tiling_flags(abo, &tiling_flags);
-	amdgpu_bo_unreserve(abo);
-
-	pipe_config = AMDGPU_TILING_GET(tiling_flags, PIPE_CONFIG);
-
-	switch (target_fb->format->format) {
-	case DRM_FORMAT_C8:
-		fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 0);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 0);
-		break;
-	case DRM_FORMAT_XRGB4444:
-	case DRM_FORMAT_ARGB4444:
-		fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 1);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 2);
-#ifdef __BIG_ENDIAN
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP,
-					ENDIAN_8IN16);
-#endif
-		break;
-	case DRM_FORMAT_XRGB1555:
-	case DRM_FORMAT_ARGB1555:
-		fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 1);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 0);
-#ifdef __BIG_ENDIAN
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP,
-					ENDIAN_8IN16);
-#endif
-		break;
-	case DRM_FORMAT_BGRX5551:
-	case DRM_FORMAT_BGRA5551:
-		fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 1);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 5);
-#ifdef __BIG_ENDIAN
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP,
-					ENDIAN_8IN16);
-#endif
-		break;
-	case DRM_FORMAT_RGB565:
-		fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 1);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 1);
-#ifdef __BIG_ENDIAN
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP,
-					ENDIAN_8IN16);
-#endif
-		break;
-	case DRM_FORMAT_XRGB8888:
-	case DRM_FORMAT_ARGB8888:
-		fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 2);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 0);
-#ifdef __BIG_ENDIAN
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP,
-					ENDIAN_8IN32);
-#endif
-		break;
-	case DRM_FORMAT_XRGB2101010:
-	case DRM_FORMAT_ARGB2101010:
-		fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 2);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 1);
-#ifdef __BIG_ENDIAN
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP,
-					ENDIAN_8IN32);
-#endif
-		/* Greater 8 bpc fb needs to bypass hw-lut to retain precision */
-		bypass_lut = true;
-		break;
-	case DRM_FORMAT_BGRX1010102:
-	case DRM_FORMAT_BGRA1010102:
-		fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 2);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 4);
-#ifdef __BIG_ENDIAN
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP,
-					ENDIAN_8IN32);
-#endif
-		/* Greater 8 bpc fb needs to bypass hw-lut to retain precision */
-		bypass_lut = true;
-		break;
-	case DRM_FORMAT_XBGR8888:
-	case DRM_FORMAT_ABGR8888:
-		fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 2);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 0);
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_RED_CROSSBAR, 2);
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_BLUE_CROSSBAR, 2);
-#ifdef __BIG_ENDIAN
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP,
-					ENDIAN_8IN32);
-#endif
-		break;
-	default:
-		DRM_ERROR("Unsupported screen format %p4cc\n",
-			  &target_fb->format->format);
-		return -EINVAL;
-	}
-
-	if (AMDGPU_TILING_GET(tiling_flags, ARRAY_MODE) == ARRAY_2D_TILED_THIN1) {
-		unsigned bankw, bankh, mtaspect, tile_split, num_banks;
-
-		bankw = AMDGPU_TILING_GET(tiling_flags, BANK_WIDTH);
-		bankh = AMDGPU_TILING_GET(tiling_flags, BANK_HEIGHT);
-		mtaspect = AMDGPU_TILING_GET(tiling_flags, MACRO_TILE_ASPECT);
-		tile_split = AMDGPU_TILING_GET(tiling_flags, TILE_SPLIT);
-		num_banks = AMDGPU_TILING_GET(tiling_flags, NUM_BANKS);
-
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_NUM_BANKS, num_banks);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_ARRAY_MODE,
-					  ARRAY_2D_TILED_THIN1);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_TILE_SPLIT,
-					  tile_split);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_BANK_WIDTH, bankw);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_BANK_HEIGHT, bankh);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_MACRO_TILE_ASPECT,
-					  mtaspect);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_MICRO_TILE_MODE,
-					  ADDR_SURF_MICRO_TILING_DISPLAY);
-	} else if (AMDGPU_TILING_GET(tiling_flags, ARRAY_MODE) == ARRAY_1D_TILED_THIN1) {
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_ARRAY_MODE,
-					  ARRAY_1D_TILED_THIN1);
-	}
-
-	fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_PIPE_CONFIG,
-				  pipe_config);
-
-	dce_v11_0_vga_enable(crtc, false);
-
-	/* Make sure surface address is updated at vertical blank rather than
-	 * horizontal blank
-	 */
-	tmp = RREG32(mmGRPH_FLIP_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, GRPH_FLIP_CONTROL,
-			    GRPH_SURFACE_UPDATE_H_RETRACE_EN, 0);
-	WREG32(mmGRPH_FLIP_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-
-	WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS_HIGH + amdgpu_crtc->crtc_offset,
-	       upper_32_bits(fb_location));
-	WREG32(mmGRPH_SECONDARY_SURFACE_ADDRESS_HIGH + amdgpu_crtc->crtc_offset,
-	       upper_32_bits(fb_location));
-	WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
-	       (u32)fb_location & GRPH_PRIMARY_SURFACE_ADDRESS__GRPH_PRIMARY_SURFACE_ADDRESS_MASK);
-	WREG32(mmGRPH_SECONDARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
-	       (u32) fb_location & GRPH_SECONDARY_SURFACE_ADDRESS__GRPH_SECONDARY_SURFACE_ADDRESS_MASK);
-	WREG32(mmGRPH_CONTROL + amdgpu_crtc->crtc_offset, fb_format);
-	WREG32(mmGRPH_SWAP_CNTL + amdgpu_crtc->crtc_offset, fb_swap);
-
-	/*
-	 * The LUT only has 256 slots for indexing by a 8 bpc fb. Bypass the LUT
-	 * for > 8 bpc scanout to avoid truncation of fb indices to 8 msb's, to
-	 * retain the full precision throughout the pipeline.
-	 */
-	tmp = RREG32(mmGRPH_LUT_10BIT_BYPASS + amdgpu_crtc->crtc_offset);
-	if (bypass_lut)
-		tmp = REG_SET_FIELD(tmp, GRPH_LUT_10BIT_BYPASS, GRPH_LUT_10BIT_BYPASS_EN, 1);
-	else
-		tmp = REG_SET_FIELD(tmp, GRPH_LUT_10BIT_BYPASS, GRPH_LUT_10BIT_BYPASS_EN, 0);
-	WREG32(mmGRPH_LUT_10BIT_BYPASS + amdgpu_crtc->crtc_offset, tmp);
-
-	if (bypass_lut)
-		DRM_DEBUG_KMS("Bypassing hardware LUT due to 10 bit fb scanout.\n");
-
-	WREG32(mmGRPH_SURFACE_OFFSET_X + amdgpu_crtc->crtc_offset, 0);
-	WREG32(mmGRPH_SURFACE_OFFSET_Y + amdgpu_crtc->crtc_offset, 0);
-	WREG32(mmGRPH_X_START + amdgpu_crtc->crtc_offset, 0);
-	WREG32(mmGRPH_Y_START + amdgpu_crtc->crtc_offset, 0);
-	WREG32(mmGRPH_X_END + amdgpu_crtc->crtc_offset, target_fb->width);
-	WREG32(mmGRPH_Y_END + amdgpu_crtc->crtc_offset, target_fb->height);
-
-	fb_pitch_pixels = target_fb->pitches[0] / target_fb->format->cpp[0];
-	WREG32(mmGRPH_PITCH + amdgpu_crtc->crtc_offset, fb_pitch_pixels);
-
-	dce_v11_0_grph_enable(crtc, true);
-
-	WREG32(mmLB_DESKTOP_HEIGHT + amdgpu_crtc->crtc_offset,
-	       target_fb->height);
-
-	x &= ~3;
-	y &= ~1;
-	WREG32(mmVIEWPORT_START + amdgpu_crtc->crtc_offset,
-	       (x << 16) | y);
-	viewport_w = crtc->mode.hdisplay;
-	viewport_h = (crtc->mode.vdisplay + 1) & ~1;
-	WREG32(mmVIEWPORT_SIZE + amdgpu_crtc->crtc_offset,
-	       (viewport_w << 16) | viewport_h);
-
-	/* set pageflip to happen anywhere in vblank interval */
-	WREG32(mmCRTC_MASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 0);
-
-	if (!atomic && fb && fb != crtc->primary->fb) {
-		abo = gem_to_amdgpu_bo(fb->obj[0]);
-		r = amdgpu_bo_reserve(abo, true);
-		if (unlikely(r != 0))
-			return r;
-		amdgpu_bo_unpin(abo);
-		amdgpu_bo_unreserve(abo);
-	}
-
-	/* Bytes per pixel may have changed */
-	dce_v11_0_bandwidth_update(adev);
-
-	return 0;
-}
-
-static void dce_v11_0_set_interleave(struct drm_crtc *crtc,
-				     struct drm_display_mode *mode)
-{
-	struct drm_device *dev = crtc->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	u32 tmp;
-
-	tmp = RREG32(mmLB_DATA_FORMAT + amdgpu_crtc->crtc_offset);
-	if (mode->flags & DRM_MODE_FLAG_INTERLACE)
-		tmp = REG_SET_FIELD(tmp, LB_DATA_FORMAT, INTERLEAVE_EN, 1);
-	else
-		tmp = REG_SET_FIELD(tmp, LB_DATA_FORMAT, INTERLEAVE_EN, 0);
-	WREG32(mmLB_DATA_FORMAT + amdgpu_crtc->crtc_offset, tmp);
-}
-
-static void dce_v11_0_crtc_load_lut(struct drm_crtc *crtc)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct drm_device *dev = crtc->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	u16 *r, *g, *b;
-	int i;
-	u32 tmp;
-
-	DRM_DEBUG_KMS("%d\n", amdgpu_crtc->crtc_id);
-
-	tmp = RREG32(mmINPUT_CSC_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, INPUT_CSC_CONTROL, INPUT_CSC_GRPH_MODE, 0);
-	WREG32(mmINPUT_CSC_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-
-	tmp = RREG32(mmPRESCALE_GRPH_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, PRESCALE_GRPH_CONTROL, GRPH_PRESCALE_BYPASS, 1);
-	WREG32(mmPRESCALE_GRPH_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-
-	tmp = RREG32(mmINPUT_GAMMA_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, INPUT_GAMMA_CONTROL, GRPH_INPUT_GAMMA_MODE, 0);
-	WREG32(mmINPUT_GAMMA_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-
-	WREG32(mmDC_LUT_CONTROL + amdgpu_crtc->crtc_offset, 0);
-
-	WREG32(mmDC_LUT_BLACK_OFFSET_BLUE + amdgpu_crtc->crtc_offset, 0);
-	WREG32(mmDC_LUT_BLACK_OFFSET_GREEN + amdgpu_crtc->crtc_offset, 0);
-	WREG32(mmDC_LUT_BLACK_OFFSET_RED + amdgpu_crtc->crtc_offset, 0);
-
-	WREG32(mmDC_LUT_WHITE_OFFSET_BLUE + amdgpu_crtc->crtc_offset, 0xffff);
-	WREG32(mmDC_LUT_WHITE_OFFSET_GREEN + amdgpu_crtc->crtc_offset, 0xffff);
-	WREG32(mmDC_LUT_WHITE_OFFSET_RED + amdgpu_crtc->crtc_offset, 0xffff);
-
-	WREG32(mmDC_LUT_RW_MODE + amdgpu_crtc->crtc_offset, 0);
-	WREG32(mmDC_LUT_WRITE_EN_MASK + amdgpu_crtc->crtc_offset, 0x00000007);
-
-	WREG32(mmDC_LUT_RW_INDEX + amdgpu_crtc->crtc_offset, 0);
-	r = crtc->gamma_store;
-	g = r + crtc->gamma_size;
-	b = g + crtc->gamma_size;
-	for (i = 0; i < 256; i++) {
-		WREG32(mmDC_LUT_30_COLOR + amdgpu_crtc->crtc_offset,
-		       ((*r++ & 0xffc0) << 14) |
-		       ((*g++ & 0xffc0) << 4) |
-		       (*b++ >> 6));
-	}
-
-	tmp = RREG32(mmDEGAMMA_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, DEGAMMA_CONTROL, GRPH_DEGAMMA_MODE, 0);
-	tmp = REG_SET_FIELD(tmp, DEGAMMA_CONTROL, CURSOR_DEGAMMA_MODE, 0);
-	tmp = REG_SET_FIELD(tmp, DEGAMMA_CONTROL, CURSOR2_DEGAMMA_MODE, 0);
-	WREG32(mmDEGAMMA_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-
-	tmp = RREG32(mmGAMUT_REMAP_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, GAMUT_REMAP_CONTROL, GRPH_GAMUT_REMAP_MODE, 0);
-	WREG32(mmGAMUT_REMAP_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-
-	tmp = RREG32(mmREGAMMA_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, REGAMMA_CONTROL, GRPH_REGAMMA_MODE, 0);
-	WREG32(mmREGAMMA_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-
-	tmp = RREG32(mmOUTPUT_CSC_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, OUTPUT_CSC_CONTROL, OUTPUT_CSC_GRPH_MODE, 0);
-	WREG32(mmOUTPUT_CSC_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-
-	/* XXX match this to the depth of the crtc fmt block, move to modeset? */
-	WREG32(mmDENORM_CONTROL + amdgpu_crtc->crtc_offset, 0);
-	/* XXX this only needs to be programmed once per crtc at startup,
-	 * not sure where the best place for it is
-	 */
-	tmp = RREG32(mmALPHA_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, ALPHA_CONTROL, CURSOR_ALPHA_BLND_ENA, 1);
-	WREG32(mmALPHA_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-}
-
-static int dce_v11_0_pick_dig_encoder(struct drm_encoder *encoder)
-{
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-
-	switch (amdgpu_encoder->encoder_id) {
-	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
-		if (dig->linkb)
-			return 1;
-		else
-			return 0;
-	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
-		if (dig->linkb)
-			return 3;
-		else
-			return 2;
-	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
-		if (dig->linkb)
-			return 5;
-		else
-			return 4;
-	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY3:
-		return 6;
-	default:
-		DRM_ERROR("invalid encoder_id: 0x%x\n", amdgpu_encoder->encoder_id);
-		return 0;
-	}
-}
-
-/**
- * dce_v11_0_pick_pll - Allocate a PPLL for use by the crtc.
- *
- * @crtc: drm crtc
- *
- * Returns the PPLL (Pixel PLL) to be used by the crtc.  For DP monitors
- * a single PPLL can be used for all DP crtcs/encoders.  For non-DP
- * monitors a dedicated PPLL must be used.  If a particular board has
- * an external DP PLL, return ATOM_PPLL_INVALID to skip PLL programming
- * as there is no need to program the PLL itself.  If we are not able to
- * allocate a PLL, return ATOM_PPLL_INVALID to skip PLL programming to
- * avoid messing up an existing monitor.
- *
- * Asic specific PLL information
- *
- * DCE 10.x
- * Tonga
- * - PPLL1, PPLL2 are available for all UNIPHY (both DP and non-DP)
- * CI
- * - PPLL0, PPLL1, PPLL2 are available for all UNIPHY (both DP and non-DP) and DAC
- *
- */
-static u32 dce_v11_0_pick_pll(struct drm_crtc *crtc)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct drm_device *dev = crtc->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	u32 pll_in_use;
-	int pll;
-
-	if ((adev->asic_type == CHIP_POLARIS10) ||
-	    (adev->asic_type == CHIP_POLARIS11) ||
-	    (adev->asic_type == CHIP_POLARIS12) ||
-	    (adev->asic_type == CHIP_VEGAM)) {
-		struct amdgpu_encoder *amdgpu_encoder =
-			to_amdgpu_encoder(amdgpu_crtc->encoder);
-		struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-
-		if (ENCODER_MODE_IS_DP(amdgpu_atombios_encoder_get_encoder_mode(amdgpu_crtc->encoder)))
-			return ATOM_DP_DTO;
-
-		switch (amdgpu_encoder->encoder_id) {
-		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
-			if (dig->linkb)
-				return ATOM_COMBOPHY_PLL1;
-			else
-				return ATOM_COMBOPHY_PLL0;
-		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
-			if (dig->linkb)
-				return ATOM_COMBOPHY_PLL3;
-			else
-				return ATOM_COMBOPHY_PLL2;
-		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
-			if (dig->linkb)
-				return ATOM_COMBOPHY_PLL5;
-			else
-				return ATOM_COMBOPHY_PLL4;
-		default:
-			DRM_ERROR("invalid encoder_id: 0x%x\n", amdgpu_encoder->encoder_id);
-			return ATOM_PPLL_INVALID;
-		}
-	}
-
-	if (ENCODER_MODE_IS_DP(amdgpu_atombios_encoder_get_encoder_mode(amdgpu_crtc->encoder))) {
-		if (adev->clock.dp_extclk)
-			/* skip PPLL programming if using ext clock */
-			return ATOM_PPLL_INVALID;
-		else {
-			/* use the same PPLL for all DP monitors */
-			pll = amdgpu_pll_get_shared_dp_ppll(crtc);
-			if (pll != ATOM_PPLL_INVALID)
-				return pll;
-		}
-	} else {
-		/* use the same PPLL for all monitors with the same clock */
-		pll = amdgpu_pll_get_shared_nondp_ppll(crtc);
-		if (pll != ATOM_PPLL_INVALID)
-			return pll;
-	}
-
-	/* XXX need to determine what plls are available on each DCE11 part */
-	pll_in_use = amdgpu_pll_get_use_mask(crtc);
-	if (adev->flags & AMD_IS_APU) {
-		if (!(pll_in_use & (1 << ATOM_PPLL1)))
-			return ATOM_PPLL1;
-		if (!(pll_in_use & (1 << ATOM_PPLL0)))
-			return ATOM_PPLL0;
-		DRM_ERROR("unable to allocate a PPLL\n");
-		return ATOM_PPLL_INVALID;
-	} else {
-		if (!(pll_in_use & (1 << ATOM_PPLL2)))
-			return ATOM_PPLL2;
-		if (!(pll_in_use & (1 << ATOM_PPLL1)))
-			return ATOM_PPLL1;
-		if (!(pll_in_use & (1 << ATOM_PPLL0)))
-			return ATOM_PPLL0;
-		DRM_ERROR("unable to allocate a PPLL\n");
-		return ATOM_PPLL_INVALID;
-	}
-	return ATOM_PPLL_INVALID;
-}
-
-static void dce_v11_0_lock_cursor(struct drm_crtc *crtc, bool lock)
-{
-	struct amdgpu_device *adev = drm_to_adev(crtc->dev);
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	uint32_t cur_lock;
-
-	cur_lock = RREG32(mmCUR_UPDATE + amdgpu_crtc->crtc_offset);
-	if (lock)
-		cur_lock = REG_SET_FIELD(cur_lock, CUR_UPDATE, CURSOR_UPDATE_LOCK, 1);
-	else
-		cur_lock = REG_SET_FIELD(cur_lock, CUR_UPDATE, CURSOR_UPDATE_LOCK, 0);
-	WREG32(mmCUR_UPDATE + amdgpu_crtc->crtc_offset, cur_lock);
-}
-
-static void dce_v11_0_hide_cursor(struct drm_crtc *crtc)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct amdgpu_device *adev = drm_to_adev(crtc->dev);
-	u32 tmp;
-
-	tmp = RREG32(mmCUR_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, CUR_CONTROL, CURSOR_EN, 0);
-	WREG32(mmCUR_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-}
-
-static void dce_v11_0_show_cursor(struct drm_crtc *crtc)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct amdgpu_device *adev = drm_to_adev(crtc->dev);
-	u32 tmp;
-
-	WREG32(mmCUR_SURFACE_ADDRESS_HIGH + amdgpu_crtc->crtc_offset,
-	       upper_32_bits(amdgpu_crtc->cursor_addr));
-	WREG32(mmCUR_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
-	       lower_32_bits(amdgpu_crtc->cursor_addr));
-
-	tmp = RREG32(mmCUR_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, CUR_CONTROL, CURSOR_EN, 1);
-	tmp = REG_SET_FIELD(tmp, CUR_CONTROL, CURSOR_MODE, 2);
-	WREG32(mmCUR_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-}
-
-static int dce_v11_0_cursor_move_locked(struct drm_crtc *crtc,
-					int x, int y)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct amdgpu_device *adev = drm_to_adev(crtc->dev);
-	int xorigin = 0, yorigin = 0;
-
-	amdgpu_crtc->cursor_x = x;
-	amdgpu_crtc->cursor_y = y;
-
-	/* avivo cursor are offset into the total surface */
-	x += crtc->x;
-	y += crtc->y;
-	DRM_DEBUG("x %d y %d c->x %d c->y %d\n", x, y, crtc->x, crtc->y);
-
-	if (x < 0) {
-		xorigin = min(-x, amdgpu_crtc->max_cursor_width - 1);
-		x = 0;
-	}
-	if (y < 0) {
-		yorigin = min(-y, amdgpu_crtc->max_cursor_height - 1);
-		y = 0;
-	}
-
-	WREG32(mmCUR_POSITION + amdgpu_crtc->crtc_offset, (x << 16) | y);
-	WREG32(mmCUR_HOT_SPOT + amdgpu_crtc->crtc_offset, (xorigin << 16) | yorigin);
-	WREG32(mmCUR_SIZE + amdgpu_crtc->crtc_offset,
-	       ((amdgpu_crtc->cursor_width - 1) << 16) | (amdgpu_crtc->cursor_height - 1));
-
-	return 0;
-}
-
-static int dce_v11_0_crtc_cursor_move(struct drm_crtc *crtc,
-				      int x, int y)
-{
-	int ret;
-
-	dce_v11_0_lock_cursor(crtc, true);
-	ret = dce_v11_0_cursor_move_locked(crtc, x, y);
-	dce_v11_0_lock_cursor(crtc, false);
-
-	return ret;
-}
-
-static int dce_v11_0_crtc_cursor_set2(struct drm_crtc *crtc,
-				      struct drm_file *file_priv,
-				      uint32_t handle,
-				      uint32_t width,
-				      uint32_t height,
-				      int32_t hot_x,
-				      int32_t hot_y)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct drm_gem_object *obj;
-	struct amdgpu_bo *aobj;
-	int ret;
-
-	if (!handle) {
-		/* turn off cursor */
-		dce_v11_0_hide_cursor(crtc);
-		obj = NULL;
-		goto unpin;
-	}
-
-	if ((width > amdgpu_crtc->max_cursor_width) ||
-	    (height > amdgpu_crtc->max_cursor_height)) {
-		DRM_ERROR("bad cursor width or height %d x %d\n", width, height);
-		return -EINVAL;
-	}
-
-	obj = drm_gem_object_lookup(file_priv, handle);
-	if (!obj) {
-		DRM_ERROR("Cannot find cursor object %x for crtc %d\n", handle, amdgpu_crtc->crtc_id);
-		return -ENOENT;
-	}
-
-	aobj = gem_to_amdgpu_bo(obj);
-	ret = amdgpu_bo_reserve(aobj, false);
-	if (ret != 0) {
-		drm_gem_object_put(obj);
-		return ret;
-	}
-
-	ret = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM);
-	amdgpu_bo_unreserve(aobj);
-	if (ret) {
-		DRM_ERROR("Failed to pin new cursor BO (%d)\n", ret);
-		drm_gem_object_put(obj);
-		return ret;
-	}
-	amdgpu_crtc->cursor_addr = amdgpu_bo_gpu_offset(aobj);
-
-	dce_v11_0_lock_cursor(crtc, true);
-
-	if (width != amdgpu_crtc->cursor_width ||
-	    height != amdgpu_crtc->cursor_height ||
-	    hot_x != amdgpu_crtc->cursor_hot_x ||
-	    hot_y != amdgpu_crtc->cursor_hot_y) {
-		int x, y;
-
-		x = amdgpu_crtc->cursor_x + amdgpu_crtc->cursor_hot_x - hot_x;
-		y = amdgpu_crtc->cursor_y + amdgpu_crtc->cursor_hot_y - hot_y;
-
-		dce_v11_0_cursor_move_locked(crtc, x, y);
-
-		amdgpu_crtc->cursor_width = width;
-		amdgpu_crtc->cursor_height = height;
-		amdgpu_crtc->cursor_hot_x = hot_x;
-		amdgpu_crtc->cursor_hot_y = hot_y;
-	}
-
-	dce_v11_0_show_cursor(crtc);
-	dce_v11_0_lock_cursor(crtc, false);
-
-unpin:
-	if (amdgpu_crtc->cursor_bo) {
-		struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo);
-		ret = amdgpu_bo_reserve(aobj, true);
-		if (likely(ret == 0)) {
-			amdgpu_bo_unpin(aobj);
-			amdgpu_bo_unreserve(aobj);
-		}
-		drm_gem_object_put(amdgpu_crtc->cursor_bo);
-	}
-
-	amdgpu_crtc->cursor_bo = obj;
-	return 0;
-}
-
-static void dce_v11_0_cursor_reset(struct drm_crtc *crtc)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-
-	if (amdgpu_crtc->cursor_bo) {
-		dce_v11_0_lock_cursor(crtc, true);
-
-		dce_v11_0_cursor_move_locked(crtc, amdgpu_crtc->cursor_x,
-					     amdgpu_crtc->cursor_y);
-
-		dce_v11_0_show_cursor(crtc);
-
-		dce_v11_0_lock_cursor(crtc, false);
-	}
-}
-
-static int dce_v11_0_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, u16 *green,
-				    u16 *blue, uint32_t size,
-				    struct drm_modeset_acquire_ctx *ctx)
-{
-	dce_v11_0_crtc_load_lut(crtc);
-
-	return 0;
-}
-
-static void dce_v11_0_crtc_destroy(struct drm_crtc *crtc)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-
-	drm_crtc_cleanup(crtc);
-	kfree(amdgpu_crtc);
-}
-
-static const struct drm_crtc_funcs dce_v11_0_crtc_funcs = {
-	.cursor_set2 = dce_v11_0_crtc_cursor_set2,
-	.cursor_move = dce_v11_0_crtc_cursor_move,
-	.gamma_set = dce_v11_0_crtc_gamma_set,
-	.set_config = amdgpu_display_crtc_set_config,
-	.destroy = dce_v11_0_crtc_destroy,
-	.page_flip_target = amdgpu_display_crtc_page_flip_target,
-	.get_vblank_counter = amdgpu_get_vblank_counter_kms,
-	.enable_vblank = amdgpu_enable_vblank_kms,
-	.disable_vblank = amdgpu_disable_vblank_kms,
-	.get_vblank_timestamp = drm_crtc_vblank_helper_get_vblank_timestamp,
-};
-
-static void dce_v11_0_crtc_dpms(struct drm_crtc *crtc, int mode)
-{
-	struct drm_device *dev = crtc->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	unsigned type;
-
-	switch (mode) {
-	case DRM_MODE_DPMS_ON:
-		amdgpu_crtc->enabled = true;
-		amdgpu_atombios_crtc_enable(crtc, ATOM_ENABLE);
-		dce_v11_0_vga_enable(crtc, true);
-		amdgpu_atombios_crtc_blank(crtc, ATOM_DISABLE);
-		dce_v11_0_vga_enable(crtc, false);
-		/* Make sure VBLANK and PFLIP interrupts are still enabled */
-		type = amdgpu_display_crtc_idx_to_irq_type(adev,
-						amdgpu_crtc->crtc_id);
-		amdgpu_irq_update(adev, &adev->crtc_irq, type);
-		amdgpu_irq_update(adev, &adev->pageflip_irq, type);
-		drm_crtc_vblank_on(crtc);
-		dce_v11_0_crtc_load_lut(crtc);
-		break;
-	case DRM_MODE_DPMS_STANDBY:
-	case DRM_MODE_DPMS_SUSPEND:
-	case DRM_MODE_DPMS_OFF:
-		drm_crtc_vblank_off(crtc);
-		if (amdgpu_crtc->enabled) {
-			dce_v11_0_vga_enable(crtc, true);
-			amdgpu_atombios_crtc_blank(crtc, ATOM_ENABLE);
-			dce_v11_0_vga_enable(crtc, false);
-		}
-		amdgpu_atombios_crtc_enable(crtc, ATOM_DISABLE);
-		amdgpu_crtc->enabled = false;
-		break;
-	}
-	/* adjust pm to dpms */
-	amdgpu_dpm_compute_clocks(adev);
-}
-
-static void dce_v11_0_crtc_prepare(struct drm_crtc *crtc)
-{
-	/* disable crtc pair power gating before programming */
-	amdgpu_atombios_crtc_powergate(crtc, ATOM_DISABLE);
-	amdgpu_atombios_crtc_lock(crtc, ATOM_ENABLE);
-	dce_v11_0_crtc_dpms(crtc, DRM_MODE_DPMS_OFF);
-}
-
-static void dce_v11_0_crtc_commit(struct drm_crtc *crtc)
-{
-	dce_v11_0_crtc_dpms(crtc, DRM_MODE_DPMS_ON);
-	amdgpu_atombios_crtc_lock(crtc, ATOM_DISABLE);
-}
-
-static void dce_v11_0_crtc_disable(struct drm_crtc *crtc)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct drm_device *dev = crtc->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_atom_ss ss;
-	int i;
-
-	dce_v11_0_crtc_dpms(crtc, DRM_MODE_DPMS_OFF);
-	if (crtc->primary->fb) {
-		int r;
-		struct amdgpu_bo *abo;
-
-		abo = gem_to_amdgpu_bo(crtc->primary->fb->obj[0]);
-		r = amdgpu_bo_reserve(abo, true);
-		if (unlikely(r))
-			DRM_ERROR("failed to reserve abo before unpin\n");
-		else {
-			amdgpu_bo_unpin(abo);
-			amdgpu_bo_unreserve(abo);
-		}
-	}
-	/* disable the GRPH */
-	dce_v11_0_grph_enable(crtc, false);
-
-	amdgpu_atombios_crtc_powergate(crtc, ATOM_ENABLE);
-
-	for (i = 0; i < adev->mode_info.num_crtc; i++) {
-		if (adev->mode_info.crtcs[i] &&
-		    adev->mode_info.crtcs[i]->enabled &&
-		    i != amdgpu_crtc->crtc_id &&
-		    amdgpu_crtc->pll_id == adev->mode_info.crtcs[i]->pll_id) {
-			/* one other crtc is using this pll don't turn
-			 * off the pll
-			 */
-			goto done;
-		}
-	}
-
-	switch (amdgpu_crtc->pll_id) {
-	case ATOM_PPLL0:
-	case ATOM_PPLL1:
-	case ATOM_PPLL2:
-		/* disable the ppll */
-		amdgpu_atombios_crtc_program_pll(crtc, amdgpu_crtc->crtc_id, amdgpu_crtc->pll_id,
-						 0, 0, ATOM_DISABLE, 0, 0, 0, 0, 0, false, &ss);
-		break;
-	case ATOM_COMBOPHY_PLL0:
-	case ATOM_COMBOPHY_PLL1:
-	case ATOM_COMBOPHY_PLL2:
-	case ATOM_COMBOPHY_PLL3:
-	case ATOM_COMBOPHY_PLL4:
-	case ATOM_COMBOPHY_PLL5:
-		/* disable the ppll */
-		amdgpu_atombios_crtc_program_pll(crtc, ATOM_CRTC_INVALID, amdgpu_crtc->pll_id,
-						 0, 0, ATOM_DISABLE, 0, 0, 0, 0, 0, false, &ss);
-		break;
-	default:
-		break;
-	}
-done:
-	amdgpu_crtc->pll_id = ATOM_PPLL_INVALID;
-	amdgpu_crtc->adjusted_clock = 0;
-	amdgpu_crtc->encoder = NULL;
-	amdgpu_crtc->connector = NULL;
-}
-
-static int dce_v11_0_crtc_mode_set(struct drm_crtc *crtc,
-				  struct drm_display_mode *mode,
-				  struct drm_display_mode *adjusted_mode,
-				  int x, int y, struct drm_framebuffer *old_fb)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct drm_device *dev = crtc->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-
-	if (!amdgpu_crtc->adjusted_clock)
-		return -EINVAL;
-
-	if ((adev->asic_type == CHIP_POLARIS10) ||
-	    (adev->asic_type == CHIP_POLARIS11) ||
-	    (adev->asic_type == CHIP_POLARIS12) ||
-	    (adev->asic_type == CHIP_VEGAM)) {
-		struct amdgpu_encoder *amdgpu_encoder =
-			to_amdgpu_encoder(amdgpu_crtc->encoder);
-		int encoder_mode =
-			amdgpu_atombios_encoder_get_encoder_mode(amdgpu_crtc->encoder);
-
-		/* SetPixelClock calculates the plls and ss values now */
-		amdgpu_atombios_crtc_program_pll(crtc, amdgpu_crtc->crtc_id,
-						 amdgpu_crtc->pll_id,
-						 encoder_mode, amdgpu_encoder->encoder_id,
-						 adjusted_mode->clock, 0, 0, 0, 0,
-						 amdgpu_crtc->bpc, amdgpu_crtc->ss_enabled, &amdgpu_crtc->ss);
-	} else {
-		amdgpu_atombios_crtc_set_pll(crtc, adjusted_mode);
-	}
-	amdgpu_atombios_crtc_set_dtd_timing(crtc, adjusted_mode);
-	dce_v11_0_crtc_do_set_base(crtc, old_fb, x, y, 0);
-	amdgpu_atombios_crtc_overscan_setup(crtc, mode, adjusted_mode);
-	amdgpu_atombios_crtc_scaler_setup(crtc);
-	dce_v11_0_cursor_reset(crtc);
-	/* update the hw version fpr dpm */
-	amdgpu_crtc->hw_mode = *adjusted_mode;
-
-	return 0;
-}
-
-static bool dce_v11_0_crtc_mode_fixup(struct drm_crtc *crtc,
-				     const struct drm_display_mode *mode,
-				     struct drm_display_mode *adjusted_mode)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct drm_device *dev = crtc->dev;
-	struct drm_encoder *encoder;
-
-	/* assign the encoder to the amdgpu crtc to avoid repeated lookups later */
-	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
-		if (encoder->crtc == crtc) {
-			amdgpu_crtc->encoder = encoder;
-			amdgpu_crtc->connector = amdgpu_get_connector_for_encoder(encoder);
-			break;
-		}
-	}
-	if ((amdgpu_crtc->encoder == NULL) || (amdgpu_crtc->connector == NULL)) {
-		amdgpu_crtc->encoder = NULL;
-		amdgpu_crtc->connector = NULL;
-		return false;
-	}
-	if (!amdgpu_display_crtc_scaling_mode_fixup(crtc, mode, adjusted_mode))
-		return false;
-	if (amdgpu_atombios_crtc_prepare_pll(crtc, adjusted_mode))
-		return false;
-	/* pick pll */
-	amdgpu_crtc->pll_id = dce_v11_0_pick_pll(crtc);
-	/* if we can't get a PPLL for a non-DP encoder, fail */
-	if ((amdgpu_crtc->pll_id == ATOM_PPLL_INVALID) &&
-	    !ENCODER_MODE_IS_DP(amdgpu_atombios_encoder_get_encoder_mode(amdgpu_crtc->encoder)))
-		return false;
-
-	return true;
-}
-
-static int dce_v11_0_crtc_set_base(struct drm_crtc *crtc, int x, int y,
-				  struct drm_framebuffer *old_fb)
-{
-	return dce_v11_0_crtc_do_set_base(crtc, old_fb, x, y, 0);
-}
-
-static int dce_v11_0_crtc_set_base_atomic(struct drm_crtc *crtc,
-					 struct drm_framebuffer *fb,
-					 int x, int y, enum mode_set_atomic state)
-{
-	return dce_v11_0_crtc_do_set_base(crtc, fb, x, y, 1);
-}
-
-static const struct drm_crtc_helper_funcs dce_v11_0_crtc_helper_funcs = {
-	.dpms = dce_v11_0_crtc_dpms,
-	.mode_fixup = dce_v11_0_crtc_mode_fixup,
-	.mode_set = dce_v11_0_crtc_mode_set,
-	.mode_set_base = dce_v11_0_crtc_set_base,
-	.mode_set_base_atomic = dce_v11_0_crtc_set_base_atomic,
-	.prepare = dce_v11_0_crtc_prepare,
-	.commit = dce_v11_0_crtc_commit,
-	.disable = dce_v11_0_crtc_disable,
-	.get_scanout_position = amdgpu_crtc_get_scanout_position,
-};
-
-static int dce_v11_0_crtc_init(struct amdgpu_device *adev, int index)
-{
-	struct amdgpu_crtc *amdgpu_crtc;
-
-	amdgpu_crtc = kzalloc(sizeof(struct amdgpu_crtc) +
-			      (AMDGPUFB_CONN_LIMIT * sizeof(struct drm_connector *)), GFP_KERNEL);
-	if (amdgpu_crtc == NULL)
-		return -ENOMEM;
-
-	drm_crtc_init(adev_to_drm(adev), &amdgpu_crtc->base, &dce_v11_0_crtc_funcs);
-
-	drm_mode_crtc_set_gamma_size(&amdgpu_crtc->base, 256);
-	amdgpu_crtc->crtc_id = index;
-	adev->mode_info.crtcs[index] = amdgpu_crtc;
-
-	amdgpu_crtc->max_cursor_width = 128;
-	amdgpu_crtc->max_cursor_height = 128;
-	adev_to_drm(adev)->mode_config.cursor_width = amdgpu_crtc->max_cursor_width;
-	adev_to_drm(adev)->mode_config.cursor_height = amdgpu_crtc->max_cursor_height;
-
-	switch (amdgpu_crtc->crtc_id) {
-	case 0:
-	default:
-		amdgpu_crtc->crtc_offset = CRTC0_REGISTER_OFFSET;
-		break;
-	case 1:
-		amdgpu_crtc->crtc_offset = CRTC1_REGISTER_OFFSET;
-		break;
-	case 2:
-		amdgpu_crtc->crtc_offset = CRTC2_REGISTER_OFFSET;
-		break;
-	case 3:
-		amdgpu_crtc->crtc_offset = CRTC3_REGISTER_OFFSET;
-		break;
-	case 4:
-		amdgpu_crtc->crtc_offset = CRTC4_REGISTER_OFFSET;
-		break;
-	case 5:
-		amdgpu_crtc->crtc_offset = CRTC5_REGISTER_OFFSET;
-		break;
-	}
-
-	amdgpu_crtc->pll_id = ATOM_PPLL_INVALID;
-	amdgpu_crtc->adjusted_clock = 0;
-	amdgpu_crtc->encoder = NULL;
-	amdgpu_crtc->connector = NULL;
-	drm_crtc_helper_add(&amdgpu_crtc->base, &dce_v11_0_crtc_helper_funcs);
-
-	return 0;
-}
-
-static int dce_v11_0_early_init(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	adev->audio_endpt_rreg = &dce_v11_0_audio_endpt_rreg;
-	adev->audio_endpt_wreg = &dce_v11_0_audio_endpt_wreg;
-
-	dce_v11_0_set_display_funcs(adev);
-
-	adev->mode_info.num_crtc = dce_v11_0_get_num_crtc(adev);
-
-	switch (adev->asic_type) {
-	case CHIP_CARRIZO:
-		adev->mode_info.num_hpd = 6;
-		adev->mode_info.num_dig = 9;
-		break;
-	case CHIP_STONEY:
-		adev->mode_info.num_hpd = 6;
-		adev->mode_info.num_dig = 9;
-		break;
-	case CHIP_POLARIS10:
-	case CHIP_VEGAM:
-		adev->mode_info.num_hpd = 6;
-		adev->mode_info.num_dig = 6;
-		break;
-	case CHIP_POLARIS11:
-	case CHIP_POLARIS12:
-		adev->mode_info.num_hpd = 5;
-		adev->mode_info.num_dig = 5;
-		break;
-	default:
-		/* FIXME: not supported yet */
-		return -EINVAL;
-	}
-
-	dce_v11_0_set_irq_funcs(adev);
-
-	return 0;
-}
-
-static int dce_v11_0_sw_init(void *handle)
-{
-	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	for (i = 0; i < adev->mode_info.num_crtc; i++) {
-		r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i + 1, &adev->crtc_irq);
-		if (r)
-			return r;
-	}
-
-	for (i = VISLANDS30_IV_SRCID_D1_GRPH_PFLIP; i < 20; i += 2) {
-		r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i, &adev->pageflip_irq);
-		if (r)
-			return r;
-	}
-
-	/* HPD hotplug */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_HOTPLUG_DETECT_A, &adev->hpd_irq);
-	if (r)
-		return r;
-
-	adev_to_drm(adev)->mode_config.funcs = &amdgpu_mode_funcs;
-
-	adev_to_drm(adev)->mode_config.async_page_flip = true;
-
-	adev_to_drm(adev)->mode_config.max_width = 16384;
-	adev_to_drm(adev)->mode_config.max_height = 16384;
-
-	adev_to_drm(adev)->mode_config.preferred_depth = 24;
-	adev_to_drm(adev)->mode_config.prefer_shadow = 1;
-
-	adev_to_drm(adev)->mode_config.fb_modifiers_not_supported = true;
-
-	r = amdgpu_display_modeset_create_props(adev);
-	if (r)
-		return r;
-
-	adev_to_drm(adev)->mode_config.max_width = 16384;
-	adev_to_drm(adev)->mode_config.max_height = 16384;
-
-
-	/* allocate crtcs */
-	for (i = 0; i < adev->mode_info.num_crtc; i++) {
-		r = dce_v11_0_crtc_init(adev, i);
-		if (r)
-			return r;
-	}
-
-	if (amdgpu_atombios_get_connector_info_from_object_table(adev))
-		amdgpu_display_print_display_setup(adev_to_drm(adev));
-	else
-		return -EINVAL;
-
-	/* setup afmt */
-	r = dce_v11_0_afmt_init(adev);
-	if (r)
-		return r;
-
-	r = dce_v11_0_audio_init(adev);
-	if (r)
-		return r;
-
-	/* Disable vblank IRQs aggressively for power-saving */
-	/* XXX: can this be enabled for DC? */
-	adev_to_drm(adev)->vblank_disable_immediate = true;
-
-	r = drm_vblank_init(adev_to_drm(adev), adev->mode_info.num_crtc);
-	if (r)
-		return r;
-
-	INIT_DELAYED_WORK(&adev->hotplug_work,
-		  amdgpu_display_hotplug_work_func);
-
-	drm_kms_helper_poll_init(adev_to_drm(adev));
-
-	adev->mode_info.mode_config_initialized = true;
-	return 0;
-}
-
-static int dce_v11_0_sw_fini(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	kfree(adev->mode_info.bios_hardcoded_edid);
-
-	drm_kms_helper_poll_fini(adev_to_drm(adev));
-
-	dce_v11_0_audio_fini(adev);
-
-	dce_v11_0_afmt_fini(adev);
-
-	drm_mode_config_cleanup(adev_to_drm(adev));
-	adev->mode_info.mode_config_initialized = false;
-
-	return 0;
-}
-
-static int dce_v11_0_hw_init(void *handle)
-{
-	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	dce_v11_0_init_golden_registers(adev);
-
-	/* disable vga render */
-	dce_v11_0_set_vga_render_state(adev, false);
-	/* init dig PHYs, disp eng pll */
-	amdgpu_atombios_crtc_powergate_init(adev);
-	amdgpu_atombios_encoder_init_dig(adev);
-	if ((adev->asic_type == CHIP_POLARIS10) ||
-	    (adev->asic_type == CHIP_POLARIS11) ||
-	    (adev->asic_type == CHIP_POLARIS12) ||
-	    (adev->asic_type == CHIP_VEGAM)) {
-		amdgpu_atombios_crtc_set_dce_clock(adev, adev->clock.default_dispclk,
-						   DCE_CLOCK_TYPE_DISPCLK, ATOM_GCK_DFS);
-		amdgpu_atombios_crtc_set_dce_clock(adev, 0,
-						   DCE_CLOCK_TYPE_DPREFCLK, ATOM_GCK_DFS);
-	} else {
-		amdgpu_atombios_crtc_set_disp_eng_pll(adev, adev->clock.default_dispclk);
-	}
-
-	/* initialize hpd */
-	dce_v11_0_hpd_init(adev);
-
-	for (i = 0; i < adev->mode_info.audio.num_pins; i++) {
-		dce_v11_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
-	}
-
-	dce_v11_0_pageflip_interrupt_init(adev);
-
-	return 0;
-}
-
-static int dce_v11_0_hw_fini(void *handle)
-{
-	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	dce_v11_0_hpd_fini(adev);
-
-	for (i = 0; i < adev->mode_info.audio.num_pins; i++) {
-		dce_v11_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
-	}
-
-	dce_v11_0_pageflip_interrupt_fini(adev);
-
-	flush_delayed_work(&adev->hotplug_work);
-
-	return 0;
-}
-
-static int dce_v11_0_suspend(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	int r;
-
-	r = amdgpu_display_suspend_helper(adev);
-	if (r)
-		return r;
-
-	adev->mode_info.bl_level =
-		amdgpu_atombios_encoder_get_backlight_level_from_reg(adev);
-
-	return dce_v11_0_hw_fini(handle);
-}
-
-static int dce_v11_0_resume(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	int ret;
-
-	amdgpu_atombios_encoder_set_backlight_level_to_reg(adev,
-							   adev->mode_info.bl_level);
-
-	ret = dce_v11_0_hw_init(handle);
-
-	/* turn on the BL */
-	if (adev->mode_info.bl_encoder) {
-		u8 bl_level = amdgpu_display_backlight_get_level(adev,
-								  adev->mode_info.bl_encoder);
-		amdgpu_display_backlight_set_level(adev, adev->mode_info.bl_encoder,
-						    bl_level);
-	}
-	if (ret)
-		return ret;
-
-	return amdgpu_display_resume_helper(adev);
-}
-
-static bool dce_v11_0_is_idle(void *handle)
-{
-	return true;
-}
-
-static int dce_v11_0_wait_for_idle(void *handle)
-{
-	return 0;
-}
-
-static int dce_v11_0_soft_reset(void *handle)
-{
-	u32 srbm_soft_reset = 0, tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	if (dce_v11_0_is_display_hung(adev))
-		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_DC_MASK;
-
-	if (srbm_soft_reset) {
-		tmp = RREG32(mmSRBM_SOFT_RESET);
-		tmp |= srbm_soft_reset;
-		dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp);
-		WREG32(mmSRBM_SOFT_RESET, tmp);
-		tmp = RREG32(mmSRBM_SOFT_RESET);
-
-		udelay(50);
-
-		tmp &= ~srbm_soft_reset;
-		WREG32(mmSRBM_SOFT_RESET, tmp);
-		tmp = RREG32(mmSRBM_SOFT_RESET);
-
-		/* Wait a little for things to settle down */
-		udelay(50);
-	}
-	return 0;
-}
-
-static void dce_v11_0_set_crtc_vblank_interrupt_state(struct amdgpu_device *adev,
-						     int crtc,
-						     enum amdgpu_interrupt_state state)
-{
-	u32 lb_interrupt_mask;
-
-	if (crtc >= adev->mode_info.num_crtc) {
-		DRM_DEBUG("invalid crtc %d\n", crtc);
-		return;
-	}
-
-	switch (state) {
-	case AMDGPU_IRQ_STATE_DISABLE:
-		lb_interrupt_mask = RREG32(mmLB_INTERRUPT_MASK + crtc_offsets[crtc]);
-		lb_interrupt_mask = REG_SET_FIELD(lb_interrupt_mask, LB_INTERRUPT_MASK,
-						  VBLANK_INTERRUPT_MASK, 0);
-		WREG32(mmLB_INTERRUPT_MASK + crtc_offsets[crtc], lb_interrupt_mask);
-		break;
-	case AMDGPU_IRQ_STATE_ENABLE:
-		lb_interrupt_mask = RREG32(mmLB_INTERRUPT_MASK + crtc_offsets[crtc]);
-		lb_interrupt_mask = REG_SET_FIELD(lb_interrupt_mask, LB_INTERRUPT_MASK,
-						  VBLANK_INTERRUPT_MASK, 1);
-		WREG32(mmLB_INTERRUPT_MASK + crtc_offsets[crtc], lb_interrupt_mask);
-		break;
-	default:
-		break;
-	}
-}
-
-static void dce_v11_0_set_crtc_vline_interrupt_state(struct amdgpu_device *adev,
-						    int crtc,
-						    enum amdgpu_interrupt_state state)
-{
-	u32 lb_interrupt_mask;
-
-	if (crtc >= adev->mode_info.num_crtc) {
-		DRM_DEBUG("invalid crtc %d\n", crtc);
-		return;
-	}
-
-	switch (state) {
-	case AMDGPU_IRQ_STATE_DISABLE:
-		lb_interrupt_mask = RREG32(mmLB_INTERRUPT_MASK + crtc_offsets[crtc]);
-		lb_interrupt_mask = REG_SET_FIELD(lb_interrupt_mask, LB_INTERRUPT_MASK,
-						  VLINE_INTERRUPT_MASK, 0);
-		WREG32(mmLB_INTERRUPT_MASK + crtc_offsets[crtc], lb_interrupt_mask);
-		break;
-	case AMDGPU_IRQ_STATE_ENABLE:
-		lb_interrupt_mask = RREG32(mmLB_INTERRUPT_MASK + crtc_offsets[crtc]);
-		lb_interrupt_mask = REG_SET_FIELD(lb_interrupt_mask, LB_INTERRUPT_MASK,
-						  VLINE_INTERRUPT_MASK, 1);
-		WREG32(mmLB_INTERRUPT_MASK + crtc_offsets[crtc], lb_interrupt_mask);
-		break;
-	default:
-		break;
-	}
-}
-
-static int dce_v11_0_set_hpd_irq_state(struct amdgpu_device *adev,
-					struct amdgpu_irq_src *source,
-					unsigned hpd,
-					enum amdgpu_interrupt_state state)
-{
-	u32 tmp;
-
-	if (hpd >= adev->mode_info.num_hpd) {
-		DRM_DEBUG("invalid hdp %d\n", hpd);
-		return 0;
-	}
-
-	switch (state) {
-	case AMDGPU_IRQ_STATE_DISABLE:
-		tmp = RREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[hpd]);
-		tmp = REG_SET_FIELD(tmp, DC_HPD_INT_CONTROL, DC_HPD_INT_EN, 0);
-		WREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[hpd], tmp);
-		break;
-	case AMDGPU_IRQ_STATE_ENABLE:
-		tmp = RREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[hpd]);
-		tmp = REG_SET_FIELD(tmp, DC_HPD_INT_CONTROL, DC_HPD_INT_EN, 1);
-		WREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[hpd], tmp);
-		break;
-	default:
-		break;
-	}
-
-	return 0;
-}
-
-static int dce_v11_0_set_crtc_irq_state(struct amdgpu_device *adev,
-					struct amdgpu_irq_src *source,
-					unsigned type,
-					enum amdgpu_interrupt_state state)
-{
-	switch (type) {
-	case AMDGPU_CRTC_IRQ_VBLANK1:
-		dce_v11_0_set_crtc_vblank_interrupt_state(adev, 0, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VBLANK2:
-		dce_v11_0_set_crtc_vblank_interrupt_state(adev, 1, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VBLANK3:
-		dce_v11_0_set_crtc_vblank_interrupt_state(adev, 2, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VBLANK4:
-		dce_v11_0_set_crtc_vblank_interrupt_state(adev, 3, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VBLANK5:
-		dce_v11_0_set_crtc_vblank_interrupt_state(adev, 4, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VBLANK6:
-		dce_v11_0_set_crtc_vblank_interrupt_state(adev, 5, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VLINE1:
-		dce_v11_0_set_crtc_vline_interrupt_state(adev, 0, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VLINE2:
-		dce_v11_0_set_crtc_vline_interrupt_state(adev, 1, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VLINE3:
-		dce_v11_0_set_crtc_vline_interrupt_state(adev, 2, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VLINE4:
-		dce_v11_0_set_crtc_vline_interrupt_state(adev, 3, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VLINE5:
-		dce_v11_0_set_crtc_vline_interrupt_state(adev, 4, state);
-		break;
-	 case AMDGPU_CRTC_IRQ_VLINE6:
-		dce_v11_0_set_crtc_vline_interrupt_state(adev, 5, state);
-		break;
-	default:
-		break;
-	}
-	return 0;
-}
-
-static int dce_v11_0_set_pageflip_irq_state(struct amdgpu_device *adev,
-					    struct amdgpu_irq_src *src,
-					    unsigned type,
-					    enum amdgpu_interrupt_state state)
-{
-	u32 reg;
-
-	if (type >= adev->mode_info.num_crtc) {
-		DRM_ERROR("invalid pageflip crtc %d\n", type);
-		return -EINVAL;
-	}
-
-	reg = RREG32(mmGRPH_INTERRUPT_CONTROL + crtc_offsets[type]);
-	if (state == AMDGPU_IRQ_STATE_DISABLE)
-		WREG32(mmGRPH_INTERRUPT_CONTROL + crtc_offsets[type],
-		       reg & ~GRPH_INTERRUPT_CONTROL__GRPH_PFLIP_INT_MASK_MASK);
-	else
-		WREG32(mmGRPH_INTERRUPT_CONTROL + crtc_offsets[type],
-		       reg | GRPH_INTERRUPT_CONTROL__GRPH_PFLIP_INT_MASK_MASK);
-
-	return 0;
-}
-
-static int dce_v11_0_pageflip_irq(struct amdgpu_device *adev,
-				  struct amdgpu_irq_src *source,
-				  struct amdgpu_iv_entry *entry)
-{
-	unsigned long flags;
-	unsigned crtc_id;
-	struct amdgpu_crtc *amdgpu_crtc;
-	struct amdgpu_flip_work *works;
-
-	crtc_id = (entry->src_id - 8) >> 1;
-	amdgpu_crtc = adev->mode_info.crtcs[crtc_id];
-
-	if (crtc_id >= adev->mode_info.num_crtc) {
-		DRM_ERROR("invalid pageflip crtc %d\n", crtc_id);
-		return -EINVAL;
-	}
-
-	if (RREG32(mmGRPH_INTERRUPT_STATUS + crtc_offsets[crtc_id]) &
-	    GRPH_INTERRUPT_STATUS__GRPH_PFLIP_INT_OCCURRED_MASK)
-		WREG32(mmGRPH_INTERRUPT_STATUS + crtc_offsets[crtc_id],
-		       GRPH_INTERRUPT_STATUS__GRPH_PFLIP_INT_CLEAR_MASK);
-
-	/* IRQ could occur when in initial stage */
-	if(amdgpu_crtc == NULL)
-		return 0;
-
-	spin_lock_irqsave(&adev_to_drm(adev)->event_lock, flags);
-	works = amdgpu_crtc->pflip_works;
-	if (amdgpu_crtc->pflip_status != AMDGPU_FLIP_SUBMITTED){
-		DRM_DEBUG_DRIVER("amdgpu_crtc->pflip_status = %d != "
-						 "AMDGPU_FLIP_SUBMITTED(%d)\n",
-						 amdgpu_crtc->pflip_status,
-						 AMDGPU_FLIP_SUBMITTED);
-		spin_unlock_irqrestore(&adev_to_drm(adev)->event_lock, flags);
-		return 0;
-	}
-
-	/* page flip completed. clean up */
-	amdgpu_crtc->pflip_status = AMDGPU_FLIP_NONE;
-	amdgpu_crtc->pflip_works = NULL;
-
-	/* wakeup usersapce */
-	if(works->event)
-		drm_crtc_send_vblank_event(&amdgpu_crtc->base, works->event);
-
-	spin_unlock_irqrestore(&adev_to_drm(adev)->event_lock, flags);
-
-	drm_crtc_vblank_put(&amdgpu_crtc->base);
-	schedule_work(&works->unpin_work);
-
-	return 0;
-}
-
-static void dce_v11_0_hpd_int_ack(struct amdgpu_device *adev,
-				  int hpd)
-{
-	u32 tmp;
-
-	if (hpd >= adev->mode_info.num_hpd) {
-		DRM_DEBUG("invalid hdp %d\n", hpd);
-		return;
-	}
-
-	tmp = RREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[hpd]);
-	tmp = REG_SET_FIELD(tmp, DC_HPD_INT_CONTROL, DC_HPD_INT_ACK, 1);
-	WREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[hpd], tmp);
-}
-
-static void dce_v11_0_crtc_vblank_int_ack(struct amdgpu_device *adev,
-					  int crtc)
-{
-	u32 tmp;
-
-	if (crtc < 0 || crtc >= adev->mode_info.num_crtc) {
-		DRM_DEBUG("invalid crtc %d\n", crtc);
-		return;
-	}
-
-	tmp = RREG32(mmLB_VBLANK_STATUS + crtc_offsets[crtc]);
-	tmp = REG_SET_FIELD(tmp, LB_VBLANK_STATUS, VBLANK_ACK, 1);
-	WREG32(mmLB_VBLANK_STATUS + crtc_offsets[crtc], tmp);
-}
-
-static void dce_v11_0_crtc_vline_int_ack(struct amdgpu_device *adev,
-					 int crtc)
-{
-	u32 tmp;
-
-	if (crtc < 0 || crtc >= adev->mode_info.num_crtc) {
-		DRM_DEBUG("invalid crtc %d\n", crtc);
-		return;
-	}
-
-	tmp = RREG32(mmLB_VLINE_STATUS + crtc_offsets[crtc]);
-	tmp = REG_SET_FIELD(tmp, LB_VLINE_STATUS, VLINE_ACK, 1);
-	WREG32(mmLB_VLINE_STATUS + crtc_offsets[crtc], tmp);
-}
-
-static int dce_v11_0_crtc_irq(struct amdgpu_device *adev,
-				struct amdgpu_irq_src *source,
-				struct amdgpu_iv_entry *entry)
-{
-	unsigned crtc = entry->src_id - 1;
-	uint32_t disp_int = RREG32(interrupt_status_offsets[crtc].reg);
-	unsigned int irq_type = amdgpu_display_crtc_idx_to_irq_type(adev,
-								    crtc);
-
-	switch (entry->src_data[0]) {
-	case 0: /* vblank */
-		if (disp_int & interrupt_status_offsets[crtc].vblank)
-			dce_v11_0_crtc_vblank_int_ack(adev, crtc);
-		else
-			DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
-
-		if (amdgpu_irq_enabled(adev, source, irq_type)) {
-			drm_handle_vblank(adev_to_drm(adev), crtc);
-		}
-		DRM_DEBUG("IH: D%d vblank\n", crtc + 1);
-
-		break;
-	case 1: /* vline */
-		if (disp_int & interrupt_status_offsets[crtc].vline)
-			dce_v11_0_crtc_vline_int_ack(adev, crtc);
-		else
-			DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
-
-		DRM_DEBUG("IH: D%d vline\n", crtc + 1);
-
-		break;
-	default:
-		DRM_DEBUG("Unhandled interrupt: %d %d\n", entry->src_id, entry->src_data[0]);
-		break;
-	}
-
-	return 0;
-}
-
-static int dce_v11_0_hpd_irq(struct amdgpu_device *adev,
-			     struct amdgpu_irq_src *source,
-			     struct amdgpu_iv_entry *entry)
-{
-	uint32_t disp_int, mask;
-	unsigned hpd;
-
-	if (entry->src_data[0] >= adev->mode_info.num_hpd) {
-		DRM_DEBUG("Unhandled interrupt: %d %d\n", entry->src_id, entry->src_data[0]);
-		return 0;
-	}
-
-	hpd = entry->src_data[0];
-	disp_int = RREG32(interrupt_status_offsets[hpd].reg);
-	mask = interrupt_status_offsets[hpd].hpd;
-
-	if (disp_int & mask) {
-		dce_v11_0_hpd_int_ack(adev, hpd);
-		schedule_delayed_work(&adev->hotplug_work, 0);
-		DRM_DEBUG("IH: HPD%d\n", hpd + 1);
-	}
-
-	return 0;
-}
-
-static int dce_v11_0_set_clockgating_state(void *handle,
-					  enum amd_clockgating_state state)
-{
-	return 0;
-}
-
-static int dce_v11_0_set_powergating_state(void *handle,
-					  enum amd_powergating_state state)
-{
-	return 0;
-}
-
-static const struct amd_ip_funcs dce_v11_0_ip_funcs = {
-	.name = "dce_v11_0",
-	.early_init = dce_v11_0_early_init,
-	.late_init = NULL,
-	.sw_init = dce_v11_0_sw_init,
-	.sw_fini = dce_v11_0_sw_fini,
-	.hw_init = dce_v11_0_hw_init,
-	.hw_fini = dce_v11_0_hw_fini,
-	.suspend = dce_v11_0_suspend,
-	.resume = dce_v11_0_resume,
-	.is_idle = dce_v11_0_is_idle,
-	.wait_for_idle = dce_v11_0_wait_for_idle,
-	.soft_reset = dce_v11_0_soft_reset,
-	.set_clockgating_state = dce_v11_0_set_clockgating_state,
-	.set_powergating_state = dce_v11_0_set_powergating_state,
-};
-
-static void
-dce_v11_0_encoder_mode_set(struct drm_encoder *encoder,
-			  struct drm_display_mode *mode,
-			  struct drm_display_mode *adjusted_mode)
-{
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-
-	amdgpu_encoder->pixel_clock = adjusted_mode->clock;
-
-	/* need to call this here rather than in prepare() since we need some crtc info */
-	amdgpu_atombios_encoder_dpms(encoder, DRM_MODE_DPMS_OFF);
-
-	/* set scaler clears this on some chips */
-	dce_v11_0_set_interleave(encoder->crtc, mode);
-
-	if (amdgpu_atombios_encoder_get_encoder_mode(encoder) == ATOM_ENCODER_MODE_HDMI) {
-		dce_v11_0_afmt_enable(encoder, true);
-		dce_v11_0_afmt_setmode(encoder, adjusted_mode);
-	}
-}
-
-static void dce_v11_0_encoder_prepare(struct drm_encoder *encoder)
-{
-	struct amdgpu_device *adev = drm_to_adev(encoder->dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct drm_connector *connector = amdgpu_get_connector_for_encoder(encoder);
-
-	if ((amdgpu_encoder->active_device &
-	     (ATOM_DEVICE_DFP_SUPPORT | ATOM_DEVICE_LCD_SUPPORT)) ||
-	    (amdgpu_encoder_get_dp_bridge_encoder_id(encoder) !=
-	     ENCODER_OBJECT_ID_NONE)) {
-		struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-		if (dig) {
-			dig->dig_encoder = dce_v11_0_pick_dig_encoder(encoder);
-			if (amdgpu_encoder->active_device & ATOM_DEVICE_DFP_SUPPORT)
-				dig->afmt = adev->mode_info.afmt[dig->dig_encoder];
-		}
-	}
-
-	amdgpu_atombios_scratch_regs_lock(adev, true);
-
-	if (connector) {
-		struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
-
-		/* select the clock/data port if it uses a router */
-		if (amdgpu_connector->router.cd_valid)
-			amdgpu_i2c_router_select_cd_port(amdgpu_connector);
-
-		/* turn eDP panel on for mode set */
-		if (connector->connector_type == DRM_MODE_CONNECTOR_eDP)
-			amdgpu_atombios_encoder_set_edp_panel_power(connector,
-							     ATOM_TRANSMITTER_ACTION_POWER_ON);
-	}
-
-	/* this is needed for the pll/ss setup to work correctly in some cases */
-	amdgpu_atombios_encoder_set_crtc_source(encoder);
-	/* set up the FMT blocks */
-	dce_v11_0_program_fmt(encoder);
-}
-
-static void dce_v11_0_encoder_commit(struct drm_encoder *encoder)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-
-	/* need to call this here as we need the crtc set up */
-	amdgpu_atombios_encoder_dpms(encoder, DRM_MODE_DPMS_ON);
-	amdgpu_atombios_scratch_regs_lock(adev, false);
-}
-
-static void dce_v11_0_encoder_disable(struct drm_encoder *encoder)
-{
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig;
-
-	amdgpu_atombios_encoder_dpms(encoder, DRM_MODE_DPMS_OFF);
-
-	if (amdgpu_atombios_encoder_is_digital(encoder)) {
-		if (amdgpu_atombios_encoder_get_encoder_mode(encoder) == ATOM_ENCODER_MODE_HDMI)
-			dce_v11_0_afmt_enable(encoder, false);
-		dig = amdgpu_encoder->enc_priv;
-		dig->dig_encoder = -1;
-	}
-	amdgpu_encoder->active_device = 0;
-}
-
-/* these are handled by the primary encoders */
-static void dce_v11_0_ext_prepare(struct drm_encoder *encoder)
-{
-
-}
-
-static void dce_v11_0_ext_commit(struct drm_encoder *encoder)
-{
-
-}
-
-static void
-dce_v11_0_ext_mode_set(struct drm_encoder *encoder,
-		      struct drm_display_mode *mode,
-		      struct drm_display_mode *adjusted_mode)
-{
-
-}
-
-static void dce_v11_0_ext_disable(struct drm_encoder *encoder)
-{
-
-}
-
-static void
-dce_v11_0_ext_dpms(struct drm_encoder *encoder, int mode)
-{
-
-}
-
-static const struct drm_encoder_helper_funcs dce_v11_0_ext_helper_funcs = {
-	.dpms = dce_v11_0_ext_dpms,
-	.prepare = dce_v11_0_ext_prepare,
-	.mode_set = dce_v11_0_ext_mode_set,
-	.commit = dce_v11_0_ext_commit,
-	.disable = dce_v11_0_ext_disable,
-	/* no detect for TMDS/LVDS yet */
-};
-
-static const struct drm_encoder_helper_funcs dce_v11_0_dig_helper_funcs = {
-	.dpms = amdgpu_atombios_encoder_dpms,
-	.mode_fixup = amdgpu_atombios_encoder_mode_fixup,
-	.prepare = dce_v11_0_encoder_prepare,
-	.mode_set = dce_v11_0_encoder_mode_set,
-	.commit = dce_v11_0_encoder_commit,
-	.disable = dce_v11_0_encoder_disable,
-	.detect = amdgpu_atombios_encoder_dig_detect,
-};
-
-static const struct drm_encoder_helper_funcs dce_v11_0_dac_helper_funcs = {
-	.dpms = amdgpu_atombios_encoder_dpms,
-	.mode_fixup = amdgpu_atombios_encoder_mode_fixup,
-	.prepare = dce_v11_0_encoder_prepare,
-	.mode_set = dce_v11_0_encoder_mode_set,
-	.commit = dce_v11_0_encoder_commit,
-	.detect = amdgpu_atombios_encoder_dac_detect,
-};
-
-static void dce_v11_0_encoder_destroy(struct drm_encoder *encoder)
-{
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	if (amdgpu_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT))
-		amdgpu_atombios_encoder_fini_backlight(amdgpu_encoder);
-	kfree(amdgpu_encoder->enc_priv);
-	drm_encoder_cleanup(encoder);
-	kfree(amdgpu_encoder);
-}
-
-static const struct drm_encoder_funcs dce_v11_0_encoder_funcs = {
-	.destroy = dce_v11_0_encoder_destroy,
-};
-
-static void dce_v11_0_encoder_add(struct amdgpu_device *adev,
-				 uint32_t encoder_enum,
-				 uint32_t supported_device,
-				 u16 caps)
-{
-	struct drm_device *dev = adev_to_drm(adev);
-	struct drm_encoder *encoder;
-	struct amdgpu_encoder *amdgpu_encoder;
-
-	/* see if we already added it */
-	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
-		amdgpu_encoder = to_amdgpu_encoder(encoder);
-		if (amdgpu_encoder->encoder_enum == encoder_enum) {
-			amdgpu_encoder->devices |= supported_device;
-			return;
-		}
-
-	}
-
-	/* add a new one */
-	amdgpu_encoder = kzalloc(sizeof(struct amdgpu_encoder), GFP_KERNEL);
-	if (!amdgpu_encoder)
-		return;
-
-	encoder = &amdgpu_encoder->base;
-	switch (adev->mode_info.num_crtc) {
-	case 1:
-		encoder->possible_crtcs = 0x1;
-		break;
-	case 2:
-	default:
-		encoder->possible_crtcs = 0x3;
-		break;
-	case 3:
-		encoder->possible_crtcs = 0x7;
-		break;
-	case 4:
-		encoder->possible_crtcs = 0xf;
-		break;
-	case 5:
-		encoder->possible_crtcs = 0x1f;
-		break;
-	case 6:
-		encoder->possible_crtcs = 0x3f;
-		break;
-	}
-
-	amdgpu_encoder->enc_priv = NULL;
-
-	amdgpu_encoder->encoder_enum = encoder_enum;
-	amdgpu_encoder->encoder_id = (encoder_enum & OBJECT_ID_MASK) >> OBJECT_ID_SHIFT;
-	amdgpu_encoder->devices = supported_device;
-	amdgpu_encoder->rmx_type = RMX_OFF;
-	amdgpu_encoder->underscan_type = UNDERSCAN_OFF;
-	amdgpu_encoder->is_ext_encoder = false;
-	amdgpu_encoder->caps = caps;
-
-	switch (amdgpu_encoder->encoder_id) {
-	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC1:
-	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC2:
-		drm_encoder_init(dev, encoder, &dce_v11_0_encoder_funcs,
-				 DRM_MODE_ENCODER_DAC, NULL);
-		drm_encoder_helper_add(encoder, &dce_v11_0_dac_helper_funcs);
-		break;
-	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DVO1:
-	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
-	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
-	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
-	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY3:
-		if (amdgpu_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT)) {
-			amdgpu_encoder->rmx_type = RMX_FULL;
-			drm_encoder_init(dev, encoder, &dce_v11_0_encoder_funcs,
-					 DRM_MODE_ENCODER_LVDS, NULL);
-			amdgpu_encoder->enc_priv = amdgpu_atombios_encoder_get_lcd_info(amdgpu_encoder);
-		} else if (amdgpu_encoder->devices & (ATOM_DEVICE_CRT_SUPPORT)) {
-			drm_encoder_init(dev, encoder, &dce_v11_0_encoder_funcs,
-					 DRM_MODE_ENCODER_DAC, NULL);
-			amdgpu_encoder->enc_priv = amdgpu_atombios_encoder_get_dig_info(amdgpu_encoder);
-		} else {
-			drm_encoder_init(dev, encoder, &dce_v11_0_encoder_funcs,
-					 DRM_MODE_ENCODER_TMDS, NULL);
-			amdgpu_encoder->enc_priv = amdgpu_atombios_encoder_get_dig_info(amdgpu_encoder);
-		}
-		drm_encoder_helper_add(encoder, &dce_v11_0_dig_helper_funcs);
-		break;
-	case ENCODER_OBJECT_ID_SI170B:
-	case ENCODER_OBJECT_ID_CH7303:
-	case ENCODER_OBJECT_ID_EXTERNAL_SDVOA:
-	case ENCODER_OBJECT_ID_EXTERNAL_SDVOB:
-	case ENCODER_OBJECT_ID_TITFP513:
-	case ENCODER_OBJECT_ID_VT1623:
-	case ENCODER_OBJECT_ID_HDMI_SI1930:
-	case ENCODER_OBJECT_ID_TRAVIS:
-	case ENCODER_OBJECT_ID_NUTMEG:
-		/* these are handled by the primary encoders */
-		amdgpu_encoder->is_ext_encoder = true;
-		if (amdgpu_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT))
-			drm_encoder_init(dev, encoder, &dce_v11_0_encoder_funcs,
-					 DRM_MODE_ENCODER_LVDS, NULL);
-		else if (amdgpu_encoder->devices & (ATOM_DEVICE_CRT_SUPPORT))
-			drm_encoder_init(dev, encoder, &dce_v11_0_encoder_funcs,
-					 DRM_MODE_ENCODER_DAC, NULL);
-		else
-			drm_encoder_init(dev, encoder, &dce_v11_0_encoder_funcs,
-					 DRM_MODE_ENCODER_TMDS, NULL);
-		drm_encoder_helper_add(encoder, &dce_v11_0_ext_helper_funcs);
-		break;
-	}
-}
-
-static const struct amdgpu_display_funcs dce_v11_0_display_funcs = {
-	.bandwidth_update = &dce_v11_0_bandwidth_update,
-	.vblank_get_counter = &dce_v11_0_vblank_get_counter,
-	.backlight_set_level = &amdgpu_atombios_encoder_set_backlight_level,
-	.backlight_get_level = &amdgpu_atombios_encoder_get_backlight_level,
-	.hpd_sense = &dce_v11_0_hpd_sense,
-	.hpd_set_polarity = &dce_v11_0_hpd_set_polarity,
-	.hpd_get_gpio_reg = &dce_v11_0_hpd_get_gpio_reg,
-	.page_flip = &dce_v11_0_page_flip,
-	.page_flip_get_scanoutpos = &dce_v11_0_crtc_get_scanoutpos,
-	.add_encoder = &dce_v11_0_encoder_add,
-	.add_connector = &amdgpu_connector_add,
-};
-
-static void dce_v11_0_set_display_funcs(struct amdgpu_device *adev)
-{
-	adev->mode_info.funcs = &dce_v11_0_display_funcs;
-}
-
-static const struct amdgpu_irq_src_funcs dce_v11_0_crtc_irq_funcs = {
-	.set = dce_v11_0_set_crtc_irq_state,
-	.process = dce_v11_0_crtc_irq,
-};
-
-static const struct amdgpu_irq_src_funcs dce_v11_0_pageflip_irq_funcs = {
-	.set = dce_v11_0_set_pageflip_irq_state,
-	.process = dce_v11_0_pageflip_irq,
-};
-
-static const struct amdgpu_irq_src_funcs dce_v11_0_hpd_irq_funcs = {
-	.set = dce_v11_0_set_hpd_irq_state,
-	.process = dce_v11_0_hpd_irq,
-};
-
-static void dce_v11_0_set_irq_funcs(struct amdgpu_device *adev)
-{
-	if (adev->mode_info.num_crtc > 0)
-		adev->crtc_irq.num_types = AMDGPU_CRTC_IRQ_VLINE1 + adev->mode_info.num_crtc;
-	else
-		adev->crtc_irq.num_types = 0;
-	adev->crtc_irq.funcs = &dce_v11_0_crtc_irq_funcs;
-
-	adev->pageflip_irq.num_types = adev->mode_info.num_crtc;
-	adev->pageflip_irq.funcs = &dce_v11_0_pageflip_irq_funcs;
-
-	adev->hpd_irq.num_types = adev->mode_info.num_hpd;
-	adev->hpd_irq.funcs = &dce_v11_0_hpd_irq_funcs;
-}
-
-const struct amdgpu_ip_block_version dce_v11_0_ip_block =
-{
-	.type = AMD_IP_BLOCK_TYPE_DCE,
-	.major = 11,
-	.minor = 0,
-	.rev = 0,
-	.funcs = &dce_v11_0_ip_funcs,
-};
-
-const struct amdgpu_ip_block_version dce_v11_2_ip_block =
-{
-	.type = AMD_IP_BLOCK_TYPE_DCE,
-	.major = 11,
-	.minor = 2,
-	.rev = 0,
-	.funcs = &dce_v11_0_ip_funcs,
-};
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c
index 7f85ba5b726f..acc887a58518 100644
--- a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c
@@ -23,6 +23,7 @@
 
 #include <linux/pci.h>
 
+#include <drm/drm_edid.h>
 #include <drm/drm_fourcc.h>
 #include <drm/drm_modeset_helper.h>
 #include <drm/drm_modeset_helper_vtables.h>
@@ -39,18 +40,25 @@
 #include "amdgpu_connectors.h"
 #include "amdgpu_display.h"
 
+#include "dce_v6_0.h"
+#include "sid.h"
+
 #include "bif/bif_3_0_d.h"
 #include "bif/bif_3_0_sh_mask.h"
+
 #include "oss/oss_1_0_d.h"
 #include "oss/oss_1_0_sh_mask.h"
+
 #include "gca/gfx_6_0_d.h"
 #include "gca/gfx_6_0_sh_mask.h"
+#include "gca/gfx_7_2_enum.h"
+
 #include "gmc/gmc_6_0_d.h"
 #include "gmc/gmc_6_0_sh_mask.h"
+
 #include "dce/dce_6_0_d.h"
 #include "dce/dce_6_0_sh_mask.h"
-#include "gca/gfx_7_2_enum.h"
-#include "dce_v6_0.h"
+
 #include "si_enums.h"
 
 static void dce_v6_0_set_display_funcs(struct amdgpu_device *adev);
@@ -58,31 +66,31 @@ static void dce_v6_0_set_irq_funcs(struct amdgpu_device *adev);
 
 static const u32 crtc_offsets[6] =
 {
-	SI_CRTC0_REGISTER_OFFSET,
-	SI_CRTC1_REGISTER_OFFSET,
-	SI_CRTC2_REGISTER_OFFSET,
-	SI_CRTC3_REGISTER_OFFSET,
-	SI_CRTC4_REGISTER_OFFSET,
-	SI_CRTC5_REGISTER_OFFSET
+	CRTC0_REGISTER_OFFSET,
+	CRTC1_REGISTER_OFFSET,
+	CRTC2_REGISTER_OFFSET,
+	CRTC3_REGISTER_OFFSET,
+	CRTC4_REGISTER_OFFSET,
+	CRTC5_REGISTER_OFFSET
 };
 
 static const u32 hpd_offsets[] =
 {
-	mmDC_HPD1_INT_STATUS - mmDC_HPD1_INT_STATUS,
-	mmDC_HPD2_INT_STATUS - mmDC_HPD1_INT_STATUS,
-	mmDC_HPD3_INT_STATUS - mmDC_HPD1_INT_STATUS,
-	mmDC_HPD4_INT_STATUS - mmDC_HPD1_INT_STATUS,
-	mmDC_HPD5_INT_STATUS - mmDC_HPD1_INT_STATUS,
-	mmDC_HPD6_INT_STATUS - mmDC_HPD1_INT_STATUS,
+	HPD0_REGISTER_OFFSET,
+	HPD1_REGISTER_OFFSET,
+	HPD2_REGISTER_OFFSET,
+	HPD3_REGISTER_OFFSET,
+	HPD4_REGISTER_OFFSET,
+	HPD5_REGISTER_OFFSET
 };
 
 static const uint32_t dig_offsets[] = {
-	SI_CRTC0_REGISTER_OFFSET,
-	SI_CRTC1_REGISTER_OFFSET,
-	SI_CRTC2_REGISTER_OFFSET,
-	SI_CRTC3_REGISTER_OFFSET,
-	SI_CRTC4_REGISTER_OFFSET,
-	SI_CRTC5_REGISTER_OFFSET,
+	CRTC0_REGISTER_OFFSET,
+	CRTC1_REGISTER_OFFSET,
+	CRTC2_REGISTER_OFFSET,
+	CRTC3_REGISTER_OFFSET,
+	CRTC4_REGISTER_OFFSET,
+	CRTC5_REGISTER_OFFSET,
 	(0x13830 - 0x7030) >> 2,
 };
 
@@ -205,9 +213,9 @@ static void dce_v6_0_page_flip(struct amdgpu_device *adev,
 	/* update the scanout addresses */
 	WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS_HIGH + amdgpu_crtc->crtc_offset,
 	       upper_32_bits(crtc_base));
+	/* writing to the low address triggers the update */
 	WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
 	       (u32)crtc_base);
-
 	/* post the write */
 	RREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset);
 }
@@ -217,11 +225,11 @@ static int dce_v6_0_crtc_get_scanoutpos(struct amdgpu_device *adev, int crtc,
 {
 	if ((crtc < 0) || (crtc >= adev->mode_info.num_crtc))
 		return -EINVAL;
+
 	*vbl = RREG32(mmCRTC_V_BLANK_START_END + crtc_offsets[crtc]);
 	*position = RREG32(mmCRTC_STATUS_POSITION + crtc_offsets[crtc]);
 
 	return 0;
-
 }
 
 /**
@@ -241,7 +249,8 @@ static bool dce_v6_0_hpd_sense(struct amdgpu_device *adev,
 	if (hpd >= adev->mode_info.num_hpd)
 		return connected;
 
-	if (RREG32(mmDC_HPD1_INT_STATUS + hpd_offsets[hpd]) & DC_HPD1_INT_STATUS__DC_HPD1_SENSE_MASK)
+	if (RREG32(mmDC_HPD1_INT_STATUS + hpd_offsets[hpd]) &
+	    DC_HPD1_INT_STATUS__DC_HPD1_SENSE_MASK)
 		connected = true;
 
 	return connected;
@@ -272,6 +281,21 @@ static void dce_v6_0_hpd_set_polarity(struct amdgpu_device *adev,
 	WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd], tmp);
 }
 
+static void dce_v6_0_hpd_int_ack(struct amdgpu_device *adev,
+				 int hpd)
+{
+	u32 tmp;
+
+	if (hpd >= adev->mode_info.num_hpd) {
+		DRM_DEBUG("invalid hpd %d\n", hpd);
+		return;
+	}
+
+	tmp = RREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd]);
+	tmp |= DC_HPD1_INT_CONTROL__DC_HPD1_INT_ACK_MASK;
+	WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd], tmp);
+}
+
 /**
  * dce_v6_0_hpd_init - hpd setup callback.
  *
@@ -311,6 +335,7 @@ static void dce_v6_0_hpd_init(struct amdgpu_device *adev)
 			continue;
 		}
 
+		dce_v6_0_hpd_int_ack(adev, amdgpu_connector->hpd.hpd);
 		dce_v6_0_hpd_set_polarity(adev, amdgpu_connector->hpd.hpd);
 		amdgpu_irq_get(adev, &adev->hpd_irq, amdgpu_connector->hpd.hpd);
 	}
@@ -353,13 +378,41 @@ static u32 dce_v6_0_hpd_get_gpio_reg(struct amdgpu_device *adev)
 	return mmDC_GPIO_HPD_A;
 }
 
+static bool dce_v6_0_is_display_hung(struct amdgpu_device *adev)
+{
+	u32 crtc_hung = 0;
+	u32 crtc_status[6];
+	u32 i, j, tmp;
+
+	for (i = 0; i < adev->mode_info.num_crtc; i++) {
+		if (RREG32(mmCRTC_CONTROL + crtc_offsets[i]) & CRTC_CONTROL__CRTC_MASTER_EN_MASK) {
+			crtc_status[i] = RREG32(mmCRTC_STATUS_HV_COUNT + crtc_offsets[i]);
+			crtc_hung |= (1 << i);
+		}
+	}
+
+	for (j = 0; j < 10; j++) {
+		for (i = 0; i < adev->mode_info.num_crtc; i++) {
+			if (crtc_hung & (1 << i)) {
+				tmp = RREG32(mmCRTC_STATUS_HV_COUNT + crtc_offsets[i]);
+				if (tmp != crtc_status[i])
+					crtc_hung &= ~(1 << i);
+			}
+		}
+		if (crtc_hung == 0)
+			return false;
+		udelay(100);
+	}
+
+	return true;
+}
+
 static void dce_v6_0_set_vga_render_state(struct amdgpu_device *adev,
 					  bool render)
 {
 	if (!render)
 		WREG32(mmVGA_RENDER_CONTROL,
-			RREG32(mmVGA_RENDER_CONTROL) & VGA_VSTATUS_CNTL);
-
+		       RREG32(mmVGA_RENDER_CONTROL) & ~VGA_RENDER_CONTROL__VGA_VSTATUS_CNTL_MASK);
 }
 
 static int dce_v6_0_get_num_crtc(struct amdgpu_device *adev)
@@ -402,7 +455,6 @@ void dce_v6_0_disable_dce(struct amdgpu_device *adev)
 
 static void dce_v6_0_program_fmt(struct drm_encoder *encoder)
 {
-
 	struct drm_device *dev = encoder->dev;
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
@@ -845,7 +897,7 @@ static void dce_v6_0_program_watermarks(struct amdgpu_device *adev,
 					    (u32)mode->clock);
 		line_time = (u32) div_u64((u64)mode->crtc_htotal * 1000000,
 					  (u32)mode->clock);
-		line_time = min(line_time, (u32)65535);
+		line_time = min_t(u32, line_time, 65535);
 		priority_a_cnt = 0;
 		priority_b_cnt = 0;
 
@@ -878,8 +930,8 @@ static void dce_v6_0_program_watermarks(struct amdgpu_device *adev,
 		wm_high.dram_channels = dram_channels;
 		wm_high.num_heads = num_heads;
 
-		if (adev->pm.dpm_enabled) {
 		/* watermark for low clocks */
+		if (adev->pm.dpm_enabled) {
 			wm_low.yclk =
 				amdgpu_dpm_get_mclk(adev, true) * 10;
 			wm_low.sclk =
@@ -906,9 +958,9 @@ static void dce_v6_0_program_watermarks(struct amdgpu_device *adev,
 		wm_low.num_heads = num_heads;
 
 		/* set for high clocks */
-		latency_watermark_a = min(dce_v6_0_latency_watermark(&wm_high), (u32)65535);
+		latency_watermark_a = min_t(u32, dce_v6_0_latency_watermark(&wm_high), 65535);
 		/* set for low clocks */
-		latency_watermark_b = min(dce_v6_0_latency_watermark(&wm_low), (u32)65535);
+		latency_watermark_b = min_t(u32, dce_v6_0_latency_watermark(&wm_low), 65535);
 
 		/* possibly force display priority to high */
 		/* should really do this at mode validation time... */
@@ -959,16 +1011,16 @@ static void dce_v6_0_program_watermarks(struct amdgpu_device *adev,
 	/* select wm A */
 	arb_control3 = RREG32(mmDPG_PIPE_ARBITRATION_CONTROL3 + amdgpu_crtc->crtc_offset);
 	tmp = arb_control3;
-	tmp &= ~LATENCY_WATERMARK_MASK(3);
-	tmp |= LATENCY_WATERMARK_MASK(1);
+	tmp &= ~(3 << DPG_PIPE_ARBITRATION_CONTROL3__URGENCY_WATERMARK_MASK__SHIFT);
+	tmp |= (1 << DPG_PIPE_ARBITRATION_CONTROL3__URGENCY_WATERMARK_MASK__SHIFT);
 	WREG32(mmDPG_PIPE_ARBITRATION_CONTROL3 + amdgpu_crtc->crtc_offset, tmp);
 	WREG32(mmDPG_PIPE_URGENCY_CONTROL + amdgpu_crtc->crtc_offset,
 	       ((latency_watermark_a << DPG_PIPE_URGENCY_CONTROL__URGENCY_LOW_WATERMARK__SHIFT)  |
 		(line_time << DPG_PIPE_URGENCY_CONTROL__URGENCY_HIGH_WATERMARK__SHIFT)));
 	/* select wm B */
 	tmp = RREG32(mmDPG_PIPE_ARBITRATION_CONTROL3 + amdgpu_crtc->crtc_offset);
-	tmp &= ~LATENCY_WATERMARK_MASK(3);
-	tmp |= LATENCY_WATERMARK_MASK(2);
+	tmp &= ~(3 << DPG_PIPE_ARBITRATION_CONTROL3__URGENCY_WATERMARK_MASK__SHIFT);
+	tmp |= (2 << DPG_PIPE_ARBITRATION_CONTROL3__URGENCY_WATERMARK_MASK__SHIFT);
 	WREG32(mmDPG_PIPE_ARBITRATION_CONTROL3 + amdgpu_crtc->crtc_offset, tmp);
 	WREG32(mmDPG_PIPE_URGENCY_CONTROL + amdgpu_crtc->crtc_offset,
 	       ((latency_watermark_b << DPG_PIPE_URGENCY_CONTROL__URGENCY_LOW_WATERMARK__SHIFT) |
@@ -982,13 +1034,26 @@ static void dce_v6_0_program_watermarks(struct amdgpu_device *adev,
 
 	/* save values for DPM */
 	amdgpu_crtc->line_time = line_time;
-	amdgpu_crtc->wm_high = latency_watermark_a;
 
 	/* Save number of lines the linebuffer leads before the scanout */
 	amdgpu_crtc->lb_vblank_lead_lines = lb_vblank_lead_lines;
 }
 
 /* watermark setup */
+/**
+ * dce_v6_0_line_buffer_adjust - Set up the line buffer
+ *
+ * @adev: amdgpu_device pointer
+ * @amdgpu_crtc: the selected display controller
+ * @mode: the current display mode on the selected display
+ * controller
+ * @other_mode: the display mode of another display controller
+ *              that may be sharing the line buffer
+ *
+ * Setup up the line buffer allocation for
+ * the selected display controller (CIK).
+ * Returns the line buffer size in pixels.
+ */
 static u32 dce_v6_0_line_buffer_adjust(struct amdgpu_device *adev,
 				   struct amdgpu_crtc *amdgpu_crtc,
 				   struct drm_display_mode *mode,
@@ -1023,7 +1088,7 @@ static u32 dce_v6_0_line_buffer_adjust(struct amdgpu_device *adev,
 	}
 
 	WREG32(mmDC_LB_MEMORY_SPLIT + amdgpu_crtc->crtc_offset,
-	       DC_LB_MEMORY_CONFIG(tmp));
+	       (tmp << DC_LB_MEMORY_SPLIT__DC_LB_MEMORY_CONFIG__SHIFT));
 
 	WREG32(mmPIPE0_DMIF_BUFFER_CONTROL + pipe_offset,
 	       (buffer_alloc << PIPE0_DMIF_BUFFER_CONTROL__DMIF_BUFFERS_ALLOCATED__SHIFT));
@@ -1200,7 +1265,7 @@ static void dce_v6_0_audio_write_speaker_allocation(struct drm_encoder *encoder)
 		return;
 	}
 
-	sad_count = drm_edid_to_speaker_allocation(amdgpu_connector_edid(connector), &sadb);
+	sad_count = drm_edid_to_speaker_allocation(amdgpu_connector->edid, &sadb);
 	if (sad_count < 0) {
 		DRM_ERROR("Couldn't read Speaker Allocation Data Block: %d\n", sad_count);
 		sad_count = 0;
@@ -1240,6 +1305,7 @@ static void dce_v6_0_audio_write_sad_regs(struct drm_encoder *encoder)
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
 	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
+	u32 offset;
 	struct drm_connector *connector;
 	struct drm_connector_list_iter iter;
 	struct amdgpu_connector *amdgpu_connector = NULL;
@@ -1261,6 +1327,11 @@ static void dce_v6_0_audio_write_sad_regs(struct drm_encoder *encoder)
 		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR13, HDMI_AUDIO_CODING_TYPE_WMA_PRO },
 	};
 
+	if (!dig || !dig->afmt || !dig->afmt->pin)
+		return;
+
+	offset = dig->afmt->pin->offset;
+
 	drm_connector_list_iter_begin(dev, &iter);
 	drm_for_each_connector_iter(connector, &iter) {
 		if (connector->encoder == encoder) {
@@ -1275,14 +1346,14 @@ static void dce_v6_0_audio_write_sad_regs(struct drm_encoder *encoder)
 		return;
 	}
 
-	sad_count = drm_edid_to_sad(amdgpu_connector_edid(connector), &sads);
+	sad_count = drm_edid_to_sad(amdgpu_connector->edid, &sads);
 	if (sad_count < 0)
 		DRM_ERROR("Couldn't read SADs: %d\n", sad_count);
 	if (sad_count <= 0)
 		return;
 
 	for (i = 0; i < ARRAY_SIZE(eld_reg_to_type); i++) {
-		u32 tmp = 0;
+		u32 value = 0;
 		u8 stereo_freqs = 0;
 		int max_channels = -1;
 		int j;
@@ -1292,12 +1363,12 @@ static void dce_v6_0_audio_write_sad_regs(struct drm_encoder *encoder)
 
 			if (sad->format == eld_reg_to_type[i][1]) {
 				if (sad->channels > max_channels) {
-					tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0,
-							MAX_CHANNELS, sad->channels);
-					tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0,
-							DESCRIPTOR_BYTE_2, sad->byte2);
-					tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0,
-							SUPPORTED_FREQUENCIES, sad->freq);
+					value = (sad->channels <<
+						AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__MAX_CHANNELS__SHIFT) |
+					       (sad->byte2 <<
+						AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__DESCRIPTOR_BYTE_2__SHIFT) |
+					       (sad->freq <<
+						AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__SUPPORTED_FREQUENCIES__SHIFT);
 					max_channels = sad->channels;
 				}
 
@@ -1308,13 +1379,13 @@ static void dce_v6_0_audio_write_sad_regs(struct drm_encoder *encoder)
 			}
 		}
 
-		tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0,
-				SUPPORTED_FREQUENCIES_STEREO, stereo_freqs);
-		WREG32_AUDIO_ENDPT(dig->afmt->pin->offset, eld_reg_to_type[i][0], tmp);
+		value |= (stereo_freqs <<
+			AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__SUPPORTED_FREQUENCIES_STEREO__SHIFT);
+
+		WREG32_AUDIO_ENDPT(offset, eld_reg_to_type[i][0], value);
 	}
 
 	kfree(sads);
-
 }
 
 static void dce_v6_0_audio_enable(struct amdgpu_device *adev,
@@ -1330,13 +1401,13 @@ static void dce_v6_0_audio_enable(struct amdgpu_device *adev,
 
 static const u32 pin_offsets[7] =
 {
-	(0x1780 - 0x1780),
-	(0x1786 - 0x1780),
-	(0x178c - 0x1780),
-	(0x1792 - 0x1780),
-	(0x1798 - 0x1780),
-	(0x179d - 0x1780),
-	(0x17a4 - 0x1780),
+	AUD0_REGISTER_OFFSET,
+	AUD1_REGISTER_OFFSET,
+	AUD2_REGISTER_OFFSET,
+	AUD3_REGISTER_OFFSET,
+	AUD4_REGISTER_OFFSET,
+	AUD5_REGISTER_OFFSET,
+	AUD6_REGISTER_OFFSET,
 };
 
 static int dce_v6_0_audio_init(struct amdgpu_device *adev)
@@ -1369,6 +1440,8 @@ static int dce_v6_0_audio_init(struct amdgpu_device *adev)
 		adev->mode_info.audio.pin[i].connected = false;
 		adev->mode_info.audio.pin[i].offset = pin_offsets[i];
 		adev->mode_info.audio.pin[i].id = i;
+		/* disable audio.  it will be set up later */
+		/* XXX remove once we switch to ip funcs */
 		dce_v6_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
 	}
 
@@ -1377,17 +1450,12 @@ static int dce_v6_0_audio_init(struct amdgpu_device *adev)
 
 static void dce_v6_0_audio_fini(struct amdgpu_device *adev)
 {
-	int i;
-
 	if (!amdgpu_audio)
 		return;
 
 	if (!adev->mode_info.audio.enabled)
 		return;
 
-	for (i = 0; i < adev->mode_info.audio.num_pins; i++)
-		dce_v6_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
-
 	adev->mode_info.audio.enabled = false;
 }
 
@@ -1818,7 +1886,7 @@ static int dce_v6_0_crtc_do_set_base(struct drm_crtc *crtc,
 	struct amdgpu_bo *abo;
 	uint64_t fb_location, tiling_flags;
 	uint32_t fb_format, fb_pitch_pixels, pipe_config;
-	u32 fb_swap = GRPH_ENDIAN_SWAP(GRPH_ENDIAN_NONE);
+	u32 fb_swap = (GRPH_ENDIAN_NONE << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 	u32 viewport_w, viewport_h;
 	int r;
 	bool bypass_lut = false;
@@ -1844,6 +1912,7 @@ static int dce_v6_0_crtc_do_set_base(struct drm_crtc *crtc,
 		return r;
 
 	if (!atomic) {
+		abo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 		r = amdgpu_bo_pin(abo, AMDGPU_GEM_DOMAIN_VRAM);
 		if (unlikely(r != 0)) {
 			amdgpu_bo_unreserve(abo);
@@ -1857,76 +1926,76 @@ static int dce_v6_0_crtc_do_set_base(struct drm_crtc *crtc,
 
 	switch (target_fb->format->format) {
 	case DRM_FORMAT_C8:
-		fb_format = (GRPH_DEPTH(GRPH_DEPTH_8BPP) |
-			     GRPH_FORMAT(GRPH_FORMAT_INDEXED));
+		fb_format = ((GRPH_DEPTH_8BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_INDEXED << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
 		break;
 	case DRM_FORMAT_XRGB4444:
 	case DRM_FORMAT_ARGB4444:
-		fb_format = (GRPH_DEPTH(GRPH_DEPTH_16BPP) |
-			     GRPH_FORMAT(GRPH_FORMAT_ARGB4444));
+		fb_format = ((GRPH_DEPTH_16BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_ARGB4444 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
 #ifdef __BIG_ENDIAN
-		fb_swap = GRPH_ENDIAN_SWAP(GRPH_ENDIAN_8IN16);
+		fb_swap = (GRPH_ENDIAN_8IN16 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 #endif
 		break;
 	case DRM_FORMAT_XRGB1555:
 	case DRM_FORMAT_ARGB1555:
-		fb_format = (GRPH_DEPTH(GRPH_DEPTH_16BPP) |
-			     GRPH_FORMAT(GRPH_FORMAT_ARGB1555));
+		fb_format = ((GRPH_DEPTH_16BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_ARGB1555 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
 #ifdef __BIG_ENDIAN
-		fb_swap = GRPH_ENDIAN_SWAP(GRPH_ENDIAN_8IN16);
+		fb_swap = (GRPH_ENDIAN_8IN16 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 #endif
 		break;
 	case DRM_FORMAT_BGRX5551:
 	case DRM_FORMAT_BGRA5551:
-		fb_format = (GRPH_DEPTH(GRPH_DEPTH_16BPP) |
-			     GRPH_FORMAT(GRPH_FORMAT_BGRA5551));
+		fb_format = ((GRPH_DEPTH_16BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_BGRA5551 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
 #ifdef __BIG_ENDIAN
-		fb_swap = GRPH_ENDIAN_SWAP(GRPH_ENDIAN_8IN16);
+		fb_swap = (GRPH_ENDIAN_8IN16 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 #endif
 		break;
 	case DRM_FORMAT_RGB565:
-		fb_format = (GRPH_DEPTH(GRPH_DEPTH_16BPP) |
-			     GRPH_FORMAT(GRPH_FORMAT_ARGB565));
+		fb_format = ((GRPH_DEPTH_16BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_ARGB565 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
 #ifdef __BIG_ENDIAN
-		fb_swap = GRPH_ENDIAN_SWAP(GRPH_ENDIAN_8IN16);
+		fb_swap = (GRPH_ENDIAN_8IN16 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 #endif
 		break;
 	case DRM_FORMAT_XRGB8888:
 	case DRM_FORMAT_ARGB8888:
-		fb_format = (GRPH_DEPTH(GRPH_DEPTH_32BPP) |
-			     GRPH_FORMAT(GRPH_FORMAT_ARGB8888));
+		fb_format = ((GRPH_DEPTH_32BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_ARGB8888 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
 #ifdef __BIG_ENDIAN
-		fb_swap = GRPH_ENDIAN_SWAP(GRPH_ENDIAN_8IN32);
+		fb_swap = (GRPH_ENDIAN_8IN32 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 #endif
 		break;
 	case DRM_FORMAT_XRGB2101010:
 	case DRM_FORMAT_ARGB2101010:
-		fb_format = (GRPH_DEPTH(GRPH_DEPTH_32BPP) |
-			     GRPH_FORMAT(GRPH_FORMAT_ARGB2101010));
+		fb_format = ((GRPH_DEPTH_32BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_ARGB2101010 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
 #ifdef __BIG_ENDIAN
-		fb_swap = GRPH_ENDIAN_SWAP(GRPH_ENDIAN_8IN32);
+		fb_swap = (GRPH_ENDIAN_8IN32 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 #endif
 		/* Greater 8 bpc fb needs to bypass hw-lut to retain precision */
 		bypass_lut = true;
 		break;
 	case DRM_FORMAT_BGRX1010102:
 	case DRM_FORMAT_BGRA1010102:
-		fb_format = (GRPH_DEPTH(GRPH_DEPTH_32BPP) |
-			     GRPH_FORMAT(GRPH_FORMAT_BGRA1010102));
+		fb_format = ((GRPH_DEPTH_32BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_BGRA1010102 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
 #ifdef __BIG_ENDIAN
-		fb_swap = GRPH_ENDIAN_SWAP(GRPH_ENDIAN_8IN32);
+		fb_swap = (GRPH_ENDIAN_8IN32 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 #endif
 		/* Greater 8 bpc fb needs to bypass hw-lut to retain precision */
 		bypass_lut = true;
 		break;
 	case DRM_FORMAT_XBGR8888:
 	case DRM_FORMAT_ABGR8888:
-		fb_format = (GRPH_DEPTH(GRPH_DEPTH_32BPP) |
-			     GRPH_FORMAT(GRPH_FORMAT_ARGB8888));
-		fb_swap = (GRPH_RED_CROSSBAR(GRPH_RED_SEL_B) |
-			   GRPH_BLUE_CROSSBAR(GRPH_BLUE_SEL_R));
+		fb_format = ((GRPH_DEPTH_32BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_ARGB8888 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
+		fb_swap = ((GRPH_RED_SEL_B << GRPH_SWAP_CNTL__GRPH_RED_CROSSBAR__SHIFT) |
+			   (GRPH_BLUE_SEL_R << GRPH_SWAP_CNTL__GRPH_BLUE_CROSSBAR__SHIFT));
 #ifdef __BIG_ENDIAN
-		fb_swap |= GRPH_ENDIAN_SWAP(GRPH_ENDIAN_8IN32);
+		fb_swap |= (GRPH_ENDIAN_8IN32 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 #endif
 		break;
 	default:
@@ -1944,18 +2013,18 @@ static int dce_v6_0_crtc_do_set_base(struct drm_crtc *crtc,
 		tile_split = AMDGPU_TILING_GET(tiling_flags, TILE_SPLIT);
 		num_banks = AMDGPU_TILING_GET(tiling_flags, NUM_BANKS);
 
-		fb_format |= GRPH_NUM_BANKS(num_banks);
-		fb_format |= GRPH_ARRAY_MODE(GRPH_ARRAY_2D_TILED_THIN1);
-		fb_format |= GRPH_TILE_SPLIT(tile_split);
-		fb_format |= GRPH_BANK_WIDTH(bankw);
-		fb_format |= GRPH_BANK_HEIGHT(bankh);
-		fb_format |= GRPH_MACRO_TILE_ASPECT(mtaspect);
+		fb_format |= (num_banks << GRPH_CONTROL__GRPH_NUM_BANKS__SHIFT);
+		fb_format |= (GRPH_ARRAY_2D_TILED_THIN1 << GRPH_CONTROL__GRPH_ARRAY_MODE__SHIFT);
+		fb_format |= (tile_split << GRPH_CONTROL__GRPH_TILE_SPLIT__SHIFT);
+		fb_format |= (bankw << GRPH_CONTROL__GRPH_BANK_WIDTH__SHIFT);
+		fb_format |= (bankh << GRPH_CONTROL__GRPH_BANK_HEIGHT__SHIFT);
+		fb_format |= (mtaspect << GRPH_CONTROL__GRPH_MACRO_TILE_ASPECT__SHIFT);
 	} else if (AMDGPU_TILING_GET(tiling_flags, ARRAY_MODE) == ARRAY_1D_TILED_THIN1) {
-		fb_format |= GRPH_ARRAY_MODE(GRPH_ARRAY_1D_TILED_THIN1);
+		fb_format |= (GRPH_ARRAY_1D_TILED_THIN1 << GRPH_CONTROL__GRPH_ARRAY_MODE__SHIFT);
 	}
 
 	pipe_config = AMDGPU_TILING_GET(tiling_flags, PIPE_CONFIG);
-	fb_format |= GRPH_PIPE_CONFIG(pipe_config);
+	fb_format |= (pipe_config << GRPH_CONTROL__GRPH_PIPE_CONFIG__SHIFT);
 
 	dce_v6_0_vga_enable(crtc, false);
 
@@ -1971,7 +2040,7 @@ static int dce_v6_0_crtc_do_set_base(struct drm_crtc *crtc,
 	WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
 	       (u32)fb_location & GRPH_PRIMARY_SURFACE_ADDRESS__GRPH_PRIMARY_SURFACE_ADDRESS_MASK);
 	WREG32(mmGRPH_SECONDARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
-	       (u32) fb_location & GRPH_PRIMARY_SURFACE_ADDRESS__GRPH_PRIMARY_SURFACE_ADDRESS_MASK);
+	       (u32) fb_location & GRPH_SECONDARY_SURFACE_ADDRESS__GRPH_SECONDARY_SURFACE_ADDRESS_MASK);
 	WREG32(mmGRPH_CONTROL + amdgpu_crtc->crtc_offset, fb_format);
 	WREG32(mmGRPH_SWAP_CNTL + amdgpu_crtc->crtc_offset, fb_swap);
 
@@ -2039,14 +2108,13 @@ static void dce_v6_0_set_interleave(struct drm_crtc *crtc,
 
 	if (mode->flags & DRM_MODE_FLAG_INTERLACE)
 		WREG32(mmDATA_FORMAT + amdgpu_crtc->crtc_offset,
-		       INTERLEAVE_EN);
+			DATA_FORMAT__INTERLEAVE_EN_MASK);
 	else
 		WREG32(mmDATA_FORMAT + amdgpu_crtc->crtc_offset, 0);
 }
 
 static void dce_v6_0_crtc_load_lut(struct drm_crtc *crtc)
 {
-
 	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
 	struct drm_device *dev = crtc->dev;
 	struct amdgpu_device *adev = drm_to_adev(dev);
@@ -2056,15 +2124,15 @@ static void dce_v6_0_crtc_load_lut(struct drm_crtc *crtc)
 	DRM_DEBUG_KMS("%d\n", amdgpu_crtc->crtc_id);
 
 	WREG32(mmINPUT_CSC_CONTROL + amdgpu_crtc->crtc_offset,
-	       ((0 << INPUT_CSC_CONTROL__INPUT_CSC_GRPH_MODE__SHIFT) |
-		(0 << INPUT_CSC_CONTROL__INPUT_CSC_OVL_MODE__SHIFT)));
+	       ((INPUT_CSC_BYPASS << INPUT_CSC_CONTROL__INPUT_CSC_GRPH_MODE__SHIFT) |
+		(INPUT_CSC_BYPASS << INPUT_CSC_CONTROL__INPUT_CSC_OVL_MODE__SHIFT)));
 	WREG32(mmPRESCALE_GRPH_CONTROL + amdgpu_crtc->crtc_offset,
 	       PRESCALE_GRPH_CONTROL__GRPH_PRESCALE_BYPASS_MASK);
 	WREG32(mmPRESCALE_OVL_CONTROL + amdgpu_crtc->crtc_offset,
 	       PRESCALE_OVL_CONTROL__OVL_PRESCALE_BYPASS_MASK);
 	WREG32(mmINPUT_GAMMA_CONTROL + amdgpu_crtc->crtc_offset,
-	       ((0 << INPUT_GAMMA_CONTROL__GRPH_INPUT_GAMMA_MODE__SHIFT) |
-		(0 << INPUT_GAMMA_CONTROL__OVL_INPUT_GAMMA_MODE__SHIFT)));
+	       ((INPUT_GAMMA_USE_LUT << INPUT_GAMMA_CONTROL__GRPH_INPUT_GAMMA_MODE__SHIFT) |
+		(INPUT_GAMMA_USE_LUT << INPUT_GAMMA_CONTROL__OVL_INPUT_GAMMA_MODE__SHIFT)));
 
 	WREG32(mmDC_LUT_CONTROL + amdgpu_crtc->crtc_offset, 0);
 
@@ -2091,19 +2159,19 @@ static void dce_v6_0_crtc_load_lut(struct drm_crtc *crtc)
 	}
 
 	WREG32(mmDEGAMMA_CONTROL + amdgpu_crtc->crtc_offset,
-	       ((0 << DEGAMMA_CONTROL__GRPH_DEGAMMA_MODE__SHIFT) |
-		(0 << DEGAMMA_CONTROL__OVL_DEGAMMA_MODE__SHIFT) |
-		ICON_DEGAMMA_MODE(0) |
-		(0 << DEGAMMA_CONTROL__CURSOR_DEGAMMA_MODE__SHIFT)));
+	       ((DEGAMMA_BYPASS << DEGAMMA_CONTROL__GRPH_DEGAMMA_MODE__SHIFT) |
+		(DEGAMMA_BYPASS << DEGAMMA_CONTROL__OVL_DEGAMMA_MODE__SHIFT) |
+		(DEGAMMA_BYPASS << DEGAMMA_CONTROL__ICON_DEGAMMA_MODE__SHIFT) |
+		(DEGAMMA_BYPASS << DEGAMMA_CONTROL__CURSOR_DEGAMMA_MODE__SHIFT)));
 	WREG32(mmGAMUT_REMAP_CONTROL + amdgpu_crtc->crtc_offset,
-	       ((0 << GAMUT_REMAP_CONTROL__GRPH_GAMUT_REMAP_MODE__SHIFT) |
-		(0 << GAMUT_REMAP_CONTROL__OVL_GAMUT_REMAP_MODE__SHIFT)));
+	       ((GAMUT_REMAP_BYPASS << GAMUT_REMAP_CONTROL__GRPH_GAMUT_REMAP_MODE__SHIFT) |
+		(GAMUT_REMAP_BYPASS << GAMUT_REMAP_CONTROL__OVL_GAMUT_REMAP_MODE__SHIFT)));
 	WREG32(mmREGAMMA_CONTROL + amdgpu_crtc->crtc_offset,
-	       ((0 << REGAMMA_CONTROL__GRPH_REGAMMA_MODE__SHIFT) |
-		(0 << REGAMMA_CONTROL__OVL_REGAMMA_MODE__SHIFT)));
+	       ((REGAMMA_BYPASS << REGAMMA_CONTROL__GRPH_REGAMMA_MODE__SHIFT) |
+		(REGAMMA_BYPASS << REGAMMA_CONTROL__OVL_REGAMMA_MODE__SHIFT)));
 	WREG32(mmOUTPUT_CSC_CONTROL + amdgpu_crtc->crtc_offset,
-	       ((0 << OUTPUT_CSC_CONTROL__OUTPUT_CSC_GRPH_MODE__SHIFT) |
-		(0 << OUTPUT_CSC_CONTROL__OUTPUT_CSC_OVL_MODE__SHIFT)));
+	       ((OUTPUT_CSC_BYPASS << OUTPUT_CSC_CONTROL__OUTPUT_CSC_GRPH_MODE__SHIFT) |
+		(OUTPUT_CSC_BYPASS << OUTPUT_CSC_CONTROL__OUTPUT_CSC_OVL_MODE__SHIFT)));
 	/* XXX match this to the depth of the crtc fmt block, move to modeset? */
 	WREG32(0x1a50 + amdgpu_crtc->crtc_offset, 0);
 
@@ -2198,8 +2266,6 @@ static void dce_v6_0_hide_cursor(struct drm_crtc *crtc)
 	WREG32(mmCUR_CONTROL + amdgpu_crtc->crtc_offset,
 	       (CURSOR_24_8_PRE_MULT << CUR_CONTROL__CURSOR_MODE__SHIFT) |
 	       (CURSOR_URGENT_1_2 << CUR_CONTROL__CURSOR_URGENT_CONTROL__SHIFT));
-
-
 }
 
 static void dce_v6_0_show_cursor(struct drm_crtc *crtc)
@@ -2216,7 +2282,6 @@ static void dce_v6_0_show_cursor(struct drm_crtc *crtc)
 	       CUR_CONTROL__CURSOR_EN_MASK |
 	       (CURSOR_24_8_PRE_MULT << CUR_CONTROL__CURSOR_MODE__SHIFT) |
 	       (CURSOR_URGENT_1_2 << CUR_CONTROL__CURSOR_URGENT_CONTROL__SHIFT));
-
 }
 
 static int dce_v6_0_cursor_move_locked(struct drm_crtc *crtc,
@@ -2304,6 +2369,7 @@ static int dce_v6_0_crtc_cursor_set2(struct drm_crtc *crtc,
 		return ret;
 	}
 
+	aobj->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 	ret = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM);
 	amdgpu_bo_unreserve(aobj);
 	if (ret) {
@@ -2526,7 +2592,6 @@ static bool dce_v6_0_crtc_mode_fixup(struct drm_crtc *crtc,
 				     const struct drm_display_mode *mode,
 				     struct drm_display_mode *adjusted_mode)
 {
-
 	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
 	struct drm_device *dev = crtc->dev;
 	struct drm_encoder *encoder;
@@ -2583,6 +2648,32 @@ static const struct drm_crtc_helper_funcs dce_v6_0_crtc_helper_funcs = {
 	.get_scanout_position = amdgpu_crtc_get_scanout_position,
 };
 
+static void dce_v6_0_panic_flush(struct drm_plane *plane)
+{
+	struct drm_framebuffer *fb;
+	struct amdgpu_crtc *amdgpu_crtc;
+	struct amdgpu_device *adev;
+	uint32_t fb_format;
+
+	if (!plane->fb)
+		return;
+
+	fb = plane->fb;
+	amdgpu_crtc = to_amdgpu_crtc(plane->crtc);
+	adev = drm_to_adev(fb->dev);
+
+	/* Disable DC tiling */
+	fb_format = RREG32(mmGRPH_CONTROL + amdgpu_crtc->crtc_offset);
+	fb_format &= ~GRPH_CONTROL__GRPH_ARRAY_MODE_MASK;
+	WREG32(mmGRPH_CONTROL + amdgpu_crtc->crtc_offset, fb_format);
+
+}
+
+static const struct drm_plane_helper_funcs dce_v6_0_drm_primary_plane_helper_funcs = {
+	.get_scanout_buffer = amdgpu_display_get_scanout_buffer,
+	.panic_flush = dce_v6_0_panic_flush,
+};
+
 static int dce_v6_0_crtc_init(struct amdgpu_device *adev, int index)
 {
 	struct amdgpu_crtc *amdgpu_crtc;
@@ -2610,13 +2701,14 @@ static int dce_v6_0_crtc_init(struct amdgpu_device *adev, int index)
 	amdgpu_crtc->encoder = NULL;
 	amdgpu_crtc->connector = NULL;
 	drm_crtc_helper_add(&amdgpu_crtc->base, &dce_v6_0_crtc_helper_funcs);
+	drm_plane_helper_add(amdgpu_crtc->base.primary, &dce_v6_0_drm_primary_plane_helper_funcs);
 
 	return 0;
 }
 
-static int dce_v6_0_early_init(void *handle)
+static int dce_v6_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->audio_endpt_rreg = &dce_v6_0_audio_endpt_rreg;
 	adev->audio_endpt_wreg = &dce_v6_0_audio_endpt_wreg;
@@ -2645,11 +2737,10 @@ static int dce_v6_0_early_init(void *handle)
 	return 0;
 }
 
-static int dce_v6_0_sw_init(void *handle)
+static int dce_v6_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r, i;
-	bool ret;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->mode_info.num_crtc; i++) {
 		r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i + 1, &adev->crtc_irq);
@@ -2692,8 +2783,7 @@ static int dce_v6_0_sw_init(void *handle)
 			return r;
 	}
 
-	ret = amdgpu_atombios_get_connector_info_from_object_table(adev);
-	if (ret)
+	if (amdgpu_atombios_get_connector_info_from_object_table(adev))
 		amdgpu_display_print_display_setup(adev_to_drm(adev));
 	else
 		return -EINVAL;
@@ -2724,11 +2814,11 @@ static int dce_v6_0_sw_init(void *handle)
 	return r;
 }
 
-static int dce_v6_0_sw_fini(void *handle)
+static int dce_v6_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	kfree(adev->mode_info.bios_hardcoded_edid);
+	drm_edid_free(adev->mode_info.bios_hardcoded_edid);
 
 	drm_kms_helper_poll_fini(adev_to_drm(adev));
 
@@ -2741,10 +2831,10 @@ static int dce_v6_0_sw_fini(void *handle)
 	return 0;
 }
 
-static int dce_v6_0_hw_init(void *handle)
+static int dce_v6_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* disable vga render */
 	dce_v6_0_set_vga_render_state(adev, false);
@@ -2764,10 +2854,10 @@ static int dce_v6_0_hw_init(void *handle)
 	return 0;
 }
 
-static int dce_v6_0_hw_fini(void *handle)
+static int dce_v6_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	dce_v6_0_hpd_fini(adev);
 
@@ -2782,9 +2872,9 @@ static int dce_v6_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int dce_v6_0_suspend(void *handle)
+static int dce_v6_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_display_suspend_helper(adev);
@@ -2793,18 +2883,18 @@ static int dce_v6_0_suspend(void *handle)
 	adev->mode_info.bl_level =
 		amdgpu_atombios_encoder_get_backlight_level_from_reg(adev);
 
-	return dce_v6_0_hw_fini(handle);
+	return dce_v6_0_hw_fini(ip_block);
 }
 
-static int dce_v6_0_resume(void *handle)
+static int dce_v6_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	amdgpu_atombios_encoder_set_backlight_level_to_reg(adev,
 							   adev->mode_info.bl_level);
 
-	ret = dce_v6_0_hw_init(handle);
+	ret = dce_v6_0_hw_init(ip_block);
 
 	/* turn on the BL */
 	if (adev->mode_info.bl_encoder) {
@@ -2819,19 +2909,35 @@ static int dce_v6_0_resume(void *handle)
 	return amdgpu_display_resume_helper(adev);
 }
 
-static bool dce_v6_0_is_idle(void *handle)
+static bool dce_v6_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	return true;
 }
 
-static int dce_v6_0_wait_for_idle(void *handle)
+static int dce_v6_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	return 0;
-}
+	u32 srbm_soft_reset = 0, tmp;
+	struct amdgpu_device *adev = ip_block->adev;
 
-static int dce_v6_0_soft_reset(void *handle)
-{
-	DRM_INFO("xxxx: dce_v6_0_soft_reset --- no impl!!\n");
+	if (dce_v6_0_is_display_hung(adev))
+		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_DC_MASK;
+
+	if (srbm_soft_reset) {
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+		tmp |= srbm_soft_reset;
+		dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp);
+		WREG32(mmSRBM_SOFT_RESET, tmp);
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+
+		udelay(50);
+
+		tmp &= ~srbm_soft_reset;
+		WREG32(mmSRBM_SOFT_RESET, tmp);
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+
+		/* Wait a little for things to settle down */
+		udelay(50);
+	}
 	return 0;
 }
 
@@ -2848,22 +2954,22 @@ static void dce_v6_0_set_crtc_vblank_interrupt_state(struct amdgpu_device *adev,
 
 	switch (crtc) {
 	case 0:
-		reg_block = SI_CRTC0_REGISTER_OFFSET;
+		reg_block = CRTC0_REGISTER_OFFSET;
 		break;
 	case 1:
-		reg_block = SI_CRTC1_REGISTER_OFFSET;
+		reg_block = CRTC1_REGISTER_OFFSET;
 		break;
 	case 2:
-		reg_block = SI_CRTC2_REGISTER_OFFSET;
+		reg_block = CRTC2_REGISTER_OFFSET;
 		break;
 	case 3:
-		reg_block = SI_CRTC3_REGISTER_OFFSET;
+		reg_block = CRTC3_REGISTER_OFFSET;
 		break;
 	case 4:
-		reg_block = SI_CRTC4_REGISTER_OFFSET;
+		reg_block = CRTC4_REGISTER_OFFSET;
 		break;
 	case 5:
-		reg_block = SI_CRTC5_REGISTER_OFFSET;
+		reg_block = CRTC5_REGISTER_OFFSET;
 		break;
 	default:
 		DRM_DEBUG("invalid crtc %d\n", crtc);
@@ -2873,12 +2979,12 @@ static void dce_v6_0_set_crtc_vblank_interrupt_state(struct amdgpu_device *adev,
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
 		interrupt_mask = RREG32(mmINT_MASK + reg_block);
-		interrupt_mask &= ~VBLANK_INT_MASK;
+		interrupt_mask &= ~INT_MASK__VBLANK_INT_MASK;
 		WREG32(mmINT_MASK + reg_block, interrupt_mask);
 		break;
 	case AMDGPU_IRQ_STATE_ENABLE:
 		interrupt_mask = RREG32(mmINT_MASK + reg_block);
-		interrupt_mask |= VBLANK_INT_MASK;
+		interrupt_mask |= INT_MASK__VBLANK_INT_MASK;
 		WREG32(mmINT_MASK + reg_block, interrupt_mask);
 		break;
 	default:
@@ -2893,28 +2999,28 @@ static void dce_v6_0_set_crtc_vline_interrupt_state(struct amdgpu_device *adev,
 
 }
 
-static int dce_v6_0_set_hpd_interrupt_state(struct amdgpu_device *adev,
+static int dce_v6_0_set_hpd_irq_state(struct amdgpu_device *adev,
 					    struct amdgpu_irq_src *src,
-					    unsigned type,
+					    unsigned hpd,
 					    enum amdgpu_interrupt_state state)
 {
 	u32 dc_hpd_int_cntl;
 
-	if (type >= adev->mode_info.num_hpd) {
-		DRM_DEBUG("invalid hdp %d\n", type);
+	if (hpd >= adev->mode_info.num_hpd) {
+		DRM_DEBUG("invalid hpd %d\n", hpd);
 		return 0;
 	}
 
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
-		dc_hpd_int_cntl = RREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[type]);
-		dc_hpd_int_cntl &= ~DC_HPDx_INT_EN;
-		WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[type], dc_hpd_int_cntl);
+		dc_hpd_int_cntl = RREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd]);
+		dc_hpd_int_cntl &= ~DC_HPD1_INT_CONTROL__DC_HPD1_INT_EN_MASK;
+		WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd], dc_hpd_int_cntl);
 		break;
 	case AMDGPU_IRQ_STATE_ENABLE:
-		dc_hpd_int_cntl = RREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[type]);
-		dc_hpd_int_cntl |= DC_HPDx_INT_EN;
-		WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[type], dc_hpd_int_cntl);
+		dc_hpd_int_cntl = RREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd]);
+		dc_hpd_int_cntl |= DC_HPD1_INT_CONTROL__DC_HPD1_INT_EN_MASK;
+		WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd], dc_hpd_int_cntl);
 		break;
 	default:
 		break;
@@ -2923,7 +3029,7 @@ static int dce_v6_0_set_hpd_interrupt_state(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int dce_v6_0_set_crtc_interrupt_state(struct amdgpu_device *adev,
+static int dce_v6_0_set_crtc_irq_state(struct amdgpu_device *adev,
 					     struct amdgpu_irq_src *src,
 					     unsigned type,
 					     enum amdgpu_interrupt_state state)
@@ -2983,7 +3089,7 @@ static int dce_v6_0_crtc_irq(struct amdgpu_device *adev,
 	switch (entry->src_data[0]) {
 	case 0: /* vblank */
 		if (disp_int & interrupt_status_offsets[crtc].vblank)
-			WREG32(mmVBLANK_STATUS + crtc_offsets[crtc], VBLANK_ACK);
+			WREG32(mmVBLANK_STATUS + crtc_offsets[crtc], VBLANK_STATUS__VBLANK_ACK_MASK);
 		else
 			DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
 
@@ -2994,7 +3100,7 @@ static int dce_v6_0_crtc_irq(struct amdgpu_device *adev,
 		break;
 	case 1: /* vline */
 		if (disp_int & interrupt_status_offsets[crtc].vline)
-			WREG32(mmVLINE_STATUS + crtc_offsets[crtc], VLINE_ACK);
+			WREG32(mmVLINE_STATUS + crtc_offsets[crtc], VLINE_STATUS__VLINE_ACK_MASK);
 		else
 			DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
 
@@ -3008,7 +3114,7 @@ static int dce_v6_0_crtc_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int dce_v6_0_set_pageflip_interrupt_state(struct amdgpu_device *adev,
+static int dce_v6_0_set_pageflip_irq_state(struct amdgpu_device *adev,
 						 struct amdgpu_irq_src *src,
 						 unsigned type,
 						 enum amdgpu_interrupt_state state)
@@ -3059,7 +3165,7 @@ static int dce_v6_0_pageflip_irq(struct amdgpu_device *adev,
 
 	spin_lock_irqsave(&adev_to_drm(adev)->event_lock, flags);
 	works = amdgpu_crtc->pflip_works;
-	if (amdgpu_crtc->pflip_status != AMDGPU_FLIP_SUBMITTED){
+	if (amdgpu_crtc->pflip_status != AMDGPU_FLIP_SUBMITTED) {
 		DRM_DEBUG_DRIVER("amdgpu_crtc->pflip_status = %d != "
 						"AMDGPU_FLIP_SUBMITTED(%d)\n",
 						amdgpu_crtc->pflip_status,
@@ -3088,7 +3194,7 @@ static int dce_v6_0_hpd_irq(struct amdgpu_device *adev,
 			    struct amdgpu_irq_src *source,
 			    struct amdgpu_iv_entry *entry)
 {
-	uint32_t disp_int, mask, tmp;
+	uint32_t disp_int, mask;
 	unsigned hpd;
 
 	if (entry->src_data[0] >= adev->mode_info.num_hpd) {
@@ -3101,24 +3207,21 @@ static int dce_v6_0_hpd_irq(struct amdgpu_device *adev,
 	mask = interrupt_status_offsets[hpd].hpd;
 
 	if (disp_int & mask) {
-		tmp = RREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd]);
-		tmp |= DC_HPD1_INT_CONTROL__DC_HPD1_INT_ACK_MASK;
-		WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd], tmp);
+		dce_v6_0_hpd_int_ack(adev, hpd);
 		schedule_delayed_work(&adev->hotplug_work, 0);
 		DRM_DEBUG("IH: HPD%d\n", hpd + 1);
 	}
 
 	return 0;
-
 }
 
-static int dce_v6_0_set_clockgating_state(void *handle,
+static int dce_v6_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int dce_v6_0_set_powergating_state(void *handle,
+static int dce_v6_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -3127,7 +3230,6 @@ static int dce_v6_0_set_powergating_state(void *handle,
 static const struct amd_ip_funcs dce_v6_0_ip_funcs = {
 	.name = "dce_v6_0",
 	.early_init = dce_v6_0_early_init,
-	.late_init = NULL,
 	.sw_init = dce_v6_0_sw_init,
 	.sw_fini = dce_v6_0_sw_fini,
 	.hw_init = dce_v6_0_hw_init,
@@ -3135,18 +3237,15 @@ static const struct amd_ip_funcs dce_v6_0_ip_funcs = {
 	.suspend = dce_v6_0_suspend,
 	.resume = dce_v6_0_resume,
 	.is_idle = dce_v6_0_is_idle,
-	.wait_for_idle = dce_v6_0_wait_for_idle,
 	.soft_reset = dce_v6_0_soft_reset,
 	.set_clockgating_state = dce_v6_0_set_clockgating_state,
 	.set_powergating_state = dce_v6_0_set_powergating_state,
 };
 
-static void
-dce_v6_0_encoder_mode_set(struct drm_encoder *encoder,
+static void dce_v6_0_encoder_mode_set(struct drm_encoder *encoder,
 			  struct drm_display_mode *mode,
 			  struct drm_display_mode *adjusted_mode)
 {
-
 	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
 	int em = amdgpu_atombios_encoder_get_encoder_mode(encoder);
 
@@ -3166,7 +3265,6 @@ dce_v6_0_encoder_mode_set(struct drm_encoder *encoder,
 
 static void dce_v6_0_encoder_prepare(struct drm_encoder *encoder)
 {
-
 	struct amdgpu_device *adev = drm_to_adev(encoder->dev);
 	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
 	struct drm_connector *connector = amdgpu_get_connector_for_encoder(encoder);
@@ -3206,7 +3304,6 @@ static void dce_v6_0_encoder_prepare(struct drm_encoder *encoder)
 
 static void dce_v6_0_encoder_commit(struct drm_encoder *encoder)
 {
-
 	struct drm_device *dev = encoder->dev;
 	struct amdgpu_device *adev = drm_to_adev(dev);
 
@@ -3217,7 +3314,6 @@ static void dce_v6_0_encoder_commit(struct drm_encoder *encoder)
 
 static void dce_v6_0_encoder_disable(struct drm_encoder *encoder)
 {
-
 	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
 	struct amdgpu_encoder_atom_dig *dig;
 	int em = amdgpu_atombios_encoder_get_encoder_mode(encoder);
@@ -3244,8 +3340,7 @@ static void dce_v6_0_ext_commit(struct drm_encoder *encoder)
 
 }
 
-static void
-dce_v6_0_ext_mode_set(struct drm_encoder *encoder,
+static void dce_v6_0_ext_mode_set(struct drm_encoder *encoder,
 		      struct drm_display_mode *mode,
 		      struct drm_display_mode *adjusted_mode)
 {
@@ -3257,8 +3352,7 @@ static void dce_v6_0_ext_disable(struct drm_encoder *encoder)
 
 }
 
-static void
-dce_v6_0_ext_dpms(struct drm_encoder *encoder, int mode)
+static void dce_v6_0_ext_dpms(struct drm_encoder *encoder, int mode)
 {
 
 }
@@ -3329,7 +3423,6 @@ static void dce_v6_0_encoder_add(struct amdgpu_device *adev,
 			amdgpu_encoder->devices |= supported_device;
 			return;
 		}
-
 	}
 
 	/* add a new one */
@@ -3436,17 +3529,17 @@ static void dce_v6_0_set_display_funcs(struct amdgpu_device *adev)
 }
 
 static const struct amdgpu_irq_src_funcs dce_v6_0_crtc_irq_funcs = {
-	.set = dce_v6_0_set_crtc_interrupt_state,
+	.set = dce_v6_0_set_crtc_irq_state,
 	.process = dce_v6_0_crtc_irq,
 };
 
 static const struct amdgpu_irq_src_funcs dce_v6_0_pageflip_irq_funcs = {
-	.set = dce_v6_0_set_pageflip_interrupt_state,
+	.set = dce_v6_0_set_pageflip_irq_state,
 	.process = dce_v6_0_pageflip_irq,
 };
 
 static const struct amdgpu_irq_src_funcs dce_v6_0_hpd_irq_funcs = {
-	.set = dce_v6_0_set_hpd_interrupt_state,
+	.set = dce_v6_0_set_hpd_irq_state,
 	.process = dce_v6_0_hpd_irq,
 };
 
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c
index d421a268c9ff..2ccd6aad8dd6 100644
--- a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c
@@ -21,6 +21,7 @@
  *
  */
 
+#include <drm/drm_edid.h>
 #include <drm/drm_fourcc.h>
 #include <drm/drm_modeset_helper.h>
 #include <drm/drm_modeset_helper_vtables.h>
@@ -53,8 +54,7 @@
 static void dce_v8_0_set_display_funcs(struct amdgpu_device *adev);
 static void dce_v8_0_set_irq_funcs(struct amdgpu_device *adev);
 
-static const u32 crtc_offsets[6] =
-{
+static const u32 crtc_offsets[6] = {
 	CRTC0_REGISTER_OFFSET,
 	CRTC1_REGISTER_OFFSET,
 	CRTC2_REGISTER_OFFSET,
@@ -63,8 +63,7 @@ static const u32 crtc_offsets[6] =
 	CRTC5_REGISTER_OFFSET
 };
 
-static const u32 hpd_offsets[] =
-{
+static const u32 hpd_offsets[] = {
 	HPD0_REGISTER_OFFSET,
 	HPD1_REGISTER_OFFSET,
 	HPD2_REGISTER_OFFSET,
@@ -266,6 +265,21 @@ static void dce_v8_0_hpd_set_polarity(struct amdgpu_device *adev,
 	WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd], tmp);
 }
 
+static void dce_v8_0_hpd_int_ack(struct amdgpu_device *adev,
+				 int hpd)
+{
+	u32 tmp;
+
+	if (hpd >= adev->mode_info.num_hpd) {
+		DRM_DEBUG("invalid hpd %d\n", hpd);
+		return;
+	}
+
+	tmp = RREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd]);
+	tmp |= DC_HPD1_INT_CONTROL__DC_HPD1_INT_ACK_MASK;
+	WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd], tmp);
+}
+
 /**
  * dce_v8_0_hpd_init - hpd setup callback.
  *
@@ -305,6 +319,7 @@ static void dce_v8_0_hpd_init(struct amdgpu_device *adev)
 			continue;
 		}
 
+		dce_v8_0_hpd_int_ack(adev, amdgpu_connector->hpd.hpd);
 		dce_v8_0_hpd_set_polarity(adev, amdgpu_connector->hpd.hpd);
 		amdgpu_irq_get(adev, &adev->hpd_irq, amdgpu_connector->hpd.hpd);
 	}
@@ -977,7 +992,7 @@ static void dce_v8_0_program_watermarks(struct amdgpu_device *adev,
 					    (u32)mode->clock);
 		line_time = (u32) div_u64((u64)mode->crtc_htotal * 1000000,
 					  (u32)mode->clock);
-		line_time = min(line_time, (u32)65535);
+		line_time = min_t(u32, line_time, 65535);
 
 		/* watermark for high clocks */
 		if (adev->pm.dpm_enabled) {
@@ -1007,7 +1022,7 @@ static void dce_v8_0_program_watermarks(struct amdgpu_device *adev,
 		wm_high.num_heads = num_heads;
 
 		/* set for high clocks */
-		latency_watermark_a = min(dce_v8_0_latency_watermark(&wm_high), (u32)65535);
+		latency_watermark_a = min_t(u32, dce_v8_0_latency_watermark(&wm_high), 65535);
 
 		/* possibly force display priority to high */
 		/* should really do this at mode validation time... */
@@ -1046,7 +1061,7 @@ static void dce_v8_0_program_watermarks(struct amdgpu_device *adev,
 		wm_low.num_heads = num_heads;
 
 		/* set for low clocks */
-		latency_watermark_b = min(dce_v8_0_latency_watermark(&wm_low), (u32)65535);
+		latency_watermark_b = min_t(u32, dce_v8_0_latency_watermark(&wm_low), 65535);
 
 		/* possibly force display priority to high */
 		/* should really do this at mode validation time... */
@@ -1081,8 +1096,7 @@ static void dce_v8_0_program_watermarks(struct amdgpu_device *adev,
 
 	/* save values for DPM */
 	amdgpu_crtc->line_time = line_time;
-	amdgpu_crtc->wm_high = latency_watermark_a;
-	amdgpu_crtc->wm_low = latency_watermark_b;
+
 	/* Save number of lines the linebuffer leads before the scanout */
 	amdgpu_crtc->lb_vblank_lead_lines = lb_vblank_lead_lines;
 }
@@ -1257,7 +1271,7 @@ static void dce_v8_0_audio_write_speaker_allocation(struct drm_encoder *encoder)
 		return;
 	}
 
-	sad_count = drm_edid_to_speaker_allocation(amdgpu_connector_edid(connector), &sadb);
+	sad_count = drm_edid_to_speaker_allocation(amdgpu_connector->edid, &sadb);
 	if (sad_count < 0) {
 		DRM_ERROR("Couldn't read Speaker Allocation Data Block: %d\n", sad_count);
 		sad_count = 0;
@@ -1325,7 +1339,7 @@ static void dce_v8_0_audio_write_sad_regs(struct drm_encoder *encoder)
 		return;
 	}
 
-	sad_count = drm_edid_to_sad(amdgpu_connector_edid(connector), &sads);
+	sad_count = drm_edid_to_sad(amdgpu_connector->edid, &sads);
 	if (sad_count < 0)
 		DRM_ERROR("Couldn't read SADs: %d\n", sad_count);
 	if (sad_count <= 0)
@@ -1345,9 +1359,9 @@ static void dce_v8_0_audio_write_sad_regs(struct drm_encoder *encoder)
 				if (sad->channels > max_channels) {
 					value = (sad->channels <<
 						 AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__MAX_CHANNELS__SHIFT) |
-					        (sad->byte2 <<
+						(sad->byte2 <<
 						 AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__DESCRIPTOR_BYTE_2__SHIFT) |
-					        (sad->freq <<
+						(sad->freq <<
 						 AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__SUPPORTED_FREQUENCIES__SHIFT);
 					max_channels = sad->channels;
 				}
@@ -1379,15 +1393,14 @@ static void dce_v8_0_audio_enable(struct amdgpu_device *adev,
 		enable ? AZALIA_F0_CODEC_PIN_CONTROL_HOT_PLUG_CONTROL__AUDIO_ENABLED_MASK : 0);
 }
 
-static const u32 pin_offsets[7] =
-{
-	(0x1780 - 0x1780),
-	(0x1786 - 0x1780),
-	(0x178c - 0x1780),
-	(0x1792 - 0x1780),
-	(0x1798 - 0x1780),
-	(0x179d - 0x1780),
-	(0x17a4 - 0x1780),
+static const u32 pin_offsets[7] = {
+	AUD0_REGISTER_OFFSET,
+	AUD1_REGISTER_OFFSET,
+	AUD2_REGISTER_OFFSET,
+	AUD3_REGISTER_OFFSET,
+	AUD4_REGISTER_OFFSET,
+	AUD5_REGISTER_OFFSET,
+	AUD6_REGISTER_OFFSET,
 };
 
 static int dce_v8_0_audio_init(struct amdgpu_device *adev)
@@ -1429,17 +1442,12 @@ static int dce_v8_0_audio_init(struct amdgpu_device *adev)
 
 static void dce_v8_0_audio_fini(struct amdgpu_device *adev)
 {
-	int i;
-
 	if (!amdgpu_audio)
 		return;
 
 	if (!adev->mode_info.audio.enabled)
 		return;
 
-	for (i = 0; i < adev->mode_info.audio.num_pins; i++)
-		dce_v8_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
-
 	adev->mode_info.audio.enabled = false;
 }
 
@@ -1740,8 +1748,7 @@ static void dce_v8_0_afmt_fini(struct amdgpu_device *adev)
 	}
 }
 
-static const u32 vga_control_regs[6] =
-{
+static const u32 vga_control_regs[6] = {
 	mmD1VGA_CONTROL,
 	mmD2VGA_CONTROL,
 	mmD3VGA_CONTROL,
@@ -1815,6 +1822,7 @@ static int dce_v8_0_crtc_do_set_base(struct drm_crtc *crtc,
 		return r;
 
 	if (!atomic) {
+		abo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 		r = amdgpu_bo_pin(abo, AMDGPU_GEM_DOMAIN_VRAM);
 		if (unlikely(r != 0)) {
 			amdgpu_bo_unreserve(abo);
@@ -1895,9 +1903,9 @@ static int dce_v8_0_crtc_do_set_base(struct drm_crtc *crtc,
 	case DRM_FORMAT_XBGR8888:
 	case DRM_FORMAT_ABGR8888:
 		fb_format = ((GRPH_DEPTH_32BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
-		             (GRPH_FORMAT_ARGB8888 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
+				(GRPH_FORMAT_ARGB8888 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
 		fb_swap = ((GRPH_RED_SEL_B << GRPH_SWAP_CNTL__GRPH_RED_CROSSBAR__SHIFT) |
-		           (GRPH_BLUE_SEL_R << GRPH_SWAP_CNTL__GRPH_BLUE_CROSSBAR__SHIFT));
+			(GRPH_BLUE_SEL_R << GRPH_SWAP_CNTL__GRPH_BLUE_CROSSBAR__SHIFT));
 #ifdef __BIG_ENDIAN
 		fb_swap |= (GRPH_ENDIAN_8IN32 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 #endif
@@ -2307,6 +2315,7 @@ static int dce_v8_0_crtc_cursor_set2(struct drm_crtc *crtc,
 		return ret;
 	}
 
+	aobj->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 	ret = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM);
 	amdgpu_bo_unreserve(aobj);
 	if (ret) {
@@ -2598,6 +2607,31 @@ static const struct drm_crtc_helper_funcs dce_v8_0_crtc_helper_funcs = {
 	.get_scanout_position = amdgpu_crtc_get_scanout_position,
 };
 
+static void dce_v8_0_panic_flush(struct drm_plane *plane)
+{
+	struct drm_framebuffer *fb;
+	struct amdgpu_crtc *amdgpu_crtc;
+	struct amdgpu_device *adev;
+	uint32_t fb_format;
+
+	if (!plane->fb)
+		return;
+
+	fb = plane->fb;
+	amdgpu_crtc = to_amdgpu_crtc(plane->crtc);
+	adev = drm_to_adev(fb->dev);
+
+	/* Disable DC tiling */
+	fb_format = RREG32(mmGRPH_CONTROL + amdgpu_crtc->crtc_offset);
+	fb_format &= ~GRPH_CONTROL__GRPH_ARRAY_MODE_MASK;
+	WREG32(mmGRPH_CONTROL + amdgpu_crtc->crtc_offset, fb_format);
+}
+
+static const struct drm_plane_helper_funcs dce_v8_0_drm_primary_plane_helper_funcs = {
+	.get_scanout_buffer = amdgpu_display_get_scanout_buffer,
+	.panic_flush = dce_v8_0_panic_flush,
+};
+
 static int dce_v8_0_crtc_init(struct amdgpu_device *adev, int index)
 {
 	struct amdgpu_crtc *amdgpu_crtc;
@@ -2625,13 +2659,14 @@ static int dce_v8_0_crtc_init(struct amdgpu_device *adev, int index)
 	amdgpu_crtc->encoder = NULL;
 	amdgpu_crtc->connector = NULL;
 	drm_crtc_helper_add(&amdgpu_crtc->base, &dce_v8_0_crtc_helper_funcs);
+	drm_plane_helper_add(amdgpu_crtc->base.primary, &dce_v8_0_drm_primary_plane_helper_funcs);
 
 	return 0;
 }
 
-static int dce_v8_0_early_init(void *handle)
+static int dce_v8_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->audio_endpt_rreg = &dce_v8_0_audio_endpt_rreg;
 	adev->audio_endpt_wreg = &dce_v8_0_audio_endpt_wreg;
@@ -2665,10 +2700,10 @@ static int dce_v8_0_early_init(void *handle)
 	return 0;
 }
 
-static int dce_v8_0_sw_init(void *handle)
+static int dce_v8_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->mode_info.num_crtc; i++) {
 		r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i + 1, &adev->crtc_irq);
@@ -2749,11 +2784,11 @@ static int dce_v8_0_sw_init(void *handle)
 	return 0;
 }
 
-static int dce_v8_0_sw_fini(void *handle)
+static int dce_v8_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	kfree(adev->mode_info.bios_hardcoded_edid);
+	drm_edid_free(adev->mode_info.bios_hardcoded_edid);
 
 	drm_kms_helper_poll_fini(adev_to_drm(adev));
 
@@ -2767,10 +2802,10 @@ static int dce_v8_0_sw_fini(void *handle)
 	return 0;
 }
 
-static int dce_v8_0_hw_init(void *handle)
+static int dce_v8_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* disable vga render */
 	dce_v8_0_set_vga_render_state(adev, false);
@@ -2790,10 +2825,10 @@ static int dce_v8_0_hw_init(void *handle)
 	return 0;
 }
 
-static int dce_v8_0_hw_fini(void *handle)
+static int dce_v8_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	dce_v8_0_hpd_fini(adev);
 
@@ -2808,9 +2843,9 @@ static int dce_v8_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int dce_v8_0_suspend(void *handle)
+static int dce_v8_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_display_suspend_helper(adev);
@@ -2820,18 +2855,18 @@ static int dce_v8_0_suspend(void *handle)
 	adev->mode_info.bl_level =
 		amdgpu_atombios_encoder_get_backlight_level_from_reg(adev);
 
-	return dce_v8_0_hw_fini(handle);
+	return dce_v8_0_hw_fini(ip_block);
 }
 
-static int dce_v8_0_resume(void *handle)
+static int dce_v8_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	amdgpu_atombios_encoder_set_backlight_level_to_reg(adev,
 							   adev->mode_info.bl_level);
 
-	ret = dce_v8_0_hw_init(handle);
+	ret = dce_v8_0_hw_init(ip_block);
 
 	/* turn on the BL */
 	if (adev->mode_info.bl_encoder) {
@@ -2846,20 +2881,15 @@ static int dce_v8_0_resume(void *handle)
 	return amdgpu_display_resume_helper(adev);
 }
 
-static bool dce_v8_0_is_idle(void *handle)
+static bool dce_v8_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	return true;
 }
 
-static int dce_v8_0_wait_for_idle(void *handle)
-{
-	return 0;
-}
-
-static int dce_v8_0_soft_reset(void *handle)
+static int dce_v8_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 srbm_soft_reset = 0, tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (dce_v8_0_is_display_hung(adev))
 		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_DC_MASK;
@@ -2985,7 +3015,7 @@ static void dce_v8_0_set_crtc_vline_interrupt_state(struct amdgpu_device *adev,
 	}
 }
 
-static int dce_v8_0_set_hpd_interrupt_state(struct amdgpu_device *adev,
+static int dce_v8_0_set_hpd_irq_state(struct amdgpu_device *adev,
 					    struct amdgpu_irq_src *src,
 					    unsigned type,
 					    enum amdgpu_interrupt_state state)
@@ -2993,7 +3023,7 @@ static int dce_v8_0_set_hpd_interrupt_state(struct amdgpu_device *adev,
 	u32 dc_hpd_int_cntl;
 
 	if (type >= adev->mode_info.num_hpd) {
-		DRM_DEBUG("invalid hdp %d\n", type);
+		DRM_DEBUG("invalid hpd %d\n", type);
 		return 0;
 	}
 
@@ -3015,7 +3045,7 @@ static int dce_v8_0_set_hpd_interrupt_state(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int dce_v8_0_set_crtc_interrupt_state(struct amdgpu_device *adev,
+static int dce_v8_0_set_crtc_irq_state(struct amdgpu_device *adev,
 					     struct amdgpu_irq_src *src,
 					     unsigned type,
 					     enum amdgpu_interrupt_state state)
@@ -3100,7 +3130,7 @@ static int dce_v8_0_crtc_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int dce_v8_0_set_pageflip_interrupt_state(struct amdgpu_device *adev,
+static int dce_v8_0_set_pageflip_irq_state(struct amdgpu_device *adev,
 						 struct amdgpu_irq_src *src,
 						 unsigned type,
 						 enum amdgpu_interrupt_state state)
@@ -3151,7 +3181,7 @@ static int dce_v8_0_pageflip_irq(struct amdgpu_device *adev,
 
 	spin_lock_irqsave(&adev_to_drm(adev)->event_lock, flags);
 	works = amdgpu_crtc->pflip_works;
-	if (amdgpu_crtc->pflip_status != AMDGPU_FLIP_SUBMITTED){
+	if (amdgpu_crtc->pflip_status != AMDGPU_FLIP_SUBMITTED) {
 		DRM_DEBUG_DRIVER("amdgpu_crtc->pflip_status = %d != "
 						"AMDGPU_FLIP_SUBMITTED(%d)\n",
 						amdgpu_crtc->pflip_status,
@@ -3180,7 +3210,7 @@ static int dce_v8_0_hpd_irq(struct amdgpu_device *adev,
 			    struct amdgpu_irq_src *source,
 			    struct amdgpu_iv_entry *entry)
 {
-	uint32_t disp_int, mask, tmp;
+	uint32_t disp_int, mask;
 	unsigned hpd;
 
 	if (entry->src_data[0] >= adev->mode_info.num_hpd) {
@@ -3193,9 +3223,7 @@ static int dce_v8_0_hpd_irq(struct amdgpu_device *adev,
 	mask = interrupt_status_offsets[hpd].hpd;
 
 	if (disp_int & mask) {
-		tmp = RREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd]);
-		tmp |= DC_HPD1_INT_CONTROL__DC_HPD1_INT_ACK_MASK;
-		WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd], tmp);
+		dce_v8_0_hpd_int_ack(adev, hpd);
 		schedule_delayed_work(&adev->hotplug_work, 0);
 		DRM_DEBUG("IH: HPD%d\n", hpd + 1);
 	}
@@ -3204,13 +3232,13 @@ static int dce_v8_0_hpd_irq(struct amdgpu_device *adev,
 
 }
 
-static int dce_v8_0_set_clockgating_state(void *handle,
+static int dce_v8_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int dce_v8_0_set_powergating_state(void *handle,
+static int dce_v8_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -3219,7 +3247,6 @@ static int dce_v8_0_set_powergating_state(void *handle,
 static const struct amd_ip_funcs dce_v8_0_ip_funcs = {
 	.name = "dce_v8_0",
 	.early_init = dce_v8_0_early_init,
-	.late_init = NULL,
 	.sw_init = dce_v8_0_sw_init,
 	.sw_fini = dce_v8_0_sw_fini,
 	.hw_init = dce_v8_0_hw_init,
@@ -3227,7 +3254,6 @@ static const struct amd_ip_funcs dce_v8_0_ip_funcs = {
 	.suspend = dce_v8_0_suspend,
 	.resume = dce_v8_0_resume,
 	.is_idle = dce_v8_0_is_idle,
-	.wait_for_idle = dce_v8_0_wait_for_idle,
 	.soft_reset = dce_v8_0_soft_reset,
 	.set_clockgating_state = dce_v8_0_set_clockgating_state,
 	.set_powergating_state = dce_v8_0_set_powergating_state,
@@ -3515,17 +3541,17 @@ static void dce_v8_0_set_display_funcs(struct amdgpu_device *adev)
 }
 
 static const struct amdgpu_irq_src_funcs dce_v8_0_crtc_irq_funcs = {
-	.set = dce_v8_0_set_crtc_interrupt_state,
+	.set = dce_v8_0_set_crtc_irq_state,
 	.process = dce_v8_0_crtc_irq,
 };
 
 static const struct amdgpu_irq_src_funcs dce_v8_0_pageflip_irq_funcs = {
-	.set = dce_v8_0_set_pageflip_interrupt_state,
+	.set = dce_v8_0_set_pageflip_irq_state,
 	.process = dce_v8_0_pageflip_irq,
 };
 
 static const struct amdgpu_irq_src_funcs dce_v8_0_hpd_irq_funcs = {
-	.set = dce_v8_0_set_hpd_interrupt_state,
+	.set = dce_v8_0_set_hpd_irq_state,
 	.process = dce_v8_0_hpd_irq,
 };
 
@@ -3544,8 +3570,7 @@ static void dce_v8_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->hpd_irq.funcs = &dce_v8_0_hpd_irq_funcs;
 }
 
-const struct amdgpu_ip_block_version dce_v8_0_ip_block =
-{
+const struct amdgpu_ip_block_version dce_v8_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_DCE,
 	.major = 8,
 	.minor = 0,
@@ -3553,8 +3578,7 @@ const struct amdgpu_ip_block_version dce_v8_0_ip_block =
 	.funcs = &dce_v8_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version dce_v8_1_ip_block =
-{
+const struct amdgpu_ip_block_version dce_v8_1_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_DCE,
 	.major = 8,
 	.minor = 1,
@@ -3562,8 +3586,7 @@ const struct amdgpu_ip_block_version dce_v8_1_ip_block =
 	.funcs = &dce_v8_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version dce_v8_2_ip_block =
-{
+const struct amdgpu_ip_block_version dce_v8_2_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_DCE,
 	.major = 8,
 	.minor = 2,
@@ -3571,8 +3594,7 @@ const struct amdgpu_ip_block_version dce_v8_2_ip_block =
 	.funcs = &dce_v8_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version dce_v8_3_ip_block =
-{
+const struct amdgpu_ip_block_version dce_v8_3_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_DCE,
 	.major = 8,
 	.minor = 3,
@@ -3580,8 +3602,7 @@ const struct amdgpu_ip_block_version dce_v8_3_ip_block =
 	.funcs = &dce_v8_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version dce_v8_5_ip_block =
-{
+const struct amdgpu_ip_block_version dce_v8_5_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_DCE,
 	.major = 8,
 	.minor = 5,
diff --git a/drivers/gpu/drm/amd/amdgpu/df_v1_7.c b/drivers/gpu/drm/amd/amdgpu/df_v1_7.c
index 5dfab80ffff2..cd298556f7a6 100644
--- a/drivers/gpu/drm/amd/amdgpu/df_v1_7.c
+++ b/drivers/gpu/drm/amd/amdgpu/df_v1_7.c
@@ -70,6 +70,8 @@ static u32 df_v1_7_get_hbm_channel_number(struct amdgpu_device *adev)
 	int fb_channel_number;
 
 	fb_channel_number = adev->df.funcs->get_fb_channel_number(adev);
+	if (fb_channel_number >= ARRAY_SIZE(df_v1_7_channel_number))
+		fb_channel_number = 0;
 
 	return df_v1_7_channel_number[fb_channel_number];
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/df_v3_6.c b/drivers/gpu/drm/amd/amdgpu/df_v3_6.c
index 483a441b46aa..621aeca53880 100644
--- a/drivers/gpu/drm/amd/amdgpu/df_v3_6.c
+++ b/drivers/gpu/drm/amd/amdgpu/df_v3_6.c
@@ -254,8 +254,8 @@ static void df_v3_6_sw_init(struct amdgpu_device *adev)
 
 static void df_v3_6_sw_fini(struct amdgpu_device *adev)
 {
-
-	device_remove_file(adev->dev, &dev_attr_df_cntr_avail);
+	if (adev->dev->kobj.sd)
+		device_remove_file(adev->dev, &dev_attr_df_cntr_avail);
 
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/df_v4_15.c b/drivers/gpu/drm/amd/amdgpu/df_v4_15.c
new file mode 100644
index 000000000000..2a573e33908b
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/df_v4_15.c
@@ -0,0 +1,45 @@
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "amdgpu.h"
+#include "df_v4_15.h"
+
+#include "df/df_4_15_offset.h"
+#include "df/df_4_15_sh_mask.h"
+
+static void df_v4_15_hw_init(struct amdgpu_device *adev)
+{
+	if (adev->have_atomics_support) {
+		uint32_t tmp;
+		uint32_t dis_lcl_proc = (1 <<  1 |
+					1 <<  2 |
+					1 << 13);
+
+		tmp = RREG32_SOC15(DF, 0, regNCSConfigurationRegister1);
+		tmp |= (dis_lcl_proc << NCSConfigurationRegister1__DisIntAtomicsLclProcessing__SHIFT);
+		WREG32_SOC15(DF, 0, regNCSConfigurationRegister1, tmp);
+	}
+}
+
+const struct amdgpu_df_funcs df_v4_15_funcs = {
+	.hw_init = df_v4_15_hw_init
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/df_v4_15.h b/drivers/gpu/drm/amd/amdgpu/df_v4_15.h
new file mode 100644
index 000000000000..dddf2422112a
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/df_v4_15.h
@@ -0,0 +1,30 @@
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __DF_V4_15_H__
+#define __DF_V4_15_H__
+
+extern const struct amdgpu_df_funcs df_v4_15_funcs;
+
+#endif /* __DF_V4_15_H__ */
+
diff --git a/drivers/gpu/drm/amd/amdgpu/df_v4_6_2.c b/drivers/gpu/drm/amd/amdgpu/df_v4_6_2.c
new file mode 100644
index 000000000000..a47960a0babd
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/df_v4_6_2.c
@@ -0,0 +1,34 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "amdgpu.h"
+#include "df_v4_6_2.h"
+
+static bool df_v4_6_2_query_ras_poison_mode(struct amdgpu_device *adev)
+{
+	/* return true since related regs are inaccessible */
+	return true;
+}
+
+const struct amdgpu_df_funcs df_v4_6_2_funcs = {
+	.query_ras_poison_mode = df_v4_6_2_query_ras_poison_mode,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.h b/drivers/gpu/drm/amd/amdgpu/df_v4_6_2.h
index 0d878ca3acba..3bc3e6d216e2 100644
--- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.h
+++ b/drivers/gpu/drm/amd/amdgpu/df_v4_6_2.h
@@ -1,5 +1,5 @@
 /*
- * Copyright 2014 Advanced Micro Devices, Inc.
+ * Copyright 2023 Advanced Micro Devices, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -21,12 +21,11 @@
  *
  */
 
-#ifndef __DCE_V11_0_H__
-#define __DCE_V11_0_H__
+#ifndef __DF_V4_6_2_H__
+#define __DF_V4_6_2_H__
 
-extern const struct amdgpu_ip_block_version dce_v11_0_ip_block;
-extern const struct amdgpu_ip_block_version dce_v11_2_ip_block;
+#include "soc15_common.h"
 
-void dce_v11_0_disable_dce(struct amdgpu_device *adev);
+extern const struct amdgpu_df_funcs df_v4_6_2_funcs;
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index f5b5ce1051a2..d75b9940f248 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
@@ -40,11 +40,11 @@
 #include "ivsrcid/gfx/irqsrcs_gfx_10_1.h"
 
 #include "soc15.h"
-#include "soc15d.h"
 #include "soc15_common.h"
 #include "clearstate_gfx10.h"
 #include "v10_structs.h"
 #include "gfx_v10_0.h"
+#include "gfx_v10_0_cleaner_shader.h"
 #include "nbio_v2_3.h"
 
 /*
@@ -102,6 +102,11 @@
 #define mmGCR_GENERAL_CNTL_Sienna_Cichlid			0x1580
 #define mmGCR_GENERAL_CNTL_Sienna_Cichlid_BASE_IDX	0
 
+#define mmGOLDEN_TSC_COUNT_UPPER_Cyan_Skillfish                0x0105
+#define mmGOLDEN_TSC_COUNT_UPPER_Cyan_Skillfish_BASE_IDX       1
+#define mmGOLDEN_TSC_COUNT_LOWER_Cyan_Skillfish                0x0106
+#define mmGOLDEN_TSC_COUNT_LOWER_Cyan_Skillfish_BASE_IDX       1
+
 #define mmGOLDEN_TSC_COUNT_UPPER_Vangogh                0x0025
 #define mmGOLDEN_TSC_COUNT_UPPER_Vangogh_BASE_IDX       1
 #define mmGOLDEN_TSC_COUNT_LOWER_Vangogh                0x0026
@@ -271,8 +276,211 @@ MODULE_FIRMWARE("amdgpu/gc_10_3_7_mec.bin");
 MODULE_FIRMWARE("amdgpu/gc_10_3_7_mec2.bin");
 MODULE_FIRMWARE("amdgpu/gc_10_3_7_rlc.bin");
 
-static const struct soc15_reg_golden golden_settings_gc_10_1[] =
-{
+static const struct amdgpu_hwip_reg_entry gc_reg_list_10_1[] = {
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS3),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_STALLED_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPC_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPF_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPC_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPF_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPC_BUSY_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPF_BUSY_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPF_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HPD_STATUS0),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB0_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB0_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB0_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB1_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB1_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB1_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB2_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB2_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB2_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCPF_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCPC_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCPG_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGDS_PROTECTION_FAULT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGDS_VM_PROTECTION_FAULT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmIA_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmIA_UTCL1_STATUS_2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmPA_CL_CNTL_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRMI_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSQC_DCACHE_UTCL0_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSQC_ICACHE_UTCL0_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSQG_UTCL0_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmTCP_UTCL0_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmWD_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGCVM_L2_PROTECTION_FAULT_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGCVM_L2_PROTECTION_FAULT_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_DEBUG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MES_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC1_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC2_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MES_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPC_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_COMMAND),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_MESSAGE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_ARGUMENT_1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_ARGUMENT_2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_ARGUMENT_3),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_ARGUMENT_4),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSMU_RLC_RESPONSE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SAFE_MODE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_SAFE_MODE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_RLCS_GPM_STAT_2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SPP_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_RLCS_BOOTLOAD_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_INT_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_GPM_GENERAL_6),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_GPM_DEBUG_INST_A),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_GPM_DEBUG_INST_B),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_GPM_DEBUG_INST_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_LX6_CORE_PDEBUG_INST),
+	/* cp header registers */
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MES_HEADER_DUMP),
+	/* SE status registers */
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE0),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE3)
+};
+
+static const struct amdgpu_hwip_reg_entry gc_cp_reg_list_10[] = {
+	/* compute registers */
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_VMID),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PERSISTENT_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PIPE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_QUEUE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_QUANTUM),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_DEQUEUE_REQUEST),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_EVENTS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_BASE_ADDR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CNTL_STACK_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CNTL_STACK_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_WG_STATE_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_GDS_RESOURCE_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_WPTR_MEM),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_SUSPEND_CNTL_STACK_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_SUSPEND_CNTL_STACK_DW_CNT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_SUSPEND_WG_STATE_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_DEQUEUE_STATUS),
+	/* cp header registers */
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+};
+
+static const struct amdgpu_hwip_reg_entry gc_gfx_queue_reg_list_10[] = {
+	/* gfx queue registers */
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_ACTIVE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_QUEUE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_CSMD_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_DEQUEUE_REQUEST),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_MAPPED),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_QUE_MGR_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_HQ_CONTROL0),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_HQ_STATUS0),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_CE_WPTR_POLL_ADDR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_CE_WPTR_POLL_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_CE_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_CE_CSMD_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_CE_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_CE_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_MQD_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_MQD_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_WPTR_POLL_ADDR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_WPTR_POLL_ADDR_HI),
+	/* gfx header registers */
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+};
+
+static const struct soc15_reg_golden golden_settings_gc_10_1[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCB_HW_CONTROL_4, 0xffffffff, 0x00400014),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_CPF_CLK_CTRL, 0xfcff8fff, 0xf8000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_CLK_CTRL, 0xcd000000, 0x0d000100),
@@ -315,13 +523,11 @@ static const struct soc15_reg_golden golden_settings_gc_10_1[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0x00c00000, 0x00c00000)
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_0_nv10[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_0_nv10[] = {
 	/* Pending on emulation bring up */
 };
 
-static const struct soc15_reg_golden golden_settings_gc_rlc_spm_10_0_nv10[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_rlc_spm_10_0_nv10[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_GFX_INDEX, 0xe0000000, 0x0),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_GFX_INDEX, 0xffffff, 0x0),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_SPM_GLB_SAMPLEDELAY_IND_ADDR, 0xFFFFFFFF, 0x28),
@@ -1376,8 +1582,7 @@ static const struct soc15_reg_golden golden_settings_gc_rlc_spm_10_0_nv10[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_GFX_INDEX, 0xFFFFFFFF, 0xe0000000)
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_1_1[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_1_1[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCB_HW_CONTROL_4, 0xffffffff, 0x003c0014),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_GS_NGG_CLK_CTRL, 0xffff8fff, 0xffff8100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_IA_CLK_CTRL, 0xffff0fff, 0xffff0100),
@@ -1418,8 +1623,7 @@ static const struct soc15_reg_golden golden_settings_gc_10_1_1[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0x00c00000, 0x00c00000),
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_1_2[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_1_2[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCB_HW_CONTROL_4, 0x003e001f, 0x003c0014),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_GS_NGG_CLK_CTRL, 0xffff8fff, 0xffff8100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_IA_CLK_CTRL, 0xffff0fff, 0xffff0100),
@@ -1464,13 +1668,11 @@ static const struct soc15_reg_golden golden_settings_gc_10_1_2[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0xffffffff, 0x00c00000)
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_1_nv14[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_1_nv14[] = {
 	/* Pending on emulation bring up */
 };
 
-static const struct soc15_reg_golden golden_settings_gc_rlc_spm_10_1_nv14[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_rlc_spm_10_1_nv14[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_GFX_INDEX, 0xE0000000L, 0x0),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_GFX_INDEX, 0xffffff, 0x0),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_SPM_GLB_SAMPLEDELAY_IND_ADDR, 0xFFFFFFFF, 0x28),
@@ -2093,13 +2295,11 @@ static const struct soc15_reg_golden golden_settings_gc_rlc_spm_10_1_nv14[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_GFX_INDEX, 0xFFFFFFFF, 0xe0000000)
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_1_2_nv12[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_1_2_nv12[] = {
 	/* Pending on emulation bring up */
 };
 
-static const struct soc15_reg_golden golden_settings_gc_rlc_spm_10_1_2_nv12[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_rlc_spm_10_1_2_nv12[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_GFX_INDEX, 0xe0000000L, 0x0),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_GFX_INDEX, 0xffffff, 0x0),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_SPM_GLB_SAMPLEDELAY_IND_ADDR, 0xFFFFFFFF, 0x28),
@@ -3154,8 +3354,7 @@ static const struct soc15_reg_golden golden_settings_gc_rlc_spm_10_1_2_nv12[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_GFX_INDEX, 0xFFFFFFFF, 0xe0000000)
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_3[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_3[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_CS_CLK_CTRL, 0x78000000, 0x78000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_PS_CLK_CTRL, 0xff7f0fff, 0x78000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_RA0_CLK_CTRL, 0xff7f0fff, 0x30000100),
@@ -3164,7 +3363,7 @@ static const struct soc15_reg_golden golden_settings_gc_10_3[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG3, 0xffffffff, 0x00000280),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG4, 0xffffffff, 0x00800000),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_EXCEPTION_CONTROL, 0x7fff0f1f, 0x00b80000),
-	SOC15_REG_GOLDEN_VALUE(GC, 0 ,mmGCEA_SDP_TAG_RESERVE0, 0xffffffff, 0x10100100),
+	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGCEA_SDP_TAG_RESERVE0, 0xffffffff, 0x10100100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGCEA_SDP_TAG_RESERVE1, 0xffffffff, 0x17000088),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGCR_GENERAL_CNTL_Sienna_Cichlid, 0x1ff1ffff, 0x00000500),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGCUTCL2_CGTT_CLK_CTRL_Sienna_Cichlid, 0xff000000, 0xff008080),
@@ -3201,13 +3400,11 @@ static const struct soc15_reg_golden golden_settings_gc_10_3[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0xffbfffff, 0x00a00000)
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_3_sienna_cichlid[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_3_sienna_cichlid[] = {
 	/* Pending on emulation bring up */
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_3_2[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_3_2[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_CS_CLK_CTRL, 0xff7f0fff, 0x78000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_PS_CLK_CTRL, 0xff7f0fff, 0x78000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_RA0_CLK_CTRL, 0xff7f0fff, 0x30000100),
@@ -3254,8 +3451,7 @@ static const struct soc15_reg_golden golden_settings_gc_10_3_2[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmLDS_CONFIG,  0x00000020, 0x00000020),
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_3_vangogh[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_3_vangogh[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_RA0_CLK_CTRL, 0xff7f0fff, 0x30000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_RA1_CLK_CTRL, 0xff7f0fff, 0x7e000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCH_PIPE_STEER, 0x000000ff, 0x000000e4),
@@ -3285,8 +3481,7 @@ static const struct soc15_reg_golden golden_settings_gc_10_3_vangogh[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmLDS_CONFIG,  0x00000020, 0x00000020),
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_3_3[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_3_3[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_CS_CLK_CTRL, 0xff7f0fff, 0x78000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCH_PIPE_STEER, 0x000000ff, 0x000000e4),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCPF_GCR_CNTL, 0x0007ffff, 0x0000c200),
@@ -3309,8 +3504,7 @@ static const struct soc15_reg_golden golden_settings_gc_10_3_3[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0xffffffff, 0x00100000)
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_3_4[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_3_4[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_CS_CLK_CTRL, 0x78000000, 0x78000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_RA0_CLK_CTRL, 0x30000000, 0x30000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_RA1_CLK_CTRL, 0x7e000000, 0x7e000100),
@@ -3380,7 +3574,7 @@ static const struct soc15_reg_golden golden_settings_gc_10_3_5[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER7_SELECT, 0xf0f001ff, 0x00000000),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER8_SELECT, 0xf0f001ff, 0x00000000),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER9_SELECT, 0xf0f001ff, 0x00000000),
-	SOC15_REG_GOLDEN_VALUE(GC, 0, mmTA_CNTL_AUX,0xfff7ffff, 0x01030000),
+	SOC15_REG_GOLDEN_VALUE(GC, 0, mmTA_CNTL_AUX, 0xfff7ffff, 0x01030000),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0xffbfffff, 0x00a00000)
 };
 
@@ -3421,8 +3615,7 @@ static const struct soc15_reg_golden golden_settings_gc_10_0_cyan_skillfish[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0x00800000, 0x00800000)
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_3_6[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_3_6[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_CS_CLK_CTRL, 0xff7f0fff, 0x78000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCH_PIPE_STEER, 0x000000ff, 0x00000044),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCPF_GCR_CNTL, 0x0007ffff, 0x0000c200),
@@ -3490,7 +3683,7 @@ static int gfx_v10_0_get_cu_info(struct amdgpu_device *adev,
 				 struct amdgpu_cu_info *cu_info);
 static uint64_t gfx_v10_0_get_gpu_clock_counter(struct amdgpu_device *adev);
 static void gfx_v10_0_select_se_sh(struct amdgpu_device *adev, u32 se_num,
-				   u32 sh_num, u32 instance);
+				   u32 sh_num, u32 instance, int xcc_id);
 static u32 gfx_v10_0_get_wgp_active_bitmap_per_sh(struct amdgpu_device *adev);
 
 static int gfx_v10_0_rlc_backdoor_autoload_buffer_init(struct amdgpu_device *adev);
@@ -3506,16 +3699,26 @@ static void gfx_v10_3_set_power_brake_sequence(struct amdgpu_device *adev);
 static void gfx_v10_0_ring_invalidate_tlbs(struct amdgpu_ring *ring,
 					   uint16_t pasid, uint32_t flush_type,
 					   bool all_hub, uint8_t dst_sel);
+static void gfx_v10_0_update_spm_vmid_internal(struct amdgpu_device *adev,
+					       unsigned int vmid);
 
+static int gfx_v10_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_powergating_state state);
 static void gfx10_kiq_set_resources(struct amdgpu_ring *kiq_ring, uint64_t queue_mask)
 {
+	struct amdgpu_device *adev = kiq_ring->adev;
+	u64 shader_mc_addr;
+
+	/* Cleaner shader MC address */
+	shader_mc_addr = adev->gfx.cleaner_shader_gpu_addr >> 8;
+
 	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_RESOURCES, 6));
 	amdgpu_ring_write(kiq_ring, PACKET3_SET_RESOURCES_VMID_MASK(0) |
 			  PACKET3_SET_RESOURCES_QUEUE_TYPE(0));	/* vmid_mask:0 queue_type:0 (KIQ) */
 	amdgpu_ring_write(kiq_ring, lower_32_bits(queue_mask));	/* queue mask lo */
 	amdgpu_ring_write(kiq_ring, upper_32_bits(queue_mask));	/* queue mask hi */
-	amdgpu_ring_write(kiq_ring, 0);	/* gws mask lo */
-	amdgpu_ring_write(kiq_ring, 0);	/* gws mask hi */
+	amdgpu_ring_write(kiq_ring, lower_32_bits(shader_mc_addr)); /* cleaner shader addr lo */
+	amdgpu_ring_write(kiq_ring, upper_32_bits(shader_mc_addr)); /* cleaner shader addr hi */
 	amdgpu_ring_write(kiq_ring, 0);	/* oac mask */
 	amdgpu_ring_write(kiq_ring, 0);	/* gds heap base:0, gds heap size:0 */
 }
@@ -3565,14 +3768,8 @@ static void gfx10_kiq_unmap_queues(struct amdgpu_ring *kiq_ring,
 				   enum amdgpu_unmap_queues_action action,
 				   u64 gpu_addr, u64 seq)
 {
-	struct amdgpu_device *adev = kiq_ring->adev;
 	uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0;
 
-	if (adev->enable_mes && !adev->gfx.kiq.ring.sched.ready) {
-		amdgpu_mes_unmap_legacy_queue(adev, ring, action, gpu_addr, seq);
-		return;
-	}
-
 	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4));
 	amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */
 			  PACKET3_UNMAP_QUEUES_ACTION(action) |
@@ -3621,12 +3818,65 @@ static void gfx10_kiq_invalidate_tlbs(struct amdgpu_ring *kiq_ring,
 	gfx_v10_0_ring_invalidate_tlbs(kiq_ring, pasid, flush_type, all_hub, 1);
 }
 
+static void gfx_v10_0_kiq_reset_hw_queue(struct amdgpu_ring *kiq_ring, uint32_t queue_type,
+					uint32_t me_id, uint32_t pipe_id, uint32_t queue_id,
+					uint32_t xcc_id, uint32_t vmid)
+{
+	struct amdgpu_device *adev = kiq_ring->adev;
+	unsigned i;
+	uint32_t tmp;
+
+	/* enter save mode */
+	amdgpu_gfx_rlc_enter_safe_mode(adev, xcc_id);
+	mutex_lock(&adev->srbm_mutex);
+	nv_grbm_select(adev, me_id, pipe_id, queue_id, 0);
+
+	if (queue_type == AMDGPU_RING_TYPE_COMPUTE) {
+		WREG32_SOC15(GC, 0, mmCP_HQD_DEQUEUE_REQUEST, 0x2);
+		WREG32_SOC15(GC, 0, mmSPI_COMPUTE_QUEUE_RESET, 0x1);
+		/* wait till dequeue take effects */
+		for (i = 0; i < adev->usec_timeout; i++) {
+			if (!(RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+		if (i >= adev->usec_timeout)
+			dev_err(adev->dev, "fail to wait on hqd deactive\n");
+	} else if (queue_type == AMDGPU_RING_TYPE_GFX) {
+		WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX,
+			     (uint32_t)(0x1 << GRBM_GFX_INDEX__SE_BROADCAST_WRITES__SHIFT));
+		tmp = REG_SET_FIELD(0, CP_VMID_RESET, RESET_REQUEST, 1 << vmid);
+		if (pipe_id == 0)
+			tmp = REG_SET_FIELD(tmp, CP_VMID_RESET, PIPE0_QUEUES, 1 << queue_id);
+		else
+			tmp = REG_SET_FIELD(tmp, CP_VMID_RESET, PIPE1_QUEUES, 1 << queue_id);
+		WREG32_SOC15(GC, 0, mmCP_VMID_RESET, tmp);
+
+		/* wait till dequeue take effects */
+		for (i = 0; i < adev->usec_timeout; i++) {
+			if (!(RREG32_SOC15(GC, 0, mmCP_GFX_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+		if (i >= adev->usec_timeout)
+			dev_err(adev->dev, "failed to wait on gfx hqd deactivate\n");
+	} else {
+		dev_err(adev->dev, "reset queue_type(%d) not supported\n", queue_type);
+	}
+
+	nv_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	/* exit safe mode */
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
+}
+
 static const struct kiq_pm4_funcs gfx_v10_0_kiq_pm4_funcs = {
 	.kiq_set_resources = gfx10_kiq_set_resources,
 	.kiq_map_queues = gfx10_kiq_map_queues,
 	.kiq_unmap_queues = gfx10_kiq_unmap_queues,
 	.kiq_query_status = gfx10_kiq_query_status,
 	.kiq_invalidate_tlbs = gfx10_kiq_invalidate_tlbs,
+	.kiq_reset_hw_queue = gfx_v10_0_kiq_reset_hw_queue,
 	.set_resources_size = 8,
 	.map_queues_size = 7,
 	.unmap_queues_size = 6,
@@ -3636,12 +3886,12 @@ static const struct kiq_pm4_funcs gfx_v10_0_kiq_pm4_funcs = {
 
 static void gfx_v10_0_set_kiq_pm4_funcs(struct amdgpu_device *adev)
 {
-	adev->gfx.kiq.pmf = &gfx_v10_0_kiq_pm4_funcs;
+	adev->gfx.kiq[0].pmf = &gfx_v10_0_kiq_pm4_funcs;
 }
 
 static void gfx_v10_0_init_spm_golden_registers(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 		soc15_program_register_sequence(adev,
 						golden_settings_gc_rlc_spm_10_0_nv10,
@@ -3664,7 +3914,10 @@ static void gfx_v10_0_init_spm_golden_registers(struct amdgpu_device *adev)
 
 static void gfx_v10_0_init_golden_registers(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 		soc15_program_register_sequence(adev,
 						golden_settings_gc_10_1,
@@ -3714,8 +3967,8 @@ static void gfx_v10_0_init_golden_registers(struct amdgpu_device *adev)
 		break;
 	case IP_VERSION(10, 3, 4):
 		soc15_program_register_sequence(adev,
-                                                golden_settings_gc_10_3_4,
-                                                (const u32)ARRAY_SIZE(golden_settings_gc_10_3_4));
+						golden_settings_gc_10_3_4,
+						(const u32)ARRAY_SIZE(golden_settings_gc_10_3_4));
 		break;
 	case IP_VERSION(10, 3, 5):
 		soc15_program_register_sequence(adev,
@@ -3782,7 +4035,7 @@ static int gfx_v10_0_ring_test_ring(struct amdgpu_ring *ring)
 	struct amdgpu_device *adev = ring->adev;
 	uint32_t scratch = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0);
 	uint32_t tmp = 0;
-	unsigned i;
+	unsigned int i;
 	int r;
 
 	WREG32(scratch, 0xCAFEDEAD);
@@ -3820,40 +4073,25 @@ static int gfx_v10_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	struct amdgpu_device *adev = ring->adev;
 	struct amdgpu_ib ib;
 	struct dma_fence *f = NULL;
-	unsigned index;
+	unsigned int index;
 	uint64_t gpu_addr;
-	volatile uint32_t *cpu_ptr;
+	uint32_t *cpu_ptr;
 	long r;
 
 	memset(&ib, 0, sizeof(ib));
 
-	if (ring->is_mes_queue) {
-		uint32_t padding, offset;
-
-		offset = amdgpu_mes_ctx_get_offs(ring, AMDGPU_MES_CTX_IB_OFFS);
-		padding = amdgpu_mes_ctx_get_offs(ring,
-						  AMDGPU_MES_CTX_PADDING_OFFS);
-
-		ib.gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		ib.ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-
-		gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, padding);
-		cpu_ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, padding);
-		*cpu_ptr = cpu_to_le32(0xCAFEDEAD);
-	} else {
-		r = amdgpu_device_wb_get(adev, &index);
-		if (r)
-			return r;
+	r = amdgpu_device_wb_get(adev, &index);
+	if (r)
+		return r;
 
-		gpu_addr = adev->wb.gpu_addr + (index * 4);
-		adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD);
-		cpu_ptr = &adev->wb.wb[index];
+	gpu_addr = adev->wb.gpu_addr + (index * 4);
+	adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD);
+	cpu_ptr = &adev->wb.wb[index];
 
-		r = amdgpu_ib_get(adev, NULL, 20, AMDGPU_IB_POOL_DIRECT, &ib);
-		if (r) {
-			DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
-			goto err1;
-		}
+	r = amdgpu_ib_get(adev, NULL, 20, AMDGPU_IB_POOL_DIRECT, &ib);
+	if (r) {
+		DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
+		goto err1;
 	}
 
 	ib.ptr[0] = PACKET3(PACKET3_WRITE_DATA, 3);
@@ -3880,12 +4118,10 @@ static int gfx_v10_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	else
 		r = -EINVAL;
 err2:
-	if (!ring->is_mes_queue)
-		amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err1:
-	if (!ring->is_mes_queue)
-		amdgpu_device_wb_free(adev, index);
+	amdgpu_device_wb_free(adev, index);
 	return r;
 }
 
@@ -3905,7 +4141,7 @@ static void gfx_v10_0_check_fw_write_wait(struct amdgpu_device *adev)
 {
 	adev->gfx.cp_fw_write_wait = false;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 	case IP_VERSION(10, 1, 2):
 	case IP_VERSION(10, 1, 1):
@@ -3951,12 +4187,12 @@ static bool gfx_v10_0_navi10_gfxoff_should_enable(struct amdgpu_device *adev)
 		break;
 	}
 
-	return ret ;
+	return ret;
 }
 
 static void gfx_v10_0_check_gfxoff_flag(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 		if (!gfx_v10_0_navi10_gfxoff_should_enable(adev))
 			adev->pm.pp_feature &= ~PP_GFXOFF_MASK;
@@ -3968,7 +4204,7 @@ static void gfx_v10_0_check_gfxoff_flag(struct amdgpu_device *adev)
 
 static int gfx_v10_0_init_microcode(struct amdgpu_device *adev)
 {
-	char fw_name[40];
+	char fw_name[53];
 	char ucode_prefix[30];
 	const char *wks = "";
 	int err;
@@ -3978,41 +4214,41 @@ static int gfx_v10_0_init_microcode(struct amdgpu_device *adev)
 
 	DRM_DEBUG("\n");
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 1) &&
-	   (!(adev->pdev->device == 0x7340 && adev->pdev->revision != 0x00)))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 1, 1) &&
+	    (!(adev->pdev->device == 0x7340 && adev->pdev->revision != 0x00)))
 		wks = "_wks";
 	amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix));
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp%s.bin", ucode_prefix, wks);
-	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_pfp%s.bin", ucode_prefix, wks);
 	if (err)
 		goto out;
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_PFP);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me%s.bin", ucode_prefix, wks);
-	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_me%s.bin", ucode_prefix, wks);
 	if (err)
 		goto out;
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_ME);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce%s.bin", ucode_prefix, wks);
-	err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_ce%s.bin", ucode_prefix, wks);
 	if (err)
 		goto out;
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_CE);
 
 	if (!amdgpu_sriov_vf(adev)) {
 		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", ucode_prefix);
-		err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, fw_name);
-		/* don't check this.  There are apparently firmwares in the wild with
-		 * incorrect size in the header
-		 */
-		if (err == -ENODEV)
-			goto out;
+		err = request_firmware(&adev->gfx.rlc_fw, fw_name, adev->dev);
 		if (err)
-			dev_dbg(adev->dev,
-				"gfx10: amdgpu_ucode_request() failed \"%s\"\n",
-				fw_name);
+			goto out;
+
+		/* don't validate this firmware. There are apparently firmwares
+		 * in the wild with incorrect size in the header
+		 */
 		rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
 		version_major = le16_to_cpu(rlc_hdr->header.header_version_major);
 		version_minor = le16_to_cpu(rlc_hdr->header.header_version_minor);
@@ -4021,15 +4257,17 @@ static int gfx_v10_0_init_microcode(struct amdgpu_device *adev)
 			goto out;
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec%s.bin", ucode_prefix, wks);
-	err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_mec%s.bin", ucode_prefix, wks);
 	if (err)
 		goto out;
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1);
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1_JT);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2%s.bin", ucode_prefix, wks);
-	err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_mec2%s.bin", ucode_prefix, wks);
 	if (!err) {
 		amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2);
 		amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2_JT);
@@ -4037,8 +4275,6 @@ static int gfx_v10_0_init_microcode(struct amdgpu_device *adev)
 		err = 0;
 		adev->gfx.mec2_fw = NULL;
 	}
-	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2);
-	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2_JT);
 
 	gfx_v10_0_check_fw_write_wait(adev);
 out:
@@ -4086,12 +4322,9 @@ static u32 gfx_v10_0_get_csb_size(struct amdgpu_device *adev)
 	return count;
 }
 
-static void gfx_v10_0_get_csb_buffer(struct amdgpu_device *adev,
-				    volatile u32 *buffer)
+static void gfx_v10_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer)
 {
-	u32 count = 0, i;
-	const struct cs_section_def *sect = NULL;
-	const struct cs_extent_def *ext = NULL;
+	u32 count = 0;
 	int ctx_reg_offset;
 
 	if (adev->gfx.rlc.cs_data == NULL)
@@ -4099,39 +4332,15 @@ static void gfx_v10_0_get_csb_buffer(struct amdgpu_device *adev,
 	if (buffer == NULL)
 		return;
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1));
-	buffer[count++] = cpu_to_le32(0x80000000);
-	buffer[count++] = cpu_to_le32(0x80000000);
-
-	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
-		for (ext = sect->section; ext->extent != NULL; ++ext) {
-			if (sect->id == SECT_CONTEXT) {
-				buffer[count++] =
-					cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count));
-				buffer[count++] = cpu_to_le32(ext->reg_index -
-						PACKET3_SET_CONTEXT_REG_START);
-				for (i = 0; i < ext->reg_count; i++)
-					buffer[count++] = cpu_to_le32(ext->extent[i]);
-			} else {
-				return;
-			}
-		}
-	}
+	count = amdgpu_gfx_csb_preamble_start(buffer);
+	count = amdgpu_gfx_csb_data_parser(adev, buffer, count);
 
-	ctx_reg_offset =
-		SOC15_REG_OFFSET(GC, 0, mmPA_SC_TILE_STEERING_OVERRIDE) - PACKET3_SET_CONTEXT_REG_START;
+	ctx_reg_offset = SOC15_REG_OFFSET(GC, 0, mmPA_SC_TILE_STEERING_OVERRIDE) - PACKET3_SET_CONTEXT_REG_START;
 	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, 1));
 	buffer[count++] = cpu_to_le32(ctx_reg_offset);
 	buffer[count++] = cpu_to_le32(adev->gfx.config.pa_sc_tile_steering_override);
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0));
-	buffer[count++] = cpu_to_le32(0);
+	amdgpu_gfx_csb_preamble_end(buffer, count);
 }
 
 static void gfx_v10_0_rlc_fini(struct amdgpu_device *adev)
@@ -4151,22 +4360,22 @@ static void gfx_v10_0_init_rlcg_reg_access_ctrl(struct amdgpu_device *adev)
 {
 	struct amdgpu_rlcg_reg_access_ctrl *reg_access_ctrl;
 
-	reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl;
+	reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl[0];
 	reg_access_ctrl->scratch_reg0 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0);
 	reg_access_ctrl->scratch_reg1 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG1);
 	reg_access_ctrl->scratch_reg2 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG2);
 	reg_access_ctrl->scratch_reg3 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG3);
 	reg_access_ctrl->grbm_cntl = SOC15_REG_OFFSET(GC, 0, mmGRBM_GFX_CNTL);
 	reg_access_ctrl->grbm_idx = SOC15_REG_OFFSET(GC, 0, mmGRBM_GFX_INDEX);
-	switch (adev->ip_versions[GC_HWIP][0]) {
-		case IP_VERSION(10, 3, 0):
-			reg_access_ctrl->spare_int =
-				SOC15_REG_OFFSET(GC, 0, mmRLC_SPARE_INT_0_Sienna_Cichlid);
-			break;
-		default:
-			reg_access_ctrl->spare_int =
-				SOC15_REG_OFFSET(GC, 0, mmRLC_SPARE_INT);
-			break;
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(10, 3, 0):
+		reg_access_ctrl->spare_int =
+			SOC15_REG_OFFSET(GC, 0, mmRLC_SPARE_INT_0_Sienna_Cichlid);
+		break;
+	default:
+		reg_access_ctrl->spare_int =
+			SOC15_REG_OFFSET(GC, 0, mmRLC_SPARE_INT);
+		break;
 	}
 	adev->gfx.rlc.rlcg_reg_access_supported = true;
 }
@@ -4187,11 +4396,6 @@ static int gfx_v10_0_rlc_init(struct amdgpu_device *adev)
 			return r;
 	}
 
-	/* init spm vmid with 0xf */
-	if (adev->gfx.rlc.funcs->update_spm_vmid)
-		adev->gfx.rlc.funcs->update_spm_vmid(adev, 0xf);
-
-
 	return 0;
 }
 
@@ -4213,13 +4417,13 @@ static int gfx_v10_0_mec_init(struct amdgpu_device *adev)
 	int r;
 	u32 *hpd;
 	const __le32 *fw_data = NULL;
-	unsigned fw_size;
+	unsigned int fw_size;
 	u32 *fw = NULL;
 	size_t mec_hpd_size;
 
 	const struct gfx_firmware_header_v1_0 *mec_hdr = NULL;
 
-	bitmap_zero(adev->gfx.mec.queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
+	bitmap_zero(adev->gfx.mec_bitmap[0].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
 
 	/* take ownership of the relevant compute queues */
 	amdgpu_gfx_compute_queue_acquire(adev);
@@ -4291,11 +4495,12 @@ static void wave_read_regs(struct amdgpu_device *adev, uint32_t wave,
 		*(out++) = RREG32_SOC15(GC, 0, mmSQ_IND_DATA);
 }
 
-static void gfx_v10_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
+static void gfx_v10_0_read_wave_data(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
 {
 	/* in gfx10 the SIMD_ID is specified as part of the INSTANCE
 	 * field when performing a select_se_sh so it should be
-	 * zero here */
+	 * zero here
+	 */
 	WARN_ON(simd != 0);
 
 	/* type 2 wave data */
@@ -4318,7 +4523,7 @@ static void gfx_v10_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd,
 	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_MODE);
 }
 
-static void gfx_v10_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v10_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				     uint32_t wave, uint32_t start,
 				     uint32_t size, uint32_t *dst)
 {
@@ -4329,7 +4534,7 @@ static void gfx_v10_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
 		dst);
 }
 
-static void gfx_v10_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v10_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				      uint32_t wave, uint32_t thread,
 				      uint32_t start, uint32_t size,
 				      uint32_t *dst)
@@ -4340,7 +4545,7 @@ static void gfx_v10_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd,
 }
 
 static void gfx_v10_0_select_me_pipe_q(struct amdgpu_device *adev,
-				       u32 me, u32 pipe, u32 q, u32 vm)
+				       u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id)
 {
 	nv_grbm_select(adev, me, pipe, q, vm);
 }
@@ -4376,7 +4581,7 @@ static void gfx_v10_0_gpu_early_init(struct amdgpu_device *adev)
 {
 	u32 gb_addr_config;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 	case IP_VERSION(10, 1, 1):
 	case IP_VERSION(10, 1, 2):
@@ -4461,7 +4666,7 @@ static int gfx_v10_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id,
 		ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
 	else
 		ring->doorbell_index = adev->doorbell_index.gfx_ring1 << 1;
-	ring->vm_hub = AMDGPU_GFXHUB_0;
+	ring->vm_hub = AMDGPU_GFXHUB(0);
 	sprintf(ring->name, "gfx_%d.%d.%d", ring->me, ring->pipe, ring->queue);
 
 	irq_type = AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP + ring->pipe;
@@ -4474,7 +4679,7 @@ static int gfx_v10_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id,
 static int gfx_v10_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
 				       int mec, int pipe, int queue)
 {
-	unsigned irq_type;
+	unsigned int irq_type;
 	struct amdgpu_ring *ring;
 	unsigned int hw_prio;
 
@@ -4490,7 +4695,7 @@ static int gfx_v10_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
 	ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
 	ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
 				+ (ring_id * GFX10_MEC_HPD_SIZE);
-	ring->vm_hub = AMDGPU_GFXHUB_0;
+	ring->vm_hub = AMDGPU_GFXHUB(0);
 	sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
 
 	irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
@@ -4503,13 +4708,57 @@ static int gfx_v10_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
 			     hw_prio, NULL);
 }
 
-static int gfx_v10_0_sw_init(void *handle)
+static void gfx_v10_0_alloc_ip_dump(struct amdgpu_device *adev)
+{
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_10_1);
+	uint32_t *ptr;
+	uint32_t inst;
+
+	ptr = kcalloc(reg_count, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for GFX IP Dump\n");
+		adev->gfx.ip_dump_core = NULL;
+	} else {
+		adev->gfx.ip_dump_core = ptr;
+	}
+
+	/* Allocate memory for compute queue registers for all the instances */
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_10);
+	inst = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe_per_mec *
+		adev->gfx.mec.num_queue_per_pipe;
+
+	ptr = kcalloc(reg_count * inst, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for Compute Queues IP Dump\n");
+		adev->gfx.ip_dump_compute_queues = NULL;
+	} else {
+		adev->gfx.ip_dump_compute_queues = ptr;
+	}
+
+	/* Allocate memory for gfx queue registers for all the instances */
+	reg_count = ARRAY_SIZE(gc_gfx_queue_reg_list_10);
+	inst = adev->gfx.me.num_me * adev->gfx.me.num_pipe_per_me *
+		adev->gfx.me.num_queue_per_pipe;
+
+	ptr = kcalloc(reg_count * inst, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for GFX Queues IP Dump\n");
+		adev->gfx.ip_dump_gfx_queues = NULL;
+	} else {
+		adev->gfx.ip_dump_gfx_queues = ptr;
+	}
+}
+
+static int gfx_v10_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int i, j, k, r, ring_id = 0;
-	struct amdgpu_kiq *kiq;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	int xcc_id = 0;
+	struct amdgpu_device *adev = ip_block->adev;
+	int num_queue_per_pipe = 1; /* we only enable 1 KGQ per pipe */
+
+	INIT_DELAYED_WORK(&adev->gfx.idle_work, amdgpu_gfx_profile_idle_work_handler);
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 	case IP_VERSION(10, 1, 1):
 	case IP_VERSION(10, 1, 2):
@@ -4517,7 +4766,7 @@ static int gfx_v10_0_sw_init(void *handle)
 	case IP_VERSION(10, 1, 4):
 		adev->gfx.me.num_me = 1;
 		adev->gfx.me.num_pipe_per_me = 1;
-		adev->gfx.me.num_queue_per_pipe = 1;
+		adev->gfx.me.num_queue_per_pipe = 8;
 		adev->gfx.mec.num_mec = 2;
 		adev->gfx.mec.num_pipe_per_mec = 4;
 		adev->gfx.mec.num_queue_per_pipe = 8;
@@ -4531,8 +4780,8 @@ static int gfx_v10_0_sw_init(void *handle)
 	case IP_VERSION(10, 3, 3):
 	case IP_VERSION(10, 3, 7):
 		adev->gfx.me.num_me = 1;
-		adev->gfx.me.num_pipe_per_me = 1;
-		adev->gfx.me.num_queue_per_pipe = 1;
+		adev->gfx.me.num_pipe_per_me = 2;
+		adev->gfx.me.num_queue_per_pipe = 2;
 		adev->gfx.mec.num_mec = 2;
 		adev->gfx.mec.num_pipe_per_mec = 4;
 		adev->gfx.mec.num_queue_per_pipe = 4;
@@ -4546,11 +4795,79 @@ static int gfx_v10_0_sw_init(void *handle)
 		adev->gfx.mec.num_queue_per_pipe = 8;
 		break;
 	}
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(10, 1, 10):
+	case IP_VERSION(10, 1, 1):
+	case IP_VERSION(10, 1, 2):
+		adev->gfx.cleaner_shader_ptr = gfx_10_1_10_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_10_1_10_cleaner_shader_hex);
+		if (adev->gfx.me_fw_version >= 101 &&
+		    adev->gfx.pfp_fw_version  >= 158 &&
+		    adev->gfx.mec_fw_version >= 151) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	case IP_VERSION(10, 3, 0):
+	case IP_VERSION(10, 3, 1):
+	case IP_VERSION(10, 3, 2):
+	case IP_VERSION(10, 3, 3):
+	case IP_VERSION(10, 3, 4):
+	case IP_VERSION(10, 3, 5):
+		adev->gfx.cleaner_shader_ptr = gfx_10_3_0_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_10_3_0_cleaner_shader_hex);
+		if (adev->gfx.me_fw_version >= 64 &&
+		    adev->gfx.pfp_fw_version >= 100 &&
+		    adev->gfx.mec_fw_version >= 122) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	case IP_VERSION(10, 3, 6):
+		adev->gfx.cleaner_shader_ptr = gfx_10_3_0_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_10_3_0_cleaner_shader_hex);
+		if (adev->gfx.me_fw_version >= 14 &&
+		    adev->gfx.pfp_fw_version >= 17 &&
+		    adev->gfx.mec_fw_version >= 24) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	case IP_VERSION(10, 3, 7):
+		adev->gfx.cleaner_shader_ptr = gfx_10_3_0_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_10_3_0_cleaner_shader_hex);
+		if (adev->gfx.me_fw_version >= 4 &&
+		    adev->gfx.pfp_fw_version >= 9 &&
+		    adev->gfx.mec_fw_version >= 12) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	default:
+		adev->gfx.enable_cleaner_shader = false;
+		break;
+	}
 
 	/* KIQ event */
 	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP,
 			      GFX_10_1__SRCID__CP_IB2_INTERRUPT_PKT,
-			      &adev->gfx.kiq.irq);
+			      &adev->gfx.kiq[0].irq);
 	if (r)
 		return r;
 
@@ -4561,6 +4878,13 @@ static int gfx_v10_0_sw_init(void *handle)
 	if (r)
 		return r;
 
+	/* Bad opcode Event */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP,
+			      GFX_10_1__SRCID__CP_BAD_OPCODE_ERROR,
+			      &adev->gfx.bad_op_irq);
+	if (r)
+		return r;
+
 	/* Privileged reg */
 	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, GFX_10_1__SRCID__CP_PRIV_REG_FAULT,
 			      &adev->gfx.priv_reg_irq);
@@ -4595,7 +4919,7 @@ static int gfx_v10_0_sw_init(void *handle)
 
 	/* set up the gfx ring */
 	for (i = 0; i < adev->gfx.me.num_me; i++) {
-		for (j = 0; j < adev->gfx.me.num_queue_per_pipe; j++) {
+		for (j = 0; j < num_queue_per_pipe; j++) {
 			for (k = 0; k < adev->gfx.me.num_pipe_per_me; k++) {
 				if (!amdgpu_gfx_is_me_queue_enabled(adev, i, k, j))
 					continue;
@@ -4614,8 +4938,8 @@ static int gfx_v10_0_sw_init(void *handle)
 	for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
 		for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
 			for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
-				if (!amdgpu_gfx_is_mec_queue_enabled(adev, i, k,
-								     j))
+				if (!amdgpu_gfx_is_mec_queue_enabled(adev, 0, i,
+								     k, j))
 					continue;
 
 				r = gfx_v10_0_compute_ring_init(adev, ring_id,
@@ -4628,20 +4952,27 @@ static int gfx_v10_0_sw_init(void *handle)
 		}
 	}
 
-	if (!adev->enable_mes_kiq) {
-		r = amdgpu_gfx_kiq_init(adev, GFX10_MEC_HPD_SIZE);
-		if (r) {
-			DRM_ERROR("Failed to init KIQ BOs!\n");
-			return r;
-		}
+	adev->gfx.gfx_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]);
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+	if (!amdgpu_sriov_vf(adev) &&
+	    !adev->debug_disable_gpu_ring_reset) {
+		adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+		adev->gfx.gfx_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+	}
 
-		kiq = &adev->gfx.kiq;
-		r = amdgpu_gfx_kiq_init_ring(adev, &kiq->ring, &kiq->irq);
-		if (r)
-			return r;
+	r = amdgpu_gfx_kiq_init(adev, GFX10_MEC_HPD_SIZE, 0);
+	if (r) {
+		DRM_ERROR("Failed to init KIQ BOs!\n");
+		return r;
 	}
 
-	r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct v10_compute_mqd));
+	r = amdgpu_gfx_kiq_init_ring(adev, xcc_id);
+	if (r)
+		return r;
+
+	r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct v10_compute_mqd), 0);
 	if (r)
 		return r;
 
@@ -4656,6 +4987,12 @@ static int gfx_v10_0_sw_init(void *handle)
 
 	gfx_v10_0_gpu_early_init(adev);
 
+	gfx_v10_0_alloc_ip_dump(adev);
+
+	r = amdgpu_gfx_sysfs_init(adev);
+	if (r)
+		return r;
+
 	return 0;
 }
 
@@ -4680,22 +5017,22 @@ static void gfx_v10_0_me_fini(struct amdgpu_device *adev)
 			      (void **)&adev->gfx.me.me_fw_ptr);
 }
 
-static int gfx_v10_0_sw_fini(void *handle)
+static int gfx_v10_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
 		amdgpu_ring_fini(&adev->gfx.gfx_ring[i]);
 	for (i = 0; i < adev->gfx.num_compute_rings; i++)
 		amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
 
-	amdgpu_gfx_mqd_sw_fini(adev);
+	amdgpu_gfx_mqd_sw_fini(adev, 0);
 
-	if (!adev->enable_mes_kiq) {
-		amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring);
-		amdgpu_gfx_kiq_fini(adev);
-	}
+	amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq[0].ring);
+	amdgpu_gfx_kiq_fini(adev, 0);
+
+	amdgpu_gfx_cleaner_shader_sw_fini(adev);
 
 	gfx_v10_0_pfp_fini(adev);
 	gfx_v10_0_ce_fini(adev);
@@ -4707,12 +5044,17 @@ static int gfx_v10_0_sw_fini(void *handle)
 		gfx_v10_0_rlc_backdoor_autoload_buffer_fini(adev);
 
 	gfx_v10_0_free_microcode(adev);
+	amdgpu_gfx_sysfs_fini(adev);
+
+	kfree(adev->gfx.ip_dump_core);
+	kfree(adev->gfx.ip_dump_compute_queues);
+	kfree(adev->gfx.ip_dump_gfx_queues);
 
 	return 0;
 }
 
 static void gfx_v10_0_select_se_sh(struct amdgpu_device *adev, u32 se_num,
-				   u32 sh_num, u32 instance)
+				   u32 sh_num, u32 instance, int xcc_id)
 {
 	u32 data;
 
@@ -4767,18 +5109,21 @@ static void gfx_v10_0_setup_rb(struct amdgpu_device *adev)
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
 			bitmap = i * adev->gfx.config.max_sh_per_se + j;
-			if (((adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 0)) ||
-				(adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 3)) ||
-				(adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 6))) &&
+			if (((amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			      IP_VERSION(10, 3, 0)) ||
+			     (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			      IP_VERSION(10, 3, 3)) ||
+			     (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			      IP_VERSION(10, 3, 6))) &&
 			    ((gfx_v10_3_get_disabled_sa(adev) >> bitmap) & 1))
 				continue;
-			gfx_v10_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v10_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			data = gfx_v10_0_get_rb_active_bitmap(adev);
 			active_rbs |= data << ((i * adev->gfx.config.max_sh_per_se + j) *
 					       rb_bitmap_width_per_sh);
 		}
 	}
-	gfx_v10_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v10_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	adev->gfx.config.backend_enable_mask = active_rbs;
@@ -4795,8 +5140,9 @@ static u32 gfx_v10_0_init_pa_sc_tile_steering_override(struct amdgpu_device *ade
 	uint32_t pa_sc_tile_steering_override;
 
 	/* for ASICs that integrates GFX v10.3
-	 * pa_sc_tile_steering_override should be set to 0 */
-	if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0))
+	 * pa_sc_tile_steering_override should be set to 0
+	 */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 3, 0))
 		return 0;
 
 	/* init num_sc */
@@ -4825,6 +5171,29 @@ static u32 gfx_v10_0_init_pa_sc_tile_steering_override(struct amdgpu_device *ade
 
 #define DEFAULT_SH_MEM_BASES	(0x6000)
 
+static void gfx_v10_0_debug_trap_config_init(struct amdgpu_device *adev,
+				uint32_t first_vmid,
+				uint32_t last_vmid)
+{
+	uint32_t data;
+	uint32_t trap_config_vmid_mask = 0;
+	int i;
+
+	/* Calculate trap config vmid mask */
+	for (i = first_vmid; i < last_vmid; i++)
+		trap_config_vmid_mask |= (1 << i);
+
+	data = REG_SET_FIELD(0, SPI_GDBG_TRAP_CONFIG,
+			VMID_SEL, trap_config_vmid_mask);
+	data = REG_SET_FIELD(data, SPI_GDBG_TRAP_CONFIG,
+			TRAP_EN, 1);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_CONFIG), data);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_DATA0), 0);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_DATA1), 0);
+}
+
 static void gfx_v10_0_init_compute_vmid(struct amdgpu_device *adev)
 {
 	int i;
@@ -4848,14 +5217,19 @@ static void gfx_v10_0_init_compute_vmid(struct amdgpu_device *adev)
 	nv_grbm_select(adev, 0, 0, 0, 0);
 	mutex_unlock(&adev->srbm_mutex);
 
-	/* Initialize all compute VMIDs to have no GDS, GWS, or OA
-	   access. These should be enabled by FW for target VMIDs. */
+	/*
+	 * Initialize all compute VMIDs to have no GDS, GWS, or OA
+	 * access. These should be enabled by FW for target VMIDs.
+	 */
 	for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) {
 		WREG32_SOC15_OFFSET(GC, 0, mmGDS_VMID0_BASE, 2 * i, 0);
 		WREG32_SOC15_OFFSET(GC, 0, mmGDS_VMID0_SIZE, 2 * i, 0);
 		WREG32_SOC15_OFFSET(GC, 0, mmGDS_GWS_VMID0, i, 0);
 		WREG32_SOC15_OFFSET(GC, 0, mmGDS_OA_VMID0, i, 0);
 	}
+
+	gfx_v10_0_debug_trap_config_init(adev, adev->vm_manager.first_kfd_vmid,
+					AMDGPU_NUM_VMID);
 }
 
 static void gfx_v10_0_init_gds_vmid(struct amdgpu_device *adev)
@@ -4907,7 +5281,7 @@ static void gfx_v10_0_tcp_harvest(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v10_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v10_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			wgp_active_bitmap = gfx_v10_0_get_wgp_active_bitmap_per_sh(adev);
 			/*
 			 * Set corresponding TCP bits for the inactive WGPs in
@@ -4940,7 +5314,7 @@ static void gfx_v10_0_tcp_harvest(struct amdgpu_device *adev)
 		}
 	}
 
-	gfx_v10_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v10_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 }
 
@@ -4949,7 +5323,7 @@ static void gfx_v10_0_get_tcc_info(struct amdgpu_device *adev)
 	/* TCCs are global (not instanced). */
 	uint32_t tcc_disable;
 
-	if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0)) {
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 3, 0)) {
 		tcc_disable = RREG32_SOC15(GC, 0, mmCGTS_TCC_DISABLE_gc_10_3) |
 			      RREG32_SOC15(GC, 0, mmCGTS_USER_TCC_DISABLE_gc_10_3);
 	} else {
@@ -4967,7 +5341,8 @@ static void gfx_v10_0_constants_init(struct amdgpu_device *adev)
 	u32 tmp;
 	int i;
 
-	WREG32_FIELD15(GC, 0, GRBM_CNTL, READ_TIMEOUT, 0xff);
+	if (!amdgpu_sriov_vf(adev))
+		WREG32_FIELD15(GC, 0, GRBM_CNTL, READ_TIMEOUT, 0xff);
 
 	gfx_v10_0_setup_rb(adev);
 	gfx_v10_0_get_cu_info(adev, &adev->gfx.cu_info);
@@ -4978,7 +5353,7 @@ static void gfx_v10_0_constants_init(struct amdgpu_device *adev)
 	/* XXX SH_MEM regs */
 	/* where to put LDS, scratch, GPUVM in FSA64 space */
 	mutex_lock(&adev->srbm_mutex);
-	for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB_0].num_ids; i++) {
+	for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
 		nv_grbm_select(adev, 0, 0, 0, i);
 		/* CP and shaders */
 		WREG32_SOC15(GC, 0, mmSH_MEM_CONFIG, DEFAULT_SH_MEM_CONFIG);
@@ -4999,26 +5374,74 @@ static void gfx_v10_0_constants_init(struct amdgpu_device *adev)
 
 }
 
+static u32 gfx_v10_0_get_cpg_int_cntl(struct amdgpu_device *adev,
+				      int me, int pipe)
+{
+	if (me != 0)
+		return 0;
+
+	switch (pipe) {
+	case 0:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_INT_CNTL_RING0);
+	case 1:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_INT_CNTL_RING1);
+	default:
+		return 0;
+	}
+}
+
+static u32 gfx_v10_0_get_cpc_int_cntl(struct amdgpu_device *adev,
+				      int me, int pipe)
+{
+	/*
+	 * amdgpu controls only the first MEC. That's why this function only
+	 * handles the setting of interrupts for this specific MEC. All other
+	 * pipes' interrupts are set by amdkfd.
+	 */
+	if (me != 1)
+		return 0;
+
+	switch (pipe) {
+	case 0:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE0_INT_CNTL);
+	case 1:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE1_INT_CNTL);
+	case 2:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE2_INT_CNTL);
+	case 3:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE3_INT_CNTL);
+	default:
+		return 0;
+	}
+}
+
 static void gfx_v10_0_enable_gui_idle_interrupt(struct amdgpu_device *adev,
 					       bool enable)
 {
-	u32 tmp;
+	u32 tmp, cp_int_cntl_reg;
+	int i, j;
 
 	if (amdgpu_sriov_vf(adev))
 		return;
 
-	tmp = RREG32_SOC15(GC, 0, mmCP_INT_CNTL_RING0);
-
-	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE,
-			    enable ? 1 : 0);
-	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_EMPTY_INT_ENABLE,
-			    enable ? 1 : 0);
-	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE,
-			    enable ? 1 : 0);
-	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, GFX_IDLE_INT_ENABLE,
-			    enable ? 1 : 0);
-
-	WREG32_SOC15(GC, 0, mmCP_INT_CNTL_RING0, tmp);
+	for (i = 0; i < adev->gfx.me.num_me; i++) {
+		for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+			cp_int_cntl_reg = gfx_v10_0_get_cpg_int_cntl(adev, i, j);
+
+			if (cp_int_cntl_reg) {
+				tmp = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE,
+						    enable ? 1 : 0);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_EMPTY_INT_ENABLE,
+						    enable ? 1 : 0);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE,
+						    enable ? 1 : 0);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, GFX_IDLE_INT_ENABLE,
+						    enable ? 1 : 0);
+				WREG32_SOC15_IP(GC, cp_int_cntl_reg, tmp);
+			}
+		}
+	}
 }
 
 static int gfx_v10_0_init_csb(struct amdgpu_device *adev)
@@ -5026,7 +5449,7 @@ static int gfx_v10_0_init_csb(struct amdgpu_device *adev)
 	adev->gfx.rlc.funcs->get_csb_buffer(adev, adev->gfx.rlc.cs_ptr);
 
 	/* csib */
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 2)) {
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 1, 2)) {
 		WREG32_SOC15_RLC(GC, 0, mmRLC_CSIB_ADDR_HI,
 				adev->gfx.rlc.clear_state_gpu_addr >> 32);
 		WREG32_SOC15_RLC(GC, 0, mmRLC_CSIB_ADDR_LO,
@@ -5082,8 +5505,10 @@ static void gfx_v10_0_rlc_smu_handshake_cntl(struct amdgpu_device *adev,
 
 static void gfx_v10_0_rlc_start(struct amdgpu_device *adev)
 {
-	/* TODO: enable rlc & smu handshake until smu
-	 * and gfxoff feature works as expected */
+	/*
+	 * TODO: enable rlc & smu handshake until smu
+	 * and gfxoff feature works as expected
+	 */
 	if (!(amdgpu_pp_feature_mask & PP_GFXOFF_MASK))
 		gfx_v10_0_rlc_smu_handshake_cntl(adev, false);
 
@@ -5106,7 +5531,7 @@ static int gfx_v10_0_rlc_load_microcode(struct amdgpu_device *adev)
 {
 	const struct rlc_firmware_header_v2_0 *hdr;
 	const __le32 *fw_data;
-	unsigned i, fw_size;
+	unsigned int i, fw_size;
 
 	if (!adev->gfx.rlc_fw)
 		return -EINVAL;
@@ -5143,6 +5568,8 @@ static int gfx_v10_0_rlc_resume(struct amdgpu_device *adev)
 
 		gfx_v10_0_init_csb(adev);
 
+		gfx_v10_0_update_spm_vmid_internal(adev, 0xf);
+
 		if (!amdgpu_sriov_vf(adev)) /* enable RLC SRM */
 			gfx_v10_0_rlc_enable_srm(adev);
 	} else {
@@ -5173,6 +5600,8 @@ static int gfx_v10_0_rlc_resume(struct amdgpu_device *adev)
 
 		gfx_v10_0_init_csb(adev);
 
+		gfx_v10_0_update_spm_vmid_internal(adev, 0xf);
+
 		adev->gfx.rlc.funcs->start(adev);
 
 		if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) {
@@ -5181,6 +5610,7 @@ static int gfx_v10_0_rlc_resume(struct amdgpu_device *adev)
 				return r;
 		}
 	}
+
 	return 0;
 }
 
@@ -5648,13 +6078,12 @@ static int gfx_v10_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable)
 	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_HALT, enable ? 0 : 1);
 	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_HALT, enable ? 0 : 1);
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 2)) {
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 1, 2))
 		WREG32_SOC15_RLC(GC, 0, mmCP_ME_CNTL, tmp);
-	} else {
+	else
 		WREG32_SOC15(GC, 0, mmCP_ME_CNTL, tmp);
-	}
 
-	if (adev->job_hang && !enable)
+	if (amdgpu_in_reset(adev) && !enable)
 		return 0;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
@@ -5674,7 +6103,7 @@ static int gfx_v10_0_cp_gfx_load_pfp_microcode(struct amdgpu_device *adev)
 	int r;
 	const struct gfx_firmware_header_v1_0 *pfp_hdr;
 	const __le32 *fw_data;
-	unsigned i, fw_size;
+	unsigned int i, fw_size;
 	uint32_t tmp;
 	uint32_t usec_timeout = 50000;  /* wait for 50ms */
 
@@ -5723,7 +6152,7 @@ static int gfx_v10_0_cp_gfx_load_pfp_microcode(struct amdgpu_device *adev)
 	}
 
 	if (amdgpu_emu_mode == 1)
-		adev->hdp.funcs->flush_hdp(adev, NULL);
+		amdgpu_device_flush_hdp(adev, NULL);
 
 	tmp = RREG32_SOC15(GC, 0, mmCP_PFP_IC_BASE_CNTL);
 	tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, VMID, 0);
@@ -5752,7 +6181,7 @@ static int gfx_v10_0_cp_gfx_load_ce_microcode(struct amdgpu_device *adev)
 	int r;
 	const struct gfx_firmware_header_v1_0 *ce_hdr;
 	const __le32 *fw_data;
-	unsigned i, fw_size;
+	unsigned int i, fw_size;
 	uint32_t tmp;
 	uint32_t usec_timeout = 50000;  /* wait for 50ms */
 
@@ -5801,7 +6230,7 @@ static int gfx_v10_0_cp_gfx_load_ce_microcode(struct amdgpu_device *adev)
 	}
 
 	if (amdgpu_emu_mode == 1)
-		adev->hdp.funcs->flush_hdp(adev, NULL);
+		amdgpu_device_flush_hdp(adev, NULL);
 
 	tmp = RREG32_SOC15(GC, 0, mmCP_CE_IC_BASE_CNTL);
 	tmp = REG_SET_FIELD(tmp, CP_CE_IC_BASE_CNTL, VMID, 0);
@@ -5829,7 +6258,7 @@ static int gfx_v10_0_cp_gfx_load_me_microcode(struct amdgpu_device *adev)
 	int r;
 	const struct gfx_firmware_header_v1_0 *me_hdr;
 	const __le32 *fw_data;
-	unsigned i, fw_size;
+	unsigned int i, fw_size;
 	uint32_t tmp;
 	uint32_t usec_timeout = 50000;  /* wait for 50ms */
 
@@ -5878,7 +6307,7 @@ static int gfx_v10_0_cp_gfx_load_me_microcode(struct amdgpu_device *adev)
 	}
 
 	if (amdgpu_emu_mode == 1)
-		adev->hdp.funcs->flush_hdp(adev, NULL);
+		amdgpu_device_flush_hdp(adev, NULL);
 
 	tmp = RREG32_SOC15(GC, 0, mmCP_ME_IC_BASE_CNTL);
 	tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, VMID, 0);
@@ -6040,7 +6469,7 @@ static void gfx_v10_0_cp_gfx_set_doorbell(struct amdgpu_device *adev,
 		}
 		WREG32_SOC15(GC, 0, mmCP_RB_DOORBELL_CONTROL, tmp);
 	}
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 0):
 	case IP_VERSION(10, 3, 2):
 	case IP_VERSION(10, 3, 1):
@@ -6073,7 +6502,6 @@ static int gfx_v10_0_cp_gfx_resume(struct amdgpu_device *adev)
 	u32 tmp;
 	u32 rb_bufsz;
 	u64 rb_addr, rptr_addr, wptr_gpu_addr;
-	u32 i;
 
 	/* Set the write pointer delay */
 	WREG32_SOC15(GC, 0, mmCP_RB_WPTR_DELAY, 0);
@@ -6100,7 +6528,7 @@ static int gfx_v10_0_cp_gfx_resume(struct amdgpu_device *adev)
 	WREG32_SOC15(GC, 0, mmCP_RB0_WPTR, lower_32_bits(ring->wptr));
 	WREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI, upper_32_bits(ring->wptr));
 
-	/* set the wb address wether it's enabled or not */
+	/* set the wb address whether it's enabled or not */
 	rptr_addr = ring->rptr_gpu_addr;
 	WREG32_SOC15(GC, 0, mmCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr));
 	WREG32_SOC15(GC, 0, mmCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) &
@@ -6138,7 +6566,7 @@ static int gfx_v10_0_cp_gfx_resume(struct amdgpu_device *adev)
 		ring->wptr = 0;
 		WREG32_SOC15(GC, 0, mmCP_RB1_WPTR, lower_32_bits(ring->wptr));
 		WREG32_SOC15(GC, 0, mmCP_RB1_WPTR_HI, upper_32_bits(ring->wptr));
-		/* Set the wb address wether it's enabled or not */
+		/* Set the wb address whether it's enabled or not */
 		rptr_addr = ring->rptr_gpu_addr;
 		WREG32_SOC15(GC, 0, mmCP_RB1_RPTR_ADDR, lower_32_bits(rptr_addr));
 		WREG32_SOC15(GC, 0, mmCP_RB1_RPTR_ADDR_HI, upper_32_bits(rptr_addr) &
@@ -6168,18 +6596,13 @@ static int gfx_v10_0_cp_gfx_resume(struct amdgpu_device *adev)
 	/* start the ring */
 	gfx_v10_0_cp_gfx_start(adev);
 
-	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
-		ring = &adev->gfx.gfx_ring[i];
-		ring->sched.ready = true;
-	}
-
 	return 0;
 }
 
 static void gfx_v10_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
 {
 	if (enable) {
-		switch (adev->ip_versions[GC_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 		case IP_VERSION(10, 3, 0):
 		case IP_VERSION(10, 3, 2):
 		case IP_VERSION(10, 3, 1):
@@ -6195,7 +6618,7 @@ static void gfx_v10_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
 			break;
 		}
 	} else {
-		switch (adev->ip_versions[GC_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 		case IP_VERSION(10, 3, 0):
 		case IP_VERSION(10, 3, 2):
 		case IP_VERSION(10, 3, 1):
@@ -6214,7 +6637,7 @@ static void gfx_v10_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
 				      CP_MEC_CNTL__MEC_ME2_HALT_MASK));
 			break;
 		}
-		adev->gfx.kiq.ring.sched.ready = false;
+		adev->gfx.kiq[0].ring.sched.ready = false;
 	}
 	udelay(50);
 }
@@ -6223,7 +6646,7 @@ static int gfx_v10_0_cp_compute_load_microcode(struct amdgpu_device *adev)
 {
 	const struct gfx_firmware_header_v1_0 *mec_hdr;
 	const __le32 *fw_data;
-	unsigned i;
+	unsigned int i;
 	u32 tmp;
 	u32 usec_timeout = 50000; /* Wait for 50 ms */
 
@@ -6259,7 +6682,7 @@ static int gfx_v10_0_cp_compute_load_microcode(struct amdgpu_device *adev)
 	}
 
 	if (amdgpu_emu_mode == 1)
-		adev->hdp.funcs->flush_hdp(adev, NULL);
+		amdgpu_device_flush_hdp(adev, NULL);
 
 	tmp = RREG32_SOC15(GC, 0, mmCP_CPC_IC_BASE_CNTL);
 	tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, CACHE_POLICY, 0);
@@ -6295,7 +6718,7 @@ static void gfx_v10_0_kiq_setting(struct amdgpu_ring *ring)
 	struct amdgpu_device *adev = ring->adev;
 
 	/* tell RLC which is KIQ queue */
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 0):
 	case IP_VERSION(10, 3, 2):
 	case IP_VERSION(10, 3, 1):
@@ -6307,17 +6730,13 @@ static void gfx_v10_0_kiq_setting(struct amdgpu_ring *ring)
 		tmp = RREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS_Sienna_Cichlid);
 		tmp &= 0xffffff00;
 		tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
-		WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS_Sienna_Cichlid, tmp);
-		tmp |= 0x80;
-		WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS_Sienna_Cichlid, tmp);
+		WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS_Sienna_Cichlid, tmp | 0x80);
 		break;
 	default:
 		tmp = RREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS);
 		tmp &= 0xffffff00;
 		tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
-		WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS, tmp);
-		tmp |= 0x80;
-		WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS, tmp);
+		WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS, tmp | 0x80);
 		break;
 	}
 }
@@ -6423,62 +6842,13 @@ static int gfx_v10_0_gfx_mqd_init(struct amdgpu_device *adev, void *m,
 	return 0;
 }
 
-#ifdef BRING_UP_DEBUG
-static int gfx_v10_0_gfx_queue_init_register(struct amdgpu_ring *ring)
-{
-	struct amdgpu_device *adev = ring->adev;
-	struct v10_gfx_mqd *mqd = ring->mqd_ptr;
-
-	/* set mmCP_GFX_HQD_WPTR/_HI to 0 */
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_WPTR, mqd->cp_gfx_hqd_wptr);
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_WPTR_HI, mqd->cp_gfx_hqd_wptr_hi);
-
-	/* set GFX_MQD_BASE */
-	WREG32_SOC15(GC, 0, mmCP_MQD_BASE_ADDR, mqd->cp_mqd_base_addr);
-	WREG32_SOC15(GC, 0, mmCP_MQD_BASE_ADDR_HI, mqd->cp_mqd_base_addr_hi);
-
-	/* set GFX_MQD_CONTROL */
-	WREG32_SOC15(GC, 0, mmCP_GFX_MQD_CONTROL, mqd->cp_gfx_mqd_control);
-
-	/* set GFX_HQD_VMID to 0 */
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_VMID, mqd->cp_gfx_hqd_vmid);
-
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_QUEUE_PRIORITY,
-			mqd->cp_gfx_hqd_queue_priority);
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_QUANTUM, mqd->cp_gfx_hqd_quantum);
-
-	/* set GFX_HQD_BASE, similar as CP_RB_BASE */
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_BASE, mqd->cp_gfx_hqd_base);
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_BASE_HI, mqd->cp_gfx_hqd_base_hi);
-
-	/* set GFX_HQD_RPTR_ADDR, similar as CP_RB_RPTR */
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_RPTR_ADDR, mqd->cp_gfx_hqd_rptr_addr);
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_RPTR_ADDR_HI, mqd->cp_gfx_hqd_rptr_addr_hi);
-
-	/* set GFX_HQD_CNTL, similar as CP_RB_CNTL */
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_CNTL, mqd->cp_gfx_hqd_cntl);
-
-	/* set RB_WPTR_POLL_ADDR */
-	WREG32_SOC15(GC, 0, mmCP_RB_WPTR_POLL_ADDR_LO, mqd->cp_rb_wptr_poll_addr_lo);
-	WREG32_SOC15(GC, 0, mmCP_RB_WPTR_POLL_ADDR_HI, mqd->cp_rb_wptr_poll_addr_hi);
-
-	/* set RB_DOORBELL_CONTROL */
-	WREG32_SOC15(GC, 0, mmCP_RB_DOORBELL_CONTROL, mqd->cp_rb_doorbell_control);
-
-	/* active the queue */
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_ACTIVE, mqd->cp_gfx_hqd_active);
-
-	return 0;
-}
-#endif
-
-static int gfx_v10_0_gfx_init_queue(struct amdgpu_ring *ring)
+static int gfx_v10_0_kgq_init_queue(struct amdgpu_ring *ring, bool reset)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct v10_gfx_mqd *mqd = ring->mqd_ptr;
 	int mqd_idx = ring - &adev->gfx.gfx_ring[0];
 
-	if (!amdgpu_in_reset(adev) && !adev->in_suspend) {
+	if (!reset && !amdgpu_in_reset(adev) && !adev->in_suspend) {
 		memset((void *)mqd, 0, sizeof(*mqd));
 		mutex_lock(&adev->srbm_mutex);
 		nv_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
@@ -6492,96 +6862,45 @@ static int gfx_v10_0_gfx_init_queue(struct amdgpu_ring *ring)
 		if (ring->doorbell_index == adev->doorbell_index.gfx_ring0 << 1)
 			gfx_v10_0_cp_gfx_set_doorbell(adev, ring);
 
-#ifdef BRING_UP_DEBUG
-		gfx_v10_0_gfx_queue_init_register(ring);
-#endif
 		nv_grbm_select(adev, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 		if (adev->gfx.me.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.me.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
-	} else if (amdgpu_in_reset(adev)) {
-		/* reset mqd with the backup copy */
-		if (adev->gfx.me.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.me.mqd_backup[mqd_idx], sizeof(*mqd));
-		/* reset the ring */
-		ring->wptr = 0;
-		*ring->wptr_cpu_addr = 0;
-		amdgpu_ring_clear_ring(ring);
-#ifdef BRING_UP_DEBUG
+			memcpy_fromio(adev->gfx.me.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+	} else {
 		mutex_lock(&adev->srbm_mutex);
 		nv_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
-		gfx_v10_0_gfx_queue_init_register(ring);
+		if (ring->doorbell_index == adev->doorbell_index.gfx_ring0 << 1)
+			gfx_v10_0_cp_gfx_set_doorbell(adev, ring);
+
 		nv_grbm_select(adev, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
-#endif
-	} else {
+		/* restore mqd with the backup copy */
+		if (adev->gfx.me.mqd_backup[mqd_idx])
+			memcpy_toio(mqd, adev->gfx.me.mqd_backup[mqd_idx], sizeof(*mqd));
+		/* reset the ring */
+		ring->wptr = 0;
+		*ring->wptr_cpu_addr = 0;
 		amdgpu_ring_clear_ring(ring);
 	}
 
 	return 0;
 }
 
-#ifndef BRING_UP_DEBUG
-static int gfx_v10_0_kiq_enable_kgq(struct amdgpu_device *adev)
-{
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
-	int r, i;
-
-	if (!kiq->pmf || !kiq->pmf->kiq_map_queues)
-		return -EINVAL;
-
-	r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size *
-					adev->gfx.num_gfx_rings);
-	if (r) {
-		DRM_ERROR("Failed to lock KIQ (%d).\n", r);
-		return r;
-	}
-
-	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
-		kiq->pmf->kiq_map_queues(kiq_ring, &adev->gfx.gfx_ring[i]);
-
-	return amdgpu_ring_test_helper(kiq_ring);
-}
-#endif
-
 static int gfx_v10_0_cp_async_gfx_ring_resume(struct amdgpu_device *adev)
 {
 	int r, i;
-	struct amdgpu_ring *ring;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
-		ring = &adev->gfx.gfx_ring[i];
-
-		r = amdgpu_bo_reserve(ring->mqd_obj, false);
-		if (unlikely(r != 0))
-			goto done;
-
-		r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr);
-		if (!r) {
-			r = gfx_v10_0_gfx_init_queue(ring);
-			amdgpu_bo_kunmap(ring->mqd_obj);
-			ring->mqd_ptr = NULL;
-		}
-		amdgpu_bo_unreserve(ring->mqd_obj);
+		r = gfx_v10_0_kgq_init_queue(&adev->gfx.gfx_ring[i], false);
 		if (r)
-			goto done;
+			return r;
 	}
-#ifndef BRING_UP_DEBUG
-	r = gfx_v10_0_kiq_enable_kgq(adev);
-	if (r)
-		goto done;
-#endif
-	r = gfx_v10_0_cp_gfx_start(adev);
+
+	r = amdgpu_gfx_enable_kgq(adev, 0);
 	if (r)
-		goto done;
+		return r;
 
-	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
-		ring = &adev->gfx.gfx_ring[i];
-		ring->sched.ready = true;
-	}
-done:
-	return r;
+	return gfx_v10_0_cp_gfx_start(adev);
 }
 
 static int gfx_v10_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
@@ -6658,8 +6977,9 @@ static int gfx_v10_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
 #ifdef __BIG_ENDIAN
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, ENDIAN_SWAP, 1);
 #endif
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 0);
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TUNNEL_DISPATCH, 0);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 1);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TUNNEL_DISPATCH,
+			    prop->allow_tunneling);
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, PRIV_STATE, 1);
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, KMD_QUEUE, 1);
 	mqd->cp_hqd_pq_control = tmp;
@@ -6812,14 +7132,13 @@ static int gfx_v10_0_kiq_init_queue(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct v10_compute_mqd *mqd = ring->mqd_ptr;
-	int mqd_idx = AMDGPU_MAX_COMPUTE_RINGS;
 
 	gfx_v10_0_kiq_setting(ring);
 
 	if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */
 		/* reset MQD to a clean status */
-		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
+		if (adev->gfx.kiq[0].mqd_backup)
+			memcpy_toio(mqd, adev->gfx.kiq[0].mqd_backup, sizeof(*mqd));
 
 		/* reset ring buffer */
 		ring->wptr = 0;
@@ -6841,20 +7160,20 @@ static int gfx_v10_0_kiq_init_queue(struct amdgpu_ring *ring)
 		nv_grbm_select(adev, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 
-		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+		if (adev->gfx.kiq[0].mqd_backup)
+			memcpy_fromio(adev->gfx.kiq[0].mqd_backup, mqd, sizeof(*mqd));
 	}
 
 	return 0;
 }
 
-static int gfx_v10_0_kcq_init_queue(struct amdgpu_ring *ring)
+static int gfx_v10_0_kcq_init_queue(struct amdgpu_ring *ring, bool restore)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct v10_compute_mqd *mqd = ring->mqd_ptr;
 	int mqd_idx = ring - &adev->gfx.compute_ring[0];
 
-	if (!amdgpu_in_reset(adev) && !adev->in_suspend) {
+	if (!restore && !amdgpu_in_reset(adev) && !adev->in_suspend) {
 		memset((void *)mqd, 0, sizeof(*mqd));
 		mutex_lock(&adev->srbm_mutex);
 		nv_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
@@ -6863,18 +7182,15 @@ static int gfx_v10_0_kcq_init_queue(struct amdgpu_ring *ring)
 		mutex_unlock(&adev->srbm_mutex);
 
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
-	} else if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */
-		/* reset MQD to a clean status */
+			memcpy_fromio(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+	} else {
+		/* restore MQD to a clean status */
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
-
+			memcpy_toio(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
 		/* reset ring buffer */
 		ring->wptr = 0;
 		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, 0);
 		amdgpu_ring_clear_ring(ring);
-	} else {
-		amdgpu_ring_clear_ring(ring);
 	}
 
 	return 0;
@@ -6882,54 +7198,24 @@ static int gfx_v10_0_kcq_init_queue(struct amdgpu_ring *ring)
 
 static int gfx_v10_0_kiq_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring;
-	int r;
-
-	ring = &adev->gfx.kiq.ring;
-
-	r = amdgpu_bo_reserve(ring->mqd_obj, false);
-	if (unlikely(r != 0))
-		return r;
-
-	r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr);
-	if (unlikely(r != 0))
-		return r;
-
-	gfx_v10_0_kiq_init_queue(ring);
-	amdgpu_bo_kunmap(ring->mqd_obj);
-	ring->mqd_ptr = NULL;
-	amdgpu_bo_unreserve(ring->mqd_obj);
-	ring->sched.ready = true;
+	gfx_v10_0_kiq_init_queue(&adev->gfx.kiq[0].ring);
 	return 0;
 }
 
 static int gfx_v10_0_kcq_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring = NULL;
-	int r = 0, i;
+	int i, r;
 
 	gfx_v10_0_cp_compute_enable(adev, true);
 
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-		ring = &adev->gfx.compute_ring[i];
-
-		r = amdgpu_bo_reserve(ring->mqd_obj, false);
-		if (unlikely(r != 0))
-			goto done;
-		r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr);
-		if (!r) {
-			r = gfx_v10_0_kcq_init_queue(ring);
-			amdgpu_bo_kunmap(ring->mqd_obj);
-			ring->mqd_ptr = NULL;
-		}
-		amdgpu_bo_unreserve(ring->mqd_obj);
+		r = gfx_v10_0_kcq_init_queue(&adev->gfx.compute_ring[i],
+					     false);
 		if (r)
-			goto done;
+			return r;
 	}
 
-	r = amdgpu_gfx_enable_kcq(adev);
-done:
-	return r;
+	return amdgpu_gfx_enable_kcq(adev, 0);
 }
 
 static int gfx_v10_0_cp_resume(struct amdgpu_device *adev)
@@ -6951,10 +7237,7 @@ static int gfx_v10_0_cp_resume(struct amdgpu_device *adev)
 			return r;
 	}
 
-	if (adev->enable_mes_kiq && adev->mes.kiq_hw_init)
-		r = amdgpu_mes_kiq_hw_init(adev);
-	else
-		r = gfx_v10_0_kiq_resume(adev);
+	r = gfx_v10_0_kiq_resume(adev);
 	if (r)
 		return r;
 
@@ -6999,9 +7282,11 @@ static bool gfx_v10_0_check_grbm_cam_remapping(struct amdgpu_device *adev)
 {
 	uint32_t data, pattern = 0xDEADBEEF;
 
-	/* check if mmVGT_ESGS_RING_SIZE_UMD
-	 * has been remapped to mmVGT_ESGS_RING_SIZE */
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	/*
+	 * check if mmVGT_ESGS_RING_SIZE_UMD
+	 * has been remapped to mmVGT_ESGS_RING_SIZE
+	 */
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 0):
 	case IP_VERSION(10, 3, 2):
 	case IP_VERSION(10, 3, 4):
@@ -7011,12 +7296,10 @@ static bool gfx_v10_0_check_grbm_cam_remapping(struct amdgpu_device *adev)
 		WREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE_UMD, pattern);
 
 		if (RREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE_Sienna_Cichlid) == pattern) {
-			WREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE_UMD , data);
+			WREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE_UMD, data);
 			return true;
-		} else {
-			WREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE_Sienna_Cichlid, data);
-			return false;
 		}
+		WREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE_Sienna_Cichlid, data);
 		break;
 	case IP_VERSION(10, 3, 1):
 	case IP_VERSION(10, 3, 3):
@@ -7031,12 +7314,12 @@ static bool gfx_v10_0_check_grbm_cam_remapping(struct amdgpu_device *adev)
 		if (RREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE) == pattern) {
 			WREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE_UMD, data);
 			return true;
-		} else {
-			WREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE, data);
-			return false;
 		}
+		WREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE, data);
 		break;
 	}
+
+	return false;
 }
 
 static void gfx_v10_0_setup_grbm_cam_remapping(struct amdgpu_device *adev)
@@ -7046,11 +7329,13 @@ static void gfx_v10_0_setup_grbm_cam_remapping(struct amdgpu_device *adev)
 	if (amdgpu_sriov_vf(adev))
 		return;
 
-	/* initialize cam_index to 0
-	 * index will auto-inc after each data writting */
+	/*
+	 * Initialize cam_index to 0
+	 * index will auto-inc after each data writing
+	 */
 	WREG32_SOC15(GC, 0, mmGRBM_CAM_INDEX, 0);
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 0):
 	case IP_VERSION(10, 3, 2):
 	case IP_VERSION(10, 3, 1):
@@ -7177,6 +7462,7 @@ static void gfx_v10_0_setup_grbm_cam_remapping(struct amdgpu_device *adev)
 static void gfx_v10_0_disable_gpa_mode(struct amdgpu_device *adev)
 {
 	uint32_t data;
+
 	data = RREG32_SOC15(GC, 0, mmCPC_PSP_DEBUG);
 	data |= CPC_PSP_DEBUG__GPA_OVERRIDE_MASK;
 	WREG32_SOC15(GC, 0, mmCPC_PSP_DEBUG, data);
@@ -7186,25 +7472,26 @@ static void gfx_v10_0_disable_gpa_mode(struct amdgpu_device *adev)
 	WREG32_SOC15(GC, 0, mmCPG_PSP_DEBUG, data);
 }
 
-static int gfx_v10_0_hw_init(void *handle)
+static int gfx_v10_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (!amdgpu_emu_mode)
 		gfx_v10_0_init_golden_registers(adev);
 
+	amdgpu_gfx_cleaner_shader_init(adev, adev->gfx.cleaner_shader_size,
+				       adev->gfx.cleaner_shader_ptr);
+
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
 		/**
 		 * For gfx 10, rlc firmware loading relies on smu firmware is
 		 * loaded firstly, so in direct type, it has to load smc ucode
 		 * here before rlc.
 		 */
-		if (!(adev->flags & AMD_IS_APU)) {
-			r = amdgpu_pm_load_smu_firmware(adev, NULL);
-			if (r)
-				return r;
-		}
+		r = amdgpu_pm_load_smu_firmware(adev, NULL);
+		if (r)
+			return r;
 		gfx_v10_0_disable_gpa_mode(adev);
 	}
 
@@ -7222,65 +7509,48 @@ static int gfx_v10_0_hw_init(void *handle)
 	 * init golden registers and rlc resume may override some registers,
 	 * reconfig them here
 	 */
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 10) ||
-	    adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 1) ||
-	    adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 2))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 1, 10) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 1, 1) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 1, 2))
 		gfx_v10_0_tcp_harvest(adev);
 
 	r = gfx_v10_0_cp_resume(adev);
 	if (r)
 		return r;
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 0))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 3, 0))
 		gfx_v10_3_program_pbb_mode(adev);
 
-	if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 3, 0) && !amdgpu_sriov_vf(adev))
 		gfx_v10_3_set_power_brake_sequence(adev);
 
 	return r;
 }
 
-#ifndef BRING_UP_DEBUG
-static int gfx_v10_0_kiq_disable_kgq(struct amdgpu_device *adev)
+static int gfx_v10_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-	struct amdgpu_ring *kiq_ring = &kiq->ring;
-	int i;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
-		return -EINVAL;
-
-	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size *
-					adev->gfx.num_gfx_rings))
-		return -ENOMEM;
-
-	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
-		kiq->pmf->kiq_unmap_queues(kiq_ring, &adev->gfx.gfx_ring[i],
-					   PREEMPT_QUEUES, 0, 0);
-	if (!adev->job_hang)
-		return amdgpu_ring_test_helper(kiq_ring);
-	else
-		return 0;
-}
-#endif
-
-static int gfx_v10_0_hw_fini(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	int r;
+	cancel_delayed_work_sync(&adev->gfx.idle_work);
 
 	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
+	amdgpu_irq_put(adev, &adev->gfx.bad_op_irq, 0);
+
+	/* WA added for Vangogh asic fixing the SMU suspend failure
+	 * It needs to set power gating again during gfxoff control
+	 * otherwise the gfxoff disallowing will be failed to set.
+	 */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 3, 1))
+		gfx_v10_0_set_powergating_state(ip_block, AMD_PG_STATE_UNGATE);
 
 	if (!adev->no_hw_access) {
-#ifndef BRING_UP_DEBUG
 		if (amdgpu_async_gfx_ring) {
-			r = gfx_v10_0_kiq_disable_kgq(adev);
-			if (r)
+			if (amdgpu_gfx_disable_kgq(adev, 0))
 				DRM_ERROR("KGQ disable failed\n");
 		}
-#endif
-		if (amdgpu_gfx_disable_kcq(adev))
+
+		if (amdgpu_gfx_disable_kcq(adev, 0))
 			DRM_ERROR("KCQ disable failed\n");
 	}
 
@@ -7297,19 +7567,19 @@ static int gfx_v10_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int gfx_v10_0_suspend(void *handle)
+static int gfx_v10_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	return gfx_v10_0_hw_fini(handle);
+	return gfx_v10_0_hw_fini(ip_block);
 }
 
-static int gfx_v10_0_resume(void *handle)
+static int gfx_v10_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	return gfx_v10_0_hw_init(handle);
+	return gfx_v10_0_hw_init(ip_block);
 }
 
-static bool gfx_v10_0_is_idle(void *handle)
+static bool gfx_v10_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (REG_GET_FIELD(RREG32_SOC15(GC, 0, mmGRBM_STATUS),
 				GRBM_STATUS, GUI_ACTIVE))
@@ -7318,11 +7588,11 @@ static bool gfx_v10_0_is_idle(void *handle)
 		return true;
 }
 
-static int gfx_v10_0_wait_for_idle(void *handle)
+static int gfx_v10_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
-	unsigned i;
+	unsigned int i;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		/* read MC_STATUS */
@@ -7336,11 +7606,11 @@ static int gfx_v10_0_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int gfx_v10_0_soft_reset(void *handle)
+static int gfx_v10_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 grbm_soft_reset = 0;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* GRBM_STATUS */
 	tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS);
@@ -7365,7 +7635,7 @@ static int gfx_v10_0_soft_reset(void *handle)
 
 	/* GRBM_STATUS2 */
 	tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS2);
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 0):
 	case IP_VERSION(10, 3, 2):
 	case IP_VERSION(10, 3, 1):
@@ -7398,19 +7668,17 @@ static int gfx_v10_0_soft_reset(void *handle)
 		/* Disable MEC parsing/prefetching */
 		gfx_v10_0_cp_compute_enable(adev, false);
 
-		if (grbm_soft_reset) {
-			tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
-			tmp |= grbm_soft_reset;
-			dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp);
-			WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
-			tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
+		tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
+		tmp |= grbm_soft_reset;
+		dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp);
+		WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
+		tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
 
-			udelay(50);
+		udelay(50);
 
-			tmp &= ~grbm_soft_reset;
-			WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
-			tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
-		}
+		tmp &= ~grbm_soft_reset;
+		WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
+		tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
 
 		/* Wait a little for things to settle down */
 		udelay(50);
@@ -7422,7 +7690,23 @@ static uint64_t gfx_v10_0_get_gpu_clock_counter(struct amdgpu_device *adev)
 {
 	uint64_t clock, clock_lo, clock_hi, hi_check;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(10, 1, 3):
+	case IP_VERSION(10, 1, 4):
+		preempt_disable();
+		clock_hi = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Cyan_Skillfish);
+		clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Cyan_Skillfish);
+		hi_check = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Cyan_Skillfish);
+		/* The SMUIO TSC clock frequency is 100MHz, which sets 32-bit carry over
+		 * roughly every 42 seconds.
+		 */
+		if (hi_check != clock_hi) {
+			clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Cyan_Skillfish);
+			clock_hi = hi_check;
+		}
+		preempt_enable();
+		clock = clock_lo | (clock_hi << 32ULL);
+		break;
 	case IP_VERSION(10, 3, 1):
 	case IP_VERSION(10, 3, 3):
 	case IP_VERSION(10, 3, 7):
@@ -7503,13 +7787,13 @@ static void gfx_v10_0_ring_emit_gds_switch(struct amdgpu_ring *ring,
 				    (1 << (oa_size + oa_base)) - (1 << oa_base));
 }
 
-static int gfx_v10_0_early_init(void *handle)
+static int gfx_v10_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->gfx.funcs = &gfx_v10_0_gfx_funcs;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 	case IP_VERSION(10, 1, 1):
 	case IP_VERSION(10, 1, 2):
@@ -7547,9 +7831,9 @@ static int gfx_v10_0_early_init(void *handle)
 	return gfx_v10_0_init_microcode(adev);
 }
 
-static int gfx_v10_0_late_init(void *handle)
+static int gfx_v10_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_irq_get(adev, &adev->gfx.priv_reg_irq, 0);
@@ -7560,6 +7844,10 @@ static int gfx_v10_0_late_init(void *handle)
 	if (r)
 		return r;
 
+	r = amdgpu_irq_get(adev, &adev->gfx.bad_op_irq, 0);
+	if (r)
+		return r;
+
 	return 0;
 }
 
@@ -7572,15 +7860,15 @@ static bool gfx_v10_0_is_rlc_enabled(struct amdgpu_device *adev)
 	return (REG_GET_FIELD(rlc_cntl, RLC_CNTL, RLC_ENABLE_F32)) ? true : false;
 }
 
-static void gfx_v10_0_set_safe_mode(struct amdgpu_device *adev)
+static void gfx_v10_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	uint32_t data;
-	unsigned i;
+	unsigned int i;
 
 	data = RLC_SAFE_MODE__CMD_MASK;
 	data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT);
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 0):
 	case IP_VERSION(10, 3, 2):
 	case IP_VERSION(10, 3, 1):
@@ -7613,12 +7901,12 @@ static void gfx_v10_0_set_safe_mode(struct amdgpu_device *adev)
 	}
 }
 
-static void gfx_v10_0_unset_safe_mode(struct amdgpu_device *adev)
+static void gfx_v10_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	uint32_t data;
 
 	data = RLC_SAFE_MODE__CMD_MASK;
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 0):
 	case IP_VERSION(10, 3, 2):
 	case IP_VERSION(10, 3, 1):
@@ -7929,7 +8217,7 @@ static void gfx_v10_0_apply_medium_grain_clock_gating_workaround(struct amdgpu_d
 		mmCGTS_SA1_QUAD1_SM_CTRL_REG
 	};
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 2)) {
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 1, 2)) {
 		for (i = 0; i < ARRAY_SIZE(tcp_ctrl_regs_nv12); i++) {
 			reg_idx = adev->reg_offset[GC_HWIP][0][mmCGTS_SA0_WGP00_CU0_TCP_CTRL_REG_BASE_IDX] +
 				  tcp_ctrl_regs_nv12[i];
@@ -7960,7 +8248,7 @@ static void gfx_v10_0_apply_medium_grain_clock_gating_workaround(struct amdgpu_d
 static int gfx_v10_0_update_gfx_clock_gating(struct amdgpu_device *adev,
 					    bool enable)
 {
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 
 	if (enable) {
 		/* enable FGCG firstly*/
@@ -7974,9 +8262,12 @@ static int gfx_v10_0_update_gfx_clock_gating(struct amdgpu_device *adev,
 		/* ===  CGCG + CGLS === */
 		gfx_v10_0_update_coarse_grain_clock_gating(adev, enable);
 
-		if ((adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 10)) ||
-		    (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 1)) ||
-		    (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 2)))
+		if ((amdgpu_ip_version(adev, GC_HWIP, 0) ==
+		     IP_VERSION(10, 1, 10)) ||
+		    (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+		     IP_VERSION(10, 1, 1)) ||
+		    (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+		     IP_VERSION(10, 1, 2)))
 			gfx_v10_0_apply_medium_grain_clock_gating_workaround(adev);
 	} else {
 		/* CGCG/CGLS should be disabled before MGCG/MGLS
@@ -7999,31 +8290,39 @@ static int gfx_v10_0_update_gfx_clock_gating(struct amdgpu_device *adev,
 	     AMD_CG_SUPPORT_GFX_3D_CGLS))
 		gfx_v10_0_enable_gui_idle_interrupt(adev, enable);
 
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 
 	return 0;
 }
 
-static void gfx_v10_0_update_spm_vmid(struct amdgpu_device *adev, unsigned vmid)
+static void gfx_v10_0_update_spm_vmid_internal(struct amdgpu_device *adev,
+					       unsigned int vmid)
 {
-	u32 reg, data;
-
-	amdgpu_gfx_off_ctrl(adev, false);
+	u32 reg, pre_data, data;
 
-	/* not for *_SOC15 */
 	reg = SOC15_REG_OFFSET(GC, 0, mmRLC_SPM_MC_CNTL);
-	if (amdgpu_sriov_is_pp_one_vf(adev))
-		data = RREG32_NO_KIQ(reg);
+	/* not for *_SOC15 */
+	if (amdgpu_sriov_is_pp_one_vf(adev) && !amdgpu_sriov_runtime(adev))
+		pre_data = RREG32_NO_KIQ(reg);
 	else
-		data = RREG32_SOC15(GC, 0, mmRLC_SPM_MC_CNTL);
+		pre_data = RREG32(reg);
 
-	data &= ~RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK;
+	data = pre_data & (~RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK);
 	data |= (vmid & RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK) << RLC_SPM_MC_CNTL__RLC_SPM_VMID__SHIFT;
 
-	if (amdgpu_sriov_is_pp_one_vf(adev))
-		WREG32_SOC15_NO_KIQ(GC, 0, mmRLC_SPM_MC_CNTL, data);
-	else
-		WREG32_SOC15(GC, 0, mmRLC_SPM_MC_CNTL, data);
+	if (pre_data != data) {
+		if (amdgpu_sriov_is_pp_one_vf(adev) && !amdgpu_sriov_runtime(adev)) {
+			WREG32_SOC15_NO_KIQ(GC, 0, mmRLC_SPM_MC_CNTL, data);
+		} else
+			WREG32_SOC15(GC, 0, mmRLC_SPM_MC_CNTL, data);
+	}
+}
+
+static void gfx_v10_0_update_spm_vmid(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned int vmid)
+{
+	amdgpu_gfx_off_ctrl(adev, false);
+
+	gfx_v10_0_update_spm_vmid_internal(adev, vmid);
 
 	amdgpu_gfx_off_ctrl(adev, true);
 }
@@ -8077,7 +8376,7 @@ static void gfx_v10_cntl_power_gating(struct amdgpu_device *adev, bool enable)
 	 * Power/performance team will optimize it and might give a new value later.
 	 */
 	if (enable && (adev->pg_flags & AMD_PG_SUPPORT_GFX_PG)) {
-		switch (adev->ip_versions[GC_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 		case IP_VERSION(10, 3, 1):
 		case IP_VERSION(10, 3, 3):
 		case IP_VERSION(10, 3, 6):
@@ -8093,11 +8392,11 @@ static void gfx_v10_cntl_power_gating(struct amdgpu_device *adev, bool enable)
 
 static void gfx_v10_cntl_pg(struct amdgpu_device *adev, bool enable)
 {
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 
 	gfx_v10_cntl_power_gating(adev, enable);
 
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 }
 
 static const struct amdgpu_rlc_funcs gfx_v10_0_rlc_funcs = {
@@ -8129,16 +8428,16 @@ static const struct amdgpu_rlc_funcs gfx_v10_0_rlc_funcs_sriov = {
 	.is_rlcg_access_range = gfx_v10_0_is_rlcg_access_range,
 };
 
-static int gfx_v10_0_set_powergating_state(void *handle,
+static int gfx_v10_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_PG_STATE_GATE);
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 	case IP_VERSION(10, 1, 1):
 	case IP_VERSION(10, 1, 2):
@@ -8152,8 +8451,14 @@ static int gfx_v10_0_set_powergating_state(void *handle,
 	case IP_VERSION(10, 3, 3):
 	case IP_VERSION(10, 3, 6):
 	case IP_VERSION(10, 3, 7):
+		if (!enable)
+			amdgpu_gfx_off_ctrl(adev, false);
+
 		gfx_v10_cntl_pg(adev, enable);
-		amdgpu_gfx_off_ctrl(adev, enable);
+
+		if (enable)
+			amdgpu_gfx_off_ctrl(adev, true);
+
 		break;
 	default:
 		break;
@@ -8161,15 +8466,15 @@ static int gfx_v10_0_set_powergating_state(void *handle,
 	return 0;
 }
 
-static int gfx_v10_0_set_clockgating_state(void *handle,
+static int gfx_v10_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 	case IP_VERSION(10, 1, 1):
 	case IP_VERSION(10, 1, 2):
@@ -8190,9 +8495,9 @@ static int gfx_v10_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static void gfx_v10_0_get_clockgating_state(void *handle, u64 *flags)
+static void gfx_v10_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	/* AMD_CG_SUPPORT_GFX_FGCG */
@@ -8259,45 +8564,17 @@ static u64 gfx_v10_0_ring_get_wptr_gfx(struct amdgpu_ring *ring)
 static void gfx_v10_0_ring_set_wptr_gfx(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	uint32_t *wptr_saved;
-	uint32_t *is_queue_unmap;
-	uint64_t aggregated_db_index;
-	uint32_t mqd_size = adev->mqds[AMDGPU_HW_IP_GFX].mqd_size;
-	uint64_t wptr_tmp;
-
-	if (ring->is_mes_queue) {
-		wptr_saved = (uint32_t *)(ring->mqd_ptr + mqd_size);
-		is_queue_unmap = (uint32_t *)(ring->mqd_ptr + mqd_size +
-					      sizeof(uint32_t));
-		aggregated_db_index =
-			amdgpu_mes_get_aggregated_doorbell_index(adev,
-			AMDGPU_MES_PRIORITY_LEVEL_NORMAL);
-
-		wptr_tmp = ring->wptr & ring->buf_mask;
-		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, wptr_tmp);
-		*wptr_saved = wptr_tmp;
-		/* assume doorbell always being used by mes mapped queue */
-		if (*is_queue_unmap) {
-			WDOORBELL64(aggregated_db_index, wptr_tmp);
-			WDOORBELL64(ring->doorbell_index, wptr_tmp);
-		} else {
-			WDOORBELL64(ring->doorbell_index, wptr_tmp);
 
-			if (*is_queue_unmap)
-				WDOORBELL64(aggregated_db_index, wptr_tmp);
-		}
+	if (ring->use_doorbell) {
+		/* XXX check if swapping is necessary on BE */
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
+			     ring->wptr);
+		WDOORBELL64(ring->doorbell_index, ring->wptr);
 	} else {
-		if (ring->use_doorbell) {
-			/* XXX check if swapping is necessary on BE */
-			atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
-				     ring->wptr);
-			WDOORBELL64(ring->doorbell_index, ring->wptr);
-		} else {
-			WREG32_SOC15(GC, 0, mmCP_RB0_WPTR,
-				     lower_32_bits(ring->wptr));
-			WREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI,
-				     upper_32_bits(ring->wptr));
-		}
+		WREG32_SOC15(GC, 0, mmCP_RB0_WPTR,
+			     lower_32_bits(ring->wptr));
+		WREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI,
+			     upper_32_bits(ring->wptr));
 	}
 }
 
@@ -8322,42 +8599,13 @@ static u64 gfx_v10_0_ring_get_wptr_compute(struct amdgpu_ring *ring)
 static void gfx_v10_0_ring_set_wptr_compute(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	uint32_t *wptr_saved;
-	uint32_t *is_queue_unmap;
-	uint64_t aggregated_db_index;
-	uint32_t mqd_size = adev->mqds[AMDGPU_HW_IP_COMPUTE].mqd_size;
-	uint64_t wptr_tmp;
-
-	if (ring->is_mes_queue) {
-		wptr_saved = (uint32_t *)(ring->mqd_ptr + mqd_size);
-		is_queue_unmap = (uint32_t *)(ring->mqd_ptr + mqd_size +
-					      sizeof(uint32_t));
-		aggregated_db_index =
-			amdgpu_mes_get_aggregated_doorbell_index(adev,
-			AMDGPU_MES_PRIORITY_LEVEL_NORMAL);
-
-		wptr_tmp = ring->wptr & ring->buf_mask;
-		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, wptr_tmp);
-		*wptr_saved = wptr_tmp;
-		/* assume doorbell always used by mes mapped queue */
-		if (*is_queue_unmap) {
-			WDOORBELL64(aggregated_db_index, wptr_tmp);
-			WDOORBELL64(ring->doorbell_index, wptr_tmp);
-		} else {
-			WDOORBELL64(ring->doorbell_index, wptr_tmp);
 
-			if (*is_queue_unmap)
-				WDOORBELL64(aggregated_db_index, wptr_tmp);
-		}
+	if (ring->use_doorbell) {
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
+			     ring->wptr);
+		WDOORBELL64(ring->doorbell_index, ring->wptr);
 	} else {
-		/* XXX check if swapping is necessary on BE */
-		if (ring->use_doorbell) {
-			atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
-				     ring->wptr);
-			WDOORBELL64(ring->doorbell_index, ring->wptr);
-		} else {
-			BUG(); /* only DOORBELL method supported on gfx10 now */
-		}
+		BUG(); /* only DOORBELL method supported on gfx10 now */
 	}
 }
 
@@ -8380,7 +8628,7 @@ static void gfx_v10_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
 		}
 		reg_mem_engine = 0;
 	} else {
-		ref_and_mask = nbio_hf_reg->ref_and_mask_cp0;
+		ref_and_mask = nbio_hf_reg->ref_and_mask_cp0 << ring->pipe;
 		reg_mem_engine = 1; /* pfp */
 	}
 
@@ -8395,7 +8643,7 @@ static void gfx_v10_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
 				       struct amdgpu_ib *ib,
 				       uint32_t flags)
 {
-	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
+	unsigned int vmid = AMDGPU_JOB_GET_VMID(job);
 	u32 header, control = 0;
 
 	if (ib->flags & AMDGPU_IB_FLAG_CE)
@@ -8405,7 +8653,7 @@ static void gfx_v10_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
 
 	control |= ib->length_dw | (vmid << 24);
 
-	if (amdgpu_mcbp && (ib->flags & AMDGPU_IB_FLAG_PREEMPT)) {
+	if (ring->adev->gfx.mcbp && (ib->flags & AMDGPU_IB_FLAG_PREEMPT)) {
 		control |= INDIRECT_BUFFER_PRE_ENB(1);
 
 		if (flags & AMDGPU_IB_PREEMPTED)
@@ -8416,10 +8664,6 @@ static void gfx_v10_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
 				    (!amdgpu_sriov_vf(ring->adev) && flags & AMDGPU_IB_PREEMPTED) ? true : false);
 	}
 
-	if (ring->is_mes_queue)
-		/* inherit vmid from mqd */
-		control |= 0x400000;
-
 	amdgpu_ring_write(ring, header);
 	BUG_ON(ib->gpu_addr & 0x3); /* Dword align */
 	amdgpu_ring_write(ring,
@@ -8436,13 +8680,9 @@ static void gfx_v10_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
 					   struct amdgpu_ib *ib,
 					   uint32_t flags)
 {
-	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
+	unsigned int vmid = AMDGPU_JOB_GET_VMID(job);
 	u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vmid << 24);
 
-	if (ring->is_mes_queue)
-		/* inherit vmid from mqd */
-		control |= 0x40000000;
-
 	/* Currently, there is a high possibility to get wave ID mismatch
 	 * between ME and GDS, leading to a hw deadlock, because ME generates
 	 * different wave IDs than the GDS expects. This situation happens
@@ -8471,7 +8711,7 @@ static void gfx_v10_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
 }
 
 static void gfx_v10_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr,
-				     u64 seq, unsigned flags)
+				     u64 seq, unsigned int flags)
 {
 	bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT;
 	bool int_sel = flags & AMDGPU_FENCE_FLAG_INT;
@@ -8500,8 +8740,7 @@ static void gfx_v10_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr,
 	amdgpu_ring_write(ring, upper_32_bits(addr));
 	amdgpu_ring_write(ring, lower_32_bits(seq));
 	amdgpu_ring_write(ring, upper_32_bits(seq));
-	amdgpu_ring_write(ring, ring->is_mes_queue ?
-			 (ring->hw_queue_id | AMDGPU_FENCE_MES_QUEUE_FLAG) : 0);
+	amdgpu_ring_write(ring, 0);
 }
 
 static void gfx_v10_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
@@ -8527,12 +8766,9 @@ static void gfx_v10_0_ring_invalidate_tlbs(struct amdgpu_ring *ring,
 }
 
 static void gfx_v10_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
-					 unsigned vmid, uint64_t pd_addr)
+					 unsigned int vmid, uint64_t pd_addr)
 {
-	if (ring->is_mes_queue)
-		gfx_v10_0_ring_invalidate_tlbs(ring, 0, 0, false, 0);
-	else
-		amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
+	amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
 
 	/* compute doesn't have PFP */
 	if (ring->funcs->type == AMDGPU_RING_TYPE_GFX) {
@@ -8580,7 +8816,7 @@ static void gfx_v10_0_ring_emit_cntxcntl(struct amdgpu_ring *ring,
 {
 	uint32_t dw2 = 0;
 
-	if (amdgpu_mcbp)
+	if (ring->adev->gfx.mcbp)
 		gfx_v10_0_ring_emit_ce_meta(ring,
 				    (!amdgpu_sriov_vf(ring->adev) && flags & AMDGPU_IB_PREEMPTED) ? true : false);
 
@@ -8609,38 +8845,28 @@ static void gfx_v10_0_ring_emit_cntxcntl(struct amdgpu_ring *ring,
 	amdgpu_ring_write(ring, 0);
 }
 
-static unsigned gfx_v10_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring)
+static unsigned int gfx_v10_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring,
+						       uint64_t addr)
 {
-	unsigned ret;
+	unsigned int ret;
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, 0); /* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	/* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, 0);
 	ret = ring->wptr & ring->buf_mask;
-	amdgpu_ring_write(ring, 0x55aa55aa); /* patch dummy value later */
+	/* patch dummy value later */
+	amdgpu_ring_write(ring, 0);
 
 	return ret;
 }
 
-static void gfx_v10_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigned offset)
-{
-	unsigned cur;
-	BUG_ON(offset > ring->buf_mask);
-	BUG_ON(ring->ring[offset] != 0x55aa55aa);
-
-	cur = (ring->wptr - 1) & ring->buf_mask;
-	if (likely(cur > offset))
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->buf_mask + 1) - offset + cur;
-}
-
 static int gfx_v10_0_ring_preempt_ib(struct amdgpu_ring *ring)
 {
 	int i, r = 0;
 	struct amdgpu_device *adev = ring->adev;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
 	struct amdgpu_ring *kiq_ring = &kiq->ring;
 	unsigned long flags;
 
@@ -8693,19 +8919,9 @@ static void gfx_v10_0_ring_emit_ce_meta(struct amdgpu_ring *ring, bool resume)
 
 	cnt = (sizeof(ce_payload) >> 2) + 4 - 2;
 
-	if (ring->is_mes_queue) {
-		offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				  gfx[0].gfx_meta_data) +
-			offsetof(struct v10_gfx_meta_data, ce_payload);
-		ce_payload_gpu_addr =
-			amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		ce_payload_cpu_addr =
-			amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-	} else {
-		offset = offsetof(struct v10_gfx_meta_data, ce_payload);
-		ce_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
-		ce_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
-	}
+	offset = offsetof(struct v10_gfx_meta_data, ce_payload);
+	ce_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
+	ce_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, cnt));
 	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(2) |
@@ -8731,28 +8947,13 @@ static void gfx_v10_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume)
 	void *de_payload_cpu_addr;
 	int cnt;
 
-	if (ring->is_mes_queue) {
-		offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				  gfx[0].gfx_meta_data) +
-			offsetof(struct v10_gfx_meta_data, de_payload);
-		de_payload_gpu_addr =
-			amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		de_payload_cpu_addr =
-			amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-
-		offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				  gfx[0].gds_backup) +
-			offsetof(struct v10_gfx_meta_data, de_payload);
-		gds_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-	} else {
-		offset = offsetof(struct v10_gfx_meta_data, de_payload);
-		de_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
-		de_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
+	offset = offsetof(struct v10_gfx_meta_data, de_payload);
+	de_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
+	de_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
 
-		gds_addr = ALIGN(amdgpu_csa_vaddr(ring->adev) +
-				 AMDGPU_CSA_SIZE - adev->gds.gds_size,
-				 PAGE_SIZE);
-	}
+	gds_addr = ALIGN(amdgpu_csa_vaddr(ring->adev) +
+			 AMDGPU_CSA_SIZE - adev->gds.gds_size,
+			 PAGE_SIZE);
 
 	de_payload.gds_backup_addrlo = lower_32_bits(gds_addr);
 	de_payload.gds_backup_addrhi = upper_32_bits(gds_addr);
@@ -8847,19 +9048,6 @@ static void gfx_v10_0_ring_emit_reg_write_reg_wait(struct amdgpu_ring *ring,
 							   ref, mask);
 }
 
-static void gfx_v10_0_ring_soft_recovery(struct amdgpu_ring *ring,
-					 unsigned vmid)
-{
-	struct amdgpu_device *adev = ring->adev;
-	uint32_t value = 0;
-
-	value = REG_SET_FIELD(value, SQ_CMD, CMD, 0x03);
-	value = REG_SET_FIELD(value, SQ_CMD, MODE, 0x01);
-	value = REG_SET_FIELD(value, SQ_CMD, CHECK_VMID, 1);
-	value = REG_SET_FIELD(value, SQ_CMD, VM_ID, vmid);
-	WREG32_SOC15(GC, 0, mmSQ_CMD, value);
-}
-
 static void
 gfx_v10_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev,
 				      uint32_t me, uint32_t pipe,
@@ -8957,7 +9145,7 @@ static void gfx_v10_0_set_compute_eop_interrupt_state(struct amdgpu_device *adev
 
 static int gfx_v10_0_set_eop_interrupt_state(struct amdgpu_device *adev,
 					    struct amdgpu_irq_src *src,
-					    unsigned type,
+					    unsigned int type,
 					    enum amdgpu_interrupt_state state)
 {
 	switch (type) {
@@ -9004,49 +9192,34 @@ static int gfx_v10_0_eop_irq(struct amdgpu_device *adev,
 	int i;
 	u8 me_id, pipe_id, queue_id;
 	struct amdgpu_ring *ring;
-	uint32_t mes_queue_id = entry->src_data[0];
 
 	DRM_DEBUG("IH: CP EOP\n");
 
-	if (adev->enable_mes && (mes_queue_id & AMDGPU_FENCE_MES_QUEUE_FLAG)) {
-		struct amdgpu_mes_queue *queue;
-
-		mes_queue_id &= AMDGPU_FENCE_MES_QUEUE_ID_MASK;
-
-		spin_lock(&adev->mes.queue_id_lock);
-		queue = idr_find(&adev->mes.queue_id_idr, mes_queue_id);
-		if (queue) {
-			DRM_DEBUG("process mes queue id = %d\n", mes_queue_id);
-			amdgpu_fence_process(queue->ring);
-		}
-		spin_unlock(&adev->mes.queue_id_lock);
-	} else {
-		me_id = (entry->ring_id & 0x0c) >> 2;
-		pipe_id = (entry->ring_id & 0x03) >> 0;
-		queue_id = (entry->ring_id & 0x70) >> 4;
+	me_id = (entry->ring_id & 0x0c) >> 2;
+	pipe_id = (entry->ring_id & 0x03) >> 0;
+	queue_id = (entry->ring_id & 0x70) >> 4;
 
-		switch (me_id) {
-		case 0:
-			if (pipe_id == 0)
-				amdgpu_fence_process(&adev->gfx.gfx_ring[0]);
-			else
-				amdgpu_fence_process(&adev->gfx.gfx_ring[1]);
-			break;
-		case 1:
-		case 2:
-			for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-				ring = &adev->gfx.compute_ring[i];
-				/* Per-queue interrupt is supported for MEC starting from VI.
-				 * The interrupt can only be enabled/disabled per pipe instead
-				 * of per queue.
-				 */
-				if ((ring->me == me_id) &&
-				    (ring->pipe == pipe_id) &&
-				    (ring->queue == queue_id))
-					amdgpu_fence_process(ring);
-			}
-			break;
+	switch (me_id) {
+	case 0:
+		if (pipe_id == 0)
+			amdgpu_fence_process(&adev->gfx.gfx_ring[0]);
+		else
+			amdgpu_fence_process(&adev->gfx.gfx_ring[1]);
+		break;
+	case 1:
+	case 2:
+		for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+			ring = &adev->gfx.compute_ring[i];
+			/* Per-queue interrupt is supported for MEC starting from VI.
+			 * The interrupt can only be enabled/disabled per pipe instead
+			 * of per queue.
+			 */
+			if ((ring->me == me_id) &&
+			    (ring->pipe == pipe_id) &&
+			    (ring->queue == queue_id))
+				amdgpu_fence_process(ring);
 		}
+		break;
 	}
 
 	return 0;
@@ -9054,15 +9227,42 @@ static int gfx_v10_0_eop_irq(struct amdgpu_device *adev,
 
 static int gfx_v10_0_set_priv_reg_fault_state(struct amdgpu_device *adev,
 					      struct amdgpu_irq_src *source,
-					      unsigned type,
+					      unsigned int type,
 					      enum amdgpu_interrupt_state state)
 {
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
 	case AMDGPU_IRQ_STATE_ENABLE:
-		WREG32_FIELD15(GC, 0, CP_INT_CNTL_RING0,
-			       PRIV_REG_INT_ENABLE,
-			       state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+				cp_int_cntl_reg = gfx_v10_0_get_cpg_int_cntl(adev, i, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+								    PRIV_REG_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				/* MECs start at 1 */
+				cp_int_cntl_reg = gfx_v10_0_get_cpc_int_cntl(adev, i + 1, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+								    PRIV_REG_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
 		break;
 	default:
 		break;
@@ -9071,17 +9271,75 @@ static int gfx_v10_0_set_priv_reg_fault_state(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int gfx_v10_0_set_bad_op_fault_state(struct amdgpu_device *adev,
+					    struct amdgpu_irq_src *source,
+					    unsigned type,
+					    enum amdgpu_interrupt_state state)
+{
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+	case AMDGPU_IRQ_STATE_ENABLE:
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+				cp_int_cntl_reg = gfx_v10_0_get_cpg_int_cntl(adev, i, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+								    OPCODE_ERROR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				/* MECs start at 1 */
+				cp_int_cntl_reg = gfx_v10_0_get_cpc_int_cntl(adev, i + 1, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+								    OPCODE_ERROR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		break;
+	default:
+		break;
+	}
+	return 0;
+}
+
 static int gfx_v10_0_set_priv_inst_fault_state(struct amdgpu_device *adev,
 					       struct amdgpu_irq_src *source,
-					       unsigned type,
+					       unsigned int type,
 					       enum amdgpu_interrupt_state state)
 {
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
 	case AMDGPU_IRQ_STATE_ENABLE:
-		WREG32_FIELD15(GC, 0, CP_INT_CNTL_RING0,
-			       PRIV_INSTR_INT_ENABLE,
-			       state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+				cp_int_cntl_reg = gfx_v10_0_get_cpg_int_cntl(adev, i, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+								    PRIV_INSTR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
 		break;
 	default:
 		break;
@@ -9105,8 +9363,8 @@ static void gfx_v10_0_handle_priv_fault(struct amdgpu_device *adev,
 	case 0:
 		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
 			ring = &adev->gfx.gfx_ring[i];
-			/* we only enabled 1 gfx queue per pipe for now */
-			if (ring->me == me_id && ring->pipe == pipe_id)
+			if (ring->me == me_id && ring->pipe == pipe_id &&
+			    ring->queue == queue_id)
 				drm_sched_fault(&ring->sched);
 		}
 		break;
@@ -9133,6 +9391,15 @@ static int gfx_v10_0_priv_reg_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int gfx_v10_0_bad_op_irq(struct amdgpu_device *adev,
+				struct amdgpu_irq_src *source,
+				struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal opcode in command stream \n");
+	gfx_v10_0_handle_priv_fault(adev, entry);
+	return 0;
+}
+
 static int gfx_v10_0_priv_inst_irq(struct amdgpu_device *adev,
 				   struct amdgpu_irq_src *source,
 				   struct amdgpu_iv_entry *entry)
@@ -9148,7 +9415,7 @@ static int gfx_v10_0_kiq_set_interrupt_state(struct amdgpu_device *adev,
 					     enum amdgpu_interrupt_state state)
 {
 	uint32_t tmp, target;
-	struct amdgpu_ring *ring = &(adev->gfx.kiq.ring);
+	struct amdgpu_ring *ring = &(adev->gfx.kiq[0].ring);
 
 	if (ring->me == 1)
 		target = SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE0_INT_CNTL);
@@ -9192,7 +9459,7 @@ static int gfx_v10_0_kiq_irq(struct amdgpu_device *adev,
 			     struct amdgpu_iv_entry *entry)
 {
 	u8 me_id, pipe_id, queue_id;
-	struct amdgpu_ring *ring = &(adev->gfx.kiq.ring);
+	struct amdgpu_ring *ring = &(adev->gfx.kiq[0].ring);
 
 	me_id = (entry->ring_id & 0x0c) >> 2;
 	pipe_id = (entry->ring_id & 0x03) >> 0;
@@ -9227,6 +9494,319 @@ static void gfx_v10_0_emit_mem_sync(struct amdgpu_ring *ring)
 	amdgpu_ring_write(ring, gcr_cntl); /* GCR_CNTL */
 }
 
+static void gfx_v10_ring_insert_nop(struct amdgpu_ring *ring, uint32_t num_nop)
+{
+	/* Header itself is a NOP packet */
+	if (num_nop == 1) {
+		amdgpu_ring_write(ring, ring->funcs->nop);
+		return;
+	}
+
+	/* Max HW optimization till 0x3ffe, followed by remaining one NOP at a time*/
+	amdgpu_ring_write(ring, PACKET3(PACKET3_NOP, min(num_nop - 2, 0x3ffe)));
+
+	/* Header is at index 0, followed by num_nops - 1 NOP packet's */
+	amdgpu_ring_insert_nop(ring, num_nop - 1);
+}
+
+static int gfx_v10_0_reset_kgq(struct amdgpu_ring *ring,
+			       unsigned int vmid,
+			       struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
+	unsigned long flags;
+	u32 tmp;
+	u64 addr;
+	int r;
+
+	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
+		return -EINVAL;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+
+	if (amdgpu_ring_alloc(kiq_ring, 5 + 7 + 7)) {
+		spin_unlock_irqrestore(&kiq->ring_lock, flags);
+		return -ENOMEM;
+	}
+
+	addr = amdgpu_bo_gpu_offset(ring->mqd_obj) +
+		offsetof(struct v10_gfx_mqd, cp_gfx_hqd_active);
+	tmp = REG_SET_FIELD(0, CP_VMID_RESET, RESET_REQUEST, 1 << vmid);
+	if (ring->pipe == 0)
+		tmp = REG_SET_FIELD(tmp, CP_VMID_RESET, PIPE0_QUEUES, 1 << ring->queue);
+	else
+		tmp = REG_SET_FIELD(tmp, CP_VMID_RESET, PIPE1_QUEUES, 1 << ring->queue);
+
+	gfx_v10_0_ring_emit_wreg(kiq_ring,
+				 SOC15_REG_OFFSET(GC, 0, mmCP_VMID_RESET), tmp);
+	gfx_v10_0_wait_reg_mem(kiq_ring, 0, 1, 0,
+			       lower_32_bits(addr), upper_32_bits(addr),
+			       0, 1, 0x20);
+	gfx_v10_0_ring_emit_reg_wait(kiq_ring,
+				     SOC15_REG_OFFSET(GC, 0, mmCP_VMID_RESET), 0, 0xffffffff);
+	amdgpu_ring_commit(kiq_ring);
+	r = amdgpu_ring_test_ring(kiq_ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+	if (r)
+		return r;
+
+	r = gfx_v10_0_kgq_init_queue(ring, true);
+	if (r) {
+		DRM_ERROR("fail to init kgq\n");
+		return r;
+	}
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+
+	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size)) {
+		spin_unlock_irqrestore(&kiq->ring_lock, flags);
+		return -ENOMEM;
+	}
+	kiq->pmf->kiq_map_queues(kiq_ring, ring);
+	amdgpu_ring_commit(kiq_ring);
+	r = amdgpu_ring_test_ring(kiq_ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+	if (r)
+		return r;
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static int gfx_v10_0_reset_kcq(struct amdgpu_ring *ring,
+			       unsigned int vmid,
+			       struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
+	unsigned long flags;
+	int i, r;
+
+	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
+		return -EINVAL;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+
+	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size)) {
+		spin_unlock_irqrestore(&kiq->ring_lock, flags);
+		return -ENOMEM;
+	}
+
+	kiq->pmf->kiq_unmap_queues(kiq_ring, ring, RESET_QUEUES,
+				   0, 0);
+	amdgpu_ring_commit(kiq_ring);
+	r = amdgpu_ring_test_ring(kiq_ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+	if (r)
+		return r;
+
+	/* make sure dequeue is complete*/
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
+	mutex_lock(&adev->srbm_mutex);
+	nv_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (!(RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE) & 1))
+			break;
+		udelay(1);
+	}
+	if (i >= adev->usec_timeout)
+		r = -ETIMEDOUT;
+	nv_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
+	if (r) {
+		dev_err(adev->dev, "fail to wait on hqd deactivate\n");
+		return r;
+	}
+
+	r = gfx_v10_0_kcq_init_queue(ring, true);
+	if (r) {
+		dev_err(adev->dev, "fail to init kcq\n");
+		return r;
+	}
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size)) {
+		spin_unlock_irqrestore(&kiq->ring_lock, flags);
+		return -ENOMEM;
+	}
+	kiq->pmf->kiq_map_queues(kiq_ring, ring);
+	amdgpu_ring_commit(kiq_ring);
+	r = amdgpu_ring_test_ring(kiq_ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+	if (r)
+		return r;
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static void gfx_v10_ip_print(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k, reg, index = 0;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_10_1);
+
+	if (!adev->gfx.ip_dump_core)
+		return;
+
+	for (i = 0; i < reg_count; i++)
+		drm_printf(p, "%-50s \t 0x%08x\n",
+			   gc_reg_list_10_1[i].reg_name,
+			   adev->gfx.ip_dump_core[i]);
+
+	/* print compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_10);
+	drm_printf(p, "\nnum_mec: %d num_pipe: %d num_queue: %d\n",
+		   adev->gfx.mec.num_mec,
+		   adev->gfx.mec.num_pipe_per_mec,
+		   adev->gfx.mec.num_queue_per_pipe);
+
+	for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+		for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+			for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+				drm_printf(p, "\nmec %d, pipe %d, queue %d\n", i, j, k);
+				for (reg = 0; reg < reg_count; reg++) {
+					if (i && gc_cp_reg_list_10[reg].reg_offset == mmCP_MEC_ME1_HEADER_DUMP)
+						drm_printf(p, "%-50s \t 0x%08x\n",
+							   "mmCP_MEC_ME2_HEADER_DUMP",
+							   adev->gfx.ip_dump_compute_queues[index + reg]);
+					else
+						drm_printf(p, "%-50s \t 0x%08x\n",
+							   gc_cp_reg_list_10[reg].reg_name,
+							   adev->gfx.ip_dump_compute_queues[index + reg]);
+				}
+				index += reg_count;
+			}
+		}
+	}
+
+	/* print gfx queue registers for all instances */
+	if (!adev->gfx.ip_dump_gfx_queues)
+		return;
+
+	index = 0;
+	reg_count = ARRAY_SIZE(gc_gfx_queue_reg_list_10);
+	drm_printf(p, "\nnum_me: %d num_pipe: %d num_queue: %d\n",
+		   adev->gfx.me.num_me,
+		   adev->gfx.me.num_pipe_per_me,
+		   adev->gfx.me.num_queue_per_pipe);
+
+	for (i = 0; i < adev->gfx.me.num_me; i++) {
+		for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+			for (k = 0; k < adev->gfx.me.num_queue_per_pipe; k++) {
+				drm_printf(p, "\nme %d, pipe %d, queue %d\n", i, j, k);
+				for (reg = 0; reg < reg_count; reg++) {
+					drm_printf(p, "%-50s \t 0x%08x\n",
+						   gc_gfx_queue_reg_list_10[reg].reg_name,
+						   adev->gfx.ip_dump_gfx_queues[index + reg]);
+				}
+				index += reg_count;
+			}
+		}
+	}
+}
+
+static void gfx_v10_ip_dump(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k, reg, index = 0;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_10_1);
+
+	if (!adev->gfx.ip_dump_core)
+		return;
+
+	amdgpu_gfx_off_ctrl(adev, false);
+	for (i = 0; i < reg_count; i++)
+		adev->gfx.ip_dump_core[i] = RREG32(SOC15_REG_ENTRY_OFFSET(gc_reg_list_10_1[i]));
+	amdgpu_gfx_off_ctrl(adev, true);
+
+	/* dump compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_10);
+	amdgpu_gfx_off_ctrl(adev, false);
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+		for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+			for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+				/* ME0 is for GFX so start from 1 for CP */
+				nv_grbm_select(adev, adev->gfx.me.num_me + i, j, k, 0);
+
+				for (reg = 0; reg < reg_count; reg++) {
+					if (i && gc_cp_reg_list_10[reg].reg_offset == mmCP_MEC_ME1_HEADER_DUMP)
+						adev->gfx.ip_dump_compute_queues[index + reg] =
+							RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEC_ME2_HEADER_DUMP));
+					else
+						adev->gfx.ip_dump_compute_queues[index + reg] =
+							RREG32(SOC15_REG_ENTRY_OFFSET(
+								       gc_cp_reg_list_10[reg]));
+				}
+				index += reg_count;
+			}
+		}
+	}
+	nv_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	amdgpu_gfx_off_ctrl(adev, true);
+
+	/* dump gfx queue registers for all instances */
+	if (!adev->gfx.ip_dump_gfx_queues)
+		return;
+
+	index = 0;
+	reg_count = ARRAY_SIZE(gc_gfx_queue_reg_list_10);
+	amdgpu_gfx_off_ctrl(adev, false);
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->gfx.me.num_me; i++) {
+		for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+			for (k = 0; k < adev->gfx.me.num_queue_per_pipe; k++) {
+				nv_grbm_select(adev, i, j, k, 0);
+
+				for (reg = 0; reg < reg_count; reg++) {
+					adev->gfx.ip_dump_gfx_queues[index + reg] =
+						RREG32(SOC15_REG_ENTRY_OFFSET(
+							gc_gfx_queue_reg_list_10[reg]));
+				}
+				index += reg_count;
+			}
+		}
+	}
+	nv_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	amdgpu_gfx_off_ctrl(adev, true);
+}
+
+static void gfx_v10_0_ring_emit_cleaner_shader(struct amdgpu_ring *ring)
+{
+	/* Emit the cleaner shader */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_RUN_CLEANER_SHADER, 0));
+	amdgpu_ring_write(ring, 0);  /* RESERVED field, programmed to zero */
+}
+
+static void gfx_v10_0_ring_begin_use(struct amdgpu_ring *ring)
+{
+	amdgpu_gfx_profile_ring_begin_use(ring);
+
+	amdgpu_gfx_enforce_isolation_ring_begin_use(ring);
+}
+
+static void gfx_v10_0_ring_end_use(struct amdgpu_ring *ring)
+{
+	amdgpu_gfx_profile_ring_end_use(ring);
+
+	amdgpu_gfx_enforce_isolation_ring_end_use(ring);
+}
+
 static const struct amd_ip_funcs gfx_v10_0_ip_funcs = {
 	.name = "gfx_v10_0",
 	.early_init = gfx_v10_0_early_init,
@@ -9243,6 +9823,8 @@ static const struct amd_ip_funcs gfx_v10_0_ip_funcs = {
 	.set_clockgating_state = gfx_v10_0_set_clockgating_state,
 	.set_powergating_state = gfx_v10_0_set_powergating_state,
 	.get_clockgating_state = gfx_v10_0_get_clockgating_state,
+	.dump_ip_state = gfx_v10_ip_dump,
+	.print_ip_state = gfx_v10_ip_print,
 };
 
 static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {
@@ -9259,7 +9841,7 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {
 		7 + /* PIPELINE_SYNC */
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
-		2 + /* VM_FLUSH */
+		4 + /* VM_FLUSH */
 		8 + /* FENCE for VM_FLUSH */
 		20 + /* GDS switch */
 		4 + /* double SWITCH_BUFFER,
@@ -9275,7 +9857,8 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {
 		5 + /* HDP_INVL */
 		8 + 8 + /* FENCE x2 */
 		2 + /* SWITCH_BUFFER */
-		8, /* gfx_v10_0_emit_mem_sync */
+		8 + /* gfx_v10_0_emit_mem_sync */
+		2, /* gfx_v10_0_ring_emit_cleaner_shader */
 	.emit_ib_size =	4, /* gfx_v10_0_ring_emit_ib_gfx */
 	.emit_ib = gfx_v10_0_ring_emit_ib_gfx,
 	.emit_fence = gfx_v10_0_ring_emit_fence,
@@ -9285,19 +9868,21 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {
 	.emit_hdp_flush = gfx_v10_0_ring_emit_hdp_flush,
 	.test_ring = gfx_v10_0_ring_test_ring,
 	.test_ib = gfx_v10_0_ring_test_ib,
-	.insert_nop = amdgpu_ring_insert_nop,
+	.insert_nop = gfx_v10_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_switch_buffer = gfx_v10_0_ring_emit_sb,
 	.emit_cntxcntl = gfx_v10_0_ring_emit_cntxcntl,
 	.init_cond_exec = gfx_v10_0_ring_emit_init_cond_exec,
-	.patch_cond_exec = gfx_v10_0_ring_emit_patch_cond_exec,
 	.preempt_ib = gfx_v10_0_ring_preempt_ib,
 	.emit_frame_cntl = gfx_v10_0_ring_emit_frame_cntl,
 	.emit_wreg = gfx_v10_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v10_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = gfx_v10_0_ring_emit_reg_write_reg_wait,
-	.soft_recovery = gfx_v10_0_ring_soft_recovery,
 	.emit_mem_sync = gfx_v10_0_emit_mem_sync,
+	.reset = gfx_v10_0_reset_kgq,
+	.emit_cleaner_shader = gfx_v10_0_ring_emit_cleaner_shader,
+	.begin_use = gfx_v10_0_ring_begin_use,
+	.end_use = gfx_v10_0_ring_end_use,
 };
 
 static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_compute = {
@@ -9317,7 +9902,8 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_compute = {
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
 		2 + /* gfx_v10_0_ring_emit_vm_flush */
 		8 + 8 + 8 + /* gfx_v10_0_ring_emit_fence x3 for user fence, vm fence */
-		8, /* gfx_v10_0_emit_mem_sync */
+		8 + /* gfx_v10_0_emit_mem_sync */
+		2, /* gfx_v10_0_ring_emit_cleaner_shader */
 	.emit_ib_size =	7, /* gfx_v10_0_ring_emit_ib_compute */
 	.emit_ib = gfx_v10_0_ring_emit_ib_compute,
 	.emit_fence = gfx_v10_0_ring_emit_fence,
@@ -9327,12 +9913,16 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_compute = {
 	.emit_hdp_flush = gfx_v10_0_ring_emit_hdp_flush,
 	.test_ring = gfx_v10_0_ring_test_ring,
 	.test_ib = gfx_v10_0_ring_test_ib,
-	.insert_nop = amdgpu_ring_insert_nop,
+	.insert_nop = gfx_v10_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_wreg = gfx_v10_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v10_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = gfx_v10_0_ring_emit_reg_write_reg_wait,
 	.emit_mem_sync = gfx_v10_0_emit_mem_sync,
+	.reset = gfx_v10_0_reset_kcq,
+	.emit_cleaner_shader = gfx_v10_0_ring_emit_cleaner_shader,
+	.begin_use = gfx_v10_0_ring_begin_use,
+	.end_use = gfx_v10_0_ring_end_use,
 };
 
 static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_kiq = {
@@ -9350,7 +9940,6 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_kiq = {
 		7 + /* gfx_v10_0_ring_emit_pipeline_sync */
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
-		2 + /* gfx_v10_0_ring_emit_vm_flush */
 		8 + 8 + 8, /* gfx_v10_0_ring_emit_fence_kiq x3 for user fence, vm fence */
 	.emit_ib_size =	7, /* gfx_v10_0_ring_emit_ib_compute */
 	.emit_ib = gfx_v10_0_ring_emit_ib_compute,
@@ -9363,13 +9952,14 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_kiq = {
 	.emit_wreg = gfx_v10_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v10_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = gfx_v10_0_ring_emit_reg_write_reg_wait,
+	.emit_hdp_flush = gfx_v10_0_ring_emit_hdp_flush,
 };
 
 static void gfx_v10_0_set_ring_funcs(struct amdgpu_device *adev)
 {
 	int i;
 
-	adev->gfx.kiq.ring.funcs = &gfx_v10_0_ring_funcs_kiq;
+	adev->gfx.kiq[0].ring.funcs = &gfx_v10_0_ring_funcs_kiq;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
 		adev->gfx.gfx_ring[i].funcs = &gfx_v10_0_ring_funcs_gfx;
@@ -9388,6 +9978,11 @@ static const struct amdgpu_irq_src_funcs gfx_v10_0_priv_reg_irq_funcs = {
 	.process = gfx_v10_0_priv_reg_irq,
 };
 
+static const struct amdgpu_irq_src_funcs gfx_v10_0_bad_op_irq_funcs = {
+	.set = gfx_v10_0_set_bad_op_fault_state,
+	.process = gfx_v10_0_bad_op_irq,
+};
+
 static const struct amdgpu_irq_src_funcs gfx_v10_0_priv_inst_irq_funcs = {
 	.set = gfx_v10_0_set_priv_inst_fault_state,
 	.process = gfx_v10_0_priv_inst_irq,
@@ -9403,19 +9998,22 @@ static void gfx_v10_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->gfx.eop_irq.num_types = AMDGPU_CP_IRQ_LAST;
 	adev->gfx.eop_irq.funcs = &gfx_v10_0_eop_irq_funcs;
 
-	adev->gfx.kiq.irq.num_types = AMDGPU_CP_KIQ_IRQ_LAST;
-	adev->gfx.kiq.irq.funcs = &gfx_v10_0_kiq_irq_funcs;
+	adev->gfx.kiq[0].irq.num_types = AMDGPU_CP_KIQ_IRQ_LAST;
+	adev->gfx.kiq[0].irq.funcs = &gfx_v10_0_kiq_irq_funcs;
 
 	adev->gfx.priv_reg_irq.num_types = 1;
 	adev->gfx.priv_reg_irq.funcs = &gfx_v10_0_priv_reg_irq_funcs;
 
+	adev->gfx.bad_op_irq.num_types = 1;
+	adev->gfx.bad_op_irq.funcs = &gfx_v10_0_bad_op_irq_funcs;
+
 	adev->gfx.priv_inst_irq.num_types = 1;
 	adev->gfx.priv_inst_irq.funcs = &gfx_v10_0_priv_inst_irq_funcs;
 }
 
 static void gfx_v10_0_set_rlc_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 	case IP_VERSION(10, 1, 1):
 	case IP_VERSION(10, 1, 3):
@@ -9440,7 +10038,7 @@ static void gfx_v10_0_set_rlc_funcs(struct amdgpu_device *adev)
 
 static void gfx_v10_0_set_gds_init(struct amdgpu_device *adev)
 {
-	unsigned total_cu = adev->gfx.config.max_cu_per_sh *
+	unsigned int total_cu = adev->gfx.config.max_cu_per_sh *
 			    adev->gfx.config.max_sh_per_se *
 			    adev->gfx.config.max_shader_engines;
 
@@ -9521,7 +10119,7 @@ static int gfx_v10_0_get_cu_info(struct amdgpu_device *adev,
 {
 	int i, j, k, counter, active_cu_number = 0;
 	u32 mask, bitmap, ao_bitmap, ao_cu_mask = 0;
-	unsigned disable_masks[4 * 2];
+	unsigned int disable_masks[4 * 2];
 
 	if (!adev || !cu_info)
 		return -EINVAL;
@@ -9532,21 +10130,25 @@ static int gfx_v10_0_get_cu_info(struct amdgpu_device *adev,
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
 			bitmap = i * adev->gfx.config.max_sh_per_se + j;
-			if (((adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 0)) ||
-			     (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 3)) ||
-			     (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 6)) ||
-			     (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 7))) &&
+			if (((amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			      IP_VERSION(10, 3, 0)) ||
+			     (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			      IP_VERSION(10, 3, 3)) ||
+			     (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			      IP_VERSION(10, 3, 6)) ||
+			     (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			      IP_VERSION(10, 3, 7))) &&
 			    ((gfx_v10_3_get_disabled_sa(adev) >> bitmap) & 1))
 				continue;
 			mask = 1;
 			ao_bitmap = 0;
 			counter = 0;
-			gfx_v10_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v10_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			if (i < 4 && j < 2)
 				gfx_v10_0_set_user_wgp_inactive_bitmap_per_sh(
 					adev, disable_masks[i * 2 + j]);
 			bitmap = gfx_v10_0_get_cu_active_bitmap_per_sh(adev);
-			cu_info->bitmap[i][j] = bitmap;
+			cu_info->bitmap[0][i][j] = bitmap;
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k++) {
 				if (bitmap & mask) {
@@ -9562,7 +10164,7 @@ static int gfx_v10_0_get_cu_info(struct amdgpu_device *adev,
 			cu_info->ao_cu_bitmap[i][j] = ao_bitmap;
 		}
 	}
-	gfx_v10_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v10_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	cu_info->number = active_cu_number;
@@ -9638,8 +10240,7 @@ static void gfx_v10_3_set_power_brake_sequence(struct amdgpu_device *adev)
 		     (0x1 << DIDT_SQ_THROTTLE_CTRL__PWRBRK_STALL_EN__SHIFT));
 }
 
-const struct amdgpu_ip_block_version gfx_v10_0_ip_block =
-{
+const struct amdgpu_ip_block_version gfx_v10_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GFX,
 	.major = 10,
 	.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0_cleaner_shader.h b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0_cleaner_shader.h
new file mode 100644
index 000000000000..f67569ccf9f6
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0_cleaner_shader.h
@@ -0,0 +1,91 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/* Define the cleaner shader gfx_10_1_10 */
+static const u32 gfx_10_1_10_cleaner_shader_hex[] = {
+	0xb0804004, 0xbf8a0000,
+	0xbf068100, 0xbf840023,
+	0xbe8203b8, 0xbefc0380,
+	0x7e008480, 0x7e028480,
+	0x7e048480, 0x7e068480,
+	0x7e088480, 0x7e0a8480,
+	0x7e0c8480, 0x7e0e8480,
+	0xbefc0302, 0x80828802,
+	0xbf84fff5, 0xbe8203ff,
+	0x80000000, 0x87020102,
+	0xbf840012, 0xbefe03c1,
+	0xbeff03c1, 0xd7650001,
+	0x0001007f, 0xd7660001,
+	0x0002027e, 0x16020288,
+	0xbe8203bf, 0xbefc03c1,
+	0xd9382000, 0x00020201,
+	0xd9386040, 0x00040401,
+	0xd70f6a01, 0x000202ff,
+	0x00000400, 0x80828102,
+	0xbf84fff7, 0xbefc03ff,
+	0x00000068, 0xbe803000,
+	0xbe813000, 0xbe823000,
+	0xbe833000, 0x80fc847c,
+	0xbf84fffa, 0xbeea0480,
+	0xbeec0480, 0xbeee0480,
+	0xbef00480, 0xbef20480,
+	0xbef40480, 0xbef60480,
+	0xbef80480, 0xbefa0480,
+	0xbf810000, 0xbf9f0000,
+	0xbf9f0000, 0xbf9f0000,
+	0xbf9f0000, 0xbf9f0000,
+};
+
+/* Define the cleaner shader gfx_10_3_0 */
+static const u32 gfx_10_3_0_cleaner_shader_hex[] = {
+	0xb0804004, 0xbf8a0000,
+	0xbe8203b8, 0xbefc0380,
+	0x7e008480, 0x7e028480,
+	0x7e048480, 0x7e068480,
+	0x7e088480, 0x7e0a8480,
+	0x7e0c8480, 0x7e0e8480,
+	0xbefc0302, 0x80828802,
+	0xbf84fff5, 0xbe8203ff,
+	0x80000000, 0x87020002,
+	0xbf840012, 0xbefe03c1,
+	0xbeff03c1, 0xd7650001,
+	0x0001007f, 0xd7660001,
+	0x0002027e, 0x16020288,
+	0xbe8203bf, 0xbefc03c1,
+	0xd9382000, 0x00020201,
+	0xd9386040, 0x00040401,
+	0xd70f6a01, 0x000202ff,
+	0x00000400, 0x80828102,
+	0xbf84fff7, 0xbefc03ff,
+	0x00000068, 0xbe803080,
+	0xbe813080, 0xbe823080,
+	0xbe833080, 0x80fc847c,
+	0xbf84fffa, 0xbeea0480,
+	0xbeec0480, 0xbeee0480,
+	0xbef00480, 0xbef20480,
+	0xbef40480, 0xbef60480,
+	0xbef80480, 0xbefa0480,
+	0xbf810000, 0xbf9f0000,
+	0xbf9f0000, 0xbf9f0000,
+	0xbf9f0000, 0xbf9f0000,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_1_10_cleaner_shader.asm b/drivers/gpu/drm/amd/amdgpu/gfx_v10_1_10_cleaner_shader.asm
new file mode 100644
index 000000000000..54f7ed9e2801
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_1_10_cleaner_shader.asm
@@ -0,0 +1,125 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+// This shader is to clean LDS, SGPRs and VGPRs. It is  first 64 Dwords or 256 bytes of 256 Dwords cleaner shader.
+
+// GFX10.1 : Clear SGPRs, VGPRs and LDS
+//   Launch 32 waves per CU (16 per SIMD) as a workgroup (threadgroup) to fill every wave slot
+//   Waves are "wave32" and have 64 VGPRs each, which uses all 1024 VGPRs per SIMD
+//   Waves are launched in "CU" mode, and the workgroup shares 64KB of LDS (half of the WGP's LDS)
+//      It takes 2 workgroups to use all of LDS: one on each CU of the WGP
+//   Each wave clears SGPRs 0 - 107
+//   Each wave clears VGPRs 0 - 63
+//   The first wave of the workgroup clears its 64KB of LDS
+//   The shader starts with "S_BARRIER" to ensure SPI has launched all waves of the workgroup
+//       before any wave in the workgroup could end.  Without this, it is possible not all SGPRs get cleared.
+
+
+shader main
+  asic(GFX10.1)
+  type(CS)
+  wave_size(32)
+// Note: original source code from SQ team
+//
+// Create 32 waves in a threadgroup (CS waves)
+// Each allocates 64 VGPRs
+// The workgroup allocates all of LDS (64kbytes)
+//
+// Takes about 2500 clocks to run.
+//   (theorhetical fastest = 1024clks vgpr + 640lds = 1660 clks)
+//
+  S_BARRIER
+  s_cmp_eq_u32 s0, 1                                // Bit0 is set, sgpr0 is set then clear VGPRS and LDS as FW set COMPUTE_USER_DATA_0
+  s_cbranch_scc0  label_0023                        // Clean VGPRs and LDS if sgpr0 of wave is set, scc = (s0 == 1)
+
+  s_mov_b32     s2, 0x00000038  // Loop 64/8=8 times  (loop unrolled for performance)
+  s_mov_b32     m0, 0
+  //
+  // CLEAR VGPRs
+  //
+label_0005:
+  v_movreld_b32     v0, 0
+  v_movreld_b32     v1, 0
+  v_movreld_b32     v2, 0
+  v_movreld_b32     v3, 0
+  v_movreld_b32     v4, 0
+  v_movreld_b32     v5, 0
+  v_movreld_b32     v6, 0
+  v_movreld_b32     v7, 0
+  s_mov_b32         m0, s2
+  s_sub_u32     s2, s2, 8
+  s_cbranch_scc0  label_0005
+  //
+  s_mov_b32     s2, 0x80000000                       // Bit31 is first_wave
+  s_and_b32     s2, s2, s1                           // sgpr0 has tg_size (first_wave) term as in ucode only COMPUTE_PGM_RSRC2.tg_size_en is set
+  s_cbranch_scc0  label_0023                         // Clean LDS if its first wave of ThreadGroup/WorkGroup
+  // CLEAR LDS
+  //
+  s_mov_b32 exec_lo, 0xffffffff
+  s_mov_b32 exec_hi, 0xffffffff
+  v_mbcnt_lo_u32_b32  v1, exec_hi, 0          // Set V1 to thread-ID (0..63)
+  v_mbcnt_hi_u32_b32  v1, exec_lo, v1        // Set V1 to thread-ID (0..63)
+  v_mul_u32_u24  v1, 0x00000008, v1          // * 8, so each thread is a double-dword address (8byte)
+  s_mov_b32     s2, 0x00000003f                    // 64 loop iterations
+  s_mov_b32     m0, 0xffffffff
+  // Clear all of LDS space
+  // Each FirstWave of WorkGroup clears 64kbyte block
+
+label_001F:
+  ds_write2_b64  v1, v[2:3], v[2:3] offset1:32
+  ds_write2_b64  v1, v[4:5], v[4:5] offset0:64 offset1:96
+  v_add_co_u32     v1, vcc, 0x00000400, v1
+  s_sub_u32     s2, s2, 1
+  s_cbranch_scc0  label_001F
+
+  //
+  // CLEAR SGPRs
+  //
+label_0023:
+  s_mov_b32     m0, 0x00000068  // Loop 108/4=27 times  (loop unrolled for performance)
+label_sgpr_loop:
+  s_movreld_b32     s0, s0
+  s_movreld_b32     s1, s0
+  s_movreld_b32     s2, s0
+  s_movreld_b32     s3, s0
+  s_sub_u32         m0, m0, 4
+  s_cbranch_scc0  label_sgpr_loop
+
+  //clear vcc
+  s_mov_b64 vcc, 0          //clear vcc
+  //s_setreg_imm32_b32 hw_reg_shader_flat_scratch_lo, 0   //clear  flat scratch lo SGPR
+  //s_setreg_imm32_b32 hw_reg_shader_flat_scratch_hi, 0    //clear  flat scratch hi SGPR
+  s_mov_b64 ttmp0, 0        //Clear ttmp0 and ttmp1
+  s_mov_b64 ttmp2, 0        //Clear ttmp2 and ttmp3
+  s_mov_b64 ttmp4, 0        //Clear ttmp4 and ttmp5
+  s_mov_b64 ttmp6, 0        //Clear ttmp6 and ttmp7
+  s_mov_b64 ttmp8, 0        //Clear ttmp8 and ttmp9
+  s_mov_b64 ttmp10, 0       //Clear ttmp10 and ttmp11
+  s_mov_b64 ttmp12, 0       //Clear ttmp12 and ttmp13
+  s_mov_b64 ttmp14, 0       //Clear ttmp14 and ttmp15
+
+ s_endpgm
+
+end
+
+
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_3_0_cleaner_shader.asm b/drivers/gpu/drm/amd/amdgpu/gfx_v10_3_0_cleaner_shader.asm
new file mode 100644
index 000000000000..0e1c246166c0
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_3_0_cleaner_shader.asm
@@ -0,0 +1,124 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+// This shader is to clean LDS, SGPRs and VGPRs. It is  first 64 Dwords or 256 bytes of 192 Dwords cleaner shader.
+//To turn this shader program on for complitaion change this to main and lower shader main to main_1
+
+// GFX10.3 : Clear SGPRs, VGPRs and LDS
+//   Launch 32 waves per CU (16 per SIMD) as a workgroup (threadgroup) to fill every wave slot
+//   Waves are "wave32" and have 64 VGPRs each, which uses all 1024 VGPRs per SIMD
+//   Waves are launched in "CU" mode, and the workgroup shares 64KB of LDS (half of the WGP's LDS)
+//      It takes 2 workgroups to use all of LDS: one on each CU of the WGP
+//   Each wave clears SGPRs 0 - 107
+//   Each wave clears VGPRs 0 - 63
+//   The first wave of the workgroup clears its 64KB of LDS
+//   The shader starts with "S_BARRIER" to ensure SPI has launched all waves of the workgroup
+//       before any wave in the workgroup could end.  Without this, it is possible not all SGPRs get cleared.
+
+
+shader main
+  asic(GFX10)
+  type(CS)
+  wave_size(32)
+// Note: original source code from SQ team
+
+//
+// Create 32 waves in a threadgroup (CS waves)
+// Each allocates 64 VGPRs
+// The workgroup allocates all of LDS (64kbytes)
+//
+// Takes about 2500 clocks to run.
+//   (theorhetical fastest = 1024clks vgpr + 640lds = 1660 clks)
+//
+  S_BARRIER
+  s_mov_b32     s2, 0x00000038  // Loop 64/8=8 times  (loop unrolled for performance)
+  s_mov_b32     m0, 0
+  //
+  // CLEAR VGPRs
+  //
+label_0005:
+  v_movreld_b32     v0, 0
+  v_movreld_b32     v1, 0
+  v_movreld_b32     v2, 0
+  v_movreld_b32     v3, 0
+  v_movreld_b32     v4, 0
+  v_movreld_b32     v5, 0
+  v_movreld_b32     v6, 0
+  v_movreld_b32     v7, 0
+  s_mov_b32         m0, s2
+  s_sub_u32     s2, s2, 8
+  s_cbranch_scc0  label_0005
+  //
+  s_mov_b32     s2, 0x80000000                     // Bit31 is first_wave
+  s_and_b32     s2, s2, s0                                  // sgpr0 has tg_size (first_wave) term as in ucode only COMPUTE_PGM_RSRC2.tg_size_en is set
+  s_cbranch_scc0  label_0023                         // Clean LDS if its first wave of ThreadGroup/WorkGroup
+  // CLEAR LDS
+  //
+  s_mov_b32 exec_lo, 0xffffffff
+  s_mov_b32 exec_hi, 0xffffffff
+  v_mbcnt_lo_u32_b32  v1, exec_hi, 0          // Set V1 to thread-ID (0..63)
+  v_mbcnt_hi_u32_b32  v1, exec_lo, v1        // Set V1 to thread-ID (0..63)
+  v_mul_u32_u24  v1, 0x00000008, v1          // * 8, so each thread is a double-dword address (8byte)
+  s_mov_b32     s2, 0x00000003f                    // 64 loop iterations
+  s_mov_b32     m0, 0xffffffff
+  // Clear all of LDS space
+  // Each FirstWave of WorkGroup clears 64kbyte block
+
+label_001F:
+  ds_write2_b64  v1, v[2:3], v[2:3] offset1:32
+  ds_write2_b64  v1, v[4:5], v[4:5] offset0:64 offset1:96
+  v_add_co_u32     v1, vcc, 0x00000400, v1
+  s_sub_u32     s2, s2, 1
+  s_cbranch_scc0  label_001F
+
+  //
+  // CLEAR SGPRs
+  //
+label_0023:
+  s_mov_b32     m0, 0x00000068  // Loop 108/4=27 times  (loop unrolled for performance)
+label_sgpr_loop:
+  s_movreld_b32     s0, 0
+  s_movreld_b32     s1, 0
+  s_movreld_b32     s2, 0
+  s_movreld_b32     s3, 0
+  s_sub_u32         m0, m0, 4
+  s_cbranch_scc0  label_sgpr_loop
+
+  //clear vcc
+  s_mov_b32 flat_scratch_lo, 0   //clear  flat scratch lo SGPR
+  s_mov_b32 flat_scratch_hi, 0   //clear  flat scratch hi SGPR
+  s_mov_b64 vcc, 0          //clear vcc
+  s_mov_b64 ttmp0, 0        //Clear ttmp0 and ttmp1
+  s_mov_b64 ttmp2, 0        //Clear ttmp2 and ttmp3
+  s_mov_b64 ttmp4, 0        //Clear ttmp4 and ttmp5
+  s_mov_b64 ttmp6, 0        //Clear ttmp6 and ttmp7
+  s_mov_b64 ttmp8, 0        //Clear ttmp8 and ttmp9
+  s_mov_b64 ttmp10, 0       //Clear ttmp10 and ttmp11
+  s_mov_b64 ttmp12, 0       //Clear ttmp12 and ttmp13
+  s_mov_b64 ttmp14, 0       //Clear ttmp14 and ttmp15
+
+ s_endpgm
+
+end
+
+
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
index a9da0486467a..8a2ee2de390f 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
@@ -29,7 +29,6 @@
 #include "amdgpu_gfx.h"
 #include "amdgpu_psp.h"
 #include "amdgpu_smu.h"
-#include "amdgpu_atomfirmware.h"
 #include "imu_v11_0.h"
 #include "soc21.h"
 #include "nvd.h"
@@ -42,13 +41,15 @@
 #include "ivsrcid/gfx/irqsrcs_gfx_11_0_0.h"
 
 #include "soc15.h"
-#include "soc15d.h"
 #include "clearstate_gfx11.h"
 #include "v11_structs.h"
 #include "gfx_v11_0.h"
+#include "gfx_v11_0_cleaner_shader.h"
 #include "gfx_v11_0_3.h"
 #include "nbio_v4_3.h"
 #include "mes_v11_0.h"
+#include "mes_userqueue.h"
+#include "amdgpu_userq_fence.h"
 
 #define GFX11_NUM_GFX_RINGS		1
 #define GFX11_MEC_HPD_SIZE	2048
@@ -60,11 +61,32 @@
 #define regCGTT_WD_CLK_CTRL_BASE_IDX	1
 #define regRLC_RLCS_BOOTLOAD_STATUS_gc_11_0_1	0x4e7e
 #define regRLC_RLCS_BOOTLOAD_STATUS_gc_11_0_1_BASE_IDX	1
+#define regPC_CONFIG_CNTL_1		0x194d
+#define regPC_CONFIG_CNTL_1_BASE_IDX	1
+
+#define regCP_GFX_MQD_CONTROL_DEFAULT                                             0x00000100
+#define regCP_GFX_HQD_VMID_DEFAULT                                                0x00000000
+#define regCP_GFX_HQD_QUEUE_PRIORITY_DEFAULT                                      0x00000000
+#define regCP_GFX_HQD_QUANTUM_DEFAULT                                             0x00000a01
+#define regCP_GFX_HQD_CNTL_DEFAULT                                                0x00a00000
+#define regCP_RB_DOORBELL_CONTROL_DEFAULT                                         0x00000000
+#define regCP_GFX_HQD_RPTR_DEFAULT                                                0x00000000
+
+#define regCP_HQD_EOP_CONTROL_DEFAULT                                             0x00000006
+#define regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT                                     0x00000000
+#define regCP_MQD_CONTROL_DEFAULT                                                 0x00000100
+#define regCP_HQD_PQ_CONTROL_DEFAULT                                              0x00308509
+#define regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT                                     0x00000000
+#define regCP_HQD_PQ_RPTR_DEFAULT                                                 0x00000000
+#define regCP_HQD_PERSISTENT_STATE_DEFAULT                                        0x0be05501
+#define regCP_HQD_IB_CONTROL_DEFAULT                                              0x00300000
 
 MODULE_FIRMWARE("amdgpu/gc_11_0_0_pfp.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_0_me.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_0_mec.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_0_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_0_0_rlc_kicker.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_0_0_rlc_1.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_0_toc.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_1_pfp.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_1_me.bin");
@@ -82,6 +104,200 @@ MODULE_FIRMWARE("amdgpu/gc_11_0_4_pfp.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_4_me.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_4_mec.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_4_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_0_pfp.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_0_me.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_0_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_0_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_1_pfp.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_1_me.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_1_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_1_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_2_pfp.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_2_me.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_2_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_2_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_3_pfp.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_3_me.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_3_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_3_rlc.bin");
+
+static const struct amdgpu_hwip_reg_entry gc_reg_list_11_0[] = {
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS2),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS3),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_STALLED_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_STALLED_STAT3),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_BUSY_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_BUSY_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HPD_STATUS0),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB0_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB0_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB0_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB1_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB1_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB1_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB2_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB2_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB2_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB2_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCPF_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCPC_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCPG_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regGDS_PROTECTION_FAULT),
+	SOC15_REG_ENTRY_STR(GC, 0, regGDS_VM_PROTECTION_FAULT),
+	SOC15_REG_ENTRY_STR(GC, 0, regIA_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regIA_UTCL1_STATUS_2),
+	SOC15_REG_ENTRY_STR(GC, 0, regPA_CL_CNTL_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regRMI_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regSQC_CACHES),
+	SOC15_REG_ENTRY_STR(GC, 0, regSQG_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regWD_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regGCVM_L2_PROTECTION_FAULT_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_DEBUG),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC1_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_DEBUG_INTERRUPT_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_STATUS),
+	/* cp header registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	/* SE status registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE0),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE1),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE2),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE3),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE4),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE5)
+};
+
+static const struct amdgpu_hwip_reg_entry gc_cp_reg_list_11[] = {
+	/* compute registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_VMID),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PERSISTENT_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PIPE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_QUEUE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_QUANTUM),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_DEQUEUE_REQUEST),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_EVENTS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_BASE_ADDR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CNTL_STACK_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CNTL_STACK_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_WG_STATE_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_GDS_RESOURCE_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_WPTR_MEM),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_SUSPEND_CNTL_STACK_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_SUSPEND_CNTL_STACK_DW_CNT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_SUSPEND_WG_STATE_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_DEQUEUE_STATUS),
+	/* cp header registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+};
+
+static const struct amdgpu_hwip_reg_entry gc_gfx_queue_reg_list_11[] = {
+	/* gfx queue registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_ACTIVE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_VMID),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_QUEUE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_QUANTUM),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_CSMD_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_DEQUEUE_REQUEST),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_MAPPED),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_QUE_MGR_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_HQ_CONTROL0),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_HQ_STATUS0),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_MQD_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_MQD_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_WPTR_POLL_ADDR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_WPTR_POLL_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BUFSZ),
+	/* cp header registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+};
+
+static const struct soc15_reg_golden golden_settings_gc_11_0[] = {
+	SOC15_REG_GOLDEN_VALUE(GC, 0, regTCP_CNTL, 0x20000000, 0x20000000)
+};
 
 static const struct soc15_reg_golden golden_settings_gc_11_0_1[] =
 {
@@ -112,7 +328,7 @@ static int gfx_v11_0_get_cu_info(struct amdgpu_device *adev,
                                  struct amdgpu_cu_info *cu_info);
 static uint64_t gfx_v11_0_get_gpu_clock_counter(struct amdgpu_device *adev);
 static void gfx_v11_0_select_se_sh(struct amdgpu_device *adev, u32 se_num,
-				   u32 sh_num, u32 instance);
+				   u32 sh_num, u32 instance, int xcc_id);
 static u32 gfx_v11_0_get_wgp_active_bitmap_per_sh(struct amdgpu_device *adev);
 
 static void gfx_v11_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume);
@@ -123,20 +339,27 @@ static int gfx_v11_0_wait_for_rlc_autoload_complete(struct amdgpu_device *adev);
 static void gfx_v11_0_ring_invalidate_tlbs(struct amdgpu_ring *ring,
 					   uint16_t pasid, uint32_t flush_type,
 					   bool all_hub, uint8_t dst_sel);
-static void gfx_v11_0_set_safe_mode(struct amdgpu_device *adev);
-static void gfx_v11_0_unset_safe_mode(struct amdgpu_device *adev);
+static void gfx_v11_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id);
+static void gfx_v11_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id);
 static void gfx_v11_0_update_perf_clk(struct amdgpu_device *adev,
 				      bool enable);
 
 static void gfx11_kiq_set_resources(struct amdgpu_ring *kiq_ring, uint64_t queue_mask)
 {
+	struct amdgpu_device *adev = kiq_ring->adev;
+	u64 shader_mc_addr;
+
+	/* Cleaner shader MC address */
+	shader_mc_addr = adev->gfx.cleaner_shader_gpu_addr >> 8;
+
 	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_RESOURCES, 6));
 	amdgpu_ring_write(kiq_ring, PACKET3_SET_RESOURCES_VMID_MASK(0) |
+			  PACKET3_SET_RESOURCES_UNMAP_LATENTY(0xa) | /* unmap_latency: 0xa (~ 1s) */
 			  PACKET3_SET_RESOURCES_QUEUE_TYPE(0));	/* vmid_mask:0 queue_type:0 (KIQ) */
 	amdgpu_ring_write(kiq_ring, lower_32_bits(queue_mask));	/* queue mask lo */
 	amdgpu_ring_write(kiq_ring, upper_32_bits(queue_mask));	/* queue mask hi */
-	amdgpu_ring_write(kiq_ring, 0);	/* gws mask lo */
-	amdgpu_ring_write(kiq_ring, 0);	/* gws mask hi */
+	amdgpu_ring_write(kiq_ring, lower_32_bits(shader_mc_addr)); /* cleaner shader addr lo */
+	amdgpu_ring_write(kiq_ring, upper_32_bits(shader_mc_addr)); /* cleaner shader addr hi */
 	amdgpu_ring_write(kiq_ring, 0);	/* oac mask */
 	amdgpu_ring_write(kiq_ring, 0);	/* gds heap base:0, gds heap size:0 */
 }
@@ -192,7 +415,7 @@ static void gfx11_kiq_unmap_queues(struct amdgpu_ring *kiq_ring,
 	struct amdgpu_device *adev = kiq_ring->adev;
 	uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0;
 
-	if (adev->enable_mes && !adev->gfx.kiq.ring.sched.ready) {
+	if (adev->enable_mes && !adev->gfx.kiq[0].ring.sched.ready) {
 		amdgpu_mes_unmap_legacy_queue(adev, ring, action, gpu_addr, seq);
 		return;
 	}
@@ -260,12 +483,15 @@ static const struct kiq_pm4_funcs gfx_v11_0_kiq_pm4_funcs = {
 
 static void gfx_v11_0_set_kiq_pm4_funcs(struct amdgpu_device *adev)
 {
-	adev->gfx.kiq.pmf = &gfx_v11_0_kiq_pm4_funcs;
+	adev->gfx.kiq[0].pmf = &gfx_v11_0_kiq_pm4_funcs;
 }
 
 static void gfx_v11_0_init_golden_registers(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 4):
 		soc15_program_register_sequence(adev,
@@ -275,6 +501,10 @@ static void gfx_v11_0_init_golden_registers(struct amdgpu_device *adev)
 	default:
 		break;
 	}
+	soc15_program_register_sequence(adev,
+					golden_settings_gc_11_0,
+					(const u32)ARRAY_SIZE(golden_settings_gc_11_0));
+
 }
 
 static void gfx_v11_0_write_data_to_reg(struct amdgpu_ring *ring, int eng_sel,
@@ -310,6 +540,21 @@ static void gfx_v11_0_wait_reg_mem(struct amdgpu_ring *ring, int eng_sel,
 	amdgpu_ring_write(ring, inv); /* poll interval */
 }
 
+static void gfx_v11_ring_insert_nop(struct amdgpu_ring *ring, uint32_t num_nop)
+{
+	/* Header itself is a NOP packet */
+	if (num_nop == 1) {
+		amdgpu_ring_write(ring, ring->funcs->nop);
+		return;
+	}
+
+	/* Max HW optimization till 0x3ffe, followed by remaining one NOP at a time*/
+	amdgpu_ring_write(ring, PACKET3(PACKET3_NOP, min(num_nop - 2, 0x3ffe)));
+
+	/* Header is at index 0, followed by num_nops - 1 NOP packet's */
+	amdgpu_ring_insert_nop(ring, num_nop - 1);
+}
+
 static int gfx_v11_0_ring_test_ring(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
@@ -358,7 +603,7 @@ static int gfx_v11_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	struct dma_fence *f = NULL;
 	unsigned index;
 	uint64_t gpu_addr;
-	volatile uint32_t *cpu_ptr;
+	uint32_t *cpu_ptr;
 	long r;
 
 	/* MES KIQ fw hasn't indirect buffer support for now */
@@ -368,33 +613,18 @@ static int gfx_v11_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 
 	memset(&ib, 0, sizeof(ib));
 
-	if (ring->is_mes_queue) {
-		uint32_t padding, offset;
-
-		offset = amdgpu_mes_ctx_get_offs(ring, AMDGPU_MES_CTX_IB_OFFS);
-		padding = amdgpu_mes_ctx_get_offs(ring,
-						  AMDGPU_MES_CTX_PADDING_OFFS);
-
-		ib.gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		ib.ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-
-		gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, padding);
-		cpu_ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, padding);
-		*cpu_ptr = cpu_to_le32(0xCAFEDEAD);
-	} else {
-		r = amdgpu_device_wb_get(adev, &index);
-		if (r)
-			return r;
+	r = amdgpu_device_wb_get(adev, &index);
+	if (r)
+		return r;
 
-		gpu_addr = adev->wb.gpu_addr + (index * 4);
-		adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD);
-		cpu_ptr = &adev->wb.wb[index];
+	gpu_addr = adev->wb.gpu_addr + (index * 4);
+	adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD);
+	cpu_ptr = &adev->wb.wb[index];
 
-		r = amdgpu_ib_get(adev, NULL, 16, AMDGPU_IB_POOL_DIRECT, &ib);
-		if (r) {
-			DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
-			goto err1;
-		}
+	r = amdgpu_ib_get(adev, NULL, 20, AMDGPU_IB_POOL_DIRECT, &ib);
+	if (r) {
+		DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
+		goto err1;
 	}
 
 	ib.ptr[0] = PACKET3(PACKET3_WRITE_DATA, 3);
@@ -421,12 +651,10 @@ static int gfx_v11_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	else
 		r = -EINVAL;
 err2:
-	if (!ring->is_mes_queue)
-		amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err1:
-	if (!ring->is_mes_queue)
-		amdgpu_device_wb_free(adev, index);
+	amdgpu_device_wb_free(adev, index);
 	return r;
 }
 
@@ -444,10 +672,10 @@ static int gfx_v11_0_init_toc_microcode(struct amdgpu_device *adev, const char *
 {
 	const struct psp_firmware_header_v1_0 *toc_hdr;
 	int err = 0;
-	char fw_name[40];
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_toc.bin", ucode_prefix);
-	err = amdgpu_ucode_request(adev, &adev->psp.toc_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->psp.toc_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_toc.bin", ucode_prefix);
 	if (err)
 		goto out;
 
@@ -463,10 +691,30 @@ out:
 	return err;
 }
 
+static void gfx_v11_0_check_fw_cp_gfx_shadow(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(11, 0, 0):
+	case IP_VERSION(11, 0, 2):
+	case IP_VERSION(11, 0, 3):
+		if ((adev->gfx.me_fw_version >= 1505) &&
+		    (adev->gfx.pfp_fw_version >= 1600) &&
+		    (adev->gfx.mec_fw_version >= 512)) {
+			if (amdgpu_sriov_vf(adev))
+				adev->gfx.cp_gfx_shadow = true;
+			else
+				adev->gfx.cp_gfx_shadow = false;
+		}
+		break;
+	default:
+		adev->gfx.cp_gfx_shadow = false;
+		break;
+	}
+}
+
 static int gfx_v11_0_init_microcode(struct amdgpu_device *adev)
 {
-	char fw_name[40];
-	char ucode_prefix[30];
+	char ucode_prefix[25];
 	int err;
 	const struct rlc_firmware_header_v2_0 *rlc_hdr;
 	uint16_t version_major;
@@ -475,9 +723,9 @@ static int gfx_v11_0_init_microcode(struct amdgpu_device *adev)
 	DRM_DEBUG("\n");
 
 	amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix));
-
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", ucode_prefix);
-	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_pfp.bin", ucode_prefix);
 	if (err)
 		goto out;
 	/* check pfp fw hdr version to decide if enable rs64 for gfx11.*/
@@ -493,8 +741,9 @@ static int gfx_v11_0_init_microcode(struct amdgpu_device *adev)
 		amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_PFP);
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", ucode_prefix);
-	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_me.bin", ucode_prefix);
 	if (err)
 		goto out;
 	if (adev->gfx.rs64_enable) {
@@ -506,8 +755,19 @@ static int gfx_v11_0_init_microcode(struct amdgpu_device *adev)
 	}
 
 	if (!amdgpu_sriov_vf(adev)) {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", ucode_prefix);
-		err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, fw_name);
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(11, 0, 0) &&
+		    adev->pdev->revision == 0xCE)
+			err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/gc_11_0_0_rlc_1.bin");
+		else if (amdgpu_is_kicker_fw(adev))
+			err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_rlc_kicker.bin", ucode_prefix);
+		else
+			err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_rlc.bin", ucode_prefix);
 		if (err)
 			goto out;
 		rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
@@ -518,8 +778,9 @@ static int gfx_v11_0_init_microcode(struct amdgpu_device *adev)
 			goto out;
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", ucode_prefix);
-	err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_mec.bin", ucode_prefix);
 	if (err)
 		goto out;
 	if (adev->gfx.rs64_enable) {
@@ -539,6 +800,15 @@ static int gfx_v11_0_init_microcode(struct amdgpu_device *adev)
 	/* only one MEC for gfx 11.0.0. */
 	adev->gfx.mec2_fw = NULL;
 
+	gfx_v11_0_check_fw_cp_gfx_shadow(adev);
+
+	if (adev->gfx.imu.funcs && adev->gfx.imu.funcs->init_microcode) {
+		err = adev->gfx.imu.funcs->init_microcode(adev);
+		if (err)
+			DRM_ERROR("Failed to init imu firmware!\n");
+		return err;
+	}
+
 out:
 	if (err) {
 		amdgpu_ucode_release(&adev->gfx.pfp_fw);
@@ -580,12 +850,9 @@ static u32 gfx_v11_0_get_csb_size(struct amdgpu_device *adev)
 	return count;
 }
 
-static void gfx_v11_0_get_csb_buffer(struct amdgpu_device *adev,
-				    volatile u32 *buffer)
+static void gfx_v11_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer)
 {
-	u32 count = 0, i;
-	const struct cs_section_def *sect = NULL;
-	const struct cs_extent_def *ext = NULL;
+	u32 count = 0;
 	int ctx_reg_offset;
 
 	if (adev->gfx.rlc.cs_data == NULL)
@@ -593,39 +860,15 @@ static void gfx_v11_0_get_csb_buffer(struct amdgpu_device *adev,
 	if (buffer == NULL)
 		return;
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1));
-	buffer[count++] = cpu_to_le32(0x80000000);
-	buffer[count++] = cpu_to_le32(0x80000000);
-
-	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
-		for (ext = sect->section; ext->extent != NULL; ++ext) {
-			if (sect->id == SECT_CONTEXT) {
-				buffer[count++] =
-					cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count));
-				buffer[count++] = cpu_to_le32(ext->reg_index -
-						PACKET3_SET_CONTEXT_REG_START);
-				for (i = 0; i < ext->reg_count; i++)
-					buffer[count++] = cpu_to_le32(ext->extent[i]);
-			} else {
-				return;
-			}
-		}
-	}
+	count = amdgpu_gfx_csb_preamble_start(buffer);
+	count = amdgpu_gfx_csb_data_parser(adev, buffer, count);
 
-	ctx_reg_offset =
-		SOC15_REG_OFFSET(GC, 0, regPA_SC_TILE_STEERING_OVERRIDE) - PACKET3_SET_CONTEXT_REG_START;
+	ctx_reg_offset = SOC15_REG_OFFSET(GC, 0, regPA_SC_TILE_STEERING_OVERRIDE) - PACKET3_SET_CONTEXT_REG_START;
 	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, 1));
 	buffer[count++] = cpu_to_le32(ctx_reg_offset);
 	buffer[count++] = cpu_to_le32(adev->gfx.config.pa_sc_tile_steering_override);
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0));
-	buffer[count++] = cpu_to_le32(0);
+	amdgpu_gfx_csb_preamble_end(buffer, count);
 }
 
 static void gfx_v11_0_rlc_fini(struct amdgpu_device *adev)
@@ -645,7 +888,7 @@ static void gfx_v11_0_init_rlcg_reg_access_ctrl(struct amdgpu_device *adev)
 {
 	struct amdgpu_rlcg_reg_access_ctrl *reg_access_ctrl;
 
-	reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl;
+	reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl[0];
 	reg_access_ctrl->scratch_reg0 = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG0);
 	reg_access_ctrl->scratch_reg1 = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG1);
 	reg_access_ctrl->scratch_reg2 = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG2);
@@ -674,7 +917,7 @@ static int gfx_v11_0_rlc_init(struct amdgpu_device *adev)
 
 	/* init spm vmid with 0xf */
 	if (adev->gfx.rlc.funcs->update_spm_vmid)
-		adev->gfx.rlc.funcs->update_spm_vmid(adev, 0xf);
+		adev->gfx.rlc.funcs->update_spm_vmid(adev, NULL, 0xf);
 
 	return 0;
 }
@@ -699,7 +942,7 @@ static int gfx_v11_0_mec_init(struct amdgpu_device *adev)
 	u32 *hpd;
 	size_t mec_hpd_size;
 
-	bitmap_zero(adev->gfx.mec.queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
+	bitmap_zero(adev->gfx.mec_bitmap[0].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
 
 	/* take ownership of the relevant compute queues */
 	amdgpu_gfx_compute_queue_acquire(adev);
@@ -747,7 +990,7 @@ static void wave_read_regs(struct amdgpu_device *adev, uint32_t wave,
 		*(out++) = RREG32_SOC15(GC, 0, regSQ_IND_DATA);
 }
 
-static void gfx_v11_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
+static void gfx_v11_0_read_wave_data(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
 {
 	/* in gfx11 the SIMD_ID is specified as part of the INSTANCE
 	 * field when performing a select_se_sh so it should be
@@ -773,7 +1016,7 @@ static void gfx_v11_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd,
 	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_MODE);
 }
 
-static void gfx_v11_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v11_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				     uint32_t wave, uint32_t start,
 				     uint32_t size, uint32_t *dst)
 {
@@ -784,7 +1027,7 @@ static void gfx_v11_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
 		dst);
 }
 
-static void gfx_v11_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v11_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				      uint32_t wave, uint32_t thread,
 				      uint32_t start, uint32_t size,
 				      uint32_t *dst)
@@ -795,11 +1038,39 @@ static void gfx_v11_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd,
 }
 
 static void gfx_v11_0_select_me_pipe_q(struct amdgpu_device *adev,
-									  u32 me, u32 pipe, u32 q, u32 vm)
+					u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id)
 {
 	soc21_grbm_select(adev, me, pipe, q, vm);
 }
 
+/* all sizes are in bytes */
+#define MQD_SHADOW_BASE_SIZE      73728
+#define MQD_SHADOW_BASE_ALIGNMENT 256
+#define MQD_FWWORKAREA_SIZE       484
+#define MQD_FWWORKAREA_ALIGNMENT  256
+
+static void gfx_v11_0_get_gfx_shadow_info_nocheck(struct amdgpu_device *adev,
+					 struct amdgpu_gfx_shadow_info *shadow_info)
+{
+	shadow_info->shadow_size = MQD_SHADOW_BASE_SIZE;
+	shadow_info->shadow_alignment = MQD_SHADOW_BASE_ALIGNMENT;
+	shadow_info->csa_size = MQD_FWWORKAREA_SIZE;
+	shadow_info->csa_alignment = MQD_FWWORKAREA_ALIGNMENT;
+}
+
+static int gfx_v11_0_get_gfx_shadow_info(struct amdgpu_device *adev,
+					 struct amdgpu_gfx_shadow_info *shadow_info,
+					 bool skip_check)
+{
+	if (adev->gfx.cp_gfx_shadow || skip_check) {
+		gfx_v11_0_get_gfx_shadow_info_nocheck(adev, shadow_info);
+		return 0;
+	} else {
+		memset(shadow_info, 0, sizeof(struct amdgpu_gfx_shadow_info));
+		return -ENOTSUPP;
+	}
+}
+
 static const struct amdgpu_gfx_funcs gfx_v11_0_gfx_funcs = {
 	.get_gpu_clock_counter = &gfx_v11_0_get_gpu_clock_counter,
 	.select_se_sh = &gfx_v11_0_select_se_sh,
@@ -808,12 +1079,12 @@ static const struct amdgpu_gfx_funcs gfx_v11_0_gfx_funcs = {
 	.read_wave_vgprs = &gfx_v11_0_read_wave_vgprs,
 	.select_me_pipe_q = &gfx_v11_0_select_me_pipe_q,
 	.update_perfmon_mgcg = &gfx_v11_0_update_perf_clk,
+	.get_gfx_shadow_info = &gfx_v11_0_get_gfx_shadow_info,
 };
 
 static int gfx_v11_0_gpu_early_init(struct amdgpu_device *adev)
 {
-
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 0):
 	case IP_VERSION(11, 0, 2):
 		adev->gfx.config.max_hw_contexts = 8;
@@ -832,6 +1103,10 @@ static int gfx_v11_0_gpu_early_init(struct amdgpu_device *adev)
 		break;
 	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
 		adev->gfx.config.max_hw_contexts = 8;
 		adev->gfx.config.sc_prim_fifo_size_frontend = 0x20;
 		adev->gfx.config.sc_prim_fifo_size_backend = 0x100;
@@ -849,9 +1124,9 @@ static int gfx_v11_0_gpu_early_init(struct amdgpu_device *adev)
 static int gfx_v11_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id,
 				   int me, int pipe, int queue)
 {
-	int r;
 	struct amdgpu_ring *ring;
 	unsigned int irq_type;
+	unsigned int hw_prio;
 
 	ring = &adev->gfx.gfx_ring[ring_id];
 
@@ -861,20 +1136,23 @@ static int gfx_v11_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id,
 
 	ring->ring_obj = NULL;
 	ring->use_doorbell = true;
+	if (adev->gfx.disable_kq) {
+		ring->no_scheduler = true;
+		ring->no_user_submission = true;
+	}
 
 	if (!ring_id)
 		ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
 	else
 		ring->doorbell_index = adev->doorbell_index.gfx_ring1 << 1;
-	ring->vm_hub = AMDGPU_GFXHUB_0;
+	ring->vm_hub = AMDGPU_GFXHUB(0);
 	sprintf(ring->name, "gfx_%d.%d.%d", ring->me, ring->pipe, ring->queue);
 
 	irq_type = AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP + ring->pipe;
-	r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type,
-			     AMDGPU_RING_PRIO_DEFAULT, NULL);
-	if (r)
-		return r;
-	return 0;
+	hw_prio = amdgpu_gfx_is_high_priority_graphics_queue(adev, ring) ?
+		AMDGPU_GFX_PIPE_PRIO_HIGH : AMDGPU_GFX_PIPE_PRIO_NORMAL;
+	return amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type,
+				hw_prio, NULL);
 }
 
 static int gfx_v11_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
@@ -897,7 +1175,7 @@ static int gfx_v11_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
 	ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
 	ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
 				+ (ring_id * GFX11_MEC_HPD_SIZE);
-	ring->vm_hub = AMDGPU_GFXHUB_0;
+	ring->vm_hub = AMDGPU_GFXHUB(0);
 	sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
 
 	irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
@@ -1251,30 +1529,69 @@ static int gfx_v11_0_rlc_backdoor_autoload_enable(struct amdgpu_device *adev)
 	return 0;
 }
 
-static int gfx_v11_0_sw_init(void *handle)
+static void gfx_v11_0_alloc_ip_dump(struct amdgpu_device *adev)
+{
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_11_0);
+	uint32_t *ptr;
+	uint32_t inst;
+
+	ptr = kcalloc(reg_count, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for GFX IP Dump\n");
+		adev->gfx.ip_dump_core = NULL;
+	} else {
+		adev->gfx.ip_dump_core = ptr;
+	}
+
+	/* Allocate memory for compute queue registers for all the instances */
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_11);
+	inst = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe_per_mec *
+		adev->gfx.mec.num_queue_per_pipe;
+
+	ptr = kcalloc(reg_count * inst, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for Compute Queues IP Dump\n");
+		adev->gfx.ip_dump_compute_queues = NULL;
+	} else {
+		adev->gfx.ip_dump_compute_queues = ptr;
+	}
+
+	/* Allocate memory for gfx queue registers for all the instances */
+	reg_count = ARRAY_SIZE(gc_gfx_queue_reg_list_11);
+	inst = adev->gfx.me.num_me * adev->gfx.me.num_pipe_per_me *
+		adev->gfx.me.num_queue_per_pipe;
+
+	ptr = kcalloc(reg_count * inst, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for GFX Queues IP Dump\n");
+		adev->gfx.ip_dump_gfx_queues = NULL;
+	} else {
+		adev->gfx.ip_dump_gfx_queues = ptr;
+	}
+}
+
+static int gfx_v11_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	int i, j, k, r, ring_id = 0;
-	struct amdgpu_kiq *kiq;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	int i, j, k, r, ring_id;
+	int xcc_id = 0;
+	struct amdgpu_device *adev = ip_block->adev;
+	int num_queue_per_pipe = 1; /* we only enable 1 KGQ per pipe */
 
-	adev->gfxhub.funcs->init(adev);
+	INIT_DELAYED_WORK(&adev->gfx.idle_work, amdgpu_gfx_profile_idle_work_handler);
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 0):
+	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 3):
-		adev->gfx.me.num_me = 1;
-		adev->gfx.me.num_pipe_per_me = 1;
-		adev->gfx.me.num_queue_per_pipe = 1;
-		adev->gfx.mec.num_mec = 2;
-		adev->gfx.mec.num_pipe_per_mec = 4;
-		adev->gfx.mec.num_queue_per_pipe = 4;
-		break;
-	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
 		adev->gfx.me.num_me = 1;
 		adev->gfx.me.num_pipe_per_me = 1;
-		adev->gfx.me.num_queue_per_pipe = 1;
+		adev->gfx.me.num_queue_per_pipe = 2;
 		adev->gfx.mec.num_mec = 1;
 		adev->gfx.mec.num_pipe_per_mec = 4;
 		adev->gfx.mec.num_queue_per_pipe = 4;
@@ -1289,9 +1606,118 @@ static int gfx_v11_0_sw_init(void *handle)
 		break;
 	}
 
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(11, 0, 0):
+	case IP_VERSION(11, 0, 2):
+	case IP_VERSION(11, 0, 3):
+		if (!adev->gfx.disable_uq &&
+		    adev->gfx.me_fw_version  >= 2420 &&
+		    adev->gfx.pfp_fw_version >= 2580 &&
+		    adev->gfx.mec_fw_version >= 2650 &&
+		    adev->mes.fw_version[0] >= 120) {
+			adev->userq_funcs[AMDGPU_HW_IP_GFX] = &userq_mes_funcs;
+			adev->userq_funcs[AMDGPU_HW_IP_COMPUTE] = &userq_mes_funcs;
+		}
+		break;
+	case IP_VERSION(11, 0, 1):
+	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
+		/* add firmware version checks here */
+		if (0 && !adev->gfx.disable_uq) {
+			adev->userq_funcs[AMDGPU_HW_IP_GFX] = &userq_mes_funcs;
+			adev->userq_funcs[AMDGPU_HW_IP_COMPUTE] = &userq_mes_funcs;
+		}
+		break;
+	default:
+		break;
+	}
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(11, 0, 0):
+	case IP_VERSION(11, 0, 2):
+	case IP_VERSION(11, 0, 3):
+		adev->gfx.cleaner_shader_ptr = gfx_11_0_3_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_11_0_3_cleaner_shader_hex);
+		if (adev->gfx.me_fw_version  >= 2280 &&
+		    adev->gfx.pfp_fw_version >= 2370 &&
+		    adev->gfx.mec_fw_version >= 2450  &&
+		    adev->mes.fw_version[0] >= 99) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	case IP_VERSION(11, 0, 1):
+	case IP_VERSION(11, 0, 4):
+		adev->gfx.cleaner_shader_ptr = gfx_11_0_3_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_11_0_3_cleaner_shader_hex);
+		if (adev->gfx.pfp_fw_version >= 102 &&
+		    adev->gfx.mec_fw_version >= 66 &&
+		    adev->mes.fw_version[0] >= 128) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+		adev->gfx.cleaner_shader_ptr = gfx_11_0_3_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_11_0_3_cleaner_shader_hex);
+		if (adev->gfx.mec_fw_version >= 26 &&
+		    adev->mes.fw_version[0] >= 114) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	case IP_VERSION(11, 5, 2):
+		adev->gfx.cleaner_shader_ptr = gfx_11_0_3_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_11_0_3_cleaner_shader_hex);
+		if (adev->gfx.me_fw_version  >= 12 &&
+		    adev->gfx.pfp_fw_version >= 15 &&
+		    adev->gfx.mec_fw_version >= 15) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	case IP_VERSION(11, 5, 3):
+		adev->gfx.cleaner_shader_ptr = gfx_11_0_3_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_11_0_3_cleaner_shader_hex);
+		if (adev->gfx.me_fw_version  >= 7 &&
+		    adev->gfx.pfp_fw_version >= 8 &&
+		    adev->gfx.mec_fw_version >= 8) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	default:
+		adev->gfx.enable_cleaner_shader = false;
+		break;
+	}
+
 	/* Enable CG flag in one VF mode for enabling RLC safe mode enter/exit */
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(11, 0, 3) &&
-		amdgpu_sriov_is_pp_one_vf(adev))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(11, 0, 3) &&
+	    amdgpu_sriov_is_pp_one_vf(adev))
 		adev->cg_flags = AMD_CG_SUPPORT_GFX_CGCG;
 
 	/* EOP Event */
@@ -1301,6 +1727,13 @@ static int gfx_v11_0_sw_init(void *handle)
 	if (r)
 		return r;
 
+	/* Bad opcode Event */
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP,
+			      GFX_11_0_0__SRCID__CP_BAD_OPCODE_ERROR,
+			      &adev->gfx.bad_op_irq);
+	if (r)
+		return r;
+
 	/* Privileged reg */
 	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP,
 			      GFX_11_0_0__SRCID__CP_PRIV_REG_FAULT,
@@ -1315,13 +1748,6 @@ static int gfx_v11_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	/* ECC error */
-	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP,
-				  GFX_11_0_0__SRCID__CP_ECC_ERROR,
-				  &adev->gfx.cp_ecc_error_irq);
-	if (r)
-		return r;
-
 	/* FED error */
 	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GFX,
 				  GFX_11_0_0__SRCID__RLC_GC_FED_INTERRUPT,
@@ -1331,14 +1757,6 @@ static int gfx_v11_0_sw_init(void *handle)
 
 	adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE;
 
-	if (adev->gfx.imu.funcs) {
-		if (adev->gfx.imu.funcs->init_microcode) {
-			r = adev->gfx.imu.funcs->init_microcode(adev);
-			if (r)
-				DRM_ERROR("Failed to load imu firmware!\n");
-		}
-	}
-
 	gfx_v11_0_me_init(adev);
 
 	r = gfx_v11_0_rlc_init(adev);
@@ -1353,55 +1771,84 @@ static int gfx_v11_0_sw_init(void *handle)
 		return r;
 	}
 
-	/* set up the gfx ring */
-	for (i = 0; i < adev->gfx.me.num_me; i++) {
-		for (j = 0; j < adev->gfx.me.num_queue_per_pipe; j++) {
-			for (k = 0; k < adev->gfx.me.num_pipe_per_me; k++) {
-				if (!amdgpu_gfx_is_me_queue_enabled(adev, i, k, j))
-					continue;
-
-				r = gfx_v11_0_gfx_ring_init(adev, ring_id,
-							    i, k, j);
-				if (r)
-					return r;
-				ring_id++;
+	if (adev->gfx.num_gfx_rings) {
+		ring_id = 0;
+		/* set up the gfx ring */
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < num_queue_per_pipe; j++) {
+				for (k = 0; k < adev->gfx.me.num_pipe_per_me; k++) {
+					if (!amdgpu_gfx_is_me_queue_enabled(adev, i, k, j))
+						continue;
+
+					r = gfx_v11_0_gfx_ring_init(adev, ring_id,
+								    i, k, j);
+					if (r)
+						return r;
+					ring_id++;
+				}
 			}
 		}
 	}
 
-	ring_id = 0;
-	/* set up the compute queues - allocate horizontally across pipes */
-	for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
-		for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
-			for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
-				if (!amdgpu_gfx_is_mec_queue_enabled(adev, i, k,
-								     j))
-					continue;
+	if (adev->gfx.num_compute_rings) {
+		ring_id = 0;
+		/* set up the compute queues - allocate horizontally across pipes */
+		for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
+			for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
+				for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
+					if (!amdgpu_gfx_is_mec_queue_enabled(adev, 0, i,
+									     k, j))
+						continue;
 
-				r = gfx_v11_0_compute_ring_init(adev, ring_id,
-								i, k, j);
-				if (r)
-					return r;
+					r = gfx_v11_0_compute_ring_init(adev, ring_id,
+									i, k, j);
+					if (r)
+						return r;
 
-				ring_id++;
+					ring_id++;
+				}
 			}
 		}
 	}
 
+	adev->gfx.gfx_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]);
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(11, 0, 0):
+	case IP_VERSION(11, 0, 2):
+	case IP_VERSION(11, 0, 3):
+		if ((adev->gfx.me_fw_version >= 2280) &&
+		    (adev->gfx.mec_fw_version >= 2410) &&
+		    !amdgpu_sriov_vf(adev) &&
+		    !adev->debug_disable_gpu_ring_reset) {
+			adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+			adev->gfx.gfx_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+		}
+		break;
+	default:
+		if (!amdgpu_sriov_vf(adev) &&
+		    !adev->debug_disable_gpu_ring_reset) {
+			adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+			adev->gfx.gfx_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+		}
+		break;
+	}
+
 	if (!adev->enable_mes_kiq) {
-		r = amdgpu_gfx_kiq_init(adev, GFX11_MEC_HPD_SIZE);
+		r = amdgpu_gfx_kiq_init(adev, GFX11_MEC_HPD_SIZE, 0);
 		if (r) {
 			DRM_ERROR("Failed to init KIQ BOs!\n");
 			return r;
 		}
 
-		kiq = &adev->gfx.kiq;
-		r = amdgpu_gfx_kiq_init_ring(adev, &kiq->ring, &kiq->irq);
+		r = amdgpu_gfx_kiq_init_ring(adev, xcc_id);
 		if (r)
 			return r;
 	}
 
-	r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct v11_compute_mqd));
+	r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct v11_compute_mqd), 0);
 	if (r)
 		return r;
 
@@ -1421,6 +1868,12 @@ static int gfx_v11_0_sw_init(void *handle)
 		return -EINVAL;
 	}
 
+	gfx_v11_0_alloc_ip_dump(adev);
+
+	r = amdgpu_gfx_sysfs_init(adev);
+	if (r)
+		return r;
+
 	return 0;
 }
 
@@ -1453,23 +1906,25 @@ static void gfx_v11_0_rlc_autoload_buffer_fini(struct amdgpu_device *adev)
 			(void **)&adev->gfx.rlc.rlc_autoload_ptr);
 }
 
-static int gfx_v11_0_sw_fini(void *handle)
+static int gfx_v11_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
 		amdgpu_ring_fini(&adev->gfx.gfx_ring[i]);
 	for (i = 0; i < adev->gfx.num_compute_rings; i++)
 		amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
 
-	amdgpu_gfx_mqd_sw_fini(adev);
+	amdgpu_gfx_mqd_sw_fini(adev, 0);
 
 	if (!adev->enable_mes_kiq) {
-		amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring);
-		amdgpu_gfx_kiq_fini(adev);
+		amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq[0].ring);
+		amdgpu_gfx_kiq_fini(adev, 0);
 	}
 
+	amdgpu_gfx_cleaner_shader_sw_fini(adev);
+
 	gfx_v11_0_pfp_fini(adev);
 	gfx_v11_0_me_fini(adev);
 	gfx_v11_0_rlc_fini(adev);
@@ -1480,11 +1935,17 @@ static int gfx_v11_0_sw_fini(void *handle)
 
 	gfx_v11_0_free_microcode(adev);
 
+	amdgpu_gfx_sysfs_fini(adev);
+
+	kfree(adev->gfx.ip_dump_core);
+	kfree(adev->gfx.ip_dump_compute_queues);
+	kfree(adev->gfx.ip_dump_gfx_queues);
+
 	return 0;
 }
 
 static void gfx_v11_0_select_se_sh(struct amdgpu_device *adev, u32 se_num,
-				   u32 sh_num, u32 instance)
+				   u32 sh_num, u32 instance, int xcc_id)
 {
 	u32 data;
 
@@ -1549,6 +2010,7 @@ static u32 gfx_v11_0_get_rb_active_bitmap(struct amdgpu_device *adev)
 
 static void gfx_v11_0_setup_rb(struct amdgpu_device *adev)
 {
+	u32 rb_bitmap_per_sa;
 	u32 rb_bitmap_width_per_sa;
 	u32 max_sa;
 	u32 active_sa_bitmap;
@@ -1566,12 +2028,14 @@ static void gfx_v11_0_setup_rb(struct amdgpu_device *adev)
 		 adev->gfx.config.max_sh_per_se;
 	rb_bitmap_width_per_sa = adev->gfx.config.max_backends_per_se /
 				 adev->gfx.config.max_sh_per_se;
+	rb_bitmap_per_sa = amdgpu_gfx_create_bitmask(rb_bitmap_width_per_sa);
+
 	for (i = 0; i < max_sa; i++) {
 		if (active_sa_bitmap & (1 << i))
-			active_rb_bitmap |= (0x3 << (i * rb_bitmap_width_per_sa));
+			active_rb_bitmap |= (rb_bitmap_per_sa << (i * rb_bitmap_width_per_sa));
 	}
 
-	active_rb_bitmap |= global_active_rb_bitmap;
+	active_rb_bitmap &= global_active_rb_bitmap;
 	adev->gfx.config.backend_enable_mask = active_rb_bitmap;
 	adev->gfx.config.num_rbs = hweight32(active_rb_bitmap);
 }
@@ -1605,12 +2069,15 @@ static void gfx_v11_0_init_compute_vmid(struct amdgpu_device *adev)
 		/* Enable trap for each kfd vmid. */
 		data = RREG32_SOC15(GC, 0, regSPI_GDBG_PER_VMID_CNTL);
 		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+		WREG32_SOC15(GC, 0, regSPI_GDBG_PER_VMID_CNTL, data);
 	}
 	soc21_grbm_select(adev, 0, 0, 0, 0);
 	mutex_unlock(&adev->srbm_mutex);
 
-	/* Initialize all compute VMIDs to have no GDS, GWS, or OA
-	   acccess. These should be enabled by FW for target VMIDs. */
+	/*
+	 * Initialize all compute VMIDs to have no GDS, GWS, or OA
+	 * access. These should be enabled by FW for target VMIDs.
+	 */
 	for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) {
 		WREG32_SOC15_OFFSET(GC, 0, regGDS_VMID0_BASE, 2 * i, 0);
 		WREG32_SOC15_OFFSET(GC, 0, regGDS_VMID0_SIZE, 2 * i, 0);
@@ -1674,7 +2141,7 @@ static void gfx_v11_0_constants_init(struct amdgpu_device *adev)
 	/* XXX SH_MEM regs */
 	/* where to put LDS, scratch, GPUVM in FSA64 space */
 	mutex_lock(&adev->srbm_mutex);
-	for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB_0].num_ids; i++) {
+	for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
 		soc21_grbm_select(adev, 0, 0, 0, i);
 		/* CP and shaders */
 		WREG32_SOC15(GC, 0, regSH_MEM_CONFIG, DEFAULT_SH_MEM_CONFIG);
@@ -1694,26 +2161,74 @@ static void gfx_v11_0_constants_init(struct amdgpu_device *adev)
 	gfx_v11_0_init_gds_vmid(adev);
 }
 
+static u32 gfx_v11_0_get_cpg_int_cntl(struct amdgpu_device *adev,
+				      int me, int pipe)
+{
+	if (me != 0)
+		return 0;
+
+	switch (pipe) {
+	case 0:
+		return SOC15_REG_OFFSET(GC, 0, regCP_INT_CNTL_RING0);
+	case 1:
+		return SOC15_REG_OFFSET(GC, 0, regCP_INT_CNTL_RING1);
+	default:
+		return 0;
+	}
+}
+
+static u32 gfx_v11_0_get_cpc_int_cntl(struct amdgpu_device *adev,
+				      int me, int pipe)
+{
+	/*
+	 * amdgpu controls only the first MEC. That's why this function only
+	 * handles the setting of interrupts for this specific MEC. All other
+	 * pipes' interrupts are set by amdkfd.
+	 */
+	if (me != 1)
+		return 0;
+
+	switch (pipe) {
+	case 0:
+		return SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE0_INT_CNTL);
+	case 1:
+		return SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE1_INT_CNTL);
+	case 2:
+		return SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE2_INT_CNTL);
+	case 3:
+		return SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE3_INT_CNTL);
+	default:
+		return 0;
+	}
+}
+
 static void gfx_v11_0_enable_gui_idle_interrupt(struct amdgpu_device *adev,
 					       bool enable)
 {
-	u32 tmp;
+	u32 tmp, cp_int_cntl_reg;
+	int i, j;
 
 	if (amdgpu_sriov_vf(adev))
 		return;
 
-	tmp = RREG32_SOC15(GC, 0, regCP_INT_CNTL_RING0);
-
-	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE,
-			    enable ? 1 : 0);
-	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_EMPTY_INT_ENABLE,
-			    enable ? 1 : 0);
-	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE,
-			    enable ? 1 : 0);
-	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, GFX_IDLE_INT_ENABLE,
-			    enable ? 1 : 0);
-
-	WREG32_SOC15(GC, 0, regCP_INT_CNTL_RING0, tmp);
+	for (i = 0; i < adev->gfx.me.num_me; i++) {
+		for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+			cp_int_cntl_reg = gfx_v11_0_get_cpg_int_cntl(adev, i, j);
+
+			if (cp_int_cntl_reg) {
+				tmp = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE,
+						    enable ? 1 : 0);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_EMPTY_INT_ENABLE,
+						    enable ? 1 : 0);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE,
+						    enable ? 1 : 0);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, GFX_IDLE_INT_ENABLE,
+						    enable ? 1 : 0);
+				WREG32_SOC15_IP(GC, cp_int_cntl_reg, tmp);
+			}
+		}
+	}
 }
 
 static int gfx_v11_0_init_csb(struct amdgpu_device *adev)
@@ -1925,7 +2440,7 @@ static int gfx_v11_0_rlc_load_microcode(struct amdgpu_device *adev)
 			if (version_minor == 3)
 				gfx_v11_0_load_rlcp_rlcv_microcode(adev);
 		}
-		
+
 		return 0;
 	}
 
@@ -1995,7 +2510,7 @@ static int gfx_v11_0_config_me_cache(struct amdgpu_device *adev, uint64_t addr)
 	}
 
 	if (amdgpu_emu_mode == 1)
-		adev->hdp.funcs->flush_hdp(adev, NULL);
+		amdgpu_device_flush_hdp(adev, NULL);
 
 	tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_BASE_CNTL);
 	tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, VMID, 0);
@@ -2039,7 +2554,7 @@ static int gfx_v11_0_config_pfp_cache(struct amdgpu_device *adev, uint64_t addr)
 	}
 
 	if (amdgpu_emu_mode == 1)
-		adev->hdp.funcs->flush_hdp(adev, NULL);
+		amdgpu_device_flush_hdp(adev, NULL);
 
 	tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_CNTL);
 	tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, VMID, 0);
@@ -2084,7 +2599,7 @@ static int gfx_v11_0_config_mec_cache(struct amdgpu_device *adev, uint64_t addr)
 	}
 
 	if (amdgpu_emu_mode == 1)
-		adev->hdp.funcs->flush_hdp(adev, NULL);
+		amdgpu_device_flush_hdp(adev, NULL);
 
 	tmp = RREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_CNTL);
 	tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, CACHE_POLICY, 0);
@@ -2524,8 +3039,14 @@ static int gfx_v11_0_wait_for_rlc_autoload_complete(struct amdgpu_device *adev)
 	for (i = 0; i < adev->usec_timeout; i++) {
 		cp_status = RREG32_SOC15(GC, 0, regCP_STAT);
 
-		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(11, 0, 1) ||
-				adev->ip_versions[GC_HWIP][0] == IP_VERSION(11, 0, 4))
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			    IP_VERSION(11, 0, 1) ||
+		    amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			    IP_VERSION(11, 0, 4) ||
+		    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(11, 5, 0) ||
+		    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(11, 5, 1) ||
+		    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(11, 5, 2) ||
+		    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(11, 5, 3))
 			bootload_status = RREG32_SOC15(GC, 0,
 					regRLC_RLCS_BOOTLOAD_STATUS_gc_11_0_1);
 		else
@@ -2714,7 +3235,7 @@ static int gfx_v11_0_cp_gfx_load_pfp_microcode_rs64(struct amdgpu_device *adev)
 	amdgpu_bo_unreserve(adev->gfx.pfp.pfp_fw_data_obj);
 
 	if (amdgpu_emu_mode == 1)
-		adev->hdp.funcs->flush_hdp(adev, NULL);
+		amdgpu_device_flush_hdp(adev, NULL);
 
 	WREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_LO,
 		lower_32_bits(adev->gfx.pfp.pfp_fw_gpu_addr));
@@ -2932,7 +3453,7 @@ static int gfx_v11_0_cp_gfx_load_me_microcode_rs64(struct amdgpu_device *adev)
 	amdgpu_bo_unreserve(adev->gfx.me.me_fw_data_obj);
 
 	if (amdgpu_emu_mode == 1)
-		adev->hdp.funcs->flush_hdp(adev, NULL);
+		amdgpu_device_flush_hdp(adev, NULL);
 
 	WREG32_SOC15(GC, 0, regCP_ME_IC_BASE_LO,
 		lower_32_bits(adev->gfx.me.me_fw_gpu_addr));
@@ -3195,7 +3716,6 @@ static int gfx_v11_0_cp_gfx_resume(struct amdgpu_device *adev)
 	u32 tmp;
 	u32 rb_bufsz;
 	u64 rb_addr, rptr_addr, wptr_gpu_addr;
-	u32 i;
 
 	/* Set the write pointer delay */
 	WREG32_SOC15(GC, 0, regCP_RB_WPTR_DELAY, 0);
@@ -3219,7 +3739,7 @@ static int gfx_v11_0_cp_gfx_resume(struct amdgpu_device *adev)
 	WREG32_SOC15(GC, 0, regCP_RB0_WPTR, lower_32_bits(ring->wptr));
 	WREG32_SOC15(GC, 0, regCP_RB0_WPTR_HI, upper_32_bits(ring->wptr));
 
-	/* set the wb address wether it's enabled or not */
+	/* set the wb address whether it's enabled or not */
 	rptr_addr = ring->rptr_gpu_addr;
 	WREG32_SOC15(GC, 0, regCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr));
 	WREG32_SOC15(GC, 0, regCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) &
@@ -3257,7 +3777,7 @@ static int gfx_v11_0_cp_gfx_resume(struct amdgpu_device *adev)
 		ring->wptr = 0;
 		WREG32_SOC15(GC, 0, regCP_RB1_WPTR, lower_32_bits(ring->wptr));
 		WREG32_SOC15(GC, 0, regCP_RB1_WPTR_HI, upper_32_bits(ring->wptr));
-		/* Set the wb address wether it's enabled or not */
+		/* Set the wb address whether it's enabled or not */
 		rptr_addr = ring->rptr_gpu_addr;
 		WREG32_SOC15(GC, 0, regCP_RB1_RPTR_ADDR, lower_32_bits(rptr_addr));
 		WREG32_SOC15(GC, 0, regCP_RB1_RPTR_ADDR_HI, upper_32_bits(rptr_addr) &
@@ -3287,11 +3807,6 @@ static int gfx_v11_0_cp_gfx_resume(struct amdgpu_device *adev)
 	/* start the ring */
 	gfx_v11_0_cp_gfx_start(adev);
 
-	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
-		ring = &adev->gfx.gfx_ring[i];
-		ring->sched.ready = true;
-	}
-
 	return 0;
 }
 
@@ -3337,8 +3852,6 @@ static void gfx_v11_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
 		WREG32_SOC15(GC, 0, regCP_MEC_CNTL, data);
 	}
 
-	adev->gfx.kiq.ring.sched.ready = enable;
-
 	udelay(50);
 }
 
@@ -3375,7 +3888,7 @@ static int gfx_v11_0_cp_compute_load_microcode(struct amdgpu_device *adev)
 	}
 
 	memcpy(fw, fw_data, fw_size);
-	
+
 	amdgpu_bo_kunmap(adev->gfx.mec.mec_fw_obj);
 	amdgpu_bo_unreserve(adev->gfx.mec.mec_fw_obj);
 
@@ -3534,9 +4047,7 @@ static void gfx_v11_0_kiq_setting(struct amdgpu_ring *ring)
 	tmp = RREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS);
 	tmp &= 0xffffff00;
 	tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
-	WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp);
-	tmp |= 0x80;
-	WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp);
+	WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp | 0x80);
 }
 
 static void gfx_v11_0_cp_set_doorbell_range(struct amdgpu_device *adev)
@@ -3554,6 +4065,24 @@ static void gfx_v11_0_cp_set_doorbell_range(struct amdgpu_device *adev)
 		     (adev->doorbell_index.userqueue_end * 2) << 2);
 }
 
+static void gfx_v11_0_gfx_mqd_set_priority(struct amdgpu_device *adev,
+					   struct v11_gfx_mqd *mqd,
+					   struct amdgpu_mqd_prop *prop)
+{
+	bool priority = 0;
+	u32 tmp;
+
+	/* set up default queue priority level
+	 * 0x0 = low priority, 0x1 = high priority
+	 */
+	if (prop->hqd_pipe_priority == AMDGPU_GFX_PIPE_PRIO_HIGH)
+		priority = 1;
+
+	tmp = regCP_GFX_HQD_QUEUE_PRIORITY_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_QUEUE_PRIORITY, PRIORITY_LEVEL, priority);
+	mqd->cp_gfx_hqd_queue_priority = tmp;
+}
+
 static int gfx_v11_0_gfx_mqd_init(struct amdgpu_device *adev, void *m,
 				  struct amdgpu_mqd_prop *prop)
 {
@@ -3571,25 +4100,22 @@ static int gfx_v11_0_gfx_mqd_init(struct amdgpu_device *adev, void *m,
 	mqd->cp_mqd_base_addr_hi = upper_32_bits(prop->mqd_gpu_addr);
 
 	/* set up mqd control */
-	tmp = RREG32_SOC15(GC, 0, regCP_GFX_MQD_CONTROL);
+	tmp = regCP_GFX_MQD_CONTROL_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, VMID, 0);
 	tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, PRIV_STATE, 1);
 	tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, CACHE_POLICY, 0);
 	mqd->cp_gfx_mqd_control = tmp;
 
 	/* set up gfx_hqd_vimd with 0x0 to indicate the ring buffer's vmid */
-	tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_VMID);
+	tmp = regCP_GFX_HQD_VMID_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_VMID, VMID, 0);
 	mqd->cp_gfx_hqd_vmid = 0;
 
-	/* set up default queue priority level
-	 * 0x0 = low priority, 0x1 = high priority */
-	tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_QUEUE_PRIORITY);
-	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_QUEUE_PRIORITY, PRIORITY_LEVEL, 0);
-	mqd->cp_gfx_hqd_queue_priority = tmp;
+	/* set up gfx queue priority */
+	gfx_v11_0_gfx_mqd_set_priority(adev, mqd, prop);
 
 	/* set up time quantum */
-	tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_QUANTUM);
+	tmp = regCP_GFX_HQD_QUANTUM_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_QUANTUM, QUANTUM_EN, 1);
 	mqd->cp_gfx_hqd_quantum = tmp;
 
@@ -3611,16 +4137,20 @@ static int gfx_v11_0_gfx_mqd_init(struct amdgpu_device *adev, void *m,
 
 	/* set up the gfx_hqd_control, similar as CP_RB0_CNTL */
 	rb_bufsz = order_base_2(prop->queue_size / 4) - 1;
-	tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_CNTL);
+	tmp = regCP_GFX_HQD_CNTL_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_BUFSZ, rb_bufsz);
 	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_BLKSZ, rb_bufsz - 2);
 #ifdef __BIG_ENDIAN
 	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, BUF_SWAP, 1);
 #endif
+	if (prop->tmz_queue)
+		tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, TMZ_MATCH, 1);
+	if (!prop->kernel_queue)
+		tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_NON_PRIV, 1);
 	mqd->cp_gfx_hqd_cntl = tmp;
 
 	/* set up cp_doorbell_control */
-	tmp = RREG32_SOC15(GC, 0, regCP_RB_DOORBELL_CONTROL);
+	tmp = regCP_RB_DOORBELL_CONTROL_DEFAULT;
 	if (prop->use_doorbell) {
 		tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL,
 				    DOORBELL_OFFSET, prop->doorbell_index);
@@ -3632,164 +4162,67 @@ static int gfx_v11_0_gfx_mqd_init(struct amdgpu_device *adev, void *m,
 	mqd->cp_rb_doorbell_control = tmp;
 
 	/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
-	mqd->cp_gfx_hqd_rptr = RREG32_SOC15(GC, 0, regCP_GFX_HQD_RPTR);
+	mqd->cp_gfx_hqd_rptr = regCP_GFX_HQD_RPTR_DEFAULT;
 
 	/* active the queue */
 	mqd->cp_gfx_hqd_active = 1;
 
-	return 0;
-}
-
-#ifdef BRING_UP_DEBUG
-static int gfx_v11_0_gfx_queue_init_register(struct amdgpu_ring *ring)
-{
-	struct amdgpu_device *adev = ring->adev;
-	struct v11_gfx_mqd *mqd = ring->mqd_ptr;
-
-	/* set mmCP_GFX_HQD_WPTR/_HI to 0 */
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_WPTR, mqd->cp_gfx_hqd_wptr);
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_WPTR_HI, mqd->cp_gfx_hqd_wptr_hi);
-
-	/* set GFX_MQD_BASE */
-	WREG32_SOC15(GC, 0, regCP_MQD_BASE_ADDR, mqd->cp_mqd_base_addr);
-	WREG32_SOC15(GC, 0, regCP_MQD_BASE_ADDR_HI, mqd->cp_mqd_base_addr_hi);
-
-	/* set GFX_MQD_CONTROL */
-	WREG32_SOC15(GC, 0, regCP_GFX_MQD_CONTROL, mqd->cp_gfx_mqd_control);
-
-	/* set GFX_HQD_VMID to 0 */
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_VMID, mqd->cp_gfx_hqd_vmid);
-
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_QUEUE_PRIORITY,
-			mqd->cp_gfx_hqd_queue_priority);
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_QUANTUM, mqd->cp_gfx_hqd_quantum);
-
-	/* set GFX_HQD_BASE, similar as CP_RB_BASE */
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_BASE, mqd->cp_gfx_hqd_base);
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_BASE_HI, mqd->cp_gfx_hqd_base_hi);
-
-	/* set GFX_HQD_RPTR_ADDR, similar as CP_RB_RPTR */
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_RPTR_ADDR, mqd->cp_gfx_hqd_rptr_addr);
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_RPTR_ADDR_HI, mqd->cp_gfx_hqd_rptr_addr_hi);
-
-	/* set GFX_HQD_CNTL, similar as CP_RB_CNTL */
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_CNTL, mqd->cp_gfx_hqd_cntl);
-
-	/* set RB_WPTR_POLL_ADDR */
-	WREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_ADDR_LO, mqd->cp_rb_wptr_poll_addr_lo);
-	WREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_ADDR_HI, mqd->cp_rb_wptr_poll_addr_hi);
-
-	/* set RB_DOORBELL_CONTROL */
-	WREG32_SOC15(GC, 0, regCP_RB_DOORBELL_CONTROL, mqd->cp_rb_doorbell_control);
-
-	/* active the queue */
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_ACTIVE, mqd->cp_gfx_hqd_active);
+	/* set gfx UQ items */
+	mqd->shadow_base_lo = lower_32_bits(prop->shadow_addr);
+	mqd->shadow_base_hi = upper_32_bits(prop->shadow_addr);
+	mqd->gds_bkup_base_lo = lower_32_bits(prop->gds_bkup_addr);
+	mqd->gds_bkup_base_hi = upper_32_bits(prop->gds_bkup_addr);
+	mqd->fw_work_area_base_lo = lower_32_bits(prop->csa_addr);
+	mqd->fw_work_area_base_hi = upper_32_bits(prop->csa_addr);
+	mqd->fence_address_lo = lower_32_bits(prop->fence_address);
+	mqd->fence_address_hi = upper_32_bits(prop->fence_address);
 
 	return 0;
 }
-#endif
 
-static int gfx_v11_0_gfx_init_queue(struct amdgpu_ring *ring)
+static int gfx_v11_0_kgq_init_queue(struct amdgpu_ring *ring, bool reset)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct v11_gfx_mqd *mqd = ring->mqd_ptr;
 	int mqd_idx = ring - &adev->gfx.gfx_ring[0];
 
-	if (!amdgpu_in_reset(adev) && !adev->in_suspend) {
+	if (!reset && !amdgpu_in_reset(adev) && !adev->in_suspend) {
 		memset((void *)mqd, 0, sizeof(*mqd));
 		mutex_lock(&adev->srbm_mutex);
 		soc21_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
 		amdgpu_ring_init_mqd(ring);
-#ifdef BRING_UP_DEBUG
-		gfx_v11_0_gfx_queue_init_register(ring);
-#endif
 		soc21_grbm_select(adev, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 		if (adev->gfx.me.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.me.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
-	} else if (amdgpu_in_reset(adev)) {
-		/* reset mqd with the backup copy */
+			memcpy_fromio(adev->gfx.me.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+	} else {
+		/* restore mqd with the backup copy */
 		if (adev->gfx.me.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.me.mqd_backup[mqd_idx], sizeof(*mqd));
+			memcpy_toio(mqd, adev->gfx.me.mqd_backup[mqd_idx], sizeof(*mqd));
 		/* reset the ring */
 		ring->wptr = 0;
 		*ring->wptr_cpu_addr = 0;
 		amdgpu_ring_clear_ring(ring);
-#ifdef BRING_UP_DEBUG
-		mutex_lock(&adev->srbm_mutex);
-		soc21_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
-		gfx_v11_0_gfx_queue_init_register(ring);
-		soc21_grbm_select(adev, 0, 0, 0, 0);
-		mutex_unlock(&adev->srbm_mutex);
-#endif
-	} else {
-		amdgpu_ring_clear_ring(ring);
 	}
 
 	return 0;
 }
 
-#ifndef BRING_UP_DEBUG
-static int gfx_v11_0_kiq_enable_kgq(struct amdgpu_device *adev)
-{
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
-	int r, i;
-
-	if (!kiq->pmf || !kiq->pmf->kiq_map_queues)
-		return -EINVAL;
-
-	r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size *
-					adev->gfx.num_gfx_rings);
-	if (r) {
-		DRM_ERROR("Failed to lock KIQ (%d).\n", r);
-		return r;
-	}
-
-	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
-		kiq->pmf->kiq_map_queues(kiq_ring, &adev->gfx.gfx_ring[i]);
-
-	return amdgpu_ring_test_helper(kiq_ring);
-}
-#endif
-
 static int gfx_v11_0_cp_async_gfx_ring_resume(struct amdgpu_device *adev)
 {
 	int r, i;
-	struct amdgpu_ring *ring;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
-		ring = &adev->gfx.gfx_ring[i];
-
-		r = amdgpu_bo_reserve(ring->mqd_obj, false);
-		if (unlikely(r != 0))
-			goto done;
-
-		r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr);
-		if (!r) {
-			r = gfx_v11_0_gfx_init_queue(ring);
-			amdgpu_bo_kunmap(ring->mqd_obj);
-			ring->mqd_ptr = NULL;
-		}
-		amdgpu_bo_unreserve(ring->mqd_obj);
+		r = gfx_v11_0_kgq_init_queue(&adev->gfx.gfx_ring[i], false);
 		if (r)
-			goto done;
+			return r;
 	}
-#ifndef BRING_UP_DEBUG
-	r = gfx_v11_0_kiq_enable_kgq(adev);
-	if (r)
-		goto done;
-#endif
-	r = gfx_v11_0_cp_gfx_start(adev);
+
+	r = amdgpu_gfx_enable_kgq(adev, 0);
 	if (r)
-		goto done;
+		return r;
 
-	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
-		ring = &adev->gfx.gfx_ring[i];
-		ring->sched.ready = true;
-	}
-done:
-	return r;
+	return gfx_v11_0_cp_gfx_start(adev);
 }
 
 static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
@@ -3812,14 +4245,14 @@ static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
 	mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr);
 
 	/* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */
-	tmp = RREG32_SOC15(GC, 0, regCP_HQD_EOP_CONTROL);
+	tmp = regCP_HQD_EOP_CONTROL_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, CP_HQD_EOP_CONTROL, EOP_SIZE,
 			(order_base_2(GFX11_MEC_HPD_SIZE / 4) - 1));
 
 	mqd->cp_hqd_eop_control = tmp;
 
 	/* enable doorbell? */
-	tmp = RREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL);
+	tmp = regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT;
 
 	if (prop->use_doorbell) {
 		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
@@ -3848,7 +4281,7 @@ static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
 	mqd->cp_mqd_base_addr_hi = upper_32_bits(prop->mqd_gpu_addr);
 
 	/* set MQD vmid to 0 */
-	tmp = RREG32_SOC15(GC, 0, regCP_MQD_CONTROL);
+	tmp = regCP_MQD_CONTROL_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, CP_MQD_CONTROL, VMID, 0);
 	mqd->cp_mqd_control = tmp;
 
@@ -3858,15 +4291,20 @@ static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
 	mqd->cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr);
 
 	/* set up the HQD, this is similar to CP_RB0_CNTL */
-	tmp = RREG32_SOC15(GC, 0, regCP_HQD_PQ_CONTROL);
+	tmp = regCP_HQD_PQ_CONTROL_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE,
 			    (order_base_2(prop->queue_size / 4) - 1));
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE,
 			    (order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1));
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 0);
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TUNNEL_DISPATCH, 0);
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, PRIV_STATE, 1);
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, KMD_QUEUE, 1);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 1);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TUNNEL_DISPATCH,
+			    prop->allow_tunneling);
+	if (prop->kernel_queue) {
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, PRIV_STATE, 1);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, KMD_QUEUE, 1);
+	}
+	if (prop->tmz_queue)
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TMZ, 1);
 	mqd->cp_hqd_pq_control = tmp;
 
 	/* set the wb address whether it's enabled or not */
@@ -3883,7 +4321,7 @@ static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
 	tmp = 0;
 	/* enable the doorbell if requested */
 	if (prop->use_doorbell) {
-		tmp = RREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL);
+		tmp = regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT;
 		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
 				DOORBELL_OFFSET, prop->doorbell_index);
 
@@ -3898,17 +4336,17 @@ static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
 	mqd->cp_hqd_pq_doorbell_control = tmp;
 
 	/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
-	mqd->cp_hqd_pq_rptr = RREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR);
+	mqd->cp_hqd_pq_rptr = regCP_HQD_PQ_RPTR_DEFAULT;
 
 	/* set the vmid for the queue */
 	mqd->cp_hqd_vmid = 0;
 
-	tmp = RREG32_SOC15(GC, 0, regCP_HQD_PERSISTENT_STATE);
+	tmp = regCP_HQD_PERSISTENT_STATE_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PERSISTENT_STATE, PRELOAD_SIZE, 0x55);
 	mqd->cp_hqd_persistent_state = tmp;
 
 	/* set MIN_IB_AVAIL_SIZE */
-	tmp = RREG32_SOC15(GC, 0, regCP_HQD_IB_CONTROL);
+	tmp = regCP_HQD_IB_CONTROL_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, CP_HQD_IB_CONTROL, MIN_IB_AVAIL_SIZE, 3);
 	mqd->cp_hqd_ib_control = tmp;
 
@@ -3918,6 +4356,10 @@ static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
 
 	mqd->cp_hqd_active = prop->hqd_active;
 
+	/* set UQ fenceaddress */
+	mqd->fence_address_lo = lower_32_bits(prop->fence_address);
+	mqd->fence_address_hi = upper_32_bits(prop->fence_address);
+
 	return 0;
 }
 
@@ -4035,14 +4477,13 @@ static int gfx_v11_0_kiq_init_queue(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct v11_compute_mqd *mqd = ring->mqd_ptr;
-	int mqd_idx = AMDGPU_MAX_COMPUTE_RINGS;
 
 	gfx_v11_0_kiq_setting(ring);
 
 	if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */
 		/* reset MQD to a clean status */
-		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
+		if (adev->gfx.kiq[0].mqd_backup)
+			memcpy_toio(mqd, adev->gfx.kiq[0].mqd_backup, sizeof(*mqd));
 
 		/* reset ring buffer */
 		ring->wptr = 0;
@@ -4064,20 +4505,20 @@ static int gfx_v11_0_kiq_init_queue(struct amdgpu_ring *ring)
 		soc21_grbm_select(adev, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 
-		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+		if (adev->gfx.kiq[0].mqd_backup)
+			memcpy_fromio(adev->gfx.kiq[0].mqd_backup, mqd, sizeof(*mqd));
 	}
 
 	return 0;
 }
 
-static int gfx_v11_0_kcq_init_queue(struct amdgpu_ring *ring)
+static int gfx_v11_0_kcq_init_queue(struct amdgpu_ring *ring, bool reset)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct v11_compute_mqd *mqd = ring->mqd_ptr;
 	int mqd_idx = ring - &adev->gfx.compute_ring[0];
 
-	if (!amdgpu_in_reset(adev) && !adev->in_suspend) {
+	if (!reset && !amdgpu_in_reset(adev) && !adev->in_suspend) {
 		memset((void *)mqd, 0, sizeof(*mqd));
 		mutex_lock(&adev->srbm_mutex);
 		soc21_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
@@ -4086,18 +4527,15 @@ static int gfx_v11_0_kcq_init_queue(struct amdgpu_ring *ring)
 		mutex_unlock(&adev->srbm_mutex);
 
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
-	} else if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */
-		/* reset MQD to a clean status */
+			memcpy_fromio(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+	} else {
+		/* restore MQD to a clean status */
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
-
+			memcpy_toio(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
 		/* reset ring buffer */
 		ring->wptr = 0;
 		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, 0);
 		amdgpu_ring_clear_ring(ring);
-	} else {
-		amdgpu_ring_clear_ring(ring);
 	}
 
 	return 0;
@@ -4105,57 +4543,24 @@ static int gfx_v11_0_kcq_init_queue(struct amdgpu_ring *ring)
 
 static int gfx_v11_0_kiq_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring;
-	int r;
-
-	ring = &adev->gfx.kiq.ring;
-
-	r = amdgpu_bo_reserve(ring->mqd_obj, false);
-	if (unlikely(r != 0))
-		return r;
-
-	r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr);
-	if (unlikely(r != 0)) {
-		amdgpu_bo_unreserve(ring->mqd_obj);
-		return r;
-	}
-
-	gfx_v11_0_kiq_init_queue(ring);
-	amdgpu_bo_kunmap(ring->mqd_obj);
-	ring->mqd_ptr = NULL;
-	amdgpu_bo_unreserve(ring->mqd_obj);
-	ring->sched.ready = true;
+	gfx_v11_0_kiq_init_queue(&adev->gfx.kiq[0].ring);
 	return 0;
 }
 
 static int gfx_v11_0_kcq_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring = NULL;
-	int r = 0, i;
+	int i, r;
 
 	if (!amdgpu_async_gfx_ring)
 		gfx_v11_0_cp_compute_enable(adev, true);
 
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-		ring = &adev->gfx.compute_ring[i];
-
-		r = amdgpu_bo_reserve(ring->mqd_obj, false);
-		if (unlikely(r != 0))
-			goto done;
-		r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr);
-		if (!r) {
-			r = gfx_v11_0_kcq_init_queue(ring);
-			amdgpu_bo_kunmap(ring->mqd_obj);
-			ring->mqd_ptr = NULL;
-		}
-		amdgpu_bo_unreserve(ring->mqd_obj);
+		r = gfx_v11_0_kcq_init_queue(&adev->gfx.compute_ring[i], false);
 		if (r)
-			goto done;
+			return r;
 	}
 
-	r = amdgpu_gfx_enable_kcq(adev);
-done:
-	return r;
+	return amdgpu_gfx_enable_kcq(adev, 0);
 }
 
 static int gfx_v11_0_cp_resume(struct amdgpu_device *adev)
@@ -4208,11 +4613,23 @@ static int gfx_v11_0_cp_resume(struct amdgpu_device *adev)
 			return r;
 	}
 
-	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
-		ring = &adev->gfx.gfx_ring[i];
-		r = amdgpu_ring_test_helper(ring);
-		if (r)
-			return r;
+	if (adev->gfx.disable_kq) {
+		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+			ring = &adev->gfx.gfx_ring[i];
+			/* we don't want to set ring->ready */
+			r = amdgpu_ring_test_ring(ring);
+			if (r)
+				return r;
+		}
+		if (amdgpu_async_gfx_ring)
+			amdgpu_gfx_disable_kgq(adev, 0);
+	} else {
+		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+			ring = &adev->gfx.gfx_ring[i];
+			r = amdgpu_ring_test_helper(ring);
+			if (r)
+				return r;
+		}
 	}
 
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
@@ -4240,13 +4657,14 @@ static int gfx_v11_0_gfxhub_enable(struct amdgpu_device *adev)
 	if (r)
 		return r;
 
-	adev->hdp.funcs->flush_hdp(adev, NULL);
+	amdgpu_device_flush_hdp(adev, NULL);
 
-	value = (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS) ?
-		false : true;
+	value = amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS;
 
 	adev->gfxhub.funcs->set_fault_enable_default(adev, value);
-	amdgpu_gmc_flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB_0, 0);
+	/* TODO investigate why this and the hdp flush above is needed,
+	 * are we missing a flush somewhere else? */
+	adev->gmc.gmc_funcs->flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB(0), 0);
 
 	return 0;
 }
@@ -4319,21 +4737,24 @@ static void gfx_v11_0_disable_gpa_mode(struct amdgpu_device *adev)
 	WREG32_SOC15(GC, 0, regCPG_PSP_DEBUG, data);
 }
 
-static int gfx_v11_0_hw_init(void *handle)
+static int gfx_v11_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	amdgpu_gfx_cleaner_shader_init(adev, adev->gfx.cleaner_shader_size,
+				       adev->gfx.cleaner_shader_ptr);
 
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) {
 		if (adev->gfx.imu.funcs) {
 			/* RLC autoload sequence 1: Program rlc ram */
 			if (adev->gfx.imu.funcs->program_rlc_ram)
 				adev->gfx.imu.funcs->program_rlc_ram(adev);
+			/* rlc autoload firmware */
+			r = gfx_v11_0_rlc_backdoor_autoload_enable(adev);
+			if (r)
+				return r;
 		}
-		/* rlc autoload firmware */
-		r = gfx_v11_0_rlc_backdoor_autoload_enable(adev);
-		if (r)
-			return r;
 	} else {
 		if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
 			if (adev->gfx.imu.funcs && (amdgpu_dpm > 0)) {
@@ -4382,11 +4803,9 @@ static int gfx_v11_0_hw_init(void *handle)
 		 * loaded firstly, so in direct type, it has to load smc ucode
 		 * here before rlc.
 		 */
-		if (!(adev->flags & AMD_IS_APU)) {
-			r = amdgpu_pm_load_smu_firmware(adev, NULL);
-			if (r)
-				return r;
-		}
+		r = amdgpu_pm_load_smu_firmware(adev, NULL);
+		if (r)
+			return r;
 	}
 
 	gfx_v11_0_constants_init(adev);
@@ -4411,52 +4830,75 @@ static int gfx_v11_0_hw_init(void *handle)
 	if (r)
 		return r;
 
+	/* get IMU version from HW if it's not set */
+	if (!adev->gfx.imu_fw_version)
+		adev->gfx.imu_fw_version = RREG32_SOC15(GC, 0, regGFX_IMU_SCRATCH_0);
+
 	return r;
 }
 
-#ifndef BRING_UP_DEBUG
-static int gfx_v11_0_kiq_disable_kgq(struct amdgpu_device *adev)
+static int gfx_v11_0_set_userq_eop_interrupts(struct amdgpu_device *adev,
+					      bool enable)
 {
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-	struct amdgpu_ring *kiq_ring = &kiq->ring;
-	int i, r = 0;
-
-	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
-		return -EINVAL;
-
-	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size *
-					adev->gfx.num_gfx_rings))
-		return -ENOMEM;
-
-	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
-		kiq->pmf->kiq_unmap_queues(kiq_ring, &adev->gfx.gfx_ring[i],
-					   PREEMPT_QUEUES, 0, 0);
+	unsigned int irq_type;
+	int m, p, r;
+
+	if (adev->userq_funcs[AMDGPU_HW_IP_GFX]) {
+		for (m = 0; m < adev->gfx.me.num_me; m++) {
+			for (p = 0; p < adev->gfx.me.num_pipe_per_me; p++) {
+				irq_type = AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP + p;
+				if (enable)
+					r = amdgpu_irq_get(adev, &adev->gfx.eop_irq,
+							   irq_type);
+				else
+					r = amdgpu_irq_put(adev, &adev->gfx.eop_irq,
+							   irq_type);
+				if (r)
+					return r;
+			}
+		}
+	}
 
-	if (adev->gfx.kiq.ring.sched.ready)
-		r = amdgpu_ring_test_helper(kiq_ring);
+	if (adev->userq_funcs[AMDGPU_HW_IP_COMPUTE]) {
+		for (m = 0; m < adev->gfx.mec.num_mec; ++m) {
+			for (p = 0; p < adev->gfx.mec.num_pipe_per_mec; p++) {
+				irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
+					+ (m * adev->gfx.mec.num_pipe_per_mec)
+					+ p;
+				if (enable)
+					r = amdgpu_irq_get(adev, &adev->gfx.eop_irq,
+							   irq_type);
+				else
+					r = amdgpu_irq_put(adev, &adev->gfx.eop_irq,
+							   irq_type);
+				if (r)
+					return r;
+			}
+		}
+	}
 
-	return r;
+	return 0;
 }
-#endif
 
-static int gfx_v11_0_hw_fini(void *handle)
+static int gfx_v11_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	cancel_delayed_work_sync(&adev->gfx.idle_work);
 
-	amdgpu_irq_put(adev, &adev->gfx.cp_ecc_error_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
+	amdgpu_irq_put(adev, &adev->gfx.bad_op_irq, 0);
+	gfx_v11_0_set_userq_eop_interrupts(adev, false);
 
 	if (!adev->no_hw_access) {
-#ifndef BRING_UP_DEBUG
-		if (amdgpu_async_gfx_ring) {
-			r = gfx_v11_0_kiq_disable_kgq(adev);
-			if (r)
+		if (amdgpu_async_gfx_ring &&
+		    !adev->gfx.disable_kq) {
+			if (amdgpu_gfx_disable_kgq(adev, 0))
 				DRM_ERROR("KGQ disable failed\n");
 		}
-#endif
-		if (amdgpu_gfx_disable_kcq(adev))
+
+		if (amdgpu_gfx_disable_kcq(adev, 0))
 			DRM_ERROR("KCQ disable failed\n");
 
 		amdgpu_mes_kiq_hw_fini(adev);
@@ -4480,19 +4922,19 @@ static int gfx_v11_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int gfx_v11_0_suspend(void *handle)
+static int gfx_v11_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	return gfx_v11_0_hw_fini(handle);
+	return gfx_v11_0_hw_fini(ip_block);
 }
 
-static int gfx_v11_0_resume(void *handle)
+static int gfx_v11_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	return gfx_v11_0_hw_init(handle);
+	return gfx_v11_0_hw_init(ip_block);
 }
 
-static bool gfx_v11_0_is_idle(void *handle)
+static bool gfx_v11_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (REG_GET_FIELD(RREG32_SOC15(GC, 0, regGRBM_STATUS),
 				GRBM_STATUS, GUI_ACTIVE))
@@ -4501,11 +4943,11 @@ static bool gfx_v11_0_is_idle(void *handle)
 		return true;
 }
 
-static int gfx_v11_0_wait_for_idle(void *handle)
+static int gfx_v11_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		/* read MC_STATUS */
@@ -4519,12 +4961,46 @@ static int gfx_v11_0_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int gfx_v11_0_soft_reset(void *handle)
+int gfx_v11_0_request_gfx_index_mutex(struct amdgpu_device *adev,
+				      bool req)
+{
+	u32 i, tmp, val;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		/* Request with MeId=2, PipeId=0 */
+		tmp = REG_SET_FIELD(0, CP_GFX_INDEX_MUTEX, REQUEST, req);
+		tmp = REG_SET_FIELD(tmp, CP_GFX_INDEX_MUTEX, CLIENTID, 4);
+		WREG32_SOC15(GC, 0, regCP_GFX_INDEX_MUTEX, tmp);
+
+		val = RREG32_SOC15(GC, 0, regCP_GFX_INDEX_MUTEX);
+		if (req) {
+			if (val == tmp)
+				break;
+		} else {
+			tmp = REG_SET_FIELD(tmp, CP_GFX_INDEX_MUTEX,
+					    REQUEST, 1);
+
+			/* unlocked or locked by firmware */
+			if (val != tmp)
+				break;
+		}
+		udelay(1);
+	}
+
+	if (i >= adev->usec_timeout)
+		return -EINVAL;
+
+	return 0;
+}
+
+static int gfx_v11_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 grbm_soft_reset = 0;
 	u32 tmp;
-	int i, j, k;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	int r, i, j, k;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 
 	tmp = RREG32_SOC15(GC, 0, regCP_INT_CNTL);
 	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL, CMP_BUSY_INT_ENABLE, 0);
@@ -4533,16 +5009,11 @@ static int gfx_v11_0_soft_reset(void *handle)
 	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL, GFX_IDLE_INT_ENABLE, 0);
 	WREG32_SOC15(GC, 0, regCP_INT_CNTL, tmp);
 
-	gfx_v11_0_set_safe_mode(adev);
-
+	mutex_lock(&adev->srbm_mutex);
 	for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
 		for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
 			for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
-				tmp = RREG32_SOC15(GC, 0, regGRBM_GFX_CNTL);
-				tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, MEID, i);
-				tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, QUEUEID, j);
-				tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, PIPEID, k);
-				WREG32_SOC15(GC, 0, regGRBM_GFX_CNTL, tmp);
+				soc21_grbm_select(adev, i, k, j, 0);
 
 				WREG32_SOC15(GC, 0, regCP_HQD_DEQUEUE_REQUEST, 0x2);
 				WREG32_SOC15(GC, 0, regSPI_COMPUTE_QUEUE_RESET, 0x1);
@@ -4552,16 +5023,23 @@ static int gfx_v11_0_soft_reset(void *handle)
 	for (i = 0; i < adev->gfx.me.num_me; ++i) {
 		for (j = 0; j < adev->gfx.me.num_queue_per_pipe; j++) {
 			for (k = 0; k < adev->gfx.me.num_pipe_per_me; k++) {
-				tmp = RREG32_SOC15(GC, 0, regGRBM_GFX_CNTL);
-				tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, MEID, i);
-				tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, QUEUEID, j);
-				tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, PIPEID, k);
-				WREG32_SOC15(GC, 0, regGRBM_GFX_CNTL, tmp);
+				soc21_grbm_select(adev, i, k, j, 0);
 
 				WREG32_SOC15(GC, 0, regCP_GFX_HQD_DEQUEUE_REQUEST, 0x1);
 			}
 		}
 	}
+	soc21_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+
+	/* Try to acquire the gfx mutex before access to CP_VMID_RESET */
+	mutex_lock(&adev->gfx.reset_sem_mutex);
+	r = gfx_v11_0_request_gfx_index_mutex(adev, true);
+	if (r) {
+		mutex_unlock(&adev->gfx.reset_sem_mutex);
+		DRM_ERROR("Failed to acquire the gfx mutex during soft reset\n");
+		return r;
+	}
 
 	WREG32_SOC15(GC, 0, regCP_VMID_RESET, 0xfffffffe);
 
@@ -4571,6 +5049,14 @@ static int gfx_v11_0_soft_reset(void *handle)
 	RREG32_SOC15(GC, 0, regCP_VMID_RESET);
 	RREG32_SOC15(GC, 0, regCP_VMID_RESET);
 
+	/* release the gfx mutex */
+	r = gfx_v11_0_request_gfx_index_mutex(adev, false);
+	mutex_unlock(&adev->gfx.reset_sem_mutex);
+	if (r) {
+		DRM_ERROR("Failed to release the gfx mutex during soft reset\n");
+		return r;
+	}
+
 	for (i = 0; i < adev->usec_timeout; i++) {
 		if (!RREG32_SOC15(GC, 0, regCP_HQD_ACTIVE) &&
 		    !RREG32_SOC15(GC, 0, regCP_GFX_HQD_ACTIVE))
@@ -4633,15 +5119,15 @@ static int gfx_v11_0_soft_reset(void *handle)
 	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL, GFX_IDLE_INT_ENABLE, 1);
 	WREG32_SOC15(GC, 0, regCP_INT_CNTL, tmp);
 
-	gfx_v11_0_unset_safe_mode(adev);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 
 	return gfx_v11_0_cp_resume(adev);
 }
 
-static bool gfx_v11_0_check_soft_reset(void *handle)
+static bool gfx_v11_0_check_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	int i, r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring;
 	long tmo = msecs_to_jiffies(1000);
 
@@ -4662,12 +5148,13 @@ static bool gfx_v11_0_check_soft_reset(void *handle)
 	return false;
 }
 
-static int gfx_v11_0_post_soft_reset(void *handle)
+static int gfx_v11_0_post_soft_reset(struct amdgpu_ip_block *ip_block)
 {
+	struct amdgpu_device *adev = ip_block->adev;
 	/**
 	 * GFX soft reset will impact MES, need resume MES when do GFX soft reset
 	 */
-	return amdgpu_mes_resume((struct amdgpu_device *)handle);
+	return amdgpu_mes_resume(adev);
 }
 
 static uint64_t gfx_v11_0_get_gpu_clock_counter(struct amdgpu_device *adev)
@@ -4675,24 +5162,27 @@ static uint64_t gfx_v11_0_get_gpu_clock_counter(struct amdgpu_device *adev)
 	uint64_t clock;
 	uint64_t clock_counter_lo, clock_counter_hi_pre, clock_counter_hi_after;
 
-	amdgpu_gfx_off_ctrl(adev, false);
-	mutex_lock(&adev->gfx.gpu_clock_mutex);
 	if (amdgpu_sriov_vf(adev)) {
+		amdgpu_gfx_off_ctrl(adev, false);
+		mutex_lock(&adev->gfx.gpu_clock_mutex);
 		clock_counter_hi_pre = (uint64_t)RREG32_SOC15(GC, 0, regCP_MES_MTIME_HI);
 		clock_counter_lo = (uint64_t)RREG32_SOC15(GC, 0, regCP_MES_MTIME_LO);
 		clock_counter_hi_after = (uint64_t)RREG32_SOC15(GC, 0, regCP_MES_MTIME_HI);
 		if (clock_counter_hi_pre != clock_counter_hi_after)
 			clock_counter_lo = (uint64_t)RREG32_SOC15(GC, 0, regCP_MES_MTIME_LO);
+		mutex_unlock(&adev->gfx.gpu_clock_mutex);
+		amdgpu_gfx_off_ctrl(adev, true);
 	} else {
+		preempt_disable();
 		clock_counter_hi_pre = (uint64_t)RREG32_SOC15(SMUIO, 0, regGOLDEN_TSC_COUNT_UPPER);
 		clock_counter_lo = (uint64_t)RREG32_SOC15(SMUIO, 0, regGOLDEN_TSC_COUNT_LOWER);
 		clock_counter_hi_after = (uint64_t)RREG32_SOC15(SMUIO, 0, regGOLDEN_TSC_COUNT_UPPER);
 		if (clock_counter_hi_pre != clock_counter_hi_after)
 			clock_counter_lo = (uint64_t)RREG32_SOC15(SMUIO, 0, regGOLDEN_TSC_COUNT_LOWER);
+		preempt_enable();
 	}
 	clock = clock_counter_lo | (clock_counter_hi_after << 32ULL);
-	mutex_unlock(&adev->gfx.gpu_clock_mutex);
-	amdgpu_gfx_off_ctrl(adev, true);
+
 	return clock;
 }
 
@@ -4725,15 +5215,40 @@ static void gfx_v11_0_ring_emit_gds_switch(struct amdgpu_ring *ring,
 				    (1 << (oa_size + oa_base)) - (1 << oa_base));
 }
 
-static int gfx_v11_0_early_init(void *handle)
+static int gfx_v11_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	switch (amdgpu_user_queue) {
+	case -1:
+	case 0:
+	default:
+		adev->gfx.disable_kq = false;
+		adev->gfx.disable_uq = true;
+		break;
+	case 1:
+		adev->gfx.disable_kq = false;
+		adev->gfx.disable_uq = false;
+		break;
+	case 2:
+		adev->gfx.disable_kq = true;
+		adev->gfx.disable_uq = false;
+		break;
+	}
 
 	adev->gfx.funcs = &gfx_v11_0_gfx_funcs;
 
-	adev->gfx.num_gfx_rings = GFX11_NUM_GFX_RINGS;
-	adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
-					  AMDGPU_MAX_COMPUTE_RINGS);
+	if (adev->gfx.disable_kq) {
+		/* We need one GFX ring temporarily to set up
+		 * the clear state.
+		 */
+		adev->gfx.num_gfx_rings = 1;
+		adev->gfx.num_compute_rings = 0;
+	} else {
+		adev->gfx.num_gfx_rings = GFX11_NUM_GFX_RINGS;
+		adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
+						  AMDGPU_MAX_COMPUTE_RINGS);
+	}
 
 	gfx_v11_0_set_kiq_pm4_funcs(adev);
 	gfx_v11_0_set_ring_funcs(adev);
@@ -4748,29 +5263,9 @@ static int gfx_v11_0_early_init(void *handle)
 	return gfx_v11_0_init_microcode(adev);
 }
 
-static int gfx_v11_0_ras_late_init(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	struct ras_common_if *gfx_common_if;
-	int ret;
-
-	gfx_common_if = kzalloc(sizeof(struct ras_common_if), GFP_KERNEL);
-	if (!gfx_common_if)
-		return -ENOMEM;
-
-	gfx_common_if->block = AMDGPU_RAS_BLOCK__GFX;
-
-	ret = amdgpu_ras_feature_enable(adev, gfx_common_if, true);
-	if (ret)
-		dev_warn(adev->dev, "Failed to enable gfx11 ras feature\n");
-
-	kfree(gfx_common_if);
-	return 0;
-}
-
-static int gfx_v11_0_late_init(void *handle)
+static int gfx_v11_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_irq_get(adev, &adev->gfx.priv_reg_irq, 0);
@@ -4781,11 +5276,13 @@ static int gfx_v11_0_late_init(void *handle)
 	if (r)
 		return r;
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(11, 0, 3)) {
-		r = gfx_v11_0_ras_late_init(handle);
-		if (r)
-			return r;
-	}
+	r = amdgpu_irq_get(adev, &adev->gfx.bad_op_irq, 0);
+	if (r)
+		return r;
+
+	r = gfx_v11_0_set_userq_eop_interrupts(adev, true);
+	if (r)
+		return r;
 
 	return 0;
 }
@@ -4799,7 +5296,7 @@ static bool gfx_v11_0_is_rlc_enabled(struct amdgpu_device *adev)
 	return (REG_GET_FIELD(rlc_cntl, RLC_CNTL, RLC_ENABLE_F32)) ? true : false;
 }
 
-static void gfx_v11_0_set_safe_mode(struct amdgpu_device *adev)
+static void gfx_v11_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	uint32_t data;
 	unsigned i;
@@ -4818,7 +5315,7 @@ static void gfx_v11_0_set_safe_mode(struct amdgpu_device *adev)
 	}
 }
 
-static void gfx_v11_0_unset_safe_mode(struct amdgpu_device *adev)
+static void gfx_v11_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	WREG32_SOC15(GC, 0, regRLC_SAFE_MODE, RLC_SAFE_MODE__CMD_MASK);
 }
@@ -5046,7 +5543,7 @@ static void gfx_v11_0_update_coarse_grain_clock_gating(struct amdgpu_device *ade
 static int gfx_v11_0_update_gfx_clock_gating(struct amdgpu_device *adev,
 					    bool enable)
 {
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 
 	gfx_v11_0_update_coarse_grain_clock_gating(adev, enable);
 
@@ -5066,32 +5563,40 @@ static int gfx_v11_0_update_gfx_clock_gating(struct amdgpu_device *adev,
 	     AMD_CG_SUPPORT_GFX_3D_CGLS))
 	        gfx_v11_0_enable_gui_idle_interrupt(adev, enable);
 
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 
 	return 0;
 }
 
-static void gfx_v11_0_update_spm_vmid(struct amdgpu_device *adev, unsigned vmid)
+static void gfx_v11_0_update_spm_vmid(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned vmid)
 {
-	u32 reg, data;
+	u32 reg, pre_data, data;
 
 	amdgpu_gfx_off_ctrl(adev, false);
-
 	reg = SOC15_REG_OFFSET(GC, 0, regRLC_SPM_MC_CNTL);
-	if (amdgpu_sriov_is_pp_one_vf(adev))
-		data = RREG32_NO_KIQ(reg);
+	if (amdgpu_sriov_is_pp_one_vf(adev) && !amdgpu_sriov_runtime(adev))
+		pre_data = RREG32_NO_KIQ(reg);
 	else
-		data = RREG32(reg);
+		pre_data = RREG32(reg);
 
-	data &= ~RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK;
+	data = pre_data & (~RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK);
 	data |= (vmid & RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK) << RLC_SPM_MC_CNTL__RLC_SPM_VMID__SHIFT;
 
-	if (amdgpu_sriov_is_pp_one_vf(adev))
-		WREG32_SOC15_NO_KIQ(GC, 0, regRLC_SPM_MC_CNTL, data);
-	else
-		WREG32_SOC15(GC, 0, regRLC_SPM_MC_CNTL, data);
-
+	if (pre_data != data) {
+		if (amdgpu_sriov_is_pp_one_vf(adev) && !amdgpu_sriov_runtime(adev)) {
+			WREG32_SOC15_NO_KIQ(GC, 0, regRLC_SPM_MC_CNTL, data);
+		} else
+			WREG32_SOC15(GC, 0, regRLC_SPM_MC_CNTL, data);
+	}
 	amdgpu_gfx_off_ctrl(adev, true);
+
+	if (ring
+		&& amdgpu_sriov_is_pp_one_vf(adev)
+		&& (pre_data != data)
+		&& ((ring->funcs->type == AMDGPU_RING_TYPE_GFX)
+			|| (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE))) {
+		amdgpu_ring_emit_wreg(ring, reg, data);
+	}
 }
 
 static const struct amdgpu_rlc_funcs gfx_v11_0_rlc_funcs = {
@@ -5121,9 +5626,13 @@ static void gfx_v11_cntl_power_gating(struct amdgpu_device *adev, bool enable)
 
 	// Program RLC_PG_DELAY3 for CGPG hysteresis
 	if (enable && (adev->pg_flags & AMD_PG_SUPPORT_GFX_PG)) {
-		switch (adev->ip_versions[GC_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 		case IP_VERSION(11, 0, 1):
 		case IP_VERSION(11, 0, 4):
+		case IP_VERSION(11, 5, 0):
+		case IP_VERSION(11, 5, 1):
+		case IP_VERSION(11, 5, 2):
+		case IP_VERSION(11, 5, 3):
 			WREG32_SOC15(GC, 0, regRLC_PG_DELAY_3, RLC_PG_DELAY_3_DEFAULT_GC_11_0_1);
 			break;
 		default:
@@ -5134,23 +5643,23 @@ static void gfx_v11_cntl_power_gating(struct amdgpu_device *adev, bool enable)
 
 static void gfx_v11_cntl_pg(struct amdgpu_device *adev, bool enable)
 {
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 
 	gfx_v11_cntl_power_gating(adev, enable);
 
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 }
 
-static int gfx_v11_0_set_powergating_state(void *handle,
+static int gfx_v11_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_PG_STATE_GATE);
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 0):
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 3):
@@ -5158,8 +5667,18 @@ static int gfx_v11_0_set_powergating_state(void *handle,
 		break;
 	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
+		if (!enable)
+			amdgpu_gfx_off_ctrl(adev, false);
+
 		gfx_v11_cntl_pg(adev, enable);
-		amdgpu_gfx_off_ctrl(adev, enable);
+
+		if (enable)
+			amdgpu_gfx_off_ctrl(adev, true);
+
 		break;
 	default:
 		break;
@@ -5168,20 +5687,24 @@ static int gfx_v11_0_set_powergating_state(void *handle,
 	return 0;
 }
 
-static int gfx_v11_0_set_clockgating_state(void *handle,
+static int gfx_v11_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 	        return 0;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 0):
 	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 3):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
 	        gfx_v11_0_update_gfx_clock_gating(adev,
 	                        state ==  AMD_CG_STATE_GATE);
 	        break;
@@ -5192,9 +5715,9 @@ static int gfx_v11_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static void gfx_v11_0_get_clockgating_state(void *handle, u64 *flags)
+static void gfx_v11_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	/* AMD_CG_SUPPORT_GFX_MGCG */
@@ -5258,45 +5781,17 @@ static u64 gfx_v11_0_ring_get_wptr_gfx(struct amdgpu_ring *ring)
 static void gfx_v11_0_ring_set_wptr_gfx(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	uint32_t *wptr_saved;
-	uint32_t *is_queue_unmap;
-	uint64_t aggregated_db_index;
-	uint32_t mqd_size = adev->mqds[AMDGPU_HW_IP_GFX].mqd_size;
-	uint64_t wptr_tmp;
-
-	if (ring->is_mes_queue) {
-		wptr_saved = (uint32_t *)(ring->mqd_ptr + mqd_size);
-		is_queue_unmap = (uint32_t *)(ring->mqd_ptr + mqd_size +
-					      sizeof(uint32_t));
-		aggregated_db_index =
-			amdgpu_mes_get_aggregated_doorbell_index(adev,
-								 ring->hw_prio);
-
-		wptr_tmp = ring->wptr & ring->buf_mask;
-		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, wptr_tmp);
-		*wptr_saved = wptr_tmp;
-		/* assume doorbell always being used by mes mapped queue */
-		if (*is_queue_unmap) {
-			WDOORBELL64(aggregated_db_index, wptr_tmp);
-			WDOORBELL64(ring->doorbell_index, wptr_tmp);
-		} else {
-			WDOORBELL64(ring->doorbell_index, wptr_tmp);
 
-			if (*is_queue_unmap)
-				WDOORBELL64(aggregated_db_index, wptr_tmp);
-		}
+	if (ring->use_doorbell) {
+		/* XXX check if swapping is necessary on BE */
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
+			     ring->wptr);
+		WDOORBELL64(ring->doorbell_index, ring->wptr);
 	} else {
-		if (ring->use_doorbell) {
-			/* XXX check if swapping is necessary on BE */
-			atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
-				     ring->wptr);
-			WDOORBELL64(ring->doorbell_index, ring->wptr);
-		} else {
-			WREG32_SOC15(GC, 0, regCP_RB0_WPTR,
-				     lower_32_bits(ring->wptr));
-			WREG32_SOC15(GC, 0, regCP_RB0_WPTR_HI,
-				     upper_32_bits(ring->wptr));
-		}
+		WREG32_SOC15(GC, 0, regCP_RB0_WPTR,
+			     lower_32_bits(ring->wptr));
+		WREG32_SOC15(GC, 0, regCP_RB0_WPTR_HI,
+			     upper_32_bits(ring->wptr));
 	}
 }
 
@@ -5321,42 +5816,14 @@ static u64 gfx_v11_0_ring_get_wptr_compute(struct amdgpu_ring *ring)
 static void gfx_v11_0_ring_set_wptr_compute(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	uint32_t *wptr_saved;
-	uint32_t *is_queue_unmap;
-	uint64_t aggregated_db_index;
-	uint32_t mqd_size = adev->mqds[AMDGPU_HW_IP_COMPUTE].mqd_size;
-	uint64_t wptr_tmp;
-
-	if (ring->is_mes_queue) {
-		wptr_saved = (uint32_t *)(ring->mqd_ptr + mqd_size);
-		is_queue_unmap = (uint32_t *)(ring->mqd_ptr + mqd_size +
-					      sizeof(uint32_t));
-		aggregated_db_index =
-			amdgpu_mes_get_aggregated_doorbell_index(adev,
-								 ring->hw_prio);
-
-		wptr_tmp = ring->wptr & ring->buf_mask;
-		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, wptr_tmp);
-		*wptr_saved = wptr_tmp;
-		/* assume doorbell always used by mes mapped queue */
-		if (*is_queue_unmap) {
-			WDOORBELL64(aggregated_db_index, wptr_tmp);
-			WDOORBELL64(ring->doorbell_index, wptr_tmp);
-		} else {
-			WDOORBELL64(ring->doorbell_index, wptr_tmp);
 
-			if (*is_queue_unmap)
-				WDOORBELL64(aggregated_db_index, wptr_tmp);
-		}
+	/* XXX check if swapping is necessary on BE */
+	if (ring->use_doorbell) {
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
+			     ring->wptr);
+		WDOORBELL64(ring->doorbell_index, ring->wptr);
 	} else {
-		/* XXX check if swapping is necessary on BE */
-		if (ring->use_doorbell) {
-			atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
-				     ring->wptr);
-			WDOORBELL64(ring->doorbell_index, ring->wptr);
-		} else {
-			BUG(); /* only DOORBELL method supported on gfx11 now */
-		}
+		BUG(); /* only DOORBELL method supported on gfx11 now */
 	}
 }
 
@@ -5379,7 +5846,7 @@ static void gfx_v11_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
 		}
 		reg_mem_engine = 0;
 	} else {
-		ref_and_mask = nbio_hf_reg->ref_and_mask_cp0;
+		ref_and_mask = nbio_hf_reg->ref_and_mask_cp0 << ring->pipe;
 		reg_mem_engine = 1; /* pfp */
 	}
 
@@ -5397,27 +5864,21 @@ static void gfx_v11_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
 	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
 	u32 header, control = 0;
 
-	BUG_ON(ib->flags & AMDGPU_IB_FLAG_CE);
-
 	header = PACKET3(PACKET3_INDIRECT_BUFFER, 2);
 
 	control |= ib->length_dw | (vmid << 24);
 
-	if (amdgpu_mcbp && (ib->flags & AMDGPU_IB_FLAG_PREEMPT)) {
+	if (ring->adev->gfx.mcbp && (ib->flags & AMDGPU_IB_FLAG_PREEMPT)) {
 		control |= INDIRECT_BUFFER_PRE_ENB(1);
 
 		if (flags & AMDGPU_IB_PREEMPTED)
 			control |= INDIRECT_BUFFER_PRE_RESUME(1);
 
-		if (vmid)
+		if (vmid && !ring->adev->gfx.rs64_enable)
 			gfx_v11_0_ring_emit_de_meta(ring,
-				    (!amdgpu_sriov_vf(ring->adev) && flags & AMDGPU_IB_PREEMPTED) ? true : false);
+				!amdgpu_sriov_vf(ring->adev) && (flags & AMDGPU_IB_PREEMPTED));
 	}
 
-	if (ring->is_mes_queue)
-		/* inherit vmid from mqd */
-		control |= 0x400000;
-
 	amdgpu_ring_write(ring, header);
 	BUG_ON(ib->gpu_addr & 0x3); /* Dword align */
 	amdgpu_ring_write(ring,
@@ -5437,10 +5898,6 @@ static void gfx_v11_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
 	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
 	u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vmid << 24);
 
-	if (ring->is_mes_queue)
-		/* inherit vmid from mqd */
-		control |= 0x40000000;
-
 	/* Currently, there is a high possibility to get wave ID mismatch
 	 * between ME and GDS, leading to a hw deadlock, because ME generates
 	 * different wave IDs than the GDS expects. This situation happens
@@ -5478,11 +5935,7 @@ static void gfx_v11_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr,
 	amdgpu_ring_write(ring, PACKET3(PACKET3_RELEASE_MEM, 6));
 	amdgpu_ring_write(ring, (PACKET3_RELEASE_MEM_GCR_SEQ |
 				 PACKET3_RELEASE_MEM_GCR_GL2_WB |
-				 PACKET3_RELEASE_MEM_GCR_GL2_INV |
-				 PACKET3_RELEASE_MEM_GCR_GL2_US |
-				 PACKET3_RELEASE_MEM_GCR_GL1_INV |
-				 PACKET3_RELEASE_MEM_GCR_GLV_INV |
-				 PACKET3_RELEASE_MEM_GCR_GLM_INV |
+				 PACKET3_RELEASE_MEM_GCR_GLM_INV | /* must be set with GLM_WB */
 				 PACKET3_RELEASE_MEM_GCR_GLM_WB |
 				 PACKET3_RELEASE_MEM_CACHE_POLICY(3) |
 				 PACKET3_RELEASE_MEM_EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) |
@@ -5502,8 +5955,7 @@ static void gfx_v11_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr,
 	amdgpu_ring_write(ring, upper_32_bits(addr));
 	amdgpu_ring_write(ring, lower_32_bits(seq));
 	amdgpu_ring_write(ring, upper_32_bits(seq));
-	amdgpu_ring_write(ring, ring->is_mes_queue ?
-			 (ring->hw_queue_id | AMDGPU_FENCE_MES_QUEUE_FLAG) : 0);
+	amdgpu_ring_write(ring, 0);
 }
 
 static void gfx_v11_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
@@ -5531,10 +5983,7 @@ static void gfx_v11_0_ring_invalidate_tlbs(struct amdgpu_ring *ring,
 static void gfx_v11_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
 					 unsigned vmid, uint64_t pd_addr)
 {
-	if (ring->is_mes_queue)
-		gfx_v11_0_ring_invalidate_tlbs(ring, 0, 0, false, 0);
-	else
-		amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
+	amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
 
 	/* compute doesn't have PFP */
 	if (ring->funcs->type == AMDGPU_RING_TYPE_GFX) {
@@ -5542,6 +5991,12 @@ static void gfx_v11_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
 		amdgpu_ring_write(ring, PACKET3(PACKET3_PFP_SYNC_ME, 0));
 		amdgpu_ring_write(ring, 0x0);
 	}
+
+	/* Make sure that we can't skip the SET_Q_MODE packets when the VM
+	 * changed in any way.
+	 */
+	ring->set_q_mode_offs = 0;
+	ring->set_q_mode_ptr = NULL;
 }
 
 static void gfx_v11_0_ring_emit_fence_kiq(struct amdgpu_ring *ring, u64 addr,
@@ -5591,41 +6046,125 @@ static void gfx_v11_0_ring_emit_cntxcntl(struct amdgpu_ring *ring,
 	amdgpu_ring_write(ring, 0);
 }
 
-static unsigned gfx_v11_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring)
+static unsigned gfx_v11_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring,
+						   uint64_t addr)
 {
 	unsigned ret;
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, 0); /* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	/* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, 0);
 	ret = ring->wptr & ring->buf_mask;
-	amdgpu_ring_write(ring, 0x55aa55aa); /* patch dummy value later */
+	/* patch dummy value later */
+	amdgpu_ring_write(ring, 0);
 
 	return ret;
 }
 
-static void gfx_v11_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigned offset)
+static void gfx_v11_0_ring_emit_gfx_shadow(struct amdgpu_ring *ring,
+					   u64 shadow_va, u64 csa_va,
+					   u64 gds_va, bool init_shadow,
+					   int vmid)
 {
-	unsigned cur;
-	BUG_ON(offset > ring->buf_mask);
-	BUG_ON(ring->ring[offset] != 0x55aa55aa);
+	struct amdgpu_device *adev = ring->adev;
+	unsigned int offs, end;
 
-	cur = (ring->wptr - 1) & ring->buf_mask;
-	if (likely(cur > offset))
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->buf_mask + 1) - offset + cur;
+	if (!adev->gfx.cp_gfx_shadow || !ring->ring_obj)
+		return;
+
+	/*
+	 * The logic here isn't easy to understand because we need to keep state
+	 * accross multiple executions of the function as well as between the
+	 * CPU and GPU. The general idea is that the newly written GPU command
+	 * has a condition on the previous one and only executed if really
+	 * necessary.
+	 */
+
+	/*
+	 * The dw in the NOP controls if the next SET_Q_MODE packet should be
+	 * executed or not. Reserve 64bits just to be on the save side.
+	 */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_NOP, 1));
+	offs = ring->wptr & ring->buf_mask;
+
+	/*
+	 * We start with skipping the prefix SET_Q_MODE and always executing
+	 * the postfix SET_Q_MODE packet. This is changed below with a
+	 * WRITE_DATA command when the postfix executed.
+	 */
+	amdgpu_ring_write(ring, shadow_va ? 1 : 0);
+	amdgpu_ring_write(ring, 0);
+
+	if (ring->set_q_mode_offs) {
+		uint64_t addr;
+
+		addr = amdgpu_bo_gpu_offset(ring->ring_obj);
+		addr += ring->set_q_mode_offs << 2;
+		end = gfx_v11_0_ring_emit_init_cond_exec(ring, addr);
+	}
+
+	/*
+	 * When the postfix SET_Q_MODE packet executes we need to make sure that the
+	 * next prefix SET_Q_MODE packet executes as well.
+	 */
+	if (!shadow_va) {
+		uint64_t addr;
+
+		addr = amdgpu_bo_gpu_offset(ring->ring_obj);
+		addr += offs << 2;
+		amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+		amdgpu_ring_write(ring, WRITE_DATA_DST_SEL(5) | WR_CONFIRM);
+		amdgpu_ring_write(ring, lower_32_bits(addr));
+		amdgpu_ring_write(ring, upper_32_bits(addr));
+		amdgpu_ring_write(ring, 0x1);
+	}
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_SET_Q_PREEMPTION_MODE, 7));
+	amdgpu_ring_write(ring, lower_32_bits(shadow_va));
+	amdgpu_ring_write(ring, upper_32_bits(shadow_va));
+	amdgpu_ring_write(ring, lower_32_bits(gds_va));
+	amdgpu_ring_write(ring, upper_32_bits(gds_va));
+	amdgpu_ring_write(ring, lower_32_bits(csa_va));
+	amdgpu_ring_write(ring, upper_32_bits(csa_va));
+	amdgpu_ring_write(ring, shadow_va ?
+			  PACKET3_SET_Q_PREEMPTION_MODE_IB_VMID(vmid) : 0);
+	amdgpu_ring_write(ring, init_shadow ?
+			  PACKET3_SET_Q_PREEMPTION_MODE_INIT_SHADOW_MEM : 0);
+
+	if (ring->set_q_mode_offs)
+		amdgpu_ring_patch_cond_exec(ring, end);
+
+	if (shadow_va) {
+		uint64_t token = shadow_va ^ csa_va ^ gds_va ^ vmid;
+
+		/*
+		 * If the tokens match try to skip the last postfix SET_Q_MODE
+		 * packet to avoid saving/restoring the state all the time.
+		 */
+		if (ring->set_q_mode_ptr && ring->set_q_mode_token == token)
+			*ring->set_q_mode_ptr = 0;
+
+		ring->set_q_mode_token = token;
+	} else {
+		ring->set_q_mode_ptr = &ring->ring[ring->set_q_mode_offs];
+	}
+
+	ring->set_q_mode_offs = offs;
 }
 
 static int gfx_v11_0_ring_preempt_ib(struct amdgpu_ring *ring)
 {
 	int i, r = 0;
 	struct amdgpu_device *adev = ring->adev;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
 	struct amdgpu_ring *kiq_ring = &kiq->ring;
 	unsigned long flags;
 
+	if (adev->enable_mes)
+		return -EINVAL;
+
 	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
 		return -EINVAL;
 
@@ -5673,28 +6212,13 @@ static void gfx_v11_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume)
 	void *de_payload_cpu_addr;
 	int cnt;
 
-	if (ring->is_mes_queue) {
-		offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				  gfx[0].gfx_meta_data) +
-			offsetof(struct v10_gfx_meta_data, de_payload);
-		de_payload_gpu_addr =
-			amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		de_payload_cpu_addr =
-			amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-
-		offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				  gfx[0].gds_backup) +
-			offsetof(struct v10_gfx_meta_data, de_payload);
-		gds_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-	} else {
-		offset = offsetof(struct v10_gfx_meta_data, de_payload);
-		de_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
-		de_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
+	offset = offsetof(struct v10_gfx_meta_data, de_payload);
+	de_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
+	de_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
 
-		gds_addr = ALIGN(amdgpu_csa_vaddr(ring->adev) +
-				 AMDGPU_CSA_SIZE - adev->gds.gds_size,
-				 PAGE_SIZE);
-	}
+	gds_addr = ALIGN(amdgpu_csa_vaddr(ring->adev) +
+			 AMDGPU_CSA_SIZE - adev->gds.gds_size,
+			 PAGE_SIZE);
 
 	de_payload.gds_backup_addrlo = lower_32_bits(gds_addr);
 	de_payload.gds_backup_addrhi = upper_32_bits(gds_addr);
@@ -5781,19 +6305,6 @@ static void gfx_v11_0_ring_emit_reg_write_reg_wait(struct amdgpu_ring *ring,
 			       ref, mask, 0x20);
 }
 
-static void gfx_v11_0_ring_soft_recovery(struct amdgpu_ring *ring,
-					 unsigned vmid)
-{
-	struct amdgpu_device *adev = ring->adev;
-	uint32_t value = 0;
-
-	value = REG_SET_FIELD(value, SQ_CMD, CMD, 0x03);
-	value = REG_SET_FIELD(value, SQ_CMD, MODE, 0x01);
-	value = REG_SET_FIELD(value, SQ_CMD, CHECK_VMID, 1);
-	value = REG_SET_FIELD(value, SQ_CMD, VM_ID, vmid);
-	WREG32_SOC15(GC, 0, regSQ_CMD, value);
-}
-
 static void
 gfx_v11_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev,
 				      uint32_t me, uint32_t pipe,
@@ -5897,36 +6408,6 @@ static void gfx_v11_0_set_compute_eop_interrupt_state(struct amdgpu_device *adev
 	}
 }
 
-#define CP_ME1_PIPE_INST_ADDR_INTERVAL  0x1
-#define SET_ECC_ME_PIPE_STATE(reg_addr, state) \
-	do { \
-		uint32_t tmp = RREG32_SOC15_IP(GC, reg_addr); \
-		tmp = REG_SET_FIELD(tmp, CP_ME1_PIPE0_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, state); \
-		WREG32_SOC15_IP(GC, reg_addr, tmp); \
-	} while (0)
-
-static int gfx_v11_0_set_cp_ecc_error_state(struct amdgpu_device *adev,
-							struct amdgpu_irq_src *source,
-							unsigned type,
-							enum amdgpu_interrupt_state state)
-{
-	uint32_t ecc_irq_state = 0;
-	uint32_t pipe0_int_cntl_addr = 0;
-	int i = 0;
-
-	ecc_irq_state = (state == AMDGPU_IRQ_STATE_ENABLE) ? 1 : 0;
-
-	pipe0_int_cntl_addr = SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE0_INT_CNTL);
-
-	WREG32_FIELD15_PREREG(GC, 0, CP_INT_CNTL_RING0, CP_ECC_ERROR_INT_ENABLE, ecc_irq_state);
-
-	for (i = 0; i < adev->gfx.mec.num_pipe_per_mec; i++)
-		SET_ECC_ME_PIPE_STATE(pipe0_int_cntl_addr + i * CP_ME1_PIPE_INST_ADDR_INTERVAL,
-					ecc_irq_state);
-
-	return 0;
-}
-
 static int gfx_v11_0_set_eop_interrupt_state(struct amdgpu_device *adev,
 					    struct amdgpu_irq_src *src,
 					    unsigned type,
@@ -5961,25 +6442,23 @@ static int gfx_v11_0_eop_irq(struct amdgpu_device *adev,
 			     struct amdgpu_irq_src *source,
 			     struct amdgpu_iv_entry *entry)
 {
-	int i;
+	u32 doorbell_offset = entry->src_data[0];
 	u8 me_id, pipe_id, queue_id;
 	struct amdgpu_ring *ring;
-	uint32_t mes_queue_id = entry->src_data[0];
+	int i;
 
 	DRM_DEBUG("IH: CP EOP\n");
 
-	if (adev->enable_mes && (mes_queue_id & AMDGPU_FENCE_MES_QUEUE_FLAG)) {
-		struct amdgpu_mes_queue *queue;
-
-		mes_queue_id &= AMDGPU_FENCE_MES_QUEUE_ID_MASK;
+	if (adev->enable_mes && doorbell_offset) {
+		struct amdgpu_userq_fence_driver *fence_drv = NULL;
+		struct xarray *xa = &adev->userq_xa;
+		unsigned long flags;
 
-		spin_lock(&adev->mes.queue_id_lock);
-		queue = idr_find(&adev->mes.queue_id_idr, mes_queue_id);
-		if (queue) {
-			DRM_DEBUG("process mes queue id = %d\n", mes_queue_id);
-			amdgpu_fence_process(queue->ring);
-		}
-		spin_unlock(&adev->mes.queue_id_lock);
+		xa_lock_irqsave(xa, flags);
+		fence_drv = xa_load(xa, doorbell_offset);
+		if (fence_drv)
+			amdgpu_userq_fence_driver_process(fence_drv);
+		xa_unlock_irqrestore(xa, flags);
 	} else {
 		me_id = (entry->ring_id & 0x0c) >> 2;
 		pipe_id = (entry->ring_id & 0x03) >> 0;
@@ -6014,15 +6493,42 @@ static int gfx_v11_0_eop_irq(struct amdgpu_device *adev,
 
 static int gfx_v11_0_set_priv_reg_fault_state(struct amdgpu_device *adev,
 					      struct amdgpu_irq_src *source,
-					      unsigned type,
+					      unsigned int type,
 					      enum amdgpu_interrupt_state state)
 {
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
 	case AMDGPU_IRQ_STATE_ENABLE:
-		WREG32_FIELD15_PREREG(GC, 0, CP_INT_CNTL_RING0,
-			       PRIV_REG_INT_ENABLE,
-			       state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+				cp_int_cntl_reg = gfx_v11_0_get_cpg_int_cntl(adev, i, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+								    PRIV_REG_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				/* MECs start at 1 */
+				cp_int_cntl_reg = gfx_v11_0_get_cpc_int_cntl(adev, i + 1, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+								    PRIV_REG_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
 		break;
 	default:
 		break;
@@ -6031,17 +6537,75 @@ static int gfx_v11_0_set_priv_reg_fault_state(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int gfx_v11_0_set_bad_op_fault_state(struct amdgpu_device *adev,
+					    struct amdgpu_irq_src *source,
+					    unsigned type,
+					    enum amdgpu_interrupt_state state)
+{
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+	case AMDGPU_IRQ_STATE_ENABLE:
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+				cp_int_cntl_reg = gfx_v11_0_get_cpg_int_cntl(adev, i, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+								    OPCODE_ERROR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				/* MECs start at 1 */
+				cp_int_cntl_reg = gfx_v11_0_get_cpc_int_cntl(adev, i + 1, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+								    OPCODE_ERROR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		break;
+	default:
+		break;
+	}
+	return 0;
+}
+
 static int gfx_v11_0_set_priv_inst_fault_state(struct amdgpu_device *adev,
 					       struct amdgpu_irq_src *source,
-					       unsigned type,
+					       unsigned int type,
 					       enum amdgpu_interrupt_state state)
 {
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
 	case AMDGPU_IRQ_STATE_ENABLE:
-		WREG32_FIELD15_PREREG(GC, 0, CP_INT_CNTL_RING0,
-			       PRIV_INSTR_INT_ENABLE,
-			       state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+				cp_int_cntl_reg = gfx_v11_0_get_cpg_int_cntl(adev, i, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+								    PRIV_INSTR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
 		break;
 	default:
 		break;
@@ -6061,27 +6625,29 @@ static void gfx_v11_0_handle_priv_fault(struct amdgpu_device *adev,
 	pipe_id = (entry->ring_id & 0x03) >> 0;
 	queue_id = (entry->ring_id & 0x70) >> 4;
 
-	switch (me_id) {
-	case 0:
-		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
-			ring = &adev->gfx.gfx_ring[i];
-			/* we only enabled 1 gfx queue per pipe for now */
-			if (ring->me == me_id && ring->pipe == pipe_id)
-				drm_sched_fault(&ring->sched);
-		}
-		break;
-	case 1:
-	case 2:
-		for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-			ring = &adev->gfx.compute_ring[i];
-			if (ring->me == me_id && ring->pipe == pipe_id &&
-			    ring->queue == queue_id)
-				drm_sched_fault(&ring->sched);
+	if (!adev->gfx.disable_kq) {
+		switch (me_id) {
+		case 0:
+			for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+				ring = &adev->gfx.gfx_ring[i];
+				if (ring->me == me_id && ring->pipe == pipe_id &&
+				    ring->queue == queue_id)
+					drm_sched_fault(&ring->sched);
+			}
+			break;
+		case 1:
+		case 2:
+			for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+				ring = &adev->gfx.compute_ring[i];
+				if (ring->me == me_id && ring->pipe == pipe_id &&
+				    ring->queue == queue_id)
+					drm_sched_fault(&ring->sched);
+			}
+			break;
+		default:
+			BUG();
+			break;
 		}
-		break;
-	default:
-		BUG();
-		break;
 	}
 }
 
@@ -6094,6 +6660,15 @@ static int gfx_v11_0_priv_reg_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int gfx_v11_0_bad_op_irq(struct amdgpu_device *adev,
+				struct amdgpu_irq_src *source,
+				struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal opcode in command stream \n");
+	gfx_v11_0_handle_priv_fault(adev, entry);
+	return 0;
+}
+
 static int gfx_v11_0_priv_inst_irq(struct amdgpu_device *adev,
 				   struct amdgpu_irq_src *source,
 				   struct amdgpu_iv_entry *entry)
@@ -6120,7 +6695,7 @@ static int gfx_v11_0_kiq_set_interrupt_state(struct amdgpu_device *adev,
 					     enum amdgpu_interrupt_state state)
 {
 	uint32_t tmp, target;
-	struct amdgpu_ring *ring = &(adev->gfx.kiq.ring);
+	struct amdgpu_ring *ring = &(adev->gfx.kiq[0].ring);
 
 	target = SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE0_INT_CNTL);
 	target += ring->pipe;
@@ -6180,6 +6755,426 @@ static void gfx_v11_0_emit_mem_sync(struct amdgpu_ring *ring)
 	amdgpu_ring_write(ring, gcr_cntl); /* GCR_CNTL */
 }
 
+static bool gfx_v11_pipe_reset_support(struct amdgpu_device *adev)
+{
+	/* Disable the pipe reset until the CPFW fully support it.*/
+	dev_warn_once(adev->dev, "The CPFW hasn't support pipe reset yet.\n");
+	return false;
+}
+
+
+static int gfx_v11_reset_gfx_pipe(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t reset_pipe = 0, clean_pipe = 0;
+	int r;
+
+	if (!gfx_v11_pipe_reset_support(adev))
+		return -EOPNOTSUPP;
+
+	gfx_v11_0_set_safe_mode(adev, 0);
+	mutex_lock(&adev->srbm_mutex);
+	soc21_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+
+	switch (ring->pipe) {
+	case 0:
+		reset_pipe = REG_SET_FIELD(reset_pipe, CP_ME_CNTL,
+					   PFP_PIPE0_RESET, 1);
+		reset_pipe = REG_SET_FIELD(reset_pipe, CP_ME_CNTL,
+					   ME_PIPE0_RESET, 1);
+		clean_pipe = REG_SET_FIELD(clean_pipe, CP_ME_CNTL,
+					   PFP_PIPE0_RESET, 0);
+		clean_pipe = REG_SET_FIELD(clean_pipe, CP_ME_CNTL,
+					   ME_PIPE0_RESET, 0);
+		break;
+	case 1:
+		reset_pipe = REG_SET_FIELD(reset_pipe, CP_ME_CNTL,
+					   PFP_PIPE1_RESET, 1);
+		reset_pipe = REG_SET_FIELD(reset_pipe, CP_ME_CNTL,
+					   ME_PIPE1_RESET, 1);
+		clean_pipe = REG_SET_FIELD(clean_pipe, CP_ME_CNTL,
+					   PFP_PIPE1_RESET, 0);
+		clean_pipe = REG_SET_FIELD(clean_pipe, CP_ME_CNTL,
+					   ME_PIPE1_RESET, 0);
+		break;
+	default:
+		break;
+	}
+
+	WREG32_SOC15(GC, 0, regCP_ME_CNTL, reset_pipe);
+	WREG32_SOC15(GC, 0, regCP_ME_CNTL, clean_pipe);
+
+	r = (RREG32(SOC15_REG_OFFSET(GC, 0, regCP_GFX_RS64_INSTR_PNTR1)) << 2) -
+						RS64_FW_UC_START_ADDR_LO;
+	soc21_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	gfx_v11_0_unset_safe_mode(adev, 0);
+
+	dev_info(adev->dev, "The ring %s pipe reset to the ME firmware start PC: %s\n", ring->name,
+			r == 0 ? "successfully" : "failed");
+	/* FIXME: Sometimes driver can't cache the ME firmware start PC correctly,
+	 * so the pipe reset status relies on the later gfx ring test result.
+	 */
+	return 0;
+}
+
+static int gfx_v11_0_reset_kgq(struct amdgpu_ring *ring,
+			       unsigned int vmid,
+			       struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int r;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	r = amdgpu_mes_reset_legacy_queue(ring->adev, ring, vmid, false);
+	if (r) {
+
+		dev_warn(adev->dev, "reset via MES failed and try pipe reset %d\n", r);
+		r = gfx_v11_reset_gfx_pipe(ring);
+		if (r)
+			return r;
+	}
+
+	r = gfx_v11_0_kgq_init_queue(ring, true);
+	if (r) {
+		dev_err(adev->dev, "failed to init kgq\n");
+		return r;
+	}
+
+	r = amdgpu_mes_map_legacy_queue(adev, ring);
+	if (r) {
+		dev_err(adev->dev, "failed to remap kgq\n");
+		return r;
+	}
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static int gfx_v11_0_reset_compute_pipe(struct amdgpu_ring *ring)
+{
+
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t reset_pipe = 0, clean_pipe = 0;
+	int r;
+
+	if (!gfx_v11_pipe_reset_support(adev))
+		return -EOPNOTSUPP;
+
+	gfx_v11_0_set_safe_mode(adev, 0);
+	mutex_lock(&adev->srbm_mutex);
+	soc21_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+
+	reset_pipe = RREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL);
+	clean_pipe = reset_pipe;
+
+	if (adev->gfx.rs64_enable) {
+
+		switch (ring->pipe) {
+		case 0:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE0_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE0_RESET, 0);
+			break;
+		case 1:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE1_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE1_RESET, 0);
+			break;
+		case 2:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE2_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE2_RESET, 0);
+			break;
+		case 3:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE3_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE3_RESET, 0);
+			break;
+		default:
+			break;
+		}
+		WREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL, reset_pipe);
+		WREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL, clean_pipe);
+		r = (RREG32_SOC15(GC, 0, regCP_MEC_RS64_INSTR_PNTR) << 2) -
+					RS64_FW_UC_START_ADDR_LO;
+	} else {
+		if (ring->me == 1) {
+			switch (ring->pipe) {
+			case 0:
+				reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE0_RESET, 1);
+				clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE0_RESET, 0);
+				break;
+			case 1:
+				reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE1_RESET, 1);
+				clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE1_RESET, 0);
+				break;
+			case 2:
+				reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE2_RESET, 1);
+				clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE2_RESET, 0);
+				break;
+			case 3:
+				reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE3_RESET, 1);
+				clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE3_RESET, 0);
+				break;
+			default:
+				break;
+			}
+			/* mec1 fw pc: CP_MEC1_INSTR_PNTR */
+		} else {
+			switch (ring->pipe) {
+			case 0:
+				reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME2_PIPE0_RESET, 1);
+				clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME2_PIPE0_RESET, 0);
+				break;
+			case 1:
+				reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME2_PIPE1_RESET, 1);
+				clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME2_PIPE1_RESET, 0);
+				break;
+			case 2:
+				reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME2_PIPE2_RESET, 1);
+				clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME2_PIPE2_RESET, 0);
+				break;
+			case 3:
+				reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME2_PIPE3_RESET, 1);
+				clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME2_PIPE3_RESET, 0);
+				break;
+			default:
+				break;
+			}
+			/* mec2 fw pc: CP:CP_MEC2_INSTR_PNTR */
+		}
+		WREG32_SOC15(GC, 0, regCP_MEC_CNTL, reset_pipe);
+		WREG32_SOC15(GC, 0, regCP_MEC_CNTL, clean_pipe);
+		r = RREG32(SOC15_REG_OFFSET(GC, 0, regCP_MEC1_INSTR_PNTR));
+	}
+
+	soc21_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	gfx_v11_0_unset_safe_mode(adev, 0);
+
+	dev_info(adev->dev, "The ring %s pipe resets to MEC FW start PC: %s\n", ring->name,
+			r == 0 ? "successfully" : "failed");
+	/*FIXME:Sometimes driver can't cache the MEC firmware start PC correctly, so the pipe
+	 * reset status relies on the compute ring test result.
+	 */
+	return 0;
+}
+
+static int gfx_v11_0_reset_kcq(struct amdgpu_ring *ring,
+			       unsigned int vmid,
+			       struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int r = 0;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	r = amdgpu_mes_reset_legacy_queue(ring->adev, ring, vmid, true);
+	if (r) {
+		dev_warn(adev->dev, "fail(%d) to reset kcq and try pipe reset\n", r);
+		r = gfx_v11_0_reset_compute_pipe(ring);
+		if (r)
+			return r;
+	}
+
+	r = gfx_v11_0_kcq_init_queue(ring, true);
+	if (r) {
+		dev_err(adev->dev, "fail to init kcq\n");
+		return r;
+	}
+	r = amdgpu_mes_map_legacy_queue(adev, ring);
+	if (r) {
+		dev_err(adev->dev, "failed to remap kcq\n");
+		return r;
+	}
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static void gfx_v11_ip_print(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k, reg, index = 0;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_11_0);
+
+	if (!adev->gfx.ip_dump_core)
+		return;
+
+	for (i = 0; i < reg_count; i++)
+		drm_printf(p, "%-50s \t 0x%08x\n",
+			   gc_reg_list_11_0[i].reg_name,
+			   adev->gfx.ip_dump_core[i]);
+
+	/* print compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_11);
+	drm_printf(p, "\nnum_mec: %d num_pipe: %d num_queue: %d\n",
+		   adev->gfx.mec.num_mec,
+		   adev->gfx.mec.num_pipe_per_mec,
+		   adev->gfx.mec.num_queue_per_pipe);
+
+	for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+		for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+			for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+				drm_printf(p, "\nmec %d, pipe %d, queue %d\n", i, j, k);
+				for (reg = 0; reg < reg_count; reg++) {
+					if (i && gc_cp_reg_list_11[reg].reg_offset == regCP_MEC_ME1_HEADER_DUMP)
+						drm_printf(p, "%-50s \t 0x%08x\n",
+							   "regCP_MEC_ME2_HEADER_DUMP",
+							   adev->gfx.ip_dump_compute_queues[index + reg]);
+					else
+						drm_printf(p, "%-50s \t 0x%08x\n",
+							   gc_cp_reg_list_11[reg].reg_name,
+							   adev->gfx.ip_dump_compute_queues[index + reg]);
+				}
+				index += reg_count;
+			}
+		}
+	}
+
+	/* print gfx queue registers for all instances */
+	if (!adev->gfx.ip_dump_gfx_queues)
+		return;
+
+	index = 0;
+	reg_count = ARRAY_SIZE(gc_gfx_queue_reg_list_11);
+	drm_printf(p, "\nnum_me: %d num_pipe: %d num_queue: %d\n",
+		   adev->gfx.me.num_me,
+		   adev->gfx.me.num_pipe_per_me,
+		   adev->gfx.me.num_queue_per_pipe);
+
+	for (i = 0; i < adev->gfx.me.num_me; i++) {
+		for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+			for (k = 0; k < adev->gfx.me.num_queue_per_pipe; k++) {
+				drm_printf(p, "\nme %d, pipe %d, queue %d\n", i, j, k);
+				for (reg = 0; reg < reg_count; reg++) {
+					drm_printf(p, "%-50s \t 0x%08x\n",
+						   gc_gfx_queue_reg_list_11[reg].reg_name,
+						   adev->gfx.ip_dump_gfx_queues[index + reg]);
+				}
+				index += reg_count;
+			}
+		}
+	}
+}
+
+static void gfx_v11_ip_dump(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k, reg, index = 0;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_11_0);
+
+	if (!adev->gfx.ip_dump_core)
+		return;
+
+	amdgpu_gfx_off_ctrl(adev, false);
+	for (i = 0; i < reg_count; i++)
+		adev->gfx.ip_dump_core[i] = RREG32(SOC15_REG_ENTRY_OFFSET(gc_reg_list_11_0[i]));
+	amdgpu_gfx_off_ctrl(adev, true);
+
+	/* dump compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_11);
+	amdgpu_gfx_off_ctrl(adev, false);
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+		for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+			for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+				/* ME0 is for GFX so start from 1 for CP */
+				soc21_grbm_select(adev, adev->gfx.me.num_me + i, j, k, 0);
+				for (reg = 0; reg < reg_count; reg++) {
+					if (i &&
+					    gc_cp_reg_list_11[reg].reg_offset ==
+						    regCP_MEC_ME1_HEADER_DUMP)
+						adev->gfx.ip_dump_compute_queues[index + reg] =
+							RREG32(SOC15_REG_OFFSET(GC, 0,
+							       regCP_MEC_ME2_HEADER_DUMP));
+					else
+						adev->gfx.ip_dump_compute_queues[index + reg] =
+							RREG32(SOC15_REG_ENTRY_OFFSET(
+								       gc_cp_reg_list_11[reg]));
+				}
+				index += reg_count;
+			}
+		}
+	}
+	soc21_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	amdgpu_gfx_off_ctrl(adev, true);
+
+	/* dump gfx queue registers for all instances */
+	if (!adev->gfx.ip_dump_gfx_queues)
+		return;
+
+	index = 0;
+	reg_count = ARRAY_SIZE(gc_gfx_queue_reg_list_11);
+	amdgpu_gfx_off_ctrl(adev, false);
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->gfx.me.num_me; i++) {
+		for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+			for (k = 0; k < adev->gfx.me.num_queue_per_pipe; k++) {
+				soc21_grbm_select(adev, i, j, k, 0);
+
+				for (reg = 0; reg < reg_count; reg++) {
+					adev->gfx.ip_dump_gfx_queues[index + reg] =
+						RREG32(SOC15_REG_ENTRY_OFFSET(
+							gc_gfx_queue_reg_list_11[reg]));
+				}
+				index += reg_count;
+			}
+		}
+	}
+	soc21_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	amdgpu_gfx_off_ctrl(adev, true);
+}
+
+static void gfx_v11_0_ring_emit_cleaner_shader(struct amdgpu_ring *ring)
+{
+	/* Emit the cleaner shader */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_RUN_CLEANER_SHADER, 0));
+	amdgpu_ring_write(ring, 0);  /* RESERVED field, programmed to zero */
+}
+
+static void gfx_v11_0_ring_begin_use(struct amdgpu_ring *ring)
+{
+	amdgpu_gfx_profile_ring_begin_use(ring);
+
+	amdgpu_gfx_enforce_isolation_ring_begin_use(ring);
+}
+
+static void gfx_v11_0_ring_end_use(struct amdgpu_ring *ring)
+{
+	amdgpu_gfx_profile_ring_end_use(ring);
+
+	amdgpu_gfx_enforce_isolation_ring_end_use(ring);
+}
+
 static const struct amd_ip_funcs gfx_v11_0_ip_funcs = {
 	.name = "gfx_v11_0",
 	.early_init = gfx_v11_0_early_init,
@@ -6198,6 +7193,8 @@ static const struct amd_ip_funcs gfx_v11_0_ip_funcs = {
 	.set_clockgating_state = gfx_v11_0_set_clockgating_state,
 	.set_powergating_state = gfx_v11_0_set_powergating_state,
 	.get_clockgating_state = gfx_v11_0_get_clockgating_state,
+	.dump_ip_state = gfx_v11_ip_dump,
+	.print_ip_state = gfx_v11_ip_print,
 };
 
 static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_gfx = {
@@ -6209,12 +7206,14 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_gfx = {
 	.get_rptr = gfx_v11_0_ring_get_rptr_gfx,
 	.get_wptr = gfx_v11_0_ring_get_wptr_gfx,
 	.set_wptr = gfx_v11_0_ring_set_wptr_gfx,
-	.emit_frame_size = /* totally 242 maximum if 16 IBs */
+	.emit_frame_size = /* totally 247 maximum if 16 IBs */
+		5 + /* update_spm_vmid */
 		5 + /* COND_EXEC */
+		22 + /* SET_Q_PREEMPTION_MODE */
 		7 + /* PIPELINE_SYNC */
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
-		2 + /* VM_FLUSH */
+		4 + /* VM_FLUSH */
 		8 + /* FENCE for VM_FLUSH */
 		20 + /* GDS switch */
 		5 + /* COND_EXEC */
@@ -6223,8 +7222,10 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_gfx = {
 		31 + /*	DE_META */
 		3 + /* CNTX_CTRL */
 		5 + /* HDP_INVL */
+		22 + /* SET_Q_PREEMPTION_MODE */
 		8 + 8 + /* FENCE x2 */
-		8, /* gfx_v11_0_emit_mem_sync */
+		8 + /* gfx_v11_0_emit_mem_sync */
+		2, /* gfx_v11_0_ring_emit_cleaner_shader */
 	.emit_ib_size =	4, /* gfx_v11_0_ring_emit_ib_gfx */
 	.emit_ib = gfx_v11_0_ring_emit_ib_gfx,
 	.emit_fence = gfx_v11_0_ring_emit_fence,
@@ -6234,18 +7235,21 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_gfx = {
 	.emit_hdp_flush = gfx_v11_0_ring_emit_hdp_flush,
 	.test_ring = gfx_v11_0_ring_test_ring,
 	.test_ib = gfx_v11_0_ring_test_ib,
-	.insert_nop = amdgpu_ring_insert_nop,
+	.insert_nop = gfx_v11_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_cntxcntl = gfx_v11_0_ring_emit_cntxcntl,
+	.emit_gfx_shadow = gfx_v11_0_ring_emit_gfx_shadow,
 	.init_cond_exec = gfx_v11_0_ring_emit_init_cond_exec,
-	.patch_cond_exec = gfx_v11_0_ring_emit_patch_cond_exec,
 	.preempt_ib = gfx_v11_0_ring_preempt_ib,
 	.emit_frame_cntl = gfx_v11_0_ring_emit_frame_cntl,
 	.emit_wreg = gfx_v11_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v11_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = gfx_v11_0_ring_emit_reg_write_reg_wait,
-	.soft_recovery = gfx_v11_0_ring_soft_recovery,
 	.emit_mem_sync = gfx_v11_0_emit_mem_sync,
+	.reset = gfx_v11_0_reset_kgq,
+	.emit_cleaner_shader = gfx_v11_0_ring_emit_cleaner_shader,
+	.begin_use = gfx_v11_0_ring_begin_use,
+	.end_use = gfx_v11_0_ring_end_use,
 };
 
 static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_compute = {
@@ -6257,6 +7261,7 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_compute = {
 	.get_wptr = gfx_v11_0_ring_get_wptr_compute,
 	.set_wptr = gfx_v11_0_ring_set_wptr_compute,
 	.emit_frame_size =
+		5 + /* update_spm_vmid */
 		20 + /* gfx_v11_0_ring_emit_gds_switch */
 		7 + /* gfx_v11_0_ring_emit_hdp_flush */
 		5 + /* hdp invalidate */
@@ -6265,7 +7270,8 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_compute = {
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
 		2 + /* gfx_v11_0_ring_emit_vm_flush */
 		8 + 8 + 8 + /* gfx_v11_0_ring_emit_fence x3 for user fence, vm fence */
-		8, /* gfx_v11_0_emit_mem_sync */
+		8 + /* gfx_v11_0_emit_mem_sync */
+		2, /* gfx_v11_0_ring_emit_cleaner_shader */
 	.emit_ib_size =	7, /* gfx_v11_0_ring_emit_ib_compute */
 	.emit_ib = gfx_v11_0_ring_emit_ib_compute,
 	.emit_fence = gfx_v11_0_ring_emit_fence,
@@ -6275,12 +7281,16 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_compute = {
 	.emit_hdp_flush = gfx_v11_0_ring_emit_hdp_flush,
 	.test_ring = gfx_v11_0_ring_test_ring,
 	.test_ib = gfx_v11_0_ring_test_ib,
-	.insert_nop = amdgpu_ring_insert_nop,
+	.insert_nop = gfx_v11_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_wreg = gfx_v11_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v11_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = gfx_v11_0_ring_emit_reg_write_reg_wait,
 	.emit_mem_sync = gfx_v11_0_emit_mem_sync,
+	.reset = gfx_v11_0_reset_kcq,
+	.emit_cleaner_shader = gfx_v11_0_ring_emit_cleaner_shader,
+	.begin_use = gfx_v11_0_ring_begin_use,
+	.end_use = gfx_v11_0_ring_end_use,
 };
 
 static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_kiq = {
@@ -6298,7 +7308,6 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_kiq = {
 		7 + /* gfx_v11_0_ring_emit_pipeline_sync */
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
-		2 + /* gfx_v11_0_ring_emit_vm_flush */
 		8 + 8 + 8, /* gfx_v11_0_ring_emit_fence_kiq x3 for user fence, vm fence */
 	.emit_ib_size =	7, /* gfx_v11_0_ring_emit_ib_compute */
 	.emit_ib = gfx_v11_0_ring_emit_ib_compute,
@@ -6311,13 +7320,14 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_kiq = {
 	.emit_wreg = gfx_v11_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v11_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = gfx_v11_0_ring_emit_reg_write_reg_wait,
+	.emit_hdp_flush = gfx_v11_0_ring_emit_hdp_flush,
 };
 
 static void gfx_v11_0_set_ring_funcs(struct amdgpu_device *adev)
 {
 	int i;
 
-	adev->gfx.kiq.ring.funcs = &gfx_v11_0_ring_funcs_kiq;
+	adev->gfx.kiq[0].ring.funcs = &gfx_v11_0_ring_funcs_kiq;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
 		adev->gfx.gfx_ring[i].funcs = &gfx_v11_0_ring_funcs_gfx;
@@ -6336,16 +7346,16 @@ static const struct amdgpu_irq_src_funcs gfx_v11_0_priv_reg_irq_funcs = {
 	.process = gfx_v11_0_priv_reg_irq,
 };
 
+static const struct amdgpu_irq_src_funcs gfx_v11_0_bad_op_irq_funcs = {
+	.set = gfx_v11_0_set_bad_op_fault_state,
+	.process = gfx_v11_0_bad_op_irq,
+};
+
 static const struct amdgpu_irq_src_funcs gfx_v11_0_priv_inst_irq_funcs = {
 	.set = gfx_v11_0_set_priv_inst_fault_state,
 	.process = gfx_v11_0_priv_inst_irq,
 };
 
-static const struct amdgpu_irq_src_funcs gfx_v11_0_cp_ecc_error_irq_funcs = {
-	.set = gfx_v11_0_set_cp_ecc_error_state,
-	.process = amdgpu_gfx_cp_ecc_error_irq,
-};
-
 static const struct amdgpu_irq_src_funcs gfx_v11_0_rlc_gc_fed_irq_funcs = {
 	.process = gfx_v11_0_rlc_gc_fed_irq,
 };
@@ -6358,12 +7368,12 @@ static void gfx_v11_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->gfx.priv_reg_irq.num_types = 1;
 	adev->gfx.priv_reg_irq.funcs = &gfx_v11_0_priv_reg_irq_funcs;
 
+	adev->gfx.bad_op_irq.num_types = 1;
+	adev->gfx.bad_op_irq.funcs = &gfx_v11_0_bad_op_irq_funcs;
+
 	adev->gfx.priv_inst_irq.num_types = 1;
 	adev->gfx.priv_inst_irq.funcs = &gfx_v11_0_priv_inst_irq_funcs;
 
-	adev->gfx.cp_ecc_error_irq.num_types = 1; /* CP ECC error */
-	adev->gfx.cp_ecc_error_irq.funcs = &gfx_v11_0_cp_ecc_error_irq_funcs;
-
 	adev->gfx.rlc_gc_fed_irq.num_types = 1; /* 0x80 FED error */
 	adev->gfx.rlc_gc_fed_irq.funcs = &gfx_v11_0_rlc_gc_fed_irq_funcs;
 
@@ -6472,9 +7482,12 @@ static int gfx_v11_0_get_cu_info(struct amdgpu_device *adev,
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
+			bitmap = i * adev->gfx.config.max_sh_per_se + j;
+			if (!((gfx_v11_0_get_sa_active_bitmap(adev) >> bitmap) & 1))
+				continue;
 			mask = 1;
 			counter = 0;
-			gfx_v11_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v11_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			if (i < 8 && j < 2)
 				gfx_v11_0_set_user_wgp_inactive_bitmap_per_sh(
 					adev, disable_masks[i * 2 + j]);
@@ -6495,7 +7508,7 @@ static int gfx_v11_0_get_cu_info(struct amdgpu_device *adev,
 			 *    SE6: {SH0,SH1} --> {bitmap[2][2], bitmap[2][3]}
 			 *    SE7: {SH0,SH1} --> {bitmap[3][2], bitmap[3][3]}
 			 */
-			cu_info->bitmap[i % 4][j + (i / 4) * 2] = bitmap;
+			cu_info->bitmap[0][i % 4][j + (i / 4) * 2] = bitmap;
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k++) {
 				if (bitmap & mask)
@@ -6506,7 +7519,7 @@ static int gfx_v11_0_get_cu_info(struct amdgpu_device *adev,
 			active_cu_number += counter;
 		}
 	}
-	gfx_v11_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v11_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	cu_info->number = active_cu_number;
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.h b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.h
index 10cfc29c27c9..157a5c812259 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.h
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.h
@@ -26,4 +26,7 @@
 
 extern const struct amdgpu_ip_block_version gfx_v11_0_ip_block;
 
+int gfx_v11_0_request_gfx_index_mutex(struct amdgpu_device *adev,
+				      bool req);
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_3.c
index 068b9586a223..999bb3cc88b7 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_3.c
@@ -69,7 +69,7 @@ static int gfx_v11_0_3_rlc_gc_fed_irq(struct amdgpu_device *adev,
 		amdgpu_ras_interrupt_dispatch(adev, &ih_data);
 	} else {
 		if (adev->virt.ops && adev->virt.ops->ras_poison_handler)
-			adev->virt.ops->ras_poison_handler(adev);
+			adev->virt.ops->ras_poison_handler(adev, ras_if->block);
 		else
 			dev_warn(adev->dev,
 				"No ras_poison_handler interface in SRIOV for %s!\n", ras_if->name);
@@ -84,8 +84,22 @@ static int gfx_v11_0_3_poison_consumption_handler(struct amdgpu_device *adev,
 	/* Workaround: when vmid and pasid are both zero, trigger gpu reset in KGD. */
 	if (entry && (entry->client_id == SOC21_IH_CLIENTID_GFX) &&
 	    (entry->src_id == GFX_11_0_0__SRCID__RLC_GC_FED_INTERRUPT) &&
-	     !entry->vmid && !entry->pasid)
-		amdgpu_ras_reset_gpu(adev);
+	     !entry->vmid && !entry->pasid) {
+		struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+		uint32_t rlc_status0 = 0;
+
+		rlc_status0 = RREG32_SOC15(GC, 0, regRLC_RLCS_FED_STATUS_0);
+
+		if (REG_GET_FIELD(rlc_status0, RLC_RLCS_FED_STATUS_0, SDMA0_FED_ERR) ||
+		    REG_GET_FIELD(rlc_status0, RLC_RLCS_FED_STATUS_0, SDMA1_FED_ERR)) {
+			struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+
+			ras->gpu_reset_flags |= AMDGPU_RAS_GPU_RESET_MODE2_RESET;
+		}
+
+		if (con && !amdgpu_ras_is_rma(adev))
+			amdgpu_ras_reset_gpu(adev);
+	}
 
 	return 0;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_3_cleaner_shader.asm b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_3_cleaner_shader.asm
new file mode 100644
index 000000000000..9b90b66368c7
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_3_cleaner_shader.asm
@@ -0,0 +1,118 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+// This shader is to clean LDS, SGPRs and VGPRs. It is  first 64 Dwords or 256 bytes of 192 Dwords cleaner shader.
+//To turn this shader program on for complitaion change this to main and lower shader main to main_1
+
+// Navi3 : Clear SGPRs, VGPRs and LDS
+//   Launch 32 waves per CU (16 per SIMD) as a workgroup (threadgroup) to fill every wave slot
+//   Waves are "wave32" and have 64 VGPRs each, which uses all 1024 VGPRs per SIMD
+//   Waves are launched in "CU" mode, and the workgroup shares 64KB of LDS (half of the WGP's LDS)
+//      It takes 2 workgroups to use all of LDS: one on each CU of the WGP
+//   Each wave clears SGPRs 0 - 107
+//   Each wave clears VGPRs 0 - 63
+//   The first wave of the workgroup clears its 64KB of LDS
+//   The shader starts with "S_BARRIER" to ensure SPI has launched all waves of the workgroup
+//       before any wave in the workgroup could end.  Without this, it is possible not all SGPRs get cleared.
+
+shader main
+  asic(GFX11)
+  type(CS)
+  wave_size(32)
+// Note: original source code from SQ team
+
+// Takes about 2500 clocks to run.
+//   (theorhetical fastest = 1024clks vgpr + 640lds = 1660 clks)
+//
+  S_BARRIER
+
+  //
+  // CLEAR VGPRs
+  //
+  s_mov_b32     m0, 0x00000058  // Loop 96/8=12 times  (loop unrolled for performance)
+
+label_0005:
+  v_movreld_b32     v0, 0
+  v_movreld_b32     v1, 0
+  v_movreld_b32     v2, 0
+  v_movreld_b32     v3, 0
+  v_movreld_b32     v4, 0
+  v_movreld_b32     v5, 0
+  v_movreld_b32     v6, 0
+  v_movreld_b32     v7, 0
+  s_sub_u32     m0, m0, 8
+  s_cbranch_scc0  label_0005
+  //
+  //
+
+  s_mov_b32     s2, 0x80000000                      // Bit31 is first_wave
+  s_and_b32     s2, s2, s0                          // sgpr0 has tg_size (first_wave) term as in ucode only COMPUTE_PGM_RSRC2.tg_size_en is set
+  s_cbranch_scc0  label_0023                        // Clean LDS if its first wave of ThreadGroup/WorkGroup
+  // CLEAR LDS
+  //
+  s_mov_b32 exec_lo, 0xffffffff
+  s_mov_b32 exec_hi, 0xffffffff
+  v_mbcnt_lo_u32_b32  v1, exec_hi, 0          // Set V1 to thread-ID (0..63)
+  v_mbcnt_hi_u32_b32  v1, exec_lo, v1        // Set V1 to thread-ID (0..63)
+  v_mul_u32_u24  v1, 0x00000008, v1          // * 8, so each thread is a double-dword address (8byte)
+  s_mov_b32     s2, 0x00000003f                    // 64 loop iterations
+  s_mov_b32     m0, 0xffffffff
+  // Clear all of LDS space
+  // Each FirstWave of WorkGroup clears 64kbyte block
+
+label_001F:
+  ds_write2_b64  v1, v[2:3], v[2:3] offset1:32
+  ds_write2_b64  v1, v[4:5], v[4:5] offset0:64 offset1:96
+  v_add_co_u32     v1, vcc, 0x00000400, v1
+  s_sub_u32     s2, s2, 1
+  s_cbranch_scc0  label_001F
+  //
+  // CLEAR SGPRs
+  //
+label_0023:
+  s_mov_b32     m0, 0x00000068  // Loop 108/4=27 times  (loop unrolled for performance)
+label_sgpr_loop:
+  s_movreld_b32     s0, 0
+  s_movreld_b32     s1, 0
+  s_movreld_b32     s2, 0
+  s_movreld_b32     s3, 0
+  s_sub_u32         m0, m0, 4
+  s_cbranch_scc0  label_sgpr_loop
+
+  //clear vcc
+  s_mov_b64 vcc, 0          //clear vcc
+  s_mov_b32 flat_scratch_lo, 0   //clear  flat scratch lo SGPR
+  s_mov_b32 flat_scratch_hi, 0   //clear  flat scratch hi SGPR
+  s_mov_b64 ttmp0, 0        //Clear ttmp0 and ttmp1
+  s_mov_b64 ttmp2, 0        //Clear ttmp2 and ttmp3
+  s_mov_b64 ttmp4, 0        //Clear ttmp4 and ttmp5
+  s_mov_b64 ttmp6, 0        //Clear ttmp6 and ttmp7
+  s_mov_b64 ttmp8, 0        //Clear ttmp8 and ttmp9
+  s_mov_b64 ttmp10, 0       //Clear ttmp10 and ttmp11
+  s_mov_b64 ttmp12, 0       //Clear ttmp12 and ttmp13
+  s_mov_b64 ttmp14, 0       //Clear ttmp14 and ttmp15
+
+ s_endpgm
+
+end
+
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_cleaner_shader.h b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_cleaner_shader.h
new file mode 100644
index 000000000000..3218cc04f543
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_cleaner_shader.h
@@ -0,0 +1,56 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/* Define the cleaner shader gfx_11_0_3 */
+static const u32 gfx_11_0_3_cleaner_shader_hex[] = {
+	0xb0804006, 0xbe8200ff,
+	0x00000058, 0xbefd0080,
+	0x7e008480, 0x7e028480,
+	0x7e048480, 0x7e068480,
+	0x7e088480, 0x7e0a8480,
+	0x7e0c8480, 0x7e0e8480,
+	0xbefd0002, 0x80828802,
+	0xbfa1fff5, 0xbe8200ff,
+	0x80000000, 0x8b020002,
+	0xbfa10012, 0xbefe00c1,
+	0xbeff00c1, 0xd71f0001,
+	0x0001007f, 0xd7200001,
+	0x0002027e, 0x16020288,
+	0xbe8200bf, 0xbefd00c1,
+	0xd9382000, 0x00020201,
+	0xd9386040, 0x00040401,
+	0xd7006a01, 0x000202ff,
+	0x00000400, 0x80828102,
+	0xbfa1fff7, 0xbefd00ff,
+	0x00000068, 0xbe804280,
+	0xbe814280, 0xbe824280,
+	0xbe834280, 0x80fd847d,
+	0xbfa1fffa, 0xbeea0180,
+	0xbeec0180, 0xbeee0180,
+	0xbef00180, 0xbef20180,
+	0xbef40180, 0xbef60180,
+	0xbef80180, 0xbefa0180,
+	0xbfb00000, 0xbf9f0000,
+	0xbf9f0000, 0xbf9f0000,
+	0xbf9f0000, 0xbf9f0000,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c
new file mode 100644
index 000000000000..d01d2712cf57
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c
@@ -0,0 +1,5793 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include <linux/delay.h>
+#include <linux/kernel.h>
+#include <linux/firmware.h>
+#include <linux/module.h>
+#include <linux/pci.h>
+#include "amdgpu.h"
+#include "amdgpu_gfx.h"
+#include "amdgpu_psp.h"
+#include "amdgpu_smu.h"
+#include "imu_v12_0.h"
+#include "soc24.h"
+#include "nvd.h"
+
+#include "gc/gc_12_0_0_offset.h"
+#include "gc/gc_12_0_0_sh_mask.h"
+#include "soc24_enum.h"
+#include "ivsrcid/gfx/irqsrcs_gfx_12_0_0.h"
+
+#include "soc15.h"
+#include "clearstate_gfx12.h"
+#include "v12_structs.h"
+#include "gfx_v12_0.h"
+#include "nbif_v6_3_1.h"
+#include "mes_v12_0.h"
+#include "mes_userqueue.h"
+#include "amdgpu_userq_fence.h"
+
+#define GFX12_NUM_GFX_RINGS	1
+#define GFX12_MEC_HPD_SIZE	2048
+
+#define RLCG_UCODE_LOADING_START_ADDRESS	0x00002000L
+
+#define regCP_GFX_MQD_CONTROL_DEFAULT                                             0x00000100
+#define regCP_GFX_HQD_VMID_DEFAULT                                                0x00000000
+#define regCP_GFX_HQD_QUEUE_PRIORITY_DEFAULT                                      0x00000000
+#define regCP_GFX_HQD_QUANTUM_DEFAULT                                             0x00000a01
+#define regCP_GFX_HQD_CNTL_DEFAULT                                                0x00f00000
+#define regCP_RB_DOORBELL_CONTROL_DEFAULT                                         0x00000000
+#define regCP_GFX_HQD_RPTR_DEFAULT                                                0x00000000
+
+#define regCP_HQD_EOP_CONTROL_DEFAULT                                             0x00000006
+#define regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT                                     0x00000000
+#define regCP_MQD_CONTROL_DEFAULT                                                 0x00000100
+#define regCP_HQD_PQ_CONTROL_DEFAULT                                              0x00308509
+#define regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT                                     0x00000000
+#define regCP_HQD_PQ_RPTR_DEFAULT                                                 0x00000000
+#define regCP_HQD_PERSISTENT_STATE_DEFAULT                                        0x0be05501
+#define regCP_HQD_IB_CONTROL_DEFAULT                                              0x00300000
+
+
+MODULE_FIRMWARE("amdgpu/gc_12_0_0_pfp.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_0_me.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_0_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_0_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_0_toc.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_pfp.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_me.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_rlc_kicker.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_toc.bin");
+
+static const struct amdgpu_hwip_reg_entry gc_reg_list_12_0[] = {
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS2),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS3),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_STALLED_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_STALLED_STAT3),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_BUSY_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_BUSY_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HPD_STATUS0),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB0_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB0_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB0_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB2_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB2_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB2_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB2_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCPF_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCPC_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCPG_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regIA_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regIA_UTCL1_STATUS_2),
+	SOC15_REG_ENTRY_STR(GC, 0, regPA_CL_CNTL_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regRMI_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regSQC_CACHES),
+	SOC15_REG_ENTRY_STR(GC, 0, regSQG_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regWD_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regGCVM_L2_PROTECTION_FAULT_STATUS_LO32),
+	SOC15_REG_ENTRY_STR(GC, 0, regGCVM_L2_PROTECTION_FAULT_STATUS_HI32),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_DEBUG),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_RS64_INSTR_PNTR0),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_RS64_INSTR_PNTR1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_RS64_INSTR_PNTR),
+	/* cp header registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	/* SE status registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE0),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE1),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE2),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE3)
+};
+
+static const struct amdgpu_hwip_reg_entry gc_cp_reg_list_12[] = {
+	/* compute registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_VMID),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PERSISTENT_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PIPE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_QUEUE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_QUANTUM),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_DEQUEUE_REQUEST),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_EVENTS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_BASE_ADDR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CNTL_STACK_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CNTL_STACK_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_WG_STATE_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_GDS_RESOURCE_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_WPTR_MEM),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_SUSPEND_CNTL_STACK_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_SUSPEND_CNTL_STACK_DW_CNT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_SUSPEND_WG_STATE_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_DEQUEUE_STATUS),
+	/* cp header registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+};
+
+static const struct amdgpu_hwip_reg_entry gc_gfx_queue_reg_list_12[] = {
+	/* gfx queue registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_ACTIVE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_VMID),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_QUEUE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_QUANTUM),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_CSMD_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_DEQUEUE_REQUEST),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_MAPPED),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_QUE_MGR_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_HQ_CONTROL0),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_HQ_STATUS0),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_MQD_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_MQD_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_WPTR_POLL_ADDR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_WPTR_POLL_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BUFSZ),
+	/* cp header registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+};
+
+static const struct soc15_reg_golden golden_settings_gc_12_0_rev0[] = {
+	SOC15_REG_GOLDEN_VALUE(GC, 0, regDB_MEM_CONFIG, 0x0000000f, 0x0000000f),
+	SOC15_REG_GOLDEN_VALUE(GC, 0, regCB_HW_CONTROL_1, 0x03000000, 0x03000000),
+	SOC15_REG_GOLDEN_VALUE(GC, 0, regGL2C_CTRL5, 0x00000070, 0x00000020)
+};
+
+static const struct soc15_reg_golden golden_settings_gc_12_0[] = {
+	SOC15_REG_GOLDEN_VALUE(GC, 0, regDB_MEM_CONFIG, 0x00008000, 0x00008000),
+};
+
+#define DEFAULT_SH_MEM_CONFIG \
+	((SH_MEM_ADDRESS_MODE_64 << SH_MEM_CONFIG__ADDRESS_MODE__SHIFT) | \
+	 (SH_MEM_ALIGNMENT_MODE_UNALIGNED << SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT) | \
+	 (3 << SH_MEM_CONFIG__INITIAL_INST_PREFETCH__SHIFT))
+
+static void gfx_v12_0_disable_gpa_mode(struct amdgpu_device *adev);
+static void gfx_v12_0_set_ring_funcs(struct amdgpu_device *adev);
+static void gfx_v12_0_set_irq_funcs(struct amdgpu_device *adev);
+static void gfx_v12_0_set_rlc_funcs(struct amdgpu_device *adev);
+static void gfx_v12_0_set_mqd_funcs(struct amdgpu_device *adev);
+static void gfx_v12_0_set_imu_funcs(struct amdgpu_device *adev);
+static int gfx_v12_0_get_cu_info(struct amdgpu_device *adev,
+				 struct amdgpu_cu_info *cu_info);
+static uint64_t gfx_v12_0_get_gpu_clock_counter(struct amdgpu_device *adev);
+static void gfx_v12_0_select_se_sh(struct amdgpu_device *adev, u32 se_num,
+				   u32 sh_num, u32 instance, int xcc_id);
+static u32 gfx_v12_0_get_wgp_active_bitmap_per_sh(struct amdgpu_device *adev);
+
+static void gfx_v12_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start, bool secure);
+static void gfx_v12_0_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg,
+				     uint32_t val);
+static int gfx_v12_0_wait_for_rlc_autoload_complete(struct amdgpu_device *adev);
+static void gfx_v12_0_ring_invalidate_tlbs(struct amdgpu_ring *ring,
+					   uint16_t pasid, uint32_t flush_type,
+					   bool all_hub, uint8_t dst_sel);
+static void gfx_v12_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id);
+static void gfx_v12_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id);
+static void gfx_v12_0_update_perf_clk(struct amdgpu_device *adev,
+				      bool enable);
+
+static void gfx_v12_0_kiq_set_resources(struct amdgpu_ring *kiq_ring,
+					uint64_t queue_mask)
+{
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_RESOURCES, 6));
+	amdgpu_ring_write(kiq_ring, PACKET3_SET_RESOURCES_VMID_MASK(0) |
+			  PACKET3_SET_RESOURCES_QUEUE_TYPE(0));	/* vmid_mask:0 queue_type:0 (KIQ) */
+	amdgpu_ring_write(kiq_ring, lower_32_bits(queue_mask));	/* queue mask lo */
+	amdgpu_ring_write(kiq_ring, upper_32_bits(queue_mask));	/* queue mask hi */
+	amdgpu_ring_write(kiq_ring, 0);	/* gws mask lo */
+	amdgpu_ring_write(kiq_ring, 0);	/* gws mask hi */
+	amdgpu_ring_write(kiq_ring, 0);	/* oac mask */
+	amdgpu_ring_write(kiq_ring, 0);
+}
+
+static void gfx_v12_0_kiq_map_queues(struct amdgpu_ring *kiq_ring,
+				     struct amdgpu_ring *ring)
+{
+	uint64_t mqd_addr = amdgpu_bo_gpu_offset(ring->mqd_obj);
+	uint64_t wptr_addr = ring->wptr_gpu_addr;
+	uint32_t me = 0, eng_sel = 0;
+
+	switch (ring->funcs->type) {
+	case AMDGPU_RING_TYPE_COMPUTE:
+		me = 1;
+		eng_sel = 0;
+		break;
+	case AMDGPU_RING_TYPE_GFX:
+		me = 0;
+		eng_sel = 4;
+		break;
+	case AMDGPU_RING_TYPE_MES:
+		me = 2;
+		eng_sel = 5;
+		break;
+	default:
+		WARN_ON(1);
+	}
+
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_MAP_QUEUES, 5));
+	/* Q_sel:0, vmid:0, vidmem: 1, engine:0, num_Q:1*/
+	amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */
+			  PACKET3_MAP_QUEUES_QUEUE_SEL(0) | /* Queue_Sel */
+			  PACKET3_MAP_QUEUES_VMID(0) | /* VMID */
+			  PACKET3_MAP_QUEUES_QUEUE(ring->queue) |
+			  PACKET3_MAP_QUEUES_PIPE(ring->pipe) |
+			  PACKET3_MAP_QUEUES_ME((me)) |
+			  PACKET3_MAP_QUEUES_QUEUE_TYPE(0) | /*queue_type: normal compute queue */
+			  PACKET3_MAP_QUEUES_ALLOC_FORMAT(0) | /* alloc format: all_on_one_pipe */
+			  PACKET3_MAP_QUEUES_ENGINE_SEL(eng_sel) |
+			  PACKET3_MAP_QUEUES_NUM_QUEUES(1)); /* num_queues: must be 1 */
+	amdgpu_ring_write(kiq_ring, PACKET3_MAP_QUEUES_DOORBELL_OFFSET(ring->doorbell_index));
+	amdgpu_ring_write(kiq_ring, lower_32_bits(mqd_addr));
+	amdgpu_ring_write(kiq_ring, upper_32_bits(mqd_addr));
+	amdgpu_ring_write(kiq_ring, lower_32_bits(wptr_addr));
+	amdgpu_ring_write(kiq_ring, upper_32_bits(wptr_addr));
+}
+
+static void gfx_v12_0_kiq_unmap_queues(struct amdgpu_ring *kiq_ring,
+				       struct amdgpu_ring *ring,
+				       enum amdgpu_unmap_queues_action action,
+				       u64 gpu_addr, u64 seq)
+{
+	struct amdgpu_device *adev = kiq_ring->adev;
+	uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0;
+
+	if (adev->enable_mes && !adev->gfx.kiq[0].ring.sched.ready) {
+		amdgpu_mes_unmap_legacy_queue(adev, ring, action, gpu_addr, seq);
+		return;
+	}
+
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4));
+	amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */
+			  PACKET3_UNMAP_QUEUES_ACTION(action) |
+			  PACKET3_UNMAP_QUEUES_QUEUE_SEL(0) |
+			  PACKET3_UNMAP_QUEUES_ENGINE_SEL(eng_sel) |
+			  PACKET3_UNMAP_QUEUES_NUM_QUEUES(1));
+	amdgpu_ring_write(kiq_ring,
+		  PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index));
+
+	if (action == PREEMPT_QUEUES_NO_UNMAP) {
+		amdgpu_ring_write(kiq_ring, lower_32_bits(gpu_addr));
+		amdgpu_ring_write(kiq_ring, upper_32_bits(gpu_addr));
+		amdgpu_ring_write(kiq_ring, seq);
+	} else {
+		amdgpu_ring_write(kiq_ring, 0);
+		amdgpu_ring_write(kiq_ring, 0);
+		amdgpu_ring_write(kiq_ring, 0);
+	}
+}
+
+static void gfx_v12_0_kiq_query_status(struct amdgpu_ring *kiq_ring,
+				       struct amdgpu_ring *ring,
+				       u64 addr, u64 seq)
+{
+	uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0;
+
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_QUERY_STATUS, 5));
+	amdgpu_ring_write(kiq_ring,
+			  PACKET3_QUERY_STATUS_CONTEXT_ID(0) |
+			  PACKET3_QUERY_STATUS_INTERRUPT_SEL(0) |
+			  PACKET3_QUERY_STATUS_COMMAND(2));
+	amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */
+			  PACKET3_QUERY_STATUS_DOORBELL_OFFSET(ring->doorbell_index) |
+			  PACKET3_QUERY_STATUS_ENG_SEL(eng_sel));
+	amdgpu_ring_write(kiq_ring, lower_32_bits(addr));
+	amdgpu_ring_write(kiq_ring, upper_32_bits(addr));
+	amdgpu_ring_write(kiq_ring, lower_32_bits(seq));
+	amdgpu_ring_write(kiq_ring, upper_32_bits(seq));
+}
+
+static void gfx_v12_0_kiq_invalidate_tlbs(struct amdgpu_ring *kiq_ring,
+					  uint16_t pasid,
+					  uint32_t flush_type,
+					  bool all_hub)
+{
+	gfx_v12_0_ring_invalidate_tlbs(kiq_ring, pasid, flush_type, all_hub, 1);
+}
+
+static const struct kiq_pm4_funcs gfx_v12_0_kiq_pm4_funcs = {
+	.kiq_set_resources = gfx_v12_0_kiq_set_resources,
+	.kiq_map_queues = gfx_v12_0_kiq_map_queues,
+	.kiq_unmap_queues = gfx_v12_0_kiq_unmap_queues,
+	.kiq_query_status = gfx_v12_0_kiq_query_status,
+	.kiq_invalidate_tlbs = gfx_v12_0_kiq_invalidate_tlbs,
+	.set_resources_size = 8,
+	.map_queues_size = 7,
+	.unmap_queues_size = 6,
+	.query_status_size = 7,
+	.invalidate_tlbs_size = 2,
+};
+
+static void gfx_v12_0_set_kiq_pm4_funcs(struct amdgpu_device *adev)
+{
+	adev->gfx.kiq[0].pmf = &gfx_v12_0_kiq_pm4_funcs;
+}
+
+static void gfx_v12_0_wait_reg_mem(struct amdgpu_ring *ring, int eng_sel,
+				   int mem_space, int opt, uint32_t addr0,
+				   uint32_t addr1, uint32_t ref,
+				   uint32_t mask, uint32_t inv)
+{
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5));
+	amdgpu_ring_write(ring,
+			  /* memory (1) or register (0) */
+			  (WAIT_REG_MEM_MEM_SPACE(mem_space) |
+			   WAIT_REG_MEM_OPERATION(opt) | /* wait */
+			   WAIT_REG_MEM_FUNCTION(3) |  /* equal */
+			   WAIT_REG_MEM_ENGINE(eng_sel)));
+
+	if (mem_space)
+		BUG_ON(addr0 & 0x3); /* Dword align */
+	amdgpu_ring_write(ring, addr0);
+	amdgpu_ring_write(ring, addr1);
+	amdgpu_ring_write(ring, ref);
+	amdgpu_ring_write(ring, mask);
+	amdgpu_ring_write(ring, inv); /* poll interval */
+}
+
+static int gfx_v12_0_ring_test_ring(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t scratch = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG0);
+	uint32_t tmp = 0;
+	unsigned i;
+	int r;
+
+	WREG32(scratch, 0xCAFEDEAD);
+	r = amdgpu_ring_alloc(ring, 5);
+	if (r) {
+		dev_err(adev->dev,
+			"amdgpu: cp failed to lock ring %d (%d).\n",
+			ring->idx, r);
+		return r;
+	}
+
+	if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ) {
+		gfx_v12_0_ring_emit_wreg(ring, scratch, 0xDEADBEEF);
+	} else {
+		amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
+		amdgpu_ring_write(ring, scratch -
+				  PACKET3_SET_UCONFIG_REG_START);
+		amdgpu_ring_write(ring, 0xDEADBEEF);
+	}
+	amdgpu_ring_commit(ring);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = RREG32(scratch);
+		if (tmp == 0xDEADBEEF)
+			break;
+		if (amdgpu_emu_mode == 1)
+			msleep(1);
+		else
+			udelay(1);
+	}
+
+	if (i >= adev->usec_timeout)
+		r = -ETIMEDOUT;
+	return r;
+}
+
+static int gfx_v12_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_ib ib;
+	struct dma_fence *f = NULL;
+	unsigned index;
+	uint64_t gpu_addr;
+	uint32_t *cpu_ptr;
+	long r;
+
+	/* MES KIQ fw hasn't indirect buffer support for now */
+	if (adev->enable_mes_kiq &&
+	    ring->funcs->type == AMDGPU_RING_TYPE_KIQ)
+		return 0;
+
+	memset(&ib, 0, sizeof(ib));
+
+	r = amdgpu_device_wb_get(adev, &index);
+	if (r)
+		return r;
+
+	gpu_addr = adev->wb.gpu_addr + (index * 4);
+	adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD);
+	cpu_ptr = &adev->wb.wb[index];
+
+	r = amdgpu_ib_get(adev, NULL, 16, AMDGPU_IB_POOL_DIRECT, &ib);
+	if (r) {
+		dev_err(adev->dev, "amdgpu: failed to get ib (%ld).\n", r);
+		goto err1;
+	}
+
+	ib.ptr[0] = PACKET3(PACKET3_WRITE_DATA, 3);
+	ib.ptr[1] = WRITE_DATA_DST_SEL(5) | WR_CONFIRM;
+	ib.ptr[2] = lower_32_bits(gpu_addr);
+	ib.ptr[3] = upper_32_bits(gpu_addr);
+	ib.ptr[4] = 0xDEADBEEF;
+	ib.length_dw = 5;
+
+	r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f);
+	if (r)
+		goto err2;
+
+	r = dma_fence_wait_timeout(f, false, timeout);
+	if (r == 0) {
+		r = -ETIMEDOUT;
+		goto err2;
+	} else if (r < 0) {
+		goto err2;
+	}
+
+	if (le32_to_cpu(*cpu_ptr) == 0xDEADBEEF)
+		r = 0;
+	else
+		r = -EINVAL;
+err2:
+	amdgpu_ib_free(&ib, NULL);
+	dma_fence_put(f);
+err1:
+	amdgpu_device_wb_free(adev, index);
+	return r;
+}
+
+static void gfx_v12_0_free_microcode(struct amdgpu_device *adev)
+{
+	amdgpu_ucode_release(&adev->gfx.pfp_fw);
+	amdgpu_ucode_release(&adev->gfx.me_fw);
+	amdgpu_ucode_release(&adev->gfx.rlc_fw);
+	amdgpu_ucode_release(&adev->gfx.mec_fw);
+
+	kfree(adev->gfx.rlc.register_list_format);
+}
+
+static int gfx_v12_0_init_toc_microcode(struct amdgpu_device *adev, const char *ucode_prefix)
+{
+	const struct psp_firmware_header_v1_0 *toc_hdr;
+	int err = 0;
+
+	err = amdgpu_ucode_request(adev, &adev->psp.toc_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_toc.bin", ucode_prefix);
+	if (err)
+		goto out;
+
+	toc_hdr = (const struct psp_firmware_header_v1_0 *)adev->psp.toc_fw->data;
+	adev->psp.toc.fw_version = le32_to_cpu(toc_hdr->header.ucode_version);
+	adev->psp.toc.feature_version = le32_to_cpu(toc_hdr->sos.fw_version);
+	adev->psp.toc.size_bytes = le32_to_cpu(toc_hdr->header.ucode_size_bytes);
+	adev->psp.toc.start_addr = (uint8_t *)toc_hdr +
+			le32_to_cpu(toc_hdr->header.ucode_array_offset_bytes);
+	return 0;
+out:
+	amdgpu_ucode_release(&adev->psp.toc_fw);
+	return err;
+}
+
+static int gfx_v12_0_init_microcode(struct amdgpu_device *adev)
+{
+	char ucode_prefix[30];
+	int err;
+	const struct rlc_firmware_header_v2_0 *rlc_hdr;
+	uint16_t version_major;
+	uint16_t version_minor;
+
+	DRM_DEBUG("\n");
+
+	amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix));
+
+	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_pfp.bin", ucode_prefix);
+	if (err)
+		goto out;
+	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_PFP);
+	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_PFP_P0_STACK);
+
+	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_me.bin", ucode_prefix);
+	if (err)
+		goto out;
+	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_ME);
+	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_ME_P0_STACK);
+
+	if (!amdgpu_sriov_vf(adev)) {
+		if (amdgpu_is_kicker_fw(adev))
+			err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_rlc_kicker.bin", ucode_prefix);
+		else
+			err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_rlc.bin", ucode_prefix);
+		if (err)
+			goto out;
+		rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
+		version_major = le16_to_cpu(rlc_hdr->header.header_version_major);
+		version_minor = le16_to_cpu(rlc_hdr->header.header_version_minor);
+		err = amdgpu_gfx_rlc_init_microcode(adev, version_major, version_minor);
+		if (err)
+			goto out;
+	}
+
+	err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_mec.bin", ucode_prefix);
+	if (err)
+		goto out;
+	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_MEC);
+	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_MEC_P0_STACK);
+	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_MEC_P1_STACK);
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO)
+		err = gfx_v12_0_init_toc_microcode(adev, ucode_prefix);
+
+	/* only one MEC for gfx 12 */
+	adev->gfx.mec2_fw = NULL;
+
+	if (adev->gfx.imu.funcs) {
+		if (adev->gfx.imu.funcs->init_microcode) {
+			err = adev->gfx.imu.funcs->init_microcode(adev);
+			if (err)
+				dev_err(adev->dev, "Failed to load imu firmware!\n");
+		}
+	}
+
+out:
+	if (err) {
+		amdgpu_ucode_release(&adev->gfx.pfp_fw);
+		amdgpu_ucode_release(&adev->gfx.me_fw);
+		amdgpu_ucode_release(&adev->gfx.rlc_fw);
+		amdgpu_ucode_release(&adev->gfx.mec_fw);
+	}
+
+	return err;
+}
+
+static u32 gfx_v12_0_get_csb_size(struct amdgpu_device *adev)
+{
+	u32 count = 0;
+	const struct cs_section_def *sect = NULL;
+	const struct cs_extent_def *ext = NULL;
+
+	count += 1;
+
+	for (sect = gfx12_cs_data; sect->section != NULL; ++sect) {
+		if (sect->id == SECT_CONTEXT) {
+			for (ext = sect->section; ext->extent != NULL; ++ext)
+				count += 2 + ext->reg_count;
+		} else
+			return 0;
+	}
+
+	return count;
+}
+
+static void gfx_v12_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer)
+{
+	u32 count = 0, clustercount = 0, i;
+	const struct cs_section_def *sect = NULL;
+	const struct cs_extent_def *ext = NULL;
+
+	if (adev->gfx.rlc.cs_data == NULL)
+		return;
+	if (buffer == NULL)
+		return;
+
+	count += 1;
+
+	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
+		if (sect->id == SECT_CONTEXT) {
+			for (ext = sect->section; ext->extent != NULL; ++ext) {
+				clustercount++;
+				buffer[count++] = ext->reg_count;
+				buffer[count++] = ext->reg_index;
+
+				for (i = 0; i < ext->reg_count; i++)
+					buffer[count++] = cpu_to_le32(ext->extent[i]);
+			}
+		} else
+			return;
+	}
+
+	buffer[0] = clustercount;
+}
+
+static void gfx_v12_0_rlc_fini(struct amdgpu_device *adev)
+{
+	/* clear state block */
+	amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj,
+			&adev->gfx.rlc.clear_state_gpu_addr,
+			(void **)&adev->gfx.rlc.cs_ptr);
+
+	/* jump table block */
+	amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj,
+			&adev->gfx.rlc.cp_table_gpu_addr,
+			(void **)&adev->gfx.rlc.cp_table_ptr);
+}
+
+static void gfx_v12_0_init_rlcg_reg_access_ctrl(struct amdgpu_device *adev)
+{
+	struct amdgpu_rlcg_reg_access_ctrl *reg_access_ctrl;
+
+	reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl[0];
+	reg_access_ctrl->scratch_reg0 = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG0);
+	reg_access_ctrl->scratch_reg1 = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG1);
+	reg_access_ctrl->scratch_reg2 = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG2);
+	reg_access_ctrl->scratch_reg3 = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG3);
+	reg_access_ctrl->grbm_cntl = SOC15_REG_OFFSET(GC, 0, regGRBM_GFX_CNTL);
+	reg_access_ctrl->grbm_idx = SOC15_REG_OFFSET(GC, 0, regGRBM_GFX_INDEX);
+	reg_access_ctrl->spare_int = SOC15_REG_OFFSET(GC, 0, regRLC_SPARE_INT_0);
+	adev->gfx.rlc.rlcg_reg_access_supported = true;
+}
+
+static int gfx_v12_0_rlc_init(struct amdgpu_device *adev)
+{
+	const struct cs_section_def *cs_data;
+	int r;
+
+	adev->gfx.rlc.cs_data = gfx12_cs_data;
+
+	cs_data = adev->gfx.rlc.cs_data;
+
+	if (cs_data) {
+		/* init clear state block */
+		r = amdgpu_gfx_rlc_init_csb(adev);
+		if (r)
+			return r;
+	}
+
+	/* init spm vmid with 0xf */
+	if (adev->gfx.rlc.funcs->update_spm_vmid)
+		adev->gfx.rlc.funcs->update_spm_vmid(adev, NULL, 0xf);
+
+	return 0;
+}
+
+static void gfx_v12_0_mec_fini(struct amdgpu_device *adev)
+{
+	amdgpu_bo_free_kernel(&adev->gfx.mec.hpd_eop_obj, NULL, NULL);
+	amdgpu_bo_free_kernel(&adev->gfx.mec.mec_fw_obj, NULL, NULL);
+	amdgpu_bo_free_kernel(&adev->gfx.mec.mec_fw_data_obj, NULL, NULL);
+}
+
+static void gfx_v12_0_me_init(struct amdgpu_device *adev)
+{
+	bitmap_zero(adev->gfx.me.queue_bitmap, AMDGPU_MAX_GFX_QUEUES);
+
+	amdgpu_gfx_graphics_queue_acquire(adev);
+}
+
+static int gfx_v12_0_mec_init(struct amdgpu_device *adev)
+{
+	int r;
+	u32 *hpd;
+	size_t mec_hpd_size;
+
+	bitmap_zero(adev->gfx.mec_bitmap[0].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
+
+	/* take ownership of the relevant compute queues */
+	amdgpu_gfx_compute_queue_acquire(adev);
+	mec_hpd_size = adev->gfx.num_compute_rings * GFX12_MEC_HPD_SIZE;
+
+	if (mec_hpd_size) {
+		r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE,
+					      AMDGPU_GEM_DOMAIN_GTT,
+					      &adev->gfx.mec.hpd_eop_obj,
+					      &adev->gfx.mec.hpd_eop_gpu_addr,
+					      (void **)&hpd);
+		if (r) {
+			dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r);
+			gfx_v12_0_mec_fini(adev);
+			return r;
+		}
+
+		memset(hpd, 0, mec_hpd_size);
+
+		amdgpu_bo_kunmap(adev->gfx.mec.hpd_eop_obj);
+		amdgpu_bo_unreserve(adev->gfx.mec.hpd_eop_obj);
+	}
+
+	return 0;
+}
+
+static uint32_t wave_read_ind(struct amdgpu_device *adev, uint32_t wave, uint32_t address)
+{
+	WREG32_SOC15(GC, 0, regSQ_IND_INDEX,
+		(wave << SQ_IND_INDEX__WAVE_ID__SHIFT) |
+		(address << SQ_IND_INDEX__INDEX__SHIFT));
+	return RREG32_SOC15(GC, 0, regSQ_IND_DATA);
+}
+
+static void wave_read_regs(struct amdgpu_device *adev, uint32_t wave,
+			   uint32_t thread, uint32_t regno,
+			   uint32_t num, uint32_t *out)
+{
+	WREG32_SOC15(GC, 0, regSQ_IND_INDEX,
+		(wave << SQ_IND_INDEX__WAVE_ID__SHIFT) |
+		(regno << SQ_IND_INDEX__INDEX__SHIFT) |
+		(thread << SQ_IND_INDEX__WORKITEM_ID__SHIFT) |
+		(SQ_IND_INDEX__AUTO_INCR_MASK));
+	while (num--)
+		*(out++) = RREG32_SOC15(GC, 0, regSQ_IND_DATA);
+}
+
+static void gfx_v12_0_read_wave_data(struct amdgpu_device *adev,
+				     uint32_t xcc_id,
+				     uint32_t simd, uint32_t wave,
+				     uint32_t *dst, int *no_fields)
+{
+	/* in gfx12 the SIMD_ID is specified as part of the INSTANCE
+	 * field when performing a select_se_sh so it should be
+	 * zero here */
+	WARN_ON(simd != 0);
+
+	/* type 4 wave data */
+	dst[(*no_fields)++] = 4;
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_STATUS);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_PC_LO);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_PC_HI);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_EXEC_LO);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_EXEC_HI);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_HW_ID1);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_HW_ID2);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_GPR_ALLOC);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_LDS_ALLOC);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_IB_STS);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_IB_STS2);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_IB_DBG1);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_M0);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_MODE);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_STATE_PRIV);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_EXCP_FLAG_PRIV);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_EXCP_FLAG_USER);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_TRAP_CTRL);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_ACTIVE);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_VALID_AND_IDLE);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_DVGPR_ALLOC_LO);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_DVGPR_ALLOC_HI);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_SCHED_MODE);
+}
+
+static void gfx_v12_0_read_wave_sgprs(struct amdgpu_device *adev,
+				      uint32_t xcc_id, uint32_t simd,
+				      uint32_t wave, uint32_t start,
+				      uint32_t size, uint32_t *dst)
+{
+	WARN_ON(simd != 0);
+
+	wave_read_regs(
+		adev, wave, 0, start + SQIND_WAVE_SGPRS_OFFSET, size,
+		dst);
+}
+
+static void gfx_v12_0_read_wave_vgprs(struct amdgpu_device *adev,
+				      uint32_t xcc_id, uint32_t simd,
+				      uint32_t wave, uint32_t thread,
+				      uint32_t start, uint32_t size,
+				      uint32_t *dst)
+{
+	wave_read_regs(
+		adev, wave, thread,
+		start + SQIND_WAVE_VGPRS_OFFSET, size, dst);
+}
+
+static void gfx_v12_0_select_me_pipe_q(struct amdgpu_device *adev,
+				       u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id)
+{
+	soc24_grbm_select(adev, me, pipe, q, vm);
+}
+
+/* all sizes are in bytes */
+#define MQD_SHADOW_BASE_SIZE      73728
+#define MQD_SHADOW_BASE_ALIGNMENT 256
+#define MQD_FWWORKAREA_SIZE       484
+#define MQD_FWWORKAREA_ALIGNMENT  256
+
+static void gfx_v12_0_get_gfx_shadow_info_nocheck(struct amdgpu_device *adev,
+						  struct amdgpu_gfx_shadow_info *shadow_info)
+{
+	shadow_info->shadow_size = MQD_SHADOW_BASE_SIZE;
+	shadow_info->shadow_alignment = MQD_SHADOW_BASE_ALIGNMENT;
+	shadow_info->csa_size = MQD_FWWORKAREA_SIZE;
+	shadow_info->csa_alignment = MQD_FWWORKAREA_ALIGNMENT;
+}
+
+static int gfx_v12_0_get_gfx_shadow_info(struct amdgpu_device *adev,
+					 struct amdgpu_gfx_shadow_info *shadow_info,
+					 bool skip_check)
+{
+	if (adev->gfx.cp_gfx_shadow || skip_check) {
+		gfx_v12_0_get_gfx_shadow_info_nocheck(adev, shadow_info);
+		return 0;
+	}
+
+	memset(shadow_info, 0, sizeof(struct amdgpu_gfx_shadow_info));
+	return -EINVAL;
+}
+
+static const struct amdgpu_gfx_funcs gfx_v12_0_gfx_funcs = {
+	.get_gpu_clock_counter = &gfx_v12_0_get_gpu_clock_counter,
+	.select_se_sh = &gfx_v12_0_select_se_sh,
+	.read_wave_data = &gfx_v12_0_read_wave_data,
+	.read_wave_sgprs = &gfx_v12_0_read_wave_sgprs,
+	.read_wave_vgprs = &gfx_v12_0_read_wave_vgprs,
+	.select_me_pipe_q = &gfx_v12_0_select_me_pipe_q,
+	.update_perfmon_mgcg = &gfx_v12_0_update_perf_clk,
+	.get_gfx_shadow_info = &gfx_v12_0_get_gfx_shadow_info,
+};
+
+static int gfx_v12_0_gpu_early_init(struct amdgpu_device *adev)
+{
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		adev->gfx.config.max_hw_contexts = 8;
+		adev->gfx.config.sc_prim_fifo_size_frontend = 0x20;
+		adev->gfx.config.sc_prim_fifo_size_backend = 0x100;
+		adev->gfx.config.sc_hiz_tile_fifo_size = 0;
+		adev->gfx.config.sc_earlyz_tile_fifo_size = 0x4C0;
+		break;
+	default:
+		BUG();
+		break;
+	}
+
+	return 0;
+}
+
+static int gfx_v12_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id,
+				   int me, int pipe, int queue)
+{
+	int r;
+	struct amdgpu_ring *ring;
+	unsigned int irq_type;
+
+	ring = &adev->gfx.gfx_ring[ring_id];
+
+	ring->me = me;
+	ring->pipe = pipe;
+	ring->queue = queue;
+
+	ring->ring_obj = NULL;
+	ring->use_doorbell = true;
+
+	if (!ring_id)
+		ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
+	else
+		ring->doorbell_index = adev->doorbell_index.gfx_ring1 << 1;
+	ring->vm_hub = AMDGPU_GFXHUB(0);
+	sprintf(ring->name, "gfx_%d.%d.%d", ring->me, ring->pipe, ring->queue);
+
+	irq_type = AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP + ring->pipe;
+	r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type,
+			     AMDGPU_RING_PRIO_DEFAULT, NULL);
+	if (r)
+		return r;
+	return 0;
+}
+
+static int gfx_v12_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
+				       int mec, int pipe, int queue)
+{
+	int r;
+	unsigned irq_type;
+	struct amdgpu_ring *ring;
+	unsigned int hw_prio;
+
+	ring = &adev->gfx.compute_ring[ring_id];
+
+	/* mec0 is me1 */
+	ring->me = mec + 1;
+	ring->pipe = pipe;
+	ring->queue = queue;
+
+	ring->ring_obj = NULL;
+	ring->use_doorbell = true;
+	ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
+	ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
+				+ (ring_id * GFX12_MEC_HPD_SIZE);
+	ring->vm_hub = AMDGPU_GFXHUB(0);
+	sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
+
+	irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
+		+ ((ring->me - 1) * adev->gfx.mec.num_pipe_per_mec)
+		+ ring->pipe;
+	hw_prio = amdgpu_gfx_is_high_priority_compute_queue(adev, ring) ?
+			AMDGPU_GFX_PIPE_PRIO_HIGH : AMDGPU_GFX_PIPE_PRIO_NORMAL;
+	/* type-2 packets are deprecated on MEC, use type-3 instead */
+	r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type,
+			     hw_prio, NULL);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static struct {
+	SOC24_FIRMWARE_ID	id;
+	unsigned int		offset;
+	unsigned int		size;
+	unsigned int		size_x16;
+} rlc_autoload_info[SOC24_FIRMWARE_ID_MAX];
+
+#define RLC_TOC_OFFSET_DWUNIT   8
+#define RLC_SIZE_MULTIPLE       1024
+#define RLC_TOC_UMF_SIZE_inM	23ULL
+#define RLC_TOC_FORMAT_API	165ULL
+
+static void gfx_v12_0_parse_rlc_toc(struct amdgpu_device *adev, void *rlc_toc)
+{
+	RLC_TABLE_OF_CONTENT_V2 *ucode = rlc_toc;
+
+	while (ucode && (ucode->id > SOC24_FIRMWARE_ID_INVALID)) {
+		rlc_autoload_info[ucode->id].id = ucode->id;
+		rlc_autoload_info[ucode->id].offset =
+			ucode->offset * RLC_TOC_OFFSET_DWUNIT * 4;
+		rlc_autoload_info[ucode->id].size =
+			ucode->size_x16 ? ucode->size * RLC_SIZE_MULTIPLE * 4 :
+					  ucode->size * 4;
+		ucode++;
+	}
+}
+
+static uint32_t gfx_v12_0_calc_toc_total_size(struct amdgpu_device *adev)
+{
+	uint32_t total_size = 0;
+	SOC24_FIRMWARE_ID id;
+
+	gfx_v12_0_parse_rlc_toc(adev, adev->psp.toc.start_addr);
+
+	for (id = SOC24_FIRMWARE_ID_RLC_G_UCODE; id < SOC24_FIRMWARE_ID_MAX; id++)
+		total_size += rlc_autoload_info[id].size;
+
+	/* In case the offset in rlc toc ucode is aligned */
+	if (total_size < rlc_autoload_info[SOC24_FIRMWARE_ID_MAX-1].offset)
+		total_size = rlc_autoload_info[SOC24_FIRMWARE_ID_MAX-1].offset +
+			rlc_autoload_info[SOC24_FIRMWARE_ID_MAX-1].size;
+	if (total_size < (RLC_TOC_UMF_SIZE_inM << 20))
+		total_size = RLC_TOC_UMF_SIZE_inM << 20;
+
+	return total_size;
+}
+
+static int gfx_v12_0_rlc_autoload_buffer_init(struct amdgpu_device *adev)
+{
+	int r;
+	uint32_t total_size;
+
+	total_size = gfx_v12_0_calc_toc_total_size(adev);
+
+	r = amdgpu_bo_create_reserved(adev, total_size, 64 * 1024,
+				      AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->gfx.rlc.rlc_autoload_bo,
+				      &adev->gfx.rlc.rlc_autoload_gpu_addr,
+				      (void **)&adev->gfx.rlc.rlc_autoload_ptr);
+
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create fw autoload bo\n", r);
+		return r;
+	}
+
+	return 0;
+}
+
+static void gfx_v12_0_rlc_backdoor_autoload_copy_ucode(struct amdgpu_device *adev,
+						       SOC24_FIRMWARE_ID id,
+						       const void *fw_data,
+						       uint32_t fw_size)
+{
+	uint32_t toc_offset;
+	uint32_t toc_fw_size;
+	char *ptr = adev->gfx.rlc.rlc_autoload_ptr;
+
+	if (id <= SOC24_FIRMWARE_ID_INVALID || id >= SOC24_FIRMWARE_ID_MAX)
+		return;
+
+	toc_offset = rlc_autoload_info[id].offset;
+	toc_fw_size = rlc_autoload_info[id].size;
+
+	if (fw_size == 0)
+		fw_size = toc_fw_size;
+
+	if (fw_size > toc_fw_size)
+		fw_size = toc_fw_size;
+
+	memcpy(ptr + toc_offset, fw_data, fw_size);
+
+	if (fw_size < toc_fw_size)
+		memset(ptr + toc_offset + fw_size, 0, toc_fw_size - fw_size);
+}
+
+static void
+gfx_v12_0_rlc_backdoor_autoload_copy_toc_ucode(struct amdgpu_device *adev)
+{
+	void *data;
+	uint32_t size;
+	uint32_t *toc_ptr;
+
+	data = adev->psp.toc.start_addr;
+	size = rlc_autoload_info[SOC24_FIRMWARE_ID_RLC_TOC].size;
+
+	toc_ptr = (uint32_t *)data + size / 4 - 2;
+	*toc_ptr = (RLC_TOC_FORMAT_API << 24) | 0x1;
+
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RLC_TOC,
+						   data, size);
+}
+
+static void
+gfx_v12_0_rlc_backdoor_autoload_copy_gfx_ucode(struct amdgpu_device *adev)
+{
+	const __le32 *fw_data;
+	uint32_t fw_size;
+	const struct gfx_firmware_header_v2_0 *cpv2_hdr;
+	const struct rlc_firmware_header_v2_0 *rlc_hdr;
+	const struct rlc_firmware_header_v2_1 *rlcv21_hdr;
+	const struct rlc_firmware_header_v2_2 *rlcv22_hdr;
+	uint16_t version_major, version_minor;
+
+	/* pfp ucode */
+	cpv2_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.pfp_fw->data;
+	/* instruction */
+	fw_data = (const __le32 *)(adev->gfx.pfp_fw->data +
+		le32_to_cpu(cpv2_hdr->ucode_offset_bytes));
+	fw_size = le32_to_cpu(cpv2_hdr->ucode_size_bytes);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_PFP,
+						   fw_data, fw_size);
+	/* data */
+	fw_data = (const __le32 *)(adev->gfx.pfp_fw->data +
+		le32_to_cpu(cpv2_hdr->data_offset_bytes));
+	fw_size = le32_to_cpu(cpv2_hdr->data_size_bytes);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_PFP_P0_STACK,
+						   fw_data, fw_size);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_PFP_P1_STACK,
+						   fw_data, fw_size);
+	/* me ucode */
+	cpv2_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.me_fw->data;
+	/* instruction */
+	fw_data = (const __le32 *)(adev->gfx.me_fw->data +
+		le32_to_cpu(cpv2_hdr->ucode_offset_bytes));
+	fw_size = le32_to_cpu(cpv2_hdr->ucode_size_bytes);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_ME,
+						   fw_data, fw_size);
+	/* data */
+	fw_data = (const __le32 *)(adev->gfx.me_fw->data +
+		le32_to_cpu(cpv2_hdr->data_offset_bytes));
+	fw_size = le32_to_cpu(cpv2_hdr->data_size_bytes);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_ME_P0_STACK,
+						   fw_data, fw_size);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_ME_P1_STACK,
+						   fw_data, fw_size);
+	/* mec ucode */
+	cpv2_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.mec_fw->data;
+	/* instruction */
+	fw_data = (const __le32 *) (adev->gfx.mec_fw->data +
+		le32_to_cpu(cpv2_hdr->ucode_offset_bytes));
+	fw_size = le32_to_cpu(cpv2_hdr->ucode_size_bytes);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_MEC,
+						   fw_data, fw_size);
+	/* data */
+	fw_data = (const __le32 *) (adev->gfx.mec_fw->data +
+		le32_to_cpu(cpv2_hdr->data_offset_bytes));
+	fw_size = le32_to_cpu(cpv2_hdr->data_size_bytes);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_MEC_P0_STACK,
+						   fw_data, fw_size);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_MEC_P1_STACK,
+						   fw_data, fw_size);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_MEC_P2_STACK,
+						   fw_data, fw_size);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_MEC_P3_STACK,
+						   fw_data, fw_size);
+
+	/* rlc ucode */
+	rlc_hdr = (const struct rlc_firmware_header_v2_0 *)
+		adev->gfx.rlc_fw->data;
+	fw_data = (const __le32 *)(adev->gfx.rlc_fw->data +
+			le32_to_cpu(rlc_hdr->header.ucode_array_offset_bytes));
+	fw_size = le32_to_cpu(rlc_hdr->header.ucode_size_bytes);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RLC_G_UCODE,
+						   fw_data, fw_size);
+
+	version_major = le16_to_cpu(rlc_hdr->header.header_version_major);
+	version_minor = le16_to_cpu(rlc_hdr->header.header_version_minor);
+	if (version_major == 2) {
+		if (version_minor >= 1) {
+			rlcv21_hdr = (const struct rlc_firmware_header_v2_1 *)adev->gfx.rlc_fw->data;
+
+			fw_data = (const __le32 *)(adev->gfx.rlc_fw->data +
+					le32_to_cpu(rlcv21_hdr->save_restore_list_gpm_offset_bytes));
+			fw_size = le32_to_cpu(rlcv21_hdr->save_restore_list_gpm_size_bytes);
+			gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RLCG_SCRATCH,
+						   fw_data, fw_size);
+
+			fw_data = (const __le32 *)(adev->gfx.rlc_fw->data +
+					le32_to_cpu(rlcv21_hdr->save_restore_list_srm_offset_bytes));
+			fw_size = le32_to_cpu(rlcv21_hdr->save_restore_list_srm_size_bytes);
+			gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RLC_SRM_ARAM,
+						   fw_data, fw_size);
+		}
+		if (version_minor >= 2) {
+			rlcv22_hdr = (const struct rlc_firmware_header_v2_2 *)adev->gfx.rlc_fw->data;
+
+			fw_data = (const __le32 *)(adev->gfx.rlc_fw->data +
+					le32_to_cpu(rlcv22_hdr->rlc_iram_ucode_offset_bytes));
+			fw_size = le32_to_cpu(rlcv22_hdr->rlc_iram_ucode_size_bytes);
+			gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RLX6_UCODE,
+						   fw_data, fw_size);
+
+			fw_data = (const __le32 *)(adev->gfx.rlc_fw->data +
+					le32_to_cpu(rlcv22_hdr->rlc_dram_ucode_offset_bytes));
+			fw_size = le32_to_cpu(rlcv22_hdr->rlc_dram_ucode_size_bytes);
+			gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RLX6_DRAM_BOOT,
+						   fw_data, fw_size);
+		}
+	}
+}
+
+static void
+gfx_v12_0_rlc_backdoor_autoload_copy_sdma_ucode(struct amdgpu_device *adev)
+{
+	const __le32 *fw_data;
+	uint32_t fw_size;
+	const struct sdma_firmware_header_v3_0 *sdma_hdr;
+
+	sdma_hdr = (const struct sdma_firmware_header_v3_0 *)
+		adev->sdma.instance[0].fw->data;
+	fw_data = (const __le32 *) (adev->sdma.instance[0].fw->data +
+			le32_to_cpu(sdma_hdr->ucode_offset_bytes));
+	fw_size = le32_to_cpu(sdma_hdr->ucode_size_bytes);
+
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_SDMA_UCODE_TH0,
+						   fw_data, fw_size);
+}
+
+static void
+gfx_v12_0_rlc_backdoor_autoload_copy_mes_ucode(struct amdgpu_device *adev)
+{
+	const __le32 *fw_data;
+	unsigned fw_size;
+	const struct mes_firmware_header_v1_0 *mes_hdr;
+	int pipe, ucode_id, data_id;
+
+	for (pipe = 0; pipe < 2; pipe++) {
+		if (pipe == 0) {
+			ucode_id = SOC24_FIRMWARE_ID_RS64_MES_P0;
+			data_id  = SOC24_FIRMWARE_ID_RS64_MES_P0_STACK;
+		} else {
+			ucode_id = SOC24_FIRMWARE_ID_RS64_MES_P1;
+			data_id  = SOC24_FIRMWARE_ID_RS64_MES_P1_STACK;
+		}
+
+		mes_hdr = (const struct mes_firmware_header_v1_0 *)
+			adev->mes.fw[pipe]->data;
+
+		fw_data = (const __le32 *)(adev->mes.fw[pipe]->data +
+				le32_to_cpu(mes_hdr->mes_ucode_offset_bytes));
+		fw_size = le32_to_cpu(mes_hdr->mes_ucode_size_bytes);
+
+		gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, ucode_id, fw_data, fw_size);
+
+		fw_data = (const __le32 *)(adev->mes.fw[pipe]->data +
+				le32_to_cpu(mes_hdr->mes_ucode_data_offset_bytes));
+		fw_size = le32_to_cpu(mes_hdr->mes_ucode_data_size_bytes);
+
+		gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, data_id, fw_data, fw_size);
+	}
+}
+
+static int gfx_v12_0_rlc_backdoor_autoload_enable(struct amdgpu_device *adev)
+{
+	uint32_t rlc_g_offset, rlc_g_size;
+	uint64_t gpu_addr;
+	uint32_t data;
+
+	/* RLC autoload sequence 2: copy ucode */
+	gfx_v12_0_rlc_backdoor_autoload_copy_sdma_ucode(adev);
+	gfx_v12_0_rlc_backdoor_autoload_copy_gfx_ucode(adev);
+	gfx_v12_0_rlc_backdoor_autoload_copy_mes_ucode(adev);
+	gfx_v12_0_rlc_backdoor_autoload_copy_toc_ucode(adev);
+
+	rlc_g_offset = rlc_autoload_info[SOC24_FIRMWARE_ID_RLC_G_UCODE].offset;
+	rlc_g_size = rlc_autoload_info[SOC24_FIRMWARE_ID_RLC_G_UCODE].size;
+	gpu_addr = adev->gfx.rlc.rlc_autoload_gpu_addr + rlc_g_offset - adev->gmc.vram_start;
+
+	WREG32_SOC15(GC, 0, regGFX_IMU_RLC_BOOTLOADER_ADDR_HI, upper_32_bits(gpu_addr));
+	WREG32_SOC15(GC, 0, regGFX_IMU_RLC_BOOTLOADER_ADDR_LO, lower_32_bits(gpu_addr));
+
+	WREG32_SOC15(GC, 0, regGFX_IMU_RLC_BOOTLOADER_SIZE, rlc_g_size);
+
+	if (adev->gfx.imu.funcs && (amdgpu_dpm > 0)) {
+		/* RLC autoload sequence 3: load IMU fw */
+		if (adev->gfx.imu.funcs->load_microcode)
+			adev->gfx.imu.funcs->load_microcode(adev);
+		/* RLC autoload sequence 4 init IMU fw */
+		if (adev->gfx.imu.funcs->setup_imu)
+			adev->gfx.imu.funcs->setup_imu(adev);
+		if (adev->gfx.imu.funcs->start_imu)
+			adev->gfx.imu.funcs->start_imu(adev);
+
+		/* RLC autoload sequence 5 disable gpa mode */
+		gfx_v12_0_disable_gpa_mode(adev);
+	} else {
+		/* unhalt rlc to start autoload without imu */
+		data = RREG32_SOC15(GC, 0, regRLC_GPM_THREAD_ENABLE);
+		data = REG_SET_FIELD(data, RLC_GPM_THREAD_ENABLE, THREAD0_ENABLE, 1);
+		data = REG_SET_FIELD(data, RLC_GPM_THREAD_ENABLE, THREAD1_ENABLE, 1);
+		WREG32_SOC15(GC, 0, regRLC_GPM_THREAD_ENABLE, data);
+		WREG32_SOC15(GC, 0, regRLC_CNTL, RLC_CNTL__RLC_ENABLE_F32_MASK);
+	}
+
+	return 0;
+}
+
+static void gfx_v12_0_alloc_ip_dump(struct amdgpu_device *adev)
+{
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_12_0);
+	uint32_t *ptr;
+	uint32_t inst;
+
+	ptr = kcalloc(reg_count, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for GFX IP Dump\n");
+		adev->gfx.ip_dump_core = NULL;
+	} else {
+		adev->gfx.ip_dump_core = ptr;
+	}
+
+	/* Allocate memory for compute queue registers for all the instances */
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_12);
+	inst = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe_per_mec *
+		adev->gfx.mec.num_queue_per_pipe;
+
+	ptr = kcalloc(reg_count * inst, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for Compute Queues IP Dump\n");
+		adev->gfx.ip_dump_compute_queues = NULL;
+	} else {
+		adev->gfx.ip_dump_compute_queues = ptr;
+	}
+
+	/* Allocate memory for gfx queue registers for all the instances */
+	reg_count = ARRAY_SIZE(gc_gfx_queue_reg_list_12);
+	inst = adev->gfx.me.num_me * adev->gfx.me.num_pipe_per_me *
+		adev->gfx.me.num_queue_per_pipe;
+
+	ptr = kcalloc(reg_count * inst, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for GFX Queues IP Dump\n");
+		adev->gfx.ip_dump_gfx_queues = NULL;
+	} else {
+		adev->gfx.ip_dump_gfx_queues = ptr;
+	}
+}
+
+static int gfx_v12_0_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	int i, j, k, r, ring_id = 0;
+	unsigned num_compute_rings;
+	int xcc_id = 0;
+	struct amdgpu_device *adev = ip_block->adev;
+	int num_queue_per_pipe = 1; /* we only enable 1 KGQ per pipe */
+
+	INIT_DELAYED_WORK(&adev->gfx.idle_work, amdgpu_gfx_profile_idle_work_handler);
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		adev->gfx.me.num_me = 1;
+		adev->gfx.me.num_pipe_per_me = 1;
+		adev->gfx.me.num_queue_per_pipe = 8;
+		adev->gfx.mec.num_mec = 1;
+		adev->gfx.mec.num_pipe_per_mec = 2;
+		adev->gfx.mec.num_queue_per_pipe = 4;
+		break;
+	default:
+		adev->gfx.me.num_me = 1;
+		adev->gfx.me.num_pipe_per_me = 1;
+		adev->gfx.me.num_queue_per_pipe = 1;
+		adev->gfx.mec.num_mec = 1;
+		adev->gfx.mec.num_pipe_per_mec = 4;
+		adev->gfx.mec.num_queue_per_pipe = 8;
+		break;
+	}
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		if (!adev->gfx.disable_uq &&
+		    adev->gfx.me_fw_version  >= 2780 &&
+		    adev->gfx.pfp_fw_version >= 2840 &&
+		    adev->gfx.mec_fw_version >= 3050 &&
+		    adev->mes.fw_version[0] >= 123) {
+			adev->userq_funcs[AMDGPU_HW_IP_GFX] = &userq_mes_funcs;
+			adev->userq_funcs[AMDGPU_HW_IP_COMPUTE] = &userq_mes_funcs;
+		}
+		break;
+	default:
+		break;
+	}
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		if (adev->gfx.me_fw_version  >= 2480 &&
+		    adev->gfx.pfp_fw_version >= 2530 &&
+		    adev->gfx.mec_fw_version >= 2680 &&
+		    adev->mes.fw_version[0] >= 100)
+			adev->gfx.enable_cleaner_shader = true;
+		break;
+	default:
+		adev->gfx.enable_cleaner_shader = false;
+		break;
+	}
+
+	if (adev->gfx.num_compute_rings) {
+		/* recalculate compute rings to use based on hardware configuration */
+		num_compute_rings = (adev->gfx.mec.num_pipe_per_mec *
+				     adev->gfx.mec.num_queue_per_pipe) / 2;
+		adev->gfx.num_compute_rings = min(adev->gfx.num_compute_rings,
+						  num_compute_rings);
+	}
+
+	/* EOP Event */
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP,
+			      GFX_12_0_0__SRCID__CP_EOP_INTERRUPT,
+			      &adev->gfx.eop_irq);
+	if (r)
+		return r;
+
+	/* Bad opcode Event */
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP,
+			      GFX_12_0_0__SRCID__CP_BAD_OPCODE_ERROR,
+			      &adev->gfx.bad_op_irq);
+	if (r)
+		return r;
+
+	/* Privileged reg */
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP,
+			      GFX_12_0_0__SRCID__CP_PRIV_REG_FAULT,
+			      &adev->gfx.priv_reg_irq);
+	if (r)
+		return r;
+
+	/* Privileged inst */
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP,
+			      GFX_12_0_0__SRCID__CP_PRIV_INSTR_FAULT,
+			      &adev->gfx.priv_inst_irq);
+	if (r)
+		return r;
+
+	adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE;
+
+	gfx_v12_0_me_init(adev);
+
+	r = gfx_v12_0_rlc_init(adev);
+	if (r) {
+		dev_err(adev->dev, "Failed to init rlc BOs!\n");
+		return r;
+	}
+
+	r = gfx_v12_0_mec_init(adev);
+	if (r) {
+		dev_err(adev->dev, "Failed to init MEC BOs!\n");
+		return r;
+	}
+
+	if (adev->gfx.num_gfx_rings) {
+		/* set up the gfx ring */
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < num_queue_per_pipe; j++) {
+				for (k = 0; k < adev->gfx.me.num_pipe_per_me; k++) {
+					if (!amdgpu_gfx_is_me_queue_enabled(adev, i, k, j))
+						continue;
+
+					r = gfx_v12_0_gfx_ring_init(adev, ring_id,
+								    i, k, j);
+					if (r)
+						return r;
+					ring_id++;
+				}
+			}
+		}
+	}
+
+	if (adev->gfx.num_compute_rings) {
+		ring_id = 0;
+		/* set up the compute queues - allocate horizontally across pipes */
+		for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
+			for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
+				for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
+					if (!amdgpu_gfx_is_mec_queue_enabled(adev,
+									     0, i, k, j))
+						continue;
+
+					r = gfx_v12_0_compute_ring_init(adev, ring_id,
+									i, k, j);
+					if (r)
+						return r;
+
+					ring_id++;
+				}
+			}
+		}
+	}
+
+	adev->gfx.gfx_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]);
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		if ((adev->gfx.me_fw_version >= 2660) &&
+		    (adev->gfx.mec_fw_version >= 2920) &&
+		    !amdgpu_sriov_vf(adev) &&
+		    !adev->debug_disable_gpu_ring_reset) {
+			adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+			adev->gfx.gfx_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+		}
+		break;
+	default:
+		break;
+	}
+
+	if (!adev->enable_mes_kiq) {
+		r = amdgpu_gfx_kiq_init(adev, GFX12_MEC_HPD_SIZE, 0);
+		if (r) {
+			dev_err(adev->dev, "Failed to init KIQ BOs!\n");
+			return r;
+		}
+
+		r = amdgpu_gfx_kiq_init_ring(adev, xcc_id);
+		if (r)
+			return r;
+	}
+
+	r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct v12_compute_mqd), 0);
+	if (r)
+		return r;
+
+	/* allocate visible FB for rlc auto-loading fw */
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) {
+		r = gfx_v12_0_rlc_autoload_buffer_init(adev);
+		if (r)
+			return r;
+	}
+
+	r = gfx_v12_0_gpu_early_init(adev);
+	if (r)
+		return r;
+
+	gfx_v12_0_alloc_ip_dump(adev);
+
+	r = amdgpu_gfx_sysfs_init(adev);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static void gfx_v12_0_pfp_fini(struct amdgpu_device *adev)
+{
+	amdgpu_bo_free_kernel(&adev->gfx.pfp.pfp_fw_obj,
+			      &adev->gfx.pfp.pfp_fw_gpu_addr,
+			      (void **)&adev->gfx.pfp.pfp_fw_ptr);
+
+	amdgpu_bo_free_kernel(&adev->gfx.pfp.pfp_fw_data_obj,
+			      &adev->gfx.pfp.pfp_fw_data_gpu_addr,
+			      (void **)&adev->gfx.pfp.pfp_fw_data_ptr);
+}
+
+static void gfx_v12_0_me_fini(struct amdgpu_device *adev)
+{
+	amdgpu_bo_free_kernel(&adev->gfx.me.me_fw_obj,
+			      &adev->gfx.me.me_fw_gpu_addr,
+			      (void **)&adev->gfx.me.me_fw_ptr);
+
+	amdgpu_bo_free_kernel(&adev->gfx.me.me_fw_data_obj,
+			       &adev->gfx.me.me_fw_data_gpu_addr,
+			       (void **)&adev->gfx.me.me_fw_data_ptr);
+}
+
+static void gfx_v12_0_rlc_autoload_buffer_fini(struct amdgpu_device *adev)
+{
+	amdgpu_bo_free_kernel(&adev->gfx.rlc.rlc_autoload_bo,
+			&adev->gfx.rlc.rlc_autoload_gpu_addr,
+			(void **)&adev->gfx.rlc.rlc_autoload_ptr);
+}
+
+static int gfx_v12_0_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	int i;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
+		amdgpu_ring_fini(&adev->gfx.gfx_ring[i]);
+	for (i = 0; i < adev->gfx.num_compute_rings; i++)
+		amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
+
+	amdgpu_gfx_mqd_sw_fini(adev, 0);
+
+	if (!adev->enable_mes_kiq) {
+		amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq[0].ring);
+		amdgpu_gfx_kiq_fini(adev, 0);
+	}
+
+	gfx_v12_0_pfp_fini(adev);
+	gfx_v12_0_me_fini(adev);
+	gfx_v12_0_rlc_fini(adev);
+	gfx_v12_0_mec_fini(adev);
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO)
+		gfx_v12_0_rlc_autoload_buffer_fini(adev);
+
+	gfx_v12_0_free_microcode(adev);
+
+	amdgpu_gfx_sysfs_fini(adev);
+
+	kfree(adev->gfx.ip_dump_core);
+	kfree(adev->gfx.ip_dump_compute_queues);
+	kfree(adev->gfx.ip_dump_gfx_queues);
+
+	return 0;
+}
+
+static void gfx_v12_0_select_se_sh(struct amdgpu_device *adev, u32 se_num,
+				   u32 sh_num, u32 instance, int xcc_id)
+{
+	u32 data;
+
+	if (instance == 0xffffffff)
+		data = REG_SET_FIELD(0, GRBM_GFX_INDEX,
+				     INSTANCE_BROADCAST_WRITES, 1);
+	else
+		data = REG_SET_FIELD(0, GRBM_GFX_INDEX, INSTANCE_INDEX,
+				     instance);
+
+	if (se_num == 0xffffffff)
+		data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_BROADCAST_WRITES,
+				     1);
+	else
+		data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_INDEX, se_num);
+
+	if (sh_num == 0xffffffff)
+		data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SA_BROADCAST_WRITES,
+				     1);
+	else
+		data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SA_INDEX, sh_num);
+
+	WREG32_SOC15(GC, 0, regGRBM_GFX_INDEX, data);
+}
+
+static u32 gfx_v12_0_get_sa_active_bitmap(struct amdgpu_device *adev)
+{
+	u32 gc_disabled_sa_mask, gc_user_disabled_sa_mask, sa_mask;
+
+	gc_disabled_sa_mask = RREG32_SOC15(GC, 0, regGRBM_CC_GC_SA_UNIT_DISABLE);
+	gc_disabled_sa_mask = REG_GET_FIELD(gc_disabled_sa_mask,
+					    GRBM_CC_GC_SA_UNIT_DISABLE,
+					    SA_DISABLE);
+	gc_user_disabled_sa_mask = RREG32_SOC15(GC, 0, regGRBM_GC_USER_SA_UNIT_DISABLE);
+	gc_user_disabled_sa_mask = REG_GET_FIELD(gc_user_disabled_sa_mask,
+						 GRBM_GC_USER_SA_UNIT_DISABLE,
+						 SA_DISABLE);
+	sa_mask = amdgpu_gfx_create_bitmask(adev->gfx.config.max_sh_per_se *
+					    adev->gfx.config.max_shader_engines);
+
+	return sa_mask & (~(gc_disabled_sa_mask | gc_user_disabled_sa_mask));
+}
+
+static u32 gfx_v12_0_get_rb_active_bitmap(struct amdgpu_device *adev)
+{
+	u32 gc_disabled_rb_mask, gc_user_disabled_rb_mask;
+	u32 rb_mask;
+
+	gc_disabled_rb_mask = RREG32_SOC15(GC, 0, regCC_RB_BACKEND_DISABLE);
+	gc_disabled_rb_mask = REG_GET_FIELD(gc_disabled_rb_mask,
+					    CC_RB_BACKEND_DISABLE,
+					    BACKEND_DISABLE);
+	gc_user_disabled_rb_mask = RREG32_SOC15(GC, 0, regGC_USER_RB_BACKEND_DISABLE);
+	gc_user_disabled_rb_mask = REG_GET_FIELD(gc_user_disabled_rb_mask,
+						 GC_USER_RB_BACKEND_DISABLE,
+						 BACKEND_DISABLE);
+	rb_mask = amdgpu_gfx_create_bitmask(adev->gfx.config.max_backends_per_se *
+					    adev->gfx.config.max_shader_engines);
+
+	return rb_mask & (~(gc_disabled_rb_mask | gc_user_disabled_rb_mask));
+}
+
+static void gfx_v12_0_setup_rb(struct amdgpu_device *adev)
+{
+	u32 rb_bitmap_per_sa;
+	u32 rb_bitmap_width_per_sa;
+	u32 max_sa;
+	u32 active_sa_bitmap;
+	u32 global_active_rb_bitmap;
+	u32 active_rb_bitmap = 0;
+	u32 i;
+
+	/* query sa bitmap from SA_UNIT_DISABLE registers */
+	active_sa_bitmap = gfx_v12_0_get_sa_active_bitmap(adev);
+	/* query rb bitmap from RB_BACKEND_DISABLE registers */
+	global_active_rb_bitmap = gfx_v12_0_get_rb_active_bitmap(adev);
+
+	/* generate active rb bitmap according to active sa bitmap */
+	max_sa = adev->gfx.config.max_shader_engines *
+		 adev->gfx.config.max_sh_per_se;
+	rb_bitmap_width_per_sa = adev->gfx.config.max_backends_per_se /
+				 adev->gfx.config.max_sh_per_se;
+	rb_bitmap_per_sa = amdgpu_gfx_create_bitmask(rb_bitmap_width_per_sa);
+
+	for (i = 0; i < max_sa; i++) {
+		if (active_sa_bitmap & (1 << i))
+			active_rb_bitmap |= (rb_bitmap_per_sa << (i * rb_bitmap_width_per_sa));
+	}
+
+	active_rb_bitmap &= global_active_rb_bitmap;
+	adev->gfx.config.backend_enable_mask = active_rb_bitmap;
+	adev->gfx.config.num_rbs = hweight32(active_rb_bitmap);
+}
+
+#define LDS_APP_BASE           0x1
+#define SCRATCH_APP_BASE       0x2
+
+static void gfx_v12_0_init_compute_vmid(struct amdgpu_device *adev)
+{
+	int i;
+	uint32_t sh_mem_bases;
+	uint32_t data;
+
+	/*
+	 * Configure apertures:
+	 * LDS:         0x60000000'00000000 - 0x60000001'00000000 (4GB)
+	 * Scratch:     0x60000001'00000000 - 0x60000002'00000000 (4GB)
+	 * GPUVM:       0x60010000'00000000 - 0x60020000'00000000 (1TB)
+	 */
+	sh_mem_bases = (LDS_APP_BASE << SH_MEM_BASES__SHARED_BASE__SHIFT) |
+			SCRATCH_APP_BASE;
+
+	mutex_lock(&adev->srbm_mutex);
+	for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) {
+		soc24_grbm_select(adev, 0, 0, 0, i);
+		/* CP and shaders */
+		WREG32_SOC15(GC, 0, regSH_MEM_CONFIG, DEFAULT_SH_MEM_CONFIG);
+		WREG32_SOC15(GC, 0, regSH_MEM_BASES, sh_mem_bases);
+
+		/* Enable trap for each kfd vmid. */
+		data = RREG32_SOC15(GC, 0, regSPI_GDBG_PER_VMID_CNTL);
+		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+		WREG32_SOC15(GC, 0, regSPI_GDBG_PER_VMID_CNTL, data);
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+}
+
+static void gfx_v12_0_tcp_harvest(struct amdgpu_device *adev)
+{
+	/* TODO: harvest feature to be added later. */
+}
+
+static void gfx_v12_0_get_tcc_info(struct amdgpu_device *adev)
+{
+}
+
+static void gfx_v12_0_constants_init(struct amdgpu_device *adev)
+{
+	u32 tmp;
+	int i;
+
+	if (!amdgpu_sriov_vf(adev))
+		WREG32_FIELD15_PREREG(GC, 0, GRBM_CNTL, READ_TIMEOUT, 0xff);
+
+	gfx_v12_0_setup_rb(adev);
+	gfx_v12_0_get_cu_info(adev, &adev->gfx.cu_info);
+	gfx_v12_0_get_tcc_info(adev);
+	adev->gfx.config.pa_sc_tile_steering_override = 0;
+
+	/* XXX SH_MEM regs */
+	/* where to put LDS, scratch, GPUVM in FSA64 space */
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
+		soc24_grbm_select(adev, 0, 0, 0, i);
+		/* CP and shaders */
+		WREG32_SOC15(GC, 0, regSH_MEM_CONFIG, DEFAULT_SH_MEM_CONFIG);
+		if (i != 0) {
+			tmp = REG_SET_FIELD(0, SH_MEM_BASES, PRIVATE_BASE,
+				(adev->gmc.private_aperture_start >> 48));
+			tmp = REG_SET_FIELD(tmp, SH_MEM_BASES, SHARED_BASE,
+				(adev->gmc.shared_aperture_start >> 48));
+			WREG32_SOC15(GC, 0, regSH_MEM_BASES, tmp);
+		}
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+
+	mutex_unlock(&adev->srbm_mutex);
+
+	gfx_v12_0_init_compute_vmid(adev);
+}
+
+static u32 gfx_v12_0_get_cpg_int_cntl(struct amdgpu_device *adev,
+				      int me, int pipe)
+{
+	if (me != 0)
+		return 0;
+
+	switch (pipe) {
+	case 0:
+		return SOC15_REG_OFFSET(GC, 0, regCP_INT_CNTL_RING0);
+	default:
+		return 0;
+	}
+}
+
+static u32 gfx_v12_0_get_cpc_int_cntl(struct amdgpu_device *adev,
+				      int me, int pipe)
+{
+	/*
+	 * amdgpu controls only the first MEC. That's why this function only
+	 * handles the setting of interrupts for this specific MEC. All other
+	 * pipes' interrupts are set by amdkfd.
+	 */
+	if (me != 1)
+		return 0;
+
+	switch (pipe) {
+	case 0:
+		return SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE0_INT_CNTL);
+	case 1:
+		return SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE1_INT_CNTL);
+	default:
+		return 0;
+	}
+}
+
+static void gfx_v12_0_enable_gui_idle_interrupt(struct amdgpu_device *adev,
+					       bool enable)
+{
+	u32 tmp, cp_int_cntl_reg;
+	int i, j;
+
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	for (i = 0; i < adev->gfx.me.num_me; i++) {
+		for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+			cp_int_cntl_reg = gfx_v12_0_get_cpg_int_cntl(adev, i, j);
+
+			if (cp_int_cntl_reg) {
+				tmp = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE,
+						    enable ? 1 : 0);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_EMPTY_INT_ENABLE,
+						    enable ? 1 : 0);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE,
+						    enable ? 1 : 0);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, GFX_IDLE_INT_ENABLE,
+						    enable ? 1 : 0);
+				WREG32_SOC15_IP(GC, cp_int_cntl_reg, tmp);
+			}
+		}
+	}
+}
+
+static int gfx_v12_0_init_csb(struct amdgpu_device *adev)
+{
+	adev->gfx.rlc.funcs->get_csb_buffer(adev, adev->gfx.rlc.cs_ptr);
+
+	WREG32_SOC15(GC, 0, regRLC_CSIB_ADDR_HI,
+			adev->gfx.rlc.clear_state_gpu_addr >> 32);
+	WREG32_SOC15(GC, 0, regRLC_CSIB_ADDR_LO,
+			adev->gfx.rlc.clear_state_gpu_addr & 0xfffffffc);
+	WREG32_SOC15(GC, 0, regRLC_CSIB_LENGTH, adev->gfx.rlc.clear_state_size);
+
+	return 0;
+}
+
+static void gfx_v12_0_rlc_stop(struct amdgpu_device *adev)
+{
+	u32 tmp = RREG32_SOC15(GC, 0, regRLC_CNTL);
+
+	tmp = REG_SET_FIELD(tmp, RLC_CNTL, RLC_ENABLE_F32, 0);
+	WREG32_SOC15(GC, 0, regRLC_CNTL, tmp);
+}
+
+static void gfx_v12_0_rlc_reset(struct amdgpu_device *adev)
+{
+	WREG32_FIELD15_PREREG(GC, 0, GRBM_SOFT_RESET, SOFT_RESET_RLC, 1);
+	udelay(50);
+	WREG32_FIELD15_PREREG(GC, 0, GRBM_SOFT_RESET, SOFT_RESET_RLC, 0);
+	udelay(50);
+}
+
+static void gfx_v12_0_rlc_smu_handshake_cntl(struct amdgpu_device *adev,
+					     bool enable)
+{
+	uint32_t rlc_pg_cntl;
+
+	rlc_pg_cntl = RREG32_SOC15(GC, 0, regRLC_PG_CNTL);
+
+	if (!enable) {
+		/* RLC_PG_CNTL[23] = 0 (default)
+		 * RLC will wait for handshake acks with SMU
+		 * GFXOFF will be enabled
+		 * RLC_PG_CNTL[23] = 1
+		 * RLC will not issue any message to SMU
+		 * hence no handshake between SMU & RLC
+		 * GFXOFF will be disabled
+		 */
+		rlc_pg_cntl |= RLC_PG_CNTL__SMU_HANDSHAKE_DISABLE_MASK;
+	} else
+		rlc_pg_cntl &= ~RLC_PG_CNTL__SMU_HANDSHAKE_DISABLE_MASK;
+	WREG32_SOC15(GC, 0, regRLC_PG_CNTL, rlc_pg_cntl);
+}
+
+static void gfx_v12_0_rlc_start(struct amdgpu_device *adev)
+{
+	/* TODO: enable rlc & smu handshake until smu
+	 * and gfxoff feature works as expected */
+	if (!(amdgpu_pp_feature_mask & PP_GFXOFF_MASK))
+		gfx_v12_0_rlc_smu_handshake_cntl(adev, false);
+
+	WREG32_FIELD15_PREREG(GC, 0, RLC_CNTL, RLC_ENABLE_F32, 1);
+	udelay(50);
+}
+
+static void gfx_v12_0_rlc_enable_srm(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	/* enable Save Restore Machine */
+	tmp = RREG32(SOC15_REG_OFFSET(GC, 0, regRLC_SRM_CNTL));
+	tmp |= RLC_SRM_CNTL__AUTO_INCR_ADDR_MASK;
+	tmp |= RLC_SRM_CNTL__SRM_ENABLE_MASK;
+	WREG32(SOC15_REG_OFFSET(GC, 0, regRLC_SRM_CNTL), tmp);
+}
+
+static void gfx_v12_0_load_rlcg_microcode(struct amdgpu_device *adev)
+{
+	const struct rlc_firmware_header_v2_0 *hdr;
+	const __le32 *fw_data;
+	unsigned i, fw_size;
+
+	hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
+	fw_data = (const __le32 *)(adev->gfx.rlc_fw->data +
+			   le32_to_cpu(hdr->header.ucode_array_offset_bytes));
+	fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4;
+
+	WREG32_SOC15(GC, 0, regRLC_GPM_UCODE_ADDR,
+		     RLCG_UCODE_LOADING_START_ADDRESS);
+
+	for (i = 0; i < fw_size; i++)
+		WREG32_SOC15(GC, 0, regRLC_GPM_UCODE_DATA,
+			     le32_to_cpup(fw_data++));
+
+	WREG32_SOC15(GC, 0, regRLC_GPM_UCODE_ADDR, adev->gfx.rlc_fw_version);
+}
+
+static void gfx_v12_0_load_rlc_iram_dram_microcode(struct amdgpu_device *adev)
+{
+	const struct rlc_firmware_header_v2_2 *hdr;
+	const __le32 *fw_data;
+	unsigned i, fw_size;
+	u32 tmp;
+
+	hdr = (const struct rlc_firmware_header_v2_2 *)adev->gfx.rlc_fw->data;
+
+	fw_data = (const __le32 *)(adev->gfx.rlc_fw->data +
+			le32_to_cpu(hdr->rlc_iram_ucode_offset_bytes));
+	fw_size = le32_to_cpu(hdr->rlc_iram_ucode_size_bytes) / 4;
+
+	WREG32_SOC15(GC, 0, regRLC_LX6_IRAM_ADDR, 0);
+
+	for (i = 0; i < fw_size; i++) {
+		if ((amdgpu_emu_mode == 1) && (i % 100 == 99))
+			msleep(1);
+		WREG32_SOC15(GC, 0, regRLC_LX6_IRAM_DATA,
+				le32_to_cpup(fw_data++));
+	}
+
+	WREG32_SOC15(GC, 0, regRLC_LX6_IRAM_ADDR, adev->gfx.rlc_fw_version);
+
+	fw_data = (const __le32 *)(adev->gfx.rlc_fw->data +
+			le32_to_cpu(hdr->rlc_dram_ucode_offset_bytes));
+	fw_size = le32_to_cpu(hdr->rlc_dram_ucode_size_bytes) / 4;
+
+	WREG32_SOC15(GC, 0, regRLC_LX6_DRAM_ADDR, 0);
+	for (i = 0; i < fw_size; i++) {
+		if ((amdgpu_emu_mode == 1) && (i % 100 == 99))
+			msleep(1);
+		WREG32_SOC15(GC, 0, regRLC_LX6_DRAM_DATA,
+				le32_to_cpup(fw_data++));
+	}
+
+	WREG32_SOC15(GC, 0, regRLC_LX6_IRAM_ADDR, adev->gfx.rlc_fw_version);
+
+	tmp = RREG32_SOC15(GC, 0, regRLC_LX6_CNTL);
+	tmp = REG_SET_FIELD(tmp, RLC_LX6_CNTL, PDEBUG_ENABLE, 1);
+	tmp = REG_SET_FIELD(tmp, RLC_LX6_CNTL, BRESET, 0);
+	WREG32_SOC15(GC, 0, regRLC_LX6_CNTL, tmp);
+}
+
+static int gfx_v12_0_rlc_load_microcode(struct amdgpu_device *adev)
+{
+	const struct rlc_firmware_header_v2_0 *hdr;
+	uint16_t version_major;
+	uint16_t version_minor;
+
+	if (!adev->gfx.rlc_fw)
+		return -EINVAL;
+
+	hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
+	amdgpu_ucode_print_rlc_hdr(&hdr->header);
+
+	version_major = le16_to_cpu(hdr->header.header_version_major);
+	version_minor = le16_to_cpu(hdr->header.header_version_minor);
+
+	if (version_major == 2) {
+		gfx_v12_0_load_rlcg_microcode(adev);
+		if (amdgpu_dpm == 1) {
+			if (version_minor >= 2)
+				gfx_v12_0_load_rlc_iram_dram_microcode(adev);
+		}
+
+		return 0;
+	}
+
+	return -EINVAL;
+}
+
+static int gfx_v12_0_rlc_resume(struct amdgpu_device *adev)
+{
+	int r;
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		gfx_v12_0_init_csb(adev);
+
+		if (!amdgpu_sriov_vf(adev)) /* enable RLC SRM */
+			gfx_v12_0_rlc_enable_srm(adev);
+	} else {
+		if (amdgpu_sriov_vf(adev)) {
+			gfx_v12_0_init_csb(adev);
+			return 0;
+		}
+
+		adev->gfx.rlc.funcs->stop(adev);
+
+		/* disable CG */
+		WREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL, 0);
+
+		/* disable PG */
+		WREG32_SOC15(GC, 0, regRLC_PG_CNTL, 0);
+
+		if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
+			/* legacy rlc firmware loading */
+			r = gfx_v12_0_rlc_load_microcode(adev);
+			if (r)
+				return r;
+		}
+
+		gfx_v12_0_init_csb(adev);
+
+		adev->gfx.rlc.funcs->start(adev);
+	}
+
+	return 0;
+}
+
+static void gfx_v12_0_config_gfx_rs64(struct amdgpu_device *adev)
+{
+	const struct gfx_firmware_header_v2_0 *pfp_hdr;
+	const struct gfx_firmware_header_v2_0 *me_hdr;
+	const struct gfx_firmware_header_v2_0 *mec_hdr;
+	uint32_t pipe_id, tmp;
+
+	mec_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.mec_fw->data;
+	me_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.me_fw->data;
+	pfp_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.pfp_fw->data;
+
+	/* config pfp program start addr */
+	for (pipe_id = 0; pipe_id < 2; pipe_id++) {
+		soc24_grbm_select(adev, 0, pipe_id, 0, 0);
+		WREG32_SOC15(GC, 0, regCP_PFP_PRGRM_CNTR_START,
+			(pfp_hdr->ucode_start_addr_hi << 30) |
+			(pfp_hdr->ucode_start_addr_lo >> 2));
+		WREG32_SOC15(GC, 0, regCP_PFP_PRGRM_CNTR_START_HI,
+			pfp_hdr->ucode_start_addr_hi >> 2);
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+
+	/* reset pfp pipe */
+	tmp = RREG32_SOC15(GC, 0, regCP_ME_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE0_RESET, 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE1_RESET, 1);
+	WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp);
+
+	/* clear pfp pipe reset */
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE0_RESET, 0);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE1_RESET, 0);
+	WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp);
+
+	/* config me program start addr */
+	for (pipe_id = 0; pipe_id < 2; pipe_id++) {
+		soc24_grbm_select(adev, 0, pipe_id, 0, 0);
+		WREG32_SOC15(GC, 0, regCP_ME_PRGRM_CNTR_START,
+			(me_hdr->ucode_start_addr_hi << 30) |
+			(me_hdr->ucode_start_addr_lo >> 2));
+		WREG32_SOC15(GC, 0, regCP_ME_PRGRM_CNTR_START_HI,
+			me_hdr->ucode_start_addr_hi>>2);
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+
+	/* reset me pipe */
+	tmp = RREG32_SOC15(GC, 0, regCP_ME_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE0_RESET, 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE1_RESET, 1);
+	WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp);
+
+	/* clear me pipe reset */
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE0_RESET, 0);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE1_RESET, 0);
+	WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp);
+
+	/* config mec program start addr */
+	for (pipe_id = 0; pipe_id < 4; pipe_id++) {
+		soc24_grbm_select(adev, 1, pipe_id, 0, 0);
+		WREG32_SOC15(GC, 0, regCP_MEC_RS64_PRGRM_CNTR_START,
+					mec_hdr->ucode_start_addr_lo >> 2 |
+					mec_hdr->ucode_start_addr_hi << 30);
+		WREG32_SOC15(GC, 0, regCP_MEC_RS64_PRGRM_CNTR_START_HI,
+					mec_hdr->ucode_start_addr_hi >> 2);
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+
+	/* reset mec pipe */
+	tmp = RREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE0_RESET, 1);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE1_RESET, 1);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE2_RESET, 1);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE3_RESET, 1);
+	WREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL, tmp);
+
+	/* clear mec pipe reset */
+	tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE0_RESET, 0);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE1_RESET, 0);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE2_RESET, 0);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE3_RESET, 0);
+	WREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL, tmp);
+}
+
+static void gfx_v12_0_set_pfp_ucode_start_addr(struct amdgpu_device *adev)
+{
+	const struct gfx_firmware_header_v2_0 *cp_hdr;
+	unsigned pipe_id, tmp;
+
+	cp_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.pfp_fw->data;
+	mutex_lock(&adev->srbm_mutex);
+	for (pipe_id = 0; pipe_id < adev->gfx.me.num_pipe_per_me; pipe_id++) {
+		soc24_grbm_select(adev, 0, pipe_id, 0, 0);
+		WREG32_SOC15(GC, 0, regCP_PFP_PRGRM_CNTR_START,
+			     (cp_hdr->ucode_start_addr_hi << 30) |
+			     (cp_hdr->ucode_start_addr_lo >> 2));
+		WREG32_SOC15(GC, 0, regCP_PFP_PRGRM_CNTR_START_HI,
+			     cp_hdr->ucode_start_addr_hi>>2);
+
+		/*
+		 * Program CP_ME_CNTL to reset given PIPE to take
+		 * effect of CP_PFP_PRGRM_CNTR_START.
+		 */
+		tmp = RREG32_SOC15(GC, 0, regCP_ME_CNTL);
+		if (pipe_id == 0)
+			tmp = REG_SET_FIELD(tmp, CP_ME_CNTL,
+					PFP_PIPE0_RESET, 1);
+		else
+			tmp = REG_SET_FIELD(tmp, CP_ME_CNTL,
+					PFP_PIPE1_RESET, 1);
+		WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp);
+
+		/* Clear pfp pipe0 reset bit. */
+		if (pipe_id == 0)
+			tmp = REG_SET_FIELD(tmp, CP_ME_CNTL,
+					PFP_PIPE0_RESET, 0);
+		else
+			tmp = REG_SET_FIELD(tmp, CP_ME_CNTL,
+					PFP_PIPE1_RESET, 0);
+		WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp);
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+}
+
+static void gfx_v12_0_set_me_ucode_start_addr(struct amdgpu_device *adev)
+{
+	const struct gfx_firmware_header_v2_0 *cp_hdr;
+	unsigned pipe_id, tmp;
+
+	cp_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.me_fw->data;
+	mutex_lock(&adev->srbm_mutex);
+	for (pipe_id = 0; pipe_id < adev->gfx.me.num_pipe_per_me; pipe_id++) {
+		soc24_grbm_select(adev, 0, pipe_id, 0, 0);
+		WREG32_SOC15(GC, 0, regCP_ME_PRGRM_CNTR_START,
+			     (cp_hdr->ucode_start_addr_hi << 30) |
+			     (cp_hdr->ucode_start_addr_lo >> 2) );
+		WREG32_SOC15(GC, 0, regCP_ME_PRGRM_CNTR_START_HI,
+			     cp_hdr->ucode_start_addr_hi>>2);
+
+		/*
+		 * Program CP_ME_CNTL to reset given PIPE to take
+		 * effect of CP_ME_PRGRM_CNTR_START.
+		 */
+		tmp = RREG32_SOC15(GC, 0, regCP_ME_CNTL);
+		if (pipe_id == 0)
+			tmp = REG_SET_FIELD(tmp, CP_ME_CNTL,
+					ME_PIPE0_RESET, 1);
+		else
+			tmp = REG_SET_FIELD(tmp, CP_ME_CNTL,
+					ME_PIPE1_RESET, 1);
+		WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp);
+
+		/* Clear pfp pipe0 reset bit. */
+		if (pipe_id == 0)
+			tmp = REG_SET_FIELD(tmp, CP_ME_CNTL,
+					ME_PIPE0_RESET, 0);
+		else
+			tmp = REG_SET_FIELD(tmp, CP_ME_CNTL,
+					ME_PIPE1_RESET, 0);
+		WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp);
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+}
+
+static void gfx_v12_0_set_mec_ucode_start_addr(struct amdgpu_device *adev)
+{
+	const struct gfx_firmware_header_v2_0 *cp_hdr;
+	unsigned pipe_id;
+
+	cp_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.mec_fw->data;
+	mutex_lock(&adev->srbm_mutex);
+	for (pipe_id = 0; pipe_id < adev->gfx.mec.num_pipe_per_mec; pipe_id++) {
+		soc24_grbm_select(adev, 1, pipe_id, 0, 0);
+		WREG32_SOC15(GC, 0, regCP_MEC_RS64_PRGRM_CNTR_START,
+			     cp_hdr->ucode_start_addr_lo >> 2 |
+			     cp_hdr->ucode_start_addr_hi << 30);
+		WREG32_SOC15(GC, 0, regCP_MEC_RS64_PRGRM_CNTR_START_HI,
+			     cp_hdr->ucode_start_addr_hi >> 2);
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+}
+
+static int gfx_v12_0_wait_for_rlc_autoload_complete(struct amdgpu_device *adev)
+{
+	uint32_t cp_status;
+	uint32_t bootload_status;
+	int i;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		cp_status = RREG32_SOC15(GC, 0, regCP_STAT);
+		bootload_status = RREG32_SOC15(GC, 0, regRLC_RLCS_BOOTLOAD_STATUS);
+
+		if ((cp_status == 0) &&
+		    (REG_GET_FIELD(bootload_status,
+			RLC_RLCS_BOOTLOAD_STATUS, BOOTLOAD_COMPLETE) == 1)) {
+			break;
+		}
+		udelay(1);
+		if (amdgpu_emu_mode)
+			msleep(10);
+	}
+
+	if (i >= adev->usec_timeout) {
+		dev_err(adev->dev, "rlc autoload: gc ucode autoload timeout\n");
+		return -ETIMEDOUT;
+	}
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) {
+		gfx_v12_0_set_pfp_ucode_start_addr(adev);
+		gfx_v12_0_set_me_ucode_start_addr(adev);
+		gfx_v12_0_set_mec_ucode_start_addr(adev);
+	}
+
+	return 0;
+}
+
+static int gfx_v12_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable)
+{
+	int i;
+	u32 tmp = RREG32_SOC15(GC, 0, regCP_ME_CNTL);
+
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_HALT, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_HALT, enable ? 0 : 1);
+	WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (RREG32_SOC15(GC, 0, regCP_STAT) == 0)
+			break;
+		udelay(1);
+	}
+
+	if (i >= adev->usec_timeout)
+		DRM_ERROR("failed to %s cp gfx\n", enable ? "unhalt" : "halt");
+
+	return 0;
+}
+
+static int gfx_v12_0_cp_gfx_load_pfp_microcode_rs64(struct amdgpu_device *adev)
+{
+	int r;
+	const struct gfx_firmware_header_v2_0 *pfp_hdr;
+	const __le32 *fw_ucode, *fw_data;
+	unsigned i, pipe_id, fw_ucode_size, fw_data_size;
+	uint32_t tmp;
+	uint32_t usec_timeout = 50000;  /* wait for 50ms */
+
+	pfp_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.pfp_fw->data;
+
+	amdgpu_ucode_print_gfx_hdr(&pfp_hdr->header);
+
+	/* instruction */
+	fw_ucode = (const __le32 *)(adev->gfx.pfp_fw->data +
+		le32_to_cpu(pfp_hdr->ucode_offset_bytes));
+	fw_ucode_size = le32_to_cpu(pfp_hdr->ucode_size_bytes);
+	/* data */
+	fw_data = (const __le32 *)(adev->gfx.pfp_fw->data +
+		le32_to_cpu(pfp_hdr->data_offset_bytes));
+	fw_data_size = le32_to_cpu(pfp_hdr->data_size_bytes);
+
+	/* 64kb align */
+	r = amdgpu_bo_create_reserved(adev, fw_ucode_size,
+				      64 * 1024, AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->gfx.pfp.pfp_fw_obj,
+				      &adev->gfx.pfp.pfp_fw_gpu_addr,
+				      (void **)&adev->gfx.pfp.pfp_fw_ptr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create pfp ucode fw bo\n", r);
+		gfx_v12_0_pfp_fini(adev);
+		return r;
+	}
+
+	r = amdgpu_bo_create_reserved(adev, fw_data_size,
+				      64 * 1024, AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->gfx.pfp.pfp_fw_data_obj,
+				      &adev->gfx.pfp.pfp_fw_data_gpu_addr,
+				      (void **)&adev->gfx.pfp.pfp_fw_data_ptr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create pfp data fw bo\n", r);
+		gfx_v12_0_pfp_fini(adev);
+		return r;
+	}
+
+	memcpy(adev->gfx.pfp.pfp_fw_ptr, fw_ucode, fw_ucode_size);
+	memcpy(adev->gfx.pfp.pfp_fw_data_ptr, fw_data, fw_data_size);
+
+	amdgpu_bo_kunmap(adev->gfx.pfp.pfp_fw_obj);
+	amdgpu_bo_kunmap(adev->gfx.pfp.pfp_fw_data_obj);
+	amdgpu_bo_unreserve(adev->gfx.pfp.pfp_fw_obj);
+	amdgpu_bo_unreserve(adev->gfx.pfp.pfp_fw_data_obj);
+
+	if (amdgpu_emu_mode == 1)
+		amdgpu_device_flush_hdp(adev, NULL);
+
+	WREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_LO,
+		lower_32_bits(adev->gfx.pfp.pfp_fw_gpu_addr));
+	WREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_HI,
+		upper_32_bits(adev->gfx.pfp.pfp_fw_gpu_addr));
+
+	tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, VMID, 0);
+	tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, CACHE_POLICY, 0);
+	tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, EXE_DISABLE, 0);
+	WREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_CNTL, tmp);
+
+	/*
+	 * Programming any of the CP_PFP_IC_BASE registers
+	 * forces invalidation of the ME L1 I$. Wait for the
+	 * invalidation complete
+	 */
+	for (i = 0; i < usec_timeout; i++) {
+		tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL);
+		if (1 == REG_GET_FIELD(tmp, CP_PFP_IC_OP_CNTL,
+			INVALIDATE_CACHE_COMPLETE))
+			break;
+		udelay(1);
+	}
+
+	if (i >= usec_timeout) {
+		dev_err(adev->dev, "failed to invalidate instruction cache\n");
+		return -EINVAL;
+	}
+
+	/* Prime the L1 instruction caches */
+	tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_PFP_IC_OP_CNTL, PRIME_ICACHE, 1);
+	WREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL, tmp);
+	/* Waiting for cache primed*/
+	for (i = 0; i < usec_timeout; i++) {
+		tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL);
+		if (1 == REG_GET_FIELD(tmp, CP_PFP_IC_OP_CNTL,
+			ICACHE_PRIMED))
+			break;
+		udelay(1);
+	}
+
+	if (i >= usec_timeout) {
+		dev_err(adev->dev, "failed to prime instruction cache\n");
+		return -EINVAL;
+	}
+
+	mutex_lock(&adev->srbm_mutex);
+	for (pipe_id = 0; pipe_id < adev->gfx.me.num_pipe_per_me; pipe_id++) {
+		soc24_grbm_select(adev, 0, pipe_id, 0, 0);
+
+		WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE0_LO,
+			lower_32_bits(adev->gfx.pfp.pfp_fw_data_gpu_addr));
+		WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE0_HI,
+			upper_32_bits(adev->gfx.pfp.pfp_fw_data_gpu_addr));
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+
+	tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_BASE_CNTL, VMID, 0);
+	tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_BASE_CNTL, CACHE_POLICY, 0);
+	WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE_CNTL, tmp);
+
+	/* Invalidate the data caches */
+	tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_OP_CNTL, INVALIDATE_DCACHE, 1);
+	WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL, tmp);
+
+	for (i = 0; i < usec_timeout; i++) {
+		tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL);
+		if (1 == REG_GET_FIELD(tmp, CP_GFX_RS64_DC_OP_CNTL,
+			INVALIDATE_DCACHE_COMPLETE))
+			break;
+		udelay(1);
+	}
+
+	if (i >= usec_timeout) {
+		dev_err(adev->dev, "failed to invalidate RS64 data cache\n");
+		return -EINVAL;
+	}
+
+	gfx_v12_0_set_pfp_ucode_start_addr(adev);
+
+	return 0;
+}
+
+static int gfx_v12_0_cp_gfx_load_me_microcode_rs64(struct amdgpu_device *adev)
+{
+	int r;
+	const struct gfx_firmware_header_v2_0 *me_hdr;
+	const __le32 *fw_ucode, *fw_data;
+	unsigned i, pipe_id, fw_ucode_size, fw_data_size;
+	uint32_t tmp;
+	uint32_t usec_timeout = 50000;  /* wait for 50ms */
+
+	me_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.me_fw->data;
+
+	amdgpu_ucode_print_gfx_hdr(&me_hdr->header);
+
+	/* instruction */
+	fw_ucode = (const __le32 *)(adev->gfx.me_fw->data +
+		le32_to_cpu(me_hdr->ucode_offset_bytes));
+	fw_ucode_size = le32_to_cpu(me_hdr->ucode_size_bytes);
+	/* data */
+	fw_data = (const __le32 *)(adev->gfx.me_fw->data +
+		le32_to_cpu(me_hdr->data_offset_bytes));
+	fw_data_size = le32_to_cpu(me_hdr->data_size_bytes);
+
+	/* 64kb align*/
+	r = amdgpu_bo_create_reserved(adev, fw_ucode_size,
+				      64 * 1024, AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->gfx.me.me_fw_obj,
+				      &adev->gfx.me.me_fw_gpu_addr,
+				      (void **)&adev->gfx.me.me_fw_ptr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create me ucode bo\n", r);
+		gfx_v12_0_me_fini(adev);
+		return r;
+	}
+
+	r = amdgpu_bo_create_reserved(adev, fw_data_size,
+				      64 * 1024, AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->gfx.me.me_fw_data_obj,
+				      &adev->gfx.me.me_fw_data_gpu_addr,
+				      (void **)&adev->gfx.me.me_fw_data_ptr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create me data bo\n", r);
+		gfx_v12_0_me_fini(adev);
+		return r;
+	}
+
+	memcpy(adev->gfx.me.me_fw_ptr, fw_ucode, fw_ucode_size);
+	memcpy(adev->gfx.me.me_fw_data_ptr, fw_data, fw_data_size);
+
+	amdgpu_bo_kunmap(adev->gfx.me.me_fw_obj);
+	amdgpu_bo_kunmap(adev->gfx.me.me_fw_data_obj);
+	amdgpu_bo_unreserve(adev->gfx.me.me_fw_obj);
+	amdgpu_bo_unreserve(adev->gfx.me.me_fw_data_obj);
+
+	if (amdgpu_emu_mode == 1)
+		amdgpu_device_flush_hdp(adev, NULL);
+
+	WREG32_SOC15(GC, 0, regCP_ME_IC_BASE_LO,
+		lower_32_bits(adev->gfx.me.me_fw_gpu_addr));
+	WREG32_SOC15(GC, 0, regCP_ME_IC_BASE_HI,
+		upper_32_bits(adev->gfx.me.me_fw_gpu_addr));
+
+	tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_BASE_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, VMID, 0);
+	tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, CACHE_POLICY, 0);
+	tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, EXE_DISABLE, 0);
+	WREG32_SOC15(GC, 0, regCP_ME_IC_BASE_CNTL, tmp);
+
+	/*
+	 * Programming any of the CP_ME_IC_BASE registers
+	 * forces invalidation of the ME L1 I$. Wait for the
+	 * invalidation complete
+	 */
+	for (i = 0; i < usec_timeout; i++) {
+		tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL);
+		if (1 == REG_GET_FIELD(tmp, CP_ME_IC_OP_CNTL,
+			INVALIDATE_CACHE_COMPLETE))
+			break;
+		udelay(1);
+	}
+
+	if (i >= usec_timeout) {
+		dev_err(adev->dev, "failed to invalidate instruction cache\n");
+		return -EINVAL;
+	}
+
+	/* Prime the instruction caches */
+	tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_ME_IC_OP_CNTL, PRIME_ICACHE, 1);
+	WREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL, tmp);
+
+	/* Waiting for instruction cache primed*/
+	for (i = 0; i < usec_timeout; i++) {
+		tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL);
+		if (1 == REG_GET_FIELD(tmp, CP_ME_IC_OP_CNTL,
+			ICACHE_PRIMED))
+			break;
+		udelay(1);
+	}
+
+	if (i >= usec_timeout) {
+		dev_err(adev->dev, "failed to prime instruction cache\n");
+		return -EINVAL;
+	}
+
+	mutex_lock(&adev->srbm_mutex);
+	for (pipe_id = 0; pipe_id < adev->gfx.me.num_pipe_per_me; pipe_id++) {
+		soc24_grbm_select(adev, 0, pipe_id, 0, 0);
+
+		WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE1_LO,
+			lower_32_bits(adev->gfx.me.me_fw_data_gpu_addr));
+		WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE1_HI,
+			upper_32_bits(adev->gfx.me.me_fw_data_gpu_addr));
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+
+	tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_BASE_CNTL, VMID, 0);
+	tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_BASE_CNTL, CACHE_POLICY, 0);
+	WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE_CNTL, tmp);
+
+	/* Invalidate the data caches */
+	tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_OP_CNTL, INVALIDATE_DCACHE, 1);
+	WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL, tmp);
+
+	for (i = 0; i < usec_timeout; i++) {
+		tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL);
+		if (1 == REG_GET_FIELD(tmp, CP_GFX_RS64_DC_OP_CNTL,
+			INVALIDATE_DCACHE_COMPLETE))
+			break;
+		udelay(1);
+	}
+
+	if (i >= usec_timeout) {
+		dev_err(adev->dev, "failed to invalidate RS64 data cache\n");
+		return -EINVAL;
+	}
+
+	gfx_v12_0_set_me_ucode_start_addr(adev);
+
+	return 0;
+}
+
+static int gfx_v12_0_cp_gfx_load_microcode(struct amdgpu_device *adev)
+{
+	int r;
+
+	if (!adev->gfx.me_fw || !adev->gfx.pfp_fw)
+		return -EINVAL;
+
+	gfx_v12_0_cp_gfx_enable(adev, false);
+
+	r = gfx_v12_0_cp_gfx_load_pfp_microcode_rs64(adev);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to load pfp fw\n", r);
+		return r;
+	}
+
+	r = gfx_v12_0_cp_gfx_load_me_microcode_rs64(adev);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to load me fw\n", r);
+		return r;
+	}
+
+	return 0;
+}
+
+static int gfx_v12_0_cp_gfx_start(struct amdgpu_device *adev)
+{
+	/* init the CP */
+	WREG32_SOC15(GC, 0, regCP_MAX_CONTEXT,
+		     adev->gfx.config.max_hw_contexts - 1);
+	WREG32_SOC15(GC, 0, regCP_DEVICE_ID, 1);
+
+	if (!amdgpu_async_gfx_ring)
+		gfx_v12_0_cp_gfx_enable(adev, true);
+
+	return 0;
+}
+
+static void gfx_v12_0_cp_gfx_switch_pipe(struct amdgpu_device *adev,
+					 CP_PIPE_ID pipe)
+{
+	u32 tmp;
+
+	tmp = RREG32_SOC15(GC, 0, regGRBM_GFX_CNTL);
+	tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, PIPEID, pipe);
+
+	WREG32_SOC15(GC, 0, regGRBM_GFX_CNTL, tmp);
+}
+
+static void gfx_v12_0_cp_gfx_set_doorbell(struct amdgpu_device *adev,
+					  struct amdgpu_ring *ring)
+{
+	u32 tmp;
+
+	tmp = RREG32_SOC15(GC, 0, regCP_RB_DOORBELL_CONTROL);
+	if (ring->use_doorbell) {
+		tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL,
+				    DOORBELL_OFFSET, ring->doorbell_index);
+		tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL,
+				    DOORBELL_EN, 1);
+	} else {
+		tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL,
+				    DOORBELL_EN, 0);
+	}
+	WREG32_SOC15(GC, 0, regCP_RB_DOORBELL_CONTROL, tmp);
+
+	tmp = REG_SET_FIELD(0, CP_RB_DOORBELL_RANGE_LOWER,
+			    DOORBELL_RANGE_LOWER, ring->doorbell_index);
+	WREG32_SOC15(GC, 0, regCP_RB_DOORBELL_RANGE_LOWER, tmp);
+
+	WREG32_SOC15(GC, 0, regCP_RB_DOORBELL_RANGE_UPPER,
+		     CP_RB_DOORBELL_RANGE_UPPER__DOORBELL_RANGE_UPPER_MASK);
+}
+
+static int gfx_v12_0_cp_gfx_resume(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	u32 tmp;
+	u32 rb_bufsz;
+	u64 rb_addr, rptr_addr, wptr_gpu_addr;
+
+	/* Set the write pointer delay */
+	WREG32_SOC15(GC, 0, regCP_RB_WPTR_DELAY, 0);
+
+	/* set the RB to use vmid 0 */
+	WREG32_SOC15(GC, 0, regCP_RB_VMID, 0);
+
+	/* Init gfx ring 0 for pipe 0 */
+	mutex_lock(&adev->srbm_mutex);
+	gfx_v12_0_cp_gfx_switch_pipe(adev, PIPE_ID0);
+
+	/* Set ring buffer size */
+	ring = &adev->gfx.gfx_ring[0];
+	rb_bufsz = order_base_2(ring->ring_size / 8);
+	tmp = REG_SET_FIELD(0, CP_RB0_CNTL, RB_BUFSZ, rb_bufsz);
+	tmp = REG_SET_FIELD(tmp, CP_RB0_CNTL, RB_BLKSZ, rb_bufsz - 2);
+	WREG32_SOC15(GC, 0, regCP_RB0_CNTL, tmp);
+
+	/* Initialize the ring buffer's write pointers */
+	ring->wptr = 0;
+	WREG32_SOC15(GC, 0, regCP_RB0_WPTR, lower_32_bits(ring->wptr));
+	WREG32_SOC15(GC, 0, regCP_RB0_WPTR_HI, upper_32_bits(ring->wptr));
+
+	/* set the wb address whether it's enabled or not */
+	rptr_addr = ring->rptr_gpu_addr;
+	WREG32_SOC15(GC, 0, regCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr));
+	WREG32_SOC15(GC, 0, regCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) &
+		     CP_RB_RPTR_ADDR_HI__RB_RPTR_ADDR_HI_MASK);
+
+	wptr_gpu_addr = ring->wptr_gpu_addr;
+	WREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_ADDR_LO,
+		     lower_32_bits(wptr_gpu_addr));
+	WREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_ADDR_HI,
+		     upper_32_bits(wptr_gpu_addr));
+
+	mdelay(1);
+	WREG32_SOC15(GC, 0, regCP_RB0_CNTL, tmp);
+
+	rb_addr = ring->gpu_addr >> 8;
+	WREG32_SOC15(GC, 0, regCP_RB0_BASE, rb_addr);
+	WREG32_SOC15(GC, 0, regCP_RB0_BASE_HI, upper_32_bits(rb_addr));
+
+	WREG32_SOC15(GC, 0, regCP_RB_ACTIVE, 1);
+
+	gfx_v12_0_cp_gfx_set_doorbell(adev, ring);
+	mutex_unlock(&adev->srbm_mutex);
+
+	/* Switch to pipe 0 */
+	mutex_lock(&adev->srbm_mutex);
+	gfx_v12_0_cp_gfx_switch_pipe(adev, PIPE_ID0);
+	mutex_unlock(&adev->srbm_mutex);
+
+	/* start the ring */
+	gfx_v12_0_cp_gfx_start(adev);
+	return 0;
+}
+
+static void gfx_v12_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
+{
+	u32 data;
+
+	data = RREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_INVALIDATE_ICACHE,
+						 enable ? 0 : 1);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE0_RESET,
+						 enable ? 0 : 1);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE1_RESET,
+						 enable ? 0 : 1);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE2_RESET,
+						 enable ? 0 : 1);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE3_RESET,
+						 enable ? 0 : 1);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE0_ACTIVE,
+						 enable ? 1 : 0);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE1_ACTIVE,
+			                         enable ? 1 : 0);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE2_ACTIVE,
+						 enable ? 1 : 0);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE3_ACTIVE,
+						 enable ? 1 : 0);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_HALT,
+						 enable ? 0 : 1);
+	WREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL, data);
+
+	adev->gfx.kiq[0].ring.sched.ready = enable;
+
+	udelay(50);
+}
+
+static int gfx_v12_0_cp_compute_load_microcode_rs64(struct amdgpu_device *adev)
+{
+	const struct gfx_firmware_header_v2_0 *mec_hdr;
+	const __le32 *fw_ucode, *fw_data;
+	u32 tmp, fw_ucode_size, fw_data_size;
+	u32 i, usec_timeout = 50000; /* Wait for 50 ms */
+	u32 *fw_ucode_ptr, *fw_data_ptr;
+	int r;
+
+	if (!adev->gfx.mec_fw)
+		return -EINVAL;
+
+	gfx_v12_0_cp_compute_enable(adev, false);
+
+	mec_hdr = (const struct gfx_firmware_header_v2_0 *)adev->gfx.mec_fw->data;
+	amdgpu_ucode_print_gfx_hdr(&mec_hdr->header);
+
+	fw_ucode = (const __le32 *) (adev->gfx.mec_fw->data +
+				le32_to_cpu(mec_hdr->ucode_offset_bytes));
+	fw_ucode_size = le32_to_cpu(mec_hdr->ucode_size_bytes);
+
+	fw_data = (const __le32 *) (adev->gfx.mec_fw->data +
+				le32_to_cpu(mec_hdr->data_offset_bytes));
+	fw_data_size = le32_to_cpu(mec_hdr->data_size_bytes);
+
+	r = amdgpu_bo_create_reserved(adev, fw_ucode_size,
+				      64 * 1024, AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->gfx.mec.mec_fw_obj,
+				      &adev->gfx.mec.mec_fw_gpu_addr,
+				      (void **)&fw_ucode_ptr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create mec fw ucode bo\n", r);
+		gfx_v12_0_mec_fini(adev);
+		return r;
+	}
+
+	r = amdgpu_bo_create_reserved(adev,
+				      ALIGN(fw_data_size, 64 * 1024) *
+				      adev->gfx.mec.num_pipe_per_mec,
+				      64 * 1024, AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->gfx.mec.mec_fw_data_obj,
+				      &adev->gfx.mec.mec_fw_data_gpu_addr,
+				      (void **)&fw_data_ptr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create mec fw ucode bo\n", r);
+		gfx_v12_0_mec_fini(adev);
+		return r;
+	}
+
+	memcpy(fw_ucode_ptr, fw_ucode, fw_ucode_size);
+	for (i = 0; i < adev->gfx.mec.num_pipe_per_mec; i++) {
+		memcpy(fw_data_ptr + i * ALIGN(fw_data_size, 64 * 1024) / 4, fw_data, fw_data_size);
+	}
+
+	amdgpu_bo_kunmap(adev->gfx.mec.mec_fw_obj);
+	amdgpu_bo_kunmap(adev->gfx.mec.mec_fw_data_obj);
+	amdgpu_bo_unreserve(adev->gfx.mec.mec_fw_obj);
+	amdgpu_bo_unreserve(adev->gfx.mec.mec_fw_data_obj);
+
+	tmp = RREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, VMID, 0);
+	tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, EXE_DISABLE, 0);
+	tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, CACHE_POLICY, 0);
+	WREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_CNTL, tmp);
+
+	tmp = RREG32_SOC15(GC, 0, regCP_MEC_DC_BASE_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_DC_BASE_CNTL, VMID, 0);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_DC_BASE_CNTL, CACHE_POLICY, 0);
+	WREG32_SOC15(GC, 0, regCP_MEC_DC_BASE_CNTL, tmp);
+
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->gfx.mec.num_pipe_per_mec; i++) {
+		soc24_grbm_select(adev, 1, i, 0, 0);
+
+		WREG32_SOC15(GC, 0, regCP_MEC_MDBASE_LO,
+			     lower_32_bits(adev->gfx.mec.mec_fw_data_gpu_addr +
+					   i * ALIGN(fw_data_size, 64 * 1024)));
+		WREG32_SOC15(GC, 0, regCP_MEC_MDBASE_HI,
+			     upper_32_bits(adev->gfx.mec.mec_fw_data_gpu_addr +
+					   i * ALIGN(fw_data_size, 64 * 1024)));
+
+		WREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_LO,
+			     lower_32_bits(adev->gfx.mec.mec_fw_gpu_addr));
+		WREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_HI,
+			     upper_32_bits(adev->gfx.mec.mec_fw_gpu_addr));
+	}
+	mutex_unlock(&adev->srbm_mutex);
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+
+	/* Trigger an invalidation of the L1 instruction caches */
+	tmp = RREG32_SOC15(GC, 0, regCP_MEC_DC_OP_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_DC_OP_CNTL, INVALIDATE_DCACHE, 1);
+	WREG32_SOC15(GC, 0, regCP_MEC_DC_OP_CNTL, tmp);
+
+	/* Wait for invalidation complete */
+	for (i = 0; i < usec_timeout; i++) {
+		tmp = RREG32_SOC15(GC, 0, regCP_MEC_DC_OP_CNTL);
+		if (1 == REG_GET_FIELD(tmp, CP_MEC_DC_OP_CNTL,
+				       INVALIDATE_DCACHE_COMPLETE))
+			break;
+		udelay(1);
+	}
+
+	if (i >= usec_timeout) {
+		dev_err(adev->dev, "failed to invalidate instruction cache\n");
+		return -EINVAL;
+	}
+
+	/* Trigger an invalidation of the L1 instruction caches */
+	tmp = RREG32_SOC15(GC, 0, regCP_CPC_IC_OP_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_CPC_IC_OP_CNTL, INVALIDATE_CACHE, 1);
+	WREG32_SOC15(GC, 0, regCP_CPC_IC_OP_CNTL, tmp);
+
+	/* Wait for invalidation complete */
+	for (i = 0; i < usec_timeout; i++) {
+		tmp = RREG32_SOC15(GC, 0, regCP_CPC_IC_OP_CNTL);
+		if (1 == REG_GET_FIELD(tmp, CP_CPC_IC_OP_CNTL,
+				       INVALIDATE_CACHE_COMPLETE))
+			break;
+		udelay(1);
+	}
+
+	if (i >= usec_timeout) {
+		dev_err(adev->dev, "failed to invalidate instruction cache\n");
+		return -EINVAL;
+	}
+
+	gfx_v12_0_set_mec_ucode_start_addr(adev);
+
+	return 0;
+}
+
+static void gfx_v12_0_kiq_setting(struct amdgpu_ring *ring)
+{
+	uint32_t tmp;
+	struct amdgpu_device *adev = ring->adev;
+
+	/* tell RLC which is KIQ queue */
+	tmp = RREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS);
+	tmp &= 0xffffff00;
+	tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
+	WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp | 0x80);
+}
+
+static void gfx_v12_0_cp_set_doorbell_range(struct amdgpu_device *adev)
+{
+	/* set graphics engine doorbell range */
+	WREG32_SOC15(GC, 0, regCP_RB_DOORBELL_RANGE_LOWER,
+		     (adev->doorbell_index.gfx_ring0 * 2) << 2);
+	WREG32_SOC15(GC, 0, regCP_RB_DOORBELL_RANGE_UPPER,
+		     (adev->doorbell_index.gfx_userqueue_end * 2) << 2);
+
+	/* set compute engine doorbell range */
+	WREG32_SOC15(GC, 0, regCP_MEC_DOORBELL_RANGE_LOWER,
+		     (adev->doorbell_index.kiq * 2) << 2);
+	WREG32_SOC15(GC, 0, regCP_MEC_DOORBELL_RANGE_UPPER,
+		     (adev->doorbell_index.userqueue_end * 2) << 2);
+}
+
+static int gfx_v12_0_gfx_mqd_init(struct amdgpu_device *adev, void *m,
+				  struct amdgpu_mqd_prop *prop)
+{
+	struct v12_gfx_mqd *mqd = m;
+	uint64_t hqd_gpu_addr, wb_gpu_addr;
+	uint32_t tmp;
+	uint32_t rb_bufsz;
+
+	/* set up gfx hqd wptr */
+	mqd->cp_gfx_hqd_wptr = 0;
+	mqd->cp_gfx_hqd_wptr_hi = 0;
+
+	/* set the pointer to the MQD */
+	mqd->cp_mqd_base_addr = prop->mqd_gpu_addr & 0xfffffffc;
+	mqd->cp_mqd_base_addr_hi = upper_32_bits(prop->mqd_gpu_addr);
+
+	/* set up mqd control */
+	tmp = regCP_GFX_MQD_CONTROL_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, VMID, 0);
+	tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, PRIV_STATE, 1);
+	tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, CACHE_POLICY, 0);
+	mqd->cp_gfx_mqd_control = tmp;
+
+	/* set up gfx_hqd_vimd with 0x0 to indicate the ring buffer's vmid */
+	tmp = regCP_GFX_HQD_VMID_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_VMID, VMID, 0);
+	mqd->cp_gfx_hqd_vmid = 0;
+
+	/* set up default queue priority level
+	 * 0x0 = low priority, 0x1 = high priority */
+	tmp = regCP_GFX_HQD_QUEUE_PRIORITY_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_QUEUE_PRIORITY, PRIORITY_LEVEL, 0);
+	mqd->cp_gfx_hqd_queue_priority = tmp;
+
+	/* set up time quantum */
+	tmp = regCP_GFX_HQD_QUANTUM_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_QUANTUM, QUANTUM_EN, 1);
+	mqd->cp_gfx_hqd_quantum = tmp;
+
+	/* set up gfx hqd base. this is similar as CP_RB_BASE */
+	hqd_gpu_addr = prop->hqd_base_gpu_addr >> 8;
+	mqd->cp_gfx_hqd_base = hqd_gpu_addr;
+	mqd->cp_gfx_hqd_base_hi = upper_32_bits(hqd_gpu_addr);
+
+	/* set up hqd_rptr_addr/_hi, similar as CP_RB_RPTR */
+	wb_gpu_addr = prop->rptr_gpu_addr;
+	mqd->cp_gfx_hqd_rptr_addr = wb_gpu_addr & 0xfffffffc;
+	mqd->cp_gfx_hqd_rptr_addr_hi =
+		upper_32_bits(wb_gpu_addr) & 0xffff;
+
+	/* set up rb_wptr_poll addr */
+	wb_gpu_addr = prop->wptr_gpu_addr;
+	mqd->cp_rb_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffffc;
+	mqd->cp_rb_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff;
+
+	/* set up the gfx_hqd_control, similar as CP_RB0_CNTL */
+	rb_bufsz = order_base_2(prop->queue_size / 4) - 1;
+	tmp = regCP_GFX_HQD_CNTL_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_BUFSZ, rb_bufsz);
+	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_BLKSZ, rb_bufsz - 2);
+#ifdef __BIG_ENDIAN
+	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, BUF_SWAP, 1);
+#endif
+	if (prop->tmz_queue)
+		tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, TMZ_MATCH, 1);
+	if (!prop->kernel_queue)
+		tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_NON_PRIV, 1);
+	mqd->cp_gfx_hqd_cntl = tmp;
+
+	/* set up cp_doorbell_control */
+	tmp = regCP_RB_DOORBELL_CONTROL_DEFAULT;
+	if (prop->use_doorbell) {
+		tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL,
+				    DOORBELL_OFFSET, prop->doorbell_index);
+		tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL,
+				    DOORBELL_EN, 1);
+	} else
+		tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL,
+				    DOORBELL_EN, 0);
+	mqd->cp_rb_doorbell_control = tmp;
+
+	/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
+	mqd->cp_gfx_hqd_rptr = regCP_GFX_HQD_RPTR_DEFAULT;
+
+	/* active the queue */
+	mqd->cp_gfx_hqd_active = 1;
+
+	/* set gfx UQ items */
+	mqd->shadow_base_lo = lower_32_bits(prop->shadow_addr);
+	mqd->shadow_base_hi = upper_32_bits(prop->shadow_addr);
+	mqd->fw_work_area_base_lo = lower_32_bits(prop->csa_addr);
+	mqd->fw_work_area_base_hi = upper_32_bits(prop->csa_addr);
+	mqd->fence_address_lo = lower_32_bits(prop->fence_address);
+	mqd->fence_address_hi = upper_32_bits(prop->fence_address);
+
+	return 0;
+}
+
+static int gfx_v12_0_kgq_init_queue(struct amdgpu_ring *ring, bool reset)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct v12_gfx_mqd *mqd = ring->mqd_ptr;
+	int mqd_idx = ring - &adev->gfx.gfx_ring[0];
+
+	if (!reset && !amdgpu_in_reset(adev) && !adev->in_suspend) {
+		memset((void *)mqd, 0, sizeof(*mqd));
+		mutex_lock(&adev->srbm_mutex);
+		soc24_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		amdgpu_ring_init_mqd(ring);
+		soc24_grbm_select(adev, 0, 0, 0, 0);
+		mutex_unlock(&adev->srbm_mutex);
+		if (adev->gfx.me.mqd_backup[mqd_idx])
+			memcpy_fromio(adev->gfx.me.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+	} else {
+		/* restore mqd with the backup copy */
+		if (adev->gfx.me.mqd_backup[mqd_idx])
+			memcpy_toio(mqd, adev->gfx.me.mqd_backup[mqd_idx], sizeof(*mqd));
+		/* reset the ring */
+		ring->wptr = 0;
+		*ring->wptr_cpu_addr = 0;
+		amdgpu_ring_clear_ring(ring);
+	}
+
+	return 0;
+}
+
+static int gfx_v12_0_cp_async_gfx_ring_resume(struct amdgpu_device *adev)
+{
+	int i, r;
+
+	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+		r = gfx_v12_0_kgq_init_queue(&adev->gfx.gfx_ring[i], false);
+		if (r)
+			return r;
+	}
+
+	r = amdgpu_gfx_enable_kgq(adev, 0);
+	if (r)
+		return r;
+
+	return gfx_v12_0_cp_gfx_start(adev);
+}
+
+static int gfx_v12_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
+				      struct amdgpu_mqd_prop *prop)
+{
+	struct v12_compute_mqd *mqd = m;
+	uint64_t hqd_gpu_addr, wb_gpu_addr, eop_base_addr;
+	uint32_t tmp;
+
+	mqd->header = 0xC0310800;
+	mqd->compute_pipelinestat_enable = 0x00000001;
+	mqd->compute_static_thread_mgmt_se0 = 0xffffffff;
+	mqd->compute_static_thread_mgmt_se1 = 0xffffffff;
+	mqd->compute_static_thread_mgmt_se2 = 0xffffffff;
+	mqd->compute_static_thread_mgmt_se3 = 0xffffffff;
+	mqd->compute_misc_reserved = 0x00000007;
+
+	eop_base_addr = prop->eop_gpu_addr >> 8;
+	mqd->cp_hqd_eop_base_addr_lo = eop_base_addr;
+	mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr);
+
+	/* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */
+	tmp = regCP_HQD_EOP_CONTROL_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_HQD_EOP_CONTROL, EOP_SIZE,
+			(order_base_2(GFX12_MEC_HPD_SIZE / 4) - 1));
+
+	mqd->cp_hqd_eop_control = tmp;
+
+	/* enable doorbell? */
+	tmp = regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT;
+
+	if (prop->use_doorbell) {
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_OFFSET, prop->doorbell_index);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_EN, 1);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_SOURCE, 0);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_HIT, 0);
+	} else {
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_EN, 0);
+	}
+
+	mqd->cp_hqd_pq_doorbell_control = tmp;
+
+	/* disable the queue if it's active */
+	mqd->cp_hqd_dequeue_request = 0;
+	mqd->cp_hqd_pq_rptr = 0;
+	mqd->cp_hqd_pq_wptr_lo = 0;
+	mqd->cp_hqd_pq_wptr_hi = 0;
+
+	/* set the pointer to the MQD */
+	mqd->cp_mqd_base_addr_lo = prop->mqd_gpu_addr & 0xfffffffc;
+	mqd->cp_mqd_base_addr_hi = upper_32_bits(prop->mqd_gpu_addr);
+
+	/* set MQD vmid to 0 */
+	tmp = regCP_MQD_CONTROL_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_MQD_CONTROL, VMID, 0);
+	mqd->cp_mqd_control = tmp;
+
+	/* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */
+	hqd_gpu_addr = prop->hqd_base_gpu_addr >> 8;
+	mqd->cp_hqd_pq_base_lo = hqd_gpu_addr;
+	mqd->cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr);
+
+	/* set up the HQD, this is similar to CP_RB0_CNTL */
+	tmp = regCP_HQD_PQ_CONTROL_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE,
+			    (order_base_2(prop->queue_size / 4) - 1));
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE,
+			    (order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1));
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 1);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TUNNEL_DISPATCH, 0);
+	if (prop->kernel_queue) {
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, PRIV_STATE, 1);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, KMD_QUEUE, 1);
+	}
+	if (prop->tmz_queue)
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TMZ, 1);
+	mqd->cp_hqd_pq_control = tmp;
+
+	/* set the wb address whether it's enabled or not */
+	wb_gpu_addr = prop->rptr_gpu_addr;
+	mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc;
+	mqd->cp_hqd_pq_rptr_report_addr_hi =
+		upper_32_bits(wb_gpu_addr) & 0xffff;
+
+	/* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */
+	wb_gpu_addr = prop->wptr_gpu_addr;
+	mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffffc;
+	mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff;
+
+	tmp = 0;
+	/* enable the doorbell if requested */
+	if (prop->use_doorbell) {
+		tmp = regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT;
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				DOORBELL_OFFSET, prop->doorbell_index);
+
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_EN, 1);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_SOURCE, 0);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_HIT, 0);
+	}
+
+	mqd->cp_hqd_pq_doorbell_control = tmp;
+
+	/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
+	mqd->cp_hqd_pq_rptr = regCP_HQD_PQ_RPTR_DEFAULT;
+
+	/* set the vmid for the queue */
+	mqd->cp_hqd_vmid = 0;
+
+	tmp = regCP_HQD_PERSISTENT_STATE_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PERSISTENT_STATE, PRELOAD_SIZE, 0x55);
+	mqd->cp_hqd_persistent_state = tmp;
+
+	/* set MIN_IB_AVAIL_SIZE */
+	tmp = regCP_HQD_IB_CONTROL_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_HQD_IB_CONTROL, MIN_IB_AVAIL_SIZE, 3);
+	mqd->cp_hqd_ib_control = tmp;
+
+	/* set static priority for a compute queue/ring */
+	mqd->cp_hqd_pipe_priority = prop->hqd_pipe_priority;
+	mqd->cp_hqd_queue_priority = prop->hqd_queue_priority;
+
+	mqd->cp_hqd_active = prop->hqd_active;
+
+	/* set UQ fenceaddress */
+	mqd->fence_address_lo = lower_32_bits(prop->fence_address);
+	mqd->fence_address_hi = upper_32_bits(prop->fence_address);
+
+	return 0;
+}
+
+static int gfx_v12_0_kiq_init_register(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct v12_compute_mqd *mqd = ring->mqd_ptr;
+	int j;
+
+	/* inactivate the queue */
+	if (amdgpu_sriov_vf(adev))
+		WREG32_SOC15(GC, 0, regCP_HQD_ACTIVE, 0);
+
+	/* disable wptr polling */
+	WREG32_FIELD15_PREREG(GC, 0, CP_PQ_WPTR_POLL_CNTL, EN, 0);
+
+	/* write the EOP addr */
+	WREG32_SOC15(GC, 0, regCP_HQD_EOP_BASE_ADDR,
+	       mqd->cp_hqd_eop_base_addr_lo);
+	WREG32_SOC15(GC, 0, regCP_HQD_EOP_BASE_ADDR_HI,
+	       mqd->cp_hqd_eop_base_addr_hi);
+
+	/* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */
+	WREG32_SOC15(GC, 0, regCP_HQD_EOP_CONTROL,
+	       mqd->cp_hqd_eop_control);
+
+	/* enable doorbell? */
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL,
+	       mqd->cp_hqd_pq_doorbell_control);
+
+	/* disable the queue if it's active */
+	if (RREG32_SOC15(GC, 0, regCP_HQD_ACTIVE) & 1) {
+		WREG32_SOC15(GC, 0, regCP_HQD_DEQUEUE_REQUEST, 1);
+		for (j = 0; j < adev->usec_timeout; j++) {
+			if (!(RREG32_SOC15(GC, 0, regCP_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+		WREG32_SOC15(GC, 0, regCP_HQD_DEQUEUE_REQUEST,
+		       mqd->cp_hqd_dequeue_request);
+		WREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR,
+		       mqd->cp_hqd_pq_rptr);
+		WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_LO,
+		       mqd->cp_hqd_pq_wptr_lo);
+		WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_HI,
+		       mqd->cp_hqd_pq_wptr_hi);
+	}
+
+	/* set the pointer to the MQD */
+	WREG32_SOC15(GC, 0, regCP_MQD_BASE_ADDR,
+	       mqd->cp_mqd_base_addr_lo);
+	WREG32_SOC15(GC, 0, regCP_MQD_BASE_ADDR_HI,
+	       mqd->cp_mqd_base_addr_hi);
+
+	/* set MQD vmid to 0 */
+	WREG32_SOC15(GC, 0, regCP_MQD_CONTROL,
+	       mqd->cp_mqd_control);
+
+	/* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_BASE,
+	       mqd->cp_hqd_pq_base_lo);
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_BASE_HI,
+	       mqd->cp_hqd_pq_base_hi);
+
+	/* set up the HQD, this is similar to CP_RB0_CNTL */
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_CONTROL,
+	       mqd->cp_hqd_pq_control);
+
+	/* set the wb address whether it's enabled or not */
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR_REPORT_ADDR,
+		mqd->cp_hqd_pq_rptr_report_addr_lo);
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR_REPORT_ADDR_HI,
+		mqd->cp_hqd_pq_rptr_report_addr_hi);
+
+	/* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR,
+	       mqd->cp_hqd_pq_wptr_poll_addr_lo);
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR_HI,
+	       mqd->cp_hqd_pq_wptr_poll_addr_hi);
+
+	/* enable the doorbell if requested */
+	if (ring->use_doorbell) {
+		WREG32_SOC15(GC, 0, regCP_MEC_DOORBELL_RANGE_LOWER,
+			(adev->doorbell_index.kiq * 2) << 2);
+		WREG32_SOC15(GC, 0, regCP_MEC_DOORBELL_RANGE_UPPER,
+			(adev->doorbell_index.userqueue_end * 2) << 2);
+	}
+
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL,
+	       mqd->cp_hqd_pq_doorbell_control);
+
+	/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_LO,
+	       mqd->cp_hqd_pq_wptr_lo);
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_HI,
+	       mqd->cp_hqd_pq_wptr_hi);
+
+	/* set the vmid for the queue */
+	WREG32_SOC15(GC, 0, regCP_HQD_VMID, mqd->cp_hqd_vmid);
+
+	WREG32_SOC15(GC, 0, regCP_HQD_PERSISTENT_STATE,
+	       mqd->cp_hqd_persistent_state);
+
+	/* activate the queue */
+	WREG32_SOC15(GC, 0, regCP_HQD_ACTIVE,
+	       mqd->cp_hqd_active);
+
+	if (ring->use_doorbell)
+		WREG32_FIELD15_PREREG(GC, 0, CP_PQ_STATUS, DOORBELL_ENABLE, 1);
+
+	return 0;
+}
+
+static int gfx_v12_0_kiq_init_queue(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct v12_compute_mqd *mqd = ring->mqd_ptr;
+	int mqd_idx = AMDGPU_MAX_COMPUTE_RINGS;
+
+	gfx_v12_0_kiq_setting(ring);
+
+	if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */
+		/* reset MQD to a clean status */
+		if (adev->gfx.mec.mqd_backup[mqd_idx])
+			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
+
+		/* reset ring buffer */
+		ring->wptr = 0;
+		amdgpu_ring_clear_ring(ring);
+
+		mutex_lock(&adev->srbm_mutex);
+		soc24_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		gfx_v12_0_kiq_init_register(ring);
+		soc24_grbm_select(adev, 0, 0, 0, 0);
+		mutex_unlock(&adev->srbm_mutex);
+	} else {
+		memset((void *)mqd, 0, sizeof(*mqd));
+		if (amdgpu_sriov_vf(adev) && adev->in_suspend)
+			amdgpu_ring_clear_ring(ring);
+		mutex_lock(&adev->srbm_mutex);
+		soc24_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		amdgpu_ring_init_mqd(ring);
+		gfx_v12_0_kiq_init_register(ring);
+		soc24_grbm_select(adev, 0, 0, 0, 0);
+		mutex_unlock(&adev->srbm_mutex);
+
+		if (adev->gfx.mec.mqd_backup[mqd_idx])
+			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+	}
+
+	return 0;
+}
+
+static int gfx_v12_0_kcq_init_queue(struct amdgpu_ring *ring, bool reset)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct v12_compute_mqd *mqd = ring->mqd_ptr;
+	int mqd_idx = ring - &adev->gfx.compute_ring[0];
+
+	if (!reset && !amdgpu_in_reset(adev) && !adev->in_suspend) {
+		memset((void *)mqd, 0, sizeof(*mqd));
+		mutex_lock(&adev->srbm_mutex);
+		soc24_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		amdgpu_ring_init_mqd(ring);
+		soc24_grbm_select(adev, 0, 0, 0, 0);
+		mutex_unlock(&adev->srbm_mutex);
+
+		if (adev->gfx.mec.mqd_backup[mqd_idx])
+			memcpy_fromio(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+	} else {
+		/* restore MQD to a clean status */
+		if (adev->gfx.mec.mqd_backup[mqd_idx])
+			memcpy_toio(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
+		/* reset ring buffer */
+		ring->wptr = 0;
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, 0);
+		amdgpu_ring_clear_ring(ring);
+	}
+
+	return 0;
+}
+
+static int gfx_v12_0_kiq_resume(struct amdgpu_device *adev)
+{
+	gfx_v12_0_kiq_init_queue(&adev->gfx.kiq[0].ring);
+	adev->gfx.kiq[0].ring.sched.ready = true;
+	return 0;
+}
+
+static int gfx_v12_0_kcq_resume(struct amdgpu_device *adev)
+{
+	int i, r;
+
+	if (!amdgpu_async_gfx_ring)
+		gfx_v12_0_cp_compute_enable(adev, true);
+
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		r = gfx_v12_0_kcq_init_queue(&adev->gfx.compute_ring[i], false);
+		if (r)
+			return r;
+	}
+
+	return amdgpu_gfx_enable_kcq(adev, 0);
+}
+
+static int gfx_v12_0_cp_resume(struct amdgpu_device *adev)
+{
+	int r, i;
+	struct amdgpu_ring *ring;
+
+	if (!(adev->flags & AMD_IS_APU))
+		gfx_v12_0_enable_gui_idle_interrupt(adev, false);
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
+		/* legacy firmware loading */
+		r = gfx_v12_0_cp_gfx_load_microcode(adev);
+		if (r)
+			return r;
+
+		r = gfx_v12_0_cp_compute_load_microcode_rs64(adev);
+		if (r)
+			return r;
+	}
+
+	gfx_v12_0_cp_set_doorbell_range(adev);
+
+	if (amdgpu_async_gfx_ring) {
+		gfx_v12_0_cp_compute_enable(adev, true);
+		gfx_v12_0_cp_gfx_enable(adev, true);
+	}
+
+	if (adev->enable_mes_kiq && adev->mes.kiq_hw_init)
+		r = amdgpu_mes_kiq_hw_init(adev);
+	else
+		r = gfx_v12_0_kiq_resume(adev);
+	if (r)
+		return r;
+
+	r = gfx_v12_0_kcq_resume(adev);
+	if (r)
+		return r;
+
+	if (!amdgpu_async_gfx_ring) {
+		r = gfx_v12_0_cp_gfx_resume(adev);
+		if (r)
+			return r;
+	} else {
+		r = gfx_v12_0_cp_async_gfx_ring_resume(adev);
+		if (r)
+			return r;
+	}
+
+	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+		ring = &adev->gfx.gfx_ring[i];
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
+	}
+
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		ring = &adev->gfx.compute_ring[i];
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+static void gfx_v12_0_cp_enable(struct amdgpu_device *adev, bool enable)
+{
+	gfx_v12_0_cp_gfx_enable(adev, enable);
+	gfx_v12_0_cp_compute_enable(adev, enable);
+}
+
+static int gfx_v12_0_gfxhub_enable(struct amdgpu_device *adev)
+{
+	int r;
+	bool value;
+
+	r = adev->gfxhub.funcs->gart_enable(adev);
+	if (r)
+		return r;
+
+	amdgpu_device_flush_hdp(adev, NULL);
+
+	value = amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS;
+
+	adev->gfxhub.funcs->set_fault_enable_default(adev, value);
+	/* TODO investigate why this and the hdp flush above is needed,
+	 * are we missing a flush somewhere else? */
+	adev->gmc.gmc_funcs->flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB(0), 0);
+
+	return 0;
+}
+
+static int get_gb_addr_config(struct amdgpu_device *adev)
+{
+	u32 gb_addr_config;
+
+	gb_addr_config = RREG32_SOC15(GC, 0, regGB_ADDR_CONFIG);
+	if (gb_addr_config == 0)
+		return -EINVAL;
+
+	adev->gfx.config.gb_addr_config_fields.num_pkrs =
+		1 << REG_GET_FIELD(gb_addr_config, GB_ADDR_CONFIG, NUM_PKRS);
+
+	adev->gfx.config.gb_addr_config = gb_addr_config;
+
+	adev->gfx.config.gb_addr_config_fields.num_pipes = 1 <<
+			REG_GET_FIELD(adev->gfx.config.gb_addr_config,
+				      GB_ADDR_CONFIG, NUM_PIPES);
+
+	adev->gfx.config.max_tile_pipes =
+		adev->gfx.config.gb_addr_config_fields.num_pipes;
+
+	adev->gfx.config.gb_addr_config_fields.max_compress_frags = 1 <<
+			REG_GET_FIELD(adev->gfx.config.gb_addr_config,
+				      GB_ADDR_CONFIG, MAX_COMPRESSED_FRAGS);
+	adev->gfx.config.gb_addr_config_fields.num_rb_per_se = 1 <<
+			REG_GET_FIELD(adev->gfx.config.gb_addr_config,
+				      GB_ADDR_CONFIG, NUM_RB_PER_SE);
+	adev->gfx.config.gb_addr_config_fields.num_se = 1 <<
+			REG_GET_FIELD(adev->gfx.config.gb_addr_config,
+				      GB_ADDR_CONFIG, NUM_SHADER_ENGINES);
+	adev->gfx.config.gb_addr_config_fields.pipe_interleave_size = 1 << (8 +
+			REG_GET_FIELD(adev->gfx.config.gb_addr_config,
+				      GB_ADDR_CONFIG, PIPE_INTERLEAVE_SIZE));
+
+	return 0;
+}
+
+static void gfx_v12_0_disable_gpa_mode(struct amdgpu_device *adev)
+{
+	uint32_t data;
+
+	data = RREG32_SOC15(GC, 0, regCPC_PSP_DEBUG);
+	data |= CPC_PSP_DEBUG__GPA_OVERRIDE_MASK;
+	WREG32_SOC15(GC, 0, regCPC_PSP_DEBUG, data);
+
+	data = RREG32_SOC15(GC, 0, regCPG_PSP_DEBUG);
+	data |= CPG_PSP_DEBUG__GPA_OVERRIDE_MASK;
+	WREG32_SOC15(GC, 0, regCPG_PSP_DEBUG, data);
+}
+
+static void gfx_v12_0_init_golden_registers(struct amdgpu_device *adev)
+{
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		soc15_program_register_sequence(adev,
+						golden_settings_gc_12_0,
+						(const u32)ARRAY_SIZE(golden_settings_gc_12_0));
+
+		if (adev->rev_id == 0)
+			soc15_program_register_sequence(adev,
+					golden_settings_gc_12_0_rev0,
+					(const u32)ARRAY_SIZE(golden_settings_gc_12_0_rev0));
+		break;
+	default:
+		break;
+	}
+}
+
+static int gfx_v12_0_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) {
+		if (adev->gfx.imu.funcs && (amdgpu_dpm > 0)) {
+			/* RLC autoload sequence 1: Program rlc ram */
+			if (adev->gfx.imu.funcs->program_rlc_ram)
+				adev->gfx.imu.funcs->program_rlc_ram(adev);
+		}
+		/* rlc autoload firmware */
+		r = gfx_v12_0_rlc_backdoor_autoload_enable(adev);
+		if (r)
+			return r;
+	} else {
+		if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
+			if (adev->gfx.imu.funcs && (amdgpu_dpm > 0)) {
+				if (adev->gfx.imu.funcs->load_microcode)
+					adev->gfx.imu.funcs->load_microcode(adev);
+				if (adev->gfx.imu.funcs->setup_imu)
+					adev->gfx.imu.funcs->setup_imu(adev);
+				if (adev->gfx.imu.funcs->start_imu)
+					adev->gfx.imu.funcs->start_imu(adev);
+			}
+
+			/* disable gpa mode in backdoor loading */
+			gfx_v12_0_disable_gpa_mode(adev);
+		}
+	}
+
+	if ((adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) ||
+	    (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP)) {
+		r = gfx_v12_0_wait_for_rlc_autoload_complete(adev);
+		if (r) {
+			dev_err(adev->dev, "(%d) failed to wait rlc autoload complete\n", r);
+			return r;
+		}
+	}
+
+	if (!amdgpu_emu_mode)
+		gfx_v12_0_init_golden_registers(adev);
+
+	adev->gfx.is_poweron = true;
+
+	if (get_gb_addr_config(adev))
+		DRM_WARN("Invalid gb_addr_config !\n");
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP)
+		gfx_v12_0_config_gfx_rs64(adev);
+
+	r = gfx_v12_0_gfxhub_enable(adev);
+	if (r)
+		return r;
+
+	if ((adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT ||
+	     adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) &&
+	     (amdgpu_dpm == 1)) {
+		/**
+		 * For gfx 12, rlc firmware loading relies on smu firmware is
+		 * loaded firstly, so in direct type, it has to load smc ucode
+		 * here before rlc.
+		 */
+		r = amdgpu_pm_load_smu_firmware(adev, NULL);
+		if (r)
+			return r;
+	}
+
+	gfx_v12_0_constants_init(adev);
+
+	if (adev->nbio.funcs->gc_doorbell_init)
+		adev->nbio.funcs->gc_doorbell_init(adev);
+
+	r = gfx_v12_0_rlc_resume(adev);
+	if (r)
+		return r;
+
+	/*
+	 * init golden registers and rlc resume may override some registers,
+	 * reconfig them here
+	 */
+	gfx_v12_0_tcp_harvest(adev);
+
+	r = gfx_v12_0_cp_resume(adev);
+	if (r)
+		return r;
+
+	return r;
+}
+
+static int gfx_v12_0_set_userq_eop_interrupts(struct amdgpu_device *adev,
+					      bool enable)
+{
+	unsigned int irq_type;
+	int m, p, r;
+
+	if (adev->userq_funcs[AMDGPU_HW_IP_GFX]) {
+		for (m = 0; m < adev->gfx.me.num_me; m++) {
+			for (p = 0; p < adev->gfx.me.num_pipe_per_me; p++) {
+				irq_type = AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP + p;
+				if (enable)
+					r = amdgpu_irq_get(adev, &adev->gfx.eop_irq,
+							   irq_type);
+				else
+					r = amdgpu_irq_put(adev, &adev->gfx.eop_irq,
+							   irq_type);
+				if (r)
+					return r;
+			}
+		}
+	}
+
+	if (adev->userq_funcs[AMDGPU_HW_IP_COMPUTE]) {
+		for (m = 0; m < adev->gfx.mec.num_mec; ++m) {
+			for (p = 0; p < adev->gfx.mec.num_pipe_per_mec; p++) {
+				irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
+					+ (m * adev->gfx.mec.num_pipe_per_mec)
+					+ p;
+				if (enable)
+					r = amdgpu_irq_get(adev, &adev->gfx.eop_irq,
+							   irq_type);
+				else
+					r = amdgpu_irq_put(adev, &adev->gfx.eop_irq,
+							   irq_type);
+				if (r)
+					return r;
+			}
+		}
+	}
+
+	return 0;
+}
+
+static int gfx_v12_0_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t tmp;
+
+	cancel_delayed_work_sync(&adev->gfx.idle_work);
+
+	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
+	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
+	amdgpu_irq_put(adev, &adev->gfx.bad_op_irq, 0);
+	gfx_v12_0_set_userq_eop_interrupts(adev, false);
+
+	if (!adev->no_hw_access) {
+		if (amdgpu_async_gfx_ring) {
+			if (amdgpu_gfx_disable_kgq(adev, 0))
+				DRM_ERROR("KGQ disable failed\n");
+		}
+
+		if (amdgpu_gfx_disable_kcq(adev, 0))
+			DRM_ERROR("KCQ disable failed\n");
+
+		amdgpu_mes_kiq_hw_fini(adev);
+	}
+
+	if (amdgpu_sriov_vf(adev)) {
+		gfx_v12_0_cp_gfx_enable(adev, false);
+		/* Program KIQ position of RLC_CP_SCHEDULERS during destroy */
+		tmp = RREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS);
+		tmp &= 0xffffff00;
+		WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp);
+
+		return 0;
+	}
+	gfx_v12_0_cp_enable(adev, false);
+	gfx_v12_0_enable_gui_idle_interrupt(adev, false);
+
+	adev->gfxhub.funcs->gart_disable(adev);
+
+	adev->gfx.is_poweron = false;
+
+	return 0;
+}
+
+static int gfx_v12_0_suspend(struct amdgpu_ip_block *ip_block)
+{
+	return gfx_v12_0_hw_fini(ip_block);
+}
+
+static int gfx_v12_0_resume(struct amdgpu_ip_block *ip_block)
+{
+	return gfx_v12_0_hw_init(ip_block);
+}
+
+static bool gfx_v12_0_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (REG_GET_FIELD(RREG32_SOC15(GC, 0, regGRBM_STATUS),
+				GRBM_STATUS, GUI_ACTIVE))
+		return false;
+	else
+		return true;
+}
+
+static int gfx_v12_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	unsigned i;
+	u32 tmp;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		/* read MC_STATUS */
+		tmp = RREG32_SOC15(GC, 0, regGRBM_STATUS) &
+			GRBM_STATUS__GUI_ACTIVE_MASK;
+
+		if (!REG_GET_FIELD(tmp, GRBM_STATUS, GUI_ACTIVE))
+			return 0;
+		udelay(1);
+	}
+	return -ETIMEDOUT;
+}
+
+static uint64_t gfx_v12_0_get_gpu_clock_counter(struct amdgpu_device *adev)
+{
+	uint64_t clock = 0;
+
+	if (adev->smuio.funcs &&
+	    adev->smuio.funcs->get_gpu_clock_counter)
+		clock = adev->smuio.funcs->get_gpu_clock_counter(adev);
+	else
+		dev_warn(adev->dev, "query gpu clock counter is not supported\n");
+
+	return clock;
+}
+
+static int gfx_v12_0_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	switch (amdgpu_user_queue) {
+	case -1:
+	case 0:
+	default:
+		adev->gfx.disable_kq = false;
+		adev->gfx.disable_uq = true;
+		break;
+	case 1:
+		adev->gfx.disable_kq = false;
+		adev->gfx.disable_uq = false;
+		break;
+	case 2:
+		adev->gfx.disable_kq = true;
+		adev->gfx.disable_uq = false;
+		break;
+	}
+
+	adev->gfx.funcs = &gfx_v12_0_gfx_funcs;
+
+	if (adev->gfx.disable_kq) {
+		adev->gfx.num_gfx_rings = 0;
+		adev->gfx.num_compute_rings = 0;
+	} else {
+		adev->gfx.num_gfx_rings = GFX12_NUM_GFX_RINGS;
+		adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
+						  AMDGPU_MAX_COMPUTE_RINGS);
+	}
+
+	gfx_v12_0_set_kiq_pm4_funcs(adev);
+	gfx_v12_0_set_ring_funcs(adev);
+	gfx_v12_0_set_irq_funcs(adev);
+	gfx_v12_0_set_rlc_funcs(adev);
+	gfx_v12_0_set_mqd_funcs(adev);
+	gfx_v12_0_set_imu_funcs(adev);
+
+	gfx_v12_0_init_rlcg_reg_access_ctrl(adev);
+
+	return gfx_v12_0_init_microcode(adev);
+}
+
+static int gfx_v12_0_late_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = amdgpu_irq_get(adev, &adev->gfx.priv_reg_irq, 0);
+	if (r)
+		return r;
+
+	r = amdgpu_irq_get(adev, &adev->gfx.priv_inst_irq, 0);
+	if (r)
+		return r;
+
+	r = amdgpu_irq_get(adev, &adev->gfx.bad_op_irq, 0);
+	if (r)
+		return r;
+
+	r = gfx_v12_0_set_userq_eop_interrupts(adev, true);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static bool gfx_v12_0_is_rlc_enabled(struct amdgpu_device *adev)
+{
+	uint32_t rlc_cntl;
+
+	/* if RLC is not enabled, do nothing */
+	rlc_cntl = RREG32_SOC15(GC, 0, regRLC_CNTL);
+	return (REG_GET_FIELD(rlc_cntl, RLC_CNTL, RLC_ENABLE_F32)) ? true : false;
+}
+
+static void gfx_v12_0_set_safe_mode(struct amdgpu_device *adev,
+				    int xcc_id)
+{
+	uint32_t data;
+	unsigned i;
+
+	data = RLC_SAFE_MODE__CMD_MASK;
+	data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT);
+
+	WREG32_SOC15(GC, 0, regRLC_SAFE_MODE, data);
+
+	/* wait for RLC_SAFE_MODE */
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (!REG_GET_FIELD(RREG32_SOC15(GC, 0, regRLC_SAFE_MODE),
+				   RLC_SAFE_MODE, CMD))
+			break;
+		udelay(1);
+	}
+}
+
+static void gfx_v12_0_unset_safe_mode(struct amdgpu_device *adev,
+				      int xcc_id)
+{
+	WREG32_SOC15(GC, 0, regRLC_SAFE_MODE, RLC_SAFE_MODE__CMD_MASK);
+}
+
+static void gfx_v12_0_update_perf_clk(struct amdgpu_device *adev,
+				      bool enable)
+{
+	uint32_t def, data;
+
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_GFX_PERF_CLK))
+		return;
+
+	def = data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE);
+
+	if (enable)
+		data &= ~RLC_CGTT_MGCG_OVERRIDE__PERFMON_CLOCK_STATE_MASK;
+	else
+		data |= RLC_CGTT_MGCG_OVERRIDE__PERFMON_CLOCK_STATE_MASK;
+
+	if (def != data)
+		WREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE, data);
+}
+
+static void gfx_v12_0_update_spm_vmid(struct amdgpu_device *adev,
+				      struct amdgpu_ring *ring,
+				      unsigned vmid)
+{
+	u32 reg, data;
+
+	reg = SOC15_REG_OFFSET(GC, 0, regRLC_SPM_MC_CNTL);
+	if (amdgpu_sriov_is_pp_one_vf(adev))
+		data = RREG32_NO_KIQ(reg);
+	else
+		data = RREG32(reg);
+
+	data &= ~RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK;
+	data |= (vmid & RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK) << RLC_SPM_MC_CNTL__RLC_SPM_VMID__SHIFT;
+
+	if (amdgpu_sriov_is_pp_one_vf(adev))
+		WREG32_SOC15_NO_KIQ(GC, 0, regRLC_SPM_MC_CNTL, data);
+	else
+		WREG32_SOC15(GC, 0, regRLC_SPM_MC_CNTL, data);
+
+	if (ring
+	    && amdgpu_sriov_is_pp_one_vf(adev)
+	    && ((ring->funcs->type == AMDGPU_RING_TYPE_GFX)
+		|| (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE))) {
+		uint32_t reg = SOC15_REG_OFFSET(GC, 0, regRLC_SPM_MC_CNTL);
+		amdgpu_ring_emit_wreg(ring, reg, data);
+	}
+}
+
+static const struct amdgpu_rlc_funcs gfx_v12_0_rlc_funcs = {
+	.is_rlc_enabled = gfx_v12_0_is_rlc_enabled,
+	.set_safe_mode = gfx_v12_0_set_safe_mode,
+	.unset_safe_mode = gfx_v12_0_unset_safe_mode,
+	.init = gfx_v12_0_rlc_init,
+	.get_csb_size = gfx_v12_0_get_csb_size,
+	.get_csb_buffer = gfx_v12_0_get_csb_buffer,
+	.resume = gfx_v12_0_rlc_resume,
+	.stop = gfx_v12_0_rlc_stop,
+	.reset = gfx_v12_0_rlc_reset,
+	.start = gfx_v12_0_rlc_start,
+	.update_spm_vmid = gfx_v12_0_update_spm_vmid,
+};
+
+#if 0
+static void gfx_v12_cntl_power_gating(struct amdgpu_device *adev, bool enable)
+{
+	/* TODO */
+}
+
+static void gfx_v12_cntl_pg(struct amdgpu_device *adev, bool enable)
+{
+	/* TODO */
+}
+#endif
+
+static int gfx_v12_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					   enum amd_powergating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	bool enable = (state == AMD_PG_STATE_GATE);
+
+	if (amdgpu_sriov_vf(adev))
+		return 0;
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		amdgpu_gfx_off_ctrl(adev, enable);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static void gfx_v12_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev,
+						       bool enable)
+{
+	uint32_t def, data;
+
+	if (!(adev->cg_flags &
+	      (AMD_CG_SUPPORT_GFX_CGCG |
+	      AMD_CG_SUPPORT_GFX_CGLS |
+	      AMD_CG_SUPPORT_GFX_3D_CGCG |
+	      AMD_CG_SUPPORT_GFX_3D_CGLS)))
+		return;
+
+	if (enable) {
+		def = data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE);
+
+		/* unset CGCG override */
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)
+			data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_CGCG_OVERRIDE_MASK;
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS)
+			data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_CGLS_OVERRIDE_MASK;
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGCG ||
+		    adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGLS)
+			data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_GFX3D_CG_OVERRIDE_MASK;
+
+		/* update CGCG override bits */
+		if (def != data)
+			WREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE, data);
+
+		/* enable cgcg FSM(0x0000363F) */
+		def = data = RREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL);
+
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG) {
+			data &= ~RLC_CGCG_CGLS_CTRL__CGCG_GFX_IDLE_THRESHOLD_MASK;
+			data |= (0x36 << RLC_CGCG_CGLS_CTRL__CGCG_GFX_IDLE_THRESHOLD__SHIFT) |
+				 RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK;
+		}
+
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS) {
+			data &= ~RLC_CGCG_CGLS_CTRL__CGLS_REP_COMPANSAT_DELAY_MASK;
+			data |= (0x000F << RLC_CGCG_CGLS_CTRL__CGLS_REP_COMPANSAT_DELAY__SHIFT) |
+				 RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK;
+		}
+
+		if (def != data)
+			WREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL, data);
+
+		/* Program RLC_CGCG_CGLS_CTRL_3D */
+		def = data = RREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL_3D);
+
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGCG) {
+			data &= ~RLC_CGCG_CGLS_CTRL_3D__CGCG_GFX_IDLE_THRESHOLD_MASK;
+			data |= (0x36 << RLC_CGCG_CGLS_CTRL_3D__CGCG_GFX_IDLE_THRESHOLD__SHIFT) |
+				 RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK;
+		}
+
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGLS) {
+			data &= ~RLC_CGCG_CGLS_CTRL_3D__CGLS_REP_COMPANSAT_DELAY_MASK;
+			data |= (0xf << RLC_CGCG_CGLS_CTRL_3D__CGLS_REP_COMPANSAT_DELAY__SHIFT) |
+				 RLC_CGCG_CGLS_CTRL_3D__CGLS_EN_MASK;
+		}
+
+		if (def != data)
+			WREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL_3D, data);
+
+		/* set IDLE_POLL_COUNT(0x00900100) */
+		def = data = RREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_CNTL);
+
+		data &= ~(CP_RB_WPTR_POLL_CNTL__POLL_FREQUENCY_MASK | CP_RB_WPTR_POLL_CNTL__IDLE_POLL_COUNT_MASK);
+		data |= (0x0100 << CP_RB_WPTR_POLL_CNTL__POLL_FREQUENCY__SHIFT) |
+			(0x0090 << CP_RB_WPTR_POLL_CNTL__IDLE_POLL_COUNT__SHIFT);
+
+		if (def != data)
+			WREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_CNTL, data);
+
+		data = RREG32_SOC15(GC, 0, regCP_INT_CNTL);
+		data = REG_SET_FIELD(data, CP_INT_CNTL, CNTX_BUSY_INT_ENABLE, 1);
+		data = REG_SET_FIELD(data, CP_INT_CNTL, CNTX_EMPTY_INT_ENABLE, 1);
+		data = REG_SET_FIELD(data, CP_INT_CNTL, CMP_BUSY_INT_ENABLE, 1);
+		data = REG_SET_FIELD(data, CP_INT_CNTL, GFX_IDLE_INT_ENABLE, 1);
+		WREG32_SOC15(GC, 0, regCP_INT_CNTL, data);
+
+		data = RREG32_SOC15(GC, 0, regSDMA0_RLC_CGCG_CTRL);
+		data = REG_SET_FIELD(data, SDMA0_RLC_CGCG_CTRL, CGCG_INT_ENABLE, 1);
+		WREG32_SOC15(GC, 0, regSDMA0_RLC_CGCG_CTRL, data);
+
+		/* Some ASICs only have one SDMA instance, not need to configure SDMA1 */
+		if (adev->sdma.num_instances > 1) {
+			data = RREG32_SOC15(GC, 0, regSDMA1_RLC_CGCG_CTRL);
+			data = REG_SET_FIELD(data, SDMA1_RLC_CGCG_CTRL, CGCG_INT_ENABLE, 1);
+			WREG32_SOC15(GC, 0, regSDMA1_RLC_CGCG_CTRL, data);
+		}
+	} else {
+		/* Program RLC_CGCG_CGLS_CTRL */
+		def = data = RREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL);
+
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)
+			data &= ~RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK;
+
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS)
+			data &= ~RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK;
+
+		if (def != data)
+			WREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL, data);
+
+		/* Program RLC_CGCG_CGLS_CTRL_3D */
+		def = data = RREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL_3D);
+
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGCG)
+			data &= ~RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK;
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGLS)
+			data &= ~RLC_CGCG_CGLS_CTRL_3D__CGLS_EN_MASK;
+
+		if (def != data)
+			WREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL_3D, data);
+	}
+}
+
+static void gfx_v12_0_update_medium_grain_clock_gating(struct amdgpu_device *adev,
+						       bool enable)
+{
+	uint32_t data, def;
+	if (!(adev->cg_flags & (AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_MGLS)))
+		return;
+
+	/* It is disabled by HW by default */
+	if (enable) {
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG) {
+			/* 1 - RLC_CGTT_MGCG_OVERRIDE */
+			def = data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE);
+
+			data &= ~(RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK |
+				  RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK |
+				  RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK);
+
+			if (def != data)
+				WREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE, data);
+		}
+	} else {
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG) {
+			def = data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE);
+
+			data |= (RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK |
+				 RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK |
+				 RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK);
+
+			if (def != data)
+				WREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE, data);
+		}
+	}
+}
+
+static void gfx_v12_0_update_repeater_fgcg(struct amdgpu_device *adev,
+					   bool enable)
+{
+	uint32_t def, data;
+
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_REPEATER_FGCG))
+		return;
+
+	def = data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE);
+
+	if (enable)
+		data &= ~(RLC_CGTT_MGCG_OVERRIDE__GFXIP_REPEATER_FGCG_OVERRIDE_MASK |
+				  RLC_CGTT_MGCG_OVERRIDE__RLC_REPEATER_FGCG_OVERRIDE_MASK);
+	else
+		data |= RLC_CGTT_MGCG_OVERRIDE__GFXIP_REPEATER_FGCG_OVERRIDE_MASK |
+				RLC_CGTT_MGCG_OVERRIDE__RLC_REPEATER_FGCG_OVERRIDE_MASK;
+
+	if (def != data)
+		WREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE, data);
+}
+
+static void gfx_v12_0_update_sram_fgcg(struct amdgpu_device *adev,
+				       bool enable)
+{
+	uint32_t def, data;
+
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_GFX_FGCG))
+		return;
+
+	def = data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE);
+
+	if (enable)
+		data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_FGCG_OVERRIDE_MASK;
+	else
+		data |= RLC_CGTT_MGCG_OVERRIDE__GFXIP_FGCG_OVERRIDE_MASK;
+
+	if (def != data)
+		WREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE, data);
+}
+
+static int gfx_v12_0_update_gfx_clock_gating(struct amdgpu_device *adev,
+					    bool enable)
+{
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
+
+	gfx_v12_0_update_coarse_grain_clock_gating(adev, enable);
+
+	gfx_v12_0_update_medium_grain_clock_gating(adev, enable);
+
+	gfx_v12_0_update_repeater_fgcg(adev, enable);
+
+	gfx_v12_0_update_sram_fgcg(adev, enable);
+
+	gfx_v12_0_update_perf_clk(adev, enable);
+
+	if (adev->cg_flags &
+	    (AMD_CG_SUPPORT_GFX_MGCG |
+	     AMD_CG_SUPPORT_GFX_CGLS |
+	     AMD_CG_SUPPORT_GFX_CGCG |
+	     AMD_CG_SUPPORT_GFX_3D_CGCG |
+	     AMD_CG_SUPPORT_GFX_3D_CGLS))
+		gfx_v12_0_enable_gui_idle_interrupt(adev, enable);
+
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
+
+	return 0;
+}
+
+static int gfx_v12_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					   enum amd_clockgating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (amdgpu_sriov_vf(adev))
+		return 0;
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		gfx_v12_0_update_gfx_clock_gating(adev,
+						  state == AMD_CG_STATE_GATE);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static void gfx_v12_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int data;
+
+	/* AMD_CG_SUPPORT_GFX_MGCG */
+	data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE);
+	if (!(data & RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK))
+		*flags |= AMD_CG_SUPPORT_GFX_MGCG;
+
+	/* AMD_CG_SUPPORT_REPEATER_FGCG */
+	if (!(data & RLC_CGTT_MGCG_OVERRIDE__GFXIP_REPEATER_FGCG_OVERRIDE_MASK))
+		*flags |= AMD_CG_SUPPORT_REPEATER_FGCG;
+
+	/* AMD_CG_SUPPORT_GFX_FGCG */
+	if (!(data & RLC_CGTT_MGCG_OVERRIDE__GFXIP_FGCG_OVERRIDE_MASK))
+		*flags |= AMD_CG_SUPPORT_GFX_FGCG;
+
+	/* AMD_CG_SUPPORT_GFX_PERF_CLK */
+	if (!(data & RLC_CGTT_MGCG_OVERRIDE__PERFMON_CLOCK_STATE_MASK))
+		*flags |= AMD_CG_SUPPORT_GFX_PERF_CLK;
+
+	/* AMD_CG_SUPPORT_GFX_CGCG */
+	data = RREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL);
+	if (data & RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_GFX_CGCG;
+
+	/* AMD_CG_SUPPORT_GFX_CGLS */
+	if (data & RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_GFX_CGLS;
+
+	/* AMD_CG_SUPPORT_GFX_3D_CGCG */
+	data = RREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL_3D);
+	if (data & RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_GFX_3D_CGCG;
+
+	/* AMD_CG_SUPPORT_GFX_3D_CGLS */
+	if (data & RLC_CGCG_CGLS_CTRL_3D__CGLS_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_GFX_3D_CGLS;
+}
+
+static u64 gfx_v12_0_ring_get_rptr_gfx(struct amdgpu_ring *ring)
+{
+	/* gfx12 is 32bit rptr*/
+	return *(uint32_t *)ring->rptr_cpu_addr;
+}
+
+static u64 gfx_v12_0_ring_get_wptr_gfx(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u64 wptr;
+
+	/* XXX check if swapping is necessary on BE */
+	if (ring->use_doorbell) {
+		wptr = atomic64_read((atomic64_t *)ring->wptr_cpu_addr);
+	} else {
+		wptr = RREG32_SOC15(GC, 0, regCP_RB0_WPTR);
+		wptr += (u64)RREG32_SOC15(GC, 0, regCP_RB0_WPTR_HI) << 32;
+	}
+
+	return wptr;
+}
+
+static void gfx_v12_0_ring_set_wptr_gfx(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->use_doorbell) {
+		/* XXX check if swapping is necessary on BE */
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
+			     ring->wptr);
+		WDOORBELL64(ring->doorbell_index, ring->wptr);
+	} else {
+		WREG32_SOC15(GC, 0, regCP_RB0_WPTR,
+			     lower_32_bits(ring->wptr));
+		WREG32_SOC15(GC, 0, regCP_RB0_WPTR_HI,
+			     upper_32_bits(ring->wptr));
+	}
+}
+
+static u64 gfx_v12_0_ring_get_rptr_compute(struct amdgpu_ring *ring)
+{
+	/* gfx12 hardware is 32bit rptr */
+	return *(uint32_t *)ring->rptr_cpu_addr;
+}
+
+static u64 gfx_v12_0_ring_get_wptr_compute(struct amdgpu_ring *ring)
+{
+	u64 wptr;
+
+	/* XXX check if swapping is necessary on BE */
+	if (ring->use_doorbell)
+		wptr = atomic64_read((atomic64_t *)ring->wptr_cpu_addr);
+	else
+		BUG();
+	return wptr;
+}
+
+static void gfx_v12_0_ring_set_wptr_compute(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	/* XXX check if swapping is necessary on BE */
+	if (ring->use_doorbell) {
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
+			     ring->wptr);
+		WDOORBELL64(ring->doorbell_index, ring->wptr);
+	} else {
+		BUG(); /* only DOORBELL method supported on gfx12 now */
+	}
+}
+
+static void gfx_v12_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 ref_and_mask, reg_mem_engine;
+	const struct nbio_hdp_flush_reg *nbio_hf_reg = adev->nbio.hdp_flush_reg;
+
+	if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE) {
+		switch (ring->me) {
+		case 1:
+			ref_and_mask = nbio_hf_reg->ref_and_mask_cp2 << ring->pipe;
+			break;
+		case 2:
+			ref_and_mask = nbio_hf_reg->ref_and_mask_cp6 << ring->pipe;
+			break;
+		default:
+			return;
+		}
+		reg_mem_engine = 0;
+	} else {
+		ref_and_mask = nbio_hf_reg->ref_and_mask_cp0;
+		reg_mem_engine = 1; /* pfp */
+	}
+
+	gfx_v12_0_wait_reg_mem(ring, reg_mem_engine, 0, 1,
+			       adev->nbio.funcs->get_hdp_flush_req_offset(adev),
+			       adev->nbio.funcs->get_hdp_flush_done_offset(adev),
+			       ref_and_mask, ref_and_mask, 0x20);
+}
+
+static void gfx_v12_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
+				       struct amdgpu_job *job,
+				       struct amdgpu_ib *ib,
+				       uint32_t flags)
+{
+	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
+	u32 header, control = 0;
+
+	header = PACKET3(PACKET3_INDIRECT_BUFFER, 2);
+
+	control |= ib->length_dw | (vmid << 24);
+
+	amdgpu_ring_write(ring, header);
+	BUG_ON(ib->gpu_addr & 0x3); /* Dword align */
+	amdgpu_ring_write(ring,
+#ifdef __BIG_ENDIAN
+		(2 << 0) |
+#endif
+		lower_32_bits(ib->gpu_addr));
+	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
+	amdgpu_ring_write(ring, control);
+}
+
+static void gfx_v12_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
+					   struct amdgpu_job *job,
+					   struct amdgpu_ib *ib,
+					   uint32_t flags)
+{
+	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
+	u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vmid << 24);
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
+	BUG_ON(ib->gpu_addr & 0x3); /* Dword align */
+	amdgpu_ring_write(ring,
+#ifdef __BIG_ENDIAN
+				(2 << 0) |
+#endif
+				lower_32_bits(ib->gpu_addr));
+	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
+	amdgpu_ring_write(ring, control);
+}
+
+static void gfx_v12_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr,
+				     u64 seq, unsigned flags)
+{
+	bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT;
+	bool int_sel = flags & AMDGPU_FENCE_FLAG_INT;
+
+	/* RELEASE_MEM - flush caches, send int */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_RELEASE_MEM, 6));
+	amdgpu_ring_write(ring, (PACKET3_RELEASE_MEM_GCR_SEQ |
+				 PACKET3_RELEASE_MEM_GCR_GL2_WB |
+				 PACKET3_RELEASE_MEM_CACHE_POLICY(3) |
+				 PACKET3_RELEASE_MEM_EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) |
+				 PACKET3_RELEASE_MEM_EVENT_INDEX(5)));
+	amdgpu_ring_write(ring, (PACKET3_RELEASE_MEM_DATA_SEL(write64bit ? 2 : 1) |
+				 PACKET3_RELEASE_MEM_INT_SEL(int_sel ? 2 : 0)));
+
+	/*
+	 * the address should be Qword aligned if 64bit write, Dword
+	 * aligned if only send 32bit data low (discard data high)
+	 */
+	if (write64bit)
+		BUG_ON(addr & 0x7);
+	else
+		BUG_ON(addr & 0x3);
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	amdgpu_ring_write(ring, lower_32_bits(seq));
+	amdgpu_ring_write(ring, upper_32_bits(seq));
+	amdgpu_ring_write(ring, 0);
+}
+
+static void gfx_v12_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
+{
+	int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX);
+	uint32_t seq = ring->fence_drv.sync_seq;
+	uint64_t addr = ring->fence_drv.gpu_addr;
+
+	gfx_v12_0_wait_reg_mem(ring, usepfp, 1, 0, lower_32_bits(addr),
+			       upper_32_bits(addr), seq, 0xffffffff, 4);
+}
+
+static void gfx_v12_0_ring_invalidate_tlbs(struct amdgpu_ring *ring,
+				   uint16_t pasid, uint32_t flush_type,
+				   bool all_hub, uint8_t dst_sel)
+{
+	amdgpu_ring_write(ring, PACKET3(PACKET3_INVALIDATE_TLBS, 0));
+	amdgpu_ring_write(ring,
+			  PACKET3_INVALIDATE_TLBS_DST_SEL(dst_sel) |
+			  PACKET3_INVALIDATE_TLBS_ALL_HUB(all_hub) |
+			  PACKET3_INVALIDATE_TLBS_PASID(pasid) |
+			  PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(flush_type));
+}
+
+static void gfx_v12_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
+					 unsigned vmid, uint64_t pd_addr)
+{
+	amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
+
+	/* compute doesn't have PFP */
+	if (ring->funcs->type == AMDGPU_RING_TYPE_GFX) {
+		/* sync PFP to ME, otherwise we might get invalid PFP reads */
+		amdgpu_ring_write(ring, PACKET3(PACKET3_PFP_SYNC_ME, 0));
+		amdgpu_ring_write(ring, 0x0);
+	}
+}
+
+static void gfx_v12_0_ring_emit_fence_kiq(struct amdgpu_ring *ring, u64 addr,
+					  u64 seq, unsigned int flags)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	/* we only allocate 32bit for each seq wb address */
+	BUG_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
+
+	/* write fence seq to the "addr" */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+				 WRITE_DATA_DST_SEL(5) | WR_CONFIRM));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	amdgpu_ring_write(ring, lower_32_bits(seq));
+
+	if (flags & AMDGPU_FENCE_FLAG_INT) {
+		/* set register to trigger INT */
+		amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+		amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+					 WRITE_DATA_DST_SEL(0) | WR_CONFIRM));
+		amdgpu_ring_write(ring, SOC15_REG_OFFSET(GC, 0, regCPC_INT_STATUS));
+		amdgpu_ring_write(ring, 0);
+		amdgpu_ring_write(ring, 0x20000000); /* src_id is 178 */
+	}
+}
+
+static void gfx_v12_0_ring_emit_cntxcntl(struct amdgpu_ring *ring,
+					 uint32_t flags)
+{
+	uint32_t dw2 = 0;
+
+	dw2 |= 0x80000000; /* set load_enable otherwise this package is just NOPs */
+	if (flags & AMDGPU_HAVE_CTX_SWITCH) {
+		/* set load_global_config & load_global_uconfig */
+		dw2 |= 0x8001;
+		/* set load_cs_sh_regs */
+		dw2 |= 0x01000000;
+		/* set load_per_context_state & load_gfx_sh_regs for GFX */
+		dw2 |= 0x10002;
+	}
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_CONTEXT_CONTROL, 1));
+	amdgpu_ring_write(ring, dw2);
+	amdgpu_ring_write(ring, 0);
+}
+
+static unsigned gfx_v12_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring,
+						   uint64_t addr)
+{
+	unsigned ret;
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	/* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, 0);
+	ret = ring->wptr & ring->buf_mask;
+	/* patch dummy value later */
+	amdgpu_ring_write(ring, 0);
+
+	return ret;
+}
+
+static int gfx_v12_0_ring_preempt_ib(struct amdgpu_ring *ring)
+{
+	int i, r = 0;
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
+	unsigned long flags;
+
+	if (adev->enable_mes)
+		return -EINVAL;
+
+	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
+		return -EINVAL;
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+
+	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size)) {
+		spin_unlock_irqrestore(&kiq->ring_lock, flags);
+		return -ENOMEM;
+	}
+
+	/* assert preemption condition */
+	amdgpu_ring_set_preempt_cond_exec(ring, false);
+
+	/* assert IB preemption, emit the trailing fence */
+	kiq->pmf->kiq_unmap_queues(kiq_ring, ring, PREEMPT_QUEUES_NO_UNMAP,
+				   ring->trail_fence_gpu_addr,
+				   ++ring->trail_seq);
+	amdgpu_ring_commit(kiq_ring);
+
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+
+	/* poll the trailing fence */
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (ring->trail_seq ==
+		    le32_to_cpu(*(ring->trail_fence_cpu_addr)))
+			break;
+		udelay(1);
+	}
+
+	if (i >= adev->usec_timeout) {
+		r = -EINVAL;
+		DRM_ERROR("ring %d failed to preempt ib\n", ring->idx);
+	}
+
+	/* deassert preemption condition */
+	amdgpu_ring_set_preempt_cond_exec(ring, true);
+	return r;
+}
+
+static void gfx_v12_0_ring_emit_frame_cntl(struct amdgpu_ring *ring,
+					   bool start,
+					   bool secure)
+{
+	uint32_t v = secure ? FRAME_TMZ : 0;
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_FRAME_CONTROL, 0));
+	amdgpu_ring_write(ring, v | FRAME_CMD(start ? 0 : 1));
+}
+
+static void gfx_v12_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg,
+				     uint32_t reg_val_offs)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_COPY_DATA, 4));
+	amdgpu_ring_write(ring, 0 |	/* src: register*/
+				(5 << 8) |	/* dst: memory */
+				(1 << 20));	/* write confirm */
+	amdgpu_ring_write(ring, reg);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +
+				reg_val_offs * 4));
+	amdgpu_ring_write(ring, upper_32_bits(adev->wb.gpu_addr +
+				reg_val_offs * 4));
+}
+
+static void gfx_v12_0_ring_emit_wreg(struct amdgpu_ring *ring,
+				     uint32_t reg,
+				     uint32_t val)
+{
+	uint32_t cmd = 0;
+
+	switch (ring->funcs->type) {
+	case AMDGPU_RING_TYPE_GFX:
+		cmd = WRITE_DATA_ENGINE_SEL(1) | WR_CONFIRM;
+		break;
+	case AMDGPU_RING_TYPE_KIQ:
+		cmd = (1 << 16); /* no inc addr */
+		break;
+	default:
+		cmd = WR_CONFIRM;
+		break;
+	}
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, cmd);
+	amdgpu_ring_write(ring, reg);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, val);
+}
+
+static void gfx_v12_0_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
+					uint32_t val, uint32_t mask)
+{
+	gfx_v12_0_wait_reg_mem(ring, 0, 0, 0, reg, 0, val, mask, 0x20);
+}
+
+static void gfx_v12_0_ring_emit_reg_write_reg_wait(struct amdgpu_ring *ring,
+						   uint32_t reg0, uint32_t reg1,
+						   uint32_t ref, uint32_t mask)
+{
+	int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX);
+
+	gfx_v12_0_wait_reg_mem(ring, usepfp, 0, 1, reg0, reg1,
+			       ref, mask, 0x20);
+}
+
+static void
+gfx_v12_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev,
+				      uint32_t me, uint32_t pipe,
+				      enum amdgpu_interrupt_state state)
+{
+	uint32_t cp_int_cntl, cp_int_cntl_reg;
+
+	if (!me) {
+		switch (pipe) {
+		case 0:
+			cp_int_cntl_reg = SOC15_REG_OFFSET(GC, 0, regCP_INT_CNTL_RING0);
+			break;
+		default:
+			DRM_DEBUG("invalid pipe %d\n", pipe);
+			return;
+		}
+	} else {
+		DRM_DEBUG("invalid me %d\n", me);
+		return;
+	}
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+		cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+					    TIME_STAMP_INT_ENABLE, 0);
+		cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+					    GENERIC0_INT_ENABLE, 0);
+		WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+		break;
+	case AMDGPU_IRQ_STATE_ENABLE:
+		cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+		cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+					    TIME_STAMP_INT_ENABLE, 1);
+		cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+					    GENERIC0_INT_ENABLE, 1);
+		WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+		break;
+	default:
+		break;
+	}
+}
+
+static void gfx_v12_0_set_compute_eop_interrupt_state(struct amdgpu_device *adev,
+						     int me, int pipe,
+						     enum amdgpu_interrupt_state state)
+{
+	u32 mec_int_cntl, mec_int_cntl_reg;
+
+	/*
+	 * amdgpu controls only the first MEC. That's why this function only
+	 * handles the setting of interrupts for this specific MEC. All other
+	 * pipes' interrupts are set by amdkfd.
+	 */
+
+	if (me == 1) {
+		switch (pipe) {
+		case 0:
+			mec_int_cntl_reg = SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE0_INT_CNTL);
+			break;
+		case 1:
+			mec_int_cntl_reg = SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE1_INT_CNTL);
+			break;
+		default:
+			DRM_DEBUG("invalid pipe %d\n", pipe);
+			return;
+		}
+	} else {
+		DRM_DEBUG("invalid me %d\n", me);
+		return;
+	}
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		mec_int_cntl = RREG32_SOC15_IP(GC, mec_int_cntl_reg);
+		mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+					     TIME_STAMP_INT_ENABLE, 0);
+		mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+					     GENERIC0_INT_ENABLE, 0);
+		WREG32_SOC15_IP(GC, mec_int_cntl_reg, mec_int_cntl);
+		break;
+	case AMDGPU_IRQ_STATE_ENABLE:
+		mec_int_cntl = RREG32_SOC15_IP(GC, mec_int_cntl_reg);
+		mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+					     TIME_STAMP_INT_ENABLE, 1);
+		mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+					     GENERIC0_INT_ENABLE, 1);
+		WREG32_SOC15_IP(GC, mec_int_cntl_reg, mec_int_cntl);
+		break;
+	default:
+		break;
+	}
+}
+
+static int gfx_v12_0_set_eop_interrupt_state(struct amdgpu_device *adev,
+					    struct amdgpu_irq_src *src,
+					    unsigned type,
+					    enum amdgpu_interrupt_state state)
+{
+	switch (type) {
+	case AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP:
+		gfx_v12_0_set_gfx_eop_interrupt_state(adev, 0, 0, state);
+		break;
+	case AMDGPU_CP_IRQ_GFX_ME0_PIPE1_EOP:
+		gfx_v12_0_set_gfx_eop_interrupt_state(adev, 0, 1, state);
+		break;
+	case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP:
+		gfx_v12_0_set_compute_eop_interrupt_state(adev, 1, 0, state);
+		break;
+	case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE1_EOP:
+		gfx_v12_0_set_compute_eop_interrupt_state(adev, 1, 1, state);
+		break;
+	case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE2_EOP:
+		gfx_v12_0_set_compute_eop_interrupt_state(adev, 1, 2, state);
+		break;
+	case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE3_EOP:
+		gfx_v12_0_set_compute_eop_interrupt_state(adev, 1, 3, state);
+		break;
+	default:
+		break;
+	}
+	return 0;
+}
+
+static int gfx_v12_0_eop_irq(struct amdgpu_device *adev,
+			     struct amdgpu_irq_src *source,
+			     struct amdgpu_iv_entry *entry)
+{
+	u32 doorbell_offset = entry->src_data[0];
+	u8 me_id, pipe_id, queue_id;
+	struct amdgpu_ring *ring;
+	int i;
+
+	DRM_DEBUG("IH: CP EOP\n");
+
+	if (adev->enable_mes && doorbell_offset) {
+		struct amdgpu_userq_fence_driver *fence_drv = NULL;
+		struct xarray *xa = &adev->userq_xa;
+		unsigned long flags;
+
+		xa_lock_irqsave(xa, flags);
+		fence_drv = xa_load(xa, doorbell_offset);
+		if (fence_drv)
+			amdgpu_userq_fence_driver_process(fence_drv);
+		xa_unlock_irqrestore(xa, flags);
+	} else {
+		me_id = (entry->ring_id & 0x0c) >> 2;
+		pipe_id = (entry->ring_id & 0x03) >> 0;
+		queue_id = (entry->ring_id & 0x70) >> 4;
+
+		switch (me_id) {
+		case 0:
+			if (pipe_id == 0)
+				amdgpu_fence_process(&adev->gfx.gfx_ring[0]);
+			else
+				amdgpu_fence_process(&adev->gfx.gfx_ring[1]);
+			break;
+		case 1:
+		case 2:
+			for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+				ring = &adev->gfx.compute_ring[i];
+				/* Per-queue interrupt is supported for MEC starting from VI.
+				 * The interrupt can only be enabled/disabled per pipe instead
+				 * of per queue.
+				 */
+				if ((ring->me == me_id) &&
+				    (ring->pipe == pipe_id) &&
+				    (ring->queue == queue_id))
+					amdgpu_fence_process(ring);
+			}
+			break;
+		}
+	}
+
+	return 0;
+}
+
+static int gfx_v12_0_set_priv_reg_fault_state(struct amdgpu_device *adev,
+					      struct amdgpu_irq_src *source,
+					      unsigned int type,
+					      enum amdgpu_interrupt_state state)
+{
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+	case AMDGPU_IRQ_STATE_ENABLE:
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+				cp_int_cntl_reg = gfx_v12_0_get_cpg_int_cntl(adev, i, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+								    PRIV_REG_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				/* MECs start at 1 */
+				cp_int_cntl_reg = gfx_v12_0_get_cpc_int_cntl(adev, i + 1, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+								    PRIV_REG_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int gfx_v12_0_set_bad_op_fault_state(struct amdgpu_device *adev,
+					    struct amdgpu_irq_src *source,
+					    unsigned type,
+					    enum amdgpu_interrupt_state state)
+{
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+	case AMDGPU_IRQ_STATE_ENABLE:
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+				cp_int_cntl_reg = gfx_v12_0_get_cpg_int_cntl(adev, i, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+								    OPCODE_ERROR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				/* MECs start at 1 */
+				cp_int_cntl_reg = gfx_v12_0_get_cpc_int_cntl(adev, i + 1, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+								    OPCODE_ERROR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		break;
+	default:
+		break;
+	}
+	return 0;
+}
+
+static int gfx_v12_0_set_priv_inst_fault_state(struct amdgpu_device *adev,
+					       struct amdgpu_irq_src *source,
+					       unsigned int type,
+					       enum amdgpu_interrupt_state state)
+{
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+	case AMDGPU_IRQ_STATE_ENABLE:
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+				cp_int_cntl_reg = gfx_v12_0_get_cpg_int_cntl(adev, i, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+								    PRIV_INSTR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static void gfx_v12_0_handle_priv_fault(struct amdgpu_device *adev,
+					struct amdgpu_iv_entry *entry)
+{
+	u8 me_id, pipe_id, queue_id;
+	struct amdgpu_ring *ring;
+	int i;
+
+	me_id = (entry->ring_id & 0x0c) >> 2;
+	pipe_id = (entry->ring_id & 0x03) >> 0;
+	queue_id = (entry->ring_id & 0x70) >> 4;
+
+	if (!adev->gfx.disable_kq) {
+		switch (me_id) {
+		case 0:
+			for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+				ring = &adev->gfx.gfx_ring[i];
+				if (ring->me == me_id && ring->pipe == pipe_id &&
+				    ring->queue == queue_id)
+					drm_sched_fault(&ring->sched);
+			}
+			break;
+		case 1:
+		case 2:
+			for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+				ring = &adev->gfx.compute_ring[i];
+				if (ring->me == me_id && ring->pipe == pipe_id &&
+				    ring->queue == queue_id)
+					drm_sched_fault(&ring->sched);
+			}
+			break;
+		default:
+			BUG();
+			break;
+		}
+	}
+}
+
+static int gfx_v12_0_priv_reg_irq(struct amdgpu_device *adev,
+				  struct amdgpu_irq_src *source,
+				  struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal register access in command stream\n");
+	gfx_v12_0_handle_priv_fault(adev, entry);
+	return 0;
+}
+
+static int gfx_v12_0_bad_op_irq(struct amdgpu_device *adev,
+				struct amdgpu_irq_src *source,
+				struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal opcode in command stream \n");
+	gfx_v12_0_handle_priv_fault(adev, entry);
+	return 0;
+}
+
+static int gfx_v12_0_priv_inst_irq(struct amdgpu_device *adev,
+				   struct amdgpu_irq_src *source,
+				   struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal instruction in command stream\n");
+	gfx_v12_0_handle_priv_fault(adev, entry);
+	return 0;
+}
+
+static void gfx_v12_0_emit_mem_sync(struct amdgpu_ring *ring)
+{
+	const unsigned int gcr_cntl =
+			PACKET3_ACQUIRE_MEM_GCR_CNTL_GL2_INV(1) |
+			PACKET3_ACQUIRE_MEM_GCR_CNTL_GL2_WB(1) |
+			PACKET3_ACQUIRE_MEM_GCR_CNTL_GLM_INV(1) |
+			PACKET3_ACQUIRE_MEM_GCR_CNTL_GLM_WB(1) |
+			PACKET3_ACQUIRE_MEM_GCR_CNTL_GL1_INV(1) |
+			PACKET3_ACQUIRE_MEM_GCR_CNTL_GLV_INV(1) |
+			PACKET3_ACQUIRE_MEM_GCR_CNTL_GLK_INV(1) |
+			PACKET3_ACQUIRE_MEM_GCR_CNTL_GLI_INV(1);
+
+	/* ACQUIRE_MEM - make one or more surfaces valid for use by the subsequent operations */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_ACQUIRE_MEM, 6));
+	amdgpu_ring_write(ring, 0); /* CP_COHER_CNTL */
+	amdgpu_ring_write(ring, 0xffffffff);  /* CP_COHER_SIZE */
+	amdgpu_ring_write(ring, 0xffffff);  /* CP_COHER_SIZE_HI */
+	amdgpu_ring_write(ring, 0); /* CP_COHER_BASE */
+	amdgpu_ring_write(ring, 0);  /* CP_COHER_BASE_HI */
+	amdgpu_ring_write(ring, 0x0000000A); /* POLL_INTERVAL */
+	amdgpu_ring_write(ring, gcr_cntl); /* GCR_CNTL */
+}
+
+static void gfx_v12_ring_insert_nop(struct amdgpu_ring *ring, uint32_t num_nop)
+{
+	/* Header itself is a NOP packet */
+	if (num_nop == 1) {
+		amdgpu_ring_write(ring, ring->funcs->nop);
+		return;
+	}
+
+	/* Max HW optimization till 0x3ffe, followed by remaining one NOP at a time*/
+	amdgpu_ring_write(ring, PACKET3(PACKET3_NOP, min(num_nop - 2, 0x3ffe)));
+
+	/* Header is at index 0, followed by num_nops - 1 NOP packet's */
+	amdgpu_ring_insert_nop(ring, num_nop - 1);
+}
+
+static void gfx_v12_0_ring_emit_cleaner_shader(struct amdgpu_ring *ring)
+{
+	/* Emit the cleaner shader */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_RUN_CLEANER_SHADER, 0));
+	amdgpu_ring_write(ring, 0);  /* RESERVED field, programmed to zero */
+}
+
+static void gfx_v12_ip_print(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k, reg, index = 0;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_12_0);
+
+	if (!adev->gfx.ip_dump_core)
+		return;
+
+	for (i = 0; i < reg_count; i++)
+		drm_printf(p, "%-50s \t 0x%08x\n",
+			   gc_reg_list_12_0[i].reg_name,
+			   adev->gfx.ip_dump_core[i]);
+
+	/* print compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_12);
+	drm_printf(p, "\nnum_mec: %d num_pipe: %d num_queue: %d\n",
+		   adev->gfx.mec.num_mec,
+		   adev->gfx.mec.num_pipe_per_mec,
+		   adev->gfx.mec.num_queue_per_pipe);
+
+	for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+		for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+			for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+				drm_printf(p, "\nmec %d, pipe %d, queue %d\n", i, j, k);
+				for (reg = 0; reg < reg_count; reg++) {
+					drm_printf(p, "%-50s \t 0x%08x\n",
+						   gc_cp_reg_list_12[reg].reg_name,
+						   adev->gfx.ip_dump_compute_queues[index + reg]);
+				}
+				index += reg_count;
+			}
+		}
+	}
+
+	/* print gfx queue registers for all instances */
+	if (!adev->gfx.ip_dump_gfx_queues)
+		return;
+
+	index = 0;
+	reg_count = ARRAY_SIZE(gc_gfx_queue_reg_list_12);
+	drm_printf(p, "\nnum_me: %d num_pipe: %d num_queue: %d\n",
+		   adev->gfx.me.num_me,
+		   adev->gfx.me.num_pipe_per_me,
+		   adev->gfx.me.num_queue_per_pipe);
+
+	for (i = 0; i < adev->gfx.me.num_me; i++) {
+		for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+			for (k = 0; k < adev->gfx.me.num_queue_per_pipe; k++) {
+				drm_printf(p, "\nme %d, pipe %d, queue %d\n", i, j, k);
+				for (reg = 0; reg < reg_count; reg++) {
+					drm_printf(p, "%-50s \t 0x%08x\n",
+						   gc_gfx_queue_reg_list_12[reg].reg_name,
+						   adev->gfx.ip_dump_gfx_queues[index + reg]);
+				}
+				index += reg_count;
+			}
+		}
+	}
+}
+
+static void gfx_v12_ip_dump(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k, reg, index = 0;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_12_0);
+
+	if (!adev->gfx.ip_dump_core)
+		return;
+
+	amdgpu_gfx_off_ctrl(adev, false);
+	for (i = 0; i < reg_count; i++)
+		adev->gfx.ip_dump_core[i] = RREG32(SOC15_REG_ENTRY_OFFSET(gc_reg_list_12_0[i]));
+	amdgpu_gfx_off_ctrl(adev, true);
+
+	/* dump compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_12);
+	amdgpu_gfx_off_ctrl(adev, false);
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+		for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+			for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+				/* ME0 is for GFX so start from 1 for CP */
+				soc24_grbm_select(adev, adev->gfx.me.num_me + i, j, k, 0);
+				for (reg = 0; reg < reg_count; reg++) {
+					adev->gfx.ip_dump_compute_queues[index + reg] =
+						RREG32(SOC15_REG_ENTRY_OFFSET(
+							gc_cp_reg_list_12[reg]));
+				}
+				index += reg_count;
+			}
+		}
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	amdgpu_gfx_off_ctrl(adev, true);
+
+	/* dump gfx queue registers for all instances */
+	if (!adev->gfx.ip_dump_gfx_queues)
+		return;
+
+	index = 0;
+	reg_count = ARRAY_SIZE(gc_gfx_queue_reg_list_12);
+	amdgpu_gfx_off_ctrl(adev, false);
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->gfx.me.num_me; i++) {
+		for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+			for (k = 0; k < adev->gfx.me.num_queue_per_pipe; k++) {
+				soc24_grbm_select(adev, i, j, k, 0);
+
+				for (reg = 0; reg < reg_count; reg++) {
+					adev->gfx.ip_dump_gfx_queues[index + reg] =
+						RREG32(SOC15_REG_ENTRY_OFFSET(
+							gc_gfx_queue_reg_list_12[reg]));
+				}
+				index += reg_count;
+			}
+		}
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	amdgpu_gfx_off_ctrl(adev, true);
+}
+
+static bool gfx_v12_pipe_reset_support(struct amdgpu_device *adev)
+{
+	/* Disable the pipe reset until the CPFW fully support it.*/
+	dev_warn_once(adev->dev, "The CPFW hasn't support pipe reset yet.\n");
+	return false;
+}
+
+static int gfx_v12_reset_gfx_pipe(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t reset_pipe = 0, clean_pipe = 0;
+	int r;
+
+	if (!gfx_v12_pipe_reset_support(adev))
+		return -EOPNOTSUPP;
+
+	gfx_v12_0_set_safe_mode(adev, 0);
+	mutex_lock(&adev->srbm_mutex);
+	soc24_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+
+	switch (ring->pipe) {
+	case 0:
+		reset_pipe = REG_SET_FIELD(reset_pipe, CP_ME_CNTL,
+					   PFP_PIPE0_RESET, 1);
+		reset_pipe = REG_SET_FIELD(reset_pipe, CP_ME_CNTL,
+					   ME_PIPE0_RESET, 1);
+		clean_pipe = REG_SET_FIELD(clean_pipe, CP_ME_CNTL,
+					   PFP_PIPE0_RESET, 0);
+		clean_pipe = REG_SET_FIELD(clean_pipe, CP_ME_CNTL,
+					   ME_PIPE0_RESET, 0);
+		break;
+	case 1:
+		reset_pipe = REG_SET_FIELD(reset_pipe, CP_ME_CNTL,
+					   PFP_PIPE1_RESET, 1);
+		reset_pipe = REG_SET_FIELD(reset_pipe, CP_ME_CNTL,
+					   ME_PIPE1_RESET, 1);
+		clean_pipe = REG_SET_FIELD(clean_pipe, CP_ME_CNTL,
+					   PFP_PIPE1_RESET, 0);
+		clean_pipe = REG_SET_FIELD(clean_pipe, CP_ME_CNTL,
+					   ME_PIPE1_RESET, 0);
+		break;
+	default:
+		break;
+	}
+
+	WREG32_SOC15(GC, 0, regCP_ME_CNTL, reset_pipe);
+	WREG32_SOC15(GC, 0, regCP_ME_CNTL, clean_pipe);
+
+	r = (RREG32(SOC15_REG_OFFSET(GC, 0, regCP_GFX_RS64_INSTR_PNTR1)) << 2) -
+					RS64_FW_UC_START_ADDR_LO;
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	gfx_v12_0_unset_safe_mode(adev, 0);
+
+	dev_info(adev->dev, "The ring %s pipe reset: %s\n", ring->name,
+			r == 0 ? "successfully" : "failed");
+	/* Sometimes the ME start pc counter can't cache correctly, so the
+	 * PC check only as a reference and pipe reset result rely on the
+	 * later ring test.
+	 */
+	return 0;
+}
+
+static int gfx_v12_0_reset_kgq(struct amdgpu_ring *ring,
+			       unsigned int vmid,
+			       struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int r;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	r = amdgpu_mes_reset_legacy_queue(ring->adev, ring, vmid, false);
+	if (r) {
+		dev_warn(adev->dev, "reset via MES failed and try pipe reset %d\n", r);
+		r = gfx_v12_reset_gfx_pipe(ring);
+		if (r)
+			return r;
+	}
+
+	r = gfx_v12_0_kgq_init_queue(ring, true);
+	if (r) {
+		dev_err(adev->dev, "failed to init kgq\n");
+		return r;
+	}
+
+	r = amdgpu_mes_map_legacy_queue(adev, ring);
+	if (r) {
+		dev_err(adev->dev, "failed to remap kgq\n");
+		return r;
+	}
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static int gfx_v12_0_reset_compute_pipe(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t reset_pipe = 0, clean_pipe = 0;
+	int r = 0;
+
+	if (!gfx_v12_pipe_reset_support(adev))
+		return -EOPNOTSUPP;
+
+	gfx_v12_0_set_safe_mode(adev, 0);
+	mutex_lock(&adev->srbm_mutex);
+	soc24_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+
+	reset_pipe = RREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL);
+	clean_pipe = reset_pipe;
+
+	if (adev->gfx.rs64_enable) {
+		switch (ring->pipe) {
+		case 0:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE0_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE0_RESET, 0);
+			break;
+		case 1:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE1_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE1_RESET, 0);
+			break;
+		case 2:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE2_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE2_RESET, 0);
+			break;
+		case 3:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE3_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE3_RESET, 0);
+			break;
+		default:
+			break;
+		}
+		WREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL, reset_pipe);
+		WREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL, clean_pipe);
+		r = (RREG32_SOC15(GC, 0, regCP_MEC_RS64_INSTR_PNTR) << 2) -
+				RS64_FW_UC_START_ADDR_LO;
+	} else {
+		switch (ring->pipe) {
+		case 0:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE0_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE0_RESET, 0);
+			break;
+		case 1:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE1_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE1_RESET, 0);
+			break;
+		default:
+		break;
+		}
+		WREG32_SOC15(GC, 0, regCP_MEC_CNTL, reset_pipe);
+		WREG32_SOC15(GC, 0, regCP_MEC_CNTL, clean_pipe);
+		/* Doesn't find the F32 MEC instruction pointer register, and suppose
+		 * the driver won't run into the F32 mode.
+		 */
+	}
+
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	gfx_v12_0_unset_safe_mode(adev, 0);
+
+	dev_info(adev->dev, "The ring %s pipe resets: %s\n", ring->name,
+			r == 0 ? "successfully" : "failed");
+	/* Need the ring test to verify the pipe reset result.*/
+	return 0;
+}
+
+static int gfx_v12_0_reset_kcq(struct amdgpu_ring *ring,
+			       unsigned int vmid,
+			       struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int r;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	r = amdgpu_mes_reset_legacy_queue(ring->adev, ring, vmid, true);
+	if (r) {
+		dev_warn(adev->dev, "fail(%d) to reset kcq  and try pipe reset\n", r);
+		r = gfx_v12_0_reset_compute_pipe(ring);
+		if (r)
+			return r;
+	}
+
+	r = gfx_v12_0_kcq_init_queue(ring, true);
+	if (r) {
+		dev_err(adev->dev, "failed to init kcq\n");
+		return r;
+	}
+	r = amdgpu_mes_map_legacy_queue(adev, ring);
+	if (r) {
+		dev_err(adev->dev, "failed to remap kcq\n");
+		return r;
+	}
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static void gfx_v12_0_ring_begin_use(struct amdgpu_ring *ring)
+{
+	amdgpu_gfx_profile_ring_begin_use(ring);
+
+	amdgpu_gfx_enforce_isolation_ring_begin_use(ring);
+}
+
+static void gfx_v12_0_ring_end_use(struct amdgpu_ring *ring)
+{
+	amdgpu_gfx_profile_ring_end_use(ring);
+
+	amdgpu_gfx_enforce_isolation_ring_end_use(ring);
+}
+
+static const struct amd_ip_funcs gfx_v12_0_ip_funcs = {
+	.name = "gfx_v12_0",
+	.early_init = gfx_v12_0_early_init,
+	.late_init = gfx_v12_0_late_init,
+	.sw_init = gfx_v12_0_sw_init,
+	.sw_fini = gfx_v12_0_sw_fini,
+	.hw_init = gfx_v12_0_hw_init,
+	.hw_fini = gfx_v12_0_hw_fini,
+	.suspend = gfx_v12_0_suspend,
+	.resume = gfx_v12_0_resume,
+	.is_idle = gfx_v12_0_is_idle,
+	.wait_for_idle = gfx_v12_0_wait_for_idle,
+	.set_clockgating_state = gfx_v12_0_set_clockgating_state,
+	.set_powergating_state = gfx_v12_0_set_powergating_state,
+	.get_clockgating_state = gfx_v12_0_get_clockgating_state,
+	.dump_ip_state = gfx_v12_ip_dump,
+	.print_ip_state = gfx_v12_ip_print,
+};
+
+static const struct amdgpu_ring_funcs gfx_v12_0_ring_funcs_gfx = {
+	.type = AMDGPU_RING_TYPE_GFX,
+	.align_mask = 0xff,
+	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
+	.support_64bit_ptrs = true,
+	.secure_submission_supported = true,
+	.get_rptr = gfx_v12_0_ring_get_rptr_gfx,
+	.get_wptr = gfx_v12_0_ring_get_wptr_gfx,
+	.set_wptr = gfx_v12_0_ring_set_wptr_gfx,
+	.emit_frame_size = /* totally 242 maximum if 16 IBs */
+		5 + /* COND_EXEC */
+		7 + /* PIPELINE_SYNC */
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
+		2 + /* VM_FLUSH */
+		8 + /* FENCE for VM_FLUSH */
+		5 + /* COND_EXEC */
+		7 + /* HDP_flush */
+		4 + /* VGT_flush */
+		31 + /*	DE_META */
+		3 + /* CNTX_CTRL */
+		5 + /* HDP_INVL */
+		8 + 8 + /* FENCE x2 */
+		8 + /* gfx_v12_0_emit_mem_sync */
+		2, /* gfx_v12_0_ring_emit_cleaner_shader */
+	.emit_ib_size =	4, /* gfx_v12_0_ring_emit_ib_gfx */
+	.emit_ib = gfx_v12_0_ring_emit_ib_gfx,
+	.emit_fence = gfx_v12_0_ring_emit_fence,
+	.emit_pipeline_sync = gfx_v12_0_ring_emit_pipeline_sync,
+	.emit_vm_flush = gfx_v12_0_ring_emit_vm_flush,
+	.emit_hdp_flush = gfx_v12_0_ring_emit_hdp_flush,
+	.test_ring = gfx_v12_0_ring_test_ring,
+	.test_ib = gfx_v12_0_ring_test_ib,
+	.insert_nop = gfx_v12_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.emit_cntxcntl = gfx_v12_0_ring_emit_cntxcntl,
+	.init_cond_exec = gfx_v12_0_ring_emit_init_cond_exec,
+	.preempt_ib = gfx_v12_0_ring_preempt_ib,
+	.emit_frame_cntl = gfx_v12_0_ring_emit_frame_cntl,
+	.emit_wreg = gfx_v12_0_ring_emit_wreg,
+	.emit_reg_wait = gfx_v12_0_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = gfx_v12_0_ring_emit_reg_write_reg_wait,
+	.emit_mem_sync = gfx_v12_0_emit_mem_sync,
+	.reset = gfx_v12_0_reset_kgq,
+	.emit_cleaner_shader = gfx_v12_0_ring_emit_cleaner_shader,
+	.begin_use = gfx_v12_0_ring_begin_use,
+	.end_use = gfx_v12_0_ring_end_use,
+};
+
+static const struct amdgpu_ring_funcs gfx_v12_0_ring_funcs_compute = {
+	.type = AMDGPU_RING_TYPE_COMPUTE,
+	.align_mask = 0xff,
+	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
+	.support_64bit_ptrs = true,
+	.get_rptr = gfx_v12_0_ring_get_rptr_compute,
+	.get_wptr = gfx_v12_0_ring_get_wptr_compute,
+	.set_wptr = gfx_v12_0_ring_set_wptr_compute,
+	.emit_frame_size =
+		7 + /* gfx_v12_0_ring_emit_hdp_flush */
+		5 + /* hdp invalidate */
+		7 + /* gfx_v12_0_ring_emit_pipeline_sync */
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
+		2 + /* gfx_v12_0_ring_emit_vm_flush */
+		8 + 8 + 8 + /* gfx_v12_0_ring_emit_fence x3 for user fence, vm fence */
+		8 + /* gfx_v12_0_emit_mem_sync */
+		2, /* gfx_v12_0_ring_emit_cleaner_shader */
+	.emit_ib_size =	7, /* gfx_v12_0_ring_emit_ib_compute */
+	.emit_ib = gfx_v12_0_ring_emit_ib_compute,
+	.emit_fence = gfx_v12_0_ring_emit_fence,
+	.emit_pipeline_sync = gfx_v12_0_ring_emit_pipeline_sync,
+	.emit_vm_flush = gfx_v12_0_ring_emit_vm_flush,
+	.emit_hdp_flush = gfx_v12_0_ring_emit_hdp_flush,
+	.test_ring = gfx_v12_0_ring_test_ring,
+	.test_ib = gfx_v12_0_ring_test_ib,
+	.insert_nop = gfx_v12_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.emit_wreg = gfx_v12_0_ring_emit_wreg,
+	.emit_reg_wait = gfx_v12_0_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = gfx_v12_0_ring_emit_reg_write_reg_wait,
+	.emit_mem_sync = gfx_v12_0_emit_mem_sync,
+	.reset = gfx_v12_0_reset_kcq,
+	.emit_cleaner_shader = gfx_v12_0_ring_emit_cleaner_shader,
+	.begin_use = gfx_v12_0_ring_begin_use,
+	.end_use = gfx_v12_0_ring_end_use,
+};
+
+static const struct amdgpu_ring_funcs gfx_v12_0_ring_funcs_kiq = {
+	.type = AMDGPU_RING_TYPE_KIQ,
+	.align_mask = 0xff,
+	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
+	.support_64bit_ptrs = true,
+	.get_rptr = gfx_v12_0_ring_get_rptr_compute,
+	.get_wptr = gfx_v12_0_ring_get_wptr_compute,
+	.set_wptr = gfx_v12_0_ring_set_wptr_compute,
+	.emit_frame_size =
+		7 + /* gfx_v12_0_ring_emit_hdp_flush */
+		5 + /*hdp invalidate */
+		7 + /* gfx_v12_0_ring_emit_pipeline_sync */
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
+		2 + /* gfx_v12_0_ring_emit_vm_flush */
+		8 + 8 + 8, /* gfx_v12_0_ring_emit_fence_kiq x3 for user fence, vm fence */
+	.emit_ib_size =	7, /* gfx_v12_0_ring_emit_ib_compute */
+	.emit_ib = gfx_v12_0_ring_emit_ib_compute,
+	.emit_fence = gfx_v12_0_ring_emit_fence_kiq,
+	.test_ring = gfx_v12_0_ring_test_ring,
+	.test_ib = gfx_v12_0_ring_test_ib,
+	.insert_nop = amdgpu_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.emit_rreg = gfx_v12_0_ring_emit_rreg,
+	.emit_wreg = gfx_v12_0_ring_emit_wreg,
+	.emit_reg_wait = gfx_v12_0_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = gfx_v12_0_ring_emit_reg_write_reg_wait,
+	.emit_hdp_flush = gfx_v12_0_ring_emit_hdp_flush,
+};
+
+static void gfx_v12_0_set_ring_funcs(struct amdgpu_device *adev)
+{
+	int i;
+
+	adev->gfx.kiq[0].ring.funcs = &gfx_v12_0_ring_funcs_kiq;
+
+	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
+		adev->gfx.gfx_ring[i].funcs = &gfx_v12_0_ring_funcs_gfx;
+
+	for (i = 0; i < adev->gfx.num_compute_rings; i++)
+		adev->gfx.compute_ring[i].funcs = &gfx_v12_0_ring_funcs_compute;
+}
+
+static const struct amdgpu_irq_src_funcs gfx_v12_0_eop_irq_funcs = {
+	.set = gfx_v12_0_set_eop_interrupt_state,
+	.process = gfx_v12_0_eop_irq,
+};
+
+static const struct amdgpu_irq_src_funcs gfx_v12_0_priv_reg_irq_funcs = {
+	.set = gfx_v12_0_set_priv_reg_fault_state,
+	.process = gfx_v12_0_priv_reg_irq,
+};
+
+static const struct amdgpu_irq_src_funcs gfx_v12_0_bad_op_irq_funcs = {
+	.set = gfx_v12_0_set_bad_op_fault_state,
+	.process = gfx_v12_0_bad_op_irq,
+};
+
+static const struct amdgpu_irq_src_funcs gfx_v12_0_priv_inst_irq_funcs = {
+	.set = gfx_v12_0_set_priv_inst_fault_state,
+	.process = gfx_v12_0_priv_inst_irq,
+};
+
+static void gfx_v12_0_set_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->gfx.eop_irq.num_types = AMDGPU_CP_IRQ_LAST;
+	adev->gfx.eop_irq.funcs = &gfx_v12_0_eop_irq_funcs;
+
+	adev->gfx.priv_reg_irq.num_types = 1;
+	adev->gfx.priv_reg_irq.funcs = &gfx_v12_0_priv_reg_irq_funcs;
+
+	adev->gfx.bad_op_irq.num_types = 1;
+	adev->gfx.bad_op_irq.funcs = &gfx_v12_0_bad_op_irq_funcs;
+
+	adev->gfx.priv_inst_irq.num_types = 1;
+	adev->gfx.priv_inst_irq.funcs = &gfx_v12_0_priv_inst_irq_funcs;
+}
+
+static void gfx_v12_0_set_imu_funcs(struct amdgpu_device *adev)
+{
+	if (adev->flags & AMD_IS_APU)
+		adev->gfx.imu.mode = MISSION_MODE;
+	else
+		adev->gfx.imu.mode = DEBUG_MODE;
+
+	adev->gfx.imu.funcs = &gfx_v12_0_imu_funcs;
+}
+
+static void gfx_v12_0_set_rlc_funcs(struct amdgpu_device *adev)
+{
+	adev->gfx.rlc.funcs = &gfx_v12_0_rlc_funcs;
+}
+
+static void gfx_v12_0_set_mqd_funcs(struct amdgpu_device *adev)
+{
+	/* set gfx eng mqd */
+	adev->mqds[AMDGPU_HW_IP_GFX].mqd_size =
+		sizeof(struct v12_gfx_mqd);
+	adev->mqds[AMDGPU_HW_IP_GFX].init_mqd =
+		gfx_v12_0_gfx_mqd_init;
+	/* set compute eng mqd */
+	adev->mqds[AMDGPU_HW_IP_COMPUTE].mqd_size =
+		sizeof(struct v12_compute_mqd);
+	adev->mqds[AMDGPU_HW_IP_COMPUTE].init_mqd =
+		gfx_v12_0_compute_mqd_init;
+}
+
+static void gfx_v12_0_set_user_wgp_inactive_bitmap_per_sh(struct amdgpu_device *adev,
+							  u32 bitmap)
+{
+	u32 data;
+
+	if (!bitmap)
+		return;
+
+	data = bitmap << GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_WGPS__SHIFT;
+	data &= GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_WGPS_MASK;
+
+	WREG32_SOC15(GC, 0, regGC_USER_SHADER_ARRAY_CONFIG, data);
+}
+
+static u32 gfx_v12_0_get_wgp_active_bitmap_per_sh(struct amdgpu_device *adev)
+{
+	u32 data, wgp_bitmask;
+	data = RREG32_SOC15(GC, 0, regCC_GC_SHADER_ARRAY_CONFIG);
+	data |= RREG32_SOC15(GC, 0, regGC_USER_SHADER_ARRAY_CONFIG);
+
+	data &= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_WGPS_MASK;
+	data >>= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_WGPS__SHIFT;
+
+	wgp_bitmask =
+		amdgpu_gfx_create_bitmask(adev->gfx.config.max_cu_per_sh >> 1);
+
+	return (~data) & wgp_bitmask;
+}
+
+static u32 gfx_v12_0_get_cu_active_bitmap_per_sh(struct amdgpu_device *adev)
+{
+	u32 wgp_idx, wgp_active_bitmap;
+	u32 cu_bitmap_per_wgp, cu_active_bitmap;
+
+	wgp_active_bitmap = gfx_v12_0_get_wgp_active_bitmap_per_sh(adev);
+	cu_active_bitmap = 0;
+
+	for (wgp_idx = 0; wgp_idx < 16; wgp_idx++) {
+		/* if there is one WGP enabled, it means 2 CUs will be enabled */
+		cu_bitmap_per_wgp = 3 << (2 * wgp_idx);
+		if (wgp_active_bitmap & (1 << wgp_idx))
+			cu_active_bitmap |= cu_bitmap_per_wgp;
+	}
+
+	return cu_active_bitmap;
+}
+
+static int gfx_v12_0_get_cu_info(struct amdgpu_device *adev,
+				 struct amdgpu_cu_info *cu_info)
+{
+	int i, j, k, counter, active_cu_number = 0;
+	u32 mask, bitmap;
+	unsigned disable_masks[8 * 2];
+
+	if (!adev || !cu_info)
+		return -EINVAL;
+
+	amdgpu_gfx_parse_disable_cu(disable_masks, 8, 2);
+
+	mutex_lock(&adev->grbm_idx_mutex);
+	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
+		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
+			bitmap = i * adev->gfx.config.max_sh_per_se + j;
+			if (!((gfx_v12_0_get_sa_active_bitmap(adev) >> bitmap) & 1))
+				continue;
+			mask = 1;
+			counter = 0;
+			gfx_v12_0_select_se_sh(adev, i, j, 0xffffffff, 0);
+			if (i < 8 && j < 2)
+				gfx_v12_0_set_user_wgp_inactive_bitmap_per_sh(
+					adev, disable_masks[i * 2 + j]);
+			bitmap = gfx_v12_0_get_cu_active_bitmap_per_sh(adev);
+
+			/**
+			 * GFX12 could support more than 4 SEs, while the bitmap
+			 * in cu_info struct is 4x4 and ioctl interface struct
+			 * drm_amdgpu_info_device should keep stable.
+			 * So we use last two columns of bitmap to store cu mask for
+			 * SEs 4 to 7, the layout of the bitmap is as below:
+			 *    SE0: {SH0,SH1} --> {bitmap[0][0], bitmap[0][1]}
+			 *    SE1: {SH0,SH1} --> {bitmap[1][0], bitmap[1][1]}
+			 *    SE2: {SH0,SH1} --> {bitmap[2][0], bitmap[2][1]}
+			 *    SE3: {SH0,SH1} --> {bitmap[3][0], bitmap[3][1]}
+			 *    SE4: {SH0,SH1} --> {bitmap[0][2], bitmap[0][3]}
+			 *    SE5: {SH0,SH1} --> {bitmap[1][2], bitmap[1][3]}
+			 *    SE6: {SH0,SH1} --> {bitmap[2][2], bitmap[2][3]}
+			 *    SE7: {SH0,SH1} --> {bitmap[3][2], bitmap[3][3]}
+			 */
+			cu_info->bitmap[0][i % 4][j + (i / 4) * 2] = bitmap;
+
+			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k++) {
+				if (bitmap & mask)
+					counter++;
+
+				mask <<= 1;
+			}
+			active_cu_number += counter;
+		}
+	}
+	gfx_v12_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	cu_info->number = active_cu_number;
+	cu_info->simd_per_cu = NUM_SIMD_PER_CU;
+
+	return 0;
+}
+
+const struct amdgpu_ip_block_version gfx_v12_0_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_GFX,
+	.major = 12,
+	.minor = 0,
+	.rev = 0,
+	.funcs = &gfx_v12_0_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.h b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.h
new file mode 100644
index 000000000000..f7184b2dc4e8
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.h
@@ -0,0 +1,31 @@
+/*
+ * Copyright 2023 dvanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __GFX_V12_0_H__
+#define __GFX_V12_0_H__
+
+extern const struct amdgpu_ip_block_version gfx_v12_0_ip_block;
+
+int gfx_v12_0_request_gfx_index_mutex(struct amdgpu_device *adev,
+				      bool req);
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c
index c41219e23151..80565392313f 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c
@@ -28,19 +28,33 @@
 #include "amdgpu_gfx.h"
 #include "amdgpu_ucode.h"
 #include "clearstate_si.h"
+#include "si.h"
+#include "sid.h"
+
 #include "bif/bif_3_0_d.h"
 #include "bif/bif_3_0_sh_mask.h"
+
 #include "oss/oss_1_0_d.h"
 #include "oss/oss_1_0_sh_mask.h"
+
 #include "gca/gfx_6_0_d.h"
 #include "gca/gfx_6_0_sh_mask.h"
+#include "gca/gfx_7_2_enum.h"
+
 #include "gmc/gmc_6_0_d.h"
 #include "gmc/gmc_6_0_sh_mask.h"
+
 #include "dce/dce_6_0_d.h"
 #include "dce/dce_6_0_sh_mask.h"
-#include "gca/gfx_7_2_enum.h"
+
 #include "si_enums.h"
-#include "si.h"
+
+#define TAHITI_GB_ADDR_CONFIG_GOLDEN        0x12011003
+#define VERDE_GB_ADDR_CONFIG_GOLDEN         0x12010002
+#define HAINAN_GB_ADDR_CONFIG_GOLDEN        0x02010001
+
+#define GFX6_NUM_GFX_RINGS     1
+#define GFX6_NUM_COMPUTE_RINGS 2
 
 static void gfx_v6_0_set_ring_funcs(struct amdgpu_device *adev);
 static void gfx_v6_0_set_irq_funcs(struct amdgpu_device *adev);
@@ -72,7 +86,7 @@ MODULE_FIRMWARE("amdgpu/hainan_ce.bin");
 MODULE_FIRMWARE("amdgpu/hainan_rlc.bin");
 
 static u32 gfx_v6_0_get_csb_size(struct amdgpu_device *adev);
-static void gfx_v6_0_get_csb_buffer(struct amdgpu_device *adev, volatile u32 *buffer);
+static void gfx_v6_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer);
 //static void gfx_v6_0_init_cp_pg_table(struct amdgpu_device *adev);
 static void gfx_v6_0_init_pg(struct amdgpu_device *adev);
 
@@ -311,7 +325,6 @@ static const u32 verde_rlc_save_restore_register_list[] =
 static int gfx_v6_0_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[30];
 	int err;
 	const struct gfx_firmware_header_v1_0 *cp_hdr;
 	const struct rlc_firmware_header_v1_0 *rlc_hdr;
@@ -337,32 +350,36 @@ static int gfx_v6_0_init_microcode(struct amdgpu_device *adev)
 	default: BUG();
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_pfp.bin", chip_name);
 	if (err)
 		goto out;
 	cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.pfp_fw->data;
 	adev->gfx.pfp_fw_version = le32_to_cpu(cp_hdr->header.ucode_version);
 	adev->gfx.pfp_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_me.bin", chip_name);
 	if (err)
 		goto out;
 	cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.me_fw->data;
 	adev->gfx.me_fw_version = le32_to_cpu(cp_hdr->header.ucode_version);
 	adev->gfx.me_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_ce.bin", chip_name);
 	if (err)
 		goto out;
 	cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data;
 	adev->gfx.ce_fw_version = le32_to_cpu(cp_hdr->header.ucode_version);
 	adev->gfx.ce_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_rlc.bin", chip_name);
 	if (err)
 		goto out;
 	rlc_hdr = (const struct rlc_firmware_header_v1_0 *)adev->gfx.rlc_fw->data;
@@ -371,7 +388,7 @@ static int gfx_v6_0_init_microcode(struct amdgpu_device *adev)
 
 out:
 	if (err) {
-		pr_err("gfx6: Failed to load firmware \"%s\"\n", fw_name);
+		pr_err("gfx6: Failed to load firmware %s gfx firmware\n", chip_name);
 		amdgpu_ucode_release(&adev->gfx.pfp_fw);
 		amdgpu_ucode_release(&adev->gfx.me_fw);
 		amdgpu_ucode_release(&adev->gfx.ce_fw);
@@ -1285,7 +1302,7 @@ static void gfx_v6_0_tiling_mode_table_init(struct amdgpu_device *adev)
 }
 
 static void gfx_v6_0_select_se_sh(struct amdgpu_device *adev, u32 se_num,
-				  u32 sh_num, u32 instance)
+				  u32 sh_num, u32 instance, int xcc_id)
 {
 	u32 data;
 
@@ -1438,12 +1455,12 @@ static void gfx_v6_0_write_harvested_raster_configs(struct amdgpu_device *adev,
 		}
 
 		/* GRBM_GFX_INDEX has a different offset on SI */
-		gfx_v6_0_select_se_sh(adev, se, 0xffffffff, 0xffffffff);
+		gfx_v6_0_select_se_sh(adev, se, 0xffffffff, 0xffffffff, 0);
 		WREG32(mmPA_SC_RASTER_CONFIG, raster_config_se);
 	}
 
 	/* GRBM_GFX_INDEX has a different offset on SI */
-	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 }
 
 static void gfx_v6_0_setup_rb(struct amdgpu_device *adev)
@@ -1459,14 +1476,14 @@ static void gfx_v6_0_setup_rb(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v6_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v6_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			data = gfx_v6_0_get_rb_active_bitmap(adev);
 			active_rbs |= data <<
 				((i * adev->gfx.config.max_sh_per_se + j) *
 				 rb_bitmap_width_per_sh);
 		}
 	}
-	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 
 	adev->gfx.config.backend_enable_mask = active_rbs;
 	adev->gfx.config.num_rbs = hweight32(active_rbs);
@@ -1487,7 +1504,7 @@ static void gfx_v6_0_setup_rb(struct amdgpu_device *adev)
 	/* cache the values for userspace */
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v6_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v6_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			adev->gfx.config.rb_config[i][j].rb_backend_disable =
 				RREG32(mmCC_RB_BACKEND_DISABLE);
 			adev->gfx.config.rb_config[i][j].user_rb_backend_disable =
@@ -1496,7 +1513,7 @@ static void gfx_v6_0_setup_rb(struct amdgpu_device *adev)
 				RREG32(mmPA_SC_RASTER_CONFIG);
 		}
 	}
-	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 }
 
@@ -1535,7 +1552,7 @@ static void gfx_v6_0_setup_spi(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v6_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v6_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			data = RREG32(mmSPI_STATIC_THREAD_MGMT_3);
 			active_cu = gfx_v6_0_get_cu_enabled(adev);
 
@@ -1550,7 +1567,7 @@ static void gfx_v6_0_setup_spi(struct amdgpu_device *adev)
 			}
 		}
 	}
-	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 }
 
@@ -1718,10 +1735,14 @@ static void gfx_v6_0_constants_init(struct amdgpu_device *adev)
 	gfx_v6_0_get_cu_info(adev);
 	gfx_v6_0_config_init(adev);
 
-	WREG32(mmCP_QUEUE_THRESHOLDS, ((0x16 << CP_QUEUE_THRESHOLDS__ROQ_IB1_START__SHIFT) |
-				       (0x2b << CP_QUEUE_THRESHOLDS__ROQ_IB2_START__SHIFT)));
-	WREG32(mmCP_MEQ_THRESHOLDS, (0x30 << CP_MEQ_THRESHOLDS__MEQ1_START__SHIFT) |
-				    (0x60 << CP_MEQ_THRESHOLDS__MEQ2_START__SHIFT));
+	WREG32(mmCP_QUEUE_THRESHOLDS,
+		((0x16 << CP_QUEUE_THRESHOLDS__ROQ_IB1_START__SHIFT) |
+		(0x2b << CP_QUEUE_THRESHOLDS__ROQ_IB2_START__SHIFT)));
+
+	/* set HW defaults for 3D engine */
+	WREG32(mmCP_MEQ_THRESHOLDS,
+		(0x30 << CP_MEQ_THRESHOLDS__MEQ1_START__SHIFT) |
+		(0x60 << CP_MEQ_THRESHOLDS__MEQ2_START__SHIFT));
 
 	sx_debug_1 = RREG32(mmSX_DEBUG_1);
 	WREG32(mmSX_DEBUG_1, sx_debug_1);
@@ -1907,7 +1928,7 @@ static int gfx_v6_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = -EINVAL;
 
 error:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 	return r;
 }
@@ -2333,7 +2354,7 @@ static void gfx_v6_0_ring_emit_wreg(struct amdgpu_ring *ring,
 static int gfx_v6_0_rlc_init(struct amdgpu_device *adev)
 {
 	const u32 *src_ptr;
-	volatile u32 *dst_ptr;
+	u32 *dst_ptr;
 	u32 dws;
 	u64 reg_list_mc_addr;
 	const struct cs_section_def *cs_data;
@@ -2391,7 +2412,7 @@ static void gfx_v6_0_enable_lbpw(struct amdgpu_device *adev, bool enable)
 	WREG32_FIELD(RLC_LB_CNTL, LOAD_BALANCE_ENABLE, enable ? 1 : 0);
 
 	if (!enable) {
-		gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+		gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 		WREG32(mmSPI_LB_CU_MASK, 0x00ff);
 	}
 }
@@ -2834,47 +2855,23 @@ static u32 gfx_v6_0_get_csb_size(struct amdgpu_device *adev)
 	return count;
 }
 
-static void gfx_v6_0_get_csb_buffer(struct amdgpu_device *adev,
-				    volatile u32 *buffer)
+static void gfx_v6_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer)
 {
-	u32 count = 0, i;
-	const struct cs_section_def *sect = NULL;
-	const struct cs_extent_def *ext = NULL;
+	u32 count = 0;
 
 	if (adev->gfx.rlc.cs_data == NULL)
 		return;
 	if (buffer == NULL)
 		return;
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1));
-	buffer[count++] = cpu_to_le32(0x80000000);
-	buffer[count++] = cpu_to_le32(0x80000000);
-
-	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
-		for (ext = sect->section; ext->extent != NULL; ++ext) {
-			if (sect->id == SECT_CONTEXT) {
-				buffer[count++] =
-					cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count));
-				buffer[count++] = cpu_to_le32(ext->reg_index - 0xa000);
-				for (i = 0; i < ext->reg_count; i++)
-					buffer[count++] = cpu_to_le32(ext->extent[i]);
-			} else {
-				return;
-			}
-		}
-	}
+	count = amdgpu_gfx_csb_preamble_start(buffer);
+	count = amdgpu_gfx_csb_data_parser(adev, buffer, count);
 
 	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, 1));
 	buffer[count++] = cpu_to_le32(mmPA_SC_RASTER_CONFIG - PACKET3_SET_CONTEXT_REG_START);
 	buffer[count++] = cpu_to_le32(adev->gfx.config.rb_config[0][0].raster_config);
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0));
-	buffer[count++] = cpu_to_le32(0);
+	amdgpu_gfx_csb_preamble_end(buffer, count);
 }
 
 static void gfx_v6_0_init_pg(struct amdgpu_device *adev)
@@ -2968,7 +2965,7 @@ static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd,
 		*(out++) = RREG32(mmSQ_IND_DATA);
 }
 
-static void gfx_v6_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
+static void gfx_v6_0_read_wave_data(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
 {
 	/* type 0 wave data */
 	dst[(*no_fields)++] = 0;
@@ -2993,7 +2990,7 @@ static void gfx_v6_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, u
 	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_MODE);
 }
 
-static void gfx_v6_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v6_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				     uint32_t wave, uint32_t start,
 				     uint32_t size, uint32_t *dst)
 {
@@ -3003,7 +3000,7 @@ static void gfx_v6_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
 }
 
 static void gfx_v6_0_select_me_pipe_q(struct amdgpu_device *adev,
-				  u32 me, u32 pipe, u32 q, u32 vm)
+				  u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id)
 {
 	DRM_INFO("Not implemented\n");
 }
@@ -3024,10 +3021,11 @@ static const struct amdgpu_rlc_funcs gfx_v6_0_rlc_funcs = {
 	.start = gfx_v6_0_rlc_start
 };
 
-static int gfx_v6_0_early_init(void *handle)
+static int gfx_v6_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
+	adev->gfx.xcc_mask = 1;
 	adev->gfx.num_gfx_rings = GFX6_NUM_GFX_RINGS;
 	adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
 					  GFX6_NUM_COMPUTE_RINGS);
@@ -3039,10 +3037,10 @@ static int gfx_v6_0_early_init(void *handle)
 	return 0;
 }
 
-static int gfx_v6_0_sw_init(void *handle)
+static int gfx_v6_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i, r;
 
 	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 181, &adev->gfx.eop_irq);
@@ -3073,7 +3071,7 @@ static int gfx_v6_0_sw_init(void *handle)
 		ring = &adev->gfx.gfx_ring[i];
 		ring->ring_obj = NULL;
 		sprintf(ring->name, "gfx");
-		r = amdgpu_ring_init(adev, ring, 1024,
+		r = amdgpu_ring_init(adev, ring, 2048,
 				     &adev->gfx.eop_irq,
 				     AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP,
 				     AMDGPU_RING_PRIO_DEFAULT, NULL);
@@ -3104,13 +3102,18 @@ static int gfx_v6_0_sw_init(void *handle)
 			return r;
 	}
 
+	adev->gfx.gfx_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]);
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+
 	return r;
 }
 
-static int gfx_v6_0_sw_fini(void *handle)
+static int gfx_v6_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
 		amdgpu_ring_fini(&adev->gfx.gfx_ring[i]);
@@ -3122,10 +3125,10 @@ static int gfx_v6_0_sw_fini(void *handle)
 	return 0;
 }
 
-static int gfx_v6_0_hw_init(void *handle)
+static int gfx_v6_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gfx_v6_0_constants_init(adev);
 
@@ -3142,9 +3145,9 @@ static int gfx_v6_0_hw_init(void *handle)
 	return r;
 }
 
-static int gfx_v6_0_hw_fini(void *handle)
+static int gfx_v6_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gfx_v6_0_cp_enable(adev, false);
 	adev->gfx.rlc.funcs->stop(adev);
@@ -3153,23 +3156,19 @@ static int gfx_v6_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int gfx_v6_0_suspend(void *handle)
+static int gfx_v6_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return gfx_v6_0_hw_fini(adev);
+	return gfx_v6_0_hw_fini(ip_block);
 }
 
-static int gfx_v6_0_resume(void *handle)
+static int gfx_v6_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return gfx_v6_0_hw_init(adev);
+	return gfx_v6_0_hw_init(ip_block);
 }
 
-static bool gfx_v6_0_is_idle(void *handle)
+static bool gfx_v6_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (RREG32(mmGRBM_STATUS) & GRBM_STATUS__GUI_ACTIVE_MASK)
 		return false;
@@ -3177,24 +3176,19 @@ static bool gfx_v6_0_is_idle(void *handle)
 		return true;
 }
 
-static int gfx_v6_0_wait_for_idle(void *handle)
+static int gfx_v6_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		if (gfx_v6_0_is_idle(handle))
+		if (gfx_v6_0_is_idle(ip_block))
 			return 0;
 		udelay(1);
 	}
 	return -ETIMEDOUT;
 }
 
-static int gfx_v6_0_soft_reset(void *handle)
-{
-	return 0;
-}
-
 static void gfx_v6_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev,
 						 enum amdgpu_interrupt_state state)
 {
@@ -3382,11 +3376,11 @@ static int gfx_v6_0_priv_inst_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int gfx_v6_0_set_clockgating_state(void *handle,
+static int gfx_v6_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	bool gate = false;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (state == AMD_CG_STATE_GATE)
 		gate = true;
@@ -3404,11 +3398,11 @@ static int gfx_v6_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int gfx_v6_0_set_powergating_state(void *handle,
+static int gfx_v6_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	bool gate = false;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (state == AMD_PG_STATE_GATE)
 		gate = true;
@@ -3444,7 +3438,6 @@ static void gfx_v6_0_emit_mem_sync(struct amdgpu_ring *ring)
 static const struct amd_ip_funcs gfx_v6_0_ip_funcs = {
 	.name = "gfx_v6_0",
 	.early_init = gfx_v6_0_early_init,
-	.late_init = NULL,
 	.sw_init = gfx_v6_0_sw_init,
 	.sw_fini = gfx_v6_0_sw_fini,
 	.hw_init = gfx_v6_0_hw_init,
@@ -3453,7 +3446,6 @@ static const struct amd_ip_funcs gfx_v6_0_ip_funcs = {
 	.resume = gfx_v6_0_resume,
 	.is_idle = gfx_v6_0_is_idle,
 	.wait_for_idle = gfx_v6_0_wait_for_idle,
-	.soft_reset = gfx_v6_0_soft_reset,
 	.set_clockgating_state = gfx_v6_0_set_clockgating_state,
 	.set_powergating_state = gfx_v6_0_set_powergating_state,
 };
@@ -3571,12 +3563,12 @@ static void gfx_v6_0_get_cu_info(struct amdgpu_device *adev)
 			mask = 1;
 			ao_bitmap = 0;
 			counter = 0;
-			gfx_v6_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v6_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			if (i < 4 && j < 2)
 				gfx_v6_0_set_user_cu_inactive_bitmap(
 					adev, disable_masks[i * 2 + j]);
 			bitmap = gfx_v6_0_get_cu_enabled(adev);
-			cu_info->bitmap[i][j] = bitmap;
+			cu_info->bitmap[0][i][j] = bitmap;
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k++) {
 				if (bitmap & mask) {
@@ -3593,7 +3585,7 @@ static void gfx_v6_0_get_cu_info(struct amdgpu_device *adev)
 		}
 	}
 
-	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	cu_info->number = active_cu_number;
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
index 9d5c1e29b4a3..2b7aba22ecc1 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
@@ -55,6 +55,9 @@
 #define GFX7_NUM_GFX_RINGS     1
 #define GFX7_MEC_HPD_SIZE      2048
 
+#define BONAIRE_GB_ADDR_CONFIG_GOLDEN        0x12010001
+#define HAWAII_GB_ADDR_CONFIG_GOLDEN         0x12011003
+
 static void gfx_v7_0_set_ring_funcs(struct amdgpu_device *adev);
 static void gfx_v7_0_set_irq_funcs(struct amdgpu_device *adev);
 static void gfx_v7_0_set_gds_init(struct amdgpu_device *adev);
@@ -90,8 +93,7 @@ MODULE_FIRMWARE("amdgpu/mullins_ce.bin");
 MODULE_FIRMWARE("amdgpu/mullins_rlc.bin");
 MODULE_FIRMWARE("amdgpu/mullins_mec.bin");
 
-static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] =
-{
+static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] = {
 	{mmGDS_VMID0_BASE, mmGDS_VMID0_SIZE, mmGDS_GWS_VMID0, mmGDS_OA_VMID0},
 	{mmGDS_VMID1_BASE, mmGDS_VMID1_SIZE, mmGDS_GWS_VMID1, mmGDS_OA_VMID1},
 	{mmGDS_VMID2_BASE, mmGDS_VMID2_SIZE, mmGDS_GWS_VMID2, mmGDS_OA_VMID2},
@@ -110,8 +112,7 @@ static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] =
 	{mmGDS_VMID15_BASE, mmGDS_VMID15_SIZE, mmGDS_GWS_VMID15, mmGDS_OA_VMID15}
 };
 
-static const u32 spectre_rlc_save_restore_register_list[] =
-{
+static const u32 spectre_rlc_save_restore_register_list[] = {
 	(0x0e00 << 16) | (0xc12c >> 2),
 	0x00000000,
 	(0x0e00 << 16) | (0xc140 >> 2),
@@ -557,8 +558,7 @@ static const u32 spectre_rlc_save_restore_register_list[] =
 	(0x0e00 << 16) | (0x9600 >> 2),
 };
 
-static const u32 kalindi_rlc_save_restore_register_list[] =
-{
+static const u32 kalindi_rlc_save_restore_register_list[] = {
 	(0x0e00 << 16) | (0xc12c >> 2),
 	0x00000000,
 	(0x0e00 << 16) | (0xc140 >> 2),
@@ -883,7 +883,7 @@ static const u32 kalindi_rlc_save_restore_register_list[] =
 };
 
 static u32 gfx_v7_0_get_csb_size(struct amdgpu_device *adev);
-static void gfx_v7_0_get_csb_buffer(struct amdgpu_device *adev, volatile u32 *buffer);
+static void gfx_v7_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer);
 static void gfx_v7_0_init_pg(struct amdgpu_device *adev);
 static void gfx_v7_0_get_cu_info(struct amdgpu_device *adev);
 
@@ -912,7 +912,6 @@ static void gfx_v7_0_free_microcode(struct amdgpu_device *adev)
 static int gfx_v7_0_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[30];
 	int err;
 
 	DRM_DEBUG("\n");
@@ -933,43 +932,48 @@ static int gfx_v7_0_init_microcode(struct amdgpu_device *adev)
 	case CHIP_MULLINS:
 		chip_name = "mullins";
 		break;
-	default: BUG();
+	default:
+		BUG();
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_pfp.bin", chip_name);
 	if (err)
 		goto out;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_me.bin", chip_name);
 	if (err)
 		goto out;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_ce.bin", chip_name);
 	if (err)
 		goto out;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_mec.bin", chip_name);
 	if (err)
 		goto out;
 
 	if (adev->asic_type == CHIP_KAVERI) {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_mec2.bin", chip_name);
 		if (err)
 			goto out;
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, fw_name);
-	if (err)
-		goto out;
+	err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_rlc.bin", chip_name);
 out:
 	if (err) {
-		pr_err("gfx7: Failed to load firmware \"%s\"\n", fw_name);
+		pr_err("gfx7: Failed to load firmware %s gfx firmware\n", chip_name);
 		gfx_v7_0_free_microcode(adev);
 	}
 	return err;
@@ -1548,11 +1552,12 @@ static void gfx_v7_0_tiling_mode_table_init(struct amdgpu_device *adev)
  * @sh_num: sh block to address
  * @instance: Certain registers are instanced per SE or SH.
  *            0xffffffff means broadcast to all SEs or SHs (CIK).
- *
+ * @xcc_id: xcc accelerated compute core id
  * Select which SE, SH combinations to address.
  */
 static void gfx_v7_0_select_se_sh(struct amdgpu_device *adev,
-				  u32 se_num, u32 sh_num, u32 instance)
+				  u32 se_num, u32 sh_num, u32 instance,
+				  int xcc_id)
 {
 	u32 data;
 
@@ -1732,13 +1737,13 @@ gfx_v7_0_write_harvested_raster_configs(struct amdgpu_device *adev,
 		}
 
 		/* GRBM_GFX_INDEX has a different offset on CI+ */
-		gfx_v7_0_select_se_sh(adev, se, 0xffffffff, 0xffffffff);
+		gfx_v7_0_select_se_sh(adev, se, 0xffffffff, 0xffffffff, 0);
 		WREG32(mmPA_SC_RASTER_CONFIG, raster_config_se);
 		WREG32(mmPA_SC_RASTER_CONFIG_1, raster_config_1);
 	}
 
 	/* GRBM_GFX_INDEX has a different offset on CI+ */
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 }
 
 /**
@@ -1761,13 +1766,13 @@ static void gfx_v7_0_setup_rb(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			data = gfx_v7_0_get_rb_active_bitmap(adev);
 			active_rbs |= data << ((i * adev->gfx.config.max_sh_per_se + j) *
 					       rb_bitmap_width_per_sh);
 		}
 	}
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 
 	adev->gfx.config.backend_enable_mask = active_rbs;
 	adev->gfx.config.num_rbs = hweight32(active_rbs);
@@ -1790,7 +1795,7 @@ static void gfx_v7_0_setup_rb(struct amdgpu_device *adev)
 	/* cache the values for userspace */
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			adev->gfx.config.rb_config[i][j].rb_backend_disable =
 				RREG32(mmCC_RB_BACKEND_DISABLE);
 			adev->gfx.config.rb_config[i][j].user_rb_backend_disable =
@@ -1801,7 +1806,7 @@ static void gfx_v7_0_setup_rb(struct amdgpu_device *adev)
 				RREG32(mmPA_SC_RASTER_CONFIG_1);
 		}
 	}
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 }
 
@@ -1911,7 +1916,7 @@ static void gfx_v7_0_constants_init(struct amdgpu_device *adev)
 	 * making sure that the following register writes will be broadcasted
 	 * to all the shaders
 	 */
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 
 	/* XXX SH_MEM regs */
 	/* where to put LDS, scratch, GPUVM in FSA64 space */
@@ -2118,6 +2123,8 @@ static void gfx_v7_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr,
 {
 	bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT;
 	bool int_sel = flags & AMDGPU_FENCE_FLAG_INT;
+	bool exec = flags & AMDGPU_FENCE_FLAG_EXEC;
+
 	/* Workaround for cache flush problems. First send a dummy EOP
 	 * event down the pipe with seq one below.
 	 */
@@ -2137,7 +2144,8 @@ static void gfx_v7_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr,
 	amdgpu_ring_write(ring, (EOP_TCL1_ACTION_EN |
 				 EOP_TC_ACTION_EN |
 				 EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) |
-				 EVENT_INDEX(5)));
+				 EVENT_INDEX(5) |
+				 (exec ? EOP_EXEC : 0)));
 	amdgpu_ring_write(ring, addr & 0xfffffffc);
 	amdgpu_ring_write(ring, (upper_32_bits(addr) & 0xffff) |
 				DATA_SEL(write64bit ? 2 : 1) | INT_SEL(int_sel ? 2 : 0));
@@ -2325,7 +2333,7 @@ static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = -EINVAL;
 
 error:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 	return r;
 }
@@ -2560,7 +2568,7 @@ static int gfx_v7_0_cp_gfx_resume(struct amdgpu_device *adev)
 	ring->wptr = 0;
 	WREG32(mmCP_RB0_WPTR, lower_32_bits(ring->wptr));
 
-	/* set the wb address wether it's enabled or not */
+	/* set the wb address whether it's enabled or not */
 	rptr_addr = ring->rptr_gpu_addr;
 	WREG32(mmCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr));
 	WREG32(mmCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & 0xFF);
@@ -2728,7 +2736,7 @@ static int gfx_v7_0_mec_init(struct amdgpu_device *adev)
 	u32 *hpd;
 	size_t mec_hpd_size;
 
-	bitmap_zero(adev->gfx.mec.queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
+	bitmap_zero(adev->gfx.mec_bitmap[0].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
 
 	/* take ownership of the relevant compute queues */
 	amdgpu_gfx_compute_queue_acquire(adev);
@@ -2758,45 +2766,6 @@ static int gfx_v7_0_mec_init(struct amdgpu_device *adev)
 	return 0;
 }
 
-struct hqd_registers
-{
-	u32 cp_mqd_base_addr;
-	u32 cp_mqd_base_addr_hi;
-	u32 cp_hqd_active;
-	u32 cp_hqd_vmid;
-	u32 cp_hqd_persistent_state;
-	u32 cp_hqd_pipe_priority;
-	u32 cp_hqd_queue_priority;
-	u32 cp_hqd_quantum;
-	u32 cp_hqd_pq_base;
-	u32 cp_hqd_pq_base_hi;
-	u32 cp_hqd_pq_rptr;
-	u32 cp_hqd_pq_rptr_report_addr;
-	u32 cp_hqd_pq_rptr_report_addr_hi;
-	u32 cp_hqd_pq_wptr_poll_addr;
-	u32 cp_hqd_pq_wptr_poll_addr_hi;
-	u32 cp_hqd_pq_doorbell_control;
-	u32 cp_hqd_pq_wptr;
-	u32 cp_hqd_pq_control;
-	u32 cp_hqd_ib_base_addr;
-	u32 cp_hqd_ib_base_addr_hi;
-	u32 cp_hqd_ib_rptr;
-	u32 cp_hqd_ib_control;
-	u32 cp_hqd_iq_timer;
-	u32 cp_hqd_iq_rptr;
-	u32 cp_hqd_dequeue_request;
-	u32 cp_hqd_dma_offload;
-	u32 cp_hqd_sema_cmd;
-	u32 cp_hqd_msg_type;
-	u32 cp_hqd_atomic0_preop_lo;
-	u32 cp_hqd_atomic0_preop_hi;
-	u32 cp_hqd_atomic1_preop_lo;
-	u32 cp_hqd_atomic1_preop_hi;
-	u32 cp_hqd_hq_scheduler0;
-	u32 cp_hqd_hq_scheduler1;
-	u32 cp_mqd_control;
-};
-
 static void gfx_v7_0_compute_pipe_init(struct amdgpu_device *adev,
 				       int mec, int pipe)
 {
@@ -2916,7 +2885,7 @@ static void gfx_v7_0_mqd_init(struct amdgpu_device *adev,
 	mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffffc;
 	mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff;
 
-	/* set the wb address wether it's enabled or not */
+	/* set the wb address whether it's enabled or not */
 	wb_gpu_addr = ring->rptr_gpu_addr;
 	mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc;
 	mqd->cp_hqd_pq_rptr_report_addr_hi =
@@ -3276,7 +3245,7 @@ static int gfx_v7_0_rlc_init(struct amdgpu_device *adev)
 
 	/* init spm vmid with 0xf */
 	if (adev->gfx.rlc.funcs->update_spm_vmid)
-		adev->gfx.rlc.funcs->update_spm_vmid(adev, 0xf);
+		adev->gfx.rlc.funcs->update_spm_vmid(adev, NULL, 0xf);
 
 	return 0;
 }
@@ -3301,7 +3270,7 @@ static void gfx_v7_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			for (k = 0; k < adev->usec_timeout; k++) {
 				if (RREG32(mmRLC_SERDES_CU_MASTER_BUSY) == 0)
 					break;
@@ -3309,7 +3278,7 @@ static void gfx_v7_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
 			}
 		}
 	}
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	mask = RLC_SERDES_NONCU_MASTER_BUSY__SE_MASTER_BUSY_MASK |
@@ -3361,7 +3330,7 @@ static bool gfx_v7_0_is_rlc_enabled(struct amdgpu_device *adev)
 	return true;
 }
 
-static void gfx_v7_0_set_safe_mode(struct amdgpu_device *adev)
+static void gfx_v7_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	u32 tmp, i, mask;
 
@@ -3383,7 +3352,7 @@ static void gfx_v7_0_set_safe_mode(struct amdgpu_device *adev)
 	}
 }
 
-static void gfx_v7_0_unset_safe_mode(struct amdgpu_device *adev)
+static void gfx_v7_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	u32 tmp;
 
@@ -3474,7 +3443,7 @@ static int gfx_v7_0_rlc_resume(struct amdgpu_device *adev)
 	WREG32(mmRLC_LB_CNTR_MAX, 0x00008000);
 
 	mutex_lock(&adev->grbm_idx_mutex);
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	WREG32(mmRLC_LB_INIT_CU_MASK, 0xffffffff);
 	WREG32(mmRLC_LB_PARAMS, 0x00600408);
 	WREG32(mmRLC_LB_CNTL, 0x80000004);
@@ -3502,7 +3471,7 @@ static int gfx_v7_0_rlc_resume(struct amdgpu_device *adev)
 	return 0;
 }
 
-static void gfx_v7_0_update_spm_vmid(struct amdgpu_device *adev, unsigned vmid)
+static void gfx_v7_0_update_spm_vmid(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned vmid)
 {
 	u32 data;
 
@@ -3530,7 +3499,7 @@ static void gfx_v7_0_enable_cgcg(struct amdgpu_device *adev, bool enable)
 		tmp = gfx_v7_0_halt_rlc(adev);
 
 		mutex_lock(&adev->grbm_idx_mutex);
-		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 		WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff);
 		WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff);
 		tmp2 = RLC_SERDES_WR_CTRL__BPM_ADDR_MASK |
@@ -3584,7 +3553,7 @@ static void gfx_v7_0_enable_mgcg(struct amdgpu_device *adev, bool enable)
 		tmp = gfx_v7_0_halt_rlc(adev);
 
 		mutex_lock(&adev->grbm_idx_mutex);
-		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 		WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff);
 		WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff);
 		data = RLC_SERDES_WR_CTRL__BPM_ADDR_MASK |
@@ -3635,7 +3604,7 @@ static void gfx_v7_0_enable_mgcg(struct amdgpu_device *adev, bool enable)
 		tmp = gfx_v7_0_halt_rlc(adev);
 
 		mutex_lock(&adev->grbm_idx_mutex);
-		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 		WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff);
 		WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff);
 		data = RLC_SERDES_WR_CTRL__BPM_ADDR_MASK | RLC_SERDES_WR_CTRL__MGCG_OVERRIDE_1_MASK;
@@ -3913,70 +3882,24 @@ static u32 gfx_v7_0_get_csb_size(struct amdgpu_device *adev)
 	return count;
 }
 
-static void gfx_v7_0_get_csb_buffer(struct amdgpu_device *adev,
-				    volatile u32 *buffer)
+static void gfx_v7_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer)
 {
-	u32 count = 0, i;
-	const struct cs_section_def *sect = NULL;
-	const struct cs_extent_def *ext = NULL;
+	u32 count = 0;
 
 	if (adev->gfx.rlc.cs_data == NULL)
 		return;
 	if (buffer == NULL)
 		return;
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1));
-	buffer[count++] = cpu_to_le32(0x80000000);
-	buffer[count++] = cpu_to_le32(0x80000000);
-
-	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
-		for (ext = sect->section; ext->extent != NULL; ++ext) {
-			if (sect->id == SECT_CONTEXT) {
-				buffer[count++] =
-					cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count));
-				buffer[count++] = cpu_to_le32(ext->reg_index - PACKET3_SET_CONTEXT_REG_START);
-				for (i = 0; i < ext->reg_count; i++)
-					buffer[count++] = cpu_to_le32(ext->extent[i]);
-			} else {
-				return;
-			}
-		}
-	}
+	count = amdgpu_gfx_csb_preamble_start(buffer);
+	count = amdgpu_gfx_csb_data_parser(adev, buffer, count);
 
 	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, 2));
 	buffer[count++] = cpu_to_le32(mmPA_SC_RASTER_CONFIG - PACKET3_SET_CONTEXT_REG_START);
-	switch (adev->asic_type) {
-	case CHIP_BONAIRE:
-		buffer[count++] = cpu_to_le32(0x16000012);
-		buffer[count++] = cpu_to_le32(0x00000000);
-		break;
-	case CHIP_KAVERI:
-		buffer[count++] = cpu_to_le32(0x00000000); /* XXX */
-		buffer[count++] = cpu_to_le32(0x00000000);
-		break;
-	case CHIP_KABINI:
-	case CHIP_MULLINS:
-		buffer[count++] = cpu_to_le32(0x00000000); /* XXX */
-		buffer[count++] = cpu_to_le32(0x00000000);
-		break;
-	case CHIP_HAWAII:
-		buffer[count++] = cpu_to_le32(0x3a00161a);
-		buffer[count++] = cpu_to_le32(0x0000002e);
-		break;
-	default:
-		buffer[count++] = cpu_to_le32(0x00000000);
-		buffer[count++] = cpu_to_le32(0x00000000);
-		break;
-	}
+	buffer[count++] = cpu_to_le32(adev->gfx.config.rb_config[0][0].raster_config);
+	buffer[count++] = cpu_to_le32(adev->gfx.config.rb_config[0][0].raster_config_1);
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0));
-	buffer[count++] = cpu_to_le32(0);
+	amdgpu_gfx_csb_preamble_end(buffer, count);
 }
 
 static void gfx_v7_0_init_pg(struct amdgpu_device *adev)
@@ -4111,7 +4034,7 @@ static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd,
 		*(out++) = RREG32(mmSQ_IND_DATA);
 }
 
-static void gfx_v7_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
+static void gfx_v7_0_read_wave_data(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
 {
 	/* type 0 wave data */
 	dst[(*no_fields)++] = 0;
@@ -4136,7 +4059,7 @@ static void gfx_v7_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, u
 	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_MODE);
 }
 
-static void gfx_v7_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v7_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				     uint32_t wave, uint32_t start,
 				     uint32_t size, uint32_t *dst)
 {
@@ -4146,7 +4069,7 @@ static void gfx_v7_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
 }
 
 static void gfx_v7_0_select_me_pipe_q(struct amdgpu_device *adev,
-				  u32 me, u32 pipe, u32 q, u32 vm)
+				  u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id)
 {
 	cik_srbm_select(adev, me, pipe, q, vm);
 }
@@ -4174,10 +4097,11 @@ static const struct amdgpu_rlc_funcs gfx_v7_0_rlc_funcs = {
 	.update_spm_vmid = gfx_v7_0_update_spm_vmid
 };
 
-static int gfx_v7_0_early_init(void *handle)
+static int gfx_v7_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
+	adev->gfx.xcc_mask = 1;
 	adev->gfx.num_gfx_rings = GFX7_NUM_GFX_RINGS;
 	adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
 					  AMDGPU_MAX_COMPUTE_RINGS);
@@ -4190,9 +4114,9 @@ static int gfx_v7_0_early_init(void *handle)
 	return 0;
 }
 
-static int gfx_v7_0_late_init(void *handle)
+static int gfx_v7_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_irq_get(adev, &adev->gfx.priv_reg_irq, 0);
@@ -4382,10 +4306,10 @@ static int gfx_v7_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
 	return 0;
 }
 
-static int gfx_v7_0_sw_init(void *handle)
+static int gfx_v7_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i, j, k, r, ring_id;
 
 	switch (adev->asic_type) {
@@ -4456,7 +4380,8 @@ static int gfx_v7_0_sw_init(void *handle)
 	for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
 		for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
 			for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
-				if (!amdgpu_gfx_is_mec_queue_enabled(adev, i, k, j))
+				if (!amdgpu_gfx_is_mec_queue_enabled(adev, 0, i,
+								     k, j))
 					continue;
 
 				r = gfx_v7_0_compute_ring_init(adev,
@@ -4474,12 +4399,17 @@ static int gfx_v7_0_sw_init(void *handle)
 
 	gfx_v7_0_gpu_early_init(adev);
 
+	adev->gfx.gfx_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]);
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+
 	return r;
 }
 
-static int gfx_v7_0_sw_fini(void *handle)
+static int gfx_v7_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
@@ -4503,10 +4433,10 @@ static int gfx_v7_0_sw_fini(void *handle)
 	return 0;
 }
 
-static int gfx_v7_0_hw_init(void *handle)
+static int gfx_v7_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gfx_v7_0_constants_init(adev);
 
@@ -4524,9 +4454,9 @@ static int gfx_v7_0_hw_init(void *handle)
 	return r;
 }
 
-static int gfx_v7_0_hw_fini(void *handle)
+static int gfx_v7_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
@@ -4537,23 +4467,19 @@ static int gfx_v7_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int gfx_v7_0_suspend(void *handle)
+static int gfx_v7_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return gfx_v7_0_hw_fini(adev);
+	return gfx_v7_0_hw_fini(ip_block);
 }
 
-static int gfx_v7_0_resume(void *handle)
+static int gfx_v7_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return gfx_v7_0_hw_init(adev);
+	return gfx_v7_0_hw_init(ip_block);
 }
 
-static bool gfx_v7_0_is_idle(void *handle)
+static bool gfx_v7_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (RREG32(mmGRBM_STATUS) & GRBM_STATUS__GUI_ACTIVE_MASK)
 		return false;
@@ -4561,11 +4487,11 @@ static bool gfx_v7_0_is_idle(void *handle)
 		return true;
 }
 
-static int gfx_v7_0_wait_for_idle(void *handle)
+static int gfx_v7_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		/* read MC_STATUS */
@@ -4578,11 +4504,11 @@ static int gfx_v7_0_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int gfx_v7_0_soft_reset(void *handle)
+static int gfx_v7_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 grbm_soft_reset = 0, srbm_soft_reset = 0;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* GRBM_STATUS */
 	tmp = RREG32(mmGRBM_STATUS);
@@ -4888,11 +4814,11 @@ static int gfx_v7_0_priv_inst_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int gfx_v7_0_set_clockgating_state(void *handle,
+static int gfx_v7_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	bool gate = false;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (state == AMD_CG_STATE_GATE)
 		gate = true;
@@ -4911,11 +4837,11 @@ static int gfx_v7_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int gfx_v7_0_set_powergating_state(void *handle,
+static int gfx_v7_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	bool gate = false;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (state == AMD_PG_STATE_GATE)
 		gate = true;
@@ -5041,6 +4967,7 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = {
 	.insert_nop = amdgpu_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_wreg = gfx_v7_0_ring_emit_wreg,
+	.soft_recovery = gfx_v7_0_ring_soft_recovery,
 	.emit_mem_sync = gfx_v7_0_emit_mem_sync_compute,
 };
 
@@ -5114,18 +5041,18 @@ static void gfx_v7_0_get_cu_info(struct amdgpu_device *adev)
 			mask = 1;
 			ao_bitmap = 0;
 			counter = 0;
-			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			if (i < 4 && j < 2)
 				gfx_v7_0_set_user_cu_inactive_bitmap(
 					adev, disable_masks[i * 2 + j]);
 			bitmap = gfx_v7_0_get_cu_active_bitmap(adev);
-			cu_info->bitmap[i][j] = bitmap;
+			cu_info->bitmap[0][i][j] = bitmap;
 
-			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k ++) {
+			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k++) {
 				if (bitmap & mask) {
 					if (counter < ao_cu_num)
 						ao_bitmap |= mask;
-					counter ++;
+					counter++;
 				}
 				mask <<= 1;
 			}
@@ -5135,7 +5062,7 @@ static void gfx_v7_0_get_cu_info(struct amdgpu_device *adev)
 			cu_info->ao_cu_bitmap[i][j] = ao_bitmap;
 		}
 	}
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	cu_info->number = active_cu_number;
@@ -5147,8 +5074,7 @@ static void gfx_v7_0_get_cu_info(struct amdgpu_device *adev)
 	cu_info->lds_size = 64;
 }
 
-const struct amdgpu_ip_block_version gfx_v7_1_ip_block =
-{
+const struct amdgpu_ip_block_version gfx_v7_1_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GFX,
 	.major = 7,
 	.minor = 1,
@@ -5156,8 +5082,7 @@ const struct amdgpu_ip_block_version gfx_v7_1_ip_block =
 	.funcs = &gfx_v7_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version gfx_v7_2_ip_block =
-{
+const struct amdgpu_ip_block_version gfx_v7_2_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GFX,
 	.major = 7,
 	.minor = 2,
@@ -5165,8 +5090,7 @@ const struct amdgpu_ip_block_version gfx_v7_2_ip_block =
 	.funcs = &gfx_v7_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version gfx_v7_3_ip_block =
-{
+const struct amdgpu_ip_block_version gfx_v7_3_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GFX,
 	.major = 7,
 	.minor = 3,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
index b1f2684d854a..1c87375e1dd5 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
@@ -883,8 +883,8 @@ static int gfx_v8_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	gpu_addr = adev->wb.gpu_addr + (index * 4);
 	adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD);
 	memset(&ib, 0, sizeof(ib));
-	r = amdgpu_ib_get(adev, NULL, 16,
-					AMDGPU_IB_POOL_DIRECT, &ib);
+
+	r = amdgpu_ib_get(adev, NULL, 20, AMDGPU_IB_POOL_DIRECT, &ib);
 	if (r)
 		goto err1;
 
@@ -914,7 +914,7 @@ static int gfx_v8_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = -EINVAL;
 
 err2:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err1:
 	amdgpu_device_wb_free(adev, index);
@@ -939,7 +939,6 @@ static void gfx_v8_0_free_microcode(struct amdgpu_device *adev)
 static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[30];
 	int err;
 	struct amdgpu_firmware_info *info = NULL;
 	const struct common_firmware_header *header = NULL;
@@ -982,15 +981,18 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 	}
 
 	if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp_2.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw,
+					   AMDGPU_UCODE_OPTIONAL,
+					   "amdgpu/%s_pfp_2.bin", chip_name);
 		if (err == -ENODEV) {
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name);
-			err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_pfp.bin", chip_name);
 		}
 	} else {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_pfp.bin", chip_name);
 	}
 	if (err)
 		goto out;
@@ -999,15 +1001,18 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 	adev->gfx.pfp_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
 	if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me_2.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.me_fw,
+					   AMDGPU_UCODE_OPTIONAL,
+					   "amdgpu/%s_me_2.bin", chip_name);
 		if (err == -ENODEV) {
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name);
-			err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->gfx.me_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_me.bin", chip_name);
 		}
 	} else {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.me_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_me.bin", chip_name);
 	}
 	if (err)
 		goto out;
@@ -1017,15 +1022,18 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 	adev->gfx.me_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
 	if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce_2.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw,
+					   AMDGPU_UCODE_OPTIONAL,
+					   "amdgpu/%s_ce_2.bin", chip_name);
 		if (err == -ENODEV) {
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name);
-			err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_ce.bin", chip_name);
 		}
 	} else {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_ce.bin", chip_name);
 	}
 	if (err)
 		goto out;
@@ -1044,8 +1052,9 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 	} else
 		adev->virt.chained_ib_support = false;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_rlc.bin", chip_name);
 	if (err)
 		goto out;
 	rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
@@ -1093,15 +1102,18 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 		adev->gfx.rlc.register_restore[i] = le32_to_cpu(tmp[i]);
 
 	if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec_2.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+					   AMDGPU_UCODE_OPTIONAL,
+					   "amdgpu/%s_mec_2.bin", chip_name);
 		if (err == -ENODEV) {
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name);
-			err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_mec.bin", chip_name);
 		}
 	} else {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_mec.bin", chip_name);
 	}
 	if (err)
 		goto out;
@@ -1112,15 +1124,18 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 	if ((adev->asic_type != CHIP_STONEY) &&
 	    (adev->asic_type != CHIP_TOPAZ)) {
 		if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2_2.bin", chip_name);
-			err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw,
+						   AMDGPU_UCODE_OPTIONAL,
+						   "amdgpu/%s_mec2_2.bin", chip_name);
 			if (err == -ENODEV) {
-				snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name);
-				err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, fw_name);
+				err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw,
+							   AMDGPU_UCODE_REQUIRED,
+							   "amdgpu/%s_mec2.bin", chip_name);
 			}
 		} else {
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name);
-			err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_mec2.bin", chip_name);
 		}
 		if (!err) {
 			cp_hdr = (const struct gfx_firmware_header_v1_0 *)
@@ -1194,9 +1209,7 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 
 out:
 	if (err) {
-		dev_err(adev->dev,
-			"gfx8: Failed to load firmware \"%s\"\n",
-			fw_name);
+		dev_err(adev->dev, "gfx8: Failed to load firmware %s gfx firmware\n", chip_name);
 		amdgpu_ucode_release(&adev->gfx.pfp_fw);
 		amdgpu_ucode_release(&adev->gfx.me_fw);
 		amdgpu_ucode_release(&adev->gfx.ce_fw);
@@ -1207,51 +1220,24 @@ out:
 	return err;
 }
 
-static void gfx_v8_0_get_csb_buffer(struct amdgpu_device *adev,
-				    volatile u32 *buffer)
+static void gfx_v8_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer)
 {
-	u32 count = 0, i;
-	const struct cs_section_def *sect = NULL;
-	const struct cs_extent_def *ext = NULL;
+	u32 count = 0;
 
 	if (adev->gfx.rlc.cs_data == NULL)
 		return;
 	if (buffer == NULL)
 		return;
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1));
-	buffer[count++] = cpu_to_le32(0x80000000);
-	buffer[count++] = cpu_to_le32(0x80000000);
-
-	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
-		for (ext = sect->section; ext->extent != NULL; ++ext) {
-			if (sect->id == SECT_CONTEXT) {
-				buffer[count++] =
-					cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count));
-				buffer[count++] = cpu_to_le32(ext->reg_index -
-						PACKET3_SET_CONTEXT_REG_START);
-				for (i = 0; i < ext->reg_count; i++)
-					buffer[count++] = cpu_to_le32(ext->extent[i]);
-			} else {
-				return;
-			}
-		}
-	}
+	count = amdgpu_gfx_csb_preamble_start(buffer);
+	count = amdgpu_gfx_csb_data_parser(adev, buffer, count);
 
 	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, 2));
-	buffer[count++] = cpu_to_le32(mmPA_SC_RASTER_CONFIG -
-			PACKET3_SET_CONTEXT_REG_START);
+	buffer[count++] = cpu_to_le32(mmPA_SC_RASTER_CONFIG - PACKET3_SET_CONTEXT_REG_START);
 	buffer[count++] = cpu_to_le32(adev->gfx.config.rb_config[0][0].raster_config);
 	buffer[count++] = cpu_to_le32(adev->gfx.config.rb_config[0][0].raster_config_1);
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0));
-	buffer[count++] = cpu_to_le32(0);
+	amdgpu_gfx_csb_preamble_end(buffer, count);
 }
 
 static int gfx_v8_0_cp_jump_table_num(struct amdgpu_device *adev)
@@ -1288,7 +1274,7 @@ static int gfx_v8_0_rlc_init(struct amdgpu_device *adev)
 
 	/* init spm vmid with 0xf */
 	if (adev->gfx.rlc.funcs->update_spm_vmid)
-		adev->gfx.rlc.funcs->update_spm_vmid(adev, 0xf);
+		adev->gfx.rlc.funcs->update_spm_vmid(adev, NULL, 0xf);
 
 	return 0;
 }
@@ -1304,7 +1290,7 @@ static int gfx_v8_0_mec_init(struct amdgpu_device *adev)
 	u32 *hpd;
 	size_t mec_hpd_size;
 
-	bitmap_zero(adev->gfx.mec.queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
+	bitmap_zero(adev->gfx.mec_bitmap[0].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
 
 	/* take ownership of the relevant compute queues */
 	amdgpu_gfx_compute_queue_acquire(adev);
@@ -1643,7 +1629,7 @@ static int gfx_v8_0_do_edc_gpr_workarounds(struct amdgpu_device *adev)
 		RREG32(sec_ded_counter_registers[i]);
 
 fail:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 
 	return r;
@@ -1897,12 +1883,12 @@ static int gfx_v8_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
 
 static void gfx_v8_0_sq_irq_work_func(struct work_struct *work);
 
-static int gfx_v8_0_sw_init(void *handle)
+static int gfx_v8_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int i, j, k, r, ring_id;
+	int xcc_id = 0;
 	struct amdgpu_ring *ring;
-	struct amdgpu_kiq *kiq;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	switch (adev->asic_type) {
 	case CHIP_TONGA:
@@ -2001,7 +1987,8 @@ static int gfx_v8_0_sw_init(void *handle)
 	for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
 		for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
 			for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
-				if (!amdgpu_gfx_is_mec_queue_enabled(adev, i, k, j))
+				if (!amdgpu_gfx_is_mec_queue_enabled(adev, 0, i,
+								     k, j))
 					continue;
 
 				r = gfx_v8_0_compute_ring_init(adev,
@@ -2015,19 +2002,18 @@ static int gfx_v8_0_sw_init(void *handle)
 		}
 	}
 
-	r = amdgpu_gfx_kiq_init(adev, GFX8_MEC_HPD_SIZE);
+	r = amdgpu_gfx_kiq_init(adev, GFX8_MEC_HPD_SIZE, 0);
 	if (r) {
 		DRM_ERROR("Failed to init KIQ BOs!\n");
 		return r;
 	}
 
-	kiq = &adev->gfx.kiq;
-	r = amdgpu_gfx_kiq_init_ring(adev, &kiq->ring, &kiq->irq);
+	r = amdgpu_gfx_kiq_init_ring(adev, xcc_id);
 	if (r)
 		return r;
 
 	/* create MQD for all compute queues as well as KIQ for SRIOV case */
-	r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct vi_mqd_allocation));
+	r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct vi_mqd_allocation), 0);
 	if (r)
 		return r;
 
@@ -2037,12 +2023,17 @@ static int gfx_v8_0_sw_init(void *handle)
 	if (r)
 		return r;
 
+	adev->gfx.gfx_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]);
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+
 	return 0;
 }
 
-static int gfx_v8_0_sw_fini(void *handle)
+static int gfx_v8_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
@@ -2050,9 +2041,9 @@ static int gfx_v8_0_sw_fini(void *handle)
 	for (i = 0; i < adev->gfx.num_compute_rings; i++)
 		amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
 
-	amdgpu_gfx_mqd_sw_fini(adev);
-	amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring);
-	amdgpu_gfx_kiq_fini(adev);
+	amdgpu_gfx_mqd_sw_fini(adev, 0);
+	amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq[0].ring);
+	amdgpu_gfx_kiq_fini(adev, 0);
 
 	gfx_v8_0_mec_fini(adev);
 	amdgpu_gfx_rlc_fini(adev);
@@ -3394,7 +3385,8 @@ static void gfx_v8_0_tiling_mode_table_init(struct amdgpu_device *adev)
 }
 
 static void gfx_v8_0_select_se_sh(struct amdgpu_device *adev,
-				  u32 se_num, u32 sh_num, u32 instance)
+				  u32 se_num, u32 sh_num, u32 instance,
+				  int xcc_id)
 {
 	u32 data;
 
@@ -3417,7 +3409,7 @@ static void gfx_v8_0_select_se_sh(struct amdgpu_device *adev,
 }
 
 static void gfx_v8_0_select_me_pipe_q(struct amdgpu_device *adev,
-				  u32 me, u32 pipe, u32 q, u32 vm)
+				  u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id)
 {
 	vi_srbm_select(adev, me, pipe, q, vm);
 }
@@ -3578,13 +3570,13 @@ gfx_v8_0_write_harvested_raster_configs(struct amdgpu_device *adev,
 		}
 
 		/* GRBM_GFX_INDEX has a different offset on VI */
-		gfx_v8_0_select_se_sh(adev, se, 0xffffffff, 0xffffffff);
+		gfx_v8_0_select_se_sh(adev, se, 0xffffffff, 0xffffffff, 0);
 		WREG32(mmPA_SC_RASTER_CONFIG, raster_config_se);
 		WREG32(mmPA_SC_RASTER_CONFIG_1, raster_config_1);
 	}
 
 	/* GRBM_GFX_INDEX has a different offset on VI */
-	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 }
 
 static void gfx_v8_0_setup_rb(struct amdgpu_device *adev)
@@ -3600,13 +3592,13 @@ static void gfx_v8_0_setup_rb(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			data = gfx_v8_0_get_rb_active_bitmap(adev);
 			active_rbs |= data << ((i * adev->gfx.config.max_sh_per_se + j) *
 					       rb_bitmap_width_per_sh);
 		}
 	}
-	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 
 	adev->gfx.config.backend_enable_mask = active_rbs;
 	adev->gfx.config.num_rbs = hweight32(active_rbs);
@@ -3629,7 +3621,7 @@ static void gfx_v8_0_setup_rb(struct amdgpu_device *adev)
 	/* cache the values for userspace */
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			adev->gfx.config.rb_config[i][j].rb_backend_disable =
 				RREG32(mmCC_RB_BACKEND_DISABLE);
 			adev->gfx.config.rb_config[i][j].user_rb_backend_disable =
@@ -3640,7 +3632,7 @@ static void gfx_v8_0_setup_rb(struct amdgpu_device *adev)
 				RREG32(mmPA_SC_RASTER_CONFIG_1);
 		}
 	}
-	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 }
 
@@ -3787,7 +3779,7 @@ static void gfx_v8_0_constants_init(struct amdgpu_device *adev)
 	 * making sure that the following register writes will be broadcasted
 	 * to all the shaders
 	 */
-	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 
 	WREG32(mmPA_SC_FIFO_SIZE,
 		   (adev->gfx.config.sc_prim_fifo_size_frontend <<
@@ -3818,7 +3810,7 @@ static void gfx_v8_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			for (k = 0; k < adev->usec_timeout; k++) {
 				if (RREG32(mmRLC_SERDES_CU_MASTER_BUSY) == 0)
 					break;
@@ -3826,7 +3818,7 @@ static void gfx_v8_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
 			}
 			if (k == adev->usec_timeout) {
 				gfx_v8_0_select_se_sh(adev, 0xffffffff,
-						      0xffffffff, 0xffffffff);
+						      0xffffffff, 0xffffffff, 0);
 				mutex_unlock(&adev->grbm_idx_mutex);
 				DRM_INFO("Timeout wait for RLC serdes %u,%u\n",
 					 i, j);
@@ -3834,7 +3826,7 @@ static void gfx_v8_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
 			}
 		}
 	}
-	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	mask = RLC_SERDES_NONCU_MASTER_BUSY__SE_MASTER_BUSY_MASK |
@@ -4262,7 +4254,7 @@ static int gfx_v8_0_cp_gfx_resume(struct amdgpu_device *adev)
 	ring->wptr = 0;
 	WREG32(mmCP_RB0_WPTR, lower_32_bits(ring->wptr));
 
-	/* set the wb address wether it's enabled or not */
+	/* set the wb address whether it's enabled or not */
 	rptr_addr = ring->rptr_gpu_addr;
 	WREG32(mmCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr));
 	WREG32(mmCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & 0xFF);
@@ -4281,7 +4273,6 @@ static int gfx_v8_0_cp_gfx_resume(struct amdgpu_device *adev)
 	/* start the ring */
 	amdgpu_ring_clear_ring(ring);
 	gfx_v8_0_cp_gfx_start(adev);
-	ring->sched.ready = true;
 
 	return 0;
 }
@@ -4292,7 +4283,7 @@ static void gfx_v8_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
 		WREG32(mmCP_MEC_CNTL, 0);
 	} else {
 		WREG32(mmCP_MEC_CNTL, (CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK));
-		adev->gfx.kiq.ring.sched.ready = false;
+		adev->gfx.kiq[0].ring.sched.ready = false;
 	}
 	udelay(50);
 }
@@ -4307,19 +4298,17 @@ static void gfx_v8_0_kiq_setting(struct amdgpu_ring *ring)
 	tmp = RREG32(mmRLC_CP_SCHEDULERS);
 	tmp &= 0xffffff00;
 	tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
-	WREG32(mmRLC_CP_SCHEDULERS, tmp);
-	tmp |= 0x80;
-	WREG32(mmRLC_CP_SCHEDULERS, tmp);
+	WREG32(mmRLC_CP_SCHEDULERS, tmp | 0x80);
 }
 
 static int gfx_v8_0_kiq_kcq_enable(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
+	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq[0].ring;
 	uint64_t queue_mask = 0;
 	int r, i;
 
 	for (i = 0; i < AMDGPU_MAX_COMPUTE_QUEUES; ++i) {
-		if (!test_bit(i, adev->gfx.mec.queue_bitmap))
+		if (!test_bit(i, adev->gfx.mec_bitmap[0].queue_bitmap))
 			continue;
 
 		/* This situation may be hit in the future if a new HW
@@ -4595,14 +4584,13 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct vi_mqd *mqd = ring->mqd_ptr;
-	int mqd_idx = AMDGPU_MAX_COMPUTE_RINGS;
 
 	gfx_v8_0_kiq_setting(ring);
 
 	if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */
 		/* reset MQD to a clean status */
-		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct vi_mqd_allocation));
+		if (adev->gfx.kiq[0].mqd_backup)
+			memcpy(mqd, adev->gfx.kiq[0].mqd_backup, sizeof(struct vi_mqd_allocation));
 
 		/* reset ring buffer */
 		ring->wptr = 0;
@@ -4625,8 +4613,8 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring)
 		vi_srbm_select(adev, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 
-		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct vi_mqd_allocation));
+		if (adev->gfx.kiq[0].mqd_backup)
+			memcpy(adev->gfx.kiq[0].mqd_backup, mqd, sizeof(struct vi_mqd_allocation));
 	}
 
 	return 0;
@@ -4650,14 +4638,13 @@ static int gfx_v8_0_kcq_init_queue(struct amdgpu_ring *ring)
 
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
 			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct vi_mqd_allocation));
-	} else if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */
-		/* reset MQD to a clean status */
+	} else {
+		/* restore MQD to a clean status */
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
 			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct vi_mqd_allocation));
 		/* reset ring buffer */
 		ring->wptr = 0;
-		amdgpu_ring_clear_ring(ring);
-	} else {
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, 0);
 		amdgpu_ring_clear_ring(ring);
 	}
 	return 0;
@@ -4675,59 +4662,25 @@ static void gfx_v8_0_set_mec_doorbell_range(struct amdgpu_device *adev)
 
 static int gfx_v8_0_kiq_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring;
-	int r;
-
-	ring = &adev->gfx.kiq.ring;
-
-	r = amdgpu_bo_reserve(ring->mqd_obj, false);
-	if (unlikely(r != 0))
-		return r;
-
-	r = amdgpu_bo_kmap(ring->mqd_obj, &ring->mqd_ptr);
-	if (unlikely(r != 0))
-		return r;
-
-	gfx_v8_0_kiq_init_queue(ring);
-	amdgpu_bo_kunmap(ring->mqd_obj);
-	ring->mqd_ptr = NULL;
-	amdgpu_bo_unreserve(ring->mqd_obj);
-	ring->sched.ready = true;
+	gfx_v8_0_kiq_init_queue(&adev->gfx.kiq[0].ring);
 	return 0;
 }
 
 static int gfx_v8_0_kcq_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring = NULL;
-	int r = 0, i;
+	int i, r;
 
 	gfx_v8_0_cp_compute_enable(adev, true);
 
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-		ring = &adev->gfx.compute_ring[i];
-
-		r = amdgpu_bo_reserve(ring->mqd_obj, false);
-		if (unlikely(r != 0))
-			goto done;
-		r = amdgpu_bo_kmap(ring->mqd_obj, &ring->mqd_ptr);
-		if (!r) {
-			r = gfx_v8_0_kcq_init_queue(ring);
-			amdgpu_bo_kunmap(ring->mqd_obj);
-			ring->mqd_ptr = NULL;
-		}
-		amdgpu_bo_unreserve(ring->mqd_obj);
+		r = gfx_v8_0_kcq_init_queue(&adev->gfx.compute_ring[i]);
 		if (r)
-			goto done;
+			return r;
 	}
 
 	gfx_v8_0_set_mec_doorbell_range(adev);
 
-	r = gfx_v8_0_kiq_kcq_enable(adev);
-	if (r)
-		goto done;
-
-done:
-	return r;
+	return gfx_v8_0_kiq_kcq_enable(adev);
 }
 
 static int gfx_v8_0_cp_test_all_rings(struct amdgpu_device *adev)
@@ -4741,7 +4694,7 @@ static int gfx_v8_0_cp_test_all_rings(struct amdgpu_device *adev)
 	if (r)
 		return r;
 
-	ring = &adev->gfx.kiq.ring;
+	ring = &adev->gfx.kiq[0].ring;
 	r = amdgpu_ring_test_helper(ring);
 	if (r)
 		return r;
@@ -4788,10 +4741,10 @@ static void gfx_v8_0_cp_enable(struct amdgpu_device *adev, bool enable)
 	gfx_v8_0_cp_compute_enable(adev, enable);
 }
 
-static int gfx_v8_0_hw_init(void *handle)
+static int gfx_v8_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gfx_v8_0_init_golden_registers(adev);
 	gfx_v8_0_constants_init(adev);
@@ -4808,7 +4761,7 @@ static int gfx_v8_0_hw_init(void *handle)
 static int gfx_v8_0_kcq_disable(struct amdgpu_device *adev)
 {
 	int r, i;
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
+	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq[0].ring;
 
 	r = amdgpu_ring_alloc(kiq_ring, 6 * adev->gfx.num_compute_rings);
 	if (r)
@@ -4828,6 +4781,13 @@ static int gfx_v8_0_kcq_disable(struct amdgpu_device *adev)
 		amdgpu_ring_write(kiq_ring, 0);
 		amdgpu_ring_write(kiq_ring, 0);
 	}
+	/* Submit unmap queue packet */
+	amdgpu_ring_commit(kiq_ring);
+	/*
+	 * Ring test will do a basic scratch register change check. Just run
+	 * this to ensure that unmap queues that is submitted before got
+	 * processed successfully before returning.
+	 */
 	r = amdgpu_ring_test_helper(kiq_ring);
 	if (r)
 		DRM_ERROR("KCQ disable failed\n");
@@ -4835,9 +4795,9 @@ static int gfx_v8_0_kcq_disable(struct amdgpu_device *adev)
 	return r;
 }
 
-static bool gfx_v8_0_is_idle(void *handle)
+static bool gfx_v8_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (REG_GET_FIELD(RREG32(mmGRBM_STATUS), GRBM_STATUS, GUI_ACTIVE)
 		|| RREG32(mmGRBM_STATUS2) != 0x8)
@@ -4870,13 +4830,13 @@ static int gfx_v8_0_wait_for_rlc_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int gfx_v8_0_wait_for_idle(void *handle)
+static int gfx_v8_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		if (gfx_v8_0_is_idle(handle))
+		if (gfx_v8_0_is_idle(ip_block))
 			return 0;
 
 		udelay(1);
@@ -4884,9 +4844,9 @@ static int gfx_v8_0_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int gfx_v8_0_hw_fini(void *handle)
+static int gfx_v8_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
@@ -4902,8 +4862,9 @@ static int gfx_v8_0_hw_fini(void *handle)
 		pr_debug("For SRIOV client, shouldn't do anything.\n");
 		return 0;
 	}
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
-	if (!gfx_v8_0_wait_for_idle(adev))
+
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
+	if (!gfx_v8_0_wait_for_idle(ip_block))
 		gfx_v8_0_cp_enable(adev, false);
 	else
 		pr_err("cp is busy, skip halt cp\n");
@@ -4911,24 +4872,24 @@ static int gfx_v8_0_hw_fini(void *handle)
 		adev->gfx.rlc.funcs->stop(adev);
 	else
 		pr_err("rlc is busy, skip halt rlc\n");
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 
 	return 0;
 }
 
-static int gfx_v8_0_suspend(void *handle)
+static int gfx_v8_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	return gfx_v8_0_hw_fini(handle);
+	return gfx_v8_0_hw_fini(ip_block);
 }
 
-static int gfx_v8_0_resume(void *handle)
+static int gfx_v8_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	return gfx_v8_0_hw_init(handle);
+	return gfx_v8_0_hw_init(ip_block);
 }
 
-static bool gfx_v8_0_check_soft_reset(void *handle)
+static bool gfx_v8_0_check_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 grbm_soft_reset = 0, srbm_soft_reset = 0;
 	u32 tmp;
 
@@ -4988,9 +4949,9 @@ static bool gfx_v8_0_check_soft_reset(void *handle)
 	}
 }
 
-static int gfx_v8_0_pre_soft_reset(void *handle)
+static int gfx_v8_0_pre_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 grbm_soft_reset = 0;
 
 	if ((!adev->gfx.grbm_soft_reset) &&
@@ -5029,9 +4990,9 @@ static int gfx_v8_0_pre_soft_reset(void *handle)
 	return 0;
 }
 
-static int gfx_v8_0_soft_reset(void *handle)
+static int gfx_v8_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 grbm_soft_reset = 0, srbm_soft_reset = 0;
 	u32 tmp;
 
@@ -5091,9 +5052,9 @@ static int gfx_v8_0_soft_reset(void *handle)
 	return 0;
 }
 
-static int gfx_v8_0_post_soft_reset(void *handle)
+static int gfx_v8_0_post_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 grbm_soft_reset = 0;
 
 	if ((!adev->gfx.grbm_soft_reset) &&
@@ -5216,7 +5177,7 @@ static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd,
 		*(out++) = RREG32(mmSQ_IND_DATA);
 }
 
-static void gfx_v8_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
+static void gfx_v8_0_read_wave_data(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
 {
 	/* type 0 wave data */
 	dst[(*no_fields)++] = 0;
@@ -5241,7 +5202,7 @@ static void gfx_v8_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, u
 	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_MODE);
 }
 
-static void gfx_v8_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v8_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				     uint32_t wave, uint32_t start,
 				     uint32_t size, uint32_t *dst)
 {
@@ -5259,10 +5220,11 @@ static const struct amdgpu_gfx_funcs gfx_v8_0_gfx_funcs = {
 	.select_me_pipe_q = &gfx_v8_0_select_me_pipe_q
 };
 
-static int gfx_v8_0_early_init(void *handle)
+static int gfx_v8_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
+	adev->gfx.xcc_mask = 1;
 	adev->gfx.num_gfx_rings = GFX8_NUM_GFX_RINGS;
 	adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
 					  AMDGPU_MAX_COMPUTE_RINGS);
@@ -5275,9 +5237,9 @@ static int gfx_v8_0_early_init(void *handle)
 	return 0;
 }
 
-static int gfx_v8_0_late_init(void *handle)
+static int gfx_v8_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_irq_get(adev, &adev->gfx.priv_reg_irq, 0);
@@ -5317,7 +5279,7 @@ static void gfx_v8_0_enable_gfx_static_mg_power_gating(struct amdgpu_device *ade
 	    (adev->asic_type == CHIP_POLARIS12) ||
 	    (adev->asic_type == CHIP_VEGAM))
 		/* Send msg to SMU via Powerplay */
-		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, enable);
+		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, enable, 0);
 
 	WREG32_FIELD(RLC_PG_CNTL, STATIC_PER_CU_PG_ENABLE, enable ? 1 : 0);
 }
@@ -5363,10 +5325,10 @@ static void cz_update_gfx_cg_power_gating(struct amdgpu_device *adev,
 	}
 }
 
-static int gfx_v8_0_set_powergating_state(void *handle,
+static int gfx_v8_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_PG_STATE_GATE);
 
 	if (amdgpu_sriov_vf(adev))
@@ -5376,7 +5338,7 @@ static int gfx_v8_0_set_powergating_state(void *handle,
 				AMD_PG_SUPPORT_RLC_SMU_HS |
 				AMD_PG_SUPPORT_CP |
 				AMD_PG_SUPPORT_GFX_DMG))
-		amdgpu_gfx_rlc_enter_safe_mode(adev);
+		amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 	switch (adev->asic_type) {
 	case CHIP_CARRIZO:
 	case CHIP_STONEY:
@@ -5430,13 +5392,13 @@ static int gfx_v8_0_set_powergating_state(void *handle,
 				AMD_PG_SUPPORT_RLC_SMU_HS |
 				AMD_PG_SUPPORT_CP |
 				AMD_PG_SUPPORT_GFX_DMG))
-		amdgpu_gfx_rlc_exit_safe_mode(adev);
+		amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 	return 0;
 }
 
-static void gfx_v8_0_get_clockgating_state(void *handle, u64 *flags)
+static void gfx_v8_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	if (amdgpu_sriov_vf(adev))
@@ -5481,7 +5443,7 @@ static void gfx_v8_0_send_serdes_cmd(struct amdgpu_device *adev,
 {
 	uint32_t data;
 
-	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 
 	WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff);
 	WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff);
@@ -5535,7 +5497,7 @@ static bool gfx_v8_0_is_rlc_enabled(struct amdgpu_device *adev)
 	return true;
 }
 
-static void gfx_v8_0_set_safe_mode(struct amdgpu_device *adev)
+static void gfx_v8_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	uint32_t data;
 	unsigned i;
@@ -5562,7 +5524,7 @@ static void gfx_v8_0_set_safe_mode(struct amdgpu_device *adev)
 	}
 }
 
-static void gfx_v8_0_unset_safe_mode(struct amdgpu_device *adev)
+static void gfx_v8_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	uint32_t data;
 	unsigned i;
@@ -5579,7 +5541,7 @@ static void gfx_v8_0_unset_safe_mode(struct amdgpu_device *adev)
 	}
 }
 
-static void gfx_v8_0_update_spm_vmid(struct amdgpu_device *adev, unsigned vmid)
+static void gfx_v8_0_update_spm_vmid(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned vmid)
 {
 	u32 data;
 
@@ -5621,8 +5583,6 @@ static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev
 {
 	uint32_t temp, data;
 
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
-
 	/* It is disabled by HW by default */
 	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) {
 		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGLS) {
@@ -5716,8 +5676,6 @@ static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev
 		/* 7- wait for RLC_SERDES_CU_MASTER & RLC_SERDES_NONCU_MASTER idle */
 		gfx_v8_0_wait_for_rlc_serdes(adev);
 	}
-
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
 }
 
 static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev,
@@ -5727,8 +5685,6 @@ static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev
 
 	temp = data = RREG32(mmRLC_CGCG_CGLS_CTRL);
 
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
-
 	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)) {
 		temp1 = data1 =	RREG32(mmRLC_CGTT_MGCG_OVERRIDE);
 		data1 &= ~RLC_CGTT_MGCG_OVERRIDE__CGCG_MASK;
@@ -5809,12 +5765,12 @@ static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev
 	}
 
 	gfx_v8_0_wait_for_rlc_serdes(adev);
-
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
 }
 static int gfx_v8_0_update_gfx_clock_gating(struct amdgpu_device *adev,
 					    bool enable)
 {
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
+
 	if (enable) {
 		/* CGCG/CGLS should be enabled after MGCG/MGLS/TS(CG/LS)
 		 * ===  MGCG + MGLS + TS(CG/LS) ===
@@ -5828,6 +5784,8 @@ static int gfx_v8_0_update_gfx_clock_gating(struct amdgpu_device *adev,
 		gfx_v8_0_update_coarse_grain_clock_gating(adev, enable);
 		gfx_v8_0_update_medium_grain_clock_gating(adev, enable);
 	}
+
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 	return 0;
 }
 
@@ -5978,10 +5936,10 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int gfx_v8_0_set_clockgating_state(void *handle,
+static int gfx_v8_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
@@ -6153,6 +6111,7 @@ static void gfx_v8_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr,
 {
 	bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT;
 	bool int_sel = flags & AMDGPU_FENCE_FLAG_INT;
+	bool exec = flags & AMDGPU_FENCE_FLAG_EXEC;
 
 	/* Workaround for cache flush problems. First send a dummy EOP
 	 * event down the pipe with seq one below.
@@ -6176,7 +6135,8 @@ static void gfx_v8_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr,
 				 EOP_TC_ACTION_EN |
 				 EOP_TC_WB_ACTION_EN |
 				 EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) |
-				 EVENT_INDEX(5)));
+				 EVENT_INDEX(5) |
+				 (exec ? EOP_EXEC : 0)));
 	amdgpu_ring_write(ring, addr & 0xfffffffc);
 	amdgpu_ring_write(ring, (upper_32_bits(addr) & 0xffff) |
 			  DATA_SEL(write64bit ? 2 : 1) | INT_SEL(int_sel ? 2 : 0));
@@ -6327,33 +6287,22 @@ static void gfx_v8_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags)
 	amdgpu_ring_write(ring, 0);
 }
 
-static unsigned gfx_v8_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring)
+static unsigned gfx_v8_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring,
+						  uint64_t addr)
 {
 	unsigned ret;
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, 0); /* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	/* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, 0);
 	ret = ring->wptr & ring->buf_mask;
-	amdgpu_ring_write(ring, 0x55aa55aa); /* patch dummy value later */
+	/* patch dummy value later */
+	amdgpu_ring_write(ring, 0);
 	return ret;
 }
 
-static void gfx_v8_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigned offset)
-{
-	unsigned cur;
-
-	BUG_ON(offset > ring->buf_mask);
-	BUG_ON(ring->ring[offset] != 0x55aa55aa);
-
-	cur = (ring->wptr & ring->buf_mask) - 1;
-	if (likely(cur > offset))
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->ring_size >> 2) - offset + cur;
-}
-
 static void gfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg,
 				    uint32_t reg_val_offs)
 {
@@ -6723,11 +6672,11 @@ static void gfx_v8_0_parse_sq_irq(struct amdgpu_device *adev, unsigned ih_data,
 			 */
 			if (from_wq) {
 				mutex_lock(&adev->grbm_idx_mutex);
-				gfx_v8_0_select_se_sh(adev, se_id, sh_id, cu_id);
+				gfx_v8_0_select_se_sh(adev, se_id, sh_id, cu_id, 0);
 
 				sq_edc_source = REG_GET_FIELD(RREG32(mmSQ_EDC_INFO), SQ_EDC_INFO, SOURCE);
 
-				gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+				gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 				mutex_unlock(&adev->grbm_idx_mutex);
 			}
 
@@ -6933,7 +6882,6 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = {
 	.emit_switch_buffer = gfx_v8_ring_emit_sb,
 	.emit_cntxcntl = gfx_v8_ring_emit_cntxcntl,
 	.init_cond_exec = gfx_v8_0_ring_emit_init_cond_exec,
-	.patch_cond_exec = gfx_v8_0_ring_emit_patch_cond_exec,
 	.emit_wreg = gfx_v8_0_ring_emit_wreg,
 	.soft_recovery = gfx_v8_0_ring_soft_recovery,
 	.emit_mem_sync = gfx_v8_0_emit_mem_sync,
@@ -6969,6 +6917,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = {
 	.insert_nop = amdgpu_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_wreg = gfx_v8_0_ring_emit_wreg,
+	.soft_recovery = gfx_v8_0_ring_soft_recovery,
 	.emit_mem_sync = gfx_v8_0_emit_mem_sync_compute,
 	.emit_wave_limit = gfx_v8_0_emit_wave_limit,
 };
@@ -6995,13 +6944,14 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_kiq = {
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_rreg = gfx_v8_0_ring_emit_rreg,
 	.emit_wreg = gfx_v8_0_ring_emit_wreg,
+	.emit_hdp_flush = gfx_v8_0_ring_emit_hdp_flush,
 };
 
 static void gfx_v8_0_set_ring_funcs(struct amdgpu_device *adev)
 {
 	int i;
 
-	adev->gfx.kiq.ring.funcs = &gfx_v8_0_ring_funcs_kiq;
+	adev->gfx.kiq[0].ring.funcs = &gfx_v8_0_ring_funcs_kiq;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
 		adev->gfx.gfx_ring[i].funcs = &gfx_v8_0_ring_funcs_gfx;
@@ -7116,12 +7066,12 @@ static void gfx_v8_0_get_cu_info(struct amdgpu_device *adev)
 			mask = 1;
 			ao_bitmap = 0;
 			counter = 0;
-			gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			if (i < 4 && j < 2)
 				gfx_v8_0_set_user_cu_inactive_bitmap(
 					adev, disable_masks[i * 2 + j]);
 			bitmap = gfx_v8_0_get_cu_active_bitmap(adev);
-			cu_info->bitmap[i][j] = bitmap;
+			cu_info->bitmap[0][i][j] = bitmap;
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k ++) {
 				if (bitmap & mask) {
@@ -7137,7 +7087,7 @@ static void gfx_v8_0_get_cu_info(struct amdgpu_device *adev)
 			cu_info->ao_cu_bitmap[i][j] = ao_bitmap;
 		}
 	}
-	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	cu_info->number = active_cu_number;
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index adbcd8127c82..0148d7ff34d9 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
@@ -50,6 +50,7 @@
 #include "amdgpu_ring_mux.h"
 #include "gfx_v9_4.h"
 #include "gfx_v9_0.h"
+#include "gfx_v9_0_cleaner_shader.h"
 #include "gfx_v9_4_2.h"
 
 #include "asic_reg/pwr/pwr_10_0_offset.h"
@@ -149,15 +150,161 @@ MODULE_FIRMWARE("amdgpu/aldebaran_sjt_mec2.bin");
 #define mmGOLDEN_TSC_COUNT_LOWER_Renoir                0x0026
 #define mmGOLDEN_TSC_COUNT_LOWER_Renoir_BASE_IDX       1
 
-#define mmGOLDEN_TSC_COUNT_UPPER_Raven   0x007a
-#define mmGOLDEN_TSC_COUNT_UPPER_Raven_BASE_IDX 0
-#define mmGOLDEN_TSC_COUNT_LOWER_Raven   0x007b
-#define mmGOLDEN_TSC_COUNT_LOWER_Raven_BASE_IDX 0
+static const struct amdgpu_hwip_reg_entry gc_reg_list_9[] = {
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_STALLED_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPC_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPF_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPC_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPF_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPF_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB0_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB0_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB0_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB1_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB1_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB1_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB2_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB2_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB2_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCPF_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCPC_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCPG_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGDS_PROTECTION_FAULT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGDS_VM_PROTECTION_FAULT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmIA_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmIA_UTCL1_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmPA_CL_CNTL_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRMI_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSQC_DCACHE_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSQC_ICACHE_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSQ_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmTCP_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmWD_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmVM_L2_PROTECTION_FAULT_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmVM_L2_PROTECTION_FAULT_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_DEBUG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC1_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC2_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPC_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_COMMAND),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_MESSAGE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_ARGUMENT_1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_ARGUMENT_2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSMU_RLC_RESPONSE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SAFE_MODE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_SAFE_MODE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_INT_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_GPM_GENERAL_6),
+	/* SE status registers */
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE0),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE3),
+	/* packet headers */
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP)
+};
 
-#define mmGOLDEN_TSC_COUNT_UPPER_Raven2   0x0068
-#define mmGOLDEN_TSC_COUNT_UPPER_Raven2_BASE_IDX 0
-#define mmGOLDEN_TSC_COUNT_LOWER_Raven2   0x0069
-#define mmGOLDEN_TSC_COUNT_LOWER_Raven2_BASE_IDX 0
+static const struct amdgpu_hwip_reg_entry gc_cp_reg_list_9[] = {
+	/* compute queue registers */
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_VMID),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_ACTIVE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PERSISTENT_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PIPE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_QUEUE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_QUANTUM),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_DEQUEUE_REQUEST),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_EVENTS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_BASE_ADDR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CNTL_STACK_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CNTL_STACK_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_WG_STATE_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_GDS_RESOURCE_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_WPTR_MEM),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_GFX_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP)
+};
 
 enum ta_ras_gfx_subblock {
 	/*CPC*/
@@ -765,17 +912,27 @@ static void gfx_v9_0_set_rlc_funcs(struct amdgpu_device *adev);
 static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev,
 				struct amdgpu_cu_info *cu_info);
 static uint64_t gfx_v9_0_get_gpu_clock_counter(struct amdgpu_device *adev);
-static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume);
+static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume, bool usegds);
 static u64 gfx_v9_0_ring_get_rptr_compute(struct amdgpu_ring *ring);
 static void gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev,
 					  void *ras_error_status);
 static int gfx_v9_0_ras_error_inject(struct amdgpu_device *adev,
-				     void *inject_if);
+				     void *inject_if, uint32_t instance_mask);
 static void gfx_v9_0_reset_ras_error_count(struct amdgpu_device *adev);
+static void gfx_v9_0_update_spm_vmid_internal(struct amdgpu_device *adev,
+					      unsigned int vmid);
+static void gfx_v9_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id);
+static void gfx_v9_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id);
 
 static void gfx_v9_0_kiq_set_resources(struct amdgpu_ring *kiq_ring,
 				uint64_t queue_mask)
 {
+	struct amdgpu_device *adev = kiq_ring->adev;
+	u64 shader_mc_addr;
+
+	/* Cleaner shader MC address */
+	shader_mc_addr = adev->gfx.cleaner_shader_gpu_addr >> 8;
+
 	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_RESOURCES, 6));
 	amdgpu_ring_write(kiq_ring,
 		PACKET3_SET_RESOURCES_VMID_MASK(0) |
@@ -785,8 +942,8 @@ static void gfx_v9_0_kiq_set_resources(struct amdgpu_ring *kiq_ring,
 			lower_32_bits(queue_mask));	/* queue mask lo */
 	amdgpu_ring_write(kiq_ring,
 			upper_32_bits(queue_mask));	/* queue mask hi */
-	amdgpu_ring_write(kiq_ring, 0);	/* gws mask lo */
-	amdgpu_ring_write(kiq_ring, 0);	/* gws mask hi */
+	amdgpu_ring_write(kiq_ring, lower_32_bits(shader_mc_addr)); /* cleaner shader addr lo */
+	amdgpu_ring_write(kiq_ring, upper_32_bits(shader_mc_addr)); /* cleaner shader addr hi */
 	amdgpu_ring_write(kiq_ring, 0);	/* oac mask */
 	amdgpu_ring_write(kiq_ring, 0);	/* gds heap base:0, gds heap size:0 */
 }
@@ -883,12 +1040,47 @@ static void gfx_v9_0_kiq_invalidate_tlbs(struct amdgpu_ring *kiq_ring,
 			PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(flush_type));
 }
 
+
+static void gfx_v9_0_kiq_reset_hw_queue(struct amdgpu_ring *kiq_ring, uint32_t queue_type,
+					uint32_t me_id, uint32_t pipe_id, uint32_t queue_id,
+					uint32_t xcc_id, uint32_t vmid)
+{
+	struct amdgpu_device *adev = kiq_ring->adev;
+	unsigned i;
+
+	/* enter save mode */
+	amdgpu_gfx_rlc_enter_safe_mode(adev, xcc_id);
+	mutex_lock(&adev->srbm_mutex);
+	soc15_grbm_select(adev, me_id, pipe_id, queue_id, 0, 0);
+
+	if (queue_type == AMDGPU_RING_TYPE_COMPUTE) {
+		WREG32_SOC15(GC, 0, mmCP_HQD_DEQUEUE_REQUEST, 0x2);
+		WREG32_SOC15(GC, 0, mmSPI_COMPUTE_QUEUE_RESET, 0x1);
+		/* wait till dequeue take effects */
+		for (i = 0; i < adev->usec_timeout; i++) {
+			if (!(RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+		if (i >= adev->usec_timeout)
+			dev_err(adev->dev, "fail to wait on hqd deactive\n");
+	} else {
+		dev_err(adev->dev, "reset queue_type(%d) not supported\n", queue_type);
+	}
+
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	/* exit safe mode */
+	amdgpu_gfx_rlc_exit_safe_mode(adev, xcc_id);
+}
+
 static const struct kiq_pm4_funcs gfx_v9_0_kiq_pm4_funcs = {
 	.kiq_set_resources = gfx_v9_0_kiq_set_resources,
 	.kiq_map_queues = gfx_v9_0_kiq_map_queues,
 	.kiq_unmap_queues = gfx_v9_0_kiq_unmap_queues,
 	.kiq_query_status = gfx_v9_0_kiq_query_status,
 	.kiq_invalidate_tlbs = gfx_v9_0_kiq_invalidate_tlbs,
+	.kiq_reset_hw_queue = gfx_v9_0_kiq_reset_hw_queue,
 	.set_resources_size = 8,
 	.map_queues_size = 7,
 	.unmap_queues_size = 6,
@@ -898,12 +1090,12 @@ static const struct kiq_pm4_funcs gfx_v9_0_kiq_pm4_funcs = {
 
 static void gfx_v9_0_set_kiq_pm4_funcs(struct amdgpu_device *adev)
 {
-	adev->gfx.kiq.pmf = &gfx_v9_0_kiq_pm4_funcs;
+	adev->gfx.kiq[0].pmf = &gfx_v9_0_kiq_pm4_funcs;
 }
 
 static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 		soc15_program_register_sequence(adev,
 						golden_settings_gc_9_0,
@@ -959,8 +1151,8 @@ static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev)
 		break;
 	}
 
-	if ((adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 1)) &&
-	    (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 2)))
+	if ((amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 1)) &&
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 2)))
 		soc15_program_register_sequence(adev, golden_settings_gc_9_x_common,
 						(const u32)ARRAY_SIZE(golden_settings_gc_9_x_common));
 }
@@ -1047,8 +1239,8 @@ static int gfx_v9_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	gpu_addr = adev->wb.gpu_addr + (index * 4);
 	adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD);
 	memset(&ib, 0, sizeof(ib));
-	r = amdgpu_ib_get(adev, NULL, 16,
-					AMDGPU_IB_POOL_DIRECT, &ib);
+
+	r = amdgpu_ib_get(adev, NULL, 20, AMDGPU_IB_POOL_DIRECT, &ib);
 	if (r)
 		goto err1;
 
@@ -1078,7 +1270,7 @@ static int gfx_v9_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = -EINVAL;
 
 err2:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err1:
 	amdgpu_device_wb_free(adev, index);
@@ -1103,14 +1295,15 @@ static void gfx_v9_0_check_fw_write_wait(struct amdgpu_device *adev)
 	adev->gfx.me_fw_write_wait = false;
 	adev->gfx.mec_fw_write_wait = false;
 
-	if ((adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 1)) &&
+	if ((amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 1)) &&
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 2)) &&
 	    ((adev->gfx.mec_fw_version < 0x000001a5) ||
-	    (adev->gfx.mec_feature_version < 46) ||
-	    (adev->gfx.pfp_fw_version < 0x000000b7) ||
-	    (adev->gfx.pfp_feature_version < 46)))
+	     (adev->gfx.mec_feature_version < 46) ||
+	     (adev->gfx.pfp_fw_version < 0x000000b7) ||
+	     (adev->gfx.pfp_feature_version < 46)))
 		DRM_WARN_ONCE("CP firmware version too old, please update!");
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 		if ((adev->gfx.me_fw_version >= 0x0000009c) &&
 		    (adev->gfx.me_feature_version >= 42) &&
@@ -1180,6 +1373,10 @@ static const struct amdgpu_gfxoff_quirk amdgpu_gfxoff_quirk_list[] = {
 	{ 0x1002, 0x15dd, 0x1002, 0x15dd, 0xc6 },
 	/* Apple MacBook Pro (15-inch, 2019) Radeon Pro Vega 20 4 GB */
 	{ 0x1002, 0x69af, 0x106b, 0x019a, 0xc0 },
+	/* https://bbs.openkylin.top/t/topic/171497 */
+	{ 0x1002, 0x15d8, 0x19e5, 0x3e14, 0xc2 },
+	/* HP 705G4 DM with R5 2400G */
+	{ 0x1002, 0x15dd, 0x103c, 0x8464, 0xd6 },
 	{ 0, 0, 0, 0, 0 },
 };
 
@@ -1210,7 +1407,7 @@ static bool is_raven_kicker(struct amdgpu_device *adev)
 
 static bool check_if_enlarge_doorbell_range(struct amdgpu_device *adev)
 {
-	if ((adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 3, 0)) &&
+	if ((amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 3, 0)) &&
 	    (adev->gfx.me_fw_version >= 0x000000a5) &&
 	    (adev->gfx.me_feature_version >= 52))
 		return true;
@@ -1223,7 +1420,7 @@ static void gfx_v9_0_check_if_need_gfxoff(struct amdgpu_device *adev)
 	if (gfx_v9_0_should_disable_gfxoff(adev->pdev))
 		adev->pm.pp_feature &= ~PP_GFXOFF_MASK;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 2, 1):
 	case IP_VERSION(9, 4, 0):
@@ -1257,23 +1454,25 @@ static void gfx_v9_0_check_if_need_gfxoff(struct amdgpu_device *adev)
 static int gfx_v9_0_init_cp_gfx_microcode(struct amdgpu_device *adev,
 					  char *chip_name)
 {
-	char fw_name[30];
 	int err;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_pfp.bin", chip_name);
 	if (err)
 		goto out;
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_PFP);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_me.bin", chip_name);
 	if (err)
 		goto out;
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_ME);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_ce.bin", chip_name);
 	if (err)
 		goto out;
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_CE);
@@ -1290,7 +1489,6 @@ out:
 static int gfx_v9_0_init_rlc_microcode(struct amdgpu_device *adev,
 				       char *chip_name)
 {
-	char fw_name[30];
 	int err;
 	const struct rlc_firmware_header_v2_0 *rlc_hdr;
 	uint16_t version_major;
@@ -1308,20 +1506,25 @@ static int gfx_v9_0_init_rlc_microcode(struct amdgpu_device *adev,
 	if (!strcmp(chip_name, "picasso") &&
 		(((adev->pdev->revision >= 0xC8) && (adev->pdev->revision <= 0xCF)) ||
 		((adev->pdev->revision >= 0xD8) && (adev->pdev->revision <= 0xDF))))
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc_am4.bin", chip_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_rlc_am4.bin", chip_name);
 	else if (!strcmp(chip_name, "raven") && (amdgpu_pm_load_smu_firmware(adev, &smu_version) == 0) &&
 		(smu_version >= 0x41e2b))
 		/**
 		*SMC is loaded by SBIOS on APU and it's able to get the SMU version directly.
 		*/
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_kicker_rlc.bin", chip_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_kicker_rlc.bin", chip_name);
 	else
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_rlc.bin", chip_name);
 	if (err)
 		goto out;
-	rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
 
+	rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
 	version_major = le16_to_cpu(rlc_hdr->header.header_version_major);
 	version_minor = le16_to_cpu(rlc_hdr->header.header_version_minor);
 	err = amdgpu_gfx_rlc_init_microcode(adev, version_major, version_minor);
@@ -1334,9 +1537,9 @@ out:
 
 static bool gfx_v9_0_load_mec2_fw_bin_support(struct amdgpu_device *adev)
 {
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2) ||
-	    adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1) ||
-	    adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 3, 0))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 3, 0))
 		return false;
 
 	return true;
@@ -1345,28 +1548,31 @@ static bool gfx_v9_0_load_mec2_fw_bin_support(struct amdgpu_device *adev)
 static int gfx_v9_0_init_cp_compute_microcode(struct amdgpu_device *adev,
 					      char *chip_name)
 {
-	char fw_name[30];
 	int err;
 
 	if (amdgpu_sriov_vf(adev) && (adev->asic_type == CHIP_ALDEBARAN))
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sjt_mec.bin", chip_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_sjt_mec.bin", chip_name);
 	else
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name);
-
-	err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_mec.bin", chip_name);
 	if (err)
 		goto out;
+
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1);
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1_JT);
 
 	if (gfx_v9_0_load_mec2_fw_bin_support(adev)) {
 		if (amdgpu_sriov_vf(adev) && (adev->asic_type == CHIP_ALDEBARAN))
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sjt_mec2.bin", chip_name);
+			err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_sjt_mec2.bin", chip_name);
 		else
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name);
-
-		/* ignore failures to load */
-		err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_mec2.bin", chip_name);
 		if (!err) {
 			amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2);
 			amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2_JT);
@@ -1442,45 +1648,18 @@ static u32 gfx_v9_0_get_csb_size(struct amdgpu_device *adev)
 	return count;
 }
 
-static void gfx_v9_0_get_csb_buffer(struct amdgpu_device *adev,
-				    volatile u32 *buffer)
+static void gfx_v9_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer)
 {
-	u32 count = 0, i;
-	const struct cs_section_def *sect = NULL;
-	const struct cs_extent_def *ext = NULL;
+	u32 count = 0;
 
 	if (adev->gfx.rlc.cs_data == NULL)
 		return;
 	if (buffer == NULL)
 		return;
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1));
-	buffer[count++] = cpu_to_le32(0x80000000);
-	buffer[count++] = cpu_to_le32(0x80000000);
-
-	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
-		for (ext = sect->section; ext->extent != NULL; ++ext) {
-			if (sect->id == SECT_CONTEXT) {
-				buffer[count++] =
-					cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count));
-				buffer[count++] = cpu_to_le32(ext->reg_index -
-						PACKET3_SET_CONTEXT_REG_START);
-				for (i = 0; i < ext->reg_count; i++)
-					buffer[count++] = cpu_to_le32(ext->extent[i]);
-			} else {
-				return;
-			}
-		}
-	}
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0));
-	buffer[count++] = cpu_to_le32(0);
+	count = amdgpu_gfx_csb_preamble_start(buffer);
+	count = amdgpu_gfx_csb_data_parser(adev, buffer, count);
+	amdgpu_gfx_csb_preamble_end(buffer, count);
 }
 
 static void gfx_v9_0_init_always_on_cu_mask(struct amdgpu_device *adev)
@@ -1493,7 +1672,7 @@ static void gfx_v9_0_init_always_on_cu_mask(struct amdgpu_device *adev)
 
 	if (adev->flags & AMD_IS_APU)
 		always_on_cu_num = 4;
-	else if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 2, 1))
+	else if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 2, 1))
 		always_on_cu_num = 8;
 	else
 		always_on_cu_num = 12;
@@ -1504,10 +1683,10 @@ static void gfx_v9_0_init_always_on_cu_mask(struct amdgpu_device *adev)
 			mask = 1;
 			cu_bitmap = 0;
 			counter = 0;
-			amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff);
+			amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff, 0);
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k ++) {
-				if (cu_info->bitmap[i][j] & mask) {
+				if (cu_info->bitmap[0][i][j] & mask) {
 					if (counter == pg_always_on_cu_num)
 						WREG32_SOC15(GC, 0, mmRLC_PG_ALWAYS_ON_CU_MASK, cu_bitmap);
 					if (counter < always_on_cu_num)
@@ -1523,7 +1702,7 @@ static void gfx_v9_0_init_always_on_cu_mask(struct amdgpu_device *adev)
 			cu_info->ao_cu_bitmap[i][j] = cu_bitmap;
 		}
 	}
-	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 }
 
@@ -1545,7 +1724,7 @@ static void gfx_v9_0_init_lbpw(struct amdgpu_device *adev)
 
 	mutex_lock(&adev->grbm_idx_mutex);
 	/* set mmRLC_LB_INIT_CU_MASK thru broadcast mode to enable all SE/SH*/
-	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	WREG32_SOC15(GC, 0, mmRLC_LB_INIT_CU_MASK, 0xffffffff);
 
 	/* set mmRLC_LB_PARAMS = 0x003F_1006 */
@@ -1594,7 +1773,7 @@ static void gfx_v9_4_init_lbpw(struct amdgpu_device *adev)
 
 	mutex_lock(&adev->grbm_idx_mutex);
 	/* set mmRLC_LB_INIT_CU_MASK thru broadcast mode to enable all SE/SH*/
-	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	WREG32_SOC15(GC, 0, mmRLC_LB_INIT_CU_MASK, 0xffffffff);
 
 	/* set mmRLC_LB_PARAMS = 0x003F_1006 */
@@ -1642,7 +1821,7 @@ static void gfx_v9_0_init_rlcg_reg_access_ctrl(struct amdgpu_device *adev)
 {
 	struct amdgpu_rlcg_reg_access_ctrl *reg_access_ctrl;
 
-	reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl;
+	reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl[0];
 	reg_access_ctrl->scratch_reg0 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0);
 	reg_access_ctrl->scratch_reg1 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG1);
 	reg_access_ctrl->scratch_reg2 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG2);
@@ -1677,22 +1856,6 @@ static int gfx_v9_0_rlc_init(struct amdgpu_device *adev)
 			return r;
 	}
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
-	case IP_VERSION(9, 2, 2):
-	case IP_VERSION(9, 1, 0):
-		gfx_v9_0_init_lbpw(adev);
-		break;
-	case IP_VERSION(9, 4, 0):
-		gfx_v9_4_init_lbpw(adev);
-		break;
-	default:
-		break;
-	}
-
-	/* init spm vmid with 0xf */
-	if (adev->gfx.rlc.funcs->update_spm_vmid)
-		adev->gfx.rlc.funcs->update_spm_vmid(adev, 0xf);
-
 	return 0;
 }
 
@@ -1713,7 +1876,7 @@ static int gfx_v9_0_mec_init(struct amdgpu_device *adev)
 
 	const struct gfx_firmware_header_v1_0 *mec_hdr;
 
-	bitmap_zero(adev->gfx.mec.queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
+	bitmap_zero(adev->gfx.mec_bitmap[0].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
 
 	/* take ownership of the relevant compute queues */
 	amdgpu_gfx_compute_queue_acquire(adev);
@@ -1788,7 +1951,7 @@ static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd,
 		*(out++) = RREG32_SOC15(GC, 0, mmSQ_IND_DATA);
 }
 
-static void gfx_v9_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
+static void gfx_v9_0_read_wave_data(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
 {
 	/* type 1 wave data */
 	dst[(*no_fields)++] = 1;
@@ -1809,7 +1972,7 @@ static void gfx_v9_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, u
 	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_MODE);
 }
 
-static void gfx_v9_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v9_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				     uint32_t wave, uint32_t start,
 				     uint32_t size, uint32_t *dst)
 {
@@ -1818,7 +1981,7 @@ static void gfx_v9_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
 		start + SQIND_WAVE_SGPRS_OFFSET, size, dst);
 }
 
-static void gfx_v9_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v9_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				     uint32_t wave, uint32_t thread,
 				     uint32_t start, uint32_t size,
 				     uint32_t *dst)
@@ -1829,9 +1992,9 @@ static void gfx_v9_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd,
 }
 
 static void gfx_v9_0_select_me_pipe_q(struct amdgpu_device *adev,
-				  u32 me, u32 pipe, u32 q, u32 vm)
+				  u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id)
 {
-	soc15_grbm_select(adev, me, pipe, q, vm);
+	soc15_grbm_select(adev, me, pipe, q, vm, 0);
 }
 
 static const struct amdgpu_gfx_funcs gfx_v9_0_gfx_funcs = {
@@ -1860,7 +2023,7 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev)
 	u32 gb_addr_config;
 	int err;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 		adev->gfx.config.max_hw_contexts = 8;
 		adev->gfx.config.sc_prim_fifo_size_frontend = 0x20;
@@ -2005,7 +2168,7 @@ static int gfx_v9_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
 	ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
 	ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
 				+ (ring_id * GFX9_MEC_HPD_SIZE);
-	ring->vm_hub = AMDGPU_GFXHUB_0;
+	ring->vm_hub = AMDGPU_GFXHUB(0);
 	sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
 
 	irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
@@ -2018,15 +2181,43 @@ static int gfx_v9_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
 				hw_prio, NULL);
 }
 
-static int gfx_v9_0_sw_init(void *handle)
+static void gfx_v9_0_alloc_ip_dump(struct amdgpu_device *adev)
+{
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_9);
+	uint32_t *ptr;
+	uint32_t inst;
+
+	ptr = kcalloc(reg_count, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for GFX IP Dump\n");
+		adev->gfx.ip_dump_core = NULL;
+	} else {
+		adev->gfx.ip_dump_core = ptr;
+	}
+
+	/* Allocate memory for compute queue registers for all the instances */
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_9);
+	inst = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe_per_mec *
+		adev->gfx.mec.num_queue_per_pipe;
+
+	ptr = kcalloc(reg_count * inst, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for Compute Queues IP Dump\n");
+		adev->gfx.ip_dump_compute_queues = NULL;
+	} else {
+		adev->gfx.ip_dump_compute_queues = ptr;
+	}
+}
+
+static int gfx_v9_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int i, j, k, r, ring_id;
+	int xcc_id = 0;
 	struct amdgpu_ring *ring;
-	struct amdgpu_kiq *kiq;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	unsigned int hw_prio;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 2, 1):
 	case IP_VERSION(9, 4, 0):
@@ -2042,6 +2233,43 @@ static int gfx_v9_0_sw_init(void *handle)
 		break;
 	}
 
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(9, 0, 1):
+	case IP_VERSION(9, 2, 1):
+	case IP_VERSION(9, 4, 0):
+	case IP_VERSION(9, 2, 2):
+	case IP_VERSION(9, 1, 0):
+	case IP_VERSION(9, 3, 0):
+		adev->gfx.cleaner_shader_ptr = gfx_9_4_2_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_9_4_2_cleaner_shader_hex);
+		if (adev->gfx.me_fw_version  >= 167 &&
+		    adev->gfx.pfp_fw_version >= 196 &&
+		    adev->gfx.mec_fw_version >= 474) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	case IP_VERSION(9, 4, 2):
+		adev->gfx.cleaner_shader_ptr = gfx_9_4_2_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_9_4_2_cleaner_shader_hex);
+		if (adev->gfx.mec_fw_version >= 88) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	default:
+		adev->gfx.enable_cleaner_shader = false;
+		break;
+	}
+
 	adev->gfx.mec.num_pipe_per_mec = 4;
 	adev->gfx.mec.num_queue_per_pipe = 8;
 
@@ -2050,6 +2278,13 @@ static int gfx_v9_0_sw_init(void *handle)
 	if (r)
 		return r;
 
+	/* Bad opcode Event */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP,
+			      GFX_9_0__SRCID__CP_BAD_OPCODE_ERROR,
+			      &adev->gfx.bad_op_irq);
+	if (r)
+		return r;
+
 	/* Privileged reg */
 	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, GFX_9_0__SRCID__CP_PRIV_REG_FAULT,
 			      &adev->gfx.priv_reg_irq);
@@ -2104,8 +2339,8 @@ static int gfx_v9_0_sw_init(void *handle)
 		ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
 
 		/* disable scheduler on the real ring */
-		ring->no_scheduler = true;
-		ring->vm_hub = AMDGPU_GFXHUB_0;
+		ring->no_scheduler = adev->gfx.mcbp;
+		ring->vm_hub = AMDGPU_GFXHUB(0);
 		r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq,
 				     AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP,
 				     AMDGPU_RING_PRIO_DEFAULT, NULL);
@@ -2114,7 +2349,7 @@ static int gfx_v9_0_sw_init(void *handle)
 	}
 
 	/* set up the software rings */
-	if (adev->gfx.num_gfx_rings) {
+	if (adev->gfx.mcbp && adev->gfx.num_gfx_rings) {
 		for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++) {
 			ring = &adev->gfx.sw_gfx_ring[i];
 			ring->ring_obj = NULL;
@@ -2123,7 +2358,7 @@ static int gfx_v9_0_sw_init(void *handle)
 			ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
 			ring->is_sw_ring = true;
 			hw_prio = amdgpu_sw_ring_priority(i);
-			ring->vm_hub = AMDGPU_GFXHUB_0;
+			ring->vm_hub = AMDGPU_GFXHUB(0);
 			r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq,
 					     AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP, hw_prio,
 					     NULL);
@@ -2154,7 +2389,8 @@ static int gfx_v9_0_sw_init(void *handle)
 	for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
 		for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
 			for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
-				if (!amdgpu_gfx_is_mec_queue_enabled(adev, i, k, j))
+				if (!amdgpu_gfx_is_mec_queue_enabled(adev, 0, i,
+								     k, j))
 					continue;
 
 				r = gfx_v9_0_compute_ring_init(adev,
@@ -2168,19 +2404,26 @@ static int gfx_v9_0_sw_init(void *handle)
 		}
 	}
 
-	r = amdgpu_gfx_kiq_init(adev, GFX9_MEC_HPD_SIZE);
+	/* TODO: Add queue reset mask when FW fully supports it */
+	adev->gfx.gfx_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]);
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+	if (!amdgpu_sriov_vf(adev) && !adev->debug_disable_gpu_ring_reset)
+		adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+
+	r = amdgpu_gfx_kiq_init(adev, GFX9_MEC_HPD_SIZE, 0);
 	if (r) {
 		DRM_ERROR("Failed to init KIQ BOs!\n");
 		return r;
 	}
 
-	kiq = &adev->gfx.kiq;
-	r = amdgpu_gfx_kiq_init_ring(adev, &kiq->ring, &kiq->irq);
+	r = amdgpu_gfx_kiq_init_ring(adev, xcc_id);
 	if (r)
 		return r;
 
 	/* create MQD for all compute queues as wel as KIQ for SRIOV case */
-	r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct v9_mqd_allocation));
+	r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct v9_mqd_allocation), 0);
 	if (r)
 		return r;
 
@@ -2195,16 +2438,22 @@ static int gfx_v9_0_sw_init(void *handle)
 		return -EINVAL;
 	}
 
+	gfx_v9_0_alloc_ip_dump(adev);
+
+	r = amdgpu_gfx_sysfs_init(adev);
+	if (r)
+		return r;
+
 	return 0;
 }
 
 
-static int gfx_v9_0_sw_fini(void *handle)
+static int gfx_v9_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	if (adev->gfx.num_gfx_rings) {
+	if (adev->gfx.mcbp && adev->gfx.num_gfx_rings) {
 		for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++)
 			amdgpu_ring_fini(&adev->gfx.sw_gfx_ring[i]);
 		amdgpu_ring_mux_fini(&adev->gfx.muxer);
@@ -2215,9 +2464,11 @@ static int gfx_v9_0_sw_fini(void *handle)
 	for (i = 0; i < adev->gfx.num_compute_rings; i++)
 		amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
 
-	amdgpu_gfx_mqd_sw_fini(adev);
-	amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring);
-	amdgpu_gfx_kiq_fini(adev);
+	amdgpu_gfx_mqd_sw_fini(adev, 0);
+	amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq[0].ring);
+	amdgpu_gfx_kiq_fini(adev, 0);
+
+	amdgpu_gfx_cleaner_shader_sw_fini(adev);
 
 	gfx_v9_0_mec_fini(adev);
 	amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj,
@@ -2230,6 +2481,11 @@ static int gfx_v9_0_sw_fini(void *handle)
 	}
 	gfx_v9_0_free_microcode(adev);
 
+	amdgpu_gfx_sysfs_fini(adev);
+
+	kfree(adev->gfx.ip_dump_core);
+	kfree(adev->gfx.ip_dump_compute_queues);
+
 	return 0;
 }
 
@@ -2240,7 +2496,7 @@ static void gfx_v9_0_tiling_mode_table_init(struct amdgpu_device *adev)
 }
 
 void gfx_v9_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num,
-			   u32 instance)
+			   u32 instance, int xcc_id)
 {
 	u32 data;
 
@@ -2289,19 +2545,42 @@ static void gfx_v9_0_setup_rb(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff);
+			amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff, 0);
 			data = gfx_v9_0_get_rb_active_bitmap(adev);
 			active_rbs |= data << ((i * adev->gfx.config.max_sh_per_se + j) *
 					       rb_bitmap_width_per_sh);
 		}
 	}
-	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	adev->gfx.config.backend_enable_mask = active_rbs;
 	adev->gfx.config.num_rbs = hweight32(active_rbs);
 }
 
+static void gfx_v9_0_debug_trap_config_init(struct amdgpu_device *adev,
+				uint32_t first_vmid,
+				uint32_t last_vmid)
+{
+	uint32_t data;
+	uint32_t trap_config_vmid_mask = 0;
+	int i;
+
+	/* Calculate trap config vmid mask */
+	for (i = first_vmid; i < last_vmid; i++)
+		trap_config_vmid_mask |= (1 << i);
+
+	data = REG_SET_FIELD(0, SPI_GDBG_TRAP_CONFIG,
+			VMID_SEL, trap_config_vmid_mask);
+	data = REG_SET_FIELD(data, SPI_GDBG_TRAP_CONFIG,
+			TRAP_EN, 1);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_CONFIG), data);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_DATA0), 0);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_DATA1), 0);
+}
+
 #define DEFAULT_SH_MEM_BASES	(0x6000)
 static void gfx_v9_0_init_compute_vmid(struct amdgpu_device *adev)
 {
@@ -2323,12 +2602,12 @@ static void gfx_v9_0_init_compute_vmid(struct amdgpu_device *adev)
 
 	mutex_lock(&adev->srbm_mutex);
 	for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) {
-		soc15_grbm_select(adev, 0, 0, 0, i);
+		soc15_grbm_select(adev, 0, 0, 0, i, 0);
 		/* CP and shaders */
 		WREG32_SOC15_RLC(GC, 0, mmSH_MEM_CONFIG, sh_mem_config);
 		WREG32_SOC15_RLC(GC, 0, mmSH_MEM_BASES, sh_mem_bases);
 	}
-	soc15_grbm_select(adev, 0, 0, 0, 0);
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 	mutex_unlock(&adev->srbm_mutex);
 
 	/* Initialize all compute VMIDs to have no GDS, GWS, or OA
@@ -2363,13 +2642,16 @@ static void gfx_v9_0_init_sq_config(struct amdgpu_device *adev)
 {
 	uint32_t tmp;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 4, 1):
 		tmp = RREG32_SOC15(GC, 0, mmSQ_CONFIG);
-		tmp = REG_SET_FIELD(tmp, SQ_CONFIG,
-					DISABLE_BARRIER_WAITCNT, 1);
+		tmp = REG_SET_FIELD(tmp, SQ_CONFIG, DISABLE_BARRIER_WAITCNT,
+				!READ_ONCE(adev->barrier_has_auto_waitcnt));
 		WREG32_SOC15(GC, 0, mmSQ_CONFIG, tmp);
 		break;
+	case IP_VERSION(9, 4, 2):
+		gfx_v9_4_2_init_sq(adev);
+		break;
 	default:
 		break;
 	}
@@ -2380,7 +2662,10 @@ static void gfx_v9_0_constants_init(struct amdgpu_device *adev)
 	u32 tmp;
 	int i;
 
-	WREG32_FIELD15_RLC(GC, 0, GRBM_CNTL, READ_TIMEOUT, 0xff);
+	if (!amdgpu_sriov_vf(adev) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 2)) {
+		WREG32_FIELD15_RLC(GC, 0, GRBM_CNTL, READ_TIMEOUT, 0xff);
+	}
 
 	gfx_v9_0_tiling_mode_table_init(adev);
 
@@ -2392,8 +2677,8 @@ static void gfx_v9_0_constants_init(struct amdgpu_device *adev)
 	/* XXX SH_MEM regs */
 	/* where to put LDS, scratch, GPUVM in FSA64 space */
 	mutex_lock(&adev->srbm_mutex);
-	for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB_0].num_ids; i++) {
-		soc15_grbm_select(adev, 0, 0, 0, i);
+	for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
+		soc15_grbm_select(adev, 0, 0, 0, i, 0);
 		/* CP and shaders */
 		if (i == 0) {
 			tmp = REG_SET_FIELD(0, SH_MEM_CONFIG, ALIGNMENT_MODE,
@@ -2415,7 +2700,7 @@ static void gfx_v9_0_constants_init(struct amdgpu_device *adev)
 			WREG32_SOC15_RLC(GC, 0, mmSH_MEM_BASES, tmp);
 		}
 	}
-	soc15_grbm_select(adev, 0, 0, 0, 0);
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 
 	mutex_unlock(&adev->srbm_mutex);
 
@@ -2432,7 +2717,7 @@ static void gfx_v9_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff);
+			amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff, 0);
 			for (k = 0; k < adev->usec_timeout; k++) {
 				if (RREG32_SOC15(GC, 0, mmRLC_SERDES_CU_MASTER_BUSY) == 0)
 					break;
@@ -2440,7 +2725,7 @@ static void gfx_v9_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
 			}
 			if (k == adev->usec_timeout) {
 				amdgpu_gfx_select_se_sh(adev, 0xffffffff,
-						      0xffffffff, 0xffffffff);
+						      0xffffffff, 0xffffffff, 0);
 				mutex_unlock(&adev->grbm_idx_mutex);
 				DRM_INFO("Timeout wait for RLC serdes %u,%u\n",
 					 i, j);
@@ -2448,7 +2733,7 @@ static void gfx_v9_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
 			}
 		}
 	}
-	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	mask = RLC_SERDES_NONCU_MASTER_BUSY__SE_MASTER_BUSY_MASK |
@@ -2474,7 +2759,7 @@ static void gfx_v9_0_enable_gui_idle_interrupt(struct amdgpu_device *adev,
 	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE, enable ? 1 : 0);
 	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_EMPTY_INT_ENABLE, enable ? 1 : 0);
 	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE, enable ? 1 : 0);
-	if(adev->gfx.num_gfx_rings)
+	if (adev->gfx.num_gfx_rings)
 		tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, GFX_IDLE_INT_ENABLE, enable ? 1 : 0);
 
 	WREG32_SOC15(GC, 0, mmCP_INT_CNTL_RING0, tmp);
@@ -2700,7 +2985,7 @@ static void gfx_v9_0_init_gfx_power_gating(struct amdgpu_device *adev)
 		/* program GRBM_REG_SAVE_GFX_IDLE_THRESHOLD to 0x55f0 */
 		data |= (0x55f0 << RLC_AUTO_PG_CTRL__GRBM_REG_SAVE_GFX_IDLE_THRESHOLD__SHIFT);
 		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_AUTO_PG_CTRL), data);
-		if (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 3, 0))
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 3, 0))
 			pwr_10_0_gfxip_control_over_cgpg(adev, true);
 	}
 }
@@ -2812,7 +3097,8 @@ static void gfx_v9_0_init_pg(struct amdgpu_device *adev)
 	 * And it's needed by gfxoff feature.
 	 */
 	if (adev->gfx.rlc.is_rlc_v2_1) {
-		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 2, 1) ||
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			    IP_VERSION(9, 2, 1) ||
 		    (adev->apu_flags & AMD_APU_IS_RAVEN2))
 			gfx_v9_1_init_rlc_save_restore_list(adev);
 		gfx_v9_0_enable_save_restore_machine(adev);
@@ -2925,15 +3211,17 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev)
 			return r;
 	}
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 2, 2):
 	case IP_VERSION(9, 1, 0):
+		gfx_v9_0_init_lbpw(adev);
 		if (amdgpu_lbpw == 0)
 			gfx_v9_0_enable_lbpw(adev, false);
 		else
 			gfx_v9_0_enable_lbpw(adev, true);
 		break;
 	case IP_VERSION(9, 4, 0):
+		gfx_v9_4_init_lbpw(adev);
 		if (amdgpu_lbpw > 0)
 			gfx_v9_0_enable_lbpw(adev, true);
 		else
@@ -2943,6 +3231,8 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev)
 		break;
 	}
 
+	gfx_v9_0_update_spm_vmid_internal(adev, 0xf);
+
 	adev->gfx.rlc.funcs->start(adev);
 
 	return 0;
@@ -2952,6 +3242,15 @@ static void gfx_v9_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable)
 {
 	u32 tmp = RREG32_SOC15(GC, 0, mmCP_ME_CNTL);
 
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_INVALIDATE_ICACHE, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_INVALIDATE_ICACHE, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_INVALIDATE_ICACHE, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_PIPE0_RESET, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_PIPE1_RESET, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE0_RESET, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE1_RESET, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE0_RESET, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE1_RESET, enable ? 0 : 1);
 	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_HALT, enable ? 0 : 1);
 	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_HALT, enable ? 0 : 1);
 	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_HALT, enable ? 0 : 1);
@@ -3029,6 +3328,14 @@ static int gfx_v9_0_cp_gfx_start(struct amdgpu_device *adev)
 
 	gfx_v9_0_cp_gfx_enable(adev, true);
 
+	/* Now only limit the quirk on the APU gfx9 series and already
+	 * confirmed that the APU gfx10/gfx11 needn't such update.
+	 */
+	if (adev->flags & AMD_IS_APU &&
+			adev->in_s3 && !pm_resume_via_firmware()) {
+		DRM_INFO("Will skip the CSB packet resubmit\n");
+		return 0;
+	}
 	r = amdgpu_ring_alloc(ring, gfx_v9_0_get_csb_size(adev) + 4 + 3);
 	if (r) {
 		DRM_ERROR("amdgpu: cp failed to lock ring (%d).\n", r);
@@ -3106,7 +3413,7 @@ static int gfx_v9_0_cp_gfx_resume(struct amdgpu_device *adev)
 	WREG32_SOC15(GC, 0, mmCP_RB0_WPTR, lower_32_bits(ring->wptr));
 	WREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI, upper_32_bits(ring->wptr));
 
-	/* set the wb address wether it's enabled or not */
+	/* set the wb address whether it's enabled or not */
 	rptr_addr = ring->rptr_gpu_addr;
 	WREG32_SOC15(GC, 0, mmCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr));
 	WREG32_SOC15(GC, 0, mmCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & CP_RB_RPTR_ADDR_HI__RB_RPTR_ADDR_HI_MASK);
@@ -3143,7 +3450,6 @@ static int gfx_v9_0_cp_gfx_resume(struct amdgpu_device *adev)
 
 	/* start the ring */
 	gfx_v9_0_cp_gfx_start(adev);
-	ring->sched.ready = true;
 
 	return 0;
 }
@@ -3154,8 +3460,16 @@ static void gfx_v9_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
 		WREG32_SOC15_RLC(GC, 0, mmCP_MEC_CNTL, 0);
 	} else {
 		WREG32_SOC15_RLC(GC, 0, mmCP_MEC_CNTL,
-			(CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK));
-		adev->gfx.kiq.ring.sched.ready = false;
+				 (CP_MEC_CNTL__MEC_INVALIDATE_ICACHE_MASK |
+				  CP_MEC_CNTL__MEC_ME1_PIPE0_RESET_MASK |
+				  CP_MEC_CNTL__MEC_ME1_PIPE1_RESET_MASK |
+				  CP_MEC_CNTL__MEC_ME1_PIPE2_RESET_MASK |
+				  CP_MEC_CNTL__MEC_ME1_PIPE3_RESET_MASK |
+				  CP_MEC_CNTL__MEC_ME2_PIPE0_RESET_MASK |
+				  CP_MEC_CNTL__MEC_ME2_PIPE1_RESET_MASK |
+				  CP_MEC_CNTL__MEC_ME1_HALT_MASK |
+				  CP_MEC_CNTL__MEC_ME2_HALT_MASK));
+		adev->gfx.kiq[0].ring.sched.ready = false;
 	}
 	udelay(50);
 }
@@ -3212,9 +3526,7 @@ static void gfx_v9_0_kiq_setting(struct amdgpu_ring *ring)
 	tmp = RREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS);
 	tmp &= 0xffffff00;
 	tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
-	WREG32_SOC15_RLC(GC, 0, mmRLC_CP_SCHEDULERS, tmp);
-	tmp |= 0x80;
-	WREG32_SOC15_RLC(GC, 0, mmRLC_CP_SCHEDULERS, tmp);
+	WREG32_SOC15_RLC(GC, 0, mmRLC_CP_SCHEDULERS, tmp | 0x80);
 }
 
 static void gfx_v9_0_mqd_set_priority(struct amdgpu_ring *ring, struct v9_mqd *mqd)
@@ -3519,7 +3831,6 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct v9_mqd *mqd = ring->mqd_ptr;
-	int mqd_idx = AMDGPU_MAX_COMPUTE_RINGS;
 	struct v9_mqd *tmp_mqd;
 
 	gfx_v9_0_kiq_setting(ring);
@@ -3529,20 +3840,20 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring)
 	 * driver need to re-init the mqd.
 	 * check mqd->cp_hqd_pq_control since this value should not be 0
 	 */
-	tmp_mqd = (struct v9_mqd *)adev->gfx.mec.mqd_backup[mqd_idx];
+	tmp_mqd = (struct v9_mqd *)adev->gfx.kiq[0].mqd_backup;
 	if (amdgpu_in_reset(adev) && tmp_mqd->cp_hqd_pq_control){
 		/* for GPU_RESET case , reset MQD to a clean status */
-		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation));
+		if (adev->gfx.kiq[0].mqd_backup)
+			memcpy(mqd, adev->gfx.kiq[0].mqd_backup, sizeof(struct v9_mqd_allocation));
 
 		/* reset ring buffer */
 		ring->wptr = 0;
 		amdgpu_ring_clear_ring(ring);
 
 		mutex_lock(&adev->srbm_mutex);
-		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0);
 		gfx_v9_0_kiq_init_register(ring);
-		soc15_grbm_select(adev, 0, 0, 0, 0);
+		soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 	} else {
 		memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation));
@@ -3551,20 +3862,20 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring)
 		if (amdgpu_sriov_vf(adev) && adev->in_suspend)
 			amdgpu_ring_clear_ring(ring);
 		mutex_lock(&adev->srbm_mutex);
-		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0);
 		gfx_v9_0_mqd_init(ring);
 		gfx_v9_0_kiq_init_register(ring);
-		soc15_grbm_select(adev, 0, 0, 0, 0);
+		soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 
-		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation));
+		if (adev->gfx.kiq[0].mqd_backup)
+			memcpy(adev->gfx.kiq[0].mqd_backup, mqd, sizeof(struct v9_mqd_allocation));
 	}
 
 	return 0;
 }
 
-static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring)
+static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring, bool restore)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct v9_mqd *mqd = ring->mqd_ptr;
@@ -3576,30 +3887,27 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring)
 	 */
 	tmp_mqd = (struct v9_mqd *)adev->gfx.mec.mqd_backup[mqd_idx];
 
-	if (!tmp_mqd->cp_hqd_pq_control ||
-	    (!amdgpu_in_reset(adev) && !adev->in_suspend)) {
+	if (!restore && (!tmp_mqd->cp_hqd_pq_control ||
+	    (!amdgpu_in_reset(adev) && !adev->in_suspend))) {
 		memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation));
 		((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
 		((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF;
 		mutex_lock(&adev->srbm_mutex);
-		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0);
 		gfx_v9_0_mqd_init(ring);
-		soc15_grbm_select(adev, 0, 0, 0, 0);
+		soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
 			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation));
-	} else if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */
-		/* reset MQD to a clean status */
+	} else {
+		/* restore MQD to a clean status */
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
 			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation));
-
 		/* reset ring buffer */
 		ring->wptr = 0;
 		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, 0);
 		amdgpu_ring_clear_ring(ring);
-	} else {
-		amdgpu_ring_clear_ring(ring);
 	}
 
 	return 0;
@@ -3607,54 +3915,23 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring)
 
 static int gfx_v9_0_kiq_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring;
-	int r;
-
-	ring = &adev->gfx.kiq.ring;
-
-	r = amdgpu_bo_reserve(ring->mqd_obj, false);
-	if (unlikely(r != 0))
-		return r;
-
-	r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr);
-	if (unlikely(r != 0))
-		return r;
-
-	gfx_v9_0_kiq_init_queue(ring);
-	amdgpu_bo_kunmap(ring->mqd_obj);
-	ring->mqd_ptr = NULL;
-	amdgpu_bo_unreserve(ring->mqd_obj);
-	ring->sched.ready = true;
+	gfx_v9_0_kiq_init_queue(&adev->gfx.kiq[0].ring);
 	return 0;
 }
 
 static int gfx_v9_0_kcq_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring = NULL;
-	int r = 0, i;
+	int i, r;
 
 	gfx_v9_0_cp_compute_enable(adev, true);
 
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-		ring = &adev->gfx.compute_ring[i];
-
-		r = amdgpu_bo_reserve(ring->mqd_obj, false);
-		if (unlikely(r != 0))
-			goto done;
-		r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr);
-		if (!r) {
-			r = gfx_v9_0_kcq_init_queue(ring);
-			amdgpu_bo_kunmap(ring->mqd_obj);
-			ring->mqd_ptr = NULL;
-		}
-		amdgpu_bo_unreserve(ring->mqd_obj);
+		r = gfx_v9_0_kcq_init_queue(&adev->gfx.compute_ring[i], false);
 		if (r)
-			goto done;
+			return r;
 	}
 
-	r = amdgpu_gfx_enable_kcq(adev);
-done:
-	return r;
+	return amdgpu_gfx_enable_kcq(adev, 0);
 }
 
 static int gfx_v9_0_cp_resume(struct amdgpu_device *adev)
@@ -3678,6 +3955,10 @@ static int gfx_v9_0_cp_resume(struct amdgpu_device *adev)
 			return r;
 	}
 
+	if (adev->gfx.num_gfx_rings)
+		gfx_v9_0_cp_gfx_enable(adev, false);
+	gfx_v9_0_cp_compute_enable(adev, false);
+
 	r = gfx_v9_0_kiq_resume(adev);
 	if (r)
 		return r;
@@ -3713,8 +3994,8 @@ static void gfx_v9_0_init_tcp_config(struct amdgpu_device *adev)
 {
 	u32 tmp;
 
-	if (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 1) &&
-	    adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 2))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 1) &&
+	    amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 2))
 		return;
 
 	tmp = RREG32_SOC15(GC, 0, mmTCP_ADDR_CONFIG);
@@ -3734,10 +4015,13 @@ static void gfx_v9_0_cp_enable(struct amdgpu_device *adev, bool enable)
 	gfx_v9_0_cp_compute_enable(adev, enable);
 }
 
-static int gfx_v9_0_hw_init(void *handle)
+static int gfx_v9_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	amdgpu_gfx_cleaner_shader_init(adev, adev->gfx.cleaner_shader_size,
+				       adev->gfx.cleaner_shader_ptr);
 
 	if (!amdgpu_sriov_vf(adev))
 		gfx_v9_0_init_golden_registers(adev);
@@ -3754,24 +4038,27 @@ static int gfx_v9_0_hw_init(void *handle)
 	if (r)
 		return r;
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2) &&
+	    !amdgpu_sriov_vf(adev))
 		gfx_v9_4_2_set_power_brake_sequence(adev);
 
 	return r;
 }
 
-static int gfx_v9_0_hw_fini(void *handle)
+static int gfx_v9_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	amdgpu_irq_put(adev, &adev->gfx.cp_ecc_error_irq, 0);
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX))
+		amdgpu_irq_put(adev, &adev->gfx.cp_ecc_error_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
+	amdgpu_irq_put(adev, &adev->gfx.bad_op_irq, 0);
 
 	/* DF freeze and kcq disable will fail */
 	if (!amdgpu_ras_intr_triggered())
 		/* disable KCQ to avoid CPC touch memory not valid anymore */
-		amdgpu_gfx_disable_kcq(adev);
+		amdgpu_gfx_disable_kcq(adev, 0);
 
 	if (amdgpu_sriov_vf(adev)) {
 		gfx_v9_0_cp_gfx_enable(adev, false);
@@ -3789,11 +4076,11 @@ static int gfx_v9_0_hw_fini(void *handle)
 	 */
 	if (!amdgpu_in_reset(adev) && !adev->in_suspend) {
 		mutex_lock(&adev->srbm_mutex);
-		soc15_grbm_select(adev, adev->gfx.kiq.ring.me,
-				adev->gfx.kiq.ring.pipe,
-				adev->gfx.kiq.ring.queue, 0);
-		gfx_v9_0_kiq_fini_register(&adev->gfx.kiq.ring);
-		soc15_grbm_select(adev, 0, 0, 0, 0);
+		soc15_grbm_select(adev, adev->gfx.kiq[0].ring.me,
+				adev->gfx.kiq[0].ring.pipe,
+				adev->gfx.kiq[0].ring.queue, 0, 0);
+		gfx_v9_0_kiq_fini_register(&adev->gfx.kiq[0].ring);
+		soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 	}
 
@@ -3801,7 +4088,7 @@ static int gfx_v9_0_hw_fini(void *handle)
 
 	/* Skip stopping RLC with A+A reset or when RLC controls GFX clock */
 	if ((adev->gmc.xgmi.connected_to_cpu && amdgpu_in_reset(adev)) ||
-	    (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(9, 4, 2))) {
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(9, 4, 2))) {
 		dev_dbg(adev->dev, "Skipping RLC halt\n");
 		return 0;
 	}
@@ -3810,19 +4097,19 @@ static int gfx_v9_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int gfx_v9_0_suspend(void *handle)
+static int gfx_v9_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	return gfx_v9_0_hw_fini(handle);
+	return gfx_v9_0_hw_fini(ip_block);
 }
 
-static int gfx_v9_0_resume(void *handle)
+static int gfx_v9_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	return gfx_v9_0_hw_init(handle);
+	return gfx_v9_0_hw_init(ip_block);
 }
 
-static bool gfx_v9_0_is_idle(void *handle)
+static bool gfx_v9_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (REG_GET_FIELD(RREG32_SOC15(GC, 0, mmGRBM_STATUS),
 				GRBM_STATUS, GUI_ACTIVE))
@@ -3831,24 +4118,24 @@ static bool gfx_v9_0_is_idle(void *handle)
 		return true;
 }
 
-static int gfx_v9_0_wait_for_idle(void *handle)
+static int gfx_v9_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		if (gfx_v9_0_is_idle(handle))
+		if (gfx_v9_0_is_idle(ip_block))
 			return 0;
 		udelay(1);
 	}
 	return -ETIMEDOUT;
 }
 
-static int gfx_v9_0_soft_reset(void *handle)
+static int gfx_v9_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 grbm_soft_reset = 0;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* GRBM_STATUS */
 	tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS);
@@ -3887,19 +4174,17 @@ static int gfx_v9_0_soft_reset(void *handle)
 		/* Disable MEC parsing/prefetching */
 		gfx_v9_0_cp_compute_enable(adev, false);
 
-		if (grbm_soft_reset) {
-			tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
-			tmp |= grbm_soft_reset;
-			dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp);
-			WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
-			tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
+		tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
+		tmp |= grbm_soft_reset;
+		dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp);
+		WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
+		tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
 
-			udelay(50);
+		udelay(50);
 
-			tmp &= ~grbm_soft_reset;
-			WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
-			tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
-		}
+		tmp &= ~grbm_soft_reset;
+		WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
+		tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
 
 		/* Wait a little for things to settle down */
 		udelay(50);
@@ -3913,7 +4198,7 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)
 	unsigned long flags;
 	uint32_t seq, reg_val_offs = 0;
 	uint64_t value = 0;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
 	struct amdgpu_ring *ring = &kiq->ring;
 
 	BUG_ON(!ring->funcs->emit_rreg);
@@ -3985,7 +4270,7 @@ static uint64_t gfx_v9_0_get_gpu_clock_counter(struct amdgpu_device *adev)
 {
 	uint64_t clock, clock_lo, clock_hi, hi_check;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 3, 0):
 		preempt_disable();
 		clock_hi = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Renoir);
@@ -4001,40 +4286,12 @@ static uint64_t gfx_v9_0_get_gpu_clock_counter(struct amdgpu_device *adev)
 		preempt_enable();
 		clock = clock_lo | (clock_hi << 32ULL);
 		break;
-	case IP_VERSION(9, 1, 0):
-		preempt_disable();
-		clock_hi = RREG32_SOC15_NO_KIQ(PWR, 0, mmGOLDEN_TSC_COUNT_UPPER_Raven);
-		clock_lo = RREG32_SOC15_NO_KIQ(PWR, 0, mmGOLDEN_TSC_COUNT_LOWER_Raven);
-		hi_check = RREG32_SOC15_NO_KIQ(PWR, 0, mmGOLDEN_TSC_COUNT_UPPER_Raven);
-		/* The PWR TSC clock frequency is 100MHz, which sets 32-bit carry over
-		 * roughly every 42 seconds.
-		 */
-		if (hi_check != clock_hi) {
-			clock_lo = RREG32_SOC15_NO_KIQ(PWR, 0, mmGOLDEN_TSC_COUNT_LOWER_Raven);
-			clock_hi = hi_check;
-		}
-		preempt_enable();
-		clock = clock_lo | (clock_hi << 32ULL);
-		break;
-	case IP_VERSION(9, 2, 2):
-		preempt_disable();
-		clock_hi = RREG32_SOC15_NO_KIQ(PWR, 0, mmGOLDEN_TSC_COUNT_UPPER_Raven2);
-		clock_lo = RREG32_SOC15_NO_KIQ(PWR, 0, mmGOLDEN_TSC_COUNT_LOWER_Raven2);
-		hi_check = RREG32_SOC15_NO_KIQ(PWR, 0, mmGOLDEN_TSC_COUNT_UPPER_Raven2);
-		/* The PWR TSC clock frequency is 100MHz, which sets 32-bit carry over
-		 * roughly every 42 seconds.
-		 */
-		if (hi_check != clock_hi) {
-			clock_lo = RREG32_SOC15_NO_KIQ(PWR, 0, mmGOLDEN_TSC_COUNT_LOWER_Raven2);
-			clock_hi = hi_check;
-		}
-		preempt_enable();
-		clock = clock_lo | (clock_hi << 32ULL);
-		break;
 	default:
 		amdgpu_gfx_off_ctrl(adev, false);
 		mutex_lock(&adev->gfx.gpu_clock_mutex);
-		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 0, 1) && amdgpu_sriov_runtime(adev)) {
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			    IP_VERSION(9, 0, 1) &&
+		    amdgpu_sriov_runtime(adev)) {
 			clock = gfx_v9_0_kiq_read_clock(adev);
 		} else {
 			WREG32_SOC15(GC, 0, mmRLC_CAPTURE_GPU_CLOCK_COUNT, 1);
@@ -4386,7 +4643,7 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev)
 	if (!ring->sched.ready)
 		return 0;
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1)) {
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1)) {
 		vgpr_init_shader_ptr = vgpr_init_compute_shader_arcturus;
 		vgpr_init_shader_size = sizeof(vgpr_init_compute_shader_arcturus);
 		vgpr_init_regs_ptr = vgpr_init_regs_arcturus;
@@ -4526,23 +4783,24 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev)
 	}
 
 fail:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 
 	return r;
 }
 
-static int gfx_v9_0_early_init(void *handle)
+static int gfx_v9_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->gfx.funcs = &gfx_v9_0_gfx_funcs;
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1) ||
-	    adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2))
 		adev->gfx.num_gfx_rings = 0;
 	else
 		adev->gfx.num_gfx_rings = GFX9_NUM_GFX_RINGS;
+	adev->gfx.xcc_mask = 1;
 	adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
 					  AMDGPU_MAX_COMPUTE_RINGS);
 	gfx_v9_0_set_kiq_pm4_funcs(adev);
@@ -4557,9 +4815,9 @@ static int gfx_v9_0_early_init(void *handle)
 	return gfx_v9_0_init_microcode(adev);
 }
 
-static int gfx_v9_0_ecc_late_init(void *handle)
+static int gfx_v9_0_ecc_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	/*
@@ -4576,7 +4834,7 @@ static int gfx_v9_0_ecc_late_init(void *handle)
 	}
 
 	/* requires IBs so do in late init after IB pool is initialized */
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2))
 		r = gfx_v9_4_2_do_edc_gpr_workarounds(adev);
 	else
 		r = gfx_v9_0_do_edc_gpr_workarounds(adev);
@@ -4591,9 +4849,9 @@ static int gfx_v9_0_ecc_late_init(void *handle)
 	return 0;
 }
 
-static int gfx_v9_0_late_init(void *handle)
+static int gfx_v9_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_irq_get(adev, &adev->gfx.priv_reg_irq, 0);
@@ -4604,10 +4862,21 @@ static int gfx_v9_0_late_init(void *handle)
 	if (r)
 		return r;
 
-	r = gfx_v9_0_ecc_late_init(handle);
+	r = amdgpu_irq_get(adev, &adev->gfx.bad_op_irq, 0);
 	if (r)
 		return r;
 
+	r = gfx_v9_0_ecc_late_init(ip_block);
+	if (r)
+		return r;
+
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2))
+		gfx_v9_4_2_debug_trap_config_init(adev,
+			adev->vm_manager.first_kfd_vmid, AMDGPU_NUM_VMID);
+	else
+		gfx_v9_0_debug_trap_config_init(adev,
+			adev->vm_manager.first_kfd_vmid, AMDGPU_NUM_VMID);
+
 	return 0;
 }
 
@@ -4623,7 +4892,7 @@ static bool gfx_v9_0_is_rlc_enabled(struct amdgpu_device *adev)
 	return true;
 }
 
-static void gfx_v9_0_set_safe_mode(struct amdgpu_device *adev)
+static void gfx_v9_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	uint32_t data;
 	unsigned i;
@@ -4640,7 +4909,7 @@ static void gfx_v9_0_set_safe_mode(struct amdgpu_device *adev)
 	}
 }
 
-static void gfx_v9_0_unset_safe_mode(struct amdgpu_device *adev)
+static void gfx_v9_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	uint32_t data;
 
@@ -4651,7 +4920,7 @@ static void gfx_v9_0_unset_safe_mode(struct amdgpu_device *adev)
 static void gfx_v9_0_update_gfx_cg_power_gating(struct amdgpu_device *adev,
 						bool enable)
 {
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 
 	if ((adev->pg_flags & AMD_PG_SUPPORT_GFX_PG) && enable) {
 		gfx_v9_0_enable_gfx_cg_power_gating(adev, true);
@@ -4663,7 +4932,7 @@ static void gfx_v9_0_update_gfx_cg_power_gating(struct amdgpu_device *adev,
 			gfx_v9_0_enable_gfx_pipeline_powergating(adev, false);
 	}
 
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 }
 
 static void gfx_v9_0_update_gfx_mg_power_gating(struct amdgpu_device *adev,
@@ -4690,14 +4959,12 @@ static void gfx_v9_0_update_medium_grain_clock_gating(struct amdgpu_device *adev
 {
 	uint32_t data, def;
 
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
-
 	/* It is disabled by HW by default */
 	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) {
 		/* 1 - RLC_CGTT_MGCG_OVERRIDE */
 		def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE);
 
-		if (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 2, 1))
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 2, 1))
 			data &= ~RLC_CGTT_MGCG_OVERRIDE__CPF_CGTT_SCLK_OVERRIDE_MASK;
 
 		data &= ~(RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK |
@@ -4731,7 +4998,7 @@ static void gfx_v9_0_update_medium_grain_clock_gating(struct amdgpu_device *adev
 		/* 1 - MGCG_OVERRIDE */
 		def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE);
 
-		if (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 2, 1))
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 2, 1))
 			data |= RLC_CGTT_MGCG_OVERRIDE__CPF_CGTT_SCLK_OVERRIDE_MASK;
 
 		data |= (RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK |
@@ -4756,8 +5023,6 @@ static void gfx_v9_0_update_medium_grain_clock_gating(struct amdgpu_device *adev
 			WREG32_SOC15(GC, 0, mmCP_MEM_SLP_CNTL, data);
 		}
 	}
-
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
 }
 
 static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev,
@@ -4768,8 +5033,6 @@ static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev,
 	if (!adev->gfx.num_gfx_rings)
 		return;
 
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
-
 	/* Enable 3D CGCG/CGLS */
 	if (enable) {
 		/* write cmd to clear cgcg/cgls ov */
@@ -4811,8 +5074,6 @@ static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev,
 		if (def != data)
 			WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D, data);
 	}
-
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
 }
 
 static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev,
@@ -4820,8 +5081,6 @@ static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev
 {
 	uint32_t def, data;
 
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
-
 	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)) {
 		def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE);
 		/* unset CGCG override */
@@ -4837,7 +5096,7 @@ static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev
 		/* enable cgcg FSM(0x0000363F) */
 		def = RREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL);
 
-		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1))
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1))
 			data = (0x2000 << RLC_CGCG_CGLS_CTRL__CGCG_GFX_IDLE_THRESHOLD__SHIFT) |
 				RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK;
 		else
@@ -4863,13 +5122,12 @@ static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev
 		if (def != data)
 			WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL, data);
 	}
-
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
 }
 
 static int gfx_v9_0_update_gfx_clock_gating(struct amdgpu_device *adev,
 					    bool enable)
 {
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 	if (enable) {
 		/* CGCG/CGLS should be enabled after MGCG/MGLS
 		 * ===  MGCG + MGLS ===
@@ -4889,15 +5147,15 @@ static int gfx_v9_0_update_gfx_clock_gating(struct amdgpu_device *adev,
 		/* ===  MGCG + MGLS === */
 		gfx_v9_0_update_medium_grain_clock_gating(adev, enable);
 	}
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 	return 0;
 }
 
-static void gfx_v9_0_update_spm_vmid(struct amdgpu_device *adev, unsigned vmid)
+static void gfx_v9_0_update_spm_vmid_internal(struct amdgpu_device *adev,
+					      unsigned int vmid)
 {
 	u32 reg, data;
 
-	amdgpu_gfx_off_ctrl(adev, false);
-
 	reg = SOC15_REG_OFFSET(GC, 0, mmRLC_SPM_MC_CNTL);
 	if (amdgpu_sriov_is_pp_one_vf(adev))
 		data = RREG32_NO_KIQ(reg);
@@ -4911,6 +5169,13 @@ static void gfx_v9_0_update_spm_vmid(struct amdgpu_device *adev, unsigned vmid)
 		WREG32_SOC15_NO_KIQ(GC, 0, mmRLC_SPM_MC_CNTL, data);
 	else
 		WREG32_SOC15(GC, 0, mmRLC_SPM_MC_CNTL, data);
+}
+
+static void gfx_v9_0_update_spm_vmid(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned int vmid)
+{
+	amdgpu_gfx_off_ctrl(adev, false);
+
+	gfx_v9_0_update_spm_vmid_internal(adev, vmid);
 
 	amdgpu_gfx_off_ctrl(adev, true);
 }
@@ -4960,18 +5225,18 @@ static const struct amdgpu_rlc_funcs gfx_v9_0_rlc_funcs = {
 	.is_rlcg_access_range = gfx_v9_0_is_rlcg_access_range,
 };
 
-static int gfx_v9_0_set_powergating_state(void *handle,
+static int gfx_v9_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_PG_STATE_GATE);
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 2, 2):
 	case IP_VERSION(9, 1, 0):
 	case IP_VERSION(9, 3, 0):
 		if (!enable)
-			amdgpu_gfx_off_ctrl(adev, false);
+			amdgpu_gfx_off_ctrl_immediate(adev, false);
 
 		if (adev->pg_flags & AMD_PG_SUPPORT_RLC_SMU_HS) {
 			gfx_v9_0_enable_sck_slow_down_on_power_up(adev, true);
@@ -4993,10 +5258,10 @@ static int gfx_v9_0_set_powergating_state(void *handle,
 		gfx_v9_0_update_gfx_mg_power_gating(adev, enable);
 
 		if (enable)
-			amdgpu_gfx_off_ctrl(adev, true);
+			amdgpu_gfx_off_ctrl_immediate(adev, true);
 		break;
 	case IP_VERSION(9, 2, 1):
-		amdgpu_gfx_off_ctrl(adev, enable);
+		amdgpu_gfx_off_ctrl_immediate(adev, enable);
 		break;
 	default:
 		break;
@@ -5005,15 +5270,15 @@ static int gfx_v9_0_set_powergating_state(void *handle,
 	return 0;
 }
 
-static int gfx_v9_0_set_clockgating_state(void *handle,
+static int gfx_v9_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 2, 1):
 	case IP_VERSION(9, 4, 0):
@@ -5031,9 +5296,9 @@ static int gfx_v9_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static void gfx_v9_0_get_clockgating_state(void *handle, u64 *flags)
+static void gfx_v9_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	if (amdgpu_sriov_vf(adev))
@@ -5063,7 +5328,7 @@ static void gfx_v9_0_get_clockgating_state(void *handle, u64 *flags)
 	if (data & CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK)
 		*flags |= AMD_CG_SUPPORT_GFX_CP_LS | AMD_CG_SUPPORT_GFX_MGLS;
 
-	if (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 1)) {
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 1)) {
 		/* AMD_CG_SUPPORT_GFX_3D_CGCG */
 		data = RREG32_KIQ(SOC15_REG_OFFSET(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D));
 		if (data & RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK)
@@ -5164,7 +5429,8 @@ static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
 			gfx_v9_0_ring_emit_de_meta(ring,
 						   (!amdgpu_sriov_vf(ring->adev) &&
 						   flags & AMDGPU_IB_PREEMPTED) ?
-						   true : false);
+						   true : false,
+						   job->gds_size > 0 && job->gds_base != 0);
 	}
 
 	amdgpu_ring_write(ring, header);
@@ -5175,9 +5441,70 @@ static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
 #endif
 		lower_32_bits(ib->gpu_addr));
 	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
+	amdgpu_ring_ib_on_emit_cntl(ring);
 	amdgpu_ring_write(ring, control);
 }
 
+static void gfx_v9_0_ring_patch_cntl(struct amdgpu_ring *ring,
+				     unsigned offset)
+{
+	u32 control = ring->ring[offset];
+
+	control |= INDIRECT_BUFFER_PRE_RESUME(1);
+	ring->ring[offset] = control;
+}
+
+static void gfx_v9_0_ring_patch_ce_meta(struct amdgpu_ring *ring,
+					unsigned offset)
+{
+	struct amdgpu_device *adev = ring->adev;
+	void *ce_payload_cpu_addr;
+	uint64_t payload_offset, payload_size;
+
+	payload_size = sizeof(struct v9_ce_ib_state);
+
+	payload_offset = offsetof(struct v9_gfx_meta_data, ce_payload);
+	ce_payload_cpu_addr = adev->virt.csa_cpu_addr + payload_offset;
+
+	if (offset + (payload_size >> 2) <= ring->buf_mask + 1) {
+		memcpy((void *)&ring->ring[offset], ce_payload_cpu_addr, payload_size);
+	} else {
+		memcpy((void *)&ring->ring[offset], ce_payload_cpu_addr,
+		       (ring->buf_mask + 1 - offset) << 2);
+		payload_size -= (ring->buf_mask + 1 - offset) << 2;
+		memcpy((void *)&ring->ring[0],
+		       ce_payload_cpu_addr + ((ring->buf_mask + 1 - offset) << 2),
+		       payload_size);
+	}
+}
+
+static void gfx_v9_0_ring_patch_de_meta(struct amdgpu_ring *ring,
+					unsigned offset)
+{
+	struct amdgpu_device *adev = ring->adev;
+	void *de_payload_cpu_addr;
+	uint64_t payload_offset, payload_size;
+
+	payload_size = sizeof(struct v9_de_ib_state);
+
+	payload_offset = offsetof(struct v9_gfx_meta_data, de_payload);
+	de_payload_cpu_addr = adev->virt.csa_cpu_addr + payload_offset;
+
+	((struct v9_de_ib_state *)de_payload_cpu_addr)->ib_completion_status =
+		IB_COMPLETION_STATUS_PREEMPTED;
+
+	if (offset + (payload_size >> 2) <= ring->buf_mask + 1) {
+		memcpy((void *)&ring->ring[offset], de_payload_cpu_addr, payload_size);
+	} else {
+		memcpy((void *)&ring->ring[offset], de_payload_cpu_addr,
+		       (ring->buf_mask + 1 - offset) << 2);
+		payload_size -= (ring->buf_mask + 1 - offset) << 2;
+		memcpy((void *)&ring->ring[0],
+		       de_payload_cpu_addr + ((ring->buf_mask + 1 - offset) << 2),
+		       payload_size);
+	}
+}
+
 static void gfx_v9_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
 					  struct amdgpu_job *job,
 					  struct amdgpu_ib *ib,
@@ -5351,19 +5678,9 @@ static void gfx_v9_0_ring_emit_ce_meta(struct amdgpu_ring *ring, bool resume)
 
 	cnt = (sizeof(ce_payload) >> 2) + 4 - 2;
 
-	if (ring->is_mes_queue) {
-		offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				  gfx[0].gfx_meta_data) +
-			offsetof(struct v9_gfx_meta_data, ce_payload);
-		ce_payload_gpu_addr =
-			amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		ce_payload_cpu_addr =
-			amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-	} else {
-		offset = offsetof(struct v9_gfx_meta_data, ce_payload);
-		ce_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
-		ce_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
-	}
+	offset = offsetof(struct v9_gfx_meta_data, ce_payload);
+	ce_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
+	ce_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, cnt));
 	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(2) |
@@ -5373,6 +5690,8 @@ static void gfx_v9_0_ring_emit_ce_meta(struct amdgpu_ring *ring, bool resume)
 	amdgpu_ring_write(ring, lower_32_bits(ce_payload_gpu_addr));
 	amdgpu_ring_write(ring, upper_32_bits(ce_payload_gpu_addr));
 
+	amdgpu_ring_ib_on_emit_ce(ring);
+
 	if (resume)
 		amdgpu_ring_write_multiple(ring, ce_payload_cpu_addr,
 					   sizeof(ce_payload) >> 2);
@@ -5385,7 +5704,7 @@ static int gfx_v9_0_ring_preempt_ib(struct amdgpu_ring *ring)
 {
 	int i, r = 0;
 	struct amdgpu_device *adev = ring->adev;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
 	struct amdgpu_ring *kiq_ring = &kiq->ring;
 	unsigned long flags;
 
@@ -5406,10 +5725,6 @@ static int gfx_v9_0_ring_preempt_ib(struct amdgpu_ring *ring)
 	amdgpu_ring_alloc(ring, 13);
 	gfx_v9_0_ring_emit_fence(ring, ring->trail_fence_gpu_addr,
 				 ring->trail_seq, AMDGPU_FENCE_FLAG_EXEC | AMDGPU_FENCE_FLAG_INT);
-	/*reset the CP_VMID_PREEMPT after trailing fence*/
-	amdgpu_ring_emit_wreg(ring,
-			      SOC15_REG_OFFSET(GC, 0, mmCP_VMID_PREEMPT),
-			      0x0);
 
 	/* assert IB preemption, emit the trailing fence */
 	kiq->pmf->kiq_unmap_queues(kiq_ring, ring, PREEMPT_QUEUES_NO_UNMAP,
@@ -5432,6 +5747,10 @@ static int gfx_v9_0_ring_preempt_ib(struct amdgpu_ring *ring)
 		DRM_WARN("ring %d timeout to preempt ib\n", ring->idx);
 	}
 
+	/*reset the CP_VMID_PREEMPT after trailing fence*/
+	amdgpu_ring_emit_wreg(ring,
+			      SOC15_REG_OFFSET(GC, 0, mmCP_VMID_PREEMPT),
+			      0x0);
 	amdgpu_ring_commit(ring);
 
 	/* deassert preemption condition */
@@ -5439,7 +5758,7 @@ static int gfx_v9_0_ring_preempt_ib(struct amdgpu_ring *ring)
 	return r;
 }
 
-static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume)
+static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume, bool usegds)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct v9_de_ib_state de_payload = {0};
@@ -5447,31 +5766,18 @@ static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume)
 	void *de_payload_cpu_addr;
 	int cnt;
 
-	if (ring->is_mes_queue) {
-		offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				  gfx[0].gfx_meta_data) +
-			offsetof(struct v9_gfx_meta_data, de_payload);
-		de_payload_gpu_addr =
-			amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		de_payload_cpu_addr =
-			amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-
-		offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				  gfx[0].gds_backup) +
-			offsetof(struct v9_gfx_meta_data, de_payload);
-		gds_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-	} else {
-		offset = offsetof(struct v9_gfx_meta_data, de_payload);
-		de_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
-		de_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
+	offset = offsetof(struct v9_gfx_meta_data, de_payload);
+	de_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
+	de_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
 
-		gds_addr = ALIGN(amdgpu_csa_vaddr(ring->adev) +
-				 AMDGPU_CSA_SIZE - adev->gds.gds_size,
-				 PAGE_SIZE);
-	}
+	gds_addr = ALIGN(amdgpu_csa_vaddr(ring->adev) +
+			 AMDGPU_CSA_SIZE - adev->gds.gds_size,
+			 PAGE_SIZE);
 
-	de_payload.gds_backup_addrlo = lower_32_bits(gds_addr);
-	de_payload.gds_backup_addrhi = upper_32_bits(gds_addr);
+	if (usegds) {
+		de_payload.gds_backup_addrlo = lower_32_bits(gds_addr);
+		de_payload.gds_backup_addrhi = upper_32_bits(gds_addr);
+	}
 
 	cnt = (sizeof(de_payload) >> 2) + 4 - 2;
 	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, cnt));
@@ -5482,6 +5788,7 @@ static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume)
 	amdgpu_ring_write(ring, lower_32_bits(de_payload_gpu_addr));
 	amdgpu_ring_write(ring, upper_32_bits(de_payload_gpu_addr));
 
+	amdgpu_ring_ib_on_emit_de(ring);
 	if (resume)
 		amdgpu_ring_write_multiple(ring, de_payload_cpu_addr,
 					   sizeof(de_payload) >> 2);
@@ -5532,31 +5839,21 @@ static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags)
 	amdgpu_ring_write(ring, 0);
 }
 
-static unsigned gfx_v9_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring)
+static unsigned gfx_v9_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring,
+						  uint64_t addr)
 {
 	unsigned ret;
 	amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, 0); /* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	/* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, 0);
 	ret = ring->wptr & ring->buf_mask;
-	amdgpu_ring_write(ring, 0x55aa55aa); /* patch dummy value later */
+	/* patch dummy value later */
+	amdgpu_ring_write(ring, 0);
 	return ret;
 }
 
-static void gfx_v9_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigned offset)
-{
-	unsigned cur;
-	BUG_ON(offset > ring->buf_mask);
-	BUG_ON(ring->ring[offset] != 0x55aa55aa);
-
-	cur = (ring->wptr - 1) & ring->buf_mask;
-	if (likely(cur > offset))
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->ring_size>>2) - offset + cur;
-}
-
 static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg,
 				    uint32_t reg_val_offs)
 {
@@ -5629,7 +5926,9 @@ static void gfx_v9_0_ring_soft_recovery(struct amdgpu_ring *ring, unsigned vmid)
 	value = REG_SET_FIELD(value, SQ_CMD, MODE, 0x01);
 	value = REG_SET_FIELD(value, SQ_CMD, CHECK_VMID, 1);
 	value = REG_SET_FIELD(value, SQ_CMD, VM_ID, vmid);
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 	WREG32_SOC15(GC, 0, mmSQ_CMD, value);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 }
 
 static void gfx_v9_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev,
@@ -5700,17 +5999,95 @@ static void gfx_v9_0_set_compute_eop_interrupt_state(struct amdgpu_device *adev,
 	}
 }
 
+static u32 gfx_v9_0_get_cpc_int_cntl(struct amdgpu_device *adev,
+				     int me, int pipe)
+{
+	/*
+	 * amdgpu controls only the first MEC. That's why this function only
+	 * handles the setting of interrupts for this specific MEC. All other
+	 * pipes' interrupts are set by amdkfd.
+	 */
+	if (me != 1)
+		return 0;
+
+	switch (pipe) {
+	case 0:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE0_INT_CNTL);
+	case 1:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE1_INT_CNTL);
+	case 2:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE2_INT_CNTL);
+	case 3:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE3_INT_CNTL);
+	default:
+		return 0;
+	}
+}
+
 static int gfx_v9_0_set_priv_reg_fault_state(struct amdgpu_device *adev,
 					     struct amdgpu_irq_src *source,
 					     unsigned type,
 					     enum amdgpu_interrupt_state state)
 {
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
 	case AMDGPU_IRQ_STATE_ENABLE:
 		WREG32_FIELD15(GC, 0, CP_INT_CNTL_RING0,
 			       PRIV_REG_INT_ENABLE,
 			       state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				/* MECs start at 1 */
+				cp_int_cntl_reg = gfx_v9_0_get_cpc_int_cntl(adev, i + 1, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+								    PRIV_REG_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int gfx_v9_0_set_bad_op_fault_state(struct amdgpu_device *adev,
+					   struct amdgpu_irq_src *source,
+					   unsigned type,
+					   enum amdgpu_interrupt_state state)
+{
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+	case AMDGPU_IRQ_STATE_ENABLE:
+		WREG32_FIELD15(GC, 0, CP_INT_CNTL_RING0,
+			       OPCODE_ERROR_INT_ENABLE,
+			       state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				/* MECs start at 1 */
+				cp_int_cntl_reg = gfx_v9_0_get_cpc_int_cntl(adev, i + 1, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+								    OPCODE_ERROR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
 		break;
 	default:
 		break;
@@ -5831,11 +6208,14 @@ static int gfx_v9_0_eop_irq(struct amdgpu_device *adev,
 
 	switch (me_id) {
 	case 0:
-		if (adev->gfx.num_gfx_rings &&
-		    !amdgpu_mcbp_handle_trailing_fence_irq(&adev->gfx.muxer)) {
-			/* Fence signals are handled on the software rings*/
-			for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++)
-				amdgpu_fence_process(&adev->gfx.sw_gfx_ring[i]);
+		if (adev->gfx.num_gfx_rings) {
+			if (!adev->gfx.mcbp) {
+				amdgpu_fence_process(&adev->gfx.gfx_ring[0]);
+			} else if (!amdgpu_mcbp_handle_trailing_fence_irq(&adev->gfx.muxer)) {
+				/* Fence signals are handled on the software rings*/
+				for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++)
+					amdgpu_fence_process(&adev->gfx.sw_gfx_ring[i]);
+			}
 		}
 		break;
 	case 1:
@@ -5889,6 +6269,15 @@ static int gfx_v9_0_priv_reg_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int gfx_v9_0_bad_op_irq(struct amdgpu_device *adev,
+			       struct amdgpu_irq_src *source,
+			       struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal opcode in command stream\n");
+	gfx_v9_0_fault(adev, entry);
+	return 0;
+}
+
 static int gfx_v9_0_priv_inst_irq(struct amdgpu_device *adev,
 				  struct amdgpu_irq_src *source,
 				  struct amdgpu_iv_entry *entry)
@@ -6341,7 +6730,7 @@ static const struct soc15_ras_field_entry gfx_v9_0_ras_fields[] = {
 };
 
 static int gfx_v9_0_ras_error_inject(struct amdgpu_device *adev,
-				     void *inject_if)
+				     void *inject_if, uint32_t instance_mask)
 {
 	struct ras_inject_if *info = (struct ras_inject_if *)inject_if;
 	int ret;
@@ -6380,13 +6769,13 @@ static int gfx_v9_0_ras_error_inject(struct amdgpu_device *adev,
 	block_info.value = info->value;
 
 	mutex_lock(&adev->grbm_idx_mutex);
-	ret = psp_ras_trigger_error(&adev->psp, &block_info);
+	ret = psp_ras_trigger_error(&adev->psp, &block_info, instance_mask);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	return ret;
 }
 
-static const char *vml2_mems[] = {
+static const char * const vml2_mems[] = {
 	"UTC_VML2_BANK_CACHE_0_BIGK_MEM0",
 	"UTC_VML2_BANK_CACHE_0_BIGK_MEM1",
 	"UTC_VML2_BANK_CACHE_0_4K_MEM0",
@@ -6405,7 +6794,7 @@ static const char *vml2_mems[] = {
 	"UTC_VML2_BANK_CACHE_3_4K_MEM1",
 };
 
-static const char *vml2_walker_mems[] = {
+static const char * const vml2_walker_mems[] = {
 	"UTC_VML2_CACHE_PDE0_MEM0",
 	"UTC_VML2_CACHE_PDE0_MEM1",
 	"UTC_VML2_CACHE_PDE1_MEM0",
@@ -6415,7 +6804,7 @@ static const char *vml2_walker_mems[] = {
 	"UTC_VML2_RDIF_LOG_FIFO",
 };
 
-static const char *atc_l2_cache_2m_mems[] = {
+static const char * const atc_l2_cache_2m_mems[] = {
 	"UTC_ATCL2_CACHE_2M_BANK0_WAY0_MEM",
 	"UTC_ATCL2_CACHE_2M_BANK0_WAY1_MEM",
 	"UTC_ATCL2_CACHE_2M_BANK1_WAY0_MEM",
@@ -6608,7 +6997,7 @@ static void gfx_v9_0_reset_ras_error_count(struct amdgpu_device *adev)
 	for (i = 0; i < ARRAY_SIZE(gfx_v9_0_edc_counter_regs); i++) {
 		for (j = 0; j < gfx_v9_0_edc_counter_regs[i].se_num; j++) {
 			for (k = 0; k < gfx_v9_0_edc_counter_regs[i].instance; k++) {
-				amdgpu_gfx_select_se_sh(adev, j, 0x0, k);
+				amdgpu_gfx_select_se_sh(adev, j, 0x0, k, 0);
 				RREG32(SOC15_REG_ENTRY_OFFSET(gfx_v9_0_edc_counter_regs[i]));
 			}
 		}
@@ -6670,7 +7059,7 @@ static void gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev,
 	for (i = 0; i < ARRAY_SIZE(gfx_v9_0_edc_counter_regs); i++) {
 		for (j = 0; j < gfx_v9_0_edc_counter_regs[i].se_num; j++) {
 			for (k = 0; k < gfx_v9_0_edc_counter_regs[i].instance; k++) {
-				amdgpu_gfx_select_se_sh(adev, j, 0, k);
+				amdgpu_gfx_select_se_sh(adev, j, 0, k, 0);
 				reg_value =
 					RREG32(SOC15_REG_ENTRY_OFFSET(gfx_v9_0_edc_counter_regs[i]));
 				if (reg_value)
@@ -6685,7 +7074,7 @@ static void gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev,
 	err_data->ce_count += sec_count;
 	err_data->ue_count += ded_count;
 
-	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	gfx_v9_0_query_utc_edc_status(adev, err_data);
@@ -6769,6 +7158,230 @@ static void gfx_v9_0_emit_wave_limit(struct amdgpu_ring *ring, bool enable)
 	}
 }
 
+static void gfx_v9_ring_insert_nop(struct amdgpu_ring *ring, uint32_t num_nop)
+{
+	/* Header itself is a NOP packet */
+	if (num_nop == 1) {
+		amdgpu_ring_write(ring, ring->funcs->nop);
+		return;
+	}
+
+	/* Max HW optimization till 0x3ffe, followed by remaining one NOP at a time*/
+	amdgpu_ring_write(ring, PACKET3(PACKET3_NOP, min(num_nop - 2, 0x3ffe)));
+
+	/* Header is at index 0, followed by num_nops - 1 NOP packet's */
+	amdgpu_ring_insert_nop(ring, num_nop - 1);
+}
+
+static int gfx_v9_0_reset_kcq(struct amdgpu_ring *ring,
+			      unsigned int vmid,
+			      struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
+	unsigned long flags;
+	int i, r;
+
+	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
+		return -EINVAL;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+
+	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size)) {
+		spin_unlock_irqrestore(&kiq->ring_lock, flags);
+		return -ENOMEM;
+	}
+
+	kiq->pmf->kiq_unmap_queues(kiq_ring, ring, RESET_QUEUES,
+				   0, 0);
+	amdgpu_ring_commit(kiq_ring);
+
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+
+	r = amdgpu_ring_test_ring(kiq_ring);
+	if (r)
+		return r;
+
+	/* make sure dequeue is complete*/
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
+	mutex_lock(&adev->srbm_mutex);
+	soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0);
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (!(RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE) & 1))
+			break;
+		udelay(1);
+	}
+	if (i >= adev->usec_timeout)
+		r = -ETIMEDOUT;
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
+	if (r) {
+		dev_err(adev->dev, "fail to wait on hqd deactive\n");
+		return r;
+	}
+
+	r = gfx_v9_0_kcq_init_queue(ring, true);
+	if (r) {
+		dev_err(adev->dev, "fail to init kcq\n");
+		return r;
+	}
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+	r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size);
+	if (r) {
+		spin_unlock_irqrestore(&kiq->ring_lock, flags);
+		return -ENOMEM;
+	}
+	kiq->pmf->kiq_map_queues(kiq_ring, ring);
+	amdgpu_ring_commit(kiq_ring);
+	r = amdgpu_ring_test_ring(kiq_ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+	if (r) {
+		DRM_ERROR("fail to remap queue\n");
+		return r;
+	}
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static void gfx_v9_ip_print(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k, reg, index = 0;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_9);
+
+	if (!adev->gfx.ip_dump_core)
+		return;
+
+	for (i = 0; i < reg_count; i++)
+		drm_printf(p, "%-50s \t 0x%08x\n",
+			   gc_reg_list_9[i].reg_name,
+			   adev->gfx.ip_dump_core[i]);
+
+	/* print compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_9);
+	drm_printf(p, "\nnum_mec: %d num_pipe: %d num_queue: %d\n",
+		   adev->gfx.mec.num_mec,
+		   adev->gfx.mec.num_pipe_per_mec,
+		   adev->gfx.mec.num_queue_per_pipe);
+
+	for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+		for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+			for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+				drm_printf(p, "\nmec %d, pipe %d, queue %d\n", i, j, k);
+				for (reg = 0; reg < reg_count; reg++) {
+					if (i && gc_cp_reg_list_9[reg].reg_offset == mmCP_MEC_ME1_HEADER_DUMP)
+						drm_printf(p, "%-50s \t 0x%08x\n",
+							   "mmCP_MEC_ME2_HEADER_DUMP",
+							   adev->gfx.ip_dump_compute_queues[index + reg]);
+					else
+						drm_printf(p, "%-50s \t 0x%08x\n",
+							   gc_cp_reg_list_9[reg].reg_name,
+							   adev->gfx.ip_dump_compute_queues[index + reg]);
+				}
+				index += reg_count;
+			}
+		}
+	}
+
+}
+
+static void gfx_v9_ip_dump(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k, reg, index = 0;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_9);
+
+	if (!adev->gfx.ip_dump_core || !adev->gfx.num_gfx_rings)
+		return;
+
+	amdgpu_gfx_off_ctrl(adev, false);
+	for (i = 0; i < reg_count; i++)
+		adev->gfx.ip_dump_core[i] = RREG32(SOC15_REG_ENTRY_OFFSET(gc_reg_list_9[i]));
+	amdgpu_gfx_off_ctrl(adev, true);
+
+	/* dump compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_9);
+	amdgpu_gfx_off_ctrl(adev, false);
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+		for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+			for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+				/* ME0 is for GFX so start from 1 for CP */
+				soc15_grbm_select(adev, 1 + i, j, k, 0, 0);
+
+				for (reg = 0; reg < reg_count; reg++) {
+					if (i && gc_cp_reg_list_9[reg].reg_offset == mmCP_MEC_ME1_HEADER_DUMP)
+						adev->gfx.ip_dump_compute_queues[index + reg] =
+							RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEC_ME2_HEADER_DUMP));
+					else
+						adev->gfx.ip_dump_compute_queues[index + reg] =
+							RREG32(SOC15_REG_ENTRY_OFFSET(
+								       gc_cp_reg_list_9[reg]));
+				}
+				index += reg_count;
+			}
+		}
+	}
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	amdgpu_gfx_off_ctrl(adev, true);
+
+}
+
+static void gfx_v9_0_ring_emit_cleaner_shader(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	/* Emit the cleaner shader */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2))
+		amdgpu_ring_write(ring, PACKET3(PACKET3_RUN_CLEANER_SHADER, 0));
+	else
+		amdgpu_ring_write(ring, PACKET3(PACKET3_RUN_CLEANER_SHADER_9_0, 0));
+
+	amdgpu_ring_write(ring, 0);  /* RESERVED field, programmed to zero */
+}
+
+static void gfx_v9_0_ring_begin_use_compute(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_ip_block *gfx_block =
+		amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GFX);
+
+	amdgpu_gfx_enforce_isolation_ring_begin_use(ring);
+
+	/* Raven and PCO APUs seem to have stability issues
+	 * with compute and gfxoff and gfx pg.  Disable gfx pg during
+	 * submission and allow again afterwards.
+	 */
+	if (gfx_block && amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 1, 0))
+		gfx_v9_0_set_powergating_state(gfx_block, AMD_PG_STATE_UNGATE);
+}
+
+static void gfx_v9_0_ring_end_use_compute(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_ip_block *gfx_block =
+		amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GFX);
+
+	/* Raven and PCO APUs seem to have stability issues
+	 * with compute and gfxoff and gfx pg.  Disable gfx pg during
+	 * submission and allow again afterwards.
+	 */
+	if (gfx_block && amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 1, 0))
+		gfx_v9_0_set_powergating_state(gfx_block, AMD_PG_STATE_GATE);
+
+	amdgpu_gfx_enforce_isolation_ring_end_use(ring);
+}
+
 static const struct amd_ip_funcs gfx_v9_0_ip_funcs = {
 	.name = "gfx_v9_0",
 	.early_init = gfx_v9_0_early_init,
@@ -6785,6 +7398,8 @@ static const struct amd_ip_funcs gfx_v9_0_ip_funcs = {
 	.set_clockgating_state = gfx_v9_0_set_clockgating_state,
 	.set_powergating_state = gfx_v9_0_set_powergating_state,
 	.get_clockgating_state = gfx_v9_0_get_clockgating_state,
+	.dump_ip_state = gfx_v9_ip_dump,
+	.print_ip_state = gfx_v9_ip_print,
 };
 
 static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = {
@@ -6816,7 +7431,8 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = {
 		5 + /* HDP_INVL */
 		8 + 8 + /* FENCE x2 */
 		2 + /* SWITCH_BUFFER */
-		7, /* gfx_v9_0_emit_mem_sync */
+		7 + /* gfx_v9_0_emit_mem_sync */
+		2, /* gfx_v9_0_ring_emit_cleaner_shader */
 	.emit_ib_size =	4, /* gfx_v9_0_ring_emit_ib_gfx */
 	.emit_ib = gfx_v9_0_ring_emit_ib_gfx,
 	.emit_fence = gfx_v9_0_ring_emit_fence,
@@ -6825,12 +7441,11 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = {
 	.emit_gds_switch = gfx_v9_0_ring_emit_gds_switch,
 	.emit_hdp_flush = gfx_v9_0_ring_emit_hdp_flush,
 	.test_ring = gfx_v9_0_ring_test_ring,
-	.insert_nop = amdgpu_ring_insert_nop,
+	.insert_nop = gfx_v9_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_switch_buffer = gfx_v9_ring_emit_sb,
 	.emit_cntxcntl = gfx_v9_ring_emit_cntxcntl,
 	.init_cond_exec = gfx_v9_0_ring_emit_init_cond_exec,
-	.patch_cond_exec = gfx_v9_0_ring_emit_patch_cond_exec,
 	.preempt_ib = gfx_v9_0_ring_preempt_ib,
 	.emit_frame_cntl = gfx_v9_0_ring_emit_frame_cntl,
 	.emit_wreg = gfx_v9_0_ring_emit_wreg,
@@ -6838,6 +7453,9 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = {
 	.emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait,
 	.soft_recovery = gfx_v9_0_ring_soft_recovery,
 	.emit_mem_sync = gfx_v9_0_emit_mem_sync,
+	.emit_cleaner_shader = gfx_v9_0_ring_emit_cleaner_shader,
+	.begin_use = amdgpu_gfx_enforce_isolation_ring_begin_use,
+	.end_use = amdgpu_gfx_enforce_isolation_ring_end_use,
 };
 
 static const struct amdgpu_ring_funcs gfx_v9_0_sw_ring_funcs_gfx = {
@@ -6870,7 +7488,8 @@ static const struct amdgpu_ring_funcs gfx_v9_0_sw_ring_funcs_gfx = {
 		5 + /* HDP_INVL */
 		8 + 8 + /* FENCE x2 */
 		2 + /* SWITCH_BUFFER */
-		7, /* gfx_v9_0_emit_mem_sync */
+		7 + /* gfx_v9_0_emit_mem_sync */
+		2, /* gfx_v9_0_ring_emit_cleaner_shader */
 	.emit_ib_size =	4, /* gfx_v9_0_ring_emit_ib_gfx */
 	.emit_ib = gfx_v9_0_ring_emit_ib_gfx,
 	.emit_fence = gfx_v9_0_ring_emit_fence,
@@ -6880,18 +7499,23 @@ static const struct amdgpu_ring_funcs gfx_v9_0_sw_ring_funcs_gfx = {
 	.emit_hdp_flush = gfx_v9_0_ring_emit_hdp_flush,
 	.test_ring = gfx_v9_0_ring_test_ring,
 	.test_ib = gfx_v9_0_ring_test_ib,
-	.insert_nop = amdgpu_sw_ring_insert_nop,
+	.insert_nop = gfx_v9_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_switch_buffer = gfx_v9_ring_emit_sb,
 	.emit_cntxcntl = gfx_v9_ring_emit_cntxcntl,
 	.init_cond_exec = gfx_v9_0_ring_emit_init_cond_exec,
-	.patch_cond_exec = gfx_v9_0_ring_emit_patch_cond_exec,
 	.emit_frame_cntl = gfx_v9_0_ring_emit_frame_cntl,
 	.emit_wreg = gfx_v9_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v9_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait,
 	.soft_recovery = gfx_v9_0_ring_soft_recovery,
 	.emit_mem_sync = gfx_v9_0_emit_mem_sync,
+	.patch_cntl = gfx_v9_0_ring_patch_cntl,
+	.patch_de = gfx_v9_0_ring_patch_de_meta,
+	.patch_ce = gfx_v9_0_ring_patch_ce_meta,
+	.emit_cleaner_shader = gfx_v9_0_ring_emit_cleaner_shader,
+	.begin_use = amdgpu_gfx_enforce_isolation_ring_begin_use,
+	.end_use = amdgpu_gfx_enforce_isolation_ring_end_use,
 };
 
 static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = {
@@ -6909,11 +7533,11 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = {
 		7 + /* gfx_v9_0_ring_emit_pipeline_sync */
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
-		2 + /* gfx_v9_0_ring_emit_vm_flush */
 		8 + 8 + 8 + /* gfx_v9_0_ring_emit_fence x3 for user fence, vm fence */
 		7 + /* gfx_v9_0_emit_mem_sync */
 		5 + /* gfx_v9_0_emit_wave_limit for updating mmSPI_WCL_PIPE_PERCENT_GFX register */
-		15, /* for updating 3 mmSPI_WCL_PIPE_PERCENT_CS registers */
+		15 + /* for updating 3 mmSPI_WCL_PIPE_PERCENT_CS registers */
+		2, /* gfx_v9_0_ring_emit_cleaner_shader */
 	.emit_ib_size =	7, /* gfx_v9_0_ring_emit_ib_compute */
 	.emit_ib = gfx_v9_0_ring_emit_ib_compute,
 	.emit_fence = gfx_v9_0_ring_emit_fence,
@@ -6923,13 +7547,18 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = {
 	.emit_hdp_flush = gfx_v9_0_ring_emit_hdp_flush,
 	.test_ring = gfx_v9_0_ring_test_ring,
 	.test_ib = gfx_v9_0_ring_test_ib,
-	.insert_nop = amdgpu_ring_insert_nop,
+	.insert_nop = gfx_v9_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_wreg = gfx_v9_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v9_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait,
+	.soft_recovery = gfx_v9_0_ring_soft_recovery,
 	.emit_mem_sync = gfx_v9_0_emit_mem_sync,
 	.emit_wave_limit = gfx_v9_0_emit_wave_limit,
+	.reset = gfx_v9_0_reset_kcq,
+	.emit_cleaner_shader = gfx_v9_0_ring_emit_cleaner_shader,
+	.begin_use = gfx_v9_0_ring_begin_use_compute,
+	.end_use = gfx_v9_0_ring_end_use_compute,
 };
 
 static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = {
@@ -6947,7 +7576,6 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = {
 		7 + /* gfx_v9_0_ring_emit_pipeline_sync */
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
-		2 + /* gfx_v9_0_ring_emit_vm_flush */
 		8 + 8 + 8, /* gfx_v9_0_ring_emit_fence_kiq x3 for user fence, vm fence */
 	.emit_ib_size =	7, /* gfx_v9_0_ring_emit_ib_compute */
 	.emit_fence = gfx_v9_0_ring_emit_fence_kiq,
@@ -6958,18 +7586,19 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = {
 	.emit_wreg = gfx_v9_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v9_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait,
+	.emit_hdp_flush = gfx_v9_0_ring_emit_hdp_flush,
 };
 
 static void gfx_v9_0_set_ring_funcs(struct amdgpu_device *adev)
 {
 	int i;
 
-	adev->gfx.kiq.ring.funcs = &gfx_v9_0_ring_funcs_kiq;
+	adev->gfx.kiq[0].ring.funcs = &gfx_v9_0_ring_funcs_kiq;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
 		adev->gfx.gfx_ring[i].funcs = &gfx_v9_0_ring_funcs_gfx;
 
-	if (adev->gfx.num_gfx_rings) {
+	if (adev->gfx.mcbp && adev->gfx.num_gfx_rings) {
 		for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++)
 			adev->gfx.sw_gfx_ring[i].funcs = &gfx_v9_0_sw_ring_funcs_gfx;
 	}
@@ -6988,6 +7617,11 @@ static const struct amdgpu_irq_src_funcs gfx_v9_0_priv_reg_irq_funcs = {
 	.process = gfx_v9_0_priv_reg_irq,
 };
 
+static const struct amdgpu_irq_src_funcs gfx_v9_0_bad_op_irq_funcs = {
+	.set = gfx_v9_0_set_bad_op_fault_state,
+	.process = gfx_v9_0_bad_op_irq,
+};
+
 static const struct amdgpu_irq_src_funcs gfx_v9_0_priv_inst_irq_funcs = {
 	.set = gfx_v9_0_set_priv_inst_fault_state,
 	.process = gfx_v9_0_priv_inst_irq,
@@ -7007,6 +7641,9 @@ static void gfx_v9_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->gfx.priv_reg_irq.num_types = 1;
 	adev->gfx.priv_reg_irq.funcs = &gfx_v9_0_priv_reg_irq_funcs;
 
+	adev->gfx.bad_op_irq.num_types = 1;
+	adev->gfx.bad_op_irq.funcs = &gfx_v9_0_bad_op_irq_funcs;
+
 	adev->gfx.priv_inst_irq.num_types = 1;
 	adev->gfx.priv_inst_irq.funcs = &gfx_v9_0_priv_inst_irq_funcs;
 
@@ -7016,7 +7653,7 @@ static void gfx_v9_0_set_irq_funcs(struct amdgpu_device *adev)
 
 static void gfx_v9_0_set_rlc_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 2, 1):
 	case IP_VERSION(9, 4, 0):
@@ -7035,7 +7672,7 @@ static void gfx_v9_0_set_rlc_funcs(struct amdgpu_device *adev)
 static void gfx_v9_0_set_gds_init(struct amdgpu_device *adev)
 {
 	/* init asci gds info */
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 2, 1):
 	case IP_VERSION(9, 4, 0):
@@ -7057,7 +7694,7 @@ static void gfx_v9_0_set_gds_init(struct amdgpu_device *adev)
 		break;
 	}
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 4, 0):
 		adev->gds.gds_compute_max_wave_id = 0x7ff;
@@ -7145,7 +7782,7 @@ static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev,
 			mask = 1;
 			ao_bitmap = 0;
 			counter = 0;
-			amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff);
+			amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff, 0);
 			gfx_v9_0_set_user_cu_inactive_bitmap(
 				adev, disable_masks[i * adev->gfx.config.max_sh_per_se + j]);
 			bitmap = gfx_v9_0_get_cu_active_bitmap(adev);
@@ -7162,7 +7799,7 @@ static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev,
 			 *    SE6,SH0 --> bitmap[2][1]
 			 *    SE7,SH0 --> bitmap[3][1]
 			 */
-			cu_info->bitmap[i % 4][j + i / 4] = bitmap;
+			cu_info->bitmap[0][i % 4][j + i / 4] = bitmap;
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k ++) {
 				if (bitmap & mask) {
@@ -7178,7 +7815,7 @@ static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev,
 			cu_info->ao_cu_bitmap[i % 4][j + i / 4] = ao_bitmap;
 		}
 	}
-	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	cu_info->number = active_cu_number;
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.h b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.h
index dfe8d4841f58..f9f6edc5e558 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.h
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.h
@@ -27,6 +27,6 @@
 extern const struct amdgpu_ip_block_version gfx_v9_0_ip_block;
 
 void gfx_v9_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num,
-			   u32 instance);
+			   u32 instance, int xcc_id);
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0_cleaner_shader.h b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0_cleaner_shader.h
new file mode 100644
index 000000000000..0b6bd09b7529
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0_cleaner_shader.h
@@ -0,0 +1,68 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+/* Define the cleaner shader gfx_9_0 */
+static const u32 __maybe_unused gfx_9_0_cleaner_shader_hex[] = {
+	/* Add the cleaner shader code here */
+};
+
+/* Define the cleaner shader gfx_9_4_2 */
+static const u32 gfx_9_4_2_cleaner_shader_hex[] = {
+	0xbf068100, 0xbf84003b,
+	0xbf8a0000, 0xb07c0000,
+	0xbe8200ff, 0x00000078,
+	0xbf110802, 0x7e000280,
+	0x7e020280, 0x7e040280,
+	0x7e060280, 0x7e080280,
+	0x7e0a0280, 0x7e0c0280,
+	0x7e0e0280, 0x80828802,
+	0xbe803202, 0xbf84fff5,
+	0xbf9c0000, 0xbe8200ff,
+	0x80000000, 0x86020102,
+	0xbf840011, 0xbefe00c1,
+	0xbeff00c1, 0xd28c0001,
+	0x0001007f, 0xd28d0001,
+	0x0002027e, 0x10020288,
+	0xbe8200bf, 0xbefc00c1,
+	0xd89c2000, 0x00020201,
+	0xd89c6040, 0x00040401,
+	0x320202ff, 0x00000400,
+	0x80828102, 0xbf84fff8,
+	0xbefc00ff, 0x0000005c,
+	0xbf800000, 0xbe802c80,
+	0xbe812c80, 0xbe822c80,
+	0xbe832c80, 0x80fc847c,
+	0xbf84fffa, 0xbee60080,
+	0xbee70080, 0xbeea0180,
+	0xbeec0180, 0xbeee0180,
+	0xbef00180, 0xbef20180,
+	0xbef40180, 0xbef60180,
+	0xbef80180, 0xbefa0180,
+	0xbf810000, 0xbf8d0001,
+	0xbefc00ff, 0x0000005c,
+	0xbf800000, 0xbe802c80,
+	0xbe812c80, 0xbe822c80,
+	0xbe832c80, 0x80fc847c,
+	0xbf84fffa, 0xbee60080,
+	0xbee70080, 0xbeea01ff,
+	0x000000ee, 0xbf810000,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c
index c67e387a97f5..6028afd81690 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c
@@ -27,7 +27,6 @@
 #include "amdgpu_gfx.h"
 #include "soc15.h"
 #include "soc15d.h"
-#include "amdgpu_atomfirmware.h"
 #include "amdgpu_pm.h"
 
 #include "gc/gc_9_4_1_offset.h"
@@ -970,31 +969,9 @@ static void gfx_v9_4_reset_ras_error_count(struct amdgpu_device *adev)
 	WREG32_SOC15(GC, 0, mmATC_L2_CACHE_4K_DSM_INDEX, 255);
 }
 
-static int gfx_v9_4_ras_error_inject(struct amdgpu_device *adev,
-				     void *inject_if)
-{
-	struct ras_inject_if *info = (struct ras_inject_if *)inject_if;
-	int ret;
-	struct ta_ras_trigger_error_input block_info = { 0 };
-
-	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX))
-		return -EINVAL;
-
-	block_info.block_id = amdgpu_ras_block_to_ta(info->head.block);
-	block_info.sub_block_index = info->head.sub_block_index;
-	block_info.inject_error_type = amdgpu_ras_error_to_ta(info->head.type);
-	block_info.address = info->address;
-	block_info.value = info->value;
-
-	mutex_lock(&adev->grbm_idx_mutex);
-	ret = psp_ras_trigger_error(&adev->psp, &block_info);
-	mutex_unlock(&adev->grbm_idx_mutex);
-
-	return ret;
-}
-
-static const struct soc15_reg_entry gfx_v9_4_ea_err_status_regs =
-	{ SOC15_REG_ENTRY(GC, 0, mmGCEA_ERR_STATUS), 0, 1, 32 };
+static const struct soc15_reg_entry gfx_v9_4_ea_err_status_regs = {
+	SOC15_REG_ENTRY(GC, 0, mmGCEA_ERR_STATUS), 0, 1, 32
+};
 
 static void gfx_v9_4_query_ras_error_status(struct amdgpu_device *adev)
 {
@@ -1030,7 +1007,6 @@ static void gfx_v9_4_query_ras_error_status(struct amdgpu_device *adev)
 
 
 const struct amdgpu_ras_block_hw_ops  gfx_v9_4_ras_ops = {
-	.ras_error_inject = &gfx_v9_4_ras_error_inject,
 	.query_ras_error_count = &gfx_v9_4_query_ras_error_count,
 	.reset_ras_error_count = &gfx_v9_4_reset_ras_error_count,
 	.query_ras_error_status = &gfx_v9_4_query_ras_error_status,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c
index 3a797424579c..8058ea91ecaf 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c
@@ -412,7 +412,7 @@ static int gfx_v9_4_2_run_shader(struct amdgpu_device *adev,
 	r = amdgpu_ib_schedule(ring, 1, ib, NULL, fence_ptr);
 	if (r) {
 		dev_err(adev->dev, "ib submit failed (%d).\n", r);
-		amdgpu_ib_free(adev, ib, NULL);
+		amdgpu_ib_free(ib, NULL);
 	}
 	return r;
 }
@@ -611,16 +611,16 @@ static int gfx_v9_4_2_do_sgprs_init(struct amdgpu_device *adev)
 	}
 
 disp2_failed:
-	amdgpu_ib_free(adev, &disp_ibs[2], NULL);
+	amdgpu_ib_free(&disp_ibs[2], NULL);
 	dma_fence_put(fences[2]);
 disp1_failed:
-	amdgpu_ib_free(adev, &disp_ibs[1], NULL);
+	amdgpu_ib_free(&disp_ibs[1], NULL);
 	dma_fence_put(fences[1]);
 disp0_failed:
-	amdgpu_ib_free(adev, &disp_ibs[0], NULL);
+	amdgpu_ib_free(&disp_ibs[0], NULL);
 	dma_fence_put(fences[0]);
 pro_end:
-	amdgpu_ib_free(adev, &wb_ib, NULL);
+	amdgpu_ib_free(&wb_ib, NULL);
 
 	if (r)
 		dev_info(adev->dev, "Init SGPRS Failed\n");
@@ -687,10 +687,10 @@ static int gfx_v9_4_2_do_vgprs_init(struct amdgpu_device *adev)
 	}
 
 disp_failed:
-	amdgpu_ib_free(adev, &disp_ib, NULL);
+	amdgpu_ib_free(&disp_ib, NULL);
 	dma_fence_put(fence);
 pro_end:
-	amdgpu_ib_free(adev, &wb_ib, NULL);
+	amdgpu_ib_free(&wb_ib, NULL);
 
 	if (r)
 		dev_info(adev->dev, "Init VGPRS Failed\n");
@@ -746,8 +746,18 @@ void gfx_v9_4_2_init_golden_registers(struct amdgpu_device *adev,
 			 die_id);
 		break;
 	}
+}
 
-	return;
+void gfx_v9_4_2_init_sq(struct amdgpu_device *adev)
+{
+	uint32_t data;
+
+	if (adev->gfx.mec_fw_version >= 98) {
+		adev->gmc.xnack_flags |= AMDGPU_GMC_XNACK_FLAG_CHAIN;
+		data = RREG32_SOC15(GC, 0, regSQ_CONFIG1);
+		data = REG_SET_FIELD(data, SQ_CONFIG1, DISABLE_XNACK_CHECK_IN_RETRY_DISABLE, 1);
+		WREG32_SOC15(GC, 0, regSQ_CONFIG1, data);
+	}
 }
 
 void gfx_v9_4_2_debug_trap_config_init(struct amdgpu_device *adev,
@@ -761,7 +771,7 @@ void gfx_v9_4_2_debug_trap_config_init(struct amdgpu_device *adev,
 
 	for (i = first_vmid; i < last_vmid; i++) {
 		data = 0;
-		soc15_grbm_select(adev, 0, 0, 0, i);
+		soc15_grbm_select(adev, 0, 0, 0, i, 0);
 		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
 		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, 0);
 		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE,
@@ -769,15 +779,18 @@ void gfx_v9_4_2_debug_trap_config_init(struct amdgpu_device *adev,
 		WREG32(SOC15_REG_OFFSET(GC, 0, regSPI_GDBG_PER_VMID_CNTL), data);
 	}
 
-	soc15_grbm_select(adev, 0, 0, 0, 0);
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 	mutex_unlock(&adev->srbm_mutex);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, regSPI_GDBG_TRAP_DATA0), 0);
+	WREG32(SOC15_REG_OFFSET(GC, 0, regSPI_GDBG_TRAP_DATA1), 0);
 }
 
 void gfx_v9_4_2_set_power_brake_sequence(struct amdgpu_device *adev)
 {
 	u32 tmp;
 
-	gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 
 	tmp = 0;
 	tmp = REG_SET_FIELD(tmp, GC_THROTTLE_CTRL, PATTERN_MODE, 1);
@@ -1545,8 +1558,8 @@ static void gfx_v9_4_2_log_utc_edc_count(struct amdgpu_device *adev,
 					 uint32_t ded_cnt)
 {
 	uint32_t bank, way, mem;
-	static const char *vml2_way_str[] = { "BIGK", "4K" };
-	static const char *utcl2_rounter_str[] = { "VMC", "APT" };
+	static const char * const vml2_way_str[] = { "BIGK", "4K" };
+	static const char * const utcl2_router_str[] = { "VMC", "APT" };
 
 	mem = instance % blk->num_mem_blocks;
 	way = (instance / blk->num_mem_blocks) % blk->num_ways;
@@ -1567,7 +1580,7 @@ static void gfx_v9_4_2_log_utc_edc_count(struct amdgpu_device *adev,
 		dev_info(
 			adev->dev,
 			"GFX SubBlock UTCL2_ROUTER_IFIF%d_GROUP0_%s, SED %d, DED %d\n",
-			bank, utcl2_rounter_str[mem], sec_cnt, ded_cnt);
+			bank, utcl2_router_str[mem], sec_cnt, ded_cnt);
 		break;
 	case ATC_L2_CACHE_2M:
 		dev_info(
@@ -1699,28 +1712,6 @@ static void gfx_v9_4_2_reset_ras_error_count(struct amdgpu_device *adev)
 	gfx_v9_4_2_query_utc_edc_count(adev, NULL, NULL);
 }
 
-static int gfx_v9_4_2_ras_error_inject(struct amdgpu_device *adev, void *inject_if)
-{
-	struct ras_inject_if *info = (struct ras_inject_if *)inject_if;
-	int ret;
-	struct ta_ras_trigger_error_input block_info = { 0 };
-
-	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX))
-		return -EINVAL;
-
-	block_info.block_id = amdgpu_ras_block_to_ta(info->head.block);
-	block_info.sub_block_index = info->head.sub_block_index;
-	block_info.inject_error_type = amdgpu_ras_error_to_ta(info->head.type);
-	block_info.address = info->address;
-	block_info.value = info->value;
-
-	mutex_lock(&adev->grbm_idx_mutex);
-	ret = psp_ras_trigger_error(&adev->psp, &block_info);
-	mutex_unlock(&adev->grbm_idx_mutex);
-
-	return ret;
-}
-
 static void gfx_v9_4_2_query_ea_err_status(struct amdgpu_device *adev)
 {
 	uint32_t i, j;
@@ -1930,21 +1921,9 @@ static void gfx_v9_4_2_reset_sq_timeout_status(struct amdgpu_device *adev)
 	mutex_unlock(&adev->grbm_idx_mutex);
 }
 
-static bool gfx_v9_4_2_query_uctl2_poison_status(struct amdgpu_device *adev)
-{
-	u32 status = 0;
-	struct amdgpu_vmhub *hub;
-
-	hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	status = RREG32(hub->vm_l2_pro_fault_status);
-	/* reset page fault status */
-	WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1);
 
-	return REG_GET_FIELD(status, VM_L2_PROTECTION_FAULT_STATUS, FED);
-}
 
 struct amdgpu_ras_block_hw_ops  gfx_v9_4_2_ras_ops = {
-		.ras_error_inject = &gfx_v9_4_2_ras_error_inject,
 		.query_ras_error_count = &gfx_v9_4_2_query_ras_error_count,
 		.reset_ras_error_count = &gfx_v9_4_2_reset_ras_error_count,
 		.query_ras_error_status = &gfx_v9_4_2_query_ras_error_status,
@@ -1956,5 +1935,4 @@ struct amdgpu_gfx_ras gfx_v9_4_2_ras = {
 		.hw_ops = &gfx_v9_4_2_ras_ops,
 	},
 	.enable_watchdog_timer = &gfx_v9_4_2_enable_watchdog_timer,
-	.query_utcl2_poison_status = gfx_v9_4_2_query_uctl2_poison_status,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h
index 7584624b641c..a603724c1dfc 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h
@@ -28,6 +28,7 @@ void gfx_v9_4_2_debug_trap_config_init(struct amdgpu_device *adev,
 				uint32_t first_vmid, uint32_t last_vmid);
 void gfx_v9_4_2_init_golden_registers(struct amdgpu_device *adev,
 				      uint32_t die_id);
+void gfx_v9_4_2_init_sq(struct amdgpu_device *adev);
 void gfx_v9_4_2_set_power_brake_sequence(struct amdgpu_device *adev);
 int gfx_v9_4_2_do_edc_gpr_workarounds(struct amdgpu_device *adev);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2_cleaner_shader.asm b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2_cleaner_shader.asm
new file mode 100644
index 000000000000..35b8cf9070bd
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2_cleaner_shader.asm
@@ -0,0 +1,153 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+// This shader is to clean LDS, SGPRs and VGPRs. It is  first 64 Dwords or 256 bytes of 192 Dwords cleaner shader.
+//To turn this shader program on for complitaion change this to main and lower shader main to main_1
+
+// MI200 : Clear SGPRs, VGPRs and LDS
+//   Uses two kernels launched separately:
+//   1. Clean VGPRs, LDS, and lower SGPRs
+//        Launches one workgroup per CU, each workgroup with 4x wave64 per SIMD in the CU
+//        Waves are "wave64" and have 128 VGPRs each, which uses all 512 VGPRs per SIMD
+//        Waves in the workgroup share the 64KB of LDS
+//        Each wave clears SGPRs 0 - 95. Because there are 4 waves/SIMD, this is physical SGPRs 0-383
+//        Each wave clears 128 VGPRs, so all 512 in the SIMD
+//        The first wave of the workgroup clears its 64KB of LDS
+//        The shader starts with "S_BARRIER" to ensure SPI has launched all waves of the workgroup
+//          before any wave in the workgroup could end.  Without this, it is possible not all SGPRs get cleared.
+//    2. Clean remaining SGPRs
+//        Launches a workgroup with 24 waves per workgroup, yielding 6 waves per SIMD in each CU
+//        Waves are allocating 96 SGPRs
+//          CP sets up SPI_RESOURCE_RESERVE_* registers to prevent these waves from allocating SGPRs 0-223.
+//          As such, these 6 waves per SIMD are allocated physical SGPRs 224-799
+//        Barriers do not work for >16 waves per workgroup, so we cannot start with S_BARRIER
+//          Instead, the shader starts with an S_SETHALT 1. Once all waves are launched CP will send unhalt command
+//        The shader then clears all SGPRs allocated to it, cleaning out physical SGPRs 224-799
+
+shader main
+  asic(MI200)
+  type(CS)
+  wave_size(64)
+// Note: original source code from SQ team
+
+//   (theorhetical fastest = ~512clks vgpr + 1536 lds + ~128 sgpr  = 2176 clks)
+
+  s_cmp_eq_u32 s0, 1                                // Bit0 is set, sgpr0 is set then clear VGPRS and LDS as FW set COMPUTE_USER_DATA_3
+  s_cbranch_scc0  label_0023                        // Clean VGPRs and LDS if sgpr0 of wave is set, scc = (s3 == 1)
+  S_BARRIER
+
+  s_movk_i32    m0, 0x0000
+  s_mov_b32     s2, 0x00000078  // Loop 128/8=16 times  (loop unrolled for performance)
+  //
+  // CLEAR VGPRs
+  //
+  s_set_gpr_idx_on  s2, 0x8    // enable Dest VGPR indexing
+label_0005:
+  v_mov_b32     v0, 0
+  v_mov_b32     v1, 0
+  v_mov_b32     v2, 0
+  v_mov_b32     v3, 0
+  v_mov_b32     v4, 0
+  v_mov_b32     v5, 0
+  v_mov_b32     v6, 0
+  v_mov_b32     v7, 0
+  s_sub_u32     s2, s2, 8
+  s_set_gpr_idx_idx  s2
+  s_cbranch_scc0  label_0005
+  s_set_gpr_idx_off
+
+  //
+  //
+
+  s_mov_b32     s2, 0x80000000                      // Bit31 is first_wave
+  s_and_b32     s2, s2, s1                          // sgpr0 has tg_size (first_wave) term as in ucode only COMPUTE_PGM_RSRC2.tg_size_en is set
+  s_cbranch_scc0  label_clean_sgpr_1                // Clean LDS if its first wave of ThreadGroup/WorkGroup
+  // CLEAR LDS
+  //
+  s_mov_b32 exec_lo, 0xffffffff
+  s_mov_b32 exec_hi, 0xffffffff
+  v_mbcnt_lo_u32_b32  v1, exec_hi, 0          // Set V1 to thread-ID (0..63)
+  v_mbcnt_hi_u32_b32  v1, exec_lo, v1         // Set V1 to thread-ID (0..63)
+  v_mul_u32_u24  v1, 0x00000008, v1           // * 8, so each thread is a double-dword address (8byte)
+  s_mov_b32     s2, 0x00000003f               // 64 loop iterations
+  s_mov_b32     m0, 0xffffffff
+  // Clear all of LDS space
+  // Each FirstWave of WorkGroup clears 64kbyte block
+
+label_001F:
+  ds_write2_b64  v1, v[2:3], v[2:3] offset1:32
+  ds_write2_b64  v1, v[4:5], v[4:5] offset0:64 offset1:96
+  v_add_co_u32     v1, vcc, 0x00000400, v1
+  s_sub_u32     s2, s2, 1
+  s_cbranch_scc0  label_001F
+  //
+  // CLEAR SGPRs
+  //
+label_clean_sgpr_1:
+  s_mov_b32     m0, 0x0000005c   // Loop 96/4=24 times  (loop unrolled for performance)
+  s_nop 0
+label_sgpr_loop:
+  s_movreld_b32     s0, 0
+  s_movreld_b32     s1, 0
+  s_movreld_b32     s2, 0
+  s_movreld_b32     s3, 0
+  s_sub_u32         m0, m0, 4
+  s_cbranch_scc0  label_sgpr_loop
+
+  //clear vcc, flat scratch
+  s_mov_b32 flat_scratch_lo, 0   //clear  flat scratch lo SGPR
+  s_mov_b32 flat_scratch_hi, 0   //clear  flat scratch hi SGPR
+  s_mov_b64 vcc, 0               //clear vcc
+  s_mov_b64 ttmp0, 0             //Clear ttmp0 and ttmp1
+  s_mov_b64 ttmp2, 0             //Clear ttmp2 and ttmp3
+  s_mov_b64 ttmp4, 0             //Clear ttmp4 and ttmp5
+  s_mov_b64 ttmp6, 0             //Clear ttmp6 and ttmp7
+  s_mov_b64 ttmp8, 0             //Clear ttmp8 and ttmp9
+  s_mov_b64 ttmp10, 0            //Clear ttmp10 and ttmp11
+  s_mov_b64 ttmp12, 0            //Clear ttmp12 and ttmp13
+  s_mov_b64 ttmp14, 0            //Clear ttmp14 and ttmp15
+s_endpgm
+
+label_0023:
+
+  s_sethalt 1
+
+  s_mov_b32     m0, 0x0000005c   // Loop 96/4=24 times  (loop unrolled for performance)
+  s_nop 0
+label_sgpr_loop1:
+
+  s_movreld_b32     s0, 0
+  s_movreld_b32     s1, 0
+  s_movreld_b32     s2, 0
+  s_movreld_b32     s3, 0
+  s_sub_u32         m0, m0, 4
+  s_cbranch_scc0  label_sgpr_loop1
+
+  //clear vcc, flat scratch
+  s_mov_b32 flat_scratch_lo, 0   //clear  flat scratch lo SGPR
+  s_mov_b32 flat_scratch_hi, 0   //clear  flat scratch hi SGPR
+  s_mov_b64 vcc, 0xee            //clear vcc
+
+s_endpgm
+end
+
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
index 5f8500577c02..cbb74ffc4792 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
@@ -25,35 +25,669 @@
 #include "amdgpu.h"
 #include "amdgpu_gfx.h"
 #include "soc15.h"
+#include "soc15d.h"
 #include "soc15_common.h"
 #include "vega10_enum.h"
 
+#include "v9_structs.h"
+
+#include "ivsrcid/gfx/irqsrcs_gfx_9_0.h"
+
 #include "gc/gc_9_4_3_offset.h"
 #include "gc/gc_9_4_3_sh_mask.h"
 
 #include "gfx_v9_4_3.h"
+#include "gfx_v9_4_3_cleaner_shader.h"
+#include "amdgpu_xcp.h"
+#include "amdgpu_aca.h"
+
+MODULE_FIRMWARE("amdgpu/gc_9_4_3_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_9_4_4_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_9_5_0_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_9_4_3_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_9_4_4_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_9_5_0_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_9_4_3_sjt_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_9_4_4_sjt_mec.bin");
 
+#define GFX9_MEC_HPD_SIZE 4096
 #define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L
 
+#define GOLDEN_GB_ADDR_CONFIG 0x2a114042
+#define CP_HQD_PERSISTENT_STATE_DEFAULT 0xbe05301
+
+#define XCC_REG_RANGE_0_LOW  0x2000     /* XCC gfxdec0 lower Bound */
+#define XCC_REG_RANGE_0_HIGH 0x3400     /* XCC gfxdec0 upper Bound */
+#define XCC_REG_RANGE_1_LOW  0xA000     /* XCC gfxdec1 lower Bound */
+#define XCC_REG_RANGE_1_HIGH 0x10000    /* XCC gfxdec1 upper Bound */
+
+#define NORMALIZE_XCC_REG_OFFSET(offset) \
+	(offset & 0xFFFF)
+
+static const struct amdgpu_hwip_reg_entry gc_reg_list_9_4_3[] = {
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS2),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_STALLED_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCPF_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCPC_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCPG_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regGDS_PROTECTION_FAULT),
+	SOC15_REG_ENTRY_STR(GC, 0, regGDS_VM_PROTECTION_FAULT),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regRMI_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regSQC_DCACHE_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regSQC_ICACHE_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regSQ_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regTCP_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regWD_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regVM_L2_PROTECTION_FAULT_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regVM_L2_PROTECTION_FAULT_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_DEBUG),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC1_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC2_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_SMU_COMMAND),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_SMU_MESSAGE),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_SMU_ARGUMENT_1),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_SMU_ARGUMENT_2),
+	SOC15_REG_ENTRY_STR(GC, 0, regSMU_RLC_RESPONSE),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_SAFE_MODE),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_SMU_SAFE_MODE),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_INT_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_GPM_GENERAL_6),
+	/* SE status registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE0),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE1),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE2),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE3)
+};
+
+static const struct amdgpu_hwip_reg_entry gc_cp_reg_list_9_4_3[] = {
+	/* compute queue registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_VMID),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_ACTIVE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PERSISTENT_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PIPE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_QUEUE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_QUANTUM),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_DEQUEUE_REQUEST),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_EVENTS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_BASE_ADDR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CNTL_STACK_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CNTL_STACK_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_WG_STATE_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_GDS_RESOURCE_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_WPTR_MEM),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_GFX_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+};
+
+struct amdgpu_gfx_ras gfx_v9_4_3_ras;
+
+static void gfx_v9_4_3_set_ring_funcs(struct amdgpu_device *adev);
+static void gfx_v9_4_3_set_irq_funcs(struct amdgpu_device *adev);
+static void gfx_v9_4_3_set_gds_init(struct amdgpu_device *adev);
+static void gfx_v9_4_3_set_rlc_funcs(struct amdgpu_device *adev);
+static int gfx_v9_4_3_get_cu_info(struct amdgpu_device *adev,
+				struct amdgpu_cu_info *cu_info);
+static void gfx_v9_4_3_xcc_set_safe_mode(struct amdgpu_device *adev, int xcc_id);
+static void gfx_v9_4_3_xcc_unset_safe_mode(struct amdgpu_device *adev, int xcc_id);
+
+static void gfx_v9_4_3_kiq_set_resources(struct amdgpu_ring *kiq_ring,
+				uint64_t queue_mask)
+{
+	struct amdgpu_device *adev = kiq_ring->adev;
+	u64 shader_mc_addr;
+
+	/* Cleaner shader MC address */
+	shader_mc_addr = adev->gfx.cleaner_shader_gpu_addr >> 8;
+
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_RESOURCES, 6));
+	amdgpu_ring_write(kiq_ring,
+		PACKET3_SET_RESOURCES_VMID_MASK(0) |
+		/* vmid_mask:0* queue_type:0 (KIQ) */
+		PACKET3_SET_RESOURCES_QUEUE_TYPE(0));
+	amdgpu_ring_write(kiq_ring,
+			lower_32_bits(queue_mask));	/* queue mask lo */
+	amdgpu_ring_write(kiq_ring,
+			upper_32_bits(queue_mask));	/* queue mask hi */
+	amdgpu_ring_write(kiq_ring, lower_32_bits(shader_mc_addr)); /* cleaner shader addr lo */
+	amdgpu_ring_write(kiq_ring, upper_32_bits(shader_mc_addr)); /* cleaner shader addr hi */
+	amdgpu_ring_write(kiq_ring, 0);	/* oac mask */
+	amdgpu_ring_write(kiq_ring, 0);	/* gds heap base:0, gds heap size:0 */
+}
+
+static void gfx_v9_4_3_kiq_map_queues(struct amdgpu_ring *kiq_ring,
+				 struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = kiq_ring->adev;
+	uint64_t mqd_addr = amdgpu_bo_gpu_offset(ring->mqd_obj);
+	uint64_t wptr_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4);
+	uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0;
+
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_MAP_QUEUES, 5));
+	/* Q_sel:0, vmid:0, vidmem: 1, engine:0, num_Q:1*/
+	amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */
+			 PACKET3_MAP_QUEUES_QUEUE_SEL(0) | /* Queue_Sel */
+			 PACKET3_MAP_QUEUES_VMID(0) | /* VMID */
+			 PACKET3_MAP_QUEUES_QUEUE(ring->queue) |
+			 PACKET3_MAP_QUEUES_PIPE(ring->pipe) |
+			 PACKET3_MAP_QUEUES_ME((ring->me == 1 ? 0 : 1)) |
+			 /*queue_type: normal compute queue */
+			 PACKET3_MAP_QUEUES_QUEUE_TYPE(0) |
+			 /* alloc format: all_on_one_pipe */
+			 PACKET3_MAP_QUEUES_ALLOC_FORMAT(0) |
+			 PACKET3_MAP_QUEUES_ENGINE_SEL(eng_sel) |
+			 /* num_queues: must be 1 */
+			 PACKET3_MAP_QUEUES_NUM_QUEUES(1));
+	amdgpu_ring_write(kiq_ring,
+			PACKET3_MAP_QUEUES_DOORBELL_OFFSET(ring->doorbell_index));
+	amdgpu_ring_write(kiq_ring, lower_32_bits(mqd_addr));
+	amdgpu_ring_write(kiq_ring, upper_32_bits(mqd_addr));
+	amdgpu_ring_write(kiq_ring, lower_32_bits(wptr_addr));
+	amdgpu_ring_write(kiq_ring, upper_32_bits(wptr_addr));
+}
+
+static void gfx_v9_4_3_kiq_unmap_queues(struct amdgpu_ring *kiq_ring,
+				   struct amdgpu_ring *ring,
+				   enum amdgpu_unmap_queues_action action,
+				   u64 gpu_addr, u64 seq)
+{
+	uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0;
+
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4));
+	amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */
+			  PACKET3_UNMAP_QUEUES_ACTION(action) |
+			  PACKET3_UNMAP_QUEUES_QUEUE_SEL(0) |
+			  PACKET3_UNMAP_QUEUES_ENGINE_SEL(eng_sel) |
+			  PACKET3_UNMAP_QUEUES_NUM_QUEUES(1));
+	amdgpu_ring_write(kiq_ring,
+			PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index));
+
+	if (action == PREEMPT_QUEUES_NO_UNMAP) {
+		amdgpu_ring_write(kiq_ring, lower_32_bits(gpu_addr));
+		amdgpu_ring_write(kiq_ring, upper_32_bits(gpu_addr));
+		amdgpu_ring_write(kiq_ring, seq);
+	} else {
+		amdgpu_ring_write(kiq_ring, 0);
+		amdgpu_ring_write(kiq_ring, 0);
+		amdgpu_ring_write(kiq_ring, 0);
+	}
+}
+
+static void gfx_v9_4_3_kiq_query_status(struct amdgpu_ring *kiq_ring,
+				   struct amdgpu_ring *ring,
+				   u64 addr,
+				   u64 seq)
+{
+	uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0;
+
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_QUERY_STATUS, 5));
+	amdgpu_ring_write(kiq_ring,
+			  PACKET3_QUERY_STATUS_CONTEXT_ID(0) |
+			  PACKET3_QUERY_STATUS_INTERRUPT_SEL(0) |
+			  PACKET3_QUERY_STATUS_COMMAND(2));
+	/* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */
+	amdgpu_ring_write(kiq_ring,
+			PACKET3_QUERY_STATUS_DOORBELL_OFFSET(ring->doorbell_index) |
+			PACKET3_QUERY_STATUS_ENG_SEL(eng_sel));
+	amdgpu_ring_write(kiq_ring, lower_32_bits(addr));
+	amdgpu_ring_write(kiq_ring, upper_32_bits(addr));
+	amdgpu_ring_write(kiq_ring, lower_32_bits(seq));
+	amdgpu_ring_write(kiq_ring, upper_32_bits(seq));
+}
+
+static void gfx_v9_4_3_kiq_invalidate_tlbs(struct amdgpu_ring *kiq_ring,
+				uint16_t pasid, uint32_t flush_type,
+				bool all_hub)
+{
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_INVALIDATE_TLBS, 0));
+	amdgpu_ring_write(kiq_ring,
+			PACKET3_INVALIDATE_TLBS_DST_SEL(1) |
+			PACKET3_INVALIDATE_TLBS_ALL_HUB(all_hub) |
+			PACKET3_INVALIDATE_TLBS_PASID(pasid) |
+			PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(flush_type));
+}
+
+static void gfx_v9_4_3_kiq_reset_hw_queue(struct amdgpu_ring *kiq_ring, uint32_t queue_type,
+					  uint32_t me_id, uint32_t pipe_id, uint32_t queue_id,
+					  uint32_t xcc_id, uint32_t vmid)
+{
+	struct amdgpu_device *adev = kiq_ring->adev;
+	unsigned i;
+
+	/* enter save mode */
+	amdgpu_gfx_rlc_enter_safe_mode(adev, xcc_id);
+	mutex_lock(&adev->srbm_mutex);
+	soc15_grbm_select(adev, me_id, pipe_id, queue_id, 0, xcc_id);
+
+	if (queue_type == AMDGPU_RING_TYPE_COMPUTE) {
+		WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_DEQUEUE_REQUEST, 0x2);
+		WREG32_SOC15(GC, GET_INST(GC, xcc_id), regSPI_COMPUTE_QUEUE_RESET, 0x1);
+		/* wait till dequeue take effects */
+		for (i = 0; i < adev->usec_timeout; i++) {
+			if (!(RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+		if (i >= adev->usec_timeout)
+			dev_err(adev->dev, "fail to wait on hqd deactive\n");
+	} else {
+		dev_err(adev->dev, "reset queue_type(%d) not supported\n\n", queue_type);
+	}
+
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	/* exit safe mode */
+	amdgpu_gfx_rlc_exit_safe_mode(adev, xcc_id);
+}
+
+static const struct kiq_pm4_funcs gfx_v9_4_3_kiq_pm4_funcs = {
+	.kiq_set_resources = gfx_v9_4_3_kiq_set_resources,
+	.kiq_map_queues = gfx_v9_4_3_kiq_map_queues,
+	.kiq_unmap_queues = gfx_v9_4_3_kiq_unmap_queues,
+	.kiq_query_status = gfx_v9_4_3_kiq_query_status,
+	.kiq_invalidate_tlbs = gfx_v9_4_3_kiq_invalidate_tlbs,
+	.kiq_reset_hw_queue = gfx_v9_4_3_kiq_reset_hw_queue,
+	.set_resources_size = 8,
+	.map_queues_size = 7,
+	.unmap_queues_size = 6,
+	.query_status_size = 7,
+	.invalidate_tlbs_size = 2,
+};
+
+static void gfx_v9_4_3_set_kiq_pm4_funcs(struct amdgpu_device *adev)
+{
+	int i, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++)
+		adev->gfx.kiq[i].pmf = &gfx_v9_4_3_kiq_pm4_funcs;
+}
+
+static void gfx_v9_4_3_init_golden_registers(struct amdgpu_device *adev)
+{
+	int i, num_xcc, dev_inst;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++) {
+		dev_inst = GET_INST(GC, i);
+
+		WREG32_SOC15(GC, dev_inst, regGB_ADDR_CONFIG,
+			     GOLDEN_GB_ADDR_CONFIG);
+		WREG32_FIELD15_PREREG(GC, dev_inst, TCP_UTCL1_CNTL2, SPARE, 0x1);
+	}
+}
+
+static uint32_t gfx_v9_4_3_normalize_xcc_reg_offset(uint32_t reg)
+{
+	uint32_t normalized_reg = NORMALIZE_XCC_REG_OFFSET(reg);
+
+	/* If it is an XCC reg, normalize the reg to keep
+	   lower 16 bits in local xcc */
+
+	if (((normalized_reg >= XCC_REG_RANGE_0_LOW) && (normalized_reg < XCC_REG_RANGE_0_HIGH)) ||
+		((normalized_reg >= XCC_REG_RANGE_1_LOW) && (normalized_reg < XCC_REG_RANGE_1_HIGH)))
+		return normalized_reg;
+	else
+		return reg;
+}
+
+static void gfx_v9_4_3_write_data_to_reg(struct amdgpu_ring *ring, int eng_sel,
+				       bool wc, uint32_t reg, uint32_t val)
+{
+	reg = gfx_v9_4_3_normalize_xcc_reg_offset(reg);
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, WRITE_DATA_ENGINE_SEL(eng_sel) |
+				WRITE_DATA_DST_SEL(0) |
+				(wc ? WR_CONFIRM : 0));
+	amdgpu_ring_write(ring, reg);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, val);
+}
+
+static void gfx_v9_4_3_wait_reg_mem(struct amdgpu_ring *ring, int eng_sel,
+				  int mem_space, int opt, uint32_t addr0,
+				  uint32_t addr1, uint32_t ref, uint32_t mask,
+				  uint32_t inv)
+{
+	/* Only do the normalization on regspace */
+	if (mem_space == 0) {
+		addr0 = gfx_v9_4_3_normalize_xcc_reg_offset(addr0);
+		addr1 = gfx_v9_4_3_normalize_xcc_reg_offset(addr1);
+	}
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5));
+	amdgpu_ring_write(ring,
+				 /* memory (1) or register (0) */
+				 (WAIT_REG_MEM_MEM_SPACE(mem_space) |
+				 WAIT_REG_MEM_OPERATION(opt) | /* wait */
+				 WAIT_REG_MEM_FUNCTION(3) |  /* equal */
+				 WAIT_REG_MEM_ENGINE(eng_sel)));
+
+	if (mem_space)
+		BUG_ON(addr0 & 0x3); /* Dword align */
+	amdgpu_ring_write(ring, addr0);
+	amdgpu_ring_write(ring, addr1);
+	amdgpu_ring_write(ring, ref);
+	amdgpu_ring_write(ring, mask);
+	amdgpu_ring_write(ring, inv); /* poll interval */
+}
+
+static int gfx_v9_4_3_ring_test_ring(struct amdgpu_ring *ring)
+{
+	uint32_t scratch_reg0_offset, xcc_offset;
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t tmp = 0;
+	unsigned i;
+	int r;
+
+	/* Use register offset which is local to XCC in the packet */
+	xcc_offset = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG0);
+	scratch_reg0_offset = SOC15_REG_OFFSET(GC, GET_INST(GC, ring->xcc_id), regSCRATCH_REG0);
+	WREG32(scratch_reg0_offset, 0xCAFEDEAD);
+	tmp = RREG32(scratch_reg0_offset);
+
+	r = amdgpu_ring_alloc(ring, 3);
+	if (r)
+		return r;
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
+	amdgpu_ring_write(ring, xcc_offset - PACKET3_SET_UCONFIG_REG_START);
+	amdgpu_ring_write(ring, 0xDEADBEEF);
+	amdgpu_ring_commit(ring);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = RREG32(scratch_reg0_offset);
+		if (tmp == 0xDEADBEEF)
+			break;
+		udelay(1);
+	}
+
+	if (i >= adev->usec_timeout)
+		r = -ETIMEDOUT;
+	return r;
+}
+
+static int gfx_v9_4_3_ring_test_ib(struct amdgpu_ring *ring, long timeout)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_ib ib;
+	struct dma_fence *f = NULL;
+
+	unsigned index;
+	uint64_t gpu_addr;
+	uint32_t tmp;
+	long r;
+
+	r = amdgpu_device_wb_get(adev, &index);
+	if (r)
+		return r;
+
+	gpu_addr = adev->wb.gpu_addr + (index * 4);
+	adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD);
+	memset(&ib, 0, sizeof(ib));
+
+	r = amdgpu_ib_get(adev, NULL, 20, AMDGPU_IB_POOL_DIRECT, &ib);
+	if (r)
+		goto err1;
+
+	ib.ptr[0] = PACKET3(PACKET3_WRITE_DATA, 3);
+	ib.ptr[1] = WRITE_DATA_DST_SEL(5) | WR_CONFIRM;
+	ib.ptr[2] = lower_32_bits(gpu_addr);
+	ib.ptr[3] = upper_32_bits(gpu_addr);
+	ib.ptr[4] = 0xDEADBEEF;
+	ib.length_dw = 5;
+
+	r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f);
+	if (r)
+		goto err2;
+
+	r = dma_fence_wait_timeout(f, false, timeout);
+	if (r == 0) {
+		r = -ETIMEDOUT;
+		goto err2;
+	} else if (r < 0) {
+		goto err2;
+	}
+
+	tmp = adev->wb.wb[index];
+	if (tmp == 0xDEADBEEF)
+		r = 0;
+	else
+		r = -EINVAL;
+
+err2:
+	amdgpu_ib_free(&ib, NULL);
+	dma_fence_put(f);
+err1:
+	amdgpu_device_wb_free(adev, index);
+	return r;
+}
+
+
+/* This value might differs per partition */
 static uint64_t gfx_v9_4_3_get_gpu_clock_counter(struct amdgpu_device *adev)
 {
 	uint64_t clock;
 
-	amdgpu_gfx_off_ctrl(adev, false);
 	mutex_lock(&adev->gfx.gpu_clock_mutex);
-	WREG32_SOC15(GC, 0, regRLC_CAPTURE_GPU_CLOCK_COUNT, 1);
-	clock = (uint64_t)RREG32_SOC15(GC, 0, regRLC_GPU_CLOCK_COUNT_LSB) |
-		((uint64_t)RREG32_SOC15(GC, 0, regRLC_GPU_CLOCK_COUNT_MSB) << 32ULL);
+	WREG32_SOC15(GC, GET_INST(GC, 0), regRLC_CAPTURE_GPU_CLOCK_COUNT, 1);
+	clock = (uint64_t)RREG32_SOC15(GC, GET_INST(GC, 0), regRLC_GPU_CLOCK_COUNT_LSB) |
+		((uint64_t)RREG32_SOC15(GC, GET_INST(GC, 0), regRLC_GPU_CLOCK_COUNT_MSB) << 32ULL);
 	mutex_unlock(&adev->gfx.gpu_clock_mutex);
-	amdgpu_gfx_off_ctrl(adev, true);
 
 	return clock;
 }
 
-static void gfx_v9_4_3_select_se_sh(struct amdgpu_device *adev,
-				    u32 se_num,
-				    u32 sh_num,
-				    u32 instance)
+static void gfx_v9_4_3_free_microcode(struct amdgpu_device *adev)
+{
+	amdgpu_ucode_release(&adev->gfx.pfp_fw);
+	amdgpu_ucode_release(&adev->gfx.me_fw);
+	amdgpu_ucode_release(&adev->gfx.ce_fw);
+	amdgpu_ucode_release(&adev->gfx.rlc_fw);
+	amdgpu_ucode_release(&adev->gfx.mec_fw);
+	amdgpu_ucode_release(&adev->gfx.mec2_fw);
+
+	kfree(adev->gfx.rlc.register_list_format);
+}
+
+static int gfx_v9_4_3_init_rlc_microcode(struct amdgpu_device *adev,
+					  const char *chip_name)
+{
+	int err;
+	const struct rlc_firmware_header_v2_0 *rlc_hdr;
+	uint16_t version_major;
+	uint16_t version_minor;
+
+
+	err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_rlc.bin", chip_name);
+	if (err)
+		goto out;
+	rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
+
+	version_major = le16_to_cpu(rlc_hdr->header.header_version_major);
+	version_minor = le16_to_cpu(rlc_hdr->header.header_version_minor);
+	err = amdgpu_gfx_rlc_init_microcode(adev, version_major, version_minor);
+out:
+	if (err)
+		amdgpu_ucode_release(&adev->gfx.rlc_fw);
+
+	return err;
+}
+
+static int gfx_v9_4_3_init_cp_compute_microcode(struct amdgpu_device *adev,
+					  const char *chip_name)
+{
+	int err;
+
+	if (amdgpu_sriov_vf(adev)) {
+		err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_sjt_mec.bin", chip_name);
+
+		if (err)
+			err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+							AMDGPU_UCODE_REQUIRED,
+							"amdgpu/%s_mec.bin", chip_name);
+	} else
+		err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_mec.bin", chip_name);
+	if (err)
+		goto out;
+	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1);
+	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1_JT);
+
+	adev->gfx.mec2_fw_version = adev->gfx.mec_fw_version;
+	adev->gfx.mec2_feature_version = adev->gfx.mec_feature_version;
+
+out:
+	if (err)
+		amdgpu_ucode_release(&adev->gfx.mec_fw);
+	return err;
+}
+
+static int gfx_v9_4_3_init_microcode(struct amdgpu_device *adev)
+{
+	char ucode_prefix[15];
+	int r;
+
+	amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix));
+
+	r = gfx_v9_4_3_init_rlc_microcode(adev, ucode_prefix);
+	if (r)
+		return r;
+
+	r = gfx_v9_4_3_init_cp_compute_microcode(adev, ucode_prefix);
+	if (r)
+		return r;
+
+	return r;
+}
+
+static void gfx_v9_4_3_mec_fini(struct amdgpu_device *adev)
+{
+	amdgpu_bo_free_kernel(&adev->gfx.mec.hpd_eop_obj, NULL, NULL);
+	amdgpu_bo_free_kernel(&adev->gfx.mec.mec_fw_obj, NULL, NULL);
+}
+
+static int gfx_v9_4_3_mec_init(struct amdgpu_device *adev)
+{
+	int r, i, num_xcc;
+	u32 *hpd;
+	const __le32 *fw_data;
+	unsigned fw_size;
+	u32 *fw;
+	size_t mec_hpd_size;
+
+	const struct gfx_firmware_header_v1_0 *mec_hdr;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++)
+		bitmap_zero(adev->gfx.mec_bitmap[i].queue_bitmap,
+			AMDGPU_MAX_COMPUTE_QUEUES);
+
+	/* take ownership of the relevant compute queues */
+	amdgpu_gfx_compute_queue_acquire(adev);
+	mec_hpd_size =
+		adev->gfx.num_compute_rings * num_xcc * GFX9_MEC_HPD_SIZE;
+	if (mec_hpd_size) {
+		r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE,
+					      AMDGPU_GEM_DOMAIN_VRAM |
+					      AMDGPU_GEM_DOMAIN_GTT,
+					      &adev->gfx.mec.hpd_eop_obj,
+					      &adev->gfx.mec.hpd_eop_gpu_addr,
+					      (void **)&hpd);
+		if (r) {
+			dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r);
+			gfx_v9_4_3_mec_fini(adev);
+			return r;
+		}
+
+		if (amdgpu_emu_mode == 1) {
+			for (i = 0; i < mec_hpd_size / 4; i++) {
+				memset((void *)(hpd + i), 0, 4);
+				if (i % 50 == 0)
+					msleep(1);
+			}
+		} else {
+			memset(hpd, 0, mec_hpd_size);
+		}
+
+		amdgpu_bo_kunmap(adev->gfx.mec.hpd_eop_obj);
+		amdgpu_bo_unreserve(adev->gfx.mec.hpd_eop_obj);
+	}
+
+	mec_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data;
+
+	fw_data = (const __le32 *)
+		(adev->gfx.mec_fw->data +
+		 le32_to_cpu(mec_hdr->header.ucode_array_offset_bytes));
+	fw_size = le32_to_cpu(mec_hdr->header.ucode_size_bytes);
+
+	r = amdgpu_bo_create_reserved(adev, mec_hdr->header.ucode_size_bytes,
+				      PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT,
+				      &adev->gfx.mec.mec_fw_obj,
+				      &adev->gfx.mec.mec_fw_gpu_addr,
+				      (void **)&fw);
+	if (r) {
+		dev_warn(adev->dev, "(%d) create mec firmware bo failed\n", r);
+		gfx_v9_4_3_mec_fini(adev);
+		return r;
+	}
+
+	memcpy(fw, fw_data, fw_size);
+
+	amdgpu_bo_kunmap(adev->gfx.mec.mec_fw_obj);
+	amdgpu_bo_unreserve(adev->gfx.mec.mec_fw_obj);
+
+	return 0;
+}
+
+static void gfx_v9_4_3_xcc_select_se_sh(struct amdgpu_device *adev, u32 se_num,
+					u32 sh_num, u32 instance, int xcc_id)
 {
 	u32 data;
 
@@ -76,24 +710,24 @@ static void gfx_v9_4_3_select_se_sh(struct amdgpu_device *adev,
 	else
 		data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_INDEX, sh_num);
 
-	WREG32_SOC15_RLC_SHADOW_EX(reg, GC, 0, regGRBM_GFX_INDEX, data);
+	WREG32_SOC15_RLC_SHADOW_EX(reg, GC, GET_INST(GC, xcc_id), regGRBM_GFX_INDEX, data);
 }
 
-static uint32_t wave_read_ind(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t address)
+static uint32_t wave_read_ind(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t address)
 {
-	WREG32_SOC15_RLC(GC, 0, regSQ_IND_INDEX,
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regSQ_IND_INDEX,
 		(wave << SQ_IND_INDEX__WAVE_ID__SHIFT) |
 		(simd << SQ_IND_INDEX__SIMD_ID__SHIFT) |
 		(address << SQ_IND_INDEX__INDEX__SHIFT) |
 		(SQ_IND_INDEX__FORCE_READ_MASK));
-	return RREG32_SOC15(GC, 0, regSQ_IND_DATA);
+	return RREG32_SOC15(GC, GET_INST(GC, xcc_id), regSQ_IND_DATA);
 }
 
-static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd,
+static void wave_read_regs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 			   uint32_t wave, uint32_t thread,
 			   uint32_t regno, uint32_t num, uint32_t *out)
 {
-	WREG32_SOC15_RLC(GC, 0, regSQ_IND_INDEX,
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regSQ_IND_INDEX,
 		(wave << SQ_IND_INDEX__WAVE_ID__SHIFT) |
 		(simd << SQ_IND_INDEX__SIMD_ID__SHIFT) |
 		(regno << SQ_IND_INDEX__INDEX__SHIFT) |
@@ -101,53 +735,663 @@ static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd,
 		(SQ_IND_INDEX__FORCE_READ_MASK) |
 		(SQ_IND_INDEX__AUTO_INCR_MASK));
 	while (num--)
-		*(out++) = RREG32_SOC15(GC, 0, regSQ_IND_DATA);
+		*(out++) = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regSQ_IND_DATA);
 }
 
 static void gfx_v9_4_3_read_wave_data(struct amdgpu_device *adev,
-				      uint32_t simd, uint32_t wave,
+				      uint32_t xcc_id, uint32_t simd, uint32_t wave,
 				      uint32_t *dst, int *no_fields)
 {
 	/* type 1 wave data */
 	dst[(*no_fields)++] = 1;
-	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_STATUS);
-	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_PC_LO);
-	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_PC_HI);
-	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_EXEC_LO);
-	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_EXEC_HI);
-	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_HW_ID);
-	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_INST_DW0);
-	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_INST_DW1);
-	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_GPR_ALLOC);
-	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_LDS_ALLOC);
-	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_TRAPSTS);
-	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_IB_STS);
-	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_IB_DBG0);
-	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_M0);
-	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_MODE);
-}
-
-static void gfx_v9_4_3_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_STATUS);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_PC_LO);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_PC_HI);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_EXEC_LO);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_EXEC_HI);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_HW_ID);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_INST_DW0);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_INST_DW1);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_GPR_ALLOC);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_LDS_ALLOC);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_TRAPSTS);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_IB_STS);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_IB_DBG0);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_M0);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_MODE);
+}
+
+static void gfx_v9_4_3_read_wave_sgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				       uint32_t wave, uint32_t start,
 				       uint32_t size, uint32_t *dst)
 {
-	wave_read_regs(adev, simd, wave, 0,
+	wave_read_regs(adev, xcc_id, simd, wave, 0,
 		       start + SQIND_WAVE_SGPRS_OFFSET, size, dst);
 }
 
-static void gfx_v9_4_3_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v9_4_3_read_wave_vgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				       uint32_t wave, uint32_t thread,
 				       uint32_t start, uint32_t size,
 				       uint32_t *dst)
 {
-	wave_read_regs(adev, simd, wave, thread,
+	wave_read_regs(adev, xcc_id, simd, wave, thread,
 		       start + SQIND_WAVE_VGPRS_OFFSET, size, dst);
 }
 
 static void gfx_v9_4_3_select_me_pipe_q(struct amdgpu_device *adev,
-					u32 me, u32 pipe, u32 q, u32 vm)
+					u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id)
+{
+	soc15_grbm_select(adev, me, pipe, q, vm, GET_INST(GC, xcc_id));
+}
+
+static int gfx_v9_4_3_get_xccs_per_xcp(struct amdgpu_device *adev)
+{
+	u32 xcp_ctl;
+
+	/* Value is expected to be the same on all, fetch from first instance */
+	xcp_ctl = RREG32_SOC15(GC, GET_INST(GC, 0), regCP_HYP_XCP_CTL);
+
+	return REG_GET_FIELD(xcp_ctl, CP_HYP_XCP_CTL, NUM_XCC_IN_XCP);
+}
+
+static int gfx_v9_4_3_switch_compute_partition(struct amdgpu_device *adev,
+						int num_xccs_per_xcp)
+{
+	int ret, i, num_xcc;
+	u32 tmp = 0;
+
+	if (adev->psp.funcs) {
+		ret = psp_spatial_partition(&adev->psp,
+					    NUM_XCC(adev->gfx.xcc_mask) /
+						    num_xccs_per_xcp);
+		if (ret)
+			return ret;
+	} else {
+		num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+
+		for (i = 0; i < num_xcc; i++) {
+			tmp = REG_SET_FIELD(tmp, CP_HYP_XCP_CTL, NUM_XCC_IN_XCP,
+					    num_xccs_per_xcp);
+			tmp = REG_SET_FIELD(tmp, CP_HYP_XCP_CTL, VIRTUAL_XCC_ID,
+					    i % num_xccs_per_xcp);
+			WREG32_SOC15(GC, GET_INST(GC, i), regCP_HYP_XCP_CTL,
+				     tmp);
+		}
+		ret = 0;
+	}
+
+	adev->gfx.num_xcc_per_xcp = num_xccs_per_xcp;
+
+	return ret;
+}
+
+static int gfx_v9_4_3_ih_to_xcc_inst(struct amdgpu_device *adev, int ih_node)
+{
+	int xcc;
+
+	xcc = hweight8(adev->gfx.xcc_mask & GENMASK(ih_node / 2, 0));
+	if (!xcc) {
+		dev_err(adev->dev, "Couldn't find xcc mapping from IH node");
+		return -EINVAL;
+	}
+
+	return xcc - 1;
+}
+
+static const struct amdgpu_gfx_funcs gfx_v9_4_3_gfx_funcs = {
+	.get_gpu_clock_counter = &gfx_v9_4_3_get_gpu_clock_counter,
+	.select_se_sh = &gfx_v9_4_3_xcc_select_se_sh,
+	.read_wave_data = &gfx_v9_4_3_read_wave_data,
+	.read_wave_sgprs = &gfx_v9_4_3_read_wave_sgprs,
+	.read_wave_vgprs = &gfx_v9_4_3_read_wave_vgprs,
+	.select_me_pipe_q = &gfx_v9_4_3_select_me_pipe_q,
+	.switch_partition_mode = &gfx_v9_4_3_switch_compute_partition,
+	.ih_node_to_logical_xcc = &gfx_v9_4_3_ih_to_xcc_inst,
+	.get_xccs_per_xcp = &gfx_v9_4_3_get_xccs_per_xcp,
+};
+
+static int gfx_v9_4_3_aca_bank_parser(struct aca_handle *handle,
+				      struct aca_bank *bank, enum aca_smu_type type,
+				      void *data)
+{
+	struct aca_bank_info info;
+	u64 misc0;
+	u32 instlo;
+	int ret;
+
+	ret = aca_bank_info_decode(bank, &info);
+	if (ret)
+		return ret;
+
+	/* NOTE: overwrite info.die_id with xcd id for gfx */
+	instlo = ACA_REG__IPID__INSTANCEIDLO(bank->regs[ACA_REG_IDX_IPID]);
+	instlo &= GENMASK(31, 1);
+	info.die_id = instlo == mmSMNAID_XCD0_MCA_SMU ? 0 : 1;
+
+	misc0 = bank->regs[ACA_REG_IDX_MISC0];
+
+	switch (type) {
+	case ACA_SMU_TYPE_UE:
+		bank->aca_err_type = ACA_ERROR_TYPE_UE;
+		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type, 1ULL);
+		break;
+	case ACA_SMU_TYPE_CE:
+		bank->aca_err_type = ACA_ERROR_TYPE_CE;
+		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type,
+						     ACA_REG__MISC0__ERRCNT(misc0));
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return ret;
+}
+
+static bool gfx_v9_4_3_aca_bank_is_valid(struct aca_handle *handle, struct aca_bank *bank,
+					 enum aca_smu_type type, void *data)
+{
+	u32 instlo;
+
+	instlo = ACA_REG__IPID__INSTANCEIDLO(bank->regs[ACA_REG_IDX_IPID]);
+	instlo &= GENMASK(31, 1);
+	switch (instlo) {
+	case mmSMNAID_XCD0_MCA_SMU:
+	case mmSMNAID_XCD1_MCA_SMU:
+	case mmSMNXCD_XCD0_MCA_SMU:
+		return true;
+	default:
+		break;
+	}
+
+	return false;
+}
+
+static const struct aca_bank_ops gfx_v9_4_3_aca_bank_ops = {
+	.aca_bank_parser = gfx_v9_4_3_aca_bank_parser,
+	.aca_bank_is_valid = gfx_v9_4_3_aca_bank_is_valid,
+};
+
+static const struct aca_info gfx_v9_4_3_aca_info = {
+	.hwip = ACA_HWIP_TYPE_SMU,
+	.mask = ACA_ERROR_UE_MASK | ACA_ERROR_CE_MASK,
+	.bank_ops = &gfx_v9_4_3_aca_bank_ops,
+};
+
+static int gfx_v9_4_3_gpu_early_init(struct amdgpu_device *adev)
+{
+	adev->gfx.funcs = &gfx_v9_4_3_gfx_funcs;
+	adev->gfx.ras = &gfx_v9_4_3_ras;
+
+	adev->gfx.config.max_hw_contexts = 8;
+	adev->gfx.config.sc_prim_fifo_size_frontend = 0x20;
+	adev->gfx.config.sc_prim_fifo_size_backend = 0x100;
+	adev->gfx.config.sc_hiz_tile_fifo_size = 0x30;
+	adev->gfx.config.sc_earlyz_tile_fifo_size = 0x4C0;
+	adev->gfx.config.gb_addr_config = GOLDEN_GB_ADDR_CONFIG;
+
+	adev->gfx.config.gb_addr_config_fields.num_pipes = 1 <<
+			REG_GET_FIELD(
+					adev->gfx.config.gb_addr_config,
+					GB_ADDR_CONFIG,
+					NUM_PIPES);
+
+	adev->gfx.config.max_tile_pipes =
+		adev->gfx.config.gb_addr_config_fields.num_pipes;
+
+	adev->gfx.config.gb_addr_config_fields.num_banks = 1 <<
+			REG_GET_FIELD(
+					adev->gfx.config.gb_addr_config,
+					GB_ADDR_CONFIG,
+					NUM_BANKS);
+	adev->gfx.config.gb_addr_config_fields.max_compress_frags = 1 <<
+			REG_GET_FIELD(
+					adev->gfx.config.gb_addr_config,
+					GB_ADDR_CONFIG,
+					MAX_COMPRESSED_FRAGS);
+	adev->gfx.config.gb_addr_config_fields.num_rb_per_se = 1 <<
+			REG_GET_FIELD(
+					adev->gfx.config.gb_addr_config,
+					GB_ADDR_CONFIG,
+					NUM_RB_PER_SE);
+	adev->gfx.config.gb_addr_config_fields.num_se = 1 <<
+			REG_GET_FIELD(
+					adev->gfx.config.gb_addr_config,
+					GB_ADDR_CONFIG,
+					NUM_SHADER_ENGINES);
+	adev->gfx.config.gb_addr_config_fields.pipe_interleave_size = 1 << (8 +
+			REG_GET_FIELD(
+					adev->gfx.config.gb_addr_config,
+					GB_ADDR_CONFIG,
+					PIPE_INTERLEAVE_SIZE));
+
+	return 0;
+}
+
+static int gfx_v9_4_3_compute_ring_init(struct amdgpu_device *adev, int ring_id,
+				        int xcc_id, int mec, int pipe, int queue)
+{
+	unsigned irq_type;
+	struct amdgpu_ring *ring = &adev->gfx.compute_ring[ring_id];
+	unsigned int hw_prio;
+	uint32_t xcc_doorbell_start;
+
+	ring = &adev->gfx.compute_ring[xcc_id * adev->gfx.num_compute_rings +
+				       ring_id];
+
+	/* mec0 is me1 */
+	ring->xcc_id = xcc_id;
+	ring->me = mec + 1;
+	ring->pipe = pipe;
+	ring->queue = queue;
+
+	ring->ring_obj = NULL;
+	ring->use_doorbell = true;
+	xcc_doorbell_start = adev->doorbell_index.mec_ring0 +
+			     xcc_id * adev->doorbell_index.xcc_doorbell_range;
+	ring->doorbell_index = (xcc_doorbell_start + ring_id) << 1;
+	ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr +
+			     (ring_id + xcc_id * adev->gfx.num_compute_rings) *
+				     GFX9_MEC_HPD_SIZE;
+	ring->vm_hub = AMDGPU_GFXHUB(xcc_id);
+	sprintf(ring->name, "comp_%d.%d.%d.%d",
+			ring->xcc_id, ring->me, ring->pipe, ring->queue);
+
+	irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
+		+ ((ring->me - 1) * adev->gfx.mec.num_pipe_per_mec)
+		+ ring->pipe;
+	hw_prio = amdgpu_gfx_is_high_priority_compute_queue(adev, ring) ?
+			AMDGPU_GFX_PIPE_PRIO_HIGH : AMDGPU_GFX_PIPE_PRIO_NORMAL;
+	/* type-2 packets are deprecated on MEC, use type-3 instead */
+	return amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type,
+				hw_prio, NULL);
+}
+
+static void gfx_v9_4_3_alloc_ip_dump(struct amdgpu_device *adev)
+{
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_9_4_3);
+	uint32_t *ptr, num_xcc, inst;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+
+	ptr = kcalloc(reg_count * num_xcc, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for GFX IP Dump\n");
+		adev->gfx.ip_dump_core = NULL;
+	} else {
+		adev->gfx.ip_dump_core = ptr;
+	}
+
+	/* Allocate memory for compute queue registers for all the instances */
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_9_4_3);
+	inst = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe_per_mec *
+		adev->gfx.mec.num_queue_per_pipe;
+
+	ptr = kcalloc(reg_count * inst * num_xcc, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for Compute Queues IP Dump\n");
+		adev->gfx.ip_dump_compute_queues = NULL;
+	} else {
+		adev->gfx.ip_dump_compute_queues = ptr;
+	}
+}
+
+static int gfx_v9_4_3_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	int i, j, k, r, ring_id, xcc_id, num_xcc;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+		adev->gfx.cleaner_shader_ptr = gfx_9_4_3_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_9_4_3_cleaner_shader_hex);
+		if (adev->gfx.mec_fw_version >= 153) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	default:
+		adev->gfx.enable_cleaner_shader = false;
+		break;
+	}
+
+	adev->gfx.mec.num_mec = 2;
+	adev->gfx.mec.num_pipe_per_mec = 4;
+	adev->gfx.mec.num_queue_per_pipe = 8;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+
+	/* EOP Event */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, GFX_9_0__SRCID__CP_EOP_INTERRUPT, &adev->gfx.eop_irq);
+	if (r)
+		return r;
+
+	/* Bad opcode Event */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP,
+			      GFX_9_0__SRCID__CP_BAD_OPCODE_ERROR,
+			      &adev->gfx.bad_op_irq);
+	if (r)
+		return r;
+
+	/* Privileged reg */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, GFX_9_0__SRCID__CP_PRIV_REG_FAULT,
+			      &adev->gfx.priv_reg_irq);
+	if (r)
+		return r;
+
+	/* Privileged inst */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, GFX_9_0__SRCID__CP_PRIV_INSTR_FAULT,
+			      &adev->gfx.priv_inst_irq);
+	if (r)
+		return r;
+
+	adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE;
+
+	r = adev->gfx.rlc.funcs->init(adev);
+	if (r) {
+		DRM_ERROR("Failed to init rlc BOs!\n");
+		return r;
+	}
+
+	r = gfx_v9_4_3_mec_init(adev);
+	if (r) {
+		DRM_ERROR("Failed to init MEC BOs!\n");
+		return r;
+	}
+
+	/* set up the compute queues - allocate horizontally across pipes */
+	for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) {
+		ring_id = 0;
+		for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
+			for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
+				for (k = 0; k < adev->gfx.mec.num_pipe_per_mec;
+				     k++) {
+					if (!amdgpu_gfx_is_mec_queue_enabled(
+							adev, xcc_id, i, k, j))
+						continue;
+
+					r = gfx_v9_4_3_compute_ring_init(adev,
+								       ring_id,
+								       xcc_id,
+								       i, k, j);
+					if (r)
+						return r;
+
+					ring_id++;
+				}
+			}
+		}
+
+		r = amdgpu_gfx_kiq_init(adev, GFX9_MEC_HPD_SIZE, xcc_id);
+		if (r) {
+			DRM_ERROR("Failed to init KIQ BOs!\n");
+			return r;
+		}
+
+		r = amdgpu_gfx_kiq_init_ring(adev, xcc_id);
+		if (r)
+			return r;
+
+		/* create MQD for all compute queues as wel as KIQ for SRIOV case */
+		r = amdgpu_gfx_mqd_sw_init(adev,
+				sizeof(struct v9_mqd_allocation), xcc_id);
+		if (r)
+			return r;
+	}
+
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+		if ((adev->gfx.mec_fw_version >= 155) &&
+		    !amdgpu_sriov_vf(adev) &&
+		    !adev->debug_disable_gpu_ring_reset) {
+			adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+			adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_PIPE;
+		}
+		break;
+	case IP_VERSION(9, 5, 0):
+		if ((adev->gfx.mec_fw_version >= 21) &&
+		    !amdgpu_sriov_vf(adev) &&
+		    !adev->debug_disable_gpu_ring_reset) {
+			adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+			adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_PIPE;
+		}
+		break;
+	default:
+		break;
+	}
+	r = gfx_v9_4_3_gpu_early_init(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_gfx_ras_sw_init(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_gfx_sysfs_init(adev);
+	if (r)
+		return r;
+
+	gfx_v9_4_3_alloc_ip_dump(adev);
+
+	return 0;
+}
+
+static int gfx_v9_4_3_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	soc15_grbm_select(adev, me, pipe, q, vm);
+	int i, num_xcc;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < adev->gfx.num_compute_rings * num_xcc; i++)
+		amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
+
+	for (i = 0; i < num_xcc; i++) {
+		amdgpu_gfx_mqd_sw_fini(adev, i);
+		amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq[i].ring);
+		amdgpu_gfx_kiq_fini(adev, i);
+	}
+
+	amdgpu_gfx_cleaner_shader_sw_fini(adev);
+
+	gfx_v9_4_3_mec_fini(adev);
+	amdgpu_bo_unref(&adev->gfx.rlc.clear_state_obj);
+	gfx_v9_4_3_free_microcode(adev);
+	amdgpu_gfx_sysfs_fini(adev);
+
+	kfree(adev->gfx.ip_dump_core);
+	kfree(adev->gfx.ip_dump_compute_queues);
+
+	return 0;
+}
+
+#define DEFAULT_SH_MEM_BASES	(0x6000)
+static void gfx_v9_4_3_xcc_init_compute_vmid(struct amdgpu_device *adev,
+					     int xcc_id)
+{
+	int i;
+	uint32_t sh_mem_config;
+	uint32_t sh_mem_bases;
+	uint32_t data;
+
+	/*
+	 * Configure apertures:
+	 * LDS:         0x60000000'00000000 - 0x60000001'00000000 (4GB)
+	 * Scratch:     0x60000001'00000000 - 0x60000002'00000000 (4GB)
+	 * GPUVM:       0x60010000'00000000 - 0x60020000'00000000 (1TB)
+	 */
+	sh_mem_bases = DEFAULT_SH_MEM_BASES | (DEFAULT_SH_MEM_BASES << 16);
+
+	sh_mem_config = SH_MEM_ADDRESS_MODE_64 |
+			SH_MEM_ALIGNMENT_MODE_UNALIGNED <<
+			SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT;
+
+	mutex_lock(&adev->srbm_mutex);
+	for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) {
+		soc15_grbm_select(adev, 0, 0, 0, i, GET_INST(GC, xcc_id));
+		/* CP and shaders */
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regSH_MEM_CONFIG, sh_mem_config);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regSH_MEM_BASES, sh_mem_bases);
+
+		/* Enable trap for each kfd vmid. */
+		data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regSPI_GDBG_PER_VMID_CNTL);
+		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regSPI_GDBG_PER_VMID_CNTL, data);
+	}
+	soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id));
+	mutex_unlock(&adev->srbm_mutex);
+
+	/*
+	 * Initialize all compute VMIDs to have no GDS, GWS, or OA
+	 * access. These should be enabled by FW for target VMIDs.
+	 */
+	for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) {
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_VMID0_BASE, 2 * i, 0);
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_VMID0_SIZE, 2 * i, 0);
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_GWS_VMID0, i, 0);
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_OA_VMID0, i, 0);
+	}
+}
+
+static void gfx_v9_4_3_xcc_init_gds_vmid(struct amdgpu_device *adev, int xcc_id)
+{
+	int vmid;
+
+	/*
+	 * Initialize all compute and user-gfx VMIDs to have no GDS, GWS, or OA
+	 * access. Compute VMIDs should be enabled by FW for target VMIDs,
+	 * the driver can enable them for graphics. VMID0 should maintain
+	 * access so that HWS firmware can save/restore entries.
+	 */
+	for (vmid = 1; vmid < AMDGPU_NUM_VMID; vmid++) {
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_VMID0_BASE, 2 * vmid, 0);
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_VMID0_SIZE, 2 * vmid, 0);
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_GWS_VMID0, vmid, 0);
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_OA_VMID0, vmid, 0);
+	}
+}
+
+/* For ASICs that needs xnack chain and MEC version supports, set SG_CONFIG1
+ * DISABLE_XNACK_CHECK_IN_RETRY_DISABLE bit and inform KFD to set xnack_chain
+ * bit in SET_RESOURCES
+ */
+static void gfx_v9_4_3_xcc_init_sq(struct amdgpu_device *adev, int xcc_id)
+{
+	uint32_t data;
+
+	if (!(adev->gmc.xnack_flags & AMDGPU_GMC_XNACK_FLAG_CHAIN))
+		return;
+
+	data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regSQ_CONFIG1);
+	data = REG_SET_FIELD(data, SQ_CONFIG1, DISABLE_XNACK_CHECK_IN_RETRY_DISABLE, 1);
+	WREG32_SOC15(GC, xcc_id, regSQ_CONFIG1, data);
+}
+
+static void gfx_v9_4_3_xcc_constants_init(struct amdgpu_device *adev,
+					  int xcc_id)
+{
+	u32 tmp;
+	int i;
+
+	/* XXX SH_MEM regs */
+	/* where to put LDS, scratch, GPUVM in FSA64 space */
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
+		soc15_grbm_select(adev, 0, 0, 0, i, GET_INST(GC, xcc_id));
+		/* CP and shaders */
+		if (i == 0) {
+			tmp = REG_SET_FIELD(0, SH_MEM_CONFIG, ALIGNMENT_MODE,
+					    SH_MEM_ALIGNMENT_MODE_UNALIGNED);
+			tmp = REG_SET_FIELD(tmp, SH_MEM_CONFIG, RETRY_DISABLE,
+					    !!adev->gmc.noretry);
+			WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id),
+					 regSH_MEM_CONFIG, tmp);
+			WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id),
+					 regSH_MEM_BASES, 0);
+		} else {
+			tmp = REG_SET_FIELD(0, SH_MEM_CONFIG, ALIGNMENT_MODE,
+					    SH_MEM_ALIGNMENT_MODE_UNALIGNED);
+			tmp = REG_SET_FIELD(tmp, SH_MEM_CONFIG, RETRY_DISABLE,
+					    !!adev->gmc.noretry);
+			WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id),
+					 regSH_MEM_CONFIG, tmp);
+			tmp = REG_SET_FIELD(0, SH_MEM_BASES, PRIVATE_BASE,
+					    (adev->gmc.private_aperture_start >>
+					     48));
+			tmp = REG_SET_FIELD(tmp, SH_MEM_BASES, SHARED_BASE,
+					    (adev->gmc.shared_aperture_start >>
+					     48));
+			WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id),
+					 regSH_MEM_BASES, tmp);
+		}
+	}
+	soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, 0));
+
+	mutex_unlock(&adev->srbm_mutex);
+
+	gfx_v9_4_3_xcc_init_compute_vmid(adev, xcc_id);
+	gfx_v9_4_3_xcc_init_gds_vmid(adev, xcc_id);
+	gfx_v9_4_3_xcc_init_sq(adev, xcc_id);
+}
+
+static void gfx_v9_4_3_constants_init(struct amdgpu_device *adev)
+{
+	int i, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+
+	gfx_v9_4_3_get_cu_info(adev, &adev->gfx.cu_info);
+	adev->gfx.config.db_debug2 =
+		RREG32_SOC15(GC, GET_INST(GC, 0), regDB_DEBUG2);
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	/* ToDo: GC 9.4.4 */
+	case IP_VERSION(9, 4, 3):
+		if (adev->gfx.mec_fw_version >= 184 &&
+		    (amdgpu_sriov_reg_access_sq_config(adev) ||
+		     !amdgpu_sriov_vf(adev)))
+			adev->gmc.xnack_flags |= AMDGPU_GMC_XNACK_FLAG_CHAIN;
+		break;
+	case IP_VERSION(9, 5, 0):
+		if (adev->gfx.mec_fw_version >= 23)
+			adev->gmc.xnack_flags |= AMDGPU_GMC_XNACK_FLAG_CHAIN;
+		break;
+	default:
+		break;
+	}
+
+	for (i = 0; i < num_xcc; i++)
+		gfx_v9_4_3_xcc_constants_init(adev, i);
+}
+
+static void
+gfx_v9_4_3_xcc_enable_save_restore_machine(struct amdgpu_device *adev,
+					   int xcc_id)
+{
+	WREG32_FIELD15_PREREG(GC, GET_INST(GC, xcc_id), RLC_SRM_CNTL, SRM_ENABLE, 1);
+}
+
+static void gfx_v9_4_3_xcc_init_pg(struct amdgpu_device *adev, int xcc_id)
+{
+	/*
+	 * Rlc save restore list is workable since v2_1.
+	 */
+	gfx_v9_4_3_xcc_enable_save_restore_machine(adev, xcc_id);
+}
+
+static void gfx_v9_4_3_xcc_disable_gpa_mode(struct amdgpu_device *adev, int xcc_id)
+{
+	uint32_t data;
+
+	data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCPC_PSP_DEBUG);
+	data |= CPC_PSP_DEBUG__UTCL2IUGPAOVERRIDE_MASK;
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCPC_PSP_DEBUG, data);
 }
 
 static bool gfx_v9_4_3_is_rlc_enabled(struct amdgpu_device *adev)
@@ -155,48 +1399,69 @@ static bool gfx_v9_4_3_is_rlc_enabled(struct amdgpu_device *adev)
 	uint32_t rlc_setting;
 
 	/* if RLC is not enabled, do nothing */
-	rlc_setting = RREG32_SOC15(GC, 0, regRLC_CNTL);
+	rlc_setting = RREG32_SOC15(GC, GET_INST(GC, 0), regRLC_CNTL);
 	if (!(rlc_setting & RLC_CNTL__RLC_ENABLE_F32_MASK))
 		return false;
 
 	return true;
 }
 
-static void gfx_v9_4_3_set_safe_mode(struct amdgpu_device *adev)
+static void gfx_v9_4_3_xcc_set_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	uint32_t data;
 	unsigned i;
 
 	data = RLC_SAFE_MODE__CMD_MASK;
 	data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT);
-	WREG32_SOC15(GC, 0, regRLC_SAFE_MODE, data);
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_SAFE_MODE, data);
 
 	/* wait for RLC_SAFE_MODE */
 	for (i = 0; i < adev->usec_timeout; i++) {
-		if (!REG_GET_FIELD(RREG32_SOC15(GC, 0, regRLC_SAFE_MODE), RLC_SAFE_MODE, CMD))
+		if (!REG_GET_FIELD(RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_SAFE_MODE), RLC_SAFE_MODE, CMD))
 			break;
 		udelay(1);
 	}
 }
 
-static void gfx_v9_4_3_unset_safe_mode(struct amdgpu_device *adev)
+static void gfx_v9_4_3_xcc_unset_safe_mode(struct amdgpu_device *adev,
+					   int xcc_id)
 {
 	uint32_t data;
 
 	data = RLC_SAFE_MODE__CMD_MASK;
-	WREG32_SOC15(GC, 0, regRLC_SAFE_MODE, data);
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_SAFE_MODE, data);
+}
+
+static void gfx_v9_4_3_init_rlcg_reg_access_ctrl(struct amdgpu_device *adev)
+{
+	int xcc_id, num_xcc;
+	struct amdgpu_rlcg_reg_access_ctrl *reg_access_ctrl;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) {
+		reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl[GET_INST(GC, xcc_id)];
+		reg_access_ctrl->scratch_reg0 = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regSCRATCH_REG0);
+		reg_access_ctrl->scratch_reg1 = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regSCRATCH_REG1);
+		reg_access_ctrl->scratch_reg2 = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regSCRATCH_REG2);
+		reg_access_ctrl->scratch_reg3 = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regSCRATCH_REG3);
+		reg_access_ctrl->grbm_cntl = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regGRBM_GFX_CNTL);
+		reg_access_ctrl->grbm_idx = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regGRBM_GFX_INDEX);
+		reg_access_ctrl->spare_int = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regRLC_SPARE_INT);
+	}
+	adev->gfx.rlc.rlcg_reg_access_supported = true;
 }
 
 static int gfx_v9_4_3_rlc_init(struct amdgpu_device *adev)
 {
 	/* init spm vmid with 0xf */
 	if (adev->gfx.rlc.funcs->update_spm_vmid)
-		adev->gfx.rlc.funcs->update_spm_vmid(adev, 0xf);
+		adev->gfx.rlc.funcs->update_spm_vmid(adev, NULL, 0xf);
 
 	return 0;
 }
 
-static void gfx_v9_4_3_wait_for_rlc_serdes(struct amdgpu_device *adev)
+static void gfx_v9_4_3_xcc_wait_for_rlc_serdes(struct amdgpu_device *adev,
+					       int xcc_id)
 {
 	u32 i, j, k;
 	u32 mask;
@@ -204,15 +1469,17 @@ static void gfx_v9_4_3_wait_for_rlc_serdes(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v9_4_3_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v9_4_3_xcc_select_se_sh(adev, i, j, 0xffffffff,
+						    xcc_id);
 			for (k = 0; k < adev->usec_timeout; k++) {
-				if (RREG32_SOC15(GC, 0, regRLC_SERDES_CU_MASTER_BUSY) == 0)
+				if (RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_SERDES_CU_MASTER_BUSY) == 0)
 					break;
 				udelay(1);
 			}
 			if (k == adev->usec_timeout) {
-				gfx_v9_4_3_select_se_sh(adev, 0xffffffff,
-						      0xffffffff, 0xffffffff);
+				gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff,
+							    0xffffffff,
+							    0xffffffff, xcc_id);
 				mutex_unlock(&adev->grbm_idx_mutex);
 				DRM_INFO("Timeout wait for RLC serdes %u,%u\n",
 					 i, j);
@@ -220,7 +1487,8 @@ static void gfx_v9_4_3_wait_for_rlc_serdes(struct amdgpu_device *adev)
 			}
 		}
 	}
-	gfx_v9_4_3_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff,
+				    xcc_id);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	mask = RLC_SERDES_NONCU_MASTER_BUSY__SE_MASTER_BUSY_MASK |
@@ -228,79 +1496,108 @@ static void gfx_v9_4_3_wait_for_rlc_serdes(struct amdgpu_device *adev)
 		RLC_SERDES_NONCU_MASTER_BUSY__TC0_MASTER_BUSY_MASK |
 		RLC_SERDES_NONCU_MASTER_BUSY__TC1_MASTER_BUSY_MASK;
 	for (k = 0; k < adev->usec_timeout; k++) {
-		if ((RREG32_SOC15(GC, 0, regRLC_SERDES_NONCU_MASTER_BUSY) & mask) == 0)
+		if ((RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_SERDES_NONCU_MASTER_BUSY) & mask) == 0)
 			break;
 		udelay(1);
 	}
 }
 
-static void gfx_v9_4_3_enable_gui_idle_interrupt(struct amdgpu_device *adev,
-					       bool enable)
+static void gfx_v9_4_3_xcc_enable_gui_idle_interrupt(struct amdgpu_device *adev,
+						     bool enable, int xcc_id)
 {
 	u32 tmp;
 
 	/* These interrupts should be enabled to drive DS clock */
 
-	tmp = RREG32_SOC15(GC, 0, regCP_INT_CNTL_RING0);
+	tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_INT_CNTL_RING0);
 
 	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE, enable ? 1 : 0);
 	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_EMPTY_INT_ENABLE, enable ? 1 : 0);
 	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE, enable ? 1 : 0);
-	if (adev->gfx.num_gfx_rings)
-		tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, GFX_IDLE_INT_ENABLE, enable ? 1 : 0);
 
-	WREG32_SOC15(GC, 0, regCP_INT_CNTL_RING0, tmp);
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_INT_CNTL_RING0, tmp);
+}
+
+static void gfx_v9_4_3_xcc_rlc_stop(struct amdgpu_device *adev, int xcc_id)
+{
+	WREG32_FIELD15_PREREG(GC, GET_INST(GC, xcc_id), RLC_CNTL,
+			      RLC_ENABLE_F32, 0);
+	gfx_v9_4_3_xcc_enable_gui_idle_interrupt(adev, false, xcc_id);
+	gfx_v9_4_3_xcc_wait_for_rlc_serdes(adev, xcc_id);
 }
 
 static void gfx_v9_4_3_rlc_stop(struct amdgpu_device *adev)
 {
-	WREG32_FIELD15_PREREG(GC, 0, RLC_CNTL, RLC_ENABLE_F32, 0);
-	gfx_v9_4_3_enable_gui_idle_interrupt(adev, false);
-	gfx_v9_4_3_wait_for_rlc_serdes(adev);
+	int i, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++)
+		gfx_v9_4_3_xcc_rlc_stop(adev, i);
 }
 
-static void gfx_v9_4_3_rlc_reset(struct amdgpu_device *adev)
+static void gfx_v9_4_3_xcc_rlc_reset(struct amdgpu_device *adev, int xcc_id)
 {
-	WREG32_FIELD15_PREREG(GC, 0, GRBM_SOFT_RESET, SOFT_RESET_RLC, 1);
+	WREG32_FIELD15_PREREG(GC, GET_INST(GC, xcc_id), GRBM_SOFT_RESET,
+			      SOFT_RESET_RLC, 1);
 	udelay(50);
-	WREG32_FIELD15_PREREG(GC, 0, GRBM_SOFT_RESET, SOFT_RESET_RLC, 0);
+	WREG32_FIELD15_PREREG(GC, GET_INST(GC, xcc_id), GRBM_SOFT_RESET,
+			      SOFT_RESET_RLC, 0);
 	udelay(50);
 }
 
-static void gfx_v9_4_3_rlc_start(struct amdgpu_device *adev)
+static void gfx_v9_4_3_rlc_reset(struct amdgpu_device *adev)
 {
-#ifdef AMDGPU_RLC_DEBUG_RETRY
-	u32 rlc_ucode_ver;
-#endif
+	int i, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++)
+		gfx_v9_4_3_xcc_rlc_reset(adev, i);
+}
 
-	WREG32_FIELD15_PREREG(GC, 0, RLC_CNTL, RLC_ENABLE_F32, 1);
+static void gfx_v9_4_3_xcc_rlc_start(struct amdgpu_device *adev, int xcc_id)
+{
+	WREG32_FIELD15_PREREG(GC, GET_INST(GC, xcc_id), RLC_CNTL,
+			      RLC_ENABLE_F32, 1);
 	udelay(50);
 
 	/* carrizo do enable cp interrupt after cp inited */
 	if (!(adev->flags & AMD_IS_APU)) {
-		gfx_v9_4_3_enable_gui_idle_interrupt(adev, true);
+		gfx_v9_4_3_xcc_enable_gui_idle_interrupt(adev, true, xcc_id);
 		udelay(50);
 	}
+}
 
+static void gfx_v9_4_3_rlc_start(struct amdgpu_device *adev)
+{
 #ifdef AMDGPU_RLC_DEBUG_RETRY
-	/* RLC_GPM_GENERAL_6 : RLC Ucode version */
-	rlc_ucode_ver = RREG32_SOC15(GC, 0, regRLC_GPM_GENERAL_6);
-	if (rlc_ucode_ver == 0x108) {
-		dev_info(adev->dev,
-			 "Using rlc debug ucode. regRLC_GPM_GENERAL_6 ==0x08%x / fw_ver == %i \n",
-			 rlc_ucode_ver, adev->gfx.rlc_fw_version);
-		/* RLC_GPM_TIMER_INT_3 : Timer interval in RefCLK cycles,
-		 * default is 0x9C4 to create a 100us interval */
-		WREG32_SOC15(GC, 0, regRLC_GPM_TIMER_INT_3, 0x9C4);
-		/* RLC_GPM_GENERAL_12 : Minimum gap between wptr and rptr
-		 * to disable the page fault retry interrupts, default is
-		 * 0x100 (256) */
-		WREG32_SOC15(GC, 0, regRLC_GPM_GENERAL_12, 0x100);
-	}
+	u32 rlc_ucode_ver;
 #endif
+	int i, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++) {
+		gfx_v9_4_3_xcc_rlc_start(adev, i);
+#ifdef AMDGPU_RLC_DEBUG_RETRY
+		/* RLC_GPM_GENERAL_6 : RLC Ucode version */
+		rlc_ucode_ver = RREG32_SOC15(GC, GET_INST(GC, i), regRLC_GPM_GENERAL_6);
+		if (rlc_ucode_ver == 0x108) {
+			dev_info(adev->dev,
+				 "Using rlc debug ucode. regRLC_GPM_GENERAL_6 ==0x08%x / fw_ver == %i \n",
+				 rlc_ucode_ver, adev->gfx.rlc_fw_version);
+			/* RLC_GPM_TIMER_INT_3 : Timer interval in RefCLK cycles,
+			 * default is 0x9C4 to create a 100us interval */
+			WREG32_SOC15(GC, GET_INST(GC, i), regRLC_GPM_TIMER_INT_3, 0x9C4);
+			/* RLC_GPM_GENERAL_12 : Minimum gap between wptr and rptr
+			 * to disable the page fault retry interrupts, default is
+			 * 0x100 (256) */
+			WREG32_SOC15(GC, GET_INST(GC, i), regRLC_GPM_GENERAL_12, 0x100);
+		}
+#endif
+	}
 }
 
-static int gfx_v9_4_3_rlc_load_microcode(struct amdgpu_device *adev)
+static int gfx_v9_4_3_xcc_rlc_load_microcode(struct amdgpu_device *adev,
+					     int xcc_id)
 {
 	const struct rlc_firmware_header_v2_0 *hdr;
 	const __le32 *fw_data;
@@ -316,61 +1613,79 @@ static int gfx_v9_4_3_rlc_load_microcode(struct amdgpu_device *adev)
 			   le32_to_cpu(hdr->header.ucode_array_offset_bytes));
 	fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4;
 
-	WREG32_SOC15(GC, 0, regRLC_GPM_UCODE_ADDR,
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_GPM_UCODE_ADDR,
 			RLCG_UCODE_LOADING_START_ADDRESS);
 	for (i = 0; i < fw_size; i++) {
 		if (amdgpu_emu_mode == 1 && i % 100 == 0) {
 			dev_info(adev->dev, "Write RLC ucode data %u DWs\n", i);
 			msleep(1);
 		}
-		WREG32_SOC15(GC, 0, regRLC_GPM_UCODE_DATA, le32_to_cpup(fw_data++));
+		WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_GPM_UCODE_DATA, le32_to_cpup(fw_data++));
 	}
-	WREG32_SOC15(GC, 0, regRLC_GPM_UCODE_ADDR, adev->gfx.rlc_fw_version);
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_GPM_UCODE_ADDR, adev->gfx.rlc_fw_version);
 
 	return 0;
 }
 
-static int gfx_v9_4_3_rlc_resume(struct amdgpu_device *adev)
+static int gfx_v9_4_3_xcc_rlc_resume(struct amdgpu_device *adev, int xcc_id)
 {
 	int r;
 
-	adev->gfx.rlc.funcs->stop(adev);
+	if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
+		gfx_v9_4_3_xcc_rlc_stop(adev, xcc_id);
+		/* legacy rlc firmware loading */
+		r = gfx_v9_4_3_xcc_rlc_load_microcode(adev, xcc_id);
+		if (r)
+			return r;
+		gfx_v9_4_3_xcc_rlc_start(adev, xcc_id);
+	}
 
+	amdgpu_gfx_rlc_enter_safe_mode(adev, xcc_id);
 	/* disable CG */
-	WREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL, 0);
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGCG_CGLS_CTRL, 0);
+	gfx_v9_4_3_xcc_init_pg(adev, xcc_id);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, xcc_id);
 
-	/* TODO: revisit pg function */
-	/* gfx_v9_4_3_init_pg(adev);*/
+	return 0;
+}
 
-	if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
-		/* legacy rlc firmware loading */
-		r = gfx_v9_4_3_rlc_load_microcode(adev);
+static int gfx_v9_4_3_rlc_resume(struct amdgpu_device *adev)
+{
+	int r, i, num_xcc;
+
+	if (amdgpu_sriov_vf(adev))
+		return 0;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++) {
+		r = gfx_v9_4_3_xcc_rlc_resume(adev, i);
 		if (r)
 			return r;
 	}
 
-	adev->gfx.rlc.funcs->start(adev);
-
 	return 0;
 }
 
-static void gfx_v9_4_3_update_spm_vmid(struct amdgpu_device *adev, unsigned vmid)
+static void gfx_v9_4_3_update_spm_vmid(struct amdgpu_device *adev, struct amdgpu_ring *ring,
+				       unsigned vmid)
 {
-	u32 reg, data;
+	u32 reg, pre_data, data;
 
-	reg = SOC15_REG_OFFSET(GC, 0, regRLC_SPM_MC_CNTL);
-	if (amdgpu_sriov_is_pp_one_vf(adev))
-		data = RREG32_NO_KIQ(reg);
+	reg = SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regRLC_SPM_MC_CNTL);
+	if (amdgpu_sriov_is_pp_one_vf(adev) && !amdgpu_sriov_runtime(adev))
+		pre_data = RREG32_NO_KIQ(reg);
 	else
-		data = RREG32(reg);
+		pre_data = RREG32(reg);
 
-	data &= ~RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK;
+	data =	pre_data & (~RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK);
 	data |= (vmid & RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK) << RLC_SPM_MC_CNTL__RLC_SPM_VMID__SHIFT;
 
-	if (amdgpu_sriov_is_pp_one_vf(adev))
-		WREG32_SOC15_NO_KIQ(GC, 0, regRLC_SPM_MC_CNTL, data);
-	else
-		WREG32_SOC15(GC, 0, regRLC_SPM_MC_CNTL, data);
+	if (pre_data != data) {
+		if (amdgpu_sriov_is_pp_one_vf(adev) && !amdgpu_sriov_runtime(adev)) {
+			WREG32_SOC15_NO_KIQ(GC, GET_INST(GC, 0), regRLC_SPM_MC_CNTL, data);
+		} else
+			WREG32_SOC15(GC, GET_INST(GC, 0), regRLC_SPM_MC_CNTL, data);
+	}
 }
 
 static const struct soc15_reg_rlcg rlcg_access_gc_9_4_3[] = {
@@ -382,7 +1697,7 @@ static bool gfx_v9_4_3_check_rlcg_range(struct amdgpu_device *adev,
 					uint32_t offset,
 					struct soc15_reg_rlcg *entries, int arr_size)
 {
-	int i;
+	int i, inst;
 	uint32_t reg;
 
 	if (!entries)
@@ -392,7 +1707,12 @@ static bool gfx_v9_4_3_check_rlcg_range(struct amdgpu_device *adev,
 		const struct soc15_reg_rlcg *entry;
 
 		entry = &entries[i];
-		reg = adev->reg_offset[entry->hwip][entry->instance][entry->segment] + entry->reg;
+		inst = adev->ip_map.logical_to_dev_inst ?
+			       adev->ip_map.logical_to_dev_inst(
+				       adev, entry->hwip, entry->instance) :
+			       entry->instance;
+		reg = adev->reg_offset[entry->hwip][inst][entry->segment] +
+		      entry->reg;
 		if (offset == reg)
 			return true;
 	}
@@ -407,19 +1727,1028 @@ static bool gfx_v9_4_3_is_rlcg_access_range(struct amdgpu_device *adev, u32 offs
 					ARRAY_SIZE(rlcg_access_gc_9_4_3));
 }
 
-const struct amdgpu_gfx_funcs gfx_v9_4_3_gfx_funcs = {
-	.get_gpu_clock_counter = &gfx_v9_4_3_get_gpu_clock_counter,
-	.select_se_sh = &gfx_v9_4_3_select_se_sh,
-	.read_wave_data = &gfx_v9_4_3_read_wave_data,
-	.read_wave_sgprs = &gfx_v9_4_3_read_wave_sgprs,
-	.read_wave_vgprs = &gfx_v9_4_3_read_wave_vgprs,
-	.select_me_pipe_q = &gfx_v9_4_3_select_me_pipe_q,
-};
+static void gfx_v9_4_3_xcc_cp_compute_enable(struct amdgpu_device *adev,
+					     bool enable, int xcc_id)
+{
+	if (enable) {
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_MEC_CNTL, 0);
+	} else {
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_MEC_CNTL,
+			(CP_MEC_CNTL__MEC_INVALIDATE_ICACHE_MASK |
+			 CP_MEC_CNTL__MEC_ME1_PIPE0_RESET_MASK |
+			 CP_MEC_CNTL__MEC_ME1_PIPE1_RESET_MASK |
+			 CP_MEC_CNTL__MEC_ME1_PIPE2_RESET_MASK |
+			 CP_MEC_CNTL__MEC_ME1_PIPE3_RESET_MASK |
+			 CP_MEC_CNTL__MEC_ME2_PIPE0_RESET_MASK |
+			 CP_MEC_CNTL__MEC_ME2_PIPE1_RESET_MASK |
+			 CP_MEC_CNTL__MEC_ME1_HALT_MASK |
+			 CP_MEC_CNTL__MEC_ME2_HALT_MASK));
+		adev->gfx.kiq[xcc_id].ring.sched.ready = false;
+	}
+	udelay(50);
+}
+
+static int gfx_v9_4_3_xcc_cp_compute_load_microcode(struct amdgpu_device *adev,
+						    int xcc_id)
+{
+	const struct gfx_firmware_header_v1_0 *mec_hdr;
+	const __le32 *fw_data;
+	unsigned i;
+	u32 tmp;
+	u32 mec_ucode_addr_offset;
+	u32 mec_ucode_data_offset;
+
+	if (!adev->gfx.mec_fw)
+		return -EINVAL;
+
+	gfx_v9_4_3_xcc_cp_compute_enable(adev, false, xcc_id);
+
+	mec_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data;
+	amdgpu_ucode_print_gfx_hdr(&mec_hdr->header);
+
+	fw_data = (const __le32 *)
+		(adev->gfx.mec_fw->data +
+		 le32_to_cpu(mec_hdr->header.ucode_array_offset_bytes));
+	tmp = 0;
+	tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, VMID, 0);
+	tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, CACHE_POLICY, 0);
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_CPC_IC_BASE_CNTL, tmp);
+
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_CPC_IC_BASE_LO,
+		adev->gfx.mec.mec_fw_gpu_addr & 0xFFFFF000);
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_CPC_IC_BASE_HI,
+		upper_32_bits(adev->gfx.mec.mec_fw_gpu_addr));
+
+	mec_ucode_addr_offset =
+		SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_MEC_ME1_UCODE_ADDR);
+	mec_ucode_data_offset =
+		SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_MEC_ME1_UCODE_DATA);
+
+	/* MEC1 */
+	WREG32(mec_ucode_addr_offset, mec_hdr->jt_offset);
+	for (i = 0; i < mec_hdr->jt_size; i++)
+		WREG32(mec_ucode_data_offset,
+		       le32_to_cpup(fw_data + mec_hdr->jt_offset + i));
+
+	WREG32(mec_ucode_addr_offset, adev->gfx.mec_fw_version);
+	/* Todo : Loading MEC2 firmware is only necessary if MEC2 should run different microcode than MEC1. */
+
+	return 0;
+}
+
+/* KIQ functions */
+static void gfx_v9_4_3_xcc_kiq_setting(struct amdgpu_ring *ring, int xcc_id)
+{
+	uint32_t tmp;
+	struct amdgpu_device *adev = ring->adev;
+
+	/* tell RLC which is KIQ queue */
+	tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CP_SCHEDULERS);
+	tmp &= 0xffffff00;
+	tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regRLC_CP_SCHEDULERS, tmp | 0x80);
+}
+
+static void gfx_v9_4_3_mqd_set_priority(struct amdgpu_ring *ring, struct v9_mqd *mqd)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE) {
+		if (amdgpu_gfx_is_high_priority_compute_queue(adev, ring)) {
+			mqd->cp_hqd_pipe_priority = AMDGPU_GFX_PIPE_PRIO_HIGH;
+			mqd->cp_hqd_queue_priority =
+				AMDGPU_GFX_QUEUE_PRIORITY_MAXIMUM;
+		}
+	}
+}
+
+static int gfx_v9_4_3_xcc_mqd_init(struct amdgpu_ring *ring, int xcc_id)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct v9_mqd *mqd = ring->mqd_ptr;
+	uint64_t hqd_gpu_addr, wb_gpu_addr, eop_base_addr;
+	uint32_t tmp;
+
+	mqd->header = 0xC0310800;
+	mqd->compute_pipelinestat_enable = 0x00000001;
+	mqd->compute_static_thread_mgmt_se0 = 0xffffffff;
+	mqd->compute_static_thread_mgmt_se1 = 0xffffffff;
+	mqd->compute_static_thread_mgmt_se2 = 0xffffffff;
+	mqd->compute_static_thread_mgmt_se3 = 0xffffffff;
+	mqd->compute_misc_reserved = 0x00000003;
+
+	mqd->dynamic_cu_mask_addr_lo =
+		lower_32_bits(ring->mqd_gpu_addr
+			      + offsetof(struct v9_mqd_allocation, dynamic_cu_mask));
+	mqd->dynamic_cu_mask_addr_hi =
+		upper_32_bits(ring->mqd_gpu_addr
+			      + offsetof(struct v9_mqd_allocation, dynamic_cu_mask));
+
+	eop_base_addr = ring->eop_gpu_addr >> 8;
+	mqd->cp_hqd_eop_base_addr_lo = eop_base_addr;
+	mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr);
+
+	/* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */
+	tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_EOP_CONTROL);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_EOP_CONTROL, EOP_SIZE,
+			(order_base_2(GFX9_MEC_HPD_SIZE / 4) - 1));
+
+	mqd->cp_hqd_eop_control = tmp;
+
+	/* enable doorbell? */
+	tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_DOORBELL_CONTROL);
+
+	if (ring->use_doorbell) {
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_OFFSET, ring->doorbell_index);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_EN, 1);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_SOURCE, 0);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_HIT, 0);
+		if (amdgpu_sriov_multi_vf_mode(adev))
+			tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+					    DOORBELL_MODE, 1);
+	} else {
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+					 DOORBELL_EN, 0);
+	}
+
+	mqd->cp_hqd_pq_doorbell_control = tmp;
+
+	/* disable the queue if it's active */
+	ring->wptr = 0;
+	mqd->cp_hqd_dequeue_request = 0;
+	mqd->cp_hqd_pq_rptr = 0;
+	mqd->cp_hqd_pq_wptr_lo = 0;
+	mqd->cp_hqd_pq_wptr_hi = 0;
+
+	/* set the pointer to the MQD */
+	mqd->cp_mqd_base_addr_lo = ring->mqd_gpu_addr & 0xfffffffc;
+	mqd->cp_mqd_base_addr_hi = upper_32_bits(ring->mqd_gpu_addr);
+
+	/* set MQD vmid to 0 */
+	tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_MQD_CONTROL);
+	tmp = REG_SET_FIELD(tmp, CP_MQD_CONTROL, VMID, 0);
+	mqd->cp_mqd_control = tmp;
+
+	/* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */
+	hqd_gpu_addr = ring->gpu_addr >> 8;
+	mqd->cp_hqd_pq_base_lo = hqd_gpu_addr;
+	mqd->cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr);
+
+	/* set up the HQD, this is similar to CP_RB0_CNTL */
+	tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_CONTROL);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE,
+			    (order_base_2(ring->ring_size / 4) - 1));
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE,
+			((order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1) << 8));
+#ifdef __BIG_ENDIAN
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, ENDIAN_SWAP, 1);
+#endif
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 0);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, ROQ_PQ_IB_FLIP, 0);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, PRIV_STATE, 1);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, KMD_QUEUE, 1);
+	mqd->cp_hqd_pq_control = tmp;
+
+	/* set the wb address whether it's enabled or not */
+	wb_gpu_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4);
+	mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc;
+	mqd->cp_hqd_pq_rptr_report_addr_hi =
+		upper_32_bits(wb_gpu_addr) & 0xffff;
+
+	/* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */
+	wb_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4);
+	mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffffc;
+	mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff;
+
+	/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
+	ring->wptr = 0;
+	mqd->cp_hqd_pq_rptr = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_RPTR);
+
+	/* set the vmid for the queue */
+	mqd->cp_hqd_vmid = 0;
+
+	tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_PERSISTENT_STATE);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PERSISTENT_STATE, PRELOAD_SIZE, 0x53);
+	mqd->cp_hqd_persistent_state = tmp;
+
+	/* set MIN_IB_AVAIL_SIZE */
+	tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_IB_CONTROL);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_IB_CONTROL, MIN_IB_AVAIL_SIZE, 3);
+	mqd->cp_hqd_ib_control = tmp;
+
+	/* set static priority for a queue/ring */
+	gfx_v9_4_3_mqd_set_priority(ring, mqd);
+	mqd->cp_hqd_quantum = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_QUANTUM);
+
+	/* map_queues packet doesn't need activate the queue,
+	 * so only kiq need set this field.
+	 */
+	if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ)
+		mqd->cp_hqd_active = 1;
+
+	return 0;
+}
+
+static int gfx_v9_4_3_xcc_kiq_init_register(struct amdgpu_ring *ring,
+					    int xcc_id)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct v9_mqd *mqd = ring->mqd_ptr;
+	int j;
+
+	/* disable wptr polling */
+	WREG32_FIELD15_PREREG(GC, GET_INST(GC, xcc_id), CP_PQ_WPTR_POLL_CNTL, EN, 0);
+
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_EOP_BASE_ADDR,
+	       mqd->cp_hqd_eop_base_addr_lo);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_EOP_BASE_ADDR_HI,
+	       mqd->cp_hqd_eop_base_addr_hi);
+
+	/* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_EOP_CONTROL,
+	       mqd->cp_hqd_eop_control);
+
+	/* enable doorbell? */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_DOORBELL_CONTROL,
+	       mqd->cp_hqd_pq_doorbell_control);
+
+	/* disable the queue if it's active */
+	if (RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE) & 1) {
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_DEQUEUE_REQUEST, 1);
+		for (j = 0; j < adev->usec_timeout; j++) {
+			if (!(RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_DEQUEUE_REQUEST,
+		       mqd->cp_hqd_dequeue_request);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_RPTR,
+		       mqd->cp_hqd_pq_rptr);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_LO,
+		       mqd->cp_hqd_pq_wptr_lo);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_HI,
+		       mqd->cp_hqd_pq_wptr_hi);
+	}
+
+	/* set the pointer to the MQD */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_MQD_BASE_ADDR,
+	       mqd->cp_mqd_base_addr_lo);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_MQD_BASE_ADDR_HI,
+	       mqd->cp_mqd_base_addr_hi);
+
+	/* set MQD vmid to 0 */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_MQD_CONTROL,
+	       mqd->cp_mqd_control);
+
+	/* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_BASE,
+	       mqd->cp_hqd_pq_base_lo);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_BASE_HI,
+	       mqd->cp_hqd_pq_base_hi);
+
+	/* set up the HQD, this is similar to CP_RB0_CNTL */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_CONTROL,
+	       mqd->cp_hqd_pq_control);
+
+	/* set the wb address whether it's enabled or not */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_RPTR_REPORT_ADDR,
+				mqd->cp_hqd_pq_rptr_report_addr_lo);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_RPTR_REPORT_ADDR_HI,
+				mqd->cp_hqd_pq_rptr_report_addr_hi);
+
+	/* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_POLL_ADDR,
+	       mqd->cp_hqd_pq_wptr_poll_addr_lo);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_POLL_ADDR_HI,
+	       mqd->cp_hqd_pq_wptr_poll_addr_hi);
+
+	/* enable the doorbell if requested */
+	if (ring->use_doorbell) {
+		WREG32_SOC15(
+			GC, GET_INST(GC, xcc_id),
+			regCP_MEC_DOORBELL_RANGE_LOWER,
+			((adev->doorbell_index.kiq +
+			  xcc_id * adev->doorbell_index.xcc_doorbell_range) *
+			 2) << 2);
+		WREG32_SOC15(
+			GC, GET_INST(GC, xcc_id),
+			regCP_MEC_DOORBELL_RANGE_UPPER,
+			((adev->doorbell_index.userqueue_end +
+			  xcc_id * adev->doorbell_index.xcc_doorbell_range) *
+			 2) << 2);
+	}
+
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_DOORBELL_CONTROL,
+	       mqd->cp_hqd_pq_doorbell_control);
+
+	/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_LO,
+	       mqd->cp_hqd_pq_wptr_lo);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_HI,
+	       mqd->cp_hqd_pq_wptr_hi);
+
+	/* set the vmid for the queue */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_VMID, mqd->cp_hqd_vmid);
+
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PERSISTENT_STATE,
+	       mqd->cp_hqd_persistent_state);
+
+	/* activate the queue */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE,
+	       mqd->cp_hqd_active);
+
+	if (ring->use_doorbell)
+		WREG32_FIELD15_PREREG(GC, GET_INST(GC, xcc_id), CP_PQ_STATUS, DOORBELL_ENABLE, 1);
+
+	return 0;
+}
+
+static int gfx_v9_4_3_xcc_q_fini_register(struct amdgpu_ring *ring,
+					    int xcc_id)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int j;
+
+	/* disable the queue if it's active */
+	if (RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE) & 1) {
+
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_DEQUEUE_REQUEST, 1);
+
+		for (j = 0; j < adev->usec_timeout; j++) {
+			if (!(RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+
+		if (j == AMDGPU_MAX_USEC_TIMEOUT) {
+			DRM_DEBUG("%s dequeue request failed.\n", ring->name);
+
+			/* Manual disable if dequeue request times out */
+			WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE, 0);
+		}
+
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_DEQUEUE_REQUEST,
+		      0);
+	}
+
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_IQ_TIMER, 0);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_IB_CONTROL, 0);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PERSISTENT_STATE, CP_HQD_PERSISTENT_STATE_DEFAULT);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_DOORBELL_CONTROL, 0x40000000);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_DOORBELL_CONTROL, 0);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_RPTR, 0);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_HI, 0);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_LO, 0);
+
+	return 0;
+}
+
+static int gfx_v9_4_3_xcc_kiq_init_queue(struct amdgpu_ring *ring, int xcc_id)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct v9_mqd *mqd = ring->mqd_ptr;
+	struct v9_mqd *tmp_mqd;
+
+	gfx_v9_4_3_xcc_kiq_setting(ring, xcc_id);
+
+	/* GPU could be in bad state during probe, driver trigger the reset
+	 * after load the SMU, in this case , the mqd is not be initialized.
+	 * driver need to re-init the mqd.
+	 * check mqd->cp_hqd_pq_control since this value should not be 0
+	 */
+	tmp_mqd = (struct v9_mqd *)adev->gfx.kiq[xcc_id].mqd_backup;
+	if (amdgpu_in_reset(adev) && tmp_mqd->cp_hqd_pq_control) {
+		/* for GPU_RESET case , reset MQD to a clean status */
+		if (adev->gfx.kiq[xcc_id].mqd_backup)
+			memcpy(mqd, adev->gfx.kiq[xcc_id].mqd_backup, sizeof(struct v9_mqd_allocation));
+
+		/* reset ring buffer */
+		ring->wptr = 0;
+		amdgpu_ring_clear_ring(ring);
+		mutex_lock(&adev->srbm_mutex);
+		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, GET_INST(GC, xcc_id));
+		gfx_v9_4_3_xcc_kiq_init_register(ring, xcc_id);
+		soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id));
+		mutex_unlock(&adev->srbm_mutex);
+	} else {
+		memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation));
+		((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
+		((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF;
+		mutex_lock(&adev->srbm_mutex);
+		if (amdgpu_sriov_vf(adev) && adev->in_suspend)
+			amdgpu_ring_clear_ring(ring);
+		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, GET_INST(GC, xcc_id));
+		gfx_v9_4_3_xcc_mqd_init(ring, xcc_id);
+		gfx_v9_4_3_xcc_kiq_init_register(ring, xcc_id);
+		soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id));
+		mutex_unlock(&adev->srbm_mutex);
+
+		if (adev->gfx.kiq[xcc_id].mqd_backup)
+			memcpy(adev->gfx.kiq[xcc_id].mqd_backup, mqd, sizeof(struct v9_mqd_allocation));
+	}
+
+	return 0;
+}
+
+static void gfx_v9_4_3_xcc_kcq_init_queue(struct amdgpu_ring *ring, int xcc_id,
+					  bool restore)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct v9_mqd *mqd = ring->mqd_ptr;
+	int mqd_idx = ring - &adev->gfx.compute_ring[0];
+	struct v9_mqd *tmp_mqd;
+
+	/* Same as above kiq init, driver need to re-init the mqd if mqd->cp_hqd_pq_control
+	 * is not be initialized before
+	 */
+	tmp_mqd = (struct v9_mqd *)adev->gfx.mec.mqd_backup[mqd_idx];
+
+	if (!restore && (!tmp_mqd->cp_hqd_pq_control ||
+	    (!amdgpu_in_reset(adev) && !adev->in_suspend))) {
+		memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation));
+		((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
+		((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF;
+		mutex_lock(&adev->srbm_mutex);
+		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, GET_INST(GC, xcc_id));
+		gfx_v9_4_3_xcc_mqd_init(ring, xcc_id);
+		soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id));
+		mutex_unlock(&adev->srbm_mutex);
+
+		if (adev->gfx.mec.mqd_backup[mqd_idx])
+			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation));
+	} else {
+		/* restore MQD to a clean status */
+		if (adev->gfx.mec.mqd_backup[mqd_idx])
+			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation));
+		/* reset ring buffer */
+		ring->wptr = 0;
+		atomic64_set((atomic64_t *)&adev->wb.wb[ring->wptr_offs], 0);
+		amdgpu_ring_clear_ring(ring);
+	}
+}
+
+static int gfx_v9_4_3_xcc_kcq_fini_register(struct amdgpu_device *adev, int xcc_id)
+{
+	struct amdgpu_ring *ring;
+	int j;
+
+	for (j = 0; j < adev->gfx.num_compute_rings; j++) {
+		ring = &adev->gfx.compute_ring[j +  xcc_id * adev->gfx.num_compute_rings];
+		if (!amdgpu_in_reset(adev) && !adev->in_suspend) {
+			mutex_lock(&adev->srbm_mutex);
+			soc15_grbm_select(adev, ring->me,
+					ring->pipe,
+					ring->queue, 0, GET_INST(GC, xcc_id));
+			gfx_v9_4_3_xcc_q_fini_register(ring, xcc_id);
+			soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id));
+			mutex_unlock(&adev->srbm_mutex);
+		}
+	}
+
+	return 0;
+}
+
+static int gfx_v9_4_3_xcc_kiq_resume(struct amdgpu_device *adev, int xcc_id)
+{
+	gfx_v9_4_3_xcc_kiq_init_queue(&adev->gfx.kiq[xcc_id].ring, xcc_id);
+	return 0;
+}
+
+static int gfx_v9_4_3_xcc_kcq_resume(struct amdgpu_device *adev, int xcc_id)
+{
+	struct amdgpu_ring *ring;
+	int i;
+
+	gfx_v9_4_3_xcc_cp_compute_enable(adev, true, xcc_id);
+
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		ring = &adev->gfx.compute_ring[i + xcc_id *
+			adev->gfx.num_compute_rings];
+
+		gfx_v9_4_3_xcc_kcq_init_queue(ring, xcc_id, false);
+	}
+
+	return amdgpu_gfx_enable_kcq(adev, xcc_id);
+}
+
+static int gfx_v9_4_3_xcc_cp_resume(struct amdgpu_device *adev, int xcc_id)
+{
+	struct amdgpu_ring *ring;
+	int r, j;
+
+	gfx_v9_4_3_xcc_enable_gui_idle_interrupt(adev, false, xcc_id);
+
+	if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
+		gfx_v9_4_3_xcc_disable_gpa_mode(adev, xcc_id);
+
+		r = gfx_v9_4_3_xcc_cp_compute_load_microcode(adev, xcc_id);
+		if (r)
+			return r;
+	} else {
+		gfx_v9_4_3_xcc_cp_compute_enable(adev, false, xcc_id);
+	}
+
+	r = gfx_v9_4_3_xcc_kiq_resume(adev, xcc_id);
+	if (r)
+		return r;
+
+	r = gfx_v9_4_3_xcc_kcq_resume(adev, xcc_id);
+	if (r)
+		return r;
+
+	for (j = 0; j < adev->gfx.num_compute_rings; j++) {
+		ring = &adev->gfx.compute_ring
+				[j + xcc_id * adev->gfx.num_compute_rings];
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
+	}
+
+	gfx_v9_4_3_xcc_enable_gui_idle_interrupt(adev, true, xcc_id);
+
+	return 0;
+}
+
+static int gfx_v9_4_3_cp_resume(struct amdgpu_device *adev)
+{
+	int r = 0, i, num_xcc, num_xcp, num_xcc_per_xcp;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	if (amdgpu_sriov_vf(adev)) {
+		enum amdgpu_gfx_partition mode;
+
+		mode = amdgpu_xcp_query_partition_mode(adev->xcp_mgr,
+						       AMDGPU_XCP_FL_NONE);
+		if (mode == AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE)
+			return -EINVAL;
+		num_xcc_per_xcp = gfx_v9_4_3_get_xccs_per_xcp(adev);
+		adev->gfx.num_xcc_per_xcp = num_xcc_per_xcp;
+		num_xcp = num_xcc / num_xcc_per_xcp;
+		r = amdgpu_xcp_init(adev->xcp_mgr, num_xcp, mode);
+
+	} else {
+		if (adev->in_suspend)
+			amdgpu_xcp_restore_partition_mode(adev->xcp_mgr);
+		else if (amdgpu_xcp_query_partition_mode(adev->xcp_mgr,
+						    AMDGPU_XCP_FL_NONE) ==
+		    AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE)
+			r = amdgpu_xcp_switch_partition_mode(
+				adev->xcp_mgr, amdgpu_user_partt_mode);
+	}
+	if (r)
+		return r;
+
+	for (i = 0; i < num_xcc; i++) {
+		r = gfx_v9_4_3_xcc_cp_resume(adev, i);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+static void gfx_v9_4_3_xcc_fini(struct amdgpu_device *adev, int xcc_id)
+{
+	if (amdgpu_gfx_disable_kcq(adev, xcc_id))
+		DRM_ERROR("XCD %d KCQ disable failed\n", xcc_id);
+
+	if (amdgpu_sriov_vf(adev)) {
+		/* must disable polling for SRIOV when hw finished, otherwise
+		 * CPC engine may still keep fetching WB address which is already
+		 * invalid after sw finished and trigger DMAR reading error in
+		 * hypervisor side.
+		 */
+		WREG32_FIELD15_PREREG(GC, GET_INST(GC, xcc_id), CP_PQ_WPTR_POLL_CNTL, EN, 0);
+		return;
+	}
+
+	/* Use deinitialize sequence from CAIL when unbinding device
+	 * from driver, otherwise KIQ is hanging when binding back
+	 */
+	if (!amdgpu_in_reset(adev) && !adev->in_suspend) {
+		mutex_lock(&adev->srbm_mutex);
+		soc15_grbm_select(adev, adev->gfx.kiq[xcc_id].ring.me,
+				  adev->gfx.kiq[xcc_id].ring.pipe,
+				  adev->gfx.kiq[xcc_id].ring.queue, 0,
+				  GET_INST(GC, xcc_id));
+		gfx_v9_4_3_xcc_q_fini_register(&adev->gfx.kiq[xcc_id].ring,
+						 xcc_id);
+		soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id));
+		mutex_unlock(&adev->srbm_mutex);
+	}
+
+	gfx_v9_4_3_xcc_kcq_fini_register(adev, xcc_id);
+	gfx_v9_4_3_xcc_cp_compute_enable(adev, false, xcc_id);
+}
+
+static int gfx_v9_4_3_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	amdgpu_gfx_cleaner_shader_init(adev, adev->gfx.cleaner_shader_size,
+				       adev->gfx.cleaner_shader_ptr);
+
+	if (!amdgpu_sriov_vf(adev))
+		gfx_v9_4_3_init_golden_registers(adev);
+
+	gfx_v9_4_3_constants_init(adev);
+
+	r = adev->gfx.rlc.funcs->resume(adev);
+	if (r)
+		return r;
+
+	r = gfx_v9_4_3_cp_resume(adev);
+	if (r)
+		return r;
+
+	return r;
+}
+
+static int gfx_v9_4_3_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, num_xcc;
+
+	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
+	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
+	amdgpu_irq_put(adev, &adev->gfx.bad_op_irq, 0);
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++) {
+		gfx_v9_4_3_xcc_fini(adev, i);
+	}
+
+	return 0;
+}
+
+static int gfx_v9_4_3_suspend(struct amdgpu_ip_block *ip_block)
+{
+	return gfx_v9_4_3_hw_fini(ip_block);
+}
+
+static int gfx_v9_4_3_resume(struct amdgpu_ip_block *ip_block)
+{
+	return gfx_v9_4_3_hw_init(ip_block);
+}
+
+static bool gfx_v9_4_3_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++) {
+		if (REG_GET_FIELD(RREG32_SOC15(GC, GET_INST(GC, i), regGRBM_STATUS),
+					GRBM_STATUS, GUI_ACTIVE))
+			return false;
+	}
+	return true;
+}
+
+static int gfx_v9_4_3_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	unsigned i;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (gfx_v9_4_3_is_idle(ip_block))
+			return 0;
+		udelay(1);
+	}
+	return -ETIMEDOUT;
+}
+
+static int gfx_v9_4_3_soft_reset(struct amdgpu_ip_block *ip_block)
+{
+	u32 grbm_soft_reset = 0;
+	u32 tmp;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	/* GRBM_STATUS */
+	tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_STATUS);
+	if (tmp & (GRBM_STATUS__PA_BUSY_MASK | GRBM_STATUS__SC_BUSY_MASK |
+		   GRBM_STATUS__BCI_BUSY_MASK | GRBM_STATUS__SX_BUSY_MASK |
+		   GRBM_STATUS__TA_BUSY_MASK | GRBM_STATUS__VGT_BUSY_MASK |
+		   GRBM_STATUS__DB_BUSY_MASK | GRBM_STATUS__CB_BUSY_MASK |
+		   GRBM_STATUS__GDS_BUSY_MASK | GRBM_STATUS__SPI_BUSY_MASK |
+		   GRBM_STATUS__IA_BUSY_MASK | GRBM_STATUS__IA_BUSY_NO_DMA_MASK)) {
+		grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset,
+						GRBM_SOFT_RESET, SOFT_RESET_CP, 1);
+		grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset,
+						GRBM_SOFT_RESET, SOFT_RESET_GFX, 1);
+	}
+
+	if (tmp & (GRBM_STATUS__CP_BUSY_MASK | GRBM_STATUS__CP_COHERENCY_BUSY_MASK)) {
+		grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset,
+						GRBM_SOFT_RESET, SOFT_RESET_CP, 1);
+	}
+
+	/* GRBM_STATUS2 */
+	tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_STATUS2);
+	if (REG_GET_FIELD(tmp, GRBM_STATUS2, RLC_BUSY))
+		grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset,
+						GRBM_SOFT_RESET, SOFT_RESET_RLC, 1);
+
+
+	if (grbm_soft_reset) {
+		/* stop the rlc */
+		adev->gfx.rlc.funcs->stop(adev);
+
+		/* Disable MEC parsing/prefetching */
+		gfx_v9_4_3_xcc_cp_compute_enable(adev, false, 0);
+
+		tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_SOFT_RESET);
+		tmp |= grbm_soft_reset;
+		dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp);
+		WREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_SOFT_RESET, tmp);
+		tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_SOFT_RESET);
+
+		udelay(50);
+
+		tmp &= ~grbm_soft_reset;
+		WREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_SOFT_RESET, tmp);
+		tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_SOFT_RESET);
+
+		/* Wait a little for things to settle down */
+		udelay(50);
+	}
+	return 0;
+}
+
+static void gfx_v9_4_3_ring_emit_gds_switch(struct amdgpu_ring *ring,
+					  uint32_t vmid,
+					  uint32_t gds_base, uint32_t gds_size,
+					  uint32_t gws_base, uint32_t gws_size,
+					  uint32_t oa_base, uint32_t oa_size)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	/* GDS Base */
+	gfx_v9_4_3_write_data_to_reg(ring, 0, false,
+				   SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regGDS_VMID0_BASE) + 2 * vmid,
+				   gds_base);
+
+	/* GDS Size */
+	gfx_v9_4_3_write_data_to_reg(ring, 0, false,
+				   SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regGDS_VMID0_SIZE) + 2 * vmid,
+				   gds_size);
+
+	/* GWS */
+	gfx_v9_4_3_write_data_to_reg(ring, 0, false,
+				   SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regGDS_GWS_VMID0) + vmid,
+				   gws_size << GDS_GWS_VMID0__SIZE__SHIFT | gws_base);
+
+	/* OA */
+	gfx_v9_4_3_write_data_to_reg(ring, 0, false,
+				   SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regGDS_OA_VMID0) + vmid,
+				   (1 << (oa_size + oa_base)) - (1 << oa_base));
+}
+
+static int gfx_v9_4_3_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
+					  AMDGPU_MAX_COMPUTE_RINGS);
+	gfx_v9_4_3_set_kiq_pm4_funcs(adev);
+	gfx_v9_4_3_set_ring_funcs(adev);
+	gfx_v9_4_3_set_irq_funcs(adev);
+	gfx_v9_4_3_set_gds_init(adev);
+	gfx_v9_4_3_set_rlc_funcs(adev);
+
+	/* init rlcg reg access ctrl */
+	gfx_v9_4_3_init_rlcg_reg_access_ctrl(adev);
 
-const struct amdgpu_rlc_funcs gfx_v9_4_3_rlc_funcs = {
+	return gfx_v9_4_3_init_microcode(adev);
+}
+
+static int gfx_v9_4_3_late_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = amdgpu_irq_get(adev, &adev->gfx.priv_reg_irq, 0);
+	if (r)
+		return r;
+
+	r = amdgpu_irq_get(adev, &adev->gfx.priv_inst_irq, 0);
+	if (r)
+		return r;
+
+	r = amdgpu_irq_get(adev, &adev->gfx.bad_op_irq, 0);
+	if (r)
+		return r;
+
+	if (adev->gfx.ras &&
+	    adev->gfx.ras->enable_watchdog_timer)
+		adev->gfx.ras->enable_watchdog_timer(adev);
+
+	return 0;
+}
+
+static void gfx_v9_4_3_xcc_update_sram_fgcg(struct amdgpu_device *adev,
+					    bool enable, int xcc_id)
+{
+	uint32_t def, data;
+
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_GFX_FGCG))
+		return;
+
+	def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id),
+				  regRLC_CGTT_MGCG_OVERRIDE);
+
+	if (enable)
+		data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_FGCG_OVERRIDE_MASK;
+	else
+		data |= RLC_CGTT_MGCG_OVERRIDE__GFXIP_FGCG_OVERRIDE_MASK;
+
+	if (def != data)
+		WREG32_SOC15(GC, GET_INST(GC, xcc_id),
+			     regRLC_CGTT_MGCG_OVERRIDE, data);
+
+}
+
+static void gfx_v9_4_3_xcc_update_repeater_fgcg(struct amdgpu_device *adev,
+						bool enable, int xcc_id)
+{
+	uint32_t def, data;
+
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_REPEATER_FGCG))
+		return;
+
+	def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id),
+				  regRLC_CGTT_MGCG_OVERRIDE);
+
+	if (enable)
+		data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_REP_FGCG_OVERRIDE_MASK;
+	else
+		data |= RLC_CGTT_MGCG_OVERRIDE__GFXIP_REP_FGCG_OVERRIDE_MASK;
+
+	if (def != data)
+		WREG32_SOC15(GC, GET_INST(GC, xcc_id),
+			     regRLC_CGTT_MGCG_OVERRIDE, data);
+}
+
+static void
+gfx_v9_4_3_xcc_update_medium_grain_clock_gating(struct amdgpu_device *adev,
+						bool enable, int xcc_id)
+{
+	uint32_t data, def;
+
+	/* It is disabled by HW by default */
+	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) {
+		/* 1 - RLC_CGTT_MGCG_OVERRIDE */
+		def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGTT_MGCG_OVERRIDE);
+
+		data &= ~(RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK |
+			  RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK |
+			  RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK |
+			  RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGLS_OVERRIDE_MASK);
+
+		if (def != data)
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGTT_MGCG_OVERRIDE, data);
+
+		/* MGLS is a global flag to control all MGLS in GFX */
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGLS) {
+			/* 2 - RLC memory Light sleep */
+			if (adev->cg_flags & AMD_CG_SUPPORT_GFX_RLC_LS) {
+				def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_MEM_SLP_CNTL);
+				data |= RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK;
+				if (def != data)
+					WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_MEM_SLP_CNTL, data);
+			}
+			/* 3 - CP memory Light sleep */
+			if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CP_LS) {
+				def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_MEM_SLP_CNTL);
+				data |= CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK;
+				if (def != data)
+					WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_MEM_SLP_CNTL, data);
+			}
+		}
+	} else {
+		/* 1 - MGCG_OVERRIDE */
+		def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGTT_MGCG_OVERRIDE);
+
+		data |= (RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK |
+			 RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK |
+			 RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK |
+			 RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGLS_OVERRIDE_MASK);
+
+		if (def != data)
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGTT_MGCG_OVERRIDE, data);
+
+		/* 2 - disable MGLS in RLC */
+		data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_MEM_SLP_CNTL);
+		if (data & RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK) {
+			data &= ~RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK;
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_MEM_SLP_CNTL, data);
+		}
+
+		/* 3 - disable MGLS in CP */
+		data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_MEM_SLP_CNTL);
+		if (data & CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK) {
+			data &= ~CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK;
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_MEM_SLP_CNTL, data);
+		}
+	}
+
+}
+
+static void
+gfx_v9_4_3_xcc_update_coarse_grain_clock_gating(struct amdgpu_device *adev,
+						bool enable, int xcc_id)
+{
+	uint32_t def, data;
+
+	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)) {
+
+		def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGTT_MGCG_OVERRIDE);
+		/* unset CGCG override */
+		data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_CGCG_OVERRIDE_MASK;
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS)
+			data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_CGLS_OVERRIDE_MASK;
+		else
+			data |= RLC_CGTT_MGCG_OVERRIDE__GFXIP_CGLS_OVERRIDE_MASK;
+		/* update CGCG and CGLS override bits */
+		if (def != data)
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGTT_MGCG_OVERRIDE, data);
+
+		/* CGCG Hysteresis: 400us */
+		def = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGCG_CGLS_CTRL);
+
+		data = (0x2710
+			<< RLC_CGCG_CGLS_CTRL__CGCG_GFX_IDLE_THRESHOLD__SHIFT) |
+		       RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK;
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS)
+			data |= (0x000F << RLC_CGCG_CGLS_CTRL__CGLS_REP_COMPANSAT_DELAY__SHIFT) |
+				RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK;
+		if (def != data)
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGCG_CGLS_CTRL, data);
+
+		/* set IDLE_POLL_COUNT(0x33450100)*/
+		def = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_RB_WPTR_POLL_CNTL);
+		data = (0x0100 << CP_RB_WPTR_POLL_CNTL__POLL_FREQUENCY__SHIFT) |
+			(0x3345 << CP_RB_WPTR_POLL_CNTL__IDLE_POLL_COUNT__SHIFT);
+		if (def != data)
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_RB_WPTR_POLL_CNTL, data);
+	} else {
+		def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGCG_CGLS_CTRL);
+		/* reset CGCG/CGLS bits */
+		data &= ~(RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK | RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK);
+		/* disable cgcg and cgls in FSM */
+		if (def != data)
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGCG_CGLS_CTRL, data);
+	}
+
+}
+
+static int gfx_v9_4_3_xcc_update_gfx_clock_gating(struct amdgpu_device *adev,
+						  bool enable, int xcc_id)
+{
+	amdgpu_gfx_rlc_enter_safe_mode(adev, xcc_id);
+
+	if (enable) {
+		/* FGCG */
+		gfx_v9_4_3_xcc_update_sram_fgcg(adev, enable, xcc_id);
+		gfx_v9_4_3_xcc_update_repeater_fgcg(adev, enable, xcc_id);
+
+		/* CGCG/CGLS should be enabled after MGCG/MGLS
+		 * ===  MGCG + MGLS ===
+		 */
+		gfx_v9_4_3_xcc_update_medium_grain_clock_gating(adev, enable,
+								xcc_id);
+		/* ===  CGCG + CGLS === */
+		gfx_v9_4_3_xcc_update_coarse_grain_clock_gating(adev, enable,
+								xcc_id);
+	} else {
+		/* CGCG/CGLS should be disabled before MGCG/MGLS
+		 * ===  CGCG + CGLS ===
+		 */
+		gfx_v9_4_3_xcc_update_coarse_grain_clock_gating(adev, enable,
+								xcc_id);
+		/* ===  MGCG + MGLS === */
+		gfx_v9_4_3_xcc_update_medium_grain_clock_gating(adev, enable,
+								xcc_id);
+
+		/* FGCG */
+		gfx_v9_4_3_xcc_update_sram_fgcg(adev, enable, xcc_id);
+		gfx_v9_4_3_xcc_update_repeater_fgcg(adev, enable, xcc_id);
+	}
+
+	amdgpu_gfx_rlc_exit_safe_mode(adev, xcc_id);
+
+	return 0;
+}
+
+static const struct amdgpu_rlc_funcs gfx_v9_4_3_rlc_funcs = {
 	.is_rlc_enabled = gfx_v9_4_3_is_rlc_enabled,
-	.set_safe_mode = gfx_v9_4_3_set_safe_mode,
-	.unset_safe_mode = gfx_v9_4_3_unset_safe_mode,
+	.set_safe_mode = gfx_v9_4_3_xcc_set_safe_mode,
+	.unset_safe_mode = gfx_v9_4_3_xcc_unset_safe_mode,
 	.init = gfx_v9_4_3_rlc_init,
 	.resume = gfx_v9_4_3_rlc_resume,
 	.stop = gfx_v9_4_3_rlc_stop,
@@ -428,3 +2757,2306 @@ const struct amdgpu_rlc_funcs gfx_v9_4_3_rlc_funcs = {
 	.update_spm_vmid = gfx_v9_4_3_update_spm_vmid,
 	.is_rlcg_access_range = gfx_v9_4_3_is_rlcg_access_range,
 };
+
+static int gfx_v9_4_3_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_powergating_state state)
+{
+	return 0;
+}
+
+static int gfx_v9_4_3_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_clockgating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, num_xcc;
+
+	if (amdgpu_sriov_vf(adev))
+		return 0;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++)
+		gfx_v9_4_3_xcc_update_gfx_clock_gating(
+			adev, state == AMD_CG_STATE_GATE, i);
+
+	return 0;
+}
+
+static void gfx_v9_4_3_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int data;
+
+	if (amdgpu_sriov_vf(adev))
+		*flags = 0;
+
+	/* AMD_CG_SUPPORT_GFX_MGCG */
+	data = RREG32_KIQ(SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regRLC_CGTT_MGCG_OVERRIDE));
+	if (!(data & RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK))
+		*flags |= AMD_CG_SUPPORT_GFX_MGCG;
+
+	/* AMD_CG_SUPPORT_GFX_CGCG */
+	data = RREG32_KIQ(SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regRLC_CGCG_CGLS_CTRL));
+	if (data & RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_GFX_CGCG;
+
+	/* AMD_CG_SUPPORT_GFX_CGLS */
+	if (data & RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_GFX_CGLS;
+
+	/* AMD_CG_SUPPORT_GFX_RLC_LS */
+	data = RREG32_KIQ(SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regRLC_MEM_SLP_CNTL));
+	if (data & RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_GFX_RLC_LS | AMD_CG_SUPPORT_GFX_MGLS;
+
+	/* AMD_CG_SUPPORT_GFX_CP_LS */
+	data = RREG32_KIQ(SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regCP_MEM_SLP_CNTL));
+	if (data & CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_GFX_CP_LS | AMD_CG_SUPPORT_GFX_MGLS;
+}
+
+static void gfx_v9_4_3_ring_emit_hdp_flush(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 ref_and_mask, reg_mem_engine;
+	const struct nbio_hdp_flush_reg *nbio_hf_reg = adev->nbio.hdp_flush_reg;
+
+	if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE) {
+		switch (ring->me) {
+		case 1:
+			ref_and_mask = nbio_hf_reg->ref_and_mask_cp2 << ring->pipe;
+			break;
+		case 2:
+			ref_and_mask = nbio_hf_reg->ref_and_mask_cp6 << ring->pipe;
+			break;
+		default:
+			return;
+		}
+		reg_mem_engine = 0;
+	} else {
+		ref_and_mask = nbio_hf_reg->ref_and_mask_cp0;
+		reg_mem_engine = 1; /* pfp */
+	}
+
+	gfx_v9_4_3_wait_reg_mem(ring, reg_mem_engine, 0, 1,
+			      adev->nbio.funcs->get_hdp_flush_req_offset(adev),
+			      adev->nbio.funcs->get_hdp_flush_done_offset(adev),
+			      ref_and_mask, ref_and_mask, 0x20);
+}
+
+static void gfx_v9_4_3_ring_emit_ib_compute(struct amdgpu_ring *ring,
+					  struct amdgpu_job *job,
+					  struct amdgpu_ib *ib,
+					  uint32_t flags)
+{
+	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
+	u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vmid << 24);
+
+	/* Currently, there is a high possibility to get wave ID mismatch
+	 * between ME and GDS, leading to a hw deadlock, because ME generates
+	 * different wave IDs than the GDS expects. This situation happens
+	 * randomly when at least 5 compute pipes use GDS ordered append.
+	 * The wave IDs generated by ME are also wrong after suspend/resume.
+	 * Those are probably bugs somewhere else in the kernel driver.
+	 *
+	 * Writing GDS_COMPUTE_MAX_WAVE_ID resets wave ID counters in ME and
+	 * GDS to 0 for this ring (me/pipe).
+	 */
+	if (ib->flags & AMDGPU_IB_FLAG_RESET_GDS_MAX_WAVE_ID) {
+		amdgpu_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
+		amdgpu_ring_write(ring, regGDS_COMPUTE_MAX_WAVE_ID);
+		amdgpu_ring_write(ring, ring->adev->gds.gds_compute_max_wave_id);
+	}
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
+	BUG_ON(ib->gpu_addr & 0x3); /* Dword align */
+	amdgpu_ring_write(ring,
+#ifdef __BIG_ENDIAN
+				(2 << 0) |
+#endif
+				lower_32_bits(ib->gpu_addr));
+	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
+	amdgpu_ring_write(ring, control);
+}
+
+static void gfx_v9_4_3_ring_emit_fence(struct amdgpu_ring *ring, u64 addr,
+				     u64 seq, unsigned flags)
+{
+	bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT;
+	bool int_sel = flags & AMDGPU_FENCE_FLAG_INT;
+	bool writeback = flags & AMDGPU_FENCE_FLAG_TC_WB_ONLY;
+
+	/* RELEASE_MEM - flush caches, send int */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_RELEASE_MEM, 6));
+	amdgpu_ring_write(ring, ((writeback ? (EOP_TC_WB_ACTION_EN |
+					       EOP_TC_NC_ACTION_EN) :
+					      (EOP_TCL1_ACTION_EN |
+					       EOP_TC_ACTION_EN |
+					       EOP_TC_WB_ACTION_EN |
+					       EOP_TC_MD_ACTION_EN)) |
+				 EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) |
+				 EVENT_INDEX(5)));
+	amdgpu_ring_write(ring, DATA_SEL(write64bit ? 2 : 1) | INT_SEL(int_sel ? 2 : 0));
+
+	/*
+	 * the address should be Qword aligned if 64bit write, Dword
+	 * aligned if only send 32bit data low (discard data high)
+	 */
+	if (write64bit)
+		BUG_ON(addr & 0x7);
+	else
+		BUG_ON(addr & 0x3);
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	amdgpu_ring_write(ring, lower_32_bits(seq));
+	amdgpu_ring_write(ring, upper_32_bits(seq));
+	amdgpu_ring_write(ring, 0);
+}
+
+static void gfx_v9_4_3_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
+{
+	int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX);
+	uint32_t seq = ring->fence_drv.sync_seq;
+	uint64_t addr = ring->fence_drv.gpu_addr;
+
+	gfx_v9_4_3_wait_reg_mem(ring, usepfp, 1, 0,
+			      lower_32_bits(addr), upper_32_bits(addr),
+			      seq, 0xffffffff, 4);
+}
+
+static void gfx_v9_4_3_ring_emit_vm_flush(struct amdgpu_ring *ring,
+					unsigned vmid, uint64_t pd_addr)
+{
+	amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
+}
+
+static u64 gfx_v9_4_3_ring_get_rptr_compute(struct amdgpu_ring *ring)
+{
+	return ring->adev->wb.wb[ring->rptr_offs]; /* gfx9 hardware is 32bit rptr */
+}
+
+static u64 gfx_v9_4_3_ring_get_wptr_compute(struct amdgpu_ring *ring)
+{
+	u64 wptr;
+
+	/* XXX check if swapping is necessary on BE */
+	if (ring->use_doorbell)
+		wptr = atomic64_read((atomic64_t *)&ring->adev->wb.wb[ring->wptr_offs]);
+	else
+		BUG();
+	return wptr;
+}
+
+static void gfx_v9_4_3_ring_set_wptr_compute(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	/* XXX check if swapping is necessary on BE */
+	if (ring->use_doorbell) {
+		atomic64_set((atomic64_t *)&adev->wb.wb[ring->wptr_offs], ring->wptr);
+		WDOORBELL64(ring->doorbell_index, ring->wptr);
+	} else {
+		BUG(); /* only DOORBELL method supported on gfx9 now */
+	}
+}
+
+static void gfx_v9_4_3_ring_emit_fence_kiq(struct amdgpu_ring *ring, u64 addr,
+					 u64 seq, unsigned int flags)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	/* we only allocate 32bit for each seq wb address */
+	BUG_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
+
+	/* write fence seq to the "addr" */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+				 WRITE_DATA_DST_SEL(5) | WR_CONFIRM));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	amdgpu_ring_write(ring, lower_32_bits(seq));
+
+	if (flags & AMDGPU_FENCE_FLAG_INT) {
+		/* set register to trigger INT */
+		amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+		amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+					 WRITE_DATA_DST_SEL(0) | WR_CONFIRM));
+		amdgpu_ring_write(ring, SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regCPC_INT_STATUS));
+		amdgpu_ring_write(ring, 0);
+		amdgpu_ring_write(ring, 0x20000000); /* src_id is 178 */
+	}
+}
+
+static void gfx_v9_4_3_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg,
+				    uint32_t reg_val_offs)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	reg = gfx_v9_4_3_normalize_xcc_reg_offset(reg);
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_COPY_DATA, 4));
+	amdgpu_ring_write(ring, 0 |	/* src: register*/
+				(5 << 8) |	/* dst: memory */
+				(1 << 20));	/* write confirm */
+	amdgpu_ring_write(ring, reg);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +
+				reg_val_offs * 4));
+	amdgpu_ring_write(ring, upper_32_bits(adev->wb.gpu_addr +
+				reg_val_offs * 4));
+}
+
+static void gfx_v9_4_3_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg,
+				    uint32_t val)
+{
+	uint32_t cmd = 0;
+
+	reg = gfx_v9_4_3_normalize_xcc_reg_offset(reg);
+
+	switch (ring->funcs->type) {
+	case AMDGPU_RING_TYPE_GFX:
+		cmd = WRITE_DATA_ENGINE_SEL(1) | WR_CONFIRM;
+		break;
+	case AMDGPU_RING_TYPE_KIQ:
+		cmd = (1 << 16); /* no inc addr */
+		break;
+	default:
+		cmd = WR_CONFIRM;
+		break;
+	}
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, cmd);
+	amdgpu_ring_write(ring, reg);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, val);
+}
+
+static void gfx_v9_4_3_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
+					uint32_t val, uint32_t mask)
+{
+	gfx_v9_4_3_wait_reg_mem(ring, 0, 0, 0, reg, 0, val, mask, 0x20);
+}
+
+static void gfx_v9_4_3_ring_emit_reg_write_reg_wait(struct amdgpu_ring *ring,
+						  uint32_t reg0, uint32_t reg1,
+						  uint32_t ref, uint32_t mask)
+{
+	amdgpu_ring_emit_reg_write_reg_wait_helper(ring, reg0, reg1,
+						   ref, mask);
+}
+
+static void gfx_v9_4_3_ring_soft_recovery(struct amdgpu_ring *ring,
+					  unsigned vmid)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t value = 0;
+
+	value = REG_SET_FIELD(value, SQ_CMD, CMD, 0x03);
+	value = REG_SET_FIELD(value, SQ_CMD, MODE, 0x01);
+	value = REG_SET_FIELD(value, SQ_CMD, CHECK_VMID, 1);
+	value = REG_SET_FIELD(value, SQ_CMD, VM_ID, vmid);
+	amdgpu_gfx_rlc_enter_safe_mode(adev, ring->xcc_id);
+	WREG32_SOC15(GC, GET_INST(GC, ring->xcc_id), regSQ_CMD, value);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, ring->xcc_id);
+}
+
+static void gfx_v9_4_3_xcc_set_compute_eop_interrupt_state(
+	struct amdgpu_device *adev, int me, int pipe,
+	enum amdgpu_interrupt_state state, int xcc_id)
+{
+	u32 mec_int_cntl, mec_int_cntl_reg;
+
+	/*
+	 * amdgpu controls only the first MEC. That's why this function only
+	 * handles the setting of interrupts for this specific MEC. All other
+	 * pipes' interrupts are set by amdkfd.
+	 */
+
+	if (me == 1) {
+		switch (pipe) {
+		case 0:
+			mec_int_cntl_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE0_INT_CNTL);
+			break;
+		case 1:
+			mec_int_cntl_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE1_INT_CNTL);
+			break;
+		case 2:
+			mec_int_cntl_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE2_INT_CNTL);
+			break;
+		case 3:
+			mec_int_cntl_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE3_INT_CNTL);
+			break;
+		default:
+			DRM_DEBUG("invalid pipe %d\n", pipe);
+			return;
+		}
+	} else {
+		DRM_DEBUG("invalid me %d\n", me);
+		return;
+	}
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		mec_int_cntl = RREG32_XCC(mec_int_cntl_reg, xcc_id);
+		mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+					     TIME_STAMP_INT_ENABLE, 0);
+		WREG32_XCC(mec_int_cntl_reg, mec_int_cntl, xcc_id);
+		break;
+	case AMDGPU_IRQ_STATE_ENABLE:
+		mec_int_cntl = RREG32_XCC(mec_int_cntl_reg, xcc_id);
+		mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+					     TIME_STAMP_INT_ENABLE, 1);
+		WREG32_XCC(mec_int_cntl_reg, mec_int_cntl, xcc_id);
+		break;
+	default:
+		break;
+	}
+}
+
+static u32 gfx_v9_4_3_get_cpc_int_cntl(struct amdgpu_device *adev,
+				     int xcc_id, int me, int pipe)
+{
+	/*
+	 * amdgpu controls only the first MEC. That's why this function only
+	 * handles the setting of interrupts for this specific MEC. All other
+	 * pipes' interrupts are set by amdkfd.
+	 */
+	if (me != 1)
+		return 0;
+
+	switch (pipe) {
+	case 0:
+		return SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE0_INT_CNTL);
+	case 1:
+		return SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE1_INT_CNTL);
+	case 2:
+		return SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE2_INT_CNTL);
+	case 3:
+		return SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE3_INT_CNTL);
+	default:
+		return 0;
+	}
+}
+
+static int gfx_v9_4_3_set_priv_reg_fault_state(struct amdgpu_device *adev,
+					     struct amdgpu_irq_src *source,
+					     unsigned type,
+					     enum amdgpu_interrupt_state state)
+{
+	u32 mec_int_cntl_reg, mec_int_cntl;
+	int i, j, k, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+	case AMDGPU_IRQ_STATE_ENABLE:
+		for (i = 0; i < num_xcc; i++) {
+			WREG32_FIELD15_PREREG(GC, GET_INST(GC, i), CP_INT_CNTL_RING0,
+					      PRIV_REG_INT_ENABLE,
+					      state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+			for (j = 0; j < adev->gfx.mec.num_mec; j++) {
+				for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
+					/* MECs start at 1 */
+					mec_int_cntl_reg = gfx_v9_4_3_get_cpc_int_cntl(adev, i, j + 1, k);
+
+					if (mec_int_cntl_reg) {
+						mec_int_cntl = RREG32_XCC(mec_int_cntl_reg, i);
+						mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+									     PRIV_REG_INT_ENABLE,
+									     state == AMDGPU_IRQ_STATE_ENABLE ?
+									     1 : 0);
+						WREG32_XCC(mec_int_cntl_reg, mec_int_cntl, i);
+					}
+				}
+			}
+		}
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int gfx_v9_4_3_set_bad_op_fault_state(struct amdgpu_device *adev,
+					     struct amdgpu_irq_src *source,
+					     unsigned type,
+					     enum amdgpu_interrupt_state state)
+{
+	u32 mec_int_cntl_reg, mec_int_cntl;
+	int i, j, k, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+	case AMDGPU_IRQ_STATE_ENABLE:
+		for (i = 0; i < num_xcc; i++) {
+			WREG32_FIELD15_PREREG(GC, GET_INST(GC, i), CP_INT_CNTL_RING0,
+					      OPCODE_ERROR_INT_ENABLE,
+					      state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+			for (j = 0; j < adev->gfx.mec.num_mec; j++) {
+				for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
+					/* MECs start at 1 */
+					mec_int_cntl_reg = gfx_v9_4_3_get_cpc_int_cntl(adev, i, j + 1, k);
+
+					if (mec_int_cntl_reg) {
+						mec_int_cntl = RREG32_XCC(mec_int_cntl_reg, i);
+						mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+									     OPCODE_ERROR_INT_ENABLE,
+									     state == AMDGPU_IRQ_STATE_ENABLE ?
+									     1 : 0);
+						WREG32_XCC(mec_int_cntl_reg, mec_int_cntl, i);
+					}
+				}
+			}
+		}
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int gfx_v9_4_3_set_priv_inst_fault_state(struct amdgpu_device *adev,
+					      struct amdgpu_irq_src *source,
+					      unsigned type,
+					      enum amdgpu_interrupt_state state)
+{
+	int i, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+	case AMDGPU_IRQ_STATE_ENABLE:
+		for (i = 0; i < num_xcc; i++)
+			WREG32_FIELD15_PREREG(GC, GET_INST(GC, i), CP_INT_CNTL_RING0,
+				PRIV_INSTR_INT_ENABLE,
+				state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int gfx_v9_4_3_set_eop_interrupt_state(struct amdgpu_device *adev,
+					    struct amdgpu_irq_src *src,
+					    unsigned type,
+					    enum amdgpu_interrupt_state state)
+{
+	int i, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++) {
+		switch (type) {
+		case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP:
+			gfx_v9_4_3_xcc_set_compute_eop_interrupt_state(
+				adev, 1, 0, state, i);
+			break;
+		case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE1_EOP:
+			gfx_v9_4_3_xcc_set_compute_eop_interrupt_state(
+				adev, 1, 1, state, i);
+			break;
+		case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE2_EOP:
+			gfx_v9_4_3_xcc_set_compute_eop_interrupt_state(
+				adev, 1, 2, state, i);
+			break;
+		case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE3_EOP:
+			gfx_v9_4_3_xcc_set_compute_eop_interrupt_state(
+				adev, 1, 3, state, i);
+			break;
+		case AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE0_EOP:
+			gfx_v9_4_3_xcc_set_compute_eop_interrupt_state(
+				adev, 2, 0, state, i);
+			break;
+		case AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE1_EOP:
+			gfx_v9_4_3_xcc_set_compute_eop_interrupt_state(
+				adev, 2, 1, state, i);
+			break;
+		case AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE2_EOP:
+			gfx_v9_4_3_xcc_set_compute_eop_interrupt_state(
+				adev, 2, 2, state, i);
+			break;
+		case AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE3_EOP:
+			gfx_v9_4_3_xcc_set_compute_eop_interrupt_state(
+				adev, 2, 3, state, i);
+			break;
+		default:
+			break;
+		}
+	}
+
+	return 0;
+}
+
+static int gfx_v9_4_3_eop_irq(struct amdgpu_device *adev,
+			    struct amdgpu_irq_src *source,
+			    struct amdgpu_iv_entry *entry)
+{
+	int i, xcc_id;
+	u8 me_id, pipe_id, queue_id;
+	struct amdgpu_ring *ring;
+
+	DRM_DEBUG("IH: CP EOP\n");
+	me_id = (entry->ring_id & 0x0c) >> 2;
+	pipe_id = (entry->ring_id & 0x03) >> 0;
+	queue_id = (entry->ring_id & 0x70) >> 4;
+
+	xcc_id = gfx_v9_4_3_ih_to_xcc_inst(adev, entry->node_id);
+
+	if (xcc_id == -EINVAL)
+		return -EINVAL;
+
+	switch (me_id) {
+	case 0:
+	case 1:
+	case 2:
+		for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+			ring = &adev->gfx.compute_ring
+					[i +
+					 xcc_id * adev->gfx.num_compute_rings];
+			/* Per-queue interrupt is supported for MEC starting from VI.
+			  * The interrupt can only be enabled/disabled per pipe instead of per queue.
+			  */
+
+			if ((ring->me == me_id) && (ring->pipe == pipe_id) && (ring->queue == queue_id))
+				amdgpu_fence_process(ring);
+		}
+		break;
+	}
+	return 0;
+}
+
+static void gfx_v9_4_3_fault(struct amdgpu_device *adev,
+			   struct amdgpu_iv_entry *entry)
+{
+	u8 me_id, pipe_id, queue_id;
+	struct amdgpu_ring *ring;
+	int i, xcc_id;
+
+	me_id = (entry->ring_id & 0x0c) >> 2;
+	pipe_id = (entry->ring_id & 0x03) >> 0;
+	queue_id = (entry->ring_id & 0x70) >> 4;
+
+	xcc_id = gfx_v9_4_3_ih_to_xcc_inst(adev, entry->node_id);
+
+	if (xcc_id == -EINVAL)
+		return;
+
+	switch (me_id) {
+	case 0:
+	case 1:
+	case 2:
+		for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+			ring = &adev->gfx.compute_ring
+					[i +
+					 xcc_id * adev->gfx.num_compute_rings];
+			if (ring->me == me_id && ring->pipe == pipe_id &&
+			    ring->queue == queue_id)
+				drm_sched_fault(&ring->sched);
+		}
+		break;
+	}
+}
+
+static int gfx_v9_4_3_priv_reg_irq(struct amdgpu_device *adev,
+				 struct amdgpu_irq_src *source,
+				 struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal register access in command stream\n");
+	gfx_v9_4_3_fault(adev, entry);
+	return 0;
+}
+
+static int gfx_v9_4_3_bad_op_irq(struct amdgpu_device *adev,
+				 struct amdgpu_irq_src *source,
+				 struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal opcode in command stream\n");
+	gfx_v9_4_3_fault(adev, entry);
+	return 0;
+}
+
+static int gfx_v9_4_3_priv_inst_irq(struct amdgpu_device *adev,
+				  struct amdgpu_irq_src *source,
+				  struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal instruction in command stream\n");
+	gfx_v9_4_3_fault(adev, entry);
+	return 0;
+}
+
+static void gfx_v9_4_3_emit_mem_sync(struct amdgpu_ring *ring)
+{
+	const unsigned int cp_coher_cntl =
+			PACKET3_ACQUIRE_MEM_CP_COHER_CNTL_SH_ICACHE_ACTION_ENA(1) |
+			PACKET3_ACQUIRE_MEM_CP_COHER_CNTL_SH_KCACHE_ACTION_ENA(1) |
+			PACKET3_ACQUIRE_MEM_CP_COHER_CNTL_TC_ACTION_ENA(1) |
+			PACKET3_ACQUIRE_MEM_CP_COHER_CNTL_TCL1_ACTION_ENA(1) |
+			PACKET3_ACQUIRE_MEM_CP_COHER_CNTL_TC_WB_ACTION_ENA(1);
+
+	/* ACQUIRE_MEM -make one or more surfaces valid for use by the subsequent operations */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_ACQUIRE_MEM, 5));
+	amdgpu_ring_write(ring, cp_coher_cntl); /* CP_COHER_CNTL */
+	amdgpu_ring_write(ring, 0xffffffff);  /* CP_COHER_SIZE */
+	amdgpu_ring_write(ring, 0xffffff);  /* CP_COHER_SIZE_HI */
+	amdgpu_ring_write(ring, 0); /* CP_COHER_BASE */
+	amdgpu_ring_write(ring, 0);  /* CP_COHER_BASE_HI */
+	amdgpu_ring_write(ring, 0x0000000A); /* POLL_INTERVAL */
+}
+
+static void gfx_v9_4_3_emit_wave_limit_cs(struct amdgpu_ring *ring,
+					uint32_t pipe, bool enable)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t val;
+	uint32_t wcl_cs_reg;
+
+	/* regSPI_WCL_PIPE_PERCENT_CS[0-7]_DEFAULT values are same */
+	val = enable ? 0x1 : 0x7f;
+
+	switch (pipe) {
+	case 0:
+		wcl_cs_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regSPI_WCL_PIPE_PERCENT_CS0);
+		break;
+	case 1:
+		wcl_cs_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regSPI_WCL_PIPE_PERCENT_CS1);
+		break;
+	case 2:
+		wcl_cs_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regSPI_WCL_PIPE_PERCENT_CS2);
+		break;
+	case 3:
+		wcl_cs_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regSPI_WCL_PIPE_PERCENT_CS3);
+		break;
+	default:
+		DRM_DEBUG("invalid pipe %d\n", pipe);
+		return;
+	}
+
+	amdgpu_ring_emit_wreg(ring, wcl_cs_reg, val);
+
+}
+static void gfx_v9_4_3_emit_wave_limit(struct amdgpu_ring *ring, bool enable)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t val;
+	int i;
+
+	/* regSPI_WCL_PIPE_PERCENT_GFX is 7 bit multiplier register to limit
+	 * number of gfx waves. Setting 5 bit will make sure gfx only gets
+	 * around 25% of gpu resources.
+	 */
+	val = enable ? 0x1f : 0x07ffffff;
+	amdgpu_ring_emit_wreg(ring,
+			      SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regSPI_WCL_PIPE_PERCENT_GFX),
+			      val);
+
+	/* Restrict waves for normal/low priority compute queues as well
+	 * to get best QoS for high priority compute jobs.
+	 *
+	 * amdgpu controls only 1st ME(0-3 CS pipes).
+	 */
+	for (i = 0; i < adev->gfx.mec.num_pipe_per_mec; i++) {
+		if (i != ring->pipe)
+			gfx_v9_4_3_emit_wave_limit_cs(ring, i, enable);
+
+	}
+}
+
+static int gfx_v9_4_3_unmap_done(struct amdgpu_device *adev, uint32_t me,
+				uint32_t pipe, uint32_t queue,
+				uint32_t xcc_id)
+{
+	int i, r;
+	/* make sure dequeue is complete*/
+	gfx_v9_4_3_xcc_set_safe_mode(adev, xcc_id);
+	mutex_lock(&adev->srbm_mutex);
+	soc15_grbm_select(adev, me, pipe, queue, 0, GET_INST(GC, xcc_id));
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (!(RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE) & 1))
+			break;
+		udelay(1);
+	}
+	if (i >= adev->usec_timeout)
+		r = -ETIMEDOUT;
+	else
+		r = 0;
+	soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id));
+	mutex_unlock(&adev->srbm_mutex);
+	gfx_v9_4_3_xcc_unset_safe_mode(adev, xcc_id);
+
+	return r;
+
+}
+
+static bool gfx_v9_4_3_pipe_reset_support(struct amdgpu_device *adev)
+{
+	if (!!(adev->gfx.compute_supported_reset & AMDGPU_RESET_TYPE_PER_PIPE))
+		return true;
+	else
+		dev_warn_once(adev->dev, "Please use the latest MEC version to see whether support pipe reset\n");
+
+	return false;
+}
+
+static int gfx_v9_4_3_reset_hw_pipe(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t reset_pipe, clean_pipe;
+	int r;
+
+	if (!gfx_v9_4_3_pipe_reset_support(adev))
+		return -EINVAL;
+
+	gfx_v9_4_3_xcc_set_safe_mode(adev, ring->xcc_id);
+	mutex_lock(&adev->srbm_mutex);
+
+	reset_pipe = RREG32_SOC15(GC, GET_INST(GC, ring->xcc_id), regCP_MEC_CNTL);
+	clean_pipe = reset_pipe;
+
+	if (ring->me == 1) {
+		switch (ring->pipe) {
+		case 0:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+						   MEC_ME1_PIPE0_RESET, 1);
+			break;
+		case 1:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+						   MEC_ME1_PIPE1_RESET, 1);
+			break;
+		case 2:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+						   MEC_ME1_PIPE2_RESET, 1);
+			break;
+		case 3:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+						   MEC_ME1_PIPE3_RESET, 1);
+			break;
+		default:
+			break;
+		}
+	} else {
+		if (ring->pipe)
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+						   MEC_ME2_PIPE1_RESET, 1);
+		else
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+						   MEC_ME2_PIPE0_RESET, 1);
+	}
+
+	WREG32_SOC15(GC, GET_INST(GC, ring->xcc_id), regCP_MEC_CNTL, reset_pipe);
+	WREG32_SOC15(GC, GET_INST(GC, ring->xcc_id), regCP_MEC_CNTL, clean_pipe);
+	mutex_unlock(&adev->srbm_mutex);
+	gfx_v9_4_3_xcc_unset_safe_mode(adev, ring->xcc_id);
+
+	r = gfx_v9_4_3_unmap_done(adev, ring->me, ring->pipe, ring->queue, ring->xcc_id);
+	return r;
+}
+
+static int gfx_v9_4_3_reset_kcq(struct amdgpu_ring *ring,
+				unsigned int vmid,
+				struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[ring->xcc_id];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
+	int reset_mode = AMDGPU_RESET_TYPE_PER_QUEUE;
+	unsigned long flags;
+	int r;
+
+	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
+		return -EINVAL;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+
+	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size)) {
+		spin_unlock_irqrestore(&kiq->ring_lock, flags);
+		return -ENOMEM;
+	}
+
+	kiq->pmf->kiq_unmap_queues(kiq_ring, ring, RESET_QUEUES,
+				   0, 0);
+	amdgpu_ring_commit(kiq_ring);
+
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+
+	r = amdgpu_ring_test_ring(kiq_ring);
+	if (r) {
+		dev_err(adev->dev, "kiq ring test failed after ring: %s queue reset\n",
+				ring->name);
+		goto pipe_reset;
+	}
+
+	r = gfx_v9_4_3_unmap_done(adev, ring->me, ring->pipe, ring->queue, ring->xcc_id);
+	if (r)
+		dev_err(adev->dev, "fail to wait on hqd deactive and will try pipe reset\n");
+
+pipe_reset:
+	if (r) {
+		if (!(adev->gfx.compute_supported_reset & AMDGPU_RESET_TYPE_PER_PIPE))
+			return -EOPNOTSUPP;
+		r = gfx_v9_4_3_reset_hw_pipe(ring);
+		reset_mode = AMDGPU_RESET_TYPE_PER_PIPE;
+		dev_info(adev->dev, "ring: %s pipe reset :%s\n", ring->name,
+				r ? "failed" : "successfully");
+		if (r)
+			return r;
+	}
+
+	gfx_v9_4_3_xcc_kcq_init_queue(ring, ring->xcc_id, true);
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+	r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size);
+	if (r) {
+		spin_unlock_irqrestore(&kiq->ring_lock, flags);
+		return -ENOMEM;
+	}
+	kiq->pmf->kiq_map_queues(kiq_ring, ring);
+	amdgpu_ring_commit(kiq_ring);
+	r = amdgpu_ring_test_ring(kiq_ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+	if (r) {
+		if (reset_mode == AMDGPU_RESET_TYPE_PER_QUEUE)
+			goto pipe_reset;
+
+		dev_err(adev->dev, "fail to remap queue\n");
+		return r;
+	}
+
+	if (reset_mode == AMDGPU_RESET_TYPE_PER_QUEUE) {
+		r = amdgpu_ring_test_ring(ring);
+		if (r)
+			goto pipe_reset;
+	}
+
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+enum amdgpu_gfx_cp_ras_mem_id {
+	AMDGPU_GFX_CP_MEM1 = 1,
+	AMDGPU_GFX_CP_MEM2,
+	AMDGPU_GFX_CP_MEM3,
+	AMDGPU_GFX_CP_MEM4,
+	AMDGPU_GFX_CP_MEM5,
+};
+
+enum amdgpu_gfx_gcea_ras_mem_id {
+	AMDGPU_GFX_GCEA_IOWR_CMDMEM = 4,
+	AMDGPU_GFX_GCEA_IORD_CMDMEM,
+	AMDGPU_GFX_GCEA_GMIWR_CMDMEM,
+	AMDGPU_GFX_GCEA_GMIRD_CMDMEM,
+	AMDGPU_GFX_GCEA_DRAMWR_CMDMEM,
+	AMDGPU_GFX_GCEA_DRAMRD_CMDMEM,
+	AMDGPU_GFX_GCEA_MAM_DMEM0,
+	AMDGPU_GFX_GCEA_MAM_DMEM1,
+	AMDGPU_GFX_GCEA_MAM_DMEM2,
+	AMDGPU_GFX_GCEA_MAM_DMEM3,
+	AMDGPU_GFX_GCEA_MAM_AMEM0,
+	AMDGPU_GFX_GCEA_MAM_AMEM1,
+	AMDGPU_GFX_GCEA_MAM_AMEM2,
+	AMDGPU_GFX_GCEA_MAM_AMEM3,
+	AMDGPU_GFX_GCEA_MAM_AFLUSH_BUFFER,
+	AMDGPU_GFX_GCEA_WRET_TAGMEM,
+	AMDGPU_GFX_GCEA_RRET_TAGMEM,
+	AMDGPU_GFX_GCEA_IOWR_DATAMEM,
+	AMDGPU_GFX_GCEA_GMIWR_DATAMEM,
+	AMDGPU_GFX_GCEA_DRAM_DATAMEM,
+};
+
+enum amdgpu_gfx_gc_cane_ras_mem_id {
+	AMDGPU_GFX_GC_CANE_MEM0 = 0,
+};
+
+enum amdgpu_gfx_gcutcl2_ras_mem_id {
+	AMDGPU_GFX_GCUTCL2_MEM2P512X95 = 160,
+};
+
+enum amdgpu_gfx_gds_ras_mem_id {
+	AMDGPU_GFX_GDS_MEM0 = 0,
+};
+
+enum amdgpu_gfx_lds_ras_mem_id {
+	AMDGPU_GFX_LDS_BANK0 = 0,
+	AMDGPU_GFX_LDS_BANK1,
+	AMDGPU_GFX_LDS_BANK2,
+	AMDGPU_GFX_LDS_BANK3,
+	AMDGPU_GFX_LDS_BANK4,
+	AMDGPU_GFX_LDS_BANK5,
+	AMDGPU_GFX_LDS_BANK6,
+	AMDGPU_GFX_LDS_BANK7,
+	AMDGPU_GFX_LDS_BANK8,
+	AMDGPU_GFX_LDS_BANK9,
+	AMDGPU_GFX_LDS_BANK10,
+	AMDGPU_GFX_LDS_BANK11,
+	AMDGPU_GFX_LDS_BANK12,
+	AMDGPU_GFX_LDS_BANK13,
+	AMDGPU_GFX_LDS_BANK14,
+	AMDGPU_GFX_LDS_BANK15,
+	AMDGPU_GFX_LDS_BANK16,
+	AMDGPU_GFX_LDS_BANK17,
+	AMDGPU_GFX_LDS_BANK18,
+	AMDGPU_GFX_LDS_BANK19,
+	AMDGPU_GFX_LDS_BANK20,
+	AMDGPU_GFX_LDS_BANK21,
+	AMDGPU_GFX_LDS_BANK22,
+	AMDGPU_GFX_LDS_BANK23,
+	AMDGPU_GFX_LDS_BANK24,
+	AMDGPU_GFX_LDS_BANK25,
+	AMDGPU_GFX_LDS_BANK26,
+	AMDGPU_GFX_LDS_BANK27,
+	AMDGPU_GFX_LDS_BANK28,
+	AMDGPU_GFX_LDS_BANK29,
+	AMDGPU_GFX_LDS_BANK30,
+	AMDGPU_GFX_LDS_BANK31,
+	AMDGPU_GFX_LDS_SP_BUFFER_A,
+	AMDGPU_GFX_LDS_SP_BUFFER_B,
+};
+
+enum amdgpu_gfx_rlc_ras_mem_id {
+	AMDGPU_GFX_RLC_GPMF32 = 1,
+	AMDGPU_GFX_RLC_RLCVF32,
+	AMDGPU_GFX_RLC_SCRATCH,
+	AMDGPU_GFX_RLC_SRM_ARAM,
+	AMDGPU_GFX_RLC_SRM_DRAM,
+	AMDGPU_GFX_RLC_TCTAG,
+	AMDGPU_GFX_RLC_SPM_SE,
+	AMDGPU_GFX_RLC_SPM_GRBMT,
+};
+
+enum amdgpu_gfx_sp_ras_mem_id {
+	AMDGPU_GFX_SP_SIMDID0 = 0,
+};
+
+enum amdgpu_gfx_spi_ras_mem_id {
+	AMDGPU_GFX_SPI_MEM0 = 0,
+	AMDGPU_GFX_SPI_MEM1,
+	AMDGPU_GFX_SPI_MEM2,
+	AMDGPU_GFX_SPI_MEM3,
+};
+
+enum amdgpu_gfx_sqc_ras_mem_id {
+	AMDGPU_GFX_SQC_INST_CACHE_A = 100,
+	AMDGPU_GFX_SQC_INST_CACHE_B = 101,
+	AMDGPU_GFX_SQC_INST_CACHE_TAG_A = 102,
+	AMDGPU_GFX_SQC_INST_CACHE_TAG_B = 103,
+	AMDGPU_GFX_SQC_INST_CACHE_MISS_FIFO_A = 104,
+	AMDGPU_GFX_SQC_INST_CACHE_MISS_FIFO_B = 105,
+	AMDGPU_GFX_SQC_INST_CACHE_GATCL1_MISS_FIFO_A = 106,
+	AMDGPU_GFX_SQC_INST_CACHE_GATCL1_MISS_FIFO_B = 107,
+	AMDGPU_GFX_SQC_DATA_CACHE_A = 200,
+	AMDGPU_GFX_SQC_DATA_CACHE_B = 201,
+	AMDGPU_GFX_SQC_DATA_CACHE_TAG_A = 202,
+	AMDGPU_GFX_SQC_DATA_CACHE_TAG_B = 203,
+	AMDGPU_GFX_SQC_DATA_CACHE_MISS_FIFO_A = 204,
+	AMDGPU_GFX_SQC_DATA_CACHE_MISS_FIFO_B = 205,
+	AMDGPU_GFX_SQC_DATA_CACHE_HIT_FIFO_A = 206,
+	AMDGPU_GFX_SQC_DATA_CACHE_HIT_FIFO_B = 207,
+	AMDGPU_GFX_SQC_DIRTY_BIT_A = 208,
+	AMDGPU_GFX_SQC_DIRTY_BIT_B = 209,
+	AMDGPU_GFX_SQC_WRITE_DATA_BUFFER_CU0 = 210,
+	AMDGPU_GFX_SQC_WRITE_DATA_BUFFER_CU1 = 211,
+	AMDGPU_GFX_SQC_UTCL1_MISS_LFIFO_DATA_CACHE_A = 212,
+	AMDGPU_GFX_SQC_UTCL1_MISS_LFIFO_DATA_CACHE_B = 213,
+	AMDGPU_GFX_SQC_UTCL1_MISS_LFIFO_INST_CACHE = 108,
+};
+
+enum amdgpu_gfx_sq_ras_mem_id {
+	AMDGPU_GFX_SQ_SGPR_MEM0 = 0,
+	AMDGPU_GFX_SQ_SGPR_MEM1,
+	AMDGPU_GFX_SQ_SGPR_MEM2,
+	AMDGPU_GFX_SQ_SGPR_MEM3,
+};
+
+enum amdgpu_gfx_ta_ras_mem_id {
+	AMDGPU_GFX_TA_FS_AFIFO_RAM_LO = 1,
+	AMDGPU_GFX_TA_FS_AFIFO_RAM_HI,
+	AMDGPU_GFX_TA_FS_CFIFO_RAM,
+	AMDGPU_GFX_TA_FSX_LFIFO,
+	AMDGPU_GFX_TA_FS_DFIFO_RAM,
+};
+
+enum amdgpu_gfx_tcc_ras_mem_id {
+	AMDGPU_GFX_TCC_MEM1 = 1,
+};
+
+enum amdgpu_gfx_tca_ras_mem_id {
+	AMDGPU_GFX_TCA_MEM1 = 1,
+};
+
+enum amdgpu_gfx_tci_ras_mem_id {
+	AMDGPU_GFX_TCIW_MEM = 1,
+};
+
+enum amdgpu_gfx_tcp_ras_mem_id {
+	AMDGPU_GFX_TCP_LFIFO0 = 1,
+	AMDGPU_GFX_TCP_SET0BANK0_RAM,
+	AMDGPU_GFX_TCP_SET0BANK1_RAM,
+	AMDGPU_GFX_TCP_SET0BANK2_RAM,
+	AMDGPU_GFX_TCP_SET0BANK3_RAM,
+	AMDGPU_GFX_TCP_SET1BANK0_RAM,
+	AMDGPU_GFX_TCP_SET1BANK1_RAM,
+	AMDGPU_GFX_TCP_SET1BANK2_RAM,
+	AMDGPU_GFX_TCP_SET1BANK3_RAM,
+	AMDGPU_GFX_TCP_SET2BANK0_RAM,
+	AMDGPU_GFX_TCP_SET2BANK1_RAM,
+	AMDGPU_GFX_TCP_SET2BANK2_RAM,
+	AMDGPU_GFX_TCP_SET2BANK3_RAM,
+	AMDGPU_GFX_TCP_SET3BANK0_RAM,
+	AMDGPU_GFX_TCP_SET3BANK1_RAM,
+	AMDGPU_GFX_TCP_SET3BANK2_RAM,
+	AMDGPU_GFX_TCP_SET3BANK3_RAM,
+	AMDGPU_GFX_TCP_VM_FIFO,
+	AMDGPU_GFX_TCP_DB_TAGRAM0,
+	AMDGPU_GFX_TCP_DB_TAGRAM1,
+	AMDGPU_GFX_TCP_DB_TAGRAM2,
+	AMDGPU_GFX_TCP_DB_TAGRAM3,
+	AMDGPU_GFX_TCP_UTCL1_LFIFO_PROBE0,
+	AMDGPU_GFX_TCP_UTCL1_LFIFO_PROBE1,
+	AMDGPU_GFX_TCP_CMD_FIFO,
+};
+
+enum amdgpu_gfx_td_ras_mem_id {
+	AMDGPU_GFX_TD_UTD_CS_FIFO_MEM = 1,
+	AMDGPU_GFX_TD_UTD_SS_FIFO_LO_MEM,
+	AMDGPU_GFX_TD_UTD_SS_FIFO_HI_MEM,
+};
+
+enum amdgpu_gfx_tcx_ras_mem_id {
+	AMDGPU_GFX_TCX_FIFOD0 = 0,
+	AMDGPU_GFX_TCX_FIFOD1,
+	AMDGPU_GFX_TCX_FIFOD2,
+	AMDGPU_GFX_TCX_FIFOD3,
+	AMDGPU_GFX_TCX_FIFOD4,
+	AMDGPU_GFX_TCX_FIFOD5,
+	AMDGPU_GFX_TCX_FIFOD6,
+	AMDGPU_GFX_TCX_FIFOD7,
+	AMDGPU_GFX_TCX_FIFOB0,
+	AMDGPU_GFX_TCX_FIFOB1,
+	AMDGPU_GFX_TCX_FIFOB2,
+	AMDGPU_GFX_TCX_FIFOB3,
+	AMDGPU_GFX_TCX_FIFOB4,
+	AMDGPU_GFX_TCX_FIFOB5,
+	AMDGPU_GFX_TCX_FIFOB6,
+	AMDGPU_GFX_TCX_FIFOB7,
+	AMDGPU_GFX_TCX_FIFOA0,
+	AMDGPU_GFX_TCX_FIFOA1,
+	AMDGPU_GFX_TCX_FIFOA2,
+	AMDGPU_GFX_TCX_FIFOA3,
+	AMDGPU_GFX_TCX_FIFOA4,
+	AMDGPU_GFX_TCX_FIFOA5,
+	AMDGPU_GFX_TCX_FIFOA6,
+	AMDGPU_GFX_TCX_FIFOA7,
+	AMDGPU_GFX_TCX_CFIFO0,
+	AMDGPU_GFX_TCX_CFIFO1,
+	AMDGPU_GFX_TCX_CFIFO2,
+	AMDGPU_GFX_TCX_CFIFO3,
+	AMDGPU_GFX_TCX_CFIFO4,
+	AMDGPU_GFX_TCX_CFIFO5,
+	AMDGPU_GFX_TCX_CFIFO6,
+	AMDGPU_GFX_TCX_CFIFO7,
+	AMDGPU_GFX_TCX_FIFO_ACKB0,
+	AMDGPU_GFX_TCX_FIFO_ACKB1,
+	AMDGPU_GFX_TCX_FIFO_ACKB2,
+	AMDGPU_GFX_TCX_FIFO_ACKB3,
+	AMDGPU_GFX_TCX_FIFO_ACKB4,
+	AMDGPU_GFX_TCX_FIFO_ACKB5,
+	AMDGPU_GFX_TCX_FIFO_ACKB6,
+	AMDGPU_GFX_TCX_FIFO_ACKB7,
+	AMDGPU_GFX_TCX_FIFO_ACKD0,
+	AMDGPU_GFX_TCX_FIFO_ACKD1,
+	AMDGPU_GFX_TCX_FIFO_ACKD2,
+	AMDGPU_GFX_TCX_FIFO_ACKD3,
+	AMDGPU_GFX_TCX_FIFO_ACKD4,
+	AMDGPU_GFX_TCX_FIFO_ACKD5,
+	AMDGPU_GFX_TCX_FIFO_ACKD6,
+	AMDGPU_GFX_TCX_FIFO_ACKD7,
+	AMDGPU_GFX_TCX_DST_FIFOA0,
+	AMDGPU_GFX_TCX_DST_FIFOA1,
+	AMDGPU_GFX_TCX_DST_FIFOA2,
+	AMDGPU_GFX_TCX_DST_FIFOA3,
+	AMDGPU_GFX_TCX_DST_FIFOA4,
+	AMDGPU_GFX_TCX_DST_FIFOA5,
+	AMDGPU_GFX_TCX_DST_FIFOA6,
+	AMDGPU_GFX_TCX_DST_FIFOA7,
+	AMDGPU_GFX_TCX_DST_FIFOB0,
+	AMDGPU_GFX_TCX_DST_FIFOB1,
+	AMDGPU_GFX_TCX_DST_FIFOB2,
+	AMDGPU_GFX_TCX_DST_FIFOB3,
+	AMDGPU_GFX_TCX_DST_FIFOB4,
+	AMDGPU_GFX_TCX_DST_FIFOB5,
+	AMDGPU_GFX_TCX_DST_FIFOB6,
+	AMDGPU_GFX_TCX_DST_FIFOB7,
+	AMDGPU_GFX_TCX_DST_FIFOD0,
+	AMDGPU_GFX_TCX_DST_FIFOD1,
+	AMDGPU_GFX_TCX_DST_FIFOD2,
+	AMDGPU_GFX_TCX_DST_FIFOD3,
+	AMDGPU_GFX_TCX_DST_FIFOD4,
+	AMDGPU_GFX_TCX_DST_FIFOD5,
+	AMDGPU_GFX_TCX_DST_FIFOD6,
+	AMDGPU_GFX_TCX_DST_FIFOD7,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKB0,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKB1,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKB2,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKB3,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKB4,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKB5,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKB6,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKB7,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKD0,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKD1,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKD2,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKD3,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKD4,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKD5,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKD6,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKD7,
+};
+
+enum amdgpu_gfx_atc_l2_ras_mem_id {
+	AMDGPU_GFX_ATC_L2_MEM0 = 0,
+};
+
+enum amdgpu_gfx_utcl2_ras_mem_id {
+	AMDGPU_GFX_UTCL2_MEM0 = 0,
+};
+
+enum amdgpu_gfx_vml2_ras_mem_id {
+	AMDGPU_GFX_VML2_MEM0 = 0,
+};
+
+enum amdgpu_gfx_vml2_walker_ras_mem_id {
+	AMDGPU_GFX_VML2_WALKER_MEM0 = 0,
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_cp_mem_list[] = {
+	{AMDGPU_GFX_CP_MEM1, "CP_MEM1"},
+	{AMDGPU_GFX_CP_MEM2, "CP_MEM2"},
+	{AMDGPU_GFX_CP_MEM3, "CP_MEM3"},
+	{AMDGPU_GFX_CP_MEM4, "CP_MEM4"},
+	{AMDGPU_GFX_CP_MEM5, "CP_MEM5"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_gcea_mem_list[] = {
+	{AMDGPU_GFX_GCEA_IOWR_CMDMEM, "GCEA_IOWR_CMDMEM"},
+	{AMDGPU_GFX_GCEA_IORD_CMDMEM, "GCEA_IORD_CMDMEM"},
+	{AMDGPU_GFX_GCEA_GMIWR_CMDMEM, "GCEA_GMIWR_CMDMEM"},
+	{AMDGPU_GFX_GCEA_GMIRD_CMDMEM, "GCEA_GMIRD_CMDMEM"},
+	{AMDGPU_GFX_GCEA_DRAMWR_CMDMEM, "GCEA_DRAMWR_CMDMEM"},
+	{AMDGPU_GFX_GCEA_DRAMRD_CMDMEM, "GCEA_DRAMRD_CMDMEM"},
+	{AMDGPU_GFX_GCEA_MAM_DMEM0, "GCEA_MAM_DMEM0"},
+	{AMDGPU_GFX_GCEA_MAM_DMEM1, "GCEA_MAM_DMEM1"},
+	{AMDGPU_GFX_GCEA_MAM_DMEM2, "GCEA_MAM_DMEM2"},
+	{AMDGPU_GFX_GCEA_MAM_DMEM3, "GCEA_MAM_DMEM3"},
+	{AMDGPU_GFX_GCEA_MAM_AMEM0, "GCEA_MAM_AMEM0"},
+	{AMDGPU_GFX_GCEA_MAM_AMEM1, "GCEA_MAM_AMEM1"},
+	{AMDGPU_GFX_GCEA_MAM_AMEM2, "GCEA_MAM_AMEM2"},
+	{AMDGPU_GFX_GCEA_MAM_AMEM3, "GCEA_MAM_AMEM3"},
+	{AMDGPU_GFX_GCEA_MAM_AFLUSH_BUFFER, "GCEA_MAM_AFLUSH_BUFFER"},
+	{AMDGPU_GFX_GCEA_WRET_TAGMEM, "GCEA_WRET_TAGMEM"},
+	{AMDGPU_GFX_GCEA_RRET_TAGMEM, "GCEA_RRET_TAGMEM"},
+	{AMDGPU_GFX_GCEA_IOWR_DATAMEM, "GCEA_IOWR_DATAMEM"},
+	{AMDGPU_GFX_GCEA_GMIWR_DATAMEM, "GCEA_GMIWR_DATAMEM"},
+	{AMDGPU_GFX_GCEA_DRAM_DATAMEM, "GCEA_DRAM_DATAMEM"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_gc_cane_mem_list[] = {
+	{AMDGPU_GFX_GC_CANE_MEM0, "GC_CANE_MEM0"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_gcutcl2_mem_list[] = {
+	{AMDGPU_GFX_GCUTCL2_MEM2P512X95, "GCUTCL2_MEM2P512X95"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_gds_mem_list[] = {
+	{AMDGPU_GFX_GDS_MEM0, "GDS_MEM"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_lds_mem_list[] = {
+	{AMDGPU_GFX_LDS_BANK0, "LDS_BANK0"},
+	{AMDGPU_GFX_LDS_BANK1, "LDS_BANK1"},
+	{AMDGPU_GFX_LDS_BANK2, "LDS_BANK2"},
+	{AMDGPU_GFX_LDS_BANK3, "LDS_BANK3"},
+	{AMDGPU_GFX_LDS_BANK4, "LDS_BANK4"},
+	{AMDGPU_GFX_LDS_BANK5, "LDS_BANK5"},
+	{AMDGPU_GFX_LDS_BANK6, "LDS_BANK6"},
+	{AMDGPU_GFX_LDS_BANK7, "LDS_BANK7"},
+	{AMDGPU_GFX_LDS_BANK8, "LDS_BANK8"},
+	{AMDGPU_GFX_LDS_BANK9, "LDS_BANK9"},
+	{AMDGPU_GFX_LDS_BANK10, "LDS_BANK10"},
+	{AMDGPU_GFX_LDS_BANK11, "LDS_BANK11"},
+	{AMDGPU_GFX_LDS_BANK12, "LDS_BANK12"},
+	{AMDGPU_GFX_LDS_BANK13, "LDS_BANK13"},
+	{AMDGPU_GFX_LDS_BANK14, "LDS_BANK14"},
+	{AMDGPU_GFX_LDS_BANK15, "LDS_BANK15"},
+	{AMDGPU_GFX_LDS_BANK16, "LDS_BANK16"},
+	{AMDGPU_GFX_LDS_BANK17, "LDS_BANK17"},
+	{AMDGPU_GFX_LDS_BANK18, "LDS_BANK18"},
+	{AMDGPU_GFX_LDS_BANK19, "LDS_BANK19"},
+	{AMDGPU_GFX_LDS_BANK20, "LDS_BANK20"},
+	{AMDGPU_GFX_LDS_BANK21, "LDS_BANK21"},
+	{AMDGPU_GFX_LDS_BANK22, "LDS_BANK22"},
+	{AMDGPU_GFX_LDS_BANK23, "LDS_BANK23"},
+	{AMDGPU_GFX_LDS_BANK24, "LDS_BANK24"},
+	{AMDGPU_GFX_LDS_BANK25, "LDS_BANK25"},
+	{AMDGPU_GFX_LDS_BANK26, "LDS_BANK26"},
+	{AMDGPU_GFX_LDS_BANK27, "LDS_BANK27"},
+	{AMDGPU_GFX_LDS_BANK28, "LDS_BANK28"},
+	{AMDGPU_GFX_LDS_BANK29, "LDS_BANK29"},
+	{AMDGPU_GFX_LDS_BANK30, "LDS_BANK30"},
+	{AMDGPU_GFX_LDS_BANK31, "LDS_BANK31"},
+	{AMDGPU_GFX_LDS_SP_BUFFER_A, "LDS_SP_BUFFER_A"},
+	{AMDGPU_GFX_LDS_SP_BUFFER_B, "LDS_SP_BUFFER_B"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_rlc_mem_list[] = {
+	{AMDGPU_GFX_RLC_GPMF32, "RLC_GPMF32"},
+	{AMDGPU_GFX_RLC_RLCVF32, "RLC_RLCVF32"},
+	{AMDGPU_GFX_RLC_SCRATCH, "RLC_SCRATCH"},
+	{AMDGPU_GFX_RLC_SRM_ARAM, "RLC_SRM_ARAM"},
+	{AMDGPU_GFX_RLC_SRM_DRAM, "RLC_SRM_DRAM"},
+	{AMDGPU_GFX_RLC_TCTAG, "RLC_TCTAG"},
+	{AMDGPU_GFX_RLC_SPM_SE, "RLC_SPM_SE"},
+	{AMDGPU_GFX_RLC_SPM_GRBMT, "RLC_SPM_GRBMT"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_sp_mem_list[] = {
+	{AMDGPU_GFX_SP_SIMDID0, "SP_SIMDID0"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_spi_mem_list[] = {
+	{AMDGPU_GFX_SPI_MEM0, "SPI_MEM0"},
+	{AMDGPU_GFX_SPI_MEM1, "SPI_MEM1"},
+	{AMDGPU_GFX_SPI_MEM2, "SPI_MEM2"},
+	{AMDGPU_GFX_SPI_MEM3, "SPI_MEM3"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_sqc_mem_list[] = {
+	{AMDGPU_GFX_SQC_INST_CACHE_A, "SQC_INST_CACHE_A"},
+	{AMDGPU_GFX_SQC_INST_CACHE_B, "SQC_INST_CACHE_B"},
+	{AMDGPU_GFX_SQC_INST_CACHE_TAG_A, "SQC_INST_CACHE_TAG_A"},
+	{AMDGPU_GFX_SQC_INST_CACHE_TAG_B, "SQC_INST_CACHE_TAG_B"},
+	{AMDGPU_GFX_SQC_INST_CACHE_MISS_FIFO_A, "SQC_INST_CACHE_MISS_FIFO_A"},
+	{AMDGPU_GFX_SQC_INST_CACHE_MISS_FIFO_B, "SQC_INST_CACHE_MISS_FIFO_B"},
+	{AMDGPU_GFX_SQC_INST_CACHE_GATCL1_MISS_FIFO_A, "SQC_INST_CACHE_GATCL1_MISS_FIFO_A"},
+	{AMDGPU_GFX_SQC_INST_CACHE_GATCL1_MISS_FIFO_B, "SQC_INST_CACHE_GATCL1_MISS_FIFO_B"},
+	{AMDGPU_GFX_SQC_DATA_CACHE_A, "SQC_DATA_CACHE_A"},
+	{AMDGPU_GFX_SQC_DATA_CACHE_B, "SQC_DATA_CACHE_B"},
+	{AMDGPU_GFX_SQC_DATA_CACHE_TAG_A, "SQC_DATA_CACHE_TAG_A"},
+	{AMDGPU_GFX_SQC_DATA_CACHE_TAG_B, "SQC_DATA_CACHE_TAG_B"},
+	{AMDGPU_GFX_SQC_DATA_CACHE_MISS_FIFO_A, "SQC_DATA_CACHE_MISS_FIFO_A"},
+	{AMDGPU_GFX_SQC_DATA_CACHE_MISS_FIFO_B, "SQC_DATA_CACHE_MISS_FIFO_B"},
+	{AMDGPU_GFX_SQC_DATA_CACHE_HIT_FIFO_A, "SQC_DATA_CACHE_HIT_FIFO_A"},
+	{AMDGPU_GFX_SQC_DATA_CACHE_HIT_FIFO_B, "SQC_DATA_CACHE_HIT_FIFO_B"},
+	{AMDGPU_GFX_SQC_DIRTY_BIT_A, "SQC_DIRTY_BIT_A"},
+	{AMDGPU_GFX_SQC_DIRTY_BIT_B, "SQC_DIRTY_BIT_B"},
+	{AMDGPU_GFX_SQC_WRITE_DATA_BUFFER_CU0, "SQC_WRITE_DATA_BUFFER_CU0"},
+	{AMDGPU_GFX_SQC_WRITE_DATA_BUFFER_CU1, "SQC_WRITE_DATA_BUFFER_CU1"},
+	{AMDGPU_GFX_SQC_UTCL1_MISS_LFIFO_DATA_CACHE_A, "SQC_UTCL1_MISS_LFIFO_DATA_CACHE_A"},
+	{AMDGPU_GFX_SQC_UTCL1_MISS_LFIFO_DATA_CACHE_B, "SQC_UTCL1_MISS_LFIFO_DATA_CACHE_B"},
+	{AMDGPU_GFX_SQC_UTCL1_MISS_LFIFO_INST_CACHE, "SQC_UTCL1_MISS_LFIFO_INST_CACHE"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_sq_mem_list[] = {
+	{AMDGPU_GFX_SQ_SGPR_MEM0, "SQ_SGPR_MEM0"},
+	{AMDGPU_GFX_SQ_SGPR_MEM1, "SQ_SGPR_MEM1"},
+	{AMDGPU_GFX_SQ_SGPR_MEM2, "SQ_SGPR_MEM2"},
+	{AMDGPU_GFX_SQ_SGPR_MEM3, "SQ_SGPR_MEM3"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_ta_mem_list[] = {
+	{AMDGPU_GFX_TA_FS_AFIFO_RAM_LO, "TA_FS_AFIFO_RAM_LO"},
+	{AMDGPU_GFX_TA_FS_AFIFO_RAM_HI, "TA_FS_AFIFO_RAM_HI"},
+	{AMDGPU_GFX_TA_FS_CFIFO_RAM, "TA_FS_CFIFO_RAM"},
+	{AMDGPU_GFX_TA_FSX_LFIFO, "TA_FSX_LFIFO"},
+	{AMDGPU_GFX_TA_FS_DFIFO_RAM, "TA_FS_DFIFO_RAM"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_tcc_mem_list[] = {
+	{AMDGPU_GFX_TCC_MEM1, "TCC_MEM1"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_tca_mem_list[] = {
+	{AMDGPU_GFX_TCA_MEM1, "TCA_MEM1"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_tci_mem_list[] = {
+	{AMDGPU_GFX_TCIW_MEM, "TCIW_MEM"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_tcp_mem_list[] = {
+	{AMDGPU_GFX_TCP_LFIFO0, "TCP_LFIFO0"},
+	{AMDGPU_GFX_TCP_SET0BANK0_RAM, "TCP_SET0BANK0_RAM"},
+	{AMDGPU_GFX_TCP_SET0BANK1_RAM, "TCP_SET0BANK1_RAM"},
+	{AMDGPU_GFX_TCP_SET0BANK2_RAM, "TCP_SET0BANK2_RAM"},
+	{AMDGPU_GFX_TCP_SET0BANK3_RAM, "TCP_SET0BANK3_RAM"},
+	{AMDGPU_GFX_TCP_SET1BANK0_RAM, "TCP_SET1BANK0_RAM"},
+	{AMDGPU_GFX_TCP_SET1BANK1_RAM, "TCP_SET1BANK1_RAM"},
+	{AMDGPU_GFX_TCP_SET1BANK2_RAM, "TCP_SET1BANK2_RAM"},
+	{AMDGPU_GFX_TCP_SET1BANK3_RAM, "TCP_SET1BANK3_RAM"},
+	{AMDGPU_GFX_TCP_SET2BANK0_RAM, "TCP_SET2BANK0_RAM"},
+	{AMDGPU_GFX_TCP_SET2BANK1_RAM, "TCP_SET2BANK1_RAM"},
+	{AMDGPU_GFX_TCP_SET2BANK2_RAM, "TCP_SET2BANK2_RAM"},
+	{AMDGPU_GFX_TCP_SET2BANK3_RAM, "TCP_SET2BANK3_RAM"},
+	{AMDGPU_GFX_TCP_SET3BANK0_RAM, "TCP_SET3BANK0_RAM"},
+	{AMDGPU_GFX_TCP_SET3BANK1_RAM, "TCP_SET3BANK1_RAM"},
+	{AMDGPU_GFX_TCP_SET3BANK2_RAM, "TCP_SET3BANK2_RAM"},
+	{AMDGPU_GFX_TCP_SET3BANK3_RAM, "TCP_SET3BANK3_RAM"},
+	{AMDGPU_GFX_TCP_VM_FIFO, "TCP_VM_FIFO"},
+	{AMDGPU_GFX_TCP_DB_TAGRAM0, "TCP_DB_TAGRAM0"},
+	{AMDGPU_GFX_TCP_DB_TAGRAM1, "TCP_DB_TAGRAM1"},
+	{AMDGPU_GFX_TCP_DB_TAGRAM2, "TCP_DB_TAGRAM2"},
+	{AMDGPU_GFX_TCP_DB_TAGRAM3, "TCP_DB_TAGRAM3"},
+	{AMDGPU_GFX_TCP_UTCL1_LFIFO_PROBE0, "TCP_UTCL1_LFIFO_PROBE0"},
+	{AMDGPU_GFX_TCP_UTCL1_LFIFO_PROBE1, "TCP_UTCL1_LFIFO_PROBE1"},
+	{AMDGPU_GFX_TCP_CMD_FIFO, "TCP_CMD_FIFO"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_td_mem_list[] = {
+	{AMDGPU_GFX_TD_UTD_CS_FIFO_MEM, "TD_UTD_CS_FIFO_MEM"},
+	{AMDGPU_GFX_TD_UTD_SS_FIFO_LO_MEM, "TD_UTD_SS_FIFO_LO_MEM"},
+	{AMDGPU_GFX_TD_UTD_SS_FIFO_HI_MEM, "TD_UTD_SS_FIFO_HI_MEM"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_tcx_mem_list[] = {
+	{AMDGPU_GFX_TCX_FIFOD0, "TCX_FIFOD0"},
+	{AMDGPU_GFX_TCX_FIFOD1, "TCX_FIFOD1"},
+	{AMDGPU_GFX_TCX_FIFOD2, "TCX_FIFOD2"},
+	{AMDGPU_GFX_TCX_FIFOD3, "TCX_FIFOD3"},
+	{AMDGPU_GFX_TCX_FIFOD4, "TCX_FIFOD4"},
+	{AMDGPU_GFX_TCX_FIFOD5, "TCX_FIFOD5"},
+	{AMDGPU_GFX_TCX_FIFOD6, "TCX_FIFOD6"},
+	{AMDGPU_GFX_TCX_FIFOD7, "TCX_FIFOD7"},
+	{AMDGPU_GFX_TCX_FIFOB0, "TCX_FIFOB0"},
+	{AMDGPU_GFX_TCX_FIFOB1, "TCX_FIFOB1"},
+	{AMDGPU_GFX_TCX_FIFOB2, "TCX_FIFOB2"},
+	{AMDGPU_GFX_TCX_FIFOB3, "TCX_FIFOB3"},
+	{AMDGPU_GFX_TCX_FIFOB4, "TCX_FIFOB4"},
+	{AMDGPU_GFX_TCX_FIFOB5, "TCX_FIFOB5"},
+	{AMDGPU_GFX_TCX_FIFOB6, "TCX_FIFOB6"},
+	{AMDGPU_GFX_TCX_FIFOB7, "TCX_FIFOB7"},
+	{AMDGPU_GFX_TCX_FIFOA0, "TCX_FIFOA0"},
+	{AMDGPU_GFX_TCX_FIFOA1, "TCX_FIFOA1"},
+	{AMDGPU_GFX_TCX_FIFOA2, "TCX_FIFOA2"},
+	{AMDGPU_GFX_TCX_FIFOA3, "TCX_FIFOA3"},
+	{AMDGPU_GFX_TCX_FIFOA4, "TCX_FIFOA4"},
+	{AMDGPU_GFX_TCX_FIFOA5, "TCX_FIFOA5"},
+	{AMDGPU_GFX_TCX_FIFOA6, "TCX_FIFOA6"},
+	{AMDGPU_GFX_TCX_FIFOA7, "TCX_FIFOA7"},
+	{AMDGPU_GFX_TCX_CFIFO0, "TCX_CFIFO0"},
+	{AMDGPU_GFX_TCX_CFIFO1, "TCX_CFIFO1"},
+	{AMDGPU_GFX_TCX_CFIFO2, "TCX_CFIFO2"},
+	{AMDGPU_GFX_TCX_CFIFO3, "TCX_CFIFO3"},
+	{AMDGPU_GFX_TCX_CFIFO4, "TCX_CFIFO4"},
+	{AMDGPU_GFX_TCX_CFIFO5, "TCX_CFIFO5"},
+	{AMDGPU_GFX_TCX_CFIFO6, "TCX_CFIFO6"},
+	{AMDGPU_GFX_TCX_CFIFO7, "TCX_CFIFO7"},
+	{AMDGPU_GFX_TCX_FIFO_ACKB0, "TCX_FIFO_ACKB0"},
+	{AMDGPU_GFX_TCX_FIFO_ACKB1, "TCX_FIFO_ACKB1"},
+	{AMDGPU_GFX_TCX_FIFO_ACKB2, "TCX_FIFO_ACKB2"},
+	{AMDGPU_GFX_TCX_FIFO_ACKB3, "TCX_FIFO_ACKB3"},
+	{AMDGPU_GFX_TCX_FIFO_ACKB4, "TCX_FIFO_ACKB4"},
+	{AMDGPU_GFX_TCX_FIFO_ACKB5, "TCX_FIFO_ACKB5"},
+	{AMDGPU_GFX_TCX_FIFO_ACKB6, "TCX_FIFO_ACKB6"},
+	{AMDGPU_GFX_TCX_FIFO_ACKB7, "TCX_FIFO_ACKB7"},
+	{AMDGPU_GFX_TCX_FIFO_ACKD0, "TCX_FIFO_ACKD0"},
+	{AMDGPU_GFX_TCX_FIFO_ACKD1, "TCX_FIFO_ACKD1"},
+	{AMDGPU_GFX_TCX_FIFO_ACKD2, "TCX_FIFO_ACKD2"},
+	{AMDGPU_GFX_TCX_FIFO_ACKD3, "TCX_FIFO_ACKD3"},
+	{AMDGPU_GFX_TCX_FIFO_ACKD4, "TCX_FIFO_ACKD4"},
+	{AMDGPU_GFX_TCX_FIFO_ACKD5, "TCX_FIFO_ACKD5"},
+	{AMDGPU_GFX_TCX_FIFO_ACKD6, "TCX_FIFO_ACKD6"},
+	{AMDGPU_GFX_TCX_FIFO_ACKD7, "TCX_FIFO_ACKD7"},
+	{AMDGPU_GFX_TCX_DST_FIFOA0, "TCX_DST_FIFOA0"},
+	{AMDGPU_GFX_TCX_DST_FIFOA1, "TCX_DST_FIFOA1"},
+	{AMDGPU_GFX_TCX_DST_FIFOA2, "TCX_DST_FIFOA2"},
+	{AMDGPU_GFX_TCX_DST_FIFOA3, "TCX_DST_FIFOA3"},
+	{AMDGPU_GFX_TCX_DST_FIFOA4, "TCX_DST_FIFOA4"},
+	{AMDGPU_GFX_TCX_DST_FIFOA5, "TCX_DST_FIFOA5"},
+	{AMDGPU_GFX_TCX_DST_FIFOA6, "TCX_DST_FIFOA6"},
+	{AMDGPU_GFX_TCX_DST_FIFOA7, "TCX_DST_FIFOA7"},
+	{AMDGPU_GFX_TCX_DST_FIFOB0, "TCX_DST_FIFOB0"},
+	{AMDGPU_GFX_TCX_DST_FIFOB1, "TCX_DST_FIFOB1"},
+	{AMDGPU_GFX_TCX_DST_FIFOB2, "TCX_DST_FIFOB2"},
+	{AMDGPU_GFX_TCX_DST_FIFOB3, "TCX_DST_FIFOB3"},
+	{AMDGPU_GFX_TCX_DST_FIFOB4, "TCX_DST_FIFOB4"},
+	{AMDGPU_GFX_TCX_DST_FIFOB5, "TCX_DST_FIFOB5"},
+	{AMDGPU_GFX_TCX_DST_FIFOB6, "TCX_DST_FIFOB6"},
+	{AMDGPU_GFX_TCX_DST_FIFOB7, "TCX_DST_FIFOB7"},
+	{AMDGPU_GFX_TCX_DST_FIFOD0, "TCX_DST_FIFOD0"},
+	{AMDGPU_GFX_TCX_DST_FIFOD1, "TCX_DST_FIFOD1"},
+	{AMDGPU_GFX_TCX_DST_FIFOD2, "TCX_DST_FIFOD2"},
+	{AMDGPU_GFX_TCX_DST_FIFOD3, "TCX_DST_FIFOD3"},
+	{AMDGPU_GFX_TCX_DST_FIFOD4, "TCX_DST_FIFOD4"},
+	{AMDGPU_GFX_TCX_DST_FIFOD5, "TCX_DST_FIFOD5"},
+	{AMDGPU_GFX_TCX_DST_FIFOD6, "TCX_DST_FIFOD6"},
+	{AMDGPU_GFX_TCX_DST_FIFOD7, "TCX_DST_FIFOD7"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKB0, "TCX_DST_FIFO_ACKB0"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKB1, "TCX_DST_FIFO_ACKB1"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKB2, "TCX_DST_FIFO_ACKB2"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKB3, "TCX_DST_FIFO_ACKB3"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKB4, "TCX_DST_FIFO_ACKB4"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKB5, "TCX_DST_FIFO_ACKB5"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKB6, "TCX_DST_FIFO_ACKB6"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKB7, "TCX_DST_FIFO_ACKB7"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKD0, "TCX_DST_FIFO_ACKD0"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKD1, "TCX_DST_FIFO_ACKD1"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKD2, "TCX_DST_FIFO_ACKD2"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKD3, "TCX_DST_FIFO_ACKD3"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKD4, "TCX_DST_FIFO_ACKD4"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKD5, "TCX_DST_FIFO_ACKD5"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKD6, "TCX_DST_FIFO_ACKD6"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKD7, "TCX_DST_FIFO_ACKD7"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_atc_l2_mem_list[] = {
+	{AMDGPU_GFX_ATC_L2_MEM, "ATC_L2_MEM"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_utcl2_mem_list[] = {
+	{AMDGPU_GFX_UTCL2_MEM, "UTCL2_MEM"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_vml2_mem_list[] = {
+	{AMDGPU_GFX_VML2_MEM, "VML2_MEM"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_vml2_walker_mem_list[] = {
+	{AMDGPU_GFX_VML2_WALKER_MEM, "VML2_WALKER_MEM"},
+};
+
+static const struct amdgpu_gfx_ras_mem_id_entry gfx_v9_4_3_ras_mem_list_array[AMDGPU_GFX_MEM_TYPE_NUM] = {
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_cp_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_gcea_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_gc_cane_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_gcutcl2_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_gds_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_lds_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_rlc_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_sp_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_spi_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_sqc_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_sq_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_ta_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_tcc_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_tca_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_tci_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_tcp_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_td_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_tcx_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_atc_l2_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_utcl2_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_vml2_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_vml2_walker_mem_list)
+};
+
+static const struct amdgpu_gfx_ras_reg_entry gfx_v9_4_3_ce_reg_list[] = {
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regRLC_CE_ERR_STATUS_LOW, regRLC_CE_ERR_STATUS_HIGH),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "RLC"},
+	    AMDGPU_GFX_RLC_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regCPC_CE_ERR_STATUS_LO, regCPC_CE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "CPC"},
+	    AMDGPU_GFX_CP_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regCPF_CE_ERR_STATUS_LO, regCPF_CE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "CPF"},
+	    AMDGPU_GFX_CP_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regCPG_CE_ERR_STATUS_LO, regCPG_CE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "CPG"},
+	    AMDGPU_GFX_CP_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regGDS_CE_ERR_STATUS_LO, regGDS_CE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "GDS"},
+	    AMDGPU_GFX_GDS_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regGC_CANE_CE_ERR_STATUS_LO, regGC_CANE_CE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "CANE"},
+	    AMDGPU_GFX_GC_CANE_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSPI_CE_ERR_STATUS_LO, regSPI_CE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SPI"},
+	    AMDGPU_GFX_SPI_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSP0_CE_ERR_STATUS_LO, regSP0_CE_ERR_STATUS_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SP0"},
+	    AMDGPU_GFX_SP_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSP1_CE_ERR_STATUS_LO, regSP1_CE_ERR_STATUS_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SP1"},
+	    AMDGPU_GFX_SP_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSQ_CE_ERR_STATUS_LO, regSQ_CE_ERR_STATUS_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SQ"},
+	    AMDGPU_GFX_SQ_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSQC_CE_EDC_LO, regSQC_CE_EDC_HI),
+	    5, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SQC"},
+	    AMDGPU_GFX_SQC_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTCX_CE_ERR_STATUS_LO, regTCX_CE_ERR_STATUS_HI),
+	    2, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TCX"},
+	    AMDGPU_GFX_TCX_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTCC_CE_ERR_STATUS_LO, regTCC_CE_ERR_STATUS_HI),
+	    16, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TCC"},
+	    AMDGPU_GFX_TCC_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTA_CE_EDC_LO, regTA_CE_EDC_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TA"},
+	    AMDGPU_GFX_TA_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTCI_CE_EDC_LO_REG, regTCI_CE_EDC_HI_REG),
+	    27, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TCI"},
+	    AMDGPU_GFX_TCI_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTCP_CE_EDC_LO_REG, regTCP_CE_EDC_HI_REG),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TCP"},
+	    AMDGPU_GFX_TCP_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTD_CE_EDC_LO, regTD_CE_EDC_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TD"},
+	    AMDGPU_GFX_TD_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regGCEA_CE_ERR_STATUS_LO, regGCEA_CE_ERR_STATUS_HI),
+	    16, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "GCEA"},
+	    AMDGPU_GFX_GCEA_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regLDS_CE_ERR_STATUS_LO, regLDS_CE_ERR_STATUS_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "LDS"},
+	    AMDGPU_GFX_LDS_MEM, 4},
+};
+
+static const struct amdgpu_gfx_ras_reg_entry gfx_v9_4_3_ue_reg_list[] = {
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regRLC_UE_ERR_STATUS_LOW, regRLC_UE_ERR_STATUS_HIGH),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "RLC"},
+	    AMDGPU_GFX_RLC_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regCPC_UE_ERR_STATUS_LO, regCPC_UE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "CPC"},
+	    AMDGPU_GFX_CP_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regCPF_UE_ERR_STATUS_LO, regCPF_UE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "CPF"},
+	    AMDGPU_GFX_CP_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regCPG_UE_ERR_STATUS_LO, regCPG_UE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "CPG"},
+	    AMDGPU_GFX_CP_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regGDS_UE_ERR_STATUS_LO, regGDS_UE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "GDS"},
+	    AMDGPU_GFX_GDS_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regGC_CANE_UE_ERR_STATUS_LO, regGC_CANE_UE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "CANE"},
+	    AMDGPU_GFX_GC_CANE_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSPI_UE_ERR_STATUS_LO, regSPI_UE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SPI"},
+	    AMDGPU_GFX_SPI_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSP0_UE_ERR_STATUS_LO, regSP0_UE_ERR_STATUS_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SP0"},
+	    AMDGPU_GFX_SP_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSP1_UE_ERR_STATUS_LO, regSP1_UE_ERR_STATUS_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SP1"},
+	    AMDGPU_GFX_SP_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSQ_UE_ERR_STATUS_LO, regSQ_UE_ERR_STATUS_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SQ"},
+	    AMDGPU_GFX_SQ_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSQC_UE_EDC_LO, regSQC_UE_EDC_HI),
+	    5, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SQC"},
+	    AMDGPU_GFX_SQC_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTCX_UE_ERR_STATUS_LO, regTCX_UE_ERR_STATUS_HI),
+	    2, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TCX"},
+	    AMDGPU_GFX_TCX_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTCC_UE_ERR_STATUS_LO, regTCC_UE_ERR_STATUS_HI),
+	    16, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TCC"},
+	    AMDGPU_GFX_TCC_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTA_UE_EDC_LO, regTA_UE_EDC_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TA"},
+	    AMDGPU_GFX_TA_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTCI_UE_EDC_LO_REG, regTCI_UE_EDC_HI_REG),
+	    27, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TCI"},
+	    AMDGPU_GFX_TCI_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTCP_UE_EDC_LO_REG, regTCP_UE_EDC_HI_REG),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TCP"},
+	    AMDGPU_GFX_TCP_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTD_UE_EDC_LO, regTD_UE_EDC_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TD"},
+	    AMDGPU_GFX_TD_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTCA_UE_ERR_STATUS_LO, regTCA_UE_ERR_STATUS_HI),
+	    2, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TCA"},
+	    AMDGPU_GFX_TCA_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regGCEA_UE_ERR_STATUS_LO, regGCEA_UE_ERR_STATUS_HI),
+	    16, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "GCEA"},
+	    AMDGPU_GFX_GCEA_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regLDS_UE_ERR_STATUS_LO, regLDS_UE_ERR_STATUS_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "LDS"},
+	    AMDGPU_GFX_LDS_MEM, 4},
+};
+
+static void gfx_v9_4_3_inst_query_ras_err_count(struct amdgpu_device *adev,
+					void *ras_error_status, int xcc_id)
+{
+	struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
+	unsigned long ce_count = 0, ue_count = 0;
+	uint32_t i, j, k;
+
+	/* NOTE: convert xcc_id to physical XCD ID (XCD0 or XCD1) */
+	struct amdgpu_smuio_mcm_config_info mcm_info = {
+		.socket_id = adev->smuio.funcs->get_socket_id(adev),
+		.die_id = xcc_id & 0x01 ? 1 : 0,
+	};
+
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	for (i = 0; i < ARRAY_SIZE(gfx_v9_4_3_ce_reg_list); i++) {
+		for (j = 0; j < gfx_v9_4_3_ce_reg_list[i].se_num; j++) {
+			for (k = 0; k < gfx_v9_4_3_ce_reg_list[i].reg_entry.reg_inst; k++) {
+				/* no need to select if instance number is 1 */
+				if (gfx_v9_4_3_ce_reg_list[i].se_num > 1 ||
+				    gfx_v9_4_3_ce_reg_list[i].reg_entry.reg_inst > 1)
+					gfx_v9_4_3_xcc_select_se_sh(adev, j, 0, k, xcc_id);
+
+				amdgpu_ras_inst_query_ras_error_count(adev,
+					&(gfx_v9_4_3_ce_reg_list[i].reg_entry),
+					1,
+					gfx_v9_4_3_ras_mem_list_array[gfx_v9_4_3_ce_reg_list[i].mem_id_type].mem_id_ent,
+					gfx_v9_4_3_ras_mem_list_array[gfx_v9_4_3_ce_reg_list[i].mem_id_type].size,
+					GET_INST(GC, xcc_id),
+					AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE,
+					&ce_count);
+
+				amdgpu_ras_inst_query_ras_error_count(adev,
+					&(gfx_v9_4_3_ue_reg_list[i].reg_entry),
+					1,
+					gfx_v9_4_3_ras_mem_list_array[gfx_v9_4_3_ue_reg_list[i].mem_id_type].mem_id_ent,
+					gfx_v9_4_3_ras_mem_list_array[gfx_v9_4_3_ue_reg_list[i].mem_id_type].size,
+					GET_INST(GC, xcc_id),
+					AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE,
+					&ue_count);
+			}
+		}
+	}
+
+	/* handle extra register entries of UE */
+	for (; i < ARRAY_SIZE(gfx_v9_4_3_ue_reg_list); i++) {
+		for (j = 0; j < gfx_v9_4_3_ue_reg_list[i].se_num; j++) {
+			for (k = 0; k < gfx_v9_4_3_ue_reg_list[i].reg_entry.reg_inst; k++) {
+				/* no need to select if instance number is 1 */
+				if (gfx_v9_4_3_ue_reg_list[i].se_num > 1 ||
+					gfx_v9_4_3_ue_reg_list[i].reg_entry.reg_inst > 1)
+					gfx_v9_4_3_xcc_select_se_sh(adev, j, 0, k, xcc_id);
+
+				amdgpu_ras_inst_query_ras_error_count(adev,
+					&(gfx_v9_4_3_ue_reg_list[i].reg_entry),
+					1,
+					gfx_v9_4_3_ras_mem_list_array[gfx_v9_4_3_ue_reg_list[i].mem_id_type].mem_id_ent,
+					gfx_v9_4_3_ras_mem_list_array[gfx_v9_4_3_ue_reg_list[i].mem_id_type].size,
+					GET_INST(GC, xcc_id),
+					AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE,
+					&ue_count);
+			}
+		}
+	}
+
+	gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff,
+			xcc_id);
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	/* the caller should make sure initialize value of
+	 * err_data->ue_count and err_data->ce_count
+	 */
+	amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, ue_count);
+	amdgpu_ras_error_statistic_ce_count(err_data, &mcm_info, ce_count);
+}
+
+static void gfx_v9_4_3_inst_reset_ras_err_count(struct amdgpu_device *adev,
+					void *ras_error_status, int xcc_id)
+{
+	uint32_t i, j, k;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	for (i = 0; i < ARRAY_SIZE(gfx_v9_4_3_ce_reg_list); i++) {
+		for (j = 0; j < gfx_v9_4_3_ce_reg_list[i].se_num; j++) {
+			for (k = 0; k < gfx_v9_4_3_ce_reg_list[i].reg_entry.reg_inst; k++) {
+				/* no need to select if instance number is 1 */
+				if (gfx_v9_4_3_ce_reg_list[i].se_num > 1 ||
+				    gfx_v9_4_3_ce_reg_list[i].reg_entry.reg_inst > 1)
+					gfx_v9_4_3_xcc_select_se_sh(adev, j, 0, k, xcc_id);
+
+				amdgpu_ras_inst_reset_ras_error_count(adev,
+					&(gfx_v9_4_3_ce_reg_list[i].reg_entry),
+					1,
+					GET_INST(GC, xcc_id));
+
+				amdgpu_ras_inst_reset_ras_error_count(adev,
+					&(gfx_v9_4_3_ue_reg_list[i].reg_entry),
+					1,
+					GET_INST(GC, xcc_id));
+			}
+		}
+	}
+
+	/* handle extra register entries of UE */
+	for (; i < ARRAY_SIZE(gfx_v9_4_3_ue_reg_list); i++) {
+		for (j = 0; j < gfx_v9_4_3_ue_reg_list[i].se_num; j++) {
+			for (k = 0; k < gfx_v9_4_3_ue_reg_list[i].reg_entry.reg_inst; k++) {
+				/* no need to select if instance number is 1 */
+				if (gfx_v9_4_3_ue_reg_list[i].se_num > 1 ||
+					gfx_v9_4_3_ue_reg_list[i].reg_entry.reg_inst > 1)
+					gfx_v9_4_3_xcc_select_se_sh(adev, j, 0, k, xcc_id);
+
+				amdgpu_ras_inst_reset_ras_error_count(adev,
+					&(gfx_v9_4_3_ue_reg_list[i].reg_entry),
+					1,
+					GET_INST(GC, xcc_id));
+			}
+		}
+	}
+
+	gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff,
+			xcc_id);
+	mutex_unlock(&adev->grbm_idx_mutex);
+}
+
+static void gfx_v9_4_3_inst_enable_watchdog_timer(struct amdgpu_device *adev,
+					void *ras_error_status, int xcc_id)
+{
+	uint32_t i;
+	uint32_t data;
+
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	data = RREG32_SOC15(GC, GET_INST(GC, 0), regSQ_TIMEOUT_CONFIG);
+	data = REG_SET_FIELD(data, SQ_TIMEOUT_CONFIG, TIMEOUT_FATAL_DISABLE,
+			     amdgpu_watchdog_timer.timeout_fatal_disable ? 1 : 0);
+
+	if (amdgpu_watchdog_timer.timeout_fatal_disable &&
+	    (amdgpu_watchdog_timer.period < 1 ||
+	     amdgpu_watchdog_timer.period > 0x23)) {
+		dev_warn(adev->dev, "Watchdog period range is 1 to 0x23\n");
+		amdgpu_watchdog_timer.period = 0x23;
+	}
+	data = REG_SET_FIELD(data, SQ_TIMEOUT_CONFIG, PERIOD_SEL,
+			     amdgpu_watchdog_timer.period);
+
+	mutex_lock(&adev->grbm_idx_mutex);
+	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
+		gfx_v9_4_3_xcc_select_se_sh(adev, i, 0xffffffff, 0xffffffff, xcc_id);
+		WREG32_SOC15(GC, GET_INST(GC, xcc_id), regSQ_TIMEOUT_CONFIG, data);
+	}
+	gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff,
+			xcc_id);
+	mutex_unlock(&adev->grbm_idx_mutex);
+}
+
+static void gfx_v9_4_3_query_ras_error_count(struct amdgpu_device *adev,
+					void *ras_error_status)
+{
+	amdgpu_gfx_ras_error_func(adev, ras_error_status,
+			gfx_v9_4_3_inst_query_ras_err_count);
+}
+
+static void gfx_v9_4_3_reset_ras_error_count(struct amdgpu_device *adev)
+{
+	amdgpu_gfx_ras_error_func(adev, NULL, gfx_v9_4_3_inst_reset_ras_err_count);
+}
+
+static void gfx_v9_4_3_enable_watchdog_timer(struct amdgpu_device *adev)
+{
+	amdgpu_gfx_ras_error_func(adev, NULL, gfx_v9_4_3_inst_enable_watchdog_timer);
+}
+
+static void gfx_v9_4_3_ring_insert_nop(struct amdgpu_ring *ring, uint32_t num_nop)
+{
+	/* Header itself is a NOP packet */
+	if (num_nop == 1) {
+		amdgpu_ring_write(ring, ring->funcs->nop);
+		return;
+	}
+
+	/* Max HW optimization till 0x3ffe, followed by remaining one NOP at a time*/
+	amdgpu_ring_write(ring, PACKET3(PACKET3_NOP, min(num_nop - 2, 0x3ffe)));
+
+	/* Header is at index 0, followed by num_nops - 1 NOP packet's */
+	amdgpu_ring_insert_nop(ring, num_nop - 1);
+}
+
+static void gfx_v9_4_3_ip_print(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k;
+	uint32_t xcc_id, xcc_offset, inst_offset;
+	uint32_t num_xcc, reg, num_inst;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_9_4_3);
+
+	if (!adev->gfx.ip_dump_core)
+		return;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	drm_printf(p, "Number of Instances:%d\n", num_xcc);
+	for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) {
+		xcc_offset = xcc_id * reg_count;
+		drm_printf(p, "\nInstance id:%d\n", xcc_id);
+		for (i = 0; i < reg_count; i++)
+			drm_printf(p, "%-50s \t 0x%08x\n",
+				   gc_reg_list_9_4_3[i].reg_name,
+				   adev->gfx.ip_dump_core[xcc_offset + i]);
+	}
+
+	/* print compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	num_inst = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe_per_mec *
+		adev->gfx.mec.num_queue_per_pipe;
+
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_9_4_3);
+	drm_printf(p, "\nnum_xcc: %d num_mec: %d num_pipe: %d num_queue: %d\n",
+		   num_xcc,
+		   adev->gfx.mec.num_mec,
+		   adev->gfx.mec.num_pipe_per_mec,
+		   adev->gfx.mec.num_queue_per_pipe);
+
+	for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) {
+		xcc_offset = xcc_id * reg_count * num_inst;
+		inst_offset = 0;
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+					drm_printf(p,
+						   "\nxcc:%d mec:%d, pipe:%d, queue:%d\n",
+						    xcc_id, i, j, k);
+					for (reg = 0; reg < reg_count; reg++) {
+						if (i && gc_cp_reg_list_9_4_3[reg].reg_offset ==
+						    regCP_MEC_ME1_HEADER_DUMP)
+							drm_printf(p,
+								   "%-50s \t 0x%08x\n",
+								   "regCP_MEC_ME2_HEADER_DUMP",
+								   adev->gfx.ip_dump_compute_queues
+								   [xcc_offset + inst_offset +
+								    reg]);
+						else
+							drm_printf(p,
+								   "%-50s \t 0x%08x\n",
+								   gc_cp_reg_list_9_4_3[reg].reg_name,
+								   adev->gfx.ip_dump_compute_queues
+								   [xcc_offset + inst_offset +
+								    reg]);
+					}
+					inst_offset += reg_count;
+				}
+			}
+		}
+	}
+}
+
+static void gfx_v9_4_3_ip_dump(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k;
+	uint32_t num_xcc, reg, num_inst;
+	uint32_t xcc_id, xcc_offset, inst_offset;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_9_4_3);
+
+	if (!adev->gfx.ip_dump_core)
+		return;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+
+	for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) {
+		xcc_offset = xcc_id * reg_count;
+		for (i = 0; i < reg_count; i++)
+			adev->gfx.ip_dump_core[xcc_offset + i] =
+				RREG32(SOC15_REG_ENTRY_OFFSET_INST(gc_reg_list_9_4_3[i],
+								   GET_INST(GC, xcc_id)));
+	}
+
+	/* dump compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	num_inst = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe_per_mec *
+		adev->gfx.mec.num_queue_per_pipe;
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_9_4_3);
+	mutex_lock(&adev->srbm_mutex);
+	for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) {
+		xcc_offset = xcc_id * reg_count * num_inst;
+		inst_offset = 0;
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+					/* ME0 is for GFX so start from 1 for CP */
+					soc15_grbm_select(adev, 1 + i, j, k, 0,
+							  GET_INST(GC, xcc_id));
+
+					for (reg = 0; reg < reg_count; reg++) {
+						if (i && gc_cp_reg_list_9_4_3[reg].reg_offset ==
+						    regCP_MEC_ME1_HEADER_DUMP)
+							adev->gfx.ip_dump_compute_queues
+								[xcc_offset +
+								 inst_offset + reg] =
+								RREG32(SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id),
+											regCP_MEC_ME2_HEADER_DUMP));
+						else
+							adev->gfx.ip_dump_compute_queues
+								[xcc_offset +
+								 inst_offset + reg] =
+								RREG32(SOC15_REG_ENTRY_OFFSET_INST(
+									       gc_cp_reg_list_9_4_3[reg],
+									       GET_INST(GC, xcc_id)));
+					}
+					inst_offset += reg_count;
+				}
+			}
+		}
+	}
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+}
+
+static void gfx_v9_4_3_ring_emit_cleaner_shader(struct amdgpu_ring *ring)
+{
+	/* Emit the cleaner shader */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_RUN_CLEANER_SHADER, 0));
+	amdgpu_ring_write(ring, 0);  /* RESERVED field, programmed to zero */
+}
+
+static const struct amd_ip_funcs gfx_v9_4_3_ip_funcs = {
+	.name = "gfx_v9_4_3",
+	.early_init = gfx_v9_4_3_early_init,
+	.late_init = gfx_v9_4_3_late_init,
+	.sw_init = gfx_v9_4_3_sw_init,
+	.sw_fini = gfx_v9_4_3_sw_fini,
+	.hw_init = gfx_v9_4_3_hw_init,
+	.hw_fini = gfx_v9_4_3_hw_fini,
+	.suspend = gfx_v9_4_3_suspend,
+	.resume = gfx_v9_4_3_resume,
+	.is_idle = gfx_v9_4_3_is_idle,
+	.wait_for_idle = gfx_v9_4_3_wait_for_idle,
+	.soft_reset = gfx_v9_4_3_soft_reset,
+	.set_clockgating_state = gfx_v9_4_3_set_clockgating_state,
+	.set_powergating_state = gfx_v9_4_3_set_powergating_state,
+	.get_clockgating_state = gfx_v9_4_3_get_clockgating_state,
+	.dump_ip_state = gfx_v9_4_3_ip_dump,
+	.print_ip_state = gfx_v9_4_3_ip_print,
+};
+
+static const struct amdgpu_ring_funcs gfx_v9_4_3_ring_funcs_compute = {
+	.type = AMDGPU_RING_TYPE_COMPUTE,
+	.align_mask = 0xff,
+	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
+	.support_64bit_ptrs = true,
+	.get_rptr = gfx_v9_4_3_ring_get_rptr_compute,
+	.get_wptr = gfx_v9_4_3_ring_get_wptr_compute,
+	.set_wptr = gfx_v9_4_3_ring_set_wptr_compute,
+	.emit_frame_size =
+		20 + /* gfx_v9_4_3_ring_emit_gds_switch */
+		7 + /* gfx_v9_4_3_ring_emit_hdp_flush */
+		5 + /* hdp invalidate */
+		7 + /* gfx_v9_4_3_ring_emit_pipeline_sync */
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
+		2 + /* gfx_v9_4_3_ring_emit_vm_flush */
+		8 + 8 + 8 + /* gfx_v9_4_3_ring_emit_fence x3 for user fence, vm fence */
+		7 + /* gfx_v9_4_3_emit_mem_sync */
+		5 + /* gfx_v9_4_3_emit_wave_limit for updating regSPI_WCL_PIPE_PERCENT_GFX register */
+		15 + /* for updating 3 regSPI_WCL_PIPE_PERCENT_CS registers */
+		2, /* gfx_v9_4_3_ring_emit_cleaner_shader */
+	.emit_ib_size =	7, /* gfx_v9_4_3_ring_emit_ib_compute */
+	.emit_ib = gfx_v9_4_3_ring_emit_ib_compute,
+	.emit_fence = gfx_v9_4_3_ring_emit_fence,
+	.emit_pipeline_sync = gfx_v9_4_3_ring_emit_pipeline_sync,
+	.emit_vm_flush = gfx_v9_4_3_ring_emit_vm_flush,
+	.emit_gds_switch = gfx_v9_4_3_ring_emit_gds_switch,
+	.emit_hdp_flush = gfx_v9_4_3_ring_emit_hdp_flush,
+	.test_ring = gfx_v9_4_3_ring_test_ring,
+	.test_ib = gfx_v9_4_3_ring_test_ib,
+	.insert_nop = gfx_v9_4_3_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.emit_wreg = gfx_v9_4_3_ring_emit_wreg,
+	.emit_reg_wait = gfx_v9_4_3_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = gfx_v9_4_3_ring_emit_reg_write_reg_wait,
+	.soft_recovery = gfx_v9_4_3_ring_soft_recovery,
+	.emit_mem_sync = gfx_v9_4_3_emit_mem_sync,
+	.emit_wave_limit = gfx_v9_4_3_emit_wave_limit,
+	.reset = gfx_v9_4_3_reset_kcq,
+	.emit_cleaner_shader = gfx_v9_4_3_ring_emit_cleaner_shader,
+	.begin_use = amdgpu_gfx_enforce_isolation_ring_begin_use,
+	.end_use = amdgpu_gfx_enforce_isolation_ring_end_use,
+};
+
+static const struct amdgpu_ring_funcs gfx_v9_4_3_ring_funcs_kiq = {
+	.type = AMDGPU_RING_TYPE_KIQ,
+	.align_mask = 0xff,
+	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
+	.support_64bit_ptrs = true,
+	.get_rptr = gfx_v9_4_3_ring_get_rptr_compute,
+	.get_wptr = gfx_v9_4_3_ring_get_wptr_compute,
+	.set_wptr = gfx_v9_4_3_ring_set_wptr_compute,
+	.emit_frame_size =
+		20 + /* gfx_v9_4_3_ring_emit_gds_switch */
+		7 + /* gfx_v9_4_3_ring_emit_hdp_flush */
+		5 + /* hdp invalidate */
+		7 + /* gfx_v9_4_3_ring_emit_pipeline_sync */
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
+		2 + /* gfx_v9_4_3_ring_emit_vm_flush */
+		8 + 8 + 8, /* gfx_v9_4_3_ring_emit_fence_kiq x3 for user fence, vm fence */
+	.emit_ib_size =	7, /* gfx_v9_4_3_ring_emit_ib_compute */
+	.emit_fence = gfx_v9_4_3_ring_emit_fence_kiq,
+	.test_ring = gfx_v9_4_3_ring_test_ring,
+	.insert_nop = amdgpu_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.emit_rreg = gfx_v9_4_3_ring_emit_rreg,
+	.emit_wreg = gfx_v9_4_3_ring_emit_wreg,
+	.emit_reg_wait = gfx_v9_4_3_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = gfx_v9_4_3_ring_emit_reg_write_reg_wait,
+	.emit_hdp_flush = gfx_v9_4_3_ring_emit_hdp_flush,
+};
+
+static void gfx_v9_4_3_set_ring_funcs(struct amdgpu_device *adev)
+{
+	int i, j, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++) {
+		adev->gfx.kiq[i].ring.funcs = &gfx_v9_4_3_ring_funcs_kiq;
+
+		for (j = 0; j < adev->gfx.num_compute_rings; j++)
+			adev->gfx.compute_ring[j + i * adev->gfx.num_compute_rings].funcs
+					= &gfx_v9_4_3_ring_funcs_compute;
+	}
+}
+
+static const struct amdgpu_irq_src_funcs gfx_v9_4_3_eop_irq_funcs = {
+	.set = gfx_v9_4_3_set_eop_interrupt_state,
+	.process = gfx_v9_4_3_eop_irq,
+};
+
+static const struct amdgpu_irq_src_funcs gfx_v9_4_3_priv_reg_irq_funcs = {
+	.set = gfx_v9_4_3_set_priv_reg_fault_state,
+	.process = gfx_v9_4_3_priv_reg_irq,
+};
+
+static const struct amdgpu_irq_src_funcs gfx_v9_4_3_bad_op_irq_funcs = {
+	.set = gfx_v9_4_3_set_bad_op_fault_state,
+	.process = gfx_v9_4_3_bad_op_irq,
+};
+
+static const struct amdgpu_irq_src_funcs gfx_v9_4_3_priv_inst_irq_funcs = {
+	.set = gfx_v9_4_3_set_priv_inst_fault_state,
+	.process = gfx_v9_4_3_priv_inst_irq,
+};
+
+static void gfx_v9_4_3_set_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->gfx.eop_irq.num_types = AMDGPU_CP_IRQ_LAST;
+	adev->gfx.eop_irq.funcs = &gfx_v9_4_3_eop_irq_funcs;
+
+	adev->gfx.priv_reg_irq.num_types = 1;
+	adev->gfx.priv_reg_irq.funcs = &gfx_v9_4_3_priv_reg_irq_funcs;
+
+	adev->gfx.bad_op_irq.num_types = 1;
+	adev->gfx.bad_op_irq.funcs = &gfx_v9_4_3_bad_op_irq_funcs;
+
+	adev->gfx.priv_inst_irq.num_types = 1;
+	adev->gfx.priv_inst_irq.funcs = &gfx_v9_4_3_priv_inst_irq_funcs;
+}
+
+static void gfx_v9_4_3_set_rlc_funcs(struct amdgpu_device *adev)
+{
+	adev->gfx.rlc.funcs = &gfx_v9_4_3_rlc_funcs;
+}
+
+
+static void gfx_v9_4_3_set_gds_init(struct amdgpu_device *adev)
+{
+	/* 9.4.3 variants removed all the GDS internal memory,
+	 * only support GWS opcode in kernel, like barrier
+	 * semaphore.etc */
+
+	/* init asic gds info */
+	adev->gds.gds_size = 0;
+	adev->gds.gds_compute_max_wave_id = 0;
+	adev->gds.gws_size = 64;
+	adev->gds.oa_size = 16;
+}
+
+static void gfx_v9_4_3_set_user_cu_inactive_bitmap(struct amdgpu_device *adev,
+						 u32 bitmap, int xcc_id)
+{
+	u32 data;
+
+	if (!bitmap)
+		return;
+
+	data = bitmap << GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_CUS__SHIFT;
+	data &= GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_CUS_MASK;
+
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regGC_USER_SHADER_ARRAY_CONFIG, data);
+}
+
+static u32 gfx_v9_4_3_get_cu_active_bitmap(struct amdgpu_device *adev, int xcc_id)
+{
+	u32 data, mask;
+
+	data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCC_GC_SHADER_ARRAY_CONFIG);
+	data |= RREG32_SOC15(GC, GET_INST(GC, xcc_id), regGC_USER_SHADER_ARRAY_CONFIG);
+
+	data &= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_CUS_MASK;
+	data >>= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_CUS__SHIFT;
+
+	mask = amdgpu_gfx_create_bitmask(adev->gfx.config.max_cu_per_sh);
+
+	return (~data) & mask;
+}
+
+static int gfx_v9_4_3_get_cu_info(struct amdgpu_device *adev,
+				 struct amdgpu_cu_info *cu_info)
+{
+	int i, j, k, prev_counter, counter, xcc_id, active_cu_number = 0;
+	u32 mask, bitmap, ao_bitmap, ao_cu_mask = 0, tmp;
+	unsigned disable_masks[4 * 4];
+	bool is_symmetric_cus;
+
+	if (!adev || !cu_info)
+		return -EINVAL;
+
+	/*
+	 * 16 comes from bitmap array size 4*4, and it can cover all gfx9 ASICs
+	 */
+	if (adev->gfx.config.max_shader_engines *
+		adev->gfx.config.max_sh_per_se > 16)
+		return -EINVAL;
+
+	amdgpu_gfx_parse_disable_cu(disable_masks,
+				    adev->gfx.config.max_shader_engines,
+				    adev->gfx.config.max_sh_per_se);
+
+	mutex_lock(&adev->grbm_idx_mutex);
+	for (xcc_id = 0; xcc_id < NUM_XCC(adev->gfx.xcc_mask); xcc_id++) {
+		is_symmetric_cus = true;
+		for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
+			for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
+				mask = 1;
+				ao_bitmap = 0;
+				counter = 0;
+				gfx_v9_4_3_xcc_select_se_sh(adev, i, j, 0xffffffff, xcc_id);
+				gfx_v9_4_3_set_user_cu_inactive_bitmap(
+					adev,
+					disable_masks[i * adev->gfx.config.max_sh_per_se + j],
+					xcc_id);
+				bitmap = gfx_v9_4_3_get_cu_active_bitmap(adev, xcc_id);
+
+				cu_info->bitmap[xcc_id][i][j] = bitmap;
+
+				for (k = 0; k < adev->gfx.config.max_cu_per_sh; k++) {
+					if (bitmap & mask) {
+						if (counter < adev->gfx.config.max_cu_per_sh)
+							ao_bitmap |= mask;
+						counter++;
+					}
+					mask <<= 1;
+				}
+				active_cu_number += counter;
+				if (i < 2 && j < 2)
+					ao_cu_mask |= (ao_bitmap << (i * 16 + j * 8));
+				cu_info->ao_cu_bitmap[i][j] = ao_bitmap;
+			}
+			if (i && is_symmetric_cus && prev_counter != counter)
+				is_symmetric_cus = false;
+			prev_counter = counter;
+		}
+		if (is_symmetric_cus) {
+			tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_CPC_DEBUG);
+			tmp = REG_SET_FIELD(tmp, CP_CPC_DEBUG, CPC_HARVESTING_RELAUNCH_DISABLE, 1);
+			tmp = REG_SET_FIELD(tmp, CP_CPC_DEBUG, CPC_HARVESTING_DISPATCH_DISABLE, 1);
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_CPC_DEBUG, tmp);
+		}
+		gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff,
+					    xcc_id);
+	}
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	cu_info->number = active_cu_number;
+	cu_info->ao_cu_mask = ao_cu_mask;
+	cu_info->simd_per_cu = NUM_SIMD_PER_CU;
+
+	return 0;
+}
+
+const struct amdgpu_ip_block_version gfx_v9_4_3_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_GFX,
+	.major = 9,
+	.minor = 4,
+	.rev = 3,
+	.funcs = &gfx_v9_4_3_ip_funcs,
+};
+
+static int gfx_v9_4_3_xcp_resume(void *handle, uint32_t inst_mask)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	uint32_t tmp_mask;
+	int i, r;
+
+	/* TODO : Initialize golden regs */
+	/* gfx_v9_4_3_init_golden_registers(adev); */
+
+	tmp_mask = inst_mask;
+	for_each_inst(i, tmp_mask)
+		gfx_v9_4_3_xcc_constants_init(adev, i);
+
+	if (!amdgpu_sriov_vf(adev)) {
+		tmp_mask = inst_mask;
+		for_each_inst(i, tmp_mask) {
+			r = gfx_v9_4_3_xcc_rlc_resume(adev, i);
+			if (r)
+				return r;
+		}
+	}
+
+	tmp_mask = inst_mask;
+	for_each_inst(i, tmp_mask) {
+		r = gfx_v9_4_3_xcc_cp_resume(adev, i);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+static int gfx_v9_4_3_xcp_suspend(void *handle, uint32_t inst_mask)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	int i;
+
+	for_each_inst(i, inst_mask)
+		gfx_v9_4_3_xcc_fini(adev, i);
+
+	return 0;
+}
+
+struct amdgpu_xcp_ip_funcs gfx_v9_4_3_xcp_funcs = {
+	.suspend = &gfx_v9_4_3_xcp_suspend,
+	.resume = &gfx_v9_4_3_xcp_resume
+};
+
+struct amdgpu_ras_block_hw_ops  gfx_v9_4_3_ras_ops = {
+	.query_ras_error_count = &gfx_v9_4_3_query_ras_error_count,
+	.reset_ras_error_count = &gfx_v9_4_3_reset_ras_error_count,
+};
+
+static int gfx_v9_4_3_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
+{
+	int r;
+
+	r = amdgpu_ras_block_late_init(adev, ras_block);
+	if (r)
+		return r;
+
+	r = amdgpu_ras_bind_aca(adev, AMDGPU_RAS_BLOCK__GFX,
+				&gfx_v9_4_3_aca_info,
+				NULL);
+	if (r)
+		goto late_fini;
+
+	return 0;
+
+late_fini:
+	amdgpu_ras_block_late_fini(adev, ras_block);
+
+	return r;
+}
+
+struct amdgpu_gfx_ras gfx_v9_4_3_ras = {
+	.ras_block = {
+		.hw_ops = &gfx_v9_4_3_ras_ops,
+		.ras_late_init = &gfx_v9_4_3_ras_late_init,
+	},
+	.enable_watchdog_timer = &gfx_v9_4_3_enable_watchdog_timer,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.h b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.h
index 84e69701b81a..42d67ee0e7ef 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.h
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.h
@@ -24,7 +24,8 @@
 #ifndef __GFX_V9_4_3_H__
 #define __GFX_V9_4_3_H__
 
-extern const struct amdgpu_gfx_funcs gfx_v9_4_3_gfx_funcs;
-extern const struct amdgpu_rlc_funcs gfx_v9_4_3_rlc_funcs;
+extern const struct amdgpu_ip_block_version gfx_v9_4_3_ip_block;
+
+extern struct amdgpu_xcp_ip_funcs gfx_v9_4_3_xcp_funcs;
 
 #endif /* __GFX_V9_4_3_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3_cleaner_shader.asm b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3_cleaner_shader.asm
new file mode 100644
index 000000000000..d5325ef80ab0
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3_cleaner_shader.asm
@@ -0,0 +1,153 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+// This shader is to clean LDS, SGPRs and VGPRs. It is  first 64 Dwords or 256 bytes of 192 Dwords cleaner shader.
+//To turn this shader program on for complitaion change this to main and lower shader main to main_1
+ 
+// MI300 : Clear SGPRs, VGPRs and LDS
+//   Uses two kernels launched separately:
+//   1. Clean VGPRs, LDS, and lower SGPRs
+//        Launches one workgroup per CU, each workgroup with 4x wave64 per SIMD in the CU
+//        Waves are "wave64" and have 128 VGPRs each, which uses all 512 VGPRs per SIMD
+//        Waves in the workgroup share the 64KB of LDS
+//        Each wave clears SGPRs 0 - 95. Because there are 4 waves/SIMD, this is physical SGPRs 0-383
+//        Each wave clears 128 VGPRs, so all 512 in the SIMD
+//        The first wave of the workgroup clears its 64KB of LDS
+//        The shader starts with "S_BARRIER" to ensure SPI has launched all waves of the workgroup
+//          before any wave in the workgroup could end.  Without this, it is possible not all SGPRs get cleared.
+//    2. Clean remaining SGPRs
+//        Launches a workgroup with 24 waves per workgroup, yielding 6 waves per SIMD in each CU
+//        Waves are allocating 96 SGPRs
+//          CP sets up SPI_RESOURCE_RESERVE_* registers to prevent these waves from allocating SGPRs 0-223.
+//          As such, these 6 waves per SIMD are allocated physical SGPRs 224-799
+//        Barriers do not work for >16 waves per workgroup, so we cannot start with S_BARRIER
+//          Instead, the shader starts with an S_SETHALT 1. Once all waves are launched CP will send unhalt command
+//        The shader then clears all SGPRs allocated to it, cleaning out physical SGPRs 224-799
+ 
+shader main
+  asic(MI300)
+  type(CS)
+  wave_size(64)
+// Note: original source code from SQ team
+
+//   (theorhetical fastest = ~512clks vgpr + 1536 lds + ~128 sgpr  = 2176 clks)
+
+  s_cmp_eq_u32 s0, 1                                // Bit0 is set, sgpr0 is set then clear VGPRS and LDS as FW set COMPUTE_USER_DATA_3
+  s_cbranch_scc0  label_0023                        // Clean VGPRs and LDS if sgpr0 of wave is set, scc = (s3 == 1)
+  S_BARRIER
+
+  s_movk_i32    m0, 0x0000
+  s_mov_b32     s2, 0x00000078  // Loop 128/8=16 times  (loop unrolled for performance)
+  //
+  // CLEAR VGPRs
+  //
+  s_set_gpr_idx_on  s2, 0x8    // enable Dest VGPR indexing
+label_0005:
+  v_mov_b32     v0, 0
+  v_mov_b32     v1, 0
+  v_mov_b32     v2, 0
+  v_mov_b32     v3, 0
+  v_mov_b32     v4, 0
+  v_mov_b32     v5, 0
+  v_mov_b32     v6, 0
+  v_mov_b32     v7, 0
+  s_sub_u32     s2, s2, 8
+  s_set_gpr_idx_idx  s2
+  s_cbranch_scc0  label_0005
+  s_set_gpr_idx_off
+ 
+  //
+  //
+ 
+  s_mov_b32     s2, 0x80000000                      // Bit31 is first_wave
+  s_and_b32     s2, s2, s1                          // sgpr0 has tg_size (first_wave) term as in ucode only COMPUTE_PGM_RSRC2.tg_size_en is set
+  s_cbranch_scc0  label_clean_sgpr_1                // Clean LDS if its first wave of ThreadGroup/WorkGroup
+  // CLEAR LDS
+  //
+  s_mov_b32 exec_lo, 0xffffffff
+  s_mov_b32 exec_hi, 0xffffffff
+  v_mbcnt_lo_u32_b32  v1, exec_hi, 0          // Set V1 to thread-ID (0..63)
+  v_mbcnt_hi_u32_b32  v1, exec_lo, v1         // Set V1 to thread-ID (0..63)
+  v_mul_u32_u24  v1, 0x00000008, v1           // * 8, so each thread is a double-dword address (8byte)
+  s_mov_b32     s2, 0x00000003f               // 64 loop iteraions
+  s_mov_b32     m0, 0xffffffff
+  // Clear all of LDS space
+  // Each FirstWave of WorkGroup clears 64kbyte block
+ 
+label_001F:
+  ds_write2_b64  v1, v[2:3], v[2:3] offset1:32
+  ds_write2_b64  v1, v[4:5], v[4:5] offset0:64 offset1:96
+  v_add_co_u32     v1, vcc, 0x00000400, v1
+  s_sub_u32     s2, s2, 1
+  s_cbranch_scc0  label_001F
+  //
+  // CLEAR SGPRs
+  //
+label_clean_sgpr_1:
+  s_mov_b32     m0, 0x0000005c   // Loop 96/4=24 times  (loop unrolled for performance)
+  s_nop 0
+label_sgpr_loop:
+  s_movreld_b32     s0, 0
+  s_movreld_b32     s1, 0
+  s_movreld_b32     s2, 0
+  s_movreld_b32     s3, 0
+  s_sub_u32         m0, m0, 4
+  s_cbranch_scc0  label_sgpr_loop
+ 
+  //clear vcc, flat scratch
+  s_mov_b32 flat_scratch_lo, 0   //clear  flat scratch lo SGPR 
+  s_mov_b32 flat_scratch_hi, 0   //clear  flat scratch hi SGPR 
+  s_mov_b64 vcc, 0               //clear vcc
+  s_mov_b64 ttmp0, 0             //Clear ttmp0 and ttmp1 
+  s_mov_b64 ttmp2, 0             //Clear ttmp2 and ttmp3 
+  s_mov_b64 ttmp4, 0             //Clear ttmp4 and ttmp5 
+  s_mov_b64 ttmp6, 0             //Clear ttmp6 and ttmp7 
+  s_mov_b64 ttmp8, 0             //Clear ttmp8 and ttmp9 
+  s_mov_b64 ttmp10, 0            //Clear ttmp10 and ttmp11 
+  s_mov_b64 ttmp12, 0            //Clear ttmp12 and ttmp13 
+  s_mov_b64 ttmp14, 0            //Clear ttmp14 and ttmp15 
+s_endpgm
+
+label_0023:
+
+  s_sethalt 1
+
+  s_mov_b32     m0, 0x0000005c   // Loop 96/4=24 times  (loop unrolled for performance)
+  s_nop 0
+label_sgpr_loop1:
+
+  s_movreld_b32     s0, 0
+  s_movreld_b32     s1, 0
+  s_movreld_b32     s2, 0
+  s_movreld_b32     s3, 0
+  s_sub_u32         m0, m0, 4
+  s_cbranch_scc0  label_sgpr_loop1
+ 
+  //clear vcc, flat scratch
+  s_mov_b32 flat_scratch_lo, 0   //clear  flat scratch lo SGPR 
+  s_mov_b32 flat_scratch_hi, 0   //clear  flat scratch hi SGPR 
+  s_mov_b64 vcc, 0xee            //clear vcc
+
+s_endpgm
+end  
+  
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3_cleaner_shader.h b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3_cleaner_shader.h
new file mode 100644
index 000000000000..69aa567c6c1d
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3_cleaner_shader.h
@@ -0,0 +1,64 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/* Define the cleaner shader gfx_9_4_3 */
+static const u32 gfx_9_4_3_cleaner_shader_hex[] = {
+	0xbf068100, 0xbf84003b,
+	0xbf8a0000, 0xb07c0000,
+	0xbe8200ff, 0x00000078,
+	0xbf110802, 0x7e000280,
+	0x7e020280, 0x7e040280,
+	0x7e060280, 0x7e080280,
+	0x7e0a0280, 0x7e0c0280,
+	0x7e0e0280, 0x80828802,
+	0xbe803202, 0xbf84fff5,
+	0xbf9c0000, 0xbe8200ff,
+	0x80000000, 0x86020102,
+	0xbf840011, 0xbefe00c1,
+	0xbeff00c1, 0xd28c0001,
+	0x0001007f, 0xd28d0001,
+	0x0002027e, 0x10020288,
+	0xbe8200bf, 0xbefc00c1,
+	0xd89c2000, 0x00020201,
+	0xd89c6040, 0x00040401,
+	0x320202ff, 0x00000400,
+	0x80828102, 0xbf84fff8,
+	0xbefc00ff, 0x0000005c,
+	0xbf800000, 0xbe802c80,
+	0xbe812c80, 0xbe822c80,
+	0xbe832c80, 0x80fc847c,
+	0xbf84fffa, 0xbee60080,
+	0xbee70080, 0xbeea0180,
+	0xbeec0180, 0xbeee0180,
+	0xbef00180, 0xbef20180,
+	0xbef40180, 0xbef60180,
+	0xbef80180, 0xbefa0180,
+	0xbf810000, 0xbf8d0001,
+	0xbefc00ff, 0x0000005c,
+	0xbf800000, 0xbe802c80,
+	0xbe812c80, 0xbe822c80,
+	0xbe832c80, 0x80fc847c,
+	0xbf84fffa, 0xbee60080,
+	0xbee70080, 0xbeea01ff,
+	0x000000ee, 0xbf810000,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v11_5_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v11_5_0.c
new file mode 100644
index 000000000000..f9949fedfbb9
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v11_5_0.c
@@ -0,0 +1,516 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "amdgpu.h"
+#include "gfxhub_v11_5_0.h"
+
+#include "gc/gc_11_5_0_offset.h"
+#include "gc/gc_11_5_0_sh_mask.h"
+
+#include "navi10_enum.h"
+#include "soc15_common.h"
+
+#define regGCVM_L2_CNTL3_DEFAULT                0x80100007
+#define regGCVM_L2_CNTL4_DEFAULT                0x000000c1
+#define regGCVM_L2_CNTL5_DEFAULT                0x00003fe0
+
+
+static const char *gfxhub_client_ids[] = {
+	"CB/DB",
+	"Reserved",
+	"GE1",
+	"GE2",
+	"CPF",
+	"CPC",
+	"CPG",
+	"RLC",
+	"TCP",
+	"SQC (inst)",
+	"SQC (data)",
+	"SQG",
+	"Reserved",
+	"SDMA0",
+	"SDMA1",
+	"GCR",
+	"SDMA2",
+	"SDMA3",
+};
+
+static uint32_t gfxhub_v11_5_0_get_invalidate_req(unsigned int vmid,
+					       uint32_t flush_type)
+{
+	u32 req = 0;
+
+	/* invalidate using legacy mode on vmid*/
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ,
+			    PER_VMID_INVALIDATE_REQ, 1 << vmid);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, flush_type);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PTES, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE0, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE1, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE2, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L1_PTES, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ,
+			    CLEAR_PROTECTION_FAULT_STATUS_ADDR,	0);
+
+	return req;
+}
+
+static void
+gfxhub_v11_5_0_print_l2_protection_fault_status(struct amdgpu_device *adev,
+					     uint32_t status)
+{
+	u32 cid = REG_GET_FIELD(status,
+				GCVM_L2_PROTECTION_FAULT_STATUS, CID);
+
+	dev_err(adev->dev,
+		"GCVM_L2_PROTECTION_FAULT_STATUS:0x%08X\n",
+		status);
+	dev_err(adev->dev, "\t Faulty UTCL2 client ID: %s (0x%x)\n",
+		cid >= ARRAY_SIZE(gfxhub_client_ids) ? "unknown" : gfxhub_client_ids[cid],
+		cid);
+	dev_err(adev->dev, "\t MORE_FAULTS: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS, MORE_FAULTS));
+	dev_err(adev->dev, "\t WALKER_ERROR: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS, WALKER_ERROR));
+	dev_err(adev->dev, "\t PERMISSION_FAULTS: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS, PERMISSION_FAULTS));
+	dev_err(adev->dev, "\t MAPPING_ERROR: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS, MAPPING_ERROR));
+	dev_err(adev->dev, "\t RW: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS, RW));
+}
+
+static u64 gfxhub_v11_5_0_get_fb_location(struct amdgpu_device *adev)
+{
+	u64 base = RREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_BASE);
+
+	base &= GCMC_VM_FB_LOCATION_BASE__FB_BASE_MASK;
+	base <<= 24;
+
+	return base;
+}
+
+static u64 gfxhub_v11_5_0_get_mc_fb_offset(struct amdgpu_device *adev)
+{
+	return (u64)RREG32_SOC15(GC, 0, regGCMC_VM_FB_OFFSET) << 24;
+}
+
+static void gfxhub_v11_5_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
+				uint64_t page_table_base)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+
+	WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
+			    hub->ctx_addr_distance * vmid,
+			    lower_32_bits(page_table_base));
+
+	WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
+			    hub->ctx_addr_distance * vmid,
+			    upper_32_bits(page_table_base));
+}
+
+static void gfxhub_v11_5_0_init_gart_aperture_regs(struct amdgpu_device *adev)
+{
+	uint64_t pt_base = amdgpu_gmc_pd_addr(adev->gart.bo);
+
+	gfxhub_v11_5_0_setup_vm_pt_regs(adev, 0, pt_base);
+
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
+		     (u32)(adev->gmc.gart_start >> 12));
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
+		     (u32)(adev->gmc.gart_start >> 44));
+
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
+		     (u32)(adev->gmc.gart_end >> 12));
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
+		     (u32)(adev->gmc.gart_end >> 44));
+}
+
+static void gfxhub_v11_5_0_init_system_aperture_regs(struct amdgpu_device *adev)
+{
+	uint64_t value;
+
+	WREG32_SOC15(GC, 0, regGCMC_VM_AGP_BASE, 0);
+	WREG32_SOC15(GC, 0, regGCMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
+	WREG32_SOC15(GC, 0, regGCMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
+
+	/* Program the system aperture low logical page number. */
+	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_LOW_ADDR,
+		     min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18);
+
+	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+		     max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
+
+	/* Set default page address. */
+	value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
+	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
+		     (u32)(value >> 12));
+	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
+		     (u32)(value >> 44));
+
+	/* Program "protection fault". */
+	WREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32,
+		     (u32)(adev->dummy_page_addr >> 12));
+	WREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32,
+		     (u32)((u64)adev->dummy_page_addr >> 44));
+
+	WREG32_FIELD15_PREREG(GC, 0, GCVM_L2_PROTECTION_FAULT_CNTL2,
+		       ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1);
+}
+
+static void gfxhub_v11_5_0_init_tlb_regs(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	/* Setup TLB control */
+	tmp = RREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL);
+
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE, 3);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL,
+			    ENABLE_ADVANCED_DRIVER_MODEL, 1);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL,
+			    SYSTEM_APERTURE_UNMAPPED_ACCESS, 0);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, ECO_BITS, 0);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL,
+			    MTYPE, MTYPE_UC); /* UC, uncached */
+
+	WREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, tmp);
+}
+
+static void gfxhub_v11_5_0_init_cache_regs(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	/* These registers are not accessible to VF-SRIOV.
+	 * The PF will program them instead.
+	 */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	/* Setup L2 cache */
+	tmp = RREG32_SOC15(GC, 0, regGCVM_L2_CNTL);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, ENABLE_L2_CACHE, 1);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL,
+			    ENABLE_DEFAULT_PAGE_OUT_TO_SYSTEM_MEMORY, 1);
+	/* XXX for emulation, Refer to closed source code.*/
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL,
+			    L2_PDE0_CACHE_TAG_GENERATION_MODE, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL, tmp);
+
+	tmp = RREG32_SOC15(GC, 0, regGCVM_L2_CNTL2);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL2, INVALIDATE_L2_CACHE, 1);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL2, tmp);
+
+	tmp = regGCVM_L2_CNTL3_DEFAULT;
+	if (adev->gmc.translate_further) {
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3, BANK_SELECT, 12);
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3,
+				    L2_CACHE_BIGK_FRAGMENT_SIZE, 9);
+	} else {
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3, BANK_SELECT, 9);
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3,
+				    L2_CACHE_BIGK_FRAGMENT_SIZE, 6);
+	}
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL3, tmp);
+
+	tmp = regGCVM_L2_CNTL4_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL4, tmp);
+
+	tmp = regGCVM_L2_CNTL5_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL5, L2_CACHE_SMALLK_FRAGMENT_SIZE, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL5, tmp);
+}
+
+static void gfxhub_v11_5_0_enable_system_domain(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	tmp = RREG32_SOC15(GC, 0, regGCVM_CONTEXT0_CNTL);
+	tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
+	tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT0_CNTL,
+			    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0);
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_CNTL, tmp);
+}
+
+static void gfxhub_v11_5_0_disable_identity_aperture(struct amdgpu_device *adev)
+{
+	/* These registers are not accessible to VF-SRIOV.
+	 * The PF will program them instead.
+	 */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32,
+		     0xFFFFFFFF);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32,
+		     0x0000000F);
+
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32,
+		     0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32,
+		     0);
+
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32, 0);
+
+}
+
+static void gfxhub_v11_5_0_setup_vmid_config(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	int i;
+	uint32_t tmp;
+
+	for (i = 0; i <= 14; i++) {
+		tmp = RREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL, i * hub->ctx_distance);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
+				    adev->vm_manager.num_level);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				PAGE_TABLE_BLOCK_SIZE,
+				adev->vm_manager.block_size - 9);
+		/* Send no-retry XNACK on fault to suppress VM fault storm. */
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT,
+				    !amdgpu_noretry);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL,
+				    i * hub->ctx_distance, tmp);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32,
+				    i * hub->ctx_addr_distance, 0);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32,
+				    i * hub->ctx_addr_distance, 0);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32,
+				    i * hub->ctx_addr_distance,
+				    lower_32_bits(adev->vm_manager.max_pfn - 1));
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32,
+				    i * hub->ctx_addr_distance,
+				    upper_32_bits(adev->vm_manager.max_pfn - 1));
+	}
+
+	hub->vm_cntx_cntl = tmp;
+}
+
+static void gfxhub_v11_5_0_program_invalidation(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	unsigned i;
+
+	for (i = 0 ; i < 18; ++i) {
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
+				    i * hub->eng_addr_distance, 0xffffffff);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ADDR_RANGE_HI32,
+				    i * hub->eng_addr_distance, 0x1f);
+	}
+}
+
+static int gfxhub_v11_5_0_gart_enable(struct amdgpu_device *adev)
+{
+	if (amdgpu_sriov_vf(adev)) {
+		/*
+		 * GCMC_VM_FB_LOCATION_BASE/TOP is NULL for VF, becuase they are
+		 * VF copy registers so vbios post doesn't program them, for
+		 * SRIOV driver need to program them
+		 */
+		WREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_BASE,
+			     adev->gmc.vram_start >> 24);
+		WREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_TOP,
+			     adev->gmc.vram_end >> 24);
+	}
+
+	/* GART Enable. */
+	gfxhub_v11_5_0_init_gart_aperture_regs(adev);
+	gfxhub_v11_5_0_init_system_aperture_regs(adev);
+	gfxhub_v11_5_0_init_tlb_regs(adev);
+	gfxhub_v11_5_0_init_cache_regs(adev);
+
+	gfxhub_v11_5_0_enable_system_domain(adev);
+	gfxhub_v11_5_0_disable_identity_aperture(adev);
+	gfxhub_v11_5_0_setup_vmid_config(adev);
+	gfxhub_v11_5_0_program_invalidation(adev);
+
+	return 0;
+}
+
+static void gfxhub_v11_5_0_gart_disable(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	u32 tmp;
+	u32 i;
+
+	/* Disable all tables */
+	for (i = 0; i < 16; i++)
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_CNTL,
+				    i * hub->ctx_distance, 0);
+
+	/* Setup TLB control */
+	tmp = RREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL,
+			    ENABLE_ADVANCED_DRIVER_MODEL, 0);
+	WREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, tmp);
+
+	/* Setup L2 cache */
+	WREG32_FIELD15_PREREG(GC, 0, GCVM_L2_CNTL, ENABLE_L2_CACHE, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL3, 0);
+}
+
+/**
+ * gfxhub_v11_5_0_set_fault_enable_default - update GART/VM fault handling
+ *
+ * @adev: amdgpu_device pointer
+ * @value: true redirects VM faults to the default page
+ */
+static void gfxhub_v11_5_0_set_fault_enable_default(struct amdgpu_device *adev,
+					  bool value)
+{
+	u32 tmp;
+
+	/* NO halt CP when page fault */
+	tmp = RREG32_SOC15(GC, 0, regCP_DEBUG);
+	tmp = REG_SET_FIELD(tmp, CP_DEBUG, CPG_UTCL1_ERROR_HALT_DISABLE, 1);
+	WREG32_SOC15(GC, 0, regCP_DEBUG, tmp);
+
+	/* These registers are not accessible to VF-SRIOV.
+	 * The PF will program them instead.
+	 */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	tmp = RREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT,
+			    value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    READ_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	if (!value) {
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_NO_RETRY_FAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_RETRY_FAULT, 1);
+	}
+	WREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL, tmp);
+}
+
+static const struct amdgpu_vmhub_funcs gfxhub_v11_5_0_vmhub_funcs = {
+	.print_l2_protection_fault_status = gfxhub_v11_5_0_print_l2_protection_fault_status,
+	.get_invalidate_req = gfxhub_v11_5_0_get_invalidate_req,
+};
+
+static void gfxhub_v11_5_0_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+
+	hub->ctx0_ptb_addr_lo32 =
+		SOC15_REG_OFFSET(GC, 0,
+				 regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32);
+	hub->ctx0_ptb_addr_hi32 =
+		SOC15_REG_OFFSET(GC, 0,
+				 regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32);
+	hub->vm_inv_eng0_sem =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_SEM);
+	hub->vm_inv_eng0_req =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_REQ);
+	hub->vm_inv_eng0_ack =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ACK);
+	hub->vm_context0_cntl =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_CONTEXT0_CNTL);
+	hub->vm_l2_pro_fault_status =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_L2_PROTECTION_FAULT_STATUS);
+	hub->vm_l2_pro_fault_cntl =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL);
+
+	hub->ctx_distance = regGCVM_CONTEXT1_CNTL - regGCVM_CONTEXT0_CNTL;
+	hub->ctx_addr_distance = regGCVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 -
+		regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32;
+	hub->eng_distance = regGCVM_INVALIDATE_ENG1_REQ -
+		regGCVM_INVALIDATE_ENG0_REQ;
+	hub->eng_addr_distance = regGCVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 -
+		regGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32;
+
+	hub->vm_cntx_cntl_vm_fault = GCVM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__VALID_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__READ_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__EXECUTE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK;
+
+	hub->vmhub_funcs = &gfxhub_v11_5_0_vmhub_funcs;
+}
+
+const struct amdgpu_gfxhub_funcs gfxhub_v11_5_0_funcs = {
+	.get_fb_location = gfxhub_v11_5_0_get_fb_location,
+	.get_mc_fb_offset = gfxhub_v11_5_0_get_mc_fb_offset,
+	.setup_vm_pt_regs = gfxhub_v11_5_0_setup_vm_pt_regs,
+	.gart_enable = gfxhub_v11_5_0_gart_enable,
+	.gart_disable = gfxhub_v11_5_0_gart_disable,
+	.set_fault_enable_default = gfxhub_v11_5_0_set_fault_enable_default,
+	.init = gfxhub_v11_5_0_init,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v11_5_0.h b/drivers/gpu/drm/amd/amdgpu/gfxhub_v11_5_0.h
new file mode 100644
index 000000000000..265ab631b3d0
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v11_5_0.h
@@ -0,0 +1,29 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __GFXHUB_V11_5_0_H__
+#define __GFXHUB_V11_5_0_H__
+
+extern const struct amdgpu_gfxhub_funcs gfxhub_v11_5_0_funcs;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v12_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v12_0.c
new file mode 100644
index 000000000000..7609b9cecae8
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v12_0.c
@@ -0,0 +1,521 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "amdgpu.h"
+#include "gfxhub_v12_0.h"
+
+#include "gc/gc_12_0_0_offset.h"
+#include "gc/gc_12_0_0_sh_mask.h"
+#include "soc24_enum.h"
+#include "soc15_common.h"
+
+#define regGCVM_L2_CNTL3_DEFAULT	0x80120007
+#define regGCVM_L2_CNTL4_DEFAULT	0x000000c1
+#define regGCVM_L2_CNTL5_DEFAULT	0x00003fe0
+#define regGRBM_GFX_INDEX_DEFAULT	0xe0000000
+
+static const char *gfxhub_client_ids[] = {
+	"CB",
+	"DB",
+	"GE1",
+	"GE2",
+	"CPF",
+	"CPC",
+	"CPG",
+	"RLC",
+	"TCP",
+	"SQC (inst)",
+	"SQC (data)",
+	"SQG/PC/SC",
+	"Reserved",
+	"SDMA0",
+	"SDMA1",
+	"GCR",
+	"Reserved",
+	"Reserved",
+	"WGS",
+	"DSM",
+	"PA"
+};
+
+static uint32_t gfxhub_v12_0_get_invalidate_req(unsigned int vmid,
+					        uint32_t flush_type)
+{
+	u32 req = 0;
+
+	/* invalidate using legacy mode on vmid*/
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ,
+			    PER_VMID_INVALIDATE_REQ, 1 << vmid);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, flush_type);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PTES, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE0, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE1, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE2, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L1_PTES, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ,
+			    CLEAR_PROTECTION_FAULT_STATUS_ADDR,	0);
+
+	return req;
+}
+
+static void
+gfxhub_v12_0_print_l2_protection_fault_status(struct amdgpu_device *adev,
+					      uint32_t status)
+{
+	u32 cid = REG_GET_FIELD(status,
+				GCVM_L2_PROTECTION_FAULT_STATUS_LO32, CID);
+
+	dev_err(adev->dev,
+		"GCVM_L2_PROTECTION_FAULT_STATUS:0x%08X\n",
+		status);
+	dev_err(adev->dev, "\t Faulty UTCL2 client ID: %s (0x%x)\n",
+		cid >= ARRAY_SIZE(gfxhub_client_ids) ? "unknown" : gfxhub_client_ids[cid],
+		cid);
+	dev_err(adev->dev, "\t MORE_FAULTS: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS_LO32, MORE_FAULTS));
+	dev_err(adev->dev, "\t WALKER_ERROR: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS_LO32, WALKER_ERROR));
+	dev_err(adev->dev, "\t PERMISSION_FAULTS: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS_LO32, PERMISSION_FAULTS));
+	dev_err(adev->dev, "\t MAPPING_ERROR: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS_LO32, MAPPING_ERROR));
+	dev_err(adev->dev, "\t RW: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS_LO32, RW));
+}
+
+static u64 gfxhub_v12_0_get_fb_location(struct amdgpu_device *adev)
+{
+	u64 base = RREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_BASE);
+
+	base &= GCMC_VM_FB_LOCATION_BASE__FB_BASE_MASK;
+	base <<= 24;
+
+	return base;
+}
+
+static u64 gfxhub_v12_0_get_mc_fb_offset(struct amdgpu_device *adev)
+{
+	return (u64)RREG32_SOC15(GC, 0, regGCMC_VM_FB_OFFSET) << 24;
+}
+
+static void gfxhub_v12_0_setup_vm_pt_regs(struct amdgpu_device *adev,
+					  uint32_t vmid,
+					  uint64_t page_table_base)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+
+	WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
+			    hub->ctx_addr_distance * vmid,
+			    lower_32_bits(page_table_base));
+
+	WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
+			    hub->ctx_addr_distance * vmid,
+			    upper_32_bits(page_table_base));
+}
+
+static void gfxhub_v12_0_init_gart_aperture_regs(struct amdgpu_device *adev)
+{
+	uint64_t pt_base = amdgpu_gmc_pd_addr(adev->gart.bo);
+
+	gfxhub_v12_0_setup_vm_pt_regs(adev, 0, pt_base);
+
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
+		     (u32)(adev->gmc.gart_start >> 12));
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
+		     (u32)(adev->gmc.gart_start >> 44));
+
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
+		     (u32)(adev->gmc.gart_end >> 12));
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
+		     (u32)(adev->gmc.gart_end >> 44));
+}
+
+static void gfxhub_v12_0_init_system_aperture_regs(struct amdgpu_device *adev)
+{
+	uint64_t value;
+
+	/* Program the AGP BAR */
+	WREG32_SOC15(GC, 0, regGCMC_VM_AGP_BASE, 0);
+	WREG32_SOC15(GC, 0, regGCMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
+	WREG32_SOC15(GC, 0, regGCMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
+
+	/* Program the system aperture low logical page number. */
+	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_LOW_ADDR,
+		     min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18);
+	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+		     max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
+
+	/* Set default page address. */
+	value = adev->mem_scratch.gpu_addr - adev->gmc.vram_start
+		+ adev->vm_manager.vram_base_offset;
+	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
+		     (u32)(value >> 12));
+	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
+		     (u32)(value >> 44));
+
+	/* Program "protection fault". */
+	WREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32,
+		     (u32)(adev->dummy_page_addr >> 12));
+	WREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32,
+		     (u32)((u64)adev->dummy_page_addr >> 44));
+
+	WREG32_FIELD15_PREREG(GC, 0, GCVM_L2_PROTECTION_FAULT_CNTL2,
+		       ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1);
+}
+
+
+static void gfxhub_v12_0_init_tlb_regs(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	/* Setup TLB control */
+	tmp = RREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL);
+
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE, 3);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL,
+			    ENABLE_ADVANCED_DRIVER_MODEL, 1);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL,
+			    SYSTEM_APERTURE_UNMAPPED_ACCESS, 0);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, ECO_BITS, 0);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL,
+			    MTYPE, MTYPE_UC); /* UC, uncached */
+
+	WREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, tmp);
+}
+
+static void gfxhub_v12_0_init_cache_regs(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	/* These registers are not accessible to VF-SRIOV.
+	 * The PF will program them instead.
+	 */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	/* Setup L2 cache */
+	tmp = RREG32_SOC15(GC, 0, regGCVM_L2_CNTL);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, ENABLE_L2_CACHE, 1);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL,
+			    ENABLE_DEFAULT_PAGE_OUT_TO_SYSTEM_MEMORY, 1);
+	/* XXX for emulation, Refer to closed source code.*/
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL,
+			    L2_PDE0_CACHE_TAG_GENERATION_MODE, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL, tmp);
+
+	tmp = RREG32_SOC15(GC, 0, regGCVM_L2_CNTL2);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL2, INVALIDATE_L2_CACHE, 1);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL2, tmp);
+
+	tmp = regGCVM_L2_CNTL3_DEFAULT;
+	if (adev->gmc.translate_further) {
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3, BANK_SELECT, 12);
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3,
+				    L2_CACHE_BIGK_FRAGMENT_SIZE, 9);
+	} else {
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3, BANK_SELECT, 9);
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3,
+				    L2_CACHE_BIGK_FRAGMENT_SIZE, 6);
+	}
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL3, tmp);
+
+	tmp = regGCVM_L2_CNTL4_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL4, tmp);
+
+	tmp = regGCVM_L2_CNTL5_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL5, L2_CACHE_SMALLK_FRAGMENT_SIZE, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL5, tmp);
+}
+
+static void gfxhub_v12_0_enable_system_domain(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	tmp = RREG32_SOC15(GC, 0, regGCVM_CONTEXT0_CNTL);
+	tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
+	tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT0_CNTL,
+			    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0);
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_CNTL, tmp);
+}
+
+static void gfxhub_v12_0_disable_identity_aperture(struct amdgpu_device *adev)
+{
+	/* These registers are not accessible to VF-SRIOV.
+	 * The PF will program them instead.
+	 */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32,
+		     0xFFFFFFFF);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32,
+		     0x0000000F);
+
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32,
+		     0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32,
+		     0);
+
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32, 0);
+
+}
+
+static void gfxhub_v12_0_setup_vmid_config(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	int i;
+	uint32_t tmp;
+
+	for (i = 0; i <= 14; i++) {
+		tmp = RREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL, i);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
+				    adev->vm_manager.num_level);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				PAGE_TABLE_BLOCK_SIZE,
+				adev->vm_manager.block_size - 9);
+		/* Send no-retry XNACK on fault to suppress VM fault storm. */
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT,
+				    !amdgpu_noretry);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL,
+				    i * hub->ctx_distance, tmp);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32,
+				    i * hub->ctx_addr_distance, 0);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32,
+				    i * hub->ctx_addr_distance, 0);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32,
+				    i * hub->ctx_addr_distance,
+				    lower_32_bits(adev->vm_manager.max_pfn - 1));
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32,
+				    i * hub->ctx_addr_distance,
+				    upper_32_bits(adev->vm_manager.max_pfn - 1));
+	}
+
+	hub->vm_cntx_cntl = tmp;
+}
+
+static void gfxhub_v12_0_program_invalidation(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	unsigned i;
+
+	for (i = 0 ; i < 18; ++i) {
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
+				    i * hub->eng_addr_distance, 0xffffffff);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ADDR_RANGE_HI32,
+				    i * hub->eng_addr_distance, 0x1f);
+	}
+}
+
+static int gfxhub_v12_0_gart_enable(struct amdgpu_device *adev)
+{
+	if (amdgpu_sriov_vf(adev)) {
+		/*
+		 * GCMC_VM_FB_LOCATION_BASE/TOP is NULL for VF, becuase they are
+		 * VF copy registers so vbios post doesn't program them, for
+		 * SRIOV driver need to program them
+		 */
+		WREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_BASE,
+			     adev->gmc.vram_start >> 24);
+		WREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_TOP,
+			     adev->gmc.vram_end >> 24);
+	}
+
+	/* GART Enable. */
+	gfxhub_v12_0_init_gart_aperture_regs(adev);
+	gfxhub_v12_0_init_system_aperture_regs(adev);
+	gfxhub_v12_0_init_tlb_regs(adev);
+	gfxhub_v12_0_init_cache_regs(adev);
+
+	gfxhub_v12_0_enable_system_domain(adev);
+	gfxhub_v12_0_disable_identity_aperture(adev);
+	gfxhub_v12_0_setup_vmid_config(adev);
+	gfxhub_v12_0_program_invalidation(adev);
+
+	return 0;
+}
+
+static void gfxhub_v12_0_gart_disable(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	u32 tmp;
+	u32 i;
+
+	/* Disable all tables */
+	for (i = 0; i < 16; i++)
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_CNTL,
+				    i * hub->ctx_distance, 0);
+
+	/* Setup TLB control */
+	tmp = RREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL,
+			    ENABLE_ADVANCED_DRIVER_MODEL, 0);
+	WREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, tmp);
+
+	/* Setup L2 cache */
+	WREG32_FIELD15_PREREG(GC, 0, GCVM_L2_CNTL, ENABLE_L2_CACHE, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL3, 0);
+}
+
+/**
+ * gfxhub_v12_0_set_fault_enable_default - update GART/VM fault handling
+ *
+ * @adev: amdgpu_device pointer
+ * @value: true redirects VM faults to the default page
+ */
+static void gfxhub_v12_0_set_fault_enable_default(struct amdgpu_device *adev,
+						  bool value)
+{
+	u32 tmp;
+
+	/* NO halt CP when page fault */
+	tmp = RREG32_SOC15(GC, 0, regCP_DEBUG);
+	tmp = REG_SET_FIELD(tmp, CP_DEBUG, CPG_UTCL1_ERROR_HALT_DISABLE, 1);
+	WREG32_SOC15(GC, 0, regCP_DEBUG, tmp);
+
+	/* These registers are not accessible to VF-SRIOV.
+	 * The PF will program them instead.
+	 */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	tmp = RREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT,
+			    value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    READ_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	if (!value) {
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_NO_RETRY_FAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_RETRY_FAULT, 1);
+	}
+	WREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL, tmp);
+}
+
+static const struct amdgpu_vmhub_funcs gfxhub_v12_0_vmhub_funcs = {
+	.print_l2_protection_fault_status = gfxhub_v12_0_print_l2_protection_fault_status,
+	.get_invalidate_req = gfxhub_v12_0_get_invalidate_req,
+};
+
+static void gfxhub_v12_0_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+
+	hub->ctx0_ptb_addr_lo32 =
+		SOC15_REG_OFFSET(GC, 0,
+				 regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32);
+	hub->ctx0_ptb_addr_hi32 =
+		SOC15_REG_OFFSET(GC, 0,
+				 regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32);
+	hub->vm_inv_eng0_sem =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_SEM);
+	hub->vm_inv_eng0_req =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_REQ);
+	hub->vm_inv_eng0_ack =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ACK);
+	hub->vm_context0_cntl =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_CONTEXT0_CNTL);
+	hub->vm_l2_pro_fault_status =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_L2_PROTECTION_FAULT_STATUS_LO32);
+	hub->vm_l2_pro_fault_cntl =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL);
+
+	hub->ctx_distance = regGCVM_CONTEXT1_CNTL - regGCVM_CONTEXT0_CNTL;
+	hub->ctx_addr_distance = regGCVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 -
+		regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32;
+	hub->eng_distance = regGCVM_INVALIDATE_ENG1_REQ -
+		regGCVM_INVALIDATE_ENG0_REQ;
+	hub->eng_addr_distance = regGCVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 -
+		regGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32;
+
+	hub->vm_cntx_cntl_vm_fault = GCVM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__VALID_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__READ_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__EXECUTE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK;
+
+	hub->vmhub_funcs = &gfxhub_v12_0_vmhub_funcs;
+}
+
+const struct amdgpu_gfxhub_funcs gfxhub_v12_0_funcs = {
+	.get_fb_location = gfxhub_v12_0_get_fb_location,
+	.get_mc_fb_offset = gfxhub_v12_0_get_mc_fb_offset,
+	.setup_vm_pt_regs = gfxhub_v12_0_setup_vm_pt_regs,
+	.gart_enable = gfxhub_v12_0_gart_enable,
+	.gart_disable = gfxhub_v12_0_gart_disable,
+	.set_fault_enable_default = gfxhub_v12_0_set_fault_enable_default,
+	.init = gfxhub_v12_0_init,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v12_0.h b/drivers/gpu/drm/amd/amdgpu/gfxhub_v12_0.h
new file mode 100644
index 000000000000..f1258265f802
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v12_0.h
@@ -0,0 +1,29 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __GFXHUB_V12_0_H__
+#define __GFXHUB_V12_0_H__
+
+extern const struct amdgpu_gfxhub_funcs gfxhub_v12_0_funcs;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c
index ab2325f6c7ac..a7bfc9f41d0e 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c
@@ -40,7 +40,7 @@ static void gfxhub_v1_0_setup_vm_pt_regs(struct amdgpu_device *adev,
 					 uint32_t vmid,
 					 uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	WREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
 			    hub->ctx_addr_distance * vmid,
@@ -92,18 +92,20 @@ static void gfxhub_v1_0_init_system_aperture_regs(struct amdgpu_device *adev)
 {
 	uint64_t value;
 
-	/* Program the AGP BAR */
-	WREG32_SOC15_RLC(GC, 0, mmMC_VM_AGP_BASE, 0);
-	WREG32_SOC15_RLC(GC, 0, mmMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
-	WREG32_SOC15_RLC(GC, 0, mmMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
-
 	if (!amdgpu_sriov_vf(adev) || adev->asic_type <= CHIP_VEGA10) {
+		/* Program the AGP BAR */
+		WREG32_SOC15_RLC(GC, 0, mmMC_VM_AGP_BASE, 0);
+		WREG32_SOC15_RLC(GC, 0, mmMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
+		WREG32_SOC15_RLC(GC, 0, mmMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
+
 		/* Program the system aperture low logical page number. */
 		WREG32_SOC15_RLC(GC, 0, mmMC_VM_SYSTEM_APERTURE_LOW_ADDR,
 			min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18);
 
-		if (adev->apu_flags & AMD_APU_IS_RAVEN2)
-			/*
+		if (adev->apu_flags & (AMD_APU_IS_RAVEN2 |
+				       AMD_APU_IS_RENOIR |
+				       AMD_APU_IS_GREEN_SARDINE))
+		       /*
 			* Raven2 has a HW issue that it is unable to use the
 			* vram which is out of MC_VM_SYSTEM_APERTURE_HIGH_ADDR.
 			* So here is the workaround that increase system
@@ -247,8 +249,8 @@ static void gfxhub_v1_0_disable_identity_aperture(struct amdgpu_device *adev)
 
 static void gfxhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	unsigned num_level, block_size;
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	unsigned int num_level, block_size;
 	uint32_t tmp;
 	int i;
 
@@ -260,7 +262,7 @@ static void gfxhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
 		block_size -= 9;
 
 	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT1_CNTL, i);
+		tmp = RREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT1_CNTL, i * hub->ctx_distance);
 		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
 		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
 				    num_level);
@@ -307,8 +309,8 @@ static void gfxhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
 
 static void gfxhub_v1_0_program_invalidation(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	unsigned i;
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	unsigned int i;
 
 	for (i = 0 ; i < 18; ++i) {
 		WREG32_SOC15_OFFSET(GC, 0, mmVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
@@ -338,7 +340,7 @@ static int gfxhub_v1_0_gart_enable(struct amdgpu_device *adev)
 
 static void gfxhub_v1_0_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	u32 tmp;
 	u32 i;
 
@@ -375,6 +377,7 @@ static void gfxhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev,
 						 bool value)
 {
 	u32 tmp;
+
 	tmp = RREG32_SOC15(GC, 0, mmVM_L2_PROTECTION_FAULT_CNTL);
 	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
 			RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
@@ -411,7 +414,7 @@ static void gfxhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev,
 
 static void gfxhub_v1_0_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	hub->ctx0_ptb_addr_lo32 =
 		SOC15_REG_OFFSET(GC, 0,
@@ -440,7 +443,6 @@ static void gfxhub_v1_0_init(struct amdgpu_device *adev)
 		mmVM_INVALIDATE_ENG0_ADDR_RANGE_LO32;
 }
 
-
 const struct amdgpu_gfxhub_funcs gfxhub_v1_0_funcs = {
 	.get_mc_fb_offset = gfxhub_v1_0_get_mc_fb_offset,
 	.setup_vm_pt_regs = gfxhub_v1_0_setup_vm_pt_regs,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
index c59c6c85fbff..6c03bf9f1ae8 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
@@ -21,6 +21,7 @@
  *
  */
 #include "amdgpu.h"
+#include "amdgpu_xcp.h"
 #include "gfxhub_v1_2.h"
 #include "gfxhub_v1_1.h"
 
@@ -35,227 +36,302 @@
 
 static u64 gfxhub_v1_2_get_mc_fb_offset(struct amdgpu_device *adev)
 {
-	return (u64)RREG32_SOC15(GC, 0, regMC_VM_FB_OFFSET) << 24;
+	return (u64)RREG32_SOC15(GC, GET_INST(GC, 0), regMC_VM_FB_OFFSET) << 24;
+}
+
+static void gfxhub_v1_2_xcc_setup_vm_pt_regs(struct amdgpu_device *adev,
+					     uint32_t vmid,
+					     uint64_t page_table_base,
+					     uint32_t xcc_mask)
+{
+	struct amdgpu_vmhub *hub;
+	int i;
+
+	for_each_inst(i, xcc_mask) {
+		hub = &adev->vmhub[AMDGPU_GFXHUB(i)];
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, i),
+				    regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
+				    hub->ctx_addr_distance * vmid,
+				    lower_32_bits(page_table_base));
+
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, i),
+				    regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
+				    hub->ctx_addr_distance * vmid,
+				    upper_32_bits(page_table_base));
+	}
 }
 
 static void gfxhub_v1_2_setup_vm_pt_regs(struct amdgpu_device *adev,
 					 uint32_t vmid,
 					 uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-
-	WREG32_SOC15_OFFSET(GC, 0, regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
-			    hub->ctx_addr_distance * vmid,
-			    lower_32_bits(page_table_base));
+	uint32_t xcc_mask;
 
-	WREG32_SOC15_OFFSET(GC, 0, regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
-			    hub->ctx_addr_distance * vmid,
-			    upper_32_bits(page_table_base));
+	xcc_mask = GENMASK(NUM_XCC(adev->gfx.xcc_mask) - 1, 0);
+	gfxhub_v1_2_xcc_setup_vm_pt_regs(adev, vmid, page_table_base, xcc_mask);
 }
 
-static void gfxhub_v1_2_init_gart_aperture_regs(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_init_gart_aperture_regs(struct amdgpu_device *adev,
+						    uint32_t xcc_mask)
 {
+	uint64_t gart_start = amdgpu_virt_xgmi_migrate_enabled(adev) ?
+			adev->gmc.vram_start : adev->gmc.fb_start;
 	uint64_t pt_base;
+	int i;
 
 	if (adev->gmc.pdb0_bo)
 		pt_base = amdgpu_gmc_pd_addr(adev->gmc.pdb0_bo);
 	else
 		pt_base = amdgpu_gmc_pd_addr(adev->gart.bo);
 
-	gfxhub_v1_2_setup_vm_pt_regs(adev, 0, pt_base);
+	gfxhub_v1_2_xcc_setup_vm_pt_regs(adev, 0, pt_base, xcc_mask);
 
 	/* If use GART for FB translation, vmid0 page table covers both
 	 * vram and system memory (gart)
 	 */
-	if (adev->gmc.pdb0_bo) {
-		WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
-				(u32)(adev->gmc.fb_start >> 12));
-		WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
-				(u32)(adev->gmc.fb_start >> 44));
-
-		WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
-				(u32)(adev->gmc.gart_end >> 12));
-		WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
-				(u32)(adev->gmc.gart_end >> 44));
-	} else {
-		WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
-				(u32)(adev->gmc.gart_start >> 12));
-		WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
-				(u32)(adev->gmc.gart_start >> 44));
-
-		WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
-				(u32)(adev->gmc.gart_end >> 12));
-		WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
-				(u32)(adev->gmc.gart_end >> 44));
+	for_each_inst(i, xcc_mask) {
+		if (adev->gmc.pdb0_bo) {
+			WREG32_SOC15(GC, GET_INST(GC, i),
+				     regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
+				     (u32)(gart_start >> 12));
+			WREG32_SOC15(GC, GET_INST(GC, i),
+				     regVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
+				     (u32)(gart_start >> 44));
+
+			WREG32_SOC15(GC, GET_INST(GC, i),
+				     regVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
+				     (u32)(adev->gmc.gart_end >> 12));
+			WREG32_SOC15(GC, GET_INST(GC, i),
+				     regVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
+				     (u32)(adev->gmc.gart_end >> 44));
+		} else {
+			WREG32_SOC15(GC, GET_INST(GC, i),
+				     regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
+				     (u32)(adev->gmc.gart_start >> 12));
+			WREG32_SOC15(GC, GET_INST(GC, i),
+				     regVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
+				     (u32)(adev->gmc.gart_start >> 44));
+
+			WREG32_SOC15(GC, GET_INST(GC, i),
+				     regVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
+				     (u32)(adev->gmc.gart_end >> 12));
+			WREG32_SOC15(GC, GET_INST(GC, i),
+				     regVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
+				     (u32)(adev->gmc.gart_end >> 44));
+		}
 	}
 }
 
-static void gfxhub_v1_2_init_system_aperture_regs(struct amdgpu_device *adev)
+static void
+gfxhub_v1_2_xcc_init_system_aperture_regs(struct amdgpu_device *adev,
+					  uint32_t xcc_mask)
 {
 	uint64_t value;
 	uint32_t tmp;
+	int i;
 
-	/* Program the AGP BAR */
-	WREG32_SOC15_RLC(GC, 0, regMC_VM_AGP_BASE, 0);
-	WREG32_SOC15_RLC(GC, 0, regMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
-	WREG32_SOC15_RLC(GC, 0, regMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
-
-	if (!amdgpu_sriov_vf(adev) || adev->asic_type <= CHIP_VEGA10) {
-		/* Program the system aperture low logical page number. */
-		WREG32_SOC15_RLC(GC, 0, regMC_VM_SYSTEM_APERTURE_LOW_ADDR,
-			min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18);
-
-		if (adev->apu_flags & AMD_APU_IS_RAVEN2)
-			/*
-			* Raven2 has a HW issue that it is unable to use the
-			* vram which is out of MC_VM_SYSTEM_APERTURE_HIGH_ADDR.
-			* So here is the workaround that increase system
-			* aperture high address (add 1) to get rid of the VM
-			* fault and hardware hang.
-			*/
-			WREG32_SOC15_RLC(GC, 0,
-					 regMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
-					 max((adev->gmc.fb_end >> 18) + 0x1,
-					     adev->gmc.agp_end >> 18));
-		else
-			WREG32_SOC15_RLC(GC, 0,
-				regMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
-				max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
-
-		/* Set default page address. */
-		value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
-		WREG32_SOC15(GC, 0, regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
-			     (u32)(value >> 12));
-		WREG32_SOC15(GC, 0, regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
-			     (u32)(value >> 44));
-
-		/* Program "protection fault". */
-		WREG32_SOC15(GC, 0, regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32,
-			     (u32)(adev->dummy_page_addr >> 12));
-		WREG32_SOC15(GC, 0, regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32,
-			     (u32)((u64)adev->dummy_page_addr >> 44));
-
-		tmp = RREG32_SOC15(GC, 0, regVM_L2_PROTECTION_FAULT_CNTL2);
-		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL2,
-				    ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1);
-		WREG32_SOC15(GC, 0, regVM_L2_PROTECTION_FAULT_CNTL2, tmp);
-	}
-
-	/* In the case squeezing vram into GART aperture, we don't use
-	 * FB aperture and AGP aperture. Disable them.
-	 */
-	if (adev->gmc.pdb0_bo) {
-		WREG32_SOC15(GC, 0, regMC_VM_FB_LOCATION_TOP, 0);
-		WREG32_SOC15(GC, 0, regMC_VM_FB_LOCATION_BASE, 0x00FFFFFF);
-		WREG32_SOC15(GC, 0, regMC_VM_AGP_TOP, 0);
-		WREG32_SOC15(GC, 0, regMC_VM_AGP_BOT, 0xFFFFFF);
-		WREG32_SOC15(GC, 0, regMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x3FFFFFFF);
-		WREG32_SOC15(GC, 0, regMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0);
+	for_each_inst(i, xcc_mask) {
+		/* Program the AGP BAR */
+		WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_AGP_BASE, 0);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
+
+		if (!amdgpu_sriov_vf(adev) || adev->asic_type <= CHIP_VEGA10) {
+			/* Program the system aperture low logical page number. */
+			WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_SYSTEM_APERTURE_LOW_ADDR,
+				min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18);
+
+			if (adev->apu_flags & (AMD_APU_IS_RAVEN2 |
+					       AMD_APU_IS_RENOIR |
+					       AMD_APU_IS_GREEN_SARDINE))
+			       /*
+				* Raven2 has a HW issue that it is unable to use the
+				* vram which is out of MC_VM_SYSTEM_APERTURE_HIGH_ADDR.
+				* So here is the workaround that increase system
+				* aperture high address (add 1) to get rid of the VM
+				* fault and hardware hang.
+				*/
+				WREG32_SOC15_RLC(GC, GET_INST(GC, i),
+						 regMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+						 max((adev->gmc.fb_end >> 18) + 0x1,
+						     adev->gmc.agp_end >> 18));
+			else
+				WREG32_SOC15_RLC(GC, GET_INST(GC, i),
+					regMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+					max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
+
+			/* Set default page address. */
+			value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
+			WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
+				     (u32)(value >> 12));
+			WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
+				     (u32)(value >> 44));
+
+			/* Program "protection fault". */
+			WREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32,
+				     (u32)(adev->dummy_page_addr >> 12));
+			WREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32,
+				     (u32)((u64)adev->dummy_page_addr >> 44));
+
+			tmp = RREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_CNTL2);
+			tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL2,
+					    ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1);
+			WREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_CNTL2, tmp);
+		}
+
+		/* In the case squeezing vram into GART aperture, we don't use
+		 * FB aperture and AGP aperture. Disable them.
+		 */
+		if (adev->gmc.pdb0_bo && adev->gmc.xgmi.connected_to_cpu) {
+			WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_FB_LOCATION_TOP, 0);
+			WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_FB_LOCATION_BASE, 0x00FFFFFF);
+			WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_AGP_TOP, 0);
+			WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_AGP_BOT, 0xFFFFFF);
+			WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x3FFFFFFF);
+			WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0);
+		}
 	}
 }
 
-static void gfxhub_v1_2_init_tlb_regs(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_init_tlb_regs(struct amdgpu_device *adev,
+					  uint32_t xcc_mask)
 {
 	uint32_t tmp;
+	int i;
 
-	/* Setup TLB control */
-	tmp = RREG32_SOC15(GC, 0, regMC_VM_MX_L1_TLB_CNTL);
-
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1);
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE, 3);
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
-			    ENABLE_ADVANCED_DRIVER_MODEL, 1);
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
-			    SYSTEM_APERTURE_UNMAPPED_ACCESS, 0);
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
-			    MTYPE, MTYPE_UC);/* XXX for emulation. */
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ATC_EN, 1);
-
-	WREG32_SOC15_RLC(GC, 0, regMC_VM_MX_L1_TLB_CNTL, tmp);
+	for_each_inst(i, xcc_mask) {
+		/* Setup TLB control */
+		tmp = RREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_MX_L1_TLB_CNTL);
+
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+				    ENABLE_L1_TLB, 1);
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+				    SYSTEM_ACCESS_MODE, 3);
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+				    ENABLE_ADVANCED_DRIVER_MODEL, 1);
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+				    SYSTEM_APERTURE_UNMAPPED_ACCESS, 0);
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+				    MTYPE, MTYPE_UC);/* XXX for emulation. */
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ATC_EN, 1);
+
+		WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_MX_L1_TLB_CNTL, tmp);
+	}
 }
 
-static void gfxhub_v1_2_init_cache_regs(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_init_cache_regs(struct amdgpu_device *adev,
+					    uint32_t xcc_mask)
 {
 	uint32_t tmp;
+	int i;
 
-	/* Setup L2 cache */
-	tmp = RREG32_SOC15(GC, 0, regVM_L2_CNTL);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 1);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 1);
-	/* XXX for emulation, Refer to closed source code.*/
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, L2_PDE0_CACHE_TAG_GENERATION_MODE,
-			    0);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0);
-	WREG32_SOC15_RLC(GC, 0, regVM_L2_CNTL, tmp);
-
-	tmp = RREG32_SOC15(GC, 0, regVM_L2_CNTL2);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_L2_CACHE, 1);
-	WREG32_SOC15_RLC(GC, 0, regVM_L2_CNTL2, tmp);
-
-	tmp = regVM_L2_CNTL3_DEFAULT;
-	if (adev->gmc.translate_further) {
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 12);
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3,
-				    L2_CACHE_BIGK_FRAGMENT_SIZE, 9);
-	} else {
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 9);
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3,
-				    L2_CACHE_BIGK_FRAGMENT_SIZE, 6);
+	for_each_inst(i, xcc_mask) {
+		/* Setup L2 cache */
+		tmp = RREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_CNTL);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 1);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 1);
+		/* XXX for emulation, Refer to closed source code.*/
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, L2_PDE0_CACHE_TAG_GENERATION_MODE,
+				    0);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, i), regVM_L2_CNTL, tmp);
+
+		tmp = RREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_CNTL2);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_L2_CACHE, 1);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, i), regVM_L2_CNTL2, tmp);
+
+		tmp = regVM_L2_CNTL3_DEFAULT;
+		if (adev->gmc.translate_further) {
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 12);
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3,
+					    L2_CACHE_BIGK_FRAGMENT_SIZE, 9);
+		} else {
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 9);
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3,
+					    L2_CACHE_BIGK_FRAGMENT_SIZE, 6);
+		}
+		WREG32_SOC15_RLC(GC, GET_INST(GC, i), regVM_L2_CNTL3, tmp);
+
+		tmp = regVM_L2_CNTL4_DEFAULT;
+		/* For AMD APP APUs setup WC memory */
+		if (adev->gmc.xgmi.connected_to_cpu || adev->gmc.is_app_apu) {
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 1);
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 1);
+		} else {
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0);
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 0);
+		}
+		WREG32_SOC15_RLC(GC, GET_INST(GC, i), regVM_L2_CNTL4, tmp);
 	}
-	WREG32_SOC15_RLC(GC, 0, regVM_L2_CNTL3, tmp);
-
-	tmp = regVM_L2_CNTL4_DEFAULT;
-	if (adev->gmc.xgmi.connected_to_cpu) {
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 1);
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 1);
-	} else {
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0);
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 0);
-	}
-	WREG32_SOC15_RLC(GC, 0, regVM_L2_CNTL4, tmp);
 }
 
-static void gfxhub_v1_2_enable_system_domain(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_enable_system_domain(struct amdgpu_device *adev,
+						 uint32_t xcc_mask)
 {
 	uint32_t tmp;
+	int i;
 
-	tmp = RREG32_SOC15(GC, 0, regVM_CONTEXT0_CNTL);
-	tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
-	tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH,
-			adev->gmc.vmid0_page_table_depth);
-	tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_BLOCK_SIZE,
-			adev->gmc.vmid0_page_table_block_size);
-	tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL,
-			    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0);
-	WREG32_SOC15(GC, 0, regVM_CONTEXT0_CNTL, tmp);
+	for_each_inst(i, xcc_mask) {
+		tmp = RREG32_SOC15(GC, GET_INST(GC, i), regVM_CONTEXT0_CNTL);
+		tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
+		tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH,
+				adev->gmc.vmid0_page_table_depth);
+		tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_BLOCK_SIZE,
+				adev->gmc.vmid0_page_table_block_size);
+		tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL,
+				    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0);
+		WREG32_SOC15(GC, GET_INST(GC, i), regVM_CONTEXT0_CNTL, tmp);
+	}
 }
 
-static void gfxhub_v1_2_disable_identity_aperture(struct amdgpu_device *adev)
+static void
+gfxhub_v1_2_xcc_disable_identity_aperture(struct amdgpu_device *adev,
+					  uint32_t xcc_mask)
 {
-	WREG32_SOC15(GC, 0, regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32,
-		     0XFFFFFFFF);
-	WREG32_SOC15(GC, 0, regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32,
-		     0x0000000F);
-
-	WREG32_SOC15(GC, 0, regVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32,
-		     0);
-	WREG32_SOC15(GC, 0, regVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32,
-		     0);
+	int i;
 
-	WREG32_SOC15(GC, 0, regVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32, 0);
-	WREG32_SOC15(GC, 0, regVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32, 0);
+	for_each_inst(i, xcc_mask) {
+		WREG32_SOC15(GC, GET_INST(GC, i),
+			     regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32,
+			     0XFFFFFFFF);
+		WREG32_SOC15(GC, GET_INST(GC, i),
+			     regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32,
+			     0x0000000F);
+
+		WREG32_SOC15(GC, GET_INST(GC, i),
+			     regVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32,
+			     0);
+		WREG32_SOC15(GC, GET_INST(GC, i),
+			     regVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32,
+			     0);
+
+		WREG32_SOC15(GC, GET_INST(GC, i),
+			     regVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32, 0);
+		WREG32_SOC15(GC, GET_INST(GC, i),
+			     regVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32, 0);
+	}
+}
 
+static inline bool
+gfxhub_v1_2_per_process_xnack_support(struct amdgpu_device *adev)
+{
+	/*
+	 * TODO: Check if this function is really needed, so far only 9.4.3
+	 * variants use GFXHUB 1.2
+	 */
+	return !!adev->aid_mask;
 }
 
-static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_setup_vmid_config(struct amdgpu_device *adev,
+					      uint32_t xcc_mask)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	unsigned num_level, block_size;
+	struct amdgpu_vmhub *hub;
+	unsigned int num_level, block_size;
 	uint32_t tmp;
-	int i;
+	int i, j;
 
 	num_level = adev->vm_manager.num_level;
 	block_size = adev->vm_manager.block_size;
@@ -264,124 +340,190 @@ static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
 	else
 		block_size -= 9;
 
-	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(GC, 0, regVM_CONTEXT1_CNTL, i);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
-				    num_level);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT,
-				    1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    PAGE_TABLE_BLOCK_SIZE,
-				    block_size);
-		/* Send no-retry XNACK on fault to suppress VM fault storm.
-		 * On Aldebaran, XNACK can be enabled in the SQ per-process.
-		 * Retry faults need to be enabled for that to work.
-		 */
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT,
-				    !adev->gmc.noretry ||
-				    adev->asic_type == CHIP_ALDEBARAN);
-		WREG32_SOC15_OFFSET(GC, 0, regVM_CONTEXT1_CNTL,
-				    i * hub->ctx_distance, tmp);
-		WREG32_SOC15_OFFSET(GC, 0,
-				    regVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32,
-				    i * hub->ctx_addr_distance, 0);
-		WREG32_SOC15_OFFSET(GC, 0,
-				    regVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32,
-				    i * hub->ctx_addr_distance, 0);
-		WREG32_SOC15_OFFSET(GC, 0,
-				    regVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32,
-				    i * hub->ctx_addr_distance,
-				    lower_32_bits(adev->vm_manager.max_pfn - 1));
-		WREG32_SOC15_OFFSET(GC, 0,
-				    regVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32,
-				    i * hub->ctx_addr_distance,
-				    upper_32_bits(adev->vm_manager.max_pfn - 1));
+	for_each_inst(j, xcc_mask) {
+		hub = &adev->vmhub[AMDGPU_GFXHUB(j)];
+		for (i = 0; i <= 14; i++) {
+			tmp = RREG32_SOC15_OFFSET(GC, GET_INST(GC, j), regVM_CONTEXT1_CNTL,
+					i * hub->ctx_distance);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
+					    num_level);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT,
+					    1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    PAGE_TABLE_BLOCK_SIZE,
+					    block_size);
+			/* Send no-retry XNACK on fault to suppress VM fault storm.
+			 * On 9.4.3 variants, XNACK can be enabled in
+			 * the SQ per-process.
+			 * Retry faults need to be enabled for that to work.
+			 */
+			tmp = REG_SET_FIELD(
+				tmp, VM_CONTEXT1_CNTL,
+				RETRY_PERMISSION_OR_INVALID_PAGE_FAULT,
+				!adev->gmc.noretry ||
+					gfxhub_v1_2_per_process_xnack_support(
+						adev));
+			WREG32_SOC15_OFFSET(GC, GET_INST(GC, j), regVM_CONTEXT1_CNTL,
+					    i * hub->ctx_distance, tmp);
+			WREG32_SOC15_OFFSET(GC, GET_INST(GC, j),
+					    regVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32,
+					    i * hub->ctx_addr_distance, 0);
+			WREG32_SOC15_OFFSET(GC, GET_INST(GC, j),
+					    regVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32,
+					    i * hub->ctx_addr_distance, 0);
+			WREG32_SOC15_OFFSET(GC, GET_INST(GC, j),
+					    regVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32,
+					    i * hub->ctx_addr_distance,
+					    lower_32_bits(adev->vm_manager.max_pfn - 1));
+			WREG32_SOC15_OFFSET(GC, GET_INST(GC, j),
+					    regVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32,
+					    i * hub->ctx_addr_distance,
+					    upper_32_bits(adev->vm_manager.max_pfn - 1));
+		}
 	}
 }
 
-static void gfxhub_v1_2_program_invalidation(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_program_invalidation(struct amdgpu_device *adev,
+						 uint32_t xcc_mask)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	unsigned i;
-
-	for (i = 0 ; i < 18; ++i) {
-		WREG32_SOC15_OFFSET(GC, 0, regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
-				    i * hub->eng_addr_distance, 0xffffffff);
-		WREG32_SOC15_OFFSET(GC, 0, regVM_INVALIDATE_ENG0_ADDR_RANGE_HI32,
-				    i * hub->eng_addr_distance, 0x1f);
+	struct amdgpu_vmhub *hub;
+	unsigned int i, j;
+
+	for_each_inst(j, xcc_mask) {
+		hub = &adev->vmhub[AMDGPU_GFXHUB(j)];
+
+		for (i = 0 ; i < 18; ++i) {
+			WREG32_SOC15_OFFSET(GC, GET_INST(GC, j), regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
+					    i * hub->eng_addr_distance, 0xffffffff);
+			WREG32_SOC15_OFFSET(GC, GET_INST(GC, j), regVM_INVALIDATE_ENG0_ADDR_RANGE_HI32,
+					    i * hub->eng_addr_distance, 0x1f);
+		}
 	}
 }
 
-static int gfxhub_v1_2_gart_enable(struct amdgpu_device *adev)
+static int gfxhub_v1_2_xcc_gart_enable(struct amdgpu_device *adev,
+				       uint32_t xcc_mask)
 {
-	if (amdgpu_sriov_vf(adev) && adev->asic_type != CHIP_ARCTURUS) {
-		/*
-		 * MC_VM_FB_LOCATION_BASE/TOP is NULL for VF, becuase they are
-		 * VF copy registers so vbios post doesn't program them, for
-		 * SRIOV driver need to program them
-		 */
-		WREG32_SOC15_RLC(GC, 0, regMC_VM_FB_LOCATION_BASE,
-			     adev->gmc.vram_start >> 24);
-		WREG32_SOC15_RLC(GC, 0, regMC_VM_FB_LOCATION_TOP,
-			     adev->gmc.vram_end >> 24);
-	}
-
 	/* GART Enable. */
-	gfxhub_v1_2_init_gart_aperture_regs(adev);
-	gfxhub_v1_2_init_system_aperture_regs(adev);
-	gfxhub_v1_2_init_tlb_regs(adev);
+	gfxhub_v1_2_xcc_init_gart_aperture_regs(adev, xcc_mask);
+	gfxhub_v1_2_xcc_init_system_aperture_regs(adev, xcc_mask);
+	gfxhub_v1_2_xcc_init_tlb_regs(adev, xcc_mask);
 	if (!amdgpu_sriov_vf(adev))
-		gfxhub_v1_2_init_cache_regs(adev);
+		gfxhub_v1_2_xcc_init_cache_regs(adev, xcc_mask);
 
-	gfxhub_v1_2_enable_system_domain(adev);
+	gfxhub_v1_2_xcc_enable_system_domain(adev, xcc_mask);
 	if (!amdgpu_sriov_vf(adev))
-		gfxhub_v1_2_disable_identity_aperture(adev);
-	gfxhub_v1_2_setup_vmid_config(adev);
-	gfxhub_v1_2_program_invalidation(adev);
+		gfxhub_v1_2_xcc_disable_identity_aperture(adev, xcc_mask);
+	gfxhub_v1_2_xcc_setup_vmid_config(adev, xcc_mask);
+	gfxhub_v1_2_xcc_program_invalidation(adev, xcc_mask);
 
 	return 0;
 }
 
+static int gfxhub_v1_2_gart_enable(struct amdgpu_device *adev)
+{
+	uint32_t xcc_mask;
+
+	xcc_mask = GENMASK(NUM_XCC(adev->gfx.xcc_mask) - 1, 0);
+	return gfxhub_v1_2_xcc_gart_enable(adev, xcc_mask);
+}
+
+static void gfxhub_v1_2_xcc_gart_disable(struct amdgpu_device *adev,
+					 uint32_t xcc_mask)
+{
+	struct amdgpu_vmhub *hub;
+	u32 tmp;
+	u32 i, j;
+
+	for_each_inst(j, xcc_mask) {
+		hub = &adev->vmhub[AMDGPU_GFXHUB(j)];
+		/* Disable all tables */
+		for (i = 0; i < 16; i++)
+			WREG32_SOC15_OFFSET(GC, GET_INST(GC, j), regVM_CONTEXT0_CNTL,
+					    i * hub->ctx_distance, 0);
+
+		/* Setup TLB control */
+		tmp = RREG32_SOC15(GC, GET_INST(GC, j), regMC_VM_MX_L1_TLB_CNTL);
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0);
+		tmp = REG_SET_FIELD(tmp,
+					MC_VM_MX_L1_TLB_CNTL,
+					ENABLE_ADVANCED_DRIVER_MODEL,
+					0);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, j), regMC_VM_MX_L1_TLB_CNTL, tmp);
+
+		/* Setup L2 cache */
+		if (!amdgpu_sriov_vf(adev)) {
+			tmp = RREG32_SOC15(GC, GET_INST(GC, j), regVM_L2_CNTL);
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 0);
+			WREG32_SOC15(GC, GET_INST(GC, j), regVM_L2_CNTL, tmp);
+			WREG32_SOC15(GC, GET_INST(GC, j), regVM_L2_CNTL3, 0);
+		}
+	}
+}
+
 static void gfxhub_v1_2_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	uint32_t xcc_mask;
+
+	xcc_mask = GENMASK(NUM_XCC(adev->gfx.xcc_mask) - 1, 0);
+	gfxhub_v1_2_xcc_gart_disable(adev, xcc_mask);
+}
+
+static void gfxhub_v1_2_xcc_set_fault_enable_default(struct amdgpu_device *adev,
+						     bool value,
+						     uint32_t xcc_mask)
+{
 	u32 tmp;
-	u32 i;
-
-	/* Disable all tables */
-	for (i = 0; i < 16; i++)
-		WREG32_SOC15_OFFSET(GC, 0, regVM_CONTEXT0_CNTL,
-				    i * hub->ctx_distance, 0);
-
-	/* Setup TLB control */
-	tmp = RREG32_SOC15(GC, 0, regMC_VM_MX_L1_TLB_CNTL);
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0);
-	tmp = REG_SET_FIELD(tmp,
-				MC_VM_MX_L1_TLB_CNTL,
-				ENABLE_ADVANCED_DRIVER_MODEL,
-				0);
-	WREG32_SOC15_RLC(GC, 0, regMC_VM_MX_L1_TLB_CNTL, tmp);
-
-	/* Setup L2 cache */
-	tmp = RREG32_SOC15(GC, 0, regVM_L2_CNTL);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 0);
-	WREG32_SOC15(GC, 0, regVM_L2_CNTL, tmp);
-	WREG32_SOC15(GC, 0, regVM_L2_CNTL3, 0);
+	int i;
+
+	for_each_inst(i, xcc_mask) {
+		tmp = RREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_CNTL);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp,
+				VM_L2_PROTECTION_FAULT_CNTL,
+				TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT,
+				value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				READ_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		if (!value) {
+			tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+					CRASH_ON_NO_RETRY_FAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+					CRASH_ON_RETRY_FAULT, 1);
+		}
+		WREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_CNTL, tmp);
+	}
 }
 
 /**
@@ -393,72 +535,100 @@ static void gfxhub_v1_2_gart_disable(struct amdgpu_device *adev)
 static void gfxhub_v1_2_set_fault_enable_default(struct amdgpu_device *adev,
 						 bool value)
 {
-	u32 tmp;
-	tmp = RREG32_SOC15(GC, 0, regVM_L2_PROTECTION_FAULT_CNTL);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp,
-			VM_L2_PROTECTION_FAULT_CNTL,
-			TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT,
-			value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			READ_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	if (!value) {
-		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-				CRASH_ON_NO_RETRY_FAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-				CRASH_ON_RETRY_FAULT, 1);
-	}
-	WREG32_SOC15(GC, 0, regVM_L2_PROTECTION_FAULT_CNTL, tmp);
+	uint32_t xcc_mask;
+
+	xcc_mask = GENMASK(NUM_XCC(adev->gfx.xcc_mask) - 1, 0);
+	gfxhub_v1_2_xcc_set_fault_enable_default(adev, value, xcc_mask);
 }
 
-static void gfxhub_v1_2_init(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_init(struct amdgpu_device *adev, uint32_t xcc_mask)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub;
+	int i;
+
+	for_each_inst(i, xcc_mask) {
+		hub = &adev->vmhub[AMDGPU_GFXHUB(i)];
 
-	hub->ctx0_ptb_addr_lo32 =
-		SOC15_REG_OFFSET(GC, 0,
+		hub->ctx0_ptb_addr_lo32 =
+			SOC15_REG_OFFSET(GC, GET_INST(GC, i),
 				regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32);
-	hub->ctx0_ptb_addr_hi32 =
-		SOC15_REG_OFFSET(GC, 0,
+		hub->ctx0_ptb_addr_hi32 =
+			SOC15_REG_OFFSET(GC, GET_INST(GC, i),
 				regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32);
-	hub->vm_inv_eng0_sem =
-		SOC15_REG_OFFSET(GC, 0, regVM_INVALIDATE_ENG0_SEM);
-	hub->vm_inv_eng0_req =
-		SOC15_REG_OFFSET(GC, 0, regVM_INVALIDATE_ENG0_REQ);
-	hub->vm_inv_eng0_ack =
-		SOC15_REG_OFFSET(GC, 0, regVM_INVALIDATE_ENG0_ACK);
-	hub->vm_context0_cntl =
-		SOC15_REG_OFFSET(GC, 0, regVM_CONTEXT0_CNTL);
-	hub->vm_l2_pro_fault_status =
-		SOC15_REG_OFFSET(GC, 0, regVM_L2_PROTECTION_FAULT_STATUS);
-	hub->vm_l2_pro_fault_cntl =
-		SOC15_REG_OFFSET(GC, 0, regVM_L2_PROTECTION_FAULT_CNTL);
-
-	hub->ctx_distance = regVM_CONTEXT1_CNTL - regVM_CONTEXT0_CNTL;
-	hub->ctx_addr_distance = regVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 -
-		regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32;
-	hub->eng_distance = regVM_INVALIDATE_ENG1_REQ - regVM_INVALIDATE_ENG0_REQ;
-	hub->eng_addr_distance = regVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 -
-		regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32;
+		hub->vm_inv_eng0_sem =
+			SOC15_REG_OFFSET(GC, GET_INST(GC, i), regVM_INVALIDATE_ENG0_SEM);
+		hub->vm_inv_eng0_req =
+			SOC15_REG_OFFSET(GC, GET_INST(GC, i), regVM_INVALIDATE_ENG0_REQ);
+		hub->vm_inv_eng0_ack =
+			SOC15_REG_OFFSET(GC, GET_INST(GC, i), regVM_INVALIDATE_ENG0_ACK);
+		hub->vm_context0_cntl =
+			SOC15_REG_OFFSET(GC, GET_INST(GC, i), regVM_CONTEXT0_CNTL);
+		hub->vm_l2_pro_fault_status =
+			SOC15_REG_OFFSET(GC, GET_INST(GC, i),
+				regVM_L2_PROTECTION_FAULT_STATUS);
+		hub->vm_l2_pro_fault_cntl =
+			SOC15_REG_OFFSET(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_CNTL);
+
+		hub->ctx_distance = regVM_CONTEXT1_CNTL -
+				regVM_CONTEXT0_CNTL;
+		hub->ctx_addr_distance =
+				regVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 -
+				regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32;
+		hub->eng_distance = regVM_INVALIDATE_ENG1_REQ -
+				regVM_INVALIDATE_ENG0_REQ;
+		hub->eng_addr_distance =
+				regVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 -
+				regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32;
+	}
 }
 
+static void gfxhub_v1_2_init(struct amdgpu_device *adev)
+{
+	uint32_t xcc_mask;
+
+	xcc_mask = GENMASK(NUM_XCC(adev->gfx.xcc_mask) - 1, 0);
+	gfxhub_v1_2_xcc_init(adev, xcc_mask);
+}
+
+static int gfxhub_v1_2_get_xgmi_info(struct amdgpu_device *adev)
+{
+	u32 max_num_physical_nodes;
+	u32 max_physical_node_id;
+	u32 xgmi_lfb_cntl;
+	u32 max_region;
+	u64 seg_size;
+
+	xgmi_lfb_cntl = RREG32_SOC15(GC, GET_INST(GC, 0), regMC_VM_XGMI_LFB_CNTL);
+	seg_size = REG_GET_FIELD(
+		RREG32_SOC15(GC, GET_INST(GC, 0), regMC_VM_XGMI_LFB_SIZE),
+		MC_VM_XGMI_LFB_SIZE, PF_LFB_SIZE) << 24;
+	max_region =
+		REG_GET_FIELD(xgmi_lfb_cntl, MC_VM_XGMI_LFB_CNTL, PF_MAX_REGION);
+
+
+
+	max_num_physical_nodes   = 8;
+	max_physical_node_id     = 7;
+
+	/* PF_MAX_REGION=0 means xgmi is disabled */
+	if (max_region || adev->gmc.xgmi.connected_to_cpu) {
+		adev->gmc.xgmi.num_physical_nodes = max_region + 1;
+
+		if (adev->gmc.xgmi.num_physical_nodes > max_num_physical_nodes)
+			return -EINVAL;
+
+		adev->gmc.xgmi.physical_node_id =
+			REG_GET_FIELD(xgmi_lfb_cntl, MC_VM_XGMI_LFB_CNTL,
+					PF_LFB_REGION);
+
+		if (adev->gmc.xgmi.physical_node_id > max_physical_node_id)
+			return -EINVAL;
+
+		adev->gmc.xgmi.node_segment_size = seg_size;
+	}
+
+	return 0;
+}
 
 const struct amdgpu_gfxhub_funcs gfxhub_v1_2_funcs = {
 	.get_mc_fb_offset = gfxhub_v1_2_get_mc_fb_offset,
@@ -467,5 +637,38 @@ const struct amdgpu_gfxhub_funcs gfxhub_v1_2_funcs = {
 	.gart_disable = gfxhub_v1_2_gart_disable,
 	.set_fault_enable_default = gfxhub_v1_2_set_fault_enable_default,
 	.init = gfxhub_v1_2_init,
-	.get_xgmi_info = gfxhub_v1_1_get_xgmi_info,
+	.get_xgmi_info = gfxhub_v1_2_get_xgmi_info,
+};
+
+static int gfxhub_v1_2_xcp_resume(void *handle, uint32_t inst_mask)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	bool value;
+
+	if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS)
+		value = false;
+	else
+		value = true;
+
+	gfxhub_v1_2_xcc_set_fault_enable_default(adev, value, inst_mask);
+
+	if (!amdgpu_sriov_vf(adev))
+		return gfxhub_v1_2_xcc_gart_enable(adev, inst_mask);
+
+	return 0;
+}
+
+static int gfxhub_v1_2_xcp_suspend(void *handle, uint32_t inst_mask)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+	if (!amdgpu_sriov_vf(adev))
+		gfxhub_v1_2_xcc_gart_disable(adev, inst_mask);
+
+	return 0;
+}
+
+struct amdgpu_xcp_ip_funcs gfxhub_v1_2_xcp_funcs = {
+	.suspend = &gfxhub_v1_2_xcp_suspend,
+	.resume = &gfxhub_v1_2_xcp_resume
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.h b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.h
index e2d508f5a7ee..997e9f90c990 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.h
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.h
@@ -26,4 +26,6 @@
 
 extern const struct amdgpu_gfxhub_funcs gfxhub_v1_2_funcs;
 
+extern struct amdgpu_xcp_ip_funcs gfxhub_v1_2_xcp_funcs;
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c
index 9b3a02527318..793faf62cb07 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c
@@ -31,7 +31,7 @@
 
 #include "soc15_common.h"
 
-static const char *gfxhub_client_ids[] = {
+static const char * const gfxhub_client_ids[] = {
 	"CB/DB",
 	"Reserved",
 	"GE1",
@@ -120,7 +120,7 @@ static u64 gfxhub_v2_0_get_mc_fb_offset(struct amdgpu_device *adev)
 static void gfxhub_v2_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
 				uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
 			    hub->ctx_addr_distance * vmid,
@@ -282,12 +282,12 @@ static void gfxhub_v2_0_disable_identity_aperture(struct amdgpu_device *adev)
 
 static void gfxhub_v2_0_setup_vmid_config(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	int i;
 	uint32_t tmp;
 
 	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT1_CNTL, i);
+		tmp = RREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT1_CNTL, i * hub->ctx_distance);
 		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
 		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
 				    adev->vm_manager.num_level);
@@ -331,8 +331,8 @@ static void gfxhub_v2_0_setup_vmid_config(struct amdgpu_device *adev)
 
 static void gfxhub_v2_0_program_invalidation(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	unsigned i;
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	unsigned int i;
 
 	for (i = 0 ; i < 18; ++i) {
 		WREG32_SOC15_OFFSET(GC, 0, mmGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
@@ -360,7 +360,7 @@ static int gfxhub_v2_0_gart_enable(struct amdgpu_device *adev)
 
 static void gfxhub_v2_0_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	u32 tmp;
 	u32 i;
 
@@ -393,6 +393,7 @@ static void gfxhub_v2_0_set_fault_enable_default(struct amdgpu_device *adev,
 					  bool value)
 {
 	u32 tmp;
+
 	tmp = RREG32_SOC15(GC, 0, mmGCVM_L2_PROTECTION_FAULT_CNTL);
 	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
 			    RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
@@ -433,7 +434,7 @@ static const struct amdgpu_vmhub_funcs gfxhub_v2_0_vmhub_funcs = {
 
 static void gfxhub_v2_0_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	hub->ctx0_ptb_addr_lo32 =
 		SOC15_REG_OFFSET(GC, 0,
@@ -470,6 +471,9 @@ static void gfxhub_v2_0_init(struct amdgpu_device *adev)
 		GCVM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
 		GCVM_CONTEXT1_CNTL__EXECUTE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK;
 
+	/* TODO: This is only needed on some Navi 1x revisions */
+	hub->sdma_invalidation_workaround = true;
+
 	hub->vmhub_funcs = &gfxhub_v2_0_vmhub_funcs;
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c
index 4aacbbec31e2..deb95fab02df 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c
@@ -34,7 +34,7 @@
 #define mmGCUTCL2_HARVEST_BYPASS_GROUPS_YELLOW_CARP				0x16f8
 #define mmGCUTCL2_HARVEST_BYPASS_GROUPS_YELLOW_CARP_BASE_IDX	0
 
-static const char *gfxhub_client_ids[] = {
+static const char * const gfxhub_client_ids[] = {
 	"CB/DB",
 	"Reserved",
 	"GE1",
@@ -123,7 +123,7 @@ static u64 gfxhub_v2_1_get_mc_fb_offset(struct amdgpu_device *adev)
 static void gfxhub_v2_1_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
 				uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
 			    hub->ctx_addr_distance * vmid,
@@ -155,6 +155,9 @@ static void gfxhub_v2_1_init_system_aperture_regs(struct amdgpu_device *adev)
 {
 	uint64_t value;
 
+	if (amdgpu_sriov_vf(adev))
+		return;
+
 	/* Program the AGP BAR */
 	WREG32_SOC15(GC, 0, mmGCMC_VM_AGP_BASE, 0);
 	WREG32_SOC15(GC, 0, mmGCMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
@@ -291,12 +294,12 @@ static void gfxhub_v2_1_disable_identity_aperture(struct amdgpu_device *adev)
 
 static void gfxhub_v2_1_setup_vmid_config(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	int i;
 	uint32_t tmp;
 
 	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT1_CNTL, i);
+		tmp = RREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT1_CNTL, i * hub->ctx_distance);
 		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
 		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
 				    adev->vm_manager.num_level);
@@ -340,8 +343,8 @@ static void gfxhub_v2_1_setup_vmid_config(struct amdgpu_device *adev)
 
 static void gfxhub_v2_1_program_invalidation(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	unsigned i;
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	unsigned int i;
 
 	for (i = 0 ; i < 18; ++i) {
 		WREG32_SOC15_OFFSET(GC, 0, mmGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
@@ -381,7 +384,7 @@ static int gfxhub_v2_1_gart_enable(struct amdgpu_device *adev)
 
 static void gfxhub_v2_1_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	u32 tmp;
 	u32 i;
 
@@ -462,7 +465,7 @@ static const struct amdgpu_vmhub_funcs gfxhub_v2_1_vmhub_funcs = {
 
 static void gfxhub_v2_1_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	hub->ctx0_ptb_addr_lo32 =
 		SOC15_REG_OFFSET(GC, 0,
@@ -502,42 +505,6 @@ static void gfxhub_v2_1_init(struct amdgpu_device *adev)
 	hub->vmhub_funcs = &gfxhub_v2_1_vmhub_funcs;
 }
 
-static int gfxhub_v2_1_get_xgmi_info(struct amdgpu_device *adev)
-{
-	u32 xgmi_lfb_cntl = RREG32_SOC15(GC, 0, mmGCMC_VM_XGMI_LFB_CNTL);
-	u32 max_region =
-		REG_GET_FIELD(xgmi_lfb_cntl, GCMC_VM_XGMI_LFB_CNTL, PF_MAX_REGION);
-	u32 max_num_physical_nodes   = 0;
-	u32 max_physical_node_id     = 0;
-
-	switch (adev->ip_versions[XGMI_HWIP][0]) {
-	case IP_VERSION(4, 8, 0):
-		max_num_physical_nodes   = 4;
-		max_physical_node_id     = 3;
-		break;
-	default:
-		return -EINVAL;
-	}
-
-	/* PF_MAX_REGION=0 means xgmi is disabled */
-	if (max_region) {
-		adev->gmc.xgmi.num_physical_nodes = max_region + 1;
-		if (adev->gmc.xgmi.num_physical_nodes > max_num_physical_nodes)
-			return -EINVAL;
-
-		adev->gmc.xgmi.physical_node_id =
-			REG_GET_FIELD(xgmi_lfb_cntl, GCMC_VM_XGMI_LFB_CNTL, PF_LFB_REGION);
-		if (adev->gmc.xgmi.physical_node_id > max_physical_node_id)
-			return -EINVAL;
-
-		adev->gmc.xgmi.node_segment_size = REG_GET_FIELD(
-			RREG32_SOC15(GC, 0, mmGCMC_VM_XGMI_LFB_SIZE),
-			GCMC_VM_XGMI_LFB_SIZE, PF_LFB_SIZE) << 24;
-	}
-
-	return 0;
-}
-
 static void gfxhub_v2_1_utcl2_harvest(struct amdgpu_device *adev)
 {
 	int i;
@@ -548,7 +515,7 @@ static void gfxhub_v2_1_utcl2_harvest(struct amdgpu_device *adev)
 		adev->gfx.config.max_sh_per_se *
 		adev->gfx.config.max_shader_engines);
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 1):
 	case IP_VERSION(10, 3, 3):
 		/* Get SA disabled bitmap from eFuse setting */
@@ -582,6 +549,7 @@ static void gfxhub_v2_1_utcl2_harvest(struct amdgpu_device *adev)
 static void gfxhub_v2_1_save_regs(struct amdgpu_device *adev)
 {
 	int i;
+
 	adev->gmc.VM_L2_CNTL = RREG32_SOC15(GC, 0, mmGCVM_L2_CNTL);
 	adev->gmc.VM_L2_CNTL2 = RREG32_SOC15(GC, 0, mmGCVM_L2_CNTL2);
 	adev->gmc.VM_DUMMY_PAGE_FAULT_CNTL = RREG32_SOC15(GC, 0, mmGCVM_DUMMY_PAGE_FAULT_CNTL);
@@ -616,6 +584,7 @@ static void gfxhub_v2_1_save_regs(struct amdgpu_device *adev)
 static void gfxhub_v2_1_restore_regs(struct amdgpu_device *adev)
 {
 	int i;
+
 	WREG32_SOC15(GC, 0, mmGCVM_L2_CNTL, adev->gmc.VM_L2_CNTL);
 	WREG32_SOC15(GC, 0, mmGCVM_L2_CNTL2, adev->gmc.VM_L2_CNTL2);
 	WREG32_SOC15(GC, 0, mmGCVM_DUMMY_PAGE_FAULT_CNTL, adev->gmc.VM_DUMMY_PAGE_FAULT_CNTL);
@@ -651,7 +620,7 @@ static void gfxhub_v2_1_restore_regs(struct amdgpu_device *adev)
 
 static void gfxhub_v2_1_halt(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	int i;
 	uint32_t tmp;
 	int time = 1000;
@@ -679,9 +648,8 @@ static void gfxhub_v2_1_halt(struct amdgpu_device *adev)
 		tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS2);
 	}
 
-	if (!time) {
+	if (!time)
 		DRM_WARN("failed to wait for GRBM(EA) idle\n");
-	}
 }
 
 const struct amdgpu_gfxhub_funcs gfxhub_v2_1_funcs = {
@@ -692,7 +660,6 @@ const struct amdgpu_gfxhub_funcs gfxhub_v2_1_funcs = {
 	.gart_disable = gfxhub_v2_1_gart_disable,
 	.set_fault_enable_default = gfxhub_v2_1_set_fault_enable_default,
 	.init = gfxhub_v2_1_init,
-	.get_xgmi_info = gfxhub_v2_1_get_xgmi_info,
 	.utcl2_harvest = gfxhub_v2_1_utcl2_harvest,
 	.mode2_save_regs = gfxhub_v2_1_save_regs,
 	.mode2_restore_regs = gfxhub_v2_1_restore_regs,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c
index 13712640fa46..abe30c8bd2ba 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c
@@ -30,7 +30,7 @@
 #include "navi10_enum.h"
 #include "soc15_common.h"
 
-static const char *gfxhub_client_ids[] = {
+static const char * const gfxhub_client_ids[] = {
 	"CB/DB",
 	"Reserved",
 	"GE1",
@@ -119,7 +119,7 @@ static u64 gfxhub_v3_0_get_mc_fb_offset(struct amdgpu_device *adev)
 static void gfxhub_v3_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
 				uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
 			    hub->ctx_addr_distance * vmid,
@@ -164,8 +164,7 @@ static void gfxhub_v3_0_init_system_aperture_regs(struct amdgpu_device *adev)
 		     max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
 
 	/* Set default page address. */
-	value = adev->mem_scratch.gpu_addr - adev->gmc.vram_start
-		+ adev->vm_manager.vram_base_offset;
+	value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
 	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
 		     (u32)(value >> 12));
 	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
@@ -290,12 +289,12 @@ static void gfxhub_v3_0_disable_identity_aperture(struct amdgpu_device *adev)
 
 static void gfxhub_v3_0_setup_vmid_config(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	int i;
 	uint32_t tmp;
 
 	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL, i);
+		tmp = RREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL, i * hub->ctx_distance);
 		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
 		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
 				    adev->vm_manager.num_level);
@@ -339,8 +338,8 @@ static void gfxhub_v3_0_setup_vmid_config(struct amdgpu_device *adev)
 
 static void gfxhub_v3_0_program_invalidation(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	unsigned i;
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	unsigned int i;
 
 	for (i = 0 ; i < 18; ++i) {
 		WREG32_SOC15_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
@@ -380,7 +379,7 @@ static int gfxhub_v3_0_gart_enable(struct amdgpu_device *adev)
 
 static void gfxhub_v3_0_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	u32 tmp;
 	u32 i;
 
@@ -463,7 +462,7 @@ static const struct amdgpu_vmhub_funcs gfxhub_v3_0_vmhub_funcs = {
 
 static void gfxhub_v3_0_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	hub->ctx0_ptb_addr_lo32 =
 		SOC15_REG_OFFSET(GC, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c
index 6e0bd628c889..b3ef6e71811f 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c
@@ -33,7 +33,7 @@
 #define regGCVM_L2_CNTL4_DEFAULT		0x000000c1
 #define regGCVM_L2_CNTL5_DEFAULT		0x00003fe0
 
-static const char *gfxhub_client_ids[] = {
+static const char * const gfxhub_client_ids[] = {
 	"CB/DB",
 	"Reserved",
 	"GE1",
@@ -122,7 +122,7 @@ static u64 gfxhub_v3_0_3_get_mc_fb_offset(struct amdgpu_device *adev)
 static void gfxhub_v3_0_3_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
 				uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
 			    hub->ctx_addr_distance * vmid,
@@ -169,8 +169,7 @@ static void gfxhub_v3_0_3_init_system_aperture_regs(struct amdgpu_device *adev)
 		     max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
 
 	/* Set default page address. */
-	value = adev->mem_scratch.gpu_addr - adev->gmc.vram_start
-		+ adev->vm_manager.vram_base_offset;
+	value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
 	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
 		     (u32)(value >> 12));
 	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
@@ -295,12 +294,12 @@ static void gfxhub_v3_0_3_disable_identity_aperture(struct amdgpu_device *adev)
 
 static void gfxhub_v3_0_3_setup_vmid_config(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	int i;
 	uint32_t tmp;
 
 	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL, i);
+		tmp = RREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL, i * hub->ctx_distance);
 		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
 		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
 				    adev->vm_manager.num_level);
@@ -344,8 +343,8 @@ static void gfxhub_v3_0_3_setup_vmid_config(struct amdgpu_device *adev)
 
 static void gfxhub_v3_0_3_program_invalidation(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	unsigned i;
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	unsigned int i;
 
 	for (i = 0 ; i < 18; ++i) {
 		WREG32_SOC15_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
@@ -373,7 +372,7 @@ static int gfxhub_v3_0_3_gart_enable(struct amdgpu_device *adev)
 
 static void gfxhub_v3_0_3_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	u32 tmp;
 	u32 i;
 
@@ -451,7 +450,7 @@ static const struct amdgpu_vmhub_funcs gfxhub_v3_0_3_vmhub_funcs = {
 
 static void gfxhub_v3_0_3_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	hub->ctx0_ptb_addr_lo32 =
 		SOC15_REG_OFFSET(GC, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
index 7d6f4a68f416..ce6e04242c52 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
@@ -51,18 +51,9 @@
 #include "athub_v2_0.h"
 #include "athub_v2_1.h"
 
-#include "amdgpu_reset.h"
-
-#if 0
-static const struct soc15_reg_golden golden_settings_navi10_hdp[] =
-{
-	/* TODO add golden setting for hdp */
-};
-#endif
-
 static int gmc_v10_0_ecc_interrupt_state(struct amdgpu_device *adev,
 					 struct amdgpu_irq_src *src,
-					 unsigned type,
+					 unsigned int type,
 					 enum amdgpu_interrupt_state state)
 {
 	return 0;
@@ -70,13 +61,13 @@ static int gmc_v10_0_ecc_interrupt_state(struct amdgpu_device *adev,
 
 static int
 gmc_v10_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
-				   struct amdgpu_irq_src *src, unsigned type,
+				   struct amdgpu_irq_src *src, unsigned int type,
 				   enum amdgpu_interrupt_state state)
 {
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
 		/* MM HUB */
-		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB_0, false);
+		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), false);
 		/* GFX HUB */
 		/* This works because this interrupt is only
 		 * enabled at init/resume and disabled in
@@ -84,11 +75,11 @@ gmc_v10_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 		 * change over the course of suspend/resume.
 		 */
 		if (!adev->in_s0ix)
-			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB_0, false);
+			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), false);
 		break;
 	case AMDGPU_IRQ_STATE_ENABLE:
 		/* MM HUB */
-		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB_0, true);
+		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), true);
 		/* GFX HUB */
 		/* This works because this interrupt is only
 		 * enabled at init/resume and disabled in
@@ -96,7 +87,7 @@ gmc_v10_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 		 * change over the course of suspend/resume.
 		 */
 		if (!adev->in_s0ix)
-			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB_0, true);
+			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), true);
 		break;
 	default:
 		break;
@@ -109,10 +100,14 @@ static int gmc_v10_0_process_interrupt(struct amdgpu_device *adev,
 				       struct amdgpu_irq_src *source,
 				       struct amdgpu_iv_entry *entry)
 {
-	bool retry_fault = !!(entry->src_data[1] & 0x80);
-	bool write_fault = !!(entry->src_data[1] & 0x20);
-	struct amdgpu_vmhub *hub = &adev->vmhub[entry->vmid_src];
-	struct amdgpu_task_info task_info;
+	uint32_t vmhub_index = entry->client_id == SOC15_IH_CLIENTID_VMC ?
+			       AMDGPU_MMHUB0(0) : AMDGPU_GFXHUB(0);
+	struct amdgpu_vmhub *hub = &adev->vmhub[vmhub_index];
+	bool retry_fault = !!(entry->src_data[1] &
+			      AMDGPU_GMC9_FAULT_SOURCE_DATA_RETRY);
+	bool write_fault = !!(entry->src_data[1] &
+			      AMDGPU_GMC9_FAULT_SOURCE_DATA_WRITE);
+	struct amdgpu_task_info *task_info;
 	uint32_t status = 0;
 	u64 addr;
 
@@ -139,7 +134,8 @@ static int gmc_v10_0_process_interrupt(struct amdgpu_device *adev,
 		/* Try to handle the recoverable page faults by filling page
 		 * tables
 		 */
-		if (amdgpu_vm_handle_fault(adev, entry->pasid, addr, write_fault))
+		if (amdgpu_vm_handle_fault(adev, entry->pasid, 0, 0, addr,
+					   entry->timestamp, write_fault))
 			return 1;
 	}
 
@@ -149,32 +145,39 @@ static int gmc_v10_0_process_interrupt(struct amdgpu_device *adev,
 		 * be updated to avoid reading an incorrect value due to
 		 * the new fast GRBM interface.
 		 */
-		if ((entry->vmid_src == AMDGPU_GFXHUB_0) &&
-		    (adev->ip_versions[GC_HWIP][0] < IP_VERSION(10, 3, 0)))
+		if ((entry->vmid_src == AMDGPU_GFXHUB(0)) &&
+		    (amdgpu_ip_version(adev, GC_HWIP, 0) <
+		     IP_VERSION(10, 3, 0)))
 			RREG32(hub->vm_l2_pro_fault_status);
 
 		status = RREG32(hub->vm_l2_pro_fault_status);
 		WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1);
+
+		amdgpu_vm_update_fault_cache(adev, entry->pasid, addr, status,
+					     entry->vmid_src ? AMDGPU_MMHUB0(0) : AMDGPU_GFXHUB(0));
 	}
 
 	if (!printk_ratelimit())
 		return 0;
 
-	memset(&task_info, 0, sizeof(struct amdgpu_task_info));
-	amdgpu_vm_get_task_info(adev, entry->pasid, &task_info);
-
 	dev_err(adev->dev,
-		"[%s] page fault (src_id:%u ring:%u vmid:%u pasid:%u, "
-		"for process %s pid %d thread %s pid %d)\n",
+		"[%s] page fault (src_id:%u ring:%u vmid:%u pasid:%u)\n",
 		entry->vmid_src ? "mmhub" : "gfxhub",
-		entry->src_id, entry->ring_id, entry->vmid,
-		entry->pasid, task_info.process_name, task_info.tgid,
-		task_info.task_name, task_info.pid);
+		entry->src_id, entry->ring_id, entry->vmid, entry->pasid);
+	task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid);
+	if (task_info) {
+		amdgpu_vm_print_task_info(adev, task_info);
+		amdgpu_vm_put_task_info(task_info);
+	}
+
 	dev_err(adev->dev, "  in page starting at address 0x%016llx from client 0x%x (%s)\n",
-		addr, entry->client_id,
-		soc15_ih_clientid_name[entry->client_id]);
+			addr, entry->client_id,
+			soc15_ih_clientid_name[entry->client_id]);
 
-	if (!amdgpu_sriov_vf(adev))
+	/* Only print L2 fault status if the status register could be read and
+	 * contains useful information
+	 */
+	if (status != 0)
 		hub->vmhub_funcs->print_l2_protection_fault_status(adev,
 								   status);
 
@@ -212,8 +215,7 @@ static void gmc_v10_0_set_irq_funcs(struct amdgpu_device *adev)
 static bool gmc_v10_0_use_invalidate_semaphore(struct amdgpu_device *adev,
 				       uint32_t vmhub)
 {
-	return ((vmhub == AMDGPU_MMHUB_0 ||
-		 vmhub == AMDGPU_MMHUB_1) &&
+	return ((vmhub == AMDGPU_MMHUB0(0)) &&
 		(!amdgpu_sriov_vf(adev)));
 }
 
@@ -237,20 +239,48 @@ static bool gmc_v10_0_get_atc_vmid_pasid_mapping_info(
  * by the amdgpu vm/hsa code.
  */
 
-static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
-				   unsigned int vmhub, uint32_t flush_type)
+/**
+ * gmc_v10_0_flush_gpu_tlb - gart tlb flush callback
+ *
+ * @adev: amdgpu_device pointer
+ * @vmid: vm instance to flush
+ * @vmhub: vmhub type
+ * @flush_type: the flush type
+ *
+ * Flush the TLB for the requested page table.
+ */
+static void gmc_v10_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
+					uint32_t vmhub, uint32_t flush_type)
 {
 	bool use_semaphore = gmc_v10_0_use_invalidate_semaphore(adev, vmhub);
 	struct amdgpu_vmhub *hub = &adev->vmhub[vmhub];
 	u32 inv_req = hub->vmhub_funcs->get_invalidate_req(vmid, flush_type);
-	u32 tmp;
 	/* Use register 17 for GART */
-	const unsigned eng = 17;
-	unsigned int i;
+	const unsigned int eng = 17;
 	unsigned char hub_ip = 0;
+	u32 sem, req, ack;
+	unsigned int i;
+	u32 tmp;
+
+	sem = hub->vm_inv_eng0_sem + hub->eng_distance * eng;
+	req = hub->vm_inv_eng0_req + hub->eng_distance * eng;
+	ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng;
+
+	/* flush hdp cache */
+	amdgpu_device_flush_hdp(adev, NULL);
+
+	/* This is necessary for SRIOV as well as for GFXOFF to function
+	 * properly under bare metal
+	 */
+	if (adev->gfx.kiq[0].ring.sched.ready && !adev->enable_mes &&
+	    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) {
+		amdgpu_gmc_fw_reg_write_reg_wait(adev, req, ack, inv_req,
+						 1 << vmid, GET_INST(GC, 0));
+		return;
+	}
 
-	hub_ip = (vmhub == AMDGPU_GFXHUB_0) ?
-		   GC_HWIP : MMHUB_HWIP;
+	/* This path is needed before KIQ/MES/GFXOFF are set up */
+	hub_ip = (vmhub == AMDGPU_GFXHUB(0)) ? GC_HWIP : MMHUB_HWIP;
 
 	spin_lock(&adev->gmc.invalidate_lock);
 	/*
@@ -264,9 +294,7 @@ static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
 	if (use_semaphore) {
 		for (i = 0; i < adev->usec_timeout; i++) {
 			/* a read return value of 1 means semaphore acuqire */
-			tmp = RREG32_RLC_NO_KIQ(hub->vm_inv_eng0_sem +
-					 hub->eng_distance * eng, hub_ip);
-
+			tmp = RREG32_RLC_NO_KIQ(sem, hub_ip);
 			if (tmp & 0x1)
 				break;
 			udelay(1);
@@ -276,24 +304,19 @@ static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
 			DRM_ERROR("Timeout waiting for sem acquire in VM flush!\n");
 	}
 
-	WREG32_RLC_NO_KIQ(hub->vm_inv_eng0_req +
-			  hub->eng_distance * eng,
-			  inv_req, hub_ip);
+	WREG32_RLC_NO_KIQ(req, inv_req, hub_ip);
 
 	/*
 	 * Issue a dummy read to wait for the ACK register to be cleared
 	 * to avoid a false ACK due to the new fast GRBM interface.
 	 */
-	if ((vmhub == AMDGPU_GFXHUB_0) &&
-	    (adev->ip_versions[GC_HWIP][0] < IP_VERSION(10, 3, 0)))
-		RREG32_RLC_NO_KIQ(hub->vm_inv_eng0_req +
-				  hub->eng_distance * eng, hub_ip);
+	if ((vmhub == AMDGPU_GFXHUB(0)) &&
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(10, 3, 0)))
+		RREG32_RLC_NO_KIQ(req, hub_ip);
 
 	/* Wait for ACK with a delay.*/
 	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32_RLC_NO_KIQ(hub->vm_inv_eng0_ack +
-				  hub->eng_distance * eng, hub_ip);
-
+		tmp = RREG32_RLC_NO_KIQ(ack, hub_ip);
 		tmp &= 1 << vmid;
 		if (tmp)
 			break;
@@ -303,109 +326,13 @@ static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
 
 	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
 	if (use_semaphore)
-		/*
-		 * add semaphore release after invalidation,
-		 * write with 0 means semaphore release
-		 */
-		WREG32_RLC_NO_KIQ(hub->vm_inv_eng0_sem +
-				  hub->eng_distance * eng, 0, hub_ip);
+		WREG32_RLC_NO_KIQ(sem, 0, hub_ip);
 
 	spin_unlock(&adev->gmc.invalidate_lock);
 
-	if (i < adev->usec_timeout)
-		return;
-
-	DRM_ERROR("Timeout waiting for VM flush hub: %d!\n", vmhub);
-}
-
-/**
- * gmc_v10_0_flush_gpu_tlb - gart tlb flush callback
- *
- * @adev: amdgpu_device pointer
- * @vmid: vm instance to flush
- * @vmhub: vmhub type
- * @flush_type: the flush type
- *
- * Flush the TLB for the requested page table.
- */
-static void gmc_v10_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
-					uint32_t vmhub, uint32_t flush_type)
-{
-	struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
-	struct dma_fence *fence;
-	struct amdgpu_job *job;
-
-	int r;
-
-	/* flush hdp cache */
-	adev->hdp.funcs->flush_hdp(adev, NULL);
-
-	/* For SRIOV run time, driver shouldn't access the register through MMIO
-	 * Directly use kiq to do the vm invalidation instead
-	 */
-	if (adev->gfx.kiq.ring.sched.ready && !adev->enable_mes &&
-	    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev)) &&
-	    down_read_trylock(&adev->reset_domain->sem)) {
-		struct amdgpu_vmhub *hub = &adev->vmhub[vmhub];
-		const unsigned eng = 17;
-		u32 inv_req = hub->vmhub_funcs->get_invalidate_req(vmid, flush_type);
-		u32 req = hub->vm_inv_eng0_req + hub->eng_distance * eng;
-		u32 ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng;
-
-		amdgpu_virt_kiq_reg_write_reg_wait(adev, req, ack, inv_req,
-				1 << vmid);
-
-		up_read(&adev->reset_domain->sem);
-		return;
-	}
-
-	mutex_lock(&adev->mman.gtt_window_lock);
-
-	if (vmhub == AMDGPU_MMHUB_0) {
-		gmc_v10_0_flush_vm_hub(adev, vmid, AMDGPU_MMHUB_0, 0);
-		mutex_unlock(&adev->mman.gtt_window_lock);
-		return;
-	}
-
-	BUG_ON(vmhub != AMDGPU_GFXHUB_0);
-
-	if (!adev->mman.buffer_funcs_enabled ||
-	    !adev->ib_pool_ready ||
-	    amdgpu_in_reset(adev) ||
-	    ring->sched.ready == false) {
-		gmc_v10_0_flush_vm_hub(adev, vmid, AMDGPU_GFXHUB_0, 0);
-		mutex_unlock(&adev->mman.gtt_window_lock);
-		return;
-	}
-
-	/* The SDMA on Navi has a bug which can theoretically result in memory
-	 * corruption if an invalidation happens at the same time as an VA
-	 * translation. Avoid this by doing the invalidation from the SDMA
-	 * itself.
-	 */
-	r = amdgpu_job_alloc_with_ib(ring->adev, &adev->mman.entity,
-				     AMDGPU_FENCE_OWNER_UNDEFINED,
-				     16 * 4, AMDGPU_IB_POOL_IMMEDIATE,
-				     &job);
-	if (r)
-		goto error_alloc;
-
-	job->vm_pd_addr = amdgpu_gmc_pd_addr(adev->gart.bo);
-	job->vm_needs_flush = true;
-	job->ibs->ptr[job->ibs->length_dw++] = ring->funcs->nop;
-	amdgpu_ring_pad_ib(ring, &job->ibs[0]);
-	fence = amdgpu_job_submit(job);
-
-	mutex_unlock(&adev->mman.gtt_window_lock);
-
-	dma_fence_wait(fence, false);
-	dma_fence_put(fence);
-
-	return;
-
-error_alloc:
-	mutex_unlock(&adev->mman.gtt_window_lock);
-	DRM_ERROR("Error flushing GPU TLB using the SDMA (%d)!\n", r);
+	if (i >= adev->usec_timeout)
+		dev_err(adev->dev, "Timeout waiting for VM flush hub: %d!\n",
+			vmhub);
 }
 
 /**
@@ -415,74 +342,44 @@ error_alloc:
  * @pasid: pasid to be flush
  * @flush_type: the flush type
  * @all_hub: Used with PACKET3_INVALIDATE_TLBS_ALL_HUB()
+ * @inst: is used to select which instance of KIQ to use for the invalidation
  *
  * Flush the TLB for the requested pasid.
  */
-static int gmc_v10_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
-					uint16_t pasid, uint32_t flush_type,
-					bool all_hub)
+static void gmc_v10_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
+					  uint16_t pasid, uint32_t flush_type,
+					  bool all_hub, uint32_t inst)
 {
+	uint16_t queried;
 	int vmid, i;
-	signed long r;
-	uint32_t seq;
-	uint16_t queried_pasid;
-	bool ret;
-	u32 usec_timeout = amdgpu_sriov_vf(adev) ? SRIOV_USEC_TIMEOUT : adev->usec_timeout;
-	struct amdgpu_ring *ring = &adev->gfx.kiq.ring;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-
-	if (amdgpu_emu_mode == 0 && ring->sched.ready) {
-		spin_lock(&adev->gfx.kiq.ring_lock);
-		/* 2 dwords flush + 8 dwords fence */
-		amdgpu_ring_alloc(ring, kiq->pmf->invalidate_tlbs_size + 8);
-		kiq->pmf->kiq_invalidate_tlbs(ring,
-					pasid, flush_type, all_hub);
-		r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
-		if (r) {
-			amdgpu_ring_undo(ring);
-			spin_unlock(&adev->gfx.kiq.ring_lock);
-			return -ETIME;
-		}
-
-		amdgpu_ring_commit(ring);
-		spin_unlock(&adev->gfx.kiq.ring_lock);
-		r = amdgpu_fence_wait_polling(ring, seq, usec_timeout);
-		if (r < 1) {
-			dev_err(adev->dev, "wait for kiq fence error: %ld.\n", r);
-			return -ETIME;
-		}
-
-		return 0;
-	}
 
 	for (vmid = 1; vmid < AMDGPU_NUM_VMID; vmid++) {
-
-		ret = gmc_v10_0_get_atc_vmid_pasid_mapping_info(adev, vmid,
-				&queried_pasid);
-		if (ret	&& queried_pasid == pasid) {
-			if (all_hub) {
-				for (i = 0; i < adev->num_vmhubs; i++)
-					gmc_v10_0_flush_gpu_tlb(adev, vmid,
-							i, flush_type);
-			} else {
-				gmc_v10_0_flush_gpu_tlb(adev, vmid,
-						AMDGPU_GFXHUB_0, flush_type);
-			}
-			if (!adev->enable_mes)
-				break;
+		bool valid;
+
+		valid = gmc_v10_0_get_atc_vmid_pasid_mapping_info(adev, vmid,
+								  &queried);
+		if (!valid || queried != pasid)
+			continue;
+
+		if (all_hub) {
+			for_each_set_bit(i, adev->vmhubs_mask,
+					 AMDGPU_MAX_VMHUBS)
+				gmc_v10_0_flush_gpu_tlb(adev, vmid, i,
+							flush_type);
+		} else {
+			gmc_v10_0_flush_gpu_tlb(adev, vmid, AMDGPU_GFXHUB(0),
+						flush_type);
 		}
 	}
-
-	return 0;
 }
 
 static uint64_t gmc_v10_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
-					     unsigned vmid, uint64_t pd_addr)
+					     unsigned int vmid, uint64_t pd_addr)
 {
 	bool use_semaphore = gmc_v10_0_use_invalidate_semaphore(ring->adev, ring->vm_hub);
 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
 	uint32_t req = hub->vmhub_funcs->get_invalidate_req(vmid, 0);
-	unsigned eng = ring->vm_inv_eng;
+	unsigned int eng = ring->vm_inv_eng;
 
 	/*
 	 * It may lose gpuvm invalidate acknowldege state across power-gating
@@ -524,17 +421,13 @@ static uint64_t gmc_v10_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
 	return pd_addr;
 }
 
-static void gmc_v10_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
-					 unsigned pasid)
+static void gmc_v10_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned int vmid,
+					 unsigned int pasid)
 {
 	struct amdgpu_device *adev = ring->adev;
 	uint32_t reg;
 
-	/* MES fw manages IH_VMID_x_LUT updating */
-	if (ring->is_mes_queue)
-		return;
-
-	if (ring->vm_hub == AMDGPU_GFXHUB_0)
+	if (ring->vm_hub == AMDGPU_GFXHUB(0))
 		reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT) + vmid;
 	else
 		reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT_MM) + vmid;
@@ -575,24 +468,6 @@ static void gmc_v10_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid
  * 0 valid
  */
 
-static uint64_t gmc_v10_0_map_mtype(struct amdgpu_device *adev, uint32_t flags)
-{
-	switch (flags) {
-	case AMDGPU_VM_MTYPE_DEFAULT:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_NC);
-	case AMDGPU_VM_MTYPE_NC:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_NC);
-	case AMDGPU_VM_MTYPE_WC:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_WC);
-	case AMDGPU_VM_MTYPE_CC:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_CC);
-	case AMDGPU_VM_MTYPE_UC:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_UC);
-	default:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_NC);
-	}
-}
-
 static void gmc_v10_0_get_vm_pde(struct amdgpu_device *adev, int level,
 				 uint64_t *addr, uint64_t *flags)
 {
@@ -617,21 +492,39 @@ static void gmc_v10_0_get_vm_pde(struct amdgpu_device *adev, int level,
 }
 
 static void gmc_v10_0_get_vm_pte(struct amdgpu_device *adev,
-				 struct amdgpu_bo_va_mapping *mapping,
+				 struct amdgpu_vm *vm,
+				 struct amdgpu_bo *bo,
+				 uint32_t vm_flags,
 				 uint64_t *flags)
 {
-	struct amdgpu_bo *bo = mapping->bo_va->base.bo;
-
-	*flags &= ~AMDGPU_PTE_EXECUTABLE;
-	*flags |= mapping->flags & AMDGPU_PTE_EXECUTABLE;
+	if (vm_flags & AMDGPU_VM_PAGE_EXECUTABLE)
+		*flags |= AMDGPU_PTE_EXECUTABLE;
+	else
+		*flags &= ~AMDGPU_PTE_EXECUTABLE;
 
-	*flags &= ~AMDGPU_PTE_MTYPE_NV10_MASK;
-	*flags |= (mapping->flags & AMDGPU_PTE_MTYPE_NV10_MASK);
+	switch (vm_flags & AMDGPU_VM_MTYPE_MASK) {
+	case AMDGPU_VM_MTYPE_DEFAULT:
+	case AMDGPU_VM_MTYPE_NC:
+	default:
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_NC);
+		break;
+	case AMDGPU_VM_MTYPE_WC:
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_WC);
+		break;
+	case AMDGPU_VM_MTYPE_CC:
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_CC);
+		break;
+	case AMDGPU_VM_MTYPE_UC:
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_UC);
+		break;
+	}
 
-	*flags &= ~AMDGPU_PTE_NOALLOC;
-	*flags |= (mapping->flags & AMDGPU_PTE_NOALLOC);
+	if (vm_flags & AMDGPU_VM_PAGE_NOALLOC)
+		*flags |= AMDGPU_PTE_NOALLOC;
+	else
+		*flags &= ~AMDGPU_PTE_NOALLOC;
 
-	if (mapping->flags & AMDGPU_PTE_PRT) {
+	if (vm_flags & AMDGPU_VM_PAGE_PRT) {
 		*flags |= AMDGPU_PTE_PRT;
 		*flags |= AMDGPU_PTE_SNOOPED;
 		*flags |= AMDGPU_PTE_LOG;
@@ -640,15 +533,15 @@ static void gmc_v10_0_get_vm_pte(struct amdgpu_device *adev,
 	}
 
 	if (bo && bo->flags & (AMDGPU_GEM_CREATE_COHERENT |
+			       AMDGPU_GEM_CREATE_EXT_COHERENT |
 			       AMDGPU_GEM_CREATE_UNCACHED))
-		*flags = (*flags & ~AMDGPU_PTE_MTYPE_NV10_MASK) |
-			 AMDGPU_PTE_MTYPE_NV10(MTYPE_UC);
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_UC);
 }
 
-static unsigned gmc_v10_0_get_vbios_fb_size(struct amdgpu_device *adev)
+static unsigned int gmc_v10_0_get_vbios_fb_size(struct amdgpu_device *adev)
 {
 	u32 d1vga_control = RREG32_SOC15(DCE, 0, mmD1VGA_CONTROL);
-	unsigned size;
+	unsigned int size;
 
 	if (REG_GET_FIELD(d1vga_control, D1VGA_CONTROL, D1VGA_MODE_ENABLE)) {
 		size = AMDGPU_VBIOS_VGA_ALLOCATION;
@@ -672,7 +565,6 @@ static const struct amdgpu_gmc_funcs gmc_v10_0_gmc_funcs = {
 	.flush_gpu_tlb_pasid = gmc_v10_0_flush_gpu_tlb_pasid,
 	.emit_flush_gpu_tlb = gmc_v10_0_emit_flush_gpu_tlb,
 	.emit_pasid_mapping = gmc_v10_0_emit_pasid_mapping,
-	.map_mtype = gmc_v10_0_map_mtype,
 	.get_vm_pde = gmc_v10_0_get_vm_pde,
 	.get_vm_pte = gmc_v10_0_get_vm_pte,
 	.get_vbios_fb_size = gmc_v10_0_get_vbios_fb_size,
@@ -686,7 +578,7 @@ static void gmc_v10_0_set_gmc_funcs(struct amdgpu_device *adev)
 
 static void gmc_v10_0_set_umc_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[UMC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, UMC_HWIP, 0)) {
 	case IP_VERSION(8, 7, 0):
 		adev->umc.max_ras_err_cnt_per_query = UMC_V8_7_TOTAL_CHANNEL_NUM;
 		adev->umc.channel_inst_num = UMC_V8_7_CHANNEL_INSTANCE_NUM;
@@ -703,7 +595,7 @@ static void gmc_v10_0_set_umc_funcs(struct amdgpu_device *adev)
 
 static void gmc_v10_0_set_mmhub_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(2, 3, 0):
 	case IP_VERSION(2, 4, 0):
 	case IP_VERSION(2, 4, 1):
@@ -717,7 +609,7 @@ static void gmc_v10_0_set_mmhub_funcs(struct amdgpu_device *adev)
 
 static void gmc_v10_0_set_gfxhub_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 0):
 	case IP_VERSION(10, 3, 2):
 	case IP_VERSION(10, 3, 1):
@@ -735,9 +627,9 @@ static void gmc_v10_0_set_gfxhub_funcs(struct amdgpu_device *adev)
 }
 
 
-static int gmc_v10_0_early_init(void *handle)
+static int gmc_v10_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v10_0_set_mmhub_funcs(adev);
 	gmc_v10_0_set_gfxhub_funcs(adev);
@@ -751,13 +643,14 @@ static int gmc_v10_0_early_init(void *handle)
 	adev->gmc.private_aperture_start = 0x1000000000000000ULL;
 	adev->gmc.private_aperture_end =
 		adev->gmc.private_aperture_start + (4ULL << 30) - 1;
+	adev->gmc.noretry_flags = AMDGPU_VM_NORETRY_FLAGS_TF;
 
 	return 0;
 }
 
-static int gmc_v10_0_late_init(void *handle)
+static int gmc_v10_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_gmc_allocate_vm_inv_eng(adev);
@@ -781,9 +674,11 @@ static void gmc_v10_0_vram_gtt_location(struct amdgpu_device *adev,
 	/* add the xgmi offset of the physical node */
 	base += adev->gmc.xgmi.physical_node_id * adev->gmc.xgmi.node_segment_size;
 
+	amdgpu_gmc_set_agp_default(adev, mc);
 	amdgpu_gmc_vram_location(adev, &adev->gmc, base);
-	amdgpu_gmc_gart_location(adev, mc);
-	amdgpu_gmc_agp_location(adev, mc);
+	amdgpu_gmc_gart_location(adev, mc, AMDGPU_GART_PLACEMENT_BEST_FIT);
+	if (!amdgpu_sriov_vf(adev) && (amdgpu_agp == 1))
+		amdgpu_gmc_agp_location(adev, mc);
 
 	/* base offset of vram pages */
 	adev->vm_manager.vram_base_offset = adev->gfxhub.funcs->get_mc_fb_offset(adev);
@@ -830,7 +725,7 @@ static int gmc_v10_0_mc_init(struct amdgpu_device *adev)
 
 	/* set the gart size */
 	if (amdgpu_gart_size == -1) {
-		switch (adev->ip_versions[GC_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 		default:
 			adev->gmc.gart_size = 512ULL << 20;
 			break;
@@ -865,16 +760,16 @@ static int gmc_v10_0_gart_init(struct amdgpu_device *adev)
 		return r;
 
 	adev->gart.table_size = adev->gart.num_gpu_pages * 8;
-	adev->gart.gart_pte_flags = AMDGPU_PTE_MTYPE_NV10(MTYPE_UC) |
+	adev->gart.gart_pte_flags = AMDGPU_PTE_MTYPE_NV10(0ULL, MTYPE_UC) |
 				 AMDGPU_PTE_EXECUTABLE;
 
 	return amdgpu_gart_table_vram_alloc(adev);
 }
 
-static int gmc_v10_0_sw_init(void *handle)
+static int gmc_v10_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r, vram_width = 0, vram_type = 0, vram_vendor = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->gfxhub.funcs->init(adev);
 
@@ -897,7 +792,7 @@ static int gmc_v10_0_sw_init(void *handle)
 		adev->gmc.vram_vendor = vram_vendor;
 	}
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 0):
 		adev->gmc.mall_size = 128 * 1024 * 1024;
 		break;
@@ -915,7 +810,7 @@ static int gmc_v10_0_sw_init(void *handle)
 		break;
 	}
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 	case IP_VERSION(10, 1, 1):
 	case IP_VERSION(10, 1, 2):
@@ -929,7 +824,8 @@ static int gmc_v10_0_sw_init(void *handle)
 	case IP_VERSION(10, 3, 6):
 	case IP_VERSION(10, 3, 3):
 	case IP_VERSION(10, 3, 7):
-		adev->num_vmhubs = 2;
+		set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
+		set_bit(AMDGPU_MMHUB0(0), adev->vmhubs_mask);
 		/*
 		 * To fulfill 4-level page support,
 		 * vm size is 256TB (48bit), maximum size of Navi10/Navi14/Navi12,
@@ -971,7 +867,7 @@ static int gmc_v10_0_sw_init(void *handle)
 
 	r = dma_set_mask_and_coherent(adev->dev, DMA_BIT_MASK(44));
 	if (r) {
-		printk(KERN_WARNING "amdgpu: No suitable DMA available.\n");
+		dev_warn(adev->dev, "amdgpu: No suitable DMA available.\n");
 		return r;
 	}
 
@@ -1021,9 +917,9 @@ static void gmc_v10_0_gart_fini(struct amdgpu_device *adev)
 	amdgpu_gart_table_vram_free(adev);
 }
 
-static int gmc_v10_0_sw_fini(void *handle)
+static int gmc_v10_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_vm_manager_fini(adev);
 	gmc_v10_0_gart_fini(adev);
@@ -1067,29 +963,30 @@ static int gmc_v10_0_gart_enable(struct amdgpu_device *adev)
 	adev->hdp.funcs->init_registers(adev);
 
 	/* Flush HDP after it is initialized */
-	adev->hdp.funcs->flush_hdp(adev, NULL);
+	amdgpu_device_flush_hdp(adev, NULL);
 
-	value = (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS) ?
-		false : true;
+	value = amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS;
 
 	if (!adev->in_s0ix)
 		adev->gfxhub.funcs->set_fault_enable_default(adev, value);
 	adev->mmhub.funcs->set_fault_enable_default(adev, value);
-	gmc_v10_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB_0, 0);
+	gmc_v10_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB0(0), 0);
 	if (!adev->in_s0ix)
-		gmc_v10_0_flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB_0, 0);
+		gmc_v10_0_flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB(0), 0);
 
 	DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
-		 (unsigned)(adev->gmc.gart_size >> 20),
+		 (unsigned int)(adev->gmc.gart_size >> 20),
 		 (unsigned long long)amdgpu_bo_gpu_offset(adev->gart.bo));
 
 	return 0;
 }
 
-static int gmc_v10_0_hw_init(void *handle)
+static int gmc_v10_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+	adev->gmc.flush_pasid_uses_kiq = !amdgpu_emu_mode;
 
 	/* The sequence of these two function calls matters.*/
 	gmc_v10_0_init_golden_registers(adev);
@@ -1131,9 +1028,9 @@ static void gmc_v10_0_gart_disable(struct amdgpu_device *adev)
 	adev->mmhub.funcs->gart_disable(adev);
 }
 
-static int gmc_v10_0_hw_fini(void *handle)
+static int gmc_v10_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v10_0_gart_disable(adev);
 
@@ -1143,64 +1040,60 @@ static int gmc_v10_0_hw_fini(void *handle)
 		return 0;
 	}
 
-	amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0);
 	amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
 
+	if (adev->gmc.ecc_irq.funcs &&
+		amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC))
+		amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0);
+
 	return 0;
 }
 
-static int gmc_v10_0_suspend(void *handle)
+static int gmc_v10_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	gmc_v10_0_hw_fini(adev);
+	gmc_v10_0_hw_fini(ip_block);
 
 	return 0;
 }
 
-static int gmc_v10_0_resume(void *handle)
+static int gmc_v10_0_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = gmc_v10_0_hw_init(adev);
+	r = gmc_v10_0_hw_init(ip_block);
 	if (r)
 		return r;
 
-	amdgpu_vmid_reset_all(adev);
+	amdgpu_vmid_reset_all(ip_block->adev);
 
 	return 0;
 }
 
-static bool gmc_v10_0_is_idle(void *handle)
+static bool gmc_v10_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* MC is always ready in GMC v10.*/
 	return true;
 }
 
-static int gmc_v10_0_wait_for_idle(void *handle)
+static int gmc_v10_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* There is no need to wait for MC idle in GMC v10.*/
 	return 0;
 }
 
-static int gmc_v10_0_soft_reset(void *handle)
-{
-	return 0;
-}
-
-static int gmc_v10_0_set_clockgating_state(void *handle,
+static int gmc_v10_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_clockgating_state state)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/*
 	 * The issue mmhub can't disconnect from DF with MMHUB clock gating being disabled
 	 * is a new problem observed at DF 3.0.3, however with the same suspend sequence not
 	 * seen any issue on the DF 3.0.2 series platform.
 	 */
-	if (adev->in_s0ix && adev->ip_versions[DF_HWIP][0] > IP_VERSION(3, 0, 2)) {
+	if (adev->in_s0ix &&
+	    amdgpu_ip_version(adev, DF_HWIP, 0) > IP_VERSION(3, 0, 2)) {
 		dev_dbg(adev->dev, "keep mmhub clock gating being enabled for s0ix\n");
 		return 0;
 	}
@@ -1209,29 +1102,29 @@ static int gmc_v10_0_set_clockgating_state(void *handle,
 	if (r)
 		return r;
 
-	if (adev->ip_versions[ATHUB_HWIP][0] >= IP_VERSION(2, 1, 0))
+	if (amdgpu_ip_version(adev, ATHUB_HWIP, 0) >= IP_VERSION(2, 1, 0))
 		return athub_v2_1_set_clockgating(adev, state);
 	else
 		return athub_v2_0_set_clockgating(adev, state);
 }
 
-static void gmc_v10_0_get_clockgating_state(void *handle, u64 *flags)
+static void gmc_v10_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 3) ||
-	    adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 4))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 1, 3) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 1, 4))
 		return;
 
 	adev->mmhub.funcs->get_clockgating(adev, flags);
 
-	if (adev->ip_versions[ATHUB_HWIP][0] >= IP_VERSION(2, 1, 0))
+	if (amdgpu_ip_version(adev, ATHUB_HWIP, 0) >= IP_VERSION(2, 1, 0))
 		athub_v2_1_get_clockgating(adev, flags);
 	else
 		athub_v2_0_get_clockgating(adev, flags);
 }
 
-static int gmc_v10_0_set_powergating_state(void *handle,
+static int gmc_v10_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_powergating_state state)
 {
 	return 0;
@@ -1249,14 +1142,12 @@ const struct amd_ip_funcs gmc_v10_0_ip_funcs = {
 	.resume = gmc_v10_0_resume,
 	.is_idle = gmc_v10_0_is_idle,
 	.wait_for_idle = gmc_v10_0_wait_for_idle,
-	.soft_reset = gmc_v10_0_soft_reset,
 	.set_clockgating_state = gmc_v10_0_set_clockgating_state,
 	.set_powergating_state = gmc_v10_0_set_powergating_state,
 	.get_clockgating_state = gmc_v10_0_get_clockgating_state,
 };
 
-const struct amdgpu_ip_block_version gmc_v10_0_ip_block =
-{
+const struct amdgpu_ip_block_version gmc_v10_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GMC,
 	.major = 10,
 	.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
index d809f2ed5600..ba59ee8e398a 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
@@ -31,6 +31,8 @@
 #include "umc_v8_10.h"
 #include "athub/athub_3_0_0_sh_mask.h"
 #include "athub/athub_3_0_0_offset.h"
+#include "dcn/dcn_3_2_0_offset.h"
+#include "dcn/dcn_3_2_0_sh_mask.h"
 #include "oss/osssys_6_0_0_offset.h"
 #include "ivsrcid/vmc/irqsrcs_vmc_1_0.h"
 #include "navi10_enum.h"
@@ -40,15 +42,17 @@
 #include "nbio_v4_3.h"
 #include "gfxhub_v3_0.h"
 #include "gfxhub_v3_0_3.h"
+#include "gfxhub_v11_5_0.h"
 #include "mmhub_v3_0.h"
 #include "mmhub_v3_0_1.h"
 #include "mmhub_v3_0_2.h"
+#include "mmhub_v3_3.h"
 #include "athub_v3_0.h"
 
 
 static int gmc_v11_0_ecc_interrupt_state(struct amdgpu_device *adev,
 					 struct amdgpu_irq_src *src,
-					 unsigned type,
+					 unsigned int type,
 					 enum amdgpu_interrupt_state state)
 {
 	return 0;
@@ -56,25 +60,26 @@ static int gmc_v11_0_ecc_interrupt_state(struct amdgpu_device *adev,
 
 static int
 gmc_v11_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
-				   struct amdgpu_irq_src *src, unsigned type,
+				   struct amdgpu_irq_src *src, unsigned int type,
 				   enum amdgpu_interrupt_state state)
 {
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
 		/* MM HUB */
-		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB_0, false);
+		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), false);
 		/* GFX HUB */
 		/* This works because this interrupt is only
 		 * enabled at init/resume and disabled in
 		 * fini/suspend, so the overall state doesn't
 		 * change over the course of suspend/resume.
 		 */
-		if (!adev->in_s0ix)
-			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB_0, false);
+		if (!adev->in_s0ix && (adev->in_runpm || adev->in_suspend ||
+							   amdgpu_in_reset(adev)))
+			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), false);
 		break;
 	case AMDGPU_IRQ_STATE_ENABLE:
 		/* MM HUB */
-		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB_0, true);
+		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), true);
 		/* GFX HUB */
 		/* This works because this interrupt is only
 		 * enabled at init/resume and disabled in
@@ -82,7 +87,7 @@ gmc_v11_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 		 * change over the course of suspend/resume.
 		 */
 		if (!adev->in_s0ix)
-			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB_0, true);
+			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), true);
 		break;
 	default:
 		break;
@@ -95,42 +100,80 @@ static int gmc_v11_0_process_interrupt(struct amdgpu_device *adev,
 				       struct amdgpu_irq_src *source,
 				       struct amdgpu_iv_entry *entry)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[entry->vmid_src];
+	uint32_t vmhub_index = entry->client_id == SOC21_IH_CLIENTID_VMC ?
+			       AMDGPU_MMHUB0(0) : AMDGPU_GFXHUB(0);
+	struct amdgpu_vmhub *hub = &adev->vmhub[vmhub_index];
+	bool retry_fault = !!(entry->src_data[1] &
+			      AMDGPU_GMC9_FAULT_SOURCE_DATA_RETRY);
+	bool write_fault = !!(entry->src_data[1] &
+			      AMDGPU_GMC9_FAULT_SOURCE_DATA_WRITE);
 	uint32_t status = 0;
 	u64 addr;
 
 	addr = (u64)entry->src_data[0] << 12;
 	addr |= ((u64)entry->src_data[1] & 0xf) << 44;
 
+	if (retry_fault) {
+		/* Returning 1 here also prevents sending the IV to the KFD */
+
+		/* Process it only if it's the first fault for this address */
+		if (entry->ih != &adev->irq.ih_soft &&
+		    amdgpu_gmc_filter_faults(adev, entry->ih, addr, entry->pasid,
+					     entry->timestamp))
+			return 1;
+
+		/* Delegate it to a different ring if the hardware hasn't
+		 * already done it.
+		 */
+		if (entry->ih == &adev->irq.ih) {
+			amdgpu_irq_delegate(adev, entry, 8);
+			return 1;
+		}
+
+		/* Try to handle the recoverable page faults by filling page
+		 * tables
+		 */
+		if (amdgpu_vm_handle_fault(adev, entry->pasid, 0, 0, addr,
+					   entry->timestamp, write_fault))
+			return 1;
+	}
+
 	if (!amdgpu_sriov_vf(adev)) {
 		/*
 		 * Issue a dummy read to wait for the status register to
 		 * be updated to avoid reading an incorrect value due to
 		 * the new fast GRBM interface.
 		 */
-		if (entry->vmid_src == AMDGPU_GFXHUB_0)
+		if (entry->vmid_src == AMDGPU_GFXHUB(0))
 			RREG32(hub->vm_l2_pro_fault_status);
 
 		status = RREG32(hub->vm_l2_pro_fault_status);
 		WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1);
+
+		amdgpu_vm_update_fault_cache(adev, entry->pasid, addr, status,
+					     entry->vmid_src ? AMDGPU_MMHUB0(0) : AMDGPU_GFXHUB(0));
 	}
 
 	if (printk_ratelimit()) {
-		struct amdgpu_task_info task_info;
-
-		memset(&task_info, 0, sizeof(struct amdgpu_task_info));
-		amdgpu_vm_get_task_info(adev, entry->pasid, &task_info);
+		struct amdgpu_task_info *task_info;
 
 		dev_err(adev->dev,
-			"[%s] page fault (src_id:%u ring:%u vmid:%u pasid:%u, "
-			"for process %s pid %d thread %s pid %d)\n",
+			"[%s] page fault (src_id:%u ring:%u vmid:%u pasid:%u)\n",
 			entry->vmid_src ? "mmhub" : "gfxhub",
-			entry->src_id, entry->ring_id, entry->vmid,
-			entry->pasid, task_info.process_name, task_info.tgid,
-			task_info.task_name, task_info.pid);
+			entry->src_id, entry->ring_id, entry->vmid, entry->pasid);
+		task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid);
+		if (task_info) {
+			amdgpu_vm_print_task_info(adev, task_info);
+			amdgpu_vm_put_task_info(task_info);
+		}
+
 		dev_err(adev->dev, "  in page starting at address 0x%016llx from client %d\n",
-			addr, entry->client_id);
-		if (!amdgpu_sriov_vf(adev))
+				addr, entry->client_id);
+
+		/* Only print L2 fault status if the status register could be read and
+		 * contains useful information
+		 */
+		if (status != 0)
 			hub->vmhub_funcs->print_l2_protection_fault_status(adev, status);
 	}
 
@@ -168,7 +211,7 @@ static void gmc_v11_0_set_irq_funcs(struct amdgpu_device *adev)
 static bool gmc_v11_0_use_invalidate_semaphore(struct amdgpu_device *adev,
 				       uint32_t vmhub)
 {
-	return ((vmhub == AMDGPU_MMHUB_0) &&
+	return ((vmhub == AMDGPU_MMHUB0(0)) &&
 		(!amdgpu_sriov_vf(adev)));
 }
 
@@ -181,27 +224,51 @@ static bool gmc_v11_0_get_vmid_pasid_mapping_info(
 	return !!(*p_pasid);
 }
 
-/*
- * GART
- * VMID 0 is the physical GPU addresses as used by the kernel.
- * VMIDs 1-15 are used for userspace clients and are handled
- * by the amdgpu vm/hsa code.
+/**
+ * gmc_v11_0_flush_gpu_tlb - gart tlb flush callback
+ *
+ * @adev: amdgpu_device pointer
+ * @vmid: vm instance to flush
+ * @vmhub: which hub to flush
+ * @flush_type: the flush type
+ *
+ * Flush the TLB for the requested page table.
  */
-
-static void gmc_v11_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
-				   unsigned int vmhub, uint32_t flush_type)
+static void gmc_v11_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
+					uint32_t vmhub, uint32_t flush_type)
 {
 	bool use_semaphore = gmc_v11_0_use_invalidate_semaphore(adev, vmhub);
 	struct amdgpu_vmhub *hub = &adev->vmhub[vmhub];
 	u32 inv_req = hub->vmhub_funcs->get_invalidate_req(vmid, flush_type);
-	u32 tmp;
 	/* Use register 17 for GART */
-	const unsigned eng = 17;
+	const unsigned int eng = 17;
+	unsigned char hub_ip;
+	u32 sem, req, ack;
 	unsigned int i;
-	unsigned char hub_ip = 0;
+	u32 tmp;
+
+	if ((vmhub == AMDGPU_GFXHUB(0)) && !adev->gfx.is_poweron)
+		return;
 
-	hub_ip = (vmhub == AMDGPU_GFXHUB_0) ?
-		   GC_HWIP : MMHUB_HWIP;
+	sem = hub->vm_inv_eng0_sem + hub->eng_distance * eng;
+	req = hub->vm_inv_eng0_req + hub->eng_distance * eng;
+	ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng;
+
+	/* flush hdp cache */
+	amdgpu_device_flush_hdp(adev, NULL);
+
+	/* This is necessary for SRIOV as well as for GFXOFF to function
+	 * properly under bare metal
+	 */
+	if ((adev->gfx.kiq[0].ring.sched.ready || adev->mes.ring[0].sched.ready) &&
+	    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) {
+		amdgpu_gmc_fw_reg_write_reg_wait(adev, req, ack, inv_req,
+						 1 << vmid, GET_INST(GC, 0));
+		return;
+	}
+
+	/* This path is needed before KIQ/MES/GFXOFF are set up */
+	hub_ip = (vmhub == AMDGPU_GFXHUB(0)) ? GC_HWIP : MMHUB_HWIP;
 
 	spin_lock(&adev->gmc.invalidate_lock);
 	/*
@@ -215,8 +282,7 @@ static void gmc_v11_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
 	if (use_semaphore) {
 		for (i = 0; i < adev->usec_timeout; i++) {
 			/* a read return value of 1 means semaphore acuqire */
-			tmp = RREG32_RLC_NO_KIQ(hub->vm_inv_eng0_sem +
-					    hub->eng_distance * eng, hub_ip);
+			tmp = RREG32_RLC_NO_KIQ(sem, hub_ip);
 			if (tmp & 0x1)
 				break;
 			udelay(1);
@@ -226,12 +292,11 @@ static void gmc_v11_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
 			DRM_ERROR("Timeout waiting for sem acquire in VM flush!\n");
 	}
 
-	WREG32_RLC_NO_KIQ(hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req, hub_ip);
+	WREG32_RLC_NO_KIQ(req, inv_req, hub_ip);
 
 	/* Wait for ACK with a delay.*/
 	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32_RLC_NO_KIQ(hub->vm_inv_eng0_ack +
-				    hub->eng_distance * eng, hub_ip);
+		tmp = RREG32_RLC_NO_KIQ(ack, hub_ip);
 		tmp &= 1 << vmid;
 		if (tmp)
 			break;
@@ -241,15 +306,10 @@ static void gmc_v11_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
 
 	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
 	if (use_semaphore)
-		/*
-		 * add semaphore release after invalidation,
-		 * write with 0 means semaphore release
-		 */
-		WREG32_RLC_NO_KIQ(hub->vm_inv_eng0_sem +
-			      hub->eng_distance * eng, 0, hub_ip);
+		WREG32_RLC_NO_KIQ(sem, 0, hub_ip);
 
 	/* Issue additional private vm invalidation to MMHUB */
-	if ((vmhub != AMDGPU_GFXHUB_0) &&
+	if ((vmhub != AMDGPU_GFXHUB(0)) &&
 	    (hub->vm_l2_bank_select_reserved_cid2) &&
 		!amdgpu_sriov_vf(adev)) {
 		inv_req = RREG32_NO_KIQ(hub->vm_l2_bank_select_reserved_cid2);
@@ -263,51 +323,8 @@ static void gmc_v11_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
 
 	spin_unlock(&adev->gmc.invalidate_lock);
 
-	if (i < adev->usec_timeout)
-		return;
-
-	DRM_ERROR("Timeout waiting for VM flush ACK!\n");
-}
-
-/**
- * gmc_v11_0_flush_gpu_tlb - gart tlb flush callback
- *
- * @adev: amdgpu_device pointer
- * @vmid: vm instance to flush
- * @vmhub: which hub to flush
- * @flush_type: the flush type
- *
- * Flush the TLB for the requested page table.
- */
-static void gmc_v11_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
-					uint32_t vmhub, uint32_t flush_type)
-{
-	if ((vmhub == AMDGPU_GFXHUB_0) && !adev->gfx.is_poweron)
-		return;
-
-	/* flush hdp cache */
-	adev->hdp.funcs->flush_hdp(adev, NULL);
-
-	/* For SRIOV run time, driver shouldn't access the register through MMIO
-	 * Directly use kiq to do the vm invalidation instead
-	 */
-	if ((adev->gfx.kiq.ring.sched.ready || adev->mes.ring.sched.ready) &&
-	    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) {
-		struct amdgpu_vmhub *hub = &adev->vmhub[vmhub];
-		const unsigned eng = 17;
-		u32 inv_req = hub->vmhub_funcs->get_invalidate_req(vmid, flush_type);
-		u32 req = hub->vm_inv_eng0_req + hub->eng_distance * eng;
-		u32 ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng;
-
-		amdgpu_virt_kiq_reg_write_reg_wait(adev, req, ack, inv_req,
-				1 << vmid);
-		return;
-	}
-
-	mutex_lock(&adev->mman.gtt_window_lock);
-	gmc_v11_0_flush_vm_hub(adev, vmid, vmhub, 0);
-	mutex_unlock(&adev->mman.gtt_window_lock);
-	return;
+	if (i >= adev->usec_timeout)
+		dev_err(adev->dev, "Timeout waiting for VM flush ACK!\n");
 }
 
 /**
@@ -317,71 +334,44 @@ static void gmc_v11_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
  * @pasid: pasid to be flush
  * @flush_type: the flush type
  * @all_hub: flush all hubs
+ * @inst: is used to select which instance of KIQ to use for the invalidation
  *
  * Flush the TLB for the requested pasid.
  */
-static int gmc_v11_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
-					uint16_t pasid, uint32_t flush_type,
-					bool all_hub)
+static void gmc_v11_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
+					  uint16_t pasid, uint32_t flush_type,
+					  bool all_hub, uint32_t inst)
 {
+	uint16_t queried;
 	int vmid, i;
-	signed long r;
-	uint32_t seq;
-	uint16_t queried_pasid;
-	bool ret;
-	struct amdgpu_ring *ring = &adev->gfx.kiq.ring;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-
-	if (amdgpu_emu_mode == 0 && ring->sched.ready) {
-		spin_lock(&adev->gfx.kiq.ring_lock);
-		/* 2 dwords flush + 8 dwords fence */
-		amdgpu_ring_alloc(ring, kiq->pmf->invalidate_tlbs_size + 8);
-		kiq->pmf->kiq_invalidate_tlbs(ring,
-					pasid, flush_type, all_hub);
-		r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
-		if (r) {
-			amdgpu_ring_undo(ring);
-			spin_unlock(&adev->gfx.kiq.ring_lock);
-			return -ETIME;
-		}
-
-		amdgpu_ring_commit(ring);
-		spin_unlock(&adev->gfx.kiq.ring_lock);
-		r = amdgpu_fence_wait_polling(ring, seq, adev->usec_timeout);
-		if (r < 1) {
-			dev_err(adev->dev, "wait for kiq fence error: %ld.\n", r);
-			return -ETIME;
-		}
-
-		return 0;
-	}
 
 	for (vmid = 1; vmid < 16; vmid++) {
-
-		ret = gmc_v11_0_get_vmid_pasid_mapping_info(adev, vmid,
-				&queried_pasid);
-		if (ret	&& queried_pasid == pasid) {
-			if (all_hub) {
-				for (i = 0; i < adev->num_vmhubs; i++)
-					gmc_v11_0_flush_gpu_tlb(adev, vmid,
-							i, flush_type);
-			} else {
-				gmc_v11_0_flush_gpu_tlb(adev, vmid,
-						AMDGPU_GFXHUB_0, flush_type);
-			}
+		bool valid;
+
+		valid = gmc_v11_0_get_vmid_pasid_mapping_info(adev, vmid,
+							      &queried);
+		if (!valid || queried != pasid)
+			continue;
+
+		if (all_hub) {
+			for_each_set_bit(i, adev->vmhubs_mask,
+					 AMDGPU_MAX_VMHUBS)
+				gmc_v11_0_flush_gpu_tlb(adev, vmid, i,
+							flush_type);
+		} else {
+			gmc_v11_0_flush_gpu_tlb(adev, vmid, AMDGPU_GFXHUB(0),
+						flush_type);
 		}
 	}
-
-	return 0;
 }
 
 static uint64_t gmc_v11_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
-					     unsigned vmid, uint64_t pd_addr)
+					     unsigned int vmid, uint64_t pd_addr)
 {
 	bool use_semaphore = gmc_v11_0_use_invalidate_semaphore(ring->adev, ring->vm_hub);
 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
 	uint32_t req = hub->vmhub_funcs->get_invalidate_req(vmid, 0);
-	unsigned eng = ring->vm_inv_eng;
+	unsigned int eng = ring->vm_inv_eng;
 
 	/*
 	 * It may lose gpuvm invalidate acknowldege state across power-gating
@@ -423,17 +413,13 @@ static uint64_t gmc_v11_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
 	return pd_addr;
 }
 
-static void gmc_v11_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
-					 unsigned pasid)
+static void gmc_v11_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned int vmid,
+					 unsigned int pasid)
 {
 	struct amdgpu_device *adev = ring->adev;
 	uint32_t reg;
 
-	/* MES fw manages IH_VMID_x_LUT updating */
-	if (ring->is_mes_queue)
-		return;
-
-	if (ring->vm_hub == AMDGPU_GFXHUB_0)
+	if (ring->vm_hub == AMDGPU_GFXHUB(0))
 		reg = SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT) + vmid;
 	else
 		reg = SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT_MM) + vmid;
@@ -473,30 +459,11 @@ static void gmc_v11_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid
  * 0 valid
  */
 
-static uint64_t gmc_v11_0_map_mtype(struct amdgpu_device *adev, uint32_t flags)
-{
-	switch (flags) {
-	case AMDGPU_VM_MTYPE_DEFAULT:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_NC);
-	case AMDGPU_VM_MTYPE_NC:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_NC);
-	case AMDGPU_VM_MTYPE_WC:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_WC);
-	case AMDGPU_VM_MTYPE_CC:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_CC);
-	case AMDGPU_VM_MTYPE_UC:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_UC);
-	default:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_NC);
-	}
-}
-
 static void gmc_v11_0_get_vm_pde(struct amdgpu_device *adev, int level,
 				 uint64_t *addr, uint64_t *flags)
 {
 	if (!(*flags & AMDGPU_PDE_PTE) && !(*flags & AMDGPU_PTE_SYSTEM))
-		*addr = adev->vm_manager.vram_base_offset + *addr -
-			adev->gmc.vram_start;
+		*addr = amdgpu_gmc_vram_mc2pa(adev, *addr);
 	BUG_ON(*addr & 0xFFFF00000000003FULL);
 
 	if (!adev->gmc.translate_further)
@@ -516,21 +483,39 @@ static void gmc_v11_0_get_vm_pde(struct amdgpu_device *adev, int level,
 }
 
 static void gmc_v11_0_get_vm_pte(struct amdgpu_device *adev,
-				 struct amdgpu_bo_va_mapping *mapping,
+				 struct amdgpu_vm *vm,
+				 struct amdgpu_bo *bo,
+				 uint32_t vm_flags,
 				 uint64_t *flags)
 {
-	struct amdgpu_bo *bo = mapping->bo_va->base.bo;
-
-	*flags &= ~AMDGPU_PTE_EXECUTABLE;
-	*flags |= mapping->flags & AMDGPU_PTE_EXECUTABLE;
+	if (vm_flags & AMDGPU_VM_PAGE_EXECUTABLE)
+		*flags |= AMDGPU_PTE_EXECUTABLE;
+	else
+		*flags &= ~AMDGPU_PTE_EXECUTABLE;
 
-	*flags &= ~AMDGPU_PTE_MTYPE_NV10_MASK;
-	*flags |= (mapping->flags & AMDGPU_PTE_MTYPE_NV10_MASK);
+	switch (vm_flags & AMDGPU_VM_MTYPE_MASK) {
+	case AMDGPU_VM_MTYPE_DEFAULT:
+	case AMDGPU_VM_MTYPE_NC:
+	default:
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_NC);
+		break;
+	case AMDGPU_VM_MTYPE_WC:
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_WC);
+		break;
+	case AMDGPU_VM_MTYPE_CC:
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_CC);
+		break;
+	case AMDGPU_VM_MTYPE_UC:
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_UC);
+		break;
+	}
 
-	*flags &= ~AMDGPU_PTE_NOALLOC;
-	*flags |= (mapping->flags & AMDGPU_PTE_NOALLOC);
+	if (vm_flags & AMDGPU_VM_PAGE_NOALLOC)
+		*flags |= AMDGPU_PTE_NOALLOC;
+	else
+		*flags &= ~AMDGPU_PTE_NOALLOC;
 
-	if (mapping->flags & AMDGPU_PTE_PRT) {
+	if (vm_flags & AMDGPU_VM_PAGE_PRT) {
 		*flags |= AMDGPU_PTE_PRT;
 		*flags |= AMDGPU_PTE_SNOOPED;
 		*flags |= AMDGPU_PTE_LOG;
@@ -539,14 +524,31 @@ static void gmc_v11_0_get_vm_pte(struct amdgpu_device *adev,
 	}
 
 	if (bo && bo->flags & (AMDGPU_GEM_CREATE_COHERENT |
+			       AMDGPU_GEM_CREATE_EXT_COHERENT |
 			       AMDGPU_GEM_CREATE_UNCACHED))
-		*flags = (*flags & ~AMDGPU_PTE_MTYPE_NV10_MASK) |
-			 AMDGPU_PTE_MTYPE_NV10(MTYPE_UC);
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_UC);
 }
 
-static unsigned gmc_v11_0_get_vbios_fb_size(struct amdgpu_device *adev)
+static unsigned int gmc_v11_0_get_vbios_fb_size(struct amdgpu_device *adev)
 {
-	return 0;
+	u32 d1vga_control = RREG32_SOC15(DCE, 0, regD1VGA_CONTROL);
+	unsigned int size;
+
+	if (REG_GET_FIELD(d1vga_control, D1VGA_CONTROL, D1VGA_MODE_ENABLE)) {
+		size = AMDGPU_VBIOS_VGA_ALLOCATION;
+	} else {
+		u32 viewport;
+		u32 pitch;
+
+		viewport = RREG32_SOC15(DCE, 0, regHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION);
+		pitch = RREG32_SOC15(DCE, 0, regHUBPREQ0_DCSURF_SURFACE_PITCH);
+		size = (REG_GET_FIELD(viewport,
+					HUBP0_DCSURF_PRI_VIEWPORT_DIMENSION, PRI_VIEWPORT_HEIGHT) *
+				REG_GET_FIELD(pitch, HUBPREQ0_DCSURF_SURFACE_PITCH, PITCH) *
+				4);
+	}
+
+	return size;
 }
 
 static const struct amdgpu_gmc_funcs gmc_v11_0_gmc_funcs = {
@@ -554,7 +556,6 @@ static const struct amdgpu_gmc_funcs gmc_v11_0_gmc_funcs = {
 	.flush_gpu_tlb_pasid = gmc_v11_0_flush_gpu_tlb_pasid,
 	.emit_flush_gpu_tlb = gmc_v11_0_emit_flush_gpu_tlb,
 	.emit_pasid_mapping = gmc_v11_0_emit_pasid_mapping,
-	.map_mtype = gmc_v11_0_map_mtype,
 	.get_vm_pde = gmc_v11_0_get_vm_pde,
 	.get_vm_pte = gmc_v11_0_get_vm_pte,
 	.get_vbios_fb_size = gmc_v11_0_get_vbios_fb_size,
@@ -567,7 +568,7 @@ static void gmc_v11_0_set_gmc_funcs(struct amdgpu_device *adev)
 
 static void gmc_v11_0_set_umc_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[UMC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, UMC_HWIP, 0)) {
 	case IP_VERSION(8, 10, 0):
 		adev->umc.channel_inst_num = UMC_V8_10_CHANNEL_INSTANCE_NUM;
 		adev->umc.umc_inst_num = UMC_V8_10_UMC_INSTANCE_NUM;
@@ -590,13 +591,18 @@ static void gmc_v11_0_set_umc_funcs(struct amdgpu_device *adev)
 
 static void gmc_v11_0_set_mmhub_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(3, 0, 1):
 		adev->mmhub.funcs = &mmhub_v3_0_1_funcs;
 		break;
 	case IP_VERSION(3, 0, 2):
 		adev->mmhub.funcs = &mmhub_v3_0_2_funcs;
 		break;
+	case IP_VERSION(3, 3, 0):
+	case IP_VERSION(3, 3, 1):
+	case IP_VERSION(3, 3, 2):
+		adev->mmhub.funcs = &mmhub_v3_3_funcs;
+		break;
 	default:
 		adev->mmhub.funcs = &mmhub_v3_0_funcs;
 		break;
@@ -605,19 +611,25 @@ static void gmc_v11_0_set_mmhub_funcs(struct amdgpu_device *adev)
 
 static void gmc_v11_0_set_gfxhub_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 3):
 		adev->gfxhub.funcs = &gfxhub_v3_0_3_funcs;
 		break;
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
+		adev->gfxhub.funcs = &gfxhub_v11_5_0_funcs;
+		break;
 	default:
 		adev->gfxhub.funcs = &gfxhub_v3_0_funcs;
 		break;
 	}
 }
 
-static int gmc_v11_0_early_init(void *handle)
+static int gmc_v11_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v11_0_set_gfxhub_funcs(adev);
 	gmc_v11_0_set_mmhub_funcs(adev);
@@ -631,13 +643,14 @@ static int gmc_v11_0_early_init(void *handle)
 	adev->gmc.private_aperture_start = 0x1000000000000000ULL;
 	adev->gmc.private_aperture_end =
 		adev->gmc.private_aperture_start + (4ULL << 30) - 1;
+	adev->gmc.noretry_flags = AMDGPU_VM_NORETRY_FLAGS_TF;
 
 	return 0;
 }
 
-static int gmc_v11_0_late_init(void *handle)
+static int gmc_v11_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_gmc_allocate_vm_inv_eng(adev);
@@ -658,9 +671,13 @@ static void gmc_v11_0_vram_gtt_location(struct amdgpu_device *adev,
 
 	base = adev->mmhub.funcs->get_fb_location(adev);
 
+	amdgpu_gmc_set_agp_default(adev, mc);
 	amdgpu_gmc_vram_location(adev, &adev->gmc, base);
-	amdgpu_gmc_gart_location(adev, mc);
-	amdgpu_gmc_agp_location(adev, mc);
+	amdgpu_gmc_gart_location(adev, mc, AMDGPU_GART_PLACEMENT_HIGH);
+	if (!amdgpu_sriov_vf(adev) &&
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(11, 5, 0)) &&
+	    (amdgpu_agp == 1))
+		amdgpu_gmc_agp_location(adev, mc);
 
 	/* base offset of vram pages */
 	if (amdgpu_sriov_vf(adev))
@@ -707,9 +724,9 @@ static int gmc_v11_0_mc_init(struct amdgpu_device *adev)
 		adev->gmc.visible_vram_size = adev->gmc.real_vram_size;
 
 	/* set the gart size */
-	if (amdgpu_gart_size == -1) {
+	if (amdgpu_gart_size == -1)
 		adev->gmc.gart_size = 512ULL << 20;
-	} else
+	else
 		adev->gmc.gart_size = (u64)amdgpu_gart_size << 20;
 
 	gmc_v11_0_vram_gtt_location(adev, &adev->gmc);
@@ -732,19 +749,21 @@ static int gmc_v11_0_gart_init(struct amdgpu_device *adev)
 		return r;
 
 	adev->gart.table_size = adev->gart.num_gpu_pages * 8;
-	adev->gart.gart_pte_flags = AMDGPU_PTE_MTYPE_NV10(MTYPE_UC) |
+	adev->gart.gart_pte_flags = AMDGPU_PTE_MTYPE_NV10(0ULL, MTYPE_UC) |
 				 AMDGPU_PTE_EXECUTABLE;
 
 	return amdgpu_gart_table_vram_alloc(adev);
 }
 
-static int gmc_v11_0_sw_init(void *handle)
+static int gmc_v11_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r, vram_width = 0, vram_type = 0, vram_vendor = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->mmhub.funcs->init(adev);
 
+	adev->gfxhub.funcs->init(adev);
+
 	spin_lock_init(&adev->gmc.invalidate_lock);
 
 	r = amdgpu_atomfirmware_get_vram_info(adev,
@@ -754,13 +773,30 @@ static int gmc_v11_0_sw_init(void *handle)
 	adev->gmc.vram_type = vram_type;
 	adev->gmc.vram_vendor = vram_vendor;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	/* The mall_size is already calculated as mall_size_per_umc * num_umc.
+	 * However, for gfx1151, which features a 2-to-1 UMC mapping,
+	 * the result must be multiplied by 2 to determine the actual mall size.
+	 */
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(11, 5, 1):
+		adev->gmc.mall_size *= 2;
+		break;
+	default:
+		break;
+	}
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 0):
 	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 3):
 	case IP_VERSION(11, 0, 4):
-		adev->num_vmhubs = 2;
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
+		set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
+		set_bit(AMDGPU_MMHUB0(0), adev->vmhubs_mask);
 		/*
 		 * To fulfill 4-level page support,
 		 * vm size is 256TB (48bit), maximum size,
@@ -802,7 +838,7 @@ static int gmc_v11_0_sw_init(void *handle)
 
 	r = dma_set_mask_and_coherent(adev->dev, DMA_BIT_MASK(44));
 	if (r) {
-		printk(KERN_WARNING "amdgpu: No suitable DMA available.\n");
+		dev_warn(adev->dev, "amdgpu: No suitable DMA available.\n");
 		return r;
 	}
 
@@ -829,7 +865,7 @@ static int gmc_v11_0_sw_init(void *handle)
 	 * amdgpu graphics/compute will use VMIDs 1-7
 	 * amdkfd will use VMIDs 8-15
 	 */
-	adev->vm_manager.first_kfd_vmid = 8;
+	adev->vm_manager.first_kfd_vmid = adev->gfx.disable_kq ? 1 : 8;
 
 	amdgpu_vm_manager_init(adev);
 
@@ -852,9 +888,9 @@ static void gmc_v11_0_gart_fini(struct amdgpu_device *adev)
 	amdgpu_gart_table_vram_free(adev);
 }
 
-static int gmc_v11_0_sw_fini(void *handle)
+static int gmc_v11_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_vm_manager_fini(adev);
 	gmc_v11_0_gart_fini(adev);
@@ -867,7 +903,7 @@ static int gmc_v11_0_sw_fini(void *handle)
 static void gmc_v11_0_init_golden_registers(struct amdgpu_device *adev)
 {
 	if (amdgpu_sriov_vf(adev)) {
-		struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+		struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 
 		WREG32(hub->vm_contexts_disable, 0);
 		return;
@@ -896,25 +932,26 @@ static int gmc_v11_0_gart_enable(struct amdgpu_device *adev)
 		return r;
 
 	/* Flush HDP after it is initialized */
-	adev->hdp.funcs->flush_hdp(adev, NULL);
+	amdgpu_device_flush_hdp(adev, NULL);
 
-	value = (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS) ?
-		false : true;
+	value = amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS;
 
 	adev->mmhub.funcs->set_fault_enable_default(adev, value);
-	gmc_v11_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB_0, 0);
+	gmc_v11_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB0(0), 0);
 
 	DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
-		 (unsigned)(adev->gmc.gart_size >> 20),
+		 (unsigned int)(adev->gmc.gart_size >> 20),
 		 (unsigned long long)amdgpu_bo_gpu_offset(adev->gart.bo));
 
 	return 0;
 }
 
-static int gmc_v11_0_hw_init(void *handle)
+static int gmc_v11_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+	adev->gmc.flush_pasid_uses_kiq = !amdgpu_emu_mode;
 
 	/* The sequence of these two function calls matters.*/
 	gmc_v11_0_init_golden_registers(adev);
@@ -941,9 +978,9 @@ static void gmc_v11_0_gart_disable(struct amdgpu_device *adev)
 	adev->mmhub.funcs->gart_disable(adev);
 }
 
-static int gmc_v11_0_hw_fini(void *handle)
+static int gmc_v11_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev)) {
 		/* full access mode, so don't touch any GMC register */
@@ -951,58 +988,54 @@ static int gmc_v11_0_hw_fini(void *handle)
 		return 0;
 	}
 
-	amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0);
 	amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
+
+	if (adev->gmc.ecc_irq.funcs &&
+		amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC))
+		amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0);
+
 	gmc_v11_0_gart_disable(adev);
 
 	return 0;
 }
 
-static int gmc_v11_0_suspend(void *handle)
+static int gmc_v11_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	gmc_v11_0_hw_fini(adev);
+	gmc_v11_0_hw_fini(ip_block);
 
 	return 0;
 }
 
-static int gmc_v11_0_resume(void *handle)
+static int gmc_v11_0_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = gmc_v11_0_hw_init(adev);
+	r = gmc_v11_0_hw_init(ip_block);
 	if (r)
 		return r;
 
-	amdgpu_vmid_reset_all(adev);
+	amdgpu_vmid_reset_all(ip_block->adev);
 
 	return 0;
 }
 
-static bool gmc_v11_0_is_idle(void *handle)
+static bool gmc_v11_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* MC is always ready in GMC v11.*/
 	return true;
 }
 
-static int gmc_v11_0_wait_for_idle(void *handle)
+static int gmc_v11_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* There is no need to wait for MC idle in GMC v11.*/
 	return 0;
 }
 
-static int gmc_v11_0_soft_reset(void *handle)
-{
-	return 0;
-}
-
-static int gmc_v11_0_set_clockgating_state(void *handle,
+static int gmc_v11_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_clockgating_state state)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = adev->mmhub.funcs->set_clockgating(adev, state);
 	if (r)
@@ -1011,16 +1044,16 @@ static int gmc_v11_0_set_clockgating_state(void *handle,
 	return athub_v3_0_set_clockgating(adev, state);
 }
 
-static void gmc_v11_0_get_clockgating_state(void *handle, u64 *flags)
+static void gmc_v11_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->mmhub.funcs->get_clockgating(adev, flags);
 
 	athub_v3_0_get_clockgating(adev, flags);
 }
 
-static int gmc_v11_0_set_powergating_state(void *handle,
+static int gmc_v11_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_powergating_state state)
 {
 	return 0;
@@ -1038,7 +1071,6 @@ const struct amd_ip_funcs gmc_v11_0_ip_funcs = {
 	.resume = gmc_v11_0_resume,
 	.is_idle = gmc_v11_0_is_idle,
 	.wait_for_idle = gmc_v11_0_wait_for_idle,
-	.soft_reset = gmc_v11_0_soft_reset,
 	.set_clockgating_state = gmc_v11_0_set_clockgating_state,
 	.set_powergating_state = gmc_v11_0_set_powergating_state,
 	.get_clockgating_state = gmc_v11_0_get_clockgating_state,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c
new file mode 100644
index 000000000000..7a9d6894e321
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c
@@ -0,0 +1,1070 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include <linux/firmware.h>
+#include <linux/pci.h>
+
+#include <drm/drm_cache.h>
+
+#include "amdgpu.h"
+#include "amdgpu_atomfirmware.h"
+#include "gmc_v12_0.h"
+#include "athub/athub_4_1_0_sh_mask.h"
+#include "athub/athub_4_1_0_offset.h"
+#include "oss/osssys_7_0_0_offset.h"
+#include "ivsrcid/vmc/irqsrcs_vmc_1_0.h"
+#include "soc24_enum.h"
+#include "soc24.h"
+#include "soc15d.h"
+#include "soc15_common.h"
+#include "nbif_v6_3_1.h"
+#include "gfxhub_v12_0.h"
+#include "mmhub_v4_1_0.h"
+#include "athub_v4_1_0.h"
+#include "umc_v8_14.h"
+
+static int gmc_v12_0_ecc_interrupt_state(struct amdgpu_device *adev,
+					 struct amdgpu_irq_src *src,
+					 unsigned type,
+					 enum amdgpu_interrupt_state state)
+{
+	return 0;
+}
+
+static int gmc_v12_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
+					      struct amdgpu_irq_src *src, unsigned type,
+					      enum amdgpu_interrupt_state state)
+{
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		/* MM HUB */
+		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), false);
+		/* GFX HUB */
+		/* This works because this interrupt is only
+		 * enabled at init/resume and disabled in
+		 * fini/suspend, so the overall state doesn't
+		 * change over the course of suspend/resume.
+		 */
+		if (!adev->in_s0ix)
+			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), false);
+		break;
+	case AMDGPU_IRQ_STATE_ENABLE:
+		/* MM HUB */
+		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), true);
+		/* GFX HUB */
+		/* This works because this interrupt is only
+		 * enabled at init/resume and disabled in
+		 * fini/suspend, so the overall state doesn't
+		 * change over the course of suspend/resume.
+		 */
+		if (!adev->in_s0ix)
+			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), true);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int gmc_v12_0_process_interrupt(struct amdgpu_device *adev,
+				       struct amdgpu_irq_src *source,
+				       struct amdgpu_iv_entry *entry)
+{
+	struct amdgpu_vmhub *hub;
+	bool retry_fault = !!(entry->src_data[1] &
+			      AMDGPU_GMC9_FAULT_SOURCE_DATA_RETRY);
+	bool write_fault = !!(entry->src_data[1] &
+			      AMDGPU_GMC9_FAULT_SOURCE_DATA_WRITE);
+	uint32_t status = 0;
+	u64 addr;
+
+	addr = (u64)entry->src_data[0] << 12;
+	addr |= ((u64)entry->src_data[1] & 0xf) << 44;
+
+	if (entry->client_id == SOC21_IH_CLIENTID_VMC)
+		hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
+	else
+		hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+
+	if (retry_fault) {
+		/* Returning 1 here also prevents sending the IV to the KFD */
+
+		/* Process it only if it's the first fault for this address */
+		if (entry->ih != &adev->irq.ih_soft &&
+		    amdgpu_gmc_filter_faults(adev, entry->ih, addr, entry->pasid,
+					     entry->timestamp))
+			return 1;
+
+		/* Delegate it to a different ring if the hardware hasn't
+		 * already done it.
+		 */
+		if (entry->ih == &adev->irq.ih) {
+			amdgpu_irq_delegate(adev, entry, 8);
+			return 1;
+		}
+
+		/* Try to handle the recoverable page faults by filling page
+		 * tables
+		 */
+		if (amdgpu_vm_handle_fault(adev, entry->pasid, 0, 0, addr,
+					   entry->timestamp, write_fault))
+			return 1;
+	}
+
+	if (!amdgpu_sriov_vf(adev)) {
+		/*
+		 * Issue a dummy read to wait for the status register to
+		 * be updated to avoid reading an incorrect value due to
+		 * the new fast GRBM interface.
+		 */
+		if (entry->vmid_src == AMDGPU_GFXHUB(0))
+			RREG32(hub->vm_l2_pro_fault_status);
+
+		status = RREG32(hub->vm_l2_pro_fault_status);
+		WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1);
+
+		amdgpu_vm_update_fault_cache(adev, entry->pasid, addr, status,
+					     entry->vmid_src ? AMDGPU_MMHUB0(0) : AMDGPU_GFXHUB(0));
+	}
+
+	if (printk_ratelimit()) {
+		struct amdgpu_task_info *task_info;
+
+		dev_err(adev->dev,
+			"[%s] page fault (src_id:%u ring:%u vmid:%u pasid:%u)\n",
+			entry->vmid_src ? "mmhub" : "gfxhub",
+			entry->src_id, entry->ring_id, entry->vmid, entry->pasid);
+		task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid);
+		if (task_info) {
+			amdgpu_vm_print_task_info(adev, task_info);
+			amdgpu_vm_put_task_info(task_info);
+		}
+
+		dev_err(adev->dev, "  in page starting at address 0x%016llx from client %d\n",
+				addr, entry->client_id);
+
+		/* Only print L2 fault status if the status register could be read and
+		 * contains useful information
+		 */
+		if (status != 0)
+			hub->vmhub_funcs->print_l2_protection_fault_status(adev, status);
+	}
+
+	return 0;
+}
+
+static const struct amdgpu_irq_src_funcs gmc_v12_0_irq_funcs = {
+	.set = gmc_v12_0_vm_fault_interrupt_state,
+	.process = gmc_v12_0_process_interrupt,
+};
+
+static const struct amdgpu_irq_src_funcs gmc_v12_0_ecc_funcs = {
+	.set = gmc_v12_0_ecc_interrupt_state,
+	.process = amdgpu_umc_process_ecc_irq,
+};
+
+static void gmc_v12_0_set_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->gmc.vm_fault.num_types = 1;
+	adev->gmc.vm_fault.funcs = &gmc_v12_0_irq_funcs;
+
+	if (!amdgpu_sriov_vf(adev)) {
+		adev->gmc.ecc_irq.num_types = 1;
+		adev->gmc.ecc_irq.funcs = &gmc_v12_0_ecc_funcs;
+	}
+}
+
+/**
+ * gmc_v12_0_use_invalidate_semaphore - judge whether to use semaphore
+ *
+ * @adev: amdgpu_device pointer
+ * @vmhub: vmhub type
+ *
+ */
+static bool gmc_v12_0_use_invalidate_semaphore(struct amdgpu_device *adev,
+				       uint32_t vmhub)
+{
+	return ((vmhub == AMDGPU_MMHUB0(0)) &&
+		(!amdgpu_sriov_vf(adev)));
+}
+
+static bool gmc_v12_0_get_vmid_pasid_mapping_info(
+					struct amdgpu_device *adev,
+					uint8_t vmid, uint16_t *p_pasid)
+{
+	*p_pasid = RREG32(SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT) + vmid) & 0xffff;
+
+	return !!(*p_pasid);
+}
+
+/*
+ * GART
+ * VMID 0 is the physical GPU addresses as used by the kernel.
+ * VMIDs 1-15 are used for userspace clients and are handled
+ * by the amdgpu vm/hsa code.
+ */
+
+static void gmc_v12_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
+				   unsigned int vmhub, uint32_t flush_type)
+{
+	bool use_semaphore = gmc_v12_0_use_invalidate_semaphore(adev, vmhub);
+	struct amdgpu_vmhub *hub = &adev->vmhub[vmhub];
+	u32 inv_req = hub->vmhub_funcs->get_invalidate_req(vmid, flush_type);
+	u32 tmp;
+	/* Use register 17 for GART */
+	const unsigned eng = 17;
+	unsigned int i;
+	unsigned char hub_ip = 0;
+
+	hub_ip = (vmhub == AMDGPU_GFXHUB(0)) ?
+		   GC_HWIP : MMHUB_HWIP;
+
+	spin_lock(&adev->gmc.invalidate_lock);
+	/*
+	 * It may lose gpuvm invalidate acknowldege state across power-gating
+	 * off cycle, add semaphore acquire before invalidation and semaphore
+	 * release after invalidation to avoid entering power gated state
+	 * to WA the Issue
+	 */
+
+	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
+	if (use_semaphore) {
+		for (i = 0; i < adev->usec_timeout; i++) {
+			/* a read return value of 1 means semaphore acuqire */
+			tmp = RREG32_RLC_NO_KIQ(hub->vm_inv_eng0_sem +
+					    hub->eng_distance * eng, hub_ip);
+			if (tmp & 0x1)
+				break;
+			udelay(1);
+		}
+
+		if (i >= adev->usec_timeout)
+			dev_err(adev->dev,
+				"Timeout waiting for sem acquire in VM flush!\n");
+	}
+
+	WREG32_RLC_NO_KIQ(hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req, hub_ip);
+
+	/* Wait for ACK with a delay.*/
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = RREG32_RLC_NO_KIQ(hub->vm_inv_eng0_ack +
+				    hub->eng_distance * eng, hub_ip);
+		tmp &= 1 << vmid;
+		if (tmp)
+			break;
+
+		udelay(1);
+	}
+
+	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
+	if (use_semaphore)
+		/*
+		 * add semaphore release after invalidation,
+		 * write with 0 means semaphore release
+		 */
+		WREG32_RLC_NO_KIQ(hub->vm_inv_eng0_sem +
+			      hub->eng_distance * eng, 0, hub_ip);
+
+	/* Issue additional private vm invalidation to MMHUB */
+	if ((vmhub != AMDGPU_GFXHUB(0)) &&
+	    (hub->vm_l2_bank_select_reserved_cid2) &&
+		!amdgpu_sriov_vf(adev)) {
+		inv_req = RREG32_NO_KIQ(hub->vm_l2_bank_select_reserved_cid2);
+		/* bit 25: RSERVED_CACHE_PRIVATE_INVALIDATION */
+		inv_req |= (1 << 25);
+		/* Issue private invalidation */
+		WREG32_NO_KIQ(hub->vm_l2_bank_select_reserved_cid2, inv_req);
+		/* Read back to ensure invalidation is done*/
+		RREG32_NO_KIQ(hub->vm_l2_bank_select_reserved_cid2);
+	}
+
+	spin_unlock(&adev->gmc.invalidate_lock);
+
+	if (i < adev->usec_timeout)
+		return;
+
+	dev_err(adev->dev, "Timeout waiting for VM flush ACK!\n");
+}
+
+/**
+ * gmc_v12_0_flush_gpu_tlb - gart tlb flush callback
+ *
+ * @adev: amdgpu_device pointer
+ * @vmid: vm instance to flush
+ * @vmhub: which hub to flush
+ * @flush_type: the flush type
+ *
+ * Flush the TLB for the requested page table.
+ */
+static void gmc_v12_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
+					uint32_t vmhub, uint32_t flush_type)
+{
+	if ((vmhub == AMDGPU_GFXHUB(0)) && !adev->gfx.is_poweron)
+		return;
+
+	/* flush hdp cache */
+	amdgpu_device_flush_hdp(adev, NULL);
+
+	/* This is necessary for SRIOV as well as for GFXOFF to function
+	 * properly under bare metal
+	 */
+	if ((adev->gfx.kiq[0].ring.sched.ready || adev->mes.ring[0].sched.ready) &&
+	    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) {
+		struct amdgpu_vmhub *hub = &adev->vmhub[vmhub];
+		const unsigned eng = 17;
+		u32 inv_req = hub->vmhub_funcs->get_invalidate_req(vmid, flush_type);
+		u32 req = hub->vm_inv_eng0_req + hub->eng_distance * eng;
+		u32 ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng;
+
+		amdgpu_gmc_fw_reg_write_reg_wait(adev, req, ack, inv_req,
+				1 << vmid, GET_INST(GC, 0));
+		return;
+	}
+
+	gmc_v12_0_flush_vm_hub(adev, vmid, vmhub, 0);
+	return;
+}
+
+/**
+ * gmc_v12_0_flush_gpu_tlb_pasid - tlb flush via pasid
+ *
+ * @adev: amdgpu_device pointer
+ * @pasid: pasid to be flush
+ * @flush_type: the flush type
+ * @all_hub: flush all hubs
+ * @inst: is used to select which instance of KIQ to use for the invalidation
+ *
+ * Flush the TLB for the requested pasid.
+ */
+static void gmc_v12_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
+					  uint16_t pasid, uint32_t flush_type,
+					  bool all_hub, uint32_t inst)
+{
+	uint16_t queried;
+	int vmid, i;
+
+	if (adev->enable_uni_mes && adev->mes.ring[AMDGPU_MES_SCHED_PIPE].sched.ready &&
+	    (adev->mes.sched_version & AMDGPU_MES_VERSION_MASK) >= 0x84) {
+		struct mes_inv_tlbs_pasid_input input = {0};
+		input.pasid = pasid;
+		input.flush_type = flush_type;
+		input.hub_id = AMDGPU_GFXHUB(0);
+		/* MES will invalidate all gc_hub for the device from master */
+		adev->mes.funcs->invalidate_tlbs_pasid(&adev->mes, &input);
+		if (all_hub) {
+			/* Only need to invalidate mm_hub now, gfx12 only support one mmhub */
+			input.hub_id = AMDGPU_MMHUB0(0);
+			adev->mes.funcs->invalidate_tlbs_pasid(&adev->mes, &input);
+		}
+		return;
+	}
+
+	for (vmid = 1; vmid < 16; vmid++) {
+		bool valid;
+
+		valid = gmc_v12_0_get_vmid_pasid_mapping_info(adev, vmid,
+							      &queried);
+		if (!valid || queried != pasid)
+			continue;
+
+		if (all_hub) {
+			for_each_set_bit(i, adev->vmhubs_mask,
+					 AMDGPU_MAX_VMHUBS)
+				gmc_v12_0_flush_gpu_tlb(adev, vmid, i,
+							flush_type);
+		} else {
+			gmc_v12_0_flush_gpu_tlb(adev, vmid, AMDGPU_GFXHUB(0),
+						flush_type);
+		}
+	}
+}
+
+static uint64_t gmc_v12_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
+					     unsigned vmid, uint64_t pd_addr)
+{
+	bool use_semaphore = gmc_v12_0_use_invalidate_semaphore(ring->adev, ring->vm_hub);
+	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
+	uint32_t req = hub->vmhub_funcs->get_invalidate_req(vmid, 0);
+	unsigned eng = ring->vm_inv_eng;
+
+	/*
+	 * It may lose gpuvm invalidate acknowldege state across power-gating
+	 * off cycle, add semaphore acquire before invalidation and semaphore
+	 * release after invalidation to avoid entering power gated state
+	 * to WA the Issue
+	 */
+
+	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
+	if (use_semaphore)
+		/* a read return value of 1 means semaphore acuqire */
+		amdgpu_ring_emit_reg_wait(ring,
+					  hub->vm_inv_eng0_sem +
+					  hub->eng_distance * eng, 0x1, 0x1);
+
+	amdgpu_ring_emit_wreg(ring, hub->ctx0_ptb_addr_lo32 +
+			      (hub->ctx_addr_distance * vmid),
+			      lower_32_bits(pd_addr));
+
+	amdgpu_ring_emit_wreg(ring, hub->ctx0_ptb_addr_hi32 +
+			      (hub->ctx_addr_distance * vmid),
+			      upper_32_bits(pd_addr));
+
+	amdgpu_ring_emit_reg_write_reg_wait(ring, hub->vm_inv_eng0_req +
+					    hub->eng_distance * eng,
+					    hub->vm_inv_eng0_ack +
+					    hub->eng_distance * eng,
+					    req, 1 << vmid);
+
+	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
+	if (use_semaphore)
+		/*
+		 * add semaphore release after invalidation,
+		 * write with 0 means semaphore release
+		 */
+		amdgpu_ring_emit_wreg(ring, hub->vm_inv_eng0_sem +
+				      hub->eng_distance * eng, 0);
+
+	return pd_addr;
+}
+
+static void gmc_v12_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
+					 unsigned pasid)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t reg;
+
+	if (ring->vm_hub == AMDGPU_GFXHUB(0))
+		reg = SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT) + vmid;
+	else
+		reg = SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT_MM) + vmid;
+
+	amdgpu_ring_emit_wreg(ring, reg, pasid);
+}
+
+/*
+ * PTE format:
+ * 63 P
+ * 62:59 reserved
+ * 58 D
+ * 57 G
+ * 56 T
+ * 55:54 M
+ * 53:52 SW
+ * 51:48 reserved for future
+ * 47:12 4k physical page base address
+ * 11:7 fragment
+ * 6 write
+ * 5 read
+ * 4 exe
+ * 3 Z
+ * 2 snooped
+ * 1 system
+ * 0 valid
+ *
+ * PDE format:
+ * 63 P
+ * 62:58 block fragment size
+ * 57 reserved
+ * 56 A
+ * 55:54 M
+ * 53:52 reserved
+ * 51:48 reserved for future
+ * 47:6 physical base address of PD or PTE
+ * 5:3 reserved
+ * 2 C
+ * 1 system
+ * 0 valid
+ */
+
+static void gmc_v12_0_get_vm_pde(struct amdgpu_device *adev, int level,
+				 uint64_t *addr, uint64_t *flags)
+{
+	if (!(*flags & AMDGPU_PDE_PTE_GFX12) && !(*flags & AMDGPU_PTE_SYSTEM))
+		*addr = adev->vm_manager.vram_base_offset + *addr -
+			adev->gmc.vram_start;
+	BUG_ON(*addr & 0xFFFF00000000003FULL);
+
+	if (!adev->gmc.translate_further)
+		return;
+
+	if (level == AMDGPU_VM_PDB1) {
+		/* Set the block fragment size */
+		if (!(*flags & AMDGPU_PDE_PTE_GFX12))
+			*flags |= AMDGPU_PDE_BFS_GFX12(0x9);
+
+	} else if (level == AMDGPU_VM_PDB0) {
+		if (*flags & AMDGPU_PDE_PTE_GFX12)
+			*flags &= ~AMDGPU_PDE_PTE_GFX12;
+	}
+}
+
+static void gmc_v12_0_get_vm_pte(struct amdgpu_device *adev,
+				 struct amdgpu_vm *vm,
+				 struct amdgpu_bo *bo,
+				 uint32_t vm_flags,
+				 uint64_t *flags)
+{
+	if (vm_flags & AMDGPU_VM_PAGE_EXECUTABLE)
+		*flags |= AMDGPU_PTE_EXECUTABLE;
+	else
+		*flags &= ~AMDGPU_PTE_EXECUTABLE;
+
+	switch (vm_flags & AMDGPU_VM_MTYPE_MASK) {
+	case AMDGPU_VM_MTYPE_DEFAULT:
+		*flags = AMDGPU_PTE_MTYPE_GFX12(*flags, MTYPE_NC);
+		break;
+	case AMDGPU_VM_MTYPE_NC:
+	default:
+		*flags = AMDGPU_PTE_MTYPE_GFX12(*flags, MTYPE_NC);
+		break;
+	case AMDGPU_VM_MTYPE_UC:
+		*flags = AMDGPU_PTE_MTYPE_GFX12(*flags, MTYPE_UC);
+		break;
+	}
+
+	if (vm_flags & AMDGPU_VM_PAGE_NOALLOC)
+		*flags |= AMDGPU_PTE_NOALLOC;
+	else
+		*flags &= ~AMDGPU_PTE_NOALLOC;
+
+	if (vm_flags & AMDGPU_VM_PAGE_PRT) {
+		*flags |= AMDGPU_PTE_PRT_GFX12;
+		*flags |= AMDGPU_PTE_SNOOPED;
+		*flags |= AMDGPU_PTE_SYSTEM;
+		*flags |= AMDGPU_PTE_IS_PTE;
+		*flags &= ~AMDGPU_PTE_VALID;
+	}
+
+	if (bo && bo->flags & AMDGPU_GEM_CREATE_GFX12_DCC)
+		*flags |= AMDGPU_PTE_DCC;
+
+	if (bo && bo->flags & AMDGPU_GEM_CREATE_UNCACHED)
+		*flags = AMDGPU_PTE_MTYPE_GFX12(*flags, MTYPE_UC);
+}
+
+static unsigned gmc_v12_0_get_vbios_fb_size(struct amdgpu_device *adev)
+{
+	return 0;
+}
+
+static unsigned int gmc_v12_0_get_dcc_alignment(struct amdgpu_device *adev)
+{
+	unsigned int max_tex_channel_caches, alignment;
+
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(12, 0, 0) &&
+	    amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(12, 0, 1))
+		return 0;
+
+	max_tex_channel_caches = adev->gfx.config.max_texture_channel_caches;
+	if (is_power_of_2(max_tex_channel_caches))
+		alignment = (unsigned int)(max_tex_channel_caches / SZ_4);
+	else
+		alignment = roundup_pow_of_two(max_tex_channel_caches);
+
+	return (unsigned int)(alignment * max_tex_channel_caches * SZ_1K);
+}
+
+static const struct amdgpu_gmc_funcs gmc_v12_0_gmc_funcs = {
+	.flush_gpu_tlb = gmc_v12_0_flush_gpu_tlb,
+	.flush_gpu_tlb_pasid = gmc_v12_0_flush_gpu_tlb_pasid,
+	.emit_flush_gpu_tlb = gmc_v12_0_emit_flush_gpu_tlb,
+	.emit_pasid_mapping = gmc_v12_0_emit_pasid_mapping,
+	.get_vm_pde = gmc_v12_0_get_vm_pde,
+	.get_vm_pte = gmc_v12_0_get_vm_pte,
+	.get_vbios_fb_size = gmc_v12_0_get_vbios_fb_size,
+	.get_dcc_alignment = gmc_v12_0_get_dcc_alignment,
+};
+
+static void gmc_v12_0_set_gmc_funcs(struct amdgpu_device *adev)
+{
+	adev->gmc.gmc_funcs = &gmc_v12_0_gmc_funcs;
+}
+
+static void gmc_v12_0_set_umc_funcs(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, UMC_HWIP, 0)) {
+	case IP_VERSION(8, 14, 0):
+		adev->umc.channel_inst_num = UMC_V8_14_CHANNEL_INSTANCE_NUM;
+		adev->umc.umc_inst_num = UMC_V8_14_UMC_INSTANCE_NUM(adev);
+		adev->umc.node_inst_num = 0;
+		adev->umc.max_ras_err_cnt_per_query = UMC_V8_14_TOTAL_CHANNEL_NUM(adev);
+		adev->umc.channel_offs = UMC_V8_14_PER_CHANNEL_OFFSET;
+		adev->umc.ras = &umc_v8_14_ras;
+		break;
+	default:
+		break;
+	}
+}
+
+
+static void gmc_v12_0_set_mmhub_funcs(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
+	case IP_VERSION(4, 1, 0):
+		adev->mmhub.funcs = &mmhub_v4_1_0_funcs;
+		break;
+	default:
+		break;
+	}
+}
+
+static void gmc_v12_0_set_gfxhub_funcs(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		adev->gfxhub.funcs = &gfxhub_v12_0_funcs;
+		break;
+	default:
+		break;
+	}
+}
+
+static int gmc_v12_0_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	gmc_v12_0_set_gfxhub_funcs(adev);
+	gmc_v12_0_set_mmhub_funcs(adev);
+	gmc_v12_0_set_gmc_funcs(adev);
+	gmc_v12_0_set_irq_funcs(adev);
+	gmc_v12_0_set_umc_funcs(adev);
+
+	adev->gmc.shared_aperture_start = 0x2000000000000000ULL;
+	adev->gmc.shared_aperture_end =
+		adev->gmc.shared_aperture_start + (4ULL << 30) - 1;
+	adev->gmc.private_aperture_start = 0x1000000000000000ULL;
+	adev->gmc.private_aperture_end =
+		adev->gmc.private_aperture_start + (4ULL << 30) - 1;
+
+	return 0;
+}
+
+static int gmc_v12_0_late_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = amdgpu_gmc_allocate_vm_inv_eng(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_gmc_ras_late_init(adev);
+	if (r)
+		return r;
+
+	return amdgpu_irq_get(adev, &adev->gmc.vm_fault, 0);
+}
+
+static void gmc_v12_0_vram_gtt_location(struct amdgpu_device *adev,
+					struct amdgpu_gmc *mc)
+{
+	u64 base = 0;
+
+	base = adev->mmhub.funcs->get_fb_location(adev);
+
+	amdgpu_gmc_set_agp_default(adev, mc);
+	amdgpu_gmc_vram_location(adev, &adev->gmc, base);
+	amdgpu_gmc_gart_location(adev, mc, AMDGPU_GART_PLACEMENT_LOW);
+	if (!amdgpu_sriov_vf(adev) && (amdgpu_agp == 1))
+		amdgpu_gmc_agp_location(adev, mc);
+
+	/* base offset of vram pages */
+	if (amdgpu_sriov_vf(adev))
+		adev->vm_manager.vram_base_offset = 0;
+	else
+		adev->vm_manager.vram_base_offset = adev->mmhub.funcs->get_mc_fb_offset(adev);
+}
+
+/**
+ * gmc_v12_0_mc_init - initialize the memory controller driver params
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Look up the amount of vram, vram width, and decide how to place
+ * vram and gart within the GPU's physical address space.
+ * Returns 0 for success.
+ */
+static int gmc_v12_0_mc_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	/* size in MB on si */
+	adev->gmc.mc_vram_size =
+		adev->nbio.funcs->get_memsize(adev) * 1024ULL * 1024ULL;
+	adev->gmc.real_vram_size = adev->gmc.mc_vram_size;
+
+	if (!(adev->flags & AMD_IS_APU)) {
+		r = amdgpu_device_resize_fb_bar(adev);
+		if (r)
+			return r;
+	}
+
+	adev->gmc.aper_base = pci_resource_start(adev->pdev, 0);
+	adev->gmc.aper_size = pci_resource_len(adev->pdev, 0);
+
+#ifdef CONFIG_X86_64
+	if ((adev->flags & AMD_IS_APU) && !amdgpu_passthrough(adev)) {
+		adev->gmc.aper_base = adev->mmhub.funcs->get_mc_fb_offset(adev);
+		adev->gmc.aper_size = adev->gmc.real_vram_size;
+	}
+#endif
+	/* In case the PCI BAR is larger than the actual amount of vram */
+	adev->gmc.visible_vram_size = adev->gmc.aper_size;
+	if (adev->gmc.visible_vram_size > adev->gmc.real_vram_size)
+		adev->gmc.visible_vram_size = adev->gmc.real_vram_size;
+
+	/* set the gart size */
+	if (amdgpu_gart_size == -1) {
+		adev->gmc.gart_size = 512ULL << 20;
+	} else
+		adev->gmc.gart_size = (u64)amdgpu_gart_size << 20;
+
+	gmc_v12_0_vram_gtt_location(adev, &adev->gmc);
+
+	return 0;
+}
+
+static int gmc_v12_0_gart_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	if (adev->gart.bo) {
+		WARN(1, "PCIE GART already initialized\n");
+		return 0;
+	}
+
+	/* Initialize common gart structure */
+	r = amdgpu_gart_init(adev);
+	if (r)
+		return r;
+
+	adev->gart.table_size = adev->gart.num_gpu_pages * 8;
+	adev->gart.gart_pte_flags = AMDGPU_PTE_MTYPE_GFX12(0ULL, MTYPE_UC) |
+				    AMDGPU_PTE_EXECUTABLE |
+				    AMDGPU_PTE_IS_PTE;
+
+	return amdgpu_gart_table_vram_alloc(adev);
+}
+
+static int gmc_v12_0_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	int r, vram_width = 0, vram_type = 0, vram_vendor = 0;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	adev->mmhub.funcs->init(adev);
+
+	adev->gfxhub.funcs->init(adev);
+
+	spin_lock_init(&adev->gmc.invalidate_lock);
+
+	r = amdgpu_atomfirmware_get_vram_info(adev,
+					      &vram_width, &vram_type, &vram_vendor);
+	adev->gmc.vram_width = vram_width;
+
+	adev->gmc.vram_type = vram_type;
+	adev->gmc.vram_vendor = vram_vendor;
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
+		set_bit(AMDGPU_MMHUB0(0), adev->vmhubs_mask);
+		/*
+		 * To fulfill 4-level page support,
+		 * vm size is 256TB (48bit), maximum size,
+		 * block size 512 (9bit)
+		 */
+		amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48);
+		break;
+	default:
+		break;
+	}
+
+	/* This interrupt is VMC page fault.*/
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_VMC,
+			      VMC_1_0__SRCID__VM_FAULT,
+			      &adev->gmc.vm_fault);
+
+	if (r)
+		return r;
+
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GFX,
+			      UTCL2_1_0__SRCID__FAULT,
+			      &adev->gmc.vm_fault);
+	if (r)
+		return r;
+
+	if (!amdgpu_sriov_vf(adev)) {
+		/* interrupt sent to DF. */
+		r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_DF, 0,
+				      &adev->gmc.ecc_irq);
+		if (r)
+			return r;
+	}
+
+	/*
+	 * Set the internal MC address mask This is the max address of the GPU's
+	 * internal address space.
+	 */
+	adev->gmc.mc_mask = 0xffffffffffffULL; /* 48 bit MC */
+
+	r = dma_set_mask_and_coherent(adev->dev, DMA_BIT_MASK(44));
+	if (r) {
+		printk(KERN_WARNING "amdgpu: No suitable DMA available.\n");
+		return r;
+	}
+
+	adev->need_swiotlb = drm_need_swiotlb(44);
+
+	r = gmc_v12_0_mc_init(adev);
+	if (r)
+		return r;
+
+	amdgpu_gmc_get_vbios_allocations(adev);
+
+	/* Memory manager */
+	r = amdgpu_bo_init(adev);
+	if (r)
+		return r;
+
+	r = gmc_v12_0_gart_init(adev);
+	if (r)
+		return r;
+
+	/*
+	 * number of VMs
+	 * VMID 0 is reserved for System
+	 * amdgpu graphics/compute will use VMIDs 1-7
+	 * amdkfd will use VMIDs 8-15
+	 */
+	adev->vm_manager.first_kfd_vmid = adev->gfx.disable_kq ? 1 : 8;
+
+	amdgpu_vm_manager_init(adev);
+
+	r = amdgpu_gmc_ras_sw_init(adev);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+/**
+ * gmc_v12_0_gart_fini - vm fini callback
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Tears down the driver GART/VM setup (CIK).
+ */
+static void gmc_v12_0_gart_fini(struct amdgpu_device *adev)
+{
+	amdgpu_gart_table_vram_free(adev);
+}
+
+static int gmc_v12_0_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	amdgpu_vm_manager_fini(adev);
+	gmc_v12_0_gart_fini(adev);
+	amdgpu_gem_force_release(adev);
+	amdgpu_bo_fini(adev);
+
+	return 0;
+}
+
+static void gmc_v12_0_init_golden_registers(struct amdgpu_device *adev)
+{
+}
+
+/**
+ * gmc_v12_0_gart_enable - gart enable
+ *
+ * @adev: amdgpu_device pointer
+ */
+static int gmc_v12_0_gart_enable(struct amdgpu_device *adev)
+{
+	int r;
+	bool value;
+
+	if (adev->gart.bo == NULL) {
+		dev_err(adev->dev, "No VRAM object for PCIE GART.\n");
+		return -EINVAL;
+	}
+
+	amdgpu_gtt_mgr_recover(&adev->mman.gtt_mgr);
+
+	r = adev->mmhub.funcs->gart_enable(adev);
+	if (r)
+		return r;
+
+	/* Flush HDP after it is initialized */
+	amdgpu_device_flush_hdp(adev, NULL);
+
+	value = amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS;
+
+	adev->mmhub.funcs->set_fault_enable_default(adev, value);
+	gmc_v12_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB0(0), 0);
+
+	dev_info(adev->dev, "PCIE GART of %uM enabled (table at 0x%016llX).\n",
+		 (unsigned)(adev->gmc.gart_size >> 20),
+		 (unsigned long long)amdgpu_bo_gpu_offset(adev->gart.bo));
+
+	return 0;
+}
+
+static int gmc_v12_0_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	/* The sequence of these two function calls matters.*/
+	gmc_v12_0_init_golden_registers(adev);
+
+	r = gmc_v12_0_gart_enable(adev);
+	if (r)
+		return r;
+
+	if (adev->umc.funcs && adev->umc.funcs->init_registers)
+		adev->umc.funcs->init_registers(adev);
+
+	return 0;
+}
+
+/**
+ * gmc_v12_0_gart_disable - gart disable
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * This disables all VM page table.
+ */
+static void gmc_v12_0_gart_disable(struct amdgpu_device *adev)
+{
+	adev->mmhub.funcs->gart_disable(adev);
+}
+
+static int gmc_v12_0_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (amdgpu_sriov_vf(adev)) {
+		/* full access mode, so don't touch any GMC register */
+		DRM_DEBUG("For SRIOV client, shouldn't do anything.\n");
+		return 0;
+	}
+
+	amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
+
+	if (adev->gmc.ecc_irq.funcs &&
+		amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC))
+		amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0);
+
+	gmc_v12_0_gart_disable(adev);
+
+	return 0;
+}
+
+static int gmc_v12_0_suspend(struct amdgpu_ip_block *ip_block)
+{
+	gmc_v12_0_hw_fini(ip_block);
+
+	return 0;
+}
+
+static int gmc_v12_0_resume(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+
+	r = gmc_v12_0_hw_init(ip_block);
+	if (r)
+		return r;
+
+	amdgpu_vmid_reset_all(ip_block->adev);
+
+	return 0;
+}
+
+static bool gmc_v12_0_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	/* MC is always ready in GMC v11.*/
+	return true;
+}
+
+static int gmc_v12_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	/* There is no need to wait for MC idle in GMC v11.*/
+	return 0;
+}
+
+static int gmc_v12_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					   enum amd_clockgating_state state)
+{
+	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	r = adev->mmhub.funcs->set_clockgating(adev, state);
+	if (r)
+		return r;
+
+	return athub_v4_1_0_set_clockgating(adev, state);
+}
+
+static void gmc_v12_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	adev->mmhub.funcs->get_clockgating(adev, flags);
+
+	athub_v4_1_0_get_clockgating(adev, flags);
+}
+
+static int gmc_v12_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					   enum amd_powergating_state state)
+{
+	return 0;
+}
+
+const struct amd_ip_funcs gmc_v12_0_ip_funcs = {
+	.name = "gmc_v12_0",
+	.early_init = gmc_v12_0_early_init,
+	.sw_init = gmc_v12_0_sw_init,
+	.hw_init = gmc_v12_0_hw_init,
+	.late_init = gmc_v12_0_late_init,
+	.sw_fini = gmc_v12_0_sw_fini,
+	.hw_fini = gmc_v12_0_hw_fini,
+	.suspend = gmc_v12_0_suspend,
+	.resume = gmc_v12_0_resume,
+	.is_idle = gmc_v12_0_is_idle,
+	.wait_for_idle = gmc_v12_0_wait_for_idle,
+	.set_clockgating_state = gmc_v12_0_set_clockgating_state,
+	.set_powergating_state = gmc_v12_0_set_powergating_state,
+	.get_clockgating_state = gmc_v12_0_get_clockgating_state,
+};
+
+const struct amdgpu_ip_block_version gmc_v12_0_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_GMC,
+	.major = 12,
+	.minor = 0,
+	.rev = 0,
+	.funcs = &gmc_v12_0_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.h b/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.h
new file mode 100644
index 000000000000..deca93e4a156
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.h
@@ -0,0 +1,30 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __GMC_V12_0_H__
+#define __GMC_V12_0_H__
+
+extern const struct amd_ip_funcs gmc_v12_0_ip_funcs;
+extern const struct amdgpu_ip_block_version gmc_v12_0_ip_block;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
index b7dad4e67813..a8ec95f42926 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
@@ -43,7 +43,7 @@
 
 static void gmc_v6_0_set_gmc_funcs(struct amdgpu_device *adev);
 static void gmc_v6_0_set_irq_funcs(struct amdgpu_device *adev);
-static int gmc_v6_0_wait_for_idle(void *handle);
+static int gmc_v6_0_wait_for_idle(struct amdgpu_ip_block *ip_block);
 
 MODULE_FIRMWARE("amdgpu/tahiti_mc.bin");
 MODULE_FIRMWARE("amdgpu/pitcairn_mc.bin");
@@ -64,8 +64,13 @@ MODULE_FIRMWARE("amdgpu/si58_mc.bin");
 static void gmc_v6_0_mc_stop(struct amdgpu_device *adev)
 {
 	u32 blackout;
+	struct amdgpu_ip_block *ip_block;
 
-	gmc_v6_0_wait_for_idle((void *)adev);
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GMC);
+	if (!ip_block)
+		return;
+
+	gmc_v6_0_wait_for_idle(ip_block);
 
 	blackout = RREG32(mmMC_SHARED_BLACKOUT_CNTL);
 	if (REG_GET_FIELD(blackout, MC_SHARED_BLACKOUT_CNTL, BLACKOUT_MODE) != 1) {
@@ -98,9 +103,7 @@ static void gmc_v6_0_mc_resume(struct amdgpu_device *adev)
 static int gmc_v6_0_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[30];
 	int err;
-	bool is_58_fw = false;
 
 	DRM_DEBUG("\n");
 
@@ -120,22 +123,20 @@ static int gmc_v6_0_init_microcode(struct amdgpu_device *adev)
 	case CHIP_HAINAN:
 		chip_name = "hainan";
 		break;
-	default: BUG();
+	default:
+		BUG();
 	}
 
 	/* this memory configuration requires special firmware */
 	if (((RREG32(mmMC_SEQ_MISC0) & 0xff000000) >> 24) == 0x58)
-		is_58_fw = true;
+		chip_name = "si58";
 
-	if (is_58_fw)
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/si58_mc.bin");
-	else
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mc.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gmc.fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gmc.fw, AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_mc.bin", chip_name);
 	if (err) {
 		dev_err(adev->dev,
-		       "si_mc: Failed to load firmware \"%s\"\n",
-		       fw_name);
+		       "si_mc: Failed to load firmware \"%s_mc.bin\"\n",
+		       chip_name);
 		amdgpu_ucode_release(&adev->gmc.fw);
 	}
 	return err;
@@ -178,9 +179,8 @@ static int gmc_v6_0_mc_load_microcode(struct amdgpu_device *adev)
 			WREG32(mmMC_SEQ_IO_DEBUG_DATA, le32_to_cpup(new_io_mc_regs++));
 		}
 		/* load the MC ucode */
-		for (i = 0; i < ucode_size; i++) {
+		for (i = 0; i < ucode_size; i++)
 			WREG32(mmMC_SEQ_SUP_PGM, le32_to_cpup(new_fw_data++));
-		}
 
 		/* put the engine back into the active state */
 		WREG32(mmMC_SEQ_SUP_CNTL, 0x00000008);
@@ -208,15 +208,19 @@ static void gmc_v6_0_vram_gtt_location(struct amdgpu_device *adev,
 				       struct amdgpu_gmc *mc)
 {
 	u64 base = RREG32(mmMC_VM_FB_LOCATION) & 0xFFFF;
+
 	base <<= 24;
 
+	amdgpu_gmc_set_agp_default(adev, mc);
 	amdgpu_gmc_vram_location(adev, mc, base);
-	amdgpu_gmc_gart_location(adev, mc);
+	amdgpu_gmc_gart_location(adev, mc, AMDGPU_GART_PLACEMENT_LOW);
 }
 
 static void gmc_v6_0_mc_program(struct amdgpu_device *adev)
 {
 	int i, j;
+	struct amdgpu_ip_block *ip_block;
+
 
 	/* Initialize HDP */
 	for (i = 0, j = 0; i < 32; i++, j += 0x6) {
@@ -228,9 +232,12 @@ static void gmc_v6_0_mc_program(struct amdgpu_device *adev)
 	}
 	WREG32(mmHDP_REG_COHERENCY_FLUSH_CNTL, 0);
 
-	if (gmc_v6_0_wait_for_idle((void *)adev)) {
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GMC);
+	if (!ip_block)
+		return;
+
+	if (gmc_v6_0_wait_for_idle(ip_block))
 		dev_warn(adev->dev, "Wait for MC idle timedout !\n");
-	}
 
 	if (adev->mode_info.num_crtc) {
 		u32 tmp;
@@ -242,7 +249,7 @@ static void gmc_v6_0_mc_program(struct amdgpu_device *adev)
 
 		/* disable VGA render */
 		tmp = RREG32(mmVGA_RENDER_CONTROL);
-		tmp &= ~VGA_VSTATUS_CNTL;
+		tmp &= VGA_RENDER_CONTROL__VGA_VSTATUS_CNTL_MASK;
 		WREG32(mmVGA_RENDER_CONTROL, tmp);
 	}
 	/* Update configuration */
@@ -253,12 +260,11 @@ static void gmc_v6_0_mc_program(struct amdgpu_device *adev)
 	WREG32(mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR,
 	       adev->mem_scratch.gpu_addr >> 12);
 	WREG32(mmMC_VM_AGP_BASE, 0);
-	WREG32(mmMC_VM_AGP_TOP, 0x0FFFFFFF);
-	WREG32(mmMC_VM_AGP_BOT, 0x0FFFFFFF);
+	WREG32(mmMC_VM_AGP_TOP, adev->gmc.agp_end >> 22);
+	WREG32(mmMC_VM_AGP_BOT, adev->gmc.agp_start >> 22);
 
-	if (gmc_v6_0_wait_for_idle((void *)adev)) {
+	if (gmc_v6_0_wait_for_idle(ip_block))
 		dev_warn(adev->dev, "Wait for MC idle timedout !\n");
-	}
 }
 
 static int gmc_v6_0_mc_init(struct amdgpu_device *adev)
@@ -269,13 +275,13 @@ static int gmc_v6_0_mc_init(struct amdgpu_device *adev)
 	int r;
 
 	tmp = RREG32(mmMC_ARB_RAMCFG);
-	if (tmp & (1 << 11)) {
+	if (tmp & (1 << 11))
 		chansize = 16;
-	} else if (tmp & MC_ARB_RAMCFG__CHANSIZE_MASK) {
+	else if (tmp & MC_ARB_RAMCFG__CHANSIZE_MASK)
 		chansize = 64;
-	} else {
+	else
 		chansize = 32;
-	}
+
 	tmp = RREG32(mmMC_SHARED_CHMAP);
 	switch ((tmp & MC_SHARED_CHMAP__NOOFCHAN_MASK) >> MC_SHARED_CHMAP__NOOFCHAN__SHIFT) {
 	case 0:
@@ -352,7 +358,7 @@ static void gmc_v6_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 }
 
 static uint64_t gmc_v6_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
-					    unsigned vmid, uint64_t pd_addr)
+					    unsigned int vmid, uint64_t pd_addr)
 {
 	uint32_t reg;
 
@@ -376,7 +382,9 @@ static void gmc_v6_0_get_vm_pde(struct amdgpu_device *adev, int level,
 }
 
 static void gmc_v6_0_get_vm_pte(struct amdgpu_device *adev,
-				struct amdgpu_bo_va_mapping *mapping,
+				struct amdgpu_vm *vm,
+				struct amdgpu_bo *bo,
+				uint32_t vm_flags,
 				uint64_t *flags)
 {
 	*flags &= ~AMDGPU_PTE_EXECUTABLE;
@@ -405,11 +413,11 @@ static void gmc_v6_0_set_fault_enable_default(struct amdgpu_device *adev,
 }
 
  /**
-   + * gmc_v8_0_set_prt - set PRT VM fault
-   + *
-   + * @adev: amdgpu_device pointer
-   + * @enable: enable/disable VM fault handling for PRT
-   +*/
+  * gmc_v8_0_set_prt() - set PRT VM fault
+  *
+  * @adev: amdgpu_device pointer
+  * @enable: enable/disable VM fault handling for PRT
+  */
 static void gmc_v6_0_set_prt(struct amdgpu_device *adev, bool enable)
 {
 	u32 tmp;
@@ -435,9 +443,10 @@ static void gmc_v6_0_set_prt(struct amdgpu_device *adev, bool enable)
 	WREG32(mmVM_PRT_CNTL, tmp);
 
 	if (enable) {
-		uint32_t low = AMDGPU_VA_RESERVED_SIZE >> AMDGPU_GPU_PAGE_SHIFT;
+		uint32_t low = AMDGPU_VA_RESERVED_BOTTOM >>
+			AMDGPU_GPU_PAGE_SHIFT;
 		uint32_t high = adev->vm_manager.max_pfn -
-			(AMDGPU_VA_RESERVED_SIZE >> AMDGPU_GPU_PAGE_SHIFT);
+			(AMDGPU_VA_RESERVED_TOP >> AMDGPU_GPU_PAGE_SHIFT);
 
 		WREG32(mmVM_PRT_APERTURE0_LOW_ADDR, low);
 		WREG32(mmVM_PRT_APERTURE1_LOW_ADDR, low);
@@ -547,7 +556,7 @@ static int gmc_v6_0_gart_enable(struct amdgpu_device *adev)
 
 	gmc_v6_0_flush_gpu_tlb(adev, 0, 0, 0);
 	dev_info(adev->dev, "PCIE GART of %uM enabled (table at 0x%016llX).\n",
-		 (unsigned)(adev->gmc.gart_size >> 20),
+		 (unsigned int)(adev->gmc.gart_size >> 20),
 		 (unsigned long long)table_addr);
 	return 0;
 }
@@ -601,36 +610,33 @@ static void gmc_v6_0_gart_disable(struct amdgpu_device *adev)
 }
 
 static void gmc_v6_0_vm_decode_fault(struct amdgpu_device *adev,
-				     u32 status, u32 addr, u32 mc_client)
+				     u32 status, u32 addr)
 {
 	u32 mc_id;
 	u32 vmid = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS, VMID);
 	u32 protections = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
 					PROTECTIONS);
-	char block[5] = { mc_client >> 24, (mc_client >> 16) & 0xff,
-		(mc_client >> 8) & 0xff, mc_client & 0xff, 0 };
 
 	mc_id = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
 			      MEMORY_CLIENT_ID);
 
-	dev_err(adev->dev, "VM fault (0x%02x, vmid %d) at page %u, %s from '%s' (0x%08x) (%d)\n",
+	dev_err(adev->dev, "VM fault (0x%02x, vmid %d) at page %u, %s from %d\n",
 	       protections, vmid, addr,
 	       REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
 			     MEMORY_CLIENT_RW) ?
-	       "write" : "read", block, mc_client, mc_id);
+	       "write" : "read", mc_id);
 }
 
-/*
 static const u32 mc_cg_registers[] = {
-	MC_HUB_MISC_HUB_CG,
-	MC_HUB_MISC_SIP_CG,
-	MC_HUB_MISC_VM_CG,
-	MC_XPB_CLK_GAT,
-	ATC_MISC_CG,
-	MC_CITF_MISC_WR_CG,
-	MC_CITF_MISC_RD_CG,
-	MC_CITF_MISC_VM_CG,
-	VM_L2_CG,
+	mmMC_HUB_MISC_HUB_CG,
+	mmMC_HUB_MISC_SIP_CG,
+	mmMC_HUB_MISC_VM_CG,
+	mmMC_XPB_CLK_GAT,
+	mmATC_MISC_CG,
+	mmMC_CITF_MISC_WR_CG,
+	mmMC_CITF_MISC_RD_CG,
+	mmMC_CITF_MISC_VM_CG,
+	mmVM_L2_CG,
 };
 
 static const u32 mc_cg_ls_en[] = {
@@ -665,7 +671,7 @@ static void gmc_v6_0_enable_mc_ls(struct amdgpu_device *adev,
 
 	for (i = 0; i < ARRAY_SIZE(mc_cg_registers); i++) {
 		orig = data = RREG32(mc_cg_registers[i]);
-		if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_MC_LS))
+		if (enable && (adev->cg_flags & AMD_CG_SUPPORT_MC_LS))
 			data |= mc_cg_ls_en[i];
 		else
 			data &= ~mc_cg_ls_en[i];
@@ -682,7 +688,7 @@ static void gmc_v6_0_enable_mc_mgcg(struct amdgpu_device *adev,
 
 	for (i = 0; i < ARRAY_SIZE(mc_cg_registers); i++) {
 		orig = data = RREG32(mc_cg_registers[i]);
-		if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_MC_MGCG))
+		if (enable && (adev->cg_flags & AMD_CG_SUPPORT_MC_MGCG))
 			data |= mc_cg_en[i];
 		else
 			data &= ~mc_cg_en[i];
@@ -698,7 +704,7 @@ static void gmc_v6_0_enable_bif_mgls(struct amdgpu_device *adev,
 
 	orig = data = RREG32_PCIE(ixPCIE_CNTL2);
 
-	if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_BIF_LS)) {
+	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_LS)) {
 		data = REG_SET_FIELD(data, PCIE_CNTL2, SLV_MEM_LS_EN, 1);
 		data = REG_SET_FIELD(data, PCIE_CNTL2, MST_MEM_LS_EN, 1);
 		data = REG_SET_FIELD(data, PCIE_CNTL2, REPLAY_MEM_LS_EN, 1);
@@ -721,7 +727,7 @@ static void gmc_v6_0_enable_hdp_mgcg(struct amdgpu_device *adev,
 
 	orig = data = RREG32(mmHDP_HOST_PATH_CNTL);
 
-	if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_HDP_MGCG))
+	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_HDP_MGCG))
 		data = REG_SET_FIELD(data, HDP_HOST_PATH_CNTL, CLOCK_GATING_DIS, 0);
 	else
 		data = REG_SET_FIELD(data, HDP_HOST_PATH_CNTL, CLOCK_GATING_DIS, 1);
@@ -737,7 +743,7 @@ static void gmc_v6_0_enable_hdp_ls(struct amdgpu_device *adev,
 
 	orig = data = RREG32(mmHDP_MEM_POWER_LS);
 
-	if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_HDP_LS))
+	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_HDP_LS))
 		data = REG_SET_FIELD(data, HDP_MEM_POWER_LS, LS_ENABLE, 1);
 	else
 		data = REG_SET_FIELD(data, HDP_MEM_POWER_LS, LS_ENABLE, 0);
@@ -745,7 +751,6 @@ static void gmc_v6_0_enable_hdp_ls(struct amdgpu_device *adev,
 	if (orig != data)
 		WREG32(mmHDP_MEM_POWER_LS, data);
 }
-*/
 
 static int gmc_v6_0_convert_vram_type(int mc_seq_vram_type)
 {
@@ -767,9 +772,9 @@ static int gmc_v6_0_convert_vram_type(int mc_seq_vram_type)
 	}
 }
 
-static int gmc_v6_0_early_init(void *handle)
+static int gmc_v6_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v6_0_set_gmc_funcs(adev);
 	gmc_v6_0_set_irq_funcs(adev);
@@ -777,9 +782,9 @@ static int gmc_v6_0_early_init(void *handle)
 	return 0;
 }
 
-static int gmc_v6_0_late_init(void *handle)
+static int gmc_v6_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS)
 		return amdgpu_irq_get(adev, &adev->gmc.vm_fault, 0);
@@ -787,15 +792,16 @@ static int gmc_v6_0_late_init(void *handle)
 		return 0;
 }
 
-static unsigned gmc_v6_0_get_vbios_fb_size(struct amdgpu_device *adev)
+static unsigned int gmc_v6_0_get_vbios_fb_size(struct amdgpu_device *adev)
 {
 	u32 d1vga_control = RREG32(mmD1VGA_CONTROL);
-	unsigned size;
+	unsigned int size;
 
 	if (REG_GET_FIELD(d1vga_control, D1VGA_CONTROL, D1VGA_MODE_ENABLE)) {
 		size = AMDGPU_VBIOS_VGA_ALLOCATION;
 	} else {
 		u32 viewport = RREG32(mmVIEWPORT_SIZE);
+
 		size = (REG_GET_FIELD(viewport, VIEWPORT_SIZE, VIEWPORT_HEIGHT) *
 			REG_GET_FIELD(viewport, VIEWPORT_SIZE, VIEWPORT_WIDTH) *
 			4);
@@ -803,17 +809,18 @@ static unsigned gmc_v6_0_get_vbios_fb_size(struct amdgpu_device *adev)
 	return size;
 }
 
-static int gmc_v6_0_sw_init(void *handle)
+static int gmc_v6_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	adev->num_vmhubs = 1;
+	set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
 
 	if (adev->flags & AMD_IS_APU) {
 		adev->gmc.vram_type = AMDGPU_VRAM_TYPE_UNKNOWN;
 	} else {
 		u32 tmp = RREG32(mmMC_SEQ_MISC0);
+
 		tmp &= MC_SEQ_MISC0__MT__MASK;
 		adev->gmc.vram_type = gmc_v6_0_convert_vram_type(tmp);
 	}
@@ -879,9 +886,9 @@ static int gmc_v6_0_sw_init(void *handle)
 	return 0;
 }
 
-static int gmc_v6_0_sw_fini(void *handle)
+static int gmc_v6_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_gem_force_release(adev);
 	amdgpu_vm_manager_fini(adev);
@@ -892,10 +899,10 @@ static int gmc_v6_0_sw_fini(void *handle)
 	return 0;
 }
 
-static int gmc_v6_0_hw_init(void *handle)
+static int gmc_v6_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v6_0_mc_program(adev);
 
@@ -913,13 +920,13 @@ static int gmc_v6_0_hw_init(void *handle)
 
 	if (amdgpu_emu_mode == 1)
 		return amdgpu_gmc_vram_checking(adev);
-	else
-		return r;
+
+	return 0;
 }
 
-static int gmc_v6_0_hw_fini(void *handle)
+static int gmc_v6_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
 	gmc_v6_0_gart_disable(adev);
@@ -927,21 +934,19 @@ static int gmc_v6_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int gmc_v6_0_suspend(void *handle)
+static int gmc_v6_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	gmc_v6_0_hw_fini(adev);
+	gmc_v6_0_hw_fini(ip_block);
 
 	return 0;
 }
 
-static int gmc_v6_0_resume(void *handle)
+static int gmc_v6_0_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	r = gmc_v6_0_hw_init(adev);
+	r = gmc_v6_0_hw_init(ip_block);
 	if (r)
 		return r;
 
@@ -950,9 +955,10 @@ static int gmc_v6_0_resume(void *handle)
 	return 0;
 }
 
-static bool gmc_v6_0_is_idle(void *handle)
+static bool gmc_v6_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (tmp & (SRBM_STATUS__MCB_BUSY_MASK | SRBM_STATUS__MCB_NON_DISPLAY_BUSY_MASK |
@@ -962,13 +968,13 @@ static bool gmc_v6_0_is_idle(void *handle)
 	return true;
 }
 
-static int gmc_v6_0_wait_for_idle(void *handle)
+static int gmc_v6_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
-	unsigned i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	unsigned int i;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		if (gmc_v6_0_is_idle(handle))
+		if (gmc_v6_0_is_idle(ip_block))
 			return 0;
 		udelay(1);
 	}
@@ -976,9 +982,10 @@ static int gmc_v6_0_wait_for_idle(void *handle)
 
 }
 
-static int gmc_v6_0_soft_reset(void *handle)
+static int gmc_v6_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+
 	u32 srbm_soft_reset = 0;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
@@ -995,10 +1002,9 @@ static int gmc_v6_0_soft_reset(void *handle)
 
 	if (srbm_soft_reset) {
 		gmc_v6_0_mc_stop(adev);
-		if (gmc_v6_0_wait_for_idle(adev)) {
-			dev_warn(adev->dev, "Wait for GMC idle timed out !\n");
-		}
 
+		if (gmc_v6_0_wait_for_idle(ip_block))
+			dev_warn(adev->dev, "Wait for GMC idle timed out !\n");
 
 		tmp = RREG32(mmSRBM_SOFT_RESET);
 		tmp |= srbm_soft_reset;
@@ -1023,7 +1029,7 @@ static int gmc_v6_0_soft_reset(void *handle)
 
 static int gmc_v6_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 					     struct amdgpu_irq_src *src,
-					     unsigned type,
+					     unsigned int type,
 					     enum amdgpu_interrupt_state state)
 {
 	u32 tmp;
@@ -1064,6 +1070,12 @@ static int gmc_v6_0_process_interrupt(struct amdgpu_device *adev,
 {
 	u32 addr, status;
 
+	/* Delegate to the soft IRQ handler ring */
+	if (adev->irq.ih_soft.enabled && entry->ih != &adev->irq.ih_soft) {
+		amdgpu_irq_delegate(adev, entry, 4);
+		return 1;
+	}
+
 	addr = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_ADDR);
 	status = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_STATUS);
 	WREG32_P(mmVM_CONTEXT1_CNTL2, 1, ~1);
@@ -1071,6 +1083,10 @@ static int gmc_v6_0_process_interrupt(struct amdgpu_device *adev,
 	if (!addr && !status)
 		return 0;
 
+	amdgpu_vm_update_fault_cache(adev, entry->pasid,
+				     ((u64)addr) << AMDGPU_GPU_PAGE_SHIFT,
+				     status, AMDGPU_GFXHUB(0));
+
 	if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_FIRST)
 		gmc_v6_0_set_fault_enable_default(adev, false);
 
@@ -1081,19 +1097,33 @@ static int gmc_v6_0_process_interrupt(struct amdgpu_device *adev,
 			addr);
 		dev_err(adev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n",
 			status);
-		gmc_v6_0_vm_decode_fault(adev, status, addr, 0);
+		gmc_v6_0_vm_decode_fault(adev, status, addr);
 	}
 
 	return 0;
 }
 
-static int gmc_v6_0_set_clockgating_state(void *handle,
+static int gmc_v6_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
+	struct amdgpu_device *adev = ip_block->adev;
+	bool gate = false;
+
+	if (state == AMD_CG_STATE_GATE)
+		gate = true;
+
+	if (!(adev->flags & AMD_IS_APU)) {
+		gmc_v6_0_enable_mc_mgcg(adev, gate);
+		gmc_v6_0_enable_mc_ls(adev, gate);
+	}
+	gmc_v6_0_enable_bif_mgls(adev, gate);
+	gmc_v6_0_enable_hdp_mgcg(adev, gate);
+	gmc_v6_0_enable_hdp_ls(adev, gate);
+
 	return 0;
 }
 
-static int gmc_v6_0_set_powergating_state(void *handle,
+static int gmc_v6_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -1141,8 +1171,7 @@ static void gmc_v6_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->gmc.vm_fault.funcs = &gmc_v6_0_irq_funcs;
 }
 
-const struct amdgpu_ip_block_version gmc_v6_0_ip_block =
-{
+const struct amdgpu_ip_block_version gmc_v6_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GMC,
 	.major = 6,
 	.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
index 402960b0174e..fbd0bf147f50 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
@@ -52,22 +52,20 @@
 
 static void gmc_v7_0_set_gmc_funcs(struct amdgpu_device *adev);
 static void gmc_v7_0_set_irq_funcs(struct amdgpu_device *adev);
-static int gmc_v7_0_wait_for_idle(void *handle);
+static int gmc_v7_0_wait_for_idle(struct amdgpu_ip_block *ip_block);
 
 MODULE_FIRMWARE("amdgpu/bonaire_mc.bin");
 MODULE_FIRMWARE("amdgpu/hawaii_mc.bin");
 MODULE_FIRMWARE("amdgpu/topaz_mc.bin");
 
-static const u32 golden_settings_iceland_a11[] =
-{
+static const u32 golden_settings_iceland_a11[] = {
 	mmVM_PRT_APERTURE0_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE1_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE2_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE3_LOW_ADDR, 0x0fffffff, 0x0fffffff
 };
 
-static const u32 iceland_mgcg_cgcg_init[] =
-{
+static const u32 iceland_mgcg_cgcg_init[] = {
 	mmMC_MEM_POWER_LS, 0xffffffff, 0x00000104
 };
 
@@ -89,9 +87,14 @@ static void gmc_v7_0_init_golden_registers(struct amdgpu_device *adev)
 
 static void gmc_v7_0_mc_stop(struct amdgpu_device *adev)
 {
+	struct amdgpu_ip_block *ip_block;
 	u32 blackout;
 
-	gmc_v7_0_wait_for_idle((void *)adev);
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GMC);
+	if (!ip_block)
+		return;
+
+	gmc_v7_0_wait_for_idle(ip_block);
 
 	blackout = RREG32(mmMC_SHARED_BLACKOUT_CNTL);
 	if (REG_GET_FIELD(blackout, MC_SHARED_BLACKOUT_CNTL, BLACKOUT_MODE) != 1) {
@@ -132,7 +135,6 @@ static void gmc_v7_0_mc_resume(struct amdgpu_device *adev)
 static int gmc_v7_0_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[30];
 	int err;
 
 	DRM_DEBUG("\n");
@@ -151,14 +153,14 @@ static int gmc_v7_0_init_microcode(struct amdgpu_device *adev)
 	case CHIP_KABINI:
 	case CHIP_MULLINS:
 		return 0;
-	default: BUG();
+	default:
+		return -EINVAL;
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mc.bin", chip_name);
-
-	err = amdgpu_ucode_request(adev, &adev->gmc.fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gmc.fw, AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_mc.bin", chip_name);
 	if (err) {
-		pr_err("cik_mc: Failed to load firmware \"%s\"\n", fw_name);
+		pr_err("cik_mc: Failed to load firmware \"%s_mc.bin\"\n", chip_name);
 		amdgpu_ucode_release(&adev->gmc.fw);
 	}
 	return err;
@@ -237,10 +239,12 @@ static void gmc_v7_0_vram_gtt_location(struct amdgpu_device *adev,
 				       struct amdgpu_gmc *mc)
 {
 	u64 base = RREG32(mmMC_VM_FB_LOCATION) & 0xFFFF;
+
 	base <<= 24;
 
+	amdgpu_gmc_set_agp_default(adev, mc);
 	amdgpu_gmc_vram_location(adev, mc, base);
-	amdgpu_gmc_gart_location(adev, mc);
+	amdgpu_gmc_gart_location(adev, mc, AMDGPU_GART_PLACEMENT_BEST_FIT);
 }
 
 /**
@@ -253,9 +257,14 @@ static void gmc_v7_0_vram_gtt_location(struct amdgpu_device *adev,
  */
 static void gmc_v7_0_mc_program(struct amdgpu_device *adev)
 {
+	struct amdgpu_ip_block *ip_block;
 	u32 tmp;
 	int i, j;
 
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GMC);
+	if (!ip_block)
+		return;
+
 	/* Initialize HDP */
 	for (i = 0, j = 0; i < 32; i++, j += 0x6) {
 		WREG32((0xb05 + j), 0x00000000);
@@ -266,9 +275,9 @@ static void gmc_v7_0_mc_program(struct amdgpu_device *adev)
 	}
 	WREG32(mmHDP_REG_COHERENCY_FLUSH_CNTL, 0);
 
-	if (gmc_v7_0_wait_for_idle((void *)adev)) {
+	if (gmc_v7_0_wait_for_idle(ip_block))
 		dev_warn(adev->dev, "Wait for MC idle timedout !\n");
-	}
+
 	if (adev->mode_info.num_crtc) {
 		/* Lockout access through VGA aperture*/
 		tmp = RREG32(mmVGA_HDP_CONTROL);
@@ -288,11 +297,10 @@ static void gmc_v7_0_mc_program(struct amdgpu_device *adev)
 	WREG32(mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR,
 	       adev->mem_scratch.gpu_addr >> 12);
 	WREG32(mmMC_VM_AGP_BASE, 0);
-	WREG32(mmMC_VM_AGP_TOP, 0x0FFFFFFF);
-	WREG32(mmMC_VM_AGP_BOT, 0x0FFFFFFF);
-	if (gmc_v7_0_wait_for_idle((void *)adev)) {
+	WREG32(mmMC_VM_AGP_TOP, adev->gmc.agp_end >> 22);
+	WREG32(mmMC_VM_AGP_BOT, adev->gmc.agp_start >> 22);
+	if (gmc_v7_0_wait_for_idle(ip_block))
 		dev_warn(adev->dev, "Wait for MC idle timedout !\n");
-	}
 
 	WREG32(mmBIF_FB_EN, BIF_FB_EN__FB_READ_EN_MASK | BIF_FB_EN__FB_WRITE_EN_MASK);
 
@@ -324,11 +332,11 @@ static int gmc_v7_0_mc_init(struct amdgpu_device *adev)
 
 		/* Get VRAM informations */
 		tmp = RREG32(mmMC_ARB_RAMCFG);
-		if (REG_GET_FIELD(tmp, MC_ARB_RAMCFG, CHANSIZE)) {
+		if (REG_GET_FIELD(tmp, MC_ARB_RAMCFG, CHANSIZE))
 			chansize = 64;
-		} else {
+		else
 			chansize = 32;
-		}
+
 		tmp = RREG32(mmMC_SHARED_CHMAP);
 		switch (REG_GET_FIELD(tmp, MC_SHARED_CHMAP, NOOFCHAN)) {
 		case 0:
@@ -419,31 +427,27 @@ static int gmc_v7_0_mc_init(struct amdgpu_device *adev)
  * @pasid: pasid to be flush
  * @flush_type: type of flush
  * @all_hub: flush all hubs
+ * @inst: is used to select which instance of KIQ to use for the invalidation
  *
  * Flush the TLB for the requested pasid.
  */
-static int gmc_v7_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
-					uint16_t pasid, uint32_t flush_type,
-					bool all_hub)
+static void gmc_v7_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
+					 uint16_t pasid, uint32_t flush_type,
+					 bool all_hub, uint32_t inst)
 {
+	u32 mask = 0x0;
 	int vmid;
-	unsigned int tmp;
-
-	if (amdgpu_in_reset(adev))
-		return -EIO;
 
 	for (vmid = 1; vmid < 16; vmid++) {
+		u32 tmp = RREG32(mmATC_VMID0_PASID_MAPPING + vmid);
 
-		tmp = RREG32(mmATC_VMID0_PASID_MAPPING + vmid);
 		if ((tmp & ATC_VMID0_PASID_MAPPING__VALID_MASK) &&
-			(tmp & ATC_VMID0_PASID_MAPPING__PASID_MASK) == pasid) {
-			WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid);
-			RREG32(mmVM_INVALIDATE_RESPONSE);
-			break;
-		}
+		    (tmp & ATC_VMID0_PASID_MAPPING__PASID_MASK) == pasid)
+			mask |= 1 << vmid;
 	}
 
-	return 0;
+	WREG32(mmVM_INVALIDATE_REQUEST, mask);
+	RREG32(mmVM_INVALIDATE_RESPONSE);
 }
 
 /*
@@ -471,7 +475,7 @@ static void gmc_v7_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 }
 
 static uint64_t gmc_v7_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
-					    unsigned vmid, uint64_t pd_addr)
+					    unsigned int vmid, uint64_t pd_addr)
 {
 	uint32_t reg;
 
@@ -487,8 +491,8 @@ static uint64_t gmc_v7_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
 	return pd_addr;
 }
 
-static void gmc_v7_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
-					unsigned pasid)
+static void gmc_v7_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned int vmid,
+					unsigned int pasid)
 {
 	amdgpu_ring_emit_wreg(ring, mmIH_VMID_0_LUT + vmid, pasid);
 }
@@ -500,7 +504,9 @@ static void gmc_v7_0_get_vm_pde(struct amdgpu_device *adev, int level,
 }
 
 static void gmc_v7_0_get_vm_pte(struct amdgpu_device *adev,
-				struct amdgpu_bo_va_mapping *mapping,
+				struct amdgpu_vm *vm,
+				struct amdgpu_bo *bo,
+				uint32_t vm_flags,
 				uint64_t *flags)
 {
 	*flags &= ~AMDGPU_PTE_EXECUTABLE;
@@ -567,9 +573,10 @@ static void gmc_v7_0_set_prt(struct amdgpu_device *adev, bool enable)
 	WREG32(mmVM_PRT_CNTL, tmp);
 
 	if (enable) {
-		uint32_t low = AMDGPU_VA_RESERVED_SIZE >> AMDGPU_GPU_PAGE_SHIFT;
+		uint32_t low = AMDGPU_VA_RESERVED_BOTTOM >>
+			AMDGPU_GPU_PAGE_SHIFT;
 		uint32_t high = adev->vm_manager.max_pfn -
-			(AMDGPU_VA_RESERVED_SIZE >> AMDGPU_GPU_PAGE_SHIFT);
+			(AMDGPU_VA_RESERVED_TOP >> AMDGPU_GPU_PAGE_SHIFT);
 
 		WREG32(mmVM_PRT_APERTURE0_LOW_ADDR, low);
 		WREG32(mmVM_PRT_APERTURE1_LOW_ADDR, low);
@@ -699,7 +706,7 @@ static int gmc_v7_0_gart_enable(struct amdgpu_device *adev)
 
 	gmc_v7_0_flush_gpu_tlb(adev, 0, 0, 0);
 	DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
-		 (unsigned)(adev->gmc.gart_size >> 20),
+		 (unsigned int)(adev->gmc.gart_size >> 20),
 		 (unsigned long long)table_addr);
 	return 0;
 }
@@ -760,7 +767,7 @@ static void gmc_v7_0_gart_disable(struct amdgpu_device *adev)
  * Print human readable fault information (CIK).
  */
 static void gmc_v7_0_vm_decode_fault(struct amdgpu_device *adev, u32 status,
-				     u32 addr, u32 mc_client, unsigned pasid)
+				     u32 addr, u32 mc_client, unsigned int pasid)
 {
 	u32 vmid = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS, VMID);
 	u32 protections = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
@@ -927,9 +934,9 @@ static int gmc_v7_0_convert_vram_type(int mc_seq_vram_type)
 	}
 }
 
-static int gmc_v7_0_early_init(void *handle)
+static int gmc_v7_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v7_0_set_gmc_funcs(adev);
 	gmc_v7_0_set_irq_funcs(adev);
@@ -941,13 +948,14 @@ static int gmc_v7_0_early_init(void *handle)
 		adev->gmc.shared_aperture_end + 1;
 	adev->gmc.private_aperture_end =
 		adev->gmc.private_aperture_start + (4ULL << 30) - 1;
+	adev->gmc.noretry_flags = AMDGPU_VM_NORETRY_FLAGS_TF;
 
 	return 0;
 }
 
-static int gmc_v7_0_late_init(void *handle)
+static int gmc_v7_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS)
 		return amdgpu_irq_get(adev, &adev->gmc.vm_fault, 0);
@@ -955,15 +963,16 @@ static int gmc_v7_0_late_init(void *handle)
 		return 0;
 }
 
-static unsigned gmc_v7_0_get_vbios_fb_size(struct amdgpu_device *adev)
+static unsigned int gmc_v7_0_get_vbios_fb_size(struct amdgpu_device *adev)
 {
 	u32 d1vga_control = RREG32(mmD1VGA_CONTROL);
-	unsigned size;
+	unsigned int size;
 
 	if (REG_GET_FIELD(d1vga_control, D1VGA_CONTROL, D1VGA_MODE_ENABLE)) {
 		size = AMDGPU_VBIOS_VGA_ALLOCATION;
 	} else {
 		u32 viewport = RREG32(mmVIEWPORT_SIZE);
+
 		size = (REG_GET_FIELD(viewport, VIEWPORT_SIZE, VIEWPORT_HEIGHT) *
 			REG_GET_FIELD(viewport, VIEWPORT_SIZE, VIEWPORT_WIDTH) *
 			4);
@@ -972,17 +981,18 @@ static unsigned gmc_v7_0_get_vbios_fb_size(struct amdgpu_device *adev)
 	return size;
 }
 
-static int gmc_v7_0_sw_init(void *handle)
+static int gmc_v7_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	adev->num_vmhubs = 1;
+	set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
 
 	if (adev->flags & AMD_IS_APU) {
 		adev->gmc.vram_type = AMDGPU_VRAM_TYPE_UNKNOWN;
 	} else {
 		u32 tmp = RREG32(mmMC_SEQ_MISC0);
+
 		tmp &= MC_SEQ_MISC0__MT__MASK;
 		adev->gmc.vram_type = gmc_v7_0_convert_vram_type(tmp);
 	}
@@ -1058,14 +1068,14 @@ static int gmc_v7_0_sw_init(void *handle)
 					GFP_KERNEL);
 	if (!adev->gmc.vm_fault_info)
 		return -ENOMEM;
-	atomic_set(&adev->gmc.vm_fault_info_updated, 0);
+	atomic_set_release(&adev->gmc.vm_fault_info_updated, 0);
 
 	return 0;
 }
 
-static int gmc_v7_0_sw_fini(void *handle)
+static int gmc_v7_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_gem_force_release(adev);
 	amdgpu_vm_manager_fini(adev);
@@ -1077,10 +1087,10 @@ static int gmc_v7_0_sw_fini(void *handle)
 	return 0;
 }
 
-static int gmc_v7_0_hw_init(void *handle)
+static int gmc_v7_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v7_0_init_golden_registers(adev);
 
@@ -1100,13 +1110,13 @@ static int gmc_v7_0_hw_init(void *handle)
 
 	if (amdgpu_emu_mode == 1)
 		return amdgpu_gmc_vram_checking(adev);
-	else
-		return r;
+
+	return 0;
 }
 
-static int gmc_v7_0_hw_fini(void *handle)
+static int gmc_v7_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
 	gmc_v7_0_gart_disable(adev);
@@ -1114,32 +1124,29 @@ static int gmc_v7_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int gmc_v7_0_suspend(void *handle)
+static int gmc_v7_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	gmc_v7_0_hw_fini(adev);
+	gmc_v7_0_hw_fini(ip_block);
 
 	return 0;
 }
 
-static int gmc_v7_0_resume(void *handle)
+static int gmc_v7_0_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = gmc_v7_0_hw_init(adev);
+	r = gmc_v7_0_hw_init(ip_block);
 	if (r)
 		return r;
 
-	amdgpu_vmid_reset_all(adev);
+	amdgpu_vmid_reset_all(ip_block->adev);
 
 	return 0;
 }
 
-static bool gmc_v7_0_is_idle(void *handle)
+static bool gmc_v7_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (tmp & (SRBM_STATUS__MCB_BUSY_MASK | SRBM_STATUS__MCB_NON_DISPLAY_BUSY_MASK |
@@ -1149,20 +1156,13 @@ static bool gmc_v7_0_is_idle(void *handle)
 	return true;
 }
 
-static int gmc_v7_0_wait_for_idle(void *handle)
+static int gmc_v7_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
-	unsigned i;
-	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	unsigned int i;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		/* read MC_STATUS */
-		tmp = RREG32(mmSRBM_STATUS) & (SRBM_STATUS__MCB_BUSY_MASK |
-					       SRBM_STATUS__MCB_NON_DISPLAY_BUSY_MASK |
-					       SRBM_STATUS__MCC_BUSY_MASK |
-					       SRBM_STATUS__MCD_BUSY_MASK |
-					       SRBM_STATUS__VMC_BUSY_MASK);
-		if (!tmp)
+		if (gmc_v7_0_is_idle(ip_block))
 			return 0;
 		udelay(1);
 	}
@@ -1170,9 +1170,9 @@ static int gmc_v7_0_wait_for_idle(void *handle)
 
 }
 
-static int gmc_v7_0_soft_reset(void *handle)
+static int gmc_v7_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 srbm_soft_reset = 0;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
@@ -1189,10 +1189,8 @@ static int gmc_v7_0_soft_reset(void *handle)
 
 	if (srbm_soft_reset) {
 		gmc_v7_0_mc_stop(adev);
-		if (gmc_v7_0_wait_for_idle((void *)adev)) {
+		if (gmc_v7_0_wait_for_idle(ip_block))
 			dev_warn(adev->dev, "Wait for GMC idle timed out !\n");
-		}
-
 
 		tmp = RREG32(mmSRBM_SOFT_RESET);
 		tmp |= srbm_soft_reset;
@@ -1218,7 +1216,7 @@ static int gmc_v7_0_soft_reset(void *handle)
 
 static int gmc_v7_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 					     struct amdgpu_irq_src *src,
-					     unsigned type,
+					     unsigned int type,
 					     enum amdgpu_interrupt_state state)
 {
 	u32 tmp;
@@ -1263,6 +1261,12 @@ static int gmc_v7_0_process_interrupt(struct amdgpu_device *adev,
 {
 	u32 addr, status, mc_client, vmid;
 
+	/* Delegate to the soft IRQ handler ring */
+	if (adev->irq.ih_soft.enabled && entry->ih != &adev->irq.ih_soft) {
+		amdgpu_irq_delegate(adev, entry, 4);
+		return 1;
+	}
+
 	addr = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_ADDR);
 	status = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_STATUS);
 	mc_client = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_MCCLIENT);
@@ -1272,6 +1276,9 @@ static int gmc_v7_0_process_interrupt(struct amdgpu_device *adev,
 	if (!addr && !status)
 		return 0;
 
+	amdgpu_vm_update_fault_cache(adev, entry->pasid,
+				     ((u64)addr) << AMDGPU_GPU_PAGE_SHIFT, status, AMDGPU_GFXHUB(0));
+
 	if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_FIRST)
 		gmc_v7_0_set_fault_enable_default(adev, false);
 
@@ -1289,7 +1296,7 @@ static int gmc_v7_0_process_interrupt(struct amdgpu_device *adev,
 	vmid = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
 			     VMID);
 	if (amdgpu_amdkfd_is_kfd_vmid(adev, vmid)
-		&& !atomic_read(&adev->gmc.vm_fault_info_updated)) {
+		&& !atomic_read_acquire(&adev->gmc.vm_fault_info_updated)) {
 		struct kfd_vm_fault_info *info = adev->gmc.vm_fault_info;
 		u32 protections = REG_GET_FIELD(status,
 					VM_CONTEXT1_PROTECTION_FAULT_STATUS,
@@ -1305,18 +1312,17 @@ static int gmc_v7_0_process_interrupt(struct amdgpu_device *adev,
 		info->prot_read = protections & 0x8 ? true : false;
 		info->prot_write = protections & 0x10 ? true : false;
 		info->prot_exec = protections & 0x20 ? true : false;
-		mb();
-		atomic_set(&adev->gmc.vm_fault_info_updated, 1);
+		atomic_set_release(&adev->gmc.vm_fault_info_updated, 1);
 	}
 
 	return 0;
 }
 
-static int gmc_v7_0_set_clockgating_state(void *handle,
+static int gmc_v7_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	bool gate = false;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (state == AMD_CG_STATE_GATE)
 		gate = true;
@@ -1332,7 +1338,7 @@ static int gmc_v7_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int gmc_v7_0_set_powergating_state(void *handle,
+static int gmc_v7_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -1382,8 +1388,7 @@ static void gmc_v7_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->gmc.vm_fault.funcs = &gmc_v7_0_irq_funcs;
 }
 
-const struct amdgpu_ip_block_version gmc_v7_0_ip_block =
-{
+const struct amdgpu_ip_block_version gmc_v7_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GMC,
 	.major = 7,
 	.minor = 0,
@@ -1391,8 +1396,7 @@ const struct amdgpu_ip_block_version gmc_v7_0_ip_block =
 	.funcs = &gmc_v7_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version gmc_v7_4_ip_block =
-{
+const struct amdgpu_ip_block_version gmc_v7_4_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GMC,
 	.major = 7,
 	.minor = 4,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
index 504c1b34dab7..6551b60f2584 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
@@ -53,7 +53,7 @@
 
 static void gmc_v8_0_set_gmc_funcs(struct amdgpu_device *adev);
 static void gmc_v8_0_set_irq_funcs(struct amdgpu_device *adev);
-static int gmc_v8_0_wait_for_idle(void *handle);
+static int gmc_v8_0_wait_for_idle(struct amdgpu_ip_block *ip_block);
 
 MODULE_FIRMWARE("amdgpu/tonga_mc.bin");
 MODULE_FIRMWARE("amdgpu/polaris11_mc.bin");
@@ -64,8 +64,7 @@ MODULE_FIRMWARE("amdgpu/polaris11_k_mc.bin");
 MODULE_FIRMWARE("amdgpu/polaris10_k_mc.bin");
 MODULE_FIRMWARE("amdgpu/polaris12_k_mc.bin");
 
-static const u32 golden_settings_tonga_a11[] =
-{
+static const u32 golden_settings_tonga_a11[] = {
 	mmMC_ARB_WTM_GRPWT_RD, 0x00000003, 0x00000000,
 	mmMC_HUB_RDREQ_DMIF_LIMIT, 0x0000007f, 0x00000028,
 	mmMC_HUB_WDP_UMC, 0x00007fb6, 0x00000991,
@@ -75,34 +74,29 @@ static const u32 golden_settings_tonga_a11[] =
 	mmVM_PRT_APERTURE3_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 };
 
-static const u32 tonga_mgcg_cgcg_init[] =
-{
+static const u32 tonga_mgcg_cgcg_init[] = {
 	mmMC_MEM_POWER_LS, 0xffffffff, 0x00000104
 };
 
-static const u32 golden_settings_fiji_a10[] =
-{
+static const u32 golden_settings_fiji_a10[] = {
 	mmVM_PRT_APERTURE0_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE1_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE2_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE3_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 };
 
-static const u32 fiji_mgcg_cgcg_init[] =
-{
+static const u32 fiji_mgcg_cgcg_init[] = {
 	mmMC_MEM_POWER_LS, 0xffffffff, 0x00000104
 };
 
-static const u32 golden_settings_polaris11_a11[] =
-{
+static const u32 golden_settings_polaris11_a11[] = {
 	mmVM_PRT_APERTURE0_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE1_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE2_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE3_LOW_ADDR, 0x0fffffff, 0x0fffffff
 };
 
-static const u32 golden_settings_polaris10_a11[] =
-{
+static const u32 golden_settings_polaris10_a11[] = {
 	mmMC_ARB_WTM_GRPWT_RD, 0x00000003, 0x00000000,
 	mmVM_PRT_APERTURE0_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE1_LOW_ADDR, 0x0fffffff, 0x0fffffff,
@@ -110,19 +104,16 @@ static const u32 golden_settings_polaris10_a11[] =
 	mmVM_PRT_APERTURE3_LOW_ADDR, 0x0fffffff, 0x0fffffff
 };
 
-static const u32 cz_mgcg_cgcg_init[] =
-{
+static const u32 cz_mgcg_cgcg_init[] = {
 	mmMC_MEM_POWER_LS, 0xffffffff, 0x00000104
 };
 
-static const u32 stoney_mgcg_cgcg_init[] =
-{
+static const u32 stoney_mgcg_cgcg_init[] = {
 	mmATC_MISC_CG, 0xffffffff, 0x000c0200,
 	mmMC_MEM_POWER_LS, 0xffffffff, 0x00000104
 };
 
-static const u32 golden_settings_stoney_common[] =
-{
+static const u32 golden_settings_stoney_common[] = {
 	mmMC_HUB_RDREQ_UVD, MC_HUB_RDREQ_UVD__PRESCALE_MASK, 0x00000004,
 	mmMC_RD_GRP_OTH, MC_RD_GRP_OTH__UVD_MASK, 0x00600000
 };
@@ -179,8 +170,13 @@ static void gmc_v8_0_init_golden_registers(struct amdgpu_device *adev)
 static void gmc_v8_0_mc_stop(struct amdgpu_device *adev)
 {
 	u32 blackout;
+	struct amdgpu_ip_block *ip_block;
+
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GMC);
+	if (!ip_block)
+		return;
 
-	gmc_v8_0_wait_for_idle(adev);
+	gmc_v8_0_wait_for_idle(ip_block);
 
 	blackout = RREG32(mmMC_SHARED_BLACKOUT_CNTL);
 	if (REG_GET_FIELD(blackout, MC_SHARED_BLACKOUT_CNTL, BLACKOUT_MODE) != 1) {
@@ -221,7 +217,6 @@ static void gmc_v8_0_mc_resume(struct amdgpu_device *adev)
 static int gmc_v8_0_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[30];
 	int err;
 
 	DRM_DEBUG("\n");
@@ -260,13 +255,14 @@ static int gmc_v8_0_init_microcode(struct amdgpu_device *adev)
 	case CHIP_STONEY:
 	case CHIP_VEGAM:
 		return 0;
-	default: BUG();
+	default:
+		return -EINVAL;
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mc.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gmc.fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gmc.fw, AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_mc.bin", chip_name);
 	if (err) {
-		pr_err("mc: Failed to load firmware \"%s\"\n", fw_name);
+		pr_err("mc: Failed to load firmware \"%s_mc.bin\"\n", chip_name);
 		amdgpu_ucode_release(&adev->gmc.fw);
 	}
 	return err;
@@ -421,8 +417,9 @@ static void gmc_v8_0_vram_gtt_location(struct amdgpu_device *adev,
 		base = RREG32(mmMC_VM_FB_LOCATION) & 0xFFFF;
 	base <<= 24;
 
+	amdgpu_gmc_set_agp_default(adev, mc);
 	amdgpu_gmc_vram_location(adev, mc, base);
-	amdgpu_gmc_gart_location(adev, mc);
+	amdgpu_gmc_gart_location(adev, mc, AMDGPU_GART_PLACEMENT_BEST_FIT);
 }
 
 /**
@@ -435,6 +432,7 @@ static void gmc_v8_0_vram_gtt_location(struct amdgpu_device *adev,
  */
 static void gmc_v8_0_mc_program(struct amdgpu_device *adev)
 {
+	struct amdgpu_ip_block *ip_block;
 	u32 tmp;
 	int i, j;
 
@@ -448,9 +446,13 @@ static void gmc_v8_0_mc_program(struct amdgpu_device *adev)
 	}
 	WREG32(mmHDP_REG_COHERENCY_FLUSH_CNTL, 0);
 
-	if (gmc_v8_0_wait_for_idle((void *)adev)) {
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GMC);
+	if (!ip_block)
+		return;
+
+	if (gmc_v8_0_wait_for_idle(ip_block))
 		dev_warn(adev->dev, "Wait for MC idle timedout !\n");
-	}
+
 	if (adev->mode_info.num_crtc) {
 		/* Lockout access through VGA aperture*/
 		tmp = RREG32(mmVGA_HDP_CONTROL);
@@ -481,11 +483,10 @@ static void gmc_v8_0_mc_program(struct amdgpu_device *adev)
 	}
 
 	WREG32(mmMC_VM_AGP_BASE, 0);
-	WREG32(mmMC_VM_AGP_TOP, 0x0FFFFFFF);
-	WREG32(mmMC_VM_AGP_BOT, 0x0FFFFFFF);
-	if (gmc_v8_0_wait_for_idle((void *)adev)) {
+	WREG32(mmMC_VM_AGP_TOP, adev->gmc.agp_end >> 22);
+	WREG32(mmMC_VM_AGP_BOT, adev->gmc.agp_start >> 22);
+	if (gmc_v8_0_wait_for_idle(ip_block))
 		dev_warn(adev->dev, "Wait for MC idle timedout !\n");
-	}
 
 	WREG32(mmBIF_FB_EN, BIF_FB_EN__FB_READ_EN_MASK | BIF_FB_EN__FB_WRITE_EN_MASK);
 
@@ -517,11 +518,11 @@ static int gmc_v8_0_mc_init(struct amdgpu_device *adev)
 
 		/* Get VRAM informations */
 		tmp = RREG32(mmMC_ARB_RAMCFG);
-		if (REG_GET_FIELD(tmp, MC_ARB_RAMCFG, CHANSIZE)) {
+		if (REG_GET_FIELD(tmp, MC_ARB_RAMCFG, CHANSIZE))
 			chansize = 64;
-		} else {
+		else
 			chansize = 32;
-		}
+
 		tmp = RREG32(mmMC_SHARED_CHMAP);
 		switch (REG_GET_FIELD(tmp, MC_SHARED_CHMAP, NOOFCHAN)) {
 		case 0:
@@ -617,32 +618,27 @@ static int gmc_v8_0_mc_init(struct amdgpu_device *adev)
  * @pasid: pasid to be flush
  * @flush_type: type of flush
  * @all_hub: flush all hubs
+ * @inst: is used to select which instance of KIQ to use for the invalidation
  *
  * Flush the TLB for the requested pasid.
  */
-static int gmc_v8_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
-					uint16_t pasid, uint32_t flush_type,
-					bool all_hub)
+static void gmc_v8_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
+					 uint16_t pasid, uint32_t flush_type,
+					 bool all_hub, uint32_t inst)
 {
+	u32 mask = 0x0;
 	int vmid;
-	unsigned int tmp;
-
-	if (amdgpu_in_reset(adev))
-		return -EIO;
 
 	for (vmid = 1; vmid < 16; vmid++) {
+		u32 tmp = RREG32(mmATC_VMID0_PASID_MAPPING + vmid);
 
-		tmp = RREG32(mmATC_VMID0_PASID_MAPPING + vmid);
 		if ((tmp & ATC_VMID0_PASID_MAPPING__VALID_MASK) &&
-			(tmp & ATC_VMID0_PASID_MAPPING__PASID_MASK) == pasid) {
-			WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid);
-			RREG32(mmVM_INVALIDATE_RESPONSE);
-			break;
-		}
+		    (tmp & ATC_VMID0_PASID_MAPPING__PASID_MASK) == pasid)
+			mask |= 1 << vmid;
 	}
 
-	return 0;
-
+	WREG32(mmVM_INVALIDATE_REQUEST, mask);
+	RREG32(mmVM_INVALIDATE_RESPONSE);
 }
 
 /*
@@ -670,7 +666,7 @@ static void gmc_v8_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 }
 
 static uint64_t gmc_v8_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
-					    unsigned vmid, uint64_t pd_addr)
+					    unsigned int vmid, uint64_t pd_addr)
 {
 	uint32_t reg;
 
@@ -686,8 +682,8 @@ static uint64_t gmc_v8_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
 	return pd_addr;
 }
 
-static void gmc_v8_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
-					unsigned pasid)
+static void gmc_v8_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned int vmid,
+					unsigned int pasid)
 {
 	amdgpu_ring_emit_wreg(ring, mmIH_VMID_0_LUT + vmid, pasid);
 }
@@ -720,11 +716,15 @@ static void gmc_v8_0_get_vm_pde(struct amdgpu_device *adev, int level,
 }
 
 static void gmc_v8_0_get_vm_pte(struct amdgpu_device *adev,
-				struct amdgpu_bo_va_mapping *mapping,
+				struct amdgpu_vm *vm,
+				struct amdgpu_bo *bo,
+				uint32_t vm_flags,
 				uint64_t *flags)
 {
-	*flags &= ~AMDGPU_PTE_EXECUTABLE;
-	*flags |= mapping->flags & AMDGPU_PTE_EXECUTABLE;
+	if (vm_flags & AMDGPU_VM_PAGE_EXECUTABLE)
+		*flags |= AMDGPU_PTE_EXECUTABLE;
+	else
+		*flags &= ~AMDGPU_PTE_EXECUTABLE;
 	*flags &= ~AMDGPU_PTE_PRT;
 }
 
@@ -758,11 +758,11 @@ static void gmc_v8_0_set_fault_enable_default(struct amdgpu_device *adev,
 }
 
 /**
- * gmc_v8_0_set_prt - set PRT VM fault
+ * gmc_v8_0_set_prt() - set PRT VM fault
  *
  * @adev: amdgpu_device pointer
  * @enable: enable/disable VM fault handling for PRT
-*/
+ */
 static void gmc_v8_0_set_prt(struct amdgpu_device *adev, bool enable)
 {
 	u32 tmp;
@@ -790,9 +790,10 @@ static void gmc_v8_0_set_prt(struct amdgpu_device *adev, bool enable)
 	WREG32(mmVM_PRT_CNTL, tmp);
 
 	if (enable) {
-		uint32_t low = AMDGPU_VA_RESERVED_SIZE >> AMDGPU_GPU_PAGE_SHIFT;
+		uint32_t low = AMDGPU_VA_RESERVED_BOTTOM >>
+			AMDGPU_GPU_PAGE_SHIFT;
 		uint32_t high = adev->vm_manager.max_pfn -
-			(AMDGPU_VA_RESERVED_SIZE >> AMDGPU_GPU_PAGE_SHIFT);
+			(AMDGPU_VA_RESERVED_TOP >> AMDGPU_GPU_PAGE_SHIFT);
 
 		WREG32(mmVM_PRT_APERTURE0_LOW_ADDR, low);
 		WREG32(mmVM_PRT_APERTURE1_LOW_ADDR, low);
@@ -939,7 +940,7 @@ static int gmc_v8_0_gart_enable(struct amdgpu_device *adev)
 
 	gmc_v8_0_flush_gpu_tlb(adev, 0, 0, 0);
 	DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
-		 (unsigned)(adev->gmc.gart_size >> 20),
+		 (unsigned int)(adev->gmc.gart_size >> 20),
 		 (unsigned long long)table_addr);
 	return 0;
 }
@@ -1000,7 +1001,7 @@ static void gmc_v8_0_gart_disable(struct amdgpu_device *adev)
  * Print human readable fault information (VI).
  */
 static void gmc_v8_0_vm_decode_fault(struct amdgpu_device *adev, u32 status,
-				     u32 addr, u32 mc_client, unsigned pasid)
+				     u32 addr, u32 mc_client, unsigned int pasid)
 {
 	u32 vmid = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS, VMID);
 	u32 protections = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
@@ -1041,9 +1042,9 @@ static int gmc_v8_0_convert_vram_type(int mc_seq_vram_type)
 	}
 }
 
-static int gmc_v8_0_early_init(void *handle)
+static int gmc_v8_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v8_0_set_gmc_funcs(adev);
 	gmc_v8_0_set_irq_funcs(adev);
@@ -1055,13 +1056,14 @@ static int gmc_v8_0_early_init(void *handle)
 		adev->gmc.shared_aperture_end + 1;
 	adev->gmc.private_aperture_end =
 		adev->gmc.private_aperture_start + (4ULL << 30) - 1;
+	adev->gmc.noretry_flags = AMDGPU_VM_NORETRY_FLAGS_TF;
 
 	return 0;
 }
 
-static int gmc_v8_0_late_init(void *handle)
+static int gmc_v8_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS)
 		return amdgpu_irq_get(adev, &adev->gmc.vm_fault, 0);
@@ -1069,15 +1071,16 @@ static int gmc_v8_0_late_init(void *handle)
 		return 0;
 }
 
-static unsigned gmc_v8_0_get_vbios_fb_size(struct amdgpu_device *adev)
+static unsigned int gmc_v8_0_get_vbios_fb_size(struct amdgpu_device *adev)
 {
 	u32 d1vga_control = RREG32(mmD1VGA_CONTROL);
-	unsigned size;
+	unsigned int size;
 
 	if (REG_GET_FIELD(d1vga_control, D1VGA_CONTROL, D1VGA_MODE_ENABLE)) {
 		size = AMDGPU_VBIOS_VGA_ALLOCATION;
 	} else {
 		u32 viewport = RREG32(mmVIEWPORT_SIZE);
+
 		size = (REG_GET_FIELD(viewport, VIEWPORT_SIZE, VIEWPORT_HEIGHT) *
 			REG_GET_FIELD(viewport, VIEWPORT_SIZE, VIEWPORT_WIDTH) *
 			4);
@@ -1088,12 +1091,12 @@ static unsigned gmc_v8_0_get_vbios_fb_size(struct amdgpu_device *adev)
 
 #define mmMC_SEQ_MISC0_FIJI 0xA71
 
-static int gmc_v8_0_sw_init(void *handle)
+static int gmc_v8_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	adev->num_vmhubs = 1;
+	set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
 
 	if (adev->flags & AMD_IS_APU) {
 		adev->gmc.vram_type = AMDGPU_VRAM_TYPE_UNKNOWN;
@@ -1180,14 +1183,14 @@ static int gmc_v8_0_sw_init(void *handle)
 					GFP_KERNEL);
 	if (!adev->gmc.vm_fault_info)
 		return -ENOMEM;
-	atomic_set(&adev->gmc.vm_fault_info_updated, 0);
+	atomic_set_release(&adev->gmc.vm_fault_info_updated, 0);
 
 	return 0;
 }
 
-static int gmc_v8_0_sw_fini(void *handle)
+static int gmc_v8_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_gem_force_release(adev);
 	amdgpu_vm_manager_fini(adev);
@@ -1199,10 +1202,10 @@ static int gmc_v8_0_sw_fini(void *handle)
 	return 0;
 }
 
-static int gmc_v8_0_hw_init(void *handle)
+static int gmc_v8_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v8_0_init_golden_registers(adev);
 
@@ -1230,13 +1233,13 @@ static int gmc_v8_0_hw_init(void *handle)
 
 	if (amdgpu_emu_mode == 1)
 		return amdgpu_gmc_vram_checking(adev);
-	else
-		return r;
+
+	return 0;
 }
 
-static int gmc_v8_0_hw_fini(void *handle)
+static int gmc_v8_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
 	gmc_v8_0_gart_disable(adev);
@@ -1244,32 +1247,29 @@ static int gmc_v8_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int gmc_v8_0_suspend(void *handle)
+static int gmc_v8_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	gmc_v8_0_hw_fini(adev);
+	gmc_v8_0_hw_fini(ip_block);
 
 	return 0;
 }
 
-static int gmc_v8_0_resume(void *handle)
+static int gmc_v8_0_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = gmc_v8_0_hw_init(adev);
+	r = gmc_v8_0_hw_init(ip_block);
 	if (r)
 		return r;
 
-	amdgpu_vmid_reset_all(adev);
+	amdgpu_vmid_reset_all(ip_block->adev);
 
 	return 0;
 }
 
-static bool gmc_v8_0_is_idle(void *handle)
+static bool gmc_v8_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (tmp & (SRBM_STATUS__MCB_BUSY_MASK | SRBM_STATUS__MCB_NON_DISPLAY_BUSY_MASK |
@@ -1279,11 +1279,11 @@ static bool gmc_v8_0_is_idle(void *handle)
 	return true;
 }
 
-static int gmc_v8_0_wait_for_idle(void *handle)
+static int gmc_v8_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
-	unsigned i;
+	unsigned int i;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		/* read MC_STATUS */
@@ -1301,10 +1301,10 @@ static int gmc_v8_0_wait_for_idle(void *handle)
 
 }
 
-static bool gmc_v8_0_check_soft_reset(void *handle)
+static bool gmc_v8_0_check_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 srbm_soft_reset = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (tmp & SRBM_STATUS__VMC_BUSY_MASK)
@@ -1317,33 +1317,34 @@ static bool gmc_v8_0_check_soft_reset(void *handle)
 			srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset,
 							SRBM_SOFT_RESET, SOFT_RESET_MC, 1);
 	}
+
 	if (srbm_soft_reset) {
 		adev->gmc.srbm_soft_reset = srbm_soft_reset;
 		return true;
-	} else {
-		adev->gmc.srbm_soft_reset = 0;
-		return false;
 	}
+
+	adev->gmc.srbm_soft_reset = 0;
+
+	return false;
 }
 
-static int gmc_v8_0_pre_soft_reset(void *handle)
+static int gmc_v8_0_pre_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (!adev->gmc.srbm_soft_reset)
 		return 0;
 
 	gmc_v8_0_mc_stop(adev);
-	if (gmc_v8_0_wait_for_idle(adev)) {
+	if (gmc_v8_0_wait_for_idle(ip_block))
 		dev_warn(adev->dev, "Wait for GMC idle timed out !\n");
-	}
 
 	return 0;
 }
 
-static int gmc_v8_0_soft_reset(void *handle)
+static int gmc_v8_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 srbm_soft_reset;
 
 	if (!adev->gmc.srbm_soft_reset)
@@ -1372,9 +1373,9 @@ static int gmc_v8_0_soft_reset(void *handle)
 	return 0;
 }
 
-static int gmc_v8_0_post_soft_reset(void *handle)
+static int gmc_v8_0_post_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (!adev->gmc.srbm_soft_reset)
 		return 0;
@@ -1385,7 +1386,7 @@ static int gmc_v8_0_post_soft_reset(void *handle)
 
 static int gmc_v8_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 					     struct amdgpu_irq_src *src,
-					     unsigned type,
+					     unsigned int type,
 					     enum amdgpu_interrupt_state state)
 {
 	u32 tmp;
@@ -1438,6 +1439,12 @@ static int gmc_v8_0_process_interrupt(struct amdgpu_device *adev,
 		return 0;
 	}
 
+	/* Delegate to the soft IRQ handler ring */
+	if (adev->irq.ih_soft.enabled && entry->ih != &adev->irq.ih_soft) {
+		amdgpu_irq_delegate(adev, entry, 4);
+		return 1;
+	}
+
 	addr = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_ADDR);
 	status = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_STATUS);
 	mc_client = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_MCCLIENT);
@@ -1447,22 +1454,29 @@ static int gmc_v8_0_process_interrupt(struct amdgpu_device *adev,
 	if (!addr && !status)
 		return 0;
 
+	amdgpu_vm_update_fault_cache(adev, entry->pasid,
+				     ((u64)addr) << AMDGPU_GPU_PAGE_SHIFT, status, AMDGPU_GFXHUB(0));
+
 	if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_FIRST)
 		gmc_v8_0_set_fault_enable_default(adev, false);
 
 	if (printk_ratelimit()) {
-		struct amdgpu_task_info task_info;
+		struct amdgpu_task_info *task_info;
+
+		dev_err(adev->dev, "GPU fault detected: %d 0x%08x\n",
+			entry->src_id, entry->src_data[0]);
 
-		memset(&task_info, 0, sizeof(struct amdgpu_task_info));
-		amdgpu_vm_get_task_info(adev, entry->pasid, &task_info);
+		task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid);
+		if (task_info) {
+			amdgpu_vm_print_task_info(adev, task_info);
+			amdgpu_vm_put_task_info(task_info);
+		}
 
-		dev_err(adev->dev, "GPU fault detected: %d 0x%08x for process %s pid %d thread %s pid %d\n",
-			entry->src_id, entry->src_data[0], task_info.process_name,
-			task_info.tgid, task_info.task_name, task_info.pid);
 		dev_err(adev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x%08X\n",
-			addr);
+				addr);
 		dev_err(adev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n",
 			status);
+
 		gmc_v8_0_vm_decode_fault(adev, status, addr, mc_client,
 					 entry->pasid);
 	}
@@ -1470,7 +1484,7 @@ static int gmc_v8_0_process_interrupt(struct amdgpu_device *adev,
 	vmid = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
 			     VMID);
 	if (amdgpu_amdkfd_is_kfd_vmid(adev, vmid)
-		&& !atomic_read(&adev->gmc.vm_fault_info_updated)) {
+		&& !atomic_read_acquire(&adev->gmc.vm_fault_info_updated)) {
 		struct kfd_vm_fault_info *info = adev->gmc.vm_fault_info;
 		u32 protections = REG_GET_FIELD(status,
 					VM_CONTEXT1_PROTECTION_FAULT_STATUS,
@@ -1486,8 +1500,7 @@ static int gmc_v8_0_process_interrupt(struct amdgpu_device *adev,
 		info->prot_read = protections & 0x8 ? true : false;
 		info->prot_write = protections & 0x10 ? true : false;
 		info->prot_exec = protections & 0x20 ? true : false;
-		mb();
-		atomic_set(&adev->gmc.vm_fault_info_updated, 1);
+		atomic_set_release(&adev->gmc.vm_fault_info_updated, 1);
 	}
 
 	return 0;
@@ -1653,10 +1666,10 @@ static void fiji_update_mc_light_sleep(struct amdgpu_device *adev,
 	}
 }
 
-static int gmc_v8_0_set_clockgating_state(void *handle,
+static int gmc_v8_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
@@ -1674,15 +1687,15 @@ static int gmc_v8_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int gmc_v8_0_set_powergating_state(void *handle,
+static int gmc_v8_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
 }
 
-static void gmc_v8_0_get_clockgating_state(void *handle, u64 *flags)
+static void gmc_v8_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	if (amdgpu_sriov_vf(adev))
@@ -1746,8 +1759,7 @@ static void gmc_v8_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->gmc.vm_fault.funcs = &gmc_v8_0_irq_funcs;
 }
 
-const struct amdgpu_ip_block_version gmc_v8_0_ip_block =
-{
+const struct amdgpu_ip_block_version gmc_v8_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GMC,
 	.major = 8,
 	.minor = 0,
@@ -1755,8 +1767,7 @@ const struct amdgpu_ip_block_version gmc_v8_0_ip_block =
 	.funcs = &gmc_v8_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version gmc_v8_1_ip_block =
-{
+const struct amdgpu_ip_block_version gmc_v8_1_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GMC,
 	.major = 8,
 	.minor = 1,
@@ -1764,8 +1775,7 @@ const struct amdgpu_ip_block_version gmc_v8_1_ip_block =
 	.funcs = &gmc_v8_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version gmc_v8_5_ip_block =
-{
+const struct amdgpu_ip_block_version gmc_v8_5_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GMC,
 	.major = 8,
 	.minor = 5,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
index 64ab1a306dfe..8ad7519f7b58 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
@@ -56,6 +56,7 @@
 #include "umc_v6_1.h"
 #include "umc_v6_0.h"
 #include "umc_v6_7.h"
+#include "umc_v12_0.h"
 #include "hdp_v4_0.h"
 #include "mca_v3_0.h"
 
@@ -64,8 +65,6 @@
 #include "amdgpu_ras.h"
 #include "amdgpu_xgmi.h"
 
-#include "amdgpu_reset.h"
-
 /* add these here since we already include dce12 headers and these are for DCN */
 #define mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION                                                          0x055d
 #define mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION_BASE_IDX                                                 2
@@ -79,8 +78,7 @@
 #define mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION_DCN2                                                          0x05ea
 #define mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION_DCN2_BASE_IDX                                                 2
 
-
-static const char *gfxhub_client_ids[] = {
+static const char * const gfxhub_client_ids[] = {
 	"CB",
 	"DB",
 	"IA",
@@ -331,14 +329,12 @@ static const char *mmhub_client_ids_aldebaran[][2] = {
 	[384+0][1] = "OSS",
 };
 
-static const struct soc15_reg_golden golden_settings_mmhub_1_0_0[] =
-{
+static const struct soc15_reg_golden golden_settings_mmhub_1_0_0[] = {
 	SOC15_REG_GOLDEN_VALUE(MMHUB, 0, mmDAGB1_WRCLI2, 0x00000007, 0xfe5fe0fa),
 	SOC15_REG_GOLDEN_VALUE(MMHUB, 0, mmMMEA1_DRAM_WR_CLI2GRP_MAP0, 0x00000030, 0x55555565)
 };
 
-static const struct soc15_reg_golden golden_settings_athub_1_0_0[] =
-{
+static const struct soc15_reg_golden golden_settings_athub_1_0_0[] = {
 	SOC15_REG_GOLDEN_VALUE(ATHUB, 0, mmRPB_ARB_CNTL, 0x0000ff00, 0x00000800),
 	SOC15_REG_GOLDEN_VALUE(ATHUB, 0, mmRPB_ARB_CNTL2, 0x00ff00ff, 0x00080008)
 };
@@ -415,13 +411,14 @@ static const uint32_t ecc_umc_mcumc_ctrl_mask_addrs[] = {
 
 static int gmc_v9_0_ecc_interrupt_state(struct amdgpu_device *adev,
 		struct amdgpu_irq_src *src,
-		unsigned type,
+		unsigned int type,
 		enum amdgpu_interrupt_state state)
 {
 	u32 bits, i, tmp, reg;
 
 	/* Devices newer then VEGA10/12 shall have these programming
-	     sequences performed by PSP BL */
+	 * sequences performed by PSP BL
+	 */
 	if (adev->asic_type >= CHIP_VEGA20)
 		return 0;
 
@@ -465,7 +462,7 @@ static int gmc_v9_0_ecc_interrupt_state(struct amdgpu_device *adev,
 
 static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 					struct amdgpu_irq_src *src,
-					unsigned type,
+					unsigned int type,
 					enum amdgpu_interrupt_state state)
 {
 	struct amdgpu_vmhub *hub;
@@ -481,7 +478,7 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
-		for (j = 0; j < adev->num_vmhubs; j++) {
+		for_each_set_bit(j, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS) {
 			hub = &adev->vmhub[j];
 			for (i = 0; i < 16; i++) {
 				reg = hub->vm_context0_cntl + i;
@@ -491,25 +488,25 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 				 * fini/suspend, so the overall state doesn't
 				 * change over the course of suspend/resume.
 				 */
-				if (adev->in_s0ix && (j == AMDGPU_GFXHUB_0))
+				if (adev->in_s0ix && (j == AMDGPU_GFXHUB(0)))
 					continue;
 
-				if (j == AMDGPU_GFXHUB_0)
-					tmp = RREG32_SOC15_IP(GC, reg);
-				else
+				if (j >= AMDGPU_MMHUB0(0))
 					tmp = RREG32_SOC15_IP(MMHUB, reg);
+				else
+					tmp = RREG32_XCC(reg, j);
 
 				tmp &= ~bits;
 
-				if (j == AMDGPU_GFXHUB_0)
-					WREG32_SOC15_IP(GC, reg, tmp);
-				else
+				if (j >= AMDGPU_MMHUB0(0))
 					WREG32_SOC15_IP(MMHUB, reg, tmp);
+				else
+					WREG32_XCC(reg, tmp, j);
 			}
 		}
 		break;
 	case AMDGPU_IRQ_STATE_ENABLE:
-		for (j = 0; j < adev->num_vmhubs; j++) {
+		for_each_set_bit(j, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS) {
 			hub = &adev->vmhub[j];
 			for (i = 0; i < 16; i++) {
 				reg = hub->vm_context0_cntl + i;
@@ -519,20 +516,20 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 				 * fini/suspend, so the overall state doesn't
 				 * change over the course of suspend/resume.
 				 */
-				if (adev->in_s0ix && (j == AMDGPU_GFXHUB_0))
+				if (adev->in_s0ix && (j == AMDGPU_GFXHUB(0)))
 					continue;
 
-				if (j == AMDGPU_GFXHUB_0)
-					tmp = RREG32_SOC15_IP(GC, reg);
-				else
+				if (j >= AMDGPU_MMHUB0(0))
 					tmp = RREG32_SOC15_IP(MMHUB, reg);
+				else
+					tmp = RREG32_XCC(reg, j);
 
 				tmp |= bits;
 
-				if (j == AMDGPU_GFXHUB_0)
-					WREG32_SOC15_IP(GC, reg, tmp);
-				else
+				if (j >= AMDGPU_MMHUB0(0))
 					WREG32_SOC15_IP(MMHUB, reg, tmp);
+				else
+					WREG32_XCC(reg, tmp, j);
 			}
 		}
 		break;
@@ -547,20 +544,44 @@ static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev,
 				      struct amdgpu_irq_src *source,
 				      struct amdgpu_iv_entry *entry)
 {
-	bool retry_fault = !!(entry->src_data[1] & 0x80);
-	bool write_fault = !!(entry->src_data[1] & 0x20);
-	uint32_t status = 0, cid = 0, rw = 0;
-	struct amdgpu_task_info task_info;
+	bool retry_fault = !!(entry->src_data[1] &
+			      AMDGPU_GMC9_FAULT_SOURCE_DATA_RETRY);
+	bool write_fault = !!(entry->src_data[1] &
+			      AMDGPU_GMC9_FAULT_SOURCE_DATA_WRITE);
+	uint32_t status = 0, cid = 0, rw = 0, fed = 0;
+	struct amdgpu_task_info *task_info;
 	struct amdgpu_vmhub *hub;
 	const char *mmhub_cid;
 	const char *hub_name;
+	unsigned int vmhub;
 	u64 addr;
 	uint32_t cam_index = 0;
-	int ret;
+	int ret, xcc_id = 0;
+	uint32_t node_id;
+
+	node_id = entry->node_id;
 
 	addr = (u64)entry->src_data[0] << 12;
 	addr |= ((u64)entry->src_data[1] & 0xf) << 44;
 
+	if (entry->client_id == SOC15_IH_CLIENTID_VMC) {
+		hub_name = "mmhub0";
+		vmhub = AMDGPU_MMHUB0(node_id / 4);
+	} else if (entry->client_id == SOC15_IH_CLIENTID_VMC1) {
+		hub_name = "mmhub1";
+		vmhub = AMDGPU_MMHUB1(0);
+	} else {
+		hub_name = "gfxhub0";
+		if (adev->gfx.funcs->ih_node_to_logical_xcc) {
+			xcc_id = adev->gfx.funcs->ih_node_to_logical_xcc(adev,
+				node_id);
+			if (xcc_id < 0)
+				xcc_id = 0;
+		}
+		vmhub = xcc_id;
+	}
+	hub = &adev->vmhub[vmhub];
+
 	if (retry_fault) {
 		if (adev->irq.retry_cam_enabled) {
 			/* Delegate it to a different ring if the hardware hasn't
@@ -573,7 +594,8 @@ static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev,
 
 			cam_index = entry->src_data[2] & 0x3ff;
 
-			ret = amdgpu_vm_handle_fault(adev, entry->pasid, addr, write_fault);
+			ret = amdgpu_vm_handle_fault(adev, entry->pasid, entry->vmid, node_id,
+						     addr, entry->timestamp, write_fault);
 			WDOORBELL32(adev->irq.retry_cam_doorbell_index, cam_index);
 			if (ret)
 				return 1;
@@ -595,39 +617,38 @@ static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev,
 			/* Try to handle the recoverable page faults by filling page
 			 * tables
 			 */
-			if (amdgpu_vm_handle_fault(adev, entry->pasid, addr, write_fault))
+			if (amdgpu_vm_handle_fault(adev, entry->pasid, entry->vmid, node_id,
+						   addr, entry->timestamp, write_fault))
 				return 1;
 		}
 	}
 
+	if (kgd2kfd_vmfault_fast_path(adev, entry, retry_fault))
+		return 1;
+
 	if (!printk_ratelimit())
 		return 0;
 
-	if (entry->client_id == SOC15_IH_CLIENTID_VMC) {
-		hub_name = "mmhub0";
-		hub = &adev->vmhub[AMDGPU_MMHUB_0];
-	} else if (entry->client_id == SOC15_IH_CLIENTID_VMC1) {
-		hub_name = "mmhub1";
-		hub = &adev->vmhub[AMDGPU_MMHUB_1];
-	} else {
-		hub_name = "gfxhub0";
-		hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	dev_err(adev->dev,
+		"[%s] %s page fault (src_id:%u ring:%u vmid:%u pasid:%u)\n", hub_name,
+		retry_fault ? "retry" : "no-retry",
+		entry->src_id, entry->ring_id, entry->vmid, entry->pasid);
+
+	task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid);
+	if (task_info) {
+		amdgpu_vm_print_task_info(adev, task_info);
+		amdgpu_vm_put_task_info(task_info);
 	}
 
-	memset(&task_info, 0, sizeof(struct amdgpu_task_info));
-	amdgpu_vm_get_task_info(adev, entry->pasid, &task_info);
-
-	dev_err(adev->dev,
-		"[%s] %s page fault (src_id:%u ring:%u vmid:%u "
-		"pasid:%u, for process %s pid %d thread %s pid %d)\n",
-		hub_name, retry_fault ? "retry" : "no-retry",
-		entry->src_id, entry->ring_id, entry->vmid,
-		entry->pasid, task_info.process_name, task_info.tgid,
-		task_info.task_name, task_info.pid);
 	dev_err(adev->dev, "  in page starting at address 0x%016llx from IH client 0x%x (%s)\n",
 		addr, entry->client_id,
 		soc15_ih_clientid_name[entry->client_id]);
 
+	if (amdgpu_is_multi_aid(adev))
+		dev_err(adev->dev, "  cookie node_id %d fault from die %s%d%s\n",
+			node_id, node_id % 4 == 3 ? "RSV" : "AID", node_id / 4,
+			node_id % 4 == 1 ? ".XCD0" : node_id % 4 == 2 ? ".XCD1" : "");
+
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
@@ -636,26 +657,41 @@ static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev,
 	 * be updated to avoid reading an incorrect value due to
 	 * the new fast GRBM interface.
 	 */
-	if ((entry->vmid_src == AMDGPU_GFXHUB_0) &&
-	    (adev->ip_versions[GC_HWIP][0] < IP_VERSION(9, 4, 2)))
+	if ((entry->vmid_src == AMDGPU_GFXHUB(0)) &&
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(9, 4, 2)))
 		RREG32(hub->vm_l2_pro_fault_status);
 
 	status = RREG32(hub->vm_l2_pro_fault_status);
 	cid = REG_GET_FIELD(status, VM_L2_PROTECTION_FAULT_STATUS, CID);
 	rw = REG_GET_FIELD(status, VM_L2_PROTECTION_FAULT_STATUS, RW);
-	WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1);
+	fed = REG_GET_FIELD(status, VM_L2_PROTECTION_FAULT_STATUS, FED);
+
+	/* for fed error, kfd will handle it, return directly */
+	if (fed && amdgpu_ras_is_poison_mode_supported(adev) &&
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(9, 4, 2)))
+		return 0;
+
+	/* Only print L2 fault status if the status register could be read and
+	 * contains useful information
+	 */
+	if (!status)
+		return 0;
 
+	if (!amdgpu_sriov_vf(adev))
+		WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1);
+
+	amdgpu_vm_update_fault_cache(adev, entry->pasid, addr, status, vmhub);
 
 	dev_err(adev->dev,
 		"VM_L2_PROTECTION_FAULT_STATUS:0x%08X\n",
 		status);
-	if (hub == &adev->vmhub[AMDGPU_GFXHUB_0]) {
+	if (entry->vmid_src == AMDGPU_GFXHUB(0)) {
 		dev_err(adev->dev, "\t Faulty UTCL2 client ID: %s (0x%x)\n",
 			cid >= ARRAY_SIZE(gfxhub_client_ids) ? "unknown" :
 			gfxhub_client_ids[cid],
 			cid);
 	} else {
-		switch (adev->ip_versions[MMHUB_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 		case IP_VERSION(9, 0, 0):
 			mmhub_cid = mmhub_client_ids_vega10[cid][rw];
 			break;
@@ -720,7 +756,8 @@ static void gmc_v9_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->gmc.vm_fault.funcs = &gmc_v9_0_irq_funcs;
 
 	if (!amdgpu_sriov_vf(adev) &&
-	    !adev->gmc.xgmi.connected_to_cpu) {
+	    !adev->gmc.xgmi.connected_to_cpu &&
+	    !adev->gmc.is_app_apu) {
 		adev->gmc.ecc_irq.num_types = 1;
 		adev->gmc.ecc_irq.funcs = &gmc_v9_0_ecc_funcs;
 	}
@@ -755,12 +792,12 @@ static uint32_t gmc_v9_0_get_invalidate_req(unsigned int vmid,
 static bool gmc_v9_0_use_invalidate_semaphore(struct amdgpu_device *adev,
 				       uint32_t vmhub)
 {
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2) ||
-	    adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2) ||
+	    amdgpu_is_multi_aid(adev))
 		return false;
 
-	return ((vmhub == AMDGPU_MMHUB_0 ||
-		 vmhub == AMDGPU_MMHUB_1) &&
+	return ((vmhub == AMDGPU_MMHUB0(0) ||
+		 vmhub == AMDGPU_MMHUB1(0)) &&
 		(!amdgpu_sriov_vf(adev)) &&
 		(!(!(adev->apu_flags & AMD_APU_IS_RAVEN2) &&
 		   (adev->apu_flags & AMD_APU_IS_PICASSO))));
@@ -799,43 +836,37 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 					uint32_t vmhub, uint32_t flush_type)
 {
 	bool use_semaphore = gmc_v9_0_use_invalidate_semaphore(adev, vmhub);
-	const unsigned eng = 17;
-	u32 j, inv_req, inv_req2, tmp;
+	u32 j, inv_req, tmp, sem, req, ack, inst;
+	const unsigned int eng = 17;
 	struct amdgpu_vmhub *hub;
 
-	BUG_ON(vmhub >= adev->num_vmhubs);
+	BUG_ON(vmhub >= AMDGPU_MAX_VMHUBS);
 
 	hub = &adev->vmhub[vmhub];
-	if (adev->gmc.xgmi.num_physical_nodes &&
-	    adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 0)) {
-		/* Vega20+XGMI caches PTEs in TC and TLB. Add a
-		 * heavy-weight TLB flush (type 2), which flushes
-		 * both. Due to a race condition with concurrent
-		 * memory accesses using the same TLB cache line, we
-		 * still need a second TLB flush after this.
-		 */
-		inv_req = gmc_v9_0_get_invalidate_req(vmid, 2);
-		inv_req2 = gmc_v9_0_get_invalidate_req(vmid, flush_type);
-	} else {
-		inv_req = gmc_v9_0_get_invalidate_req(vmid, flush_type);
-		inv_req2 = 0;
-	}
+	inv_req = gmc_v9_0_get_invalidate_req(vmid, flush_type);
+	sem = hub->vm_inv_eng0_sem + hub->eng_distance * eng;
+	req = hub->vm_inv_eng0_req + hub->eng_distance * eng;
+	ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng;
+
+	if (vmhub >= AMDGPU_MMHUB0(0))
+		inst = 0;
+	else
+		inst = vmhub;
 
-	/* This is necessary for a HW workaround under SRIOV as well
-	 * as GFXOFF under bare metal
+	/* This is necessary for SRIOV as well as for GFXOFF to function
+	 * properly under bare metal
 	 */
-	if (adev->gfx.kiq.ring.sched.ready &&
-	    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev)) &&
-	    down_read_trylock(&adev->reset_domain->sem)) {
+	if (adev->gfx.kiq[inst].ring.sched.ready &&
+	    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) {
 		uint32_t req = hub->vm_inv_eng0_req + hub->eng_distance * eng;
 		uint32_t ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng;
 
-		amdgpu_virt_kiq_reg_write_reg_wait(adev, req, ack, inv_req,
-						   1 << vmid);
-		up_read(&adev->reset_domain->sem);
+		amdgpu_gmc_fw_reg_write_reg_wait(adev, req, ack, inv_req,
+						 1 << vmid, inst);
 		return;
 	}
 
+	/* This path is needed before KIQ/MES/GFXOFF are set up */
 	spin_lock(&adev->gmc.invalidate_lock);
 
 	/*
@@ -849,11 +880,10 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 	if (use_semaphore) {
 		for (j = 0; j < adev->usec_timeout; j++) {
 			/* a read return value of 1 means semaphore acquire */
-			if (vmhub == AMDGPU_GFXHUB_0)
-				tmp = RREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_sem + hub->eng_distance * eng);
+			if (vmhub >= AMDGPU_MMHUB0(0))
+				tmp = RREG32_SOC15_IP_NO_KIQ(MMHUB, sem, GET_INST(GC, inst));
 			else
-				tmp = RREG32_SOC15_IP_NO_KIQ(MMHUB, hub->vm_inv_eng0_sem + hub->eng_distance * eng);
-
+				tmp = RREG32_SOC15_IP_NO_KIQ(GC, sem, GET_INST(GC, inst));
 			if (tmp & 0x1)
 				break;
 			udelay(1);
@@ -863,36 +893,29 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 			DRM_ERROR("Timeout waiting for sem acquire in VM flush!\n");
 	}
 
-	do {
-		if (vmhub == AMDGPU_GFXHUB_0)
-			WREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req);
-		else
-			WREG32_SOC15_IP_NO_KIQ(MMHUB, hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req);
-
-		/*
-		 * Issue a dummy read to wait for the ACK register to
-		 * be cleared to avoid a false ACK due to the new fast
-		 * GRBM interface.
-		 */
-		if ((vmhub == AMDGPU_GFXHUB_0) &&
-		    (adev->ip_versions[GC_HWIP][0] < IP_VERSION(9, 4, 2)))
-			RREG32_NO_KIQ(hub->vm_inv_eng0_req +
-				      hub->eng_distance * eng);
-
-		for (j = 0; j < adev->usec_timeout; j++) {
-			if (vmhub == AMDGPU_GFXHUB_0)
-				tmp = RREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_ack + hub->eng_distance * eng);
-			else
-				tmp = RREG32_SOC15_IP_NO_KIQ(MMHUB, hub->vm_inv_eng0_ack + hub->eng_distance * eng);
+	if (vmhub >= AMDGPU_MMHUB0(0))
+		WREG32_SOC15_IP_NO_KIQ(MMHUB, req, inv_req, GET_INST(GC, inst));
+	else
+		WREG32_SOC15_IP_NO_KIQ(GC, req, inv_req, GET_INST(GC, inst));
 
-			if (tmp & (1 << vmid))
-				break;
-			udelay(1);
-		}
+	/*
+	 * Issue a dummy read to wait for the ACK register to
+	 * be cleared to avoid a false ACK due to the new fast
+	 * GRBM interface.
+	 */
+	if ((vmhub == AMDGPU_GFXHUB(0)) &&
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(9, 4, 2)))
+		RREG32_NO_KIQ(req);
 
-		inv_req = inv_req2;
-		inv_req2 = 0;
-	} while (inv_req);
+	for (j = 0; j < adev->usec_timeout; j++) {
+		if (vmhub >= AMDGPU_MMHUB0(0))
+			tmp = RREG32_SOC15_IP_NO_KIQ(MMHUB, ack, GET_INST(GC, inst));
+		else
+			tmp = RREG32_SOC15_IP_NO_KIQ(GC, ack, GET_INST(GC, inst));
+		if (tmp & (1 << vmid))
+			break;
+		udelay(1);
+	}
 
 	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
 	if (use_semaphore) {
@@ -900,10 +923,10 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 		 * add semaphore release after invalidation,
 		 * write with 0 means semaphore release
 		 */
-		if (vmhub == AMDGPU_GFXHUB_0)
-			WREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_sem + hub->eng_distance * eng, 0);
+		if (vmhub >= AMDGPU_MMHUB0(0))
+			WREG32_SOC15_IP_NO_KIQ(MMHUB, sem, 0, GET_INST(GC, inst));
 		else
-			WREG32_SOC15_IP_NO_KIQ(MMHUB, hub->vm_inv_eng0_sem + hub->eng_distance * eng, 0);
+			WREG32_SOC15_IP_NO_KIQ(GC, sem, 0, GET_INST(GC, inst));
 	}
 
 	spin_unlock(&adev->gmc.invalidate_lock);
@@ -921,97 +944,46 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
  * @pasid: pasid to be flush
  * @flush_type: the flush type
  * @all_hub: flush all hubs
+ * @inst: is used to select which instance of KIQ to use for the invalidation
  *
  * Flush the TLB for the requested pasid.
  */
-static int gmc_v9_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
-					uint16_t pasid, uint32_t flush_type,
-					bool all_hub)
+static void gmc_v9_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
+					 uint16_t pasid, uint32_t flush_type,
+					 bool all_hub, uint32_t inst)
 {
-	int vmid, i;
-	signed long r;
-	uint32_t seq;
-	uint16_t queried_pasid;
-	bool ret;
-	u32 usec_timeout = amdgpu_sriov_vf(adev) ? SRIOV_USEC_TIMEOUT : adev->usec_timeout;
-	struct amdgpu_ring *ring = &adev->gfx.kiq.ring;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-
-	if (amdgpu_in_reset(adev))
-		return -EIO;
-
-	if (ring->sched.ready && down_read_trylock(&adev->reset_domain->sem)) {
-		/* Vega20+XGMI caches PTEs in TC and TLB. Add a
-		 * heavy-weight TLB flush (type 2), which flushes
-		 * both. Due to a race condition with concurrent
-		 * memory accesses using the same TLB cache line, we
-		 * still need a second TLB flush after this.
-		 */
-		bool vega20_xgmi_wa = (adev->gmc.xgmi.num_physical_nodes &&
-				       adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 0));
-		/* 2 dwords flush + 8 dwords fence */
-		unsigned int ndw = kiq->pmf->invalidate_tlbs_size + 8;
-
-		if (vega20_xgmi_wa)
-			ndw += kiq->pmf->invalidate_tlbs_size;
-
-		spin_lock(&adev->gfx.kiq.ring_lock);
-		/* 2 dwords flush + 8 dwords fence */
-		amdgpu_ring_alloc(ring, ndw);
-		if (vega20_xgmi_wa)
-			kiq->pmf->kiq_invalidate_tlbs(ring,
-						      pasid, 2, all_hub);
-		kiq->pmf->kiq_invalidate_tlbs(ring,
-					pasid, flush_type, all_hub);
-		r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
-		if (r) {
-			amdgpu_ring_undo(ring);
-			spin_unlock(&adev->gfx.kiq.ring_lock);
-			up_read(&adev->reset_domain->sem);
-			return -ETIME;
-		}
-
-		amdgpu_ring_commit(ring);
-		spin_unlock(&adev->gfx.kiq.ring_lock);
-		r = amdgpu_fence_wait_polling(ring, seq, usec_timeout);
-		if (r < 1) {
-			dev_err(adev->dev, "wait for kiq fence error: %ld.\n", r);
-			up_read(&adev->reset_domain->sem);
-			return -ETIME;
-		}
-		up_read(&adev->reset_domain->sem);
-		return 0;
-	}
+	uint16_t queried;
+	int i, vmid;
 
 	for (vmid = 1; vmid < 16; vmid++) {
+		bool valid;
 
-		ret = gmc_v9_0_get_atc_vmid_pasid_mapping_info(adev, vmid,
-				&queried_pasid);
-		if (ret && queried_pasid == pasid) {
-			if (all_hub) {
-				for (i = 0; i < adev->num_vmhubs; i++)
-					gmc_v9_0_flush_gpu_tlb(adev, vmid,
-							i, flush_type);
-			} else {
-				gmc_v9_0_flush_gpu_tlb(adev, vmid,
-						AMDGPU_GFXHUB_0, flush_type);
-			}
-			break;
+		valid = gmc_v9_0_get_atc_vmid_pasid_mapping_info(adev, vmid,
+								 &queried);
+		if (!valid || queried != pasid)
+			continue;
+
+		if (all_hub) {
+			for_each_set_bit(i, adev->vmhubs_mask,
+					 AMDGPU_MAX_VMHUBS)
+				gmc_v9_0_flush_gpu_tlb(adev, vmid, i,
+						       flush_type);
+		} else {
+			gmc_v9_0_flush_gpu_tlb(adev, vmid,
+					       AMDGPU_GFXHUB(0),
+					       flush_type);
 		}
 	}
-
-	return 0;
-
 }
 
 static uint64_t gmc_v9_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
-					    unsigned vmid, uint64_t pd_addr)
+					    unsigned int vmid, uint64_t pd_addr)
 {
 	bool use_semaphore = gmc_v9_0_use_invalidate_semaphore(ring->adev, ring->vm_hub);
 	struct amdgpu_device *adev = ring->adev;
 	struct amdgpu_vmhub *hub = &adev->vmhub[ring->vm_hub];
 	uint32_t req = gmc_v9_0_get_invalidate_req(vmid, 0);
-	unsigned eng = ring->vm_inv_eng;
+	unsigned int eng = ring->vm_inv_eng;
 
 	/*
 	 * It may lose gpuvm invalidate acknowldege state across power-gating
@@ -1053,17 +1025,17 @@ static uint64_t gmc_v9_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
 	return pd_addr;
 }
 
-static void gmc_v9_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
-					unsigned pasid)
+static void gmc_v9_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned int vmid,
+					unsigned int pasid)
 {
 	struct amdgpu_device *adev = ring->adev;
 	uint32_t reg;
 
 	/* Do nothing because there's no lut register for mmhub1. */
-	if (ring->vm_hub == AMDGPU_MMHUB_1)
+	if (ring->vm_hub == AMDGPU_MMHUB1(0))
 		return;
 
-	if (ring->vm_hub == AMDGPU_GFXHUB_0)
+	if (ring->vm_hub == AMDGPU_GFXHUB(0))
 		reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT) + vmid;
 	else
 		reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT_MM) + vmid;
@@ -1103,27 +1075,6 @@ static void gmc_v9_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
  * 0 valid
  */
 
-static uint64_t gmc_v9_0_map_mtype(struct amdgpu_device *adev, uint32_t flags)
-
-{
-	switch (flags) {
-	case AMDGPU_VM_MTYPE_DEFAULT:
-		return AMDGPU_PTE_MTYPE_VG10(MTYPE_NC);
-	case AMDGPU_VM_MTYPE_NC:
-		return AMDGPU_PTE_MTYPE_VG10(MTYPE_NC);
-	case AMDGPU_VM_MTYPE_WC:
-		return AMDGPU_PTE_MTYPE_VG10(MTYPE_WC);
-	case AMDGPU_VM_MTYPE_RW:
-		return AMDGPU_PTE_MTYPE_VG10(MTYPE_RW);
-	case AMDGPU_VM_MTYPE_CC:
-		return AMDGPU_PTE_MTYPE_VG10(MTYPE_CC);
-	case AMDGPU_VM_MTYPE_UC:
-		return AMDGPU_PTE_MTYPE_VG10(MTYPE_UC);
-	default:
-		return AMDGPU_PTE_MTYPE_VG10(MTYPE_NC);
-	}
-}
-
 static void gmc_v9_0_get_vm_pde(struct amdgpu_device *adev, int level,
 				uint64_t *addr, uint64_t *flags)
 {
@@ -1151,21 +1102,28 @@ static void gmc_v9_0_get_vm_pde(struct amdgpu_device *adev, int level,
 }
 
 static void gmc_v9_0_get_coherence_flags(struct amdgpu_device *adev,
+					 struct amdgpu_vm *vm,
 					 struct amdgpu_bo *bo,
-					 struct amdgpu_bo_va_mapping *mapping,
+					 uint32_t vm_flags,
 					 uint64_t *flags)
 {
 	struct amdgpu_device *bo_adev = amdgpu_ttm_adev(bo->tbo.bdev);
-	bool is_vram = bo->tbo.resource->mem_type == TTM_PL_VRAM;
-	bool coherent = bo->flags & AMDGPU_GEM_CREATE_COHERENT;
+	bool is_vram = bo->tbo.resource &&
+		bo->tbo.resource->mem_type == TTM_PL_VRAM;
+	bool coherent = bo->flags & (AMDGPU_GEM_CREATE_COHERENT |
+				     AMDGPU_GEM_CREATE_EXT_COHERENT);
+	bool ext_coherent = bo->flags & AMDGPU_GEM_CREATE_EXT_COHERENT;
 	bool uncached = bo->flags & AMDGPU_GEM_CREATE_UNCACHED;
-	unsigned int mtype;
+	unsigned int mtype_local, mtype;
+	uint32_t gc_ip_version = amdgpu_ip_version(adev, GC_HWIP, 0);
 	bool snoop = false;
+	bool is_local;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	dma_resv_assert_held(bo->tbo.base.resv);
+
+	switch (gc_ip_version) {
 	case IP_VERSION(9, 4, 1):
 	case IP_VERSION(9, 4, 2):
-	case IP_VERSION(9, 4, 3):
 		if (is_vram) {
 			if (bo_adev == adev) {
 				if (uncached)
@@ -1177,8 +1135,7 @@ static void gmc_v9_0_get_coherence_flags(struct amdgpu_device *adev,
 				/* FIXME: is this still needed? Or does
 				 * amdgpu_ttm_tt_pde_flags already handle this?
 				 */
-				if ((adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2) ||
-				     adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3)) &&
+				if (gc_ip_version == IP_VERSION(9, 4, 2) &&
 				    adev->gmc.xgmi.connected_to_cpu)
 					snoop = true;
 			} else {
@@ -1186,7 +1143,7 @@ static void gmc_v9_0_get_coherence_flags(struct amdgpu_device *adev,
 					mtype = MTYPE_UC;
 				else
 					mtype = MTYPE_NC;
-				if (mapping->bo_va->is_xgmi)
+				if (amdgpu_xgmi_same_hive(adev, bo_adev))
 					snoop = true;
 			}
 		} else {
@@ -1200,6 +1157,47 @@ static void gmc_v9_0_get_coherence_flags(struct amdgpu_device *adev,
 			snoop = true;
 		}
 		break;
+	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+	case IP_VERSION(9, 5, 0):
+		/* Only local VRAM BOs or system memory on non-NUMA APUs
+		 * can be assumed to be local in their entirety. Choose
+		 * MTYPE_NC as safe fallback for all system memory BOs on
+		 * NUMA systems. Their MTYPE can be overridden per-page in
+		 * gmc_v9_0_override_vm_pte_flags.
+		 */
+		mtype_local = MTYPE_RW;
+		if (amdgpu_mtype_local == 1) {
+			DRM_INFO_ONCE("Using MTYPE_NC for local memory\n");
+			mtype_local = MTYPE_NC;
+		} else if (amdgpu_mtype_local == 2) {
+			DRM_INFO_ONCE("Using MTYPE_CC for local memory\n");
+			mtype_local = MTYPE_CC;
+		} else {
+			DRM_INFO_ONCE("Using MTYPE_RW for local memory\n");
+		}
+		is_local = (!is_vram && (adev->flags & AMD_IS_APU) &&
+			    num_possible_nodes() <= 1) ||
+			   (is_vram && adev == bo_adev &&
+			    KFD_XCP_MEM_ID(adev, bo->xcp_id) == vm->mem_id);
+		snoop = true;
+		if (uncached) {
+			mtype = MTYPE_UC;
+		} else if (ext_coherent) {
+			mtype = is_local ? MTYPE_CC : MTYPE_UC;
+		} else if (adev->flags & AMD_IS_APU) {
+			mtype = is_local ? mtype_local : MTYPE_NC;
+		} else {
+			/* dGPU */
+			if (is_local)
+				mtype = mtype_local;
+			else if (gc_ip_version < IP_VERSION(9, 5, 0) && !is_vram)
+				mtype = MTYPE_UC;
+			else
+				mtype = MTYPE_NC;
+		}
+
+		break;
 	default:
 		if (uncached || coherent)
 			mtype = MTYPE_UC;
@@ -1214,37 +1212,131 @@ static void gmc_v9_0_get_coherence_flags(struct amdgpu_device *adev,
 	}
 
 	if (mtype != MTYPE_NC)
-		*flags = (*flags & ~AMDGPU_PTE_MTYPE_VG10_MASK) |
-			 AMDGPU_PTE_MTYPE_VG10(mtype);
+		*flags = AMDGPU_PTE_MTYPE_VG10(*flags, mtype);
+
 	*flags |= snoop ? AMDGPU_PTE_SNOOPED : 0;
 }
 
 static void gmc_v9_0_get_vm_pte(struct amdgpu_device *adev,
-				struct amdgpu_bo_va_mapping *mapping,
+				struct amdgpu_vm *vm,
+				struct amdgpu_bo *bo,
+				uint32_t vm_flags,
 				uint64_t *flags)
 {
-	struct amdgpu_bo *bo = mapping->bo_va->base.bo;
-
-	*flags &= ~AMDGPU_PTE_EXECUTABLE;
-	*flags |= mapping->flags & AMDGPU_PTE_EXECUTABLE;
+	if (vm_flags & AMDGPU_VM_PAGE_EXECUTABLE)
+		*flags |= AMDGPU_PTE_EXECUTABLE;
+	else
+		*flags &= ~AMDGPU_PTE_EXECUTABLE;
 
-	*flags &= ~AMDGPU_PTE_MTYPE_VG10_MASK;
-	*flags |= mapping->flags & AMDGPU_PTE_MTYPE_VG10_MASK;
+	switch (vm_flags & AMDGPU_VM_MTYPE_MASK) {
+	case AMDGPU_VM_MTYPE_DEFAULT:
+	case AMDGPU_VM_MTYPE_NC:
+	default:
+		*flags = AMDGPU_PTE_MTYPE_VG10(*flags, MTYPE_NC);
+		break;
+	case AMDGPU_VM_MTYPE_WC:
+		*flags |= AMDGPU_PTE_MTYPE_VG10(*flags, MTYPE_WC);
+		break;
+	case AMDGPU_VM_MTYPE_RW:
+		*flags |= AMDGPU_PTE_MTYPE_VG10(*flags, MTYPE_RW);
+		break;
+	case AMDGPU_VM_MTYPE_CC:
+		*flags |= AMDGPU_PTE_MTYPE_VG10(*flags, MTYPE_CC);
+		break;
+	case AMDGPU_VM_MTYPE_UC:
+		*flags |= AMDGPU_PTE_MTYPE_VG10(*flags, MTYPE_UC);
+		break;
+	}
 
-	if (mapping->flags & AMDGPU_PTE_PRT) {
+	if (vm_flags & AMDGPU_VM_PAGE_PRT) {
 		*flags |= AMDGPU_PTE_PRT;
 		*flags &= ~AMDGPU_PTE_VALID;
 	}
 
-	if (bo && bo->tbo.resource)
-		gmc_v9_0_get_coherence_flags(adev, mapping->bo_va->base.bo,
-					     mapping, flags);
+	if ((*flags & AMDGPU_PTE_VALID) && bo)
+		gmc_v9_0_get_coherence_flags(adev, vm, bo, vm_flags, flags);
+}
+
+static void gmc_v9_0_override_vm_pte_flags(struct amdgpu_device *adev,
+					   struct amdgpu_vm *vm,
+					   uint64_t addr, uint64_t *flags)
+{
+	int local_node, nid;
+
+	/* Only GFX 9.4.3 APUs associate GPUs with NUMA nodes. Local system
+	 * memory can use more efficient MTYPEs.
+	 */
+	if (!(adev->flags & AMD_IS_APU) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 3))
+		return;
+
+	/* Only direct-mapped memory allows us to determine the NUMA node from
+	 * the DMA address.
+	 */
+	if (!adev->ram_is_direct_mapped) {
+		dev_dbg_ratelimited(adev->dev, "RAM is not direct mapped\n");
+		return;
+	}
+
+	/* MTYPE_NC is the same default and can be overridden.
+	 * MTYPE_UC will be present if the memory is extended-coherent
+	 * and can also be overridden.
+	 */
+	if ((*flags & AMDGPU_PTE_MTYPE_VG10_MASK) !=
+	    AMDGPU_PTE_MTYPE_VG10(0ULL, MTYPE_NC) &&
+	    (*flags & AMDGPU_PTE_MTYPE_VG10_MASK) !=
+	    AMDGPU_PTE_MTYPE_VG10(0ULL, MTYPE_UC)) {
+		dev_dbg_ratelimited(adev->dev, "MTYPE is not NC or UC\n");
+		return;
+	}
+
+	/* FIXME: Only supported on native mode for now. For carve-out, the
+	 * NUMA affinity of the GPU/VM needs to come from the PCI info because
+	 * memory partitions are not associated with different NUMA nodes.
+	 */
+	if (adev->gmc.is_app_apu && vm->mem_id >= 0) {
+		local_node = adev->gmc.mem_partitions[vm->mem_id].numa.node;
+	} else {
+		dev_dbg_ratelimited(adev->dev, "Only native mode APU is supported.\n");
+		return;
+	}
+
+	/* Only handle real RAM. Mappings of PCIe resources don't have struct
+	 * page or NUMA nodes.
+	 */
+	if (!page_is_ram(addr >> PAGE_SHIFT)) {
+		dev_dbg_ratelimited(adev->dev, "Page is not RAM.\n");
+		return;
+	}
+	nid = pfn_to_nid(addr >> PAGE_SHIFT);
+	dev_dbg_ratelimited(adev->dev, "vm->mem_id=%d, local_node=%d, nid=%d\n",
+			    vm->mem_id, local_node, nid);
+	if (nid == local_node) {
+		uint64_t old_flags = *flags;
+		if ((*flags & AMDGPU_PTE_MTYPE_VG10_MASK) ==
+			AMDGPU_PTE_MTYPE_VG10(0ULL, MTYPE_NC)) {
+			unsigned int mtype_local = MTYPE_RW;
+
+			if (amdgpu_mtype_local == 1)
+				mtype_local = MTYPE_NC;
+			else if (amdgpu_mtype_local == 2)
+				mtype_local = MTYPE_CC;
+
+			*flags = AMDGPU_PTE_MTYPE_VG10(*flags, mtype_local);
+		} else {
+			/* MTYPE_UC case */
+			*flags = AMDGPU_PTE_MTYPE_VG10(*flags, MTYPE_CC);
+		}
+
+		dev_dbg_ratelimited(adev->dev, "flags updated from %llx to %llx\n",
+				    old_flags, *flags);
+	}
 }
 
-static unsigned gmc_v9_0_get_vbios_fb_size(struct amdgpu_device *adev)
+static unsigned int gmc_v9_0_get_vbios_fb_size(struct amdgpu_device *adev)
 {
 	u32 d1vga_control = RREG32_SOC15(DCE, 0, mmD1VGA_CONTROL);
-	unsigned size;
+	unsigned int size;
 
 	/* TODO move to DC so GMC doesn't need to hard-code DCN registers */
 
@@ -1253,7 +1345,7 @@ static unsigned gmc_v9_0_get_vbios_fb_size(struct amdgpu_device *adev)
 	} else {
 		u32 viewport;
 
-		switch (adev->ip_versions[DCE_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
 		case IP_VERSION(1, 0, 0):
 		case IP_VERSION(1, 0, 1):
 			viewport = RREG32_SOC15(DCE, 0, mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION);
@@ -1283,15 +1375,29 @@ static unsigned gmc_v9_0_get_vbios_fb_size(struct amdgpu_device *adev)
 	return size;
 }
 
+static bool gmc_v9_0_need_reset_on_init(struct amdgpu_device *adev)
+{
+	if (adev->nbio.funcs && adev->nbio.funcs->is_nps_switch_requested &&
+	    adev->nbio.funcs->is_nps_switch_requested(adev)) {
+		adev->gmc.reset_flags |= AMDGPU_GMC_INIT_RESET_NPS;
+		return true;
+	}
+
+	return false;
+}
+
 static const struct amdgpu_gmc_funcs gmc_v9_0_gmc_funcs = {
 	.flush_gpu_tlb = gmc_v9_0_flush_gpu_tlb,
 	.flush_gpu_tlb_pasid = gmc_v9_0_flush_gpu_tlb_pasid,
 	.emit_flush_gpu_tlb = gmc_v9_0_emit_flush_gpu_tlb,
 	.emit_pasid_mapping = gmc_v9_0_emit_pasid_mapping,
-	.map_mtype = gmc_v9_0_map_mtype,
 	.get_vm_pde = gmc_v9_0_get_vm_pde,
 	.get_vm_pte = gmc_v9_0_get_vm_pte,
+	.override_vm_pte_flags = gmc_v9_0_override_vm_pte_flags,
 	.get_vbios_fb_size = gmc_v9_0_get_vbios_fb_size,
+	.query_mem_partition_mode = &amdgpu_gmc_query_memory_partition,
+	.request_mem_partition_mode = &amdgpu_gmc_request_memory_partition,
+	.need_reset_on_init = &gmc_v9_0_need_reset_on_init,
 };
 
 static void gmc_v9_0_set_gmc_funcs(struct amdgpu_device *adev)
@@ -1301,7 +1407,7 @@ static void gmc_v9_0_set_gmc_funcs(struct amdgpu_device *adev)
 
 static void gmc_v9_0_set_umc_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[UMC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, UMC_HWIP, 0)) {
 	case IP_VERSION(6, 0, 0):
 		adev->umc.funcs = &umc_v6_0_funcs;
 		break;
@@ -1337,6 +1443,17 @@ static void gmc_v9_0_set_umc_funcs(struct amdgpu_device *adev)
 		else
 			adev->umc.channel_idx_tbl = &umc_v6_7_channel_idx_tbl_second[0][0];
 		break;
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 5, 0):
+		adev->umc.max_ras_err_cnt_per_query =
+			UMC_V12_0_TOTAL_CHANNEL_NUM(adev) * UMC_V12_0_BAD_PAGE_NUM_PER_CHANNEL;
+		adev->umc.channel_inst_num = UMC_V12_0_CHANNEL_INSTANCE_NUM;
+		adev->umc.umc_inst_num = UMC_V12_0_UMC_INSTANCE_NUM;
+		adev->umc.node_inst_num /= UMC_V12_0_UMC_INSTANCE_NUM;
+		adev->umc.channel_offs = UMC_V12_0_PER_CHANNEL_OFFSET;
+		if (!adev->gmc.xgmi.connected_to_cpu && !adev->gmc.is_app_apu)
+			adev->umc.ras = &umc_v12_0_ras;
+		break;
 	default:
 		break;
 	}
@@ -1344,7 +1461,7 @@ static void gmc_v9_0_set_umc_funcs(struct amdgpu_device *adev)
 
 static void gmc_v9_0_set_mmhub_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(9, 4, 1):
 		adev->mmhub.funcs = &mmhub_v9_4_funcs;
 		break;
@@ -1352,6 +1469,7 @@ static void gmc_v9_0_set_mmhub_funcs(struct amdgpu_device *adev)
 		adev->mmhub.funcs = &mmhub_v1_7_funcs;
 		break;
 	case IP_VERSION(1, 8, 0):
+	case IP_VERSION(1, 8, 1):
 		adev->mmhub.funcs = &mmhub_v1_8_funcs;
 		break;
 	default:
@@ -1362,7 +1480,7 @@ static void gmc_v9_0_set_mmhub_funcs(struct amdgpu_device *adev)
 
 static void gmc_v9_0_set_mmhub_ras_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(9, 4, 0):
 		adev->mmhub.ras = &mmhub_v1_0_ras;
 		break;
@@ -1372,6 +1490,10 @@ static void gmc_v9_0_set_mmhub_ras_funcs(struct amdgpu_device *adev)
 	case IP_VERSION(9, 4, 2):
 		adev->mmhub.ras = &mmhub_v1_7_ras;
 		break;
+	case IP_VERSION(1, 8, 0):
+	case IP_VERSION(1, 8, 1):
+		adev->mmhub.ras = &mmhub_v1_8_ras;
+		break;
 	default:
 		/* mmhub ras is not available */
 		break;
@@ -1380,7 +1502,7 @@ static void gmc_v9_0_set_mmhub_ras_funcs(struct amdgpu_device *adev)
 
 static void gmc_v9_0_set_gfxhub_funcs(struct amdgpu_device *adev)
 {
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3))
+	if (amdgpu_is_multi_aid(adev))
 		adev->gfxhub.funcs = &gfxhub_v1_2_funcs;
 	else
 		adev->gfxhub.funcs = &gfxhub_v1_0_funcs;
@@ -1396,7 +1518,7 @@ static void gmc_v9_0_set_mca_ras_funcs(struct amdgpu_device *adev)
 	struct amdgpu_mca *mca = &adev->mca;
 
 	/* is UMC the right IP to check for MCA?  Maybe DF? */
-	switch (adev->ip_versions[UMC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, UMC_HWIP, 0)) {
 	case IP_VERSION(6, 7, 0):
 		if (!adev->gmc.xgmi.connected_to_cpu) {
 			mca->mp0.ras = &mca_v3_0_mp0_ras;
@@ -1415,21 +1537,76 @@ static void gmc_v9_0_set_xgmi_ras_funcs(struct amdgpu_device *adev)
 		adev->gmc.xgmi.ras = &xgmi_ras;
 }
 
-static int gmc_v9_0_early_init(void *handle)
+static void gmc_v9_0_init_nps_details(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	enum amdgpu_memory_partition mode;
+	uint32_t supp_modes;
+	int i;
 
-	/* ARCT and VEGA20 don't have XGMI defined in their IP discovery tables */
-	if (adev->asic_type == CHIP_VEGA20 ||
-	    adev->asic_type == CHIP_ARCTURUS)
+	adev->gmc.supported_nps_modes = 0;
+
+	if (amdgpu_sriov_vf(adev) || (adev->flags & AMD_IS_APU))
+		return;
+
+	mode = amdgpu_gmc_get_memory_partition(adev, &supp_modes);
+
+	/* Mode detected by hardware and supported modes available */
+	if ((mode != UNKNOWN_MEMORY_PARTITION_MODE) && supp_modes) {
+		while ((i = ffs(supp_modes))) {
+			if (AMDGPU_ALL_NPS_MASK & BIT(i))
+				adev->gmc.supported_nps_modes |= BIT(i);
+			supp_modes &= supp_modes - 1;
+		}
+	} else {
+		/*TODO: Check PSP version also which supports NPS switch. Otherwise keep
+	 * supported modes as 0.
+	 */
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+		case IP_VERSION(9, 4, 3):
+		case IP_VERSION(9, 4, 4):
+			adev->gmc.supported_nps_modes =
+				BIT(AMDGPU_NPS1_PARTITION_MODE) |
+				BIT(AMDGPU_NPS4_PARTITION_MODE);
+			break;
+		default:
+			break;
+		}
+	}
+}
+
+static int gmc_v9_0_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	/*
+	 * 9.4.0, 9.4.1 and 9.4.3 don't have XGMI defined
+	 * in their IP discovery tables
+	 */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 0) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1) ||
+	    amdgpu_is_multi_aid(adev))
 		adev->gmc.xgmi.supported = true;
 
-	if (adev->ip_versions[XGMI_HWIP][0] == IP_VERSION(6, 1, 0)) {
+	if (amdgpu_ip_version(adev, XGMI_HWIP, 0) == IP_VERSION(6, 1, 0)) {
 		adev->gmc.xgmi.supported = true;
 		adev->gmc.xgmi.connected_to_cpu =
 			adev->smuio.funcs->is_host_gpu_xgmi_supported(adev);
 	}
 
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3)) {
+		enum amdgpu_pkg_type pkg_type =
+			adev->smuio.funcs->get_pkg_type(adev);
+		/* On GFXIP 9.4.3. APU, there is no physical VRAM domain present
+		 * and the APU, can be in used two possible modes:
+		 *  - carveout mode
+		 *  - native APU mode
+		 * "is_app_apu" can be used to identify the APU in the native
+		 * mode.
+		 */
+		adev->gmc.is_app_apu = (pkg_type == AMDGPU_PKG_TYPE_APU &&
+					!pci_resource_len(adev->pdev, 0));
+	}
+
 	gmc_v9_0_set_gmc_funcs(adev);
 	gmc_v9_0_set_irq_funcs(adev);
 	gmc_v9_0_set_umc_funcs(adev);
@@ -1446,13 +1623,14 @@ static int gmc_v9_0_early_init(void *handle)
 	adev->gmc.private_aperture_start = 0x1000000000000000ULL;
 	adev->gmc.private_aperture_end =
 		adev->gmc.private_aperture_start + (4ULL << 30) - 1;
+	adev->gmc.noretry_flags = AMDGPU_VM_NORETRY_FLAGS_TF;
 
 	return 0;
 }
 
-static int gmc_v9_0_late_init(void *handle)
+static int gmc_v9_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_gmc_allocate_vm_inv_eng(adev);
@@ -1464,7 +1642,7 @@ static int gmc_v9_0_late_init(void *handle)
 	 * writes, while disables HBM ECC for vega10.
 	 */
 	if (!amdgpu_sriov_vf(adev) &&
-	    (adev->ip_versions[UMC_HWIP][0] == IP_VERSION(6, 0, 0))) {
+	    (amdgpu_ip_version(adev, UMC_HWIP, 0) == IP_VERSION(6, 0, 0))) {
 		if (!(adev->ras_enabled & (1 << AMDGPU_RAS_BLOCK__UMC))) {
 			if (adev->df.funcs &&
 			    adev->df.funcs->enable_ecc_force_par_wr_rmw)
@@ -1473,13 +1651,8 @@ static int gmc_v9_0_late_init(void *handle)
 	}
 
 	if (!amdgpu_persistent_edc_harvesting_supported(adev)) {
-		if (adev->mmhub.ras && adev->mmhub.ras->ras_block.hw_ops &&
-		    adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count)
-			adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count(adev);
-
-		if (adev->hdp.ras && adev->hdp.ras->ras_block.hw_ops &&
-		    adev->hdp.ras->ras_block.hw_ops->reset_ras_error_count)
-			adev->hdp.ras->ras_block.hw_ops->reset_ras_error_count(adev);
+		amdgpu_ras_reset_error_count(adev, AMDGPU_RAS_BLOCK__MMHUB);
+		amdgpu_ras_reset_error_count(adev, AMDGPU_RAS_BLOCK__HDP);
 	}
 
 	r = amdgpu_gmc_ras_late_init(adev);
@@ -1494,14 +1667,17 @@ static void gmc_v9_0_vram_gtt_location(struct amdgpu_device *adev,
 {
 	u64 base = adev->mmhub.funcs->get_fb_location(adev);
 
+	amdgpu_gmc_set_agp_default(adev, mc);
+
 	/* add the xgmi offset of the physical node */
 	base += adev->gmc.xgmi.physical_node_id * adev->gmc.xgmi.node_segment_size;
-	if (adev->gmc.xgmi.connected_to_cpu) {
+	if (amdgpu_gmc_is_pdb0_enabled(adev)) {
 		amdgpu_gmc_sysvm_location(adev, mc);
 	} else {
 		amdgpu_gmc_vram_location(adev, mc, base);
-		amdgpu_gmc_gart_location(adev, mc);
-		amdgpu_gmc_agp_location(adev, mc);
+		amdgpu_gmc_gart_location(adev, mc, AMDGPU_GART_PLACEMENT_BEST_FIT);
+		if (!amdgpu_sriov_vf(adev) && (amdgpu_agp == 1))
+			amdgpu_gmc_agp_location(adev, mc);
 	}
 	/* base offset of vram pages */
 	adev->vm_manager.vram_base_offset = adev->gfxhub.funcs->get_mc_fb_offset(adev);
@@ -1525,8 +1701,13 @@ static int gmc_v9_0_mc_init(struct amdgpu_device *adev)
 	int r;
 
 	/* size in MB on si */
-	adev->gmc.mc_vram_size =
-		adev->nbio.funcs->get_memsize(adev) * 1024ULL * 1024ULL;
+	if (!adev->gmc.is_app_apu) {
+		adev->gmc.mc_vram_size =
+			adev->nbio.funcs->get_memsize(adev) * 1024ULL * 1024ULL;
+	} else {
+		DRM_DEBUG("Set mc_vram_size = 0 for APP APU\n");
+		adev->gmc.mc_vram_size = 0;
+	}
 	adev->gmc.real_vram_size = adev->gmc.mc_vram_size;
 
 	if (!(adev->flags & AMD_IS_APU) &&
@@ -1551,7 +1732,8 @@ static int gmc_v9_0_mc_init(struct amdgpu_device *adev)
 	 */
 
 	/* check whether both host-gpu and gpu-gpu xgmi links exist */
-	if (((adev->flags & AMD_IS_APU) && !amdgpu_passthrough(adev)) ||
+	if ((!amdgpu_sriov_vf(adev) &&
+		(adev->flags & AMD_IS_APU) && !amdgpu_passthrough(adev)) ||
 	    (adev->gmc.xgmi.supported &&
 	     adev->gmc.xgmi.connected_to_cpu)) {
 		adev->gmc.aper_base =
@@ -1566,13 +1748,15 @@ static int gmc_v9_0_mc_init(struct amdgpu_device *adev)
 
 	/* set the gart size */
 	if (amdgpu_gart_size == -1) {
-		switch (adev->ip_versions[GC_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 		case IP_VERSION(9, 0, 1):  /* all engines support GPUVM */
 		case IP_VERSION(9, 2, 1):  /* all engines support GPUVM */
 		case IP_VERSION(9, 4, 0):
 		case IP_VERSION(9, 4, 1):
 		case IP_VERSION(9, 4, 2):
 		case IP_VERSION(9, 4, 3):
+		case IP_VERSION(9, 4, 4):
+		case IP_VERSION(9, 5, 0):
 		default:
 			adev->gmc.gart_size = 512ULL << 20;
 			break;
@@ -1602,7 +1786,7 @@ static int gmc_v9_0_gart_init(struct amdgpu_device *adev)
 		return 0;
 	}
 
-	if (adev->gmc.xgmi.connected_to_cpu) {
+	if (amdgpu_gmc_is_pdb0_enabled(adev)) {
 		adev->gmc.vmid0_page_table_depth = 1;
 		adev->gmc.vmid0_page_table_block_size = 12;
 	} else {
@@ -1615,15 +1799,21 @@ static int gmc_v9_0_gart_init(struct amdgpu_device *adev)
 	if (r)
 		return r;
 	adev->gart.table_size = adev->gart.num_gpu_pages * 8;
-	adev->gart.gart_pte_flags = AMDGPU_PTE_MTYPE_VG10(MTYPE_UC) |
+	adev->gart.gart_pte_flags = AMDGPU_PTE_MTYPE_VG10(0ULL, MTYPE_UC) |
 				 AMDGPU_PTE_EXECUTABLE;
 
-	r = amdgpu_gart_table_vram_alloc(adev);
-	if (r)
-		return r;
+	if (!adev->gmc.real_vram_size) {
+		dev_info(adev->dev, "Put GART in system memory for APU\n");
+		r = amdgpu_gart_table_ram_alloc(adev);
+		if (r)
+			dev_err(adev->dev, "Failed to allocate GART in system memory\n");
+	} else {
+		r = amdgpu_gart_table_vram_alloc(adev);
+		if (r)
+			return r;
 
-	if (adev->gmc.xgmi.connected_to_cpu) {
-		r = amdgpu_gmc_pdb0_alloc(adev);
+		if (amdgpu_gmc_is_pdb0_enabled(adev))
+			r = amdgpu_gmc_pdb0_alloc(adev);
 	}
 
 	return r;
@@ -1639,15 +1829,37 @@ static int gmc_v9_0_gart_init(struct amdgpu_device *adev)
  */
 static void gmc_v9_0_save_registers(struct amdgpu_device *adev)
 {
-	if ((adev->ip_versions[DCE_HWIP][0] == IP_VERSION(1, 0, 0)) ||
-	    (adev->ip_versions[DCE_HWIP][0] == IP_VERSION(1, 0, 1)))
+	if ((amdgpu_ip_version(adev, DCE_HWIP, 0) == IP_VERSION(1, 0, 0)) ||
+	    (amdgpu_ip_version(adev, DCE_HWIP, 0) == IP_VERSION(1, 0, 1)))
 		adev->gmc.sdpif_register = RREG32_SOC15(DCE, 0, mmDCHUBBUB_SDPIF_MMIO_CNTRL_0);
 }
 
-static int gmc_v9_0_sw_init(void *handle)
+static void gmc_v9_4_3_init_vram_info(struct amdgpu_device *adev)
+{
+	static const u32 regBIF_BIOS_SCRATCH_4 = 0x50;
+	u32 vram_info;
+
+	adev->gmc.vram_type = AMDGPU_VRAM_TYPE_HBM;
+	adev->gmc.vram_width = 128 * 64;
+
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 5, 0))
+		adev->gmc.vram_type = AMDGPU_VRAM_TYPE_HBM3E;
+
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 4) &&
+		adev->rev_id == 0x3)
+		adev->gmc.vram_type = AMDGPU_VRAM_TYPE_HBM3E;
+
+	if (!(adev->flags & AMD_IS_APU) && !amdgpu_sriov_vf(adev)) {
+		vram_info = RREG32(regBIF_BIOS_SCRATCH_4);
+		adev->gmc.vram_vendor = vram_info & 0xF;
+	}
+}
+
+static int gmc_v9_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r, vram_width = 0, vram_type = 0, vram_vendor = 0, dma_addr_bits;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	unsigned long inst_mask = adev->aid_mask;
 
 	adev->gfxhub.funcs->init(adev);
 
@@ -1655,38 +1867,51 @@ static int gmc_v9_0_sw_init(void *handle)
 
 	spin_lock_init(&adev->gmc.invalidate_lock);
 
-	r = amdgpu_atomfirmware_get_vram_info(adev,
-		&vram_width, &vram_type, &vram_vendor);
-	if (amdgpu_sriov_vf(adev))
-		/* For Vega10 SR-IOV, vram_width can't be read from ATOM as RAVEN,
-		 * and DF related registers is not readable, seems hardcord is the
-		 * only way to set the correct vram_width
-		 */
-		adev->gmc.vram_width = 2048;
-	else if (amdgpu_emu_mode != 1)
-		adev->gmc.vram_width = vram_width;
+	if (amdgpu_is_multi_aid(adev)) {
+		gmc_v9_4_3_init_vram_info(adev);
+	} else if (!adev->bios) {
+		if (adev->flags & AMD_IS_APU) {
+			adev->gmc.vram_type = AMDGPU_VRAM_TYPE_DDR4;
+			adev->gmc.vram_width = 64 * 64;
+		} else {
+			adev->gmc.vram_type = AMDGPU_VRAM_TYPE_HBM;
+			adev->gmc.vram_width = 128 * 64;
+		}
+	} else {
+		r = amdgpu_atomfirmware_get_vram_info(adev,
+			&vram_width, &vram_type, &vram_vendor);
+		if (amdgpu_sriov_vf(adev))
+			/* For Vega10 SR-IOV, vram_width can't be read from ATOM as RAVEN,
+			 * and DF related registers is not readable, seems hardcord is the
+			 * only way to set the correct vram_width
+			 */
+			adev->gmc.vram_width = 2048;
+		else if (amdgpu_emu_mode != 1)
+			adev->gmc.vram_width = vram_width;
 
-	if (!adev->gmc.vram_width) {
-		int chansize, numchan;
+		if (!adev->gmc.vram_width) {
+			int chansize, numchan;
 
-		/* hbm memory channel size */
-		if (adev->flags & AMD_IS_APU)
-			chansize = 64;
-		else
-			chansize = 128;
-		if (adev->df.funcs &&
-		    adev->df.funcs->get_hbm_channel_number) {
-			numchan = adev->df.funcs->get_hbm_channel_number(adev);
-			adev->gmc.vram_width = numchan * chansize;
+			/* hbm memory channel size */
+			if (adev->flags & AMD_IS_APU)
+				chansize = 64;
+			else
+				chansize = 128;
+			if (adev->df.funcs &&
+			    adev->df.funcs->get_hbm_channel_number) {
+				numchan = adev->df.funcs->get_hbm_channel_number(adev);
+				adev->gmc.vram_width = numchan * chansize;
+			}
 		}
-	}
 
-	adev->gmc.vram_type = vram_type;
-	adev->gmc.vram_vendor = vram_vendor;
-	switch (adev->ip_versions[GC_HWIP][0]) {
+		adev->gmc.vram_type = vram_type;
+		adev->gmc.vram_vendor = vram_vendor;
+	}
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 1, 0):
 	case IP_VERSION(9, 2, 2):
-		adev->num_vmhubs = 2;
+		set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
+		set_bit(AMDGPU_MMHUB0(0), adev->vmhubs_mask);
 
 		if (adev->rev_id == 0x0 || adev->rev_id == 0x1) {
 			amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48);
@@ -1702,30 +1927,40 @@ static int gmc_v9_0_sw_init(void *handle)
 	case IP_VERSION(9, 4, 0):
 	case IP_VERSION(9, 3, 0):
 	case IP_VERSION(9, 4, 2):
-	case IP_VERSION(9, 4, 3):
-		adev->num_vmhubs = 2;
-
+		set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
+		set_bit(AMDGPU_MMHUB0(0), adev->vmhubs_mask);
 
 		/*
 		 * To fulfill 4-level page support,
 		 * vm size is 256TB (48bit), maximum size of Vega10,
 		 * block size 512 (9bit)
 		 */
-		/* sriov restrict max_pfn below AMDGPU_GMC_HOLE */
-		if (amdgpu_sriov_vf(adev))
-			amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 47);
-		else
-			amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48);
-		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2))
+
+		amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48);
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2))
 			adev->gmc.translate_further = adev->vm_manager.num_level > 1;
 		break;
 	case IP_VERSION(9, 4, 1):
-		adev->num_vmhubs = 3;
+		set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
+		set_bit(AMDGPU_MMHUB0(0), adev->vmhubs_mask);
+		set_bit(AMDGPU_MMHUB1(0), adev->vmhubs_mask);
 
 		/* Keep the vm size same with Vega20 */
 		amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48);
 		adev->gmc.translate_further = adev->vm_manager.num_level > 1;
 		break;
+	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+	case IP_VERSION(9, 5, 0):
+		bitmap_set(adev->vmhubs_mask, AMDGPU_GFXHUB(0),
+				  NUM_XCC(adev->gfx.xcc_mask));
+
+		inst_mask <<= AMDGPU_MMHUB0(0);
+		bitmap_or(adev->vmhubs_mask, adev->vmhubs_mask, &inst_mask, 32);
+
+		amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48);
+		adev->gmc.translate_further = adev->vm_manager.num_level > 1;
+		break;
 	default:
 		break;
 	}
@@ -1736,7 +1971,7 @@ static int gmc_v9_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1)) {
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1)) {
 		r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VMC1, VMC_1_0__SRCID__VM_FAULT,
 					&adev->gmc.vm_fault);
 		if (r)
@@ -1750,7 +1985,8 @@ static int gmc_v9_0_sw_init(void *handle)
 		return r;
 
 	if (!amdgpu_sriov_vf(adev) &&
-	    !adev->gmc.xgmi.connected_to_cpu) {
+	    !adev->gmc.xgmi.connected_to_cpu &&
+	    !adev->gmc.is_app_apu) {
 		/* interrupt sent to DF. */
 		r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_DF, 0,
 				      &adev->gmc.ecc_irq);
@@ -1764,10 +2000,13 @@ static int gmc_v9_0_sw_init(void *handle)
 	 */
 	adev->gmc.mc_mask = 0xffffffffffffULL; /* 48 bit MC */
 
-	dma_addr_bits = adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2) ? 48:44;
+	dma_addr_bits = amdgpu_ip_version(adev, GC_HWIP, 0) >=
+					IP_VERSION(9, 4, 2) ?
+				48 :
+				44;
 	r = dma_set_mask_and_coherent(adev->dev, DMA_BIT_MASK(dma_addr_bits));
 	if (r) {
-		printk(KERN_WARNING "amdgpu: No suitable DMA available.\n");
+		dev_warn(adev->dev, "amdgpu: No suitable DMA available.\n");
 		return r;
 	}
 	adev->need_swiotlb = drm_need_swiotlb(dma_addr_bits);
@@ -1778,6 +2017,12 @@ static int gmc_v9_0_sw_init(void *handle)
 
 	amdgpu_gmc_get_vbios_allocations(adev);
 
+	if (amdgpu_is_multi_aid(adev)) {
+		r = amdgpu_gmc_init_mem_ranges(adev);
+		if (r)
+			return r;
+	}
+
 	/* Memory manager */
 	r = amdgpu_bo_init(adev);
 	if (r)
@@ -1787,6 +2032,7 @@ static int gmc_v9_0_sw_init(void *handle)
 	if (r)
 		return r;
 
+	gmc_v9_0_init_nps_details(adev);
 	/*
 	 * number of VMs
 	 * VMID 0 is reserved for System
@@ -1798,9 +2044,11 @@ static int gmc_v9_0_sw_init(void *handle)
 	 * for video processing.
 	 */
 	adev->vm_manager.first_kfd_vmid =
-		(adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1) ||
-		 adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2) ||
-		 adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3)) ? 3 : 8;
+		(amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1) ||
+		 amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2) ||
+		 amdgpu_is_multi_aid(adev)) ?
+			3 :
+			8;
 
 	amdgpu_vm_manager_init(adev);
 
@@ -1810,27 +2058,40 @@ static int gmc_v9_0_sw_init(void *handle)
 	if (r)
 		return r;
 
+	if (amdgpu_is_multi_aid(adev))
+		amdgpu_gmc_sysfs_init(adev);
+
 	return 0;
 }
 
-static int gmc_v9_0_sw_fini(void *handle)
+static int gmc_v9_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (amdgpu_is_multi_aid(adev))
+		amdgpu_gmc_sysfs_fini(adev);
 
 	amdgpu_gmc_ras_fini(adev);
 	amdgpu_gem_force_release(adev);
 	amdgpu_vm_manager_fini(adev);
-	amdgpu_gart_table_vram_free(adev);
+	if (!adev->gmc.real_vram_size) {
+		dev_info(adev->dev, "Put GART in system memory for APU free\n");
+		amdgpu_gart_table_ram_free(adev);
+	} else {
+		amdgpu_gart_table_vram_free(adev);
+	}
 	amdgpu_bo_free_kernel(&adev->gmc.pdb0_bo, NULL, &adev->gmc.ptr_pdb0);
 	amdgpu_bo_fini(adev);
 
+	adev->gmc.num_mem_partitions = 0;
+	kfree(adev->gmc.mem_partitions);
+
 	return 0;
 }
 
 static void gmc_v9_0_init_golden_registers(struct amdgpu_device *adev)
 {
-
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(9, 0, 0):
 		if (amdgpu_sriov_vf(adev))
 			break;
@@ -1864,8 +2125,8 @@ static void gmc_v9_0_init_golden_registers(struct amdgpu_device *adev)
  */
 void gmc_v9_0_restore_registers(struct amdgpu_device *adev)
 {
-	if ((adev->ip_versions[DCE_HWIP][0] == IP_VERSION(1, 0, 0)) ||
-	    (adev->ip_versions[DCE_HWIP][0] == IP_VERSION(1, 0, 1))) {
+	if ((amdgpu_ip_version(adev, DCE_HWIP, 0) == IP_VERSION(1, 0, 0)) ||
+	    (amdgpu_ip_version(adev, DCE_HWIP, 0) == IP_VERSION(1, 0, 1))) {
 		WREG32_SOC15(DCE, 0, mmDCHUBBUB_SDPIF_MMIO_CNTRL_0, adev->gmc.sdpif_register);
 		WARN_ON(adev->gmc.sdpif_register !=
 			RREG32_SOC15(DCE, 0, mmDCHUBBUB_SDPIF_MMIO_CNTRL_0));
@@ -1881,7 +2142,7 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev)
 {
 	int r;
 
-	if (adev->gmc.xgmi.connected_to_cpu)
+	if (amdgpu_gmc_is_pdb0_enabled(adev))
 		amdgpu_gmc_init_pdb0(adev);
 
 	if (adev->gart.bo == NULL) {
@@ -1902,7 +2163,7 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev)
 		return r;
 
 	DRM_INFO("PCIE GART of %uM enabled.\n",
-		 (unsigned)(adev->gmc.gart_size >> 20));
+		 (unsigned int)(adev->gmc.gart_size >> 20));
 	if (adev->gmc.pdb0_bo)
 		DRM_INFO("PDB0 located at 0x%016llX\n",
 				(unsigned long long)amdgpu_bo_gpu_offset(adev->gmc.pdb0_bo));
@@ -1912,12 +2173,23 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev)
 	return 0;
 }
 
-static int gmc_v9_0_hw_init(void *handle)
+static int gmc_v9_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool value;
 	int i, r;
 
+	adev->gmc.flush_pasid_uses_kiq = true;
+
+	/* Vega20+XGMI caches PTEs in TC and TLB. Add a heavy-weight TLB flush
+	 * (type 2), which flushes both. Due to a race condition with
+	 * concurrent memory accesses using the same TLB cache line, we still
+	 * need a second TLB flush after this.
+	 */
+	adev->gmc.flush_tlb_needs_extra_type_2 =
+		amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 0) &&
+		adev->gmc.xgmi.num_physical_nodes;
+
 	/* The sequence of these two function calls matters.*/
 	gmc_v9_0_init_golden_registers(adev);
 
@@ -1934,7 +2206,7 @@ static int gmc_v9_0_hw_init(void *handle)
 	adev->hdp.funcs->init_registers(adev);
 
 	/* After HDP is initialized, flush HDP.*/
-	adev->hdp.funcs->flush_hdp(adev, NULL);
+	amdgpu_device_flush_hdp(adev, NULL);
 
 	if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS)
 		value = false;
@@ -1946,8 +2218,8 @@ static int gmc_v9_0_hw_init(void *handle)
 			adev->gfxhub.funcs->set_fault_enable_default(adev, value);
 		adev->mmhub.funcs->set_fault_enable_default(adev, value);
 	}
-	for (i = 0; i < adev->num_vmhubs; ++i) {
-		if (adev->in_s0ix && (i == AMDGPU_GFXHUB_0))
+	for_each_set_bit(i, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS) {
+		if (adev->in_s0ix && (i == AMDGPU_GFXHUB(0)))
 			continue;
 		gmc_v9_0_flush_gpu_tlb(adev, 0, i, 0);
 	}
@@ -1961,8 +2233,8 @@ static int gmc_v9_0_hw_init(void *handle)
 
 	if (amdgpu_emu_mode == 1)
 		return amdgpu_gmc_vram_checking(adev);
-	else
-		return r;
+
+	return 0;
 }
 
 /**
@@ -1979,9 +2251,9 @@ static void gmc_v9_0_gart_disable(struct amdgpu_device *adev)
 	adev->mmhub.funcs->gart_disable(adev);
 }
 
-static int gmc_v9_0_hw_fini(void *handle)
+static int gmc_v9_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v9_0_gart_disable(adev);
 
@@ -1999,55 +2271,70 @@ static int gmc_v9_0_hw_fini(void *handle)
 	if (adev->mmhub.funcs->update_power_gating)
 		adev->mmhub.funcs->update_power_gating(adev, false);
 
-	amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0);
-	amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
+	/*
+	 * For minimal init, late_init is not called, hence VM fault/RAS irqs
+	 * are not enabled.
+	 */
+	if (adev->init_lvl->level != AMDGPU_INIT_LEVEL_MINIMAL_XGMI) {
+		amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
+
+		if (adev->gmc.ecc_irq.funcs &&
+		    amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC))
+			amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0);
+	}
 
 	return 0;
 }
 
-static int gmc_v9_0_suspend(void *handle)
+static int gmc_v9_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return gmc_v9_0_hw_fini(adev);
+	return gmc_v9_0_hw_fini(ip_block);
 }
 
-static int gmc_v9_0_resume(void *handle)
+static int gmc_v9_0_resume(struct amdgpu_ip_block *ip_block)
 {
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = gmc_v9_0_hw_init(adev);
+	/* If a reset is done for NPS mode switch, read the memory range
+	 * information again.
+	 */
+	if (adev->gmc.reset_flags & AMDGPU_GMC_INIT_RESET_NPS) {
+		amdgpu_gmc_init_sw_mem_ranges(adev, adev->gmc.mem_partitions);
+		adev->gmc.reset_flags &= ~AMDGPU_GMC_INIT_RESET_NPS;
+	}
+
+	r = gmc_v9_0_hw_init(ip_block);
 	if (r)
 		return r;
 
-	amdgpu_vmid_reset_all(adev);
+	amdgpu_vmid_reset_all(ip_block->adev);
 
 	return 0;
 }
 
-static bool gmc_v9_0_is_idle(void *handle)
+static bool gmc_v9_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* MC is always ready in GMC v9.*/
 	return true;
 }
 
-static int gmc_v9_0_wait_for_idle(void *handle)
+static int gmc_v9_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* There is no need to wait for MC idle in GMC v9.*/
 	return 0;
 }
 
-static int gmc_v9_0_soft_reset(void *handle)
+static int gmc_v9_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	/* XXX for emulation.*/
 	return 0;
 }
 
-static int gmc_v9_0_set_clockgating_state(void *handle,
+static int gmc_v9_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->mmhub.funcs->set_clockgating(adev, state);
 
@@ -2056,16 +2343,16 @@ static int gmc_v9_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static void gmc_v9_0_get_clockgating_state(void *handle, u64 *flags)
+static void gmc_v9_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->mmhub.funcs->get_clockgating(adev, flags);
 
 	athub_v1_0_get_clockgating(adev, flags);
 }
 
-static int gmc_v9_0_set_powergating_state(void *handle,
+static int gmc_v9_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					enum amd_powergating_state state)
 {
 	return 0;
@@ -2089,8 +2376,7 @@ const struct amd_ip_funcs gmc_v9_0_ip_funcs = {
 	.get_clockgating_state = gmc_v9_0_get_clockgating_state,
 };
 
-const struct amdgpu_ip_block_version gmc_v9_0_ip_block =
-{
+const struct amdgpu_ip_block_version gmc_v9_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GMC,
 	.major = 9,
 	.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c
index 71d1a2e3bac9..e6c0d86d3486 100644
--- a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c
@@ -21,7 +21,6 @@
  *
  */
 #include "amdgpu.h"
-#include "amdgpu_atombios.h"
 #include "hdp_v4_0.h"
 #include "amdgpu_ras.h"
 
@@ -37,27 +36,21 @@
 #define HDP_MEM_POWER_CTRL__RC_MEM_POWER_LS_EN_MASK     0x00020000L
 #define mmHDP_MEM_POWER_CTRL_BASE_IDX   0
 
-static void hdp_v4_0_flush_hdp(struct amdgpu_device *adev,
-				struct amdgpu_ring *ring)
-{
-	if (!ring || !ring->funcs->emit_wreg)
-		WREG32_NO_KIQ((adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2, 0);
-	else
-		amdgpu_ring_emit_wreg(ring, (adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2, 0);
-}
-
 static void hdp_v4_0_invalidate_hdp(struct amdgpu_device *adev,
 				    struct amdgpu_ring *ring)
 {
-	if (adev->ip_versions[HDP_HWIP][0] == IP_VERSION(4, 4, 0) ||
-	    adev->ip_versions[HDP_HWIP][0] == IP_VERSION(4, 4, 2))
+	if (amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 4, 0) ||
+	    amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 4, 2) ||
+	    amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 4, 5))
 		return;
 
-	if (!ring || !ring->funcs->emit_wreg)
+	if (!ring || !ring->funcs->emit_wreg) {
 		WREG32_SOC15_NO_KIQ(HDP, 0, mmHDP_READ_CACHE_INVALIDATE, 1);
-	else
+		RREG32_SOC15_NO_KIQ(HDP, 0, mmHDP_READ_CACHE_INVALIDATE);
+	} else {
 		amdgpu_ring_emit_wreg(ring, SOC15_REG_OFFSET(
 			HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 1);
+	}
 }
 
 static void hdp_v4_0_query_ras_error_count(struct amdgpu_device *adev,
@@ -80,7 +73,7 @@ static void hdp_v4_0_reset_ras_error_count(struct amdgpu_device *adev)
 	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__HDP))
 		return;
 
-	if (adev->ip_versions[HDP_HWIP][0] >= IP_VERSION(4, 4, 0))
+	if (amdgpu_ip_version(adev, HDP_HWIP, 0) >= IP_VERSION(4, 4, 0))
 		WREG32_SOC15(HDP, 0, mmHDP_EDC_CNT, 0);
 	else
 		/*read back hdp ras counter to reset it to 0 */
@@ -92,10 +85,10 @@ static void hdp_v4_0_update_clock_gating(struct amdgpu_device *adev,
 {
 	uint32_t def, data;
 
-	if (adev->ip_versions[HDP_HWIP][0] == IP_VERSION(4, 0, 0) ||
-	    adev->ip_versions[HDP_HWIP][0] == IP_VERSION(4, 0, 1) ||
-	    adev->ip_versions[HDP_HWIP][0] == IP_VERSION(4, 1, 1) ||
-	    adev->ip_versions[HDP_HWIP][0] == IP_VERSION(4, 1, 0)) {
+	if (amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 0, 0) ||
+	    amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 0, 1) ||
+	    amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 1, 1) ||
+	    amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 1, 0)) {
 		def = data = RREG32(SOC15_REG_OFFSET(HDP, 0, mmHDP_MEM_POWER_LS));
 
 		if (enable && (adev->cg_flags & AMD_CG_SUPPORT_HDP_LS))
@@ -129,6 +122,12 @@ static void hdp_v4_0_get_clockgating_state(struct amdgpu_device *adev,
 {
 	int data;
 
+	if (amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 4, 2) ||
+	    amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 4, 5)) {
+		/* Default enabled */
+		*flags |= AMD_CG_SUPPORT_HDP_MGCG;
+		return;
+	}
 	/* AMD_CG_SUPPORT_HDP_LS */
 	data = RREG32(SOC15_REG_OFFSET(HDP, 0, mmHDP_MEM_POWER_LS));
 	if (data & HDP_MEM_POWER_LS__LS_ENABLE_MASK)
@@ -137,7 +136,7 @@ static void hdp_v4_0_get_clockgating_state(struct amdgpu_device *adev,
 
 static void hdp_v4_0_init_registers(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[HDP_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, HDP_HWIP, 0)) {
 	case IP_VERSION(4, 2, 1):
 		WREG32_FIELD15(HDP, 0, HDP_MMHUB_CNTL, HDP_MMHUB_GCC, 1);
 		break;
@@ -145,9 +144,13 @@ static void hdp_v4_0_init_registers(struct amdgpu_device *adev)
 		break;
 	}
 
+	/* Do not program registers if VF */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
 	WREG32_FIELD15(HDP, 0, HDP_MISC_CNTL, FLUSH_INVALIDATE_CACHE, 1);
 
-	if (adev->ip_versions[HDP_HWIP][0] == IP_VERSION(4, 4, 0))
+	if (amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 4, 0))
 		WREG32_FIELD15(HDP, 0, HDP_MISC_CNTL, READ_BUFFER_WATERMARK, 2);
 
 	WREG32_SOC15(HDP, 0, mmHDP_NONSURFACE_BASE, (adev->gmc.vram_start >> 8));
@@ -166,7 +169,7 @@ struct amdgpu_hdp_ras hdp_v4_0_ras = {
 };
 
 const struct amdgpu_hdp_funcs hdp_v4_0_funcs = {
-	.flush_hdp = hdp_v4_0_flush_hdp,
+	.flush_hdp = amdgpu_hdp_generic_flush,
 	.invalidate_hdp = hdp_v4_0_invalidate_hdp,
 	.update_clock_gating = hdp_v4_0_update_clock_gating,
 	.get_clock_gating_state = hdp_v4_0_get_clockgating_state,
diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v5_0.c b/drivers/gpu/drm/amd/amdgpu/hdp_v5_0.c
index a9ea23fa0def..8bc001dc9f63 100644
--- a/drivers/gpu/drm/amd/amdgpu/hdp_v5_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/hdp_v5_0.c
@@ -21,27 +21,18 @@
  *
  */
 #include "amdgpu.h"
-#include "amdgpu_atombios.h"
 #include "hdp_v5_0.h"
 
 #include "hdp/hdp_5_0_0_offset.h"
 #include "hdp/hdp_5_0_0_sh_mask.h"
 #include <uapi/linux/kfd_ioctl.h>
 
-static void hdp_v5_0_flush_hdp(struct amdgpu_device *adev,
-				struct amdgpu_ring *ring)
-{
-	if (!ring || !ring->funcs->emit_wreg)
-		WREG32_NO_KIQ((adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2, 0);
-	else
-		amdgpu_ring_emit_wreg(ring, (adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2, 0);
-}
-
 static void hdp_v5_0_invalidate_hdp(struct amdgpu_device *adev,
 				    struct amdgpu_ring *ring)
 {
 	if (!ring || !ring->funcs->emit_wreg) {
 		WREG32_SOC15_NO_KIQ(HDP, 0, mmHDP_READ_CACHE_INVALIDATE, 1);
+		RREG32_SOC15_NO_KIQ(HDP, 0, mmHDP_READ_CACHE_INVALIDATE);
 	} else {
 		amdgpu_ring_emit_wreg(ring, SOC15_REG_OFFSET(
 					HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 1);
@@ -215,7 +206,7 @@ static void hdp_v5_0_init_registers(struct amdgpu_device *adev)
 }
 
 const struct amdgpu_hdp_funcs hdp_v5_0_funcs = {
-	.flush_hdp = hdp_v5_0_flush_hdp,
+	.flush_hdp = amdgpu_hdp_generic_flush,
 	.invalidate_hdp = hdp_v5_0_invalidate_hdp,
 	.update_clock_gating = hdp_v5_0_update_clock_gating,
 	.get_clock_gating_state = hdp_v5_0_get_clockgating_state,
diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v5_2.c b/drivers/gpu/drm/amd/amdgpu/hdp_v5_2.c
index 29c3484ae1f1..40940b4ab400 100644
--- a/drivers/gpu/drm/amd/amdgpu/hdp_v5_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/hdp_v5_2.c
@@ -21,7 +21,6 @@
  *
  */
 #include "amdgpu.h"
-#include "amdgpu_atombios.h"
 #include "hdp_v5_2.h"
 
 #include "hdp/hdp_5_2_1_offset.h"
@@ -31,13 +30,25 @@
 static void hdp_v5_2_flush_hdp(struct amdgpu_device *adev,
 				struct amdgpu_ring *ring)
 {
-	if (!ring || !ring->funcs->emit_wreg)
+	if (!ring || !ring->funcs->emit_wreg) {
 		WREG32_NO_KIQ((adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2,
 			0);
-	else
+		if (amdgpu_sriov_vf(adev)) {
+			/* this is fine because SR_IOV doesn't remap the register */
+			RREG32_NO_KIQ((adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2);
+		} else {
+			/* We just need to read back a register to post the write.
+			 * Reading back the remapped register causes problems on
+			 * some platforms so just read back the memory size register.
+			 */
+			if (adev->nbio.funcs->get_memsize)
+				adev->nbio.funcs->get_memsize(adev);
+		}
+	} else {
 		amdgpu_ring_emit_wreg(ring,
 			(adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2,
 			0);
+	}
 }
 
 static void hdp_v5_2_update_mem_power_gating(struct amdgpu_device *adev,
diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v6_0.c b/drivers/gpu/drm/amd/amdgpu/hdp_v6_0.c
index 063eba619f2f..ec20daf4272c 100644
--- a/drivers/gpu/drm/amd/amdgpu/hdp_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/hdp_v6_0.c
@@ -21,26 +21,19 @@
  *
  */
 #include "amdgpu.h"
-#include "amdgpu_atombios.h"
 #include "hdp_v6_0.h"
 
 #include "hdp/hdp_6_0_0_offset.h"
 #include "hdp/hdp_6_0_0_sh_mask.h"
 #include <uapi/linux/kfd_ioctl.h>
 
-static void hdp_v6_0_flush_hdp(struct amdgpu_device *adev,
-				struct amdgpu_ring *ring)
-{
-	if (!ring || !ring->funcs->emit_wreg)
-		WREG32_NO_KIQ((adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2, 0);
-	else
-		amdgpu_ring_emit_wreg(ring, (adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2, 0);
-}
+#define regHDP_CLK_CNTL_V6_1	0xd5
+#define regHDP_CLK_CNTL_V6_1_BASE_IDX 0
 
 static void hdp_v6_0_update_clock_gating(struct amdgpu_device *adev,
 					 bool enable)
 {
-	uint32_t hdp_clk_cntl, hdp_clk_cntl1;
+	uint32_t hdp_clk_cntl;
 	uint32_t hdp_mem_pwr_cntl;
 
 	if (!(adev->cg_flags & (AMD_CG_SUPPORT_HDP_LS |
@@ -48,14 +41,20 @@ static void hdp_v6_0_update_clock_gating(struct amdgpu_device *adev,
 				AMD_CG_SUPPORT_HDP_SD)))
 		return;
 
-	hdp_clk_cntl = hdp_clk_cntl1 = RREG32_SOC15(HDP, 0,regHDP_CLK_CNTL);
+	if (amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(6, 1, 0))
+		hdp_clk_cntl = RREG32_SOC15(HDP, 0, regHDP_CLK_CNTL_V6_1);
+	else
+		hdp_clk_cntl = RREG32_SOC15(HDP, 0, regHDP_CLK_CNTL);
 	hdp_mem_pwr_cntl = RREG32_SOC15(HDP, 0, regHDP_MEM_POWER_CTRL);
 
 	/* Before doing clock/power mode switch,
 	 * forced on IPH & RC clock */
 	hdp_clk_cntl = REG_SET_FIELD(hdp_clk_cntl, HDP_CLK_CNTL,
 				     RC_MEM_CLK_SOFT_OVERRIDE, 1);
-	WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL, hdp_clk_cntl);
+	if (amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(6, 1, 0))
+		WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL_V6_1, hdp_clk_cntl);
+	else
+		WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL, hdp_clk_cntl);
 
 	/* disable clock and power gating before any changing */
 	hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
@@ -117,7 +116,10 @@ static void hdp_v6_0_update_clock_gating(struct amdgpu_device *adev,
 	/* disable IPH & RC clock override after clock/power mode changing */
 	hdp_clk_cntl = REG_SET_FIELD(hdp_clk_cntl, HDP_CLK_CNTL,
 				     RC_MEM_CLK_SOFT_OVERRIDE, 0);
-	WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL, hdp_clk_cntl);
+	if (amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(6, 1, 0))
+		WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL_V6_1, hdp_clk_cntl);
+	else
+		WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL, hdp_clk_cntl);
 }
 
 static void hdp_v6_0_get_clockgating_state(struct amdgpu_device *adev,
@@ -136,7 +138,7 @@ static void hdp_v6_0_get_clockgating_state(struct amdgpu_device *adev,
 }
 
 const struct amdgpu_hdp_funcs hdp_v6_0_funcs = {
-	.flush_hdp = hdp_v6_0_flush_hdp,
+	.flush_hdp = amdgpu_hdp_generic_flush,
 	.update_clock_gating = hdp_v6_0_update_clock_gating,
 	.get_clock_gating_state = hdp_v6_0_get_clockgating_state,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v7_0.c b/drivers/gpu/drm/amd/amdgpu/hdp_v7_0.c
new file mode 100644
index 000000000000..ed1debc03507
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/hdp_v7_0.c
@@ -0,0 +1,132 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "amdgpu.h"
+#include "hdp_v7_0.h"
+
+#include "hdp/hdp_7_0_0_offset.h"
+#include "hdp/hdp_7_0_0_sh_mask.h"
+#include <uapi/linux/kfd_ioctl.h>
+
+static void hdp_v7_0_update_clock_gating(struct amdgpu_device *adev,
+					 bool enable)
+{
+	uint32_t hdp_clk_cntl, hdp_clk_cntl1;
+	uint32_t hdp_mem_pwr_cntl;
+
+	if (!(adev->cg_flags & (AMD_CG_SUPPORT_HDP_LS |
+				AMD_CG_SUPPORT_HDP_DS |
+				AMD_CG_SUPPORT_HDP_SD)))
+		return;
+
+	hdp_clk_cntl = hdp_clk_cntl1 = RREG32_SOC15(HDP, 0,regHDP_CLK_CNTL);
+	hdp_mem_pwr_cntl = RREG32_SOC15(HDP, 0, regHDP_MEM_POWER_CTRL);
+
+	/* Before doing clock/power mode switch,
+	 * forced on IPH & RC clock */
+	hdp_clk_cntl = REG_SET_FIELD(hdp_clk_cntl, HDP_CLK_CNTL,
+				     RC_MEM_CLK_SOFT_OVERRIDE, 1);
+	WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL, hdp_clk_cntl);
+
+	/* disable clock and power gating before any changing */
+	hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+					 ATOMIC_MEM_POWER_CTRL_EN, 0);
+	hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+					 ATOMIC_MEM_POWER_LS_EN, 0);
+	hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+					 ATOMIC_MEM_POWER_DS_EN, 0);
+	hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+					 ATOMIC_MEM_POWER_SD_EN, 0);
+	hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+					 RC_MEM_POWER_CTRL_EN, 0);
+	hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+					 RC_MEM_POWER_LS_EN, 0);
+	hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+					 RC_MEM_POWER_DS_EN, 0);
+	hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+					 RC_MEM_POWER_SD_EN, 0);
+	WREG32_SOC15(HDP, 0, regHDP_MEM_POWER_CTRL, hdp_mem_pwr_cntl);
+
+	/* Already disabled above. The actions below are for "enabled" only */
+	if (enable) {
+		/* only one clock gating mode (LS/DS/SD) can be enabled */
+		if (adev->cg_flags & AMD_CG_SUPPORT_HDP_SD) {
+			hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl,
+							 HDP_MEM_POWER_CTRL,
+							 ATOMIC_MEM_POWER_SD_EN, 1);
+			hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl,
+							 HDP_MEM_POWER_CTRL,
+							 RC_MEM_POWER_SD_EN, 1);
+		} else if (adev->cg_flags & AMD_CG_SUPPORT_HDP_LS) {
+			hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl,
+							 HDP_MEM_POWER_CTRL,
+							 ATOMIC_MEM_POWER_LS_EN, 1);
+			hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl,
+							 HDP_MEM_POWER_CTRL,
+							 RC_MEM_POWER_LS_EN, 1);
+		} else if (adev->cg_flags & AMD_CG_SUPPORT_HDP_DS) {
+			hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl,
+							 HDP_MEM_POWER_CTRL,
+							 ATOMIC_MEM_POWER_DS_EN, 1);
+			hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl,
+							 HDP_MEM_POWER_CTRL,
+							 RC_MEM_POWER_DS_EN, 1);
+		}
+
+		/* confirmed that IPH_MEM_POWER_CTRL_EN and RC_MEM_POWER_CTRL_EN have to
+		 * be set for SRAM LS/DS/SD */
+		if (adev->cg_flags & (AMD_CG_SUPPORT_HDP_LS | AMD_CG_SUPPORT_HDP_DS |
+				      AMD_CG_SUPPORT_HDP_SD)) {
+			hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+							 ATOMIC_MEM_POWER_CTRL_EN, 1);
+			hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+							 RC_MEM_POWER_CTRL_EN, 1);
+			WREG32_SOC15(HDP, 0, regHDP_MEM_POWER_CTRL, hdp_mem_pwr_cntl);
+		}
+	}
+
+	/* disable IPH & RC clock override after clock/power mode changing */
+	hdp_clk_cntl = REG_SET_FIELD(hdp_clk_cntl, HDP_CLK_CNTL,
+				     RC_MEM_CLK_SOFT_OVERRIDE, 0);
+	WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL, hdp_clk_cntl);
+}
+
+static void hdp_v7_0_get_clockgating_state(struct amdgpu_device *adev,
+					    u64 *flags)
+{
+	uint32_t tmp;
+
+	/* AMD_CG_SUPPORT_HDP_LS/DS/SD */
+	tmp = RREG32_SOC15(HDP, 0, regHDP_MEM_POWER_CTRL);
+	if (tmp & HDP_MEM_POWER_CTRL__ATOMIC_MEM_POWER_LS_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_HDP_LS;
+	else if (tmp & HDP_MEM_POWER_CTRL__ATOMIC_MEM_POWER_DS_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_HDP_DS;
+	else if (tmp & HDP_MEM_POWER_CTRL__ATOMIC_MEM_POWER_SD_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_HDP_SD;
+}
+
+const struct amdgpu_hdp_funcs hdp_v7_0_funcs = {
+	.flush_hdp = amdgpu_hdp_generic_flush,
+	.update_clock_gating = hdp_v7_0_update_clock_gating,
+	.get_clock_gating_state = hdp_v7_0_get_clockgating_state,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v7_0.h b/drivers/gpu/drm/amd/amdgpu/hdp_v7_0.h
new file mode 100644
index 000000000000..25b69201402d
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/hdp_v7_0.h
@@ -0,0 +1,31 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __HDP_V7_0_H__
+#define __HDP_V7_0_H__
+
+#include "soc15_common.h"
+
+extern const struct amdgpu_hdp_funcs hdp_v7_0_funcs;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c
index aecad530b10a..01cadf898c00 100644
--- a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c
@@ -157,6 +157,9 @@ static int iceland_ih_irq_init(struct amdgpu_device *adev)
 	/* enable interrupts */
 	iceland_ih_enable_interrupts(adev);
 
+	if (adev->irq.ih_soft.ring_size)
+		adev->irq.ih_soft.enabled = true;
+
 	return 0;
 }
 
@@ -194,6 +197,9 @@ static u32 iceland_ih_get_wptr(struct amdgpu_device *adev,
 
 	wptr = le32_to_cpu(*ih->wptr_cpu);
 
+	if (ih == &adev->irq.ih_soft)
+		goto out;
+
 	if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
 		goto out;
 
@@ -215,6 +221,11 @@ static u32 iceland_ih_get_wptr(struct amdgpu_device *adev,
 	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
 	WREG32(mmIH_RB_CNTL, tmp);
 
+	/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+	 * can be detected.
+	 */
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+	WREG32(mmIH_RB_CNTL, tmp);
 
 out:
 	return (wptr & ih->ptr_mask);
@@ -268,9 +279,9 @@ static void iceland_ih_set_rptr(struct amdgpu_device *adev,
 	WREG32(mmIH_RB_RPTR, ih->rptr);
 }
 
-static int iceland_ih_early_init(void *handle)
+static int iceland_ih_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	ret = amdgpu_irq_add_domain(adev);
@@ -282,23 +293,27 @@ static int iceland_ih_early_init(void *handle)
 	return 0;
 }
 
-static int iceland_ih_sw_init(void *handle)
+static int iceland_ih_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 64 * 1024, false);
 	if (r)
 		return r;
 
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, IH_SW_RING_SIZE, true);
+	if (r)
+		return r;
+
 	r = amdgpu_irq_init(adev);
 
 	return r;
 }
 
-static int iceland_ih_sw_fini(void *handle)
+static int iceland_ih_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_fini_sw(adev);
 	amdgpu_irq_remove_domain(adev);
@@ -306,39 +321,33 @@ static int iceland_ih_sw_fini(void *handle)
 	return 0;
 }
 
-static int iceland_ih_hw_init(void *handle)
+static int iceland_ih_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return iceland_ih_irq_init(adev);
 }
 
-static int iceland_ih_hw_fini(void *handle)
+static int iceland_ih_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	iceland_ih_irq_disable(adev);
+	iceland_ih_irq_disable(ip_block->adev);
 
 	return 0;
 }
 
-static int iceland_ih_suspend(void *handle)
+static int iceland_ih_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return iceland_ih_hw_fini(adev);
+	return iceland_ih_hw_fini(ip_block);
 }
 
-static int iceland_ih_resume(void *handle)
+static int iceland_ih_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return iceland_ih_hw_init(adev);
+	return iceland_ih_hw_init(ip_block);
 }
 
-static bool iceland_ih_is_idle(void *handle)
+static bool iceland_ih_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (REG_GET_FIELD(tmp, SRBM_STATUS, IH_BUSY))
@@ -347,11 +356,11 @@ static bool iceland_ih_is_idle(void *handle)
 	return true;
 }
 
-static int iceland_ih_wait_for_idle(void *handle)
+static int iceland_ih_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		/* read MC_STATUS */
@@ -363,10 +372,10 @@ static int iceland_ih_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int iceland_ih_soft_reset(void *handle)
+static int iceland_ih_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 srbm_soft_reset = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (tmp & SRBM_STATUS__IH_BUSY_MASK)
@@ -393,13 +402,13 @@ static int iceland_ih_soft_reset(void *handle)
 	return 0;
 }
 
-static int iceland_ih_set_clockgating_state(void *handle,
+static int iceland_ih_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int iceland_ih_set_powergating_state(void *handle,
+static int iceland_ih_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -408,7 +417,6 @@ static int iceland_ih_set_powergating_state(void *handle,
 static const struct amd_ip_funcs iceland_ih_ip_funcs = {
 	.name = "iceland_ih",
 	.early_init = iceland_ih_early_init,
-	.late_init = NULL,
 	.sw_init = iceland_ih_sw_init,
 	.sw_fini = iceland_ih_sw_fini,
 	.hw_init = iceland_ih_hw_init,
diff --git a/drivers/gpu/drm/amd/amdgpu/ih_v6_0.c b/drivers/gpu/drm/amd/amdgpu/ih_v6_0.c
index b02e1cef78a7..333e9c30c091 100644
--- a/drivers/gpu/drm/amd/amdgpu/ih_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/ih_v6_0.c
@@ -135,6 +135,34 @@ static int ih_v6_0_toggle_ring_interrupts(struct amdgpu_device *adev,
 
 	tmp = RREG32(ih_regs->ih_rb_cntl);
 	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_ENABLE, (enable ? 1 : 0));
+
+	if (enable) {
+		/* Unset the CLEAR_OVERFLOW bit to make sure the next step
+		 * is switching the bit from 0 to 1
+		 */
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+		if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) {
+			if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp))
+				return -ETIMEDOUT;
+		} else {
+			WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+		}
+
+		/* Clear RB_OVERFLOW bit */
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
+		if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) {
+			if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp))
+				return -ETIMEDOUT;
+		} else {
+			WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+		}
+
+		/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+		 * can be detected.
+		 */
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+	}
+
 	/* enable_intr field is only valid in ring0 */
 	if (ih == &adev->irq.ih)
 		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, ENABLE_INTR, (enable ? 1 : 0));
@@ -321,6 +349,7 @@ static int ih_v6_0_irq_init(struct amdgpu_device *adev)
 			if (ret)
 				return ret;
 		}
+		ih[i]->overflow = false;
 	}
 
 	/* update doorbell range for ih ring 0 */
@@ -346,6 +375,21 @@ static int ih_v6_0_irq_init(struct amdgpu_device *adev)
 			    DELAY, 3);
 	WREG32_SOC15(OSSSYS, 0, regIH_MSI_STORM_CTRL, tmp);
 
+	/* Redirect the interrupts to IH RB1 for dGPU */
+	if (adev->irq.ih1.ring_size) {
+		tmp = RREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_INDEX);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_INDEX, INDEX, 0);
+		WREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_INDEX, tmp);
+
+		tmp = RREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_DATA);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA, CLIENT_ID, 0xa);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA, SOURCE_ID, 0x0);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA,
+				    SOURCE_ID_MATCH_ENABLE, 0x1);
+
+		WREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_DATA, tmp);
+	}
+
 	pci_set_master(adev->pdev);
 
 	/* enable interrupts */
@@ -403,7 +447,10 @@ static u32 ih_v6_0_get_wptr(struct amdgpu_device *adev,
 	wptr = RREG32_NO_KIQ(ih_regs->ih_rb_wptr);
 	if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
 		goto out;
-	wptr = REG_SET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW, 0);
+	if (!amdgpu_sriov_vf(adev))
+		wptr = REG_SET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW, 0);
+	else
+		ih->overflow = true;
 
 	/* When a ring buffer overflow happen start parsing interrupt
 	 * from the last not overwritten vector (wptr + 32). Hopefully
@@ -418,6 +465,12 @@ static u32 ih_v6_0_get_wptr(struct amdgpu_device *adev,
 	tmp = RREG32_NO_KIQ(ih_regs->ih_rb_cntl);
 	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
 	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+
+	/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+	 * can be detected.
+	 */
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
 out:
 	return (wptr & ih->ptr_mask);
 }
@@ -494,7 +547,8 @@ static int ih_v6_0_self_irq(struct amdgpu_device *adev,
 		*adev->irq.ih1.wptr_cpu = wptr;
 		schedule_work(&adev->irq.ih1_work);
 		break;
-	default: break;
+	default:
+		break;
 	}
 	return 0;
 }
@@ -509,19 +563,19 @@ static void ih_v6_0_set_self_irq_funcs(struct amdgpu_device *adev)
 	adev->irq.self_irq.funcs = &ih_v6_0_self_irq_funcs;
 }
 
-static int ih_v6_0_early_init(void *handle)
+static int ih_v6_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	ih_v6_0_set_interrupt_funcs(adev);
 	ih_v6_0_set_self_irq_funcs(adev);
 	return 0;
 }
 
-static int ih_v6_0_sw_init(void *handle)
+static int ih_v6_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool use_bus_addr;
 
 	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_IH, 0,
@@ -533,22 +587,28 @@ static int ih_v6_0_sw_init(void *handle)
 	/* use gpu virtual address for ih ring
 	 * until ih_checken is programmed to allow
 	 * use bus address for ih ring by psp bl */
-	use_bus_addr =
-		(adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) ? false : true;
-	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 256 * 1024, use_bus_addr);
+	use_bus_addr = adev->firmware.load_type != AMDGPU_FW_LOAD_PSP;
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, IH_RING_SIZE, use_bus_addr);
 	if (r)
 		return r;
 
 	adev->irq.ih.use_doorbell = true;
 	adev->irq.ih.doorbell_index = adev->doorbell_index.ih << 1;
 
-	adev->irq.ih1.ring_size = 0;
-	adev->irq.ih2.ring_size = 0;
+	if (!(adev->flags & AMD_IS_APU)) {
+		r = amdgpu_ih_ring_init(adev, &adev->irq.ih1, IH_RING_SIZE,
+					use_bus_addr);
+		if (r)
+			return r;
+
+		adev->irq.ih1.use_doorbell = true;
+		adev->irq.ih1.doorbell_index = (adev->doorbell_index.ih + 1) << 1;
+	}
 
 	/* initialize ih control register offset */
 	ih_v6_0_init_register_offset(adev);
 
-	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, PAGE_SIZE, true);
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, IH_SW_RING_SIZE, true);
 	if (r)
 		return r;
 
@@ -557,19 +617,19 @@ static int ih_v6_0_sw_init(void *handle)
 	return r;
 }
 
-static int ih_v6_0_sw_fini(void *handle)
+static int ih_v6_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_fini_sw(adev);
 
 	return 0;
 }
 
-static int ih_v6_0_hw_init(void *handle)
+static int ih_v6_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = ih_v6_0_irq_init(adev);
 	if (r)
@@ -578,42 +638,36 @@ static int ih_v6_0_hw_init(void *handle)
 	return 0;
 }
 
-static int ih_v6_0_hw_fini(void *handle)
+static int ih_v6_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	ih_v6_0_irq_disable(adev);
+	ih_v6_0_irq_disable(ip_block->adev);
 
 	return 0;
 }
 
-static int ih_v6_0_suspend(void *handle)
+static int ih_v6_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return ih_v6_0_hw_fini(adev);
+	return ih_v6_0_hw_fini(ip_block);
 }
 
-static int ih_v6_0_resume(void *handle)
+static int ih_v6_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return ih_v6_0_hw_init(adev);
+	return ih_v6_0_hw_init(ip_block);
 }
 
-static bool ih_v6_0_is_idle(void *handle)
+static bool ih_v6_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* todo */
 	return true;
 }
 
-static int ih_v6_0_wait_for_idle(void *handle)
+static int ih_v6_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* todo */
 	return -ETIMEDOUT;
 }
 
-static int ih_v6_0_soft_reset(void *handle)
+static int ih_v6_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	/* todo */
 	return 0;
@@ -640,14 +694,12 @@ static void ih_v6_0_update_clockgating_state(struct amdgpu_device *adev,
 		if (def != data)
 			WREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL, data);
 	}
-
-	return;
 }
 
-static int ih_v6_0_set_clockgating_state(void *handle,
+static int ih_v6_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	ih_v6_0_update_clockgating_state(adev,
 				state == AMD_CG_STATE_GATE);
@@ -707,10 +759,10 @@ static void ih_v6_0_update_ih_mem_power_gating(struct amdgpu_device *adev,
 	WREG32_SOC15(OSSSYS, 0, regIH_MEM_POWER_CTRL, ih_mem_pwr_cntl);
 }
 
-static int ih_v6_0_set_powergating_state(void *handle,
+static int ih_v6_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					 enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_PG_STATE_GATE);
 
 	if (adev->pg_flags & AMD_PG_SUPPORT_IH_SRAM_PG)
@@ -719,20 +771,17 @@ static int ih_v6_0_set_powergating_state(void *handle,
 	return 0;
 }
 
-static void ih_v6_0_get_clockgating_state(void *handle, u64 *flags)
+static void ih_v6_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (!RREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL))
 		*flags |= AMD_CG_SUPPORT_IH_CG;
-
-	return;
 }
 
 static const struct amd_ip_funcs ih_v6_0_ip_funcs = {
 	.name = "ih_v6_0",
 	.early_init = ih_v6_0_early_init,
-	.late_init = NULL,
 	.sw_init = ih_v6_0_sw_init,
 	.sw_fini = ih_v6_0_sw_fini,
 	.hw_init = ih_v6_0_hw_init,
@@ -759,8 +808,7 @@ static void ih_v6_0_set_interrupt_funcs(struct amdgpu_device *adev)
 	adev->irq.ih_funcs = &ih_v6_0_funcs;
 }
 
-const struct amdgpu_ip_block_version ih_v6_0_ip_block =
-{
+const struct amdgpu_ip_block_version ih_v6_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_IH,
 	.major = 6,
 	.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/ih_v6_1.c b/drivers/gpu/drm/amd/amdgpu/ih_v6_1.c
new file mode 100644
index 000000000000..95b3f4e55ec3
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/ih_v6_1.c
@@ -0,0 +1,796 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/pci.h>
+
+#include "amdgpu.h"
+#include "amdgpu_ih.h"
+
+#include "oss/osssys_6_1_0_offset.h"
+#include "oss/osssys_6_1_0_sh_mask.h"
+
+#include "soc15_common.h"
+#include "ih_v6_1.h"
+
+#define MAX_REARM_RETRY 10
+
+static void ih_v6_1_set_interrupt_funcs(struct amdgpu_device *adev);
+
+/**
+ * ih_v6_1_init_register_offset - Initialize register offset for ih rings
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Initialize register offset ih rings (IH_V6_0).
+ */
+static void ih_v6_1_init_register_offset(struct amdgpu_device *adev)
+{
+	struct amdgpu_ih_regs *ih_regs;
+
+	/* ih ring 2 is removed
+	 * ih ring and ih ring 1 are available */
+	if (adev->irq.ih.ring_size) {
+		ih_regs = &adev->irq.ih.ih_regs;
+		ih_regs->ih_rb_base = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE);
+		ih_regs->ih_rb_base_hi = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE_HI);
+		ih_regs->ih_rb_cntl = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_CNTL);
+		ih_regs->ih_rb_wptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR);
+		ih_regs->ih_rb_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_RPTR);
+		ih_regs->ih_doorbell_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_DOORBELL_RPTR);
+		ih_regs->ih_rb_wptr_addr_lo = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR_ADDR_LO);
+		ih_regs->ih_rb_wptr_addr_hi = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR_ADDR_HI);
+		ih_regs->psp_reg_id = PSP_REG_IH_RB_CNTL;
+	}
+
+	if (adev->irq.ih1.ring_size) {
+		ih_regs = &adev->irq.ih1.ih_regs;
+		ih_regs->ih_rb_base = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE_RING1);
+		ih_regs->ih_rb_base_hi = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE_HI_RING1);
+		ih_regs->ih_rb_cntl = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_CNTL_RING1);
+		ih_regs->ih_rb_wptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR_RING1);
+		ih_regs->ih_rb_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_RPTR_RING1);
+		ih_regs->ih_doorbell_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_DOORBELL_RPTR_RING1);
+		ih_regs->psp_reg_id = PSP_REG_IH_RB_CNTL_RING1;
+	}
+}
+
+/**
+ * force_update_wptr_for_self_int - Force update the wptr for self interrupt
+ *
+ * @adev: amdgpu_device pointer
+ * @threshold: threshold to trigger the wptr reporting
+ * @timeout: timeout to trigger the wptr reporting
+ * @enabled: Enable/disable timeout flush mechanism
+ *
+ * threshold input range: 0 ~ 15, default 0,
+ * real_threshold = 2^threshold
+ * timeout input range: 0 ~ 20, default 8,
+ * real_timeout = (2^timeout) * 1024 / (socclk_freq)
+ *
+ * Force update wptr for self interrupt ( >= SIENNA_CICHLID).
+ */
+static void
+force_update_wptr_for_self_int(struct amdgpu_device *adev,
+			       u32 threshold, u32 timeout, bool enabled)
+{
+	u32 ih_cntl, ih_rb_cntl;
+
+	ih_cntl = RREG32_SOC15(OSSSYS, 0, regIH_CNTL2);
+	ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, regIH_RB_CNTL_RING1);
+
+	ih_cntl = REG_SET_FIELD(ih_cntl, IH_CNTL2,
+				SELF_IV_FORCE_WPTR_UPDATE_TIMEOUT, timeout);
+	ih_cntl = REG_SET_FIELD(ih_cntl, IH_CNTL2,
+				SELF_IV_FORCE_WPTR_UPDATE_ENABLE, enabled);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL_RING1,
+				   RB_USED_INT_THRESHOLD, threshold);
+
+	if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) {
+		if (psp_reg_program(&adev->psp, PSP_REG_IH_RB_CNTL_RING1, ih_rb_cntl))
+			return;
+	} else {
+		WREG32_SOC15(OSSSYS, 0, regIH_RB_CNTL_RING1, ih_rb_cntl);
+	}
+
+	WREG32_SOC15(OSSSYS, 0, regIH_CNTL2, ih_cntl);
+}
+
+/**
+ * ih_v6_1_toggle_ring_interrupts - toggle the interrupt ring buffer
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: amdgpu_ih_ring pointer
+ * @enable: true - enable the interrupts, false - disable the interrupts
+ *
+ * Toggle the interrupt ring buffer (IH_V6_0)
+ */
+static int ih_v6_1_toggle_ring_interrupts(struct amdgpu_device *adev,
+					  struct amdgpu_ih_ring *ih,
+					  bool enable)
+{
+	struct amdgpu_ih_regs *ih_regs;
+	uint32_t tmp;
+
+	ih_regs = &ih->ih_regs;
+
+	tmp = RREG32(ih_regs->ih_rb_cntl);
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_ENABLE, (enable ? 1 : 0));
+	/* enable_intr field is only valid in ring0 */
+	if (ih == &adev->irq.ih)
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, ENABLE_INTR, (enable ? 1 : 0));
+
+	if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) {
+		if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp))
+			return -ETIMEDOUT;
+	} else {
+		WREG32(ih_regs->ih_rb_cntl, tmp);
+	}
+
+	if (enable) {
+		ih->enabled = true;
+	} else {
+		/* set rptr, wptr to 0 */
+		WREG32(ih_regs->ih_rb_rptr, 0);
+		WREG32(ih_regs->ih_rb_wptr, 0);
+		ih->enabled = false;
+		ih->rptr = 0;
+	}
+
+	return 0;
+}
+
+/**
+ * ih_v6_1_toggle_interrupts - Toggle all the available interrupt ring buffers
+ *
+ * @adev: amdgpu_device pointer
+ * @enable: enable or disable interrupt ring buffers
+ *
+ * Toggle all the available interrupt ring buffers (IH_V6_0).
+ */
+static int ih_v6_1_toggle_interrupts(struct amdgpu_device *adev, bool enable)
+{
+	struct amdgpu_ih_ring *ih[] = {&adev->irq.ih, &adev->irq.ih1};
+	int i;
+	int r;
+
+	for (i = 0; i < ARRAY_SIZE(ih); i++) {
+		if (ih[i]->ring_size) {
+			r = ih_v6_1_toggle_ring_interrupts(adev, ih[i], enable);
+			if (r)
+				return r;
+		}
+	}
+
+	return 0;
+}
+
+static uint32_t ih_v6_1_rb_cntl(struct amdgpu_ih_ring *ih, uint32_t ih_rb_cntl)
+{
+	int rb_bufsz = order_base_2(ih->ring_size / 4);
+
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+				   MC_SPACE, ih->use_bus_addr ? 2 : 4);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+				   WPTR_OVERFLOW_CLEAR, 1);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+				   WPTR_OVERFLOW_ENABLE, 1);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, RB_SIZE, rb_bufsz);
+	/* Ring Buffer write pointer writeback. If enabled, IH_RB_WPTR register
+	 * value is written to memory
+	 */
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+				   WPTR_WRITEBACK_ENABLE, 1);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_SNOOP, 1);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_RO, 0);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_VMID, 0);
+
+	return ih_rb_cntl;
+}
+
+static uint32_t ih_v6_1_doorbell_rptr(struct amdgpu_ih_ring *ih)
+{
+	u32 ih_doorbell_rtpr = 0;
+
+	if (ih->use_doorbell) {
+		ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
+						 IH_DOORBELL_RPTR, OFFSET,
+						 ih->doorbell_index);
+		ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
+						 IH_DOORBELL_RPTR,
+						 ENABLE, 1);
+	} else {
+		ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
+						 IH_DOORBELL_RPTR,
+						 ENABLE, 0);
+	}
+	return ih_doorbell_rtpr;
+}
+
+/**
+ * ih_v6_1_enable_ring - enable an ih ring buffer
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: amdgpu_ih_ring pointer
+ *
+ * Enable an ih ring buffer (IH_V6_0)
+ */
+static int ih_v6_1_enable_ring(struct amdgpu_device *adev,
+				      struct amdgpu_ih_ring *ih)
+{
+	struct amdgpu_ih_regs *ih_regs;
+	uint32_t tmp;
+
+	ih_regs = &ih->ih_regs;
+
+	/* Ring Buffer base. [39:8] of 40-bit address of the beginning of the ring buffer*/
+	WREG32(ih_regs->ih_rb_base, ih->gpu_addr >> 8);
+	WREG32(ih_regs->ih_rb_base_hi, (ih->gpu_addr >> 40) & 0xff);
+
+	tmp = RREG32(ih_regs->ih_rb_cntl);
+	tmp = ih_v6_1_rb_cntl(ih, tmp);
+	if (ih == &adev->irq.ih)
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RPTR_REARM, !!adev->irq.msi_enabled);
+	if (ih == &adev->irq.ih1) {
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_ENABLE, 0);
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_FULL_DRAIN_ENABLE, 1);
+	}
+
+	if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) {
+		if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp)) {
+			DRM_ERROR("PSP program IH_RB_CNTL failed!\n");
+			return -ETIMEDOUT;
+		}
+	} else {
+		WREG32(ih_regs->ih_rb_cntl, tmp);
+	}
+
+	if (ih == &adev->irq.ih) {
+		/* set the ih ring 0 writeback address whether it's enabled or not */
+		WREG32(ih_regs->ih_rb_wptr_addr_lo, lower_32_bits(ih->wptr_addr));
+		WREG32(ih_regs->ih_rb_wptr_addr_hi, upper_32_bits(ih->wptr_addr) & 0xFFFF);
+	}
+
+	/* set rptr, wptr to 0 */
+	WREG32(ih_regs->ih_rb_wptr, 0);
+	WREG32(ih_regs->ih_rb_rptr, 0);
+
+	WREG32(ih_regs->ih_doorbell_rptr, ih_v6_1_doorbell_rptr(ih));
+
+	return 0;
+}
+
+/**
+ * ih_v6_1_irq_init - init and enable the interrupt ring
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Allocate a ring buffer for the interrupt controller,
+ * enable the RLC, disable interrupts, enable the IH
+ * ring buffer and enable it.
+ * Called at device load and reume.
+ * Returns 0 for success, errors for failure.
+ */
+static int ih_v6_1_irq_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_ih_ring *ih[] = {&adev->irq.ih, &adev->irq.ih1};
+	u32 ih_chicken;
+	u32 tmp;
+	int ret;
+	int i;
+
+	/* disable irqs */
+	ret = ih_v6_1_toggle_interrupts(adev, false);
+	if (ret)
+		return ret;
+
+	adev->nbio.funcs->ih_control(adev);
+
+	if (unlikely((adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) ||
+		     (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO))) {
+		if (ih[0]->use_bus_addr) {
+			ih_chicken = RREG32_SOC15(OSSSYS, 0, regIH_CHICKEN);
+			ih_chicken = REG_SET_FIELD(ih_chicken,
+					IH_CHICKEN, MC_SPACE_GPA_ENABLE, 1);
+			WREG32_SOC15(OSSSYS, 0, regIH_CHICKEN, ih_chicken);
+		}
+	}
+
+	for (i = 0; i < ARRAY_SIZE(ih); i++) {
+		if (ih[i]->ring_size) {
+			ret = ih_v6_1_enable_ring(adev, ih[i]);
+			if (ret)
+				return ret;
+		}
+	}
+
+	/* update doorbell range for ih ring 0 */
+	adev->nbio.funcs->ih_doorbell_range(adev, ih[0]->use_doorbell,
+					    ih[0]->doorbell_index);
+
+	tmp = RREG32_SOC15(OSSSYS, 0, regIH_STORM_CLIENT_LIST_CNTL);
+	tmp = REG_SET_FIELD(tmp, IH_STORM_CLIENT_LIST_CNTL,
+			    CLIENT18_IS_STORM_CLIENT, 1);
+	WREG32_SOC15(OSSSYS, 0, regIH_STORM_CLIENT_LIST_CNTL, tmp);
+
+	tmp = RREG32_SOC15(OSSSYS, 0, regIH_INT_FLOOD_CNTL);
+	tmp = REG_SET_FIELD(tmp, IH_INT_FLOOD_CNTL, FLOOD_CNTL_ENABLE, 1);
+	WREG32_SOC15(OSSSYS, 0, regIH_INT_FLOOD_CNTL, tmp);
+
+	/* GC/MMHUB UTCL2 page fault interrupts are configured as
+	 * MSI storm capable interrupts by deafult. The delay is
+	 * used to avoid ISR being called too frequently
+	 * when page fault happens on several continuous page
+	 * and thus avoid MSI storm */
+	tmp = RREG32_SOC15(OSSSYS, 0, regIH_MSI_STORM_CTRL);
+	tmp = REG_SET_FIELD(tmp, IH_MSI_STORM_CTRL,
+			    DELAY, 3);
+	WREG32_SOC15(OSSSYS, 0, regIH_MSI_STORM_CTRL, tmp);
+
+	/* Redirect the interrupts to IH RB1 for dGPU */
+	if (adev->irq.ih1.ring_size) {
+		tmp = RREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_INDEX);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_INDEX, INDEX, 0);
+		WREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_INDEX, tmp);
+
+		tmp = RREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_DATA);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA, CLIENT_ID, 0xa);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA, SOURCE_ID, 0x0);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA,
+				    SOURCE_ID_MATCH_ENABLE, 0x1);
+
+		WREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_DATA, tmp);
+	}
+
+	pci_set_master(adev->pdev);
+
+	/* enable interrupts */
+	ret = ih_v6_1_toggle_interrupts(adev, true);
+	if (ret)
+		return ret;
+	/* enable wptr force update for self int */
+	force_update_wptr_for_self_int(adev, 0, 8, true);
+
+	if (adev->irq.ih_soft.ring_size)
+		adev->irq.ih_soft.enabled = true;
+
+	return 0;
+}
+
+/**
+ * ih_v6_1_irq_disable - disable interrupts
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Disable interrupts on the hw.
+ */
+static void ih_v6_1_irq_disable(struct amdgpu_device *adev)
+{
+	force_update_wptr_for_self_int(adev, 0, 8, false);
+	ih_v6_1_toggle_interrupts(adev, false);
+
+	/* Wait and acknowledge irq */
+	mdelay(1);
+}
+
+/**
+ * ih_v6_1_get_wptr - get the IH ring buffer wptr
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: amdgpu_ih_ring pointer
+ *
+ * Get the IH ring buffer wptr from either the register
+ * or the writeback memory buffer.  Also check for
+ * ring buffer overflow and deal with it.
+ * Returns the value of the wptr.
+ */
+static u32 ih_v6_1_get_wptr(struct amdgpu_device *adev,
+			      struct amdgpu_ih_ring *ih)
+{
+	u32 wptr, tmp;
+	struct amdgpu_ih_regs *ih_regs;
+
+	wptr = le32_to_cpu(*ih->wptr_cpu);
+	ih_regs = &ih->ih_regs;
+
+	if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
+		goto out;
+
+	wptr = RREG32_NO_KIQ(ih_regs->ih_rb_wptr);
+	if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
+		goto out;
+	wptr = REG_SET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW, 0);
+
+	/* When a ring buffer overflow happen start parsing interrupt
+	 * from the last not overwritten vector (wptr + 32). Hopefully
+	 * this should allow us to catch up.
+	 */
+	tmp = (wptr + 32) & ih->ptr_mask;
+	dev_warn(adev->dev, "IH ring buffer overflow "
+		 "(0x%08X, 0x%08X, 0x%08X)\n",
+		 wptr, ih->rptr, tmp);
+	ih->rptr = tmp;
+
+	tmp = RREG32_NO_KIQ(ih_regs->ih_rb_cntl);
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
+	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+
+	/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+	 * can be detected.
+	 */
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+
+out:
+	return (wptr & ih->ptr_mask);
+}
+
+/**
+ * ih_v6_1_irq_rearm - rearm IRQ if lost
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: amdgpu_ih_ring pointer
+ *
+ */
+static void ih_v6_1_irq_rearm(struct amdgpu_device *adev,
+			       struct amdgpu_ih_ring *ih)
+{
+	uint32_t v = 0;
+	uint32_t i = 0;
+	struct amdgpu_ih_regs *ih_regs;
+
+	ih_regs = &ih->ih_regs;
+
+	/* Rearm IRQ / re-write doorbell if doorbell write is lost */
+	for (i = 0; i < MAX_REARM_RETRY; i++) {
+		v = RREG32_NO_KIQ(ih_regs->ih_rb_rptr);
+		if ((v < ih->ring_size) && (v != ih->rptr))
+			WDOORBELL32(ih->doorbell_index, ih->rptr);
+		else
+			break;
+	}
+}
+
+/**
+ * ih_v6_1_set_rptr - set the IH ring buffer rptr
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: amdgpu_ih_ring pointer
+ *
+ * Set the IH ring buffer rptr.
+ */
+static void ih_v6_1_set_rptr(struct amdgpu_device *adev,
+			       struct amdgpu_ih_ring *ih)
+{
+	struct amdgpu_ih_regs *ih_regs;
+
+	if (ih->use_doorbell) {
+		/* XXX check if swapping is necessary on BE */
+		*ih->rptr_cpu = ih->rptr;
+		WDOORBELL32(ih->doorbell_index, ih->rptr);
+
+		if (amdgpu_sriov_vf(adev))
+			ih_v6_1_irq_rearm(adev, ih);
+	} else {
+		ih_regs = &ih->ih_regs;
+		WREG32(ih_regs->ih_rb_rptr, ih->rptr);
+	}
+}
+
+/**
+ * ih_v6_1_self_irq - dispatch work for ring 1
+ *
+ * @adev: amdgpu_device pointer
+ * @source: irq source
+ * @entry: IV with WPTR update
+ *
+ * Update the WPTR from the IV and schedule work to handle the entries.
+ */
+static int ih_v6_1_self_irq(struct amdgpu_device *adev,
+			      struct amdgpu_irq_src *source,
+			      struct amdgpu_iv_entry *entry)
+{
+	uint32_t wptr = cpu_to_le32(entry->src_data[0]);
+
+	switch (entry->ring_id) {
+	case 1:
+		*adev->irq.ih1.wptr_cpu = wptr;
+		schedule_work(&adev->irq.ih1_work);
+		break;
+	default:
+		break;
+	}
+	return 0;
+}
+
+static const struct amdgpu_irq_src_funcs ih_v6_1_self_irq_funcs = {
+	.process = ih_v6_1_self_irq,
+};
+
+static void ih_v6_1_set_self_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->irq.self_irq.num_types = 0;
+	adev->irq.self_irq.funcs = &ih_v6_1_self_irq_funcs;
+}
+
+static int ih_v6_1_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int ret;
+
+	ret = amdgpu_irq_add_domain(adev);
+	if (ret) {
+		return ret;
+	}
+
+	ih_v6_1_set_interrupt_funcs(adev);
+	ih_v6_1_set_self_irq_funcs(adev);
+	return 0;
+}
+
+static int ih_v6_1_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+	bool use_bus_addr;
+
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_IH, 0,
+			      &adev->irq.self_irq);
+
+	if (r)
+		return r;
+
+	/* use gpu virtual address for ih ring
+	 * until ih_checken is programmed to allow
+	 * use bus address for ih ring by psp bl */
+	use_bus_addr = adev->firmware.load_type != AMDGPU_FW_LOAD_PSP;
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 256 * 1024, use_bus_addr);
+	if (r)
+		return r;
+
+	adev->irq.ih.use_doorbell = true;
+	adev->irq.ih.doorbell_index = adev->doorbell_index.ih << 1;
+
+	if (!(adev->flags & AMD_IS_APU)) {
+		r = amdgpu_ih_ring_init(adev, &adev->irq.ih1, IH_RING_SIZE,
+					use_bus_addr);
+		if (r)
+			return r;
+
+		adev->irq.ih1.use_doorbell = true;
+		adev->irq.ih1.doorbell_index = (adev->doorbell_index.ih + 1) << 1;
+	}
+
+	/* initialize ih control register offset */
+	ih_v6_1_init_register_offset(adev);
+
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, PAGE_SIZE, true);
+	if (r)
+		return r;
+
+	r = amdgpu_irq_init(adev);
+
+	return r;
+}
+
+static int ih_v6_1_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	amdgpu_irq_fini_sw(adev);
+
+	return 0;
+}
+
+static int ih_v6_1_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	r = ih_v6_1_irq_init(adev);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static int ih_v6_1_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	ih_v6_1_irq_disable(ip_block->adev);
+
+	return 0;
+}
+
+static int ih_v6_1_suspend(struct amdgpu_ip_block *ip_block)
+{
+	return ih_v6_1_hw_fini(ip_block);
+}
+
+static int ih_v6_1_resume(struct amdgpu_ip_block *ip_block)
+{
+	return ih_v6_1_hw_init(ip_block);
+}
+
+static bool ih_v6_1_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	/* todo */
+	return true;
+}
+
+static int ih_v6_1_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	/* todo */
+	return -ETIMEDOUT;
+}
+
+static int ih_v6_1_soft_reset(struct amdgpu_ip_block *ip_block)
+{
+	/* todo */
+	return 0;
+}
+
+static void ih_v6_1_update_clockgating_state(struct amdgpu_device *adev,
+					       bool enable)
+{
+	uint32_t data, def, field_val;
+
+	if (adev->cg_flags & AMD_CG_SUPPORT_IH_CG) {
+		def = data = RREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL);
+		field_val = enable ? 0 : 1;
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     DBUS_MUX_CLK_SOFT_OVERRIDE, field_val);
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     OSSSYS_SHARE_CLK_SOFT_OVERRIDE, field_val);
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     LIMIT_SMN_CLK_SOFT_OVERRIDE, field_val);
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     DYN_CLK_SOFT_OVERRIDE, field_val);
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     REG_CLK_SOFT_OVERRIDE, field_val);
+		if (def != data)
+			WREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL, data);
+	}
+
+	return;
+}
+
+static int ih_v6_1_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					   enum amd_clockgating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	ih_v6_1_update_clockgating_state(adev,
+				state == AMD_CG_STATE_GATE);
+	return 0;
+}
+
+static void ih_v6_1_update_ih_mem_power_gating(struct amdgpu_device *adev,
+					       bool enable)
+{
+	uint32_t ih_mem_pwr_cntl;
+
+	/* Disable ih sram power cntl before switch powergating mode */
+	ih_mem_pwr_cntl = RREG32_SOC15(OSSSYS, 0, regIH_MEM_POWER_CTRL);
+	ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+					IH_BUFFER_MEM_POWER_CTRL_EN, 0);
+	WREG32_SOC15(OSSSYS, 0, regIH_MEM_POWER_CTRL, ih_mem_pwr_cntl);
+
+	/* It is recommended to set mem powergating mode to DS mode */
+	if (enable) {
+		/* mem power mode */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_LS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_DS_EN, 1);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_SD_EN, 0);
+		/* cam mem power mode */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_LS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_DS_EN, 1);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_SD_EN, 0);
+		/* re-enable power cntl */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_CTRL_EN, 1);
+	} else {
+		/* mem power mode */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_LS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_DS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_SD_EN, 0);
+		/* cam mem power mode */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_LS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_DS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_SD_EN, 0);
+		/* re-enable power cntl*/
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_CTRL_EN, 1);
+	}
+
+	WREG32_SOC15(OSSSYS, 0, regIH_MEM_POWER_CTRL, ih_mem_pwr_cntl);
+}
+
+static int ih_v6_1_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					 enum amd_powergating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	bool enable = (state == AMD_PG_STATE_GATE);
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_IH_SRAM_PG)
+		ih_v6_1_update_ih_mem_power_gating(adev, enable);
+
+	return 0;
+}
+
+static void ih_v6_1_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (!RREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL))
+		*flags |= AMD_CG_SUPPORT_IH_CG;
+
+	return;
+}
+
+static const struct amd_ip_funcs ih_v6_1_ip_funcs = {
+	.name = "ih_v6_1",
+	.early_init = ih_v6_1_early_init,
+	.sw_init = ih_v6_1_sw_init,
+	.sw_fini = ih_v6_1_sw_fini,
+	.hw_init = ih_v6_1_hw_init,
+	.hw_fini = ih_v6_1_hw_fini,
+	.suspend = ih_v6_1_suspend,
+	.resume = ih_v6_1_resume,
+	.is_idle = ih_v6_1_is_idle,
+	.wait_for_idle = ih_v6_1_wait_for_idle,
+	.soft_reset = ih_v6_1_soft_reset,
+	.set_clockgating_state = ih_v6_1_set_clockgating_state,
+	.set_powergating_state = ih_v6_1_set_powergating_state,
+	.get_clockgating_state = ih_v6_1_get_clockgating_state,
+};
+
+static const struct amdgpu_ih_funcs ih_v6_1_funcs = {
+	.get_wptr = ih_v6_1_get_wptr,
+	.decode_iv = amdgpu_ih_decode_iv_helper,
+	.decode_iv_ts = amdgpu_ih_decode_iv_ts_helper,
+	.set_rptr = ih_v6_1_set_rptr
+};
+
+static void ih_v6_1_set_interrupt_funcs(struct amdgpu_device *adev)
+{
+	adev->irq.ih_funcs = &ih_v6_1_funcs;
+}
+
+const struct amdgpu_ip_block_version ih_v6_1_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_IH,
+	.major = 6,
+	.minor = 0,
+	.rev = 0,
+	.funcs = &ih_v6_1_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/ih_v6_1.h b/drivers/gpu/drm/amd/amdgpu/ih_v6_1.h
new file mode 100644
index 000000000000..2232bc5cbd09
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/ih_v6_1.h
@@ -0,0 +1,28 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __IH_V6_1_IH_H__
+#define __IH_V6_1_IH_H__
+
+extern const struct amdgpu_ip_block_version ih_v6_1_ip_block;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/ih_v7_0.c b/drivers/gpu/drm/amd/amdgpu/ih_v7_0.c
new file mode 100644
index 000000000000..b32ea4129c61
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/ih_v7_0.c
@@ -0,0 +1,787 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/pci.h>
+
+#include "amdgpu.h"
+#include "amdgpu_ih.h"
+
+#include "oss/osssys_7_0_0_offset.h"
+#include "oss/osssys_7_0_0_sh_mask.h"
+
+#include "soc15_common.h"
+#include "ih_v7_0.h"
+
+#define MAX_REARM_RETRY 10
+
+static void ih_v7_0_set_interrupt_funcs(struct amdgpu_device *adev);
+
+/**
+ * ih_v7_0_init_register_offset - Initialize register offset for ih rings
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Initialize register offset ih rings (IH_V7_0).
+ */
+static void ih_v7_0_init_register_offset(struct amdgpu_device *adev)
+{
+	struct amdgpu_ih_regs *ih_regs;
+
+	/* ih ring 2 is removed
+	 * ih ring and ih ring 1 are available */
+	if (adev->irq.ih.ring_size) {
+		ih_regs = &adev->irq.ih.ih_regs;
+		ih_regs->ih_rb_base = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE);
+		ih_regs->ih_rb_base_hi = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE_HI);
+		ih_regs->ih_rb_cntl = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_CNTL);
+		ih_regs->ih_rb_wptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR);
+		ih_regs->ih_rb_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_RPTR);
+		ih_regs->ih_doorbell_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_DOORBELL_RPTR);
+		ih_regs->ih_rb_wptr_addr_lo = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR_ADDR_LO);
+		ih_regs->ih_rb_wptr_addr_hi = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR_ADDR_HI);
+		ih_regs->psp_reg_id = PSP_REG_IH_RB_CNTL;
+	}
+
+	if (adev->irq.ih1.ring_size) {
+		ih_regs = &adev->irq.ih1.ih_regs;
+		ih_regs->ih_rb_base = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE_RING1);
+		ih_regs->ih_rb_base_hi = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE_HI_RING1);
+		ih_regs->ih_rb_cntl = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_CNTL_RING1);
+		ih_regs->ih_rb_wptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR_RING1);
+		ih_regs->ih_rb_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_RPTR_RING1);
+		ih_regs->ih_doorbell_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_DOORBELL_RPTR_RING1);
+		ih_regs->psp_reg_id = PSP_REG_IH_RB_CNTL_RING1;
+	}
+}
+
+/**
+ * force_update_wptr_for_self_int - Force update the wptr for self interrupt
+ *
+ * @adev: amdgpu_device pointer
+ * @threshold: threshold to trigger the wptr reporting
+ * @timeout: timeout to trigger the wptr reporting
+ * @enabled: Enable/disable timeout flush mechanism
+ *
+ * threshold input range: 0 ~ 15, default 0,
+ * real_threshold = 2^threshold
+ * timeout input range: 0 ~ 20, default 8,
+ * real_timeout = (2^timeout) * 1024 / (socclk_freq)
+ *
+ * Force update wptr for self interrupt ( >= SIENNA_CICHLID).
+ */
+static void
+force_update_wptr_for_self_int(struct amdgpu_device *adev,
+			       u32 threshold, u32 timeout, bool enabled)
+{
+	u32 ih_cntl, ih_rb_cntl;
+
+	ih_cntl = RREG32_SOC15(OSSSYS, 0, regIH_CNTL2);
+	ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, regIH_RB_CNTL_RING1);
+
+	ih_cntl = REG_SET_FIELD(ih_cntl, IH_CNTL2,
+				SELF_IV_FORCE_WPTR_UPDATE_TIMEOUT, timeout);
+	ih_cntl = REG_SET_FIELD(ih_cntl, IH_CNTL2,
+				SELF_IV_FORCE_WPTR_UPDATE_ENABLE, enabled);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL_RING1,
+				   RB_USED_INT_THRESHOLD, threshold);
+
+	if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) {
+		if (psp_reg_program(&adev->psp, PSP_REG_IH_RB_CNTL_RING1, ih_rb_cntl))
+			return;
+	} else {
+		WREG32_SOC15(OSSSYS, 0, regIH_RB_CNTL_RING1, ih_rb_cntl);
+	}
+
+	WREG32_SOC15(OSSSYS, 0, regIH_CNTL2, ih_cntl);
+}
+
+/**
+ * ih_v7_0_toggle_ring_interrupts - toggle the interrupt ring buffer
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: amdgpu_ih_ring pointet
+ * @enable: true - enable the interrupts, false - disable the interrupts
+ *
+ * Toggle the interrupt ring buffer (IH_V7_0)
+ */
+static int ih_v7_0_toggle_ring_interrupts(struct amdgpu_device *adev,
+					  struct amdgpu_ih_ring *ih,
+					  bool enable)
+{
+	struct amdgpu_ih_regs *ih_regs;
+	uint32_t tmp;
+
+	ih_regs = &ih->ih_regs;
+
+	tmp = RREG32(ih_regs->ih_rb_cntl);
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_ENABLE, (enable ? 1 : 0));
+	/* enable_intr field is only valid in ring0 */
+	if (ih == &adev->irq.ih)
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, ENABLE_INTR, (enable ? 1 : 0));
+
+	if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) {
+		if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp))
+			return -ETIMEDOUT;
+	} else {
+		WREG32(ih_regs->ih_rb_cntl, tmp);
+	}
+
+	if (enable) {
+		ih->enabled = true;
+	} else {
+		/* set rptr, wptr to 0 */
+		WREG32(ih_regs->ih_rb_rptr, 0);
+		WREG32(ih_regs->ih_rb_wptr, 0);
+		ih->enabled = false;
+		ih->rptr = 0;
+	}
+
+	return 0;
+}
+
+/**
+ * ih_v7_0_toggle_interrupts - Toggle all the available interrupt ring buffers
+ *
+ * @adev: amdgpu_device pointer
+ * @enable: enable or disable interrupt ring buffers
+ *
+ * Toggle all the available interrupt ring buffers (IH_V7_0).
+ */
+static int ih_v7_0_toggle_interrupts(struct amdgpu_device *adev, bool enable)
+{
+	struct amdgpu_ih_ring *ih[] = {&adev->irq.ih, &adev->irq.ih1};
+	int i;
+	int r;
+
+	for (i = 0; i < ARRAY_SIZE(ih); i++) {
+		if (ih[i]->ring_size) {
+			r = ih_v7_0_toggle_ring_interrupts(adev, ih[i], enable);
+			if (r)
+				return r;
+		}
+	}
+
+	return 0;
+}
+
+static uint32_t ih_v7_0_rb_cntl(struct amdgpu_ih_ring *ih, uint32_t ih_rb_cntl)
+{
+	int rb_bufsz = order_base_2(ih->ring_size / 4);
+
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+				   MC_SPACE, ih->use_bus_addr ? 2 : 4);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+				   WPTR_OVERFLOW_CLEAR, 1);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+				   WPTR_OVERFLOW_ENABLE, 1);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, RB_SIZE, rb_bufsz);
+	/* Ring Buffer write pointer writeback. If enabled, IH_RB_WPTR register
+	 * value is written to memory
+	 */
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+				   WPTR_WRITEBACK_ENABLE, 1);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_SNOOP, 1);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_RO, 0);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_VMID, 0);
+
+	return ih_rb_cntl;
+}
+
+static uint32_t ih_v7_0_doorbell_rptr(struct amdgpu_ih_ring *ih)
+{
+	u32 ih_doorbell_rtpr = 0;
+
+	if (ih->use_doorbell) {
+		ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
+						 IH_DOORBELL_RPTR, OFFSET,
+						 ih->doorbell_index);
+		ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
+						 IH_DOORBELL_RPTR,
+						 ENABLE, 1);
+	} else {
+		ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
+						 IH_DOORBELL_RPTR,
+						 ENABLE, 0);
+	}
+	return ih_doorbell_rtpr;
+}
+
+/**
+ * ih_v7_0_enable_ring - enable an ih ring buffer
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: amdgpu_ih_ring pointer
+ *
+ * Enable an ih ring buffer (IH_V7_0)
+ */
+static int ih_v7_0_enable_ring(struct amdgpu_device *adev,
+				      struct amdgpu_ih_ring *ih)
+{
+	struct amdgpu_ih_regs *ih_regs;
+	uint32_t tmp;
+
+	ih_regs = &ih->ih_regs;
+
+	/* Ring Buffer base. [39:8] of 40-bit address of the beginning of the ring buffer*/
+	WREG32(ih_regs->ih_rb_base, ih->gpu_addr >> 8);
+	WREG32(ih_regs->ih_rb_base_hi, (ih->gpu_addr >> 40) & 0xff);
+
+	tmp = RREG32(ih_regs->ih_rb_cntl);
+	tmp = ih_v7_0_rb_cntl(ih, tmp);
+	if (ih == &adev->irq.ih)
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RPTR_REARM, !!adev->irq.msi_enabled);
+	if (ih == &adev->irq.ih1) {
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_ENABLE, 0);
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_FULL_DRAIN_ENABLE, 1);
+	}
+
+	if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) {
+		if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp)) {
+			DRM_ERROR("PSP program IH_RB_CNTL failed!\n");
+			return -ETIMEDOUT;
+		}
+	} else {
+		WREG32(ih_regs->ih_rb_cntl, tmp);
+	}
+
+	if (ih == &adev->irq.ih) {
+		/* set the ih ring 0 writeback address whether it's enabled or not */
+		WREG32(ih_regs->ih_rb_wptr_addr_lo, lower_32_bits(ih->wptr_addr));
+		WREG32(ih_regs->ih_rb_wptr_addr_hi, upper_32_bits(ih->wptr_addr) & 0xFFFF);
+	}
+
+	/* set rptr, wptr to 0 */
+	WREG32(ih_regs->ih_rb_wptr, 0);
+	WREG32(ih_regs->ih_rb_rptr, 0);
+
+	WREG32(ih_regs->ih_doorbell_rptr, ih_v7_0_doorbell_rptr(ih));
+
+	return 0;
+}
+
+/**
+ * ih_v7_0_irq_init - init and enable the interrupt ring
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Allocate a ring buffer for the interrupt controller,
+ * enable the RLC, disable interrupts, enable the IH
+ * ring buffer and enable it.
+ * Called at device load and reume.
+ * Returns 0 for success, errors for failure.
+ */
+static int ih_v7_0_irq_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_ih_ring *ih[] = {&adev->irq.ih, &adev->irq.ih1};
+	u32 ih_chicken;
+	u32 tmp;
+	int ret;
+	int i;
+
+	/* disable irqs */
+	ret = ih_v7_0_toggle_interrupts(adev, false);
+	if (ret)
+		return ret;
+
+	adev->nbio.funcs->ih_control(adev);
+
+	if (unlikely((adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) ||
+		     (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO))) {
+		if (ih[0]->use_bus_addr) {
+			ih_chicken = RREG32_SOC15(OSSSYS, 0, regIH_CHICKEN);
+			ih_chicken = REG_SET_FIELD(ih_chicken,
+					IH_CHICKEN, MC_SPACE_GPA_ENABLE, 1);
+			WREG32_SOC15(OSSSYS, 0, regIH_CHICKEN, ih_chicken);
+		}
+	}
+
+	for (i = 0; i < ARRAY_SIZE(ih); i++) {
+		if (ih[i]->ring_size) {
+			ret = ih_v7_0_enable_ring(adev, ih[i]);
+			if (ret)
+				return ret;
+		}
+	}
+
+	/* update doorbell range for ih ring 0 */
+	adev->nbio.funcs->ih_doorbell_range(adev, ih[0]->use_doorbell,
+					    ih[0]->doorbell_index);
+
+	tmp = RREG32_SOC15(OSSSYS, 0, regIH_STORM_CLIENT_LIST_CNTL);
+	tmp = REG_SET_FIELD(tmp, IH_STORM_CLIENT_LIST_CNTL,
+			    CLIENT18_IS_STORM_CLIENT, 1);
+	WREG32_SOC15(OSSSYS, 0, regIH_STORM_CLIENT_LIST_CNTL, tmp);
+
+	tmp = RREG32_SOC15(OSSSYS, 0, regIH_INT_FLOOD_CNTL);
+	tmp = REG_SET_FIELD(tmp, IH_INT_FLOOD_CNTL, FLOOD_CNTL_ENABLE, 1);
+	WREG32_SOC15(OSSSYS, 0, regIH_INT_FLOOD_CNTL, tmp);
+
+	/* GC/MMHUB UTCL2 page fault interrupts are configured as
+	 * MSI storm capable interrupts by deafult. The delay is
+	 * used to avoid ISR being called too frequently
+	 * when page fault happens on several continuous page
+	 * and thus avoid MSI storm */
+	tmp = RREG32_SOC15(OSSSYS, 0, regIH_MSI_STORM_CTRL);
+	tmp = REG_SET_FIELD(tmp, IH_MSI_STORM_CTRL,
+			    DELAY, 3);
+	WREG32_SOC15(OSSSYS, 0, regIH_MSI_STORM_CTRL, tmp);
+
+	/* Redirect the interrupts to IH RB1 for dGPU */
+	if (adev->irq.ih1.ring_size) {
+		tmp = RREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_INDEX);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_INDEX, INDEX, 0);
+		WREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_INDEX, tmp);
+
+		tmp = RREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_DATA);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA, CLIENT_ID, 0xa);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA, SOURCE_ID, 0x0);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA,
+				    SOURCE_ID_MATCH_ENABLE, 0x1);
+
+		WREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_DATA, tmp);
+	}
+
+	pci_set_master(adev->pdev);
+
+	/* enable interrupts */
+	ret = ih_v7_0_toggle_interrupts(adev, true);
+	if (ret)
+		return ret;
+	/* enable wptr force update for self int */
+	force_update_wptr_for_self_int(adev, 0, 8, true);
+
+	if (adev->irq.ih_soft.ring_size)
+		adev->irq.ih_soft.enabled = true;
+
+	return 0;
+}
+
+/**
+ * ih_v7_0_irq_disable - disable interrupts
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Disable interrupts on the hw.
+ */
+static void ih_v7_0_irq_disable(struct amdgpu_device *adev)
+{
+	force_update_wptr_for_self_int(adev, 0, 8, false);
+	ih_v7_0_toggle_interrupts(adev, false);
+
+	/* Wait and acknowledge irq */
+	mdelay(1);
+}
+
+/**
+ * ih_v7_0_get_wptr() - get the IH ring buffer wptr
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: IH ring buffer to fetch wptr
+ *
+ * Get the IH ring buffer wptr from either the register
+ * or the writeback memory buffer.  Also check for
+ * ring buffer overflow and deal with it.
+ * Returns the value of the wptr.
+ */
+static u32 ih_v7_0_get_wptr(struct amdgpu_device *adev,
+			      struct amdgpu_ih_ring *ih)
+{
+	u32 wptr, tmp;
+	struct amdgpu_ih_regs *ih_regs;
+
+	wptr = le32_to_cpu(*ih->wptr_cpu);
+	ih_regs = &ih->ih_regs;
+
+	if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
+		goto out;
+
+	wptr = RREG32_NO_KIQ(ih_regs->ih_rb_wptr);
+	if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
+		goto out;
+	wptr = REG_SET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW, 0);
+
+	/* When a ring buffer overflow happen start parsing interrupt
+	 * from the last not overwritten vector (wptr + 32). Hopefully
+	 * this should allow us to catch up.
+	 */
+	tmp = (wptr + 32) & ih->ptr_mask;
+	dev_warn(adev->dev, "IH ring buffer overflow "
+		 "(0x%08X, 0x%08X, 0x%08X)\n",
+		 wptr, ih->rptr, tmp);
+	ih->rptr = tmp;
+
+	tmp = RREG32_NO_KIQ(ih_regs->ih_rb_cntl);
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
+	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+
+	/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+	 * can be detected.
+	 */
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+out:
+	return (wptr & ih->ptr_mask);
+}
+
+/**
+ * ih_v7_0_irq_rearm - rearm IRQ if lost
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: IH ring to match
+ *
+ */
+static void ih_v7_0_irq_rearm(struct amdgpu_device *adev,
+			       struct amdgpu_ih_ring *ih)
+{
+	uint32_t v = 0;
+	uint32_t i = 0;
+	struct amdgpu_ih_regs *ih_regs;
+
+	ih_regs = &ih->ih_regs;
+
+	/* Rearm IRQ / re-write doorbell if doorbell write is lost */
+	for (i = 0; i < MAX_REARM_RETRY; i++) {
+		v = RREG32_NO_KIQ(ih_regs->ih_rb_rptr);
+		if ((v < ih->ring_size) && (v != ih->rptr))
+			WDOORBELL32(ih->doorbell_index, ih->rptr);
+		else
+			break;
+	}
+}
+
+/**
+ * ih_v7_0_set_rptr - set the IH ring buffer rptr
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: IH ring buffer to set rptr
+ */
+static void ih_v7_0_set_rptr(struct amdgpu_device *adev,
+			       struct amdgpu_ih_ring *ih)
+{
+	struct amdgpu_ih_regs *ih_regs;
+
+	if (ih->use_doorbell) {
+		/* XXX check if swapping is necessary on BE */
+		*ih->rptr_cpu = ih->rptr;
+		WDOORBELL32(ih->doorbell_index, ih->rptr);
+
+		if (amdgpu_sriov_vf(adev))
+			ih_v7_0_irq_rearm(adev, ih);
+	} else {
+		ih_regs = &ih->ih_regs;
+		WREG32(ih_regs->ih_rb_rptr, ih->rptr);
+	}
+}
+
+/**
+ * ih_v7_0_self_irq - dispatch work for ring 1
+ *
+ * @adev: amdgpu_device pointer
+ * @source: irq source
+ * @entry: IV with WPTR update
+ *
+ * Update the WPTR from the IV and schedule work to handle the entries.
+ */
+static int ih_v7_0_self_irq(struct amdgpu_device *adev,
+			      struct amdgpu_irq_src *source,
+			      struct amdgpu_iv_entry *entry)
+{
+	uint32_t wptr = cpu_to_le32(entry->src_data[0]);
+
+	switch (entry->ring_id) {
+	case 1:
+		*adev->irq.ih1.wptr_cpu = wptr;
+		schedule_work(&adev->irq.ih1_work);
+		break;
+	default: break;
+	}
+	return 0;
+}
+
+static const struct amdgpu_irq_src_funcs ih_v7_0_self_irq_funcs = {
+	.process = ih_v7_0_self_irq,
+};
+
+static void ih_v7_0_set_self_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->irq.self_irq.num_types = 0;
+	adev->irq.self_irq.funcs = &ih_v7_0_self_irq_funcs;
+}
+
+static int ih_v7_0_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	ih_v7_0_set_interrupt_funcs(adev);
+	ih_v7_0_set_self_irq_funcs(adev);
+	return 0;
+}
+
+static int ih_v7_0_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+	bool use_bus_addr;
+
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_IH, 0,
+			      &adev->irq.self_irq);
+
+	if (r)
+		return r;
+
+	/* use gpu virtual address for ih ring
+	 * until ih_checken is programmed to allow
+	 * use bus address for ih ring by psp bl */
+	use_bus_addr = adev->firmware.load_type != AMDGPU_FW_LOAD_PSP;
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 256 * 1024, use_bus_addr);
+	if (r)
+		return r;
+
+	adev->irq.ih.use_doorbell = true;
+	adev->irq.ih.doorbell_index = adev->doorbell_index.ih << 1;
+
+	if (!(adev->flags & AMD_IS_APU)) {
+		r = amdgpu_ih_ring_init(adev, &adev->irq.ih1, IH_RING_SIZE,
+					use_bus_addr);
+		if (r)
+			return r;
+
+		adev->irq.ih1.use_doorbell = true;
+		adev->irq.ih1.doorbell_index = (adev->doorbell_index.ih + 1) << 1;
+	}
+
+	/* initialize ih control register offset */
+	ih_v7_0_init_register_offset(adev);
+
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, PAGE_SIZE, true);
+	if (r)
+		return r;
+
+	r = amdgpu_irq_init(adev);
+
+	return r;
+}
+
+static int ih_v7_0_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	amdgpu_irq_fini_sw(adev);
+
+	return 0;
+}
+
+static int ih_v7_0_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	r = ih_v7_0_irq_init(adev);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static int ih_v7_0_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	ih_v7_0_irq_disable(ip_block->adev);
+
+	return 0;
+}
+
+static int ih_v7_0_suspend(struct amdgpu_ip_block *ip_block)
+{
+	return ih_v7_0_hw_fini(ip_block);
+}
+
+static int ih_v7_0_resume(struct amdgpu_ip_block *ip_block)
+{
+	return ih_v7_0_hw_init(ip_block);
+}
+
+static bool ih_v7_0_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	/* todo */
+	return true;
+}
+
+static int ih_v7_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	/* todo */
+	return -ETIMEDOUT;
+}
+
+static int ih_v7_0_soft_reset(struct amdgpu_ip_block *ip_block)
+{
+	/* todo */
+	return 0;
+}
+
+static void ih_v7_0_update_clockgating_state(struct amdgpu_device *adev,
+					       bool enable)
+{
+	uint32_t data, def, field_val;
+
+	if (adev->cg_flags & AMD_CG_SUPPORT_IH_CG) {
+		def = data = RREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL);
+		field_val = enable ? 0 : 1;
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     DBUS_MUX_CLK_SOFT_OVERRIDE, field_val);
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     OSSSYS_SHARE_CLK_SOFT_OVERRIDE, field_val);
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     LIMIT_SMN_CLK_SOFT_OVERRIDE, field_val);
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     DYN_CLK_SOFT_OVERRIDE, field_val);
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     REG_CLK_SOFT_OVERRIDE, field_val);
+		if (def != data)
+			WREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL, data);
+	}
+
+	return;
+}
+
+static int ih_v7_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					   enum amd_clockgating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	ih_v7_0_update_clockgating_state(adev,
+				state == AMD_CG_STATE_GATE);
+	return 0;
+}
+
+static void ih_v7_0_update_ih_mem_power_gating(struct amdgpu_device *adev,
+					       bool enable)
+{
+	uint32_t ih_mem_pwr_cntl;
+
+	/* Disable ih sram power cntl before switch powergating mode */
+	ih_mem_pwr_cntl = RREG32_SOC15(OSSSYS, 0, regIH_MEM_POWER_CTRL);
+	ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+					IH_BUFFER_MEM_POWER_CTRL_EN, 0);
+	WREG32_SOC15(OSSSYS, 0, regIH_MEM_POWER_CTRL, ih_mem_pwr_cntl);
+
+	/* It is recommended to set mem powergating mode to DS mode */
+	if (enable) {
+		/* mem power mode */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_LS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_DS_EN, 1);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_SD_EN, 0);
+		/* cam mem power mode */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_LS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_DS_EN, 1);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_SD_EN, 0);
+		/* re-enable power cntl */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_CTRL_EN, 1);
+	} else {
+		/* mem power mode */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_LS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_DS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_SD_EN, 0);
+		/* cam mem power mode */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_LS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_DS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_SD_EN, 0);
+		/* re-enable power cntl*/
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_CTRL_EN, 1);
+	}
+
+	WREG32_SOC15(OSSSYS, 0, regIH_MEM_POWER_CTRL, ih_mem_pwr_cntl);
+}
+
+static int ih_v7_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					 enum amd_powergating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	bool enable = (state == AMD_PG_STATE_GATE);
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_IH_SRAM_PG)
+		ih_v7_0_update_ih_mem_power_gating(adev, enable);
+
+	return 0;
+}
+
+static void ih_v7_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (!RREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL))
+		*flags |= AMD_CG_SUPPORT_IH_CG;
+
+	return;
+}
+
+static const struct amd_ip_funcs ih_v7_0_ip_funcs = {
+	.name = "ih_v7_0",
+	.early_init = ih_v7_0_early_init,
+	.sw_init = ih_v7_0_sw_init,
+	.sw_fini = ih_v7_0_sw_fini,
+	.hw_init = ih_v7_0_hw_init,
+	.hw_fini = ih_v7_0_hw_fini,
+	.suspend = ih_v7_0_suspend,
+	.resume = ih_v7_0_resume,
+	.is_idle = ih_v7_0_is_idle,
+	.wait_for_idle = ih_v7_0_wait_for_idle,
+	.soft_reset = ih_v7_0_soft_reset,
+	.set_clockgating_state = ih_v7_0_set_clockgating_state,
+	.set_powergating_state = ih_v7_0_set_powergating_state,
+	.get_clockgating_state = ih_v7_0_get_clockgating_state,
+};
+
+static const struct amdgpu_ih_funcs ih_v7_0_funcs = {
+	.get_wptr = ih_v7_0_get_wptr,
+	.decode_iv = amdgpu_ih_decode_iv_helper,
+	.decode_iv_ts = amdgpu_ih_decode_iv_ts_helper,
+	.set_rptr = ih_v7_0_set_rptr
+};
+
+static void ih_v7_0_set_interrupt_funcs(struct amdgpu_device *adev)
+{
+	adev->irq.ih_funcs = &ih_v7_0_funcs;
+}
+
+const struct amdgpu_ip_block_version ih_v7_0_ip_block =
+{
+	.type = AMD_IP_BLOCK_TYPE_IH,
+	.major = 7,
+	.minor = 0,
+	.rev = 0,
+	.funcs = &ih_v7_0_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/ih_v7_0.h b/drivers/gpu/drm/amd/amdgpu/ih_v7_0.h
new file mode 100644
index 000000000000..af9dcbc451fd
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/ih_v7_0.h
@@ -0,0 +1,28 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __IH_V7_0_IH_H__
+#define __IH_V7_0_IH_H__
+
+extern const struct amdgpu_ip_block_version ih_v7_0_ip_block;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/imu_v11_0.c b/drivers/gpu/drm/amd/amdgpu/imu_v11_0.c
index 4ab90c7852c3..cc626036ed9c 100644
--- a/drivers/gpu/drm/amd/amdgpu/imu_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/imu_v11_0.c
@@ -32,14 +32,18 @@
 #include "gc/gc_11_0_0_sh_mask.h"
 
 MODULE_FIRMWARE("amdgpu/gc_11_0_0_imu.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_0_0_imu_kicker.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_1_imu.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_2_imu.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_3_imu.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_4_imu.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_0_imu.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_1_imu.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_2_imu.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_3_imu.bin");
 
 static int imu_v11_0_init_microcode(struct amdgpu_device *adev)
 {
-	char fw_name[40];
 	char ucode_prefix[30];
 	int err;
 	const struct imu_firmware_header_v1_0 *imu_hdr;
@@ -48,13 +52,16 @@ static int imu_v11_0_init_microcode(struct amdgpu_device *adev)
 	DRM_DEBUG("\n");
 
 	amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix));
-
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_imu.bin", ucode_prefix);
-	err = amdgpu_ucode_request(adev, &adev->gfx.imu_fw, fw_name);
+	if (amdgpu_is_kicker_fw(adev))
+		err = amdgpu_ucode_request(adev, &adev->gfx.imu_fw, AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_imu_kicker.bin", ucode_prefix);
+	else
+		err = amdgpu_ucode_request(adev, &adev->gfx.imu_fw, AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_imu.bin", ucode_prefix);
 	if (err)
 		goto out;
+
 	imu_hdr = (const struct imu_firmware_header_v1_0 *)adev->gfx.imu_fw->data;
-	adev->gfx.imu_fw_version = le32_to_cpu(imu_hdr->header.ucode_version);
 	//adev->gfx.imu_feature_version = le32_to_cpu(imu_hdr->ucode_feature_version);
 	
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
@@ -68,13 +75,14 @@ static int imu_v11_0_init_microcode(struct amdgpu_device *adev)
 		info->fw = adev->gfx.imu_fw;
 		adev->firmware.fw_size +=
 			ALIGN(le32_to_cpu(imu_hdr->imu_dram_ucode_size_bytes), PAGE_SIZE);
-	}
+	} else
+		adev->gfx.imu_fw_version = le32_to_cpu(imu_hdr->header.ucode_version);
 
 out:
 	if (err) {
 		dev_err(adev->dev,
-			"gfx11: Failed to load firmware \"%s\"\n",
-			fw_name);
+			"gfx11: Failed to load firmware \"%s_imu.bin\"\n",
+			ucode_prefix);
 		amdgpu_ucode_release(&adev->gfx.imu_fw);
 	}
 
@@ -152,7 +160,7 @@ static void imu_v11_0_setup(struct amdgpu_device *adev)
 		WREG32_SOC15(GC, 0, regGFX_IMU_C2PMSG_16, imu_reg_val);
 	}
 
-	//disble imu Rtavfs, SmsRepair, DfllBTC, and ClkB
+	//disable imu Rtavfs, SmsRepair, DfllBTC, and ClkB
 	imu_reg_val = RREG32_SOC15(GC, 0, regGFX_IMU_SCRATCH_10);
 	imu_reg_val |= 0x10007;
 	WREG32_SOC15(GC, 0, regGFX_IMU_SCRATCH_10, imu_reg_val);
@@ -352,7 +360,7 @@ static void imu_v11_0_program_rlc_ram(struct amdgpu_device *adev)
 
 	WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_INDEX, 0x2);
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 0):
 		program_imu_rlc_ram(adev, imu_rlc_ram_golden_11,
 				(const u32)ARRAY_SIZE(imu_rlc_ram_golden_11));
diff --git a/drivers/gpu/drm/amd/amdgpu/imu_v12_0.c b/drivers/gpu/drm/amd/amdgpu/imu_v12_0.c
new file mode 100644
index 000000000000..58cd87db8061
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/imu_v12_0.c
@@ -0,0 +1,406 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/firmware.h>
+#include "amdgpu.h"
+#include "amdgpu_imu.h"
+#include "amdgpu_dpm.h"
+
+#include "imu_v12_0.h"
+
+#include "gc/gc_12_0_0_offset.h"
+#include "gc/gc_12_0_0_sh_mask.h"
+#include "mmhub/mmhub_4_1_0_offset.h"
+
+MODULE_FIRMWARE("amdgpu/gc_12_0_0_imu.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_imu.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_imu_kicker.bin");
+
+#define TRANSFER_RAM_MASK	0x001c0000
+
+static int imu_v12_0_init_microcode(struct amdgpu_device *adev)
+{
+	char ucode_prefix[30];
+	int err;
+	const struct imu_firmware_header_v1_0 *imu_hdr;
+	struct amdgpu_firmware_info *info = NULL;
+
+	DRM_DEBUG("\n");
+
+	amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix));
+	if (amdgpu_is_kicker_fw(adev))
+		err = amdgpu_ucode_request(adev, &adev->gfx.imu_fw, AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_imu_kicker.bin", ucode_prefix);
+	else
+		err = amdgpu_ucode_request(adev, &adev->gfx.imu_fw, AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_imu.bin", ucode_prefix);
+	if (err)
+		goto out;
+
+	imu_hdr = (const struct imu_firmware_header_v1_0 *)adev->gfx.imu_fw->data;
+	adev->gfx.imu_fw_version = le32_to_cpu(imu_hdr->header.ucode_version);
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		info = &adev->firmware.ucode[AMDGPU_UCODE_ID_IMU_I];
+		info->ucode_id = AMDGPU_UCODE_ID_IMU_I;
+		info->fw = adev->gfx.imu_fw;
+		adev->firmware.fw_size +=
+			ALIGN(le32_to_cpu(imu_hdr->imu_iram_ucode_size_bytes), PAGE_SIZE);
+		info = &adev->firmware.ucode[AMDGPU_UCODE_ID_IMU_D];
+		info->ucode_id = AMDGPU_UCODE_ID_IMU_D;
+		info->fw = adev->gfx.imu_fw;
+		adev->firmware.fw_size +=
+			ALIGN(le32_to_cpu(imu_hdr->imu_dram_ucode_size_bytes), PAGE_SIZE);
+	}
+
+out:
+	if (err) {
+		dev_err(adev->dev,
+			"gfx12: Failed to load firmware \"%s_imu.bin\"\n",
+			ucode_prefix);
+		amdgpu_ucode_release(&adev->gfx.imu_fw);
+	}
+
+	return err;
+}
+
+static int imu_v12_0_load_microcode(struct amdgpu_device *adev)
+{
+	const struct imu_firmware_header_v1_0 *hdr;
+	const __le32 *fw_data;
+	unsigned i, fw_size;
+
+	if (!adev->gfx.imu_fw)
+		return -EINVAL;
+
+	hdr = (const struct imu_firmware_header_v1_0 *)adev->gfx.imu_fw->data;
+
+	fw_data = (const __le32 *)(adev->gfx.imu_fw->data +
+			le32_to_cpu(hdr->header.ucode_array_offset_bytes));
+	fw_size = le32_to_cpu(hdr->imu_iram_ucode_size_bytes) / 4;
+
+	WREG32_SOC15(GC, 0, regGFX_IMU_I_RAM_ADDR, 0);
+
+	for (i = 0; i < fw_size; i++)
+		WREG32_SOC15(GC, 0, regGFX_IMU_I_RAM_DATA, le32_to_cpup(fw_data++));
+
+	WREG32_SOC15(GC, 0, regGFX_IMU_I_RAM_ADDR, adev->gfx.imu_fw_version);
+
+	fw_data = (const __le32 *)(adev->gfx.imu_fw->data +
+			le32_to_cpu(hdr->header.ucode_array_offset_bytes) +
+			le32_to_cpu(hdr->imu_iram_ucode_size_bytes));
+	fw_size = le32_to_cpu(hdr->imu_dram_ucode_size_bytes) / 4;
+
+	WREG32_SOC15(GC, 0, regGFX_IMU_D_RAM_ADDR, 0);
+
+	for (i = 0; i < fw_size; i++)
+		WREG32_SOC15(GC, 0, regGFX_IMU_D_RAM_DATA, le32_to_cpup(fw_data++));
+
+	WREG32_SOC15(GC, 0, regGFX_IMU_D_RAM_ADDR, adev->gfx.imu_fw_version);
+
+	return 0;
+}
+
+static int imu_v12_0_wait_for_reset_status(struct amdgpu_device *adev)
+{
+	u32 imu_reg_val = 0;
+	int i;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		imu_reg_val = RREG32_SOC15(GC, 0, regGFX_IMU_GFX_RESET_CTRL);
+		if ((imu_reg_val & 0x1f) == 0x1f)
+			break;
+		udelay(1);
+	}
+
+	if (i >= adev->usec_timeout) {
+		dev_err(adev->dev, "init imu: IMU start timeout\n");
+		return -ETIMEDOUT;
+	}
+
+	return 0;
+}
+
+static void imu_v12_0_setup(struct amdgpu_device *adev)
+{
+	u32 imu_reg_val;
+
+	WREG32_SOC15(GC, 0, regGFX_IMU_C2PMSG_ACCESS_CTRL0, 0xffffff);
+	WREG32_SOC15(GC, 0, regGFX_IMU_C2PMSG_ACCESS_CTRL1, 0xffff);
+
+	if (adev->gfx.imu.mode == DEBUG_MODE) {
+		imu_reg_val = RREG32_SOC15(GC, 0, regGFX_IMU_C2PMSG_16);
+		imu_reg_val |= 0x1;
+		WREG32_SOC15(GC, 0, regGFX_IMU_C2PMSG_16, imu_reg_val);
+
+		imu_reg_val = RREG32_SOC15(GC, 0, regGFX_IMU_SCRATCH_10);
+		imu_reg_val |= 0x20010007;
+		WREG32_SOC15(GC, 0, regGFX_IMU_SCRATCH_10, imu_reg_val);
+
+	}
+}
+
+static int imu_v12_0_start(struct amdgpu_device *adev)
+{
+	u32 imu_reg_val;
+
+	imu_reg_val = RREG32_SOC15(GC, 0, regGFX_IMU_CORE_CTRL);
+	imu_reg_val &= 0xfffffffe;
+	WREG32_SOC15(GC, 0, regGFX_IMU_CORE_CTRL, imu_reg_val);
+
+	if (adev->flags & AMD_IS_APU)
+		amdgpu_dpm_set_gfx_power_up_by_imu(adev);
+
+	return imu_v12_0_wait_for_reset_status(adev);
+}
+
+static const struct imu_rlc_ram_golden imu_rlc_ram_golden_12_0_1[] = {
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCH_PIPE_STEER, 0x1e4, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL1X_PIPE_STEER, 0x1e4, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL1_PIPE_STEER, 0x1e4, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL2_PIPE_STEER_0, 0x13571357, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL2_PIPE_STEER_1, 0x64206420, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL2_PIPE_STEER_2, 0x2460246, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL2_PIPE_STEER_3, 0x75317531, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL2C_CTRL3, 0xc0d41183, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regSDMA0_CHICKEN_BITS, 0x507d1c0, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regSDMA1_CHICKEN_BITS, 0x507d1c0, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCP_RB_WPTR_POLL_CNTL, 0x600100, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_SDP_CREDITS, 0x3f7fff, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_SDP_CREDITS, 0x3f7ebf, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_SDP_TAG_RESERVE0, 0x2e00000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_SDP_TAG_RESERVE1, 0x1a078, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_SDP_TAG_RESERVE2, 0x0, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_SDP_TAG_RESERVE0, 0x0, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_SDP_TAG_RESERVE1, 0x12030, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_SDP_TAG_RESERVE2, 0x0, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_SDP_VCC_RESERVE0, 0x19041000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_SDP_VCC_RESERVE1, 0x80000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_SDP_VCC_RESERVE0, 0x1e080000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_SDP_VCC_RESERVE1, 0x80000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_SDP_PRIORITY, 0x880, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_SDP_PRIORITY, 0x8880, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_SDP_ARB_FINAL, 0x17, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_SDP_ARB_FINAL, 0x77, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_SDP_ENABLE, 0x00000001, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_SDP_ENABLE, 0x00000001, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL2, 0x20000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_APT_CNTL, 0x0c, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_CACHEABLE_DRAM_ADDRESS_END, 0xfffff, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_MISC, 0x0091, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_MISC, 0x0091, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGRBM_GFX_INDEX, 0xe0000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCR_GENERAL_CNTL, 0x00008500, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regPA_CL_ENHANCE, 0x00880007, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regTD_CNTL, 0x00000001, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGRBM_GFX_INDEX, 0x00000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regRMI_GENERAL_CNTL, 0x01e00000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGRBM_GFX_INDEX, 0x00000001, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regRMI_GENERAL_CNTL, 0x01e00000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGRBM_GFX_INDEX, 0x00000100, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regRMI_GENERAL_CNTL, 0x01e00000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGRBM_GFX_INDEX, 0x00000101, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regRMI_GENERAL_CNTL, 0x01e00000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGRBM_GFX_INDEX, 0xe0000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGB_ADDR_CONFIG, 0x08200545, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGRBMH_CP_PERFMON_CNTL, 0x00000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCB_PERFCOUNTER0_SELECT1, 0x000fffff, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCP_DEBUG_2, 0x00020000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCP_CPC_DEBUG, 0x00500010, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, 0x00000500, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x00000001, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0x00000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_LOCAL_FB_ADDRESS_START, 0x00000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_LOCAL_FB_ADDRESS_END, 0x0000000f, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_FB_LOCATION_BASE, 0x00006000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_FB_LOCATION_TOP, 0x0000600f, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_CONTEXT0_CNTL, 0x00000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_CONTEXT1_CNTL, 0x00000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_NB_TOP_OF_DRAM_SLOT1, 0xff800000, 0xe0000000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_NB_LOWER_TOP_OF_DRAM2, 0x00000001, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_NB_UPPER_TOP_OF_DRAM2, 0x0000ffff, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_AGP_BASE, 0x00000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_AGP_BOT, 0x00000002, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_AGP_TOP, 0x00000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL, 0x00001ffc, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, 0x00000551, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL, 0x00080603, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL2, 0x00000003, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL3, 0x00100003, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL5, 0x00003fe0, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x0003d000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0x0003d7ff, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB, 0, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB, 0, 0x1c0000)
+};
+
+static void program_imu_rlc_ram_old(struct amdgpu_device *adev,
+				    const struct imu_rlc_ram_golden *regs,
+				    const u32 array_size)
+{
+	const struct imu_rlc_ram_golden *entry;
+	u32 reg, data;
+	int i;
+
+	for (i = 0; i < array_size; ++i) {
+		entry = &regs[i];
+		reg =  adev->reg_offset[entry->hwip][entry->instance][entry->segment] + entry->reg;
+		reg |= entry->addr_mask;
+		data = entry->data;
+		if (entry->reg == regGCMC_VM_AGP_BASE)
+			data = 0x00ffffff;
+		else if (entry->reg == regGCMC_VM_AGP_TOP)
+			data = 0x0;
+		else if (entry->reg == regGCMC_VM_FB_LOCATION_BASE)
+			data = adev->gmc.vram_start >> 24;
+		else if (entry->reg == regGCMC_VM_FB_LOCATION_TOP)
+			data = adev->gmc.vram_end >> 24;
+
+		WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_HIGH, 0);
+		WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_LOW, reg);
+		WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_DATA, data);
+	}
+}
+
+static u32 imu_v12_0_grbm_gfx_index_remap(struct amdgpu_device *adev,
+					  u32 data, bool high)
+{
+	u32 val, inst_index;
+
+	inst_index = REG_GET_FIELD(data, GRBM_GFX_INDEX, INSTANCE_INDEX);
+
+	if (high)
+		val = inst_index >> 5;
+	else
+		val = REG_GET_FIELD(data, GRBM_GFX_INDEX, SE_BROADCAST_WRITES) << 18 |
+		      REG_GET_FIELD(data, GRBM_GFX_INDEX, SA_BROADCAST_WRITES) << 19 |
+		      REG_GET_FIELD(data, GRBM_GFX_INDEX, INSTANCE_BROADCAST_WRITES) << 20 |
+		      REG_GET_FIELD(data, GRBM_GFX_INDEX, SE_INDEX) << 21 |
+		      REG_GET_FIELD(data, GRBM_GFX_INDEX, SA_INDEX) << 25 |
+		      (inst_index & 0x1f);
+
+	return val;
+}
+
+static u32 imu_v12_init_gfxhub_settings(struct amdgpu_device *adev,
+					u32 reg, u32 data)
+{
+	if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_FB_LOCATION_BASE))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_FB_LOCATION_BASE);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_FB_LOCATION_TOP))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_FB_LOCATION_TOP);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_FB_OFFSET))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_FB_OFFSET);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_AGP_BASE))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_BASE);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_AGP_BOT))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_BOT);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_AGP_TOP))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_TOP);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_SYSTEM_APERTURE_LOW_ADDR))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_LOW_ADDR);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_SYSTEM_APERTURE_HIGH_ADDR))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_HIGH_ADDR);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_LOCAL_FB_ADDRESS_START))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_LOCAL_FB_ADDRESS_START);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_LOCAL_FB_ADDRESS_END))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_LOCAL_FB_ADDRESS_END);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_LOCAL_SYSMEM_ADDRESS_START))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_LOCAL_SYSMEM_ADDRESS_START);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_LOCAL_SYSMEM_ADDRESS_END))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_LOCAL_SYSMEM_ADDRESS_END);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB);
+	else
+		return data;
+}
+
+static void program_imu_rlc_ram(struct amdgpu_device *adev,
+				const u32 *regs,
+				const u32 array_size)
+{
+	u32 reg, data, val_h = 0, val_l = TRANSFER_RAM_MASK;
+	int i;
+
+	if (array_size % 3)
+		return;
+
+	for (i = 0; i < array_size; i += 3) {
+		reg = regs[i + 0];
+		data = regs[i + 2];
+		data = imu_v12_init_gfxhub_settings(adev, reg, data);
+		if (reg == SOC15_REG_OFFSET(GC, 0, regGRBM_GFX_INDEX)) {
+			val_l = imu_v12_0_grbm_gfx_index_remap(adev, data, false);
+			val_h = imu_v12_0_grbm_gfx_index_remap(adev, data, true);
+		} else {
+			WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_HIGH, val_h);
+			WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_LOW, reg | val_l);
+			WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_DATA, data);
+		}
+	}
+}
+
+static void imu_v12_0_program_rlc_ram(struct amdgpu_device *adev)
+{
+	u32 reg_data, size = 0;
+	const u32 *data = NULL;
+	int r = -EINVAL;
+
+	WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_INDEX, 0x2);
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		if (!r)
+			program_imu_rlc_ram(adev, data, (const u32)size);
+		else
+			program_imu_rlc_ram_old(adev, imu_rlc_ram_golden_12_0_1,
+				(const u32)ARRAY_SIZE(imu_rlc_ram_golden_12_0_1));
+		break;
+	default:
+		BUG();
+		break;
+	}
+
+	//Indicate the latest entry
+	WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_HIGH, 0);
+	WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_LOW, 0);
+	WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_DATA, 0);
+
+	reg_data = RREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_INDEX);
+	reg_data |= GFX_IMU_RLC_RAM_INDEX__RAM_VALID_MASK;
+	WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_INDEX, reg_data);
+}
+
+const struct amdgpu_imu_funcs gfx_v12_0_imu_funcs = {
+	.init_microcode = imu_v12_0_init_microcode,
+	.load_microcode = imu_v12_0_load_microcode,
+	.setup_imu = imu_v12_0_setup,
+	.start_imu = imu_v12_0_start,
+	.program_rlc_ram = imu_v12_0_program_rlc_ram,
+	.wait_for_reset_status = imu_v12_0_wait_for_reset_status,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v10_1.h b/drivers/gpu/drm/amd/amdgpu/imu_v12_0.h
index 9afd6ddb01e9..a1f50cb1aeab 100644
--- a/drivers/gpu/drm/amd/amdgpu/mes_v10_1.h
+++ b/drivers/gpu/drm/amd/amdgpu/imu_v12_0.h
@@ -1,5 +1,5 @@
 /*
- * Copyright 2019 Advanced Micro Devices, Inc.
+ * Copyright 2021 Advanced Micro Devices, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -21,9 +21,10 @@
  *
  */
 
-#ifndef __MES_V10_1_H__
-#define __MES_V10_1_H__
+#ifndef __IMU_V12_0_H__
+#define __IMU_V12_0_H__
 
-extern const struct amdgpu_ip_block_version mes_v10_1_ip_block;
+extern const struct amdgpu_imu_funcs gfx_v12_0_imu_funcs;
 
 #endif
+
diff --git a/drivers/gpu/drm/amd/amdgpu/isp_v4_1_0.c b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_0.c
new file mode 100644
index 000000000000..0027a639c7e6
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_0.c
@@ -0,0 +1,191 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright (C) 2024 Advanced Micro Devices, Inc. All rights reserved.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ */
+
+#include "amdgpu.h"
+#include "isp_v4_1_0.h"
+
+static const unsigned int isp_4_1_0_int_srcid[MAX_ISP410_INT_SRC] = {
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT9,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT10,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT11,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT12,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT13,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT14,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT15,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT16
+};
+
+static int isp_v4_1_0_hw_init(struct amdgpu_isp *isp)
+{
+	struct amdgpu_device *adev = isp->adev;
+	int idx, int_idx, num_res, r;
+	u64 isp_base;
+
+	if (adev->rmmio_size == 0 || adev->rmmio_size < 0x5289)
+		return -EINVAL;
+
+	isp_base = adev->rmmio_base;
+
+	isp->isp_cell = kcalloc(3, sizeof(struct mfd_cell), GFP_KERNEL);
+	if (!isp->isp_cell) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev,
+			"%s: isp mfd cell alloc failed\n", __func__);
+		goto failure;
+	}
+
+	num_res = MAX_ISP410_MEM_RES + MAX_ISP410_INT_SRC;
+	isp->isp_res = kcalloc(num_res, sizeof(struct resource),
+			       GFP_KERNEL);
+	if (!isp->isp_res) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev,
+			"%s: isp mfd res alloc failed\n", __func__);
+		goto failure;
+	}
+
+	isp->isp_pdata = kzalloc(sizeof(*isp->isp_pdata), GFP_KERNEL);
+	if (!isp->isp_pdata) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev,
+			"%s: isp platform data alloc failed\n", __func__);
+		goto failure;
+	}
+
+	/* initialize isp platform data */
+	isp->isp_pdata->adev = (void *)adev;
+	isp->isp_pdata->asic_type = adev->asic_type;
+	isp->isp_pdata->base_rmmio_size = adev->rmmio_size;
+
+	isp->isp_res[0].name = "isp_4_1_0_reg";
+	isp->isp_res[0].flags = IORESOURCE_MEM;
+	isp->isp_res[0].start = isp_base;
+	isp->isp_res[0].end = isp_base + ISP_REGS_OFFSET_END;
+
+	isp->isp_res[1].name = "isp_4_1_phy0_reg";
+	isp->isp_res[1].flags = IORESOURCE_MEM;
+	isp->isp_res[1].start = isp_base + ISP410_PHY0_OFFSET;
+	isp->isp_res[1].end = isp_base + ISP410_PHY0_OFFSET + ISP410_PHY0_SIZE;
+
+	for (idx = MAX_ISP410_MEM_RES, int_idx = 0; idx < num_res; idx++, int_idx++) {
+		isp->isp_res[idx].name = "isp_4_1_0_irq";
+		isp->isp_res[idx].flags = IORESOURCE_IRQ;
+		isp->isp_res[idx].start =
+			amdgpu_irq_create_mapping(adev, isp_4_1_0_int_srcid[int_idx]);
+		isp->isp_res[idx].end =
+			isp->isp_res[idx].start;
+	}
+
+	isp->isp_cell[0].name = "amd_isp_capture";
+	isp->isp_cell[0].num_resources = num_res;
+	isp->isp_cell[0].resources = &isp->isp_res[0];
+	isp->isp_cell[0].platform_data = isp->isp_pdata;
+	isp->isp_cell[0].pdata_size = sizeof(struct isp_platform_data);
+
+	/* initialize isp i2c platform data */
+	isp->isp_i2c_res = kcalloc(1, sizeof(struct resource), GFP_KERNEL);
+	if (!isp->isp_i2c_res) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev,
+			"%s: isp mfd res alloc failed\n", __func__);
+		goto failure;
+	}
+
+	isp->isp_i2c_res[0].name = "isp_i2c0_reg";
+	isp->isp_i2c_res[0].flags = IORESOURCE_MEM;
+	isp->isp_i2c_res[0].start = isp_base + ISP410_I2C0_OFFSET;
+	isp->isp_i2c_res[0].end = isp_base + ISP410_I2C0_OFFSET + ISP410_I2C0_SIZE;
+
+	isp->isp_cell[1].name = "amd_isp_i2c_designware";
+	isp->isp_cell[1].num_resources = 1;
+	isp->isp_cell[1].resources = &isp->isp_i2c_res[0];
+	isp->isp_cell[1].platform_data = isp->isp_pdata;
+	isp->isp_cell[1].pdata_size = sizeof(struct isp_platform_data);
+
+	/* initialize isp gpiochip platform data */
+	isp->isp_gpio_res = kcalloc(1, sizeof(struct resource), GFP_KERNEL);
+	if (!isp->isp_gpio_res) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev,
+			"%s: isp gpio res alloc failed\n", __func__);
+		goto failure;
+	}
+
+	isp->isp_gpio_res[0].name = "isp_gpio_reg";
+	isp->isp_gpio_res[0].flags = IORESOURCE_MEM;
+	isp->isp_gpio_res[0].start = isp_base + ISP410_GPIO_SENSOR_OFFSET;
+	isp->isp_gpio_res[0].end = isp_base + ISP410_GPIO_SENSOR_OFFSET +
+				   ISP410_GPIO_SENSOR_SIZE;
+
+	isp->isp_cell[2].name = "amdisp-pinctrl";
+	isp->isp_cell[2].num_resources = 1;
+	isp->isp_cell[2].resources = &isp->isp_gpio_res[0];
+	isp->isp_cell[2].platform_data = isp->isp_pdata;
+	isp->isp_cell[2].pdata_size = sizeof(struct isp_platform_data);
+
+	r = mfd_add_hotplug_devices(isp->parent, isp->isp_cell, 3);
+	if (r) {
+		drm_err(&adev->ddev,
+			"%s: add mfd hotplug device failed\n", __func__);
+		goto failure;
+	}
+
+	return 0;
+
+failure:
+
+	kfree(isp->isp_pdata);
+	kfree(isp->isp_res);
+	kfree(isp->isp_cell);
+	kfree(isp->isp_i2c_res);
+	kfree(isp->isp_gpio_res);
+
+	return r;
+}
+
+static int isp_v4_1_0_hw_fini(struct amdgpu_isp *isp)
+{
+	mfd_remove_devices(isp->parent);
+
+	kfree(isp->isp_res);
+	kfree(isp->isp_cell);
+	kfree(isp->isp_pdata);
+	kfree(isp->isp_i2c_res);
+	kfree(isp->isp_gpio_res);
+
+	return 0;
+}
+
+static const struct isp_funcs isp_v4_1_0_funcs = {
+	.hw_init = isp_v4_1_0_hw_init,
+	.hw_fini = isp_v4_1_0_hw_fini,
+};
+
+void isp_v4_1_0_set_isp_funcs(struct amdgpu_isp *isp)
+{
+	isp->funcs = &isp_v4_1_0_funcs;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/isp_v4_1_0.h b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_0.h
new file mode 100644
index 000000000000..4d239198edd0
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_0.h
@@ -0,0 +1,50 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright (C) 2024 Advanced Micro Devices, Inc. All rights reserved.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ */
+
+#ifndef __ISP_V4_1_0_H__
+#define __ISP_V4_1_0_H__
+
+#include "amdgpu_isp.h"
+
+#include "ivsrcid/isp/irqsrcs_isp_4_1.h"
+
+#define MAX_ISP410_MEM_RES 2
+#define MAX_ISP410_SENSOR_RES 1
+#define MAX_ISP410_INT_SRC 8
+
+#define ISP410_PHY0_OFFSET 0x66700
+#define ISP410_PHY0_SIZE   0xD30
+
+#define ISP410_I2C0_OFFSET 0x66400
+#define ISP410_I2C0_SIZE 0x100
+
+#define ISP410_GPIO_SENSOR_OFFSET 0x6613C
+#define ISP410_GPIO_SENSOR_SIZE 0x54
+
+void isp_v4_1_0_set_isp_funcs(struct amdgpu_isp *isp);
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.c b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.c
new file mode 100644
index 000000000000..4258d3e0b706
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.c
@@ -0,0 +1,377 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright (C) 2024 Advanced Micro Devices, Inc. All rights reserved.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ */
+
+#include <linux/gpio/machine.h>
+#include "amdgpu.h"
+#include "isp_v4_1_1.h"
+
+MODULE_FIRMWARE("amdgpu/isp_4_1_1.bin");
+
+#define ISP_PERFORMANCE_STATE_LOW 0
+#define ISP_PERFORMANCE_STATE_HIGH 1
+
+#define ISP_HIGH_PERFORMANC_XCLK 788
+#define ISP_HIGH_PERFORMANC_ICLK 788
+
+static const unsigned int isp_4_1_1_int_srcid[MAX_ISP411_INT_SRC] = {
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT9,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT10,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT11,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT12,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT13,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT14,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT15,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT16
+};
+
+static struct gpiod_lookup_table isp_gpio_table = {
+	.dev_id = "amd_isp_capture",
+	.table = {
+		GPIO_LOOKUP("AMDI0030:00", 85, "enable_isp", GPIO_ACTIVE_HIGH),
+		{ }
+	},
+};
+
+static struct gpiod_lookup_table isp_sensor_gpio_table = {
+	.dev_id = "i2c-ov05c10",
+	.table = {
+		GPIO_LOOKUP("amdisp-pinctrl", 0, "enable", GPIO_ACTIVE_HIGH),
+		{ }
+	},
+};
+
+static int isp_poweroff(struct generic_pm_domain *genpd)
+{
+	struct amdgpu_isp *isp = container_of(genpd, struct amdgpu_isp, ispgpd);
+	struct amdgpu_device *adev = isp->adev;
+
+	return amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ISP, true, 0);
+}
+
+static int isp_poweron(struct generic_pm_domain *genpd)
+{
+	struct amdgpu_isp *isp = container_of(genpd, struct amdgpu_isp, ispgpd);
+	struct amdgpu_device *adev = isp->adev;
+
+	return amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ISP, false, 0);
+}
+
+static int isp_set_performance_state(struct generic_pm_domain *genpd,
+				     unsigned int state)
+{
+	struct amdgpu_isp *isp = container_of(genpd, struct amdgpu_isp, ispgpd);
+	struct amdgpu_device *adev = isp->adev;
+	u32 iclk, xclk;
+	int ret;
+
+	switch (state) {
+	case ISP_PERFORMANCE_STATE_HIGH:
+		xclk = ISP_HIGH_PERFORMANC_XCLK;
+		iclk = ISP_HIGH_PERFORMANC_ICLK;
+		break;
+	case ISP_PERFORMANCE_STATE_LOW:
+		/* isp runs at default lowest clock-rate on power-on, do nothing */
+		return 0;
+	default:
+		return -EINVAL;
+	}
+
+	ret = amdgpu_dpm_set_soft_freq_range(adev, PP_ISPXCLK, xclk, 0);
+	if (ret) {
+		drm_err(&adev->ddev, "failed to set xclk %u to %u: %d\n",
+			xclk, state, ret);
+		return ret;
+	}
+
+	ret = amdgpu_dpm_set_soft_freq_range(adev, PP_ISPICLK, iclk, 0);
+	if (ret) {
+		drm_err(&adev->ddev, "failed to set iclk %u to %u: %d\n",
+			iclk, state, ret);
+		return ret;
+	}
+
+	return 0;
+}
+
+static int isp_genpd_add_device(struct device *dev, void *data)
+{
+	struct generic_pm_domain *gpd = data;
+	struct platform_device *pdev = container_of(dev, struct platform_device, dev);
+	struct amdgpu_isp *isp = container_of(gpd, struct amdgpu_isp, ispgpd);
+	struct amdgpu_device *adev = isp->adev;
+	int ret;
+
+	if (!pdev)
+		return -EINVAL;
+
+	if (!dev->type->name) {
+		drm_dbg(&adev->ddev, "Invalid device type to add\n");
+		goto exit;
+	}
+
+	if (strcmp(dev->type->name, "mfd_device")) {
+		drm_dbg(&adev->ddev, "Invalid isp mfd device %s to add\n", pdev->mfd_cell->name);
+		goto exit;
+	}
+
+	ret = pm_genpd_add_device(gpd, dev);
+	if (ret) {
+		drm_err(&adev->ddev, "Failed to add dev %s to genpd %d\n",
+			pdev->mfd_cell->name, ret);
+		return -ENODEV;
+	}
+
+exit:
+	/* Continue to add */
+	return 0;
+}
+
+static int isp_genpd_remove_device(struct device *dev, void *data)
+{
+	struct generic_pm_domain *gpd = data;
+	struct platform_device *pdev = container_of(dev, struct platform_device, dev);
+	struct amdgpu_isp *isp = container_of(gpd, struct amdgpu_isp, ispgpd);
+	struct amdgpu_device *adev = isp->adev;
+	int ret;
+
+	if (!pdev)
+		return -EINVAL;
+
+	if (!dev->type->name) {
+		drm_dbg(&adev->ddev, "Invalid device type to remove\n");
+		goto exit;
+	}
+
+	if (strcmp(dev->type->name, "mfd_device")) {
+		drm_dbg(&adev->ddev, "Invalid isp mfd device %s to remove\n",
+			pdev->mfd_cell->name);
+		goto exit;
+	}
+
+	ret = pm_genpd_remove_device(dev);
+	if (ret) {
+		drm_err(&adev->ddev, "Failed to remove dev from genpd %d\n", ret);
+		return -ENODEV;
+	}
+
+exit:
+	/* Continue to remove */
+	return 0;
+}
+
+static int isp_v4_1_1_hw_init(struct amdgpu_isp *isp)
+{
+	const struct software_node *amd_camera_node, *isp4_node;
+	struct amdgpu_device *adev = isp->adev;
+	struct acpi_device *acpi_dev;
+	int idx, int_idx, num_res, r;
+	u64 isp_base;
+
+	if (adev->rmmio_size == 0 || adev->rmmio_size < 0x5289)
+		return -EINVAL;
+
+	r = amdgpu_acpi_get_isp4_dev(&acpi_dev);
+	if (r) {
+		drm_dbg(&adev->ddev, "Invalid isp platform detected (%d)", r);
+		/* allow GPU init to progress */
+		return 0;
+	}
+
+	/* add GPIO resources required for OMNI5C10 sensor */
+	if (!strcmp("OMNI5C10", acpi_device_hid(acpi_dev))) {
+		gpiod_add_lookup_table(&isp_gpio_table);
+		gpiod_add_lookup_table(&isp_sensor_gpio_table);
+	}
+
+	isp_base = adev->rmmio_base;
+
+	isp->ispgpd.name = "ISP_v_4_1_1";
+	isp->ispgpd.power_off = isp_poweroff;
+	isp->ispgpd.power_on = isp_poweron;
+	isp->ispgpd.set_performance_state = isp_set_performance_state;
+
+	r = pm_genpd_init(&isp->ispgpd, NULL, true);
+	if (r) {
+		drm_err(&adev->ddev, "failed to initialize genpd (%d)\n", r);
+		return -EINVAL;
+	}
+
+	isp->isp_cell = kcalloc(3, sizeof(struct mfd_cell), GFP_KERNEL);
+	if (!isp->isp_cell) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev, "isp mfd cell alloc failed (%d)\n", r);
+		goto failure;
+	}
+
+	num_res = MAX_ISP411_MEM_RES + MAX_ISP411_INT_SRC;
+
+	isp->isp_res = kcalloc(num_res, sizeof(struct resource),
+			       GFP_KERNEL);
+	if (!isp->isp_res) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev, "isp mfd resource alloc failed (%d)\n", r);
+		goto failure;
+	}
+
+	isp->isp_pdata = kzalloc(sizeof(*isp->isp_pdata), GFP_KERNEL);
+	if (!isp->isp_pdata) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev, "isp platform data alloc failed (%d)\n", r);
+		goto failure;
+	}
+
+	amd_camera_node = (const struct software_node *)acpi_dev->driver_data;
+	isp4_node = software_node_find_by_name(amd_camera_node, "isp4");
+
+	/* initialize isp platform data */
+	isp->isp_pdata->adev = (void *)adev;
+	isp->isp_pdata->asic_type = adev->asic_type;
+	isp->isp_pdata->base_rmmio_size = adev->rmmio_size;
+
+	isp->isp_res[0].name = "isp_4_1_1_reg";
+	isp->isp_res[0].flags = IORESOURCE_MEM;
+	isp->isp_res[0].start = isp_base;
+	isp->isp_res[0].end = isp_base + ISP_REGS_OFFSET_END;
+
+	isp->isp_res[1].name = "isp_4_1_1_phy0_reg";
+	isp->isp_res[1].flags = IORESOURCE_MEM;
+	isp->isp_res[1].start = isp_base + ISP411_PHY0_OFFSET;
+	isp->isp_res[1].end = isp_base + ISP411_PHY0_OFFSET + ISP411_PHY0_SIZE;
+
+	for (idx = MAX_ISP411_MEM_RES, int_idx = 0; idx < num_res; idx++, int_idx++) {
+		isp->isp_res[idx].name = "isp_4_1_1_irq";
+		isp->isp_res[idx].flags = IORESOURCE_IRQ;
+		isp->isp_res[idx].start =
+			amdgpu_irq_create_mapping(adev, isp_4_1_1_int_srcid[int_idx]);
+		isp->isp_res[idx].end =
+			isp->isp_res[idx].start;
+	}
+
+	isp->isp_cell[0].name = "amd_isp_capture";
+	isp->isp_cell[0].num_resources = num_res;
+	isp->isp_cell[0].resources = &isp->isp_res[0];
+	isp->isp_cell[0].platform_data = isp->isp_pdata;
+	isp->isp_cell[0].swnode = isp4_node;
+	isp->isp_cell[0].pdata_size = sizeof(struct isp_platform_data);
+
+	/* initialize isp i2c platform data */
+	isp->isp_i2c_res = kcalloc(1, sizeof(struct resource), GFP_KERNEL);
+	if (!isp->isp_i2c_res) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev, "isp mfd res alloc failed (%d)\n", r);
+		goto failure;
+	}
+
+	isp->isp_i2c_res[0].name = "isp_i2c0_reg";
+	isp->isp_i2c_res[0].flags = IORESOURCE_MEM;
+	isp->isp_i2c_res[0].start = isp_base + ISP411_I2C0_OFFSET;
+	isp->isp_i2c_res[0].end = isp_base + ISP411_I2C0_OFFSET + ISP411_I2C0_SIZE;
+
+	isp->isp_cell[1].name = "amd_isp_i2c_designware";
+	isp->isp_cell[1].num_resources = 1;
+	isp->isp_cell[1].resources = &isp->isp_i2c_res[0];
+	isp->isp_cell[1].platform_data = isp->isp_pdata;
+	isp->isp_cell[1].pdata_size = sizeof(struct isp_platform_data);
+
+	/* initialize isp gpiochip platform data */
+	isp->isp_gpio_res = kcalloc(1, sizeof(struct resource), GFP_KERNEL);
+	if (!isp->isp_gpio_res) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev, "isp gpio resource alloc failed (%d)\n", r);
+		goto failure;
+	}
+
+	isp->isp_gpio_res[0].name = "isp_gpio_reg";
+	isp->isp_gpio_res[0].flags = IORESOURCE_MEM;
+	isp->isp_gpio_res[0].start = isp_base + ISP411_GPIO_SENSOR_OFFSET;
+	isp->isp_gpio_res[0].end = isp_base + ISP411_GPIO_SENSOR_OFFSET +
+				   ISP411_GPIO_SENSOR_SIZE;
+
+	isp->isp_cell[2].name = "amdisp-pinctrl";
+	isp->isp_cell[2].num_resources = 1;
+	isp->isp_cell[2].resources = &isp->isp_gpio_res[0];
+	isp->isp_cell[2].platform_data = isp->isp_pdata;
+	isp->isp_cell[2].pdata_size = sizeof(struct isp_platform_data);
+
+	/* add only amd_isp_capture and amd_isp_i2c_designware to genpd */
+	r = mfd_add_hotplug_devices(isp->parent, isp->isp_cell, 2);
+	if (r) {
+		drm_err(&adev->ddev, "add mfd hotplug device failed (%d)\n", r);
+		goto failure;
+	}
+
+	r = device_for_each_child(isp->parent, &isp->ispgpd,
+				  isp_genpd_add_device);
+	if (r) {
+		drm_err(&adev->ddev, "failed to add devices to genpd (%d)\n", r);
+		goto failure;
+	}
+
+	r = mfd_add_hotplug_devices(isp->parent, &isp->isp_cell[2], 1);
+	if (r) {
+		drm_err(&adev->ddev, "add pinctl hotplug device failed (%d)\n", r);
+		goto failure;
+	}
+
+	return 0;
+
+failure:
+
+	kfree(isp->isp_pdata);
+	kfree(isp->isp_res);
+	kfree(isp->isp_cell);
+	kfree(isp->isp_i2c_res);
+	kfree(isp->isp_gpio_res);
+
+	return r;
+}
+
+static int isp_v4_1_1_hw_fini(struct amdgpu_isp *isp)
+{
+	device_for_each_child(isp->parent, NULL,
+			      isp_genpd_remove_device);
+
+	mfd_remove_devices(isp->parent);
+
+	kfree(isp->isp_res);
+	kfree(isp->isp_cell);
+	kfree(isp->isp_pdata);
+	kfree(isp->isp_i2c_res);
+	kfree(isp->isp_gpio_res);
+
+	return 0;
+}
+
+static const struct isp_funcs isp_v4_1_1_funcs = {
+	.hw_init = isp_v4_1_1_hw_init,
+	.hw_fini = isp_v4_1_1_hw_fini,
+};
+
+void isp_v4_1_1_set_isp_funcs(struct amdgpu_isp *isp)
+{
+	isp->funcs = &isp_v4_1_1_funcs;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.h b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.h
new file mode 100644
index 000000000000..fe45d70d87f1
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.h
@@ -0,0 +1,49 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright (C) 2024 Advanced Micro Devices, Inc. All rights reserved.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ */
+
+#ifndef __ISP_V4_1_1_H__
+#define __ISP_V4_1_1_H__
+
+#include "amdgpu_isp.h"
+
+#include "ivsrcid/isp/irqsrcs_isp_4_1.h"
+
+#define MAX_ISP411_MEM_RES 2
+#define MAX_ISP411_INT_SRC 8
+
+#define ISP411_PHY0_OFFSET 0x66700
+#define ISP411_PHY0_SIZE   0xD30
+
+#define ISP411_I2C0_OFFSET 0x66400
+#define ISP411_I2C0_SIZE 0x100
+
+#define ISP411_GPIO_SENSOR_OFFSET 0x6613C
+#define ISP411_GPIO_SENSOR_SIZE 0x54
+
+void isp_v4_1_1_set_isp_funcs(struct amdgpu_isp *isp);
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c
index a3076eb8af6a..b5bb7f4d607c 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c
@@ -23,6 +23,7 @@
 
 #include "amdgpu.h"
 #include "amdgpu_jpeg.h"
+#include "amdgpu_cs.h"
 #include "soc15.h"
 #include "soc15d.h"
 #include "vcn_v1_0.h"
@@ -34,6 +35,9 @@
 static void jpeg_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev);
 static void jpeg_v1_0_set_irq_funcs(struct amdgpu_device *adev);
 static void jpeg_v1_0_ring_begin_use(struct amdgpu_ring *ring);
+static int jpeg_v1_dec_ring_parse_cs(struct amdgpu_cs_parser *parser,
+				     struct amdgpu_job *job,
+				     struct amdgpu_ib *ib);
 
 static void jpeg_v1_0_decode_ring_patch_wreg(struct amdgpu_ring *ring, uint32_t *ptr, uint32_t reg_offset, uint32_t val)
 {
@@ -300,7 +304,10 @@ static void jpeg_v1_0_decode_ring_emit_ib(struct amdgpu_ring *ring,
 
 	amdgpu_ring_write(ring,
 		PACKETJ(SOC15_REG_OFFSET(JPEG, 0, mmUVD_LMI_JRBC_IB_VMID), 0, 0, PACKETJ_TYPE0));
-	amdgpu_ring_write(ring, (vmid | (vmid << 4)));
+	if (ring->funcs->parse_cs)
+		amdgpu_ring_write(ring, 0);
+	else
+		amdgpu_ring_write(ring, (vmid | (vmid << 4)));
 
 	amdgpu_ring_write(ring,
 		PACKETJ(SOC15_REG_OFFSET(JPEG, 0, mmUVD_LMI_JPEG_VMID), 0, 0, PACKETJ_TYPE0));
@@ -437,7 +444,7 @@ static int jpeg_v1_0_process_interrupt(struct amdgpu_device *adev,
 
 	switch (entry->src_id) {
 	case 126:
-		amdgpu_fence_process(&adev->jpeg.inst->ring_dec);
+		amdgpu_fence_process(adev->jpeg.inst->ring_dec);
 		break;
 	default:
 		DRM_ERROR("Unhandled interrupt: %d %d\n",
@@ -451,15 +458,16 @@ static int jpeg_v1_0_process_interrupt(struct amdgpu_device *adev,
 /**
  * jpeg_v1_0_early_init - set function pointers
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Set ring and irq function pointers
  */
-int jpeg_v1_0_early_init(void *handle)
+int jpeg_v1_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->jpeg.num_jpeg_inst = 1;
+	adev->jpeg.num_jpeg_rings = 1;
 
 	jpeg_v1_0_set_dec_ring_funcs(adev);
 	jpeg_v1_0_set_irq_funcs(adev);
@@ -470,12 +478,12 @@ int jpeg_v1_0_early_init(void *handle)
 /**
  * jpeg_v1_0_sw_init - sw init for JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  */
-int jpeg_v1_0_sw_init(void *handle)
+int jpeg_v1_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring;
 	int r;
 
@@ -484,15 +492,15 @@ int jpeg_v1_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	ring = &adev->jpeg.inst->ring_dec;
-	ring->vm_hub = AMDGPU_MMHUB_0;
+	ring = adev->jpeg.inst->ring_dec;
+	ring->vm_hub = AMDGPU_MMHUB0(0);
 	sprintf(ring->name, "jpeg_dec");
 	r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq,
 			     0, AMDGPU_RING_PRIO_DEFAULT, NULL);
 	if (r)
 		return r;
 
-	adev->jpeg.internal.jpeg_pitch = adev->jpeg.inst->external.jpeg_pitch =
+	adev->jpeg.internal.jpeg_pitch[0] = adev->jpeg.inst->external.jpeg_pitch[0] =
 		SOC15_REG_OFFSET(JPEG, 0, mmUVD_JPEG_PITCH);
 
 	return 0;
@@ -501,15 +509,15 @@ int jpeg_v1_0_sw_init(void *handle)
 /**
  * jpeg_v1_0_sw_fini - sw fini for JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * JPEG free up sw allocation
  */
-void jpeg_v1_0_sw_fini(void *handle)
+void jpeg_v1_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	amdgpu_ring_fini(&adev->jpeg.inst[0].ring_dec);
+	amdgpu_ring_fini(adev->jpeg.inst->ring_dec);
 }
 
 /**
@@ -522,7 +530,7 @@ void jpeg_v1_0_sw_fini(void *handle)
  */
 void jpeg_v1_0_start(struct amdgpu_device *adev, int mode)
 {
-	struct amdgpu_ring *ring = &adev->jpeg.inst->ring_dec;
+	struct amdgpu_ring *ring = adev->jpeg.inst->ring_dec;
 
 	if (mode == 0) {
 		WREG32_SOC15(JPEG, 0, mmUVD_LMI_JRBC_RB_VMID, 0);
@@ -549,10 +557,11 @@ static const struct amdgpu_ring_funcs jpeg_v1_0_decode_ring_vm_funcs = {
 	.nop = PACKET0(0x81ff, 0),
 	.support_64bit_ptrs = false,
 	.no_user_fence = true,
-	.extra_dw = 64,
+	.extra_bytes = 256,
 	.get_rptr = jpeg_v1_0_decode_ring_get_rptr,
 	.get_wptr = jpeg_v1_0_decode_ring_get_wptr,
 	.set_wptr = jpeg_v1_0_decode_ring_set_wptr,
+	.parse_cs = jpeg_v1_dec_ring_parse_cs,
 	.emit_frame_size =
 		6 + 6 + /* hdp invalidate / flush */
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 +
@@ -579,8 +588,7 @@ static const struct amdgpu_ring_funcs jpeg_v1_0_decode_ring_vm_funcs = {
 
 static void jpeg_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev)
 {
-	adev->jpeg.inst->ring_dec.funcs = &jpeg_v1_0_decode_ring_vm_funcs;
-	DRM_INFO("JPEG decode is enabled in VM mode\n");
+	adev->jpeg.inst->ring_dec->funcs = &jpeg_v1_0_decode_ring_vm_funcs;
 }
 
 static const struct amdgpu_irq_src_funcs jpeg_v1_0_irq_funcs = {
@@ -596,18 +604,84 @@ static void jpeg_v1_0_set_irq_funcs(struct amdgpu_device *adev)
 static void jpeg_v1_0_ring_begin_use(struct amdgpu_ring *ring)
 {
 	struct	amdgpu_device *adev = ring->adev;
-	bool	set_clocks = !cancel_delayed_work_sync(&adev->vcn.idle_work);
+	bool	set_clocks = !cancel_delayed_work_sync(&adev->vcn.inst[0].idle_work);
 	int		cnt = 0;
 
-	mutex_lock(&adev->vcn.vcn1_jpeg1_workaround);
+	mutex_lock(&adev->vcn.inst[0].vcn1_jpeg1_workaround);
 
 	if (amdgpu_fence_wait_empty(&adev->vcn.inst->ring_dec))
 		DRM_ERROR("JPEG dec: vcn dec ring may not be empty\n");
 
-	for (cnt = 0; cnt < adev->vcn.num_enc_rings; cnt++) {
+	for (cnt = 0; cnt < adev->vcn.inst[0].num_enc_rings; cnt++) {
 		if (amdgpu_fence_wait_empty(&adev->vcn.inst->ring_enc[cnt]))
 			DRM_ERROR("JPEG dec: vcn enc ring[%d] may not be empty\n", cnt);
 	}
 
 	vcn_v1_0_set_pg_for_begin_use(ring, set_clocks);
 }
+
+/**
+ * jpeg_v1_dec_ring_parse_cs - command submission parser
+ *
+ * @parser: Command submission parser context
+ * @job: the job to parse
+ * @ib: the IB to parse
+ *
+ * Parse the command stream, return -EINVAL for invalid packet,
+ * 0 otherwise
+ */
+static int jpeg_v1_dec_ring_parse_cs(struct amdgpu_cs_parser *parser,
+				     struct amdgpu_job *job,
+				     struct amdgpu_ib *ib)
+{
+	u32 i, reg, res, cond, type;
+	int ret = 0;
+	struct amdgpu_device *adev = parser->adev;
+
+	for (i = 0; i < ib->length_dw ; i += 2) {
+		reg  = CP_PACKETJ_GET_REG(ib->ptr[i]);
+		res  = CP_PACKETJ_GET_RES(ib->ptr[i]);
+		cond = CP_PACKETJ_GET_COND(ib->ptr[i]);
+		type = CP_PACKETJ_GET_TYPE(ib->ptr[i]);
+
+		if (res || cond != PACKETJ_CONDITION_CHECK0) /* only allow 0 for now */
+			return -EINVAL;
+
+		if (reg >= JPEG_V1_REG_RANGE_START && reg <= JPEG_V1_REG_RANGE_END)
+			continue;
+
+		switch (type) {
+		case PACKETJ_TYPE0:
+			if (reg != JPEG_V1_LMI_JPEG_WRITE_64BIT_BAR_HIGH &&
+			    reg != JPEG_V1_LMI_JPEG_WRITE_64BIT_BAR_LOW &&
+			    reg != JPEG_V1_LMI_JPEG_READ_64BIT_BAR_HIGH &&
+			    reg != JPEG_V1_LMI_JPEG_READ_64BIT_BAR_LOW &&
+			    reg != JPEG_V1_REG_CTX_INDEX &&
+			    reg != JPEG_V1_REG_CTX_DATA) {
+				ret = -EINVAL;
+			}
+			break;
+		case PACKETJ_TYPE1:
+			if (reg != JPEG_V1_REG_CTX_DATA)
+				ret = -EINVAL;
+			break;
+		case PACKETJ_TYPE3:
+			if (reg != JPEG_V1_REG_SOFT_RESET)
+				ret = -EINVAL;
+			break;
+		case PACKETJ_TYPE6:
+			if (ib->ptr[i] != CP_PACKETJ_NOP)
+				ret = -EINVAL;
+			break;
+		default:
+			ret = -EINVAL;
+		}
+
+		if (ret) {
+			dev_err(adev->dev, "Invalid packet [0x%08x]!\n", ib->ptr[i]);
+			break;
+		}
+	}
+
+	return ret;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.h b/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.h
index bbf33a6a3972..097328635083 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.h
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.h
@@ -24,9 +24,20 @@
 #ifndef __JPEG_V1_0_H__
 #define __JPEG_V1_0_H__
 
-int jpeg_v1_0_early_init(void *handle);
-int jpeg_v1_0_sw_init(void *handle);
-void jpeg_v1_0_sw_fini(void *handle);
+int jpeg_v1_0_early_init(struct amdgpu_ip_block *ip_block);
+int jpeg_v1_0_sw_init(struct amdgpu_ip_block *ip_block);
+void jpeg_v1_0_sw_fini(struct amdgpu_ip_block *ip_block);
 void jpeg_v1_0_start(struct amdgpu_device *adev, int mode);
 
+#define JPEG_V1_REG_RANGE_START	0x8000
+#define JPEG_V1_REG_RANGE_END	0x803f
+
+#define JPEG_V1_LMI_JPEG_WRITE_64BIT_BAR_HIGH	0x8238
+#define JPEG_V1_LMI_JPEG_WRITE_64BIT_BAR_LOW	0x8239
+#define JPEG_V1_LMI_JPEG_READ_64BIT_BAR_HIGH	0x825a
+#define JPEG_V1_LMI_JPEG_READ_64BIT_BAR_LOW	0x825b
+#define JPEG_V1_REG_CTX_INDEX			0x8328
+#define JPEG_V1_REG_CTX_DATA			0x8329
+#define JPEG_V1_REG_SOFT_RESET			0x83a0
+
 #endif /*__JPEG_V1_0_H__*/
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c
index 0eddf7c824a7..27c76bd424cf 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c
@@ -32,23 +32,40 @@
 #include "vcn/vcn_2_0_0_sh_mask.h"
 #include "ivsrcid/vcn/irqsrcs_vcn_2_0.h"
 
+static const struct amdgpu_hwip_reg_entry jpeg_reg_list_2_0[] = {
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JPEG_POWER_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JPEG_INT_STAT),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JRBC_RB_CNTL),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JRBC_RB_SIZE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmJPEG_DEC_ADDR_MODE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmJPEG_DEC_GFX10_ADDR_CONFIG),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmJPEG_DEC_Y_GFX10_TILING_SURFACE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmJPEG_DEC_UV_GFX10_TILING_SURFACE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JPEG_PITCH),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JPEG_UV_PITCH),
+};
+
 static void jpeg_v2_0_set_dec_ring_funcs(struct amdgpu_device *adev);
 static void jpeg_v2_0_set_irq_funcs(struct amdgpu_device *adev);
-static int jpeg_v2_0_set_powergating_state(void *handle,
+static int jpeg_v2_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 				enum amd_powergating_state state);
 
 /**
  * jpeg_v2_0_early_init - set function pointers
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Set ring and irq function pointers
  */
-static int jpeg_v2_0_early_init(void *handle)
+static int jpeg_v2_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->jpeg.num_jpeg_inst = 1;
+	adev->jpeg.num_jpeg_rings = 1;
 
 	jpeg_v2_0_set_dec_ring_funcs(adev);
 	jpeg_v2_0_set_irq_funcs(adev);
@@ -59,13 +76,13 @@ static int jpeg_v2_0_early_init(void *handle)
 /**
  * jpeg_v2_0_sw_init - sw init for JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Load firmware and sw initialization
  */
-static int jpeg_v2_0_sw_init(void *handle)
+static int jpeg_v2_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring;
 	int r;
 
@@ -83,38 +100,50 @@ static int jpeg_v2_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	ring = &adev->jpeg.inst->ring_dec;
+	ring = adev->jpeg.inst->ring_dec;
 	ring->use_doorbell = true;
 	ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1;
-	ring->vm_hub = AMDGPU_MMHUB_0;
+	ring->vm_hub = AMDGPU_MMHUB0(0);
 	sprintf(ring->name, "jpeg_dec");
 	r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq,
 			     0, AMDGPU_RING_PRIO_DEFAULT, NULL);
 	if (r)
 		return r;
 
-	adev->jpeg.internal.jpeg_pitch = mmUVD_JPEG_PITCH_INTERNAL_OFFSET;
-	adev->jpeg.inst->external.jpeg_pitch = SOC15_REG_OFFSET(JPEG, 0, mmUVD_JPEG_PITCH);
+	adev->jpeg.internal.jpeg_pitch[0] = mmUVD_JPEG_PITCH_INTERNAL_OFFSET;
+	adev->jpeg.inst->external.jpeg_pitch[0] = SOC15_REG_OFFSET(JPEG, 0, mmUVD_JPEG_PITCH);
 
-	return 0;
+	r = amdgpu_jpeg_reg_dump_init(adev, jpeg_reg_list_2_0, ARRAY_SIZE(jpeg_reg_list_2_0));
+	if (r)
+		return r;
+
+	adev->jpeg.supported_reset =
+		amdgpu_get_soft_full_reset_mask(adev->jpeg.inst[0].ring_dec);
+	if (!amdgpu_sriov_vf(adev))
+		adev->jpeg.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+	r = amdgpu_jpeg_sysfs_reset_mask_init(adev);
+
+	return r;
 }
 
 /**
  * jpeg_v2_0_sw_fini - sw fini for JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * JPEG suspend and free up sw allocation
  */
-static int jpeg_v2_0_sw_fini(void *handle)
+static int jpeg_v2_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = amdgpu_jpeg_suspend(adev);
 	if (r)
 		return r;
 
+	amdgpu_jpeg_sysfs_reset_mask_fini(adev);
+
 	r = amdgpu_jpeg_sw_fini(adev);
 
 	return r;
@@ -123,41 +152,36 @@ static int jpeg_v2_0_sw_fini(void *handle)
 /**
  * jpeg_v2_0_hw_init - start and test JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  */
-static int jpeg_v2_0_hw_init(void *handle)
+static int jpeg_v2_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	struct amdgpu_ring *ring = &adev->jpeg.inst->ring_dec;
-	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring = adev->jpeg.inst->ring_dec;
 
 	adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell,
 		(adev->doorbell_index.vcn.vcn_ring0_1 << 1), 0);
 
-	r = amdgpu_ring_test_helper(ring);
-	if (!r)
-		DRM_INFO("JPEG decode initialized successfully.\n");
-
-	return r;
+	return amdgpu_ring_test_helper(ring);
 }
 
 /**
  * jpeg_v2_0_hw_fini - stop the hardware block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Stop the JPEG block, mark ring as not ready any more
  */
-static int jpeg_v2_0_hw_fini(void *handle)
+static int jpeg_v2_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	cancel_delayed_work_sync(&adev->vcn.idle_work);
+	cancel_delayed_work_sync(&adev->jpeg.idle_work);
 
 	if (adev->jpeg.cur_state != AMD_PG_STATE_GATE &&
 	      RREG32_SOC15(JPEG, 0, mmUVD_JRBC_STATUS))
-		jpeg_v2_0_set_powergating_state(adev, AMD_PG_STATE_GATE);
+		jpeg_v2_0_set_powergating_state(ip_block, AMD_PG_STATE_GATE);
 
 	return 0;
 }
@@ -165,20 +189,19 @@ static int jpeg_v2_0_hw_fini(void *handle)
 /**
  * jpeg_v2_0_suspend - suspend JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * HW fini and suspend JPEG block
  */
-static int jpeg_v2_0_suspend(void *handle)
+static int jpeg_v2_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	int r;
 
-	r = jpeg_v2_0_hw_fini(adev);
+	r = jpeg_v2_0_hw_fini(ip_block);
 	if (r)
 		return r;
 
-	r = amdgpu_jpeg_suspend(adev);
+	r = amdgpu_jpeg_suspend(ip_block->adev);
 
 	return r;
 }
@@ -186,20 +209,19 @@ static int jpeg_v2_0_suspend(void *handle)
 /**
  * jpeg_v2_0_resume - resume JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Resume firmware and hw init JPEG block
  */
-static int jpeg_v2_0_resume(void *handle)
+static int jpeg_v2_0_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = amdgpu_jpeg_resume(adev);
+	r = amdgpu_jpeg_resume(ip_block->adev);
 	if (r)
 		return r;
 
-	r = jpeg_v2_0_hw_init(adev);
+	r = jpeg_v2_0_hw_init(ip_block);
 
 	return r;
 }
@@ -312,7 +334,7 @@ static void jpeg_v2_0_enable_clock_gating(struct amdgpu_device *adev)
  */
 static int jpeg_v2_0_start(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring = &adev->jpeg.inst->ring_dec;
+	struct amdgpu_ring *ring = adev->jpeg.inst->ring_dec;
 	int r;
 
 	if (adev->pm.dpm_enabled)
@@ -542,11 +564,15 @@ void jpeg_v2_0_dec_ring_emit_ib(struct amdgpu_ring *ring,
 
 	amdgpu_ring_write(ring, PACKETJ(mmUVD_LMI_JRBC_IB_VMID_INTERNAL_OFFSET,
 		0, 0, PACKETJ_TYPE0));
-	amdgpu_ring_write(ring, (vmid | (vmid << 4)));
+
+	if (ring->funcs->parse_cs)
+		amdgpu_ring_write(ring, 0);
+	else
+		amdgpu_ring_write(ring, (vmid | (vmid << 4) | (vmid << 8)));
 
 	amdgpu_ring_write(ring, PACKETJ(mmUVD_LMI_JPEG_VMID_INTERNAL_OFFSET,
 		0, 0, PACKETJ_TYPE0));
-	amdgpu_ring_write(ring, (vmid | (vmid << 4)));
+	amdgpu_ring_write(ring, (vmid | (vmid << 4) | (vmid << 8)));
 
 	amdgpu_ring_write(ring,	PACKETJ(mmUVD_LMI_JRBC_IB_64BIT_BAR_LOW_INTERNAL_OFFSET,
 		0, 0, PACKETJ_TYPE0));
@@ -656,18 +682,18 @@ void jpeg_v2_0_dec_ring_nop(struct amdgpu_ring *ring, uint32_t count)
 	}
 }
 
-static bool jpeg_v2_0_is_idle(void *handle)
+static bool jpeg_v2_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return ((RREG32_SOC15(JPEG, 0, mmUVD_JRBC_STATUS) &
 		UVD_JRBC_STATUS__RB_JOB_DONE_MASK) ==
 		UVD_JRBC_STATUS__RB_JOB_DONE_MASK);
 }
 
-static int jpeg_v2_0_wait_for_idle(void *handle)
+static int jpeg_v2_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	ret = SOC15_WAIT_ON_RREG(JPEG, 0, mmUVD_JRBC_STATUS, UVD_JRBC_STATUS__RB_JOB_DONE_MASK,
@@ -676,14 +702,14 @@ static int jpeg_v2_0_wait_for_idle(void *handle)
 	return ret;
 }
 
-static int jpeg_v2_0_set_clockgating_state(void *handle,
+static int jpeg_v2_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_CG_STATE_GATE);
 
 	if (enable) {
-		if (!jpeg_v2_0_is_idle(handle))
+		if (!jpeg_v2_0_is_idle(ip_block))
 			return -EBUSY;
 		jpeg_v2_0_enable_clock_gating(adev);
 	} else {
@@ -693,10 +719,10 @@ static int jpeg_v2_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int jpeg_v2_0_set_powergating_state(void *handle,
+static int jpeg_v2_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	if (state == adev->jpeg.cur_state)
@@ -729,7 +755,7 @@ static int jpeg_v2_0_process_interrupt(struct amdgpu_device *adev,
 
 	switch (entry->src_id) {
 	case VCN_2_0__SRCID__JPEG_DECODE:
-		amdgpu_fence_process(&adev->jpeg.inst->ring_dec);
+		amdgpu_fence_process(adev->jpeg.inst->ring_dec);
 		break;
 	default:
 		DRM_ERROR("Unhandled interrupt: %d %d\n",
@@ -740,10 +766,25 @@ static int jpeg_v2_0_process_interrupt(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int jpeg_v2_0_ring_reset(struct amdgpu_ring *ring,
+				unsigned int vmid,
+				struct amdgpu_fence *timedout_fence)
+{
+	int r;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+	r = jpeg_v2_0_stop(ring->adev);
+	if (r)
+		return r;
+	r = jpeg_v2_0_start(ring->adev);
+	if (r)
+		return r;
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
 static const struct amd_ip_funcs jpeg_v2_0_ip_funcs = {
 	.name = "jpeg_v2_0",
 	.early_init = jpeg_v2_0_early_init,
-	.late_init = NULL,
 	.sw_init = jpeg_v2_0_sw_init,
 	.sw_fini = jpeg_v2_0_sw_fini,
 	.hw_init = jpeg_v2_0_hw_init,
@@ -752,12 +793,10 @@ static const struct amd_ip_funcs jpeg_v2_0_ip_funcs = {
 	.resume = jpeg_v2_0_resume,
 	.is_idle = jpeg_v2_0_is_idle,
 	.wait_for_idle = jpeg_v2_0_wait_for_idle,
-	.check_soft_reset = NULL,
-	.pre_soft_reset = NULL,
-	.soft_reset = NULL,
-	.post_soft_reset = NULL,
 	.set_clockgating_state = jpeg_v2_0_set_clockgating_state,
 	.set_powergating_state = jpeg_v2_0_set_powergating_state,
+	.dump_ip_state = amdgpu_jpeg_dump_ip_state,
+	.print_ip_state = amdgpu_jpeg_print_ip_state,
 };
 
 static const struct amdgpu_ring_funcs jpeg_v2_0_dec_ring_vm_funcs = {
@@ -766,6 +805,7 @@ static const struct amdgpu_ring_funcs jpeg_v2_0_dec_ring_vm_funcs = {
 	.get_rptr = jpeg_v2_0_dec_ring_get_rptr,
 	.get_wptr = jpeg_v2_0_dec_ring_get_wptr,
 	.set_wptr = jpeg_v2_0_dec_ring_set_wptr,
+	.parse_cs = amdgpu_jpeg_dec_parse_cs,
 	.emit_frame_size =
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 +
@@ -787,12 +827,12 @@ static const struct amdgpu_ring_funcs jpeg_v2_0_dec_ring_vm_funcs = {
 	.emit_wreg = jpeg_v2_0_dec_ring_emit_wreg,
 	.emit_reg_wait = jpeg_v2_0_dec_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = jpeg_v2_0_ring_reset,
 };
 
 static void jpeg_v2_0_set_dec_ring_funcs(struct amdgpu_device *adev)
 {
-	adev->jpeg.inst->ring_dec.funcs = &jpeg_v2_0_dec_ring_vm_funcs;
-	DRM_INFO("JPEG decode is enabled in VM mode\n");
+	adev->jpeg.inst->ring_dec->funcs = &jpeg_v2_0_dec_ring_vm_funcs;
 }
 
 static const struct amdgpu_irq_src_funcs jpeg_v2_0_irq_funcs = {
@@ -806,8 +846,7 @@ static void jpeg_v2_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->jpeg.inst->irq.funcs = &jpeg_v2_0_irq_funcs;
 }
 
-const struct amdgpu_ip_block_version jpeg_v2_0_ip_block =
-{
+const struct amdgpu_ip_block_version jpeg_v2_0_ip_block = {
 		.type = AMD_IP_BLOCK_TYPE_JPEG,
 		.major = 2,
 		.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c
index b040f51d9aa9..20983f126b49 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c
@@ -36,9 +36,25 @@
 
 #define JPEG25_MAX_HW_INSTANCES_ARCTURUS			2
 
+static const struct amdgpu_hwip_reg_entry jpeg_reg_list_2_5[] = {
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JPEG_POWER_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JPEG_INT_STAT),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JRBC_RB_CNTL),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JRBC_RB_SIZE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmJPEG_DEC_ADDR_MODE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmJPEG_DEC_GFX10_ADDR_CONFIG),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmJPEG_DEC_Y_GFX10_TILING_SURFACE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmJPEG_DEC_UV_GFX10_TILING_SURFACE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JPEG_PITCH),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JPEG_UV_PITCH),
+};
+
 static void jpeg_v2_5_set_dec_ring_funcs(struct amdgpu_device *adev);
 static void jpeg_v2_5_set_irq_funcs(struct amdgpu_device *adev);
-static int jpeg_v2_5_set_powergating_state(void *handle,
+static int jpeg_v2_5_set_powergating_state(struct amdgpu_ip_block *ip_block,
 				enum amd_powergating_state state);
 static void jpeg_v2_5_set_ras_funcs(struct amdgpu_device *adev);
 
@@ -50,16 +66,17 @@ static int amdgpu_ih_clientid_jpeg[] = {
 /**
  * jpeg_v2_5_early_init - set function pointers
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Set ring and irq function pointers
  */
-static int jpeg_v2_5_early_init(void *handle)
+static int jpeg_v2_5_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 harvest;
 	int i;
 
+	adev->jpeg.num_jpeg_rings = 1;
 	adev->jpeg.num_jpeg_inst = JPEG25_MAX_HW_INSTANCES_ARCTURUS;
 	for (i = 0; i < adev->jpeg.num_jpeg_inst; i++) {
 		harvest = RREG32_SOC15(JPEG, i, mmCC_UVD_HARVESTING);
@@ -80,15 +97,15 @@ static int jpeg_v2_5_early_init(void *handle)
 /**
  * jpeg_v2_5_sw_init - sw init for JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Load firmware and sw initialization
  */
-static int jpeg_v2_5_sw_init(void *handle)
+static int jpeg_v2_5_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
 	int i, r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
 		if (adev->jpeg.harvest_config & (1 << i))
@@ -102,13 +119,13 @@ static int jpeg_v2_5_sw_init(void *handle)
 
 		/* JPEG DJPEG POISON EVENT */
 		r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_jpeg[i],
-			VCN_2_6__SRCID_DJPEG0_POISON, &adev->jpeg.inst[i].irq);
+			VCN_2_6__SRCID_DJPEG0_POISON, &adev->jpeg.inst[i].ras_poison_irq);
 		if (r)
 			return r;
 
 		/* JPEG EJPEG POISON EVENT */
 		r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_jpeg[i],
-			VCN_2_6__SRCID_EJPEG0_POISON, &adev->jpeg.inst[i].irq);
+			VCN_2_6__SRCID_EJPEG0_POISON, &adev->jpeg.inst[i].ras_poison_irq);
 		if (r)
 			return r;
 	}
@@ -125,12 +142,12 @@ static int jpeg_v2_5_sw_init(void *handle)
 		if (adev->jpeg.harvest_config & (1 << i))
 			continue;
 
-		ring = &adev->jpeg.inst[i].ring_dec;
+		ring = adev->jpeg.inst[i].ring_dec;
 		ring->use_doorbell = true;
-		if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(2, 5, 0))
-			ring->vm_hub = AMDGPU_MMHUB_1;
+		if (amdgpu_ip_version(adev, UVD_HWIP, 0) == IP_VERSION(2, 5, 0))
+			ring->vm_hub = AMDGPU_MMHUB1(0);
 		else
-			ring->vm_hub = AMDGPU_MMHUB_0;
+			ring->vm_hub = AMDGPU_MMHUB0(0);
 		ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1 + 8 * i;
 		sprintf(ring->name, "jpeg_dec_%d", i);
 		r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst[i].irq,
@@ -138,33 +155,45 @@ static int jpeg_v2_5_sw_init(void *handle)
 		if (r)
 			return r;
 
-		adev->jpeg.internal.jpeg_pitch = mmUVD_JPEG_PITCH_INTERNAL_OFFSET;
-		adev->jpeg.inst[i].external.jpeg_pitch = SOC15_REG_OFFSET(JPEG, i, mmUVD_JPEG_PITCH);
+		adev->jpeg.internal.jpeg_pitch[0] = mmUVD_JPEG_PITCH_INTERNAL_OFFSET;
+		adev->jpeg.inst[i].external.jpeg_pitch[0] = SOC15_REG_OFFSET(JPEG, i, mmUVD_JPEG_PITCH);
 	}
 
 	r = amdgpu_jpeg_ras_sw_init(adev);
 	if (r)
 		return r;
 
-	return 0;
+	r = amdgpu_jpeg_reg_dump_init(adev, jpeg_reg_list_2_5, ARRAY_SIZE(jpeg_reg_list_2_5));
+	if (r)
+		return r;
+
+	adev->jpeg.supported_reset =
+		amdgpu_get_soft_full_reset_mask(adev->jpeg.inst[0].ring_dec);
+	if (!amdgpu_sriov_vf(adev))
+		adev->jpeg.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+	r = amdgpu_jpeg_sysfs_reset_mask_init(adev);
+
+	return r;
 }
 
 /**
  * jpeg_v2_5_sw_fini - sw fini for JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * JPEG suspend and free up sw allocation
  */
-static int jpeg_v2_5_sw_fini(void *handle)
+static int jpeg_v2_5_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = amdgpu_jpeg_suspend(adev);
 	if (r)
 		return r;
 
+	amdgpu_jpeg_sysfs_reset_mask_fini(adev);
+
 	r = amdgpu_jpeg_sw_fini(adev);
 
 	return r;
@@ -173,12 +202,12 @@ static int jpeg_v2_5_sw_fini(void *handle)
 /**
  * jpeg_v2_5_hw_init - start and test JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  */
-static int jpeg_v2_5_hw_init(void *handle)
+static int jpeg_v2_5_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring;
 	int i, r;
 
@@ -186,7 +215,7 @@ static int jpeg_v2_5_hw_init(void *handle)
 		if (adev->jpeg.harvest_config & (1 << i))
 			continue;
 
-		ring = &adev->jpeg.inst[i].ring_dec;
+		ring = adev->jpeg.inst[i].ring_dec;
 		adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell,
 			(adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 8 * i, i);
 
@@ -195,24 +224,22 @@ static int jpeg_v2_5_hw_init(void *handle)
 			return r;
 	}
 
-	DRM_INFO("JPEG decode initialized successfully.\n");
-
 	return 0;
 }
 
 /**
  * jpeg_v2_5_hw_fini - stop the hardware block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Stop the JPEG block, mark ring as not ready any more
  */
-static int jpeg_v2_5_hw_fini(void *handle)
+static int jpeg_v2_5_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
-	cancel_delayed_work_sync(&adev->vcn.idle_work);
+	cancel_delayed_work_sync(&adev->jpeg.idle_work);
 
 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
 		if (adev->jpeg.harvest_config & (1 << i))
@@ -220,7 +247,10 @@ static int jpeg_v2_5_hw_fini(void *handle)
 
 		if (adev->jpeg.cur_state != AMD_PG_STATE_GATE &&
 		      RREG32_SOC15(JPEG, i, mmUVD_JRBC_STATUS))
-			jpeg_v2_5_set_powergating_state(adev, AMD_PG_STATE_GATE);
+			jpeg_v2_5_set_powergating_state(ip_block, AMD_PG_STATE_GATE);
+
+		if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__JPEG))
+			amdgpu_irq_put(adev, &adev->jpeg.inst[i].ras_poison_irq, 0);
 	}
 
 	return 0;
@@ -229,20 +259,19 @@ static int jpeg_v2_5_hw_fini(void *handle)
 /**
  * jpeg_v2_5_suspend - suspend JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * HW fini and suspend JPEG block
  */
-static int jpeg_v2_5_suspend(void *handle)
+static int jpeg_v2_5_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	int r;
 
-	r = jpeg_v2_5_hw_fini(adev);
+	r = jpeg_v2_5_hw_fini(ip_block);
 	if (r)
 		return r;
 
-	r = amdgpu_jpeg_suspend(adev);
+	r = amdgpu_jpeg_suspend(ip_block->adev);
 
 	return r;
 }
@@ -250,20 +279,19 @@ static int jpeg_v2_5_suspend(void *handle)
 /**
  * jpeg_v2_5_resume - resume JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Resume firmware and hw init JPEG block
  */
-static int jpeg_v2_5_resume(void *handle)
+static int jpeg_v2_5_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	int r;
 
-	r = amdgpu_jpeg_resume(adev);
+	r = amdgpu_jpeg_resume(ip_block->adev);
 	if (r)
 		return r;
 
-	r = jpeg_v2_5_hw_init(adev);
+	r = jpeg_v2_5_hw_init(ip_block);
 
 	return r;
 }
@@ -310,6 +338,44 @@ static void jpeg_v2_5_enable_clock_gating(struct amdgpu_device *adev, int inst)
 	WREG32_SOC15(JPEG, inst, mmJPEG_CGC_GATE, data);
 }
 
+static void jpeg_v2_5_start_inst(struct amdgpu_device *adev, int i)
+{
+	struct amdgpu_ring *ring = adev->jpeg.inst[i].ring_dec;
+	/* disable anti hang mechanism */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, i, mmUVD_JPEG_POWER_STATUS), 0,
+		~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK);
+
+	/* JPEG disable CGC */
+	jpeg_v2_5_disable_clock_gating(adev, i);
+
+	/* MJPEG global tiling registers */
+	WREG32_SOC15(JPEG, i, mmJPEG_DEC_GFX8_ADDR_CONFIG,
+		adev->gfx.config.gb_addr_config);
+	WREG32_SOC15(JPEG, i, mmJPEG_DEC_GFX10_ADDR_CONFIG,
+		adev->gfx.config.gb_addr_config);
+
+	/* enable JMI channel */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, i, mmUVD_JMI_CNTL), 0,
+		~UVD_JMI_CNTL__SOFT_RESET_MASK);
+
+	/* enable System Interrupt for JRBC */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, i, mmJPEG_SYS_INT_EN),
+		JPEG_SYS_INT_EN__DJRBC_MASK,
+		~JPEG_SYS_INT_EN__DJRBC_MASK);
+
+	WREG32_SOC15(JPEG, i, mmUVD_LMI_JRBC_RB_VMID, 0);
+	WREG32_SOC15(JPEG, i, mmUVD_JRBC_RB_CNTL, (0x00000001L | 0x00000002L));
+	WREG32_SOC15(JPEG, i, mmUVD_LMI_JRBC_RB_64BIT_BAR_LOW,
+		lower_32_bits(ring->gpu_addr));
+	WREG32_SOC15(JPEG, i, mmUVD_LMI_JRBC_RB_64BIT_BAR_HIGH,
+		upper_32_bits(ring->gpu_addr));
+	WREG32_SOC15(JPEG, i, mmUVD_JRBC_RB_RPTR, 0);
+	WREG32_SOC15(JPEG, i, mmUVD_JRBC_RB_WPTR, 0);
+	WREG32_SOC15(JPEG, i, mmUVD_JRBC_RB_CNTL, 0x00000002L);
+	WREG32_SOC15(JPEG, i, mmUVD_JRBC_RB_SIZE, ring->ring_size / 4);
+	ring->wptr = RREG32_SOC15(JPEG, i, mmUVD_JRBC_RB_WPTR);
+}
+
 /**
  * jpeg_v2_5_start - start JPEG block
  *
@@ -319,52 +385,33 @@ static void jpeg_v2_5_enable_clock_gating(struct amdgpu_device *adev, int inst)
  */
 static int jpeg_v2_5_start(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring;
 	int i;
 
 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
 		if (adev->jpeg.harvest_config & (1 << i))
 			continue;
+		jpeg_v2_5_start_inst(adev, i);
 
-		ring = &adev->jpeg.inst[i].ring_dec;
-		/* disable anti hang mechanism */
-		WREG32_P(SOC15_REG_OFFSET(JPEG, i, mmUVD_JPEG_POWER_STATUS), 0,
-			~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK);
-
-		/* JPEG disable CGC */
-		jpeg_v2_5_disable_clock_gating(adev, i);
-
-		/* MJPEG global tiling registers */
-		WREG32_SOC15(JPEG, i, mmJPEG_DEC_GFX8_ADDR_CONFIG,
-			adev->gfx.config.gb_addr_config);
-		WREG32_SOC15(JPEG, i, mmJPEG_DEC_GFX10_ADDR_CONFIG,
-			adev->gfx.config.gb_addr_config);
-
-		/* enable JMI channel */
-		WREG32_P(SOC15_REG_OFFSET(JPEG, i, mmUVD_JMI_CNTL), 0,
-			~UVD_JMI_CNTL__SOFT_RESET_MASK);
-
-		/* enable System Interrupt for JRBC */
-		WREG32_P(SOC15_REG_OFFSET(JPEG, i, mmJPEG_SYS_INT_EN),
-			JPEG_SYS_INT_EN__DJRBC_MASK,
-			~JPEG_SYS_INT_EN__DJRBC_MASK);
-
-		WREG32_SOC15(JPEG, i, mmUVD_LMI_JRBC_RB_VMID, 0);
-		WREG32_SOC15(JPEG, i, mmUVD_JRBC_RB_CNTL, (0x00000001L | 0x00000002L));
-		WREG32_SOC15(JPEG, i, mmUVD_LMI_JRBC_RB_64BIT_BAR_LOW,
-			lower_32_bits(ring->gpu_addr));
-		WREG32_SOC15(JPEG, i, mmUVD_LMI_JRBC_RB_64BIT_BAR_HIGH,
-			upper_32_bits(ring->gpu_addr));
-		WREG32_SOC15(JPEG, i, mmUVD_JRBC_RB_RPTR, 0);
-		WREG32_SOC15(JPEG, i, mmUVD_JRBC_RB_WPTR, 0);
-		WREG32_SOC15(JPEG, i, mmUVD_JRBC_RB_CNTL, 0x00000002L);
-		WREG32_SOC15(JPEG, i, mmUVD_JRBC_RB_SIZE, ring->ring_size / 4);
-		ring->wptr = RREG32_SOC15(JPEG, i, mmUVD_JRBC_RB_WPTR);
 	}
 
 	return 0;
 }
 
+static void jpeg_v2_5_stop_inst(struct amdgpu_device *adev, int i)
+{
+	/* reset JMI */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, i, mmUVD_JMI_CNTL),
+		UVD_JMI_CNTL__SOFT_RESET_MASK,
+		~UVD_JMI_CNTL__SOFT_RESET_MASK);
+
+	jpeg_v2_5_enable_clock_gating(adev, i);
+
+	/* enable anti hang mechanism */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, i, mmUVD_JPEG_POWER_STATUS),
+		UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK,
+		~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK);
+}
+
 /**
  * jpeg_v2_5_stop - stop JPEG block
  *
@@ -379,18 +426,7 @@ static int jpeg_v2_5_stop(struct amdgpu_device *adev)
 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
 		if (adev->jpeg.harvest_config & (1 << i))
 			continue;
-
-		/* reset JMI */
-		WREG32_P(SOC15_REG_OFFSET(JPEG, i, mmUVD_JMI_CNTL),
-			UVD_JMI_CNTL__SOFT_RESET_MASK,
-			~UVD_JMI_CNTL__SOFT_RESET_MASK);
-
-		jpeg_v2_5_enable_clock_gating(adev, i);
-
-		/* enable anti hang mechanism */
-		WREG32_P(SOC15_REG_OFFSET(JPEG, i, mmUVD_JPEG_POWER_STATUS),
-			UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK,
-			~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK);
+		jpeg_v2_5_stop_inst(adev, i);
 	}
 
 	return 0;
@@ -482,9 +518,9 @@ static void jpeg_v2_6_dec_ring_insert_end(struct amdgpu_ring *ring)
 	amdgpu_ring_write(ring, (1 << (ring->me * 2 + 14)));
 }
 
-static bool jpeg_v2_5_is_idle(void *handle)
+static bool jpeg_v2_5_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i, ret = 1;
 
 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
@@ -499,9 +535,9 @@ static bool jpeg_v2_5_is_idle(void *handle)
 	return ret;
 }
 
-static int jpeg_v2_5_wait_for_idle(void *handle)
+static int jpeg_v2_5_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i, ret;
 
 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
@@ -518,10 +554,10 @@ static int jpeg_v2_5_wait_for_idle(void *handle)
 	return 0;
 }
 
-static int jpeg_v2_5_set_clockgating_state(void *handle,
+static int jpeg_v2_5_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_CG_STATE_GATE);
 	int i;
 
@@ -530,7 +566,7 @@ static int jpeg_v2_5_set_clockgating_state(void *handle,
 			continue;
 
 		if (enable) {
-			if (!jpeg_v2_5_is_idle(handle))
+			if (!jpeg_v2_5_is_idle(ip_block))
 				return -EBUSY;
 			jpeg_v2_5_enable_clock_gating(adev, i);
 		} else {
@@ -541,13 +577,13 @@ static int jpeg_v2_5_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int jpeg_v2_5_set_powergating_state(void *handle,
+static int jpeg_v2_5_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
-	if(state == adev->jpeg.cur_state)
+	if (state == adev->jpeg.cur_state)
 		return 0;
 
 	if (state == AMD_PG_STATE_GATE)
@@ -555,7 +591,7 @@ static int jpeg_v2_5_set_powergating_state(void *handle,
 	else
 		ret = jpeg_v2_5_start(adev);
 
-	if(!ret)
+	if (!ret)
 		adev->jpeg.cur_state = state;
 
 	return ret;
@@ -569,6 +605,14 @@ static int jpeg_v2_5_set_interrupt_state(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int jpeg_v2_6_set_ras_interrupt_state(struct amdgpu_device *adev,
+					struct amdgpu_irq_src *source,
+					unsigned int type,
+					enum amdgpu_interrupt_state state)
+{
+	return 0;
+}
+
 static int jpeg_v2_5_process_interrupt(struct amdgpu_device *adev,
 				      struct amdgpu_irq_src *source,
 				      struct amdgpu_iv_entry *entry)
@@ -591,11 +635,7 @@ static int jpeg_v2_5_process_interrupt(struct amdgpu_device *adev,
 
 	switch (entry->src_id) {
 	case VCN_2_0__SRCID__JPEG_DECODE:
-		amdgpu_fence_process(&adev->jpeg.inst[ip_instance].ring_dec);
-		break;
-	case VCN_2_6__SRCID_DJPEG0_POISON:
-	case VCN_2_6__SRCID_EJPEG0_POISON:
-		amdgpu_jpeg_process_poison_irq(adev, source, entry);
+		amdgpu_fence_process(adev->jpeg.inst[ip_instance].ring_dec);
 		break;
 	default:
 		DRM_ERROR("Unhandled interrupt: %d %d\n",
@@ -606,10 +646,19 @@ static int jpeg_v2_5_process_interrupt(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int jpeg_v2_5_ring_reset(struct amdgpu_ring *ring,
+				unsigned int vmid,
+				struct amdgpu_fence *timedout_fence)
+{
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+	jpeg_v2_5_stop_inst(ring->adev, ring->me);
+	jpeg_v2_5_start_inst(ring->adev, ring->me);
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
 static const struct amd_ip_funcs jpeg_v2_5_ip_funcs = {
 	.name = "jpeg_v2_5",
 	.early_init = jpeg_v2_5_early_init,
-	.late_init = NULL,
 	.sw_init = jpeg_v2_5_sw_init,
 	.sw_fini = jpeg_v2_5_sw_fini,
 	.hw_init = jpeg_v2_5_hw_init,
@@ -618,18 +667,15 @@ static const struct amd_ip_funcs jpeg_v2_5_ip_funcs = {
 	.resume = jpeg_v2_5_resume,
 	.is_idle = jpeg_v2_5_is_idle,
 	.wait_for_idle = jpeg_v2_5_wait_for_idle,
-	.check_soft_reset = NULL,
-	.pre_soft_reset = NULL,
-	.soft_reset = NULL,
-	.post_soft_reset = NULL,
 	.set_clockgating_state = jpeg_v2_5_set_clockgating_state,
 	.set_powergating_state = jpeg_v2_5_set_powergating_state,
+	.dump_ip_state = amdgpu_jpeg_dump_ip_state,
+	.print_ip_state = amdgpu_jpeg_print_ip_state,
 };
 
 static const struct amd_ip_funcs jpeg_v2_6_ip_funcs = {
 	.name = "jpeg_v2_6",
 	.early_init = jpeg_v2_5_early_init,
-	.late_init = NULL,
 	.sw_init = jpeg_v2_5_sw_init,
 	.sw_fini = jpeg_v2_5_sw_fini,
 	.hw_init = jpeg_v2_5_hw_init,
@@ -638,12 +684,10 @@ static const struct amd_ip_funcs jpeg_v2_6_ip_funcs = {
 	.resume = jpeg_v2_5_resume,
 	.is_idle = jpeg_v2_5_is_idle,
 	.wait_for_idle = jpeg_v2_5_wait_for_idle,
-	.check_soft_reset = NULL,
-	.pre_soft_reset = NULL,
-	.soft_reset = NULL,
-	.post_soft_reset = NULL,
 	.set_clockgating_state = jpeg_v2_5_set_clockgating_state,
 	.set_powergating_state = jpeg_v2_5_set_powergating_state,
+	.dump_ip_state = amdgpu_jpeg_dump_ip_state,
+	.print_ip_state = amdgpu_jpeg_print_ip_state,
 };
 
 static const struct amdgpu_ring_funcs jpeg_v2_5_dec_ring_vm_funcs = {
@@ -652,6 +696,7 @@ static const struct amdgpu_ring_funcs jpeg_v2_5_dec_ring_vm_funcs = {
 	.get_rptr = jpeg_v2_5_dec_ring_get_rptr,
 	.get_wptr = jpeg_v2_5_dec_ring_get_wptr,
 	.set_wptr = jpeg_v2_5_dec_ring_set_wptr,
+	.parse_cs = amdgpu_jpeg_dec_parse_cs,
 	.emit_frame_size =
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 +
@@ -673,6 +718,7 @@ static const struct amdgpu_ring_funcs jpeg_v2_5_dec_ring_vm_funcs = {
 	.emit_wreg = jpeg_v2_0_dec_ring_emit_wreg,
 	.emit_reg_wait = jpeg_v2_0_dec_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = jpeg_v2_5_ring_reset,
 };
 
 static const struct amdgpu_ring_funcs jpeg_v2_6_dec_ring_vm_funcs = {
@@ -681,6 +727,7 @@ static const struct amdgpu_ring_funcs jpeg_v2_6_dec_ring_vm_funcs = {
 	.get_rptr = jpeg_v2_5_dec_ring_get_rptr,
 	.get_wptr = jpeg_v2_5_dec_ring_get_wptr,
 	.set_wptr = jpeg_v2_5_dec_ring_set_wptr,
+	.parse_cs = amdgpu_jpeg_dec_parse_cs,
 	.emit_frame_size =
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 +
@@ -702,6 +749,7 @@ static const struct amdgpu_ring_funcs jpeg_v2_6_dec_ring_vm_funcs = {
 	.emit_wreg = jpeg_v2_0_dec_ring_emit_wreg,
 	.emit_reg_wait = jpeg_v2_0_dec_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = jpeg_v2_5_ring_reset,
 };
 
 static void jpeg_v2_5_set_dec_ring_funcs(struct amdgpu_device *adev)
@@ -712,11 +760,10 @@ static void jpeg_v2_5_set_dec_ring_funcs(struct amdgpu_device *adev)
 		if (adev->jpeg.harvest_config & (1 << i))
 			continue;
 		if (adev->asic_type == CHIP_ARCTURUS)
-			adev->jpeg.inst[i].ring_dec.funcs = &jpeg_v2_5_dec_ring_vm_funcs;
+			adev->jpeg.inst[i].ring_dec->funcs = &jpeg_v2_5_dec_ring_vm_funcs;
 		else  /* CHIP_ALDEBARAN */
-			adev->jpeg.inst[i].ring_dec.funcs = &jpeg_v2_6_dec_ring_vm_funcs;
-		adev->jpeg.inst[i].ring_dec.me = i;
-		DRM_INFO("JPEG(%d) JPEG decode is enabled in VM mode\n", i);
+			adev->jpeg.inst[i].ring_dec->funcs = &jpeg_v2_6_dec_ring_vm_funcs;
+		adev->jpeg.inst[i].ring_dec->me = i;
 	}
 }
 
@@ -725,6 +772,11 @@ static const struct amdgpu_irq_src_funcs jpeg_v2_5_irq_funcs = {
 	.process = jpeg_v2_5_process_interrupt,
 };
 
+static const struct amdgpu_irq_src_funcs jpeg_v2_6_ras_irq_funcs = {
+	.set = jpeg_v2_6_set_ras_interrupt_state,
+	.process = amdgpu_jpeg_process_poison_irq,
+};
+
 static void jpeg_v2_5_set_irq_funcs(struct amdgpu_device *adev)
 {
 	int i;
@@ -735,11 +787,13 @@ static void jpeg_v2_5_set_irq_funcs(struct amdgpu_device *adev)
 
 		adev->jpeg.inst[i].irq.num_types = 1;
 		adev->jpeg.inst[i].irq.funcs = &jpeg_v2_5_irq_funcs;
+
+		adev->jpeg.inst[i].ras_poison_irq.num_types = 1;
+		adev->jpeg.inst[i].ras_poison_irq.funcs = &jpeg_v2_6_ras_irq_funcs;
 	}
 }
 
-const struct amdgpu_ip_block_version jpeg_v2_5_ip_block =
-{
+const struct amdgpu_ip_block_version jpeg_v2_5_ip_block = {
 		.type = AMD_IP_BLOCK_TYPE_JPEG,
 		.major = 2,
 		.minor = 5,
@@ -747,8 +801,7 @@ const struct amdgpu_ip_block_version jpeg_v2_5_ip_block =
 		.funcs = &jpeg_v2_5_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version jpeg_v2_6_ip_block =
-{
+const struct amdgpu_ip_block_version jpeg_v2_6_ip_block = {
 		.type = AMD_IP_BLOCK_TYPE_JPEG,
 		.major = 2,
 		.minor = 6,
@@ -800,12 +853,13 @@ const struct amdgpu_ras_block_hw_ops jpeg_v2_6_ras_hw_ops = {
 static struct amdgpu_jpeg_ras jpeg_v2_6_ras = {
 	.ras_block = {
 		.hw_ops = &jpeg_v2_6_ras_hw_ops,
+		.ras_late_init = amdgpu_jpeg_ras_late_init,
 	},
 };
 
 static void jpeg_v2_5_set_ras_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[JPEG_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, JPEG_HWIP, 0)) {
 	case IP_VERSION(2, 6, 0):
 		adev->jpeg.ras = &jpeg_v2_6_ras;
 		break;
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c
index c55e09432e26..d1a011c40ba2 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c
@@ -34,26 +34,43 @@
 
 #define mmUVD_JPEG_PITCH_INTERNAL_OFFSET	0x401f
 
+static const struct amdgpu_hwip_reg_entry jpeg_reg_list_3_0[] = {
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JPEG_POWER_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JPEG_INT_STAT),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JRBC_RB_CNTL),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JRBC_RB_SIZE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmJPEG_DEC_ADDR_MODE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmJPEG_DEC_GFX10_ADDR_CONFIG),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmJPEG_DEC_Y_GFX10_TILING_SURFACE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmJPEG_DEC_UV_GFX10_TILING_SURFACE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JPEG_PITCH),
+	SOC15_REG_ENTRY_STR(JPEG, 0, mmUVD_JPEG_UV_PITCH),
+};
+
 static void jpeg_v3_0_set_dec_ring_funcs(struct amdgpu_device *adev);
 static void jpeg_v3_0_set_irq_funcs(struct amdgpu_device *adev);
-static int jpeg_v3_0_set_powergating_state(void *handle,
+static int jpeg_v3_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 				enum amd_powergating_state state);
 
 /**
  * jpeg_v3_0_early_init - set function pointers
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Set ring and irq function pointers
  */
-static int jpeg_v3_0_early_init(void *handle)
+static int jpeg_v3_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	u32 harvest;
 
-	switch (adev->ip_versions[UVD_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, UVD_HWIP, 0)) {
 	case IP_VERSION(3, 1, 1):
+	case IP_VERSION(3, 1, 2):
 		break;
 	default:
 		harvest = RREG32_SOC15(JPEG, 0, mmCC_UVD_HARVESTING);
@@ -63,6 +80,7 @@ static int jpeg_v3_0_early_init(void *handle)
 	}
 
 	adev->jpeg.num_jpeg_inst = 1;
+	adev->jpeg.num_jpeg_rings = 1;
 
 	jpeg_v3_0_set_dec_ring_funcs(adev);
 	jpeg_v3_0_set_irq_funcs(adev);
@@ -73,13 +91,13 @@ static int jpeg_v3_0_early_init(void *handle)
 /**
  * jpeg_v3_0_sw_init - sw init for JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Load firmware and sw initialization
  */
-static int jpeg_v3_0_sw_init(void *handle)
+static int jpeg_v3_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring;
 	int r;
 
@@ -97,38 +115,50 @@ static int jpeg_v3_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	ring = &adev->jpeg.inst->ring_dec;
+	ring = adev->jpeg.inst->ring_dec;
 	ring->use_doorbell = true;
 	ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1;
-	ring->vm_hub = AMDGPU_MMHUB_0;
+	ring->vm_hub = AMDGPU_MMHUB0(0);
 	sprintf(ring->name, "jpeg_dec");
 	r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq, 0,
 			     AMDGPU_RING_PRIO_DEFAULT, NULL);
 	if (r)
 		return r;
 
-	adev->jpeg.internal.jpeg_pitch = mmUVD_JPEG_PITCH_INTERNAL_OFFSET;
-	adev->jpeg.inst->external.jpeg_pitch = SOC15_REG_OFFSET(JPEG, 0, mmUVD_JPEG_PITCH);
+	adev->jpeg.internal.jpeg_pitch[0] = mmUVD_JPEG_PITCH_INTERNAL_OFFSET;
+	adev->jpeg.inst->external.jpeg_pitch[0] = SOC15_REG_OFFSET(JPEG, 0, mmUVD_JPEG_PITCH);
 
-	return 0;
+	r = amdgpu_jpeg_reg_dump_init(adev, jpeg_reg_list_3_0, ARRAY_SIZE(jpeg_reg_list_3_0));
+	if (r)
+		return r;
+
+	adev->jpeg.supported_reset =
+		amdgpu_get_soft_full_reset_mask(adev->jpeg.inst[0].ring_dec);
+	if (!amdgpu_sriov_vf(adev))
+		adev->jpeg.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+	r = amdgpu_jpeg_sysfs_reset_mask_init(adev);
+
+	return r;
 }
 
 /**
  * jpeg_v3_0_sw_fini - sw fini for JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * JPEG suspend and free up sw allocation
  */
-static int jpeg_v3_0_sw_fini(void *handle)
+static int jpeg_v3_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_jpeg_suspend(adev);
 	if (r)
 		return r;
 
+	amdgpu_jpeg_sysfs_reset_mask_fini(adev);
+
 	r = amdgpu_jpeg_sw_fini(adev);
 
 	return r;
@@ -137,43 +167,36 @@ static int jpeg_v3_0_sw_fini(void *handle)
 /**
  * jpeg_v3_0_hw_init - start and test JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  */
-static int jpeg_v3_0_hw_init(void *handle)
+static int jpeg_v3_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	struct amdgpu_ring *ring = &adev->jpeg.inst->ring_dec;
-	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring = adev->jpeg.inst->ring_dec;
 
 	adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell,
 		(adev->doorbell_index.vcn.vcn_ring0_1 << 1), 0);
 
-	r = amdgpu_ring_test_helper(ring);
-	if (r)
-		return r;
-
-	DRM_INFO("JPEG decode initialized successfully.\n");
-
-	return 0;
+	return amdgpu_ring_test_helper(ring);
 }
 
 /**
  * jpeg_v3_0_hw_fini - stop the hardware block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Stop the JPEG block, mark ring as not ready any more
  */
-static int jpeg_v3_0_hw_fini(void *handle)
+static int jpeg_v3_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	cancel_delayed_work_sync(&adev->vcn.idle_work);
+	cancel_delayed_work_sync(&adev->jpeg.idle_work);
 
 	if (adev->jpeg.cur_state != AMD_PG_STATE_GATE &&
 	      RREG32_SOC15(JPEG, 0, mmUVD_JRBC_STATUS))
-		jpeg_v3_0_set_powergating_state(adev, AMD_PG_STATE_GATE);
+		jpeg_v3_0_set_powergating_state(ip_block, AMD_PG_STATE_GATE);
 
 	return 0;
 }
@@ -181,20 +204,19 @@ static int jpeg_v3_0_hw_fini(void *handle)
 /**
  * jpeg_v3_0_suspend - suspend JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * HW fini and suspend JPEG block
  */
-static int jpeg_v3_0_suspend(void *handle)
+static int jpeg_v3_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	int r;
 
-	r = jpeg_v3_0_hw_fini(adev);
+	r = jpeg_v3_0_hw_fini(ip_block);
 	if (r)
 		return r;
 
-	r = amdgpu_jpeg_suspend(adev);
+	r = amdgpu_jpeg_suspend(ip_block->adev);
 
 	return r;
 }
@@ -202,20 +224,19 @@ static int jpeg_v3_0_suspend(void *handle)
 /**
  * jpeg_v3_0_resume - resume JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Resume firmware and hw init JPEG block
  */
-static int jpeg_v3_0_resume(void *handle)
+static int jpeg_v3_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	int r;
 
-	r = amdgpu_jpeg_resume(adev);
+	r = amdgpu_jpeg_resume(ip_block->adev);
 	if (r)
 		return r;
 
-	r = jpeg_v3_0_hw_init(adev);
+	r = jpeg_v3_0_hw_init(ip_block);
 
 	return r;
 }
@@ -329,7 +350,7 @@ static int jpeg_v3_0_enable_static_power_gating(struct amdgpu_device *adev)
  */
 static int jpeg_v3_0_start(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring = &adev->jpeg.inst->ring_dec;
+	struct amdgpu_ring *ring = adev->jpeg.inst->ring_dec;
 	int r;
 
 	if (adev->pm.dpm_enabled)
@@ -452,9 +473,9 @@ static void jpeg_v3_0_dec_ring_set_wptr(struct amdgpu_ring *ring)
 	}
 }
 
-static bool jpeg_v3_0_is_idle(void *handle)
+static bool jpeg_v3_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret = 1;
 
 	ret &= (((RREG32_SOC15(JPEG, 0, mmUVD_JRBC_STATUS) &
@@ -464,23 +485,23 @@ static bool jpeg_v3_0_is_idle(void *handle)
 	return ret;
 }
 
-static int jpeg_v3_0_wait_for_idle(void *handle)
+static int jpeg_v3_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return SOC15_WAIT_ON_RREG(JPEG, 0, mmUVD_JRBC_STATUS,
 		UVD_JRBC_STATUS__RB_JOB_DONE_MASK,
 		UVD_JRBC_STATUS__RB_JOB_DONE_MASK);
 }
 
-static int jpeg_v3_0_set_clockgating_state(void *handle,
+static int jpeg_v3_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	bool enable = (state == AMD_CG_STATE_GATE) ? true : false;
+	struct amdgpu_device *adev = ip_block->adev;
+	bool enable = state == AMD_CG_STATE_GATE;
 
 	if (enable) {
-		if (!jpeg_v3_0_is_idle(handle))
+		if (!jpeg_v3_0_is_idle(ip_block))
 			return -EBUSY;
 		jpeg_v3_0_enable_clock_gating(adev);
 	} else {
@@ -490,10 +511,10 @@ static int jpeg_v3_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int jpeg_v3_0_set_powergating_state(void *handle,
+static int jpeg_v3_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	if(state == adev->jpeg.cur_state)
@@ -526,7 +547,7 @@ static int jpeg_v3_0_process_interrupt(struct amdgpu_device *adev,
 
 	switch (entry->src_id) {
 	case VCN_2_0__SRCID__JPEG_DECODE:
-		amdgpu_fence_process(&adev->jpeg.inst->ring_dec);
+		amdgpu_fence_process(adev->jpeg.inst->ring_dec);
 		break;
 	default:
 		DRM_ERROR("Unhandled interrupt: %d %d\n",
@@ -537,10 +558,25 @@ static int jpeg_v3_0_process_interrupt(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int jpeg_v3_0_ring_reset(struct amdgpu_ring *ring,
+				unsigned int vmid,
+				struct amdgpu_fence *timedout_fence)
+{
+	int r;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+	r = jpeg_v3_0_stop(ring->adev);
+	if (r)
+		return r;
+	r = jpeg_v3_0_start(ring->adev);
+	if (r)
+		return r;
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
 static const struct amd_ip_funcs jpeg_v3_0_ip_funcs = {
 	.name = "jpeg_v3_0",
 	.early_init = jpeg_v3_0_early_init,
-	.late_init = NULL,
 	.sw_init = jpeg_v3_0_sw_init,
 	.sw_fini = jpeg_v3_0_sw_fini,
 	.hw_init = jpeg_v3_0_hw_init,
@@ -549,12 +585,10 @@ static const struct amd_ip_funcs jpeg_v3_0_ip_funcs = {
 	.resume = jpeg_v3_0_resume,
 	.is_idle = jpeg_v3_0_is_idle,
 	.wait_for_idle = jpeg_v3_0_wait_for_idle,
-	.check_soft_reset = NULL,
-	.pre_soft_reset = NULL,
-	.soft_reset = NULL,
-	.post_soft_reset = NULL,
 	.set_clockgating_state = jpeg_v3_0_set_clockgating_state,
 	.set_powergating_state = jpeg_v3_0_set_powergating_state,
+	.dump_ip_state = amdgpu_jpeg_dump_ip_state,
+	.print_ip_state = amdgpu_jpeg_print_ip_state,
 };
 
 static const struct amdgpu_ring_funcs jpeg_v3_0_dec_ring_vm_funcs = {
@@ -563,6 +597,7 @@ static const struct amdgpu_ring_funcs jpeg_v3_0_dec_ring_vm_funcs = {
 	.get_rptr = jpeg_v3_0_dec_ring_get_rptr,
 	.get_wptr = jpeg_v3_0_dec_ring_get_wptr,
 	.set_wptr = jpeg_v3_0_dec_ring_set_wptr,
+	.parse_cs = amdgpu_jpeg_dec_parse_cs,
 	.emit_frame_size =
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 +
@@ -584,12 +619,12 @@ static const struct amdgpu_ring_funcs jpeg_v3_0_dec_ring_vm_funcs = {
 	.emit_wreg = jpeg_v2_0_dec_ring_emit_wreg,
 	.emit_reg_wait = jpeg_v2_0_dec_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = jpeg_v3_0_ring_reset,
 };
 
 static void jpeg_v3_0_set_dec_ring_funcs(struct amdgpu_device *adev)
 {
-	adev->jpeg.inst->ring_dec.funcs = &jpeg_v3_0_dec_ring_vm_funcs;
-	DRM_INFO("JPEG decode is enabled in VM mode\n");
+	adev->jpeg.inst->ring_dec->funcs = &jpeg_v3_0_dec_ring_vm_funcs;
 }
 
 static const struct amdgpu_irq_src_funcs jpeg_v3_0_irq_funcs = {
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c
index a6ad678fd507..33db2c1ae6cc 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c
@@ -36,28 +36,44 @@
 
 #define regUVD_JPEG_PITCH_INTERNAL_OFFSET                  0x401f
 
+static const struct amdgpu_hwip_reg_entry jpeg_reg_list_4_0[] = {
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_POWER_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_INT_STAT),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC_RB_CNTL),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC_RB_SIZE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_ADDR_MODE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_GFX10_ADDR_CONFIG),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_Y_GFX10_TILING_SURFACE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_UV_GFX10_TILING_SURFACE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_PITCH),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_UV_PITCH),
+};
+
 static int jpeg_v4_0_start_sriov(struct amdgpu_device *adev);
 static void jpeg_v4_0_set_dec_ring_funcs(struct amdgpu_device *adev);
 static void jpeg_v4_0_set_irq_funcs(struct amdgpu_device *adev);
-static int jpeg_v4_0_set_powergating_state(void *handle,
+static int jpeg_v4_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 				enum amd_powergating_state state);
 static void jpeg_v4_0_set_ras_funcs(struct amdgpu_device *adev);
-
 static void jpeg_v4_0_dec_ring_set_wptr(struct amdgpu_ring *ring);
 
 /**
  * jpeg_v4_0_early_init - set function pointers
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Set ring and irq function pointers
  */
-static int jpeg_v4_0_early_init(void *handle)
+static int jpeg_v4_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 
 	adev->jpeg.num_jpeg_inst = 1;
+	adev->jpeg.num_jpeg_rings = 1;
 
 	jpeg_v4_0_set_dec_ring_funcs(adev);
 	jpeg_v4_0_set_irq_funcs(adev);
@@ -69,13 +85,13 @@ static int jpeg_v4_0_early_init(void *handle)
 /**
  * jpeg_v4_0_sw_init - sw init for JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Load firmware and sw initialization
  */
-static int jpeg_v4_0_sw_init(void *handle)
+static int jpeg_v4_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring;
 	int r;
 
@@ -87,13 +103,13 @@ static int jpeg_v4_0_sw_init(void *handle)
 
 	/* JPEG DJPEG POISON EVENT */
 	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
-			VCN_4_0__SRCID_DJPEG0_POISON, &adev->jpeg.inst->irq);
+			VCN_4_0__SRCID_DJPEG0_POISON, &adev->jpeg.inst->ras_poison_irq);
 	if (r)
 		return r;
 
 	/* JPEG EJPEG POISON EVENT */
 	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
-			VCN_4_0__SRCID_EJPEG0_POISON, &adev->jpeg.inst->irq);
+			VCN_4_0__SRCID_EJPEG0_POISON, &adev->jpeg.inst->ras_poison_irq);
 	if (r)
 		return r;
 
@@ -105,10 +121,10 @@ static int jpeg_v4_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	ring = &adev->jpeg.inst->ring_dec;
+	ring = adev->jpeg.inst->ring_dec;
 	ring->use_doorbell = true;
 	ring->doorbell_index = amdgpu_sriov_vf(adev) ? (((adev->doorbell_index.vcn.vcn_ring0_1) << 1) + 4) : ((adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1);
-	ring->vm_hub = AMDGPU_MMHUB_0;
+	ring->vm_hub = AMDGPU_MMHUB0(0);
 
 	sprintf(ring->name, "jpeg_dec");
 	r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq, 0,
@@ -116,32 +132,43 @@ static int jpeg_v4_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	adev->jpeg.internal.jpeg_pitch = regUVD_JPEG_PITCH_INTERNAL_OFFSET;
-	adev->jpeg.inst->external.jpeg_pitch = SOC15_REG_OFFSET(JPEG, 0, regUVD_JPEG_PITCH);
+	adev->jpeg.internal.jpeg_pitch[0] = regUVD_JPEG_PITCH_INTERNAL_OFFSET;
+	adev->jpeg.inst->external.jpeg_pitch[0] = SOC15_REG_OFFSET(JPEG, 0, regUVD_JPEG_PITCH);
 
 	r = amdgpu_jpeg_ras_sw_init(adev);
 	if (r)
 		return r;
 
-	return 0;
+	r = amdgpu_jpeg_reg_dump_init(adev, jpeg_reg_list_4_0, ARRAY_SIZE(jpeg_reg_list_4_0));
+	if (r)
+		return r;
+
+	adev->jpeg.supported_reset =
+		amdgpu_get_soft_full_reset_mask(adev->jpeg.inst[0].ring_dec);
+	if (!amdgpu_sriov_vf(adev))
+		adev->jpeg.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+	r = amdgpu_jpeg_sysfs_reset_mask_init(adev);
+
+	return r;
 }
 
 /**
  * jpeg_v4_0_sw_fini - sw fini for JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * JPEG suspend and free up sw allocation
  */
-static int jpeg_v4_0_sw_fini(void *handle)
+static int jpeg_v4_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_jpeg_suspend(adev);
 	if (r)
 		return r;
 
+	amdgpu_jpeg_sysfs_reset_mask_fini(adev);
 	r = amdgpu_jpeg_sw_fini(adev);
 
 	return r;
@@ -150,13 +177,13 @@ static int jpeg_v4_0_sw_fini(void *handle)
 /**
  * jpeg_v4_0_hw_init - start and test JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  */
-static int jpeg_v4_0_hw_init(void *handle)
+static int jpeg_v4_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	struct amdgpu_ring *ring = &adev->jpeg.inst->ring_dec;
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring = adev->jpeg.inst->ring_dec;
 	int r;
 
 	if (amdgpu_sriov_vf(adev)) {
@@ -180,29 +207,28 @@ static int jpeg_v4_0_hw_init(void *handle)
 			return r;
 	}
 
-	DRM_DEV_INFO(adev->dev, "JPEG decode initialized successfully.\n");
-
 	return 0;
 }
 
 /**
  * jpeg_v4_0_hw_fini - stop the hardware block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Stop the JPEG block, mark ring as not ready any more
  */
-static int jpeg_v4_0_hw_fini(void *handle)
+static int jpeg_v4_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	cancel_delayed_work_sync(&adev->vcn.idle_work);
+	cancel_delayed_work_sync(&adev->jpeg.idle_work);
 	if (!amdgpu_sriov_vf(adev)) {
 		if (adev->jpeg.cur_state != AMD_PG_STATE_GATE &&
 			RREG32_SOC15(JPEG, 0, regUVD_JRBC_STATUS))
-			jpeg_v4_0_set_powergating_state(adev, AMD_PG_STATE_GATE);
+			jpeg_v4_0_set_powergating_state(ip_block, AMD_PG_STATE_GATE);
 	}
-	amdgpu_irq_put(adev, &adev->jpeg.inst->irq, 0);
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__JPEG))
+		amdgpu_irq_put(adev, &adev->jpeg.inst->ras_poison_irq, 0);
 
 	return 0;
 }
@@ -210,20 +236,19 @@ static int jpeg_v4_0_hw_fini(void *handle)
 /**
  * jpeg_v4_0_suspend - suspend JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * HW fini and suspend JPEG block
  */
-static int jpeg_v4_0_suspend(void *handle)
+static int jpeg_v4_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	int r;
 
-	r = jpeg_v4_0_hw_fini(adev);
+	r = jpeg_v4_0_hw_fini(ip_block);
 	if (r)
 		return r;
 
-	r = amdgpu_jpeg_suspend(adev);
+	r = amdgpu_jpeg_suspend(ip_block->adev);
 
 	return r;
 }
@@ -231,20 +256,19 @@ static int jpeg_v4_0_suspend(void *handle)
 /**
  * jpeg_v4_0_resume - resume JPEG block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Resume firmware and hw init JPEG block
  */
-static int jpeg_v4_0_resume(void *handle)
+static int jpeg_v4_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	int r;
 
-	r = amdgpu_jpeg_resume(adev);
+	r = amdgpu_jpeg_resume(ip_block->adev);
 	if (r)
 		return r;
 
-	r = jpeg_v4_0_hw_init(adev);
+	r = jpeg_v4_0_hw_init(ip_block);
 
 	return r;
 }
@@ -363,7 +387,7 @@ static int jpeg_v4_0_enable_static_power_gating(struct amdgpu_device *adev)
  */
 static int jpeg_v4_0_start(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring = &adev->jpeg.inst->ring_dec;
+	struct amdgpu_ring *ring = adev->jpeg.inst->ring_dec;
 	int r;
 
 	if (adev->pm.dpm_enabled)
@@ -429,8 +453,12 @@ static int jpeg_v4_0_start_sriov(struct amdgpu_device *adev)
 	end.cmd_header.command_type =
 		MMSCH_COMMAND__END;
 
+	size = sizeof(struct mmsch_v4_0_init_header);
+	table_loc = (uint32_t *)table->cpu_addr;
+	memcpy(&header, (void *)table_loc, size);
+
 	header.version = MMSCH_VERSION;
-	header.total_size = sizeof(struct mmsch_v4_0_init_header) >> 2;
+	header.total_size = RREG32_SOC15(VCN, 0, regMMSCH_VF_CTX_SIZE);
 
 	header.jpegdec.init_status = 0;
 	header.jpegdec.table_offset = 0;
@@ -441,7 +469,7 @@ static int jpeg_v4_0_start_sriov(struct amdgpu_device *adev)
 
 	table_size = 0;
 
-	ring = &adev->jpeg.inst->ring_dec;
+	ring = adev->jpeg.inst->ring_dec;
 
 	MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(JPEG, 0,
 		regUVD_LMI_JRBC_RB_64BIT_BAR_LOW),
@@ -466,6 +494,9 @@ static int jpeg_v4_0_start_sriov(struct amdgpu_device *adev)
 	table_loc = (uint32_t *)table->cpu_addr;
 	memcpy((void *)table_loc, &header, size);
 
+	/* Perform HDP flush before writing to MMSCH registers */
+	amdgpu_device_flush_hdp(adev, NULL);
+
 	/* message MMSCH (in VCN[0]) to initialize this client
 	 * 1, write to mmsch_vf_ctx_addr_lo/hi register with GPU mc addr
 	 * of memory descriptor location
@@ -513,8 +544,11 @@ static int jpeg_v4_0_start_sriov(struct amdgpu_device *adev)
 			return -EBUSY;
 		}
 	}
-	if (resp != expected && resp != MMSCH_VF_MAILBOX_RESP__INCOMPLETE && init_status != MMSCH_VF_ENGINE_STATUS__PASS)
+	if (resp != expected && resp != MMSCH_VF_MAILBOX_RESP__INCOMPLETE
+			&& init_status != MMSCH_VF_ENGINE_STATUS__PASS) {
 		DRM_ERROR("MMSCH init status is incorrect! readback=0x%08x, header init status for jpeg: %x\n", resp, init_status);
+		return -EINVAL;
+	}
 
 	return 0;
 
@@ -599,9 +633,9 @@ static void jpeg_v4_0_dec_ring_set_wptr(struct amdgpu_ring *ring)
 	}
 }
 
-static bool jpeg_v4_0_is_idle(void *handle)
+static bool jpeg_v4_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret = 1;
 
 	ret &= (((RREG32_SOC15(JPEG, 0, regUVD_JRBC_STATUS) &
@@ -611,23 +645,23 @@ static bool jpeg_v4_0_is_idle(void *handle)
 	return ret;
 }
 
-static int jpeg_v4_0_wait_for_idle(void *handle)
+static int jpeg_v4_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return SOC15_WAIT_ON_RREG(JPEG, 0, regUVD_JRBC_STATUS,
 		UVD_JRBC_STATUS__RB_JOB_DONE_MASK,
 		UVD_JRBC_STATUS__RB_JOB_DONE_MASK);
 }
 
-static int jpeg_v4_0_set_clockgating_state(void *handle,
+static int jpeg_v4_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	bool enable = (state == AMD_CG_STATE_GATE) ? true : false;
+	struct amdgpu_device *adev = ip_block->adev;
+	bool enable = state == AMD_CG_STATE_GATE;
 
 	if (enable) {
-		if (!jpeg_v4_0_is_idle(handle))
+		if (!jpeg_v4_0_is_idle(ip_block))
 			return -EBUSY;
 		jpeg_v4_0_enable_clock_gating(adev);
 	} else {
@@ -637,10 +671,10 @@ static int jpeg_v4_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int jpeg_v4_0_set_powergating_state(void *handle,
+static int jpeg_v4_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	if (amdgpu_sriov_vf(adev)) {
@@ -662,9 +696,9 @@ static int jpeg_v4_0_set_powergating_state(void *handle,
 	return ret;
 }
 
-static int jpeg_v4_0_set_interrupt_state(struct amdgpu_device *adev,
+static int jpeg_v4_0_set_ras_interrupt_state(struct amdgpu_device *adev,
 					struct amdgpu_irq_src *source,
-					unsigned type,
+					unsigned int type,
 					enum amdgpu_interrupt_state state)
 {
 	return 0;
@@ -678,11 +712,7 @@ static int jpeg_v4_0_process_interrupt(struct amdgpu_device *adev,
 
 	switch (entry->src_id) {
 	case VCN_4_0__SRCID__JPEG_DECODE:
-		amdgpu_fence_process(&adev->jpeg.inst->ring_dec);
-		break;
-	case VCN_4_0__SRCID_DJPEG0_POISON:
-	case VCN_4_0__SRCID_EJPEG0_POISON:
-		amdgpu_jpeg_process_poison_irq(adev, source, entry);
+		amdgpu_fence_process(adev->jpeg.inst->ring_dec);
 		break;
 	default:
 		DRM_DEV_ERROR(adev->dev, "Unhandled interrupt: %d %d\n",
@@ -693,10 +723,25 @@ static int jpeg_v4_0_process_interrupt(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int jpeg_v4_0_ring_reset(struct amdgpu_ring *ring,
+				unsigned int vmid,
+				struct amdgpu_fence *timedout_fence)
+{
+	int r;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+	r = jpeg_v4_0_stop(ring->adev);
+	if (r)
+		return r;
+	r = jpeg_v4_0_start(ring->adev);
+	if (r)
+		return r;
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
 static const struct amd_ip_funcs jpeg_v4_0_ip_funcs = {
 	.name = "jpeg_v4_0",
 	.early_init = jpeg_v4_0_early_init,
-	.late_init = NULL,
 	.sw_init = jpeg_v4_0_sw_init,
 	.sw_fini = jpeg_v4_0_sw_fini,
 	.hw_init = jpeg_v4_0_hw_init,
@@ -705,12 +750,10 @@ static const struct amd_ip_funcs jpeg_v4_0_ip_funcs = {
 	.resume = jpeg_v4_0_resume,
 	.is_idle = jpeg_v4_0_is_idle,
 	.wait_for_idle = jpeg_v4_0_wait_for_idle,
-	.check_soft_reset = NULL,
-	.pre_soft_reset = NULL,
-	.soft_reset = NULL,
-	.post_soft_reset = NULL,
 	.set_clockgating_state = jpeg_v4_0_set_clockgating_state,
 	.set_powergating_state = jpeg_v4_0_set_powergating_state,
+	.dump_ip_state = amdgpu_jpeg_dump_ip_state,
+	.print_ip_state = amdgpu_jpeg_print_ip_state,
 };
 
 static const struct amdgpu_ring_funcs jpeg_v4_0_dec_ring_vm_funcs = {
@@ -719,6 +762,7 @@ static const struct amdgpu_ring_funcs jpeg_v4_0_dec_ring_vm_funcs = {
 	.get_rptr = jpeg_v4_0_dec_ring_get_rptr,
 	.get_wptr = jpeg_v4_0_dec_ring_get_wptr,
 	.set_wptr = jpeg_v4_0_dec_ring_set_wptr,
+	.parse_cs = amdgpu_jpeg_dec_parse_cs,
 	.emit_frame_size =
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 +
@@ -740,23 +784,30 @@ static const struct amdgpu_ring_funcs jpeg_v4_0_dec_ring_vm_funcs = {
 	.emit_wreg = jpeg_v2_0_dec_ring_emit_wreg,
 	.emit_reg_wait = jpeg_v2_0_dec_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = jpeg_v4_0_ring_reset,
 };
 
 static void jpeg_v4_0_set_dec_ring_funcs(struct amdgpu_device *adev)
 {
-	adev->jpeg.inst->ring_dec.funcs = &jpeg_v4_0_dec_ring_vm_funcs;
-	DRM_DEV_INFO(adev->dev, "JPEG decode is enabled in VM mode\n");
+	adev->jpeg.inst->ring_dec->funcs = &jpeg_v4_0_dec_ring_vm_funcs;
 }
 
 static const struct amdgpu_irq_src_funcs jpeg_v4_0_irq_funcs = {
-	.set = jpeg_v4_0_set_interrupt_state,
 	.process = jpeg_v4_0_process_interrupt,
 };
 
+static const struct amdgpu_irq_src_funcs jpeg_v4_0_ras_irq_funcs = {
+	.set = jpeg_v4_0_set_ras_interrupt_state,
+	.process = amdgpu_jpeg_process_poison_irq,
+};
+
 static void jpeg_v4_0_set_irq_funcs(struct amdgpu_device *adev)
 {
 	adev->jpeg.inst->irq.num_types = 1;
 	adev->jpeg.inst->irq.funcs = &jpeg_v4_0_irq_funcs;
+
+	adev->jpeg.inst->ras_poison_irq.num_types = 1;
+	adev->jpeg.inst->ras_poison_irq.funcs = &jpeg_v4_0_ras_irq_funcs;
 }
 
 const struct amdgpu_ip_block_version jpeg_v4_0_ip_block = {
@@ -811,12 +862,13 @@ const struct amdgpu_ras_block_hw_ops jpeg_v4_0_ras_hw_ops = {
 static struct amdgpu_jpeg_ras jpeg_v4_0_ras = {
 	.ras_block = {
 		.hw_ops = &jpeg_v4_0_ras_hw_ops,
+		.ras_late_init = amdgpu_jpeg_ras_late_init,
 	},
 };
 
 static void jpeg_v4_0_set_ras_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[JPEG_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, JPEG_HWIP, 0)) {
 	case IP_VERSION(4, 0, 0):
 		adev->jpeg.ras = &jpeg_v4_0_ras;
 		break;
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.h b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.h
index 07d36c2abd6b..47638fd4d4e2 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.h
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.h
@@ -32,5 +32,4 @@ enum amdgpu_jpeg_v4_0_sub_block {
 };
 
 extern const struct amdgpu_ip_block_version jpeg_v4_0_ip_block;
-
 #endif /* __JPEG_V4_0_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c
new file mode 100644
index 000000000000..aae7328973d1
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c
@@ -0,0 +1,1486 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "amdgpu.h"
+#include "amdgpu_jpeg.h"
+#include "soc15.h"
+#include "soc15d.h"
+#include "jpeg_v2_0.h"
+#include "jpeg_v4_0_3.h"
+#include "mmsch_v4_0_3.h"
+
+#include "vcn/vcn_4_0_3_offset.h"
+#include "vcn/vcn_4_0_3_sh_mask.h"
+#include "ivsrcid/vcn/irqsrcs_vcn_4_0.h"
+
+#define NORMALIZE_JPEG_REG_OFFSET(offset) \
+		(offset & 0x1FFFF)
+
+enum jpeg_engin_status {
+	UVD_PGFSM_STATUS__UVDJ_PWR_ON  = 0,
+	UVD_PGFSM_STATUS__UVDJ_PWR_OFF = 2,
+};
+
+static void jpeg_v4_0_3_set_dec_ring_funcs(struct amdgpu_device *adev);
+static void jpeg_v4_0_3_set_irq_funcs(struct amdgpu_device *adev);
+static int jpeg_v4_0_3_set_powergating_state(struct amdgpu_ip_block *ip_block,
+				enum amd_powergating_state state);
+static void jpeg_v4_0_3_set_ras_funcs(struct amdgpu_device *adev);
+static void jpeg_v4_0_3_dec_ring_set_wptr(struct amdgpu_ring *ring);
+
+static int amdgpu_ih_srcid_jpeg[] = {
+	VCN_4_0__SRCID__JPEG_DECODE,
+	VCN_4_0__SRCID__JPEG1_DECODE,
+	VCN_4_0__SRCID__JPEG2_DECODE,
+	VCN_4_0__SRCID__JPEG3_DECODE,
+	VCN_4_0__SRCID__JPEG4_DECODE,
+	VCN_4_0__SRCID__JPEG5_DECODE,
+	VCN_4_0__SRCID__JPEG6_DECODE,
+	VCN_4_0__SRCID__JPEG7_DECODE
+};
+
+static const struct amdgpu_hwip_reg_entry jpeg_reg_list_4_0_3[] = {
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_POWER_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_INT_STAT),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_SYS_INT_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC0_UVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC0_UVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC0_UVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_ADDR_MODE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_GFX10_ADDR_CONFIG),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_Y_GFX10_TILING_SURFACE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_UV_GFX10_TILING_SURFACE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_PITCH),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_UV_PITCH),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC1_UVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC1_UVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC1_UVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC2_UVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC2_UVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC2_UVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC3_UVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC3_UVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC3_UVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC4_UVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC4_UVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC4_UVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC5_UVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC5_UVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC5_UVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC6_UVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC6_UVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC6_UVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC7_UVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC7_UVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC7_UVD_JRBC_STATUS),
+};
+
+static inline bool jpeg_v4_0_3_normalizn_reqd(struct amdgpu_device *adev)
+{
+	return (adev->jpeg.caps & AMDGPU_JPEG_CAPS(RRMT_ENABLED)) == 0;
+}
+
+static inline int jpeg_v4_0_3_core_reg_offset(u32 pipe)
+{
+	if (pipe)
+		return ((0x40 * pipe) - 0xc80);
+	else
+		return 0;
+}
+
+/**
+ * jpeg_v4_0_3_early_init - set function pointers
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Set ring and irq function pointers
+ */
+static int jpeg_v4_0_3_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	adev->jpeg.num_jpeg_rings = AMDGPU_MAX_JPEG_RINGS_4_0_3;
+
+	jpeg_v4_0_3_set_dec_ring_funcs(adev);
+	jpeg_v4_0_3_set_irq_funcs(adev);
+	jpeg_v4_0_3_set_ras_funcs(adev);
+
+	return 0;
+}
+
+/**
+ * jpeg_v4_0_3_sw_init - sw init for JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Load firmware and sw initialization
+ */
+static int jpeg_v4_0_3_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring;
+	int i, j, r, jpeg_inst;
+
+	for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+		/* JPEG TRAP */
+		r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
+				amdgpu_ih_srcid_jpeg[j], &adev->jpeg.inst->irq);
+		if (r)
+			return r;
+	}
+
+	/* JPEG DJPEG POISON EVENT */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
+			VCN_4_0__SRCID_DJPEG0_POISON, &adev->jpeg.inst->ras_poison_irq);
+	if (r)
+		return r;
+
+	/* JPEG EJPEG POISON EVENT */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
+			VCN_4_0__SRCID_EJPEG0_POISON, &adev->jpeg.inst->ras_poison_irq);
+	if (r)
+		return r;
+
+	r = amdgpu_jpeg_sw_init(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_jpeg_resume(adev);
+	if (r)
+		return r;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		jpeg_inst = GET_INST(JPEG, i);
+
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+			ring = &adev->jpeg.inst[i].ring_dec[j];
+			ring->use_doorbell = true;
+			ring->vm_hub = AMDGPU_MMHUB0(adev->jpeg.inst[i].aid_id);
+			if (!amdgpu_sriov_vf(adev)) {
+				ring->doorbell_index =
+					(adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
+					1 + j + 9 * jpeg_inst;
+			} else {
+				if (j < 4)
+					ring->doorbell_index =
+						(adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
+						4 + j + 32 * jpeg_inst;
+				else
+					ring->doorbell_index =
+						(adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
+						8 + j + 32 * jpeg_inst;
+			}
+			sprintf(ring->name, "jpeg_dec_%d.%d", adev->jpeg.inst[i].aid_id, j);
+			r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq, 0,
+						AMDGPU_RING_PRIO_DEFAULT, NULL);
+			if (r)
+				return r;
+
+			adev->jpeg.internal.jpeg_pitch[j] =
+				regUVD_JRBC0_UVD_JRBC_SCRATCH0_INTERNAL_OFFSET;
+			adev->jpeg.inst[i].external.jpeg_pitch[j] =
+				SOC15_REG_OFFSET1(JPEG, jpeg_inst, regUVD_JRBC0_UVD_JRBC_SCRATCH0,
+						  jpeg_v4_0_3_core_reg_offset(j));
+		}
+	}
+
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__JPEG)) {
+		r = amdgpu_jpeg_ras_sw_init(adev);
+		if (r) {
+			dev_err(adev->dev, "Failed to initialize jpeg ras block!\n");
+			return r;
+		}
+	}
+
+	r = amdgpu_jpeg_reg_dump_init(adev, jpeg_reg_list_4_0_3, ARRAY_SIZE(jpeg_reg_list_4_0_3));
+	if (r)
+		return r;
+
+	adev->jpeg.supported_reset =
+		amdgpu_get_soft_full_reset_mask(adev->jpeg.inst[0].ring_dec);
+	if (!amdgpu_sriov_vf(adev))
+		adev->jpeg.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+	r = amdgpu_jpeg_sysfs_reset_mask_init(adev);
+
+	return 0;
+}
+
+/**
+ * jpeg_v4_0_3_sw_fini - sw fini for JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * JPEG suspend and free up sw allocation
+ */
+static int jpeg_v4_0_3_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = amdgpu_jpeg_suspend(adev);
+	if (r)
+		return r;
+
+	amdgpu_jpeg_sysfs_reset_mask_fini(adev);
+
+	r = amdgpu_jpeg_sw_fini(adev);
+
+	return r;
+}
+
+static int jpeg_v4_0_3_start_sriov(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	uint64_t ctx_addr;
+	uint32_t param, resp, expected;
+	uint32_t tmp, timeout;
+
+	struct amdgpu_mm_table *table = &adev->virt.mm_table;
+	uint32_t *table_loc;
+	uint32_t table_size;
+	uint32_t size, size_dw, item_offset;
+	uint32_t init_status;
+	int i, j, jpeg_inst;
+
+	struct mmsch_v4_0_cmd_direct_write
+		direct_wt = { {0} };
+	struct mmsch_v4_0_cmd_end end = { {0} };
+	struct mmsch_v4_0_3_init_header header;
+
+	direct_wt.cmd_header.command_type =
+		MMSCH_COMMAND__DIRECT_REG_WRITE;
+	end.cmd_header.command_type =
+		MMSCH_COMMAND__END;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; i++) {
+		jpeg_inst = GET_INST(JPEG, i);
+
+		memset(&header, 0, sizeof(struct mmsch_v4_0_3_init_header));
+		header.version = MMSCH_VERSION;
+		header.total_size = sizeof(struct mmsch_v4_0_3_init_header) >> 2;
+
+		table_loc = (uint32_t *)table->cpu_addr;
+		table_loc += header.total_size;
+
+		item_offset = header.total_size;
+
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; j++) {
+			ring = &adev->jpeg.inst[i].ring_dec[j];
+			table_size = 0;
+
+			tmp = SOC15_REG_OFFSET(JPEG, 0, regUVD_JMI0_UVD_LMI_JRBC_RB_64BIT_BAR_LOW);
+			MMSCH_V4_0_INSERT_DIRECT_WT(tmp, lower_32_bits(ring->gpu_addr));
+			tmp = SOC15_REG_OFFSET(JPEG, 0, regUVD_JMI0_UVD_LMI_JRBC_RB_64BIT_BAR_HIGH);
+			MMSCH_V4_0_INSERT_DIRECT_WT(tmp, upper_32_bits(ring->gpu_addr));
+			tmp = SOC15_REG_OFFSET(JPEG, 0, regUVD_JRBC0_UVD_JRBC_RB_SIZE);
+			MMSCH_V4_0_INSERT_DIRECT_WT(tmp, ring->ring_size / 4);
+
+			if (j <= 3) {
+				header.mjpegdec0[j].table_offset = item_offset;
+				header.mjpegdec0[j].init_status = 0;
+				header.mjpegdec0[j].table_size = table_size;
+			} else {
+				header.mjpegdec1[j - 4].table_offset = item_offset;
+				header.mjpegdec1[j - 4].init_status = 0;
+				header.mjpegdec1[j - 4].table_size = table_size;
+			}
+			header.total_size += table_size;
+			item_offset += table_size;
+		}
+
+		MMSCH_V4_0_INSERT_END();
+
+		/* send init table to MMSCH */
+		size = sizeof(struct mmsch_v4_0_3_init_header);
+		table_loc = (uint32_t *)table->cpu_addr;
+		memcpy((void *)table_loc, &header, size);
+
+		ctx_addr = table->gpu_addr;
+		WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_CTX_ADDR_LO, lower_32_bits(ctx_addr));
+		WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_CTX_ADDR_HI, upper_32_bits(ctx_addr));
+
+		tmp = RREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_VMID);
+		tmp &= ~MMSCH_VF_VMID__VF_CTX_VMID_MASK;
+		tmp |= (0 << MMSCH_VF_VMID__VF_CTX_VMID__SHIFT);
+		WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_VMID, tmp);
+
+		size = header.total_size;
+		WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_CTX_SIZE, size);
+
+		WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_MAILBOX_RESP, 0);
+
+		param = 0x00000001;
+		WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_MAILBOX_HOST, param);
+		tmp = 0;
+		timeout = 1000;
+		resp = 0;
+		expected = MMSCH_VF_MAILBOX_RESP__OK;
+		init_status =
+			((struct mmsch_v4_0_3_init_header *)(table_loc))->mjpegdec0[i].init_status;
+		while (resp != expected) {
+			resp = RREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_MAILBOX_RESP);
+
+			if (resp != 0)
+				break;
+			udelay(10);
+			tmp = tmp + 10;
+			if (tmp >= timeout) {
+				DRM_ERROR("failed to init MMSCH. TIME-OUT after %d usec"\
+					" waiting for regMMSCH_VF_MAILBOX_RESP "\
+					"(expected=0x%08x, readback=0x%08x)\n",
+					tmp, expected, resp);
+				return -EBUSY;
+			}
+		}
+		if (resp != expected && resp != MMSCH_VF_MAILBOX_RESP__INCOMPLETE &&
+				init_status != MMSCH_VF_ENGINE_STATUS__PASS)
+			DRM_ERROR("MMSCH init status is incorrect! readback=0x%08x, header init status for jpeg: %x\n",
+					resp, init_status);
+
+	}
+	return 0;
+}
+
+/**
+ * jpeg_v4_0_3_hw_init - start and test JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ */
+static int jpeg_v4_0_3_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring;
+	int i, j, r, jpeg_inst;
+
+	if (amdgpu_sriov_vf(adev)) {
+		r = jpeg_v4_0_3_start_sriov(adev);
+		if (r)
+			return r;
+
+		for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+			for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+				ring = &adev->jpeg.inst[i].ring_dec[j];
+				ring->wptr = 0;
+				ring->wptr_old = 0;
+				jpeg_v4_0_3_dec_ring_set_wptr(ring);
+				ring->sched.ready = true;
+			}
+		}
+	} else {
+		/* This flag is not set for VF, assumed to be disabled always */
+		if (RREG32_SOC15(VCN, GET_INST(VCN, 0), regVCN_RRMT_CNTL) &
+		    0x100)
+			adev->jpeg.caps |= AMDGPU_JPEG_CAPS(RRMT_ENABLED);
+
+		for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+			jpeg_inst = GET_INST(JPEG, i);
+
+			ring = adev->jpeg.inst[i].ring_dec;
+
+			if (ring->use_doorbell)
+				adev->nbio.funcs->vcn_doorbell_range(
+					adev, ring->use_doorbell,
+					(adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
+						9 * jpeg_inst,
+					adev->jpeg.inst[i].aid_id);
+
+			for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+				ring = &adev->jpeg.inst[i].ring_dec[j];
+				if (ring->use_doorbell)
+					WREG32_SOC15_OFFSET(
+						VCN, GET_INST(VCN, i),
+						regVCN_JPEG_DB_CTRL,
+						(ring->pipe ? (ring->pipe - 0x15) : 0),
+						ring->doorbell_index
+							<< VCN_JPEG_DB_CTRL__OFFSET__SHIFT |
+							VCN_JPEG_DB_CTRL__EN_MASK);
+				r = amdgpu_ring_test_helper(ring);
+				if (r)
+					return r;
+			}
+		}
+	}
+
+	return 0;
+}
+
+/**
+ * jpeg_v4_0_3_hw_fini - stop the hardware block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Stop the JPEG block, mark ring as not ready any more
+ */
+static int jpeg_v4_0_3_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int ret = 0;
+
+	cancel_delayed_work_sync(&adev->jpeg.idle_work);
+
+	if (!amdgpu_sriov_vf(adev)) {
+		if (adev->jpeg.cur_state != AMD_PG_STATE_GATE)
+			ret = jpeg_v4_0_3_set_powergating_state(ip_block, AMD_PG_STATE_GATE);
+	}
+
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__JPEG) && !amdgpu_sriov_vf(adev))
+		amdgpu_irq_put(adev, &adev->jpeg.inst->ras_poison_irq, 0);
+
+	return ret;
+}
+
+/**
+ * jpeg_v4_0_3_suspend - suspend JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * HW fini and suspend JPEG block
+ */
+static int jpeg_v4_0_3_suspend(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+
+	r = jpeg_v4_0_3_hw_fini(ip_block);
+	if (r)
+		return r;
+
+	r = amdgpu_jpeg_suspend(ip_block->adev);
+
+	return r;
+}
+
+/**
+ * jpeg_v4_0_3_resume - resume JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Resume firmware and hw init JPEG block
+ */
+static int jpeg_v4_0_3_resume(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+
+	r = amdgpu_jpeg_resume(ip_block->adev);
+	if (r)
+		return r;
+
+	r = jpeg_v4_0_3_hw_init(ip_block);
+
+	return r;
+}
+
+static void jpeg_v4_0_3_disable_clock_gating(struct amdgpu_device *adev, int inst_idx)
+{
+	int i, jpeg_inst;
+	uint32_t data;
+
+	jpeg_inst = GET_INST(JPEG, inst_idx);
+	data = RREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_CTRL);
+	if (adev->cg_flags & AMD_CG_SUPPORT_JPEG_MGCG) {
+		data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+		data &= (~(JPEG_CGC_CTRL__JPEG0_DEC_MODE_MASK << 1));
+	} else {
+		data &= ~JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+	}
+
+	data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
+	data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
+	WREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_CTRL, data);
+
+	data = RREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_GATE);
+	data &= ~(JPEG_CGC_GATE__JMCIF_MASK | JPEG_CGC_GATE__JRBBM_MASK);
+	for (i = 0; i < adev->jpeg.num_jpeg_rings; ++i)
+		data &= ~(JPEG_CGC_GATE__JPEG0_DEC_MASK << i);
+	WREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_GATE, data);
+}
+
+static void jpeg_v4_0_3_enable_clock_gating(struct amdgpu_device *adev, int inst_idx)
+{
+	int i, jpeg_inst;
+	uint32_t data;
+
+	jpeg_inst = GET_INST(JPEG, inst_idx);
+	data = RREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_CTRL);
+	if (adev->cg_flags & AMD_CG_SUPPORT_JPEG_MGCG) {
+		data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+		data |= (JPEG_CGC_CTRL__JPEG0_DEC_MODE_MASK << 1);
+	} else {
+		data &= ~JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+	}
+
+	data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
+	data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
+	WREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_CTRL, data);
+
+	data = RREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_GATE);
+	data |= (JPEG_CGC_GATE__JMCIF_MASK | JPEG_CGC_GATE__JRBBM_MASK);
+	for (i = 0; i < adev->jpeg.num_jpeg_rings; ++i)
+		data |= (JPEG_CGC_GATE__JPEG0_DEC_MASK << i);
+	WREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_GATE, data);
+}
+
+static void jpeg_v4_0_3_start_inst(struct amdgpu_device *adev, int inst)
+{
+	int jpeg_inst = GET_INST(JPEG, inst);
+
+	WREG32_SOC15(JPEG, jpeg_inst, regUVD_PGFSM_CONFIG,
+		     1 << UVD_PGFSM_CONFIG__UVDJ_PWR_CONFIG__SHIFT);
+	SOC15_WAIT_ON_RREG(JPEG, jpeg_inst, regUVD_PGFSM_STATUS,
+			   UVD_PGFSM_STATUS__UVDJ_PWR_ON <<
+			   UVD_PGFSM_STATUS__UVDJ_PWR_STATUS__SHIFT,
+			   UVD_PGFSM_STATUS__UVDJ_PWR_STATUS_MASK);
+
+	/* disable anti hang mechanism */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst, regUVD_JPEG_POWER_STATUS),
+		 0, ~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK);
+
+	/* JPEG disable CGC */
+	jpeg_v4_0_3_disable_clock_gating(adev, inst);
+
+	/* MJPEG global tiling registers */
+	WREG32_SOC15(JPEG, jpeg_inst, regJPEG_DEC_GFX8_ADDR_CONFIG,
+		     adev->gfx.config.gb_addr_config);
+	WREG32_SOC15(JPEG, jpeg_inst, regJPEG_DEC_GFX10_ADDR_CONFIG,
+		     adev->gfx.config.gb_addr_config);
+
+	/* enable JMI channel */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst, regUVD_JMI_CNTL), 0,
+		 ~UVD_JMI_CNTL__SOFT_RESET_MASK);
+}
+
+static void jpeg_v4_0_3_start_jrbc(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int jpeg_inst = GET_INST(JPEG, ring->me);
+	int reg_offset = jpeg_v4_0_3_core_reg_offset(ring->pipe);
+
+	/* enable System Interrupt for JRBC */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst, regJPEG_SYS_INT_EN),
+		 JPEG_SYS_INT_EN__DJRBC0_MASK << ring->pipe,
+		 ~(JPEG_SYS_INT_EN__DJRBC0_MASK << ring->pipe));
+
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JMI0_UVD_LMI_JRBC_RB_VMID,
+			    reg_offset, 0);
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JRBC0_UVD_JRBC_RB_CNTL,
+			    reg_offset,
+			    (0x00000001L | 0x00000002L));
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JMI0_UVD_LMI_JRBC_RB_64BIT_BAR_LOW,
+			    reg_offset, lower_32_bits(ring->gpu_addr));
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JMI0_UVD_LMI_JRBC_RB_64BIT_BAR_HIGH,
+			    reg_offset, upper_32_bits(ring->gpu_addr));
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JRBC0_UVD_JRBC_RB_RPTR,
+			    reg_offset, 0);
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JRBC0_UVD_JRBC_RB_WPTR,
+			    reg_offset, 0);
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JRBC0_UVD_JRBC_RB_CNTL,
+			    reg_offset, 0x00000002L);
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JRBC0_UVD_JRBC_RB_SIZE,
+			    reg_offset, ring->ring_size / 4);
+	ring->wptr = RREG32_SOC15_OFFSET(JPEG, jpeg_inst, regUVD_JRBC0_UVD_JRBC_RB_WPTR,
+					 reg_offset);
+}
+
+/**
+ * jpeg_v4_0_3_start - start JPEG block
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Setup and start the JPEG block
+ */
+static int jpeg_v4_0_3_start(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	int i, j;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		jpeg_v4_0_3_start_inst(adev, i);
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+			ring = &adev->jpeg.inst[i].ring_dec[j];
+			jpeg_v4_0_3_start_jrbc(ring);
+		}
+	}
+
+	return 0;
+}
+
+static void jpeg_v4_0_3_stop_inst(struct amdgpu_device *adev, int inst)
+{
+	int jpeg_inst = GET_INST(JPEG, inst);
+	/* reset JMI */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst, regUVD_JMI_CNTL),
+		 UVD_JMI_CNTL__SOFT_RESET_MASK,
+		 ~UVD_JMI_CNTL__SOFT_RESET_MASK);
+
+	jpeg_v4_0_3_enable_clock_gating(adev, inst);
+
+	/* enable anti hang mechanism */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst, regUVD_JPEG_POWER_STATUS),
+		 UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK,
+		 ~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK);
+
+}
+
+/**
+ * jpeg_v4_0_3_stop - stop JPEG block
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * stop the JPEG block
+ */
+static int jpeg_v4_0_3_stop(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i)
+		jpeg_v4_0_3_stop_inst(adev, i);
+
+	return 0;
+}
+
+/**
+ * jpeg_v4_0_3_dec_ring_get_rptr - get read pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware read pointer
+ */
+static uint64_t jpeg_v4_0_3_dec_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	return RREG32_SOC15_OFFSET(JPEG, GET_INST(JPEG, ring->me), regUVD_JRBC0_UVD_JRBC_RB_RPTR,
+				   jpeg_v4_0_3_core_reg_offset(ring->pipe));
+}
+
+/**
+ * jpeg_v4_0_3_dec_ring_get_wptr - get write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware write pointer
+ */
+static uint64_t jpeg_v4_0_3_dec_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->use_doorbell)
+		return adev->wb.wb[ring->wptr_offs];
+
+	return RREG32_SOC15_OFFSET(JPEG, GET_INST(JPEG, ring->me), regUVD_JRBC0_UVD_JRBC_RB_WPTR,
+				   jpeg_v4_0_3_core_reg_offset(ring->pipe));
+}
+
+void jpeg_v4_0_3_ring_emit_hdp_flush(struct amdgpu_ring *ring)
+{
+	/* JPEG engine access for HDP flush doesn't work when RRMT is enabled.
+	 * This is a workaround to avoid any HDP flush through JPEG ring.
+	 */
+}
+
+/**
+ * jpeg_v4_0_3_dec_ring_set_wptr - set write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Commits the write pointer to the hardware
+ */
+static void jpeg_v4_0_3_dec_ring_set_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->use_doorbell) {
+		adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr);
+		WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr));
+	} else {
+		WREG32_SOC15_OFFSET(JPEG, GET_INST(JPEG, ring->me), regUVD_JRBC0_UVD_JRBC_RB_WPTR,
+				    jpeg_v4_0_3_core_reg_offset(ring->pipe),
+				    lower_32_bits(ring->wptr));
+	}
+}
+
+/**
+ * jpeg_v4_0_3_dec_ring_insert_start - insert a start command
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Write a start command to the ring.
+ */
+void jpeg_v4_0_3_dec_ring_insert_start(struct amdgpu_ring *ring)
+{
+	if (!amdgpu_sriov_vf(ring->adev)) {
+		amdgpu_ring_write(ring, PACKETJ(regUVD_JRBC_EXTERNAL_REG_INTERNAL_OFFSET,
+			0, 0, PACKETJ_TYPE0));
+		amdgpu_ring_write(ring, 0x62a04); /* PCTL0_MMHUB_DEEPSLEEP_IB */
+
+		amdgpu_ring_write(ring,
+				  PACKETJ(JRBC_DEC_EXTERNAL_REG_WRITE_ADDR, 0,
+					  0, PACKETJ_TYPE0));
+		amdgpu_ring_write(ring, 0x80004000);
+	}
+}
+
+/**
+ * jpeg_v4_0_3_dec_ring_insert_end - insert a end command
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Write a end command to the ring.
+ */
+void jpeg_v4_0_3_dec_ring_insert_end(struct amdgpu_ring *ring)
+{
+	if (!amdgpu_sriov_vf(ring->adev)) {
+		amdgpu_ring_write(ring, PACKETJ(regUVD_JRBC_EXTERNAL_REG_INTERNAL_OFFSET,
+			0, 0, PACKETJ_TYPE0));
+		amdgpu_ring_write(ring, 0x62a04);
+
+		amdgpu_ring_write(ring,
+				  PACKETJ(JRBC_DEC_EXTERNAL_REG_WRITE_ADDR, 0,
+					  0, PACKETJ_TYPE0));
+		amdgpu_ring_write(ring, 0x00004000);
+	}
+}
+
+/**
+ * jpeg_v4_0_3_dec_ring_emit_fence - emit an fence & trap command
+ *
+ * @ring: amdgpu_ring pointer
+ * @addr: address
+ * @seq: sequence number
+ * @flags: fence related flags
+ *
+ * Write a fence and a trap command to the ring.
+ */
+void jpeg_v4_0_3_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq,
+				unsigned int flags)
+{
+	WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
+
+	amdgpu_ring_write(ring, PACKETJ(regUVD_JPEG_GPCOM_DATA0_INTERNAL_OFFSET,
+		0, 0, PACKETJ_TYPE0));
+	amdgpu_ring_write(ring, seq);
+
+	amdgpu_ring_write(ring,	PACKETJ(regUVD_JPEG_GPCOM_DATA1_INTERNAL_OFFSET,
+		0, 0, PACKETJ_TYPE0));
+	amdgpu_ring_write(ring, seq);
+
+	amdgpu_ring_write(ring,	PACKETJ(regUVD_LMI_JRBC_RB_MEM_WR_64BIT_BAR_LOW_INTERNAL_OFFSET,
+		0, 0, PACKETJ_TYPE0));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+
+	amdgpu_ring_write(ring,	PACKETJ(regUVD_LMI_JRBC_RB_MEM_WR_64BIT_BAR_HIGH_INTERNAL_OFFSET,
+		0, 0, PACKETJ_TYPE0));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+
+	amdgpu_ring_write(ring,	PACKETJ(regUVD_JPEG_GPCOM_CMD_INTERNAL_OFFSET,
+		0, 0, PACKETJ_TYPE0));
+	amdgpu_ring_write(ring, 0x8);
+
+	amdgpu_ring_write(ring,	PACKETJ(regUVD_JPEG_GPCOM_CMD_INTERNAL_OFFSET,
+		0, PACKETJ_CONDITION_CHECK0, PACKETJ_TYPE4));
+	amdgpu_ring_write(ring, 0);
+
+	amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE6));
+	amdgpu_ring_write(ring, 0);
+
+	amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE6));
+	amdgpu_ring_write(ring, 0);
+
+	amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE7));
+	amdgpu_ring_write(ring, 0);
+}
+
+/**
+ * jpeg_v4_0_3_dec_ring_emit_ib - execute indirect buffer
+ *
+ * @ring: amdgpu_ring pointer
+ * @job: job to retrieve vmid from
+ * @ib: indirect buffer to execute
+ * @flags: unused
+ *
+ * Write ring commands to execute the indirect buffer.
+ */
+void jpeg_v4_0_3_dec_ring_emit_ib(struct amdgpu_ring *ring,
+				struct amdgpu_job *job,
+				struct amdgpu_ib *ib,
+				uint32_t flags)
+{
+	unsigned int vmid = AMDGPU_JOB_GET_VMID(job);
+
+	amdgpu_ring_write(ring, PACKETJ(regUVD_LMI_JRBC_IB_VMID_INTERNAL_OFFSET,
+		0, 0, PACKETJ_TYPE0));
+
+	if (ring->funcs->parse_cs)
+		amdgpu_ring_write(ring, 0);
+	else
+		amdgpu_ring_write(ring, (vmid | (vmid << 4) | (vmid << 8)));
+
+	amdgpu_ring_write(ring, PACKETJ(regUVD_LMI_JPEG_VMID_INTERNAL_OFFSET,
+		0, 0, PACKETJ_TYPE0));
+	amdgpu_ring_write(ring, (vmid | (vmid << 4) | (vmid << 8)));
+
+	amdgpu_ring_write(ring,	PACKETJ(regUVD_LMI_JRBC_IB_64BIT_BAR_LOW_INTERNAL_OFFSET,
+		0, 0, PACKETJ_TYPE0));
+	amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));
+
+	amdgpu_ring_write(ring,	PACKETJ(regUVD_LMI_JRBC_IB_64BIT_BAR_HIGH_INTERNAL_OFFSET,
+		0, 0, PACKETJ_TYPE0));
+	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
+
+	amdgpu_ring_write(ring,	PACKETJ(regUVD_JRBC_IB_SIZE_INTERNAL_OFFSET,
+		0, 0, PACKETJ_TYPE0));
+	amdgpu_ring_write(ring, ib->length_dw);
+
+	amdgpu_ring_write(ring,	PACKETJ(regUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_LOW_INTERNAL_OFFSET,
+		0, 0, PACKETJ_TYPE0));
+	amdgpu_ring_write(ring, lower_32_bits(ring->gpu_addr));
+
+	amdgpu_ring_write(ring,	PACKETJ(regUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_HIGH_INTERNAL_OFFSET,
+		0, 0, PACKETJ_TYPE0));
+	amdgpu_ring_write(ring, upper_32_bits(ring->gpu_addr));
+
+	amdgpu_ring_write(ring,	PACKETJ(0, 0, PACKETJ_CONDITION_CHECK0, PACKETJ_TYPE2));
+	amdgpu_ring_write(ring, 0);
+
+	amdgpu_ring_write(ring,	PACKETJ(regUVD_JRBC_RB_COND_RD_TIMER_INTERNAL_OFFSET,
+		0, 0, PACKETJ_TYPE0));
+	amdgpu_ring_write(ring, 0x01400200);
+
+	amdgpu_ring_write(ring, PACKETJ(regUVD_JRBC_RB_REF_DATA_INTERNAL_OFFSET,
+		0, 0, PACKETJ_TYPE0));
+	amdgpu_ring_write(ring, 0x2);
+
+	amdgpu_ring_write(ring,	PACKETJ(regUVD_JRBC_STATUS_INTERNAL_OFFSET,
+		0, PACKETJ_CONDITION_CHECK3, PACKETJ_TYPE3));
+	amdgpu_ring_write(ring, 0x2);
+}
+
+void jpeg_v4_0_3_dec_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
+				uint32_t val, uint32_t mask)
+{
+	uint32_t reg_offset;
+
+	/* Use normalized offsets if required */
+	if (jpeg_v4_0_3_normalizn_reqd(ring->adev))
+		reg = NORMALIZE_JPEG_REG_OFFSET(reg);
+
+	reg_offset = (reg << 2);
+
+	amdgpu_ring_write(ring, PACKETJ(regUVD_JRBC_RB_COND_RD_TIMER_INTERNAL_OFFSET,
+		0, 0, PACKETJ_TYPE0));
+	amdgpu_ring_write(ring, 0x01400200);
+
+	amdgpu_ring_write(ring,	PACKETJ(regUVD_JRBC_RB_REF_DATA_INTERNAL_OFFSET,
+		0, 0, PACKETJ_TYPE0));
+	amdgpu_ring_write(ring, val);
+
+	amdgpu_ring_write(ring, PACKETJ(regUVD_JRBC_EXTERNAL_REG_INTERNAL_OFFSET,
+		0, 0, PACKETJ_TYPE0));
+	if (reg_offset >= 0x10000 && reg_offset <= 0x105ff) {
+		amdgpu_ring_write(ring, 0);
+		amdgpu_ring_write(ring,
+			PACKETJ((reg_offset >> 2), 0, 0, PACKETJ_TYPE3));
+	} else {
+		amdgpu_ring_write(ring, reg_offset);
+		amdgpu_ring_write(ring,	PACKETJ(JRBC_DEC_EXTERNAL_REG_WRITE_ADDR,
+			0, 0, PACKETJ_TYPE3));
+	}
+	amdgpu_ring_write(ring, mask);
+}
+
+void jpeg_v4_0_3_dec_ring_emit_vm_flush(struct amdgpu_ring *ring,
+				unsigned int vmid, uint64_t pd_addr)
+{
+	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
+	uint32_t data0, data1, mask;
+
+	pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
+
+	/* wait for register write */
+	data0 = hub->ctx0_ptb_addr_lo32 + vmid * hub->ctx_addr_distance;
+	data1 = lower_32_bits(pd_addr);
+	mask = 0xffffffff;
+	jpeg_v4_0_3_dec_ring_emit_reg_wait(ring, data0, data1, mask);
+}
+
+void jpeg_v4_0_3_dec_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg, uint32_t val)
+{
+	uint32_t reg_offset;
+
+	/* Use normalized offsets if required */
+	if (jpeg_v4_0_3_normalizn_reqd(ring->adev))
+		reg = NORMALIZE_JPEG_REG_OFFSET(reg);
+
+	reg_offset = (reg << 2);
+
+	amdgpu_ring_write(ring,	PACKETJ(regUVD_JRBC_EXTERNAL_REG_INTERNAL_OFFSET,
+		0, 0, PACKETJ_TYPE0));
+	if (reg_offset >= 0x10000 && reg_offset <= 0x105ff) {
+		amdgpu_ring_write(ring, 0);
+		amdgpu_ring_write(ring,
+			PACKETJ((reg_offset >> 2), 0, 0, PACKETJ_TYPE0));
+	} else {
+		amdgpu_ring_write(ring, reg_offset);
+		amdgpu_ring_write(ring,	PACKETJ(JRBC_DEC_EXTERNAL_REG_WRITE_ADDR,
+			0, 0, PACKETJ_TYPE0));
+	}
+	amdgpu_ring_write(ring, val);
+}
+
+void jpeg_v4_0_3_dec_ring_nop(struct amdgpu_ring *ring, uint32_t count)
+{
+	int i;
+
+	WARN_ON(ring->wptr % 2 || count % 2);
+
+	for (i = 0; i < count / 2; i++) {
+		amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE6));
+		amdgpu_ring_write(ring, 0);
+	}
+}
+
+static bool jpeg_v4_0_3_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	bool ret = false;
+	int i, j;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+			ret &= ((RREG32_SOC15_OFFSET(JPEG, GET_INST(JPEG, i),
+				regUVD_JRBC0_UVD_JRBC_STATUS, jpeg_v4_0_3_core_reg_offset(j)) &
+				UVD_JRBC0_UVD_JRBC_STATUS__RB_JOB_DONE_MASK) ==
+				UVD_JRBC0_UVD_JRBC_STATUS__RB_JOB_DONE_MASK);
+		}
+	}
+
+	return ret;
+}
+
+static int jpeg_v4_0_3_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int ret = 0;
+	int i, j;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+			ret &= (SOC15_WAIT_ON_RREG_OFFSET(JPEG, GET_INST(JPEG, i),
+				regUVD_JRBC0_UVD_JRBC_STATUS, jpeg_v4_0_3_core_reg_offset(j),
+				UVD_JRBC0_UVD_JRBC_STATUS__RB_JOB_DONE_MASK,
+				UVD_JRBC0_UVD_JRBC_STATUS__RB_JOB_DONE_MASK));
+		}
+	}
+	return ret;
+}
+
+static int jpeg_v4_0_3_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_clockgating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	bool enable = state == AMD_CG_STATE_GATE;
+	int i;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		if (enable) {
+			if (!jpeg_v4_0_3_is_idle(ip_block))
+				return -EBUSY;
+			jpeg_v4_0_3_enable_clock_gating(adev, i);
+		} else {
+			jpeg_v4_0_3_disable_clock_gating(adev, i);
+		}
+	}
+	return 0;
+}
+
+static int jpeg_v4_0_3_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_powergating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int ret;
+
+	if (amdgpu_sriov_vf(adev)) {
+		adev->jpeg.cur_state = AMD_PG_STATE_UNGATE;
+		return 0;
+	}
+
+	if (state == adev->jpeg.cur_state)
+		return 0;
+
+	if (state == AMD_PG_STATE_GATE)
+		ret = jpeg_v4_0_3_stop(adev);
+	else
+		ret = jpeg_v4_0_3_start(adev);
+
+	if (!ret)
+		adev->jpeg.cur_state = state;
+
+	return ret;
+}
+
+static int jpeg_v4_0_3_set_interrupt_state(struct amdgpu_device *adev,
+					struct amdgpu_irq_src *source,
+					unsigned int type,
+					enum amdgpu_interrupt_state state)
+{
+	return 0;
+}
+
+static int jpeg_v4_0_3_set_ras_interrupt_state(struct amdgpu_device *adev,
+					struct amdgpu_irq_src *source,
+					unsigned int type,
+					enum amdgpu_interrupt_state state)
+{
+	return 0;
+}
+
+static int jpeg_v4_0_3_process_interrupt(struct amdgpu_device *adev,
+				      struct amdgpu_irq_src *source,
+				      struct amdgpu_iv_entry *entry)
+{
+	uint32_t i, inst;
+
+	i = node_id_to_phys_map[entry->node_id];
+	DRM_DEV_DEBUG(adev->dev, "IH: JPEG TRAP\n");
+
+	for (inst = 0; inst < adev->jpeg.num_jpeg_inst; ++inst)
+		if (adev->jpeg.inst[inst].aid_id == i)
+			break;
+
+	if (inst >= adev->jpeg.num_jpeg_inst) {
+		dev_WARN_ONCE(adev->dev, 1,
+			      "Interrupt received for unknown JPEG instance %d",
+			      entry->node_id);
+		return 0;
+	}
+
+	switch (entry->src_id) {
+	case VCN_4_0__SRCID__JPEG_DECODE:
+		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[0]);
+		break;
+	case VCN_4_0__SRCID__JPEG1_DECODE:
+		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[1]);
+		break;
+	case VCN_4_0__SRCID__JPEG2_DECODE:
+		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[2]);
+		break;
+	case VCN_4_0__SRCID__JPEG3_DECODE:
+		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[3]);
+		break;
+	case VCN_4_0__SRCID__JPEG4_DECODE:
+		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[4]);
+		break;
+	case VCN_4_0__SRCID__JPEG5_DECODE:
+		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[5]);
+		break;
+	case VCN_4_0__SRCID__JPEG6_DECODE:
+		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[6]);
+		break;
+	case VCN_4_0__SRCID__JPEG7_DECODE:
+		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[7]);
+		break;
+	default:
+		DRM_DEV_ERROR(adev->dev, "Unhandled interrupt: %d %d\n",
+			  entry->src_id, entry->src_data[0]);
+		break;
+	}
+
+	return 0;
+}
+
+static void jpeg_v4_0_3_core_stall_reset(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int jpeg_inst = GET_INST(JPEG, ring->me);
+	int reg_offset = jpeg_v4_0_3_core_reg_offset(ring->pipe);
+
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JMI0_UVD_JMI_CLIENT_STALL,
+			    reg_offset, 0x1F);
+	SOC15_WAIT_ON_RREG_OFFSET(JPEG, jpeg_inst,
+				  regUVD_JMI0_UVD_JMI_CLIENT_CLEAN_STATUS,
+				  reg_offset, 0x1F, 0x1F);
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JMI0_JPEG_LMI_DROP,
+			    reg_offset, 0x1F);
+	WREG32_SOC15(JPEG, jpeg_inst, regJPEG_CORE_RST_CTRL, 1 << ring->pipe);
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JMI0_UVD_JMI_CLIENT_STALL,
+			    reg_offset, 0x00);
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JMI0_JPEG_LMI_DROP,
+			    reg_offset, 0x00);
+	WREG32_SOC15(JPEG, jpeg_inst, regJPEG_CORE_RST_CTRL, 0x00);
+}
+
+static int jpeg_v4_0_3_ring_reset(struct amdgpu_ring *ring,
+				  unsigned int vmid,
+				  struct amdgpu_fence *timedout_fence)
+{
+	if (amdgpu_sriov_vf(ring->adev))
+		return -EOPNOTSUPP;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+	jpeg_v4_0_3_core_stall_reset(ring);
+	jpeg_v4_0_3_start_jrbc(ring);
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static const struct amd_ip_funcs jpeg_v4_0_3_ip_funcs = {
+	.name = "jpeg_v4_0_3",
+	.early_init = jpeg_v4_0_3_early_init,
+	.sw_init = jpeg_v4_0_3_sw_init,
+	.sw_fini = jpeg_v4_0_3_sw_fini,
+	.hw_init = jpeg_v4_0_3_hw_init,
+	.hw_fini = jpeg_v4_0_3_hw_fini,
+	.suspend = jpeg_v4_0_3_suspend,
+	.resume = jpeg_v4_0_3_resume,
+	.is_idle = jpeg_v4_0_3_is_idle,
+	.wait_for_idle = jpeg_v4_0_3_wait_for_idle,
+	.set_clockgating_state = jpeg_v4_0_3_set_clockgating_state,
+	.set_powergating_state = jpeg_v4_0_3_set_powergating_state,
+	.dump_ip_state = amdgpu_jpeg_dump_ip_state,
+	.print_ip_state = amdgpu_jpeg_print_ip_state,
+};
+
+static const struct amdgpu_ring_funcs jpeg_v4_0_3_dec_ring_vm_funcs = {
+	.type = AMDGPU_RING_TYPE_VCN_JPEG,
+	.align_mask = 0xf,
+	.get_rptr = jpeg_v4_0_3_dec_ring_get_rptr,
+	.get_wptr = jpeg_v4_0_3_dec_ring_get_wptr,
+	.set_wptr = jpeg_v4_0_3_dec_ring_set_wptr,
+	.parse_cs = amdgpu_jpeg_dec_parse_cs,
+	.emit_frame_size =
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 +
+		8 + /* jpeg_v4_0_3_dec_ring_emit_vm_flush */
+		18 + 18 + /* jpeg_v4_0_3_dec_ring_emit_fence x2 vm fence */
+		8 + 16,
+	.emit_ib_size = 22, /* jpeg_v4_0_3_dec_ring_emit_ib */
+	.emit_ib = jpeg_v4_0_3_dec_ring_emit_ib,
+	.emit_fence = jpeg_v4_0_3_dec_ring_emit_fence,
+	.emit_vm_flush = jpeg_v4_0_3_dec_ring_emit_vm_flush,
+	.emit_hdp_flush = jpeg_v4_0_3_ring_emit_hdp_flush,
+	.test_ring = amdgpu_jpeg_dec_ring_test_ring,
+	.test_ib = amdgpu_jpeg_dec_ring_test_ib,
+	.insert_nop = jpeg_v4_0_3_dec_ring_nop,
+	.insert_start = jpeg_v4_0_3_dec_ring_insert_start,
+	.insert_end = jpeg_v4_0_3_dec_ring_insert_end,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.begin_use = amdgpu_jpeg_ring_begin_use,
+	.end_use = amdgpu_jpeg_ring_end_use,
+	.emit_wreg = jpeg_v4_0_3_dec_ring_emit_wreg,
+	.emit_reg_wait = jpeg_v4_0_3_dec_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = jpeg_v4_0_3_ring_reset,
+};
+
+static void jpeg_v4_0_3_set_dec_ring_funcs(struct amdgpu_device *adev)
+{
+	int i, j, jpeg_inst;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+			adev->jpeg.inst[i].ring_dec[j].funcs = &jpeg_v4_0_3_dec_ring_vm_funcs;
+			adev->jpeg.inst[i].ring_dec[j].me = i;
+			adev->jpeg.inst[i].ring_dec[j].pipe = j;
+		}
+		jpeg_inst = GET_INST(JPEG, i);
+		adev->jpeg.inst[i].aid_id =
+			jpeg_inst / adev->jpeg.num_inst_per_aid;
+	}
+}
+
+static const struct amdgpu_irq_src_funcs jpeg_v4_0_3_irq_funcs = {
+	.set = jpeg_v4_0_3_set_interrupt_state,
+	.process = jpeg_v4_0_3_process_interrupt,
+};
+
+static const struct amdgpu_irq_src_funcs jpeg_v4_0_3_ras_irq_funcs = {
+	.set = jpeg_v4_0_3_set_ras_interrupt_state,
+	.process = amdgpu_jpeg_process_poison_irq,
+};
+
+static void jpeg_v4_0_3_set_irq_funcs(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		adev->jpeg.inst->irq.num_types += adev->jpeg.num_jpeg_rings;
+	}
+	adev->jpeg.inst->irq.funcs = &jpeg_v4_0_3_irq_funcs;
+
+	adev->jpeg.inst->ras_poison_irq.num_types = 1;
+	adev->jpeg.inst->ras_poison_irq.funcs = &jpeg_v4_0_3_ras_irq_funcs;
+}
+
+const struct amdgpu_ip_block_version jpeg_v4_0_3_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_JPEG,
+	.major = 4,
+	.minor = 0,
+	.rev = 3,
+	.funcs = &jpeg_v4_0_3_ip_funcs,
+};
+
+static const struct amdgpu_ras_err_status_reg_entry jpeg_v4_0_3_ue_reg_list[] = {
+	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG0S, regVCN_UE_ERR_STATUS_HI_JPEG0S),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG0S"},
+	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG0D, regVCN_UE_ERR_STATUS_HI_JPEG0D),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG0D"},
+	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG1S, regVCN_UE_ERR_STATUS_HI_JPEG1S),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG1S"},
+	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG1D, regVCN_UE_ERR_STATUS_HI_JPEG1D),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG1D"},
+	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG2S, regVCN_UE_ERR_STATUS_HI_JPEG2S),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG2S"},
+	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG2D, regVCN_UE_ERR_STATUS_HI_JPEG2D),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG2D"},
+	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG3S, regVCN_UE_ERR_STATUS_HI_JPEG3S),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG3S"},
+	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG3D, regVCN_UE_ERR_STATUS_HI_JPEG3D),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG3D"},
+	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG4S, regVCN_UE_ERR_STATUS_HI_JPEG4S),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG4S"},
+	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG4D, regVCN_UE_ERR_STATUS_HI_JPEG4D),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG4D"},
+	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG5S, regVCN_UE_ERR_STATUS_HI_JPEG5S),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG5S"},
+	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG5D, regVCN_UE_ERR_STATUS_HI_JPEG5D),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG5D"},
+	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG6S, regVCN_UE_ERR_STATUS_HI_JPEG6S),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG6S"},
+	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG6D, regVCN_UE_ERR_STATUS_HI_JPEG6D),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG6D"},
+	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG7S, regVCN_UE_ERR_STATUS_HI_JPEG7S),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG7S"},
+	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG7D, regVCN_UE_ERR_STATUS_HI_JPEG7D),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG7D"},
+};
+
+static void jpeg_v4_0_3_inst_query_ras_error_count(struct amdgpu_device *adev,
+						   uint32_t jpeg_inst,
+						   void *ras_err_status)
+{
+	struct ras_err_data *err_data = (struct ras_err_data *)ras_err_status;
+
+	/* jpeg v4_0_3 only support uncorrectable errors */
+	amdgpu_ras_inst_query_ras_error_count(adev,
+			jpeg_v4_0_3_ue_reg_list,
+			ARRAY_SIZE(jpeg_v4_0_3_ue_reg_list),
+			NULL, 0, GET_INST(VCN, jpeg_inst),
+			AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE,
+			&err_data->ue_count);
+}
+
+static void jpeg_v4_0_3_query_ras_error_count(struct amdgpu_device *adev,
+					      void *ras_err_status)
+{
+	uint32_t i;
+
+	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__JPEG)) {
+		dev_warn(adev->dev, "JPEG RAS is not supported\n");
+		return;
+	}
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; i++)
+		jpeg_v4_0_3_inst_query_ras_error_count(adev, i, ras_err_status);
+}
+
+static void jpeg_v4_0_3_inst_reset_ras_error_count(struct amdgpu_device *adev,
+						   uint32_t jpeg_inst)
+{
+	amdgpu_ras_inst_reset_ras_error_count(adev,
+			jpeg_v4_0_3_ue_reg_list,
+			ARRAY_SIZE(jpeg_v4_0_3_ue_reg_list),
+			GET_INST(VCN, jpeg_inst));
+}
+
+static void jpeg_v4_0_3_reset_ras_error_count(struct amdgpu_device *adev)
+{
+	uint32_t i;
+
+	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__JPEG)) {
+		dev_warn(adev->dev, "JPEG RAS is not supported\n");
+		return;
+	}
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; i++)
+		jpeg_v4_0_3_inst_reset_ras_error_count(adev, i);
+}
+
+static uint32_t jpeg_v4_0_3_query_poison_by_instance(struct amdgpu_device *adev,
+		uint32_t instance, uint32_t sub_block)
+{
+	uint32_t poison_stat = 0, reg_value = 0;
+
+	switch (sub_block) {
+	case AMDGPU_JPEG_V4_0_3_JPEG0:
+		reg_value = RREG32_SOC15(JPEG, instance, regUVD_RAS_JPEG0_STATUS);
+		poison_stat = REG_GET_FIELD(reg_value, UVD_RAS_JPEG0_STATUS, POISONED_PF);
+		break;
+	case AMDGPU_JPEG_V4_0_3_JPEG1:
+		reg_value = RREG32_SOC15(JPEG, instance, regUVD_RAS_JPEG1_STATUS);
+		poison_stat = REG_GET_FIELD(reg_value, UVD_RAS_JPEG1_STATUS, POISONED_PF);
+		break;
+	default:
+		break;
+	}
+
+	if (poison_stat)
+		dev_info(adev->dev, "Poison detected in JPEG%d sub_block%d\n",
+			instance, sub_block);
+
+	return poison_stat;
+}
+
+static bool jpeg_v4_0_3_query_ras_poison_status(struct amdgpu_device *adev)
+{
+	uint32_t inst = 0, sub = 0, poison_stat = 0;
+
+	for (inst = 0; inst < adev->jpeg.num_jpeg_inst; inst++)
+		for (sub = 0; sub < AMDGPU_JPEG_V4_0_3_MAX_SUB_BLOCK; sub++)
+			poison_stat +=
+			jpeg_v4_0_3_query_poison_by_instance(adev, inst, sub);
+
+	return !!poison_stat;
+}
+
+static const struct amdgpu_ras_block_hw_ops jpeg_v4_0_3_ras_hw_ops = {
+	.query_ras_error_count = jpeg_v4_0_3_query_ras_error_count,
+	.reset_ras_error_count = jpeg_v4_0_3_reset_ras_error_count,
+	.query_poison_status = jpeg_v4_0_3_query_ras_poison_status,
+};
+
+static int jpeg_v4_0_3_aca_bank_parser(struct aca_handle *handle, struct aca_bank *bank,
+				      enum aca_smu_type type, void *data)
+{
+	struct aca_bank_info info;
+	u64 misc0;
+	int ret;
+
+	ret = aca_bank_info_decode(bank, &info);
+	if (ret)
+		return ret;
+
+	misc0 = bank->regs[ACA_REG_IDX_MISC0];
+	switch (type) {
+	case ACA_SMU_TYPE_UE:
+		bank->aca_err_type = ACA_ERROR_TYPE_UE;
+		ret = aca_error_cache_log_bank_error(handle, &info, ACA_ERROR_TYPE_UE,
+						     1ULL);
+		break;
+	case ACA_SMU_TYPE_CE:
+		bank->aca_err_type = ACA_ERROR_TYPE_CE;
+		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type,
+						     ACA_REG__MISC0__ERRCNT(misc0));
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return ret;
+}
+
+/* reference to smu driver if header file */
+static int jpeg_v4_0_3_err_codes[] = {
+	16, 17, 18, 19, 20, 21, 22, 23, /* JPEG[0-7][S|D] */
+	24, 25, 26, 27, 28, 29, 30, 31
+};
+
+static bool jpeg_v4_0_3_aca_bank_is_valid(struct aca_handle *handle, struct aca_bank *bank,
+					 enum aca_smu_type type, void *data)
+{
+	u32 instlo;
+
+	instlo = ACA_REG__IPID__INSTANCEIDLO(bank->regs[ACA_REG_IDX_IPID]);
+	instlo &= GENMASK(31, 1);
+
+	if (instlo != mmSMNAID_AID0_MCA_SMU)
+		return false;
+
+	if (aca_bank_check_error_codes(handle->adev, bank,
+				       jpeg_v4_0_3_err_codes,
+				       ARRAY_SIZE(jpeg_v4_0_3_err_codes)))
+		return false;
+
+	return true;
+}
+
+static const struct aca_bank_ops jpeg_v4_0_3_aca_bank_ops = {
+	.aca_bank_parser = jpeg_v4_0_3_aca_bank_parser,
+	.aca_bank_is_valid = jpeg_v4_0_3_aca_bank_is_valid,
+};
+
+static const struct aca_info jpeg_v4_0_3_aca_info = {
+	.hwip = ACA_HWIP_TYPE_SMU,
+	.mask = ACA_ERROR_UE_MASK,
+	.bank_ops = &jpeg_v4_0_3_aca_bank_ops,
+};
+
+static int jpeg_v4_0_3_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
+{
+	int r;
+
+	r = amdgpu_ras_block_late_init(adev, ras_block);
+	if (r)
+		return r;
+
+	if (amdgpu_ras_is_supported(adev, ras_block->block) &&
+		adev->jpeg.inst->ras_poison_irq.funcs) {
+		r = amdgpu_irq_get(adev, &adev->jpeg.inst->ras_poison_irq, 0);
+		if (r)
+			goto late_fini;
+	}
+
+	r = amdgpu_ras_bind_aca(adev, AMDGPU_RAS_BLOCK__JPEG,
+				&jpeg_v4_0_3_aca_info, NULL);
+	if (r)
+		goto late_fini;
+
+	return 0;
+
+late_fini:
+	amdgpu_ras_block_late_fini(adev, ras_block);
+
+	return r;
+}
+
+static struct amdgpu_jpeg_ras jpeg_v4_0_3_ras = {
+	.ras_block = {
+		.hw_ops = &jpeg_v4_0_3_ras_hw_ops,
+		.ras_late_init = jpeg_v4_0_3_ras_late_init,
+	},
+};
+
+static void jpeg_v4_0_3_set_ras_funcs(struct amdgpu_device *adev)
+{
+	adev->jpeg.ras = &jpeg_v4_0_3_ras;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.h b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.h
new file mode 100644
index 000000000000..2e110d04af84
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.h
@@ -0,0 +1,74 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __JPEG_V4_0_3_H__
+#define __JPEG_V4_0_3_H__
+
+#define regUVD_JRBC_EXTERNAL_REG_INTERNAL_OFFSET			0x1bfff
+#define regUVD_JPEG_GPCOM_CMD_INTERNAL_OFFSET				0x404d
+#define regUVD_JPEG_GPCOM_DATA0_INTERNAL_OFFSET				0x404e
+#define regUVD_JPEG_GPCOM_DATA1_INTERNAL_OFFSET				0x404f
+#define regUVD_LMI_JRBC_RB_MEM_WR_64BIT_BAR_LOW_INTERNAL_OFFSET		0x40ab
+#define regUVD_LMI_JRBC_RB_MEM_WR_64BIT_BAR_HIGH_INTERNAL_OFFSET	0x40ac
+#define regUVD_LMI_JRBC_IB_VMID_INTERNAL_OFFSET				0x40a4
+#define regUVD_LMI_JPEG_VMID_INTERNAL_OFFSET				0x40a6
+#define regUVD_LMI_JRBC_IB_64BIT_BAR_LOW_INTERNAL_OFFSET		0x40b6
+#define regUVD_LMI_JRBC_IB_64BIT_BAR_HIGH_INTERNAL_OFFSET		0x40b7
+#define regUVD_JRBC_IB_SIZE_INTERNAL_OFFSET				0x4082
+#define regUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_LOW_INTERNAL_OFFSET		0x42d4
+#define regUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_HIGH_INTERNAL_OFFSET	0x42d5
+#define regUVD_JRBC_RB_COND_RD_TIMER_INTERNAL_OFFSET			0x4085
+#define regUVD_JRBC_RB_REF_DATA_INTERNAL_OFFSET				0x4084
+#define regUVD_JRBC_STATUS_INTERNAL_OFFSET				0x4089
+#define regUVD_JPEG_PITCH_INTERNAL_OFFSET				0x4043
+#define regUVD_JRBC0_UVD_JRBC_SCRATCH0_INTERNAL_OFFSET			0x4094
+#define regUVD_JRBC_EXTERNAL_MCM_ADDR_INTERNAL_OFFSET			0x1bffe
+
+#define JRBC_DEC_EXTERNAL_REG_WRITE_ADDR				0x18000
+
+enum amdgpu_jpeg_v4_0_3_sub_block {
+	AMDGPU_JPEG_V4_0_3_JPEG0 = 0,
+	AMDGPU_JPEG_V4_0_3_JPEG1,
+
+	AMDGPU_JPEG_V4_0_3_MAX_SUB_BLOCK,
+};
+
+extern const struct amdgpu_ip_block_version jpeg_v4_0_3_ip_block;
+
+void jpeg_v4_0_3_dec_ring_emit_ib(struct amdgpu_ring *ring,
+				struct amdgpu_job *job,
+				struct amdgpu_ib *ib,
+				uint32_t flags);
+void jpeg_v4_0_3_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq,
+				unsigned int flags);
+void jpeg_v4_0_3_dec_ring_emit_vm_flush(struct amdgpu_ring *ring,
+					unsigned int vmid, uint64_t pd_addr);
+void jpeg_v4_0_3_ring_emit_hdp_flush(struct amdgpu_ring *ring);
+void jpeg_v4_0_3_dec_ring_nop(struct amdgpu_ring *ring, uint32_t count);
+void jpeg_v4_0_3_dec_ring_insert_start(struct amdgpu_ring *ring);
+void jpeg_v4_0_3_dec_ring_insert_end(struct amdgpu_ring *ring);
+void jpeg_v4_0_3_dec_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg, uint32_t val);
+void jpeg_v4_0_3_dec_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
+					uint32_t val, uint32_t mask);
+
+#endif /* __JPEG_V4_0_3_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_5.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_5.c
new file mode 100644
index 000000000000..54fd9c800c40
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_5.c
@@ -0,0 +1,872 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "amdgpu.h"
+#include "amdgpu_jpeg.h"
+#include "amdgpu_pm.h"
+#include "soc15.h"
+#include "soc15d.h"
+#include "jpeg_v2_0.h"
+#include "jpeg_v4_0_5.h"
+#include "mmsch_v4_0.h"
+
+#include "vcn/vcn_4_0_5_offset.h"
+#include "vcn/vcn_4_0_5_sh_mask.h"
+#include "ivsrcid/vcn/irqsrcs_vcn_4_0.h"
+
+#define mmUVD_DPG_LMA_CTL						regUVD_DPG_LMA_CTL
+#define mmUVD_DPG_LMA_CTL_BASE_IDX					regUVD_DPG_LMA_CTL_BASE_IDX
+#define mmUVD_DPG_LMA_DATA						regUVD_DPG_LMA_DATA
+#define mmUVD_DPG_LMA_DATA_BASE_IDX					regUVD_DPG_LMA_DATA_BASE_IDX
+
+#define regUVD_JPEG_PITCH_INTERNAL_OFFSET		0x401f
+#define regJPEG_DEC_GFX10_ADDR_CONFIG_INTERNAL_OFFSET	0x4026
+#define regJPEG_SYS_INT_EN_INTERNAL_OFFSET		0x4141
+#define regJPEG_CGC_CTRL_INTERNAL_OFFSET		0x4161
+#define regJPEG_CGC_GATE_INTERNAL_OFFSET		0x4160
+#define regUVD_NO_OP_INTERNAL_OFFSET			0x0029
+
+static const struct amdgpu_hwip_reg_entry jpeg_reg_list_4_0_5[] = {
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_POWER_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_INT_STAT),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC_RB_CNTL),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC_RB_SIZE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_ADDR_MODE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_GFX10_ADDR_CONFIG),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_Y_GFX10_TILING_SURFACE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_UV_GFX10_TILING_SURFACE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_PITCH),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_UV_PITCH),
+};
+
+static void jpeg_v4_0_5_set_dec_ring_funcs(struct amdgpu_device *adev);
+static void jpeg_v4_0_5_set_irq_funcs(struct amdgpu_device *adev);
+static int jpeg_v4_0_5_set_powergating_state(struct amdgpu_ip_block *ip_block,
+				enum amd_powergating_state state);
+static void jpeg_v4_0_5_dec_ring_set_wptr(struct amdgpu_ring *ring);
+
+static int amdgpu_ih_clientid_jpeg[] = {
+	SOC15_IH_CLIENTID_VCN,
+	SOC15_IH_CLIENTID_VCN1
+};
+
+
+
+/**
+ * jpeg_v4_0_5_early_init - set function pointers
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Set ring and irq function pointers
+ */
+static int jpeg_v4_0_5_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	switch (amdgpu_ip_version(adev, UVD_HWIP, 0)) {
+	case IP_VERSION(4, 0, 5):
+		adev->jpeg.num_jpeg_inst = 1;
+		break;
+	case IP_VERSION(4, 0, 6):
+		adev->jpeg.num_jpeg_inst = 2;
+		break;
+	default:
+		DRM_DEV_ERROR(adev->dev,
+			"Failed to init vcn ip block(UVD_HWIP:0x%x)\n",
+			amdgpu_ip_version(adev, UVD_HWIP, 0));
+		return -EINVAL;
+	}
+
+	adev->jpeg.num_jpeg_rings = 1;
+
+	jpeg_v4_0_5_set_dec_ring_funcs(adev);
+	jpeg_v4_0_5_set_irq_funcs(adev);
+
+	return 0;
+}
+
+/**
+ * jpeg_v4_0_5_sw_init - sw init for JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Load firmware and sw initialization
+ */
+static int jpeg_v4_0_5_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring;
+	int r, i;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		if (adev->jpeg.harvest_config & (1 << i))
+			continue;
+
+		/* JPEG TRAP */
+		r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_jpeg[i],
+				VCN_4_0__SRCID__JPEG_DECODE, &adev->jpeg.inst[i].irq);
+		if (r)
+			return r;
+
+		/* JPEG DJPEG POISON EVENT */
+		r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_jpeg[i],
+			VCN_4_0__SRCID_DJPEG0_POISON, &adev->jpeg.inst[i].irq);
+		if (r)
+			return r;
+
+		/* JPEG EJPEG POISON EVENT */
+		r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_jpeg[i],
+			VCN_4_0__SRCID_EJPEG0_POISON, &adev->jpeg.inst[i].irq);
+		if (r)
+			return r;
+	}
+
+	r = amdgpu_jpeg_sw_init(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_jpeg_resume(adev);
+	if (r)
+		return r;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		if (adev->jpeg.harvest_config & (1 << i))
+			continue;
+
+		ring = adev->jpeg.inst[i].ring_dec;
+		ring->use_doorbell = true;
+		ring->vm_hub = AMDGPU_MMHUB0(0);
+		ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1 + 8 * i;
+		sprintf(ring->name, "jpeg_dec_%d", i);
+		r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst[i].irq,
+				     0, AMDGPU_RING_PRIO_DEFAULT, NULL);
+		if (r)
+			return r;
+
+		adev->jpeg.internal.jpeg_pitch[0] = regUVD_JPEG_PITCH_INTERNAL_OFFSET;
+		adev->jpeg.inst[i].external.jpeg_pitch[0] = SOC15_REG_OFFSET(JPEG, i, regUVD_JPEG_PITCH);
+	}
+
+	r = amdgpu_jpeg_reg_dump_init(adev, jpeg_reg_list_4_0_5, ARRAY_SIZE(jpeg_reg_list_4_0_5));
+	if (r)
+		return r;
+
+	adev->jpeg.supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->jpeg.inst[0].ring_dec[0]);
+	if (!amdgpu_sriov_vf(adev))
+		adev->jpeg.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+	r = amdgpu_jpeg_sysfs_reset_mask_init(adev);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+/**
+ * jpeg_v4_0_5_sw_fini - sw fini for JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * JPEG suspend and free up sw allocation
+ */
+static int jpeg_v4_0_5_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = amdgpu_jpeg_suspend(adev);
+	if (r)
+		return r;
+
+	amdgpu_jpeg_sysfs_reset_mask_fini(adev);
+	r = amdgpu_jpeg_sw_fini(adev);
+
+	return r;
+}
+
+/**
+ * jpeg_v4_0_5_hw_init - start and test JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ */
+static int jpeg_v4_0_5_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring;
+	int i, r = 0;
+
+	// TODO: Enable ring test with DPG support
+	if (adev->pg_flags & AMD_PG_SUPPORT_JPEG_DPG) {
+		return 0;
+	}
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		if (adev->jpeg.harvest_config & (1 << i))
+			continue;
+
+		ring = adev->jpeg.inst[i].ring_dec;
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+/**
+ * jpeg_v4_0_5_hw_fini - stop the hardware block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Stop the JPEG block, mark ring as not ready any more
+ */
+static int jpeg_v4_0_5_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i;
+
+	cancel_delayed_work_sync(&adev->jpeg.idle_work);
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		if (adev->jpeg.harvest_config & (1 << i))
+			continue;
+
+		if (!amdgpu_sriov_vf(adev)) {
+			if (adev->jpeg.cur_state != AMD_PG_STATE_GATE &&
+			    RREG32_SOC15(JPEG, i, regUVD_JRBC_STATUS))
+				jpeg_v4_0_5_set_powergating_state(ip_block, AMD_PG_STATE_GATE);
+		}
+	}
+	return 0;
+}
+
+/**
+ * jpeg_v4_0_5_suspend - suspend JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * HW fini and suspend JPEG block
+ */
+static int jpeg_v4_0_5_suspend(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+
+	r = jpeg_v4_0_5_hw_fini(ip_block);
+	if (r)
+		return r;
+
+	r = amdgpu_jpeg_suspend(ip_block->adev);
+
+	return r;
+}
+
+/**
+ * jpeg_v4_0_5_resume - resume JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Resume firmware and hw init JPEG block
+ */
+static int jpeg_v4_0_5_resume(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+
+	r = amdgpu_jpeg_resume(ip_block->adev);
+	if (r)
+		return r;
+
+	r = jpeg_v4_0_5_hw_init(ip_block);
+
+	return r;
+}
+
+static void jpeg_v4_0_5_disable_clock_gating(struct amdgpu_device *adev, int inst)
+{
+	uint32_t data = 0;
+
+	data = RREG32_SOC15(JPEG, inst, regJPEG_CGC_CTRL);
+	if (adev->cg_flags & AMD_CG_SUPPORT_JPEG_MGCG) {
+		data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+		data &= (~JPEG_CGC_CTRL__JPEG_DEC_MODE_MASK);
+	} else {
+		data &= ~JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+	}
+
+	data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
+	data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
+	WREG32_SOC15(JPEG, inst, regJPEG_CGC_CTRL, data);
+
+	data = RREG32_SOC15(JPEG, inst, regJPEG_CGC_GATE);
+	data &= ~(JPEG_CGC_GATE__JPEG_DEC_MASK
+		| JPEG_CGC_GATE__JPEG2_DEC_MASK
+		| JPEG_CGC_GATE__JMCIF_MASK
+		| JPEG_CGC_GATE__JRBBM_MASK);
+	WREG32_SOC15(JPEG, inst, regJPEG_CGC_GATE, data);
+}
+
+static void jpeg_v4_0_5_enable_clock_gating(struct amdgpu_device *adev, int inst)
+{
+	uint32_t data = 0;
+
+	data = RREG32_SOC15(JPEG, inst, regJPEG_CGC_CTRL);
+	if (adev->cg_flags & AMD_CG_SUPPORT_JPEG_MGCG) {
+		data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+		data |= JPEG_CGC_CTRL__JPEG_DEC_MODE_MASK;
+	} else {
+		data &= ~JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+	}
+
+	data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
+	data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
+	WREG32_SOC15(JPEG, inst, regJPEG_CGC_CTRL, data);
+
+	data = RREG32_SOC15(JPEG, inst, regJPEG_CGC_GATE);
+	data |= (JPEG_CGC_GATE__JPEG_DEC_MASK
+		|JPEG_CGC_GATE__JPEG2_DEC_MASK
+		|JPEG_CGC_GATE__JMCIF_MASK
+		|JPEG_CGC_GATE__JRBBM_MASK);
+	WREG32_SOC15(JPEG, inst, regJPEG_CGC_GATE, data);
+}
+
+static void jpeg_engine_4_0_5_dpg_clock_gating_mode(struct amdgpu_device *adev,
+			int inst_idx, uint8_t indirect)
+{
+	uint32_t data = 0;
+
+	if (adev->cg_flags & AMD_CG_SUPPORT_JPEG_MGCG)
+		data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+	else
+		data |= 0 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+
+	data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
+	data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
+	WREG32_SOC15_JPEG_DPG_MODE(inst_idx, regJPEG_CGC_CTRL_INTERNAL_OFFSET, data, indirect);
+
+	data = 0;
+	WREG32_SOC15_JPEG_DPG_MODE(inst_idx, regJPEG_CGC_GATE_INTERNAL_OFFSET,
+				data, indirect);
+}
+
+static int jpeg_v4_0_5_disable_static_power_gating(struct amdgpu_device *adev, int inst)
+{
+	if (adev->pg_flags & AMD_PG_SUPPORT_JPEG) {
+		WREG32(SOC15_REG_OFFSET(JPEG, inst, regUVD_IPX_DLDO_CONFIG),
+			1 << UVD_IPX_DLDO_CONFIG__ONO1_PWR_CONFIG__SHIFT);
+		SOC15_WAIT_ON_RREG(JPEG, inst, regUVD_IPX_DLDO_STATUS,
+			0, UVD_IPX_DLDO_STATUS__ONO1_PWR_STATUS_MASK);
+	}
+
+	/* disable anti hang mechanism */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, inst, regUVD_JPEG_POWER_STATUS), 0,
+		~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK);
+
+	/* keep the JPEG in static PG mode */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, inst, regUVD_JPEG_POWER_STATUS), 0,
+		~UVD_JPEG_POWER_STATUS__JPEG_PG_MODE_MASK);
+
+	return 0;
+}
+
+static int jpeg_v4_0_5_enable_static_power_gating(struct amdgpu_device *adev, int inst)
+{
+	/* enable anti hang mechanism */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, inst, regUVD_JPEG_POWER_STATUS),
+		UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK,
+		~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK);
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_JPEG) {
+		WREG32(SOC15_REG_OFFSET(JPEG, inst, regUVD_IPX_DLDO_CONFIG),
+			2 << UVD_IPX_DLDO_CONFIG__ONO1_PWR_CONFIG__SHIFT);
+		SOC15_WAIT_ON_RREG(JPEG, inst, regUVD_IPX_DLDO_STATUS,
+			1 << UVD_IPX_DLDO_STATUS__ONO1_PWR_STATUS__SHIFT,
+			UVD_IPX_DLDO_STATUS__ONO1_PWR_STATUS_MASK);
+	}
+
+	return 0;
+}
+
+/**
+ * jpeg_v4_0_5_start_dpg_mode - Jpeg start with dpg mode
+ *
+ * @adev: amdgpu_device pointer
+ * @inst_idx: instance number index
+ * @indirect: indirectly write sram
+ *
+ * Start JPEG block with dpg mode
+ */
+static void jpeg_v4_0_5_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, bool indirect)
+{
+	struct amdgpu_ring *ring = adev->jpeg.inst[inst_idx].ring_dec;
+	uint32_t reg_data = 0;
+
+	/* enable anti hang mechanism */
+	reg_data = RREG32_SOC15(JPEG, inst_idx, regUVD_JPEG_POWER_STATUS);
+	reg_data &= ~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK;
+	reg_data |=  0x1;
+	WREG32_SOC15(JPEG, inst_idx, regUVD_JPEG_POWER_STATUS, reg_data);
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_JPEG) {
+		WREG32(SOC15_REG_OFFSET(JPEG, inst_idx, regUVD_IPX_DLDO_CONFIG),
+			2 << UVD_IPX_DLDO_CONFIG__ONO1_PWR_CONFIG__SHIFT);
+		SOC15_WAIT_ON_RREG(JPEG, inst_idx, regUVD_IPX_DLDO_STATUS,
+			1 << UVD_IPX_DLDO_STATUS__ONO1_PWR_STATUS__SHIFT,
+			UVD_IPX_DLDO_STATUS__ONO1_PWR_STATUS_MASK);
+	}
+
+	reg_data = RREG32_SOC15(JPEG, inst_idx, regUVD_JPEG_POWER_STATUS);
+	reg_data |= UVD_JPEG_POWER_STATUS__JPEG_PG_MODE_MASK;
+	WREG32_SOC15(JPEG, inst_idx, regUVD_JPEG_POWER_STATUS, reg_data);
+
+	if (indirect)
+		adev->jpeg.inst[inst_idx].dpg_sram_curr_addr =
+					(uint32_t *)adev->jpeg.inst[inst_idx].dpg_sram_cpu_addr;
+
+	jpeg_engine_4_0_5_dpg_clock_gating_mode(adev, inst_idx, indirect);
+
+	/* MJPEG global tiling registers */
+	WREG32_SOC15_JPEG_DPG_MODE(inst_idx, regJPEG_DEC_GFX10_ADDR_CONFIG_INTERNAL_OFFSET,
+	adev->gfx.config.gb_addr_config, indirect);
+	/* enable System Interrupt for JRBC */
+	WREG32_SOC15_JPEG_DPG_MODE(inst_idx, regJPEG_SYS_INT_EN_INTERNAL_OFFSET,
+	JPEG_SYS_INT_EN__DJRBC_MASK, indirect);
+
+	/* add nop to workaround PSP size check */
+	WREG32_SOC15_JPEG_DPG_MODE(inst_idx, regUVD_NO_OP_INTERNAL_OFFSET, 0, indirect);
+
+	if (indirect)
+		amdgpu_jpeg_psp_update_sram(adev, inst_idx, 0);
+
+	WREG32_SOC15(JPEG, inst_idx, regUVD_LMI_JRBC_RB_VMID, 0);
+	WREG32_SOC15(JPEG, inst_idx, regUVD_JRBC_RB_CNTL, (0x00000001L | 0x00000002L));
+	WREG32_SOC15(JPEG, inst_idx, regUVD_LMI_JRBC_RB_64BIT_BAR_LOW,
+		lower_32_bits(ring->gpu_addr));
+	WREG32_SOC15(JPEG, inst_idx, regUVD_LMI_JRBC_RB_64BIT_BAR_HIGH,
+		upper_32_bits(ring->gpu_addr));
+	WREG32_SOC15(JPEG, inst_idx, regUVD_JRBC_RB_RPTR, 0);
+	WREG32_SOC15(JPEG, inst_idx, regUVD_JRBC_RB_WPTR, 0);
+	WREG32_SOC15(JPEG, inst_idx, regUVD_JRBC_RB_CNTL, 0x00000002L);
+	WREG32_SOC15(JPEG, inst_idx, regUVD_JRBC_RB_SIZE, ring->ring_size / 4);
+	ring->wptr = RREG32_SOC15(JPEG, inst_idx, regUVD_JRBC_RB_WPTR);
+}
+
+/**
+ * jpeg_v4_0_5_stop_dpg_mode - Jpeg stop with dpg mode
+ *
+ * @adev: amdgpu_device pointer
+ * @inst_idx: instance number index
+ *
+ * Stop JPEG block with dpg mode
+ */
+static void jpeg_v4_0_5_stop_dpg_mode(struct amdgpu_device *adev, int inst_idx)
+{
+	uint32_t reg_data = 0;
+
+	reg_data = RREG32_SOC15(JPEG, inst_idx, regUVD_JPEG_POWER_STATUS);
+	reg_data &= ~UVD_JPEG_POWER_STATUS__JPEG_PG_MODE_MASK;
+	WREG32_SOC15(JPEG, inst_idx, regUVD_JPEG_POWER_STATUS, reg_data);
+
+}
+
+/**
+ * jpeg_v4_0_5_start - start JPEG block
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Setup and start the JPEG block
+ */
+static int jpeg_v4_0_5_start(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	int r, i;
+
+	if (adev->pm.dpm_enabled)
+		amdgpu_dpm_enable_jpeg(adev, true);
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		if (adev->jpeg.harvest_config & (1 << i))
+			continue;
+
+		ring = adev->jpeg.inst[i].ring_dec;
+		/* doorbell programming is done for every playback */
+		adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell,
+				(adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 8 * i, i);
+
+		WREG32_SOC15(VCN, i, regVCN_JPEG_DB_CTRL,
+			ring->doorbell_index << VCN_JPEG_DB_CTRL__OFFSET__SHIFT |
+			VCN_JPEG_DB_CTRL__EN_MASK);
+
+		if (adev->pg_flags & AMD_PG_SUPPORT_JPEG_DPG) {
+			jpeg_v4_0_5_start_dpg_mode(adev, i, adev->jpeg.indirect_sram);
+			continue;
+		}
+
+		/* disable power gating */
+		r = jpeg_v4_0_5_disable_static_power_gating(adev, i);
+		if (r)
+			return r;
+
+		/* JPEG disable CGC */
+		jpeg_v4_0_5_disable_clock_gating(adev, i);
+
+		/* MJPEG global tiling registers */
+		WREG32_SOC15(JPEG, i, regJPEG_DEC_GFX10_ADDR_CONFIG,
+			adev->gfx.config.gb_addr_config);
+
+		/* enable JMI channel */
+		WREG32_P(SOC15_REG_OFFSET(JPEG, i, regUVD_JMI_CNTL), 0,
+			~UVD_JMI_CNTL__SOFT_RESET_MASK);
+
+		/* enable System Interrupt for JRBC */
+		WREG32_P(SOC15_REG_OFFSET(JPEG, i, regJPEG_SYS_INT_EN),
+			JPEG_SYS_INT_EN__DJRBC_MASK,
+			~JPEG_SYS_INT_EN__DJRBC_MASK);
+
+		WREG32_SOC15(JPEG, i, regUVD_LMI_JRBC_RB_VMID, 0);
+		WREG32_SOC15(JPEG, i, regUVD_JRBC_RB_CNTL, (0x00000001L | 0x00000002L));
+		WREG32_SOC15(JPEG, i, regUVD_LMI_JRBC_RB_64BIT_BAR_LOW,
+			lower_32_bits(ring->gpu_addr));
+		WREG32_SOC15(JPEG, i, regUVD_LMI_JRBC_RB_64BIT_BAR_HIGH,
+			upper_32_bits(ring->gpu_addr));
+		WREG32_SOC15(JPEG, i, regUVD_JRBC_RB_RPTR, 0);
+		WREG32_SOC15(JPEG, i, regUVD_JRBC_RB_WPTR, 0);
+		WREG32_SOC15(JPEG, i, regUVD_JRBC_RB_CNTL, 0x00000002L);
+		WREG32_SOC15(JPEG, i, regUVD_JRBC_RB_SIZE, ring->ring_size / 4);
+		ring->wptr = RREG32_SOC15(JPEG, i, regUVD_JRBC_RB_WPTR);
+	}
+
+	return 0;
+}
+
+/**
+ * jpeg_v4_0_5_stop - stop JPEG block
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * stop the JPEG block
+ */
+static int jpeg_v4_0_5_stop(struct amdgpu_device *adev)
+{
+	int r, i;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		if (adev->jpeg.harvest_config & (1 << i))
+			continue;
+
+		if (adev->pg_flags & AMD_PG_SUPPORT_JPEG_DPG) {
+			jpeg_v4_0_5_stop_dpg_mode(adev, i);
+			continue;
+		}
+
+		/* reset JMI */
+		WREG32_P(SOC15_REG_OFFSET(JPEG, i, regUVD_JMI_CNTL),
+			UVD_JMI_CNTL__SOFT_RESET_MASK,
+			~UVD_JMI_CNTL__SOFT_RESET_MASK);
+
+		jpeg_v4_0_5_enable_clock_gating(adev, i);
+
+		/* enable power gating */
+		r = jpeg_v4_0_5_enable_static_power_gating(adev, i);
+		if (r)
+			return r;
+	}
+	if (adev->pm.dpm_enabled)
+		amdgpu_dpm_enable_jpeg(adev, false);
+
+	return 0;
+}
+
+/**
+ * jpeg_v4_0_5_dec_ring_get_rptr - get read pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware read pointer
+ */
+static uint64_t jpeg_v4_0_5_dec_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	return RREG32_SOC15(JPEG, ring->me, regUVD_JRBC_RB_RPTR);
+}
+
+/**
+ * jpeg_v4_0_5_dec_ring_get_wptr - get write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware write pointer
+ */
+static uint64_t jpeg_v4_0_5_dec_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->use_doorbell)
+		return *ring->wptr_cpu_addr;
+	else
+		return RREG32_SOC15(JPEG, ring->me, regUVD_JRBC_RB_WPTR);
+}
+
+/**
+ * jpeg_v4_0_5_dec_ring_set_wptr - set write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Commits the write pointer to the hardware
+ */
+static void jpeg_v4_0_5_dec_ring_set_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->use_doorbell) {
+		*ring->wptr_cpu_addr = lower_32_bits(ring->wptr);
+		WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr));
+	} else {
+		WREG32_SOC15(JPEG, ring->me, regUVD_JRBC_RB_WPTR, lower_32_bits(ring->wptr));
+	}
+}
+
+static bool jpeg_v4_0_5_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, ret = 1;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		if (adev->jpeg.harvest_config & (1 << i))
+			continue;
+
+		ret &= (((RREG32_SOC15(JPEG, i, regUVD_JRBC_STATUS) &
+			UVD_JRBC_STATUS__RB_JOB_DONE_MASK) ==
+			UVD_JRBC_STATUS__RB_JOB_DONE_MASK));
+	}
+	return ret;
+}
+
+static int jpeg_v4_0_5_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		if (adev->jpeg.harvest_config & (1 << i))
+			continue;
+
+		return SOC15_WAIT_ON_RREG(JPEG, i, regUVD_JRBC_STATUS,
+			UVD_JRBC_STATUS__RB_JOB_DONE_MASK,
+			UVD_JRBC_STATUS__RB_JOB_DONE_MASK);
+	}
+
+	return 0;
+}
+
+static int jpeg_v4_0_5_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_clockgating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	bool enable = state == AMD_CG_STATE_GATE;
+	int i;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		if (adev->jpeg.harvest_config & (1 << i))
+			continue;
+
+		if (enable) {
+			if (!jpeg_v4_0_5_is_idle(ip_block))
+				return -EBUSY;
+
+			jpeg_v4_0_5_enable_clock_gating(adev, i);
+		} else {
+			jpeg_v4_0_5_disable_clock_gating(adev, i);
+		}
+	}
+
+	return 0;
+}
+
+static int jpeg_v4_0_5_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_powergating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int ret;
+
+	if (amdgpu_sriov_vf(adev)) {
+		adev->jpeg.cur_state = AMD_PG_STATE_UNGATE;
+		return 0;
+	}
+
+	if (state == adev->jpeg.cur_state)
+		return 0;
+
+	if (state == AMD_PG_STATE_GATE)
+		ret = jpeg_v4_0_5_stop(adev);
+	else
+		ret = jpeg_v4_0_5_start(adev);
+
+	if (!ret)
+		adev->jpeg.cur_state = state;
+
+	return ret;
+}
+
+static int jpeg_v4_0_5_process_interrupt(struct amdgpu_device *adev,
+				      struct amdgpu_irq_src *source,
+				      struct amdgpu_iv_entry *entry)
+{
+	uint32_t ip_instance;
+
+	DRM_DEBUG("IH: JPEG TRAP\n");
+
+	switch (entry->client_id) {
+	case SOC15_IH_CLIENTID_VCN:
+		ip_instance = 0;
+		break;
+	case SOC15_IH_CLIENTID_VCN1:
+		ip_instance = 1;
+		break;
+	default:
+		DRM_ERROR("Unhandled client id: %d\n", entry->client_id);
+		return 0;
+	}
+
+	switch (entry->src_id) {
+	case VCN_4_0__SRCID__JPEG_DECODE:
+		amdgpu_fence_process(adev->jpeg.inst[ip_instance].ring_dec);
+		break;
+	case VCN_4_0__SRCID_DJPEG0_POISON:
+	case VCN_4_0__SRCID_EJPEG0_POISON:
+		amdgpu_jpeg_process_poison_irq(adev, source, entry);
+		break;
+	default:
+		DRM_DEV_ERROR(adev->dev, "Unhandled interrupt: %d %d\n",
+			  entry->src_id, entry->src_data[0]);
+		break;
+	}
+
+	return 0;
+}
+
+static int jpeg_v4_0_5_ring_reset(struct amdgpu_ring *ring,
+				  unsigned int vmid,
+				  struct amdgpu_fence *timedout_fence)
+{
+	int r;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+	r = jpeg_v4_0_5_stop(ring->adev);
+	if (r)
+		return r;
+	r = jpeg_v4_0_5_start(ring->adev);
+	if (r)
+		return r;
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static const struct amd_ip_funcs jpeg_v4_0_5_ip_funcs = {
+	.name = "jpeg_v4_0_5",
+	.early_init = jpeg_v4_0_5_early_init,
+	.sw_init = jpeg_v4_0_5_sw_init,
+	.sw_fini = jpeg_v4_0_5_sw_fini,
+	.hw_init = jpeg_v4_0_5_hw_init,
+	.hw_fini = jpeg_v4_0_5_hw_fini,
+	.suspend = jpeg_v4_0_5_suspend,
+	.resume = jpeg_v4_0_5_resume,
+	.is_idle = jpeg_v4_0_5_is_idle,
+	.wait_for_idle = jpeg_v4_0_5_wait_for_idle,
+	.set_clockgating_state = jpeg_v4_0_5_set_clockgating_state,
+	.set_powergating_state = jpeg_v4_0_5_set_powergating_state,
+	.dump_ip_state = amdgpu_jpeg_dump_ip_state,
+	.print_ip_state = amdgpu_jpeg_print_ip_state,
+};
+
+static const struct amdgpu_ring_funcs jpeg_v4_0_5_dec_ring_vm_funcs = {
+	.type = AMDGPU_RING_TYPE_VCN_JPEG,
+	.align_mask = 0xf,
+	.get_rptr = jpeg_v4_0_5_dec_ring_get_rptr,
+	.get_wptr = jpeg_v4_0_5_dec_ring_get_wptr,
+	.set_wptr = jpeg_v4_0_5_dec_ring_set_wptr,
+	.parse_cs = amdgpu_jpeg_dec_parse_cs,
+	.emit_frame_size =
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 +
+		8 + /* jpeg_v4_0_5_dec_ring_emit_vm_flush */
+		18 + 18 + /* jpeg_v4_0_5_dec_ring_emit_fence x2 vm fence */
+		8 + 16,
+	.emit_ib_size = 22, /* jpeg_v4_0_5_dec_ring_emit_ib */
+	.emit_ib = jpeg_v2_0_dec_ring_emit_ib,
+	.emit_fence = jpeg_v2_0_dec_ring_emit_fence,
+	.emit_vm_flush = jpeg_v2_0_dec_ring_emit_vm_flush,
+	.test_ring = amdgpu_jpeg_dec_ring_test_ring,
+	.test_ib = amdgpu_jpeg_dec_ring_test_ib,
+	.insert_nop = jpeg_v2_0_dec_ring_nop,
+	.insert_start = jpeg_v2_0_dec_ring_insert_start,
+	.insert_end = jpeg_v2_0_dec_ring_insert_end,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.begin_use = amdgpu_jpeg_ring_begin_use,
+	.end_use = amdgpu_jpeg_ring_end_use,
+	.emit_wreg = jpeg_v2_0_dec_ring_emit_wreg,
+	.emit_reg_wait = jpeg_v2_0_dec_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = jpeg_v4_0_5_ring_reset,
+};
+
+static void jpeg_v4_0_5_set_dec_ring_funcs(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		if (adev->jpeg.harvest_config & (1 << i))
+			continue;
+
+		adev->jpeg.inst[i].ring_dec->funcs = &jpeg_v4_0_5_dec_ring_vm_funcs;
+		adev->jpeg.inst[i].ring_dec->me = i;
+	}
+}
+
+static const struct amdgpu_irq_src_funcs jpeg_v4_0_5_irq_funcs = {
+	.process = jpeg_v4_0_5_process_interrupt,
+};
+
+static void jpeg_v4_0_5_set_irq_funcs(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		if (adev->jpeg.harvest_config & (1 << i))
+			continue;
+
+		adev->jpeg.inst[i].irq.num_types = 1;
+		adev->jpeg.inst[i].irq.funcs = &jpeg_v4_0_5_irq_funcs;
+	}
+}
+
+const struct amdgpu_ip_block_version jpeg_v4_0_5_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_JPEG,
+	.major = 4,
+	.minor = 0,
+	.rev = 5,
+	.funcs = &jpeg_v4_0_5_ip_funcs,
+};
+
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_5.h b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_5.h
new file mode 100644
index 000000000000..c5eee572079c
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_5.h
@@ -0,0 +1,35 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __JPEG_V4_0_5_H__
+#define __JPEG_V4_0_5_H__
+
+enum amdgpu_jpeg_v4_0_5_sub_block {
+	AMDGPU_JPEG_V4_0_5_JPEG0 = 0,
+
+	AMDGPU_JPEG_V4_0_5_MAX_SUB_BLOCK,
+};
+
+extern const struct amdgpu_ip_block_version jpeg_v4_0_5_ip_block;
+
+#endif /* __JPEG_V4_0_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_0.c
new file mode 100644
index 000000000000..46bf15dce2bd
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_0.c
@@ -0,0 +1,733 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "amdgpu.h"
+#include "amdgpu_jpeg.h"
+#include "amdgpu_pm.h"
+#include "soc15.h"
+#include "soc15d.h"
+#include "jpeg_v2_0.h"
+#include "jpeg_v4_0_3.h"
+
+#include "vcn/vcn_5_0_0_offset.h"
+#include "vcn/vcn_5_0_0_sh_mask.h"
+#include "ivsrcid/vcn/irqsrcs_vcn_5_0.h"
+#include "jpeg_v5_0_0.h"
+
+static const struct amdgpu_hwip_reg_entry jpeg_reg_list_5_0[] = {
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_POWER_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_INT_STAT),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC_RB_CNTL),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC_RB_SIZE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_ADDR_MODE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_GFX10_ADDR_CONFIG),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_Y_GFX10_TILING_SURFACE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_UV_GFX10_TILING_SURFACE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_PITCH),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_UV_PITCH),
+};
+
+static void jpeg_v5_0_0_set_dec_ring_funcs(struct amdgpu_device *adev);
+static void jpeg_v5_0_0_set_irq_funcs(struct amdgpu_device *adev);
+static int jpeg_v5_0_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
+				enum amd_powergating_state state);
+
+/**
+ * jpeg_v5_0_0_early_init - set function pointers
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Set ring and irq function pointers
+ */
+static int jpeg_v5_0_0_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	adev->jpeg.num_jpeg_inst = 1;
+	adev->jpeg.num_jpeg_rings = 1;
+
+	jpeg_v5_0_0_set_dec_ring_funcs(adev);
+	jpeg_v5_0_0_set_irq_funcs(adev);
+
+	return 0;
+}
+
+/**
+ * jpeg_v5_0_0_sw_init - sw init for JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Load firmware and sw initialization
+ */
+static int jpeg_v5_0_0_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring;
+	int r;
+
+	/* JPEG TRAP */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
+		VCN_5_0__SRCID__JPEG_DECODE, &adev->jpeg.inst->irq);
+	if (r)
+		return r;
+
+	r = amdgpu_jpeg_sw_init(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_jpeg_resume(adev);
+	if (r)
+		return r;
+
+	ring = adev->jpeg.inst->ring_dec;
+	ring->use_doorbell = true;
+	ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1;
+	ring->vm_hub = AMDGPU_MMHUB0(0);
+
+	sprintf(ring->name, "jpeg_dec");
+	r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq, 0,
+			     AMDGPU_RING_PRIO_DEFAULT, NULL);
+	if (r)
+		return r;
+
+	adev->jpeg.internal.jpeg_pitch[0] = regUVD_JPEG_PITCH_INTERNAL_OFFSET;
+	adev->jpeg.inst->external.jpeg_pitch[0] = SOC15_REG_OFFSET(JPEG, 0, regUVD_JPEG_PITCH);
+
+	r = amdgpu_jpeg_reg_dump_init(adev, jpeg_reg_list_5_0, ARRAY_SIZE(jpeg_reg_list_5_0));
+	if (r)
+		return r;
+
+	adev->jpeg.supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->jpeg.inst[0].ring_dec[0]);
+	if (!amdgpu_sriov_vf(adev))
+		adev->jpeg.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+	r = amdgpu_jpeg_sysfs_reset_mask_init(adev);
+
+	return r;
+}
+
+/**
+ * jpeg_v5_0_0_sw_fini - sw fini for JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * JPEG suspend and free up sw allocation
+ */
+static int jpeg_v5_0_0_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = amdgpu_jpeg_suspend(adev);
+	if (r)
+		return r;
+
+	amdgpu_jpeg_sysfs_reset_mask_fini(adev);
+	r = amdgpu_jpeg_sw_fini(adev);
+
+	return r;
+}
+
+/**
+ * jpeg_v5_0_0_hw_init - start and test JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ */
+static int jpeg_v5_0_0_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring = adev->jpeg.inst->ring_dec;
+	int r;
+
+	adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell,
+			(adev->doorbell_index.vcn.vcn_ring0_1 << 1), 0);
+
+	/* Skip ring test because pause DPG is not implemented. */
+	if (adev->pg_flags & AMD_PG_SUPPORT_JPEG_DPG)
+		return 0;
+
+	r = amdgpu_ring_test_helper(ring);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+/**
+ * jpeg_v5_0_0_hw_fini - stop the hardware block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Stop the JPEG block, mark ring as not ready any more
+ */
+static int jpeg_v5_0_0_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	cancel_delayed_work_sync(&adev->jpeg.idle_work);
+
+	if (adev->jpeg.cur_state != AMD_PG_STATE_GATE &&
+	      RREG32_SOC15(JPEG, 0, regUVD_JRBC_STATUS))
+		jpeg_v5_0_0_set_powergating_state(ip_block, AMD_PG_STATE_GATE);
+
+	return 0;
+}
+
+/**
+ * jpeg_v5_0_0_suspend - suspend JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * HW fini and suspend JPEG block
+ */
+static int jpeg_v5_0_0_suspend(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+
+	r = jpeg_v5_0_0_hw_fini(ip_block);
+	if (r)
+		return r;
+
+	r = amdgpu_jpeg_suspend(ip_block->adev);
+
+	return r;
+}
+
+/**
+ * jpeg_v5_0_0_resume - resume JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Resume firmware and hw init JPEG block
+ */
+static int jpeg_v5_0_0_resume(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+
+	r = amdgpu_jpeg_resume(ip_block->adev);
+	if (r)
+		return r;
+
+	r = jpeg_v5_0_0_hw_init(ip_block);
+
+	return r;
+}
+
+static void jpeg_v5_0_0_disable_clock_gating(struct amdgpu_device *adev)
+{
+	uint32_t data = 0;
+
+	WREG32_SOC15(JPEG, 0, regJPEG_CGC_GATE, data);
+
+	data = RREG32_SOC15(JPEG, 0, regJPEG_CGC_CTRL);
+	data &= ~(JPEG_CGC_CTRL__JPEG0_DEC_MODE_MASK
+		| JPEG_CGC_CTRL__JPEG_ENC_MODE_MASK);
+	WREG32_SOC15(JPEG, 0, regJPEG_CGC_CTRL, data);
+}
+
+static void jpeg_v5_0_0_enable_clock_gating(struct amdgpu_device *adev)
+{
+	uint32_t data = 0;
+
+	data = RREG32_SOC15(JPEG, 0, regJPEG_CGC_CTRL);
+
+	data |= 1 << JPEG_CGC_CTRL__JPEG0_DEC_MODE__SHIFT;
+	WREG32_SOC15(JPEG, 0, regJPEG_CGC_CTRL, data);
+
+	data = RREG32_SOC15(JPEG, 0, regJPEG_CGC_GATE);
+	data |= (JPEG_CGC_GATE__JPEG0_DEC_MASK
+		|JPEG_CGC_GATE__JPEG_ENC_MASK
+		|JPEG_CGC_GATE__JMCIF_MASK
+		|JPEG_CGC_GATE__JRBBM_MASK);
+	WREG32_SOC15(JPEG, 0, regJPEG_CGC_GATE, data);
+}
+
+static int jpeg_v5_0_0_disable_power_gating(struct amdgpu_device *adev)
+{
+	uint32_t data = 0;
+
+	data = 1 << UVD_IPX_DLDO_CONFIG__ONO1_PWR_CONFIG__SHIFT;
+	WREG32_SOC15(JPEG, 0, regUVD_IPX_DLDO_CONFIG, data);
+	SOC15_WAIT_ON_RREG(JPEG, 0, regUVD_IPX_DLDO_STATUS, 0,
+			UVD_IPX_DLDO_STATUS__ONO1_PWR_STATUS_MASK);
+
+	/* disable anti hang mechanism */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, 0, regUVD_JPEG_POWER_STATUS), 0,
+		~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK);
+
+	return 0;
+}
+
+static int jpeg_v5_0_0_enable_power_gating(struct amdgpu_device *adev)
+{
+	/* enable anti hang mechanism */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, 0, regUVD_JPEG_POWER_STATUS),
+		UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK,
+		~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK);
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_JPEG) {
+		WREG32(SOC15_REG_OFFSET(JPEG, 0, regUVD_IPX_DLDO_CONFIG),
+			2 << UVD_IPX_DLDO_CONFIG__ONO1_PWR_CONFIG__SHIFT);
+		SOC15_WAIT_ON_RREG(JPEG, 0, regUVD_IPX_DLDO_STATUS,
+			1 << UVD_IPX_DLDO_STATUS__ONO1_PWR_STATUS__SHIFT,
+			UVD_IPX_DLDO_STATUS__ONO1_PWR_STATUS_MASK);
+	}
+
+	return 0;
+}
+
+static void jpeg_engine_5_0_0_dpg_clock_gating_mode(struct amdgpu_device *adev,
+	       int inst_idx, uint8_t indirect)
+{
+	uint32_t data = 0;
+
+	// JPEG disable CGC
+	if (adev->cg_flags & AMD_CG_SUPPORT_JPEG_MGCG)
+		data = 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+	else
+		data = 0 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+
+	data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
+	data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
+
+	if (indirect) {
+		ADD_SOC24_JPEG_TO_DPG_SRAM(inst_idx, vcnipJPEG_CGC_CTRL, data, indirect);
+
+		// Turn on All JPEG clocks
+		data = 0;
+		ADD_SOC24_JPEG_TO_DPG_SRAM(inst_idx, vcnipJPEG_CGC_GATE, data, indirect);
+	} else {
+		WREG32_SOC24_JPEG_DPG_MODE(inst_idx, vcnipJPEG_CGC_CTRL, data, indirect);
+
+		// Turn on All JPEG clocks
+		data = 0;
+		WREG32_SOC24_JPEG_DPG_MODE(inst_idx, vcnipJPEG_CGC_GATE, data, indirect);
+	}
+}
+
+/**
+ * jpeg_v5_0_0_start_dpg_mode - Jpeg start with dpg mode
+ *
+ * @adev: amdgpu_device pointer
+ * @inst_idx: instance number index
+ * @indirect: indirectly write sram
+ *
+ * Start JPEG block with dpg mode
+ */
+static int jpeg_v5_0_0_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, bool indirect)
+{
+	struct amdgpu_ring *ring = adev->jpeg.inst[inst_idx].ring_dec;
+	uint32_t reg_data = 0;
+
+	jpeg_v5_0_0_enable_power_gating(adev);
+
+	// enable dynamic power gating mode
+	reg_data = RREG32_SOC15(JPEG, inst_idx, regUVD_JPEG_POWER_STATUS);
+	reg_data |= UVD_JPEG_POWER_STATUS__JPEG_PG_MODE_MASK;
+	WREG32_SOC15(JPEG, inst_idx, regUVD_JPEG_POWER_STATUS, reg_data);
+
+	if (indirect)
+		adev->jpeg.inst[inst_idx].dpg_sram_curr_addr =
+			(uint32_t *)adev->jpeg.inst[inst_idx].dpg_sram_cpu_addr;
+
+	jpeg_engine_5_0_0_dpg_clock_gating_mode(adev, inst_idx, indirect);
+
+	/* MJPEG global tiling registers */
+	if (indirect)
+		ADD_SOC24_JPEG_TO_DPG_SRAM(inst_idx, vcnipJPEG_DEC_GFX10_ADDR_CONFIG,
+			adev->gfx.config.gb_addr_config, indirect);
+	else
+		WREG32_SOC24_JPEG_DPG_MODE(inst_idx, vcnipJPEG_DEC_GFX10_ADDR_CONFIG,
+			adev->gfx.config.gb_addr_config, 1);
+
+	/* enable System Interrupt for JRBC */
+	if (indirect)
+		ADD_SOC24_JPEG_TO_DPG_SRAM(inst_idx, vcnipJPEG_SYS_INT_EN,
+			JPEG_SYS_INT_EN__DJRBC0_MASK, indirect);
+	else
+		WREG32_SOC24_JPEG_DPG_MODE(inst_idx, vcnipJPEG_SYS_INT_EN,
+			JPEG_SYS_INT_EN__DJRBC0_MASK, 1);
+
+	if (indirect) {
+		/* add nop to workaround PSP size check */
+		ADD_SOC24_JPEG_TO_DPG_SRAM(inst_idx, vcnipUVD_NO_OP, 0, indirect);
+
+		amdgpu_jpeg_psp_update_sram(adev, inst_idx, 0);
+	}
+
+	WREG32_SOC15(VCN, 0, regVCN_JPEG_DB_CTRL,
+		ring->doorbell_index << VCN_JPEG_DB_CTRL__OFFSET__SHIFT |
+		VCN_JPEG_DB_CTRL__EN_MASK);
+
+	WREG32_SOC15(JPEG, inst_idx, regUVD_LMI_JRBC_RB_VMID, 0);
+	WREG32_SOC15(JPEG, inst_idx, regUVD_JRBC_RB_CNTL, (0x00000001L | 0x00000002L));
+	WREG32_SOC15(JPEG, inst_idx, regUVD_LMI_JRBC_RB_64BIT_BAR_LOW,
+		lower_32_bits(ring->gpu_addr));
+	WREG32_SOC15(JPEG, inst_idx, regUVD_LMI_JRBC_RB_64BIT_BAR_HIGH,
+		upper_32_bits(ring->gpu_addr));
+	WREG32_SOC15(JPEG, inst_idx, regUVD_JRBC_RB_RPTR, 0);
+	WREG32_SOC15(JPEG, inst_idx, regUVD_JRBC_RB_WPTR, 0);
+	WREG32_SOC15(JPEG, inst_idx, regUVD_JRBC_RB_CNTL, 0x00000002L);
+	WREG32_SOC15(JPEG, inst_idx, regUVD_JRBC_RB_SIZE, ring->ring_size / 4);
+	ring->wptr = RREG32_SOC15(JPEG, inst_idx, regUVD_JRBC_RB_WPTR);
+
+	return 0;
+}
+
+/**
+ * jpeg_v5_0_0_stop_dpg_mode - Jpeg stop with dpg mode
+ *
+ * @adev: amdgpu_device pointer
+ * @inst_idx: instance number index
+ *
+ * Stop JPEG block with dpg mode
+ */
+static void jpeg_v5_0_0_stop_dpg_mode(struct amdgpu_device *adev, int inst_idx)
+{
+	uint32_t reg_data = 0;
+
+	reg_data = RREG32_SOC15(JPEG, inst_idx, regUVD_JPEG_POWER_STATUS);
+	reg_data &= ~UVD_JPEG_POWER_STATUS__JPEG_PG_MODE_MASK;
+	WREG32_SOC15(JPEG, inst_idx, regUVD_JPEG_POWER_STATUS, reg_data);
+}
+
+/**
+ * jpeg_v5_0_0_start - start JPEG block
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Setup and start the JPEG block
+ */
+static int jpeg_v5_0_0_start(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring = adev->jpeg.inst->ring_dec;
+	int r;
+
+	if (adev->pm.dpm_enabled)
+		amdgpu_dpm_enable_jpeg(adev, true);
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_JPEG_DPG) {
+		r = jpeg_v5_0_0_start_dpg_mode(adev, 0, adev->jpeg.indirect_sram);
+		return r;
+	}
+
+	/* disable power gating */
+	r = jpeg_v5_0_0_disable_power_gating(adev);
+	if (r)
+		return r;
+
+	/* JPEG disable CGC */
+	jpeg_v5_0_0_disable_clock_gating(adev);
+
+	/* MJPEG global tiling registers */
+	WREG32_SOC15(JPEG, 0, regJPEG_DEC_GFX10_ADDR_CONFIG,
+		adev->gfx.config.gb_addr_config);
+
+	/* enable JMI channel */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, 0, regUVD_JMI_CNTL), 0,
+		~UVD_JMI_CNTL__SOFT_RESET_MASK);
+
+	/* enable System Interrupt for JRBC */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, 0, regJPEG_SYS_INT_EN),
+		JPEG_SYS_INT_EN__DJRBC0_MASK,
+		~JPEG_SYS_INT_EN__DJRBC0_MASK);
+
+	WREG32_SOC15(VCN, 0, regVCN_JPEG_DB_CTRL,
+		ring->doorbell_index << VCN_JPEG_DB_CTRL__OFFSET__SHIFT |
+		VCN_JPEG_DB_CTRL__EN_MASK);
+
+	WREG32_SOC15(JPEG, 0, regUVD_LMI_JRBC_RB_VMID, 0);
+	WREG32_SOC15(JPEG, 0, regUVD_JRBC_RB_CNTL, (0x00000001L | 0x00000002L));
+	WREG32_SOC15(JPEG, 0, regUVD_LMI_JRBC_RB_64BIT_BAR_LOW,
+		lower_32_bits(ring->gpu_addr));
+	WREG32_SOC15(JPEG, 0, regUVD_LMI_JRBC_RB_64BIT_BAR_HIGH,
+		upper_32_bits(ring->gpu_addr));
+	WREG32_SOC15(JPEG, 0, regUVD_JRBC_RB_RPTR, 0);
+	WREG32_SOC15(JPEG, 0, regUVD_JRBC_RB_WPTR, 0);
+	WREG32_SOC15(JPEG, 0, regUVD_JRBC_RB_CNTL, 0x00000002L);
+	WREG32_SOC15(JPEG, 0, regUVD_JRBC_RB_SIZE, ring->ring_size / 4);
+	ring->wptr = RREG32_SOC15(JPEG, 0, regUVD_JRBC_RB_WPTR);
+
+	return 0;
+}
+
+/**
+ * jpeg_v5_0_0_stop - stop JPEG block
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * stop the JPEG block
+ */
+static int jpeg_v5_0_0_stop(struct amdgpu_device *adev)
+{
+	int r;
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_JPEG_DPG) {
+		jpeg_v5_0_0_stop_dpg_mode(adev, 0);
+	} else {
+
+		/* reset JMI */
+		WREG32_P(SOC15_REG_OFFSET(JPEG, 0, regUVD_JMI_CNTL),
+			UVD_JMI_CNTL__SOFT_RESET_MASK,
+			~UVD_JMI_CNTL__SOFT_RESET_MASK);
+
+		jpeg_v5_0_0_enable_clock_gating(adev);
+
+		/* enable power gating */
+		r = jpeg_v5_0_0_enable_power_gating(adev);
+		if (r)
+			return r;
+	}
+
+	if (adev->pm.dpm_enabled)
+		amdgpu_dpm_enable_jpeg(adev, false);
+
+	return 0;
+}
+
+/**
+ * jpeg_v5_0_0_dec_ring_get_rptr - get read pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware read pointer
+ */
+static uint64_t jpeg_v5_0_0_dec_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	return RREG32_SOC15(JPEG, 0, regUVD_JRBC_RB_RPTR);
+}
+
+/**
+ * jpeg_v5_0_0_dec_ring_get_wptr - get write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware write pointer
+ */
+static uint64_t jpeg_v5_0_0_dec_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->use_doorbell)
+		return *ring->wptr_cpu_addr;
+	else
+		return RREG32_SOC15(JPEG, 0, regUVD_JRBC_RB_WPTR);
+}
+
+/**
+ * jpeg_v5_0_0_dec_ring_set_wptr - set write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Commits the write pointer to the hardware
+ */
+static void jpeg_v5_0_0_dec_ring_set_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->use_doorbell) {
+		*ring->wptr_cpu_addr = lower_32_bits(ring->wptr);
+		WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr));
+	} else {
+		WREG32_SOC15(JPEG, 0, regUVD_JRBC_RB_WPTR, lower_32_bits(ring->wptr));
+	}
+}
+
+static bool jpeg_v5_0_0_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int ret = 1;
+
+	ret &= (((RREG32_SOC15(JPEG, 0, regUVD_JRBC_STATUS) &
+		UVD_JRBC_STATUS__RB_JOB_DONE_MASK) ==
+		UVD_JRBC_STATUS__RB_JOB_DONE_MASK));
+
+	return ret;
+}
+
+static int jpeg_v5_0_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	return SOC15_WAIT_ON_RREG(JPEG, 0, regUVD_JRBC_STATUS,
+		UVD_JRBC_STATUS__RB_JOB_DONE_MASK,
+		UVD_JRBC_STATUS__RB_JOB_DONE_MASK);
+}
+
+static int jpeg_v5_0_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_clockgating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	bool enable = state == AMD_CG_STATE_GATE;
+
+	if (enable) {
+		if (!jpeg_v5_0_0_is_idle(ip_block))
+			return -EBUSY;
+		jpeg_v5_0_0_enable_clock_gating(adev);
+	} else {
+		jpeg_v5_0_0_disable_clock_gating(adev);
+	}
+
+	return 0;
+}
+
+static int jpeg_v5_0_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_powergating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int ret;
+
+	if (state == adev->jpeg.cur_state)
+		return 0;
+
+	if (state == AMD_PG_STATE_GATE)
+		ret = jpeg_v5_0_0_stop(adev);
+	else
+		ret = jpeg_v5_0_0_start(adev);
+
+	if (!ret)
+		adev->jpeg.cur_state = state;
+
+	return ret;
+}
+
+static int jpeg_v5_0_0_set_interrupt_state(struct amdgpu_device *adev,
+					struct amdgpu_irq_src *source,
+					unsigned int type,
+					enum amdgpu_interrupt_state state)
+{
+	return 0;
+}
+
+static int jpeg_v5_0_0_process_interrupt(struct amdgpu_device *adev,
+				      struct amdgpu_irq_src *source,
+				      struct amdgpu_iv_entry *entry)
+{
+	DRM_DEBUG("IH: JPEG TRAP\n");
+
+	switch (entry->src_id) {
+	case VCN_5_0__SRCID__JPEG_DECODE:
+		amdgpu_fence_process(adev->jpeg.inst->ring_dec);
+		break;
+	default:
+		DRM_DEV_ERROR(adev->dev, "Unhandled interrupt: %d %d\n",
+			  entry->src_id, entry->src_data[0]);
+		break;
+	}
+
+	return 0;
+}
+
+static int jpeg_v5_0_0_ring_reset(struct amdgpu_ring *ring,
+				  unsigned int vmid,
+				  struct amdgpu_fence *timedout_fence)
+{
+	int r;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+	r = jpeg_v5_0_0_stop(ring->adev);
+	if (r)
+		return r;
+	r = jpeg_v5_0_0_start(ring->adev);
+	if (r)
+		return r;
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static const struct amd_ip_funcs jpeg_v5_0_0_ip_funcs = {
+	.name = "jpeg_v5_0_0",
+	.early_init = jpeg_v5_0_0_early_init,
+	.sw_init = jpeg_v5_0_0_sw_init,
+	.sw_fini = jpeg_v5_0_0_sw_fini,
+	.hw_init = jpeg_v5_0_0_hw_init,
+	.hw_fini = jpeg_v5_0_0_hw_fini,
+	.suspend = jpeg_v5_0_0_suspend,
+	.resume = jpeg_v5_0_0_resume,
+	.is_idle = jpeg_v5_0_0_is_idle,
+	.wait_for_idle = jpeg_v5_0_0_wait_for_idle,
+	.set_clockgating_state = jpeg_v5_0_0_set_clockgating_state,
+	.set_powergating_state = jpeg_v5_0_0_set_powergating_state,
+	.dump_ip_state = amdgpu_jpeg_dump_ip_state,
+	.print_ip_state = amdgpu_jpeg_print_ip_state,
+};
+
+static const struct amdgpu_ring_funcs jpeg_v5_0_0_dec_ring_vm_funcs = {
+	.type = AMDGPU_RING_TYPE_VCN_JPEG,
+	.align_mask = 0xf,
+	.get_rptr = jpeg_v5_0_0_dec_ring_get_rptr,
+	.get_wptr = jpeg_v5_0_0_dec_ring_get_wptr,
+	.set_wptr = jpeg_v5_0_0_dec_ring_set_wptr,
+	.parse_cs = amdgpu_jpeg_dec_parse_cs,
+	.emit_frame_size =
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 +
+		8 + /* jpeg_v5_0_0_dec_ring_emit_vm_flush */
+		22 + 22 + /* jpeg_v5_0_0_dec_ring_emit_fence x2 vm fence */
+		8 + 16,
+	.emit_ib_size = 22, /* jpeg_v5_0_0_dec_ring_emit_ib */
+	.emit_ib = jpeg_v4_0_3_dec_ring_emit_ib,
+	.emit_fence = jpeg_v4_0_3_dec_ring_emit_fence,
+	.emit_vm_flush = jpeg_v4_0_3_dec_ring_emit_vm_flush,
+	.test_ring = amdgpu_jpeg_dec_ring_test_ring,
+	.test_ib = amdgpu_jpeg_dec_ring_test_ib,
+	.insert_nop = jpeg_v4_0_3_dec_ring_nop,
+	.insert_start = jpeg_v4_0_3_dec_ring_insert_start,
+	.insert_end = jpeg_v4_0_3_dec_ring_insert_end,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.begin_use = amdgpu_jpeg_ring_begin_use,
+	.end_use = amdgpu_jpeg_ring_end_use,
+	.emit_wreg = jpeg_v4_0_3_dec_ring_emit_wreg,
+	.emit_reg_wait = jpeg_v4_0_3_dec_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = jpeg_v5_0_0_ring_reset,
+};
+
+static void jpeg_v5_0_0_set_dec_ring_funcs(struct amdgpu_device *adev)
+{
+	adev->jpeg.inst->ring_dec->funcs = &jpeg_v5_0_0_dec_ring_vm_funcs;
+}
+
+static const struct amdgpu_irq_src_funcs jpeg_v5_0_0_irq_funcs = {
+	.set = jpeg_v5_0_0_set_interrupt_state,
+	.process = jpeg_v5_0_0_process_interrupt,
+};
+
+static void jpeg_v5_0_0_set_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->jpeg.inst->irq.num_types = 1;
+	adev->jpeg.inst->irq.funcs = &jpeg_v5_0_0_irq_funcs;
+}
+
+const struct amdgpu_ip_block_version jpeg_v5_0_0_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_JPEG,
+	.major = 5,
+	.minor = 0,
+	.rev = 0,
+	.funcs = &jpeg_v5_0_0_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_0.h b/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_0.h
new file mode 100644
index 000000000000..5abb96159814
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_0.h
@@ -0,0 +1,35 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __JPEG_V5_0_0_H__
+#define __JPEG_V5_0_0_H__
+
+#define vcnipJPEG_CGC_GATE                                 0x4160
+#define vcnipJPEG_CGC_CTRL                                 0x4161
+#define vcnipJPEG_SYS_INT_EN                               0x4141
+#define vcnipUVD_NO_OP                                     0x0029
+#define vcnipJPEG_DEC_GFX10_ADDR_CONFIG                    0x404A
+
+extern const struct amdgpu_ip_block_version jpeg_v5_0_0_ip_block;
+
+#endif /* __JPEG_V5_0_0_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_1.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_1.c
new file mode 100644
index 000000000000..ab0bf880d3d8
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_1.c
@@ -0,0 +1,1101 @@
+// SPDX-License-Identifier: GPL-2.0 OR MIT
+/*
+ * Copyright 2014-2024 Advanced Micro Devices, Inc. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "amdgpu.h"
+#include "amdgpu_jpeg.h"
+#include "amdgpu_pm.h"
+#include "soc15.h"
+#include "soc15d.h"
+#include "jpeg_v4_0_3.h"
+#include "jpeg_v5_0_1.h"
+#include "mmsch_v5_0.h"
+
+#include "vcn/vcn_5_0_0_offset.h"
+#include "vcn/vcn_5_0_0_sh_mask.h"
+#include "ivsrcid/vcn/irqsrcs_vcn_5_0.h"
+
+static int jpeg_v5_0_1_start_sriov(struct amdgpu_device *adev);
+static void jpeg_v5_0_1_set_dec_ring_funcs(struct amdgpu_device *adev);
+static void jpeg_v5_0_1_set_irq_funcs(struct amdgpu_device *adev);
+static int jpeg_v5_0_1_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					     enum amd_powergating_state state);
+static void jpeg_v5_0_1_set_ras_funcs(struct amdgpu_device *adev);
+static void jpeg_v5_0_1_dec_ring_set_wptr(struct amdgpu_ring *ring);
+
+static int amdgpu_ih_srcid_jpeg[] = {
+	VCN_5_0__SRCID__JPEG_DECODE,
+	VCN_5_0__SRCID__JPEG1_DECODE,
+	VCN_5_0__SRCID__JPEG2_DECODE,
+	VCN_5_0__SRCID__JPEG3_DECODE,
+	VCN_5_0__SRCID__JPEG4_DECODE,
+	VCN_5_0__SRCID__JPEG5_DECODE,
+	VCN_5_0__SRCID__JPEG6_DECODE,
+	VCN_5_0__SRCID__JPEG7_DECODE,
+	VCN_5_0__SRCID__JPEG8_DECODE,
+	VCN_5_0__SRCID__JPEG9_DECODE,
+};
+
+static const struct amdgpu_hwip_reg_entry jpeg_reg_list_5_0_1[] = {
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_POWER_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_INT_STAT),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC0_UVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC0_UVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC0_UVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_ADDR_MODE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_GFX10_ADDR_CONFIG),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_Y_GFX10_TILING_SURFACE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_UV_GFX10_TILING_SURFACE),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_PITCH),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_UV_PITCH),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC1_UVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC1_UVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC1_UVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC2_UVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC2_UVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC2_UVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC3_UVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC3_UVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC3_UVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC4_UVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC4_UVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC4_UVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC5_UVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC5_UVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC5_UVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC6_UVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC6_UVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC6_UVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC7_UVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC7_UVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC7_UVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC8_UVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC8_UVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC8_UVD_JRBC_STATUS),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC9_UVD_JRBC_RB_RPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC9_UVD_JRBC_RB_WPTR),
+	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC9_UVD_JRBC_STATUS),
+};
+
+static int jpeg_v5_0_1_core_reg_offset(u32 pipe)
+{
+	if (pipe <= AMDGPU_MAX_JPEG_RINGS_4_0_3)
+		return ((0x40 * pipe) - 0xc80);
+	else
+		return ((0x40 * pipe) - 0x440);
+}
+
+/**
+ * jpeg_v5_0_1_early_init - set function pointers
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Set ring and irq function pointers
+ */
+static int jpeg_v5_0_1_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (!adev->jpeg.num_jpeg_inst || adev->jpeg.num_jpeg_inst > AMDGPU_MAX_JPEG_INSTANCES)
+		return -ENOENT;
+
+	adev->jpeg.num_jpeg_rings = AMDGPU_MAX_JPEG_RINGS;
+	jpeg_v5_0_1_set_dec_ring_funcs(adev);
+	jpeg_v5_0_1_set_irq_funcs(adev);
+	jpeg_v5_0_1_set_ras_funcs(adev);
+
+	return 0;
+}
+
+/**
+ * jpeg_v5_0_1_sw_init - sw init for JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Load firmware and sw initialization
+ */
+static int jpeg_v5_0_1_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring;
+	int i, j, r, jpeg_inst;
+
+	for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+		/* JPEG TRAP */
+		r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
+				      amdgpu_ih_srcid_jpeg[j], &adev->jpeg.inst->irq);
+		if (r)
+			return r;
+	}
+	/* JPEG DJPEG POISON EVENT */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
+			VCN_5_0__SRCID_DJPEG0_POISON, &adev->jpeg.inst->ras_poison_irq);
+	if (r)
+		return r;
+
+	/* JPEG EJPEG POISON EVENT */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
+			VCN_5_0__SRCID_EJPEG0_POISON, &adev->jpeg.inst->ras_poison_irq);
+	if (r)
+		return r;
+
+	r = amdgpu_jpeg_sw_init(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_jpeg_resume(adev);
+	if (r)
+		return r;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		jpeg_inst = GET_INST(JPEG, i);
+
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+			ring = &adev->jpeg.inst[i].ring_dec[j];
+			ring->use_doorbell = true;
+			ring->vm_hub = AMDGPU_MMHUB0(adev->jpeg.inst[i].aid_id);
+			if (!amdgpu_sriov_vf(adev)) {
+				ring->doorbell_index =
+					(adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
+					1 + j + 11 * jpeg_inst;
+			} else {
+				ring->doorbell_index =
+					(adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
+					2 + j + 32 * jpeg_inst;
+			}
+			sprintf(ring->name, "jpeg_dec_%d.%d", adev->jpeg.inst[i].aid_id, j);
+			r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq, 0,
+					     AMDGPU_RING_PRIO_DEFAULT, NULL);
+			if (r)
+				return r;
+
+			adev->jpeg.internal.jpeg_pitch[j] =
+				regUVD_JRBC0_UVD_JRBC_SCRATCH0_INTERNAL_OFFSET;
+			adev->jpeg.inst[i].external.jpeg_pitch[j] =
+				SOC15_REG_OFFSET1(JPEG, jpeg_inst, regUVD_JRBC_SCRATCH0,
+						  (j ? jpeg_v5_0_1_core_reg_offset(j) : 0));
+		}
+	}
+
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__JPEG)) {
+		r = amdgpu_jpeg_ras_sw_init(adev);
+		if (r) {
+			dev_err(adev->dev, "Failed to initialize jpeg ras block!\n");
+			return r;
+		}
+	}
+
+	r = amdgpu_jpeg_reg_dump_init(adev, jpeg_reg_list_5_0_1, ARRAY_SIZE(jpeg_reg_list_5_0_1));
+	if (r)
+		return r;
+
+	adev->jpeg.supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->jpeg.inst[0].ring_dec[0]);
+	if (!amdgpu_sriov_vf(adev))
+		adev->jpeg.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+	r = amdgpu_jpeg_sysfs_reset_mask_init(adev);
+
+	return r;
+}
+
+/**
+ * jpeg_v5_0_1_sw_fini - sw fini for JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * JPEG suspend and free up sw allocation
+ */
+static int jpeg_v5_0_1_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = amdgpu_jpeg_suspend(adev);
+	if (r)
+		return r;
+
+	amdgpu_jpeg_sysfs_reset_mask_fini(adev);
+
+	r = amdgpu_jpeg_sw_fini(adev);
+
+	return r;
+}
+
+/**
+ * jpeg_v5_0_1_hw_init - start and test JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ */
+static int jpeg_v5_0_1_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring;
+	int i, j, r, jpeg_inst;
+
+	if (amdgpu_sriov_vf(adev)) {
+		r = jpeg_v5_0_1_start_sriov(adev);
+		if (r)
+			return r;
+
+		for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+			for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+				ring = &adev->jpeg.inst[i].ring_dec[j];
+				ring->wptr = 0;
+				ring->wptr_old = 0;
+				jpeg_v5_0_1_dec_ring_set_wptr(ring);
+				ring->sched.ready = true;
+			}
+		}
+		return 0;
+	}
+	if (RREG32_SOC15(VCN, GET_INST(VCN, 0), regVCN_RRMT_CNTL) & 0x100)
+		adev->jpeg.caps |= AMDGPU_JPEG_CAPS(RRMT_ENABLED);
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		jpeg_inst = GET_INST(JPEG, i);
+		ring = adev->jpeg.inst[i].ring_dec;
+		if (ring->use_doorbell)
+			adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell,
+				 (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 11 * jpeg_inst,
+				 adev->jpeg.inst[i].aid_id);
+
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+			ring = &adev->jpeg.inst[i].ring_dec[j];
+			if (ring->use_doorbell)
+				WREG32_SOC15_OFFSET(VCN, GET_INST(VCN, i), regVCN_JPEG_DB_CTRL,
+						    ring->pipe,
+						    ring->doorbell_index <<
+						    VCN_JPEG_DB_CTRL__OFFSET__SHIFT |
+						    VCN_JPEG_DB_CTRL__EN_MASK);
+			r = amdgpu_ring_test_helper(ring);
+			if (r)
+				return r;
+		}
+	}
+
+	return 0;
+}
+
+/**
+ * jpeg_v5_0_1_hw_fini - stop the hardware block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Stop the JPEG block, mark ring as not ready any more
+ */
+static int jpeg_v5_0_1_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int ret = 0;
+
+	cancel_delayed_work_sync(&adev->jpeg.idle_work);
+
+	if (!amdgpu_sriov_vf(adev)) {
+		if (adev->jpeg.cur_state != AMD_PG_STATE_GATE)
+			ret = jpeg_v5_0_1_set_powergating_state(ip_block, AMD_PG_STATE_GATE);
+	}
+
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__JPEG) && !amdgpu_sriov_vf(adev))
+		amdgpu_irq_put(adev, &adev->jpeg.inst->ras_poison_irq, 0);
+
+	return ret;
+}
+
+/**
+ * jpeg_v5_0_1_suspend - suspend JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * HW fini and suspend JPEG block
+ */
+static int jpeg_v5_0_1_suspend(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = jpeg_v5_0_1_hw_fini(ip_block);
+	if (r)
+		return r;
+
+	r = amdgpu_jpeg_suspend(adev);
+
+	return r;
+}
+
+/**
+ * jpeg_v5_0_1_resume - resume JPEG block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Resume firmware and hw init JPEG block
+ */
+static int jpeg_v5_0_1_resume(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = amdgpu_jpeg_resume(adev);
+	if (r)
+		return r;
+
+	r = jpeg_v5_0_1_hw_init(ip_block);
+
+	return r;
+}
+
+static void jpeg_v5_0_1_init_inst(struct amdgpu_device *adev, int i)
+{
+	int jpeg_inst = GET_INST(JPEG, i);
+
+	/* disable anti hang mechanism */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst, regUVD_JPEG_POWER_STATUS), 0,
+		 ~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK);
+
+	/* keep the JPEG in static PG mode */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst, regUVD_JPEG_POWER_STATUS), 0,
+		 ~UVD_JPEG_POWER_STATUS__JPEG_PG_MODE_MASK);
+
+	/* MJPEG global tiling registers */
+	WREG32_SOC15(JPEG, 0, regJPEG_DEC_GFX10_ADDR_CONFIG,
+		     adev->gfx.config.gb_addr_config);
+
+	/* enable JMI channel */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst, regUVD_JMI_CNTL), 0,
+		 ~UVD_JMI_CNTL__SOFT_RESET_MASK);
+}
+
+static void jpeg_v5_0_1_deinit_inst(struct amdgpu_device *adev, int i)
+{
+	int jpeg_inst = GET_INST(JPEG, i);
+	/* reset JMI */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst, regUVD_JMI_CNTL),
+		 UVD_JMI_CNTL__SOFT_RESET_MASK,
+		 ~UVD_JMI_CNTL__SOFT_RESET_MASK);
+
+	/* enable anti hang mechanism */
+	WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst, regUVD_JPEG_POWER_STATUS),
+		 UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK,
+		 ~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK);
+}
+
+static void jpeg_v5_0_1_init_jrbc(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 reg, data, mask;
+	int jpeg_inst = GET_INST(JPEG, ring->me);
+	int reg_offset = ring->pipe ? jpeg_v5_0_1_core_reg_offset(ring->pipe) : 0;
+
+	/* enable System Interrupt for JRBC */
+	reg = SOC15_REG_OFFSET(JPEG, jpeg_inst, regJPEG_SYS_INT_EN);
+	if (ring->pipe < AMDGPU_MAX_JPEG_RINGS_4_0_3) {
+		data = JPEG_SYS_INT_EN__DJRBC0_MASK << ring->pipe;
+		mask = ~(JPEG_SYS_INT_EN__DJRBC0_MASK << ring->pipe);
+		WREG32_P(reg, data, mask);
+	} else {
+		data = JPEG_SYS_INT_EN__DJRBC0_MASK << (ring->pipe+12);
+		mask = ~(JPEG_SYS_INT_EN__DJRBC0_MASK << (ring->pipe+12));
+		WREG32_P(reg, data, mask);
+	}
+
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_LMI_JRBC_RB_VMID,
+			    reg_offset, 0);
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JRBC_RB_CNTL,
+			    reg_offset,
+			    (0x00000001L | 0x00000002L));
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_LMI_JRBC_RB_64BIT_BAR_LOW,
+			    reg_offset, lower_32_bits(ring->gpu_addr));
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_LMI_JRBC_RB_64BIT_BAR_HIGH,
+			    reg_offset, upper_32_bits(ring->gpu_addr));
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JRBC_RB_RPTR,
+			    reg_offset, 0);
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JRBC_RB_WPTR,
+			    reg_offset, 0);
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JRBC_RB_CNTL,
+			    reg_offset, 0x00000002L);
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JRBC_RB_SIZE,
+			    reg_offset, ring->ring_size / 4);
+	ring->wptr = RREG32_SOC15_OFFSET(JPEG, jpeg_inst, regUVD_JRBC_RB_WPTR,
+					 reg_offset);
+}
+
+static int jpeg_v5_0_1_start_sriov(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	uint64_t ctx_addr;
+	uint32_t param, resp, expected;
+	uint32_t tmp, timeout;
+
+	struct amdgpu_mm_table *table = &adev->virt.mm_table;
+	uint32_t *table_loc;
+	uint32_t table_size;
+	uint32_t size, size_dw, item_offset;
+	uint32_t init_status;
+	int i, j, jpeg_inst;
+
+	struct mmsch_v5_0_cmd_direct_write
+		direct_wt = { {0} };
+	struct mmsch_v5_0_cmd_end end = { {0} };
+	struct mmsch_v5_0_init_header header;
+
+	direct_wt.cmd_header.command_type =
+		MMSCH_COMMAND__DIRECT_REG_WRITE;
+	end.cmd_header.command_type =
+		MMSCH_COMMAND__END;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; i++) {
+		jpeg_inst = GET_INST(JPEG, i);
+
+		memset(&header, 0, sizeof(struct mmsch_v5_0_init_header));
+		header.version = MMSCH_VERSION;
+		header.total_size = sizeof(struct mmsch_v5_0_init_header) >> 2;
+
+		table_loc = (uint32_t *)table->cpu_addr;
+		table_loc += header.total_size;
+
+		item_offset = header.total_size;
+
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; j++) {
+			ring = &adev->jpeg.inst[i].ring_dec[j];
+			table_size = 0;
+
+			tmp = SOC15_REG_OFFSET(JPEG, 0, regUVD_LMI_JRBC_RB_64BIT_BAR_LOW);
+			MMSCH_V5_0_INSERT_DIRECT_WT(tmp, lower_32_bits(ring->gpu_addr));
+			tmp = SOC15_REG_OFFSET(JPEG, 0, regUVD_LMI_JRBC_RB_64BIT_BAR_HIGH);
+			MMSCH_V5_0_INSERT_DIRECT_WT(tmp, upper_32_bits(ring->gpu_addr));
+			tmp = SOC15_REG_OFFSET(JPEG, 0, regUVD_JRBC_RB_SIZE);
+			MMSCH_V5_0_INSERT_DIRECT_WT(tmp, ring->ring_size / 4);
+
+			if (j < 5) {
+				header.mjpegdec0[j].table_offset = item_offset;
+				header.mjpegdec0[j].init_status = 0;
+				header.mjpegdec0[j].table_size = table_size;
+			} else {
+				header.mjpegdec1[j - 5].table_offset = item_offset;
+				header.mjpegdec1[j - 5].init_status = 0;
+				header.mjpegdec1[j - 5].table_size = table_size;
+			}
+			header.total_size += table_size;
+			item_offset += table_size;
+		}
+
+		MMSCH_V5_0_INSERT_END();
+
+		/* send init table to MMSCH */
+		size = sizeof(struct mmsch_v5_0_init_header);
+		table_loc = (uint32_t *)table->cpu_addr;
+		memcpy((void *)table_loc, &header, size);
+
+		ctx_addr = table->gpu_addr;
+		WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_CTX_ADDR_LO, lower_32_bits(ctx_addr));
+		WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_CTX_ADDR_HI, upper_32_bits(ctx_addr));
+
+		tmp = RREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_VMID);
+		tmp &= ~MMSCH_VF_VMID__VF_CTX_VMID_MASK;
+		tmp |= (0 << MMSCH_VF_VMID__VF_CTX_VMID__SHIFT);
+		WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_VMID, tmp);
+
+		size = header.total_size;
+		WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_CTX_SIZE, size);
+
+		WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_MAILBOX_RESP, 0);
+
+		param = 0x00000001;
+		WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_MAILBOX_HOST, param);
+		tmp = 0;
+		timeout = 1000;
+		resp = 0;
+		expected = MMSCH_VF_MAILBOX_RESP__OK;
+		init_status =
+			((struct mmsch_v5_0_init_header *)(table_loc))->mjpegdec0[i].init_status;
+		while (resp != expected) {
+			resp = RREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_MAILBOX_RESP);
+
+			if (resp != 0)
+				break;
+			udelay(10);
+			tmp = tmp + 10;
+			if (tmp >= timeout) {
+				DRM_ERROR("failed to init MMSCH. TIME-OUT after %d usec"\
+					" waiting for regMMSCH_VF_MAILBOX_RESP "\
+					"(expected=0x%08x, readback=0x%08x)\n",
+					tmp, expected, resp);
+				return -EBUSY;
+			}
+		}
+		if (resp != expected && resp != MMSCH_VF_MAILBOX_RESP__INCOMPLETE &&
+				init_status != MMSCH_VF_ENGINE_STATUS__PASS)
+			DRM_ERROR("MMSCH init status is incorrect! readback=0x%08x, header init status for jpeg: %x\n",
+					resp, init_status);
+
+	}
+	return 0;
+}
+
+/**
+ * jpeg_v5_0_1_start - start JPEG block
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Setup and start the JPEG block
+ */
+static int jpeg_v5_0_1_start(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	int i, j;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		jpeg_v5_0_1_init_inst(adev, i);
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+			ring = &adev->jpeg.inst[i].ring_dec[j];
+			jpeg_v5_0_1_init_jrbc(ring);
+		}
+	}
+
+	return 0;
+}
+
+/**
+ * jpeg_v5_0_1_stop - stop JPEG block
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * stop the JPEG block
+ */
+static int jpeg_v5_0_1_stop(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i)
+		jpeg_v5_0_1_deinit_inst(adev, i);
+
+	return 0;
+}
+
+/**
+ * jpeg_v5_0_1_dec_ring_get_rptr - get read pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware read pointer
+ */
+static uint64_t jpeg_v5_0_1_dec_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	return RREG32_SOC15_OFFSET(JPEG, GET_INST(JPEG, ring->me), regUVD_JRBC_RB_RPTR,
+				   ring->pipe ? jpeg_v5_0_1_core_reg_offset(ring->pipe) : 0);
+}
+
+/**
+ * jpeg_v5_0_1_dec_ring_get_wptr - get write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware write pointer
+ */
+static uint64_t jpeg_v5_0_1_dec_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->use_doorbell)
+		return adev->wb.wb[ring->wptr_offs];
+
+	return RREG32_SOC15_OFFSET(JPEG, GET_INST(JPEG, ring->me), regUVD_JRBC_RB_WPTR,
+				   ring->pipe ? jpeg_v5_0_1_core_reg_offset(ring->pipe) : 0);
+}
+
+/**
+ * jpeg_v5_0_1_dec_ring_set_wptr - set write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Commits the write pointer to the hardware
+ */
+static void jpeg_v5_0_1_dec_ring_set_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->use_doorbell) {
+		adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr);
+		WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr));
+	} else {
+		WREG32_SOC15_OFFSET(JPEG, GET_INST(JPEG, ring->me),
+				    regUVD_JRBC_RB_WPTR,
+				    (ring->pipe ? jpeg_v5_0_1_core_reg_offset(ring->pipe) : 0),
+				    lower_32_bits(ring->wptr));
+	}
+}
+
+static bool jpeg_v5_0_1_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	bool ret = false;
+	int i, j;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+			int reg_offset = (j ? jpeg_v5_0_1_core_reg_offset(j) : 0);
+
+			ret &= ((RREG32_SOC15_OFFSET(JPEG, GET_INST(JPEG, i),
+				regUVD_JRBC_STATUS, reg_offset) &
+				UVD_JRBC_STATUS__RB_JOB_DONE_MASK) ==
+				UVD_JRBC_STATUS__RB_JOB_DONE_MASK);
+		}
+	}
+
+	return ret;
+}
+
+static int jpeg_v5_0_1_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int ret = 0;
+	int i, j;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+			int reg_offset = (j ? jpeg_v5_0_1_core_reg_offset(j) : 0);
+
+			ret &= SOC15_WAIT_ON_RREG_OFFSET(JPEG, GET_INST(JPEG, i),
+							 regUVD_JRBC_STATUS, reg_offset,
+							 UVD_JRBC_STATUS__RB_JOB_DONE_MASK,
+							 UVD_JRBC_STATUS__RB_JOB_DONE_MASK);
+		}
+	}
+	return ret;
+}
+
+static int jpeg_v5_0_1_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					     enum amd_clockgating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	bool enable = state == AMD_CG_STATE_GATE;
+
+	int i;
+
+	if (!enable)
+		return 0;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		if (!jpeg_v5_0_1_is_idle(ip_block))
+			return -EBUSY;
+	}
+
+	return 0;
+}
+
+static int jpeg_v5_0_1_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					     enum amd_powergating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int ret;
+
+	if (amdgpu_sriov_vf(adev)) {
+		adev->jpeg.cur_state = AMD_PG_STATE_UNGATE;
+		return 0;
+	}
+
+	if (state == adev->jpeg.cur_state)
+		return 0;
+
+	if (state == AMD_PG_STATE_GATE)
+		ret = jpeg_v5_0_1_stop(adev);
+	else
+		ret = jpeg_v5_0_1_start(adev);
+
+	if (!ret)
+		adev->jpeg.cur_state = state;
+
+	return ret;
+}
+
+static int jpeg_v5_0_1_set_interrupt_state(struct amdgpu_device *adev,
+					   struct amdgpu_irq_src *source,
+					   unsigned int type,
+					   enum amdgpu_interrupt_state state)
+{
+	return 0;
+}
+
+static int jpeg_v5_0_1_set_ras_interrupt_state(struct amdgpu_device *adev,
+					struct amdgpu_irq_src *source,
+					unsigned int type,
+					enum amdgpu_interrupt_state state)
+{
+	return 0;
+}
+
+
+
+static int jpeg_v5_0_1_process_interrupt(struct amdgpu_device *adev,
+					 struct amdgpu_irq_src *source,
+					 struct amdgpu_iv_entry *entry)
+{
+	u32 i, inst;
+
+	i = node_id_to_phys_map[entry->node_id];
+	DRM_DEV_DEBUG(adev->dev, "IH: JPEG TRAP\n");
+
+	for (inst = 0; inst < adev->jpeg.num_jpeg_inst; ++inst)
+		if (adev->jpeg.inst[inst].aid_id == i)
+			break;
+
+	if (inst >= adev->jpeg.num_jpeg_inst) {
+		dev_WARN_ONCE(adev->dev, 1,
+			      "Interrupt received for unknown JPEG instance %d",
+			      entry->node_id);
+		return 0;
+	}
+
+	switch (entry->src_id) {
+	case VCN_5_0__SRCID__JPEG_DECODE:
+		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[0]);
+		break;
+	case VCN_5_0__SRCID__JPEG1_DECODE:
+		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[1]);
+		break;
+	case VCN_5_0__SRCID__JPEG2_DECODE:
+		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[2]);
+		break;
+	case VCN_5_0__SRCID__JPEG3_DECODE:
+		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[3]);
+		break;
+	case VCN_5_0__SRCID__JPEG4_DECODE:
+		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[4]);
+		break;
+	case VCN_5_0__SRCID__JPEG5_DECODE:
+		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[5]);
+		break;
+	case VCN_5_0__SRCID__JPEG6_DECODE:
+		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[6]);
+		break;
+	case VCN_5_0__SRCID__JPEG7_DECODE:
+		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[7]);
+		break;
+	case VCN_5_0__SRCID__JPEG8_DECODE:
+		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[8]);
+		break;
+	case VCN_5_0__SRCID__JPEG9_DECODE:
+		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[9]);
+		break;
+	default:
+		DRM_DEV_ERROR(adev->dev, "Unhandled interrupt: %d %d\n",
+			      entry->src_id, entry->src_data[0]);
+		break;
+	}
+
+	return 0;
+}
+
+static void jpeg_v5_0_1_core_stall_reset(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int jpeg_inst = GET_INST(JPEG, ring->me);
+	int reg_offset = ring->pipe ? jpeg_v5_0_1_core_reg_offset(ring->pipe) : 0;
+
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JMI0_UVD_JMI_CLIENT_STALL,
+			    reg_offset, 0x1F);
+	SOC15_WAIT_ON_RREG_OFFSET(JPEG, jpeg_inst,
+				  regUVD_JMI0_UVD_JMI_CLIENT_CLEAN_STATUS,
+				  reg_offset, 0x1F, 0x1F);
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JMI0_JPEG_LMI_DROP,
+			    reg_offset, 0x1F);
+	WREG32_SOC15(JPEG, jpeg_inst, regJPEG_CORE_RST_CTRL, 1 << ring->pipe);
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JMI0_UVD_JMI_CLIENT_STALL,
+			    reg_offset, 0x00);
+	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
+			    regUVD_JMI0_JPEG_LMI_DROP,
+			    reg_offset, 0x00);
+	WREG32_SOC15(JPEG, jpeg_inst, regJPEG_CORE_RST_CTRL, 0x00);
+}
+
+static int jpeg_v5_0_1_ring_reset(struct amdgpu_ring *ring,
+				  unsigned int vmid,
+				  struct amdgpu_fence *timedout_fence)
+{
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+	jpeg_v5_0_1_core_stall_reset(ring);
+	jpeg_v5_0_1_init_jrbc(ring);
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static const struct amd_ip_funcs jpeg_v5_0_1_ip_funcs = {
+	.name = "jpeg_v5_0_1",
+	.early_init = jpeg_v5_0_1_early_init,
+	.late_init = NULL,
+	.sw_init = jpeg_v5_0_1_sw_init,
+	.sw_fini = jpeg_v5_0_1_sw_fini,
+	.hw_init = jpeg_v5_0_1_hw_init,
+	.hw_fini = jpeg_v5_0_1_hw_fini,
+	.suspend = jpeg_v5_0_1_suspend,
+	.resume = jpeg_v5_0_1_resume,
+	.is_idle = jpeg_v5_0_1_is_idle,
+	.wait_for_idle = jpeg_v5_0_1_wait_for_idle,
+	.check_soft_reset = NULL,
+	.pre_soft_reset = NULL,
+	.soft_reset = NULL,
+	.post_soft_reset = NULL,
+	.set_clockgating_state = jpeg_v5_0_1_set_clockgating_state,
+	.set_powergating_state = jpeg_v5_0_1_set_powergating_state,
+	.dump_ip_state = amdgpu_jpeg_dump_ip_state,
+	.print_ip_state = amdgpu_jpeg_print_ip_state,
+};
+
+static const struct amdgpu_ring_funcs jpeg_v5_0_1_dec_ring_vm_funcs = {
+	.type = AMDGPU_RING_TYPE_VCN_JPEG,
+	.align_mask = 0xf,
+	.get_rptr = jpeg_v5_0_1_dec_ring_get_rptr,
+	.get_wptr = jpeg_v5_0_1_dec_ring_get_wptr,
+	.set_wptr = jpeg_v5_0_1_dec_ring_set_wptr,
+	.parse_cs = amdgpu_jpeg_dec_parse_cs,
+	.emit_frame_size =
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 +
+		8 + /* jpeg_v5_0_1_dec_ring_emit_vm_flush */
+		22 + 22 + /* jpeg_v5_0_1_dec_ring_emit_fence x2 vm fence */
+		8 + 16,
+	.emit_ib_size = 22, /* jpeg_v5_0_1_dec_ring_emit_ib */
+	.emit_ib = jpeg_v4_0_3_dec_ring_emit_ib,
+	.emit_fence = jpeg_v4_0_3_dec_ring_emit_fence,
+	.emit_vm_flush = jpeg_v4_0_3_dec_ring_emit_vm_flush,
+	.emit_hdp_flush = jpeg_v4_0_3_ring_emit_hdp_flush,
+	.test_ring = amdgpu_jpeg_dec_ring_test_ring,
+	.test_ib = amdgpu_jpeg_dec_ring_test_ib,
+	.insert_nop = jpeg_v4_0_3_dec_ring_nop,
+	.insert_start = jpeg_v4_0_3_dec_ring_insert_start,
+	.insert_end = jpeg_v4_0_3_dec_ring_insert_end,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.begin_use = amdgpu_jpeg_ring_begin_use,
+	.end_use = amdgpu_jpeg_ring_end_use,
+	.emit_wreg = jpeg_v4_0_3_dec_ring_emit_wreg,
+	.emit_reg_wait = jpeg_v4_0_3_dec_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = jpeg_v5_0_1_ring_reset,
+};
+
+static void jpeg_v5_0_1_set_dec_ring_funcs(struct amdgpu_device *adev)
+{
+	int i, j, jpeg_inst;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+			adev->jpeg.inst[i].ring_dec[j].funcs = &jpeg_v5_0_1_dec_ring_vm_funcs;
+			adev->jpeg.inst[i].ring_dec[j].me = i;
+			adev->jpeg.inst[i].ring_dec[j].pipe = j;
+		}
+		jpeg_inst = GET_INST(JPEG, i);
+		adev->jpeg.inst[i].aid_id =
+			jpeg_inst / adev->jpeg.num_inst_per_aid;
+	}
+}
+
+static const struct amdgpu_irq_src_funcs jpeg_v5_0_1_irq_funcs = {
+	.set = jpeg_v5_0_1_set_interrupt_state,
+	.process = jpeg_v5_0_1_process_interrupt,
+};
+
+static const struct amdgpu_irq_src_funcs jpeg_v5_0_1_ras_irq_funcs = {
+	.set = jpeg_v5_0_1_set_ras_interrupt_state,
+	.process = amdgpu_jpeg_process_poison_irq,
+};
+
+static void jpeg_v5_0_1_set_irq_funcs(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i)
+		adev->jpeg.inst->irq.num_types += adev->jpeg.num_jpeg_rings;
+
+	adev->jpeg.inst->irq.funcs = &jpeg_v5_0_1_irq_funcs;
+
+	adev->jpeg.inst->ras_poison_irq.num_types = 1;
+	adev->jpeg.inst->ras_poison_irq.funcs = &jpeg_v5_0_1_ras_irq_funcs;
+
+}
+
+const struct amdgpu_ip_block_version jpeg_v5_0_1_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_JPEG,
+	.major = 5,
+	.minor = 0,
+	.rev = 1,
+	.funcs = &jpeg_v5_0_1_ip_funcs,
+};
+
+static uint32_t jpeg_v5_0_1_query_poison_by_instance(struct amdgpu_device *adev,
+		uint32_t instance, uint32_t sub_block)
+{
+	uint32_t poison_stat = 0, reg_value = 0;
+
+	switch (sub_block) {
+	case AMDGPU_JPEG_V5_0_1_JPEG0:
+		reg_value = RREG32_SOC15(JPEG, instance, regUVD_RAS_JPEG0_STATUS);
+		poison_stat = REG_GET_FIELD(reg_value, UVD_RAS_JPEG0_STATUS, POISONED_PF);
+		break;
+	case AMDGPU_JPEG_V5_0_1_JPEG1:
+		reg_value = RREG32_SOC15(JPEG, instance, regUVD_RAS_JPEG1_STATUS);
+		poison_stat = REG_GET_FIELD(reg_value, UVD_RAS_JPEG1_STATUS, POISONED_PF);
+		break;
+	default:
+		break;
+	}
+
+	if (poison_stat)
+		dev_info(adev->dev, "Poison detected in JPEG%d sub_block%d\n",
+			instance, sub_block);
+
+	return poison_stat;
+}
+
+static bool jpeg_v5_0_1_query_ras_poison_status(struct amdgpu_device *adev)
+{
+	uint32_t inst = 0, sub = 0, poison_stat = 0;
+
+	for (inst = 0; inst < adev->jpeg.num_jpeg_inst; inst++)
+		for (sub = 0; sub < AMDGPU_JPEG_V5_0_1_MAX_SUB_BLOCK; sub++)
+			poison_stat +=
+			jpeg_v5_0_1_query_poison_by_instance(adev, inst, sub);
+
+	return !!poison_stat;
+}
+
+static const struct amdgpu_ras_block_hw_ops jpeg_v5_0_1_ras_hw_ops = {
+	.query_poison_status = jpeg_v5_0_1_query_ras_poison_status,
+};
+
+static int jpeg_v5_0_1_aca_bank_parser(struct aca_handle *handle, struct aca_bank *bank,
+				      enum aca_smu_type type, void *data)
+{
+	struct aca_bank_info info;
+	u64 misc0;
+	int ret;
+
+	ret = aca_bank_info_decode(bank, &info);
+	if (ret)
+		return ret;
+
+	misc0 = bank->regs[ACA_REG_IDX_MISC0];
+	switch (type) {
+	case ACA_SMU_TYPE_UE:
+		bank->aca_err_type = ACA_ERROR_TYPE_UE;
+		ret = aca_error_cache_log_bank_error(handle, &info, ACA_ERROR_TYPE_UE,
+						     1ULL);
+		break;
+	case ACA_SMU_TYPE_CE:
+		bank->aca_err_type = ACA_ERROR_TYPE_CE;
+		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type,
+						     ACA_REG__MISC0__ERRCNT(misc0));
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return ret;
+}
+
+/* reference to smu driver if header file */
+static int jpeg_v5_0_1_err_codes[] = {
+	16, 17, 18, 19, 20, 21, 22, 23, /* JPEG[0-9][S|D] */
+	24, 25, 26, 27, 28, 29, 30, 31,
+	48, 49, 50, 51,
+};
+
+static bool jpeg_v5_0_1_aca_bank_is_valid(struct aca_handle *handle, struct aca_bank *bank,
+					 enum aca_smu_type type, void *data)
+{
+	u32 instlo;
+
+	instlo = ACA_REG__IPID__INSTANCEIDLO(bank->regs[ACA_REG_IDX_IPID]);
+	instlo &= GENMASK(31, 1);
+
+	if (instlo != mmSMNAID_AID0_MCA_SMU)
+		return false;
+
+	if (aca_bank_check_error_codes(handle->adev, bank,
+				       jpeg_v5_0_1_err_codes,
+				       ARRAY_SIZE(jpeg_v5_0_1_err_codes)))
+		return false;
+
+	return true;
+}
+
+static const struct aca_bank_ops jpeg_v5_0_1_aca_bank_ops = {
+	.aca_bank_parser = jpeg_v5_0_1_aca_bank_parser,
+	.aca_bank_is_valid = jpeg_v5_0_1_aca_bank_is_valid,
+};
+
+static const struct aca_info jpeg_v5_0_1_aca_info = {
+	.hwip = ACA_HWIP_TYPE_SMU,
+	.mask = ACA_ERROR_UE_MASK,
+	.bank_ops = &jpeg_v5_0_1_aca_bank_ops,
+};
+
+static int jpeg_v5_0_1_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
+{
+	int r;
+
+	r = amdgpu_ras_block_late_init(adev, ras_block);
+	if (r)
+		return r;
+
+	r = amdgpu_ras_bind_aca(adev, AMDGPU_RAS_BLOCK__JPEG,
+				&jpeg_v5_0_1_aca_info, NULL);
+	if (r)
+		goto late_fini;
+
+	if (amdgpu_ras_is_supported(adev, ras_block->block) &&
+		adev->jpeg.inst->ras_poison_irq.funcs) {
+		r = amdgpu_irq_get(adev, &adev->jpeg.inst->ras_poison_irq, 0);
+		if (r)
+			goto late_fini;
+	}
+
+	return 0;
+
+late_fini:
+	amdgpu_ras_block_late_fini(adev, ras_block);
+
+	return r;
+}
+
+static struct amdgpu_jpeg_ras jpeg_v5_0_1_ras = {
+	.ras_block = {
+		.hw_ops = &jpeg_v5_0_1_ras_hw_ops,
+		.ras_late_init = jpeg_v5_0_1_ras_late_init,
+	},
+};
+
+static void jpeg_v5_0_1_set_ras_funcs(struct amdgpu_device *adev)
+{
+	adev->jpeg.ras = &jpeg_v5_0_1_ras;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_1.h b/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_1.h
new file mode 100644
index 000000000000..a7e58d5fb246
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_1.h
@@ -0,0 +1,111 @@
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __JPEG_V5_0_1_H__
+#define __JPEG_V5_0_1_H__
+
+extern const struct amdgpu_ip_block_version jpeg_v5_0_1_ip_block;
+
+#define regUVD_JRBC0_UVD_JRBC_SCRATCH0_INTERNAL_OFFSET			0x4094
+#define regUVD_JRBC_EXTERNAL_MCM_ADDR_INTERNAL_OFFSET			0x1bffe
+
+#define regUVD_JRBC0_UVD_JRBC_RB_WPTR                                                         0x0640
+#define regUVD_JRBC0_UVD_JRBC_RB_WPTR_BASE_IDX                                                1
+#define regUVD_JRBC0_UVD_JRBC_STATUS                                                          0x0649
+#define regUVD_JRBC0_UVD_JRBC_STATUS_BASE_IDX                                                 1
+#define regUVD_JRBC0_UVD_JRBC_RB_RPTR                                                         0x064a
+#define regUVD_JRBC0_UVD_JRBC_RB_RPTR_BASE_IDX                                                1
+#define regUVD_JRBC1_UVD_JRBC_RB_WPTR                                                         0x0000
+#define regUVD_JRBC1_UVD_JRBC_RB_WPTR_BASE_IDX                                                0
+#define regUVD_JRBC1_UVD_JRBC_STATUS                                                          0x0009
+#define regUVD_JRBC1_UVD_JRBC_STATUS_BASE_IDX                                                 0
+#define regUVD_JRBC1_UVD_JRBC_RB_RPTR                                                         0x000a
+#define regUVD_JRBC1_UVD_JRBC_RB_RPTR_BASE_IDX                                                0
+#define regUVD_JRBC2_UVD_JRBC_RB_WPTR                                                         0x0040
+#define regUVD_JRBC2_UVD_JRBC_RB_WPTR_BASE_IDX                                                0
+#define regUVD_JRBC2_UVD_JRBC_STATUS                                                          0x0049
+#define regUVD_JRBC2_UVD_JRBC_STATUS_BASE_IDX                                                 0
+#define regUVD_JRBC2_UVD_JRBC_RB_RPTR                                                         0x004a
+#define regUVD_JRBC2_UVD_JRBC_RB_RPTR_BASE_IDX                                                0
+#define regUVD_JRBC3_UVD_JRBC_RB_WPTR                                                         0x0080
+#define regUVD_JRBC3_UVD_JRBC_RB_WPTR_BASE_IDX                                                0
+#define regUVD_JRBC3_UVD_JRBC_STATUS                                                          0x0089
+#define regUVD_JRBC3_UVD_JRBC_STATUS_BASE_IDX                                                 0
+#define regUVD_JRBC3_UVD_JRBC_RB_RPTR                                                         0x008a
+#define regUVD_JRBC3_UVD_JRBC_RB_RPTR_BASE_IDX                                                0
+#define regUVD_JRBC4_UVD_JRBC_RB_WPTR                                                         0x00c0
+#define regUVD_JRBC4_UVD_JRBC_RB_WPTR_BASE_IDX                                                0
+#define regUVD_JRBC4_UVD_JRBC_STATUS                                                          0x00c9
+#define regUVD_JRBC4_UVD_JRBC_STATUS_BASE_IDX                                                 0
+#define regUVD_JRBC4_UVD_JRBC_RB_RPTR                                                         0x00ca
+#define regUVD_JRBC4_UVD_JRBC_RB_RPTR_BASE_IDX                                                0
+#define regUVD_JRBC5_UVD_JRBC_RB_WPTR                                                         0x0100
+#define regUVD_JRBC5_UVD_JRBC_RB_WPTR_BASE_IDX                                                0
+#define regUVD_JRBC5_UVD_JRBC_STATUS                                                          0x0109
+#define regUVD_JRBC5_UVD_JRBC_STATUS_BASE_IDX                                                 0
+#define regUVD_JRBC5_UVD_JRBC_RB_RPTR                                                         0x010a
+#define regUVD_JRBC5_UVD_JRBC_RB_RPTR_BASE_IDX                                                0
+#define regUVD_JRBC6_UVD_JRBC_RB_WPTR                                                         0x0140
+#define regUVD_JRBC6_UVD_JRBC_RB_WPTR_BASE_IDX                                                0
+#define regUVD_JRBC6_UVD_JRBC_STATUS                                                          0x0149
+#define regUVD_JRBC6_UVD_JRBC_STATUS_BASE_IDX                                                 0
+#define regUVD_JRBC6_UVD_JRBC_RB_RPTR                                                         0x014a
+#define regUVD_JRBC6_UVD_JRBC_RB_RPTR_BASE_IDX                                                0
+#define regUVD_JRBC7_UVD_JRBC_RB_WPTR                                                         0x0180
+#define regUVD_JRBC7_UVD_JRBC_RB_WPTR_BASE_IDX                                                0
+#define regUVD_JRBC7_UVD_JRBC_STATUS                                                          0x0189
+#define regUVD_JRBC7_UVD_JRBC_STATUS_BASE_IDX                                                 0
+#define regUVD_JRBC7_UVD_JRBC_RB_RPTR                                                         0x018a
+#define regUVD_JRBC7_UVD_JRBC_RB_RPTR_BASE_IDX                                                0
+#define regUVD_JRBC8_UVD_JRBC_RB_WPTR                                                         0x01c0
+#define regUVD_JRBC8_UVD_JRBC_RB_WPTR_BASE_IDX                                                0
+#define regUVD_JRBC8_UVD_JRBC_STATUS                                                          0x01c9
+#define regUVD_JRBC8_UVD_JRBC_STATUS_BASE_IDX                                                 0
+#define regUVD_JRBC8_UVD_JRBC_RB_RPTR                                                         0x01ca
+#define regUVD_JRBC8_UVD_JRBC_RB_RPTR_BASE_IDX                                                0
+#define regUVD_JRBC9_UVD_JRBC_RB_WPTR                                                         0x0440
+#define regUVD_JRBC9_UVD_JRBC_RB_WPTR_BASE_IDX                                                1
+#define regUVD_JRBC9_UVD_JRBC_STATUS                                                          0x0449
+#define regUVD_JRBC9_UVD_JRBC_STATUS_BASE_IDX                                                 1
+#define regUVD_JRBC9_UVD_JRBC_RB_RPTR                                                         0x044a
+#define regUVD_JRBC9_UVD_JRBC_RB_RPTR_BASE_IDX                                                1
+#define regUVD_JMI0_JPEG_LMI_DROP                                                             0x0663
+#define regUVD_JMI0_JPEG_LMI_DROP_BASE_IDX                                                    1
+#define regUVD_JMI0_UVD_JMI_CLIENT_STALL                                                      0x067a
+#define regUVD_JMI0_UVD_JMI_CLIENT_STALL_BASE_IDX                                             1
+#define regUVD_JMI0_UVD_JMI_CLIENT_CLEAN_STATUS                                               0x067b
+#define regUVD_JMI0_UVD_JMI_CLIENT_CLEAN_STATUS_BASE_IDX                                      1
+#define regJPEG_CORE_RST_CTRL                                                                 0x072e
+#define regJPEG_CORE_RST_CTRL_BASE_IDX                                                        1
+
+#define regVCN_RRMT_CNTL                          0x0940
+#define regVCN_RRMT_CNTL_BASE_IDX                 1
+
+enum amdgpu_jpeg_v5_0_1_sub_block {
+	AMDGPU_JPEG_V5_0_1_JPEG0 = 0,
+	AMDGPU_JPEG_V5_0_1_JPEG1,
+
+	AMDGPU_JPEG_V5_0_1_MAX_SUB_BLOCK,
+};
+
+#endif /* __JPEG_V5_0_1_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/lsdma_v7_0.c b/drivers/gpu/drm/amd/amdgpu/lsdma_v7_0.c
new file mode 100644
index 000000000000..396262044ea8
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/lsdma_v7_0.c
@@ -0,0 +1,121 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/delay.h>
+#include "amdgpu.h"
+#include "lsdma_v7_0.h"
+#include "amdgpu_lsdma.h"
+
+#include "lsdma/lsdma_7_0_0_offset.h"
+#include "lsdma/lsdma_7_0_0_sh_mask.h"
+
+static int lsdma_v7_0_wait_pio_status(struct amdgpu_device *adev)
+{
+	return amdgpu_lsdma_wait_for(adev, SOC15_REG_OFFSET(LSDMA, 0, regLSDMA_PIO_STATUS),
+			LSDMA_PIO_STATUS__PIO_IDLE_MASK | LSDMA_PIO_STATUS__PIO_FIFO_EMPTY_MASK,
+			LSDMA_PIO_STATUS__PIO_IDLE_MASK | LSDMA_PIO_STATUS__PIO_FIFO_EMPTY_MASK);
+}
+
+static int lsdma_v7_0_copy_mem(struct amdgpu_device *adev,
+			       uint64_t src_addr,
+			       uint64_t dst_addr,
+			       uint64_t size)
+{
+	int ret;
+	uint32_t tmp;
+
+	WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_SRC_ADDR_LO, lower_32_bits(src_addr));
+	WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_SRC_ADDR_HI, upper_32_bits(src_addr));
+
+	WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_DST_ADDR_LO, lower_32_bits(dst_addr));
+	WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_DST_ADDR_HI, upper_32_bits(dst_addr));
+
+	WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_CONTROL, 0x0);
+
+	tmp = RREG32_SOC15(LSDMA, 0, regLSDMA_PIO_COMMAND);
+	tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, BYTE_COUNT, size);
+	tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, SRC_LOCATION, 0);
+	tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, DST_LOCATION, 0);
+	tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, SRC_ADDR_INC, 0);
+	tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, DST_ADDR_INC, 0);
+	tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, OVERLAP_DISABLE, 0);
+	tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, CONSTANT_FILL, 0);
+	WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_COMMAND, tmp);
+
+	ret = lsdma_v7_0_wait_pio_status(adev);
+	if (ret)
+		dev_err(adev->dev, "LSDMA PIO failed to copy memory!\n");
+
+	return ret;
+}
+
+static int lsdma_v7_0_fill_mem(struct amdgpu_device *adev,
+			       uint64_t dst_addr,
+			       uint32_t data,
+			       uint64_t size)
+{
+	int ret;
+	uint32_t tmp;
+
+	WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_CONSTFILL_DATA, data);
+
+	WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_DST_ADDR_LO, lower_32_bits(dst_addr));
+	WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_DST_ADDR_HI, upper_32_bits(dst_addr));
+
+	WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_CONTROL, 0x0);
+
+	tmp = RREG32_SOC15(LSDMA, 0, regLSDMA_PIO_COMMAND);
+	tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, BYTE_COUNT, size);
+	tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, SRC_LOCATION, 0);
+	tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, DST_LOCATION, 0);
+	tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, SRC_ADDR_INC, 0);
+	tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, DST_ADDR_INC, 0);
+	tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, OVERLAP_DISABLE, 0);
+	tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, CONSTANT_FILL, 1);
+	WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_COMMAND, tmp);
+
+	ret = lsdma_v7_0_wait_pio_status(adev);
+	if (ret)
+		dev_err(adev->dev, "LSDMA PIO failed to fill memory!\n");
+
+	return ret;
+}
+
+static void lsdma_v7_0_update_memory_power_gating(struct amdgpu_device *adev,
+						 bool enable)
+{
+	uint32_t tmp;
+
+	tmp = RREG32_SOC15(LSDMA, 0, regLSDMA_MEM_POWER_CTRL);
+	tmp = REG_SET_FIELD(tmp, LSDMA_MEM_POWER_CTRL, MEM_POWER_CTRL_EN, 0);
+	WREG32_SOC15(LSDMA, 0, regLSDMA_MEM_POWER_CTRL, tmp);
+
+	tmp = REG_SET_FIELD(tmp, LSDMA_MEM_POWER_CTRL, MEM_POWER_CTRL_EN, enable);
+	WREG32_SOC15(LSDMA, 0, regLSDMA_MEM_POWER_CTRL, tmp);
+}
+
+const struct amdgpu_lsdma_funcs lsdma_v7_0_funcs = {
+	.copy_mem = lsdma_v7_0_copy_mem,
+	.fill_mem = lsdma_v7_0_fill_mem,
+	.update_memory_power_gating = lsdma_v7_0_update_memory_power_gating
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/lsdma_v7_0.h b/drivers/gpu/drm/amd/amdgpu/lsdma_v7_0.h
new file mode 100644
index 000000000000..52b4485cdd98
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/lsdma_v7_0.h
@@ -0,0 +1,31 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __LSDMA_V7_0_H__
+#define __LSDMA_V7_0_H__
+
+#include "soc15_common.h"
+
+extern const struct amdgpu_lsdma_funcs lsdma_v7_0_funcs;
+
+#endif /* __LSDMA_V7_0_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/mes_userqueue.c b/drivers/gpu/drm/amd/amdgpu/mes_userqueue.c
new file mode 100644
index 000000000000..64cae89357b6
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/mes_userqueue.c
@@ -0,0 +1,501 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include <drm/drm_drv.h>
+#include "amdgpu.h"
+#include "amdgpu_gfx.h"
+#include "mes_userqueue.h"
+#include "amdgpu_userq_fence.h"
+
+#define AMDGPU_USERQ_PROC_CTX_SZ PAGE_SIZE
+#define AMDGPU_USERQ_GANG_CTX_SZ PAGE_SIZE
+
+static int
+mes_userq_map_gtt_bo_to_gart(struct amdgpu_bo *bo)
+{
+	int ret;
+
+	ret = amdgpu_bo_reserve(bo, true);
+	if (ret) {
+		DRM_ERROR("Failed to reserve bo. ret %d\n", ret);
+		goto err_reserve_bo_failed;
+	}
+
+	ret = amdgpu_ttm_alloc_gart(&bo->tbo);
+	if (ret) {
+		DRM_ERROR("Failed to bind bo to GART. ret %d\n", ret);
+		goto err_map_bo_gart_failed;
+	}
+
+	amdgpu_bo_unreserve(bo);
+	bo = amdgpu_bo_ref(bo);
+
+	return 0;
+
+err_map_bo_gart_failed:
+	amdgpu_bo_unreserve(bo);
+err_reserve_bo_failed:
+	return ret;
+}
+
+static int
+mes_userq_create_wptr_mapping(struct amdgpu_userq_mgr *uq_mgr,
+			      struct amdgpu_usermode_queue *queue,
+			      uint64_t wptr)
+{
+	struct amdgpu_bo_va_mapping *wptr_mapping;
+	struct amdgpu_vm *wptr_vm;
+	struct amdgpu_userq_obj *wptr_obj = &queue->wptr_obj;
+	int ret;
+
+	wptr_vm = queue->vm;
+	ret = amdgpu_bo_reserve(wptr_vm->root.bo, false);
+	if (ret)
+		return ret;
+
+	wptr &= AMDGPU_GMC_HOLE_MASK;
+	wptr_mapping = amdgpu_vm_bo_lookup_mapping(wptr_vm, wptr >> PAGE_SHIFT);
+	amdgpu_bo_unreserve(wptr_vm->root.bo);
+	if (!wptr_mapping) {
+		DRM_ERROR("Failed to lookup wptr bo\n");
+		return -EINVAL;
+	}
+
+	wptr_obj->obj = wptr_mapping->bo_va->base.bo;
+	if (wptr_obj->obj->tbo.base.size > PAGE_SIZE) {
+		DRM_ERROR("Requested GART mapping for wptr bo larger than one page\n");
+		return -EINVAL;
+	}
+
+	ret = mes_userq_map_gtt_bo_to_gart(wptr_obj->obj);
+	if (ret) {
+		DRM_ERROR("Failed to map wptr bo to GART\n");
+		return ret;
+	}
+
+	queue->wptr_obj.gpu_addr = amdgpu_bo_gpu_offset_no_check(wptr_obj->obj);
+	return 0;
+}
+
+static int convert_to_mes_priority(int priority)
+{
+	switch (priority) {
+	case AMDGPU_USERQ_CREATE_FLAGS_QUEUE_PRIORITY_NORMAL_LOW:
+	default:
+		return AMDGPU_MES_PRIORITY_LEVEL_NORMAL;
+	case AMDGPU_USERQ_CREATE_FLAGS_QUEUE_PRIORITY_LOW:
+		return AMDGPU_MES_PRIORITY_LEVEL_LOW;
+	case AMDGPU_USERQ_CREATE_FLAGS_QUEUE_PRIORITY_NORMAL_HIGH:
+		return AMDGPU_MES_PRIORITY_LEVEL_MEDIUM;
+	case AMDGPU_USERQ_CREATE_FLAGS_QUEUE_PRIORITY_HIGH:
+		return AMDGPU_MES_PRIORITY_LEVEL_HIGH;
+	}
+}
+
+static int mes_userq_map(struct amdgpu_userq_mgr *uq_mgr,
+			 struct amdgpu_usermode_queue *queue)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	struct amdgpu_userq_obj *ctx = &queue->fw_obj;
+	struct amdgpu_mqd_prop *userq_props = queue->userq_prop;
+	struct mes_add_queue_input queue_input;
+	int r;
+
+	memset(&queue_input, 0x0, sizeof(struct mes_add_queue_input));
+
+	queue_input.process_va_start = 0;
+	queue_input.process_va_end = adev->vm_manager.max_pfn - 1;
+
+	/* set process quantum to 10 ms and gang quantum to 1 ms as default */
+	queue_input.process_quantum = 100000;
+	queue_input.gang_quantum = 10000;
+	queue_input.paging = false;
+
+	queue_input.process_context_addr = ctx->gpu_addr;
+	queue_input.gang_context_addr = ctx->gpu_addr + AMDGPU_USERQ_PROC_CTX_SZ;
+	queue_input.inprocess_gang_priority = AMDGPU_MES_PRIORITY_LEVEL_NORMAL;
+	queue_input.gang_global_priority_level = convert_to_mes_priority(queue->priority);
+
+	queue_input.process_id = queue->vm->pasid;
+	queue_input.queue_type = queue->queue_type;
+	queue_input.mqd_addr = queue->mqd.gpu_addr;
+	queue_input.wptr_addr = userq_props->wptr_gpu_addr;
+	queue_input.queue_size = userq_props->queue_size >> 2;
+	queue_input.doorbell_offset = userq_props->doorbell_index;
+	queue_input.page_table_base_addr = amdgpu_gmc_pd_addr(queue->vm->root.bo);
+	queue_input.wptr_mc_addr = queue->wptr_obj.gpu_addr;
+
+	amdgpu_mes_lock(&adev->mes);
+	r = adev->mes.funcs->add_hw_queue(&adev->mes, &queue_input);
+	amdgpu_mes_unlock(&adev->mes);
+	if (r) {
+		DRM_ERROR("Failed to map queue in HW, err (%d)\n", r);
+		return r;
+	}
+
+	DRM_DEBUG_DRIVER("Queue (doorbell:%d) mapped successfully\n", userq_props->doorbell_index);
+	return 0;
+}
+
+static int mes_userq_unmap(struct amdgpu_userq_mgr *uq_mgr,
+			   struct amdgpu_usermode_queue *queue)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	struct mes_remove_queue_input queue_input;
+	struct amdgpu_userq_obj *ctx = &queue->fw_obj;
+	int r;
+
+	memset(&queue_input, 0x0, sizeof(struct mes_remove_queue_input));
+	queue_input.doorbell_offset = queue->doorbell_index;
+	queue_input.gang_context_addr = ctx->gpu_addr + AMDGPU_USERQ_PROC_CTX_SZ;
+
+	amdgpu_mes_lock(&adev->mes);
+	r = adev->mes.funcs->remove_hw_queue(&adev->mes, &queue_input);
+	amdgpu_mes_unlock(&adev->mes);
+	if (r)
+		DRM_ERROR("Failed to unmap queue in HW, err (%d)\n", r);
+	return r;
+}
+
+static int mes_userq_create_ctx_space(struct amdgpu_userq_mgr *uq_mgr,
+				      struct amdgpu_usermode_queue *queue,
+				      struct drm_amdgpu_userq_in *mqd_user)
+{
+	struct amdgpu_userq_obj *ctx = &queue->fw_obj;
+	int r, size;
+
+	/*
+	 * The FW expects at least one page space allocated for
+	 * process ctx and gang ctx each. Create an object
+	 * for the same.
+	 */
+	size = AMDGPU_USERQ_PROC_CTX_SZ + AMDGPU_USERQ_GANG_CTX_SZ;
+	r = amdgpu_userq_create_object(uq_mgr, ctx, size);
+	if (r) {
+		DRM_ERROR("Failed to allocate ctx space bo for userqueue, err:%d\n", r);
+		return r;
+	}
+
+	return 0;
+}
+
+static int mes_userq_detect_and_reset(struct amdgpu_device *adev,
+				      int queue_type)
+{
+	int db_array_size = amdgpu_mes_get_hung_queue_db_array_size(adev);
+	struct mes_detect_and_reset_queue_input input;
+	struct amdgpu_usermode_queue *queue;
+	unsigned int hung_db_num = 0;
+	unsigned long queue_id;
+	u32 db_array[8];
+	bool found_hung_queue = false;
+	int r, i;
+
+	if (db_array_size > 8) {
+		dev_err(adev->dev, "DB array size (%d vs 8) too small\n",
+			db_array_size);
+		return -EINVAL;
+	}
+
+	memset(&input, 0x0, sizeof(struct mes_detect_and_reset_queue_input));
+
+	input.queue_type = queue_type;
+
+	amdgpu_mes_lock(&adev->mes);
+	r = amdgpu_mes_detect_and_reset_hung_queues(adev, queue_type, false,
+						    &hung_db_num, db_array);
+	amdgpu_mes_unlock(&adev->mes);
+	if (r) {
+		dev_err(adev->dev, "Failed to detect and reset queues, err (%d)\n", r);
+	} else if (hung_db_num) {
+		xa_for_each(&adev->userq_doorbell_xa, queue_id, queue) {
+			if (queue->queue_type == queue_type) {
+				for (i = 0; i < hung_db_num; i++) {
+					if (queue->doorbell_index == db_array[i]) {
+						queue->state = AMDGPU_USERQ_STATE_HUNG;
+						found_hung_queue = true;
+						atomic_inc(&adev->gpu_reset_counter);
+						amdgpu_userq_fence_driver_force_completion(queue);
+						drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE, NULL);
+					}
+				}
+			}
+		}
+	}
+
+	if (found_hung_queue) {
+		/* Resume scheduling after hang recovery */
+		r = amdgpu_mes_resume(adev);
+	}
+
+	return r;
+}
+
+static int mes_userq_mqd_create(struct amdgpu_userq_mgr *uq_mgr,
+				struct drm_amdgpu_userq_in *args_in,
+				struct amdgpu_usermode_queue *queue)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	struct amdgpu_mqd *mqd_hw_default = &adev->mqds[queue->queue_type];
+	struct drm_amdgpu_userq_in *mqd_user = args_in;
+	struct amdgpu_mqd_prop *userq_props;
+	int r;
+
+	/* Structure to initialize MQD for userqueue using generic MQD init function */
+	userq_props = kzalloc(sizeof(struct amdgpu_mqd_prop), GFP_KERNEL);
+	if (!userq_props) {
+		DRM_ERROR("Failed to allocate memory for userq_props\n");
+		return -ENOMEM;
+	}
+
+	r = amdgpu_userq_create_object(uq_mgr, &queue->mqd, mqd_hw_default->mqd_size);
+	if (r) {
+		DRM_ERROR("Failed to create MQD object for userqueue\n");
+		goto free_props;
+	}
+
+	/* Initialize the MQD BO with user given values */
+	userq_props->wptr_gpu_addr = mqd_user->wptr_va;
+	userq_props->rptr_gpu_addr = mqd_user->rptr_va;
+	userq_props->queue_size = mqd_user->queue_size;
+	userq_props->hqd_base_gpu_addr = mqd_user->queue_va;
+	userq_props->mqd_gpu_addr = queue->mqd.gpu_addr;
+	userq_props->use_doorbell = true;
+	userq_props->doorbell_index = queue->doorbell_index;
+	userq_props->fence_address = queue->fence_drv->gpu_addr;
+
+	if (queue->queue_type == AMDGPU_HW_IP_COMPUTE) {
+		struct drm_amdgpu_userq_mqd_compute_gfx11 *compute_mqd;
+
+		if (mqd_user->mqd_size != sizeof(*compute_mqd)) {
+			DRM_ERROR("Invalid compute IP MQD size\n");
+			r = -EINVAL;
+			goto free_mqd;
+		}
+
+		compute_mqd = memdup_user(u64_to_user_ptr(mqd_user->mqd), mqd_user->mqd_size);
+		if (IS_ERR(compute_mqd)) {
+			DRM_ERROR("Failed to read user MQD\n");
+			r = -ENOMEM;
+			goto free_mqd;
+		}
+
+		r = amdgpu_userq_input_va_validate(queue, compute_mqd->eop_va,
+						   2048);
+		if (r)
+			goto free_mqd;
+
+		userq_props->eop_gpu_addr = compute_mqd->eop_va;
+		userq_props->hqd_pipe_priority = AMDGPU_GFX_PIPE_PRIO_NORMAL;
+		userq_props->hqd_queue_priority = AMDGPU_GFX_QUEUE_PRIORITY_MINIMUM;
+		userq_props->hqd_active = false;
+		userq_props->tmz_queue =
+			mqd_user->flags & AMDGPU_USERQ_CREATE_FLAGS_QUEUE_SECURE;
+		kfree(compute_mqd);
+	} else if (queue->queue_type == AMDGPU_HW_IP_GFX) {
+		struct drm_amdgpu_userq_mqd_gfx11 *mqd_gfx_v11;
+		struct amdgpu_gfx_shadow_info shadow_info;
+
+		if (adev->gfx.funcs->get_gfx_shadow_info) {
+			adev->gfx.funcs->get_gfx_shadow_info(adev, &shadow_info, true);
+		} else {
+			r = -EINVAL;
+			goto free_mqd;
+		}
+
+		if (mqd_user->mqd_size != sizeof(*mqd_gfx_v11) || !mqd_user->mqd) {
+			DRM_ERROR("Invalid GFX MQD\n");
+			r = -EINVAL;
+			goto free_mqd;
+		}
+
+		mqd_gfx_v11 = memdup_user(u64_to_user_ptr(mqd_user->mqd), mqd_user->mqd_size);
+		if (IS_ERR(mqd_gfx_v11)) {
+			DRM_ERROR("Failed to read user MQD\n");
+			r = -ENOMEM;
+			goto free_mqd;
+		}
+
+		userq_props->shadow_addr = mqd_gfx_v11->shadow_va;
+		userq_props->csa_addr = mqd_gfx_v11->csa_va;
+		userq_props->tmz_queue =
+			mqd_user->flags & AMDGPU_USERQ_CREATE_FLAGS_QUEUE_SECURE;
+
+		r = amdgpu_userq_input_va_validate(queue, mqd_gfx_v11->shadow_va,
+						   shadow_info.shadow_size);
+		if (r)
+			goto free_mqd;
+		r = amdgpu_userq_input_va_validate(queue, mqd_gfx_v11->csa_va,
+						   shadow_info.csa_size);
+		if (r)
+			goto free_mqd;
+
+		kfree(mqd_gfx_v11);
+	} else if (queue->queue_type == AMDGPU_HW_IP_DMA) {
+		struct drm_amdgpu_userq_mqd_sdma_gfx11 *mqd_sdma_v11;
+
+		if (mqd_user->mqd_size != sizeof(*mqd_sdma_v11) || !mqd_user->mqd) {
+			DRM_ERROR("Invalid SDMA MQD\n");
+			r = -EINVAL;
+			goto free_mqd;
+		}
+
+		mqd_sdma_v11 = memdup_user(u64_to_user_ptr(mqd_user->mqd), mqd_user->mqd_size);
+		if (IS_ERR(mqd_sdma_v11)) {
+			DRM_ERROR("Failed to read sdma user MQD\n");
+			r = -ENOMEM;
+			goto free_mqd;
+		}
+		r = amdgpu_userq_input_va_validate(queue, mqd_sdma_v11->csa_va,
+						   32);
+		if (r)
+			goto free_mqd;
+
+		userq_props->csa_addr = mqd_sdma_v11->csa_va;
+		kfree(mqd_sdma_v11);
+	}
+
+	queue->userq_prop = userq_props;
+
+	r = mqd_hw_default->init_mqd(adev, (void *)queue->mqd.cpu_ptr, userq_props);
+	if (r) {
+		DRM_ERROR("Failed to initialize MQD for userqueue\n");
+		goto free_mqd;
+	}
+
+	/* Create BO for FW operations */
+	r = mes_userq_create_ctx_space(uq_mgr, queue, mqd_user);
+	if (r) {
+		DRM_ERROR("Failed to allocate BO for userqueue (%d)", r);
+		goto free_mqd;
+	}
+
+	/* FW expects WPTR BOs to be mapped into GART */
+	r = mes_userq_create_wptr_mapping(uq_mgr, queue, userq_props->wptr_gpu_addr);
+	if (r) {
+		DRM_ERROR("Failed to create WPTR mapping\n");
+		goto free_ctx;
+	}
+
+	return 0;
+
+free_ctx:
+	amdgpu_userq_destroy_object(uq_mgr, &queue->fw_obj);
+
+free_mqd:
+	amdgpu_userq_destroy_object(uq_mgr, &queue->mqd);
+
+free_props:
+	kfree(userq_props);
+
+	return r;
+}
+
+static void
+mes_userq_mqd_destroy(struct amdgpu_userq_mgr *uq_mgr,
+		      struct amdgpu_usermode_queue *queue)
+{
+	amdgpu_userq_destroy_object(uq_mgr, &queue->fw_obj);
+	kfree(queue->userq_prop);
+	amdgpu_userq_destroy_object(uq_mgr, &queue->mqd);
+}
+
+static int mes_userq_preempt(struct amdgpu_userq_mgr *uq_mgr,
+				struct amdgpu_usermode_queue *queue)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	struct mes_suspend_gang_input queue_input;
+	struct amdgpu_userq_obj *ctx = &queue->fw_obj;
+	signed long timeout = 2100000; /* 2100 ms */
+	u64 fence_gpu_addr;
+	u32 fence_offset;
+	u64 *fence_ptr;
+	int i, r;
+
+	if (queue->state != AMDGPU_USERQ_STATE_MAPPED)
+		return 0;
+	r = amdgpu_device_wb_get(adev, &fence_offset);
+	if (r)
+		return r;
+
+	fence_gpu_addr = adev->wb.gpu_addr + (fence_offset * 4);
+	fence_ptr = (u64 *)&adev->wb.wb[fence_offset];
+	*fence_ptr = 0;
+
+	memset(&queue_input, 0x0, sizeof(struct mes_suspend_gang_input));
+	queue_input.gang_context_addr = ctx->gpu_addr + AMDGPU_USERQ_PROC_CTX_SZ;
+	queue_input.suspend_fence_addr = fence_gpu_addr;
+	queue_input.suspend_fence_value = 1;
+	amdgpu_mes_lock(&adev->mes);
+	r = adev->mes.funcs->suspend_gang(&adev->mes, &queue_input);
+	amdgpu_mes_unlock(&adev->mes);
+	if (r) {
+		DRM_ERROR("Failed to suspend gang: %d\n", r);
+		goto out;
+	}
+
+	for (i = 0; i < timeout; i++) {
+		if (*fence_ptr == 1)
+			goto out;
+		udelay(1);
+	}
+	r = -ETIMEDOUT;
+
+out:
+	amdgpu_device_wb_free(adev, fence_offset);
+	return r;
+}
+
+static int mes_userq_restore(struct amdgpu_userq_mgr *uq_mgr,
+				struct amdgpu_usermode_queue *queue)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	struct mes_resume_gang_input queue_input;
+	struct amdgpu_userq_obj *ctx = &queue->fw_obj;
+	int r;
+
+	if (queue->state == AMDGPU_USERQ_STATE_HUNG)
+		return -EINVAL;
+	if (queue->state != AMDGPU_USERQ_STATE_PREEMPTED)
+		return 0;
+
+	memset(&queue_input, 0x0, sizeof(struct mes_resume_gang_input));
+	queue_input.gang_context_addr = ctx->gpu_addr + AMDGPU_USERQ_PROC_CTX_SZ;
+
+	amdgpu_mes_lock(&adev->mes);
+	r = adev->mes.funcs->resume_gang(&adev->mes, &queue_input);
+	amdgpu_mes_unlock(&adev->mes);
+	if (r)
+		dev_err(adev->dev, "Failed to resume queue, err (%d)\n", r);
+	return r;
+}
+
+const struct amdgpu_userq_funcs userq_mes_funcs = {
+	.mqd_create = mes_userq_mqd_create,
+	.mqd_destroy = mes_userq_mqd_destroy,
+	.unmap = mes_userq_unmap,
+	.map = mes_userq_map,
+	.detect_and_reset = mes_userq_detect_and_reset,
+	.preempt = mes_userq_preempt,
+	.restore = mes_userq_restore,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/mes_userqueue.h b/drivers/gpu/drm/amd/amdgpu/mes_userqueue.h
new file mode 100644
index 000000000000..090ae8897770
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/mes_userqueue.h
@@ -0,0 +1,30 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef MES_USERQ_H
+#define MES_USERQ_H
+#include "amdgpu_userq.h"
+
+extern const struct amdgpu_userq_funcs userq_mes_funcs;
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c b/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c
deleted file mode 100644
index 2e2062636d5f..000000000000
--- a/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c
+++ /dev/null
@@ -1,1187 +0,0 @@
-/*
- * Copyright 2019 Advanced Micro Devices, Inc.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
- * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
- * OTHER DEALINGS IN THE SOFTWARE.
- *
- */
-
-#include <linux/firmware.h>
-#include <linux/module.h>
-#include "amdgpu.h"
-#include "soc15_common.h"
-#include "nv.h"
-#include "gc/gc_10_1_0_offset.h"
-#include "gc/gc_10_1_0_sh_mask.h"
-#include "gc/gc_10_1_0_default.h"
-#include "v10_structs.h"
-#include "mes_api_def.h"
-
-#define mmCP_MES_IC_OP_CNTL_Sienna_Cichlid               0x2820
-#define mmCP_MES_IC_OP_CNTL_Sienna_Cichlid_BASE_IDX      1
-#define mmRLC_CP_SCHEDULERS_Sienna_Cichlid		0x4ca1
-#define mmRLC_CP_SCHEDULERS_Sienna_Cichlid_BASE_IDX	1
-
-MODULE_FIRMWARE("amdgpu/navi10_mes.bin");
-MODULE_FIRMWARE("amdgpu/sienna_cichlid_mes.bin");
-MODULE_FIRMWARE("amdgpu/sienna_cichlid_mes1.bin");
-
-static int mes_v10_1_hw_fini(void *handle);
-static int mes_v10_1_kiq_hw_init(struct amdgpu_device *adev);
-
-#define MES_EOP_SIZE   2048
-
-static void mes_v10_1_ring_set_wptr(struct amdgpu_ring *ring)
-{
-	struct amdgpu_device *adev = ring->adev;
-
-	if (ring->use_doorbell) {
-		atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
-			     ring->wptr);
-		WDOORBELL64(ring->doorbell_index, ring->wptr);
-	} else {
-		BUG();
-	}
-}
-
-static u64 mes_v10_1_ring_get_rptr(struct amdgpu_ring *ring)
-{
-	return *ring->rptr_cpu_addr;
-}
-
-static u64 mes_v10_1_ring_get_wptr(struct amdgpu_ring *ring)
-{
-	u64 wptr;
-
-	if (ring->use_doorbell)
-		wptr = atomic64_read((atomic64_t *)ring->wptr_cpu_addr);
-	else
-		BUG();
-	return wptr;
-}
-
-static const struct amdgpu_ring_funcs mes_v10_1_ring_funcs = {
-	.type = AMDGPU_RING_TYPE_MES,
-	.align_mask = 1,
-	.nop = 0,
-	.support_64bit_ptrs = true,
-	.get_rptr = mes_v10_1_ring_get_rptr,
-	.get_wptr = mes_v10_1_ring_get_wptr,
-	.set_wptr = mes_v10_1_ring_set_wptr,
-	.insert_nop = amdgpu_ring_insert_nop,
-};
-
-static int mes_v10_1_submit_pkt_and_poll_completion(struct amdgpu_mes *mes,
-						    void *pkt, int size,
-						    int api_status_off)
-{
-	int ndw = size / 4;
-	signed long r;
-	union MESAPI__ADD_QUEUE *x_pkt = pkt;
-	struct MES_API_STATUS *api_status;
-	struct amdgpu_device *adev = mes->adev;
-	struct amdgpu_ring *ring = &mes->ring;
-	unsigned long flags;
-
-	BUG_ON(size % 4 != 0);
-
-	spin_lock_irqsave(&mes->ring_lock, flags);
-	if (amdgpu_ring_alloc(ring, ndw)) {
-		spin_unlock_irqrestore(&mes->ring_lock, flags);
-		return -ENOMEM;
-	}
-
-	api_status = (struct MES_API_STATUS *)((char *)pkt + api_status_off);
-	api_status->api_completion_fence_addr = mes->ring.fence_drv.gpu_addr;
-	api_status->api_completion_fence_value = ++mes->ring.fence_drv.sync_seq;
-
-	amdgpu_ring_write_multiple(ring, pkt, ndw);
-	amdgpu_ring_commit(ring);
-	spin_unlock_irqrestore(&mes->ring_lock, flags);
-
-	DRM_DEBUG("MES msg=%d was emitted\n", x_pkt->header.opcode);
-
-	r = amdgpu_fence_wait_polling(ring, ring->fence_drv.sync_seq,
-				      adev->usec_timeout);
-	if (r < 1) {
-		DRM_ERROR("MES failed to response msg=%d\n",
-			  x_pkt->header.opcode);
-
-		while (halt_if_hws_hang)
-			schedule();
-
-		return -ETIMEDOUT;
-	}
-
-	return 0;
-}
-
-static int convert_to_mes_queue_type(int queue_type)
-{
-	if (queue_type == AMDGPU_RING_TYPE_GFX)
-		return MES_QUEUE_TYPE_GFX;
-	else if (queue_type == AMDGPU_RING_TYPE_COMPUTE)
-		return MES_QUEUE_TYPE_COMPUTE;
-	else if (queue_type == AMDGPU_RING_TYPE_SDMA)
-		return MES_QUEUE_TYPE_SDMA;
-	else
-		BUG();
-	return -1;
-}
-
-static int mes_v10_1_add_hw_queue(struct amdgpu_mes *mes,
-				  struct mes_add_queue_input *input)
-{
-	struct amdgpu_device *adev = mes->adev;
-	union MESAPI__ADD_QUEUE mes_add_queue_pkt;
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	uint32_t vm_cntx_cntl = hub->vm_cntx_cntl;
-
-	memset(&mes_add_queue_pkt, 0, sizeof(mes_add_queue_pkt));
-
-	mes_add_queue_pkt.header.type = MES_API_TYPE_SCHEDULER;
-	mes_add_queue_pkt.header.opcode = MES_SCH_API_ADD_QUEUE;
-	mes_add_queue_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
-
-	mes_add_queue_pkt.process_id = input->process_id;
-	mes_add_queue_pkt.page_table_base_addr = input->page_table_base_addr;
-	mes_add_queue_pkt.process_va_start = input->process_va_start;
-	mes_add_queue_pkt.process_va_end = input->process_va_end;
-	mes_add_queue_pkt.process_quantum = input->process_quantum;
-	mes_add_queue_pkt.process_context_addr = input->process_context_addr;
-	mes_add_queue_pkt.gang_quantum = input->gang_quantum;
-	mes_add_queue_pkt.gang_context_addr = input->gang_context_addr;
-	mes_add_queue_pkt.inprocess_gang_priority =
-		input->inprocess_gang_priority;
-	mes_add_queue_pkt.gang_global_priority_level =
-		input->gang_global_priority_level;
-	mes_add_queue_pkt.doorbell_offset = input->doorbell_offset;
-	mes_add_queue_pkt.mqd_addr = input->mqd_addr;
-	mes_add_queue_pkt.wptr_addr = input->wptr_addr;
-	mes_add_queue_pkt.queue_type =
-		convert_to_mes_queue_type(input->queue_type);
-	mes_add_queue_pkt.paging = input->paging;
-	mes_add_queue_pkt.vm_context_cntl = vm_cntx_cntl;
-	mes_add_queue_pkt.gws_base = input->gws_base;
-	mes_add_queue_pkt.gws_size = input->gws_size;
-	mes_add_queue_pkt.trap_handler_addr = input->tba_addr;
-
-	return mes_v10_1_submit_pkt_and_poll_completion(mes,
-			&mes_add_queue_pkt, sizeof(mes_add_queue_pkt),
-			offsetof(union MESAPI__ADD_QUEUE, api_status));
-}
-
-static int mes_v10_1_remove_hw_queue(struct amdgpu_mes *mes,
-				     struct mes_remove_queue_input *input)
-{
-	union MESAPI__REMOVE_QUEUE mes_remove_queue_pkt;
-
-	memset(&mes_remove_queue_pkt, 0, sizeof(mes_remove_queue_pkt));
-
-	mes_remove_queue_pkt.header.type = MES_API_TYPE_SCHEDULER;
-	mes_remove_queue_pkt.header.opcode = MES_SCH_API_REMOVE_QUEUE;
-	mes_remove_queue_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
-
-	mes_remove_queue_pkt.doorbell_offset = input->doorbell_offset;
-	mes_remove_queue_pkt.gang_context_addr = input->gang_context_addr;
-
-	return mes_v10_1_submit_pkt_and_poll_completion(mes,
-			&mes_remove_queue_pkt, sizeof(mes_remove_queue_pkt),
-			offsetof(union MESAPI__REMOVE_QUEUE, api_status));
-}
-
-static int mes_v10_1_unmap_legacy_queue(struct amdgpu_mes *mes,
-				 struct mes_unmap_legacy_queue_input *input)
-{
-	union MESAPI__REMOVE_QUEUE mes_remove_queue_pkt;
-
-	memset(&mes_remove_queue_pkt, 0, sizeof(mes_remove_queue_pkt));
-
-	mes_remove_queue_pkt.header.type = MES_API_TYPE_SCHEDULER;
-	mes_remove_queue_pkt.header.opcode = MES_SCH_API_REMOVE_QUEUE;
-	mes_remove_queue_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
-
-	mes_remove_queue_pkt.doorbell_offset = input->doorbell_offset;
-	mes_remove_queue_pkt.gang_context_addr = 0;
-
-	mes_remove_queue_pkt.pipe_id = input->pipe_id;
-	mes_remove_queue_pkt.queue_id = input->queue_id;
-
-	if (input->action == PREEMPT_QUEUES_NO_UNMAP) {
-		mes_remove_queue_pkt.preempt_legacy_gfx_queue = 1;
-		mes_remove_queue_pkt.tf_addr = input->trail_fence_addr;
-		mes_remove_queue_pkt.tf_data =
-			lower_32_bits(input->trail_fence_data);
-	} else {
-		if (input->queue_type == AMDGPU_RING_TYPE_GFX)
-			mes_remove_queue_pkt.unmap_legacy_gfx_queue = 1;
-		else
-			mes_remove_queue_pkt.unmap_kiq_utility_queue = 1;
-	}
-
-	return mes_v10_1_submit_pkt_and_poll_completion(mes,
-			&mes_remove_queue_pkt, sizeof(mes_remove_queue_pkt),
-			offsetof(union MESAPI__REMOVE_QUEUE, api_status));
-}
-
-static int mes_v10_1_suspend_gang(struct amdgpu_mes *mes,
-				  struct mes_suspend_gang_input *input)
-{
-	return 0;
-}
-
-static int mes_v10_1_resume_gang(struct amdgpu_mes *mes,
-				 struct mes_resume_gang_input *input)
-{
-	return 0;
-}
-
-static int mes_v10_1_query_sched_status(struct amdgpu_mes *mes)
-{
-	union MESAPI__QUERY_MES_STATUS mes_status_pkt;
-
-	memset(&mes_status_pkt, 0, sizeof(mes_status_pkt));
-
-	mes_status_pkt.header.type = MES_API_TYPE_SCHEDULER;
-	mes_status_pkt.header.opcode = MES_SCH_API_QUERY_SCHEDULER_STATUS;
-	mes_status_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
-
-	return mes_v10_1_submit_pkt_and_poll_completion(mes,
-			&mes_status_pkt, sizeof(mes_status_pkt),
-			offsetof(union MESAPI__QUERY_MES_STATUS, api_status));
-}
-
-static int mes_v10_1_set_hw_resources(struct amdgpu_mes *mes)
-{
-	int i;
-	struct amdgpu_device *adev = mes->adev;
-	union MESAPI_SET_HW_RESOURCES mes_set_hw_res_pkt;
-
-	memset(&mes_set_hw_res_pkt, 0, sizeof(mes_set_hw_res_pkt));
-
-	mes_set_hw_res_pkt.header.type = MES_API_TYPE_SCHEDULER;
-	mes_set_hw_res_pkt.header.opcode = MES_SCH_API_SET_HW_RSRC;
-	mes_set_hw_res_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
-
-	mes_set_hw_res_pkt.vmid_mask_mmhub = mes->vmid_mask_mmhub;
-	mes_set_hw_res_pkt.vmid_mask_gfxhub = mes->vmid_mask_gfxhub;
-	mes_set_hw_res_pkt.gds_size = adev->gds.gds_size;
-	mes_set_hw_res_pkt.paging_vmid = 0;
-	mes_set_hw_res_pkt.g_sch_ctx_gpu_mc_ptr = mes->sch_ctx_gpu_addr;
-	mes_set_hw_res_pkt.query_status_fence_gpu_mc_ptr =
-		mes->query_status_fence_gpu_addr;
-
-	for (i = 0; i < MAX_COMPUTE_PIPES; i++)
-		mes_set_hw_res_pkt.compute_hqd_mask[i] =
-			mes->compute_hqd_mask[i];
-
-	for (i = 0; i < MAX_GFX_PIPES; i++)
-		mes_set_hw_res_pkt.gfx_hqd_mask[i] = mes->gfx_hqd_mask[i];
-
-	for (i = 0; i < MAX_SDMA_PIPES; i++)
-		mes_set_hw_res_pkt.sdma_hqd_mask[i] = mes->sdma_hqd_mask[i];
-
-	for (i = 0; i < AMD_PRIORITY_NUM_LEVELS; i++)
-		mes_set_hw_res_pkt.aggregated_doorbells[i] =
-			mes->aggregated_doorbells[i];
-
-	for (i = 0; i < 5; i++) {
-		mes_set_hw_res_pkt.gc_base[i] = adev->reg_offset[GC_HWIP][0][i];
-		mes_set_hw_res_pkt.mmhub_base[i] =
-			adev->reg_offset[MMHUB_HWIP][0][i];
-		mes_set_hw_res_pkt.osssys_base[i] =
-			adev->reg_offset[OSSSYS_HWIP][0][i];
-	}
-
-	mes_set_hw_res_pkt.disable_reset = 1;
-	mes_set_hw_res_pkt.disable_mes_log = 1;
-	mes_set_hw_res_pkt.use_different_vmid_compute = 1;
-
-	return mes_v10_1_submit_pkt_and_poll_completion(mes,
-			&mes_set_hw_res_pkt, sizeof(mes_set_hw_res_pkt),
-			offsetof(union MESAPI_SET_HW_RESOURCES, api_status));
-}
-
-static void mes_v10_1_init_aggregated_doorbell(struct amdgpu_mes *mes)
-{
-	struct amdgpu_device *adev = mes->adev;
-	uint32_t data;
-
-	data = RREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL1);
-	data &= ~(CP_MES_DOORBELL_CONTROL1__DOORBELL_OFFSET_MASK |
-		  CP_MES_DOORBELL_CONTROL1__DOORBELL_EN_MASK |
-		  CP_MES_DOORBELL_CONTROL1__DOORBELL_HIT_MASK);
-	data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_LOW] <<
-		CP_MES_DOORBELL_CONTROL1__DOORBELL_OFFSET__SHIFT;
-	data |= 1 << CP_MES_DOORBELL_CONTROL1__DOORBELL_EN__SHIFT;
-	WREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL1, data);
-
-	data = RREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL2);
-	data &= ~(CP_MES_DOORBELL_CONTROL2__DOORBELL_OFFSET_MASK |
-		  CP_MES_DOORBELL_CONTROL2__DOORBELL_EN_MASK |
-		  CP_MES_DOORBELL_CONTROL2__DOORBELL_HIT_MASK);
-	data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_NORMAL] <<
-		CP_MES_DOORBELL_CONTROL2__DOORBELL_OFFSET__SHIFT;
-	data |= 1 << CP_MES_DOORBELL_CONTROL2__DOORBELL_EN__SHIFT;
-	WREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL2, data);
-
-	data = RREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL3);
-	data &= ~(CP_MES_DOORBELL_CONTROL3__DOORBELL_OFFSET_MASK |
-		  CP_MES_DOORBELL_CONTROL3__DOORBELL_EN_MASK |
-		  CP_MES_DOORBELL_CONTROL3__DOORBELL_HIT_MASK);
-	data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_MEDIUM] <<
-		CP_MES_DOORBELL_CONTROL3__DOORBELL_OFFSET__SHIFT;
-	data |= 1 << CP_MES_DOORBELL_CONTROL3__DOORBELL_EN__SHIFT;
-	WREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL3, data);
-
-	data = RREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL4);
-	data &= ~(CP_MES_DOORBELL_CONTROL4__DOORBELL_OFFSET_MASK |
-		  CP_MES_DOORBELL_CONTROL4__DOORBELL_EN_MASK |
-		  CP_MES_DOORBELL_CONTROL4__DOORBELL_HIT_MASK);
-	data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_HIGH] <<
-		CP_MES_DOORBELL_CONTROL4__DOORBELL_OFFSET__SHIFT;
-	data |= 1 << CP_MES_DOORBELL_CONTROL4__DOORBELL_EN__SHIFT;
-	WREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL4, data);
-
-	data = RREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL5);
-	data &= ~(CP_MES_DOORBELL_CONTROL5__DOORBELL_OFFSET_MASK |
-		  CP_MES_DOORBELL_CONTROL5__DOORBELL_EN_MASK |
-		  CP_MES_DOORBELL_CONTROL5__DOORBELL_HIT_MASK);
-	data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_REALTIME] <<
-		CP_MES_DOORBELL_CONTROL5__DOORBELL_OFFSET__SHIFT;
-	data |= 1 << CP_MES_DOORBELL_CONTROL5__DOORBELL_EN__SHIFT;
-	WREG32_SOC15(GC, 0, mmCP_MES_DOORBELL_CONTROL5, data);
-
-	data = 1 << CP_HQD_GFX_CONTROL__DB_UPDATED_MSG_EN__SHIFT;
-	WREG32_SOC15(GC, 0, mmCP_HQD_GFX_CONTROL, data);
-}
-
-static const struct amdgpu_mes_funcs mes_v10_1_funcs = {
-	.add_hw_queue = mes_v10_1_add_hw_queue,
-	.remove_hw_queue = mes_v10_1_remove_hw_queue,
-	.unmap_legacy_queue = mes_v10_1_unmap_legacy_queue,
-	.suspend_gang = mes_v10_1_suspend_gang,
-	.resume_gang = mes_v10_1_resume_gang,
-};
-
-static int mes_v10_1_allocate_ucode_buffer(struct amdgpu_device *adev,
-					   enum admgpu_mes_pipe pipe)
-{
-	int r;
-	const struct mes_firmware_header_v1_0 *mes_hdr;
-	const __le32 *fw_data;
-	unsigned fw_size;
-
-	mes_hdr = (const struct mes_firmware_header_v1_0 *)
-		adev->mes.fw[pipe]->data;
-
-	fw_data = (const __le32 *)(adev->mes.fw[pipe]->data +
-		   le32_to_cpu(mes_hdr->mes_ucode_offset_bytes));
-	fw_size = le32_to_cpu(mes_hdr->mes_ucode_size_bytes);
-
-	r = amdgpu_bo_create_reserved(adev, fw_size,
-				      PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT,
-				      &adev->mes.ucode_fw_obj[pipe],
-				      &adev->mes.ucode_fw_gpu_addr[pipe],
-				      (void **)&adev->mes.ucode_fw_ptr[pipe]);
-	if (r) {
-		dev_err(adev->dev, "(%d) failed to create mes fw bo\n", r);
-		return r;
-	}
-
-	memcpy(adev->mes.ucode_fw_ptr[pipe], fw_data, fw_size);
-
-	amdgpu_bo_kunmap(adev->mes.ucode_fw_obj[pipe]);
-	amdgpu_bo_unreserve(adev->mes.ucode_fw_obj[pipe]);
-
-	return 0;
-}
-
-static int mes_v10_1_allocate_ucode_data_buffer(struct amdgpu_device *adev,
-						enum admgpu_mes_pipe pipe)
-{
-	int r;
-	const struct mes_firmware_header_v1_0 *mes_hdr;
-	const __le32 *fw_data;
-	unsigned fw_size;
-
-	mes_hdr = (const struct mes_firmware_header_v1_0 *)
-		adev->mes.fw[pipe]->data;
-
-	fw_data = (const __le32 *)(adev->mes.fw[pipe]->data +
-		   le32_to_cpu(mes_hdr->mes_ucode_data_offset_bytes));
-	fw_size = le32_to_cpu(mes_hdr->mes_ucode_data_size_bytes);
-
-	r = amdgpu_bo_create_reserved(adev, fw_size,
-				      64 * 1024, AMDGPU_GEM_DOMAIN_GTT,
-				      &adev->mes.data_fw_obj[pipe],
-				      &adev->mes.data_fw_gpu_addr[pipe],
-				      (void **)&adev->mes.data_fw_ptr[pipe]);
-	if (r) {
-		dev_err(adev->dev, "(%d) failed to create mes data fw bo\n", r);
-		return r;
-	}
-
-	memcpy(adev->mes.data_fw_ptr[pipe], fw_data, fw_size);
-
-	amdgpu_bo_kunmap(adev->mes.data_fw_obj[pipe]);
-	amdgpu_bo_unreserve(adev->mes.data_fw_obj[pipe]);
-
-	return 0;
-}
-
-static void mes_v10_1_free_ucode_buffers(struct amdgpu_device *adev,
-					 enum admgpu_mes_pipe pipe)
-{
-	amdgpu_bo_free_kernel(&adev->mes.data_fw_obj[pipe],
-			      &adev->mes.data_fw_gpu_addr[pipe],
-			      (void **)&adev->mes.data_fw_ptr[pipe]);
-
-	amdgpu_bo_free_kernel(&adev->mes.ucode_fw_obj[pipe],
-			      &adev->mes.ucode_fw_gpu_addr[pipe],
-			      (void **)&adev->mes.ucode_fw_ptr[pipe]);
-}
-
-static void mes_v10_1_enable(struct amdgpu_device *adev, bool enable)
-{
-	uint32_t pipe, data = 0;
-
-	if (enable) {
-		data = RREG32_SOC15(GC, 0, mmCP_MES_CNTL);
-		data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE0_RESET, 1);
-		data = REG_SET_FIELD(data, CP_MES_CNTL,
-			     MES_PIPE1_RESET, adev->enable_mes_kiq ? 1 : 0);
-		WREG32_SOC15(GC, 0, mmCP_MES_CNTL, data);
-
-		mutex_lock(&adev->srbm_mutex);
-		for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) {
-			if (!adev->enable_mes_kiq &&
-			    pipe == AMDGPU_MES_KIQ_PIPE)
-				continue;
-
-			nv_grbm_select(adev, 3, pipe, 0, 0);
-			WREG32_SOC15(GC, 0, mmCP_MES_PRGRM_CNTR_START,
-			     (uint32_t)(adev->mes.uc_start_addr[pipe]) >> 2);
-		}
-		nv_grbm_select(adev, 0, 0, 0, 0);
-		mutex_unlock(&adev->srbm_mutex);
-
-		/* clear BYPASS_UNCACHED to avoid hangs after interrupt. */
-		data = RREG32_SOC15(GC, 0, mmCP_MES_DC_OP_CNTL);
-		data = REG_SET_FIELD(data, CP_MES_DC_OP_CNTL,
-				     BYPASS_UNCACHED, 0);
-		WREG32_SOC15(GC, 0, mmCP_MES_DC_OP_CNTL, data);
-
-		/* unhalt MES and activate pipe0 */
-		data = REG_SET_FIELD(0, CP_MES_CNTL, MES_PIPE0_ACTIVE, 1);
-		data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_ACTIVE,
-				     adev->enable_mes_kiq ? 1 : 0);
-		WREG32_SOC15(GC, 0, mmCP_MES_CNTL, data);
-		udelay(100);
-	} else {
-		data = RREG32_SOC15(GC, 0, mmCP_MES_CNTL);
-		data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE0_ACTIVE, 0);
-		data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_ACTIVE, 0);
-		data = REG_SET_FIELD(data, CP_MES_CNTL,
-				     MES_INVALIDATE_ICACHE, 1);
-		data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE0_RESET, 1);
-		data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_RESET,
-				     adev->enable_mes_kiq ? 1 : 0);
-		data = REG_SET_FIELD(data, CP_MES_CNTL, MES_HALT, 1);
-		WREG32_SOC15(GC, 0, mmCP_MES_CNTL, data);
-	}
-}
-
-/* This function is for backdoor MES firmware */
-static int mes_v10_1_load_microcode(struct amdgpu_device *adev,
-				    enum admgpu_mes_pipe pipe)
-{
-	int r;
-	uint32_t data;
-
-	mes_v10_1_enable(adev, false);
-
-	if (!adev->mes.fw[pipe])
-		return -EINVAL;
-
-	r = mes_v10_1_allocate_ucode_buffer(adev, pipe);
-	if (r)
-		return r;
-
-	r = mes_v10_1_allocate_ucode_data_buffer(adev, pipe);
-	if (r) {
-		mes_v10_1_free_ucode_buffers(adev, pipe);
-		return r;
-	}
-
-	WREG32_SOC15(GC, 0, mmCP_MES_IC_BASE_CNTL, 0);
-
-	mutex_lock(&adev->srbm_mutex);
-	/* me=3, pipe=0, queue=0 */
-	nv_grbm_select(adev, 3, pipe, 0, 0);
-
-	/* set ucode start address */
-	WREG32_SOC15(GC, 0, mmCP_MES_PRGRM_CNTR_START,
-		     (uint32_t)(adev->mes.uc_start_addr[pipe]) >> 2);
-
-	/* set ucode fimrware address */
-	WREG32_SOC15(GC, 0, mmCP_MES_IC_BASE_LO,
-		     lower_32_bits(adev->mes.ucode_fw_gpu_addr[pipe]));
-	WREG32_SOC15(GC, 0, mmCP_MES_IC_BASE_HI,
-		     upper_32_bits(adev->mes.ucode_fw_gpu_addr[pipe]));
-
-	/* set ucode instruction cache boundary to 2M-1 */
-	WREG32_SOC15(GC, 0, mmCP_MES_MIBOUND_LO, 0x1FFFFF);
-
-	/* set ucode data firmware address */
-	WREG32_SOC15(GC, 0, mmCP_MES_MDBASE_LO,
-		     lower_32_bits(adev->mes.data_fw_gpu_addr[pipe]));
-	WREG32_SOC15(GC, 0, mmCP_MES_MDBASE_HI,
-		     upper_32_bits(adev->mes.data_fw_gpu_addr[pipe]));
-
-	/* Set 0x3FFFF (256K-1) to CP_MES_MDBOUND_LO */
-	WREG32_SOC15(GC, 0, mmCP_MES_MDBOUND_LO, 0x3FFFF);
-
-	/* invalidate ICACHE */
-	switch (adev->ip_versions[GC_HWIP][0]) {
-	case IP_VERSION(10, 3, 0):
-		data = RREG32_SOC15(GC, 0, mmCP_MES_IC_OP_CNTL_Sienna_Cichlid);
-		break;
-	default:
-		data = RREG32_SOC15(GC, 0, mmCP_MES_IC_OP_CNTL);
-		break;
-	}
-	data = REG_SET_FIELD(data, CP_MES_IC_OP_CNTL, PRIME_ICACHE, 0);
-	data = REG_SET_FIELD(data, CP_MES_IC_OP_CNTL, INVALIDATE_CACHE, 1);
-	switch (adev->ip_versions[GC_HWIP][0]) {
-	case IP_VERSION(10, 3, 0):
-		WREG32_SOC15(GC, 0, mmCP_MES_IC_OP_CNTL_Sienna_Cichlid, data);
-		break;
-	default:
-		WREG32_SOC15(GC, 0, mmCP_MES_IC_OP_CNTL, data);
-		break;
-	}
-
-	/* prime the ICACHE. */
-	switch (adev->ip_versions[GC_HWIP][0]) {
-	case IP_VERSION(10, 3, 0):
-		data = RREG32_SOC15(GC, 0, mmCP_MES_IC_OP_CNTL_Sienna_Cichlid);
-		break;
-	default:
-		data = RREG32_SOC15(GC, 0, mmCP_MES_IC_OP_CNTL);
-		break;
-	}
-	data = REG_SET_FIELD(data, CP_MES_IC_OP_CNTL, PRIME_ICACHE, 1);
-	switch (adev->ip_versions[GC_HWIP][0]) {
-	case IP_VERSION(10, 3, 0):
-		WREG32_SOC15(GC, 0, mmCP_MES_IC_OP_CNTL_Sienna_Cichlid, data);
-		break;
-	default:
-		WREG32_SOC15(GC, 0, mmCP_MES_IC_OP_CNTL, data);
-		break;
-	}
-
-	nv_grbm_select(adev, 0, 0, 0, 0);
-	mutex_unlock(&adev->srbm_mutex);
-
-	return 0;
-}
-
-static int mes_v10_1_allocate_eop_buf(struct amdgpu_device *adev,
-				      enum admgpu_mes_pipe pipe)
-{
-	int r;
-	u32 *eop;
-
-	r = amdgpu_bo_create_reserved(adev, MES_EOP_SIZE, PAGE_SIZE,
-			      AMDGPU_GEM_DOMAIN_GTT,
-			      &adev->mes.eop_gpu_obj[pipe],
-			      &adev->mes.eop_gpu_addr[pipe],
-			      (void **)&eop);
-	if (r) {
-		dev_warn(adev->dev, "(%d) create EOP bo failed\n", r);
-		return r;
-	}
-
-	memset(eop, 0, adev->mes.eop_gpu_obj[pipe]->tbo.base.size);
-
-	amdgpu_bo_kunmap(adev->mes.eop_gpu_obj[pipe]);
-	amdgpu_bo_unreserve(adev->mes.eop_gpu_obj[pipe]);
-
-	return 0;
-}
-
-static int mes_v10_1_mqd_init(struct amdgpu_ring *ring)
-{
-	struct v10_compute_mqd *mqd = ring->mqd_ptr;
-	uint64_t hqd_gpu_addr, wb_gpu_addr, eop_base_addr;
-	uint32_t tmp;
-
-	mqd->header = 0xC0310800;
-	mqd->compute_pipelinestat_enable = 0x00000001;
-	mqd->compute_static_thread_mgmt_se0 = 0xffffffff;
-	mqd->compute_static_thread_mgmt_se1 = 0xffffffff;
-	mqd->compute_static_thread_mgmt_se2 = 0xffffffff;
-	mqd->compute_static_thread_mgmt_se3 = 0xffffffff;
-	mqd->compute_misc_reserved = 0x00000003;
-
-	eop_base_addr = ring->eop_gpu_addr >> 8;
-
-	/* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */
-	tmp = mmCP_HQD_EOP_CONTROL_DEFAULT;
-	tmp = REG_SET_FIELD(tmp, CP_HQD_EOP_CONTROL, EOP_SIZE,
-			(order_base_2(MES_EOP_SIZE / 4) - 1));
-
-	mqd->cp_hqd_eop_base_addr_lo = lower_32_bits(eop_base_addr);
-	mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr);
-	mqd->cp_hqd_eop_control = tmp;
-
-	/* disable the queue if it's active */
-	ring->wptr = 0;
-	mqd->cp_hqd_pq_rptr = 0;
-	mqd->cp_hqd_pq_wptr_lo = 0;
-	mqd->cp_hqd_pq_wptr_hi = 0;
-
-	/* set the pointer to the MQD */
-	mqd->cp_mqd_base_addr_lo = ring->mqd_gpu_addr & 0xfffffffc;
-	mqd->cp_mqd_base_addr_hi = upper_32_bits(ring->mqd_gpu_addr);
-
-	/* set MQD vmid to 0 */
-	tmp = mmCP_MQD_CONTROL_DEFAULT;
-	tmp = REG_SET_FIELD(tmp, CP_MQD_CONTROL, VMID, 0);
-	mqd->cp_mqd_control = tmp;
-
-	/* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */
-	hqd_gpu_addr = ring->gpu_addr >> 8;
-	mqd->cp_hqd_pq_base_lo = lower_32_bits(hqd_gpu_addr);
-	mqd->cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr);
-
-	/* set the wb address whether it's enabled or not */
-	wb_gpu_addr = ring->rptr_gpu_addr;
-	mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc;
-	mqd->cp_hqd_pq_rptr_report_addr_hi =
-		upper_32_bits(wb_gpu_addr) & 0xffff;
-
-	/* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */
-	wb_gpu_addr = ring->wptr_gpu_addr;
-	mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffff8;
-	mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff;
-
-	/* set up the HQD, this is similar to CP_RB0_CNTL */
-	tmp = mmCP_HQD_PQ_CONTROL_DEFAULT;
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE,
-			    (order_base_2(ring->ring_size / 4) - 1));
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE,
-			    ((order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1) << 8));
-#ifdef __BIG_ENDIAN
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, ENDIAN_SWAP, 1);
-#endif
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 1);
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TUNNEL_DISPATCH, 0);
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, PRIV_STATE, 1);
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, KMD_QUEUE, 1);
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, NO_UPDATE_RPTR, 1);
-	mqd->cp_hqd_pq_control = tmp;
-
-	/* enable doorbell? */
-	tmp = 0;
-	if (ring->use_doorbell) {
-		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
-				    DOORBELL_OFFSET, ring->doorbell_index);
-		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
-				    DOORBELL_EN, 1);
-		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
-				    DOORBELL_SOURCE, 0);
-		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
-				    DOORBELL_HIT, 0);
-	}
-	else
-		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
-				    DOORBELL_EN, 0);
-	mqd->cp_hqd_pq_doorbell_control = tmp;
-
-	mqd->cp_hqd_vmid = 0;
-	/* activate the queue */
-	mqd->cp_hqd_active = 1;
-	mqd->cp_hqd_persistent_state = mmCP_HQD_PERSISTENT_STATE_DEFAULT;
-	mqd->cp_hqd_ib_control = mmCP_HQD_IB_CONTROL_DEFAULT;
-	mqd->cp_hqd_iq_timer = mmCP_HQD_IQ_TIMER_DEFAULT;
-	mqd->cp_hqd_quantum = mmCP_HQD_QUANTUM_DEFAULT;
-
-	tmp = mmCP_HQD_GFX_CONTROL_DEFAULT;
-	tmp = REG_SET_FIELD(tmp, CP_HQD_GFX_CONTROL, DB_UPDATED_MSG_EN, 1);
-	/* offset: 184 - this is used for CP_HQD_GFX_CONTROL */
-	mqd->cp_hqd_suspend_cntl_stack_offset = tmp;
-
-	return 0;
-}
-
-#if 0
-static void mes_v10_1_queue_init_register(struct amdgpu_ring *ring)
-{
-	struct v10_compute_mqd *mqd = ring->mqd_ptr;
-	struct amdgpu_device *adev = ring->adev;
-	uint32_t data = 0;
-
-	mutex_lock(&adev->srbm_mutex);
-	nv_grbm_select(adev, 3, ring->pipe, 0, 0);
-
-	/* set CP_HQD_VMID.VMID = 0. */
-	data = RREG32_SOC15(GC, 0, mmCP_HQD_VMID);
-	data = REG_SET_FIELD(data, CP_HQD_VMID, VMID, 0);
-	WREG32_SOC15(GC, 0, mmCP_HQD_VMID, data);
-
-	/* set CP_HQD_PQ_DOORBELL_CONTROL.DOORBELL_EN=0 */
-	data = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL);
-	data = REG_SET_FIELD(data, CP_HQD_PQ_DOORBELL_CONTROL,
-			     DOORBELL_EN, 0);
-	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL, data);
-
-	/* set CP_MQD_BASE_ADDR/HI with the MQD base address */
-	WREG32_SOC15(GC, 0, mmCP_MQD_BASE_ADDR, mqd->cp_mqd_base_addr_lo);
-	WREG32_SOC15(GC, 0, mmCP_MQD_BASE_ADDR_HI, mqd->cp_mqd_base_addr_hi);
-
-	/* set CP_MQD_CONTROL.VMID=0 */
-	data = RREG32_SOC15(GC, 0, mmCP_MQD_CONTROL);
-	data = REG_SET_FIELD(data, CP_MQD_CONTROL, VMID, 0);
-	WREG32_SOC15(GC, 0, mmCP_MQD_CONTROL, 0);
-
-	/* set CP_HQD_PQ_BASE/HI with the ring buffer base address */
-	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_BASE, mqd->cp_hqd_pq_base_lo);
-	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_BASE_HI, mqd->cp_hqd_pq_base_hi);
-
-	/* set CP_HQD_PQ_RPTR_REPORT_ADDR/HI */
-	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_RPTR_REPORT_ADDR,
-		     mqd->cp_hqd_pq_rptr_report_addr_lo);
-	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_RPTR_REPORT_ADDR_HI,
-		     mqd->cp_hqd_pq_rptr_report_addr_hi);
-
-	/* set CP_HQD_PQ_CONTROL */
-	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_CONTROL, mqd->cp_hqd_pq_control);
-
-	/* set CP_HQD_PQ_WPTR_POLL_ADDR/HI */
-	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR,
-		     mqd->cp_hqd_pq_wptr_poll_addr_lo);
-	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR_HI,
-		     mqd->cp_hqd_pq_wptr_poll_addr_hi);
-
-	/* set CP_HQD_PQ_DOORBELL_CONTROL */
-	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL,
-		     mqd->cp_hqd_pq_doorbell_control);
-
-	/* set CP_HQD_PERSISTENT_STATE.PRELOAD_SIZE=0x53 */
-	WREG32_SOC15(GC, 0, mmCP_HQD_PERSISTENT_STATE, mqd->cp_hqd_persistent_state);
-
-	/* set CP_HQD_ACTIVE.ACTIVE=1 */
-	WREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE, mqd->cp_hqd_active);
-
-	nv_grbm_select(adev, 0, 0, 0, 0);
-	mutex_unlock(&adev->srbm_mutex);
-}
-#endif
-
-static int mes_v10_1_kiq_enable_queue(struct amdgpu_device *adev)
-{
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
-	int r;
-
-	if (!kiq->pmf || !kiq->pmf->kiq_map_queues)
-		return -EINVAL;
-
-	r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size);
-	if (r) {
-		DRM_ERROR("Failed to lock KIQ (%d).\n", r);
-		return r;
-	}
-
-	kiq->pmf->kiq_map_queues(kiq_ring, &adev->mes.ring);
-
-	r = amdgpu_ring_test_ring(kiq_ring);
-	if (r) {
-		DRM_ERROR("kfq enable failed\n");
-		kiq_ring->sched.ready = false;
-	}
-
-	return r;
-}
-
-static int mes_v10_1_queue_init(struct amdgpu_device *adev)
-{
-	int r;
-
-	r = mes_v10_1_mqd_init(&adev->mes.ring);
-	if (r)
-		return r;
-
-	r = mes_v10_1_kiq_enable_queue(adev);
-	if (r)
-		return r;
-
-	return 0;
-}
-
-static int mes_v10_1_ring_init(struct amdgpu_device *adev)
-{
-	struct amdgpu_ring *ring;
-
-	ring = &adev->mes.ring;
-
-	ring->funcs = &mes_v10_1_ring_funcs;
-
-	ring->me = 3;
-	ring->pipe = 0;
-	ring->queue = 0;
-
-	ring->ring_obj = NULL;
-	ring->use_doorbell = true;
-	ring->doorbell_index = adev->doorbell_index.mes_ring0 << 1;
-	ring->eop_gpu_addr = adev->mes.eop_gpu_addr[AMDGPU_MES_SCHED_PIPE];
-	ring->no_scheduler = true;
-	sprintf(ring->name, "mes_%d.%d.%d", ring->me, ring->pipe, ring->queue);
-
-	return amdgpu_ring_init(adev, ring, 1024, NULL, 0,
-				AMDGPU_RING_PRIO_DEFAULT, NULL);
-}
-
-static int mes_v10_1_kiq_ring_init(struct amdgpu_device *adev)
-{
-	struct amdgpu_ring *ring;
-
-	spin_lock_init(&adev->gfx.kiq.ring_lock);
-
-	ring = &adev->gfx.kiq.ring;
-
-	ring->me = 3;
-	ring->pipe = 1;
-	ring->queue = 0;
-
-	ring->adev = NULL;
-	ring->ring_obj = NULL;
-	ring->use_doorbell = true;
-	ring->doorbell_index = adev->doorbell_index.mes_ring1 << 1;
-	ring->eop_gpu_addr = adev->mes.eop_gpu_addr[AMDGPU_MES_KIQ_PIPE];
-	ring->no_scheduler = true;
-	sprintf(ring->name, "mes_kiq_%d.%d.%d",
-		ring->me, ring->pipe, ring->queue);
-
-	return amdgpu_ring_init(adev, ring, 1024, NULL, 0,
-				AMDGPU_RING_PRIO_DEFAULT, NULL);
-}
-
-static int mes_v10_1_mqd_sw_init(struct amdgpu_device *adev,
-				 enum admgpu_mes_pipe pipe)
-{
-	int r, mqd_size = sizeof(struct v10_compute_mqd);
-	struct amdgpu_ring *ring;
-
-	if (pipe == AMDGPU_MES_KIQ_PIPE)
-		ring = &adev->gfx.kiq.ring;
-	else if (pipe == AMDGPU_MES_SCHED_PIPE)
-		ring = &adev->mes.ring;
-	else
-		BUG();
-
-	if (ring->mqd_obj)
-		return 0;
-
-	r = amdgpu_bo_create_kernel(adev, mqd_size, PAGE_SIZE,
-				    AMDGPU_GEM_DOMAIN_GTT, &ring->mqd_obj,
-				    &ring->mqd_gpu_addr, &ring->mqd_ptr);
-	if (r) {
-		dev_warn(adev->dev, "failed to create ring mqd bo (%d)", r);
-		return r;
-	}
-	memset(ring->mqd_ptr, 0, mqd_size);
-
-	/* prepare MQD backup */
-	adev->mes.mqd_backup[pipe] = kmalloc(mqd_size, GFP_KERNEL);
-	if (!adev->mes.mqd_backup[pipe])
-		dev_warn(adev->dev,
-			 "no memory to create MQD backup for ring %s\n",
-			 ring->name);
-
-	return 0;
-}
-
-static int mes_v10_1_sw_init(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	int pipe, r;
-
-	adev->mes.funcs = &mes_v10_1_funcs;
-	adev->mes.kiq_hw_init = &mes_v10_1_kiq_hw_init;
-
-	r = amdgpu_mes_init(adev);
-	if (r)
-		return r;
-
-	for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) {
-		if (!adev->enable_mes_kiq && pipe == AMDGPU_MES_KIQ_PIPE)
-			continue;
-
-		r = mes_v10_1_allocate_eop_buf(adev, pipe);
-		if (r)
-			return r;
-
-		r = mes_v10_1_mqd_sw_init(adev, pipe);
-		if (r)
-			return r;
-	}
-
-	if (adev->enable_mes_kiq) {
-		r = mes_v10_1_kiq_ring_init(adev);
-		if (r)
-			return r;
-	}
-
-	r = mes_v10_1_ring_init(adev);
-	if (r)
-		return r;
-
-	return 0;
-}
-
-static int mes_v10_1_sw_fini(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	int pipe;
-
-	amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs);
-	amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs);
-
-	for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) {
-		kfree(adev->mes.mqd_backup[pipe]);
-
-		amdgpu_bo_free_kernel(&adev->mes.eop_gpu_obj[pipe],
-				      &adev->mes.eop_gpu_addr[pipe],
-				      NULL);
-		amdgpu_ucode_release(&adev->mes.fw[pipe]);
-	}
-
-	amdgpu_bo_free_kernel(&adev->gfx.kiq.ring.mqd_obj,
-			      &adev->gfx.kiq.ring.mqd_gpu_addr,
-			      &adev->gfx.kiq.ring.mqd_ptr);
-
-	amdgpu_bo_free_kernel(&adev->mes.ring.mqd_obj,
-			      &adev->mes.ring.mqd_gpu_addr,
-			      &adev->mes.ring.mqd_ptr);
-
-	amdgpu_ring_fini(&adev->gfx.kiq.ring);
-	amdgpu_ring_fini(&adev->mes.ring);
-
-	amdgpu_mes_fini(adev);
-	return 0;
-}
-
-static void mes_v10_1_kiq_setting(struct amdgpu_ring *ring)
-{
-	uint32_t tmp;
-	struct amdgpu_device *adev = ring->adev;
-
-	/* tell RLC which is KIQ queue */
-	switch (adev->ip_versions[GC_HWIP][0]) {
-	case IP_VERSION(10, 3, 0):
-	case IP_VERSION(10, 3, 2):
-	case IP_VERSION(10, 3, 1):
-	case IP_VERSION(10, 3, 4):
-		tmp = RREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS_Sienna_Cichlid);
-		tmp &= 0xffffff00;
-		tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
-		WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS_Sienna_Cichlid, tmp);
-		tmp |= 0x80;
-		WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS_Sienna_Cichlid, tmp);
-		break;
-	default:
-		tmp = RREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS);
-		tmp &= 0xffffff00;
-		tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
-		WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS, tmp);
-		tmp |= 0x80;
-		WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS, tmp);
-		break;
-	}
-}
-
-static int mes_v10_1_kiq_hw_init(struct amdgpu_device *adev)
-{
-	int r = 0;
-
-	if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
-		r = mes_v10_1_load_microcode(adev, AMDGPU_MES_KIQ_PIPE);
-		if (r) {
-			DRM_ERROR("failed to load MES kiq fw, r=%d\n", r);
-			return r;
-		}
-
-		r = mes_v10_1_load_microcode(adev, AMDGPU_MES_SCHED_PIPE);
-		if (r) {
-			DRM_ERROR("failed to load MES fw, r=%d\n", r);
-			return r;
-		}
-	}
-
-	mes_v10_1_enable(adev, true);
-
-	mes_v10_1_kiq_setting(&adev->gfx.kiq.ring);
-
-	r = mes_v10_1_queue_init(adev);
-	if (r)
-		goto failure;
-
-	return r;
-
-failure:
-	mes_v10_1_hw_fini(adev);
-	return r;
-}
-
-static int mes_v10_1_hw_init(void *handle)
-{
-	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	if (!adev->enable_mes_kiq) {
-		if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
-			r = mes_v10_1_load_microcode(adev,
-					     AMDGPU_MES_SCHED_PIPE);
-			if (r) {
-				DRM_ERROR("failed to MES fw, r=%d\n", r);
-				return r;
-			}
-		}
-
-		mes_v10_1_enable(adev, true);
-	}
-
-	r = mes_v10_1_queue_init(adev);
-	if (r)
-		goto failure;
-
-	r = mes_v10_1_set_hw_resources(&adev->mes);
-	if (r)
-		goto failure;
-
-	mes_v10_1_init_aggregated_doorbell(&adev->mes);
-
-	r = mes_v10_1_query_sched_status(&adev->mes);
-	if (r) {
-		DRM_ERROR("MES is busy\n");
-		goto failure;
-	}
-
-	/*
-	 * Disable KIQ ring usage from the driver once MES is enabled.
-	 * MES uses KIQ ring exclusively so driver cannot access KIQ ring
-	 * with MES enabled.
-	 */
-	adev->gfx.kiq.ring.sched.ready = false;
-	adev->mes.ring.sched.ready = true;
-
-	return 0;
-
-failure:
-	mes_v10_1_hw_fini(adev);
-	return r;
-}
-
-static int mes_v10_1_hw_fini(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	adev->mes.ring.sched.ready = false;
-
-	mes_v10_1_enable(adev, false);
-
-	if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
-		mes_v10_1_free_ucode_buffers(adev, AMDGPU_MES_KIQ_PIPE);
-		mes_v10_1_free_ucode_buffers(adev, AMDGPU_MES_SCHED_PIPE);
-	}
-
-	return 0;
-}
-
-static int mes_v10_1_suspend(void *handle)
-{
-	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	r = amdgpu_mes_suspend(adev);
-	if (r)
-		return r;
-
-	return mes_v10_1_hw_fini(adev);
-}
-
-static int mes_v10_1_resume(void *handle)
-{
-	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	r = mes_v10_1_hw_init(adev);
-	if (r)
-		return r;
-
-	return amdgpu_mes_resume(adev);
-}
-
-static int mes_v10_0_early_init(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	int pipe, r;
-
-	for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) {
-		if (!adev->enable_mes_kiq && pipe == AMDGPU_MES_KIQ_PIPE)
-			continue;
-		r = amdgpu_mes_init_microcode(adev, pipe);
-		if (r)
-			return r;
-	}
-
-	return 0;
-}
-
-static int mes_v10_0_late_init(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	if (!amdgpu_in_reset(adev))
-		amdgpu_mes_self_test(adev);
-
-	return 0;
-}
-
-static const struct amd_ip_funcs mes_v10_1_ip_funcs = {
-	.name = "mes_v10_1",
-	.early_init = mes_v10_0_early_init,
-	.late_init = mes_v10_0_late_init,
-	.sw_init = mes_v10_1_sw_init,
-	.sw_fini = mes_v10_1_sw_fini,
-	.hw_init = mes_v10_1_hw_init,
-	.hw_fini = mes_v10_1_hw_fini,
-	.suspend = mes_v10_1_suspend,
-	.resume = mes_v10_1_resume,
-};
-
-const struct amdgpu_ip_block_version mes_v10_1_ip_block = {
-	.type = AMD_IP_BLOCK_TYPE_MES,
-	.major = 10,
-	.minor = 1,
-	.rev = 0,
-	.funcs = &mes_v10_1_ip_funcs,
-};
diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c
index 45280f047180..3a52754b5cad 100644
--- a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c
@@ -26,6 +26,7 @@
 #include "amdgpu.h"
 #include "soc15_common.h"
 #include "soc21.h"
+#include "gfx_v11_0.h"
 #include "gc/gc_11_0_0_offset.h"
 #include "gc/gc_11_0_0_sh_mask.h"
 #include "gc/gc_11_0_0_default.h"
@@ -47,12 +48,26 @@ MODULE_FIRMWARE("amdgpu/gc_11_0_3_mes1.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_4_mes.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_4_mes_2.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_4_mes1.bin");
-
-static int mes_v11_0_hw_fini(void *handle);
+MODULE_FIRMWARE("amdgpu/gc_11_5_0_mes_2.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_0_mes1.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_1_mes_2.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_1_mes1.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_2_mes_2.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_2_mes1.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_3_mes_2.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_3_mes1.bin");
+
+static int mes_v11_0_hw_init(struct amdgpu_ip_block *ip_block);
+static int mes_v11_0_hw_fini(struct amdgpu_ip_block *ip_block);
 static int mes_v11_0_kiq_hw_init(struct amdgpu_device *adev);
 static int mes_v11_0_kiq_hw_fini(struct amdgpu_device *adev);
 
 #define MES_EOP_SIZE   2048
+#define GFX_MES_DRAM_SIZE	0x80000
+#define MES11_HW_RESOURCE_1_SIZE (128 * AMDGPU_GPU_PAGE_SIZE)
+
+#define MES11_HUNG_DB_OFFSET_ARRAY_SIZE 8 /* [0:3] = db offset, [4:7] = hqd info */
+#define MES11_HUNG_HQD_INFO_OFFSET	4
 
 static void mes_v11_0_ring_set_wptr(struct amdgpu_ring *ring)
 {
@@ -94,18 +109,79 @@ static const struct amdgpu_ring_funcs mes_v11_0_ring_funcs = {
 	.insert_nop = amdgpu_ring_insert_nop,
 };
 
+static const char *mes_v11_0_opcodes[] = {
+	"SET_HW_RSRC",
+	"SET_SCHEDULING_CONFIG",
+	"ADD_QUEUE",
+	"REMOVE_QUEUE",
+	"PERFORM_YIELD",
+	"SET_GANG_PRIORITY_LEVEL",
+	"SUSPEND",
+	"RESUME",
+	"RESET",
+	"SET_LOG_BUFFER",
+	"CHANGE_GANG_PRORITY",
+	"QUERY_SCHEDULER_STATUS",
+	"PROGRAM_GDS",
+	"SET_DEBUG_VMID",
+	"MISC",
+	"UPDATE_ROOT_PAGE_TABLE",
+	"AMD_LOG",
+	"unused",
+	"unused",
+	"SET_HW_RSRC_1",
+};
+
+static const char *mes_v11_0_misc_opcodes[] = {
+	"WRITE_REG",
+	"INV_GART",
+	"QUERY_STATUS",
+	"READ_REG",
+	"WAIT_REG_MEM",
+	"SET_SHADER_DEBUGGER",
+};
+
+static const char *mes_v11_0_get_op_string(union MESAPI__MISC *x_pkt)
+{
+	const char *op_str = NULL;
+
+	if (x_pkt->header.opcode < ARRAY_SIZE(mes_v11_0_opcodes))
+		op_str = mes_v11_0_opcodes[x_pkt->header.opcode];
+
+	return op_str;
+}
+
+static const char *mes_v11_0_get_misc_op_string(union MESAPI__MISC *x_pkt)
+{
+	const char *op_str = NULL;
+
+	if ((x_pkt->header.opcode == MES_SCH_API_MISC) &&
+	    (x_pkt->opcode < ARRAY_SIZE(mes_v11_0_misc_opcodes)))
+		op_str = mes_v11_0_misc_opcodes[x_pkt->opcode];
+
+	return op_str;
+}
+
 static int mes_v11_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes,
 						    void *pkt, int size,
 						    int api_status_off)
 {
-	int ndw = size / 4;
-	signed long r;
-	union MESAPI__ADD_QUEUE *x_pkt = pkt;
-	struct MES_API_STATUS *api_status;
+	union MESAPI__QUERY_MES_STATUS mes_status_pkt;
+	signed long timeout = 2100000; /* 2100 ms */
 	struct amdgpu_device *adev = mes->adev;
-	struct amdgpu_ring *ring = &mes->ring;
+	struct amdgpu_ring *ring = &mes->ring[0];
+	struct MES_API_STATUS *api_status;
+	union MESAPI__MISC *x_pkt = pkt;
+	const char *op_str, *misc_op_str;
 	unsigned long flags;
-	signed long timeout = adev->usec_timeout;
+	u64 status_gpu_addr;
+	u32 seq, status_offset;
+	u64 *status_ptr;
+	signed long r;
+	int ret;
+
+	if (x_pkt->header.opcode >= MES_SCH_API_MAX)
+		return -EINVAL;
 
 	if (amdgpu_emu_mode) {
 		timeout *= 100;
@@ -113,37 +189,92 @@ static int mes_v11_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes,
 		/* Worst case in sriov where all other 15 VF timeout, each VF needs about 600ms */
 		timeout = 15 * 600 * 1000;
 	}
-	BUG_ON(size % 4 != 0);
 
-	spin_lock_irqsave(&mes->ring_lock, flags);
-	if (amdgpu_ring_alloc(ring, ndw)) {
-		spin_unlock_irqrestore(&mes->ring_lock, flags);
-		return -ENOMEM;
-	}
+	ret = amdgpu_device_wb_get(adev, &status_offset);
+	if (ret)
+		return ret;
+
+	status_gpu_addr = adev->wb.gpu_addr + (status_offset * 4);
+	status_ptr = (u64 *)&adev->wb.wb[status_offset];
+	*status_ptr = 0;
+
+	spin_lock_irqsave(&mes->ring_lock[0], flags);
+	r = amdgpu_ring_alloc(ring, (size + sizeof(mes_status_pkt)) / 4);
+	if (r)
+		goto error_unlock_free;
+
+	seq = ++ring->fence_drv.sync_seq;
+	r = amdgpu_fence_wait_polling(ring,
+				      seq - ring->fence_drv.num_fences_mask,
+				      timeout);
+	if (r < 1)
+		goto error_undo;
 
 	api_status = (struct MES_API_STATUS *)((char *)pkt + api_status_off);
-	api_status->api_completion_fence_addr = mes->ring.fence_drv.gpu_addr;
-	api_status->api_completion_fence_value = ++mes->ring.fence_drv.sync_seq;
+	api_status->api_completion_fence_addr = status_gpu_addr;
+	api_status->api_completion_fence_value = 1;
+
+	amdgpu_ring_write_multiple(ring, pkt, size / 4);
+
+	memset(&mes_status_pkt, 0, sizeof(mes_status_pkt));
+	mes_status_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_status_pkt.header.opcode = MES_SCH_API_QUERY_SCHEDULER_STATUS;
+	mes_status_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+	mes_status_pkt.api_status.api_completion_fence_addr =
+		ring->fence_drv.gpu_addr;
+	mes_status_pkt.api_status.api_completion_fence_value = seq;
+
+	amdgpu_ring_write_multiple(ring, &mes_status_pkt,
+				   sizeof(mes_status_pkt) / 4);
 
-	amdgpu_ring_write_multiple(ring, pkt, ndw);
 	amdgpu_ring_commit(ring);
-	spin_unlock_irqrestore(&mes->ring_lock, flags);
+	spin_unlock_irqrestore(&mes->ring_lock[0], flags);
 
-	DRM_DEBUG("MES msg=%d was emitted\n", x_pkt->header.opcode);
+	op_str = mes_v11_0_get_op_string(x_pkt);
+	misc_op_str = mes_v11_0_get_misc_op_string(x_pkt);
 
-	r = amdgpu_fence_wait_polling(ring, ring->fence_drv.sync_seq,
-		      timeout);
-	if (r < 1) {
-		DRM_ERROR("MES failed to response msg=%d\n",
-			  x_pkt->header.opcode);
+	if (misc_op_str)
+		dev_dbg(adev->dev, "MES msg=%s (%s) was emitted\n", op_str,
+			misc_op_str);
+	else if (op_str)
+		dev_dbg(adev->dev, "MES msg=%s was emitted\n", op_str);
+	else
+		dev_dbg(adev->dev, "MES msg=%d was emitted\n",
+			x_pkt->header.opcode);
+
+	r = amdgpu_fence_wait_polling(ring, seq, timeout);
+	if (r < 1 || !*status_ptr) {
+
+		if (misc_op_str)
+			dev_err(adev->dev, "MES failed to respond to msg=%s (%s)\n",
+				op_str, misc_op_str);
+		else if (op_str)
+			dev_err(adev->dev, "MES failed to respond to msg=%s\n",
+				op_str);
+		else
+			dev_err(adev->dev, "MES failed to respond to msg=%d\n",
+				x_pkt->header.opcode);
 
 		while (halt_if_hws_hang)
 			schedule();
 
-		return -ETIMEDOUT;
+		r = -ETIMEDOUT;
+		goto error_wb_free;
 	}
 
+	amdgpu_device_wb_free(adev, status_offset);
 	return 0;
+
+error_undo:
+	dev_err(adev->dev, "MES ring buffer is full.\n");
+	amdgpu_ring_undo(ring);
+
+error_unlock_free:
+	spin_unlock_irqrestore(&mes->ring_lock[0], flags);
+
+error_wb_free:
+	amdgpu_device_wb_free(adev, status_offset);
+	return r;
 }
 
 static int convert_to_mes_queue_type(int queue_type)
@@ -159,12 +290,29 @@ static int convert_to_mes_queue_type(int queue_type)
 	return -1;
 }
 
+static int convert_to_mes_priority_level(int priority_level)
+{
+	switch (priority_level) {
+	case AMDGPU_MES_PRIORITY_LEVEL_LOW:
+		return AMD_PRIORITY_LEVEL_LOW;
+	case AMDGPU_MES_PRIORITY_LEVEL_NORMAL:
+	default:
+		return AMD_PRIORITY_LEVEL_NORMAL;
+	case AMDGPU_MES_PRIORITY_LEVEL_MEDIUM:
+		return AMD_PRIORITY_LEVEL_MEDIUM;
+	case AMDGPU_MES_PRIORITY_LEVEL_HIGH:
+		return AMD_PRIORITY_LEVEL_HIGH;
+	case AMDGPU_MES_PRIORITY_LEVEL_REALTIME:
+		return AMD_PRIORITY_LEVEL_REALTIME;
+	}
+}
+
 static int mes_v11_0_add_hw_queue(struct amdgpu_mes *mes,
 				  struct mes_add_queue_input *input)
 {
 	struct amdgpu_device *adev = mes->adev;
 	union MESAPI__ADD_QUEUE mes_add_queue_pkt;
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	uint32_t vm_cntx_cntl = hub->vm_cntx_cntl;
 
 	memset(&mes_add_queue_pkt, 0, sizeof(mes_add_queue_pkt));
@@ -182,9 +330,9 @@ static int mes_v11_0_add_hw_queue(struct amdgpu_mes *mes,
 	mes_add_queue_pkt.gang_quantum = input->gang_quantum;
 	mes_add_queue_pkt.gang_context_addr = input->gang_context_addr;
 	mes_add_queue_pkt.inprocess_gang_priority =
-		input->inprocess_gang_priority;
+		convert_to_mes_priority_level(input->inprocess_gang_priority);
 	mes_add_queue_pkt.gang_global_priority_level =
-		input->gang_global_priority_level;
+		convert_to_mes_priority_level(input->gang_global_priority_level);
 	mes_add_queue_pkt.doorbell_offset = input->doorbell_offset;
 	mes_add_queue_pkt.mqd_addr = input->mqd_addr;
 
@@ -202,20 +350,15 @@ static int mes_v11_0_add_hw_queue(struct amdgpu_mes *mes,
 	mes_add_queue_pkt.gws_size = input->gws_size;
 	mes_add_queue_pkt.trap_handler_addr = input->tba_addr;
 	mes_add_queue_pkt.tma_addr = input->tma_addr;
+	mes_add_queue_pkt.trap_en = input->trap_en;
+	mes_add_queue_pkt.skip_process_ctx_clear = input->skip_process_ctx_clear;
 	mes_add_queue_pkt.is_kfd_process = input->is_kfd_process;
 
 	/* For KFD, gds_size is re-used for queue size (needed in MES for AQL queues) */
 	mes_add_queue_pkt.is_aql_queue = input->is_aql_queue;
 	mes_add_queue_pkt.gds_size = input->queue_size;
 
-	if (!(((adev->mes.sched_version & AMDGPU_MES_VERSION_MASK) >= 4) &&
-		  (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(11, 0, 0)) &&
-		  (adev->ip_versions[GC_HWIP][0] <= IP_VERSION(11, 0, 3))))
-		mes_add_queue_pkt.trap_en = 1;
-
-	/* For KFD, gds_size is re-used for queue size (needed in MES for AQL queues) */
-	mes_add_queue_pkt.is_aql_queue = input->is_aql_queue;
-	mes_add_queue_pkt.gds_size = input->queue_size;
+	mes_add_queue_pkt.exclusively_scheduled = input->exclusively_scheduled;
 
 	return mes_v11_0_submit_pkt_and_poll_completion(mes,
 			&mes_add_queue_pkt, sizeof(mes_add_queue_pkt),
@@ -226,6 +369,7 @@ static int mes_v11_0_remove_hw_queue(struct amdgpu_mes *mes,
 				     struct mes_remove_queue_input *input)
 {
 	union MESAPI__REMOVE_QUEUE mes_remove_queue_pkt;
+	uint32_t mes_rev = mes->sched_version & AMDGPU_MES_VERSION_MASK;
 
 	memset(&mes_remove_queue_pkt, 0, sizeof(mes_remove_queue_pkt));
 
@@ -236,11 +380,133 @@ static int mes_v11_0_remove_hw_queue(struct amdgpu_mes *mes,
 	mes_remove_queue_pkt.doorbell_offset = input->doorbell_offset;
 	mes_remove_queue_pkt.gang_context_addr = input->gang_context_addr;
 
+	if (mes_rev >= 0x60)
+		mes_remove_queue_pkt.remove_queue_after_reset = input->remove_queue_after_reset;
+
 	return mes_v11_0_submit_pkt_and_poll_completion(mes,
 			&mes_remove_queue_pkt, sizeof(mes_remove_queue_pkt),
 			offsetof(union MESAPI__REMOVE_QUEUE, api_status));
 }
 
+static int mes_v11_0_reset_queue_mmio(struct amdgpu_mes *mes, uint32_t queue_type,
+				      uint32_t me_id, uint32_t pipe_id,
+				      uint32_t queue_id, uint32_t vmid)
+{
+	struct amdgpu_device *adev = mes->adev;
+	uint32_t value, reg;
+	int i, r = 0;
+
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
+
+	if (queue_type == AMDGPU_RING_TYPE_GFX) {
+		dev_info(adev->dev, "reset gfx queue (%d:%d:%d: vmid:%d)\n",
+			 me_id, pipe_id, queue_id, vmid);
+
+		mutex_lock(&adev->gfx.reset_sem_mutex);
+		gfx_v11_0_request_gfx_index_mutex(adev, true);
+		/* all se allow writes */
+		WREG32_SOC15(GC, 0, regGRBM_GFX_INDEX,
+			     (uint32_t)(0x1 << GRBM_GFX_INDEX__SE_BROADCAST_WRITES__SHIFT));
+		value = REG_SET_FIELD(0, CP_VMID_RESET, RESET_REQUEST, 1 << vmid);
+		if (pipe_id == 0)
+			value = REG_SET_FIELD(value, CP_VMID_RESET, PIPE0_QUEUES, 1 << queue_id);
+		else
+			value = REG_SET_FIELD(value, CP_VMID_RESET, PIPE1_QUEUES, 1 << queue_id);
+		WREG32_SOC15(GC, 0, regCP_VMID_RESET, value);
+		gfx_v11_0_request_gfx_index_mutex(adev, false);
+		mutex_unlock(&adev->gfx.reset_sem_mutex);
+
+		mutex_lock(&adev->srbm_mutex);
+		soc21_grbm_select(adev, me_id, pipe_id, queue_id, 0);
+		/* wait till dequeue take effects */
+		for (i = 0; i < adev->usec_timeout; i++) {
+			if (!(RREG32_SOC15(GC, 0, regCP_GFX_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+		if (i >= adev->usec_timeout) {
+			dev_err(adev->dev, "failed to wait on gfx hqd deactivate\n");
+			r = -ETIMEDOUT;
+		}
+
+		soc21_grbm_select(adev, 0, 0, 0, 0);
+		mutex_unlock(&adev->srbm_mutex);
+	} else if (queue_type == AMDGPU_RING_TYPE_COMPUTE) {
+		dev_info(adev->dev, "reset compute queue (%d:%d:%d)\n",
+			 me_id, pipe_id, queue_id);
+		mutex_lock(&adev->srbm_mutex);
+		soc21_grbm_select(adev, me_id, pipe_id, queue_id, 0);
+		WREG32_SOC15(GC, 0, regCP_HQD_DEQUEUE_REQUEST, 0x2);
+		WREG32_SOC15(GC, 0, regSPI_COMPUTE_QUEUE_RESET, 0x1);
+
+		/* wait till dequeue take effects */
+		for (i = 0; i < adev->usec_timeout; i++) {
+			if (!(RREG32_SOC15(GC, 0, regCP_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+		if (i >= adev->usec_timeout) {
+			dev_err(adev->dev, "failed to wait on hqd deactivate\n");
+			r = -ETIMEDOUT;
+		}
+		soc21_grbm_select(adev, 0, 0, 0, 0);
+		mutex_unlock(&adev->srbm_mutex);
+	} else if (queue_type == AMDGPU_RING_TYPE_SDMA) {
+		dev_info(adev->dev, "reset sdma queue (%d:%d:%d)\n",
+			 me_id, pipe_id, queue_id);
+		switch (me_id) {
+		case 1:
+			reg = SOC15_REG_OFFSET(GC, 0, regSDMA1_QUEUE_RESET_REQ);
+			break;
+		case 0:
+		default:
+			reg = SOC15_REG_OFFSET(GC, 0, regSDMA0_QUEUE_RESET_REQ);
+			break;
+		}
+
+		value = 1 << queue_id;
+		WREG32(reg, value);
+		/* wait for queue reset done */
+		for (i = 0; i < adev->usec_timeout; i++) {
+			if (!(RREG32(reg) & value))
+				break;
+			udelay(1);
+		}
+		if (i >= adev->usec_timeout) {
+			dev_err(adev->dev, "failed to wait on sdma queue reset done\n");
+			r = -ETIMEDOUT;
+		}
+	}
+
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
+	return r;
+}
+
+static int mes_v11_0_map_legacy_queue(struct amdgpu_mes *mes,
+				      struct mes_map_legacy_queue_input *input)
+{
+	union MESAPI__ADD_QUEUE mes_add_queue_pkt;
+
+	memset(&mes_add_queue_pkt, 0, sizeof(mes_add_queue_pkt));
+
+	mes_add_queue_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_add_queue_pkt.header.opcode = MES_SCH_API_ADD_QUEUE;
+	mes_add_queue_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	mes_add_queue_pkt.pipe_id = input->pipe_id;
+	mes_add_queue_pkt.queue_id = input->queue_id;
+	mes_add_queue_pkt.doorbell_offset = input->doorbell_offset;
+	mes_add_queue_pkt.mqd_addr = input->mqd_addr;
+	mes_add_queue_pkt.wptr_addr = input->wptr_addr;
+	mes_add_queue_pkt.queue_type =
+		convert_to_mes_queue_type(input->queue_type);
+	mes_add_queue_pkt.map_legacy_kq = 1;
+
+	return mes_v11_0_submit_pkt_and_poll_completion(mes,
+			&mes_add_queue_pkt, sizeof(mes_add_queue_pkt),
+			offsetof(union MESAPI__ADD_QUEUE, api_status));
+}
+
 static int mes_v11_0_unmap_legacy_queue(struct amdgpu_mes *mes,
 			struct mes_unmap_legacy_queue_input *input)
 {
@@ -277,13 +543,41 @@ static int mes_v11_0_unmap_legacy_queue(struct amdgpu_mes *mes,
 static int mes_v11_0_suspend_gang(struct amdgpu_mes *mes,
 				  struct mes_suspend_gang_input *input)
 {
-	return 0;
+	union MESAPI__SUSPEND mes_suspend_gang_pkt;
+
+	memset(&mes_suspend_gang_pkt, 0, sizeof(mes_suspend_gang_pkt));
+
+	mes_suspend_gang_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_suspend_gang_pkt.header.opcode = MES_SCH_API_SUSPEND;
+	mes_suspend_gang_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	mes_suspend_gang_pkt.suspend_all_gangs = input->suspend_all_gangs;
+	mes_suspend_gang_pkt.gang_context_addr = input->gang_context_addr;
+	mes_suspend_gang_pkt.suspend_fence_addr = input->suspend_fence_addr;
+	mes_suspend_gang_pkt.suspend_fence_value = input->suspend_fence_value;
+
+	return mes_v11_0_submit_pkt_and_poll_completion(mes,
+			&mes_suspend_gang_pkt, sizeof(mes_suspend_gang_pkt),
+			offsetof(union MESAPI__SUSPEND, api_status));
 }
 
 static int mes_v11_0_resume_gang(struct amdgpu_mes *mes,
 				 struct mes_resume_gang_input *input)
 {
-	return 0;
+	union MESAPI__RESUME mes_resume_gang_pkt;
+
+	memset(&mes_resume_gang_pkt, 0, sizeof(mes_resume_gang_pkt));
+
+	mes_resume_gang_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_resume_gang_pkt.header.opcode = MES_SCH_API_RESUME;
+	mes_resume_gang_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	mes_resume_gang_pkt.resume_all_gangs = input->resume_all_gangs;
+	mes_resume_gang_pkt.gang_context_addr = input->gang_context_addr;
+
+	return mes_v11_0_submit_pkt_and_poll_completion(mes,
+			&mes_resume_gang_pkt, sizeof(mes_resume_gang_pkt),
+			offsetof(union MESAPI__RESUME, api_status));
 }
 
 static int mes_v11_0_query_sched_status(struct amdgpu_mes *mes)
@@ -339,6 +633,32 @@ static int mes_v11_0_misc_op(struct amdgpu_mes *mes,
 		misc_pkt.wait_reg_mem.reg_offset1 = input->wrm_reg.reg0;
 		misc_pkt.wait_reg_mem.reg_offset2 = input->wrm_reg.reg1;
 		break;
+	case MES_MISC_OP_SET_SHADER_DEBUGGER:
+		misc_pkt.opcode = MESAPI_MISC__SET_SHADER_DEBUGGER;
+		misc_pkt.set_shader_debugger.process_context_addr =
+				input->set_shader_debugger.process_context_addr;
+		misc_pkt.set_shader_debugger.flags.u32all =
+				input->set_shader_debugger.flags.u32all;
+		misc_pkt.set_shader_debugger.spi_gdbg_per_vmid_cntl =
+				input->set_shader_debugger.spi_gdbg_per_vmid_cntl;
+		memcpy(misc_pkt.set_shader_debugger.tcp_watch_cntl,
+				input->set_shader_debugger.tcp_watch_cntl,
+				sizeof(misc_pkt.set_shader_debugger.tcp_watch_cntl));
+		misc_pkt.set_shader_debugger.trap_en = input->set_shader_debugger.trap_en;
+		break;
+	case MES_MISC_OP_CHANGE_CONFIG:
+		if ((mes->adev->mes.sched_version & AMDGPU_MES_VERSION_MASK) < 0x63) {
+			dev_warn_once(mes->adev->dev,
+				      "MES FW version must be larger than 0x63 to support limit single process feature.\n");
+			return 0;
+		}
+		misc_pkt.opcode = MESAPI_MISC__CHANGE_CONFIG;
+		misc_pkt.change_config.opcode =
+				MESAPI_MISC__CHANGE_CONFIG_OPTION_LIMIT_SINGLE_PROCESS;
+		misc_pkt.change_config.option.bits.limit_single_process =
+				input->change_config.option.limit_single_process;
+		break;
+
 	default:
 		DRM_ERROR("unsupported misc op (%d) \n", input->op);
 		return -EINVAL;
@@ -365,16 +685,17 @@ static int mes_v11_0_set_hw_resources(struct amdgpu_mes *mes)
 	mes_set_hw_res_pkt.vmid_mask_gfxhub = mes->vmid_mask_gfxhub;
 	mes_set_hw_res_pkt.gds_size = adev->gds.gds_size;
 	mes_set_hw_res_pkt.paging_vmid = 0;
-	mes_set_hw_res_pkt.g_sch_ctx_gpu_mc_ptr = mes->sch_ctx_gpu_addr;
+	mes_set_hw_res_pkt.g_sch_ctx_gpu_mc_ptr = mes->sch_ctx_gpu_addr[0];
 	mes_set_hw_res_pkt.query_status_fence_gpu_mc_ptr =
-		mes->query_status_fence_gpu_addr;
+		mes->query_status_fence_gpu_addr[0];
 
 	for (i = 0; i < MAX_COMPUTE_PIPES; i++)
 		mes_set_hw_res_pkt.compute_hqd_mask[i] =
 			mes->compute_hqd_mask[i];
 
 	for (i = 0; i < MAX_GFX_PIPES; i++)
-		mes_set_hw_res_pkt.gfx_hqd_mask[i] = mes->gfx_hqd_mask[i];
+		mes_set_hw_res_pkt.gfx_hqd_mask[i] =
+			mes->gfx_hqd_mask[i];
 
 	for (i = 0; i < MAX_SDMA_PIPES; i++)
 		mes_set_hw_res_pkt.sdma_hqd_mask[i] = mes->sdma_hqd_mask[i];
@@ -395,78 +716,127 @@ static int mes_v11_0_set_hw_resources(struct amdgpu_mes *mes)
 	mes_set_hw_res_pkt.disable_mes_log = 1;
 	mes_set_hw_res_pkt.use_different_vmid_compute = 1;
 	mes_set_hw_res_pkt.enable_reg_active_poll = 1;
+	mes_set_hw_res_pkt.enable_level_process_quantum_check = 1;
 	mes_set_hw_res_pkt.oversubscription_timer = 50;
+	if ((mes->adev->mes.sched_version & AMDGPU_MES_VERSION_MASK) >= 0x7f)
+		mes_set_hw_res_pkt.enable_lr_compute_wa = 1;
+	else
+		dev_info_once(mes->adev->dev,
+			      "MES FW version must be >= 0x7f to enable LR compute workaround.\n");
+
+	if (amdgpu_mes_log_enable) {
+		mes_set_hw_res_pkt.enable_mes_event_int_logging = 1;
+		mes_set_hw_res_pkt.event_intr_history_gpu_mc_ptr =
+					mes->event_log_gpu_addr;
+	}
+
+	if (adev->enforce_isolation[0] == AMDGPU_ENFORCE_ISOLATION_ENABLE)
+		mes_set_hw_res_pkt.limit_single_process = 1;
 
 	return mes_v11_0_submit_pkt_and_poll_completion(mes,
 			&mes_set_hw_res_pkt, sizeof(mes_set_hw_res_pkt),
 			offsetof(union MESAPI_SET_HW_RESOURCES, api_status));
 }
 
-static void mes_v11_0_init_aggregated_doorbell(struct amdgpu_mes *mes)
+static int mes_v11_0_set_hw_resources_1(struct amdgpu_mes *mes)
 {
-	struct amdgpu_device *adev = mes->adev;
-	uint32_t data;
+	union MESAPI_SET_HW_RESOURCES_1 mes_set_hw_res_pkt;
+	memset(&mes_set_hw_res_pkt, 0, sizeof(mes_set_hw_res_pkt));
+
+	mes_set_hw_res_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_set_hw_res_pkt.header.opcode = MES_SCH_API_SET_HW_RSRC_1;
+	mes_set_hw_res_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+	mes_set_hw_res_pkt.enable_mes_info_ctx = 1;
+
+	mes_set_hw_res_pkt.cleaner_shader_fence_mc_addr = mes->resource_1_gpu_addr[0];
+	if (amdgpu_sriov_is_mes_info_enable(mes->adev)) {
+		mes_set_hw_res_pkt.mes_info_ctx_mc_addr =
+			mes->resource_1_gpu_addr[0] + AMDGPU_GPU_PAGE_SIZE;
+		mes_set_hw_res_pkt.mes_info_ctx_size = MES11_HW_RESOURCE_1_SIZE;
+	}
 
-	data = RREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL1);
-	data &= ~(CP_MES_DOORBELL_CONTROL1__DOORBELL_OFFSET_MASK |
-		  CP_MES_DOORBELL_CONTROL1__DOORBELL_EN_MASK |
-		  CP_MES_DOORBELL_CONTROL1__DOORBELL_HIT_MASK);
-	data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_LOW] <<
-		CP_MES_DOORBELL_CONTROL1__DOORBELL_OFFSET__SHIFT;
-	data |= 1 << CP_MES_DOORBELL_CONTROL1__DOORBELL_EN__SHIFT;
-	WREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL1, data);
-
-	data = RREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL2);
-	data &= ~(CP_MES_DOORBELL_CONTROL2__DOORBELL_OFFSET_MASK |
-		  CP_MES_DOORBELL_CONTROL2__DOORBELL_EN_MASK |
-		  CP_MES_DOORBELL_CONTROL2__DOORBELL_HIT_MASK);
-	data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_NORMAL] <<
-		CP_MES_DOORBELL_CONTROL2__DOORBELL_OFFSET__SHIFT;
-	data |= 1 << CP_MES_DOORBELL_CONTROL2__DOORBELL_EN__SHIFT;
-	WREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL2, data);
-
-	data = RREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL3);
-	data &= ~(CP_MES_DOORBELL_CONTROL3__DOORBELL_OFFSET_MASK |
-		  CP_MES_DOORBELL_CONTROL3__DOORBELL_EN_MASK |
-		  CP_MES_DOORBELL_CONTROL3__DOORBELL_HIT_MASK);
-	data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_MEDIUM] <<
-		CP_MES_DOORBELL_CONTROL3__DOORBELL_OFFSET__SHIFT;
-	data |= 1 << CP_MES_DOORBELL_CONTROL3__DOORBELL_EN__SHIFT;
-	WREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL3, data);
-
-	data = RREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL4);
-	data &= ~(CP_MES_DOORBELL_CONTROL4__DOORBELL_OFFSET_MASK |
-		  CP_MES_DOORBELL_CONTROL4__DOORBELL_EN_MASK |
-		  CP_MES_DOORBELL_CONTROL4__DOORBELL_HIT_MASK);
-	data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_HIGH] <<
-		CP_MES_DOORBELL_CONTROL4__DOORBELL_OFFSET__SHIFT;
-	data |= 1 << CP_MES_DOORBELL_CONTROL4__DOORBELL_EN__SHIFT;
-	WREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL4, data);
-
-	data = RREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL5);
-	data &= ~(CP_MES_DOORBELL_CONTROL5__DOORBELL_OFFSET_MASK |
-		  CP_MES_DOORBELL_CONTROL5__DOORBELL_EN_MASK |
-		  CP_MES_DOORBELL_CONTROL5__DOORBELL_HIT_MASK);
-	data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_REALTIME] <<
-		CP_MES_DOORBELL_CONTROL5__DOORBELL_OFFSET__SHIFT;
-	data |= 1 << CP_MES_DOORBELL_CONTROL5__DOORBELL_EN__SHIFT;
-	WREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL5, data);
-
-	data = 1 << CP_HQD_GFX_CONTROL__DB_UPDATED_MSG_EN__SHIFT;
-	WREG32_SOC15(GC, 0, regCP_HQD_GFX_CONTROL, data);
+	return mes_v11_0_submit_pkt_and_poll_completion(mes,
+			&mes_set_hw_res_pkt, sizeof(mes_set_hw_res_pkt),
+			offsetof(union MESAPI_SET_HW_RESOURCES_1, api_status));
+}
+
+static int mes_v11_0_reset_hw_queue(struct amdgpu_mes *mes,
+				    struct mes_reset_queue_input *input)
+{
+	union MESAPI__RESET mes_reset_queue_pkt;
+
+	if (input->use_mmio)
+		return mes_v11_0_reset_queue_mmio(mes, input->queue_type,
+						  input->me_id, input->pipe_id,
+						  input->queue_id, input->vmid);
+
+	memset(&mes_reset_queue_pkt, 0, sizeof(mes_reset_queue_pkt));
+
+	mes_reset_queue_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_reset_queue_pkt.header.opcode = MES_SCH_API_RESET;
+	mes_reset_queue_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	mes_reset_queue_pkt.queue_type =
+		convert_to_mes_queue_type(input->queue_type);
+
+	if (input->legacy_gfx) {
+		mes_reset_queue_pkt.reset_legacy_gfx = 1;
+		mes_reset_queue_pkt.pipe_id_lp = input->pipe_id;
+		mes_reset_queue_pkt.queue_id_lp = input->queue_id;
+		mes_reset_queue_pkt.mqd_mc_addr_lp = input->mqd_addr;
+		mes_reset_queue_pkt.doorbell_offset_lp = input->doorbell_offset;
+		mes_reset_queue_pkt.wptr_addr_lp = input->wptr_addr;
+		mes_reset_queue_pkt.vmid_id_lp = input->vmid;
+	} else {
+		mes_reset_queue_pkt.reset_queue_only = 1;
+		mes_reset_queue_pkt.doorbell_offset = input->doorbell_offset;
+	}
+
+	return mes_v11_0_submit_pkt_and_poll_completion(mes,
+			&mes_reset_queue_pkt, sizeof(mes_reset_queue_pkt),
+			offsetof(union MESAPI__RESET, api_status));
+}
+
+static int mes_v11_0_detect_and_reset_hung_queues(struct amdgpu_mes *mes,
+						  struct mes_detect_and_reset_queue_input *input)
+{
+	union MESAPI__RESET mes_reset_queue_pkt;
+
+	memset(&mes_reset_queue_pkt, 0, sizeof(mes_reset_queue_pkt));
+
+	mes_reset_queue_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_reset_queue_pkt.header.opcode = MES_SCH_API_RESET;
+	mes_reset_queue_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	mes_reset_queue_pkt.queue_type =
+		convert_to_mes_queue_type(input->queue_type);
+	mes_reset_queue_pkt.doorbell_offset_addr =
+		mes->hung_queue_db_array_gpu_addr;
+
+	if (input->detect_only)
+		mes_reset_queue_pkt.hang_detect_only = 1;
+	else
+		mes_reset_queue_pkt.hang_detect_then_reset = 1;
+
+	return mes_v11_0_submit_pkt_and_poll_completion(mes,
+			&mes_reset_queue_pkt, sizeof(mes_reset_queue_pkt),
+			offsetof(union MESAPI__RESET, api_status));
 }
 
 static const struct amdgpu_mes_funcs mes_v11_0_funcs = {
 	.add_hw_queue = mes_v11_0_add_hw_queue,
 	.remove_hw_queue = mes_v11_0_remove_hw_queue,
+	.map_legacy_queue = mes_v11_0_map_legacy_queue,
 	.unmap_legacy_queue = mes_v11_0_unmap_legacy_queue,
 	.suspend_gang = mes_v11_0_suspend_gang,
 	.resume_gang = mes_v11_0_resume_gang,
 	.misc_op = mes_v11_0_misc_op,
+	.reset_hw_queue = mes_v11_0_reset_hw_queue,
+	.detect_and_reset_hung_queues = mes_v11_0_detect_and_reset_hung_queues,
 };
 
 static int mes_v11_0_allocate_ucode_buffer(struct amdgpu_device *adev,
-					   enum admgpu_mes_pipe pipe)
+					   enum amdgpu_mes_pipe pipe)
 {
 	int r;
 	const struct mes_firmware_header_v1_0 *mes_hdr;
@@ -501,7 +871,7 @@ static int mes_v11_0_allocate_ucode_buffer(struct amdgpu_device *adev,
 }
 
 static int mes_v11_0_allocate_ucode_data_buffer(struct amdgpu_device *adev,
-						enum admgpu_mes_pipe pipe)
+						enum amdgpu_mes_pipe pipe)
 {
 	int r;
 	const struct mes_firmware_header_v1_0 *mes_hdr;
@@ -515,7 +885,13 @@ static int mes_v11_0_allocate_ucode_data_buffer(struct amdgpu_device *adev,
 		   le32_to_cpu(mes_hdr->mes_ucode_data_offset_bytes));
 	fw_size = le32_to_cpu(mes_hdr->mes_ucode_data_size_bytes);
 
-	r = amdgpu_bo_create_reserved(adev, fw_size,
+	if (fw_size > GFX_MES_DRAM_SIZE) {
+		dev_err(adev->dev, "PIPE%d ucode data fw size (%d) is greater than dram size (%d)\n",
+			pipe, fw_size, GFX_MES_DRAM_SIZE);
+		return -EINVAL;
+	}
+
+	r = amdgpu_bo_create_reserved(adev, GFX_MES_DRAM_SIZE,
 				      64 * 1024,
 				      AMDGPU_GEM_DOMAIN_VRAM |
 				      AMDGPU_GEM_DOMAIN_GTT,
@@ -536,7 +912,7 @@ static int mes_v11_0_allocate_ucode_data_buffer(struct amdgpu_device *adev,
 }
 
 static void mes_v11_0_free_ucode_buffers(struct amdgpu_device *adev,
-					 enum admgpu_mes_pipe pipe)
+					 enum amdgpu_mes_pipe pipe)
 {
 	amdgpu_bo_free_kernel(&adev->mes.data_fw_obj[pipe],
 			      &adev->mes.data_fw_gpu_addr[pipe],
@@ -547,12 +923,48 @@ static void mes_v11_0_free_ucode_buffers(struct amdgpu_device *adev,
 			      (void **)&adev->mes.ucode_fw_ptr[pipe]);
 }
 
+static void mes_v11_0_get_fw_version(struct amdgpu_device *adev)
+{
+	int pipe;
+
+	/* return early if we have already fetched these */
+	if (adev->mes.sched_version && adev->mes.kiq_version)
+		return;
+
+	/* get MES scheduler/KIQ versions */
+	mutex_lock(&adev->srbm_mutex);
+
+	for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) {
+		soc21_grbm_select(adev, 3, pipe, 0, 0);
+
+		if (pipe == AMDGPU_MES_SCHED_PIPE)
+			adev->mes.sched_version =
+				RREG32_SOC15(GC, 0, regCP_MES_GP3_LO);
+		else if (pipe == AMDGPU_MES_KIQ_PIPE && adev->enable_mes_kiq)
+			adev->mes.kiq_version =
+				RREG32_SOC15(GC, 0, regCP_MES_GP3_LO);
+	}
+
+	soc21_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+}
+
 static void mes_v11_0_enable(struct amdgpu_device *adev, bool enable)
 {
 	uint64_t ucode_addr;
 	uint32_t pipe, data = 0;
 
 	if (enable) {
+		if (amdgpu_mes_log_enable) {
+			WREG32_SOC15(GC, 0, regCP_MES_MSCRATCH_LO,
+				lower_32_bits(adev->mes.event_log_gpu_addr + AMDGPU_MES_LOG_BUFFER_SIZE));
+			WREG32_SOC15(GC, 0, regCP_MES_MSCRATCH_HI,
+				upper_32_bits(adev->mes.event_log_gpu_addr + AMDGPU_MES_LOG_BUFFER_SIZE));
+			dev_info(adev->dev, "Setup CP MES MSCRATCH address : 0x%x. 0x%x\n",
+				RREG32_SOC15(GC, 0, regCP_MES_MSCRATCH_HI),
+				RREG32_SOC15(GC, 0, regCP_MES_MSCRATCH_LO));
+		}
+
 		data = RREG32_SOC15(GC, 0, regCP_MES_CNTL);
 		data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE0_RESET, 1);
 		data = REG_SET_FIELD(data, CP_MES_CNTL,
@@ -585,7 +997,7 @@ static void mes_v11_0_enable(struct amdgpu_device *adev, bool enable)
 		if (amdgpu_emu_mode)
 			msleep(100);
 		else
-			udelay(50);
+			udelay(500);
 	} else {
 		data = RREG32_SOC15(GC, 0, regCP_MES_CNTL);
 		data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE0_ACTIVE, 0);
@@ -602,7 +1014,7 @@ static void mes_v11_0_enable(struct amdgpu_device *adev, bool enable)
 
 /* This function is for backdoor MES firmware */
 static int mes_v11_0_load_microcode(struct amdgpu_device *adev,
-				    enum admgpu_mes_pipe pipe, bool prime_icache)
+				    enum amdgpu_mes_pipe pipe, bool prime_icache)
 {
 	int r;
 	uint32_t data;
@@ -651,8 +1063,8 @@ static int mes_v11_0_load_microcode(struct amdgpu_device *adev,
 	WREG32_SOC15(GC, 0, regCP_MES_MDBASE_HI,
 		     upper_32_bits(adev->mes.data_fw_gpu_addr[pipe]));
 
-	/* Set 0x3FFFF (256K-1) to CP_MES_MDBOUND_LO */
-	WREG32_SOC15(GC, 0, regCP_MES_MDBOUND_LO, 0x3FFFF);
+	/* Set 0x7FFFF (512K-1) to CP_MES_MDBOUND_LO */
+	WREG32_SOC15(GC, 0, regCP_MES_MDBOUND_LO, 0x7FFFF);
 
 	if (prime_icache) {
 		/* invalidate ICACHE */
@@ -674,7 +1086,7 @@ static int mes_v11_0_load_microcode(struct amdgpu_device *adev,
 }
 
 static int mes_v11_0_allocate_eop_buf(struct amdgpu_device *adev,
-				      enum admgpu_mes_pipe pipe)
+				      enum amdgpu_mes_pipe pipe)
 {
 	int r;
 	u32 *eop;
@@ -704,6 +1116,8 @@ static int mes_v11_0_mqd_init(struct amdgpu_ring *ring)
 	uint64_t hqd_gpu_addr, wb_gpu_addr, eop_base_addr;
 	uint32_t tmp;
 
+	memset(mqd, 0, sizeof(*mqd));
+
 	mqd->header = 0xC0310800;
 	mqd->compute_pipelinestat_enable = 0x00000001;
 	mqd->compute_static_thread_mgmt_se0 = 0xffffffff;
@@ -778,8 +1192,7 @@ static int mes_v11_0_mqd_init(struct amdgpu_ring *ring)
 				    DOORBELL_SOURCE, 0);
 		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
 				    DOORBELL_HIT, 0);
-	}
-	else
+	} else
 		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
 				    DOORBELL_EN, 0);
 	mqd->cp_hqd_pq_doorbell_control = tmp;
@@ -797,6 +1210,7 @@ static int mes_v11_0_mqd_init(struct amdgpu_ring *ring)
 	mqd->cp_hqd_iq_timer = regCP_HQD_IQ_TIMER_DEFAULT;
 	mqd->cp_hqd_quantum = regCP_HQD_QUANTUM_DEFAULT;
 
+	amdgpu_device_flush_hdp(ring->adev, NULL);
 	return 0;
 }
 
@@ -864,8 +1278,8 @@ static void mes_v11_0_queue_init_register(struct amdgpu_ring *ring)
 
 static int mes_v11_0_kiq_enable_queue(struct amdgpu_device *adev)
 {
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
+	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq[0].ring;
 	int r;
 
 	if (!kiq->pmf || !kiq->pmf->kiq_map_queues)
@@ -877,26 +1291,21 @@ static int mes_v11_0_kiq_enable_queue(struct amdgpu_device *adev)
 		return r;
 	}
 
-	kiq->pmf->kiq_map_queues(kiq_ring, &adev->mes.ring);
+	kiq->pmf->kiq_map_queues(kiq_ring, &adev->mes.ring[0]);
 
-	r = amdgpu_ring_test_ring(kiq_ring);
-	if (r) {
-		DRM_ERROR("kfq enable failed\n");
-		kiq_ring->sched.ready = false;
-	}
-	return r;
+	return amdgpu_ring_test_helper(kiq_ring);
 }
 
 static int mes_v11_0_queue_init(struct amdgpu_device *adev,
-				enum admgpu_mes_pipe pipe)
+				enum amdgpu_mes_pipe pipe)
 {
 	struct amdgpu_ring *ring;
 	int r;
 
 	if (pipe == AMDGPU_MES_KIQ_PIPE)
-		ring = &adev->gfx.kiq.ring;
+		ring = &adev->gfx.kiq[0].ring;
 	else if (pipe == AMDGPU_MES_SCHED_PIPE)
-		ring = &adev->mes.ring;
+		ring = &adev->mes.ring[0];
 	else
 		BUG();
 
@@ -919,18 +1328,6 @@ static int mes_v11_0_queue_init(struct amdgpu_device *adev,
 		mes_v11_0_queue_init_register(ring);
 	}
 
-	/* get MES scheduler/KIQ versions */
-	mutex_lock(&adev->srbm_mutex);
-	soc21_grbm_select(adev, 3, pipe, 0, 0);
-
-	if (pipe == AMDGPU_MES_SCHED_PIPE)
-		adev->mes.sched_version = RREG32_SOC15(GC, 0, regCP_MES_GP3_LO);
-	else if (pipe == AMDGPU_MES_KIQ_PIPE && adev->enable_mes_kiq)
-		adev->mes.kiq_version = RREG32_SOC15(GC, 0, regCP_MES_GP3_LO);
-
-	soc21_grbm_select(adev, 0, 0, 0, 0);
-	mutex_unlock(&adev->srbm_mutex);
-
 	return 0;
 }
 
@@ -938,7 +1335,7 @@ static int mes_v11_0_ring_init(struct amdgpu_device *adev)
 {
 	struct amdgpu_ring *ring;
 
-	ring = &adev->mes.ring;
+	ring = &adev->mes.ring[0];
 
 	ring->funcs = &mes_v11_0_ring_funcs;
 
@@ -961,9 +1358,9 @@ static int mes_v11_0_kiq_ring_init(struct amdgpu_device *adev)
 {
 	struct amdgpu_ring *ring;
 
-	spin_lock_init(&adev->gfx.kiq.ring_lock);
+	spin_lock_init(&adev->gfx.kiq[0].ring_lock);
 
-	ring = &adev->gfx.kiq.ring;
+	ring = &adev->gfx.kiq[0].ring;
 
 	ring->me = 3;
 	ring->pipe = 1;
@@ -983,15 +1380,15 @@ static int mes_v11_0_kiq_ring_init(struct amdgpu_device *adev)
 }
 
 static int mes_v11_0_mqd_sw_init(struct amdgpu_device *adev,
-				 enum admgpu_mes_pipe pipe)
+				 enum amdgpu_mes_pipe pipe)
 {
 	int r, mqd_size = sizeof(struct v11_compute_mqd);
 	struct amdgpu_ring *ring;
 
 	if (pipe == AMDGPU_MES_KIQ_PIPE)
-		ring = &adev->gfx.kiq.ring;
+		ring = &adev->gfx.kiq[0].ring;
 	else if (pipe == AMDGPU_MES_SCHED_PIPE)
-		ring = &adev->mes.ring;
+		ring = &adev->mes.ring[0];
 	else
 		BUG();
 
@@ -999,6 +1396,7 @@ static int mes_v11_0_mqd_sw_init(struct amdgpu_device *adev,
 		return 0;
 
 	r = amdgpu_bo_create_kernel(adev, mqd_size, PAGE_SIZE,
+				    AMDGPU_GEM_DOMAIN_VRAM |
 				    AMDGPU_GEM_DOMAIN_GTT, &ring->mqd_obj,
 				    &ring->mqd_gpu_addr, &ring->mqd_ptr);
 	if (r) {
@@ -1010,23 +1408,27 @@ static int mes_v11_0_mqd_sw_init(struct amdgpu_device *adev,
 
 	/* prepare MQD backup */
 	adev->mes.mqd_backup[pipe] = kmalloc(mqd_size, GFP_KERNEL);
-	if (!adev->mes.mqd_backup[pipe])
+	if (!adev->mes.mqd_backup[pipe]) {
 		dev_warn(adev->dev,
 			 "no memory to create MQD backup for ring %s\n",
 			 ring->name);
+		return -ENOMEM;
+	}
 
 	return 0;
 }
 
-static int mes_v11_0_sw_init(void *handle)
+static int mes_v11_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	int pipe, r;
+	struct amdgpu_device *adev = ip_block->adev;
+	int pipe, r, bo_size;
 
 	adev->mes.funcs = &mes_v11_0_funcs;
 	adev->mes.kiq_hw_init = &mes_v11_0_kiq_hw_init;
 	adev->mes.kiq_hw_fini = &mes_v11_0_kiq_hw_fini;
 
+	adev->mes.event_log_size = AMDGPU_MES_LOG_BUFFER_SIZE + AMDGPU_MES_MSCRATCH_SIZE;
+
 	r = amdgpu_mes_init(adev);
 	if (r)
 		return r;
@@ -1054,16 +1456,33 @@ static int mes_v11_0_sw_init(void *handle)
 	if (r)
 		return r;
 
+	bo_size = AMDGPU_GPU_PAGE_SIZE;
+	if (amdgpu_sriov_is_mes_info_enable(adev))
+		bo_size += MES11_HW_RESOURCE_1_SIZE;
+
+	/* Only needed for AMDGPU_MES_SCHED_PIPE on MES 11*/
+	r = amdgpu_bo_create_kernel(adev,
+				    bo_size,
+				    PAGE_SIZE,
+				    AMDGPU_GEM_DOMAIN_VRAM,
+				    &adev->mes.resource_1[0],
+				    &adev->mes.resource_1_gpu_addr[0],
+				    &adev->mes.resource_1_addr[0]);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create mes resource_1 bo\n", r);
+		return r;
+	}
+
 	return 0;
 }
 
-static int mes_v11_0_sw_fini(void *handle)
+static int mes_v11_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int pipe;
 
-	amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs);
-	amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs);
+	amdgpu_bo_free_kernel(&adev->mes.resource_1[0], &adev->mes.resource_1_gpu_addr[0],
+			      &adev->mes.resource_1_addr[0]);
 
 	for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) {
 		kfree(adev->mes.mqd_backup[pipe]);
@@ -1074,16 +1493,16 @@ static int mes_v11_0_sw_fini(void *handle)
 		amdgpu_ucode_release(&adev->mes.fw[pipe]);
 	}
 
-	amdgpu_bo_free_kernel(&adev->gfx.kiq.ring.mqd_obj,
-			      &adev->gfx.kiq.ring.mqd_gpu_addr,
-			      &adev->gfx.kiq.ring.mqd_ptr);
+	amdgpu_bo_free_kernel(&adev->gfx.kiq[0].ring.mqd_obj,
+			      &adev->gfx.kiq[0].ring.mqd_gpu_addr,
+			      &adev->gfx.kiq[0].ring.mqd_ptr);
 
-	amdgpu_bo_free_kernel(&adev->mes.ring.mqd_obj,
-			      &adev->mes.ring.mqd_gpu_addr,
-			      &adev->mes.ring.mqd_ptr);
+	amdgpu_bo_free_kernel(&adev->mes.ring[0].mqd_obj,
+			      &adev->mes.ring[0].mqd_gpu_addr,
+			      &adev->mes.ring[0].mqd_ptr);
 
-	amdgpu_ring_fini(&adev->gfx.kiq.ring);
-	amdgpu_ring_fini(&adev->mes.ring);
+	amdgpu_ring_fini(&adev->gfx.kiq[0].ring);
+	amdgpu_ring_fini(&adev->mes.ring[0]);
 
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
 		mes_v11_0_free_ucode_buffers(adev, AMDGPU_MES_KIQ_PIPE);
@@ -1138,9 +1557,7 @@ static void mes_v11_0_kiq_setting(struct amdgpu_ring *ring)
 	tmp = RREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS);
 	tmp &= 0xffffff00;
 	tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
-	WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp);
-	tmp |= 0x80;
-	WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp);
+	WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp | 0x80);
 }
 
 static void mes_v11_0_kiq_clear(struct amdgpu_device *adev)
@@ -1156,6 +1573,7 @@ static void mes_v11_0_kiq_clear(struct amdgpu_device *adev)
 static int mes_v11_0_kiq_hw_init(struct amdgpu_device *adev)
 {
 	int r = 0;
+	struct amdgpu_ip_block *ip_block;
 
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
 
@@ -1175,28 +1593,47 @@ static int mes_v11_0_kiq_hw_init(struct amdgpu_device *adev)
 
 	mes_v11_0_enable(adev, true);
 
-	mes_v11_0_kiq_setting(&adev->gfx.kiq.ring);
+	mes_v11_0_get_fw_version(adev);
+
+	mes_v11_0_kiq_setting(&adev->gfx.kiq[0].ring);
+
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_MES);
+	if (unlikely(!ip_block)) {
+		dev_err(adev->dev, "Failed to get MES handle\n");
+		return -EINVAL;
+	}
 
 	r = mes_v11_0_queue_init(adev, AMDGPU_MES_KIQ_PIPE);
 	if (r)
 		goto failure;
 
+	if ((adev->mes.sched_version & AMDGPU_MES_VERSION_MASK) >= 0x47)
+		adev->mes.enable_legacy_queue_map = true;
+	else
+		adev->mes.enable_legacy_queue_map = false;
+
+	if (adev->mes.enable_legacy_queue_map) {
+		r = mes_v11_0_hw_init(ip_block);
+		if (r)
+			goto failure;
+	}
+
 	return r;
 
 failure:
-	mes_v11_0_hw_fini(adev);
+	mes_v11_0_hw_fini(ip_block);
 	return r;
 }
 
 static int mes_v11_0_kiq_hw_fini(struct amdgpu_device *adev)
 {
-	if (adev->mes.ring.sched.ready) {
-		mes_v11_0_kiq_dequeue(&adev->mes.ring);
-		adev->mes.ring.sched.ready = false;
+	if (adev->mes.ring[0].sched.ready) {
+		mes_v11_0_kiq_dequeue(&adev->mes.ring[0]);
+		adev->mes.ring[0].sched.ready = false;
 	}
 
 	if (amdgpu_sriov_vf(adev)) {
-		mes_v11_0_kiq_dequeue(&adev->gfx.kiq.ring);
+		mes_v11_0_kiq_dequeue(&adev->gfx.kiq[0].ring);
 		mes_v11_0_kiq_clear(adev);
 	}
 
@@ -1205,10 +1642,13 @@ static int mes_v11_0_kiq_hw_fini(struct amdgpu_device *adev)
 	return 0;
 }
 
-static int mes_v11_0_hw_init(void *handle)
+static int mes_v11_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (adev->mes.ring[0].sched.ready)
+		goto out;
 
 	if (!adev->enable_mes_kiq) {
 		if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
@@ -1231,7 +1671,13 @@ static int mes_v11_0_hw_init(void *handle)
 	if (r)
 		goto failure;
 
-	mes_v11_0_init_aggregated_doorbell(&adev->mes);
+	if ((adev->mes.sched_version & AMDGPU_MES_VERSION_MASK) >= 0x50) {
+		r = mes_v11_0_set_hw_resources_1(&adev->mes);
+		if (r) {
+			DRM_ERROR("failed mes_v11_0_set_hw_resources_1, r=%d\n", r);
+			goto failure;
+		}
+	}
 
 	r = mes_v11_0_query_sched_status(&adev->mes);
 	if (r) {
@@ -1239,55 +1685,49 @@ static int mes_v11_0_hw_init(void *handle)
 		goto failure;
 	}
 
+	r = amdgpu_mes_update_enforce_isolation(adev);
+	if (r)
+		goto failure;
+
+out:
 	/*
 	 * Disable KIQ ring usage from the driver once MES is enabled.
 	 * MES uses KIQ ring exclusively so driver cannot access KIQ ring
 	 * with MES enabled.
 	 */
-	adev->gfx.kiq.ring.sched.ready = false;
-	adev->mes.ring.sched.ready = true;
+	adev->gfx.kiq[0].ring.sched.ready = false;
+	adev->mes.ring[0].sched.ready = true;
 
 	return 0;
 
 failure:
-	mes_v11_0_hw_fini(adev);
+	mes_v11_0_hw_fini(ip_block);
 	return r;
 }
 
-static int mes_v11_0_hw_fini(void *handle)
+static int mes_v11_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
 	return 0;
 }
 
-static int mes_v11_0_suspend(void *handle)
+static int mes_v11_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	r = amdgpu_mes_suspend(adev);
-	if (r)
-		return r;
-
-	return mes_v11_0_hw_fini(adev);
+	return mes_v11_0_hw_fini(ip_block);
 }
 
-static int mes_v11_0_resume(void *handle)
+static int mes_v11_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	r = mes_v11_0_hw_init(adev);
-	if (r)
-		return r;
-
-	return amdgpu_mes_resume(adev);
+	return mes_v11_0_hw_init(ip_block);
 }
 
-static int mes_v11_0_early_init(void *handle)
+static int mes_v11_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int pipe, r;
 
+	adev->mes.hung_queue_db_array_size = MES11_HUNG_DB_OFFSET_ARRAY_SIZE;
+	adev->mes.hung_queue_hqd_info_offset = MES11_HUNG_HQD_INFO_OFFSET;
+
 	for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) {
 		if (!adev->enable_mes_kiq && pipe == AMDGPU_MES_KIQ_PIPE)
 			continue;
@@ -1299,22 +1739,10 @@ static int mes_v11_0_early_init(void *handle)
 	return 0;
 }
 
-static int mes_v11_0_late_init(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	/* it's only intended for use in mes_self_test case, not for s0ix and reset */
-	if (!amdgpu_in_reset(adev) && !adev->in_s0ix && !adev->in_suspend &&
-	    (adev->ip_versions[GC_HWIP][0] != IP_VERSION(11, 0, 3)))
-		amdgpu_mes_self_test(adev);
-
-	return 0;
-}
-
 static const struct amd_ip_funcs mes_v11_0_ip_funcs = {
 	.name = "mes_v11_0",
 	.early_init = mes_v11_0_early_init,
-	.late_init = mes_v11_0_late_init,
+	.late_init = NULL,
 	.sw_init = mes_v11_0_sw_init,
 	.sw_fini = mes_v11_0_sw_fini,
 	.hw_init = mes_v11_0_hw_init,
diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c b/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c
new file mode 100644
index 000000000000..744e95d3984a
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c
@@ -0,0 +1,1942 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/firmware.h>
+#include <linux/module.h>
+#include "amdgpu.h"
+#include "gfx_v12_0.h"
+#include "soc15_common.h"
+#include "soc21.h"
+#include "gc/gc_12_0_0_offset.h"
+#include "gc/gc_12_0_0_sh_mask.h"
+#include "gc/gc_11_0_0_default.h"
+#include "v12_structs.h"
+#include "mes_v12_api_def.h"
+
+MODULE_FIRMWARE("amdgpu/gc_12_0_0_mes.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_0_mes1.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_0_uni_mes.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_mes.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_mes1.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_uni_mes.bin");
+
+static int mes_v12_0_hw_init(struct amdgpu_ip_block *ip_block);
+static int mes_v12_0_hw_fini(struct amdgpu_ip_block *ip_block);
+static int mes_v12_0_kiq_hw_init(struct amdgpu_device *adev);
+static int mes_v12_0_kiq_hw_fini(struct amdgpu_device *adev);
+
+#define MES_EOP_SIZE   2048
+
+#define MES12_HUNG_DB_OFFSET_ARRAY_SIZE 8 /* [0:3] = db offset [4:7] hqd info */
+#define MES12_HUNG_HQD_INFO_OFFSET	4
+
+static void mes_v12_0_ring_set_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->use_doorbell) {
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
+			     ring->wptr);
+		WDOORBELL64(ring->doorbell_index, ring->wptr);
+	} else {
+		BUG();
+	}
+}
+
+static u64 mes_v12_0_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	return *ring->rptr_cpu_addr;
+}
+
+static u64 mes_v12_0_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	u64 wptr;
+
+	if (ring->use_doorbell)
+		wptr = atomic64_read((atomic64_t *)ring->wptr_cpu_addr);
+	else
+		BUG();
+	return wptr;
+}
+
+static const struct amdgpu_ring_funcs mes_v12_0_ring_funcs = {
+	.type = AMDGPU_RING_TYPE_MES,
+	.align_mask = 1,
+	.nop = 0,
+	.support_64bit_ptrs = true,
+	.get_rptr = mes_v12_0_ring_get_rptr,
+	.get_wptr = mes_v12_0_ring_get_wptr,
+	.set_wptr = mes_v12_0_ring_set_wptr,
+	.insert_nop = amdgpu_ring_insert_nop,
+};
+
+static const char *mes_v12_0_opcodes[] = {
+	"SET_HW_RSRC",
+	"SET_SCHEDULING_CONFIG",
+	"ADD_QUEUE",
+	"REMOVE_QUEUE",
+	"PERFORM_YIELD",
+	"SET_GANG_PRIORITY_LEVEL",
+	"SUSPEND",
+	"RESUME",
+	"RESET",
+	"SET_LOG_BUFFER",
+	"CHANGE_GANG_PRORITY",
+	"QUERY_SCHEDULER_STATUS",
+	"unused",
+	"SET_DEBUG_VMID",
+	"MISC",
+	"UPDATE_ROOT_PAGE_TABLE",
+	"AMD_LOG",
+	"SET_SE_MODE",
+	"SET_GANG_SUBMIT",
+	"SET_HW_RSRC_1",
+	"INVALIDATE_TLBS",
+};
+
+static const char *mes_v12_0_misc_opcodes[] = {
+	"WRITE_REG",
+	"INV_GART",
+	"QUERY_STATUS",
+	"READ_REG",
+	"WAIT_REG_MEM",
+	"SET_SHADER_DEBUGGER",
+	"NOTIFY_WORK_ON_UNMAPPED_QUEUE",
+	"NOTIFY_TO_UNMAP_PROCESSES",
+};
+
+static const char *mes_v12_0_get_op_string(union MESAPI__MISC *x_pkt)
+{
+	const char *op_str = NULL;
+
+	if (x_pkt->header.opcode < ARRAY_SIZE(mes_v12_0_opcodes))
+		op_str = mes_v12_0_opcodes[x_pkt->header.opcode];
+
+	return op_str;
+}
+
+static const char *mes_v12_0_get_misc_op_string(union MESAPI__MISC *x_pkt)
+{
+	const char *op_str = NULL;
+
+	if ((x_pkt->header.opcode == MES_SCH_API_MISC) &&
+	    (x_pkt->opcode < ARRAY_SIZE(mes_v12_0_misc_opcodes)))
+		op_str = mes_v12_0_misc_opcodes[x_pkt->opcode];
+
+	return op_str;
+}
+
+static int mes_v12_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes,
+					    int pipe, void *pkt, int size,
+					    int api_status_off)
+{
+	union MESAPI__QUERY_MES_STATUS mes_status_pkt;
+	signed long timeout = 2100000; /* 2100 ms */
+	struct amdgpu_device *adev = mes->adev;
+	struct amdgpu_ring *ring = &mes->ring[pipe];
+	spinlock_t *ring_lock = &mes->ring_lock[pipe];
+	struct MES_API_STATUS *api_status;
+	union MESAPI__MISC *x_pkt = pkt;
+	const char *op_str, *misc_op_str;
+	unsigned long flags;
+	u64 status_gpu_addr;
+	u32 seq, status_offset;
+	u64 *status_ptr;
+	signed long r;
+	int ret;
+
+	if (x_pkt->header.opcode >= MES_SCH_API_MAX)
+		return -EINVAL;
+
+	if (amdgpu_emu_mode) {
+		timeout *= 100;
+	} else if (amdgpu_sriov_vf(adev)) {
+		/* Worst case in sriov where all other 15 VF timeout, each VF needs about 600ms */
+		timeout = 15 * 600 * 1000;
+	}
+
+	ret = amdgpu_device_wb_get(adev, &status_offset);
+	if (ret)
+		return ret;
+
+	status_gpu_addr = adev->wb.gpu_addr + (status_offset * 4);
+	status_ptr = (u64 *)&adev->wb.wb[status_offset];
+	*status_ptr = 0;
+
+	spin_lock_irqsave(ring_lock, flags);
+	r = amdgpu_ring_alloc(ring, (size + sizeof(mes_status_pkt)) / 4);
+	if (r)
+		goto error_unlock_free;
+
+	seq = ++ring->fence_drv.sync_seq;
+	r = amdgpu_fence_wait_polling(ring,
+				      seq - ring->fence_drv.num_fences_mask,
+				      timeout);
+	if (r < 1)
+		goto error_undo;
+
+	api_status = (struct MES_API_STATUS *)((char *)pkt + api_status_off);
+	api_status->api_completion_fence_addr = status_gpu_addr;
+	api_status->api_completion_fence_value = 1;
+
+	amdgpu_ring_write_multiple(ring, pkt, size / 4);
+
+	memset(&mes_status_pkt, 0, sizeof(mes_status_pkt));
+	mes_status_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_status_pkt.header.opcode = MES_SCH_API_QUERY_SCHEDULER_STATUS;
+	mes_status_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+	mes_status_pkt.api_status.api_completion_fence_addr =
+		ring->fence_drv.gpu_addr;
+	mes_status_pkt.api_status.api_completion_fence_value = seq;
+
+	amdgpu_ring_write_multiple(ring, &mes_status_pkt,
+				   sizeof(mes_status_pkt) / 4);
+
+	amdgpu_ring_commit(ring);
+	spin_unlock_irqrestore(ring_lock, flags);
+
+	op_str = mes_v12_0_get_op_string(x_pkt);
+	misc_op_str = mes_v12_0_get_misc_op_string(x_pkt);
+
+	if (misc_op_str)
+		dev_dbg(adev->dev, "MES(%d) msg=%s (%s) was emitted\n",
+			pipe, op_str, misc_op_str);
+	else if (op_str)
+		dev_dbg(adev->dev, "MES(%d) msg=%s was emitted\n",
+			pipe, op_str);
+	else
+		dev_dbg(adev->dev, "MES(%d) msg=%d was emitted\n",
+			pipe, x_pkt->header.opcode);
+
+	r = amdgpu_fence_wait_polling(ring, seq, timeout);
+
+	/*
+	 * status_ptr[31:0] == 0 (fail) or status_ptr[63:0] == 1 (success).
+	 * If status_ptr[31:0] == 0 then status_ptr[63:32] will have debug error information.
+	 */
+	if (r < 1 || !(lower_32_bits(*status_ptr))) {
+
+		if (misc_op_str)
+			dev_err(adev->dev, "MES(%d) failed to respond to msg=%s (%s)\n",
+				pipe, op_str, misc_op_str);
+		else if (op_str)
+			dev_err(adev->dev, "MES(%d) failed to respond to msg=%s\n",
+				pipe, op_str);
+		else
+			dev_err(adev->dev, "MES(%d) failed to respond to msg=%d\n",
+				pipe, x_pkt->header.opcode);
+
+		while (halt_if_hws_hang)
+			schedule();
+
+		r = -ETIMEDOUT;
+		goto error_wb_free;
+	}
+
+	amdgpu_device_wb_free(adev, status_offset);
+	return 0;
+
+error_undo:
+	dev_err(adev->dev, "MES ring buffer is full.\n");
+	amdgpu_ring_undo(ring);
+
+error_unlock_free:
+	spin_unlock_irqrestore(ring_lock, flags);
+
+error_wb_free:
+	amdgpu_device_wb_free(adev, status_offset);
+	return r;
+}
+
+static int convert_to_mes_queue_type(int queue_type)
+{
+	if (queue_type == AMDGPU_RING_TYPE_GFX)
+		return MES_QUEUE_TYPE_GFX;
+	else if (queue_type == AMDGPU_RING_TYPE_COMPUTE)
+		return MES_QUEUE_TYPE_COMPUTE;
+	else if (queue_type == AMDGPU_RING_TYPE_SDMA)
+		return MES_QUEUE_TYPE_SDMA;
+	else if (queue_type == AMDGPU_RING_TYPE_MES)
+		return MES_QUEUE_TYPE_SCHQ;
+	else
+		BUG();
+	return -1;
+}
+
+static int convert_to_mes_priority_level(int priority_level)
+{
+	switch (priority_level) {
+	case AMDGPU_MES_PRIORITY_LEVEL_LOW:
+		return AMD_PRIORITY_LEVEL_LOW;
+	case AMDGPU_MES_PRIORITY_LEVEL_NORMAL:
+	default:
+		return AMD_PRIORITY_LEVEL_NORMAL;
+	case AMDGPU_MES_PRIORITY_LEVEL_MEDIUM:
+		return AMD_PRIORITY_LEVEL_MEDIUM;
+	case AMDGPU_MES_PRIORITY_LEVEL_HIGH:
+		return AMD_PRIORITY_LEVEL_HIGH;
+	case AMDGPU_MES_PRIORITY_LEVEL_REALTIME:
+		return AMD_PRIORITY_LEVEL_REALTIME;
+	}
+}
+
+static int mes_v12_0_add_hw_queue(struct amdgpu_mes *mes,
+				  struct mes_add_queue_input *input)
+{
+	struct amdgpu_device *adev = mes->adev;
+	union MESAPI__ADD_QUEUE mes_add_queue_pkt;
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	uint32_t vm_cntx_cntl = hub->vm_cntx_cntl;
+
+	memset(&mes_add_queue_pkt, 0, sizeof(mes_add_queue_pkt));
+
+	mes_add_queue_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_add_queue_pkt.header.opcode = MES_SCH_API_ADD_QUEUE;
+	mes_add_queue_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	mes_add_queue_pkt.process_id = input->process_id;
+	mes_add_queue_pkt.page_table_base_addr = input->page_table_base_addr;
+	mes_add_queue_pkt.process_va_start = input->process_va_start;
+	mes_add_queue_pkt.process_va_end = input->process_va_end;
+	mes_add_queue_pkt.process_quantum = input->process_quantum;
+	mes_add_queue_pkt.process_context_addr = input->process_context_addr;
+	mes_add_queue_pkt.gang_quantum = input->gang_quantum;
+	mes_add_queue_pkt.gang_context_addr = input->gang_context_addr;
+	mes_add_queue_pkt.inprocess_gang_priority =
+		convert_to_mes_priority_level(input->inprocess_gang_priority);
+	mes_add_queue_pkt.gang_global_priority_level =
+		convert_to_mes_priority_level(input->gang_global_priority_level);
+	mes_add_queue_pkt.doorbell_offset = input->doorbell_offset;
+	mes_add_queue_pkt.mqd_addr = input->mqd_addr;
+
+	mes_add_queue_pkt.wptr_addr = input->wptr_mc_addr;
+
+	mes_add_queue_pkt.queue_type =
+		convert_to_mes_queue_type(input->queue_type);
+	mes_add_queue_pkt.paging = input->paging;
+	mes_add_queue_pkt.vm_context_cntl = vm_cntx_cntl;
+	mes_add_queue_pkt.gws_base = input->gws_base;
+	mes_add_queue_pkt.gws_size = input->gws_size;
+	mes_add_queue_pkt.trap_handler_addr = input->tba_addr;
+	mes_add_queue_pkt.tma_addr = input->tma_addr;
+	mes_add_queue_pkt.trap_en = input->trap_en;
+	mes_add_queue_pkt.skip_process_ctx_clear = input->skip_process_ctx_clear;
+	mes_add_queue_pkt.is_kfd_process = input->is_kfd_process;
+
+	/* For KFD, gds_size is re-used for queue size (needed in MES for AQL queues) */
+	mes_add_queue_pkt.is_aql_queue = input->is_aql_queue;
+	mes_add_queue_pkt.gds_size = input->queue_size;
+
+	/* For KFD, gds_size is re-used for queue size (needed in MES for AQL queues) */
+	mes_add_queue_pkt.is_aql_queue = input->is_aql_queue;
+	mes_add_queue_pkt.gds_size = input->queue_size;
+
+	return mes_v12_0_submit_pkt_and_poll_completion(mes,
+			AMDGPU_MES_SCHED_PIPE,
+			&mes_add_queue_pkt, sizeof(mes_add_queue_pkt),
+			offsetof(union MESAPI__ADD_QUEUE, api_status));
+}
+
+static int mes_v12_0_remove_hw_queue(struct amdgpu_mes *mes,
+				     struct mes_remove_queue_input *input)
+{
+	union MESAPI__REMOVE_QUEUE mes_remove_queue_pkt;
+	uint32_t mes_rev = mes->sched_version & AMDGPU_MES_VERSION_MASK;
+
+	memset(&mes_remove_queue_pkt, 0, sizeof(mes_remove_queue_pkt));
+
+	mes_remove_queue_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_remove_queue_pkt.header.opcode = MES_SCH_API_REMOVE_QUEUE;
+	mes_remove_queue_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	mes_remove_queue_pkt.doorbell_offset = input->doorbell_offset;
+	mes_remove_queue_pkt.gang_context_addr = input->gang_context_addr;
+
+	if (mes_rev >= 0x5a)
+		mes_remove_queue_pkt.remove_queue_after_reset = input->remove_queue_after_reset;
+
+	return mes_v12_0_submit_pkt_and_poll_completion(mes,
+			AMDGPU_MES_SCHED_PIPE,
+			&mes_remove_queue_pkt, sizeof(mes_remove_queue_pkt),
+			offsetof(union MESAPI__REMOVE_QUEUE, api_status));
+}
+
+int gfx_v12_0_request_gfx_index_mutex(struct amdgpu_device *adev,
+				      bool req)
+{
+	u32 i, tmp, val;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		/* Request with MeId=2, PipeId=0 */
+		tmp = REG_SET_FIELD(0, CP_GFX_INDEX_MUTEX, REQUEST, req);
+		tmp = REG_SET_FIELD(tmp, CP_GFX_INDEX_MUTEX, CLIENTID, 4);
+		WREG32_SOC15(GC, 0, regCP_GFX_INDEX_MUTEX, tmp);
+
+		val = RREG32_SOC15(GC, 0, regCP_GFX_INDEX_MUTEX);
+		if (req) {
+			if (val == tmp)
+				break;
+		} else {
+			tmp = REG_SET_FIELD(tmp, CP_GFX_INDEX_MUTEX,
+					    REQUEST, 1);
+
+			/* unlocked or locked by firmware */
+			if (val != tmp)
+				break;
+		}
+		udelay(1);
+	}
+
+	if (i >= adev->usec_timeout)
+		return -EINVAL;
+
+	return 0;
+}
+
+static int mes_v12_0_reset_queue_mmio(struct amdgpu_mes *mes, uint32_t queue_type,
+				      uint32_t me_id, uint32_t pipe_id,
+				      uint32_t queue_id, uint32_t vmid)
+{
+	struct amdgpu_device *adev = mes->adev;
+	uint32_t value, reg;
+	int i, r = 0;
+
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
+
+	if (queue_type == AMDGPU_RING_TYPE_GFX) {
+		dev_info(adev->dev, "reset gfx queue (%d:%d:%d: vmid:%d)\n",
+			 me_id, pipe_id, queue_id, vmid);
+
+		mutex_lock(&adev->gfx.reset_sem_mutex);
+		gfx_v12_0_request_gfx_index_mutex(adev, true);
+		/* all se allow writes */
+		WREG32_SOC15(GC, 0, regGRBM_GFX_INDEX,
+			     (uint32_t)(0x1 << GRBM_GFX_INDEX__SE_BROADCAST_WRITES__SHIFT));
+		value = REG_SET_FIELD(0, CP_VMID_RESET, RESET_REQUEST, 1 << vmid);
+		if (pipe_id == 0)
+			value = REG_SET_FIELD(value, CP_VMID_RESET, PIPE0_QUEUES, 1 << queue_id);
+		else
+			value = REG_SET_FIELD(value, CP_VMID_RESET, PIPE1_QUEUES, 1 << queue_id);
+		WREG32_SOC15(GC, 0, regCP_VMID_RESET, value);
+		gfx_v12_0_request_gfx_index_mutex(adev, false);
+		mutex_unlock(&adev->gfx.reset_sem_mutex);
+
+		mutex_lock(&adev->srbm_mutex);
+		soc21_grbm_select(adev, me_id, pipe_id, queue_id, 0);
+		/* wait till dequeue take effects */
+		for (i = 0; i < adev->usec_timeout; i++) {
+			if (!(RREG32_SOC15(GC, 0, regCP_GFX_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+		if (i >= adev->usec_timeout) {
+			dev_err(adev->dev, "failed to wait on gfx hqd deactivate\n");
+			r = -ETIMEDOUT;
+		}
+
+		soc21_grbm_select(adev, 0, 0, 0, 0);
+		mutex_unlock(&adev->srbm_mutex);
+	} else if (queue_type == AMDGPU_RING_TYPE_COMPUTE) {
+		dev_info(adev->dev, "reset compute queue (%d:%d:%d)\n",
+			 me_id, pipe_id, queue_id);
+		mutex_lock(&adev->srbm_mutex);
+		soc21_grbm_select(adev, me_id, pipe_id, queue_id, 0);
+		WREG32_SOC15(GC, 0, regCP_HQD_DEQUEUE_REQUEST, 0x2);
+		WREG32_SOC15(GC, 0, regSPI_COMPUTE_QUEUE_RESET, 0x1);
+
+		/* wait till dequeue take effects */
+		for (i = 0; i < adev->usec_timeout; i++) {
+			if (!(RREG32_SOC15(GC, 0, regCP_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+		if (i >= adev->usec_timeout) {
+			dev_err(adev->dev, "failed to wait on hqd deactivate\n");
+			r = -ETIMEDOUT;
+		}
+		soc21_grbm_select(adev, 0, 0, 0, 0);
+		mutex_unlock(&adev->srbm_mutex);
+	} else if (queue_type == AMDGPU_RING_TYPE_SDMA) {
+		dev_info(adev->dev, "reset sdma queue (%d:%d:%d)\n",
+			 me_id, pipe_id, queue_id);
+		switch (me_id) {
+		case 1:
+			reg = SOC15_REG_OFFSET(GC, 0, regSDMA1_QUEUE_RESET_REQ);
+			break;
+		case 0:
+		default:
+			reg = SOC15_REG_OFFSET(GC, 0, regSDMA0_QUEUE_RESET_REQ);
+			break;
+		}
+
+		value = 1 << queue_id;
+		WREG32(reg, value);
+		/* wait for queue reset done */
+		for (i = 0; i < adev->usec_timeout; i++) {
+			if (!(RREG32(reg) & value))
+				break;
+			udelay(1);
+		}
+		if (i >= adev->usec_timeout) {
+			dev_err(adev->dev, "failed to wait on sdma queue reset done\n");
+			r = -ETIMEDOUT;
+		}
+	}
+
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
+	return r;
+}
+
+static int mes_v12_0_map_legacy_queue(struct amdgpu_mes *mes,
+				      struct mes_map_legacy_queue_input *input)
+{
+	union MESAPI__ADD_QUEUE mes_add_queue_pkt;
+	int pipe;
+
+	memset(&mes_add_queue_pkt, 0, sizeof(mes_add_queue_pkt));
+
+	mes_add_queue_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_add_queue_pkt.header.opcode = MES_SCH_API_ADD_QUEUE;
+	mes_add_queue_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	mes_add_queue_pkt.pipe_id = input->pipe_id;
+	mes_add_queue_pkt.queue_id = input->queue_id;
+	mes_add_queue_pkt.doorbell_offset = input->doorbell_offset;
+	mes_add_queue_pkt.mqd_addr = input->mqd_addr;
+	mes_add_queue_pkt.wptr_addr = input->wptr_addr;
+	mes_add_queue_pkt.queue_type =
+		convert_to_mes_queue_type(input->queue_type);
+	mes_add_queue_pkt.map_legacy_kq = 1;
+
+	if (mes->adev->enable_uni_mes)
+		pipe = AMDGPU_MES_KIQ_PIPE;
+	else
+		pipe = AMDGPU_MES_SCHED_PIPE;
+
+	return mes_v12_0_submit_pkt_and_poll_completion(mes, pipe,
+			&mes_add_queue_pkt, sizeof(mes_add_queue_pkt),
+			offsetof(union MESAPI__ADD_QUEUE, api_status));
+}
+
+static int mes_v12_0_unmap_legacy_queue(struct amdgpu_mes *mes,
+			struct mes_unmap_legacy_queue_input *input)
+{
+	union MESAPI__REMOVE_QUEUE mes_remove_queue_pkt;
+	int pipe;
+
+	memset(&mes_remove_queue_pkt, 0, sizeof(mes_remove_queue_pkt));
+
+	mes_remove_queue_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_remove_queue_pkt.header.opcode = MES_SCH_API_REMOVE_QUEUE;
+	mes_remove_queue_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	mes_remove_queue_pkt.doorbell_offset = input->doorbell_offset;
+	mes_remove_queue_pkt.gang_context_addr = 0;
+
+	mes_remove_queue_pkt.pipe_id = input->pipe_id;
+	mes_remove_queue_pkt.queue_id = input->queue_id;
+
+	if (input->action == PREEMPT_QUEUES_NO_UNMAP) {
+		mes_remove_queue_pkt.preempt_legacy_gfx_queue = 1;
+		mes_remove_queue_pkt.tf_addr = input->trail_fence_addr;
+		mes_remove_queue_pkt.tf_data =
+			lower_32_bits(input->trail_fence_data);
+	} else {
+		mes_remove_queue_pkt.unmap_legacy_queue = 1;
+		mes_remove_queue_pkt.queue_type =
+			convert_to_mes_queue_type(input->queue_type);
+	}
+
+	if (mes->adev->enable_uni_mes)
+		pipe = AMDGPU_MES_KIQ_PIPE;
+	else
+		pipe = AMDGPU_MES_SCHED_PIPE;
+
+	return mes_v12_0_submit_pkt_and_poll_completion(mes, pipe,
+			&mes_remove_queue_pkt, sizeof(mes_remove_queue_pkt),
+			offsetof(union MESAPI__REMOVE_QUEUE, api_status));
+}
+
+static int mes_v12_0_suspend_gang(struct amdgpu_mes *mes,
+				  struct mes_suspend_gang_input *input)
+{
+	union MESAPI__SUSPEND mes_suspend_gang_pkt;
+
+	memset(&mes_suspend_gang_pkt, 0, sizeof(mes_suspend_gang_pkt));
+
+	mes_suspend_gang_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_suspend_gang_pkt.header.opcode = MES_SCH_API_SUSPEND;
+	mes_suspend_gang_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	mes_suspend_gang_pkt.suspend_all_gangs = input->suspend_all_gangs;
+	mes_suspend_gang_pkt.gang_context_addr = input->gang_context_addr;
+	mes_suspend_gang_pkt.suspend_fence_addr = input->suspend_fence_addr;
+	mes_suspend_gang_pkt.suspend_fence_value = input->suspend_fence_value;
+
+	return mes_v12_0_submit_pkt_and_poll_completion(mes, AMDGPU_MES_SCHED_PIPE,
+			&mes_suspend_gang_pkt, sizeof(mes_suspend_gang_pkt),
+			offsetof(union MESAPI__SUSPEND, api_status));
+}
+
+static int mes_v12_0_resume_gang(struct amdgpu_mes *mes,
+				 struct mes_resume_gang_input *input)
+{
+	union MESAPI__RESUME mes_resume_gang_pkt;
+
+	memset(&mes_resume_gang_pkt, 0, sizeof(mes_resume_gang_pkt));
+
+	mes_resume_gang_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_resume_gang_pkt.header.opcode = MES_SCH_API_RESUME;
+	mes_resume_gang_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	mes_resume_gang_pkt.resume_all_gangs = input->resume_all_gangs;
+	mes_resume_gang_pkt.gang_context_addr = input->gang_context_addr;
+
+	return mes_v12_0_submit_pkt_and_poll_completion(mes, AMDGPU_MES_SCHED_PIPE,
+			&mes_resume_gang_pkt, sizeof(mes_resume_gang_pkt),
+			offsetof(union MESAPI__RESUME, api_status));
+}
+
+static int mes_v12_0_query_sched_status(struct amdgpu_mes *mes, int pipe)
+{
+	union MESAPI__QUERY_MES_STATUS mes_status_pkt;
+
+	memset(&mes_status_pkt, 0, sizeof(mes_status_pkt));
+
+	mes_status_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_status_pkt.header.opcode = MES_SCH_API_QUERY_SCHEDULER_STATUS;
+	mes_status_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	return mes_v12_0_submit_pkt_and_poll_completion(mes, pipe,
+			&mes_status_pkt, sizeof(mes_status_pkt),
+			offsetof(union MESAPI__QUERY_MES_STATUS, api_status));
+}
+
+static int mes_v12_0_misc_op(struct amdgpu_mes *mes,
+			     struct mes_misc_op_input *input)
+{
+	union MESAPI__MISC misc_pkt;
+	int pipe;
+
+	if (mes->adev->enable_uni_mes)
+		pipe = AMDGPU_MES_KIQ_PIPE;
+	else
+		pipe = AMDGPU_MES_SCHED_PIPE;
+
+	memset(&misc_pkt, 0, sizeof(misc_pkt));
+
+	misc_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	misc_pkt.header.opcode = MES_SCH_API_MISC;
+	misc_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	switch (input->op) {
+	case MES_MISC_OP_READ_REG:
+		misc_pkt.opcode = MESAPI_MISC__READ_REG;
+		misc_pkt.read_reg.reg_offset = input->read_reg.reg_offset;
+		misc_pkt.read_reg.buffer_addr = input->read_reg.buffer_addr;
+		break;
+	case MES_MISC_OP_WRITE_REG:
+		misc_pkt.opcode = MESAPI_MISC__WRITE_REG;
+		misc_pkt.write_reg.reg_offset = input->write_reg.reg_offset;
+		misc_pkt.write_reg.reg_value = input->write_reg.reg_value;
+		break;
+	case MES_MISC_OP_WRM_REG_WAIT:
+		misc_pkt.opcode = MESAPI_MISC__WAIT_REG_MEM;
+		misc_pkt.wait_reg_mem.op = WRM_OPERATION__WAIT_REG_MEM;
+		misc_pkt.wait_reg_mem.reference = input->wrm_reg.ref;
+		misc_pkt.wait_reg_mem.mask = input->wrm_reg.mask;
+		misc_pkt.wait_reg_mem.reg_offset1 = input->wrm_reg.reg0;
+		misc_pkt.wait_reg_mem.reg_offset2 = 0;
+		break;
+	case MES_MISC_OP_WRM_REG_WR_WAIT:
+		misc_pkt.opcode = MESAPI_MISC__WAIT_REG_MEM;
+		misc_pkt.wait_reg_mem.op = WRM_OPERATION__WR_WAIT_WR_REG;
+		misc_pkt.wait_reg_mem.reference = input->wrm_reg.ref;
+		misc_pkt.wait_reg_mem.mask = input->wrm_reg.mask;
+		misc_pkt.wait_reg_mem.reg_offset1 = input->wrm_reg.reg0;
+		misc_pkt.wait_reg_mem.reg_offset2 = input->wrm_reg.reg1;
+		break;
+	case MES_MISC_OP_SET_SHADER_DEBUGGER:
+		pipe = AMDGPU_MES_SCHED_PIPE;
+		misc_pkt.opcode = MESAPI_MISC__SET_SHADER_DEBUGGER;
+		misc_pkt.set_shader_debugger.process_context_addr =
+				input->set_shader_debugger.process_context_addr;
+		misc_pkt.set_shader_debugger.flags.u32all =
+				input->set_shader_debugger.flags.u32all;
+		misc_pkt.set_shader_debugger.spi_gdbg_per_vmid_cntl =
+				input->set_shader_debugger.spi_gdbg_per_vmid_cntl;
+		memcpy(misc_pkt.set_shader_debugger.tcp_watch_cntl,
+				input->set_shader_debugger.tcp_watch_cntl,
+				sizeof(misc_pkt.set_shader_debugger.tcp_watch_cntl));
+		misc_pkt.set_shader_debugger.trap_en = input->set_shader_debugger.trap_en;
+		break;
+	case MES_MISC_OP_CHANGE_CONFIG:
+		misc_pkt.opcode = MESAPI_MISC__CHANGE_CONFIG;
+		misc_pkt.change_config.opcode =
+				MESAPI_MISC__CHANGE_CONFIG_OPTION_LIMIT_SINGLE_PROCESS;
+		misc_pkt.change_config.option.bits.limit_single_process =
+				input->change_config.option.limit_single_process;
+		break;
+
+	default:
+		DRM_ERROR("unsupported misc op (%d) \n", input->op);
+		return -EINVAL;
+	}
+
+	return mes_v12_0_submit_pkt_and_poll_completion(mes, pipe,
+			&misc_pkt, sizeof(misc_pkt),
+			offsetof(union MESAPI__MISC, api_status));
+}
+
+static int mes_v12_0_set_hw_resources_1(struct amdgpu_mes *mes, int pipe)
+{
+	union MESAPI_SET_HW_RESOURCES_1 mes_set_hw_res_1_pkt;
+
+	memset(&mes_set_hw_res_1_pkt, 0, sizeof(mes_set_hw_res_1_pkt));
+
+	mes_set_hw_res_1_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_set_hw_res_1_pkt.header.opcode = MES_SCH_API_SET_HW_RSRC_1;
+	mes_set_hw_res_1_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+	mes_set_hw_res_1_pkt.mes_kiq_unmap_timeout = 0xa;
+	mes_set_hw_res_1_pkt.cleaner_shader_fence_mc_addr =
+		mes->resource_1_gpu_addr[pipe];
+
+	return mes_v12_0_submit_pkt_and_poll_completion(mes, pipe,
+			&mes_set_hw_res_1_pkt, sizeof(mes_set_hw_res_1_pkt),
+			offsetof(union MESAPI_SET_HW_RESOURCES_1, api_status));
+}
+
+static int mes_v12_0_set_hw_resources(struct amdgpu_mes *mes, int pipe)
+{
+	int i;
+	struct amdgpu_device *adev = mes->adev;
+	union MESAPI_SET_HW_RESOURCES mes_set_hw_res_pkt;
+
+	memset(&mes_set_hw_res_pkt, 0, sizeof(mes_set_hw_res_pkt));
+
+	mes_set_hw_res_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_set_hw_res_pkt.header.opcode = MES_SCH_API_SET_HW_RSRC;
+	mes_set_hw_res_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	if (pipe == AMDGPU_MES_SCHED_PIPE) {
+		mes_set_hw_res_pkt.vmid_mask_mmhub = mes->vmid_mask_mmhub;
+		mes_set_hw_res_pkt.vmid_mask_gfxhub = mes->vmid_mask_gfxhub;
+		mes_set_hw_res_pkt.gds_size = adev->gds.gds_size;
+		mes_set_hw_res_pkt.paging_vmid = 0;
+
+		for (i = 0; i < MAX_COMPUTE_PIPES; i++)
+			mes_set_hw_res_pkt.compute_hqd_mask[i] =
+				mes->compute_hqd_mask[i];
+
+		for (i = 0; i < MAX_GFX_PIPES; i++)
+			mes_set_hw_res_pkt.gfx_hqd_mask[i] =
+				mes->gfx_hqd_mask[i];
+
+		for (i = 0; i < MAX_SDMA_PIPES; i++)
+			mes_set_hw_res_pkt.sdma_hqd_mask[i] =
+				mes->sdma_hqd_mask[i];
+
+		for (i = 0; i < AMD_PRIORITY_NUM_LEVELS; i++)
+			mes_set_hw_res_pkt.aggregated_doorbells[i] =
+				mes->aggregated_doorbells[i];
+	}
+
+	mes_set_hw_res_pkt.g_sch_ctx_gpu_mc_ptr =
+		mes->sch_ctx_gpu_addr[pipe];
+	mes_set_hw_res_pkt.query_status_fence_gpu_mc_ptr =
+		mes->query_status_fence_gpu_addr[pipe];
+
+	for (i = 0; i < 5; i++) {
+		mes_set_hw_res_pkt.gc_base[i] = adev->reg_offset[GC_HWIP][0][i];
+		mes_set_hw_res_pkt.mmhub_base[i] =
+				adev->reg_offset[MMHUB_HWIP][0][i];
+		mes_set_hw_res_pkt.osssys_base[i] =
+		adev->reg_offset[OSSSYS_HWIP][0][i];
+	}
+
+	mes_set_hw_res_pkt.disable_reset = 1;
+	mes_set_hw_res_pkt.disable_mes_log = 1;
+	mes_set_hw_res_pkt.use_different_vmid_compute = 1;
+	mes_set_hw_res_pkt.enable_reg_active_poll = 1;
+	mes_set_hw_res_pkt.enable_level_process_quantum_check = 1;
+	if ((mes->adev->mes.sched_version & AMDGPU_MES_VERSION_MASK) >= 0x82)
+		mes_set_hw_res_pkt.enable_lr_compute_wa = 1;
+	else
+		dev_info_once(adev->dev,
+			      "MES FW version must be >= 0x82 to enable LR compute workaround.\n");
+
+	/*
+	 * Keep oversubscribe timer for sdma . When we have unmapped doorbell
+	 * handling support, other queue will not use the oversubscribe timer.
+	 * handling  mode - 0: disabled; 1: basic version; 2: basic+ version
+	 */
+	mes_set_hw_res_pkt.oversubscription_timer = 50;
+	mes_set_hw_res_pkt.unmapped_doorbell_handling = 1;
+
+	if (amdgpu_mes_log_enable) {
+		mes_set_hw_res_pkt.enable_mes_event_int_logging = 1;
+		mes_set_hw_res_pkt.event_intr_history_gpu_mc_ptr = mes->event_log_gpu_addr +
+				pipe * (AMDGPU_MES_LOG_BUFFER_SIZE + AMDGPU_MES_MSCRATCH_SIZE);
+	}
+
+	if (adev->enforce_isolation[0] == AMDGPU_ENFORCE_ISOLATION_ENABLE)
+		mes_set_hw_res_pkt.limit_single_process = 1;
+
+	return mes_v12_0_submit_pkt_and_poll_completion(mes, pipe,
+			&mes_set_hw_res_pkt, sizeof(mes_set_hw_res_pkt),
+			offsetof(union MESAPI_SET_HW_RESOURCES, api_status));
+}
+
+static void mes_v12_0_init_aggregated_doorbell(struct amdgpu_mes *mes)
+{
+	struct amdgpu_device *adev = mes->adev;
+	uint32_t data;
+
+	data = RREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL1);
+	data &= ~(CP_MES_DOORBELL_CONTROL1__DOORBELL_OFFSET_MASK |
+		  CP_MES_DOORBELL_CONTROL1__DOORBELL_EN_MASK |
+		  CP_MES_DOORBELL_CONTROL1__DOORBELL_HIT_MASK);
+	data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_LOW] <<
+		CP_MES_DOORBELL_CONTROL1__DOORBELL_OFFSET__SHIFT;
+	data |= 1 << CP_MES_DOORBELL_CONTROL1__DOORBELL_EN__SHIFT;
+	WREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL1, data);
+
+	data = RREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL2);
+	data &= ~(CP_MES_DOORBELL_CONTROL2__DOORBELL_OFFSET_MASK |
+		  CP_MES_DOORBELL_CONTROL2__DOORBELL_EN_MASK |
+		  CP_MES_DOORBELL_CONTROL2__DOORBELL_HIT_MASK);
+	data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_NORMAL] <<
+		CP_MES_DOORBELL_CONTROL2__DOORBELL_OFFSET__SHIFT;
+	data |= 1 << CP_MES_DOORBELL_CONTROL2__DOORBELL_EN__SHIFT;
+	WREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL2, data);
+
+	data = RREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL3);
+	data &= ~(CP_MES_DOORBELL_CONTROL3__DOORBELL_OFFSET_MASK |
+		  CP_MES_DOORBELL_CONTROL3__DOORBELL_EN_MASK |
+		  CP_MES_DOORBELL_CONTROL3__DOORBELL_HIT_MASK);
+	data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_MEDIUM] <<
+		CP_MES_DOORBELL_CONTROL3__DOORBELL_OFFSET__SHIFT;
+	data |= 1 << CP_MES_DOORBELL_CONTROL3__DOORBELL_EN__SHIFT;
+	WREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL3, data);
+
+	data = RREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL4);
+	data &= ~(CP_MES_DOORBELL_CONTROL4__DOORBELL_OFFSET_MASK |
+		  CP_MES_DOORBELL_CONTROL4__DOORBELL_EN_MASK |
+		  CP_MES_DOORBELL_CONTROL4__DOORBELL_HIT_MASK);
+	data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_HIGH] <<
+		CP_MES_DOORBELL_CONTROL4__DOORBELL_OFFSET__SHIFT;
+	data |= 1 << CP_MES_DOORBELL_CONTROL4__DOORBELL_EN__SHIFT;
+	WREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL4, data);
+
+	data = RREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL5);
+	data &= ~(CP_MES_DOORBELL_CONTROL5__DOORBELL_OFFSET_MASK |
+		  CP_MES_DOORBELL_CONTROL5__DOORBELL_EN_MASK |
+		  CP_MES_DOORBELL_CONTROL5__DOORBELL_HIT_MASK);
+	data |= mes->aggregated_doorbells[AMDGPU_MES_PRIORITY_LEVEL_REALTIME] <<
+		CP_MES_DOORBELL_CONTROL5__DOORBELL_OFFSET__SHIFT;
+	data |= 1 << CP_MES_DOORBELL_CONTROL5__DOORBELL_EN__SHIFT;
+	WREG32_SOC15(GC, 0, regCP_MES_DOORBELL_CONTROL5, data);
+
+	data = 1 << CP_HQD_GFX_CONTROL__DB_UPDATED_MSG_EN__SHIFT;
+	WREG32_SOC15(GC, 0, regCP_HQD_GFX_CONTROL, data);
+}
+
+
+static void mes_v12_0_enable_unmapped_doorbell_handling(
+		struct amdgpu_mes *mes, bool enable)
+{
+	struct amdgpu_device *adev = mes->adev;
+	uint32_t data = RREG32_SOC15(GC, 0, regCP_UNMAPPED_DOORBELL);
+
+	/*
+	 * The default PROC_LSB settng is 0xc which means doorbell
+	 * addr[16:12] gives the doorbell page number. For kfd, each
+	 * process will use 2 pages of doorbell, we need to change the
+	 * setting to 0xd
+	 */
+	data &= ~CP_UNMAPPED_DOORBELL__PROC_LSB_MASK;
+	data |= 0xd <<  CP_UNMAPPED_DOORBELL__PROC_LSB__SHIFT;
+
+	data |= (enable ? 1 : 0) << CP_UNMAPPED_DOORBELL__ENABLE__SHIFT;
+
+	WREG32_SOC15(GC, 0, regCP_UNMAPPED_DOORBELL, data);
+}
+
+static int mes_v12_0_reset_hw_queue(struct amdgpu_mes *mes,
+				    struct mes_reset_queue_input *input)
+{
+	union MESAPI__RESET mes_reset_queue_pkt;
+	int pipe;
+
+	if (input->use_mmio)
+		return mes_v12_0_reset_queue_mmio(mes, input->queue_type,
+						  input->me_id, input->pipe_id,
+						  input->queue_id, input->vmid);
+
+	memset(&mes_reset_queue_pkt, 0, sizeof(mes_reset_queue_pkt));
+
+	mes_reset_queue_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_reset_queue_pkt.header.opcode = MES_SCH_API_RESET;
+	mes_reset_queue_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	mes_reset_queue_pkt.queue_type =
+		convert_to_mes_queue_type(input->queue_type);
+
+	if (input->legacy_gfx) {
+		mes_reset_queue_pkt.reset_legacy_gfx = 1;
+		mes_reset_queue_pkt.pipe_id_lp = input->pipe_id;
+		mes_reset_queue_pkt.queue_id_lp = input->queue_id;
+		mes_reset_queue_pkt.mqd_mc_addr_lp = input->mqd_addr;
+		mes_reset_queue_pkt.doorbell_offset_lp = input->doorbell_offset;
+		mes_reset_queue_pkt.wptr_addr_lp = input->wptr_addr;
+		mes_reset_queue_pkt.vmid_id_lp = input->vmid;
+	} else {
+		mes_reset_queue_pkt.reset_queue_only = 1;
+		mes_reset_queue_pkt.doorbell_offset = input->doorbell_offset;
+	}
+
+	if (input->is_kq)
+		pipe = AMDGPU_MES_KIQ_PIPE;
+	else
+		pipe = AMDGPU_MES_SCHED_PIPE;
+
+	return mes_v12_0_submit_pkt_and_poll_completion(mes, pipe,
+			&mes_reset_queue_pkt, sizeof(mes_reset_queue_pkt),
+			offsetof(union MESAPI__RESET, api_status));
+}
+
+static int mes_v12_0_detect_and_reset_hung_queues(struct amdgpu_mes *mes,
+						  struct mes_detect_and_reset_queue_input *input)
+{
+	union MESAPI__RESET mes_reset_queue_pkt;
+
+	memset(&mes_reset_queue_pkt, 0, sizeof(mes_reset_queue_pkt));
+
+	mes_reset_queue_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_reset_queue_pkt.header.opcode = MES_SCH_API_RESET;
+	mes_reset_queue_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	mes_reset_queue_pkt.queue_type =
+		convert_to_mes_queue_type(input->queue_type);
+	mes_reset_queue_pkt.doorbell_offset_addr =
+		mes->hung_queue_db_array_gpu_addr;
+
+	if (input->detect_only)
+		mes_reset_queue_pkt.hang_detect_only = 1;
+	else
+		mes_reset_queue_pkt.hang_detect_then_reset = 1;
+
+	return mes_v12_0_submit_pkt_and_poll_completion(mes, AMDGPU_MES_SCHED_PIPE,
+			&mes_reset_queue_pkt, sizeof(mes_reset_queue_pkt),
+			offsetof(union MESAPI__RESET, api_status));
+}
+
+static int mes_v12_inv_tlb_convert_hub_id(uint8_t id)
+{
+	/*
+	 * MES doesn't support invalidate gc_hub on slave xcc individually
+	 * master xcc will invalidate all gc_hub for the partition
+	 */
+	if (AMDGPU_IS_GFXHUB(id))
+		return 0;
+	else if (AMDGPU_IS_MMHUB0(id))
+		return 1;
+	else
+		return -EINVAL;
+
+}
+
+static int mes_v12_0_inv_tlbs_pasid(struct amdgpu_mes *mes,
+				    struct mes_inv_tlbs_pasid_input *input)
+{
+	union MESAPI__INV_TLBS mes_inv_tlbs;
+	int ret;
+
+	memset(&mes_inv_tlbs, 0, sizeof(mes_inv_tlbs));
+
+	mes_inv_tlbs.header.type = MES_API_TYPE_SCHEDULER;
+	mes_inv_tlbs.header.opcode = MES_SCH_API_INV_TLBS;
+	mes_inv_tlbs.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	mes_inv_tlbs.invalidate_tlbs.inv_sel = 0;
+	mes_inv_tlbs.invalidate_tlbs.flush_type = input->flush_type;
+	mes_inv_tlbs.invalidate_tlbs.inv_sel_id = input->pasid;
+
+	/*convert amdgpu_mes_hub_id to mes expected hub_id */
+	ret = mes_v12_inv_tlb_convert_hub_id(input->hub_id);
+	if (ret < 0)
+		return -EINVAL;
+	mes_inv_tlbs.invalidate_tlbs.hub_id = ret;
+	return mes_v12_0_submit_pkt_and_poll_completion(mes, AMDGPU_MES_KIQ_PIPE,
+			&mes_inv_tlbs, sizeof(mes_inv_tlbs),
+			offsetof(union MESAPI__INV_TLBS, api_status));
+
+}
+
+static const struct amdgpu_mes_funcs mes_v12_0_funcs = {
+	.add_hw_queue = mes_v12_0_add_hw_queue,
+	.remove_hw_queue = mes_v12_0_remove_hw_queue,
+	.map_legacy_queue = mes_v12_0_map_legacy_queue,
+	.unmap_legacy_queue = mes_v12_0_unmap_legacy_queue,
+	.suspend_gang = mes_v12_0_suspend_gang,
+	.resume_gang = mes_v12_0_resume_gang,
+	.misc_op = mes_v12_0_misc_op,
+	.reset_hw_queue = mes_v12_0_reset_hw_queue,
+	.invalidate_tlbs_pasid = mes_v12_0_inv_tlbs_pasid,
+	.detect_and_reset_hung_queues = mes_v12_0_detect_and_reset_hung_queues,
+};
+
+static int mes_v12_0_allocate_ucode_buffer(struct amdgpu_device *adev,
+					   enum amdgpu_mes_pipe pipe)
+{
+	int r;
+	const struct mes_firmware_header_v1_0 *mes_hdr;
+	const __le32 *fw_data;
+	unsigned fw_size;
+
+	mes_hdr = (const struct mes_firmware_header_v1_0 *)
+		adev->mes.fw[pipe]->data;
+
+	fw_data = (const __le32 *)(adev->mes.fw[pipe]->data +
+		   le32_to_cpu(mes_hdr->mes_ucode_offset_bytes));
+	fw_size = le32_to_cpu(mes_hdr->mes_ucode_size_bytes);
+
+	r = amdgpu_bo_create_reserved(adev, fw_size,
+				      PAGE_SIZE,
+				      AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->mes.ucode_fw_obj[pipe],
+				      &adev->mes.ucode_fw_gpu_addr[pipe],
+				      (void **)&adev->mes.ucode_fw_ptr[pipe]);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create mes fw bo\n", r);
+		return r;
+	}
+
+	memcpy(adev->mes.ucode_fw_ptr[pipe], fw_data, fw_size);
+
+	amdgpu_bo_kunmap(adev->mes.ucode_fw_obj[pipe]);
+	amdgpu_bo_unreserve(adev->mes.ucode_fw_obj[pipe]);
+
+	return 0;
+}
+
+static int mes_v12_0_allocate_ucode_data_buffer(struct amdgpu_device *adev,
+						enum amdgpu_mes_pipe pipe)
+{
+	int r;
+	const struct mes_firmware_header_v1_0 *mes_hdr;
+	const __le32 *fw_data;
+	unsigned fw_size;
+
+	mes_hdr = (const struct mes_firmware_header_v1_0 *)
+		adev->mes.fw[pipe]->data;
+
+	fw_data = (const __le32 *)(adev->mes.fw[pipe]->data +
+		   le32_to_cpu(mes_hdr->mes_ucode_data_offset_bytes));
+	fw_size = le32_to_cpu(mes_hdr->mes_ucode_data_size_bytes);
+
+	r = amdgpu_bo_create_reserved(adev, fw_size,
+				      64 * 1024,
+				      AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->mes.data_fw_obj[pipe],
+				      &adev->mes.data_fw_gpu_addr[pipe],
+				      (void **)&adev->mes.data_fw_ptr[pipe]);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create mes data fw bo\n", r);
+		return r;
+	}
+
+	memcpy(adev->mes.data_fw_ptr[pipe], fw_data, fw_size);
+
+	amdgpu_bo_kunmap(adev->mes.data_fw_obj[pipe]);
+	amdgpu_bo_unreserve(adev->mes.data_fw_obj[pipe]);
+
+	return 0;
+}
+
+static void mes_v12_0_free_ucode_buffers(struct amdgpu_device *adev,
+					 enum amdgpu_mes_pipe pipe)
+{
+	amdgpu_bo_free_kernel(&adev->mes.data_fw_obj[pipe],
+			      &adev->mes.data_fw_gpu_addr[pipe],
+			      (void **)&adev->mes.data_fw_ptr[pipe]);
+
+	amdgpu_bo_free_kernel(&adev->mes.ucode_fw_obj[pipe],
+			      &adev->mes.ucode_fw_gpu_addr[pipe],
+			      (void **)&adev->mes.ucode_fw_ptr[pipe]);
+}
+
+static void mes_v12_0_enable(struct amdgpu_device *adev, bool enable)
+{
+	uint64_t ucode_addr;
+	uint32_t pipe, data = 0;
+
+	if (enable) {
+		mutex_lock(&adev->srbm_mutex);
+		for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) {
+			soc21_grbm_select(adev, 3, pipe, 0, 0);
+			if (amdgpu_mes_log_enable) {
+				u32 log_size = AMDGPU_MES_LOG_BUFFER_SIZE + AMDGPU_MES_MSCRATCH_SIZE;
+				/* In case uni mes is not enabled, only program for pipe 0 */
+				if (adev->mes.event_log_size >= (pipe + 1) * log_size) {
+					WREG32_SOC15(GC, 0, regCP_MES_MSCRATCH_LO,
+						     lower_32_bits(adev->mes.event_log_gpu_addr +
+						     pipe * log_size + AMDGPU_MES_LOG_BUFFER_SIZE));
+					WREG32_SOC15(GC, 0, regCP_MES_MSCRATCH_HI,
+						     upper_32_bits(adev->mes.event_log_gpu_addr +
+						     pipe * log_size + AMDGPU_MES_LOG_BUFFER_SIZE));
+					dev_info(adev->dev, "Setup CP MES MSCRATCH address : 0x%x. 0x%x\n",
+						 RREG32_SOC15(GC, 0, regCP_MES_MSCRATCH_HI),
+						 RREG32_SOC15(GC, 0, regCP_MES_MSCRATCH_LO));
+				}
+			}
+
+			data = RREG32_SOC15(GC, 0, regCP_MES_CNTL);
+			if (pipe == 0)
+				data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE0_RESET, 1);
+			else
+				data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_RESET, 1);
+			WREG32_SOC15(GC, 0, regCP_MES_CNTL, data);
+
+			ucode_addr = adev->mes.uc_start_addr[pipe] >> 2;
+			WREG32_SOC15(GC, 0, regCP_MES_PRGRM_CNTR_START,
+				     lower_32_bits(ucode_addr));
+			WREG32_SOC15(GC, 0, regCP_MES_PRGRM_CNTR_START_HI,
+				     upper_32_bits(ucode_addr));
+
+			/* unhalt MES and activate one pipe each loop */
+			data = REG_SET_FIELD(0, CP_MES_CNTL, MES_PIPE0_ACTIVE, 1);
+			if (pipe)
+				data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_ACTIVE, 1);
+			dev_info(adev->dev, "program CP_MES_CNTL : 0x%x\n", data);
+
+			WREG32_SOC15(GC, 0, regCP_MES_CNTL, data);
+
+		}
+		soc21_grbm_select(adev, 0, 0, 0, 0);
+		mutex_unlock(&adev->srbm_mutex);
+
+		if (amdgpu_emu_mode)
+			msleep(100);
+		else if (adev->enable_uni_mes)
+			udelay(500);
+		else
+			udelay(50);
+	} else {
+		data = RREG32_SOC15(GC, 0, regCP_MES_CNTL);
+		data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE0_ACTIVE, 0);
+		data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_ACTIVE, 0);
+		data = REG_SET_FIELD(data, CP_MES_CNTL,
+				     MES_INVALIDATE_ICACHE, 1);
+		data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE0_RESET, 1);
+		data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_RESET, 1);
+		data = REG_SET_FIELD(data, CP_MES_CNTL, MES_HALT, 1);
+		WREG32_SOC15(GC, 0, regCP_MES_CNTL, data);
+	}
+}
+
+static void mes_v12_0_set_ucode_start_addr(struct amdgpu_device *adev)
+{
+	uint64_t ucode_addr;
+	int pipe;
+
+	mes_v12_0_enable(adev, false);
+
+	mutex_lock(&adev->srbm_mutex);
+	for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) {
+		/* me=3, queue=0 */
+		soc21_grbm_select(adev, 3, pipe, 0, 0);
+
+		/* set ucode start address */
+		ucode_addr = adev->mes.uc_start_addr[pipe] >> 2;
+		WREG32_SOC15(GC, 0, regCP_MES_PRGRM_CNTR_START,
+				lower_32_bits(ucode_addr));
+		WREG32_SOC15(GC, 0, regCP_MES_PRGRM_CNTR_START_HI,
+				upper_32_bits(ucode_addr));
+
+		soc21_grbm_select(adev, 0, 0, 0, 0);
+	}
+	mutex_unlock(&adev->srbm_mutex);
+}
+
+/* This function is for backdoor MES firmware */
+static int mes_v12_0_load_microcode(struct amdgpu_device *adev,
+				    enum amdgpu_mes_pipe pipe, bool prime_icache)
+{
+	int r;
+	uint32_t data;
+
+	mes_v12_0_enable(adev, false);
+
+	if (!adev->mes.fw[pipe])
+		return -EINVAL;
+
+	r = mes_v12_0_allocate_ucode_buffer(adev, pipe);
+	if (r)
+		return r;
+
+	r = mes_v12_0_allocate_ucode_data_buffer(adev, pipe);
+	if (r) {
+		mes_v12_0_free_ucode_buffers(adev, pipe);
+		return r;
+	}
+
+	mutex_lock(&adev->srbm_mutex);
+	/* me=3, pipe=0, queue=0 */
+	soc21_grbm_select(adev, 3, pipe, 0, 0);
+
+	WREG32_SOC15(GC, 0, regCP_MES_IC_BASE_CNTL, 0);
+
+	/* set ucode fimrware address */
+	WREG32_SOC15(GC, 0, regCP_MES_IC_BASE_LO,
+		     lower_32_bits(adev->mes.ucode_fw_gpu_addr[pipe]));
+	WREG32_SOC15(GC, 0, regCP_MES_IC_BASE_HI,
+		     upper_32_bits(adev->mes.ucode_fw_gpu_addr[pipe]));
+
+	/* set ucode instruction cache boundary to 2M-1 */
+	WREG32_SOC15(GC, 0, regCP_MES_MIBOUND_LO, 0x1FFFFF);
+
+	/* set ucode data firmware address */
+	WREG32_SOC15(GC, 0, regCP_MES_MDBASE_LO,
+		     lower_32_bits(adev->mes.data_fw_gpu_addr[pipe]));
+	WREG32_SOC15(GC, 0, regCP_MES_MDBASE_HI,
+		     upper_32_bits(adev->mes.data_fw_gpu_addr[pipe]));
+
+	/* Set data cache boundary CP_MES_MDBOUND_LO */
+	WREG32_SOC15(GC, 0, regCP_MES_MDBOUND_LO, 0x7FFFF);
+
+	if (prime_icache) {
+		/* invalidate ICACHE */
+		data = RREG32_SOC15(GC, 0, regCP_MES_IC_OP_CNTL);
+		data = REG_SET_FIELD(data, CP_MES_IC_OP_CNTL, PRIME_ICACHE, 0);
+		data = REG_SET_FIELD(data, CP_MES_IC_OP_CNTL, INVALIDATE_CACHE, 1);
+		WREG32_SOC15(GC, 0, regCP_MES_IC_OP_CNTL, data);
+
+		/* prime the ICACHE. */
+		data = RREG32_SOC15(GC, 0, regCP_MES_IC_OP_CNTL);
+		data = REG_SET_FIELD(data, CP_MES_IC_OP_CNTL, PRIME_ICACHE, 1);
+		WREG32_SOC15(GC, 0, regCP_MES_IC_OP_CNTL, data);
+	}
+
+	soc21_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+
+	return 0;
+}
+
+static int mes_v12_0_allocate_eop_buf(struct amdgpu_device *adev,
+				      enum amdgpu_mes_pipe pipe)
+{
+	int r;
+	u32 *eop;
+
+	r = amdgpu_bo_create_reserved(adev, MES_EOP_SIZE, PAGE_SIZE,
+			      AMDGPU_GEM_DOMAIN_GTT,
+			      &adev->mes.eop_gpu_obj[pipe],
+			      &adev->mes.eop_gpu_addr[pipe],
+			      (void **)&eop);
+	if (r) {
+		dev_warn(adev->dev, "(%d) create EOP bo failed\n", r);
+		return r;
+	}
+
+	memset(eop, 0,
+	       adev->mes.eop_gpu_obj[pipe]->tbo.base.size);
+
+	amdgpu_bo_kunmap(adev->mes.eop_gpu_obj[pipe]);
+	amdgpu_bo_unreserve(adev->mes.eop_gpu_obj[pipe]);
+
+	return 0;
+}
+
+static int mes_v12_0_mqd_init(struct amdgpu_ring *ring)
+{
+	struct v12_compute_mqd *mqd = ring->mqd_ptr;
+	uint64_t hqd_gpu_addr, wb_gpu_addr, eop_base_addr;
+	uint32_t tmp;
+
+	mqd->header = 0xC0310800;
+	mqd->compute_pipelinestat_enable = 0x00000001;
+	mqd->compute_static_thread_mgmt_se0 = 0xffffffff;
+	mqd->compute_static_thread_mgmt_se1 = 0xffffffff;
+	mqd->compute_static_thread_mgmt_se2 = 0xffffffff;
+	mqd->compute_static_thread_mgmt_se3 = 0xffffffff;
+	mqd->compute_misc_reserved = 0x00000007;
+
+	eop_base_addr = ring->eop_gpu_addr >> 8;
+
+	/* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */
+	tmp = regCP_HQD_EOP_CONTROL_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_HQD_EOP_CONTROL, EOP_SIZE,
+			(order_base_2(MES_EOP_SIZE / 4) - 1));
+
+	mqd->cp_hqd_eop_base_addr_lo = lower_32_bits(eop_base_addr);
+	mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr);
+	mqd->cp_hqd_eop_control = tmp;
+
+	/* disable the queue if it's active */
+	ring->wptr = 0;
+	mqd->cp_hqd_pq_rptr = 0;
+	mqd->cp_hqd_pq_wptr_lo = 0;
+	mqd->cp_hqd_pq_wptr_hi = 0;
+
+	/* set the pointer to the MQD */
+	mqd->cp_mqd_base_addr_lo = ring->mqd_gpu_addr & 0xfffffffc;
+	mqd->cp_mqd_base_addr_hi = upper_32_bits(ring->mqd_gpu_addr);
+
+	/* set MQD vmid to 0 */
+	tmp = regCP_MQD_CONTROL_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_MQD_CONTROL, VMID, 0);
+	mqd->cp_mqd_control = tmp;
+
+	/* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */
+	hqd_gpu_addr = ring->gpu_addr >> 8;
+	mqd->cp_hqd_pq_base_lo = lower_32_bits(hqd_gpu_addr);
+	mqd->cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr);
+
+	/* set the wb address whether it's enabled or not */
+	wb_gpu_addr = ring->rptr_gpu_addr;
+	mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc;
+	mqd->cp_hqd_pq_rptr_report_addr_hi =
+		upper_32_bits(wb_gpu_addr) & 0xffff;
+
+	/* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */
+	wb_gpu_addr = ring->wptr_gpu_addr;
+	mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffff8;
+	mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff;
+
+	/* set up the HQD, this is similar to CP_RB0_CNTL */
+	tmp = regCP_HQD_PQ_CONTROL_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE,
+			    (order_base_2(ring->ring_size / 4) - 1));
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE,
+			    ((order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1) << 8));
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 1);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TUNNEL_DISPATCH, 0);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, PRIV_STATE, 1);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, KMD_QUEUE, 1);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, NO_UPDATE_RPTR, 1);
+	mqd->cp_hqd_pq_control = tmp;
+
+	/* enable doorbell */
+	tmp = 0;
+	if (ring->use_doorbell) {
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_OFFSET, ring->doorbell_index);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_EN, 1);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_SOURCE, 0);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_HIT, 0);
+	} else {
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_EN, 0);
+	}
+	mqd->cp_hqd_pq_doorbell_control = tmp;
+
+	mqd->cp_hqd_vmid = 0;
+	/* activate the queue */
+	mqd->cp_hqd_active = 1;
+
+	tmp = regCP_HQD_PERSISTENT_STATE_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PERSISTENT_STATE,
+			    PRELOAD_SIZE, 0x55);
+	mqd->cp_hqd_persistent_state = tmp;
+
+	mqd->cp_hqd_ib_control = regCP_HQD_IB_CONTROL_DEFAULT;
+	mqd->cp_hqd_iq_timer = regCP_HQD_IQ_TIMER_DEFAULT;
+	mqd->cp_hqd_quantum = regCP_HQD_QUANTUM_DEFAULT;
+
+	/*
+	 * Set CP_HQD_GFX_CONTROL.DB_UPDATED_MSG_EN[15] to enable unmapped
+	 * doorbell handling. This is a reserved CP internal register can
+	 * not be accesss by others
+	 */
+	mqd->reserved_184 = BIT(15);
+
+	return 0;
+}
+
+static void mes_v12_0_queue_init_register(struct amdgpu_ring *ring)
+{
+	struct v12_compute_mqd *mqd = ring->mqd_ptr;
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t data = 0;
+
+	mutex_lock(&adev->srbm_mutex);
+	soc21_grbm_select(adev, 3, ring->pipe, 0, 0);
+
+	/* set CP_HQD_VMID.VMID = 0. */
+	data = RREG32_SOC15(GC, 0, regCP_HQD_VMID);
+	data = REG_SET_FIELD(data, CP_HQD_VMID, VMID, 0);
+	WREG32_SOC15(GC, 0, regCP_HQD_VMID, data);
+
+	/* set CP_HQD_PQ_DOORBELL_CONTROL.DOORBELL_EN=0 */
+	data = RREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL);
+	data = REG_SET_FIELD(data, CP_HQD_PQ_DOORBELL_CONTROL,
+			     DOORBELL_EN, 0);
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL, data);
+
+	/* set CP_MQD_BASE_ADDR/HI with the MQD base address */
+	WREG32_SOC15(GC, 0, regCP_MQD_BASE_ADDR, mqd->cp_mqd_base_addr_lo);
+	WREG32_SOC15(GC, 0, regCP_MQD_BASE_ADDR_HI, mqd->cp_mqd_base_addr_hi);
+
+	/* set CP_MQD_CONTROL.VMID=0 */
+	data = RREG32_SOC15(GC, 0, regCP_MQD_CONTROL);
+	data = REG_SET_FIELD(data, CP_MQD_CONTROL, VMID, 0);
+	WREG32_SOC15(GC, 0, regCP_MQD_CONTROL, 0);
+
+	/* set CP_HQD_PQ_BASE/HI with the ring buffer base address */
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_BASE, mqd->cp_hqd_pq_base_lo);
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_BASE_HI, mqd->cp_hqd_pq_base_hi);
+
+	/* set CP_HQD_PQ_RPTR_REPORT_ADDR/HI */
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR_REPORT_ADDR,
+		     mqd->cp_hqd_pq_rptr_report_addr_lo);
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR_REPORT_ADDR_HI,
+		     mqd->cp_hqd_pq_rptr_report_addr_hi);
+
+	/* set CP_HQD_PQ_CONTROL */
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_CONTROL, mqd->cp_hqd_pq_control);
+
+	/* set CP_HQD_PQ_WPTR_POLL_ADDR/HI */
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR,
+		     mqd->cp_hqd_pq_wptr_poll_addr_lo);
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR_HI,
+		     mqd->cp_hqd_pq_wptr_poll_addr_hi);
+
+	/* set CP_HQD_PQ_DOORBELL_CONTROL */
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL,
+		     mqd->cp_hqd_pq_doorbell_control);
+
+	/* set CP_HQD_PERSISTENT_STATE.PRELOAD_SIZE=0x53 */
+	WREG32_SOC15(GC, 0, regCP_HQD_PERSISTENT_STATE, mqd->cp_hqd_persistent_state);
+
+	/* set CP_HQD_ACTIVE.ACTIVE=1 */
+	WREG32_SOC15(GC, 0, regCP_HQD_ACTIVE, mqd->cp_hqd_active);
+
+	soc21_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+}
+
+static int mes_v12_0_kiq_enable_queue(struct amdgpu_device *adev)
+{
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
+	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq[0].ring;
+	int r;
+
+	if (!kiq->pmf || !kiq->pmf->kiq_map_queues)
+		return -EINVAL;
+
+	r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size);
+	if (r) {
+		DRM_ERROR("Failed to lock KIQ (%d).\n", r);
+		return r;
+	}
+
+	kiq->pmf->kiq_map_queues(kiq_ring, &adev->mes.ring[0]);
+
+	r = amdgpu_ring_test_ring(kiq_ring);
+	if (r) {
+		DRM_ERROR("kfq enable failed\n");
+		kiq_ring->sched.ready = false;
+	}
+	return r;
+}
+
+static int mes_v12_0_queue_init(struct amdgpu_device *adev,
+				enum amdgpu_mes_pipe pipe)
+{
+	struct amdgpu_ring *ring;
+	int r;
+
+	if (!adev->enable_uni_mes && pipe == AMDGPU_MES_KIQ_PIPE)
+		ring = &adev->gfx.kiq[0].ring;
+	else
+		ring = &adev->mes.ring[pipe];
+
+	if ((adev->enable_uni_mes || pipe == AMDGPU_MES_SCHED_PIPE) &&
+	    (amdgpu_in_reset(adev) || adev->in_suspend)) {
+		*(ring->wptr_cpu_addr) = 0;
+		*(ring->rptr_cpu_addr) = 0;
+		amdgpu_ring_clear_ring(ring);
+	}
+
+	r = mes_v12_0_mqd_init(ring);
+	if (r)
+		return r;
+
+	if (pipe == AMDGPU_MES_SCHED_PIPE) {
+		if (adev->enable_uni_mes)
+			r = amdgpu_mes_map_legacy_queue(adev, ring);
+		else
+			r = mes_v12_0_kiq_enable_queue(adev);
+		if (r)
+			return r;
+	} else {
+		mes_v12_0_queue_init_register(ring);
+	}
+
+	if (((pipe == AMDGPU_MES_SCHED_PIPE) && !adev->mes.sched_version) ||
+	    ((pipe == AMDGPU_MES_KIQ_PIPE) && !adev->mes.kiq_version)) {
+		/* get MES scheduler/KIQ versions */
+		mutex_lock(&adev->srbm_mutex);
+		soc21_grbm_select(adev, 3, pipe, 0, 0);
+
+		if (pipe == AMDGPU_MES_SCHED_PIPE)
+			adev->mes.sched_version = RREG32_SOC15(GC, 0, regCP_MES_GP3_LO);
+		else if (pipe == AMDGPU_MES_KIQ_PIPE && adev->enable_mes_kiq)
+			adev->mes.kiq_version = RREG32_SOC15(GC, 0, regCP_MES_GP3_LO);
+
+		soc21_grbm_select(adev, 0, 0, 0, 0);
+		mutex_unlock(&adev->srbm_mutex);
+	}
+
+	return 0;
+}
+
+static int mes_v12_0_ring_init(struct amdgpu_device *adev, int pipe)
+{
+	struct amdgpu_ring *ring;
+
+	ring = &adev->mes.ring[pipe];
+
+	ring->funcs = &mes_v12_0_ring_funcs;
+
+	ring->me = 3;
+	ring->pipe = pipe;
+	ring->queue = 0;
+
+	ring->ring_obj = NULL;
+	ring->use_doorbell = true;
+	ring->eop_gpu_addr = adev->mes.eop_gpu_addr[pipe];
+	ring->no_scheduler = true;
+	sprintf(ring->name, "mes_%d.%d.%d", ring->me, ring->pipe, ring->queue);
+
+	if (pipe == AMDGPU_MES_SCHED_PIPE)
+		ring->doorbell_index = adev->doorbell_index.mes_ring0 << 1;
+	else
+		ring->doorbell_index = adev->doorbell_index.mes_ring1 << 1;
+
+	return amdgpu_ring_init(adev, ring, 1024, NULL, 0,
+				AMDGPU_RING_PRIO_DEFAULT, NULL);
+}
+
+static int mes_v12_0_kiq_ring_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+
+	spin_lock_init(&adev->gfx.kiq[0].ring_lock);
+
+	ring = &adev->gfx.kiq[0].ring;
+
+	ring->me = 3;
+	ring->pipe = 1;
+	ring->queue = 0;
+
+	ring->adev = NULL;
+	ring->ring_obj = NULL;
+	ring->use_doorbell = true;
+	ring->doorbell_index = adev->doorbell_index.mes_ring1 << 1;
+	ring->eop_gpu_addr = adev->mes.eop_gpu_addr[AMDGPU_MES_KIQ_PIPE];
+	ring->no_scheduler = true;
+	sprintf(ring->name, "mes_kiq_%d.%d.%d",
+		ring->me, ring->pipe, ring->queue);
+
+	return amdgpu_ring_init(adev, ring, 1024, NULL, 0,
+				AMDGPU_RING_PRIO_DEFAULT, NULL);
+}
+
+static int mes_v12_0_mqd_sw_init(struct amdgpu_device *adev,
+				 enum amdgpu_mes_pipe pipe)
+{
+	int r, mqd_size = sizeof(struct v12_compute_mqd);
+	struct amdgpu_ring *ring;
+
+	if (!adev->enable_uni_mes && pipe == AMDGPU_MES_KIQ_PIPE)
+		ring = &adev->gfx.kiq[0].ring;
+	else
+		ring = &adev->mes.ring[pipe];
+
+	if (ring->mqd_obj)
+		return 0;
+
+	r = amdgpu_bo_create_kernel(adev, mqd_size, PAGE_SIZE,
+				    AMDGPU_GEM_DOMAIN_GTT, &ring->mqd_obj,
+				    &ring->mqd_gpu_addr, &ring->mqd_ptr);
+	if (r) {
+		dev_warn(adev->dev, "failed to create ring mqd bo (%d)", r);
+		return r;
+	}
+
+	memset(ring->mqd_ptr, 0, mqd_size);
+
+	/* prepare MQD backup */
+	adev->mes.mqd_backup[pipe] = kmalloc(mqd_size, GFP_KERNEL);
+	if (!adev->mes.mqd_backup[pipe])
+		dev_warn(adev->dev,
+			 "no memory to create MQD backup for ring %s\n",
+			 ring->name);
+
+	return 0;
+}
+
+static int mes_v12_0_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int pipe, r;
+
+	adev->mes.funcs = &mes_v12_0_funcs;
+	adev->mes.kiq_hw_init = &mes_v12_0_kiq_hw_init;
+	adev->mes.kiq_hw_fini = &mes_v12_0_kiq_hw_fini;
+	adev->mes.enable_legacy_queue_map = true;
+
+	adev->mes.event_log_size = adev->enable_uni_mes ?
+		(AMDGPU_MAX_MES_PIPES * (AMDGPU_MES_LOG_BUFFER_SIZE + AMDGPU_MES_MSCRATCH_SIZE)) :
+		(AMDGPU_MES_LOG_BUFFER_SIZE + AMDGPU_MES_MSCRATCH_SIZE);
+	r = amdgpu_mes_init(adev);
+	if (r)
+		return r;
+
+	for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) {
+		r = mes_v12_0_allocate_eop_buf(adev, pipe);
+		if (r)
+			return r;
+
+		r = mes_v12_0_mqd_sw_init(adev, pipe);
+		if (r)
+			return r;
+
+		if (!adev->enable_uni_mes && pipe == AMDGPU_MES_KIQ_PIPE) {
+			r = mes_v12_0_kiq_ring_init(adev);
+		}
+		else {
+			r = mes_v12_0_ring_init(adev, pipe);
+			if (r)
+				return r;
+			r = amdgpu_bo_create_kernel(adev, AMDGPU_GPU_PAGE_SIZE, PAGE_SIZE,
+						    AMDGPU_GEM_DOMAIN_VRAM,
+						    &adev->mes.resource_1[pipe],
+						    &adev->mes.resource_1_gpu_addr[pipe],
+						    &adev->mes.resource_1_addr[pipe]);
+			if (r) {
+				dev_err(adev->dev, "(%d) failed to create mes resource_1 bo pipe[%d]\n", r, pipe);
+				return r;
+			}
+		}
+	}
+
+	return 0;
+}
+
+static int mes_v12_0_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int pipe;
+
+	for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) {
+		amdgpu_bo_free_kernel(&adev->mes.resource_1[pipe],
+				      &adev->mes.resource_1_gpu_addr[pipe],
+				      &adev->mes.resource_1_addr[pipe]);
+
+		kfree(adev->mes.mqd_backup[pipe]);
+
+		amdgpu_bo_free_kernel(&adev->mes.eop_gpu_obj[pipe],
+				      &adev->mes.eop_gpu_addr[pipe],
+				      NULL);
+		amdgpu_ucode_release(&adev->mes.fw[pipe]);
+
+		if (adev->enable_uni_mes || pipe == AMDGPU_MES_SCHED_PIPE) {
+			amdgpu_bo_free_kernel(&adev->mes.ring[pipe].mqd_obj,
+					      &adev->mes.ring[pipe].mqd_gpu_addr,
+					      &adev->mes.ring[pipe].mqd_ptr);
+			amdgpu_ring_fini(&adev->mes.ring[pipe]);
+		}
+	}
+
+	if (!adev->enable_uni_mes) {
+		amdgpu_bo_free_kernel(&adev->gfx.kiq[0].ring.mqd_obj,
+				      &adev->gfx.kiq[0].ring.mqd_gpu_addr,
+				      &adev->gfx.kiq[0].ring.mqd_ptr);
+		amdgpu_ring_fini(&adev->gfx.kiq[0].ring);
+	}
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
+		mes_v12_0_free_ucode_buffers(adev, AMDGPU_MES_KIQ_PIPE);
+		mes_v12_0_free_ucode_buffers(adev, AMDGPU_MES_SCHED_PIPE);
+	}
+
+	amdgpu_mes_fini(adev);
+	return 0;
+}
+
+static void mes_v12_0_kiq_dequeue_sched(struct amdgpu_device *adev)
+{
+	uint32_t data;
+	int i;
+
+	mutex_lock(&adev->srbm_mutex);
+	soc21_grbm_select(adev, 3, AMDGPU_MES_SCHED_PIPE, 0, 0);
+
+	/* disable the queue if it's active */
+	if (RREG32_SOC15(GC, 0, regCP_HQD_ACTIVE) & 1) {
+		WREG32_SOC15(GC, 0, regCP_HQD_DEQUEUE_REQUEST, 1);
+		for (i = 0; i < adev->usec_timeout; i++) {
+			if (!(RREG32_SOC15(GC, 0, regCP_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+	}
+	data = RREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL);
+	data = REG_SET_FIELD(data, CP_HQD_PQ_DOORBELL_CONTROL,
+				DOORBELL_EN, 0);
+	data = REG_SET_FIELD(data, CP_HQD_PQ_DOORBELL_CONTROL,
+				DOORBELL_HIT, 1);
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL, data);
+
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL, 0);
+
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_LO, 0);
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_HI, 0);
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR, 0);
+
+	soc21_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+
+	adev->mes.ring[0].sched.ready = false;
+}
+
+static void mes_v12_0_kiq_setting(struct amdgpu_ring *ring)
+{
+	uint32_t tmp;
+	struct amdgpu_device *adev = ring->adev;
+
+	/* tell RLC which is KIQ queue */
+	tmp = RREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS);
+	tmp &= 0xffffff00;
+	tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
+	WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp | 0x80);
+}
+
+static int mes_v12_0_kiq_hw_init(struct amdgpu_device *adev)
+{
+	int r = 0;
+	struct amdgpu_ip_block *ip_block;
+
+	if (adev->enable_uni_mes)
+		mes_v12_0_kiq_setting(&adev->mes.ring[AMDGPU_MES_KIQ_PIPE]);
+	else
+		mes_v12_0_kiq_setting(&adev->gfx.kiq[0].ring);
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
+
+		r = mes_v12_0_load_microcode(adev, AMDGPU_MES_SCHED_PIPE, false);
+		if (r) {
+			DRM_ERROR("failed to load MES fw, r=%d\n", r);
+			return r;
+		}
+
+		r = mes_v12_0_load_microcode(adev, AMDGPU_MES_KIQ_PIPE, true);
+		if (r) {
+			DRM_ERROR("failed to load MES kiq fw, r=%d\n", r);
+			return r;
+		}
+
+		mes_v12_0_set_ucode_start_addr(adev);
+
+	} else if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO)
+		mes_v12_0_set_ucode_start_addr(adev);
+
+	mes_v12_0_enable(adev, true);
+
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_MES);
+	if (unlikely(!ip_block)) {
+		dev_err(adev->dev, "Failed to get MES handle\n");
+		return -EINVAL;
+	}
+
+	r = mes_v12_0_queue_init(adev, AMDGPU_MES_KIQ_PIPE);
+	if (r)
+		goto failure;
+
+	if (adev->enable_uni_mes) {
+		r = mes_v12_0_set_hw_resources(&adev->mes, AMDGPU_MES_KIQ_PIPE);
+		if (r)
+			goto failure;
+
+		mes_v12_0_set_hw_resources_1(&adev->mes, AMDGPU_MES_KIQ_PIPE);
+	}
+
+	if (adev->mes.enable_legacy_queue_map) {
+		r = mes_v12_0_hw_init(ip_block);
+		if (r)
+			goto failure;
+	}
+
+	return r;
+
+failure:
+	mes_v12_0_hw_fini(ip_block);
+	return r;
+}
+
+static int mes_v12_0_kiq_hw_fini(struct amdgpu_device *adev)
+{
+	if (adev->mes.ring[0].sched.ready) {
+		if (adev->enable_uni_mes)
+			amdgpu_mes_unmap_legacy_queue(adev,
+				      &adev->mes.ring[AMDGPU_MES_SCHED_PIPE],
+				      RESET_QUEUES, 0, 0);
+		else
+			mes_v12_0_kiq_dequeue_sched(adev);
+
+		adev->mes.ring[0].sched.ready = false;
+	}
+
+	mes_v12_0_enable(adev, false);
+
+	return 0;
+}
+
+static int mes_v12_0_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (adev->mes.ring[0].sched.ready)
+		goto out;
+
+	if (!adev->enable_mes_kiq) {
+		if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
+			r = mes_v12_0_load_microcode(adev,
+					     AMDGPU_MES_SCHED_PIPE, true);
+			if (r) {
+				DRM_ERROR("failed to MES fw, r=%d\n", r);
+				return r;
+			}
+
+			mes_v12_0_set_ucode_start_addr(adev);
+
+		} else if (adev->firmware.load_type ==
+			   AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) {
+
+			mes_v12_0_set_ucode_start_addr(adev);
+		}
+
+		mes_v12_0_enable(adev, true);
+	}
+
+	/* Enable the MES to handle doorbell ring on unmapped queue */
+	mes_v12_0_enable_unmapped_doorbell_handling(&adev->mes, true);
+
+	r = mes_v12_0_queue_init(adev, AMDGPU_MES_SCHED_PIPE);
+	if (r)
+		goto failure;
+
+	r = mes_v12_0_set_hw_resources(&adev->mes, AMDGPU_MES_SCHED_PIPE);
+	if (r)
+		goto failure;
+
+	if ((adev->mes.sched_version & AMDGPU_MES_VERSION_MASK) >= 0x4b)
+		mes_v12_0_set_hw_resources_1(&adev->mes, AMDGPU_MES_SCHED_PIPE);
+
+	mes_v12_0_init_aggregated_doorbell(&adev->mes);
+
+	r = mes_v12_0_query_sched_status(&adev->mes, AMDGPU_MES_SCHED_PIPE);
+	if (r) {
+		DRM_ERROR("MES is busy\n");
+		goto failure;
+	}
+
+	r = amdgpu_mes_update_enforce_isolation(adev);
+	if (r)
+		goto failure;
+
+out:
+	/*
+	 * Disable KIQ ring usage from the driver once MES is enabled.
+	 * MES uses KIQ ring exclusively so driver cannot access KIQ ring
+	 * with MES enabled.
+	 */
+	adev->gfx.kiq[0].ring.sched.ready = false;
+	adev->mes.ring[0].sched.ready = true;
+
+	return 0;
+
+failure:
+	mes_v12_0_hw_fini(ip_block);
+	return r;
+}
+
+static int mes_v12_0_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	return 0;
+}
+
+static int mes_v12_0_suspend(struct amdgpu_ip_block *ip_block)
+{
+	return mes_v12_0_hw_fini(ip_block);
+}
+
+static int mes_v12_0_resume(struct amdgpu_ip_block *ip_block)
+{
+	return mes_v12_0_hw_init(ip_block);
+}
+
+static int mes_v12_0_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int pipe, r;
+
+	adev->mes.hung_queue_db_array_size = MES12_HUNG_DB_OFFSET_ARRAY_SIZE;
+	adev->mes.hung_queue_hqd_info_offset = MES12_HUNG_HQD_INFO_OFFSET;
+
+	for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) {
+		r = amdgpu_mes_init_microcode(adev, pipe);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+static const struct amd_ip_funcs mes_v12_0_ip_funcs = {
+	.name = "mes_v12_0",
+	.early_init = mes_v12_0_early_init,
+	.late_init = NULL,
+	.sw_init = mes_v12_0_sw_init,
+	.sw_fini = mes_v12_0_sw_fini,
+	.hw_init = mes_v12_0_hw_init,
+	.hw_fini = mes_v12_0_hw_fini,
+	.suspend = mes_v12_0_suspend,
+	.resume = mes_v12_0_resume,
+};
+
+const struct amdgpu_ip_block_version mes_v12_0_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_MES,
+	.major = 12,
+	.minor = 0,
+	.rev = 0,
+	.funcs = &mes_v12_0_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v12_0.h b/drivers/gpu/drm/amd/amdgpu/mes_v12_0.h
new file mode 100644
index 000000000000..ac3740f353aa
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/mes_v12_0.h
@@ -0,0 +1,29 @@
+/*
+ * Copyright 2021 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __MES_V12_0_H__
+#define __MES_V12_0_H__
+
+extern const struct amdgpu_ip_block_version mes_v12_0_ip_block;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c
index 15e7cbeae75b..243eabda0607 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c
@@ -54,7 +54,7 @@ static u64 mmhub_v1_0_get_fb_location(struct amdgpu_device *adev)
 static void mmhub_v1_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
 				uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 
 	WREG32_SOC15_OFFSET(MMHUB, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
 			    hub->ctx_addr_distance * vmid,
@@ -96,7 +96,9 @@ static void mmhub_v1_0_init_system_aperture_regs(struct amdgpu_device *adev)
 	WREG32_SOC15(MMHUB, 0, mmMC_VM_SYSTEM_APERTURE_LOW_ADDR,
 		     min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18);
 
-	if (adev->apu_flags & AMD_APU_IS_RAVEN2)
+	if (adev->apu_flags & (AMD_APU_IS_RAVEN2 |
+			       AMD_APU_IS_RENOIR |
+			       AMD_APU_IS_GREEN_SARDINE))
 		/*
 		 * Raven2 has a HW issue that it is unable to use the vram which
 		 * is out of MC_VM_SYSTEM_APERTURE_HIGH_ADDR. So here is the
@@ -227,9 +229,55 @@ static void mmhub_v1_0_disable_identity_aperture(struct amdgpu_device *adev)
 		     0);
 }
 
+static void mmhub_v1_0_init_saw(struct amdgpu_device *adev)
+{
+	uint64_t pt_base = amdgpu_gmc_pd_addr(adev->gart.bo);
+	uint32_t tmp;
+
+	/* VM_9_X_REGISTER_VM_L2_SAW_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32 */
+	WREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
+			    lower_32_bits(pt_base >> 12));
+
+	/* VM_9_X_REGISTER_VM_L2_SAW_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32 */
+	WREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
+			    upper_32_bits(pt_base >> 12));
+
+	/* VM_9_X_REGISTER_VM_L2_SAW_CONTEXT0_PAGE_TABLE_START_ADDR_LO32 */
+	WREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
+			    (u32)(adev->gmc.gart_start >> 12));
+
+	/* VM_9_X_REGISTER_VM_L2_SAW_CONTEXT0_PAGE_TABLE_START_ADDR_HI32 */
+	WREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
+			    (u32)(adev->gmc.gart_start >> 44));
+
+	/* VM_9_X_REGISTER_VM_L2_SAW_CONTEXT0_PAGE_TABLE_END_ADDR_LO32 */
+	WREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
+			    (u32)(adev->gmc.gart_end >> 12));
+
+	/* VM_9_X_REGISTER_VM_L2_SAW_CONTEXT0_PAGE_TABLE_END_ADDR_HI32 */
+	WREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
+			    (u32)(adev->gmc.gart_end >> 44));
+
+	/* Program SAW CONTEXT0 CNTL */
+	tmp = RREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CONTEXT0_CNTL);
+	tmp |= 1 << CONTEXT0_CNTL_ENABLE_OFFSET;
+	tmp &= ~(3 << CONTEXT0_CNTL_PAGE_TABLE_DEPTH_OFFSET);
+	WREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CONTEXT0_CNTL, tmp);
+
+	/* Disable all Contexts except Context0 */
+	tmp = 0xfffe;
+	WREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CONTEXTS_DISABLE, tmp);
+
+	/* Program SAW CNTL4 */
+	tmp = RREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CNTL4);
+	tmp |= 1 << VMC_TAP_PDE_REQUEST_SNOOP_OFFSET;
+	tmp |= 1 << VMC_TAP_PTE_REQUEST_SNOOP_OFFSET;
+	WREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CNTL4, tmp);
+}
+
 static void mmhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	unsigned num_level, block_size;
 	uint32_t tmp;
 	int i;
@@ -242,7 +290,7 @@ static void mmhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
 		block_size -= 9;
 
 	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(MMHUB, 0, mmVM_CONTEXT1_CNTL, i);
+		tmp = RREG32_SOC15_OFFSET(MMHUB, 0, mmVM_CONTEXT1_CNTL, i * hub->ctx_distance);
 		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
 		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
 				    num_level);
@@ -281,11 +329,14 @@ static void mmhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
 				    i * hub->ctx_addr_distance,
 				    upper_32_bits(adev->vm_manager.max_pfn - 1));
 	}
+
+	if (amdgpu_ip_version(adev, ISP_HWIP, 0))
+		mmhub_v1_0_init_saw(adev);
 }
 
 static void mmhub_v1_0_program_invalidation(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	unsigned i;
 
 	for (i = 0; i < 18; ++i) {
@@ -305,7 +356,7 @@ static void mmhub_v1_0_update_power_gating(struct amdgpu_device *adev,
 	if (adev->pg_flags & AMD_PG_SUPPORT_MMHUB)
 		amdgpu_dpm_set_powergating_by_smu(adev,
 						  AMD_IP_BLOCK_TYPE_GMC,
-						  enable);
+						  enable, 0);
 }
 
 static int mmhub_v1_0_gart_enable(struct amdgpu_device *adev)
@@ -338,7 +389,7 @@ static int mmhub_v1_0_gart_enable(struct amdgpu_device *adev)
 
 static void mmhub_v1_0_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	u32 tmp;
 	u32 i;
 
@@ -415,7 +466,7 @@ static void mmhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev, bool
 
 static void mmhub_v1_0_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 
 	hub->ctx0_ptb_addr_lo32 =
 		SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c
index 73afbf2facc9..2adee2b94c37 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c
@@ -54,7 +54,7 @@ static u64 mmhub_v1_7_get_fb_location(struct amdgpu_device *adev)
 static void mmhub_v1_7_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
 				uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 
 	WREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
 			hub->ctx_addr_distance * vmid, lower_32_bits(page_table_base));
@@ -172,6 +172,30 @@ static void mmhub_v1_7_init_tlb_regs(struct amdgpu_device *adev)
 	WREG32_SOC15(MMHUB, 0, regMC_VM_MX_L1_TLB_CNTL, tmp);
 }
 
+/* Set snoop bit for SDMA so that SDMA writes probe-invalidates RW lines */
+static void mmhub_v1_7_init_snoop_override_regs(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+	int i;
+	uint32_t distance = regDAGB1_WRCLI_GPU_SNOOP_OVERRIDE -
+			    regDAGB0_WRCLI_GPU_SNOOP_OVERRIDE;
+
+	for (i = 0; i < 5; i++) { /* DAGB instances */
+		tmp = RREG32_SOC15_OFFSET(MMHUB, 0,
+			regDAGB0_WRCLI_GPU_SNOOP_OVERRIDE, i * distance);
+		tmp |= (1 << 15); /* SDMA client is BIT15 */
+		WREG32_SOC15_OFFSET(MMHUB, 0,
+			regDAGB0_WRCLI_GPU_SNOOP_OVERRIDE, i * distance, tmp);
+
+		tmp = RREG32_SOC15_OFFSET(MMHUB, 0,
+			regDAGB0_WRCLI_GPU_SNOOP_OVERRIDE_VALUE, i * distance);
+		tmp |= (1 << 15);
+		WREG32_SOC15_OFFSET(MMHUB, 0,
+			regDAGB0_WRCLI_GPU_SNOOP_OVERRIDE_VALUE, i * distance, tmp);
+	}
+
+}
+
 static void mmhub_v1_7_init_cache_regs(struct amdgpu_device *adev)
 {
 	uint32_t tmp;
@@ -261,7 +285,7 @@ static void mmhub_v1_7_disable_identity_aperture(struct amdgpu_device *adev)
 
 static void mmhub_v1_7_setup_vmid_config(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	unsigned num_level, block_size;
 	uint32_t tmp;
 	int i;
@@ -274,7 +298,7 @@ static void mmhub_v1_7_setup_vmid_config(struct amdgpu_device *adev)
 		block_size -= 9;
 
 	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT1_CNTL, i);
+		tmp = RREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT1_CNTL, i * hub->ctx_distance);
 		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
 		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
 				    num_level);
@@ -319,7 +343,7 @@ static void mmhub_v1_7_setup_vmid_config(struct amdgpu_device *adev)
 
 static void mmhub_v1_7_program_invalidation(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	unsigned i;
 
 	for (i = 0; i < 18; ++i) {
@@ -337,6 +361,7 @@ static int mmhub_v1_7_gart_enable(struct amdgpu_device *adev)
 	mmhub_v1_7_init_system_aperture_regs(adev);
 	mmhub_v1_7_init_tlb_regs(adev);
 	mmhub_v1_7_init_cache_regs(adev);
+	mmhub_v1_7_init_snoop_override_regs(adev);
 
 	mmhub_v1_7_enable_system_domain(adev);
 	mmhub_v1_7_disable_identity_aperture(adev);
@@ -348,7 +373,7 @@ static int mmhub_v1_7_gart_enable(struct amdgpu_device *adev)
 
 static void mmhub_v1_7_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	u32 tmp;
 	u32 i;
 
@@ -425,7 +450,7 @@ static void mmhub_v1_7_set_fault_enable_default(struct amdgpu_device *adev, bool
 
 static void mmhub_v1_7_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 
 	hub->ctx0_ptb_addr_lo32 =
 		SOC15_REG_OFFSET(MMHUB, 0,
@@ -544,7 +569,7 @@ static int mmhub_v1_7_set_clockgating(struct amdgpu_device *adev,
 
 static void mmhub_v1_7_get_clockgating(struct amdgpu_device *adev, u64 *flags)
 {
-	int data, data1;
+	u32 data, data1;
 
 	if (amdgpu_sriov_vf(adev))
 		*flags = 0;
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c
index 342d1702104c..cc688ae79e84 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c
@@ -29,6 +29,8 @@
 
 #include "soc15_common.h"
 #include "soc15.h"
+#include "amdgpu_ras.h"
+#include "amdgpu_psp.h"
 
 #define regVM_L2_CNTL3_DEFAULT	0x80100007
 #define regVM_L2_CNTL4_DEFAULT	0x000000c1
@@ -53,18 +55,32 @@ static u64 mmhub_v1_8_get_fb_location(struct amdgpu_device *adev)
 static void mmhub_v1_8_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
 				uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
-
-	WREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
-			    hub->ctx_addr_distance * vmid, lower_32_bits(page_table_base));
+	struct amdgpu_vmhub *hub;
+	u32 inst_mask;
+	int i;
 
-	WREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
-			    hub->ctx_addr_distance * vmid, upper_32_bits(page_table_base));
+	inst_mask = adev->aid_mask;
+	for_each_inst(i, inst_mask) {
+		hub = &adev->vmhub[AMDGPU_MMHUB0(i)];
+		WREG32_SOC15_OFFSET(MMHUB, i,
+				    regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
+				    hub->ctx_addr_distance * vmid,
+				    lower_32_bits(page_table_base));
+
+		WREG32_SOC15_OFFSET(MMHUB, i,
+				    regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
+				    hub->ctx_addr_distance * vmid,
+				    upper_32_bits(page_table_base));
+	}
 }
 
 static void mmhub_v1_8_init_gart_aperture_regs(struct amdgpu_device *adev)
 {
+	uint64_t gart_start = amdgpu_virt_xgmi_migrate_enabled(adev) ?
+			adev->gmc.vram_start : adev->gmc.fb_start;
 	uint64_t pt_base;
+	u32 inst_mask;
+	int i;
 
 	if (adev->gmc.pdb0_bo)
 		pt_base = amdgpu_gmc_pd_addr(adev->gmc.pdb0_bo);
@@ -76,187 +92,294 @@ static void mmhub_v1_8_init_gart_aperture_regs(struct amdgpu_device *adev)
 	/* If use GART for FB translation, vmid0 page table covers both
 	 * vram and system memory (gart)
 	 */
-	if (adev->gmc.pdb0_bo) {
-		WREG32_SOC15(MMHUB, 0, regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
-			     (u32)(adev->gmc.fb_start >> 12));
-		WREG32_SOC15(MMHUB, 0, regVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
-			     (u32)(adev->gmc.fb_start >> 44));
-
-		WREG32_SOC15(MMHUB, 0, regVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
-			     (u32)(adev->gmc.gart_end >> 12));
-		WREG32_SOC15(MMHUB, 0, regVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
-			     (u32)(adev->gmc.gart_end >> 44));
-
-	} else {
-		WREG32_SOC15(MMHUB, 0, regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
-			     (u32)(adev->gmc.gart_start >> 12));
-		WREG32_SOC15(MMHUB, 0, regVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
-			     (u32)(adev->gmc.gart_start >> 44));
-
-		WREG32_SOC15(MMHUB, 0, regVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
-			     (u32)(adev->gmc.gart_end >> 12));
-		WREG32_SOC15(MMHUB, 0, regVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
-			     (u32)(adev->gmc.gart_end >> 44));
+	inst_mask = adev->aid_mask;
+	for_each_inst(i, inst_mask) {
+		if (adev->gmc.pdb0_bo) {
+			WREG32_SOC15(MMHUB, i,
+				     regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
+				     (u32)(gart_start >> 12));
+			WREG32_SOC15(MMHUB, i,
+				     regVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
+				     (u32)(gart_start >> 44));
+
+			WREG32_SOC15(MMHUB, i,
+				     regVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
+				     (u32)(adev->gmc.gart_end >> 12));
+			WREG32_SOC15(MMHUB, i,
+				     regVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
+				     (u32)(adev->gmc.gart_end >> 44));
+
+		} else {
+			WREG32_SOC15(MMHUB, i,
+				     regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
+				     (u32)(adev->gmc.gart_start >> 12));
+			WREG32_SOC15(MMHUB, i,
+				     regVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
+				     (u32)(adev->gmc.gart_start >> 44));
+
+			WREG32_SOC15(MMHUB, i,
+				     regVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
+				     (u32)(adev->gmc.gart_end >> 12));
+			WREG32_SOC15(MMHUB, i,
+				     regVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
+				     (u32)(adev->gmc.gart_end >> 44));
+		}
 	}
 }
 
 static void mmhub_v1_8_init_system_aperture_regs(struct amdgpu_device *adev)
 {
+	uint32_t tmp, inst_mask;
 	uint64_t value;
-	uint32_t tmp;
+	int i;
 
-	/* Program the AGP BAR */
-	WREG32_SOC15(MMHUB, 0, regMC_VM_AGP_BASE, 0);
-	WREG32_SOC15(MMHUB, 0, regMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
-	WREG32_SOC15(MMHUB, 0, regMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
+	if (amdgpu_sriov_vf(adev))
+		return;
 
-	/* Program the system aperture low logical page number. */
-	WREG32_SOC15(MMHUB, 0, regMC_VM_SYSTEM_APERTURE_LOW_ADDR,
-		     min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18);
+	inst_mask = adev->aid_mask;
+	for_each_inst(i, inst_mask) {
+		/* Program the AGP BAR */
+		WREG32_SOC15(MMHUB, i, regMC_VM_AGP_BASE, 0);
+		WREG32_SOC15(MMHUB, i, regMC_VM_AGP_BOT,
+			     adev->gmc.agp_start >> 24);
+		WREG32_SOC15(MMHUB, i, regMC_VM_AGP_TOP,
+			     adev->gmc.agp_end >> 24);
 
-	WREG32_SOC15(MMHUB, 0, regMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
-		     max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
+		/* Program the system aperture low logical page number. */
+		WREG32_SOC15(MMHUB, i, regMC_VM_SYSTEM_APERTURE_LOW_ADDR,
+			min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18);
 
-	/* In the case squeezing vram into GART aperture, we don't use
-	 * FB aperture and AGP aperture. Disable them.
-	 */
-	if (adev->gmc.pdb0_bo) {
-		WREG32_SOC15(MMHUB, 0, regMC_VM_AGP_BOT, 0xFFFFFF);
-		WREG32_SOC15(MMHUB, 0, regMC_VM_AGP_TOP, 0);
-		WREG32_SOC15(MMHUB, 0, regMC_VM_FB_LOCATION_TOP, 0);
-		WREG32_SOC15(MMHUB, 0, regMC_VM_FB_LOCATION_BASE, 0x00FFFFFF);
-		WREG32_SOC15(MMHUB, 0, regMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x3FFFFFFF);
-		WREG32_SOC15(MMHUB, 0, regMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0);
-	}
-	if (amdgpu_sriov_vf(adev))
-		return;
+		WREG32_SOC15(MMHUB, i, regMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+			max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
 
-	/* Set default page address. */
-	value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
-	WREG32_SOC15(MMHUB, 0, regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
-		     (u32)(value >> 12));
-	WREG32_SOC15(MMHUB, 0, regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
-		     (u32)(value >> 44));
-
-	/* Program "protection fault". */
-	WREG32_SOC15(MMHUB, 0, regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32,
-		     (u32)(adev->dummy_page_addr >> 12));
-	WREG32_SOC15(MMHUB, 0, regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32,
-		     (u32)((u64)adev->dummy_page_addr >> 44));
-
-	tmp = RREG32_SOC15(MMHUB, 0, regVM_L2_PROTECTION_FAULT_CNTL2);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL2,
-			    ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1);
-	WREG32_SOC15(MMHUB, 0, regVM_L2_PROTECTION_FAULT_CNTL2, tmp);
+		/* In the case squeezing vram into GART aperture, we don't use
+		 * FB aperture and AGP aperture. Disable them.
+		 */
+		if (adev->gmc.pdb0_bo) {
+			WREG32_SOC15(MMHUB, i, regMC_VM_AGP_BOT, 0xFFFFFF);
+			WREG32_SOC15(MMHUB, i, regMC_VM_AGP_TOP, 0);
+			WREG32_SOC15(MMHUB, i, regMC_VM_FB_LOCATION_TOP, 0);
+			WREG32_SOC15(MMHUB, i, regMC_VM_FB_LOCATION_BASE,
+				     0x00FFFFFF);
+			WREG32_SOC15(MMHUB, i,
+				     regMC_VM_SYSTEM_APERTURE_LOW_ADDR,
+				     0x3FFFFFFF);
+			WREG32_SOC15(MMHUB, i,
+				     regMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0);
+		}
+
+		/* Set default page address. */
+		value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
+		WREG32_SOC15(MMHUB, i, regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
+			     (u32)(value >> 12));
+		WREG32_SOC15(MMHUB, i, regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
+			     (u32)(value >> 44));
+
+		/* Program "protection fault". */
+		WREG32_SOC15(MMHUB, i,
+			     regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32,
+			     (u32)(adev->dummy_page_addr >> 12));
+		WREG32_SOC15(MMHUB, i,
+			     regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32,
+			     (u32)((u64)adev->dummy_page_addr >> 44));
+
+		tmp = RREG32_SOC15(MMHUB, i, regVM_L2_PROTECTION_FAULT_CNTL2);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL2,
+				    ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1);
+		WREG32_SOC15(MMHUB, i, regVM_L2_PROTECTION_FAULT_CNTL2, tmp);
+	}
 }
 
 static void mmhub_v1_8_init_tlb_regs(struct amdgpu_device *adev)
 {
-	uint32_t tmp;
-
-	/* Setup TLB control */
-	tmp = RREG32_SOC15(MMHUB, 0, regMC_VM_MX_L1_TLB_CNTL);
-
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1);
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE, 3);
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
-			    ENABLE_ADVANCED_DRIVER_MODEL, 1);
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
-			    SYSTEM_APERTURE_UNMAPPED_ACCESS, 0);
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
-			    MTYPE, MTYPE_UC);/* XXX for emulation. */
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ATC_EN, 1);
-
-	WREG32_SOC15(MMHUB, 0, regMC_VM_MX_L1_TLB_CNTL, tmp);
+	uint32_t tmp, inst_mask;
+	int i;
+
+	if (amdgpu_sriov_reg_indirect_l1_tlb_cntl(adev)) {
+		tmp = RREG32_SOC15(MMHUB, 0, regMC_VM_MX_L1_TLB_CNTL);
+
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB,
+				    1);
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+				    SYSTEM_ACCESS_MODE, 3);
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+				    ENABLE_ADVANCED_DRIVER_MODEL, 1);
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+				    SYSTEM_APERTURE_UNMAPPED_ACCESS, 0);
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+				    MTYPE, MTYPE_UC);/* XXX for emulation. */
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ATC_EN, 1);
+
+		psp_reg_program_no_ring(&adev->psp, tmp, PSP_REG_MMHUB_L1_TLB_CNTL);
+	} else {
+		inst_mask = adev->aid_mask;
+		for_each_inst(i, inst_mask) {
+			tmp = RREG32_SOC15(MMHUB, i, regMC_VM_MX_L1_TLB_CNTL);
+
+			tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB,
+					    1);
+			tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+					    SYSTEM_ACCESS_MODE, 3);
+			tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+					    ENABLE_ADVANCED_DRIVER_MODEL, 1);
+			tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+					    SYSTEM_APERTURE_UNMAPPED_ACCESS, 0);
+			tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+					    MTYPE, MTYPE_UC);/* XXX for emulation. */
+			tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ATC_EN, 1);
+
+			WREG32_SOC15(MMHUB, i, regMC_VM_MX_L1_TLB_CNTL, tmp);
+		}
+	}
+}
+
+/* Set snoop bit for SDMA so that SDMA writes probe-invalidates RW lines */
+static void mmhub_v1_8_init_snoop_override_regs(struct amdgpu_device *adev)
+{
+	uint32_t tmp, inst_mask;
+	int i, j;
+	uint32_t distance = regDAGB1_WRCLI_GPU_SNOOP_OVERRIDE -
+			    regDAGB0_WRCLI_GPU_SNOOP_OVERRIDE;
+
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	inst_mask = adev->aid_mask;
+	for_each_inst(i, inst_mask) {
+		for (j = 0; j < 5; j++) { /* DAGB instances */
+			tmp = RREG32_SOC15_OFFSET(MMHUB, i,
+				regDAGB0_WRCLI_GPU_SNOOP_OVERRIDE, j * distance);
+			tmp |= (1 << 15); /* SDMA client is BIT15 */
+			WREG32_SOC15_OFFSET(MMHUB, i,
+				regDAGB0_WRCLI_GPU_SNOOP_OVERRIDE, j * distance, tmp);
+
+			tmp = RREG32_SOC15_OFFSET(MMHUB, i,
+				regDAGB0_WRCLI_GPU_SNOOP_OVERRIDE_VALUE, j * distance);
+			tmp |= (1 << 15);
+			WREG32_SOC15_OFFSET(MMHUB, i,
+				regDAGB0_WRCLI_GPU_SNOOP_OVERRIDE_VALUE, j * distance, tmp);
+		}
+	}
 }
 
 static void mmhub_v1_8_init_cache_regs(struct amdgpu_device *adev)
 {
-	uint32_t tmp;
+	uint32_t tmp, inst_mask;
+	int i;
 
 	if (amdgpu_sriov_vf(adev))
 		return;
 
 	/* Setup L2 cache */
-	tmp = RREG32_SOC15(MMHUB, 0, regVM_L2_CNTL);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 1);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 1);
-	/* XXX for emulation, Refer to closed source code.*/
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, L2_PDE0_CACHE_TAG_GENERATION_MODE,
-			    0);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0);
-	WREG32_SOC15(MMHUB, 0, regVM_L2_CNTL, tmp);
-
-	tmp = RREG32_SOC15(MMHUB, 0, regVM_L2_CNTL2);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_L2_CACHE, 1);
-	WREG32_SOC15(MMHUB, 0, regVM_L2_CNTL2, tmp);
-
-	tmp = regVM_L2_CNTL3_DEFAULT;
-	if (adev->gmc.translate_further) {
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 12);
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3,
-				    L2_CACHE_BIGK_FRAGMENT_SIZE, 9);
-	} else {
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 9);
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3,
-				    L2_CACHE_BIGK_FRAGMENT_SIZE, 6);
-	}
-	WREG32_SOC15(MMHUB, 0, regVM_L2_CNTL3, tmp);
-
-	tmp = regVM_L2_CNTL4_DEFAULT;
-	if (adev->gmc.xgmi.connected_to_cpu) {
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4,
-				    VMC_TAP_PDE_REQUEST_PHYSICAL, 1);
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4,
-				    VMC_TAP_PTE_REQUEST_PHYSICAL, 1);
-	} else {
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4,
-				    VMC_TAP_PDE_REQUEST_PHYSICAL, 0);
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4,
-				    VMC_TAP_PTE_REQUEST_PHYSICAL, 0);
+	inst_mask = adev->aid_mask;
+	for_each_inst(i, inst_mask) {
+		tmp = RREG32_SOC15(MMHUB, i, regVM_L2_CNTL);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 1);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL,
+				    ENABLE_L2_FRAGMENT_PROCESSING, 1);
+		/* XXX for emulation, Refer to closed source code.*/
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL,
+				    L2_PDE0_CACHE_TAG_GENERATION_MODE, 0);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, PDE_FAULT_CLASSIFICATION,
+				    0);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL,
+				    CONTEXT1_IDENTITY_ACCESS_MODE, 1);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL,
+				    IDENTITY_MODE_FRAGMENT_SIZE, 0);
+		WREG32_SOC15(MMHUB, i, regVM_L2_CNTL, tmp);
+
+		tmp = RREG32_SOC15(MMHUB, i, regVM_L2_CNTL2);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS,
+				    1);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_L2_CACHE, 1);
+		WREG32_SOC15(MMHUB, i, regVM_L2_CNTL2, tmp);
+
+		tmp = regVM_L2_CNTL3_DEFAULT;
+		if (adev->gmc.translate_further) {
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 12);
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3,
+					    L2_CACHE_BIGK_FRAGMENT_SIZE, 9);
+		} else {
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 9);
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3,
+					    L2_CACHE_BIGK_FRAGMENT_SIZE, 6);
+		}
+		WREG32_SOC15(MMHUB, i, regVM_L2_CNTL3, tmp);
+
+		tmp = regVM_L2_CNTL4_DEFAULT;
+		/* For AMD APP APUs setup WC memory */
+		if (adev->gmc.xgmi.connected_to_cpu || adev->gmc.is_app_apu) {
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4,
+					    VMC_TAP_PDE_REQUEST_PHYSICAL, 1);
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4,
+					    VMC_TAP_PTE_REQUEST_PHYSICAL, 1);
+		} else {
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4,
+					    VMC_TAP_PDE_REQUEST_PHYSICAL, 0);
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4,
+					    VMC_TAP_PTE_REQUEST_PHYSICAL, 0);
+		}
+		WREG32_SOC15(MMHUB, i, regVM_L2_CNTL4, tmp);
 	}
-	WREG32_SOC15(MMHUB, 0, regVM_L2_CNTL4, tmp);
 }
 
 static void mmhub_v1_8_enable_system_domain(struct amdgpu_device *adev)
 {
-	uint32_t tmp;
-
-	tmp = RREG32_SOC15(MMHUB, 0, regVM_CONTEXT0_CNTL);
-	tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
-	tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH,
-			adev->gmc.vmid0_page_table_depth);
-	tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_BLOCK_SIZE,
-			adev->gmc.vmid0_page_table_block_size);
-	tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL,
-			    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0);
-	WREG32_SOC15(MMHUB, 0, regVM_CONTEXT0_CNTL, tmp);
+	uint32_t tmp, inst_mask;
+	int i;
+
+	inst_mask = adev->aid_mask;
+	for_each_inst(i, inst_mask) {
+		tmp = RREG32_SOC15(MMHUB, i, regVM_CONTEXT0_CNTL);
+		tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
+		tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH,
+				adev->gmc.vmid0_page_table_depth);
+		tmp = REG_SET_FIELD(tmp,
+				    VM_CONTEXT0_CNTL, PAGE_TABLE_BLOCK_SIZE,
+				    adev->gmc.vmid0_page_table_block_size);
+		tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL,
+				    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0);
+		WREG32_SOC15(MMHUB, i, regVM_CONTEXT0_CNTL, tmp);
+	}
 }
 
 static void mmhub_v1_8_disable_identity_aperture(struct amdgpu_device *adev)
 {
+	u32 inst_mask;
+	int i;
+
 	if (amdgpu_sriov_vf(adev))
 		return;
 
-	WREG32_SOC15(MMHUB, 0, regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32, 0xFFFFFFFF);
-	WREG32_SOC15(MMHUB, 0, regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32, 0x0000000F);
-
-	WREG32_SOC15(MMHUB, 0, regVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32, 0);
-	WREG32_SOC15(MMHUB, 0, regVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32, 0);
-
-	WREG32_SOC15(MMHUB, 0, regVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32, 0);
-	WREG32_SOC15(MMHUB, 0, regVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32, 0);
+	inst_mask = adev->aid_mask;
+	for_each_inst(i, inst_mask) {
+		WREG32_SOC15(MMHUB, i,
+			     regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32,
+			     0XFFFFFFFF);
+		WREG32_SOC15(MMHUB, i,
+			     regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32,
+			     0x0000000F);
+
+		WREG32_SOC15(MMHUB, i,
+			     regVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32,
+			     0);
+		WREG32_SOC15(MMHUB, i,
+			     regVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32,
+			     0);
+
+		WREG32_SOC15(MMHUB, i,
+			     regVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32, 0);
+		WREG32_SOC15(MMHUB, i,
+			     regVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32, 0);
+	}
 }
 
 static void mmhub_v1_8_setup_vmid_config(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
-	unsigned num_level, block_size;
-	uint32_t tmp;
-	int i;
+	struct amdgpu_vmhub *hub;
+	unsigned int num_level, block_size;
+	uint32_t tmp, inst_mask;
+	int i, j;
 
 	num_level = adev->vm_manager.num_level;
 	block_size = adev->vm_manager.block_size;
@@ -265,82 +388,86 @@ static void mmhub_v1_8_setup_vmid_config(struct amdgpu_device *adev)
 	else
 		block_size -= 9;
 
-	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT1_CNTL, i);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
-				    num_level);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT,
-				    1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    PAGE_TABLE_BLOCK_SIZE,
-				    block_size);
-		/* On Aldebaran, XNACK can be enabled in the SQ per-process.
-		 * Retry faults need to be enabled for that to work.
-		 */
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT,
-				    1);
-		WREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT1_CNTL,
-				    i * hub->ctx_distance, tmp);
-		WREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32,
-				    i * hub->ctx_addr_distance, 0);
-		WREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32,
-				    i * hub->ctx_addr_distance, 0);
-		WREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32,
-				    i * hub->ctx_addr_distance,
-				    lower_32_bits(adev->vm_manager.max_pfn - 1));
-		WREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32,
-				    i * hub->ctx_addr_distance,
-				    upper_32_bits(adev->vm_manager.max_pfn - 1));
+	inst_mask = adev->aid_mask;
+	for_each_inst(j, inst_mask) {
+		hub = &adev->vmhub[AMDGPU_MMHUB0(j)];
+		for (i = 0; i <= 14; i++) {
+			tmp = RREG32_SOC15_OFFSET(MMHUB, j, regVM_CONTEXT1_CNTL,
+						  i * hub->ctx_distance);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    ENABLE_CONTEXT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    PAGE_TABLE_DEPTH, num_level);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+				DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+				PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+				VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+				READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+				WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+				EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    PAGE_TABLE_BLOCK_SIZE,
+					    block_size);
+			/* On 9.4.3, XNACK can be enabled in the SQ
+			 * per-process. Retry faults need to be enabled for
+			 * that to work.
+			 */
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+				RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 1);
+			WREG32_SOC15_OFFSET(MMHUB, j, regVM_CONTEXT1_CNTL,
+					    i * hub->ctx_distance, tmp);
+			WREG32_SOC15_OFFSET(MMHUB, j,
+				regVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32,
+				i * hub->ctx_addr_distance, 0);
+			WREG32_SOC15_OFFSET(MMHUB, j,
+				regVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32,
+				i * hub->ctx_addr_distance, 0);
+			WREG32_SOC15_OFFSET(MMHUB, j,
+				regVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32,
+				i * hub->ctx_addr_distance,
+				lower_32_bits(adev->vm_manager.max_pfn - 1));
+			WREG32_SOC15_OFFSET(MMHUB, j,
+				regVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32,
+				i * hub->ctx_addr_distance,
+				upper_32_bits(adev->vm_manager.max_pfn - 1));
+		}
 	}
 }
 
 static void mmhub_v1_8_program_invalidation(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
-	unsigned i;
-
-	for (i = 0; i < 18; ++i) {
-		WREG32_SOC15_OFFSET(MMHUB, 0, regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
-				    i * hub->eng_addr_distance, 0xffffffff);
-		WREG32_SOC15_OFFSET(MMHUB, 0, regVM_INVALIDATE_ENG0_ADDR_RANGE_HI32,
-				    i * hub->eng_addr_distance, 0x1f);
+	struct amdgpu_vmhub *hub;
+	u32 i, j, inst_mask;
+
+	inst_mask = adev->aid_mask;
+	for_each_inst(j, inst_mask) {
+		hub = &adev->vmhub[AMDGPU_MMHUB0(j)];
+		for (i = 0; i < 18; ++i) {
+			WREG32_SOC15_OFFSET(MMHUB, j,
+					regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
+					i * hub->eng_addr_distance, 0xffffffff);
+			WREG32_SOC15_OFFSET(MMHUB, j,
+					regVM_INVALIDATE_ENG0_ADDR_RANGE_HI32,
+					i * hub->eng_addr_distance, 0x1f);
+		}
 	}
 }
 
 static int mmhub_v1_8_gart_enable(struct amdgpu_device *adev)
 {
-	if (amdgpu_sriov_vf(adev)) {
-		/*
-		 * MC_VM_FB_LOCATION_BASE/TOP is NULL for VF, becuase they are
-		 * VF copy registers so vbios post doesn't program them, for
-		 * SRIOV driver need to program them
-		 */
-		WREG32_SOC15(MMHUB, 0, regMC_VM_FB_LOCATION_BASE,
-			     adev->gmc.vram_start >> 24);
-		WREG32_SOC15(MMHUB, 0, regMC_VM_FB_LOCATION_TOP,
-			     adev->gmc.vram_end >> 24);
-	}
-
 	/* GART Enable. */
 	mmhub_v1_8_init_gart_aperture_regs(adev);
 	mmhub_v1_8_init_system_aperture_regs(adev);
 	mmhub_v1_8_init_tlb_regs(adev);
 	mmhub_v1_8_init_cache_regs(adev);
+	mmhub_v1_8_init_snoop_override_regs(adev);
 
 	mmhub_v1_8_enable_system_domain(adev);
 	mmhub_v1_8_disable_identity_aperture(adev);
@@ -350,31 +477,54 @@ static int mmhub_v1_8_gart_enable(struct amdgpu_device *adev)
 	return 0;
 }
 
+static void mmhub_v1_8_disable_l1_tlb(struct amdgpu_device *adev)
+{
+	u32 tmp;
+	u32 i, inst_mask;
+
+	if (amdgpu_sriov_reg_indirect_l1_tlb_cntl(adev)) {
+		tmp = RREG32_SOC15(MMHUB, 0, regMC_VM_MX_L1_TLB_CNTL);
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0);
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+				    ENABLE_ADVANCED_DRIVER_MODEL, 0);
+		psp_reg_program_no_ring(&adev->psp, tmp, PSP_REG_MMHUB_L1_TLB_CNTL);
+	} else {
+		inst_mask = adev->aid_mask;
+		for_each_inst(i, inst_mask) {
+			tmp = RREG32_SOC15(MMHUB, i, regMC_VM_MX_L1_TLB_CNTL);
+			tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB,
+					    0);
+			tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+					    ENABLE_ADVANCED_DRIVER_MODEL, 0);
+			WREG32_SOC15(MMHUB, i, regMC_VM_MX_L1_TLB_CNTL, tmp);
+		}
+	}
+}
+
 static void mmhub_v1_8_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub;
 	u32 tmp;
-	u32 i;
+	u32 i, j, inst_mask;
 
 	/* Disable all tables */
-	for (i = 0; i < 16; i++)
-		WREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT0_CNTL,
-				    i * hub->ctx_distance, 0);
-
-	/* Setup TLB control */
-	tmp = RREG32_SOC15(MMHUB, 0, regMC_VM_MX_L1_TLB_CNTL);
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0);
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
-			    ENABLE_ADVANCED_DRIVER_MODEL, 0);
-	WREG32_SOC15(MMHUB, 0, regMC_VM_MX_L1_TLB_CNTL, tmp);
-
-	if (!amdgpu_sriov_vf(adev)) {
-		/* Setup L2 cache */
-		tmp = RREG32_SOC15(MMHUB, 0, regVM_L2_CNTL);
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 0);
-		WREG32_SOC15(MMHUB, 0, regVM_L2_CNTL, tmp);
-		WREG32_SOC15(MMHUB, 0, regVM_L2_CNTL3, 0);
+	inst_mask = adev->aid_mask;
+	for_each_inst(j, inst_mask) {
+		hub = &adev->vmhub[AMDGPU_MMHUB0(j)];
+		for (i = 0; i < 16; i++)
+			WREG32_SOC15_OFFSET(MMHUB, j, regVM_CONTEXT0_CNTL,
+					    i * hub->ctx_distance, 0);
+		if (!amdgpu_sriov_vf(adev)) {
+			/* Setup L2 cache */
+			tmp = RREG32_SOC15(MMHUB, j, regVM_L2_CNTL);
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE,
+					    0);
+			WREG32_SOC15(MMHUB, j, regVM_L2_CNTL, tmp);
+			WREG32_SOC15(MMHUB, j, regVM_L2_CNTL3, 0);
+		}
 	}
+
+	mmhub_v1_8_disable_l1_tlb(adev);
 }
 
 /**
@@ -385,73 +535,83 @@ static void mmhub_v1_8_gart_disable(struct amdgpu_device *adev)
  */
 static void mmhub_v1_8_set_fault_enable_default(struct amdgpu_device *adev, bool value)
 {
-	u32 tmp;
+	u32 tmp, inst_mask;
+	int i;
 
 	if (amdgpu_sriov_vf(adev))
 		return;
 
-	tmp = RREG32_SOC15(MMHUB, 0, regVM_L2_PROTECTION_FAULT_CNTL);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			    RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			    PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			    PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			    PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			    TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT,
-			    value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			    NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			    DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			    VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			    READ_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			    WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			    EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	if (!value) {
+	inst_mask = adev->aid_mask;
+	for_each_inst(i, inst_mask) {
+		tmp = RREG32_SOC15(MMHUB, i, regVM_L2_PROTECTION_FAULT_CNTL);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+			TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT,
+			value);
 		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-				    CRASH_ON_NO_RETRY_FAULT, 1);
+				NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value);
 		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-				    CRASH_ON_RETRY_FAULT, 1);
+				DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				READ_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		if (!value) {
+			tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+					    CRASH_ON_NO_RETRY_FAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+					    CRASH_ON_RETRY_FAULT, 1);
+		}
+
+		WREG32_SOC15(MMHUB, i, regVM_L2_PROTECTION_FAULT_CNTL, tmp);
 	}
-
-	WREG32_SOC15(MMHUB, 0, regVM_L2_PROTECTION_FAULT_CNTL, tmp);
 }
 
 static void mmhub_v1_8_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
-
-	hub->ctx0_ptb_addr_lo32 =
-		SOC15_REG_OFFSET(MMHUB, 0,
-				 regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32);
-	hub->ctx0_ptb_addr_hi32 =
-		SOC15_REG_OFFSET(MMHUB, 0,
-				 regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32);
-	hub->vm_inv_eng0_req =
-		SOC15_REG_OFFSET(MMHUB, 0, regVM_INVALIDATE_ENG0_REQ);
-	hub->vm_inv_eng0_ack =
-		SOC15_REG_OFFSET(MMHUB, 0, regVM_INVALIDATE_ENG0_ACK);
-	hub->vm_context0_cntl =
-		SOC15_REG_OFFSET(MMHUB, 0, regVM_CONTEXT0_CNTL);
-	hub->vm_l2_pro_fault_status =
-		SOC15_REG_OFFSET(MMHUB, 0, regVM_L2_PROTECTION_FAULT_STATUS);
-	hub->vm_l2_pro_fault_cntl =
-		SOC15_REG_OFFSET(MMHUB, 0, regVM_L2_PROTECTION_FAULT_CNTL);
-
-	hub->ctx_distance = regVM_CONTEXT1_CNTL - regVM_CONTEXT0_CNTL;
-	hub->ctx_addr_distance = regVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 -
-		regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32;
-	hub->eng_distance = regVM_INVALIDATE_ENG1_REQ - regVM_INVALIDATE_ENG0_REQ;
-	hub->eng_addr_distance = regVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 -
-		regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32;
+	struct amdgpu_vmhub *hub;
+	u32 inst_mask;
+	int i;
 
+	inst_mask = adev->aid_mask;
+	for_each_inst(i, inst_mask) {
+		hub = &adev->vmhub[AMDGPU_MMHUB0(i)];
+
+		hub->ctx0_ptb_addr_lo32 = SOC15_REG_OFFSET(MMHUB, i,
+			regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32);
+		hub->ctx0_ptb_addr_hi32 = SOC15_REG_OFFSET(MMHUB, i,
+			regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32);
+		hub->vm_inv_eng0_req =
+			SOC15_REG_OFFSET(MMHUB, i, regVM_INVALIDATE_ENG0_REQ);
+		hub->vm_inv_eng0_ack =
+			SOC15_REG_OFFSET(MMHUB, i, regVM_INVALIDATE_ENG0_ACK);
+		hub->vm_context0_cntl =
+			SOC15_REG_OFFSET(MMHUB, i, regVM_CONTEXT0_CNTL);
+		hub->vm_l2_pro_fault_status = SOC15_REG_OFFSET(MMHUB, i,
+			regVM_L2_PROTECTION_FAULT_STATUS);
+		hub->vm_l2_pro_fault_cntl = SOC15_REG_OFFSET(MMHUB, i,
+			regVM_L2_PROTECTION_FAULT_CNTL);
+
+		hub->ctx_distance = regVM_CONTEXT1_CNTL - regVM_CONTEXT0_CNTL;
+		hub->ctx_addr_distance =
+			regVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 -
+			regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32;
+		hub->eng_distance = regVM_INVALIDATE_ENG1_REQ -
+			regVM_INVALIDATE_ENG0_REQ;
+		hub->eng_addr_distance = regVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 -
+			regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32;
+	}
 }
 
 static int mmhub_v1_8_set_clockgating(struct amdgpu_device *adev,
@@ -475,3 +635,237 @@ const struct amdgpu_mmhub_funcs mmhub_v1_8_funcs = {
 	.set_clockgating = mmhub_v1_8_set_clockgating,
 	.get_clockgating = mmhub_v1_8_get_clockgating,
 };
+
+static const struct amdgpu_ras_err_status_reg_entry mmhub_v1_8_ce_reg_list[] = {
+	{AMDGPU_RAS_REG_ENTRY(MMHUB, 0, regMMEA0_CE_ERR_STATUS_LO, regMMEA0_CE_ERR_STATUS_HI),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "MMEA0"},
+	{AMDGPU_RAS_REG_ENTRY(MMHUB, 0, regMMEA1_CE_ERR_STATUS_LO, regMMEA1_CE_ERR_STATUS_HI),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "MMEA1"},
+	{AMDGPU_RAS_REG_ENTRY(MMHUB, 0, regMMEA2_CE_ERR_STATUS_LO, regMMEA2_CE_ERR_STATUS_HI),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "MMEA2"},
+	{AMDGPU_RAS_REG_ENTRY(MMHUB, 0, regMMEA3_CE_ERR_STATUS_LO, regMMEA3_CE_ERR_STATUS_HI),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "MMEA3"},
+	{AMDGPU_RAS_REG_ENTRY(MMHUB, 0, regMMEA4_CE_ERR_STATUS_LO, regMMEA4_CE_ERR_STATUS_HI),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "MMEA4"},
+	{AMDGPU_RAS_REG_ENTRY(MMHUB, 0, regMM_CANE_CE_ERR_STATUS_LO, regMM_CANE_CE_ERR_STATUS_HI),
+	1, 0, "MM_CANE"},
+};
+
+static const struct amdgpu_ras_err_status_reg_entry mmhub_v1_8_ue_reg_list[] = {
+	{AMDGPU_RAS_REG_ENTRY(MMHUB, 0, regMMEA0_UE_ERR_STATUS_LO, regMMEA0_UE_ERR_STATUS_HI),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "MMEA0"},
+	{AMDGPU_RAS_REG_ENTRY(MMHUB, 0, regMMEA1_UE_ERR_STATUS_LO, regMMEA1_UE_ERR_STATUS_HI),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "MMEA1"},
+	{AMDGPU_RAS_REG_ENTRY(MMHUB, 0, regMMEA2_UE_ERR_STATUS_LO, regMMEA2_UE_ERR_STATUS_HI),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "MMEA2"},
+	{AMDGPU_RAS_REG_ENTRY(MMHUB, 0, regMMEA3_UE_ERR_STATUS_LO, regMMEA3_UE_ERR_STATUS_HI),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "MMEA3"},
+	{AMDGPU_RAS_REG_ENTRY(MMHUB, 0, regMMEA4_UE_ERR_STATUS_LO, regMMEA4_UE_ERR_STATUS_HI),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "MMEA4"},
+	{AMDGPU_RAS_REG_ENTRY(MMHUB, 0, regMM_CANE_UE_ERR_STATUS_LO, regMM_CANE_UE_ERR_STATUS_HI),
+	1, 0, "MM_CANE"},
+};
+
+static const struct amdgpu_ras_memory_id_entry mmhub_v1_8_ras_memory_list[] = {
+	{AMDGPU_MMHUB_WGMI_PAGEMEM, "MMEA_WGMI_PAGEMEM"},
+	{AMDGPU_MMHUB_RGMI_PAGEMEM, "MMEA_RGMI_PAGEMEM"},
+	{AMDGPU_MMHUB_WDRAM_PAGEMEM, "MMEA_WDRAM_PAGEMEM"},
+	{AMDGPU_MMHUB_RDRAM_PAGEMEM, "MMEA_RDRAM_PAGEMEM"},
+	{AMDGPU_MMHUB_WIO_CMDMEM, "MMEA_WIO_CMDMEM"},
+	{AMDGPU_MMHUB_RIO_CMDMEM, "MMEA_RIO_CMDMEM"},
+	{AMDGPU_MMHUB_WGMI_CMDMEM, "MMEA_WGMI_CMDMEM"},
+	{AMDGPU_MMHUB_RGMI_CMDMEM, "MMEA_RGMI_CMDMEM"},
+	{AMDGPU_MMHUB_WDRAM_CMDMEM, "MMEA_WDRAM_CMDMEM"},
+	{AMDGPU_MMHUB_RDRAM_CMDMEM, "MMEA_RDRAM_CMDMEM"},
+	{AMDGPU_MMHUB_MAM_DMEM0, "MMEA_MAM_DMEM0"},
+	{AMDGPU_MMHUB_MAM_DMEM1, "MMEA_MAM_DMEM1"},
+	{AMDGPU_MMHUB_MAM_DMEM2, "MMEA_MAM_DMEM2"},
+	{AMDGPU_MMHUB_MAM_DMEM3, "MMEA_MAM_DMEM3"},
+	{AMDGPU_MMHUB_WRET_TAGMEM, "MMEA_WRET_TAGMEM"},
+	{AMDGPU_MMHUB_RRET_TAGMEM, "MMEA_RRET_TAGMEM"},
+	{AMDGPU_MMHUB_WIO_DATAMEM, "MMEA_WIO_DATAMEM"},
+	{AMDGPU_MMHUB_WGMI_DATAMEM, "MMEA_WGMI_DATAMEM"},
+	{AMDGPU_MMHUB_WDRAM_DATAMEM, "MMEA_WDRAM_DATAMEM"},
+};
+
+static void mmhub_v1_8_inst_query_ras_error_count(struct amdgpu_device *adev,
+						  uint32_t mmhub_inst,
+						  void *ras_err_status)
+{
+	struct ras_err_data *err_data = (struct ras_err_data *)ras_err_status;
+	unsigned long ue_count = 0, ce_count = 0;
+
+	/* NOTE: mmhub is converted by aid_mask and the range is 0-3,
+	 * which can be used as die ID directly */
+	struct amdgpu_smuio_mcm_config_info mcm_info = {
+		.socket_id = adev->smuio.funcs->get_socket_id(adev),
+		.die_id = mmhub_inst,
+	};
+
+	amdgpu_ras_inst_query_ras_error_count(adev,
+					mmhub_v1_8_ce_reg_list,
+					ARRAY_SIZE(mmhub_v1_8_ce_reg_list),
+					mmhub_v1_8_ras_memory_list,
+					ARRAY_SIZE(mmhub_v1_8_ras_memory_list),
+					mmhub_inst,
+					AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE,
+					&ce_count);
+	amdgpu_ras_inst_query_ras_error_count(adev,
+					mmhub_v1_8_ue_reg_list,
+					ARRAY_SIZE(mmhub_v1_8_ue_reg_list),
+					mmhub_v1_8_ras_memory_list,
+					ARRAY_SIZE(mmhub_v1_8_ras_memory_list),
+					mmhub_inst,
+					AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE,
+					&ue_count);
+
+	amdgpu_ras_error_statistic_ce_count(err_data, &mcm_info, ce_count);
+	amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, ue_count);
+}
+
+static void mmhub_v1_8_query_ras_error_count(struct amdgpu_device *adev,
+					     void *ras_err_status)
+{
+	uint32_t inst_mask;
+	uint32_t i;
+
+	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__MMHUB)) {
+		dev_warn(adev->dev, "MMHUB RAS is not supported\n");
+		return;
+	}
+
+	inst_mask = adev->aid_mask;
+	for_each_inst(i, inst_mask)
+		mmhub_v1_8_inst_query_ras_error_count(adev, i, ras_err_status);
+}
+
+static void mmhub_v1_8_inst_reset_ras_error_count(struct amdgpu_device *adev,
+						  uint32_t mmhub_inst)
+{
+	amdgpu_ras_inst_reset_ras_error_count(adev,
+					mmhub_v1_8_ce_reg_list,
+					ARRAY_SIZE(mmhub_v1_8_ce_reg_list),
+					mmhub_inst);
+	amdgpu_ras_inst_reset_ras_error_count(adev,
+					mmhub_v1_8_ue_reg_list,
+					ARRAY_SIZE(mmhub_v1_8_ue_reg_list),
+					mmhub_inst);
+}
+
+static void mmhub_v1_8_reset_ras_error_count(struct amdgpu_device *adev)
+{
+	uint32_t inst_mask;
+	uint32_t i;
+
+	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__MMHUB)) {
+		dev_warn(adev->dev, "MMHUB RAS is not supported\n");
+		return;
+	}
+
+	inst_mask = adev->aid_mask;
+	for_each_inst(i, inst_mask)
+		mmhub_v1_8_inst_reset_ras_error_count(adev, i);
+}
+
+static const struct amdgpu_ras_block_hw_ops mmhub_v1_8_ras_hw_ops = {
+	.query_ras_error_count = mmhub_v1_8_query_ras_error_count,
+	.reset_ras_error_count = mmhub_v1_8_reset_ras_error_count,
+};
+
+static int mmhub_v1_8_aca_bank_parser(struct aca_handle *handle, struct aca_bank *bank,
+				      enum aca_smu_type type, void *data)
+{
+	struct aca_bank_info info;
+	u64 misc0;
+	int ret;
+
+	ret = aca_bank_info_decode(bank, &info);
+	if (ret)
+		return ret;
+
+	misc0 = bank->regs[ACA_REG_IDX_MISC0];
+	switch (type) {
+	case ACA_SMU_TYPE_UE:
+		bank->aca_err_type = ACA_ERROR_TYPE_UE;
+		ret = aca_error_cache_log_bank_error(handle, &info, ACA_ERROR_TYPE_UE,
+						     1ULL);
+		break;
+	case ACA_SMU_TYPE_CE:
+		bank->aca_err_type = ACA_ERROR_TYPE_CE;
+		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type,
+						     ACA_REG__MISC0__ERRCNT(misc0));
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return ret;
+}
+
+/* reference to smu driver if header file */
+static int mmhub_v1_8_err_codes[] = {
+	0, 1, 2, 3, 4, /* CODE_DAGB0 - 4 */
+	5, 6, 7, 8, 9, /* CODE_EA0 - 4 */
+	10, /* CODE_UTCL2_ROUTER */
+	11, /* CODE_VML2 */
+	12, /* CODE_VML2_WALKER */
+	13, /* CODE_MMCANE */
+};
+
+static bool mmhub_v1_8_aca_bank_is_valid(struct aca_handle *handle, struct aca_bank *bank,
+					 enum aca_smu_type type, void *data)
+{
+	u32 instlo;
+
+	instlo = ACA_REG__IPID__INSTANCEIDLO(bank->regs[ACA_REG_IDX_IPID]);
+	instlo &= GENMASK(31, 1);
+
+	if (instlo != mmSMNAID_AID0_MCA_SMU)
+		return false;
+
+	if (aca_bank_check_error_codes(handle->adev, bank,
+				       mmhub_v1_8_err_codes,
+				       ARRAY_SIZE(mmhub_v1_8_err_codes)))
+		return false;
+
+	return true;
+}
+
+static const struct aca_bank_ops mmhub_v1_8_aca_bank_ops = {
+	.aca_bank_parser = mmhub_v1_8_aca_bank_parser,
+	.aca_bank_is_valid = mmhub_v1_8_aca_bank_is_valid,
+};
+
+static const struct aca_info mmhub_v1_8_aca_info = {
+	.hwip = ACA_HWIP_TYPE_SMU,
+	.mask = ACA_ERROR_UE_MASK,
+	.bank_ops = &mmhub_v1_8_aca_bank_ops,
+};
+
+static int mmhub_v1_8_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
+{
+	int r;
+
+	r = amdgpu_ras_block_late_init(adev, ras_block);
+	if (r)
+		return r;
+
+	r = amdgpu_ras_bind_aca(adev, AMDGPU_RAS_BLOCK__MMHUB,
+				&mmhub_v1_8_aca_info, NULL);
+	if (r)
+		goto late_fini;
+
+	return 0;
+
+late_fini:
+	amdgpu_ras_block_late_fini(adev, ras_block);
+
+	return r;
+}
+
+struct amdgpu_mmhub_ras mmhub_v1_8_ras = {
+	.ras_block = {
+		.hw_ops = &mmhub_v1_8_ras_hw_ops,
+		.ras_late_init = mmhub_v1_8_ras_late_init,
+	},
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.h b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.h
index 0bb36200e4e5..126f0075ac50 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.h
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.h
@@ -24,5 +24,6 @@
 #define __MMHUB_V1_8_H__
 
 extern const struct amdgpu_mmhub_funcs mmhub_v1_8_funcs;
+extern struct amdgpu_mmhub_ras mmhub_v1_8_ras;
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c
index 278e32db878d..a0cc8e218ca1 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c
@@ -151,7 +151,7 @@ mmhub_v2_0_print_l2_protection_fault_status(struct amdgpu_device *adev,
 	dev_err(adev->dev,
 		"MMVM_L2_PROTECTION_FAULT_STATUS:0x%08X\n",
 		status);
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(2, 0, 0):
 	case IP_VERSION(2, 0, 2):
 		mmhub_cid = mmhub_client_ids_navi1x[cid][rw];
@@ -187,7 +187,7 @@ mmhub_v2_0_print_l2_protection_fault_status(struct amdgpu_device *adev,
 static void mmhub_v2_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
 				uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 
 	WREG32_SOC15_OFFSET_RLC(MMHUB, 0, mmMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
 			    hub->ctx_addr_distance * vmid,
@@ -362,12 +362,12 @@ static void mmhub_v2_0_disable_identity_aperture(struct amdgpu_device *adev)
 
 static void mmhub_v2_0_setup_vmid_config(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	int i;
 	uint32_t tmp;
 
 	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(MMHUB, 0, mmMMVM_CONTEXT1_CNTL, i);
+		tmp = RREG32_SOC15_OFFSET(MMHUB, 0, mmMMVM_CONTEXT1_CNTL, i * hub->ctx_distance);
 		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
 		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
 				    adev->vm_manager.num_level);
@@ -412,7 +412,7 @@ static void mmhub_v2_0_setup_vmid_config(struct amdgpu_device *adev)
 
 static void mmhub_v2_0_program_invalidation(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	unsigned i;
 
 	for (i = 0; i < 18; ++i) {
@@ -441,7 +441,7 @@ static int mmhub_v2_0_gart_enable(struct amdgpu_device *adev)
 
 static void mmhub_v2_0_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	u32 tmp;
 	u32 i;
 
@@ -520,7 +520,7 @@ static const struct amdgpu_vmhub_funcs mmhub_v2_0_vmhub_funcs = {
 
 static void mmhub_v2_0_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 
 	hub->ctx0_ptb_addr_lo32 =
 		SOC15_REG_OFFSET(MMHUB, 0,
@@ -568,7 +568,7 @@ static void mmhub_v2_0_update_medium_grain_clock_gating(struct amdgpu_device *ad
 	if (!(adev->cg_flags & AMD_CG_SUPPORT_MC_MGCG))
 		return;
 
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(2, 1, 0):
 	case IP_VERSION(2, 1, 1):
 	case IP_VERSION(2, 1, 2):
@@ -601,7 +601,7 @@ static void mmhub_v2_0_update_medium_grain_clock_gating(struct amdgpu_device *ad
 			  DAGB0_CNTL_MISC2__DISABLE_TLBRD_CG_MASK);
 	}
 
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(2, 1, 0):
 	case IP_VERSION(2, 1, 1):
 	case IP_VERSION(2, 1, 2):
@@ -625,7 +625,7 @@ static void mmhub_v2_0_update_medium_grain_light_sleep(struct amdgpu_device *ade
 	if (!(adev->cg_flags & AMD_CG_SUPPORT_MC_LS))
 		return;
 
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(2, 1, 0):
 	case IP_VERSION(2, 1, 1):
 	case IP_VERSION(2, 1, 2):
@@ -651,7 +651,7 @@ static int mmhub_v2_0_set_clockgating(struct amdgpu_device *adev,
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(2, 0, 0):
 	case IP_VERSION(2, 0, 2):
 	case IP_VERSION(2, 1, 0):
@@ -671,12 +671,12 @@ static int mmhub_v2_0_set_clockgating(struct amdgpu_device *adev,
 
 static void mmhub_v2_0_get_clockgating(struct amdgpu_device *adev, u64 *flags)
 {
-	int data, data1;
+	u32 data, data1;
 
 	if (amdgpu_sriov_vf(adev))
 		*flags = 0;
 
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(2, 1, 0):
 	case IP_VERSION(2, 1, 1):
 	case IP_VERSION(2, 1, 2):
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c
index fcf2813e70db..5eb8122e2746 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c
@@ -90,7 +90,7 @@ mmhub_v2_3_print_l2_protection_fault_status(struct amdgpu_device *adev,
 	dev_err(adev->dev,
 		"MMVM_L2_PROTECTION_FAULT_STATUS:0x%08X\n",
 		status);
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(2, 3, 0):
 	case IP_VERSION(2, 4, 0):
 	case IP_VERSION(2, 4, 1):
@@ -121,7 +121,7 @@ static void mmhub_v2_3_setup_vm_pt_regs(struct amdgpu_device *adev,
 					uint32_t vmid,
 					uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 
 	WREG32_SOC15_OFFSET(MMHUB, 0, mmMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
 			    hub->ctx_addr_distance * vmid, lower_32_bits(page_table_base));
@@ -280,12 +280,12 @@ static void mmhub_v2_3_disable_identity_aperture(struct amdgpu_device *adev)
 
 static void mmhub_v2_3_setup_vmid_config(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	int i;
 	uint32_t tmp;
 
 	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(MMHUB, 0, mmMMVM_CONTEXT1_CNTL, i);
+		tmp = RREG32_SOC15_OFFSET(MMHUB, 0, mmMMVM_CONTEXT1_CNTL, i * hub->ctx_distance);
 		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
 		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
 				    adev->vm_manager.num_level);
@@ -330,8 +330,8 @@ static void mmhub_v2_3_setup_vmid_config(struct amdgpu_device *adev)
 
 static void mmhub_v2_3_program_invalidation(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
-	unsigned i;
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
+	unsigned int i;
 
 	for (i = 0; i < 18; ++i) {
 		WREG32_SOC15_OFFSET(MMHUB, 0,
@@ -373,7 +373,7 @@ static int mmhub_v2_3_gart_enable(struct amdgpu_device *adev)
 
 static void mmhub_v2_3_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	u32 tmp;
 	u32 i;
 
@@ -406,6 +406,7 @@ static void mmhub_v2_3_set_fault_enable_default(struct amdgpu_device *adev,
 						bool value)
 {
 	u32 tmp;
+
 	tmp = RREG32_SOC15(MMHUB, 0, mmMMVM_L2_PROTECTION_FAULT_CNTL);
 	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
 			    RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
@@ -446,7 +447,7 @@ static const struct amdgpu_vmhub_funcs mmhub_v2_3_vmhub_funcs = {
 
 static void mmhub_v2_3_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 
 	hub->ctx0_ptb_addr_lo32 =
 		SOC15_REG_OFFSET(MMHUB, 0,
@@ -499,11 +500,11 @@ mmhub_v2_3_update_medium_grain_clock_gating(struct amdgpu_device *adev,
 	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_MC_MGCG)) {
 		data &= ~MM_ATC_L2_CGTT_CLK_CTRL__SOFT_OVERRIDE_MASK;
 		data1 &= ~(DAGB0_CNTL_MISC2__DISABLE_WRREQ_CG_MASK |
-		           DAGB0_CNTL_MISC2__DISABLE_WRRET_CG_MASK |
-		           DAGB0_CNTL_MISC2__DISABLE_RDREQ_CG_MASK |
-		           DAGB0_CNTL_MISC2__DISABLE_RDRET_CG_MASK |
-		           DAGB0_CNTL_MISC2__DISABLE_TLBWR_CG_MASK |
-		           DAGB0_CNTL_MISC2__DISABLE_TLBRD_CG_MASK);
+			   DAGB0_CNTL_MISC2__DISABLE_WRRET_CG_MASK |
+			   DAGB0_CNTL_MISC2__DISABLE_RDREQ_CG_MASK |
+			   DAGB0_CNTL_MISC2__DISABLE_RDRET_CG_MASK |
+			   DAGB0_CNTL_MISC2__DISABLE_TLBWR_CG_MASK |
+			   DAGB0_CNTL_MISC2__DISABLE_TLBRD_CG_MASK);
 
 	} else {
 		data |= MM_ATC_L2_CGTT_CLK_CTRL__SOFT_OVERRIDE_MASK;
@@ -593,13 +594,13 @@ static void mmhub_v2_3_get_clockgating(struct amdgpu_device *adev, u64 *flags)
 
 	/* AMD_CG_SUPPORT_MC_MGCG */
 	if (!(data & (DAGB0_CNTL_MISC2__DISABLE_WRREQ_CG_MASK |
-		       DAGB0_CNTL_MISC2__DISABLE_WRRET_CG_MASK |
-		       DAGB0_CNTL_MISC2__DISABLE_RDREQ_CG_MASK |
-		       DAGB0_CNTL_MISC2__DISABLE_RDRET_CG_MASK |
-		       DAGB0_CNTL_MISC2__DISABLE_TLBWR_CG_MASK |
-		       DAGB0_CNTL_MISC2__DISABLE_TLBRD_CG_MASK))
+			DAGB0_CNTL_MISC2__DISABLE_WRRET_CG_MASK |
+			DAGB0_CNTL_MISC2__DISABLE_RDREQ_CG_MASK |
+			DAGB0_CNTL_MISC2__DISABLE_RDRET_CG_MASK |
+			DAGB0_CNTL_MISC2__DISABLE_TLBWR_CG_MASK |
+			DAGB0_CNTL_MISC2__DISABLE_TLBRD_CG_MASK))
 		&& !(data1 & MM_ATC_L2_CGTT_CLK_CTRL__SOFT_OVERRIDE_MASK)) {
-			*flags |= AMD_CG_SUPPORT_MC_MGCG;
+		*flags |= AMD_CG_SUPPORT_MC_MGCG;
 	}
 
 	/* AMD_CG_SUPPORT_MC_LS */
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c
index 17a792616979..7d5242df58a5 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c
@@ -107,7 +107,7 @@ mmhub_v3_0_print_l2_protection_fault_status(struct amdgpu_device *adev,
 	dev_err(adev->dev,
 		"MMVM_L2_PROTECTION_FAULT_STATUS:0x%08X\n",
 		status);
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(3, 0, 0):
 	case IP_VERSION(3, 0, 1):
 		mmhub_cid = mmhub_client_ids_v3_0_0[cid][rw];
@@ -136,7 +136,7 @@ mmhub_v3_0_print_l2_protection_fault_status(struct amdgpu_device *adev,
 static void mmhub_v3_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
 				uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 
 	WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
 			    hub->ctx_addr_distance * vmid,
@@ -189,8 +189,7 @@ static void mmhub_v3_0_init_system_aperture_regs(struct amdgpu_device *adev)
 		     max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
 
 	/* Set default page address. */
-	value = adev->mem_scratch.gpu_addr - adev->gmc.vram_start +
-		adev->vm_manager.vram_base_offset;
+	value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
 	WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
 		     (u32)(value >> 12));
 	WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
@@ -319,12 +318,12 @@ static void mmhub_v3_0_disable_identity_aperture(struct amdgpu_device *adev)
 
 static void mmhub_v3_0_setup_vmid_config(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	int i;
 	uint32_t tmp;
 
 	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_CNTL, i);
+		tmp = RREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_CNTL, i * hub->ctx_distance);
 		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
 		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
 				    adev->vm_manager.num_level);
@@ -369,7 +368,7 @@ static void mmhub_v3_0_setup_vmid_config(struct amdgpu_device *adev)
 
 static void mmhub_v3_0_program_invalidation(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	unsigned i;
 
 	for (i = 0; i < 18; ++i) {
@@ -398,7 +397,7 @@ static int mmhub_v3_0_gart_enable(struct amdgpu_device *adev)
 
 static void mmhub_v3_0_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	u32 tmp;
 	u32 i;
 
@@ -477,7 +476,7 @@ static const struct amdgpu_vmhub_funcs mmhub_v3_0_vmhub_funcs = {
 
 static void mmhub_v3_0_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 
 	hub->ctx0_ptb_addr_lo32 =
 		SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c
index 26509b6b8c24..910337dc28d1 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c
@@ -36,40 +36,47 @@
 
 static const char *mmhub_client_ids_v3_0_1[][2] = {
 	[0][0] = "VMC",
+	[1][0] = "ISPXT",
+	[2][0] = "ISPIXT",
 	[4][0] = "DCEDMC",
 	[5][0] = "DCEVGA",
 	[6][0] = "MP0",
 	[7][0] = "MP1",
-	[8][0] = "MPIO",
-	[16][0] = "HDP",
-	[17][0] = "LSDMA",
-	[18][0] = "JPEG",
-	[19][0] = "VCNU0",
-	[21][0] = "VSCH",
-	[22][0] = "VCNU1",
-	[23][0] = "VCN1",
-	[32+20][0] = "VCN0",
-	[2][1] = "DBGUNBIO",
+	[8][0] = "MPM",
+	[12][0] = "ISPTNR",
+	[14][0] = "ISPCRD0",
+	[15][0] = "ISPCRD1",
+	[16][0] = "ISPCRD2",
+	[22][0] = "HDP",
+	[23][0] = "LSDMA",
+	[24][0] = "JPEG",
+	[27][0] = "VSCH",
+	[28][0] = "VCNU",
+	[29][0] = "VCN",
+	[1][1] = "ISPXT",
+	[2][1] = "ISPIXT",
 	[3][1] = "DCEDWB",
 	[4][1] = "DCEDMC",
 	[5][1] = "DCEVGA",
 	[6][1] = "MP0",
 	[7][1] = "MP1",
-	[8][1] = "MPIO",
-	[10][1] = "DBGU0",
-	[11][1] = "DBGU1",
-	[12][1] = "DBGU2",
-	[13][1] = "DBGU3",
-	[14][1] = "XDP",
-	[15][1] = "OSSSYS",
-	[16][1] = "HDP",
-	[17][1] = "LSDMA",
-	[18][1] = "JPEG",
-	[19][1] = "VCNU0",
-	[20][1] = "VCN0",
-	[21][1] = "VSCH",
-	[22][1] = "VCNU1",
-	[23][1] = "VCN1",
+	[8][1] = "MPM",
+	[10][1] = "ISPMWR0",
+	[11][1] = "ISPMWR1",
+	[12][1] = "ISPTNR",
+	[13][1] = "ISPSWR",
+	[14][1] = "ISPCWR0",
+	[15][1] = "ISPCWR1",
+	[16][1] = "ISPCWR2",
+	[17][1] = "ISPCWR3",
+	[18][1] = "XDP",
+	[21][1] = "OSSSYS",
+	[22][1] = "HDP",
+	[23][1] = "LSDMA",
+	[24][1] = "JPEG",
+	[27][1] = "VSCH",
+	[28][1] = "VCNU",
+	[29][1] = "VCN",
 };
 
 static uint32_t mmhub_v3_0_1_get_invalidate_req(unsigned int vmid,
@@ -108,7 +115,7 @@ mmhub_v3_0_1_print_l2_protection_fault_status(struct amdgpu_device *adev,
 		"MMVM_L2_PROTECTION_FAULT_STATUS:0x%08X\n",
 		status);
 
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(3, 0, 1):
 		mmhub_cid = mmhub_client_ids_v3_0_1[cid][rw];
 		break;
@@ -138,7 +145,7 @@ static void mmhub_v3_0_1_setup_vm_pt_regs(struct amdgpu_device *adev,
 					  uint32_t vmid,
 					  uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 
 	WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
 			    hub->ctx_addr_distance * vmid,
@@ -188,8 +195,7 @@ static void mmhub_v3_0_1_init_system_aperture_regs(struct amdgpu_device *adev)
 		     max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
 
 	/* Set default page address. */
-	value = adev->mem_scratch.gpu_addr - adev->gmc.vram_start +
-		adev->vm_manager.vram_base_offset;
+	value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
 	WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
 		     (u32)(value >> 12));
 	WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
@@ -306,12 +312,12 @@ static void mmhub_v3_0_1_disable_identity_aperture(struct amdgpu_device *adev)
 
 static void mmhub_v3_0_1_setup_vmid_config(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	int i;
 	uint32_t tmp;
 
 	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_CNTL, i);
+		tmp = RREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_CNTL, i * hub->ctx_distance);
 		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
 		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
 				    adev->vm_manager.num_level);
@@ -356,7 +362,7 @@ static void mmhub_v3_0_1_setup_vmid_config(struct amdgpu_device *adev)
 
 static void mmhub_v3_0_1_program_invalidation(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	unsigned i;
 
 	for (i = 0; i < 18; ++i) {
@@ -385,7 +391,7 @@ static int mmhub_v3_0_1_gart_enable(struct amdgpu_device *adev)
 
 static void mmhub_v3_0_1_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	u32 tmp;
 	u32 i;
 
@@ -459,7 +465,7 @@ static const struct amdgpu_vmhub_funcs mmhub_v3_0_1_vmhub_funcs = {
 
 static void mmhub_v3_0_1_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 
 	hub->ctx0_ptb_addr_lo32 =
 		SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c
index 26abbc6a47ab..f0f182f033b9 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c
@@ -129,7 +129,7 @@ mmhub_v3_0_2_print_l2_protection_fault_status(struct amdgpu_device *adev,
 static void mmhub_v3_0_2_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
 				uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 
 	WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
 			    hub->ctx_addr_distance * vmid,
@@ -181,8 +181,7 @@ static void mmhub_v3_0_2_init_system_aperture_regs(struct amdgpu_device *adev)
 	}
 
 	/* Set default page address. */
-	value = adev->mem_scratch.gpu_addr - adev->gmc.vram_start +
-		adev->vm_manager.vram_base_offset;
+	value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
 	WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
 		     (u32)(value >> 12));
 	WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
@@ -311,12 +310,12 @@ static void mmhub_v3_0_2_disable_identity_aperture(struct amdgpu_device *adev)
 
 static void mmhub_v3_0_2_setup_vmid_config(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	int i;
 	uint32_t tmp;
 
 	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_CNTL, i);
+		tmp = RREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_CNTL, i * hub->ctx_distance);
 		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
 		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
 				    adev->vm_manager.num_level);
@@ -361,7 +360,7 @@ static void mmhub_v3_0_2_setup_vmid_config(struct amdgpu_device *adev)
 
 static void mmhub_v3_0_2_program_invalidation(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	unsigned i;
 
 	for (i = 0; i < 18; ++i) {
@@ -390,7 +389,7 @@ static int mmhub_v3_0_2_gart_enable(struct amdgpu_device *adev)
 
 static void mmhub_v3_0_2_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	u32 tmp;
 	u32 i;
 
@@ -469,7 +468,7 @@ static const struct amdgpu_vmhub_funcs mmhub_v3_0_2_vmhub_funcs = {
 
 static void mmhub_v3_0_2_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 
 	hub->ctx0_ptb_addr_lo32 =
 		SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_3.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_3.c
new file mode 100644
index 000000000000..f6fc9778bc30
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_3.c
@@ -0,0 +1,746 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "amdgpu.h"
+#include "mmhub_v3_3.h"
+
+#include "mmhub/mmhub_3_3_0_offset.h"
+#include "mmhub/mmhub_3_3_0_sh_mask.h"
+
+#include "navi10_enum.h"
+#include "soc15_common.h"
+
+#define regMMVM_L2_CNTL3_DEFAULT				0x80100007
+#define regMMVM_L2_CNTL4_DEFAULT				0x000000c1
+#define regMMVM_L2_CNTL5_DEFAULT				0x00003fe0
+#define regDAGB0_L1TLB_REG_RW_3_3                   0x00a4
+#define regDAGB0_L1TLB_REG_RW_3_3_BASE_IDX          1
+#define regDAGB1_L1TLB_REG_RW_3_3                   0x0163
+#define regDAGB1_L1TLB_REG_RW_3_3_BASE_IDX          1
+
+static const char *mmhub_client_ids_v3_3[][2] = {
+	[0][0] = "VMC",
+	[1][0] = "ISPXT",
+	[2][0] = "ISPIXT",
+	[4][0] = "DCEDMC",
+	[6][0] = "MP0",
+	[7][0] = "MP1",
+	[8][0] = "MPM",
+	[9][0] = "ISPPDPRD",
+	[10][0] = "ISPCSTATRD",
+	[11][0] = "ISPBYRPRD",
+	[12][0] = "ISPRGBPRD",
+	[13][0] = "ISPMCFPRD",
+	[14][0] = "ISPMCFPRD1",
+	[15][0] = "ISPYUVPRD",
+	[16][0] = "ISPMCSCRD",
+	[17][0] = "ISPGDCRD",
+	[18][0] = "ISPLMERD",
+	[22][0] = "ISPXT1",
+	[23][0] = "ISPIXT1",
+	[24][0] = "HDP",
+	[25][0] = "LSDMA",
+	[26][0] = "JPEG",
+	[27][0] = "VPE",
+	[28][0] = "VSCH",
+	[29][0] = "VCNU",
+	[30][0] = "VCN",
+	[1][1] = "ISPXT",
+	[2][1] = "ISPIXT",
+	[3][1] = "DCEDWB",
+	[4][1] = "DCEDMC",
+	[5][1] = "ISPCSISWR",
+	[6][1] = "MP0",
+	[7][1] = "MP1",
+	[8][1] = "MPM",
+	[9][1] = "ISPPDPWR",
+	[10][1] = "ISPCSTATWR",
+	[11][1] = "ISPBYRPWR",
+	[12][1] = "ISPRGBPWR",
+	[13][1] = "ISPMCFPWR",
+	[14][1] = "ISPMWR0",
+	[15][1] = "ISPYUVPWR",
+	[16][1] = "ISPMCSCWR",
+	[17][1] = "ISPGDCWR",
+	[18][1] = "ISPLMEWR",
+	[20][1] = "ISPMWR2",
+	[21][1] = "OSSSYS",
+	[22][1] = "ISPXT1",
+	[23][1] = "ISPIXT1",
+	[24][1] = "HDP",
+	[25][1] = "LSDMA",
+	[26][1] = "JPEG",
+	[27][1] = "VPE",
+	[28][1] = "VSCH",
+	[29][1] = "VCNU",
+	[30][1] = "VCN",
+};
+
+static const char *mmhub_client_ids_v3_3_1[][2] = {
+	[0][0] = "VMC",
+	[4][0] = "DCEDMC",
+	[6][0] = "MP0",
+	[7][0] = "MP1",
+	[8][0] = "MPM",
+	[24][0] = "HDP",
+	[25][0] = "LSDMA",
+	[26][0] = "JPEG0",
+	[27][0] = "VPE0",
+	[28][0] = "VSCH",
+	[29][0] = "VCNU0",
+	[30][0] = "VCN0",
+	[32+1][0] = "ISPXT",
+	[32+2][0] = "ISPIXT",
+	[32+9][0] = "ISPPDPRD",
+	[32+10][0] = "ISPCSTATRD",
+	[32+11][0] = "ISPBYRPRD",
+	[32+12][0] = "ISPRGBPRD",
+	[32+13][0] = "ISPMCFPRD",
+	[32+14][0] = "ISPMCFPRD1",
+	[32+15][0] = "ISPYUVPRD",
+	[32+16][0] = "ISPMCSCRD",
+	[32+17][0] = "ISPGDCRD",
+	[32+18][0] = "ISPLMERD",
+	[32+22][0] = "ISPXT1",
+	[32+23][0] = "ISPIXT1",
+	[32+26][0] = "JPEG1",
+	[32+27][0] = "VPE1",
+	[32+29][0] = "VCNU1",
+	[32+30][0] = "VCN1",
+	[3][1] = "DCEDWB",
+	[4][1] = "DCEDMC",
+	[6][1] = "MP0",
+	[7][1] = "MP1",
+	[8][1] = "MPM",
+	[21][1] = "OSSSYS",
+	[24][1] = "HDP",
+	[25][1] = "LSDMA",
+	[26][1] = "JPEG0",
+	[27][1] = "VPE0",
+	[28][1] = "VSCH",
+	[29][1] = "VCNU0",
+	[30][1] = "VCN0",
+	[32+1][1] = "ISPXT",
+	[32+2][1] = "ISPIXT",
+	[32+5][1] = "ISPCSISWR",
+	[32+9][1] = "ISPPDPWR",
+	[32+10][1] = "ISPCSTATWR",
+	[32+11][1] = "ISPBYRPWR",
+	[32+12][1] = "ISPRGBPWR",
+	[32+13][1] = "ISPMCFPWR",
+	[32+14][1] = "ISPMWR0",
+	[32+15][1] = "ISPYUVPWR",
+	[32+16][1] = "ISPMCSCWR",
+	[32+17][1] = "ISPGDCWR",
+	[32+18][1] = "ISPLMEWR",
+	[32+19][1] = "ISPMWR1",
+	[32+20][1] = "ISPMWR2",
+	[32+22][1] = "ISPXT1",
+	[32+23][1] = "ISPIXT1",
+	[32+26][1] = "JPEG1",
+	[32+27][1] = "VPE1",
+	[32+29][1] = "VCNU1",
+	[32+30][1] = "VCN1",
+};
+
+static uint32_t mmhub_v3_3_get_invalidate_req(unsigned int vmid,
+						uint32_t flush_type)
+{
+	u32 req = 0;
+
+	/* invalidate using legacy mode on vmid*/
+	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ,
+			    PER_VMID_INVALIDATE_REQ, 1 << vmid);
+	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, flush_type ? : 1);
+	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PTES, 1);
+	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE0, 1);
+	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE1, 1);
+	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE2, 1);
+	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L1_PTES, 1);
+	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ,
+			    CLEAR_PROTECTION_FAULT_STATUS_ADDR,	0);
+
+	return req;
+}
+
+static void
+mmhub_v3_3_print_l2_protection_fault_status(struct amdgpu_device *adev,
+					      uint32_t status)
+{
+	uint32_t cid, rw;
+	const char *mmhub_cid = NULL;
+
+	cid = REG_GET_FIELD(status,
+			    MMVM_L2_PROTECTION_FAULT_STATUS, CID);
+	rw = REG_GET_FIELD(status,
+			   MMVM_L2_PROTECTION_FAULT_STATUS, RW);
+
+	dev_err(adev->dev,
+		"MMVM_L2_PROTECTION_FAULT_STATUS:0x%08X\n",
+		status);
+
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
+	case IP_VERSION(3, 3, 0):
+	case IP_VERSION(3, 3, 2):
+		mmhub_cid = cid < ARRAY_SIZE(mmhub_client_ids_v3_3) ?
+			    mmhub_client_ids_v3_3[cid][rw] :
+			    cid == 0x140 ? "UMSCH" : NULL;
+		break;
+	case IP_VERSION(3, 3, 1):
+		mmhub_cid = cid < ARRAY_SIZE(mmhub_client_ids_v3_3_1) ?
+			    mmhub_client_ids_v3_3_1[cid][rw] :
+			    cid == 0x140 ? "UMSCH" : NULL;
+		break;
+	default:
+		mmhub_cid = NULL;
+		break;
+	}
+
+	dev_err(adev->dev, "\t Faulty UTCL2 client ID: %s (0x%x)\n",
+		mmhub_cid ? mmhub_cid : "unknown", cid);
+	dev_err(adev->dev, "\t MORE_FAULTS: 0x%lx\n",
+		REG_GET_FIELD(status,
+		MMVM_L2_PROTECTION_FAULT_STATUS, MORE_FAULTS));
+	dev_err(adev->dev, "\t WALKER_ERROR: 0x%lx\n",
+		REG_GET_FIELD(status,
+		MMVM_L2_PROTECTION_FAULT_STATUS, WALKER_ERROR));
+	dev_err(adev->dev, "\t PERMISSION_FAULTS: 0x%lx\n",
+		REG_GET_FIELD(status,
+		MMVM_L2_PROTECTION_FAULT_STATUS, PERMISSION_FAULTS));
+	dev_err(adev->dev, "\t MAPPING_ERROR: 0x%lx\n",
+		REG_GET_FIELD(status,
+		MMVM_L2_PROTECTION_FAULT_STATUS, MAPPING_ERROR));
+	dev_err(adev->dev, "\t RW: 0x%x\n", rw);
+}
+
+static void mmhub_v3_3_setup_vm_pt_regs(struct amdgpu_device *adev,
+					  uint32_t vmid,
+					  uint64_t page_table_base)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
+
+	WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
+			    hub->ctx_addr_distance * vmid,
+			    lower_32_bits(page_table_base));
+
+	WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
+			    hub->ctx_addr_distance * vmid,
+			    upper_32_bits(page_table_base));
+
+}
+
+static void mmhub_v3_3_init_gart_aperture_regs(struct amdgpu_device *adev)
+{
+	uint64_t pt_base = amdgpu_gmc_pd_addr(adev->gart.bo);
+
+	mmhub_v3_3_setup_vm_pt_regs(adev, 0, pt_base);
+
+	WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
+		     (u32)(adev->gmc.gart_start >> 12));
+	WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
+		     (u32)(adev->gmc.gart_start >> 44));
+
+	WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
+		     (u32)(adev->gmc.gart_end >> 12));
+	WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
+		     (u32)(adev->gmc.gart_end >> 44));
+}
+
+static void mmhub_v3_3_init_system_aperture_regs(struct amdgpu_device *adev)
+{
+	uint64_t value;
+	uint32_t tmp;
+
+	/* Program the AGP BAR */
+	WREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_BASE, 0);
+	WREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
+	WREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
+
+	/*
+	 * the new L1 policy will block SRIOV guest from writing
+	 * these regs, and they will be programed at host.
+	 * so skip programing these regs.
+	 */
+	/* Program the system aperture low logical page number. */
+	WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_LOW_ADDR,
+		     min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18);
+	WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+		     max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
+
+	/* Set default page address. */
+	value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
+	WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
+		     (u32)(value >> 12));
+	WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
+		     (u32)(value >> 44));
+
+	/* Program "protection fault". */
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32,
+		     (u32)(adev->dummy_page_addr >> 12));
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32,
+		     (u32)((u64)adev->dummy_page_addr >> 44));
+
+	tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL2);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL2,
+			    ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1);
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL2, tmp);
+}
+
+static void mmhub_v3_3_init_tlb_regs(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	/* Setup TLB control */
+	tmp = RREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL);
+
+	tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1);
+	tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE, 3);
+	tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL,
+			    ENABLE_ADVANCED_DRIVER_MODEL, 1);
+	tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL,
+			    SYSTEM_APERTURE_UNMAPPED_ACCESS, 0);
+	tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, ECO_BITS, 0);
+	tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL,
+			    MTYPE, MTYPE_UC); /* UC, uncached */
+
+	WREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL, tmp);
+}
+
+static void mmhub_v3_3_init_cache_regs(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	/* Setup L2 cache */
+	tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, ENABLE_L2_CACHE, 1);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 0);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL,
+			    ENABLE_DEFAULT_PAGE_OUT_TO_SYSTEM_MEMORY, 1);
+	/* XXX for emulation, Refer to closed source code.*/
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, L2_PDE0_CACHE_TAG_GENERATION_MODE,
+			    0);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0);
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL, tmp);
+
+	tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL2);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL2, INVALIDATE_L2_CACHE, 1);
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL2, tmp);
+
+	tmp = regMMVM_L2_CNTL3_DEFAULT;
+	if (adev->gmc.translate_further) {
+		tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3, BANK_SELECT, 12);
+		tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3,
+				    L2_CACHE_BIGK_FRAGMENT_SIZE, 9);
+	} else {
+		tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3, BANK_SELECT, 9);
+		tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3,
+				    L2_CACHE_BIGK_FRAGMENT_SIZE, 6);
+	}
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL3, tmp);
+
+	tmp = regMMVM_L2_CNTL4_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 0);
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL4, tmp);
+
+	tmp = regMMVM_L2_CNTL5_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL5, L2_CACHE_SMALLK_FRAGMENT_SIZE, 0);
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL5, tmp);
+}
+
+static void mmhub_v3_3_enable_system_domain(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	tmp = RREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_CNTL);
+	tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
+	tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH, 0);
+	tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT0_CNTL,
+			    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0);
+
+	WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_CNTL, tmp);
+}
+
+static void mmhub_v3_3_disable_identity_aperture(struct amdgpu_device *adev)
+{
+	WREG32_SOC15(MMHUB, 0,
+		     regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32,
+		     0xFFFFFFFF);
+	WREG32_SOC15(MMHUB, 0,
+		     regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32,
+		     0x0000000F);
+
+	WREG32_SOC15(MMHUB, 0,
+		     regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32, 0);
+	WREG32_SOC15(MMHUB, 0,
+		     regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32, 0);
+
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32,
+		     0);
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32,
+		     0);
+}
+
+static void mmhub_v3_3_setup_vmid_config(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
+	int i;
+	uint32_t tmp;
+
+	for (i = 0; i <= 14; i++) {
+		tmp = RREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_CNTL, i * hub->ctx_distance);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
+				    adev->vm_manager.num_level);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL,
+				    RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL,
+				    DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT,
+				    1);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL,
+				    PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL,
+				    VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL,
+				    READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL,
+				    WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL,
+				    EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL,
+				    PAGE_TABLE_BLOCK_SIZE,
+				    adev->vm_manager.block_size - 9);
+		/* Send no-retry XNACK on fault to suppress VM fault storm. */
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL,
+				    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT,
+				    !amdgpu_noretry);
+		WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_CNTL,
+				    i * hub->ctx_distance, tmp);
+		WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32,
+				    i * hub->ctx_addr_distance, 0);
+		WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32,
+				    i * hub->ctx_addr_distance, 0);
+		WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32,
+				    i * hub->ctx_addr_distance,
+				    lower_32_bits(adev->vm_manager.max_pfn - 1));
+		WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32,
+				    i * hub->ctx_addr_distance,
+				    upper_32_bits(adev->vm_manager.max_pfn - 1));
+	}
+
+	hub->vm_cntx_cntl = tmp;
+}
+
+static void mmhub_v3_3_program_invalidation(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
+	unsigned int i;
+
+	for (i = 0; i < 18; ++i) {
+		WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
+				    i * hub->eng_addr_distance, 0xffffffff);
+		WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_ADDR_RANGE_HI32,
+				    i * hub->eng_addr_distance, 0x1f);
+	}
+}
+
+static void mmhub_v3_3_init_saw_regs(struct amdgpu_device *adev)
+{
+	uint64_t pt_base = amdgpu_gmc_pd_addr(adev->gart.bo);
+	uint32_t tmp;
+
+	/* Program page table base, gart start, gart end */
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_SAW_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
+			lower_32_bits(pt_base >> 12));
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_SAW_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
+			upper_32_bits(pt_base >> 12));
+
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_SAW_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
+		     (u32)(adev->gmc.gart_start >> 12));
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_SAW_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
+		     (u32)(adev->gmc.gart_start >> 44));
+
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_SAW_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
+		     (u32)(adev->gmc.gart_end >> 12));
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_SAW_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
+		     (u32)(adev->gmc.gart_end >> 44));
+
+	tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_SAW_CONTEXT0_CNTL);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_SAW_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_SAW_CONTEXT0_CNTL, PAGE_TABLE_DEPTH, 0);
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_SAW_CONTEXT0_CNTL, tmp);
+
+	/* Disable all contexts except context 0 */
+	tmp = 0xfffe;
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_SAW_CONTEXTS_DISABLE, tmp);
+
+	/* Program saw cntl4 */
+	tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_SAW_CNTL4);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_SAW_CNTL4, VMC_TAP_CONTEXT0_PDE_REQUEST_SNOOP, 1);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_SAW_CNTL4, VMC_TAP_CONTEXT0_PTE_REQUEST_SNOOP, 1);
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_SAW_CNTL4, tmp);
+}
+
+static void mmhub_v3_3_enable_tls(struct amdgpu_device *adev)
+{
+	WREG32_SOC15(MMHUB, 0, regDAGB0_L1TLB_REG_RW_3_3, 0);
+	WREG32_SOC15(MMHUB, 0, regDAGB1_L1TLB_REG_RW_3_3, 3);
+}
+
+static int mmhub_v3_3_gart_enable(struct amdgpu_device *adev)
+{
+	/* GART Enable. */
+	mmhub_v3_3_init_gart_aperture_regs(adev);
+	mmhub_v3_3_init_system_aperture_regs(adev);
+	mmhub_v3_3_init_tlb_regs(adev);
+	mmhub_v3_3_init_cache_regs(adev);
+
+	mmhub_v3_3_enable_system_domain(adev);
+	mmhub_v3_3_disable_identity_aperture(adev);
+	mmhub_v3_3_setup_vmid_config(adev);
+	mmhub_v3_3_program_invalidation(adev);
+
+	/* standalone alone walker init */
+	mmhub_v3_3_init_saw_regs(adev);
+
+	/* enable mmhub tls */
+	mmhub_v3_3_enable_tls(adev);
+
+	return 0;
+}
+
+static void mmhub_v3_3_gart_disable(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
+	u32 tmp;
+	u32 i;
+
+	/* Disable all tables */
+	for (i = 0; i < 16; i++)
+		WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_CNTL,
+				    i * hub->ctx_distance, 0);
+
+	/* Setup TLB control */
+	tmp = RREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL);
+	tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0);
+	tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL,
+			    ENABLE_ADVANCED_DRIVER_MODEL, 0);
+	WREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL, tmp);
+
+	/* Setup L2 cache */
+	tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, ENABLE_L2_CACHE, 0);
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL, tmp);
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL3, 0);
+}
+
+/**
+ * mmhub_v3_3_set_fault_enable_default - update GART/VM fault handling
+ *
+ * @adev: amdgpu_device pointer
+ * @value: true redirects VM faults to the default page
+ */
+static void mmhub_v3_3_set_fault_enable_default(struct amdgpu_device *adev,
+						  bool value)
+{
+	u32 tmp;
+
+	tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT,
+			    value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    READ_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	if (!value) {
+		tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_NO_RETRY_FAULT, 1);
+		tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_RETRY_FAULT, 1);
+	}
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL, tmp);
+}
+
+static const struct amdgpu_vmhub_funcs mmhub_v3_3_vmhub_funcs = {
+	.print_l2_protection_fault_status = mmhub_v3_3_print_l2_protection_fault_status,
+	.get_invalidate_req = mmhub_v3_3_get_invalidate_req,
+};
+
+static void mmhub_v3_3_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
+
+	hub->ctx0_ptb_addr_lo32 =
+		SOC15_REG_OFFSET(MMHUB, 0,
+				 regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32);
+	hub->ctx0_ptb_addr_hi32 =
+		SOC15_REG_OFFSET(MMHUB, 0,
+				 regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32);
+	hub->vm_inv_eng0_sem =
+		SOC15_REG_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_SEM);
+	hub->vm_inv_eng0_req =
+		SOC15_REG_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_REQ);
+	hub->vm_inv_eng0_ack =
+		SOC15_REG_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_ACK);
+	hub->vm_context0_cntl =
+		SOC15_REG_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_CNTL);
+	hub->vm_l2_pro_fault_status =
+		SOC15_REG_OFFSET(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_STATUS);
+	hub->vm_l2_pro_fault_cntl =
+		SOC15_REG_OFFSET(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL);
+
+	hub->ctx_distance = regMMVM_CONTEXT1_CNTL - regMMVM_CONTEXT0_CNTL;
+	hub->ctx_addr_distance = regMMVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 -
+		regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32;
+	hub->eng_distance = regMMVM_INVALIDATE_ENG1_REQ -
+		regMMVM_INVALIDATE_ENG0_REQ;
+	hub->eng_addr_distance = regMMVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 -
+		regMMVM_INVALIDATE_ENG0_ADDR_RANGE_LO32;
+
+	hub->vm_cntx_cntl_vm_fault = MMVM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		MMVM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		MMVM_CONTEXT1_CNTL__PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		MMVM_CONTEXT1_CNTL__VALID_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		MMVM_CONTEXT1_CNTL__READ_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		MMVM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		MMVM_CONTEXT1_CNTL__EXECUTE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK;
+
+	hub->vmhub_funcs = &mmhub_v3_3_vmhub_funcs;
+}
+
+static u64 mmhub_v3_3_get_fb_location(struct amdgpu_device *adev)
+{
+	u64 base;
+
+	base = RREG32_SOC15(MMHUB, 0, regMMMC_VM_FB_LOCATION_BASE);
+	base &= MMMC_VM_FB_LOCATION_BASE__FB_BASE_MASK;
+	base <<= 24;
+
+	return base;
+}
+
+static u64 mmhub_v3_3_get_mc_fb_offset(struct amdgpu_device *adev)
+{
+	u64 offset;
+
+	offset = RREG32_SOC15(MMHUB, 0, regMMMC_VM_FB_OFFSET);
+	offset &= MMMC_VM_FB_OFFSET__FB_OFFSET_MASK;
+	offset <<= 24;
+
+	return offset;
+}
+
+static void mmhub_v3_3_update_medium_grain_clock_gating(struct amdgpu_device *adev,
+							  bool enable)
+{
+	uint32_t def, data;
+
+	def = data = RREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG);
+
+	if (enable)
+		data |= MM_ATC_L2_MISC_CG__ENABLE_MASK;
+	else
+		data &= ~MM_ATC_L2_MISC_CG__ENABLE_MASK;
+
+	if (def != data)
+		WREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG, data);
+}
+
+static void mmhub_v3_3_update_medium_grain_light_sleep(struct amdgpu_device *adev,
+							 bool enable)
+{
+	uint32_t def, data;
+
+	def = data = RREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG);
+
+	if (enable)
+		data |= MM_ATC_L2_MISC_CG__MEM_LS_ENABLE_MASK;
+	else
+		data &= ~MM_ATC_L2_MISC_CG__MEM_LS_ENABLE_MASK;
+
+	if (def != data)
+		WREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG, data);
+}
+
+static int mmhub_v3_3_set_clockgating(struct amdgpu_device *adev,
+					enum amd_clockgating_state state)
+{
+	if (amdgpu_sriov_vf(adev))
+		return 0;
+
+	mmhub_v3_3_update_medium_grain_clock_gating(adev,
+			state == AMD_CG_STATE_GATE);
+	mmhub_v3_3_update_medium_grain_light_sleep(adev,
+			state == AMD_CG_STATE_GATE);
+	return 0;
+}
+
+static void mmhub_v3_3_get_clockgating(struct amdgpu_device *adev, u64 *flags)
+{
+	u32 data;
+
+	if (amdgpu_sriov_vf(adev))
+		*flags = 0;
+
+	data = RREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG);
+
+	/* AMD_CG_SUPPORT_MC_MGCG */
+	if (data & MM_ATC_L2_MISC_CG__ENABLE_MASK)
+		*flags |= AMD_CG_SUPPORT_MC_MGCG;
+
+	/* AMD_CG_SUPPORT_MC_LS */
+	if (data & MM_ATC_L2_MISC_CG__MEM_LS_ENABLE_MASK)
+		*flags |= AMD_CG_SUPPORT_MC_LS;
+}
+
+const struct amdgpu_mmhub_funcs mmhub_v3_3_funcs = {
+	.init = mmhub_v3_3_init,
+	.get_fb_location = mmhub_v3_3_get_fb_location,
+	.get_mc_fb_offset = mmhub_v3_3_get_mc_fb_offset,
+	.gart_enable = mmhub_v3_3_gart_enable,
+	.set_fault_enable_default = mmhub_v3_3_set_fault_enable_default,
+	.gart_disable = mmhub_v3_3_gart_disable,
+	.set_clockgating = mmhub_v3_3_set_clockgating,
+	.get_clockgating = mmhub_v3_3_get_clockgating,
+	.setup_vm_pt_regs = mmhub_v3_3_setup_vm_pt_regs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_3.h b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_3.h
new file mode 100644
index 000000000000..37b62c7e5a4a
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_3.h
@@ -0,0 +1,29 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __MMHUB_V3_3_H__
+#define __MMHUB_V3_3_H__
+
+extern const struct amdgpu_mmhub_funcs mmhub_v3_3_funcs;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v4_1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v4_1_0.c
new file mode 100644
index 000000000000..951998454b25
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v4_1_0.c
@@ -0,0 +1,647 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "amdgpu.h"
+#include "mmhub_v4_1_0.h"
+
+#include "mmhub/mmhub_4_1_0_offset.h"
+#include "mmhub/mmhub_4_1_0_sh_mask.h"
+
+#include "soc15_common.h"
+#include "soc24_enum.h"
+
+#define regMMVM_L2_CNTL3_DEFAULT				0x80100007
+#define regMMVM_L2_CNTL4_DEFAULT				0x000000c1
+#define regMMVM_L2_CNTL5_DEFAULT				0x00003fe0
+
+static const char *mmhub_client_ids_v4_1_0[][2] = {
+	[0][0] = "VMC",
+	[4][0] = "DCEDMC",
+	[6][0] = "MP0",
+	[7][0] = "MP1",
+	[8][0] = "MPIO",
+	[16][0] = "LSDMA",
+	[17][0] = "JPEG",
+	[19][0] = "VCNU",
+	[22][0] = "VSCH",
+	[23][0] = "HDP",
+	[32+23][0] = "VCNRD",
+	[3][1] = "DCEDWB",
+	[4][1] = "DCEDMC",
+	[6][1] = "MP0",
+	[7][1] = "MP1",
+	[8][1] = "MPIO",
+	[10][1] = "DBGU0",
+	[11][1] = "DBGU1",
+	[12][1] = "DBGUNBIO",
+	[14][1] = "XDP",
+	[15][1] = "OSSSYS",
+	[16][1] = "LSDMA",
+	[17][1] = "JPEG",
+	[18][1] = "VCNWR",
+	[19][1] = "VCNU",
+	[22][1] = "VSCH",
+	[23][1] = "HDP",
+};
+
+static uint32_t mmhub_v4_1_0_get_invalidate_req(unsigned int vmid,
+						uint32_t flush_type)
+{
+	u32 req = 0;
+
+	/* invalidate using legacy mode on vmid*/
+	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ,
+			    PER_VMID_INVALIDATE_REQ, 1 << vmid);
+	/* Only use legacy inv on mmhub side */
+	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, 0);
+	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PTES, 1);
+	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE0, 1);
+	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE1, 1);
+	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE2, 1);
+	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L1_PTES, 1);
+	req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ,
+			    CLEAR_PROTECTION_FAULT_STATUS_ADDR,	0);
+
+	return req;
+}
+
+static void
+mmhub_v4_1_0_print_l2_protection_fault_status(struct amdgpu_device *adev,
+					      uint32_t status)
+{
+	uint32_t cid, rw;
+	const char *mmhub_cid = NULL;
+
+	cid = REG_GET_FIELD(status,
+			    MMVM_L2_PROTECTION_FAULT_STATUS_LO32, CID);
+	rw = REG_GET_FIELD(status,
+			   MMVM_L2_PROTECTION_FAULT_STATUS_LO32, RW);
+
+	dev_err(adev->dev,
+		"MMVM_L2_PROTECTION_FAULT_STATUS_LO32:0x%08X\n",
+		status);
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
+	case IP_VERSION(4, 1, 0):
+		mmhub_cid = mmhub_client_ids_v4_1_0[cid][rw];
+		break;
+	default:
+		mmhub_cid = NULL;
+		break;
+	}
+	dev_err(adev->dev, "\t Faulty UTCL2 client ID: %s (0x%x)\n",
+		mmhub_cid ? mmhub_cid : "unknown", cid);
+	dev_err(adev->dev, "\t MORE_FAULTS: 0x%lx\n",
+		REG_GET_FIELD(status,
+		MMVM_L2_PROTECTION_FAULT_STATUS_LO32, MORE_FAULTS));
+	dev_err(adev->dev, "\t WALKER_ERROR: 0x%lx\n",
+		REG_GET_FIELD(status,
+		MMVM_L2_PROTECTION_FAULT_STATUS_LO32, WALKER_ERROR));
+	dev_err(adev->dev, "\t PERMISSION_FAULTS: 0x%lx\n",
+		REG_GET_FIELD(status,
+		MMVM_L2_PROTECTION_FAULT_STATUS_LO32, PERMISSION_FAULTS));
+	dev_err(adev->dev, "\t MAPPING_ERROR: 0x%lx\n",
+		REG_GET_FIELD(status,
+		MMVM_L2_PROTECTION_FAULT_STATUS_LO32, MAPPING_ERROR));
+	dev_err(adev->dev, "\t RW: 0x%x\n", rw);
+}
+
+static void mmhub_v4_1_0_setup_vm_pt_regs(struct amdgpu_device *adev,
+					  uint32_t vmid, uint64_t page_table_base)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
+
+	WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
+			    hub->ctx_addr_distance * vmid,
+			    lower_32_bits(page_table_base));
+
+	WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
+			    hub->ctx_addr_distance * vmid,
+			    upper_32_bits(page_table_base));
+}
+
+static void mmhub_v4_1_0_init_gart_aperture_regs(struct amdgpu_device *adev)
+{
+	uint64_t pt_base = amdgpu_gmc_pd_addr(adev->gart.bo);
+
+	mmhub_v4_1_0_setup_vm_pt_regs(adev, 0, pt_base);
+
+	WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
+		     (u32)(adev->gmc.gart_start >> 12));
+	WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
+		     (u32)(adev->gmc.gart_start >> 44));
+
+	WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
+		     (u32)(adev->gmc.gart_end >> 12));
+	WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
+		     (u32)(adev->gmc.gart_end >> 44));
+}
+
+static void mmhub_v4_1_0_init_system_aperture_regs(struct amdgpu_device *adev)
+{
+	uint64_t value;
+	uint32_t tmp;
+
+	/*
+	 * the new L1 policy will block SRIOV guest from writing
+	 * these regs, and they will be programed at host.
+	 * so skip programing these regs.
+	 */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	/* Program the AGP BAR */
+	WREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_BASE, 0);
+	WREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
+	WREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
+
+	/* Program the system aperture low logical page number. */
+	WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_LOW_ADDR,
+		     min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18);
+	WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+		     max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
+
+	/* Set default page address. */
+	value = adev->mem_scratch.gpu_addr - adev->gmc.vram_start +
+		adev->vm_manager.vram_base_offset;
+	WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
+		     (u32)(value >> 12));
+	WREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
+		     (u32)(value >> 44));
+
+	/* Program "protection fault". */
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32,
+		     (u32)(adev->dummy_page_addr >> 12));
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32,
+		     (u32)((u64)adev->dummy_page_addr >> 44));
+
+	tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL2);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL2,
+			    ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1);
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL2, tmp);
+}
+
+static void mmhub_v4_1_0_init_tlb_regs(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	/* Setup TLB control */
+	tmp = RREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL);
+
+	tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1);
+	tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE, 3);
+	tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL,
+			    ENABLE_ADVANCED_DRIVER_MODEL, 1);
+	tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL,
+			    SYSTEM_APERTURE_UNMAPPED_ACCESS, 0);
+	tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, ECO_BITS, 0);
+	tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL,
+			    MTYPE, MTYPE_UC); /* UC, uncached */
+
+	WREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL, tmp);
+}
+
+static void mmhub_v4_1_0_init_cache_regs(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	/* These registers are not accessible to VF-SRIOV.
+	 * The PF will program them instead.
+	 */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	/* Setup L2 cache */
+	tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, ENABLE_L2_CACHE, 1);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 0);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL,
+			    ENABLE_DEFAULT_PAGE_OUT_TO_SYSTEM_MEMORY, 1);
+	/* XXX for emulation, Refer to closed source code.*/
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, L2_PDE0_CACHE_TAG_GENERATION_MODE,
+			    0);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0);
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL, tmp);
+
+	tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL2);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL2, INVALIDATE_L2_CACHE, 1);
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL2, tmp);
+
+	tmp = regMMVM_L2_CNTL3_DEFAULT;
+	if (adev->gmc.translate_further) {
+		tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3, BANK_SELECT, 12);
+		tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3,
+				    L2_CACHE_BIGK_FRAGMENT_SIZE, 9);
+	} else {
+		tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3, BANK_SELECT, 9);
+		tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL3,
+				    L2_CACHE_BIGK_FRAGMENT_SIZE, 6);
+	}
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL3, tmp);
+
+	tmp = regMMVM_L2_CNTL4_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 0);
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL4, tmp);
+
+	tmp = regMMVM_L2_CNTL5_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL5, L2_CACHE_SMALLK_FRAGMENT_SIZE, 0);
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL5, tmp);
+}
+
+static void mmhub_v4_1_0_enable_system_domain(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	tmp = RREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_CNTL);
+	tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
+	tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH, 0);
+	tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT0_CNTL,
+			    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0);
+	WREG32_SOC15(MMHUB, 0, regMMVM_CONTEXT0_CNTL, tmp);
+}
+
+static void mmhub_v4_1_0_disable_identity_aperture(struct amdgpu_device *adev)
+{
+	/* These registers are not accessible to VF-SRIOV.
+	 * The PF will program them instead.
+	 */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	WREG32_SOC15(MMHUB, 0,
+		     regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32,
+		     0xFFFFFFFF);
+	WREG32_SOC15(MMHUB, 0,
+		     regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32,
+		     0x0000000F);
+
+	WREG32_SOC15(MMHUB, 0,
+		     regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32, 0);
+	WREG32_SOC15(MMHUB, 0,
+		     regMMVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32, 0);
+
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32,
+		     0);
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32,
+		     0);
+}
+
+static void mmhub_v4_1_0_setup_vmid_config(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
+	int i;
+	uint32_t tmp;
+
+	for (i = 0; i <= 14; i++) {
+		tmp = RREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_CNTL, i);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
+				    adev->vm_manager.num_level);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL,
+				    RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL,
+				    DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT,
+				    1);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL,
+				    PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL,
+				    VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL,
+				    READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL,
+				    WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL,
+				    EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL,
+				    PAGE_TABLE_BLOCK_SIZE,
+				    adev->vm_manager.block_size - 9);
+		/* Send no-retry XNACK on fault to suppress VM fault storm. */
+		tmp = REG_SET_FIELD(tmp, MMVM_CONTEXT1_CNTL,
+				    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT,
+				    !amdgpu_noretry);
+		WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_CNTL,
+				    i * hub->ctx_distance, tmp);
+		WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32,
+				    i * hub->ctx_addr_distance, 0);
+		WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32,
+				    i * hub->ctx_addr_distance, 0);
+		WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32,
+				    i * hub->ctx_addr_distance,
+				    lower_32_bits(adev->vm_manager.max_pfn - 1));
+		WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32,
+				    i * hub->ctx_addr_distance,
+				    upper_32_bits(adev->vm_manager.max_pfn - 1));
+	}
+
+	hub->vm_cntx_cntl = tmp;
+}
+
+static void mmhub_v4_1_0_program_invalidation(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
+	unsigned i;
+
+	for (i = 0; i < 18; ++i) {
+		WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
+				    i * hub->eng_addr_distance, 0xffffffff);
+		WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_ADDR_RANGE_HI32,
+				    i * hub->eng_addr_distance, 0x1f);
+	}
+}
+
+static int mmhub_v4_1_0_gart_enable(struct amdgpu_device *adev)
+{
+	/* GART Enable. */
+	mmhub_v4_1_0_init_gart_aperture_regs(adev);
+	mmhub_v4_1_0_init_system_aperture_regs(adev);
+	mmhub_v4_1_0_init_tlb_regs(adev);
+	mmhub_v4_1_0_init_cache_regs(adev);
+
+	mmhub_v4_1_0_enable_system_domain(adev);
+	mmhub_v4_1_0_disable_identity_aperture(adev);
+	mmhub_v4_1_0_setup_vmid_config(adev);
+	mmhub_v4_1_0_program_invalidation(adev);
+
+	return 0;
+}
+
+static void mmhub_v4_1_0_gart_disable(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
+	u32 tmp;
+	u32 i;
+
+	/* Disable all tables */
+	for (i = 0; i < 16; i++)
+		WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_CNTL,
+				    i * hub->ctx_distance, 0);
+
+	/* Setup TLB control */
+	tmp = RREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL);
+	tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0);
+	tmp = REG_SET_FIELD(tmp, MMMC_VM_MX_L1_TLB_CNTL,
+			    ENABLE_ADVANCED_DRIVER_MODEL, 0);
+	WREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL, tmp);
+
+	/* Setup L2 cache */
+	tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL, ENABLE_L2_CACHE, 0);
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL, tmp);
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL3, 0);
+}
+
+/**
+ * mmhub_v4_1_0_set_fault_enable_default - update GART/VM fault handling
+ *
+ * @adev: amdgpu_device pointer
+ * @value: true redirects VM faults to the default page
+ */
+static void
+mmhub_v4_1_0_set_fault_enable_default(struct amdgpu_device *adev, bool value)
+{
+	u32 tmp;
+
+	/* These registers are not accessible to VF-SRIOV.
+	 * The PF will program them instead.
+	 */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	tmp = RREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT,
+			    value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    READ_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+			    EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	if (!value) {
+		tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_NO_RETRY_FAULT, 1);
+		tmp = REG_SET_FIELD(tmp, MMVM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_RETRY_FAULT, 1);
+	}
+	WREG32_SOC15(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL, tmp);
+}
+
+static const struct amdgpu_vmhub_funcs mmhub_v4_1_0_vmhub_funcs = {
+	.print_l2_protection_fault_status = mmhub_v4_1_0_print_l2_protection_fault_status,
+	.get_invalidate_req = mmhub_v4_1_0_get_invalidate_req,
+};
+
+static void mmhub_v4_1_0_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
+
+	hub->ctx0_ptb_addr_lo32 =
+		SOC15_REG_OFFSET(MMHUB, 0,
+				 regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32);
+	hub->ctx0_ptb_addr_hi32 =
+		SOC15_REG_OFFSET(MMHUB, 0,
+				 regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32);
+	hub->vm_inv_eng0_sem =
+		SOC15_REG_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_SEM);
+	hub->vm_inv_eng0_req =
+		SOC15_REG_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_REQ);
+	hub->vm_inv_eng0_ack =
+		SOC15_REG_OFFSET(MMHUB, 0, regMMVM_INVALIDATE_ENG0_ACK);
+	hub->vm_context0_cntl =
+		SOC15_REG_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_CNTL);
+	hub->vm_l2_pro_fault_status =
+		SOC15_REG_OFFSET(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_STATUS_LO32);
+	hub->vm_l2_pro_fault_cntl =
+		SOC15_REG_OFFSET(MMHUB, 0, regMMVM_L2_PROTECTION_FAULT_CNTL);
+
+	hub->ctx_distance = regMMVM_CONTEXT1_CNTL - regMMVM_CONTEXT0_CNTL;
+	hub->ctx_addr_distance = regMMVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 -
+		regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32;
+	hub->eng_distance = regMMVM_INVALIDATE_ENG1_REQ -
+		regMMVM_INVALIDATE_ENG0_REQ;
+	hub->eng_addr_distance = regMMVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 -
+		regMMVM_INVALIDATE_ENG0_ADDR_RANGE_LO32;
+
+	hub->vm_cntx_cntl_vm_fault = MMVM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		MMVM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		MMVM_CONTEXT1_CNTL__PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		MMVM_CONTEXT1_CNTL__VALID_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		MMVM_CONTEXT1_CNTL__READ_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		MMVM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		MMVM_CONTEXT1_CNTL__EXECUTE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK;
+
+	hub->vm_l2_bank_select_reserved_cid2 =
+		SOC15_REG_OFFSET(MMHUB, 0, regMMVM_L2_BANK_SELECT_RESERVED_CID2);
+
+	hub->vm_contexts_disable =
+		SOC15_REG_OFFSET(MMHUB, 0, regMMVM_CONTEXTS_DISABLE);
+
+	hub->vmhub_funcs = &mmhub_v4_1_0_vmhub_funcs;
+}
+
+static u64 mmhub_v4_1_0_get_fb_location(struct amdgpu_device *adev)
+{
+	u64 base;
+
+	base = RREG32_SOC15(MMHUB, 0, regMMMC_VM_FB_LOCATION_BASE);
+
+	base &= MMMC_VM_FB_LOCATION_BASE__FB_BASE_MASK;
+	base <<= 24;
+
+	return base;
+}
+
+static u64 mmhub_v4_1_0_get_mc_fb_offset(struct amdgpu_device *adev)
+{
+	return (u64)RREG32_SOC15(MMHUB, 0, regMMMC_VM_FB_OFFSET) << 24;
+}
+
+static void
+mmhub_v4_1_0_update_medium_grain_clock_gating(struct amdgpu_device *adev,
+					      bool enable)
+{
+#if 0
+	uint32_t def, data;
+#endif
+	uint32_t def1, data1, def2 = 0, data2 = 0;
+#if 0
+	def  = data  = RREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG);
+#endif
+	def1 = data1 = RREG32_SOC15(MMHUB, 0, regDAGB0_CNTL_MISC2);
+	def2 = data2 = RREG32_SOC15(MMHUB, 0, regDAGB1_CNTL_MISC2);
+
+	if (enable) {
+#if 0
+		data |= MM_ATC_L2_MISC_CG__ENABLE_MASK;
+#endif
+		data1 &= ~(DAGB0_CNTL_MISC2__DISABLE_RDRET_TAP_CHAIN_FGCG_MASK |
+			   DAGB0_CNTL_MISC2__DISABLE_WRRET_TAP_CHAIN_FGCG_MASK);
+
+		data2 &= ~(DAGB1_CNTL_MISC2__DISABLE_RDRET_TAP_CHAIN_FGCG_MASK |
+			   DAGB1_CNTL_MISC2__DISABLE_WRRET_TAP_CHAIN_FGCG_MASK);
+	} else {
+#if 0
+		data &= ~MM_ATC_L2_MISC_CG__ENABLE_MASK;
+#endif
+		data1 |= (DAGB0_CNTL_MISC2__DISABLE_RDRET_TAP_CHAIN_FGCG_MASK |
+			  DAGB0_CNTL_MISC2__DISABLE_WRRET_TAP_CHAIN_FGCG_MASK);
+
+		data2 |= (DAGB1_CNTL_MISC2__DISABLE_RDRET_TAP_CHAIN_FGCG_MASK |
+			  DAGB1_CNTL_MISC2__DISABLE_WRRET_TAP_CHAIN_FGCG_MASK);
+	}
+
+#if 0
+	if (def != data)
+		WREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG, data);
+#endif
+	if (def1 != data1)
+		WREG32_SOC15(MMHUB, 0, regDAGB0_CNTL_MISC2, data1);
+
+	if (def2 != data2)
+		WREG32_SOC15(MMHUB, 0, regDAGB1_CNTL_MISC2, data2);
+}
+
+static void
+mmhub_v4_1_0_update_medium_grain_light_sleep(struct amdgpu_device *adev,
+					     bool enable)
+{
+#if 0
+	uint32_t def, data;
+
+	def = data = RREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG);
+
+	if (enable)
+		data |= MM_ATC_L2_MISC_CG__MEM_LS_ENABLE_MASK;
+	else
+		data &= ~MM_ATC_L2_MISC_CG__MEM_LS_ENABLE_MASK;
+
+	if (def != data)
+		WREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG, data);
+#endif
+}
+
+static int mmhub_v4_1_0_set_clockgating(struct amdgpu_device *adev,
+					enum amd_clockgating_state state)
+{
+	if (amdgpu_sriov_vf(adev))
+		return 0;
+
+	if (adev->cg_flags & AMD_CG_SUPPORT_MC_MGCG)
+		mmhub_v4_1_0_update_medium_grain_clock_gating(adev,
+				state == AMD_CG_STATE_GATE);
+
+	if (adev->cg_flags & AMD_CG_SUPPORT_MC_LS)
+		mmhub_v4_1_0_update_medium_grain_light_sleep(adev,
+				state == AMD_CG_STATE_GATE);
+
+	return 0;
+}
+
+static void mmhub_v4_1_0_get_clockgating(struct amdgpu_device *adev, u64 *flags)
+{
+#if 0
+	int data;
+
+	if (amdgpu_sriov_vf(adev))
+		*flags = 0;
+
+	data = RREG32_SOC15(MMHUB, 0, regMM_ATC_L2_MISC_CG);
+
+	/* AMD_CG_SUPPORT_MC_MGCG */
+	if (data & MM_ATC_L2_MISC_CG__ENABLE_MASK)
+		*flags |= AMD_CG_SUPPORT_MC_MGCG;
+
+	/* AMD_CG_SUPPORT_MC_LS */
+	if (data & MM_ATC_L2_MISC_CG__MEM_LS_ENABLE_MASK)
+		*flags |= AMD_CG_SUPPORT_MC_LS;
+#endif
+}
+
+const struct amdgpu_mmhub_funcs mmhub_v4_1_0_funcs = {
+	.init = mmhub_v4_1_0_init,
+	.get_fb_location = mmhub_v4_1_0_get_fb_location,
+	.get_mc_fb_offset = mmhub_v4_1_0_get_mc_fb_offset,
+	.gart_enable = mmhub_v4_1_0_gart_enable,
+	.set_fault_enable_default = mmhub_v4_1_0_set_fault_enable_default,
+	.gart_disable = mmhub_v4_1_0_gart_disable,
+	.set_clockgating = mmhub_v4_1_0_set_clockgating,
+	.get_clockgating = mmhub_v4_1_0_get_clockgating,
+	.setup_vm_pt_regs = mmhub_v4_1_0_setup_vm_pt_regs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v4_1_0.h b/drivers/gpu/drm/amd/amdgpu/mmhub_v4_1_0.h
new file mode 100644
index 000000000000..3902d653353c
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v4_1_0.h
@@ -0,0 +1,28 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __MMHUB_V4_1_0_H__
+#define __MMHUB_V4_1_0_H__
+
+extern const struct amdgpu_mmhub_funcs mmhub_v4_1_0_funcs;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c
index 72083e96222f..fe0710b55c3a 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c
@@ -57,7 +57,7 @@ static u64 mmhub_v9_4_get_fb_location(struct amdgpu_device *adev)
 static void mmhub_v9_4_setup_hubid_vm_pt_regs(struct amdgpu_device *adev, int hubid,
 				uint32_t vmid, uint64_t value)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 
 	WREG32_SOC15_OFFSET(MMHUB, 0,
 			    mmVML2VC0_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
@@ -108,7 +108,7 @@ static void mmhub_v9_4_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmi
 }
 
 static void mmhub_v9_4_init_system_aperture_regs(struct amdgpu_device *adev,
-					         int hubid)
+						int hubid)
 {
 	uint64_t value;
 	uint32_t tmp;
@@ -198,6 +198,36 @@ static void mmhub_v9_4_init_tlb_regs(struct amdgpu_device *adev, int hubid)
 			    hubid * MMHUB_INSTANCE_REGISTER_OFFSET, tmp);
 }
 
+/* Set snoop bit for SDMA so that SDMA writes probe-invalidates RW lines */
+static void mmhub_v9_4_init_snoop_override_regs(struct amdgpu_device *adev, int hubid)
+{
+	uint32_t tmp;
+	int i;
+	uint32_t distance = mmDAGB1_WRCLI_GPU_SNOOP_OVERRIDE -
+			    mmDAGB0_WRCLI_GPU_SNOOP_OVERRIDE;
+	uint32_t huboffset = hubid * MMHUB_INSTANCE_REGISTER_OFFSET;
+
+	for (i = 0; i < 5 - (2 * hubid); i++) {
+		/* DAGB instances 0 to 4 are in hub0 and 5 to 7 are in hub1 */
+		tmp = RREG32_SOC15_OFFSET(MMHUB, 0,
+			mmDAGB0_WRCLI_GPU_SNOOP_OVERRIDE,
+			huboffset + i * distance);
+		tmp |= (1 << 15); /* SDMA client is BIT15 */
+		WREG32_SOC15_OFFSET(MMHUB, 0,
+			mmDAGB0_WRCLI_GPU_SNOOP_OVERRIDE,
+			huboffset + i * distance, tmp);
+
+		tmp = RREG32_SOC15_OFFSET(MMHUB, 0,
+			mmDAGB0_WRCLI_GPU_SNOOP_OVERRIDE_VALUE,
+			huboffset + i * distance);
+		tmp |= (1 << 15);
+		WREG32_SOC15_OFFSET(MMHUB, 0,
+			mmDAGB0_WRCLI_GPU_SNOOP_OVERRIDE_VALUE,
+			huboffset + i * distance, tmp);
+	}
+
+}
+
 static void mmhub_v9_4_init_cache_regs(struct amdgpu_device *adev, int hubid)
 {
 	uint32_t tmp;
@@ -294,7 +324,7 @@ static void mmhub_v9_4_disable_identity_aperture(struct amdgpu_device *adev,
 
 static void mmhub_v9_4_setup_vmid_config(struct amdgpu_device *adev, int hubid)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	unsigned int num_level, block_size;
 	uint32_t tmp;
 	int i;
@@ -308,7 +338,7 @@ static void mmhub_v9_4_setup_vmid_config(struct amdgpu_device *adev, int hubid)
 
 	for (i = 0; i <= 14; i++) {
 		tmp = RREG32_SOC15_OFFSET(MMHUB, 0, mmVML2VC0_VM_CONTEXT1_CNTL,
-				hubid * MMHUB_INSTANCE_REGISTER_OFFSET + i);
+				hubid * MMHUB_INSTANCE_REGISTER_OFFSET + i * hub->ctx_distance);
 		tmp = REG_SET_FIELD(tmp, VML2VC0_VM_CONTEXT1_CNTL,
 				    ENABLE_CONTEXT, 1);
 		tmp = REG_SET_FIELD(tmp, VML2VC0_VM_CONTEXT1_CNTL,
@@ -363,7 +393,7 @@ static void mmhub_v9_4_setup_vmid_config(struct amdgpu_device *adev, int hubid)
 static void mmhub_v9_4_program_invalidation(struct amdgpu_device *adev,
 					    int hubid)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	unsigned i;
 
 	for (i = 0; i < 18; ++i) {
@@ -392,6 +422,7 @@ static int mmhub_v9_4_gart_enable(struct amdgpu_device *adev)
 		if (!amdgpu_sriov_vf(adev))
 			mmhub_v9_4_init_cache_regs(adev, i);
 
+		mmhub_v9_4_init_snoop_override_regs(adev, i);
 		mmhub_v9_4_enable_system_domain(adev, i);
 		if (!amdgpu_sriov_vf(adev))
 			mmhub_v9_4_disable_identity_aperture(adev, i);
@@ -404,7 +435,7 @@ static int mmhub_v9_4_gart_enable(struct amdgpu_device *adev)
 
 static void mmhub_v9_4_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 	u32 tmp;
 	u32 i, j;
 
@@ -507,8 +538,8 @@ static void mmhub_v9_4_set_fault_enable_default(struct amdgpu_device *adev, bool
 
 static void mmhub_v9_4_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub[MMHUB_NUM_INSTANCES] =
-		{&adev->vmhub[AMDGPU_MMHUB_0], &adev->vmhub[AMDGPU_MMHUB_1]};
+	struct amdgpu_vmhub *hub[MMHUB_NUM_INSTANCES] = {
+		&adev->vmhub[AMDGPU_MMHUB0(0)], &adev->vmhub[AMDGPU_MMHUB1(0)]};
 	int i;
 
 	for (i = 0; i < MMHUB_NUM_INSTANCES; i++) {
@@ -657,7 +688,7 @@ static int mmhub_v9_4_set_clockgating(struct amdgpu_device *adev,
 
 static void mmhub_v9_4_get_clockgating(struct amdgpu_device *adev, u64 *flags)
 {
-	int data, data1;
+	u32 data, data1;
 
 	if (amdgpu_sriov_vf(adev))
 		*flags = 0;
@@ -1568,7 +1599,7 @@ static int mmhub_v9_4_get_ras_error_count(struct amdgpu_device *adev,
 	uint32_t sec_cnt, ded_cnt;
 
 	for (i = 0; i < ARRAY_SIZE(mmhub_v9_4_ras_fields); i++) {
-		if(mmhub_v9_4_ras_fields[i].reg_offset != reg->reg_offset)
+		if (mmhub_v9_4_ras_fields[i].reg_offset != reg->reg_offset)
 			continue;
 
 		sec_cnt = (value &
diff --git a/drivers/gpu/drm/amd/amdgpu/mmsch_v3_0.h b/drivers/gpu/drm/amd/amdgpu/mmsch_v3_0.h
index 3e4e858a6965..a773ef61b78c 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmsch_v3_0.h
+++ b/drivers/gpu/drm/amd/amdgpu/mmsch_v3_0.h
@@ -30,6 +30,8 @@
 #define MMSCH_VERSION_MINOR	0
 #define MMSCH_VERSION	(MMSCH_VERSION_MAJOR << 16 | MMSCH_VERSION_MINOR)
 
+#define MMSCH_V3_0_VCN_INSTANCES 0x2
+
 enum mmsch_v3_0_command_type {
 	MMSCH_COMMAND__DIRECT_REG_WRITE = 0,
 	MMSCH_COMMAND__DIRECT_REG_POLLING = 2,
@@ -47,7 +49,7 @@ struct mmsch_v3_0_table_info {
 struct mmsch_v3_0_init_header {
 	uint32_t version;
 	uint32_t total_size;
-	struct mmsch_v3_0_table_info inst[AMDGPU_MAX_VCN_INSTANCES];
+	struct mmsch_v3_0_table_info inst[MMSCH_V3_0_VCN_INSTANCES];
 };
 
 struct mmsch_v3_0_cmd_direct_reg_header {
diff --git a/drivers/gpu/drm/amd/amdgpu/mmsch_v4_0.h b/drivers/gpu/drm/amd/amdgpu/mmsch_v4_0.h
index 83653a50a1a2..ced26cc5123a 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmsch_v4_0.h
+++ b/drivers/gpu/drm/amd/amdgpu/mmsch_v4_0.h
@@ -35,13 +35,13 @@
 
 #define MMSCH_VF_ENGINE_STATUS__PASS 0x1
 
-#define MMSCH_VF_MAILBOX_RESP__OK 0x1
-#define MMSCH_VF_MAILBOX_RESP__INCOMPLETE 0x2
+#define MMSCH_VF_MAILBOX_RESP__OK                      0x1
+#define MMSCH_VF_MAILBOX_RESP__INCOMPLETE              0x2
+#define MMSCH_VF_MAILBOX_RESP__FAILED                  0x3
+#define MMSCH_VF_MAILBOX_RESP__FAILED_SMALL_CTX_SIZE   0x4
+#define MMSCH_VF_MAILBOX_RESP__UNKNOWN_CMD             0x5
 
-#define MMSCH_VF_ENGINE_STATUS__PASS 0x1
-
-#define MMSCH_VF_MAILBOX_RESP__OK 0x1
-#define MMSCH_VF_MAILBOX_RESP__INCOMPLETE 0x2
+#define MMSCH_V4_0_VCN_INSTANCES 0x2
 
 enum mmsch_v4_0_command_type {
 	MMSCH_COMMAND__DIRECT_REG_WRITE = 0,
@@ -60,7 +60,7 @@ struct mmsch_v4_0_table_info {
 struct mmsch_v4_0_init_header {
 	uint32_t version;
 	uint32_t total_size;
-	struct mmsch_v4_0_table_info inst[AMDGPU_MAX_VCN_INSTANCES];
+	struct mmsch_v4_0_table_info inst[MMSCH_V4_0_VCN_INSTANCES];
 	struct mmsch_v4_0_table_info jpegdec;
 };
 
diff --git a/drivers/gpu/drm/amd/amdgpu/mmsch_v4_0_3.h b/drivers/gpu/drm/amd/amdgpu/mmsch_v4_0_3.h
new file mode 100644
index 000000000000..db7eb5260295
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/mmsch_v4_0_3.h
@@ -0,0 +1,37 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __MMSCH_V4_0_3_H__
+#define __MMSCH_V4_0_3_H__
+
+#include "amdgpu_vcn.h"
+#include "mmsch_v4_0.h"
+
+struct mmsch_v4_0_3_init_header {
+	uint32_t version;
+	uint32_t total_size;
+	struct mmsch_v4_0_table_info vcn0;
+	struct mmsch_v4_0_table_info mjpegdec0[4];
+	struct mmsch_v4_0_table_info mjpegdec1[4];
+};
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/mmsch_v5_0.h b/drivers/gpu/drm/amd/amdgpu/mmsch_v5_0.h
new file mode 100644
index 000000000000..6f749814929f
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/mmsch_v5_0.h
@@ -0,0 +1,144 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __MMSCH_V5_0_H__
+#define __MMSCH_V5_0_H__
+
+#include "amdgpu_vcn.h"
+
+#define MMSCH_VERSION_MAJOR	5
+#define MMSCH_VERSION_MINOR	0
+#define MMSCH_VERSION	(MMSCH_VERSION_MAJOR << 16 | MMSCH_VERSION_MINOR)
+
+#define RB_ENABLED (1 << 0)
+#define RB4_ENABLED (1 << 1)
+
+#define MMSCH_VF_ENGINE_STATUS__PASS 0x1
+
+#define MMSCH_VF_MAILBOX_RESP__OK                      0x1
+#define MMSCH_VF_MAILBOX_RESP__INCOMPLETE              0x2
+#define MMSCH_VF_MAILBOX_RESP__FAILED                  0x3
+#define MMSCH_VF_MAILBOX_RESP__FAILED_SMALL_CTX_SIZE   0x4
+#define MMSCH_VF_MAILBOX_RESP__UNKNOWN_CMD             0x5
+
+enum mmsch_v5_0_command_type {
+	MMSCH_COMMAND__DIRECT_REG_WRITE = 0,
+	MMSCH_COMMAND__DIRECT_REG_POLLING = 2,
+	MMSCH_COMMAND__DIRECT_REG_READ_MODIFY_WRITE = 3,
+	MMSCH_COMMAND__INDIRECT_REG_WRITE = 8,
+	MMSCH_COMMAND__END = 0xf
+};
+
+struct mmsch_v5_0_table_info {
+	uint32_t init_status;
+	uint32_t table_offset;
+	uint32_t table_size;
+};
+
+struct mmsch_v5_0_init_header {
+	uint32_t version;
+	uint32_t total_size;
+	struct mmsch_v5_0_table_info vcn0;
+	struct mmsch_v5_0_table_info mjpegdec0[5];
+	struct mmsch_v5_0_table_info mjpegdec1[5];
+};
+
+struct mmsch_v5_0_cmd_direct_reg_header {
+	uint32_t reg_offset   : 28;
+	uint32_t command_type : 4;
+};
+
+struct mmsch_v5_0_cmd_indirect_reg_header {
+	uint32_t reg_offset    : 20;
+	uint32_t reg_idx_space : 8;
+	uint32_t command_type  : 4;
+};
+
+struct mmsch_v5_0_cmd_direct_write {
+	struct mmsch_v5_0_cmd_direct_reg_header cmd_header;
+	uint32_t reg_value;
+};
+
+struct mmsch_v5_0_cmd_direct_read_modify_write {
+	struct mmsch_v5_0_cmd_direct_reg_header cmd_header;
+	uint32_t write_data;
+	uint32_t mask_value;
+};
+
+struct mmsch_v5_0_cmd_direct_polling {
+	struct mmsch_v5_0_cmd_direct_reg_header cmd_header;
+	uint32_t mask_value;
+	uint32_t wait_value;
+};
+
+struct mmsch_v5_0_cmd_end {
+	struct mmsch_v5_0_cmd_direct_reg_header cmd_header;
+};
+
+struct mmsch_v5_0_cmd_indirect_write {
+	struct mmsch_v5_0_cmd_indirect_reg_header cmd_header;
+	uint32_t reg_value;
+};
+
+#define MMSCH_V5_0_INSERT_DIRECT_RD_MOD_WT(reg, mask, data) { \
+	size = sizeof(struct mmsch_v5_0_cmd_direct_read_modify_write); \
+	size_dw = size / 4; \
+	direct_rd_mod_wt.cmd_header.reg_offset = reg; \
+	direct_rd_mod_wt.mask_value = mask; \
+	direct_rd_mod_wt.write_data = data; \
+	memcpy((void *)table_loc, &direct_rd_mod_wt, size); \
+	table_loc += size_dw; \
+	table_size += size_dw; \
+}
+
+#define MMSCH_V5_0_INSERT_DIRECT_WT(reg, value) { \
+	size = sizeof(struct mmsch_v5_0_cmd_direct_write); \
+	size_dw = size / 4; \
+	direct_wt.cmd_header.reg_offset = reg; \
+	direct_wt.reg_value = value; \
+	memcpy((void *)table_loc, &direct_wt, size); \
+	table_loc += size_dw; \
+	table_size += size_dw; \
+}
+
+#define MMSCH_V5_0_INSERT_DIRECT_POLL(reg, mask, wait) { \
+	size = sizeof(struct mmsch_v5_0_cmd_direct_polling); \
+	size_dw = size / 4; \
+	direct_poll.cmd_header.reg_offset = reg; \
+	direct_poll.mask_value = mask; \
+	direct_poll.wait_value = wait; \
+	memcpy((void *)table_loc, &direct_poll, size); \
+	table_loc += size_dw; \
+	table_size += size_dw; \
+}
+
+#define MMSCH_V5_0_INSERT_END() { \
+	size = sizeof(struct mmsch_v5_0_cmd_end); \
+	size_dw = size / 4; \
+	memcpy((void *)table_loc, &end, size); \
+	table_loc += size_dw; \
+	table_size += size_dw; \
+}
+
+#endif
+
diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c
index 63725b2ebc03..9a40107a0869 100644
--- a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c
+++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c
@@ -93,7 +93,7 @@ static int xgpu_ai_poll_ack(struct amdgpu_device *adev)
 		timeout -= 5;
 	} while (timeout > 1);
 
-	pr_err("Doesn't get TRN_MSG_ACK from pf in %d msec\n", AI_MAILBOX_POLL_ACK_TIMEDOUT);
+	dev_err(adev->dev, "Doesn't get TRN_MSG_ACK from pf in %d msec\n", AI_MAILBOX_POLL_ACK_TIMEDOUT);
 
 	return -ETIME;
 }
@@ -111,7 +111,7 @@ static int xgpu_ai_poll_msg(struct amdgpu_device *adev, enum idh_event event)
 		timeout -= 10;
 	} while (timeout > 1);
 
-	pr_err("Doesn't get msg:%d from pf, error=%d\n", event, r);
+	dev_err(adev->dev, "Doesn't get msg:%d from pf, error=%d\n", event, r);
 
 	return -ETIME;
 }
@@ -132,7 +132,7 @@ static void xgpu_ai_mailbox_trans_msg (struct amdgpu_device *adev,
 		xgpu_ai_mailbox_set_valid(adev, false);
 		trn = xgpu_ai_peek_ack(adev);
 		if (trn) {
-			pr_err("trn=%x ACK should not assert! wait again !\n", trn);
+			dev_err_ratelimited(adev->dev, "trn=%x ACK should not assert! wait again !\n", trn);
 			msleep(1);
 		}
 	} while(trn);
@@ -155,7 +155,7 @@ static void xgpu_ai_mailbox_trans_msg (struct amdgpu_device *adev,
 	/* start to poll ack */
 	r = xgpu_ai_poll_ack(adev);
 	if (r)
-		pr_err("Doesn't get ack from pf, continue\n");
+		dev_err(adev->dev, "Doesn't get ack from pf, continue\n");
 
 	xgpu_ai_mailbox_set_valid(adev, false);
 }
@@ -173,7 +173,7 @@ static int xgpu_ai_send_access_requests(struct amdgpu_device *adev,
 		req == IDH_REQ_GPU_RESET_ACCESS) {
 		r = xgpu_ai_poll_msg(adev, IDH_READY_TO_ACCESS_GPU);
 		if (r) {
-			pr_err("Doesn't get READY_TO_ACCESS_GPU from pf, give up\n");
+			dev_err(adev->dev, "Doesn't get READY_TO_ACCESS_GPU from pf, give up\n");
 			return r;
 		}
 		/* Retrieve checksum from mailbox2 */
@@ -231,7 +231,7 @@ static int xgpu_ai_mailbox_ack_irq(struct amdgpu_device *adev,
 					struct amdgpu_irq_src *source,
 					struct amdgpu_iv_entry *entry)
 {
-	DRM_DEBUG("get ack intr and do nothing.\n");
+	dev_dbg(adev->dev, "get ack intr and do nothing.\n");
 	return 0;
 }
 
@@ -249,52 +249,80 @@ static int xgpu_ai_set_mailbox_ack_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static void xgpu_ai_mailbox_flr_work(struct work_struct *work)
+static void xgpu_ai_ready_to_reset(struct amdgpu_device *adev)
 {
-	struct amdgpu_virt *virt = container_of(work, struct amdgpu_virt, flr_work);
-	struct amdgpu_device *adev = container_of(virt, struct amdgpu_device, virt);
-	int timeout = AI_MAILBOX_POLL_FLR_TIMEDOUT;
-
-	/* block amdgpu_gpu_recover till msg FLR COMPLETE received,
-	 * otherwise the mailbox msg will be ruined/reseted by
-	 * the VF FLR.
-	 */
-	if (atomic_cmpxchg(&adev->reset_domain->in_gpu_reset, 0, 1) != 0)
-		return;
-
-	down_write(&adev->reset_domain->sem);
-
-	amdgpu_virt_fini_data_exchange(adev);
-
 	xgpu_ai_mailbox_trans_msg(adev, IDH_READY_TO_RESET, 0, 0, 0);
+}
 
+static int xgpu_ai_wait_reset(struct amdgpu_device *adev)
+{
+	int timeout = AI_MAILBOX_POLL_FLR_TIMEDOUT;
 	do {
-		if (xgpu_ai_mailbox_peek_msg(adev) == IDH_FLR_NOTIFICATION_CMPL)
-			goto flr_done;
-
+		if (xgpu_ai_mailbox_peek_msg(adev) == IDH_FLR_NOTIFICATION_CMPL) {
+			dev_dbg(adev->dev, "Got AI IDH_FLR_NOTIFICATION_CMPL after %d ms\n", AI_MAILBOX_POLL_FLR_TIMEDOUT - timeout);
+			return 0;
+		}
 		msleep(10);
 		timeout -= 10;
 	} while (timeout > 1);
 
-flr_done:
-	atomic_set(&adev->reset_domain->in_gpu_reset, 0);
-	up_write(&adev->reset_domain->sem);
+	dev_dbg(adev->dev, "waiting AI IDH_FLR_NOTIFICATION_CMPL timeout\n");
+	return -ETIME;
+}
+
+static void xgpu_ai_mailbox_flr_work(struct work_struct *work)
+{
+	struct amdgpu_virt *virt = container_of(work, struct amdgpu_virt, flr_work);
+	struct amdgpu_device *adev = container_of(virt, struct amdgpu_device, virt);
+	struct amdgpu_reset_context reset_context = { 0 };
+
+	amdgpu_virt_fini_data_exchange(adev);
 
 	/* Trigger recovery for world switch failure if no TDR */
 	if (amdgpu_device_should_recover_gpu(adev)
 		&& (!amdgpu_device_has_job_running(adev) ||
 			adev->sdma_timeout == MAX_SCHEDULE_TIMEOUT)) {
-		struct amdgpu_reset_context reset_context;
-		memset(&reset_context, 0, sizeof(reset_context));
 
 		reset_context.method = AMD_RESET_METHOD_NONE;
 		reset_context.reset_req_dev = adev;
 		clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
+		set_bit(AMDGPU_HOST_FLR, &reset_context.flags);
 
 		amdgpu_device_gpu_recover(adev, NULL, &reset_context);
 	}
 }
 
+static void xgpu_ai_mailbox_req_bad_pages_work(struct work_struct *work)
+{
+	struct amdgpu_virt *virt = container_of(work, struct amdgpu_virt, req_bad_pages_work);
+	struct amdgpu_device *adev = container_of(virt, struct amdgpu_device, virt);
+
+	if (down_read_trylock(&adev->reset_domain->sem)) {
+		amdgpu_virt_fini_data_exchange(adev);
+		amdgpu_virt_request_bad_pages(adev);
+		up_read(&adev->reset_domain->sem);
+	}
+}
+
+/**
+ * xgpu_ai_mailbox_handle_bad_pages_work - Reinitialize the data exchange region to get fresh bad page information
+ * @work: pointer to the work_struct
+ *
+ * This work handler is triggered when bad pages are ready, and it reinitializes
+ * the data exchange region to retrieve updated bad page information from the host.
+ */
+static void xgpu_ai_mailbox_handle_bad_pages_work(struct work_struct *work)
+{
+	struct amdgpu_virt *virt = container_of(work, struct amdgpu_virt, handle_bad_pages_work);
+	struct amdgpu_device *adev = container_of(virt, struct amdgpu_device, virt);
+
+	if (down_read_trylock(&adev->reset_domain->sem)) {
+		amdgpu_virt_fini_data_exchange(adev);
+		amdgpu_virt_init_data_exchange(adev);
+		up_read(&adev->reset_domain->sem);
+	}
+}
+
 static int xgpu_ai_set_mailbox_rcv_irq(struct amdgpu_device *adev,
 				       struct amdgpu_irq_src *src,
 				       unsigned type,
@@ -314,26 +342,47 @@ static int xgpu_ai_mailbox_rcv_irq(struct amdgpu_device *adev,
 				   struct amdgpu_iv_entry *entry)
 {
 	enum idh_event event = xgpu_ai_mailbox_peek_msg(adev);
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 
 	switch (event) {
-		case IDH_FLR_NOTIFICATION:
-		if (amdgpu_sriov_runtime(adev) && !amdgpu_in_reset(adev))
+	case IDH_RAS_BAD_PAGES_READY:
+		xgpu_ai_mailbox_send_ack(adev);
+		if (amdgpu_sriov_runtime(adev))
+			schedule_work(&adev->virt.handle_bad_pages_work);
+		break;
+	case IDH_RAS_BAD_PAGES_NOTIFICATION:
+		xgpu_ai_mailbox_send_ack(adev);
+		if (amdgpu_sriov_runtime(adev))
+			schedule_work(&adev->virt.req_bad_pages_work);
+		break;
+	case IDH_UNRECOV_ERR_NOTIFICATION:
+		xgpu_ai_mailbox_send_ack(adev);
+		ras->is_rma = true;
+		dev_err(adev->dev, "VF is in an unrecoverable state. Runtime Services are halted.\n");
+		if (amdgpu_sriov_runtime(adev))
 			WARN_ONCE(!amdgpu_reset_domain_schedule(adev->reset_domain,
-								&adev->virt.flr_work),
-				  "Failed to queue work! at %s",
-				  __func__);
+					&adev->virt.flr_work),
+					"Failed to queue work! at %s",
+					__func__);
 		break;
-		case IDH_QUERY_ALIVE:
-			xgpu_ai_mailbox_send_ack(adev);
-			break;
-		/* READY_TO_ACCESS_GPU is fetched by kernel polling, IRQ can ignore
-		 * it byfar since that polling thread will handle it,
-		 * other msg like flr complete is not handled here.
-		 */
-		case IDH_CLR_MSG_BUF:
-		case IDH_FLR_NOTIFICATION_CMPL:
-		case IDH_READY_TO_ACCESS_GPU:
-		default:
+	case IDH_FLR_NOTIFICATION:
+		if (amdgpu_sriov_runtime(adev))
+			WARN_ONCE(!amdgpu_reset_domain_schedule(adev->reset_domain,
+						&adev->virt.flr_work),
+					"Failed to queue work! at %s",
+					__func__);
+		break;
+	case IDH_QUERY_ALIVE:
+		xgpu_ai_mailbox_send_ack(adev);
+		break;
+	/* READY_TO_ACCESS_GPU is fetched by kernel polling, IRQ can ignore
+	 * it byfar since that polling thread will handle it,
+	 * other msg like flr complete is not handled here.
+	 */
+	case IDH_CLR_MSG_BUF:
+	case IDH_FLR_NOTIFICATION_CMPL:
+	case IDH_READY_TO_ACCESS_GPU:
+	default:
 		break;
 	}
 
@@ -389,6 +438,8 @@ int xgpu_ai_mailbox_get_irq(struct amdgpu_device *adev)
 	}
 
 	INIT_WORK(&adev->virt.flr_work, xgpu_ai_mailbox_flr_work);
+	INIT_WORK(&adev->virt.req_bad_pages_work, xgpu_ai_mailbox_req_bad_pages_work);
+	INIT_WORK(&adev->virt.handle_bad_pages_work, xgpu_ai_mailbox_handle_bad_pages_work);
 
 	return 0;
 }
@@ -404,17 +455,27 @@ static int xgpu_ai_request_init_data(struct amdgpu_device *adev)
 	return xgpu_ai_send_access_requests(adev, IDH_REQ_GPU_INIT_DATA);
 }
 
-static void xgpu_ai_ras_poison_handler(struct amdgpu_device *adev)
+static void xgpu_ai_ras_poison_handler(struct amdgpu_device *adev,
+					enum amdgpu_ras_block block)
 {
 	xgpu_ai_send_access_requests(adev, IDH_RAS_POISON);
 }
 
+static bool xgpu_ai_rcvd_ras_intr(struct amdgpu_device *adev)
+{
+	enum idh_event msg = xgpu_ai_mailbox_peek_msg(adev);
+
+	return (msg == IDH_RAS_ERROR_DETECTED || msg == 0xFFFFFFFF);
+}
+
 const struct amdgpu_virt_ops xgpu_ai_virt_ops = {
 	.req_full_gpu	= xgpu_ai_request_full_gpu_access,
 	.rel_full_gpu	= xgpu_ai_release_full_gpu_access,
 	.reset_gpu = xgpu_ai_request_reset,
-	.wait_reset = NULL,
+	.ready_to_reset = xgpu_ai_ready_to_reset,
+	.wait_reset = xgpu_ai_wait_reset,
 	.trans_msg = xgpu_ai_mailbox_trans_msg,
 	.req_init_data  = xgpu_ai_request_init_data,
 	.ras_poison_handler = xgpu_ai_ras_poison_handler,
+	.rcvd_ras_intr = xgpu_ai_rcvd_ras_intr,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h
index af1a784696bd..874b9f8f9804 100644
--- a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h
+++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h
@@ -40,6 +40,7 @@ enum idh_request {
 	IDH_LOG_VF_ERROR       = 200,
 	IDH_READY_TO_RESET 	= 201,
 	IDH_RAS_POISON  = 202,
+	IDH_REQ_RAS_BAD_PAGES = 205,
 };
 
 enum idh_event {
@@ -51,7 +52,12 @@ enum idh_event {
 	IDH_FAIL,
 	IDH_QUERY_ALIVE,
 	IDH_REQ_GPU_INIT_DATA_READY,
-
+	IDH_RAS_POISON_READY,
+	IDH_PF_SOFT_FLR_NOTIFICATION,
+	IDH_RAS_ERROR_DETECTED,
+	IDH_RAS_BAD_PAGES_READY = 15,
+	IDH_RAS_BAD_PAGES_NOTIFICATION = 16,
+	IDH_UNRECOV_ERR_NOTIFICATION = 17,
 	IDH_TEXT_MESSAGE = 255,
 };
 
@@ -62,7 +68,9 @@ int xgpu_ai_mailbox_add_irq_id(struct amdgpu_device *adev);
 int xgpu_ai_mailbox_get_irq(struct amdgpu_device *adev);
 void xgpu_ai_mailbox_put_irq(struct amdgpu_device *adev);
 
-#define AI_MAIBOX_CONTROL_TRN_OFFSET_BYTE SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_MAILBOX_CONTROL) * 4
-#define AI_MAIBOX_CONTROL_RCV_OFFSET_BYTE SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_MAILBOX_CONTROL) * 4 + 1
+#define AI_MAIBOX_CONTROL_TRN_OFFSET_BYTE \
+	(SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_MAILBOX_CONTROL) * 4)
+#define AI_MAIBOX_CONTROL_RCV_OFFSET_BYTE \
+	(SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_MAILBOX_CONTROL) * 4 + 1)
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c
index cae1aaa4ddb6..e7cd07383d56 100644
--- a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c
+++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c
@@ -61,15 +61,20 @@ static enum idh_event xgpu_nv_mailbox_peek_msg(struct amdgpu_device *adev)
 static int xgpu_nv_mailbox_rcv_msg(struct amdgpu_device *adev,
 				   enum idh_event event)
 {
+	int r = 0;
 	u32 reg;
 
 	reg = RREG32_NO_KIQ(mmMAILBOX_MSGBUF_RCV_DW0);
-	if (reg != event)
+	if (reg == IDH_FAIL)
+		r = -EINVAL;
+	if (reg == IDH_UNRECOV_ERR_NOTIFICATION)
+		r = -ENODEV;
+	else if (reg != event)
 		return -ENOENT;
 
 	xgpu_nv_mailbox_send_ack(adev);
 
-	return 0;
+	return r;
 }
 
 static uint8_t xgpu_nv_peek_ack(struct amdgpu_device *adev)
@@ -91,7 +96,7 @@ static int xgpu_nv_poll_ack(struct amdgpu_device *adev)
 		timeout -= 5;
 	} while (timeout > 1);
 
-	pr_err("Doesn't get TRN_MSG_ACK from pf in %d msec\n", NV_MAILBOX_POLL_ACK_TIMEDOUT);
+	dev_err(adev->dev, "Doesn't get TRN_MSG_ACK from pf in %d msec \n", NV_MAILBOX_POLL_ACK_TIMEDOUT);
 
 	return -ETIME;
 }
@@ -100,19 +105,31 @@ static int xgpu_nv_poll_msg(struct amdgpu_device *adev, enum idh_event event)
 {
 	int r;
 	uint64_t timeout, now;
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 
 	now = (uint64_t)ktime_to_ms(ktime_get());
 	timeout = now + NV_MAILBOX_POLL_MSG_TIMEDOUT;
 
 	do {
 		r = xgpu_nv_mailbox_rcv_msg(adev, event);
-		if (!r)
+		if (!r) {
+			dev_dbg(adev->dev, "rcv_msg 0x%x after %llu ms\n",
+					event, NV_MAILBOX_POLL_MSG_TIMEDOUT - timeout + now);
 			return 0;
+		} else if (r == -ENODEV) {
+			if (!amdgpu_ras_is_rma(adev)) {
+				ras->is_rma = true;
+				dev_err(adev->dev, "VF is in an unrecoverable state. "
+						"Runtime Services are halted.\n");
+			}
+			return r;
+		}
 
 		msleep(10);
 		now = (uint64_t)ktime_to_ms(ktime_get());
 	} while (timeout > now);
 
+	dev_dbg(adev->dev, "nv_poll_msg timed out\n");
 
 	return -ETIME;
 }
@@ -133,11 +150,12 @@ static void xgpu_nv_mailbox_trans_msg (struct amdgpu_device *adev,
 		xgpu_nv_mailbox_set_valid(adev, false);
 		trn = xgpu_nv_peek_ack(adev);
 		if (trn) {
-			pr_err("trn=%x ACK should not assert! wait again !\n", trn);
+			dev_err_ratelimited(adev->dev, "trn=%x ACK should not assert! wait again !\n", trn);
 			msleep(1);
 		}
 	} while (trn);
 
+	dev_dbg(adev->dev, "trans_msg req = 0x%x, data1 = 0x%x\n", req, data1);
 	WREG32_NO_KIQ(mmMAILBOX_MSGBUF_TRN_DW0, req);
 	WREG32_NO_KIQ(mmMAILBOX_MSGBUF_TRN_DW1, data1);
 	WREG32_NO_KIQ(mmMAILBOX_MSGBUF_TRN_DW2, data2);
@@ -147,19 +165,27 @@ static void xgpu_nv_mailbox_trans_msg (struct amdgpu_device *adev,
 	/* start to poll ack */
 	r = xgpu_nv_poll_ack(adev);
 	if (r)
-		pr_err("Doesn't get ack from pf, continue\n");
+		dev_err(adev->dev, "Doesn't get ack from pf, continue\n");
 
 	xgpu_nv_mailbox_set_valid(adev, false);
 }
 
-static int xgpu_nv_send_access_requests(struct amdgpu_device *adev,
-					enum idh_request req)
+static int xgpu_nv_send_access_requests_with_param(struct amdgpu_device *adev,
+			enum idh_request req, u32 data1, u32 data2, u32 data3)
 {
-	int r, retry = 1;
+	struct amdgpu_virt *virt = &adev->virt;
+	int r = 0, retry = 1;
 	enum idh_event event = -1;
 
+	mutex_lock(&virt->access_req_mutex);
 send_request:
-	xgpu_nv_mailbox_trans_msg(adev, req, 0, 0, 0);
+
+	if (amdgpu_ras_is_rma(adev)) {
+		r = -ENODEV;
+		goto out;
+	}
+
+	xgpu_nv_mailbox_trans_msg(adev, req, data1, data2, data3);
 
 	switch (req) {
 	case IDH_REQ_GPU_INIT_ACCESS:
@@ -170,6 +196,19 @@ send_request:
 	case IDH_REQ_GPU_INIT_DATA:
 		event = IDH_REQ_GPU_INIT_DATA_READY;
 		break;
+	case IDH_RAS_POISON:
+		if (data1 != 0)
+			event = IDH_RAS_POISON_READY;
+		break;
+	case IDH_REQ_RAS_ERROR_COUNT:
+		event = IDH_RAS_ERROR_COUNT_READY;
+		break;
+	case IDH_REQ_RAS_CPER_DUMP:
+		event = IDH_RAS_CPER_DUMP_READY;
+		break;
+	case IDH_REQ_RAS_CHK_CRITI:
+		event = IDH_REQ_RAS_CHK_CRITI_READY;
+		break;
 	default:
 		break;
 	}
@@ -177,24 +216,30 @@ send_request:
 	if (event != -1) {
 		r = xgpu_nv_poll_msg(adev, event);
 		if (r) {
-			if (retry++ < 2)
+			if (retry++ < 5)
 				goto send_request;
 
 			if (req != IDH_REQ_GPU_INIT_DATA) {
-				pr_err("Doesn't get msg:%d from pf, error=%d\n", event, r);
-				return r;
-			}
-			else /* host doesn't support REQ_GPU_INIT_DATA handshake */
+				dev_err(adev->dev, "Doesn't get msg:%d from pf, error=%d\n", event, r);
+				goto out;
+			} else /* host doesn't support REQ_GPU_INIT_DATA handshake */
 				adev->virt.req_init_data_ver = 0;
 		} else {
-			if (req == IDH_REQ_GPU_INIT_DATA)
-			{
-				adev->virt.req_init_data_ver =
-					RREG32_NO_KIQ(mmMAILBOX_MSGBUF_RCV_DW1);
-
-				/* assume V1 in case host doesn't set version number */
-				if (adev->virt.req_init_data_ver < 1)
-					adev->virt.req_init_data_ver = 1;
+			if (req == IDH_REQ_GPU_INIT_DATA) {
+				switch (RREG32_NO_KIQ(mmMAILBOX_MSGBUF_RCV_DW1)) {
+				case GPU_CRIT_REGION_V2:
+					adev->virt.req_init_data_ver = GPU_CRIT_REGION_V2;
+					adev->virt.init_data_header.offset =
+						RREG32_NO_KIQ(mmMAILBOX_MSGBUF_RCV_DW2);
+					adev->virt.init_data_header.size_kb =
+						RREG32_NO_KIQ(mmMAILBOX_MSGBUF_RCV_DW3);
+					break;
+				default:
+					adev->virt.req_init_data_ver = GPU_CRIT_REGION_V1;
+					adev->virt.init_data_header.offset = -1;
+					adev->virt.init_data_header.size_kb = 0;
+					break;
+				}
 			}
 		}
 
@@ -205,7 +250,17 @@ send_request:
 		}
 	}
 
-	return 0;
+out:
+	mutex_unlock(&virt->access_req_mutex);
+
+	return r;
+}
+
+static int xgpu_nv_send_access_requests(struct amdgpu_device *adev,
+					enum idh_request req)
+{
+	return xgpu_nv_send_access_requests_with_param(adev,
+						req, 0, 0, 0);
 }
 
 static int xgpu_nv_request_reset(struct amdgpu_device *adev)
@@ -245,14 +300,15 @@ static int xgpu_nv_release_full_gpu_access(struct amdgpu_device *adev,
 
 static int xgpu_nv_request_init_data(struct amdgpu_device *adev)
 {
-	return xgpu_nv_send_access_requests(adev, IDH_REQ_GPU_INIT_DATA);
+	return xgpu_nv_send_access_requests_with_param(adev, IDH_REQ_GPU_INIT_DATA,
+			0, GPU_CRIT_REGION_V2, 0);
 }
 
 static int xgpu_nv_mailbox_ack_irq(struct amdgpu_device *adev,
 					struct amdgpu_irq_src *source,
 					struct amdgpu_iv_entry *entry)
 {
-	DRM_DEBUG("get ack intr and do nothing.\n");
+	dev_dbg(adev->dev, "get ack intr and do nothing.\n");
 	return 0;
 }
 
@@ -273,36 +329,34 @@ static int xgpu_nv_set_mailbox_ack_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static void xgpu_nv_mailbox_flr_work(struct work_struct *work)
+static void xgpu_nv_ready_to_reset(struct amdgpu_device *adev)
 {
-	struct amdgpu_virt *virt = container_of(work, struct amdgpu_virt, flr_work);
-	struct amdgpu_device *adev = container_of(virt, struct amdgpu_device, virt);
-	int timeout = NV_MAILBOX_POLL_FLR_TIMEDOUT;
-
-	/* block amdgpu_gpu_recover till msg FLR COMPLETE received,
-	 * otherwise the mailbox msg will be ruined/reseted by
-	 * the VF FLR.
-	 */
-	if (atomic_cmpxchg(&adev->reset_domain->in_gpu_reset, 0, 1) != 0)
-		return;
-
-	down_write(&adev->reset_domain->sem);
-
-	amdgpu_virt_fini_data_exchange(adev);
-
 	xgpu_nv_mailbox_trans_msg(adev, IDH_READY_TO_RESET, 0, 0, 0);
+}
 
+static int xgpu_nv_wait_reset(struct amdgpu_device *adev)
+{
+	int timeout = NV_MAILBOX_POLL_FLR_TIMEDOUT;
 	do {
-		if (xgpu_nv_mailbox_peek_msg(adev) == IDH_FLR_NOTIFICATION_CMPL)
-			goto flr_done;
-
+		if (xgpu_nv_mailbox_peek_msg(adev) == IDH_FLR_NOTIFICATION_CMPL) {
+			dev_dbg(adev->dev, "Got NV IDH_FLR_NOTIFICATION_CMPL after %d ms\n", NV_MAILBOX_POLL_FLR_TIMEDOUT - timeout);
+			return 0;
+		}
 		msleep(10);
 		timeout -= 10;
 	} while (timeout > 1);
 
-flr_done:
-	atomic_set(&adev->reset_domain->in_gpu_reset, 0);
-	up_write(&adev->reset_domain->sem);
+	dev_dbg(adev->dev, "waiting NV IDH_FLR_NOTIFICATION_CMPL timeout\n");
+	return -ETIME;
+}
+
+static void xgpu_nv_mailbox_flr_work(struct work_struct *work)
+{
+	struct amdgpu_virt *virt = container_of(work, struct amdgpu_virt, flr_work);
+	struct amdgpu_device *adev = container_of(virt, struct amdgpu_device, virt);
+	struct amdgpu_reset_context reset_context = { 0 };
+
+	amdgpu_virt_fini_data_exchange(adev);
 
 	/* Trigger recovery for world switch failure if no TDR */
 	if (amdgpu_device_should_recover_gpu(adev)
@@ -311,17 +365,47 @@ flr_done:
 		adev->gfx_timeout == MAX_SCHEDULE_TIMEOUT ||
 		adev->compute_timeout == MAX_SCHEDULE_TIMEOUT ||
 		adev->video_timeout == MAX_SCHEDULE_TIMEOUT)) {
-		struct amdgpu_reset_context reset_context;
-		memset(&reset_context, 0, sizeof(reset_context));
 
 		reset_context.method = AMD_RESET_METHOD_NONE;
 		reset_context.reset_req_dev = adev;
 		clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
+		set_bit(AMDGPU_HOST_FLR, &reset_context.flags);
 
 		amdgpu_device_gpu_recover(adev, NULL, &reset_context);
 	}
 }
 
+static void xgpu_nv_mailbox_req_bad_pages_work(struct work_struct *work)
+{
+	struct amdgpu_virt *virt = container_of(work, struct amdgpu_virt, req_bad_pages_work);
+	struct amdgpu_device *adev = container_of(virt, struct amdgpu_device, virt);
+
+	if (down_read_trylock(&adev->reset_domain->sem)) {
+		amdgpu_virt_fini_data_exchange(adev);
+		amdgpu_virt_request_bad_pages(adev);
+		up_read(&adev->reset_domain->sem);
+	}
+}
+
+/**
+ * xgpu_nv_mailbox_handle_bad_pages_work - Reinitialize the data exchange region to get fresh bad page information
+ * @work: pointer to the work_struct
+ *
+ * This work handler is triggered when bad pages are ready, and it reinitializes
+ * the data exchange region to retrieve updated bad page information from the host.
+ */
+static void xgpu_nv_mailbox_handle_bad_pages_work(struct work_struct *work)
+{
+	struct amdgpu_virt *virt = container_of(work, struct amdgpu_virt, handle_bad_pages_work);
+	struct amdgpu_device *adev = container_of(virt, struct amdgpu_device, virt);
+
+	if (down_read_trylock(&adev->reset_domain->sem)) {
+		amdgpu_virt_fini_data_exchange(adev);
+		amdgpu_virt_init_data_exchange(adev);
+		up_read(&adev->reset_domain->sem);
+	}
+}
+
 static int xgpu_nv_set_mailbox_rcv_irq(struct amdgpu_device *adev,
 				       struct amdgpu_irq_src *src,
 				       unsigned type,
@@ -344,10 +428,34 @@ static int xgpu_nv_mailbox_rcv_irq(struct amdgpu_device *adev,
 				   struct amdgpu_iv_entry *entry)
 {
 	enum idh_event event = xgpu_nv_mailbox_peek_msg(adev);
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 
 	switch (event) {
+	case IDH_RAS_BAD_PAGES_READY:
+		xgpu_nv_mailbox_send_ack(adev);
+		if (amdgpu_sriov_runtime(adev))
+			schedule_work(&adev->virt.handle_bad_pages_work);
+		break;
+	case IDH_RAS_BAD_PAGES_NOTIFICATION:
+		xgpu_nv_mailbox_send_ack(adev);
+		if (amdgpu_sriov_runtime(adev))
+			schedule_work(&adev->virt.req_bad_pages_work);
+		break;
+	case IDH_UNRECOV_ERR_NOTIFICATION:
+		xgpu_nv_mailbox_send_ack(adev);
+		if (!amdgpu_ras_is_rma(adev)) {
+			ras->is_rma = true;
+			dev_err(adev->dev, "VF is in an unrecoverable state. Runtime Services are halted.\n");
+		}
+
+		if (amdgpu_sriov_runtime(adev))
+			WARN_ONCE(!amdgpu_reset_domain_schedule(adev->reset_domain,
+						&adev->virt.flr_work),
+					"Failed to queue work! at %s",
+					__func__);
+		break;
 	case IDH_FLR_NOTIFICATION:
-		if (amdgpu_sriov_runtime(adev) && !amdgpu_in_reset(adev))
+		if (amdgpu_sriov_runtime(adev))
 			WARN_ONCE(!amdgpu_reset_domain_schedule(adev->reset_domain,
 				   &adev->virt.flr_work),
 				  "Failed to queue work! at %s",
@@ -416,6 +524,8 @@ int xgpu_nv_mailbox_get_irq(struct amdgpu_device *adev)
 	}
 
 	INIT_WORK(&adev->virt.flr_work, xgpu_nv_mailbox_flr_work);
+	INIT_WORK(&adev->virt.req_bad_pages_work, xgpu_nv_mailbox_req_bad_pages_work);
+	INIT_WORK(&adev->virt.handle_bad_pages_work, xgpu_nv_mailbox_handle_bad_pages_work);
 
 	return 0;
 }
@@ -426,9 +536,53 @@ void xgpu_nv_mailbox_put_irq(struct amdgpu_device *adev)
 	amdgpu_irq_put(adev, &adev->virt.rcv_irq, 0);
 }
 
-static void xgpu_nv_ras_poison_handler(struct amdgpu_device *adev)
+static void xgpu_nv_ras_poison_handler(struct amdgpu_device *adev,
+		enum amdgpu_ras_block block)
+{
+	if (amdgpu_ip_version(adev, UMC_HWIP, 0) < IP_VERSION(12, 0, 0)) {
+		xgpu_nv_send_access_requests(adev, IDH_RAS_POISON);
+	} else {
+		amdgpu_virt_fini_data_exchange(adev);
+		xgpu_nv_send_access_requests_with_param(adev,
+					IDH_RAS_POISON,	block, 0, 0);
+	}
+}
+
+static bool xgpu_nv_rcvd_ras_intr(struct amdgpu_device *adev)
 {
-	xgpu_nv_send_access_requests(adev, IDH_RAS_POISON);
+	enum idh_event msg = xgpu_nv_mailbox_peek_msg(adev);
+
+	return (msg == IDH_RAS_ERROR_DETECTED || msg == 0xFFFFFFFF);
+}
+
+static int xgpu_nv_req_ras_err_count(struct amdgpu_device *adev)
+{
+	return xgpu_nv_send_access_requests(adev, IDH_REQ_RAS_ERROR_COUNT);
+}
+
+static int xgpu_nv_req_ras_cper_dump(struct amdgpu_device *adev, u64 vf_rptr)
+{
+	uint32_t vf_rptr_hi, vf_rptr_lo;
+
+	vf_rptr_hi = (uint32_t)(vf_rptr >> 32);
+	vf_rptr_lo = (uint32_t)(vf_rptr & 0xFFFFFFFF);
+	return xgpu_nv_send_access_requests_with_param(
+		adev, IDH_REQ_RAS_CPER_DUMP, vf_rptr_hi, vf_rptr_lo, 0);
+}
+
+static int xgpu_nv_req_ras_bad_pages(struct amdgpu_device *adev)
+{
+	return xgpu_nv_send_access_requests(adev, IDH_REQ_RAS_BAD_PAGES);
+}
+
+static int xgpu_nv_check_vf_critical_region(struct amdgpu_device *adev, u64 addr)
+{
+	uint32_t addr_hi, addr_lo;
+
+	addr_hi = (uint32_t)(addr >> 32);
+	addr_lo = (uint32_t)(addr & 0xFFFFFFFF);
+	return xgpu_nv_send_access_requests_with_param(
+		adev, IDH_REQ_RAS_CHK_CRITI, addr_hi, addr_lo, 0);
 }
 
 const struct amdgpu_virt_ops xgpu_nv_virt_ops = {
@@ -436,7 +590,13 @@ const struct amdgpu_virt_ops xgpu_nv_virt_ops = {
 	.rel_full_gpu	= xgpu_nv_release_full_gpu_access,
 	.req_init_data  = xgpu_nv_request_init_data,
 	.reset_gpu = xgpu_nv_request_reset,
-	.wait_reset = NULL,
+	.ready_to_reset = xgpu_nv_ready_to_reset,
+	.wait_reset = xgpu_nv_wait_reset,
 	.trans_msg = xgpu_nv_mailbox_trans_msg,
 	.ras_poison_handler = xgpu_nv_ras_poison_handler,
+	.rcvd_ras_intr = xgpu_nv_rcvd_ras_intr,
+	.req_ras_err_count = xgpu_nv_req_ras_err_count,
+	.req_ras_cper_dump = xgpu_nv_req_ras_cper_dump,
+	.req_bad_pages = xgpu_nv_req_ras_bad_pages,
+	.req_ras_chk_criti = xgpu_nv_check_vf_critical_region
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.h b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.h
index d0221ce08769..c1083e5e41e0 100644
--- a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.h
+++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.h
@@ -25,8 +25,8 @@
 #define __MXGPU_NV_H__
 
 #define NV_MAILBOX_POLL_ACK_TIMEDOUT	500
-#define NV_MAILBOX_POLL_MSG_TIMEDOUT	6000
-#define NV_MAILBOX_POLL_FLR_TIMEDOUT	5000
+#define NV_MAILBOX_POLL_MSG_TIMEDOUT	15000
+#define NV_MAILBOX_POLL_FLR_TIMEDOUT	10000
 #define NV_MAILBOX_POLL_MSG_REP_MAX	11
 
 enum idh_request {
@@ -40,6 +40,10 @@ enum idh_request {
 	IDH_LOG_VF_ERROR	= 200,
 	IDH_READY_TO_RESET 	= 201,
 	IDH_RAS_POISON	= 202,
+	IDH_REQ_RAS_ERROR_COUNT = 203,
+	IDH_REQ_RAS_CPER_DUMP = 204,
+	IDH_REQ_RAS_BAD_PAGES = 205,
+	IDH_REQ_RAS_CHK_CRITI = 206
 };
 
 enum idh_event {
@@ -51,6 +55,15 @@ enum idh_event {
 	IDH_FAIL,
 	IDH_QUERY_ALIVE,
 	IDH_REQ_GPU_INIT_DATA_READY,
+	IDH_RAS_POISON_READY,
+	IDH_PF_SOFT_FLR_NOTIFICATION,
+	IDH_RAS_ERROR_DETECTED,
+	IDH_RAS_ERROR_COUNT_READY = 11,
+	IDH_RAS_CPER_DUMP_READY = 14,
+	IDH_RAS_BAD_PAGES_READY = 15,
+	IDH_RAS_BAD_PAGES_NOTIFICATION = 16,
+	IDH_UNRECOV_ERR_NOTIFICATION = 17,
+	IDH_REQ_RAS_CHK_CRITI_READY = 18,
 
 	IDH_TEXT_MESSAGE = 255,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c
index 288c414babdf..e1d63bed84bf 100644
--- a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c
+++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c
@@ -334,7 +334,7 @@ static void xgpu_vi_mailbox_send_ack(struct amdgpu_device *adev)
 			break;
 		}
 		mdelay(1);
-		timeout -=1;
+		timeout -= 1;
 
 		reg = RREG32_NO_KIQ(mmMAILBOX_CONTROL);
 	}
@@ -515,12 +515,6 @@ static void xgpu_vi_mailbox_flr_work(struct work_struct *work)
 	struct amdgpu_virt *virt = container_of(work, struct amdgpu_virt, flr_work);
 	struct amdgpu_device *adev = container_of(virt, struct amdgpu_device, virt);
 
-	/* wait until RCV_MSG become 3 */
-	if (xgpu_vi_poll_msg(adev, IDH_FLR_NOTIFICATION_CMPL)) {
-		pr_err("failed to receive FLR_CMPL\n");
-		return;
-	}
-
 	/* Trigger recovery due to world switch failure */
 	if (amdgpu_device_should_recover_gpu(adev)) {
 		struct amdgpu_reset_context reset_context;
@@ -529,6 +523,7 @@ static void xgpu_vi_mailbox_flr_work(struct work_struct *work)
 		reset_context.method = AMD_RESET_METHOD_NONE;
 		reset_context.reset_req_dev = adev;
 		clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
+		set_bit(AMDGPU_HOST_FLR, &reset_context.flags);
 
 		amdgpu_device_gpu_recover(adev, NULL, &reset_context);
 	}
@@ -560,7 +555,7 @@ static int xgpu_vi_mailbox_rcv_irq(struct amdgpu_device *adev,
 		r = xgpu_vi_mailbox_rcv_msg(adev, IDH_FLR_NOTIFICATION);
 
 		/* only handle FLR_NOTIFY now */
-		if (!r && !amdgpu_in_reset(adev))
+		if (!r)
 			WARN_ONCE(!amdgpu_reset_domain_schedule(adev->reset_domain,
 								&adev->virt.flr_work),
 				  "Failed to queue work! at %s",
diff --git a/drivers/gpu/drm/amd/amdgpu/navi10_ih.c b/drivers/gpu/drm/amd/amdgpu/navi10_ih.c
index eec13cb5bf75..4cd325149b63 100644
--- a/drivers/gpu/drm/amd/amdgpu/navi10_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/navi10_ih.c
@@ -107,7 +107,7 @@ force_update_wptr_for_self_int(struct amdgpu_device *adev,
 {
 	u32 ih_cntl, ih_rb_cntl;
 
-	if (adev->ip_versions[OSSSYS_HWIP][0] < IP_VERSION(5, 0, 3))
+	if (amdgpu_ip_version(adev, OSSSYS_HWIP, 0) < IP_VERSION(5, 0, 3))
 		return;
 
 	ih_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_CNTL2);
@@ -330,7 +330,7 @@ static int navi10_ih_irq_init(struct amdgpu_device *adev)
 
 	if (unlikely(adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT)) {
 		if (ih[0]->use_bus_addr) {
-			switch (adev->ip_versions[OSSSYS_HWIP][0]) {
+			switch (amdgpu_ip_version(adev, OSSSYS_HWIP, 0)) {
 			case IP_VERSION(5, 0, 3):
 			case IP_VERSION(5, 2, 0):
 			case IP_VERSION(5, 2, 1):
@@ -434,14 +434,19 @@ static u32 navi10_ih_get_wptr(struct amdgpu_device *adev,
 	 * this should allow us to catch up.
 	 */
 	tmp = (wptr + 32) & ih->ptr_mask;
-	dev_warn(adev->dev, "IH ring buffer overflow "
-		 "(0x%08X, 0x%08X, 0x%08X)\n",
-		 wptr, ih->rptr, tmp);
+	dev_warn(adev->dev, "%s ring buffer overflow (0x%08X, 0x%08X, 0x%08X)\n",
+		 amdgpu_ih_ring_name(adev, ih), wptr, ih->rptr, tmp);
 	ih->rptr = tmp;
 
 	tmp = RREG32_NO_KIQ(ih_regs->ih_rb_cntl);
 	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
 	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+
+	/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+	 * can be detected.
+	 */
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
 out:
 	return (wptr & ih->ptr_mask);
 }
@@ -536,19 +541,19 @@ static void navi10_ih_set_self_irq_funcs(struct amdgpu_device *adev)
 	adev->irq.self_irq.funcs = &navi10_ih_self_irq_funcs;
 }
 
-static int navi10_ih_early_init(void *handle)
+static int navi10_ih_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	navi10_ih_set_interrupt_funcs(adev);
 	navi10_ih_set_self_irq_funcs(adev);
 	return 0;
 }
 
-static int navi10_ih_sw_init(void *handle)
+static int navi10_ih_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool use_bus_addr;
 
 	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_IH, 0,
@@ -565,7 +570,7 @@ static int navi10_ih_sw_init(void *handle)
 		use_bus_addr = false;
 	else
 		use_bus_addr = true;
-	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 256 * 1024, use_bus_addr);
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, IH_RING_SIZE, use_bus_addr);
 	if (r)
 		return r;
 
@@ -578,7 +583,7 @@ static int navi10_ih_sw_init(void *handle)
 	/* initialize ih control registers offset */
 	navi10_ih_init_register_offset(adev);
 
-	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, PAGE_SIZE, true);
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, IH_SW_RING_SIZE, true);
 	if (r)
 		return r;
 
@@ -587,58 +592,52 @@ static int navi10_ih_sw_init(void *handle)
 	return r;
 }
 
-static int navi10_ih_sw_fini(void *handle)
+static int navi10_ih_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_fini_sw(adev);
 
 	return 0;
 }
 
-static int navi10_ih_hw_init(void *handle)
+static int navi10_ih_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return navi10_ih_irq_init(adev);
 }
 
-static int navi10_ih_hw_fini(void *handle)
+static int navi10_ih_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	navi10_ih_irq_disable(adev);
+	navi10_ih_irq_disable(ip_block->adev);
 
 	return 0;
 }
 
-static int navi10_ih_suspend(void *handle)
+static int navi10_ih_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return navi10_ih_hw_fini(adev);
+	return navi10_ih_hw_fini(ip_block);
 }
 
-static int navi10_ih_resume(void *handle)
+static int navi10_ih_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return navi10_ih_hw_init(adev);
+	return navi10_ih_hw_init(ip_block);
 }
 
-static bool navi10_ih_is_idle(void *handle)
+static bool navi10_ih_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* todo */
 	return true;
 }
 
-static int navi10_ih_wait_for_idle(void *handle)
+static int navi10_ih_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* todo */
 	return -ETIMEDOUT;
 }
 
-static int navi10_ih_soft_reset(void *handle)
+static int navi10_ih_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	/* todo */
 	return 0;
@@ -665,40 +664,35 @@ static void navi10_ih_update_clockgating_state(struct amdgpu_device *adev,
 		if (def != data)
 			WREG32_SOC15(OSSSYS, 0, mmIH_CLK_CTRL, data);
 	}
-
-	return;
 }
 
-static int navi10_ih_set_clockgating_state(void *handle,
+static int navi10_ih_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	navi10_ih_update_clockgating_state(adev,
 				state == AMD_CG_STATE_GATE);
 	return 0;
 }
 
-static int navi10_ih_set_powergating_state(void *handle,
+static int navi10_ih_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_powergating_state state)
 {
 	return 0;
 }
 
-static void navi10_ih_get_clockgating_state(void *handle, u64 *flags)
+static void navi10_ih_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (!RREG32_SOC15(OSSSYS, 0, mmIH_CLK_CTRL))
 		*flags |= AMD_CG_SUPPORT_IH_CG;
-
-	return;
 }
 
 static const struct amd_ip_funcs navi10_ih_ip_funcs = {
 	.name = "navi10_ih",
 	.early_init = navi10_ih_early_init,
-	.late_init = NULL,
 	.sw_init = navi10_ih_sw_init,
 	.sw_fini = navi10_ih_sw_fini,
 	.hw_init = navi10_ih_hw_init,
@@ -726,8 +720,7 @@ static void navi10_ih_set_interrupt_funcs(struct amdgpu_device *adev)
 		adev->irq.ih_funcs = &navi10_ih_funcs;
 }
 
-const struct amdgpu_ip_block_version navi10_ih_ip_block =
-{
+const struct amdgpu_ip_block_version navi10_ih_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_IH,
 	.major = 5,
 	.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/navi10_sdma_pkt_open.h b/drivers/gpu/drm/amd/amdgpu/navi10_sdma_pkt_open.h
index a5b60c9a2418..c88284ff92d8 100644
--- a/drivers/gpu/drm/amd/amdgpu/navi10_sdma_pkt_open.h
+++ b/drivers/gpu/drm/amd/amdgpu/navi10_sdma_pkt_open.h
@@ -68,6 +68,7 @@
 #define SDMA_SUBOP_POLL_REG_WRITE_MEM  1
 #define SDMA_SUBOP_POLL_DBIT_WRITE_MEM  2
 #define SDMA_SUBOP_POLL_MEM_VERIFY  3
+#define SDMA_SUBOP_VM_INVALIDATION  4
 #define HEADER_AGENT_DISPATCH  4
 #define HEADER_BARRIER  5
 #define SDMA_OP_AQL_COPY  0
@@ -4041,6 +4042,69 @@
 
 
 /*
+** Definitions for SDMA_PKT_VM_INVALIDATION packet
+*/
+
+/*define for HEADER word*/
+/*define for op field*/
+#define SDMA_PKT_VM_INVALIDATION_HEADER_op_offset 0
+#define SDMA_PKT_VM_INVALIDATION_HEADER_op_mask   0x000000FF
+#define SDMA_PKT_VM_INVALIDATION_HEADER_op_shift  0
+#define SDMA_PKT_VM_INVALIDATION_HEADER_OP(x) (((x) & SDMA_PKT_VM_INVALIDATION_HEADER_op_mask) << SDMA_PKT_VM_INVALIDATION_HEADER_op_shift)
+
+/*define for sub_op field*/
+#define SDMA_PKT_VM_INVALIDATION_HEADER_sub_op_offset 0
+#define SDMA_PKT_VM_INVALIDATION_HEADER_sub_op_mask   0x000000FF
+#define SDMA_PKT_VM_INVALIDATION_HEADER_sub_op_shift  8
+#define SDMA_PKT_VM_INVALIDATION_HEADER_SUB_OP(x) (((x) & SDMA_PKT_VM_INVALIDATION_HEADER_sub_op_mask) << SDMA_PKT_VM_INVALIDATION_HEADER_sub_op_shift)
+
+/*define for gfx_eng_id field*/
+#define SDMA_PKT_VM_INVALIDATION_HEADER_gfx_eng_id_offset 0
+#define SDMA_PKT_VM_INVALIDATION_HEADER_gfx_eng_id_mask   0x0000001F
+#define SDMA_PKT_VM_INVALIDATION_HEADER_gfx_eng_id_shift  16
+#define SDMA_PKT_VM_INVALIDATION_HEADER_GFX_ENG_ID(x) (((x) & SDMA_PKT_VM_INVALIDATION_HEADER_gfx_eng_id_mask) << SDMA_PKT_VM_INVALIDATION_HEADER_gfx_eng_id_shift)
+
+/*define for mm_eng_id field*/
+#define SDMA_PKT_VM_INVALIDATION_HEADER_mm_eng_id_offset 0
+#define SDMA_PKT_VM_INVALIDATION_HEADER_mm_eng_id_mask   0x0000001F
+#define SDMA_PKT_VM_INVALIDATION_HEADER_mm_eng_id_shift  24
+#define SDMA_PKT_VM_INVALIDATION_HEADER_MM_ENG_ID(x) (((x) & SDMA_PKT_VM_INVALIDATION_HEADER_mm_eng_id_mask) << SDMA_PKT_VM_INVALIDATION_HEADER_mm_eng_id_shift)
+
+/*define for INVALIDATEREQ word*/
+/*define for invalidatereq field*/
+#define SDMA_PKT_VM_INVALIDATION_INVALIDATEREQ_invalidatereq_offset 1
+#define SDMA_PKT_VM_INVALIDATION_INVALIDATEREQ_invalidatereq_mask   0xFFFFFFFF
+#define SDMA_PKT_VM_INVALIDATION_INVALIDATEREQ_invalidatereq_shift  0
+#define SDMA_PKT_VM_INVALIDATION_INVALIDATEREQ_INVALIDATEREQ(x) (((x) & SDMA_PKT_VM_INVALIDATION_INVALIDATEREQ_invalidatereq_mask) << SDMA_PKT_VM_INVALIDATION_INVALIDATEREQ_invalidatereq_shift)
+
+/*define for ADDRESSRANGELO word*/
+/*define for addressrangelo field*/
+#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGELO_addressrangelo_offset 2
+#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGELO_addressrangelo_mask   0xFFFFFFFF
+#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGELO_addressrangelo_shift  0
+#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGELO_ADDRESSRANGELO(x) (((x) & SDMA_PKT_VM_INVALIDATION_ADDRESSRANGELO_addressrangelo_mask) << SDMA_PKT_VM_INVALIDATION_ADDRESSRANGELO_addressrangelo_shift)
+
+/*define for ADDRESSRANGEHI word*/
+/*define for invalidateack field*/
+#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_invalidateack_offset 3
+#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_invalidateack_mask   0x0000FFFF
+#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_invalidateack_shift  0
+#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_INVALIDATEACK(x) (((x) & SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_invalidateack_mask) << SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_invalidateack_shift)
+
+/*define for addressrangehi field*/
+#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_addressrangehi_offset 3
+#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_addressrangehi_mask   0x0000001F
+#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_addressrangehi_shift  16
+#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_ADDRESSRANGEHI(x) (((x) & SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_addressrangehi_mask) << SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_addressrangehi_shift)
+
+/*define for reserved field*/
+#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_reserved_offset 3
+#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_reserved_mask   0x000001FF
+#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_reserved_shift  23
+#define SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_RESERVED(x) (((x) & SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_reserved_mask) << SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_reserved_shift)
+
+
+/*
 ** Definitions for SDMA_PKT_ATOMIC packet
 */
 
diff --git a/drivers/gpu/drm/amd/amdgpu/nbif_v6_3_1.c b/drivers/gpu/drm/amd/amdgpu/nbif_v6_3_1.c
new file mode 100644
index 000000000000..9b4025c39e44
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/nbif_v6_3_1.c
@@ -0,0 +1,554 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "amdgpu.h"
+#include "nbif_v6_3_1.h"
+
+#include "nbif/nbif_6_3_1_offset.h"
+#include "nbif/nbif_6_3_1_sh_mask.h"
+#include "pcie/pcie_6_1_0_offset.h"
+#include "pcie/pcie_6_1_0_sh_mask.h"
+#include "ivsrcid/nbio/irqsrcs_nbif_7_4.h"
+#include <uapi/linux/kfd_ioctl.h>
+
+static void nbif_v6_3_1_remap_hdp_registers(struct amdgpu_device *adev)
+{
+	WREG32_SOC15(NBIO, 0, regBIF_BX0_REMAP_HDP_MEM_FLUSH_CNTL,
+		adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL);
+	WREG32_SOC15(NBIO, 0, regBIF_BX0_REMAP_HDP_REG_FLUSH_CNTL,
+		adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_REG_FLUSH_CNTL);
+}
+
+static u32 nbif_v6_3_1_get_rev_id(struct amdgpu_device *adev)
+{
+	u32 tmp = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_DEV0_EPF0_STRAP0);
+
+	tmp &= RCC_STRAP0_RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0_MASK;
+	tmp >>= RCC_STRAP0_RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0__SHIFT;
+
+	return tmp;
+}
+
+static void nbif_v6_3_1_mc_access_enable(struct amdgpu_device *adev, bool enable)
+{
+	if (enable)
+		WREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_FB_EN,
+			     BIF_BX0_BIF_FB_EN__FB_READ_EN_MASK |
+			     BIF_BX0_BIF_FB_EN__FB_WRITE_EN_MASK);
+	else
+		WREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_FB_EN, 0);
+}
+
+static u32 nbif_v6_3_1_get_memsize(struct amdgpu_device *adev)
+{
+	return RREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF0_RCC_CONFIG_MEMSIZE);
+}
+
+static void nbif_v6_3_1_sdma_doorbell_range(struct amdgpu_device *adev,
+					    int instance, bool use_doorbell,
+					    int doorbell_index,
+					    int doorbell_size)
+{
+	if (instance == 0) {
+		u32 doorbell_range = RREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_2_CTRL);
+
+		if (use_doorbell) {
+			doorbell_range = REG_SET_FIELD(doorbell_range,
+						       GDC_S2A0_S2A_DOORBELL_ENTRY_2_CTRL,
+						       S2A_DOORBELL_PORT2_ENABLE,
+						       0x1);
+			doorbell_range = REG_SET_FIELD(doorbell_range,
+						       GDC_S2A0_S2A_DOORBELL_ENTRY_2_CTRL,
+						       S2A_DOORBELL_PORT2_AWID,
+						       0xe);
+			doorbell_range = REG_SET_FIELD(doorbell_range,
+						       GDC_S2A0_S2A_DOORBELL_ENTRY_2_CTRL,
+						       S2A_DOORBELL_PORT2_RANGE_OFFSET,
+						       doorbell_index);
+			doorbell_range = REG_SET_FIELD(doorbell_range,
+						       GDC_S2A0_S2A_DOORBELL_ENTRY_2_CTRL,
+						       S2A_DOORBELL_PORT2_RANGE_SIZE,
+						       doorbell_size);
+			doorbell_range = REG_SET_FIELD(doorbell_range,
+						       GDC_S2A0_S2A_DOORBELL_ENTRY_2_CTRL,
+						       S2A_DOORBELL_PORT2_AWADDR_31_28_VALUE,
+						       0x3);
+		} else
+			doorbell_range = REG_SET_FIELD(doorbell_range,
+						       GDC_S2A0_S2A_DOORBELL_ENTRY_2_CTRL,
+						       S2A_DOORBELL_PORT2_RANGE_SIZE,
+						       0);
+
+		WREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_2_CTRL, doorbell_range);
+	}
+}
+
+static void nbif_v6_3_1_vcn_doorbell_range(struct amdgpu_device *adev,
+					   bool use_doorbell, int doorbell_index,
+					   int instance)
+{
+	u32 doorbell_range;
+
+	if (instance)
+		doorbell_range = RREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_5_CTRL);
+	else
+		doorbell_range = RREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_4_CTRL);
+
+	if (use_doorbell) {
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC_S2A0_S2A_DOORBELL_ENTRY_4_CTRL,
+					       S2A_DOORBELL_PORT4_ENABLE,
+					       0x1);
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC_S2A0_S2A_DOORBELL_ENTRY_4_CTRL,
+					       S2A_DOORBELL_PORT4_AWID,
+					       instance ? 0x7 : 0x4);
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC_S2A0_S2A_DOORBELL_ENTRY_4_CTRL,
+					       S2A_DOORBELL_PORT4_RANGE_OFFSET,
+					       doorbell_index);
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC_S2A0_S2A_DOORBELL_ENTRY_4_CTRL,
+					       S2A_DOORBELL_PORT4_RANGE_SIZE,
+					       8);
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC_S2A0_S2A_DOORBELL_ENTRY_4_CTRL,
+					       S2A_DOORBELL_PORT4_AWADDR_31_28_VALUE,
+					       instance ? 0x7 : 0x4);
+	} else
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC_S2A0_S2A_DOORBELL_ENTRY_4_CTRL,
+					       S2A_DOORBELL_PORT4_RANGE_SIZE,
+					       0);
+
+	if (instance)
+		WREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_5_CTRL, doorbell_range);
+	else
+		WREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_4_CTRL, doorbell_range);
+}
+
+static void nbif_v6_3_1_gc_doorbell_init(struct amdgpu_device *adev)
+{
+	WREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_0_CTRL, 0x30000007);
+	WREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_3_CTRL, 0x3000000d);
+}
+
+static void nbif_v6_3_1_enable_doorbell_aperture(struct amdgpu_device *adev,
+						 bool enable)
+{
+	WREG32_FIELD15_PREREG(NBIO, 0, RCC_DEV0_EPF0_RCC_DOORBELL_APER_EN,
+			BIF_DOORBELL_APER_EN, enable ? 1 : 0);
+}
+
+static void
+nbif_v6_3_1_enable_doorbell_selfring_aperture(struct amdgpu_device *adev,
+					      bool enable)
+{
+	u32 tmp = 0;
+
+	if (enable) {
+		tmp = REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL,
+				    DOORBELL_SELFRING_GPA_APER_EN, 1) |
+		      REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL,
+				    DOORBELL_SELFRING_GPA_APER_MODE, 1) |
+		      REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL,
+				    DOORBELL_SELFRING_GPA_APER_SIZE, 0);
+
+		WREG32_SOC15(NBIO, 0, regBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_BASE_LOW,
+			     lower_32_bits(adev->doorbell.base));
+		WREG32_SOC15(NBIO, 0, regBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_BASE_HIGH,
+			     upper_32_bits(adev->doorbell.base));
+	}
+
+	WREG32_SOC15(NBIO, 0, regBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, tmp);
+}
+
+static void nbif_v6_3_1_ih_doorbell_range(struct amdgpu_device *adev,
+					  bool use_doorbell, int doorbell_index)
+{
+	u32 ih_doorbell_range = RREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_1_CTRL);
+
+	if (use_doorbell) {
+		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
+						  GDC_S2A0_S2A_DOORBELL_ENTRY_1_CTRL,
+						  S2A_DOORBELL_PORT1_ENABLE,
+						  0x1);
+		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
+						  GDC_S2A0_S2A_DOORBELL_ENTRY_1_CTRL,
+						  S2A_DOORBELL_PORT1_AWID,
+						  0x0);
+		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
+						  GDC_S2A0_S2A_DOORBELL_ENTRY_1_CTRL,
+						  S2A_DOORBELL_PORT1_RANGE_OFFSET,
+						  doorbell_index);
+		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
+						  GDC_S2A0_S2A_DOORBELL_ENTRY_1_CTRL,
+						  S2A_DOORBELL_PORT1_RANGE_SIZE,
+						  2);
+		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
+						  GDC_S2A0_S2A_DOORBELL_ENTRY_1_CTRL,
+						  S2A_DOORBELL_PORT1_AWADDR_31_28_VALUE,
+						  0x0);
+	} else
+		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
+						  GDC_S2A0_S2A_DOORBELL_ENTRY_1_CTRL,
+						  S2A_DOORBELL_PORT1_RANGE_SIZE,
+						  0);
+
+	WREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_1_CTRL, ih_doorbell_range);
+}
+
+static void nbif_v6_3_1_ih_control(struct amdgpu_device *adev)
+{
+	u32 interrupt_cntl;
+
+	/* setup interrupt control */
+	WREG32_SOC15(NBIO, 0, regBIF_BX0_INTERRUPT_CNTL2, adev->dummy_page_addr >> 8);
+
+	interrupt_cntl = RREG32_SOC15(NBIO, 0, regBIF_BX0_INTERRUPT_CNTL);
+	/*
+	 * BIF_BX0_INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=0 - dummy read disabled with msi, enabled without msi
+	 * BIF_BX0_INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=1 - dummy read controlled by IH_DUMMY_RD_EN
+	 */
+	interrupt_cntl = REG_SET_FIELD(interrupt_cntl, BIF_BX0_INTERRUPT_CNTL,
+				       IH_DUMMY_RD_OVERRIDE, 0);
+
+	/* BIF_BX0_INTERRUPT_CNTL__IH_REQ_NONSNOOP_EN_MASK=1 if ring is in non-cacheable memory, e.g., vram */
+	interrupt_cntl = REG_SET_FIELD(interrupt_cntl, BIF_BX0_INTERRUPT_CNTL,
+				       IH_REQ_NONSNOOP_EN, 0);
+
+	WREG32_SOC15(NBIO, 0, regBIF_BX0_INTERRUPT_CNTL, interrupt_cntl);
+}
+
+static void
+nbif_v6_3_1_update_medium_grain_clock_gating(struct amdgpu_device *adev,
+					     bool enable)
+{
+}
+
+static void
+nbif_v6_3_1_update_medium_grain_light_sleep(struct amdgpu_device *adev,
+					    bool enable)
+{
+}
+
+static void
+nbif_v6_3_1_get_clockgating_state(struct amdgpu_device *adev,
+				  u64 *flags)
+{
+}
+
+static u32 nbif_v6_3_1_get_hdp_flush_req_offset(struct amdgpu_device *adev)
+{
+	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_GPU_HDP_FLUSH_REQ);
+}
+
+static u32 nbif_v6_3_1_get_hdp_flush_done_offset(struct amdgpu_device *adev)
+{
+	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_GPU_HDP_FLUSH_DONE);
+}
+
+static u32 nbif_v6_3_1_get_pcie_index_offset(struct amdgpu_device *adev)
+{
+	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_RSMU_INDEX);
+}
+
+static u32 nbif_v6_3_1_get_pcie_data_offset(struct amdgpu_device *adev)
+{
+	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_RSMU_DATA);
+}
+
+const struct nbio_hdp_flush_reg nbif_v6_3_1_hdp_flush_reg = {
+	.ref_and_mask_cp0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP0_MASK,
+	.ref_and_mask_cp1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP1_MASK,
+	.ref_and_mask_cp2 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP2_MASK,
+	.ref_and_mask_cp3 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP3_MASK,
+	.ref_and_mask_cp4 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP4_MASK,
+	.ref_and_mask_cp5 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP5_MASK,
+	.ref_and_mask_cp6 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP6_MASK,
+	.ref_and_mask_cp7 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP7_MASK,
+	.ref_and_mask_cp8 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP8_MASK,
+	.ref_and_mask_cp9 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP9_MASK,
+	.ref_and_mask_sdma0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA0_MASK,
+	.ref_and_mask_sdma1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA1_MASK,
+};
+
+static void nbif_v6_3_1_init_registers(struct amdgpu_device *adev)
+{
+	uint32_t data;
+
+	data = RREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF2_STRAP2);
+	data &= ~RCC_DEV0_EPF2_STRAP2__STRAP_NO_SOFT_RESET_DEV0_F2_MASK;
+	WREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF2_STRAP2, data);
+}
+
+static u32 nbif_v6_3_1_get_rom_offset(struct amdgpu_device *adev)
+{
+	u32 data, rom_offset;
+
+	data = RREG32_SOC15(NBIO, 0, regREGS_ROM_OFFSET_CTRL);
+	rom_offset = REG_GET_FIELD(data, REGS_ROM_OFFSET_CTRL, ROM_OFFSET);
+
+	return rom_offset;
+}
+
+#ifdef CONFIG_PCIEASPM
+static void nbif_v6_3_1_program_ltr(struct amdgpu_device *adev)
+{
+	uint32_t def, data;
+	u16 devctl2;
+
+	def = RREG32_SOC15(NBIO, 0, regRCC_EP_DEV0_0_EP_PCIE_TX_LTR_CNTL);
+	data = 0x35EB;
+	data &= ~RCC_EP_DEV0_0_EP_PCIE_TX_LTR_CNTL__LTR_PRIV_MSG_DIS_IN_PM_NON_D0_MASK;
+	data &= ~RCC_EP_DEV0_0_EP_PCIE_TX_LTR_CNTL__LTR_PRIV_RST_LTR_IN_DL_DOWN_MASK;
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regRCC_EP_DEV0_0_EP_PCIE_TX_LTR_CNTL, data);
+
+	def = data = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP2);
+	data &= ~RCC_STRAP0_RCC_BIF_STRAP2__STRAP_LTR_IN_ASPML1_DIS_MASK;
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP2, data);
+
+	pcie_capability_read_word(adev->pdev, PCI_EXP_DEVCTL2, &devctl2);
+
+	if (adev->pdev->ltr_path == (devctl2 & PCI_EXP_DEVCTL2_LTR_EN))
+		return;
+
+	if (adev->pdev->ltr_path)
+		pcie_capability_set_word(adev->pdev, PCI_EXP_DEVCTL2, PCI_EXP_DEVCTL2_LTR_EN);
+	else
+		pcie_capability_clear_word(adev->pdev, PCI_EXP_DEVCTL2, PCI_EXP_DEVCTL2_LTR_EN);
+}
+#endif
+
+static void nbif_v6_3_1_program_aspm(struct amdgpu_device *adev)
+{
+#ifdef CONFIG_PCIEASPM
+	uint32_t def, data;
+	u16 devctl2, ltr;
+
+	def = data = RREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL);
+	data &= ~PCIE_LC_CNTL__LC_L1_INACTIVITY_MASK;
+	data &= ~PCIE_LC_CNTL__LC_L0S_INACTIVITY_MASK;
+	data |= PCIE_LC_CNTL__LC_PMI_TO_L1_DIS_MASK;
+	if (def != data)
+		WREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL, data);
+
+	def = data = RREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL7);
+	data |= PCIE_LC_CNTL7__LC_NBIF_ASPM_INPUT_EN_MASK;
+	if (def != data)
+		WREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL7, data);
+
+	def = data = RREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL3);
+	data |= PCIE_LC_CNTL3__LC_DSC_DONT_ENTER_L23_AFTER_PME_ACK_MASK;
+	if (def != data)
+		WREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL3, data);
+
+	def = data = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP3);
+	data &= ~RCC_STRAP0_RCC_BIF_STRAP3__STRAP_VLINK_ASPM_IDLE_TIMER_MASK;
+	data &= ~RCC_STRAP0_RCC_BIF_STRAP3__STRAP_VLINK_PM_L1_ENTRY_TIMER_MASK;
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP3, data);
+
+	def = data = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP5);
+	data &= ~RCC_STRAP0_RCC_BIF_STRAP5__STRAP_VLINK_LDN_ENTRY_TIMER_MASK;
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP5, data);
+
+	pcie_capability_read_word(adev->pdev, PCI_EXP_DEVCTL2, &devctl2);
+	data = def = devctl2;
+	data &= ~PCI_EXP_DEVCTL2_LTR_EN;
+	if (def != data)
+		pcie_capability_set_word(adev->pdev, PCI_EXP_DEVCTL2, (u16)data);
+
+	ltr = pci_find_ext_capability(adev->pdev, PCI_EXT_CAP_ID_LTR);
+
+	if (ltr) {
+		pci_write_config_dword(adev->pdev, ltr + PCI_LTR_MAX_SNOOP_LAT, 0x10011001);
+	}
+
+#if 0
+	/* regPSWUSP0_PCIE_LC_CNTL2 should be replace by PCIE_LC_CNTL2 or someone else ? */
+	def = data = RREG32_SOC15(NBIO, 0, regPSWUSP0_PCIE_LC_CNTL2);
+	data |= PSWUSP0_PCIE_LC_CNTL2__LC_ALLOW_PDWN_IN_L1_MASK |
+		PSWUSP0_PCIE_LC_CNTL2__LC_ALLOW_PDWN_IN_L23_MASK;
+	data &= ~PSWUSP0_PCIE_LC_CNTL2__LC_RCV_L0_TO_RCV_L0S_DIS_MASK;
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regPSWUSP0_PCIE_LC_CNTL2, data);
+#endif
+	def = data = RREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL4);
+	data |= PCIE_LC_CNTL4__LC_L1_POWERDOWN_MASK;
+	if (def != data)
+		WREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL4, data);
+
+	def = data = RREG32_SOC15(PCIE, 0, regPCIE_LC_RXRECOVER_RXSTANDBY_CNTL);
+	data |= PCIE_LC_RXRECOVER_RXSTANDBY_CNTL__LC_RX_L0S_STANDBY_EN_MASK;
+	if (def != data)
+		WREG32_SOC15(PCIE, 0, regPCIE_LC_RXRECOVER_RXSTANDBY_CNTL, data);
+
+	nbif_v6_3_1_program_ltr(adev);
+
+	def = data = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP3);
+	data |= 0x5DE0 << RCC_STRAP0_RCC_BIF_STRAP3__STRAP_VLINK_ASPM_IDLE_TIMER__SHIFT;
+	data |= 0x0010 << RCC_STRAP0_RCC_BIF_STRAP3__STRAP_VLINK_PM_L1_ENTRY_TIMER__SHIFT;
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP3, data);
+
+	def = data = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP5);
+	data |= 0x0010 << RCC_STRAP0_RCC_BIF_STRAP5__STRAP_VLINK_LDN_ENTRY_TIMER__SHIFT;
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP5, data);
+
+	def = data = RREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL);
+	data |= 0x0 << PCIE_LC_CNTL__LC_L0S_INACTIVITY__SHIFT;
+	data |= 0x9 << PCIE_LC_CNTL__LC_L1_INACTIVITY__SHIFT;
+	data &= ~PCIE_LC_CNTL__LC_PMI_TO_L1_DIS_MASK;
+	if (def != data)
+		WREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL, data);
+
+	def = data = RREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL3);
+	data &= ~PCIE_LC_CNTL3__LC_DSC_DONT_ENTER_L23_AFTER_PME_ACK_MASK;
+	if (def != data)
+		WREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL3, data);
+#endif
+}
+
+#define MMIO_REG_HOLE_OFFSET (0x80000 - PAGE_SIZE)
+
+static void nbif_v6_3_1_set_reg_remap(struct amdgpu_device *adev)
+{
+	if (!amdgpu_sriov_vf(adev) && (PAGE_SIZE <= 4096)) {
+		adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET;
+		adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET;
+	} else {
+		adev->rmmio_remap.reg_offset = SOC15_REG_OFFSET(NBIO, 0,
+			regBIF_BX_PF0_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2;
+		adev->rmmio_remap.bus_addr = 0;
+	}
+}
+
+const struct amdgpu_nbio_funcs nbif_v6_3_1_funcs = {
+	.get_hdp_flush_req_offset = nbif_v6_3_1_get_hdp_flush_req_offset,
+	.get_hdp_flush_done_offset = nbif_v6_3_1_get_hdp_flush_done_offset,
+	.get_pcie_index_offset = nbif_v6_3_1_get_pcie_index_offset,
+	.get_pcie_data_offset = nbif_v6_3_1_get_pcie_data_offset,
+	.get_rev_id = nbif_v6_3_1_get_rev_id,
+	.mc_access_enable = nbif_v6_3_1_mc_access_enable,
+	.get_memsize = nbif_v6_3_1_get_memsize,
+	.sdma_doorbell_range = nbif_v6_3_1_sdma_doorbell_range,
+	.vcn_doorbell_range = nbif_v6_3_1_vcn_doorbell_range,
+	.gc_doorbell_init = nbif_v6_3_1_gc_doorbell_init,
+	.enable_doorbell_aperture = nbif_v6_3_1_enable_doorbell_aperture,
+	.enable_doorbell_selfring_aperture = nbif_v6_3_1_enable_doorbell_selfring_aperture,
+	.ih_doorbell_range = nbif_v6_3_1_ih_doorbell_range,
+	.update_medium_grain_clock_gating = nbif_v6_3_1_update_medium_grain_clock_gating,
+	.update_medium_grain_light_sleep = nbif_v6_3_1_update_medium_grain_light_sleep,
+	.get_clockgating_state = nbif_v6_3_1_get_clockgating_state,
+	.ih_control = nbif_v6_3_1_ih_control,
+	.init_registers = nbif_v6_3_1_init_registers,
+	.remap_hdp_registers = nbif_v6_3_1_remap_hdp_registers,
+	.get_rom_offset = nbif_v6_3_1_get_rom_offset,
+	.program_aspm = nbif_v6_3_1_program_aspm,
+	.set_reg_remap = nbif_v6_3_1_set_reg_remap,
+};
+
+
+static int nbif_v6_3_1_set_ras_err_event_athub_irq_state(struct amdgpu_device *adev,
+						       struct amdgpu_irq_src *src,
+						       unsigned type,
+						       enum amdgpu_interrupt_state state)
+{
+	/* The ras_controller_irq enablement should be done in psp bl when it
+	 * tries to enable ras feature. Driver only need to set the correct interrupt
+	 * vector for bare-metal and sriov use case respectively
+	 */
+	uint32_t bif_doorbell_int_cntl;
+
+	bif_doorbell_int_cntl = RREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_DOORBELL_INT_CNTL);
+	bif_doorbell_int_cntl = REG_SET_FIELD(bif_doorbell_int_cntl,
+					      BIF_BX0_BIF_DOORBELL_INT_CNTL,
+					      RAS_ATHUB_ERR_EVENT_INTERRUPT_DISABLE,
+					      (state == AMDGPU_IRQ_STATE_ENABLE) ? 0 : 1);
+	WREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_DOORBELL_INT_CNTL, bif_doorbell_int_cntl);
+
+	return 0;
+}
+
+static int nbif_v6_3_1_process_err_event_athub_irq(struct amdgpu_device *adev,
+						 struct amdgpu_irq_src *source,
+						 struct amdgpu_iv_entry *entry)
+{
+	/* By design, the ih cookie for err_event_athub_irq should be written
+	 * to bif ring. since bif ring is not enabled, just leave process callback
+	 * as a dummy one.
+	 */
+	return 0;
+}
+
+static const struct amdgpu_irq_src_funcs nbif_v6_3_1_ras_err_event_athub_irq_funcs = {
+	.set = nbif_v6_3_1_set_ras_err_event_athub_irq_state,
+	.process = nbif_v6_3_1_process_err_event_athub_irq,
+};
+
+static void nbif_v6_3_1_handle_ras_err_event_athub_intr_no_bifring(struct amdgpu_device *adev)
+{
+	uint32_t bif_doorbell_int_cntl;
+
+	bif_doorbell_int_cntl = RREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_DOORBELL_INT_CNTL);
+	if (REG_GET_FIELD(bif_doorbell_int_cntl,
+			  BIF_BX0_BIF_DOORBELL_INT_CNTL,
+			  RAS_ATHUB_ERR_EVENT_INTERRUPT_STATUS)) {
+		/* driver has to clear the interrupt status when bif ring is disabled */
+		bif_doorbell_int_cntl = REG_SET_FIELD(bif_doorbell_int_cntl,
+						BIF_BX0_BIF_DOORBELL_INT_CNTL,
+						RAS_ATHUB_ERR_EVENT_INTERRUPT_CLEAR, 1);
+		WREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_DOORBELL_INT_CNTL, bif_doorbell_int_cntl);
+		amdgpu_ras_global_ras_isr(adev);
+	}
+}
+
+static int nbif_v6_3_1_init_ras_err_event_athub_interrupt(struct amdgpu_device *adev)
+{
+	int r;
+
+	/* init the irq funcs */
+	adev->nbio.ras_err_event_athub_irq.funcs =
+		&nbif_v6_3_1_ras_err_event_athub_irq_funcs;
+	adev->nbio.ras_err_event_athub_irq.num_types = 1;
+
+	/* register ras err event athub interrupt
+	 * nbif v6_3_1 uses the same irq source as nbio v7_4
+	 */
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_BIF,
+			      NBIF_7_4__SRCID__ERREVENT_ATHUB_INTERRUPT,
+			      &adev->nbio.ras_err_event_athub_irq);
+
+	return r;
+}
+
+struct amdgpu_nbio_ras nbif_v6_3_1_ras = {
+	.handle_ras_err_event_athub_intr_no_bifring =
+		nbif_v6_3_1_handle_ras_err_event_athub_intr_no_bifring,
+	.init_ras_err_event_athub_interrupt =
+		nbif_v6_3_1_init_ras_err_event_athub_interrupt,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/nbif_v6_3_1.h b/drivers/gpu/drm/amd/amdgpu/nbif_v6_3_1.h
new file mode 100644
index 000000000000..3afec715a9fe
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/nbif_v6_3_1.h
@@ -0,0 +1,33 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __NBIO_V6_3_1_H__
+#define __NBIO_V6_3_1_H__
+
+#include "soc15_common.h"
+
+extern const struct nbio_hdp_flush_reg nbif_v6_3_1_hdp_flush_reg;
+extern const struct amdgpu_nbio_funcs nbif_v6_3_1_funcs;
+extern struct amdgpu_nbio_ras nbif_v6_3_1_ras;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v2_3.c b/drivers/gpu/drm/amd/amdgpu/nbio_v2_3.c
index aa761ff3a5fa..04041b398781 100644
--- a/drivers/gpu/drm/amd/amdgpu/nbio_v2_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/nbio_v2_3.c
@@ -21,13 +21,13 @@
  *
  */
 #include "amdgpu.h"
-#include "amdgpu_atombios.h"
 #include "nbio_v2_3.h"
 
 #include "nbio/nbio_2_3_default.h"
 #include "nbio/nbio_2_3_offset.h"
 #include "nbio/nbio_2_3_sh_mask.h"
 #include <uapi/linux/kfd_ioctl.h>
+#include <linux/device.h>
 #include <linux/pci.h>
 
 #define smnPCIE_CONFIG_CNTL	0x11180044
@@ -338,15 +338,11 @@ static void nbio_v2_3_init_registers(struct amdgpu_device *adev)
 
 	if (def != data)
 		WREG32_PCIE(smnPCIE_CONFIG_CNTL, data);
-
-	if (amdgpu_sriov_vf(adev))
-		adev->rmmio_remap.reg_offset = SOC15_REG_OFFSET(NBIO, 0,
-			mmBIF_BX_DEV0_EPF0_VF0_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2;
 }
 
 #define NAVI10_PCIE__LC_L0S_INACTIVITY_DEFAULT		0x00000000 // off by default, no gains over L1
-#define NAVI10_PCIE__LC_L1_INACTIVITY_DEFAULT		0x00000009 // 1=1us, 9=1ms
-#define NAVI10_PCIE__LC_L1_INACTIVITY_TBT_DEFAULT	0x0000000E // 4ms
+#define NAVI10_PCIE__LC_L1_INACTIVITY_DEFAULT		0x0000000A // 1=1us, 9=1ms, 10=4ms
+#define NAVI10_PCIE__LC_L1_INACTIVITY_TBT_DEFAULT	0x0000000E // 400ms
 
 static void nbio_v2_3_enable_aspm(struct amdgpu_device *adev,
 				  bool enable)
@@ -361,14 +357,14 @@ static void nbio_v2_3_enable_aspm(struct amdgpu_device *adev,
 
 		data |= NAVI10_PCIE__LC_L0S_INACTIVITY_DEFAULT << PCIE_LC_CNTL__LC_L0S_INACTIVITY__SHIFT;
 
-		if (pci_is_thunderbolt_attached(adev->pdev))
+		if (dev_is_removable(&adev->pdev->dev))
 			data |= NAVI10_PCIE__LC_L1_INACTIVITY_TBT_DEFAULT  << PCIE_LC_CNTL__LC_L1_INACTIVITY__SHIFT;
 		else
 			data |= NAVI10_PCIE__LC_L1_INACTIVITY_DEFAULT << PCIE_LC_CNTL__LC_L1_INACTIVITY__SHIFT;
 
 		data &= ~PCIE_LC_CNTL__LC_PMI_TO_L1_DIS_MASK;
 	} else {
-		/* Disbale ASPM L1 */
+		/* Disable ASPM L1 */
 		data &= ~PCIE_LC_CNTL__LC_L1_INACTIVITY_MASK;
 		/* Disable ASPM TxL0s */
 		data &= ~PCIE_LC_CNTL__LC_L0S_INACTIVITY_MASK;
@@ -479,9 +475,12 @@ static void nbio_v2_3_program_aspm(struct amdgpu_device *adev)
 		WREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP5, data);
 
 	def = data = RREG32_PCIE(smnPCIE_LC_CNTL);
-	data &= ~PCIE_LC_CNTL__LC_L0S_INACTIVITY_MASK;
-	data |= 0x9 << PCIE_LC_CNTL__LC_L1_INACTIVITY__SHIFT;
-	data |= 0x1 << PCIE_LC_CNTL__LC_PMI_TO_L1_DIS__SHIFT;
+	data |= NAVI10_PCIE__LC_L0S_INACTIVITY_DEFAULT << PCIE_LC_CNTL__LC_L0S_INACTIVITY__SHIFT;
+	if (dev_is_removable(&adev->pdev->dev))
+		data |= NAVI10_PCIE__LC_L1_INACTIVITY_TBT_DEFAULT  << PCIE_LC_CNTL__LC_L1_INACTIVITY__SHIFT;
+	else
+		data |= NAVI10_PCIE__LC_L1_INACTIVITY_DEFAULT << PCIE_LC_CNTL__LC_L1_INACTIVITY__SHIFT;
+	data &= ~PCIE_LC_CNTL__LC_PMI_TO_L1_DIS_MASK;
 	if (def != data)
 		WREG32_PCIE(smnPCIE_LC_CNTL, data);
 
@@ -533,7 +532,7 @@ static void nbio_v2_3_clear_doorbell_interrupt(struct amdgpu_device *adev)
 {
 	uint32_t reg, reg_data;
 
-	if (adev->ip_versions[NBIO_HWIP][0] != IP_VERSION(3, 3, 0))
+	if (amdgpu_ip_version(adev, NBIO_HWIP, 0) != IP_VERSION(3, 3, 0))
 		return;
 
 	reg = RREG32_SOC15(NBIO, 0, mmBIF_RB_CNTL);
@@ -549,6 +548,20 @@ static void nbio_v2_3_clear_doorbell_interrupt(struct amdgpu_device *adev)
 	}
 }
 
+#define MMIO_REG_HOLE_OFFSET (0x80000 - PAGE_SIZE)
+
+static void nbio_v2_3_set_reg_remap(struct amdgpu_device *adev)
+{
+	if (!amdgpu_sriov_vf(adev) && (PAGE_SIZE <= 4096)) {
+		adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET;
+		adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET;
+	} else {
+		adev->rmmio_remap.reg_offset = SOC15_REG_OFFSET(NBIO, 0,
+			mmBIF_BX_DEV0_EPF0_VF0_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2;
+		adev->rmmio_remap.bus_addr = 0;
+	}
+}
+
 const struct amdgpu_nbio_funcs nbio_v2_3_funcs = {
 	.get_hdp_flush_req_offset = nbio_v2_3_get_hdp_flush_req_offset,
 	.get_hdp_flush_done_offset = nbio_v2_3_get_hdp_flush_done_offset,
@@ -573,4 +586,5 @@ const struct amdgpu_nbio_funcs nbio_v2_3_funcs = {
 	.apply_lc_spc_mode_wa = nbio_v2_3_apply_lc_spc_mode_wa,
 	.apply_l1_link_width_reconfig_wa = nbio_v2_3_apply_l1_link_width_reconfig_wa,
 	.clear_doorbell_interrupt = nbio_v2_3_clear_doorbell_interrupt,
+	.set_reg_remap = nbio_v2_3_set_reg_remap,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v4_3.c b/drivers/gpu/drm/amd/amdgpu/nbio_v4_3.c
index d5ed9e0e1a5f..f89e5f40e1a5 100644
--- a/drivers/gpu/drm/amd/amdgpu/nbio_v4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/nbio_v4_3.c
@@ -21,7 +21,6 @@
  *
  */
 #include "amdgpu.h"
-#include "amdgpu_atombios.h"
 #include "nbio_v4_3.h"
 
 #include "nbio/nbio_4_3_0_offset.h"
@@ -338,7 +337,7 @@ const struct nbio_hdp_flush_reg nbio_v4_3_hdp_flush_reg = {
 
 static void nbio_v4_3_init_registers(struct amdgpu_device *adev)
 {
-	if (adev->ip_versions[NBIO_HWIP][0] == IP_VERSION(4, 3, 0)) {
+	if (amdgpu_ip_version(adev, NBIO_HWIP, 0) == IP_VERSION(4, 3, 0)) {
 		uint32_t data;
 
 		data = RREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF2_STRAP2);
@@ -389,8 +388,8 @@ static void nbio_v4_3_program_aspm(struct amdgpu_device *adev)
 #ifdef CONFIG_PCIEASPM
 	uint32_t def, data;
 
-	if (!(adev->ip_versions[PCIE_HWIP][0] == IP_VERSION(7, 4, 0)) &&
-	      !(adev->ip_versions[PCIE_HWIP][0] == IP_VERSION(7, 6, 0)))
+	if (!(amdgpu_ip_version(adev, PCIE_HWIP, 0) == IP_VERSION(7, 4, 0)) &&
+	    !(amdgpu_ip_version(adev, PCIE_HWIP, 0) == IP_VERSION(7, 6, 0)))
 		return;
 
 	def = data = RREG32_SOC15(NBIO, 0, regPCIE_LC_CNTL);
@@ -472,6 +471,20 @@ static void nbio_v4_3_program_aspm(struct amdgpu_device *adev)
 #endif
 }
 
+#define MMIO_REG_HOLE_OFFSET (0x80000 - PAGE_SIZE)
+
+static void nbio_v4_3_set_reg_remap(struct amdgpu_device *adev)
+{
+	if (!amdgpu_sriov_vf(adev) && (PAGE_SIZE <= 4096)) {
+		adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET;
+		adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET;
+	} else {
+		adev->rmmio_remap.reg_offset = SOC15_REG_OFFSET(NBIO, 0,
+			regBIF_BX_DEV0_EPF0_VF0_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2;
+		adev->rmmio_remap.bus_addr = 0;
+	}
+}
+
 const struct amdgpu_nbio_funcs nbio_v4_3_funcs = {
 	.get_hdp_flush_req_offset = nbio_v4_3_get_hdp_flush_req_offset,
 	.get_hdp_flush_done_offset = nbio_v4_3_get_hdp_flush_done_offset,
@@ -494,6 +507,7 @@ const struct amdgpu_nbio_funcs nbio_v4_3_funcs = {
 	.remap_hdp_registers = nbio_v4_3_remap_hdp_registers,
 	.get_rom_offset = nbio_v4_3_get_rom_offset,
 	.program_aspm = nbio_v4_3_program_aspm,
+	.set_reg_remap = nbio_v4_3_set_reg_remap,
 };
 
 
@@ -538,6 +552,7 @@ const struct amdgpu_nbio_funcs nbio_v4_3_sriov_funcs = {
 	.init_registers = nbio_v4_3_init_registers,
 	.remap_hdp_registers = nbio_v4_3_remap_hdp_registers,
 	.get_rom_offset = nbio_v4_3_get_rom_offset,
+	.set_reg_remap = nbio_v4_3_set_reg_remap,
 };
 
 static int nbio_v4_3_set_ras_err_event_athub_irq_state(struct amdgpu_device *adev,
diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c b/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c
index 37615a77287b..e911368c1aeb 100644
--- a/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c
@@ -21,7 +21,6 @@
  *
  */
 #include "amdgpu.h"
-#include "amdgpu_atombios.h"
 #include "nbio_v6_1.h"
 
 #include "nbio/nbio_6_1_default.h"
@@ -276,10 +275,6 @@ static void nbio_v6_1_init_registers(struct amdgpu_device *adev)
 
 	if (def != data)
 		WREG32_PCIE(smnPCIE_CI_CNTL, data);
-
-	if (amdgpu_sriov_vf(adev))
-		adev->rmmio_remap.reg_offset = SOC15_REG_OFFSET(NBIO, 0,
-			mmBIF_BX_DEV0_EPF0_VF0_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2;
 }
 
 #ifdef CONFIG_PCIEASPM
@@ -394,6 +389,21 @@ static void nbio_v6_1_program_aspm(struct amdgpu_device *adev)
 #endif
 }
 
+#define MMIO_REG_HOLE_OFFSET (0x80000 - PAGE_SIZE)
+
+static void nbio_v6_1_set_reg_remap(struct amdgpu_device *adev)
+{
+	if (!amdgpu_sriov_vf(adev) && (PAGE_SIZE <= 4096)) {
+		adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET;
+		adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET;
+	} else {
+		adev->rmmio_remap.reg_offset =
+			SOC15_REG_OFFSET(NBIO, 0,
+					 mmBIF_BX_DEV0_EPF0_VF0_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2;
+		adev->rmmio_remap.bus_addr = 0;
+	}
+}
+
 const struct amdgpu_nbio_funcs nbio_v6_1_funcs = {
 	.get_hdp_flush_req_offset = nbio_v6_1_get_hdp_flush_req_offset,
 	.get_hdp_flush_done_offset = nbio_v6_1_get_hdp_flush_done_offset,
@@ -412,5 +422,6 @@ const struct amdgpu_nbio_funcs nbio_v6_1_funcs = {
 	.ih_control = nbio_v6_1_ih_control,
 	.init_registers = nbio_v6_1_init_registers,
 	.remap_hdp_registers = nbio_v6_1_remap_hdp_registers,
-	.program_aspm =  nbio_v6_1_program_aspm,
+	.program_aspm = nbio_v6_1_program_aspm,
+	.set_reg_remap = nbio_v6_1_set_reg_remap,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c
index aa0326d00c72..1569a1e934ec 100644
--- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c
@@ -21,7 +21,6 @@
  *
  */
 #include "amdgpu.h"
-#include "amdgpu_atombios.h"
 #include "nbio_v7_0.h"
 
 #include "nbio/nbio_7_0_default.h"
@@ -271,11 +270,33 @@ const struct nbio_hdp_flush_reg nbio_v7_0_hdp_flush_reg = {
 	.ref_and_mask_sdma1 = GPU_HDP_FLUSH_DONE__SDMA1_MASK,
 };
 
+#define regRCC_DEV0_EPF6_STRAP4                                                                         0xd304
+#define regRCC_DEV0_EPF6_STRAP4_BASE_IDX                                                                5
+
 static void nbio_v7_0_init_registers(struct amdgpu_device *adev)
 {
-	if (amdgpu_sriov_vf(adev))
+	uint32_t data;
+
+	switch (amdgpu_ip_version(adev, NBIO_HWIP, 0)) {
+	case IP_VERSION(2, 5, 0):
+		data = RREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF6_STRAP4) & ~BIT(23);
+		WREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF6_STRAP4, data);
+		break;
+	}
+}
+
+#define MMIO_REG_HOLE_OFFSET (0x80000 - PAGE_SIZE)
+
+static void nbio_v7_0_set_reg_remap(struct amdgpu_device *adev)
+{
+	if (!amdgpu_sriov_vf(adev) && (PAGE_SIZE <= 4096)) {
+		adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET;
+		adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET;
+	} else {
 		adev->rmmio_remap.reg_offset =
 			SOC15_REG_OFFSET(NBIO, 0, mmHDP_MEM_COHERENCY_FLUSH_CNTL) << 2;
+		adev->rmmio_remap.bus_addr = 0;
+	}
 }
 
 const struct amdgpu_nbio_funcs nbio_v7_0_funcs = {
@@ -297,4 +318,5 @@ const struct amdgpu_nbio_funcs nbio_v7_0_funcs = {
 	.ih_control = nbio_v7_0_ih_control,
 	.init_registers = nbio_v7_0_init_registers,
 	.remap_hdp_registers = nbio_v7_0_remap_hdp_registers,
+	.set_reg_remap = nbio_v7_0_set_reg_remap,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_11.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_11.c
new file mode 100644
index 000000000000..bed5ef4d8788
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_11.c
@@ -0,0 +1,400 @@
+/*
+ * Copyright 2021 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "amdgpu.h"
+#include "nbio_v7_11.h"
+
+#include "nbio/nbio_7_11_0_offset.h"
+#include "nbio/nbio_7_11_0_sh_mask.h"
+#include <uapi/linux/kfd_ioctl.h>
+
+static void nbio_v7_11_remap_hdp_registers(struct amdgpu_device *adev)
+{
+	WREG32_SOC15(NBIO, 0, regBIF_BX0_REMAP_HDP_MEM_FLUSH_CNTL,
+		     adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL);
+	WREG32_SOC15(NBIO, 0, regBIF_BX0_REMAP_HDP_REG_FLUSH_CNTL,
+		     adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_REG_FLUSH_CNTL);
+}
+
+static u32 nbio_v7_11_get_rev_id(struct amdgpu_device *adev)
+{
+	u32 tmp;
+
+	tmp = RREG32_SOC15(NBIO, 0, regRCC_STRAP1_RCC_DEV0_EPF0_STRAP0);
+	tmp &= RCC_STRAP0_RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0_MASK;
+	tmp >>= RCC_STRAP0_RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0__SHIFT;
+
+	return tmp;
+}
+
+static void nbio_v7_11_mc_access_enable(struct amdgpu_device *adev, bool enable)
+{
+	if (enable)
+		WREG32_SOC15(NBIO, 0, regBIF_BX1_BIF_FB_EN,
+			BIF_BX1_BIF_FB_EN__FB_READ_EN_MASK |
+			BIF_BX1_BIF_FB_EN__FB_WRITE_EN_MASK);
+	else
+		WREG32_SOC15(NBIO, 0, regBIF_BX1_BIF_FB_EN, 0);
+}
+
+static u32 nbio_v7_11_get_memsize(struct amdgpu_device *adev)
+{
+	return RREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF0_0_RCC_CONFIG_MEMSIZE);
+}
+
+static void nbio_v7_11_sdma_doorbell_range(struct amdgpu_device *adev, int instance,
+					  bool use_doorbell, int doorbell_index,
+					  int doorbell_size)
+{
+	u32 reg = SOC15_REG_OFFSET(NBIO, 0, regGDC0_BIF_CSDMA_DOORBELL_RANGE);
+	u32 doorbell_range = RREG32_PCIE_PORT(reg);
+
+	if (use_doorbell) {
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC0_BIF_CSDMA_DOORBELL_RANGE,
+					       OFFSET, doorbell_index);
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC0_BIF_CSDMA_DOORBELL_RANGE,
+					       SIZE, doorbell_size);
+	} else {
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC0_BIF_CSDMA_DOORBELL_RANGE,
+					       SIZE, 0);
+	}
+
+	WREG32_PCIE_PORT(reg, doorbell_range);
+}
+
+static void nbio_v7_11_vpe_doorbell_range(struct amdgpu_device *adev, int instance,
+					  bool use_doorbell, int doorbell_index,
+					  int doorbell_size)
+{
+	u32 reg = instance == 0 ?
+		  SOC15_REG_OFFSET(NBIO, 0, regGDC0_BIF_VPE_DOORBELL_RANGE) :
+		  SOC15_REG_OFFSET(NBIO, 0, regGDC0_BIF_VPE1_DOORBELL_RANGE);
+	u32 doorbell_range = RREG32_PCIE_PORT(reg);
+
+	if (use_doorbell) {
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC0_BIF_VPE_DOORBELL_RANGE,
+					       OFFSET, doorbell_index);
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC0_BIF_VPE_DOORBELL_RANGE,
+					       SIZE, doorbell_size);
+	} else {
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC0_BIF_VPE_DOORBELL_RANGE,
+					       SIZE, 0);
+	}
+
+	WREG32_PCIE_PORT(reg, doorbell_range);
+}
+
+static void nbio_v7_11_vcn_doorbell_range(struct amdgpu_device *adev,
+					  bool use_doorbell,
+					  int doorbell_index, int instance)
+{
+	u32 reg = instance == 0 ?
+		SOC15_REG_OFFSET(NBIO, 0, regGDC0_BIF_VCN0_DOORBELL_RANGE):
+		SOC15_REG_OFFSET(NBIO, 0, regGDC0_BIF_VCN1_DOORBELL_RANGE);
+
+	u32 doorbell_range = RREG32_PCIE_PORT(reg);
+
+	if (use_doorbell) {
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC0_BIF_VCN0_DOORBELL_RANGE, OFFSET,
+					       doorbell_index);
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC0_BIF_VCN0_DOORBELL_RANGE, SIZE, 8);
+	} else {
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC0_BIF_VCN0_DOORBELL_RANGE, SIZE, 0);
+	}
+
+	WREG32_PCIE_PORT(reg, doorbell_range);
+}
+
+static void nbio_v7_11_enable_doorbell_aperture(struct amdgpu_device *adev,
+					       bool enable)
+{
+	u32 reg;
+
+
+	reg = RREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF0_0_RCC_DOORBELL_APER_EN);
+	reg = REG_SET_FIELD(reg, RCC_DEV0_EPF0_0_RCC_DOORBELL_APER_EN,
+			    BIF_DOORBELL_APER_EN, enable ? 1 : 0);
+
+	WREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF0_0_RCC_DOORBELL_APER_EN, reg);
+}
+
+static void nbio_v7_11_enable_doorbell_selfring_aperture(struct amdgpu_device *adev,
+							bool enable)
+{
+	u32 tmp = 0;
+
+	if (enable) {
+		tmp = REG_SET_FIELD(tmp, BIF_BX_PF1_DOORBELL_SELFRING_GPA_APER_CNTL,
+				DOORBELL_SELFRING_GPA_APER_EN, 1) |
+		      REG_SET_FIELD(tmp, BIF_BX_PF1_DOORBELL_SELFRING_GPA_APER_CNTL,
+				DOORBELL_SELFRING_GPA_APER_MODE, 1) |
+		      REG_SET_FIELD(tmp, BIF_BX_PF1_DOORBELL_SELFRING_GPA_APER_CNTL,
+				DOORBELL_SELFRING_GPA_APER_SIZE, 0);
+
+		WREG32_SOC15(NBIO, 0,
+			regBIF_BX_PF1_DOORBELL_SELFRING_GPA_APER_BASE_LOW,
+			lower_32_bits(adev->doorbell.base));
+		WREG32_SOC15(NBIO, 0,
+			regBIF_BX_PF1_DOORBELL_SELFRING_GPA_APER_BASE_HIGH,
+			upper_32_bits(adev->doorbell.base));
+	}
+
+	WREG32_SOC15(NBIO, 0, regBIF_BX_PF1_DOORBELL_SELFRING_GPA_APER_CNTL, tmp);
+}
+
+
+static void nbio_v7_11_ih_doorbell_range(struct amdgpu_device *adev,
+					bool use_doorbell, int doorbell_index)
+{    
+	u32 ih_doorbell_range = RREG32_SOC15(NBIO, 0,regGDC0_BIF_IH_DOORBELL_RANGE);
+
+	if (use_doorbell) {
+		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
+						  GDC0_BIF_IH_DOORBELL_RANGE, OFFSET,
+						  doorbell_index);
+		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
+						  GDC0_BIF_IH_DOORBELL_RANGE, SIZE,
+						  2);
+	} else {
+		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
+						  GDC0_BIF_IH_DOORBELL_RANGE, SIZE,
+						  0);
+	}
+
+	WREG32_SOC15(NBIO, 0, regGDC0_BIF_IH_DOORBELL_RANGE,
+			 ih_doorbell_range);
+}
+
+static void nbio_v7_11_ih_control(struct amdgpu_device *adev)
+{
+	u32 interrupt_cntl;
+
+	/* setup interrupt control */
+	WREG32_SOC15(NBIO, 0, regBIF_BX1_INTERRUPT_CNTL2,
+		     adev->dummy_page_addr >> 8);
+
+	interrupt_cntl = RREG32_SOC15(NBIO, 0, regBIF_BX1_INTERRUPT_CNTL);
+	/*
+	 * INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=0 - dummy read disabled with msi, enabled without msi
+	 * INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=1 - dummy read controlled by IH_DUMMY_RD_EN
+	 */
+	interrupt_cntl = REG_SET_FIELD(interrupt_cntl, BIF_BX1_INTERRUPT_CNTL,
+				       IH_DUMMY_RD_OVERRIDE, 0);
+
+	/* INTERRUPT_CNTL__IH_REQ_NONSNOOP_EN_MASK=1 if ring is in non-cacheable memory, e.g., vram */
+	interrupt_cntl = REG_SET_FIELD(interrupt_cntl, BIF_BX1_INTERRUPT_CNTL,
+				       IH_REQ_NONSNOOP_EN, 0);
+
+	WREG32_SOC15(NBIO, 0, regBIF_BX1_INTERRUPT_CNTL, interrupt_cntl);
+}
+
+static u32 nbio_v7_11_get_hdp_flush_req_offset(struct amdgpu_device *adev)
+{
+	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF1_GPU_HDP_FLUSH_REQ);
+}
+
+static u32 nbio_v7_11_get_hdp_flush_done_offset(struct amdgpu_device *adev)
+{
+	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF1_GPU_HDP_FLUSH_DONE);
+}
+
+static u32 nbio_v7_11_get_pcie_index_offset(struct amdgpu_device *adev)
+{
+	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX1_PCIE_INDEX2);
+}
+
+static u32 nbio_v7_11_get_pcie_data_offset(struct amdgpu_device *adev)
+{
+	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX1_PCIE_DATA2);
+}
+
+static u32 nbio_v7_11_get_pcie_port_index_offset(struct amdgpu_device *adev)
+{
+	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF1_RSMU_INDEX);
+}
+
+static u32 nbio_v7_11_get_pcie_port_data_offset(struct amdgpu_device *adev)
+{
+	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF1_RSMU_DATA);
+}
+
+const struct nbio_hdp_flush_reg nbio_v7_11_hdp_flush_reg = {
+	.ref_and_mask_cp0 = BIF_BX_PF1_GPU_HDP_FLUSH_DONE__CP0_MASK,
+	.ref_and_mask_cp1 = BIF_BX_PF1_GPU_HDP_FLUSH_DONE__CP1_MASK,
+	.ref_and_mask_cp2 = BIF_BX_PF1_GPU_HDP_FLUSH_DONE__CP2_MASK,
+	.ref_and_mask_cp3 = BIF_BX_PF1_GPU_HDP_FLUSH_DONE__CP3_MASK,
+	.ref_and_mask_cp4 = BIF_BX_PF1_GPU_HDP_FLUSH_DONE__CP4_MASK,
+	.ref_and_mask_cp5 = BIF_BX_PF1_GPU_HDP_FLUSH_DONE__CP5_MASK,
+	.ref_and_mask_cp6 = BIF_BX_PF1_GPU_HDP_FLUSH_DONE__CP6_MASK,
+	.ref_and_mask_cp7 = BIF_BX_PF1_GPU_HDP_FLUSH_DONE__CP7_MASK,
+	.ref_and_mask_cp8 = BIF_BX_PF1_GPU_HDP_FLUSH_DONE__CP8_MASK,
+	.ref_and_mask_cp9 = BIF_BX_PF1_GPU_HDP_FLUSH_DONE__CP9_MASK,
+	.ref_and_mask_sdma0 = BIF_BX_PF1_GPU_HDP_FLUSH_DONE__SDMA0_MASK,
+	.ref_and_mask_sdma1 = BIF_BX_PF1_GPU_HDP_FLUSH_DONE__SDMA1_MASK,
+};
+
+static void nbio_v7_11_init_registers(struct amdgpu_device *adev)
+{
+	uint32_t def, data;
+
+	def = data = RREG32_SOC15(NBIO, 0, regBIF_BIF256_CI256_RC3X4_USB4_PCIE_MST_CTRL_3);
+	data = REG_SET_FIELD(data, BIF_BIF256_CI256_RC3X4_USB4_PCIE_MST_CTRL_3,
+				CI_SWUS_MAX_READ_REQUEST_SIZE_MODE, 1);
+	data = REG_SET_FIELD(data, BIF_BIF256_CI256_RC3X4_USB4_PCIE_MST_CTRL_3,
+				CI_SWUS_MAX_READ_REQUEST_SIZE_PRIV, 1);
+
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regBIF_BIF256_CI256_RC3X4_USB4_PCIE_MST_CTRL_3, data);
+
+	switch (amdgpu_ip_version(adev, NBIO_HWIP, 0)) {
+	case IP_VERSION(7, 11, 0):
+	case IP_VERSION(7, 11, 1):
+	case IP_VERSION(7, 11, 2):
+	case IP_VERSION(7, 11, 3):
+		data = RREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF5_STRAP4) & ~BIT(23);
+		WREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF5_STRAP4, data);
+		break;
+	}
+}
+
+static void nbio_v7_11_update_medium_grain_clock_gating(struct amdgpu_device *adev,
+						       bool enable)
+{
+	uint32_t def, data;
+
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_BIF_MGCG))
+		return;
+
+	def = data = RREG32_SOC15(NBIO, 0, regBIF_BIF256_CI256_RC3X4_USB4_CPM_CONTROL);
+	if (enable) {
+		data |= (BIF_BIF256_CI256_RC3X4_USB4_CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK |
+			 BIF_BIF256_CI256_RC3X4_USB4_CPM_CONTROL__TXCLK_DYN_GATE_ENABLE_MASK |
+			 BIF_BIF256_CI256_RC3X4_USB4_CPM_CONTROL__TXCLK_LCNT_GATE_ENABLE_MASK |
+			 BIF_BIF256_CI256_RC3X4_USB4_CPM_CONTROL__TXCLK_REGS_GATE_ENABLE_MASK |
+			 BIF_BIF256_CI256_RC3X4_USB4_CPM_CONTROL__TXCLK_PRBS_GATE_ENABLE_MASK |
+			 BIF_BIF256_CI256_RC3X4_USB4_CPM_CONTROL__REFCLK_REGS_GATE_ENABLE_MASK);
+	} else {
+		data &= ~(BIF_BIF256_CI256_RC3X4_USB4_CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK |
+			  BIF_BIF256_CI256_RC3X4_USB4_CPM_CONTROL__TXCLK_DYN_GATE_ENABLE_MASK |
+			  BIF_BIF256_CI256_RC3X4_USB4_CPM_CONTROL__TXCLK_LCNT_GATE_ENABLE_MASK |
+			  BIF_BIF256_CI256_RC3X4_USB4_CPM_CONTROL__TXCLK_REGS_GATE_ENABLE_MASK |
+			  BIF_BIF256_CI256_RC3X4_USB4_CPM_CONTROL__TXCLK_PRBS_GATE_ENABLE_MASK |
+			  BIF_BIF256_CI256_RC3X4_USB4_CPM_CONTROL__REFCLK_REGS_GATE_ENABLE_MASK);
+	}
+
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regBIF_BIF256_CI256_RC3X4_USB4_CPM_CONTROL, data);
+}
+
+static void nbio_v7_11_update_medium_grain_light_sleep(struct amdgpu_device *adev,
+						      bool enable)
+{
+	uint32_t def, data;
+
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_BIF_LS))
+		return;
+
+	def = data = RREG32_SOC15(NBIO, 0, regBIF_BIF256_CI256_RC3X4_USB4_PCIE_CNTL2);
+	if (enable)
+		data |= BIF_BIF256_CI256_RC3X4_USB4_PCIE_CNTL2__SLV_MEM_LS_EN_MASK;
+	else
+		data &= ~BIF_BIF256_CI256_RC3X4_USB4_PCIE_CNTL2__SLV_MEM_LS_EN_MASK;
+
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regBIF_BIF256_CI256_RC3X4_USB4_PCIE_CNTL2, data);
+
+	def = data = RREG32_SOC15(NBIO, 0, regBIF_BIF256_CI256_RC3X4_USB4_PCIE_TX_POWER_CTRL_1);
+	if (enable) {
+		data |= (BIF_BIF256_CI256_RC3X4_USB4_PCIE_TX_POWER_CTRL_1__MST_MEM_LS_EN_MASK |
+			BIF_BIF256_CI256_RC3X4_USB4_PCIE_TX_POWER_CTRL_1__REPLAY_MEM_LS_EN_MASK);
+	} else {
+		data &= ~(BIF_BIF256_CI256_RC3X4_USB4_PCIE_TX_POWER_CTRL_1__MST_MEM_LS_EN_MASK |
+			BIF_BIF256_CI256_RC3X4_USB4_PCIE_TX_POWER_CTRL_1__REPLAY_MEM_LS_EN_MASK);
+	}
+
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regBIF_BIF256_CI256_RC3X4_USB4_PCIE_TX_POWER_CTRL_1, data);
+}
+
+static void nbio_v7_11_get_clockgating_state(struct amdgpu_device *adev,
+					    u64 *flags)
+{
+	uint32_t data;
+
+	/* AMD_CG_SUPPORT_BIF_MGCG */
+	data = RREG32_SOC15(NBIO, 0, regBIF_BIF256_CI256_RC3X4_USB4_CPM_CONTROL);
+	if (data & BIF_BIF256_CI256_RC3X4_USB4_CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK)
+		*flags |= AMD_CG_SUPPORT_BIF_MGCG;
+
+	/* AMD_CG_SUPPORT_BIF_LS */
+	data = RREG32_SOC15(NBIO, 0, regBIF_BIF256_CI256_RC3X4_USB4_PCIE_CNTL2);
+	if (data & BIF_BIF256_CI256_RC3X4_USB4_PCIE_CNTL2__SLV_MEM_LS_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_BIF_LS;
+}
+
+#define MMIO_REG_HOLE_OFFSET 0x44000
+
+static void nbio_v7_11_set_reg_remap(struct amdgpu_device *adev)
+{
+	if (!amdgpu_sriov_vf(adev) && (PAGE_SIZE <= 4096)) {
+		adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET;
+		adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET;
+	} else {
+		adev->rmmio_remap.reg_offset =
+			SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF1_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2;
+		adev->rmmio_remap.bus_addr = 0;
+	}
+}
+
+const struct amdgpu_nbio_funcs nbio_v7_11_funcs = {
+	.get_hdp_flush_req_offset = nbio_v7_11_get_hdp_flush_req_offset,
+	.get_hdp_flush_done_offset = nbio_v7_11_get_hdp_flush_done_offset,
+	.get_pcie_index_offset = nbio_v7_11_get_pcie_index_offset,
+	.get_pcie_data_offset = nbio_v7_11_get_pcie_data_offset,
+	.get_pcie_port_index_offset = nbio_v7_11_get_pcie_port_index_offset,
+	.get_pcie_port_data_offset = nbio_v7_11_get_pcie_port_data_offset,
+	.get_rev_id = nbio_v7_11_get_rev_id,
+	.mc_access_enable = nbio_v7_11_mc_access_enable,
+	.get_memsize = nbio_v7_11_get_memsize,
+	.sdma_doorbell_range = nbio_v7_11_sdma_doorbell_range,
+	.vcn_doorbell_range = nbio_v7_11_vcn_doorbell_range,
+	.vpe_doorbell_range = nbio_v7_11_vpe_doorbell_range,
+	.enable_doorbell_aperture = nbio_v7_11_enable_doorbell_aperture,
+	.enable_doorbell_selfring_aperture = nbio_v7_11_enable_doorbell_selfring_aperture,
+	.ih_doorbell_range = nbio_v7_11_ih_doorbell_range,
+	.update_medium_grain_clock_gating = nbio_v7_11_update_medium_grain_clock_gating,
+	.update_medium_grain_light_sleep = nbio_v7_11_update_medium_grain_light_sleep,
+	.get_clockgating_state = nbio_v7_11_get_clockgating_state,
+	.ih_control = nbio_v7_11_ih_control,
+	.init_registers = nbio_v7_11_init_registers,
+	.remap_hdp_registers = nbio_v7_11_remap_hdp_registers,
+	.set_reg_remap = nbio_v7_11_set_reg_remap,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_11.h b/drivers/gpu/drm/amd/amdgpu/nbio_v7_11.h
new file mode 100644
index 000000000000..9d8258ed3f0a
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_11.h
@@ -0,0 +1,33 @@
+/*
+ * Copyright 2021 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __NBIO_V7_11_H__
+#define __NBIO_V7_11_H__
+
+#include "soc15_common.h"
+
+extern const struct nbio_hdp_flush_reg nbio_v7_11_hdp_flush_reg;
+extern const struct amdgpu_nbio_funcs nbio_v7_11_funcs;
+extern const struct amdgpu_nbio_ras_funcs nbio_v7_11_ras_funcs;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_2.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_2.c
index 4ef1fa4603c8..acc5f363684a 100644
--- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_2.c
@@ -21,7 +21,6 @@
  *
  */
 #include "amdgpu.h"
-#include "amdgpu_atombios.h"
 #include "nbio_v7_2.h"
 
 #include "nbio/nbio_7_2_0_offset.h"
@@ -59,7 +58,7 @@ static u32 nbio_v7_2_get_rev_id(struct amdgpu_device *adev)
 {
 	u32 tmp;
 
-	switch (adev->ip_versions[NBIO_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, NBIO_HWIP, 0)) {
 	case IP_VERSION(7, 2, 1):
 	case IP_VERSION(7, 3, 0):
 	case IP_VERSION(7, 5, 0):
@@ -78,7 +77,7 @@ static u32 nbio_v7_2_get_rev_id(struct amdgpu_device *adev)
 
 static void nbio_v7_2_mc_access_enable(struct amdgpu_device *adev, bool enable)
 {
-	switch (adev->ip_versions[NBIO_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, NBIO_HWIP, 0)) {
 	case IP_VERSION(7, 2, 1):
 	case IP_VERSION(7, 3, 0):
 	case IP_VERSION(7, 5, 0):
@@ -262,7 +261,7 @@ static void nbio_v7_2_update_medium_grain_light_sleep(struct amdgpu_device *adev
 {
 	uint32_t def, data;
 
-	switch (adev->ip_versions[NBIO_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, NBIO_HWIP, 0)) {
 	case IP_VERSION(7, 2, 1):
 	case IP_VERSION(7, 3, 0):
 	case IP_VERSION(7, 5, 0):
@@ -369,7 +368,7 @@ const struct nbio_hdp_flush_reg nbio_v7_2_hdp_flush_reg = {
 static void nbio_v7_2_init_registers(struct amdgpu_device *adev)
 {
 	uint32_t def, data;
-	switch (adev->ip_versions[NBIO_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, NBIO_HWIP, 0)) {
 	case IP_VERSION(7, 2, 1):
 	case IP_VERSION(7, 3, 0):
 	case IP_VERSION(7, 5, 0):
@@ -394,7 +393,7 @@ static void nbio_v7_2_init_registers(struct amdgpu_device *adev)
 		break;
 	}
 
-	switch (adev->ip_versions[NBIO_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, NBIO_HWIP, 0)) {
 	case IP_VERSION(7, 3, 0):
 	case IP_VERSION(7, 5, 1):
 		data = RREG32_SOC15(NBIO, 0, regRCC_DEV2_EPF0_STRAP2);
@@ -402,10 +401,21 @@ static void nbio_v7_2_init_registers(struct amdgpu_device *adev)
 		WREG32_SOC15(NBIO, 0, regRCC_DEV2_EPF0_STRAP2, data);
 		break;
 	}
+}
+
+#define MMIO_REG_HOLE_OFFSET (0x80000 - PAGE_SIZE)
 
-	if (amdgpu_sriov_vf(adev))
-		adev->rmmio_remap.reg_offset = SOC15_REG_OFFSET(NBIO, 0,
-			regBIF_BX_PF0_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2;
+static void nbio_v7_2_set_reg_remap(struct amdgpu_device *adev)
+{
+	if (!amdgpu_sriov_vf(adev) && (PAGE_SIZE <= 4096)) {
+		adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET;
+		adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET;
+	} else {
+		adev->rmmio_remap.reg_offset =
+			SOC15_REG_OFFSET(NBIO, 0,
+					 regBIF_BX_PF0_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2;
+		adev->rmmio_remap.bus_addr = 0;
+	}
 }
 
 const struct amdgpu_nbio_funcs nbio_v7_2_funcs = {
@@ -429,4 +439,5 @@ const struct amdgpu_nbio_funcs nbio_v7_2_funcs = {
 	.ih_control = nbio_v7_2_ih_control,
 	.init_registers = nbio_v7_2_init_registers,
 	.remap_hdp_registers = nbio_v7_2_remap_hdp_registers,
+	.set_reg_remap = nbio_v7_2_set_reg_remap,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c
index 685abf57ffdd..860bc5cb03c8 100644
--- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c
+++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c
@@ -21,7 +21,6 @@
  *
  */
 #include "amdgpu.h"
-#include "amdgpu_atombios.h"
 #include "nbio_v7_4.h"
 #include "amdgpu_ras.h"
 
@@ -152,9 +151,9 @@ static void nbio_v7_4_sdma_doorbell_range(struct amdgpu_device *adev, int instan
 		 *   BIF_SDMA0_DOORBELL_RANGE:  0x3bc0
 		 *   BIF_SDMA1_DOORBELL_RANGE:  0x3bc4
 		 *   BIF_SDMA2_DOORBELL_RANGE:  0x3bd8
-+		 *   BIF_SDMA4_DOORBELL_RANGE:
-+		 *     ARCTURUS:  0x3be0
-+		 *     ALDEBARAN: 0x3be4
+		 *   BIF_SDMA4_DOORBELL_RANGE:
+		 *     ARCTURUS:  0x3be0
+		 *     ALDEBARAN: 0x3be4
 		 */
 		if (adev->asic_type == CHIP_ALDEBARAN && instance == 4)
 			reg = instance + 0x4 + 0x1 +
@@ -343,11 +342,7 @@ static void nbio_v7_4_init_registers(struct amdgpu_device *adev)
 {
 	uint32_t baco_cntl;
 
-	if (amdgpu_sriov_vf(adev))
-		adev->rmmio_remap.reg_offset = SOC15_REG_OFFSET(NBIO, 0,
-			mmBIF_BX_DEV0_EPF0_VF0_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2;
-
-	if (adev->ip_versions[NBIO_HWIP][0] == IP_VERSION(7, 4, 4) &&
+	if (amdgpu_ip_version(adev, NBIO_HWIP, 0) == IP_VERSION(7, 4, 4) &&
 	    !amdgpu_sriov_vf(adev)) {
 		baco_cntl = RREG32_SOC15(NBIO, 0, mmBACO_CNTL);
 		if (baco_cntl &
@@ -365,9 +360,12 @@ static void nbio_v7_4_handle_ras_controller_intr_no_bifring(struct amdgpu_device
 {
 	uint32_t bif_doorbell_intr_cntl;
 	struct ras_manager *obj = amdgpu_ras_find_obj(adev, adev->nbio.ras_if);
-	struct ras_err_data err_data = {0, 0, 0, NULL};
+	struct ras_err_data err_data;
 	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 
+	if (amdgpu_ras_error_data_init(&err_data))
+		return;
+
 	if (adev->asic_type == CHIP_ALDEBARAN)
 		bif_doorbell_intr_cntl = RREG32_SOC15(NBIO, 0, mmBIF_DOORBELL_INT_CNTL_ALDE);
 	else
@@ -384,7 +382,7 @@ static void nbio_v7_4_handle_ras_controller_intr_no_bifring(struct amdgpu_device
 		else
 			WREG32_SOC15(NBIO, 0, mmBIF_DOORBELL_INT_CNTL, bif_doorbell_intr_cntl);
 
-		if (!ras->disable_ras_err_cnt_harvest) {
+		if (ras && !ras->disable_ras_err_cnt_harvest && obj) {
 			/*
 			 * clear error status after ras_controller_intr
 			 * according to hw team and count ue number
@@ -398,8 +396,7 @@ static void nbio_v7_4_handle_ras_controller_intr_no_bifring(struct amdgpu_device
 
 			if (err_data.ce_count)
 				dev_info(adev->dev, "%ld correctable hardware "
-						"errors detected in %s block, "
-						"no user action is needed.\n",
+						"errors detected in %s block\n",
 						obj->err_data.ce_count,
 						get_ras_block_str(adev->nbio.ras_if));
 
@@ -416,8 +413,10 @@ static void nbio_v7_4_handle_ras_controller_intr_no_bifring(struct amdgpu_device
 		/* ras_controller_int is dedicated for nbif ras error,
 		 * not the global interrupt for sync flood
 		 */
-		amdgpu_ras_reset_gpu(adev);
+		amdgpu_ras_global_ras_isr(adev);
 	}
+
+	amdgpu_ras_error_data_fini(&err_data);
 }
 
 static void nbio_v7_4_handle_ras_err_event_athub_intr_no_bifring(struct amdgpu_device *adev)
@@ -702,7 +701,7 @@ static void nbio_v7_4_program_aspm(struct amdgpu_device *adev)
 #ifdef CONFIG_PCIEASPM
 	uint32_t def, data;
 
-	if (adev->ip_versions[NBIO_HWIP][0] == IP_VERSION(7, 4, 4))
+	if (amdgpu_ip_version(adev, NBIO_HWIP, 0) == IP_VERSION(7, 4, 4))
 		return;
 
 	def = data = RREG32_PCIE(smnPCIE_LC_CNTL);
@@ -788,6 +787,21 @@ static void nbio_v7_4_program_aspm(struct amdgpu_device *adev)
 #endif
 }
 
+#define MMIO_REG_HOLE_OFFSET (0x80000 - PAGE_SIZE)
+
+static void nbio_v7_4_set_reg_remap(struct amdgpu_device *adev)
+{
+	if (!amdgpu_sriov_vf(adev) && (PAGE_SIZE <= 4096)) {
+		adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET;
+		adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET;
+	} else {
+		adev->rmmio_remap.reg_offset =
+			SOC15_REG_OFFSET(NBIO, 0,
+					 mmBIF_BX_DEV0_EPF0_VF0_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2;
+		adev->rmmio_remap.bus_addr = 0;
+	}
+}
+
 const struct amdgpu_nbio_funcs nbio_v7_4_funcs = {
 	.get_hdp_flush_req_offset = nbio_v7_4_get_hdp_flush_req_offset,
 	.get_hdp_flush_done_offset = nbio_v7_4_get_hdp_flush_done_offset,
@@ -809,4 +823,5 @@ const struct amdgpu_nbio_funcs nbio_v7_4_funcs = {
 	.init_registers = nbio_v7_4_init_registers,
 	.remap_hdp_registers = nbio_v7_4_remap_hdp_registers,
 	.program_aspm =  nbio_v7_4_program_aspm,
+	.set_reg_remap = nbio_v7_4_set_reg_remap,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_7.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_7.c
index def89379b51a..2ee60b8746a6 100644
--- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_7.c
+++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_7.c
@@ -21,7 +21,6 @@
  *
  */
 #include "amdgpu.h"
-#include "amdgpu_atombios.h"
 #include "nbio_v7_7.h"
 
 #include "nbio/nbio_7_7_0_offset.h"
@@ -247,6 +246,12 @@ static void nbio_v7_7_init_registers(struct amdgpu_device *adev)
 	if (def != data)
 		WREG32_SOC15(NBIO, 0, regBIF0_PCIE_MST_CTRL_3, data);
 
+	switch (amdgpu_ip_version(adev, NBIO_HWIP, 0)) {
+	case IP_VERSION(7, 7, 0):
+		data = RREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF5_STRAP4) & ~BIT(23);
+		WREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF5_STRAP4, data);
+		break;
+	}
 }
 
 static void nbio_v7_7_update_medium_grain_clock_gating(struct amdgpu_device *adev,
@@ -254,7 +259,7 @@ static void nbio_v7_7_update_medium_grain_clock_gating(struct amdgpu_device *ade
 {
 	uint32_t def, data;
 
-	if (enable && !(adev->cg_flags & AMD_CG_SUPPORT_BIF_MGCG))
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_BIF_MGCG))
 		return;
 
 	def = data = RREG32_SOC15(NBIO, 0, regBIF0_CPM_CONTROL);
@@ -283,7 +288,7 @@ static void nbio_v7_7_update_medium_grain_light_sleep(struct amdgpu_device *adev
 {
 	uint32_t def, data;
 
-	if (enable && !(adev->cg_flags & AMD_CG_SUPPORT_BIF_LS))
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_BIF_LS))
 		return;
 
 	def = data = RREG32_SOC15(NBIO, 0, regBIF0_PCIE_CNTL2);
@@ -324,6 +329,21 @@ static void nbio_v7_7_get_clockgating_state(struct amdgpu_device *adev,
 		*flags |= AMD_CG_SUPPORT_BIF_LS;
 }
 
+#define MMIO_REG_HOLE_OFFSET (0x80000 - PAGE_SIZE)
+
+static void nbio_v7_7_set_reg_remap(struct amdgpu_device *adev)
+{
+	if (!amdgpu_sriov_vf(adev) && (PAGE_SIZE <= 4096)) {
+		adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET;
+		adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET;
+	} else {
+		adev->rmmio_remap.reg_offset =
+			SOC15_REG_OFFSET(NBIO, 0,
+					 regBIF_BX_PF0_HDP_MEM_COHERENCY_FLUSH_CNTL) << 2;
+		adev->rmmio_remap.bus_addr = 0;
+	}
+}
+
 const struct amdgpu_nbio_funcs nbio_v7_7_funcs = {
 	.get_hdp_flush_req_offset = nbio_v7_7_get_hdp_flush_req_offset,
 	.get_hdp_flush_done_offset = nbio_v7_7_get_hdp_flush_done_offset,
@@ -345,4 +365,5 @@ const struct amdgpu_nbio_funcs nbio_v7_7_funcs = {
 	.ih_control = nbio_v7_7_ih_control,
 	.init_registers = nbio_v7_7_init_registers,
 	.remap_hdp_registers = nbio_v7_7_remap_hdp_registers,
+	.set_reg_remap = nbio_v7_7_set_reg_remap,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_9.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_9.c
index 24d12075ca3a..bdfd2917e3ca 100644
--- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_9.c
+++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_9.c
@@ -21,7 +21,6 @@
  *
  */
 #include "amdgpu.h"
-#include "amdgpu_atombios.h"
 #include "nbio_v7_9.h"
 #include "amdgpu_ras.h"
 
@@ -30,6 +29,8 @@
 #include "ivsrcid/nbio/irqsrcs_nbif_7_4.h"
 #include <uapi/linux/kfd_ioctl.h>
 
+#define NPS_MODE_MASK 0x000000FFL
+
 static void nbio_v7_9_remap_hdp_registers(struct amdgpu_device *adev)
 {
 	WREG32_SOC15(NBIO, 0, regBIF_BX0_REMAP_HDP_MEM_FLUSH_CNTL,
@@ -40,12 +41,21 @@ static void nbio_v7_9_remap_hdp_registers(struct amdgpu_device *adev)
 
 static u32 nbio_v7_9_get_rev_id(struct amdgpu_device *adev)
 {
-	u32 tmp;
+	u32 rev_id;
 
-	tmp = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_DEV0_EPF0_STRAP0);
-	tmp = REG_GET_FIELD(tmp, RCC_STRAP0_RCC_DEV0_EPF0_STRAP0, STRAP_ATI_REV_ID_DEV0_F0);
+	/*
+	 * fetch the sub-revision field from the IP-discovery table
+	 * (returns zero if the table entry is not populated).
+	 */
+	if (amdgpu_sriov_vf(adev)) {
+		rev_id = IP_VERSION_SUBREV(amdgpu_ip_version_full(adev, NBIO_HWIP, 0));
+	} else {
+		rev_id = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_DEV0_EPF0_STRAP0);
+		rev_id = REG_GET_FIELD(rev_id, RCC_STRAP0_RCC_DEV0_EPF0_STRAP0,
+				STRAP_ATI_REV_ID_DEV0_F0);
+	}
 
-	return tmp;
+	return rev_id;
 }
 
 static void nbio_v7_9_mc_access_enable(struct amdgpu_device *adev, bool enable)
@@ -66,6 +76,13 @@ static void nbio_v7_9_sdma_doorbell_range(struct amdgpu_device *adev, int instan
 			bool use_doorbell, int doorbell_index, int doorbell_size)
 {
 	u32 doorbell_range = 0, doorbell_ctrl = 0;
+	int aid_id, dev_inst;
+
+	dev_inst = GET_INST(SDMA0, instance);
+	aid_id = adev->sdma.instance[instance].aid_id;
+
+	if (use_doorbell == false)
+		return;
 
 	doorbell_range =
 		REG_SET_FIELD(doorbell_range, DOORBELL0_CTRL_ENTRY_0,
@@ -80,9 +97,10 @@ static void nbio_v7_9_sdma_doorbell_range(struct amdgpu_device *adev, int instan
 		REG_SET_FIELD(doorbell_ctrl, S2A_DOORBELL_ENTRY_1_CTRL,
 			S2A_DOORBELL_PORT1_RANGE_SIZE, doorbell_size);
 
-	switch (instance) {
+	switch (dev_inst % adev->sdma.num_inst_per_aid) {
 	case 0:
-		WREG32_SOC15(NBIO, 0, regDOORBELL0_CTRL_ENTRY_1, doorbell_range);
+		WREG32_SOC15_OFFSET(NBIO, 0, regDOORBELL0_CTRL_ENTRY_1,
+			4 * aid_id, doorbell_range);
 
 		doorbell_ctrl = REG_SET_FIELD(doorbell_ctrl,
 					S2A_DOORBELL_ENTRY_1_CTRL,
@@ -94,10 +112,12 @@ static void nbio_v7_9_sdma_doorbell_range(struct amdgpu_device *adev, int instan
 					S2A_DOORBELL_ENTRY_1_CTRL,
 					S2A_DOORBELL_PORT1_AWADDR_31_28_VALUE,
 					0x1);
-		WREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_1_CTRL, doorbell_ctrl);
+		WREG32_SOC15_EXT(NBIO, aid_id, regS2A_DOORBELL_ENTRY_1_CTRL,
+			aid_id, doorbell_ctrl);
 		break;
 	case 1:
-		WREG32_SOC15(NBIO, 0, regDOORBELL0_CTRL_ENTRY_2, doorbell_range);
+		WREG32_SOC15_OFFSET(NBIO, 0, regDOORBELL0_CTRL_ENTRY_2,
+			4 * aid_id, doorbell_range);
 
 		doorbell_ctrl = REG_SET_FIELD(doorbell_ctrl,
 					S2A_DOORBELL_ENTRY_1_CTRL,
@@ -109,10 +129,12 @@ static void nbio_v7_9_sdma_doorbell_range(struct amdgpu_device *adev, int instan
 					S2A_DOORBELL_ENTRY_1_CTRL,
 					S2A_DOORBELL_PORT1_AWADDR_31_28_VALUE,
 					0x2);
-		WREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_2_CTRL, doorbell_ctrl);
+		WREG32_SOC15_EXT(NBIO, aid_id, regS2A_DOORBELL_ENTRY_2_CTRL,
+			aid_id, doorbell_ctrl);
 		break;
 	case 2:
-		WREG32_SOC15(NBIO, 0, regDOORBELL0_CTRL_ENTRY_3, doorbell_range);
+		WREG32_SOC15_OFFSET(NBIO, 0, regDOORBELL0_CTRL_ENTRY_3,
+			4 * aid_id, doorbell_range);
 
 		doorbell_ctrl = REG_SET_FIELD(doorbell_ctrl,
 					S2A_DOORBELL_ENTRY_1_CTRL,
@@ -124,10 +146,12 @@ static void nbio_v7_9_sdma_doorbell_range(struct amdgpu_device *adev, int instan
 					S2A_DOORBELL_ENTRY_1_CTRL,
 					S2A_DOORBELL_PORT1_AWADDR_31_28_VALUE,
 					0x8);
-		WREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_5_CTRL, doorbell_ctrl);
+		WREG32_SOC15_EXT(NBIO, aid_id, regS2A_DOORBELL_ENTRY_5_CTRL,
+			aid_id, doorbell_ctrl);
 		break;
 	case 3:
-		WREG32_SOC15(NBIO, 0, regDOORBELL0_CTRL_ENTRY_4, doorbell_range);
+		WREG32_SOC15_OFFSET(NBIO, 0, regDOORBELL0_CTRL_ENTRY_4,
+			4 * aid_id, doorbell_range);
 
 		doorbell_ctrl = REG_SET_FIELD(doorbell_ctrl,
 					S2A_DOORBELL_ENTRY_1_CTRL,
@@ -139,21 +163,25 @@ static void nbio_v7_9_sdma_doorbell_range(struct amdgpu_device *adev, int instan
 					S2A_DOORBELL_ENTRY_1_CTRL,
 					S2A_DOORBELL_PORT1_AWADDR_31_28_VALUE,
 					0x9);
-		WREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_6_CTRL, doorbell_ctrl);
+		WREG32_SOC15_EXT(NBIO, aid_id, regS2A_DOORBELL_ENTRY_6_CTRL,
+			aid_id, doorbell_ctrl);
 		break;
 	default:
 		break;
-	};
-
-	return;
+	}
 }
 
 static void nbio_v7_9_vcn_doorbell_range(struct amdgpu_device *adev, bool use_doorbell,
 					 int doorbell_index, int instance)
 {
 	u32 doorbell_range = 0, doorbell_ctrl = 0;
+	u32 aid_id = instance;
+	u32 range_size;
 
 	if (use_doorbell) {
+		range_size = (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+						IP_VERSION(9, 5, 0)) ?
+						0xb : 0x9;
 		doorbell_range = REG_SET_FIELD(doorbell_range,
 				DOORBELL0_CTRL_ENTRY_0,
 				BIF_DOORBELL0_RANGE_OFFSET_ENTRY,
@@ -161,7 +189,12 @@ static void nbio_v7_9_vcn_doorbell_range(struct amdgpu_device *adev, bool use_do
 		doorbell_range = REG_SET_FIELD(doorbell_range,
 				DOORBELL0_CTRL_ENTRY_0,
 				BIF_DOORBELL0_RANGE_SIZE_ENTRY,
-				0x8);
+				range_size);
+		if (aid_id)
+			doorbell_range = REG_SET_FIELD(doorbell_range,
+					DOORBELL0_CTRL_ENTRY_0,
+					DOORBELL0_FENCE_ENABLE_ENTRY,
+					0x4);
 
 		doorbell_ctrl = REG_SET_FIELD(doorbell_ctrl,
 				S2A_DOORBELL_ENTRY_1_CTRL,
@@ -174,10 +207,15 @@ static void nbio_v7_9_vcn_doorbell_range(struct amdgpu_device *adev, bool use_do
 				S2A_DOORBELL_PORT1_RANGE_OFFSET, 0x4);
 		doorbell_ctrl = REG_SET_FIELD(doorbell_ctrl,
 				S2A_DOORBELL_ENTRY_1_CTRL,
-				S2A_DOORBELL_PORT1_RANGE_SIZE, 0x8);
+				S2A_DOORBELL_PORT1_RANGE_SIZE, range_size);
 		doorbell_ctrl = REG_SET_FIELD(doorbell_ctrl,
 				S2A_DOORBELL_ENTRY_1_CTRL,
 				S2A_DOORBELL_PORT1_AWADDR_31_28_VALUE, 0x4);
+
+		WREG32_SOC15_OFFSET(NBIO, 0, regDOORBELL0_CTRL_ENTRY_17,
+					aid_id, doorbell_range);
+		WREG32_SOC15_EXT(NBIO, aid_id, regS2A_DOORBELL_ENTRY_4_CTRL,
+				aid_id, doorbell_ctrl);
 	} else {
 		doorbell_range = REG_SET_FIELD(doorbell_range,
 				DOORBELL0_CTRL_ENTRY_0,
@@ -185,10 +223,12 @@ static void nbio_v7_9_vcn_doorbell_range(struct amdgpu_device *adev, bool use_do
 		doorbell_ctrl = REG_SET_FIELD(doorbell_ctrl,
 				S2A_DOORBELL_ENTRY_1_CTRL,
 				S2A_DOORBELL_PORT1_RANGE_SIZE, 0);
-	}
 
-	WREG32_SOC15(NBIO, 0, regDOORBELL0_CTRL_ENTRY_17, doorbell_range);
-	WREG32_SOC15(NBIO, 0, regS2A_DOORBELL_ENTRY_4_CTRL, doorbell_ctrl);
+		WREG32_SOC15_OFFSET(NBIO, 0, regDOORBELL0_CTRL_ENTRY_17,
+					aid_id, doorbell_range);
+		WREG32_SOC15_EXT(NBIO, aid_id, regS2A_DOORBELL_ENTRY_4_CTRL,
+				aid_id, doorbell_ctrl);
+	}
 }
 
 static void nbio_v7_9_enable_doorbell_aperture(struct amdgpu_device *adev,
@@ -235,7 +275,7 @@ static void nbio_v7_9_ih_doorbell_range(struct amdgpu_device *adev,
 		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
 				DOORBELL0_CTRL_ENTRY_0,
 				BIF_DOORBELL0_RANGE_SIZE_ENTRY,
-				0x4);
+				0x8);
 
 		ih_doorbell_ctrl = REG_SET_FIELD(ih_doorbell_ctrl,
 				S2A_DOORBELL_ENTRY_1_CTRL,
@@ -248,7 +288,7 @@ static void nbio_v7_9_ih_doorbell_range(struct amdgpu_device *adev,
 				S2A_DOORBELL_PORT1_RANGE_OFFSET, 0);
 		ih_doorbell_ctrl = REG_SET_FIELD(ih_doorbell_ctrl,
 				S2A_DOORBELL_ENTRY_1_CTRL,
-				S2A_DOORBELL_PORT1_RANGE_SIZE, 0x4);
+				S2A_DOORBELL_PORT1_RANGE_SIZE, 0x8);
 		ih_doorbell_ctrl = REG_SET_FIELD(ih_doorbell_ctrl,
 				S2A_DOORBELL_ENTRY_1_CTRL,
 				S2A_DOORBELL_PORT1_AWADDR_31_28_VALUE, 0);
@@ -319,6 +359,11 @@ static u32 nbio_v7_9_get_pcie_data_offset(struct amdgpu_device *adev)
 	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX0_PCIE_DATA2);
 }
 
+static u32 nbio_v7_9_get_pcie_index_hi_offset(struct amdgpu_device *adev)
+{
+	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX0_PCIE_INDEX2_HI);
+}
+
 const struct nbio_hdp_flush_reg nbio_v7_9_hdp_flush_reg = {
 	.ref_and_mask_cp0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP0_MASK,
 	.ref_and_mask_cp1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP1_MASK,
@@ -347,11 +392,103 @@ static void nbio_v7_9_enable_doorbell_interrupt(struct amdgpu_device *adev,
 			      DOORBELL_INTERRUPT_DISABLE, enable ? 0 : 1);
 }
 
+static int nbio_v7_9_get_compute_partition_mode(struct amdgpu_device *adev)
+{
+	u32 tmp, px;
+
+	tmp = RREG32_SOC15(NBIO, 0, regBIF_BX_PF0_PARTITION_COMPUTE_STATUS);
+	px = REG_GET_FIELD(tmp, BIF_BX_PF0_PARTITION_COMPUTE_STATUS,
+			   PARTITION_MODE);
+
+	return px;
+}
+
+static bool nbio_v7_9_is_nps_switch_requested(struct amdgpu_device *adev)
+{
+	u32 tmp;
+
+	tmp = RREG32_SOC15(NBIO, 0, regBIF_BX_PF0_PARTITION_MEM_STATUS);
+	tmp = REG_GET_FIELD(tmp, BIF_BX_PF0_PARTITION_MEM_STATUS,
+			    CHANGE_STATUE);
+
+	/* 0x8 - NPS switch requested */
+	return (tmp == 0x8);
+}
+static u32 nbio_v7_9_get_memory_partition_mode(struct amdgpu_device *adev,
+					       u32 *supp_modes)
+{
+	u32 tmp;
+
+	tmp = RREG32_SOC15(NBIO, 0, regBIF_BX_PF0_PARTITION_MEM_STATUS);
+	tmp = REG_GET_FIELD(tmp, BIF_BX_PF0_PARTITION_MEM_STATUS, NPS_MODE);
+
+	if (supp_modes) {
+		*supp_modes =
+			RREG32_SOC15(NBIO, 0, regBIF_BX_PF0_PARTITION_MEM_CAP);
+	}
+
+	return ffs(tmp);
+}
+
+static void nbio_v7_9_init_registers(struct amdgpu_device *adev)
+{
+	u32 inst_mask;
+	int i;
+
+	WREG32_SOC15(NBIO, 0, regXCC_DOORBELL_FENCE,
+		0xff & ~(adev->gfx.xcc_mask));
+
+	WREG32_SOC15(NBIO, 0, regBIFC_GFX_INT_MONITOR_MASK, 0x7ff);
+
+	inst_mask = adev->aid_mask & ~1U;
+	for_each_inst(i, inst_mask) {
+		WREG32_SOC15_EXT(NBIO, i, regXCC_DOORBELL_FENCE, i,
+			XCC_DOORBELL_FENCE__SHUB_SLV_MODE_MASK);
+
+	}
+
+	if (!amdgpu_sriov_vf(adev)) {
+		u32 baco_cntl;
+		for_each_inst(i, adev->aid_mask) {
+			baco_cntl = RREG32_SOC15(NBIO, i, regBIF_BX0_BACO_CNTL);
+			if (baco_cntl & (BIF_BX0_BACO_CNTL__BACO_DUMMY_EN_MASK |
+					 BIF_BX0_BACO_CNTL__BACO_EN_MASK)) {
+				baco_cntl &= ~(
+					BIF_BX0_BACO_CNTL__BACO_DUMMY_EN_MASK |
+					BIF_BX0_BACO_CNTL__BACO_EN_MASK);
+				dev_dbg(adev->dev,
+					"Unsetting baco dummy mode %x",
+					baco_cntl);
+				WREG32_SOC15(NBIO, i, regBIF_BX0_BACO_CNTL,
+					     baco_cntl);
+			}
+		}
+	}
+}
+
+#define MMIO_REG_HOLE_OFFSET 0x1A000
+
+static void nbio_v7_9_set_reg_remap(struct amdgpu_device *adev)
+{
+	if (!amdgpu_sriov_vf(adev) && (PAGE_SIZE <= 4096)) {
+		adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET;
+		adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET;
+	} else {
+		adev->rmmio_remap.reg_offset =
+			SOC15_REG_OFFSET(
+				NBIO, 0,
+				regBIF_BX_DEV0_EPF0_VF0_HDP_MEM_COHERENCY_FLUSH_CNTL)
+			<< 2;
+		adev->rmmio_remap.bus_addr = 0;
+	}
+}
+
 const struct amdgpu_nbio_funcs nbio_v7_9_funcs = {
 	.get_hdp_flush_req_offset = nbio_v7_9_get_hdp_flush_req_offset,
 	.get_hdp_flush_done_offset = nbio_v7_9_get_hdp_flush_done_offset,
 	.get_pcie_index_offset = nbio_v7_9_get_pcie_index_offset,
 	.get_pcie_data_offset = nbio_v7_9_get_pcie_data_offset,
+	.get_pcie_index_hi_offset = nbio_v7_9_get_pcie_index_hi_offset,
 	.get_rev_id = nbio_v7_9_get_rev_id,
 	.mc_access_enable = nbio_v7_9_mc_access_enable,
 	.get_memsize = nbio_v7_9_get_memsize,
@@ -366,4 +503,194 @@ const struct amdgpu_nbio_funcs nbio_v7_9_funcs = {
 	.get_clockgating_state = nbio_v7_9_get_clockgating_state,
 	.ih_control = nbio_v7_9_ih_control,
 	.remap_hdp_registers = nbio_v7_9_remap_hdp_registers,
+	.get_compute_partition_mode = nbio_v7_9_get_compute_partition_mode,
+	.get_memory_partition_mode = nbio_v7_9_get_memory_partition_mode,
+	.is_nps_switch_requested = nbio_v7_9_is_nps_switch_requested,
+	.init_registers = nbio_v7_9_init_registers,
+	.set_reg_remap = nbio_v7_9_set_reg_remap,
+};
+
+static void nbio_v7_9_query_ras_error_count(struct amdgpu_device *adev,
+					void *ras_error_status)
+{
+}
+
+static void nbio_v7_9_handle_ras_controller_intr_no_bifring(struct amdgpu_device *adev)
+{
+	uint32_t bif_doorbell_intr_cntl;
+	struct ras_manager *obj = amdgpu_ras_find_obj(adev, adev->nbio.ras_if);
+	struct ras_err_data err_data;
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+
+	if (amdgpu_ras_error_data_init(&err_data))
+		return;
+
+	bif_doorbell_intr_cntl = RREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_DOORBELL_INT_CNTL);
+
+	if (REG_GET_FIELD(bif_doorbell_intr_cntl,
+		BIF_BX0_BIF_DOORBELL_INT_CNTL, RAS_CNTLR_INTERRUPT_STATUS)) {
+		/* driver has to clear the interrupt status when bif ring is disabled */
+		bif_doorbell_intr_cntl = REG_SET_FIELD(bif_doorbell_intr_cntl,
+						BIF_BX0_BIF_DOORBELL_INT_CNTL,
+						RAS_CNTLR_INTERRUPT_CLEAR, 1);
+		WREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_DOORBELL_INT_CNTL, bif_doorbell_intr_cntl);
+
+		if (!ras->disable_ras_err_cnt_harvest) {
+			/*
+			 * clear error status after ras_controller_intr
+			 * according to hw team and count ue number
+			 * for query
+			 */
+			nbio_v7_9_query_ras_error_count(adev, &err_data);
+
+			/* logging on error cnt and printing for awareness */
+			obj->err_data.ue_count += err_data.ue_count;
+			obj->err_data.ce_count += err_data.ce_count;
+
+			if (err_data.ce_count)
+				dev_info(adev->dev, "%ld correctable hardware "
+						"errors detected in %s block\n",
+						obj->err_data.ce_count,
+						get_ras_block_str(adev->nbio.ras_if));
+
+			if (err_data.ue_count)
+				dev_info(adev->dev, "%ld uncorrectable hardware "
+						"errors detected in %s block\n",
+						obj->err_data.ue_count,
+						get_ras_block_str(adev->nbio.ras_if));
+		}
+
+		dev_info(adev->dev, "RAS controller interrupt triggered "
+					"by NBIF error\n");
+	}
+
+	amdgpu_ras_error_data_fini(&err_data);
+}
+
+static void nbio_v7_9_handle_ras_err_event_athub_intr_no_bifring(struct amdgpu_device *adev)
+{
+	uint32_t bif_doorbell_intr_cntl;
+
+	bif_doorbell_intr_cntl = RREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_DOORBELL_INT_CNTL);
+
+	if (REG_GET_FIELD(bif_doorbell_intr_cntl,
+		BIF_BX0_BIF_DOORBELL_INT_CNTL, RAS_ATHUB_ERR_EVENT_INTERRUPT_STATUS)) {
+		/* driver has to clear the interrupt status when bif ring is disabled */
+		bif_doorbell_intr_cntl = REG_SET_FIELD(bif_doorbell_intr_cntl,
+						BIF_BX0_BIF_DOORBELL_INT_CNTL,
+						RAS_ATHUB_ERR_EVENT_INTERRUPT_CLEAR, 1);
+
+		WREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_DOORBELL_INT_CNTL, bif_doorbell_intr_cntl);
+
+		amdgpu_ras_global_ras_isr(adev);
+	}
+}
+
+static int nbio_v7_9_set_ras_controller_irq_state(struct amdgpu_device *adev,
+						  struct amdgpu_irq_src *src,
+						  unsigned type,
+						  enum amdgpu_interrupt_state state)
+{
+	/* Dummy function, there is no initialization operation in driver */
+
+	return 0;
+}
+
+static int nbio_v7_9_process_ras_controller_irq(struct amdgpu_device *adev,
+						struct amdgpu_irq_src *source,
+						struct amdgpu_iv_entry *entry)
+{
+	/* By design, the ih cookie for ras_controller_irq should be written
+	 * to BIFring instead of general iv ring. However, due to known bif ring
+	 * hw bug, it has to be disabled. There is no chance the process function
+	 * will be involked. Just left it as a dummy one.
+	 */
+	return 0;
+}
+
+static int nbio_v7_9_set_ras_err_event_athub_irq_state(struct amdgpu_device *adev,
+						       struct amdgpu_irq_src *src,
+						       unsigned type,
+						       enum amdgpu_interrupt_state state)
+{
+	/* Dummy function, there is no initialization operation in driver */
+
+	return 0;
+}
+
+static int nbio_v7_9_process_err_event_athub_irq(struct amdgpu_device *adev,
+						 struct amdgpu_irq_src *source,
+						 struct amdgpu_iv_entry *entry)
+{
+	/* By design, the ih cookie for err_event_athub_irq should be written
+	 * to BIFring instead of general iv ring. However, due to known bif ring
+	 * hw bug, it has to be disabled. There is no chance the process function
+	 * will be involked. Just left it as a dummy one.
+	 */
+	return 0;
+}
+
+static const struct amdgpu_irq_src_funcs nbio_v7_9_ras_controller_irq_funcs = {
+	.set = nbio_v7_9_set_ras_controller_irq_state,
+	.process = nbio_v7_9_process_ras_controller_irq,
+};
+
+static const struct amdgpu_irq_src_funcs nbio_v7_9_ras_err_event_athub_irq_funcs = {
+	.set = nbio_v7_9_set_ras_err_event_athub_irq_state,
+	.process = nbio_v7_9_process_err_event_athub_irq,
+};
+
+static int nbio_v7_9_init_ras_controller_interrupt (struct amdgpu_device *adev)
+{
+	int r;
+
+	/* init the irq funcs */
+	adev->nbio.ras_controller_irq.funcs =
+		&nbio_v7_9_ras_controller_irq_funcs;
+	adev->nbio.ras_controller_irq.num_types = 1;
+
+	/* register ras controller interrupt */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_BIF,
+			      NBIF_7_4__SRCID__RAS_CONTROLLER_INTERRUPT,
+			      &adev->nbio.ras_controller_irq);
+
+	return r;
+}
+
+static int nbio_v7_9_init_ras_err_event_athub_interrupt (struct amdgpu_device *adev)
+{
+
+	int r;
+
+	/* init the irq funcs */
+	adev->nbio.ras_err_event_athub_irq.funcs =
+		&nbio_v7_9_ras_err_event_athub_irq_funcs;
+	adev->nbio.ras_err_event_athub_irq.num_types = 1;
+
+	/* register ras err event athub interrupt */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_BIF,
+			      NBIF_7_4__SRCID__ERREVENT_ATHUB_INTERRUPT,
+			      &adev->nbio.ras_err_event_athub_irq);
+
+	return r;
+}
+
+const struct amdgpu_ras_block_hw_ops nbio_v7_9_ras_hw_ops = {
+	.query_ras_error_count = nbio_v7_9_query_ras_error_count,
+};
+
+struct amdgpu_nbio_ras nbio_v7_9_ras = {
+	.ras_block = {
+		.ras_comm = {
+			.name = "pcie_bif",
+			.block = AMDGPU_RAS_BLOCK__PCIE_BIF,
+			.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE,
+		},
+		.hw_ops = &nbio_v7_9_ras_hw_ops,
+		.ras_late_init = amdgpu_nbio_ras_late_init,
+	},
+	.handle_ras_controller_intr_no_bifring = nbio_v7_9_handle_ras_controller_intr_no_bifring,
+	.handle_ras_err_event_athub_intr_no_bifring = nbio_v7_9_handle_ras_err_event_athub_intr_no_bifring,
+	.init_ras_controller_interrupt = nbio_v7_9_init_ras_controller_interrupt,
+	.init_ras_err_event_athub_interrupt = nbio_v7_9_init_ras_err_event_athub_interrupt,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_9.h b/drivers/gpu/drm/amd/amdgpu/nbio_v7_9.h
index 8e04eb484328..73709771950d 100644
--- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_9.h
+++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_9.h
@@ -28,5 +28,6 @@
 
 extern const struct nbio_hdp_flush_reg nbio_v7_9_hdp_flush_reg;
 extern const struct amdgpu_nbio_funcs nbio_v7_9_funcs;
+extern struct amdgpu_nbio_ras nbio_v7_9_ras;
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/nv.c b/drivers/gpu/drm/amd/amdgpu/nv.c
index 47420b403871..50e77d9b30af 100644
--- a/drivers/gpu/drm/amd/amdgpu/nv.c
+++ b/drivers/gpu/drm/amd/amdgpu/nv.c
@@ -59,7 +59,6 @@
 #include "vcn_v3_0.h"
 #include "jpeg_v3_0.h"
 #include "amdgpu_vkms.h"
-#include "mes_v10_1.h"
 #include "mxgpu_nv.h"
 #include "smuio_v11_0.h"
 #include "smuio_v11_0_6.h"
@@ -67,116 +66,112 @@
 static const struct amd_ip_funcs nv_common_ip_funcs;
 
 /* Navi */
-static const struct amdgpu_video_codec_info nv_video_codecs_encode_array[] =
-{
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 2304, 0)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 4096, 2304, 0)},
+static const struct amdgpu_video_codec_info nv_video_codecs_encode_array[] = {
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 4096, 4096, 0)},
 };
 
-static const struct amdgpu_video_codecs nv_video_codecs_encode =
-{
+static const struct amdgpu_video_codecs nv_video_codecs_encode = {
 	.codec_count = ARRAY_SIZE(nv_video_codecs_encode_array),
 	.codec_array = nv_video_codecs_encode_array,
 };
 
 /* Navi1x */
-static const struct amdgpu_video_codec_info nv_video_codecs_decode_array[] =
-{
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)},
+static const struct amdgpu_video_codec_info nv_video_codecs_decode_array[] = {
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 1920, 1088, 3)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 1920, 1088, 5)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 1920, 1088, 4)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 8192, 8192, 0)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)},
 };
 
-static const struct amdgpu_video_codecs nv_video_codecs_decode =
-{
+static const struct amdgpu_video_codecs nv_video_codecs_decode = {
 	.codec_count = ARRAY_SIZE(nv_video_codecs_decode_array),
 	.codec_array = nv_video_codecs_decode_array,
 };
 
 /* Sienna Cichlid */
-static const struct amdgpu_video_codec_info sc_video_codecs_decode_array_vcn0[] =
-{
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)},
+static const struct amdgpu_video_codec_info sc_video_codecs_encode_array[] = {
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 0)},
+};
+
+static const struct amdgpu_video_codecs sc_video_codecs_encode = {
+	.codec_count = ARRAY_SIZE(sc_video_codecs_encode_array),
+	.codec_array = sc_video_codecs_encode_array,
+};
+
+static const struct amdgpu_video_codec_info sc_video_codecs_decode_array_vcn0[] = {
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 1920, 1088, 3)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 1920, 1088, 5)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 1920, 1088, 4)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 16384, 16384, 0)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_AV1, 8192, 4352, 0)},
 };
 
-static const struct amdgpu_video_codec_info sc_video_codecs_decode_array_vcn1[] =
-{
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)},
+static const struct amdgpu_video_codec_info sc_video_codecs_decode_array_vcn1[] = {
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 1920, 1088, 3)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 1920, 1088, 5)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 1920, 1088, 4)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 16384, 16384, 0)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)},
 };
 
-static const struct amdgpu_video_codecs sc_video_codecs_decode_vcn0 =
-{
+static const struct amdgpu_video_codecs sc_video_codecs_decode_vcn0 = {
 	.codec_count = ARRAY_SIZE(sc_video_codecs_decode_array_vcn0),
 	.codec_array = sc_video_codecs_decode_array_vcn0,
 };
 
-static const struct amdgpu_video_codecs sc_video_codecs_decode_vcn1 =
-{
+static const struct amdgpu_video_codecs sc_video_codecs_decode_vcn1 = {
 	.codec_count = ARRAY_SIZE(sc_video_codecs_decode_array_vcn1),
 	.codec_array = sc_video_codecs_decode_array_vcn1,
 };
 
 /* SRIOV Sienna Cichlid, not const since data is controlled by host */
-static struct amdgpu_video_codec_info sriov_sc_video_codecs_encode_array[] =
-{
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 2304, 0)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 4096, 2304, 0)},
+static struct amdgpu_video_codec_info sriov_sc_video_codecs_encode_array[] = {
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 0)},
 };
 
-static struct amdgpu_video_codec_info sriov_sc_video_codecs_decode_array_vcn0[] =
-{
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)},
+static struct amdgpu_video_codec_info sriov_sc_video_codecs_decode_array_vcn0[] = {
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 1920, 1088, 3)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 1920, 1088, 5)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 1920, 1088, 4)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 16384, 16384, 0)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_AV1, 8192, 4352, 0)},
 };
 
-static struct amdgpu_video_codec_info sriov_sc_video_codecs_decode_array_vcn1[] =
-{
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)},
+static struct amdgpu_video_codec_info sriov_sc_video_codecs_decode_array_vcn1[] = {
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 1920, 1088, 3)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 1920, 1088, 5)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 1920, 1088, 4)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 16384, 16384, 0)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)},
 };
 
-static struct amdgpu_video_codecs sriov_sc_video_codecs_encode =
-{
+static struct amdgpu_video_codecs sriov_sc_video_codecs_encode = {
 	.codec_count = ARRAY_SIZE(sriov_sc_video_codecs_encode_array),
 	.codec_array = sriov_sc_video_codecs_encode_array,
 };
 
-static struct amdgpu_video_codecs sriov_sc_video_codecs_decode_vcn0 =
-{
+static struct amdgpu_video_codecs sriov_sc_video_codecs_decode_vcn0 = {
 	.codec_count = ARRAY_SIZE(sriov_sc_video_codecs_decode_array_vcn0),
 	.codec_array = sriov_sc_video_codecs_decode_array_vcn0,
 };
 
-static struct amdgpu_video_codecs sriov_sc_video_codecs_decode_vcn1 =
-{
+static struct amdgpu_video_codecs sriov_sc_video_codecs_decode_vcn1 = {
 	.codec_count = ARRAY_SIZE(sriov_sc_video_codecs_decode_array_vcn1),
 	.codec_array = sriov_sc_video_codecs_decode_array_vcn1,
 };
@@ -203,7 +198,7 @@ static const struct amdgpu_video_codec_info yc_video_codecs_decode_array[] = {
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 16384, 16384, 0)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_AV1, 8192, 4352, 0)},
 };
 
@@ -218,7 +213,7 @@ static int nv_query_video_codecs(struct amdgpu_device *adev, bool encode,
 	if (adev->vcn.num_vcn_inst == hweight8(adev->vcn.harvest_config))
 		return -EINVAL;
 
-	switch (adev->ip_versions[UVD_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, UVD_HWIP, 0)) {
 	case IP_VERSION(3, 0, 0):
 	case IP_VERSION(3, 0, 64):
 	case IP_VERSION(3, 0, 192):
@@ -237,12 +232,12 @@ static int nv_query_video_codecs(struct amdgpu_device *adev, bool encode,
 		} else {
 			if (adev->vcn.harvest_config & AMDGPU_VCN_HARVEST_VCN0) {
 				if (encode)
-					*codecs = &nv_video_codecs_encode;
+					*codecs = &sc_video_codecs_encode;
 				else
 					*codecs = &sc_video_codecs_decode_vcn1;
 			} else {
 				if (encode)
-					*codecs = &nv_video_codecs_encode;
+					*codecs = &sc_video_codecs_encode;
 				else
 					*codecs = &sc_video_codecs_decode_vcn0;
 			}
@@ -251,14 +246,14 @@ static int nv_query_video_codecs(struct amdgpu_device *adev, bool encode,
 	case IP_VERSION(3, 0, 16):
 	case IP_VERSION(3, 0, 2):
 		if (encode)
-			*codecs = &nv_video_codecs_encode;
+			*codecs = &sc_video_codecs_encode;
 		else
 			*codecs = &sc_video_codecs_decode_vcn0;
 		return 0;
 	case IP_VERSION(3, 1, 1):
 	case IP_VERSION(3, 1, 2):
 		if (encode)
-			*codecs = &nv_video_codecs_encode;
+			*codecs = &sc_video_codecs_encode;
 		else
 			*codecs = &yc_video_codecs_decode;
 		return 0;
@@ -331,11 +326,6 @@ void nv_grbm_select(struct amdgpu_device *adev,
 	WREG32_SOC15(GC, 0, mmGRBM_GFX_CNTL, grbm_gfx_cntl);
 }
 
-static void nv_vga_set_state(struct amdgpu_device *adev, bool state)
-{
-	/* todo */
-}
-
 static bool nv_read_disabled_bios(struct amdgpu_device *adev)
 {
 	/* todo */
@@ -371,12 +361,12 @@ static uint32_t nv_read_indexed_register(struct amdgpu_device *adev, u32 se_num,
 
 	mutex_lock(&adev->grbm_idx_mutex);
 	if (se_num != 0xffffffff || sh_num != 0xffffffff)
-		amdgpu_gfx_select_se_sh(adev, se_num, sh_num, 0xffffffff);
+		amdgpu_gfx_select_se_sh(adev, se_num, sh_num, 0xffffffff, 0);
 
 	val = RREG32(reg_offset);
 
 	if (se_num != 0xffffffff || sh_num != 0xffffffff)
-		amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+		amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 	return val;
 }
@@ -462,8 +452,9 @@ nv_asic_reset_method(struct amdgpu_device *adev)
 		dev_warn(adev->dev, "Specified reset method:%d isn't supported, using AUTO instead.\n",
 				  amdgpu_reset_method);
 
-	switch (adev->ip_versions[MP1_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
 	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 2):
 	case IP_VERSION(13, 0, 1):
 	case IP_VERSION(13, 0, 3):
 	case IP_VERSION(13, 0, 5):
@@ -522,24 +513,15 @@ static int nv_set_vce_clocks(struct amdgpu_device *adev, u32 evclk, u32 ecclk)
 
 static void nv_program_aspm(struct amdgpu_device *adev)
 {
-	if (!amdgpu_device_should_use_aspm(adev) || !amdgpu_device_aspm_support_quirk())
+	if (!amdgpu_device_should_use_aspm(adev))
 		return;
 
-	if (!(adev->flags & AMD_IS_APU) &&
-	    (adev->nbio.funcs->program_aspm))
+	if (adev->nbio.funcs->program_aspm)
 		adev->nbio.funcs->program_aspm(adev);
 
 }
 
-static void nv_enable_doorbell_aperture(struct amdgpu_device *adev,
-					bool enable)
-{
-	adev->nbio.funcs->enable_doorbell_aperture(adev, enable);
-	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, enable);
-}
-
-const struct amdgpu_ip_block_version nv_common_ip_block =
-{
+const struct amdgpu_ip_block_version nv_common_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_COMMON,
 	.major = 1,
 	.minor = 0,
@@ -574,16 +556,6 @@ static bool nv_need_reset_on_init(struct amdgpu_device *adev)
 	return false;
 }
 
-static uint64_t nv_get_pcie_replay_count(struct amdgpu_device *adev)
-{
-
-	/* TODO
-	 * dummy implement for pcie_replay_count sysfs interface
-	 * */
-
-	return 0;
-}
-
 static void nv_init_doorbell_index(struct amdgpu_device *adev)
 {
 	adev->doorbell_index.kiq = AMDGPU_NAVI10_DOORBELL_KIQ;
@@ -629,29 +601,26 @@ static int nv_update_umd_stable_pstate(struct amdgpu_device *adev,
 				       bool enter)
 {
 	if (enter)
-		amdgpu_gfx_rlc_enter_safe_mode(adev);
+		amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 	else
-		amdgpu_gfx_rlc_exit_safe_mode(adev);
+		amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 
 	if (adev->gfx.funcs->update_perfmon_mgcg)
 		adev->gfx.funcs->update_perfmon_mgcg(adev, !enter);
 
-	if (!(adev->flags & AMD_IS_APU) &&
-	    (adev->nbio.funcs->enable_aspm) &&
-	     amdgpu_device_should_use_aspm(adev))
+	if (adev->nbio.funcs->enable_aspm &&
+	    amdgpu_device_should_use_aspm(adev))
 		adev->nbio.funcs->enable_aspm(adev, !enter);
 
 	return 0;
 }
 
-static const struct amdgpu_asic_funcs nv_asic_funcs =
-{
+static const struct amdgpu_asic_funcs nv_asic_funcs = {
 	.read_disabled_bios = &nv_read_disabled_bios,
 	.read_bios_from_rom = &amdgpu_soc15_read_bios_from_rom,
 	.read_register = &nv_read_register,
 	.reset = &nv_asic_reset,
 	.reset_method = &nv_asic_reset_method,
-	.set_vga_state = &nv_vga_set_state,
 	.get_xclk = &nv_get_xclk,
 	.set_uvd_clocks = &nv_set_uvd_clocks,
 	.set_vce_clocks = &nv_set_vce_clocks,
@@ -659,22 +628,18 @@ static const struct amdgpu_asic_funcs nv_asic_funcs =
 	.init_doorbell_index = &nv_init_doorbell_index,
 	.need_full_reset = &nv_need_full_reset,
 	.need_reset_on_init = &nv_need_reset_on_init,
-	.get_pcie_replay_count = &nv_get_pcie_replay_count,
+	.get_pcie_replay_count = &amdgpu_nbio_get_pcie_replay_count,
 	.supports_baco = &amdgpu_dpm_is_baco_supported,
 	.pre_asic_init = &nv_pre_asic_init,
 	.update_umd_stable_pstate = &nv_update_umd_stable_pstate,
 	.query_video_codecs = &nv_query_video_codecs,
 };
 
-static int nv_common_early_init(void *handle)
+static int nv_common_early_init(struct amdgpu_ip_block *ip_block)
 {
-#define MMIO_REG_HOLE_OFFSET (0x80000 - PAGE_SIZE)
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	if (!amdgpu_sriov_vf(adev)) {
-		adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET;
-		adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET;
-	}
+	adev->nbio.funcs->set_reg_remap(adev);
 	adev->smc_rreg = NULL;
 	adev->smc_wreg = NULL;
 	adev->pcie_rreg = &amdgpu_device_indirect_rreg;
@@ -698,7 +663,7 @@ static int nv_common_early_init(void *handle)
 	/* TODO: split the GC and PG flags based on the relevant IP version for which
 	 * they are relevant.
 	 */
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 		adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG |
 			AMD_CG_SUPPORT_GFX_CGCG |
@@ -892,7 +857,8 @@ static int nv_common_early_init(void *handle)
 			AMD_CG_SUPPORT_ATHUB_LS |
 			AMD_CG_SUPPORT_IH_CG |
 			AMD_CG_SUPPORT_VCN_MGCG |
-			AMD_CG_SUPPORT_JPEG_MGCG;
+			AMD_CG_SUPPORT_JPEG_MGCG |
+			AMD_CG_SUPPORT_SDMA_MGCG;
 		adev->pg_flags = AMD_PG_SUPPORT_GFX_PG |
 			AMD_PG_SUPPORT_VCN |
 			AMD_PG_SUPPORT_VCN_DPG |
@@ -953,7 +919,8 @@ static int nv_common_early_init(void *handle)
 			AMD_CG_SUPPORT_ATHUB_LS |
 			AMD_CG_SUPPORT_IH_CG |
 			AMD_CG_SUPPORT_VCN_MGCG |
-			AMD_CG_SUPPORT_JPEG_MGCG;
+			AMD_CG_SUPPORT_JPEG_MGCG |
+			AMD_CG_SUPPORT_SDMA_MGCG;
 		adev->pg_flags = AMD_PG_SUPPORT_VCN |
 			AMD_PG_SUPPORT_VCN_DPG |
 			AMD_PG_SUPPORT_JPEG |
@@ -978,9 +945,9 @@ static int nv_common_early_init(void *handle)
 	return 0;
 }
 
-static int nv_common_late_init(void *handle)
+static int nv_common_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev)) {
 		xgpu_nv_mailbox_get_irq(adev);
@@ -999,12 +966,17 @@ static int nv_common_late_init(void *handle)
 		}
 	}
 
+	/* Enable selfring doorbell aperture late because doorbell BAR
+	 * aperture will change if resize BAR successfully in gmc sw_init.
+	 */
+	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, true);
+
 	return 0;
 }
 
-static int nv_common_sw_init(void *handle)
+static int nv_common_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		xgpu_nv_mailbox_add_irq_id(adev);
@@ -1012,14 +984,9 @@ static int nv_common_sw_init(void *handle)
 	return 0;
 }
 
-static int nv_common_sw_fini(void *handle)
+static int nv_common_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	return 0;
-}
-
-static int nv_common_hw_init(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (adev->nbio.funcs->apply_lc_spc_mode_wa)
 		adev->nbio.funcs->apply_lc_spc_mode_wa(adev);
@@ -1038,59 +1005,50 @@ static int nv_common_hw_init(void *handle)
 	if (adev->nbio.funcs->remap_hdp_registers && !amdgpu_sriov_vf(adev))
 		adev->nbio.funcs->remap_hdp_registers(adev);
 	/* enable the doorbell aperture */
-	nv_enable_doorbell_aperture(adev, true);
+	adev->nbio.funcs->enable_doorbell_aperture(adev, true);
 
 	return 0;
 }
 
-static int nv_common_hw_fini(void *handle)
+static int nv_common_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	/* disable the doorbell aperture */
-	nv_enable_doorbell_aperture(adev, false);
+	/* Disable the doorbell aperture and selfring doorbell aperture
+	 * separately in hw_fini because nv_enable_doorbell_aperture
+	 * has been removed and there is no need to delay disabling
+	 * selfring doorbell.
+	 */
+	adev->nbio.funcs->enable_doorbell_aperture(adev, false);
+	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, false);
 
 	return 0;
 }
 
-static int nv_common_suspend(void *handle)
+static int nv_common_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return nv_common_hw_fini(adev);
+	return nv_common_hw_fini(ip_block);
 }
 
-static int nv_common_resume(void *handle)
+static int nv_common_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return nv_common_hw_init(adev);
+	return nv_common_hw_init(ip_block);
 }
 
-static bool nv_common_is_idle(void *handle)
+static bool nv_common_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	return true;
 }
 
-static int nv_common_wait_for_idle(void *handle)
-{
-	return 0;
-}
-
-static int nv_common_soft_reset(void *handle)
-{
-	return 0;
-}
-
-static int nv_common_set_clockgating_state(void *handle,
+static int nv_common_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[NBIO_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, NBIO_HWIP, 0)) {
 	case IP_VERSION(2, 3, 0):
 	case IP_VERSION(2, 3, 1):
 	case IP_VERSION(2, 3, 2):
@@ -1113,16 +1071,16 @@ static int nv_common_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int nv_common_set_powergating_state(void *handle,
+static int nv_common_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_powergating_state state)
 {
 	/* TODO */
 	return 0;
 }
 
-static void nv_common_get_clockgating_state(void *handle, u64 *flags)
+static void nv_common_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		*flags = 0;
@@ -1132,8 +1090,6 @@ static void nv_common_get_clockgating_state(void *handle, u64 *flags)
 	adev->hdp.funcs->get_clock_gating_state(adev, flags);
 
 	adev->smuio.funcs->get_clock_gating_state(adev, flags);
-
-	return;
 }
 
 static const struct amd_ip_funcs nv_common_ip_funcs = {
@@ -1141,14 +1097,11 @@ static const struct amd_ip_funcs nv_common_ip_funcs = {
 	.early_init = nv_common_early_init,
 	.late_init = nv_common_late_init,
 	.sw_init = nv_common_sw_init,
-	.sw_fini = nv_common_sw_fini,
 	.hw_init = nv_common_hw_init,
 	.hw_fini = nv_common_hw_fini,
 	.suspend = nv_common_suspend,
 	.resume = nv_common_resume,
 	.is_idle = nv_common_is_idle,
-	.wait_for_idle = nv_common_wait_for_idle,
-	.soft_reset = nv_common_soft_reset,
 	.set_clockgating_state = nv_common_set_clockgating_state,
 	.set_powergating_state = nv_common_set_powergating_state,
 	.get_clockgating_state = nv_common_get_clockgating_state,
diff --git a/drivers/gpu/drm/amd/amdgpu/nv.h b/drivers/gpu/drm/amd/amdgpu/nv.h
index 83e9782aef39..8f4817404f10 100644
--- a/drivers/gpu/drm/amd/amdgpu/nv.h
+++ b/drivers/gpu/drm/amd/amdgpu/nv.h
@@ -31,5 +31,6 @@ extern const struct amdgpu_ip_block_version nv_common_ip_block;
 void nv_grbm_select(struct amdgpu_device *adev,
 		    u32 me, u32 pipe, u32 queue, u32 vmid);
 void nv_set_virt_ops(struct amdgpu_device *adev);
+int cyan_skillfish_reg_base_init(struct amdgpu_device *adev);
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/nvd.h b/drivers/gpu/drm/amd/amdgpu/nvd.h
index fd6b58243b03..56f1bfac0b20 100644
--- a/drivers/gpu/drm/amd/amdgpu/nvd.h
+++ b/drivers/gpu/drm/amd/amdgpu/nvd.h
@@ -64,6 +64,24 @@
 #define	PACKET3_INDIRECT_BUFFER_CNST_END		0x19
 #define	PACKET3_ATOMIC_GDS				0x1D
 #define	PACKET3_ATOMIC_MEM				0x1E
+#define 	PACKET3_ATOMIC_MEM__ATOMIC(x) ((((unsigned)(x)) & 0x7F) << 0)
+#define 	PACKET3_ATOMIC_MEM__COMMAND(x) ((((unsigned)(x)) & 0xF) << 8)
+#define 	PACKET3_ATOMIC_MEM__CACHE_POLICY(x) ((((unsigned)(x)) & 0x3) << 25)
+#define 	PACKET3_ATOMIC_MEM__ADDR_LO(x) (((unsigned)(x)))
+#define 	PACKET3_ATOMIC_MEM__ADDR_HI(x) (((unsigned)(x)))
+#define 	PACKET3_ATOMIC_MEM__SRC_DATA_LO(x) (((unsigned)(x)))
+#define 	PACKET3_ATOMIC_MEM__SRC_DATA_HI(x) (((unsigned)(x)))
+#define 	PACKET3_ATOMIC_MEM__CMP_DATA_LO(x) (((unsigned)(x)))
+#define 	PACKET3_ATOMIC_MEM__CMP_DATA_HI(x) (((unsigned)(x)))
+#define 	PACKET3_ATOMIC_MEM__LOOP_INTERVAL(x) ((((unsigned)(x)) & 0x1FFF) << 0)
+#define 	PACKET3_ATOMIC_MEM__COMMAND__SINGLE_PASS_ATOMIC 0
+#define 	PACKET3_ATOMIC_MEM__COMMAND__LOOP_UNTIL_COMPARE_SATISFIED 1
+#define 	PACKET3_ATOMIC_MEM__COMMAND__WAIT_FOR_WRITE_CONFIRMATION 2
+#define 	PACKET3_ATOMIC_MEM__COMMAND__SEND_AND_CONTINUE 3
+#define 	PACKET3_ATOMIC_MEM__CACHE_POLICY__LRU 0
+#define 	PACKET3_ATOMIC_MEM__CACHE_POLICY__STREAM 1
+#define 	PACKET3_ATOMIC_MEM__CACHE_POLICY__NOA 2
+#define 	PACKET3_ATOMIC_MEM__CACHE_POLICY__BYPASS 3
 #define	PACKET3_OCCLUSION_QUERY				0x1F
 #define	PACKET3_SET_PREDICATION				0x20
 #define	PACKET3_REG_RMW					0x21
@@ -105,6 +123,38 @@
 		 * 1 - pfp
 		 * 2 - ce
 		 */
+#define 	PACKET3_WRITE_DATA__DST_SEL(x) ((((unsigned)(x)) & 0xF) << 8)
+#define 	PACKET3_WRITE_DATA__ADDR_INCR(x) ((((unsigned)(x)) & 0x1) << 16)
+#define 	PACKET3_WRITE_DATA__WR_CONFIRM(x) ((((unsigned)(x)) & 0x1) << 20)
+#define 	PACKET3_WRITE_DATA__CACHE_POLICY(x) ((((unsigned)(x)) & 0x3) << 25)
+#define 	PACKET3_WRITE_DATA__DST_MMREG_ADDR(x) ((((unsigned)(x)) & 0x3FFFF) << 0)
+#define 	PACKET3_WRITE_DATA__DST_GDS_ADDR(x) ((((unsigned)(x)) & 0xFFFF) << 0)
+#define 	PACKET3_WRITE_DATA__DST_MEM_ADDR_LO(x) ((((unsigned)(x)) & 0x3FFFFFFF) << 2)
+#define 	PACKET3_WRITE_DATA__DST_MEM_ADDR_HI(x) ((unsigned)(x))
+#define 	PACKET3_WRITE_DATA__MODE(x) ((((unsigned)(x)) & 0x1) << 21)
+#define 	PACKET3_WRITE_DATA__AID_ID(x) ((((unsigned)(x)) & 0x3) << 22)
+#define 	PACKET3_WRITE_DATA__TEMPORAL(x) ((((unsigned)(x)) & 0x3) << 24)
+#define 	PACKET3_WRITE_DATA__DST_MMREG_ADDR_LO(x) ((unsigned)(x))
+#define 	PACKET3_WRITE_DATA__DST_MMREG_ADDR_HI(x) ((((unsigned)(x)) & 0xFF) << 0)
+#define 	PACKET3_WRITE_DATA__DST_SEL__MEM_MAPPED_REGISTER 0
+#define 	PACKET3_WRITE_DATA__DST_SEL__TC_L2 2
+#define 	PACKET3_WRITE_DATA__DST_SEL__GDS 3
+#define 	PACKET3_WRITE_DATA__DST_SEL__MEMORY 5
+#define 	PACKET3_WRITE_DATA__DST_SEL__MEMORY_MAPPED_ADC_PERSISTENT_STATE 6
+#define 	PACKET3_WRITE_DATA__ADDR_INCR__INCREMENT_ADDRESS 0
+#define 	PACKET3_WRITE_DATA__ADDR_INCR__DO_NOT_INCREMENT_ADDRESS 1
+#define 	PACKET3_WRITE_DATA__WR_CONFIRM__DO_NOT_WAIT_FOR_WRITE_CONFIRMATION 0
+#define 	PACKET3_WRITE_DATA__WR_CONFIRM__WAIT_FOR_WRITE_CONFIRMATION 1
+#define 	PACKET3_WRITE_DATA__MODE__PF_VF_DISABLED 0
+#define 	PACKET3_WRITE_DATA__MODE__PF_VF_ENABLED 1
+#define 	PACKET3_WRITE_DATA__TEMPORAL__RT 0
+#define 	PACKET3_WRITE_DATA__TEMPORAL__NT 1
+#define 	PACKET3_WRITE_DATA__TEMPORAL__HT 2
+#define 	PACKET3_WRITE_DATA__TEMPORAL__LU 3
+#define 	PACKET3_WRITE_DATA__CACHE_POLICY__LRU 0
+#define 	PACKET3_WRITE_DATA__CACHE_POLICY__STREAM 1
+#define 	PACKET3_WRITE_DATA__CACHE_POLICY__NOA 2
+#define 	PACKET3_WRITE_DATA__CACHE_POLICY__BYPASS 3
 #define	PACKET3_DRAW_INDEX_INDIRECT_MULTI		0x38
 #define	PACKET3_MEM_SEMAPHORE				0x39
 #              define PACKET3_SEM_USE_MAILBOX       (0x1 << 16)
@@ -135,6 +185,42 @@
 		/* 0 - me
 		 * 1 - pfp
 		 */
+#define		PACKET3_WAIT_REG_MEM__FUNCTION(x) ((((unsigned)(x)) & 0x7) << 0)
+#define		PACKET3_WAIT_REG_MEM__MEM_SPACE(x) ((((unsigned)(x)) & 0x3) << 4)
+#define		PACKET3_WAIT_REG_MEM__OPERATION(x) ((((unsigned)(x)) & 0x3) << 6)
+#define		PACKET3_WAIT_REG_MEM__MES_INTR_PIPE(x) ((((unsigned)(x)) & 0x3) << 22)
+#define		PACKET3_WAIT_REG_MEM__MES_ACTION(x) ((((unsigned)(x)) & 0x1) << 24)
+#define		PACKET3_WAIT_REG_MEM__CACHE_POLICY(x) ((((unsigned)(x)) & 0x3) << 25)
+#define		PACKET3_WAIT_REG_MEM__TEMPORAL(x) ((((unsigned)(x)) & 0x3) << 25)
+#define		PACKET3_WAIT_REG_MEM__MEM_POLL_ADDR_LO(x) ((((unsigned)(x)) & 0x3FFFFFFF) << 2)
+#define		PACKET3_WAIT_REG_MEM__REG_POLL_ADDR(x) ((((unsigned)(x)) & 0X3FFFF) << 0)
+#define		PACKET3_WAIT_REG_MEM__REG_WRITE_ADDR1(x) ((((unsigned)(x)) & 0X3FFFF) << 0)
+#define		PACKET3_WAIT_REG_MEM__MEM_POLL_ADDR_HI(x) ((unsigned)(x))
+#define		PACKET3_WAIT_REG_MEM__REG_WRITE_ADDR2(x) ((((unsigned)(x)) & 0x3FFFF) << 0)
+#define		PACKET3_WAIT_REG_MEM__REFERENCE(x) ((unsigned)(x))
+#define		PACKET3_WAIT_REG_MEM__MASK(x) ((unsigned)(x))
+#define		PACKET3_WAIT_REG_MEM__POLL_INTERVAL(x) ((((unsigned)(x)) & 0xFFFF) << 0)
+#define		PACKET3_WAIT_REG_MEM__OPTIMIZE_ACE_OFFLOAD_MODE(x) ((((unsigned)(x)) & 0x1) << 31)
+#define 	PACKET3_WAIT_REG_MEM__FUNCTION__ALWAYS_PASS 0
+#define 	PACKET3_WAIT_REG_MEM__FUNCTION__LESS_THAN_REF_VALUE 1
+#define 	PACKET3_WAIT_REG_MEM__FUNCTION__LESS_THAN_EQUAL_TO_THE_REF_VALUE 2
+#define 	PACKET3_WAIT_REG_MEM__FUNCTION__EQUAL_TO_THE_REFERENCE_VALUE 3
+#define 	PACKET3_WAIT_REG_MEM__FUNCTION__NOT_EQUAL_REFERENCE_VALUE 4
+#define 	PACKET3_WAIT_REG_MEM__FUNCTION__GREATER_THAN_OR_EQUAL_REFERENCE_VALUE 5
+#define 	PACKET3_WAIT_REG_MEM__FUNCTION__GREATER_THAN_REFERENCE_VALUE 6
+#define 	PACKET3_WAIT_REG_MEM__MEM_SPACE__REGISTER_SPACE 0
+#define 	PACKET3_WAIT_REG_MEM__MEM_SPACE__MEMORY_SPACE 1
+#define 	PACKET3_WAIT_REG_MEM__OPERATION__WAIT_REG_MEM 0
+#define 	PACKET3_WAIT_REG_MEM__OPERATION__WR_WAIT_WR_REG 1
+#define 	PACKET3_WAIT_REG_MEM__OPERATION__WAIT_MEM_PREEMPTABLE 3
+#define 	PACKET3_WAIT_REG_MEM__CACHE_POLICY__LRU 0
+#define 	PACKET3_WAIT_REG_MEM__CACHE_POLICY__STREAM 1
+#define 	PACKET3_WAIT_REG_MEM__CACHE_POLICY__NOA 2
+#define 	PACKET3_WAIT_REG_MEM__CACHE_POLICY__BYPASS 3
+#define 	PACKET3_WAIT_REG_MEM__TEMPORAL__RT 0
+#define 	PACKET3_WAIT_REG_MEM__TEMPORAL__NT 1
+#define 	PACKET3_WAIT_REG_MEM__TEMPORAL__HT 2
+#define 	PACKET3_WAIT_REG_MEM__TEMPORAL__LU 3
 #define	PACKET3_INDIRECT_BUFFER				0x3F
 #define		INDIRECT_BUFFER_VALID                   (1 << 23)
 #define		INDIRECT_BUFFER_CACHE_POLICY(x)         ((x) << 28)
@@ -144,8 +230,94 @@
 		 */
 #define		INDIRECT_BUFFER_PRE_ENB(x)		((x) << 21)
 #define		INDIRECT_BUFFER_PRE_RESUME(x)           ((x) << 30)
+#define 	PACKET3_INDIRECT_BUFFER__IB_BASE_LO(x) ((((unsigned)(x)) & 0x3FFFFFFF) << 2)
+#define 	PACKET3_INDIRECT_BUFFER__IB_BASE_HI(x) ((unsigned)(x))
+#define 	PACKET3_INDIRECT_BUFFER__IB_SIZE(x) ((((unsigned)(x)) & 0xFFFFF) << 0)
+#define 	PACKET3_INDIRECT_BUFFER__CHAIN(x) ((((unsigned)(x)) & 0x1) << 20)
+#define 	PACKET3_INDIRECT_BUFFER__OFFLOAD_POLLING(x) ((((unsigned)(x)) & 0x1) << 21)
+#define 	PACKET3_INDIRECT_BUFFER__VALID(x) ((((unsigned)(x)) & 0x1) << 23)
+#define 	PACKET3_INDIRECT_BUFFER__VMID(x) ((((unsigned)(x)) & 0xF) << 24)
+#define 	PACKET3_INDIRECT_BUFFER__CACHE_POLICY(x) ((((unsigned)(x)) & 0x3) << 28)
+#define 	PACKET3_INDIRECT_BUFFER__TEMPORAL(x) ((((unsigned)(x)) & 0x3) << 28)
+#define 	PACKET3_INDIRECT_BUFFER__PRIV(x) ((((unsigned)(x)) & 0x1) << 31)
+#define 	PACKET3_INDIRECT_BUFFER__TEMPORAL__RT 0
+#define 	PACKET3_INDIRECT_BUFFER__TEMPORAL__NT 1
+#define 	PACKET3_INDIRECT_BUFFER__TEMPORAL__HT 2
+#define 	PACKET3_INDIRECT_BUFFER__TEMPORAL__LU 3
+#define 	PACKET3_INDIRECT_BUFFER__CACHE_POLICY__LRU 0
+#define 	PACKET3_INDIRECT_BUFFER__CACHE_POLICY__STREAM 1
+#define 	PACKET3_INDIRECT_BUFFER__CACHE_POLICY__NOA 2
+#define 	PACKET3_INDIRECT_BUFFER__CACHE_POLICY__BYPASS 3
 #define	PACKET3_COND_INDIRECT_BUFFER			0x3F
 #define	PACKET3_COPY_DATA				0x40
+#define 	PACKET3_COPY_DATA__SRC_SEL(x) ((((unsigned)(x)) & 0xF) << 0)
+#define 	PACKET3_COPY_DATA__DST_SEL(x) ((((unsigned)(x)) & 0xF) << 8)
+#define 	PACKET3_COPY_DATA__SRC_CACHE_POLICY(x) ((((unsigned)(x)) & 0x3) << 13)
+#define 	PACKET3_COPY_DATA__SRC_TEMPORAL(x) ((((unsigned)(x)) & 0x3) << 13)
+#define 	PACKET3_COPY_DATA__COUNT_SEL(x) ((((unsigned)(x)) & 0x1) << 16)
+#define 	PACKET3_COPY_DATA__WR_CONFIRM(x) ((((unsigned)(x)) & 0x1) << 20)
+#define 	PACKET3_COPY_DATA__DST_CACHE_POLICY(x) ((((unsigned)(x)) & 0x3) << 25)
+#define 	PACKET3_COPY_DATA__PQ_EXE_STATUS(x) ((((unsigned)(x)) & 0x1) << 29)
+#define 	PACKET3_COPY_DATA__SRC_REG_OFFSET(x) ((((unsigned)(x)) & 0x3FFFF) << 0)
+#define 	PACKET3_COPY_DATA__SRC_32B_ADDR_LO(x) ((((unsigned)(x)) & 0x3FFFFFFF) << 2)
+#define 	PACKET3_COPY_DATA__SRC_64B_ADDR_LO(x) ((((unsigned)(x)) & 0x1FFFFFFF) << 3)
+#define 	PACKET3_COPY_DATA__SRC_GDS_ADDR_LO(x) ((((unsigned)(x)) & 0xFFFF) << 0)
+#define 	PACKET3_COPY_DATA__IMM_DATA(x) ((unsigned)(x))
+#define 	PACKET3_COPY_DATA__SRC_MEMTC_ADDR_HI(x) ((unsigned)(x))
+#define 	PACKET3_COPY_DATA__SRC_IMM_DATA(x) ((unsigned)(x))
+#define 	PACKET3_COPY_DATA__DST_REG_OFFSET(x) ((((unsigned)(x)) & 0x3FFFF) << 0)
+#define 	PACKET3_COPY_DATA__DST_32B_ADDR_LO(x) ((((unsigned)(x)) & 0x3FFFFFFF) << 2)
+#define 	PACKET3_COPY_DATA__DST_64B_ADDR_LO(x) ((((unsigned)(x)) & 0x1FFFFFFF) << 3)
+#define 	PACKET3_COPY_DATA__DST_GDS_ADDR_LO(x) ((((unsigned)(x)) & 0xFFFF) << 0)
+#define 	PACKET3_COPY_DATA__DST_ADDR_HI(x) ((unsigned)(x))
+#define 	PACKET3_COPY_DATA__MODE(x) ((((unsigned)(x)) & 0x1) << 21)
+#define 	PACKET3_COPY_DATA__AID_ID(x) ((((unsigned)(x)) & 0x3) << 23)
+#define 	PACKET3_COPY_DATA__DST_TEMPORAL(x) ((((unsigned)(x)) & 0x3) << 25)
+#define 	PACKET3_COPY_DATA__SRC_REG_OFFSET_LO(x) ((unsigned)(x))
+#define 	PACKET3_COPY_DATA__SRC_REG_OFFSET_HI(x) ((((unsigned)(x)) & 0xFF) << 0)
+#define 	PACKET3_COPY_DATA__DST_REG_OFFSET_LO(x) ((unsigned)(x))
+#define 	PACKET3_COPY_DATA__DST_REG_OFFSET_HI(x) ((((unsigned)(x)) & 0xFF) << 0)
+#define 	PACKET3_COPY_DATA__SRC_SEL__MEM_MAPPED_REGISTER 0
+#define 	PACKET3_COPY_DATA__SRC_SEL__TC_L2_OBSOLETE 1
+#define 	PACKET3_COPY_DATA__SRC_SEL__TC_L2 2
+#define 	PACKET3_COPY_DATA__SRC_SEL__GDS 3
+#define 	PACKET3_COPY_DATA__SRC_SEL__PERFCOUNTERS 4
+#define 	PACKET3_COPY_DATA__SRC_SEL__IMMEDIATE_DATA 5
+#define 	PACKET3_COPY_DATA__SRC_SEL__ATOMIC_RETURN_DATA 6
+#define 	PACKET3_COPY_DATA__SRC_SEL__GDS_ATOMIC_RETURN_DATA0 7
+#define 	PACKET3_COPY_DATA__SRC_SEL__GDS_ATOMIC_RETURN_DATA1 8
+#define 	PACKET3_COPY_DATA__SRC_SEL__GPU_CLOCK_COUNT 9
+#define 	PACKET3_COPY_DATA__SRC_SEL__SYSTEM_CLOCK_COUNT 10
+#define 	PACKET3_COPY_DATA__DST_SEL__MEM_MAPPED_REGISTER 0
+#define 	PACKET3_COPY_DATA__DST_SEL__TC_L2 2
+#define 	PACKET3_COPY_DATA__DST_SEL__GDS 3
+#define 	PACKET3_COPY_DATA__DST_SEL__PERFCOUNTERS 4
+#define 	PACKET3_COPY_DATA__DST_SEL__TC_L2_OBSOLETE 5
+#define 	PACKET3_COPY_DATA__DST_SEL__MEM_MAPPED_REG_DC 6
+#define 	PACKET3_COPY_DATA__SRC_TEMPORAL__RT 0
+#define 	PACKET3_COPY_DATA__SRC_TEMPORAL__NT 1
+#define 	PACKET3_COPY_DATA__SRC_TEMPORAL__HT 2
+#define 	PACKET3_COPY_DATA__SRC_TEMPORAL__LU 3
+#define 	PACKET3_COPY_DATA__SRC_CACHE_POLICY__LRU 0
+#define 	PACKET3_COPY_DATA__SRC_CACHE_POLICY__STREAM 1
+#define 	PACKET3_COPY_DATA__SRC_CACHE_POLICY__NOA 2
+#define 	PACKET3_COPY_DATA__SRC_CACHE_POLICY__BYPASS 3
+#define 	PACKET3_COPY_DATA__COUNT_SEL__32_BITS_OF_DATA 0
+#define 	PACKET3_COPY_DATA__COUNT_SEL__64_BITS_OF_DATA 1
+#define 	PACKET3_COPY_DATA__WR_CONFIRM__DO_NOT_WAIT_FOR_CONFIRMATION 0
+#define 	PACKET3_COPY_DATA__WR_CONFIRM__WAIT_FOR_CONFIRMATION 1
+#define 	PACKET3_COPY_DATA__MODE__PF_VF_DISABLED 0
+#define 	PACKET3_COPY_DATA__MODE__PF_VF_ENABLED 1
+#define 	PACKET3_COPY_DATA__DST_TEMPORAL__RT 0
+#define 	PACKET3_COPY_DATA__DST_TEMPORAL__NT 1
+#define 	PACKET3_COPY_DATA__DST_TEMPORAL__HT 2
+#define 	PACKET3_COPY_DATA__DST_TEMPORAL__LU 3
+#define 	PACKET3_COPY_DATA__DST_CACHE_POLICY__LRU 0
+#define 	PACKET3_COPY_DATA__DST_CACHE_POLICY__STREAM 1
+#define 	PACKET3_COPY_DATA__DST_CACHE_POLICY__NOA 2
+#define 	PACKET3_COPY_DATA__DST_CACHE_POLICY__BYPASS 3
+#define 	PACKET3_COPY_DATA__PQ_EXE_STATUS__DEFAULT 0
+#define 	PACKET3_COPY_DATA__PQ_EXE_STATUS__PHASE_UPDATE 1
 #define	PACKET3_CP_DMA					0x41
 #define	PACKET3_PFP_SYNC_ME				0x42
 #define	PACKET3_SURFACE_SYNC				0x43
@@ -160,6 +332,23 @@
 		 * 3 - SAMPLE_STREAMOUTSTAT*
 		 * 4 - *S_PARTIAL_FLUSH
 		 */
+#define		PACKET3_EVENT_WRITE__EVENT_TYPE(x) ((((unsigned)(x)) & 0x3F) << 0)
+#define		PACKET3_EVENT_WRITE__EVENT_INDEX(x) ((((unsigned)(x)) & 0xF) << 8)
+#define		PACKET3_EVENT_WRITE__SAMP_PLST_CNTR_MODE(x) ((((unsigned)(x)) & 0x3) << 29)
+#define		PACKET3_EVENT_WRITE__OFFLOAD_ENABLE(x) ((((unsigned)(x)) & 0x1) << 0)
+#define 	PACKET3_EVENT_WRITE__ADDRESS_LO(x) ((((unsigned)(x)) & 0x1FFFFFFF) << 3)
+#define 	PACKET3_EVENT_WRITE__ADDRESS_HI(x) ((unsigned)(x))
+#define 	PACKET3_EVENT_WRITE__EVENT_INDEX__OTHER 0
+#define 	PACKET3_EVENT_WRITE__EVENT_INDEX__SAMPLE_PIPELINESTAT 2
+#define 	PACKET3_EVENT_WRITE__EVENT_INDEX__CS_PARTIAL_FLUSH 4
+#define 	PACKET3_EVENT_WRITE__EVENT_INDEX__SAMPLE_STREAMOUTSTATS 8
+#define 	PACKET3_EVENT_WRITE__EVENT_INDEX__SAMPLE_STREAMOUTSTATS1 9
+#define 	PACKET3_EVENT_WRITE__EVENT_INDEX__SAMPLE_STREAMOUTSTATS2 10
+#define 	PACKET3_EVENT_WRITE__EVENT_INDEX__SAMPLE_STREAMOUTSTATS3 11
+#define 	PACKET3_EVENT_WRITE__SAMP_PLST_CNTR_MODE__LEGACY_MODE 0
+#define 	PACKET3_EVENT_WRITE__SAMP_PLST_CNTR_MODE__MIXED_MODE1 1
+#define 	PACKET3_EVENT_WRITE__SAMP_PLST_CNTR_MODE__NEW_MODE 2
+#define 	PACKET3_EVENT_WRITE__SAMP_PLST_CNTR_MODE__MIXED_MODE3 3
 #define	PACKET3_EVENT_WRITE_EOP				0x47
 #define	PACKET3_EVENT_WRITE_EOS				0x48
 #define	PACKET3_RELEASE_MEM				0x49
@@ -304,6 +493,12 @@
 		 * 2: REVERSE
 		 */
 #define 	PACKET3_ACQUIRE_MEM_GCR_RANGE_IS_PA  (1 << 18)
+#define 	PACKET3_ACQUIRE_MEM__COHER_SIZE(x) ((unsigned)(x))
+#define 	PACKET3_ACQUIRE_MEM__COHER_SIZE_HI(x) ((((unsigned)(x)) & 0xFF) << 0)
+#define 	PACKET3_ACQUIRE_MEM__COHER_BASE_LO(x) ((unsigned)(x))
+#define 	PACKET3_ACQUIRE_MEM__COHER_BASE_HI(x) ((((unsigned)(x)) & 0xFFFFFF) << 0)
+#define 	PACKET3_ACQUIRE_MEM__POLL_INTERVAL(x) ((((unsigned)(x)) & 0xFFFF) << 0)
+#define 	PACKET3_ACQUIRE_MEM__GCR_CNTL(x) ((((unsigned)(x)) & 0x7FFFF) << 0)
 #define	PACKET3_REWIND					0x59
 #define	PACKET3_INTERRUPT				0x5A
 #define	PACKET3_GEN_PDEPTE				0x5B
@@ -330,11 +525,17 @@
 #define	PACKET3_SET_SH_REG				0x76
 #define		PACKET3_SET_SH_REG_START			0x00002c00
 #define		PACKET3_SET_SH_REG_END				0x00003000
+#define 	PACKET3_SET_SH_REG__REG_OFFSET(x) ((((unsigned)(x)) & 0xFFFF) << 0)
+#define 	PACKET3_SET_SH_REG__VMID_SHIFT(x) ((((unsigned)(x)) & 0x1F) << 23)
+#define 	PACKET3_SET_SH_REG__INDEX(x) ((((unsigned)(x)) & 0xF) << 28)
+#define 	PACKET3_SET_SH_REG__INDEX__DEFAULT 0
+#define 	PACKET3_SET_SH_REG__INDEX__INSERT_VMID 1
 #define	PACKET3_SET_SH_REG_OFFSET			0x77
 #define	PACKET3_SET_QUEUE_REG				0x78
 #define	PACKET3_SET_UCONFIG_REG				0x79
 #define		PACKET3_SET_UCONFIG_REG_START			0x0000c000
 #define		PACKET3_SET_UCONFIG_REG_END			0x0000c400
+#define 	PACKET3_SET_UCONFIG_REG__REG_OFFSET(x) ((((unsigned)(x)) & 0xFFFF) << 0)
 #define	PACKET3_SET_UCONFIG_REG_INDEX			0x7A
 #define	PACKET3_FORWARD_HEADER				0x7C
 #define	PACKET3_SCRATCH_RAM_WRITE			0x7D
@@ -369,6 +570,7 @@
 #              define PACKET3_INVALIDATE_TLBS_DST_SEL(x)     ((x) << 0)
 #              define PACKET3_INVALIDATE_TLBS_ALL_HUB(x)     ((x) << 4)
 #              define PACKET3_INVALIDATE_TLBS_PASID(x)       ((x) << 5)
+#              define PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(x)  ((x) << 29)
 #define	PACKET3_AQL_PACKET				0x99
 #define	PACKET3_DMA_DATA_FILL_MULTI			0x9A
 #define	PACKET3_SET_SH_REG_INDEX			0x9B
@@ -463,5 +665,14 @@
 #define	PACKET3_RUN_LIST				0xA5
 #define	PACKET3_MAP_PROCESS_VM				0xA6
 
+#define PACKET3_RUN_CLEANER_SHADER                      0xD2
+/* 1. header
+ * 2. RESERVED [31:0]
+ */
+
+/* GFX11 */
+#define	PACKET3_SET_Q_PREEMPTION_MODE			0xF0
+#              define PACKET3_SET_Q_PREEMPTION_MODE_IB_VMID(x)  ((x) << 0)
+#              define PACKET3_SET_Q_PREEMPTION_MODE_INIT_SHADOW_MEM    (1 << 0)
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h
index 22c775f39119..73f87131a7e9 100644
--- a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h
+++ b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h
@@ -102,6 +102,13 @@ enum psp_gfx_cmd_id
     GFX_CMD_ID_LOAD_TOC           = 0x00000020,   /* Load TOC and obtain TMR size */
     GFX_CMD_ID_AUTOLOAD_RLC       = 0x00000021,   /* Indicates all graphics fw loaded, start RLC autoload */
     GFX_CMD_ID_BOOT_CFG           = 0x00000022,   /* Boot Config */
+    GFX_CMD_ID_SRIOV_SPATIAL_PART = 0x00000027,   /* Configure spatial partitioning mode */
+    /*IDs of performance monitoring/profiling*/
+    GFX_CMD_ID_CONFIG_SQ_PERFMON  = 0x00000046,   /* Config CGTT_SQ_CLK_CTRL */
+    /* Dynamic memory partitioninig (NPS mode change)*/
+    GFX_CMD_ID_FB_NPS_MODE        = 0x00000048,  /* Configure memory partitioning mode */
+    GFX_CMD_ID_FB_FW_RESERV_ADDR  = 0x00000050,  /* Query FW reservation addr */
+    GFX_CMD_ID_FB_FW_RESERV_EXT_ADDR = 0x00000051,  /* Query FW reservation extended addr */
 };
 
 /* PSP boot config sub-commands */
@@ -292,6 +299,11 @@ enum psp_gfx_fw_type {
 	GFX_FW_TYPE_RS64_MEC_P1_STACK               = 95,   /* RS64 MEC stack P1        SOC21   */
 	GFX_FW_TYPE_RS64_MEC_P2_STACK               = 96,   /* RS64 MEC stack P2        SOC21   */
 	GFX_FW_TYPE_RS64_MEC_P3_STACK               = 97,   /* RS64 MEC stack P3        SOC21   */
+	GFX_FW_TYPE_VPEC_FW1                        = 100,  /* VPEC FW1 To Save         VPE     */
+	GFX_FW_TYPE_VPEC_FW2                        = 101,  /* VPEC FW2 To Save         VPE     */
+	GFX_FW_TYPE_VPE                             = 102,
+	GFX_FW_TYPE_JPEG_RAM                        = 128,  /**< JPEG Command buffer */
+	GFX_FW_TYPE_P2S_TABLE                       = 129,
 	GFX_FW_TYPE_MAX
 };
 
@@ -338,6 +350,27 @@ struct psp_gfx_cmd_boot_cfg
     uint32_t                        boot_config_valid;    /* dynamic boot configuration valid bits bitmask */
 };
 
+struct psp_gfx_cmd_sriov_spatial_part {
+	uint32_t mode;
+	uint32_t override_ips;
+	uint32_t override_xcds_avail;
+	uint32_t override_this_aid;
+};
+
+/*Structure for sq performance monitoring/profiling enable/disable*/
+struct psp_gfx_cmd_config_sq_perfmon {
+	uint32_t        gfx_xcp_mask;
+	uint8_t         core_override;
+	uint8_t         reg_override;
+	uint8_t         perfmon_override;
+	uint8_t         reserved[5];
+};
+
+struct psp_gfx_cmd_fb_memory_part {
+	uint32_t mode; /* requested NPS mode */
+	uint32_t resvd;
+};
+
 /* All GFX ring buffer commands. */
 union psp_gfx_commands
 {
@@ -351,6 +384,9 @@ union psp_gfx_commands
     struct psp_gfx_cmd_setup_tmr        cmd_setup_vmr;
     struct psp_gfx_cmd_load_toc         cmd_load_toc;
     struct psp_gfx_cmd_boot_cfg         boot_cfg;
+    struct psp_gfx_cmd_sriov_spatial_part cmd_spatial_part;
+    struct psp_gfx_cmd_config_sq_perfmon config_sq_perfmon;
+    struct psp_gfx_cmd_fb_memory_part cmd_memory_part;
 };
 
 struct psp_gfx_uresp_reserved
@@ -370,11 +406,19 @@ struct psp_gfx_uresp_bootcfg {
 	uint32_t boot_cfg;	/* boot config data */
 };
 
+/* Command-specific response for fw reserve info */
+struct psp_gfx_uresp_fw_reserve_info {
+    uint32_t reserve_base_address_hi;
+    uint32_t reserve_base_address_lo;
+    uint32_t reserve_size;
+};
+
 /* Union of command-specific responses for GPCOM ring. */
 union psp_gfx_uresp {
 	struct psp_gfx_uresp_reserved		reserved;
 	struct psp_gfx_uresp_bootcfg		boot_cfg;
 	struct psp_gfx_uresp_fwar_db_info	fwar_db_info;
+	struct psp_gfx_uresp_fw_reserve_info	fw_reserve_info;
 };
 
 /* Structure of GFX Response buffer.
@@ -450,8 +494,9 @@ struct psp_gfx_rb_frame
 #define PSP_ERR_UNKNOWN_COMMAND 0x00000100
 
 enum tee_error_code {
-    TEE_SUCCESS                         = 0x00000000,
-    TEE_ERROR_NOT_SUPPORTED             = 0xFFFF000A,
+	TEE_SUCCESS			= 0x00000000,
+	TEE_ERROR_CANCEL		= 0xFFFF0002,
+	TEE_ERROR_NOT_SUPPORTED		= 0xFFFF000A,
 };
 
 #endif /* _PSP_TEE_GFX_IF_H_ */
diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c
index e1b7fca09666..3584b8c18fd9 100644
--- a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c
@@ -57,7 +57,14 @@ static int psp_v10_0_init_microcode(struct psp_context *psp)
 	if (err)
 		return err;
 
-	return psp_init_ta_microcode(psp, ucode_prefix);
+	err = psp_init_ta_microcode(psp, ucode_prefix);
+	if ((amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 1, 0)) &&
+	    (adev->pdev->revision == 0xa1) &&
+	    (psp->securedisplay_context.context.bin_desc.fw_version >=
+	     0x27000008)) {
+		adev->psp.securedisplay_context.context.bin_desc.size_bytes = 0;
+	}
+	return err;
 }
 
 static int psp_v10_0_ring_create(struct psp_context *psp,
@@ -87,7 +94,7 @@ static int psp_v10_0_ring_create(struct psp_context *psp,
 
 	/* Wait for response flag (bit 31) in C2PMSG_64 */
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
-			   0x80000000, 0x8000FFFF, false);
+			   MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 
 	return ret;
 }
@@ -108,7 +115,7 @@ static int psp_v10_0_ring_stop(struct psp_context *psp,
 
 	/* Wait for response flag (bit 31) in C2PMSG_64 */
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
-			   0x80000000, 0x80000000, false);
+			   MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 
 	return ret;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c
index 8f84fe40abbb..a9be7a505026 100644
--- a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c
@@ -95,7 +95,7 @@ static int psp_v11_0_init_microcode(struct psp_context *psp)
 
 	amdgpu_ucode_ip_version_decode(adev, MP0_HWIP, ucode_prefix, sizeof(ucode_prefix));
 
-	switch (adev->ip_versions[MP0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 4):
 		err = psp_init_sos_microcode(psp, ucode_prefix);
@@ -129,6 +129,7 @@ static int psp_v11_0_init_microcode(struct psp_context *psp)
 		err = psp_init_ta_microcode(psp, ucode_prefix);
 		break;
 	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 2):
 		err = psp_init_asd_microcode(psp, ucode_prefix);
 		if (err)
 			return err;
@@ -141,21 +142,43 @@ static int psp_v11_0_init_microcode(struct psp_context *psp)
 	return err;
 }
 
-static int psp_v11_0_wait_for_bootloader(struct psp_context *psp)
+static int psp_v11_wait_for_tos_unload(struct psp_context *psp)
 {
 	struct amdgpu_device *adev = psp->adev;
+	uint32_t sol_reg1, sol_reg2;
+	int retry_loop;
+
+	/* Wait for the TOS to be unloaded */
+	for (retry_loop = 0; retry_loop < 20; retry_loop++) {
+		sol_reg1 = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81);
+		usleep_range(1000, 2000);
+		sol_reg2 = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81);
+		if (sol_reg1 == sol_reg2)
+			return 0;
+	}
+	dev_err(adev->dev, "TOS unload failed, C2PMSG_33: %x C2PMSG_81: %x",
+		RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_33),
+		RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81));
 
+	return -ETIME;
+}
+
+static int psp_v11_0_wait_for_bootloader(struct psp_context *psp)
+{
+	struct amdgpu_device *adev = psp->adev;
 	int ret;
 	int retry_loop;
 
-	for (retry_loop = 0; retry_loop < 10; retry_loop++) {
+	/* For a reset done at the end of S3, only wait for TOS to be unloaded */
+	if (adev->in_s3 && !(adev->flags & AMD_IS_APU) && amdgpu_in_reset(adev))
+		return psp_v11_wait_for_tos_unload(psp);
+
+	for (retry_loop = 0; retry_loop < 20; retry_loop++) {
 		/* Wait for bootloader to signify that is
 		    ready having bit 31 of C2PMSG_35 set to 1 */
-		ret = psp_wait_for(psp,
-				   SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35),
-				   0x80000000,
-				   0x80000000,
-				   false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35),
+			0x80000000, 0x8000FFFF, PSP_WAITREG_NOVERBOSE);
 
 		if (ret == 0)
 			return 0;
@@ -251,8 +274,8 @@ static int psp_v11_0_bootloader_load_sos(struct psp_context *psp)
 	/* there might be handshake issue with hardware which needs delay */
 	mdelay(20);
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_81),
-			   RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81),
-			   0, true);
+			   RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81), 0,
+			   PSP_WAITREG_CHANGED);
 
 	return ret;
 }
@@ -276,11 +299,13 @@ static int psp_v11_0_ring_stop(struct psp_context *psp,
 
 	/* Wait for response flag (bit 31) */
 	if (amdgpu_sriov_vf(adev))
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_101),
-				   0x80000000, 0x80000000, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_101),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 	else
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
-				   0x80000000, 0x80000000, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 
 	return ret;
 }
@@ -316,13 +341,15 @@ static int psp_v11_0_ring_create(struct psp_context *psp,
 		mdelay(20);
 
 		/* Wait for response flag (bit 31) in C2PMSG_101 */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_101),
-				   0x80000000, 0x8000FFFF, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_101),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 
 	} else {
 		/* Wait for sOS ready for ring creation */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
-				   0x80000000, 0x80000000, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
+			MBOX_TOS_READY_FLAG, MBOX_TOS_READY_MASK, 0);
 		if (ret) {
 			DRM_ERROR("Failed to wait for sOS ready for ring creation\n");
 			return ret;
@@ -346,8 +373,9 @@ static int psp_v11_0_ring_create(struct psp_context *psp,
 		mdelay(20);
 
 		/* Wait for response flag (bit 31) in C2PMSG_64 */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
-				   0x80000000, 0x8000FFFF, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 	}
 
 	return ret;
@@ -380,7 +408,8 @@ static int psp_v11_0_mode1_reset(struct psp_context *psp)
 
 	offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64);
 
-	ret = psp_wait_for(psp, offset, 0x80000000, 0x8000FFFF, false);
+	ret = psp_wait_for(psp, offset, MBOX_TOS_READY_FLAG,
+			   MBOX_TOS_READY_MASK, 0);
 
 	if (ret) {
 		DRM_INFO("psp is not working correctly before mode1 reset!\n");
@@ -392,17 +421,6 @@ static int psp_v11_0_mode1_reset(struct psp_context *psp)
 
 	msleep(500);
 
-	offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_33);
-
-	ret = psp_wait_for(psp, offset, 0x80000000, 0x80000000, false);
-
-	if (ret) {
-		DRM_INFO("psp mode 1 reset failed!\n");
-		return -EINVAL;
-	}
-
-	DRM_INFO("psp mode1 reset succeed \n");
-
 	return 0;
 }
 
@@ -420,8 +438,9 @@ static int psp_v11_0_memory_training_send_msg(struct psp_context *psp, int msg)
 
 	max_wait = MEM_TRAIN_SEND_MSG_TIMEOUT_US / adev->usec_timeout;
 	for (i = 0; i < max_wait; i++) {
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35),
-				   0x80000000, 0x80000000, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35),
+			0x80000000, 0x80000000, PSP_WAITREG_NOVERBOSE);
 		if (ret == 0)
 			break;
 	}
@@ -506,7 +525,7 @@ static int psp_v11_0_memory_training(struct psp_context *psp, uint32_t ops)
 		 * before training, and restore it after training to avoid
 		 * VRAM corruption.
 		 */
-		sz = GDDR6_MEM_TRAINING_ENCROACHED_SIZE;
+		sz = BIST_MEM_TRAINING_ENCROACHED_SIZE;
 
 		if (adev->gmc.visible_vram_size < sz || !adev->mman.aper_base_kaddr) {
 			DRM_ERROR("visible_vram_size %llx or aper_base_kaddr %p is not initialized.\n",
@@ -532,7 +551,7 @@ static int psp_v11_0_memory_training(struct psp_context *psp, uint32_t ops)
 			}
 
 			memcpy_toio(adev->mman.aper_base_kaddr, buf, sz);
-			adev->hdp.funcs->flush_hdp(adev, NULL);
+			amdgpu_device_flush_hdp(adev, NULL);
 			vfree(buf);
 			drm_dev_exit(idx);
 		} else {
@@ -600,7 +619,7 @@ static int psp_v11_0_load_usbc_pd_fw(struct psp_context *psp, uint64_t fw_pri_mc
 	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_36, (fw_pri_mc_addr >> 20));
 
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35),
-			     0x80000000, 0x80000000, false);
+			   0x80000000, 0x80000000, 0);
 	if (ret)
 		return ret;
 
@@ -637,7 +656,7 @@ static int psp_v11_0_read_usbc_pd_fw(struct psp_context *psp, uint32_t *fw_ver)
 	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_35, C2PMSG_CMD_GFX_USB_PD_FW_VER);
 
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35),
-				     0x80000000, 0x80000000, false);
+			   0x80000000, 0x80000000, 0);
 	if (!ret)
 		*fw_ver = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_36);
 
@@ -658,7 +677,8 @@ static const struct psp_funcs psp_v11_0_funcs = {
 	.ring_get_wptr = psp_v11_0_ring_get_wptr,
 	.ring_set_wptr = psp_v11_0_ring_set_wptr,
 	.load_usbc_pd_fw = psp_v11_0_load_usbc_pd_fw,
-	.read_usbc_pd_fw = psp_v11_0_read_usbc_pd_fw
+	.read_usbc_pd_fw = psp_v11_0_read_usbc_pd_fw,
+	.wait_for_bootloader = psp_v11_0_wait_for_bootloader
 };
 
 void psp_v11_0_set_psp_funcs(struct psp_context *psp)
diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v11_0_8.c b/drivers/gpu/drm/amd/amdgpu/psp_v11_0_8.c
index 5697760a819b..93787a90d598 100644
--- a/drivers/gpu/drm/amd/amdgpu/psp_v11_0_8.c
+++ b/drivers/gpu/drm/amd/amdgpu/psp_v11_0_8.c
@@ -41,8 +41,9 @@ static int psp_v11_0_8_ring_stop(struct psp_context *psp,
 		/* there might be handshake issue with hardware which needs delay */
 		mdelay(20);
 		/* Wait for response flag (bit 31) */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_101),
-				   0x80000000, 0x80000000, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_101),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 	} else {
 		/* Write the ring destroy command*/
 		WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64,
@@ -50,8 +51,9 @@ static int psp_v11_0_8_ring_stop(struct psp_context *psp,
 		/* there might be handshake issue with hardware which needs delay */
 		mdelay(20);
 		/* Wait for response flag (bit 31) */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
-				   0x80000000, 0x80000000, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 	}
 
 	return ret;
@@ -87,13 +89,15 @@ static int psp_v11_0_8_ring_create(struct psp_context *psp,
 		mdelay(20);
 
 		/* Wait for response flag (bit 31) in C2PMSG_101 */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_101),
-				   0x80000000, 0x8000FFFF, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_101),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 
 	} else {
 		/* Wait for sOS ready for ring creation */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
-				   0x80000000, 0x80000000, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
+			MBOX_TOS_READY_FLAG, MBOX_TOS_READY_MASK, 0);
 		if (ret) {
 			DRM_ERROR("Failed to wait for trust OS ready for ring creation\n");
 			return ret;
@@ -117,8 +121,9 @@ static int psp_v11_0_8_ring_create(struct psp_context *psp,
 		mdelay(20);
 
 		/* Wait for response flag (bit 31) in C2PMSG_64 */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
-				   0x80000000, 0x8000FFFF, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 	}
 
 	return ret;
diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v12_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v12_0.c
index fcd708eae75c..4c6450d62299 100644
--- a/drivers/gpu/drm/amd/amdgpu/psp_v12_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/psp_v12_0.c
@@ -34,9 +34,6 @@
 #include "sdma0/sdma0_4_0_offset.h"
 #include "nbio/nbio_7_4_offset.h"
 
-#include "oss/osssys_4_0_offset.h"
-#include "oss/osssys_4_0_sh_mask.h"
-
 MODULE_FIRMWARE("amdgpu/renoir_asd.bin");
 MODULE_FIRMWARE("amdgpu/renoir_ta.bin");
 MODULE_FIRMWARE("amdgpu/green_sardine_asd.bin");
@@ -85,7 +82,7 @@ static int psp_v12_0_bootloader_load_sysdrv(struct psp_context *psp)
 
 	/* Wait for bootloader to signify that is ready having bit 31 of C2PMSG_35 set to 1 */
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35),
-			   0x80000000, 0x80000000, false);
+			   0x80000000, 0x80000000, 0);
 	if (ret)
 		return ret;
 
@@ -99,11 +96,8 @@ static int psp_v12_0_bootloader_load_sysdrv(struct psp_context *psp)
 	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_35,
 	       psp_gfxdrv_command_reg);
 
-	/* there might be handshake issue with hardware which needs delay */
-	mdelay(20);
-
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35),
-			   0x80000000, 0x80000000, false);
+			   0x80000000, 0x80000000, 0);
 
 	return ret;
 }
@@ -124,7 +118,7 @@ static int psp_v12_0_bootloader_load_sos(struct psp_context *psp)
 
 	/* Wait for bootloader to signify that is ready having bit 31 of C2PMSG_35 set to 1 */
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35),
-			   0x80000000, 0x80000000, false);
+			   0x80000000, 0x80000000, 0);
 	if (ret)
 		return ret;
 
@@ -138,46 +132,13 @@ static int psp_v12_0_bootloader_load_sos(struct psp_context *psp)
 	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_35,
 	       psp_gfxdrv_command_reg);
 
-	/* there might be handshake issue with hardware which needs delay */
-	mdelay(20);
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_81),
-			   RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81),
-			   0, true);
+			   RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81), 0,
+			   PSP_WAITREG_CHANGED);
 
 	return ret;
 }
 
-static void psp_v12_0_reroute_ih(struct psp_context *psp)
-{
-	struct amdgpu_device *adev = psp->adev;
-	uint32_t tmp;
-
-	/* Change IH ring for VMC */
-	tmp = REG_SET_FIELD(0, IH_CLIENT_CFG_DATA, CREDIT_RETURN_ADDR, 0x1244b);
-	tmp = REG_SET_FIELD(tmp, IH_CLIENT_CFG_DATA, CLIENT_TYPE, 1);
-	tmp = REG_SET_FIELD(tmp, IH_CLIENT_CFG_DATA, RING_ID, 1);
-
-	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_69, 3);
-	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_70, tmp);
-	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, GFX_CTRL_CMD_ID_GBR_IH_SET);
-
-	mdelay(20);
-	psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
-		     0x80000000, 0x8000FFFF, false);
-
-	/* Change IH ring for UMC */
-	tmp = REG_SET_FIELD(0, IH_CLIENT_CFG_DATA, CREDIT_RETURN_ADDR, 0x1216b);
-	tmp = REG_SET_FIELD(tmp, IH_CLIENT_CFG_DATA, RING_ID, 1);
-
-	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_69, 4);
-	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_70, tmp);
-	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, GFX_CTRL_CMD_ID_GBR_IH_SET);
-
-	mdelay(20);
-	psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
-		     0x80000000, 0x8000FFFF, false);
-}
-
 static int psp_v12_0_ring_create(struct psp_context *psp,
 				enum psp_ring_type ring_type)
 {
@@ -186,49 +147,23 @@ static int psp_v12_0_ring_create(struct psp_context *psp,
 	struct psp_ring *ring = &psp->km_ring;
 	struct amdgpu_device *adev = psp->adev;
 
-	psp_v12_0_reroute_ih(psp);
-
-	if (amdgpu_sriov_vf(psp->adev)) {
-		/* Write low address of the ring to C2PMSG_102 */
-		psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr);
-		WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_102, psp_ring_reg);
-		/* Write high address of the ring to C2PMSG_103 */
-		psp_ring_reg = upper_32_bits(ring->ring_mem_mc_addr);
-		WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_103, psp_ring_reg);
-
-		/* Write the ring initialization command to C2PMSG_101 */
-		WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_101,
-					     GFX_CTRL_CMD_ID_INIT_GPCOM_RING);
-
-		/* there might be handshake issue with hardware which needs delay */
-		mdelay(20);
-
-		/* Wait for response flag (bit 31) in C2PMSG_101 */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_101),
-				   0x80000000, 0x8000FFFF, false);
-
-	} else {
-		/* Write low address of the ring to C2PMSG_69 */
-		psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr);
-		WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_69, psp_ring_reg);
-		/* Write high address of the ring to C2PMSG_70 */
-		psp_ring_reg = upper_32_bits(ring->ring_mem_mc_addr);
-		WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_70, psp_ring_reg);
-		/* Write size of ring to C2PMSG_71 */
-		psp_ring_reg = ring->ring_size;
-		WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_71, psp_ring_reg);
-		/* Write the ring initialization command to C2PMSG_64 */
-		psp_ring_reg = ring_type;
-		psp_ring_reg = psp_ring_reg << 16;
-		WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, psp_ring_reg);
-
-		/* there might be handshake issue with hardware which needs delay */
-		mdelay(20);
-
-		/* Wait for response flag (bit 31) in C2PMSG_64 */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
-				   0x80000000, 0x8000FFFF, false);
-	}
+	/* Write low address of the ring to C2PMSG_69 */
+	psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr);
+	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_69, psp_ring_reg);
+	/* Write high address of the ring to C2PMSG_70 */
+	psp_ring_reg = upper_32_bits(ring->ring_mem_mc_addr);
+	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_70, psp_ring_reg);
+	/* Write size of ring to C2PMSG_71 */
+	psp_ring_reg = ring->ring_size;
+	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_71, psp_ring_reg);
+	/* Write the ring initialization command to C2PMSG_64 */
+	psp_ring_reg = ring_type;
+	psp_ring_reg = psp_ring_reg << 16;
+	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, psp_ring_reg);
+
+	/* Wait for response flag (bit 31) in C2PMSG_64 */
+	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
+			   MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 
 	return ret;
 }
@@ -247,16 +182,15 @@ static int psp_v12_0_ring_stop(struct psp_context *psp,
 		WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64,
 				     GFX_CTRL_CMD_ID_DESTROY_RINGS);
 
-	/* there might be handshake issue with hardware which needs delay */
-	mdelay(20);
-
 	/* Wait for response flag (bit 31) */
 	if (amdgpu_sriov_vf(adev))
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_101),
-				   0x80000000, 0x80000000, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_101),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 	else
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
-				   0x80000000, 0x80000000, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 
 	return ret;
 }
@@ -287,7 +221,8 @@ static int psp_v12_0_mode1_reset(struct psp_context *psp)
 
 	offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64);
 
-	ret = psp_wait_for(psp, offset, 0x80000000, 0x8000FFFF, false);
+	ret = psp_wait_for(psp, offset, MBOX_TOS_READY_FLAG,
+			   MBOX_TOS_READY_MASK, 0);
 
 	if (ret) {
 		DRM_INFO("psp is not working correctly before mode1 reset!\n");
@@ -301,7 +236,8 @@ static int psp_v12_0_mode1_reset(struct psp_context *psp)
 
 	offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_33);
 
-	ret = psp_wait_for(psp, offset, 0x80000000, 0x80000000, false);
+	ret = psp_wait_for(psp, offset, MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK,
+			   0);
 
 	if (ret) {
 		DRM_INFO("psp mode 1 reset failed!\n");
diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c
index caee76ab7110..af4a7d7c4abd 100644
--- a/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c
@@ -27,6 +27,7 @@
 #include "amdgpu_ucode.h"
 #include "soc15_common.h"
 #include "psp_v13_0.h"
+#include "amdgpu_ras.h"
 
 #include "mp/mp_13_0_2_offset.h"
 #include "mp/mp_13_0_2_sh_mask.h"
@@ -41,7 +42,9 @@ MODULE_FIRMWARE("amdgpu/psp_13_0_5_ta.bin");
 MODULE_FIRMWARE("amdgpu/psp_13_0_8_toc.bin");
 MODULE_FIRMWARE("amdgpu/psp_13_0_8_ta.bin");
 MODULE_FIRMWARE("amdgpu/psp_13_0_0_sos.bin");
+MODULE_FIRMWARE("amdgpu/psp_13_0_0_sos_kicker.bin");
 MODULE_FIRMWARE("amdgpu/psp_13_0_0_ta.bin");
+MODULE_FIRMWARE("amdgpu/psp_13_0_0_ta_kicker.bin");
 MODULE_FIRMWARE("amdgpu/psp_13_0_7_sos.bin");
 MODULE_FIRMWARE("amdgpu/psp_13_0_7_ta.bin");
 MODULE_FIRMWARE("amdgpu/psp_13_0_10_sos.bin");
@@ -49,6 +52,17 @@ MODULE_FIRMWARE("amdgpu/psp_13_0_10_ta.bin");
 MODULE_FIRMWARE("amdgpu/psp_13_0_11_toc.bin");
 MODULE_FIRMWARE("amdgpu/psp_13_0_11_ta.bin");
 MODULE_FIRMWARE("amdgpu/psp_13_0_6_sos.bin");
+MODULE_FIRMWARE("amdgpu/psp_13_0_6_ta.bin");
+MODULE_FIRMWARE("amdgpu/psp_13_0_12_sos.bin");
+MODULE_FIRMWARE("amdgpu/psp_13_0_12_ta.bin");
+MODULE_FIRMWARE("amdgpu/psp_13_0_14_sos.bin");
+MODULE_FIRMWARE("amdgpu/psp_13_0_14_ta.bin");
+MODULE_FIRMWARE("amdgpu/psp_14_0_0_toc.bin");
+MODULE_FIRMWARE("amdgpu/psp_14_0_0_ta.bin");
+MODULE_FIRMWARE("amdgpu/psp_14_0_1_toc.bin");
+MODULE_FIRMWARE("amdgpu/psp_14_0_1_ta.bin");
+MODULE_FIRMWARE("amdgpu/psp_14_0_4_toc.bin");
+MODULE_FIRMWARE("amdgpu/psp_14_0_4_ta.bin");
 
 /* For large FW files the time to complete can be very long */
 #define USBC_PD_POLLING_LIMIT_S 240
@@ -56,18 +70,16 @@ MODULE_FIRMWARE("amdgpu/psp_13_0_6_sos.bin");
 /* Read USB-PD from LFB */
 #define GFX_CMD_USB_PD_USE_LFB 0x480
 
-/* VBIOS gfl defines */
-#define MBOX_READY_MASK 0x80000000
-#define MBOX_STATUS_MASK 0x0000FFFF
-#define MBOX_COMMAND_MASK 0x00FF0000
-#define MBOX_READY_FLAG 0x80000000
-#define C2PMSG_CMD_SPI_UPDATE_ROM_IMAGE_ADDR_LO 0x2
-#define C2PMSG_CMD_SPI_UPDATE_ROM_IMAGE_ADDR_HI 0x3
-#define C2PMSG_CMD_SPI_UPDATE_FLASH_IMAGE 0x4
+/* Retry times for vmbx ready wait */
+#define PSP_VMBX_POLLING_LIMIT 3000
 
 /* memory training timeout define */
 #define MEM_TRAIN_SEND_MSG_TIMEOUT_US	3000000
 
+#define regMP1_PUB_SCRATCH0	0x3b10090
+
+#define PSP13_BL_STATUS_SIZE 100
+
 static int psp_v13_0_init_microcode(struct psp_context *psp)
 {
 	struct amdgpu_device *adev = psp->adev;
@@ -76,7 +88,7 @@ static int psp_v13_0_init_microcode(struct psp_context *psp)
 
 	amdgpu_ucode_ip_version_decode(adev, MP0_HWIP, ucode_prefix, sizeof(ucode_prefix));
 
-	switch (adev->ip_versions[MP0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 	case IP_VERSION(13, 0, 2):
 		err = psp_init_sos_microcode(psp, ucode_prefix);
 		if (err)
@@ -93,6 +105,9 @@ static int psp_v13_0_init_microcode(struct psp_context *psp)
 	case IP_VERSION(13, 0, 5):
 	case IP_VERSION(13, 0, 8):
 	case IP_VERSION(13, 0, 11):
+	case IP_VERSION(14, 0, 0):
+	case IP_VERSION(14, 0, 1):
+	case IP_VERSION(14, 0, 4):
 		err = psp_init_toc_microcode(psp, ucode_prefix);
 		if (err)
 			return err;
@@ -104,6 +119,8 @@ static int psp_v13_0_init_microcode(struct psp_context *psp)
 	case IP_VERSION(13, 0, 6):
 	case IP_VERSION(13, 0, 7):
 	case IP_VERSION(13, 0, 10):
+	case IP_VERSION(13, 0, 12):
+	case IP_VERSION(13, 0, 14):
 		err = psp_init_sos_microcode(psp, ucode_prefix);
 		if (err)
 			return err;
@@ -129,29 +146,107 @@ static bool psp_v13_0_is_sos_alive(struct psp_context *psp)
 	return sol_reg != 0x0;
 }
 
-static int psp_v13_0_wait_for_bootloader(struct psp_context *psp)
+static void psp_v13_0_bootloader_print_status(struct psp_context *psp,
+					      const char *msg)
 {
 	struct amdgpu_device *adev = psp->adev;
+	u32 bl_status_reg;
+	char bl_status_msg[PSP13_BL_STATUS_SIZE];
+	int i, at;
+
+	if (amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 6) ||
+	    amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 12) ||
+	    amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 14)) {
+		at = 0;
+		for_each_inst(i, adev->aid_mask) {
+			bl_status_reg =
+				(SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_92)
+				 << 2) +
+				adev->asic_funcs->encode_ext_smn_addressing(i);
+			at += snprintf(bl_status_msg + at,
+				       PSP13_BL_STATUS_SIZE - at,
+				       " status(%02i): 0x%08x", i,
+				       RREG32_PCIE_EXT(bl_status_reg));
+		}
+		dev_info(adev->dev, "%s - %s", msg, bl_status_msg);
+	}
+}
 
-	int ret;
-	int retry_loop;
+static int psp_v13_0_wait_for_vmbx_ready(struct psp_context *psp)
+{
+	struct amdgpu_device *adev = psp->adev;
+	int retry_loop, ret;
 
-	for (retry_loop = 0; retry_loop < 10; retry_loop++) {
+	for (retry_loop = 0; retry_loop < PSP_VMBX_POLLING_LIMIT; retry_loop++) {
 		/* Wait for bootloader to signify that is
-		    ready having bit 31 of C2PMSG_35 set to 1 */
-		ret = psp_wait_for(psp,
-				   SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_35),
-				   0x80000000,
-				   0x80000000,
-				   false);
+		   ready having bit 31 of C2PMSG_33 set to 1 */
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_33),
+			0x80000000, 0xffffffff, PSP_WAITREG_NOVERBOSE);
+
+		if (ret == 0)
+			break;
+	}
+
+	if (ret)
+		dev_warn(adev->dev, "Bootloader wait timed out");
+
+	return ret;
+}
+
+static int psp_v13_0_wait_for_bootloader(struct psp_context *psp)
+{
+	struct amdgpu_device *adev = psp->adev;
+	int retry_loop, retry_cnt, ret;
+
+	retry_cnt =
+		((amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 6) ||
+		  amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 12) ||
+		  amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 14))) ?
+			PSP_VMBX_POLLING_LIMIT :
+			10;
+	/* Wait for bootloader to signify that it is ready having bit 31 of
+	 * C2PMSG_35 set to 1. All other bits are expected to be cleared.
+	 * If there is an error in processing command, bits[7:0] will be set.
+	 * This is applicable for PSP v13.0.6 and newer.
+	 */
+	for (retry_loop = 0; retry_loop < retry_cnt; retry_loop++) {
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_35),
+			0x80000000, 0xffffffff, PSP_WAITREG_NOVERBOSE);
 
 		if (ret == 0)
 			return 0;
+		if (retry_loop && !(retry_loop % 10))
+			psp_v13_0_bootloader_print_status(
+				psp, "Waiting for bootloader completion");
 	}
 
 	return ret;
 }
 
+static int psp_v13_0_wait_for_bootloader_steady_state(struct psp_context *psp)
+{
+	struct amdgpu_device *adev = psp->adev;
+	int ret;
+
+	if (amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 6) ||
+	    amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 12) ||
+	    amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 14)) {
+		ret = psp_v13_0_wait_for_vmbx_ready(psp);
+		if (ret)
+			amdgpu_ras_query_boot_status(adev, 4);
+
+		ret = psp_v13_0_wait_for_bootloader(psp);
+		if (ret)
+			amdgpu_ras_query_boot_status(adev, 4);
+
+		return ret;
+	}
+
+	return 0;
+}
+
 static int psp_v13_0_bootloader_load_component(struct psp_context  	*psp,
 					       struct psp_bin_desc 	*bin_desc,
 					       enum psp_bootloader_cmd  bl_cmd)
@@ -222,6 +317,17 @@ static int psp_v13_0_bootloader_load_ras_drv(struct psp_context *psp)
 	return psp_v13_0_bootloader_load_component(psp, &psp->ras_drv, PSP_BL__LOAD_RASDRV);
 }
 
+static int psp_v13_0_bootloader_load_spdm_drv(struct psp_context *psp)
+{
+	return psp_v13_0_bootloader_load_component(psp, &psp->spdm_drv, PSP_BL__LOAD_SPDMDRV);
+}
+
+static inline void psp_v13_0_init_sos_version(struct psp_context *psp)
+{
+	struct amdgpu_device *adev = psp->adev;
+
+	psp->sos.fw_version = RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_58);
+}
 
 static int psp_v13_0_bootloader_load_sos(struct psp_context *psp)
 {
@@ -232,8 +338,10 @@ static int psp_v13_0_bootloader_load_sos(struct psp_context *psp)
 	/* Check sOS sign of life register to confirm sys driver and sOS
 	 * are already been loaded.
 	 */
-	if (psp_v13_0_is_sos_alive(psp))
+	if (psp_v13_0_is_sos_alive(psp)) {
+		psp_v13_0_init_sos_version(psp);
 		return 0;
+	}
 
 	ret = psp_v13_0_wait_for_bootloader(psp);
 	if (ret)
@@ -254,8 +362,11 @@ static int psp_v13_0_bootloader_load_sos(struct psp_context *psp)
 	/* there might be handshake issue with hardware which needs delay */
 	mdelay(20);
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_81),
-			   RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_81),
-			   0, true);
+			   RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_81), 0,
+			   PSP_WAITREG_CHANGED);
+
+	if (!ret)
+		psp_v13_0_init_sos_version(psp);
 
 	return ret;
 }
@@ -273,8 +384,9 @@ static int psp_v13_0_ring_stop(struct psp_context *psp,
 		/* there might be handshake issue with hardware which needs delay */
 		mdelay(20);
 		/* Wait for response flag (bit 31) */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_101),
-				   0x80000000, 0x80000000, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_101),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 	} else {
 		/* Write the ring destroy command*/
 		WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_64,
@@ -282,8 +394,9 @@ static int psp_v13_0_ring_stop(struct psp_context *psp,
 		/* there might be handshake issue with hardware which needs delay */
 		mdelay(20);
 		/* Wait for response flag (bit 31) */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_64),
-				   0x80000000, 0x80000000, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_64),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 	}
 
 	return ret;
@@ -319,13 +432,15 @@ static int psp_v13_0_ring_create(struct psp_context *psp,
 		mdelay(20);
 
 		/* Wait for response flag (bit 31) in C2PMSG_101 */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_101),
-				   0x80000000, 0x8000FFFF, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_101),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 
 	} else {
 		/* Wait for sOS ready for ring creation */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_64),
-				   0x80000000, 0x80000000, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_64),
+			MBOX_TOS_READY_FLAG, MBOX_TOS_READY_MASK, 0);
 		if (ret) {
 			DRM_ERROR("Failed to wait for trust OS ready for ring creation\n");
 			return ret;
@@ -349,8 +464,9 @@ static int psp_v13_0_ring_create(struct psp_context *psp,
 		mdelay(20);
 
 		/* Wait for response flag (bit 31) in C2PMSG_64 */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_64),
-				   0x80000000, 0x8000FFFF, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_64),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 	}
 
 	return ret;
@@ -413,8 +529,9 @@ static int psp_v13_0_memory_training_send_msg(struct psp_context *psp, int msg)
 
 	max_wait = MEM_TRAIN_SEND_MSG_TIMEOUT_US / adev->usec_timeout;
 	for (i = 0; i < max_wait; i++) {
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_35),
-				   0x80000000, 0x80000000, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_35),
+			0x80000000, 0x80000000, PSP_WAITREG_NOVERBOSE);
 		if (ret == 0)
 			break;
 	}
@@ -497,7 +614,7 @@ static int psp_v13_0_memory_training(struct psp_context *psp, uint32_t ops)
 		 * before training, and restore it after training to avoid
 		 * VRAM corruption.
 		 */
-		sz = GDDR6_MEM_TRAINING_ENCROACHED_SIZE;
+		sz = BIST_MEM_TRAINING_ENCROACHED_SIZE;
 
 		if (adev->gmc.visible_vram_size < sz || !adev->mman.aper_base_kaddr) {
 			dev_err(adev->dev, "visible_vram_size %llx or aper_base_kaddr %p is not initialized.\n",
@@ -523,7 +640,7 @@ static int psp_v13_0_memory_training(struct psp_context *psp, uint32_t ops)
 			}
 
 			memcpy_toio(adev->mman.aper_base_kaddr, buf, sz);
-			adev->hdp.funcs->flush_hdp(adev, NULL);
+			amdgpu_device_flush_hdp(adev, NULL);
 			vfree(buf);
 			drm_dev_exit(idx);
 		} else {
@@ -566,7 +683,7 @@ static int psp_v13_0_load_usbc_pd_fw(struct psp_context *psp, uint64_t fw_pri_mc
 	WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_36, (fw_pri_mc_addr >> 20));
 
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_35),
-			     0x80000000, 0x80000000, false);
+			   0x80000000, 0x80000000, 0);
 	if (ret)
 		return ret;
 
@@ -603,7 +720,7 @@ static int psp_v13_0_read_usbc_pd_fw(struct psp_context *psp, uint32_t *fw_ver)
 	WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_35, C2PMSG_CMD_GFX_USB_PD_FW_VER);
 
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_35),
-				     0x80000000, 0x80000000, false);
+			   0x80000000, 0x80000000, 0);
 	if (!ret)
 		*fw_ver = RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_36);
 
@@ -623,11 +740,14 @@ static int psp_v13_0_exec_spi_cmd(struct psp_context *psp, int cmd)
 	/* Ring the doorbell */
 	WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_73, 1);
 
-	if (cmd == C2PMSG_CMD_SPI_UPDATE_FLASH_IMAGE)
-		return 0;
-
-	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_115),
-				MBOX_READY_FLAG, MBOX_READY_MASK, false);
+	if (cmd == C2PMSG_CMD_SPI_UPDATE_FLASH_IMAGE ||
+	    cmd == C2PMSG_CMD_SPI_GET_FLASH_IMAGE)
+		ret = psp_wait_for_spirom_update(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_115),
+						 MBOX_READY_FLAG, MBOX_READY_MASK, PSP_SPIROM_UPDATE_TIMEOUT);
+	else
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_115),
+			MBOX_READY_FLAG, MBOX_READY_MASK, 0);
 	if (ret) {
 		dev_err(adev->dev, "SPI cmd %x timed out, ret = %d", cmd, ret);
 		return ret;
@@ -651,7 +771,7 @@ static int psp_v13_0_update_spirom(struct psp_context *psp,
 
 	/* Confirm PSP is ready to start */
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_115),
-			   MBOX_READY_FLAG, MBOX_READY_MASK, false);
+			   MBOX_READY_FLAG, MBOX_READY_MASK, 0);
 	if (ret) {
 		dev_err(adev->dev, "PSP Not ready to start processing, ret = %d", ret);
 		return ret;
@@ -678,6 +798,37 @@ static int psp_v13_0_update_spirom(struct psp_context *psp,
 	return 0;
 }
 
+static int psp_v13_0_dump_spirom(struct psp_context *psp,
+				 uint64_t fw_pri_mc_addr)
+{
+	struct amdgpu_device *adev = psp->adev;
+	int ret;
+
+	/* Confirm PSP is ready to start */
+	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_115),
+			   MBOX_READY_FLAG, MBOX_READY_MASK, 0);
+	if (ret) {
+		dev_err(adev->dev, "PSP Not ready to start processing, ret = %d", ret);
+		return ret;
+	}
+
+	WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_116, lower_32_bits(fw_pri_mc_addr));
+
+	ret = psp_v13_0_exec_spi_cmd(psp, C2PMSG_CMD_SPI_GET_ROM_IMAGE_ADDR_LO);
+	if (ret)
+		return ret;
+
+	WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_116, upper_32_bits(fw_pri_mc_addr));
+
+	ret = psp_v13_0_exec_spi_cmd(psp, C2PMSG_CMD_SPI_GET_ROM_IMAGE_ADDR_HI);
+	if (ret)
+		return ret;
+
+	ret = psp_v13_0_exec_spi_cmd(psp, C2PMSG_CMD_SPI_GET_FLASH_IMAGE);
+
+	return ret;
+}
+
 static int psp_v13_0_vbflash_status(struct psp_context *psp)
 {
 	struct amdgpu_device *adev = psp->adev;
@@ -685,8 +836,114 @@ static int psp_v13_0_vbflash_status(struct psp_context *psp)
 	return RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_115);
 }
 
+static int psp_v13_0_fatal_error_recovery_quirk(struct psp_context *psp)
+{
+	struct amdgpu_device *adev = psp->adev;
+
+	if (amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 10)) {
+		uint32_t  reg_data;
+		/* MP1 fatal error: trigger PSP dram read to unhalt PSP
+		 * during MP1 triggered sync flood.
+		 */
+		reg_data = RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_67);
+		WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_67, reg_data + 0x10);
+
+		/* delay 1000ms for the mode1 reset for fatal error
+		 * to be recovered back.
+		 */
+		msleep(1000);
+	}
+
+	return 0;
+}
+
+static bool psp_v13_0_get_ras_capability(struct psp_context *psp)
+{
+	struct amdgpu_device *adev = psp->adev;
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	u32 reg_data;
+
+	/* query ras cap should be done from host side */
+	if (amdgpu_sriov_vf(adev))
+		return false;
+
+	if (!con)
+		return false;
+
+	if ((amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 6) ||
+	     amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 12) ||
+	     amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 14)) &&
+	    (!(adev->flags & AMD_IS_APU))) {
+		reg_data = RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_127);
+		adev->ras_hw_enabled = (reg_data & GENMASK_ULL(23, 0));
+		con->poison_supported = ((reg_data & GENMASK_ULL(24, 24)) >> 24) ? true : false;
+		return true;
+	} else {
+		return false;
+	}
+}
+
+static bool psp_v13_0_is_aux_sos_load_required(struct psp_context *psp)
+{
+	struct amdgpu_device *adev = psp->adev;
+	u32 pmfw_ver;
+
+	if (amdgpu_ip_version(adev, MP0_HWIP, 0) != IP_VERSION(13, 0, 6))
+		return false;
+
+	/* load 4e version of sos if pmfw version less than 85.115.0 */
+	pmfw_ver = RREG32(regMP1_PUB_SCRATCH0 / 4);
+
+	return (pmfw_ver < 0x557300);
+}
+
+static bool psp_v13_0_is_reload_needed(struct psp_context *psp)
+{
+	uint32_t ucode_ver;
+
+	if (!psp_v13_0_is_sos_alive(psp))
+		return false;
+
+	/* Restrict reload support only to specific IP versions */
+	switch (amdgpu_ip_version(psp->adev, MP0_HWIP, 0)) {
+	case IP_VERSION(13, 0, 2):
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 14):
+		/* TOS version read from microcode header */
+		ucode_ver = psp->sos.fw_version;
+		/* Read TOS version from hardware */
+		psp_v13_0_init_sos_version(psp);
+		return (ucode_ver != psp->sos.fw_version);
+	default:
+		return false;
+	}
+
+	return false;
+}
+
+static int psp_v13_0_reg_program_no_ring(struct psp_context *psp, uint32_t val,
+					 enum psp_reg_prog_id id)
+{
+	struct amdgpu_device *adev = psp->adev;
+	int ret = -EOPNOTSUPP;
+
+	/* PSP will broadcast the value to all instances */
+	if (amdgpu_sriov_vf(adev)) {
+		WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_101, GFX_CTRL_CMD_ID_GBR_IH_SET);
+		WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_102, id);
+		WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_103, val);
+
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_101),
+			0x80000000, 0x80000000, 0);
+	}
+
+	return ret;
+}
+
 static const struct psp_funcs psp_v13_0_funcs = {
 	.init_microcode = psp_v13_0_init_microcode,
+	.wait_for_bootloader = psp_v13_0_wait_for_bootloader_steady_state,
 	.bootloader_load_kdb = psp_v13_0_bootloader_load_kdb,
 	.bootloader_load_spl = psp_v13_0_bootloader_load_spl,
 	.bootloader_load_sysdrv = psp_v13_0_bootloader_load_sysdrv,
@@ -694,6 +951,7 @@ static const struct psp_funcs psp_v13_0_funcs = {
 	.bootloader_load_intf_drv = psp_v13_0_bootloader_load_intf_drv,
 	.bootloader_load_dbg_drv = psp_v13_0_bootloader_load_dbg_drv,
 	.bootloader_load_ras_drv = psp_v13_0_bootloader_load_ras_drv,
+	.bootloader_load_spdm_drv = psp_v13_0_bootloader_load_spdm_drv,
 	.bootloader_load_sos = psp_v13_0_bootloader_load_sos,
 	.ring_create = psp_v13_0_ring_create,
 	.ring_stop = psp_v13_0_ring_stop,
@@ -704,7 +962,13 @@ static const struct psp_funcs psp_v13_0_funcs = {
 	.load_usbc_pd_fw = psp_v13_0_load_usbc_pd_fw,
 	.read_usbc_pd_fw = psp_v13_0_read_usbc_pd_fw,
 	.update_spirom = psp_v13_0_update_spirom,
-	.vbflash_stat = psp_v13_0_vbflash_status
+	.dump_spirom = psp_v13_0_dump_spirom,
+	.vbflash_stat = psp_v13_0_vbflash_status,
+	.fatal_error_recovery_quirk = psp_v13_0_fatal_error_recovery_quirk,
+	.get_ras_capability = psp_v13_0_get_ras_capability,
+	.is_aux_sos_load_required = psp_v13_0_is_aux_sos_load_required,
+	.is_reload_needed = psp_v13_0_is_reload_needed,
+	.reg_program_no_ring = psp_v13_0_reg_program_no_ring,
 };
 
 void psp_v13_0_set_psp_funcs(struct psp_context *psp)
diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v13_0.h b/drivers/gpu/drm/amd/amdgpu/psp_v13_0.h
index b2414a729ca1..de5677ce4330 100644
--- a/drivers/gpu/drm/amd/amdgpu/psp_v13_0.h
+++ b/drivers/gpu/drm/amd/amdgpu/psp_v13_0.h
@@ -25,6 +25,8 @@
 
 #include "amdgpu_psp.h"
 
+#define PSP_SPIROM_UPDATE_TIMEOUT   60000   /* 60s */
+
 void psp_v13_0_set_psp_funcs(struct psp_context *psp);
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v13_0_4.c b/drivers/gpu/drm/amd/amdgpu/psp_v13_0_4.c
index d5ba58eba3e2..5f39a2edcc95 100644
--- a/drivers/gpu/drm/amd/amdgpu/psp_v13_0_4.c
+++ b/drivers/gpu/drm/amd/amdgpu/psp_v13_0_4.c
@@ -40,7 +40,7 @@ static int psp_v13_0_4_init_microcode(struct psp_context *psp)
 
 	amdgpu_ucode_ip_version_decode(adev, MP0_HWIP, ucode_prefix, sizeof(ucode_prefix));
 
-	switch (adev->ip_versions[MP0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 	case IP_VERSION(13, 0, 4):
 		err = psp_init_toc_microcode(psp, ucode_prefix);
 		if (err)
@@ -76,11 +76,9 @@ static int psp_v13_0_4_wait_for_bootloader(struct psp_context *psp)
 	for (retry_loop = 0; retry_loop < 10; retry_loop++) {
 		/* Wait for bootloader to signify that is
 		    ready having bit 31 of C2PMSG_35 set to 1 */
-		ret = psp_wait_for(psp,
-				   SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_35),
-				   0x80000000,
-				   0x80000000,
-				   false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_35),
+			0x80000000, 0x80000000, PSP_WAITREG_NOVERBOSE);
 
 		if (ret == 0)
 			return 0;
@@ -185,8 +183,8 @@ static int psp_v13_0_4_bootloader_load_sos(struct psp_context *psp)
 	/* there might be handshake issue with hardware which needs delay */
 	mdelay(20);
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_81),
-			   RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_81),
-			   0, true);
+			   RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_81), 0,
+			   PSP_WAITREG_CHANGED);
 
 	return ret;
 }
@@ -204,8 +202,9 @@ static int psp_v13_0_4_ring_stop(struct psp_context *psp,
 		/* there might be handshake issue with hardware which needs delay */
 		mdelay(20);
 		/* Wait for response flag (bit 31) */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_101),
-				   0x80000000, 0x80000000, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_101),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 	} else {
 		/* Write the ring destroy command*/
 		WREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_64,
@@ -213,8 +212,9 @@ static int psp_v13_0_4_ring_stop(struct psp_context *psp,
 		/* there might be handshake issue with hardware which needs delay */
 		mdelay(20);
 		/* Wait for response flag (bit 31) */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_64),
-				   0x80000000, 0x80000000, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_64),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 	}
 
 	return ret;
@@ -250,13 +250,15 @@ static int psp_v13_0_4_ring_create(struct psp_context *psp,
 		mdelay(20);
 
 		/* Wait for response flag (bit 31) in C2PMSG_101 */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_101),
-				   0x80000000, 0x8000FFFF, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_101),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 
 	} else {
 		/* Wait for sOS ready for ring creation */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_64),
-				   0x80000000, 0x80000000, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_64),
+			MBOX_TOS_READY_FLAG, MBOX_TOS_READY_MASK, 0);
 		if (ret) {
 			DRM_ERROR("Failed to wait for trust OS ready for ring creation\n");
 			return ret;
@@ -280,8 +282,9 @@ static int psp_v13_0_4_ring_create(struct psp_context *psp,
 		mdelay(20);
 
 		/* Wait for response flag (bit 31) in C2PMSG_64 */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_64),
-				   0x80000000, 0x8000FFFF, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMP0_SMN_C2PMSG_64),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
 	}
 
 	return ret;
diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v14_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v14_0.c
new file mode 100644
index 000000000000..38dfc5c19f2a
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/psp_v14_0.c
@@ -0,0 +1,705 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include <drm/drm_drv.h>
+#include <linux/vmalloc.h>
+#include "amdgpu.h"
+#include "amdgpu_psp.h"
+#include "amdgpu_ucode.h"
+#include "soc15_common.h"
+#include "psp_v14_0.h"
+
+#include "mp/mp_14_0_2_offset.h"
+#include "mp/mp_14_0_2_sh_mask.h"
+
+MODULE_FIRMWARE("amdgpu/psp_14_0_2_sos.bin");
+MODULE_FIRMWARE("amdgpu/psp_14_0_2_ta.bin");
+MODULE_FIRMWARE("amdgpu/psp_14_0_3_sos.bin");
+MODULE_FIRMWARE("amdgpu/psp_14_0_3_sos_kicker.bin");
+MODULE_FIRMWARE("amdgpu/psp_14_0_3_ta.bin");
+MODULE_FIRMWARE("amdgpu/psp_14_0_3_ta_kicker.bin");
+MODULE_FIRMWARE("amdgpu/psp_14_0_5_toc.bin");
+MODULE_FIRMWARE("amdgpu/psp_14_0_5_ta.bin");
+
+/* For large FW files the time to complete can be very long */
+#define USBC_PD_POLLING_LIMIT_S 240
+
+/* Read USB-PD from LFB */
+#define GFX_CMD_USB_PD_USE_LFB 0x480
+
+/* VBIOS gfl defines */
+#define MBOX_READY_MASK 0x80000000
+#define MBOX_STATUS_MASK 0x0000FFFF
+#define MBOX_COMMAND_MASK 0x00FF0000
+#define MBOX_READY_FLAG 0x80000000
+#define C2PMSG_CMD_SPI_UPDATE_ROM_IMAGE_ADDR_LO 0x2
+#define C2PMSG_CMD_SPI_UPDATE_ROM_IMAGE_ADDR_HI 0x3
+#define C2PMSG_CMD_SPI_UPDATE_FLASH_IMAGE 0x4
+
+/* memory training timeout define */
+#define MEM_TRAIN_SEND_MSG_TIMEOUT_US	3000000
+
+static int psp_v14_0_init_microcode(struct psp_context *psp)
+{
+	struct amdgpu_device *adev = psp->adev;
+	char ucode_prefix[30];
+	int err = 0;
+
+	amdgpu_ucode_ip_version_decode(adev, MP0_HWIP, ucode_prefix, sizeof(ucode_prefix));
+
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
+	case IP_VERSION(14, 0, 2):
+	case IP_VERSION(14, 0, 3):
+		err = psp_init_sos_microcode(psp, ucode_prefix);
+		if (err)
+			return err;
+		err = psp_init_ta_microcode(psp, ucode_prefix);
+		if (err)
+			return err;
+		break;
+	case IP_VERSION(14, 0, 5):
+		err = psp_init_toc_microcode(psp, ucode_prefix);
+		if (err)
+			return err;
+		err = psp_init_ta_microcode(psp, ucode_prefix);
+		if (err)
+			return err;
+		break;
+	default:
+		BUG();
+	}
+
+	return 0;
+}
+
+static bool psp_v14_0_is_sos_alive(struct psp_context *psp)
+{
+	struct amdgpu_device *adev = psp->adev;
+	uint32_t sol_reg;
+
+	sol_reg = RREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_81);
+
+	return sol_reg != 0x0;
+}
+
+static int psp_v14_0_wait_for_bootloader(struct psp_context *psp)
+{
+	struct amdgpu_device *adev = psp->adev;
+
+	int ret;
+	int retry_loop;
+
+	for (retry_loop = 0; retry_loop < 10; retry_loop++) {
+		/* Wait for bootloader to signify that is
+		    ready having bit 31 of C2PMSG_35 set to 1 */
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMPASP_SMN_C2PMSG_35),
+			0x80000000, 0x80000000, PSP_WAITREG_NOVERBOSE);
+
+		if (ret == 0)
+			return 0;
+	}
+
+	return ret;
+}
+
+static int psp_v14_0_bootloader_load_component(struct psp_context  	*psp,
+					       struct psp_bin_desc 	*bin_desc,
+					       enum psp_bootloader_cmd  bl_cmd)
+{
+	int ret;
+	uint32_t psp_gfxdrv_command_reg = 0;
+	struct amdgpu_device *adev = psp->adev;
+
+	/* Check tOS sign of life register to confirm sys driver and sOS
+	 * are already been loaded.
+	 */
+	if (psp_v14_0_is_sos_alive(psp))
+		return 0;
+
+	ret = psp_v14_0_wait_for_bootloader(psp);
+	if (ret)
+		return ret;
+
+	memset(psp->fw_pri_buf, 0, PSP_1_MEG);
+
+	/* Copy PSP KDB binary to memory */
+	memcpy(psp->fw_pri_buf, bin_desc->start_addr, bin_desc->size_bytes);
+
+	/* Provide the PSP KDB to bootloader */
+	WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_36,
+	       (uint32_t)(psp->fw_pri_mc_addr >> 20));
+	psp_gfxdrv_command_reg = bl_cmd;
+	WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_35,
+	       psp_gfxdrv_command_reg);
+
+	ret = psp_v14_0_wait_for_bootloader(psp);
+
+	return ret;
+}
+
+static int psp_v14_0_bootloader_load_kdb(struct psp_context *psp)
+{
+	return psp_v14_0_bootloader_load_component(psp, &psp->kdb, PSP_BL__LOAD_KEY_DATABASE);
+}
+
+static int psp_v14_0_bootloader_load_spl(struct psp_context *psp)
+{
+	return psp_v14_0_bootloader_load_component(psp, &psp->spl, PSP_BL__LOAD_TOS_SPL_TABLE);
+}
+
+static int psp_v14_0_bootloader_load_sysdrv(struct psp_context *psp)
+{
+	return psp_v14_0_bootloader_load_component(psp, &psp->sys, PSP_BL__LOAD_SYSDRV);
+}
+
+static int psp_v14_0_bootloader_load_soc_drv(struct psp_context *psp)
+{
+	return psp_v14_0_bootloader_load_component(psp, &psp->soc_drv, PSP_BL__LOAD_SOCDRV);
+}
+
+static int psp_v14_0_bootloader_load_intf_drv(struct psp_context *psp)
+{
+	return psp_v14_0_bootloader_load_component(psp, &psp->intf_drv, PSP_BL__LOAD_INTFDRV);
+}
+
+static int psp_v14_0_bootloader_load_dbg_drv(struct psp_context *psp)
+{
+	/* dbg_drv was renamed to had_drv in psp v14 */
+	return psp_v14_0_bootloader_load_component(psp, &psp->dbg_drv, PSP_BL__LOAD_HADDRV);
+}
+
+static int psp_v14_0_bootloader_load_ras_drv(struct psp_context *psp)
+{
+	return psp_v14_0_bootloader_load_component(psp, &psp->ras_drv, PSP_BL__LOAD_RASDRV);
+}
+
+static int psp_v14_0_bootloader_load_ipkeymgr_drv(struct psp_context *psp)
+{
+	return psp_v14_0_bootloader_load_component(psp, &psp->ipkeymgr_drv, PSP_BL__LOAD_IPKEYMGRDRV);
+}
+
+static int psp_v14_0_bootloader_load_sos(struct psp_context *psp)
+{
+	int ret;
+	unsigned int psp_gfxdrv_command_reg = 0;
+	struct amdgpu_device *adev = psp->adev;
+
+	/* Check sOS sign of life register to confirm sys driver and sOS
+	 * are already been loaded.
+	 */
+	if (psp_v14_0_is_sos_alive(psp))
+		return 0;
+
+	ret = psp_v14_0_wait_for_bootloader(psp);
+	if (ret)
+		return ret;
+
+	memset(psp->fw_pri_buf, 0, PSP_1_MEG);
+
+	/* Copy Secure OS binary to PSP memory */
+	memcpy(psp->fw_pri_buf, psp->sos.start_addr, psp->sos.size_bytes);
+
+	/* Provide the PSP secure OS to bootloader */
+	WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_36,
+	       (uint32_t)(psp->fw_pri_mc_addr >> 20));
+	psp_gfxdrv_command_reg = PSP_BL__LOAD_SOSDRV;
+	WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_35,
+	       psp_gfxdrv_command_reg);
+
+	/* there might be handshake issue with hardware which needs delay */
+	mdelay(20);
+	ret = psp_wait_for(psp,
+			   SOC15_REG_OFFSET(MP0, 0, regMPASP_SMN_C2PMSG_81),
+			   RREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_81), 0,
+			   PSP_WAITREG_CHANGED);
+
+	return ret;
+}
+
+static int psp_v14_0_ring_stop(struct psp_context *psp,
+			       enum psp_ring_type ring_type)
+{
+	int ret = 0;
+	struct amdgpu_device *adev = psp->adev;
+
+	if (amdgpu_sriov_vf(adev)) {
+		/* Write the ring destroy command*/
+		WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_101,
+			     GFX_CTRL_CMD_ID_DESTROY_GPCOM_RING);
+		/* there might be handshake issue with hardware which needs delay */
+		mdelay(20);
+		/* Wait for response flag (bit 31) */
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMPASP_SMN_C2PMSG_101),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
+	} else {
+		/* Write the ring destroy command*/
+		WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_64,
+			     GFX_CTRL_CMD_ID_DESTROY_RINGS);
+		/* there might be handshake issue with hardware which needs delay */
+		mdelay(20);
+		/* Wait for response flag (bit 31) */
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMPASP_SMN_C2PMSG_64),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
+	}
+
+	return ret;
+}
+
+static int psp_v14_0_ring_create(struct psp_context *psp,
+				 enum psp_ring_type ring_type)
+{
+	int ret = 0;
+	unsigned int psp_ring_reg = 0;
+	struct psp_ring *ring = &psp->km_ring;
+	struct amdgpu_device *adev = psp->adev;
+
+	if (amdgpu_sriov_vf(adev)) {
+		ret = psp_v14_0_ring_stop(psp, ring_type);
+		if (ret) {
+			DRM_ERROR("psp_v14_0_ring_stop_sriov failed!\n");
+			return ret;
+		}
+
+		/* Write low address of the ring to C2PMSG_102 */
+		psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr);
+		WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_102, psp_ring_reg);
+		/* Write high address of the ring to C2PMSG_103 */
+		psp_ring_reg = upper_32_bits(ring->ring_mem_mc_addr);
+		WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_103, psp_ring_reg);
+
+		/* Write the ring initialization command to C2PMSG_101 */
+		WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_101,
+			     GFX_CTRL_CMD_ID_INIT_GPCOM_RING);
+
+		/* there might be handshake issue with hardware which needs delay */
+		mdelay(20);
+
+		/* Wait for response flag (bit 31) in C2PMSG_101 */
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMPASP_SMN_C2PMSG_101),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
+
+	} else {
+		/* Wait for sOS ready for ring creation */
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMPASP_SMN_C2PMSG_64),
+			MBOX_TOS_READY_FLAG, MBOX_TOS_READY_MASK, 0);
+		if (ret) {
+			DRM_ERROR("Failed to wait for trust OS ready for ring creation\n");
+			return ret;
+		}
+
+		/* Write low address of the ring to C2PMSG_69 */
+		psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr);
+		WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_69, psp_ring_reg);
+		/* Write high address of the ring to C2PMSG_70 */
+		psp_ring_reg = upper_32_bits(ring->ring_mem_mc_addr);
+		WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_70, psp_ring_reg);
+		/* Write size of ring to C2PMSG_71 */
+		psp_ring_reg = ring->ring_size;
+		WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_71, psp_ring_reg);
+		/* Write the ring initialization command to C2PMSG_64 */
+		psp_ring_reg = ring_type;
+		psp_ring_reg = psp_ring_reg << 16;
+		WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_64, psp_ring_reg);
+
+		/* there might be handshake issue with hardware which needs delay */
+		mdelay(20);
+
+		/* Wait for response flag (bit 31) in C2PMSG_64 */
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMPASP_SMN_C2PMSG_64),
+			MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, 0);
+	}
+
+	return ret;
+}
+
+static int psp_v14_0_ring_destroy(struct psp_context *psp,
+				  enum psp_ring_type ring_type)
+{
+	int ret = 0;
+	struct psp_ring *ring = &psp->km_ring;
+	struct amdgpu_device *adev = psp->adev;
+
+	ret = psp_v14_0_ring_stop(psp, ring_type);
+	if (ret)
+		DRM_ERROR("Fail to stop psp ring\n");
+
+	amdgpu_bo_free_kernel(&adev->firmware.rbuf,
+			      &ring->ring_mem_mc_addr,
+			      (void **)&ring->ring_mem);
+
+	return ret;
+}
+
+static uint32_t psp_v14_0_ring_get_wptr(struct psp_context *psp)
+{
+	uint32_t data;
+	struct amdgpu_device *adev = psp->adev;
+
+	if (amdgpu_sriov_vf(adev))
+		data = RREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_102);
+	else
+		data = RREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_67);
+
+	return data;
+}
+
+static void psp_v14_0_ring_set_wptr(struct psp_context *psp, uint32_t value)
+{
+	struct amdgpu_device *adev = psp->adev;
+
+	if (amdgpu_sriov_vf(adev)) {
+		WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_102, value);
+		WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_101,
+			     GFX_CTRL_CMD_ID_CONSUME_CMD);
+	} else
+		WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_67, value);
+}
+
+static int psp_v14_0_memory_training_send_msg(struct psp_context *psp, int msg)
+{
+	int ret;
+	int i;
+	uint32_t data_32;
+	int max_wait;
+	struct amdgpu_device *adev = psp->adev;
+
+	data_32 = (psp->mem_train_ctx.c2p_train_data_offset >> 20);
+	WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_36, data_32);
+	WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_35, msg);
+
+	max_wait = MEM_TRAIN_SEND_MSG_TIMEOUT_US / adev->usec_timeout;
+	for (i = 0; i < max_wait; i++) {
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMPASP_SMN_C2PMSG_35),
+			0x80000000, 0x80000000, PSP_WAITREG_NOVERBOSE);
+		if (ret == 0)
+			break;
+	}
+	if (i < max_wait)
+		ret = 0;
+	else
+		ret = -ETIME;
+
+	dev_dbg(adev->dev, "training %s %s, cost %d @ %d ms\n",
+		  (msg == PSP_BL__DRAM_SHORT_TRAIN) ? "short" : "long",
+		  (ret == 0) ? "succeed" : "failed",
+		  i, adev->usec_timeout/1000);
+	return ret;
+}
+
+
+static int psp_v14_0_memory_training(struct psp_context *psp, uint32_t ops)
+{
+	struct psp_memory_training_context *ctx = &psp->mem_train_ctx;
+	uint32_t *pcache = (uint32_t *)ctx->sys_cache;
+	struct amdgpu_device *adev = psp->adev;
+	uint32_t p2c_header[4];
+	uint32_t sz;
+	void *buf;
+	int ret, idx;
+
+	if (ctx->init == PSP_MEM_TRAIN_NOT_SUPPORT) {
+		dev_dbg(adev->dev, "Memory training is not supported.\n");
+		return 0;
+	} else if (ctx->init != PSP_MEM_TRAIN_INIT_SUCCESS) {
+		dev_err(adev->dev, "Memory training initialization failure.\n");
+		return -EINVAL;
+	}
+
+	if (psp_v14_0_is_sos_alive(psp)) {
+		dev_dbg(adev->dev, "SOS is alive, skip memory training.\n");
+		return 0;
+	}
+
+	amdgpu_device_vram_access(adev, ctx->p2c_train_data_offset, p2c_header, sizeof(p2c_header), false);
+	dev_dbg(adev->dev, "sys_cache[%08x,%08x,%08x,%08x] p2c_header[%08x,%08x,%08x,%08x]\n",
+		  pcache[0], pcache[1], pcache[2], pcache[3],
+		  p2c_header[0], p2c_header[1], p2c_header[2], p2c_header[3]);
+
+	if (ops & PSP_MEM_TRAIN_SEND_SHORT_MSG) {
+		dev_dbg(adev->dev, "Short training depends on restore.\n");
+		ops |= PSP_MEM_TRAIN_RESTORE;
+	}
+
+	if ((ops & PSP_MEM_TRAIN_RESTORE) &&
+	    pcache[0] != MEM_TRAIN_SYSTEM_SIGNATURE) {
+		dev_dbg(adev->dev, "sys_cache[0] is invalid, restore depends on save.\n");
+		ops |= PSP_MEM_TRAIN_SAVE;
+	}
+
+	if (p2c_header[0] == MEM_TRAIN_SYSTEM_SIGNATURE &&
+	    !(pcache[0] == MEM_TRAIN_SYSTEM_SIGNATURE &&
+	      pcache[3] == p2c_header[3])) {
+		dev_dbg(adev->dev, "sys_cache is invalid or out-of-date, need save training data to sys_cache.\n");
+		ops |= PSP_MEM_TRAIN_SAVE;
+	}
+
+	if ((ops & PSP_MEM_TRAIN_SAVE) &&
+	    p2c_header[0] != MEM_TRAIN_SYSTEM_SIGNATURE) {
+		dev_dbg(adev->dev, "p2c_header[0] is invalid, save depends on long training.\n");
+		ops |= PSP_MEM_TRAIN_SEND_LONG_MSG;
+	}
+
+	if (ops & PSP_MEM_TRAIN_SEND_LONG_MSG) {
+		ops &= ~PSP_MEM_TRAIN_SEND_SHORT_MSG;
+		ops |= PSP_MEM_TRAIN_SAVE;
+	}
+
+	dev_dbg(adev->dev, "Memory training ops:%x.\n", ops);
+
+	if (ops & PSP_MEM_TRAIN_SEND_LONG_MSG) {
+		/*
+		 * Long training will encroach a certain amount on the bottom of VRAM;
+		 * save the content from the bottom of VRAM to system memory
+		 * before training, and restore it after training to avoid
+		 * VRAM corruption.
+		 */
+		sz = BIST_MEM_TRAINING_ENCROACHED_SIZE;
+
+		if (adev->gmc.visible_vram_size < sz || !adev->mman.aper_base_kaddr) {
+			dev_err(adev->dev, "visible_vram_size %llx or aper_base_kaddr %p is not initialized.\n",
+				  adev->gmc.visible_vram_size,
+				  adev->mman.aper_base_kaddr);
+			return -EINVAL;
+		}
+
+		buf = vmalloc(sz);
+		if (!buf) {
+			dev_err(adev->dev, "failed to allocate system memory.\n");
+			return -ENOMEM;
+		}
+
+		if (drm_dev_enter(adev_to_drm(adev), &idx)) {
+			memcpy_fromio(buf, adev->mman.aper_base_kaddr, sz);
+			ret = psp_v14_0_memory_training_send_msg(psp, PSP_BL__DRAM_LONG_TRAIN);
+			if (ret) {
+				DRM_ERROR("Send long training msg failed.\n");
+				vfree(buf);
+				drm_dev_exit(idx);
+				return ret;
+			}
+
+			memcpy_toio(adev->mman.aper_base_kaddr, buf, sz);
+			amdgpu_device_flush_hdp(adev, NULL);
+			vfree(buf);
+			drm_dev_exit(idx);
+		} else {
+			vfree(buf);
+			return -ENODEV;
+		}
+	}
+
+	if (ops & PSP_MEM_TRAIN_SAVE) {
+		amdgpu_device_vram_access(psp->adev, ctx->p2c_train_data_offset, ctx->sys_cache, ctx->train_data_size, false);
+	}
+
+	if (ops & PSP_MEM_TRAIN_RESTORE) {
+		amdgpu_device_vram_access(psp->adev, ctx->c2p_train_data_offset, ctx->sys_cache, ctx->train_data_size, true);
+	}
+
+	if (ops & PSP_MEM_TRAIN_SEND_SHORT_MSG) {
+		ret = psp_v14_0_memory_training_send_msg(psp, (amdgpu_force_long_training > 0) ?
+							 PSP_BL__DRAM_LONG_TRAIN : PSP_BL__DRAM_SHORT_TRAIN);
+		if (ret) {
+			dev_err(adev->dev, "send training msg failed.\n");
+			return ret;
+		}
+	}
+	ctx->training_cnt++;
+	return 0;
+}
+
+static int psp_v14_0_load_usbc_pd_fw(struct psp_context *psp, uint64_t fw_pri_mc_addr)
+{
+	struct amdgpu_device *adev = psp->adev;
+	uint32_t reg_status;
+	int ret, i = 0;
+
+	/*
+	 * LFB address which is aligned to 1MB address and has to be
+	 * right-shifted by 20 so that LFB address can be passed on a 32-bit C2P
+	 * register
+	 */
+	WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_36, (fw_pri_mc_addr >> 20));
+
+	ret = psp_wait_for(psp,
+			   SOC15_REG_OFFSET(MP0, 0, regMPASP_SMN_C2PMSG_35),
+			   0x80000000, 0x80000000, 0);
+	if (ret)
+		return ret;
+
+	/* Fireup interrupt so PSP can pick up the address */
+	WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_35, (GFX_CMD_USB_PD_USE_LFB << 16));
+
+	/* FW load takes very long time */
+	do {
+		msleep(1000);
+		reg_status = RREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_35);
+
+		if (reg_status & 0x80000000)
+			goto done;
+
+	} while (++i < USBC_PD_POLLING_LIMIT_S);
+
+	return -ETIME;
+done:
+
+	if ((reg_status & 0xFFFF) != 0) {
+		DRM_ERROR("Address load failed - MP0_SMN_C2PMSG_35.Bits [15:0] = %04x\n",
+				reg_status & 0xFFFF);
+		return -EIO;
+	}
+
+	return 0;
+}
+
+static int psp_v14_0_read_usbc_pd_fw(struct psp_context *psp, uint32_t *fw_ver)
+{
+	struct amdgpu_device *adev = psp->adev;
+	int ret;
+
+	WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_35, C2PMSG_CMD_GFX_USB_PD_FW_VER);
+
+	ret = psp_wait_for(psp,
+			   SOC15_REG_OFFSET(MP0, 0, regMPASP_SMN_C2PMSG_35),
+			   0x80000000, 0x80000000, 0);
+	if (!ret)
+		*fw_ver = RREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_36);
+
+	return ret;
+}
+
+static int psp_v14_0_exec_spi_cmd(struct psp_context *psp, int cmd)
+{
+	uint32_t reg_status = 0, reg_val = 0;
+	struct amdgpu_device *adev = psp->adev;
+	int ret;
+
+	/* clear MBX ready (MBOX_READY_MASK bit is 0) and set update command */
+	reg_val |= (cmd << 16);
+	WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_115,  reg_val);
+
+	/* Ring the doorbell */
+	WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_73, 1);
+
+	if (cmd == C2PMSG_CMD_SPI_UPDATE_FLASH_IMAGE)
+		ret = psp_wait_for_spirom_update(psp, SOC15_REG_OFFSET(MP0, 0, regMPASP_SMN_C2PMSG_115),
+						 MBOX_READY_FLAG, MBOX_READY_MASK, PSP_SPIROM_UPDATE_TIMEOUT);
+	else
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, regMPASP_SMN_C2PMSG_115),
+			MBOX_READY_FLAG, MBOX_READY_MASK, 0);
+
+	ret = psp_wait_for(psp,
+			   SOC15_REG_OFFSET(MP0, 0, regMPASP_SMN_C2PMSG_115),
+			   MBOX_READY_FLAG, MBOX_READY_MASK, 0);
+	if (ret) {
+		dev_err(adev->dev, "SPI cmd %x timed out, ret = %d", cmd, ret);
+		return ret;
+	}
+
+	reg_status = RREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_115);
+	if ((reg_status & 0xFFFF) != 0) {
+		dev_err(adev->dev, "SPI cmd %x failed, fail status = %04x\n",
+				cmd, reg_status & 0xFFFF);
+		return -EIO;
+	}
+
+	return 0;
+}
+
+static int psp_v14_0_update_spirom(struct psp_context *psp,
+				   uint64_t fw_pri_mc_addr)
+{
+	struct amdgpu_device *adev = psp->adev;
+	int ret;
+
+	/* Confirm PSP is ready to start */
+	ret = psp_wait_for(psp,
+			   SOC15_REG_OFFSET(MP0, 0, regMPASP_SMN_C2PMSG_115),
+			   MBOX_READY_FLAG, MBOX_READY_MASK, 0);
+	if (ret) {
+		dev_err(adev->dev, "PSP Not ready to start processing, ret = %d", ret);
+		return ret;
+	}
+
+	WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_116, lower_32_bits(fw_pri_mc_addr));
+
+	ret = psp_v14_0_exec_spi_cmd(psp, C2PMSG_CMD_SPI_UPDATE_ROM_IMAGE_ADDR_LO);
+	if (ret)
+		return ret;
+
+	WREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_116, upper_32_bits(fw_pri_mc_addr));
+
+	ret = psp_v14_0_exec_spi_cmd(psp, C2PMSG_CMD_SPI_UPDATE_ROM_IMAGE_ADDR_HI);
+	if (ret)
+		return ret;
+
+	psp->vbflash_done = true;
+
+	ret = psp_v14_0_exec_spi_cmd(psp, C2PMSG_CMD_SPI_UPDATE_FLASH_IMAGE);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static int psp_v14_0_vbflash_status(struct psp_context *psp)
+{
+	struct amdgpu_device *adev = psp->adev;
+
+	return RREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_115);
+}
+
+static const struct psp_funcs psp_v14_0_funcs = {
+	.init_microcode = psp_v14_0_init_microcode,
+	.bootloader_load_kdb = psp_v14_0_bootloader_load_kdb,
+	.bootloader_load_spl = psp_v14_0_bootloader_load_spl,
+	.bootloader_load_sysdrv = psp_v14_0_bootloader_load_sysdrv,
+	.bootloader_load_soc_drv = psp_v14_0_bootloader_load_soc_drv,
+	.bootloader_load_intf_drv = psp_v14_0_bootloader_load_intf_drv,
+	.bootloader_load_dbg_drv = psp_v14_0_bootloader_load_dbg_drv,
+	.bootloader_load_ras_drv = psp_v14_0_bootloader_load_ras_drv,
+	.bootloader_load_ipkeymgr_drv = psp_v14_0_bootloader_load_ipkeymgr_drv,
+	.bootloader_load_sos = psp_v14_0_bootloader_load_sos,
+	.ring_create = psp_v14_0_ring_create,
+	.ring_stop = psp_v14_0_ring_stop,
+	.ring_destroy = psp_v14_0_ring_destroy,
+	.ring_get_wptr = psp_v14_0_ring_get_wptr,
+	.ring_set_wptr = psp_v14_0_ring_set_wptr,
+	.mem_training = psp_v14_0_memory_training,
+	.load_usbc_pd_fw = psp_v14_0_load_usbc_pd_fw,
+	.read_usbc_pd_fw = psp_v14_0_read_usbc_pd_fw,
+	.update_spirom = psp_v14_0_update_spirom,
+	.vbflash_stat = psp_v14_0_vbflash_status
+};
+
+void psp_v14_0_set_psp_funcs(struct psp_context *psp)
+{
+	psp->funcs = &psp_v14_0_funcs;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v14_0.h b/drivers/gpu/drm/amd/amdgpu/psp_v14_0.h
new file mode 100644
index 000000000000..dd18ba2cfad5
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/psp_v14_0.h
@@ -0,0 +1,32 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __PSP_V14_0_H__
+#define __PSP_V14_0_H__
+
+#include "amdgpu_psp.h"
+
+#define PSP_SPIROM_UPDATE_TIMEOUT   60000   /* 60s */
+
+void psp_v14_0_set_psp_funcs(struct psp_context *psp);
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c
index f6b75e3e47ff..833830bc3e2e 100644
--- a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c
@@ -91,7 +91,7 @@ static int psp_v3_1_bootloader_load_sysdrv(struct psp_context *psp)
 
 	/* Wait for bootloader to signify that is ready having bit 31 of C2PMSG_35 set to 1 */
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35),
-			   0x80000000, 0x80000000, false);
+			   0x80000000, 0x80000000, 0);
 	if (ret)
 		return ret;
 
@@ -109,7 +109,7 @@ static int psp_v3_1_bootloader_load_sysdrv(struct psp_context *psp)
 	mdelay(20);
 
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35),
-			   0x80000000, 0x80000000, false);
+			   0x80000000, 0x80000000, 0);
 
 	return ret;
 }
@@ -130,7 +130,7 @@ static int psp_v3_1_bootloader_load_sos(struct psp_context *psp)
 
 	/* Wait for bootloader to signify that is ready having bit 31 of C2PMSG_35 set to 1 */
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35),
-			   0x80000000, 0x80000000, false);
+			   0x80000000, 0x80000000, 0);
 	if (ret)
 		return ret;
 
@@ -147,8 +147,8 @@ static int psp_v3_1_bootloader_load_sos(struct psp_context *psp)
 	/* there might be handshake issue with hardware which needs delay */
 	mdelay(20);
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_81),
-			   RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81),
-			   0, true);
+			   RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81), 0,
+			   PSP_WAITREG_CHANGED);
 	return ret;
 }
 
@@ -168,7 +168,7 @@ static void psp_v3_1_reroute_ih(struct psp_context *psp)
 
 	mdelay(20);
 	psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
-		     0x80000000, 0x8000FFFF, false);
+		     0x80000000, 0x8000FFFF, 0);
 
 	/* Change IH ring for UMC */
 	tmp = REG_SET_FIELD(0, IH_CLIENT_CFG_DATA, CREDIT_RETURN_ADDR, 0x1216b);
@@ -180,7 +180,7 @@ static void psp_v3_1_reroute_ih(struct psp_context *psp)
 
 	mdelay(20);
 	psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
-		     0x80000000, 0x8000FFFF, false);
+		     0x80000000, 0x8000FFFF, 0);
 }
 
 static int psp_v3_1_ring_create(struct psp_context *psp,
@@ -217,9 +217,9 @@ static int psp_v3_1_ring_create(struct psp_context *psp,
 		mdelay(20);
 
 		/* Wait for response flag (bit 31) in C2PMSG_101 */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0,
-					mmMP0_SMN_C2PMSG_101), 0x80000000,
-					0x8000FFFF, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_101),
+			0x80000000, 0x8000FFFF, 0);
 	} else {
 
 		/* Write low address of the ring to C2PMSG_69 */
@@ -240,10 +240,9 @@ static int psp_v3_1_ring_create(struct psp_context *psp,
 		mdelay(20);
 
 		/* Wait for response flag (bit 31) in C2PMSG_64 */
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0,
-					mmMP0_SMN_C2PMSG_64), 0x80000000,
-					0x8000FFFF, false);
-
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
+			0x80000000, 0x8000FFFF, 0);
 	}
 	return ret;
 }
@@ -267,11 +266,13 @@ static int psp_v3_1_ring_stop(struct psp_context *psp,
 
 	/* Wait for response flag (bit 31) */
 	if (amdgpu_sriov_vf(adev))
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_101),
-				   0x80000000, 0x80000000, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_101),
+			0x80000000, 0x80000000, 0);
 	else
-		ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
-				   0x80000000, 0x80000000, false);
+		ret = psp_wait_for(
+			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
+			0x80000000, 0x80000000, 0);
 
 	return ret;
 }
@@ -311,7 +312,7 @@ static int psp_v3_1_mode1_reset(struct psp_context *psp)
 
 	offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64);
 
-	ret = psp_wait_for(psp, offset, 0x80000000, 0x8000FFFF, false);
+	ret = psp_wait_for(psp, offset, 0x80000000, 0x8000FFFF, 0);
 
 	if (ret) {
 		DRM_INFO("psp is not working correctly before mode1 reset!\n");
@@ -325,7 +326,7 @@ static int psp_v3_1_mode1_reset(struct psp_context *psp)
 
 	offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_33);
 
-	ret = psp_wait_for(psp, offset, 0x80000000, 0x80000000, false);
+	ret = psp_wait_for(psp, offset, 0x80000000, 0x80000000, 0);
 
 	if (ret) {
 		DRM_INFO("psp mode 1 reset failed!\n");
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
index fd2a7b66ac56..92ce580647cd 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
@@ -57,22 +57,19 @@ static void sdma_v2_4_set_irq_funcs(struct amdgpu_device *adev);
 MODULE_FIRMWARE("amdgpu/topaz_sdma.bin");
 MODULE_FIRMWARE("amdgpu/topaz_sdma1.bin");
 
-static const u32 sdma_offsets[SDMA_MAX_INSTANCE] =
-{
+static const u32 sdma_offsets[SDMA_MAX_INSTANCE] = {
 	SDMA0_REGISTER_OFFSET,
 	SDMA1_REGISTER_OFFSET
 };
 
-static const u32 golden_settings_iceland_a11[] =
-{
+static const u32 golden_settings_iceland_a11[] = {
 	mmSDMA0_CHICKEN_BITS, 0xfc910007, 0x00810007,
 	mmSDMA0_CLK_CTRL, 0xff000fff, 0x00000000,
 	mmSDMA1_CHICKEN_BITS, 0xfc910007, 0x00810007,
 	mmSDMA1_CLK_CTRL, 0xff000fff, 0x00000000,
 };
 
-static const u32 iceland_mgcg_cgcg_init[] =
-{
+static const u32 iceland_mgcg_cgcg_init[] = {
 	mmSDMA0_CLK_CTRL, 0xff000ff0, 0x00000100,
 	mmSDMA1_CLK_CTRL, 0xff000ff0, 0x00000100
 };
@@ -130,7 +127,6 @@ static void sdma_v2_4_free_microcode(struct amdgpu_device *adev)
 static int sdma_v2_4_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[30];
 	int err = 0, i;
 	struct amdgpu_firmware_info *info = NULL;
 	const struct common_firmware_header *header = NULL;
@@ -142,15 +138,19 @@ static int sdma_v2_4_init_microcode(struct amdgpu_device *adev)
 	case CHIP_TOPAZ:
 		chip_name = "topaz";
 		break;
-	default: BUG();
+	default:
+		BUG();
 	}
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		if (i == 0)
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sdma.bin", chip_name);
+			err = amdgpu_ucode_request(adev, &adev->sdma.instance[i].fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_sdma.bin", chip_name);
 		else
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sdma1.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->sdma.instance[i].fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->sdma.instance[i].fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_sdma1.bin", chip_name);
 		if (err)
 			goto out;
 		hdr = (const struct sdma_firmware_header_v1_0 *)adev->sdma.instance[i].fw->data;
@@ -171,7 +171,8 @@ static int sdma_v2_4_init_microcode(struct amdgpu_device *adev)
 
 out:
 	if (err) {
-		pr_err("sdma_v2_4: Failed to load firmware \"%s\"\n", fw_name);
+		pr_err("sdma_v2_4: Failed to load firmware \"%s_sdma%s.bin\"\n",
+		       chip_name, i == 0 ? "" : "1");
 		for (i = 0; i < adev->sdma.num_instances; i++)
 			amdgpu_ucode_release(&adev->sdma.instance[i].fw);
 	}
@@ -339,8 +340,6 @@ static void sdma_v2_4_gfx_stop(struct amdgpu_device *adev)
 	u32 rb_cntl, ib_cntl;
 	int i;
 
-	amdgpu_sdma_unset_buffer_funcs_helper(adev);
-
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		rb_cntl = RREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i]);
 		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 0);
@@ -466,8 +465,6 @@ static int sdma_v2_4_gfx_resume(struct amdgpu_device *adev)
 #endif
 		/* enable DMA IBs */
 		WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl);
-
-		ring->sched.ready = true;
 	}
 
 	sdma_v2_4_enable(adev, true);
@@ -476,9 +473,6 @@ static int sdma_v2_4_gfx_resume(struct amdgpu_device *adev)
 		r = amdgpu_ring_test_helper(ring);
 		if (r)
 			return r;
-
-		if (adev->mman.buffer_funcs_ring == ring)
-			amdgpu_ttm_set_buffer_funcs_status(adev, true);
 	}
 
 	return 0;
@@ -639,7 +633,7 @@ static int sdma_v2_4_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = -EINVAL;
 
 err1:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err0:
 	amdgpu_device_wb_free(adev, index);
@@ -815,12 +809,17 @@ static void sdma_v2_4_ring_emit_wreg(struct amdgpu_ring *ring,
 	amdgpu_ring_write(ring, val);
 }
 
-static int sdma_v2_4_early_init(void *handle)
+static int sdma_v2_4_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
 
 	adev->sdma.num_instances = SDMA_MAX_INSTANCE;
 
+	r = sdma_v2_4_init_microcode(adev);
+	if (r)
+		return r;
+
 	sdma_v2_4_set_ring_funcs(adev);
 	sdma_v2_4_set_buffer_funcs(adev);
 	sdma_v2_4_set_vm_pte_funcs(adev);
@@ -829,11 +828,11 @@ static int sdma_v2_4_early_init(void *handle)
 	return 0;
 }
 
-static int sdma_v2_4_sw_init(void *handle)
+static int sdma_v2_4_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
 	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* SDMA trap event */
 	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SDMA_TRAP,
@@ -853,12 +852,6 @@ static int sdma_v2_4_sw_init(void *handle)
 	if (r)
 		return r;
 
-	r = sdma_v2_4_init_microcode(adev);
-	if (r) {
-		DRM_ERROR("Failed to load sdma firmware!\n");
-		return r;
-	}
-
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		ring = &adev->sdma.instance[i].ring;
 		ring->ring_obj = NULL;
@@ -875,9 +868,9 @@ static int sdma_v2_4_sw_init(void *handle)
 	return r;
 }
 
-static int sdma_v2_4_sw_fini(void *handle)
+static int sdma_v2_4_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++)
@@ -887,10 +880,10 @@ static int sdma_v2_4_sw_fini(void *handle)
 	return 0;
 }
 
-static int sdma_v2_4_hw_init(void *handle)
+static int sdma_v2_4_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	sdma_v2_4_init_golden_registers(adev);
 
@@ -901,32 +894,26 @@ static int sdma_v2_4_hw_init(void *handle)
 	return r;
 }
 
-static int sdma_v2_4_hw_fini(void *handle)
+static int sdma_v2_4_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	sdma_v2_4_enable(adev, false);
+	sdma_v2_4_enable(ip_block->adev, false);
 
 	return 0;
 }
 
-static int sdma_v2_4_suspend(void *handle)
+static int sdma_v2_4_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return sdma_v2_4_hw_fini(adev);
+	return sdma_v2_4_hw_fini(ip_block);
 }
 
-static int sdma_v2_4_resume(void *handle)
+static int sdma_v2_4_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return sdma_v2_4_hw_init(adev);
+	return sdma_v2_4_hw_init(ip_block);
 }
 
-static bool sdma_v2_4_is_idle(void *handle)
+static bool sdma_v2_4_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS2);
 
 	if (tmp & (SRBM_STATUS2__SDMA_BUSY_MASK |
@@ -936,11 +923,11 @@ static bool sdma_v2_4_is_idle(void *handle)
 	return true;
 }
 
-static int sdma_v2_4_wait_for_idle(void *handle)
+static int sdma_v2_4_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		tmp = RREG32(mmSRBM_STATUS2) & (SRBM_STATUS2__SDMA_BUSY_MASK |
@@ -953,10 +940,10 @@ static int sdma_v2_4_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int sdma_v2_4_soft_reset(void *handle)
+static int sdma_v2_4_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 srbm_soft_reset = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS2);
 
 	if (tmp & SRBM_STATUS2__SDMA_BUSY_MASK) {
@@ -1095,14 +1082,14 @@ static int sdma_v2_4_process_illegal_inst_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int sdma_v2_4_set_clockgating_state(void *handle,
+static int sdma_v2_4_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	/* XXX handled via the smc on VI */
 	return 0;
 }
 
-static int sdma_v2_4_set_powergating_state(void *handle,
+static int sdma_v2_4_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -1111,7 +1098,6 @@ static int sdma_v2_4_set_powergating_state(void *handle,
 static const struct amd_ip_funcs sdma_v2_4_ip_funcs = {
 	.name = "sdma_v2_4",
 	.early_init = sdma_v2_4_early_init,
-	.late_init = NULL,
 	.sw_init = sdma_v2_4_sw_init,
 	.sw_fini = sdma_v2_4_sw_fini,
 	.hw_init = sdma_v2_4_hw_init,
@@ -1186,7 +1172,7 @@ static void sdma_v2_4_set_irq_funcs(struct amdgpu_device *adev)
  * @src_offset: src GPU address
  * @dst_offset: dst GPU address
  * @byte_count: number of bytes to xfer
- * @tmz: unused
+ * @copy_flags: unused
  *
  * Copy GPU buffers using the DMA engine (VI).
  * Used by the amdgpu ttm implementation to move pages if
@@ -1196,7 +1182,7 @@ static void sdma_v2_4_emit_copy_buffer(struct amdgpu_ib *ib,
 				       uint64_t src_offset,
 				       uint64_t dst_offset,
 				       uint32_t byte_count,
-				       bool tmz)
+				       uint32_t copy_flags)
 {
 	ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) |
 		SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR);
@@ -1266,8 +1252,7 @@ static void sdma_v2_4_set_vm_pte_funcs(struct amdgpu_device *adev)
 	adev->vm_manager.vm_pte_num_scheds = adev->sdma.num_instances;
 }
 
-const struct amdgpu_ip_block_version sdma_v2_4_ip_block =
-{
+const struct amdgpu_ip_block_version sdma_v2_4_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_SDMA,
 	.major = 2,
 	.minor = 4,
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
index e572389089d2..1c076bd1cf73 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
@@ -267,7 +267,6 @@ static void sdma_v3_0_free_microcode(struct amdgpu_device *adev)
 static int sdma_v3_0_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[30];
 	int err = 0, i;
 	struct amdgpu_firmware_info *info = NULL;
 	const struct common_firmware_header *header = NULL;
@@ -305,10 +304,13 @@ static int sdma_v3_0_init_microcode(struct amdgpu_device *adev)
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		if (i == 0)
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sdma.bin", chip_name);
+			err = amdgpu_ucode_request(adev, &adev->sdma.instance[i].fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_sdma.bin", chip_name);
 		else
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sdma1.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->sdma.instance[i].fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->sdma.instance[i].fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_sdma1.bin", chip_name);
 		if (err)
 			goto out;
 		hdr = (const struct sdma_firmware_header_v1_0 *)adev->sdma.instance[i].fw->data;
@@ -327,7 +329,8 @@ static int sdma_v3_0_init_microcode(struct amdgpu_device *adev)
 	}
 out:
 	if (err) {
-		pr_err("sdma_v3_0: Failed to load firmware \"%s\"\n", fw_name);
+		pr_err("sdma_v3_0: Failed to load firmware \"%s_sdma%s.bin\"\n",
+		       chip_name, i == 0 ? "" : "1");
 		for (i = 0; i < adev->sdma.num_instances; i++)
 			amdgpu_ucode_release(&adev->sdma.instance[i].fw);
 	}
@@ -513,8 +516,6 @@ static void sdma_v3_0_gfx_stop(struct amdgpu_device *adev)
 	u32 rb_cntl, ib_cntl;
 	int i;
 
-	amdgpu_sdma_unset_buffer_funcs_helper(adev);
-
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		rb_cntl = RREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i]);
 		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 0);
@@ -711,7 +712,7 @@ static int sdma_v3_0_gfx_resume(struct amdgpu_device *adev)
 		       upper_32_bits(wptr_gpu_addr));
 		wptr_poll_cntl = RREG32(mmSDMA0_GFX_RB_WPTR_POLL_CNTL + sdma_offsets[i]);
 		if (ring->use_pollmem) {
-			/*wptr polling is not enogh fast, directly clean the wptr register */
+			/*wptr polling is not enough fast, directly clean the wptr register */
 			WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[i], 0);
 			wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl,
 						       SDMA0_GFX_RB_WPTR_POLL_CNTL,
@@ -734,8 +735,6 @@ static int sdma_v3_0_gfx_resume(struct amdgpu_device *adev)
 #endif
 		/* enable DMA IBs */
 		WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl);
-
-		ring->sched.ready = true;
 	}
 
 	/* unhalt the MEs */
@@ -748,9 +747,6 @@ static int sdma_v3_0_gfx_resume(struct amdgpu_device *adev)
 		r = amdgpu_ring_test_helper(ring);
 		if (r)
 			return r;
-
-		if (adev->mman.buffer_funcs_ring == ring)
-			amdgpu_ttm_set_buffer_funcs_status(adev, true);
 	}
 
 	return 0;
@@ -910,7 +906,7 @@ static int sdma_v3_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	else
 		r = -EINVAL;
 err1:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err0:
 	amdgpu_device_wb_free(adev, index);
@@ -1086,9 +1082,10 @@ static void sdma_v3_0_ring_emit_wreg(struct amdgpu_ring *ring,
 	amdgpu_ring_write(ring, val);
 }
 
-static int sdma_v3_0_early_init(void *handle)
+static int sdma_v3_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
 
 	switch (adev->asic_type) {
 	case CHIP_STONEY:
@@ -1099,6 +1096,10 @@ static int sdma_v3_0_early_init(void *handle)
 		break;
 	}
 
+	r = sdma_v3_0_init_microcode(adev);
+	if (r)
+		return r;
+
 	sdma_v3_0_set_ring_funcs(adev);
 	sdma_v3_0_set_buffer_funcs(adev);
 	sdma_v3_0_set_vm_pte_funcs(adev);
@@ -1107,11 +1108,11 @@ static int sdma_v3_0_early_init(void *handle)
 	return 0;
 }
 
-static int sdma_v3_0_sw_init(void *handle)
+static int sdma_v3_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
 	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* SDMA trap event */
 	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SDMA_TRAP,
@@ -1131,12 +1132,6 @@ static int sdma_v3_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	r = sdma_v3_0_init_microcode(adev);
-	if (r) {
-		DRM_ERROR("Failed to load sdma firmware!\n");
-		return r;
-	}
-
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		ring = &adev->sdma.instance[i].ring;
 		ring->ring_obj = NULL;
@@ -1159,9 +1154,9 @@ static int sdma_v3_0_sw_init(void *handle)
 	return r;
 }
 
-static int sdma_v3_0_sw_fini(void *handle)
+static int sdma_v3_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++)
@@ -1171,10 +1166,10 @@ static int sdma_v3_0_sw_fini(void *handle)
 	return 0;
 }
 
-static int sdma_v3_0_hw_init(void *handle)
+static int sdma_v3_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	sdma_v3_0_init_golden_registers(adev);
 
@@ -1185,9 +1180,9 @@ static int sdma_v3_0_hw_init(void *handle)
 	return r;
 }
 
-static int sdma_v3_0_hw_fini(void *handle)
+static int sdma_v3_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	sdma_v3_0_ctx_switch_enable(adev, false);
 	sdma_v3_0_enable(adev, false);
@@ -1195,23 +1190,19 @@ static int sdma_v3_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int sdma_v3_0_suspend(void *handle)
+static int sdma_v3_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return sdma_v3_0_hw_fini(adev);
+	return sdma_v3_0_hw_fini(ip_block);
 }
 
-static int sdma_v3_0_resume(void *handle)
+static int sdma_v3_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return sdma_v3_0_hw_init(adev);
+	return sdma_v3_0_hw_init(ip_block);
 }
 
-static bool sdma_v3_0_is_idle(void *handle)
+static bool sdma_v3_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS2);
 
 	if (tmp & (SRBM_STATUS2__SDMA_BUSY_MASK |
@@ -1221,11 +1212,11 @@ static bool sdma_v3_0_is_idle(void *handle)
 	return true;
 }
 
-static int sdma_v3_0_wait_for_idle(void *handle)
+static int sdma_v3_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		tmp = RREG32(mmSRBM_STATUS2) & (SRBM_STATUS2__SDMA_BUSY_MASK |
@@ -1238,9 +1229,9 @@ static int sdma_v3_0_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static bool sdma_v3_0_check_soft_reset(void *handle)
+static bool sdma_v3_0_check_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 srbm_soft_reset = 0;
 	u32 tmp = RREG32(mmSRBM_STATUS2);
 
@@ -1259,9 +1250,9 @@ static bool sdma_v3_0_check_soft_reset(void *handle)
 	}
 }
 
-static int sdma_v3_0_pre_soft_reset(void *handle)
+static int sdma_v3_0_pre_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 srbm_soft_reset = 0;
 
 	if (!adev->sdma.srbm_soft_reset)
@@ -1278,9 +1269,9 @@ static int sdma_v3_0_pre_soft_reset(void *handle)
 	return 0;
 }
 
-static int sdma_v3_0_post_soft_reset(void *handle)
+static int sdma_v3_0_post_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 srbm_soft_reset = 0;
 
 	if (!adev->sdma.srbm_soft_reset)
@@ -1297,9 +1288,9 @@ static int sdma_v3_0_post_soft_reset(void *handle)
 	return 0;
 }
 
-static int sdma_v3_0_soft_reset(void *handle)
+static int sdma_v3_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 srbm_soft_reset = 0;
 	u32 tmp;
 
@@ -1494,10 +1485,10 @@ static void sdma_v3_0_update_sdma_medium_grain_light_sleep(
 	}
 }
 
-static int sdma_v3_0_set_clockgating_state(void *handle,
+static int sdma_v3_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
@@ -1517,15 +1508,15 @@ static int sdma_v3_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int sdma_v3_0_set_powergating_state(void *handle,
+static int sdma_v3_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
 }
 
-static void sdma_v3_0_get_clockgating_state(void *handle, u64 *flags)
+static void sdma_v3_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	if (amdgpu_sriov_vf(adev))
@@ -1545,7 +1536,6 @@ static void sdma_v3_0_get_clockgating_state(void *handle, u64 *flags)
 static const struct amd_ip_funcs sdma_v3_0_ip_funcs = {
 	.name = "sdma_v3_0",
 	.early_init = sdma_v3_0_early_init,
-	.late_init = NULL,
 	.sw_init = sdma_v3_0_sw_init,
 	.sw_fini = sdma_v3_0_sw_fini,
 	.hw_init = sdma_v3_0_hw_init,
@@ -1624,7 +1614,7 @@ static void sdma_v3_0_set_irq_funcs(struct amdgpu_device *adev)
  * @src_offset: src GPU address
  * @dst_offset: dst GPU address
  * @byte_count: number of bytes to xfer
- * @tmz: unused
+ * @copy_flags: unused
  *
  * Copy GPU buffers using the DMA engine (VI).
  * Used by the amdgpu ttm implementation to move pages if
@@ -1634,7 +1624,7 @@ static void sdma_v3_0_emit_copy_buffer(struct amdgpu_ib *ib,
 				       uint64_t src_offset,
 				       uint64_t dst_offset,
 				       uint32_t byte_count,
-				       bool tmz)
+				       uint32_t copy_flags)
 {
 	ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) |
 		SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR);
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
index b3cc04dd8653..f38004e6064e 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
@@ -72,6 +72,53 @@ MODULE_FIRMWARE("amdgpu/renoir_sdma.bin");
 MODULE_FIRMWARE("amdgpu/green_sardine_sdma.bin");
 MODULE_FIRMWARE("amdgpu/aldebaran_sdma.bin");
 
+static const struct amdgpu_hwip_reg_entry sdma_reg_list_4_0[] = {
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_STATUS_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_STATUS1_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_STATUS2_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_STATUS3_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UCODE_CHECKSUM),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RB_RPTR_FETCH_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RB_RPTR_FETCH),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UTCL1_RD_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UTCL1_WR_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UTCL1_RD_XNACK0),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UTCL1_RD_XNACK1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UTCL1_WR_XNACK0),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UTCL1_WR_XNACK1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_RB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_RB_RPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_RB_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_IB_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_IB_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_IB_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_IB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_IB_SUB_REMAIN),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_DUMMY_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_RB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_RB_RPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_RB_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_IB_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_IB_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_IB_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_DUMMY_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_RB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_RB_RPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_RB_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_IB_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_IB_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_IB_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_DUMMY_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_VM_CNTL)
+};
+
 #define SDMA0_POWER_CNTL__ON_OFF_CONDITION_HOLD_TIME_MASK  0x000000F8L
 #define SDMA0_POWER_CNTL__ON_OFF_STATUS_DURATION_TIME_MASK 0xFC000000L
 
@@ -469,7 +516,7 @@ static int sdma_v4_0_irq_id_to_seq(unsigned client_id)
 
 static void sdma_v4_0_init_golden_registers(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[SDMA0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
 	case IP_VERSION(4, 0, 0):
 		soc15_program_register_sequence(adev,
 						golden_settings_sdma_4,
@@ -539,7 +586,7 @@ static void sdma_v4_0_setup_ulv(struct amdgpu_device *adev)
 	 * The only chips with SDMAv4 and ULV are VG10 and VG20.
 	 * Server SKUs take a different hysteresis setting from other SKUs.
 	 */
-	switch (adev->ip_versions[SDMA0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
 	case IP_VERSION(4, 0, 0):
 		if (adev->pdev->device == 0x6860)
 			break;
@@ -578,8 +625,10 @@ static int sdma_v4_0_init_microcode(struct amdgpu_device *adev)
 	int ret, i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
-		if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 2, 2) ||
-                    adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 4, 0)) {
+		if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) ==
+			    IP_VERSION(4, 2, 2) ||
+		    amdgpu_ip_version(adev, SDMA0_HWIP, 0) ==
+			    IP_VERSION(4, 4, 0)) {
 			/* Acturus & Aldebaran will leverage the same FW memory
 			   for every SDMA instance */
 			ret = amdgpu_sdma_init_microcode(adev, 0, true);
@@ -875,8 +924,6 @@ static void sdma_v4_0_gfx_enable(struct amdgpu_device *adev, bool enable)
 	u32 rb_cntl, ib_cntl;
 	int i;
 
-	amdgpu_sdma_unset_buffer_funcs_helper(adev);
-
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		rb_cntl = RREG32_SDMA(i, mmSDMA0_GFX_RB_CNTL);
 		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, enable ? 1 : 0);
@@ -911,8 +958,6 @@ static void sdma_v4_0_page_stop(struct amdgpu_device *adev)
 	u32 rb_cntl, ib_cntl;
 	int i;
 
-	amdgpu_sdma_unset_buffer_funcs_helper(adev);
-
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		rb_cntl = RREG32_SDMA(i, mmSDMA0_PAGE_RB_CNTL);
 		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_PAGE_RB_CNTL,
@@ -978,7 +1023,8 @@ static void sdma_v4_0_ctx_switch_enable(struct amdgpu_device *adev, bool enable)
 		 * Arcturus for the moment and firmware version 14
 		 * and above.
 		 */
-		if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 2, 2) &&
+		if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) ==
+			    IP_VERSION(4, 2, 2) &&
 		    adev->sdma.instance[i].fw_version >= 14)
 			WREG32_SDMA(i, mmSDMA0_PUB_DUMMY_REG2, enable);
 		/* Extend page fault timeout to avoid interrupt storm */
@@ -1114,8 +1160,6 @@ static void sdma_v4_0_gfx_resume(struct amdgpu_device *adev, unsigned int i)
 #endif
 	/* enable DMA IBs */
 	WREG32_SDMA(i, mmSDMA0_GFX_IB_CNTL, ib_cntl);
-
-	ring->sched.ready = true;
 }
 
 /**
@@ -1202,8 +1246,6 @@ static void sdma_v4_0_page_resume(struct amdgpu_device *adev, unsigned int i)
 #endif
 	/* enable DMA IBs */
 	WREG32_SDMA(i, mmSDMA0_PAGE_IB_CNTL, ib_cntl);
-
-	ring->sched.ready = true;
 }
 
 static void
@@ -1259,7 +1301,7 @@ static void sdma_v4_0_init_pg(struct amdgpu_device *adev)
 	if (!(adev->pg_flags & AMD_PG_SUPPORT_SDMA))
 		return;
 
-	switch (adev->ip_versions[SDMA0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
 	case IP_VERSION(4, 1, 0):
         case IP_VERSION(4, 1, 1):
 	case IP_VERSION(4, 1, 2):
@@ -1403,13 +1445,7 @@ static int sdma_v4_0_start(struct amdgpu_device *adev)
 			r = amdgpu_ring_test_helper(page);
 			if (r)
 				return r;
-
-			if (adev->mman.buffer_funcs_ring == page)
-				amdgpu_ttm_set_buffer_funcs_status(adev, true);
 		}
-
-		if (adev->mman.buffer_funcs_ring == ring)
-			amdgpu_ttm_set_buffer_funcs_status(adev, true);
 	}
 
 	return r;
@@ -1529,7 +1565,7 @@ static int sdma_v4_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = -EINVAL;
 
 err1:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err0:
 	amdgpu_device_wb_free(adev, index);
@@ -1702,7 +1738,7 @@ static bool sdma_v4_0_fw_support_paging_queue(struct amdgpu_device *adev)
 {
 	uint fw_version = adev->sdma.instance[0].fw_version;
 
-	switch (adev->ip_versions[SDMA0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
 	case IP_VERSION(4, 0, 0):
 		return fw_version >= 430;
 	case IP_VERSION(4, 0, 1):
@@ -1715,19 +1751,17 @@ static bool sdma_v4_0_fw_support_paging_queue(struct amdgpu_device *adev)
 	}
 }
 
-static int sdma_v4_0_early_init(void *handle)
+static int sdma_v4_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = sdma_v4_0_init_microcode(adev);
-	if (r) {
-		DRM_ERROR("Failed to load sdma firmware!\n");
+	if (r)
 		return r;
-	}
 
 	/* TODO: Page queue breaks driver reload under SRIOV */
-	if ((adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 0, 0)) &&
+	if ((amdgpu_ip_version(adev, SDMA0_HWIP, 0) == IP_VERSION(4, 0, 0)) &&
 	    amdgpu_sriov_vf((adev)))
 		adev->sdma.has_page_queue = false;
 	else if (sdma_v4_0_fw_support_paging_queue(adev))
@@ -1746,26 +1780,25 @@ static int sdma_v4_0_process_ras_data_cb(struct amdgpu_device *adev,
 		void *err_data,
 		struct amdgpu_iv_entry *entry);
 
-static int sdma_v4_0_late_init(void *handle)
+static int sdma_v4_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	sdma_v4_0_setup_ulv(adev);
 
-	if (!amdgpu_persistent_edc_harvesting_supported(adev)) {
-		if (adev->sdma.ras && adev->sdma.ras->ras_block.hw_ops &&
-		    adev->sdma.ras->ras_block.hw_ops->reset_ras_error_count)
-			adev->sdma.ras->ras_block.hw_ops->reset_ras_error_count(adev);
-	}
+	if (!amdgpu_persistent_edc_harvesting_supported(adev))
+		amdgpu_ras_reset_error_count(adev, AMDGPU_RAS_BLOCK__SDMA);
 
 	return 0;
 }
 
-static int sdma_v4_0_sw_init(void *handle)
+static int sdma_v4_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
 	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_4_0);
+	uint32_t *ptr;
 
 	/* SDMA trap event */
 	for (i = 0; i < adev->sdma.num_instances; i++) {
@@ -1825,12 +1858,14 @@ static int sdma_v4_0_sw_init(void *handle)
 
 		/*
 		 * On Arcturus, SDMA instance 5~7 has a different vmhub
-		 * type(AMDGPU_MMHUB_1).
+		 * type(AMDGPU_MMHUB1).
 		 */
-		if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 2, 2) && i >= 5)
-			ring->vm_hub = AMDGPU_MMHUB_1;
+		if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) ==
+			    IP_VERSION(4, 2, 2) &&
+		    i >= 5)
+			ring->vm_hub = AMDGPU_MMHUB1(0);
 		else
-			ring->vm_hub = AMDGPU_MMHUB_0;
+			ring->vm_hub = AMDGPU_MMHUB0(0);
 
 		sprintf(ring->name, "sdma%d", i);
 		r = amdgpu_ring_init(adev, ring, 1024, &adev->sdma.trap_irq,
@@ -1847,13 +1882,27 @@ static int sdma_v4_0_sw_init(void *handle)
 			/* paging queue use same doorbell index/routing as gfx queue
 			 * with 0x400 (4096 dwords) offset on second doorbell page
 			 */
-			ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1;
-			ring->doorbell_index += 0x400;
-
-			if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 2, 2) && i >= 5)
-				ring->vm_hub = AMDGPU_MMHUB_1;
+			if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) >=
+				    IP_VERSION(4, 0, 0) &&
+			    amdgpu_ip_version(adev, SDMA0_HWIP, 0) <
+				    IP_VERSION(4, 2, 0)) {
+				ring->doorbell_index =
+					adev->doorbell_index.sdma_engine[i] << 1;
+				ring->doorbell_index += 0x400;
+			} else {
+				/* From vega20, the sdma_doorbell_range in 1st
+				 * doorbell page is reserved for page queue.
+				 */
+				ring->doorbell_index =
+					(adev->doorbell_index.sdma_engine[i] + 1) << 1;
+			}
+
+			if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) ==
+				    IP_VERSION(4, 2, 2) &&
+			    i >= 5)
+				ring->vm_hub = AMDGPU_MMHUB1(0);
 			else
-				ring->vm_hub = AMDGPU_MMHUB_0;
+				ring->vm_hub = AMDGPU_MMHUB0(0);
 
 			sprintf(ring->name, "page%d", i);
 			r = amdgpu_ring_init(adev, ring, 1024,
@@ -1870,12 +1919,19 @@ static int sdma_v4_0_sw_init(void *handle)
 		return -EINVAL;
 	}
 
+	/* Allocate memory for SDMA IP Dump buffer */
+	ptr = kcalloc(adev->sdma.num_instances * reg_count, sizeof(uint32_t), GFP_KERNEL);
+	if (ptr)
+		adev->sdma.ip_dump = ptr;
+	else
+		DRM_ERROR("Failed to allocated memory for SDMA IP Dump\n");
+
 	return r;
 }
 
-static int sdma_v4_0_sw_fini(void *handle)
+static int sdma_v4_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
@@ -1884,21 +1940,23 @@ static int sdma_v4_0_sw_fini(void *handle)
 			amdgpu_ring_fini(&adev->sdma.instance[i].page);
 	}
 
-	if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 2, 2) ||
-            adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 4, 0))
+	if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) == IP_VERSION(4, 2, 2) ||
+	    amdgpu_ip_version(adev, SDMA0_HWIP, 0) == IP_VERSION(4, 4, 0))
 		amdgpu_sdma_destroy_inst_ctx(adev, true);
 	else
 		amdgpu_sdma_destroy_inst_ctx(adev, false);
 
+	kfree(adev->sdma.ip_dump);
+
 	return 0;
 }
 
-static int sdma_v4_0_hw_init(void *handle)
+static int sdma_v4_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (adev->flags & AMD_IS_APU)
-		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_SDMA, false);
+		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_SDMA, false, 0);
 
 	if (!amdgpu_sriov_vf(adev))
 		sdma_v4_0_init_golden_registers(adev);
@@ -1906,34 +1964,33 @@ static int sdma_v4_0_hw_init(void *handle)
 	return sdma_v4_0_start(adev);
 }
 
-static int sdma_v4_0_hw_fini(void *handle)
+static int sdma_v4_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
-	if (amdgpu_sriov_vf(adev)) {
-		/* disable the scheduler for SDMA */
-		amdgpu_sdma_unset_buffer_funcs_helper(adev);
+	if (amdgpu_sriov_vf(adev))
 		return 0;
-	}
 
-	for (i = 0; i < adev->sdma.num_instances; i++) {
-		amdgpu_irq_put(adev, &adev->sdma.ecc_irq,
-			       AMDGPU_SDMA_IRQ_INSTANCE0 + i);
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__SDMA)) {
+		for (i = 0; i < adev->sdma.num_instances; i++) {
+			amdgpu_irq_put(adev, &adev->sdma.ecc_irq,
+				       AMDGPU_SDMA_IRQ_INSTANCE0 + i);
+		}
 	}
 
 	sdma_v4_0_ctx_switch_enable(adev, false);
 	sdma_v4_0_enable(adev, false);
 
 	if (adev->flags & AMD_IS_APU)
-		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_SDMA, true);
+		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_SDMA, true, 0);
 
 	return 0;
 }
 
-static int sdma_v4_0_suspend(void *handle)
+static int sdma_v4_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* SMU saves SDMA state for us */
 	if (adev->in_s0ix) {
@@ -1941,27 +1998,26 @@ static int sdma_v4_0_suspend(void *handle)
 		return 0;
 	}
 
-	return sdma_v4_0_hw_fini(adev);
+	return sdma_v4_0_hw_fini(ip_block);
 }
 
-static int sdma_v4_0_resume(void *handle)
+static int sdma_v4_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* SMU restores SDMA state for us */
 	if (adev->in_s0ix) {
 		sdma_v4_0_enable(adev, true);
 		sdma_v4_0_gfx_enable(adev, true);
-		amdgpu_ttm_set_buffer_funcs_status(adev, true);
 		return 0;
 	}
 
-	return sdma_v4_0_hw_init(adev);
+	return sdma_v4_0_hw_init(ip_block);
 }
 
-static bool sdma_v4_0_is_idle(void *handle)
+static bool sdma_v4_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
@@ -1974,11 +2030,11 @@ static bool sdma_v4_0_is_idle(void *handle)
 	return true;
 }
 
-static int sdma_v4_0_wait_for_idle(void *handle)
+static int sdma_v4_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i, j;
 	u32 sdma[AMDGPU_MAX_SDMA_INSTANCES];
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		for (j = 0; j < adev->sdma.num_instances; j++) {
@@ -1993,7 +2049,7 @@ static int sdma_v4_0_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int sdma_v4_0_soft_reset(void *handle)
+static int sdma_v4_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	/* todo */
 
@@ -2019,23 +2075,28 @@ static int sdma_v4_0_process_trap_irq(struct amdgpu_device *adev,
 				      struct amdgpu_irq_src *source,
 				      struct amdgpu_iv_entry *entry)
 {
-	uint32_t instance;
+	int instance;
 
 	DRM_DEBUG("IH: SDMA trap\n");
 	instance = sdma_v4_0_irq_id_to_seq(entry->client_id);
+	if (instance < 0)
+		return instance;
+
 	switch (entry->ring_id) {
 	case 0:
 		amdgpu_fence_process(&adev->sdma.instance[instance].ring);
 		break;
 	case 1:
-		if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 2, 0))
+		if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) ==
+		    IP_VERSION(4, 2, 0))
 			amdgpu_fence_process(&adev->sdma.instance[instance].page);
 		break;
 	case 2:
 		/* XXX compute */
 		break;
 	case 3:
-		if (adev->ip_versions[SDMA0_HWIP][0] != IP_VERSION(4, 2, 0))
+		if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) !=
+		    IP_VERSION(4, 2, 0))
 			amdgpu_fence_process(&adev->sdma.instance[instance].page);
 		break;
 	}
@@ -2104,7 +2165,7 @@ static int sdma_v4_0_print_iv_entry(struct amdgpu_device *adev,
 					      struct amdgpu_iv_entry *entry)
 {
 	int instance;
-	struct amdgpu_task_info task_info;
+	struct amdgpu_task_info *task_info;
 	u64 addr;
 
 	instance = sdma_v4_0_irq_id_to_seq(entry->client_id);
@@ -2116,15 +2177,20 @@ static int sdma_v4_0_print_iv_entry(struct amdgpu_device *adev,
 	addr = (u64)entry->src_data[0] << 12;
 	addr |= ((u64)entry->src_data[1] & 0xf) << 44;
 
-	memset(&task_info, 0, sizeof(struct amdgpu_task_info));
-	amdgpu_vm_get_task_info(adev, entry->pasid, &task_info);
-
 	dev_dbg_ratelimited(adev->dev,
-		   "[sdma%d] address:0x%016llx src_id:%u ring:%u vmid:%u "
-		   "pasid:%u, for process %s pid %d thread %s pid %d\n",
-		   instance, addr, entry->src_id, entry->ring_id, entry->vmid,
-		   entry->pasid, task_info.process_name, task_info.tgid,
-		   task_info.task_name, task_info.pid);
+			   "[sdma%d] address:0x%016llx src_id:%u ring:%u vmid:%u pasid:%u\n",
+			   instance, addr, entry->src_id, entry->ring_id, entry->vmid,
+			   entry->pasid);
+
+	task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid);
+	if (task_info) {
+		dev_dbg_ratelimited(adev->dev,
+				    " for process %s pid %d thread %s pid %d\n",
+				    task_info->process_name, task_info->tgid,
+				    task_info->task.comm, task_info->task.pid);
+		amdgpu_vm_put_task_info(task_info);
+	}
+
 	return 0;
 }
 
@@ -2231,10 +2297,10 @@ static void sdma_v4_0_update_medium_grain_light_sleep(
 	}
 }
 
-static int sdma_v4_0_set_clockgating_state(void *handle,
+static int sdma_v4_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
@@ -2246,12 +2312,12 @@ static int sdma_v4_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int sdma_v4_0_set_powergating_state(void *handle,
+static int sdma_v4_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	switch (adev->ip_versions[SDMA0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
 	case IP_VERSION(4, 1, 0):
 	case IP_VERSION(4, 1, 1):
 	case IP_VERSION(4, 1, 2):
@@ -2265,9 +2331,9 @@ static int sdma_v4_0_set_powergating_state(void *handle,
 	return 0;
 }
 
-static void sdma_v4_0_get_clockgating_state(void *handle, u64 *flags)
+static void sdma_v4_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	if (amdgpu_sriov_vf(adev))
@@ -2284,6 +2350,46 @@ static void sdma_v4_0_get_clockgating_state(void *handle, u64 *flags)
 		*flags |= AMD_CG_SUPPORT_SDMA_LS;
 }
 
+static void sdma_v4_0_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, j;
+	uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_4_0);
+	uint32_t instance_offset;
+
+	if (!adev->sdma.ip_dump)
+		return;
+
+	drm_printf(p, "num_instances:%d\n", adev->sdma.num_instances);
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		instance_offset = i * reg_count;
+		drm_printf(p, "\nInstance:%d\n", i);
+
+		for (j = 0; j < reg_count; j++)
+			drm_printf(p, "%-50s \t 0x%08x\n", sdma_reg_list_4_0[j].reg_name,
+				   adev->sdma.ip_dump[instance_offset + j]);
+	}
+}
+
+static void sdma_v4_0_dump_ip_state(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, j;
+	uint32_t instance_offset;
+	uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_4_0);
+
+	if (!adev->sdma.ip_dump)
+		return;
+
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		instance_offset = i * reg_count;
+		for (j = 0; j < reg_count; j++)
+			adev->sdma.ip_dump[instance_offset + j] =
+				RREG32(sdma_v4_0_get_reg_offset(adev, i,
+				       sdma_reg_list_4_0[j].reg_offset));
+	}
+}
+
 const struct amd_ip_funcs sdma_v4_0_ip_funcs = {
 	.name = "sdma_v4_0",
 	.early_init = sdma_v4_0_early_init,
@@ -2300,11 +2406,13 @@ const struct amd_ip_funcs sdma_v4_0_ip_funcs = {
 	.set_clockgating_state = sdma_v4_0_set_clockgating_state,
 	.set_powergating_state = sdma_v4_0_set_powergating_state,
 	.get_clockgating_state = sdma_v4_0_get_clockgating_state,
+	.dump_ip_state = sdma_v4_0_dump_ip_state,
+	.print_ip_state = sdma_v4_0_print_ip_state,
 };
 
 static const struct amdgpu_ring_funcs sdma_v4_0_ring_funcs = {
 	.type = AMDGPU_RING_TYPE_SDMA,
-	.align_mask = 0xf,
+	.align_mask = 0xff,
 	.nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP),
 	.support_64bit_ptrs = true,
 	.secure_submission_supported = true,
@@ -2336,7 +2444,7 @@ static const struct amdgpu_ring_funcs sdma_v4_0_ring_funcs = {
 
 static const struct amdgpu_ring_funcs sdma_v4_0_page_ring_funcs = {
 	.type = AMDGPU_RING_TYPE_SDMA,
-	.align_mask = 0xf,
+	.align_mask = 0xff,
 	.nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP),
 	.support_64bit_ptrs = true,
 	.secure_submission_supported = true,
@@ -2443,7 +2551,7 @@ static void sdma_v4_0_set_irq_funcs(struct amdgpu_device *adev)
  * @src_offset: src GPU address
  * @dst_offset: dst GPU address
  * @byte_count: number of bytes to xfer
- * @tmz: if a secure copy should be used
+ * @copy_flags: copy flags for the buffers
  *
  * Copy GPU buffers using the DMA engine (VEGA10/12).
  * Used by the amdgpu ttm implementation to move pages if
@@ -2453,11 +2561,11 @@ static void sdma_v4_0_emit_copy_buffer(struct amdgpu_ib *ib,
 				       uint64_t src_offset,
 				       uint64_t dst_offset,
 				       uint32_t byte_count,
-				       bool tmz)
+				       uint32_t copy_flags)
 {
 	ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) |
 		SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR) |
-		SDMA_PKT_COPY_LINEAR_HEADER_TMZ(tmz ? 1 : 0);
+		SDMA_PKT_COPY_LINEAR_HEADER_TMZ((copy_flags & AMDGPU_COPY_FLAGS_TMZ) ? 1 : 0);
 	ib->ptr[ib->length_dw++] = byte_count - 1;
 	ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */
 	ib->ptr[ib->length_dw++] = lower_32_bits(src_offset);
@@ -2614,7 +2722,7 @@ static struct amdgpu_sdma_ras sdma_v4_0_ras = {
 
 static void sdma_v4_0_set_ras_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[SDMA0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
 	case IP_VERSION(4, 2, 0):
 	case IP_VERSION(4, 2, 2):
 		adev->sdma.ras = &sdma_v4_0_ras;
@@ -2625,7 +2733,6 @@ static void sdma_v4_0_set_ras_funcs(struct amdgpu_device *adev)
 	default:
 		break;
 	}
-
 }
 
 const struct amdgpu_ip_block_version sdma_v4_0_ip_block = {
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.c
index 6f9895cdddb1..0ddb6955a6d3 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.c
@@ -141,6 +141,10 @@ static const struct soc15_ras_field_entry sdma_v4_4_ras_fields[] = {
 	SOC15_REG_FIELD(SDMA0_EDC_COUNTER2, SDMA_UTCL1_RDBST_FIFO_SED),
 	0, 0,
 	},
+	{ "SDMA_UTCL1_WR_FIFO_SED", SOC15_REG_ENTRY(SDMA0, 0, regSDMA0_EDC_COUNTER2),
+	SOC15_REG_FIELD(SDMA0_EDC_COUNTER2, SDMA_UTCL1_WR_FIFO_SED),
+	0, 0,
+	},
 	{ "SDMA_DATA_LUT_FIFO_SED", SOC15_REG_ENTRY(SDMA0, 0, regSDMA0_EDC_COUNTER2),
 	SOC15_REG_FIELD(SDMA0_EDC_COUNTER2, SDMA_DATA_LUT_FIFO_SED),
 	0, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 64dcaa2670dd..a1443990d5c6 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -27,8 +27,10 @@
 #include <linux/pci.h>
 
 #include "amdgpu.h"
+#include "amdgpu_xcp.h"
 #include "amdgpu_ucode.h"
 #include "amdgpu_trace.h"
+#include "amdgpu_reset.h"
 
 #include "sdma/sdma_4_4_2_offset.h"
 #include "sdma/sdma_4_4_2_sh_mask.h"
@@ -43,6 +45,57 @@
 #include "amdgpu_ras.h"
 
 MODULE_FIRMWARE("amdgpu/sdma_4_4_2.bin");
+MODULE_FIRMWARE("amdgpu/sdma_4_4_4.bin");
+MODULE_FIRMWARE("amdgpu/sdma_4_4_5.bin");
+
+static const struct amdgpu_hwip_reg_entry sdma_reg_list_4_4_2[] = {
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_STATUS_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_STATUS1_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_STATUS2_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_STATUS3_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_UCODE_CHECKSUM),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RB_RPTR_FETCH_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RB_RPTR_FETCH),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_UTCL1_RD_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_UTCL1_WR_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_UTCL1_RD_XNACK0),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_UTCL1_RD_XNACK1),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_UTCL1_WR_XNACK0),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_UTCL1_WR_XNACK1),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_RB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_RB_RPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_RB_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_IB_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_IB_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_IB_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_IB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_IB_SUB_REMAIN),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_DUMMY_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_PAGE_RB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_PAGE_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_PAGE_RB_RPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_PAGE_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_PAGE_RB_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_PAGE_IB_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_PAGE_IB_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_PAGE_IB_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_PAGE_DUMMY_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RLC0_RB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RLC0_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RLC0_RB_RPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RLC0_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RLC0_RB_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RLC0_IB_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RLC0_IB_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RLC0_IB_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RLC0_DUMMY_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA_VM_CNTL)
+};
+
+#define mmSMNAID_AID0_MCA_SMU 0x03b30400
 
 #define WREG32_SDMA(instance, offset, value) \
 	WREG32(sdma_v4_4_2_get_reg_offset(adev, (instance), (offset)), value)
@@ -53,11 +106,19 @@ static void sdma_v4_4_2_set_ring_funcs(struct amdgpu_device *adev);
 static void sdma_v4_4_2_set_buffer_funcs(struct amdgpu_device *adev);
 static void sdma_v4_4_2_set_vm_pte_funcs(struct amdgpu_device *adev);
 static void sdma_v4_4_2_set_irq_funcs(struct amdgpu_device *adev);
+static void sdma_v4_4_2_set_ras_funcs(struct amdgpu_device *adev);
+static void sdma_v4_4_2_update_reset_mask(struct amdgpu_device *adev);
+static int sdma_v4_4_2_stop_queue(struct amdgpu_ring *ring);
+static int sdma_v4_4_2_restore_queue(struct amdgpu_ring *ring);
+static int sdma_v4_4_2_soft_reset_engine(struct amdgpu_device *adev,
+					 u32 instance_id);
 
 static u32 sdma_v4_4_2_get_reg_offset(struct amdgpu_device *adev,
 		u32 instance, u32 offset)
 {
-	return (adev->reg_offset[SDMA0_HWIP][instance][0] + offset);
+	u32 dev_inst = GET_INST(SDMA0, instance);
+
+	return (adev->reg_offset[SDMA0_HWIP][dev_inst][0] + offset);
 }
 
 static unsigned sdma_v4_4_2_seq_to_irq_id(int seq_num)
@@ -76,7 +137,7 @@ static unsigned sdma_v4_4_2_seq_to_irq_id(int seq_num)
 	}
 }
 
-static int sdma_v4_4_2_irq_id_to_seq(unsigned client_id)
+static int sdma_v4_4_2_irq_id_to_seq(struct amdgpu_device *adev, unsigned client_id)
 {
 	switch (client_id) {
 	case SOC15_IH_CLIENTID_SDMA0:
@@ -84,21 +145,39 @@ static int sdma_v4_4_2_irq_id_to_seq(unsigned client_id)
 	case SOC15_IH_CLIENTID_SDMA1:
 		return 1;
 	case SOC15_IH_CLIENTID_SDMA2:
-		return 2;
+		if (amdgpu_sriov_vf(adev) && (adev->gfx.xcc_mask == 0x1))
+			return 0;
+		else
+			return 2;
 	case SOC15_IH_CLIENTID_SDMA3:
-		return 3;
+		if (amdgpu_sriov_vf(adev) && (adev->gfx.xcc_mask == 0x1))
+			return 1;
+		else
+			return 3;
 	default:
 		return -EINVAL;
 	}
 }
 
-static void sdma_v4_4_2_init_golden_registers(struct amdgpu_device *adev)
+static void sdma_v4_4_2_inst_init_golden_registers(struct amdgpu_device *adev,
+						   uint32_t inst_mask)
 {
-	switch (adev->ip_versions[SDMA0_HWIP][0]) {
-	case IP_VERSION(4, 4, 2):
-		break;
-	default:
-		break;
+	u32 val;
+	int i;
+
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		val = RREG32_SDMA(i, regSDMA_GB_ADDR_CONFIG);
+		val = REG_SET_FIELD(val, SDMA_GB_ADDR_CONFIG, NUM_BANKS, 4);
+		val = REG_SET_FIELD(val, SDMA_GB_ADDR_CONFIG,
+				    PIPE_INTERLEAVE_SIZE, 0);
+		WREG32_SDMA(i, regSDMA_GB_ADDR_CONFIG, val);
+
+		val = RREG32_SDMA(i, regSDMA_GB_ADDR_CONFIG_READ);
+		val = REG_SET_FIELD(val, SDMA_GB_ADDR_CONFIG_READ, NUM_BANKS,
+				    4);
+		val = REG_SET_FIELD(val, SDMA_GB_ADDR_CONFIG_READ,
+				    PIPE_INTERLEAVE_SIZE, 0);
+		WREG32_SDMA(i, regSDMA_GB_ADDR_CONFIG_READ, val);
 	}
 }
 
@@ -116,7 +195,9 @@ static int sdma_v4_4_2_init_microcode(struct amdgpu_device *adev)
 	int ret, i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
-		if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 4, 2)) {
+		if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) == IP_VERSION(4, 4, 2) ||
+		    amdgpu_ip_version(adev, SDMA0_HWIP, 0) == IP_VERSION(4, 4, 4) ||
+		    amdgpu_ip_version(adev, SDMA0_HWIP, 0) == IP_VERSION(4, 4, 5)) {
 			ret = amdgpu_sdma_init_microcode(adev, 0, true);
 			break;
 		} else {
@@ -138,13 +219,13 @@ static int sdma_v4_4_2_init_microcode(struct amdgpu_device *adev)
  */
 static uint64_t sdma_v4_4_2_ring_get_rptr(struct amdgpu_ring *ring)
 {
-	u64 *rptr;
+	u64 rptr;
 
 	/* XXX check if swapping is necessary on BE */
-	rptr = ((u64 *)&ring->adev->wb.wb[ring->rptr_offs]);
+	rptr = READ_ONCE(*((u64 *)&ring->adev->wb.wb[ring->rptr_offs]));
 
-	DRM_DEBUG("rptr before shift == 0x%016llx\n", *rptr);
-	return ((*rptr) >> 2);
+	DRM_DEBUG("rptr before shift == 0x%016llx\n", rptr);
+	return rptr >> 2;
 }
 
 /**
@@ -349,7 +430,8 @@ static void sdma_v4_4_2_ring_emit_hdp_flush(struct amdgpu_ring *ring)
 	u32 ref_and_mask = 0;
 	const struct nbio_hdp_flush_reg *nbio_hf_reg = adev->nbio.hdp_flush_reg;
 
-	ref_and_mask = nbio_hf_reg->ref_and_mask_sdma0 << ring->me;
+	ref_and_mask = nbio_hf_reg->ref_and_mask_sdma0
+		       << (ring->me % adev->sdma.num_inst_per_aid);
 
 	sdma_v4_4_2_wait_reg_mem(ring, 0, 1,
 			       adev->nbio.funcs->get_hdp_flush_done_offset(adev),
@@ -399,70 +481,77 @@ static void sdma_v4_4_2_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64
 
 
 /**
- * sdma_v4_4_2_gfx_stop - stop the gfx async dma engines
+ * sdma_v4_4_2_inst_gfx_stop - stop the gfx async dma engines
  *
  * @adev: amdgpu_device pointer
+ * @inst_mask: mask of dma engine instances to be disabled
  *
  * Stop the gfx async dma ring buffers.
  */
-static void sdma_v4_4_2_gfx_stop(struct amdgpu_device *adev)
+static void sdma_v4_4_2_inst_gfx_stop(struct amdgpu_device *adev,
+				      uint32_t inst_mask)
 {
 	struct amdgpu_ring *sdma[AMDGPU_MAX_SDMA_INSTANCES];
-	u32 rb_cntl, ib_cntl;
-	int i, unset = 0;
+	u32 doorbell_offset, doorbell;
+	u32 rb_cntl, ib_cntl, sdma_cntl;
+	int i;
 
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	for_each_inst(i, inst_mask) {
 		sdma[i] = &adev->sdma.instance[i].ring;
 
-		if ((adev->mman.buffer_funcs_ring == sdma[i]) && unset != 1) {
-			amdgpu_ttm_set_buffer_funcs_status(adev, false);
-			unset = 1;
-		}
-
 		rb_cntl = RREG32_SDMA(i, regSDMA_GFX_RB_CNTL);
 		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA_GFX_RB_CNTL, RB_ENABLE, 0);
 		WREG32_SDMA(i, regSDMA_GFX_RB_CNTL, rb_cntl);
 		ib_cntl = RREG32_SDMA(i, regSDMA_GFX_IB_CNTL);
 		ib_cntl = REG_SET_FIELD(ib_cntl, SDMA_GFX_IB_CNTL, IB_ENABLE, 0);
 		WREG32_SDMA(i, regSDMA_GFX_IB_CNTL, ib_cntl);
+		sdma_cntl = RREG32_SDMA(i, regSDMA_CNTL);
+		sdma_cntl = REG_SET_FIELD(sdma_cntl, SDMA_CNTL, UTC_L1_ENABLE, 0);
+		WREG32_SDMA(i, regSDMA_CNTL, sdma_cntl);
+
+		if (sdma[i]->use_doorbell) {
+			doorbell = RREG32_SDMA(i, regSDMA_GFX_DOORBELL);
+			doorbell_offset = RREG32_SDMA(i, regSDMA_GFX_DOORBELL_OFFSET);
+
+			doorbell = REG_SET_FIELD(doorbell, SDMA_GFX_DOORBELL, ENABLE, 0);
+			doorbell_offset = REG_SET_FIELD(doorbell_offset,
+					SDMA_GFX_DOORBELL_OFFSET,
+					OFFSET, 0);
+			WREG32_SDMA(i, regSDMA_GFX_DOORBELL, doorbell);
+			WREG32_SDMA(i, regSDMA_GFX_DOORBELL_OFFSET, doorbell_offset);
+		}
 	}
 }
 
 /**
- * sdma_v4_4_2_rlc_stop - stop the compute async dma engines
+ * sdma_v4_4_2_inst_rlc_stop - stop the compute async dma engines
  *
  * @adev: amdgpu_device pointer
+ * @inst_mask: mask of dma engine instances to be disabled
  *
  * Stop the compute async dma queues.
  */
-static void sdma_v4_4_2_rlc_stop(struct amdgpu_device *adev)
+static void sdma_v4_4_2_inst_rlc_stop(struct amdgpu_device *adev,
+				      uint32_t inst_mask)
 {
 	/* XXX todo */
 }
 
 /**
- * sdma_v4_4_2_page_stop - stop the page async dma engines
+ * sdma_v4_4_2_inst_page_stop - stop the page async dma engines
  *
  * @adev: amdgpu_device pointer
+ * @inst_mask: mask of dma engine instances to be disabled
  *
  * Stop the page async dma ring buffers.
  */
-static void sdma_v4_4_2_page_stop(struct amdgpu_device *adev)
+static void sdma_v4_4_2_inst_page_stop(struct amdgpu_device *adev,
+				       uint32_t inst_mask)
 {
-	struct amdgpu_ring *sdma[AMDGPU_MAX_SDMA_INSTANCES];
 	u32 rb_cntl, ib_cntl;
 	int i;
-	bool unset = false;
-
-	for (i = 0; i < adev->sdma.num_instances; i++) {
-		sdma[i] = &adev->sdma.instance[i].page;
-
-		if ((adev->mman.buffer_funcs_ring == sdma[i]) &&
-			(!unset)) {
-			amdgpu_ttm_set_buffer_funcs_status(adev, false);
-			unset = true;
-		}
 
+	for_each_inst(i, inst_mask) {
 		rb_cntl = RREG32_SDMA(i, regSDMA_PAGE_RB_CNTL);
 		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA_PAGE_RB_CNTL,
 					RB_ENABLE, 0);
@@ -475,14 +564,16 @@ static void sdma_v4_4_2_page_stop(struct amdgpu_device *adev)
 }
 
 /**
- * sdma_v4_4_2_ctx_switch_enable - stop the async dma engines context switch
+ * sdma_v4_4_2_inst_ctx_switch_enable - stop the async dma engines context switch
  *
  * @adev: amdgpu_device pointer
  * @enable: enable/disable the DMA MEs context switch.
+ * @inst_mask: mask of dma engine instances to be enabled
  *
  * Halt or unhalt the async dma engines context switch.
  */
-static void sdma_v4_4_2_ctx_switch_enable(struct amdgpu_device *adev, bool enable)
+static void sdma_v4_4_2_inst_ctx_switch_enable(struct amdgpu_device *adev,
+					       bool enable, uint32_t inst_mask)
 {
 	u32 f32_cntl, phase_quantum = 0;
 	int i;
@@ -511,7 +602,7 @@ static void sdma_v4_4_2_ctx_switch_enable(struct amdgpu_device *adev, bool enabl
 			unit  << SDMA_PHASE0_QUANTUM__UNIT__SHIFT;
 	}
 
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	for_each_inst(i, inst_mask) {
 		f32_cntl = RREG32_SDMA(i, regSDMA_CNTL);
 		f32_cntl = REG_SET_FIELD(f32_cntl, SDMA_CNTL,
 				AUTO_CTXSW_ENABLE, enable ? 1 : 0);
@@ -525,30 +616,39 @@ static void sdma_v4_4_2_ctx_switch_enable(struct amdgpu_device *adev, bool enabl
 		/* Extend page fault timeout to avoid interrupt storm */
 		WREG32_SDMA(i, regSDMA_UTCL1_TIMEOUT, 0x00800080);
 	}
-
 }
 
 /**
- * sdma_v4_4_2_enable - stop the async dma engines
+ * sdma_v4_4_2_inst_enable - stop the async dma engines
  *
  * @adev: amdgpu_device pointer
  * @enable: enable/disable the DMA MEs.
+ * @inst_mask: mask of dma engine instances to be enabled
  *
  * Halt or unhalt the async dma engines.
  */
-static void sdma_v4_4_2_enable(struct amdgpu_device *adev, bool enable)
+static void sdma_v4_4_2_inst_enable(struct amdgpu_device *adev, bool enable,
+				    uint32_t inst_mask)
 {
 	u32 f32_cntl;
 	int i;
 
 	if (!enable) {
-		sdma_v4_4_2_gfx_stop(adev);
-		sdma_v4_4_2_rlc_stop(adev);
+		sdma_v4_4_2_inst_gfx_stop(adev, inst_mask);
+		sdma_v4_4_2_inst_rlc_stop(adev, inst_mask);
 		if (adev->sdma.has_page_queue)
-			sdma_v4_4_2_page_stop(adev);
+			sdma_v4_4_2_inst_page_stop(adev, inst_mask);
+
+		/* SDMA FW needs to respond to FREEZE requests during reset.
+		 * Keep it running during reset */
+		if (!amdgpu_sriov_vf(adev) && amdgpu_in_reset(adev))
+			return;
 	}
 
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP)
+		return;
+
+	for_each_inst(i, inst_mask) {
 		f32_cntl = RREG32_SDMA(i, regSDMA_F32_CNTL);
 		f32_cntl = REG_SET_FIELD(f32_cntl, SDMA_F32_CNTL, HALT, enable ? 0 : 1);
 		WREG32_SDMA(i, regSDMA_F32_CNTL, f32_cntl);
@@ -563,7 +663,7 @@ static uint32_t sdma_v4_4_2_rb_cntl(struct amdgpu_ring *ring, uint32_t rb_cntl)
 	/* Set ring buffer size in dwords */
 	uint32_t rb_bufsz = order_base_2(ring->ring_size / 4);
 
-	barrier(); /* work around https://bugs.llvm.org/show_bug.cgi?id=42576 */
+	barrier(); /* work around https://llvm.org/pr42576 */
 	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA_GFX_RB_CNTL, RB_SIZE, rb_bufsz);
 #ifdef __BIG_ENDIAN
 	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA_GFX_RB_CNTL, RB_SWAP_ENABLE, 1);
@@ -578,11 +678,12 @@ static uint32_t sdma_v4_4_2_rb_cntl(struct amdgpu_ring *ring, uint32_t rb_cntl)
  *
  * @adev: amdgpu_device pointer
  * @i: instance to resume
+ * @restore: used to restore wptr when restart
  *
  * Set up the gfx DMA ring buffers and enable them.
  * Returns 0 for success, error for failure.
  */
-static void sdma_v4_4_2_gfx_resume(struct amdgpu_device *adev, unsigned int i)
+static void sdma_v4_4_2_gfx_resume(struct amdgpu_device *adev, unsigned int i, bool restore)
 {
 	struct amdgpu_ring *ring = &adev->sdma.instance[i].ring;
 	u32 rb_cntl, ib_cntl, wptr_poll_cntl;
@@ -590,6 +691,7 @@ static void sdma_v4_4_2_gfx_resume(struct amdgpu_device *adev, unsigned int i)
 	u32 doorbell;
 	u32 doorbell_offset;
 	u64 wptr_gpu_addr;
+	u64 rwptr;
 
 	wb_offset = (ring->rptr_offs * 4);
 
@@ -597,12 +699,6 @@ static void sdma_v4_4_2_gfx_resume(struct amdgpu_device *adev, unsigned int i)
 	rb_cntl = sdma_v4_4_2_rb_cntl(ring, rb_cntl);
 	WREG32_SDMA(i, regSDMA_GFX_RB_CNTL, rb_cntl);
 
-	/* Initialize the ring buffer's read and write pointers */
-	WREG32_SDMA(i, regSDMA_GFX_RB_RPTR, 0);
-	WREG32_SDMA(i, regSDMA_GFX_RB_RPTR_HI, 0);
-	WREG32_SDMA(i, regSDMA_GFX_RB_WPTR, 0);
-	WREG32_SDMA(i, regSDMA_GFX_RB_WPTR_HI, 0);
-
 	/* set the wb address whether it's enabled or not */
 	WREG32_SDMA(i, regSDMA_GFX_RB_RPTR_ADDR_HI,
 	       upper_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF);
@@ -615,11 +711,33 @@ static void sdma_v4_4_2_gfx_resume(struct amdgpu_device *adev, unsigned int i)
 	WREG32_SDMA(i, regSDMA_GFX_RB_BASE, ring->gpu_addr >> 8);
 	WREG32_SDMA(i, regSDMA_GFX_RB_BASE_HI, ring->gpu_addr >> 40);
 
-	ring->wptr = 0;
+	if (!restore)
+		ring->wptr = 0;
 
 	/* before programing wptr to a less value, need set minor_ptr_update first */
 	WREG32_SDMA(i, regSDMA_GFX_MINOR_PTR_UPDATE, 1);
 
+	/* For the guilty queue, set RPTR to the current wptr to skip bad commands,
+	 * It is not a guilty queue, restore cache_rptr and continue execution.
+	 */
+	if (adev->sdma.instance[i].gfx_guilty)
+		rwptr = ring->wptr;
+	else
+		rwptr = ring->cached_rptr;
+
+	/* Initialize the ring buffer's read and write pointers */
+	if (restore) {
+		WREG32_SDMA(i, regSDMA_GFX_RB_RPTR, lower_32_bits(rwptr << 2));
+		WREG32_SDMA(i, regSDMA_GFX_RB_RPTR_HI, upper_32_bits(rwptr << 2));
+		WREG32_SDMA(i, regSDMA_GFX_RB_WPTR, lower_32_bits(rwptr << 2));
+		WREG32_SDMA(i, regSDMA_GFX_RB_WPTR_HI, upper_32_bits(rwptr << 2));
+	} else {
+		WREG32_SDMA(i, regSDMA_GFX_RB_RPTR, 0);
+		WREG32_SDMA(i, regSDMA_GFX_RB_RPTR_HI, 0);
+		WREG32_SDMA(i, regSDMA_GFX_RB_WPTR, 0);
+		WREG32_SDMA(i, regSDMA_GFX_RB_WPTR_HI, 0);
+	}
+
 	doorbell = RREG32_SDMA(i, regSDMA_GFX_DOORBELL);
 	doorbell_offset = RREG32_SDMA(i, regSDMA_GFX_DOORBELL_OFFSET);
 
@@ -659,8 +777,6 @@ static void sdma_v4_4_2_gfx_resume(struct amdgpu_device *adev, unsigned int i)
 #endif
 	/* enable DMA IBs */
 	WREG32_SDMA(i, regSDMA_GFX_IB_CNTL, ib_cntl);
-
-	ring->sched.ready = true;
 }
 
 /**
@@ -668,11 +784,12 @@ static void sdma_v4_4_2_gfx_resume(struct amdgpu_device *adev, unsigned int i)
  *
  * @adev: amdgpu_device pointer
  * @i: instance to resume
+ * @restore: boolean to say restore needed or not
  *
  * Set up the page DMA ring buffers and enable them.
  * Returns 0 for success, error for failure.
  */
-static void sdma_v4_4_2_page_resume(struct amdgpu_device *adev, unsigned int i)
+static void sdma_v4_4_2_page_resume(struct amdgpu_device *adev, unsigned int i, bool restore)
 {
 	struct amdgpu_ring *ring = &adev->sdma.instance[i].page;
 	u32 rb_cntl, ib_cntl, wptr_poll_cntl;
@@ -680,6 +797,7 @@ static void sdma_v4_4_2_page_resume(struct amdgpu_device *adev, unsigned int i)
 	u32 doorbell;
 	u32 doorbell_offset;
 	u64 wptr_gpu_addr;
+	u64 rwptr;
 
 	wb_offset = (ring->rptr_offs * 4);
 
@@ -687,11 +805,26 @@ static void sdma_v4_4_2_page_resume(struct amdgpu_device *adev, unsigned int i)
 	rb_cntl = sdma_v4_4_2_rb_cntl(ring, rb_cntl);
 	WREG32_SDMA(i, regSDMA_PAGE_RB_CNTL, rb_cntl);
 
+	/* For the guilty queue, set RPTR to the current wptr to skip bad commands,
+	 * It is not a guilty queue, restore cache_rptr and continue execution.
+	 */
+	if (adev->sdma.instance[i].page_guilty)
+		rwptr = ring->wptr;
+	else
+		rwptr = ring->cached_rptr;
+
 	/* Initialize the ring buffer's read and write pointers */
-	WREG32_SDMA(i, regSDMA_PAGE_RB_RPTR, 0);
-	WREG32_SDMA(i, regSDMA_PAGE_RB_RPTR_HI, 0);
-	WREG32_SDMA(i, regSDMA_PAGE_RB_WPTR, 0);
-	WREG32_SDMA(i, regSDMA_PAGE_RB_WPTR_HI, 0);
+	if (restore) {
+		WREG32_SDMA(i, regSDMA_PAGE_RB_RPTR, lower_32_bits(rwptr << 2));
+		WREG32_SDMA(i, regSDMA_PAGE_RB_RPTR_HI, upper_32_bits(rwptr << 2));
+		WREG32_SDMA(i, regSDMA_PAGE_RB_WPTR, lower_32_bits(rwptr << 2));
+		WREG32_SDMA(i, regSDMA_PAGE_RB_WPTR_HI, upper_32_bits(rwptr << 2));
+	} else {
+		WREG32_SDMA(i, regSDMA_PAGE_RB_RPTR, 0);
+		WREG32_SDMA(i, regSDMA_PAGE_RB_RPTR_HI, 0);
+		WREG32_SDMA(i, regSDMA_PAGE_RB_WPTR, 0);
+		WREG32_SDMA(i, regSDMA_PAGE_RB_WPTR_HI, 0);
+	}
 
 	/* set the wb address whether it's enabled or not */
 	WREG32_SDMA(i, regSDMA_PAGE_RB_RPTR_ADDR_HI,
@@ -705,7 +838,8 @@ static void sdma_v4_4_2_page_resume(struct amdgpu_device *adev, unsigned int i)
 	WREG32_SDMA(i, regSDMA_PAGE_RB_BASE, ring->gpu_addr >> 8);
 	WREG32_SDMA(i, regSDMA_PAGE_RB_BASE_HI, ring->gpu_addr >> 40);
 
-	ring->wptr = 0;
+	if (!restore)
+		ring->wptr = 0;
 
 	/* before programing wptr to a less value, need set minor_ptr_update first */
 	WREG32_SDMA(i, regSDMA_PAGE_MINOR_PTR_UPDATE, 1);
@@ -750,8 +884,6 @@ static void sdma_v4_4_2_page_resume(struct amdgpu_device *adev, unsigned int i)
 #endif
 	/* enable DMA IBs */
 	WREG32_SDMA(i, regSDMA_PAGE_IB_CNTL, ib_cntl);
-
-	ring->sched.ready = true;
 }
 
 static void sdma_v4_4_2_init_pg(struct amdgpu_device *adev)
@@ -760,14 +892,16 @@ static void sdma_v4_4_2_init_pg(struct amdgpu_device *adev)
 }
 
 /**
- * sdma_v4_4_2_rlc_resume - setup and start the async dma engines
+ * sdma_v4_4_2_inst_rlc_resume - setup and start the async dma engines
  *
  * @adev: amdgpu_device pointer
+ * @inst_mask: mask of dma engine instances to be enabled
  *
  * Set up the compute DMA queues and enable them.
  * Returns 0 for success, error for failure.
  */
-static int sdma_v4_4_2_rlc_resume(struct amdgpu_device *adev)
+static int sdma_v4_4_2_inst_rlc_resume(struct amdgpu_device *adev,
+				       uint32_t inst_mask)
 {
 	sdma_v4_4_2_init_pg(adev);
 
@@ -775,14 +909,16 @@ static int sdma_v4_4_2_rlc_resume(struct amdgpu_device *adev)
 }
 
 /**
- * sdma_v4_4_2_load_microcode - load the sDMA ME ucode
+ * sdma_v4_4_2_inst_load_microcode - load the sDMA ME ucode
  *
  * @adev: amdgpu_device pointer
+ * @inst_mask: mask of dma engine instances to be enabled
  *
  * Loads the sDMA0/1 ucode.
  * Returns 0 for success, -EINVAL if the ucode is not available.
  */
-static int sdma_v4_4_2_load_microcode(struct amdgpu_device *adev)
+static int sdma_v4_4_2_inst_load_microcode(struct amdgpu_device *adev,
+					   uint32_t inst_mask)
 {
 	const struct sdma_firmware_header_v1_0 *hdr;
 	const __le32 *fw_data;
@@ -790,9 +926,9 @@ static int sdma_v4_4_2_load_microcode(struct amdgpu_device *adev)
 	int i, j;
 
 	/* halt the MEs */
-	sdma_v4_4_2_enable(adev, false);
+	sdma_v4_4_2_inst_enable(adev, false, inst_mask);
 
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	for_each_inst(i, inst_mask) {
 		if (!adev->sdma.instance[i].fw)
 			return -EINVAL;
 
@@ -818,73 +954,81 @@ static int sdma_v4_4_2_load_microcode(struct amdgpu_device *adev)
 }
 
 /**
- * sdma_v4_4_2_start - setup and start the async dma engines
+ * sdma_v4_4_2_inst_start - setup and start the async dma engines
  *
  * @adev: amdgpu_device pointer
+ * @inst_mask: mask of dma engine instances to be enabled
+ * @restore: boolean to say restore needed or not
  *
  * Set up the DMA engines and enable them.
  * Returns 0 for success, error for failure.
  */
-static int sdma_v4_4_2_start(struct amdgpu_device *adev)
+static int sdma_v4_4_2_inst_start(struct amdgpu_device *adev,
+				  uint32_t inst_mask, bool restore)
 {
 	struct amdgpu_ring *ring;
+	uint32_t tmp_mask;
 	int i, r = 0;
 
 	if (amdgpu_sriov_vf(adev)) {
-		sdma_v4_4_2_ctx_switch_enable(adev, false);
-		sdma_v4_4_2_enable(adev, false);
+		sdma_v4_4_2_inst_ctx_switch_enable(adev, false, inst_mask);
+		sdma_v4_4_2_inst_enable(adev, false, inst_mask);
 	} else {
 		/* bypass sdma microcode loading on Gopher */
-		if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP &&
-		    !(adev->pdev->device == 0x49) && !(adev->pdev->device == 0x50)) {
-			r = sdma_v4_4_2_load_microcode(adev);
+		if (!restore && adev->firmware.load_type != AMDGPU_FW_LOAD_PSP &&
+		    adev->sdma.instance[0].fw) {
+			r = sdma_v4_4_2_inst_load_microcode(adev, inst_mask);
 			if (r)
 				return r;
 		}
 
 		/* unhalt the MEs */
-		sdma_v4_4_2_enable(adev, true);
+		sdma_v4_4_2_inst_enable(adev, true, inst_mask);
 		/* enable sdma ring preemption */
-		sdma_v4_4_2_ctx_switch_enable(adev, true);
+		sdma_v4_4_2_inst_ctx_switch_enable(adev, true, inst_mask);
 	}
 
 	/* start the gfx rings and rlc compute queues */
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	tmp_mask = inst_mask;
+	for_each_inst(i, tmp_mask) {
 		uint32_t temp;
 
 		WREG32_SDMA(i, regSDMA_SEM_WAIT_FAIL_TIMER_CNTL, 0);
-		sdma_v4_4_2_gfx_resume(adev, i);
+		sdma_v4_4_2_gfx_resume(adev, i, restore);
 		if (adev->sdma.has_page_queue)
-			sdma_v4_4_2_page_resume(adev, i);
+			sdma_v4_4_2_page_resume(adev, i, restore);
 
 		/* set utc l1 enable flag always to 1 */
 		temp = RREG32_SDMA(i, regSDMA_CNTL);
 		temp = REG_SET_FIELD(temp, SDMA_CNTL, UTC_L1_ENABLE, 1);
 		WREG32_SDMA(i, regSDMA_CNTL, temp);
 
+		if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) < IP_VERSION(4, 4, 5)) {
+			/* enable context empty interrupt during initialization */
+			temp = REG_SET_FIELD(temp, SDMA_CNTL, CTXEMPTY_INT_ENABLE, 1);
+			WREG32_SDMA(i, regSDMA_CNTL, temp);
+		}
 		if (!amdgpu_sriov_vf(adev)) {
-			ring = &adev->sdma.instance[i].ring;
-			adev->nbio.funcs->sdma_doorbell_range(adev, i,
-				ring->use_doorbell, ring->doorbell_index,
-				adev->doorbell_index.sdma_doorbell_range);
-
-			/* unhalt engine */
-			temp = RREG32_SDMA(i, regSDMA_F32_CNTL);
-			temp = REG_SET_FIELD(temp, SDMA_F32_CNTL, HALT, 0);
-			WREG32_SDMA(i, regSDMA_F32_CNTL, temp);
+			if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
+				/* unhalt engine */
+				temp = RREG32_SDMA(i, regSDMA_F32_CNTL);
+				temp = REG_SET_FIELD(temp, SDMA_F32_CNTL, HALT, 0);
+				WREG32_SDMA(i, regSDMA_F32_CNTL, temp);
+			}
 		}
 	}
 
 	if (amdgpu_sriov_vf(adev)) {
-		sdma_v4_4_2_ctx_switch_enable(adev, true);
-		sdma_v4_4_2_enable(adev, true);
+		sdma_v4_4_2_inst_ctx_switch_enable(adev, true, inst_mask);
+		sdma_v4_4_2_inst_enable(adev, true, inst_mask);
 	} else {
-		r = sdma_v4_4_2_rlc_resume(adev);
+		r = sdma_v4_4_2_inst_rlc_resume(adev, inst_mask);
 		if (r)
 			return r;
 	}
 
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	tmp_mask = inst_mask;
+	for_each_inst(i, tmp_mask) {
 		ring = &adev->sdma.instance[i].ring;
 
 		r = amdgpu_ring_test_helper(ring);
@@ -897,13 +1041,7 @@ static int sdma_v4_4_2_start(struct amdgpu_device *adev)
 			r = amdgpu_ring_test_helper(page);
 			if (r)
 				return r;
-
-			if (adev->mman.buffer_funcs_ring == page)
-				amdgpu_ttm_set_buffer_funcs_status(adev, true);
 		}
-
-		if (adev->mman.buffer_funcs_ring == ring)
-			amdgpu_ttm_set_buffer_funcs_status(adev, true);
 	}
 
 	return r;
@@ -1023,7 +1161,7 @@ static int sdma_v4_4_2_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = -EINVAL;
 
 err1:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err0:
 	amdgpu_device_wb_free(adev, index);
@@ -1194,24 +1332,29 @@ static void sdma_v4_4_2_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t re
 
 static bool sdma_v4_4_2_fw_support_paging_queue(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[SDMA0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
 	case IP_VERSION(4, 4, 2):
+	case IP_VERSION(4, 4, 5):
 		return false;
 	default:
 		return false;
 	}
 }
 
-static int sdma_v4_4_2_early_init(void *handle)
+static const struct amdgpu_sdma_funcs sdma_v4_4_2_sdma_funcs = {
+	.stop_kernel_queue = &sdma_v4_4_2_stop_queue,
+	.start_kernel_queue = &sdma_v4_4_2_restore_queue,
+	.soft_reset_kernel_queue = &sdma_v4_4_2_soft_reset_engine,
+};
+
+static int sdma_v4_4_2_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = sdma_v4_4_2_init_microcode(adev);
-	if (r) {
-		DRM_ERROR("Failed to load sdma firmware!\n");
+	if (r)
 		return r;
-	}
 
 	/* TODO: Page queue breaks driver reload under SRIOV */
 	if (sdma_v4_4_2_fw_support_paging_queue(adev))
@@ -1221,7 +1364,7 @@ static int sdma_v4_4_2_early_init(void *handle)
 	sdma_v4_4_2_set_buffer_funcs(adev);
 	sdma_v4_4_2_set_vm_pte_funcs(adev);
 	sdma_v4_4_2_set_irq_funcs(adev);
-
+	sdma_v4_4_2_set_ras_funcs(adev);
 	return 0;
 }
 
@@ -1231,31 +1374,37 @@ static int sdma_v4_4_2_process_ras_data_cb(struct amdgpu_device *adev,
 		struct amdgpu_iv_entry *entry);
 #endif
 
-static int sdma_v4_4_2_late_init(void *handle)
+static int sdma_v4_4_2_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 #if 0
 	struct ras_ih_if ih_info = {
 		.cb = sdma_v4_4_2_process_ras_data_cb,
 	};
 #endif
-	if (!amdgpu_persistent_edc_harvesting_supported(adev)) {
-		if (adev->sdma.ras && adev->sdma.ras->ras_block.hw_ops &&
-		    adev->sdma.ras->ras_block.hw_ops->reset_ras_error_count)
-			adev->sdma.ras->ras_block.hw_ops->reset_ras_error_count(adev);
-	}
+	if (!amdgpu_persistent_edc_harvesting_supported(adev))
+		amdgpu_ras_reset_error_count(adev, AMDGPU_RAS_BLOCK__SDMA);
+
+	/* The initialization is done in the late_init stage to ensure that the SMU
+	 * initialization and capability setup are completed before we check the SDMA
+	 * reset capability
+	 */
+	sdma_v4_4_2_update_reset_mask(adev);
 
 	return 0;
 }
 
-static int sdma_v4_4_2_sw_init(void *handle)
+static int sdma_v4_4_2_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
 	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	u32 aid_id;
+	uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_4_4_2);
+	uint32_t *ptr;
 
 	/* SDMA trap event */
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	for (i = 0; i < adev->sdma.num_inst_per_aid; i++) {
 		r = amdgpu_irq_add_id(adev, sdma_v4_4_2_seq_to_irq_id(i),
 				      SDMA0_4_0__SRCID__SDMA_TRAP,
 				      &adev->sdma.trap_irq);
@@ -1264,7 +1413,7 @@ static int sdma_v4_4_2_sw_init(void *handle)
 	}
 
 	/* SDMA SRAM ECC event */
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	for (i = 0; i < adev->sdma.num_inst_per_aid; i++) {
 		r = amdgpu_irq_add_id(adev, sdma_v4_4_2_seq_to_irq_id(i),
 				      SDMA0_4_0__SRCID__SDMA_SRAM_ECC,
 				      &adev->sdma.ecc_irq);
@@ -1273,7 +1422,7 @@ static int sdma_v4_4_2_sw_init(void *handle)
 	}
 
 	/* SDMA VM_HOLE/DOORBELL_INV/POLL_TIMEOUT/SRBM_WRITE_PROTECTION event*/
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	for (i = 0; i < adev->sdma.num_inst_per_aid; i++) {
 		r = amdgpu_irq_add_id(adev, sdma_v4_4_2_seq_to_irq_id(i),
 				      SDMA0_4_0__SRCID__SDMA_VM_HOLE,
 				      &adev->sdma.vm_hole_irq);
@@ -1297,21 +1446,35 @@ static int sdma_v4_4_2_sw_init(void *handle)
 				      &adev->sdma.srbm_write_irq);
 		if (r)
 			return r;
+
+		r = amdgpu_irq_add_id(adev, sdma_v4_4_2_seq_to_irq_id(i),
+				      SDMA0_4_0__SRCID__SDMA_CTXEMPTY,
+				      &adev->sdma.ctxt_empty_irq);
+		if (r)
+			return r;
 	}
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
+		mutex_init(&adev->sdma.instance[i].engine_reset_mutex);
+		/* Initialize guilty flags for GFX and PAGE queues */
+		adev->sdma.instance[i].gfx_guilty = false;
+		adev->sdma.instance[i].page_guilty = false;
+		adev->sdma.instance[i].funcs = &sdma_v4_4_2_sdma_funcs;
+
 		ring = &adev->sdma.instance[i].ring;
 		ring->ring_obj = NULL;
 		ring->use_doorbell = true;
+		aid_id = adev->sdma.instance[i].aid_id;
 
 		DRM_DEBUG("SDMA %d use_doorbell being set to: [%s]\n", i,
 				ring->use_doorbell?"true":"false");
 
 		/* doorbell size is 2 dwords, get DWORD offset */
 		ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1;
-		ring->vm_hub = AMDGPU_MMHUB_0;
+		ring->vm_hub = AMDGPU_MMHUB0(aid_id);
 
-		sprintf(ring->name, "sdma%d", i);
+		sprintf(ring->name, "sdma%d.%d", aid_id,
+				i % adev->sdma.num_inst_per_aid);
 		r = amdgpu_ring_init(adev, ring, 1024, &adev->sdma.trap_irq,
 				     AMDGPU_SDMA_IRQ_INSTANCE0 + i,
 				     AMDGPU_RING_PRIO_DEFAULT, NULL);
@@ -1323,14 +1486,15 @@ static int sdma_v4_4_2_sw_init(void *handle)
 			ring->ring_obj = NULL;
 			ring->use_doorbell = true;
 
-			/* paging queue use same doorbell index/routing as gfx queue
-			 * with 0x400 (4096 dwords) offset on second doorbell page
+			/* doorbell index of page queue is assigned right after
+			 * gfx queue on the same instance
 			 */
-			ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1;
-			ring->doorbell_index += 0x400;
-			ring->vm_hub = AMDGPU_MMHUB_0;
+			ring->doorbell_index =
+				(adev->doorbell_index.sdma_engine[i] + 1) << 1;
+			ring->vm_hub = AMDGPU_MMHUB0(aid_id);
 
-			sprintf(ring->name, "page%d", i);
+			sprintf(ring->name, "page%d.%d", aid_id,
+					i % adev->sdma.num_inst_per_aid);
 			r = amdgpu_ring_init(adev, ring, 1024,
 					     &adev->sdma.trap_irq,
 					     AMDGPU_SDMA_IRQ_INSTANCE0 + i,
@@ -1340,12 +1504,31 @@ static int sdma_v4_4_2_sw_init(void *handle)
 		}
 	}
 
+	adev->sdma.supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->sdma.instance[0].ring);
+
+	if (amdgpu_sdma_ras_sw_init(adev)) {
+		dev_err(adev->dev, "fail to initialize sdma ras block\n");
+		return -EINVAL;
+	}
+
+	/* Allocate memory for SDMA IP Dump buffer */
+	ptr = kcalloc(adev->sdma.num_instances * reg_count, sizeof(uint32_t), GFP_KERNEL);
+	if (ptr)
+		adev->sdma.ip_dump = ptr;
+	else
+		DRM_ERROR("Failed to allocated memory for SDMA IP Dump\n");
+
+	r = amdgpu_sdma_sysfs_reset_mask_init(adev);
+	if (r)
+		return r;
+
 	return r;
 }
 
-static int sdma_v4_4_2_sw_fini(void *handle)
+static int sdma_v4_4_2_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
@@ -1354,66 +1537,78 @@ static int sdma_v4_4_2_sw_fini(void *handle)
 			amdgpu_ring_fini(&adev->sdma.instance[i].page);
 	}
 
-	if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 4, 2))
+	amdgpu_sdma_sysfs_reset_mask_fini(adev);
+	if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) == IP_VERSION(4, 4, 2) ||
+	    amdgpu_ip_version(adev, SDMA0_HWIP, 0) == IP_VERSION(4, 4, 4) ||
+	    amdgpu_ip_version(adev, SDMA0_HWIP, 0) == IP_VERSION(4, 4, 5))
 		amdgpu_sdma_destroy_inst_ctx(adev, true);
 	else
 		amdgpu_sdma_destroy_inst_ctx(adev, false);
 
+	kfree(adev->sdma.ip_dump);
+
 	return 0;
 }
 
-static int sdma_v4_4_2_hw_init(void *handle)
+static int sdma_v4_4_2_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	if (adev->flags & AMD_IS_APU)
-		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_SDMA, false);
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t inst_mask;
 
+	inst_mask = GENMASK(adev->sdma.num_instances - 1, 0);
 	if (!amdgpu_sriov_vf(adev))
-		sdma_v4_4_2_init_golden_registers(adev);
+		sdma_v4_4_2_inst_init_golden_registers(adev, inst_mask);
 
-	r = sdma_v4_4_2_start(adev);
+	r = sdma_v4_4_2_inst_start(adev, inst_mask, false);
 
 	return r;
 }
 
-static int sdma_v4_4_2_hw_fini(void *handle)
+static int sdma_v4_4_2_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t inst_mask;
 	int i;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	for (i = 0; i < adev->sdma.num_instances; i++) {
-		amdgpu_irq_put(adev, &adev->sdma.ecc_irq,
-			       AMDGPU_SDMA_IRQ_INSTANCE0 + i);
+	inst_mask = GENMASK(adev->sdma.num_instances - 1, 0);
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__SDMA)) {
+		for (i = 0; i < adev->sdma.num_instances; i++) {
+			amdgpu_irq_put(adev, &adev->sdma.ecc_irq,
+				       AMDGPU_SDMA_IRQ_INSTANCE0 + i);
+		}
 	}
 
-	sdma_v4_4_2_ctx_switch_enable(adev, false);
-	sdma_v4_4_2_enable(adev, false);
+	sdma_v4_4_2_inst_ctx_switch_enable(adev, false, inst_mask);
+	sdma_v4_4_2_inst_enable(adev, false, inst_mask);
 
 	return 0;
 }
 
-static int sdma_v4_4_2_suspend(void *handle)
+static int sdma_v4_4_2_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					     enum amd_clockgating_state state);
+
+static int sdma_v4_4_2_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (amdgpu_in_reset(adev))
+		sdma_v4_4_2_set_clockgating_state(ip_block, AMD_CG_STATE_UNGATE);
 
-	return sdma_v4_4_2_hw_fini(adev);
+	return sdma_v4_4_2_hw_fini(ip_block);
 }
 
-static int sdma_v4_4_2_resume(void *handle)
+static int sdma_v4_4_2_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return sdma_v4_4_2_hw_init(adev);
+	return sdma_v4_4_2_hw_init(ip_block);
 }
 
-static bool sdma_v4_4_2_is_idle(void *handle)
+static bool sdma_v4_4_2_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
@@ -1426,11 +1621,11 @@ static bool sdma_v4_4_2_is_idle(void *handle)
 	return true;
 }
 
-static int sdma_v4_4_2_wait_for_idle(void *handle)
+static int sdma_v4_4_2_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i, j;
 	u32 sdma[AMDGPU_MAX_SDMA_INSTANCES];
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		for (j = 0; j < adev->sdma.num_instances; j++) {
@@ -1445,13 +1640,109 @@ static int sdma_v4_4_2_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int sdma_v4_4_2_soft_reset(void *handle)
+static int sdma_v4_4_2_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	/* todo */
 
 	return 0;
 }
 
+static bool sdma_v4_4_2_is_queue_selected(struct amdgpu_device *adev, uint32_t instance_id, bool is_page_queue)
+{
+	uint32_t reg_offset = is_page_queue ? regSDMA_PAGE_CONTEXT_STATUS : regSDMA_GFX_CONTEXT_STATUS;
+	uint32_t context_status = RREG32(sdma_v4_4_2_get_reg_offset(adev, instance_id, reg_offset));
+
+	/* Check if the SELECTED bit is set */
+	return (context_status & SDMA_GFX_CONTEXT_STATUS__SELECTED_MASK) != 0;
+}
+
+static int sdma_v4_4_2_reset_queue(struct amdgpu_ring *ring,
+				   unsigned int vmid,
+				   struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 id = ring->me;
+	int r;
+
+	amdgpu_amdkfd_suspend(adev, true);
+	r = amdgpu_sdma_reset_engine(adev, id, false);
+	amdgpu_amdkfd_resume(adev, true);
+	return r;
+}
+
+static int sdma_v4_4_2_stop_queue(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 instance_id = ring->me;
+	u32 inst_mask;
+	uint64_t rptr;
+
+	if (amdgpu_sriov_vf(adev))
+		return -EINVAL;
+
+	/* Check if this queue is the guilty one */
+	adev->sdma.instance[instance_id].gfx_guilty =
+		sdma_v4_4_2_is_queue_selected(adev, instance_id, false);
+	if (adev->sdma.has_page_queue)
+		adev->sdma.instance[instance_id].page_guilty =
+			sdma_v4_4_2_is_queue_selected(adev, instance_id, true);
+
+	/* Cache the rptr before reset, after the reset,
+	* all of the registers will be reset to 0
+	*/
+	rptr = amdgpu_ring_get_rptr(ring);
+	ring->cached_rptr = rptr;
+	/* Cache the rptr for the page queue if it exists */
+	if (adev->sdma.has_page_queue) {
+		struct amdgpu_ring *page_ring = &adev->sdma.instance[instance_id].page;
+		rptr = amdgpu_ring_get_rptr(page_ring);
+		page_ring->cached_rptr = rptr;
+	}
+
+	/* stop queue */
+	inst_mask = 1 << ring->me;
+	sdma_v4_4_2_inst_gfx_stop(adev, inst_mask);
+	if (adev->sdma.has_page_queue)
+		sdma_v4_4_2_inst_page_stop(adev, inst_mask);
+
+	return 0;
+}
+
+static int sdma_v4_4_2_restore_queue(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 inst_mask;
+	int i, r;
+
+	inst_mask = 1 << ring->me;
+	udelay(50);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (!REG_GET_FIELD(RREG32_SDMA(ring->me, regSDMA_F32_CNTL), SDMA_F32_CNTL, HALT))
+			break;
+		udelay(1);
+	}
+
+	if (i == adev->usec_timeout) {
+		dev_err(adev->dev, "timed out waiting for SDMA%d unhalt after reset\n",
+			ring->me);
+		return -ETIMEDOUT;
+	}
+
+	r = sdma_v4_4_2_inst_start(adev, inst_mask, true);
+
+	return r;
+}
+
+static int sdma_v4_4_2_soft_reset_engine(struct amdgpu_device *adev,
+					 u32 instance_id)
+{
+	/* For SDMA 4.x, use the existing DPM interface for backward compatibility
+	 * we need to convert the logical instance ID to physical instance ID before reset.
+	 */
+	return amdgpu_dpm_reset_sdma(adev, 1 << GET_INST(SDMA0, instance_id));
+}
+
 static int sdma_v4_4_2_set_trap_irq_state(struct amdgpu_device *adev,
 					struct amdgpu_irq_src *source,
 					unsigned type,
@@ -1471,13 +1762,34 @@ static int sdma_v4_4_2_process_trap_irq(struct amdgpu_device *adev,
 				      struct amdgpu_irq_src *source,
 				      struct amdgpu_iv_entry *entry)
 {
-	uint32_t instance;
+	uint32_t instance, i;
 
 	DRM_DEBUG("IH: SDMA trap\n");
-	instance = sdma_v4_4_2_irq_id_to_seq(entry->client_id);
+	instance = sdma_v4_4_2_irq_id_to_seq(adev, entry->client_id);
+
+	/* Client id gives the SDMA instance in AID. To know the exact SDMA
+	 * instance, interrupt entry gives the node id which corresponds to the AID instance.
+	 * Match node id with the AID id associated with the SDMA instance. */
+	for (i = instance; i < adev->sdma.num_instances;
+	     i += adev->sdma.num_inst_per_aid) {
+		if (adev->sdma.instance[i].aid_id ==
+		    node_id_to_phys_map[entry->node_id])
+			break;
+	}
+
+	if (i >= adev->sdma.num_instances) {
+		dev_WARN_ONCE(
+			adev->dev, 1,
+			"Couldn't find the right sdma instance in trap handler");
+		return 0;
+	}
+
 	switch (entry->ring_id) {
 	case 0:
-		amdgpu_fence_process(&adev->sdma.instance[instance].ring);
+		amdgpu_fence_process(&adev->sdma.instance[i].ring);
+		break;
+	case 1:
+		amdgpu_fence_process(&adev->sdma.instance[i].page);
 		break;
 	default:
 		break;
@@ -1496,10 +1808,10 @@ static int sdma_v4_4_2_process_ras_data_cb(struct amdgpu_device *adev,
 	 * be disabled and the driver should only look for the aggregated
 	 * interrupt via sync flood
 	 */
-	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX))
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__SDMA))
 		goto out;
 
-	instance = sdma_v4_4_2_irq_id_to_seq(entry->client_id);
+	instance = sdma_v4_4_2_irq_id_to_seq(adev, entry->client_id);
 	if (instance < 0)
 		goto out;
 
@@ -1518,7 +1830,7 @@ static int sdma_v4_4_2_process_illegal_inst_irq(struct amdgpu_device *adev,
 
 	DRM_ERROR("Illegal instruction in SDMA command stream\n");
 
-	instance = sdma_v4_4_2_irq_id_to_seq(entry->client_id);
+	instance = sdma_v4_4_2_irq_id_to_seq(adev, entry->client_id);
 	if (instance < 0)
 		return 0;
 
@@ -1535,15 +1847,12 @@ static int sdma_v4_4_2_set_ecc_irq_state(struct amdgpu_device *adev,
 					unsigned type,
 					enum amdgpu_interrupt_state state)
 {
-	u32 sdma_edc_config;
+	u32 sdma_cntl;
 
-	sdma_edc_config = RREG32_SDMA(type, regCC_SDMA_EDC_CONFIG);
-	/*
-	 * FIXME: This was inherited from Aldebaran, but no this field
-	 * definition in the regspec of both Aldebaran and SDMA 4.4.2
-	 */
-	sdma_edc_config |= (state == AMDGPU_IRQ_STATE_ENABLE) ? (1 << 2) : 0;
-	WREG32_SDMA(type, regCC_SDMA_EDC_CONFIG, sdma_edc_config);
+	sdma_cntl = RREG32_SDMA(type, regSDMA_CNTL);
+	sdma_cntl = REG_SET_FIELD(sdma_cntl, SDMA_CNTL, DRAM_ECC_INT_ENABLE,
+					state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+	WREG32_SDMA(type, regSDMA_CNTL, sdma_cntl);
 
 	return 0;
 }
@@ -1552,10 +1861,10 @@ static int sdma_v4_4_2_print_iv_entry(struct amdgpu_device *adev,
 					      struct amdgpu_iv_entry *entry)
 {
 	int instance;
-	struct amdgpu_task_info task_info;
+	struct amdgpu_task_info *task_info;
 	u64 addr;
 
-	instance = sdma_v4_4_2_irq_id_to_seq(entry->client_id);
+	instance = sdma_v4_4_2_irq_id_to_seq(adev, entry->client_id);
 	if (instance < 0 || instance >= adev->sdma.num_instances) {
 		dev_err(adev->dev, "sdma instance invalid %d\n", instance);
 		return -EINVAL;
@@ -1564,15 +1873,19 @@ static int sdma_v4_4_2_print_iv_entry(struct amdgpu_device *adev,
 	addr = (u64)entry->src_data[0] << 12;
 	addr |= ((u64)entry->src_data[1] & 0xf) << 44;
 
-	memset(&task_info, 0, sizeof(struct amdgpu_task_info));
-	amdgpu_vm_get_task_info(adev, entry->pasid, &task_info);
-
 	dev_dbg_ratelimited(adev->dev,
-		   "[sdma%d] address:0x%016llx src_id:%u ring:%u vmid:%u "
-		   "pasid:%u, for process %s pid %d thread %s pid %d\n",
-		   instance, addr, entry->src_id, entry->ring_id, entry->vmid,
-		   entry->pasid, task_info.process_name, task_info.tgid,
-		   task_info.task_name, task_info.pid);
+			    "[sdma%d] address:0x%016llx src_id:%u ring:%u vmid:%u pasid:%u\n",
+			    instance, addr, entry->src_id, entry->ring_id, entry->vmid,
+			    entry->pasid);
+
+	task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid);
+	if (task_info) {
+		dev_dbg_ratelimited(adev->dev, " for process %s pid %d thread %s pid %d\n",
+				    task_info->process_name, task_info->tgid,
+				    task_info->task.comm, task_info->task.pid);
+		amdgpu_vm_put_task_info(task_info);
+	}
+
 	return 0;
 }
 
@@ -1615,19 +1928,59 @@ static int sdma_v4_4_2_process_srbm_write_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static void sdma_v4_4_2_update_medium_grain_clock_gating(
-		struct amdgpu_device *adev,
-		bool enable)
+static int sdma_v4_4_2_process_ctxt_empty_irq(struct amdgpu_device *adev,
+					      struct amdgpu_irq_src *source,
+					      struct amdgpu_iv_entry *entry)
+{
+	/* There is nothing useful to be done here, only kept for debug */
+	dev_dbg_ratelimited(adev->dev, "SDMA context empty interrupt");
+	sdma_v4_4_2_print_iv_entry(adev, entry);
+	return 0;
+}
+
+static void sdma_v4_4_2_inst_update_medium_grain_light_sleep(
+	struct amdgpu_device *adev, bool enable, uint32_t inst_mask)
 {
 	uint32_t data, def;
 	int i;
 
-	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_SDMA_MGCG)) {
-		for (i = 0; i < adev->sdma.num_instances; i++) {
+	/* leave as default if it is not driver controlled */
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_SDMA_LS))
+		return;
+
+	if (enable) {
+		for_each_inst(i, inst_mask) {
+			/* 1-not override: enable sdma mem light sleep */
+			def = data = RREG32_SDMA(i, regSDMA_POWER_CNTL);
+			data |= SDMA_POWER_CNTL__MEM_POWER_OVERRIDE_MASK;
+			if (def != data)
+				WREG32_SDMA(i, regSDMA_POWER_CNTL, data);
+		}
+	} else {
+		for_each_inst(i, inst_mask) {
+			/* 0-override:disable sdma mem light sleep */
+			def = data = RREG32_SDMA(i, regSDMA_POWER_CNTL);
+			data &= ~SDMA_POWER_CNTL__MEM_POWER_OVERRIDE_MASK;
+			if (def != data)
+				WREG32_SDMA(i, regSDMA_POWER_CNTL, data);
+		}
+	}
+}
+
+static void sdma_v4_4_2_inst_update_medium_grain_clock_gating(
+	struct amdgpu_device *adev, bool enable, uint32_t inst_mask)
+{
+	uint32_t data, def;
+	int i;
+
+	/* leave as default if it is not driver controlled */
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_SDMA_MGCG))
+		return;
+
+	if (enable) {
+		for_each_inst(i, inst_mask) {
 			def = data = RREG32_SDMA(i, regSDMA_CLK_CTRL);
-			data &= ~(SDMA_CLK_CTRL__SOFT_OVERRIDE7_MASK |
-				  SDMA_CLK_CTRL__SOFT_OVERRIDE6_MASK |
-				  SDMA_CLK_CTRL__SOFT_OVERRIDE5_MASK |
+			data &= ~(SDMA_CLK_CTRL__SOFT_OVERRIDE5_MASK |
 				  SDMA_CLK_CTRL__SOFT_OVERRIDE4_MASK |
 				  SDMA_CLK_CTRL__SOFT_OVERRIDE3_MASK |
 				  SDMA_CLK_CTRL__SOFT_OVERRIDE2_MASK |
@@ -1637,11 +1990,9 @@ static void sdma_v4_4_2_update_medium_grain_clock_gating(
 				WREG32_SDMA(i, regSDMA_CLK_CTRL, data);
 		}
 	} else {
-		for (i = 0; i < adev->sdma.num_instances; i++) {
+		for_each_inst(i, inst_mask) {
 			def = data = RREG32_SDMA(i, regSDMA_CLK_CTRL);
-			data |= (SDMA_CLK_CTRL__SOFT_OVERRIDE7_MASK |
-				 SDMA_CLK_CTRL__SOFT_OVERRIDE6_MASK |
-				 SDMA_CLK_CTRL__SOFT_OVERRIDE5_MASK |
+			data |= (SDMA_CLK_CTRL__SOFT_OVERRIDE5_MASK |
 				 SDMA_CLK_CTRL__SOFT_OVERRIDE4_MASK |
 				 SDMA_CLK_CTRL__SOFT_OVERRIDE3_MASK |
 				 SDMA_CLK_CTRL__SOFT_OVERRIDE2_MASK |
@@ -1653,73 +2004,89 @@ static void sdma_v4_4_2_update_medium_grain_clock_gating(
 	}
 }
 
-
-static void sdma_v4_4_2_update_medium_grain_light_sleep(
-		struct amdgpu_device *adev,
-		bool enable)
-{
-	uint32_t data, def;
-	int i;
-
-	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_SDMA_LS)) {
-		for (i = 0; i < adev->sdma.num_instances; i++) {
-			/* 1-not override: enable sdma mem light sleep */
-			def = data = RREG32_SDMA(0, regSDMA_POWER_CNTL);
-			data |= SDMA_POWER_CNTL__MEM_POWER_OVERRIDE_MASK;
-			if (def != data)
-				WREG32_SDMA(0, regSDMA_POWER_CNTL, data);
-		}
-	} else {
-		for (i = 0; i < adev->sdma.num_instances; i++) {
-		/* 0-override:disable sdma mem light sleep */
-			def = data = RREG32_SDMA(0, regSDMA_POWER_CNTL);
-			data &= ~SDMA_POWER_CNTL__MEM_POWER_OVERRIDE_MASK;
-			if (def != data)
-				WREG32_SDMA(0, regSDMA_POWER_CNTL, data);
-		}
-	}
-}
-
-static int sdma_v4_4_2_set_clockgating_state(void *handle,
+static int sdma_v4_4_2_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t inst_mask;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	sdma_v4_4_2_update_medium_grain_clock_gating(adev,
-			state == AMD_CG_STATE_GATE);
-	sdma_v4_4_2_update_medium_grain_light_sleep(adev,
-			state == AMD_CG_STATE_GATE);
+	inst_mask = GENMASK(adev->sdma.num_instances - 1, 0);
+
+	sdma_v4_4_2_inst_update_medium_grain_clock_gating(
+		adev, state == AMD_CG_STATE_GATE, inst_mask);
+	sdma_v4_4_2_inst_update_medium_grain_light_sleep(
+		adev, state == AMD_CG_STATE_GATE, inst_mask);
 	return 0;
 }
 
-static int sdma_v4_4_2_set_powergating_state(void *handle,
+static int sdma_v4_4_2_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
 }
 
-static void sdma_v4_4_2_get_clockgating_state(void *handle, u64 *flags)
+static void sdma_v4_4_2_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	if (amdgpu_sriov_vf(adev))
 		*flags = 0;
 
 	/* AMD_CG_SUPPORT_SDMA_MGCG */
-	data = RREG32(SOC15_REG_OFFSET(SDMA0, 0, regSDMA_CLK_CTRL));
-	if (!(data & SDMA_CLK_CTRL__SOFT_OVERRIDE7_MASK))
+	data = RREG32(SOC15_REG_OFFSET(SDMA0, GET_INST(SDMA0, 0), regSDMA_CLK_CTRL));
+	if (!(data & SDMA_CLK_CTRL__SOFT_OVERRIDE5_MASK))
 		*flags |= AMD_CG_SUPPORT_SDMA_MGCG;
 
 	/* AMD_CG_SUPPORT_SDMA_LS */
-	data = RREG32(SOC15_REG_OFFSET(SDMA0, 0, regSDMA_POWER_CNTL));
+	data = RREG32(SOC15_REG_OFFSET(SDMA0, GET_INST(SDMA0, 0), regSDMA_POWER_CNTL));
 	if (data & SDMA_POWER_CNTL__MEM_POWER_OVERRIDE_MASK)
 		*flags |= AMD_CG_SUPPORT_SDMA_LS;
 }
 
+static void sdma_v4_4_2_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, j;
+	uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_4_4_2);
+	uint32_t instance_offset;
+
+	if (!adev->sdma.ip_dump)
+		return;
+
+	drm_printf(p, "num_instances:%d\n", adev->sdma.num_instances);
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		instance_offset = i * reg_count;
+		drm_printf(p, "\nInstance:%d\n", i);
+
+		for (j = 0; j < reg_count; j++)
+			drm_printf(p, "%-50s \t 0x%08x\n", sdma_reg_list_4_4_2[j].reg_name,
+				   adev->sdma.ip_dump[instance_offset + j]);
+	}
+}
+
+static void sdma_v4_4_2_dump_ip_state(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, j;
+	uint32_t instance_offset;
+	uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_4_4_2);
+
+	if (!adev->sdma.ip_dump)
+		return;
+
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		instance_offset = i * reg_count;
+		for (j = 0; j < reg_count; j++)
+			adev->sdma.ip_dump[instance_offset + j] =
+				RREG32(sdma_v4_4_2_get_reg_offset(adev, i,
+				       sdma_reg_list_4_4_2[j].reg_offset));
+	}
+}
+
 const struct amd_ip_funcs sdma_v4_4_2_ip_funcs = {
 	.name = "sdma_v4_4_2",
 	.early_init = sdma_v4_4_2_early_init,
@@ -1736,11 +2103,13 @@ const struct amd_ip_funcs sdma_v4_4_2_ip_funcs = {
 	.set_clockgating_state = sdma_v4_4_2_set_clockgating_state,
 	.set_powergating_state = sdma_v4_4_2_set_powergating_state,
 	.get_clockgating_state = sdma_v4_4_2_get_clockgating_state,
+	.dump_ip_state = sdma_v4_4_2_dump_ip_state,
+	.print_ip_state = sdma_v4_4_2_print_ip_state,
 };
 
 static const struct amdgpu_ring_funcs sdma_v4_4_2_ring_funcs = {
 	.type = AMDGPU_RING_TYPE_SDMA,
-	.align_mask = 0xf,
+	.align_mask = 0xff,
 	.nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP),
 	.support_64bit_ptrs = true,
 	.get_rptr = sdma_v4_4_2_ring_get_rptr,
@@ -1767,11 +2136,12 @@ static const struct amdgpu_ring_funcs sdma_v4_4_2_ring_funcs = {
 	.emit_wreg = sdma_v4_4_2_ring_emit_wreg,
 	.emit_reg_wait = sdma_v4_4_2_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = sdma_v4_4_2_reset_queue,
 };
 
 static const struct amdgpu_ring_funcs sdma_v4_4_2_page_ring_funcs = {
 	.type = AMDGPU_RING_TYPE_SDMA,
-	.align_mask = 0xf,
+	.align_mask = 0xff,
 	.nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP),
 	.support_64bit_ptrs = true,
 	.get_rptr = sdma_v4_4_2_ring_get_rptr,
@@ -1798,11 +2168,12 @@ static const struct amdgpu_ring_funcs sdma_v4_4_2_page_ring_funcs = {
 	.emit_wreg = sdma_v4_4_2_ring_emit_wreg,
 	.emit_reg_wait = sdma_v4_4_2_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = sdma_v4_4_2_reset_queue,
 };
 
 static void sdma_v4_4_2_set_ring_funcs(struct amdgpu_device *adev)
 {
-	int i;
+	int i, dev_inst;
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		adev->sdma.instance[i].ring.funcs = &sdma_v4_4_2_ring_funcs;
@@ -1812,6 +2183,11 @@ static void sdma_v4_4_2_set_ring_funcs(struct amdgpu_device *adev)
 				&sdma_v4_4_2_page_ring_funcs;
 			adev->sdma.instance[i].page.me = i;
 		}
+
+		dev_inst = GET_INST(SDMA0, i);
+		/* AID to which SDMA belongs depends on physical instance */
+		adev->sdma.instance[i].aid_id =
+			dev_inst / adev->sdma.num_inst_per_aid;
 	}
 }
 
@@ -1845,6 +2221,10 @@ static const struct amdgpu_irq_src_funcs sdma_v4_4_2_srbm_write_irq_funcs = {
 	.process = sdma_v4_4_2_process_srbm_write_irq,
 };
 
+static const struct amdgpu_irq_src_funcs sdma_v4_4_2_ctxt_empty_irq_funcs = {
+	.process = sdma_v4_4_2_process_ctxt_empty_irq,
+};
+
 static void sdma_v4_4_2_set_irq_funcs(struct amdgpu_device *adev)
 {
 	adev->sdma.trap_irq.num_types = adev->sdma.num_instances;
@@ -1853,6 +2233,7 @@ static void sdma_v4_4_2_set_irq_funcs(struct amdgpu_device *adev)
 	adev->sdma.doorbell_invalid_irq.num_types = adev->sdma.num_instances;
 	adev->sdma.pool_timeout_irq.num_types = adev->sdma.num_instances;
 	adev->sdma.srbm_write_irq.num_types = adev->sdma.num_instances;
+	adev->sdma.ctxt_empty_irq.num_types = adev->sdma.num_instances;
 
 	adev->sdma.trap_irq.funcs = &sdma_v4_4_2_trap_irq_funcs;
 	adev->sdma.illegal_inst_irq.funcs = &sdma_v4_4_2_illegal_inst_irq_funcs;
@@ -1861,6 +2242,7 @@ static void sdma_v4_4_2_set_irq_funcs(struct amdgpu_device *adev)
 	adev->sdma.doorbell_invalid_irq.funcs = &sdma_v4_4_2_doorbell_invalid_irq_funcs;
 	adev->sdma.pool_timeout_irq.funcs = &sdma_v4_4_2_pool_timeout_irq_funcs;
 	adev->sdma.srbm_write_irq.funcs = &sdma_v4_4_2_srbm_write_irq_funcs;
+	adev->sdma.ctxt_empty_irq.funcs = &sdma_v4_4_2_ctxt_empty_irq_funcs;
 }
 
 /**
@@ -1870,7 +2252,7 @@ static void sdma_v4_4_2_set_irq_funcs(struct amdgpu_device *adev)
  * @src_offset: src GPU address
  * @dst_offset: dst GPU address
  * @byte_count: number of bytes to xfer
- * @tmz: if a secure copy should be used
+ * @copy_flags: copy flags for the buffers
  *
  * Copy GPU buffers using the DMA engine.
  * Used by the amdgpu ttm implementation to move pages if
@@ -1880,11 +2262,11 @@ static void sdma_v4_4_2_emit_copy_buffer(struct amdgpu_ib *ib,
 				       uint64_t src_offset,
 				       uint64_t dst_offset,
 				       uint32_t byte_count,
-				       bool tmz)
+				       uint32_t copy_flags)
 {
 	ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) |
 		SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR) |
-		SDMA_PKT_COPY_LINEAR_HEADER_TMZ(tmz ? 1 : 0);
+		SDMA_PKT_COPY_LINEAR_HEADER_TMZ((copy_flags & AMDGPU_COPY_FLAGS_TMZ) ? 1 : 0);
 	ib->ptr[ib->length_dw++] = byte_count - 1;
 	ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */
 	ib->ptr[ib->length_dw++] = lower_32_bits(src_offset);
@@ -1958,10 +2340,274 @@ static void sdma_v4_4_2_set_vm_pte_funcs(struct amdgpu_device *adev)
 	adev->vm_manager.vm_pte_num_scheds = adev->sdma.num_instances;
 }
 
+/**
+ * sdma_v4_4_2_update_reset_mask - update  reset mask for SDMA
+ * @adev: Pointer to the AMDGPU device structure
+ *
+ * This function update reset mask for SDMA and sets the supported
+ * reset types based on the IP version and firmware versions.
+ *
+ */
+static void sdma_v4_4_2_update_reset_mask(struct amdgpu_device *adev)
+{
+	/* per queue reset not supported for SRIOV */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	/*
+	 * the user queue relies on MEC fw and pmfw when the sdma queue do reset.
+	 * it needs to check both of them at here to skip old mec and pmfw.
+	 */
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+		if ((adev->gfx.mec_fw_version >= 0xb0) &&
+		    amdgpu_dpm_reset_sdma_is_supported(adev) &&
+		    !adev->debug_disable_gpu_ring_reset)
+			adev->sdma.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+		break;
+	case IP_VERSION(9, 5, 0):
+		if ((adev->gfx.mec_fw_version >= 0xf) &&
+		    amdgpu_dpm_reset_sdma_is_supported(adev) &&
+		    !adev->debug_disable_gpu_ring_reset)
+			adev->sdma.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+		break;
+	default:
+		break;
+	}
+
+}
+
 const struct amdgpu_ip_block_version sdma_v4_4_2_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_SDMA,
 	.major = 4,
 	.minor = 4,
-	.rev = 0,
+	.rev = 2,
 	.funcs = &sdma_v4_4_2_ip_funcs,
 };
+
+static int sdma_v4_4_2_xcp_resume(void *handle, uint32_t inst_mask)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	int r;
+
+	if (!amdgpu_sriov_vf(adev))
+		sdma_v4_4_2_inst_init_golden_registers(adev, inst_mask);
+
+	r = sdma_v4_4_2_inst_start(adev, inst_mask, false);
+
+	return r;
+}
+
+static int sdma_v4_4_2_xcp_suspend(void *handle, uint32_t inst_mask)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	uint32_t tmp_mask = inst_mask;
+	int i;
+
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__SDMA)) {
+		for_each_inst(i, tmp_mask) {
+			amdgpu_irq_put(adev, &adev->sdma.ecc_irq,
+				       AMDGPU_SDMA_IRQ_INSTANCE0 + i);
+		}
+	}
+
+	sdma_v4_4_2_inst_ctx_switch_enable(adev, false, inst_mask);
+	sdma_v4_4_2_inst_enable(adev, false, inst_mask);
+
+	return 0;
+}
+
+struct amdgpu_xcp_ip_funcs sdma_v4_4_2_xcp_funcs = {
+	.suspend = &sdma_v4_4_2_xcp_suspend,
+	.resume = &sdma_v4_4_2_xcp_resume
+};
+
+static const struct amdgpu_ras_err_status_reg_entry sdma_v4_2_2_ue_reg_list[] = {
+	{AMDGPU_RAS_REG_ENTRY(SDMA0, 0, regSDMA_UE_ERR_STATUS_LO, regSDMA_UE_ERR_STATUS_HI),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SDMA"},
+};
+
+static const struct amdgpu_ras_memory_id_entry sdma_v4_4_2_ras_memory_list[] = {
+	{AMDGPU_SDMA_MBANK_DATA_BUF0, "SDMA_MBANK_DATA_BUF0"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF1, "SDMA_MBANK_DATA_BUF1"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF2, "SDMA_MBANK_DATA_BUF2"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF3, "SDMA_MBANK_DATA_BUF3"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF4, "SDMA_MBANK_DATA_BUF4"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF5, "SDMA_MBANK_DATA_BUF5"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF6, "SDMA_MBANK_DATA_BUF6"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF7, "SDMA_MBANK_DATA_BUF7"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF8, "SDMA_MBANK_DATA_BUF8"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF9, "SDMA_MBANK_DATA_BUF9"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF10, "SDMA_MBANK_DATA_BUF10"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF11, "SDMA_MBANK_DATA_BUF11"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF12, "SDMA_MBANK_DATA_BUF12"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF13, "SDMA_MBANK_DATA_BUF13"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF14, "SDMA_MBANK_DATA_BUF14"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF15, "SDMA_MBANK_DATA_BUF15"},
+	{AMDGPU_SDMA_UCODE_BUF, "SDMA_UCODE_BUF"},
+	{AMDGPU_SDMA_RB_CMD_BUF, "SDMA_RB_CMD_BUF"},
+	{AMDGPU_SDMA_IB_CMD_BUF, "SDMA_IB_CMD_BUF"},
+	{AMDGPU_SDMA_UTCL1_RD_FIFO, "SDMA_UTCL1_RD_FIFO"},
+	{AMDGPU_SDMA_UTCL1_RDBST_FIFO, "SDMA_UTCL1_RDBST_FIFO"},
+	{AMDGPU_SDMA_UTCL1_WR_FIFO, "SDMA_UTCL1_WR_FIFO"},
+	{AMDGPU_SDMA_DATA_LUT_FIFO, "SDMA_DATA_LUT_FIFO"},
+	{AMDGPU_SDMA_SPLIT_DAT_BUF, "SDMA_SPLIT_DAT_BUF"},
+};
+
+static void sdma_v4_4_2_inst_query_ras_error_count(struct amdgpu_device *adev,
+						   uint32_t sdma_inst,
+						   void *ras_err_status)
+{
+	struct ras_err_data *err_data = (struct ras_err_data *)ras_err_status;
+	uint32_t sdma_dev_inst = GET_INST(SDMA0, sdma_inst);
+	unsigned long ue_count = 0;
+	struct amdgpu_smuio_mcm_config_info mcm_info = {
+		.socket_id = adev->smuio.funcs->get_socket_id(adev),
+		.die_id = adev->sdma.instance[sdma_inst].aid_id,
+	};
+
+	/* sdma v4_4_2 doesn't support query ce counts */
+	amdgpu_ras_inst_query_ras_error_count(adev,
+					sdma_v4_2_2_ue_reg_list,
+					ARRAY_SIZE(sdma_v4_2_2_ue_reg_list),
+					sdma_v4_4_2_ras_memory_list,
+					ARRAY_SIZE(sdma_v4_4_2_ras_memory_list),
+					sdma_dev_inst,
+					AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE,
+					&ue_count);
+
+	amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, ue_count);
+}
+
+static void sdma_v4_4_2_query_ras_error_count(struct amdgpu_device *adev,
+					      void *ras_err_status)
+{
+	uint32_t inst_mask;
+	int i = 0;
+
+	inst_mask = GENMASK(adev->sdma.num_instances - 1, 0);
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__SDMA)) {
+		for_each_inst(i, inst_mask)
+			sdma_v4_4_2_inst_query_ras_error_count(adev, i, ras_err_status);
+	} else {
+		dev_warn(adev->dev, "SDMA RAS is not supported\n");
+	}
+}
+
+static void sdma_v4_4_2_inst_reset_ras_error_count(struct amdgpu_device *adev,
+						   uint32_t sdma_inst)
+{
+	uint32_t sdma_dev_inst = GET_INST(SDMA0, sdma_inst);
+
+	amdgpu_ras_inst_reset_ras_error_count(adev,
+					sdma_v4_2_2_ue_reg_list,
+					ARRAY_SIZE(sdma_v4_2_2_ue_reg_list),
+					sdma_dev_inst);
+}
+
+static void sdma_v4_4_2_reset_ras_error_count(struct amdgpu_device *adev)
+{
+	uint32_t inst_mask;
+	int i = 0;
+
+	inst_mask = GENMASK(adev->sdma.num_instances - 1, 0);
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__SDMA)) {
+		for_each_inst(i, inst_mask)
+			sdma_v4_4_2_inst_reset_ras_error_count(adev, i);
+	} else {
+		dev_warn(adev->dev, "SDMA RAS is not supported\n");
+	}
+}
+
+static const struct amdgpu_ras_block_hw_ops sdma_v4_4_2_ras_hw_ops = {
+	.query_ras_error_count = sdma_v4_4_2_query_ras_error_count,
+	.reset_ras_error_count = sdma_v4_4_2_reset_ras_error_count,
+};
+
+static int sdma_v4_4_2_aca_bank_parser(struct aca_handle *handle, struct aca_bank *bank,
+				       enum aca_smu_type type, void *data)
+{
+	struct aca_bank_info info;
+	u64 misc0;
+	int ret;
+
+	ret = aca_bank_info_decode(bank, &info);
+	if (ret)
+		return ret;
+
+	misc0 = bank->regs[ACA_REG_IDX_MISC0];
+	switch (type) {
+	case ACA_SMU_TYPE_UE:
+		bank->aca_err_type = ACA_ERROR_TYPE_UE;
+		ret = aca_error_cache_log_bank_error(handle, &info, ACA_ERROR_TYPE_UE,
+						     1ULL);
+		break;
+	case ACA_SMU_TYPE_CE:
+		bank->aca_err_type = ACA_ERROR_TYPE_CE;
+		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type,
+						     ACA_REG__MISC0__ERRCNT(misc0));
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return ret;
+}
+
+/* CODE_SDMA0 - CODE_SDMA4, reference to smu driver if header file */
+static int sdma_v4_4_2_err_codes[] = { 33, 34, 35, 36 };
+
+static bool sdma_v4_4_2_aca_bank_is_valid(struct aca_handle *handle, struct aca_bank *bank,
+					  enum aca_smu_type type, void *data)
+{
+	u32 instlo;
+
+	instlo = ACA_REG__IPID__INSTANCEIDLO(bank->regs[ACA_REG_IDX_IPID]);
+	instlo &= GENMASK(31, 1);
+
+	if (instlo != mmSMNAID_AID0_MCA_SMU)
+		return false;
+
+	if (aca_bank_check_error_codes(handle->adev, bank,
+				       sdma_v4_4_2_err_codes,
+				       ARRAY_SIZE(sdma_v4_4_2_err_codes)))
+		return false;
+
+	return true;
+}
+
+static const struct aca_bank_ops sdma_v4_4_2_aca_bank_ops = {
+	.aca_bank_parser = sdma_v4_4_2_aca_bank_parser,
+	.aca_bank_is_valid = sdma_v4_4_2_aca_bank_is_valid,
+};
+
+static const struct aca_info sdma_v4_4_2_aca_info = {
+	.hwip = ACA_HWIP_TYPE_SMU,
+	.mask = ACA_ERROR_UE_MASK,
+	.bank_ops = &sdma_v4_4_2_aca_bank_ops,
+};
+
+static int sdma_v4_4_2_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
+{
+	int r;
+
+	r = amdgpu_sdma_ras_late_init(adev, ras_block);
+	if (r)
+		return r;
+
+	return amdgpu_ras_bind_aca(adev, AMDGPU_RAS_BLOCK__SDMA,
+				   &sdma_v4_4_2_aca_info, NULL);
+}
+
+static struct amdgpu_sdma_ras sdma_v4_4_2_ras = {
+	.ras_block = {
+		.hw_ops = &sdma_v4_4_2_ras_hw_ops,
+		.ras_late_init = sdma_v4_4_2_ras_late_init,
+	},
+};
+
+static void sdma_v4_4_2_set_ras_funcs(struct amdgpu_device *adev)
+{
+	adev->sdma.ras = &sdma_v4_4_2_ras;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.h b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.h
index 4814e8a074d6..d516145529bb 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.h
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.h
@@ -27,4 +27,6 @@
 extern const struct amd_ip_funcs sdma_v4_4_2_ip_funcs;
 extern const struct amdgpu_ip_block_version sdma_v4_4_2_ip_block;
 
+extern struct amdgpu_xcp_ip_funcs sdma_v4_4_2_xcp_funcs;
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c
index 92e1299be021..8ddc4df06a1f 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c
@@ -59,10 +59,61 @@ MODULE_FIRMWARE("amdgpu/cyan_skillfish2_sdma1.bin");
 #define SDMA0_HYP_DEC_REG_END 0x5893
 #define SDMA1_HYP_DEC_REG_OFFSET 0x20
 
+static const struct amdgpu_hwip_reg_entry sdma_reg_list_5_0[] = {
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_STATUS_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_STATUS1_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_STATUS2_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_STATUS3_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UCODE_CHECKSUM),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RB_RPTR_FETCH_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RB_RPTR_FETCH),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UTCL1_RD_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UTCL1_WR_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UTCL1_RD_XNACK0),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UTCL1_RD_XNACK1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UTCL1_WR_XNACK0),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UTCL1_WR_XNACK1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_RB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_RB_RPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_RB_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_IB_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_IB_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_IB_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_IB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_IB_SUB_REMAIN),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_DUMMY_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_RB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_RB_RPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_RB_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_IB_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_IB_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_IB_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_DUMMY_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_RB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_RB_RPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_RB_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_IB_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_IB_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_IB_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_DUMMY_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_INT_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_VM_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS2)
+};
+
 static void sdma_v5_0_set_ring_funcs(struct amdgpu_device *adev);
 static void sdma_v5_0_set_buffer_funcs(struct amdgpu_device *adev);
 static void sdma_v5_0_set_vm_pte_funcs(struct amdgpu_device *adev);
 static void sdma_v5_0_set_irq_funcs(struct amdgpu_device *adev);
+static int sdma_v5_0_stop_queue(struct amdgpu_ring *ring);
+static int sdma_v5_0_restore_queue(struct amdgpu_ring *ring);
 
 static const struct soc15_reg_golden golden_settings_sdma_5[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmSDMA0_CHICKEN_BITS, 0xffbf1f0f, 0x03ab0107),
@@ -184,7 +235,7 @@ static u32 sdma_v5_0_get_reg_offset(struct amdgpu_device *adev, u32 instance, u3
 
 static void sdma_v5_0_init_golden_registers(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[SDMA0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
 	case IP_VERSION(5, 0, 0):
 		soc15_program_register_sequence(adev,
 						golden_settings_sdma_5,
@@ -237,49 +288,35 @@ static void sdma_v5_0_init_golden_registers(struct amdgpu_device *adev)
 // emulation only, won't work on real chip
 // navi10 real chip need to use PSP to load firmware
 static int sdma_v5_0_init_microcode(struct amdgpu_device *adev)
-{	int ret, i;
-
-	if (amdgpu_sriov_vf(adev) && (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(5, 0, 5)))
-		return 0;
+{
+	int ret, i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		ret = amdgpu_sdma_init_microcode(adev, i, false);
 		if (ret)
 			return ret;
 	}
-	
+
 	return ret;
 }
 
-static unsigned sdma_v5_0_ring_init_cond_exec(struct amdgpu_ring *ring)
+static unsigned sdma_v5_0_ring_init_cond_exec(struct amdgpu_ring *ring,
+					      uint64_t addr)
 {
 	unsigned ret;
 
 	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_COND_EXE));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
 	amdgpu_ring_write(ring, 1);
-	ret = ring->wptr & ring->buf_mask;/* this is the offset we need patch later */
-	amdgpu_ring_write(ring, 0x55aa55aa);/* insert dummy here and patch it later */
+	/* this is the offset we need patch later */
+	ret = ring->wptr & ring->buf_mask;
+	/* insert dummy here and patch it later */
+	amdgpu_ring_write(ring, 0);
 
 	return ret;
 }
 
-static void sdma_v5_0_ring_patch_cond_exec(struct amdgpu_ring *ring,
-					   unsigned offset)
-{
-	unsigned cur;
-
-	BUG_ON(offset > ring->buf_mask);
-	BUG_ON(ring->ring[offset] != 0x55aa55aa);
-
-	cur = (ring->wptr - 1) & ring->buf_mask;
-	if (cur > offset)
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->buf_mask + 1) - offset + cur;
-}
-
 /**
  * sdma_v5_0_ring_get_rptr - get the current read pointer
  *
@@ -334,67 +371,36 @@ static uint64_t sdma_v5_0_ring_get_wptr(struct amdgpu_ring *ring)
 static void sdma_v5_0_ring_set_wptr(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	uint32_t *wptr_saved;
-	uint32_t *is_queue_unmap;
-	uint64_t aggregated_db_index;
-	uint32_t mqd_size = adev->mqds[AMDGPU_HW_IP_DMA].mqd_size;
 
 	DRM_DEBUG("Setting write pointer\n");
-	if (ring->is_mes_queue) {
-		wptr_saved = (uint32_t *)(ring->mqd_ptr + mqd_size);
-		is_queue_unmap = (uint32_t *)(ring->mqd_ptr + mqd_size +
-					      sizeof(uint32_t));
-		aggregated_db_index =
-			amdgpu_mes_get_aggregated_doorbell_index(adev,
-			AMDGPU_MES_PRIORITY_LEVEL_NORMAL);
-
+	if (ring->use_doorbell) {
+		DRM_DEBUG("Using doorbell -- "
+			  "wptr_offs == 0x%08x "
+			  "lower_32_bits(ring->wptr) << 2 == 0x%08x "
+			  "upper_32_bits(ring->wptr) << 2 == 0x%08x\n",
+			  ring->wptr_offs,
+			  lower_32_bits(ring->wptr << 2),
+			  upper_32_bits(ring->wptr << 2));
+		/* XXX check if swapping is necessary on BE */
 		atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
 			     ring->wptr << 2);
-		*wptr_saved = ring->wptr << 2;
-		if (*is_queue_unmap) {
-			WDOORBELL64(aggregated_db_index, ring->wptr << 2);
-			DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n",
-					ring->doorbell_index, ring->wptr << 2);
-			WDOORBELL64(ring->doorbell_index, ring->wptr << 2);
-		} else {
-			DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n",
-					ring->doorbell_index, ring->wptr << 2);
-			WDOORBELL64(ring->doorbell_index, ring->wptr << 2);
-
-			if (*is_queue_unmap)
-				WDOORBELL64(aggregated_db_index,
-					    ring->wptr << 2);
-		}
+		DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n",
+			  ring->doorbell_index, ring->wptr << 2);
+		WDOORBELL64(ring->doorbell_index, ring->wptr << 2);
 	} else {
-		if (ring->use_doorbell) {
-			DRM_DEBUG("Using doorbell -- "
-				  "wptr_offs == 0x%08x "
-				  "lower_32_bits(ring->wptr) << 2 == 0x%08x "
-				  "upper_32_bits(ring->wptr) << 2 == 0x%08x\n",
-				  ring->wptr_offs,
-				  lower_32_bits(ring->wptr << 2),
-				  upper_32_bits(ring->wptr << 2));
-			/* XXX check if swapping is necessary on BE */
-			atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
-				     ring->wptr << 2);
-			DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n",
-				  ring->doorbell_index, ring->wptr << 2);
-			WDOORBELL64(ring->doorbell_index, ring->wptr << 2);
-		} else {
-			DRM_DEBUG("Not using doorbell -- "
-				  "mmSDMA%i_GFX_RB_WPTR == 0x%08x "
-				  "mmSDMA%i_GFX_RB_WPTR_HI == 0x%08x\n",
-				  ring->me,
-				  lower_32_bits(ring->wptr << 2),
-				  ring->me,
-				  upper_32_bits(ring->wptr << 2));
-			WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev,
-					     ring->me, mmSDMA0_GFX_RB_WPTR),
-					lower_32_bits(ring->wptr << 2));
-			WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev,
-					     ring->me, mmSDMA0_GFX_RB_WPTR_HI),
-					upper_32_bits(ring->wptr << 2));
-		}
+		DRM_DEBUG("Not using doorbell -- "
+			  "mmSDMA%i_GFX_RB_WPTR == 0x%08x "
+			  "mmSDMA%i_GFX_RB_WPTR_HI == 0x%08x\n",
+			  ring->me,
+			  lower_32_bits(ring->wptr << 2),
+			  ring->me,
+			  upper_32_bits(ring->wptr << 2));
+		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev,
+							     ring->me, mmSDMA0_GFX_RB_WPTR),
+				lower_32_bits(ring->wptr << 2));
+		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev,
+							     ring->me, mmSDMA0_GFX_RB_WPTR_HI),
+				upper_32_bits(ring->wptr << 2));
 	}
 }
 
@@ -540,11 +546,9 @@ static void sdma_v5_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 se
 	}
 
 	if (flags & AMDGPU_FENCE_FLAG_INT) {
-		uint32_t ctx = ring->is_mes_queue ?
-			(ring->hw_queue_id | AMDGPU_FENCE_MES_QUEUE_FLAG) : 0;
 		/* generate an interrupt */
 		amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_TRAP));
-		amdgpu_ring_write(ring, SDMA_PKT_TRAP_INT_CONTEXT_INT_CONTEXT(ctx));
+		amdgpu_ring_write(ring, SDMA_PKT_TRAP_INT_CONTEXT_INT_CONTEXT(0));
 	}
 }
 
@@ -553,17 +557,15 @@ static void sdma_v5_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 se
  * sdma_v5_0_gfx_stop - stop the gfx async dma engines
  *
  * @adev: amdgpu_device pointer
- *
+ * @inst_mask: mask of dma engine instances to be disabled
  * Stop the gfx async dma ring buffers (NAVI10).
  */
-static void sdma_v5_0_gfx_stop(struct amdgpu_device *adev)
+static void sdma_v5_0_gfx_stop(struct amdgpu_device *adev, uint32_t inst_mask)
 {
 	u32 rb_cntl, ib_cntl;
 	int i;
 
-	amdgpu_sdma_unset_buffer_funcs_helper(adev);
-
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	for_each_inst(i, inst_mask) {
 		rb_cntl = RREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL));
 		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 0);
 		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL), rb_cntl);
@@ -655,9 +657,11 @@ static void sdma_v5_0_enable(struct amdgpu_device *adev, bool enable)
 {
 	u32 f32_cntl;
 	int i;
+	uint32_t inst_mask;
 
+	inst_mask = GENMASK(adev->sdma.num_instances - 1, 0);
 	if (!enable) {
-		sdma_v5_0_gfx_stop(adev);
+		sdma_v5_0_gfx_stop(adev, 1 << inst_mask);
 		sdma_v5_0_rlc_stop(adev);
 	}
 
@@ -672,14 +676,16 @@ static void sdma_v5_0_enable(struct amdgpu_device *adev, bool enable)
 }
 
 /**
- * sdma_v5_0_gfx_resume - setup and start the async dma engines
+ * sdma_v5_0_gfx_resume_instance - start/restart a certain sdma engine
  *
  * @adev: amdgpu_device pointer
+ * @i: instance
+ * @restore: used to restore wptr when restart
  *
- * Set up the gfx DMA ring buffers and enable them (NAVI10).
- * Returns 0 for success, error for failure.
+ * Set up the gfx DMA ring buffers and enable them. On restart, we will restore wptr and rptr.
+ * Return 0 for success.
  */
-static int sdma_v5_0_gfx_resume(struct amdgpu_device *adev)
+static int sdma_v5_0_gfx_resume_instance(struct amdgpu_device *adev, int i, bool restore)
 {
 	struct amdgpu_ring *ring;
 	u32 rb_cntl, ib_cntl;
@@ -689,149 +695,165 @@ static int sdma_v5_0_gfx_resume(struct amdgpu_device *adev)
 	u32 temp;
 	u32 wptr_poll_cntl;
 	u64 wptr_gpu_addr;
-	int i, r;
 
-	for (i = 0; i < adev->sdma.num_instances; i++) {
-		ring = &adev->sdma.instance[i].ring;
+	ring = &adev->sdma.instance[i].ring;
 
-		if (!amdgpu_sriov_vf(adev))
-			WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_SEM_WAIT_FAIL_TIMER_CNTL), 0);
+	if (!amdgpu_sriov_vf(adev))
+		WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_SEM_WAIT_FAIL_TIMER_CNTL), 0);
 
-		/* Set ring buffer size in dwords */
-		rb_bufsz = order_base_2(ring->ring_size / 4);
-		rb_cntl = RREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL));
-		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_SIZE, rb_bufsz);
+	/* Set ring buffer size in dwords */
+	rb_bufsz = order_base_2(ring->ring_size / 4);
+	rb_cntl = RREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL));
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_SIZE, rb_bufsz);
 #ifdef __BIG_ENDIAN
-		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_SWAP_ENABLE, 1);
-		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL,
-					RPTR_WRITEBACK_SWAP_ENABLE, 1);
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_SWAP_ENABLE, 1);
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL,
+				RPTR_WRITEBACK_SWAP_ENABLE, 1);
 #endif
-		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL), rb_cntl);
-
-		/* Initialize the ring buffer's read and write pointers */
+	WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL), rb_cntl);
+
+	/* Initialize the ring buffer's read and write pointers */
+	if (restore) {
+		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR), lower_32_bits(ring->wptr << 2));
+		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_HI), upper_32_bits(ring->wptr << 2));
+		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR), lower_32_bits(ring->wptr << 2));
+		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_HI), upper_32_bits(ring->wptr << 2));
+	} else {
 		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR), 0);
 		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_HI), 0);
 		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR), 0);
 		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_HI), 0);
-
-		/* setup the wptr shadow polling */
-		wptr_gpu_addr = ring->wptr_gpu_addr;
-		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_LO),
-		       lower_32_bits(wptr_gpu_addr));
-		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_HI),
-		       upper_32_bits(wptr_gpu_addr));
-		wptr_poll_cntl = RREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i,
-							 mmSDMA0_GFX_RB_WPTR_POLL_CNTL));
-		wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl,
-					       SDMA0_GFX_RB_WPTR_POLL_CNTL,
-					       F32_POLL_ENABLE, 1);
-		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_CNTL),
-		       wptr_poll_cntl);
-
-		/* set the wb address whether it's enabled or not */
-		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_ADDR_HI),
-		       upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF);
-		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_ADDR_LO),
-		       lower_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFC);
-
-		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1);
-
-		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_BASE),
-		       ring->gpu_addr >> 8);
-		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_BASE_HI),
-		       ring->gpu_addr >> 40);
-
+	}
+	/* setup the wptr shadow polling */
+	wptr_gpu_addr = ring->wptr_gpu_addr;
+	WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_LO),
+	       lower_32_bits(wptr_gpu_addr));
+	WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_HI),
+	       upper_32_bits(wptr_gpu_addr));
+	wptr_poll_cntl = RREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i,
+						 mmSDMA0_GFX_RB_WPTR_POLL_CNTL));
+	wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl,
+				       SDMA0_GFX_RB_WPTR_POLL_CNTL,
+				       F32_POLL_ENABLE, 1);
+	WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_CNTL),
+	       wptr_poll_cntl);
+
+	/* set the wb address whether it's enabled or not */
+	WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_ADDR_HI),
+	       upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF);
+	WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_ADDR_LO),
+	       lower_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFC);
+
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1);
+
+	WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_BASE),
+	       ring->gpu_addr >> 8);
+	WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_BASE_HI),
+	       ring->gpu_addr >> 40);
+
+	if (!restore)
 		ring->wptr = 0;
 
-		/* before programing wptr to a less value, need set minor_ptr_update first */
-		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_MINOR_PTR_UPDATE), 1);
+	/* before programing wptr to a less value, need set minor_ptr_update first */
+	WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_MINOR_PTR_UPDATE), 1);
 
-		if (!amdgpu_sriov_vf(adev)) { /* only bare-metal use register write for wptr */
-			WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR),
-			       lower_32_bits(ring->wptr << 2));
-			WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_HI),
-			       upper_32_bits(ring->wptr << 2));
-		}
+	if (!amdgpu_sriov_vf(adev)) { /* only bare-metal use register write for wptr */
+		WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR),
+		       lower_32_bits(ring->wptr << 2));
+		WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_HI),
+		       upper_32_bits(ring->wptr << 2));
+	}
 
-		doorbell = RREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL));
-		doorbell_offset = RREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i,
-						mmSDMA0_GFX_DOORBELL_OFFSET));
+	doorbell = RREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL));
+	doorbell_offset = RREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i,
+					mmSDMA0_GFX_DOORBELL_OFFSET));
 
-		if (ring->use_doorbell) {
-			doorbell = REG_SET_FIELD(doorbell, SDMA0_GFX_DOORBELL, ENABLE, 1);
-			doorbell_offset = REG_SET_FIELD(doorbell_offset, SDMA0_GFX_DOORBELL_OFFSET,
-					OFFSET, ring->doorbell_index);
-		} else {
-			doorbell = REG_SET_FIELD(doorbell, SDMA0_GFX_DOORBELL, ENABLE, 0);
-		}
-		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL), doorbell);
-		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL_OFFSET),
-		       doorbell_offset);
+	if (ring->use_doorbell) {
+		doorbell = REG_SET_FIELD(doorbell, SDMA0_GFX_DOORBELL, ENABLE, 1);
+		doorbell_offset = REG_SET_FIELD(doorbell_offset, SDMA0_GFX_DOORBELL_OFFSET,
+				OFFSET, ring->doorbell_index);
+	} else {
+		doorbell = REG_SET_FIELD(doorbell, SDMA0_GFX_DOORBELL, ENABLE, 0);
+	}
+	WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL), doorbell);
+	WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL_OFFSET),
+	       doorbell_offset);
 
-		adev->nbio.funcs->sdma_doorbell_range(adev, i, ring->use_doorbell,
-						      ring->doorbell_index, 20);
+	adev->nbio.funcs->sdma_doorbell_range(adev, i, ring->use_doorbell,
+					      ring->doorbell_index, 20);
 
-		if (amdgpu_sriov_vf(adev))
-			sdma_v5_0_ring_set_wptr(ring);
+	if (amdgpu_sriov_vf(adev))
+		sdma_v5_0_ring_set_wptr(ring);
 
-		/* set minor_ptr_update to 0 after wptr programed */
-		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_MINOR_PTR_UPDATE), 0);
+	/* set minor_ptr_update to 0 after wptr programed */
+	WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_MINOR_PTR_UPDATE), 0);
 
-		if (!amdgpu_sriov_vf(adev)) {
-			/* set utc l1 enable flag always to 1 */
-			temp = RREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_CNTL));
-			temp = REG_SET_FIELD(temp, SDMA0_CNTL, UTC_L1_ENABLE, 1);
-
-			/* enable MCBP */
-			temp = REG_SET_FIELD(temp, SDMA0_CNTL, MIDCMD_PREEMPT_ENABLE, 1);
-			WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_CNTL), temp);
-
-			/* Set up RESP_MODE to non-copy addresses */
-			temp = RREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_UTCL1_CNTL));
-			temp = REG_SET_FIELD(temp, SDMA0_UTCL1_CNTL, RESP_MODE, 3);
-			temp = REG_SET_FIELD(temp, SDMA0_UTCL1_CNTL, REDO_DELAY, 9);
-			WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_UTCL1_CNTL), temp);
-
-			/* program default cache read and write policy */
-			temp = RREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_UTCL1_PAGE));
-			/* clean read policy and write policy bits */
-			temp &= 0xFF0FFF;
-			temp |= ((CACHE_READ_POLICY_L2__DEFAULT << 12) | (CACHE_WRITE_POLICY_L2__DEFAULT << 14));
-			WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_UTCL1_PAGE), temp);
-		}
+	if (!amdgpu_sriov_vf(adev)) {
+		/* set utc l1 enable flag always to 1 */
+		temp = RREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_CNTL));
+		temp = REG_SET_FIELD(temp, SDMA0_CNTL, UTC_L1_ENABLE, 1);
+
+		/* enable MCBP */
+		temp = REG_SET_FIELD(temp, SDMA0_CNTL, MIDCMD_PREEMPT_ENABLE, 1);
+		WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_CNTL), temp);
+
+		/* Set up RESP_MODE to non-copy addresses */
+		temp = RREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_UTCL1_CNTL));
+		temp = REG_SET_FIELD(temp, SDMA0_UTCL1_CNTL, RESP_MODE, 3);
+		temp = REG_SET_FIELD(temp, SDMA0_UTCL1_CNTL, REDO_DELAY, 9);
+		WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_UTCL1_CNTL), temp);
+
+		/* program default cache read and write policy */
+		temp = RREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_UTCL1_PAGE));
+		/* clean read policy and write policy bits */
+		temp &= 0xFF0FFF;
+		temp |= ((CACHE_READ_POLICY_L2__DEFAULT << 12) | (CACHE_WRITE_POLICY_L2__DEFAULT << 14));
+		WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_UTCL1_PAGE), temp);
+	}
 
-		if (!amdgpu_sriov_vf(adev)) {
-			/* unhalt engine */
-			temp = RREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_F32_CNTL));
-			temp = REG_SET_FIELD(temp, SDMA0_F32_CNTL, HALT, 0);
-			WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_F32_CNTL), temp);
-		}
+	if (!amdgpu_sriov_vf(adev)) {
+		/* unhalt engine */
+		temp = RREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_F32_CNTL));
+		temp = REG_SET_FIELD(temp, SDMA0_F32_CNTL, HALT, 0);
+		WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_F32_CNTL), temp);
+	}
 
-		/* enable DMA RB */
-		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 1);
-		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL), rb_cntl);
+	/* enable DMA RB */
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 1);
+	WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL), rb_cntl);
 
-		ib_cntl = RREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL));
-		ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_ENABLE, 1);
+	ib_cntl = RREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL));
+	ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_ENABLE, 1);
 #ifdef __BIG_ENDIAN
-		ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_SWAP_ENABLE, 1);
+	ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_SWAP_ENABLE, 1);
 #endif
-		/* enable DMA IBs */
-		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL), ib_cntl);
+	/* enable DMA IBs */
+	WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL), ib_cntl);
 
-		ring->sched.ready = true;
+	if (amdgpu_sriov_vf(adev)) { /* bare-metal sequence doesn't need below to lines */
+		sdma_v5_0_ctx_switch_enable(adev, true);
+		sdma_v5_0_enable(adev, true);
+	}
 
-		if (amdgpu_sriov_vf(adev)) { /* bare-metal sequence doesn't need below to lines */
-			sdma_v5_0_ctx_switch_enable(adev, true);
-			sdma_v5_0_enable(adev, true);
-		}
+	return amdgpu_ring_test_helper(ring);
+}
+
+/**
+ * sdma_v5_0_gfx_resume - setup and start the async dma engines
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set up the gfx DMA ring buffers and enable them (NAVI10).
+ * Returns 0 for success, error for failure.
+ */
+static int sdma_v5_0_gfx_resume(struct amdgpu_device *adev)
+{
+	int i, r;
 
-		r = amdgpu_ring_test_helper(ring);
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		r = sdma_v5_0_gfx_resume_instance(adev, i, false);
 		if (r)
 			return r;
-
-		if (adev->mman.buffer_funcs_ring == ring)
-			amdgpu_ttm_set_buffer_funcs_status(adev, true);
 	}
 
 	return 0;
@@ -995,28 +1017,18 @@ static int sdma_v5_0_ring_test_ring(struct amdgpu_ring *ring)
 	int r;
 	u32 tmp;
 	u64 gpu_addr;
-	volatile uint32_t *cpu_ptr = NULL;
 
 	tmp = 0xCAFEDEAD;
 
-	if (ring->is_mes_queue) {
-		uint32_t offset = 0;
-		offset = amdgpu_mes_ctx_get_offs(ring,
-					 AMDGPU_MES_CTX_PADDING_OFFS);
-		gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		cpu_ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-		*cpu_ptr = tmp;
-	} else {
-		r = amdgpu_device_wb_get(adev, &index);
-		if (r) {
-			dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r);
-			return r;
-		}
-
-		gpu_addr = adev->wb.gpu_addr + (index * 4);
-		adev->wb.wb[index] = cpu_to_le32(tmp);
+	r = amdgpu_device_wb_get(adev, &index);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r);
+		return r;
 	}
 
+	gpu_addr = adev->wb.gpu_addr + (index * 4);
+	adev->wb.wb[index] = cpu_to_le32(tmp);
+
 	r = amdgpu_ring_alloc(ring, 20);
 	if (r) {
 		DRM_ERROR("amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, r);
@@ -1033,10 +1045,7 @@ static int sdma_v5_0_ring_test_ring(struct amdgpu_ring *ring)
 	amdgpu_ring_commit(ring);
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		if (ring->is_mes_queue)
-			tmp = le32_to_cpu(*cpu_ptr);
-		else
-			tmp = le32_to_cpu(adev->wb.wb[index]);
+		tmp = le32_to_cpu(adev->wb.wb[index]);
 		if (tmp == 0xDEADBEEF)
 			break;
 		if (amdgpu_emu_mode == 1)
@@ -1048,8 +1057,7 @@ static int sdma_v5_0_ring_test_ring(struct amdgpu_ring *ring)
 	if (i >= adev->usec_timeout)
 		r = -ETIMEDOUT;
 
-	if (!ring->is_mes_queue)
-		amdgpu_device_wb_free(adev, index);
+	amdgpu_device_wb_free(adev, index);
 
 	return r;
 }
@@ -1072,38 +1080,24 @@ static int sdma_v5_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	long r;
 	u32 tmp = 0;
 	u64 gpu_addr;
-	volatile uint32_t *cpu_ptr = NULL;
 
 	tmp = 0xCAFEDEAD;
 	memset(&ib, 0, sizeof(ib));
 
-	if (ring->is_mes_queue) {
-		uint32_t offset = 0;
-		offset = amdgpu_mes_ctx_get_offs(ring, AMDGPU_MES_CTX_IB_OFFS);
-		ib.gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		ib.ptr = (void *)amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-
-		offset = amdgpu_mes_ctx_get_offs(ring,
-					 AMDGPU_MES_CTX_PADDING_OFFS);
-		gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		cpu_ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-		*cpu_ptr = tmp;
-	} else {
-		r = amdgpu_device_wb_get(adev, &index);
-		if (r) {
-			dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r);
-			return r;
-		}
+	r = amdgpu_device_wb_get(adev, &index);
+	if (r) {
+		dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r);
+		return r;
+	}
 
-		gpu_addr = adev->wb.gpu_addr + (index * 4);
-		adev->wb.wb[index] = cpu_to_le32(tmp);
+	gpu_addr = adev->wb.gpu_addr + (index * 4);
+	adev->wb.wb[index] = cpu_to_le32(tmp);
 
-		r = amdgpu_ib_get(adev, NULL, 256,
-					AMDGPU_IB_POOL_DIRECT, &ib);
-		if (r) {
-			DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
-			goto err0;
-		}
+	r = amdgpu_ib_get(adev, NULL, 256,
+			  AMDGPU_IB_POOL_DIRECT, &ib);
+	if (r) {
+		DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
+		goto err0;
 	}
 
 	ib.ptr[0] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) |
@@ -1131,10 +1125,7 @@ static int sdma_v5_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		goto err1;
 	}
 
-	if (ring->is_mes_queue)
-		tmp = le32_to_cpu(*cpu_ptr);
-	else
-		tmp = le32_to_cpu(adev->wb.wb[index]);
+	tmp = le32_to_cpu(adev->wb.wb[index]);
 
 	if (tmp == 0xDEADBEEF)
 		r = 0;
@@ -1142,11 +1133,10 @@ static int sdma_v5_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = -EINVAL;
 
 err1:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err0:
-	if (!ring->is_mes_queue)
-		amdgpu_device_wb_free(adev, index);
+	amdgpu_device_wb_free(adev, index);
 	return r;
 }
 
@@ -1337,9 +1327,44 @@ static void sdma_v5_0_ring_emit_reg_write_reg_wait(struct amdgpu_ring *ring,
 	amdgpu_ring_emit_reg_wait(ring, reg1, mask, mask);
 }
 
-static int sdma_v5_0_early_init(void *handle)
+static int sdma_v5_0_soft_reset_engine(struct amdgpu_device *adev, u32 instance_id)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	u32 grbm_soft_reset;
+	u32 tmp;
+
+	grbm_soft_reset = REG_SET_FIELD(0,
+					GRBM_SOFT_RESET, SOFT_RESET_SDMA0,
+					1);
+	grbm_soft_reset <<= instance_id;
+
+	tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
+	tmp |= grbm_soft_reset;
+	DRM_DEBUG("GRBM_SOFT_RESET=0x%08X\n", tmp);
+	WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
+	tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
+
+	udelay(50);
+
+	tmp &= ~grbm_soft_reset;
+	WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
+	tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
+	return 0;
+}
+
+static const struct amdgpu_sdma_funcs sdma_v5_0_sdma_funcs = {
+	.stop_kernel_queue = &sdma_v5_0_stop_queue,
+	.start_kernel_queue = &sdma_v5_0_restore_queue,
+	.soft_reset_kernel_queue = &sdma_v5_0_soft_reset_engine,
+};
+
+static int sdma_v5_0_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = sdma_v5_0_init_microcode(adev);
+	if (r)
+		return r;
 
 	sdma_v5_0_set_ring_funcs(adev);
 	sdma_v5_0_set_buffer_funcs(adev);
@@ -1351,11 +1376,13 @@ static int sdma_v5_0_early_init(void *handle)
 }
 
 
-static int sdma_v5_0_sw_init(void *handle)
+static int sdma_v5_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
 	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_5_0);
+	uint32_t *ptr;
 
 	/* SDMA trap event */
 	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_SDMA0,
@@ -1371,13 +1398,9 @@ static int sdma_v5_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	r = sdma_v5_0_init_microcode(adev);
-	if (r) {
-		DRM_ERROR("Failed to load sdma firmware!\n");
-		return r;
-	}
-
 	for (i = 0; i < adev->sdma.num_instances; i++) {
+		mutex_init(&adev->sdma.instance[i].engine_reset_mutex);
+		adev->sdma.instance[i].funcs = &sdma_v5_0_sdma_funcs;
 		ring = &adev->sdma.instance[i].ring;
 		ring->ring_obj = NULL;
 		ring->use_doorbell = true;
@@ -1389,7 +1412,7 @@ static int sdma_v5_0_sw_init(void *handle)
 			(adev->doorbell_index.sdma_engine[0] << 1) //get DWORD offset
 			: (adev->doorbell_index.sdma_engine[1] << 1); // get DWORD offset
 
-		ring->vm_hub = AMDGPU_GFXHUB_0;
+		ring->vm_hub = AMDGPU_GFXHUB(0);
 		sprintf(ring->name, "sdma%d", i);
 		r = amdgpu_ring_init(adev, ring, 1024, &adev->sdma.trap_irq,
 				     (i == 0) ? AMDGPU_SDMA_IRQ_INSTANCE0 :
@@ -1399,26 +1422,55 @@ static int sdma_v5_0_sw_init(void *handle)
 			return r;
 	}
 
+	adev->sdma.supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->sdma.instance[0].ring);
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
+	case IP_VERSION(5, 0, 0):
+	case IP_VERSION(5, 0, 2):
+	case IP_VERSION(5, 0, 5):
+		if ((adev->sdma.instance[0].fw_version >= 35) &&
+		    !amdgpu_sriov_vf(adev) &&
+		    !adev->debug_disable_gpu_ring_reset)
+			adev->sdma.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+		break;
+	default:
+		break;
+	}
+
+	/* Allocate memory for SDMA IP Dump buffer */
+	ptr = kcalloc(adev->sdma.num_instances * reg_count, sizeof(uint32_t), GFP_KERNEL);
+	if (ptr)
+		adev->sdma.ip_dump = ptr;
+	else
+		DRM_ERROR("Failed to allocated memory for SDMA IP Dump\n");
+
+	r = amdgpu_sdma_sysfs_reset_mask_init(adev);
+	if (r)
+		return r;
+
 	return r;
 }
 
-static int sdma_v5_0_sw_fini(void *handle)
+static int sdma_v5_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++)
 		amdgpu_ring_fini(&adev->sdma.instance[i].ring);
 
+	amdgpu_sdma_sysfs_reset_mask_fini(adev);
 	amdgpu_sdma_destroy_inst_ctx(adev, false);
 
+	kfree(adev->sdma.ip_dump);
+
 	return 0;
 }
 
-static int sdma_v5_0_hw_init(void *handle)
+static int sdma_v5_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	sdma_v5_0_init_golden_registers(adev);
 
@@ -1427,15 +1479,12 @@ static int sdma_v5_0_hw_init(void *handle)
 	return r;
 }
 
-static int sdma_v5_0_hw_fini(void *handle)
+static int sdma_v5_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	if (amdgpu_sriov_vf(adev)) {
-		/* disable the scheduler for SDMA */
-		amdgpu_sdma_unset_buffer_funcs_helper(adev);
+	if (amdgpu_sriov_vf(adev))
 		return 0;
-	}
 
 	sdma_v5_0_ctx_switch_enable(adev, false);
 	sdma_v5_0_enable(adev, false);
@@ -1443,23 +1492,19 @@ static int sdma_v5_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int sdma_v5_0_suspend(void *handle)
+static int sdma_v5_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return sdma_v5_0_hw_fini(adev);
+	return sdma_v5_0_hw_fini(ip_block);
 }
 
-static int sdma_v5_0_resume(void *handle)
+static int sdma_v5_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return sdma_v5_0_hw_init(adev);
+	return sdma_v5_0_hw_init(ip_block);
 }
 
-static bool sdma_v5_0_is_idle(void *handle)
+static bool sdma_v5_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
@@ -1472,11 +1517,11 @@ static bool sdma_v5_0_is_idle(void *handle)
 	return true;
 }
 
-static int sdma_v5_0_wait_for_idle(void *handle)
+static int sdma_v5_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
 	u32 sdma0, sdma1;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		sdma0 = RREG32(sdma_v5_0_get_reg_offset(adev, 0, mmSDMA0_STATUS_REG));
@@ -1489,13 +1534,104 @@ static int sdma_v5_0_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int sdma_v5_0_soft_reset(void *handle)
+static int sdma_v5_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	/* todo */
 
 	return 0;
 }
 
+static int sdma_v5_0_reset_queue(struct amdgpu_ring *ring,
+				 unsigned int vmid,
+				 struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int r;
+
+	if (ring->me >= adev->sdma.num_instances) {
+		dev_err(adev->dev, "sdma instance not found\n");
+		return -EINVAL;
+	}
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	amdgpu_amdkfd_suspend(adev, true);
+	r = amdgpu_sdma_reset_engine(adev, ring->me, true);
+	amdgpu_amdkfd_resume(adev, true);
+	if (r)
+		return r;
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static int sdma_v5_0_stop_queue(struct amdgpu_ring *ring)
+{
+	u32 f32_cntl, freeze, cntl, stat1_reg;
+	struct amdgpu_device *adev = ring->adev;
+	int i, j, r = 0;
+
+	if (amdgpu_sriov_vf(adev))
+		return -EINVAL;
+
+	i = ring->me;
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
+
+	/* stop queue */
+	sdma_v5_0_gfx_stop(adev, 1 << i);
+
+	/* engine stop SDMA1_F32_CNTL.HALT to 1 and SDMAx_FREEZE freeze bit to 1 */
+	freeze = RREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_FREEZE));
+	freeze = REG_SET_FIELD(freeze, SDMA0_FREEZE, FREEZE, 1);
+	WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_FREEZE), freeze);
+
+	for (j = 0; j < adev->usec_timeout; j++) {
+		freeze = RREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_FREEZE));
+		if (REG_GET_FIELD(freeze, SDMA0_FREEZE, FROZEN) & 1)
+			break;
+		udelay(1);
+	}
+
+	/* check sdma copy engine all idle if frozen not received*/
+	if (j == adev->usec_timeout) {
+		stat1_reg = RREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_STATUS1_REG));
+		if ((stat1_reg & 0x3FF) != 0x3FF) {
+			DRM_ERROR("cannot soft reset as sdma not idle\n");
+			r = -ETIMEDOUT;
+			goto err0;
+		}
+	}
+
+	f32_cntl = RREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_F32_CNTL));
+	f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_F32_CNTL, HALT, 1);
+	WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_F32_CNTL), f32_cntl);
+
+	cntl = RREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_CNTL));
+	cntl = REG_SET_FIELD(cntl, SDMA0_CNTL, UTC_L1_ENABLE, 0);
+	WREG32(sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_CNTL), cntl);
+err0:
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
+	return r;
+}
+
+static int sdma_v5_0_restore_queue(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 inst_id = ring->me;
+	u32 freeze;
+	int r;
+
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
+	/* unfreeze*/
+	freeze = RREG32(sdma_v5_0_get_reg_offset(adev, inst_id, mmSDMA0_FREEZE));
+	freeze = REG_SET_FIELD(freeze, SDMA0_FREEZE, FREEZE, 0);
+	WREG32(sdma_v5_0_get_reg_offset(adev, inst_id, mmSDMA0_FREEZE), freeze);
+
+	r = sdma_v5_0_gfx_resume_instance(adev, inst_id, true);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
+
+	return r;
+}
+
 static int sdma_v5_0_ring_preempt_ib(struct amdgpu_ring *ring)
 {
 	int i, r = 0;
@@ -1693,15 +1829,15 @@ static void sdma_v5_0_update_medium_grain_light_sleep(struct amdgpu_device *adev
 	}
 }
 
-static int sdma_v5_0_set_clockgating_state(void *handle,
+static int sdma_v5_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[SDMA0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
 	case IP_VERSION(5, 0, 0):
 	case IP_VERSION(5, 0, 2):
 	case IP_VERSION(5, 0, 5):
@@ -1717,15 +1853,15 @@ static int sdma_v5_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int sdma_v5_0_set_powergating_state(void *handle,
+static int sdma_v5_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
 }
 
-static void sdma_v5_0_get_clockgating_state(void *handle, u64 *flags)
+static void sdma_v5_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	if (amdgpu_sriov_vf(adev))
@@ -1742,10 +1878,51 @@ static void sdma_v5_0_get_clockgating_state(void *handle, u64 *flags)
 		*flags |= AMD_CG_SUPPORT_SDMA_LS;
 }
 
-const struct amd_ip_funcs sdma_v5_0_ip_funcs = {
+static void sdma_v5_0_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, j;
+	uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_5_0);
+	uint32_t instance_offset;
+
+	if (!adev->sdma.ip_dump)
+		return;
+
+	drm_printf(p, "num_instances:%d\n", adev->sdma.num_instances);
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		instance_offset = i * reg_count;
+		drm_printf(p, "\nInstance:%d\n", i);
+
+		for (j = 0; j < reg_count; j++)
+			drm_printf(p, "%-50s \t 0x%08x\n", sdma_reg_list_5_0[j].reg_name,
+				   adev->sdma.ip_dump[instance_offset + j]);
+	}
+}
+
+static void sdma_v5_0_dump_ip_state(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, j;
+	uint32_t instance_offset;
+	uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_5_0);
+
+	if (!adev->sdma.ip_dump)
+		return;
+
+	amdgpu_gfx_off_ctrl(adev, false);
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		instance_offset = i * reg_count;
+		for (j = 0; j < reg_count; j++)
+			adev->sdma.ip_dump[instance_offset + j] =
+				RREG32(sdma_v5_0_get_reg_offset(adev, i,
+				       sdma_reg_list_5_0[j].reg_offset));
+	}
+	amdgpu_gfx_off_ctrl(adev, true);
+}
+
+static const struct amd_ip_funcs sdma_v5_0_ip_funcs = {
 	.name = "sdma_v5_0",
 	.early_init = sdma_v5_0_early_init,
-	.late_init = NULL,
 	.sw_init = sdma_v5_0_sw_init,
 	.sw_fini = sdma_v5_0_sw_fini,
 	.hw_init = sdma_v5_0_hw_init,
@@ -1758,6 +1935,8 @@ const struct amd_ip_funcs sdma_v5_0_ip_funcs = {
 	.set_clockgating_state = sdma_v5_0_set_clockgating_state,
 	.set_powergating_state = sdma_v5_0_set_powergating_state,
 	.get_clockgating_state = sdma_v5_0_get_clockgating_state,
+	.dump_ip_state = sdma_v5_0_dump_ip_state,
+	.print_ip_state = sdma_v5_0_print_ip_state,
 };
 
 static const struct amdgpu_ring_funcs sdma_v5_0_ring_funcs = {
@@ -1793,8 +1972,8 @@ static const struct amdgpu_ring_funcs sdma_v5_0_ring_funcs = {
 	.emit_reg_wait = sdma_v5_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = sdma_v5_0_ring_emit_reg_write_reg_wait,
 	.init_cond_exec = sdma_v5_0_ring_init_cond_exec,
-	.patch_cond_exec = sdma_v5_0_ring_patch_cond_exec,
 	.preempt_ib = sdma_v5_0_ring_preempt_ib,
+	.reset = sdma_v5_0_reset_queue,
 };
 
 static void sdma_v5_0_set_ring_funcs(struct amdgpu_device *adev)
@@ -1831,7 +2010,7 @@ static void sdma_v5_0_set_irq_funcs(struct amdgpu_device *adev)
  * @src_offset: src GPU address
  * @dst_offset: dst GPU address
  * @byte_count: number of bytes to xfer
- * @tmz: if a secure copy should be used
+ * @copy_flags: copy flags for the buffers
  *
  * Copy GPU buffers using the DMA engine (NAVI10).
  * Used by the amdgpu ttm implementation to move pages if
@@ -1841,11 +2020,11 @@ static void sdma_v5_0_emit_copy_buffer(struct amdgpu_ib *ib,
 				       uint64_t src_offset,
 				       uint64_t dst_offset,
 				       uint32_t byte_count,
-				       bool tmz)
+				       uint32_t copy_flags)
 {
 	ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) |
 		SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR) |
-		SDMA_PKT_COPY_LINEAR_HEADER_TMZ(tmz ? 1 : 0);
+		SDMA_PKT_COPY_LINEAR_HEADER_TMZ((copy_flags & AMDGPU_COPY_FLAGS_TMZ) ? 1 : 0);
 	ib->ptr[ib->length_dw++] = byte_count - 1;
 	ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */
 	ib->ptr[ib->length_dw++] = lower_32_bits(src_offset);
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.h b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.h
index d4e3c2e696f6..2ab71f21755a 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.h
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.h
@@ -24,7 +24,6 @@
 #ifndef __SDMA_V5_0_H__
 #define __SDMA_V5_0_H__
 
-extern const struct amd_ip_funcs sdma_v5_0_ip_funcs;
 extern const struct amdgpu_ip_block_version sdma_v5_0_ip_block;
 
 #endif /* __SDMA_V5_0_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
index ca7e8757d78e..51101b0aa2fa 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
@@ -60,10 +60,61 @@ MODULE_FIRMWARE("amdgpu/sdma_5_2_7.bin");
 #define SDMA0_HYP_DEC_REG_END 0x5893
 #define SDMA1_HYP_DEC_REG_OFFSET 0x20
 
+static const struct amdgpu_hwip_reg_entry sdma_reg_list_5_2[] = {
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_STATUS_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_STATUS1_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_STATUS2_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_STATUS3_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UCODE_CHECKSUM),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RB_RPTR_FETCH_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RB_RPTR_FETCH),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UTCL1_RD_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UTCL1_WR_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UTCL1_RD_XNACK0),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UTCL1_RD_XNACK1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UTCL1_WR_XNACK0),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_UTCL1_WR_XNACK1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_RB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_RB_RPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_RB_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_IB_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_IB_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_IB_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_IB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_IB_SUB_REMAIN),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_GFX_DUMMY_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_RB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_RB_RPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_RB_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_IB_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_IB_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_IB_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_PAGE_DUMMY_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_RB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_RB_RPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_RB_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_IB_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_IB_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_IB_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_RLC0_DUMMY_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_INT_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSDMA0_VM_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS2)
+};
+
 static void sdma_v5_2_set_ring_funcs(struct amdgpu_device *adev);
 static void sdma_v5_2_set_buffer_funcs(struct amdgpu_device *adev);
 static void sdma_v5_2_set_vm_pte_funcs(struct amdgpu_device *adev);
 static void sdma_v5_2_set_irq_funcs(struct amdgpu_device *adev);
+static int sdma_v5_2_stop_queue(struct amdgpu_ring *ring);
+static int sdma_v5_2_restore_queue(struct amdgpu_ring *ring);
 
 static u32 sdma_v5_2_get_reg_offset(struct amdgpu_device *adev, u32 instance, u32 internal_offset)
 {
@@ -89,35 +140,23 @@ static u32 sdma_v5_2_get_reg_offset(struct amdgpu_device *adev, u32 instance, u3
 	return base + internal_offset;
 }
 
-static unsigned sdma_v5_2_ring_init_cond_exec(struct amdgpu_ring *ring)
+static unsigned sdma_v5_2_ring_init_cond_exec(struct amdgpu_ring *ring,
+					      uint64_t addr)
 {
 	unsigned ret;
 
 	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_COND_EXE));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
 	amdgpu_ring_write(ring, 1);
-	ret = ring->wptr & ring->buf_mask;/* this is the offset we need patch later */
-	amdgpu_ring_write(ring, 0x55aa55aa);/* insert dummy here and patch it later */
+	/* this is the offset we need patch later */
+	ret = ring->wptr & ring->buf_mask;
+	/* insert dummy here and patch it later */
+	amdgpu_ring_write(ring, 0);
 
 	return ret;
 }
 
-static void sdma_v5_2_ring_patch_cond_exec(struct amdgpu_ring *ring,
-					   unsigned offset)
-{
-	unsigned cur;
-
-	BUG_ON(offset > ring->buf_mask);
-	BUG_ON(ring->ring[offset] != 0x55aa55aa);
-
-	cur = (ring->wptr - 1) & ring->buf_mask;
-	if (cur > offset)
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->buf_mask + 1) - offset + cur;
-}
-
 /**
  * sdma_v5_2_ring_get_rptr - get the current read pointer
  *
@@ -188,6 +227,16 @@ static void sdma_v5_2_ring_set_wptr(struct amdgpu_ring *ring)
 		DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n",
 				ring->doorbell_index, ring->wptr << 2);
 		WDOORBELL64(ring->doorbell_index, ring->wptr << 2);
+		if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) == IP_VERSION(5, 2, 1)) {
+			/* SDMA seems to miss doorbells sometimes when powergating kicks in.
+			 * Updating the wptr directly will wake it. This is only safe because
+			 * we disallow gfxoff in begin_use() and then allow it again in end_use().
+			 */
+			WREG32(sdma_v5_2_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR),
+			       lower_32_bits(ring->wptr << 2));
+			WREG32(sdma_v5_2_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR_HI),
+			       upper_32_bits(ring->wptr << 2));
+		}
 	} else {
 		DRM_DEBUG("Not using doorbell -- "
 				"mmSDMA%i_GFX_RB_WPTR == 0x%08x "
@@ -292,17 +341,21 @@ static void sdma_v5_2_ring_emit_hdp_flush(struct amdgpu_ring *ring)
 	u32 ref_and_mask = 0;
 	const struct nbio_hdp_flush_reg *nbio_hf_reg = adev->nbio.hdp_flush_reg;
 
-	ref_and_mask = nbio_hf_reg->ref_and_mask_sdma0 << ring->me;
-
-	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_POLL_REGMEM) |
-			  SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(1) |
-			  SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3)); /* == */
-	amdgpu_ring_write(ring, (adev->nbio.funcs->get_hdp_flush_done_offset(adev)) << 2);
-	amdgpu_ring_write(ring, (adev->nbio.funcs->get_hdp_flush_req_offset(adev)) << 2);
-	amdgpu_ring_write(ring, ref_and_mask); /* reference */
-	amdgpu_ring_write(ring, ref_and_mask); /* mask */
-	amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) |
-			  SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(10)); /* retry count, poll interval */
+	if (ring->me > 1) {
+		amdgpu_hdp_flush(adev, ring);
+	} else {
+		ref_and_mask = nbio_hf_reg->ref_and_mask_sdma0 << ring->me;
+
+		amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_POLL_REGMEM) |
+				  SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(1) |
+				  SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3)); /* == */
+		amdgpu_ring_write(ring, (adev->nbio.funcs->get_hdp_flush_done_offset(adev)) << 2);
+		amdgpu_ring_write(ring, (adev->nbio.funcs->get_hdp_flush_req_offset(adev)) << 2);
+		amdgpu_ring_write(ring, ref_and_mask); /* reference */
+		amdgpu_ring_write(ring, ref_and_mask); /* mask */
+		amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) |
+				  SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(10)); /* retry count, poll interval */
+	}
 }
 
 /**
@@ -343,11 +396,9 @@ static void sdma_v5_2_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 se
 	}
 
 	if ((flags & AMDGPU_FENCE_FLAG_INT)) {
-		uint32_t ctx = ring->is_mes_queue ?
-			(ring->hw_queue_id | AMDGPU_FENCE_MES_QUEUE_FLAG) : 0;
 		/* generate an interrupt */
 		amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_TRAP));
-		amdgpu_ring_write(ring, SDMA_PKT_TRAP_INT_CONTEXT_INT_CONTEXT(ctx));
+		amdgpu_ring_write(ring, SDMA_PKT_TRAP_INT_CONTEXT_INT_CONTEXT(0));
 	}
 }
 
@@ -356,17 +407,15 @@ static void sdma_v5_2_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 se
  * sdma_v5_2_gfx_stop - stop the gfx async dma engines
  *
  * @adev: amdgpu_device pointer
- *
+ * @inst_mask: mask of dma engine instances to be disabled
  * Stop the gfx async dma ring buffers.
  */
-static void sdma_v5_2_gfx_stop(struct amdgpu_device *adev)
+static void sdma_v5_2_gfx_stop(struct amdgpu_device *adev,  uint32_t inst_mask)
 {
 	u32 rb_cntl, ib_cntl;
 	int i;
 
-	amdgpu_sdma_unset_buffer_funcs_helper(adev);
-
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	for_each_inst(i, inst_mask) {
 		rb_cntl = RREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL));
 		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 0);
 		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL), rb_cntl);
@@ -457,9 +506,11 @@ static void sdma_v5_2_enable(struct amdgpu_device *adev, bool enable)
 {
 	u32 f32_cntl;
 	int i;
+	uint32_t inst_mask;
 
+	inst_mask = GENMASK(adev->sdma.num_instances - 1, 0);
 	if (!enable) {
-		sdma_v5_2_gfx_stop(adev);
+		sdma_v5_2_gfx_stop(adev, inst_mask);
 		sdma_v5_2_rlc_stop(adev);
 	}
 
@@ -473,14 +524,17 @@ static void sdma_v5_2_enable(struct amdgpu_device *adev, bool enable)
 }
 
 /**
- * sdma_v5_2_gfx_resume - setup and start the async dma engines
+ * sdma_v5_2_gfx_resume_instance - start/restart a certain sdma engine
  *
  * @adev: amdgpu_device pointer
+ * @i: instance
+ * @restore: used to restore wptr when restart
  *
- * Set up the gfx DMA ring buffers and enable them.
- * Returns 0 for success, error for failure.
+ * Set up the gfx DMA ring buffers and enable them. On restart, we will restore wptr and rptr.
+ * Return 0 for success.
  */
-static int sdma_v5_2_gfx_resume(struct amdgpu_device *adev)
+
+static int sdma_v5_2_gfx_resume_instance(struct amdgpu_device *adev, int i, bool restore)
 {
 	struct amdgpu_ring *ring;
 	u32 rb_cntl, ib_cntl;
@@ -490,148 +544,163 @@ static int sdma_v5_2_gfx_resume(struct amdgpu_device *adev)
 	u32 temp;
 	u32 wptr_poll_cntl;
 	u64 wptr_gpu_addr;
-	int i, r;
 
-	for (i = 0; i < adev->sdma.num_instances; i++) {
-		ring = &adev->sdma.instance[i].ring;
+	ring = &adev->sdma.instance[i].ring;
 
-		if (!amdgpu_sriov_vf(adev))
-			WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_SEM_WAIT_FAIL_TIMER_CNTL), 0);
+	if (!amdgpu_sriov_vf(adev))
+		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_SEM_WAIT_FAIL_TIMER_CNTL), 0);
 
-		/* Set ring buffer size in dwords */
-		rb_bufsz = order_base_2(ring->ring_size / 4);
-		rb_cntl = RREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL));
-		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_SIZE, rb_bufsz);
+	/* Set ring buffer size in dwords */
+	rb_bufsz = order_base_2(ring->ring_size / 4);
+	rb_cntl = RREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL));
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_SIZE, rb_bufsz);
 #ifdef __BIG_ENDIAN
-		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_SWAP_ENABLE, 1);
-		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL,
-					RPTR_WRITEBACK_SWAP_ENABLE, 1);
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_SWAP_ENABLE, 1);
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL,
+				RPTR_WRITEBACK_SWAP_ENABLE, 1);
 #endif
-		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL), rb_cntl);
-
-		/* Initialize the ring buffer's read and write pointers */
+	WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL), rb_cntl);
+
+	/* Initialize the ring buffer's read and write pointers */
+	if (restore) {
+		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR), lower_32_bits(ring->wptr << 2));
+		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_HI), upper_32_bits(ring->wptr << 2));
+		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR), lower_32_bits(ring->wptr << 2));
+		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_HI), upper_32_bits(ring->wptr << 2));
+	} else {
 		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR), 0);
 		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_HI), 0);
 		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR), 0);
 		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_HI), 0);
+	}
 
-		/* setup the wptr shadow polling */
-		wptr_gpu_addr = ring->wptr_gpu_addr;
-		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_LO),
-		       lower_32_bits(wptr_gpu_addr));
-		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_HI),
-		       upper_32_bits(wptr_gpu_addr));
-		wptr_poll_cntl = RREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i,
-							 mmSDMA0_GFX_RB_WPTR_POLL_CNTL));
-		wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl,
-					       SDMA0_GFX_RB_WPTR_POLL_CNTL,
-					       F32_POLL_ENABLE, 1);
-		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_CNTL),
-		       wptr_poll_cntl);
-
-		/* set the wb address whether it's enabled or not */
-		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_ADDR_HI),
-		       upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF);
-		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_ADDR_LO),
-		       lower_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFC);
-
-		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1);
-
-		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_BASE), ring->gpu_addr >> 8);
-		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_BASE_HI), ring->gpu_addr >> 40);
-
+	/* setup the wptr shadow polling */
+	wptr_gpu_addr = ring->wptr_gpu_addr;
+	WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_LO),
+	       lower_32_bits(wptr_gpu_addr));
+	WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_HI),
+	       upper_32_bits(wptr_gpu_addr));
+	wptr_poll_cntl = RREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i,
+						 mmSDMA0_GFX_RB_WPTR_POLL_CNTL));
+	wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl,
+				       SDMA0_GFX_RB_WPTR_POLL_CNTL,
+				       F32_POLL_ENABLE, 1);
+	WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_CNTL),
+	       wptr_poll_cntl);
+
+	/* set the wb address whether it's enabled or not */
+	WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_ADDR_HI),
+	       upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF);
+	WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_ADDR_LO),
+	       lower_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFC);
+
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1);
+
+	WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_BASE), ring->gpu_addr >> 8);
+	WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_BASE_HI), ring->gpu_addr >> 40);
+
+	if (!restore)
 		ring->wptr = 0;
 
-		/* before programing wptr to a less value, need set minor_ptr_update first */
-		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_MINOR_PTR_UPDATE), 1);
+	/* before programing wptr to a less value, need set minor_ptr_update first */
+	WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_MINOR_PTR_UPDATE), 1);
 
-		if (!amdgpu_sriov_vf(adev)) { /* only bare-metal use register write for wptr */
-			WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR), lower_32_bits(ring->wptr << 2));
-			WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_HI), upper_32_bits(ring->wptr << 2));
-		}
+	if (!amdgpu_sriov_vf(adev)) { /* only bare-metal use register write for wptr */
+		WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR), lower_32_bits(ring->wptr << 2));
+		WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_HI), upper_32_bits(ring->wptr << 2));
+	}
 
-		doorbell = RREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL));
-		doorbell_offset = RREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL_OFFSET));
+	doorbell = RREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL));
+	doorbell_offset = RREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL_OFFSET));
 
-		if (ring->use_doorbell) {
-			doorbell = REG_SET_FIELD(doorbell, SDMA0_GFX_DOORBELL, ENABLE, 1);
-			doorbell_offset = REG_SET_FIELD(doorbell_offset, SDMA0_GFX_DOORBELL_OFFSET,
-					OFFSET, ring->doorbell_index);
-		} else {
-			doorbell = REG_SET_FIELD(doorbell, SDMA0_GFX_DOORBELL, ENABLE, 0);
-		}
-		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL), doorbell);
-		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL_OFFSET), doorbell_offset);
+	if (ring->use_doorbell) {
+		doorbell = REG_SET_FIELD(doorbell, SDMA0_GFX_DOORBELL, ENABLE, 1);
+		doorbell_offset = REG_SET_FIELD(doorbell_offset, SDMA0_GFX_DOORBELL_OFFSET,
+				OFFSET, ring->doorbell_index);
+	} else {
+		doorbell = REG_SET_FIELD(doorbell, SDMA0_GFX_DOORBELL, ENABLE, 0);
+	}
+	WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL), doorbell);
+	WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL_OFFSET), doorbell_offset);
 
-		adev->nbio.funcs->sdma_doorbell_range(adev, i, ring->use_doorbell,
-						      ring->doorbell_index,
-						      adev->doorbell_index.sdma_doorbell_range);
+	adev->nbio.funcs->sdma_doorbell_range(adev, i, ring->use_doorbell,
+					      ring->doorbell_index,
+					      adev->doorbell_index.sdma_doorbell_range);
 
-		if (amdgpu_sriov_vf(adev))
-			sdma_v5_2_ring_set_wptr(ring);
+	if (amdgpu_sriov_vf(adev))
+		sdma_v5_2_ring_set_wptr(ring);
 
-		/* set minor_ptr_update to 0 after wptr programed */
+	/* set minor_ptr_update to 0 after wptr programed */
 
-		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_MINOR_PTR_UPDATE), 0);
+	WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_MINOR_PTR_UPDATE), 0);
 
-		/* SRIOV VF has no control of any of registers below */
-		if (!amdgpu_sriov_vf(adev)) {
-			/* set utc l1 enable flag always to 1 */
-			temp = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_CNTL));
-			temp = REG_SET_FIELD(temp, SDMA0_CNTL, UTC_L1_ENABLE, 1);
-
-			/* enable MCBP */
-			temp = REG_SET_FIELD(temp, SDMA0_CNTL, MIDCMD_PREEMPT_ENABLE, 1);
-			WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_CNTL), temp);
-
-			/* Set up RESP_MODE to non-copy addresses */
-			temp = RREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_UTCL1_CNTL));
-			temp = REG_SET_FIELD(temp, SDMA0_UTCL1_CNTL, RESP_MODE, 3);
-			temp = REG_SET_FIELD(temp, SDMA0_UTCL1_CNTL, REDO_DELAY, 9);
-			WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_UTCL1_CNTL), temp);
-
-			/* program default cache read and write policy */
-			temp = RREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_UTCL1_PAGE));
-			/* clean read policy and write policy bits */
-			temp &= 0xFF0FFF;
-			temp |= ((CACHE_READ_POLICY_L2__DEFAULT << 12) |
-				 (CACHE_WRITE_POLICY_L2__DEFAULT << 14) |
-				 SDMA0_UTCL1_PAGE__LLC_NOALLOC_MASK);
-			WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_UTCL1_PAGE), temp);
-
-			/* unhalt engine */
-			temp = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_F32_CNTL));
-			temp = REG_SET_FIELD(temp, SDMA0_F32_CNTL, HALT, 0);
-			WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_F32_CNTL), temp);
-		}
+	/* SRIOV VF has no control of any of registers below */
+	if (!amdgpu_sriov_vf(adev)) {
+		/* set utc l1 enable flag always to 1 */
+		temp = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_CNTL));
+		temp = REG_SET_FIELD(temp, SDMA0_CNTL, UTC_L1_ENABLE, 1);
+
+		/* enable MCBP */
+		temp = REG_SET_FIELD(temp, SDMA0_CNTL, MIDCMD_PREEMPT_ENABLE, 1);
+		WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_CNTL), temp);
+
+		/* Set up RESP_MODE to non-copy addresses */
+		temp = RREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_UTCL1_CNTL));
+		temp = REG_SET_FIELD(temp, SDMA0_UTCL1_CNTL, RESP_MODE, 3);
+		temp = REG_SET_FIELD(temp, SDMA0_UTCL1_CNTL, REDO_DELAY, 9);
+		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_UTCL1_CNTL), temp);
+
+		/* program default cache read and write policy */
+		temp = RREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_UTCL1_PAGE));
+		/* clean read policy and write policy bits */
+		temp &= 0xFF0FFF;
+		temp |= ((CACHE_READ_POLICY_L2__DEFAULT << 12) |
+			 (CACHE_WRITE_POLICY_L2__DEFAULT << 14) |
+			 SDMA0_UTCL1_PAGE__LLC_NOALLOC_MASK);
+		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_UTCL1_PAGE), temp);
+
+		/* unhalt engine */
+		temp = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_F32_CNTL));
+		temp = REG_SET_FIELD(temp, SDMA0_F32_CNTL, HALT, 0);
+		WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_F32_CNTL), temp);
+	}
 
-		/* enable DMA RB */
-		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 1);
-		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL), rb_cntl);
+	/* enable DMA RB */
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 1);
+	WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL), rb_cntl);
 
-		ib_cntl = RREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL));
-		ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_ENABLE, 1);
+	ib_cntl = RREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL));
+	ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_ENABLE, 1);
 #ifdef __BIG_ENDIAN
-		ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_SWAP_ENABLE, 1);
+	ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_SWAP_ENABLE, 1);
 #endif
-		/* enable DMA IBs */
-		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL), ib_cntl);
+	/* enable DMA IBs */
+	WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL), ib_cntl);
 
-		ring->sched.ready = true;
+	if (amdgpu_sriov_vf(adev)) { /* bare-metal sequence doesn't need below to lines */
+		sdma_v5_2_ctx_switch_enable(adev, true);
+		sdma_v5_2_enable(adev, true);
+	}
 
-		if (amdgpu_sriov_vf(adev)) { /* bare-metal sequence doesn't need below to lines */
-			sdma_v5_2_ctx_switch_enable(adev, true);
-			sdma_v5_2_enable(adev, true);
-		}
+	return amdgpu_ring_test_helper(ring);
+}
 
-		r = amdgpu_ring_test_ring(ring);
-		if (r) {
-			ring->sched.ready = false;
-			return r;
-		}
+/**
+ * sdma_v5_2_gfx_resume - setup and start the async dma engines
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set up the gfx DMA ring buffers and enable them.
+ * Returns 0 for success, error for failure.
+ */
+static int sdma_v5_2_gfx_resume(struct amdgpu_device *adev)
+{
+	int i, r;
 
-		if (adev->mman.buffer_funcs_ring == ring)
-			amdgpu_ttm_set_buffer_funcs_status(adev, true);
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		r = sdma_v5_2_gfx_resume_instance(adev, i, false);
+		if (r)
+			return r;
 	}
 
 	return 0;
@@ -694,37 +763,49 @@ static int sdma_v5_2_load_microcode(struct amdgpu_device *adev)
 	return 0;
 }
 
-static int sdma_v5_2_soft_reset(void *handle)
+static int sdma_v5_2_soft_reset_engine(struct amdgpu_device *adev, u32 instance_id)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	u32 grbm_soft_reset;
 	u32 tmp;
-	int i;
 
-	for (i = 0; i < adev->sdma.num_instances; i++) {
-		grbm_soft_reset = REG_SET_FIELD(0,
-						GRBM_SOFT_RESET, SOFT_RESET_SDMA0,
-						1);
-		grbm_soft_reset <<= i;
+	grbm_soft_reset = REG_SET_FIELD(0,
+					GRBM_SOFT_RESET, SOFT_RESET_SDMA0,
+					1);
+	grbm_soft_reset <<= instance_id;
 
-		tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
-		tmp |= grbm_soft_reset;
-		DRM_DEBUG("GRBM_SOFT_RESET=0x%08X\n", tmp);
-		WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
-		tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
+	tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
+	tmp |= grbm_soft_reset;
+	DRM_DEBUG("GRBM_SOFT_RESET=0x%08X\n", tmp);
+	WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
+	tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
 
-		udelay(50);
+	udelay(50);
 
-		tmp &= ~grbm_soft_reset;
-		WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
-		tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
+	tmp &= ~grbm_soft_reset;
+	WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
+	tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
+	return 0;
+}
 
+static int sdma_v5_2_soft_reset(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i;
+
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		sdma_v5_2_soft_reset_engine(adev, i);
 		udelay(50);
 	}
 
 	return 0;
 }
 
+static const struct amdgpu_sdma_funcs sdma_v5_2_sdma_funcs = {
+	.stop_kernel_queue = &sdma_v5_2_stop_queue,
+	.start_kernel_queue = &sdma_v5_2_restore_queue,
+	.soft_reset_kernel_queue = &sdma_v5_2_soft_reset_engine,
+};
+
 /**
  * sdma_v5_2_start - setup and start the async dma engines
  *
@@ -736,6 +817,7 @@ static int sdma_v5_2_soft_reset(void *handle)
 static int sdma_v5_2_start(struct amdgpu_device *adev)
 {
 	int r = 0;
+	struct amdgpu_ip_block *ip_block;
 
 	if (amdgpu_sriov_vf(adev)) {
 		sdma_v5_2_ctx_switch_enable(adev, false);
@@ -756,7 +838,11 @@ static int sdma_v5_2_start(struct amdgpu_device *adev)
 			msleep(1000);
 	}
 
-	sdma_v5_2_soft_reset(adev);
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_SDMA);
+	if (!ip_block)
+		return -EINVAL;
+
+	sdma_v5_2_soft_reset(ip_block);
 	/* unhalt the MEs */
 	sdma_v5_2_enable(adev, true);
 	/* enable sdma ring preemption */
@@ -831,28 +917,18 @@ static int sdma_v5_2_ring_test_ring(struct amdgpu_ring *ring)
 	int r;
 	u32 tmp;
 	u64 gpu_addr;
-	volatile uint32_t *cpu_ptr = NULL;
 
 	tmp = 0xCAFEDEAD;
 
-	if (ring->is_mes_queue) {
-		uint32_t offset = 0;
-		offset = amdgpu_mes_ctx_get_offs(ring,
-					 AMDGPU_MES_CTX_PADDING_OFFS);
-		gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		cpu_ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-		*cpu_ptr = tmp;
-	} else {
-		r = amdgpu_device_wb_get(adev, &index);
-		if (r) {
-			dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r);
-			return r;
-		}
-
-		gpu_addr = adev->wb.gpu_addr + (index * 4);
-		adev->wb.wb[index] = cpu_to_le32(tmp);
+	r = amdgpu_device_wb_get(adev, &index);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r);
+		return r;
 	}
 
+	gpu_addr = adev->wb.gpu_addr + (index * 4);
+	adev->wb.wb[index] = cpu_to_le32(tmp);
+
 	r = amdgpu_ring_alloc(ring, 20);
 	if (r) {
 		DRM_ERROR("amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, r);
@@ -869,10 +945,7 @@ static int sdma_v5_2_ring_test_ring(struct amdgpu_ring *ring)
 	amdgpu_ring_commit(ring);
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		if (ring->is_mes_queue)
-			tmp = le32_to_cpu(*cpu_ptr);
-		else
-			tmp = le32_to_cpu(adev->wb.wb[index]);
+		tmp = le32_to_cpu(adev->wb.wb[index]);
 		if (tmp == 0xDEADBEEF)
 			break;
 		if (amdgpu_emu_mode == 1)
@@ -884,8 +957,7 @@ static int sdma_v5_2_ring_test_ring(struct amdgpu_ring *ring)
 	if (i >= adev->usec_timeout)
 		r = -ETIMEDOUT;
 
-	if (!ring->is_mes_queue)
-		amdgpu_device_wb_free(adev, index);
+	amdgpu_device_wb_free(adev, index);
 
 	return r;
 }
@@ -908,37 +980,23 @@ static int sdma_v5_2_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	long r;
 	u32 tmp = 0;
 	u64 gpu_addr;
-	volatile uint32_t *cpu_ptr = NULL;
 
 	tmp = 0xCAFEDEAD;
 	memset(&ib, 0, sizeof(ib));
 
-	if (ring->is_mes_queue) {
-		uint32_t offset = 0;
-		offset = amdgpu_mes_ctx_get_offs(ring, AMDGPU_MES_CTX_IB_OFFS);
-		ib.gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		ib.ptr = (void *)amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-
-		offset = amdgpu_mes_ctx_get_offs(ring,
-					 AMDGPU_MES_CTX_PADDING_OFFS);
-		gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		cpu_ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-		*cpu_ptr = tmp;
-	} else {
-		r = amdgpu_device_wb_get(adev, &index);
-		if (r) {
-			dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r);
-			return r;
-		}
+	r = amdgpu_device_wb_get(adev, &index);
+	if (r) {
+		dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r);
+		return r;
+	}
 
-		gpu_addr = adev->wb.gpu_addr + (index * 4);
-		adev->wb.wb[index] = cpu_to_le32(tmp);
+	gpu_addr = adev->wb.gpu_addr + (index * 4);
+	adev->wb.wb[index] = cpu_to_le32(tmp);
 
-		r = amdgpu_ib_get(adev, NULL, 256, AMDGPU_IB_POOL_DIRECT, &ib);
-		if (r) {
-			DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
-			goto err0;
-		}
+	r = amdgpu_ib_get(adev, NULL, 256, AMDGPU_IB_POOL_DIRECT, &ib);
+	if (r) {
+		DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
+		goto err0;
 	}
 
 	ib.ptr[0] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) |
@@ -966,10 +1024,7 @@ static int sdma_v5_2_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		goto err1;
 	}
 
-	if (ring->is_mes_queue)
-		tmp = le32_to_cpu(*cpu_ptr);
-	else
-		tmp = le32_to_cpu(adev->wb.wb[index]);
+	tmp = le32_to_cpu(adev->wb.wb[index]);
 
 	if (tmp == 0xDEADBEEF)
 		r = 0;
@@ -977,11 +1032,10 @@ static int sdma_v5_2_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = -EINVAL;
 
 err1:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err0:
-	if (!ring->is_mes_queue)
-		amdgpu_device_wb_free(adev, index);
+	amdgpu_device_wb_free(adev, index);
 	return r;
 }
 
@@ -1137,7 +1191,28 @@ static void sdma_v5_2_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
 static void sdma_v5_2_ring_emit_vm_flush(struct amdgpu_ring *ring,
 					 unsigned vmid, uint64_t pd_addr)
 {
-	amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
+	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
+	uint32_t req = hub->vmhub_funcs->get_invalidate_req(vmid, 0);
+
+	/* Update the PD address for this VMID. */
+	amdgpu_ring_emit_wreg(ring, hub->ctx0_ptb_addr_lo32 +
+			      (hub->ctx_addr_distance * vmid),
+			      lower_32_bits(pd_addr));
+	amdgpu_ring_emit_wreg(ring, hub->ctx0_ptb_addr_hi32 +
+			      (hub->ctx_addr_distance * vmid),
+			      upper_32_bits(pd_addr));
+
+	/* Trigger invalidation. */
+	amdgpu_ring_write(ring,
+			  SDMA_PKT_VM_INVALIDATION_HEADER_OP(SDMA_OP_POLL_REGMEM) |
+			  SDMA_PKT_VM_INVALIDATION_HEADER_SUB_OP(SDMA_SUBOP_VM_INVALIDATION) |
+			  SDMA_PKT_VM_INVALIDATION_HEADER_GFX_ENG_ID(ring->vm_inv_eng) |
+			  SDMA_PKT_VM_INVALIDATION_HEADER_MM_ENG_ID(0x1f));
+	amdgpu_ring_write(ring, req);
+	amdgpu_ring_write(ring, 0xFFFFFFFF);
+	amdgpu_ring_write(ring,
+			  SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_INVALIDATEACK(1 << vmid) |
+			  SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_ADDRESSRANGEHI(0x1F));
 }
 
 static void sdma_v5_2_ring_emit_wreg(struct amdgpu_ring *ring,
@@ -1173,9 +1248,14 @@ static void sdma_v5_2_ring_emit_reg_write_reg_wait(struct amdgpu_ring *ring,
 	amdgpu_ring_emit_reg_wait(ring, reg1, mask, mask);
 }
 
-static int sdma_v5_2_early_init(void *handle)
+static int sdma_v5_2_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = amdgpu_sdma_init_microcode(adev, 0, true);
+	if (r)
+		return r;
 
 	sdma_v5_2_set_ring_funcs(adev);
 	sdma_v5_2_set_buffer_funcs(adev);
@@ -1220,11 +1300,13 @@ static unsigned sdma_v5_2_seq_to_trap_id(int seq_num)
 	return -EINVAL;
 }
 
-static int sdma_v5_2_sw_init(void *handle)
+static int sdma_v5_2_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
 	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_5_2);
+	uint32_t *ptr;
 
 	/* SDMA trap event */
 	for (i = 0; i < adev->sdma.num_instances; i++) {
@@ -1235,13 +1317,9 @@ static int sdma_v5_2_sw_init(void *handle)
 			return r;
 	}
 
-	r = amdgpu_sdma_init_microcode(adev, 0, true);
-	if (r) {
-		DRM_ERROR("Failed to load sdma firmware!\n");
-		return r;
-	}
-
 	for (i = 0; i < adev->sdma.num_instances; i++) {
+		mutex_init(&adev->sdma.instance[i].engine_reset_mutex);
+		adev->sdma.instance[i].funcs = &sdma_v5_2_sdma_funcs;
 		ring = &adev->sdma.instance[i].ring;
 		ring->ring_obj = NULL;
 		ring->use_doorbell = true;
@@ -1253,7 +1331,7 @@ static int sdma_v5_2_sw_init(void *handle)
 		ring->doorbell_index =
 			(adev->doorbell_index.sdma_engine[i] << 1); //get DWORD offset
 
-		ring->vm_hub = AMDGPU_GFXHUB_0;
+		ring->vm_hub = AMDGPU_GFXHUB(0);
 		sprintf(ring->name, "sdma%d", i);
 		r = amdgpu_ring_init(adev, ring, 1024, &adev->sdma.trap_irq,
 				     AMDGPU_SDMA_IRQ_INSTANCE0 + i,
@@ -1262,38 +1340,71 @@ static int sdma_v5_2_sw_init(void *handle)
 			return r;
 	}
 
+	adev->sdma.supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->sdma.instance[0].ring);
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
+	case IP_VERSION(5, 2, 0):
+	case IP_VERSION(5, 2, 2):
+	case IP_VERSION(5, 2, 3):
+	case IP_VERSION(5, 2, 4):
+		if ((adev->sdma.instance[0].fw_version >= 76) &&
+		    !amdgpu_sriov_vf(adev) &&
+		    !adev->debug_disable_gpu_ring_reset)
+			adev->sdma.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+		break;
+	case IP_VERSION(5, 2, 5):
+		if ((adev->sdma.instance[0].fw_version >= 34) &&
+		    !amdgpu_sriov_vf(adev) &&
+		    !adev->debug_disable_gpu_ring_reset)
+			adev->sdma.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+		break;
+	default:
+		break;
+	}
+
+	/* Allocate memory for SDMA IP Dump buffer */
+	ptr = kcalloc(adev->sdma.num_instances * reg_count, sizeof(uint32_t), GFP_KERNEL);
+	if (ptr)
+		adev->sdma.ip_dump = ptr;
+	else
+		DRM_ERROR("Failed to allocated memory for SDMA IP Dump\n");
+
+	r = amdgpu_sdma_sysfs_reset_mask_init(adev);
+	if (r)
+		return r;
+
 	return r;
 }
 
-static int sdma_v5_2_sw_fini(void *handle)
+static int sdma_v5_2_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++)
 		amdgpu_ring_fini(&adev->sdma.instance[i].ring);
 
+	amdgpu_sdma_sysfs_reset_mask_fini(adev);
 	amdgpu_sdma_destroy_inst_ctx(adev, true);
 
+	kfree(adev->sdma.ip_dump);
+
 	return 0;
 }
 
-static int sdma_v5_2_hw_init(void *handle)
+static int sdma_v5_2_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return sdma_v5_2_start(adev);
 }
 
-static int sdma_v5_2_hw_fini(void *handle)
+static int sdma_v5_2_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	if (amdgpu_sriov_vf(adev)) {
-		/* disable the scheduler for SDMA */
-		amdgpu_sdma_unset_buffer_funcs_helper(adev);
+	if (amdgpu_sriov_vf(adev))
 		return 0;
-	}
 
 	sdma_v5_2_ctx_switch_enable(adev, false);
 	sdma_v5_2_enable(adev, false);
@@ -1301,23 +1412,19 @@ static int sdma_v5_2_hw_fini(void *handle)
 	return 0;
 }
 
-static int sdma_v5_2_suspend(void *handle)
+static int sdma_v5_2_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return sdma_v5_2_hw_fini(adev);
+	return sdma_v5_2_hw_fini(ip_block);
 }
 
-static int sdma_v5_2_resume(void *handle)
+static int sdma_v5_2_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return sdma_v5_2_hw_init(adev);
+	return sdma_v5_2_hw_init(ip_block);
 }
 
-static bool sdma_v5_2_is_idle(void *handle)
+static bool sdma_v5_2_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
@@ -1330,11 +1437,11 @@ static bool sdma_v5_2_is_idle(void *handle)
 	return true;
 }
 
-static int sdma_v5_2_wait_for_idle(void *handle)
+static int sdma_v5_2_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
 	u32 sdma0, sdma1, sdma2, sdma3;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		sdma0 = RREG32(sdma_v5_2_get_reg_offset(adev, 0, mmSDMA0_STATUS_REG));
@@ -1349,6 +1456,100 @@ static int sdma_v5_2_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
+static int sdma_v5_2_reset_queue(struct amdgpu_ring *ring,
+				 unsigned int vmid,
+				 struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int r;
+
+	if (ring->me >= adev->sdma.num_instances) {
+		dev_err(adev->dev, "sdma instance not found\n");
+		return -EINVAL;
+	}
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	amdgpu_amdkfd_suspend(adev, true);
+	r = amdgpu_sdma_reset_engine(adev, ring->me, true);
+	amdgpu_amdkfd_resume(adev, true);
+	if (r)
+		return r;
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static int sdma_v5_2_stop_queue(struct amdgpu_ring *ring)
+{
+	u32 f32_cntl, freeze, cntl, stat1_reg;
+	struct amdgpu_device *adev = ring->adev;
+	int i, j, r = 0;
+
+	if (amdgpu_sriov_vf(adev))
+		return -EINVAL;
+
+	i = ring->me;
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
+
+	/* stop queue */
+	sdma_v5_2_gfx_stop(adev, 1 << i);
+
+	/*engine stop SDMA1_F32_CNTL.HALT to 1 and SDMAx_FREEZE freeze bit to 1 */
+	freeze = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_FREEZE));
+	freeze = REG_SET_FIELD(freeze, SDMA0_FREEZE, FREEZE, 1);
+	WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_FREEZE), freeze);
+
+	for (j = 0; j < adev->usec_timeout; j++) {
+		freeze = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_FREEZE));
+
+		if (REG_GET_FIELD(freeze, SDMA0_FREEZE, FROZEN) & 1)
+			break;
+		udelay(1);
+	}
+
+
+	if (j == adev->usec_timeout) {
+		stat1_reg = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_STATUS1_REG));
+		if ((stat1_reg & 0x3FF) != 0x3FF) {
+			DRM_ERROR("cannot soft reset as sdma not idle\n");
+			r = -ETIMEDOUT;
+			goto err0;
+		}
+	}
+
+	f32_cntl = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_F32_CNTL));
+	f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_F32_CNTL, HALT, 1);
+	WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_F32_CNTL), f32_cntl);
+
+	cntl = RREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_CNTL));
+	cntl = REG_SET_FIELD(cntl, SDMA0_CNTL, UTC_L1_ENABLE, 0);
+	WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_CNTL), cntl);
+
+err0:
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
+	return r;
+}
+
+static int sdma_v5_2_restore_queue(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 inst_id = ring->me;
+	u32 freeze;
+	int r;
+
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
+	/* unfreeze and unhalt */
+	freeze = RREG32(sdma_v5_2_get_reg_offset(adev, inst_id, mmSDMA0_FREEZE));
+	freeze = REG_SET_FIELD(freeze, SDMA0_FREEZE, FREEZE, 0);
+	WREG32(sdma_v5_2_get_reg_offset(adev, inst_id, mmSDMA0_FREEZE), freeze);
+
+	r = sdma_v5_2_gfx_resume_instance(adev, inst_id, true);
+
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
+
+	return r;
+}
+
 static int sdma_v5_2_ring_preempt_ib(struct amdgpu_ring *ring)
 {
 	int i, r = 0;
@@ -1511,6 +1712,30 @@ static int sdma_v5_2_process_illegal_inst_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
+static bool sdma_v5_2_firmware_mgcg_support(struct amdgpu_device *adev,
+						     int i)
+{
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
+	case IP_VERSION(5, 2, 1):
+		if (adev->sdma.instance[i].fw_version < 70)
+			return false;
+		break;
+	case IP_VERSION(5, 2, 3):
+		if (adev->sdma.instance[i].fw_version < 47)
+			return false;
+		break;
+	case IP_VERSION(5, 2, 7):
+		if (adev->sdma.instance[i].fw_version < 9)
+			return false;
+		break;
+	default:
+		return true;
+	}
+
+	return true;
+
+}
+
 static void sdma_v5_2_update_medium_grain_clock_gating(struct amdgpu_device *adev,
 						       bool enable)
 {
@@ -1519,7 +1744,7 @@ static void sdma_v5_2_update_medium_grain_clock_gating(struct amdgpu_device *ade
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 
-		if (adev->sdma.instance[i].fw_version < 70 && adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(5, 2, 1))
+		if (!sdma_v5_2_firmware_mgcg_support(adev, i))
 			adev->cg_flags &= ~AMD_CG_SUPPORT_SDMA_MGCG;
 
 		if (enable && (adev->cg_flags & AMD_CG_SUPPORT_SDMA_MGCG)) {
@@ -1555,8 +1780,9 @@ static void sdma_v5_2_update_medium_grain_light_sleep(struct amdgpu_device *adev
 	int i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
-
-		if (adev->sdma.instance[i].fw_version < 70 && adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(5, 2, 1))
+		if (adev->sdma.instance[i].fw_version < 70 &&
+		    amdgpu_ip_version(adev, SDMA0_HWIP, 0) ==
+			    IP_VERSION(5, 2, 1))
 			adev->cg_flags &= ~AMD_CG_SUPPORT_SDMA_LS;
 
 		if (enable && (adev->cg_flags & AMD_CG_SUPPORT_SDMA_LS)) {
@@ -1577,15 +1803,15 @@ static void sdma_v5_2_update_medium_grain_light_sleep(struct amdgpu_device *adev
 	}
 }
 
-static int sdma_v5_2_set_clockgating_state(void *handle,
+static int sdma_v5_2_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[SDMA0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
 	case IP_VERSION(5, 2, 0):
 	case IP_VERSION(5, 2, 2):
 	case IP_VERSION(5, 2, 1):
@@ -1593,6 +1819,7 @@ static int sdma_v5_2_set_clockgating_state(void *handle,
 	case IP_VERSION(5, 2, 5):
 	case IP_VERSION(5, 2, 6):
 	case IP_VERSION(5, 2, 3):
+	case IP_VERSION(5, 2, 7):
 		sdma_v5_2_update_medium_grain_clock_gating(adev,
 				state == AMD_CG_STATE_GATE);
 		sdma_v5_2_update_medium_grain_light_sleep(adev,
@@ -1605,15 +1832,15 @@ static int sdma_v5_2_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int sdma_v5_2_set_powergating_state(void *handle,
+static int sdma_v5_2_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
 }
 
-static void sdma_v5_2_get_clockgating_state(void *handle, u64 *flags)
+static void sdma_v5_2_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	if (amdgpu_sriov_vf(adev))
@@ -1630,10 +1857,81 @@ static void sdma_v5_2_get_clockgating_state(void *handle, u64 *flags)
 		*flags |= AMD_CG_SUPPORT_SDMA_LS;
 }
 
-const struct amd_ip_funcs sdma_v5_2_ip_funcs = {
+static void sdma_v5_2_ring_begin_use(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	/* SDMA 5.2.3 (RMB) FW doesn't seem to properly
+	 * disallow GFXOFF in some cases leading to
+	 * hangs in SDMA.  Disallow GFXOFF while SDMA is active.
+	 * We can probably just limit this to 5.2.3,
+	 * but it shouldn't hurt for other parts since
+	 * this GFXOFF will be disallowed anyway when SDMA is
+	 * active, this just makes it explicit.
+	 * sdma_v5_2_ring_set_wptr() takes advantage of this
+	 * to update the wptr because sometimes SDMA seems to miss
+	 * doorbells when entering PG.  If you remove this, update
+	 * sdma_v5_2_ring_set_wptr() as well!
+	 */
+	amdgpu_gfx_off_ctrl(adev, false);
+}
+
+static void sdma_v5_2_ring_end_use(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	/* SDMA 5.2.3 (RMB) FW doesn't seem to properly
+	 * disallow GFXOFF in some cases leading to
+	 * hangs in SDMA.  Allow GFXOFF when SDMA is complete.
+	 */
+	amdgpu_gfx_off_ctrl(adev, true);
+}
+
+static void sdma_v5_2_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, j;
+	uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_5_2);
+	uint32_t instance_offset;
+
+	if (!adev->sdma.ip_dump)
+		return;
+
+	drm_printf(p, "num_instances:%d\n", adev->sdma.num_instances);
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		instance_offset = i * reg_count;
+		drm_printf(p, "\nInstance:%d\n", i);
+
+		for (j = 0; j < reg_count; j++)
+			drm_printf(p, "%-50s \t 0x%08x\n", sdma_reg_list_5_2[j].reg_name,
+				   adev->sdma.ip_dump[instance_offset + j]);
+	}
+}
+
+static void sdma_v5_2_dump_ip_state(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, j;
+	uint32_t instance_offset;
+	uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_5_2);
+
+	if (!adev->sdma.ip_dump)
+		return;
+
+	amdgpu_gfx_off_ctrl(adev, false);
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		instance_offset = i * reg_count;
+		for (j = 0; j < reg_count; j++)
+			adev->sdma.ip_dump[instance_offset + j] =
+				RREG32(sdma_v5_2_get_reg_offset(adev, i,
+				       sdma_reg_list_5_2[j].reg_offset));
+	}
+	amdgpu_gfx_off_ctrl(adev, true);
+}
+
+static const struct amd_ip_funcs sdma_v5_2_ip_funcs = {
 	.name = "sdma_v5_2",
 	.early_init = sdma_v5_2_early_init,
-	.late_init = NULL,
 	.sw_init = sdma_v5_2_sw_init,
 	.sw_fini = sdma_v5_2_sw_fini,
 	.hw_init = sdma_v5_2_hw_init,
@@ -1646,6 +1944,8 @@ const struct amd_ip_funcs sdma_v5_2_ip_funcs = {
 	.set_clockgating_state = sdma_v5_2_set_clockgating_state,
 	.set_powergating_state = sdma_v5_2_set_powergating_state,
 	.get_clockgating_state = sdma_v5_2_get_clockgating_state,
+	.dump_ip_state = sdma_v5_2_dump_ip_state,
+	.print_ip_state = sdma_v5_2_print_ip_state,
 };
 
 static const struct amdgpu_ring_funcs sdma_v5_2_ring_funcs = {
@@ -1677,12 +1977,14 @@ static const struct amdgpu_ring_funcs sdma_v5_2_ring_funcs = {
 	.test_ib = sdma_v5_2_ring_test_ib,
 	.insert_nop = sdma_v5_2_ring_insert_nop,
 	.pad_ib = sdma_v5_2_ring_pad_ib,
+	.begin_use = sdma_v5_2_ring_begin_use,
+	.end_use = sdma_v5_2_ring_end_use,
 	.emit_wreg = sdma_v5_2_ring_emit_wreg,
 	.emit_reg_wait = sdma_v5_2_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = sdma_v5_2_ring_emit_reg_write_reg_wait,
 	.init_cond_exec = sdma_v5_2_ring_init_cond_exec,
-	.patch_cond_exec = sdma_v5_2_ring_patch_cond_exec,
 	.preempt_ib = sdma_v5_2_ring_preempt_ib,
+	.reset = sdma_v5_2_reset_queue,
 };
 
 static void sdma_v5_2_set_ring_funcs(struct amdgpu_device *adev)
@@ -1719,7 +2021,7 @@ static void sdma_v5_2_set_irq_funcs(struct amdgpu_device *adev)
  * @src_offset: src GPU address
  * @dst_offset: dst GPU address
  * @byte_count: number of bytes to xfer
- * @tmz: if a secure copy should be used
+ * @copy_flags: copy flags for the buffers
  *
  * Copy GPU buffers using the DMA engine.
  * Used by the amdgpu ttm implementation to move pages if
@@ -1729,11 +2031,11 @@ static void sdma_v5_2_emit_copy_buffer(struct amdgpu_ib *ib,
 				       uint64_t src_offset,
 				       uint64_t dst_offset,
 				       uint32_t byte_count,
-				       bool tmz)
+				       uint32_t copy_flags)
 {
 	ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) |
 		SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR) |
-		SDMA_PKT_COPY_LINEAR_HEADER_TMZ(tmz ? 1 : 0);
+		SDMA_PKT_COPY_LINEAR_HEADER_TMZ((copy_flags & AMDGPU_COPY_FLAGS_TMZ) ? 1 : 0);
 	ib->ptr[ib->length_dw++] = byte_count - 1;
 	ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */
 	ib->ptr[ib->length_dw++] = lower_32_bits(src_offset);
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.h b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.h
index b70414fef2a1..863145b3a77e 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.h
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.h
@@ -24,7 +24,6 @@
 #ifndef __SDMA_V5_2_H__
 #define __SDMA_V5_2_H__
 
-extern const struct amd_ip_funcs sdma_v5_2_ip_funcs;
 extern const struct amdgpu_ip_block_version sdma_v5_2_ip_block;
 
 #endif /* __SDMA_V5_2_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
index eb722830531f..217040044987 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
@@ -43,17 +43,80 @@
 #include "sdma_common.h"
 #include "sdma_v6_0.h"
 #include "v11_structs.h"
+#include "mes_userqueue.h"
+#include "amdgpu_userq_fence.h"
 
 MODULE_FIRMWARE("amdgpu/sdma_6_0_0.bin");
 MODULE_FIRMWARE("amdgpu/sdma_6_0_1.bin");
 MODULE_FIRMWARE("amdgpu/sdma_6_0_2.bin");
 MODULE_FIRMWARE("amdgpu/sdma_6_0_3.bin");
+MODULE_FIRMWARE("amdgpu/sdma_6_1_0.bin");
+MODULE_FIRMWARE("amdgpu/sdma_6_1_1.bin");
+MODULE_FIRMWARE("amdgpu/sdma_6_1_2.bin");
+MODULE_FIRMWARE("amdgpu/sdma_6_1_3.bin");
 
 #define SDMA1_REG_OFFSET 0x600
 #define SDMA0_HYP_DEC_REG_START 0x5880
 #define SDMA0_HYP_DEC_REG_END 0x589a
 #define SDMA1_HYP_DEC_REG_OFFSET 0x20
 
+static const struct amdgpu_hwip_reg_entry sdma_reg_list_6_0[] = {
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_STATUS_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_STATUS1_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_STATUS2_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_STATUS3_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_STATUS4_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_STATUS5_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_STATUS6_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_UCODE_CHECKSUM),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_RB_RPTR_FETCH_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_RB_RPTR_FETCH),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_UTCL1_RD_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_UTCL1_WR_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_UTCL1_RD_XNACK0),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_UTCL1_RD_XNACK1),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_UTCL1_WR_XNACK0),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_UTCL1_WR_XNACK1),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_RB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_RB_RPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_RB_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_IB_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_IB_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_IB_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_IB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_IB_SUB_REMAIN),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_DUMMY_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE_STATUS0),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_RB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_RB_RPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_RB_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_IB_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_IB_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_IB_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_IB_SUB_REMAIN),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_DUMMY_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_RB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_RB_RPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_RB_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_IB_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_IB_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_IB_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_IB_SUB_REMAIN),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_DUMMY_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_INT_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS2),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_CHICKEN_BITS),
+};
+
 static void sdma_v6_0_set_ring_funcs(struct amdgpu_device *adev);
 static void sdma_v6_0_set_buffer_funcs(struct amdgpu_device *adev);
 static void sdma_v6_0_set_vm_pte_funcs(struct amdgpu_device *adev);
@@ -78,35 +141,23 @@ static u32 sdma_v6_0_get_reg_offset(struct amdgpu_device *adev, u32 instance, u3
 	return base + internal_offset;
 }
 
-static unsigned sdma_v6_0_ring_init_cond_exec(struct amdgpu_ring *ring)
+static unsigned sdma_v6_0_ring_init_cond_exec(struct amdgpu_ring *ring,
+					      uint64_t addr)
 {
 	unsigned ret;
 
 	amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_COND_EXE));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
 	amdgpu_ring_write(ring, 1);
-	ret = ring->wptr & ring->buf_mask;/* this is the offset we need patch later */
-	amdgpu_ring_write(ring, 0x55aa55aa);/* insert dummy here and patch it later */
+	/* this is the offset we need patch later */
+	ret = ring->wptr & ring->buf_mask;
+	/* insert dummy here and patch it later */
+	amdgpu_ring_write(ring, 0);
 
 	return ret;
 }
 
-static void sdma_v6_0_ring_patch_cond_exec(struct amdgpu_ring *ring,
-					   unsigned offset)
-{
-	unsigned cur;
-
-	BUG_ON(offset > ring->buf_mask);
-	BUG_ON(ring->ring[offset] != 0x55aa55aa);
-
-	cur = (ring->wptr - 1) & ring->buf_mask;
-	if (cur > offset)
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->buf_mask + 1) - offset + cur;
-}
-
 /**
  * sdma_v6_0_ring_get_rptr - get the current read pointer
  *
@@ -155,68 +206,35 @@ static uint64_t sdma_v6_0_ring_get_wptr(struct amdgpu_ring *ring)
 static void sdma_v6_0_ring_set_wptr(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	uint32_t *wptr_saved;
-	uint32_t *is_queue_unmap;
-	uint64_t aggregated_db_index;
-	uint32_t mqd_size = adev->mqds[AMDGPU_HW_IP_DMA].mqd_size;
-
-	DRM_DEBUG("Setting write pointer\n");
-
-	if (ring->is_mes_queue) {
-		wptr_saved = (uint32_t *)(ring->mqd_ptr + mqd_size);
-		is_queue_unmap = (uint32_t *)(ring->mqd_ptr + mqd_size +
-					      sizeof(uint32_t));
-		aggregated_db_index =
-			amdgpu_mes_get_aggregated_doorbell_index(adev,
-							 ring->hw_prio);
 
+	if (ring->use_doorbell) {
+		DRM_DEBUG("Using doorbell -- "
+			  "wptr_offs == 0x%08x "
+			  "lower_32_bits(ring->wptr) << 2 == 0x%08x "
+			  "upper_32_bits(ring->wptr) << 2 == 0x%08x\n",
+			  ring->wptr_offs,
+			  lower_32_bits(ring->wptr << 2),
+			  upper_32_bits(ring->wptr << 2));
+		/* XXX check if swapping is necessary on BE */
 		atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
 			     ring->wptr << 2);
-		*wptr_saved = ring->wptr << 2;
-		if (*is_queue_unmap) {
-			WDOORBELL64(aggregated_db_index, ring->wptr << 2);
-			DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n",
-					ring->doorbell_index, ring->wptr << 2);
-			WDOORBELL64(ring->doorbell_index, ring->wptr << 2);
-		} else {
-			DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n",
-					ring->doorbell_index, ring->wptr << 2);
-			WDOORBELL64(ring->doorbell_index, ring->wptr << 2);
-
-			if (*is_queue_unmap)
-				WDOORBELL64(aggregated_db_index,
-					    ring->wptr << 2);
-		}
+		DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n",
+			  ring->doorbell_index, ring->wptr << 2);
+		WDOORBELL64(ring->doorbell_index, ring->wptr << 2);
 	} else {
-		if (ring->use_doorbell) {
-			DRM_DEBUG("Using doorbell -- "
-				  "wptr_offs == 0x%08x "
-				  "lower_32_bits(ring->wptr) << 2 == 0x%08x "
-				  "upper_32_bits(ring->wptr) << 2 == 0x%08x\n",
-				  ring->wptr_offs,
-				  lower_32_bits(ring->wptr << 2),
-				  upper_32_bits(ring->wptr << 2));
-			/* XXX check if swapping is necessary on BE */
-			atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
-				     ring->wptr << 2);
-			DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n",
-				  ring->doorbell_index, ring->wptr << 2);
-			WDOORBELL64(ring->doorbell_index, ring->wptr << 2);
-		} else {
-			DRM_DEBUG("Not using doorbell -- "
-				  "regSDMA%i_GFX_RB_WPTR == 0x%08x "
-				  "regSDMA%i_GFX_RB_WPTR_HI == 0x%08x\n",
-				  ring->me,
-				  lower_32_bits(ring->wptr << 2),
-				  ring->me,
-				  upper_32_bits(ring->wptr << 2));
-			WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev,
-				        ring->me, regSDMA0_QUEUE0_RB_WPTR),
-					lower_32_bits(ring->wptr << 2));
-			WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev,
-				        ring->me, regSDMA0_QUEUE0_RB_WPTR_HI),
-					upper_32_bits(ring->wptr << 2));
-		}
+		DRM_DEBUG("Not using doorbell -- "
+			  "regSDMA%i_GFX_RB_WPTR == 0x%08x "
+			  "regSDMA%i_GFX_RB_WPTR_HI == 0x%08x\n",
+			  ring->me,
+			  lower_32_bits(ring->wptr << 2),
+			  ring->me,
+			  upper_32_bits(ring->wptr << 2));
+		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev,
+							     ring->me, regSDMA0_QUEUE0_RB_WPTR),
+				lower_32_bits(ring->wptr << 2));
+		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev,
+							     ring->me, regSDMA0_QUEUE0_RB_WPTR_HI),
+				upper_32_bits(ring->wptr << 2));
 	}
 }
 
@@ -233,11 +251,13 @@ static void sdma_v6_0_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
 			amdgpu_ring_write(ring, ring->funcs->nop);
 }
 
-/**
+/*
  * sdma_v6_0_ring_emit_ib - Schedule an IB on the DMA engine
  *
  * @ring: amdgpu ring pointer
  * @ib: IB object to schedule
+ * @flags: unused
+ * @job: job to retrieve vmid from
  *
  * Schedule an IB in the DMA ring.
  */
@@ -358,11 +378,9 @@ static void sdma_v6_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 se
 	}
 
 	if (flags & AMDGPU_FENCE_FLAG_INT) {
-		uint32_t ctx = ring->is_mes_queue ?
-			(ring->hw_queue_id | AMDGPU_FENCE_MES_QUEUE_FLAG) : 0;
 		/* generate an interrupt */
 		amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_TRAP));
-		amdgpu_ring_write(ring, SDMA_PKT_TRAP_INT_CONTEXT_INT_CONTEXT(ctx));
+		amdgpu_ring_write(ring, SDMA_PKT_TRAP_INT_CONTEXT_INT_CONTEXT(0));
 	}
 }
 
@@ -378,8 +396,6 @@ static void sdma_v6_0_gfx_stop(struct amdgpu_device *adev)
 	u32 rb_cntl, ib_cntl;
 	int i;
 
-	amdgpu_sdma_unset_buffer_funcs_helper(adev);
-
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		rb_cntl = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_CNTL));
 		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_ENABLE, 0);
@@ -454,14 +470,16 @@ static void sdma_v6_0_enable(struct amdgpu_device *adev, bool enable)
 }
 
 /**
- * sdma_v6_0_gfx_resume - setup and start the async dma engines
+ * sdma_v6_0_gfx_resume_instance - start/restart a certain sdma engine
  *
  * @adev: amdgpu_device pointer
+ * @i: instance
+ * @restore: used to restore wptr when restart
  *
- * Set up the gfx DMA ring buffers and enable them.
- * Returns 0 for success, error for failure.
+ * Set up the gfx DMA ring buffers and enable them. On restart, we will restore wptr and rptr.
+ * Return 0 for success.
  */
-static int sdma_v6_0_gfx_resume(struct amdgpu_device *adev)
+static int sdma_v6_0_gfx_resume_instance(struct amdgpu_device *adev, int i, bool restore)
 {
 	struct amdgpu_ring *ring;
 	u32 rb_cntl, ib_cntl;
@@ -470,137 +488,154 @@ static int sdma_v6_0_gfx_resume(struct amdgpu_device *adev)
 	u32 doorbell_offset;
 	u32 temp;
 	u64 wptr_gpu_addr;
-	int i, r;
 
-	for (i = 0; i < adev->sdma.num_instances; i++) {
-		ring = &adev->sdma.instance[i].ring;
+	ring = &adev->sdma.instance[i].ring;
+	if (!amdgpu_sriov_vf(adev))
+		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_SEM_WAIT_FAIL_TIMER_CNTL), 0);
 
-		if (!amdgpu_sriov_vf(adev))
-			WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_SEM_WAIT_FAIL_TIMER_CNTL), 0);
-
-		/* Set ring buffer size in dwords */
-		rb_bufsz = order_base_2(ring->ring_size / 4);
-		rb_cntl = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_CNTL));
-		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_SIZE, rb_bufsz);
+	/* Set ring buffer size in dwords */
+	rb_bufsz = order_base_2(ring->ring_size / 4);
+	rb_cntl = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_CNTL));
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_SIZE, rb_bufsz);
 #ifdef __BIG_ENDIAN
-		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_SWAP_ENABLE, 1);
-		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL,
-					RPTR_WRITEBACK_SWAP_ENABLE, 1);
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_SWAP_ENABLE, 1);
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL,
+				RPTR_WRITEBACK_SWAP_ENABLE, 1);
 #endif
-		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_PRIV, 1);
-		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_CNTL), rb_cntl);
-
-		/* Initialize the ring buffer's read and write pointers */
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_PRIV, 1);
+	WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_CNTL), rb_cntl);
+
+	/* Initialize the ring buffer's read and write pointers */
+	if (restore) {
+		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_RPTR), lower_32_bits(ring->wptr << 2));
+		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_RPTR_HI), upper_32_bits(ring->wptr << 2));
+		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR), lower_32_bits(ring->wptr << 2));
+		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR_HI), upper_32_bits(ring->wptr << 2));
+	} else {
 		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_RPTR), 0);
 		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_RPTR_HI), 0);
 		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR), 0);
 		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR_HI), 0);
+	}
+	/* setup the wptr shadow polling */
+	wptr_gpu_addr = ring->wptr_gpu_addr;
+	WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR_POLL_ADDR_LO),
+	       lower_32_bits(wptr_gpu_addr));
+	WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR_POLL_ADDR_HI),
+	       upper_32_bits(wptr_gpu_addr));
+
+	/* set the wb address whether it's enabled or not */
+	WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_RPTR_ADDR_HI),
+	       upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF);
+	WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_RPTR_ADDR_LO),
+	       lower_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFC);
+
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1);
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, WPTR_POLL_ENABLE, 0);
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, F32_WPTR_POLL_ENABLE, 1);
+
+	WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_BASE), ring->gpu_addr >> 8);
+	WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_BASE_HI), ring->gpu_addr >> 40);
+
+	if (!restore)
+		ring->wptr = 0;
 
-		/* setup the wptr shadow polling */
-		wptr_gpu_addr = ring->wptr_gpu_addr;
-		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR_POLL_ADDR_LO),
-		       lower_32_bits(wptr_gpu_addr));
-		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR_POLL_ADDR_HI),
-		       upper_32_bits(wptr_gpu_addr));
-
-		/* set the wb address whether it's enabled or not */
-		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_RPTR_ADDR_HI),
-		       upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF);
-		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_RPTR_ADDR_LO),
-		       lower_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFC);
-
-		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1);
-		if (amdgpu_sriov_vf(adev))
-			rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, WPTR_POLL_ENABLE, 1);
-		else
-			rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, WPTR_POLL_ENABLE, 0);
-		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, F32_WPTR_POLL_ENABLE, 1);
+	/* before programing wptr to a less value, need set minor_ptr_update first */
+	WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_MINOR_PTR_UPDATE), 1);
 
-		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_BASE), ring->gpu_addr >> 8);
-		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_BASE_HI), ring->gpu_addr >> 40);
+	if (!amdgpu_sriov_vf(adev)) { /* only bare-metal use register write for wptr */
+		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR), lower_32_bits(ring->wptr) << 2);
+		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR_HI), upper_32_bits(ring->wptr) << 2);
+	}
 
-		ring->wptr = 0;
+	doorbell = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_DOORBELL));
+	doorbell_offset = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_DOORBELL_OFFSET));
 
-		/* before programing wptr to a less value, need set minor_ptr_update first */
-		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_MINOR_PTR_UPDATE), 1);
+	if (ring->use_doorbell) {
+		doorbell = REG_SET_FIELD(doorbell, SDMA0_QUEUE0_DOORBELL, ENABLE, 1);
+		doorbell_offset = REG_SET_FIELD(doorbell_offset, SDMA0_QUEUE0_DOORBELL_OFFSET,
+				OFFSET, ring->doorbell_index);
+	} else {
+		doorbell = REG_SET_FIELD(doorbell, SDMA0_QUEUE0_DOORBELL, ENABLE, 0);
+	}
+	WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_DOORBELL), doorbell);
+	WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_DOORBELL_OFFSET), doorbell_offset);
 
-		if (!amdgpu_sriov_vf(adev)) { /* only bare-metal use register write for wptr */
-			WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR), lower_32_bits(ring->wptr) << 2);
-			WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR_HI), upper_32_bits(ring->wptr) << 2);
-		}
+	if (i == 0)
+		adev->nbio.funcs->sdma_doorbell_range(adev, i, ring->use_doorbell,
+					      ring->doorbell_index,
+					      adev->doorbell_index.sdma_doorbell_range * adev->sdma.num_instances);
 
-		doorbell = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_DOORBELL));
-		doorbell_offset = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_DOORBELL_OFFSET));
+	if (amdgpu_sriov_vf(adev))
+		sdma_v6_0_ring_set_wptr(ring);
+
+	/* set minor_ptr_update to 0 after wptr programed */
+	WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_MINOR_PTR_UPDATE), 0);
+
+	/* Set up sdma hang watchdog */
+	temp = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_WATCHDOG_CNTL));
+	/* 100ms per unit */
+	temp = REG_SET_FIELD(temp, SDMA0_WATCHDOG_CNTL, QUEUE_HANG_COUNT,
+			     max(adev->usec_timeout/100000, 1));
+	WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_WATCHDOG_CNTL), temp);
+
+	/* Set up RESP_MODE to non-copy addresses */
+	temp = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_UTCL1_CNTL));
+	temp = REG_SET_FIELD(temp, SDMA0_UTCL1_CNTL, RESP_MODE, 3);
+	temp = REG_SET_FIELD(temp, SDMA0_UTCL1_CNTL, REDO_DELAY, 9);
+	WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_UTCL1_CNTL), temp);
+
+	/* program default cache read and write policy */
+	temp = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_UTCL1_PAGE));
+	/* clean read policy and write policy bits */
+	temp &= 0xFF0FFF;
+	temp |= ((CACHE_READ_POLICY_L2__DEFAULT << 12) |
+		 (CACHE_WRITE_POLICY_L2__DEFAULT << 14) |
+		 SDMA0_UTCL1_PAGE__LLC_NOALLOC_MASK);
+	WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_UTCL1_PAGE), temp);
 
-		if (ring->use_doorbell) {
-			doorbell = REG_SET_FIELD(doorbell, SDMA0_QUEUE0_DOORBELL, ENABLE, 1);
-			doorbell_offset = REG_SET_FIELD(doorbell_offset, SDMA0_QUEUE0_DOORBELL_OFFSET,
-					OFFSET, ring->doorbell_index);
-		} else {
-			doorbell = REG_SET_FIELD(doorbell, SDMA0_QUEUE0_DOORBELL, ENABLE, 0);
-		}
-		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_DOORBELL), doorbell);
-		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_DOORBELL_OFFSET), doorbell_offset);
-
-		if (i == 0)
-			adev->nbio.funcs->sdma_doorbell_range(adev, i, ring->use_doorbell,
-						      ring->doorbell_index,
-						      adev->doorbell_index.sdma_doorbell_range * adev->sdma.num_instances);
-
-		if (amdgpu_sriov_vf(adev))
-			sdma_v6_0_ring_set_wptr(ring);
-
-		/* set minor_ptr_update to 0 after wptr programed */
-		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_MINOR_PTR_UPDATE), 0);
-
-		/* Set up RESP_MODE to non-copy addresses */
-		temp = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_UTCL1_CNTL));
-		temp = REG_SET_FIELD(temp, SDMA0_UTCL1_CNTL, RESP_MODE, 3);
-		temp = REG_SET_FIELD(temp, SDMA0_UTCL1_CNTL, REDO_DELAY, 9);
-		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_UTCL1_CNTL), temp);
-
-		/* program default cache read and write policy */
-		temp = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_UTCL1_PAGE));
-		/* clean read policy and write policy bits */
-		temp &= 0xFF0FFF;
-		temp |= ((CACHE_READ_POLICY_L2__DEFAULT << 12) |
-			 (CACHE_WRITE_POLICY_L2__DEFAULT << 14) |
-			 SDMA0_UTCL1_PAGE__LLC_NOALLOC_MASK);
-		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_UTCL1_PAGE), temp);
-
-		if (!amdgpu_sriov_vf(adev)) {
-			/* unhalt engine */
-			temp = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_F32_CNTL));
-			temp = REG_SET_FIELD(temp, SDMA0_F32_CNTL, HALT, 0);
-			temp = REG_SET_FIELD(temp, SDMA0_F32_CNTL, TH1_RESET, 0);
-			WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_F32_CNTL), temp);
-		}
+	if (!amdgpu_sriov_vf(adev)) {
+		/* unhalt engine */
+		temp = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_F32_CNTL));
+		temp = REG_SET_FIELD(temp, SDMA0_F32_CNTL, HALT, 0);
+		temp = REG_SET_FIELD(temp, SDMA0_F32_CNTL, TH1_RESET, 0);
+		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_F32_CNTL), temp);
+	}
 
-		/* enable DMA RB */
-		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_ENABLE, 1);
-		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_CNTL), rb_cntl);
+	/* enable DMA RB */
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_ENABLE, 1);
+	WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_CNTL), rb_cntl);
 
-		ib_cntl = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_IB_CNTL));
-		ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_QUEUE0_IB_CNTL, IB_ENABLE, 1);
+	ib_cntl = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_IB_CNTL));
+	ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_QUEUE0_IB_CNTL, IB_ENABLE, 1);
 #ifdef __BIG_ENDIAN
-		ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_QUEUE0_IB_CNTL, IB_SWAP_ENABLE, 1);
+	ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_QUEUE0_IB_CNTL, IB_SWAP_ENABLE, 1);
 #endif
-		/* enable DMA IBs */
-		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_IB_CNTL), ib_cntl);
+	/* enable DMA IBs */
+	WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_IB_CNTL), ib_cntl);
 
-		ring->sched.ready = true;
+	if (amdgpu_sriov_vf(adev))
+		sdma_v6_0_enable(adev, true);
 
-		if (amdgpu_sriov_vf(adev))
-			sdma_v6_0_enable(adev, true);
+	return amdgpu_ring_test_helper(ring);
+}
 
-		r = amdgpu_ring_test_helper(ring);
-		if (r) {
-			ring->sched.ready = false;
-			return r;
-		}
+/**
+ * sdma_v6_0_gfx_resume - setup and start the async dma engines
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set up the gfx DMA ring buffers and enable them.
+ * Returns 0 for success, error for failure.
+ */
+static int sdma_v6_0_gfx_resume(struct amdgpu_device *adev)
+{
+	int i, r;
 
-		if (adev->mman.buffer_funcs_ring == ring)
-			amdgpu_ttm_set_buffer_funcs_status(adev, true);
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		r = sdma_v6_0_gfx_resume_instance(adev, i, false);
+		if (r)
+			return r;
 	}
 
 	return 0;
@@ -721,9 +756,9 @@ static int sdma_v6_0_load_microcode(struct amdgpu_device *adev)
 	return 0;
 }
 
-static int sdma_v6_0_soft_reset(void *handle)
+static int sdma_v6_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp;
 	int i;
 
@@ -757,9 +792,9 @@ static int sdma_v6_0_soft_reset(void *handle)
 	return sdma_v6_0_start(adev);
 }
 
-static bool sdma_v6_0_check_soft_reset(void *handle)
+static bool sdma_v6_0_check_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring;
 	int i, r;
 	long tmo = msecs_to_jiffies(1000);
@@ -856,6 +891,12 @@ static int sdma_v6_0_mqd_init(struct amdgpu_device *adev, void *mqd,
 	m->sdmax_rlcx_rb_aql_cntl = regSDMA0_QUEUE0_RB_AQL_CNTL_DEFAULT;
 	m->sdmax_rlcx_dummy_reg = regSDMA0_QUEUE0_DUMMY_REG_DEFAULT;
 
+	m->sdmax_rlcx_csa_addr_lo = lower_32_bits(prop->csa_addr);
+	m->sdmax_rlcx_csa_addr_hi = upper_32_bits(prop->csa_addr);
+
+	m->sdmax_rlcx_f32_dbg0 = lower_32_bits(prop->fence_address);
+	m->sdmax_rlcx_f32_dbg1 = upper_32_bits(prop->fence_address);
+
 	return 0;
 }
 
@@ -882,28 +923,18 @@ static int sdma_v6_0_ring_test_ring(struct amdgpu_ring *ring)
 	int r;
 	u32 tmp;
 	u64 gpu_addr;
-	volatile uint32_t *cpu_ptr = NULL;
 
 	tmp = 0xCAFEDEAD;
 
-	if (ring->is_mes_queue) {
-		uint32_t offset = 0;
-		offset = amdgpu_mes_ctx_get_offs(ring,
-					 AMDGPU_MES_CTX_PADDING_OFFS);
-		gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		cpu_ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-		*cpu_ptr = tmp;
-	} else {
-		r = amdgpu_device_wb_get(adev, &index);
-		if (r) {
-			dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r);
-			return r;
-		}
-
-		gpu_addr = adev->wb.gpu_addr + (index * 4);
-		adev->wb.wb[index] = cpu_to_le32(tmp);
+	r = amdgpu_device_wb_get(adev, &index);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r);
+		return r;
 	}
 
+	gpu_addr = adev->wb.gpu_addr + (index * 4);
+	adev->wb.wb[index] = cpu_to_le32(tmp);
+
 	r = amdgpu_ring_alloc(ring, 5);
 	if (r) {
 		DRM_ERROR("amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, r);
@@ -920,10 +951,7 @@ static int sdma_v6_0_ring_test_ring(struct amdgpu_ring *ring)
 	amdgpu_ring_commit(ring);
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		if (ring->is_mes_queue)
-			tmp = le32_to_cpu(*cpu_ptr);
-		else
-			tmp = le32_to_cpu(adev->wb.wb[index]);
+		tmp = le32_to_cpu(adev->wb.wb[index]);
 		if (tmp == 0xDEADBEEF)
 			break;
 		if (amdgpu_emu_mode == 1)
@@ -935,16 +963,16 @@ static int sdma_v6_0_ring_test_ring(struct amdgpu_ring *ring)
 	if (i >= adev->usec_timeout)
 		r = -ETIMEDOUT;
 
-	if (!ring->is_mes_queue)
-		amdgpu_device_wb_free(adev, index);
+	amdgpu_device_wb_free(adev, index);
 
 	return r;
 }
 
-/**
+/*
  * sdma_v6_0_ring_test_ib - test an IB on the DMA engine
  *
  * @ring: amdgpu_ring structure holding ring information
+ * @timeout: timeout value in jiffies, or MAX_SCHEDULE_TIMEOUT
  *
  * Test a simple IB in the DMA ring.
  * Returns 0 on success, error on failure.
@@ -958,37 +986,23 @@ static int sdma_v6_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	long r;
 	u32 tmp = 0;
 	u64 gpu_addr;
-	volatile uint32_t *cpu_ptr = NULL;
 
 	tmp = 0xCAFEDEAD;
 	memset(&ib, 0, sizeof(ib));
 
-	if (ring->is_mes_queue) {
-		uint32_t offset = 0;
-		offset = amdgpu_mes_ctx_get_offs(ring, AMDGPU_MES_CTX_IB_OFFS);
-		ib.gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		ib.ptr = (void *)amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-
-		offset = amdgpu_mes_ctx_get_offs(ring,
-					 AMDGPU_MES_CTX_PADDING_OFFS);
-		gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		cpu_ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-		*cpu_ptr = tmp;
-	} else {
-		r = amdgpu_device_wb_get(adev, &index);
-		if (r) {
-			dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r);
-			return r;
-		}
+	r = amdgpu_device_wb_get(adev, &index);
+	if (r) {
+		dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r);
+		return r;
+	}
 
-		gpu_addr = adev->wb.gpu_addr + (index * 4);
-		adev->wb.wb[index] = cpu_to_le32(tmp);
+	gpu_addr = adev->wb.gpu_addr + (index * 4);
+	adev->wb.wb[index] = cpu_to_le32(tmp);
 
-		r = amdgpu_ib_get(adev, NULL, 256, AMDGPU_IB_POOL_DIRECT, &ib);
-		if (r) {
-			DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
-			goto err0;
-		}
+	r = amdgpu_ib_get(adev, NULL, 256, AMDGPU_IB_POOL_DIRECT, &ib);
+	if (r) {
+		DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
+		goto err0;
 	}
 
 	ib.ptr[0] = SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_WRITE) |
@@ -1016,10 +1030,7 @@ static int sdma_v6_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		goto err1;
 	}
 
-	if (ring->is_mes_queue)
-		tmp = le32_to_cpu(*cpu_ptr);
-	else
-		tmp = le32_to_cpu(adev->wb.wb[index]);
+	tmp = le32_to_cpu(adev->wb.wb[index]);
 
 	if (tmp == 0xDEADBEEF)
 		r = 0;
@@ -1027,11 +1038,10 @@ static int sdma_v6_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = -EINVAL;
 
 err1:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err0:
-	if (!ring->is_mes_queue)
-		amdgpu_device_wb_free(adev, index);
+	amdgpu_device_wb_free(adev, index);
 	return r;
 }
 
@@ -1122,9 +1132,10 @@ static void sdma_v6_0_vm_set_pte_pde(struct amdgpu_ib *ib,
 	ib->ptr[ib->length_dw++] = count - 1; /* number of entries */
 }
 
-/**
+/*
  * sdma_v6_0_ring_pad_ib - pad the IB
  * @ib: indirect buffer to fill with padding
+ * @ring: amdgpu ring pointer
  *
  * Pad the IB with NOPs to a boundary multiple of 8.
  */
@@ -1170,10 +1181,12 @@ static void sdma_v6_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
 			  SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(4)); /* retry count, poll interval */
 }
 
-/**
+/*
  * sdma_v6_0_ring_emit_vm_flush - vm flush using sDMA
  *
  * @ring: amdgpu_ring pointer
+ * @vmid: vmid number to use
+ * @pd_addr: address
  *
  * Update the page table base and flush the VM TLB
  * using sDMA.
@@ -1246,19 +1259,40 @@ static struct amdgpu_sdma_ras sdma_v6_0_3_ras = {
 
 static void sdma_v6_0_set_ras_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[SDMA0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
 	case IP_VERSION(6, 0, 3):
 		adev->sdma.ras = &sdma_v6_0_3_ras;
 		break;
 	default:
 		break;
 	}
-
 }
 
-static int sdma_v6_0_early_init(void *handle)
+static int sdma_v6_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	switch (amdgpu_user_queue) {
+	case -1:
+	case 0:
+	default:
+		adev->sdma.no_user_submission = false;
+		adev->sdma.disable_uq = true;
+		break;
+	case 1:
+		adev->sdma.no_user_submission = false;
+		adev->sdma.disable_uq = false;
+		break;
+	case 2:
+		adev->sdma.no_user_submission = true;
+		adev->sdma.disable_uq = false;
+		break;
+	}
+
+	r = amdgpu_sdma_init_microcode(adev, 0, true);
+	if (r)
+		return r;
 
 	sdma_v6_0_set_ring_funcs(adev);
 	sdma_v6_0_set_buffer_funcs(adev);
@@ -1270,11 +1304,13 @@ static int sdma_v6_0_early_init(void *handle)
 	return 0;
 }
 
-static int sdma_v6_0_sw_init(void *handle)
+static int sdma_v6_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
 	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_6_0);
+	uint32_t *ptr;
 
 	/* SDMA trap event */
 	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GFX,
@@ -1283,17 +1319,19 @@ static int sdma_v6_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	r = amdgpu_sdma_init_microcode(adev, 0, true);
-	if (r) {
-		DRM_ERROR("Failed to load sdma firmware!\n");
+	/* SDMA user fence event */
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GFX,
+			      GFX_11_0_0__SRCID__SDMA_FENCE,
+			      &adev->sdma.fence_irq);
+	if (r)
 		return r;
-	}
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		ring = &adev->sdma.instance[i].ring;
 		ring->ring_obj = NULL;
 		ring->use_doorbell = true;
 		ring->me = i;
+		ring->no_user_submission = adev->sdma.no_user_submission;
 
 		DRM_DEBUG("SDMA %d use_doorbell being set to: [%s]\n", i,
 				ring->use_doorbell?"true":"false");
@@ -1301,7 +1339,7 @@ static int sdma_v6_0_sw_init(void *handle)
 		ring->doorbell_index =
 			(adev->doorbell_index.sdma_engine[i] << 1); // get DWORD offset
 
-		ring->vm_hub = AMDGPU_GFXHUB_0;
+		ring->vm_hub = AMDGPU_GFXHUB(0);
 		sprintf(ring->name, "sdma%d", i);
 		r = amdgpu_ring_init(adev, ring, 1024,
 				     &adev->sdma.trap_irq,
@@ -1311,67 +1349,155 @@ static int sdma_v6_0_sw_init(void *handle)
 			return r;
 	}
 
+	adev->sdma.supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->sdma.instance[0].ring);
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
+	case IP_VERSION(6, 0, 0):
+	case IP_VERSION(6, 0, 2):
+	case IP_VERSION(6, 0, 3):
+		if ((adev->sdma.instance[0].fw_version >= 21) &&
+		    !amdgpu_sriov_vf(adev) &&
+		    !adev->debug_disable_gpu_ring_reset)
+			adev->sdma.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+		break;
+	default:
+		break;
+	}
+
 	if (amdgpu_sdma_ras_sw_init(adev)) {
 		dev_err(adev->dev, "Failed to initialize sdma ras block!\n");
 		return -EINVAL;
 	}
 
+	/* Allocate memory for SDMA IP Dump buffer */
+	ptr = kcalloc(adev->sdma.num_instances * reg_count, sizeof(uint32_t), GFP_KERNEL);
+	if (ptr)
+		adev->sdma.ip_dump = ptr;
+	else
+		DRM_ERROR("Failed to allocated memory for SDMA IP Dump\n");
+
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
+	case IP_VERSION(6, 0, 0):
+		if ((adev->sdma.instance[0].fw_version >= 27) && !adev->sdma.disable_uq)
+			adev->userq_funcs[AMDGPU_HW_IP_DMA] = &userq_mes_funcs;
+		break;
+	case IP_VERSION(6, 0, 1):
+		if ((adev->sdma.instance[0].fw_version >= 18) && !adev->sdma.disable_uq)
+			adev->userq_funcs[AMDGPU_HW_IP_DMA] = &userq_mes_funcs;
+		break;
+	case IP_VERSION(6, 0, 2):
+		if ((adev->sdma.instance[0].fw_version >= 23) && !adev->sdma.disable_uq)
+			adev->userq_funcs[AMDGPU_HW_IP_DMA] = &userq_mes_funcs;
+		break;
+	case IP_VERSION(6, 0, 3):
+		if (adev->sdma.instance[0].fw_version >= 29 && !adev->sdma.disable_uq)
+			adev->userq_funcs[AMDGPU_HW_IP_DMA] = &userq_mes_funcs;
+		break;
+	case IP_VERSION(6, 1, 0):
+		if ((adev->sdma.instance[0].fw_version >= 14) && !adev->sdma.disable_uq)
+			adev->userq_funcs[AMDGPU_HW_IP_DMA] = &userq_mes_funcs;
+		break;
+	case IP_VERSION(6, 1, 1):
+		if ((adev->sdma.instance[0].fw_version >= 17) && !adev->sdma.disable_uq)
+			adev->userq_funcs[AMDGPU_HW_IP_DMA] = &userq_mes_funcs;
+		break;
+	case IP_VERSION(6, 1, 2):
+		if ((adev->sdma.instance[0].fw_version >= 15) && !adev->sdma.disable_uq)
+			adev->userq_funcs[AMDGPU_HW_IP_DMA] = &userq_mes_funcs;
+		break;
+	case IP_VERSION(6, 1, 3):
+		if ((adev->sdma.instance[0].fw_version >= 10) && !adev->sdma.disable_uq)
+			adev->userq_funcs[AMDGPU_HW_IP_DMA] = &userq_mes_funcs;
+		break;
+	default:
+		break;
+	}
+
+	r = amdgpu_sdma_sysfs_reset_mask_init(adev);
+	if (r)
+		return r;
+
 	return r;
 }
 
-static int sdma_v6_0_sw_fini(void *handle)
+static int sdma_v6_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++)
 		amdgpu_ring_fini(&adev->sdma.instance[i].ring);
 
+	amdgpu_sdma_sysfs_reset_mask_fini(adev);
 	amdgpu_sdma_destroy_inst_ctx(adev, true);
 
+	kfree(adev->sdma.ip_dump);
+
 	return 0;
 }
 
-static int sdma_v6_0_hw_init(void *handle)
+static int sdma_v6_0_set_userq_trap_interrupts(struct amdgpu_device *adev,
+					       bool enable)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	unsigned int irq_type;
+	int i, r;
 
-	return sdma_v6_0_start(adev);
+	if (adev->userq_funcs[AMDGPU_HW_IP_DMA]) {
+		for (i = 0; i < adev->sdma.num_instances; i++) {
+			irq_type = AMDGPU_SDMA_IRQ_INSTANCE0 + i;
+			if (enable)
+				r = amdgpu_irq_get(adev, &adev->sdma.trap_irq,
+						   irq_type);
+			else
+				r = amdgpu_irq_put(adev, &adev->sdma.trap_irq,
+						   irq_type);
+			if (r)
+				return r;
+		}
+	}
+
+	return 0;
 }
 
-static int sdma_v6_0_hw_fini(void *handle)
+static int sdma_v6_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
 
-	if (amdgpu_sriov_vf(adev)) {
-		/* disable the scheduler for SDMA */
-		amdgpu_sdma_unset_buffer_funcs_helper(adev);
+	r = sdma_v6_0_start(adev);
+	if (r)
+		return r;
+
+	return sdma_v6_0_set_userq_trap_interrupts(adev, true);
+}
+
+static int sdma_v6_0_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (amdgpu_sriov_vf(adev))
 		return 0;
-	}
 
 	sdma_v6_0_ctxempty_int_enable(adev, false);
 	sdma_v6_0_enable(adev, false);
+	sdma_v6_0_set_userq_trap_interrupts(adev, false);
 
 	return 0;
 }
 
-static int sdma_v6_0_suspend(void *handle)
+static int sdma_v6_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return sdma_v6_0_hw_fini(adev);
+	return sdma_v6_0_hw_fini(ip_block);
 }
 
-static int sdma_v6_0_resume(void *handle)
+static int sdma_v6_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return sdma_v6_0_hw_init(adev);
+	return sdma_v6_0_hw_init(ip_block);
 }
 
-static bool sdma_v6_0_is_idle(void *handle)
+static bool sdma_v6_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
@@ -1384,11 +1510,11 @@ static bool sdma_v6_0_is_idle(void *handle)
 	return true;
 }
 
-static int sdma_v6_0_wait_for_idle(void *handle)
+static int sdma_v6_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
 	u32 sdma0, sdma1;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		sdma0 = RREG32(sdma_v6_0_get_reg_offset(adev, 0, regSDMA0_STATUS_REG));
@@ -1446,6 +1572,31 @@ static int sdma_v6_0_ring_preempt_ib(struct amdgpu_ring *ring)
 	return r;
 }
 
+static int sdma_v6_0_reset_queue(struct amdgpu_ring *ring,
+				 unsigned int vmid,
+				 struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int r;
+
+	if (ring->me >= adev->sdma.num_instances) {
+		dev_err(adev->dev, "sdma instance not found\n");
+		return -EINVAL;
+	}
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	r = amdgpu_mes_reset_legacy_queue(adev, ring, vmid, true);
+	if (r)
+		return r;
+
+	r = sdma_v6_0_gfx_resume_instance(adev, ring->me, true);
+	if (r)
+		return r;
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
 static int sdma_v6_0_set_trap_irq_state(struct amdgpu_device *adev,
 					struct amdgpu_irq_src *source,
 					unsigned type,
@@ -1470,25 +1621,9 @@ static int sdma_v6_0_process_trap_irq(struct amdgpu_device *adev,
 				      struct amdgpu_iv_entry *entry)
 {
 	int instances, queue;
-	uint32_t mes_queue_id = entry->src_data[0];
 
 	DRM_DEBUG("IH: SDMA trap\n");
 
-	if (adev->enable_mes && (mes_queue_id & AMDGPU_FENCE_MES_QUEUE_FLAG)) {
-		struct amdgpu_mes_queue *queue;
-
-		mes_queue_id &= AMDGPU_FENCE_MES_QUEUE_ID_MASK;
-
-		spin_lock(&adev->mes.queue_id_lock);
-		queue = idr_find(&adev->mes.queue_id_idr, mes_queue_id);
-		if (queue) {
-			DRM_DEBUG("process smda queue id = %d\n", mes_queue_id);
-			amdgpu_fence_process(queue->ring);
-		}
-		spin_unlock(&adev->mes.queue_id_lock);
-		return 0;
-	}
-
 	queue = entry->ring_id & 0xf;
 	instances = (entry->ring_id & 0xf0) >> 4;
 	if (instances > 1) {
@@ -1510,6 +1645,29 @@ static int sdma_v6_0_process_trap_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int sdma_v6_0_process_fence_irq(struct amdgpu_device *adev,
+				       struct amdgpu_irq_src *source,
+				       struct amdgpu_iv_entry *entry)
+{
+	u32 doorbell_offset = entry->src_data[0];
+
+	if (adev->enable_mes && doorbell_offset) {
+		struct amdgpu_userq_fence_driver *fence_drv = NULL;
+		struct xarray *xa = &adev->userq_xa;
+		unsigned long flags;
+
+		doorbell_offset >>= SDMA0_QUEUE0_DOORBELL_OFFSET__OFFSET__SHIFT;
+
+		xa_lock_irqsave(xa, flags);
+		fence_drv = xa_load(xa, doorbell_offset);
+		if (fence_drv)
+			amdgpu_userq_fence_driver_process(fence_drv);
+		xa_unlock_irqrestore(xa, flags);
+	}
+
+	return 0;
+}
+
 static int sdma_v6_0_process_illegal_inst_irq(struct amdgpu_device *adev,
 					      struct amdgpu_irq_src *source,
 					      struct amdgpu_iv_entry *entry)
@@ -1517,26 +1675,67 @@ static int sdma_v6_0_process_illegal_inst_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int sdma_v6_0_set_clockgating_state(void *handle,
+static int sdma_v6_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int sdma_v6_0_set_powergating_state(void *handle,
+static int sdma_v6_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
 }
 
-static void sdma_v6_0_get_clockgating_state(void *handle, u64 *flags)
+static void sdma_v6_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
+{
+}
+
+static void sdma_v6_0_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, j;
+	uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_6_0);
+	uint32_t instance_offset;
+
+	if (!adev->sdma.ip_dump)
+		return;
+
+	drm_printf(p, "num_instances:%d\n", adev->sdma.num_instances);
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		instance_offset = i * reg_count;
+		drm_printf(p, "\nInstance:%d\n", i);
+
+		for (j = 0; j < reg_count; j++)
+			drm_printf(p, "%-50s \t 0x%08x\n", sdma_reg_list_6_0[j].reg_name,
+				   adev->sdma.ip_dump[instance_offset + j]);
+	}
+}
+
+static void sdma_v6_0_dump_ip_state(struct amdgpu_ip_block *ip_block)
 {
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, j;
+	uint32_t instance_offset;
+	uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_6_0);
+
+	if (!adev->sdma.ip_dump)
+		return;
+
+	amdgpu_gfx_off_ctrl(adev, false);
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		instance_offset = i * reg_count;
+		for (j = 0; j < reg_count; j++)
+			adev->sdma.ip_dump[instance_offset + j] =
+				RREG32(sdma_v6_0_get_reg_offset(adev, i,
+				       sdma_reg_list_6_0[j].reg_offset));
+	}
+	amdgpu_gfx_off_ctrl(adev, true);
 }
 
 const struct amd_ip_funcs sdma_v6_0_ip_funcs = {
 	.name = "sdma_v6_0",
 	.early_init = sdma_v6_0_early_init,
-	.late_init = NULL,
 	.sw_init = sdma_v6_0_sw_init,
 	.sw_fini = sdma_v6_0_sw_fini,
 	.hw_init = sdma_v6_0_hw_init,
@@ -1550,6 +1749,8 @@ const struct amd_ip_funcs sdma_v6_0_ip_funcs = {
 	.set_clockgating_state = sdma_v6_0_set_clockgating_state,
 	.set_powergating_state = sdma_v6_0_set_powergating_state,
 	.get_clockgating_state = sdma_v6_0_get_clockgating_state,
+	.dump_ip_state = sdma_v6_0_dump_ip_state,
+	.print_ip_state = sdma_v6_0_print_ip_state,
 };
 
 static const struct amdgpu_ring_funcs sdma_v6_0_ring_funcs = {
@@ -1584,8 +1785,8 @@ static const struct amdgpu_ring_funcs sdma_v6_0_ring_funcs = {
 	.emit_reg_wait = sdma_v6_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = sdma_v6_0_ring_emit_reg_write_reg_wait,
 	.init_cond_exec = sdma_v6_0_ring_init_cond_exec,
-	.patch_cond_exec = sdma_v6_0_ring_patch_cond_exec,
 	.preempt_ib = sdma_v6_0_ring_preempt_ib,
+	.reset = sdma_v6_0_reset_queue,
 };
 
 static void sdma_v6_0_set_ring_funcs(struct amdgpu_device *adev)
@@ -1603,6 +1804,10 @@ static const struct amdgpu_irq_src_funcs sdma_v6_0_trap_irq_funcs = {
 	.process = sdma_v6_0_process_trap_irq,
 };
 
+static const struct amdgpu_irq_src_funcs sdma_v6_0_fence_irq_funcs = {
+	.process = sdma_v6_0_process_fence_irq,
+};
+
 static const struct amdgpu_irq_src_funcs sdma_v6_0_illegal_inst_irq_funcs = {
 	.process = sdma_v6_0_process_illegal_inst_irq,
 };
@@ -1612,6 +1817,7 @@ static void sdma_v6_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->sdma.trap_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE0 +
 					adev->sdma.num_instances;
 	adev->sdma.trap_irq.funcs = &sdma_v6_0_trap_irq_funcs;
+	adev->sdma.fence_irq.funcs = &sdma_v6_0_fence_irq_funcs;
 	adev->sdma.illegal_inst_irq.funcs = &sdma_v6_0_illegal_inst_irq_funcs;
 }
 
@@ -1622,7 +1828,7 @@ static void sdma_v6_0_set_irq_funcs(struct amdgpu_device *adev)
  * @src_offset: src GPU address
  * @dst_offset: dst GPU address
  * @byte_count: number of bytes to xfer
- * @tmz: if a secure copy should be used
+ * @copy_flags: copy flags for the buffers
  *
  * Copy GPU buffers using the DMA engine.
  * Used by the amdgpu ttm implementation to move pages if
@@ -1632,11 +1838,11 @@ static void sdma_v6_0_emit_copy_buffer(struct amdgpu_ib *ib,
 				       uint64_t src_offset,
 				       uint64_t dst_offset,
 				       uint32_t byte_count,
-				       bool tmz)
+				       uint32_t copy_flags)
 {
 	ib->ptr[ib->length_dw++] = SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_COPY) |
 		SDMA_PKT_COPY_LINEAR_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR) |
-		SDMA_PKT_COPY_LINEAR_HEADER_TMZ(tmz ? 1 : 0);
+		SDMA_PKT_COPY_LINEAR_HEADER_TMZ((copy_flags & AMDGPU_COPY_FLAGS_TMZ) ? 1 : 0);
 	ib->ptr[ib->length_dw++] = byte_count - 1;
 	ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */
 	ib->ptr[ib->length_dw++] = lower_32_bits(src_offset);
@@ -1660,7 +1866,7 @@ static void sdma_v6_0_emit_fill_buffer(struct amdgpu_ib *ib,
 				       uint64_t dst_offset,
 				       uint32_t byte_count)
 {
-	ib->ptr[ib->length_dw++] = SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_CONST_FILL);
+	ib->ptr[ib->length_dw++] = SDMA_PKT_CONSTANT_FILL_HEADER_OP(SDMA_OP_CONST_FILL);
 	ib->ptr[ib->length_dw++] = lower_32_bits(dst_offset);
 	ib->ptr[ib->length_dw++] = upper_32_bits(dst_offset);
 	ib->ptr[ib->length_dw++] = src_data;
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0_0_pkt_open.h b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0_0_pkt_open.h
index 6af23e7888ca..d8cf830916b9 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0_0_pkt_open.h
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0_0_pkt_open.h
@@ -91,6 +91,14 @@
 #define SDMA_GCR_GLM_WB			(1 << 4)
 #define SDMA_GCR_GL1_RANGE(x)		(((x) & 0x3) << 2)
 #define SDMA_GCR_GLI_INV(x)		(((x) & 0x3) << 0)
+
+#define SDMA_DCC_DATA_FORMAT(x) ((x) & 0x3f)
+#define SDMA_DCC_NUM_TYPE(x) (((x) & 0x7) << 9)
+#define SDMA_DCC_READ_CM(x) (((x) & 0x3) << 16)
+#define SDMA_DCC_WRITE_CM(x) (((x) & 0x3) << 18)
+#define SDMA_DCC_MAX_COM(x) (((x) & 0x3) << 24)
+#define SDMA_DCC_MAX_UCOM(x) (((x) & 0x1) << 26)
+
 /*
 ** Definitions for SDMA_PKT_COPY_LINEAR packet
 */
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v7_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v7_0.c
new file mode 100644
index 000000000000..2b81344dcd66
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v7_0.c
@@ -0,0 +1,1859 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/delay.h>
+#include <linux/firmware.h>
+#include <linux/module.h>
+#include <linux/pci.h>
+
+#include "amdgpu.h"
+#include "amdgpu_ucode.h"
+#include "amdgpu_trace.h"
+
+#include "gc/gc_12_0_0_offset.h"
+#include "gc/gc_12_0_0_sh_mask.h"
+#include "hdp/hdp_6_0_0_offset.h"
+#include "ivsrcid/gfx/irqsrcs_gfx_12_0_0.h"
+
+#include "soc15_common.h"
+#include "soc15.h"
+#include "sdma_v6_0_0_pkt_open.h"
+#include "nbio_v4_3.h"
+#include "sdma_common.h"
+#include "sdma_v7_0.h"
+#include "v12_structs.h"
+#include "mes_userqueue.h"
+#include "amdgpu_userq_fence.h"
+
+MODULE_FIRMWARE("amdgpu/sdma_7_0_0.bin");
+MODULE_FIRMWARE("amdgpu/sdma_7_0_1.bin");
+
+#define SDMA1_REG_OFFSET 0x600
+#define SDMA0_HYP_DEC_REG_START 0x5880
+#define SDMA0_HYP_DEC_REG_END 0x589a
+#define SDMA1_HYP_DEC_REG_OFFSET 0x20
+
+/*define for compression field for sdma7*/
+#define SDMA_PKT_CONSTANT_FILL_HEADER_compress_offset 0
+#define SDMA_PKT_CONSTANT_FILL_HEADER_compress_mask   0x00000001
+#define SDMA_PKT_CONSTANT_FILL_HEADER_compress_shift  16
+#define SDMA_PKT_CONSTANT_FILL_HEADER_COMPRESS(x) (((x) & SDMA_PKT_CONSTANT_FILL_HEADER_compress_mask) << SDMA_PKT_CONSTANT_FILL_HEADER_compress_shift)
+
+static const struct amdgpu_hwip_reg_entry sdma_reg_list_7_0[] = {
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_STATUS_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_STATUS1_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_STATUS2_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_STATUS3_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_STATUS4_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_STATUS5_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_STATUS6_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_UCODE_REV),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_RB_RPTR_FETCH_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_RB_RPTR_FETCH),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_UTCL1_RD_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_UTCL1_WR_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_UTCL1_RD_XNACK0),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_UTCL1_RD_XNACK1),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_UTCL1_WR_XNACK0),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_UTCL1_WR_XNACK1),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_RB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_RB_RPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_RB_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_IB_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_IB_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_IB_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_IB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_IB_SUB_REMAIN),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE0_DUMMY_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE_STATUS0),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_RB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_RB_RPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_RB_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_IB_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_IB_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_IB_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_IB_SUB_REMAIN),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE1_DUMMY_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_RB_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_RB_RPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_RB_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_IB_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_IB_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_IB_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_IB_SUB_REMAIN),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_QUEUE2_DUMMY_REG),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_INT_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_VM_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS2),
+	SOC15_REG_ENTRY_STR(GC, 0, regSDMA0_CHICKEN_BITS),
+};
+
+static void sdma_v7_0_set_ring_funcs(struct amdgpu_device *adev);
+static void sdma_v7_0_set_buffer_funcs(struct amdgpu_device *adev);
+static void sdma_v7_0_set_vm_pte_funcs(struct amdgpu_device *adev);
+static void sdma_v7_0_set_irq_funcs(struct amdgpu_device *adev);
+static int sdma_v7_0_start(struct amdgpu_device *adev);
+
+static u32 sdma_v7_0_get_reg_offset(struct amdgpu_device *adev, u32 instance, u32 internal_offset)
+{
+	u32 base;
+
+	if (internal_offset >= SDMA0_HYP_DEC_REG_START &&
+	    internal_offset <= SDMA0_HYP_DEC_REG_END) {
+		base = adev->reg_offset[GC_HWIP][0][1];
+		if (instance != 0)
+			internal_offset += SDMA1_HYP_DEC_REG_OFFSET * instance;
+	} else {
+		base = adev->reg_offset[GC_HWIP][0][0];
+		if (instance == 1)
+			internal_offset += SDMA1_REG_OFFSET;
+	}
+
+	return base + internal_offset;
+}
+
+static unsigned sdma_v7_0_ring_init_cond_exec(struct amdgpu_ring *ring,
+					      uint64_t addr)
+{
+	unsigned ret;
+
+	amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_COND_EXE));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	amdgpu_ring_write(ring, 1);
+	/* this is the offset we need patch later */
+	ret = ring->wptr & ring->buf_mask;
+	/* insert dummy here and patch it later */
+	amdgpu_ring_write(ring, 0);
+
+	return ret;
+}
+
+/**
+ * sdma_v7_0_ring_get_rptr - get the current read pointer
+ *
+ * @ring: amdgpu ring pointer
+ *
+ * Get the current rptr from the hardware.
+ */
+static uint64_t sdma_v7_0_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	u64 *rptr;
+
+	/* XXX check if swapping is necessary on BE */
+	rptr = (u64 *)ring->rptr_cpu_addr;
+
+	DRM_DEBUG("rptr before shift == 0x%016llx\n", *rptr);
+	return ((*rptr) >> 2);
+}
+
+/**
+ * sdma_v7_0_ring_get_wptr - get the current write pointer
+ *
+ * @ring: amdgpu ring pointer
+ *
+ * Get the current wptr from the hardware.
+ */
+static uint64_t sdma_v7_0_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	u64 wptr = 0;
+
+	if (ring->use_doorbell) {
+		/* XXX check if swapping is necessary on BE */
+		wptr = READ_ONCE(*((u64 *)ring->wptr_cpu_addr));
+		DRM_DEBUG("wptr/doorbell before shift == 0x%016llx\n", wptr);
+	}
+
+	return wptr >> 2;
+}
+
+/**
+ * sdma_v7_0_ring_set_wptr - commit the write pointer
+ *
+ * @ring: amdgpu ring pointer
+ *
+ * Write the wptr back to the hardware.
+ */
+static void sdma_v7_0_ring_set_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	DRM_DEBUG("Setting write pointer\n");
+
+	if (ring->use_doorbell) {
+		DRM_DEBUG("Using doorbell -- "
+			  "wptr_offs == 0x%08x "
+			  "lower_32_bits(ring->wptr) << 2 == 0x%08x "
+			  "upper_32_bits(ring->wptr) << 2 == 0x%08x\n",
+			  ring->wptr_offs,
+			  lower_32_bits(ring->wptr << 2),
+			  upper_32_bits(ring->wptr << 2));
+		/* XXX check if swapping is necessary on BE */
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
+			     ring->wptr << 2);
+		DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n",
+			  ring->doorbell_index, ring->wptr << 2);
+		WDOORBELL64(ring->doorbell_index, ring->wptr << 2);
+	} else {
+		DRM_DEBUG("Not using doorbell -- "
+			  "regSDMA%i_GFX_RB_WPTR == 0x%08x "
+			  "regSDMA%i_GFX_RB_WPTR_HI == 0x%08x\n",
+			  ring->me,
+			  lower_32_bits(ring->wptr << 2),
+			  ring->me,
+			  upper_32_bits(ring->wptr << 2));
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev,
+							     ring->me,
+							     regSDMA0_QUEUE0_RB_WPTR),
+				lower_32_bits(ring->wptr << 2));
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev,
+							     ring->me,
+							     regSDMA0_QUEUE0_RB_WPTR_HI),
+				upper_32_bits(ring->wptr << 2));
+	}
+}
+
+static void sdma_v7_0_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
+{
+	struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring);
+	int i;
+
+	for (i = 0; i < count; i++)
+		if (sdma && sdma->burst_nop && (i == 0))
+			amdgpu_ring_write(ring, ring->funcs->nop |
+				SDMA_PKT_NOP_HEADER_COUNT(count - 1));
+		else
+			amdgpu_ring_write(ring, ring->funcs->nop);
+}
+
+/**
+ * sdma_v7_0_ring_emit_ib - Schedule an IB on the DMA engine
+ *
+ * @ring: amdgpu ring pointer
+ * @job: job to retrieve vmid from
+ * @ib: IB object to schedule
+ * @flags: unused
+ *
+ * Schedule an IB in the DMA ring.
+ */
+static void sdma_v7_0_ring_emit_ib(struct amdgpu_ring *ring,
+				   struct amdgpu_job *job,
+				   struct amdgpu_ib *ib,
+				   uint32_t flags)
+{
+	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
+	uint64_t csa_mc_addr = amdgpu_sdma_get_csa_mc_addr(ring, vmid);
+
+	/* An IB packet must end on a 8 DW boundary--the next dword
+	 * must be on a 8-dword boundary. Our IB packet below is 6
+	 * dwords long, thus add x number of NOPs, such that, in
+	 * modular arithmetic,
+	 * wptr + 6 + x = 8k, k >= 0, which in C is,
+	 * (wptr + 6 + x) % 8 = 0.
+	 * The expression below, is a solution of x.
+	 */
+	sdma_v7_0_ring_insert_nop(ring, (2 - lower_32_bits(ring->wptr)) & 7);
+
+	amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_INDIRECT) |
+			  SDMA_PKT_INDIRECT_HEADER_VMID(vmid & 0xf));
+	/* base must be 32 byte aligned */
+	amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr) & 0xffffffe0);
+	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
+	amdgpu_ring_write(ring, ib->length_dw);
+	amdgpu_ring_write(ring, lower_32_bits(csa_mc_addr));
+	amdgpu_ring_write(ring, upper_32_bits(csa_mc_addr));
+}
+
+/**
+ * sdma_v7_0_ring_emit_mem_sync - flush the IB by graphics cache rinse
+ *
+ * @ring: amdgpu ring pointer
+ *
+ * flush the IB by graphics cache rinse.
+ */
+static void sdma_v7_0_ring_emit_mem_sync(struct amdgpu_ring *ring)
+{
+	uint32_t gcr_cntl = SDMA_GCR_GL2_INV | SDMA_GCR_GL2_WB | SDMA_GCR_GLM_INV |
+		SDMA_GCR_GL1_INV | SDMA_GCR_GLV_INV | SDMA_GCR_GLK_INV |
+		SDMA_GCR_GLI_INV(1);
+
+	/* flush entire cache L0/L1/L2, this can be optimized by performance requirement */
+	amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_GCR_REQ));
+	amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD1_BASE_VA_31_7(0));
+	amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD2_GCR_CONTROL_15_0(gcr_cntl) |
+			  SDMA_PKT_GCR_REQ_PAYLOAD2_BASE_VA_47_32(0));
+	amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD3_LIMIT_VA_31_7(0) |
+			  SDMA_PKT_GCR_REQ_PAYLOAD3_GCR_CONTROL_18_16(gcr_cntl >> 16));
+	amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD4_LIMIT_VA_47_32(0) |
+			  SDMA_PKT_GCR_REQ_PAYLOAD4_VMID(0));
+}
+
+
+/**
+ * sdma_v7_0_ring_emit_hdp_flush - emit an hdp flush on the DMA ring
+ *
+ * @ring: amdgpu ring pointer
+ *
+ * Emit an hdp flush packet on the requested DMA ring.
+ */
+static void sdma_v7_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 ref_and_mask = 0;
+	const struct nbio_hdp_flush_reg *nbio_hf_reg = adev->nbio.hdp_flush_reg;
+
+	ref_and_mask = nbio_hf_reg->ref_and_mask_sdma0 << ring->me;
+
+	amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_POLL_REGMEM) |
+			  SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(1) |
+			  SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3)); /* == */
+	amdgpu_ring_write(ring, (adev->nbio.funcs->get_hdp_flush_done_offset(adev)) << 2);
+	amdgpu_ring_write(ring, (adev->nbio.funcs->get_hdp_flush_req_offset(adev)) << 2);
+	amdgpu_ring_write(ring, ref_and_mask); /* reference */
+	amdgpu_ring_write(ring, ref_and_mask); /* mask */
+	amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) |
+			  SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(10)); /* retry count, poll interval */
+}
+
+/**
+ * sdma_v7_0_ring_emit_fence - emit a fence on the DMA ring
+ *
+ * @ring: amdgpu ring pointer
+ * @addr: address
+ * @seq: fence seq number
+ * @flags: fence flags
+ *
+ * Add a DMA fence packet to the ring to write
+ * the fence seq number and DMA trap packet to generate
+ * an interrupt if needed.
+ */
+static void sdma_v7_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq,
+				      unsigned flags)
+{
+	bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT;
+	/* write the fence */
+	amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_FENCE) |
+			  SDMA_PKT_FENCE_HEADER_MTYPE(0x3)); /* Ucached(UC) */
+	/* zero in first two bits */
+	BUG_ON(addr & 0x3);
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	amdgpu_ring_write(ring, lower_32_bits(seq));
+
+	/* optionally write high bits as well */
+	if (write64bit) {
+		addr += 4;
+		amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_FENCE) |
+				  SDMA_PKT_FENCE_HEADER_MTYPE(0x3));
+		/* zero in first two bits */
+		BUG_ON(addr & 0x3);
+		amdgpu_ring_write(ring, lower_32_bits(addr));
+		amdgpu_ring_write(ring, upper_32_bits(addr));
+		amdgpu_ring_write(ring, upper_32_bits(seq));
+	}
+
+	if (flags & AMDGPU_FENCE_FLAG_INT) {
+		/* generate an interrupt */
+		amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_TRAP));
+		amdgpu_ring_write(ring, SDMA_PKT_TRAP_INT_CONTEXT_INT_CONTEXT(0));
+	}
+}
+
+/**
+ * sdma_v7_0_gfx_stop - stop the gfx async dma engines
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Stop the gfx async dma ring buffers.
+ */
+static void sdma_v7_0_gfx_stop(struct amdgpu_device *adev)
+{
+	u32 rb_cntl, ib_cntl;
+	int i;
+
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		rb_cntl = RREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_CNTL));
+		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_ENABLE, 0);
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_CNTL), rb_cntl);
+		ib_cntl = RREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_IB_CNTL));
+		ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_QUEUE0_IB_CNTL, IB_ENABLE, 0);
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_IB_CNTL), ib_cntl);
+	}
+}
+
+/**
+ * sdma_v7_0_rlc_stop - stop the compute async dma engines
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Stop the compute async dma queues.
+ */
+static void sdma_v7_0_rlc_stop(struct amdgpu_device *adev)
+{
+	/* XXX todo */
+}
+
+/**
+ * sdma_v7_0_ctx_switch_enable - stop the async dma engines context switch
+ *
+ * @adev: amdgpu_device pointer
+ * @enable: enable/disable the DMA MEs context switch.
+ *
+ * Halt or unhalt the async dma engines context switch.
+ */
+static void sdma_v7_0_ctx_switch_enable(struct amdgpu_device *adev, bool enable)
+{
+}
+
+/**
+ * sdma_v7_0_enable - stop the async dma engines
+ *
+ * @adev: amdgpu_device pointer
+ * @enable: enable/disable the DMA MEs.
+ *
+ * Halt or unhalt the async dma engines.
+ */
+static void sdma_v7_0_enable(struct amdgpu_device *adev, bool enable)
+{
+	u32 mcu_cntl;
+	int i;
+
+	if (!enable) {
+		sdma_v7_0_gfx_stop(adev);
+		sdma_v7_0_rlc_stop(adev);
+	}
+
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		mcu_cntl = RREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_MCU_CNTL));
+		mcu_cntl = REG_SET_FIELD(mcu_cntl, SDMA0_MCU_CNTL, HALT, enable ? 0 : 1);
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_MCU_CNTL), mcu_cntl);
+	}
+}
+
+/**
+ * sdma_v7_0_gfx_resume_instance - start/restart a certain sdma engine
+ *
+ * @adev: amdgpu_device pointer
+ * @i: instance
+ * @restore: used to restore wptr when restart
+ *
+ * Set up the gfx DMA ring buffers and enable them. On restart, we will restore wptr and rptr.
+ * Return 0 for success.
+ */
+static int sdma_v7_0_gfx_resume_instance(struct amdgpu_device *adev, int i, bool restore)
+{
+	struct amdgpu_ring *ring;
+	u32 rb_cntl, ib_cntl;
+	u32 rb_bufsz;
+	u32 doorbell;
+	u32 doorbell_offset;
+	u32 temp;
+	u64 wptr_gpu_addr;
+	int r;
+
+	ring = &adev->sdma.instance[i].ring;
+
+	/* Set ring buffer size in dwords */
+	rb_bufsz = order_base_2(ring->ring_size / 4);
+	rb_cntl = RREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_CNTL));
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_SIZE, rb_bufsz);
+#ifdef __BIG_ENDIAN
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_SWAP_ENABLE, 1);
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL,
+				RPTR_WRITEBACK_SWAP_ENABLE, 1);
+#endif
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_PRIV, 1);
+	WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_CNTL), rb_cntl);
+
+	/* Initialize the ring buffer's read and write pointers */
+	if (restore) {
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_RPTR), lower_32_bits(ring->wptr << 2));
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_RPTR_HI), upper_32_bits(ring->wptr << 2));
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR), lower_32_bits(ring->wptr << 2));
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR_HI), upper_32_bits(ring->wptr << 2));
+	} else {
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_RPTR), 0);
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_RPTR_HI), 0);
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR), 0);
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR_HI), 0);
+	}
+	/* setup the wptr shadow polling */
+	wptr_gpu_addr = ring->wptr_gpu_addr;
+	WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR_POLL_ADDR_LO),
+	       lower_32_bits(wptr_gpu_addr));
+	WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR_POLL_ADDR_HI),
+	       upper_32_bits(wptr_gpu_addr));
+
+	/* set the wb address whether it's enabled or not */
+	WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_RPTR_ADDR_HI),
+	       upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF);
+	WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_RPTR_ADDR_LO),
+	       lower_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFC);
+
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1);
+	if (amdgpu_sriov_vf(adev))
+		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, WPTR_POLL_ENABLE, 1);
+	else
+		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, WPTR_POLL_ENABLE, 0);
+
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, MCU_WPTR_POLL_ENABLE, 1);
+
+	WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_BASE), ring->gpu_addr >> 8);
+	WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_BASE_HI), ring->gpu_addr >> 40);
+
+	if (!restore)
+		ring->wptr = 0;
+
+	/* before programing wptr to a less value, need set minor_ptr_update first */
+	WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_MINOR_PTR_UPDATE), 1);
+
+	if (!amdgpu_sriov_vf(adev)) { /* only bare-metal use register write for wptr */
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR), lower_32_bits(ring->wptr) << 2);
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_WPTR_HI), upper_32_bits(ring->wptr) << 2);
+	}
+
+	doorbell = RREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_DOORBELL));
+	doorbell_offset = RREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_DOORBELL_OFFSET));
+
+	if (ring->use_doorbell) {
+		doorbell = REG_SET_FIELD(doorbell, SDMA0_QUEUE0_DOORBELL, ENABLE, 1);
+		doorbell_offset = REG_SET_FIELD(doorbell_offset, SDMA0_QUEUE0_DOORBELL_OFFSET,
+				OFFSET, ring->doorbell_index);
+	} else {
+		doorbell = REG_SET_FIELD(doorbell, SDMA0_QUEUE0_DOORBELL, ENABLE, 0);
+	}
+	WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_DOORBELL), doorbell);
+	WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_DOORBELL_OFFSET), doorbell_offset);
+
+	if (i == 0)
+		adev->nbio.funcs->sdma_doorbell_range(adev, i, ring->use_doorbell,
+					      ring->doorbell_index,
+					      adev->doorbell_index.sdma_doorbell_range * adev->sdma.num_instances);
+
+	if (amdgpu_sriov_vf(adev))
+		sdma_v7_0_ring_set_wptr(ring);
+
+	/* set minor_ptr_update to 0 after wptr programed */
+	WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_MINOR_PTR_UPDATE), 0);
+
+	/* Set up sdma hang watchdog */
+	temp = RREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_WATCHDOG_CNTL));
+	/* 100ms per unit */
+	temp = REG_SET_FIELD(temp, SDMA0_WATCHDOG_CNTL, QUEUE_HANG_COUNT,
+			     max(adev->usec_timeout/100000, 1));
+	WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_WATCHDOG_CNTL), temp);
+
+	/* Set up RESP_MODE to non-copy addresses */
+	temp = RREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_UTCL1_CNTL));
+	temp = REG_SET_FIELD(temp, SDMA0_UTCL1_CNTL, RESP_MODE, 3);
+	temp = REG_SET_FIELD(temp, SDMA0_UTCL1_CNTL, REDO_DELAY, 9);
+	WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_UTCL1_CNTL), temp);
+
+	/* program default cache read and write policy */
+	temp = RREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_UTCL1_PAGE));
+	/* clean read policy and write policy bits */
+	temp &= 0xFF0FFF;
+	temp |= ((CACHE_READ_POLICY_L2__DEFAULT << 12) |
+		 (CACHE_WRITE_POLICY_L2__DEFAULT << 14));
+	WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_UTCL1_PAGE), temp);
+
+	if (!amdgpu_sriov_vf(adev)) {
+		/* unhalt engine */
+		temp = RREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_MCU_CNTL));
+		temp = REG_SET_FIELD(temp, SDMA0_MCU_CNTL, HALT, 0);
+		temp = REG_SET_FIELD(temp, SDMA0_MCU_CNTL, RESET, 0);
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_MCU_CNTL), temp);
+	}
+
+	/* enable DMA RB */
+	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RB_ENABLE, 1);
+	WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_RB_CNTL), rb_cntl);
+
+	ib_cntl = RREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_IB_CNTL));
+	ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_QUEUE0_IB_CNTL, IB_ENABLE, 1);
+#ifdef __BIG_ENDIAN
+	ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_QUEUE0_IB_CNTL, IB_SWAP_ENABLE, 1);
+#endif
+	/* enable DMA IBs */
+	WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_IB_CNTL), ib_cntl);
+	ring->sched.ready = true;
+
+	if (amdgpu_sriov_vf(adev)) { /* bare-metal sequence doesn't need below to lines */
+		sdma_v7_0_ctx_switch_enable(adev, true);
+		sdma_v7_0_enable(adev, true);
+	}
+
+	r = amdgpu_ring_test_helper(ring);
+	if (r)
+		ring->sched.ready = false;
+
+	return r;
+}
+
+/**
+ * sdma_v7_0_gfx_resume - setup and start the async dma engines
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set up the gfx DMA ring buffers and enable them.
+ * Returns 0 for success, error for failure.
+ */
+static int sdma_v7_0_gfx_resume(struct amdgpu_device *adev)
+{
+	int i, r;
+
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		r = sdma_v7_0_gfx_resume_instance(adev, i, false);
+		if (r)
+			return r;
+	}
+
+	return 0;
+
+}
+
+/**
+ * sdma_v7_0_rlc_resume - setup and start the async dma engines
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set up the compute DMA queues and enable them.
+ * Returns 0 for success, error for failure.
+ */
+static int sdma_v7_0_rlc_resume(struct amdgpu_device *adev)
+{
+	return 0;
+}
+
+static void sdma_v12_0_free_ucode_buffer(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		amdgpu_bo_free_kernel(&adev->sdma.instance[i].sdma_fw_obj,
+				      &adev->sdma.instance[i].sdma_fw_gpu_addr,
+				      (void **)&adev->sdma.instance[i].sdma_fw_ptr);
+	}
+}
+
+/**
+ * sdma_v7_0_load_microcode - load the sDMA ME ucode
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Loads the sDMA0/1 ucode.
+ * Returns 0 for success, -EINVAL if the ucode is not available.
+ */
+static int sdma_v7_0_load_microcode(struct amdgpu_device *adev)
+{
+	const struct sdma_firmware_header_v3_0 *hdr;
+	const __le32 *fw_data;
+	u32 fw_size;
+	uint32_t tmp, sdma_status, ic_op_cntl;
+	int i, r, j;
+
+	/* halt the MEs */
+	sdma_v7_0_enable(adev, false);
+
+	if (!adev->sdma.instance[0].fw)
+		return -EINVAL;
+
+	hdr = (const struct sdma_firmware_header_v3_0 *)
+		adev->sdma.instance[0].fw->data;
+	amdgpu_ucode_print_sdma_hdr(&hdr->header);
+
+	fw_data = (const __le32 *)(adev->sdma.instance[0].fw->data +
+			le32_to_cpu(hdr->ucode_offset_bytes));
+	fw_size = le32_to_cpu(hdr->ucode_size_bytes);
+
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		r = amdgpu_bo_create_reserved(adev, fw_size,
+					      PAGE_SIZE,
+					      AMDGPU_GEM_DOMAIN_VRAM,
+					      &adev->sdma.instance[i].sdma_fw_obj,
+					      &adev->sdma.instance[i].sdma_fw_gpu_addr,
+					      (void **)&adev->sdma.instance[i].sdma_fw_ptr);
+		if (r) {
+			dev_err(adev->dev, "(%d) failed to create sdma ucode bo\n", r);
+			return r;
+		}
+
+		memcpy(adev->sdma.instance[i].sdma_fw_ptr, fw_data, fw_size);
+
+		amdgpu_bo_kunmap(adev->sdma.instance[i].sdma_fw_obj);
+		amdgpu_bo_unreserve(adev->sdma.instance[i].sdma_fw_obj);
+
+		tmp = RREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_IC_CNTL));
+		tmp = REG_SET_FIELD(tmp, SDMA0_IC_CNTL, GPA, 0);
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_IC_CNTL), tmp);
+
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_IC_BASE_LO),
+			lower_32_bits(adev->sdma.instance[i].sdma_fw_gpu_addr));
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_IC_BASE_HI),
+			upper_32_bits(adev->sdma.instance[i].sdma_fw_gpu_addr));
+
+		tmp = RREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_IC_OP_CNTL));
+		tmp = REG_SET_FIELD(tmp, SDMA0_IC_OP_CNTL, PRIME_ICACHE, 1);
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_IC_OP_CNTL), tmp);
+
+		/* Wait for sdma ucode init complete */
+		for (j = 0; j < adev->usec_timeout; j++) {
+			ic_op_cntl = RREG32_SOC15_IP(GC,
+					sdma_v7_0_get_reg_offset(adev, i, regSDMA0_IC_OP_CNTL));
+			sdma_status = RREG32_SOC15_IP(GC,
+					sdma_v7_0_get_reg_offset(adev, i, regSDMA0_STATUS_REG));
+			if ((REG_GET_FIELD(ic_op_cntl, SDMA0_IC_OP_CNTL, ICACHE_PRIMED) == 1) &&
+			    (REG_GET_FIELD(sdma_status, SDMA0_STATUS_REG, UCODE_INIT_DONE) == 1))
+				break;
+			udelay(1);
+		}
+
+		if (j >= adev->usec_timeout) {
+			dev_err(adev->dev, "failed to init sdma ucode\n");
+			return -EINVAL;
+		}
+	}
+
+	return 0;
+}
+
+static int sdma_v7_0_soft_reset(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	u32 tmp;
+	int i;
+
+	sdma_v7_0_gfx_stop(adev);
+
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		//tmp = RREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_FREEZE));
+		//tmp |= SDMA0_FREEZE__FREEZE_MASK;
+		//WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_FREEZE), tmp);
+		tmp = RREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_MCU_CNTL));
+		tmp |= SDMA0_MCU_CNTL__HALT_MASK;
+		tmp |= SDMA0_MCU_CNTL__RESET_MASK;
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_MCU_CNTL), tmp);
+
+		WREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_PREEMPT), 0);
+
+		udelay(100);
+
+		tmp = GRBM_SOFT_RESET__SOFT_RESET_SDMA0_MASK << i;
+		WREG32_SOC15(GC, 0, regGRBM_SOFT_RESET, tmp);
+		tmp = RREG32_SOC15(GC, 0, regGRBM_SOFT_RESET);
+
+		udelay(100);
+
+		WREG32_SOC15(GC, 0, regGRBM_SOFT_RESET, 0);
+		tmp = RREG32_SOC15(GC, 0, regGRBM_SOFT_RESET);
+
+		udelay(100);
+	}
+
+	return sdma_v7_0_start(adev);
+}
+
+static bool sdma_v7_0_check_soft_reset(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring;
+	int i, r;
+	long tmo = msecs_to_jiffies(1000);
+
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		ring = &adev->sdma.instance[i].ring;
+		r = amdgpu_ring_test_ib(ring, tmo);
+		if (r)
+			return true;
+	}
+
+	return false;
+}
+
+static int sdma_v7_0_reset_queue(struct amdgpu_ring *ring,
+				 unsigned int vmid,
+				 struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int r;
+
+	if (ring->me >= adev->sdma.num_instances) {
+		dev_err(adev->dev, "sdma instance not found\n");
+		return -EINVAL;
+	}
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	r = amdgpu_mes_reset_legacy_queue(adev, ring, vmid, true);
+	if (r)
+		return r;
+
+	r = sdma_v7_0_gfx_resume_instance(adev, ring->me, true);
+	if (r)
+		return r;
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+/**
+ * sdma_v7_0_start - setup and start the async dma engines
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set up the DMA engines and enable them.
+ * Returns 0 for success, error for failure.
+ */
+static int sdma_v7_0_start(struct amdgpu_device *adev)
+{
+	int r = 0;
+
+	if (amdgpu_sriov_vf(adev)) {
+		sdma_v7_0_ctx_switch_enable(adev, false);
+		sdma_v7_0_enable(adev, false);
+
+		/* set RB registers */
+		r = sdma_v7_0_gfx_resume(adev);
+		return r;
+	}
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
+		r = sdma_v7_0_load_microcode(adev);
+		if (r) {
+			sdma_v12_0_free_ucode_buffer(adev);
+			return r;
+		}
+
+		if (amdgpu_emu_mode == 1)
+			msleep(1000);
+	}
+
+	/* unhalt the MEs */
+	sdma_v7_0_enable(adev, true);
+	/* enable sdma ring preemption */
+	sdma_v7_0_ctx_switch_enable(adev, true);
+
+	/* start the gfx rings and rlc compute queues */
+	r = sdma_v7_0_gfx_resume(adev);
+	if (r)
+		return r;
+	r = sdma_v7_0_rlc_resume(adev);
+
+	return r;
+}
+
+static int sdma_v7_0_mqd_init(struct amdgpu_device *adev, void *mqd,
+			      struct amdgpu_mqd_prop *prop)
+{
+	struct v12_sdma_mqd *m = mqd;
+	uint64_t wb_gpu_addr;
+
+	m->sdmax_rlcx_rb_cntl =
+		order_base_2(prop->queue_size / 4) << SDMA0_QUEUE0_RB_CNTL__RB_SIZE__SHIFT |
+		1 << SDMA0_QUEUE0_RB_CNTL__RPTR_WRITEBACK_ENABLE__SHIFT |
+		4 << SDMA0_QUEUE0_RB_CNTL__RPTR_WRITEBACK_TIMER__SHIFT |
+		1 << SDMA0_QUEUE0_RB_CNTL__MCU_WPTR_POLL_ENABLE__SHIFT;
+
+	m->sdmax_rlcx_rb_base = lower_32_bits(prop->hqd_base_gpu_addr >> 8);
+	m->sdmax_rlcx_rb_base_hi = upper_32_bits(prop->hqd_base_gpu_addr >> 8);
+
+	wb_gpu_addr = prop->wptr_gpu_addr;
+	m->sdmax_rlcx_rb_wptr_poll_addr_lo = lower_32_bits(wb_gpu_addr);
+	m->sdmax_rlcx_rb_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr);
+
+	wb_gpu_addr = prop->rptr_gpu_addr;
+	m->sdmax_rlcx_rb_rptr_addr_lo = lower_32_bits(wb_gpu_addr);
+	m->sdmax_rlcx_rb_rptr_addr_hi = upper_32_bits(wb_gpu_addr);
+
+	m->sdmax_rlcx_ib_cntl = RREG32_SOC15_IP(GC, sdma_v7_0_get_reg_offset(adev, 0,
+							regSDMA0_QUEUE0_IB_CNTL));
+
+	m->sdmax_rlcx_doorbell_offset =
+		prop->doorbell_index << SDMA0_QUEUE0_DOORBELL_OFFSET__OFFSET__SHIFT;
+
+	m->sdmax_rlcx_doorbell = REG_SET_FIELD(0, SDMA0_QUEUE0_DOORBELL, ENABLE, 1);
+
+	m->sdmax_rlcx_doorbell_log = 0;
+	m->sdmax_rlcx_rb_aql_cntl = 0x4000;	//regSDMA0_QUEUE0_RB_AQL_CNTL_DEFAULT;
+	m->sdmax_rlcx_dummy_reg = 0xf;	//regSDMA0_QUEUE0_DUMMY_REG_DEFAULT;
+
+	m->sdmax_rlcx_csa_addr_lo = lower_32_bits(prop->csa_addr);
+	m->sdmax_rlcx_csa_addr_hi = upper_32_bits(prop->csa_addr);
+
+	m->sdmax_rlcx_mcu_dbg0 = lower_32_bits(prop->fence_address);
+	m->sdmax_rlcx_mcu_dbg1 = upper_32_bits(prop->fence_address);
+
+	return 0;
+}
+
+static void sdma_v7_0_set_mqd_funcs(struct amdgpu_device *adev)
+{
+	adev->mqds[AMDGPU_HW_IP_DMA].mqd_size = sizeof(struct v12_sdma_mqd);
+	adev->mqds[AMDGPU_HW_IP_DMA].init_mqd = sdma_v7_0_mqd_init;
+}
+
+/**
+ * sdma_v7_0_ring_test_ring - simple async dma engine test
+ *
+ * @ring: amdgpu_ring structure holding ring information
+ *
+ * Test the DMA engine by writing using it to write an
+ * value to memory.
+ * Returns 0 for success, error for failure.
+ */
+static int sdma_v7_0_ring_test_ring(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	unsigned i;
+	unsigned index;
+	int r;
+	u32 tmp;
+	u64 gpu_addr;
+
+	tmp = 0xCAFEDEAD;
+
+	r = amdgpu_device_wb_get(adev, &index);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r);
+		return r;
+	}
+
+	gpu_addr = adev->wb.gpu_addr + (index * 4);
+	adev->wb.wb[index] = cpu_to_le32(tmp);
+
+	r = amdgpu_ring_alloc(ring, 5);
+	if (r) {
+		DRM_ERROR("amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, r);
+		amdgpu_device_wb_free(adev, index);
+		return r;
+	}
+
+	amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_WRITE) |
+			  SDMA_PKT_COPY_LINEAR_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR));
+	amdgpu_ring_write(ring, lower_32_bits(gpu_addr));
+	amdgpu_ring_write(ring, upper_32_bits(gpu_addr));
+	amdgpu_ring_write(ring, SDMA_PKT_WRITE_UNTILED_DW_3_COUNT(0));
+	amdgpu_ring_write(ring, 0xDEADBEEF);
+	amdgpu_ring_commit(ring);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = le32_to_cpu(adev->wb.wb[index]);
+		if (tmp == 0xDEADBEEF)
+			break;
+		if (amdgpu_emu_mode == 1)
+			msleep(1);
+		else
+			udelay(1);
+	}
+
+	if (i >= adev->usec_timeout)
+		r = -ETIMEDOUT;
+
+	amdgpu_device_wb_free(adev, index);
+
+	return r;
+}
+
+/**
+ * sdma_v7_0_ring_test_ib - test an IB on the DMA engine
+ *
+ * @ring: amdgpu_ring structure holding ring information
+ * @timeout: timeout value in jiffies, or MAX_SCHEDULE_TIMEOUT
+ *
+ * Test a simple IB in the DMA ring.
+ * Returns 0 on success, error on failure.
+ */
+static int sdma_v7_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_ib ib;
+	struct dma_fence *f = NULL;
+	unsigned index;
+	long r;
+	u32 tmp = 0;
+	u64 gpu_addr;
+
+	tmp = 0xCAFEDEAD;
+	memset(&ib, 0, sizeof(ib));
+
+	r = amdgpu_device_wb_get(adev, &index);
+	if (r) {
+		dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r);
+		return r;
+	}
+
+	gpu_addr = adev->wb.gpu_addr + (index * 4);
+	adev->wb.wb[index] = cpu_to_le32(tmp);
+
+	r = amdgpu_ib_get(adev, NULL, 256, AMDGPU_IB_POOL_DIRECT, &ib);
+	if (r) {
+		DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
+		goto err0;
+	}
+
+	ib.ptr[0] = SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_WRITE) |
+		SDMA_PKT_COPY_LINEAR_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR);
+	ib.ptr[1] = lower_32_bits(gpu_addr);
+	ib.ptr[2] = upper_32_bits(gpu_addr);
+	ib.ptr[3] = SDMA_PKT_WRITE_UNTILED_DW_3_COUNT(0);
+	ib.ptr[4] = 0xDEADBEEF;
+	ib.ptr[5] = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP);
+	ib.ptr[6] = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP);
+	ib.ptr[7] = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP);
+	ib.length_dw = 8;
+
+	r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f);
+	if (r)
+		goto err1;
+
+	r = dma_fence_wait_timeout(f, false, timeout);
+	if (r == 0) {
+		DRM_ERROR("amdgpu: IB test timed out\n");
+		r = -ETIMEDOUT;
+		goto err1;
+	} else if (r < 0) {
+		DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
+		goto err1;
+	}
+
+	tmp = le32_to_cpu(adev->wb.wb[index]);
+
+	if (tmp == 0xDEADBEEF)
+		r = 0;
+	else
+		r = -EINVAL;
+
+err1:
+	amdgpu_ib_free(&ib, NULL);
+	dma_fence_put(f);
+err0:
+	amdgpu_device_wb_free(adev, index);
+	return r;
+}
+
+
+/**
+ * sdma_v7_0_vm_copy_pte - update PTEs by copying them from the GART
+ *
+ * @ib: indirect buffer to fill with commands
+ * @pe: addr of the page entry
+ * @src: src addr to copy from
+ * @count: number of page entries to update
+ *
+ * Update PTEs by copying them from the GART using sDMA.
+ */
+static void sdma_v7_0_vm_copy_pte(struct amdgpu_ib *ib,
+				  uint64_t pe, uint64_t src,
+				  unsigned count)
+{
+	unsigned bytes = count * 8;
+
+	ib->ptr[ib->length_dw++] = SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_COPY) |
+		SDMA_PKT_COPY_LINEAR_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR) |
+		SDMA_PKT_COPY_LINEAR_HEADER_CPV(1);
+
+	ib->ptr[ib->length_dw++] = bytes - 1;
+	ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */
+	ib->ptr[ib->length_dw++] = lower_32_bits(src);
+	ib->ptr[ib->length_dw++] = upper_32_bits(src);
+	ib->ptr[ib->length_dw++] = lower_32_bits(pe);
+	ib->ptr[ib->length_dw++] = upper_32_bits(pe);
+	ib->ptr[ib->length_dw++] = 0;
+
+}
+
+/**
+ * sdma_v7_0_vm_write_pte - update PTEs by writing them manually
+ *
+ * @ib: indirect buffer to fill with commands
+ * @pe: addr of the page entry
+ * @value: dst addr to write into pe
+ * @count: number of page entries to update
+ * @incr: increase next addr by incr bytes
+ *
+ * Update PTEs by writing them manually using sDMA.
+ */
+static void sdma_v7_0_vm_write_pte(struct amdgpu_ib *ib, uint64_t pe,
+				   uint64_t value, unsigned count,
+				   uint32_t incr)
+{
+	unsigned ndw = count * 2;
+
+	ib->ptr[ib->length_dw++] = SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_WRITE) |
+		SDMA_PKT_COPY_LINEAR_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR);
+	ib->ptr[ib->length_dw++] = lower_32_bits(pe);
+	ib->ptr[ib->length_dw++] = upper_32_bits(pe);
+	ib->ptr[ib->length_dw++] = ndw - 1;
+	for (; ndw > 0; ndw -= 2) {
+		ib->ptr[ib->length_dw++] = lower_32_bits(value);
+		ib->ptr[ib->length_dw++] = upper_32_bits(value);
+		value += incr;
+	}
+}
+
+/**
+ * sdma_v7_0_vm_set_pte_pde - update the page tables using sDMA
+ *
+ * @ib: indirect buffer to fill with commands
+ * @pe: addr of the page entry
+ * @addr: dst addr to write into pe
+ * @count: number of page entries to update
+ * @incr: increase next addr by incr bytes
+ * @flags: access flags
+ *
+ * Update the page tables using sDMA.
+ */
+static void sdma_v7_0_vm_set_pte_pde(struct amdgpu_ib *ib,
+				     uint64_t pe,
+				     uint64_t addr, unsigned count,
+				     uint32_t incr, uint64_t flags)
+{
+	/* for physically contiguous pages (vram) */
+	ib->ptr[ib->length_dw++] = SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_PTEPDE);
+	ib->ptr[ib->length_dw++] = lower_32_bits(pe); /* dst addr */
+	ib->ptr[ib->length_dw++] = upper_32_bits(pe);
+	ib->ptr[ib->length_dw++] = lower_32_bits(flags); /* mask */
+	ib->ptr[ib->length_dw++] = upper_32_bits(flags);
+	ib->ptr[ib->length_dw++] = lower_32_bits(addr); /* value */
+	ib->ptr[ib->length_dw++] = upper_32_bits(addr);
+	ib->ptr[ib->length_dw++] = incr; /* increment size */
+	ib->ptr[ib->length_dw++] = 0;
+	ib->ptr[ib->length_dw++] = count - 1; /* number of entries */
+}
+
+/**
+ * sdma_v7_0_ring_pad_ib - pad the IB
+ *
+ * @ring: amdgpu ring pointer
+ * @ib: indirect buffer to fill with padding
+ *
+ * Pad the IB with NOPs to a boundary multiple of 8.
+ */
+static void sdma_v7_0_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib)
+{
+	struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring);
+	u32 pad_count;
+	int i;
+
+	pad_count = (-ib->length_dw) & 0x7;
+	for (i = 0; i < pad_count; i++)
+		if (sdma && sdma->burst_nop && (i == 0))
+			ib->ptr[ib->length_dw++] =
+				SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_NOP) |
+				SDMA_PKT_NOP_HEADER_COUNT(pad_count - 1);
+		else
+			ib->ptr[ib->length_dw++] =
+				SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_NOP);
+}
+
+/**
+ * sdma_v7_0_ring_emit_pipeline_sync - sync the pipeline
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Make sure all previous operations are completed (CIK).
+ */
+static void sdma_v7_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
+{
+	uint32_t seq = ring->fence_drv.sync_seq;
+	uint64_t addr = ring->fence_drv.gpu_addr;
+
+	/* wait for idle */
+	amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_POLL_REGMEM) |
+			  SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(0) |
+			  SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3) | /* equal */
+			  SDMA_PKT_POLL_REGMEM_HEADER_MEM_POLL(1));
+	amdgpu_ring_write(ring, addr & 0xfffffffc);
+	amdgpu_ring_write(ring, upper_32_bits(addr) & 0xffffffff);
+	amdgpu_ring_write(ring, seq); /* reference */
+	amdgpu_ring_write(ring, 0xffffffff); /* mask */
+	amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) |
+			  SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(4)); /* retry count, poll interval */
+}
+
+/**
+ * sdma_v7_0_ring_emit_vm_flush - vm flush using sDMA
+ *
+ * @ring: amdgpu_ring pointer
+ * @vmid: vmid number to use
+ * @pd_addr: address
+ *
+ * Update the page table base and flush the VM TLB
+ * using sDMA.
+ */
+static void sdma_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
+					 unsigned vmid, uint64_t pd_addr)
+{
+	amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
+}
+
+static void sdma_v7_0_ring_emit_wreg(struct amdgpu_ring *ring,
+				     uint32_t reg, uint32_t val)
+{
+	/* SRBM WRITE command will not support on sdma v7.
+	 * Use Register WRITE command instead, which OPCODE is same as SRBM WRITE
+	 */
+	amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_SRBM_WRITE));
+	amdgpu_ring_write(ring, reg << 2);
+	amdgpu_ring_write(ring, val);
+}
+
+static void sdma_v7_0_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
+					 uint32_t val, uint32_t mask)
+{
+	amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_POLL_REGMEM) |
+			  SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(0) |
+			  SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3)); /* equal */
+	amdgpu_ring_write(ring, reg << 2);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, val); /* reference */
+	amdgpu_ring_write(ring, mask); /* mask */
+	amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) |
+			  SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(10));
+}
+
+static void sdma_v7_0_ring_emit_reg_write_reg_wait(struct amdgpu_ring *ring,
+						   uint32_t reg0, uint32_t reg1,
+						   uint32_t ref, uint32_t mask)
+{
+	amdgpu_ring_emit_wreg(ring, reg0, ref);
+	/* wait for a cycle to reset vm_inv_eng*_ack */
+	amdgpu_ring_emit_reg_wait(ring, reg0, 0, 0);
+	amdgpu_ring_emit_reg_wait(ring, reg1, mask, mask);
+}
+
+static int sdma_v7_0_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	switch (amdgpu_user_queue) {
+	case -1:
+	case 0:
+	default:
+		adev->sdma.no_user_submission = false;
+		adev->sdma.disable_uq = true;
+		break;
+	case 1:
+		adev->sdma.no_user_submission = false;
+		adev->sdma.disable_uq = false;
+		break;
+	case 2:
+		adev->sdma.no_user_submission = true;
+		adev->sdma.disable_uq = false;
+		break;
+	}
+
+	r = amdgpu_sdma_init_microcode(adev, 0, true);
+	if (r) {
+		DRM_ERROR("Failed to init sdma firmware!\n");
+		return r;
+	}
+
+	sdma_v7_0_set_ring_funcs(adev);
+	sdma_v7_0_set_buffer_funcs(adev);
+	sdma_v7_0_set_vm_pte_funcs(adev);
+	sdma_v7_0_set_irq_funcs(adev);
+	sdma_v7_0_set_mqd_funcs(adev);
+
+	return 0;
+}
+
+static int sdma_v7_0_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_ring *ring;
+	int r, i;
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_7_0);
+	uint32_t *ptr;
+
+	/* SDMA trap event */
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GFX,
+			      GFX_12_0_0__SRCID__SDMA_TRAP,
+			      &adev->sdma.trap_irq);
+	if (r)
+		return r;
+
+	/* SDMA user fence event */
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GFX,
+			      GFX_12_0_0__SRCID__SDMA_FENCE,
+			      &adev->sdma.fence_irq);
+	if (r)
+		return r;
+
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		ring = &adev->sdma.instance[i].ring;
+		ring->ring_obj = NULL;
+		ring->use_doorbell = true;
+		ring->me = i;
+		ring->no_user_submission = adev->sdma.no_user_submission;
+
+		DRM_DEBUG("SDMA %d use_doorbell being set to: [%s]\n", i,
+				ring->use_doorbell?"true":"false");
+
+		ring->doorbell_index =
+			(adev->doorbell_index.sdma_engine[i] << 1); // get DWORD offset
+
+		ring->vm_hub = AMDGPU_GFXHUB(0);
+		sprintf(ring->name, "sdma%d", i);
+		r = amdgpu_ring_init(adev, ring, 1024,
+				     &adev->sdma.trap_irq,
+				     AMDGPU_SDMA_IRQ_INSTANCE0 + i,
+				     AMDGPU_RING_PRIO_DEFAULT, NULL);
+		if (r)
+			return r;
+	}
+
+	adev->sdma.supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->sdma.instance[0].ring);
+	if (!amdgpu_sriov_vf(adev) &&
+	    !adev->debug_disable_gpu_ring_reset)
+		adev->sdma.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+
+	r = amdgpu_sdma_sysfs_reset_mask_init(adev);
+	if (r)
+		return r;
+	/* Allocate memory for SDMA IP Dump buffer */
+	ptr = kcalloc(adev->sdma.num_instances * reg_count, sizeof(uint32_t), GFP_KERNEL);
+	if (ptr)
+		adev->sdma.ip_dump = ptr;
+	else
+		DRM_ERROR("Failed to allocated memory for SDMA IP Dump\n");
+
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
+	case IP_VERSION(7, 0, 0):
+	case IP_VERSION(7, 0, 1):
+		if ((adev->sdma.instance[0].fw_version >= 7966358) && !adev->sdma.disable_uq)
+			adev->userq_funcs[AMDGPU_HW_IP_DMA] = &userq_mes_funcs;
+		break;
+	default:
+		break;
+	}
+
+	return r;
+}
+
+static int sdma_v7_0_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i;
+
+	for (i = 0; i < adev->sdma.num_instances; i++)
+		amdgpu_ring_fini(&adev->sdma.instance[i].ring);
+
+	amdgpu_sdma_sysfs_reset_mask_fini(adev);
+	amdgpu_sdma_destroy_inst_ctx(adev, true);
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT)
+		sdma_v12_0_free_ucode_buffer(adev);
+
+	kfree(adev->sdma.ip_dump);
+
+	return 0;
+}
+
+static int sdma_v7_0_set_userq_trap_interrupts(struct amdgpu_device *adev,
+					       bool enable)
+{
+	unsigned int irq_type;
+	int i, r;
+
+	if (adev->userq_funcs[AMDGPU_HW_IP_DMA]) {
+		for (i = 0; i < adev->sdma.num_instances; i++) {
+			irq_type = AMDGPU_SDMA_IRQ_INSTANCE0 + i;
+			if (enable)
+				r = amdgpu_irq_get(adev, &adev->sdma.trap_irq,
+						   irq_type);
+			else
+				r = amdgpu_irq_put(adev, &adev->sdma.trap_irq,
+						   irq_type);
+			if (r)
+				return r;
+		}
+	}
+
+	return 0;
+}
+
+static int sdma_v7_0_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = sdma_v7_0_start(adev);
+	if (r)
+		return r;
+
+	return sdma_v7_0_set_userq_trap_interrupts(adev, true);
+}
+
+static int sdma_v7_0_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (amdgpu_sriov_vf(adev))
+		return 0;
+
+	sdma_v7_0_ctx_switch_enable(adev, false);
+	sdma_v7_0_enable(adev, false);
+	sdma_v7_0_set_userq_trap_interrupts(adev, false);
+
+	return 0;
+}
+
+static int sdma_v7_0_suspend(struct amdgpu_ip_block *ip_block)
+{
+	return sdma_v7_0_hw_fini(ip_block);
+}
+
+static int sdma_v7_0_resume(struct amdgpu_ip_block *ip_block)
+{
+	return sdma_v7_0_hw_init(ip_block);
+}
+
+static bool sdma_v7_0_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	u32 i;
+
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		u32 tmp = RREG32(sdma_v7_0_get_reg_offset(adev, i, regSDMA0_STATUS_REG));
+
+		if (!(tmp & SDMA0_STATUS_REG__IDLE_MASK))
+			return false;
+	}
+
+	return true;
+}
+
+static int sdma_v7_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	unsigned i;
+	u32 sdma0, sdma1;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		sdma0 = RREG32(sdma_v7_0_get_reg_offset(adev, 0, regSDMA0_STATUS_REG));
+		sdma1 = RREG32(sdma_v7_0_get_reg_offset(adev, 1, regSDMA0_STATUS_REG));
+
+		if (sdma0 & sdma1 & SDMA0_STATUS_REG__IDLE_MASK)
+			return 0;
+		udelay(1);
+	}
+	return -ETIMEDOUT;
+}
+
+static int sdma_v7_0_ring_preempt_ib(struct amdgpu_ring *ring)
+{
+	int i, r = 0;
+	struct amdgpu_device *adev = ring->adev;
+	u32 index = 0;
+	u64 sdma_gfx_preempt;
+
+	amdgpu_sdma_get_index_from_ring(ring, &index);
+	sdma_gfx_preempt =
+		sdma_v7_0_get_reg_offset(adev, index, regSDMA0_QUEUE0_PREEMPT);
+
+	/* assert preemption condition */
+	amdgpu_ring_set_preempt_cond_exec(ring, false);
+
+	/* emit the trailing fence */
+	ring->trail_seq += 1;
+	r = amdgpu_ring_alloc(ring, 10);
+	if (r) {
+		DRM_ERROR("ring %d failed to be allocated \n", ring->idx);
+		return r;
+	}
+	sdma_v7_0_ring_emit_fence(ring, ring->trail_fence_gpu_addr,
+				  ring->trail_seq, 0);
+	amdgpu_ring_commit(ring);
+
+	/* assert IB preemption */
+	WREG32(sdma_gfx_preempt, 1);
+
+	/* poll the trailing fence */
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (ring->trail_seq ==
+		    le32_to_cpu(*(ring->trail_fence_cpu_addr)))
+			break;
+		udelay(1);
+	}
+
+	if (i >= adev->usec_timeout) {
+		r = -EINVAL;
+		DRM_ERROR("ring %d failed to be preempted\n", ring->idx);
+	}
+
+	/* deassert IB preemption */
+	WREG32(sdma_gfx_preempt, 0);
+
+	/* deassert the preemption condition */
+	amdgpu_ring_set_preempt_cond_exec(ring, true);
+	return r;
+}
+
+static int sdma_v7_0_set_trap_irq_state(struct amdgpu_device *adev,
+					struct amdgpu_irq_src *source,
+					unsigned type,
+					enum amdgpu_interrupt_state state)
+{
+	u32 sdma_cntl;
+
+	u32 reg_offset = sdma_v7_0_get_reg_offset(adev, type, regSDMA0_CNTL);
+
+	sdma_cntl = RREG32(reg_offset);
+	sdma_cntl = REG_SET_FIELD(sdma_cntl, SDMA0_CNTL, TRAP_ENABLE,
+		       state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+	WREG32(reg_offset, sdma_cntl);
+
+	return 0;
+}
+
+static int sdma_v7_0_process_trap_irq(struct amdgpu_device *adev,
+				      struct amdgpu_irq_src *source,
+				      struct amdgpu_iv_entry *entry)
+{
+	int instances, queue;
+
+	DRM_DEBUG("IH: SDMA trap\n");
+
+	queue = entry->ring_id & 0xf;
+	instances = (entry->ring_id & 0xf0) >> 4;
+	if (instances > 1) {
+		DRM_ERROR("IH: wrong ring_ID detected, as wrong sdma instance\n");
+		return -EINVAL;
+	}
+
+	switch (entry->client_id) {
+	case SOC21_IH_CLIENTID_GFX:
+		switch (queue) {
+		case 0:
+			amdgpu_fence_process(&adev->sdma.instance[instances].ring);
+			break;
+		default:
+			break;
+		}
+		break;
+	}
+	return 0;
+}
+
+static int sdma_v7_0_process_fence_irq(struct amdgpu_device *adev,
+				       struct amdgpu_irq_src *source,
+				       struct amdgpu_iv_entry *entry)
+{
+	u32 doorbell_offset = entry->src_data[0];
+
+	if (adev->enable_mes && doorbell_offset) {
+		struct amdgpu_userq_fence_driver *fence_drv = NULL;
+		struct xarray *xa = &adev->userq_xa;
+		unsigned long flags;
+
+		doorbell_offset >>= SDMA0_QUEUE0_DOORBELL_OFFSET__OFFSET__SHIFT;
+
+		xa_lock_irqsave(xa, flags);
+		fence_drv = xa_load(xa, doorbell_offset);
+		if (fence_drv)
+			amdgpu_userq_fence_driver_process(fence_drv);
+		xa_unlock_irqrestore(xa, flags);
+	}
+
+	return 0;
+}
+
+static int sdma_v7_0_process_illegal_inst_irq(struct amdgpu_device *adev,
+					      struct amdgpu_irq_src *source,
+					      struct amdgpu_iv_entry *entry)
+{
+	return 0;
+}
+
+static int sdma_v7_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					   enum amd_clockgating_state state)
+{
+	return 0;
+}
+
+static int sdma_v7_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_powergating_state state)
+{
+	return 0;
+}
+
+static void sdma_v7_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
+{
+}
+
+static void sdma_v7_0_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, j;
+	uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_7_0);
+	uint32_t instance_offset;
+
+	if (!adev->sdma.ip_dump)
+		return;
+
+	drm_printf(p, "num_instances:%d\n", adev->sdma.num_instances);
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		instance_offset = i * reg_count;
+		drm_printf(p, "\nInstance:%d\n", i);
+
+		for (j = 0; j < reg_count; j++)
+			drm_printf(p, "%-50s \t 0x%08x\n", sdma_reg_list_7_0[j].reg_name,
+				   adev->sdma.ip_dump[instance_offset + j]);
+	}
+}
+
+static void sdma_v7_0_dump_ip_state(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, j;
+	uint32_t instance_offset;
+	uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_7_0);
+
+	if (!adev->sdma.ip_dump)
+		return;
+
+	amdgpu_gfx_off_ctrl(adev, false);
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		instance_offset = i * reg_count;
+		for (j = 0; j < reg_count; j++)
+			adev->sdma.ip_dump[instance_offset + j] =
+				RREG32(sdma_v7_0_get_reg_offset(adev, i,
+				       sdma_reg_list_7_0[j].reg_offset));
+	}
+	amdgpu_gfx_off_ctrl(adev, true);
+}
+
+const struct amd_ip_funcs sdma_v7_0_ip_funcs = {
+	.name = "sdma_v7_0",
+	.early_init = sdma_v7_0_early_init,
+	.late_init = NULL,
+	.sw_init = sdma_v7_0_sw_init,
+	.sw_fini = sdma_v7_0_sw_fini,
+	.hw_init = sdma_v7_0_hw_init,
+	.hw_fini = sdma_v7_0_hw_fini,
+	.suspend = sdma_v7_0_suspend,
+	.resume = sdma_v7_0_resume,
+	.is_idle = sdma_v7_0_is_idle,
+	.wait_for_idle = sdma_v7_0_wait_for_idle,
+	.soft_reset = sdma_v7_0_soft_reset,
+	.check_soft_reset = sdma_v7_0_check_soft_reset,
+	.set_clockgating_state = sdma_v7_0_set_clockgating_state,
+	.set_powergating_state = sdma_v7_0_set_powergating_state,
+	.get_clockgating_state = sdma_v7_0_get_clockgating_state,
+	.dump_ip_state = sdma_v7_0_dump_ip_state,
+	.print_ip_state = sdma_v7_0_print_ip_state,
+};
+
+static const struct amdgpu_ring_funcs sdma_v7_0_ring_funcs = {
+	.type = AMDGPU_RING_TYPE_SDMA,
+	.align_mask = 0xf,
+	.nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP),
+	.support_64bit_ptrs = true,
+	.secure_submission_supported = true,
+	.get_rptr = sdma_v7_0_ring_get_rptr,
+	.get_wptr = sdma_v7_0_ring_get_wptr,
+	.set_wptr = sdma_v7_0_ring_set_wptr,
+	.emit_frame_size =
+		5 + /* sdma_v7_0_ring_init_cond_exec */
+		6 + /* sdma_v7_0_ring_emit_hdp_flush */
+		6 + /* sdma_v7_0_ring_emit_pipeline_sync */
+		/* sdma_v7_0_ring_emit_vm_flush */
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 6 +
+		10 + 10 + 10, /* sdma_v7_0_ring_emit_fence x3 for user fence, vm fence */
+	.emit_ib_size = 5 + 7 + 6, /* sdma_v7_0_ring_emit_ib */
+	.emit_ib = sdma_v7_0_ring_emit_ib,
+	.emit_mem_sync = sdma_v7_0_ring_emit_mem_sync,
+	.emit_fence = sdma_v7_0_ring_emit_fence,
+	.emit_pipeline_sync = sdma_v7_0_ring_emit_pipeline_sync,
+	.emit_vm_flush = sdma_v7_0_ring_emit_vm_flush,
+	.emit_hdp_flush = sdma_v7_0_ring_emit_hdp_flush,
+	.test_ring = sdma_v7_0_ring_test_ring,
+	.test_ib = sdma_v7_0_ring_test_ib,
+	.insert_nop = sdma_v7_0_ring_insert_nop,
+	.pad_ib = sdma_v7_0_ring_pad_ib,
+	.emit_wreg = sdma_v7_0_ring_emit_wreg,
+	.emit_reg_wait = sdma_v7_0_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = sdma_v7_0_ring_emit_reg_write_reg_wait,
+	.init_cond_exec = sdma_v7_0_ring_init_cond_exec,
+	.preempt_ib = sdma_v7_0_ring_preempt_ib,
+	.reset = sdma_v7_0_reset_queue,
+};
+
+static void sdma_v7_0_set_ring_funcs(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		adev->sdma.instance[i].ring.funcs = &sdma_v7_0_ring_funcs;
+		adev->sdma.instance[i].ring.me = i;
+	}
+}
+
+static const struct amdgpu_irq_src_funcs sdma_v7_0_trap_irq_funcs = {
+	.set = sdma_v7_0_set_trap_irq_state,
+	.process = sdma_v7_0_process_trap_irq,
+};
+
+static const struct amdgpu_irq_src_funcs sdma_v7_0_fence_irq_funcs = {
+	.process = sdma_v7_0_process_fence_irq,
+};
+
+static const struct amdgpu_irq_src_funcs sdma_v7_0_illegal_inst_irq_funcs = {
+	.process = sdma_v7_0_process_illegal_inst_irq,
+};
+
+static void sdma_v7_0_set_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->sdma.trap_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE0 +
+					adev->sdma.num_instances;
+	adev->sdma.trap_irq.funcs = &sdma_v7_0_trap_irq_funcs;
+	adev->sdma.fence_irq.funcs = &sdma_v7_0_fence_irq_funcs;
+	adev->sdma.illegal_inst_irq.funcs = &sdma_v7_0_illegal_inst_irq_funcs;
+}
+
+/**
+ * sdma_v7_0_emit_copy_buffer - copy buffer using the sDMA engine
+ *
+ * @ib: indirect buffer to fill with commands
+ * @src_offset: src GPU address
+ * @dst_offset: dst GPU address
+ * @byte_count: number of bytes to xfer
+ * @copy_flags: copy flags for the buffers
+ *
+ * Copy GPU buffers using the DMA engine.
+ * Used by the amdgpu ttm implementation to move pages if
+ * registered as the asic copy callback.
+ */
+static void sdma_v7_0_emit_copy_buffer(struct amdgpu_ib *ib,
+				       uint64_t src_offset,
+				       uint64_t dst_offset,
+				       uint32_t byte_count,
+				       uint32_t copy_flags)
+{
+	uint32_t num_type, data_format, max_com, write_cm;
+
+	max_com = AMDGPU_COPY_FLAGS_GET(copy_flags, MAX_COMPRESSED);
+	data_format = AMDGPU_COPY_FLAGS_GET(copy_flags, DATA_FORMAT);
+	num_type = AMDGPU_COPY_FLAGS_GET(copy_flags, NUMBER_TYPE);
+	write_cm = AMDGPU_COPY_FLAGS_GET(copy_flags, WRITE_COMPRESS_DISABLE) ? 2 : 1;
+
+	ib->ptr[ib->length_dw++] = SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_COPY) |
+		SDMA_PKT_COPY_LINEAR_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR) |
+		SDMA_PKT_COPY_LINEAR_HEADER_TMZ((copy_flags & AMDGPU_COPY_FLAGS_TMZ) ? 1 : 0) |
+		SDMA_PKT_COPY_LINEAR_HEADER_CPV(1);
+
+	ib->ptr[ib->length_dw++] = byte_count - 1;
+	ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */
+	ib->ptr[ib->length_dw++] = lower_32_bits(src_offset);
+	ib->ptr[ib->length_dw++] = upper_32_bits(src_offset);
+	ib->ptr[ib->length_dw++] = lower_32_bits(dst_offset);
+	ib->ptr[ib->length_dw++] = upper_32_bits(dst_offset);
+
+	if ((copy_flags & (AMDGPU_COPY_FLAGS_READ_DECOMPRESSED | AMDGPU_COPY_FLAGS_WRITE_COMPRESSED)))
+		ib->ptr[ib->length_dw++] = SDMA_DCC_DATA_FORMAT(data_format) | SDMA_DCC_NUM_TYPE(num_type) |
+			((copy_flags & AMDGPU_COPY_FLAGS_READ_DECOMPRESSED) ? SDMA_DCC_READ_CM(2) : 0) |
+			((copy_flags & AMDGPU_COPY_FLAGS_WRITE_COMPRESSED) ? SDMA_DCC_WRITE_CM(write_cm) : 0) |
+			SDMA_DCC_MAX_COM(max_com) | SDMA_DCC_MAX_UCOM(1);
+	else
+		ib->ptr[ib->length_dw++] = 0;
+}
+
+/**
+ * sdma_v7_0_emit_fill_buffer - fill buffer using the sDMA engine
+ *
+ * @ib: indirect buffer to fill
+ * @src_data: value to write to buffer
+ * @dst_offset: dst GPU address
+ * @byte_count: number of bytes to xfer
+ *
+ * Fill GPU buffers using the DMA engine.
+ */
+static void sdma_v7_0_emit_fill_buffer(struct amdgpu_ib *ib,
+				       uint32_t src_data,
+				       uint64_t dst_offset,
+				       uint32_t byte_count)
+{
+	ib->ptr[ib->length_dw++] = SDMA_PKT_CONSTANT_FILL_HEADER_OP(SDMA_OP_CONST_FILL) |
+		SDMA_PKT_CONSTANT_FILL_HEADER_COMPRESS(1);
+	ib->ptr[ib->length_dw++] = lower_32_bits(dst_offset);
+	ib->ptr[ib->length_dw++] = upper_32_bits(dst_offset);
+	ib->ptr[ib->length_dw++] = src_data;
+	ib->ptr[ib->length_dw++] = byte_count - 1;
+}
+
+static const struct amdgpu_buffer_funcs sdma_v7_0_buffer_funcs = {
+	.copy_max_bytes = 0x400000,
+	.copy_num_dw = 8,
+	.emit_copy_buffer = sdma_v7_0_emit_copy_buffer,
+	.fill_max_bytes = 0x400000,
+	.fill_num_dw = 5,
+	.emit_fill_buffer = sdma_v7_0_emit_fill_buffer,
+};
+
+static void sdma_v7_0_set_buffer_funcs(struct amdgpu_device *adev)
+{
+	adev->mman.buffer_funcs = &sdma_v7_0_buffer_funcs;
+	adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring;
+}
+
+static const struct amdgpu_vm_pte_funcs sdma_v7_0_vm_pte_funcs = {
+	.copy_pte_num_dw = 8,
+	.copy_pte = sdma_v7_0_vm_copy_pte,
+	.write_pte = sdma_v7_0_vm_write_pte,
+	.set_pte_pde = sdma_v7_0_vm_set_pte_pde,
+};
+
+static void sdma_v7_0_set_vm_pte_funcs(struct amdgpu_device *adev)
+{
+	unsigned i;
+
+	adev->vm_manager.vm_pte_funcs = &sdma_v7_0_vm_pte_funcs;
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		adev->vm_manager.vm_pte_scheds[i] =
+			&adev->sdma.instance[i].ring.sched;
+	}
+	adev->vm_manager.vm_pte_num_scheds = adev->sdma.num_instances;
+}
+
+const struct amdgpu_ip_block_version sdma_v7_0_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_SDMA,
+	.major = 7,
+	.minor = 0,
+	.rev = 0,
+	.funcs = &sdma_v7_0_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v7_0.h b/drivers/gpu/drm/amd/amdgpu/sdma_v7_0.h
new file mode 100644
index 000000000000..5af863bb39c4
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v7_0.h
@@ -0,0 +1,30 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __SDMA_V7_0_H__
+#define __SDMA_V7_0_H__
+
+extern const struct amd_ip_funcs sdma_v7_0_ip_funcs;
+extern const struct amdgpu_ip_block_version sdma_v7_0_ip_block;
+
+#endif /* __SDMA_V7_0_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/si.c b/drivers/gpu/drm/amd/amdgpu/si.c
index 7f99e130acd0..f7288372ee61 100644
--- a/drivers/gpu/drm/amd/amdgpu/si.c
+++ b/drivers/gpu/drm/amd/amdgpu/si.c
@@ -35,6 +35,7 @@
 #include "amdgpu_vce.h"
 #include "atom.h"
 #include "amd_pcie.h"
+
 #include "si_dpm.h"
 #include "sid.h"
 #include "si_ih.h"
@@ -44,17 +45,31 @@
 #include "dce_v6_0.h"
 #include "si.h"
 #include "uvd_v3_1.h"
-#include "amdgpu_vkms.h"
+#include "vce_v1_0.h"
+
+#include "uvd/uvd_4_0_d.h"
+
+#include "smu/smu_6_0_d.h"
+#include "smu/smu_6_0_sh_mask.h"
+
 #include "gca/gfx_6_0_d.h"
+#include "gca/gfx_6_0_sh_mask.h"
+
 #include "oss/oss_1_0_d.h"
 #include "oss/oss_1_0_sh_mask.h"
+
 #include "gmc/gmc_6_0_d.h"
+#include"gmc/gmc_6_0_sh_mask.h"
+
 #include "dce/dce_6_0_d.h"
-#include "uvd/uvd_4_0_d.h"
+#include "dce/dce_6_0_sh_mask.h"
+
 #include "bif/bif_3_0_d.h"
 #include "bif/bif_3_0_sh_mask.h"
+#include "si_enums.h"
 
 #include "amdgpu_dm.h"
+#include "amdgpu_vkms.h"
 
 static const u32 tahiti_golden_registers[] =
 {
@@ -907,9 +922,7 @@ static const u32 hainan_mgcg_cgcg_init[] =
 	0x3630, 0xfffffff0, 0x00000100,
 };
 
-/* XXX: update when we support VCE */
-#if 0
-/* tahiti, pitcarin, verde */
+/* tahiti, pitcairn, verde */
 static const struct amdgpu_video_codec_info tahiti_video_codecs_encode_array[] =
 {
 	{
@@ -926,13 +939,7 @@ static const struct amdgpu_video_codecs tahiti_video_codecs_encode =
 	.codec_count = ARRAY_SIZE(tahiti_video_codecs_encode_array),
 	.codec_array = tahiti_video_codecs_encode_array,
 };
-#else
-static const struct amdgpu_video_codecs tahiti_video_codecs_encode =
-{
-	.codec_count = 0,
-	.codec_array = NULL,
-};
-#endif
+
 /* oland and hainan don't support encode */
 static const struct amdgpu_video_codecs hainan_video_codecs_encode =
 {
@@ -940,7 +947,7 @@ static const struct amdgpu_video_codecs hainan_video_codecs_encode =
 	.codec_array = NULL,
 };
 
-/* tahiti, pitcarin, verde, oland */
+/* tahiti, pitcairn, verde, oland */
 static const struct amdgpu_video_codec_info tahiti_video_codecs_decode_array[] =
 {
 	{
@@ -1071,8 +1078,8 @@ static u32 si_smc_rreg(struct amdgpu_device *adev, u32 reg)
 	u32 r;
 
 	spin_lock_irqsave(&adev->smc_idx_lock, flags);
-	WREG32(SMC_IND_INDEX_0, (reg));
-	r = RREG32(SMC_IND_DATA_0);
+	WREG32(mmSMC_IND_INDEX_0, (reg));
+	r = RREG32(mmSMC_IND_DATA_0);
 	spin_unlock_irqrestore(&adev->smc_idx_lock, flags);
 	return r;
 }
@@ -1082,8 +1089,8 @@ static void si_smc_wreg(struct amdgpu_device *adev, u32 reg, u32 v)
 	unsigned long flags;
 
 	spin_lock_irqsave(&adev->smc_idx_lock, flags);
-	WREG32(SMC_IND_INDEX_0, (reg));
-	WREG32(SMC_IND_DATA_0, (v));
+	WREG32(mmSMC_IND_INDEX_0, (reg));
+	WREG32(mmSMC_IND_DATA_0, (v));
 	spin_unlock_irqrestore(&adev->smc_idx_lock, flags);
 }
 
@@ -1110,55 +1117,55 @@ static void si_uvd_ctx_wreg(struct amdgpu_device *adev, u32 reg, u32 v)
 }
 
 static struct amdgpu_allowed_register_entry si_allowed_read_registers[] = {
-	{GRBM_STATUS},
+	{mmGRBM_STATUS},
 	{mmGRBM_STATUS2},
 	{mmGRBM_STATUS_SE0},
 	{mmGRBM_STATUS_SE1},
 	{mmSRBM_STATUS},
 	{mmSRBM_STATUS2},
-	{DMA_STATUS_REG + DMA0_REGISTER_OFFSET},
-	{DMA_STATUS_REG + DMA1_REGISTER_OFFSET},
+	{mmDMA_STATUS_REG + DMA0_REGISTER_OFFSET},
+	{mmDMA_STATUS_REG + DMA1_REGISTER_OFFSET},
 	{mmCP_STAT},
 	{mmCP_STALLED_STAT1},
 	{mmCP_STALLED_STAT2},
 	{mmCP_STALLED_STAT3},
-	{GB_ADDR_CONFIG},
-	{MC_ARB_RAMCFG},
-	{GB_TILE_MODE0},
-	{GB_TILE_MODE1},
-	{GB_TILE_MODE2},
-	{GB_TILE_MODE3},
-	{GB_TILE_MODE4},
-	{GB_TILE_MODE5},
-	{GB_TILE_MODE6},
-	{GB_TILE_MODE7},
-	{GB_TILE_MODE8},
-	{GB_TILE_MODE9},
-	{GB_TILE_MODE10},
-	{GB_TILE_MODE11},
-	{GB_TILE_MODE12},
-	{GB_TILE_MODE13},
-	{GB_TILE_MODE14},
-	{GB_TILE_MODE15},
-	{GB_TILE_MODE16},
-	{GB_TILE_MODE17},
-	{GB_TILE_MODE18},
-	{GB_TILE_MODE19},
-	{GB_TILE_MODE20},
-	{GB_TILE_MODE21},
-	{GB_TILE_MODE22},
-	{GB_TILE_MODE23},
-	{GB_TILE_MODE24},
-	{GB_TILE_MODE25},
-	{GB_TILE_MODE26},
-	{GB_TILE_MODE27},
-	{GB_TILE_MODE28},
-	{GB_TILE_MODE29},
-	{GB_TILE_MODE30},
-	{GB_TILE_MODE31},
-	{CC_RB_BACKEND_DISABLE, true},
-	{GC_USER_RB_BACKEND_DISABLE, true},
-	{PA_SC_RASTER_CONFIG, true},
+	{mmGB_ADDR_CONFIG},
+	{mmMC_ARB_RAMCFG},
+	{mmGB_TILE_MODE0},
+	{mmGB_TILE_MODE1},
+	{mmGB_TILE_MODE2},
+	{mmGB_TILE_MODE3},
+	{mmGB_TILE_MODE4},
+	{mmGB_TILE_MODE5},
+	{mmGB_TILE_MODE6},
+	{mmGB_TILE_MODE7},
+	{mmGB_TILE_MODE8},
+	{mmGB_TILE_MODE9},
+	{mmGB_TILE_MODE10},
+	{mmGB_TILE_MODE11},
+	{mmGB_TILE_MODE12},
+	{mmGB_TILE_MODE13},
+	{mmGB_TILE_MODE14},
+	{mmGB_TILE_MODE15},
+	{mmGB_TILE_MODE16},
+	{mmGB_TILE_MODE17},
+	{mmGB_TILE_MODE18},
+	{mmGB_TILE_MODE19},
+	{mmGB_TILE_MODE20},
+	{mmGB_TILE_MODE21},
+	{mmGB_TILE_MODE22},
+	{mmGB_TILE_MODE23},
+	{mmGB_TILE_MODE24},
+	{mmGB_TILE_MODE25},
+	{mmGB_TILE_MODE26},
+	{mmGB_TILE_MODE27},
+	{mmGB_TILE_MODE28},
+	{mmGB_TILE_MODE29},
+	{mmGB_TILE_MODE30},
+	{mmGB_TILE_MODE31},
+	{mmCC_RB_BACKEND_DISABLE, true},
+	{mmGC_USER_RB_BACKEND_DISABLE, true},
+	{mmPA_SC_RASTER_CONFIG, true},
 };
 
 static uint32_t si_get_register_value(struct amdgpu_device *adev,
@@ -1181,12 +1188,12 @@ static uint32_t si_get_register_value(struct amdgpu_device *adev,
 
 		mutex_lock(&adev->grbm_idx_mutex);
 		if (se_num != 0xffffffff || sh_num != 0xffffffff)
-			amdgpu_gfx_select_se_sh(adev, se_num, sh_num, 0xffffffff);
+			amdgpu_gfx_select_se_sh(adev, se_num, sh_num, 0xffffffff, 0);
 
 		val = RREG32(reg_offset);
 
 		if (se_num != 0xffffffff || sh_num != 0xffffffff)
-			amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+			amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 		mutex_unlock(&adev->grbm_idx_mutex);
 		return val;
 	} else {
@@ -1264,37 +1271,37 @@ static bool si_read_disabled_bios(struct amdgpu_device *adev)
 	u32 rom_cntl;
 	bool r;
 
-	bus_cntl = RREG32(R600_BUS_CNTL);
+	bus_cntl = RREG32(mmBUS_CNTL);
 	if (adev->mode_info.num_crtc) {
-		d1vga_control = RREG32(AVIVO_D1VGA_CONTROL);
-		d2vga_control = RREG32(AVIVO_D2VGA_CONTROL);
-		vga_render_control = RREG32(VGA_RENDER_CONTROL);
+		d1vga_control = RREG32(mmD1VGA_CONTROL);
+		d2vga_control = RREG32(mmD2VGA_CONTROL);
+		vga_render_control = RREG32(mmVGA_RENDER_CONTROL);
 	}
 	rom_cntl = RREG32(R600_ROM_CNTL);
 
 	/* enable the rom */
-	WREG32(R600_BUS_CNTL, (bus_cntl & ~R600_BIOS_ROM_DIS));
+	WREG32(mmBUS_CNTL, (bus_cntl & ~BUS_CNTL__BIOS_ROM_DIS_MASK));
 	if (adev->mode_info.num_crtc) {
 		/* Disable VGA mode */
-		WREG32(AVIVO_D1VGA_CONTROL,
-		       (d1vga_control & ~(AVIVO_DVGA_CONTROL_MODE_ENABLE |
-					  AVIVO_DVGA_CONTROL_TIMING_SELECT)));
-		WREG32(AVIVO_D2VGA_CONTROL,
-		       (d2vga_control & ~(AVIVO_DVGA_CONTROL_MODE_ENABLE |
-					  AVIVO_DVGA_CONTROL_TIMING_SELECT)));
-		WREG32(VGA_RENDER_CONTROL,
-		       (vga_render_control & C_000300_VGA_VSTATUS_CNTL));
+		WREG32(mmD1VGA_CONTROL,
+		       (d1vga_control & ~(D1VGA_CONTROL__D1VGA_MODE_ENABLE_MASK |
+					  D1VGA_CONTROL__D1VGA_TIMING_SELECT_MASK)));
+		WREG32(mmD2VGA_CONTROL,
+		       (d2vga_control & ~(D1VGA_CONTROL__D1VGA_MODE_ENABLE_MASK |
+					  D1VGA_CONTROL__D1VGA_TIMING_SELECT_MASK)));
+		WREG32(mmVGA_RENDER_CONTROL,
+		       (vga_render_control & ~VGA_RENDER_CONTROL__VGA_VSTATUS_CNTL_MASK));
 	}
 	WREG32(R600_ROM_CNTL, rom_cntl | R600_SCK_OVERWRITE);
 
 	r = amdgpu_read_bios(adev);
 
 	/* restore regs */
-	WREG32(R600_BUS_CNTL, bus_cntl);
+	WREG32(mmBUS_CNTL, bus_cntl);
 	if (adev->mode_info.num_crtc) {
-		WREG32(AVIVO_D1VGA_CONTROL, d1vga_control);
-		WREG32(AVIVO_D2VGA_CONTROL, d2vga_control);
-		WREG32(VGA_RENDER_CONTROL, vga_render_control);
+		WREG32(mmD1VGA_CONTROL, d1vga_control);
+		WREG32(mmD2VGA_CONTROL, d2vga_control);
+		WREG32(mmVGA_RENDER_CONTROL, vga_render_control);
 	}
 	WREG32(R600_ROM_CNTL, rom_cntl);
 	return r;
@@ -1331,23 +1338,24 @@ static void si_set_clk_bypass_mode(struct amdgpu_device *adev)
 {
 	u32 tmp, i;
 
-	tmp = RREG32(CG_SPLL_FUNC_CNTL);
-	tmp |= SPLL_BYPASS_EN;
-	WREG32(CG_SPLL_FUNC_CNTL, tmp);
+	tmp = RREG32(mmCG_SPLL_FUNC_CNTL);
+	tmp |= CG_SPLL_FUNC_CNTL__SPLL_BYPASS_EN_MASK;
+	WREG32(mmCG_SPLL_FUNC_CNTL, tmp);
 
-	tmp = RREG32(CG_SPLL_FUNC_CNTL_2);
-	tmp |= SPLL_CTLREQ_CHG;
-	WREG32(CG_SPLL_FUNC_CNTL_2, tmp);
+	tmp = RREG32(mmCG_SPLL_FUNC_CNTL_2);
+	tmp |= CG_SPLL_FUNC_CNTL_2__SPLL_CTLREQ_CHG_MASK;
+	WREG32(mmCG_SPLL_FUNC_CNTL_2, tmp);
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		if (RREG32(SPLL_STATUS) & SPLL_CHG_STATUS)
+		if (RREG32(mmCG_SPLL_STATUS) & CG_SPLL_STATUS__SPLL_CHG_STATUS_MASK)
 			break;
 		udelay(1);
 	}
 
-	tmp = RREG32(CG_SPLL_FUNC_CNTL_2);
-	tmp &= ~(SPLL_CTLREQ_CHG | SCLK_MUX_UPDATE);
-	WREG32(CG_SPLL_FUNC_CNTL_2, tmp);
+	tmp = RREG32(mmCG_SPLL_FUNC_CNTL_2);
+	tmp &= ~(CG_SPLL_FUNC_CNTL_2__SPLL_CTLREQ_CHG_MASK |
+		CG_SPLL_FUNC_CNTL_2__SCLK_MUX_UPDATE_MASK);
+	WREG32(mmCG_SPLL_FUNC_CNTL_2, tmp);
 
 	tmp = RREG32(MPLL_CNTL_MODE);
 	tmp &= ~MPLL_MCLK_SEL;
@@ -1358,21 +1366,21 @@ static void si_spll_powerdown(struct amdgpu_device *adev)
 {
 	u32 tmp;
 
-	tmp = RREG32(SPLL_CNTL_MODE);
-	tmp |= SPLL_SW_DIR_CONTROL;
-	WREG32(SPLL_CNTL_MODE, tmp);
+	tmp = RREG32(mmSPLL_CNTL_MODE);
+	tmp |= SPLL_CNTL_MODE__SPLL_SW_DIR_CONTROL_MASK;
+	WREG32(mmSPLL_CNTL_MODE, tmp);
 
-	tmp = RREG32(CG_SPLL_FUNC_CNTL);
-	tmp |= SPLL_RESET;
-	WREG32(CG_SPLL_FUNC_CNTL, tmp);
+	tmp = RREG32(mmCG_SPLL_FUNC_CNTL);
+	tmp |= CG_SPLL_FUNC_CNTL__SPLL_RESET_MASK;
+	WREG32(mmCG_SPLL_FUNC_CNTL, tmp);
 
-	tmp = RREG32(CG_SPLL_FUNC_CNTL);
-	tmp |= SPLL_SLEEP;
-	WREG32(CG_SPLL_FUNC_CNTL, tmp);
+	tmp = RREG32(mmCG_SPLL_FUNC_CNTL);
+	tmp |= CG_SPLL_FUNC_CNTL__SPLL_SLEEP_MASK;
+	WREG32(mmCG_SPLL_FUNC_CNTL, tmp);
 
-	tmp = RREG32(SPLL_CNTL_MODE);
-	tmp &= ~SPLL_SW_DIR_CONTROL;
-	WREG32(SPLL_CNTL_MODE, tmp);
+	tmp = RREG32(mmSPLL_CNTL_MODE);
+	tmp &= ~SPLL_CNTL_MODE__SPLL_SW_DIR_CONTROL_MASK;
+	WREG32(mmSPLL_CNTL_MODE, tmp);
 }
 
 static int si_gpu_pci_config_reset(struct amdgpu_device *adev)
@@ -1409,9 +1417,9 @@ static int si_gpu_pci_config_reset(struct amdgpu_device *adev)
 	return r;
 }
 
-static bool si_asic_supports_baco(struct amdgpu_device *adev)
+static int si_asic_supports_baco(struct amdgpu_device *adev)
 {
-	return false;
+	return 0;
 }
 
 static enum amd_reset_method
@@ -1454,14 +1462,14 @@ static void si_vga_set_state(struct amdgpu_device *adev, bool state)
 {
 	uint32_t temp;
 
-	temp = RREG32(CONFIG_CNTL);
+	temp = RREG32(mmCONFIG_CNTL);
 	if (!state) {
 		temp &= ~(1<<0);
 		temp |= (1<<1);
 	} else {
 		temp &= ~(1<<1);
 	}
-	WREG32(CONFIG_CNTL, temp);
+	WREG32(mmCONFIG_CNTL, temp);
 }
 
 static u32 si_get_xclk(struct amdgpu_device *adev)
@@ -1469,12 +1477,12 @@ static u32 si_get_xclk(struct amdgpu_device *adev)
 	u32 reference_clock = adev->clock.spll.reference_freq;
 	u32 tmp;
 
-	tmp = RREG32(CG_CLKPIN_CNTL_2);
-	if (tmp & MUX_TCLK_TO_XCLK)
+	tmp = RREG32(mmCG_CLKPIN_CNTL_2);
+	if (tmp & CG_CLKPIN_CNTL_2__MUX_TCLK_TO_XCLK_MASK)
 		return TCLK;
 
-	tmp = RREG32(CG_CLKPIN_CNTL);
-	if (tmp & XTALIN_DIVIDE)
+	tmp = RREG32(mmCG_CLKPIN_CNTL);
+	if (tmp & CG_CLKPIN_CNTL__XTALIN_DIVIDE_MASK)
 		return reference_clock / 4;
 
 	return reference_clock;
@@ -1519,9 +1527,9 @@ static int si_get_pcie_lanes(struct amdgpu_device *adev)
 	if (adev->flags & AMD_IS_APU)
 		return 0;
 
-	link_width_cntl = RREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL);
+	link_width_cntl = RREG32_PCIE_PORT(ixPCIE_LC_LINK_WIDTH_CNTL);
 
-	switch ((link_width_cntl & LC_LINK_WIDTH_RD_MASK) >> LC_LINK_WIDTH_RD_SHIFT) {
+	switch ((link_width_cntl & PCIE_LC_LINK_WIDTH_CNTL__LC_LINK_WIDTH_RD_MASK) >> PCIE_LC_LINK_WIDTH_CNTL__LC_LINK_WIDTH_RD__SHIFT) {
 	case LC_LINK_WIDTH_X1:
 		return 1;
 	case LC_LINK_WIDTH_X2:
@@ -1568,13 +1576,13 @@ static void si_set_pcie_lanes(struct amdgpu_device *adev, int lanes)
 		return;
 	}
 
-	link_width_cntl = RREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL);
-	link_width_cntl &= ~LC_LINK_WIDTH_MASK;
-	link_width_cntl |= mask << LC_LINK_WIDTH_SHIFT;
-	link_width_cntl |= (LC_RECONFIG_NOW |
-			    LC_RECONFIG_ARC_MISSING_ESCAPE);
+	link_width_cntl = RREG32_PCIE_PORT(ixPCIE_LC_LINK_WIDTH_CNTL);
+	link_width_cntl &= ~PCIE_LC_LINK_WIDTH_CNTL__LC_LINK_WIDTH_MASK;
+	link_width_cntl |= mask << PCIE_LC_LINK_WIDTH_CNTL__LC_LINK_WIDTH__SHIFT;
+	link_width_cntl |= (PCIE_LC_LINK_WIDTH_CNTL__LC_RECONFIG_NOW_MASK |
+		PCIE_LC_LINK_WIDTH_CNTL__LC_RECONFIG_ARC_MISSING_ESCAPE_MASK);
 
-	WREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl);
+	WREG32_PCIE_PORT(ixPCIE_LC_LINK_WIDTH_CNTL, link_width_cntl);
 }
 
 static void si_get_pcie_usage(struct amdgpu_device *adev, uint64_t *count0,
@@ -1888,7 +1896,7 @@ static int si_vce_send_vcepll_ctlreq(struct amdgpu_device *adev)
 	WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL, 0, ~UPLL_CTLREQ_MASK);
 
 	if (i == SI_MAX_CTLACKS_ASSERTION_WAIT) {
-		DRM_ERROR("Timeout setting UVD clocks!\n");
+		DRM_ERROR("Timeout setting VCE clocks!\n");
 		return -ETIMEDOUT;
 	}
 
@@ -1910,6 +1918,14 @@ static int si_set_vce_clocks(struct amdgpu_device *adev, u32 evclk, u32 ecclk)
 		     ~VCEPLL_BYPASS_EN_MASK);
 
 	if (!evclk || !ecclk) {
+		/*
+		 * On some chips, the PLL takes way too long to get out of
+		 * sleep mode, causing a timeout waiting on CTLACK/CTLACK2.
+		 * Leave the PLL running in bypass mode.
+		 */
+		if (adev->pdev->device == 0x6780)
+			return 0;
+
 		/* Keep the Bypass mode, put PLL to sleep */
 		WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL, VCEPLL_SLEEP_MASK,
 			     ~VCEPLL_SLEEP_MASK);
@@ -2018,13 +2034,13 @@ static const struct amdgpu_asic_funcs si_asic_funcs =
 
 static uint32_t si_get_rev_id(struct amdgpu_device *adev)
 {
-	return (RREG32(CC_DRM_ID_STRAPS) & CC_DRM_ID_STRAPS__ATI_REV_ID_MASK)
+	return (RREG32(mmCC_DRM_ID_STRAPS) & CC_DRM_ID_STRAPS__ATI_REV_ID_MASK)
 		>> CC_DRM_ID_STRAPS__ATI_REV_ID__SHIFT;
 }
 
-static int si_common_early_init(void *handle)
+static int si_common_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->smc_rreg = &si_smc_rreg;
 	adev->smc_wreg = &si_smc_wreg;
@@ -2148,17 +2164,6 @@ static int si_common_early_init(void *handle)
 	return 0;
 }
 
-static int si_common_sw_init(void *handle)
-{
-	return 0;
-}
-
-static int si_common_sw_fini(void *handle)
-{
-	return 0;
-}
-
-
 static void si_init_golden_registers(struct amdgpu_device *adev)
 {
 	switch (adev->asic_type) {
@@ -2250,9 +2255,9 @@ static void si_pcie_gen3_enable(struct amdgpu_device *adev)
 					CAIL_PCIE_LINK_SPEED_SUPPORT_GEN3)))
 		return;
 
-	speed_cntl = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL);
-	current_data_rate = (speed_cntl & LC_CURRENT_DATA_RATE_MASK) >>
-		LC_CURRENT_DATA_RATE_SHIFT;
+	speed_cntl = RREG32_PCIE_PORT(ixPCIE_LC_SPEED_CNTL);
+	current_data_rate = (speed_cntl & PCIE_LC_SPEED_CNTL__LC_CURRENT_DATA_RATE_MASK) >>
+		PCIE_LC_SPEED_CNTL__LC_CURRENT_DATA_RATE__SHIFT;
 	if (adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_GEN3) {
 		if (current_data_rate == 2) {
 			DRM_INFO("PCIE gen 3 link speeds already enabled\n");
@@ -2276,29 +2281,20 @@ static void si_pcie_gen3_enable(struct amdgpu_device *adev)
 			u16 bridge_cfg2, gpu_cfg2;
 			u32 max_lw, current_lw, tmp;
 
-			pcie_capability_read_word(root, PCI_EXP_LNKCTL,
-						  &bridge_cfg);
-			pcie_capability_read_word(adev->pdev, PCI_EXP_LNKCTL,
-						  &gpu_cfg);
+			pcie_capability_set_word(root, PCI_EXP_LNKCTL, PCI_EXP_LNKCTL_HAWD);
+			pcie_capability_set_word(adev->pdev, PCI_EXP_LNKCTL, PCI_EXP_LNKCTL_HAWD);
 
-			tmp16 = bridge_cfg | PCI_EXP_LNKCTL_HAWD;
-			pcie_capability_write_word(root, PCI_EXP_LNKCTL, tmp16);
-
-			tmp16 = gpu_cfg | PCI_EXP_LNKCTL_HAWD;
-			pcie_capability_write_word(adev->pdev, PCI_EXP_LNKCTL,
-						   tmp16);
-
-			tmp = RREG32_PCIE(PCIE_LC_STATUS1);
-			max_lw = (tmp & LC_DETECTED_LINK_WIDTH_MASK) >> LC_DETECTED_LINK_WIDTH_SHIFT;
-			current_lw = (tmp & LC_OPERATING_LINK_WIDTH_MASK) >> LC_OPERATING_LINK_WIDTH_SHIFT;
+			tmp = RREG32_PCIE(ixPCIE_LC_STATUS1);
+			max_lw = (tmp & PCIE_LC_STATUS1__LC_DETECTED_LINK_WIDTH_MASK) >> PCIE_LC_STATUS1__LC_DETECTED_LINK_WIDTH__SHIFT;
+			current_lw = (tmp & PCIE_LC_STATUS1__LC_OPERATING_LINK_WIDTH_MASK) >> PCIE_LC_STATUS1__LC_OPERATING_LINK_WIDTH__SHIFT;
 
 			if (current_lw < max_lw) {
-				tmp = RREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL);
-				if (tmp & LC_RENEGOTIATION_SUPPORT) {
-					tmp &= ~(LC_LINK_WIDTH_MASK | LC_UPCONFIGURE_DIS);
-					tmp |= (max_lw << LC_LINK_WIDTH_SHIFT);
-					tmp |= LC_UPCONFIGURE_SUPPORT | LC_RENEGOTIATE_EN | LC_RECONFIG_NOW;
-					WREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL, tmp);
+				tmp = RREG32_PCIE_PORT(ixPCIE_LC_LINK_WIDTH_CNTL);
+				if (tmp & PCIE_LC_LINK_WIDTH_CNTL__LC_RENEGOTIATION_SUPPORT_MASK) {
+					tmp &= ~(PCIE_LC_LINK_WIDTH_CNTL__LC_LINK_WIDTH_MASK | PCIE_LC_LINK_WIDTH_CNTL__LC_UPCONFIGURE_DIS_MASK);
+					tmp |= (max_lw << PCIE_LC_LINK_WIDTH_CNTL__LC_LINK_WIDTH__SHIFT);
+					tmp |= PCIE_LC_LINK_WIDTH_CNTL__LC_UPCONFIGURE_SUPPORT_MASK | PCIE_LC_LINK_WIDTH_CNTL__LC_RENEGOTIATE_EN_MASK | PCIE_LC_LINK_WIDTH_CNTL__LC_RECONFIG_NOW_MASK;
+					WREG32_PCIE_PORT(ixPCIE_LC_LINK_WIDTH_CNTL, tmp);
 				}
 			}
 
@@ -2321,84 +2317,66 @@ static void si_pcie_gen3_enable(struct amdgpu_device *adev)
 							  PCI_EXP_LNKCTL2,
 							  &gpu_cfg2);
 
-				tmp = RREG32_PCIE_PORT(PCIE_LC_CNTL4);
-				tmp |= LC_SET_QUIESCE;
-				WREG32_PCIE_PORT(PCIE_LC_CNTL4, tmp);
+				tmp = RREG32_PCIE_PORT(ixPCIE_LC_CNTL4);
+				tmp |= PCIE_LC_CNTL4__LC_SET_QUIESCE_MASK;
+				WREG32_PCIE_PORT(ixPCIE_LC_CNTL4, tmp);
 
-				tmp = RREG32_PCIE_PORT(PCIE_LC_CNTL4);
-				tmp |= LC_REDO_EQ;
-				WREG32_PCIE_PORT(PCIE_LC_CNTL4, tmp);
+				tmp = RREG32_PCIE_PORT(ixPCIE_LC_CNTL4);
+				tmp |= PCIE_LC_CNTL4__LC_REDO_EQ_MASK;
+				WREG32_PCIE_PORT(ixPCIE_LC_CNTL4, tmp);
 
 				mdelay(100);
 
-				pcie_capability_read_word(root, PCI_EXP_LNKCTL,
-							  &tmp16);
-				tmp16 &= ~PCI_EXP_LNKCTL_HAWD;
-				tmp16 |= (bridge_cfg & PCI_EXP_LNKCTL_HAWD);
-				pcie_capability_write_word(root, PCI_EXP_LNKCTL,
-							   tmp16);
-
-				pcie_capability_read_word(adev->pdev,
-							  PCI_EXP_LNKCTL,
-							  &tmp16);
-				tmp16 &= ~PCI_EXP_LNKCTL_HAWD;
-				tmp16 |= (gpu_cfg & PCI_EXP_LNKCTL_HAWD);
-				pcie_capability_write_word(adev->pdev,
-							   PCI_EXP_LNKCTL,
-							   tmp16);
-
-				pcie_capability_read_word(root, PCI_EXP_LNKCTL2,
-							  &tmp16);
-				tmp16 &= ~(PCI_EXP_LNKCTL2_ENTER_COMP |
-					   PCI_EXP_LNKCTL2_TX_MARGIN);
-				tmp16 |= (bridge_cfg2 &
-					  (PCI_EXP_LNKCTL2_ENTER_COMP |
-					   PCI_EXP_LNKCTL2_TX_MARGIN));
-				pcie_capability_write_word(root,
-							   PCI_EXP_LNKCTL2,
-							   tmp16);
-
-				pcie_capability_read_word(adev->pdev,
-							  PCI_EXP_LNKCTL2,
-							  &tmp16);
-				tmp16 &= ~(PCI_EXP_LNKCTL2_ENTER_COMP |
-					   PCI_EXP_LNKCTL2_TX_MARGIN);
-				tmp16 |= (gpu_cfg2 &
-					  (PCI_EXP_LNKCTL2_ENTER_COMP |
-					   PCI_EXP_LNKCTL2_TX_MARGIN));
-				pcie_capability_write_word(adev->pdev,
-							   PCI_EXP_LNKCTL2,
-							   tmp16);
-
-				tmp = RREG32_PCIE_PORT(PCIE_LC_CNTL4);
-				tmp &= ~LC_SET_QUIESCE;
-				WREG32_PCIE_PORT(PCIE_LC_CNTL4, tmp);
+				pcie_capability_clear_and_set_word(root, PCI_EXP_LNKCTL,
+								   PCI_EXP_LNKCTL_HAWD,
+								   bridge_cfg &
+								   PCI_EXP_LNKCTL_HAWD);
+				pcie_capability_clear_and_set_word(adev->pdev, PCI_EXP_LNKCTL,
+								   PCI_EXP_LNKCTL_HAWD,
+								   gpu_cfg &
+								   PCI_EXP_LNKCTL_HAWD);
+
+				pcie_capability_clear_and_set_word(root, PCI_EXP_LNKCTL2,
+								   PCI_EXP_LNKCTL2_ENTER_COMP |
+								   PCI_EXP_LNKCTL2_TX_MARGIN,
+								   bridge_cfg2 &
+								   (PCI_EXP_LNKCTL2_ENTER_COMP |
+								    PCI_EXP_LNKCTL2_TX_MARGIN));
+				pcie_capability_clear_and_set_word(adev->pdev, PCI_EXP_LNKCTL2,
+								   PCI_EXP_LNKCTL2_ENTER_COMP |
+								   PCI_EXP_LNKCTL2_TX_MARGIN,
+								   gpu_cfg2 &
+								   (PCI_EXP_LNKCTL2_ENTER_COMP |
+								    PCI_EXP_LNKCTL2_TX_MARGIN));
+
+				tmp = RREG32_PCIE_PORT(ixPCIE_LC_CNTL4);
+				tmp &= ~PCIE_LC_CNTL4__LC_SET_QUIESCE_MASK;
+				WREG32_PCIE_PORT(ixPCIE_LC_CNTL4, tmp);
 			}
 		}
 	}
 
-	speed_cntl |= LC_FORCE_EN_SW_SPEED_CHANGE | LC_FORCE_DIS_HW_SPEED_CHANGE;
-	speed_cntl &= ~LC_FORCE_DIS_SW_SPEED_CHANGE;
-	WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, speed_cntl);
-
-	pcie_capability_read_word(adev->pdev, PCI_EXP_LNKCTL2, &tmp16);
-	tmp16 &= ~PCI_EXP_LNKCTL2_TLS;
+	speed_cntl |= PCIE_LC_SPEED_CNTL__LC_FORCE_EN_SW_SPEED_CHANGE_MASK | PCIE_LC_SPEED_CNTL__LC_FORCE_DIS_HW_SPEED_CHANGE_MASK;
+	speed_cntl &= ~PCIE_LC_SPEED_CNTL__LC_FORCE_DIS_SW_SPEED_CHANGE_MASK;
+	WREG32_PCIE_PORT(ixPCIE_LC_SPEED_CNTL, speed_cntl);
 
+	tmp16 = 0;
 	if (adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_GEN3)
 		tmp16 |= PCI_EXP_LNKCTL2_TLS_8_0GT; /* gen3 */
 	else if (adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_GEN2)
 		tmp16 |= PCI_EXP_LNKCTL2_TLS_5_0GT; /* gen2 */
 	else
 		tmp16 |= PCI_EXP_LNKCTL2_TLS_2_5GT; /* gen1 */
-	pcie_capability_write_word(adev->pdev, PCI_EXP_LNKCTL2, tmp16);
+	pcie_capability_clear_and_set_word(adev->pdev, PCI_EXP_LNKCTL2,
+					   PCI_EXP_LNKCTL2_TLS, tmp16);
 
-	speed_cntl = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL);
-	speed_cntl |= LC_INITIATE_LINK_SPEED_CHANGE;
-	WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, speed_cntl);
+	speed_cntl = RREG32_PCIE_PORT(ixPCIE_LC_SPEED_CNTL);
+	speed_cntl |= PCIE_LC_SPEED_CNTL__LC_INITIATE_LINK_SPEED_CHANGE_MASK;
+	WREG32_PCIE_PORT(ixPCIE_LC_SPEED_CNTL, speed_cntl);
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		speed_cntl = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL);
-		if ((speed_cntl & LC_INITIATE_LINK_SPEED_CHANGE) == 0)
+		speed_cntl = RREG32_PCIE_PORT(ixPCIE_LC_SPEED_CNTL);
+		if ((speed_cntl & PCIE_LC_SPEED_CNTL__LC_INITIATE_LINK_SPEED_CHANGE_MASK) == 0)
 			break;
 		udelay(1);
 	}
@@ -2456,123 +2434,121 @@ static void si_program_aspm(struct amdgpu_device *adev)
 	if (!amdgpu_device_should_use_aspm(adev))
 		return;
 
-	if (adev->flags & AMD_IS_APU)
-		return;
-	orig = data = RREG32_PCIE_PORT(PCIE_LC_N_FTS_CNTL);
-	data &= ~LC_XMIT_N_FTS_MASK;
-	data |= LC_XMIT_N_FTS(0x24) | LC_XMIT_N_FTS_OVERRIDE_EN;
+	orig = data = RREG32_PCIE_PORT(ixPCIE_LC_N_FTS_CNTL);
+	data &= ~PCIE_LC_N_FTS_CNTL__LC_XMIT_N_FTS_MASK;
+	data |= (0x24 << PCIE_LC_N_FTS_CNTL__LC_XMIT_N_FTS__SHIFT) | PCIE_LC_N_FTS_CNTL__LC_XMIT_N_FTS_OVERRIDE_EN_MASK;
 	if (orig != data)
-		WREG32_PCIE_PORT(PCIE_LC_N_FTS_CNTL, data);
+		WREG32_PCIE_PORT(ixPCIE_LC_N_FTS_CNTL, data);
 
-	orig = data = RREG32_PCIE_PORT(PCIE_LC_CNTL3);
-	data |= LC_GO_TO_RECOVERY;
+	orig = data = RREG32_PCIE_PORT(ixPCIE_LC_CNTL3);
+	data |= PCIE_LC_CNTL3__LC_GO_TO_RECOVERY_MASK;
 	if (orig != data)
-		WREG32_PCIE_PORT(PCIE_LC_CNTL3, data);
+		WREG32_PCIE_PORT(ixPCIE_LC_CNTL3, data);
 
-	orig = data = RREG32_PCIE(PCIE_P_CNTL);
-	data |= P_IGNORE_EDB_ERR;
+	orig = data = RREG32_PCIE(ixPCIE_P_CNTL);
+	data |= PCIE_P_CNTL__P_IGNORE_EDB_ERR_MASK;
 	if (orig != data)
-		WREG32_PCIE(PCIE_P_CNTL, data);
+		WREG32_PCIE(ixPCIE_P_CNTL, data);
 
-	orig = data = RREG32_PCIE_PORT(PCIE_LC_CNTL);
-	data &= ~(LC_L0S_INACTIVITY_MASK | LC_L1_INACTIVITY_MASK);
-	data |= LC_PMI_TO_L1_DIS;
+	orig = data = RREG32_PCIE_PORT(ixPCIE_LC_CNTL);
+	data &= ~(PCIE_LC_CNTL__LC_L0S_INACTIVITY_MASK | PCIE_LC_CNTL__LC_L1_INACTIVITY_MASK);
+	data |= PCIE_LC_CNTL__LC_PMI_TO_L1_DIS_MASK;
 	if (!disable_l0s)
-		data |= LC_L0S_INACTIVITY(7);
+		data |= (7 << PCIE_LC_CNTL__LC_L0S_INACTIVITY__SHIFT);
 
 	if (!disable_l1) {
-		data |= LC_L1_INACTIVITY(7);
-		data &= ~LC_PMI_TO_L1_DIS;
+		data |= (7 << PCIE_LC_CNTL__LC_L1_INACTIVITY__SHIFT);
+		data &= ~PCIE_LC_CNTL__LC_PMI_TO_L1_DIS_MASK;
 		if (orig != data)
-			WREG32_PCIE_PORT(PCIE_LC_CNTL, data);
+			WREG32_PCIE_PORT(ixPCIE_LC_CNTL, data);
 
 		if (!disable_plloff_in_l1) {
 			bool clk_req_support;
 
-			orig = data = si_pif_phy0_rreg(adev,PB0_PIF_PWRDOWN_0);
-			data &= ~(PLL_POWER_STATE_IN_OFF_0_MASK | PLL_POWER_STATE_IN_TXS2_0_MASK);
-			data |= PLL_POWER_STATE_IN_OFF_0(7) | PLL_POWER_STATE_IN_TXS2_0(7);
+			orig = data = si_pif_phy0_rreg(adev,ixPB0_PIF_PWRDOWN_0);
+			data &= ~(PB0_PIF_PWRDOWN_0__PLL_POWER_STATE_IN_OFF_0_MASK | PB0_PIF_PWRDOWN_0__PLL_POWER_STATE_IN_TXS2_0_MASK);
+			data |= (7 << PB0_PIF_PWRDOWN_0__PLL_POWER_STATE_IN_OFF_0__SHIFT) | (7 << PB0_PIF_PWRDOWN_0__PLL_POWER_STATE_IN_TXS2_0__SHIFT);
 			if (orig != data)
-				si_pif_phy0_wreg(adev,PB0_PIF_PWRDOWN_0, data);
+				si_pif_phy0_wreg(adev,ixPB0_PIF_PWRDOWN_0, data);
 
-			orig = data = si_pif_phy0_rreg(adev,PB0_PIF_PWRDOWN_1);
-			data &= ~(PLL_POWER_STATE_IN_OFF_1_MASK | PLL_POWER_STATE_IN_TXS2_1_MASK);
-			data |= PLL_POWER_STATE_IN_OFF_1(7) | PLL_POWER_STATE_IN_TXS2_1(7);
+			orig = data = si_pif_phy0_rreg(adev,ixPB0_PIF_PWRDOWN_1);
+			data &= ~(PB0_PIF_PWRDOWN_1__PLL_POWER_STATE_IN_OFF_1_MASK | PB0_PIF_PWRDOWN_1__PLL_POWER_STATE_IN_TXS2_1_MASK);
+			data |= (7 << PB0_PIF_PWRDOWN_1__PLL_POWER_STATE_IN_OFF_1__SHIFT) | (7 << PB0_PIF_PWRDOWN_1__PLL_POWER_STATE_IN_TXS2_1__SHIFT);
 			if (orig != data)
-				si_pif_phy0_wreg(adev,PB0_PIF_PWRDOWN_1, data);
+				si_pif_phy0_wreg(adev,ixPB0_PIF_PWRDOWN_1, data);
 
-			orig = data = si_pif_phy1_rreg(adev,PB1_PIF_PWRDOWN_0);
-			data &= ~(PLL_POWER_STATE_IN_OFF_0_MASK | PLL_POWER_STATE_IN_TXS2_0_MASK);
-			data |= PLL_POWER_STATE_IN_OFF_0(7) | PLL_POWER_STATE_IN_TXS2_0(7);
+			orig = data = si_pif_phy1_rreg(adev,ixPB1_PIF_PWRDOWN_0);
+			data &= ~(PB1_PIF_PWRDOWN_0__PLL_POWER_STATE_IN_OFF_0_MASK | PB1_PIF_PWRDOWN_0__PLL_POWER_STATE_IN_TXS2_0_MASK);
+			data |= (7 << PB1_PIF_PWRDOWN_0__PLL_POWER_STATE_IN_OFF_0__SHIFT) | (7 << PB1_PIF_PWRDOWN_0__PLL_POWER_STATE_IN_TXS2_0__SHIFT);
 			if (orig != data)
-				si_pif_phy1_wreg(adev,PB1_PIF_PWRDOWN_0, data);
+				si_pif_phy1_wreg(adev,ixPB1_PIF_PWRDOWN_0, data);
 
-			orig = data = si_pif_phy1_rreg(adev,PB1_PIF_PWRDOWN_1);
-			data &= ~(PLL_POWER_STATE_IN_OFF_1_MASK | PLL_POWER_STATE_IN_TXS2_1_MASK);
-			data |= PLL_POWER_STATE_IN_OFF_1(7) | PLL_POWER_STATE_IN_TXS2_1(7);
+			orig = data = si_pif_phy1_rreg(adev,ixPB1_PIF_PWRDOWN_1);
+			data &= ~(PB1_PIF_PWRDOWN_1__PLL_POWER_STATE_IN_OFF_1_MASK | PB1_PIF_PWRDOWN_1__PLL_POWER_STATE_IN_TXS2_1_MASK);
+			data |= (7 << PB1_PIF_PWRDOWN_1__PLL_POWER_STATE_IN_OFF_1__SHIFT) | (7 << PB1_PIF_PWRDOWN_1__PLL_POWER_STATE_IN_TXS2_1__SHIFT);
 			if (orig != data)
-				si_pif_phy1_wreg(adev,PB1_PIF_PWRDOWN_1, data);
+				si_pif_phy1_wreg(adev,ixPB1_PIF_PWRDOWN_1, data);
 
 			if ((adev->asic_type != CHIP_OLAND) && (adev->asic_type != CHIP_HAINAN)) {
-				orig = data = si_pif_phy0_rreg(adev,PB0_PIF_PWRDOWN_0);
-				data &= ~PLL_RAMP_UP_TIME_0_MASK;
+				orig = data = si_pif_phy0_rreg(adev,ixPB0_PIF_PWRDOWN_0);
+				data &= ~PB0_PIF_PWRDOWN_0__PLL_RAMP_UP_TIME_0_MASK;
 				if (orig != data)
-					si_pif_phy0_wreg(adev,PB0_PIF_PWRDOWN_0, data);
+					si_pif_phy0_wreg(adev,ixPB0_PIF_PWRDOWN_0, data);
 
-				orig = data = si_pif_phy0_rreg(adev,PB0_PIF_PWRDOWN_1);
-				data &= ~PLL_RAMP_UP_TIME_1_MASK;
+				orig = data = si_pif_phy0_rreg(adev,ixPB0_PIF_PWRDOWN_1);
+				data &= ~PB0_PIF_PWRDOWN_1__PLL_RAMP_UP_TIME_1_MASK;
 				if (orig != data)
-					si_pif_phy0_wreg(adev,PB0_PIF_PWRDOWN_1, data);
+					si_pif_phy0_wreg(adev,ixPB0_PIF_PWRDOWN_1, data);
 
-				orig = data = si_pif_phy0_rreg(adev,PB0_PIF_PWRDOWN_2);
-				data &= ~PLL_RAMP_UP_TIME_2_MASK;
+				orig = data = si_pif_phy0_rreg(adev,ixPB0_PIF_PWRDOWN_2);
+				data &= ~PB0_PIF_PWRDOWN_2__PLL_RAMP_UP_TIME_2_MASK;
 				if (orig != data)
-					si_pif_phy0_wreg(adev,PB0_PIF_PWRDOWN_2, data);
+					si_pif_phy0_wreg(adev,ixPB0_PIF_PWRDOWN_2, data);
 
-				orig = data = si_pif_phy0_rreg(adev,PB0_PIF_PWRDOWN_3);
-				data &= ~PLL_RAMP_UP_TIME_3_MASK;
+				orig = data = si_pif_phy0_rreg(adev,ixPB0_PIF_PWRDOWN_3);
+				data &= ~PB0_PIF_PWRDOWN_3__PLL_RAMP_UP_TIME_3_MASK;
 				if (orig != data)
-					si_pif_phy0_wreg(adev,PB0_PIF_PWRDOWN_3, data);
+					si_pif_phy0_wreg(adev,ixPB0_PIF_PWRDOWN_3, data);
 
-				orig = data = si_pif_phy1_rreg(adev,PB1_PIF_PWRDOWN_0);
-				data &= ~PLL_RAMP_UP_TIME_0_MASK;
+				orig = data = si_pif_phy1_rreg(adev,ixPB1_PIF_PWRDOWN_0);
+				data &= ~PB1_PIF_PWRDOWN_0__PLL_RAMP_UP_TIME_0_MASK;
 				if (orig != data)
-					si_pif_phy1_wreg(adev,PB1_PIF_PWRDOWN_0, data);
+					si_pif_phy1_wreg(adev,ixPB1_PIF_PWRDOWN_0, data);
 
-				orig = data = si_pif_phy1_rreg(adev,PB1_PIF_PWRDOWN_1);
-				data &= ~PLL_RAMP_UP_TIME_1_MASK;
+				orig = data = si_pif_phy1_rreg(adev,ixPB1_PIF_PWRDOWN_1);
+				data &= ~PB1_PIF_PWRDOWN_1__PLL_RAMP_UP_TIME_1_MASK;
 				if (orig != data)
-					si_pif_phy1_wreg(adev,PB1_PIF_PWRDOWN_1, data);
+					si_pif_phy1_wreg(adev,ixPB1_PIF_PWRDOWN_1, data);
 
-				orig = data = si_pif_phy1_rreg(adev,PB1_PIF_PWRDOWN_2);
-				data &= ~PLL_RAMP_UP_TIME_2_MASK;
+				orig = data = si_pif_phy1_rreg(adev,ixPB1_PIF_PWRDOWN_2);
+				data &= ~PB1_PIF_PWRDOWN_2__PLL_RAMP_UP_TIME_2_MASK;
 				if (orig != data)
-					si_pif_phy1_wreg(adev,PB1_PIF_PWRDOWN_2, data);
+					si_pif_phy1_wreg(adev,ixPB1_PIF_PWRDOWN_2, data);
 
-				orig = data = si_pif_phy1_rreg(adev,PB1_PIF_PWRDOWN_3);
-				data &= ~PLL_RAMP_UP_TIME_3_MASK;
+				orig = data = si_pif_phy1_rreg(adev,ixPB1_PIF_PWRDOWN_3);
+				data &= ~PB1_PIF_PWRDOWN_3__PLL_RAMP_UP_TIME_3_MASK;
 				if (orig != data)
-					si_pif_phy1_wreg(adev,PB1_PIF_PWRDOWN_3, data);
+					si_pif_phy1_wreg(adev,ixPB1_PIF_PWRDOWN_3, data);
 			}
-			orig = data = RREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL);
-			data &= ~LC_DYN_LANES_PWR_STATE_MASK;
-			data |= LC_DYN_LANES_PWR_STATE(3);
+			orig = data = RREG32_PCIE_PORT(ixPCIE_LC_LINK_WIDTH_CNTL);
+			data &= ~PCIE_LC_LINK_WIDTH_CNTL__LC_DYN_LANES_PWR_STATE_MASK;
+			data |= (3 << PCIE_LC_LINK_WIDTH_CNTL__LC_DYN_LANES_PWR_STATE__SHIFT);
 			if (orig != data)
-				WREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL, data);
+				WREG32_PCIE_PORT(ixPCIE_LC_LINK_WIDTH_CNTL, data);
 
-			orig = data = si_pif_phy0_rreg(adev,PB0_PIF_CNTL);
-			data &= ~LS2_EXIT_TIME_MASK;
+			orig = data = si_pif_phy0_rreg(adev,ixPB0_PIF_CNTL);
+			data &= ~PB0_PIF_CNTL__LS2_EXIT_TIME_MASK;
 			if ((adev->asic_type == CHIP_OLAND) || (adev->asic_type == CHIP_HAINAN))
-				data |= LS2_EXIT_TIME(5);
+				data |= (5 << PB0_PIF_CNTL__LS2_EXIT_TIME__SHIFT);
 			if (orig != data)
-				si_pif_phy0_wreg(adev,PB0_PIF_CNTL, data);
+				si_pif_phy0_wreg(adev,ixPB0_PIF_CNTL, data);
 
-			orig = data = si_pif_phy1_rreg(adev,PB1_PIF_CNTL);
-			data &= ~LS2_EXIT_TIME_MASK;
+			orig = data = si_pif_phy1_rreg(adev,ixPB1_PIF_CNTL);
+			data &= ~PB1_PIF_CNTL__LS2_EXIT_TIME_MASK;
 			if ((adev->asic_type == CHIP_OLAND) || (adev->asic_type == CHIP_HAINAN))
-				data |= LS2_EXIT_TIME(5);
+				data |= (5 << PB1_PIF_CNTL__LS2_EXIT_TIME__SHIFT);
 			if (orig != data)
-				si_pif_phy1_wreg(adev,PB1_PIF_CNTL, data);
+				si_pif_phy1_wreg(adev,ixPB1_PIF_CNTL, data);
 
 			if (!disable_clkreq &&
 			    !pci_is_root_bus(adev->pdev->bus)) {
@@ -2588,64 +2564,64 @@ static void si_program_aspm(struct amdgpu_device *adev)
 			}
 
 			if (clk_req_support) {
-				orig = data = RREG32_PCIE_PORT(PCIE_LC_CNTL2);
-				data |= LC_ALLOW_PDWN_IN_L1 | LC_ALLOW_PDWN_IN_L23;
+				orig = data = RREG32_PCIE_PORT(ixPCIE_LC_CNTL2);
+				data |= PCIE_LC_CNTL2__LC_ALLOW_PDWN_IN_L1_MASK | PCIE_LC_CNTL2__LC_ALLOW_PDWN_IN_L23_MASK;
 				if (orig != data)
-					WREG32_PCIE_PORT(PCIE_LC_CNTL2, data);
+					WREG32_PCIE_PORT(ixPCIE_LC_CNTL2, data);
 
-				orig = data = RREG32(THM_CLK_CNTL);
-				data &= ~(CMON_CLK_SEL_MASK | TMON_CLK_SEL_MASK);
-				data |= CMON_CLK_SEL(1) | TMON_CLK_SEL(1);
+				orig = data = RREG32(mmTHM_CLK_CNTL);
+				data &= ~(THM_CLK_CNTL__CMON_CLK_SEL_MASK | THM_CLK_CNTL__TMON_CLK_SEL_MASK);
+				data |= (1 << THM_CLK_CNTL__CMON_CLK_SEL__SHIFT) | (1 << THM_CLK_CNTL__TMON_CLK_SEL__SHIFT);
 				if (orig != data)
-					WREG32(THM_CLK_CNTL, data);
+					WREG32(mmTHM_CLK_CNTL, data);
 
-				orig = data = RREG32(MISC_CLK_CNTL);
-				data &= ~(DEEP_SLEEP_CLK_SEL_MASK | ZCLK_SEL_MASK);
-				data |= DEEP_SLEEP_CLK_SEL(1) | ZCLK_SEL(1);
+				orig = data = RREG32(mmMISC_CLK_CNTL);
+				data &= ~(MISC_CLK_CNTL__DEEP_SLEEP_CLK_SEL_MASK | MISC_CLK_CNTL__ZCLK_SEL_MASK);
+				data |= (1 << MISC_CLK_CNTL__DEEP_SLEEP_CLK_SEL__SHIFT) | (1 << MISC_CLK_CNTL__ZCLK_SEL__SHIFT);
 				if (orig != data)
-					WREG32(MISC_CLK_CNTL, data);
+					WREG32(mmMISC_CLK_CNTL, data);
 
-				orig = data = RREG32(CG_CLKPIN_CNTL);
-				data &= ~BCLK_AS_XCLK;
+				orig = data = RREG32(mmCG_CLKPIN_CNTL);
+				data &= ~CG_CLKPIN_CNTL__BCLK_AS_XCLK_MASK;
 				if (orig != data)
-					WREG32(CG_CLKPIN_CNTL, data);
+					WREG32(mmCG_CLKPIN_CNTL, data);
 
-				orig = data = RREG32(CG_CLKPIN_CNTL_2);
-				data &= ~FORCE_BIF_REFCLK_EN;
+				orig = data = RREG32(mmCG_CLKPIN_CNTL_2);
+				data &= ~CG_CLKPIN_CNTL_2__FORCE_BIF_REFCLK_EN_MASK;
 				if (orig != data)
-					WREG32(CG_CLKPIN_CNTL_2, data);
+					WREG32(mmCG_CLKPIN_CNTL_2, data);
 
-				orig = data = RREG32(MPLL_BYPASSCLK_SEL);
-				data &= ~MPLL_CLKOUT_SEL_MASK;
-				data |= MPLL_CLKOUT_SEL(4);
+				orig = data = RREG32(mmMPLL_BYPASSCLK_SEL);
+				data &= ~MPLL_BYPASSCLK_SEL__MPLL_CLKOUT_SEL_MASK;
+				data |= 4 << MPLL_BYPASSCLK_SEL__MPLL_CLKOUT_SEL__SHIFT;
 				if (orig != data)
-					WREG32(MPLL_BYPASSCLK_SEL, data);
+					WREG32(mmMPLL_BYPASSCLK_SEL, data);
 
-				orig = data = RREG32(SPLL_CNTL_MODE);
-				data &= ~SPLL_REFCLK_SEL_MASK;
+				orig = data = RREG32(mmSPLL_CNTL_MODE);
+				data &= ~SPLL_CNTL_MODE__SPLL_REFCLK_SEL_MASK;
 				if (orig != data)
-					WREG32(SPLL_CNTL_MODE, data);
+					WREG32(mmSPLL_CNTL_MODE, data);
 			}
 		}
 	} else {
 		if (orig != data)
-			WREG32_PCIE_PORT(PCIE_LC_CNTL, data);
+			WREG32_PCIE_PORT(ixPCIE_LC_CNTL, data);
 	}
 
-	orig = data = RREG32_PCIE(PCIE_CNTL2);
-	data |= SLV_MEM_LS_EN | MST_MEM_LS_EN | REPLAY_MEM_LS_EN;
+	orig = data = RREG32_PCIE(ixPCIE_CNTL2);
+	data |= PCIE_CNTL2__SLV_MEM_LS_EN_MASK | PCIE_CNTL2__MST_MEM_LS_EN_MASK | PCIE_CNTL2__REPLAY_MEM_LS_EN_MASK;
 	if (orig != data)
-		WREG32_PCIE(PCIE_CNTL2, data);
+		WREG32_PCIE(ixPCIE_CNTL2, data);
 
 	if (!disable_l0s) {
-		data = RREG32_PCIE_PORT(PCIE_LC_N_FTS_CNTL);
-		if((data & LC_N_FTS_MASK) == LC_N_FTS_MASK) {
-			data = RREG32_PCIE(PCIE_LC_STATUS1);
-			if ((data & LC_REVERSE_XMIT) && (data & LC_REVERSE_RCVR)) {
-				orig = data = RREG32_PCIE_PORT(PCIE_LC_CNTL);
-				data &= ~LC_L0S_INACTIVITY_MASK;
+		data = RREG32_PCIE_PORT(ixPCIE_LC_N_FTS_CNTL);
+		if((data & PCIE_LC_N_FTS_CNTL__LC_N_FTS_MASK) == PCIE_LC_N_FTS_CNTL__LC_N_FTS_MASK) {
+			data = RREG32_PCIE(ixPCIE_LC_STATUS1);
+			if ((data & PCIE_LC_STATUS1__LC_REVERSE_XMIT_MASK) && (data & PCIE_LC_STATUS1__LC_REVERSE_RCVR_MASK)) {
+				orig = data = RREG32_PCIE_PORT(ixPCIE_LC_CNTL);
+				data &= ~PCIE_LC_CNTL__LC_L0S_INACTIVITY_MASK;
 				if (orig != data)
-					WREG32_PCIE_PORT(PCIE_LC_CNTL, data);
+					WREG32_PCIE_PORT(ixPCIE_LC_CNTL, data);
 			}
 		}
 	}
@@ -2662,9 +2638,9 @@ static void si_fix_pci_max_read_req_size(struct amdgpu_device *adev)
 		pcie_set_readrq(adev->pdev, 512);
 }
 
-static int si_common_hw_init(void *handle)
+static int si_common_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	si_fix_pci_max_read_req_size(adev);
 	si_init_golden_registers(adev);
@@ -2674,47 +2650,28 @@ static int si_common_hw_init(void *handle)
 	return 0;
 }
 
-static int si_common_hw_fini(void *handle)
+static int si_common_hw_fini(struct amdgpu_ip_block *ip_block)
 {
 	return 0;
 }
 
-static int si_common_suspend(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return si_common_hw_fini(adev);
-}
-
-static int si_common_resume(void *handle)
+static int si_common_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return si_common_hw_init(adev);
+	return si_common_hw_init(ip_block);
 }
 
-static bool si_common_is_idle(void *handle)
+static bool si_common_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	return true;
 }
 
-static int si_common_wait_for_idle(void *handle)
-{
-	return 0;
-}
-
-static int si_common_soft_reset(void *handle)
-{
-	return 0;
-}
-
-static int si_common_set_clockgating_state(void *handle,
+static int si_common_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					    enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int si_common_set_powergating_state(void *handle,
+static int si_common_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					    enum amd_powergating_state state)
 {
 	return 0;
@@ -2723,16 +2680,10 @@ static int si_common_set_powergating_state(void *handle,
 static const struct amd_ip_funcs si_common_ip_funcs = {
 	.name = "si_common",
 	.early_init = si_common_early_init,
-	.late_init = NULL,
-	.sw_init = si_common_sw_init,
-	.sw_fini = si_common_sw_fini,
 	.hw_init = si_common_hw_init,
 	.hw_fini = si_common_hw_fini,
-	.suspend = si_common_suspend,
 	.resume = si_common_resume,
 	.is_idle = si_common_is_idle,
-	.wait_for_idle = si_common_wait_for_idle,
-	.soft_reset = si_common_soft_reset,
 	.set_clockgating_state = si_common_set_clockgating_state,
 	.set_powergating_state = si_common_set_powergating_state,
 };
@@ -2767,7 +2718,7 @@ int si_set_ip_blocks(struct amdgpu_device *adev)
 		else
 			amdgpu_device_ip_block_add(adev, &dce_v6_0_ip_block);
 		amdgpu_device_ip_block_add(adev, &uvd_v3_1_ip_block);
-		/* amdgpu_device_ip_block_add(adev, &vce_v1_0_ip_block); */
+		amdgpu_device_ip_block_add(adev, &vce_v1_0_ip_block);
 		break;
 	case CHIP_OLAND:
 		amdgpu_device_ip_block_add(adev, &si_common_ip_block);
@@ -2785,7 +2736,6 @@ int si_set_ip_blocks(struct amdgpu_device *adev)
 		else
 			amdgpu_device_ip_block_add(adev, &dce_v6_4_ip_block);
 		amdgpu_device_ip_block_add(adev, &uvd_v3_1_ip_block);
-		/* amdgpu_device_ip_block_add(adev, &vce_v1_0_ip_block); */
 		break;
 	case CHIP_HAINAN:
 		amdgpu_device_ip_block_add(adev, &si_common_ip_block);
diff --git a/drivers/gpu/drm/amd/amdgpu/si_dma.c b/drivers/gpu/drm/amd/amdgpu/si_dma.c
index abca8b529721..7f18e4875287 100644
--- a/drivers/gpu/drm/amd/amdgpu/si_dma.c
+++ b/drivers/gpu/drm/amd/amdgpu/si_dma.c
@@ -27,6 +27,8 @@
 #include "si.h"
 #include "sid.h"
 
+#include "oss/oss_1_0_d.h"
+#include "oss/oss_1_0_sh_mask.h"
 const u32 sdma_offsets[SDMA_MAX_INSTANCE] =
 {
 	DMA0_REGISTER_OFFSET,
@@ -38,17 +40,31 @@ static void si_dma_set_buffer_funcs(struct amdgpu_device *adev);
 static void si_dma_set_vm_pte_funcs(struct amdgpu_device *adev);
 static void si_dma_set_irq_funcs(struct amdgpu_device *adev);
 
+/**
+ * si_dma_ring_get_rptr - get the current read pointer
+ *
+ * @ring: amdgpu ring pointer
+ *
+ * Get the current rptr from the hardware (SI).
+ */
 static uint64_t si_dma_ring_get_rptr(struct amdgpu_ring *ring)
 {
 	return *ring->rptr_cpu_addr;
 }
 
+/**
+ * si_dma_ring_get_wptr - get the current write pointer
+ *
+ * @ring: amdgpu ring pointer
+ *
+ * Get the current wptr from the hardware (SI).
+ */
 static uint64_t si_dma_ring_get_wptr(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
 	u32 me = (ring == &adev->sdma.instance[0].ring) ? 0 : 1;
 
-	return (RREG32(DMA_RB_WPTR + sdma_offsets[me]) & 0x3fffc) >> 2;
+	return (RREG32(mmDMA_GFX_RB_WPTR + sdma_offsets[me]) & 0x3fffc) >> 2;
 }
 
 static void si_dma_ring_set_wptr(struct amdgpu_ring *ring)
@@ -56,7 +72,7 @@ static void si_dma_ring_set_wptr(struct amdgpu_ring *ring)
 	struct amdgpu_device *adev = ring->adev;
 	u32 me = (ring == &adev->sdma.instance[0].ring) ? 0 : 1;
 
-	WREG32(DMA_RB_WPTR + sdma_offsets[me], (ring->wptr << 2) & 0x3fffc);
+	WREG32(mmDMA_GFX_RB_WPTR + sdma_offsets[me], (ring->wptr << 2) & 0x3fffc);
 }
 
 static void si_dma_ring_emit_ib(struct amdgpu_ring *ring,
@@ -115,13 +131,11 @@ static void si_dma_stop(struct amdgpu_device *adev)
 	u32 rb_cntl;
 	unsigned i;
 
-	amdgpu_sdma_unset_buffer_funcs_helper(adev);
-
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		/* dma0 */
-		rb_cntl = RREG32(DMA_RB_CNTL + sdma_offsets[i]);
-		rb_cntl &= ~DMA_RB_ENABLE;
-		WREG32(DMA_RB_CNTL + sdma_offsets[i], rb_cntl);
+		rb_cntl = RREG32(mmDMA_GFX_RB_CNTL + sdma_offsets[i]);
+		rb_cntl &= ~DMA_GFX_RB_CNTL__RB_ENABLE_MASK;
+		WREG32(mmDMA_GFX_RB_CNTL + sdma_offsets[i], rb_cntl);
 	}
 }
 
@@ -135,53 +149,48 @@ static int si_dma_start(struct amdgpu_device *adev)
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		ring = &adev->sdma.instance[i].ring;
 
-		WREG32(DMA_SEM_INCOMPLETE_TIMER_CNTL + sdma_offsets[i], 0);
-		WREG32(DMA_SEM_WAIT_FAIL_TIMER_CNTL + sdma_offsets[i], 0);
+		WREG32(mmDMA_SEM_INCOMPLETE_TIMER_CNTL + sdma_offsets[i], 0);
+		WREG32(mmDMA_SEM_WAIT_FAIL_TIMER_CNTL + sdma_offsets[i], 0);
 
 		/* Set ring buffer size in dwords */
 		rb_bufsz = order_base_2(ring->ring_size / 4);
 		rb_cntl = rb_bufsz << 1;
 #ifdef __BIG_ENDIAN
-		rb_cntl |= DMA_RB_SWAP_ENABLE | DMA_RPTR_WRITEBACK_SWAP_ENABLE;
+		rb_cntl |= DMA_GFX_RB_CNTL__RB_SWAP_ENABLE_MASK | DMA_GFX_RB_CNTL__RPTR_WRITEBACK_SWAP_ENABLE_MASK;
 #endif
-		WREG32(DMA_RB_CNTL + sdma_offsets[i], rb_cntl);
+		WREG32(mmDMA_GFX_RB_CNTL + sdma_offsets[i], rb_cntl);
 
 		/* Initialize the ring buffer's read and write pointers */
-		WREG32(DMA_RB_RPTR + sdma_offsets[i], 0);
-		WREG32(DMA_RB_WPTR + sdma_offsets[i], 0);
+		WREG32(mmDMA_GFX_RB_RPTR + sdma_offsets[i], 0);
+		WREG32(mmDMA_GFX_RB_WPTR + sdma_offsets[i], 0);
 
 		rptr_addr = ring->rptr_gpu_addr;
 
-		WREG32(DMA_RB_RPTR_ADDR_LO + sdma_offsets[i], lower_32_bits(rptr_addr));
-		WREG32(DMA_RB_RPTR_ADDR_HI + sdma_offsets[i], upper_32_bits(rptr_addr) & 0xFF);
+		WREG32(mmDMA_GFX_RB_RPTR_ADDR_LO + sdma_offsets[i], lower_32_bits(rptr_addr));
+		WREG32(mmDMA_GFX_RB_RPTR_ADDR_HI + sdma_offsets[i], upper_32_bits(rptr_addr) & 0xFF);
 
-		rb_cntl |= DMA_RPTR_WRITEBACK_ENABLE;
+		rb_cntl |= DMA_GFX_RB_CNTL__RPTR_WRITEBACK_ENABLE_MASK;
 
-		WREG32(DMA_RB_BASE + sdma_offsets[i], ring->gpu_addr >> 8);
+		WREG32(mmDMA_GFX_RB_BASE + sdma_offsets[i], ring->gpu_addr >> 8);
 
 		/* enable DMA IBs */
-		ib_cntl = DMA_IB_ENABLE | CMD_VMID_FORCE;
+		ib_cntl = DMA_GFX_IB_CNTL__IB_ENABLE_MASK | DMA_GFX_IB_CNTL__CMD_VMID_FORCE_MASK;
 #ifdef __BIG_ENDIAN
-		ib_cntl |= DMA_IB_SWAP_ENABLE;
+		ib_cntl |= DMA_GFX_IB_CNTL__IB_SWAP_ENABLE_MASK;
 #endif
-		WREG32(DMA_IB_CNTL + sdma_offsets[i], ib_cntl);
+		WREG32(mmDMA_GFX_IB_CNTL + sdma_offsets[i], ib_cntl);
 
-		dma_cntl = RREG32(DMA_CNTL + sdma_offsets[i]);
-		dma_cntl &= ~CTXEMPTY_INT_ENABLE;
-		WREG32(DMA_CNTL + sdma_offsets[i], dma_cntl);
+		dma_cntl = RREG32(mmDMA_CNTL + sdma_offsets[i]);
+		dma_cntl &= ~DMA_CNTL__CTXEMPTY_INT_ENABLE_MASK;
+		WREG32(mmDMA_CNTL + sdma_offsets[i], dma_cntl);
 
 		ring->wptr = 0;
-		WREG32(DMA_RB_WPTR + sdma_offsets[i], ring->wptr << 2);
-		WREG32(DMA_RB_CNTL + sdma_offsets[i], rb_cntl | DMA_RB_ENABLE);
-
-		ring->sched.ready = true;
+		WREG32(mmDMA_GFX_RB_WPTR + sdma_offsets[i], ring->wptr << 2);
+		WREG32(mmDMA_GFX_RB_CNTL + sdma_offsets[i], rb_cntl | DMA_GFX_RB_CNTL__RB_ENABLE_MASK);
 
 		r = amdgpu_ring_test_helper(ring);
 		if (r)
 			return r;
-
-		if (adev->mman.buffer_funcs_ring == ring)
-			amdgpu_ttm_set_buffer_funcs_status(adev, true);
 	}
 
 	return 0;
@@ -293,7 +302,7 @@ static int si_dma_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = -EINVAL;
 
 err1:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err0:
 	amdgpu_device_wb_free(adev, index);
@@ -464,11 +473,11 @@ static void si_dma_ring_emit_wreg(struct amdgpu_ring *ring,
 	amdgpu_ring_write(ring, val);
 }
 
-static int si_dma_early_init(void *handle)
+static int si_dma_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	adev->sdma.num_instances = 2;
+	adev->sdma.num_instances = SDMA_MAX_INSTANCE;
 
 	si_dma_set_ring_funcs(adev);
 	si_dma_set_buffer_funcs(adev);
@@ -478,11 +487,11 @@ static int si_dma_early_init(void *handle)
 	return 0;
 }
 
-static int si_dma_sw_init(void *handle)
+static int si_dma_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
 	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* DMA0 trap event */
 	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 224,
@@ -513,9 +522,9 @@ static int si_dma_sw_init(void *handle)
 	return r;
 }
 
-static int si_dma_sw_fini(void *handle)
+static int si_dma_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++)
@@ -524,61 +533,56 @@ static int si_dma_sw_fini(void *handle)
 	return 0;
 }
 
-static int si_dma_hw_init(void *handle)
+static int si_dma_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return si_dma_start(adev);
 }
 
-static int si_dma_hw_fini(void *handle)
+static int si_dma_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	si_dma_stop(adev);
+	si_dma_stop(ip_block->adev);
 
 	return 0;
 }
 
-static int si_dma_suspend(void *handle)
+static int si_dma_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return si_dma_hw_fini(adev);
+	return si_dma_hw_fini(ip_block);
 }
 
-static int si_dma_resume(void *handle)
+static int si_dma_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return si_dma_hw_init(adev);
+	return si_dma_hw_init(ip_block);
 }
 
-static bool si_dma_is_idle(void *handle)
+static bool si_dma_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	u32 tmp = RREG32(SRBM_STATUS2);
+	struct amdgpu_device *adev = ip_block->adev;
+
+	u32 tmp = RREG32(mmSRBM_STATUS2);
 
-	if (tmp & (DMA_BUSY_MASK | DMA1_BUSY_MASK))
+	if (tmp & (SRBM_STATUS2__DMA_BUSY_MASK | SRBM_STATUS2__DMA1_BUSY_MASK))
 	    return false;
 
 	return true;
 }
 
-static int si_dma_wait_for_idle(void *handle)
+static int si_dma_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		if (si_dma_is_idle(handle))
+		if (si_dma_is_idle(ip_block))
 			return 0;
 		udelay(1);
 	}
 	return -ETIMEDOUT;
 }
 
-static int si_dma_soft_reset(void *handle)
+static int si_dma_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	DRM_INFO("si_dma_soft_reset --- not implemented !!!!!!!\n");
 	return 0;
@@ -595,14 +599,14 @@ static int si_dma_set_trap_irq_state(struct amdgpu_device *adev,
 	case AMDGPU_SDMA_IRQ_INSTANCE0:
 		switch (state) {
 		case AMDGPU_IRQ_STATE_DISABLE:
-			sdma_cntl = RREG32(DMA_CNTL + DMA0_REGISTER_OFFSET);
-			sdma_cntl &= ~TRAP_ENABLE;
-			WREG32(DMA_CNTL + DMA0_REGISTER_OFFSET, sdma_cntl);
+			sdma_cntl = RREG32(mmDMA_CNTL + DMA0_REGISTER_OFFSET);
+			sdma_cntl &= ~DMA_CNTL__TRAP_ENABLE_MASK;
+			WREG32(mmDMA_CNTL + DMA0_REGISTER_OFFSET, sdma_cntl);
 			break;
 		case AMDGPU_IRQ_STATE_ENABLE:
-			sdma_cntl = RREG32(DMA_CNTL + DMA0_REGISTER_OFFSET);
-			sdma_cntl |= TRAP_ENABLE;
-			WREG32(DMA_CNTL + DMA0_REGISTER_OFFSET, sdma_cntl);
+			sdma_cntl = RREG32(mmDMA_CNTL + DMA0_REGISTER_OFFSET);
+			sdma_cntl |= DMA_CNTL__TRAP_ENABLE_MASK;
+			WREG32(mmDMA_CNTL + DMA0_REGISTER_OFFSET, sdma_cntl);
 			break;
 		default:
 			break;
@@ -611,14 +615,14 @@ static int si_dma_set_trap_irq_state(struct amdgpu_device *adev,
 	case AMDGPU_SDMA_IRQ_INSTANCE1:
 		switch (state) {
 		case AMDGPU_IRQ_STATE_DISABLE:
-			sdma_cntl = RREG32(DMA_CNTL + DMA1_REGISTER_OFFSET);
-			sdma_cntl &= ~TRAP_ENABLE;
-			WREG32(DMA_CNTL + DMA1_REGISTER_OFFSET, sdma_cntl);
+			sdma_cntl = RREG32(mmDMA_CNTL + DMA1_REGISTER_OFFSET);
+			sdma_cntl &= ~DMA_CNTL__TRAP_ENABLE_MASK;
+			WREG32(mmDMA_CNTL + DMA1_REGISTER_OFFSET, sdma_cntl);
 			break;
 		case AMDGPU_IRQ_STATE_ENABLE:
-			sdma_cntl = RREG32(DMA_CNTL + DMA1_REGISTER_OFFSET);
-			sdma_cntl |= TRAP_ENABLE;
-			WREG32(DMA_CNTL + DMA1_REGISTER_OFFSET, sdma_cntl);
+			sdma_cntl = RREG32(mmDMA_CNTL + DMA1_REGISTER_OFFSET);
+			sdma_cntl |= DMA_CNTL__TRAP_ENABLE_MASK;
+			WREG32(mmDMA_CNTL + DMA1_REGISTER_OFFSET, sdma_cntl);
 			break;
 		default:
 			break;
@@ -641,13 +645,13 @@ static int si_dma_process_trap_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int si_dma_set_clockgating_state(void *handle,
+static int si_dma_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	u32 orig, data, offset;
 	int i;
 	bool enable;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	enable = (state == AMD_CG_STATE_GATE);
 
@@ -657,11 +661,11 @@ static int si_dma_set_clockgating_state(void *handle,
 				offset = DMA0_REGISTER_OFFSET;
 			else
 				offset = DMA1_REGISTER_OFFSET;
-			orig = data = RREG32(DMA_POWER_CNTL + offset);
-			data &= ~MEM_POWER_OVERRIDE;
+			orig = data = RREG32(mmDMA_POWER_CNTL + offset);
+			data &= ~DMA_POWER_CNTL__MEM_POWER_OVERRIDE_MASK;
 			if (data != orig)
-				WREG32(DMA_POWER_CNTL + offset, data);
-			WREG32(DMA_CLK_CTRL + offset, 0x00000100);
+				WREG32(mmDMA_POWER_CNTL + offset, data);
+			WREG32(mmDMA_CLK_CTRL + offset, 0x00000100);
 		}
 	} else {
 		for (i = 0; i < adev->sdma.num_instances; i++) {
@@ -669,33 +673,33 @@ static int si_dma_set_clockgating_state(void *handle,
 				offset = DMA0_REGISTER_OFFSET;
 			else
 				offset = DMA1_REGISTER_OFFSET;
-			orig = data = RREG32(DMA_POWER_CNTL + offset);
-			data |= MEM_POWER_OVERRIDE;
+			orig = data = RREG32(mmDMA_POWER_CNTL + offset);
+			data |= DMA_POWER_CNTL__MEM_POWER_OVERRIDE_MASK;
 			if (data != orig)
-				WREG32(DMA_POWER_CNTL + offset, data);
+				WREG32(mmDMA_POWER_CNTL + offset, data);
 
-			orig = data = RREG32(DMA_CLK_CTRL + offset);
+			orig = data = RREG32(mmDMA_CLK_CTRL + offset);
 			data = 0xff000000;
 			if (data != orig)
-				WREG32(DMA_CLK_CTRL + offset, data);
+				WREG32(mmDMA_CLK_CTRL + offset, data);
 		}
 	}
 
 	return 0;
 }
 
-static int si_dma_set_powergating_state(void *handle,
+static int si_dma_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	u32 tmp;
 
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	WREG32(DMA_PGFSM_WRITE,  0x00002000);
-	WREG32(DMA_PGFSM_CONFIG, 0x100010ff);
+	WREG32(mmDMA_PGFSM_WRITE,  0x00002000);
+	WREG32(mmDMA_PGFSM_CONFIG, 0x100010ff);
 
 	for (tmp = 0; tmp < 5; tmp++)
-		WREG32(DMA_PGFSM_WRITE, 0);
+		WREG32(mmDMA_PGFSM_WRITE, 0);
 
 	return 0;
 }
@@ -703,7 +707,6 @@ static int si_dma_set_powergating_state(void *handle,
 static const struct amd_ip_funcs si_dma_ip_funcs = {
 	.name = "si_dma",
 	.early_init = si_dma_early_init,
-	.late_init = NULL,
 	.sw_init = si_dma_sw_init,
 	.sw_fini = si_dma_sw_fini,
 	.hw_init = si_dma_hw_init,
@@ -768,7 +771,7 @@ static void si_dma_set_irq_funcs(struct amdgpu_device *adev)
  * @src_offset: src GPU address
  * @dst_offset: dst GPU address
  * @byte_count: number of bytes to xfer
- * @tmz: is this a secure operation
+ * @copy_flags: unused
  *
  * Copy GPU buffers using the DMA engine (VI).
  * Used by the amdgpu ttm implementation to move pages if
@@ -778,7 +781,7 @@ static void si_dma_emit_copy_buffer(struct amdgpu_ib *ib,
 				       uint64_t src_offset,
 				       uint64_t dst_offset,
 				       uint32_t byte_count,
-				       bool tmz)
+				       uint32_t copy_flags)
 {
 	ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_COPY,
 					      1, 0, 0, byte_count);
diff --git a/drivers/gpu/drm/amd/amdgpu/si_enums.h b/drivers/gpu/drm/amd/amdgpu/si_enums.h
index 4e935baa7b91..6da65778292b 100644
--- a/drivers/gpu/drm/amd/amdgpu/si_enums.h
+++ b/drivers/gpu/drm/amd/amdgpu/si_enums.h
@@ -23,123 +23,15 @@
 #ifndef SI_ENUMS_H
 #define SI_ENUMS_H
 
-#define VBLANK_INT_MASK                (1 << 0)
-#define DC_HPDx_INT_EN                 (1 << 16)
-#define VBLANK_ACK                     (1 << 4)
-#define VLINE_ACK                      (1 << 4)
-
-#define CURSOR_WIDTH 64
-#define CURSOR_HEIGHT 64
-
-#define VGA_VSTATUS_CNTL               0xFFFCFFFF
 #define PRIORITY_MARK_MASK             0x7fff
 #define PRIORITY_OFF                   (1 << 16)
 #define PRIORITY_ALWAYS_ON             (1 << 20)
-#define INTERLEAVE_EN                  (1 << 0)
-
-#define LATENCY_WATERMARK_MASK(x)      ((x) << 16)
-#define DC_LB_MEMORY_CONFIG(x)         ((x) << 20)
-#define ICON_DEGAMMA_MODE(x)           (((x) & 0x3) << 8)
-
-#define GRPH_ENDIAN_SWAP(x)            (((x) & 0x3) << 0)
-#define GRPH_ENDIAN_NONE               0
-#define GRPH_ENDIAN_8IN16              1
-#define GRPH_ENDIAN_8IN32              2
-#define GRPH_ENDIAN_8IN64              3
-#define GRPH_RED_CROSSBAR(x)           (((x) & 0x3) << 4)
-#define GRPH_RED_SEL_R                 0
-#define GRPH_RED_SEL_G                 1
-#define GRPH_RED_SEL_B                 2
-#define GRPH_RED_SEL_A                 3
-#define GRPH_GREEN_CROSSBAR(x)         (((x) & 0x3) << 6)
-#define GRPH_GREEN_SEL_G               0
-#define GRPH_GREEN_SEL_B               1
-#define GRPH_GREEN_SEL_A               2
-#define GRPH_GREEN_SEL_R               3
-#define GRPH_BLUE_CROSSBAR(x)          (((x) & 0x3) << 8)
-#define GRPH_BLUE_SEL_B                0
-#define GRPH_BLUE_SEL_A                1
-#define GRPH_BLUE_SEL_R                2
-#define GRPH_BLUE_SEL_G                3
-#define GRPH_ALPHA_CROSSBAR(x)         (((x) & 0x3) << 10)
-#define GRPH_ALPHA_SEL_A               0
-#define GRPH_ALPHA_SEL_R               1
-#define GRPH_ALPHA_SEL_G               2
-#define GRPH_ALPHA_SEL_B               3
-
-#define GRPH_DEPTH(x)                  (((x) & 0x3) << 0)
-#define GRPH_DEPTH_8BPP                0
-#define GRPH_DEPTH_16BPP               1
-#define GRPH_DEPTH_32BPP               2
-
-#define GRPH_FORMAT(x)                 (((x) & 0x7) << 8)
-#define GRPH_FORMAT_INDEXED            0
-#define GRPH_FORMAT_ARGB1555           0
-#define GRPH_FORMAT_ARGB565            1
-#define GRPH_FORMAT_ARGB4444           2
-#define GRPH_FORMAT_AI88               3
-#define GRPH_FORMAT_MONO16             4
-#define GRPH_FORMAT_BGRA5551           5
-#define GRPH_FORMAT_ARGB8888           0
-#define GRPH_FORMAT_ARGB2101010        1
-#define GRPH_FORMAT_32BPP_DIG          2
-#define GRPH_FORMAT_8B_ARGB2101010     3
-#define GRPH_FORMAT_BGRA1010102        4
-#define GRPH_FORMAT_8B_BGRA1010102     5
-#define GRPH_FORMAT_RGB111110          6
-#define GRPH_FORMAT_BGR101111          7
-
-#define GRPH_NUM_BANKS(x)              (((x) & 0x3) << 2)
-#define GRPH_ARRAY_MODE(x)             (((x) & 0x7) << 20)
-#define GRPH_ARRAY_LINEAR_GENERAL      0
-#define GRPH_ARRAY_LINEAR_ALIGNED      1
-#define GRPH_ARRAY_1D_TILED_THIN1      2
-#define GRPH_ARRAY_2D_TILED_THIN1      4
-#define GRPH_TILE_SPLIT(x)             (((x) & 0x7) << 13)
-#define GRPH_BANK_WIDTH(x)             (((x) & 0x3) << 6)
-#define GRPH_BANK_HEIGHT(x)            (((x) & 0x3) << 11)
-#define GRPH_MACRO_TILE_ASPECT(x)      (((x) & 0x3) << 18)
-#define GRPH_ARRAY_MODE(x)             (((x) & 0x7) << 20)
-#define GRPH_PIPE_CONFIG(x)                   (((x) & 0x1f) << 24)
-
-#define CURSOR_EN                      (1 << 0)
-#define CURSOR_MODE(x)                 (((x) & 0x3) << 8)
-#define CURSOR_MONO                    0
-#define CURSOR_24_1                    1
-#define CURSOR_24_8_PRE_MULT           2
-#define CURSOR_24_8_UNPRE_MULT         3
-#define CURSOR_2X_MAGNIFY              (1 << 16)
-#define CURSOR_FORCE_MC_ON             (1 << 20)
-#define CURSOR_URGENT_CONTROL(x)       (((x) & 0x7) << 24)
-#define CURSOR_URGENT_ALWAYS           0
-#define CURSOR_URGENT_1_8              1
-#define CURSOR_URGENT_1_4              2
-#define CURSOR_URGENT_3_8              3
-#define CURSOR_URGENT_1_2              4
-#define CURSOR_UPDATE_PENDING          (1 << 0)
-#define CURSOR_UPDATE_TAKEN            (1 << 1)
-#define CURSOR_UPDATE_LOCK             (1 << 16)
-#define CURSOR_DISABLE_MULTIPLE_UPDATE (1 << 24)
-
-#define SI_CRTC0_REGISTER_OFFSET                0
-#define SI_CRTC1_REGISTER_OFFSET                0x300
-#define SI_CRTC2_REGISTER_OFFSET                0x2600
-#define SI_CRTC3_REGISTER_OFFSET                0x2900
-#define SI_CRTC4_REGISTER_OFFSET                0x2c00
-#define SI_CRTC5_REGISTER_OFFSET                0x2f00
-
-#define DMA0_REGISTER_OFFSET 0x000
-#define DMA1_REGISTER_OFFSET 0x200
-#define ES_AND_GS_AUTO       3
-#define RADEON_PACKET_TYPE3  3
-#define CE_PARTITION_BASE    3
-#define BUF_SWAP_32BIT       (2 << 16)
 
 #define GFX_POWER_STATUS                           (1 << 1)
 #define GFX_CLOCK_STATUS                           (1 << 2)
 #define GFX_LS_STATUS                              (1 << 3)
-#define RLC_BUSY_STATUS                            (1 << 0)
 
+#define RLC_BUSY_STATUS                            (1 << 0)
 #define RLC_PUD(x)                               ((x) << 0)
 #define RLC_PUD_MASK                             (0xff << 0)
 #define RLC_PDD(x)                               ((x) << 8)
@@ -148,144 +40,8 @@
 #define RLC_TTPD_MASK                            (0xff << 16)
 #define RLC_MSD(x)                               ((x) << 24)
 #define RLC_MSD_MASK                             (0xff << 24)
-#define WRITE_DATA_ENGINE_SEL(x) ((x) << 30)
-#define WRITE_DATA_DST_SEL(x) ((x) << 8)
-#define EVENT_TYPE(x) ((x) << 0)
-#define EVENT_INDEX(x) ((x) << 8)
-#define WAIT_REG_MEM_MEM_SPACE(x)               ((x) << 4)
-#define WAIT_REG_MEM_FUNCTION(x)                ((x) << 0)
-#define WAIT_REG_MEM_ENGINE(x)                  ((x) << 8)
 
-#define GFX6_NUM_GFX_RINGS     1
-#define GFX6_NUM_COMPUTE_RINGS 2
 #define RLC_SAVE_AND_RESTORE_STARTING_OFFSET 0x90
 #define RLC_CLEAR_STATE_DESCRIPTOR_OFFSET    0x3D
 
-#define TAHITI_GB_ADDR_CONFIG_GOLDEN        0x12011003
-#define VERDE_GB_ADDR_CONFIG_GOLDEN         0x02010002
-#define HAINAN_GB_ADDR_CONFIG_GOLDEN        0x02011003
-
-#define PACKET3(op, n)  ((RADEON_PACKET_TYPE3 << 30) |                  \
-                         (((op) & 0xFF) << 8) |                         \
-                         ((n) & 0x3FFF) << 16)
-#define PACKET3_COMPUTE(op, n) (PACKET3(op, n) | 1 << 1)
-#define	PACKET3_NOP					0x10
-#define	PACKET3_SET_BASE				0x11
-#define		PACKET3_BASE_INDEX(x)                  ((x) << 0)
-#define	PACKET3_CLEAR_STATE				0x12
-#define	PACKET3_INDEX_BUFFER_SIZE			0x13
-#define	PACKET3_DISPATCH_DIRECT				0x15
-#define	PACKET3_DISPATCH_INDIRECT			0x16
-#define	PACKET3_ALLOC_GDS				0x1B
-#define	PACKET3_WRITE_GDS_RAM				0x1C
-#define	PACKET3_ATOMIC_GDS				0x1D
-#define	PACKET3_ATOMIC					0x1E
-#define	PACKET3_OCCLUSION_QUERY				0x1F
-#define	PACKET3_SET_PREDICATION				0x20
-#define	PACKET3_REG_RMW					0x21
-#define	PACKET3_COND_EXEC				0x22
-#define	PACKET3_PRED_EXEC				0x23
-#define	PACKET3_DRAW_INDIRECT				0x24
-#define	PACKET3_DRAW_INDEX_INDIRECT			0x25
-#define	PACKET3_INDEX_BASE				0x26
-#define	PACKET3_DRAW_INDEX_2				0x27
-#define	PACKET3_CONTEXT_CONTROL				0x28
-#define	PACKET3_INDEX_TYPE				0x2A
-#define	PACKET3_DRAW_INDIRECT_MULTI			0x2C
-#define	PACKET3_DRAW_INDEX_AUTO				0x2D
-#define	PACKET3_DRAW_INDEX_IMMD				0x2E
-#define	PACKET3_NUM_INSTANCES				0x2F
-#define	PACKET3_DRAW_INDEX_MULTI_AUTO			0x30
-#define	PACKET3_INDIRECT_BUFFER_CONST			0x31
-#define	PACKET3_INDIRECT_BUFFER				0x3F
-#define	PACKET3_STRMOUT_BUFFER_UPDATE			0x34
-#define	PACKET3_DRAW_INDEX_OFFSET_2			0x35
-#define	PACKET3_DRAW_INDEX_MULTI_ELEMENT		0x36
-#define	PACKET3_WRITE_DATA				0x37
-#define	PACKET3_DRAW_INDEX_INDIRECT_MULTI		0x38
-#define	PACKET3_MEM_SEMAPHORE				0x39
-#define	PACKET3_MPEG_INDEX				0x3A
-#define	PACKET3_COPY_DW					0x3B
-#define	PACKET3_WAIT_REG_MEM				0x3C
-#define	PACKET3_MEM_WRITE				0x3D
-#define	PACKET3_COPY_DATA				0x40
-#define	PACKET3_CP_DMA					0x41
-#              define PACKET3_CP_DMA_DST_SEL(x)    ((x) << 20)
-#              define PACKET3_CP_DMA_ENGINE(x)     ((x) << 27)
-#              define PACKET3_CP_DMA_SRC_SEL(x)    ((x) << 29)
-#              define PACKET3_CP_DMA_CP_SYNC       (1 << 31)
-#              define PACKET3_CP_DMA_DIS_WC        (1 << 21)
-#              define PACKET3_CP_DMA_CMD_SRC_SWAP(x) ((x) << 22)
-#              define PACKET3_CP_DMA_CMD_DST_SWAP(x) ((x) << 24)
-#              define PACKET3_CP_DMA_CMD_SAS       (1 << 26)
-#              define PACKET3_CP_DMA_CMD_DAS       (1 << 27)
-#              define PACKET3_CP_DMA_CMD_SAIC      (1 << 28)
-#              define PACKET3_CP_DMA_CMD_DAIC      (1 << 29)
-#              define PACKET3_CP_DMA_CMD_RAW_WAIT  (1 << 30)
-#define	PACKET3_PFP_SYNC_ME				0x42
-#define	PACKET3_SURFACE_SYNC				0x43
-#              define PACKET3_DEST_BASE_0_ENA      (1 << 0)
-#              define PACKET3_DEST_BASE_1_ENA      (1 << 1)
-#              define PACKET3_CB0_DEST_BASE_ENA    (1 << 6)
-#              define PACKET3_CB1_DEST_BASE_ENA    (1 << 7)
-#              define PACKET3_CB2_DEST_BASE_ENA    (1 << 8)
-#              define PACKET3_CB3_DEST_BASE_ENA    (1 << 9)
-#              define PACKET3_CB4_DEST_BASE_ENA    (1 << 10)
-#              define PACKET3_CB5_DEST_BASE_ENA    (1 << 11)
-#              define PACKET3_CB6_DEST_BASE_ENA    (1 << 12)
-#              define PACKET3_CB7_DEST_BASE_ENA    (1 << 13)
-#              define PACKET3_DB_DEST_BASE_ENA     (1 << 14)
-#              define PACKET3_DEST_BASE_2_ENA      (1 << 19)
-#              define PACKET3_DEST_BASE_3_ENA      (1 << 21)
-#              define PACKET3_TCL1_ACTION_ENA      (1 << 22)
-#              define PACKET3_TC_ACTION_ENA        (1 << 23)
-#              define PACKET3_CB_ACTION_ENA        (1 << 25)
-#              define PACKET3_DB_ACTION_ENA        (1 << 26)
-#              define PACKET3_SH_KCACHE_ACTION_ENA (1 << 27)
-#              define PACKET3_SH_ICACHE_ACTION_ENA (1 << 29)
-#define	PACKET3_ME_INITIALIZE				0x44
-#define		PACKET3_ME_INITIALIZE_DEVICE_ID(x) ((x) << 16)
-#define	PACKET3_COND_WRITE				0x45
-#define	PACKET3_EVENT_WRITE				0x46
-#define	PACKET3_EVENT_WRITE_EOP				0x47
-#define	PACKET3_EVENT_WRITE_EOS				0x48
-#define	PACKET3_PREAMBLE_CNTL				0x4A
-#              define PACKET3_PREAMBLE_BEGIN_CLEAR_STATE     (2 << 28)
-#              define PACKET3_PREAMBLE_END_CLEAR_STATE       (3 << 28)
-#define	PACKET3_ONE_REG_WRITE				0x57
-#define	PACKET3_LOAD_CONFIG_REG				0x5F
-#define	PACKET3_LOAD_CONTEXT_REG			0x60
-#define	PACKET3_LOAD_SH_REG				0x61
-#define	PACKET3_SET_CONFIG_REG				0x68
-#define		PACKET3_SET_CONFIG_REG_START			0x00002000
-#define		PACKET3_SET_CONFIG_REG_END			0x00002c00
-#define	PACKET3_SET_CONTEXT_REG				0x69
-#define		PACKET3_SET_CONTEXT_REG_START			0x000a000
-#define		PACKET3_SET_CONTEXT_REG_END			0x000a400
-#define	PACKET3_SET_CONTEXT_REG_INDIRECT		0x73
-#define	PACKET3_SET_RESOURCE_INDIRECT			0x74
-#define	PACKET3_SET_SH_REG				0x76
-#define		PACKET3_SET_SH_REG_START			0x00002c00
-#define		PACKET3_SET_SH_REG_END				0x00003000
-#define	PACKET3_SET_SH_REG_OFFSET			0x77
-#define	PACKET3_ME_WRITE				0x7A
-#define	PACKET3_SCRATCH_RAM_WRITE			0x7D
-#define	PACKET3_SCRATCH_RAM_READ			0x7E
-#define	PACKET3_CE_WRITE				0x7F
-#define	PACKET3_LOAD_CONST_RAM				0x80
-#define	PACKET3_WRITE_CONST_RAM				0x81
-#define	PACKET3_WRITE_CONST_RAM_OFFSET			0x82
-#define	PACKET3_DUMP_CONST_RAM				0x83
-#define	PACKET3_INCREMENT_CE_COUNTER			0x84
-#define	PACKET3_INCREMENT_DE_COUNTER			0x85
-#define	PACKET3_WAIT_ON_CE_COUNTER			0x86
-#define	PACKET3_WAIT_ON_DE_COUNTER			0x87
-#define	PACKET3_WAIT_ON_DE_COUNTER_DIFF			0x88
-#define	PACKET3_SET_CE_DE_COUNTERS			0x89
-#define	PACKET3_WAIT_ON_AVAIL_BUFFER			0x8A
-#define	PACKET3_SWITCH_BUFFER				0x8B
-#define PACKET3_SEM_WAIT_ON_SIGNAL    (0x1 << 12)
-#define PACKET3_SEM_SEL_SIGNAL	    (0x6 << 29)
-#define PACKET3_SEM_SEL_WAIT	    (0x7 << 29)
-
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/si_ih.c b/drivers/gpu/drm/amd/amdgpu/si_ih.c
index 9a24f17a5750..66f650f87243 100644
--- a/drivers/gpu/drm/amd/amdgpu/si_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/si_ih.c
@@ -27,6 +27,7 @@
 #include "amdgpu_ih.h"
 #include "sid.h"
 #include "si_ih.h"
+
 #include "oss/oss_1_0_d.h"
 #include "oss/oss_1_0_sh_mask.h"
 
@@ -95,6 +96,9 @@ static int si_ih_irq_init(struct amdgpu_device *adev)
 	pci_set_master(adev->pdev);
 	si_ih_enable_interrupts(adev);
 
+	if (adev->irq.ih_soft.ring_size)
+		adev->irq.ih_soft.enabled = true;
+
 	return 0;
 }
 
@@ -111,6 +115,9 @@ static u32 si_ih_get_wptr(struct amdgpu_device *adev,
 
 	wptr = le32_to_cpu(*ih->wptr_cpu);
 
+	if (ih == &adev->irq.ih_soft)
+		goto out;
+
 	if (wptr & IH_RB_WPTR__RB_OVERFLOW_MASK) {
 		wptr &= ~IH_RB_WPTR__RB_OVERFLOW_MASK;
 		dev_warn(adev->dev, "IH ring buffer overflow (0x%08X, 0x%08X, 0x%08X)\n",
@@ -119,7 +126,15 @@ static u32 si_ih_get_wptr(struct amdgpu_device *adev,
 		tmp = RREG32(IH_RB_CNTL);
 		tmp |= IH_RB_CNTL__WPTR_OVERFLOW_CLEAR_MASK;
 		WREG32(IH_RB_CNTL, tmp);
+
+		/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+		 * can be detected.
+		 */
+		tmp &= ~IH_RB_CNTL__WPTR_OVERFLOW_CLEAR_MASK;
+		WREG32(IH_RB_CNTL, tmp);
 	}
+
+out:
 	return (wptr & ih->ptr_mask);
 }
 
@@ -150,70 +165,68 @@ static void si_ih_set_rptr(struct amdgpu_device *adev,
 	WREG32(IH_RB_RPTR, ih->rptr);
 }
 
-static int si_ih_early_init(void *handle)
+static int si_ih_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	si_ih_set_interrupt_funcs(adev);
 
 	return 0;
 }
 
-static int si_ih_sw_init(void *handle)
+static int si_ih_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 64 * 1024, false);
 	if (r)
 		return r;
 
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, IH_SW_RING_SIZE, true);
+	if (r)
+		return r;
+
 	return amdgpu_irq_init(adev);
 }
 
-static int si_ih_sw_fini(void *handle)
+static int si_ih_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_fini_sw(adev);
 
 	return 0;
 }
 
-static int si_ih_hw_init(void *handle)
+static int si_ih_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return si_ih_irq_init(adev);
 }
 
-static int si_ih_hw_fini(void *handle)
+static int si_ih_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	si_ih_irq_disable(adev);
+	si_ih_irq_disable(ip_block->adev);
 
 	return 0;
 }
 
-static int si_ih_suspend(void *handle)
+static int si_ih_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return si_ih_hw_fini(adev);
+	return si_ih_hw_fini(ip_block);
 }
 
-static int si_ih_resume(void *handle)
+static int si_ih_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return si_ih_hw_init(adev);
+	return si_ih_hw_init(ip_block);
 }
 
-static bool si_ih_is_idle(void *handle)
+static bool si_ih_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	u32 tmp = RREG32(SRBM_STATUS);
+	struct amdgpu_device *adev = ip_block->adev;
+	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (tmp & SRBM_STATUS__IH_BUSY_MASK)
 		return false;
@@ -221,41 +234,41 @@ static bool si_ih_is_idle(void *handle)
 	return true;
 }
 
-static int si_ih_wait_for_idle(void *handle)
+static int si_ih_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		if (si_ih_is_idle(handle))
+		if (si_ih_is_idle(ip_block))
 			return 0;
 		udelay(1);
 	}
 	return -ETIMEDOUT;
 }
 
-static int si_ih_soft_reset(void *handle)
+static int si_ih_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	u32 srbm_soft_reset = 0;
-	u32 tmp = RREG32(SRBM_STATUS);
+	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (tmp & SRBM_STATUS__IH_BUSY_MASK)
 		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_IH_MASK;
 
 	if (srbm_soft_reset) {
-		tmp = RREG32(SRBM_SOFT_RESET);
+		tmp = RREG32(mmSRBM_SOFT_RESET);
 		tmp |= srbm_soft_reset;
-		dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp);
-		WREG32(SRBM_SOFT_RESET, tmp);
-		tmp = RREG32(SRBM_SOFT_RESET);
+		dev_info(adev->dev, "mmSRBM_SOFT_RESET=0x%08X\n", tmp);
+		WREG32(mmSRBM_SOFT_RESET, tmp);
+		tmp = RREG32(mmSRBM_SOFT_RESET);
 
 		udelay(50);
 
 		tmp &= ~srbm_soft_reset;
-		WREG32(SRBM_SOFT_RESET, tmp);
-		tmp = RREG32(SRBM_SOFT_RESET);
+		WREG32(mmSRBM_SOFT_RESET, tmp);
+		tmp = RREG32(mmSRBM_SOFT_RESET);
 
 		udelay(50);
 	}
@@ -263,13 +276,13 @@ static int si_ih_soft_reset(void *handle)
 	return 0;
 }
 
-static int si_ih_set_clockgating_state(void *handle,
+static int si_ih_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int si_ih_set_powergating_state(void *handle,
+static int si_ih_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -278,7 +291,6 @@ static int si_ih_set_powergating_state(void *handle,
 static const struct amd_ip_funcs si_ih_ip_funcs = {
 	.name = "si_ih",
 	.early_init = si_ih_early_init,
-	.late_init = NULL,
 	.sw_init = si_ih_sw_init,
 	.sw_fini = si_ih_sw_fini,
 	.hw_init = si_ih_hw_init,
diff --git a/drivers/gpu/drm/amd/amdgpu/sid.h b/drivers/gpu/drm/amd/amdgpu/sid.h
index 9a39cbfe6db9..561462a8332e 100644
--- a/drivers/gpu/drm/amd/amdgpu/sid.h
+++ b/drivers/gpu/drm/amd/amdgpu/sid.h
@@ -24,47 +24,12 @@
 #ifndef SI_H
 #define SI_H
 
-#define TAHITI_RB_BITMAP_WIDTH_PER_SH  2
-
-#define TAHITI_GB_ADDR_CONFIG_GOLDEN        0x12011003
-#define VERDE_GB_ADDR_CONFIG_GOLDEN         0x12010002
-#define HAINAN_GB_ADDR_CONFIG_GOLDEN        0x02010001
-
-#define SI_MAX_SH_GPRS		 	256
-#define SI_MAX_TEMP_GPRS         	16
-#define SI_MAX_SH_THREADS        	256
-#define SI_MAX_SH_STACK_ENTRIES  	4096
-#define SI_MAX_FRC_EOV_CNT       	16384
-#define SI_MAX_BACKENDS          	8
-#define SI_MAX_BACKENDS_MASK     	0xFF
-#define SI_MAX_BACKENDS_PER_SE_MASK     0x0F
-#define SI_MAX_SIMDS             	12
-#define SI_MAX_SIMDS_MASK        	0x0FFF
-#define SI_MAX_SIMDS_PER_SE_MASK        0x00FF
-#define SI_MAX_PIPES            	8
-#define SI_MAX_PIPES_MASK        	0xFF
-#define SI_MAX_PIPES_PER_SIMD_MASK      0x3F
-#define SI_MAX_LDS_NUM           	0xFFFF
-#define SI_MAX_TCC               	16
-#define SI_MAX_TCC_MASK          	0xFFFF
 #define SI_MAX_CTLACKS_ASSERTION_WAIT   100
 
-/* SMC IND accessor regs */
-#define SMC_IND_INDEX_0                              0x80
-#define SMC_IND_DATA_0                               0x81
-
-#define SMC_IND_ACCESS_CNTL                          0x8A
-#       define AUTO_INCREMENT_IND_0                  (1 << 0)
-#define SMC_MESSAGE_0                                0x8B
-#define SMC_RESP_0                                   0x8C
-
 /* CG IND registers are accessed via SMC indirect space + SMC_CG_IND_START */
 #define SMC_CG_IND_START                    0xc0030000
 #define SMC_CG_IND_END                      0xc0040000
 
-#define	CG_CGTT_LOCAL_0				0x400
-#define	CG_CGTT_LOCAL_1				0x401
-
 /* SMC IND registers */
 #define	SMC_SYSCON_RESET_CNTL				0x80000000
 #       define RST_REG                                  (1 << 0)
@@ -72,9 +37,6 @@
 #       define CK_DISABLE                               (1 << 0)
 #       define CKEN                                     (1 << 24)
 
-#define VGA_HDP_CONTROL  				0xCA
-#define		VGA_MEMORY_DISABLE				(1 << 4)
-
 #define DCCG_DISP_SLOW_SELECT_REG                       0x13F
 #define		DCCG_DISP1_SLOW_SELECT(x)		((x) << 0)
 #define		DCCG_DISP1_SLOW_SELECT_MASK		(7 << 0)
@@ -83,47 +45,6 @@
 #define		DCCG_DISP2_SLOW_SELECT_MASK		(7 << 4)
 #define		DCCG_DISP2_SLOW_SELECT_SHIFT		4
 
-#define	CG_SPLL_FUNC_CNTL				0x180
-#define		SPLL_RESET				(1 << 0)
-#define		SPLL_SLEEP				(1 << 1)
-#define		SPLL_BYPASS_EN				(1 << 3)
-#define		SPLL_REF_DIV(x)				((x) << 4)
-#define		SPLL_REF_DIV_MASK			(0x3f << 4)
-#define		SPLL_PDIV_A(x)				((x) << 20)
-#define		SPLL_PDIV_A_MASK			(0x7f << 20)
-#define		SPLL_PDIV_A_SHIFT			20
-#define	CG_SPLL_FUNC_CNTL_2				0x181
-#define		SCLK_MUX_SEL(x)				((x) << 0)
-#define		SCLK_MUX_SEL_MASK			(0x1ff << 0)
-#define		SPLL_CTLREQ_CHG				(1 << 23)
-#define		SCLK_MUX_UPDATE				(1 << 26)
-#define	CG_SPLL_FUNC_CNTL_3				0x182
-#define		SPLL_FB_DIV(x)				((x) << 0)
-#define		SPLL_FB_DIV_MASK			(0x3ffffff << 0)
-#define		SPLL_FB_DIV_SHIFT			0
-#define		SPLL_DITHEN				(1 << 28)
-#define	CG_SPLL_FUNC_CNTL_4				0x183
-
-#define	SPLL_STATUS					0x185
-#define		SPLL_CHG_STATUS				(1 << 1)
-#define	SPLL_CNTL_MODE					0x186
-#define		SPLL_SW_DIR_CONTROL			(1 << 0)
-#	define SPLL_REFCLK_SEL(x)			((x) << 26)
-#	define SPLL_REFCLK_SEL_MASK			(3 << 26)
-
-#define	CG_SPLL_SPREAD_SPECTRUM				0x188
-#define		SSEN					(1 << 0)
-#define		CLK_S(x)				((x) << 4)
-#define		CLK_S_MASK				(0xfff << 4)
-#define		CLK_S_SHIFT				4
-#define	CG_SPLL_SPREAD_SPECTRUM_2			0x189
-#define		CLK_V(x)				((x) << 0)
-#define		CLK_V_MASK				(0x3ffffff << 0)
-#define		CLK_V_SHIFT				0
-
-#define	CG_SPLL_AUTOSCALE_CNTL				0x18b
-#       define AUTOSCALE_ON_SS_CLEAR                    (1 << 9)
-
 /* discrete uvd clocks */
 #define	CG_UPLL_FUNC_CNTL				0x18d
 #	define UPLL_RESET_MASK				0x00000001
@@ -153,317 +74,13 @@
 #define	CG_UPLL_SPREAD_SPECTRUM				0x194
 #	define SSEN_MASK				0x00000001
 
-#define	MPLL_BYPASSCLK_SEL				0x197
-#	define MPLL_CLKOUT_SEL(x)			((x) << 8)
-#	define MPLL_CLKOUT_SEL_MASK			0xFF00
-
-#define CG_CLKPIN_CNTL                                    0x198
-#       define XTALIN_DIVIDE                              (1 << 1)
-#       define BCLK_AS_XCLK                               (1 << 2)
-#define CG_CLKPIN_CNTL_2                                  0x199
-#       define FORCE_BIF_REFCLK_EN                        (1 << 3)
-#       define MUX_TCLK_TO_XCLK                           (1 << 8)
-
-#define	THM_CLK_CNTL					0x19b
-#	define CMON_CLK_SEL(x)				((x) << 0)
-#	define CMON_CLK_SEL_MASK			0xFF
-#	define TMON_CLK_SEL(x)				((x) << 8)
-#	define TMON_CLK_SEL_MASK			0xFF00
-#define	MISC_CLK_CNTL					0x19c
-#	define DEEP_SLEEP_CLK_SEL(x)			((x) << 0)
-#	define DEEP_SLEEP_CLK_SEL_MASK			0xFF
-#	define ZCLK_SEL(x)				((x) << 8)
-#	define ZCLK_SEL_MASK				0xFF00
-
-#define	CG_THERMAL_CTRL					0x1c0
-#define 	DPM_EVENT_SRC(x)			((x) << 0)
-#define 	DPM_EVENT_SRC_MASK			(7 << 0)
-#define		DIG_THERM_DPM(x)			((x) << 14)
-#define		DIG_THERM_DPM_MASK			0x003FC000
-#define		DIG_THERM_DPM_SHIFT			14
-#define	CG_THERMAL_STATUS				0x1c1
-#define		FDO_PWM_DUTY(x)				((x) << 9)
-#define		FDO_PWM_DUTY_MASK			(0xff << 9)
-#define		FDO_PWM_DUTY_SHIFT			9
-#define	CG_THERMAL_INT					0x1c2
-#define		DIG_THERM_INTH(x)			((x) << 8)
-#define		DIG_THERM_INTH_MASK			0x0000FF00
-#define		DIG_THERM_INTH_SHIFT			8
-#define		DIG_THERM_INTL(x)			((x) << 16)
-#define		DIG_THERM_INTL_MASK			0x00FF0000
-#define		DIG_THERM_INTL_SHIFT			16
-#define 	THERM_INT_MASK_HIGH			(1 << 24)
-#define 	THERM_INT_MASK_LOW			(1 << 25)
-
-#define	CG_MULT_THERMAL_CTRL					0x1c4
-#define		TEMP_SEL(x)					((x) << 20)
-#define		TEMP_SEL_MASK					(0xff << 20)
-#define		TEMP_SEL_SHIFT					20
-#define	CG_MULT_THERMAL_STATUS					0x1c5
-#define		ASIC_MAX_TEMP(x)				((x) << 0)
-#define		ASIC_MAX_TEMP_MASK				0x000001ff
-#define		ASIC_MAX_TEMP_SHIFT				0
-#define		CTF_TEMP(x)					((x) << 9)
-#define		CTF_TEMP_MASK					0x0003fe00
-#define		CTF_TEMP_SHIFT					9
-
-#define	CG_FDO_CTRL0					0x1d5
-#define		FDO_STATIC_DUTY(x)			((x) << 0)
-#define		FDO_STATIC_DUTY_MASK			0x000000FF
-#define		FDO_STATIC_DUTY_SHIFT			0
-#define	CG_FDO_CTRL1					0x1d6
-#define		FMAX_DUTY100(x)				((x) << 0)
-#define		FMAX_DUTY100_MASK			0x000000FF
-#define		FMAX_DUTY100_SHIFT			0
-#define	CG_FDO_CTRL2					0x1d7
-#define		TMIN(x)					((x) << 0)
-#define		TMIN_MASK				0x000000FF
-#define		TMIN_SHIFT				0
-#define		FDO_PWM_MODE(x)				((x) << 11)
-#define		FDO_PWM_MODE_MASK			(7 << 11)
-#define		FDO_PWM_MODE_SHIFT			11
-#define		TACH_PWM_RESP_RATE(x)			((x) << 25)
-#define		TACH_PWM_RESP_RATE_MASK			(0x7f << 25)
-#define		TACH_PWM_RESP_RATE_SHIFT		25
-
-#define CG_TACH_CTRL                                    0x1dc
-#       define EDGE_PER_REV(x)                          ((x) << 0)
-#       define EDGE_PER_REV_MASK                        (0x7 << 0)
-#       define EDGE_PER_REV_SHIFT                       0
-#       define TARGET_PERIOD(x)                         ((x) << 3)
-#       define TARGET_PERIOD_MASK                       0xfffffff8
-#       define TARGET_PERIOD_SHIFT                      3
-#define CG_TACH_STATUS                                  0x1dd
-#       define TACH_PERIOD(x)                           ((x) << 0)
-#       define TACH_PERIOD_MASK                         0xffffffff
-#       define TACH_PERIOD_SHIFT                        0
-
-#define GENERAL_PWRMGT                                  0x1e0
-#       define GLOBAL_PWRMGT_EN                         (1 << 0)
-#       define STATIC_PM_EN                             (1 << 1)
-#       define THERMAL_PROTECTION_DIS                   (1 << 2)
-#       define THERMAL_PROTECTION_TYPE                  (1 << 3)
-#       define SW_SMIO_INDEX(x)                         ((x) << 6)
-#       define SW_SMIO_INDEX_MASK                       (1 << 6)
-#       define SW_SMIO_INDEX_SHIFT                      6
-#       define VOLT_PWRMGT_EN                           (1 << 10)
-#       define DYN_SPREAD_SPECTRUM_EN                   (1 << 23)
-#define CG_TPC                                            0x1e1
-#define SCLK_PWRMGT_CNTL                                  0x1e2
-#       define SCLK_PWRMGT_OFF                            (1 << 0)
-#       define SCLK_LOW_D1                                (1 << 1)
-#       define FIR_RESET                                  (1 << 4)
-#       define FIR_FORCE_TREND_SEL                        (1 << 5)
-#       define FIR_TREND_MODE                             (1 << 6)
-#       define DYN_GFX_CLK_OFF_EN                         (1 << 7)
-#       define GFX_CLK_FORCE_ON                           (1 << 8)
-#       define GFX_CLK_REQUEST_OFF                        (1 << 9)
-#       define GFX_CLK_FORCE_OFF                          (1 << 10)
-#       define GFX_CLK_OFF_ACPI_D1                        (1 << 11)
-#       define GFX_CLK_OFF_ACPI_D2                        (1 << 12)
-#       define GFX_CLK_OFF_ACPI_D3                        (1 << 13)
-#       define DYN_LIGHT_SLEEP_EN                         (1 << 14)
-
-#define TARGET_AND_CURRENT_PROFILE_INDEX                  0x1e6
-#       define CURRENT_STATE_INDEX_MASK                   (0xf << 4)
-#       define CURRENT_STATE_INDEX_SHIFT                  4
-
-#define CG_FTV                                            0x1ef
-
-#define CG_FFCT_0                                         0x1f0
-#       define UTC_0(x)                                   ((x) << 0)
-#       define UTC_0_MASK                                 (0x3ff << 0)
-#       define DTC_0(x)                                   ((x) << 10)
-#       define DTC_0_MASK                                 (0x3ff << 10)
-
-#define CG_BSP                                          0x1ff
-#       define BSP(x)					((x) << 0)
-#       define BSP_MASK					(0xffff << 0)
-#       define BSU(x)					((x) << 16)
-#       define BSU_MASK					(0xf << 16)
-#define CG_AT                                           0x200
-#       define CG_R(x)					((x) << 0)
-#       define CG_R_MASK				(0xffff << 0)
-#       define CG_L(x)					((x) << 16)
-#       define CG_L_MASK				(0xffff << 16)
-
-#define CG_GIT                                          0x201
-#       define CG_GICST(x)                              ((x) << 0)
-#       define CG_GICST_MASK                            (0xffff << 0)
-#       define CG_GIPOT(x)                              ((x) << 16)
-#       define CG_GIPOT_MASK                            (0xffff << 16)
-
-#define CG_SSP                                            0x203
-#       define SST(x)                                     ((x) << 0)
-#       define SST_MASK                                   (0xffff << 0)
-#       define SSTU(x)                                    ((x) << 16)
-#       define SSTU_MASK                                  (0xf << 16)
-
-#define CG_DISPLAY_GAP_CNTL                               0x20a
-#       define DISP1_GAP(x)                               ((x) << 0)
-#       define DISP1_GAP_MASK                             (3 << 0)
-#       define DISP2_GAP(x)                               ((x) << 2)
-#       define DISP2_GAP_MASK                             (3 << 2)
-#       define VBI_TIMER_COUNT(x)                         ((x) << 4)
-#       define VBI_TIMER_COUNT_MASK                       (0x3fff << 4)
-#       define VBI_TIMER_UNIT(x)                          ((x) << 20)
-#       define VBI_TIMER_UNIT_MASK                        (7 << 20)
-#       define DISP1_GAP_MCHG(x)                          ((x) << 24)
-#       define DISP1_GAP_MCHG_MASK                        (3 << 24)
-#       define DISP2_GAP_MCHG(x)                          ((x) << 26)
-#       define DISP2_GAP_MCHG_MASK                        (3 << 26)
-
-#define	CG_ULV_CONTROL					0x21e
-#define	CG_ULV_PARAMETER				0x21f
-
-#define	SMC_SCRATCH0					0x221
-
-#define	CG_CAC_CTRL					0x22e
-#	define CAC_WINDOW(x)				((x) << 0)
-#	define CAC_WINDOW_MASK				0x00ffffff
-
-#define DMIF_ADDR_CONFIG  				0x2F5
-
-#define DMIF_ADDR_CALC  				0x300
-
-#define	PIPE0_DMIF_BUFFER_CONTROL			  0x0328
-#       define DMIF_BUFFERS_ALLOCATED(x)                  ((x) << 0)
-#       define DMIF_BUFFERS_ALLOCATED_COMPLETED           (1 << 4)
-
-#define	SRBM_STATUS				        0x394
-#define		GRBM_RQ_PENDING 			(1 << 5)
-#define		VMC_BUSY 				(1 << 8)
-#define		MCB_BUSY 				(1 << 9)
-#define		MCB_NON_DISPLAY_BUSY 			(1 << 10)
-#define		MCC_BUSY 				(1 << 11)
-#define		MCD_BUSY 				(1 << 12)
-#define		SEM_BUSY 				(1 << 14)
-#define		IH_BUSY 				(1 << 17)
-
-#define	SRBM_SOFT_RESET				        0x398
-#define		SOFT_RESET_BIF				(1 << 1)
-#define		SOFT_RESET_DC				(1 << 5)
-#define		SOFT_RESET_DMA1				(1 << 6)
-#define		SOFT_RESET_GRBM				(1 << 8)
-#define		SOFT_RESET_HDP				(1 << 9)
-#define		SOFT_RESET_IH				(1 << 10)
-#define		SOFT_RESET_MC				(1 << 11)
-#define		SOFT_RESET_ROM				(1 << 14)
-#define		SOFT_RESET_SEM				(1 << 15)
-#define		SOFT_RESET_VMC				(1 << 17)
-#define		SOFT_RESET_DMA				(1 << 20)
-#define		SOFT_RESET_TST				(1 << 21)
-#define		SOFT_RESET_REGBB			(1 << 22)
-#define		SOFT_RESET_ORB				(1 << 23)
-
-#define	CC_SYS_RB_BACKEND_DISABLE			0x3A0
-#define	GC_USER_SYS_RB_BACKEND_DISABLE			0x3A1
-
-#define SRBM_READ_ERROR					0x3A6
-#define SRBM_INT_CNTL					0x3A8
-#define SRBM_INT_ACK					0x3AA
-
-#define	SRBM_STATUS2				        0x3B1
-#define		DMA_BUSY 				(1 << 5)
-#define		DMA1_BUSY 				(1 << 6)
-
-#define VM_L2_CNTL					0x500
-#define		ENABLE_L2_CACHE					(1 << 0)
-#define		ENABLE_L2_FRAGMENT_PROCESSING			(1 << 1)
-#define		L2_CACHE_PTE_ENDIAN_SWAP_MODE(x)		((x) << 2)
-#define		L2_CACHE_PDE_ENDIAN_SWAP_MODE(x)		((x) << 4)
-#define		ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE		(1 << 9)
-#define		ENABLE_L2_PDE0_CACHE_LRU_UPDATE_BY_WRITE	(1 << 10)
-#define		EFFECTIVE_L2_QUEUE_SIZE(x)			(((x) & 7) << 15)
-#define		CONTEXT1_IDENTITY_ACCESS_MODE(x)		(((x) & 3) << 19)
-#define VM_L2_CNTL2					0x501
-#define		INVALIDATE_ALL_L1_TLBS				(1 << 0)
-#define		INVALIDATE_L2_CACHE				(1 << 1)
-#define		INVALIDATE_CACHE_MODE(x)			((x) << 26)
-#define			INVALIDATE_PTE_AND_PDE_CACHES		0
-#define			INVALIDATE_ONLY_PTE_CACHES		1
-#define			INVALIDATE_ONLY_PDE_CACHES		2
-#define VM_L2_CNTL3					0x502
-#define		BANK_SELECT(x)					((x) << 0)
-#define		L2_CACHE_UPDATE_MODE(x)				((x) << 6)
-#define		L2_CACHE_BIGK_FRAGMENT_SIZE(x)			((x) << 15)
-#define		L2_CACHE_BIGK_ASSOCIATIVITY			(1 << 20)
-#define	VM_L2_STATUS					0x503
-#define		L2_BUSY						(1 << 0)
-#define VM_CONTEXT0_CNTL				0x504
-#define		ENABLE_CONTEXT					(1 << 0)
-#define		PAGE_TABLE_DEPTH(x)				(((x) & 3) << 1)
-#define		RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT		(1 << 3)
-#define		RANGE_PROTECTION_FAULT_ENABLE_DEFAULT		(1 << 4)
-#define		DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT	(1 << 6)
-#define		DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT	(1 << 7)
-#define		PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT		(1 << 9)
-#define		PDE0_PROTECTION_FAULT_ENABLE_DEFAULT		(1 << 10)
-#define		VALID_PROTECTION_FAULT_ENABLE_INTERRUPT		(1 << 12)
-#define		VALID_PROTECTION_FAULT_ENABLE_DEFAULT		(1 << 13)
-#define		READ_PROTECTION_FAULT_ENABLE_INTERRUPT		(1 << 15)
-#define		READ_PROTECTION_FAULT_ENABLE_DEFAULT		(1 << 16)
-#define		WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT		(1 << 18)
-#define		WRITE_PROTECTION_FAULT_ENABLE_DEFAULT		(1 << 19)
-#define		PAGE_TABLE_BLOCK_SIZE(x)			(((x) & 0xF) << 24)
-#define VM_CONTEXT1_CNTL				0x505
-#define VM_CONTEXT0_CNTL2				0x50C
-#define VM_CONTEXT1_CNTL2				0x50D
-#define	VM_CONTEXT8_PAGE_TABLE_BASE_ADDR		0x50E
-#define	VM_CONTEXT9_PAGE_TABLE_BASE_ADDR		0x50F
-#define	VM_CONTEXT10_PAGE_TABLE_BASE_ADDR		0x510
-#define	VM_CONTEXT11_PAGE_TABLE_BASE_ADDR		0x511
-#define	VM_CONTEXT12_PAGE_TABLE_BASE_ADDR		0x512
-#define	VM_CONTEXT13_PAGE_TABLE_BASE_ADDR		0x513
-#define	VM_CONTEXT14_PAGE_TABLE_BASE_ADDR		0x514
-#define	VM_CONTEXT15_PAGE_TABLE_BASE_ADDR		0x515
-
-#define	VM_CONTEXT1_PROTECTION_FAULT_ADDR		0x53f
-#define	VM_CONTEXT1_PROTECTION_FAULT_STATUS		0x537
-#define		PROTECTIONS_MASK			(0xf << 0)
-#define		PROTECTIONS_SHIFT			0
-		/* bit 0: range
-		 * bit 1: pde0
-		 * bit 2: valid
-		 * bit 3: read
-		 * bit 4: write
-		 */
-#define		MEMORY_CLIENT_ID_MASK			(0xff << 12)
-#define		MEMORY_CLIENT_ID_SHIFT			12
-#define		MEMORY_CLIENT_RW_MASK			(1 << 24)
-#define		MEMORY_CLIENT_RW_SHIFT			24
-#define		FAULT_VMID_MASK				(0xf << 25)
-#define		FAULT_VMID_SHIFT			25
-
 #define VM_INVALIDATE_REQUEST				0x51E
 #define VM_INVALIDATE_RESPONSE				0x51F
 
-#define VM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR	0x546
-#define VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR	0x547
-
-#define	VM_CONTEXT0_PAGE_TABLE_BASE_ADDR		0x54F
-#define	VM_CONTEXT1_PAGE_TABLE_BASE_ADDR		0x550
-#define	VM_CONTEXT2_PAGE_TABLE_BASE_ADDR		0x551
-#define	VM_CONTEXT3_PAGE_TABLE_BASE_ADDR		0x552
-#define	VM_CONTEXT4_PAGE_TABLE_BASE_ADDR		0x553
-#define	VM_CONTEXT5_PAGE_TABLE_BASE_ADDR		0x554
-#define	VM_CONTEXT6_PAGE_TABLE_BASE_ADDR		0x555
-#define	VM_CONTEXT7_PAGE_TABLE_BASE_ADDR		0x556
-#define	VM_CONTEXT0_PAGE_TABLE_START_ADDR		0x557
-#define	VM_CONTEXT1_PAGE_TABLE_START_ADDR		0x558
-
-#define	VM_CONTEXT0_PAGE_TABLE_END_ADDR			0x55F
-#define	VM_CONTEXT1_PAGE_TABLE_END_ADDR			0x560
-
 #define VM_L2_CG           				0x570
 #define		MC_CG_ENABLE				(1 << 18)
 #define		MC_LS_ENABLE				(1 << 19)
 
-#define MC_SHARED_CHMAP						0x801
-#define		NOOFCHAN_SHIFT					12
-#define		NOOFCHAN_MASK					0x0000f000
-#define MC_SHARED_CHREMAP					0x802
-
 #define	MC_VM_FB_LOCATION				0x809
 #define	MC_VM_AGP_TOP					0x80A
 #define	MC_VM_AGP_BOT					0x80B
@@ -495,21 +112,6 @@
 #define MC_CITF_MISC_WR_CG           			0x993
 #define MC_CITF_MISC_VM_CG           			0x994
 
-#define	MC_ARB_RAMCFG					0x9D8
-#define		NOOFBANK_SHIFT					0
-#define		NOOFBANK_MASK					0x00000003
-#define		NOOFRANK_SHIFT					2
-#define		NOOFRANK_MASK					0x00000004
-#define		NOOFROWS_SHIFT					3
-#define		NOOFROWS_MASK					0x00000038
-#define		NOOFCOLS_SHIFT					6
-#define		NOOFCOLS_MASK					0x000000C0
-#define		CHANSIZE_SHIFT					8
-#define		CHANSIZE_MASK					0x00000100
-#define		CHANSIZE_OVERRIDE				(1 << 11)
-#define		NOOFGROUPS_SHIFT				12
-#define		NOOFGROUPS_MASK					0x00001000
-
 #define	MC_ARB_DRAM_TIMING				0x9DD
 #define	MC_ARB_DRAM_TIMING2				0x9DE
 
@@ -635,20 +237,6 @@
 #define		CLKS(x)					((x) << 0)
 #define		CLKS_MASK				(0xfff << 0)
 
-#define	HDP_HOST_PATH_CNTL				0xB00
-#define 	CLOCK_GATING_DIS			(1 << 23)
-#define	HDP_NONSURFACE_BASE				0xB01
-#define	HDP_NONSURFACE_INFO				0xB02
-#define	HDP_NONSURFACE_SIZE				0xB03
-
-#define HDP_DEBUG0  					0xBCC
-
-#define HDP_ADDR_CONFIG  				0xBD2
-#define HDP_MISC_CNTL					0xBD3
-#define 	HDP_FLUSH_INVALIDATE_CACHE			(1 << 0)
-#define HDP_MEM_POWER_LS				0xBD4
-#define 	HDP_LS_ENABLE				(1 << 0)
-
 #define ATC_MISC_CG           				0xCD4
 
 #define IH_RB_CNTL                                        0xF80
@@ -678,8 +266,6 @@
 #       define MC_WR_CLEAN_CNT(x)                         ((x) << 20)
 #       define MC_VMID(x)                                 ((x) << 25)
 
-#define	CONFIG_MEMSIZE					0x150A
-
 #define INTERRUPT_CNTL                                    0x151A
 #       define IH_DUMMY_RD_OVERRIDE                       (1 << 0)
 #       define IH_DUMMY_RD_EN                             (1 << 1)
@@ -687,734 +273,28 @@
 #       define GEN_IH_INT_EN                              (1 << 8)
 #define INTERRUPT_CNTL2                                   0x151B
 
-#define HDP_MEM_COHERENCY_FLUSH_CNTL			0x1520
-
-#define	BIF_FB_EN						0x1524
-#define		FB_READ_EN					(1 << 0)
-#define		FB_WRITE_EN					(1 << 1)
-
-#define HDP_REG_COHERENCY_FLUSH_CNTL			0x1528
-
-/* DCE6 ELD audio interface */
-#define AZ_F0_CODEC_ENDPOINT_INDEX                       0x1780
-#       define AZ_ENDPOINT_REG_INDEX(x)                  (((x) & 0xff) << 0)
-#       define AZ_ENDPOINT_REG_WRITE_EN                  (1 << 8)
-#define AZ_F0_CODEC_ENDPOINT_DATA                        0x1781
-
-#define AZ_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER          0x25
-#define		SPEAKER_ALLOCATION(x)			(((x) & 0x7f) << 0)
-#define		SPEAKER_ALLOCATION_MASK			(0x7f << 0)
-#define		SPEAKER_ALLOCATION_SHIFT		0
-#define		HDMI_CONNECTION				(1 << 16)
-#define		DP_CONNECTION				(1 << 17)
-
-#define AZ_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0        0x28 /* LPCM */
-#define AZ_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR1        0x29 /* AC3 */
-#define AZ_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR2        0x2A /* MPEG1 */
-#define AZ_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR3        0x2B /* MP3 */
-#define AZ_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR4        0x2C /* MPEG2 */
-#define AZ_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR5        0x2D /* AAC */
-#define AZ_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR6        0x2E /* DTS */
-#define AZ_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR7        0x2F /* ATRAC */
-#define AZ_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR8        0x30 /* one bit audio - leave at 0 (default) */
-#define AZ_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR9        0x31 /* Dolby Digital */
-#define AZ_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR10       0x32 /* DTS-HD */
-#define AZ_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR11       0x33 /* MAT-MLP */
-#define AZ_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR12       0x34 /* DTS */
-#define AZ_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR13       0x35 /* WMA Pro */
-#       define MAX_CHANNELS(x)                            (((x) & 0x7) << 0)
-/* max channels minus one.  7 = 8 channels */
-#       define SUPPORTED_FREQUENCIES(x)                   (((x) & 0xff) << 8)
-#       define DESCRIPTOR_BYTE_2(x)                       (((x) & 0xff) << 16)
-#       define SUPPORTED_FREQUENCIES_STEREO(x)            (((x) & 0xff) << 24) /* LPCM only */
-/* SUPPORTED_FREQUENCIES, SUPPORTED_FREQUENCIES_STEREO
- * bit0 = 32 kHz
- * bit1 = 44.1 kHz
- * bit2 = 48 kHz
- * bit3 = 88.2 kHz
- * bit4 = 96 kHz
- * bit5 = 176.4 kHz
- * bit6 = 192 kHz
- */
-
-#define AZ_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC         0x37
-#       define VIDEO_LIPSYNC(x)                           (((x) & 0xff) << 0)
-#       define AUDIO_LIPSYNC(x)                           (((x) & 0xff) << 8)
-/* VIDEO_LIPSYNC, AUDIO_LIPSYNC
- * 0   = invalid
- * x   = legal delay value
- * 255 = sync not supported
- */
-#define AZ_F0_CODEC_PIN_CONTROL_RESPONSE_HBR             0x38
-#       define HBR_CAPABLE                                (1 << 0) /* enabled by default */
-
-#define AZ_F0_CODEC_PIN_CONTROL_SINK_INFO0               0x3a
-#       define MANUFACTURER_ID(x)                        (((x) & 0xffff) << 0)
-#       define PRODUCT_ID(x)                             (((x) & 0xffff) << 16)
-#define AZ_F0_CODEC_PIN_CONTROL_SINK_INFO1               0x3b
-#       define SINK_DESCRIPTION_LEN(x)                   (((x) & 0xff) << 0)
-#define AZ_F0_CODEC_PIN_CONTROL_SINK_INFO2               0x3c
-#       define PORT_ID0(x)                               (((x) & 0xffffffff) << 0)
-#define AZ_F0_CODEC_PIN_CONTROL_SINK_INFO3               0x3d
-#       define PORT_ID1(x)                               (((x) & 0xffffffff) << 0)
-#define AZ_F0_CODEC_PIN_CONTROL_SINK_INFO4               0x3e
-#       define DESCRIPTION0(x)                           (((x) & 0xff) << 0)
-#       define DESCRIPTION1(x)                           (((x) & 0xff) << 8)
-#       define DESCRIPTION2(x)                           (((x) & 0xff) << 16)
-#       define DESCRIPTION3(x)                           (((x) & 0xff) << 24)
-#define AZ_F0_CODEC_PIN_CONTROL_SINK_INFO5               0x3f
-#       define DESCRIPTION4(x)                           (((x) & 0xff) << 0)
-#       define DESCRIPTION5(x)                           (((x) & 0xff) << 8)
-#       define DESCRIPTION6(x)                           (((x) & 0xff) << 16)
-#       define DESCRIPTION7(x)                           (((x) & 0xff) << 24)
-#define AZ_F0_CODEC_PIN_CONTROL_SINK_INFO6               0x40
-#       define DESCRIPTION8(x)                           (((x) & 0xff) << 0)
-#       define DESCRIPTION9(x)                           (((x) & 0xff) << 8)
-#       define DESCRIPTION10(x)                          (((x) & 0xff) << 16)
-#       define DESCRIPTION11(x)                          (((x) & 0xff) << 24)
-#define AZ_F0_CODEC_PIN_CONTROL_SINK_INFO7               0x41
-#       define DESCRIPTION12(x)                          (((x) & 0xff) << 0)
-#       define DESCRIPTION13(x)                          (((x) & 0xff) << 8)
-#       define DESCRIPTION14(x)                          (((x) & 0xff) << 16)
-#       define DESCRIPTION15(x)                          (((x) & 0xff) << 24)
-#define AZ_F0_CODEC_PIN_CONTROL_SINK_INFO8               0x42
-#       define DESCRIPTION16(x)                          (((x) & 0xff) << 0)
-#       define DESCRIPTION17(x)                          (((x) & 0xff) << 8)
-
-#define AZ_F0_CODEC_PIN_CONTROL_HOT_PLUG_CONTROL         0x54
-#       define AUDIO_ENABLED                             (1 << 31)
-
-#define AZ_F0_CODEC_PIN_CONTROL_RESPONSE_CONFIGURATION_DEFAULT  0x56
-#define		PORT_CONNECTIVITY_MASK				(3 << 30)
-#define		PORT_CONNECTIVITY_SHIFT				30
-
-#define	DC_LB_MEMORY_SPLIT					0x1AC3
-#define		DC_LB_MEMORY_CONFIG(x)				((x) << 20)
-
-#define	PRIORITY_A_CNT						0x1AC6
-#define		PRIORITY_MARK_MASK				0x7fff
-#define		PRIORITY_OFF					(1 << 16)
-#define		PRIORITY_ALWAYS_ON				(1 << 20)
-#define	PRIORITY_B_CNT						0x1AC7
-
-#define	DPG_PIPE_ARBITRATION_CONTROL3				0x1B32
-#       define LATENCY_WATERMARK_MASK(x)			((x) << 16)
-#define	DPG_PIPE_LATENCY_CONTROL				0x1B33
-#       define LATENCY_LOW_WATERMARK(x)				((x) << 0)
-#       define LATENCY_HIGH_WATERMARK(x)			((x) << 16)
-
-/* 0x6bb8, 0x77b8, 0x103b8, 0x10fb8, 0x11bb8, 0x127b8 */
-#define VLINE_STATUS                                    0x1AEE
-#       define VLINE_OCCURRED                           (1 << 0)
-#       define VLINE_ACK                                (1 << 4)
-#       define VLINE_STAT                               (1 << 12)
-#       define VLINE_INTERRUPT                          (1 << 16)
-#       define VLINE_INTERRUPT_TYPE                     (1 << 17)
-/* 0x6bbc, 0x77bc, 0x103bc, 0x10fbc, 0x11bbc, 0x127bc */
-#define VBLANK_STATUS                                   0x1AEF
-#       define VBLANK_OCCURRED                          (1 << 0)
-#       define VBLANK_ACK                               (1 << 4)
-#       define VBLANK_STAT                              (1 << 12)
-#       define VBLANK_INTERRUPT                         (1 << 16)
-#       define VBLANK_INTERRUPT_TYPE                    (1 << 17)
-
-/* 0x6b40, 0x7740, 0x10340, 0x10f40, 0x11b40, 0x12740 */
-#define INT_MASK                                        0x1AD0
-#       define VBLANK_INT_MASK                          (1 << 0)
-#       define VLINE_INT_MASK                           (1 << 4)
-
-#define DISP_INTERRUPT_STATUS                           0x183D
-#       define LB_D1_VLINE_INTERRUPT                    (1 << 2)
-#       define LB_D1_VBLANK_INTERRUPT                   (1 << 3)
-#       define DC_HPD1_INTERRUPT                        (1 << 17)
-#       define DC_HPD1_RX_INTERRUPT                     (1 << 18)
-#       define DACA_AUTODETECT_INTERRUPT                (1 << 22)
-#       define DACB_AUTODETECT_INTERRUPT                (1 << 23)
-#       define DC_I2C_SW_DONE_INTERRUPT                 (1 << 24)
-#       define DC_I2C_HW_DONE_INTERRUPT                 (1 << 25)
-#define DISP_INTERRUPT_STATUS_CONTINUE                  0x183E
-#       define LB_D2_VLINE_INTERRUPT                    (1 << 2)
-#       define LB_D2_VBLANK_INTERRUPT                   (1 << 3)
-#       define DC_HPD2_INTERRUPT                        (1 << 17)
-#       define DC_HPD2_RX_INTERRUPT                     (1 << 18)
-#       define DISP_TIMER_INTERRUPT                     (1 << 24)
-#define DISP_INTERRUPT_STATUS_CONTINUE2                 0x183F
-#       define LB_D3_VLINE_INTERRUPT                    (1 << 2)
-#       define LB_D3_VBLANK_INTERRUPT                   (1 << 3)
-#       define DC_HPD3_INTERRUPT                        (1 << 17)
-#       define DC_HPD3_RX_INTERRUPT                     (1 << 18)
-#define DISP_INTERRUPT_STATUS_CONTINUE3                 0x1840
-#       define LB_D4_VLINE_INTERRUPT                    (1 << 2)
-#       define LB_D4_VBLANK_INTERRUPT                   (1 << 3)
-#       define DC_HPD4_INTERRUPT                        (1 << 17)
-#       define DC_HPD4_RX_INTERRUPT                     (1 << 18)
-#define DISP_INTERRUPT_STATUS_CONTINUE4                 0x1853
-#       define LB_D5_VLINE_INTERRUPT                    (1 << 2)
-#       define LB_D5_VBLANK_INTERRUPT                   (1 << 3)
-#       define DC_HPD5_INTERRUPT                        (1 << 17)
-#       define DC_HPD5_RX_INTERRUPT                     (1 << 18)
-#define DISP_INTERRUPT_STATUS_CONTINUE5                 0x1854
-#       define LB_D6_VLINE_INTERRUPT                    (1 << 2)
-#       define LB_D6_VBLANK_INTERRUPT                   (1 << 3)
-#       define DC_HPD6_INTERRUPT                        (1 << 17)
-#       define DC_HPD6_RX_INTERRUPT                     (1 << 18)
-
-/* 0x6858, 0x7458, 0x10058, 0x10c58, 0x11858, 0x12458 */
-#define GRPH_INT_STATUS                                 0x1A16
-#       define GRPH_PFLIP_INT_OCCURRED                  (1 << 0)
-#       define GRPH_PFLIP_INT_CLEAR                     (1 << 8)
-/* 0x685c, 0x745c, 0x1005c, 0x10c5c, 0x1185c, 0x1245c */
-#define	GRPH_INT_CONTROL			        0x1A17
-#       define GRPH_PFLIP_INT_MASK                      (1 << 0)
-#       define GRPH_PFLIP_INT_TYPE                      (1 << 8)
-
-#define	DAC_AUTODETECT_INT_CONTROL			0x19F2
-
-#define DC_HPD1_INT_STATUS                              0x1807
-#define DC_HPD2_INT_STATUS                              0x180A
-#define DC_HPD3_INT_STATUS                              0x180D
-#define DC_HPD4_INT_STATUS                              0x1810
-#define DC_HPD5_INT_STATUS                              0x1813
-#define DC_HPD6_INT_STATUS                              0x1816
-#       define DC_HPDx_INT_STATUS                       (1 << 0)
-#       define DC_HPDx_SENSE                            (1 << 1)
-#       define DC_HPDx_RX_INT_STATUS                    (1 << 8)
-
-#define DC_HPD1_INT_CONTROL                             0x1808
-#define DC_HPD2_INT_CONTROL                             0x180B
-#define DC_HPD3_INT_CONTROL                             0x180E
-#define DC_HPD4_INT_CONTROL                             0x1811
-#define DC_HPD5_INT_CONTROL                             0x1814
-#define DC_HPD6_INT_CONTROL                             0x1817
-#       define DC_HPDx_INT_ACK                          (1 << 0)
-#       define DC_HPDx_INT_POLARITY                     (1 << 8)
-#       define DC_HPDx_INT_EN                           (1 << 16)
-#       define DC_HPDx_RX_INT_ACK                       (1 << 20)
-#       define DC_HPDx_RX_INT_EN                        (1 << 24)
-
-#define DC_HPD1_CONTROL                                   0x1809
-#define DC_HPD2_CONTROL                                   0x180C
-#define DC_HPD3_CONTROL                                   0x180F
-#define DC_HPD4_CONTROL                                   0x1812
-#define DC_HPD5_CONTROL                                   0x1815
-#define DC_HPD6_CONTROL                                   0x1818
-#       define DC_HPDx_CONNECTION_TIMER(x)                ((x) << 0)
-#       define DC_HPDx_RX_INT_TIMER(x)                    ((x) << 16)
-#       define DC_HPDx_EN                                 (1 << 28)
-
-#define DPG_PIPE_STUTTER_CONTROL                          0x1B35
-#       define STUTTER_ENABLE                             (1 << 0)
-
-/* 0x6e98, 0x7a98, 0x10698, 0x11298, 0x11e98, 0x12a98 */
-#define CRTC_STATUS_FRAME_COUNT                         0x1BA6
-
-/* Audio clocks */
-#define DCCG_AUDIO_DTO_SOURCE                           0x05ac
-#       define DCCG_AUDIO_DTO0_SOURCE_SEL(x) ((x) << 0) /* crtc0 - crtc5 */
-#       define DCCG_AUDIO_DTO_SEL            (1 << 4)   /* 0=dto0 1=dto1 */
-
-#define DCCG_AUDIO_DTO0_PHASE                           0x05b0
-#define DCCG_AUDIO_DTO0_MODULE                          0x05b4
-#define DCCG_AUDIO_DTO1_PHASE                           0x05c0
-#define DCCG_AUDIO_DTO1_MODULE                          0x05c4
-
-#define AFMT_AUDIO_SRC_CONTROL                          0x1c4f
-#define		AFMT_AUDIO_SRC_SELECT(x)		(((x) & 7) << 0)
-/* AFMT_AUDIO_SRC_SELECT
- * 0 = stream0
- * 1 = stream1
- * 2 = stream2
- * 3 = stream3
- * 4 = stream4
- * 5 = stream5
- */
-
-#define	GRBM_CNTL					0x2000
-#define		GRBM_READ_TIMEOUT(x)				((x) << 0)
-
-#define	GRBM_STATUS2					0x2002
-#define		RLC_RQ_PENDING 					(1 << 0)
-#define		RLC_BUSY 					(1 << 8)
-#define		TC_BUSY 					(1 << 9)
-
-#define	GRBM_STATUS					0x2004
-#define		CMDFIFO_AVAIL_MASK				0x0000000F
-#define		RING2_RQ_PENDING				(1 << 4)
-#define		SRBM_RQ_PENDING					(1 << 5)
-#define		RING1_RQ_PENDING				(1 << 6)
-#define		CF_RQ_PENDING					(1 << 7)
-#define		PF_RQ_PENDING					(1 << 8)
-#define		GDS_DMA_RQ_PENDING				(1 << 9)
-#define		GRBM_EE_BUSY					(1 << 10)
-#define		DB_CLEAN					(1 << 12)
-#define		CB_CLEAN					(1 << 13)
-#define		TA_BUSY 					(1 << 14)
-#define		GDS_BUSY 					(1 << 15)
-#define		VGT_BUSY					(1 << 17)
-#define		IA_BUSY_NO_DMA					(1 << 18)
-#define		IA_BUSY						(1 << 19)
-#define		SX_BUSY 					(1 << 20)
-#define		SPI_BUSY					(1 << 22)
-#define		BCI_BUSY					(1 << 23)
-#define		SC_BUSY 					(1 << 24)
-#define		PA_BUSY 					(1 << 25)
-#define		DB_BUSY 					(1 << 26)
-#define		CP_COHERENCY_BUSY      				(1 << 28)
-#define		CP_BUSY 					(1 << 29)
-#define		CB_BUSY 					(1 << 30)
-#define		GUI_ACTIVE					(1 << 31)
-#define	GRBM_STATUS_SE0					0x2005
-#define	GRBM_STATUS_SE1					0x2006
-#define		SE_DB_CLEAN					(1 << 1)
-#define		SE_CB_CLEAN					(1 << 2)
-#define		SE_BCI_BUSY					(1 << 22)
-#define		SE_VGT_BUSY					(1 << 23)
-#define		SE_PA_BUSY					(1 << 24)
-#define		SE_TA_BUSY					(1 << 25)
-#define		SE_SX_BUSY					(1 << 26)
-#define		SE_SPI_BUSY					(1 << 27)
-#define		SE_SC_BUSY					(1 << 29)
-#define		SE_DB_BUSY					(1 << 30)
-#define		SE_CB_BUSY					(1 << 31)
-
-#define	GRBM_SOFT_RESET					0x2008
-#define		SOFT_RESET_CP					(1 << 0)
-#define		SOFT_RESET_CB					(1 << 1)
-#define		SOFT_RESET_RLC					(1 << 2)
-#define		SOFT_RESET_DB					(1 << 3)
-#define		SOFT_RESET_GDS					(1 << 4)
-#define		SOFT_RESET_PA					(1 << 5)
-#define		SOFT_RESET_SC					(1 << 6)
-#define		SOFT_RESET_BCI					(1 << 7)
-#define		SOFT_RESET_SPI					(1 << 8)
-#define		SOFT_RESET_SX					(1 << 10)
-#define		SOFT_RESET_TC					(1 << 11)
-#define		SOFT_RESET_TA					(1 << 12)
-#define		SOFT_RESET_VGT					(1 << 14)
-#define		SOFT_RESET_IA					(1 << 15)
-
-#define GRBM_GFX_INDEX          			0x200B
-#define		INSTANCE_INDEX(x)			((x) << 0)
-#define		SH_INDEX(x)     			((x) << 8)
-#define		SE_INDEX(x)     			((x) << 16)
-#define		SH_BROADCAST_WRITES      		(1 << 29)
-#define		INSTANCE_BROADCAST_WRITES      		(1 << 30)
-#define		SE_BROADCAST_WRITES      		(1 << 31)
-
-#define GRBM_INT_CNTL                                   0x2018
-#       define RDERR_INT_ENABLE                         (1 << 0)
-#       define GUI_IDLE_INT_ENABLE                      (1 << 19)
-
-#define	CP_STRMOUT_CNTL					0x213F
-#define	SCRATCH_REG0					0x2140
-#define	SCRATCH_REG1					0x2141
-#define	SCRATCH_REG2					0x2142
-#define	SCRATCH_REG3					0x2143
-#define	SCRATCH_REG4					0x2144
-#define	SCRATCH_REG5					0x2145
-#define	SCRATCH_REG6					0x2146
-#define	SCRATCH_REG7					0x2147
-
-#define	SCRATCH_UMSK					0x2150
-#define	SCRATCH_ADDR					0x2151
-
-#define	CP_SEM_WAIT_TIMER				0x216F
-
-#define	CP_SEM_INCOMPLETE_TIMER_CNTL			0x2172
-
-#define CP_ME_CNTL					0x21B6
-#define		CP_CE_HALT					(1 << 24)
-#define		CP_PFP_HALT					(1 << 26)
-#define		CP_ME_HALT					(1 << 28)
-
-#define	CP_COHER_CNTL2					0x217A
-
-#define	CP_RB2_RPTR					0x21BE
-#define	CP_RB1_RPTR					0x21BF
-#define	CP_RB0_RPTR					0x21C0
-#define	CP_RB_WPTR_DELAY				0x21C1
-
-#define	CP_QUEUE_THRESHOLDS				0x21D8
-#define		ROQ_IB1_START(x)				((x) << 0)
-#define		ROQ_IB2_START(x)				((x) << 8)
-#define CP_MEQ_THRESHOLDS				0x21D9
-#define		MEQ1_START(x)				((x) << 0)
-#define		MEQ2_START(x)				((x) << 8)
-
-#define	CP_PERFMON_CNTL					0x21FF
-
 #define	VGT_VTX_VECT_EJECT_REG				0x222C
-
-#define	VGT_CACHE_INVALIDATION				0x2231
-#define		CACHE_INVALIDATION(x)				((x) << 0)
-#define			VC_ONLY						0
-#define			TC_ONLY						1
-#define			VC_AND_TC					2
-#define		AUTO_INVLD_EN(x)				((x) << 6)
-#define			NO_AUTO						0
-#define			ES_AUTO						1
-#define			GS_AUTO						2
-#define			ES_AND_GS_AUTO					3
 #define	VGT_ESGS_RING_SIZE				0x2232
 #define	VGT_GSVS_RING_SIZE				0x2233
-
 #define	VGT_GS_VERTEX_REUSE				0x2235
-
 #define	VGT_PRIMITIVE_TYPE				0x2256
 #define	VGT_INDEX_TYPE					0x2257
-
 #define	VGT_NUM_INDICES					0x225C
 #define	VGT_NUM_INSTANCES				0x225D
-
 #define	VGT_TF_RING_SIZE				0x2262
-
 #define	VGT_HS_OFFCHIP_PARAM				0x226C
-
 #define	VGT_TF_MEMORY_BASE				0x226E
 
-#define CC_GC_SHADER_ARRAY_CONFIG			0x226F
-#define		INACTIVE_CUS_MASK			0xFFFF0000
-#define		INACTIVE_CUS_SHIFT			16
-#define GC_USER_SHADER_ARRAY_CONFIG			0x2270
-
-#define	PA_CL_ENHANCE					0x2285
-#define		CLIP_VTX_REORDER_ENA				(1 << 0)
-#define		NUM_CLIP_SEQ(x)					((x) << 1)
-
-#define	PA_SU_LINE_STIPPLE_VALUE			0x2298
-
-#define	PA_SC_LINE_STIPPLE_STATE			0x22C4
-
-#define	PA_SC_FORCE_EOV_MAX_CNTS			0x22C9
-#define		FORCE_EOV_MAX_CLK_CNT(x)			((x) << 0)
-#define		FORCE_EOV_MAX_REZ_CNT(x)			((x) << 16)
-
-#define	PA_SC_FIFO_SIZE					0x22F3
-#define		SC_FRONTEND_PRIM_FIFO_SIZE(x)			((x) << 0)
-#define		SC_BACKEND_PRIM_FIFO_SIZE(x)			((x) << 6)
-#define		SC_HIZ_TILE_FIFO_SIZE(x)			((x) << 15)
-#define		SC_EARLYZ_TILE_FIFO_SIZE(x)			((x) << 23)
-
 #define	PA_SC_ENHANCE					0x22FC
 
-#define	SQ_CONFIG					0x2300
-
-#define	SQC_CACHES					0x2302
-
-#define SQ_POWER_THROTTLE                               0x2396
-#define		MIN_POWER(x)				((x) << 0)
-#define		MIN_POWER_MASK				(0x3fff << 0)
-#define		MIN_POWER_SHIFT				0
-#define		MAX_POWER(x)				((x) << 16)
-#define		MAX_POWER_MASK				(0x3fff << 16)
-#define		MAX_POWER_SHIFT				0
-#define SQ_POWER_THROTTLE2                              0x2397
-#define		MAX_POWER_DELTA(x)			((x) << 0)
-#define		MAX_POWER_DELTA_MASK			(0x3fff << 0)
-#define		MAX_POWER_DELTA_SHIFT			0
-#define		STI_SIZE(x)				((x) << 16)
-#define		STI_SIZE_MASK				(0x3ff << 16)
-#define		STI_SIZE_SHIFT				16
-#define		LTI_RATIO(x)				((x) << 27)
-#define		LTI_RATIO_MASK				(0xf << 27)
-#define		LTI_RATIO_SHIFT				27
-
-#define	SX_DEBUG_1					0x2418
-
-#define	SPI_STATIC_THREAD_MGMT_1			0x2438
-#define	SPI_STATIC_THREAD_MGMT_2			0x2439
-#define	SPI_STATIC_THREAD_MGMT_3			0x243A
-#define	SPI_PS_MAX_WAVE_ID				0x243B
-
-#define	SPI_CONFIG_CNTL					0x2440
-
-#define	SPI_CONFIG_CNTL_1				0x244F
-#define		VTX_DONE_DELAY(x)				((x) << 0)
-#define		INTERP_ONE_PRIM_PER_ROW				(1 << 4)
-
-#define	CGTS_TCC_DISABLE				0x2452
-#define	CGTS_USER_TCC_DISABLE				0x2453
-#define		TCC_DISABLE_MASK				0xFFFF0000
-#define		TCC_DISABLE_SHIFT				16
-#define	CGTS_SM_CTRL_REG				0x2454
-#define		OVERRIDE				(1 << 21)
-#define		LS_OVERRIDE				(1 << 22)
-
-#define	SPI_LB_CU_MASK					0x24D5
-
 #define	TA_CNTL_AUX					0x2542
 
-#define CC_RB_BACKEND_DISABLE				0x263D
-#define		BACKEND_DISABLE(x)     			((x) << 16)
-#define GB_ADDR_CONFIG  				0x263E
-#define		NUM_PIPES(x)				((x) << 0)
-#define		NUM_PIPES_MASK				0x00000007
-#define		NUM_PIPES_SHIFT				0
-#define		PIPE_INTERLEAVE_SIZE(x)			((x) << 4)
-#define		PIPE_INTERLEAVE_SIZE_MASK		0x00000070
-#define		PIPE_INTERLEAVE_SIZE_SHIFT		4
-#define		NUM_SHADER_ENGINES(x)			((x) << 12)
-#define		NUM_SHADER_ENGINES_MASK			0x00003000
-#define		NUM_SHADER_ENGINES_SHIFT		12
-#define		SHADER_ENGINE_TILE_SIZE(x)     		((x) << 16)
-#define		SHADER_ENGINE_TILE_SIZE_MASK		0x00070000
-#define		SHADER_ENGINE_TILE_SIZE_SHIFT		16
-#define		NUM_GPUS(x)     			((x) << 20)
-#define		NUM_GPUS_MASK				0x00700000
-#define		NUM_GPUS_SHIFT				20
-#define		MULTI_GPU_TILE_SIZE(x)     		((x) << 24)
-#define		MULTI_GPU_TILE_SIZE_MASK		0x03000000
-#define		MULTI_GPU_TILE_SIZE_SHIFT		24
-#define		ROW_SIZE(x)             		((x) << 28)
-#define		ROW_SIZE_MASK				0x30000000
-#define		ROW_SIZE_SHIFT				28
-
-#define	GB_TILE_MODE0					0x2644
-#       define MICRO_TILE_MODE(x)				((x) << 0)
-#              define	ADDR_SURF_DISPLAY_MICRO_TILING		0
-#              define	ADDR_SURF_THIN_MICRO_TILING		1
-#              define	ADDR_SURF_DEPTH_MICRO_TILING		2
-#       define ARRAY_MODE(x)					((x) << 2)
-#              define	ARRAY_LINEAR_GENERAL			0
-#              define	ARRAY_LINEAR_ALIGNED			1
-#              define	ARRAY_1D_TILED_THIN1			2
-#              define	ARRAY_2D_TILED_THIN1			4
-#       define PIPE_CONFIG(x)					((x) << 6)
-#              define	ADDR_SURF_P2				0
-#              define	ADDR_SURF_P4_8x16			4
-#              define	ADDR_SURF_P4_16x16			5
-#              define	ADDR_SURF_P4_16x32			6
-#              define	ADDR_SURF_P4_32x32			7
-#              define	ADDR_SURF_P8_16x16_8x16			8
-#              define	ADDR_SURF_P8_16x32_8x16			9
-#              define	ADDR_SURF_P8_32x32_8x16			10
-#              define	ADDR_SURF_P8_16x32_16x16		11
-#              define	ADDR_SURF_P8_32x32_16x16		12
-#              define	ADDR_SURF_P8_32x32_16x32		13
-#              define	ADDR_SURF_P8_32x64_32x32		14
-#       define TILE_SPLIT(x)					((x) << 11)
-#              define	ADDR_SURF_TILE_SPLIT_64B		0
-#              define	ADDR_SURF_TILE_SPLIT_128B		1
-#              define	ADDR_SURF_TILE_SPLIT_256B		2
-#              define	ADDR_SURF_TILE_SPLIT_512B		3
-#              define	ADDR_SURF_TILE_SPLIT_1KB		4
-#              define	ADDR_SURF_TILE_SPLIT_2KB		5
-#              define	ADDR_SURF_TILE_SPLIT_4KB		6
-#       define BANK_WIDTH(x)					((x) << 14)
-#              define	ADDR_SURF_BANK_WIDTH_1			0
-#              define	ADDR_SURF_BANK_WIDTH_2			1
-#              define	ADDR_SURF_BANK_WIDTH_4			2
-#              define	ADDR_SURF_BANK_WIDTH_8			3
-#       define BANK_HEIGHT(x)					((x) << 16)
-#              define	ADDR_SURF_BANK_HEIGHT_1			0
-#              define	ADDR_SURF_BANK_HEIGHT_2			1
-#              define	ADDR_SURF_BANK_HEIGHT_4			2
-#              define	ADDR_SURF_BANK_HEIGHT_8			3
-#       define MACRO_TILE_ASPECT(x)				((x) << 18)
-#              define	ADDR_SURF_MACRO_ASPECT_1		0
-#              define	ADDR_SURF_MACRO_ASPECT_2		1
-#              define	ADDR_SURF_MACRO_ASPECT_4		2
-#              define	ADDR_SURF_MACRO_ASPECT_8		3
-#       define NUM_BANKS(x)					((x) << 20)
-#              define	ADDR_SURF_2_BANK			0
-#              define	ADDR_SURF_4_BANK			1
-#              define	ADDR_SURF_8_BANK			2
-#              define	ADDR_SURF_16_BANK			3
-#define	GB_TILE_MODE1					0x2645
-#define	GB_TILE_MODE2					0x2646
-#define	GB_TILE_MODE3					0x2647
-#define	GB_TILE_MODE4					0x2648
-#define	GB_TILE_MODE5					0x2649
-#define	GB_TILE_MODE6					0x264a
-#define	GB_TILE_MODE7					0x264b
-#define	GB_TILE_MODE8					0x264c
-#define	GB_TILE_MODE9					0x264d
-#define	GB_TILE_MODE10					0x264e
-#define	GB_TILE_MODE11					0x264f
-#define	GB_TILE_MODE12					0x2650
-#define	GB_TILE_MODE13					0x2651
-#define	GB_TILE_MODE14					0x2652
-#define	GB_TILE_MODE15					0x2653
-#define	GB_TILE_MODE16					0x2654
-#define	GB_TILE_MODE17					0x2655
-#define	GB_TILE_MODE18					0x2656
-#define	GB_TILE_MODE19					0x2657
-#define	GB_TILE_MODE20					0x2658
-#define	GB_TILE_MODE21					0x2659
-#define	GB_TILE_MODE22					0x265a
-#define	GB_TILE_MODE23					0x265b
-#define	GB_TILE_MODE24					0x265c
-#define	GB_TILE_MODE25					0x265d
-#define	GB_TILE_MODE26					0x265e
-#define	GB_TILE_MODE27					0x265f
-#define	GB_TILE_MODE28					0x2660
-#define	GB_TILE_MODE29					0x2661
-#define	GB_TILE_MODE30					0x2662
-#define	GB_TILE_MODE31					0x2663
-
-#define	CB_PERFCOUNTER0_SELECT0				0x2688
-#define	CB_PERFCOUNTER0_SELECT1				0x2689
-#define	CB_PERFCOUNTER1_SELECT0				0x268A
-#define	CB_PERFCOUNTER1_SELECT1				0x268B
-#define	CB_PERFCOUNTER2_SELECT0				0x268C
-#define	CB_PERFCOUNTER2_SELECT1				0x268D
-#define	CB_PERFCOUNTER3_SELECT0				0x268E
-#define	CB_PERFCOUNTER3_SELECT1				0x268F
-
-#define	CB_CGTT_SCLK_CTRL				0x2698
-
-#define	GC_USER_RB_BACKEND_DISABLE			0x26DF
-#define		BACKEND_DISABLE_MASK			0x00FF0000
-#define		BACKEND_DISABLE_SHIFT			16
-
-#define	TCP_CHAN_STEER_LO				0x2B03
-#define	TCP_CHAN_STEER_HI				0x2B94
-
-#define	CP_RB0_BASE					0x3040
-#define	CP_RB0_CNTL					0x3041
-#define		RB_BUFSZ(x)					((x) << 0)
-#define		RB_BLKSZ(x)					((x) << 8)
-#define		BUF_SWAP_32BIT					(2 << 16)
-#define		RB_NO_UPDATE					(1 << 27)
-#define		RB_RPTR_WR_ENA					(1 << 31)
-
-#define	CP_RB0_RPTR_ADDR				0x3043
-#define	CP_RB0_RPTR_ADDR_HI				0x3044
-#define	CP_RB0_WPTR					0x3045
-
-#define	CP_PFP_UCODE_ADDR				0x3054
-#define	CP_PFP_UCODE_DATA				0x3055
-#define	CP_ME_RAM_RADDR					0x3056
-#define	CP_ME_RAM_WADDR					0x3057
-#define	CP_ME_RAM_DATA					0x3058
-
-#define	CP_CE_UCODE_ADDR				0x305A
-#define	CP_CE_UCODE_DATA				0x305B
-
-#define	CP_RB1_BASE					0x3060
-#define	CP_RB1_CNTL					0x3061
-#define	CP_RB1_RPTR_ADDR				0x3062
-#define	CP_RB1_RPTR_ADDR_HI				0x3063
-#define	CP_RB1_WPTR					0x3064
-#define	CP_RB2_BASE					0x3065
-#define	CP_RB2_CNTL					0x3066
-#define	CP_RB2_RPTR_ADDR				0x3067
-#define	CP_RB2_RPTR_ADDR_HI				0x3068
-#define	CP_RB2_WPTR					0x3069
-#define CP_INT_CNTL_RING0                               0x306A
-#define CP_INT_CNTL_RING1                               0x306B
-#define CP_INT_CNTL_RING2                               0x306C
-#       define CNTX_BUSY_INT_ENABLE                     (1 << 19)
-#       define CNTX_EMPTY_INT_ENABLE                    (1 << 20)
-#       define WAIT_MEM_SEM_INT_ENABLE                  (1 << 21)
-#       define TIME_STAMP_INT_ENABLE                    (1 << 26)
-#       define CP_RINGID2_INT_ENABLE                    (1 << 29)
-#       define CP_RINGID1_INT_ENABLE                    (1 << 30)
-#       define CP_RINGID0_INT_ENABLE                    (1 << 31)
-#define CP_INT_STATUS_RING0                             0x306D
-#define CP_INT_STATUS_RING1                             0x306E
-#define CP_INT_STATUS_RING2                             0x306F
-#       define WAIT_MEM_SEM_INT_STAT                    (1 << 21)
-#       define TIME_STAMP_INT_STAT                      (1 << 26)
-#       define CP_RINGID2_INT_STAT                      (1 << 29)
-#       define CP_RINGID1_INT_STAT                      (1 << 30)
-#       define CP_RINGID0_INT_STAT                      (1 << 31)
-
-#define	CP_MEM_SLP_CNTL					0x3079
-#       define CP_MEM_LS_EN                             (1 << 0)
-
-#define	CP_DEBUG					0x307F
-
-#define RLC_CNTL                                          0x30C0
-#       define RLC_ENABLE                                 (1 << 0)
-#define RLC_RL_BASE                                       0x30C1
-#define RLC_RL_SIZE                                       0x30C2
-#define RLC_LB_CNTL                                       0x30C3
-#       define LOAD_BALANCE_ENABLE                        (1 << 0)
-#define RLC_SAVE_AND_RESTORE_BASE                         0x30C4
-#define RLC_LB_CNTR_MAX                                   0x30C5
-#define RLC_LB_CNTR_INIT                                  0x30C6
-
-#define RLC_CLEAR_STATE_RESTORE_BASE                      0x30C8
-
-#define RLC_UCODE_ADDR                                    0x30CB
-#define RLC_UCODE_DATA                                    0x30CC
-
-#define RLC_GPU_CLOCK_COUNT_LSB                           0x30CE
-#define RLC_GPU_CLOCK_COUNT_MSB                           0x30CF
-#define RLC_CAPTURE_GPU_CLOCK_COUNT                       0x30D0
-#define RLC_MC_CNTL                                       0x30D1
-#define RLC_UCODE_CNTL                                    0x30D2
-#define RLC_STAT                                          0x30D3
-#       define RLC_BUSY_STATUS                            (1 << 0)
-#       define GFX_POWER_STATUS                           (1 << 1)
-#       define GFX_CLOCK_STATUS                           (1 << 2)
-#       define GFX_LS_STATUS                              (1 << 3)
-
-#define	RLC_PG_CNTL					0x30D7
-#	define GFX_PG_ENABLE				(1 << 0)
-#	define GFX_PG_SRC				(1 << 1)
-
-#define	RLC_CGTT_MGCG_OVERRIDE				0x3100
-#define	RLC_CGCG_CGLS_CTRL				0x3101
-#	define CGCG_EN					(1 << 0)
-#	define CGLS_EN					(1 << 1)
-
-#define	RLC_TTOP_D					0x3105
-#	define RLC_PUD(x)				((x) << 0)
-#	define RLC_PUD_MASK				(0xff << 0)
-#	define RLC_PDD(x)				((x) << 8)
-#	define RLC_PDD_MASK				(0xff << 8)
-#	define RLC_TTPD(x)				((x) << 16)
-#	define RLC_TTPD_MASK				(0xff << 16)
-#	define RLC_MSD(x)				((x) << 24)
-#	define RLC_MSD_MASK				(0xff << 24)
-
-#define RLC_LB_INIT_CU_MASK                               0x3107
-
-#define	RLC_PG_AO_CU_MASK				0x310B
-#define	RLC_MAX_PG_CU					0x310C
-#	define MAX_PU_CU(x)				((x) << 0)
-#	define MAX_PU_CU_MASK				(0xff << 0)
-#define	RLC_AUTO_PG_CTRL				0x310C
-#	define AUTO_PG_EN				(1 << 0)
-#	define GRBM_REG_SGIT(x)				((x) << 3)
-#	define GRBM_REG_SGIT_MASK			(0xffff << 3)
-#	define PG_AFTER_GRBM_REG_ST(x)			((x) << 19)
-#	define PG_AFTER_GRBM_REG_ST_MASK		(0x1fff << 19)
-
-#define RLC_SERDES_WR_MASTER_MASK_0                       0x3115
-#define RLC_SERDES_WR_MASTER_MASK_1                       0x3116
-#define RLC_SERDES_WR_CTRL                                0x3117
-
-#define RLC_SERDES_MASTER_BUSY_0                          0x3119
-#define RLC_SERDES_MASTER_BUSY_1                          0x311A
-
-#define RLC_GCPM_GENERAL_3                                0x311E
-
-#define	DB_RENDER_CONTROL				0xA000
-
-#define DB_DEPTH_INFO                                   0xA00F
-
-#define PA_SC_RASTER_CONFIG                             0xA0D4
-#	define RB_MAP_PKR0(x)				((x) << 0)
-#	define RB_MAP_PKR0_MASK				(0x3 << 0)
-#	define RB_MAP_PKR1(x)				((x) << 2)
-#	define RB_MAP_PKR1_MASK				(0x3 << 2)
-#       define RASTER_CONFIG_RB_MAP_0                   0
-#       define RASTER_CONFIG_RB_MAP_1                   1
-#       define RASTER_CONFIG_RB_MAP_2                   2
-#       define RASTER_CONFIG_RB_MAP_3                   3
+// #define PA_SC_RASTER_CONFIG                             0xA0D4
 #	define RB_XSEL2(x)				((x) << 4)
 #	define RB_XSEL2_MASK				(0x3 << 4)
 #	define RB_XSEL					(1 << 6)
 #	define RB_YSEL					(1 << 7)
 #	define PKR_MAP(x)				((x) << 8)
-#	define PKR_MAP_MASK				(0x3 << 8)
-#       define RASTER_CONFIG_PKR_MAP_0			0
-#       define RASTER_CONFIG_PKR_MAP_1			1
-#       define RASTER_CONFIG_PKR_MAP_2			2
-#       define RASTER_CONFIG_PKR_MAP_3			3
 #	define PKR_XSEL(x)				((x) << 10)
 #	define PKR_XSEL_MASK				(0x3 << 10)
 #	define PKR_YSEL(x)				((x) << 12)
@@ -1426,221 +306,19 @@
 #	define SC_YSEL(x)				((x) << 20)
 #	define SC_YSEL_MASK				(0x3 << 20)
 #	define SE_MAP(x)				((x) << 24)
-#	define SE_MAP_MASK				(0x3 << 24)
-#       define RASTER_CONFIG_SE_MAP_0			0
-#       define RASTER_CONFIG_SE_MAP_1			1
-#       define RASTER_CONFIG_SE_MAP_2			2
-#       define RASTER_CONFIG_SE_MAP_3			3
 #	define SE_XSEL(x)				((x) << 26)
 #	define SE_XSEL_MASK				(0x3 << 26)
 #	define SE_YSEL(x)				((x) << 28)
 #	define SE_YSEL_MASK				(0x3 << 28)
 
-
-#define VGT_EVENT_INITIATOR                             0xA2A4
-#       define SAMPLE_STREAMOUTSTATS1                   (1 << 0)
-#       define SAMPLE_STREAMOUTSTATS2                   (2 << 0)
-#       define SAMPLE_STREAMOUTSTATS3                   (3 << 0)
-#       define CACHE_FLUSH_TS                           (4 << 0)
-#       define CACHE_FLUSH                              (6 << 0)
-#       define CS_PARTIAL_FLUSH                         (7 << 0)
-#       define VGT_STREAMOUT_RESET                      (10 << 0)
-#       define END_OF_PIPE_INCR_DE                      (11 << 0)
-#       define END_OF_PIPE_IB_END                       (12 << 0)
-#       define RST_PIX_CNT                              (13 << 0)
-#       define VS_PARTIAL_FLUSH                         (15 << 0)
-#       define PS_PARTIAL_FLUSH                         (16 << 0)
-#       define CACHE_FLUSH_AND_INV_TS_EVENT             (20 << 0)
-#       define ZPASS_DONE                               (21 << 0)
-#       define CACHE_FLUSH_AND_INV_EVENT                (22 << 0)
-#       define PERFCOUNTER_START                        (23 << 0)
-#       define PERFCOUNTER_STOP                         (24 << 0)
-#       define PIPELINESTAT_START                       (25 << 0)
-#       define PIPELINESTAT_STOP                        (26 << 0)
-#       define PERFCOUNTER_SAMPLE                       (27 << 0)
-#       define SAMPLE_PIPELINESTAT                      (30 << 0)
-#       define SAMPLE_STREAMOUTSTATS                    (32 << 0)
-#       define RESET_VTX_CNT                            (33 << 0)
-#       define VGT_FLUSH                                (36 << 0)
-#       define BOTTOM_OF_PIPE_TS                        (40 << 0)
-#       define DB_CACHE_FLUSH_AND_INV                   (42 << 0)
-#       define FLUSH_AND_INV_DB_DATA_TS                 (43 << 0)
-#       define FLUSH_AND_INV_DB_META                    (44 << 0)
-#       define FLUSH_AND_INV_CB_DATA_TS                 (45 << 0)
-#       define FLUSH_AND_INV_CB_META                    (46 << 0)
-#       define CS_DONE                                  (47 << 0)
-#       define PS_DONE                                  (48 << 0)
-#       define FLUSH_AND_INV_CB_PIXEL_DATA              (49 << 0)
-#       define THREAD_TRACE_START                       (51 << 0)
-#       define THREAD_TRACE_STOP                        (52 << 0)
-#       define THREAD_TRACE_FLUSH                       (54 << 0)
-#       define THREAD_TRACE_FINISH                      (55 << 0)
-
-/* PIF PHY0 registers idx/data 0x8/0xc */
-#define PB0_PIF_CNTL                                      0x10
-#       define LS2_EXIT_TIME(x)                           ((x) << 17)
-#       define LS2_EXIT_TIME_MASK                         (0x7 << 17)
-#       define LS2_EXIT_TIME_SHIFT                        17
-#define PB0_PIF_PAIRING                                   0x11
-#       define MULTI_PIF                                  (1 << 25)
-#define PB0_PIF_PWRDOWN_0                                 0x12
-#       define PLL_POWER_STATE_IN_TXS2_0(x)               ((x) << 7)
-#       define PLL_POWER_STATE_IN_TXS2_0_MASK             (0x7 << 7)
-#       define PLL_POWER_STATE_IN_TXS2_0_SHIFT            7
-#       define PLL_POWER_STATE_IN_OFF_0(x)                ((x) << 10)
-#       define PLL_POWER_STATE_IN_OFF_0_MASK              (0x7 << 10)
-#       define PLL_POWER_STATE_IN_OFF_0_SHIFT             10
-#       define PLL_RAMP_UP_TIME_0(x)                      ((x) << 24)
-#       define PLL_RAMP_UP_TIME_0_MASK                    (0x7 << 24)
-#       define PLL_RAMP_UP_TIME_0_SHIFT                   24
-#define PB0_PIF_PWRDOWN_1                                 0x13
-#       define PLL_POWER_STATE_IN_TXS2_1(x)               ((x) << 7)
-#       define PLL_POWER_STATE_IN_TXS2_1_MASK             (0x7 << 7)
-#       define PLL_POWER_STATE_IN_TXS2_1_SHIFT            7
-#       define PLL_POWER_STATE_IN_OFF_1(x)                ((x) << 10)
-#       define PLL_POWER_STATE_IN_OFF_1_MASK              (0x7 << 10)
-#       define PLL_POWER_STATE_IN_OFF_1_SHIFT             10
-#       define PLL_RAMP_UP_TIME_1(x)                      ((x) << 24)
-#       define PLL_RAMP_UP_TIME_1_MASK                    (0x7 << 24)
-#       define PLL_RAMP_UP_TIME_1_SHIFT                   24
-
-#define PB0_PIF_PWRDOWN_2                                 0x17
-#       define PLL_POWER_STATE_IN_TXS2_2(x)               ((x) << 7)
-#       define PLL_POWER_STATE_IN_TXS2_2_MASK             (0x7 << 7)
-#       define PLL_POWER_STATE_IN_TXS2_2_SHIFT            7
-#       define PLL_POWER_STATE_IN_OFF_2(x)                ((x) << 10)
-#       define PLL_POWER_STATE_IN_OFF_2_MASK              (0x7 << 10)
-#       define PLL_POWER_STATE_IN_OFF_2_SHIFT             10
-#       define PLL_RAMP_UP_TIME_2(x)                      ((x) << 24)
-#       define PLL_RAMP_UP_TIME_2_MASK                    (0x7 << 24)
-#       define PLL_RAMP_UP_TIME_2_SHIFT                   24
-#define PB0_PIF_PWRDOWN_3                                 0x18
-#       define PLL_POWER_STATE_IN_TXS2_3(x)               ((x) << 7)
-#       define PLL_POWER_STATE_IN_TXS2_3_MASK             (0x7 << 7)
-#       define PLL_POWER_STATE_IN_TXS2_3_SHIFT            7
-#       define PLL_POWER_STATE_IN_OFF_3(x)                ((x) << 10)
-#       define PLL_POWER_STATE_IN_OFF_3_MASK              (0x7 << 10)
-#       define PLL_POWER_STATE_IN_OFF_3_SHIFT             10
-#       define PLL_RAMP_UP_TIME_3(x)                      ((x) << 24)
-#       define PLL_RAMP_UP_TIME_3_MASK                    (0x7 << 24)
-#       define PLL_RAMP_UP_TIME_3_SHIFT                   24
-/* PIF PHY1 registers idx/data 0x10/0x14 */
-#define PB1_PIF_CNTL                                      0x10
-#define PB1_PIF_PAIRING                                   0x11
-#define PB1_PIF_PWRDOWN_0                                 0x12
-#define PB1_PIF_PWRDOWN_1                                 0x13
-
-#define PB1_PIF_PWRDOWN_2                                 0x17
-#define PB1_PIF_PWRDOWN_3                                 0x18
-/* PCIE registers idx/data 0x30/0x34 */
-#define PCIE_CNTL2                                        0x1c /* PCIE */
-#       define SLV_MEM_LS_EN                              (1 << 16)
-#       define SLV_MEM_AGGRESSIVE_LS_EN                   (1 << 17)
-#       define MST_MEM_LS_EN                              (1 << 18)
-#       define REPLAY_MEM_LS_EN                           (1 << 19)
-#define PCIE_LC_STATUS1                                   0x28 /* PCIE */
-#       define LC_REVERSE_RCVR                            (1 << 0)
-#       define LC_REVERSE_XMIT                            (1 << 1)
-#       define LC_OPERATING_LINK_WIDTH_MASK               (0x7 << 2)
-#       define LC_OPERATING_LINK_WIDTH_SHIFT              2
-#       define LC_DETECTED_LINK_WIDTH_MASK                (0x7 << 5)
-#       define LC_DETECTED_LINK_WIDTH_SHIFT               5
-
-#define PCIE_P_CNTL                                       0x40 /* PCIE */
-#       define P_IGNORE_EDB_ERR                           (1 << 6)
-
 /* PCIE PORT registers idx/data 0x38/0x3c */
-#define PCIE_LC_CNTL                                      0xa0
-#       define LC_L0S_INACTIVITY(x)                       ((x) << 8)
-#       define LC_L0S_INACTIVITY_MASK                     (0xf << 8)
-#       define LC_L0S_INACTIVITY_SHIFT                    8
-#       define LC_L1_INACTIVITY(x)                        ((x) << 12)
-#       define LC_L1_INACTIVITY_MASK                      (0xf << 12)
-#       define LC_L1_INACTIVITY_SHIFT                     12
-#       define LC_PMI_TO_L1_DIS                           (1 << 16)
-#       define LC_ASPM_TO_L1_DIS                          (1 << 24)
-#define PCIE_LC_LINK_WIDTH_CNTL                           0xa2 /* PCIE_P */
-#       define LC_LINK_WIDTH_SHIFT                        0
-#       define LC_LINK_WIDTH_MASK                         0x7
+// #define PCIE_LC_LINK_WIDTH_CNTL                           0xa2 /* PCIE_P */
 #       define LC_LINK_WIDTH_X0                           0
 #       define LC_LINK_WIDTH_X1                           1
 #       define LC_LINK_WIDTH_X2                           2
 #       define LC_LINK_WIDTH_X4                           3
 #       define LC_LINK_WIDTH_X8                           4
 #       define LC_LINK_WIDTH_X16                          6
-#       define LC_LINK_WIDTH_RD_SHIFT                     4
-#       define LC_LINK_WIDTH_RD_MASK                      0x70
-#       define LC_RECONFIG_ARC_MISSING_ESCAPE             (1 << 7)
-#       define LC_RECONFIG_NOW                            (1 << 8)
-#       define LC_RENEGOTIATION_SUPPORT                   (1 << 9)
-#       define LC_RENEGOTIATE_EN                          (1 << 10)
-#       define LC_SHORT_RECONFIG_EN                       (1 << 11)
-#       define LC_UPCONFIGURE_SUPPORT                     (1 << 12)
-#       define LC_UPCONFIGURE_DIS                         (1 << 13)
-#       define LC_DYN_LANES_PWR_STATE(x)                  ((x) << 21)
-#       define LC_DYN_LANES_PWR_STATE_MASK                (0x3 << 21)
-#       define LC_DYN_LANES_PWR_STATE_SHIFT               21
-#define PCIE_LC_N_FTS_CNTL                                0xa3 /* PCIE_P */
-#       define LC_XMIT_N_FTS(x)                           ((x) << 0)
-#       define LC_XMIT_N_FTS_MASK                         (0xff << 0)
-#       define LC_XMIT_N_FTS_SHIFT                        0
-#       define LC_XMIT_N_FTS_OVERRIDE_EN                  (1 << 8)
-#       define LC_N_FTS_MASK                              (0xff << 24)
-#define PCIE_LC_SPEED_CNTL                                0xa4 /* PCIE_P */
-#       define LC_GEN2_EN_STRAP                           (1 << 0)
-#       define LC_GEN3_EN_STRAP                           (1 << 1)
-#       define LC_TARGET_LINK_SPEED_OVERRIDE_EN           (1 << 2)
-#       define LC_TARGET_LINK_SPEED_OVERRIDE_MASK         (0x3 << 3)
-#       define LC_TARGET_LINK_SPEED_OVERRIDE_SHIFT        3
-#       define LC_FORCE_EN_SW_SPEED_CHANGE                (1 << 5)
-#       define LC_FORCE_DIS_SW_SPEED_CHANGE               (1 << 6)
-#       define LC_FORCE_EN_HW_SPEED_CHANGE                (1 << 7)
-#       define LC_FORCE_DIS_HW_SPEED_CHANGE               (1 << 8)
-#       define LC_INITIATE_LINK_SPEED_CHANGE              (1 << 9)
-#       define LC_SPEED_CHANGE_ATTEMPTS_ALLOWED_MASK      (0x3 << 10)
-#       define LC_SPEED_CHANGE_ATTEMPTS_ALLOWED_SHIFT     10
-#       define LC_CURRENT_DATA_RATE_MASK                  (0x3 << 13) /* 0/1/2 = gen1/2/3 */
-#       define LC_CURRENT_DATA_RATE_SHIFT                 13
-#       define LC_CLR_FAILED_SPD_CHANGE_CNT               (1 << 16)
-#       define LC_OTHER_SIDE_EVER_SENT_GEN2               (1 << 18)
-#       define LC_OTHER_SIDE_SUPPORTS_GEN2                (1 << 19)
-#       define LC_OTHER_SIDE_EVER_SENT_GEN3               (1 << 20)
-#       define LC_OTHER_SIDE_SUPPORTS_GEN3                (1 << 21)
-
-#define PCIE_LC_CNTL2                                     0xb1
-#       define LC_ALLOW_PDWN_IN_L1                        (1 << 17)
-#       define LC_ALLOW_PDWN_IN_L23                       (1 << 18)
-
-#define PCIE_LC_CNTL3                                     0xb5 /* PCIE_P */
-#       define LC_GO_TO_RECOVERY                          (1 << 30)
-#define PCIE_LC_CNTL4                                     0xb6 /* PCIE_P */
-#       define LC_REDO_EQ                                 (1 << 5)
-#       define LC_SET_QUIESCE                             (1 << 13)
-
-/*
- * UVD
- */
-#define UVD_UDEC_ADDR_CONFIG				0x3bd3
-#define UVD_UDEC_DB_ADDR_CONFIG				0x3bd4
-#define UVD_UDEC_DBW_ADDR_CONFIG			0x3bd5
-#define UVD_RBC_RB_RPTR					0x3da4
-#define UVD_RBC_RB_WPTR					0x3da5
-#define UVD_STATUS					0x3daf
-
-#define	UVD_CGC_CTRL					0x3dc2
-#	define DCM					(1 << 0)
-#	define CG_DT(x)					((x) << 2)
-#	define CG_DT_MASK				(0xf << 2)
-#	define CLK_OD(x)				((x) << 6)
-#	define CLK_OD_MASK				(0x1f << 6)
-
- /* UVD CTX indirect */
-#define	UVD_CGC_MEM_CTRL				0xC0
-#define	UVD_CGC_CTRL2					0xC1
-#	define DYN_OR_EN				(1 << 0)
-#	define DYN_RR_EN				(1 << 1)
-#	define G_DIV_ID(x)				((x) << 2)
-#	define G_DIV_ID_MASK				(0x7 << 2)
 
 /*
  * PM4
@@ -1874,45 +552,7 @@
 /* ASYNC DMA - first instance at 0xd000, second at 0xd800 */
 #define DMA0_REGISTER_OFFSET                              0x0 /* not a register */
 #define DMA1_REGISTER_OFFSET                              0x200 /* not a register */
-
-#define DMA_RB_CNTL                                       0x3400
-#       define DMA_RB_ENABLE                              (1 << 0)
-#       define DMA_RB_SIZE(x)                             ((x) << 1) /* log2 */
-#       define DMA_RB_SWAP_ENABLE                         (1 << 9) /* 8IN32 */
-#       define DMA_RPTR_WRITEBACK_ENABLE                  (1 << 12)
-#       define DMA_RPTR_WRITEBACK_SWAP_ENABLE             (1 << 13)  /* 8IN32 */
-#       define DMA_RPTR_WRITEBACK_TIMER(x)                ((x) << 16) /* log2 */
-#define DMA_RB_BASE                                       0x3401
-#define DMA_RB_RPTR                                       0x3402
-#define DMA_RB_WPTR                                       0x3403
-
-#define DMA_RB_RPTR_ADDR_HI                               0x3407
-#define DMA_RB_RPTR_ADDR_LO                               0x3408
-
-#define DMA_IB_CNTL                                       0x3409
-#       define DMA_IB_ENABLE                              (1 << 0)
-#       define DMA_IB_SWAP_ENABLE                         (1 << 4)
-#       define CMD_VMID_FORCE                             (1 << 31)
-#define DMA_IB_RPTR                                       0x340a
-#define DMA_CNTL                                          0x340b
-#       define TRAP_ENABLE                                (1 << 0)
-#       define SEM_INCOMPLETE_INT_ENABLE                  (1 << 1)
-#       define SEM_WAIT_INT_ENABLE                        (1 << 2)
-#       define DATA_SWAP_ENABLE                           (1 << 3)
-#       define FENCE_SWAP_ENABLE                          (1 << 4)
-#       define CTXEMPTY_INT_ENABLE                        (1 << 28)
-#define DMA_STATUS_REG                                    0x340d
-#       define DMA_IDLE                                   (1 << 0)
-#define DMA_TILING_CONFIG  				  0x342e
-
-#define	DMA_POWER_CNTL					0x342f
-#       define MEM_POWER_OVERRIDE                       (1 << 8)
-#define	DMA_CLK_CTRL					0x3430
-
-#define	DMA_PG						0x3435
-#	define PG_CNTL_ENABLE				(1 << 0)
-#define	DMA_PGFSM_CONFIG				0x3436
-#define	DMA_PGFSM_WRITE					0x3437
+#define SDMA_MAX_INSTANCE 2
 
 #define DMA_PACKET(cmd, b, t, s, n)	((((cmd) & 0xF) << 28) |	\
 					 (((b) & 0x1) << 26) |		\
@@ -1941,45 +581,7 @@
 #define	DMA_PACKET_POLL_REG_MEM				  0xe
 #define	DMA_PACKET_NOP					  0xf
 
-#define VCE_STATUS					0x20004
-#define VCE_VCPU_CNTL					0x20014
-#define		VCE_CLK_EN				(1 << 0)
-#define VCE_VCPU_CACHE_OFFSET0				0x20024
-#define VCE_VCPU_CACHE_SIZE0				0x20028
-#define VCE_VCPU_CACHE_OFFSET1				0x2002c
-#define VCE_VCPU_CACHE_SIZE1				0x20030
-#define VCE_VCPU_CACHE_OFFSET2				0x20034
-#define VCE_VCPU_CACHE_SIZE2				0x20038
-#define VCE_SOFT_RESET					0x20120
-#define 	VCE_ECPU_SOFT_RESET			(1 << 0)
-#define 	VCE_FME_SOFT_RESET			(1 << 2)
-#define VCE_RB_BASE_LO2					0x2016c
-#define VCE_RB_BASE_HI2					0x20170
-#define VCE_RB_SIZE2					0x20174
-#define VCE_RB_RPTR2					0x20178
-#define VCE_RB_WPTR2					0x2017c
-#define VCE_RB_BASE_LO					0x20180
-#define VCE_RB_BASE_HI					0x20184
-#define VCE_RB_SIZE					0x20188
-#define VCE_RB_RPTR					0x2018c
-#define VCE_RB_WPTR					0x20190
-#define VCE_CLOCK_GATING_A				0x202f8
-#define VCE_CLOCK_GATING_B				0x202fc
-#define VCE_UENC_CLOCK_GATING				0x205bc
-#define VCE_UENC_REG_CLOCK_GATING			0x205c0
-#define VCE_FW_REG_STATUS				0x20e10
-#	define VCE_FW_REG_STATUS_BUSY			(1 << 0)
-#	define VCE_FW_REG_STATUS_PASS			(1 << 3)
-#	define VCE_FW_REG_STATUS_DONE			(1 << 11)
-#define VCE_LMI_FW_START_KEYSEL				0x20e18
-#define VCE_LMI_FW_PERIODIC_CTRL			0x20e20
-#define VCE_LMI_CTRL2					0x20e74
-#define VCE_LMI_CTRL					0x20e98
-#define VCE_LMI_VM_CTRL					0x20ea0
-#define VCE_LMI_SWAP_CNTL				0x20eb4
-#define VCE_LMI_SWAP_CNTL1				0x20eb8
-#define VCE_LMI_CACHE_CTRL				0x20ef4
-
+/* VCE */
 #define VCE_CMD_NO_OP					0x00000000
 #define VCE_CMD_END					0x00000001
 #define VCE_CMD_IB					0x00000002
@@ -1988,434 +590,146 @@
 #define VCE_CMD_IB_AUTO					0x00000005
 #define VCE_CMD_SEMAPHORE				0x00000006
 
-
 //#dce stupp
 /* display controller offsets used for crtc/cur/lut/grph/viewport/etc. */
-#define SI_CRTC0_REGISTER_OFFSET                0 //(0x6df0 - 0x6df0)/4
-#define SI_CRTC1_REGISTER_OFFSET                0x300 //(0x79f0 - 0x6df0)/4
-#define SI_CRTC2_REGISTER_OFFSET                0x2600 //(0x105f0 - 0x6df0)/4
-#define SI_CRTC3_REGISTER_OFFSET                0x2900 //(0x111f0 - 0x6df0)/4
-#define SI_CRTC4_REGISTER_OFFSET                0x2c00 //(0x11df0 - 0x6df0)/4
-#define SI_CRTC5_REGISTER_OFFSET                0x2f00 //(0x129f0 - 0x6df0)/4
+#define CRTC0_REGISTER_OFFSET                 (0x1b7c - 0x1b7c) //(0x6df0 - 0x6df0)/4
+#define CRTC1_REGISTER_OFFSET                 (0x1e7c - 0x1b7c) //(0x79f0 - 0x6df0)/4
+#define CRTC2_REGISTER_OFFSET                 (0x417c - 0x1b7c) //(0x105f0 - 0x6df0)/4
+#define CRTC3_REGISTER_OFFSET                 (0x447c - 0x1b7c) //(0x111f0 - 0x6df0)/4
+#define CRTC4_REGISTER_OFFSET                 (0x477c - 0x1b7c) //(0x11df0 - 0x6df0)/4
+#define CRTC5_REGISTER_OFFSET                 (0x4a7c - 0x1b7c) //(0x129f0 - 0x6df0)/4
+
+/* hpd instance offsets */
+#define HPD0_REGISTER_OFFSET                 (0x1807 - 0x1807)
+#define HPD1_REGISTER_OFFSET                 (0x180a - 0x1807)
+#define HPD2_REGISTER_OFFSET                 (0x180d - 0x1807)
+#define HPD3_REGISTER_OFFSET                 (0x1810 - 0x1807)
+#define HPD4_REGISTER_OFFSET                 (0x1813 - 0x1807)
+#define HPD5_REGISTER_OFFSET                 (0x1816 - 0x1807)
+
+/* audio endpt instance offsets */
+#define AUD0_REGISTER_OFFSET                 (0x1780 - 0x1780)
+#define AUD1_REGISTER_OFFSET                 (0x1786 - 0x1780)
+#define AUD2_REGISTER_OFFSET                 (0x178c - 0x1780)
+#define AUD3_REGISTER_OFFSET                 (0x1792 - 0x1780)
+#define AUD4_REGISTER_OFFSET                 (0x1798 - 0x1780)
+#define AUD5_REGISTER_OFFSET                 (0x179d - 0x1780)
+#define AUD6_REGISTER_OFFSET                 (0x17a4 - 0x1780)
 
 #define CURSOR_WIDTH 64
 #define CURSOR_HEIGHT 64
-#define AMDGPU_MM_INDEX		        0x0000
-#define AMDGPU_MM_DATA		        0x0001
-
-#define VERDE_NUM_CRTC 6
-#define	BLACKOUT_MODE_MASK			0x00000007
-#define	VGA_RENDER_CONTROL			0xC0
-#define R_000300_VGA_RENDER_CONTROL             0xC0
-#define C_000300_VGA_VSTATUS_CNTL               0xFFFCFFFF
-#define EVERGREEN_CRTC_STATUS                   0x1BA3
-#define EVERGREEN_CRTC_V_BLANK                  (1 << 0)
-#define EVERGREEN_CRTC_STATUS_POSITION          0x1BA4
-/* CRTC blocks at 0x6df0, 0x79f0, 0x105f0, 0x111f0, 0x11df0, 0x129f0 */
-#define EVERGREEN_CRTC_V_BLANK_START_END                0x1b8d
-#define EVERGREEN_CRTC_CONTROL                          0x1b9c
-#define EVERGREEN_CRTC_MASTER_EN                 (1 << 0)
-#define EVERGREEN_CRTC_DISP_READ_REQUEST_DISABLE (1 << 24)
-#define EVERGREEN_CRTC_BLANK_CONTROL                    0x1b9d
-#define EVERGREEN_CRTC_BLANK_DATA_EN             (1 << 8)
-#define EVERGREEN_CRTC_V_BLANK                   (1 << 0)
-#define EVERGREEN_CRTC_STATUS_HV_COUNT                  0x1ba8
-#define EVERGREEN_CRTC_UPDATE_LOCK                      0x1bb5
-#define EVERGREEN_MASTER_UPDATE_LOCK                    0x1bbd
-#define EVERGREEN_MASTER_UPDATE_MODE                    0x1bbe
-#define EVERGREEN_GRPH_UPDATE_LOCK               (1 << 16)
-#define EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH     0x1a07
-#define EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH   0x1a08
-#define EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS          0x1a04
-#define EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS        0x1a05
-#define EVERGREEN_GRPH_UPDATE                           0x1a11
-#define EVERGREEN_VGA_MEMORY_BASE_ADDRESS               0xc4
-#define EVERGREEN_VGA_MEMORY_BASE_ADDRESS_HIGH          0xc9
-#define EVERGREEN_GRPH_SURFACE_UPDATE_PENDING    (1 << 2)
-
-#define EVERGREEN_DATA_FORMAT                           0x1ac0
-#       define EVERGREEN_INTERLEAVE_EN                  (1 << 0)
-
-#define MC_SHARED_CHMAP__NOOFCHAN_MASK 0xf000
-#define MC_SHARED_CHMAP__NOOFCHAN__SHIFT 0xc
-
-#define R600_D1GRPH_ARRAY_MODE_LINEAR_GENERAL            (0 << 20)
-#define R600_D1GRPH_ARRAY_MODE_LINEAR_ALIGNED            (1 << 20)
-#define R600_D1GRPH_ARRAY_MODE_1D_TILED_THIN1            (2 << 20)
-#define R600_D1GRPH_ARRAY_MODE_2D_TILED_THIN1            (4 << 20)
-
-#define R700_D1GRPH_PRIMARY_SURFACE_ADDRESS_HIGH                0x1a45
-#define R700_D2GRPH_PRIMARY_SURFACE_ADDRESS_HIGH                0x1845
-
-#define R700_D2GRPH_SECONDARY_SURFACE_ADDRESS_HIGH              0x1847
-#define R700_D1GRPH_SECONDARY_SURFACE_ADDRESS_HIGH              0x1a47
-
-#define DISP_INTERRUPT_STATUS__LB_D1_VBLANK_INTERRUPT_MASK 0x8
-#define DISP_INTERRUPT_STATUS_CONTINUE__LB_D2_VBLANK_INTERRUPT_MASK 0x8
-#define DISP_INTERRUPT_STATUS_CONTINUE2__LB_D3_VBLANK_INTERRUPT_MASK 0x8
-#define DISP_INTERRUPT_STATUS_CONTINUE3__LB_D4_VBLANK_INTERRUPT_MASK 0x8
-#define DISP_INTERRUPT_STATUS_CONTINUE4__LB_D5_VBLANK_INTERRUPT_MASK 0x8
-#define DISP_INTERRUPT_STATUS_CONTINUE5__LB_D6_VBLANK_INTERRUPT_MASK 0x8
-
-#define DISP_INTERRUPT_STATUS__LB_D1_VLINE_INTERRUPT_MASK 0x4
-#define DISP_INTERRUPT_STATUS_CONTINUE__LB_D2_VLINE_INTERRUPT_MASK 0x4
-#define DISP_INTERRUPT_STATUS_CONTINUE2__LB_D3_VLINE_INTERRUPT_MASK 0x4
-#define DISP_INTERRUPT_STATUS_CONTINUE3__LB_D4_VLINE_INTERRUPT_MASK 0x4
-#define DISP_INTERRUPT_STATUS_CONTINUE4__LB_D5_VLINE_INTERRUPT_MASK 0x4
-#define DISP_INTERRUPT_STATUS_CONTINUE5__LB_D6_VLINE_INTERRUPT_MASK 0x4
-
-#define DISP_INTERRUPT_STATUS__DC_HPD1_INTERRUPT_MASK 0x20000
-#define DISP_INTERRUPT_STATUS_CONTINUE__DC_HPD2_INTERRUPT_MASK 0x20000
-#define DISP_INTERRUPT_STATUS_CONTINUE2__DC_HPD3_INTERRUPT_MASK 0x20000
-#define DISP_INTERRUPT_STATUS_CONTINUE3__DC_HPD4_INTERRUPT_MASK 0x20000
-#define DISP_INTERRUPT_STATUS_CONTINUE4__DC_HPD5_INTERRUPT_MASK 0x20000
-#define DISP_INTERRUPT_STATUS_CONTINUE5__DC_HPD6_INTERRUPT_MASK 0x20000
-
-#define GRPH_INTERRUPT_STATUS__GRPH_PFLIP_INT_OCCURRED_MASK 0x1
-#define GRPH_INTERRUPT_STATUS__GRPH_PFLIP_INT_CLEAR_MASK 0x100
-
-#define DC_HPD1_INT_CONTROL__DC_HPD1_INT_ACK_MASK 0x1
-
-#define R600_D1GRPH_SWAP_CONTROL                               0x1843
-#define R600_D1GRPH_SWAP_ENDIAN_NONE                    (0 << 0)
-#define R600_D1GRPH_SWAP_ENDIAN_16BIT                   (1 << 0)
-#define R600_D1GRPH_SWAP_ENDIAN_32BIT                   (2 << 0)
-#define R600_D1GRPH_SWAP_ENDIAN_64BIT                   (3 << 0)
-
-#define AVIVO_D1VGA_CONTROL					0x00cc
-#       define AVIVO_DVGA_CONTROL_MODE_ENABLE            (1 << 0)
-#       define AVIVO_DVGA_CONTROL_TIMING_SELECT          (1 << 8)
-#       define AVIVO_DVGA_CONTROL_SYNC_POLARITY_SELECT   (1 << 9)
-#       define AVIVO_DVGA_CONTROL_OVERSCAN_TIMING_SELECT (1 << 10)
-#       define AVIVO_DVGA_CONTROL_OVERSCAN_COLOR_EN      (1 << 16)
-#       define AVIVO_DVGA_CONTROL_ROTATE                 (1 << 24)
-#define AVIVO_D2VGA_CONTROL					0x00ce
-
-#define R600_BUS_CNTL                                           0x1508
-#       define R600_BIOS_ROM_DIS                                (1 << 1)
+
 
 #define R600_ROM_CNTL                              0x580
 #       define R600_SCK_OVERWRITE                  (1 << 1)
 #       define R600_SCK_PRESCALE_CRYSTAL_CLK_SHIFT 28
 #       define R600_SCK_PRESCALE_CRYSTAL_CLK_MASK  (0xf << 28)
 
-#define GRPH_INTERRUPT_CONTROL__GRPH_PFLIP_INT_MASK_MASK 0x1
-
-#define FMT_BIT_DEPTH_CONTROL                0x1bf2
-#define FMT_TRUNCATE_EN               (1 << 0)
-#define FMT_TRUNCATE_DEPTH            (1 << 4)
-#define FMT_SPATIAL_DITHER_EN         (1 << 8)
-#define FMT_SPATIAL_DITHER_MODE(x)    ((x) << 9)
-#define FMT_SPATIAL_DITHER_DEPTH      (1 << 12)
-#define FMT_FRAME_RANDOM_ENABLE       (1 << 13)
-#define FMT_RGB_RANDOM_ENABLE         (1 << 14)
-#define FMT_HIGHPASS_RANDOM_ENABLE    (1 << 15)
-#define FMT_TEMPORAL_DITHER_EN        (1 << 16)
-#define FMT_TEMPORAL_DITHER_DEPTH     (1 << 20)
-#define FMT_TEMPORAL_DITHER_OFFSET(x) ((x) << 21)
-#define FMT_TEMPORAL_LEVEL            (1 << 24)
-#define FMT_TEMPORAL_DITHER_RESET     (1 << 25)
-#define FMT_25FRC_SEL(x)              ((x) << 26)
-#define FMT_50FRC_SEL(x)              ((x) << 28)
-#define FMT_75FRC_SEL(x)              ((x) << 30)
-
-#define EVERGREEN_DC_LUT_CONTROL                        0x1a80
-#define EVERGREEN_DC_LUT_BLACK_OFFSET_BLUE              0x1a81
-#define EVERGREEN_DC_LUT_BLACK_OFFSET_GREEN             0x1a82
-#define EVERGREEN_DC_LUT_BLACK_OFFSET_RED               0x1a83
-#define EVERGREEN_DC_LUT_WHITE_OFFSET_BLUE              0x1a84
-#define EVERGREEN_DC_LUT_WHITE_OFFSET_GREEN             0x1a85
-#define EVERGREEN_DC_LUT_WHITE_OFFSET_RED               0x1a86
-#define EVERGREEN_DC_LUT_30_COLOR                       0x1a7c
-#define EVERGREEN_DC_LUT_RW_INDEX                       0x1a79
-#define EVERGREEN_DC_LUT_WRITE_EN_MASK                  0x1a7e
-#define EVERGREEN_DC_LUT_RW_MODE                        0x1a78
-
-#define EVERGREEN_GRPH_ENABLE                           0x1a00
-#define EVERGREEN_GRPH_CONTROL                          0x1a01
-#define EVERGREEN_GRPH_DEPTH(x)                  (((x) & 0x3) << 0)
-#define EVERGREEN_GRPH_DEPTH_8BPP                0
-#define EVERGREEN_GRPH_DEPTH_16BPP               1
-#define EVERGREEN_GRPH_DEPTH_32BPP               2
-#define EVERGREEN_GRPH_NUM_BANKS(x)              (((x) & 0x3) << 2)
-#define EVERGREEN_ADDR_SURF_2_BANK               0
-#define EVERGREEN_ADDR_SURF_4_BANK               1
-#define EVERGREEN_ADDR_SURF_8_BANK               2
-#define EVERGREEN_ADDR_SURF_16_BANK              3
-#define EVERGREEN_GRPH_Z(x)                      (((x) & 0x3) << 4)
-#define EVERGREEN_GRPH_BANK_WIDTH(x)             (((x) & 0x3) << 6)
-#define EVERGREEN_ADDR_SURF_BANK_WIDTH_1         0
-#define EVERGREEN_ADDR_SURF_BANK_WIDTH_2         1
-#define EVERGREEN_ADDR_SURF_BANK_WIDTH_4         2
-#define EVERGREEN_ADDR_SURF_BANK_WIDTH_8         3
-#define EVERGREEN_GRPH_FORMAT(x)                 (((x) & 0x7) << 8)
-
-#define EVERGREEN_GRPH_FORMAT_INDEXED            0
-#define EVERGREEN_GRPH_FORMAT_ARGB1555           0
-#define EVERGREEN_GRPH_FORMAT_ARGB565            1
-#define EVERGREEN_GRPH_FORMAT_ARGB4444           2
-#define EVERGREEN_GRPH_FORMAT_AI88               3
-#define EVERGREEN_GRPH_FORMAT_MONO16             4
-#define EVERGREEN_GRPH_FORMAT_BGRA5551           5
+#define GRPH_ARRAY_LINEAR_GENERAL      0
+#define GRPH_ARRAY_LINEAR_ALIGNED      1
+#define GRPH_ARRAY_1D_TILED_THIN1      2
+#define GRPH_ARRAY_2D_TILED_THIN1      4
+
+#define ES_AND_GS_AUTO       3
+#define BUF_SWAP_32BIT       (2 << 16)
+
+#define GRPH_DEPTH_8BPP                0
+#define GRPH_DEPTH_16BPP               1
+#define GRPH_DEPTH_32BPP               2
+
+/* 8 BPP */
+#define GRPH_FORMAT_INDEXED            0
+
+/* 16 BPP */
+#define GRPH_FORMAT_ARGB1555           0
+#define GRPH_FORMAT_ARGB565            1
+#define GRPH_FORMAT_ARGB4444           2
+#define GRPH_FORMAT_AI88               3
+#define GRPH_FORMAT_MONO16             4
+#define GRPH_FORMAT_BGRA5551           5
 
 /* 32 BPP */
-#define EVERGREEN_GRPH_FORMAT_ARGB8888           0
-#define EVERGREEN_GRPH_FORMAT_ARGB2101010        1
-#define EVERGREEN_GRPH_FORMAT_32BPP_DIG          2
-#define EVERGREEN_GRPH_FORMAT_8B_ARGB2101010     3
-#define EVERGREEN_GRPH_FORMAT_BGRA1010102        4
-#define EVERGREEN_GRPH_FORMAT_8B_BGRA1010102     5
-#define EVERGREEN_GRPH_FORMAT_RGB111110          6
-#define EVERGREEN_GRPH_FORMAT_BGR101111          7
-#define EVERGREEN_GRPH_BANK_HEIGHT(x)            (((x) & 0x3) << 11)
-#define EVERGREEN_ADDR_SURF_BANK_HEIGHT_1        0
-#define EVERGREEN_ADDR_SURF_BANK_HEIGHT_2        1
-#define EVERGREEN_ADDR_SURF_BANK_HEIGHT_4        2
-#define EVERGREEN_ADDR_SURF_BANK_HEIGHT_8        3
-#define EVERGREEN_GRPH_TILE_SPLIT(x)             (((x) & 0x7) << 13)
-#define EVERGREEN_ADDR_SURF_TILE_SPLIT_64B       0
-#define EVERGREEN_ADDR_SURF_TILE_SPLIT_128B      1
-#define EVERGREEN_ADDR_SURF_TILE_SPLIT_256B      2
-#define EVERGREEN_ADDR_SURF_TILE_SPLIT_512B      3
-#define EVERGREEN_ADDR_SURF_TILE_SPLIT_1KB       4
-#define EVERGREEN_ADDR_SURF_TILE_SPLIT_2KB       5
-#define EVERGREEN_ADDR_SURF_TILE_SPLIT_4KB       6
-#define EVERGREEN_GRPH_MACRO_TILE_ASPECT(x)      (((x) & 0x3) << 18)
-#define EVERGREEN_ADDR_SURF_MACRO_TILE_ASPECT_1  0
-#define EVERGREEN_ADDR_SURF_MACRO_TILE_ASPECT_2  1
-#define EVERGREEN_ADDR_SURF_MACRO_TILE_ASPECT_4  2
-#define EVERGREEN_ADDR_SURF_MACRO_TILE_ASPECT_8  3
-#define EVERGREEN_GRPH_ARRAY_MODE(x)             (((x) & 0x7) << 20)
-#define EVERGREEN_GRPH_ARRAY_LINEAR_GENERAL      0
-#define EVERGREEN_GRPH_ARRAY_LINEAR_ALIGNED      1
-#define EVERGREEN_GRPH_ARRAY_1D_TILED_THIN1      2
-#define EVERGREEN_GRPH_ARRAY_2D_TILED_THIN1      4
-#define EVERGREEN_ADDR_SURF_MACRO_TILE_ASPECT_1  0
-#define EVERGREEN_ADDR_SURF_MACRO_TILE_ASPECT_2  1
-#define EVERGREEN_ADDR_SURF_MACRO_TILE_ASPECT_4  2
-#define EVERGREEN_ADDR_SURF_MACRO_TILE_ASPECT_8  3
-
-#define EVERGREEN_GRPH_SWAP_CONTROL                     0x1a03
-#define EVERGREEN_GRPH_ENDIAN_SWAP(x)            (((x) & 0x3) << 0)
-#       define EVERGREEN_GRPH_ENDIAN_NONE               0
-#       define EVERGREEN_GRPH_ENDIAN_8IN16              1
-#       define EVERGREEN_GRPH_ENDIAN_8IN32              2
-#       define EVERGREEN_GRPH_ENDIAN_8IN64              3
-#define EVERGREEN_GRPH_RED_CROSSBAR(x)           (((x) & 0x3) << 4)
-#       define EVERGREEN_GRPH_RED_SEL_R                 0
-#       define EVERGREEN_GRPH_RED_SEL_G                 1
-#       define EVERGREEN_GRPH_RED_SEL_B                 2
-#       define EVERGREEN_GRPH_RED_SEL_A                 3
-#define EVERGREEN_GRPH_GREEN_CROSSBAR(x)         (((x) & 0x3) << 6)
-#       define EVERGREEN_GRPH_GREEN_SEL_G               0
-#       define EVERGREEN_GRPH_GREEN_SEL_B               1
-#       define EVERGREEN_GRPH_GREEN_SEL_A               2
-#       define EVERGREEN_GRPH_GREEN_SEL_R               3
-#define EVERGREEN_GRPH_BLUE_CROSSBAR(x)          (((x) & 0x3) << 8)
-#       define EVERGREEN_GRPH_BLUE_SEL_B                0
-#       define EVERGREEN_GRPH_BLUE_SEL_A                1
-#       define EVERGREEN_GRPH_BLUE_SEL_R                2
-#       define EVERGREEN_GRPH_BLUE_SEL_G                3
-#define EVERGREEN_GRPH_ALPHA_CROSSBAR(x)         (((x) & 0x3) << 10)
-#       define EVERGREEN_GRPH_ALPHA_SEL_A               0
-#       define EVERGREEN_GRPH_ALPHA_SEL_R               1
-#       define EVERGREEN_GRPH_ALPHA_SEL_G               2
-#       define EVERGREEN_GRPH_ALPHA_SEL_B               3
-
-#define EVERGREEN_D3VGA_CONTROL                         0xf8
-#define EVERGREEN_D4VGA_CONTROL                         0xf9
-#define EVERGREEN_D5VGA_CONTROL                         0xfa
-#define EVERGREEN_D6VGA_CONTROL                         0xfb
-
-#define EVERGREEN_GRPH_SURFACE_ADDRESS_MASK      0xffffff00
-
-#define EVERGREEN_GRPH_LUT_10BIT_BYPASS_CONTROL         0x1a02
-#define EVERGREEN_LUT_10BIT_BYPASS_EN            (1 << 8)
-
-#define EVERGREEN_GRPH_PITCH                            0x1a06
-#define EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH     0x1a07
-#define EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH   0x1a08
-#define EVERGREEN_GRPH_SURFACE_OFFSET_X                 0x1a09
-#define EVERGREEN_GRPH_SURFACE_OFFSET_Y                 0x1a0a
-#define EVERGREEN_GRPH_X_START                          0x1a0b
-#define EVERGREEN_GRPH_Y_START                          0x1a0c
-#define EVERGREEN_GRPH_X_END                            0x1a0d
-#define EVERGREEN_GRPH_Y_END                            0x1a0e
-#define EVERGREEN_GRPH_UPDATE                           0x1a11
-#define EVERGREEN_GRPH_SURFACE_UPDATE_PENDING    (1 << 2)
-#define EVERGREEN_GRPH_UPDATE_LOCK               (1 << 16)
-#define EVERGREEN_GRPH_FLIP_CONTROL                     0x1a12
-#define EVERGREEN_GRPH_SURFACE_UPDATE_H_RETRACE_EN (1 << 0)
-
-#define EVERGREEN_VIEWPORT_START                        0x1b5c
-#define EVERGREEN_VIEWPORT_SIZE                         0x1b5d
-#define EVERGREEN_DESKTOP_HEIGHT                        0x1ac1
-
-/* CUR blocks at 0x6998, 0x7598, 0x10198, 0x10d98, 0x11998, 0x12598 */
-#define EVERGREEN_CUR_CONTROL                           0x1a66
-#       define EVERGREEN_CURSOR_EN                      (1 << 0)
-#       define EVERGREEN_CURSOR_MODE(x)                 (((x) & 0x3) << 8)
-#       define EVERGREEN_CURSOR_MONO                    0
-#       define EVERGREEN_CURSOR_24_1                    1
-#       define EVERGREEN_CURSOR_24_8_PRE_MULT           2
-#       define EVERGREEN_CURSOR_24_8_UNPRE_MULT         3
-#       define EVERGREEN_CURSOR_2X_MAGNIFY              (1 << 16)
-#       define EVERGREEN_CURSOR_FORCE_MC_ON             (1 << 20)
-#       define EVERGREEN_CURSOR_URGENT_CONTROL(x)       (((x) & 0x7) << 24)
-#       define EVERGREEN_CURSOR_URGENT_ALWAYS           0
-#       define EVERGREEN_CURSOR_URGENT_1_8              1
-#       define EVERGREEN_CURSOR_URGENT_1_4              2
-#       define EVERGREEN_CURSOR_URGENT_3_8              3
-#       define EVERGREEN_CURSOR_URGENT_1_2              4
-#define EVERGREEN_CUR_SURFACE_ADDRESS                   0x1a67
-#       define EVERGREEN_CUR_SURFACE_ADDRESS_MASK       0xfffff000
-#define EVERGREEN_CUR_SIZE                              0x1a68
-#define EVERGREEN_CUR_SURFACE_ADDRESS_HIGH              0x1a69
-#define EVERGREEN_CUR_POSITION                          0x1a6a
-#define EVERGREEN_CUR_HOT_SPOT                          0x1a6b
-#define EVERGREEN_CUR_COLOR1                            0x1a6c
-#define EVERGREEN_CUR_COLOR2                            0x1a6d
-#define EVERGREEN_CUR_UPDATE                            0x1a6e
-#       define EVERGREEN_CURSOR_UPDATE_PENDING          (1 << 0)
-#       define EVERGREEN_CURSOR_UPDATE_TAKEN            (1 << 1)
-#       define EVERGREEN_CURSOR_UPDATE_LOCK             (1 << 16)
-#       define EVERGREEN_CURSOR_DISABLE_MULTIPLE_UPDATE (1 << 24)
-
-
-#define NI_INPUT_CSC_CONTROL                           0x1a35
-#       define NI_INPUT_CSC_GRPH_MODE(x)               (((x) & 0x3) << 0)
-#       define NI_INPUT_CSC_BYPASS                     0
-#       define NI_INPUT_CSC_PROG_COEFF                 1
-#       define NI_INPUT_CSC_PROG_SHARED_MATRIXA        2
-#       define NI_INPUT_CSC_OVL_MODE(x)                (((x) & 0x3) << 4)
-
-#define NI_OUTPUT_CSC_CONTROL                          0x1a3c
-#       define NI_OUTPUT_CSC_GRPH_MODE(x)              (((x) & 0x7) << 0)
-#       define NI_OUTPUT_CSC_BYPASS                    0
-#       define NI_OUTPUT_CSC_TV_RGB                    1
-#       define NI_OUTPUT_CSC_YCBCR_601                 2
-#       define NI_OUTPUT_CSC_YCBCR_709                 3
-#       define NI_OUTPUT_CSC_PROG_COEFF                4
-#       define NI_OUTPUT_CSC_PROG_SHARED_MATRIXB       5
-#       define NI_OUTPUT_CSC_OVL_MODE(x)               (((x) & 0x7) << 4)
-
-#define NI_DEGAMMA_CONTROL                             0x1a58
-#       define NI_GRPH_DEGAMMA_MODE(x)                 (((x) & 0x3) << 0)
-#       define NI_DEGAMMA_BYPASS                       0
-#       define NI_DEGAMMA_SRGB_24                      1
-#       define NI_DEGAMMA_XVYCC_222                    2
-#       define NI_OVL_DEGAMMA_MODE(x)                  (((x) & 0x3) << 4)
-#       define NI_ICON_DEGAMMA_MODE(x)                 (((x) & 0x3) << 8)
-#       define NI_CURSOR_DEGAMMA_MODE(x)               (((x) & 0x3) << 12)
-
-#define NI_GAMUT_REMAP_CONTROL                         0x1a59
-#       define NI_GRPH_GAMUT_REMAP_MODE(x)             (((x) & 0x3) << 0)
-#       define NI_GAMUT_REMAP_BYPASS                   0
-#       define NI_GAMUT_REMAP_PROG_COEFF               1
-#       define NI_GAMUT_REMAP_PROG_SHARED_MATRIXA      2
-#       define NI_GAMUT_REMAP_PROG_SHARED_MATRIXB      3
-#       define NI_OVL_GAMUT_REMAP_MODE(x)              (((x) & 0x3) << 4)
-
-#define NI_REGAMMA_CONTROL                             0x1aa0
-#       define NI_GRPH_REGAMMA_MODE(x)                 (((x) & 0x7) << 0)
-#       define NI_REGAMMA_BYPASS                       0
-#       define NI_REGAMMA_SRGB_24                      1
-#       define NI_REGAMMA_XVYCC_222                    2
-#       define NI_REGAMMA_PROG_A                       3
-#       define NI_REGAMMA_PROG_B                       4
-#       define NI_OVL_REGAMMA_MODE(x)                  (((x) & 0x7) << 4)
-
-
-#define NI_PRESCALE_GRPH_CONTROL                       0x1a2d
-#       define NI_GRPH_PRESCALE_BYPASS                 (1 << 4)
-
-#define NI_PRESCALE_OVL_CONTROL                        0x1a31
-#       define NI_OVL_PRESCALE_BYPASS                  (1 << 4)
-
-#define NI_INPUT_GAMMA_CONTROL                         0x1a10
-#       define NI_GRPH_INPUT_GAMMA_MODE(x)             (((x) & 0x3) << 0)
-#       define NI_INPUT_GAMMA_USE_LUT                  0
-#       define NI_INPUT_GAMMA_BYPASS                   1
-#       define NI_INPUT_GAMMA_SRGB_24                  2
-#       define NI_INPUT_GAMMA_XVYCC_222                3
-#       define NI_OVL_INPUT_GAMMA_MODE(x)              (((x) & 0x3) << 4)
-
-#define	BLACKOUT_MODE_MASK			0x00000007
-#define	VGA_RENDER_CONTROL			0xC0
-#define R_000300_VGA_RENDER_CONTROL             0xC0
-#define C_000300_VGA_VSTATUS_CNTL               0xFFFCFFFF
-#define EVERGREEN_CRTC_STATUS                   0x1BA3
-#define EVERGREEN_CRTC_V_BLANK                  (1 << 0)
-#define EVERGREEN_CRTC_STATUS_POSITION          0x1BA4
-/* CRTC blocks at 0x6df0, 0x79f0, 0x105f0, 0x111f0, 0x11df0, 0x129f0 */
-#define EVERGREEN_CRTC_V_BLANK_START_END                0x1b8d
-#define EVERGREEN_CRTC_CONTROL                          0x1b9c
-#       define EVERGREEN_CRTC_MASTER_EN                 (1 << 0)
-#       define EVERGREEN_CRTC_DISP_READ_REQUEST_DISABLE (1 << 24)
-#define EVERGREEN_CRTC_BLANK_CONTROL                    0x1b9d
-#       define EVERGREEN_CRTC_BLANK_DATA_EN             (1 << 8)
-#       define EVERGREEN_CRTC_V_BLANK                   (1 << 0)
-#define EVERGREEN_CRTC_STATUS_HV_COUNT                  0x1ba8
-#define EVERGREEN_CRTC_UPDATE_LOCK                      0x1bb5
-#define EVERGREEN_MASTER_UPDATE_LOCK                    0x1bbd
-#define EVERGREEN_MASTER_UPDATE_MODE                    0x1bbe
-#define EVERGREEN_GRPH_UPDATE_LOCK               (1 << 16)
-#define EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH     0x1a07
-#define EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH   0x1a08
-#define EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS          0x1a04
-#define EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS        0x1a05
-#define EVERGREEN_GRPH_UPDATE                           0x1a11
-#define EVERGREEN_VGA_MEMORY_BASE_ADDRESS               0xc4
-#define EVERGREEN_VGA_MEMORY_BASE_ADDRESS_HIGH          0xc9
-#define EVERGREEN_GRPH_SURFACE_UPDATE_PENDING    (1 << 2)
-
-#define mmVM_CONTEXT1_CNTL__xxRANGE_PROTECTION_FAULT_ENABLE_DEFAULT_MASK 0x10
-#define mmVM_CONTEXT1_CNTL__xxRANGE_PROTECTION_FAULT_ENABLE_DEFAULT__SHIFT 0x4
-#define mmVM_CONTEXT1_CNTL__xxDUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT_MASK 0x80
-#define mmVM_CONTEXT1_CNTL__xxDUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT__SHIFT 0x7
-#define mmVM_CONTEXT1_CNTL__xxPDE0_PROTECTION_FAULT_ENABLE_DEFAULT_MASK 0x400
-#define mmVM_CONTEXT1_CNTL__xxPDE0_PROTECTION_FAULT_ENABLE_DEFAULT__SHIFT 0xa
-#define mmVM_CONTEXT1_CNTL__xxVALID_PROTECTION_FAULT_ENABLE_DEFAULT_MASK 0x2000
-#define mmVM_CONTEXT1_CNTL__xxVALID_PROTECTION_FAULT_ENABLE_DEFAULT__SHIFT 0xd
-#define mmVM_CONTEXT1_CNTL__xxREAD_PROTECTION_FAULT_ENABLE_DEFAULT_MASK 0x10000
-#define mmVM_CONTEXT1_CNTL__xxREAD_PROTECTION_FAULT_ENABLE_DEFAULT__SHIFT 0x10
-#define mmVM_CONTEXT1_CNTL__xxWRITE_PROTECTION_FAULT_ENABLE_DEFAULT_MASK 0x80000
-#define mmVM_CONTEXT1_CNTL__xxWRITE_PROTECTION_FAULT_ENABLE_DEFAULT__SHIFT 0x13
-
-#define mmVM_CONTEXT1_PROTECTION_FAULT_STATUS__xxVMID_MASK 0x1e000000
-#define mmVM_CONTEXT1_PROTECTION_FAULT_STATUS__xxVMID__SHIFT 0x19
-#define mmVM_CONTEXT1_PROTECTION_FAULT_STATUS__xxPROTECTIONS_MASK 0xff
-#define mmVM_CONTEXT1_PROTECTION_FAULT_STATUS__xxPROTECTIONS__SHIFT 0x0
-#define mmVM_CONTEXT1_PROTECTION_FAULT_STATUS__xxMEMORY_CLIENT_ID_MASK 0xff000
-#define mmVM_CONTEXT1_PROTECTION_FAULT_STATUS__xxMEMORY_CLIENT_ID__SHIFT 0xc
-#define mmVM_CONTEXT1_PROTECTION_FAULT_STATUS__xxMEMORY_CLIENT_RW_MASK 0x1000000
-#define mmVM_CONTEXT1_PROTECTION_FAULT_STATUS__xxMEMORY_CLIENT_RW__SHIFT 0x18
-
-#define mmMC_SHARED_BLACKOUT_CNTL__xxBLACKOUT_MODE_MASK 0x7
-#define mmMC_SHARED_BLACKOUT_CNTL__xxBLACKOUT_MODE__SHIFT 0x0
-
-#define mmBIF_FB_EN__xxFB_READ_EN_MASK 0x1
-#define mmBIF_FB_EN__xxFB_READ_EN__SHIFT 0x0
-#define mmBIF_FB_EN__xxFB_WRITE_EN_MASK 0x2
-#define mmBIF_FB_EN__xxFB_WRITE_EN__SHIFT 0x1
-
-#define mmSRBM_SOFT_RESET__xxSOFT_RESET_VMC_MASK 0x20000
-#define mmSRBM_SOFT_RESET__xxSOFT_RESET_VMC__SHIFT 0x11
-#define mmSRBM_SOFT_RESET__xxSOFT_RESET_MC_MASK 0x800
-#define mmSRBM_SOFT_RESET__xxSOFT_RESET_MC__SHIFT 0xb
-
-#define VM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK 0x8
-#define VM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT__SHIFT 0x3
-#define VM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK 0x40
-#define VM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT__SHIFT 0x6
-#define VM_CONTEXT1_CNTL__PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK 0x200
-#define VM_CONTEXT1_CNTL__PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT__SHIFT 0x9
-#define VM_CONTEXT1_CNTL__VALID_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK 0x1000
-#define VM_CONTEXT1_CNTL__VALID_PROTECTION_FAULT_ENABLE_INTERRUPT__SHIFT 0xc
-#define VM_CONTEXT1_CNTL__READ_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK 0x8000
-#define VM_CONTEXT1_CNTL__READ_PROTECTION_FAULT_ENABLE_INTERRUPT__SHIFT 0xf
-#define VM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK 0x40000
-#define VM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT__SHIFT 0x12
+#define GRPH_FORMAT_ARGB8888           0
+#define GRPH_FORMAT_ARGB2101010        1
+#define GRPH_FORMAT_32BPP_DIG          2
+#define GRPH_FORMAT_8B_ARGB2101010     3
+#define GRPH_FORMAT_BGRA1010102        4
+#define GRPH_FORMAT_8B_BGRA1010102     5
+#define GRPH_FORMAT_RGB111110          6
+#define GRPH_FORMAT_BGR101111          7
+
+#define GRPH_ENDIAN_NONE               0
+#define GRPH_ENDIAN_8IN16              1
+#define GRPH_ENDIAN_8IN32              2
+#define GRPH_ENDIAN_8IN64              3
+#define GRPH_RED_SEL_R                 0
+#define GRPH_RED_SEL_G                 1
+#define GRPH_RED_SEL_B                 2
+#define GRPH_RED_SEL_A                 3
+
+#define GRPH_GREEN_SEL_G               0
+#define GRPH_GREEN_SEL_B               1
+#define GRPH_GREEN_SEL_A               2
+#define GRPH_GREEN_SEL_R               3
+
+#define GRPH_BLUE_SEL_B                0
+#define GRPH_BLUE_SEL_A                1
+#define GRPH_BLUE_SEL_R                2
+#define GRPH_BLUE_SEL_G                3
+
+#define GRPH_ALPHA_SEL_A               0
+#define GRPH_ALPHA_SEL_R               1
+#define GRPH_ALPHA_SEL_G               2
+#define GRPH_ALPHA_SEL_B               3
+
+/* CUR_CONTROL */
+	#define CURSOR_MONO                    0
+	#define CURSOR_24_1                    1
+	#define CURSOR_24_8_PRE_MULT           2
+	#define CURSOR_24_8_UNPRE_MULT         3
+	#define CURSOR_URGENT_ALWAYS           0
+	#define CURSOR_URGENT_1_8              1
+	#define CURSOR_URGENT_1_4              2
+	#define CURSOR_URGENT_3_8              3
+	#define CURSOR_URGENT_1_2              4
+
+/* INPUT_CSC_CONTROL */
+#       define INPUT_CSC_BYPASS                     0
+#       define INPUT_CSC_PROG_COEFF                 1
+#       define INPUT_CSC_PROG_SHARED_MATRIXA        2
+
+/* OUTPUT_CSC_CONTROL */
+#       define OUTPUT_CSC_BYPASS                    0
+#       define OUTPUT_CSC_TV_RGB                    1
+#       define OUTPUT_CSC_YCBCR_601                 2
+#       define OUTPUT_CSC_YCBCR_709                 3
+#       define OUTPUT_CSC_PROG_COEFF                4
+#       define OUTPUT_CSC_PROG_SHARED_MATRIXB       5
+
+/* DEGAMMA_CONTROL */
+#       define DEGAMMA_BYPASS                       0
+#       define DEGAMMA_SRGB_24                      1
+#       define DEGAMMA_XVYCC_222                    2
+
+/* GAMUT_REMAP_CONTROL */
+#       define GAMUT_REMAP_BYPASS                   0
+#       define GAMUT_REMAP_PROG_COEFF               1
+#       define GAMUT_REMAP_PROG_SHARED_MATRIXA      2
+#       define GAMUT_REMAP_PROG_SHARED_MATRIXB      3
+
+/* REGAMMA_CONTROL */
+#       define REGAMMA_BYPASS                       0
+#       define REGAMMA_SRGB_24                      1
+#       define REGAMMA_XVYCC_222                    2
+#       define REGAMMA_PROG_A                       3
+#       define REGAMMA_PROG_B                       4
+
+
+/* INPUT_GAMMA_CONTROL */
+#       define INPUT_GAMMA_USE_LUT                  0
+#       define INPUT_GAMMA_BYPASS                   1
+#       define INPUT_GAMMA_SRGB_24                  2
+#       define INPUT_GAMMA_XVYCC_222                3
 
 #define MC_SEQ_MISC0__MT__MASK	0xf0000000
 #define MC_SEQ_MISC0__MT__GDDR1  0x10000000
@@ -2426,28 +740,14 @@
 #define MC_SEQ_MISC0__MT__HBM    0x60000000
 #define MC_SEQ_MISC0__MT__DDR3   0xB0000000
 
-#define GRBM_STATUS__GUI_ACTIVE_MASK 0x80000000
 #define CP_INT_CNTL_RING__TIME_STAMP_INT_ENABLE_MASK 0x4000000
-#define CP_INT_CNTL_RING0__PRIV_REG_INT_ENABLE_MASK 0x800000
-#define CP_INT_CNTL_RING0__PRIV_INSTR_INT_ENABLE_MASK 0x400000
 #define PACKET3_SEM_WAIT_ON_SIGNAL    (0x1 << 12)
 #define PACKET3_SEM_SEL_SIGNAL	    (0x6 << 29)
 #define PACKET3_SEM_SEL_WAIT	    (0x7 << 29)
 
-#define CONFIG_CNTL	0x1509
-#define CC_DRM_ID_STRAPS	0X1559
 #define AMDGPU_PCIE_INDEX	0xc
 #define AMDGPU_PCIE_DATA	0xd
 
-#define DMA_SEM_INCOMPLETE_TIMER_CNTL                     0x3411
-#define DMA_SEM_WAIT_FAIL_TIMER_CNTL                      0x3412
-#define DMA_MODE                                          0x342f
-#define DMA_RB_RPTR_ADDR_HI                               0x3407
-#define DMA_RB_RPTR_ADDR_LO                               0x3408
-#define DMA_BUSY_MASK 0x20
-#define DMA1_BUSY_MASK 0X40
-#define SDMA_MAX_INSTANCE 2
-
 #define PCIE_BUS_CLK    10000
 #define TCLK            (PCIE_BUS_CLK / 10)
 #define	PCIE_PORT_INDEX					0xe
@@ -2457,8 +757,6 @@
 #define EVERGREEN_PIF_PHY1_INDEX                        0x10
 #define EVERGREEN_PIF_PHY1_DATA				0x14
 
-#define	MC_VM_FB_OFFSET					0x81a
-
 /* Discrete VCE clocks */
 #define CG_VCEPLL_FUNC_CNTL                             0xc0030600
 #define    VCEPLL_RESET_MASK                            0x00000001
diff --git a/drivers/gpu/drm/amd/amdgpu/sienna_cichlid.c b/drivers/gpu/drm/amd/amdgpu/sienna_cichlid.c
index 81a6d5b94987..2594467bdd87 100644
--- a/drivers/gpu/drm/amd/amdgpu/sienna_cichlid.c
+++ b/drivers/gpu/drm/amd/amdgpu/sienna_cichlid.c
@@ -36,11 +36,11 @@ static bool sienna_cichlid_is_mode2_default(struct amdgpu_reset_control *reset_c
 #if 0
 	struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle;
 
-	if (adev->ip_versions[MP1_HWIP][0] == IP_VERSION(11, 0, 7) &&
+	if (amdgpu_ip_version(adev, MP1_HWIP, 0) == IP_VERSION(11, 0, 7) &&
 	    adev->pm.fw_version >= 0x3a5500 && !amdgpu_sriov_vf(adev))
 		return true;
 #endif
-	return false;
+	return amdgpu_reset_method == AMD_RESET_METHOD_MODE2;
 }
 
 static struct amdgpu_reset_handler *
@@ -48,18 +48,17 @@ sienna_cichlid_get_reset_handler(struct amdgpu_reset_control *reset_ctl,
 			    struct amdgpu_reset_context *reset_context)
 {
 	struct amdgpu_reset_handler *handler;
+	int i;
 
 	if (reset_context->method != AMD_RESET_METHOD_NONE) {
-		list_for_each_entry(handler, &reset_ctl->reset_handlers,
-				     handler_list) {
+		for_each_handler(i, handler, reset_ctl)	{
 			if (handler->reset_method == reset_context->method)
 				return handler;
 		}
 	}
 
 	if (sienna_cichlid_is_mode2_default(reset_ctl)) {
-		list_for_each_entry (handler, &reset_ctl->reset_handlers,
-				     handler_list) {
+		for_each_handler(i, handler, reset_ctl)	{
 			if (handler->reset_method == AMD_RESET_METHOD_MODE2)
 				return handler;
 		}
@@ -82,18 +81,12 @@ static int sienna_cichlid_mode2_suspend_ip(struct amdgpu_device *adev)
 			      AMD_IP_BLOCK_TYPE_SDMA))
 			continue;
 
-		r = adev->ip_blocks[i].version->funcs->suspend(adev);
-
-		if (r) {
-			dev_err(adev->dev,
-				"suspend of IP block <%s> failed %d\n",
-				adev->ip_blocks[i].version->funcs->name, r);
+		r = amdgpu_ip_block_suspend(&adev->ip_blocks[i]);
+		if (r)
 			return r;
-		}
-		adev->ip_blocks[i].status.hw = false;
 	}
 
-	return r;
+	return 0;
 }
 
 static int
@@ -120,9 +113,9 @@ static void sienna_cichlid_async_reset(struct work_struct *work)
 	struct amdgpu_reset_control *reset_ctl =
 		container_of(work, struct amdgpu_reset_control, reset_work);
 	struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle;
+	int i;
 
-	list_for_each_entry(handler, &reset_ctl->reset_handlers,
-			     handler_list) {
+	for_each_handler(i, handler, reset_ctl)	{
 		if (handler->reset_method == reset_ctl->active_reset) {
 			dev_dbg(adev->dev, "Resetting device\n");
 			handler->do_reset(adev);
@@ -176,15 +169,9 @@ static int sienna_cichlid_mode2_restore_ip(struct amdgpu_device *adev)
 
 	for (i = 0; i < adev->num_ip_blocks; i++) {
 		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH) {
-			r = adev->ip_blocks[i].version->funcs->resume(adev);
-			if (r) {
-				dev_err(adev->dev,
-					"resume of IP block <%s> failed %d\n",
-					adev->ip_blocks[i].version->funcs->name, r);
+			r = amdgpu_ip_block_resume(&adev->ip_blocks[i]);
+			if (r)
 				return r;
-			}
-
-			adev->ip_blocks[i].status.hw = true;
 		}
 	}
 
@@ -194,15 +181,9 @@ static int sienna_cichlid_mode2_restore_ip(struct amdgpu_device *adev)
 		      adev->ip_blocks[i].version->type ==
 			      AMD_IP_BLOCK_TYPE_SDMA))
 			continue;
-		r = adev->ip_blocks[i].version->funcs->resume(adev);
-		if (r) {
-			dev_err(adev->dev,
-				"resume of IP block <%s> failed %d\n",
-				adev->ip_blocks[i].version->funcs->name, r);
+		r = amdgpu_ip_block_resume(&adev->ip_blocks[i]);
+		if (r)
 			return r;
-		}
-
-		adev->ip_blocks[i].status.hw = true;
 	}
 
 	for (i = 0; i < adev->num_ip_blocks; i++) {
@@ -214,7 +195,7 @@ static int sienna_cichlid_mode2_restore_ip(struct amdgpu_device *adev)
 
 		if (adev->ip_blocks[i].version->funcs->late_init) {
 			r = adev->ip_blocks[i].version->funcs->late_init(
-				(void *)adev);
+				&adev->ip_blocks[i]);
 			if (r) {
 				dev_err(adev->dev,
 					"late_init of IP block <%s> failed %d after reset\n",
@@ -239,6 +220,7 @@ sienna_cichlid_mode2_restore_hwcontext(struct amdgpu_reset_control *reset_ctl,
 	int r;
 	struct amdgpu_device *tmp_adev = (struct amdgpu_device *)reset_ctl->handle;
 
+	amdgpu_set_init_level(tmp_adev, AMDGPU_INIT_LEVEL_RESET_RECOVERY);
 	dev_info(tmp_adev->dev,
 			"GPU reset succeeded, trying to resume\n");
 	r = sienna_cichlid_mode2_restore_ip(tmp_adev);
@@ -256,6 +238,7 @@ sienna_cichlid_mode2_restore_hwcontext(struct amdgpu_reset_control *reset_ctl,
 
 	amdgpu_irq_gpu_reset_resume_helper(tmp_adev);
 
+	amdgpu_set_init_level(tmp_adev, AMDGPU_INIT_LEVEL_DEFAULT);
 	r = amdgpu_ib_ring_tests(tmp_adev);
 	if (r) {
 		dev_err(tmp_adev->dev,
@@ -281,6 +264,11 @@ static struct amdgpu_reset_handler sienna_cichlid_mode2_handler = {
 	.do_reset		= sienna_cichlid_mode2_reset,
 };
 
+static struct amdgpu_reset_handler
+	*sienna_cichlid_rst_handlers[AMDGPU_RESET_MAX_HANDLERS] = {
+		&sienna_cichlid_mode2_handler,
+	};
+
 int sienna_cichlid_reset_init(struct amdgpu_device *adev)
 {
 	struct amdgpu_reset_control *reset_ctl;
@@ -294,11 +282,9 @@ int sienna_cichlid_reset_init(struct amdgpu_device *adev)
 	reset_ctl->active_reset = AMD_RESET_METHOD_NONE;
 	reset_ctl->get_reset_handler = sienna_cichlid_get_reset_handler;
 
-	INIT_LIST_HEAD(&reset_ctl->reset_handlers);
 	INIT_WORK(&reset_ctl->reset_work, reset_ctl->async_reset);
 	/* Only mode2 is handled through reset control now */
-	amdgpu_reset_add_handler(reset_ctl, &sienna_cichlid_mode2_handler);
-
+	reset_ctl->reset_handlers = &sienna_cichlid_rst_handlers;
 	adev->reset_cntl = reset_ctl;
 
 	return 0;
diff --git a/drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.c b/drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.c
index dd2d66090d23..68aef47254a9 100644
--- a/drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.c
+++ b/drivers/gpu/drm/amd/amdgpu/smu_v11_0_i2c.c
@@ -743,7 +743,7 @@ int smu_v11_0_i2c_control_init(struct amdgpu_device *adev)
 	adev->pm.ras_eeprom_i2c_bus = &adev->pm.smu_i2c[0].adapter;
 	adev->pm.fru_eeprom_i2c_bus = &adev->pm.smu_i2c[0].adapter;
 
-	res = i2c_add_adapter(control);
+	res = devm_i2c_add_adapter(adev->dev, control);
 	if (res)
 		DRM_ERROR("Failed to register hw i2c, err: %d\n", res);
 
@@ -752,9 +752,6 @@ int smu_v11_0_i2c_control_init(struct amdgpu_device *adev)
 
 void smu_v11_0_i2c_control_fini(struct amdgpu_device *adev)
 {
-	struct i2c_adapter *control = adev->pm.ras_eeprom_i2c_bus;
-
-	i2c_del_adapter(control);
 	adev->pm.ras_eeprom_i2c_bus = NULL;
 	adev->pm.fru_eeprom_i2c_bus = NULL;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/smu_v13_0_10.c b/drivers/gpu/drm/amd/amdgpu/smu_v13_0_10.c
index ae29620b1ea4..70569ea906bc 100644
--- a/drivers/gpu/drm/amd/amdgpu/smu_v13_0_10.c
+++ b/drivers/gpu/drm/amd/amdgpu/smu_v13_0_10.c
@@ -44,10 +44,10 @@ smu_v13_0_10_get_reset_handler(struct amdgpu_reset_control *reset_ctl,
 {
 	struct amdgpu_reset_handler *handler;
 	struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle;
+	int i;
 
 	if (reset_context->method != AMD_RESET_METHOD_NONE) {
-		list_for_each_entry(handler, &reset_ctl->reset_handlers,
-				     handler_list) {
+		for_each_handler(i, handler, reset_ctl) {
 			if (handler->reset_method == reset_context->method)
 				return handler;
 		}
@@ -55,8 +55,7 @@ smu_v13_0_10_get_reset_handler(struct amdgpu_reset_control *reset_ctl,
 
 	if (smu_v13_0_10_is_mode2_default(reset_ctl) &&
 		amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_MODE2) {
-		list_for_each_entry (handler, &reset_ctl->reset_handlers,
-				     handler_list) {
+		for_each_handler(i, handler, reset_ctl)	{
 			if (handler->reset_method == AMD_RESET_METHOD_MODE2)
 				return handler;
 		}
@@ -81,18 +80,12 @@ static int smu_v13_0_10_mode2_suspend_ip(struct amdgpu_device *adev)
 			      AMD_IP_BLOCK_TYPE_MES))
 			continue;
 
-		r = adev->ip_blocks[i].version->funcs->suspend(adev);
-
-		if (r) {
-			dev_err(adev->dev,
-				"suspend of IP block <%s> failed %d\n",
-				adev->ip_blocks[i].version->funcs->name, r);
+		r = amdgpu_ip_block_suspend(&adev->ip_blocks[i]);
+		if (r)
 			return r;
-		}
-		adev->ip_blocks[i].status.hw = false;
 	}
 
-	return r;
+	return 0;
 }
 
 static int
@@ -119,9 +112,9 @@ static void smu_v13_0_10_async_reset(struct work_struct *work)
 	struct amdgpu_reset_control *reset_ctl =
 		container_of(work, struct amdgpu_reset_control, reset_work);
 	struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle;
+	int i;
 
-	list_for_each_entry(handler, &reset_ctl->reset_handlers,
-			     handler_list) {
+	for_each_handler(i, handler, reset_ctl)	{
 		if (handler->reset_method == reset_ctl->active_reset) {
 			dev_dbg(adev->dev, "Resetting device\n");
 			handler->do_reset(adev);
@@ -187,15 +180,9 @@ static int smu_v13_0_10_mode2_restore_ip(struct amdgpu_device *adev)
 		      adev->ip_blocks[i].version->type ==
 			      AMD_IP_BLOCK_TYPE_SDMA))
 			continue;
-		r = adev->ip_blocks[i].version->funcs->resume(adev);
-		if (r) {
-			dev_err(adev->dev,
-				"resume of IP block <%s> failed %d\n",
-				adev->ip_blocks[i].version->funcs->name, r);
+		r = amdgpu_ip_block_resume(&adev->ip_blocks[i]);
+		if (r)
 			return r;
-		}
-
-		adev->ip_blocks[i].status.hw = true;
 	}
 
 	for (i = 0; i < adev->num_ip_blocks; i++) {
@@ -209,7 +196,7 @@ static int smu_v13_0_10_mode2_restore_ip(struct amdgpu_device *adev)
 
 		if (adev->ip_blocks[i].version->funcs->late_init) {
 			r = adev->ip_blocks[i].version->funcs->late_init(
-				(void *)adev);
+				&adev->ip_blocks[i]);
 			if (r) {
 				dev_err(adev->dev,
 					"late_init of IP block <%s> failed %d after reset\n",
@@ -234,6 +221,7 @@ smu_v13_0_10_mode2_restore_hwcontext(struct amdgpu_reset_control *reset_ctl,
 	int r;
 	struct amdgpu_device *tmp_adev = (struct amdgpu_device *)reset_ctl->handle;
 
+	amdgpu_set_init_level(tmp_adev, AMDGPU_INIT_LEVEL_RESET_RECOVERY);
 	dev_info(tmp_adev->dev,
 			"GPU reset succeeded, trying to resume\n");
 	r = smu_v13_0_10_mode2_restore_ip(tmp_adev);
@@ -247,6 +235,7 @@ smu_v13_0_10_mode2_restore_hwcontext(struct amdgpu_reset_control *reset_ctl,
 
 	amdgpu_irq_gpu_reset_resume_helper(tmp_adev);
 
+	amdgpu_set_init_level(tmp_adev, AMDGPU_INIT_LEVEL_DEFAULT);
 	r = amdgpu_ib_ring_tests(tmp_adev);
 	if (r) {
 		dev_err(tmp_adev->dev,
@@ -272,6 +261,11 @@ static struct amdgpu_reset_handler smu_v13_0_10_mode2_handler = {
 	.do_reset		= smu_v13_0_10_mode2_reset,
 };
 
+static struct amdgpu_reset_handler
+	*smu_v13_0_10_rst_handlers[AMDGPU_RESET_MAX_HANDLERS] = {
+		&smu_v13_0_10_mode2_handler,
+	};
+
 int smu_v13_0_10_reset_init(struct amdgpu_device *adev)
 {
 	struct amdgpu_reset_control *reset_ctl;
@@ -285,10 +279,9 @@ int smu_v13_0_10_reset_init(struct amdgpu_device *adev)
 	reset_ctl->active_reset = AMD_RESET_METHOD_NONE;
 	reset_ctl->get_reset_handler = smu_v13_0_10_get_reset_handler;
 
-	INIT_LIST_HEAD(&reset_ctl->reset_handlers);
 	INIT_WORK(&reset_ctl->reset_work, reset_ctl->async_reset);
 	/* Only mode2 is handled through reset control now */
-	amdgpu_reset_add_handler(reset_ctl, &smu_v13_0_10_mode2_handler);
+	reset_ctl->reset_handlers = &smu_v13_0_10_rst_handlers;
 
 	adev->reset_cntl = reset_ctl;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/smuio_v13_0.c b/drivers/gpu/drm/amd/amdgpu/smuio_v13_0.c
index 13e905c22592..bf8b8e5ddf5d 100644
--- a/drivers/gpu/drm/amd/amdgpu/smuio_v13_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/smuio_v13_0.c
@@ -128,6 +128,27 @@ static bool smuio_v13_0_is_host_gpu_xgmi_supported(struct amdgpu_device *adev)
 	return data ? true : false;
 }
 
+static enum amdgpu_pkg_type smuio_v13_0_get_pkg_type(struct amdgpu_device *adev)
+{
+	enum amdgpu_pkg_type pkg_type;
+	u32 data;
+
+	data = RREG32_SOC15(SMUIO, 0, regSMUIO_MCM_CONFIG);
+	data = REG_GET_FIELD(data, SMUIO_MCM_CONFIG, TOPOLOGY_ID);
+
+	switch (data) {
+	case 0x4:
+	case 0xC:
+		pkg_type = AMDGPU_PKG_TYPE_CEM;
+		break;
+	default:
+		pkg_type = AMDGPU_PKG_TYPE_OAM;
+		break;
+	}
+
+	return pkg_type;
+}
+
 const struct amdgpu_smuio_funcs smuio_v13_0_funcs = {
 	.get_rom_index_offset = smuio_v13_0_get_rom_index_offset,
 	.get_rom_data_offset = smuio_v13_0_get_rom_data_offset,
@@ -136,4 +157,5 @@ const struct amdgpu_smuio_funcs smuio_v13_0_funcs = {
 	.is_host_gpu_xgmi_supported = smuio_v13_0_is_host_gpu_xgmi_supported,
 	.update_rom_clock_gating = smuio_v13_0_update_rom_clock_gating,
 	.get_clock_gating_state = smuio_v13_0_get_clock_gating_state,
+	.get_pkg_type = smuio_v13_0_get_pkg_type,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/smuio_v13_0_3.c b/drivers/gpu/drm/amd/amdgpu/smuio_v13_0_3.c
new file mode 100644
index 000000000000..5461b5289793
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/smuio_v13_0_3.c
@@ -0,0 +1,109 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "amdgpu.h"
+#include "smuio_v13_0_3.h"
+#include "soc15_common.h"
+#include "smuio/smuio_13_0_3_offset.h"
+#include "smuio/smuio_13_0_3_sh_mask.h"
+
+#define PKG_TYPE_MASK		0x00000003L
+
+/**
+ * smuio_v13_0_3_get_die_id - query die id from FCH.
+ *
+ * @adev: amdgpu device pointer
+ *
+ * Returns die id
+ */
+static u32 smuio_v13_0_3_get_die_id(struct amdgpu_device *adev)
+{
+	u32 data, die_id;
+
+	data = RREG32_SOC15(SMUIO, 0, regSMUIO_MCM_CONFIG);
+	die_id = REG_GET_FIELD(data, SMUIO_MCM_CONFIG, DIE_ID);
+
+	return die_id;
+}
+
+/**
+ * smuio_v13_0_3_get_socket_id - query socket id from FCH
+ *
+ * @adev: amdgpu device pointer
+ *
+ * Returns socket id
+ */
+static u32 smuio_v13_0_3_get_socket_id(struct amdgpu_device *adev)
+{
+	u32 data, socket_id;
+
+	data = RREG32_SOC15(SMUIO, 0, regSMUIO_MCM_CONFIG);
+	socket_id = REG_GET_FIELD(data, SMUIO_MCM_CONFIG, SOCKET_ID);
+
+	return socket_id;
+}
+
+/**
+ * smuio_v13_0_3_get_pkg_type - query package type set by MP1/bootcode
+ *
+ * @adev: amdgpu device pointer
+ *
+ * Returns package type
+ */
+
+static enum amdgpu_pkg_type smuio_v13_0_3_get_pkg_type(struct amdgpu_device *adev)
+{
+	enum amdgpu_pkg_type pkg_type;
+	u32 data;
+
+	data = RREG32_SOC15(SMUIO, 0, regSMUIO_MCM_CONFIG);
+	data = REG_GET_FIELD(data, SMUIO_MCM_CONFIG, PKG_TYPE);
+	/* pkg_type[4:0]
+	 *
+	 * bit 1 == 1 APU form factor
+	 *
+	 * b0100 - b1111 - Reserved
+	 */
+	switch (data & PKG_TYPE_MASK) {
+	case 0x0:
+		pkg_type = AMDGPU_PKG_TYPE_CEM;
+		break;
+	case 0x1:
+		pkg_type = AMDGPU_PKG_TYPE_OAM;
+		break;
+	case 0x2:
+		pkg_type = AMDGPU_PKG_TYPE_APU;
+		break;
+	default:
+		pkg_type = AMDGPU_PKG_TYPE_UNKNOWN;
+		break;
+	}
+
+	return pkg_type;
+}
+
+
+const struct amdgpu_smuio_funcs smuio_v13_0_3_funcs = {
+	.get_die_id = smuio_v13_0_3_get_die_id,
+	.get_socket_id = smuio_v13_0_3_get_socket_id,
+	.get_pkg_type = smuio_v13_0_3_get_pkg_type,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/smuio_v13_0_3.h b/drivers/gpu/drm/amd/amdgpu/smuio_v13_0_3.h
new file mode 100644
index 000000000000..795f66c5a58b
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/smuio_v13_0_3.h
@@ -0,0 +1,30 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __SMUIO_V13_0_3_H__
+#define __SMUIO_V13_0_3_H__
+
+#include "soc15_common.h"
+
+extern const struct amdgpu_smuio_funcs smuio_v13_0_3_funcs;
+
+#endif /* __SMUIO_V13_0_3_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/smuio_v14_0_2.c b/drivers/gpu/drm/amd/amdgpu/smuio_v14_0_2.c
new file mode 100644
index 000000000000..2a51a70d4846
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/smuio_v14_0_2.c
@@ -0,0 +1,62 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "amdgpu.h"
+#include "smuio_v14_0_2.h"
+#include "smuio/smuio_14_0_2_offset.h"
+#include "smuio/smuio_14_0_2_sh_mask.h"
+#include <linux/preempt.h>
+
+static u32 smuio_v14_0_2_get_rom_index_offset(struct amdgpu_device *adev)
+{
+	return SOC15_REG_OFFSET(SMUIO, 0, regROM_INDEX);
+}
+
+static u32 smuio_v14_0_2_get_rom_data_offset(struct amdgpu_device *adev)
+{
+	return SOC15_REG_OFFSET(SMUIO, 0, regROM_DATA);
+}
+
+static u64 smuio_v14_0_2_get_gpu_clock_counter(struct amdgpu_device *adev)
+{
+	u64 clock;
+	u64 clock_counter_lo, clock_counter_hi_pre, clock_counter_hi_after;
+
+	preempt_disable();
+	clock_counter_hi_pre = (u64)RREG32_SOC15(SMUIO, 0, regGOLDEN_TSC_COUNT_UPPER);
+	clock_counter_lo = (u64)RREG32_SOC15(SMUIO, 0, regGOLDEN_TSC_COUNT_LOWER);
+	/* the clock counter may be udpated during polling the counters */
+	clock_counter_hi_after = (u64)RREG32_SOC15(SMUIO, 0, regGOLDEN_TSC_COUNT_UPPER);
+	if (clock_counter_hi_pre != clock_counter_hi_after)
+		clock_counter_lo = (u64)RREG32_SOC15(SMUIO, 0, regGOLDEN_TSC_COUNT_LOWER);
+	preempt_enable();
+
+	clock = clock_counter_lo | (clock_counter_hi_after << 32ULL);
+
+	return clock;
+}
+
+const struct amdgpu_smuio_funcs smuio_v14_0_2_funcs = {
+	.get_rom_index_offset = smuio_v14_0_2_get_rom_index_offset,
+	.get_rom_data_offset = smuio_v14_0_2_get_rom_data_offset,
+	.get_gpu_clock_counter = smuio_v14_0_2_get_gpu_clock_counter,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/smuio_v14_0_2.h b/drivers/gpu/drm/amd/amdgpu/smuio_v14_0_2.h
new file mode 100644
index 000000000000..6e617f832d90
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/smuio_v14_0_2.h
@@ -0,0 +1,30 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __SMUIO_V14_0_2_H__
+#define __SMUIO_V14_0_2_H__
+
+#include "soc15_common.h"
+
+extern const struct amdgpu_smuio_funcs smuio_v14_0_2_funcs;
+
+#endif /* __SMUIO_V14_0_2_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/smuio_v9_0.c b/drivers/gpu/drm/amd/amdgpu/smuio_v9_0.c
index e4e30b9d481b..c04fdd2d5b38 100644
--- a/drivers/gpu/drm/amd/amdgpu/smuio_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/smuio_v9_0.c
@@ -60,7 +60,7 @@ static void smuio_v9_0_get_clock_gating_state(struct amdgpu_device *adev, u64 *f
 {
 	u32 data;
 
-	/* CGTT_ROM_CLK_CTRL0 is not availabe for APUs */
+	/* CGTT_ROM_CLK_CTRL0 is not available for APUs */
 	if (adev->flags & AMD_IS_APU)
 		return;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c
index bc5dd80f10c1..42f5d9c0e3af 100644
--- a/drivers/gpu/drm/amd/amdgpu/soc15.c
+++ b/drivers/gpu/drm/amd/amdgpu/soc15.c
@@ -28,7 +28,6 @@
 #include <drm/amdgpu_drm.h>
 
 #include "amdgpu.h"
-#include "amdgpu_atombios.h"
 #include "amdgpu_ih.h"
 #include "amdgpu_uvd.h"
 #include "amdgpu_vce.h"
@@ -90,8 +89,8 @@ static const struct amd_ip_funcs soc15_common_ip_funcs;
 /* Vega, Raven, Arcturus */
 static const struct amdgpu_video_codec_info vega_video_codecs_encode_array[] =
 {
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 2304, 0)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 4096, 2304, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 4096, 4096, 0)},
 };
 
 static const struct amdgpu_video_codecs vega_video_codecs_encode =
@@ -103,12 +102,11 @@ static const struct amdgpu_video_codecs vega_video_codecs_encode =
 /* Vega */
 static const struct amdgpu_video_codec_info vega_video_codecs_decode_array[] =
 {
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 1920, 1088, 3)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 1920, 1088, 5)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 1920, 1088, 4)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 4096, 4096, 186)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)},
 };
 
 static const struct amdgpu_video_codecs vega_video_codecs_decode =
@@ -120,12 +118,12 @@ static const struct amdgpu_video_codecs vega_video_codecs_decode =
 /* Raven */
 static const struct amdgpu_video_codec_info rv_video_codecs_decode_array[] =
 {
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 1920, 1088, 3)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 1920, 1088, 5)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 1920, 1088, 4)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 4096, 4096, 186)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 8192, 8192, 0)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 4096, 4096, 0)},
 };
 
@@ -138,12 +136,12 @@ static const struct amdgpu_video_codecs rv_video_codecs_decode =
 /* Renoir, Arcturus */
 static const struct amdgpu_video_codec_info rn_video_codecs_decode_array[] =
 {
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 1920, 1088, 3)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 1920, 1088, 5)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 1920, 1088, 4)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 16384, 16384, 0)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)},
 };
 
@@ -153,11 +151,47 @@ static const struct amdgpu_video_codecs rn_video_codecs_decode =
 	.codec_array = rn_video_codecs_decode_array,
 };
 
+static const struct amdgpu_video_codec_info vcn_4_0_3_video_codecs_decode_array[] = {
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 16384, 16384, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_AV1, 8192, 4352, 0)},
+};
+
+static const struct amdgpu_video_codecs vcn_4_0_3_video_codecs_decode = {
+	.codec_count = ARRAY_SIZE(vcn_4_0_3_video_codecs_decode_array),
+	.codec_array = vcn_4_0_3_video_codecs_decode_array,
+};
+
+static const struct amdgpu_video_codecs vcn_4_0_3_video_codecs_encode = {
+	.codec_count = 0,
+	.codec_array = NULL,
+};
+
+static const struct amdgpu_video_codecs vcn_5_0_1_video_codecs_encode_vcn0 = {
+	.codec_count = 0,
+	.codec_array = NULL,
+};
+
+static const struct amdgpu_video_codec_info vcn_5_0_1_video_codecs_decode_array_vcn0[] = {
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 16384, 16384, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_AV1, 8192, 4352, 0)},
+};
+
+static const struct amdgpu_video_codecs vcn_5_0_1_video_codecs_decode_vcn0 = {
+	.codec_count = ARRAY_SIZE(vcn_5_0_1_video_codecs_decode_array_vcn0),
+	.codec_array = vcn_5_0_1_video_codecs_decode_array_vcn0,
+};
+
 static int soc15_query_video_codecs(struct amdgpu_device *adev, bool encode,
 				    const struct amdgpu_video_codecs **codecs)
 {
-	if (adev->ip_versions[VCE_HWIP][0]) {
-		switch (adev->ip_versions[VCE_HWIP][0]) {
+	if (amdgpu_ip_version(adev, VCE_HWIP, 0)) {
+		switch (amdgpu_ip_version(adev, VCE_HWIP, 0)) {
 		case IP_VERSION(4, 0, 0):
 		case IP_VERSION(4, 1, 0):
 			if (encode)
@@ -169,7 +203,7 @@ static int soc15_query_video_codecs(struct amdgpu_device *adev, bool encode,
 			return -EINVAL;
 		}
 	} else {
-		switch (adev->ip_versions[UVD_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, UVD_HWIP, 0)) {
 		case IP_VERSION(1, 0, 0):
 		case IP_VERSION(1, 0, 1):
 			if (encode)
@@ -185,6 +219,18 @@ static int soc15_query_video_codecs(struct amdgpu_device *adev, bool encode,
 			else
 				*codecs = &rn_video_codecs_decode;
 			return 0;
+		case IP_VERSION(4, 0, 3):
+			if (encode)
+				*codecs = &vcn_4_0_3_video_codecs_encode;
+			else
+				*codecs = &vcn_4_0_3_video_codecs_decode;
+			return 0;
+		case IP_VERSION(5, 0, 1):
+			if (encode)
+				*codecs = &vcn_5_0_1_video_codecs_encode_vcn0;
+			else
+				*codecs = &vcn_5_0_1_video_codecs_decode_vcn0;
+			return 0;
 		default:
 			return -EINVAL;
 		}
@@ -300,18 +346,22 @@ static u32 soc15_get_xclk(struct amdgpu_device *adev)
 {
 	u32 reference_clock = adev->clock.spll.reference_freq;
 
-	if (adev->ip_versions[MP1_HWIP][0] == IP_VERSION(12, 0, 0) ||
-	    adev->ip_versions[MP1_HWIP][0] == IP_VERSION(12, 0, 1) ||
-	    adev->ip_versions[MP1_HWIP][0] == IP_VERSION(10, 0, 0) ||
-	    adev->ip_versions[MP1_HWIP][0] == IP_VERSION(10, 0, 1))
+	if (amdgpu_ip_version(adev, MP1_HWIP, 0) == IP_VERSION(12, 0, 0) ||
+	    amdgpu_ip_version(adev, MP1_HWIP, 0) == IP_VERSION(12, 0, 1) ||
+	    amdgpu_ip_version(adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 6) ||
+	    amdgpu_ip_version(adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 12) ||
+	    amdgpu_ip_version(adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 14))
 		return 10000;
+	if (amdgpu_ip_version(adev, MP1_HWIP, 0) == IP_VERSION(10, 0, 0) ||
+	    amdgpu_ip_version(adev, MP1_HWIP, 0) == IP_VERSION(10, 0, 1))
+		return reference_clock / 4;
 
 	return reference_clock;
 }
 
 
 void soc15_grbm_select(struct amdgpu_device *adev,
-		     u32 me, u32 pipe, u32 queue, u32 vmid)
+		     u32 me, u32 pipe, u32 queue, u32 vmid, int xcc_id)
 {
 	u32 grbm_gfx_cntl = 0;
 	grbm_gfx_cntl = REG_SET_FIELD(grbm_gfx_cntl, GRBM_GFX_CNTL, PIPEID, pipe);
@@ -319,12 +369,7 @@ void soc15_grbm_select(struct amdgpu_device *adev,
 	grbm_gfx_cntl = REG_SET_FIELD(grbm_gfx_cntl, GRBM_GFX_CNTL, VMID, vmid);
 	grbm_gfx_cntl = REG_SET_FIELD(grbm_gfx_cntl, GRBM_GFX_CNTL, QUEUEID, queue);
 
-	WREG32_SOC15_RLC_SHADOW(GC, 0, mmGRBM_GFX_CNTL, grbm_gfx_cntl);
-}
-
-static void soc15_vga_set_state(struct amdgpu_device *adev, bool state)
-{
-	/* todo */
+	WREG32_SOC15_RLC_SHADOW(GC, xcc_id, mmGRBM_GFX_CNTL, grbm_gfx_cntl);
 }
 
 static bool soc15_read_disabled_bios(struct amdgpu_device *adev)
@@ -363,12 +408,12 @@ static uint32_t soc15_read_indexed_register(struct amdgpu_device *adev, u32 se_n
 
 	mutex_lock(&adev->grbm_idx_mutex);
 	if (se_num != 0xffffffff || sh_num != 0xffffffff)
-		amdgpu_gfx_select_se_sh(adev, se_num, sh_num, 0xffffffff);
+		amdgpu_gfx_select_se_sh(adev, se_num, sh_num, 0xffffffff, 0);
 
 	val = RREG32(reg_offset);
 
 	if (se_num != 0xffffffff || sh_num != 0xffffffff)
-		amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+		amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 	return val;
 }
@@ -481,7 +526,7 @@ static int soc15_asic_baco_reset(struct amdgpu_device *adev)
 static enum amd_reset_method
 soc15_asic_reset_method(struct amdgpu_device *adev)
 {
-	bool baco_reset = false;
+	int baco_reset = 0;
 	bool connected_to_cpu = false;
 	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 
@@ -502,7 +547,7 @@ soc15_asic_reset_method(struct amdgpu_device *adev)
 		dev_warn(adev->dev, "Specified reset method:%d isn't supported, using AUTO instead.\n",
 				  amdgpu_reset_method);
 
-	switch (adev->ip_versions[MP1_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
 	case IP_VERSION(10, 0, 0):
 	case IP_VERSION(10, 0, 1):
 	case IP_VERSION(12, 0, 0):
@@ -519,7 +564,7 @@ soc15_asic_reset_method(struct amdgpu_device *adev)
 			 */
 			if (ras && adev->ras_enabled &&
 			    adev->pm.fw_version <= 0x283400)
-				baco_reset = false;
+				baco_reset = 0;
 		} else {
 			baco_reset = amdgpu_dpm_is_baco_supported(adev);
 		}
@@ -532,6 +577,21 @@ soc15_asic_reset_method(struct amdgpu_device *adev)
 		if (connected_to_cpu)
 			return AMD_RESET_METHOD_MODE2;
 		break;
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 14):
+	case IP_VERSION(13, 0, 12):
+		/* Use gpu_recovery param to target a reset method.
+		 * Enable triggering of GPU reset only if specified
+		 * by module parameter.
+		 */
+		if (adev->pcie_reset_ctx.in_link_reset)
+			return AMD_RESET_METHOD_LINK;
+		if (amdgpu_gpu_recovery == 4 || amdgpu_gpu_recovery == 5)
+			return AMD_RESET_METHOD_MODE2;
+		else if (!(adev->flags & AMD_IS_APU))
+			return AMD_RESET_METHOD_MODE1;
+		else
+			return AMD_RESET_METHOD_MODE2;
 	default:
 		break;
 	}
@@ -542,13 +602,36 @@ soc15_asic_reset_method(struct amdgpu_device *adev)
 		return AMD_RESET_METHOD_MODE1;
 }
 
+static bool soc15_need_reset_on_resume(struct amdgpu_device *adev)
+{
+	/* Will reset for the following suspend abort cases.
+	 * 1) S3 suspend aborted in the normal S3 suspend
+	 * 2) S3 suspend aborted in performing pm core test.
+	 */
+	if (adev->in_s3 && !pm_resume_via_firmware())
+		return true;
+	else
+		return false;
+}
+
 static int soc15_asic_reset(struct amdgpu_device *adev)
 {
 	/* original raven doesn't have full asic reset */
+	/* On the latest Raven, the GPU reset can be performed
+	 * successfully. So now, temporarily enable it for the
+	 * S3 suspend abort case.
+	 */
+
+	if ((adev->apu_flags & AMD_APU_IS_PICASSO ||
+			!(adev->apu_flags & AMD_APU_IS_RAVEN)) &&
+			soc15_need_reset_on_resume(adev))
+		goto asic_reset;
+
 	if ((adev->apu_flags & AMD_APU_IS_RAVEN) ||
-	    (adev->apu_flags & AMD_APU_IS_RAVEN2))
+			(adev->apu_flags & AMD_APU_IS_RAVEN2))
 		return 0;
 
+asic_reset:
 	switch (soc15_asic_reset_method(adev)) {
 	case AMD_RESET_METHOD_PCI:
 		dev_info(adev->dev, "PCI reset\n");
@@ -559,27 +642,30 @@ static int soc15_asic_reset(struct amdgpu_device *adev)
 	case AMD_RESET_METHOD_MODE2:
 		dev_info(adev->dev, "MODE2 reset\n");
 		return amdgpu_dpm_mode2_reset(adev);
+	case AMD_RESET_METHOD_LINK:
+		dev_info(adev->dev, "Link reset\n");
+		return amdgpu_device_link_reset(adev);
 	default:
 		dev_info(adev->dev, "MODE1 reset\n");
 		return amdgpu_device_mode1_reset(adev);
 	}
 }
 
-static bool soc15_supports_baco(struct amdgpu_device *adev)
+static int soc15_supports_baco(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[MP1_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
 	case IP_VERSION(9, 0, 0):
 	case IP_VERSION(11, 0, 2):
 		if (adev->asic_type == CHIP_VEGA20) {
 			if (adev->psp.sos.fw_version >= 0x80067)
 				return amdgpu_dpm_is_baco_supported(adev);
-			return false;
+			return 0;
 		} else {
 			return amdgpu_dpm_is_baco_supported(adev);
 		}
 		break;
 	default:
-		return false;
+		return 0;
 	}
 }
 
@@ -614,18 +700,10 @@ static void soc15_program_aspm(struct amdgpu_device *adev)
 	if (!amdgpu_device_should_use_aspm(adev))
 		return;
 
-	if (!(adev->flags & AMD_IS_APU) &&
-	    (adev->nbio.funcs->program_aspm))
+	if (adev->nbio.funcs->program_aspm)
 		adev->nbio.funcs->program_aspm(adev);
 }
 
-static void soc15_enable_doorbell_aperture(struct amdgpu_device *adev,
-					   bool enable)
-{
-	adev->nbio.funcs->enable_doorbell_aperture(adev, enable);
-	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, enable);
-}
-
 const struct amdgpu_ip_block_version vega10_common_ip_block =
 {
 	.type = AMD_IP_BLOCK_TYPE_COMMON,
@@ -663,7 +741,6 @@ static void soc15_reg_base_init(struct amdgpu_device *adev)
 void soc15_set_virt_ops(struct amdgpu_device *adev)
 {
 	adev->virt.ops = &xgpu_ai_virt_ops;
-
 	/* init soc15 reg base early enough so we can
 	 * request request full access for sriov before
 	 * set_ip_blocks. */
@@ -776,10 +853,10 @@ static bool soc15_need_reset_on_init(struct amdgpu_device *adev)
 {
 	u32 sol_reg;
 
-	/* CP hangs in IGT reloading test on RN, reset to WA */
-	if (adev->asic_type == CHIP_RENOIR)
+	if (amdgpu_gmc_need_reset_on_init(adev))
+		return true;
+	if (amdgpu_psp_tos_reload_needed(adev))
 		return true;
-
 	/* Just return false for soc15 GPUs.  Reset does not seem to
 	 * be necessary.
 	 */
@@ -823,7 +900,6 @@ static const struct amdgpu_asic_funcs soc15_asic_funcs =
 	.read_register = &soc15_read_register,
 	.reset = &soc15_asic_reset,
 	.reset_method = &soc15_asic_reset_method,
-	.set_vga_state = &soc15_vga_set_state,
 	.get_xclk = &soc15_get_xclk,
 	.set_uvd_clocks = &soc15_set_uvd_clocks,
 	.set_vce_clocks = &soc15_set_vce_clocks,
@@ -845,7 +921,6 @@ static const struct amdgpu_asic_funcs vega20_asic_funcs =
 	.read_register = &soc15_read_register,
 	.reset = &soc15_asic_reset,
 	.reset_method = &soc15_asic_reset_method,
-	.set_vga_state = &soc15_vga_set_state,
 	.get_xclk = &soc15_get_xclk,
 	.set_uvd_clocks = &soc15_set_uvd_clocks,
 	.set_vce_clocks = &soc15_set_vce_clocks,
@@ -860,21 +935,43 @@ static const struct amdgpu_asic_funcs vega20_asic_funcs =
 	.query_video_codecs = &soc15_query_video_codecs,
 };
 
-static int soc15_common_early_init(void *handle)
+static const struct amdgpu_asic_funcs aqua_vanjaram_asic_funcs =
 {
-#define MMIO_REG_HOLE_OFFSET (0x80000 - PAGE_SIZE)
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	.read_disabled_bios = &soc15_read_disabled_bios,
+	.read_bios_from_rom = &amdgpu_soc15_read_bios_from_rom,
+	.read_register = &soc15_read_register,
+	.reset = &soc15_asic_reset,
+	.reset_method = &soc15_asic_reset_method,
+	.get_xclk = &soc15_get_xclk,
+	.set_uvd_clocks = &soc15_set_uvd_clocks,
+	.set_vce_clocks = &soc15_set_vce_clocks,
+	.get_config_memsize = &soc15_get_config_memsize,
+	.need_full_reset = &soc15_need_full_reset,
+	.init_doorbell_index = &aqua_vanjaram_doorbell_index_init,
+	.need_reset_on_init = &soc15_need_reset_on_init,
+	.get_pcie_replay_count = &amdgpu_nbio_get_pcie_replay_count,
+	.supports_baco = &soc15_supports_baco,
+	.pre_asic_init = &soc15_pre_asic_init,
+	.query_video_codecs = &soc15_query_video_codecs,
+	.encode_ext_smn_addressing = &aqua_vanjaram_encode_ext_smn_addressing,
+	.get_reg_state = &aqua_vanjaram_get_reg_state,
+};
 
-	if (!amdgpu_sriov_vf(adev)) {
-		adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET;
-		adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET;
-	}
+static int soc15_common_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	adev->nbio.funcs->set_reg_remap(adev);
 	adev->smc_rreg = NULL;
 	adev->smc_wreg = NULL;
 	adev->pcie_rreg = &amdgpu_device_indirect_rreg;
 	adev->pcie_wreg = &amdgpu_device_indirect_wreg;
+	adev->pcie_rreg_ext = &amdgpu_device_indirect_rreg_ext;
+	adev->pcie_wreg_ext = &amdgpu_device_indirect_wreg_ext;
 	adev->pcie_rreg64 = &amdgpu_device_indirect_rreg64;
 	adev->pcie_wreg64 = &amdgpu_device_indirect_wreg64;
+	adev->pcie_rreg64_ext = &amdgpu_device_indirect_rreg64_ext;
+	adev->pcie_wreg64_ext = &amdgpu_device_indirect_wreg64_ext;
 	adev->uvd_ctx_rreg = &soc15_uvd_ctx_rreg;
 	adev->uvd_ctx_wreg = &soc15_uvd_ctx_wreg;
 	adev->didt_rreg = &soc15_didt_rreg;
@@ -889,7 +986,7 @@ static int soc15_common_early_init(void *handle)
 	/* TODO: split the GC and PG flags based on the relevant IP version for which
 	 * they are relevant.
 	 */
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 		adev->asic_funcs = &soc15_asic_funcs;
 		adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG |
@@ -1101,9 +1198,23 @@ static int soc15_common_early_init(void *handle)
 		adev->external_rev_id = adev->rev_id + 0x3c;
 		break;
 	case IP_VERSION(9, 4, 3):
-		adev->asic_funcs = &vega20_asic_funcs;
-		adev->cg_flags = 0;
-		adev->pg_flags = 0;
+	case IP_VERSION(9, 4, 4):
+	case IP_VERSION(9, 5, 0):
+		adev->asic_funcs = &aqua_vanjaram_asic_funcs;
+		adev->cg_flags =
+			AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_CGCG |
+			AMD_CG_SUPPORT_GFX_CGLS | AMD_CG_SUPPORT_SDMA_MGCG |
+			AMD_CG_SUPPORT_GFX_FGCG | AMD_CG_SUPPORT_REPEATER_FGCG |
+			AMD_CG_SUPPORT_VCN_MGCG | AMD_CG_SUPPORT_JPEG_MGCG |
+			AMD_CG_SUPPORT_IH_CG;
+		adev->pg_flags =
+			AMD_PG_SUPPORT_VCN |
+			AMD_PG_SUPPORT_VCN_DPG |
+			AMD_PG_SUPPORT_JPEG;
+		/*TODO: need a new external_rev_id for GC 9.4.4? */
+		adev->external_rev_id = adev->rev_id + 0x46;
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 5, 0))
+			adev->external_rev_id = adev->rev_id + 0x50;
 		break;
 	default:
 		/* FIXME: not supported yet */
@@ -1118,19 +1229,24 @@ static int soc15_common_early_init(void *handle)
 	return 0;
 }
 
-static int soc15_common_late_init(void *handle)
+static int soc15_common_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		xgpu_ai_mailbox_get_irq(adev);
 
+	/* Enable selfring doorbell aperture late because doorbell BAR
+	 * aperture will change if resize BAR successfully in gmc sw_init.
+	 */
+	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, true);
+
 	return 0;
 }
 
-static int soc15_common_sw_init(void *handle)
+static int soc15_common_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		xgpu_ai_mailbox_add_irq_id(adev);
@@ -1142,9 +1258,9 @@ static int soc15_common_sw_init(void *handle)
 	return 0;
 }
 
-static int soc15_common_sw_fini(void *handle)
+static int soc15_common_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (adev->df.funcs &&
 	    adev->df.funcs->sw_fini)
@@ -1166,9 +1282,9 @@ static void soc15_sdma_doorbell_range_init(struct amdgpu_device *adev)
 	}
 }
 
-static int soc15_common_hw_init(void *handle)
+static int soc15_common_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* enable aspm */
 	soc15_program_aspm(adev);
@@ -1182,7 +1298,8 @@ static int soc15_common_hw_init(void *handle)
 		adev->nbio.funcs->remap_hdp_registers(adev);
 
 	/* enable the doorbell aperture */
-	soc15_enable_doorbell_aperture(adev, true);
+	adev->nbio.funcs->enable_doorbell_aperture(adev, true);
+
 	/* HW doorbell routing policy: doorbell writing not
 	 * in SDMA/IH/MM/ACV range will be routed to CP. So
 	 * we need to init SDMA doorbell range prior
@@ -1194,16 +1311,28 @@ static int soc15_common_hw_init(void *handle)
 	return 0;
 }
 
-static int soc15_common_hw_fini(void *handle)
+static int soc15_common_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	/* Disable the doorbell aperture and selfring doorbell aperture
+	 * separately in hw_fini because soc15_enable_doorbell_aperture
+	 * has been removed and there is no need to delay disabling
+	 * selfring doorbell.
+	 */
+	adev->nbio.funcs->enable_doorbell_aperture(adev, false);
+	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, false);
 
-	/* disable the doorbell aperture */
-	soc15_enable_doorbell_aperture(adev, false);
 	if (amdgpu_sriov_vf(adev))
 		xgpu_ai_mailbox_put_irq(adev);
 
-	if (adev->nbio.ras_if &&
+	/*
+	 * For minimal init, late_init is not called, hence RAS irqs are not
+	 * enabled.
+	 */
+	if ((!amdgpu_sriov_vf(adev)) &&
+	    (adev->init_lvl->level != AMDGPU_INIT_LEVEL_MINIMAL_XGMI) &&
+	    adev->nbio.ras_if &&
 	    amdgpu_ras_is_supported(adev, adev->nbio.ras_if->block)) {
 		if (adev->nbio.ras &&
 		    adev->nbio.ras->init_ras_controller_interrupt)
@@ -1216,35 +1345,27 @@ static int soc15_common_hw_fini(void *handle)
 	return 0;
 }
 
-static int soc15_common_suspend(void *handle)
+static int soc15_common_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return soc15_common_hw_fini(adev);
+	return soc15_common_hw_fini(ip_block);
 }
 
-static int soc15_common_resume(void *handle)
+static int soc15_common_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	return soc15_common_hw_init(adev);
+	if (soc15_need_reset_on_resume(adev)) {
+		dev_info(adev->dev, "S3 suspend abort case, let's reset ASIC.\n");
+		soc15_asic_reset(adev);
+	}
+	return soc15_common_hw_init(ip_block);
 }
 
-static bool soc15_common_is_idle(void *handle)
+static bool soc15_common_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	return true;
 }
 
-static int soc15_common_wait_for_idle(void *handle)
-{
-	return 0;
-}
-
-static int soc15_common_soft_reset(void *handle)
-{
-	return 0;
-}
-
 static void soc15_update_drm_clock_gating(struct amdgpu_device *adev, bool enable)
 {
 	uint32_t def, data;
@@ -1289,15 +1410,15 @@ static void soc15_update_drm_light_sleep(struct amdgpu_device *adev, bool enable
 		WREG32(SOC15_REG_OFFSET(MP0, 0, mmMP0_MISC_LIGHT_SLEEP_CTRL), data);
 }
 
-static int soc15_common_set_clockgating_state(void *handle,
+static int soc15_common_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					    enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[NBIO_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, NBIO_HWIP, 0)) {
 	case IP_VERSION(6, 1, 0):
 	case IP_VERSION(6, 2, 0):
 	case IP_VERSION(7, 4, 0):
@@ -1341,20 +1462,24 @@ static int soc15_common_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static void soc15_common_get_clockgating_state(void *handle, u64 *flags)
+static void soc15_common_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	if (amdgpu_sriov_vf(adev))
 		*flags = 0;
 
-	adev->nbio.funcs->get_clockgating_state(adev, flags);
-
-	adev->hdp.funcs->get_clock_gating_state(adev, flags);
+	if (adev->nbio.funcs && adev->nbio.funcs->get_clockgating_state)
+		adev->nbio.funcs->get_clockgating_state(adev, flags);
 
-	if (adev->ip_versions[MP0_HWIP][0] != IP_VERSION(13, 0, 2)) {
+	if (adev->hdp.funcs && adev->hdp.funcs->get_clock_gating_state)
+		adev->hdp.funcs->get_clock_gating_state(adev, flags);
 
+	if ((amdgpu_ip_version(adev, MP0_HWIP, 0) != IP_VERSION(13, 0, 2)) &&
+	    (amdgpu_ip_version(adev, MP0_HWIP, 0) != IP_VERSION(13, 0, 6)) &&
+	    (amdgpu_ip_version(adev, MP0_HWIP, 0) != IP_VERSION(13, 0, 12)) &&
+	    (amdgpu_ip_version(adev, MP0_HWIP, 0) != IP_VERSION(13, 0, 14))) {
 		/* AMD_CG_SUPPORT_DRM_MGCG */
 		data = RREG32(SOC15_REG_OFFSET(MP0, 0, mmMP0_MISC_CGTT_CTRL0));
 		if (!(data & 0x01000000))
@@ -1367,12 +1492,14 @@ static void soc15_common_get_clockgating_state(void *handle, u64 *flags)
 	}
 
 	/* AMD_CG_SUPPORT_ROM_MGCG */
-	adev->smuio.funcs->get_clock_gating_state(adev, flags);
+	if (adev->smuio.funcs && adev->smuio.funcs->get_clock_gating_state)
+		adev->smuio.funcs->get_clock_gating_state(adev, flags);
 
-	adev->df.funcs->get_clockgating_state(adev, flags);
+	if (adev->df.funcs && adev->df.funcs->get_clockgating_state)
+		adev->df.funcs->get_clockgating_state(adev, flags);
 }
 
-static int soc15_common_set_powergating_state(void *handle,
+static int soc15_common_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					    enum amd_powergating_state state)
 {
 	/* todo */
@@ -1390,8 +1517,6 @@ static const struct amd_ip_funcs soc15_common_ip_funcs = {
 	.suspend = soc15_common_suspend,
 	.resume = soc15_common_resume,
 	.is_idle = soc15_common_is_idle,
-	.wait_for_idle = soc15_common_wait_for_idle,
-	.soft_reset = soc15_common_soft_reset,
 	.set_clockgating_state = soc15_common_set_clockgating_state,
 	.set_powergating_state = soc15_common_set_powergating_state,
 	.get_clockgating_state= soc15_common_get_clockgating_state,
diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.h b/drivers/gpu/drm/amd/amdgpu/soc15.h
index efc2a253e8db..c8ac11a9cdef 100644
--- a/drivers/gpu/drm/amd/amdgpu/soc15.h
+++ b/drivers/gpu/drm/amd/amdgpu/soc15.h
@@ -27,6 +27,7 @@
 #include "nbio_v6_1.h"
 #include "nbio_v7_0.h"
 #include "nbio_v7_4.h"
+#include "amdgpu_reg_state.h"
 
 extern const struct amdgpu_ip_block_version vega10_common_ip_block;
 
@@ -87,9 +88,15 @@ struct soc15_ras_field_entry {
 };
 
 #define SOC15_REG_ENTRY(ip, inst, reg)	ip##_HWIP, inst, reg##_BASE_IDX, reg
+#define SOC15_REG_ENTRY_STR(ip, inst, reg) \
+	{ ip##_HWIP, inst, reg##_BASE_IDX, reg, #reg }
 
 #define SOC15_REG_ENTRY_OFFSET(entry)	(adev->reg_offset[entry.hwip][entry.inst][entry.seg] + entry.reg_offset)
 
+/* Over ride the instance id */
+#define SOC15_REG_ENTRY_OFFSET_INST(entry, inst) \
+	(adev->reg_offset[entry.hwip][inst][entry.seg] + entry.reg_offset)
+
 #define SOC15_REG_GOLDEN_VALUE(ip, inst, reg, and_mask, or_mask) \
 	{ ip##_HWIP, inst, reg##_BASE_IDX, reg, and_mask, or_mask }
 
@@ -100,7 +107,7 @@ struct soc15_ras_field_entry {
 #define SOC15_RAS_REG_FIELD_VAL(val, entry, field) SOC15_REG_FIELD_VAL((val), (entry).field##_count_mask, (entry).field##_count_shift)
 
 void soc15_grbm_select(struct amdgpu_device *adev,
-		    u32 me, u32 pipe, u32 queue, u32 vmid);
+		    u32 me, u32 pipe, u32 queue, u32 vmid, int xcc_id);
 void soc15_set_virt_ops(struct amdgpu_device *adev);
 
 void soc15_program_register_sequence(struct amdgpu_device *adev,
@@ -111,7 +118,13 @@ int vega10_reg_base_init(struct amdgpu_device *adev);
 int vega20_reg_base_init(struct amdgpu_device *adev);
 int arct_reg_base_init(struct amdgpu_device *adev);
 int aldebaran_reg_base_init(struct amdgpu_device *adev);
+u64 aqua_vanjaram_encode_ext_smn_addressing(int ext_id);
+int aqua_vanjaram_init_soc_config(struct amdgpu_device *adev);
+ssize_t aqua_vanjaram_get_reg_state(struct amdgpu_device *adev,
+				    enum amdgpu_reg_state reg_state, void *buf,
+				    size_t max_size);
 
 void vega10_doorbell_index_init(struct amdgpu_device *adev);
 void vega20_doorbell_index_init(struct amdgpu_device *adev);
+void aqua_vanjaram_doorbell_index_init(struct amdgpu_device *adev);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/soc15_common.h b/drivers/gpu/drm/amd/amdgpu/soc15_common.h
index 9fefd403e14f..242b24f73c17 100644
--- a/drivers/gpu/drm/amd/amdgpu/soc15_common.h
+++ b/drivers/gpu/drm/amd/amdgpu/soc15_common.h
@@ -24,105 +24,100 @@
 #ifndef __SOC15_COMMON_H__
 #define __SOC15_COMMON_H__
 
+/* GET_INST returns the physical instance corresponding to a logical instance */
+#define GET_INST(ip, inst) \
+	(adev->ip_map.logical_to_dev_inst ? \
+	adev->ip_map.logical_to_dev_inst(adev, ip##_HWIP, inst) : inst)
+#define GET_MASK(ip, mask) \
+	(adev->ip_map.logical_to_dev_mask ? \
+	adev->ip_map.logical_to_dev_mask(adev, ip##_HWIP, mask) : mask)
+
 /* Register Access Macros */
 #define SOC15_REG_OFFSET(ip, inst, reg)	(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg)
+#define SOC15_REG_OFFSET1(ip, inst, reg, offset) \
+	(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + (reg)+(offset))
 
-#define __WREG32_SOC15_RLC__(reg, value, flag, hwip) \
+#define __WREG32_SOC15_RLC__(reg, value, flag, hwip, inst) \
 	((amdgpu_sriov_vf(adev) && adev->gfx.rlc.funcs && adev->gfx.rlc.rlcg_reg_access_supported) ? \
-	 amdgpu_sriov_wreg(adev, reg, value, flag, hwip) : \
+	 amdgpu_sriov_wreg(adev, reg, value, flag, hwip, inst) : \
 	 WREG32(reg, value))
 
-#define __RREG32_SOC15_RLC__(reg, flag, hwip) \
+#define __RREG32_SOC15_RLC__(reg, flag, hwip, inst) \
 	((amdgpu_sriov_vf(adev) && adev->gfx.rlc.funcs && adev->gfx.rlc.rlcg_reg_access_supported) ? \
-	 amdgpu_sriov_rreg(adev, reg, flag, hwip) : \
+	 amdgpu_sriov_rreg(adev, reg, flag, hwip, inst) : \
 	 RREG32(reg))
 
 #define WREG32_FIELD15(ip, idx, reg, field, val)	\
 	 __WREG32_SOC15_RLC__(adev->reg_offset[ip##_HWIP][idx][mm##reg##_BASE_IDX] + mm##reg,	\
 				(__RREG32_SOC15_RLC__( \
 					adev->reg_offset[ip##_HWIP][idx][mm##reg##_BASE_IDX] + mm##reg, \
-					0, ip##_HWIP) & \
+					0, ip##_HWIP, idx) & \
 				~REG_FIELD_MASK(reg, field)) | (val) << REG_FIELD_SHIFT(reg, field), \
-			      0, ip##_HWIP)
+			      0, ip##_HWIP, idx)
 
 #define WREG32_FIELD15_PREREG(ip, idx, reg_name, field, val)        \
 	__WREG32_SOC15_RLC__(adev->reg_offset[ip##_HWIP][idx][reg##reg_name##_BASE_IDX] + reg##reg_name,   \
 			(__RREG32_SOC15_RLC__( \
 					adev->reg_offset[ip##_HWIP][idx][reg##reg_name##_BASE_IDX] + reg##reg_name, \
-					0, ip##_HWIP) & \
+					0, ip##_HWIP, idx) & \
 					~REG_FIELD_MASK(reg_name, field)) | (val) << REG_FIELD_SHIFT(reg_name, field), \
-			0, ip##_HWIP)
+			0, ip##_HWIP, idx)
 
 #define RREG32_SOC15(ip, inst, reg) \
 	__RREG32_SOC15_RLC__(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg, \
-			 0, ip##_HWIP)
+			 0, ip##_HWIP, inst)
 
-#define RREG32_SOC15_IP(ip, reg) __RREG32_SOC15_RLC__(reg, 0, ip##_HWIP)
+#define RREG32_SOC15_IP(ip, reg) __RREG32_SOC15_RLC__(reg, 0, ip##_HWIP, 0)
 
-#define RREG32_SOC15_IP_NO_KIQ(ip, reg) __RREG32_SOC15_RLC__(reg, AMDGPU_REGS_NO_KIQ, ip##_HWIP)
+#define RREG32_SOC15_IP_NO_KIQ(ip, reg, inst) __RREG32_SOC15_RLC__(reg, AMDGPU_REGS_NO_KIQ, ip##_HWIP, inst)
 
 #define RREG32_SOC15_NO_KIQ(ip, inst, reg) \
 	__RREG32_SOC15_RLC__(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg, \
-			 AMDGPU_REGS_NO_KIQ, ip##_HWIP)
+			 AMDGPU_REGS_NO_KIQ, ip##_HWIP, inst)
 
 #define RREG32_SOC15_OFFSET(ip, inst, reg, offset) \
-	 __RREG32_SOC15_RLC__((adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg) + offset, 0, ip##_HWIP)
+	 __RREG32_SOC15_RLC__((adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + (reg)) + \
+			 (offset), 0, ip##_HWIP, inst)
 
 #define WREG32_SOC15(ip, inst, reg, value) \
 	 __WREG32_SOC15_RLC__((adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg), \
-			  value, 0, ip##_HWIP)
+			  value, 0, ip##_HWIP, inst)
 
 #define WREG32_SOC15_IP(ip, reg, value) \
-	 __WREG32_SOC15_RLC__(reg, value, 0, ip##_HWIP)
+	 __WREG32_SOC15_RLC__(reg, value, 0, ip##_HWIP, 0)
 
-#define WREG32_SOC15_IP_NO_KIQ(ip, reg, value) \
-	 __WREG32_SOC15_RLC__(reg, value, AMDGPU_REGS_NO_KIQ, ip##_HWIP)
+#define WREG32_SOC15_IP_NO_KIQ(ip, reg, value, inst) \
+	 __WREG32_SOC15_RLC__(reg, value, AMDGPU_REGS_NO_KIQ, ip##_HWIP, inst)
 
 #define WREG32_SOC15_NO_KIQ(ip, inst, reg, value) \
 	__WREG32_SOC15_RLC__(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg, \
-			     value, AMDGPU_REGS_NO_KIQ, ip##_HWIP)
+			     value, AMDGPU_REGS_NO_KIQ, ip##_HWIP, inst)
 
 #define WREG32_SOC15_OFFSET(ip, inst, reg, offset, value) \
 	 __WREG32_SOC15_RLC__((adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg) + offset, \
-			  value, 0, ip##_HWIP)
+			  value, 0, ip##_HWIP, inst)
 
-#define SOC15_WAIT_ON_RREG(ip, inst, reg, expected_value, mask) \
-({	int ret = 0;						\
-	do {							\
-		uint32_t old_ = 0;				\
-		uint32_t tmp_ = RREG32(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg); \
-		uint32_t loop = adev->usec_timeout;		\
-		ret = 0;					\
-		while ((tmp_ & (mask)) != (expected_value)) {	\
-			if (old_ != tmp_) {			\
-				loop = adev->usec_timeout;	\
-				old_ = tmp_;			\
-			} else					\
-				udelay(1);			\
-			tmp_ = RREG32(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg); \
-			loop--;					\
-			if (!loop) {				\
-				DRM_WARN("Register(%d) [%s] failed to reach value 0x%08x != 0x%08x\n", \
-					  inst, #reg, (unsigned)expected_value, (unsigned)(tmp_ & (mask))); \
-				ret = -ETIMEDOUT;		\
-				break;				\
-			}					\
-		}						\
-	} while (0);						\
-	ret;							\
-})
+#define SOC15_WAIT_ON_RREG(ip, inst, reg, expected_value, mask)      \
+	amdgpu_device_wait_on_rreg(adev, inst,                       \
+	(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + (reg)), \
+	#reg, expected_value, mask)
+
+#define SOC15_WAIT_ON_RREG_OFFSET(ip, inst, reg, offset, expected_value, mask)  \
+	amdgpu_device_wait_on_rreg(adev, inst,                                  \
+	(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + (reg) + (offset)), \
+	#reg, expected_value, mask)
 
 #define WREG32_RLC(reg, value) \
-	__WREG32_SOC15_RLC__(reg, value, AMDGPU_REGS_RLC, GC_HWIP)
+	__WREG32_SOC15_RLC__(reg, value, AMDGPU_REGS_RLC, GC_HWIP, 0)
 
-#define WREG32_RLC_EX(prefix, reg, value) \
+#define WREG32_RLC_EX(prefix, reg, value, inst) \
 	do {							\
 		if (amdgpu_sriov_fullaccess(adev)) {    \
 			uint32_t i = 0;	\
 			uint32_t retries = 50000;	\
-			uint32_t r0 = adev->reg_offset[GC_HWIP][0][prefix##SCRATCH_REG0_BASE_IDX] + prefix##SCRATCH_REG0;	\
-			uint32_t r1 = adev->reg_offset[GC_HWIP][0][prefix##SCRATCH_REG1_BASE_IDX] + prefix##SCRATCH_REG1;	\
-			uint32_t spare_int = adev->reg_offset[GC_HWIP][0][prefix##RLC_SPARE_INT_BASE_IDX] + prefix##RLC_SPARE_INT;	\
+			uint32_t r0 = adev->reg_offset[GC_HWIP][inst][prefix##SCRATCH_REG0_BASE_IDX] + prefix##SCRATCH_REG0;	\
+			uint32_t r1 = adev->reg_offset[GC_HWIP][inst][prefix##SCRATCH_REG1_BASE_IDX] + prefix##SCRATCH_REG1;	\
+			uint32_t spare_int = adev->reg_offset[GC_HWIP][inst][prefix##RLC_SPARE_INT_BASE_IDX] + prefix##RLC_SPARE_INT;	\
 			WREG32(r0, value);	\
 			WREG32(r1, (reg | 0x80000000));	\
 			WREG32(spare_int, 0x1);	\
@@ -141,26 +136,26 @@
 
 /* shadow the registers in the callback function */
 #define WREG32_SOC15_RLC_SHADOW(ip, inst, reg, value) \
-	__WREG32_SOC15_RLC__((adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg), value, AMDGPU_REGS_RLC, GC_HWIP)
+	__WREG32_SOC15_RLC__((adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg), value, AMDGPU_REGS_RLC, GC_HWIP, inst)
 
 /* for GC only */
 #define RREG32_RLC(reg) \
-	__RREG32_SOC15_RLC__(reg, AMDGPU_REGS_RLC, GC_HWIP)
+	__RREG32_SOC15_RLC__(reg, AMDGPU_REGS_RLC, GC_HWIP, 0)
 
 #define WREG32_RLC_NO_KIQ(reg, value, hwip) \
-	__WREG32_SOC15_RLC__(reg, value, AMDGPU_REGS_NO_KIQ | AMDGPU_REGS_RLC, hwip)
+	__WREG32_SOC15_RLC__(reg, value, AMDGPU_REGS_NO_KIQ | AMDGPU_REGS_RLC, hwip, 0)
 
 #define RREG32_RLC_NO_KIQ(reg, hwip) \
-	__RREG32_SOC15_RLC__(reg, AMDGPU_REGS_NO_KIQ | AMDGPU_REGS_RLC, hwip)
+	__RREG32_SOC15_RLC__(reg, AMDGPU_REGS_NO_KIQ | AMDGPU_REGS_RLC, hwip, 0)
 
 #define WREG32_SOC15_RLC_SHADOW_EX(prefix, ip, inst, reg, value) \
 	do {							\
 		uint32_t target_reg = adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg;\
 		if (amdgpu_sriov_fullaccess(adev)) {    \
-			uint32_t r2 = adev->reg_offset[GC_HWIP][0][prefix##SCRATCH_REG1_BASE_IDX] + prefix##SCRATCH_REG2;	\
-			uint32_t r3 = adev->reg_offset[GC_HWIP][0][prefix##SCRATCH_REG1_BASE_IDX] + prefix##SCRATCH_REG3;	\
-			uint32_t grbm_cntl = adev->reg_offset[GC_HWIP][0][prefix##GRBM_GFX_CNTL_BASE_IDX] + prefix##GRBM_GFX_CNTL;   \
-			uint32_t grbm_idx = adev->reg_offset[GC_HWIP][0][prefix##GRBM_GFX_INDEX_BASE_IDX] + prefix##GRBM_GFX_INDEX;   \
+			uint32_t r2 = adev->reg_offset[GC_HWIP][inst][prefix##SCRATCH_REG1_BASE_IDX] + prefix##SCRATCH_REG2;	\
+			uint32_t r3 = adev->reg_offset[GC_HWIP][inst][prefix##SCRATCH_REG1_BASE_IDX] + prefix##SCRATCH_REG3;	\
+			uint32_t grbm_cntl = adev->reg_offset[GC_HWIP][inst][prefix##GRBM_GFX_CNTL_BASE_IDX] + prefix##GRBM_GFX_CNTL;   \
+			uint32_t grbm_idx = adev->reg_offset[GC_HWIP][inst][prefix##GRBM_GFX_INDEX_BASE_IDX] + prefix##GRBM_GFX_INDEX;   \
 			if (target_reg == grbm_cntl) \
 				WREG32(r2, value);	\
 			else if (target_reg == grbm_idx) \
@@ -172,31 +167,47 @@
 	} while (0)
 
 #define RREG32_SOC15_RLC(ip, inst, reg) \
-	__RREG32_SOC15_RLC__(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg, AMDGPU_REGS_RLC, ip##_HWIP)
+	__RREG32_SOC15_RLC__(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg, AMDGPU_REGS_RLC, ip##_HWIP, inst)
 
 #define WREG32_SOC15_RLC(ip, inst, reg, value) \
 	do {							\
-		uint32_t target_reg = adev->reg_offset[ip##_HWIP][0][reg##_BASE_IDX] + reg;\
-		__WREG32_SOC15_RLC__(target_reg, value, AMDGPU_REGS_RLC, ip##_HWIP); \
+		uint32_t target_reg = adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg;\
+		__WREG32_SOC15_RLC__(target_reg, value, AMDGPU_REGS_RLC, ip##_HWIP, inst); \
 	} while (0)
 
 #define WREG32_SOC15_RLC_EX(prefix, ip, inst, reg, value) \
 	do {							\
-			uint32_t target_reg = adev->reg_offset[GC_HWIP][0][reg##_BASE_IDX] + reg;\
-			WREG32_RLC_EX(prefix, target_reg, value); \
+			uint32_t target_reg = adev->reg_offset[GC_HWIP][inst][reg##_BASE_IDX] + reg;\
+			WREG32_RLC_EX(prefix, target_reg, value, inst); \
 	} while (0)
 
 #define WREG32_FIELD15_RLC(ip, idx, reg, field, val)   \
 	__WREG32_SOC15_RLC__((adev->reg_offset[ip##_HWIP][idx][mm##reg##_BASE_IDX] + mm##reg), \
 			     (__RREG32_SOC15_RLC__(adev->reg_offset[ip##_HWIP][idx][mm##reg##_BASE_IDX] + mm##reg, \
-						   AMDGPU_REGS_RLC, ip##_HWIP) & \
+						   AMDGPU_REGS_RLC, ip##_HWIP, idx) & \
 			      ~REG_FIELD_MASK(reg, field)) | (val) << REG_FIELD_SHIFT(reg, field), \
-			     AMDGPU_REGS_RLC, ip##_HWIP)
+			     AMDGPU_REGS_RLC, ip##_HWIP, idx)
 
 #define WREG32_SOC15_OFFSET_RLC(ip, inst, reg, offset, value) \
-	__WREG32_SOC15_RLC__((adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg) + offset, value, AMDGPU_REGS_RLC, ip##_HWIP)
+	__WREG32_SOC15_RLC__((adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg) + offset, value, AMDGPU_REGS_RLC, ip##_HWIP, inst)
 
 #define RREG32_SOC15_OFFSET_RLC(ip, inst, reg, offset) \
-	__RREG32_SOC15_RLC__((adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg) + offset, AMDGPU_REGS_RLC, ip##_HWIP)
+	__RREG32_SOC15_RLC__((adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg) + offset, AMDGPU_REGS_RLC, ip##_HWIP, inst)
+
+/* inst equals to ext for some IPs */
+#define RREG32_SOC15_EXT(ip, inst, reg, ext) \
+	RREG32_PCIE_EXT((adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg) * 4 \
+			+ adev->asic_funcs->encode_ext_smn_addressing(ext)) \
+
+#define WREG32_SOC15_EXT(ip, inst, reg, ext, value) \
+	WREG32_PCIE_EXT((adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg) * 4 \
+			+ adev->asic_funcs->encode_ext_smn_addressing(ext), \
+			value) \
+
+#define RREG64_MCA(ext, mca_base, idx) \
+	RREG64_PCIE_EXT(adev->asic_funcs->encode_ext_smn_addressing(ext) + mca_base + (idx * 8))
+
+#define WREG64_MCA(ext, mca_base, idx, val) \
+	WREG64_PCIE_EXT(adev->asic_funcs->encode_ext_smn_addressing(ext) + mca_base + (idx * 8), val)
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/soc15d.h b/drivers/gpu/drm/amd/amdgpu/soc15d.h
index 2357ff39323f..cf93fa477674 100644
--- a/drivers/gpu/drm/amd/amdgpu/soc15d.h
+++ b/drivers/gpu/drm/amd/amdgpu/soc15d.h
@@ -76,6 +76,12 @@
 			 ((cond & 0xF) << 24) |				\
 			 ((type & 0xF) << 28))
 
+#define CP_PACKETJ_NOP		0x60000000
+#define CP_PACKETJ_GET_REG(x)  ((x) & 0x3FFFF)
+#define CP_PACKETJ_GET_RES(x)  (((x) >> 18) & 0x3F)
+#define CP_PACKETJ_GET_COND(x) (((x) >> 24) & 0xF)
+#define CP_PACKETJ_GET_TYPE(x) (((x) >> 28) & 0xF)
+
 /* Packet 3 types */
 #define	PACKET3_NOP					0x10
 #define	PACKET3_SET_BASE				0x11
@@ -87,11 +93,25 @@
 #define	PACKET3_DISPATCH_INDIRECT			0x16
 #define	PACKET3_ATOMIC_GDS				0x1D
 #define	PACKET3_ATOMIC_MEM				0x1E
+#define		PACKET3_ATOMIC_MEM__ATOMIC(x) ((((unsigned)(x)) & 0x3F) << 0)
+#define		PACKET3_ATOMIC_MEM__COMMAND(x) ((((unsigned)(x)) & 0xF) << 8)
+#define		PACKET3_ATOMIC_MEM__CACHE_POLICY(x) ((((unsigned)(x)) & 0x3) << 25)
+#define		PACKET3_ATOMIC_MEM__ADDR_LO(x) (((unsigned)(x)) << 0)
+#define		PACKET3_ATOMIC_MEM__ADDR_HI(x) (((unsigned)(x)) << 0)
+#define		PACKET3_ATOMIC_MEM__SRC_DATA_LO(x) (((unsigned)(x)) << 0)
+#define		PACKET3_ATOMIC_MEM__SRC_DATA_HI(x) (((unsigned)(x)) << 0)
+#define		PACKET3_ATOMIC_MEM__CMP_DATA_LO(x) (((unsigned)(x)) << 0)
+#define		PACKET3_ATOMIC_MEM__CMP_DATA_HI(x) (((unsigned)(x)) << 0)
+#define		PACKET3_ATOMIC_MEM__LOOP_INTERVAL(x) ((((unsigned)(x)) & 0x1FFF) << 0)
+#define 	PACKET3_ATOMIC_MEM__COMMAND__SINGLE_PASS_ATOMIC	0
+#define 	PACKET3_ATOMIC_MEM__COMMAND__LOOP_UNTIL_COMPARE_SATISFIED 1
 #define	PACKET3_OCCLUSION_QUERY				0x1F
 #define	PACKET3_SET_PREDICATION				0x20
 #define	PACKET3_REG_RMW					0x21
 #define	PACKET3_COND_EXEC				0x22
 #define	PACKET3_PRED_EXEC				0x23
+#define		PACKET3_PRED_EXEC__EXEC_COUNT(x)	((((unsigned)(x)) & 0x3FFF) << 0)
+#define		PACKET3_PRED_EXEC__VIRTUAL_XCC_ID_SELECT(x)	((((unsigned)(x)) & 0xFF) << 24)
 #define	PACKET3_DRAW_INDIRECT				0x24
 #define	PACKET3_DRAW_INDEX_INDIRECT			0x25
 #define	PACKET3_INDEX_BASE				0x26
@@ -126,6 +146,28 @@
 		 * 1 - pfp
 		 * 2 - ce
 		 */
+#define		PACKET3_WRITE_DATA__DST_SEL(x) ((((unsigned)(x)) & 0xF) << 8)
+#define		PACKET3_WRITE_DATA__ADDR_INCR(x) ((((unsigned)(x)) & 0x1) << 16)
+#define		PACKET3_WRITE_DATA__RESUME_VF_MI300(x) ((((unsigned)(x)) & 0x1) << 19)
+#define		PACKET3_WRITE_DATA__WR_CONFIRM(x) ((((unsigned)(x)) & 0x1) << 20)
+#define		PACKET3_WRITE_DATA__CACHE_POLICY(x) ((((unsigned)(x)) & 0x3) << 25)
+#define		PACKET3_WRITE_DATA__DST_MMREG_ADDR(x) ((((unsigned)(x)) & 0x3FFFF) << 0)
+#define		PACKET3_WRITE_DATA__DST_GDS_ADDR(x) ((((unsigned)(x)) & 0xFFFF) << 0)
+#define		PACKET3_WRITE_DATA__DST_MEM_ADDR_LO(x) ((((unsigned)(x)) & 0x3FFFFFFF) << 2)
+#define		PACKET3_WRITE_DATA__DST_MEM_ADDR_HI(x) ((unsigned)(x))
+#define		PACKET3_WRITE_DATA__DST_SEL__MEM_MAPPED_REGISTER	0
+#define		PACKET3_WRITE_DATA__DST_SEL__TC_L2	2
+#define		PACKET3_WRITE_DATA__DST_SEL__GDS	3
+#define		PACKET3_WRITE_DATA__DST_SEL__MEMORY	5
+#define		PACKET3_WRITE_DATA__DST_SEL__MEMORY_MAPPED_ADC_PERSISTENT_STATE	6
+#define		PACKET3_WRITE_DATA__ADDR_INCR__INCREMENT_ADDRESS	0
+#define		PACKET3_WRITE_DATA__ADDR_INCR__DO_NOT_INCREMENT_ADDRESS	1
+#define		PACKET3_WRITE_DATA__WR_CONFIRM__DO_NOT_WAIT_FOR_WRITE_CONFIRMATION	0
+#define		PACKET3_WRITE_DATA__WR_CONFIRM__WAIT_FOR_WRITE_CONFIRMATION	1
+#define		PACKET3_WRITE_DATA__CACHE_POLICY__LRU	0
+#define		PACKET3_WRITE_DATA__CACHE_POLICY__STREAM	1
+#define		PACKET3_WRITE_DATA__CACHE_POLICY__NOA	2
+#define		PACKET3_WRITE_DATA__CACHE_POLICY__BYPASS	3
 #define	PACKET3_DRAW_INDEX_INDIRECT_MULTI		0x38
 #define	PACKET3_MEM_SEMAPHORE				0x39
 #              define PACKET3_SEM_USE_MAILBOX       (0x1 << 16)
@@ -154,6 +196,33 @@
 		/* 0 - me
 		 * 1 - pfp
 		 */
+#define		PACKET3_WAIT_REG_MEM__FUNCTION(x) ((((unsigned)(x)) & 0x7) << 0)
+#define		PACKET3_WAIT_REG_MEM__MEM_SPACE(x) ((((unsigned)(x)) & 0x3) << 4)
+#define		PACKET3_WAIT_REG_MEM__OPERATION(x) ((((unsigned)(x)) & 0x3) << 6)
+#define		PACKET3_WAIT_REG_MEM__MES_INTR_PIPE(x) ((((unsigned)(x)) & 0x3) << 22)
+#define		PACKET3_WAIT_REG_MEM__MES_ACTION(x) ((((unsigned)(x)) & 0x1) << 24)
+#define		PACKET3_WAIT_REG_MEM__CACHE_POLICY(x) ((((unsigned)(x)) & 0x3) << 25)
+#define		PACKET3_WAIT_REG_MEM__MEM_POLL_ADDR_LO(x) ((((unsigned)(x)) & 0x3FFFFFFF) << 2)
+#define		PACKET3_WAIT_REG_MEM__REG_POLL_ADDR(x) ((((unsigned)(x)) & 0x3FFFF) << 0)
+#define		PACKET3_WAIT_REG_MEM__REG_WRITE_ADDR1(x) ((((unsigned)(x)) & 0x3FFFF) << 0)
+#define		PACKET3_WAIT_REG_MEM__MEM_POLL_ADDR_HI(x) ((unsigned)(x))
+#define		PACKET3_WAIT_REG_MEM__REG_WRITE_ADDR2(x) ((((unsigned)(x)) & 0x3FFFF) << 0)
+#define		PACKET3_WAIT_REG_MEM__REFERENCE(x) ((unsigned)(x))
+#define		PACKET3_WAIT_REG_MEM__MASK(x) ((unsigned)(x))
+#define		PACKET3_WAIT_REG_MEM__POLL_INTERVAL(x) ((((unsigned)(x)) & 0xFFFF) << 0)
+#define		PACKET3_WAIT_REG_MEM__OPTIMIZE_ACE_OFFLOAD_MODE(x) ((((unsigned)(x)) & 0x1) << 31)
+#define		PACKET3_WAIT_REG_MEM__FUNCTION__ALWAYS_PASS	0
+#define		PACKET3_WAIT_REG_MEM__FUNCTION__LESS_THAN_REF_VALUE	1
+#define		PACKET3_WAIT_REG_MEM__FUNCTION__LESS_THAN_EQUAL_TO_THE_REF_VALUE	2
+#define		PACKET3_WAIT_REG_MEM__FUNCTION__EQUAL_TO_THE_REFERENCE_VALUE	3
+#define		PACKET3_WAIT_REG_MEM__FUNCTION__NOT_EQUAL_REFERENCE_VALUE	4
+#define		PACKET3_WAIT_REG_MEM__FUNCTION__GREATER_THAN_OR_EQUAL_REFERENCE_VALUE	5
+#define		PACKET3_WAIT_REG_MEM__FUNCTION__GREATER_THAN_REFERENCE_VALUE	6
+#define		PACKET3_WAIT_REG_MEM__MEM_SPACE__REGISTER_SPACE	0
+#define		PACKET3_WAIT_REG_MEM__MEM_SPACE__MEMORY_SPACE	1
+#define		PACKET3_WAIT_REG_MEM__OPERATION__WAIT_REG_MEM	0
+#define		PACKET3_WAIT_REG_MEM__OPERATION__WR_WAIT_WR_REG	1
+#define		PACKET3_WAIT_REG_MEM__OPERATION__WAIT_MEM_PREEMPTABLE	3
 #define	PACKET3_INDIRECT_BUFFER				0x3F
 #define		INDIRECT_BUFFER_VALID                   (1 << 23)
 #define		INDIRECT_BUFFER_CACHE_POLICY(x)         ((x) << 28)
@@ -163,7 +232,63 @@
 		 */
 #define     INDIRECT_BUFFER_PRE_ENB(x)		 ((x) << 21)
 #define     INDIRECT_BUFFER_PRE_RESUME(x)               ((x) << 30)
+#define		PACKET3_INDIRECT_BUFFER__IB_BASE_LO(x) ((((unsigned)(x)) & 0x3FFFFFFF) << 2)
+#define		PACKET3_INDIRECT_BUFFER__IB_BASE_HI(x) ((unsigned)(x))
+#define		PACKET3_INDIRECT_BUFFER__IB_SIZE(x) ((((unsigned)(x)) & 0xFFFFF) << 0)
+#define		PACKET3_INDIRECT_BUFFER__CHAIN(x) ((((unsigned)(x)) & 0x1) << 20)
+#define		PACKET3_INDIRECT_BUFFER__OFFLOAD_POLLING(x) ((((unsigned)(x)) & 0x1) << 21)
+#define		PACKET3_INDIRECT_BUFFER__VALID(x) ((((unsigned)(x)) & 0x1) << 23)
+#define		PACKET3_INDIRECT_BUFFER__VMID(x) ((((unsigned)(x)) & 0xF) << 24)
+#define		PACKET3_INDIRECT_BUFFER__CACHE_POLICY(x) ((((unsigned)(x)) & 0x3) << 28)
+#define		PACKET3_INDIRECT_BUFFER__PRIV(x) ((((unsigned)(x)) & 0x1) << 31)
+#define		PACKET3_INDIRECT_BUFFER__CACHE_POLICY__LRU	0
+#define		PACKET3_INDIRECT_BUFFER__CACHE_POLICY__STREAM	1
 #define	PACKET3_COPY_DATA				0x40
+#define		PACKET3_COPY_DATA__SRC_SEL(x) ((((unsigned)(x)) & 0xF) << 0)
+#define		PACKET3_COPY_DATA__DST_SEL(x) ((((unsigned)(x)) & 0xF) << 8)
+#define		PACKET3_COPY_DATA__SRC_CACHE_POLICY(x) ((((unsigned)(x)) & 0x3) << 13)
+#define		PACKET3_COPY_DATA__COUNT_SEL(x) ((((unsigned)(x)) & 0x1) << 16)
+#define		PACKET3_COPY_DATA__WR_CONFIRM(x) ((((unsigned)(x)) & 0x1) << 20)
+#define		PACKET3_COPY_DATA__DST_CACHE_POLICY(x) ((((unsigned)(x)) & 0x3) << 25)
+#define		PACKET3_COPY_DATA__PQ_EXE_STATUS(x) ((((unsigned)(x)) & 0x1) << 29)
+#define		PACKET3_COPY_DATA__SRC_REG_OFFSET(x) ((((unsigned)(x)) & 0x3FFFF) << 0)
+#define		PACKET3_COPY_DATA__SRC_32B_ADDR_LO(x) ((((unsigned)(x)) & 0x3FFFFFFF) << 2)
+#define		PACKET3_COPY_DATA__SRC_64B_ADDR_LO(x) ((((unsigned)(x)) & 0x1FFFFFFF) << 3)
+#define		PACKET3_COPY_DATA__SRC_GDS_ADDR_LO(x) ((((unsigned)(x)) & 0xFFFF) << 0)
+#define		PACKET3_COPY_DATA__IMM_DATA(x) ((unsigned)(x))
+#define		PACKET3_COPY_DATA__SRC_MEMTC_ADDR_HI(x) ((unsigned)(x))
+#define		PACKET3_COPY_DATA__SRC_IMM_DATA(x) ((unsigned)(x))
+#define		PACKET3_COPY_DATA__DST_REG_OFFSET(x) ((((unsigned)(x)) & 0x3FFFF) << 0)
+#define		PACKET3_COPY_DATA__DST_32B_ADDR_LO(x) ((((unsigned)(x)) & 0x3FFFFFFF) << 2)
+#define		PACKET3_COPY_DATA__DST_64B_ADDR_LO(x) ((((unsigned)(x)) & 0x1FFFFFFF) << 3)
+#define		PACKET3_COPY_DATA__DST_GDS_ADDR_LO(x) ((((unsigned)(x)) & 0xFFFF) << 0)
+#define		PACKET3_COPY_DATA__DST_ADDR_HI(x) ((unsigned)(x))
+#define		PACKET3_COPY_DATA__SRC_SEL__MEM_MAPPED_REGISTER	0
+#define		PACKET3_COPY_DATA__SRC_SEL__MEMORY	1
+#define		PACKET3_COPY_DATA__SRC_SEL__TC_L2	2
+#define		PACKET3_COPY_DATA__SRC_SEL__GDS	3
+#define		PACKET3_COPY_DATA__SRC_SEL__PERFCOUNTERS	4
+#define		PACKET3_COPY_DATA__SRC_SEL__IMMEDIATE_DATA	5
+#define		PACKET3_COPY_DATA__SRC_SEL__ATOMIC_RETURN_DATA	6
+#define		PACKET3_COPY_DATA__SRC_SEL__GDS_ATOMIC_RETURN_DATA0	7
+#define		PACKET3_COPY_DATA__SRC_SEL__GDS_ATOMIC_RETURN_DATA1	8
+#define		PACKET3_COPY_DATA__SRC_SEL__GPU_CLOCK_COUNT	9
+#define		PACKET3_COPY_DATA__DST_SEL__MEM_MAPPED_REGISTER	0
+#define		PACKET3_COPY_DATA__DST_SEL__TC_L2	2
+#define		PACKET3_COPY_DATA__DST_SEL__GDS	3
+#define		PACKET3_COPY_DATA__DST_SEL__PERFCOUNTERS	4
+#define		PACKET3_COPY_DATA__DST_SEL__MEMORY	5
+#define		PACKET3_COPY_DATA__DST_SEL__MEM_MAPPED_REG_DC	6
+#define		PACKET3_COPY_DATA__SRC_CACHE_POLICY__LRU	0
+#define		PACKET3_COPY_DATA__SRC_CACHE_POLICY__STREAM 1
+#define		PACKET3_COPY_DATA__COUNT_SEL__32_BITS_OF_DATA	0
+#define		PACKET3_COPY_DATA__COUNT_SEL__64_BITS_OF_DATA	1
+#define		PACKET3_COPY_DATA__WR_CONFIRM__DO_NOT_WAIT_FOR_CONFIRMATION	0
+#define		PACKET3_COPY_DATA__WR_CONFIRM__WAIT_FOR_CONFIRMATION	1
+#define		PACKET3_COPY_DATA__DST_CACHE_POLICY__LRU	0
+#define		PACKET3_COPY_DATA__DST_CACHE_POLICY__STREAM	1
+#define		PACKET3_COPY_DATA__PQ_EXE_STATUS__DEFAULT	0
+#define		PACKET3_COPY_DATA__PQ_EXE_STATUS__PHASE_UPDATE	1
 #define	PACKET3_PFP_SYNC_ME				0x42
 #define	PACKET3_COND_WRITE				0x45
 #define	PACKET3_EVENT_WRITE				0x46
@@ -175,6 +300,15 @@
 		 * 3 - SAMPLE_STREAMOUTSTAT*
 		 * 4 - *S_PARTIAL_FLUSH
 		 */
+#define		PACKET3_EVENT_WRITE__EVENT_TYPE(x) ((((unsigned)(x)) & 0x3F) << 0)
+#define		PACKET3_EVENT_WRITE__EVENT_INDEX(x) ((((unsigned)(x)) & 0xF) << 8)
+#define		PACKET3_EVENT_WRITE__OFFLOAD_ENABLE(x) ((((unsigned)(x)) & 0x1) << 31)
+#define		PACKET3_EVENT_WRITE__SAMP_PLST_CNTR_MODE(x) ((((unsigned)(x)) & 0x3) << 29)
+#define		PACKET3_EVENT_WRITE__ADDRESS_LO(x) ((((unsigned)(x)) & 0x1FFFFFFF) << 3)
+#define		PACKET3_EVENT_WRITE__ADDRESS_HI(x) (((unsigned)(x)) << 0)
+#define		PACKET3_EVENT_WRITE__EVENT_INDEX__OTHER	0
+#define		PACKET3_EVENT_WRITE__EVENT_INDEX__SAMPLE_PIPELINESTATS	2
+#define		PACKET3_EVENT_WRITE__EVENT_INDEX__CS_PARTIAL_FLUSH	4
 #define	PACKET3_RELEASE_MEM				0x49
 #define		EVENT_TYPE(x)                           ((x) << 0)
 #define		EVENT_INDEX(x)                          ((x) << 8)
@@ -280,6 +414,13 @@
 #define 	PACKET3_ACQUIRE_MEM_CP_COHER_CNTL_SH_ICACHE_ACTION_ENA(x) ((x) << 29)
 #define 	PACKET3_ACQUIRE_MEM_CP_COHER_CNTL_SH_KCACHE_WB_ACTION_ENA(x) ((x) << 30)
 #define	PACKET3_REWIND					0x59
+#define		PACKET3_ACQUIRE_MEM__COHER_SIZE(x) ((unsigned)(x))
+#define 	PACKET3_ACQUIRE_MEM__COHER_SIZE_HI(x) ((((unsigned)(x)) & 0xFF) << 0)
+#define 	PACKET3_ACQUIRE_MEM__COHER_SIZE_HI_VG10(x) ((((unsigned)(x)) & 0xFFFFFF) << 0)
+#define 	PACKET3_ACQUIRE_MEM__COHER_BASE_LO(x) ((unsigned)(x))
+#define 	PACKET3_ACQUIRE_MEM__COHER_BASE_HI(x) ((((unsigned)(x)) & 0xFFFFFF) << 0)
+#define 	PACKET3_ACQUIRE_MEM__POLL_INTERVAL(x) ((((unsigned)(x)) & 0xFFFF) << 0)
+#define 	PACKET3_ACQUIRE_MEM__GCR_CNTL(x) ((((unsigned)(x)) & 0x7FF) << 0)
 #define	PACKET3_LOAD_UCONFIG_REG			0x5E
 #define	PACKET3_LOAD_SH_REG				0x5F
 #define	PACKET3_LOAD_CONFIG_REG				0x60
@@ -294,12 +435,16 @@
 #define	PACKET3_SET_SH_REG				0x76
 #define		PACKET3_SET_SH_REG_START			0x00002c00
 #define		PACKET3_SET_SH_REG_END				0x00003000
+#define 	PACKET3_SET_SH_REG__REG_OFFSET(x) ((((unsigned)(x)) & 0xFFFF) << 0)
+#define 	PACKET3_SET_SH_REG__VMID_SHIFT(x) ((((unsigned)(x)) & 0x1F) << 23)
+#define 	PACKET3_SET_SH_REG__INDEX(x) ((((unsigned)(x)) & 0xF) << 28)
 #define	PACKET3_SET_SH_REG_OFFSET			0x77
 #define	PACKET3_SET_QUEUE_REG				0x78
 #define	PACKET3_SET_UCONFIG_REG				0x79
 #define		PACKET3_SET_UCONFIG_REG_START			0x0000c000
 #define		PACKET3_SET_UCONFIG_REG_END			0x0000c400
 #define		PACKET3_SET_UCONFIG_REG_INDEX_TYPE		(2 << 28)
+#define 	PACKET3_SET_UCONFIG_REG__REG_OFFSET(x) ((((unsigned)(x)) & 0xFFFF) << 0)
 #define	PACKET3_SCRATCH_RAM_WRITE			0x7D
 #define	PACKET3_SCRATCH_RAM_READ			0x7E
 #define	PACKET3_LOAD_CONST_RAM				0x80
@@ -407,6 +552,15 @@
 #              define PACKET3_QUERY_STATUS_DOORBELL_OFFSET(x)  ((x) << 2)
 #              define PACKET3_QUERY_STATUS_ENG_SEL(x)          ((x) << 25)
 
+#define PACKET3_RUN_CLEANER_SHADER_9_0                      0xD7
+/* 1. header
+ * 2. RESERVED [31:0]
+ */
+
+#define PACKET3_RUN_CLEANER_SHADER                      0xD2
+/* 1. header
+ * 2. RESERVED [31:0]
+ */
 
 #define VCE_CMD_NO_OP		0x00000000
 #define VCE_CMD_END		0x00000001
diff --git a/drivers/gpu/drm/amd/amdgpu/soc21.c b/drivers/gpu/drm/amd/amdgpu/soc21.c
index 514bfc705d5a..ad36c96478a8 100644
--- a/drivers/gpu/drm/amd/amdgpu/soc21.c
+++ b/drivers/gpu/drm/amd/amdgpu/soc21.c
@@ -48,70 +48,62 @@
 static const struct amd_ip_funcs soc21_common_ip_funcs;
 
 /* SOC21 */
-static const struct amdgpu_video_codec_info vcn_4_0_0_video_codecs_encode_array_vcn0[] =
-{
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 2304, 0)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 4096, 2304, 0)},
+static const struct amdgpu_video_codec_info vcn_4_0_0_video_codecs_encode_array_vcn0[] = {
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 0)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_AV1, 8192, 4352, 0)},
 };
 
-static const struct amdgpu_video_codec_info vcn_4_0_0_video_codecs_encode_array_vcn1[] =
-{
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 2304, 0)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 4096, 2304, 0)},
+static const struct amdgpu_video_codec_info vcn_4_0_0_video_codecs_encode_array_vcn1[] = {
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 0)},
 };
 
-static const struct amdgpu_video_codecs vcn_4_0_0_video_codecs_encode_vcn0 =
-{
+static const struct amdgpu_video_codecs vcn_4_0_0_video_codecs_encode_vcn0 = {
 	.codec_count = ARRAY_SIZE(vcn_4_0_0_video_codecs_encode_array_vcn0),
 	.codec_array = vcn_4_0_0_video_codecs_encode_array_vcn0,
 };
 
-static const struct amdgpu_video_codecs vcn_4_0_0_video_codecs_encode_vcn1 =
-{
+static const struct amdgpu_video_codecs vcn_4_0_0_video_codecs_encode_vcn1 = {
 	.codec_count = ARRAY_SIZE(vcn_4_0_0_video_codecs_encode_array_vcn1),
 	.codec_array = vcn_4_0_0_video_codecs_encode_array_vcn1,
 };
 
-static const struct amdgpu_video_codec_info vcn_4_0_0_video_codecs_decode_array_vcn0[] =
-{
+static const struct amdgpu_video_codec_info vcn_4_0_0_video_codecs_decode_array_vcn0[] = {
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 16384, 16384, 0)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_AV1, 8192, 4352, 0)},
 };
 
-static const struct amdgpu_video_codec_info vcn_4_0_0_video_codecs_decode_array_vcn1[] =
-{
+static const struct amdgpu_video_codec_info vcn_4_0_0_video_codecs_decode_array_vcn1[] = {
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 16384, 16384, 0)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)},
 };
 
-static const struct amdgpu_video_codecs vcn_4_0_0_video_codecs_decode_vcn0 =
-{
+static const struct amdgpu_video_codecs vcn_4_0_0_video_codecs_decode_vcn0 = {
 	.codec_count = ARRAY_SIZE(vcn_4_0_0_video_codecs_decode_array_vcn0),
 	.codec_array = vcn_4_0_0_video_codecs_decode_array_vcn0,
 };
 
-static const struct amdgpu_video_codecs vcn_4_0_0_video_codecs_decode_vcn1 =
-{
+static const struct amdgpu_video_codecs vcn_4_0_0_video_codecs_decode_vcn1 = {
 	.codec_count = ARRAY_SIZE(vcn_4_0_0_video_codecs_decode_array_vcn1),
 	.codec_array = vcn_4_0_0_video_codecs_decode_array_vcn1,
 };
 
 /* SRIOV SOC21, not const since data is controlled by host */
 static struct amdgpu_video_codec_info sriov_vcn_4_0_0_video_codecs_encode_array_vcn0[] = {
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 2304, 0)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 4096, 2304, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 0)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_AV1, 8192, 4352, 0)},
 };
 
 static struct amdgpu_video_codec_info sriov_vcn_4_0_0_video_codecs_encode_array_vcn1[] = {
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 2304, 0)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 4096, 2304, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 0)},
 };
 
 static struct amdgpu_video_codecs sriov_vcn_4_0_0_video_codecs_encode_vcn0 = {
@@ -125,23 +117,17 @@ static struct amdgpu_video_codecs sriov_vcn_4_0_0_video_codecs_encode_vcn1 = {
 };
 
 static struct amdgpu_video_codec_info sriov_vcn_4_0_0_video_codecs_decode_array_vcn0[] = {
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 16384, 16384, 0)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_AV1, 8192, 4352, 0)},
 };
 
 static struct amdgpu_video_codec_info sriov_vcn_4_0_0_video_codecs_decode_array_vcn1[] = {
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)},
-	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 16384, 16384, 0)},
 	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)},
 };
 
@@ -161,10 +147,11 @@ static int soc21_query_video_codecs(struct amdgpu_device *adev, bool encode,
 	if (adev->vcn.num_vcn_inst == hweight8(adev->vcn.harvest_config))
 		return -EINVAL;
 
-	switch (adev->ip_versions[UVD_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, UVD_HWIP, 0)) {
 	case IP_VERSION(4, 0, 0):
 	case IP_VERSION(4, 0, 2):
 	case IP_VERSION(4, 0, 4):
+	case IP_VERSION(4, 0, 5):
 		if (amdgpu_sriov_vf(adev)) {
 			if ((adev->vcn.harvest_config & AMDGPU_VCN_HARVEST_VCN0) ||
 			!amdgpu_sriov_is_av1_support(adev)) {
@@ -192,6 +179,12 @@ static int soc21_query_video_codecs(struct amdgpu_device *adev, bool encode,
 			}
 		}
 		return 0;
+	case IP_VERSION(4, 0, 6):
+		if (encode)
+			*codecs = &vcn_4_0_0_video_codecs_encode_vcn0;
+		else
+			*codecs = &vcn_4_0_0_video_codecs_decode_vcn0;
+		return 0;
 	default:
 		return -EINVAL;
 	}
@@ -248,11 +241,6 @@ void soc21_grbm_select(struct amdgpu_device *adev,
 	WREG32_SOC15(GC, 0, regGRBM_GFX_CNTL, grbm_gfx_cntl);
 }
 
-static void soc21_vga_set_state(struct amdgpu_device *adev, bool state)
-{
-	/* todo */
-}
-
 static bool soc21_read_disabled_bios(struct amdgpu_device *adev)
 {
 	/* todo */
@@ -288,12 +276,12 @@ static uint32_t soc21_read_indexed_register(struct amdgpu_device *adev, u32 se_n
 
 	mutex_lock(&adev->grbm_idx_mutex);
 	if (se_num != 0xffffffff || sh_num != 0xffffffff)
-		amdgpu_gfx_select_se_sh(adev, se_num, sh_num, 0xffffffff);
+		amdgpu_gfx_select_se_sh(adev, se_num, sh_num, 0xffffffff, 0);
 
 	val = RREG32(reg_offset);
 
 	if (se_num != 0xffffffff || sh_num != 0xffffffff)
-		amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+		amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 	return val;
 }
@@ -386,13 +374,17 @@ soc21_asic_reset_method(struct amdgpu_device *adev)
 		dev_warn(adev->dev, "Specified reset method:%d isn't supported, using AUTO instead.\n",
 				  amdgpu_reset_method);
 
-	switch (adev->ip_versions[MP1_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
 	case IP_VERSION(13, 0, 0):
 	case IP_VERSION(13, 0, 7):
 	case IP_VERSION(13, 0, 10):
 		return AMD_RESET_METHOD_MODE1;
 	case IP_VERSION(13, 0, 4):
 	case IP_VERSION(13, 0, 11):
+	case IP_VERSION(14, 0, 0):
+	case IP_VERSION(14, 0, 1):
+	case IP_VERSION(14, 0, 4):
+	case IP_VERSION(14, 0, 5):
 		return AMD_RESET_METHOD_MODE2;
 	default:
 		if (amdgpu_dpm_is_baco_supported(adev))
@@ -445,20 +437,11 @@ static void soc21_program_aspm(struct amdgpu_device *adev)
 	if (!amdgpu_device_should_use_aspm(adev))
 		return;
 
-	if (!(adev->flags & AMD_IS_APU) &&
-	    (adev->nbio.funcs->program_aspm))
+	if (adev->nbio.funcs->program_aspm)
 		adev->nbio.funcs->program_aspm(adev);
 }
 
-static void soc21_enable_doorbell_aperture(struct amdgpu_device *adev,
-					bool enable)
-{
-	adev->nbio.funcs->enable_doorbell_aperture(adev, enable);
-	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, enable);
-}
-
-const struct amdgpu_ip_block_version soc21_common_ip_block =
-{
+const struct amdgpu_ip_block_version soc21_common_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_COMMON,
 	.major = 1,
 	.minor = 0,
@@ -468,12 +451,10 @@ const struct amdgpu_ip_block_version soc21_common_ip_block =
 
 static bool soc21_need_full_reset(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 0):
-		return amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC);
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 3):
-		return false;
 	default:
 		return true;
 	}
@@ -496,16 +477,6 @@ static bool soc21_need_reset_on_init(struct amdgpu_device *adev)
 	return false;
 }
 
-static uint64_t soc21_get_pcie_replay_count(struct amdgpu_device *adev)
-{
-
-	/* TODO
-	 * dummy implement for pcie_replay_count sysfs interface
-	 * */
-
-	return 0;
-}
-
 static void soc21_init_doorbell_index(struct amdgpu_device *adev)
 {
 	adev->doorbell_index.kiq = AMDGPU_NAVI10_DOORBELL_KIQ;
@@ -534,6 +505,7 @@ static void soc21_init_doorbell_index(struct amdgpu_device *adev)
 	adev->doorbell_index.vcn.vcn_ring2_3 = AMDGPU_NAVI10_DOORBELL64_VCN2_3;
 	adev->doorbell_index.vcn.vcn_ring4_5 = AMDGPU_NAVI10_DOORBELL64_VCN4_5;
 	adev->doorbell_index.vcn.vcn_ring6_7 = AMDGPU_NAVI10_DOORBELL64_VCN6_7;
+	adev->doorbell_index.vpe_ring = AMDGPU_NAVI10_DOORBELL64_VPE;
 	adev->doorbell_index.first_non_cp = AMDGPU_NAVI10_DOORBELL64_FIRST_NON_CP;
 	adev->doorbell_index.last_non_cp = AMDGPU_NAVI10_DOORBELL64_LAST_NON_CP;
 
@@ -549,9 +521,9 @@ static int soc21_update_umd_stable_pstate(struct amdgpu_device *adev,
 					  bool enter)
 {
 	if (enter)
-		amdgpu_gfx_rlc_enter_safe_mode(adev);
+		amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 	else
-		amdgpu_gfx_rlc_exit_safe_mode(adev);
+		amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 
 	if (adev->gfx.funcs->update_perfmon_mgcg)
 		adev->gfx.funcs->update_perfmon_mgcg(adev, !enter);
@@ -559,14 +531,12 @@ static int soc21_update_umd_stable_pstate(struct amdgpu_device *adev,
 	return 0;
 }
 
-static const struct amdgpu_asic_funcs soc21_asic_funcs =
-{
+static const struct amdgpu_asic_funcs soc21_asic_funcs = {
 	.read_disabled_bios = &soc21_read_disabled_bios,
 	.read_bios_from_rom = &amdgpu_soc15_read_bios_from_rom,
 	.read_register = &soc21_read_register,
 	.reset = &soc21_asic_reset,
 	.reset_method = &soc21_asic_reset_method,
-	.set_vga_state = &soc21_vga_set_state,
 	.get_xclk = &soc21_get_xclk,
 	.set_uvd_clocks = &soc21_set_uvd_clocks,
 	.set_vce_clocks = &soc21_set_vce_clocks,
@@ -574,20 +544,18 @@ static const struct amdgpu_asic_funcs soc21_asic_funcs =
 	.init_doorbell_index = &soc21_init_doorbell_index,
 	.need_full_reset = &soc21_need_full_reset,
 	.need_reset_on_init = &soc21_need_reset_on_init,
-	.get_pcie_replay_count = &soc21_get_pcie_replay_count,
+	.get_pcie_replay_count = &amdgpu_nbio_get_pcie_replay_count,
 	.supports_baco = &amdgpu_dpm_is_baco_supported,
 	.pre_asic_init = &soc21_pre_asic_init,
 	.query_video_codecs = &soc21_query_video_codecs,
 	.update_umd_stable_pstate = &soc21_update_umd_stable_pstate,
 };
 
-static int soc21_common_early_init(void *handle)
+static int soc21_common_early_init(struct amdgpu_ip_block *ip_block)
 {
-#define MMIO_REG_HOLE_OFFSET (0x80000 - PAGE_SIZE)
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET;
-	adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET;
+	adev->nbio.funcs->set_reg_remap(adev);
 	adev->smc_rreg = NULL;
 	adev->smc_wreg = NULL;
 	adev->pcie_rreg = &amdgpu_device_indirect_rreg;
@@ -608,7 +576,7 @@ static int soc21_common_early_init(void *handle)
 
 	adev->rev_id = amdgpu_device_get_rev_id(adev);
 	adev->external_rev_id = 0xff;
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 0):
 		adev->cg_flags = AMD_CG_SUPPORT_GFX_CGCG |
 			AMD_CG_SUPPORT_GFX_CGLS |
@@ -718,9 +686,124 @@ static int soc21_common_early_init(void *handle)
 			AMD_PG_SUPPORT_VCN_DPG |
 			AMD_PG_SUPPORT_GFX_PG |
 			AMD_PG_SUPPORT_JPEG;
-		adev->external_rev_id = adev->rev_id + 0x1;
+		adev->external_rev_id = adev->rev_id + 0x80;
+		break;
+	case IP_VERSION(11, 5, 0):
+		adev->cg_flags = AMD_CG_SUPPORT_VCN_MGCG |
+			AMD_CG_SUPPORT_JPEG_MGCG |
+			AMD_CG_SUPPORT_GFX_CGCG |
+			AMD_CG_SUPPORT_GFX_CGLS |
+			AMD_CG_SUPPORT_GFX_MGCG |
+			AMD_CG_SUPPORT_GFX_FGCG |
+			AMD_CG_SUPPORT_REPEATER_FGCG |
+			AMD_CG_SUPPORT_GFX_PERF_CLK	|
+			AMD_CG_SUPPORT_GFX_3D_CGCG |
+			AMD_CG_SUPPORT_GFX_3D_CGLS	|
+			AMD_CG_SUPPORT_MC_MGCG |
+			AMD_CG_SUPPORT_MC_LS |
+			AMD_CG_SUPPORT_HDP_LS |
+			AMD_CG_SUPPORT_HDP_DS |
+			AMD_CG_SUPPORT_HDP_SD |
+			AMD_CG_SUPPORT_ATHUB_MGCG |
+			AMD_CG_SUPPORT_ATHUB_LS |
+			AMD_CG_SUPPORT_IH_CG |
+			AMD_CG_SUPPORT_BIF_MGCG |
+			AMD_CG_SUPPORT_BIF_LS;
+		adev->pg_flags = AMD_PG_SUPPORT_VCN_DPG |
+			AMD_PG_SUPPORT_JPEG_DPG |
+			AMD_PG_SUPPORT_VCN |
+			AMD_PG_SUPPORT_JPEG |
+			AMD_PG_SUPPORT_GFX_PG;
+		if (adev->rev_id == 0)
+			adev->external_rev_id = 0x1;
+		else
+			adev->external_rev_id = adev->rev_id + 0x10;
+		break;
+	case IP_VERSION(11, 5, 1):
+		adev->cg_flags =
+			AMD_CG_SUPPORT_GFX_CGCG |
+			AMD_CG_SUPPORT_GFX_CGLS |
+			AMD_CG_SUPPORT_GFX_MGCG |
+			AMD_CG_SUPPORT_GFX_FGCG |
+			AMD_CG_SUPPORT_REPEATER_FGCG |
+			AMD_CG_SUPPORT_GFX_PERF_CLK	|
+			AMD_CG_SUPPORT_GFX_3D_CGCG |
+			AMD_CG_SUPPORT_GFX_3D_CGLS	|
+			AMD_CG_SUPPORT_MC_MGCG |
+			AMD_CG_SUPPORT_MC_LS |
+			AMD_CG_SUPPORT_HDP_LS |
+			AMD_CG_SUPPORT_HDP_DS |
+			AMD_CG_SUPPORT_HDP_SD |
+			AMD_CG_SUPPORT_ATHUB_MGCG |
+			AMD_CG_SUPPORT_ATHUB_LS |
+			AMD_CG_SUPPORT_IH_CG |
+			AMD_CG_SUPPORT_BIF_MGCG |
+			AMD_CG_SUPPORT_BIF_LS |
+			AMD_CG_SUPPORT_VCN_MGCG |
+			AMD_CG_SUPPORT_JPEG_MGCG;
+		adev->pg_flags =
+			AMD_PG_SUPPORT_GFX_PG |
+			AMD_PG_SUPPORT_VCN |
+			AMD_PG_SUPPORT_VCN_DPG |
+			AMD_PG_SUPPORT_JPEG;
+		adev->external_rev_id = adev->rev_id + 0xc1;
+		break;
+	case IP_VERSION(11, 5, 2):
+		adev->cg_flags = AMD_CG_SUPPORT_VCN_MGCG |
+			AMD_CG_SUPPORT_JPEG_MGCG |
+			AMD_CG_SUPPORT_GFX_CGCG |
+			AMD_CG_SUPPORT_GFX_CGLS |
+			AMD_CG_SUPPORT_GFX_MGCG |
+			AMD_CG_SUPPORT_GFX_FGCG |
+			AMD_CG_SUPPORT_REPEATER_FGCG |
+			AMD_CG_SUPPORT_GFX_PERF_CLK	|
+			AMD_CG_SUPPORT_GFX_3D_CGCG |
+			AMD_CG_SUPPORT_GFX_3D_CGLS	|
+			AMD_CG_SUPPORT_MC_MGCG |
+			AMD_CG_SUPPORT_MC_LS |
+			AMD_CG_SUPPORT_HDP_LS |
+			AMD_CG_SUPPORT_HDP_DS |
+			AMD_CG_SUPPORT_HDP_SD |
+			AMD_CG_SUPPORT_ATHUB_MGCG |
+			AMD_CG_SUPPORT_ATHUB_LS |
+			AMD_CG_SUPPORT_IH_CG |
+			AMD_CG_SUPPORT_BIF_MGCG |
+			AMD_CG_SUPPORT_BIF_LS;
+		adev->pg_flags = AMD_PG_SUPPORT_VCN_DPG |
+			AMD_PG_SUPPORT_VCN |
+			AMD_PG_SUPPORT_JPEG_DPG |
+			AMD_PG_SUPPORT_JPEG |
+			AMD_PG_SUPPORT_GFX_PG;
+		adev->external_rev_id = adev->rev_id + 0x40;
+		break;
+	case IP_VERSION(11, 5, 3):
+		adev->cg_flags = AMD_CG_SUPPORT_VCN_MGCG |
+			AMD_CG_SUPPORT_JPEG_MGCG |
+			AMD_CG_SUPPORT_GFX_CGCG |
+			AMD_CG_SUPPORT_GFX_CGLS |
+			AMD_CG_SUPPORT_GFX_MGCG |
+			AMD_CG_SUPPORT_GFX_FGCG |
+			AMD_CG_SUPPORT_REPEATER_FGCG |
+			AMD_CG_SUPPORT_GFX_PERF_CLK |
+			AMD_CG_SUPPORT_GFX_3D_CGCG |
+			AMD_CG_SUPPORT_GFX_3D_CGLS |
+			AMD_CG_SUPPORT_MC_MGCG |
+			AMD_CG_SUPPORT_MC_LS |
+			AMD_CG_SUPPORT_HDP_LS |
+			AMD_CG_SUPPORT_HDP_DS |
+			AMD_CG_SUPPORT_HDP_SD |
+			AMD_CG_SUPPORT_ATHUB_MGCG |
+			AMD_CG_SUPPORT_ATHUB_LS |
+			AMD_CG_SUPPORT_IH_CG |
+			AMD_CG_SUPPORT_BIF_MGCG |
+			AMD_CG_SUPPORT_BIF_LS;
+		adev->pg_flags = AMD_PG_SUPPORT_VCN_DPG |
+			AMD_PG_SUPPORT_VCN |
+			AMD_PG_SUPPORT_JPEG_DPG |
+			AMD_PG_SUPPORT_JPEG |
+			AMD_PG_SUPPORT_GFX_PG;
+		adev->external_rev_id = adev->rev_id + 0x50;
 		break;
-
 	default:
 		/* FIXME: not supported yet */
 		return -EINVAL;
@@ -734,9 +817,9 @@ static int soc21_common_early_init(void *handle)
 	return 0;
 }
 
-static int soc21_common_late_init(void *handle)
+static int soc21_common_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev)) {
 		xgpu_nv_mailbox_get_irq(adev);
@@ -764,12 +847,17 @@ static int soc21_common_late_init(void *handle)
 			amdgpu_irq_get(adev, &adev->nbio.ras_err_event_athub_irq, 0);
 	}
 
+	/* Enable selfring doorbell aperture late because doorbell BAR
+	 * aperture will change if resize BAR successfully in gmc sw_init.
+	 */
+	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, true);
+
 	return 0;
 }
 
-static int soc21_common_sw_init(void *handle)
+static int soc21_common_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		xgpu_nv_mailbox_add_irq_id(adev);
@@ -777,14 +865,9 @@ static int soc21_common_sw_init(void *handle)
 	return 0;
 }
 
-static int soc21_common_sw_fini(void *handle)
-{
-	return 0;
-}
-
-static int soc21_common_hw_init(void *handle)
+static int soc21_common_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* enable aspm */
 	soc21_program_aspm(adev);
@@ -794,20 +877,25 @@ static int soc21_common_hw_init(void *handle)
 	 * for the purpose of expose those registers
 	 * to process space
 	 */
-	if (adev->nbio.funcs->remap_hdp_registers)
+	if (adev->nbio.funcs->remap_hdp_registers && !amdgpu_sriov_vf(adev))
 		adev->nbio.funcs->remap_hdp_registers(adev);
 	/* enable the doorbell aperture */
-	soc21_enable_doorbell_aperture(adev, true);
+	adev->nbio.funcs->enable_doorbell_aperture(adev, true);
 
 	return 0;
 }
 
-static int soc21_common_hw_fini(void *handle)
+static int soc21_common_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	/* disable the doorbell aperture */
-	soc21_enable_doorbell_aperture(adev, false);
+	/* Disable the doorbell aperture and selfring doorbell aperture
+	 * separately in hw_fini because soc21_enable_doorbell_aperture
+	 * has been removed and there is no need to delay disabling
+	 * selfring doorbell.
+	 */
+	adev->nbio.funcs->enable_doorbell_aperture(adev, false);
+	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, false);
 
 	if (amdgpu_sriov_vf(adev)) {
 		xgpu_nv_mailbox_put_irq(adev);
@@ -820,44 +908,63 @@ static int soc21_common_hw_fini(void *handle)
 	return 0;
 }
 
-static int soc21_common_suspend(void *handle)
+static int soc21_common_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return soc21_common_hw_fini(adev);
+	return soc21_common_hw_fini(ip_block);
 }
 
-static int soc21_common_resume(void *handle)
+static bool soc21_need_reset_on_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	u32 sol_reg1, sol_reg2;
 
-	return soc21_common_hw_init(adev);
-}
+	/* Will reset for the following suspend abort cases.
+	 * 1) Only reset dGPU side.
+	 * 2) S3 suspend got aborted and TOS is active.
+	 *    As for dGPU suspend abort cases the SOL value
+	 *    will be kept as zero at this resume point.
+	 */
+	if (!(adev->flags & AMD_IS_APU) && adev->in_s3) {
+		sol_reg1 = RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_81);
+		msleep(100);
+		sol_reg2 = RREG32_SOC15(MP0, 0, regMP0_SMN_C2PMSG_81);
 
-static bool soc21_common_is_idle(void *handle)
-{
-	return true;
+		return (sol_reg1 != sol_reg2);
+	}
+
+	return false;
 }
 
-static int soc21_common_wait_for_idle(void *handle)
+static int soc21_common_resume(struct amdgpu_ip_block *ip_block)
 {
-	return 0;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (soc21_need_reset_on_resume(adev)) {
+		dev_info(adev->dev, "S3 suspend aborted, resetting...");
+		soc21_asic_reset(adev);
+	}
+
+	return soc21_common_hw_init(ip_block);
 }
 
-static int soc21_common_soft_reset(void *handle)
+static bool soc21_common_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	return 0;
+	return true;
 }
 
-static int soc21_common_set_clockgating_state(void *handle,
+static int soc21_common_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	switch (adev->ip_versions[NBIO_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, NBIO_HWIP, 0)) {
 	case IP_VERSION(4, 3, 0):
 	case IP_VERSION(4, 3, 1):
 	case IP_VERSION(7, 7, 0):
+	case IP_VERSION(7, 7, 1):
+	case IP_VERSION(7, 11, 0):
+	case IP_VERSION(7, 11, 1):
+	case IP_VERSION(7, 11, 2):
+	case IP_VERSION(7, 11, 3):
 		adev->nbio.funcs->update_medium_grain_clock_gating(adev,
 				state == AMD_CG_STATE_GATE);
 		adev->nbio.funcs->update_medium_grain_light_sleep(adev,
@@ -871,12 +978,12 @@ static int soc21_common_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int soc21_common_set_powergating_state(void *handle,
+static int soc21_common_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	switch (adev->ip_versions[LSDMA_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, LSDMA_HWIP, 0)) {
 	case IP_VERSION(6, 0, 0):
 	case IP_VERSION(6, 0, 2):
 		adev->lsdma.funcs->update_memory_power_gating(adev,
@@ -889,15 +996,13 @@ static int soc21_common_set_powergating_state(void *handle,
 	return 0;
 }
 
-static void soc21_common_get_clockgating_state(void *handle, u64 *flags)
+static void soc21_common_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->nbio.funcs->get_clockgating_state(adev, flags);
 
 	adev->hdp.funcs->get_clock_gating_state(adev, flags);
-
-	return;
 }
 
 static const struct amd_ip_funcs soc21_common_ip_funcs = {
@@ -905,14 +1010,11 @@ static const struct amd_ip_funcs soc21_common_ip_funcs = {
 	.early_init = soc21_common_early_init,
 	.late_init = soc21_common_late_init,
 	.sw_init = soc21_common_sw_init,
-	.sw_fini = soc21_common_sw_fini,
 	.hw_init = soc21_common_hw_init,
 	.hw_fini = soc21_common_hw_fini,
 	.suspend = soc21_common_suspend,
 	.resume = soc21_common_resume,
 	.is_idle = soc21_common_is_idle,
-	.wait_for_idle = soc21_common_wait_for_idle,
-	.soft_reset = soc21_common_soft_reset,
 	.set_clockgating_state = soc21_common_set_clockgating_state,
 	.set_powergating_state = soc21_common_set_powergating_state,
 	.get_clockgating_state = soc21_common_get_clockgating_state,
diff --git a/drivers/gpu/drm/amd/amdgpu/soc24.c b/drivers/gpu/drm/amd/amdgpu/soc24.c
new file mode 100644
index 000000000000..972b449ab89f
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/soc24.c
@@ -0,0 +1,601 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include <linux/firmware.h>
+#include <linux/slab.h>
+#include <linux/module.h>
+#include <linux/pci.h>
+
+#include "amdgpu.h"
+#include "amdgpu_ih.h"
+#include "amdgpu_uvd.h"
+#include "amdgpu_vce.h"
+#include "amdgpu_ucode.h"
+#include "amdgpu_psp.h"
+#include "amdgpu_smu.h"
+#include "atom.h"
+#include "amd_pcie.h"
+
+#include "gc/gc_12_0_0_offset.h"
+#include "gc/gc_12_0_0_sh_mask.h"
+#include "mp/mp_14_0_2_offset.h"
+
+#include "soc15.h"
+#include "soc15_common.h"
+#include "soc24.h"
+#include "mxgpu_nv.h"
+
+static const struct amd_ip_funcs soc24_common_ip_funcs;
+
+static const struct amdgpu_video_codec_info vcn_5_0_0_video_codecs_encode_array_vcn0[] = {
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_AV1, 8192, 4352, 0)},
+};
+
+static const struct amdgpu_video_codecs vcn_5_0_0_video_codecs_encode_vcn0 = {
+	.codec_count = ARRAY_SIZE(vcn_5_0_0_video_codecs_encode_array_vcn0),
+	.codec_array = vcn_5_0_0_video_codecs_encode_array_vcn0,
+};
+
+static const struct amdgpu_video_codec_info vcn_5_0_0_video_codecs_decode_array_vcn0[] = {
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 16384, 16384, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)},
+	{codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_AV1, 8192, 4352, 0)},
+};
+
+static const struct amdgpu_video_codecs vcn_5_0_0_video_codecs_decode_vcn0 = {
+	.codec_count = ARRAY_SIZE(vcn_5_0_0_video_codecs_decode_array_vcn0),
+	.codec_array = vcn_5_0_0_video_codecs_decode_array_vcn0,
+};
+
+static int soc24_query_video_codecs(struct amdgpu_device *adev, bool encode,
+				 const struct amdgpu_video_codecs **codecs)
+{
+	if (adev->vcn.num_vcn_inst == hweight8(adev->vcn.harvest_config))
+		return -EINVAL;
+
+	switch (amdgpu_ip_version(adev, UVD_HWIP, 0)) {
+	case IP_VERSION(5, 0, 0):
+		if (encode)
+			*codecs = &vcn_5_0_0_video_codecs_encode_vcn0;
+		else
+			*codecs = &vcn_5_0_0_video_codecs_decode_vcn0;
+		return 0;
+	default:
+		return -EINVAL;
+	}
+}
+
+static u32 soc24_get_config_memsize(struct amdgpu_device *adev)
+{
+	return adev->nbio.funcs->get_memsize(adev);
+}
+
+static u32 soc24_get_xclk(struct amdgpu_device *adev)
+{
+	return adev->clock.spll.reference_freq;
+}
+
+void soc24_grbm_select(struct amdgpu_device *adev,
+		       u32 me, u32 pipe, u32 queue, u32 vmid)
+{
+	u32 grbm_gfx_cntl = 0;
+	grbm_gfx_cntl = REG_SET_FIELD(grbm_gfx_cntl, GRBM_GFX_CNTL, PIPEID, pipe);
+	grbm_gfx_cntl = REG_SET_FIELD(grbm_gfx_cntl, GRBM_GFX_CNTL, MEID, me);
+	grbm_gfx_cntl = REG_SET_FIELD(grbm_gfx_cntl, GRBM_GFX_CNTL, VMID, vmid);
+	grbm_gfx_cntl = REG_SET_FIELD(grbm_gfx_cntl, GRBM_GFX_CNTL, QUEUEID, queue);
+
+	WREG32_SOC15(GC, 0, regGRBM_GFX_CNTL, grbm_gfx_cntl);
+}
+
+static struct soc15_allowed_register_entry soc24_allowed_read_registers[] = {
+	{ SOC15_REG_ENTRY(GC, 0, regGRBM_STATUS)},
+	{ SOC15_REG_ENTRY(GC, 0, regGRBM_STATUS2)},
+	{ SOC15_REG_ENTRY(GC, 0, regGRBM_STATUS_SE0)},
+	{ SOC15_REG_ENTRY(GC, 0, regGRBM_STATUS_SE1)},
+	{ SOC15_REG_ENTRY(GC, 0, regGRBM_STATUS_SE2)},
+	{ SOC15_REG_ENTRY(GC, 0, regGRBM_STATUS_SE3)},
+	{ SOC15_REG_ENTRY(SDMA0, 0, regSDMA0_STATUS_REG)},
+	{ SOC15_REG_ENTRY(SDMA1, 0, regSDMA1_STATUS_REG)},
+	{ SOC15_REG_ENTRY(GC, 0, regCP_STAT)},
+	{ SOC15_REG_ENTRY(GC, 0, regCP_STALLED_STAT1)},
+	{ SOC15_REG_ENTRY(GC, 0, regCP_STALLED_STAT2)},
+	{ SOC15_REG_ENTRY(GC, 0, regCP_STALLED_STAT3)},
+	{ SOC15_REG_ENTRY(GC, 0, regCP_CPF_BUSY_STAT)},
+	{ SOC15_REG_ENTRY(GC, 0, regCP_CPF_STALLED_STAT1)},
+	{ SOC15_REG_ENTRY(GC, 0, regCP_CPF_STATUS)},
+	{ SOC15_REG_ENTRY(GC, 0, regCP_CPC_BUSY_STAT)},
+	{ SOC15_REG_ENTRY(GC, 0, regCP_CPC_STALLED_STAT1)},
+	{ SOC15_REG_ENTRY(GC, 0, regCP_CPC_STATUS)},
+	{ SOC15_REG_ENTRY(GC, 0, regGB_ADDR_CONFIG)},
+};
+
+static uint32_t soc24_read_indexed_register(struct amdgpu_device *adev,
+					    u32 se_num,
+					    u32 sh_num,
+					    u32 reg_offset)
+{
+	uint32_t val;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+	if (se_num != 0xffffffff || sh_num != 0xffffffff)
+		amdgpu_gfx_select_se_sh(adev, se_num, sh_num, 0xffffffff, 0);
+
+	val = RREG32(reg_offset);
+
+	if (se_num != 0xffffffff || sh_num != 0xffffffff)
+		amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
+	mutex_unlock(&adev->grbm_idx_mutex);
+	return val;
+}
+
+static uint32_t soc24_get_register_value(struct amdgpu_device *adev,
+					 bool indexed, u32 se_num,
+					 u32 sh_num, u32 reg_offset)
+{
+	if (indexed) {
+		return soc24_read_indexed_register(adev, se_num, sh_num, reg_offset);
+	} else {
+		if (reg_offset == SOC15_REG_OFFSET(GC, 0, regGB_ADDR_CONFIG) &&
+		    adev->gfx.config.gb_addr_config)
+			return adev->gfx.config.gb_addr_config;
+		return RREG32(reg_offset);
+	}
+}
+
+static int soc24_read_register(struct amdgpu_device *adev, u32 se_num,
+			       u32 sh_num, u32 reg_offset, u32 *value)
+{
+	uint32_t i;
+	struct soc15_allowed_register_entry  *en;
+
+	*value = 0;
+	for (i = 0; i < ARRAY_SIZE(soc24_allowed_read_registers); i++) {
+		en = &soc24_allowed_read_registers[i];
+		if (!adev->reg_offset[en->hwip][en->inst])
+			continue;
+		else if (reg_offset != (adev->reg_offset[en->hwip][en->inst][en->seg]
+					+ en->reg_offset))
+			continue;
+
+		*value = soc24_get_register_value(adev,
+				soc24_allowed_read_registers[i].grbm_indexed,
+				se_num, sh_num, reg_offset);
+		return 0;
+	}
+	return -EINVAL;
+}
+
+static enum amd_reset_method
+soc24_asic_reset_method(struct amdgpu_device *adev)
+{
+	if (amdgpu_reset_method == AMD_RESET_METHOD_MODE1 ||
+	    amdgpu_reset_method == AMD_RESET_METHOD_MODE2 ||
+	    amdgpu_reset_method == AMD_RESET_METHOD_BACO)
+		return amdgpu_reset_method;
+
+	if (amdgpu_reset_method != -1)
+		dev_warn(adev->dev,
+			 "Specified reset method:%d isn't supported, using AUTO instead.\n",
+			 amdgpu_reset_method);
+
+	switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
+	case IP_VERSION(14, 0, 2):
+	case IP_VERSION(14, 0, 3):
+		return AMD_RESET_METHOD_MODE1;
+	default:
+		if (amdgpu_dpm_is_baco_supported(adev))
+			return AMD_RESET_METHOD_BACO;
+		else
+			return AMD_RESET_METHOD_MODE1;
+	}
+}
+
+static int soc24_asic_reset(struct amdgpu_device *adev)
+{
+	int ret = 0;
+
+	switch (soc24_asic_reset_method(adev)) {
+	case AMD_RESET_METHOD_PCI:
+		dev_info(adev->dev, "PCI reset\n");
+		ret = amdgpu_device_pci_reset(adev);
+		break;
+	case AMD_RESET_METHOD_BACO:
+		dev_info(adev->dev, "BACO reset\n");
+		ret = amdgpu_dpm_baco_reset(adev);
+		break;
+	case AMD_RESET_METHOD_MODE2:
+		dev_info(adev->dev, "MODE2 reset\n");
+		ret = amdgpu_dpm_mode2_reset(adev);
+		break;
+	default:
+		dev_info(adev->dev, "MODE1 reset\n");
+		ret = amdgpu_device_mode1_reset(adev);
+		break;
+	}
+
+	return ret;
+}
+
+static void soc24_program_aspm(struct amdgpu_device *adev)
+{
+	if (!amdgpu_device_should_use_aspm(adev))
+		return;
+
+	if (!(adev->flags & AMD_IS_APU) &&
+	    (adev->nbio.funcs->program_aspm))
+		adev->nbio.funcs->program_aspm(adev);
+}
+
+const struct amdgpu_ip_block_version soc24_common_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_COMMON,
+	.major = 1,
+	.minor = 0,
+	.rev = 0,
+	.funcs = &soc24_common_ip_funcs,
+};
+
+static bool soc24_need_full_reset(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+	default:
+		return true;
+	}
+}
+
+static bool soc24_need_reset_on_init(struct amdgpu_device *adev)
+{
+	u32 sol_reg;
+
+	if (adev->flags & AMD_IS_APU)
+		return false;
+
+	/* Check sOS sign of life register to confirm sys driver and sOS
+	 * are already been loaded.
+	 */
+	sol_reg = RREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_81);
+	if (sol_reg)
+		return true;
+
+	return false;
+}
+
+static uint64_t soc24_get_pcie_replay_count(struct amdgpu_device *adev)
+{
+	/* TODO
+	 * dummy implement for pcie_replay_count sysfs interface
+	 * */
+	return 0;
+}
+
+static void soc24_init_doorbell_index(struct amdgpu_device *adev)
+{
+	adev->doorbell_index.kiq = AMDGPU_NAVI10_DOORBELL_KIQ;
+	adev->doorbell_index.mec_ring0 = AMDGPU_NAVI10_DOORBELL_MEC_RING0;
+	adev->doorbell_index.mec_ring1 = AMDGPU_NAVI10_DOORBELL_MEC_RING1;
+	adev->doorbell_index.mec_ring2 = AMDGPU_NAVI10_DOORBELL_MEC_RING2;
+	adev->doorbell_index.mec_ring3 = AMDGPU_NAVI10_DOORBELL_MEC_RING3;
+	adev->doorbell_index.mec_ring4 = AMDGPU_NAVI10_DOORBELL_MEC_RING4;
+	adev->doorbell_index.mec_ring5 = AMDGPU_NAVI10_DOORBELL_MEC_RING5;
+	adev->doorbell_index.mec_ring6 = AMDGPU_NAVI10_DOORBELL_MEC_RING6;
+	adev->doorbell_index.mec_ring7 = AMDGPU_NAVI10_DOORBELL_MEC_RING7;
+	adev->doorbell_index.userqueue_start = AMDGPU_NAVI10_DOORBELL_USERQUEUE_START;
+	adev->doorbell_index.userqueue_end = AMDGPU_NAVI10_DOORBELL_USERQUEUE_END;
+	adev->doorbell_index.gfx_ring0 = AMDGPU_NAVI10_DOORBELL_GFX_RING0;
+	adev->doorbell_index.gfx_ring1 = AMDGPU_NAVI10_DOORBELL_GFX_RING1;
+	adev->doorbell_index.gfx_userqueue_start =
+		AMDGPU_NAVI10_DOORBELL_GFX_USERQUEUE_START;
+	adev->doorbell_index.gfx_userqueue_end =
+		AMDGPU_NAVI10_DOORBELL_GFX_USERQUEUE_END;
+	adev->doorbell_index.mes_ring0 = AMDGPU_NAVI10_DOORBELL_MES_RING0;
+	adev->doorbell_index.mes_ring1 = AMDGPU_NAVI10_DOORBELL_MES_RING1;
+	adev->doorbell_index.sdma_engine[0] = AMDGPU_NAVI10_DOORBELL_sDMA_ENGINE0;
+	adev->doorbell_index.sdma_engine[1] = AMDGPU_NAVI10_DOORBELL_sDMA_ENGINE1;
+	adev->doorbell_index.ih = AMDGPU_NAVI10_DOORBELL_IH;
+	adev->doorbell_index.vcn.vcn_ring0_1 = AMDGPU_NAVI10_DOORBELL64_VCN0_1;
+	adev->doorbell_index.vcn.vcn_ring2_3 = AMDGPU_NAVI10_DOORBELL64_VCN2_3;
+	adev->doorbell_index.vcn.vcn_ring4_5 = AMDGPU_NAVI10_DOORBELL64_VCN4_5;
+	adev->doorbell_index.vcn.vcn_ring6_7 = AMDGPU_NAVI10_DOORBELL64_VCN6_7;
+	adev->doorbell_index.first_non_cp = AMDGPU_NAVI10_DOORBELL64_FIRST_NON_CP;
+	adev->doorbell_index.last_non_cp = AMDGPU_NAVI10_DOORBELL64_LAST_NON_CP;
+
+	adev->doorbell_index.max_assignment = AMDGPU_NAVI10_DOORBELL_MAX_ASSIGNMENT << 1;
+	adev->doorbell_index.sdma_doorbell_range = 20;
+}
+
+static void soc24_pre_asic_init(struct amdgpu_device *adev)
+{
+}
+
+static int soc24_update_umd_stable_pstate(struct amdgpu_device *adev,
+					  bool enter)
+{
+	if (enter)
+		amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
+	else
+		amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
+
+	if (adev->gfx.funcs->update_perfmon_mgcg)
+		adev->gfx.funcs->update_perfmon_mgcg(adev, !enter);
+
+	return 0;
+}
+
+static const struct amdgpu_asic_funcs soc24_asic_funcs = {
+	.read_bios_from_rom = &amdgpu_soc15_read_bios_from_rom,
+	.read_register = &soc24_read_register,
+	.reset = &soc24_asic_reset,
+	.reset_method = &soc24_asic_reset_method,
+	.get_xclk = &soc24_get_xclk,
+	.get_config_memsize = &soc24_get_config_memsize,
+	.init_doorbell_index = &soc24_init_doorbell_index,
+	.need_full_reset = &soc24_need_full_reset,
+	.need_reset_on_init = &soc24_need_reset_on_init,
+	.get_pcie_replay_count = &soc24_get_pcie_replay_count,
+	.supports_baco = &amdgpu_dpm_is_baco_supported,
+	.pre_asic_init = &soc24_pre_asic_init,
+	.query_video_codecs = &soc24_query_video_codecs,
+	.update_umd_stable_pstate = &soc24_update_umd_stable_pstate,
+};
+
+static int soc24_common_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	adev->nbio.funcs->set_reg_remap(adev);
+	adev->smc_rreg = NULL;
+	adev->smc_wreg = NULL;
+	adev->pcie_rreg = &amdgpu_device_indirect_rreg;
+	adev->pcie_wreg = &amdgpu_device_indirect_wreg;
+	adev->pcie_rreg64 = &amdgpu_device_indirect_rreg64;
+	adev->pcie_wreg64 = &amdgpu_device_indirect_wreg64;
+	adev->pciep_rreg = amdgpu_device_pcie_port_rreg;
+	adev->pciep_wreg = amdgpu_device_pcie_port_wreg;
+	adev->uvd_ctx_rreg = NULL;
+	adev->uvd_ctx_wreg = NULL;
+	adev->didt_rreg = NULL;
+	adev->didt_wreg = NULL;
+
+	adev->asic_funcs = &soc24_asic_funcs;
+
+	adev->rev_id = amdgpu_device_get_rev_id(adev);
+	adev->external_rev_id = 0xff;
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+		adev->cg_flags = AMD_CG_SUPPORT_GFX_CGCG |
+			AMD_CG_SUPPORT_GFX_CGLS |
+			AMD_CG_SUPPORT_GFX_MGCG |
+			AMD_CG_SUPPORT_GFX_3D_CGCG |
+			AMD_CG_SUPPORT_GFX_3D_CGLS |
+			AMD_CG_SUPPORT_REPEATER_FGCG |
+			AMD_CG_SUPPORT_GFX_FGCG |
+			AMD_CG_SUPPORT_GFX_PERF_CLK |
+			AMD_CG_SUPPORT_ATHUB_MGCG |
+			AMD_CG_SUPPORT_ATHUB_LS |
+			AMD_CG_SUPPORT_MC_MGCG |
+			AMD_CG_SUPPORT_HDP_SD |
+			AMD_CG_SUPPORT_MC_LS;
+		adev->pg_flags = AMD_PG_SUPPORT_VCN |
+			AMD_PG_SUPPORT_JPEG |
+			AMD_PG_SUPPORT_VCN_DPG;
+		adev->external_rev_id = adev->rev_id + 0x40;
+		break;
+	case IP_VERSION(12, 0, 1):
+		adev->cg_flags = AMD_CG_SUPPORT_GFX_CGCG |
+			AMD_CG_SUPPORT_GFX_CGLS |
+			AMD_CG_SUPPORT_GFX_MGCG |
+			AMD_CG_SUPPORT_GFX_3D_CGCG |
+			AMD_CG_SUPPORT_GFX_3D_CGLS |
+			AMD_CG_SUPPORT_REPEATER_FGCG |
+			AMD_CG_SUPPORT_GFX_FGCG |
+			AMD_CG_SUPPORT_GFX_PERF_CLK |
+			AMD_CG_SUPPORT_ATHUB_MGCG |
+			AMD_CG_SUPPORT_ATHUB_LS |
+			AMD_CG_SUPPORT_MC_MGCG |
+			AMD_CG_SUPPORT_HDP_SD |
+			AMD_CG_SUPPORT_MC_LS;
+
+		adev->pg_flags = AMD_PG_SUPPORT_VCN |
+			AMD_PG_SUPPORT_JPEG |
+			AMD_PG_SUPPORT_JPEG_DPG |
+			AMD_PG_SUPPORT_VCN_DPG;
+		adev->external_rev_id = adev->rev_id + 0x50;
+		break;
+	default:
+		/* FIXME: not supported yet */
+		return -EINVAL;
+	}
+
+	if (amdgpu_sriov_vf(adev)) {
+		amdgpu_virt_init_setting(adev);
+		xgpu_nv_mailbox_set_irq_funcs(adev);
+	}
+
+	return 0;
+}
+
+static int soc24_common_late_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (amdgpu_sriov_vf(adev)) {
+		xgpu_nv_mailbox_get_irq(adev);
+	} else {
+		if (adev->nbio.ras &&
+		    adev->nbio.ras_err_event_athub_irq.funcs)
+			/* don't need to fail gpu late init
+			 * if enabling athub_err_event interrupt failed
+			 * nbif v6_3_1 only support fatal error hanlding
+			 * just enable the interrupt directly
+			 */
+			amdgpu_irq_get(adev, &adev->nbio.ras_err_event_athub_irq, 0);
+	}
+
+	/* Enable selfring doorbell aperture late because doorbell BAR
+	 * aperture will change if resize BAR successfully in gmc sw_init.
+	 */
+	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, true);
+
+	return 0;
+}
+
+static int soc24_common_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (amdgpu_sriov_vf(adev))
+		xgpu_nv_mailbox_add_irq_id(adev);
+
+	return 0;
+}
+
+static int soc24_common_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	/* enable aspm */
+	soc24_program_aspm(adev);
+	/* setup nbio registers */
+	adev->nbio.funcs->init_registers(adev);
+	/* remap HDP registers to a hole in mmio space,
+	 * for the purpose of expose those registers
+	 * to process space
+	 */
+	if (adev->nbio.funcs->remap_hdp_registers)
+		adev->nbio.funcs->remap_hdp_registers(adev);
+
+	if (adev->df.funcs->hw_init)
+		adev->df.funcs->hw_init(adev);
+
+	/* enable the doorbell aperture */
+	adev->nbio.funcs->enable_doorbell_aperture(adev, true);
+
+	return 0;
+}
+
+static int soc24_common_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	/* Disable the doorbell aperture and selfring doorbell aperture
+	 * separately in hw_fini because soc21_enable_doorbell_aperture
+	 * has been removed and there is no need to delay disabling
+	 * selfring doorbell.
+	 */
+	adev->nbio.funcs->enable_doorbell_aperture(adev, false);
+	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, false);
+
+	if (amdgpu_sriov_vf(adev)) {
+		xgpu_nv_mailbox_put_irq(adev);
+	} else {
+		if (adev->nbio.ras &&
+		    adev->nbio.ras_err_event_athub_irq.funcs)
+			amdgpu_irq_put(adev, &adev->nbio.ras_err_event_athub_irq, 0);
+	}
+
+	return 0;
+}
+
+static int soc24_common_suspend(struct amdgpu_ip_block *ip_block)
+{
+	return soc24_common_hw_fini(ip_block);
+}
+
+static int soc24_common_resume(struct amdgpu_ip_block *ip_block)
+{
+	return soc24_common_hw_init(ip_block);
+}
+
+static bool soc24_common_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	return true;
+}
+
+static int soc24_common_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					      enum amd_clockgating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	switch (amdgpu_ip_version(adev, NBIO_HWIP, 0)) {
+	case IP_VERSION(6, 3, 1):
+		adev->nbio.funcs->update_medium_grain_clock_gating(adev,
+				state == AMD_CG_STATE_GATE);
+		adev->nbio.funcs->update_medium_grain_light_sleep(adev,
+				state == AMD_CG_STATE_GATE);
+		adev->hdp.funcs->update_clock_gating(adev,
+				state == AMD_CG_STATE_GATE);
+		break;
+	default:
+		break;
+	}
+	return 0;
+}
+
+static int soc24_common_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					      enum amd_powergating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	switch (amdgpu_ip_version(adev, LSDMA_HWIP, 0)) {
+	case IP_VERSION(7, 0, 0):
+	case IP_VERSION(7, 0, 1):
+		adev->lsdma.funcs->update_memory_power_gating(adev,
+				state == AMD_PG_STATE_GATE);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static void soc24_common_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	adev->nbio.funcs->get_clockgating_state(adev, flags);
+
+	adev->hdp.funcs->get_clock_gating_state(adev, flags);
+
+	return;
+}
+
+static const struct amd_ip_funcs soc24_common_ip_funcs = {
+	.name = "soc24_common",
+	.early_init = soc24_common_early_init,
+	.late_init = soc24_common_late_init,
+	.sw_init = soc24_common_sw_init,
+	.hw_init = soc24_common_hw_init,
+	.hw_fini = soc24_common_hw_fini,
+	.suspend = soc24_common_suspend,
+	.resume = soc24_common_resume,
+	.is_idle = soc24_common_is_idle,
+	.set_clockgating_state = soc24_common_set_clockgating_state,
+	.set_powergating_state = soc24_common_set_powergating_state,
+	.get_clockgating_state = soc24_common_get_clockgating_state,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/soc24.h b/drivers/gpu/drm/amd/amdgpu/soc24.h
new file mode 100644
index 000000000000..fa7e442e0b62
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/soc24.h
@@ -0,0 +1,30 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __SOC24_H__
+#define __SOC24_H__
+
+extern const struct amdgpu_ip_block_version soc24_common_ip_block;
+
+void soc24_grbm_select(struct amdgpu_device *adev,
+		       u32 me, u32 pipe, u32 queue, u32 vmid);
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h b/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h
index 30d0482ac466..8a3f326474e5 100644
--- a/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h
+++ b/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h
@@ -30,12 +30,18 @@
 #define RSP_ID_MASK (1U << 31)
 #define RSP_ID(cmdId) (((uint32_t)(cmdId)) | RSP_ID_MASK)
 
+/* invalid node instance value */
+#define TA_RAS_INV_NODE 0xffff
+
 /* RAS related enumerations */
 /**********************************************************/
 enum ras_command {
 	TA_RAS_COMMAND__ENABLE_FEATURES = 0,
 	TA_RAS_COMMAND__DISABLE_FEATURES,
 	TA_RAS_COMMAND__TRIGGER_ERROR,
+	TA_RAS_COMMAND__QUERY_BLOCK_INFO,
+	TA_RAS_COMMAND__QUERY_SUB_BLOCK_INFO,
+	TA_RAS_COMMAND__QUERY_ADDRESS,
 };
 
 enum ta_ras_status {
@@ -86,6 +92,9 @@ enum ta_ras_block {
 	TA_RAS_BLOCK__MCA,
 	TA_RAS_BLOCK__VCN,
 	TA_RAS_BLOCK__JPEG,
+	TA_RAS_BLOCK__IH,
+	TA_RAS_BLOCK__MPIO,
+	TA_RAS_BLOCK__MMSCH,
 	TA_NUM_BLOCK_MAX
 };
 
@@ -105,6 +114,19 @@ enum ta_ras_error_type {
 	TA_RAS_ERROR__POISON			= 8,
 };
 
+enum ta_ras_address_type {
+	TA_RAS_MCA_TO_PA,
+	TA_RAS_PA_TO_MCA,
+};
+
+enum ta_ras_nps_mode {
+	TA_RAS_UNKNOWN_MODE = 0,
+	TA_RAS_NPS1_MODE = 1,
+	TA_RAS_NPS2_MODE = 2,
+	TA_RAS_NPS4_MODE = 4,
+	TA_RAS_NPS8_MODE = 8,
+};
+
 /* Input/output structures for RAS commands */
 /**********************************************************/
 
@@ -129,6 +151,30 @@ struct ta_ras_trigger_error_input {
 struct ta_ras_init_flags {
 	uint8_t poison_mode_en;
 	uint8_t dgpu_mode;
+	uint16_t xcc_mask;
+	uint8_t channel_dis_num;
+	uint8_t nps_mode;
+	uint32_t active_umc_mask;
+};
+
+struct ta_ras_mca_addr {
+	uint64_t err_addr;
+	uint32_t ch_inst;
+	uint32_t umc_inst;
+	uint32_t node_inst;
+	uint32_t socket_id;
+};
+
+struct ta_ras_phy_addr {
+	uint64_t pa;
+	uint32_t bank;
+	uint32_t channel_idx;
+};
+
+struct ta_ras_query_address_input {
+	enum ta_ras_address_type addr_type;
+	struct ta_ras_mca_addr ma;
+	struct ta_ras_phy_addr pa;
 };
 
 struct ta_ras_output_flags {
@@ -137,6 +183,13 @@ struct ta_ras_output_flags {
 	uint8_t reg_access_failure_flag;
 };
 
+struct ta_ras_query_address_output {
+	/* don't use the flags here */
+	struct ta_ras_output_flags flags;
+	struct ta_ras_mca_addr ma;
+	struct ta_ras_phy_addr pa;
+};
+
 /* Common input structure for RAS callbacks */
 /**********************************************************/
 union ta_ras_cmd_input {
@@ -144,12 +197,14 @@ union ta_ras_cmd_input {
 	struct ta_ras_enable_features_input	enable_features;
 	struct ta_ras_disable_features_input	disable_features;
 	struct ta_ras_trigger_error_input	trigger_error;
+	struct ta_ras_query_address_input	address;
 
 	uint32_t reserve_pad[256];
 };
 
 union ta_ras_cmd_output {
 	struct ta_ras_output_flags flags;
+	struct ta_ras_query_address_output address;
 
 	uint32_t reserve_pad[256];
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/ta_secureDisplay_if.h b/drivers/gpu/drm/amd/amdgpu/ta_secureDisplay_if.h
index 00d8bdb8254f..9ec2e03d41c7 100644
--- a/drivers/gpu/drm/amd/amdgpu/ta_secureDisplay_if.h
+++ b/drivers/gpu/drm/amd/amdgpu/ta_secureDisplay_if.h
@@ -31,10 +31,12 @@
  *    Secure Display Command ID
  */
 enum ta_securedisplay_command {
-	/* Query whether TA is responding used only for validation purpose */
+	/* Query whether TA is responding. It is used only for validation purpose */
 	TA_SECUREDISPLAY_COMMAND__QUERY_TA              = 1,
 	/* Send region of Interest and CRC value to I2C */
 	TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC          = 2,
+	/* V2 to send multiple regions of Interest and CRC value to I2C */
+	TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC_V2       = 3,
 	/* Maximum Command ID */
 	TA_SECUREDISPLAY_COMMAND__MAX_ID                = 0x7FFFFFFF,
 };
@@ -83,6 +85,8 @@ enum ta_securedisplay_ta_query_cmd_ret {
 enum ta_securedisplay_buffer_size {
 	/* 15 bytes = 8 byte (ROI) + 6 byte(CRC) + 1 byte(phy_id) */
 	TA_SECUREDISPLAY_I2C_BUFFER_SIZE                = 15,
+	/* 16 bytes = 8 byte (ROI) + 6 byte(CRC) + 1 byte(phy_id) + 1 byte(roi_idx) */
+	TA_SECUREDISPLAY_V2_I2C_BUFFER_SIZE             = 16,
 };
 
 /** Input/output structures for Secure Display commands */
@@ -95,7 +99,15 @@ enum ta_securedisplay_buffer_size {
  *    Physical ID to determine which DIO scratch register should be used to get ROI
  */
 struct ta_securedisplay_send_roi_crc_input {
-	uint32_t  phy_id;  /* Physical ID */
+	/* Physical ID */
+	uint32_t  phy_id;
+};
+
+struct ta_securedisplay_send_roi_crc_v2_input {
+	/* Physical ID */
+	uint32_t phy_id;
+	/* Region of interest index */
+	uint8_t  roi_idx;
 };
 
 /** @union ta_securedisplay_cmd_input
@@ -104,6 +116,8 @@ struct ta_securedisplay_send_roi_crc_input {
 union ta_securedisplay_cmd_input {
 	/* send ROI and CRC input buffer format */
 	struct ta_securedisplay_send_roi_crc_input        send_roi_crc;
+	/* send ROI and CRC input buffer format, v2 adds a ROI index */
+	struct ta_securedisplay_send_roi_crc_v2_input     send_roi_crc_v2;
 	uint32_t                                          reserved[4];
 };
 
@@ -128,6 +142,10 @@ struct ta_securedisplay_send_roi_crc_output {
 	uint8_t  reserved;
 };
 
+struct ta_securedisplay_send_roi_crc_v2_output {
+	uint8_t  i2c_buf[TA_SECUREDISPLAY_V2_I2C_BUFFER_SIZE];  /* I2C buffer */
+};
+
 /** @union ta_securedisplay_cmd_output
  *    Output buffer
  */
@@ -136,6 +154,8 @@ union ta_securedisplay_cmd_output {
 	struct ta_securedisplay_query_ta_output            query_ta;
 	/* Send ROI CRC output buffer format used only for validation purpose */
 	struct ta_securedisplay_send_roi_crc_output        send_roi_crc;
+	/* Send ROI CRC output buffer format used only for validation purpose */
+	struct ta_securedisplay_send_roi_crc_v2_output     send_roi_crc_v2;
 	uint32_t                                           reserved[4];
 };
 
diff --git a/drivers/gpu/drm/amd/amdgpu/ta_xgmi_if.h b/drivers/gpu/drm/amd/amdgpu/ta_xgmi_if.h
index da815a93d46e..d5748032674e 100644
--- a/drivers/gpu/drm/amd/amdgpu/ta_xgmi_if.h
+++ b/drivers/gpu/drm/amd/amdgpu/ta_xgmi_if.h
@@ -1,5 +1,5 @@
 /*
- * Copyright 2018 Advanced Micro Devices, Inc.
+ * Copyright 2018-2022 Advanced Micro Devices, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -20,7 +20,6 @@
  * OTHER DEALINGS IN THE SOFTWARE.
  *
  */
-
 #ifndef _TA_XGMI_IF_H
 #define _TA_XGMI_IF_H
 
@@ -28,20 +27,31 @@
 #define RSP_ID_MASK (1U << 31)
 #define RSP_ID(cmdId) (((uint32_t)(cmdId)) | RSP_ID_MASK)
 
+#define EXTEND_PEER_LINK_INFO_CMD_FLAG 1
+
 enum ta_command_xgmi {
+	/* Initialize the Context and Session Topology */
 	TA_COMMAND_XGMI__INITIALIZE			= 0x00,
+	/* Gets the current GPU's node ID */
 	TA_COMMAND_XGMI__GET_NODE_ID			= 0x01,
+	/* Gets the current GPU's hive ID */
 	TA_COMMAND_XGMI__GET_HIVE_ID			= 0x02,
-	TA_COMMAND_XGMI__GET_GET_TOPOLOGY_INFO		= 0x03,
+	/* Gets the Peer's topology Information */
+	TA_COMMAND_XGMI__GET_TOPOLOGY_INFO		= 0x03,
+	/* Sets the Peer's topology Information */
 	TA_COMMAND_XGMI__SET_TOPOLOGY_INFO		= 0x04,
-	TA_COMMAND_XGMI__GET_PEER_LINKS			= 0x0B
+	/* Gets the total links between adjacent peer dies in hive */
+	TA_COMMAND_XGMI__GET_PEER_LINKS			= 0x0B,
+	/* Gets the total links and connected port numbers between adjacent peer dies in hive */
+	TA_COMMAND_XGMI__GET_EXTEND_PEER_LINKS		= 0x0C
 };
 
 /* XGMI related enumerations */
 /**********************************************************/;
-enum ta_xgmi_connected_nodes {
-	TA_XGMI__MAX_CONNECTED_NODES			= 64
-};
+enum { TA_XGMI__MAX_CONNECTED_NODES = 64 };
+enum { TA_XGMI__MAX_INTERNAL_STATE = 32 };
+enum { TA_XGMI__MAX_INTERNAL_STATE_BUFFER = 128 };
+enum { TA_XGMI__MAX_PORT_NUM = 8 };
 
 enum ta_xgmi_status {
 	TA_XGMI_STATUS__SUCCESS				= 0x00,
@@ -81,6 +91,18 @@ struct ta_xgmi_peer_link_info {
 	uint8_t					num_links;
 };
 
+struct xgmi_connected_port_num {
+	uint8_t		dst_xgmi_port_num;
+	uint8_t		src_xgmi_port_num;
+};
+
+/* support both the port num and num_links */
+struct ta_xgmi_extend_peer_link_info {
+	uint64_t				node_id;
+	uint8_t					num_links;
+	struct xgmi_connected_port_num		port_num[TA_XGMI__MAX_PORT_NUM];
+};
+
 struct ta_xgmi_cmd_initialize_output {
 	uint32_t	status;
 };
@@ -103,16 +125,21 @@ struct ta_xgmi_cmd_get_topology_info_output {
 	struct ta_xgmi_node_info	nodes[TA_XGMI__MAX_CONNECTED_NODES];
 };
 
-struct ta_xgmi_cmd_get_peer_link_info_output {
+struct ta_xgmi_cmd_set_topology_info_input {
 	uint32_t			num_nodes;
-	struct ta_xgmi_peer_link_info	nodes[TA_XGMI__MAX_CONNECTED_NODES];
+	struct ta_xgmi_node_info	nodes[TA_XGMI__MAX_CONNECTED_NODES];
 };
 
-struct ta_xgmi_cmd_set_topology_info_input {
+/* support XGMI TA w/ and w/o port_num both so two similar structs defined */
+struct ta_xgmi_cmd_get_peer_link_info {
 	uint32_t			num_nodes;
-	struct ta_xgmi_node_info	nodes[TA_XGMI__MAX_CONNECTED_NODES];
+	struct ta_xgmi_peer_link_info	nodes[TA_XGMI__MAX_CONNECTED_NODES];
 };
 
+struct ta_xgmi_cmd_get_extend_peer_link_info {
+	uint32_t				num_nodes;
+	struct ta_xgmi_extend_peer_link_info nodes[TA_XGMI__MAX_CONNECTED_NODES];
+};
 /**********************************************************/
 /* Common input structure for XGMI callbacks */
 union ta_xgmi_cmd_input {
@@ -126,16 +153,23 @@ union ta_xgmi_cmd_output {
 	struct ta_xgmi_cmd_get_node_id_output		get_node_id;
 	struct ta_xgmi_cmd_get_hive_id_output		get_hive_id;
 	struct ta_xgmi_cmd_get_topology_info_output	get_topology_info;
-	struct ta_xgmi_cmd_get_peer_link_info_output	get_link_info;
+	struct ta_xgmi_cmd_get_peer_link_info		get_link_info;
+	struct ta_xgmi_cmd_get_extend_peer_link_info	get_extend_link_info;
 };
-/**********************************************************/
 
 struct ta_xgmi_shared_memory {
 	uint32_t			cmd_id;
 	uint32_t			resp_id;
 	enum ta_xgmi_status		xgmi_status;
+
+	/* if the number of xgmi link record is more than 128, driver will set the
+	 * flag 0 to get the first 128 of the link records and will set to 1, to get
+	 * the second set
+	 */
 	uint8_t				flag_extend_link_record;
-	uint8_t				reserved0[3];
+	/* bit0: port_num info support flag for GET_EXTEND_PEER_LINKS commmand */
+	uint8_t				caps_flag;
+	uint8_t				reserved[2];
 	union ta_xgmi_cmd_input		xgmi_in_message;
 	union ta_xgmi_cmd_output	xgmi_out_message;
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c
index b08905d1c00f..ee8038df17e3 100644
--- a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c
@@ -159,6 +159,9 @@ static int tonga_ih_irq_init(struct amdgpu_device *adev)
 	/* enable interrupts */
 	tonga_ih_enable_interrupts(adev);
 
+	if (adev->irq.ih_soft.ring_size)
+		adev->irq.ih_soft.enabled = true;
+
 	return 0;
 }
 
@@ -196,6 +199,9 @@ static u32 tonga_ih_get_wptr(struct amdgpu_device *adev,
 
 	wptr = le32_to_cpu(*ih->wptr_cpu);
 
+	if (ih == &adev->irq.ih_soft)
+		goto out;
+
 	if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
 		goto out;
 
@@ -219,6 +225,12 @@ static u32 tonga_ih_get_wptr(struct amdgpu_device *adev,
 	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
 	WREG32(mmIH_RB_CNTL, tmp);
 
+	/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+	 * can be detected.
+	 */
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+	WREG32(mmIH_RB_CNTL, tmp);
+
 out:
 	return (wptr & ih->ptr_mask);
 }
@@ -277,9 +289,9 @@ static void tonga_ih_set_rptr(struct amdgpu_device *adev,
 	}
 }
 
-static int tonga_ih_early_init(void *handle)
+static int tonga_ih_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	ret = amdgpu_irq_add_domain(adev);
@@ -291,15 +303,19 @@ static int tonga_ih_early_init(void *handle)
 	return 0;
 }
 
-static int tonga_ih_sw_init(void *handle)
+static int tonga_ih_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 64 * 1024, true);
 	if (r)
 		return r;
 
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, IH_SW_RING_SIZE, true);
+	if (r)
+		return r;
+
 	adev->irq.ih.use_doorbell = true;
 	adev->irq.ih.doorbell_index = adev->doorbell_index.ih;
 
@@ -308,9 +324,9 @@ static int tonga_ih_sw_init(void *handle)
 	return r;
 }
 
-static int tonga_ih_sw_fini(void *handle)
+static int tonga_ih_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_fini_sw(adev);
 	amdgpu_irq_remove_domain(adev);
@@ -318,10 +334,10 @@ static int tonga_ih_sw_fini(void *handle)
 	return 0;
 }
 
-static int tonga_ih_hw_init(void *handle)
+static int tonga_ih_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = tonga_ih_irq_init(adev);
 	if (r)
@@ -330,32 +346,26 @@ static int tonga_ih_hw_init(void *handle)
 	return 0;
 }
 
-static int tonga_ih_hw_fini(void *handle)
+static int tonga_ih_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	tonga_ih_irq_disable(adev);
+	tonga_ih_irq_disable(ip_block->adev);
 
 	return 0;
 }
 
-static int tonga_ih_suspend(void *handle)
+static int tonga_ih_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return tonga_ih_hw_fini(adev);
+	return tonga_ih_hw_fini(ip_block);
 }
 
-static int tonga_ih_resume(void *handle)
+static int tonga_ih_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return tonga_ih_hw_init(adev);
+	return tonga_ih_hw_init(ip_block);
 }
 
-static bool tonga_ih_is_idle(void *handle)
+static bool tonga_ih_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (REG_GET_FIELD(tmp, SRBM_STATUS, IH_BUSY))
@@ -364,11 +374,11 @@ static bool tonga_ih_is_idle(void *handle)
 	return true;
 }
 
-static int tonga_ih_wait_for_idle(void *handle)
+static int tonga_ih_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		/* read MC_STATUS */
@@ -380,9 +390,9 @@ static int tonga_ih_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static bool tonga_ih_check_soft_reset(void *handle)
+static bool tonga_ih_check_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 srbm_soft_reset = 0;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
@@ -399,29 +409,27 @@ static bool tonga_ih_check_soft_reset(void *handle)
 	}
 }
 
-static int tonga_ih_pre_soft_reset(void *handle)
+static int tonga_ih_pre_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	if (!adev->irq.srbm_soft_reset)
+	if (!ip_block->adev->irq.srbm_soft_reset)
 		return 0;
 
-	return tonga_ih_hw_fini(adev);
+	return tonga_ih_hw_fini(ip_block);
 }
 
-static int tonga_ih_post_soft_reset(void *handle)
+static int tonga_ih_post_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (!adev->irq.srbm_soft_reset)
 		return 0;
 
-	return tonga_ih_hw_init(adev);
+	return tonga_ih_hw_init(ip_block);
 }
 
-static int tonga_ih_soft_reset(void *handle)
+static int tonga_ih_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 srbm_soft_reset;
 
 	if (!adev->irq.srbm_soft_reset)
@@ -450,13 +458,13 @@ static int tonga_ih_soft_reset(void *handle)
 	return 0;
 }
 
-static int tonga_ih_set_clockgating_state(void *handle,
+static int tonga_ih_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int tonga_ih_set_powergating_state(void *handle,
+static int tonga_ih_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -465,7 +473,6 @@ static int tonga_ih_set_powergating_state(void *handle,
 static const struct amd_ip_funcs tonga_ih_ip_funcs = {
 	.name = "tonga_ih",
 	.early_init = tonga_ih_early_init,
-	.late_init = NULL,
 	.sw_init = tonga_ih_sw_init,
 	.sw_fini = tonga_ih_sw_fini,
 	.hw_init = tonga_ih_hw_init,
@@ -493,8 +500,7 @@ static void tonga_ih_set_interrupt_funcs(struct amdgpu_device *adev)
 	adev->irq.ih_funcs = &tonga_ih_funcs;
 }
 
-const struct amdgpu_ip_block_version tonga_ih_ip_block =
-{
+const struct amdgpu_ip_block_version tonga_ih_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_IH,
 	.major = 3,
 	.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c
new file mode 100644
index 000000000000..0f5b1719fda5
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c
@@ -0,0 +1,742 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "umc_v12_0.h"
+#include "amdgpu_ras.h"
+#include "amdgpu_umc.h"
+#include "amdgpu.h"
+#include "umc/umc_12_0_0_offset.h"
+#include "umc/umc_12_0_0_sh_mask.h"
+#include "mp/mp_13_0_6_sh_mask.h"
+
+#define MAX_ECC_NUM_PER_RETIREMENT  32
+#define DELAYED_TIME_FOR_GPU_RESET  1000  //ms
+
+static inline uint64_t get_umc_v12_0_reg_offset(struct amdgpu_device *adev,
+					    uint32_t node_inst,
+					    uint32_t umc_inst,
+					    uint32_t ch_inst)
+{
+	uint32_t index = umc_inst * adev->umc.channel_inst_num + ch_inst;
+	uint64_t cross_node_offset = (node_inst == 0) ? 0 : UMC_V12_0_CROSS_NODE_OFFSET;
+
+	umc_inst = index / 4;
+	ch_inst = index % 4;
+
+	return adev->umc.channel_offs * ch_inst + UMC_V12_0_INST_DIST * umc_inst +
+		UMC_V12_0_NODE_DIST * node_inst + cross_node_offset;
+}
+
+static int umc_v12_0_reset_error_count_per_channel(struct amdgpu_device *adev,
+					uint32_t node_inst, uint32_t umc_inst,
+					uint32_t ch_inst, void *data)
+{
+	uint64_t odecc_err_cnt_addr;
+	uint64_t umc_reg_offset =
+		get_umc_v12_0_reg_offset(adev, node_inst, umc_inst, ch_inst);
+
+	odecc_err_cnt_addr =
+		SOC15_REG_OFFSET(UMC, 0, regUMCCH0_OdEccErrCnt);
+
+	/* clear error count */
+	WREG32_PCIE_EXT((odecc_err_cnt_addr + umc_reg_offset) * 4,
+			UMC_V12_0_CE_CNT_INIT);
+
+	return 0;
+}
+
+static void umc_v12_0_reset_error_count(struct amdgpu_device *adev)
+{
+	amdgpu_umc_loop_channels(adev,
+		umc_v12_0_reset_error_count_per_channel, NULL);
+}
+
+bool umc_v12_0_is_deferred_error(struct amdgpu_device *adev, uint64_t mc_umc_status)
+{
+	dev_dbg(adev->dev,
+		"MCA_UMC_STATUS(0x%llx): Val:%llu, Poison:%llu, Deferred:%llu, PCC:%llu, UC:%llu, TCC:%llu\n",
+		mc_umc_status,
+		REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val),
+		REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Poison),
+		REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Deferred),
+		REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, PCC),
+		REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UC),
+		REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, TCC)
+	);
+
+	return (amdgpu_ras_is_poison_mode_supported(adev) &&
+		(REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1) &&
+		((REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Deferred) == 1) ||
+		(REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Poison) == 1)));
+}
+
+bool umc_v12_0_is_uncorrectable_error(struct amdgpu_device *adev, uint64_t mc_umc_status)
+{
+	if (umc_v12_0_is_deferred_error(adev, mc_umc_status))
+		return false;
+
+	return ((REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1) &&
+		(REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, PCC) == 1 ||
+		REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UC) == 1 ||
+		REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, TCC) == 1));
+}
+
+bool umc_v12_0_is_correctable_error(struct amdgpu_device *adev, uint64_t mc_umc_status)
+{
+	if (umc_v12_0_is_deferred_error(adev, mc_umc_status))
+		return false;
+
+	return (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 &&
+		(REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, CECC) == 1 ||
+		(REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 &&
+		REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UC) == 0) ||
+		/* Identify data parity error in replay mode */
+		((REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, ErrorCodeExt) == 0x5 ||
+		REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, ErrorCodeExt) == 0xb) &&
+		!(umc_v12_0_is_uncorrectable_error(adev, mc_umc_status)))));
+}
+
+static void umc_v12_0_query_error_count_per_type(struct amdgpu_device *adev,
+						   uint64_t umc_reg_offset,
+						   unsigned long *error_count,
+						   check_error_type_func error_type_func)
+{
+	uint64_t mc_umc_status;
+	uint64_t mc_umc_status_addr;
+
+	mc_umc_status_addr =
+		SOC15_REG_OFFSET(UMC, 0, regMCA_UMC_UMC0_MCUMC_STATUST0);
+
+	/* Check MCUMC_STATUS */
+	mc_umc_status =
+		RREG64_PCIE_EXT((mc_umc_status_addr + umc_reg_offset) * 4);
+
+	if (error_type_func(adev, mc_umc_status))
+		*error_count += 1;
+}
+
+static int umc_v12_0_query_error_count(struct amdgpu_device *adev,
+					uint32_t node_inst, uint32_t umc_inst,
+					uint32_t ch_inst, void *data)
+{
+	struct ras_err_data *err_data = (struct ras_err_data *)data;
+	unsigned long ue_count = 0, ce_count = 0, de_count = 0;
+
+	/* NOTE: node_inst is converted by adev->umc.active_mask and the range is [0-3],
+	 * which can be used as die ID directly */
+	struct amdgpu_smuio_mcm_config_info mcm_info = {
+		.socket_id = adev->smuio.funcs->get_socket_id(adev),
+		.die_id = node_inst,
+	};
+
+	uint64_t umc_reg_offset =
+		get_umc_v12_0_reg_offset(adev, node_inst, umc_inst, ch_inst);
+
+	umc_v12_0_query_error_count_per_type(adev, umc_reg_offset,
+					    &ce_count, umc_v12_0_is_correctable_error);
+	umc_v12_0_query_error_count_per_type(adev, umc_reg_offset,
+					    &ue_count, umc_v12_0_is_uncorrectable_error);
+	umc_v12_0_query_error_count_per_type(adev, umc_reg_offset,
+					    &de_count, umc_v12_0_is_deferred_error);
+
+	amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, ue_count);
+	amdgpu_ras_error_statistic_ce_count(err_data, &mcm_info, ce_count);
+	amdgpu_ras_error_statistic_de_count(err_data, &mcm_info, de_count);
+
+	return 0;
+}
+
+static void umc_v12_0_query_ras_error_count(struct amdgpu_device *adev,
+					   void *ras_error_status)
+{
+	amdgpu_umc_loop_channels(adev,
+		umc_v12_0_query_error_count, ras_error_status);
+
+	umc_v12_0_reset_error_count(adev);
+}
+
+static void umc_v12_0_get_retire_flip_bits(struct amdgpu_device *adev)
+{
+	enum amdgpu_memory_partition nps = AMDGPU_NPS1_PARTITION_MODE;
+	uint32_t vram_type = adev->gmc.vram_type;
+	struct amdgpu_umc_flip_bits *flip_bits = &(adev->umc.flip_bits);
+
+	if (adev->gmc.gmc_funcs->query_mem_partition_mode)
+		nps = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
+
+	/* default setting */
+	flip_bits->flip_bits_in_pa[0] = UMC_V12_0_PA_C2_BIT;
+	flip_bits->flip_bits_in_pa[1] = UMC_V12_0_PA_C3_BIT;
+	flip_bits->flip_bits_in_pa[2] = UMC_V12_0_PA_C4_BIT;
+	flip_bits->flip_bits_in_pa[3] = UMC_V12_0_PA_R13_BIT;
+	flip_bits->flip_row_bit = 13;
+	flip_bits->bit_num = 4;
+	flip_bits->r13_in_pa = UMC_V12_0_PA_R13_BIT;
+
+	if (nps == AMDGPU_NPS2_PARTITION_MODE) {
+		flip_bits->flip_bits_in_pa[0] = UMC_V12_0_PA_CH5_BIT;
+		flip_bits->flip_bits_in_pa[1] = UMC_V12_0_PA_C2_BIT;
+		flip_bits->flip_bits_in_pa[2] = UMC_V12_0_PA_B1_BIT;
+		flip_bits->r13_in_pa = UMC_V12_0_PA_R12_BIT;
+	} else if (nps == AMDGPU_NPS4_PARTITION_MODE) {
+		flip_bits->flip_bits_in_pa[0] = UMC_V12_0_PA_CH4_BIT;
+		flip_bits->flip_bits_in_pa[1] = UMC_V12_0_PA_CH5_BIT;
+		flip_bits->flip_bits_in_pa[2] = UMC_V12_0_PA_B0_BIT;
+		flip_bits->r13_in_pa = UMC_V12_0_PA_R11_BIT;
+	}
+
+	switch (vram_type) {
+	case AMDGPU_VRAM_TYPE_HBM:
+		/* other nps modes are taken as nps1 */
+		if (nps == AMDGPU_NPS2_PARTITION_MODE)
+			flip_bits->flip_bits_in_pa[3] = UMC_V12_0_PA_R12_BIT;
+		else if (nps == AMDGPU_NPS4_PARTITION_MODE)
+			flip_bits->flip_bits_in_pa[3] = UMC_V12_0_PA_R11_BIT;
+
+		break;
+	case AMDGPU_VRAM_TYPE_HBM3E:
+		flip_bits->flip_bits_in_pa[3] = UMC_V12_0_PA_R12_BIT;
+		flip_bits->flip_row_bit = 12;
+
+		if (nps == AMDGPU_NPS2_PARTITION_MODE)
+			flip_bits->flip_bits_in_pa[3] = UMC_V12_0_PA_R11_BIT;
+		else if (nps == AMDGPU_NPS4_PARTITION_MODE)
+			flip_bits->flip_bits_in_pa[3] = UMC_V12_0_PA_R10_BIT;
+
+		break;
+	default:
+		dev_warn(adev->dev,
+			"Unknown HBM type, set RAS retire flip bits to the value in NPS1 mode.\n");
+		break;
+	}
+
+	adev->umc.retire_unit = 0x1 << flip_bits->bit_num;
+}
+
+static int umc_v12_0_convert_error_address(struct amdgpu_device *adev,
+					struct ras_err_data *err_data,
+					struct ta_ras_query_address_input *addr_in,
+					struct ta_ras_query_address_output *addr_out,
+					bool dump_addr)
+{
+	uint32_t col, col_lower, row, row_lower, row_high, bank;
+	uint32_t channel_index = 0, umc_inst = 0;
+	uint32_t i, bit_num, retire_unit, *flip_bits;
+	uint64_t soc_pa, column, err_addr;
+	struct ta_ras_query_address_output addr_out_tmp;
+	struct ta_ras_query_address_output *paddr_out;
+	int ret = 0;
+
+	if (!addr_out)
+		paddr_out = &addr_out_tmp;
+	else
+		paddr_out = addr_out;
+
+	err_addr = bank = 0;
+	if (addr_in) {
+		err_addr = addr_in->ma.err_addr;
+		addr_in->addr_type = TA_RAS_MCA_TO_PA;
+		ret = psp_ras_query_address(&adev->psp, addr_in, paddr_out);
+		if (ret) {
+			dev_warn(adev->dev, "Failed to query RAS physical address for 0x%llx",
+				err_addr);
+
+			goto out;
+		}
+
+		bank = paddr_out->pa.bank;
+		/* no need to care about umc inst if addr_in is NULL */
+		umc_inst = addr_in->ma.umc_inst;
+	}
+
+	flip_bits = adev->umc.flip_bits.flip_bits_in_pa;
+	bit_num = adev->umc.flip_bits.bit_num;
+	retire_unit = adev->umc.retire_unit;
+
+	soc_pa = paddr_out->pa.pa;
+	channel_index = paddr_out->pa.channel_idx;
+	/* clear loop bits in soc physical address */
+	for (i = 0; i < bit_num; i++)
+		soc_pa &= ~BIT_ULL(flip_bits[i]);
+
+	paddr_out->pa.pa = soc_pa;
+	/* get column bit 0 and 1 in mca address */
+	col_lower = (err_addr >> 1) & 0x3ULL;
+	/* extra row bit will be handled later */
+	row_lower = (err_addr >> UMC_V12_0_MA_R0_BIT) & 0x1fffULL;
+	row_lower &= ~BIT_ULL(adev->umc.flip_bits.flip_row_bit);
+
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(9, 5, 0)) {
+		row_high = (soc_pa >> adev->umc.flip_bits.r13_in_pa) & 0x3ULL;
+		/* it's 2.25GB in each channel, from MCA address to PA
+		 * [R14 R13] is converted if the two bits value are 0x3,
+		 * get them from PA instead of MCA address.
+		 */
+		row_lower |= (row_high << 13);
+	}
+
+	if (!err_data && !dump_addr)
+		goto out;
+
+	/* loop for all possibilities of retired bits */
+	for (column = 0; column < retire_unit; column++) {
+		soc_pa = paddr_out->pa.pa;
+		for (i = 0; i < bit_num; i++)
+			soc_pa |= (((column >> i) & 0x1ULL) << flip_bits[i]);
+
+		col = ((column & 0x7) << 2) | col_lower;
+		/* handle extra row bit */
+		if (bit_num == RETIRE_FLIP_BITS_NUM)
+			row = ((column >> 3) << adev->umc.flip_bits.flip_row_bit) |
+					row_lower;
+
+		if (dump_addr)
+			dev_info(adev->dev,
+				"Error Address(PA):0x%-10llx Row:0x%-4x Col:0x%-2x Bank:0x%x Channel:0x%x\n",
+				soc_pa, row, col, bank, channel_index);
+
+		if (err_data)
+			amdgpu_umc_fill_error_record(err_data, err_addr,
+				soc_pa, channel_index, umc_inst);
+	}
+
+out:
+	return ret;
+}
+
+static int umc_v12_0_query_error_address(struct amdgpu_device *adev,
+					uint32_t node_inst, uint32_t umc_inst,
+					uint32_t ch_inst, void *data)
+{
+	struct ras_err_data *err_data = (struct ras_err_data *)data;
+	struct ta_ras_query_address_input addr_in;
+	uint64_t mc_umc_status_addr;
+	uint64_t mc_umc_status, err_addr;
+	uint64_t mc_umc_addrt0;
+	uint64_t umc_reg_offset =
+		get_umc_v12_0_reg_offset(adev, node_inst, umc_inst, ch_inst);
+
+	mc_umc_status_addr =
+		SOC15_REG_OFFSET(UMC, 0, regMCA_UMC_UMC0_MCUMC_STATUST0);
+
+	mc_umc_status = RREG64_PCIE_EXT((mc_umc_status_addr + umc_reg_offset) * 4);
+
+	if (mc_umc_status == 0)
+		return 0;
+
+	if (!err_data->err_addr) {
+		/* clear umc status */
+		WREG64_PCIE_EXT((mc_umc_status_addr + umc_reg_offset) * 4, 0x0ULL);
+
+		return 0;
+	}
+
+	/* calculate error address if ue error is detected */
+	if (umc_v12_0_is_uncorrectable_error(adev, mc_umc_status) ||
+	    umc_v12_0_is_deferred_error(adev, mc_umc_status)) {
+		mc_umc_addrt0 =
+			SOC15_REG_OFFSET(UMC, 0, regMCA_UMC_UMC0_MCUMC_ADDRT0);
+
+		err_addr = RREG64_PCIE_EXT((mc_umc_addrt0 + umc_reg_offset) * 4);
+
+		err_addr = REG_GET_FIELD(err_addr, MCA_UMC_UMC0_MCUMC_ADDRT0, ErrorAddr);
+
+		if (!adev->aid_mask &&
+		    adev->smuio.funcs &&
+		    adev->smuio.funcs->get_socket_id)
+			addr_in.ma.socket_id = adev->smuio.funcs->get_socket_id(adev);
+		else
+			addr_in.ma.socket_id = 0;
+
+		addr_in.ma.err_addr = err_addr;
+		addr_in.ma.ch_inst = ch_inst;
+		addr_in.ma.umc_inst = umc_inst;
+		addr_in.ma.node_inst = node_inst;
+
+		umc_v12_0_convert_error_address(adev, err_data, &addr_in, NULL, true);
+	}
+
+	/* clear umc status */
+	WREG64_PCIE_EXT((mc_umc_status_addr + umc_reg_offset) * 4, 0x0ULL);
+
+	return 0;
+}
+
+static void umc_v12_0_query_ras_error_address(struct amdgpu_device *adev,
+					     void *ras_error_status)
+{
+	amdgpu_umc_loop_channels(adev,
+		umc_v12_0_query_error_address, ras_error_status);
+}
+
+static int umc_v12_0_err_cnt_init_per_channel(struct amdgpu_device *adev,
+					uint32_t node_inst, uint32_t umc_inst,
+					uint32_t ch_inst, void *data)
+{
+	uint32_t odecc_cnt_sel;
+	uint64_t odecc_cnt_sel_addr, odecc_err_cnt_addr;
+	uint64_t umc_reg_offset =
+		get_umc_v12_0_reg_offset(adev, node_inst, umc_inst, ch_inst);
+
+	odecc_cnt_sel_addr =
+		SOC15_REG_OFFSET(UMC, 0, regUMCCH0_OdEccCntSel);
+	odecc_err_cnt_addr =
+		SOC15_REG_OFFSET(UMC, 0, regUMCCH0_OdEccErrCnt);
+
+	odecc_cnt_sel = RREG32_PCIE_EXT((odecc_cnt_sel_addr + umc_reg_offset) * 4);
+
+	/* set ce error interrupt type to APIC based interrupt */
+	odecc_cnt_sel = REG_SET_FIELD(odecc_cnt_sel, UMCCH0_OdEccCntSel,
+					OdEccErrInt, 0x1);
+	WREG32_PCIE_EXT((odecc_cnt_sel_addr + umc_reg_offset) * 4, odecc_cnt_sel);
+
+	/* set error count to initial value */
+	WREG32_PCIE_EXT((odecc_err_cnt_addr + umc_reg_offset) * 4, UMC_V12_0_CE_CNT_INIT);
+
+	return 0;
+}
+
+static bool umc_v12_0_check_ecc_err_status(struct amdgpu_device *adev,
+			enum amdgpu_mca_error_type type, void *ras_error_status)
+{
+	uint64_t mc_umc_status = *(uint64_t *)ras_error_status;
+
+	switch (type) {
+	case AMDGPU_MCA_ERROR_TYPE_UE:
+		return umc_v12_0_is_uncorrectable_error(adev, mc_umc_status);
+	case AMDGPU_MCA_ERROR_TYPE_CE:
+		return umc_v12_0_is_correctable_error(adev, mc_umc_status);
+	case AMDGPU_MCA_ERROR_TYPE_DE:
+		return umc_v12_0_is_deferred_error(adev, mc_umc_status);
+	default:
+		return false;
+	}
+
+	return false;
+}
+
+static void umc_v12_0_err_cnt_init(struct amdgpu_device *adev)
+{
+	amdgpu_umc_loop_channels(adev,
+		umc_v12_0_err_cnt_init_per_channel, NULL);
+}
+
+static bool umc_v12_0_query_ras_poison_mode(struct amdgpu_device *adev)
+{
+	/*
+	 * Force return true, because regUMCCH0_EccCtrl
+	 * is not accessible from host side
+	 */
+	return true;
+}
+
+const struct amdgpu_ras_block_hw_ops umc_v12_0_ras_hw_ops = {
+	.query_ras_error_count = umc_v12_0_query_ras_error_count,
+	.query_ras_error_address = umc_v12_0_query_ras_error_address,
+};
+
+static int umc_v12_0_aca_bank_parser(struct aca_handle *handle, struct aca_bank *bank,
+				     enum aca_smu_type type, void *data)
+{
+	struct amdgpu_device *adev = handle->adev;
+	struct aca_bank_info info;
+	enum aca_error_type err_type;
+	u64 status, count;
+	u32 ext_error_code;
+	int ret;
+
+	status = bank->regs[ACA_REG_IDX_STATUS];
+	if (umc_v12_0_is_deferred_error(adev, status))
+		err_type = ACA_ERROR_TYPE_DEFERRED;
+	else if (umc_v12_0_is_uncorrectable_error(adev, status))
+		err_type = ACA_ERROR_TYPE_UE;
+	else if (umc_v12_0_is_correctable_error(adev, status))
+		err_type = ACA_ERROR_TYPE_CE;
+	else
+		return 0;
+	bank->aca_err_type = err_type;
+
+	ret = aca_bank_info_decode(bank, &info);
+	if (ret)
+		return ret;
+
+	amdgpu_umc_update_ecc_status(adev,
+		bank->regs[ACA_REG_IDX_STATUS],
+		bank->regs[ACA_REG_IDX_IPID],
+		bank->regs[ACA_REG_IDX_ADDR]);
+
+	ext_error_code = ACA_REG__STATUS__ERRORCODEEXT(status);
+	if (umc_v12_0_is_deferred_error(adev, status))
+		count = ext_error_code == 0 ?
+			adev->umc.err_addr_cnt / adev->umc.retire_unit : 1ULL;
+	else
+		count = ext_error_code == 0 ?
+			ACA_REG__MISC0__ERRCNT(bank->regs[ACA_REG_IDX_MISC0]) : 1ULL;
+
+	return aca_error_cache_log_bank_error(handle, &info, err_type, count);
+}
+
+static const struct aca_bank_ops umc_v12_0_aca_bank_ops = {
+	.aca_bank_parser = umc_v12_0_aca_bank_parser,
+};
+
+const struct aca_info umc_v12_0_aca_info = {
+	.hwip = ACA_HWIP_TYPE_UMC,
+	.mask = ACA_ERROR_UE_MASK | ACA_ERROR_CE_MASK | ACA_ERROR_DEFERRED_MASK,
+	.bank_ops = &umc_v12_0_aca_bank_ops,
+};
+
+static int umc_v12_0_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
+{
+	int ret;
+
+	ret = amdgpu_umc_ras_late_init(adev, ras_block);
+	if (ret)
+		return ret;
+
+	ret = amdgpu_ras_bind_aca(adev, AMDGPU_RAS_BLOCK__UMC,
+				  &umc_v12_0_aca_info, NULL);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static int umc_v12_0_update_ecc_status(struct amdgpu_device *adev,
+			uint64_t status, uint64_t ipid, uint64_t addr)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	uint16_t hwid, mcatype;
+	uint64_t page_pfn[UMC_V12_0_BAD_PAGE_NUM_PER_CHANNEL];
+	uint64_t err_addr, pa_addr = 0;
+	struct ras_ecc_err *ecc_err;
+	struct ta_ras_query_address_output addr_out;
+	uint32_t shift_bit = adev->umc.flip_bits.flip_bits_in_pa[2];
+	int count, ret, i;
+
+	hwid = REG_GET_FIELD(ipid, MCMP1_IPIDT0, HardwareID);
+	mcatype = REG_GET_FIELD(ipid, MCMP1_IPIDT0, McaType);
+
+	/* The IP block decode of consumption is SMU */
+	if (hwid != MCA_UMC_HWID_V12_0 || mcatype != MCA_UMC_MCATYPE_V12_0) {
+		con->umc_ecc_log.consumption_q_count++;
+		return 0;
+	}
+
+	if (!status)
+		return 0;
+
+	if (!umc_v12_0_is_deferred_error(adev, status))
+		return 0;
+
+	err_addr = REG_GET_FIELD(addr,
+				MCA_UMC_UMC0_MCUMC_ADDRT0, ErrorAddr);
+
+	dev_dbg(adev->dev,
+		"UMC:IPID:0x%llx, socket:%llu, aid:%llu, inst:%llu, ch:%llu, err_addr:0x%llx\n",
+		ipid,
+		MCA_IPID_2_SOCKET_ID(ipid),
+		MCA_IPID_2_DIE_ID(ipid),
+		MCA_IPID_2_UMC_INST(ipid),
+		MCA_IPID_2_UMC_CH(ipid),
+		err_addr);
+
+	ret = amdgpu_umc_mca_to_addr(adev,
+			err_addr, MCA_IPID_2_UMC_CH(ipid),
+			MCA_IPID_2_UMC_INST(ipid), MCA_IPID_2_DIE_ID(ipid),
+			MCA_IPID_2_SOCKET_ID(ipid), &addr_out, true);
+	if (ret)
+		return ret;
+
+	ecc_err = kzalloc(sizeof(*ecc_err), GFP_KERNEL);
+	if (!ecc_err)
+		return -ENOMEM;
+
+	pa_addr = addr_out.pa.pa;
+	ecc_err->status = status;
+	ecc_err->ipid = ipid;
+	ecc_err->addr = addr;
+	ecc_err->pa_pfn = pa_addr >> AMDGPU_GPU_PAGE_SHIFT;
+	ecc_err->channel_idx = addr_out.pa.channel_idx;
+
+	/* If converted pa_pfn is 0, use pa C4 pfn. */
+	if (!ecc_err->pa_pfn)
+		ecc_err->pa_pfn = BIT_ULL(shift_bit) >> AMDGPU_GPU_PAGE_SHIFT;
+
+	ret = amdgpu_umc_logs_ecc_err(adev, &con->umc_ecc_log.de_page_tree, ecc_err);
+	if (ret) {
+		if (ret == -EEXIST)
+			con->umc_ecc_log.de_queried_count++;
+		else
+			dev_err(adev->dev, "Fail to log ecc error! ret:%d\n", ret);
+
+		kfree(ecc_err);
+		return ret;
+	}
+
+	con->umc_ecc_log.de_queried_count++;
+
+	memset(page_pfn, 0, sizeof(page_pfn));
+	count = amdgpu_umc_lookup_bad_pages_in_a_row(adev,
+				pa_addr,
+				page_pfn, ARRAY_SIZE(page_pfn));
+	if (count <= 0) {
+		dev_warn(adev->dev, "Fail to convert error address! count:%d\n", count);
+		return 0;
+	}
+
+	/* Reserve memory */
+	for (i = 0; i < count; i++)
+		amdgpu_ras_reserve_page(adev, page_pfn[i]);
+
+	/* The problem case is as follows:
+	 * 1. GPU A triggers a gpu ras reset, and GPU A drives
+	 *    GPU B to also perform a gpu ras reset.
+	 * 2. After gpu B ras reset started, gpu B queried a DE
+	 *    data. Since the DE data was queried in the ras reset
+	 *    thread instead of the page retirement thread, bad
+	 *    page retirement work would not be triggered. Then
+	 *    even if all gpu resets are completed, the bad pages
+	 *    will be cached in RAM until GPU B's bad page retirement
+	 *    work is triggered again and then saved to eeprom.
+	 * Trigger delayed work to save the bad pages to eeprom in time
+	 * after gpu ras reset is completed.
+	 */
+	if (amdgpu_ras_in_recovery(adev))
+		schedule_delayed_work(&con->page_retirement_dwork,
+			msecs_to_jiffies(DELAYED_TIME_FOR_GPU_RESET));
+
+	return 0;
+}
+
+static int umc_v12_0_fill_error_record(struct amdgpu_device *adev,
+				struct ras_ecc_err *ecc_err, void *ras_error_status)
+{
+	struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
+	uint64_t page_pfn[UMC_V12_0_BAD_PAGE_NUM_PER_CHANNEL];
+	int ret, i, count;
+
+	if (!err_data || !ecc_err)
+		return -EINVAL;
+
+	memset(page_pfn, 0, sizeof(page_pfn));
+	count = amdgpu_umc_lookup_bad_pages_in_a_row(adev,
+				ecc_err->pa_pfn << AMDGPU_GPU_PAGE_SHIFT,
+				page_pfn, ARRAY_SIZE(page_pfn));
+
+	for (i = 0; i < count; i++) {
+		ret = amdgpu_umc_fill_error_record(err_data,
+				ecc_err->addr,
+				page_pfn[i] << AMDGPU_GPU_PAGE_SHIFT,
+				ecc_err->channel_idx,
+				MCA_IPID_2_UMC_INST(ecc_err->ipid));
+		if (ret)
+			break;
+	}
+
+	err_data->de_count++;
+
+	return ret;
+}
+
+static void umc_v12_0_query_ras_ecc_err_addr(struct amdgpu_device *adev,
+					void *ras_error_status)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct ras_ecc_err *entries[MAX_ECC_NUM_PER_RETIREMENT];
+	struct radix_tree_root *ecc_tree;
+	int new_detected, ret, i;
+
+	ecc_tree = &con->umc_ecc_log.de_page_tree;
+
+	mutex_lock(&con->umc_ecc_log.lock);
+	new_detected = radix_tree_gang_lookup_tag(ecc_tree, (void **)entries,
+			0, ARRAY_SIZE(entries), UMC_ECC_NEW_DETECTED_TAG);
+	for (i = 0; i < new_detected; i++) {
+		if (!entries[i])
+			continue;
+
+		ret = umc_v12_0_fill_error_record(adev, entries[i], ras_error_status);
+		if (ret) {
+			dev_err(adev->dev, "Fail to fill umc error record, ret:%d\n", ret);
+			break;
+		}
+		radix_tree_tag_clear(ecc_tree,
+				entries[i]->pa_pfn, UMC_ECC_NEW_DETECTED_TAG);
+	}
+	mutex_unlock(&con->umc_ecc_log.lock);
+}
+
+static uint32_t umc_v12_0_get_die_id(struct amdgpu_device *adev,
+		uint64_t mca_addr, uint64_t retired_page)
+{
+	uint32_t die = 0;
+
+	/* we only calculate die id for nps1 mode right now */
+	die += ((((retired_page >> 12) & 0x1ULL)^
+	    ((retired_page >> 20) & 0x1ULL) ^
+	    ((retired_page >> 27) & 0x1ULL) ^
+	    ((retired_page >> 34) & 0x1ULL) ^
+	    ((retired_page >> 41) & 0x1ULL)) << 0);
+
+	/* the original PA_C4 and PA_R13 may be cleared in retired_page, so
+	 * get them from mca_addr.
+	 */
+	die += ((((retired_page >> 13) & 0x1ULL) ^
+	    ((mca_addr >> 5) & 0x1ULL) ^
+	    ((retired_page >> 28) & 0x1ULL) ^
+	    ((mca_addr >> 23) & 0x1ULL) ^
+	    ((retired_page >> 42) & 0x1ULL)) << 1);
+	die &= 3;
+
+	return die;
+}
+
+static void umc_v12_0_mca_ipid_parse(struct amdgpu_device *adev, uint64_t ipid,
+		uint32_t *did, uint32_t *ch, uint32_t *umc_inst, uint32_t *sid)
+{
+	if (did)
+		*did = MCA_IPID_2_DIE_ID(ipid);
+	if (ch)
+		*ch = MCA_IPID_2_UMC_CH(ipid);
+	if (umc_inst)
+		*umc_inst = MCA_IPID_2_UMC_INST(ipid);
+	if (sid)
+		*sid = MCA_IPID_2_SOCKET_ID(ipid);
+}
+
+struct amdgpu_umc_ras umc_v12_0_ras = {
+	.ras_block = {
+		.hw_ops = &umc_v12_0_ras_hw_ops,
+		.ras_late_init = umc_v12_0_ras_late_init,
+	},
+	.err_cnt_init = umc_v12_0_err_cnt_init,
+	.query_ras_poison_mode = umc_v12_0_query_ras_poison_mode,
+	.ecc_info_query_ras_error_address = umc_v12_0_query_ras_ecc_err_addr,
+	.check_ecc_err_status = umc_v12_0_check_ecc_err_status,
+	.update_ecc_status = umc_v12_0_update_ecc_status,
+	.convert_ras_err_addr = umc_v12_0_convert_error_address,
+	.get_die_id_from_pa = umc_v12_0_get_die_id,
+	.get_retire_flip_bits = umc_v12_0_get_retire_flip_bits,
+	.mca_ipid_parse = umc_v12_0_mca_ipid_parse,
+};
+
diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v12_0.h b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.h
new file mode 100644
index 000000000000..63b7e7254526
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.h
@@ -0,0 +1,105 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __UMC_V12_0_H__
+#define __UMC_V12_0_H__
+
+#include "soc15_common.h"
+#include "amdgpu.h"
+
+#define UMC_V12_0_NODE_DIST		0x40000000
+#define UMC_V12_0_INST_DIST		0x40000
+
+/* UMC register per channel offset */
+#define UMC_V12_0_PER_CHANNEL_OFFSET	0x400
+
+/* UMC cross node offset */
+#define UMC_V12_0_CROSS_NODE_OFFSET		0x100000000
+
+/* OdEccErrCnt max value */
+#define UMC_V12_0_CE_CNT_MAX		0xffff
+/* umc ce interrupt threshold */
+#define UMC_V12_0_CE_INT_THRESHOLD	0xffff
+/* umc ce count initial value */
+#define UMC_V12_0_CE_CNT_INIT	(UMC_V12_0_CE_CNT_MAX - UMC_V12_0_CE_INT_THRESHOLD)
+
+/* number of umc channel instance with memory map register access */
+#define UMC_V12_0_CHANNEL_INSTANCE_NUM		8
+/* number of umc instance with memory map register access */
+#define UMC_V12_0_UMC_INSTANCE_NUM		4
+
+/* Total channel instances for all available umc nodes */
+#define UMC_V12_0_TOTAL_CHANNEL_NUM(adev) \
+	(UMC_V12_0_CHANNEL_INSTANCE_NUM * (adev)->gmc.num_umc)
+
+/* one piece of normalized address is mapped to 8 pieces of physical address */
+#define UMC_V12_0_NA_MAP_PA_NUM        8
+/* R13 bit shift should be considered, double the number */
+#define UMC_V12_0_BAD_PAGE_NUM_PER_CHANNEL (UMC_V12_0_NA_MAP_PA_NUM * 2)
+
+/* column bits in SOC physical address */
+#define UMC_V12_0_PA_C2_BIT 15
+#define UMC_V12_0_PA_C3_BIT 16
+#define UMC_V12_0_PA_C4_BIT 21
+/* row bits in SOC physical address */
+#define UMC_V12_0_PA_R0_BIT 22
+#define UMC_V12_0_PA_R10_BIT 32
+#define UMC_V12_0_PA_R11_BIT 33
+#define UMC_V12_0_PA_R12_BIT 34
+#define UMC_V12_0_PA_R13_BIT 35
+/* channel bit in SOC physical address */
+#define UMC_V12_0_PA_CH4_BIT 12
+#define UMC_V12_0_PA_CH5_BIT 13
+/* bank bit in SOC physical address */
+#define UMC_V12_0_PA_B0_BIT 19
+#define UMC_V12_0_PA_B1_BIT 20
+/* row bits in MCA address */
+#define UMC_V12_0_MA_R0_BIT 10
+
+#define MCA_UMC_HWID_V12_0     0x96
+#define MCA_UMC_MCATYPE_V12_0  0x0
+
+#define MCA_IPID_LO_2_UMC_CH(_ipid_lo) (((((_ipid_lo) >> 20) & 0x1) * 4) + \
+			(((_ipid_lo) >> 12) & 0xF))
+#define MCA_IPID_LO_2_UMC_INST(_ipid_lo) (((_ipid_lo) >> 21) & 0x7)
+
+#define MCA_IPID_2_DIE_ID(ipid)  ((REG_GET_FIELD(ipid, MCMP1_IPIDT0, InstanceIdHi) >> 2) & 0x03)
+
+#define MCA_IPID_2_UMC_CH(ipid) \
+	(MCA_IPID_LO_2_UMC_CH(REG_GET_FIELD(ipid, MCMP1_IPIDT0, InstanceIdLo)))
+
+#define MCA_IPID_2_UMC_INST(ipid) \
+	(MCA_IPID_LO_2_UMC_INST(REG_GET_FIELD(ipid, MCMP1_IPIDT0, InstanceIdLo)))
+
+#define MCA_IPID_2_SOCKET_ID(ipid) \
+	(((REG_GET_FIELD(ipid, MCMP1_IPIDT0, InstanceIdLo) & 0x1) << 2) | \
+	 (REG_GET_FIELD(ipid, MCMP1_IPIDT0, InstanceIdHi) & 0x03))
+
+bool umc_v12_0_is_deferred_error(struct amdgpu_device *adev, uint64_t mc_umc_status);
+bool umc_v12_0_is_uncorrectable_error(struct amdgpu_device *adev, uint64_t mc_umc_status);
+bool umc_v12_0_is_correctable_error(struct amdgpu_device *adev, uint64_t mc_umc_status);
+
+typedef bool (*check_error_type_func)(struct amdgpu_device *adev, uint64_t mc_umc_status);
+
+extern struct amdgpu_umc_ras umc_v12_0_ras;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_0.c b/drivers/gpu/drm/amd/amdgpu/umc_v6_0.c
index 0d6b50528d76..97fa88ed770c 100644
--- a/drivers/gpu/drm/amd/amdgpu/umc_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_0.c
@@ -25,7 +25,7 @@
 
 static void umc_v6_0_init_registers(struct amdgpu_device *adev)
 {
-	unsigned i,j;
+	unsigned i, j;
 
 	for (i = 0; i < 4; i++)
 		for (j = 0; j < 4; j++)
diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c
index 530549314ce4..a3ee3c4c650f 100644
--- a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c
+++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c
@@ -64,7 +64,7 @@ static void umc_v6_7_query_error_status_helper(struct amdgpu_device *adev,
 	uint64_t reg_value;
 
 	if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Deferred) == 1)
-		dev_info(adev->dev, "Deferred error, no user action is needed.\n");
+		dev_info(adev->dev, "Deferred error\n");
 
 	if (mc_umc_status)
 		dev_info(adev->dev, "MCA STATUS 0x%llx, umc_reg_offset 0x%x\n", mc_umc_status, umc_reg_offset);
diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v8_10.c b/drivers/gpu/drm/amd/amdgpu/umc_v8_10.c
index d51ae0bc36f7..a32f87992f20 100644
--- a/drivers/gpu/drm/amd/amdgpu/umc_v8_10.c
+++ b/drivers/gpu/drm/amd/amdgpu/umc_v8_10.c
@@ -336,7 +336,7 @@ static void umc_v8_10_ecc_info_query_correctable_error_count(struct amdgpu_devic
 				      uint32_t node_inst, uint32_t umc_inst, uint32_t ch_inst,
 				      unsigned long *error_count)
 {
-	uint64_t mc_umc_status;
+	uint16_t ecc_ce_cnt;
 	uint32_t eccinfo_table_idx;
 	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 
@@ -345,12 +345,10 @@ static void umc_v8_10_ecc_info_query_correctable_error_count(struct amdgpu_devic
 				  umc_inst * adev->umc.channel_inst_num +
 				  ch_inst;
 
-	/* check the MCUMC_STATUS */
-	mc_umc_status = ras->umc_ecc.ecc[eccinfo_table_idx].mca_umc_status;
-	if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 &&
-	    REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, CECC) == 1) {
-		*error_count += 1;
-	}
+	/* Retrieve CE count */
+	ecc_ce_cnt = ras->umc_ecc.ecc[eccinfo_table_idx].ce_count_lo_chip;
+	if (ecc_ce_cnt)
+		*error_count += ecc_ce_cnt;
 }
 
 static void umc_v8_10_ecc_info_query_uncorrectable_error_count(struct amdgpu_device *adev,
diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v8_10.h b/drivers/gpu/drm/amd/amdgpu/umc_v8_10.h
index c6dfd433fec7..dc12e0af5451 100644
--- a/drivers/gpu/drm/amd/amdgpu/umc_v8_10.h
+++ b/drivers/gpu/drm/amd/amdgpu/umc_v8_10.h
@@ -33,7 +33,8 @@
 
 /* Total channel instances for all available umc nodes */
 #define UMC_V8_10_TOTAL_CHANNEL_NUM(adev) \
-	(UMC_V8_10_CHANNEL_INSTANCE_NUM * UMC_V8_10_UMC_INSTANCE_NUM * (adev)->gmc.num_umc)
+	(UMC_V8_10_CHANNEL_INSTANCE_NUM * UMC_V8_10_UMC_INSTANCE_NUM * \
+	(adev)->gmc.num_umc - hweight32((adev)->gmc.m_half_use) * 2)
 
 /* UMC regiser per channel offset */
 #define UMC_V8_10_PER_CHANNEL_OFFSET	0x400
diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v8_14.c b/drivers/gpu/drm/amd/amdgpu/umc_v8_14.c
new file mode 100644
index 000000000000..eaca10a3c4a9
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/umc_v8_14.c
@@ -0,0 +1,160 @@
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "umc_v8_14.h"
+#include "amdgpu_ras.h"
+#include "amdgpu_umc.h"
+#include "amdgpu.h"
+#include "umc/umc_8_14_0_offset.h"
+#include "umc/umc_8_14_0_sh_mask.h"
+
+static inline uint32_t get_umc_v8_14_reg_offset(struct amdgpu_device *adev,
+					    uint32_t umc_inst,
+					    uint32_t ch_inst)
+{
+	return adev->umc.channel_offs * ch_inst + UMC_V8_14_INST_DIST * umc_inst;
+}
+
+static int umc_v8_14_clear_error_count_per_channel(struct amdgpu_device *adev,
+					uint32_t node_inst, uint32_t umc_inst,
+					uint32_t ch_inst, void *data)
+{
+	uint32_t ecc_err_cnt_addr;
+	uint32_t umc_reg_offset =
+		get_umc_v8_14_reg_offset(adev, umc_inst, ch_inst);
+
+	ecc_err_cnt_addr =
+		SOC15_REG_OFFSET(UMC, 0, regUMCCH0_GeccErrCnt);
+
+	/* clear error count */
+	WREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4,
+			UMC_V8_14_CE_CNT_INIT);
+
+	return 0;
+}
+
+static void umc_v8_14_clear_error_count(struct amdgpu_device *adev)
+{
+	amdgpu_umc_loop_channels(adev,
+		umc_v8_14_clear_error_count_per_channel, NULL);
+}
+
+static void umc_v8_14_query_correctable_error_count(struct amdgpu_device *adev,
+						   uint32_t umc_reg_offset,
+						   unsigned long *error_count)
+{
+	uint32_t ecc_err_cnt, ecc_err_cnt_addr;
+
+	/* UMC 8_14 registers */
+	ecc_err_cnt_addr =
+		SOC15_REG_OFFSET(UMC, 0, regUMCCH0_GeccErrCnt);
+
+	ecc_err_cnt = RREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4);
+	*error_count +=
+		(REG_GET_FIELD(ecc_err_cnt, UMCCH0_GeccErrCnt, GeccErrCnt) -
+		 UMC_V8_14_CE_CNT_INIT);
+}
+
+static void umc_v8_14_query_uncorrectable_error_count(struct amdgpu_device *adev,
+						      uint32_t umc_reg_offset,
+						      unsigned long *error_count)
+{
+	uint32_t ecc_err_cnt, ecc_err_cnt_addr;
+	/* UMC 8_14 registers */
+	ecc_err_cnt_addr =
+		SOC15_REG_OFFSET(UMC, 0, regUMCCH0_GeccErrCnt);
+
+	ecc_err_cnt = RREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4);
+	*error_count +=
+		(REG_GET_FIELD(ecc_err_cnt, UMCCH0_GeccErrCnt, GeccUnCorrErrCnt) -
+		 UMC_V8_14_CE_CNT_INIT);
+}
+
+static int umc_v8_14_query_error_count_per_channel(struct amdgpu_device *adev,
+					uint32_t node_inst, uint32_t umc_inst,
+					uint32_t ch_inst, void *data)
+{
+	struct ras_err_data *err_data = (struct ras_err_data *)data;
+	uint32_t umc_reg_offset =
+		get_umc_v8_14_reg_offset(adev, umc_inst, ch_inst);
+
+	umc_v8_14_query_correctable_error_count(adev,
+					umc_reg_offset,
+					&(err_data->ce_count));
+	umc_v8_14_query_uncorrectable_error_count(adev,
+					umc_reg_offset,
+					&(err_data->ue_count));
+
+	return 0;
+}
+
+static void umc_v8_14_query_ras_error_count(struct amdgpu_device *adev,
+					   void *ras_error_status)
+{
+	amdgpu_umc_loop_channels(adev,
+		umc_v8_14_query_error_count_per_channel, ras_error_status);
+
+	umc_v8_14_clear_error_count(adev);
+}
+
+static int umc_v8_14_err_cnt_init_per_channel(struct amdgpu_device *adev,
+					uint32_t node_inst, uint32_t umc_inst,
+					uint32_t ch_inst, void *data)
+{
+	uint32_t ecc_err_cnt_sel, ecc_err_cnt_sel_addr;
+	uint32_t ecc_err_cnt_addr;
+	uint32_t umc_reg_offset =
+		get_umc_v8_14_reg_offset(adev, umc_inst, ch_inst);
+
+	ecc_err_cnt_sel_addr =
+		SOC15_REG_OFFSET(UMC, 0, regUMCCH0_GeccErrCntSel);
+	ecc_err_cnt_addr =
+		SOC15_REG_OFFSET(UMC, 0, regUMCCH0_GeccErrCnt);
+
+	ecc_err_cnt_sel = RREG32_PCIE((ecc_err_cnt_sel_addr + umc_reg_offset) * 4);
+
+	/* set ce error interrupt type to APIC based interrupt */
+	ecc_err_cnt_sel = REG_SET_FIELD(ecc_err_cnt_sel, UMCCH0_GeccErrCntSel,
+					GeccErrInt, 0x1);
+	WREG32_PCIE((ecc_err_cnt_sel_addr + umc_reg_offset) * 4, ecc_err_cnt_sel);
+	/* set error count to initial value */
+	WREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4, UMC_V8_14_CE_CNT_INIT);
+
+	return 0;
+}
+
+static void umc_v8_14_err_cnt_init(struct amdgpu_device *adev)
+{
+	amdgpu_umc_loop_channels(adev,
+		umc_v8_14_err_cnt_init_per_channel, NULL);
+}
+
+const struct amdgpu_ras_block_hw_ops umc_v8_14_ras_hw_ops = {
+	.query_ras_error_count = umc_v8_14_query_ras_error_count,
+};
+
+struct amdgpu_umc_ras umc_v8_14_ras = {
+	.ras_block = {
+		.hw_ops = &umc_v8_14_ras_hw_ops,
+	},
+	.err_cnt_init = umc_v8_14_err_cnt_init,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v8_14.h b/drivers/gpu/drm/amd/amdgpu/umc_v8_14.h
new file mode 100644
index 000000000000..20a258f0017a
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/umc_v8_14.h
@@ -0,0 +1,51 @@
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __UMC_V8_14_H__
+#define __UMC_V8_14_H__
+
+#include "soc15_common.h"
+#include "amdgpu.h"
+
+/* number of umc channel instance with memory map register access */
+#define UMC_V8_14_CHANNEL_INSTANCE_NUM		2
+/* number of umc instance with memory map register access */
+#define UMC_V8_14_UMC_INSTANCE_NUM(adev)		((adev)->umc.node_inst_num)
+
+/* Total channel instances for all available umc nodes */
+#define UMC_V8_14_TOTAL_CHANNEL_NUM(adev) \
+	(UMC_V8_14_CHANNEL_INSTANCE_NUM * (adev)->gmc.num_umc)
+
+/* UMC register per channel offset */
+#define UMC_V8_14_PER_CHANNEL_OFFSET		0x400
+
+#define UMC_V8_14_INST_DIST		0x40000
+
+/* EccErrCnt max value */
+#define UMC_V8_14_CE_CNT_MAX		0xffff
+/* umc ce interrupt threshold */
+#define UMC_V8_14_CE_INT_THRESHOLD		0xffff
+/* umc ce count initial value */
+#define UMC_V8_14_CE_CNT_INIT	(UMC_V8_14_CE_CNT_MAX - UMC_V8_14_CE_INT_THRESHOLD)
+
+extern struct amdgpu_umc_ras umc_v8_14_ras;
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c b/drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c
new file mode 100644
index 000000000000..ce3bb12e3572
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c
@@ -0,0 +1,434 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/firmware.h>
+#include <linux/module.h>
+#include <linux/debugfs.h>
+#include "amdgpu.h"
+#include "soc15_common.h"
+#include "soc21.h"
+#include "vcn/vcn_4_0_0_offset.h"
+#include "vcn/vcn_4_0_0_sh_mask.h"
+
+#include "amdgpu_umsch_mm.h"
+#include "umsch_mm_4_0_api_def.h"
+#include "umsch_mm_v4_0.h"
+
+#define regUVD_IPX_DLDO_CONFIG                             0x0064
+#define regUVD_IPX_DLDO_CONFIG_BASE_IDX                    1
+#define regUVD_IPX_DLDO_STATUS                             0x0065
+#define regUVD_IPX_DLDO_STATUS_BASE_IDX                    1
+
+#define UVD_IPX_DLDO_CONFIG__ONO0_PWR_CONFIG__SHIFT        0x00000002
+#define UVD_IPX_DLDO_CONFIG__ONO0_PWR_CONFIG_MASK          0x0000000cUL
+#define UVD_IPX_DLDO_STATUS__ONO0_PWR_STATUS__SHIFT        0x00000001
+#define UVD_IPX_DLDO_STATUS__ONO0_PWR_STATUS_MASK          0x00000002UL
+
+static int umsch_mm_v4_0_load_microcode(struct amdgpu_umsch_mm *umsch)
+{
+	struct amdgpu_device *adev = umsch->ring.adev;
+	uint64_t data;
+	int r;
+
+	r = amdgpu_umsch_mm_allocate_ucode_buffer(umsch);
+	if (r)
+		return r;
+
+	r = amdgpu_umsch_mm_allocate_ucode_data_buffer(umsch);
+	if (r)
+		goto err_free_ucode_bo;
+
+	umsch->cmd_buf_curr_ptr = umsch->cmd_buf_ptr;
+
+	if (amdgpu_ip_version(adev, VCN_HWIP, 0) >= IP_VERSION(4, 0, 5)) {
+		WREG32_SOC15(VCN, 0, regUVD_IPX_DLDO_CONFIG,
+			1 << UVD_IPX_DLDO_CONFIG__ONO0_PWR_CONFIG__SHIFT);
+		SOC15_WAIT_ON_RREG(VCN, 0, regUVD_IPX_DLDO_STATUS,
+			0 << UVD_IPX_DLDO_STATUS__ONO0_PWR_STATUS__SHIFT,
+			UVD_IPX_DLDO_STATUS__ONO0_PWR_STATUS_MASK);
+	}
+
+	data = RREG32_SOC15(VCN, 0, regUMSCH_MES_RESET_CTRL);
+	data = REG_SET_FIELD(data, UMSCH_MES_RESET_CTRL, MES_CORE_SOFT_RESET, 0);
+	WREG32_SOC15_UMSCH(regUMSCH_MES_RESET_CTRL, data);
+
+	data = RREG32_SOC15(VCN, 0, regVCN_MES_CNTL);
+	data = REG_SET_FIELD(data, VCN_MES_CNTL, MES_INVALIDATE_ICACHE, 1);
+	data = REG_SET_FIELD(data, VCN_MES_CNTL, MES_PIPE0_RESET, 1);
+	data = REG_SET_FIELD(data, VCN_MES_CNTL, MES_PIPE0_ACTIVE, 0);
+	data = REG_SET_FIELD(data, VCN_MES_CNTL, MES_HALT, 1);
+	WREG32_SOC15_UMSCH(regVCN_MES_CNTL, data);
+
+	data = RREG32_SOC15(VCN, 0, regVCN_MES_IC_BASE_CNTL);
+	data = REG_SET_FIELD(data, VCN_MES_IC_BASE_CNTL, VMID, 0);
+	data = REG_SET_FIELD(data, VCN_MES_IC_BASE_CNTL, EXE_DISABLE, 0);
+	data = REG_SET_FIELD(data, VCN_MES_IC_BASE_CNTL, CACHE_POLICY, 0);
+	WREG32_SOC15_UMSCH(regVCN_MES_IC_BASE_CNTL, data);
+
+	WREG32_SOC15_UMSCH(regVCN_MES_INTR_ROUTINE_START,
+		lower_32_bits(adev->umsch_mm.irq_start_addr >> 2));
+	WREG32_SOC15_UMSCH(regVCN_MES_INTR_ROUTINE_START_HI,
+		upper_32_bits(adev->umsch_mm.irq_start_addr >> 2));
+
+	WREG32_SOC15_UMSCH(regVCN_MES_PRGRM_CNTR_START,
+		lower_32_bits(adev->umsch_mm.uc_start_addr >> 2));
+	WREG32_SOC15_UMSCH(regVCN_MES_PRGRM_CNTR_START_HI,
+		upper_32_bits(adev->umsch_mm.uc_start_addr >> 2));
+
+	WREG32_SOC15_UMSCH(regVCN_MES_LOCAL_INSTR_BASE_LO, 0);
+	WREG32_SOC15_UMSCH(regVCN_MES_LOCAL_INSTR_BASE_HI, 0);
+
+	data = adev->umsch_mm.uc_start_addr + adev->umsch_mm.ucode_size - 1;
+	WREG32_SOC15_UMSCH(regVCN_MES_LOCAL_INSTR_MASK_LO, lower_32_bits(data));
+	WREG32_SOC15_UMSCH(regVCN_MES_LOCAL_INSTR_MASK_HI, upper_32_bits(data));
+
+	data = adev->firmware.load_type == AMDGPU_FW_LOAD_PSP ?
+	       0 : adev->umsch_mm.ucode_fw_gpu_addr;
+	WREG32_SOC15_UMSCH(regVCN_MES_IC_BASE_LO, lower_32_bits(data));
+	WREG32_SOC15_UMSCH(regVCN_MES_IC_BASE_HI, upper_32_bits(data));
+
+	WREG32_SOC15_UMSCH(regVCN_MES_MIBOUND_LO, 0x1FFFFF);
+
+	WREG32_SOC15_UMSCH(regVCN_MES_LOCAL_BASE0_LO,
+		lower_32_bits(adev->umsch_mm.data_start_addr));
+	WREG32_SOC15_UMSCH(regVCN_MES_LOCAL_BASE0_HI,
+		upper_32_bits(adev->umsch_mm.data_start_addr));
+
+	WREG32_SOC15_UMSCH(regVCN_MES_LOCAL_MASK0_LO,
+		adev->umsch_mm.data_size - 1);
+	WREG32_SOC15_UMSCH(regVCN_MES_LOCAL_MASK0_HI, 0);
+
+	data = adev->firmware.load_type == AMDGPU_FW_LOAD_PSP ?
+	       0 : adev->umsch_mm.data_fw_gpu_addr;
+	WREG32_SOC15_UMSCH(regVCN_MES_DC_BASE_LO, lower_32_bits(data));
+	WREG32_SOC15_UMSCH(regVCN_MES_DC_BASE_HI, upper_32_bits(data));
+
+	WREG32_SOC15_UMSCH(regVCN_MES_MDBOUND_LO, 0x3FFFF);
+
+	data = RREG32_SOC15(VCN, 0, regUVD_UMSCH_FORCE);
+	data = REG_SET_FIELD(data, UVD_UMSCH_FORCE, IC_FORCE_GPUVM, 1);
+	data = REG_SET_FIELD(data, UVD_UMSCH_FORCE, DC_FORCE_GPUVM, 1);
+	WREG32_SOC15_UMSCH(regUVD_UMSCH_FORCE, data);
+
+	data = RREG32_SOC15(VCN, 0, regVCN_MES_IC_OP_CNTL);
+	data = REG_SET_FIELD(data, VCN_MES_IC_OP_CNTL, PRIME_ICACHE, 0);
+	data = REG_SET_FIELD(data, VCN_MES_IC_OP_CNTL, INVALIDATE_CACHE, 1);
+	WREG32_SOC15_UMSCH(regVCN_MES_IC_OP_CNTL, data);
+
+	data = RREG32_SOC15(VCN, 0, regVCN_MES_IC_OP_CNTL);
+	data = REG_SET_FIELD(data, VCN_MES_IC_OP_CNTL, PRIME_ICACHE, 1);
+	WREG32_SOC15_UMSCH(regVCN_MES_IC_OP_CNTL, data);
+
+	WREG32_SOC15_UMSCH(regVCN_MES_GP0_LO, 0);
+	WREG32_SOC15_UMSCH(regVCN_MES_GP0_HI, 0);
+
+#if defined(CONFIG_DEBUG_FS)
+	WREG32_SOC15_UMSCH(regVCN_MES_GP0_LO, lower_32_bits(umsch->log_gpu_addr));
+	WREG32_SOC15_UMSCH(regVCN_MES_GP0_HI, upper_32_bits(umsch->log_gpu_addr));
+#endif
+
+	WREG32_SOC15_UMSCH(regVCN_MES_GP1_LO, 0);
+	WREG32_SOC15_UMSCH(regVCN_MES_GP1_HI, 0);
+
+	data = RREG32_SOC15(VCN, 0, regVCN_MES_CNTL);
+	data = REG_SET_FIELD(data, VCN_MES_CNTL, MES_INVALIDATE_ICACHE, 0);
+	data = REG_SET_FIELD(data, VCN_MES_CNTL, MES_PIPE0_RESET, 0);
+	data = REG_SET_FIELD(data, VCN_MES_CNTL, MES_HALT, 0);
+	data = REG_SET_FIELD(data, VCN_MES_CNTL, MES_PIPE0_ACTIVE, 1);
+	WREG32_SOC15_UMSCH(regVCN_MES_CNTL, data);
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP)
+		amdgpu_umsch_mm_psp_execute_cmd_buf(umsch);
+
+	r = SOC15_WAIT_ON_RREG(VCN, 0, regVCN_MES_MSTATUS_LO, 0xAAAAAAAA, 0xFFFFFFFF);
+	if (r) {
+		dev_err(adev->dev, "UMSCH FW Load: Failed, regVCN_MES_MSTATUS_LO: 0x%08x\n",
+			RREG32_SOC15(VCN, 0, regVCN_MES_MSTATUS_LO));
+		goto err_free_data_bo;
+	}
+
+	return 0;
+
+err_free_data_bo:
+	amdgpu_bo_free_kernel(&adev->umsch_mm.data_fw_obj,
+			      &adev->umsch_mm.data_fw_gpu_addr,
+			      (void **)&adev->umsch_mm.data_fw_ptr);
+err_free_ucode_bo:
+	amdgpu_bo_free_kernel(&adev->umsch_mm.ucode_fw_obj,
+			      &adev->umsch_mm.ucode_fw_gpu_addr,
+			      (void **)&adev->umsch_mm.ucode_fw_ptr);
+	return r;
+}
+
+static void umsch_mm_v4_0_aggregated_doorbell_init(struct amdgpu_umsch_mm *umsch)
+{
+	struct amdgpu_device *adev = umsch->ring.adev;
+	uint32_t data;
+
+	data = RREG32_SOC15(VCN, 0, regVCN_AGDB_CTRL0);
+	data = REG_SET_FIELD(data, VCN_AGDB_CTRL0, OFFSET,
+	       umsch->agdb_index[CONTEXT_PRIORITY_LEVEL_REALTIME]);
+	data = REG_SET_FIELD(data, VCN_AGDB_CTRL0, EN, 1);
+	WREG32_SOC15(VCN, 0, regVCN_AGDB_CTRL0, data);
+
+	data = RREG32_SOC15(VCN, 0, regVCN_AGDB_CTRL1);
+	data = REG_SET_FIELD(data, VCN_AGDB_CTRL1, OFFSET,
+	       umsch->agdb_index[CONTEXT_PRIORITY_LEVEL_FOCUS]);
+	data = REG_SET_FIELD(data, VCN_AGDB_CTRL1, EN, 1);
+	WREG32_SOC15(VCN, 0, regVCN_AGDB_CTRL1, data);
+
+	data = RREG32_SOC15(VCN, 0, regVCN_AGDB_CTRL2);
+	data = REG_SET_FIELD(data, VCN_AGDB_CTRL2, OFFSET,
+	       umsch->agdb_index[CONTEXT_PRIORITY_LEVEL_NORMAL]);
+	data = REG_SET_FIELD(data, VCN_AGDB_CTRL2, EN, 1);
+	WREG32_SOC15(VCN, 0, regVCN_AGDB_CTRL2, data);
+
+	data = RREG32_SOC15(VCN, 0, regVCN_AGDB_CTRL3);
+	data = REG_SET_FIELD(data, VCN_AGDB_CTRL3, OFFSET,
+	       umsch->agdb_index[CONTEXT_PRIORITY_LEVEL_IDLE]);
+	data = REG_SET_FIELD(data, VCN_AGDB_CTRL3, EN, 1);
+	WREG32_SOC15(VCN, 0, regVCN_AGDB_CTRL3, data);
+}
+
+static int umsch_mm_v4_0_ring_start(struct amdgpu_umsch_mm *umsch)
+{
+	struct amdgpu_ring *ring = &umsch->ring;
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t data;
+
+	data = RREG32_SOC15(VCN, 0, regVCN_UMSCH_RB_DB_CTRL);
+	data = REG_SET_FIELD(data, VCN_UMSCH_RB_DB_CTRL, OFFSET, ring->doorbell_index);
+	data = REG_SET_FIELD(data, VCN_UMSCH_RB_DB_CTRL, EN, 1);
+	WREG32_SOC15(VCN, 0, regVCN_UMSCH_RB_DB_CTRL, data);
+
+	adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell,
+		(adev->doorbell_index.vcn.vcn_ring0_1 << 1), 0);
+
+	WREG32_SOC15(VCN, 0, regVCN_UMSCH_RB_BASE_LO, lower_32_bits(ring->gpu_addr));
+	WREG32_SOC15(VCN, 0, regVCN_UMSCH_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
+
+	WREG32_SOC15(VCN, 0, regVCN_UMSCH_RB_SIZE, ring->ring_size);
+
+	ring->wptr = 0;
+
+	data = RREG32_SOC15(VCN, 0, regVCN_RB_ENABLE);
+	data &= ~(VCN_RB_ENABLE__AUDIO_RB_EN_MASK);
+	WREG32_SOC15(VCN, 0, regVCN_RB_ENABLE, data);
+
+	umsch_mm_v4_0_aggregated_doorbell_init(umsch);
+
+	return 0;
+}
+
+static int umsch_mm_v4_0_ring_stop(struct amdgpu_umsch_mm *umsch)
+{
+	struct amdgpu_ring *ring = &umsch->ring;
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t data;
+
+	data = RREG32_SOC15(VCN, 0, regVCN_RB_ENABLE);
+	data = REG_SET_FIELD(data, VCN_RB_ENABLE, UMSCH_RB_EN, 0);
+	WREG32_SOC15(VCN, 0, regVCN_RB_ENABLE, data);
+
+	data = RREG32_SOC15(VCN, 0, regVCN_UMSCH_RB_DB_CTRL);
+	data = REG_SET_FIELD(data, VCN_UMSCH_RB_DB_CTRL, EN, 0);
+	WREG32_SOC15(VCN, 0, regVCN_UMSCH_RB_DB_CTRL, data);
+
+	if (amdgpu_ip_version(adev, VCN_HWIP, 0) >= IP_VERSION(4, 0, 5)) {
+		WREG32_SOC15(VCN, 0, regUVD_IPX_DLDO_CONFIG,
+			2 << UVD_IPX_DLDO_CONFIG__ONO0_PWR_CONFIG__SHIFT);
+		SOC15_WAIT_ON_RREG(VCN, 0, regUVD_IPX_DLDO_STATUS,
+			1 << UVD_IPX_DLDO_STATUS__ONO0_PWR_STATUS__SHIFT,
+			UVD_IPX_DLDO_STATUS__ONO0_PWR_STATUS_MASK);
+	}
+
+	return 0;
+}
+
+static int umsch_mm_v4_0_set_hw_resources(struct amdgpu_umsch_mm *umsch)
+{
+	union UMSCHAPI__SET_HW_RESOURCES set_hw_resources = {};
+	struct amdgpu_device *adev = umsch->ring.adev;
+	int r;
+
+	set_hw_resources.header.type = UMSCH_API_TYPE_SCHEDULER;
+	set_hw_resources.header.opcode = UMSCH_API_SET_HW_RSRC;
+	set_hw_resources.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	set_hw_resources.vmid_mask_mm_vcn = umsch->vmid_mask_mm_vcn;
+	set_hw_resources.vmid_mask_mm_vpe = umsch->vmid_mask_mm_vpe;
+	set_hw_resources.collaboration_mask_vpe =
+		adev->vpe.collaborate_mode ? 0x3 : 0x0;
+	set_hw_resources.engine_mask = umsch->engine_mask;
+
+	set_hw_resources.vcn0_hqd_mask[0] = umsch->vcn0_hqd_mask;
+	set_hw_resources.vcn1_hqd_mask[0] = umsch->vcn1_hqd_mask;
+	set_hw_resources.vcn_hqd_mask[0] = umsch->vcn_hqd_mask[0];
+	set_hw_resources.vcn_hqd_mask[1] = umsch->vcn_hqd_mask[1];
+	set_hw_resources.vpe_hqd_mask[0] = umsch->vpe_hqd_mask;
+
+	set_hw_resources.g_sch_ctx_gpu_mc_ptr = umsch->sch_ctx_gpu_addr;
+
+	set_hw_resources.enable_level_process_quantum_check = 1;
+
+	memcpy(set_hw_resources.mmhub_base, adev->reg_offset[MMHUB_HWIP][0],
+	       sizeof(uint32_t) * 5);
+	set_hw_resources.mmhub_version =
+		IP_VERSION_MAJ_MIN_REV(amdgpu_ip_version(adev, MMHUB_HWIP, 0));
+
+	memcpy(set_hw_resources.osssys_base, adev->reg_offset[OSSSYS_HWIP][0],
+	       sizeof(uint32_t) * 5);
+	set_hw_resources.osssys_version =
+		IP_VERSION_MAJ_MIN_REV(amdgpu_ip_version(adev, OSSSYS_HWIP, 0));
+
+	set_hw_resources.vcn_version =
+		IP_VERSION_MAJ_MIN_REV(amdgpu_ip_version(adev, VCN_HWIP, 0));
+	set_hw_resources.vpe_version =
+		IP_VERSION_MAJ_MIN_REV(amdgpu_ip_version(adev, VPE_HWIP, 0));
+
+	set_hw_resources.api_status.api_completion_fence_addr = umsch->ring.fence_drv.gpu_addr;
+	set_hw_resources.api_status.api_completion_fence_value = ++umsch->ring.fence_drv.sync_seq;
+
+	r = amdgpu_umsch_mm_submit_pkt(umsch, &set_hw_resources.max_dwords_in_api,
+				       API_FRAME_SIZE_IN_DWORDS);
+	if (r)
+		return r;
+
+	r = amdgpu_umsch_mm_query_fence(umsch);
+	if (r) {
+		dev_err(adev->dev, "UMSCH SET_HW_RESOURCES: Failed\n");
+		return r;
+	}
+
+	return 0;
+}
+
+static int umsch_mm_v4_0_add_queue(struct amdgpu_umsch_mm *umsch,
+				   struct umsch_mm_add_queue_input *input_ptr)
+{
+	struct amdgpu_device *adev = umsch->ring.adev;
+	union UMSCHAPI__ADD_QUEUE add_queue = {};
+	int r;
+
+	add_queue.header.type = UMSCH_API_TYPE_SCHEDULER;
+	add_queue.header.opcode = UMSCH_API_ADD_QUEUE;
+	add_queue.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	add_queue.process_id = input_ptr->process_id;
+	add_queue.page_table_base_addr = input_ptr->page_table_base_addr;
+	add_queue.process_va_start = input_ptr->process_va_start;
+	add_queue.process_va_end = input_ptr->process_va_end;
+	add_queue.process_quantum = input_ptr->process_quantum;
+	add_queue.process_csa_addr = input_ptr->process_csa_addr;
+	add_queue.context_quantum = input_ptr->context_quantum;
+	add_queue.context_csa_addr = input_ptr->context_csa_addr;
+	add_queue.inprocess_context_priority = input_ptr->inprocess_context_priority;
+	add_queue.context_global_priority_level =
+		(enum UMSCH_AMD_PRIORITY_LEVEL)input_ptr->context_global_priority_level;
+	add_queue.doorbell_offset_0 = input_ptr->doorbell_offset_0;
+	add_queue.doorbell_offset_1 = input_ptr->doorbell_offset_1;
+	add_queue.affinity.u32All = input_ptr->affinity;
+	add_queue.mqd_addr = input_ptr->mqd_addr;
+	add_queue.engine_type = (enum UMSCH_ENGINE_TYPE)input_ptr->engine_type;
+	add_queue.h_context = input_ptr->h_context;
+	add_queue.h_queue = input_ptr->h_queue;
+	add_queue.vm_context_cntl = input_ptr->vm_context_cntl;
+	add_queue.is_context_suspended = input_ptr->is_context_suspended;
+	add_queue.collaboration_mode = adev->vpe.collaborate_mode ? 1 : 0;
+
+	add_queue.api_status.api_completion_fence_addr = umsch->ring.fence_drv.gpu_addr;
+	add_queue.api_status.api_completion_fence_value = ++umsch->ring.fence_drv.sync_seq;
+
+	r = amdgpu_umsch_mm_submit_pkt(umsch, &add_queue.max_dwords_in_api,
+				       API_FRAME_SIZE_IN_DWORDS);
+	if (r)
+		return r;
+
+	r = amdgpu_umsch_mm_query_fence(umsch);
+	if (r) {
+		dev_err(adev->dev, "UMSCH ADD_QUEUE: Failed\n");
+		return r;
+	}
+
+	return 0;
+}
+
+static int umsch_mm_v4_0_remove_queue(struct amdgpu_umsch_mm *umsch,
+				      struct umsch_mm_remove_queue_input *input_ptr)
+{
+	union UMSCHAPI__REMOVE_QUEUE remove_queue = {};
+	struct amdgpu_device *adev = umsch->ring.adev;
+	int r;
+
+	remove_queue.header.type = UMSCH_API_TYPE_SCHEDULER;
+	remove_queue.header.opcode = UMSCH_API_REMOVE_QUEUE;
+	remove_queue.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+
+	remove_queue.doorbell_offset_0 = input_ptr->doorbell_offset_0;
+	remove_queue.doorbell_offset_1 = input_ptr->doorbell_offset_1;
+	remove_queue.context_csa_addr = input_ptr->context_csa_addr;
+
+	remove_queue.api_status.api_completion_fence_addr = umsch->ring.fence_drv.gpu_addr;
+	remove_queue.api_status.api_completion_fence_value = ++umsch->ring.fence_drv.sync_seq;
+
+	r = amdgpu_umsch_mm_submit_pkt(umsch, &remove_queue.max_dwords_in_api,
+				       API_FRAME_SIZE_IN_DWORDS);
+	if (r)
+		return r;
+
+	r = amdgpu_umsch_mm_query_fence(umsch);
+	if (r) {
+		dev_err(adev->dev, "UMSCH REMOVE_QUEUE: Failed\n");
+		return r;
+	}
+
+	return 0;
+}
+
+static int umsch_mm_v4_0_set_regs(struct amdgpu_umsch_mm *umsch)
+{
+	struct amdgpu_device *adev = container_of(umsch, struct amdgpu_device, umsch_mm);
+
+	umsch->rb_wptr = SOC15_REG_OFFSET(VCN, 0, regVCN_UMSCH_RB_WPTR);
+	umsch->rb_rptr = SOC15_REG_OFFSET(VCN, 0, regVCN_UMSCH_RB_RPTR);
+
+	return 0;
+}
+
+static const struct umsch_mm_funcs umsch_mm_v4_0_funcs = {
+	.set_hw_resources = umsch_mm_v4_0_set_hw_resources,
+	.add_queue = umsch_mm_v4_0_add_queue,
+	.remove_queue = umsch_mm_v4_0_remove_queue,
+	.set_regs = umsch_mm_v4_0_set_regs,
+	.init_microcode = amdgpu_umsch_mm_init_microcode,
+	.load_microcode = umsch_mm_v4_0_load_microcode,
+	.ring_init = amdgpu_umsch_mm_ring_init,
+	.ring_start = umsch_mm_v4_0_ring_start,
+	.ring_stop = umsch_mm_v4_0_ring_stop,
+};
+
+void umsch_mm_v4_0_set_funcs(struct amdgpu_umsch_mm *umsch)
+{
+	umsch->funcs = &umsch_mm_v4_0_funcs;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.h b/drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.h
new file mode 100644
index 000000000000..06bc0fa74996
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.h
@@ -0,0 +1,30 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __UMSCH_MM_V4_0_H__
+#define __UMSCH_MM_V4_0_H__
+
+void umsch_mm_v4_0_set_funcs(struct amdgpu_umsch_mm *umsch);
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v3_1.c b/drivers/gpu/drm/amd/amdgpu/uvd_v3_1.c
index 0fef925b6602..2e79a3afc774 100644
--- a/drivers/gpu/drm/amd/amdgpu/uvd_v3_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/uvd_v3_1.c
@@ -98,7 +98,7 @@ static void uvd_v3_1_ring_emit_ib(struct amdgpu_ring *ring,
 }
 
 /**
- * uvd_v3_1_ring_emit_fence - emit an fence & trap command
+ * uvd_v3_1_ring_emit_fence - emit a fence & trap command
  *
  * @ring: amdgpu_ring pointer
  * @addr: address
@@ -242,7 +242,7 @@ static void uvd_v3_1_mc_resume(struct amdgpu_device *adev)
 	uint64_t addr;
 	uint32_t size;
 
-	/* programm the VCPU memory controller bits 0-27 */
+	/* program the VCPU memory controller bits 0-27 */
 	addr = (adev->uvd.inst->gpu_addr + AMDGPU_UVD_FIRMWARE_OFFSET) >> 3;
 	size = AMDGPU_UVD_FIRMWARE_SIZE(adev) >> 3;
 	WREG32(mmUVD_VCPU_CACHE_OFFSET0, addr);
@@ -416,7 +416,7 @@ static int uvd_v3_1_start(struct amdgpu_device *adev)
 	/* Set the write pointer delay */
 	WREG32(mmUVD_RBC_RB_WPTR_CNTL, 0);
 
-	/* programm the 4GB memory segment for rptr and ring buffer */
+	/* Program the 4GB memory segment for rptr and ring buffer */
 	WREG32(mmUVD_LMI_EXT40_ADDR, upper_32_bits(ring->gpu_addr) |
 		   (0x7 << 16) | (0x1 << 31));
 
@@ -531,9 +531,9 @@ static void uvd_v3_1_set_irq_funcs(struct amdgpu_device *adev)
 }
 
 
-static int uvd_v3_1_early_init(void *handle)
+static int uvd_v3_1_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	adev->uvd.num_uvd_inst = 1;
 
 	uvd_v3_1_set_ring_funcs(adev);
@@ -542,10 +542,10 @@ static int uvd_v3_1_early_init(void *handle)
 	return 0;
 }
 
-static int uvd_v3_1_sw_init(void *handle)
+static int uvd_v3_1_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 	void *ptr;
 	uint32_t ucode_len;
@@ -577,15 +577,13 @@ static int uvd_v3_1_sw_init(void *handle)
 	ptr += ucode_len;
 	memcpy(&adev->uvd.keyselect, ptr, 4);
 
-	r = amdgpu_uvd_entity_init(adev);
-
 	return r;
 }
 
-static int uvd_v3_1_sw_fini(void *handle)
+static int uvd_v3_1_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = amdgpu_uvd_suspend(adev);
 	if (r)
@@ -623,18 +621,42 @@ static void uvd_v3_1_enable_mgcg(struct amdgpu_device *adev,
 /**
  * uvd_v3_1_hw_init - start and test UVD block
  *
- * @handle: handle used to pass amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Initialize the hardware, boot up the VCPU and do some testing.
  *
- * Initialize the hardware, boot up the VCPU and do some testing
+ * On SI, the UVD is meant to be used in a specific power state,
+ * or alternatively the driver can manually enable its clock.
+ * In amdgpu we use the dedicated UVD power state when DPM is enabled.
+ * Calling amdgpu_dpm_enable_uvd makes DPM select the UVD power state
+ * for the SMU and afterwards enables the UVD clock.
+ * This is automatically done by amdgpu_uvd_ring_begin_use when work
+ * is submitted to the UVD ring. Here, we have to call it manually
+ * in order to power up UVD before firmware validation.
+ *
+ * Note that we must not disable the UVD clock here, as that would
+ * cause the ring test to fail. However, UVD is powered off
+ * automatically after the ring test: amdgpu_uvd_ring_end_use calls
+ * the UVD idle work handler which will disable the UVD clock when
+ * all fences are signalled.
  */
-static int uvd_v3_1_hw_init(void *handle)
+static int uvd_v3_1_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring = &adev->uvd.inst->ring;
 	uint32_t tmp;
 	int r;
 
 	uvd_v3_1_mc_resume(adev);
+	uvd_v3_1_enable_mgcg(adev, true);
+
+	/* Make sure UVD is powered during FW validation.
+	 * It's going to be automatically powered off after the ring test.
+	 */
+	if (adev->pm.dpm_enabled)
+		amdgpu_dpm_enable_uvd(adev, true);
+	else
+		amdgpu_asic_set_uvd_clocks(adev, 53300, 40000);
 
 	r = uvd_v3_1_fw_validate(adev);
 	if (r) {
@@ -642,9 +664,6 @@ static int uvd_v3_1_hw_init(void *handle)
 		return r;
 	}
 
-	uvd_v3_1_enable_mgcg(adev, true);
-	amdgpu_asic_set_uvd_clocks(adev, 53300, 40000);
-
 	uvd_v3_1_start(adev);
 
 	r = amdgpu_ring_test_helper(ring);
@@ -690,13 +709,13 @@ done:
 /**
  * uvd_v3_1_hw_fini - stop the hardware block
  *
- * @handle: handle used to pass amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Stop the UVD block, mark ring as not ready any more
  */
-static int uvd_v3_1_hw_fini(void *handle)
+static int uvd_v3_1_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	cancel_delayed_work_sync(&adev->uvd.idle_work);
 
@@ -706,10 +725,17 @@ static int uvd_v3_1_hw_fini(void *handle)
 	return 0;
 }
 
-static int uvd_v3_1_suspend(void *handle)
+static int uvd_v3_1_prepare_suspend(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	return amdgpu_uvd_prepare_suspend(adev);
+}
+
+static int uvd_v3_1_suspend(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/*
 	 * Proper cleanups before halting the HW engine:
@@ -735,36 +761,35 @@ static int uvd_v3_1_suspend(void *handle)
 						       AMD_CG_STATE_GATE);
 	}
 
-	r = uvd_v3_1_hw_fini(adev);
+	r = uvd_v3_1_hw_fini(ip_block);
 	if (r)
 		return r;
 
 	return amdgpu_uvd_suspend(adev);
 }
 
-static int uvd_v3_1_resume(void *handle)
+static int uvd_v3_1_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = amdgpu_uvd_resume(adev);
+	r = amdgpu_uvd_resume(ip_block->adev);
 	if (r)
 		return r;
 
-	return uvd_v3_1_hw_init(adev);
+	return uvd_v3_1_hw_init(ip_block);
 }
 
-static bool uvd_v3_1_is_idle(void *handle)
+static bool uvd_v3_1_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return !(RREG32(mmSRBM_STATUS) & SRBM_STATUS__UVD_BUSY_MASK);
 }
 
-static int uvd_v3_1_wait_for_idle(void *handle)
+static int uvd_v3_1_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		if (!(RREG32(mmSRBM_STATUS) & SRBM_STATUS__UVD_BUSY_MASK))
@@ -773,9 +798,9 @@ static int uvd_v3_1_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int uvd_v3_1_soft_reset(void *handle)
+static int uvd_v3_1_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	uvd_v3_1_stop(adev);
 
@@ -786,13 +811,13 @@ static int uvd_v3_1_soft_reset(void *handle)
 	return uvd_v3_1_start(adev);
 }
 
-static int uvd_v3_1_set_clockgating_state(void *handle,
+static int uvd_v3_1_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int uvd_v3_1_set_powergating_state(void *handle,
+static int uvd_v3_1_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -801,11 +826,11 @@ static int uvd_v3_1_set_powergating_state(void *handle,
 static const struct amd_ip_funcs uvd_v3_1_ip_funcs = {
 	.name = "uvd_v3_1",
 	.early_init = uvd_v3_1_early_init,
-	.late_init = NULL,
 	.sw_init = uvd_v3_1_sw_init,
 	.sw_fini = uvd_v3_1_sw_fini,
 	.hw_init = uvd_v3_1_hw_init,
 	.hw_fini = uvd_v3_1_hw_fini,
+	.prepare_suspend = uvd_v3_1_prepare_suspend,
 	.suspend = uvd_v3_1_suspend,
 	.resume = uvd_v3_1_resume,
 	.is_idle = uvd_v3_1_is_idle,
@@ -815,8 +840,7 @@ static const struct amd_ip_funcs uvd_v3_1_ip_funcs = {
 	.set_powergating_state = uvd_v3_1_set_powergating_state,
 };
 
-const struct amdgpu_ip_block_version uvd_v3_1_ip_block =
-{
+const struct amdgpu_ip_block_version uvd_v3_1_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_UVD,
 	.major = 3,
 	.minor = 1,
diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c
index c108b8381795..4b96fd583772 100644
--- a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c
@@ -44,7 +44,7 @@ static void uvd_v4_2_set_ring_funcs(struct amdgpu_device *adev);
 static void uvd_v4_2_set_irq_funcs(struct amdgpu_device *adev);
 static int uvd_v4_2_start(struct amdgpu_device *adev);
 static void uvd_v4_2_stop(struct amdgpu_device *adev);
-static int uvd_v4_2_set_clockgating_state(void *handle,
+static int uvd_v4_2_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 				enum amd_clockgating_state state);
 static void uvd_v4_2_set_dcm(struct amdgpu_device *adev,
 			     bool sw_mode);
@@ -90,9 +90,9 @@ static void uvd_v4_2_ring_set_wptr(struct amdgpu_ring *ring)
 	WREG32(mmUVD_RBC_RB_WPTR, lower_32_bits(ring->wptr));
 }
 
-static int uvd_v4_2_early_init(void *handle)
+static int uvd_v4_2_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	adev->uvd.num_uvd_inst = 1;
 
 	uvd_v4_2_set_ring_funcs(adev);
@@ -101,10 +101,10 @@ static int uvd_v4_2_early_init(void *handle)
 	return 0;
 }
 
-static int uvd_v4_2_sw_init(void *handle)
+static int uvd_v4_2_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	/* UVD TRAP */
@@ -127,15 +127,13 @@ static int uvd_v4_2_sw_init(void *handle)
 	if (r)
 		return r;
 
-	r = amdgpu_uvd_entity_init(adev);
-
 	return r;
 }
 
-static int uvd_v4_2_sw_fini(void *handle)
+static int uvd_v4_2_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = amdgpu_uvd_suspend(adev);
 	if (r)
@@ -149,13 +147,13 @@ static void uvd_v4_2_enable_mgcg(struct amdgpu_device *adev,
 /**
  * uvd_v4_2_hw_init - start and test UVD block
  *
- * @handle: handle used to pass amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Initialize the hardware, boot up the VCPU and do some testing
  */
-static int uvd_v4_2_hw_init(void *handle)
+static int uvd_v4_2_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring = &adev->uvd.inst->ring;
 	uint32_t tmp;
 	int r;
@@ -204,13 +202,13 @@ done:
 /**
  * uvd_v4_2_hw_fini - stop the hardware block
  *
- * @handle: handle used to pass amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Stop the UVD block, mark ring as not ready any more
  */
-static int uvd_v4_2_hw_fini(void *handle)
+static int uvd_v4_2_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	cancel_delayed_work_sync(&adev->uvd.idle_work);
 
@@ -220,10 +218,17 @@ static int uvd_v4_2_hw_fini(void *handle)
 	return 0;
 }
 
-static int uvd_v4_2_suspend(void *handle)
+static int uvd_v4_2_prepare_suspend(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	return amdgpu_uvd_prepare_suspend(adev);
+}
+
+static int uvd_v4_2_suspend(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/*
 	 * Proper cleanups before halting the HW engine:
@@ -249,23 +254,22 @@ static int uvd_v4_2_suspend(void *handle)
 						       AMD_CG_STATE_GATE);
 	}
 
-	r = uvd_v4_2_hw_fini(adev);
+	r = uvd_v4_2_hw_fini(ip_block);
 	if (r)
 		return r;
 
 	return amdgpu_uvd_suspend(adev);
 }
 
-static int uvd_v4_2_resume(void *handle)
+static int uvd_v4_2_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = amdgpu_uvd_resume(adev);
+	r = amdgpu_uvd_resume(ip_block->adev);
 	if (r)
 		return r;
 
-	return uvd_v4_2_hw_init(adev);
+	return uvd_v4_2_hw_init(ip_block);
 }
 
 /**
@@ -298,7 +302,7 @@ static int uvd_v4_2_start(struct amdgpu_device *adev)
 	/* enable VCPU clock */
 	WREG32(mmUVD_VCPU_CNTL,  1 << 9);
 
-	/* disable interupt */
+	/* disable interrupt */
 	WREG32_P(mmUVD_MASTINT_EN, 0, ~(1 << 1));
 
 #ifdef __BIG_ENDIAN
@@ -308,6 +312,7 @@ static int uvd_v4_2_start(struct amdgpu_device *adev)
 #endif
 	WREG32(mmUVD_LMI_SWAP_CNTL, lmi_swap_cntl);
 	WREG32(mmUVD_MP_SWAP_CNTL, mp_swap_cntl);
+
 	/* initialize UVD memory controller */
 	WREG32(mmUVD_LMI_CTRL, 0x203108);
 
@@ -654,17 +659,17 @@ static void uvd_v4_2_set_dcm(struct amdgpu_device *adev,
 	WREG32_UVD_CTX(ixUVD_CGC_CTRL2, tmp2);
 }
 
-static bool uvd_v4_2_is_idle(void *handle)
+static bool uvd_v4_2_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return !(RREG32(mmSRBM_STATUS) & SRBM_STATUS__UVD_BUSY_MASK);
 }
 
-static int uvd_v4_2_wait_for_idle(void *handle)
+static int uvd_v4_2_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		if (!(RREG32(mmSRBM_STATUS) & SRBM_STATUS__UVD_BUSY_MASK))
@@ -673,9 +678,9 @@ static int uvd_v4_2_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int uvd_v4_2_soft_reset(void *handle)
+static int uvd_v4_2_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	uvd_v4_2_stop(adev);
 
@@ -704,13 +709,13 @@ static int uvd_v4_2_process_interrupt(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int uvd_v4_2_set_clockgating_state(void *handle,
+static int uvd_v4_2_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int uvd_v4_2_set_powergating_state(void *handle,
+static int uvd_v4_2_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	/* This doesn't actually powergate the UVD block.
@@ -720,7 +725,7 @@ static int uvd_v4_2_set_powergating_state(void *handle,
 	 * revisit this when there is a cleaner line between
 	 * the smc and the hw blocks
 	 */
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (state == AMD_PG_STATE_GATE) {
 		uvd_v4_2_stop(adev);
@@ -751,11 +756,11 @@ static int uvd_v4_2_set_powergating_state(void *handle,
 static const struct amd_ip_funcs uvd_v4_2_ip_funcs = {
 	.name = "uvd_v4_2",
 	.early_init = uvd_v4_2_early_init,
-	.late_init = NULL,
 	.sw_init = uvd_v4_2_sw_init,
 	.sw_fini = uvd_v4_2_sw_fini,
 	.hw_init = uvd_v4_2_hw_init,
 	.hw_fini = uvd_v4_2_hw_fini,
+	.prepare_suspend = uvd_v4_2_prepare_suspend,
 	.suspend = uvd_v4_2_suspend,
 	.resume = uvd_v4_2_resume,
 	.is_idle = uvd_v4_2_is_idle,
diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c
index d7e31e48a2b8..71409ad8b7ed 100644
--- a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c
@@ -42,7 +42,7 @@ static void uvd_v5_0_set_ring_funcs(struct amdgpu_device *adev);
 static void uvd_v5_0_set_irq_funcs(struct amdgpu_device *adev);
 static int uvd_v5_0_start(struct amdgpu_device *adev);
 static void uvd_v5_0_stop(struct amdgpu_device *adev);
-static int uvd_v5_0_set_clockgating_state(void *handle,
+static int uvd_v5_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state);
 static void uvd_v5_0_enable_mgcg(struct amdgpu_device *adev,
 				 bool enable);
@@ -88,9 +88,9 @@ static void uvd_v5_0_ring_set_wptr(struct amdgpu_ring *ring)
 	WREG32(mmUVD_RBC_RB_WPTR, lower_32_bits(ring->wptr));
 }
 
-static int uvd_v5_0_early_init(void *handle)
+static int uvd_v5_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	adev->uvd.num_uvd_inst = 1;
 
 	uvd_v5_0_set_ring_funcs(adev);
@@ -99,10 +99,10 @@ static int uvd_v5_0_early_init(void *handle)
 	return 0;
 }
 
-static int uvd_v5_0_sw_init(void *handle)
+static int uvd_v5_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	/* UVD TRAP */
@@ -125,15 +125,13 @@ static int uvd_v5_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	r = amdgpu_uvd_entity_init(adev);
-
 	return r;
 }
 
-static int uvd_v5_0_sw_fini(void *handle)
+static int uvd_v5_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = amdgpu_uvd_suspend(adev);
 	if (r)
@@ -145,19 +143,19 @@ static int uvd_v5_0_sw_fini(void *handle)
 /**
  * uvd_v5_0_hw_init - start and test UVD block
  *
- * @handle: handle used to pass amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Initialize the hardware, boot up the VCPU and do some testing
  */
-static int uvd_v5_0_hw_init(void *handle)
+static int uvd_v5_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring = &adev->uvd.inst->ring;
 	uint32_t tmp;
 	int r;
 
 	amdgpu_asic_set_uvd_clocks(adev, 10000, 10000);
-	uvd_v5_0_set_clockgating_state(adev, AMD_CG_STATE_UNGATE);
+	uvd_v5_0_set_clockgating_state(ip_block, AMD_CG_STATE_UNGATE);
 	uvd_v5_0_enable_mgcg(adev, true);
 
 	r = amdgpu_ring_test_helper(ring);
@@ -202,13 +200,13 @@ done:
 /**
  * uvd_v5_0_hw_fini - stop the hardware block
  *
- * @handle: handle used to pass amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Stop the UVD block, mark ring as not ready any more
  */
-static int uvd_v5_0_hw_fini(void *handle)
+static int uvd_v5_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	cancel_delayed_work_sync(&adev->uvd.idle_work);
 
@@ -218,10 +216,17 @@ static int uvd_v5_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int uvd_v5_0_suspend(void *handle)
+static int uvd_v5_0_prepare_suspend(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	return amdgpu_uvd_prepare_suspend(adev);
+}
+
+static int uvd_v5_0_suspend(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/*
 	 * Proper cleanups before halting the HW engine:
@@ -247,23 +252,22 @@ static int uvd_v5_0_suspend(void *handle)
 						       AMD_CG_STATE_GATE);
 	}
 
-	r = uvd_v5_0_hw_fini(adev);
+	r = uvd_v5_0_hw_fini(ip_block);
 	if (r)
 		return r;
 
 	return amdgpu_uvd_suspend(adev);
 }
 
-static int uvd_v5_0_resume(void *handle)
+static int uvd_v5_0_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = amdgpu_uvd_resume(adev);
+	r = amdgpu_uvd_resume(ip_block->adev);
 	if (r)
 		return r;
 
-	return uvd_v5_0_hw_init(adev);
+	return uvd_v5_0_hw_init(ip_block);
 }
 
 /**
@@ -576,17 +580,17 @@ static void uvd_v5_0_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
 	}
 }
 
-static bool uvd_v5_0_is_idle(void *handle)
+static bool uvd_v5_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return !(RREG32(mmSRBM_STATUS) & SRBM_STATUS__UVD_BUSY_MASK);
 }
 
-static int uvd_v5_0_wait_for_idle(void *handle)
+static int uvd_v5_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		if (!(RREG32(mmSRBM_STATUS) & SRBM_STATUS__UVD_BUSY_MASK))
@@ -595,9 +599,9 @@ static int uvd_v5_0_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int uvd_v5_0_soft_reset(void *handle)
+static int uvd_v5_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	uvd_v5_0_stop(adev);
 
@@ -786,15 +790,15 @@ static void uvd_v5_0_enable_mgcg(struct amdgpu_device *adev,
 	}
 }
 
-static int uvd_v5_0_set_clockgating_state(void *handle,
+static int uvd_v5_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_CG_STATE_GATE);
 
 	if (enable) {
 		/* wait for STATUS to clear */
-		if (uvd_v5_0_wait_for_idle(handle))
+		if (uvd_v5_0_wait_for_idle(ip_block))
 			return -EBUSY;
 		uvd_v5_0_enable_clock_gating(adev, true);
 
@@ -808,7 +812,7 @@ static int uvd_v5_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int uvd_v5_0_set_powergating_state(void *handle,
+static int uvd_v5_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	/* This doesn't actually powergate the UVD block.
@@ -818,7 +822,7 @@ static int uvd_v5_0_set_powergating_state(void *handle,
 	 * revisit this when there is a cleaner line between
 	 * the smc and the hw blocks
 	 */
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret = 0;
 
 	if (state == AMD_PG_STATE_GATE) {
@@ -833,9 +837,9 @@ out:
 	return ret;
 }
 
-static void uvd_v5_0_get_clockgating_state(void *handle, u64 *flags)
+static void uvd_v5_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	mutex_lock(&adev->pm.mutex);
@@ -858,11 +862,11 @@ out:
 static const struct amd_ip_funcs uvd_v5_0_ip_funcs = {
 	.name = "uvd_v5_0",
 	.early_init = uvd_v5_0_early_init,
-	.late_init = NULL,
 	.sw_init = uvd_v5_0_sw_init,
 	.sw_fini = uvd_v5_0_sw_fini,
 	.hw_init = uvd_v5_0_hw_init,
 	.hw_fini = uvd_v5_0_hw_fini,
+	.prepare_suspend = uvd_v5_0_prepare_suspend,
 	.suspend = uvd_v5_0_suspend,
 	.resume = uvd_v5_0_resume,
 	.is_idle = uvd_v5_0_is_idle,
diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c
index 5fe872f4bea7..ceb94bbb03a4 100644
--- a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c
@@ -48,7 +48,7 @@ static void uvd_v6_0_set_irq_funcs(struct amdgpu_device *adev);
 static int uvd_v6_0_start(struct amdgpu_device *adev);
 static void uvd_v6_0_stop(struct amdgpu_device *adev);
 static void uvd_v6_0_set_sw_clock_gating(struct amdgpu_device *adev);
-static int uvd_v6_0_set_clockgating_state(void *handle,
+static int uvd_v6_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state);
 static void uvd_v6_0_enable_mgcg(struct amdgpu_device *adev,
 				 bool enable);
@@ -217,7 +217,8 @@ static int uvd_v6_0_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t handle
 	int i, r;
 
 	r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, ib_size_dw * 4,
-				     AMDGPU_IB_POOL_DIRECT, &job);
+				     AMDGPU_IB_POOL_DIRECT, &job,
+				     AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		return r;
 
@@ -281,7 +282,8 @@ static int uvd_v6_0_enc_get_destroy_msg(struct amdgpu_ring *ring,
 	int i, r;
 
 	r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, ib_size_dw * 4,
-				     AMDGPU_IB_POOL_DIRECT, &job);
+				     AMDGPU_IB_POOL_DIRECT, &job,
+				     AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		return r;
 
@@ -354,9 +356,9 @@ error:
 	return r;
 }
 
-static int uvd_v6_0_early_init(void *handle)
+static int uvd_v6_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	adev->uvd.num_uvd_inst = 1;
 
 	if (!(adev->flags & AMD_IS_APU) &&
@@ -375,11 +377,11 @@ static int uvd_v6_0_early_init(void *handle)
 	return 0;
 }
 
-static int uvd_v6_0_sw_init(void *handle)
+static int uvd_v6_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
 	int i, r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* UVD TRAP */
 	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_UVD_SYSTEM_MESSAGE, &adev->uvd.inst->irq);
@@ -432,15 +434,13 @@ static int uvd_v6_0_sw_init(void *handle)
 		}
 	}
 
-	r = amdgpu_uvd_entity_init(adev);
-
 	return r;
 }
 
-static int uvd_v6_0_sw_fini(void *handle)
+static int uvd_v6_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int i, r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = amdgpu_uvd_suspend(adev);
 	if (r)
@@ -457,19 +457,19 @@ static int uvd_v6_0_sw_fini(void *handle)
 /**
  * uvd_v6_0_hw_init - start and test UVD block
  *
- * @handle: handle used to pass amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Initialize the hardware, boot up the VCPU and do some testing
  */
-static int uvd_v6_0_hw_init(void *handle)
+static int uvd_v6_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring = &adev->uvd.inst->ring;
 	uint32_t tmp;
 	int i, r;
 
 	amdgpu_asic_set_uvd_clocks(adev, 10000, 10000);
-	uvd_v6_0_set_clockgating_state(adev, AMD_CG_STATE_UNGATE);
+	uvd_v6_0_set_clockgating_state(ip_block, AMD_CG_STATE_UNGATE);
 	uvd_v6_0_enable_mgcg(adev, true);
 
 	r = amdgpu_ring_test_helper(ring);
@@ -526,13 +526,13 @@ done:
 /**
  * uvd_v6_0_hw_fini - stop the hardware block
  *
- * @handle: handle used to pass amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Stop the UVD block, mark ring as not ready any more
  */
-static int uvd_v6_0_hw_fini(void *handle)
+static int uvd_v6_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	cancel_delayed_work_sync(&adev->uvd.idle_work);
 
@@ -542,10 +542,17 @@ static int uvd_v6_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int uvd_v6_0_suspend(void *handle)
+static int uvd_v6_0_prepare_suspend(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	return amdgpu_uvd_prepare_suspend(adev);
+}
+
+static int uvd_v6_0_suspend(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/*
 	 * Proper cleanups before halting the HW engine:
@@ -571,23 +578,22 @@ static int uvd_v6_0_suspend(void *handle)
 						       AMD_CG_STATE_GATE);
 	}
 
-	r = uvd_v6_0_hw_fini(adev);
+	r = uvd_v6_0_hw_fini(ip_block);
 	if (r)
 		return r;
 
 	return amdgpu_uvd_suspend(adev);
 }
 
-static int uvd_v6_0_resume(void *handle)
+static int uvd_v6_0_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = amdgpu_uvd_resume(adev);
+	r = amdgpu_uvd_resume(ip_block->adev);
 	if (r)
 		return r;
 
-	return uvd_v6_0_hw_init(adev);
+	return uvd_v6_0_hw_init(ip_block);
 }
 
 /**
@@ -1139,29 +1145,29 @@ static void uvd_v6_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring,
 	amdgpu_ring_write(ring, vmid);
 }
 
-static bool uvd_v6_0_is_idle(void *handle)
+static bool uvd_v6_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return !(RREG32(mmSRBM_STATUS) & SRBM_STATUS__UVD_BUSY_MASK);
 }
 
-static int uvd_v6_0_wait_for_idle(void *handle)
+static int uvd_v6_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		if (uvd_v6_0_is_idle(handle))
+		if (uvd_v6_0_is_idle(ip_block))
 			return 0;
 	}
 	return -ETIMEDOUT;
 }
 
 #define AMDGPU_UVD_STATUS_BUSY_MASK    0xfd
-static bool uvd_v6_0_check_soft_reset(void *handle)
+static bool uvd_v6_0_check_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 srbm_soft_reset = 0;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
@@ -1179,9 +1185,9 @@ static bool uvd_v6_0_check_soft_reset(void *handle)
 	}
 }
 
-static int uvd_v6_0_pre_soft_reset(void *handle)
+static int uvd_v6_0_pre_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (!adev->uvd.inst->srbm_soft_reset)
 		return 0;
@@ -1190,9 +1196,9 @@ static int uvd_v6_0_pre_soft_reset(void *handle)
 	return 0;
 }
 
-static int uvd_v6_0_soft_reset(void *handle)
+static int uvd_v6_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 srbm_soft_reset;
 
 	if (!adev->uvd.inst->srbm_soft_reset)
@@ -1221,9 +1227,9 @@ static int uvd_v6_0_soft_reset(void *handle)
 	return 0;
 }
 
-static int uvd_v6_0_post_soft_reset(void *handle)
+static int uvd_v6_0_post_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (!adev->uvd.inst->srbm_soft_reset)
 		return 0;
@@ -1446,15 +1452,15 @@ static void uvd_v6_0_enable_mgcg(struct amdgpu_device *adev,
 	}
 }
 
-static int uvd_v6_0_set_clockgating_state(void *handle,
+static int uvd_v6_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_CG_STATE_GATE);
 
 	if (enable) {
 		/* wait for STATUS to clear */
-		if (uvd_v6_0_wait_for_idle(handle))
+		if (uvd_v6_0_wait_for_idle(ip_block))
 			return -EBUSY;
 		uvd_v6_0_enable_clock_gating(adev, true);
 		/* enable HW gates because UVD is idle */
@@ -1467,7 +1473,7 @@ static int uvd_v6_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int uvd_v6_0_set_powergating_state(void *handle,
+static int uvd_v6_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	/* This doesn't actually powergate the UVD block.
@@ -1477,7 +1483,7 @@ static int uvd_v6_0_set_powergating_state(void *handle,
 	 * revisit this when there is a cleaner line between
 	 * the smc and the hw blocks
 	 */
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret = 0;
 
 	WREG32(mmUVD_POWER_STATUS, UVD_POWER_STATUS__UVD_PG_EN_MASK);
@@ -1494,9 +1500,9 @@ out:
 	return ret;
 }
 
-static void uvd_v6_0_get_clockgating_state(void *handle, u64 *flags)
+static void uvd_v6_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	mutex_lock(&adev->pm.mutex);
@@ -1523,11 +1529,11 @@ out:
 static const struct amd_ip_funcs uvd_v6_0_ip_funcs = {
 	.name = "uvd_v6_0",
 	.early_init = uvd_v6_0_early_init,
-	.late_init = NULL,
 	.sw_init = uvd_v6_0_sw_init,
 	.sw_fini = uvd_v6_0_sw_fini,
 	.hw_init = uvd_v6_0_hw_init,
 	.hw_fini = uvd_v6_0_hw_fini,
+	.prepare_suspend = uvd_v6_0_prepare_suspend,
 	.suspend = uvd_v6_0_suspend,
 	.resume = uvd_v6_0_resume,
 	.is_idle = uvd_v6_0_is_idle,
diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c
index e32b656b3dab..1f8866f3f63c 100644
--- a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c
@@ -225,7 +225,8 @@ static int uvd_v7_0_enc_get_create_msg(struct amdgpu_ring *ring, u32 handle,
 	int i, r;
 
 	r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, ib_size_dw * 4,
-				     AMDGPU_IB_POOL_DIRECT, &job);
+				     AMDGPU_IB_POOL_DIRECT, &job,
+				     AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		return r;
 
@@ -288,7 +289,8 @@ static int uvd_v7_0_enc_get_destroy_msg(struct amdgpu_ring *ring, u32 handle,
 	int i, r;
 
 	r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, ib_size_dw * 4,
-				     AMDGPU_IB_POOL_DIRECT, &job);
+				     AMDGPU_IB_POOL_DIRECT, &job,
+				     AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		return r;
 
@@ -361,9 +363,9 @@ error:
 	return r;
 }
 
-static int uvd_v7_0_early_init(void *handle)
+static int uvd_v7_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (adev->asic_type == CHIP_VEGA20) {
 		u32 harvest;
@@ -395,12 +397,12 @@ static int uvd_v7_0_early_init(void *handle)
 	return 0;
 }
 
-static int uvd_v7_0_sw_init(void *handle)
+static int uvd_v7_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
 
 	int i, j, r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (j = 0; j < adev->uvd.num_uvd_inst; j++) {
 		if (adev->uvd.harvest_config & (1 << j))
@@ -444,7 +446,7 @@ static int uvd_v7_0_sw_init(void *handle)
 			continue;
 		if (!amdgpu_sriov_vf(adev)) {
 			ring = &adev->uvd.inst[j].ring;
-			ring->vm_hub = AMDGPU_MMHUB_0;
+			ring->vm_hub = AMDGPU_MMHUB0(0);
 			sprintf(ring->name, "uvd_%d", ring->me);
 			r = amdgpu_ring_init(adev, ring, 512,
 					     &adev->uvd.inst[j].irq, 0,
@@ -455,7 +457,7 @@ static int uvd_v7_0_sw_init(void *handle)
 
 		for (i = 0; i < adev->uvd.num_enc_rings; ++i) {
 			ring = &adev->uvd.inst[j].ring_enc[i];
-			ring->vm_hub = AMDGPU_MMHUB_0;
+			ring->vm_hub = AMDGPU_MMHUB0(0);
 			sprintf(ring->name, "uvd_enc_%d.%d", ring->me, i);
 			if (amdgpu_sriov_vf(adev)) {
 				ring->use_doorbell = true;
@@ -480,10 +482,6 @@ static int uvd_v7_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	r = amdgpu_uvd_entity_init(adev);
-	if (r)
-		return r;
-
 	r = amdgpu_virt_alloc_mm_table(adev);
 	if (r)
 		return r;
@@ -491,10 +489,10 @@ static int uvd_v7_0_sw_init(void *handle)
 	return r;
 }
 
-static int uvd_v7_0_sw_fini(void *handle)
+static int uvd_v7_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int i, j, r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_virt_free_mm_table(adev);
 
@@ -514,13 +512,13 @@ static int uvd_v7_0_sw_fini(void *handle)
 /**
  * uvd_v7_0_hw_init - start and test UVD block
  *
- * @handle: handle used to pass amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Initialize the hardware, boot up the VCPU and do some testing
  */
-static int uvd_v7_0_hw_init(void *handle)
+static int uvd_v7_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring;
 	uint32_t tmp;
 	int i, j, r;
@@ -592,13 +590,13 @@ done:
 /**
  * uvd_v7_0_hw_fini - stop the hardware block
  *
- * @handle: handle used to pass amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Stop the UVD block, mark ring as not ready any more
  */
-static int uvd_v7_0_hw_fini(void *handle)
+static int uvd_v7_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	cancel_delayed_work_sync(&adev->uvd.idle_work);
 
@@ -612,10 +610,17 @@ static int uvd_v7_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int uvd_v7_0_suspend(void *handle)
+static int uvd_v7_0_prepare_suspend(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	return amdgpu_uvd_prepare_suspend(adev);
+}
+
+static int uvd_v7_0_suspend(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/*
 	 * Proper cleanups before halting the HW engine:
@@ -641,23 +646,22 @@ static int uvd_v7_0_suspend(void *handle)
 						       AMD_CG_STATE_GATE);
 	}
 
-	r = uvd_v7_0_hw_fini(adev);
+	r = uvd_v7_0_hw_fini(ip_block);
 	if (r)
 		return r;
 
 	return amdgpu_uvd_suspend(adev);
 }
 
-static int uvd_v7_0_resume(void *handle)
+static int uvd_v7_0_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = amdgpu_uvd_resume(adev);
+	r = amdgpu_uvd_resume(ip_block->adev);
 	if (r)
 		return r;
 
-	return uvd_v7_0_hw_init(adev);
+	return uvd_v7_0_hw_init(ip_block);
 }
 
 /**
@@ -679,11 +683,11 @@ static void uvd_v7_0_mc_resume(struct amdgpu_device *adev)
 		if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 			WREG32_SOC15(UVD, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
 				i == 0 ?
-				adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].tmr_mc_addr_lo:
+				adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].tmr_mc_addr_lo :
 				adev->firmware.ucode[AMDGPU_UCODE_ID_UVD1].tmr_mc_addr_lo);
 			WREG32_SOC15(UVD, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
 				i == 0 ?
-				adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].tmr_mc_addr_hi:
+				adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].tmr_mc_addr_hi :
 				adev->firmware.ucode[AMDGPU_UCODE_ID_UVD1].tmr_mc_addr_hi);
 			WREG32_SOC15(UVD, i, mmUVD_VCPU_CACHE_OFFSET0, 0);
 			offset = 0;
@@ -1286,7 +1290,7 @@ static int uvd_v7_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p,
 					   struct amdgpu_job *job,
 					   struct amdgpu_ib *ib)
 {
-	struct amdgpu_ring *ring = to_amdgpu_ring(job->base.sched);
+	struct amdgpu_ring *ring = amdgpu_job_ring(job);
 	unsigned i;
 
 	/* No patching necessary for the first instance */
@@ -1460,104 +1464,6 @@ static void uvd_v7_0_enc_ring_emit_wreg(struct amdgpu_ring *ring,
 	amdgpu_ring_write(ring, val);
 }
 
-#if 0
-static bool uvd_v7_0_is_idle(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return !(RREG32(mmSRBM_STATUS) & SRBM_STATUS__UVD_BUSY_MASK);
-}
-
-static int uvd_v7_0_wait_for_idle(void *handle)
-{
-	unsigned i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	for (i = 0; i < adev->usec_timeout; i++) {
-		if (uvd_v7_0_is_idle(handle))
-			return 0;
-	}
-	return -ETIMEDOUT;
-}
-
-#define AMDGPU_UVD_STATUS_BUSY_MASK    0xfd
-static bool uvd_v7_0_check_soft_reset(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	u32 srbm_soft_reset = 0;
-	u32 tmp = RREG32(mmSRBM_STATUS);
-
-	if (REG_GET_FIELD(tmp, SRBM_STATUS, UVD_RQ_PENDING) ||
-	    REG_GET_FIELD(tmp, SRBM_STATUS, UVD_BUSY) ||
-	    (RREG32_SOC15(UVD, ring->me, mmUVD_STATUS) &
-		    AMDGPU_UVD_STATUS_BUSY_MASK))
-		srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset,
-				SRBM_SOFT_RESET, SOFT_RESET_UVD, 1);
-
-	if (srbm_soft_reset) {
-		adev->uvd.inst[ring->me].srbm_soft_reset = srbm_soft_reset;
-		return true;
-	} else {
-		adev->uvd.inst[ring->me].srbm_soft_reset = 0;
-		return false;
-	}
-}
-
-static int uvd_v7_0_pre_soft_reset(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	if (!adev->uvd.inst[ring->me].srbm_soft_reset)
-		return 0;
-
-	uvd_v7_0_stop(adev);
-	return 0;
-}
-
-static int uvd_v7_0_soft_reset(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	u32 srbm_soft_reset;
-
-	if (!adev->uvd.inst[ring->me].srbm_soft_reset)
-		return 0;
-	srbm_soft_reset = adev->uvd.inst[ring->me].srbm_soft_reset;
-
-	if (srbm_soft_reset) {
-		u32 tmp;
-
-		tmp = RREG32(mmSRBM_SOFT_RESET);
-		tmp |= srbm_soft_reset;
-		dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp);
-		WREG32(mmSRBM_SOFT_RESET, tmp);
-		tmp = RREG32(mmSRBM_SOFT_RESET);
-
-		udelay(50);
-
-		tmp &= ~srbm_soft_reset;
-		WREG32(mmSRBM_SOFT_RESET, tmp);
-		tmp = RREG32(mmSRBM_SOFT_RESET);
-
-		/* Wait a little for things to settle down */
-		udelay(50);
-	}
-
-	return 0;
-}
-
-static int uvd_v7_0_post_soft_reset(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	if (!adev->uvd.inst[ring->me].srbm_soft_reset)
-		return 0;
-
-	mdelay(5);
-
-	return uvd_v7_0_start(adev);
-}
-#endif
-
 static int uvd_v7_0_set_interrupt_state(struct amdgpu_device *adev,
 					struct amdgpu_irq_src *source,
 					unsigned type,
@@ -1607,172 +1513,7 @@ static int uvd_v7_0_process_interrupt(struct amdgpu_device *adev,
 	return 0;
 }
 
-#if 0
-static void uvd_v7_0_set_sw_clock_gating(struct amdgpu_device *adev)
-{
-	uint32_t data, data1, data2, suvd_flags;
-
-	data = RREG32_SOC15(UVD, ring->me, mmUVD_CGC_CTRL);
-	data1 = RREG32_SOC15(UVD, ring->me, mmUVD_SUVD_CGC_GATE);
-	data2 = RREG32_SOC15(UVD, ring->me, mmUVD_SUVD_CGC_CTRL);
-
-	data &= ~(UVD_CGC_CTRL__CLK_OFF_DELAY_MASK |
-		  UVD_CGC_CTRL__CLK_GATE_DLY_TIMER_MASK);
-
-	suvd_flags = UVD_SUVD_CGC_GATE__SRE_MASK |
-		     UVD_SUVD_CGC_GATE__SIT_MASK |
-		     UVD_SUVD_CGC_GATE__SMP_MASK |
-		     UVD_SUVD_CGC_GATE__SCM_MASK |
-		     UVD_SUVD_CGC_GATE__SDB_MASK;
-
-	data |= UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK |
-		(1 << REG_FIELD_SHIFT(UVD_CGC_CTRL, CLK_GATE_DLY_TIMER)) |
-		(4 << REG_FIELD_SHIFT(UVD_CGC_CTRL, CLK_OFF_DELAY));
-
-	data &= ~(UVD_CGC_CTRL__UDEC_RE_MODE_MASK |
-			UVD_CGC_CTRL__UDEC_CM_MODE_MASK |
-			UVD_CGC_CTRL__UDEC_IT_MODE_MASK |
-			UVD_CGC_CTRL__UDEC_DB_MODE_MASK |
-			UVD_CGC_CTRL__UDEC_MP_MODE_MASK |
-			UVD_CGC_CTRL__SYS_MODE_MASK |
-			UVD_CGC_CTRL__UDEC_MODE_MASK |
-			UVD_CGC_CTRL__MPEG2_MODE_MASK |
-			UVD_CGC_CTRL__REGS_MODE_MASK |
-			UVD_CGC_CTRL__RBC_MODE_MASK |
-			UVD_CGC_CTRL__LMI_MC_MODE_MASK |
-			UVD_CGC_CTRL__LMI_UMC_MODE_MASK |
-			UVD_CGC_CTRL__IDCT_MODE_MASK |
-			UVD_CGC_CTRL__MPRD_MODE_MASK |
-			UVD_CGC_CTRL__MPC_MODE_MASK |
-			UVD_CGC_CTRL__LBSI_MODE_MASK |
-			UVD_CGC_CTRL__LRBBM_MODE_MASK |
-			UVD_CGC_CTRL__WCB_MODE_MASK |
-			UVD_CGC_CTRL__VCPU_MODE_MASK |
-			UVD_CGC_CTRL__JPEG_MODE_MASK |
-			UVD_CGC_CTRL__JPEG2_MODE_MASK |
-			UVD_CGC_CTRL__SCPU_MODE_MASK);
-	data2 &= ~(UVD_SUVD_CGC_CTRL__SRE_MODE_MASK |
-			UVD_SUVD_CGC_CTRL__SIT_MODE_MASK |
-			UVD_SUVD_CGC_CTRL__SMP_MODE_MASK |
-			UVD_SUVD_CGC_CTRL__SCM_MODE_MASK |
-			UVD_SUVD_CGC_CTRL__SDB_MODE_MASK);
-	data1 |= suvd_flags;
-
-	WREG32_SOC15(UVD, ring->me, mmUVD_CGC_CTRL, data);
-	WREG32_SOC15(UVD, ring->me, mmUVD_CGC_GATE, 0);
-	WREG32_SOC15(UVD, ring->me, mmUVD_SUVD_CGC_GATE, data1);
-	WREG32_SOC15(UVD, ring->me, mmUVD_SUVD_CGC_CTRL, data2);
-}
-
-static void uvd_v7_0_set_hw_clock_gating(struct amdgpu_device *adev)
-{
-	uint32_t data, data1, cgc_flags, suvd_flags;
-
-	data = RREG32_SOC15(UVD, ring->me, mmUVD_CGC_GATE);
-	data1 = RREG32_SOC15(UVD, ring->me, mmUVD_SUVD_CGC_GATE);
-
-	cgc_flags = UVD_CGC_GATE__SYS_MASK |
-		UVD_CGC_GATE__UDEC_MASK |
-		UVD_CGC_GATE__MPEG2_MASK |
-		UVD_CGC_GATE__RBC_MASK |
-		UVD_CGC_GATE__LMI_MC_MASK |
-		UVD_CGC_GATE__IDCT_MASK |
-		UVD_CGC_GATE__MPRD_MASK |
-		UVD_CGC_GATE__MPC_MASK |
-		UVD_CGC_GATE__LBSI_MASK |
-		UVD_CGC_GATE__LRBBM_MASK |
-		UVD_CGC_GATE__UDEC_RE_MASK |
-		UVD_CGC_GATE__UDEC_CM_MASK |
-		UVD_CGC_GATE__UDEC_IT_MASK |
-		UVD_CGC_GATE__UDEC_DB_MASK |
-		UVD_CGC_GATE__UDEC_MP_MASK |
-		UVD_CGC_GATE__WCB_MASK |
-		UVD_CGC_GATE__VCPU_MASK |
-		UVD_CGC_GATE__SCPU_MASK |
-		UVD_CGC_GATE__JPEG_MASK |
-		UVD_CGC_GATE__JPEG2_MASK;
-
-	suvd_flags = UVD_SUVD_CGC_GATE__SRE_MASK |
-				UVD_SUVD_CGC_GATE__SIT_MASK |
-				UVD_SUVD_CGC_GATE__SMP_MASK |
-				UVD_SUVD_CGC_GATE__SCM_MASK |
-				UVD_SUVD_CGC_GATE__SDB_MASK;
-
-	data |= cgc_flags;
-	data1 |= suvd_flags;
-
-	WREG32_SOC15(UVD, ring->me, mmUVD_CGC_GATE, data);
-	WREG32_SOC15(UVD, ring->me, mmUVD_SUVD_CGC_GATE, data1);
-}
-
-static void uvd_v7_0_set_bypass_mode(struct amdgpu_device *adev, bool enable)
-{
-	u32 tmp = RREG32_SMC(ixGCK_DFS_BYPASS_CNTL);
-
-	if (enable)
-		tmp |= (GCK_DFS_BYPASS_CNTL__BYPASSDCLK_MASK |
-			GCK_DFS_BYPASS_CNTL__BYPASSVCLK_MASK);
-	else
-		tmp &= ~(GCK_DFS_BYPASS_CNTL__BYPASSDCLK_MASK |
-			 GCK_DFS_BYPASS_CNTL__BYPASSVCLK_MASK);
-
-	WREG32_SMC(ixGCK_DFS_BYPASS_CNTL, tmp);
-}
-
-
-static int uvd_v7_0_set_clockgating_state(void *handle,
-					  enum amd_clockgating_state state)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	bool enable = (state == AMD_CG_STATE_GATE);
-
-	uvd_v7_0_set_bypass_mode(adev, enable);
-
-	if (!(adev->cg_flags & AMD_CG_SUPPORT_UVD_MGCG))
-		return 0;
-
-	if (enable) {
-		/* disable HW gating and enable Sw gating */
-		uvd_v7_0_set_sw_clock_gating(adev);
-	} else {
-		/* wait for STATUS to clear */
-		if (uvd_v7_0_wait_for_idle(handle))
-			return -EBUSY;
-
-		/* enable HW gates because UVD is idle */
-		/* uvd_v7_0_set_hw_clock_gating(adev); */
-	}
-
-	return 0;
-}
-
-static int uvd_v7_0_set_powergating_state(void *handle,
-					  enum amd_powergating_state state)
-{
-	/* This doesn't actually powergate the UVD block.
-	 * That's done in the dpm code via the SMC.  This
-	 * just re-inits the block as necessary.  The actual
-	 * gating still happens in the dpm code.  We should
-	 * revisit this when there is a cleaner line between
-	 * the smc and the hw blocks
-	 */
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	if (!(adev->pg_flags & AMD_PG_SUPPORT_UVD))
-		return 0;
-
-	WREG32_SOC15(UVD, ring->me, mmUVD_POWER_STATUS, UVD_POWER_STATUS__UVD_PG_EN_MASK);
-
-	if (state == AMD_PG_STATE_GATE) {
-		uvd_v7_0_stop(adev);
-		return 0;
-	} else {
-		return uvd_v7_0_start(adev);
-	}
-}
-#endif
-
-static int uvd_v7_0_set_clockgating_state(void *handle,
+static int uvd_v7_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	/* needed for driver unload*/
@@ -1782,19 +1523,13 @@ static int uvd_v7_0_set_clockgating_state(void *handle,
 const struct amd_ip_funcs uvd_v7_0_ip_funcs = {
 	.name = "uvd_v7_0",
 	.early_init = uvd_v7_0_early_init,
-	.late_init = NULL,
 	.sw_init = uvd_v7_0_sw_init,
 	.sw_fini = uvd_v7_0_sw_fini,
 	.hw_init = uvd_v7_0_hw_init,
 	.hw_fini = uvd_v7_0_hw_fini,
+	.prepare_suspend = uvd_v7_0_prepare_suspend,
 	.suspend = uvd_v7_0_suspend,
 	.resume = uvd_v7_0_resume,
-	.is_idle = NULL /* uvd_v7_0_is_idle */,
-	.wait_for_idle = NULL /* uvd_v7_0_wait_for_idle */,
-	.check_soft_reset = NULL /* uvd_v7_0_check_soft_reset */,
-	.pre_soft_reset = NULL /* uvd_v7_0_pre_soft_reset */,
-	.soft_reset = NULL /* uvd_v7_0_soft_reset */,
-	.post_soft_reset = NULL /* uvd_v7_0_post_soft_reset */,
 	.set_clockgating_state = uvd_v7_0_set_clockgating_state,
 	.set_powergating_state = NULL /* uvd_v7_0_set_powergating_state */,
 };
@@ -1908,8 +1643,7 @@ static void uvd_v7_0_set_irq_funcs(struct amdgpu_device *adev)
 	}
 }
 
-const struct amdgpu_ip_block_version uvd_v7_0_ip_block =
-{
+const struct amdgpu_ip_block_version uvd_v7_0_ip_block = {
 		.type = AMD_IP_BLOCK_TYPE_UVD,
 		.major = 7,
 		.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v1_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v1_0.c
new file mode 100644
index 000000000000..9ae424618556
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/vce_v1_0.c
@@ -0,0 +1,839 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2013 Advanced Micro Devices, Inc.
+ * Copyright 2025 Valve Corporation
+ * Copyright 2025 Alexandre Demers
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * Authors: Christian König <christian.koenig@amd.com>
+ *          Timur Kristóf <timur.kristof@gmail.com>
+ *          Alexandre Demers <alexandre.f.demers@gmail.com>
+ */
+
+#include <linux/firmware.h>
+
+#include "amdgpu.h"
+#include "amdgpu_vce.h"
+#include "amdgpu_gart.h"
+#include "sid.h"
+#include "vce_v1_0.h"
+#include "vce/vce_1_0_d.h"
+#include "vce/vce_1_0_sh_mask.h"
+#include "oss/oss_1_0_d.h"
+#include "oss/oss_1_0_sh_mask.h"
+
+#define VCE_V1_0_FW_SIZE	(256 * 1024)
+#define VCE_V1_0_STACK_SIZE	(64 * 1024)
+#define VCE_V1_0_DATA_SIZE	(7808 * (AMDGPU_MAX_VCE_HANDLES + 1))
+#define VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK	0x02
+
+#define VCE_V1_0_GART_PAGE_START \
+	(AMDGPU_GTT_MAX_TRANSFER_SIZE * AMDGPU_GTT_NUM_TRANSFER_WINDOWS)
+#define VCE_V1_0_GART_ADDR_START \
+	(VCE_V1_0_GART_PAGE_START * AMDGPU_GPU_PAGE_SIZE)
+
+static void vce_v1_0_set_ring_funcs(struct amdgpu_device *adev);
+static void vce_v1_0_set_irq_funcs(struct amdgpu_device *adev);
+
+struct vce_v1_0_fw_signature {
+	int32_t offset;
+	uint32_t length;
+	int32_t number;
+	struct {
+		uint32_t chip_id;
+		uint32_t keyselect;
+		uint32_t nonce[4];
+		uint32_t sigval[4];
+	} val[8];
+};
+
+/**
+ * vce_v1_0_ring_get_rptr - get read pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware read pointer
+ */
+static uint64_t vce_v1_0_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->me == 0)
+		return RREG32(mmVCE_RB_RPTR);
+	else
+		return RREG32(mmVCE_RB_RPTR2);
+}
+
+/**
+ * vce_v1_0_ring_get_wptr - get write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware write pointer
+ */
+static uint64_t vce_v1_0_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->me == 0)
+		return RREG32(mmVCE_RB_WPTR);
+	else
+		return RREG32(mmVCE_RB_WPTR2);
+}
+
+/**
+ * vce_v1_0_ring_set_wptr - set write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Commits the write pointer to the hardware
+ */
+static void vce_v1_0_ring_set_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->me == 0)
+		WREG32(mmVCE_RB_WPTR, lower_32_bits(ring->wptr));
+	else
+		WREG32(mmVCE_RB_WPTR2, lower_32_bits(ring->wptr));
+}
+
+static int vce_v1_0_lmi_clean(struct amdgpu_device *adev)
+{
+	int i, j;
+
+	for (i = 0; i < 10; ++i) {
+		for (j = 0; j < 100; ++j) {
+			if (RREG32(mmVCE_LMI_STATUS) & 0x337f)
+				return 0;
+
+			mdelay(10);
+		}
+	}
+
+	return -ETIMEDOUT;
+}
+
+static int vce_v1_0_firmware_loaded(struct amdgpu_device *adev)
+{
+	int i, j;
+
+	for (i = 0; i < 10; ++i) {
+		for (j = 0; j < 100; ++j) {
+			if (RREG32(mmVCE_STATUS) & VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK)
+				return 0;
+			mdelay(10);
+		}
+
+		dev_err(adev->dev, "VCE not responding, trying to reset the ECPU\n");
+
+		WREG32_P(mmVCE_SOFT_RESET,
+			VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK,
+			~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+		mdelay(10);
+		WREG32_P(mmVCE_SOFT_RESET, 0,
+			~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+		mdelay(10);
+	}
+
+	return -ETIMEDOUT;
+}
+
+static void vce_v1_0_init_cg(struct amdgpu_device *adev)
+{
+	u32 tmp;
+
+	tmp = RREG32(mmVCE_CLOCK_GATING_A);
+	tmp |= VCE_CLOCK_GATING_A__CGC_DYN_CLOCK_MODE_MASK;
+	WREG32(mmVCE_CLOCK_GATING_A, tmp);
+
+	tmp = RREG32(mmVCE_CLOCK_GATING_B);
+	tmp |= 0x1e;
+	tmp &= ~0xe100e1;
+	WREG32(mmVCE_CLOCK_GATING_B, tmp);
+
+	tmp = RREG32(mmVCE_UENC_CLOCK_GATING);
+	tmp &= ~0xff9ff000;
+	WREG32(mmVCE_UENC_CLOCK_GATING, tmp);
+
+	tmp = RREG32(mmVCE_UENC_REG_CLOCK_GATING);
+	tmp &= ~0x3ff;
+	WREG32(mmVCE_UENC_REG_CLOCK_GATING, tmp);
+}
+
+/**
+ * vce_v1_0_load_fw_signature - load firmware signature into VCPU BO
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * The VCE1 firmware validation mechanism needs a firmware signature.
+ * This function finds the signature appropriate for the current
+ * ASIC and writes that into the VCPU BO.
+ */
+static int vce_v1_0_load_fw_signature(struct amdgpu_device *adev)
+{
+	const struct common_firmware_header *hdr;
+	struct vce_v1_0_fw_signature *sign;
+	unsigned int ucode_offset;
+	uint32_t chip_id;
+	u32 *cpu_addr;
+	int i;
+
+	hdr = (const struct common_firmware_header *)adev->vce.fw->data;
+	ucode_offset = le32_to_cpu(hdr->ucode_array_offset_bytes);
+	cpu_addr = adev->vce.cpu_addr;
+
+	sign = (void *)adev->vce.fw->data + ucode_offset;
+
+	switch (adev->asic_type) {
+	case CHIP_TAHITI:
+		chip_id = 0x01000014;
+		break;
+	case CHIP_VERDE:
+		chip_id = 0x01000015;
+		break;
+	case CHIP_PITCAIRN:
+		chip_id = 0x01000016;
+		break;
+	default:
+		dev_err(adev->dev, "asic_type %#010x was not found!", adev->asic_type);
+		return -EINVAL;
+	}
+
+	for (i = 0; i < le32_to_cpu(sign->number); ++i) {
+		if (le32_to_cpu(sign->val[i].chip_id) == chip_id)
+			break;
+	}
+
+	if (i == le32_to_cpu(sign->number)) {
+		dev_err(adev->dev, "chip_id 0x%x for %s was not found in VCE firmware",
+			chip_id, amdgpu_asic_name[adev->asic_type]);
+		return -EINVAL;
+	}
+
+	cpu_addr += (256 - 64) / 4;
+	memcpy_toio(&cpu_addr[0], &sign->val[i].nonce[0], 16);
+	cpu_addr[4] = cpu_to_le32(le32_to_cpu(sign->length) + 64);
+
+	memset_io(&cpu_addr[5], 0, 44);
+	memcpy_toio(&cpu_addr[16], &sign[1], hdr->ucode_size_bytes - sizeof(*sign));
+
+	cpu_addr += (le32_to_cpu(sign->length) + 64) / 4;
+	memcpy_toio(&cpu_addr[0], &sign->val[i].sigval[0], 16);
+
+	adev->vce.keyselect = le32_to_cpu(sign->val[i].keyselect);
+
+	return 0;
+}
+
+static int vce_v1_0_wait_for_fw_validation(struct amdgpu_device *adev)
+{
+	int i;
+
+	dev_dbg(adev->dev, "VCE keyselect: %d", adev->vce.keyselect);
+	WREG32(mmVCE_LMI_FW_START_KEYSEL, adev->vce.keyselect);
+
+	for (i = 0; i < 10; ++i) {
+		mdelay(10);
+		if (RREG32(mmVCE_FW_REG_STATUS) & VCE_FW_REG_STATUS__DONE_MASK)
+			break;
+	}
+
+	if (!(RREG32(mmVCE_FW_REG_STATUS) & VCE_FW_REG_STATUS__DONE_MASK)) {
+		dev_err(adev->dev, "VCE FW validation timeout\n");
+		return -ETIMEDOUT;
+	}
+
+	if (!(RREG32(mmVCE_FW_REG_STATUS) & VCE_FW_REG_STATUS__PASS_MASK)) {
+		dev_err(adev->dev, "VCE FW validation failed\n");
+		return -EINVAL;
+	}
+
+	for (i = 0; i < 10; ++i) {
+		mdelay(10);
+		if (!(RREG32(mmVCE_FW_REG_STATUS) & VCE_FW_REG_STATUS__BUSY_MASK))
+			break;
+	}
+
+	if (RREG32(mmVCE_FW_REG_STATUS) & VCE_FW_REG_STATUS__BUSY_MASK) {
+		dev_err(adev->dev, "VCE FW busy timeout\n");
+		return -ETIMEDOUT;
+	}
+
+	return 0;
+}
+
+static int vce_v1_0_mc_resume(struct amdgpu_device *adev)
+{
+	uint32_t offset;
+	uint32_t size;
+
+	/*
+	 * When the keyselect is already set, don't perturb VCE FW.
+	 * Validation seems to always fail the second time.
+	 */
+	if (RREG32(mmVCE_LMI_FW_START_KEYSEL)) {
+		dev_dbg(adev->dev, "keyselect already set: 0x%x (on CPU: 0x%x)\n",
+			RREG32(mmVCE_LMI_FW_START_KEYSEL), adev->vce.keyselect);
+
+		WREG32_P(mmVCE_LMI_CTRL2, 0x0, ~0x100);
+		return 0;
+	}
+
+	WREG32_P(mmVCE_CLOCK_GATING_A, 0, ~(1 << 16));
+	WREG32_P(mmVCE_UENC_CLOCK_GATING, 0x1FF000, ~0xFF9FF000);
+	WREG32_P(mmVCE_UENC_REG_CLOCK_GATING, 0x3F, ~0x3F);
+	WREG32(mmVCE_CLOCK_GATING_B, 0);
+
+	WREG32_P(mmVCE_LMI_FW_PERIODIC_CTRL, 0x4, ~0x4);
+
+	WREG32(mmVCE_LMI_CTRL, 0x00398000);
+
+	WREG32_P(mmVCE_LMI_CACHE_CTRL, 0x0, ~0x1);
+	WREG32(mmVCE_LMI_SWAP_CNTL, 0);
+	WREG32(mmVCE_LMI_SWAP_CNTL1, 0);
+	WREG32(mmVCE_LMI_VM_CTRL, 0);
+
+	WREG32(mmVCE_VCPU_SCRATCH7, AMDGPU_MAX_VCE_HANDLES);
+
+	offset =  adev->vce.gpu_addr + AMDGPU_VCE_FIRMWARE_OFFSET;
+	size = VCE_V1_0_FW_SIZE;
+	WREG32(mmVCE_VCPU_CACHE_OFFSET0, offset & 0x7fffffff);
+	WREG32(mmVCE_VCPU_CACHE_SIZE0, size);
+
+	offset += size;
+	size = VCE_V1_0_STACK_SIZE;
+	WREG32(mmVCE_VCPU_CACHE_OFFSET1, offset & 0x7fffffff);
+	WREG32(mmVCE_VCPU_CACHE_SIZE1, size);
+
+	offset += size;
+	size = VCE_V1_0_DATA_SIZE;
+	WREG32(mmVCE_VCPU_CACHE_OFFSET2, offset & 0x7fffffff);
+	WREG32(mmVCE_VCPU_CACHE_SIZE2, size);
+
+	WREG32_P(mmVCE_LMI_CTRL2, 0x0, ~0x100);
+
+	return vce_v1_0_wait_for_fw_validation(adev);
+}
+
+/**
+ * vce_v1_0_is_idle() - Check idle status of VCE1 IP block
+ *
+ * @ip_block: amdgpu_ip_block pointer
+ *
+ * Check whether VCE is busy according to VCE_STATUS.
+ * Also check whether the SRBM thinks VCE is busy, although
+ * SRBM_STATUS.VCE_BUSY seems to be bogus because it
+ * appears to mirror the VCE_STATUS.VCPU_REPORT_FW_LOADED bit.
+ */
+static bool vce_v1_0_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	bool busy =
+		(RREG32(mmVCE_STATUS) & (VCE_STATUS__JOB_BUSY_MASK | VCE_STATUS__UENC_BUSY_MASK)) ||
+		(RREG32(mmSRBM_STATUS2) & SRBM_STATUS2__VCE_BUSY_MASK);
+
+	return !busy;
+}
+
+static int vce_v1_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	unsigned int i;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		udelay(1);
+		if (vce_v1_0_is_idle(ip_block))
+			return 0;
+	}
+	return -ETIMEDOUT;
+}
+
+/**
+ * vce_v1_0_start - start VCE block
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Setup and start the VCE block
+ */
+static int vce_v1_0_start(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	int r;
+
+	WREG32_P(mmVCE_STATUS, 1, ~1);
+
+	r = vce_v1_0_mc_resume(adev);
+	if (r)
+		return r;
+
+	ring = &adev->vce.ring[0];
+	WREG32(mmVCE_RB_RPTR, lower_32_bits(ring->wptr));
+	WREG32(mmVCE_RB_WPTR, lower_32_bits(ring->wptr));
+	WREG32(mmVCE_RB_BASE_LO, lower_32_bits(ring->gpu_addr));
+	WREG32(mmVCE_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
+	WREG32(mmVCE_RB_SIZE, ring->ring_size / 4);
+
+	ring = &adev->vce.ring[1];
+	WREG32(mmVCE_RB_RPTR2, lower_32_bits(ring->wptr));
+	WREG32(mmVCE_RB_WPTR2, lower_32_bits(ring->wptr));
+	WREG32(mmVCE_RB_BASE_LO2, lower_32_bits(ring->gpu_addr));
+	WREG32(mmVCE_RB_BASE_HI2, upper_32_bits(ring->gpu_addr));
+	WREG32(mmVCE_RB_SIZE2, ring->ring_size / 4);
+
+	WREG32_P(mmVCE_VCPU_CNTL, VCE_VCPU_CNTL__CLK_EN_MASK,
+		 ~VCE_VCPU_CNTL__CLK_EN_MASK);
+
+	WREG32_P(mmVCE_SOFT_RESET,
+		VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK |
+		VCE_SOFT_RESET__FME_SOFT_RESET_MASK,
+		~(VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK |
+		  VCE_SOFT_RESET__FME_SOFT_RESET_MASK));
+
+	mdelay(100);
+
+	WREG32_P(mmVCE_SOFT_RESET, 0,
+		~(VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK |
+		  VCE_SOFT_RESET__FME_SOFT_RESET_MASK));
+
+	r = vce_v1_0_firmware_loaded(adev);
+
+	/* Clear VCE_STATUS, otherwise SRBM thinks VCE1 is busy. */
+	WREG32(mmVCE_STATUS, 0);
+
+	if (r) {
+		dev_err(adev->dev, "VCE not responding, giving up\n");
+		return r;
+	}
+
+	return 0;
+}
+
+static int vce_v1_0_stop(struct amdgpu_device *adev)
+{
+	struct amdgpu_ip_block *ip_block;
+	int status;
+	int i;
+
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_VCE);
+	if (!ip_block)
+		return -EINVAL;
+
+	if (vce_v1_0_lmi_clean(adev))
+		dev_warn(adev->dev, "VCE not idle\n");
+
+	if (vce_v1_0_wait_for_idle(ip_block))
+		dev_warn(adev->dev, "VCE busy: VCE_STATUS=0x%x, SRBM_STATUS2=0x%x\n",
+			RREG32(mmVCE_STATUS), RREG32(mmSRBM_STATUS2));
+
+	/* Stall UMC and register bus before resetting VCPU */
+	WREG32_P(mmVCE_LMI_CTRL2, 1 << 8, ~(1 << 8));
+
+	for (i = 0; i < 100; ++i) {
+		status = RREG32(mmVCE_LMI_STATUS);
+		if (status & 0x240)
+			break;
+		mdelay(1);
+	}
+
+	WREG32_P(mmVCE_VCPU_CNTL, 0, ~VCE_VCPU_CNTL__CLK_EN_MASK);
+
+	WREG32_P(mmVCE_SOFT_RESET,
+		VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK |
+		VCE_SOFT_RESET__FME_SOFT_RESET_MASK,
+		~(VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK |
+		  VCE_SOFT_RESET__FME_SOFT_RESET_MASK));
+
+	WREG32(mmVCE_STATUS, 0);
+
+	return 0;
+}
+
+static void vce_v1_0_enable_mgcg(struct amdgpu_device *adev, bool enable)
+{
+	u32 tmp;
+
+	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_VCE_MGCG)) {
+		tmp = RREG32(mmVCE_CLOCK_GATING_A);
+		tmp |= VCE_CLOCK_GATING_A__CGC_DYN_CLOCK_MODE_MASK;
+		WREG32(mmVCE_CLOCK_GATING_A, tmp);
+
+		tmp = RREG32(mmVCE_UENC_CLOCK_GATING);
+		tmp &= ~0x1ff000;
+		tmp |= 0xff800000;
+		WREG32(mmVCE_UENC_CLOCK_GATING, tmp);
+
+		tmp = RREG32(mmVCE_UENC_REG_CLOCK_GATING);
+		tmp &= ~0x3ff;
+		WREG32(mmVCE_UENC_REG_CLOCK_GATING, tmp);
+	} else {
+		tmp = RREG32(mmVCE_CLOCK_GATING_A);
+		tmp &= ~VCE_CLOCK_GATING_A__CGC_DYN_CLOCK_MODE_MASK;
+		WREG32(mmVCE_CLOCK_GATING_A, tmp);
+
+		tmp = RREG32(mmVCE_UENC_CLOCK_GATING);
+		tmp |= 0x1ff000;
+		tmp &= ~0xff800000;
+		WREG32(mmVCE_UENC_CLOCK_GATING, tmp);
+
+		tmp = RREG32(mmVCE_UENC_REG_CLOCK_GATING);
+		tmp |= 0x3ff;
+		WREG32(mmVCE_UENC_REG_CLOCK_GATING, tmp);
+	}
+}
+
+static int vce_v1_0_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = amdgpu_vce_early_init(adev);
+	if (r)
+		return r;
+
+	adev->vce.num_rings = 2;
+
+	vce_v1_0_set_ring_funcs(adev);
+	vce_v1_0_set_irq_funcs(adev);
+
+	return 0;
+}
+
+/**
+ * vce_v1_0_ensure_vcpu_bo_32bit_addr() - ensure the VCPU BO has a 32-bit address
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Due to various hardware limitations, the VCE1 requires
+ * the VCPU BO to be in the low 32 bit address range.
+ * Ensure that the VCPU BO has a 32-bit GPU address,
+ * or return an error code when that isn't possible.
+ *
+ * To accomodate that, we put GART to the LOW address range
+ * and reserve some GART pages where we map the VCPU BO,
+ * so that it gets a 32-bit address.
+ */
+static int vce_v1_0_ensure_vcpu_bo_32bit_addr(struct amdgpu_device *adev)
+{
+	u64 gpu_addr = amdgpu_bo_gpu_offset(adev->vce.vcpu_bo);
+	u64 bo_size = amdgpu_bo_size(adev->vce.vcpu_bo);
+	u64 max_vcpu_bo_addr = 0xffffffff - bo_size;
+	u64 num_pages = ALIGN(bo_size, AMDGPU_GPU_PAGE_SIZE) / AMDGPU_GPU_PAGE_SIZE;
+	u64 pa = amdgpu_gmc_vram_pa(adev, adev->vce.vcpu_bo);
+	u64 flags = AMDGPU_PTE_READABLE | AMDGPU_PTE_WRITEABLE | AMDGPU_PTE_VALID;
+
+	/*
+	 * Check if the VCPU BO already has a 32-bit address.
+	 * Eg. if MC is configured to put VRAM in the low address range.
+	 */
+	if (gpu_addr <= max_vcpu_bo_addr)
+		return 0;
+
+	/* Check if we can map the VCPU BO in GART to a 32-bit address. */
+	if (adev->gmc.gart_start + VCE_V1_0_GART_ADDR_START > max_vcpu_bo_addr)
+		return -EINVAL;
+
+	amdgpu_gart_map_vram_range(adev, pa, VCE_V1_0_GART_PAGE_START,
+				   num_pages, flags, adev->gart.ptr);
+	adev->vce.gpu_addr = adev->gmc.gart_start + VCE_V1_0_GART_ADDR_START;
+	if (adev->vce.gpu_addr > max_vcpu_bo_addr)
+		return -EINVAL;
+
+	return 0;
+}
+
+static int vce_v1_0_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring;
+	int r, i;
+
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 167, &adev->vce.irq);
+	if (r)
+		return r;
+
+	r = amdgpu_vce_sw_init(adev, VCE_V1_0_FW_SIZE +
+		VCE_V1_0_STACK_SIZE + VCE_V1_0_DATA_SIZE);
+	if (r)
+		return r;
+
+	r = amdgpu_vce_resume(adev);
+	if (r)
+		return r;
+	r = vce_v1_0_load_fw_signature(adev);
+	if (r)
+		return r;
+	r = vce_v1_0_ensure_vcpu_bo_32bit_addr(adev);
+	if (r)
+		return r;
+
+	for (i = 0; i < adev->vce.num_rings; i++) {
+		enum amdgpu_ring_priority_level hw_prio = amdgpu_vce_get_ring_prio(i);
+
+		ring = &adev->vce.ring[i];
+		sprintf(ring->name, "vce%d", i);
+		r = amdgpu_ring_init(adev, ring, 512, &adev->vce.irq, 0,
+				     hw_prio, NULL);
+		if (r)
+			return r;
+	}
+
+	return r;
+}
+
+static int vce_v1_0_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = amdgpu_vce_suspend(adev);
+	if (r)
+		return r;
+
+	return amdgpu_vce_sw_fini(adev);
+}
+
+/**
+ * vce_v1_0_hw_init - start and test VCE block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Initialize the hardware, boot up the VCPU and do some testing
+ */
+static int vce_v1_0_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, r;
+
+	if (adev->pm.dpm_enabled)
+		amdgpu_dpm_enable_vce(adev, true);
+	else
+		amdgpu_asic_set_vce_clocks(adev, 10000, 10000);
+
+	for (i = 0; i < adev->vce.num_rings; i++) {
+		r = amdgpu_ring_test_helper(&adev->vce.ring[i]);
+		if (r)
+			return r;
+	}
+
+	dev_info(adev->dev, "VCE initialized successfully.\n");
+
+	return 0;
+}
+
+static int vce_v1_0_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+
+	r = vce_v1_0_stop(ip_block->adev);
+	if (r)
+		return r;
+
+	cancel_delayed_work_sync(&ip_block->adev->vce.idle_work);
+	return 0;
+}
+
+static int vce_v1_0_suspend(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	/*
+	 * Proper cleanups before halting the HW engine:
+	 *   - cancel the delayed idle work
+	 *   - enable powergating
+	 *   - enable clockgating
+	 *   - disable dpm
+	 *
+	 * TODO: to align with the VCN implementation, move the
+	 * jobs for clockgating/powergating/dpm setting to
+	 * ->set_powergating_state().
+	 */
+	cancel_delayed_work_sync(&adev->vce.idle_work);
+
+	if (adev->pm.dpm_enabled) {
+		amdgpu_dpm_enable_vce(adev, false);
+	} else {
+		amdgpu_asic_set_vce_clocks(adev, 0, 0);
+		amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCE,
+						       AMD_PG_STATE_GATE);
+		amdgpu_device_ip_set_clockgating_state(adev, AMD_IP_BLOCK_TYPE_VCE,
+						       AMD_CG_STATE_GATE);
+	}
+
+	r = vce_v1_0_hw_fini(ip_block);
+	if (r) {
+		dev_err(adev->dev, "vce_v1_0_hw_fini() failed with error %i", r);
+		return r;
+	}
+
+	return amdgpu_vce_suspend(adev);
+}
+
+static int vce_v1_0_resume(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = amdgpu_vce_resume(adev);
+	if (r)
+		return r;
+	r = vce_v1_0_load_fw_signature(adev);
+	if (r)
+		return r;
+	r = vce_v1_0_ensure_vcpu_bo_32bit_addr(adev);
+	if (r)
+		return r;
+
+	return vce_v1_0_hw_init(ip_block);
+}
+
+static int vce_v1_0_set_interrupt_state(struct amdgpu_device *adev,
+					struct amdgpu_irq_src *source,
+					unsigned int type,
+					enum amdgpu_interrupt_state state)
+{
+	uint32_t val = 0;
+
+	if (state == AMDGPU_IRQ_STATE_ENABLE)
+		val |= VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK;
+
+	WREG32_P(mmVCE_SYS_INT_EN, val,
+		 ~VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK);
+	return 0;
+}
+
+static int vce_v1_0_process_interrupt(struct amdgpu_device *adev,
+				      struct amdgpu_irq_src *source,
+				      struct amdgpu_iv_entry *entry)
+{
+	dev_dbg(adev->dev, "IH: VCE\n");
+	switch (entry->src_data[0]) {
+	case 0:
+	case 1:
+		amdgpu_fence_process(&adev->vce.ring[entry->src_data[0]]);
+		break;
+	default:
+		dev_err(adev->dev, "Unhandled interrupt: %d %d\n",
+			  entry->src_id, entry->src_data[0]);
+		break;
+	}
+
+	return 0;
+}
+
+static int vce_v1_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_clockgating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	vce_v1_0_init_cg(adev);
+	vce_v1_0_enable_mgcg(adev, state == AMD_CG_STATE_GATE);
+
+	return 0;
+}
+
+static int vce_v1_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_powergating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	/*
+	 * This doesn't actually powergate the VCE block.
+	 * That's done in the dpm code via the SMC.  This
+	 * just re-inits the block as necessary.  The actual
+	 * gating still happens in the dpm code.  We should
+	 * revisit this when there is a cleaner line between
+	 * the smc and the hw blocks
+	 */
+	if (state == AMD_PG_STATE_GATE)
+		return vce_v1_0_stop(adev);
+	else
+		return vce_v1_0_start(adev);
+}
+
+static const struct amd_ip_funcs vce_v1_0_ip_funcs = {
+	.name = "vce_v1_0",
+	.early_init = vce_v1_0_early_init,
+	.sw_init = vce_v1_0_sw_init,
+	.sw_fini = vce_v1_0_sw_fini,
+	.hw_init = vce_v1_0_hw_init,
+	.hw_fini = vce_v1_0_hw_fini,
+	.suspend = vce_v1_0_suspend,
+	.resume = vce_v1_0_resume,
+	.is_idle = vce_v1_0_is_idle,
+	.wait_for_idle = vce_v1_0_wait_for_idle,
+	.set_clockgating_state = vce_v1_0_set_clockgating_state,
+	.set_powergating_state = vce_v1_0_set_powergating_state,
+};
+
+static const struct amdgpu_ring_funcs vce_v1_0_ring_funcs = {
+	.type = AMDGPU_RING_TYPE_VCE,
+	.align_mask = 0xf,
+	.nop = VCE_CMD_NO_OP,
+	.support_64bit_ptrs = false,
+	.no_user_fence = true,
+	.get_rptr = vce_v1_0_ring_get_rptr,
+	.get_wptr = vce_v1_0_ring_get_wptr,
+	.set_wptr = vce_v1_0_ring_set_wptr,
+	.parse_cs = amdgpu_vce_ring_parse_cs,
+	.emit_frame_size = 6, /* amdgpu_vce_ring_emit_fence  x1 no user fence */
+	.emit_ib_size = 4, /* amdgpu_vce_ring_emit_ib */
+	.emit_ib = amdgpu_vce_ring_emit_ib,
+	.emit_fence = amdgpu_vce_ring_emit_fence,
+	.test_ring = amdgpu_vce_ring_test_ring,
+	.test_ib = amdgpu_vce_ring_test_ib,
+	.insert_nop = amdgpu_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.begin_use = amdgpu_vce_ring_begin_use,
+	.end_use = amdgpu_vce_ring_end_use,
+};
+
+static void vce_v1_0_set_ring_funcs(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->vce.num_rings; i++) {
+		adev->vce.ring[i].funcs = &vce_v1_0_ring_funcs;
+		adev->vce.ring[i].me = i;
+	}
+};
+
+static const struct amdgpu_irq_src_funcs vce_v1_0_irq_funcs = {
+	.set = vce_v1_0_set_interrupt_state,
+	.process = vce_v1_0_process_interrupt,
+};
+
+static void vce_v1_0_set_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->vce.irq.num_types = 1;
+	adev->vce.irq.funcs = &vce_v1_0_irq_funcs;
+};
+
+const struct amdgpu_ip_block_version vce_v1_0_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_VCE,
+	.major = 1,
+	.minor = 0,
+	.rev = 0,
+	.funcs = &vce_v1_0_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v1_0.h b/drivers/gpu/drm/amd/amdgpu/vce_v1_0.h
new file mode 100644
index 000000000000..206e7bec897f
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/vce_v1_0.h
@@ -0,0 +1,32 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ * Copyright 2025 Valve Corporation
+ * Copyright 2025 Alexandre Demers
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __VCE_V1_0_H__
+#define __VCE_V1_0_H__
+
+extern const struct amdgpu_ip_block_version vce_v1_0_ip_block;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c
index 67eb01fef789..8ea8a6193492 100644
--- a/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c
@@ -201,20 +201,20 @@ static void vce_v2_0_mc_resume(struct amdgpu_device *adev)
 	WREG32_FIELD(VCE_SYS_INT_EN, VCE_SYS_INT_TRAP_INTERRUPT_EN, 1);
 }
 
-static bool vce_v2_0_is_idle(void *handle)
+static bool vce_v2_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return !(RREG32(mmSRBM_STATUS2) & SRBM_STATUS2__VCE_BUSY_MASK);
 }
 
-static int vce_v2_0_wait_for_idle(void *handle)
+static int vce_v2_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	unsigned i;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		if (vce_v2_0_is_idle(handle))
+		if (vce_v2_0_is_idle(ip_block))
 			return 0;
 	}
 	return -ETIMEDOUT;
@@ -274,15 +274,21 @@ static int vce_v2_0_start(struct amdgpu_device *adev)
 
 static int vce_v2_0_stop(struct amdgpu_device *adev)
 {
+	struct amdgpu_ip_block *ip_block;
 	int i;
 	int status;
 
+
 	if (vce_v2_0_lmi_clean(adev)) {
-		DRM_INFO("vce is not idle \n");
+		DRM_INFO("VCE is not idle \n");
 		return 0;
 	}
 
-	if (vce_v2_0_wait_for_idle(adev)) {
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_VCE);
+	if (!ip_block)
+		return -EINVAL;
+
+	if (vce_v2_0_wait_for_idle(ip_block)) {
 		DRM_INFO("VCE is busy, Can't set clock gating");
 		return 0;
 	}
@@ -398,9 +404,14 @@ static void vce_v2_0_enable_mgcg(struct amdgpu_device *adev, bool enable,
 	}
 }
 
-static int vce_v2_0_early_init(void *handle)
+static int vce_v2_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = amdgpu_vce_early_init(adev);
+	if (r)
+		return r;
 
 	adev->vce.num_rings = 2;
 
@@ -410,11 +421,11 @@ static int vce_v2_0_early_init(void *handle)
 	return 0;
 }
 
-static int vce_v2_0_sw_init(void *handle)
+static int vce_v2_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
 	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* VCE */
 	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 167, &adev->vce.irq);
@@ -441,15 +452,13 @@ static int vce_v2_0_sw_init(void *handle)
 			return r;
 	}
 
-	r = amdgpu_vce_entity_init(adev);
-
 	return r;
 }
 
-static int vce_v2_0_sw_fini(void *handle)
+static int vce_v2_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = amdgpu_vce_suspend(adev);
 	if (r)
@@ -458,10 +467,10 @@ static int vce_v2_0_sw_fini(void *handle)
 	return amdgpu_vce_sw_fini(adev);
 }
 
-static int vce_v2_0_hw_init(void *handle)
+static int vce_v2_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_asic_set_vce_clocks(adev, 10000, 10000);
 	vce_v2_0_enable_mgcg(adev, true, false);
@@ -477,19 +486,17 @@ static int vce_v2_0_hw_init(void *handle)
 	return 0;
 }
 
-static int vce_v2_0_hw_fini(void *handle)
+static int vce_v2_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	cancel_delayed_work_sync(&adev->vce.idle_work);
+	cancel_delayed_work_sync(&ip_block->adev->vce.idle_work);
 
 	return 0;
 }
 
-static int vce_v2_0_suspend(void *handle)
+static int vce_v2_0_suspend(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 
 	/*
@@ -515,28 +522,27 @@ static int vce_v2_0_suspend(void *handle)
 						       AMD_CG_STATE_GATE);
 	}
 
-	r = vce_v2_0_hw_fini(adev);
+	r = vce_v2_0_hw_fini(ip_block);
 	if (r)
 		return r;
 
 	return amdgpu_vce_suspend(adev);
 }
 
-static int vce_v2_0_resume(void *handle)
+static int vce_v2_0_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = amdgpu_vce_resume(adev);
+	r = amdgpu_vce_resume(ip_block->adev);
 	if (r)
 		return r;
 
-	return vce_v2_0_hw_init(adev);
+	return vce_v2_0_hw_init(ip_block);
 }
 
-static int vce_v2_0_soft_reset(void *handle)
+static int vce_v2_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	WREG32_FIELD(SRBM_SOFT_RESET, SOFT_RESET_VCE, 1);
 	mdelay(5);
@@ -577,13 +583,13 @@ static int vce_v2_0_process_interrupt(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int vce_v2_0_set_clockgating_state(void *handle,
+static int vce_v2_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	bool gate = false;
 	bool sw_cg = false;
 
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (state == AMD_CG_STATE_GATE) {
 		gate = true;
@@ -595,7 +601,7 @@ static int vce_v2_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int vce_v2_0_set_powergating_state(void *handle,
+static int vce_v2_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	/* This doesn't actually powergate the VCE block.
@@ -605,7 +611,7 @@ static int vce_v2_0_set_powergating_state(void *handle,
 	 * revisit this when there is a cleaner line between
 	 * the smc and the hw blocks
 	 */
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (state == AMD_PG_STATE_GATE)
 		return vce_v2_0_stop(adev);
@@ -616,7 +622,6 @@ static int vce_v2_0_set_powergating_state(void *handle,
 static const struct amd_ip_funcs vce_v2_0_ip_funcs = {
 	.name = "vce_v2_0",
 	.early_init = vce_v2_0_early_init,
-	.late_init = NULL,
 	.sw_init = vce_v2_0_sw_init,
 	.sw_fini = vce_v2_0_sw_fini,
 	.hw_init = vce_v2_0_hw_init,
diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c
index 8def62c83ffd..719e9643c43d 100644
--- a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c
@@ -64,8 +64,8 @@
 static void vce_v3_0_mc_resume(struct amdgpu_device *adev, int idx);
 static void vce_v3_0_set_ring_funcs(struct amdgpu_device *adev);
 static void vce_v3_0_set_irq_funcs(struct amdgpu_device *adev);
-static int vce_v3_0_wait_for_idle(void *handle);
-static int vce_v3_0_set_clockgating_state(void *handle,
+static int vce_v3_0_wait_for_idle(struct amdgpu_ip_block *ip_block);
+static int vce_v3_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state);
 /**
  * vce_v3_0_ring_get_rptr - get read pointer
@@ -396,9 +396,10 @@ static unsigned vce_v3_0_get_harvest_config(struct amdgpu_device *adev)
 	}
 }
 
-static int vce_v3_0_early_init(void *handle)
+static int vce_v3_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
 
 	adev->vce.harvest_config = vce_v3_0_get_harvest_config(adev);
 
@@ -407,6 +408,10 @@ static int vce_v3_0_early_init(void *handle)
 	    (AMDGPU_VCE_HARVEST_VCE0 | AMDGPU_VCE_HARVEST_VCE1))
 		return -ENOENT;
 
+	r = amdgpu_vce_early_init(adev);
+	if (r)
+		return r;
+
 	adev->vce.num_rings = 3;
 
 	vce_v3_0_set_ring_funcs(adev);
@@ -415,9 +420,9 @@ static int vce_v3_0_early_init(void *handle)
 	return 0;
 }
 
-static int vce_v3_0_sw_init(void *handle)
+static int vce_v3_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring;
 	int r, i;
 
@@ -450,15 +455,13 @@ static int vce_v3_0_sw_init(void *handle)
 			return r;
 	}
 
-	r = amdgpu_vce_entity_init(adev);
-
 	return r;
 }
 
-static int vce_v3_0_sw_fini(void *handle)
+static int vce_v3_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = amdgpu_vce_suspend(adev);
 	if (r)
@@ -467,10 +470,10 @@ static int vce_v3_0_sw_fini(void *handle)
 	return amdgpu_vce_sw_fini(adev);
 }
 
-static int vce_v3_0_hw_init(void *handle)
+static int vce_v3_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	vce_v3_0_override_vce_clock_gating(adev, true);
 
@@ -487,25 +490,25 @@ static int vce_v3_0_hw_init(void *handle)
 	return 0;
 }
 
-static int vce_v3_0_hw_fini(void *handle)
+static int vce_v3_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	cancel_delayed_work_sync(&adev->vce.idle_work);
 
-	r = vce_v3_0_wait_for_idle(handle);
+	r = vce_v3_0_wait_for_idle(ip_block);
 	if (r)
 		return r;
 
 	vce_v3_0_stop(adev);
-	return vce_v3_0_set_clockgating_state(adev, AMD_CG_STATE_GATE);
+	return vce_v3_0_set_clockgating_state(ip_block, AMD_CG_STATE_GATE);
 }
 
-static int vce_v3_0_suspend(void *handle)
+static int vce_v3_0_suspend(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/*
 	 * Proper cleanups before halting the HW engine:
@@ -530,23 +533,22 @@ static int vce_v3_0_suspend(void *handle)
 						       AMD_CG_STATE_GATE);
 	}
 
-	r = vce_v3_0_hw_fini(adev);
+	r = vce_v3_0_hw_fini(ip_block);
 	if (r)
 		return r;
 
 	return amdgpu_vce_suspend(adev);
 }
 
-static int vce_v3_0_resume(void *handle)
+static int vce_v3_0_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = amdgpu_vce_resume(adev);
+	r = amdgpu_vce_resume(ip_block->adev);
 	if (r)
 		return r;
 
-	return vce_v3_0_hw_init(adev);
+	return vce_v3_0_hw_init(ip_block);
 }
 
 static void vce_v3_0_mc_resume(struct amdgpu_device *adev, int idx)
@@ -600,9 +602,9 @@ static void vce_v3_0_mc_resume(struct amdgpu_device *adev, int idx)
 	WREG32_FIELD(VCE_SYS_INT_EN, VCE_SYS_INT_TRAP_INTERRUPT_EN, 1);
 }
 
-static bool vce_v3_0_is_idle(void *handle)
+static bool vce_v3_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 mask = 0;
 
 	mask |= (adev->vce.harvest_config & AMDGPU_VCE_HARVEST_VCE0) ? 0 : SRBM_STATUS2__VCE0_BUSY_MASK;
@@ -611,13 +613,13 @@ static bool vce_v3_0_is_idle(void *handle)
 	return !(RREG32(mmSRBM_STATUS2) & mask);
 }
 
-static int vce_v3_0_wait_for_idle(void *handle)
+static int vce_v3_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++)
-		if (vce_v3_0_is_idle(handle))
+		if (vce_v3_0_is_idle(ip_block))
 			return 0;
 
 	return -ETIMEDOUT;
@@ -629,9 +631,9 @@ static int vce_v3_0_wait_for_idle(void *handle)
 #define  AMDGPU_VCE_STATUS_BUSY_MASK (VCE_STATUS_VCPU_REPORT_AUTO_BUSY_MASK | \
 				      VCE_STATUS_VCPU_REPORT_RB0_BUSY_MASK)
 
-static bool vce_v3_0_check_soft_reset(void *handle)
+static bool vce_v3_0_check_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 srbm_soft_reset = 0;
 
 	/* According to VCE team , we should use VCE_STATUS instead
@@ -670,9 +672,9 @@ static bool vce_v3_0_check_soft_reset(void *handle)
 	}
 }
 
-static int vce_v3_0_soft_reset(void *handle)
+static int vce_v3_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 srbm_soft_reset;
 
 	if (!adev->vce.srbm_soft_reset)
@@ -701,29 +703,29 @@ static int vce_v3_0_soft_reset(void *handle)
 	return 0;
 }
 
-static int vce_v3_0_pre_soft_reset(void *handle)
+static int vce_v3_0_pre_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (!adev->vce.srbm_soft_reset)
 		return 0;
 
 	mdelay(5);
 
-	return vce_v3_0_suspend(adev);
+	return vce_v3_0_suspend(ip_block);
 }
 
 
-static int vce_v3_0_post_soft_reset(void *handle)
+static int vce_v3_0_post_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (!adev->vce.srbm_soft_reset)
 		return 0;
 
 	mdelay(5);
 
-	return vce_v3_0_resume(adev);
+	return vce_v3_0_resume(ip_block);
 }
 
 static int vce_v3_0_set_interrupt_state(struct amdgpu_device *adev,
@@ -763,10 +765,10 @@ static int vce_v3_0_process_interrupt(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int vce_v3_0_set_clockgating_state(void *handle,
+static int vce_v3_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_CG_STATE_GATE);
 	int i;
 
@@ -804,7 +806,7 @@ static int vce_v3_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int vce_v3_0_set_powergating_state(void *handle,
+static int vce_v3_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	/* This doesn't actually powergate the VCE block.
@@ -814,7 +816,7 @@ static int vce_v3_0_set_powergating_state(void *handle,
 	 * revisit this when there is a cleaner line between
 	 * the smc and the hw blocks
 	 */
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret = 0;
 
 	if (state == AMD_PG_STATE_GATE) {
@@ -831,9 +833,9 @@ out:
 	return ret;
 }
 
-static void vce_v3_0_get_clockgating_state(void *handle, u64 *flags)
+static void vce_v3_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	mutex_lock(&adev->pm.mutex);
@@ -899,7 +901,6 @@ static void vce_v3_0_emit_pipeline_sync(struct amdgpu_ring *ring)
 static const struct amd_ip_funcs vce_v3_0_ip_funcs = {
 	.name = "vce_v3_0",
 	.early_init = vce_v3_0_early_init,
-	.late_init = NULL,
 	.sw_init = vce_v3_0_sw_init,
 	.sw_fini = vce_v3_0_sw_fini,
 	.hw_init = vce_v3_0_hw_init,
@@ -950,7 +951,7 @@ static const struct amdgpu_ring_funcs vce_v3_0_ring_vm_funcs = {
 	.get_rptr = vce_v3_0_ring_get_rptr,
 	.get_wptr = vce_v3_0_ring_get_wptr,
 	.set_wptr = vce_v3_0_ring_set_wptr,
-	.parse_cs = amdgpu_vce_ring_parse_cs_vm,
+	.patch_cs_in_place = amdgpu_vce_ring_parse_cs_vm,
 	.emit_frame_size =
 		6 + /* vce_v3_0_emit_vm_flush */
 		4 + /* vce_v3_0_emit_pipeline_sync */
@@ -998,8 +999,7 @@ static void vce_v3_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->vce.irq.funcs = &vce_v3_0_irq_funcs;
 };
 
-const struct amdgpu_ip_block_version vce_v3_0_ip_block =
-{
+const struct amdgpu_ip_block_version vce_v3_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_VCE,
 	.major = 3,
 	.minor = 0,
@@ -1007,8 +1007,7 @@ const struct amdgpu_ip_block_version vce_v3_0_ip_block =
 	.funcs = &vce_v3_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version vce_v3_1_ip_block =
-{
+const struct amdgpu_ip_block_version vce_v3_1_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_VCE,
 	.major = 3,
 	.minor = 1,
@@ -1016,8 +1015,7 @@ const struct amdgpu_ip_block_version vce_v3_1_ip_block =
 	.funcs = &vce_v3_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version vce_v3_4_ip_block =
-{
+const struct amdgpu_ip_block_version vce_v3_4_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_VCE,
 	.major = 3,
 	.minor = 4,
diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c
index 57b85bb6a1e4..2d64002bed61 100644
--- a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c
@@ -407,9 +407,14 @@ static int vce_v4_0_stop(struct amdgpu_device *adev)
 	return 0;
 }
 
-static int vce_v4_0_early_init(void *handle)
+static int vce_v4_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = amdgpu_vce_early_init(adev);
+	if (r)
+		return r;
 
 	if (amdgpu_sriov_vf(adev)) /* currently only VCN0 support SRIOV */
 		adev->vce.num_rings = 1;
@@ -422,9 +427,9 @@ static int vce_v4_0_early_init(void *handle)
 	return 0;
 }
 
-static int vce_v4_0_sw_init(void *handle)
+static int vce_v4_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring;
 
 	unsigned size;
@@ -466,7 +471,7 @@ static int vce_v4_0_sw_init(void *handle)
 		enum amdgpu_ring_priority_level hw_prio = amdgpu_vce_get_ring_prio(i);
 
 		ring = &adev->vce.ring[i];
-		ring->vm_hub = AMDGPU_MMHUB_0;
+		ring->vm_hub = AMDGPU_MMHUB0(0);
 		sprintf(ring->name, "vce%d", i);
 		if (amdgpu_sriov_vf(adev)) {
 			/* DOORBELL only works under SRIOV */
@@ -486,11 +491,6 @@ static int vce_v4_0_sw_init(void *handle)
 			return r;
 	}
 
-
-	r = amdgpu_vce_entity_init(adev);
-	if (r)
-		return r;
-
 	r = amdgpu_virt_alloc_mm_table(adev);
 	if (r)
 		return r;
@@ -498,10 +498,10 @@ static int vce_v4_0_sw_init(void *handle)
 	return r;
 }
 
-static int vce_v4_0_sw_fini(void *handle)
+static int vce_v4_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* free MM table */
 	amdgpu_virt_free_mm_table(adev);
@@ -518,10 +518,10 @@ static int vce_v4_0_sw_fini(void *handle)
 	return amdgpu_vce_sw_fini(adev);
 }
 
-static int vce_v4_0_hw_init(void *handle)
+static int vce_v4_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		r = vce_v4_0_sriov_start(adev);
@@ -541,14 +541,14 @@ static int vce_v4_0_hw_init(void *handle)
 	return 0;
 }
 
-static int vce_v4_0_hw_fini(void *handle)
+static int vce_v4_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	cancel_delayed_work_sync(&adev->vce.idle_work);
 
 	if (!amdgpu_sriov_vf(adev)) {
-		/* vce_v4_0_wait_for_idle(handle); */
+		/* vce_v4_0_wait_for_idle(ip_block); */
 		vce_v4_0_stop(adev);
 	} else {
 		/* full access mode, so don't touch any VCE register */
@@ -558,9 +558,9 @@ static int vce_v4_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int vce_v4_0_suspend(void *handle)
+static int vce_v4_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r, idx;
 
 	if (adev->vce.vcpu_bo == NULL)
@@ -599,16 +599,16 @@ static int vce_v4_0_suspend(void *handle)
 						       AMD_CG_STATE_GATE);
 	}
 
-	r = vce_v4_0_hw_fini(adev);
+	r = vce_v4_0_hw_fini(ip_block);
 	if (r)
 		return r;
 
 	return amdgpu_vce_suspend(adev);
 }
 
-static int vce_v4_0_resume(void *handle)
+static int vce_v4_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r, idx;
 
 	if (adev->vce.vcpu_bo == NULL)
@@ -629,7 +629,7 @@ static int vce_v4_0_resume(void *handle)
 			return r;
 	}
 
-	return vce_v4_0_hw_init(adev);
+	return vce_v4_0_hw_init(ip_block);
 }
 
 static void vce_v4_0_mc_resume(struct amdgpu_device *adev)
@@ -689,281 +689,14 @@ static void vce_v4_0_mc_resume(struct amdgpu_device *adev)
 			~VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK);
 }
 
-static int vce_v4_0_set_clockgating_state(void *handle,
+static int vce_v4_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	/* needed for driver unload*/
 	return 0;
 }
 
-#if 0
-static bool vce_v4_0_is_idle(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	u32 mask = 0;
-
-	mask |= (adev->vce.harvest_config & AMDGPU_VCE_HARVEST_VCE0) ? 0 : SRBM_STATUS2__VCE0_BUSY_MASK;
-	mask |= (adev->vce.harvest_config & AMDGPU_VCE_HARVEST_VCE1) ? 0 : SRBM_STATUS2__VCE1_BUSY_MASK;
-
-	return !(RREG32(mmSRBM_STATUS2) & mask);
-}
-
-static int vce_v4_0_wait_for_idle(void *handle)
-{
-	unsigned i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	for (i = 0; i < adev->usec_timeout; i++)
-		if (vce_v4_0_is_idle(handle))
-			return 0;
-
-	return -ETIMEDOUT;
-}
-
-#define  VCE_STATUS_VCPU_REPORT_AUTO_BUSY_MASK  0x00000008L   /* AUTO_BUSY */
-#define  VCE_STATUS_VCPU_REPORT_RB0_BUSY_MASK   0x00000010L   /* RB0_BUSY */
-#define  VCE_STATUS_VCPU_REPORT_RB1_BUSY_MASK   0x00000020L   /* RB1_BUSY */
-#define  AMDGPU_VCE_STATUS_BUSY_MASK (VCE_STATUS_VCPU_REPORT_AUTO_BUSY_MASK | \
-				      VCE_STATUS_VCPU_REPORT_RB0_BUSY_MASK)
-
-static bool vce_v4_0_check_soft_reset(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	u32 srbm_soft_reset = 0;
-
-	/* According to VCE team , we should use VCE_STATUS instead
-	 * SRBM_STATUS.VCE_BUSY bit for busy status checking.
-	 * GRBM_GFX_INDEX.INSTANCE_INDEX is used to specify which VCE
-	 * instance's registers are accessed
-	 * (0 for 1st instance, 10 for 2nd instance).
-	 *
-	 *VCE_STATUS
-	 *|UENC|ACPI|AUTO ACTIVE|RB1 |RB0 |RB2 |          |FW_LOADED|JOB |
-	 *|----+----+-----------+----+----+----+----------+---------+----|
-	 *|bit8|bit7|    bit6   |bit5|bit4|bit3|   bit2   |  bit1   |bit0|
-	 *
-	 * VCE team suggest use bit 3--bit 6 for busy status check
-	 */
-	mutex_lock(&adev->grbm_idx_mutex);
-	WREG32_FIELD(GRBM_GFX_INDEX, INSTANCE_INDEX, 0);
-	if (RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_STATUS) & AMDGPU_VCE_STATUS_BUSY_MASK) {
-		srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_VCE0, 1);
-		srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_VCE1, 1);
-	}
-	WREG32_FIELD(GRBM_GFX_INDEX, INSTANCE_INDEX, 0x10);
-	if (RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_STATUS) & AMDGPU_VCE_STATUS_BUSY_MASK) {
-		srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_VCE0, 1);
-		srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_VCE1, 1);
-	}
-	WREG32_FIELD(GRBM_GFX_INDEX, INSTANCE_INDEX, 0);
-	mutex_unlock(&adev->grbm_idx_mutex);
-
-	if (srbm_soft_reset) {
-		adev->vce.srbm_soft_reset = srbm_soft_reset;
-		return true;
-	} else {
-		adev->vce.srbm_soft_reset = 0;
-		return false;
-	}
-}
-
-static int vce_v4_0_soft_reset(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	u32 srbm_soft_reset;
-
-	if (!adev->vce.srbm_soft_reset)
-		return 0;
-	srbm_soft_reset = adev->vce.srbm_soft_reset;
-
-	if (srbm_soft_reset) {
-		u32 tmp;
-
-		tmp = RREG32(mmSRBM_SOFT_RESET);
-		tmp |= srbm_soft_reset;
-		dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp);
-		WREG32(mmSRBM_SOFT_RESET, tmp);
-		tmp = RREG32(mmSRBM_SOFT_RESET);
-
-		udelay(50);
-
-		tmp &= ~srbm_soft_reset;
-		WREG32(mmSRBM_SOFT_RESET, tmp);
-		tmp = RREG32(mmSRBM_SOFT_RESET);
-
-		/* Wait a little for things to settle down */
-		udelay(50);
-	}
-
-	return 0;
-}
-
-static int vce_v4_0_pre_soft_reset(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	if (!adev->vce.srbm_soft_reset)
-		return 0;
-
-	mdelay(5);
-
-	return vce_v4_0_suspend(adev);
-}
-
-
-static int vce_v4_0_post_soft_reset(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	if (!adev->vce.srbm_soft_reset)
-		return 0;
-
-	mdelay(5);
-
-	return vce_v4_0_resume(adev);
-}
-
-static void vce_v4_0_override_vce_clock_gating(struct amdgpu_device *adev, bool override)
-{
-	u32 tmp, data;
-
-	tmp = data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_ARB_CTRL));
-	if (override)
-		data |= VCE_RB_ARB_CTRL__VCE_CGTT_OVERRIDE_MASK;
-	else
-		data &= ~VCE_RB_ARB_CTRL__VCE_CGTT_OVERRIDE_MASK;
-
-	if (tmp != data)
-		WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_ARB_CTRL), data);
-}
-
-static void vce_v4_0_set_vce_sw_clock_gating(struct amdgpu_device *adev,
-					     bool gated)
-{
-	u32 data;
-
-	/* Set Override to disable Clock Gating */
-	vce_v4_0_override_vce_clock_gating(adev, true);
-
-	/* This function enables MGCG which is controlled by firmware.
-	   With the clocks in the gated state the core is still
-	   accessible but the firmware will throttle the clocks on the
-	   fly as necessary.
-	*/
-	if (gated) {
-		data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_CLOCK_GATING_B));
-		data |= 0x1ff;
-		data &= ~0xef0000;
-		WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_CLOCK_GATING_B), data);
-
-		data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_CLOCK_GATING));
-		data |= 0x3ff000;
-		data &= ~0xffc00000;
-		WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_CLOCK_GATING), data);
-
-		data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_CLOCK_GATING_2));
-		data |= 0x2;
-		data &= ~0x00010000;
-		WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_CLOCK_GATING_2), data);
-
-		data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_REG_CLOCK_GATING));
-		data |= 0x37f;
-		WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_REG_CLOCK_GATING), data);
-
-		data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_DMA_DCLK_CTRL));
-		data |= VCE_UENC_DMA_DCLK_CTRL__WRDMCLK_FORCEON_MASK |
-			VCE_UENC_DMA_DCLK_CTRL__RDDMCLK_FORCEON_MASK |
-			VCE_UENC_DMA_DCLK_CTRL__REGCLK_FORCEON_MASK  |
-			0x8;
-		WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_DMA_DCLK_CTRL), data);
-	} else {
-		data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_CLOCK_GATING_B));
-		data &= ~0x80010;
-		data |= 0xe70008;
-		WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_CLOCK_GATING_B), data);
-
-		data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_CLOCK_GATING));
-		data |= 0xffc00000;
-		WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_CLOCK_GATING), data);
-
-		data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_CLOCK_GATING_2));
-		data |= 0x10000;
-		WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_CLOCK_GATING_2), data);
-
-		data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_REG_CLOCK_GATING));
-		data &= ~0xffc00000;
-		WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_REG_CLOCK_GATING), data);
-
-		data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_DMA_DCLK_CTRL));
-		data &= ~(VCE_UENC_DMA_DCLK_CTRL__WRDMCLK_FORCEON_MASK |
-			  VCE_UENC_DMA_DCLK_CTRL__RDDMCLK_FORCEON_MASK |
-			  VCE_UENC_DMA_DCLK_CTRL__REGCLK_FORCEON_MASK  |
-			  0x8);
-		WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_DMA_DCLK_CTRL), data);
-	}
-	vce_v4_0_override_vce_clock_gating(adev, false);
-}
-
-static void vce_v4_0_set_bypass_mode(struct amdgpu_device *adev, bool enable)
-{
-	u32 tmp = RREG32_SMC(ixGCK_DFS_BYPASS_CNTL);
-
-	if (enable)
-		tmp |= GCK_DFS_BYPASS_CNTL__BYPASSECLK_MASK;
-	else
-		tmp &= ~GCK_DFS_BYPASS_CNTL__BYPASSECLK_MASK;
-
-	WREG32_SMC(ixGCK_DFS_BYPASS_CNTL, tmp);
-}
-
-static int vce_v4_0_set_clockgating_state(void *handle,
-					  enum amd_clockgating_state state)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	bool enable = (state == AMD_CG_STATE_GATE);
-	int i;
-
-	if ((adev->asic_type == CHIP_POLARIS10) ||
-		(adev->asic_type == CHIP_TONGA) ||
-		(adev->asic_type == CHIP_FIJI))
-		vce_v4_0_set_bypass_mode(adev, enable);
-
-	if (!(adev->cg_flags & AMD_CG_SUPPORT_VCE_MGCG))
-		return 0;
-
-	mutex_lock(&adev->grbm_idx_mutex);
-	for (i = 0; i < 2; i++) {
-		/* Program VCE Instance 0 or 1 if not harvested */
-		if (adev->vce.harvest_config & (1 << i))
-			continue;
-
-		WREG32_FIELD(GRBM_GFX_INDEX, VCE_INSTANCE, i);
-
-		if (enable) {
-			/* initialize VCE_CLOCK_GATING_A: Clock ON/OFF delay */
-			uint32_t data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_CLOCK_GATING_A);
-			data &= ~(0xf | 0xff0);
-			data |= ((0x0 << 0) | (0x04 << 4));
-			WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_CLOCK_GATING_A, data);
-
-			/* initialize VCE_UENC_CLOCK_GATING: Clock ON/OFF delay */
-			data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_CLOCK_GATING);
-			data &= ~(0xf | 0xff0);
-			data |= ((0x0 << 0) | (0x04 << 4));
-			WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_CLOCK_GATING, data);
-		}
-
-		vce_v4_0_set_vce_sw_clock_gating(adev, enable);
-	}
-
-	WREG32_FIELD(GRBM_GFX_INDEX, VCE_INSTANCE, 0);
-	mutex_unlock(&adev->grbm_idx_mutex);
-
-	return 0;
-}
-#endif
-
-static int vce_v4_0_set_powergating_state(void *handle,
+static int vce_v4_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	/* This doesn't actually powergate the VCE block.
@@ -973,7 +706,7 @@ static int vce_v4_0_set_powergating_state(void *handle,
 	 * revisit this when there is a cleaner line between
 	 * the smc and the hw blocks
 	 */
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (state == AMD_PG_STATE_GATE)
 		return vce_v4_0_stop(adev);
@@ -1081,19 +814,12 @@ static int vce_v4_0_process_interrupt(struct amdgpu_device *adev,
 const struct amd_ip_funcs vce_v4_0_ip_funcs = {
 	.name = "vce_v4_0",
 	.early_init = vce_v4_0_early_init,
-	.late_init = NULL,
 	.sw_init = vce_v4_0_sw_init,
 	.sw_fini = vce_v4_0_sw_fini,
 	.hw_init = vce_v4_0_hw_init,
 	.hw_fini = vce_v4_0_hw_fini,
 	.suspend = vce_v4_0_suspend,
 	.resume = vce_v4_0_resume,
-	.is_idle = NULL /* vce_v4_0_is_idle */,
-	.wait_for_idle = NULL /* vce_v4_0_wait_for_idle */,
-	.check_soft_reset = NULL /* vce_v4_0_check_soft_reset */,
-	.pre_soft_reset = NULL /* vce_v4_0_pre_soft_reset */,
-	.soft_reset = NULL /* vce_v4_0_soft_reset */,
-	.post_soft_reset = NULL /* vce_v4_0_post_soft_reset */,
 	.set_clockgating_state = vce_v4_0_set_clockgating_state,
 	.set_powergating_state = vce_v4_0_set_powergating_state,
 };
@@ -1107,7 +833,7 @@ static const struct amdgpu_ring_funcs vce_v4_0_ring_vm_funcs = {
 	.get_rptr = vce_v4_0_ring_get_rptr,
 	.get_wptr = vce_v4_0_ring_get_wptr,
 	.set_wptr = vce_v4_0_ring_set_wptr,
-	.parse_cs = amdgpu_vce_ring_parse_cs_vm,
+	.patch_cs_in_place = amdgpu_vce_ring_parse_cs_vm,
 	.emit_frame_size =
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 4 +
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c
index 761c28fa6ec1..a316797875a8 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c
@@ -45,13 +45,50 @@
 #define mmUVD_REG_XX_MASK_1_0			0x05ac
 #define mmUVD_REG_XX_MASK_1_0_BASE_IDX		1
 
-static int vcn_v1_0_stop(struct amdgpu_device *adev);
+static const struct amdgpu_hwip_reg_entry vcn_reg_list_1_0[] = {
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_POWER_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_CONTEXT_ID),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_CONTEXT_ID2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_GPCOM_VCPU_DATA0),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_GPCOM_VCPU_DATA1),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_GPCOM_VCPU_CMD),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_PGFSM_CONFIG),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_PGFSM_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_LMA_CTL),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_LMA_DATA),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_LMA_MASK),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_PAUSE)
+};
+
+static int vcn_v1_0_stop(struct amdgpu_vcn_inst *vinst);
 static void vcn_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev);
 static void vcn_v1_0_set_enc_ring_funcs(struct amdgpu_device *adev);
 static void vcn_v1_0_set_irq_funcs(struct amdgpu_device *adev);
-static int vcn_v1_0_set_powergating_state(void *handle, enum amd_powergating_state state);
-static int vcn_v1_0_pause_dpg_mode(struct amdgpu_device *adev,
-				int inst_idx, struct dpg_pause_state *new_state);
+static int vcn_v1_0_set_pg_state(struct amdgpu_vcn_inst *vinst,
+				 enum amd_powergating_state state);
+static int vcn_v1_0_pause_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				   struct dpg_pause_state *new_state);
 
 static void vcn_v1_0_idle_work_handler(struct work_struct *work);
 static void vcn_v1_0_ring_begin_use(struct amdgpu_ring *ring);
@@ -59,38 +96,41 @@ static void vcn_v1_0_ring_begin_use(struct amdgpu_ring *ring);
 /**
  * vcn_v1_0_early_init - set function pointers and load microcode
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Set ring and irq function pointers
  * Load microcode from filesystem
  */
-static int vcn_v1_0_early_init(void *handle)
+static int vcn_v1_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	adev->vcn.num_enc_rings = 2;
+	adev->vcn.inst[0].num_enc_rings = 2;
+	adev->vcn.inst[0].set_pg_state = vcn_v1_0_set_pg_state;
 
 	vcn_v1_0_set_dec_ring_funcs(adev);
 	vcn_v1_0_set_enc_ring_funcs(adev);
 	vcn_v1_0_set_irq_funcs(adev);
 
-	jpeg_v1_0_early_init(handle);
+	jpeg_v1_0_early_init(ip_block);
 
-	return amdgpu_vcn_early_init(adev);
+	return amdgpu_vcn_early_init(adev, 0);
 }
 
 /**
  * vcn_v1_0_sw_init - sw init for VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Load firmware and sw initialization
  */
-static int vcn_v1_0_sw_init(void *handle)
+static int vcn_v1_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
 	int i, r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	uint32_t reg_count = ARRAY_SIZE(vcn_reg_list_1_0);
+	uint32_t *ptr;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* VCN DEC TRAP */
 	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
@@ -99,50 +139,50 @@ static int vcn_v1_0_sw_init(void *handle)
 		return r;
 
 	/* VCN ENC TRAP */
-	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
+	for (i = 0; i < adev->vcn.inst[0].num_enc_rings; ++i) {
 		r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, i + VCN_1_0__SRCID__UVD_ENC_GENERAL_PURPOSE,
 					&adev->vcn.inst->irq);
 		if (r)
 			return r;
 	}
 
-	r = amdgpu_vcn_sw_init(adev);
+	r = amdgpu_vcn_sw_init(adev, 0);
 	if (r)
 		return r;
 
 	/* Override the work func */
-	adev->vcn.idle_work.work.func = vcn_v1_0_idle_work_handler;
+	adev->vcn.inst[0].idle_work.work.func = vcn_v1_0_idle_work_handler;
 
-	amdgpu_vcn_setup_ucode(adev);
+	amdgpu_vcn_setup_ucode(adev, 0);
 
-	r = amdgpu_vcn_resume(adev);
+	r = amdgpu_vcn_resume(adev, 0);
 	if (r)
 		return r;
 
 	ring = &adev->vcn.inst->ring_dec;
-	ring->vm_hub = AMDGPU_MMHUB_0;
+	ring->vm_hub = AMDGPU_MMHUB0(0);
 	sprintf(ring->name, "vcn_dec");
 	r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0,
 			     AMDGPU_RING_PRIO_DEFAULT, NULL);
 	if (r)
 		return r;
 
-	adev->vcn.internal.scratch9 = adev->vcn.inst->external.scratch9 =
+	adev->vcn.inst[0].internal.scratch9 = adev->vcn.inst->external.scratch9 =
 		SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9);
-	adev->vcn.internal.data0 = adev->vcn.inst->external.data0 =
+	adev->vcn.inst[0].internal.data0 = adev->vcn.inst->external.data0 =
 		SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0);
-	adev->vcn.internal.data1 = adev->vcn.inst->external.data1 =
+	adev->vcn.inst[0].internal.data1 = adev->vcn.inst->external.data1 =
 		SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1);
-	adev->vcn.internal.cmd = adev->vcn.inst->external.cmd =
+	adev->vcn.inst[0].internal.cmd = adev->vcn.inst->external.cmd =
 		SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD);
-	adev->vcn.internal.nop = adev->vcn.inst->external.nop =
+	adev->vcn.inst[0].internal.nop = adev->vcn.inst->external.nop =
 		SOC15_REG_OFFSET(UVD, 0, mmUVD_NO_OP);
 
-	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
+	for (i = 0; i < adev->vcn.inst[0].num_enc_rings; ++i) {
 		enum amdgpu_ring_priority_level hw_prio = amdgpu_vcn_get_enc_ring_prio(i);
 
 		ring = &adev->vcn.inst->ring_enc[i];
-		ring->vm_hub = AMDGPU_MMHUB_0;
+		ring->vm_hub = AMDGPU_MMHUB0(0);
 		sprintf(ring->name, "vcn_enc%d", i);
 		r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0,
 				     hw_prio, NULL);
@@ -150,76 +190,79 @@ static int vcn_v1_0_sw_init(void *handle)
 			return r;
 	}
 
-	adev->vcn.pause_dpg_mode = vcn_v1_0_pause_dpg_mode;
+	adev->vcn.inst[0].pause_dpg_mode = vcn_v1_0_pause_dpg_mode;
 
 	if (amdgpu_vcnfw_log) {
-		volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared.cpu_addr;
+		struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared.cpu_addr;
 
 		fw_shared->present_flag_0 = 0;
 		amdgpu_vcn_fwlog_init(adev->vcn.inst);
 	}
 
-	r = jpeg_v1_0_sw_init(handle);
+	r = jpeg_v1_0_sw_init(ip_block);
 
+	/* Allocate memory for VCN IP Dump buffer */
+	ptr = kcalloc(adev->vcn.num_vcn_inst * reg_count, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for VCN IP Dump\n");
+		adev->vcn.ip_dump = NULL;
+	} else {
+		adev->vcn.ip_dump = ptr;
+	}
 	return r;
 }
 
 /**
  * vcn_v1_0_sw_fini - sw fini for VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * VCN suspend and free up sw allocation
  */
-static int vcn_v1_0_sw_fini(void *handle)
+static int vcn_v1_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	r = amdgpu_vcn_suspend(adev);
+	r = amdgpu_vcn_suspend(adev, 0);
 	if (r)
 		return r;
 
-	jpeg_v1_0_sw_fini(handle);
+	jpeg_v1_0_sw_fini(ip_block);
 
-	r = amdgpu_vcn_sw_fini(adev);
+	amdgpu_vcn_sw_fini(adev, 0);
 
-	return r;
+	kfree(adev->vcn.ip_dump);
+
+	return 0;
 }
 
 /**
  * vcn_v1_0_hw_init - start and test VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Initialize the hardware, boot up the VCPU and do some testing
  */
-static int vcn_v1_0_hw_init(void *handle)
+static int vcn_v1_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring = &adev->vcn.inst->ring_dec;
 	int i, r;
 
 	r = amdgpu_ring_test_helper(ring);
 	if (r)
-		goto done;
+		return r;
 
-	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
+	for (i = 0; i < adev->vcn.inst[0].num_enc_rings; ++i) {
 		ring = &adev->vcn.inst->ring_enc[i];
 		r = amdgpu_ring_test_helper(ring);
 		if (r)
-			goto done;
+			return r;
 	}
 
-	ring = &adev->jpeg.inst->ring_dec;
+	ring = adev->jpeg.inst->ring_dec;
 	r = amdgpu_ring_test_helper(ring);
-	if (r)
-		goto done;
-
-done:
-	if (!r)
-		DRM_INFO("VCN decode and encode initialized successfully(under %s).\n",
-			(adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)?"DPG Mode":"SPG Mode");
 
 	return r;
 }
@@ -227,20 +270,21 @@ done:
 /**
  * vcn_v1_0_hw_fini - stop the hardware block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Stop the VCN block, mark ring as not ready any more
  */
-static int vcn_v1_0_hw_fini(void *handle)
+static int vcn_v1_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_vcn_inst *vinst = adev->vcn.inst;
 
-	cancel_delayed_work_sync(&adev->vcn.idle_work);
+	cancel_delayed_work_sync(&vinst->idle_work);
 
 	if ((adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) ||
-		(adev->vcn.cur_state != AMD_PG_STATE_GATE &&
-		 RREG32_SOC15(VCN, 0, mmUVD_STATUS))) {
-		vcn_v1_0_set_powergating_state(adev, AMD_PG_STATE_GATE);
+	    (vinst->cur_state != AMD_PG_STATE_GATE &&
+	     RREG32_SOC15(VCN, 0, mmUVD_STATUS))) {
+		vinst->set_pg_state(vinst, AMD_PG_STATE_GATE);
 	}
 
 	return 0;
@@ -249,27 +293,27 @@ static int vcn_v1_0_hw_fini(void *handle)
 /**
  * vcn_v1_0_suspend - suspend VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * HW fini and suspend VCN block
  */
-static int vcn_v1_0_suspend(void *handle)
+static int vcn_v1_0_suspend(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool idle_work_unexecuted;
 
-	idle_work_unexecuted = cancel_delayed_work_sync(&adev->vcn.idle_work);
+	idle_work_unexecuted = cancel_delayed_work_sync(&adev->vcn.inst[0].idle_work);
 	if (idle_work_unexecuted) {
 		if (adev->pm.dpm_enabled)
-			amdgpu_dpm_enable_uvd(adev, false);
+			amdgpu_dpm_enable_vcn(adev, false, 0);
 	}
 
-	r = vcn_v1_0_hw_fini(adev);
+	r = vcn_v1_0_hw_fini(ip_block);
 	if (r)
 		return r;
 
-	r = amdgpu_vcn_suspend(adev);
+	r = amdgpu_vcn_suspend(adev, 0);
 
 	return r;
 }
@@ -277,20 +321,19 @@ static int vcn_v1_0_suspend(void *handle)
 /**
  * vcn_v1_0_resume - resume VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Resume firmware and hw init VCN block
  */
-static int vcn_v1_0_resume(void *handle)
+static int vcn_v1_0_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = amdgpu_vcn_resume(adev);
+	r = amdgpu_vcn_resume(ip_block->adev, 0);
 	if (r)
 		return r;
 
-	r = vcn_v1_0_hw_init(adev);
+	r = vcn_v1_0_hw_init(ip_block);
 
 	return r;
 }
@@ -298,13 +341,14 @@ static int vcn_v1_0_resume(void *handle)
 /**
  * vcn_v1_0_mc_resume_spg_mode - memory controller programming
  *
- * @adev: amdgpu_device pointer
+ * @vinst: VCN instance
  *
  * Let the VCN memory controller know it's offsets
  */
-static void vcn_v1_0_mc_resume_spg_mode(struct amdgpu_device *adev)
+static void vcn_v1_0_mc_resume_spg_mode(struct amdgpu_vcn_inst *vinst)
 {
-	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+	struct amdgpu_device *adev = vinst->adev;
+	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.inst[0].fw->size + 4);
 	uint32_t offset;
 
 	/* cache window 0: fw */
@@ -369,9 +413,10 @@ static void vcn_v1_0_mc_resume_spg_mode(struct amdgpu_device *adev)
 			adev->gfx.config.gb_addr_config);
 }
 
-static void vcn_v1_0_mc_resume_dpg_mode(struct amdgpu_device *adev)
+static void vcn_v1_0_mc_resume_dpg_mode(struct amdgpu_vcn_inst *vinst)
 {
-	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+	struct amdgpu_device *adev = vinst->adev;
+	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.inst[0].fw->size + 4);
 	uint32_t offset;
 
 	/* cache window 0: fw */
@@ -444,12 +489,13 @@ static void vcn_v1_0_mc_resume_dpg_mode(struct amdgpu_device *adev)
 /**
  * vcn_v1_0_disable_clock_gating - disable VCN clock gating
  *
- * @adev: amdgpu_device pointer
+ * @vinst: VCN instance
  *
  * Disable clock gating for VCN block
  */
-static void vcn_v1_0_disable_clock_gating(struct amdgpu_device *adev)
+static void vcn_v1_0_disable_clock_gating(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
 	uint32_t data;
 
 	/* JPEG disable CGC */
@@ -473,7 +519,7 @@ static void vcn_v1_0_disable_clock_gating(struct amdgpu_device *adev)
 	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
 		data |= 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
 	else
-		data &= ~ UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK;
+		data &= ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK;
 
 	data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
 	data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
@@ -570,12 +616,13 @@ static void vcn_v1_0_disable_clock_gating(struct amdgpu_device *adev)
 /**
  * vcn_v1_0_enable_clock_gating - enable VCN clock gating
  *
- * @adev: amdgpu_device pointer
+ * @vinst: Pointer to the VCN instance structure
  *
  * Enable clock gating for VCN block
  */
-static void vcn_v1_0_enable_clock_gating(struct amdgpu_device *adev)
+static void vcn_v1_0_enable_clock_gating(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
 	uint32_t data = 0;
 
 	/* enable JPEG CGC */
@@ -639,8 +686,10 @@ static void vcn_v1_0_enable_clock_gating(struct amdgpu_device *adev)
 	WREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL, data);
 }
 
-static void vcn_v1_0_clock_gating_dpg_mode(struct amdgpu_device *adev, uint8_t sram_sel)
+static void vcn_v1_0_clock_gating_dpg_mode(struct amdgpu_vcn_inst *vinst,
+					   uint8_t sram_sel)
 {
+	struct amdgpu_device *adev = vinst->adev;
 	uint32_t reg_data = 0;
 
 	/* disable JPEG CGC */
@@ -693,8 +742,9 @@ static void vcn_v1_0_clock_gating_dpg_mode(struct amdgpu_device *adev, uint8_t s
 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_SUVD_CGC_CTRL, 0, 0xFFFFFFFF, sram_sel);
 }
 
-static void vcn_1_0_disable_static_power_gating(struct amdgpu_device *adev)
+static void vcn_1_0_disable_static_power_gating(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
 	uint32_t data = 0;
 
 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN) {
@@ -738,8 +788,9 @@ static void vcn_1_0_disable_static_power_gating(struct amdgpu_device *adev)
 	WREG32_SOC15(VCN, 0, mmUVD_POWER_STATUS, data);
 }
 
-static void vcn_1_0_enable_static_power_gating(struct amdgpu_device *adev)
+static void vcn_1_0_enable_static_power_gating(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
 	uint32_t data = 0;
 
 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN) {
@@ -782,12 +833,13 @@ static void vcn_1_0_enable_static_power_gating(struct amdgpu_device *adev)
 /**
  * vcn_v1_0_start_spg_mode - start VCN block
  *
- * @adev: amdgpu_device pointer
+ * @vinst: VCN instance
  *
  * Setup and start the VCN block
  */
-static int vcn_v1_0_start_spg_mode(struct amdgpu_device *adev)
+static int vcn_v1_0_start_spg_mode(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
 	struct amdgpu_ring *ring = &adev->vcn.inst->ring_dec;
 	uint32_t rb_bufsz, tmp;
 	uint32_t lmi_swap_cntl;
@@ -796,13 +848,13 @@ static int vcn_v1_0_start_spg_mode(struct amdgpu_device *adev)
 	/* disable byte swapping */
 	lmi_swap_cntl = 0;
 
-	vcn_1_0_disable_static_power_gating(adev);
+	vcn_1_0_disable_static_power_gating(vinst);
 
 	tmp = RREG32_SOC15(UVD, 0, mmUVD_STATUS) | UVD_STATUS__UVD_BUSY;
 	WREG32_SOC15(UVD, 0, mmUVD_STATUS, tmp);
 
 	/* disable clock gating */
-	vcn_v1_0_disable_clock_gating(adev);
+	vcn_v1_0_disable_clock_gating(vinst);
 
 	/* disable interupt */
 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN), 0,
@@ -844,7 +896,7 @@ static int vcn_v1_0_start_spg_mode(struct amdgpu_device *adev)
 		(0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
 		(0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)));
 
-	vcn_v1_0_mc_resume_spg_mode(adev);
+	vcn_v1_0_mc_resume_spg_mode(vinst);
 
 	WREG32_SOC15(UVD, 0, mmUVD_REG_XX_MASK_1_0, 0x10);
 	WREG32_SOC15(UVD, 0, mmUVD_RBC_XX_IB_REG_CHECK_1_0,
@@ -957,11 +1009,17 @@ static int vcn_v1_0_start_spg_mode(struct amdgpu_device *adev)
 
 	jpeg_v1_0_start(adev, 0);
 
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(UVD, 0, mmUVD_STATUS);
+
 	return 0;
 }
 
-static int vcn_v1_0_start_dpg_mode(struct amdgpu_device *adev)
+static int vcn_v1_0_start_dpg_mode(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
 	struct amdgpu_ring *ring = &adev->vcn.inst->ring_dec;
 	uint32_t rb_bufsz, tmp;
 	uint32_t lmi_swap_cntl;
@@ -969,7 +1027,7 @@ static int vcn_v1_0_start_dpg_mode(struct amdgpu_device *adev)
 	/* disable byte swapping */
 	lmi_swap_cntl = 0;
 
-	vcn_1_0_enable_static_power_gating(adev);
+	vcn_1_0_enable_static_power_gating(vinst);
 
 	/* enable dynamic power gating mode */
 	tmp = RREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS);
@@ -978,7 +1036,7 @@ static int vcn_v1_0_start_dpg_mode(struct amdgpu_device *adev)
 	WREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS, tmp);
 
 	/* enable clock gating */
-	vcn_v1_0_clock_gating_dpg_mode(adev, 0);
+	vcn_v1_0_clock_gating_dpg_mode(vinst, 0);
 
 	/* enable VCPU clock */
 	tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
@@ -1027,7 +1085,7 @@ static int vcn_v1_0_start_dpg_mode(struct amdgpu_device *adev)
 		 (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
 		 (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)), 0xFFFFFFFF, 0);
 
-	vcn_v1_0_mc_resume_dpg_mode(adev);
+	vcn_v1_0_mc_resume_dpg_mode(vinst);
 
 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_REG_XX_MASK, 0x10, 0xFFFFFFFF, 0);
 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_RBC_XX_IB_REG_CHECK, 0x3, 0xFFFFFFFF, 0);
@@ -1044,7 +1102,7 @@ static int vcn_v1_0_start_dpg_mode(struct amdgpu_device *adev)
 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_MASTINT_EN,
 			UVD_MASTINT_EN__VCPU_EN_MASK, UVD_MASTINT_EN__VCPU_EN_MASK, 0);
 
-	vcn_v1_0_clock_gating_dpg_mode(adev, 1);
+	vcn_v1_0_clock_gating_dpg_mode(vinst, 1);
 	/* setup mmUVD_LMI_CTRL */
 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_LMI_CTRL,
 		(8 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) |
@@ -1101,24 +1159,32 @@ static int vcn_v1_0_start_dpg_mode(struct amdgpu_device *adev)
 
 	jpeg_v1_0_start(adev, 1);
 
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(UVD, 0, mmUVD_STATUS);
+
 	return 0;
 }
 
-static int vcn_v1_0_start(struct amdgpu_device *adev)
+static int vcn_v1_0_start(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
+
 	return (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) ?
-		vcn_v1_0_start_dpg_mode(adev) : vcn_v1_0_start_spg_mode(adev);
+		vcn_v1_0_start_dpg_mode(vinst) : vcn_v1_0_start_spg_mode(vinst);
 }
 
 /**
  * vcn_v1_0_stop_spg_mode - stop VCN block
  *
- * @adev: amdgpu_device pointer
+ * @vinst: VCN instance
  *
  * stop the VCN block
  */
-static int vcn_v1_0_stop_spg_mode(struct amdgpu_device *adev)
+static int vcn_v1_0_stop_spg_mode(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
 	int tmp;
 
 	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_STATUS, UVD_STATUS__IDLE, 0x7);
@@ -1158,13 +1224,20 @@ static int vcn_v1_0_stop_spg_mode(struct amdgpu_device *adev)
 
 	WREG32_SOC15(UVD, 0, mmUVD_STATUS, 0);
 
-	vcn_v1_0_enable_clock_gating(adev);
-	vcn_1_0_enable_static_power_gating(adev);
+	vcn_v1_0_enable_clock_gating(vinst);
+	vcn_1_0_enable_static_power_gating(vinst);
+
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(UVD, 0, mmUVD_STATUS);
+
 	return 0;
 }
 
-static int vcn_v1_0_stop_dpg_mode(struct amdgpu_device *adev)
+static int vcn_v1_0_stop_dpg_mode(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
 	uint32_t tmp;
 
 	/* Wait for power status to be UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF */
@@ -1193,24 +1266,32 @@ static int vcn_v1_0_stop_dpg_mode(struct amdgpu_device *adev)
 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_POWER_STATUS), 0,
 			~UVD_POWER_STATUS__UVD_PG_MODE_MASK);
 
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(UVD, 0, mmUVD_STATUS);
+
 	return 0;
 }
 
-static int vcn_v1_0_stop(struct amdgpu_device *adev)
+static int vcn_v1_0_stop(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
 	int r;
 
 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
-		r = vcn_v1_0_stop_dpg_mode(adev);
+		r = vcn_v1_0_stop_dpg_mode(vinst);
 	else
-		r = vcn_v1_0_stop_spg_mode(adev);
+		r = vcn_v1_0_stop_spg_mode(vinst);
 
 	return r;
 }
 
-static int vcn_v1_0_pause_dpg_mode(struct amdgpu_device *adev,
-				int inst_idx, struct dpg_pause_state *new_state)
+static int vcn_v1_0_pause_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				   struct dpg_pause_state *new_state)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
 	int ret_code;
 	uint32_t reg_data = 0;
 	uint32_t reg_data2 = 0;
@@ -1257,7 +1338,6 @@ static int vcn_v1_0_pause_dpg_mode(struct amdgpu_device *adev,
 				WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr));
 				WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr));
 
-				ring = &adev->vcn.inst->ring_dec;
 				WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR,
 						   RREG32_SOC15(UVD, 0, mmUVD_SCRATCH2) & 0x7FFFFFFF);
 				SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS,
@@ -1304,7 +1384,7 @@ static int vcn_v1_0_pause_dpg_mode(struct amdgpu_device *adev,
 							UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK);
 
 				/* Restore */
-				ring = &adev->jpeg.inst->ring_dec;
+				ring = adev->jpeg.inst->ring_dec;
 				WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_VMID, 0);
 				WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_CNTL,
 							UVD_JRBC_RB_CNTL__RB_NO_FETCH_MASK |
@@ -1318,7 +1398,6 @@ static int vcn_v1_0_pause_dpg_mode(struct amdgpu_device *adev,
 				WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_CNTL,
 							UVD_JRBC_RB_CNTL__RB_RPTR_WR_EN_MASK);
 
-				ring = &adev->vcn.inst->ring_dec;
 				WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR,
 						   RREG32_SOC15(UVD, 0, mmUVD_SCRATCH2) & 0x7FFFFFFF);
 				SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS,
@@ -1336,16 +1415,16 @@ static int vcn_v1_0_pause_dpg_mode(struct amdgpu_device *adev,
 	return 0;
 }
 
-static bool vcn_v1_0_is_idle(void *handle)
+static bool vcn_v1_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return (RREG32_SOC15(VCN, 0, mmUVD_STATUS) == UVD_STATUS__IDLE);
 }
 
-static int vcn_v1_0_wait_for_idle(void *handle)
+static int vcn_v1_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	ret = SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_STATUS, UVD_STATUS__IDLE,
@@ -1354,20 +1433,21 @@ static int vcn_v1_0_wait_for_idle(void *handle)
 	return ret;
 }
 
-static int vcn_v1_0_set_clockgating_state(void *handle,
+static int vcn_v1_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_vcn_inst *vinst = adev->vcn.inst;
 	bool enable = (state == AMD_CG_STATE_GATE);
 
 	if (enable) {
 		/* wait for STATUS to clear */
-		if (!vcn_v1_0_is_idle(handle))
+		if (!vcn_v1_0_is_idle(ip_block))
 			return -EBUSY;
-		vcn_v1_0_enable_clock_gating(adev);
+		vcn_v1_0_enable_clock_gating(vinst);
 	} else {
 		/* disable HW gating and enable Sw gating */
-		vcn_v1_0_disable_clock_gating(adev);
+		vcn_v1_0_disable_clock_gating(vinst);
 	}
 	return 0;
 }
@@ -1759,8 +1839,8 @@ static void vcn_v1_0_dec_ring_insert_nop(struct amdgpu_ring *ring, uint32_t coun
 	}
 }
 
-static int vcn_v1_0_set_powergating_state(void *handle,
-					  enum amd_powergating_state state)
+static int vcn_v1_0_set_pg_state(struct amdgpu_vcn_inst *vinst,
+				 enum amd_powergating_state state)
 {
 	/* This doesn't actually powergate the VCN block.
 	 * That's done in the dpm code via the SMC.  This
@@ -1770,28 +1850,29 @@ static int vcn_v1_0_set_powergating_state(void *handle,
 	 * the smc and the hw blocks
 	 */
 	int ret;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	if(state == adev->vcn.cur_state)
+	if (state == vinst->cur_state)
 		return 0;
 
 	if (state == AMD_PG_STATE_GATE)
-		ret = vcn_v1_0_stop(adev);
+		ret = vcn_v1_0_stop(vinst);
 	else
-		ret = vcn_v1_0_start(adev);
+		ret = vcn_v1_0_start(vinst);
+
+	if (!ret)
+		vinst->cur_state = state;
 
-	if(!ret)
-		adev->vcn.cur_state = state;
 	return ret;
 }
 
 static void vcn_v1_0_idle_work_handler(struct work_struct *work)
 {
-	struct amdgpu_device *adev =
-		container_of(work, struct amdgpu_device, vcn.idle_work.work);
+	struct amdgpu_vcn_inst *vcn_inst =
+		container_of(work, struct amdgpu_vcn_inst, idle_work.work);
+	struct amdgpu_device *adev = vcn_inst->adev;
 	unsigned int fences = 0, i;
 
-	for (i = 0; i < adev->vcn.num_enc_rings; ++i)
+	for (i = 0; i < adev->vcn.inst[0].num_enc_rings; ++i)
 		fences += amdgpu_fence_count_emitted(&adev->vcn.inst->ring_enc[i]);
 
 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
@@ -1802,37 +1883,37 @@ static void vcn_v1_0_idle_work_handler(struct work_struct *work)
 		else
 			new_state.fw_based = VCN_DPG_STATE__UNPAUSE;
 
-		if (amdgpu_fence_count_emitted(&adev->jpeg.inst->ring_dec))
+		if (amdgpu_fence_count_emitted(adev->jpeg.inst->ring_dec))
 			new_state.jpeg = VCN_DPG_STATE__PAUSE;
 		else
 			new_state.jpeg = VCN_DPG_STATE__UNPAUSE;
 
-		adev->vcn.pause_dpg_mode(adev, 0, &new_state);
+		adev->vcn.inst->pause_dpg_mode(vcn_inst, &new_state);
 	}
 
-	fences += amdgpu_fence_count_emitted(&adev->jpeg.inst->ring_dec);
+	fences += amdgpu_fence_count_emitted(adev->jpeg.inst->ring_dec);
 	fences += amdgpu_fence_count_emitted(&adev->vcn.inst->ring_dec);
 
 	if (fences == 0) {
 		amdgpu_gfx_off_ctrl(adev, true);
 		if (adev->pm.dpm_enabled)
-			amdgpu_dpm_enable_uvd(adev, false);
+			amdgpu_dpm_enable_vcn(adev, false, 0);
 		else
 			amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN,
 			       AMD_PG_STATE_GATE);
 	} else {
-		schedule_delayed_work(&adev->vcn.idle_work, VCN_IDLE_TIMEOUT);
+		schedule_delayed_work(&adev->vcn.inst[0].idle_work, VCN_IDLE_TIMEOUT);
 	}
 }
 
 static void vcn_v1_0_ring_begin_use(struct amdgpu_ring *ring)
 {
 	struct	amdgpu_device *adev = ring->adev;
-	bool set_clocks = !cancel_delayed_work_sync(&adev->vcn.idle_work);
+	bool set_clocks = !cancel_delayed_work_sync(&adev->vcn.inst[0].idle_work);
 
-	mutex_lock(&adev->vcn.vcn1_jpeg1_workaround);
+	mutex_lock(&adev->vcn.inst[0].vcn1_jpeg1_workaround);
 
-	if (amdgpu_fence_wait_empty(&ring->adev->jpeg.inst->ring_dec))
+	if (amdgpu_fence_wait_empty(ring->adev->jpeg.inst->ring_dec))
 		DRM_ERROR("VCN dec: jpeg dec ring may not be empty\n");
 
 	vcn_v1_0_set_pg_for_begin_use(ring, set_clocks);
@@ -1846,7 +1927,7 @@ void vcn_v1_0_set_pg_for_begin_use(struct amdgpu_ring *ring, bool set_clocks)
 	if (set_clocks) {
 		amdgpu_gfx_off_ctrl(adev, false);
 		if (adev->pm.dpm_enabled)
-			amdgpu_dpm_enable_uvd(adev, true);
+			amdgpu_dpm_enable_vcn(adev, true, 0);
 		else
 			amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN,
 			       AMD_PG_STATE_UNGATE);
@@ -1856,7 +1937,7 @@ void vcn_v1_0_set_pg_for_begin_use(struct amdgpu_ring *ring, bool set_clocks)
 		struct dpg_pause_state new_state;
 		unsigned int fences = 0, i;
 
-		for (i = 0; i < adev->vcn.num_enc_rings; ++i)
+		for (i = 0; i < adev->vcn.inst[0].num_enc_rings; ++i)
 			fences += amdgpu_fence_count_emitted(&adev->vcn.inst->ring_enc[i]);
 
 		if (fences)
@@ -1864,7 +1945,7 @@ void vcn_v1_0_set_pg_for_begin_use(struct amdgpu_ring *ring, bool set_clocks)
 		else
 			new_state.fw_based = VCN_DPG_STATE__UNPAUSE;
 
-		if (amdgpu_fence_count_emitted(&adev->jpeg.inst->ring_dec))
+		if (amdgpu_fence_count_emitted(adev->jpeg.inst->ring_dec))
 			new_state.jpeg = VCN_DPG_STATE__PAUSE;
 		else
 			new_state.jpeg = VCN_DPG_STATE__UNPAUSE;
@@ -1874,20 +1955,79 @@ void vcn_v1_0_set_pg_for_begin_use(struct amdgpu_ring *ring, bool set_clocks)
 		else if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_JPEG)
 			new_state.jpeg = VCN_DPG_STATE__PAUSE;
 
-		adev->vcn.pause_dpg_mode(adev, 0, &new_state);
+		adev->vcn.inst->pause_dpg_mode(adev->vcn.inst, &new_state);
 	}
 }
 
 void vcn_v1_0_ring_end_use(struct amdgpu_ring *ring)
 {
-	schedule_delayed_work(&ring->adev->vcn.idle_work, VCN_IDLE_TIMEOUT);
-	mutex_unlock(&ring->adev->vcn.vcn1_jpeg1_workaround);
+	schedule_delayed_work(&ring->adev->vcn.inst[0].idle_work, VCN_IDLE_TIMEOUT);
+	mutex_unlock(&ring->adev->vcn.inst[0].vcn1_jpeg1_workaround);
+}
+
+static void vcn_v1_0_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, j;
+	uint32_t reg_count = ARRAY_SIZE(vcn_reg_list_1_0);
+	uint32_t inst_off, is_powered;
+
+	if (!adev->vcn.ip_dump)
+		return;
+
+	drm_printf(p, "num_instances:%d\n", adev->vcn.num_vcn_inst);
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		if (adev->vcn.harvest_config & (1 << i)) {
+			drm_printf(p, "\nHarvested Instance:VCN%d Skipping dump\n", i);
+			continue;
+		}
+
+		inst_off = i * reg_count;
+		is_powered = (adev->vcn.ip_dump[inst_off] &
+				UVD_POWER_STATUS__UVD_POWER_STATUS_MASK) != 1;
+
+		if (is_powered) {
+			drm_printf(p, "\nActive Instance:VCN%d\n", i);
+			for (j = 0; j < reg_count; j++)
+				drm_printf(p, "%-50s \t 0x%08x\n", vcn_reg_list_1_0[j].reg_name,
+					   adev->vcn.ip_dump[inst_off + j]);
+		} else {
+			drm_printf(p, "\nInactive Instance:VCN%d\n", i);
+		}
+	}
+}
+
+static void vcn_v1_0_dump_ip_state(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, j;
+	bool is_powered;
+	uint32_t inst_off;
+	uint32_t reg_count = ARRAY_SIZE(vcn_reg_list_1_0);
+
+	if (!adev->vcn.ip_dump)
+		return;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		inst_off = i * reg_count;
+		/* mmUVD_POWER_STATUS is always readable and is first element of the array */
+		adev->vcn.ip_dump[inst_off] = RREG32_SOC15(VCN, i, mmUVD_POWER_STATUS);
+		is_powered = (adev->vcn.ip_dump[inst_off] &
+				UVD_POWER_STATUS__UVD_POWER_STATUS_MASK) != 1;
+
+		if (is_powered)
+			for (j = 1; j < reg_count; j++)
+				adev->vcn.ip_dump[inst_off + j] =
+					RREG32(SOC15_REG_ENTRY_OFFSET_INST(vcn_reg_list_1_0[j], i));
+	}
 }
 
 static const struct amd_ip_funcs vcn_v1_0_ip_funcs = {
 	.name = "vcn_v1_0",
 	.early_init = vcn_v1_0_early_init,
-	.late_init = NULL,
 	.sw_init = vcn_v1_0_sw_init,
 	.sw_fini = vcn_v1_0_sw_fini,
 	.hw_init = vcn_v1_0_hw_init,
@@ -1896,12 +2036,10 @@ static const struct amd_ip_funcs vcn_v1_0_ip_funcs = {
 	.resume = vcn_v1_0_resume,
 	.is_idle = vcn_v1_0_is_idle,
 	.wait_for_idle = vcn_v1_0_wait_for_idle,
-	.check_soft_reset = NULL /* vcn_v1_0_check_soft_reset */,
-	.pre_soft_reset = NULL /* vcn_v1_0_pre_soft_reset */,
-	.soft_reset = NULL /* vcn_v1_0_soft_reset */,
-	.post_soft_reset = NULL /* vcn_v1_0_post_soft_reset */,
 	.set_clockgating_state = vcn_v1_0_set_clockgating_state,
-	.set_powergating_state = vcn_v1_0_set_powergating_state,
+	.set_powergating_state = vcn_set_powergating_state,
+	.dump_ip_state = vcn_v1_0_dump_ip_state,
+	.print_ip_state = vcn_v1_0_print_ip_state,
 };
 
 /*
@@ -1958,11 +2096,11 @@ static int vcn_v1_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p,
 		uint32_t reg = amdgpu_ib_get_value(ib, i);
 		uint32_t val = amdgpu_ib_get_value(ib, i + 1);
 
-		if (reg == PACKET0(p->adev->vcn.internal.data0, 0)) {
+		if (reg == PACKET0(p->adev->vcn.inst[0].internal.data0, 0)) {
 			msg_lo = val;
-		} else if (reg == PACKET0(p->adev->vcn.internal.data1, 0)) {
+		} else if (reg == PACKET0(p->adev->vcn.inst[0].internal.data1, 0)) {
 			msg_hi = val;
-		} else if (reg == PACKET0(p->adev->vcn.internal.cmd, 0)) {
+		} else if (reg == PACKET0(p->adev->vcn.inst[0].internal.cmd, 0)) {
 			r = vcn_v1_0_validate_bo(p, job,
 						 ((u64)msg_hi) << 32 | msg_lo);
 			if (r)
@@ -2041,17 +2179,14 @@ static const struct amdgpu_ring_funcs vcn_v1_0_enc_ring_vm_funcs = {
 static void vcn_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev)
 {
 	adev->vcn.inst->ring_dec.funcs = &vcn_v1_0_dec_ring_vm_funcs;
-	DRM_INFO("VCN decode is enabled in VM mode\n");
 }
 
 static void vcn_v1_0_set_enc_ring_funcs(struct amdgpu_device *adev)
 {
 	int i;
 
-	for (i = 0; i < adev->vcn.num_enc_rings; ++i)
+	for (i = 0; i < adev->vcn.inst[0].num_enc_rings; ++i)
 		adev->vcn.inst->ring_enc[i].funcs = &vcn_v1_0_enc_ring_vm_funcs;
-
-	DRM_INFO("VCN encode is enabled in VM mode\n");
 }
 
 static const struct amdgpu_irq_src_funcs vcn_v1_0_irq_funcs = {
@@ -2061,12 +2196,11 @@ static const struct amdgpu_irq_src_funcs vcn_v1_0_irq_funcs = {
 
 static void vcn_v1_0_set_irq_funcs(struct amdgpu_device *adev)
 {
-	adev->vcn.inst->irq.num_types = adev->vcn.num_enc_rings + 2;
+	adev->vcn.inst->irq.num_types = adev->vcn.inst[0].num_enc_rings + 2;
 	adev->vcn.inst->irq.funcs = &vcn_v1_0_irq_funcs;
 }
 
-const struct amdgpu_ip_block_version vcn_v1_0_ip_block =
-{
+const struct amdgpu_ip_block_version vcn_v1_0_ip_block = {
 		.type = AMD_IP_BLOCK_TYPE_VCN,
 		.major = 1,
 		.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c
index 7c2b3aa48083..8897dcc9c1a0 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c
@@ -39,6 +39,7 @@
 
 #define VCN_VID_SOC_ADDRESS_2_0					0x1fa00
 #define VCN1_VID_SOC_ADDRESS_3_0				0x48200
+#define VCN1_AON_SOC_ADDRESS_3_0				0x48000
 
 #define mmUVD_CONTEXT_ID_INTERNAL_OFFSET			0x1fd
 #define mmUVD_GPCOM_VCPU_CMD_INTERNAL_OFFSET			0x503
@@ -53,51 +54,90 @@
 #define mmUVD_LMI_RBC_IB_64BIT_BAR_LOW_INTERNAL_OFFSET		0x5a7
 #define mmUVD_RBC_IB_SIZE_INTERNAL_OFFSET			0x1e2
 
+static const struct amdgpu_hwip_reg_entry vcn_reg_list_2_0[] = {
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_POWER_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_CONTEXT_ID),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_CONTEXT_ID2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_GPCOM_VCPU_DATA0),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_GPCOM_VCPU_DATA1),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_GPCOM_VCPU_CMD),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_PGFSM_CONFIG),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_PGFSM_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_LMA_CTL),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_LMA_DATA),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_LMA_MASK),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_PAUSE)
+};
+
 static void vcn_v2_0_set_dec_ring_funcs(struct amdgpu_device *adev);
 static void vcn_v2_0_set_enc_ring_funcs(struct amdgpu_device *adev);
 static void vcn_v2_0_set_irq_funcs(struct amdgpu_device *adev);
-static int vcn_v2_0_set_powergating_state(void *handle,
-				enum amd_powergating_state state);
-static int vcn_v2_0_pause_dpg_mode(struct amdgpu_device *adev,
-				int inst_idx, struct dpg_pause_state *new_state);
+static int vcn_v2_0_set_pg_state(struct amdgpu_vcn_inst *vinst,
+				 enum amd_powergating_state state);
+static int vcn_v2_0_pause_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				   struct dpg_pause_state *new_state);
 static int vcn_v2_0_start_sriov(struct amdgpu_device *adev);
+static int vcn_v2_0_reset(struct amdgpu_vcn_inst *vinst);
+
 /**
  * vcn_v2_0_early_init - set function pointers and load microcode
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Set ring and irq function pointers
  * Load microcode from filesystem
  */
-static int vcn_v2_0_early_init(void *handle)
+static int vcn_v2_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
-		adev->vcn.num_enc_rings = 1;
+		adev->vcn.inst[0].num_enc_rings = 1;
 	else
-		adev->vcn.num_enc_rings = 2;
+		adev->vcn.inst[0].num_enc_rings = 2;
 
+	adev->vcn.inst->set_pg_state = vcn_v2_0_set_pg_state;
 	vcn_v2_0_set_dec_ring_funcs(adev);
 	vcn_v2_0_set_enc_ring_funcs(adev);
 	vcn_v2_0_set_irq_funcs(adev);
 
-	return amdgpu_vcn_early_init(adev);
+	return amdgpu_vcn_early_init(adev, 0);
 }
 
 /**
  * vcn_v2_0_sw_init - sw init for VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Load firmware and sw initialization
  */
-static int vcn_v2_0_sw_init(void *handle)
+static int vcn_v2_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
 	int i, r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	volatile struct amdgpu_fw_shared *fw_shared;
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_fw_shared *fw_shared;
 
 	/* VCN DEC TRAP */
 	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
@@ -107,7 +147,7 @@ static int vcn_v2_0_sw_init(void *handle)
 		return r;
 
 	/* VCN ENC TRAP */
-	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
+	for (i = 0; i < adev->vcn.inst[0].num_enc_rings; ++i) {
 		r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
 				      i + VCN_2_0__SRCID__UVD_ENC_GENERAL_PURPOSE,
 				      &adev->vcn.inst->irq);
@@ -115,13 +155,13 @@ static int vcn_v2_0_sw_init(void *handle)
 			return r;
 	}
 
-	r = amdgpu_vcn_sw_init(adev);
+	r = amdgpu_vcn_sw_init(adev, 0);
 	if (r)
 		return r;
 
-	amdgpu_vcn_setup_ucode(adev);
+	amdgpu_vcn_setup_ucode(adev, 0);
 
-	r = amdgpu_vcn_resume(adev);
+	r = amdgpu_vcn_resume(adev, 0);
 	if (r)
 		return r;
 
@@ -129,7 +169,7 @@ static int vcn_v2_0_sw_init(void *handle)
 
 	ring->use_doorbell = true;
 	ring->doorbell_index = adev->doorbell_index.vcn.vcn_ring0_1 << 1;
-	ring->vm_hub = AMDGPU_MMHUB_0;
+	ring->vm_hub = AMDGPU_MMHUB0(0);
 
 	sprintf(ring->name, "vcn_dec");
 	r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0,
@@ -137,30 +177,30 @@ static int vcn_v2_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	adev->vcn.internal.context_id = mmUVD_CONTEXT_ID_INTERNAL_OFFSET;
-	adev->vcn.internal.ib_vmid = mmUVD_LMI_RBC_IB_VMID_INTERNAL_OFFSET;
-	adev->vcn.internal.ib_bar_low = mmUVD_LMI_RBC_IB_64BIT_BAR_LOW_INTERNAL_OFFSET;
-	adev->vcn.internal.ib_bar_high = mmUVD_LMI_RBC_IB_64BIT_BAR_HIGH_INTERNAL_OFFSET;
-	adev->vcn.internal.ib_size = mmUVD_RBC_IB_SIZE_INTERNAL_OFFSET;
-	adev->vcn.internal.gp_scratch8 = mmUVD_GP_SCRATCH8_INTERNAL_OFFSET;
+	adev->vcn.inst[0].internal.context_id = mmUVD_CONTEXT_ID_INTERNAL_OFFSET;
+	adev->vcn.inst[0].internal.ib_vmid = mmUVD_LMI_RBC_IB_VMID_INTERNAL_OFFSET;
+	adev->vcn.inst[0].internal.ib_bar_low = mmUVD_LMI_RBC_IB_64BIT_BAR_LOW_INTERNAL_OFFSET;
+	adev->vcn.inst[0].internal.ib_bar_high = mmUVD_LMI_RBC_IB_64BIT_BAR_HIGH_INTERNAL_OFFSET;
+	adev->vcn.inst[0].internal.ib_size = mmUVD_RBC_IB_SIZE_INTERNAL_OFFSET;
+	adev->vcn.inst[0].internal.gp_scratch8 = mmUVD_GP_SCRATCH8_INTERNAL_OFFSET;
 
-	adev->vcn.internal.scratch9 = mmUVD_SCRATCH9_INTERNAL_OFFSET;
+	adev->vcn.inst[0].internal.scratch9 = mmUVD_SCRATCH9_INTERNAL_OFFSET;
 	adev->vcn.inst->external.scratch9 = SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9);
-	adev->vcn.internal.data0 = mmUVD_GPCOM_VCPU_DATA0_INTERNAL_OFFSET;
+	adev->vcn.inst[0].internal.data0 = mmUVD_GPCOM_VCPU_DATA0_INTERNAL_OFFSET;
 	adev->vcn.inst->external.data0 = SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0);
-	adev->vcn.internal.data1 = mmUVD_GPCOM_VCPU_DATA1_INTERNAL_OFFSET;
+	adev->vcn.inst[0].internal.data1 = mmUVD_GPCOM_VCPU_DATA1_INTERNAL_OFFSET;
 	adev->vcn.inst->external.data1 = SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1);
-	adev->vcn.internal.cmd = mmUVD_GPCOM_VCPU_CMD_INTERNAL_OFFSET;
+	adev->vcn.inst[0].internal.cmd = mmUVD_GPCOM_VCPU_CMD_INTERNAL_OFFSET;
 	adev->vcn.inst->external.cmd = SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD);
-	adev->vcn.internal.nop = mmUVD_NO_OP_INTERNAL_OFFSET;
+	adev->vcn.inst[0].internal.nop = mmUVD_NO_OP_INTERNAL_OFFSET;
 	adev->vcn.inst->external.nop = SOC15_REG_OFFSET(UVD, 0, mmUVD_NO_OP);
 
-	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
+	for (i = 0; i < adev->vcn.inst[0].num_enc_rings; ++i) {
 		enum amdgpu_ring_priority_level hw_prio = amdgpu_vcn_get_enc_ring_prio(i);
 
 		ring = &adev->vcn.inst->ring_enc[i];
 		ring->use_doorbell = true;
-		ring->vm_hub = AMDGPU_MMHUB_0;
+		ring->vm_hub = AMDGPU_MMHUB0(0);
 		if (!amdgpu_sriov_vf(adev))
 			ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 2 + i;
 		else
@@ -172,7 +212,13 @@ static int vcn_v2_0_sw_init(void *handle)
 			return r;
 	}
 
-	adev->vcn.pause_dpg_mode = vcn_v2_0_pause_dpg_mode;
+	adev->vcn.inst[0].pause_dpg_mode = vcn_v2_0_pause_dpg_mode;
+	adev->vcn.inst[0].reset = vcn_v2_0_reset;
+
+	adev->vcn.supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->vcn.inst[0].ring_enc[0]);
+	if (!amdgpu_sriov_vf(adev))
+		adev->vcn.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
 
 	r = amdgpu_virt_alloc_mm_table(adev);
 	if (r)
@@ -184,21 +230,29 @@ static int vcn_v2_0_sw_init(void *handle)
 	if (amdgpu_vcnfw_log)
 		amdgpu_vcn_fwlog_init(adev->vcn.inst);
 
+	r = amdgpu_vcn_reg_dump_init(adev, vcn_reg_list_2_0, ARRAY_SIZE(vcn_reg_list_2_0));
+	if (r)
+		return r;
+
+	r = amdgpu_vcn_sysfs_reset_mask_init(adev);
+	if (r)
+		return r;
+
 	return 0;
 }
 
 /**
  * vcn_v2_0_sw_fini - sw fini for VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * VCN suspend and free up sw allocation
  */
-static int vcn_v2_0_sw_fini(void *handle)
+static int vcn_v2_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int r, idx;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared.cpu_addr;
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared.cpu_addr;
 
 	if (drm_dev_enter(adev_to_drm(adev), &idx)) {
 		fw_shared->present_flag_0 = 0;
@@ -207,25 +261,27 @@ static int vcn_v2_0_sw_fini(void *handle)
 
 	amdgpu_virt_free_mm_table(adev);
 
-	r = amdgpu_vcn_suspend(adev);
+	r = amdgpu_vcn_suspend(adev, 0);
 	if (r)
 		return r;
 
-	r = amdgpu_vcn_sw_fini(adev);
+	amdgpu_vcn_sysfs_reset_mask_fini(adev);
 
-	return r;
+	amdgpu_vcn_sw_fini(adev, 0);
+
+	return 0;
 }
 
 /**
  * vcn_v2_0_hw_init - start and test VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Initialize the hardware, boot up the VCPU and do some testing
  */
-static int vcn_v2_0_hw_init(void *handle)
+static int vcn_v2_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring = &adev->vcn.inst->ring_dec;
 	int i, r;
 
@@ -237,44 +293,40 @@ static int vcn_v2_0_hw_init(void *handle)
 
 	r = amdgpu_ring_test_helper(ring);
 	if (r)
-		goto done;
+		return r;
 
 	//Disable vcn decode for sriov
 	if (amdgpu_sriov_vf(adev))
 		ring->sched.ready = false;
 
-	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
+	for (i = 0; i < adev->vcn.inst[0].num_enc_rings; ++i) {
 		ring = &adev->vcn.inst->ring_enc[i];
 		r = amdgpu_ring_test_helper(ring);
 		if (r)
-			goto done;
+			return r;
 	}
 
-done:
-	if (!r)
-		DRM_INFO("VCN decode and encode initialized successfully(under %s).\n",
-			(adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)?"DPG Mode":"SPG Mode");
-
-	return r;
+	return 0;
 }
 
 /**
  * vcn_v2_0_hw_fini - stop the hardware block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Stop the VCN block, mark ring as not ready any more
  */
-static int vcn_v2_0_hw_fini(void *handle)
+static int vcn_v2_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_vcn_inst *vinst = adev->vcn.inst;
 
-	cancel_delayed_work_sync(&adev->vcn.idle_work);
+	cancel_delayed_work_sync(&vinst->idle_work);
 
 	if ((adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) ||
-	    (adev->vcn.cur_state != AMD_PG_STATE_GATE &&
-	      RREG32_SOC15(VCN, 0, mmUVD_STATUS)))
-		vcn_v2_0_set_powergating_state(adev, AMD_PG_STATE_GATE);
+	    (vinst->cur_state != AMD_PG_STATE_GATE &&
+	     RREG32_SOC15(VCN, 0, mmUVD_STATUS)))
+		vinst->set_pg_state(vinst, AMD_PG_STATE_GATE);
 
 	return 0;
 }
@@ -282,20 +334,19 @@ static int vcn_v2_0_hw_fini(void *handle)
 /**
  * vcn_v2_0_suspend - suspend VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * HW fini and suspend VCN block
  */
-static int vcn_v2_0_suspend(void *handle)
+static int vcn_v2_0_suspend(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = vcn_v2_0_hw_fini(adev);
+	r = vcn_v2_0_hw_fini(ip_block);
 	if (r)
 		return r;
 
-	r = amdgpu_vcn_suspend(adev);
+	r = amdgpu_vcn_suspend(ip_block->adev, 0);
 
 	return r;
 }
@@ -303,20 +354,19 @@ static int vcn_v2_0_suspend(void *handle)
 /**
  * vcn_v2_0_resume - resume VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Resume firmware and hw init VCN block
  */
-static int vcn_v2_0_resume(void *handle)
+static int vcn_v2_0_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = amdgpu_vcn_resume(adev);
+	r = amdgpu_vcn_resume(ip_block->adev, 0);
 	if (r)
 		return r;
 
-	r = vcn_v2_0_hw_init(adev);
+	r = vcn_v2_0_hw_init(ip_block);
 
 	return r;
 }
@@ -324,13 +374,14 @@ static int vcn_v2_0_resume(void *handle)
 /**
  * vcn_v2_0_mc_resume - memory controller programming
  *
- * @adev: amdgpu_device pointer
+ * @vinst: Pointer to the VCN instance structure
  *
  * Let the VCN memory controller know it's offsets
  */
-static void vcn_v2_0_mc_resume(struct amdgpu_device *adev)
+static void vcn_v2_0_mc_resume(struct amdgpu_vcn_inst *vinst)
 {
-	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+	struct amdgpu_device *adev = vinst->adev;
+	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.inst[0].fw->size + 4);
 	uint32_t offset;
 
 	if (amdgpu_sriov_vf(adev))
@@ -384,9 +435,11 @@ static void vcn_v2_0_mc_resume(struct amdgpu_device *adev)
 	WREG32_SOC15(UVD, 0, mmUVD_GFX10_ADDR_CONFIG, adev->gfx.config.gb_addr_config);
 }
 
-static void vcn_v2_0_mc_resume_dpg_mode(struct amdgpu_device *adev, bool indirect)
+static void vcn_v2_0_mc_resume_dpg_mode(struct amdgpu_vcn_inst *vinst,
+					bool indirect)
 {
-	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+	struct amdgpu_device *adev = vinst->adev;
+	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.inst[0].fw->size + 4);
 	uint32_t offset;
 
 	/* cache window 0: fw */
@@ -483,12 +536,13 @@ static void vcn_v2_0_mc_resume_dpg_mode(struct amdgpu_device *adev, bool indirec
 /**
  * vcn_v2_0_disable_clock_gating - disable VCN clock gating
  *
- * @adev: amdgpu_device pointer
+ * @vinst: VCN instance
  *
  * Disable clock gating for VCN block
  */
-static void vcn_v2_0_disable_clock_gating(struct amdgpu_device *adev)
+static void vcn_v2_0_disable_clock_gating(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
 	uint32_t data;
 
 	if (amdgpu_sriov_vf(adev))
@@ -592,9 +646,10 @@ static void vcn_v2_0_disable_clock_gating(struct amdgpu_device *adev)
 	WREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL, data);
 }
 
-static void vcn_v2_0_clock_gating_dpg_mode(struct amdgpu_device *adev,
+static void vcn_v2_0_clock_gating_dpg_mode(struct amdgpu_vcn_inst *vinst,
 		uint8_t sram_sel, uint8_t indirect)
 {
+	struct amdgpu_device *adev = vinst->adev;
 	uint32_t reg_data = 0;
 
 	/* enable sw clock gating control */
@@ -643,12 +698,13 @@ static void vcn_v2_0_clock_gating_dpg_mode(struct amdgpu_device *adev,
 /**
  * vcn_v2_0_enable_clock_gating - enable VCN clock gating
  *
- * @adev: amdgpu_device pointer
+ * @vinst: VCN instance
  *
  * Enable clock gating for VCN block
  */
-static void vcn_v2_0_enable_clock_gating(struct amdgpu_device *adev)
+static void vcn_v2_0_enable_clock_gating(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
 	uint32_t data = 0;
 
 	if (amdgpu_sriov_vf(adev))
@@ -701,8 +757,9 @@ static void vcn_v2_0_enable_clock_gating(struct amdgpu_device *adev)
 	WREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL, data);
 }
 
-static void vcn_v2_0_disable_static_power_gating(struct amdgpu_device *adev)
+static void vcn_v2_0_disable_static_power_gating(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
 	uint32_t data = 0;
 
 	if (amdgpu_sriov_vf(adev))
@@ -750,8 +807,9 @@ static void vcn_v2_0_disable_static_power_gating(struct amdgpu_device *adev)
 	WREG32_SOC15(VCN, 0, mmUVD_POWER_STATUS, data);
 }
 
-static void vcn_v2_0_enable_static_power_gating(struct amdgpu_device *adev)
+static void vcn_v2_0_enable_static_power_gating(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
 	uint32_t data = 0;
 
 	if (amdgpu_sriov_vf(adev))
@@ -792,13 +850,15 @@ static void vcn_v2_0_enable_static_power_gating(struct amdgpu_device *adev)
 	}
 }
 
-static int vcn_v2_0_start_dpg_mode(struct amdgpu_device *adev, bool indirect)
+static int vcn_v2_0_start_dpg_mode(struct amdgpu_vcn_inst *vinst, bool indirect)
 {
-	volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared.cpu_addr;
+	struct amdgpu_device *adev = vinst->adev;
+	struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared.cpu_addr;
 	struct amdgpu_ring *ring = &adev->vcn.inst->ring_dec;
 	uint32_t rb_bufsz, tmp;
+	int ret;
 
-	vcn_v2_0_enable_static_power_gating(adev);
+	vcn_v2_0_enable_static_power_gating(vinst);
 
 	/* enable dynamic power gating mode */
 	tmp = RREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS);
@@ -810,7 +870,7 @@ static int vcn_v2_0_start_dpg_mode(struct amdgpu_device *adev, bool indirect)
 		adev->vcn.inst->dpg_sram_curr_addr = (uint32_t *)adev->vcn.inst->dpg_sram_cpu_addr;
 
 	/* enable clock gating */
-	vcn_v2_0_clock_gating_dpg_mode(adev, 0, indirect);
+	vcn_v2_0_clock_gating_dpg_mode(vinst, 0, indirect);
 
 	/* enable VCPU clock */
 	tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
@@ -859,7 +919,7 @@ static int vcn_v2_0_start_dpg_mode(struct amdgpu_device *adev, bool indirect)
 		 (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
 		 (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)), 0, indirect);
 
-	vcn_v2_0_mc_resume_dpg_mode(adev, indirect);
+	vcn_v2_0_mc_resume_dpg_mode(vinst, indirect);
 
 	WREG32_SOC15_DPG_MODE(0, SOC15_DPG_MODE_OFFSET(
 		UVD, 0, mmUVD_REG_XX_MASK), 0x10, 0, indirect);
@@ -880,10 +940,13 @@ static int vcn_v2_0_start_dpg_mode(struct amdgpu_device *adev, bool indirect)
 		UVD, 0, mmUVD_MASTINT_EN),
 		UVD_MASTINT_EN__VCPU_EN_MASK, 0, indirect);
 
-	if (indirect)
-		psp_update_vcn_sram(adev, 0, adev->vcn.inst->dpg_sram_gpu_addr,
-				    (uint32_t)((uintptr_t)adev->vcn.inst->dpg_sram_curr_addr -
-					       (uintptr_t)adev->vcn.inst->dpg_sram_cpu_addr));
+	if (indirect) {
+		ret = amdgpu_vcn_psp_update_sram(adev, 0, 0);
+		if (ret) {
+			dev_err(adev->dev, "vcn sram load failed %d\n", ret);
+			return ret;
+		}
+	}
 
 	/* force RBC into idle state */
 	rb_bufsz = order_base_2(ring->ring_size);
@@ -926,31 +989,38 @@ static int vcn_v2_0_start_dpg_mode(struct amdgpu_device *adev, bool indirect)
 	/* Unstall DPG */
 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_POWER_STATUS),
 		0, ~UVD_POWER_STATUS__STALL_DPG_POWER_UP_MASK);
+
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(UVD, 0, mmUVD_STATUS);
+
 	return 0;
 }
 
-static int vcn_v2_0_start(struct amdgpu_device *adev)
+static int vcn_v2_0_start(struct amdgpu_vcn_inst *vinst)
 {
-	volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared.cpu_addr;
+	struct amdgpu_device *adev = vinst->adev;
+	struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared.cpu_addr;
 	struct amdgpu_ring *ring = &adev->vcn.inst->ring_dec;
 	uint32_t rb_bufsz, tmp;
 	uint32_t lmi_swap_cntl;
 	int i, j, r;
 
 	if (adev->pm.dpm_enabled)
-		amdgpu_dpm_enable_uvd(adev, true);
+		amdgpu_dpm_enable_vcn(adev, true, 0);
 
 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
-		return vcn_v2_0_start_dpg_mode(adev, adev->vcn.indirect_sram);
+		return vcn_v2_0_start_dpg_mode(vinst, adev->vcn.inst->indirect_sram);
 
-	vcn_v2_0_disable_static_power_gating(adev);
+	vcn_v2_0_disable_static_power_gating(vinst);
 
 	/* set uvd status busy */
 	tmp = RREG32_SOC15(UVD, 0, mmUVD_STATUS) | UVD_STATUS__UVD_BUSY;
 	WREG32_SOC15(UVD, 0, mmUVD_STATUS, tmp);
 
 	/*SW clock gating */
-	vcn_v2_0_disable_clock_gating(adev);
+	vcn_v2_0_disable_clock_gating(vinst);
 
 	/* enable VCPU clock */
 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CNTL),
@@ -994,7 +1064,7 @@ static int vcn_v2_0_start(struct amdgpu_device *adev)
 		(0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
 		(0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)));
 
-	vcn_v2_0_mc_resume(adev);
+	vcn_v2_0_mc_resume(vinst);
 
 	/* release VCPU reset to boot */
 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), 0,
@@ -1099,15 +1169,21 @@ static int vcn_v2_0_start(struct amdgpu_device *adev)
 	WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE2, ring->ring_size / 4);
 	fw_shared->multi_queue.encode_lowlatency_queue_mode &= ~FW_QUEUE_RING_RESET;
 
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(UVD, 0, mmUVD_STATUS);
+
 	return 0;
 }
 
-static int vcn_v2_0_stop_dpg_mode(struct amdgpu_device *adev)
+static int vcn_v2_0_stop_dpg_mode(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
 	struct dpg_pause_state state = {.fw_based = VCN_DPG_STATE__UNPAUSE};
 	uint32_t tmp;
 
-	vcn_v2_0_pause_dpg_mode(adev, 0, &state);
+	vcn_v2_0_pause_dpg_mode(vinst, &state);
 	/* Wait for power status to be 1 */
 	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, 1,
 		UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
@@ -1129,16 +1205,22 @@ static int vcn_v2_0_stop_dpg_mode(struct amdgpu_device *adev)
 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_POWER_STATUS), 0,
 			~UVD_POWER_STATUS__UVD_PG_MODE_MASK);
 
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(UVD, 0, mmUVD_STATUS);
+
 	return 0;
 }
 
-static int vcn_v2_0_stop(struct amdgpu_device *adev)
+static int vcn_v2_0_stop(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
 	uint32_t tmp;
 	int r;
 
 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
-		r = vcn_v2_0_stop_dpg_mode(adev);
+		r = vcn_v2_0_stop_dpg_mode(vinst);
 		if (r)
 			return r;
 		goto power_off;
@@ -1190,19 +1272,26 @@ static int vcn_v2_0_stop(struct amdgpu_device *adev)
 	/* clear status */
 	WREG32_SOC15(VCN, 0, mmUVD_STATUS, 0);
 
-	vcn_v2_0_enable_clock_gating(adev);
-	vcn_v2_0_enable_static_power_gating(adev);
+	vcn_v2_0_enable_clock_gating(vinst);
+	vcn_v2_0_enable_static_power_gating(vinst);
+
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, 0, mmUVD_STATUS);
 
 power_off:
 	if (adev->pm.dpm_enabled)
-		amdgpu_dpm_enable_uvd(adev, false);
+		amdgpu_dpm_enable_vcn(adev, false, 0);
 
 	return 0;
 }
 
-static int vcn_v2_0_pause_dpg_mode(struct amdgpu_device *adev,
-				int inst_idx, struct dpg_pause_state *new_state)
+static int vcn_v2_0_pause_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				   struct dpg_pause_state *new_state)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
 	struct amdgpu_ring *ring;
 	uint32_t reg_data = 0;
 	int ret_code;
@@ -1219,7 +1308,7 @@ static int vcn_v2_0_pause_dpg_mode(struct amdgpu_device *adev,
 				UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
 
 			if (!ret_code) {
-				volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared.cpu_addr;
+				struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared.cpu_addr;
 				/* pause DPG */
 				reg_data |= UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK;
 				WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data);
@@ -1277,16 +1366,26 @@ static int vcn_v2_0_pause_dpg_mode(struct amdgpu_device *adev,
 	return 0;
 }
 
-static bool vcn_v2_0_is_idle(void *handle)
+static int vcn_v2_0_reset(struct amdgpu_vcn_inst *vinst)
+{
+	int r;
+
+	r = vcn_v2_0_stop(vinst);
+	if (r)
+		return r;
+	return vcn_v2_0_start(vinst);
+}
+
+static bool vcn_v2_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return (RREG32_SOC15(VCN, 0, mmUVD_STATUS) == UVD_STATUS__IDLE);
 }
 
-static int vcn_v2_0_wait_for_idle(void *handle)
+static int vcn_v2_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	ret = SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_STATUS, UVD_STATUS__IDLE,
@@ -1295,10 +1394,10 @@ static int vcn_v2_0_wait_for_idle(void *handle)
 	return ret;
 }
 
-static int vcn_v2_0_set_clockgating_state(void *handle,
+static int vcn_v2_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_CG_STATE_GATE);
 
 	if (amdgpu_sriov_vf(adev))
@@ -1306,12 +1405,12 @@ static int vcn_v2_0_set_clockgating_state(void *handle,
 
 	if (enable) {
 		/* wait for STATUS to clear */
-		if (!vcn_v2_0_is_idle(handle))
+		if (!vcn_v2_0_is_idle(ip_block))
 			return -EBUSY;
-		vcn_v2_0_enable_clock_gating(adev);
+		vcn_v2_0_enable_clock_gating(&adev->vcn.inst[0]);
 	} else {
 		/* disable HW gating and enable Sw gating */
-		vcn_v2_0_disable_clock_gating(adev);
+		vcn_v2_0_disable_clock_gating(&adev->vcn.inst[0]);
 	}
 	return 0;
 }
@@ -1381,9 +1480,9 @@ void vcn_v2_0_dec_ring_insert_start(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
 
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.data0, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.data0, 0));
 	amdgpu_ring_write(ring, 0);
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.cmd, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.cmd, 0));
 	amdgpu_ring_write(ring, VCN_DEC_KMD_CMD | (VCN_DEC_CMD_PACKET_START << 1));
 }
 
@@ -1398,7 +1497,7 @@ void vcn_v2_0_dec_ring_insert_end(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
 
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.cmd, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[0].internal.cmd, 0));
 	amdgpu_ring_write(ring, VCN_DEC_KMD_CMD | (VCN_DEC_CMD_PACKET_END << 1));
 }
 
@@ -1418,7 +1517,7 @@ void vcn_v2_0_dec_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
 	WARN_ON(ring->wptr % 2 || count % 2);
 
 	for (i = 0; i < count / 2; i++) {
-		amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.nop, 0));
+		amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.nop, 0));
 		amdgpu_ring_write(ring, 0);
 	}
 }
@@ -1439,25 +1538,25 @@ void vcn_v2_0_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq,
 	struct amdgpu_device *adev = ring->adev;
 
 	WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.context_id, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.context_id, 0));
 	amdgpu_ring_write(ring, seq);
 
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.data0, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.data0, 0));
 	amdgpu_ring_write(ring, addr & 0xffffffff);
 
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.data1, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.data1, 0));
 	amdgpu_ring_write(ring, upper_32_bits(addr) & 0xff);
 
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.cmd, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.cmd, 0));
 	amdgpu_ring_write(ring, VCN_DEC_KMD_CMD | (VCN_DEC_CMD_FENCE << 1));
 
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.data0, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.data0, 0));
 	amdgpu_ring_write(ring, 0);
 
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.data1, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.data1, 0));
 	amdgpu_ring_write(ring, 0);
 
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.cmd, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.cmd, 0));
 
 	amdgpu_ring_write(ring, VCN_DEC_KMD_CMD | (VCN_DEC_CMD_TRAP << 1));
 }
@@ -1480,14 +1579,14 @@ void vcn_v2_0_dec_ring_emit_ib(struct amdgpu_ring *ring,
 	struct amdgpu_device *adev = ring->adev;
 	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
 
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.ib_vmid, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.ib_vmid, 0));
 	amdgpu_ring_write(ring, vmid);
 
-	amdgpu_ring_write(ring,	PACKET0(adev->vcn.internal.ib_bar_low, 0));
+	amdgpu_ring_write(ring,	PACKET0(adev->vcn.inst[ring->me].internal.ib_bar_low, 0));
 	amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));
-	amdgpu_ring_write(ring,	PACKET0(adev->vcn.internal.ib_bar_high, 0));
+	amdgpu_ring_write(ring,	PACKET0(adev->vcn.inst[ring->me].internal.ib_bar_high, 0));
 	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
-	amdgpu_ring_write(ring,	PACKET0(adev->vcn.internal.ib_size, 0));
+	amdgpu_ring_write(ring,	PACKET0(adev->vcn.inst[ring->me].internal.ib_size, 0));
 	amdgpu_ring_write(ring, ib->length_dw);
 }
 
@@ -1496,16 +1595,16 @@ void vcn_v2_0_dec_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
 {
 	struct amdgpu_device *adev = ring->adev;
 
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.data0, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.data0, 0));
 	amdgpu_ring_write(ring, reg << 2);
 
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.data1, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.data1, 0));
 	amdgpu_ring_write(ring, val);
 
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.gp_scratch8, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.gp_scratch8, 0));
 	amdgpu_ring_write(ring, mask);
 
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.cmd, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.cmd, 0));
 
 	amdgpu_ring_write(ring, VCN_DEC_KMD_CMD | (VCN_DEC_CMD_REG_READ_COND_WAIT << 1));
 }
@@ -1530,13 +1629,13 @@ void vcn_v2_0_dec_ring_emit_wreg(struct amdgpu_ring *ring,
 {
 	struct amdgpu_device *adev = ring->adev;
 
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.data0, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.data0, 0));
 	amdgpu_ring_write(ring, reg << 2);
 
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.data1, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.data1, 0));
 	amdgpu_ring_write(ring, val);
 
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.cmd, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.cmd, 0));
 
 	amdgpu_ring_write(ring, VCN_DEC_KMD_CMD | (VCN_DEC_CMD_WRITE_REG << 1));
 }
@@ -1737,9 +1836,9 @@ int vcn_v2_0_dec_ring_test_ring(struct amdgpu_ring *ring)
 	r = amdgpu_ring_alloc(ring, 4);
 	if (r)
 		return r;
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.cmd, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.cmd, 0));
 	amdgpu_ring_write(ring, VCN_DEC_KMD_CMD | (VCN_DEC_CMD_PACKET_START << 1));
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.scratch9, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.scratch9, 0));
 	amdgpu_ring_write(ring, 0xDEADBEEF);
 	amdgpu_ring_commit(ring);
 	for (i = 0; i < adev->usec_timeout; i++) {
@@ -1756,8 +1855,8 @@ int vcn_v2_0_dec_ring_test_ring(struct amdgpu_ring *ring)
 }
 
 
-static int vcn_v2_0_set_powergating_state(void *handle,
-					  enum amd_powergating_state state)
+static int vcn_v2_0_set_pg_state(struct amdgpu_vcn_inst *vinst,
+				 enum amd_powergating_state state)
 {
 	/* This doesn't actually powergate the VCN block.
 	 * That's done in the dpm code via the SMC.  This
@@ -1767,23 +1866,24 @@ static int vcn_v2_0_set_powergating_state(void *handle,
 	 * the smc and the hw blocks
 	 */
 	int ret;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = vinst->adev;
 
 	if (amdgpu_sriov_vf(adev)) {
-		adev->vcn.cur_state = AMD_PG_STATE_UNGATE;
+		vinst->cur_state = AMD_PG_STATE_UNGATE;
 		return 0;
 	}
 
-	if (state == adev->vcn.cur_state)
+	if (state == vinst->cur_state)
 		return 0;
 
 	if (state == AMD_PG_STATE_GATE)
-		ret = vcn_v2_0_stop(adev);
+		ret = vcn_v2_0_stop(vinst);
 	else
-		ret = vcn_v2_0_start(adev);
+		ret = vcn_v2_0_start(vinst);
 
 	if (!ret)
-		adev->vcn.cur_state = state;
+		vinst->cur_state = state;
+
 	return ret;
 }
 
@@ -1822,7 +1922,7 @@ static int vcn_v2_0_start_mmsch(struct amdgpu_device *adev,
 	adev->vcn.inst->ring_dec.wptr_old = 0;
 	vcn_v2_0_dec_ring_set_wptr(&adev->vcn.inst->ring_dec);
 
-	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
+	for (i = 0; i < adev->vcn.inst[0].num_enc_rings; ++i) {
 		adev->vcn.inst->ring_enc[i].wptr = 0;
 		adev->vcn.inst->ring_enc[i].wptr_old = 0;
 		vcn_v2_0_enc_ring_set_wptr(&adev->vcn.inst->ring_enc[i]);
@@ -1880,7 +1980,7 @@ static int vcn_v2_0_start_sriov(struct amdgpu_device *adev)
 
 		init_table += header->vcn_table_offset;
 
-		size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+		size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.inst[0].fw->size + 4);
 
 		MMSCH_V2_0_INSERT_DIRECT_RD_MOD_WT(
 			SOC15_REG_OFFSET(UVD, i, mmUVD_STATUS),
@@ -1948,7 +2048,7 @@ static int vcn_v2_0_start_sriov(struct amdgpu_device *adev)
 			SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_SIZE2),
 			AMDGPU_VCN_CONTEXT_SIZE);
 
-		for (r = 0; r < adev->vcn.num_enc_rings; ++r) {
+		for (r = 0; r < adev->vcn.inst[0].num_enc_rings; ++r) {
 			ring = &adev->vcn.inst->ring_enc[r];
 			ring->wptr = 0;
 			MMSCH_V2_0_INSERT_DIRECT_WT(
@@ -1995,7 +2095,6 @@ static int vcn_v2_0_start_sriov(struct amdgpu_device *adev)
 static const struct amd_ip_funcs vcn_v2_0_ip_funcs = {
 	.name = "vcn_v2_0",
 	.early_init = vcn_v2_0_early_init,
-	.late_init = NULL,
 	.sw_init = vcn_v2_0_sw_init,
 	.sw_fini = vcn_v2_0_sw_fini,
 	.hw_init = vcn_v2_0_hw_init,
@@ -2004,12 +2103,10 @@ static const struct amd_ip_funcs vcn_v2_0_ip_funcs = {
 	.resume = vcn_v2_0_resume,
 	.is_idle = vcn_v2_0_is_idle,
 	.wait_for_idle = vcn_v2_0_wait_for_idle,
-	.check_soft_reset = NULL,
-	.pre_soft_reset = NULL,
-	.soft_reset = NULL,
-	.post_soft_reset = NULL,
 	.set_clockgating_state = vcn_v2_0_set_clockgating_state,
-	.set_powergating_state = vcn_v2_0_set_powergating_state,
+	.set_powergating_state = vcn_set_powergating_state,
+	.dump_ip_state = amdgpu_vcn_dump_ip_state,
+	.print_ip_state = amdgpu_vcn_print_ip_state,
 };
 
 static const struct amdgpu_ring_funcs vcn_v2_0_dec_ring_vm_funcs = {
@@ -2040,6 +2137,7 @@ static const struct amdgpu_ring_funcs vcn_v2_0_dec_ring_vm_funcs = {
 	.emit_wreg = vcn_v2_0_dec_ring_emit_wreg,
 	.emit_reg_wait = vcn_v2_0_dec_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = amdgpu_vcn_ring_reset,
 };
 
 static const struct amdgpu_ring_funcs vcn_v2_0_enc_ring_vm_funcs = {
@@ -2069,22 +2167,20 @@ static const struct amdgpu_ring_funcs vcn_v2_0_enc_ring_vm_funcs = {
 	.emit_wreg = vcn_v2_0_enc_ring_emit_wreg,
 	.emit_reg_wait = vcn_v2_0_enc_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = amdgpu_vcn_ring_reset,
 };
 
 static void vcn_v2_0_set_dec_ring_funcs(struct amdgpu_device *adev)
 {
 	adev->vcn.inst->ring_dec.funcs = &vcn_v2_0_dec_ring_vm_funcs;
-	DRM_INFO("VCN decode is enabled in VM mode\n");
 }
 
 static void vcn_v2_0_set_enc_ring_funcs(struct amdgpu_device *adev)
 {
 	int i;
 
-	for (i = 0; i < adev->vcn.num_enc_rings; ++i)
+	for (i = 0; i < adev->vcn.inst[0].num_enc_rings; ++i)
 		adev->vcn.inst->ring_enc[i].funcs = &vcn_v2_0_enc_ring_vm_funcs;
-
-	DRM_INFO("VCN encode is enabled in VM mode\n");
 }
 
 static const struct amdgpu_irq_src_funcs vcn_v2_0_irq_funcs = {
@@ -2094,7 +2190,7 @@ static const struct amdgpu_irq_src_funcs vcn_v2_0_irq_funcs = {
 
 static void vcn_v2_0_set_irq_funcs(struct amdgpu_device *adev)
 {
-	adev->vcn.inst->irq.num_types = adev->vcn.num_enc_rings + 1;
+	adev->vcn.inst->irq.num_types = adev->vcn.inst[0].num_enc_rings + 1;
 	adev->vcn.inst->irq.funcs = &vcn_v2_0_irq_funcs;
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c b/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c
index ab0b45d0ead1..cebee453871c 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c
@@ -39,6 +39,7 @@
 
 #define VCN_VID_SOC_ADDRESS_2_0					0x1fa00
 #define VCN1_VID_SOC_ADDRESS_3_0				0x48200
+#define VCN1_AON_SOC_ADDRESS_3_0				0x48000
 
 #define mmUVD_CONTEXT_ID_INTERNAL_OFFSET			0x27
 #define mmUVD_GPCOM_VCPU_CMD_INTERNAL_OFFSET			0x0f
@@ -55,37 +56,181 @@
 
 #define VCN25_MAX_HW_INSTANCES_ARCTURUS			2
 
+static const struct amdgpu_hwip_reg_entry vcn_reg_list_2_5[] = {
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_POWER_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_POWER_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_CONTEXT_ID),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_CONTEXT_ID2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_GPCOM_VCPU_DATA0),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_GPCOM_VCPU_DATA1),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_GPCOM_VCPU_CMD),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_PGFSM_CONFIG),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_PGFSM_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_LMA_CTL),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_LMA_DATA),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_LMA_MASK),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_PAUSE)
+};
+
 static void vcn_v2_5_set_dec_ring_funcs(struct amdgpu_device *adev);
 static void vcn_v2_5_set_enc_ring_funcs(struct amdgpu_device *adev);
 static void vcn_v2_5_set_irq_funcs(struct amdgpu_device *adev);
-static int vcn_v2_5_set_powergating_state(void *handle,
-				enum amd_powergating_state state);
-static int vcn_v2_5_pause_dpg_mode(struct amdgpu_device *adev,
-				int inst_idx, struct dpg_pause_state *new_state);
+static int vcn_v2_5_set_pg_state(struct amdgpu_vcn_inst *vinst,
+				 enum amd_powergating_state state);
+static int vcn_v2_5_pause_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				   struct dpg_pause_state *new_state);
 static int vcn_v2_5_sriov_start(struct amdgpu_device *adev);
 static void vcn_v2_5_set_ras_funcs(struct amdgpu_device *adev);
+static int vcn_v2_5_reset(struct amdgpu_vcn_inst *vinst);
 
 static int amdgpu_ih_clientid_vcns[] = {
 	SOC15_IH_CLIENTID_VCN,
 	SOC15_IH_CLIENTID_VCN1
 };
 
+static void vcn_v2_5_idle_work_handler(struct work_struct *work)
+{
+	struct amdgpu_vcn_inst *vcn_inst =
+		container_of(work, struct amdgpu_vcn_inst, idle_work.work);
+	struct amdgpu_device *adev = vcn_inst->adev;
+	unsigned int fences = 0, fence[AMDGPU_MAX_VCN_INSTANCES] = {0};
+	unsigned int i, j;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		struct amdgpu_vcn_inst *v = &adev->vcn.inst[i];
+
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		for (j = 0; j < v->num_enc_rings; ++j)
+			fence[i] += amdgpu_fence_count_emitted(&v->ring_enc[j]);
+
+		/* Only set DPG pause for VCN3 or below, VCN4 and above will be handled by FW */
+		if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG &&
+		    !v->using_unified_queue) {
+			struct dpg_pause_state new_state;
+
+			if (fence[i] ||
+			    unlikely(atomic_read(&v->dpg_enc_submission_cnt)))
+				new_state.fw_based = VCN_DPG_STATE__PAUSE;
+			else
+				new_state.fw_based = VCN_DPG_STATE__UNPAUSE;
+
+			v->pause_dpg_mode(v, &new_state);
+		}
+
+		fence[i] += amdgpu_fence_count_emitted(&v->ring_dec);
+		fences += fence[i];
+
+	}
+
+	if (!fences && !atomic_read(&adev->vcn.inst[0].total_submission_cnt)) {
+		amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN,
+						       AMD_PG_STATE_GATE);
+		amdgpu_vcn_put_profile(adev);
+	} else {
+		schedule_delayed_work(&adev->vcn.inst[0].idle_work, VCN_IDLE_TIMEOUT);
+	}
+}
+
+static void vcn_v2_5_ring_begin_use(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_vcn_inst *v = &adev->vcn.inst[ring->me];
+
+	atomic_inc(&adev->vcn.inst[0].total_submission_cnt);
+
+	cancel_delayed_work_sync(&adev->vcn.inst[0].idle_work);
+
+	/* We can safely return early here because we've cancelled the
+	 * the delayed work so there is no one else to set it to false
+	 * and we don't care if someone else sets it to true.
+	 */
+	mutex_lock(&adev->vcn.inst[0].vcn_pg_lock);
+	amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN,
+					       AMD_PG_STATE_UNGATE);
+
+	/* Only set DPG pause for VCN3 or below, VCN4 and above will be handled by FW */
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG &&
+	    !v->using_unified_queue) {
+		struct dpg_pause_state new_state;
+
+		if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC) {
+			atomic_inc(&v->dpg_enc_submission_cnt);
+			new_state.fw_based = VCN_DPG_STATE__PAUSE;
+		} else {
+			unsigned int fences = 0;
+			unsigned int i;
+
+			for (i = 0; i < v->num_enc_rings; ++i)
+				fences += amdgpu_fence_count_emitted(&v->ring_enc[i]);
+
+			if (fences || atomic_read(&v->dpg_enc_submission_cnt))
+				new_state.fw_based = VCN_DPG_STATE__PAUSE;
+			else
+				new_state.fw_based = VCN_DPG_STATE__UNPAUSE;
+		}
+		v->pause_dpg_mode(v, &new_state);
+	}
+	mutex_unlock(&adev->vcn.inst[0].vcn_pg_lock);
+	amdgpu_vcn_get_profile(adev);
+}
+
+static void vcn_v2_5_ring_end_use(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	/* Only set DPG pause for VCN3 or below, VCN4 and above will be handled by FW */
+	if (ring->adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG &&
+	    ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC &&
+	    !adev->vcn.inst[ring->me].using_unified_queue)
+		atomic_dec(&adev->vcn.inst[ring->me].dpg_enc_submission_cnt);
+
+	atomic_dec(&adev->vcn.inst[0].total_submission_cnt);
+
+	schedule_delayed_work(&adev->vcn.inst[0].idle_work,
+			      VCN_IDLE_TIMEOUT);
+}
+
 /**
  * vcn_v2_5_early_init - set function pointers and load microcode
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Set ring and irq function pointers
  * Load microcode from filesystem
  */
-static int vcn_v2_5_early_init(void *handle)
+static int vcn_v2_5_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, r;
 
 	if (amdgpu_sriov_vf(adev)) {
 		adev->vcn.num_vcn_inst = 2;
 		adev->vcn.harvest_config = 0;
-		adev->vcn.num_enc_rings = 1;
+		for (i = 0; i < adev->vcn.num_vcn_inst; i++)
+			adev->vcn.inst[i].num_enc_rings = 1;
 	} else {
 		u32 harvest;
 		int i;
@@ -94,13 +239,12 @@ static int vcn_v2_5_early_init(void *handle)
 			harvest = RREG32_SOC15(VCN, i, mmCC_UVD_HARVESTING);
 			if (harvest & CC_UVD_HARVESTING__UVD_DISABLE_MASK)
 				adev->vcn.harvest_config |= 1 << i;
+			adev->vcn.inst[i].num_enc_rings = 2;
 		}
 		if (adev->vcn.harvest_config == (AMDGPU_VCN_HARVEST_VCN0 |
-					AMDGPU_VCN_HARVEST_VCN1))
+						 AMDGPU_VCN_HARVEST_VCN1))
 			/* both instances are harvested, disable the block */
 			return -ENOENT;
-
-		adev->vcn.num_enc_rings = 2;
 	}
 
 	vcn_v2_5_set_dec_ring_funcs(adev);
@@ -108,23 +252,33 @@ static int vcn_v2_5_early_init(void *handle)
 	vcn_v2_5_set_irq_funcs(adev);
 	vcn_v2_5_set_ras_funcs(adev);
 
-	return amdgpu_vcn_early_init(adev);
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		adev->vcn.inst[i].set_pg_state = vcn_v2_5_set_pg_state;
+
+		r = amdgpu_vcn_early_init(adev, i);
+		if (r)
+			return r;
+	}
+
+	return 0;
 }
 
 /**
  * vcn_v2_5_sw_init - sw init for VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Load firmware and sw initialization
  */
-static int vcn_v2_5_sw_init(void *handle)
+static int vcn_v2_5_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
 	int i, j, r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (j = 0; j < adev->vcn.num_vcn_inst; j++) {
+		struct amdgpu_fw_shared *fw_shared;
+
 		if (adev->vcn.harvest_config & (1 << j))
 			continue;
 		/* VCN DEC TRAP */
@@ -134,7 +288,7 @@ static int vcn_v2_5_sw_init(void *handle)
 			return r;
 
 		/* VCN ENC TRAP */
-		for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
+		for (i = 0; i < adev->vcn.inst[j].num_enc_rings; ++i) {
 			r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_vcns[j],
 				i + VCN_2_0__SRCID__UVD_ENC_GENERAL_PURPOSE, &adev->vcn.inst[j].irq);
 			if (r)
@@ -143,42 +297,39 @@ static int vcn_v2_5_sw_init(void *handle)
 
 		/* VCN POISON TRAP */
 		r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_vcns[j],
-			VCN_2_6__SRCID_UVD_POISON, &adev->vcn.inst[j].irq);
+			VCN_2_6__SRCID_UVD_POISON, &adev->vcn.inst[j].ras_poison_irq);
 		if (r)
 			return r;
-	}
 
-	r = amdgpu_vcn_sw_init(adev);
-	if (r)
-		return r;
+		r = amdgpu_vcn_sw_init(adev, j);
+		if (r)
+			return r;
 
-	amdgpu_vcn_setup_ucode(adev);
+		/* Override the work func */
+		adev->vcn.inst[j].idle_work.work.func = vcn_v2_5_idle_work_handler;
 
-	r = amdgpu_vcn_resume(adev);
-	if (r)
-		return r;
+		amdgpu_vcn_setup_ucode(adev, j);
 
-	for (j = 0; j < adev->vcn.num_vcn_inst; j++) {
-		volatile struct amdgpu_fw_shared *fw_shared;
+		r = amdgpu_vcn_resume(adev, j);
+		if (r)
+			return r;
 
-		if (adev->vcn.harvest_config & (1 << j))
-			continue;
-		adev->vcn.internal.context_id = mmUVD_CONTEXT_ID_INTERNAL_OFFSET;
-		adev->vcn.internal.ib_vmid = mmUVD_LMI_RBC_IB_VMID_INTERNAL_OFFSET;
-		adev->vcn.internal.ib_bar_low = mmUVD_LMI_RBC_IB_64BIT_BAR_LOW_INTERNAL_OFFSET;
-		adev->vcn.internal.ib_bar_high = mmUVD_LMI_RBC_IB_64BIT_BAR_HIGH_INTERNAL_OFFSET;
-		adev->vcn.internal.ib_size = mmUVD_RBC_IB_SIZE_INTERNAL_OFFSET;
-		adev->vcn.internal.gp_scratch8 = mmUVD_GP_SCRATCH8_INTERNAL_OFFSET;
-
-		adev->vcn.internal.scratch9 = mmUVD_SCRATCH9_INTERNAL_OFFSET;
+		adev->vcn.inst[j].internal.context_id = mmUVD_CONTEXT_ID_INTERNAL_OFFSET;
+		adev->vcn.inst[j].internal.ib_vmid = mmUVD_LMI_RBC_IB_VMID_INTERNAL_OFFSET;
+		adev->vcn.inst[j].internal.ib_bar_low = mmUVD_LMI_RBC_IB_64BIT_BAR_LOW_INTERNAL_OFFSET;
+		adev->vcn.inst[j].internal.ib_bar_high = mmUVD_LMI_RBC_IB_64BIT_BAR_HIGH_INTERNAL_OFFSET;
+		adev->vcn.inst[j].internal.ib_size = mmUVD_RBC_IB_SIZE_INTERNAL_OFFSET;
+		adev->vcn.inst[j].internal.gp_scratch8 = mmUVD_GP_SCRATCH8_INTERNAL_OFFSET;
+
+		adev->vcn.inst[j].internal.scratch9 = mmUVD_SCRATCH9_INTERNAL_OFFSET;
 		adev->vcn.inst[j].external.scratch9 = SOC15_REG_OFFSET(VCN, j, mmUVD_SCRATCH9);
-		adev->vcn.internal.data0 = mmUVD_GPCOM_VCPU_DATA0_INTERNAL_OFFSET;
+		adev->vcn.inst[j].internal.data0 = mmUVD_GPCOM_VCPU_DATA0_INTERNAL_OFFSET;
 		adev->vcn.inst[j].external.data0 = SOC15_REG_OFFSET(VCN, j, mmUVD_GPCOM_VCPU_DATA0);
-		adev->vcn.internal.data1 = mmUVD_GPCOM_VCPU_DATA1_INTERNAL_OFFSET;
+		adev->vcn.inst[j].internal.data1 = mmUVD_GPCOM_VCPU_DATA1_INTERNAL_OFFSET;
 		adev->vcn.inst[j].external.data1 = SOC15_REG_OFFSET(VCN, j, mmUVD_GPCOM_VCPU_DATA1);
-		adev->vcn.internal.cmd = mmUVD_GPCOM_VCPU_CMD_INTERNAL_OFFSET;
+		adev->vcn.inst[j].internal.cmd = mmUVD_GPCOM_VCPU_CMD_INTERNAL_OFFSET;
 		adev->vcn.inst[j].external.cmd = SOC15_REG_OFFSET(VCN, j, mmUVD_GPCOM_VCPU_CMD);
-		adev->vcn.internal.nop = mmUVD_NO_OP_INTERNAL_OFFSET;
+		adev->vcn.inst[j].internal.nop = mmUVD_NO_OP_INTERNAL_OFFSET;
 		adev->vcn.inst[j].external.nop = SOC15_REG_OFFSET(VCN, j, mmUVD_NO_OP);
 
 		ring = &adev->vcn.inst[j].ring_dec;
@@ -187,10 +338,10 @@ static int vcn_v2_5_sw_init(void *handle)
 		ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
 				(amdgpu_sriov_vf(adev) ? 2*j : 8*j);
 
-		if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(2, 5, 0))
-			ring->vm_hub = AMDGPU_MMHUB_1;
+		if (amdgpu_ip_version(adev, UVD_HWIP, 0) == IP_VERSION(2, 5, 0))
+			ring->vm_hub = AMDGPU_MMHUB1(0);
 		else
-			ring->vm_hub = AMDGPU_MMHUB_0;
+			ring->vm_hub = AMDGPU_MMHUB0(0);
 
 		sprintf(ring->name, "vcn_dec_%d", j);
 		r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[j].irq,
@@ -198,7 +349,7 @@ static int vcn_v2_5_sw_init(void *handle)
 		if (r)
 			return r;
 
-		for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
+		for (i = 0; i < adev->vcn.inst[j].num_enc_rings; ++i) {
 			enum amdgpu_ring_priority_level hw_prio = amdgpu_vcn_get_enc_ring_prio(i);
 
 			ring = &adev->vcn.inst[j].ring_enc[i];
@@ -207,10 +358,11 @@ static int vcn_v2_5_sw_init(void *handle)
 			ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
 					(amdgpu_sriov_vf(adev) ? (1 + i + 2*j) : (2 + i + 8*j));
 
-			if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(2, 5, 0))
-				ring->vm_hub = AMDGPU_MMHUB_1;
+			if (amdgpu_ip_version(adev, UVD_HWIP, 0) ==
+			    IP_VERSION(2, 5, 0))
+				ring->vm_hub = AMDGPU_MMHUB1(0);
 			else
-				ring->vm_hub = AMDGPU_MMHUB_0;
+				ring->vm_hub = AMDGPU_MMHUB0(0);
 
 			sprintf(ring->name, "vcn_enc_%d.%d", j, i);
 			r = amdgpu_ring_init(adev, ring, 512,
@@ -225,36 +377,50 @@ static int vcn_v2_5_sw_init(void *handle)
 
 		if (amdgpu_vcnfw_log)
 			amdgpu_vcn_fwlog_init(&adev->vcn.inst[i]);
+
+		if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
+			adev->vcn.inst[j].pause_dpg_mode = vcn_v2_5_pause_dpg_mode;
+		adev->vcn.inst[j].reset = vcn_v2_5_reset;
 	}
 
+	adev->vcn.supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->vcn.inst[0].ring_enc[0]);
+	if (!amdgpu_sriov_vf(adev))
+		adev->vcn.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+
 	if (amdgpu_sriov_vf(adev)) {
 		r = amdgpu_virt_alloc_mm_table(adev);
 		if (r)
 			return r;
 	}
 
-	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
-		adev->vcn.pause_dpg_mode = vcn_v2_5_pause_dpg_mode;
-
 	r = amdgpu_vcn_ras_sw_init(adev);
 	if (r)
 		return r;
 
+	r = amdgpu_vcn_reg_dump_init(adev, vcn_reg_list_2_5, ARRAY_SIZE(vcn_reg_list_2_5));
+	if (r)
+		return r;
+
+	r = amdgpu_vcn_sysfs_reset_mask_init(adev);
+	if (r)
+		return r;
+
 	return 0;
 }
 
 /**
  * vcn_v2_5_sw_fini - sw fini for VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * VCN suspend and free up sw allocation
  */
-static int vcn_v2_5_sw_fini(void *handle)
+static int vcn_v2_5_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int i, r, idx;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	volatile struct amdgpu_fw_shared *fw_shared;
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_fw_shared *fw_shared;
 
 	if (drm_dev_enter(adev_to_drm(adev), &idx)) {
 		for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
@@ -270,25 +436,28 @@ static int vcn_v2_5_sw_fini(void *handle)
 	if (amdgpu_sriov_vf(adev))
 		amdgpu_virt_free_mm_table(adev);
 
-	r = amdgpu_vcn_suspend(adev);
-	if (r)
-		return r;
+	amdgpu_vcn_sysfs_reset_mask_fini(adev);
 
-	r = amdgpu_vcn_sw_fini(adev);
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_suspend(adev, i);
+		if (r)
+			return r;
+		amdgpu_vcn_sw_fini(adev, i);
+	}
 
-	return r;
+	return 0;
 }
 
 /**
  * vcn_v2_5_hw_init - start and test VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Initialize the hardware, boot up the VCPU and do some testing
  */
-static int vcn_v2_5_hw_init(void *handle)
+static int vcn_v2_5_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring;
 	int i, j, r = 0;
 
@@ -313,47 +482,47 @@ static int vcn_v2_5_hw_init(void *handle)
 
 			r = amdgpu_ring_test_helper(ring);
 			if (r)
-				goto done;
+				return r;
 
-			for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
+			for (i = 0; i < adev->vcn.inst[j].num_enc_rings; ++i) {
 				ring = &adev->vcn.inst[j].ring_enc[i];
 				r = amdgpu_ring_test_helper(ring);
 				if (r)
-					goto done;
+					return r;
 			}
 		}
 	}
 
-done:
-	if (!r)
-		DRM_INFO("VCN decode and encode initialized successfully(under %s).\n",
-			(adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)?"DPG Mode":"SPG Mode");
-
 	return r;
 }
 
 /**
  * vcn_v2_5_hw_fini - stop the hardware block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Stop the VCN block, mark ring as not ready any more
  */
-static int vcn_v2_5_hw_fini(void *handle)
+static int vcn_v2_5_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
-	cancel_delayed_work_sync(&adev->vcn.idle_work);
-
 	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[i];
+
 		if (adev->vcn.harvest_config & (1 << i))
 			continue;
 
+		cancel_delayed_work_sync(&vinst->idle_work);
+
 		if ((adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) ||
-		    (adev->vcn.cur_state != AMD_PG_STATE_GATE &&
+		    (vinst->cur_state != AMD_PG_STATE_GATE &&
 		     RREG32_SOC15(VCN, i, mmUVD_STATUS)))
-			vcn_v2_5_set_powergating_state(adev, AMD_PG_STATE_GATE);
+			vinst->set_pg_state(vinst, AMD_PG_STATE_GATE);
+
+		if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__VCN))
+			amdgpu_irq_put(adev, &vinst->ras_poison_irq, 0);
 	}
 
 	return 0;
@@ -362,41 +531,47 @@ static int vcn_v2_5_hw_fini(void *handle)
 /**
  * vcn_v2_5_suspend - suspend VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * HW fini and suspend VCN block
  */
-static int vcn_v2_5_suspend(void *handle)
+static int vcn_v2_5_suspend(struct amdgpu_ip_block *ip_block)
 {
-	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int r, i;
 
-	r = vcn_v2_5_hw_fini(adev);
+	r = vcn_v2_5_hw_fini(ip_block);
 	if (r)
 		return r;
 
-	r = amdgpu_vcn_suspend(adev);
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_suspend(ip_block->adev, i);
+		if (r)
+			return r;
+	}
 
-	return r;
+	return 0;
 }
 
 /**
  * vcn_v2_5_resume - resume VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Resume firmware and hw init VCN block
  */
-static int vcn_v2_5_resume(void *handle)
+static int vcn_v2_5_resume(struct amdgpu_ip_block *ip_block)
 {
-	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int r, i;
 
-	r = amdgpu_vcn_resume(adev);
-	if (r)
-		return r;
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_resume(ip_block->adev, i);
+		if (r)
+			return r;
+	}
 
-	r = vcn_v2_5_hw_init(adev);
+	r = vcn_v2_5_hw_init(ip_block);
 
 	return r;
 }
@@ -404,68 +579,72 @@ static int vcn_v2_5_resume(void *handle)
 /**
  * vcn_v2_5_mc_resume - memory controller programming
  *
- * @adev: amdgpu_device pointer
+ * @vinst: VCN instance
  *
  * Let the VCN memory controller know it's offsets
  */
-static void vcn_v2_5_mc_resume(struct amdgpu_device *adev)
+static void vcn_v2_5_mc_resume(struct amdgpu_vcn_inst *vinst)
 {
-	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+	struct amdgpu_device *adev = vinst->adev;
+	int i = vinst->inst;
+	uint32_t size;
 	uint32_t offset;
-	int i;
 
-	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
-		if (adev->vcn.harvest_config & (1 << i))
-			continue;
-		/* cache window 0: fw */
-		if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
-			WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
-				(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + i].tmr_mc_addr_lo));
-			WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
-				(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + i].tmr_mc_addr_hi));
-			WREG32_SOC15(VCN, i, mmUVD_VCPU_CACHE_OFFSET0, 0);
-			offset = 0;
-		} else {
-			WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
-				lower_32_bits(adev->vcn.inst[i].gpu_addr));
-			WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
-				upper_32_bits(adev->vcn.inst[i].gpu_addr));
-			offset = size;
-			WREG32_SOC15(VCN, i, mmUVD_VCPU_CACHE_OFFSET0,
-				AMDGPU_UVD_FIRMWARE_OFFSET >> 3);
-		}
-		WREG32_SOC15(VCN, i, mmUVD_VCPU_CACHE_SIZE0, size);
+	if (adev->vcn.harvest_config & (1 << i))
+		return;
 
-		/* cache window 1: stack */
-		WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW,
-			lower_32_bits(adev->vcn.inst[i].gpu_addr + offset));
-		WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH,
-			upper_32_bits(adev->vcn.inst[i].gpu_addr + offset));
-		WREG32_SOC15(VCN, i, mmUVD_VCPU_CACHE_OFFSET1, 0);
-		WREG32_SOC15(VCN, i, mmUVD_VCPU_CACHE_SIZE1, AMDGPU_VCN_STACK_SIZE);
-
-		/* cache window 2: context */
-		WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW,
-			lower_32_bits(adev->vcn.inst[i].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE));
-		WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH,
-			upper_32_bits(adev->vcn.inst[i].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE));
-		WREG32_SOC15(VCN, i, mmUVD_VCPU_CACHE_OFFSET2, 0);
-		WREG32_SOC15(VCN, i, mmUVD_VCPU_CACHE_SIZE2, AMDGPU_VCN_CONTEXT_SIZE);
-
-		/* non-cache window */
-		WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_NC0_64BIT_BAR_LOW,
-			lower_32_bits(adev->vcn.inst[i].fw_shared.gpu_addr));
-		WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH,
-			upper_32_bits(adev->vcn.inst[i].fw_shared.gpu_addr));
-		WREG32_SOC15(VCN, i, mmUVD_VCPU_NONCACHE_OFFSET0, 0);
-		WREG32_SOC15(VCN, i, mmUVD_VCPU_NONCACHE_SIZE0,
-			AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_fw_shared)));
+	size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.inst[i].fw->size + 4);
+	/* cache window 0: fw */
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
+			     (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + i].tmr_mc_addr_lo));
+		WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
+			     (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + i].tmr_mc_addr_hi));
+		WREG32_SOC15(VCN, i, mmUVD_VCPU_CACHE_OFFSET0, 0);
+		offset = 0;
+	} else {
+		WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
+			     lower_32_bits(adev->vcn.inst[i].gpu_addr));
+		WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
+			     upper_32_bits(adev->vcn.inst[i].gpu_addr));
+		offset = size;
+		WREG32_SOC15(VCN, i, mmUVD_VCPU_CACHE_OFFSET0,
+			     AMDGPU_UVD_FIRMWARE_OFFSET >> 3);
 	}
+	WREG32_SOC15(VCN, i, mmUVD_VCPU_CACHE_SIZE0, size);
+
+	/* cache window 1: stack */
+	WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW,
+		     lower_32_bits(adev->vcn.inst[i].gpu_addr + offset));
+	WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH,
+		     upper_32_bits(adev->vcn.inst[i].gpu_addr + offset));
+	WREG32_SOC15(VCN, i, mmUVD_VCPU_CACHE_OFFSET1, 0);
+	WREG32_SOC15(VCN, i, mmUVD_VCPU_CACHE_SIZE1, AMDGPU_VCN_STACK_SIZE);
+
+	/* cache window 2: context */
+	WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW,
+		     lower_32_bits(adev->vcn.inst[i].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE));
+	WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH,
+		     upper_32_bits(adev->vcn.inst[i].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE));
+	WREG32_SOC15(VCN, i, mmUVD_VCPU_CACHE_OFFSET2, 0);
+	WREG32_SOC15(VCN, i, mmUVD_VCPU_CACHE_SIZE2, AMDGPU_VCN_CONTEXT_SIZE);
+
+	/* non-cache window */
+	WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_NC0_64BIT_BAR_LOW,
+		     lower_32_bits(adev->vcn.inst[i].fw_shared.gpu_addr));
+	WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH,
+		     upper_32_bits(adev->vcn.inst[i].fw_shared.gpu_addr));
+	WREG32_SOC15(VCN, i, mmUVD_VCPU_NONCACHE_OFFSET0, 0);
+	WREG32_SOC15(VCN, i, mmUVD_VCPU_NONCACHE_SIZE0,
+		     AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_fw_shared)));
 }
 
-static void vcn_v2_5_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_idx, bool indirect)
+static void vcn_v2_5_mc_resume_dpg_mode(struct amdgpu_vcn_inst *vinst,
+					bool indirect)
 {
-	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.inst[inst_idx].fw->size + 4);
 	uint32_t offset;
 
 	/* cache window 0: fw */
@@ -562,123 +741,124 @@ static void vcn_v2_5_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_idx
 /**
  * vcn_v2_5_disable_clock_gating - disable VCN clock gating
  *
- * @adev: amdgpu_device pointer
+ * @vinst: VCN instance
  *
  * Disable clock gating for VCN block
  */
-static void vcn_v2_5_disable_clock_gating(struct amdgpu_device *adev)
+static void vcn_v2_5_disable_clock_gating(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int i = vinst->inst;
 	uint32_t data;
-	int i;
 
-	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
-		if (adev->vcn.harvest_config & (1 << i))
-			continue;
-		/* UVD disable CGC */
-		data = RREG32_SOC15(VCN, i, mmUVD_CGC_CTRL);
-		if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
-			data |= 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
-		else
-			data &= ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK;
-		data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
-		data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
-		WREG32_SOC15(VCN, i, mmUVD_CGC_CTRL, data);
-
-		data = RREG32_SOC15(VCN, i, mmUVD_CGC_GATE);
-		data &= ~(UVD_CGC_GATE__SYS_MASK
-			| UVD_CGC_GATE__UDEC_MASK
-			| UVD_CGC_GATE__MPEG2_MASK
-			| UVD_CGC_GATE__REGS_MASK
-			| UVD_CGC_GATE__RBC_MASK
-			| UVD_CGC_GATE__LMI_MC_MASK
-			| UVD_CGC_GATE__LMI_UMC_MASK
-			| UVD_CGC_GATE__IDCT_MASK
-			| UVD_CGC_GATE__MPRD_MASK
-			| UVD_CGC_GATE__MPC_MASK
-			| UVD_CGC_GATE__LBSI_MASK
-			| UVD_CGC_GATE__LRBBM_MASK
-			| UVD_CGC_GATE__UDEC_RE_MASK
-			| UVD_CGC_GATE__UDEC_CM_MASK
-			| UVD_CGC_GATE__UDEC_IT_MASK
-			| UVD_CGC_GATE__UDEC_DB_MASK
-			| UVD_CGC_GATE__UDEC_MP_MASK
-			| UVD_CGC_GATE__WCB_MASK
-			| UVD_CGC_GATE__VCPU_MASK
-			| UVD_CGC_GATE__MMSCH_MASK);
-
-		WREG32_SOC15(VCN, i, mmUVD_CGC_GATE, data);
-
-		SOC15_WAIT_ON_RREG(VCN, i, mmUVD_CGC_GATE, 0,  0xFFFFFFFF);
-
-		data = RREG32_SOC15(VCN, i, mmUVD_CGC_CTRL);
-		data &= ~(UVD_CGC_CTRL__UDEC_RE_MODE_MASK
-			| UVD_CGC_CTRL__UDEC_CM_MODE_MASK
-			| UVD_CGC_CTRL__UDEC_IT_MODE_MASK
-			| UVD_CGC_CTRL__UDEC_DB_MODE_MASK
-			| UVD_CGC_CTRL__UDEC_MP_MODE_MASK
-			| UVD_CGC_CTRL__SYS_MODE_MASK
-			| UVD_CGC_CTRL__UDEC_MODE_MASK
-			| UVD_CGC_CTRL__MPEG2_MODE_MASK
-			| UVD_CGC_CTRL__REGS_MODE_MASK
-			| UVD_CGC_CTRL__RBC_MODE_MASK
-			| UVD_CGC_CTRL__LMI_MC_MODE_MASK
-			| UVD_CGC_CTRL__LMI_UMC_MODE_MASK
-			| UVD_CGC_CTRL__IDCT_MODE_MASK
-			| UVD_CGC_CTRL__MPRD_MODE_MASK
-			| UVD_CGC_CTRL__MPC_MODE_MASK
-			| UVD_CGC_CTRL__LBSI_MODE_MASK
-			| UVD_CGC_CTRL__LRBBM_MODE_MASK
-			| UVD_CGC_CTRL__WCB_MODE_MASK
-			| UVD_CGC_CTRL__VCPU_MODE_MASK
-			| UVD_CGC_CTRL__MMSCH_MODE_MASK);
-		WREG32_SOC15(VCN, i, mmUVD_CGC_CTRL, data);
-
-		/* turn on */
-		data = RREG32_SOC15(VCN, i, mmUVD_SUVD_CGC_GATE);
-		data |= (UVD_SUVD_CGC_GATE__SRE_MASK
-			| UVD_SUVD_CGC_GATE__SIT_MASK
-			| UVD_SUVD_CGC_GATE__SMP_MASK
-			| UVD_SUVD_CGC_GATE__SCM_MASK
-			| UVD_SUVD_CGC_GATE__SDB_MASK
-			| UVD_SUVD_CGC_GATE__SRE_H264_MASK
-			| UVD_SUVD_CGC_GATE__SRE_HEVC_MASK
-			| UVD_SUVD_CGC_GATE__SIT_H264_MASK
-			| UVD_SUVD_CGC_GATE__SIT_HEVC_MASK
-			| UVD_SUVD_CGC_GATE__SCM_H264_MASK
-			| UVD_SUVD_CGC_GATE__SCM_HEVC_MASK
-			| UVD_SUVD_CGC_GATE__SDB_H264_MASK
-			| UVD_SUVD_CGC_GATE__SDB_HEVC_MASK
-			| UVD_SUVD_CGC_GATE__SCLR_MASK
-			| UVD_SUVD_CGC_GATE__UVD_SC_MASK
-			| UVD_SUVD_CGC_GATE__ENT_MASK
-			| UVD_SUVD_CGC_GATE__SIT_HEVC_DEC_MASK
-			| UVD_SUVD_CGC_GATE__SIT_HEVC_ENC_MASK
-			| UVD_SUVD_CGC_GATE__SITE_MASK
-			| UVD_SUVD_CGC_GATE__SRE_VP9_MASK
-			| UVD_SUVD_CGC_GATE__SCM_VP9_MASK
-			| UVD_SUVD_CGC_GATE__SIT_VP9_DEC_MASK
-			| UVD_SUVD_CGC_GATE__SDB_VP9_MASK
-			| UVD_SUVD_CGC_GATE__IME_HEVC_MASK);
-		WREG32_SOC15(VCN, i, mmUVD_SUVD_CGC_GATE, data);
-
-		data = RREG32_SOC15(VCN, i, mmUVD_SUVD_CGC_CTRL);
-		data &= ~(UVD_SUVD_CGC_CTRL__SRE_MODE_MASK
-			| UVD_SUVD_CGC_CTRL__SIT_MODE_MASK
-			| UVD_SUVD_CGC_CTRL__SMP_MODE_MASK
-			| UVD_SUVD_CGC_CTRL__SCM_MODE_MASK
-			| UVD_SUVD_CGC_CTRL__SDB_MODE_MASK
-			| UVD_SUVD_CGC_CTRL__SCLR_MODE_MASK
-			| UVD_SUVD_CGC_CTRL__UVD_SC_MODE_MASK
-			| UVD_SUVD_CGC_CTRL__ENT_MODE_MASK
+	if (adev->vcn.harvest_config & (1 << i))
+		return;
+	/* UVD disable CGC */
+	data = RREG32_SOC15(VCN, i, mmUVD_CGC_CTRL);
+	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
+		data |= 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+	else
+		data &= ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK;
+	data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
+	data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
+	WREG32_SOC15(VCN, i, mmUVD_CGC_CTRL, data);
+
+	data = RREG32_SOC15(VCN, i, mmUVD_CGC_GATE);
+	data &= ~(UVD_CGC_GATE__SYS_MASK
+		  | UVD_CGC_GATE__UDEC_MASK
+		  | UVD_CGC_GATE__MPEG2_MASK
+		  | UVD_CGC_GATE__REGS_MASK
+		  | UVD_CGC_GATE__RBC_MASK
+		  | UVD_CGC_GATE__LMI_MC_MASK
+		  | UVD_CGC_GATE__LMI_UMC_MASK
+		  | UVD_CGC_GATE__IDCT_MASK
+		  | UVD_CGC_GATE__MPRD_MASK
+		  | UVD_CGC_GATE__MPC_MASK
+		  | UVD_CGC_GATE__LBSI_MASK
+		  | UVD_CGC_GATE__LRBBM_MASK
+		  | UVD_CGC_GATE__UDEC_RE_MASK
+		  | UVD_CGC_GATE__UDEC_CM_MASK
+		  | UVD_CGC_GATE__UDEC_IT_MASK
+		  | UVD_CGC_GATE__UDEC_DB_MASK
+		  | UVD_CGC_GATE__UDEC_MP_MASK
+		  | UVD_CGC_GATE__WCB_MASK
+		  | UVD_CGC_GATE__VCPU_MASK
+		  | UVD_CGC_GATE__MMSCH_MASK);
+
+	WREG32_SOC15(VCN, i, mmUVD_CGC_GATE, data);
+
+	SOC15_WAIT_ON_RREG(VCN, i, mmUVD_CGC_GATE, 0,  0xFFFFFFFF);
+
+	data = RREG32_SOC15(VCN, i, mmUVD_CGC_CTRL);
+	data &= ~(UVD_CGC_CTRL__UDEC_RE_MODE_MASK
+		  | UVD_CGC_CTRL__UDEC_CM_MODE_MASK
+		  | UVD_CGC_CTRL__UDEC_IT_MODE_MASK
+		  | UVD_CGC_CTRL__UDEC_DB_MODE_MASK
+		  | UVD_CGC_CTRL__UDEC_MP_MODE_MASK
+		  | UVD_CGC_CTRL__SYS_MODE_MASK
+		  | UVD_CGC_CTRL__UDEC_MODE_MASK
+		  | UVD_CGC_CTRL__MPEG2_MODE_MASK
+		  | UVD_CGC_CTRL__REGS_MODE_MASK
+		  | UVD_CGC_CTRL__RBC_MODE_MASK
+		  | UVD_CGC_CTRL__LMI_MC_MODE_MASK
+		  | UVD_CGC_CTRL__LMI_UMC_MODE_MASK
+		  | UVD_CGC_CTRL__IDCT_MODE_MASK
+		  | UVD_CGC_CTRL__MPRD_MODE_MASK
+		  | UVD_CGC_CTRL__MPC_MODE_MASK
+		  | UVD_CGC_CTRL__LBSI_MODE_MASK
+		  | UVD_CGC_CTRL__LRBBM_MODE_MASK
+		  | UVD_CGC_CTRL__WCB_MODE_MASK
+		  | UVD_CGC_CTRL__VCPU_MODE_MASK
+		  | UVD_CGC_CTRL__MMSCH_MODE_MASK);
+	WREG32_SOC15(VCN, i, mmUVD_CGC_CTRL, data);
+
+	/* turn on */
+	data = RREG32_SOC15(VCN, i, mmUVD_SUVD_CGC_GATE);
+	data |= (UVD_SUVD_CGC_GATE__SRE_MASK
+		 | UVD_SUVD_CGC_GATE__SIT_MASK
+		 | UVD_SUVD_CGC_GATE__SMP_MASK
+		 | UVD_SUVD_CGC_GATE__SCM_MASK
+		 | UVD_SUVD_CGC_GATE__SDB_MASK
+		 | UVD_SUVD_CGC_GATE__SRE_H264_MASK
+		 | UVD_SUVD_CGC_GATE__SRE_HEVC_MASK
+		 | UVD_SUVD_CGC_GATE__SIT_H264_MASK
+		 | UVD_SUVD_CGC_GATE__SIT_HEVC_MASK
+		 | UVD_SUVD_CGC_GATE__SCM_H264_MASK
+		 | UVD_SUVD_CGC_GATE__SCM_HEVC_MASK
+		 | UVD_SUVD_CGC_GATE__SDB_H264_MASK
+		 | UVD_SUVD_CGC_GATE__SDB_HEVC_MASK
+		 | UVD_SUVD_CGC_GATE__SCLR_MASK
+		 | UVD_SUVD_CGC_GATE__UVD_SC_MASK
+		 | UVD_SUVD_CGC_GATE__ENT_MASK
+		 | UVD_SUVD_CGC_GATE__SIT_HEVC_DEC_MASK
+		 | UVD_SUVD_CGC_GATE__SIT_HEVC_ENC_MASK
+		 | UVD_SUVD_CGC_GATE__SITE_MASK
+		 | UVD_SUVD_CGC_GATE__SRE_VP9_MASK
+		 | UVD_SUVD_CGC_GATE__SCM_VP9_MASK
+		 | UVD_SUVD_CGC_GATE__SIT_VP9_DEC_MASK
+		 | UVD_SUVD_CGC_GATE__SDB_VP9_MASK
+		 | UVD_SUVD_CGC_GATE__IME_HEVC_MASK);
+	WREG32_SOC15(VCN, i, mmUVD_SUVD_CGC_GATE, data);
+
+	data = RREG32_SOC15(VCN, i, mmUVD_SUVD_CGC_CTRL);
+	data &= ~(UVD_SUVD_CGC_CTRL__SRE_MODE_MASK
+		  | UVD_SUVD_CGC_CTRL__SIT_MODE_MASK
+		  | UVD_SUVD_CGC_CTRL__SMP_MODE_MASK
+		  | UVD_SUVD_CGC_CTRL__SCM_MODE_MASK
+		  | UVD_SUVD_CGC_CTRL__SDB_MODE_MASK
+		  | UVD_SUVD_CGC_CTRL__SCLR_MODE_MASK
+		  | UVD_SUVD_CGC_CTRL__UVD_SC_MODE_MASK
+		  | UVD_SUVD_CGC_CTRL__ENT_MODE_MASK
 			| UVD_SUVD_CGC_CTRL__IME_MODE_MASK
-			| UVD_SUVD_CGC_CTRL__SITE_MODE_MASK);
-		WREG32_SOC15(VCN, i, mmUVD_SUVD_CGC_CTRL, data);
-	}
+		  | UVD_SUVD_CGC_CTRL__SITE_MODE_MASK);
+	WREG32_SOC15(VCN, i, mmUVD_SUVD_CGC_CTRL, data);
 }
 
-static void vcn_v2_5_clock_gating_dpg_mode(struct amdgpu_device *adev,
-		uint8_t sram_sel, int inst_idx, uint8_t indirect)
+static void vcn_v2_5_clock_gating_dpg_mode(struct amdgpu_vcn_inst *vinst,
+		uint8_t sram_sel, uint8_t indirect)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
 	uint32_t reg_data = 0;
 
 	/* enable sw clock gating control */
@@ -727,71 +907,72 @@ static void vcn_v2_5_clock_gating_dpg_mode(struct amdgpu_device *adev,
 /**
  * vcn_v2_5_enable_clock_gating - enable VCN clock gating
  *
- * @adev: amdgpu_device pointer
+ * @vinst: VCN instance
  *
  * Enable clock gating for VCN block
  */
-static void vcn_v2_5_enable_clock_gating(struct amdgpu_device *adev)
+static void vcn_v2_5_enable_clock_gating(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int i = vinst->inst;
 	uint32_t data = 0;
-	int i;
 
-	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
-		if (adev->vcn.harvest_config & (1 << i))
-			continue;
-		/* enable UVD CGC */
-		data = RREG32_SOC15(VCN, i, mmUVD_CGC_CTRL);
-		if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
-			data |= 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
-		else
-			data |= 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
-		data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
-		data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
-		WREG32_SOC15(VCN, i, mmUVD_CGC_CTRL, data);
-
-		data = RREG32_SOC15(VCN, i, mmUVD_CGC_CTRL);
-		data |= (UVD_CGC_CTRL__UDEC_RE_MODE_MASK
-			| UVD_CGC_CTRL__UDEC_CM_MODE_MASK
-			| UVD_CGC_CTRL__UDEC_IT_MODE_MASK
-			| UVD_CGC_CTRL__UDEC_DB_MODE_MASK
-			| UVD_CGC_CTRL__UDEC_MP_MODE_MASK
-			| UVD_CGC_CTRL__SYS_MODE_MASK
-			| UVD_CGC_CTRL__UDEC_MODE_MASK
-			| UVD_CGC_CTRL__MPEG2_MODE_MASK
-			| UVD_CGC_CTRL__REGS_MODE_MASK
-			| UVD_CGC_CTRL__RBC_MODE_MASK
-			| UVD_CGC_CTRL__LMI_MC_MODE_MASK
-			| UVD_CGC_CTRL__LMI_UMC_MODE_MASK
-			| UVD_CGC_CTRL__IDCT_MODE_MASK
-			| UVD_CGC_CTRL__MPRD_MODE_MASK
-			| UVD_CGC_CTRL__MPC_MODE_MASK
-			| UVD_CGC_CTRL__LBSI_MODE_MASK
-			| UVD_CGC_CTRL__LRBBM_MODE_MASK
-			| UVD_CGC_CTRL__WCB_MODE_MASK
-			| UVD_CGC_CTRL__VCPU_MODE_MASK);
-		WREG32_SOC15(VCN, i, mmUVD_CGC_CTRL, data);
-
-		data = RREG32_SOC15(VCN, i, mmUVD_SUVD_CGC_CTRL);
-		data |= (UVD_SUVD_CGC_CTRL__SRE_MODE_MASK
-			| UVD_SUVD_CGC_CTRL__SIT_MODE_MASK
-			| UVD_SUVD_CGC_CTRL__SMP_MODE_MASK
-			| UVD_SUVD_CGC_CTRL__SCM_MODE_MASK
-			| UVD_SUVD_CGC_CTRL__SDB_MODE_MASK
-			| UVD_SUVD_CGC_CTRL__SCLR_MODE_MASK
-			| UVD_SUVD_CGC_CTRL__UVD_SC_MODE_MASK
-			| UVD_SUVD_CGC_CTRL__ENT_MODE_MASK
-			| UVD_SUVD_CGC_CTRL__IME_MODE_MASK
-			| UVD_SUVD_CGC_CTRL__SITE_MODE_MASK);
-		WREG32_SOC15(VCN, i, mmUVD_SUVD_CGC_CTRL, data);
-	}
+	if (adev->vcn.harvest_config & (1 << i))
+		return;
+	/* enable UVD CGC */
+	data = RREG32_SOC15(VCN, i, mmUVD_CGC_CTRL);
+	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
+		data |= 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+	else
+		data |= 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+	data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
+	data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
+	WREG32_SOC15(VCN, i, mmUVD_CGC_CTRL, data);
+
+	data = RREG32_SOC15(VCN, i, mmUVD_CGC_CTRL);
+	data |= (UVD_CGC_CTRL__UDEC_RE_MODE_MASK
+		 | UVD_CGC_CTRL__UDEC_CM_MODE_MASK
+		 | UVD_CGC_CTRL__UDEC_IT_MODE_MASK
+		 | UVD_CGC_CTRL__UDEC_DB_MODE_MASK
+		 | UVD_CGC_CTRL__UDEC_MP_MODE_MASK
+		 | UVD_CGC_CTRL__SYS_MODE_MASK
+		 | UVD_CGC_CTRL__UDEC_MODE_MASK
+		 | UVD_CGC_CTRL__MPEG2_MODE_MASK
+		 | UVD_CGC_CTRL__REGS_MODE_MASK
+		 | UVD_CGC_CTRL__RBC_MODE_MASK
+		 | UVD_CGC_CTRL__LMI_MC_MODE_MASK
+		 | UVD_CGC_CTRL__LMI_UMC_MODE_MASK
+		 | UVD_CGC_CTRL__IDCT_MODE_MASK
+		 | UVD_CGC_CTRL__MPRD_MODE_MASK
+		 | UVD_CGC_CTRL__MPC_MODE_MASK
+		 | UVD_CGC_CTRL__LBSI_MODE_MASK
+		 | UVD_CGC_CTRL__LRBBM_MODE_MASK
+		 | UVD_CGC_CTRL__WCB_MODE_MASK
+		 | UVD_CGC_CTRL__VCPU_MODE_MASK);
+	WREG32_SOC15(VCN, i, mmUVD_CGC_CTRL, data);
+
+	data = RREG32_SOC15(VCN, i, mmUVD_SUVD_CGC_CTRL);
+	data |= (UVD_SUVD_CGC_CTRL__SRE_MODE_MASK
+		 | UVD_SUVD_CGC_CTRL__SIT_MODE_MASK
+		 | UVD_SUVD_CGC_CTRL__SMP_MODE_MASK
+		 | UVD_SUVD_CGC_CTRL__SCM_MODE_MASK
+		 | UVD_SUVD_CGC_CTRL__SDB_MODE_MASK
+		 | UVD_SUVD_CGC_CTRL__SCLR_MODE_MASK
+		 | UVD_SUVD_CGC_CTRL__UVD_SC_MODE_MASK
+		 | UVD_SUVD_CGC_CTRL__ENT_MODE_MASK
+		 | UVD_SUVD_CGC_CTRL__IME_MODE_MASK
+		 | UVD_SUVD_CGC_CTRL__SITE_MODE_MASK);
+	WREG32_SOC15(VCN, i, mmUVD_SUVD_CGC_CTRL, data);
 }
 
-static void vcn_v2_6_enable_ras(struct amdgpu_device *adev, int inst_idx,
+static void vcn_v2_6_enable_ras(struct amdgpu_vcn_inst *vinst,
 				bool indirect)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
 	uint32_t tmp;
 
-	if (adev->ip_versions[UVD_HWIP][0] != IP_VERSION(2, 6, 0))
+	if (amdgpu_ip_version(adev, UVD_HWIP, 0) != IP_VERSION(2, 6, 0))
 		return;
 
 	tmp = VCN_RAS_CNTL__VCPU_VCODEC_REARM_MASK |
@@ -813,11 +994,14 @@ static void vcn_v2_6_enable_ras(struct amdgpu_device *adev, int inst_idx,
 			      tmp, 0, indirect);
 }
 
-static int vcn_v2_5_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, bool indirect)
+static int vcn_v2_5_start_dpg_mode(struct amdgpu_vcn_inst *vinst, bool indirect)
 {
-	volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr;
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	struct amdgpu_fw_shared *fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr;
 	struct amdgpu_ring *ring;
 	uint32_t rb_bufsz, tmp;
+	int ret;
 
 	/* disable register anti-hang mechanism */
 	WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, mmUVD_POWER_STATUS), 1,
@@ -832,7 +1016,7 @@ static int vcn_v2_5_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, boo
 		adev->vcn.inst[inst_idx].dpg_sram_curr_addr = (uint32_t *)adev->vcn.inst[inst_idx].dpg_sram_cpu_addr;
 
 	/* enable clock gating */
-	vcn_v2_5_clock_gating_dpg_mode(adev, 0, inst_idx, indirect);
+	vcn_v2_5_clock_gating_dpg_mode(vinst, 0, indirect);
 
 	/* enable VCPU clock */
 	tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
@@ -881,7 +1065,7 @@ static int vcn_v2_5_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, boo
 		 (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
 		 (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)), 0, indirect);
 
-	vcn_v2_5_mc_resume_dpg_mode(adev, inst_idx, indirect);
+	vcn_v2_5_mc_resume_dpg_mode(vinst, indirect);
 
 	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
 		VCN, 0, mmUVD_REG_XX_MASK), 0x10, 0, indirect);
@@ -892,7 +1076,7 @@ static int vcn_v2_5_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, boo
 	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
 		VCN, 0, mmUVD_LMI_CTRL2), 0, 0, indirect);
 
-	vcn_v2_6_enable_ras(adev, inst_idx, indirect);
+	vcn_v2_6_enable_ras(vinst, indirect);
 
 	/* unblock VCPU register access */
 	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
@@ -908,10 +1092,13 @@ static int vcn_v2_5_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, boo
 		VCN, 0, mmUVD_MASTINT_EN),
 		UVD_MASTINT_EN__VCPU_EN_MASK, 0, indirect);
 
-	if (indirect)
-		psp_update_vcn_sram(adev, inst_idx, adev->vcn.inst[inst_idx].dpg_sram_gpu_addr,
-				    (uint32_t)((uintptr_t)adev->vcn.inst[inst_idx].dpg_sram_curr_addr -
-					       (uintptr_t)adev->vcn.inst[inst_idx].dpg_sram_cpu_addr));
+	if (indirect) {
+		ret = amdgpu_vcn_psp_update_sram(adev, inst_idx, 0);
+		if (ret) {
+			dev_err(adev->dev, "vcn sram load failed %d\n", ret);
+			return ret;
+		}
+	}
 
 	ring = &adev->vcn.inst[inst_idx].ring_dec;
 	/* force RBC into idle state */
@@ -956,198 +1143,200 @@ static int vcn_v2_5_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, boo
 	WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, mmUVD_POWER_STATUS),
 		0, ~UVD_POWER_STATUS__STALL_DPG_POWER_UP_MASK);
 
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, inst_idx, mmUVD_STATUS);
+
 	return 0;
 }
 
-static int vcn_v2_5_start(struct amdgpu_device *adev)
+static int vcn_v2_5_start(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int i = vinst->inst;
+	struct amdgpu_fw_shared *fw_shared =
+		adev->vcn.inst[i].fw_shared.cpu_addr;
 	struct amdgpu_ring *ring;
 	uint32_t rb_bufsz, tmp;
-	int i, j, k, r;
+	int j, k, r;
+
+	if (adev->vcn.harvest_config & (1 << i))
+		return 0;
 
 	if (adev->pm.dpm_enabled)
-		amdgpu_dpm_enable_uvd(adev, true);
+		amdgpu_dpm_enable_vcn(adev, true, i);
 
-	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
-		if (adev->vcn.harvest_config & (1 << i))
-			continue;
-		if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
-			r = vcn_v2_5_start_dpg_mode(adev, i, adev->vcn.indirect_sram);
-			continue;
-		}
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
+		return vcn_v2_5_start_dpg_mode(vinst, adev->vcn.inst[i].indirect_sram);
 
-		/* disable register anti-hang mechanism */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_POWER_STATUS), 0,
-			~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
+	/* disable register anti-hang mechanism */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_POWER_STATUS), 0,
+		 ~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
 
-		/* set uvd status busy */
-		tmp = RREG32_SOC15(VCN, i, mmUVD_STATUS) | UVD_STATUS__UVD_BUSY;
-		WREG32_SOC15(VCN, i, mmUVD_STATUS, tmp);
-	}
+	/* set uvd status busy */
+	tmp = RREG32_SOC15(VCN, i, mmUVD_STATUS) | UVD_STATUS__UVD_BUSY;
+	WREG32_SOC15(VCN, i, mmUVD_STATUS, tmp);
 
 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
 		return 0;
 
-	/*SW clock gating */
-	vcn_v2_5_disable_clock_gating(adev);
+	/* SW clock gating */
+	vcn_v2_5_disable_clock_gating(vinst);
 
-	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
-		if (adev->vcn.harvest_config & (1 << i))
-			continue;
-		/* enable VCPU clock */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL),
-			UVD_VCPU_CNTL__CLK_EN_MASK, ~UVD_VCPU_CNTL__CLK_EN_MASK);
-
-		/* disable master interrupt */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_MASTINT_EN), 0,
-			~UVD_MASTINT_EN__VCPU_EN_MASK);
-
-		/* setup mmUVD_LMI_CTRL */
-		tmp = RREG32_SOC15(VCN, i, mmUVD_LMI_CTRL);
-		tmp &= ~0xff;
-		WREG32_SOC15(VCN, i, mmUVD_LMI_CTRL, tmp | 0x8|
-			UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK	|
-			UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
-			UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
-			UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK);
-
-		/* setup mmUVD_MPC_CNTL */
-		tmp = RREG32_SOC15(VCN, i, mmUVD_MPC_CNTL);
-		tmp &= ~UVD_MPC_CNTL__REPLACEMENT_MODE_MASK;
-		tmp |= 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT;
-		WREG32_SOC15(VCN, i, mmUVD_MPC_CNTL, tmp);
-
-		/* setup UVD_MPC_SET_MUXA0 */
-		WREG32_SOC15(VCN, i, mmUVD_MPC_SET_MUXA0,
-			((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) |
-			(0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) |
-			(0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) |
-			(0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)));
-
-		/* setup UVD_MPC_SET_MUXB0 */
-		WREG32_SOC15(VCN, i, mmUVD_MPC_SET_MUXB0,
-			((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) |
-			(0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) |
-			(0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) |
-			(0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)));
-
-		/* setup mmUVD_MPC_SET_MUX */
-		WREG32_SOC15(VCN, i, mmUVD_MPC_SET_MUX,
-			((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) |
-			(0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
-			(0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)));
-	}
+	/* enable VCPU clock */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL),
+		 UVD_VCPU_CNTL__CLK_EN_MASK, ~UVD_VCPU_CNTL__CLK_EN_MASK);
 
-	vcn_v2_5_mc_resume(adev);
+	/* disable master interrupt */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_MASTINT_EN), 0,
+		 ~UVD_MASTINT_EN__VCPU_EN_MASK);
 
-	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
-		volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
-		if (adev->vcn.harvest_config & (1 << i))
-			continue;
-		/* VCN global tiling registers */
-		WREG32_SOC15(VCN, i, mmUVD_GFX8_ADDR_CONFIG,
-			adev->gfx.config.gb_addr_config);
-		WREG32_SOC15(VCN, i, mmUVD_GFX8_ADDR_CONFIG,
-			adev->gfx.config.gb_addr_config);
+	/* setup mmUVD_LMI_CTRL */
+	tmp = RREG32_SOC15(VCN, i, mmUVD_LMI_CTRL);
+	tmp &= ~0xff;
+	WREG32_SOC15(VCN, i, mmUVD_LMI_CTRL, tmp | 0x8|
+		     UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK	|
+		     UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
+		     UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
+		     UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK);
+
+	/* setup mmUVD_MPC_CNTL */
+	tmp = RREG32_SOC15(VCN, i, mmUVD_MPC_CNTL);
+	tmp &= ~UVD_MPC_CNTL__REPLACEMENT_MODE_MASK;
+	tmp |= 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT;
+	WREG32_SOC15(VCN, i, mmUVD_MPC_CNTL, tmp);
+
+	/* setup UVD_MPC_SET_MUXA0 */
+	WREG32_SOC15(VCN, i, mmUVD_MPC_SET_MUXA0,
+		     ((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) |
+		      (0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) |
+		      (0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) |
+		      (0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)));
+
+	/* setup UVD_MPC_SET_MUXB0 */
+	WREG32_SOC15(VCN, i, mmUVD_MPC_SET_MUXB0,
+		     ((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) |
+		      (0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) |
+		      (0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) |
+		      (0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)));
+
+	/* setup mmUVD_MPC_SET_MUX */
+	WREG32_SOC15(VCN, i, mmUVD_MPC_SET_MUX,
+		     ((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) |
+		      (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
+		      (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)));
+
+	vcn_v2_5_mc_resume(vinst);
 
-		/* enable LMI MC and UMC channels */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_LMI_CTRL2), 0,
-			~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
+	/* VCN global tiling registers */
+	WREG32_SOC15(VCN, i, mmUVD_GFX8_ADDR_CONFIG,
+		     adev->gfx.config.gb_addr_config);
+	WREG32_SOC15(VCN, i, mmUVD_GFX8_ADDR_CONFIG,
+		     adev->gfx.config.gb_addr_config);
 
-		/* unblock VCPU register access */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_RB_ARB_CTRL), 0,
-			~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
+	/* enable LMI MC and UMC channels */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_LMI_CTRL2), 0,
+		 ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
 
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL), 0,
-			~UVD_VCPU_CNTL__BLK_RST_MASK);
-
-		for (k = 0; k < 10; ++k) {
-			uint32_t status;
-
-			for (j = 0; j < 100; ++j) {
-				status = RREG32_SOC15(VCN, i, mmUVD_STATUS);
-				if (status & 2)
-					break;
-				if (amdgpu_emu_mode == 1)
-					msleep(500);
-				else
-					mdelay(10);
-			}
-			r = 0;
+	/* unblock VCPU register access */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_RB_ARB_CTRL), 0,
+		 ~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
+
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL), 0,
+		 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+
+	for (k = 0; k < 10; ++k) {
+		uint32_t status;
+
+		for (j = 0; j < 100; ++j) {
+			status = RREG32_SOC15(VCN, i, mmUVD_STATUS);
 			if (status & 2)
 				break;
+			if (amdgpu_emu_mode == 1)
+				msleep(500);
+			else
+				mdelay(10);
+		}
+		r = 0;
+		if (status & 2)
+			break;
 
-			DRM_ERROR("VCN decode not responding, trying to reset the VCPU!!!\n");
-			WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL),
-				UVD_VCPU_CNTL__BLK_RST_MASK,
-				~UVD_VCPU_CNTL__BLK_RST_MASK);
-			mdelay(10);
-			WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL), 0,
-				~UVD_VCPU_CNTL__BLK_RST_MASK);
+		DRM_ERROR("VCN decode not responding, trying to reset the VCPU!!!\n");
+		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL),
+			 UVD_VCPU_CNTL__BLK_RST_MASK,
+			 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+		mdelay(10);
+		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL), 0,
+			 ~UVD_VCPU_CNTL__BLK_RST_MASK);
 
-			mdelay(10);
-			r = -1;
-		}
+		mdelay(10);
+		r = -1;
+	}
 
-		if (r) {
-			DRM_ERROR("VCN decode not responding, giving up!!!\n");
-			return r;
-		}
+	if (r) {
+		DRM_ERROR("VCN decode not responding, giving up!!!\n");
+		return r;
+	}
 
-		/* enable master interrupt */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_MASTINT_EN),
-			UVD_MASTINT_EN__VCPU_EN_MASK,
-			~UVD_MASTINT_EN__VCPU_EN_MASK);
+	/* enable master interrupt */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_MASTINT_EN),
+		 UVD_MASTINT_EN__VCPU_EN_MASK,
+		 ~UVD_MASTINT_EN__VCPU_EN_MASK);
 
-		/* clear the busy bit of VCN_STATUS */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_STATUS), 0,
-			~(2 << UVD_STATUS__VCPU_REPORT__SHIFT));
+	/* clear the busy bit of VCN_STATUS */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_STATUS), 0,
+		 ~(2 << UVD_STATUS__VCPU_REPORT__SHIFT));
 
-		WREG32_SOC15(VCN, i, mmUVD_LMI_RBC_RB_VMID, 0);
+	WREG32_SOC15(VCN, i, mmUVD_LMI_RBC_RB_VMID, 0);
 
-		ring = &adev->vcn.inst[i].ring_dec;
-		/* force RBC into idle state */
-		rb_bufsz = order_base_2(ring->ring_size);
-		tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz);
-		tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1);
-		tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1);
-		tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1);
-		tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1);
-		WREG32_SOC15(VCN, i, mmUVD_RBC_RB_CNTL, tmp);
+	ring = &adev->vcn.inst[i].ring_dec;
+	/* force RBC into idle state */
+	rb_bufsz = order_base_2(ring->ring_size);
+	tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz);
+	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1);
+	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1);
+	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1);
+	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1);
+	WREG32_SOC15(VCN, i, mmUVD_RBC_RB_CNTL, tmp);
 
-		fw_shared->multi_queue.decode_queue_mode |= FW_QUEUE_RING_RESET;
-		/* program the RB_BASE for ring buffer */
-		WREG32_SOC15(VCN, i, mmUVD_LMI_RBC_RB_64BIT_BAR_LOW,
-			lower_32_bits(ring->gpu_addr));
-		WREG32_SOC15(VCN, i, mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH,
-			upper_32_bits(ring->gpu_addr));
+	fw_shared->multi_queue.decode_queue_mode |= FW_QUEUE_RING_RESET;
+	/* program the RB_BASE for ring buffer */
+	WREG32_SOC15(VCN, i, mmUVD_LMI_RBC_RB_64BIT_BAR_LOW,
+		     lower_32_bits(ring->gpu_addr));
+	WREG32_SOC15(VCN, i, mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH,
+		     upper_32_bits(ring->gpu_addr));
 
-		/* Initialize the ring buffer's read and write pointers */
-		WREG32_SOC15(VCN, i, mmUVD_RBC_RB_RPTR, 0);
+	/* Initialize the ring buffer's read and write pointers */
+	WREG32_SOC15(VCN, i, mmUVD_RBC_RB_RPTR, 0);
 
-		ring->wptr = RREG32_SOC15(VCN, i, mmUVD_RBC_RB_RPTR);
-		WREG32_SOC15(VCN, i, mmUVD_RBC_RB_WPTR,
-				lower_32_bits(ring->wptr));
-		fw_shared->multi_queue.decode_queue_mode &= ~FW_QUEUE_RING_RESET;
+	ring->wptr = RREG32_SOC15(VCN, i, mmUVD_RBC_RB_RPTR);
+	WREG32_SOC15(VCN, i, mmUVD_RBC_RB_WPTR,
+		     lower_32_bits(ring->wptr));
+	fw_shared->multi_queue.decode_queue_mode &= ~FW_QUEUE_RING_RESET;
 
-		fw_shared->multi_queue.encode_generalpurpose_queue_mode |= FW_QUEUE_RING_RESET;
-		ring = &adev->vcn.inst[i].ring_enc[0];
-		WREG32_SOC15(VCN, i, mmUVD_RB_RPTR, lower_32_bits(ring->wptr));
-		WREG32_SOC15(VCN, i, mmUVD_RB_WPTR, lower_32_bits(ring->wptr));
-		WREG32_SOC15(VCN, i, mmUVD_RB_BASE_LO, ring->gpu_addr);
-		WREG32_SOC15(VCN, i, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
-		WREG32_SOC15(VCN, i, mmUVD_RB_SIZE, ring->ring_size / 4);
-		fw_shared->multi_queue.encode_generalpurpose_queue_mode &= ~FW_QUEUE_RING_RESET;
-
-		fw_shared->multi_queue.encode_lowlatency_queue_mode |= FW_QUEUE_RING_RESET;
-		ring = &adev->vcn.inst[i].ring_enc[1];
-		WREG32_SOC15(VCN, i, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr));
-		WREG32_SOC15(VCN, i, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr));
-		WREG32_SOC15(VCN, i, mmUVD_RB_BASE_LO2, ring->gpu_addr);
-		WREG32_SOC15(VCN, i, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr));
-		WREG32_SOC15(VCN, i, mmUVD_RB_SIZE2, ring->ring_size / 4);
-		fw_shared->multi_queue.encode_lowlatency_queue_mode &= ~FW_QUEUE_RING_RESET;
-	}
+	fw_shared->multi_queue.encode_generalpurpose_queue_mode |= FW_QUEUE_RING_RESET;
+	ring = &adev->vcn.inst[i].ring_enc[0];
+	WREG32_SOC15(VCN, i, mmUVD_RB_RPTR, lower_32_bits(ring->wptr));
+	WREG32_SOC15(VCN, i, mmUVD_RB_WPTR, lower_32_bits(ring->wptr));
+	WREG32_SOC15(VCN, i, mmUVD_RB_BASE_LO, ring->gpu_addr);
+	WREG32_SOC15(VCN, i, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
+	WREG32_SOC15(VCN, i, mmUVD_RB_SIZE, ring->ring_size / 4);
+	fw_shared->multi_queue.encode_generalpurpose_queue_mode &= ~FW_QUEUE_RING_RESET;
+
+	fw_shared->multi_queue.encode_lowlatency_queue_mode |= FW_QUEUE_RING_RESET;
+	ring = &adev->vcn.inst[i].ring_enc[1];
+	WREG32_SOC15(VCN, i, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr));
+	WREG32_SOC15(VCN, i, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr));
+	WREG32_SOC15(VCN, i, mmUVD_RB_BASE_LO2, ring->gpu_addr);
+	WREG32_SOC15(VCN, i, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr));
+	WREG32_SOC15(VCN, i, mmUVD_RB_SIZE2, ring->ring_size / 4);
+	fw_shared->multi_queue.encode_lowlatency_queue_mode &= ~FW_QUEUE_RING_RESET;
+
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, i, mmUVD_STATUS);
 
 	return 0;
 }
@@ -1238,7 +1427,7 @@ static int vcn_v2_5_sriov_start(struct amdgpu_device *adev)
 			SOC15_REG_OFFSET(VCN, i, mmUVD_STATUS),
 			~UVD_STATUS__UVD_BUSY, UVD_STATUS__UVD_BUSY);
 
-		size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+		size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.inst[i].fw->size + 4);
 		/* mc resume*/
 		if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 			MMSCH_V1_0_INSERT_DIRECT_WT(
@@ -1348,8 +1537,10 @@ static int vcn_v2_5_sriov_start(struct amdgpu_device *adev)
 	return vcn_v2_5_mmsch_start(adev, &adev->virt.mm_table);
 }
 
-static int vcn_v2_5_stop_dpg_mode(struct amdgpu_device *adev, int inst_idx)
+static int vcn_v2_5_stop_dpg_mode(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
 	uint32_t tmp;
 
 	/* Wait for power status to be 1 */
@@ -1373,80 +1564,93 @@ static int vcn_v2_5_stop_dpg_mode(struct amdgpu_device *adev, int inst_idx)
 	WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, mmUVD_POWER_STATUS), 0,
 			~UVD_POWER_STATUS__UVD_PG_MODE_MASK);
 
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, inst_idx, mmUVD_STATUS);
+
 	return 0;
 }
 
-static int vcn_v2_5_stop(struct amdgpu_device *adev)
+static int vcn_v2_5_stop(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int i = vinst->inst;
 	uint32_t tmp;
-	int i, r = 0;
+	int r;
 
-	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
-		if (adev->vcn.harvest_config & (1 << i))
-			continue;
-		if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
-			r = vcn_v2_5_stop_dpg_mode(adev, i);
-			continue;
-		}
+	if (adev->vcn.harvest_config & (1 << i))
+		return 0;
 
-		/* wait for vcn idle */
-		r = SOC15_WAIT_ON_RREG(VCN, i, mmUVD_STATUS, UVD_STATUS__IDLE, 0x7);
-		if (r)
-			return r;
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
+		r = vcn_v2_5_stop_dpg_mode(vinst);
+		goto done;
+	}
 
-		tmp = UVD_LMI_STATUS__VCPU_LMI_WRITE_CLEAN_MASK |
-			UVD_LMI_STATUS__READ_CLEAN_MASK |
-			UVD_LMI_STATUS__WRITE_CLEAN_MASK |
-			UVD_LMI_STATUS__WRITE_CLEAN_RAW_MASK;
-		r = SOC15_WAIT_ON_RREG(VCN, i, mmUVD_LMI_STATUS, tmp, tmp);
-		if (r)
-			return r;
+	/* wait for vcn idle */
+	r = SOC15_WAIT_ON_RREG(VCN, i, mmUVD_STATUS, UVD_STATUS__IDLE, 0x7);
+	if (r)
+		goto done;
 
-		/* block LMI UMC channel */
-		tmp = RREG32_SOC15(VCN, i, mmUVD_LMI_CTRL2);
-		tmp |= UVD_LMI_CTRL2__STALL_ARB_UMC_MASK;
-		WREG32_SOC15(VCN, i, mmUVD_LMI_CTRL2, tmp);
+	tmp = UVD_LMI_STATUS__VCPU_LMI_WRITE_CLEAN_MASK |
+		UVD_LMI_STATUS__READ_CLEAN_MASK |
+		UVD_LMI_STATUS__WRITE_CLEAN_MASK |
+		UVD_LMI_STATUS__WRITE_CLEAN_RAW_MASK;
+	r = SOC15_WAIT_ON_RREG(VCN, i, mmUVD_LMI_STATUS, tmp, tmp);
+	if (r)
+		goto done;
 
-		tmp = UVD_LMI_STATUS__UMC_READ_CLEAN_RAW_MASK|
-			UVD_LMI_STATUS__UMC_WRITE_CLEAN_RAW_MASK;
-		r = SOC15_WAIT_ON_RREG(VCN, i, mmUVD_LMI_STATUS, tmp, tmp);
-		if (r)
-			return r;
+	/* block LMI UMC channel */
+	tmp = RREG32_SOC15(VCN, i, mmUVD_LMI_CTRL2);
+	tmp |= UVD_LMI_CTRL2__STALL_ARB_UMC_MASK;
+	WREG32_SOC15(VCN, i, mmUVD_LMI_CTRL2, tmp);
 
-		/* block VCPU register access */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_RB_ARB_CTRL),
-			UVD_RB_ARB_CTRL__VCPU_DIS_MASK,
-			~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
+	tmp = UVD_LMI_STATUS__UMC_READ_CLEAN_RAW_MASK|
+		UVD_LMI_STATUS__UMC_WRITE_CLEAN_RAW_MASK;
+	r = SOC15_WAIT_ON_RREG(VCN, i, mmUVD_LMI_STATUS, tmp, tmp);
+	if (r)
+		goto done;
 
-		/* reset VCPU */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL),
-			UVD_VCPU_CNTL__BLK_RST_MASK,
-			~UVD_VCPU_CNTL__BLK_RST_MASK);
+	/* block VCPU register access */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_RB_ARB_CTRL),
+		 UVD_RB_ARB_CTRL__VCPU_DIS_MASK,
+		 ~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
 
-		/* disable VCPU clock */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL), 0,
-			~(UVD_VCPU_CNTL__CLK_EN_MASK));
+	/* reset VCPU */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL),
+		 UVD_VCPU_CNTL__BLK_RST_MASK,
+		 ~UVD_VCPU_CNTL__BLK_RST_MASK);
 
-		/* clear status */
-		WREG32_SOC15(VCN, i, mmUVD_STATUS, 0);
+	/* disable VCPU clock */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL), 0,
+		 ~(UVD_VCPU_CNTL__CLK_EN_MASK));
 
-		vcn_v2_5_enable_clock_gating(adev);
+	/* clear status */
+	WREG32_SOC15(VCN, i, mmUVD_STATUS, 0);
 
-		/* enable register anti-hang mechanism */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_POWER_STATUS),
-			UVD_POWER_STATUS__UVD_POWER_STATUS_MASK,
-			~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
-	}
+	vcn_v2_5_enable_clock_gating(vinst);
+
+	/* enable register anti-hang mechanism */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_POWER_STATUS),
+		 UVD_POWER_STATUS__UVD_POWER_STATUS_MASK,
+		 ~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
 
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, i, mmUVD_STATUS);
+done:
 	if (adev->pm.dpm_enabled)
-		amdgpu_dpm_enable_uvd(adev, false);
+		amdgpu_dpm_enable_vcn(adev, false, i);
 
-	return 0;
+	return r;
 }
 
-static int vcn_v2_5_pause_dpg_mode(struct amdgpu_device *adev,
-				int inst_idx, struct dpg_pause_state *new_state)
+static int vcn_v2_5_pause_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				   struct dpg_pause_state *new_state)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
 	struct amdgpu_ring *ring;
 	uint32_t reg_data = 0;
 	int ret_code = 0;
@@ -1463,7 +1667,7 @@ static int vcn_v2_5_pause_dpg_mode(struct amdgpu_device *adev,
 				UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
 
 			if (!ret_code) {
-				volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr;
+				struct amdgpu_fw_shared *fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr;
 
 				/* pause DPG */
 				reg_data |= UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK;
@@ -1592,11 +1796,12 @@ static const struct amdgpu_ring_funcs vcn_v2_5_dec_ring_vm_funcs = {
 	.insert_start = vcn_v2_0_dec_ring_insert_start,
 	.insert_end = vcn_v2_0_dec_ring_insert_end,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
-	.begin_use = amdgpu_vcn_ring_begin_use,
-	.end_use = amdgpu_vcn_ring_end_use,
+	.begin_use = vcn_v2_5_ring_begin_use,
+	.end_use = vcn_v2_5_ring_end_use,
 	.emit_wreg = vcn_v2_0_dec_ring_emit_wreg,
 	.emit_reg_wait = vcn_v2_0_dec_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = amdgpu_vcn_ring_reset,
 };
 
 /**
@@ -1690,11 +1895,12 @@ static const struct amdgpu_ring_funcs vcn_v2_5_enc_ring_vm_funcs = {
 	.insert_nop = amdgpu_ring_insert_nop,
 	.insert_end = vcn_v2_0_enc_ring_insert_end,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
-	.begin_use = amdgpu_vcn_ring_begin_use,
-	.end_use = amdgpu_vcn_ring_end_use,
+	.begin_use = vcn_v2_5_ring_begin_use,
+	.end_use = vcn_v2_5_ring_end_use,
 	.emit_wreg = vcn_v2_0_enc_ring_emit_wreg,
 	.emit_reg_wait = vcn_v2_0_enc_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = amdgpu_vcn_ring_reset,
 };
 
 static void vcn_v2_5_set_dec_ring_funcs(struct amdgpu_device *adev)
@@ -1706,7 +1912,6 @@ static void vcn_v2_5_set_dec_ring_funcs(struct amdgpu_device *adev)
 			continue;
 		adev->vcn.inst[i].ring_dec.funcs = &vcn_v2_5_dec_ring_vm_funcs;
 		adev->vcn.inst[i].ring_dec.me = i;
-		DRM_INFO("VCN(%d) decode is enabled in VM mode\n", i);
 	}
 }
 
@@ -1717,31 +1922,41 @@ static void vcn_v2_5_set_enc_ring_funcs(struct amdgpu_device *adev)
 	for (j = 0; j < adev->vcn.num_vcn_inst; ++j) {
 		if (adev->vcn.harvest_config & (1 << j))
 			continue;
-		for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
+		for (i = 0; i < adev->vcn.inst[j].num_enc_rings; ++i) {
 			adev->vcn.inst[j].ring_enc[i].funcs = &vcn_v2_5_enc_ring_vm_funcs;
 			adev->vcn.inst[j].ring_enc[i].me = j;
 		}
-		DRM_INFO("VCN(%d) encode is enabled in VM mode\n", j);
 	}
 }
 
-static bool vcn_v2_5_is_idle(void *handle)
+static int vcn_v2_5_reset(struct amdgpu_vcn_inst *vinst)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	int r;
+
+	r = vcn_v2_5_stop(vinst);
+	if (r)
+		return r;
+	return vcn_v2_5_start(vinst);
+}
+
+static bool vcn_v2_5_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
 	int i, ret = 1;
 
 	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
 		if (adev->vcn.harvest_config & (1 << i))
 			continue;
+
 		ret &= (RREG32_SOC15(VCN, i, mmUVD_STATUS) == UVD_STATUS__IDLE);
 	}
 
 	return ret;
 }
 
-static int vcn_v2_5_wait_for_idle(void *handle)
+static int vcn_v2_5_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i, ret = 0;
 
 	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
@@ -1756,45 +1971,50 @@ static int vcn_v2_5_wait_for_idle(void *handle)
 	return ret;
 }
 
-static int vcn_v2_5_set_clockgating_state(void *handle,
+static int vcn_v2_5_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_CG_STATE_GATE);
+	int i;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	if (enable) {
-		if (!vcn_v2_5_is_idle(handle))
-			return -EBUSY;
-		vcn_v2_5_enable_clock_gating(adev);
-	} else {
-		vcn_v2_5_disable_clock_gating(adev);
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[i];
+
+		if (enable) {
+			if (!vcn_v2_5_is_idle(ip_block))
+				return -EBUSY;
+			vcn_v2_5_enable_clock_gating(vinst);
+		} else {
+			vcn_v2_5_disable_clock_gating(vinst);
+		}
 	}
 
 	return 0;
 }
 
-static int vcn_v2_5_set_powergating_state(void *handle,
-					  enum amd_powergating_state state)
+static int vcn_v2_5_set_pg_state(struct amdgpu_vcn_inst *vinst,
+				 enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = vinst->adev;
 	int ret;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	if(state == adev->vcn.cur_state)
+	if (state == vinst->cur_state)
 		return 0;
 
 	if (state == AMD_PG_STATE_GATE)
-		ret = vcn_v2_5_stop(adev);
+		ret = vcn_v2_5_stop(vinst);
 	else
-		ret = vcn_v2_5_start(adev);
+		ret = vcn_v2_5_start(vinst);
 
-	if(!ret)
-		adev->vcn.cur_state = state;
+	if (!ret)
+		vinst->cur_state = state;
 
 	return ret;
 }
@@ -1807,6 +2027,14 @@ static int vcn_v2_5_set_interrupt_state(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int vcn_v2_6_set_ras_interrupt_state(struct amdgpu_device *adev,
+					struct amdgpu_irq_src *source,
+					unsigned int type,
+					enum amdgpu_interrupt_state state)
+{
+	return 0;
+}
+
 static int vcn_v2_5_process_interrupt(struct amdgpu_device *adev,
 				      struct amdgpu_irq_src *source,
 				      struct amdgpu_iv_entry *entry)
@@ -1837,9 +2065,6 @@ static int vcn_v2_5_process_interrupt(struct amdgpu_device *adev,
 	case VCN_2_0__SRCID__UVD_ENC_LOW_LATENCY:
 		amdgpu_fence_process(&adev->vcn.inst[ip_instance].ring_enc[1]);
 		break;
-	case VCN_2_6__SRCID_UVD_POISON:
-		amdgpu_vcn_process_poison_irq(adev, source, entry);
-		break;
 	default:
 		DRM_ERROR("Unhandled interrupt: %d %d\n",
 			  entry->src_id, entry->src_data[0]);
@@ -1854,6 +2079,11 @@ static const struct amdgpu_irq_src_funcs vcn_v2_5_irq_funcs = {
 	.process = vcn_v2_5_process_interrupt,
 };
 
+static const struct amdgpu_irq_src_funcs vcn_v2_6_ras_irq_funcs = {
+	.set = vcn_v2_6_set_ras_interrupt_state,
+	.process = amdgpu_vcn_process_poison_irq,
+};
+
 static void vcn_v2_5_set_irq_funcs(struct amdgpu_device *adev)
 {
 	int i;
@@ -1861,15 +2091,17 @@ static void vcn_v2_5_set_irq_funcs(struct amdgpu_device *adev)
 	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
 		if (adev->vcn.harvest_config & (1 << i))
 			continue;
-		adev->vcn.inst[i].irq.num_types = adev->vcn.num_enc_rings + 1;
+		adev->vcn.inst[i].irq.num_types = adev->vcn.inst[i].num_enc_rings + 1;
 		adev->vcn.inst[i].irq.funcs = &vcn_v2_5_irq_funcs;
+
+		adev->vcn.inst[i].ras_poison_irq.num_types = adev->vcn.inst[i].num_enc_rings + 1;
+		adev->vcn.inst[i].ras_poison_irq.funcs = &vcn_v2_6_ras_irq_funcs;
 	}
 }
 
 static const struct amd_ip_funcs vcn_v2_5_ip_funcs = {
 	.name = "vcn_v2_5",
 	.early_init = vcn_v2_5_early_init,
-	.late_init = NULL,
 	.sw_init = vcn_v2_5_sw_init,
 	.sw_fini = vcn_v2_5_sw_fini,
 	.hw_init = vcn_v2_5_hw_init,
@@ -1878,18 +2110,15 @@ static const struct amd_ip_funcs vcn_v2_5_ip_funcs = {
 	.resume = vcn_v2_5_resume,
 	.is_idle = vcn_v2_5_is_idle,
 	.wait_for_idle = vcn_v2_5_wait_for_idle,
-	.check_soft_reset = NULL,
-	.pre_soft_reset = NULL,
-	.soft_reset = NULL,
-	.post_soft_reset = NULL,
 	.set_clockgating_state = vcn_v2_5_set_clockgating_state,
-	.set_powergating_state = vcn_v2_5_set_powergating_state,
+	.set_powergating_state = vcn_set_powergating_state,
+	.dump_ip_state = amdgpu_vcn_dump_ip_state,
+	.print_ip_state = amdgpu_vcn_print_ip_state,
 };
 
 static const struct amd_ip_funcs vcn_v2_6_ip_funcs = {
         .name = "vcn_v2_6",
         .early_init = vcn_v2_5_early_init,
-        .late_init = NULL,
         .sw_init = vcn_v2_5_sw_init,
         .sw_fini = vcn_v2_5_sw_fini,
         .hw_init = vcn_v2_5_hw_init,
@@ -1898,12 +2127,10 @@ static const struct amd_ip_funcs vcn_v2_6_ip_funcs = {
         .resume = vcn_v2_5_resume,
         .is_idle = vcn_v2_5_is_idle,
         .wait_for_idle = vcn_v2_5_wait_for_idle,
-        .check_soft_reset = NULL,
-        .pre_soft_reset = NULL,
-        .soft_reset = NULL,
-        .post_soft_reset = NULL,
         .set_clockgating_state = vcn_v2_5_set_clockgating_state,
-        .set_powergating_state = vcn_v2_5_set_powergating_state,
+	.set_powergating_state = vcn_set_powergating_state,
+	.dump_ip_state = amdgpu_vcn_dump_ip_state,
+	.print_ip_state = amdgpu_vcn_print_ip_state,
 };
 
 const struct amdgpu_ip_block_version vcn_v2_5_ip_block =
@@ -1965,12 +2192,13 @@ const struct amdgpu_ras_block_hw_ops vcn_v2_6_ras_hw_ops = {
 static struct amdgpu_vcn_ras vcn_v2_6_ras = {
 	.ras_block = {
 		.hw_ops = &vcn_v2_6_ras_hw_ops,
+		.ras_late_init = amdgpu_vcn_ras_late_init,
 	},
 };
 
 static void vcn_v2_5_set_ras_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[VCN_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, VCN_HWIP, 0)) {
 	case IP_VERSION(2, 6, 0):
 		adev->vcn.ras = &vcn_v2_6_ras;
 		break;
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
index 3eab186261aa..d9cf8f0feeb3 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
@@ -40,6 +40,7 @@
 
 #define VCN_VID_SOC_ADDRESS_2_0					0x1fa00
 #define VCN1_VID_SOC_ADDRESS_3_0				0x48200
+#define VCN1_AON_SOC_ADDRESS_3_0				0x48000
 
 #define mmUVD_CONTEXT_ID_INTERNAL_OFFSET			0x27
 #define mmUVD_GPCOM_VCPU_CMD_INTERNAL_OFFSET			0x0f
@@ -60,6 +61,42 @@
 #define RDECODE_MSG_CREATE					0x00000000
 #define RDECODE_MESSAGE_CREATE					0x00000001
 
+static const struct amdgpu_hwip_reg_entry vcn_reg_list_3_0[] = {
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_POWER_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_CONTEXT_ID),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_CONTEXT_ID2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_GPCOM_VCPU_DATA0),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_GPCOM_VCPU_DATA1),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_GPCOM_VCPU_CMD),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE2),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE3),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE4),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_PGFSM_CONFIG),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_PGFSM_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_LMA_CTL),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_LMA_DATA),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_LMA_MASK),
+	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_PAUSE)
+};
+
 static int amdgpu_ih_clientid_vcns[] = {
 	SOC15_IH_CLIENTID_VCN,
 	SOC15_IH_CLIENTID_VCN1
@@ -69,10 +106,11 @@ static int vcn_v3_0_start_sriov(struct amdgpu_device *adev);
 static void vcn_v3_0_set_dec_ring_funcs(struct amdgpu_device *adev);
 static void vcn_v3_0_set_enc_ring_funcs(struct amdgpu_device *adev);
 static void vcn_v3_0_set_irq_funcs(struct amdgpu_device *adev);
-static int vcn_v3_0_set_powergating_state(void *handle,
-			enum amd_powergating_state state);
-static int vcn_v3_0_pause_dpg_mode(struct amdgpu_device *adev,
-			int inst_idx, struct dpg_pause_state *new_state);
+static int vcn_v3_0_set_pg_state(struct amdgpu_vcn_inst *vinst,
+				 enum amd_powergating_state state);
+static int vcn_v3_0_pause_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				   struct dpg_pause_state *new_state);
+static int vcn_v3_0_reset(struct amdgpu_vcn_inst *vinst);
 
 static void vcn_v3_0_dec_ring_set_wptr(struct amdgpu_ring *ring);
 static void vcn_v3_0_enc_ring_set_wptr(struct amdgpu_ring *ring);
@@ -80,19 +118,21 @@ static void vcn_v3_0_enc_ring_set_wptr(struct amdgpu_ring *ring);
 /**
  * vcn_v3_0_early_init - set function pointers and load microcode
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Set ring and irq function pointers
  * Load microcode from filesystem
  */
-static int vcn_v3_0_early_init(void *handle)
+static int vcn_v3_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, r;
 
 	if (amdgpu_sriov_vf(adev)) {
 		adev->vcn.num_vcn_inst = VCN_INSTANCES_SIENNA_CICHLID;
 		adev->vcn.harvest_config = 0;
-		adev->vcn.num_enc_rings = 1;
+		for (i = 0; i < adev->vcn.num_vcn_inst; i++)
+			adev->vcn.inst[i].num_enc_rings = 1;
 
 	} else {
 		if (adev->vcn.harvest_config == (AMDGPU_VCN_HARVEST_VCN0 |
@@ -100,42 +140,42 @@ static int vcn_v3_0_early_init(void *handle)
 			/* both instances are harvested, disable the block */
 			return -ENOENT;
 
-		if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(3, 0, 33))
-			adev->vcn.num_enc_rings = 0;
-		else
-			adev->vcn.num_enc_rings = 2;
+		for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+			if (amdgpu_ip_version(adev, UVD_HWIP, 0) ==
+			    IP_VERSION(3, 0, 33))
+				adev->vcn.inst[i].num_enc_rings = 0;
+			else
+				adev->vcn.inst[i].num_enc_rings = 2;
+		}
 	}
 
 	vcn_v3_0_set_dec_ring_funcs(adev);
 	vcn_v3_0_set_enc_ring_funcs(adev);
 	vcn_v3_0_set_irq_funcs(adev);
 
-	return amdgpu_vcn_early_init(adev);
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		adev->vcn.inst[i].set_pg_state = vcn_v3_0_set_pg_state;
+
+		r = amdgpu_vcn_early_init(adev, i);
+		if (r)
+			return r;
+	}
+	return 0;
 }
 
 /**
  * vcn_v3_0_sw_init - sw init for VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Load firmware and sw initialization
  */
-static int vcn_v3_0_sw_init(void *handle)
+static int vcn_v3_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
 	int i, j, r;
 	int vcn_doorbell_index = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	r = amdgpu_vcn_sw_init(adev);
-	if (r)
-		return r;
-
-	amdgpu_vcn_setup_ucode(adev);
-
-	r = amdgpu_vcn_resume(adev);
-	if (r)
-		return r;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/*
 	 * Note: doorbell assignment is fixed for SRIOV multiple VCN engines
@@ -151,27 +191,37 @@ static int vcn_v3_0_sw_init(void *handle)
 	}
 
 	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
-		volatile struct amdgpu_fw_shared *fw_shared;
+		struct amdgpu_fw_shared *fw_shared;
 
 		if (adev->vcn.harvest_config & (1 << i))
 			continue;
 
-		adev->vcn.internal.context_id = mmUVD_CONTEXT_ID_INTERNAL_OFFSET;
-		adev->vcn.internal.ib_vmid = mmUVD_LMI_RBC_IB_VMID_INTERNAL_OFFSET;
-		adev->vcn.internal.ib_bar_low = mmUVD_LMI_RBC_IB_64BIT_BAR_LOW_INTERNAL_OFFSET;
-		adev->vcn.internal.ib_bar_high = mmUVD_LMI_RBC_IB_64BIT_BAR_HIGH_INTERNAL_OFFSET;
-		adev->vcn.internal.ib_size = mmUVD_RBC_IB_SIZE_INTERNAL_OFFSET;
-		adev->vcn.internal.gp_scratch8 = mmUVD_GP_SCRATCH8_INTERNAL_OFFSET;
+		r = amdgpu_vcn_sw_init(adev, i);
+		if (r)
+			return r;
+
+		amdgpu_vcn_setup_ucode(adev, i);
+
+		r = amdgpu_vcn_resume(adev, i);
+		if (r)
+			return r;
+
+		adev->vcn.inst[i].internal.context_id = mmUVD_CONTEXT_ID_INTERNAL_OFFSET;
+		adev->vcn.inst[i].internal.ib_vmid = mmUVD_LMI_RBC_IB_VMID_INTERNAL_OFFSET;
+		adev->vcn.inst[i].internal.ib_bar_low = mmUVD_LMI_RBC_IB_64BIT_BAR_LOW_INTERNAL_OFFSET;
+		adev->vcn.inst[i].internal.ib_bar_high = mmUVD_LMI_RBC_IB_64BIT_BAR_HIGH_INTERNAL_OFFSET;
+		adev->vcn.inst[i].internal.ib_size = mmUVD_RBC_IB_SIZE_INTERNAL_OFFSET;
+		adev->vcn.inst[i].internal.gp_scratch8 = mmUVD_GP_SCRATCH8_INTERNAL_OFFSET;
 
-		adev->vcn.internal.scratch9 = mmUVD_SCRATCH9_INTERNAL_OFFSET;
+		adev->vcn.inst[i].internal.scratch9 = mmUVD_SCRATCH9_INTERNAL_OFFSET;
 		adev->vcn.inst[i].external.scratch9 = SOC15_REG_OFFSET(VCN, i, mmUVD_SCRATCH9);
-		adev->vcn.internal.data0 = mmUVD_GPCOM_VCPU_DATA0_INTERNAL_OFFSET;
+		adev->vcn.inst[i].internal.data0 = mmUVD_GPCOM_VCPU_DATA0_INTERNAL_OFFSET;
 		adev->vcn.inst[i].external.data0 = SOC15_REG_OFFSET(VCN, i, mmUVD_GPCOM_VCPU_DATA0);
-		adev->vcn.internal.data1 = mmUVD_GPCOM_VCPU_DATA1_INTERNAL_OFFSET;
+		adev->vcn.inst[i].internal.data1 = mmUVD_GPCOM_VCPU_DATA1_INTERNAL_OFFSET;
 		adev->vcn.inst[i].external.data1 = SOC15_REG_OFFSET(VCN, i, mmUVD_GPCOM_VCPU_DATA1);
-		adev->vcn.internal.cmd = mmUVD_GPCOM_VCPU_CMD_INTERNAL_OFFSET;
+		adev->vcn.inst[i].internal.cmd = mmUVD_GPCOM_VCPU_CMD_INTERNAL_OFFSET;
 		adev->vcn.inst[i].external.cmd = SOC15_REG_OFFSET(VCN, i, mmUVD_GPCOM_VCPU_CMD);
-		adev->vcn.internal.nop = mmUVD_NO_OP_INTERNAL_OFFSET;
+		adev->vcn.inst[i].internal.nop = mmUVD_NO_OP_INTERNAL_OFFSET;
 		adev->vcn.inst[i].external.nop = SOC15_REG_OFFSET(VCN, i, mmUVD_NO_OP);
 
 		/* VCN DEC TRAP */
@@ -185,11 +235,11 @@ static int vcn_v3_0_sw_init(void *handle)
 		ring = &adev->vcn.inst[i].ring_dec;
 		ring->use_doorbell = true;
 		if (amdgpu_sriov_vf(adev)) {
-			ring->doorbell_index = vcn_doorbell_index + i * (adev->vcn.num_enc_rings + 1);
+			ring->doorbell_index = vcn_doorbell_index + i * (adev->vcn.inst[i].num_enc_rings + 1);
 		} else {
 			ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 8 * i;
 		}
-		ring->vm_hub = AMDGPU_MMHUB_0;
+		ring->vm_hub = AMDGPU_MMHUB0(0);
 		sprintf(ring->name, "vcn_dec_%d", i);
 		r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[i].irq, 0,
 				     AMDGPU_RING_PRIO_DEFAULT,
@@ -197,7 +247,7 @@ static int vcn_v3_0_sw_init(void *handle)
 		if (r)
 			return r;
 
-		for (j = 0; j < adev->vcn.num_enc_rings; ++j) {
+		for (j = 0; j < adev->vcn.inst[i].num_enc_rings; ++j) {
 			enum amdgpu_ring_priority_level hw_prio = amdgpu_vcn_get_enc_ring_prio(j);
 
 			/* VCN ENC TRAP */
@@ -209,11 +259,11 @@ static int vcn_v3_0_sw_init(void *handle)
 			ring = &adev->vcn.inst[i].ring_enc[j];
 			ring->use_doorbell = true;
 			if (amdgpu_sriov_vf(adev)) {
-				ring->doorbell_index = vcn_doorbell_index + i * (adev->vcn.num_enc_rings + 1) + 1 + j;
+				ring->doorbell_index = vcn_doorbell_index + i * (adev->vcn.inst[i].num_enc_rings + 1) + 1 + j;
 			} else {
 				ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 2 + j + 8 * i;
 			}
-			ring->vm_hub = AMDGPU_MMHUB_0;
+			ring->vm_hub = AMDGPU_MMHUB0(0);
 			sprintf(ring->name, "vcn_enc_%d.%d", i, j);
 			r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[i].irq, 0,
 					     hw_prio, &adev->vcn.inst[i].sched_score);
@@ -227,22 +277,38 @@ static int vcn_v3_0_sw_init(void *handle)
 					     cpu_to_le32(AMDGPU_VCN_FW_SHARED_FLAG_0_RB);
 		fw_shared->sw_ring.is_enabled = cpu_to_le32(DEC_SW_RING_ENABLED);
 		fw_shared->present_flag_0 |= AMDGPU_VCN_SMU_VERSION_INFO_FLAG;
-		if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(3, 1, 2))
+		if (amdgpu_ip_version(adev, UVD_HWIP, 0) == IP_VERSION(3, 1, 2))
 			fw_shared->smu_interface_info.smu_interface_type = 2;
-		else if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(3, 1, 1))
+		else if (amdgpu_ip_version(adev, UVD_HWIP, 0) ==
+			 IP_VERSION(3, 1, 1))
 			fw_shared->smu_interface_info.smu_interface_type = 1;
 
 		if (amdgpu_vcnfw_log)
 			amdgpu_vcn_fwlog_init(&adev->vcn.inst[i]);
+
+		if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
+			adev->vcn.inst[i].pause_dpg_mode = vcn_v3_0_pause_dpg_mode;
+		adev->vcn.inst[i].reset = vcn_v3_0_reset;
 	}
 
+	adev->vcn.supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->vcn.inst[0].ring_enc[0]);
+	if (!amdgpu_sriov_vf(adev))
+		adev->vcn.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+
 	if (amdgpu_sriov_vf(adev)) {
 		r = amdgpu_virt_alloc_mm_table(adev);
 		if (r)
 			return r;
 	}
-	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
-		adev->vcn.pause_dpg_mode = vcn_v3_0_pause_dpg_mode;
+
+	r = amdgpu_vcn_reg_dump_init(adev, vcn_reg_list_3_0, ARRAY_SIZE(vcn_reg_list_3_0));
+	if (r)
+		return r;
+
+	r = amdgpu_vcn_sysfs_reset_mask_init(adev);
+	if (r)
+		return r;
 
 	return 0;
 }
@@ -250,18 +316,18 @@ static int vcn_v3_0_sw_init(void *handle)
 /**
  * vcn_v3_0_sw_fini - sw fini for VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * VCN suspend and free up sw allocation
  */
-static int vcn_v3_0_sw_fini(void *handle)
+static int vcn_v3_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i, r, idx;
 
 	if (drm_dev_enter(adev_to_drm(adev), &idx)) {
 		for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
-			volatile struct amdgpu_fw_shared *fw_shared;
+			struct amdgpu_fw_shared *fw_shared;
 
 			if (adev->vcn.harvest_config & (1 << i))
 				continue;
@@ -276,32 +342,36 @@ static int vcn_v3_0_sw_fini(void *handle)
 	if (amdgpu_sriov_vf(adev))
 		amdgpu_virt_free_mm_table(adev);
 
-	r = amdgpu_vcn_suspend(adev);
-	if (r)
-		return r;
+	amdgpu_vcn_sysfs_reset_mask_fini(adev);
 
-	r = amdgpu_vcn_sw_fini(adev);
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_suspend(adev, i);
+		if (r)
+			return r;
 
-	return r;
+		amdgpu_vcn_sw_fini(adev, i);
+	}
+
+	return 0;
 }
 
 /**
  * vcn_v3_0_hw_init - start and test VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Initialize the hardware, boot up the VCPU and do some testing
  */
-static int vcn_v3_0_hw_init(void *handle)
+static int vcn_v3_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring;
 	int i, j, r;
 
 	if (amdgpu_sriov_vf(adev)) {
 		r = vcn_v3_0_start_sriov(adev);
 		if (r)
-			goto done;
+			return r;
 
 		/* initialize VCN dec and enc ring buffers */
 		for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
@@ -320,7 +390,7 @@ static int vcn_v3_0_hw_init(void *handle)
 				ring->sched.ready = true;
 			}
 
-			for (j = 0; j < adev->vcn.num_enc_rings; ++j) {
+			for (j = 0; j < adev->vcn.inst[i].num_enc_rings; ++j) {
 				ring = &adev->vcn.inst[i].ring_enc[j];
 				if (amdgpu_vcn_is_disabled_vcn(adev, VCN_ENCODE_RING, i)) {
 					ring->sched.ready = false;
@@ -346,48 +416,45 @@ static int vcn_v3_0_hw_init(void *handle)
 
 			r = amdgpu_ring_test_helper(ring);
 			if (r)
-				goto done;
+				return r;
 
-			for (j = 0; j < adev->vcn.num_enc_rings; ++j) {
+			for (j = 0; j < adev->vcn.inst[i].num_enc_rings; ++j) {
 				ring = &adev->vcn.inst[i].ring_enc[j];
 				r = amdgpu_ring_test_helper(ring);
 				if (r)
-					goto done;
+					return r;
 			}
 		}
 	}
 
-done:
-	if (!r)
-		DRM_INFO("VCN decode and encode initialized successfully(under %s).\n",
-			(adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)?"DPG Mode":"SPG Mode");
-
-	return r;
+	return 0;
 }
 
 /**
  * vcn_v3_0_hw_fini - stop the hardware block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Stop the VCN block, mark ring as not ready any more
  */
-static int vcn_v3_0_hw_fini(void *handle)
+static int vcn_v3_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
-	cancel_delayed_work_sync(&adev->vcn.idle_work);
-
 	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[i];
+
 		if (adev->vcn.harvest_config & (1 << i))
 			continue;
 
+		cancel_delayed_work_sync(&vinst->idle_work);
+
 		if (!amdgpu_sriov_vf(adev)) {
 			if ((adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) ||
-					(adev->vcn.cur_state != AMD_PG_STATE_GATE &&
-					 RREG32_SOC15(VCN, i, mmUVD_STATUS))) {
-				vcn_v3_0_set_powergating_state(adev, AMD_PG_STATE_GATE);
+			    (vinst->cur_state != AMD_PG_STATE_GATE &&
+			     RREG32_SOC15(VCN, i, mmUVD_STATUS))) {
+				vinst->set_pg_state(vinst, AMD_PG_STATE_GATE);
 			}
 		}
 	}
@@ -398,41 +465,47 @@ static int vcn_v3_0_hw_fini(void *handle)
 /**
  * vcn_v3_0_suspend - suspend VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * HW fini and suspend VCN block
  */
-static int vcn_v3_0_suspend(void *handle)
+static int vcn_v3_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int r, i;
 
-	r = vcn_v3_0_hw_fini(adev);
+	r = vcn_v3_0_hw_fini(ip_block);
 	if (r)
 		return r;
 
-	r = amdgpu_vcn_suspend(adev);
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_suspend(ip_block->adev, i);
+		if (r)
+			return r;
+	}
 
-	return r;
+	return 0;
 }
 
 /**
  * vcn_v3_0_resume - resume VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Resume firmware and hw init VCN block
  */
-static int vcn_v3_0_resume(void *handle)
+static int vcn_v3_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int r, i;
 
-	r = amdgpu_vcn_resume(adev);
-	if (r)
-		return r;
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_resume(ip_block->adev, i);
+		if (r)
+			return r;
+	}
 
-	r = vcn_v3_0_hw_init(adev);
+	r = vcn_v3_0_hw_init(ip_block);
 
 	return r;
 }
@@ -440,14 +513,15 @@ static int vcn_v3_0_resume(void *handle)
 /**
  * vcn_v3_0_mc_resume - memory controller programming
  *
- * @adev: amdgpu_device pointer
- * @inst: instance number
+ * @vinst: VCN instance
  *
  * Let the VCN memory controller know it's offsets
  */
-static void vcn_v3_0_mc_resume(struct amdgpu_device *adev, int inst)
+static void vcn_v3_0_mc_resume(struct amdgpu_vcn_inst *vinst)
 {
-	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
+	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.inst[inst].fw->size + 4);
 	uint32_t offset;
 
 	/* cache window 0: fw */
@@ -495,9 +569,12 @@ static void vcn_v3_0_mc_resume(struct amdgpu_device *adev, int inst)
 		AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_fw_shared)));
 }
 
-static void vcn_v3_0_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_idx, bool indirect)
+static void vcn_v3_0_mc_resume_dpg_mode(struct amdgpu_vcn_inst *vinst,
+					bool indirect)
 {
-	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.inst[inst_idx].fw->size + 4);
 	uint32_t offset;
 
 	/* cache window 0: fw */
@@ -591,8 +668,10 @@ static void vcn_v3_0_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_idx
 		UVD, inst_idx, mmUVD_GFX10_ADDR_CONFIG), adev->gfx.config.gb_addr_config, 0, indirect);
 }
 
-static void vcn_v3_0_disable_static_power_gating(struct amdgpu_device *adev, int inst)
+static void vcn_v3_0_disable_static_power_gating(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
 	uint32_t data = 0;
 
 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN) {
@@ -642,8 +721,10 @@ static void vcn_v3_0_disable_static_power_gating(struct amdgpu_device *adev, int
 	WREG32_SOC15(VCN, inst, mmUVD_POWER_STATUS, data);
 }
 
-static void vcn_v3_0_enable_static_power_gating(struct amdgpu_device *adev, int inst)
+static void vcn_v3_0_enable_static_power_gating(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
 	uint32_t data;
 
 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN) {
@@ -690,13 +771,14 @@ static void vcn_v3_0_enable_static_power_gating(struct amdgpu_device *adev, int
 /**
  * vcn_v3_0_disable_clock_gating - disable VCN clock gating
  *
- * @adev: amdgpu_device pointer
- * @inst: instance number
+ * @vinst: Pointer to the VCN instance structure
  *
  * Disable clock gating for VCN block
  */
-static void vcn_v3_0_disable_clock_gating(struct amdgpu_device *adev, int inst)
+static void vcn_v3_0_disable_clock_gating(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
 	uint32_t data;
 
 	/* VCN disable CGC */
@@ -823,9 +905,12 @@ static void vcn_v3_0_disable_clock_gating(struct amdgpu_device *adev, int inst)
 	WREG32_SOC15(VCN, inst, mmUVD_SUVD_CGC_CTRL, data);
 }
 
-static void vcn_v3_0_clock_gating_dpg_mode(struct amdgpu_device *adev,
-		uint8_t sram_sel, int inst_idx, uint8_t indirect)
+static void vcn_v3_0_clock_gating_dpg_mode(struct amdgpu_vcn_inst *vinst,
+					   uint8_t sram_sel,
+					   uint8_t indirect)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
 	uint32_t reg_data = 0;
 
 	/* enable sw clock gating control */
@@ -874,13 +959,14 @@ static void vcn_v3_0_clock_gating_dpg_mode(struct amdgpu_device *adev,
 /**
  * vcn_v3_0_enable_clock_gating - enable VCN clock gating
  *
- * @adev: amdgpu_device pointer
- * @inst: instance number
+ * @vinst: Pointer to the VCN instance structure
  *
  * Enable clock gating for VCN block
  */
-static void vcn_v3_0_enable_clock_gating(struct amdgpu_device *adev, int inst)
+static void vcn_v3_0_enable_clock_gating(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
 	uint32_t data;
 
 	/* enable VCN CGC */
@@ -939,11 +1025,14 @@ static void vcn_v3_0_enable_clock_gating(struct amdgpu_device *adev, int inst)
 	WREG32_SOC15(VCN, inst, mmUVD_SUVD_CGC_CTRL, data);
 }
 
-static int vcn_v3_0_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, bool indirect)
+static int vcn_v3_0_start_dpg_mode(struct amdgpu_vcn_inst *vinst, bool indirect)
 {
-	volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr;
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	struct amdgpu_fw_shared *fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr;
 	struct amdgpu_ring *ring;
 	uint32_t rb_bufsz, tmp;
+	int ret;
 
 	/* disable register anti-hang mechanism */
 	WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, mmUVD_POWER_STATUS), 1,
@@ -958,7 +1047,7 @@ static int vcn_v3_0_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, boo
 		adev->vcn.inst[inst_idx].dpg_sram_curr_addr = (uint32_t *)adev->vcn.inst[inst_idx].dpg_sram_cpu_addr;
 
 	/* enable clock gating */
-	vcn_v3_0_clock_gating_dpg_mode(adev, 0, inst_idx, indirect);
+	vcn_v3_0_clock_gating_dpg_mode(vinst, 0, indirect);
 
 	/* enable VCPU clock */
 	tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
@@ -1007,7 +1096,7 @@ static int vcn_v3_0_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, boo
 		 (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
 		 (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)), 0, indirect);
 
-	vcn_v3_0_mc_resume_dpg_mode(adev, inst_idx, indirect);
+	vcn_v3_0_mc_resume_dpg_mode(vinst, indirect);
 
 	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
 		VCN, inst_idx, mmUVD_REG_XX_MASK), 0x10, 0, indirect);
@@ -1036,10 +1125,13 @@ static int vcn_v3_0_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, boo
 	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
 		VCN, inst_idx, mmUVD_VCPU_CNTL), tmp, 0, indirect);
 
-	if (indirect)
-		psp_update_vcn_sram(adev, inst_idx, adev->vcn.inst[inst_idx].dpg_sram_gpu_addr,
-			(uint32_t)((uintptr_t)adev->vcn.inst[inst_idx].dpg_sram_curr_addr -
-				(uintptr_t)adev->vcn.inst[inst_idx].dpg_sram_cpu_addr));
+	if (indirect) {
+		ret = amdgpu_vcn_psp_update_sram(adev, inst_idx, 0);
+		if (ret) {
+			dev_err(adev->dev, "vcn sram load failed %d\n", ret);
+			return ret;
+		}
+	}
 
 	ring = &adev->vcn.inst[inst_idx].ring_dec;
 	/* force RBC into idle state */
@@ -1090,194 +1182,203 @@ static int vcn_v3_0_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, boo
 	WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, mmUVD_POWER_STATUS),
 		0, ~UVD_POWER_STATUS__STALL_DPG_POWER_UP_MASK);
 
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, inst_idx, mmUVD_STATUS);
+
 	return 0;
 }
 
-static int vcn_v3_0_start(struct amdgpu_device *adev)
+static int vcn_v3_0_start(struct amdgpu_vcn_inst *vinst)
 {
-	volatile struct amdgpu_fw_shared *fw_shared;
+	struct amdgpu_device *adev = vinst->adev;
+	int i = vinst->inst;
+	struct amdgpu_fw_shared *fw_shared;
 	struct amdgpu_ring *ring;
 	uint32_t rb_bufsz, tmp;
-	int i, j, k, r;
+	int j, k, r;
+
+	if (adev->vcn.harvest_config & (1 << i))
+		return 0;
 
 	if (adev->pm.dpm_enabled)
-		amdgpu_dpm_enable_uvd(adev, true);
+		amdgpu_dpm_enable_vcn(adev, true, i);
 
-	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
-		if (adev->vcn.harvest_config & (1 << i))
-			continue;
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
+		return vcn_v3_0_start_dpg_mode(vinst, vinst->indirect_sram);
 
-		if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG){
-			r = vcn_v3_0_start_dpg_mode(adev, i, adev->vcn.indirect_sram);
-			continue;
-		}
+	/* disable VCN power gating */
+	vcn_v3_0_disable_static_power_gating(vinst);
 
-		/* disable VCN power gating */
-		vcn_v3_0_disable_static_power_gating(adev, i);
+	/* set VCN status busy */
+	tmp = RREG32_SOC15(VCN, i, mmUVD_STATUS) | UVD_STATUS__UVD_BUSY;
+	WREG32_SOC15(VCN, i, mmUVD_STATUS, tmp);
 
-		/* set VCN status busy */
-		tmp = RREG32_SOC15(VCN, i, mmUVD_STATUS) | UVD_STATUS__UVD_BUSY;
-		WREG32_SOC15(VCN, i, mmUVD_STATUS, tmp);
+	/* SW clock gating */
+	vcn_v3_0_disable_clock_gating(vinst);
 
-		/*SW clock gating */
-		vcn_v3_0_disable_clock_gating(adev, i);
+	/* enable VCPU clock */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL),
+		 UVD_VCPU_CNTL__CLK_EN_MASK, ~UVD_VCPU_CNTL__CLK_EN_MASK);
 
-		/* enable VCPU clock */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL),
-			UVD_VCPU_CNTL__CLK_EN_MASK, ~UVD_VCPU_CNTL__CLK_EN_MASK);
-
-		/* disable master interrupt */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_MASTINT_EN), 0,
-			~UVD_MASTINT_EN__VCPU_EN_MASK);
-
-		/* enable LMI MC and UMC channels */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_LMI_CTRL2), 0,
-			~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
-
-		tmp = RREG32_SOC15(VCN, i, mmUVD_SOFT_RESET);
-		tmp &= ~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
-		tmp &= ~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
-		WREG32_SOC15(VCN, i, mmUVD_SOFT_RESET, tmp);
-
-		/* setup mmUVD_LMI_CTRL */
-		tmp = RREG32_SOC15(VCN, i, mmUVD_LMI_CTRL);
-		WREG32_SOC15(VCN, i, mmUVD_LMI_CTRL, tmp |
-			UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK	|
-			UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
-			UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
-			UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK);
-
-		/* setup mmUVD_MPC_CNTL */
-		tmp = RREG32_SOC15(VCN, i, mmUVD_MPC_CNTL);
-		tmp &= ~UVD_MPC_CNTL__REPLACEMENT_MODE_MASK;
-		tmp |= 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT;
-		WREG32_SOC15(VCN, i, mmUVD_MPC_CNTL, tmp);
-
-		/* setup UVD_MPC_SET_MUXA0 */
-		WREG32_SOC15(VCN, i, mmUVD_MPC_SET_MUXA0,
-			((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) |
-			(0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) |
-			(0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) |
-			(0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)));
-
-		/* setup UVD_MPC_SET_MUXB0 */
-		WREG32_SOC15(VCN, i, mmUVD_MPC_SET_MUXB0,
-			((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) |
-			(0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) |
-			(0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) |
-			(0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)));
-
-		/* setup mmUVD_MPC_SET_MUX */
-		WREG32_SOC15(VCN, i, mmUVD_MPC_SET_MUX,
-			((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) |
-			(0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
-			(0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)));
-
-		vcn_v3_0_mc_resume(adev, i);
-
-		/* VCN global tiling registers */
-		WREG32_SOC15(VCN, i, mmUVD_GFX10_ADDR_CONFIG,
-			adev->gfx.config.gb_addr_config);
-
-		/* unblock VCPU register access */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_RB_ARB_CTRL), 0,
-			~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
-
-		/* release VCPU reset to boot */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL), 0,
-			~UVD_VCPU_CNTL__BLK_RST_MASK);
+	/* disable master interrupt */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_MASTINT_EN), 0,
+		 ~UVD_MASTINT_EN__VCPU_EN_MASK);
 
-		for (j = 0; j < 10; ++j) {
-			uint32_t status;
+	/* enable LMI MC and UMC channels */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_LMI_CTRL2), 0,
+		 ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
 
-			for (k = 0; k < 100; ++k) {
-				status = RREG32_SOC15(VCN, i, mmUVD_STATUS);
-				if (status & 2)
-					break;
-				mdelay(10);
-			}
-			r = 0;
-			if (status & 2)
-				break;
+	tmp = RREG32_SOC15(VCN, i, mmUVD_SOFT_RESET);
+	tmp &= ~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
+	tmp &= ~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
+	WREG32_SOC15(VCN, i, mmUVD_SOFT_RESET, tmp);
 
-			DRM_ERROR("VCN[%d] decode not responding, trying to reset the VCPU!!!\n", i);
-			WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL),
-				UVD_VCPU_CNTL__BLK_RST_MASK,
-				~UVD_VCPU_CNTL__BLK_RST_MASK);
-			mdelay(10);
-			WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL), 0,
-				~UVD_VCPU_CNTL__BLK_RST_MASK);
+	/* setup mmUVD_LMI_CTRL */
+	tmp = RREG32_SOC15(VCN, i, mmUVD_LMI_CTRL);
+	WREG32_SOC15(VCN, i, mmUVD_LMI_CTRL, tmp |
+		     UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK	|
+		     UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
+		     UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
+		     UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK);
+
+	/* setup mmUVD_MPC_CNTL */
+	tmp = RREG32_SOC15(VCN, i, mmUVD_MPC_CNTL);
+	tmp &= ~UVD_MPC_CNTL__REPLACEMENT_MODE_MASK;
+	tmp |= 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT;
+	WREG32_SOC15(VCN, i, mmUVD_MPC_CNTL, tmp);
+
+	/* setup UVD_MPC_SET_MUXA0 */
+	WREG32_SOC15(VCN, i, mmUVD_MPC_SET_MUXA0,
+		     ((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) |
+		      (0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) |
+		      (0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) |
+		      (0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)));
+
+	/* setup UVD_MPC_SET_MUXB0 */
+	WREG32_SOC15(VCN, i, mmUVD_MPC_SET_MUXB0,
+		     ((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) |
+		      (0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) |
+		      (0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) |
+		      (0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)));
+
+	/* setup mmUVD_MPC_SET_MUX */
+	WREG32_SOC15(VCN, i, mmUVD_MPC_SET_MUX,
+		     ((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) |
+		      (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
+		      (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)));
+
+	vcn_v3_0_mc_resume(vinst);
 
+	/* VCN global tiling registers */
+	WREG32_SOC15(VCN, i, mmUVD_GFX10_ADDR_CONFIG,
+		     adev->gfx.config.gb_addr_config);
+
+	/* unblock VCPU register access */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_RB_ARB_CTRL), 0,
+		 ~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
+
+	/* release VCPU reset to boot */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL), 0,
+		 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+
+	for (j = 0; j < 10; ++j) {
+		uint32_t status;
+
+		for (k = 0; k < 100; ++k) {
+			status = RREG32_SOC15(VCN, i, mmUVD_STATUS);
+			if (status & 2)
+				break;
 			mdelay(10);
-			r = -1;
 		}
+		r = 0;
+		if (status & 2)
+			break;
 
-		if (r) {
-			DRM_ERROR("VCN[%d] decode not responding, giving up!!!\n", i);
-			return r;
-		}
+		DRM_ERROR("VCN[%d] decode not responding, trying to reset the VCPU!!!\n", i);
+		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL),
+			 UVD_VCPU_CNTL__BLK_RST_MASK,
+			 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+		mdelay(10);
+		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL), 0,
+			 ~UVD_VCPU_CNTL__BLK_RST_MASK);
 
-		/* enable master interrupt */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_MASTINT_EN),
-			UVD_MASTINT_EN__VCPU_EN_MASK,
-			~UVD_MASTINT_EN__VCPU_EN_MASK);
+		mdelay(10);
+		r = -1;
+	}
 
-		/* clear the busy bit of VCN_STATUS */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_STATUS), 0,
-			~(2 << UVD_STATUS__VCPU_REPORT__SHIFT));
+	if (r) {
+		DRM_ERROR("VCN[%d] decode not responding, giving up!!!\n", i);
+		return r;
+	}
 
-		WREG32_SOC15(VCN, i, mmUVD_LMI_RBC_RB_VMID, 0);
+	/* enable master interrupt */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_MASTINT_EN),
+		 UVD_MASTINT_EN__VCPU_EN_MASK,
+		 ~UVD_MASTINT_EN__VCPU_EN_MASK);
 
-		ring = &adev->vcn.inst[i].ring_dec;
-		/* force RBC into idle state */
-		rb_bufsz = order_base_2(ring->ring_size);
-		tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz);
-		tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1);
-		tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1);
-		tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1);
-		tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1);
-		WREG32_SOC15(VCN, i, mmUVD_RBC_RB_CNTL, tmp);
+	/* clear the busy bit of VCN_STATUS */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_STATUS), 0,
+		 ~(2 << UVD_STATUS__VCPU_REPORT__SHIFT));
 
-		fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
-		fw_shared->multi_queue.decode_queue_mode |= cpu_to_le32(FW_QUEUE_RING_RESET);
+	WREG32_SOC15(VCN, i, mmUVD_LMI_RBC_RB_VMID, 0);
+
+	ring = &adev->vcn.inst[i].ring_dec;
+	/* force RBC into idle state */
+	rb_bufsz = order_base_2(ring->ring_size);
+	tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz);
+	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1);
+	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1);
+	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1);
+	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1);
+	WREG32_SOC15(VCN, i, mmUVD_RBC_RB_CNTL, tmp);
 
-		/* programm the RB_BASE for ring buffer */
-		WREG32_SOC15(VCN, i, mmUVD_LMI_RBC_RB_64BIT_BAR_LOW,
-			lower_32_bits(ring->gpu_addr));
-		WREG32_SOC15(VCN, i, mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH,
-			upper_32_bits(ring->gpu_addr));
+	fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
+	fw_shared->multi_queue.decode_queue_mode |= cpu_to_le32(FW_QUEUE_RING_RESET);
 
-		/* Initialize the ring buffer's read and write pointers */
-		WREG32_SOC15(VCN, i, mmUVD_RBC_RB_RPTR, 0);
+	/* programm the RB_BASE for ring buffer */
+	WREG32_SOC15(VCN, i, mmUVD_LMI_RBC_RB_64BIT_BAR_LOW,
+		     lower_32_bits(ring->gpu_addr));
+	WREG32_SOC15(VCN, i, mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH,
+		     upper_32_bits(ring->gpu_addr));
 
-		WREG32_SOC15(VCN, i, mmUVD_SCRATCH2, 0);
-		ring->wptr = RREG32_SOC15(VCN, i, mmUVD_RBC_RB_RPTR);
-		WREG32_SOC15(VCN, i, mmUVD_RBC_RB_WPTR,
-			lower_32_bits(ring->wptr));
-		fw_shared->rb.wptr = lower_32_bits(ring->wptr);
-		fw_shared->multi_queue.decode_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET);
-
-		if (adev->ip_versions[UVD_HWIP][0] != IP_VERSION(3, 0, 33)) {
-			fw_shared->multi_queue.encode_generalpurpose_queue_mode |= cpu_to_le32(FW_QUEUE_RING_RESET);
-			ring = &adev->vcn.inst[i].ring_enc[0];
-			WREG32_SOC15(VCN, i, mmUVD_RB_RPTR, lower_32_bits(ring->wptr));
-			WREG32_SOC15(VCN, i, mmUVD_RB_WPTR, lower_32_bits(ring->wptr));
-			WREG32_SOC15(VCN, i, mmUVD_RB_BASE_LO, ring->gpu_addr);
-			WREG32_SOC15(VCN, i, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
-			WREG32_SOC15(VCN, i, mmUVD_RB_SIZE, ring->ring_size / 4);
-			fw_shared->multi_queue.encode_generalpurpose_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET);
-
-			fw_shared->multi_queue.encode_lowlatency_queue_mode |= cpu_to_le32(FW_QUEUE_RING_RESET);
-			ring = &adev->vcn.inst[i].ring_enc[1];
-			WREG32_SOC15(VCN, i, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr));
-			WREG32_SOC15(VCN, i, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr));
-			WREG32_SOC15(VCN, i, mmUVD_RB_BASE_LO2, ring->gpu_addr);
-			WREG32_SOC15(VCN, i, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr));
-			WREG32_SOC15(VCN, i, mmUVD_RB_SIZE2, ring->ring_size / 4);
-			fw_shared->multi_queue.encode_lowlatency_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET);
-		}
+	/* Initialize the ring buffer's read and write pointers */
+	WREG32_SOC15(VCN, i, mmUVD_RBC_RB_RPTR, 0);
+
+	WREG32_SOC15(VCN, i, mmUVD_SCRATCH2, 0);
+	ring->wptr = RREG32_SOC15(VCN, i, mmUVD_RBC_RB_RPTR);
+	WREG32_SOC15(VCN, i, mmUVD_RBC_RB_WPTR,
+		     lower_32_bits(ring->wptr));
+	fw_shared->rb.wptr = lower_32_bits(ring->wptr);
+	fw_shared->multi_queue.decode_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET);
+
+	if (amdgpu_ip_version(adev, UVD_HWIP, 0) !=
+	    IP_VERSION(3, 0, 33)) {
+		fw_shared->multi_queue.encode_generalpurpose_queue_mode |= cpu_to_le32(FW_QUEUE_RING_RESET);
+		ring = &adev->vcn.inst[i].ring_enc[0];
+		WREG32_SOC15(VCN, i, mmUVD_RB_RPTR, lower_32_bits(ring->wptr));
+		WREG32_SOC15(VCN, i, mmUVD_RB_WPTR, lower_32_bits(ring->wptr));
+		WREG32_SOC15(VCN, i, mmUVD_RB_BASE_LO, ring->gpu_addr);
+		WREG32_SOC15(VCN, i, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
+		WREG32_SOC15(VCN, i, mmUVD_RB_SIZE, ring->ring_size / 4);
+		fw_shared->multi_queue.encode_generalpurpose_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET);
+
+		fw_shared->multi_queue.encode_lowlatency_queue_mode |= cpu_to_le32(FW_QUEUE_RING_RESET);
+		ring = &adev->vcn.inst[i].ring_enc[1];
+		WREG32_SOC15(VCN, i, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr));
+		WREG32_SOC15(VCN, i, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr));
+		WREG32_SOC15(VCN, i, mmUVD_RB_BASE_LO2, ring->gpu_addr);
+		WREG32_SOC15(VCN, i, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr));
+		WREG32_SOC15(VCN, i, mmUVD_RB_SIZE2, ring->ring_size / 4);
+		fw_shared->multi_queue.encode_lowlatency_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET);
 	}
 
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, i, mmUVD_STATUS);
+
 	return 0;
 }
 
@@ -1313,7 +1414,7 @@ static int vcn_v3_0_start_sriov(struct amdgpu_device *adev)
 
 	header.version = MMSCH_VERSION;
 	header.total_size = sizeof(struct mmsch_v3_0_init_header) >> 2;
-	for (i = 0; i < AMDGPU_MAX_VCN_INSTANCES; i++) {
+	for (i = 0; i < MMSCH_V3_0_VCN_INSTANCES; i++) {
 		header.inst[i].init_status = 0;
 		header.inst[i].table_offset = 0;
 		header.inst[i].table_size = 0;
@@ -1331,7 +1432,7 @@ static int vcn_v3_0_start_sriov(struct amdgpu_device *adev)
 			mmUVD_STATUS),
 			~UVD_STATUS__UVD_BUSY, UVD_STATUS__UVD_BUSY);
 
-		cache_size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+		cache_size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.inst[i].fw->size + 4);
 
 		if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 			MMSCH_V3_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i,
@@ -1390,7 +1491,7 @@ static int vcn_v3_0_start_sriov(struct amdgpu_device *adev)
 			mmUVD_VCPU_CACHE_SIZE2),
 			AMDGPU_VCN_CONTEXT_SIZE);
 
-		for (j = 0; j < adev->vcn.num_enc_rings; ++j) {
+		for (j = 0; j < adev->vcn.inst[i].num_enc_rings; ++j) {
 			ring = &adev->vcn.inst[i].ring_enc[j];
 			ring->wptr = 0;
 			rb_addr = ring->gpu_addr;
@@ -1490,12 +1591,14 @@ static int vcn_v3_0_start_sriov(struct amdgpu_device *adev)
 	return 0;
 }
 
-static int vcn_v3_0_stop_dpg_mode(struct amdgpu_device *adev, int inst_idx)
+static int vcn_v3_0_stop_dpg_mode(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
 	struct dpg_pause_state state = {.fw_based = VCN_DPG_STATE__UNPAUSE};
 	uint32_t tmp;
 
-	vcn_v3_0_pause_dpg_mode(adev, inst_idx, &state);
+	vcn_v3_0_pause_dpg_mode(vinst, &state);
 
 	/* Wait for power status to be 1 */
 	SOC15_WAIT_ON_RREG(VCN, inst_idx, mmUVD_POWER_STATUS, 1,
@@ -1518,88 +1621,101 @@ static int vcn_v3_0_stop_dpg_mode(struct amdgpu_device *adev, int inst_idx)
 	WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, mmUVD_POWER_STATUS), 0,
 		~UVD_POWER_STATUS__UVD_PG_MODE_MASK);
 
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, inst_idx, mmUVD_STATUS);
+
 	return 0;
 }
 
-static int vcn_v3_0_stop(struct amdgpu_device *adev)
+static int vcn_v3_0_stop(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int i = vinst->inst;
 	uint32_t tmp;
-	int i, r = 0;
+	int r = 0;
 
-	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
-		if (adev->vcn.harvest_config & (1 << i))
-			continue;
+	if (adev->vcn.harvest_config & (1 << i))
+		return 0;
 
-		if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
-			r = vcn_v3_0_stop_dpg_mode(adev, i);
-			continue;
-		}
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
+		r = vcn_v3_0_stop_dpg_mode(vinst);
+		goto done;
+	}
 
-		/* wait for vcn idle */
-		r = SOC15_WAIT_ON_RREG(VCN, i, mmUVD_STATUS, UVD_STATUS__IDLE, 0x7);
-		if (r)
-			return r;
+	/* wait for vcn idle */
+	r = SOC15_WAIT_ON_RREG(VCN, i, mmUVD_STATUS, UVD_STATUS__IDLE, 0x7);
+	if (r)
+		goto done;
 
-		tmp = UVD_LMI_STATUS__VCPU_LMI_WRITE_CLEAN_MASK |
-			UVD_LMI_STATUS__READ_CLEAN_MASK |
-			UVD_LMI_STATUS__WRITE_CLEAN_MASK |
-			UVD_LMI_STATUS__WRITE_CLEAN_RAW_MASK;
-		r = SOC15_WAIT_ON_RREG(VCN, i, mmUVD_LMI_STATUS, tmp, tmp);
-		if (r)
-			return r;
+	tmp = UVD_LMI_STATUS__VCPU_LMI_WRITE_CLEAN_MASK |
+		UVD_LMI_STATUS__READ_CLEAN_MASK |
+		UVD_LMI_STATUS__WRITE_CLEAN_MASK |
+		UVD_LMI_STATUS__WRITE_CLEAN_RAW_MASK;
+	r = SOC15_WAIT_ON_RREG(VCN, i, mmUVD_LMI_STATUS, tmp, tmp);
+	if (r)
+		goto done;
+
+	/* disable LMI UMC channel */
+	tmp = RREG32_SOC15(VCN, i, mmUVD_LMI_CTRL2);
+	tmp |= UVD_LMI_CTRL2__STALL_ARB_UMC_MASK;
+	WREG32_SOC15(VCN, i, mmUVD_LMI_CTRL2, tmp);
+	tmp = UVD_LMI_STATUS__UMC_READ_CLEAN_RAW_MASK|
+		UVD_LMI_STATUS__UMC_WRITE_CLEAN_RAW_MASK;
+	r = SOC15_WAIT_ON_RREG(VCN, i, mmUVD_LMI_STATUS, tmp, tmp);
+	if (r)
+		goto done;
 
-		/* disable LMI UMC channel */
-		tmp = RREG32_SOC15(VCN, i, mmUVD_LMI_CTRL2);
-		tmp |= UVD_LMI_CTRL2__STALL_ARB_UMC_MASK;
-		WREG32_SOC15(VCN, i, mmUVD_LMI_CTRL2, tmp);
-		tmp = UVD_LMI_STATUS__UMC_READ_CLEAN_RAW_MASK|
-			UVD_LMI_STATUS__UMC_WRITE_CLEAN_RAW_MASK;
-		r = SOC15_WAIT_ON_RREG(VCN, i, mmUVD_LMI_STATUS, tmp, tmp);
-		if (r)
-			return r;
+	/* block VCPU register access */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_RB_ARB_CTRL),
+		 UVD_RB_ARB_CTRL__VCPU_DIS_MASK,
+		 ~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
 
-		/* block VCPU register access */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_RB_ARB_CTRL),
-			UVD_RB_ARB_CTRL__VCPU_DIS_MASK,
-			~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
+	/* reset VCPU */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL),
+		 UVD_VCPU_CNTL__BLK_RST_MASK,
+		 ~UVD_VCPU_CNTL__BLK_RST_MASK);
 
-		/* reset VCPU */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL),
-			UVD_VCPU_CNTL__BLK_RST_MASK,
-			~UVD_VCPU_CNTL__BLK_RST_MASK);
+	/* disable VCPU clock */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL), 0,
+		 ~(UVD_VCPU_CNTL__CLK_EN_MASK));
 
-		/* disable VCPU clock */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, mmUVD_VCPU_CNTL), 0,
-			~(UVD_VCPU_CNTL__CLK_EN_MASK));
+	/* apply soft reset */
+	tmp = RREG32_SOC15(VCN, i, mmUVD_SOFT_RESET);
+	tmp |= UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
+	WREG32_SOC15(VCN, i, mmUVD_SOFT_RESET, tmp);
+	tmp = RREG32_SOC15(VCN, i, mmUVD_SOFT_RESET);
+	tmp |= UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
+	WREG32_SOC15(VCN, i, mmUVD_SOFT_RESET, tmp);
 
-		/* apply soft reset */
-		tmp = RREG32_SOC15(VCN, i, mmUVD_SOFT_RESET);
-		tmp |= UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
-		WREG32_SOC15(VCN, i, mmUVD_SOFT_RESET, tmp);
-		tmp = RREG32_SOC15(VCN, i, mmUVD_SOFT_RESET);
-		tmp |= UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
-		WREG32_SOC15(VCN, i, mmUVD_SOFT_RESET, tmp);
+	/* clear status */
+	WREG32_SOC15(VCN, i, mmUVD_STATUS, 0);
 
-		/* clear status */
-		WREG32_SOC15(VCN, i, mmUVD_STATUS, 0);
+	/* apply HW clock gating */
+	vcn_v3_0_enable_clock_gating(vinst);
 
-		/* apply HW clock gating */
-		vcn_v3_0_enable_clock_gating(adev, i);
+	/* enable VCN power gating */
+	vcn_v3_0_enable_static_power_gating(vinst);
 
-		/* enable VCN power gating */
-		vcn_v3_0_enable_static_power_gating(adev, i);
-	}
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, i, mmUVD_STATUS);
 
+done:
 	if (adev->pm.dpm_enabled)
-		amdgpu_dpm_enable_uvd(adev, false);
+		amdgpu_dpm_enable_vcn(adev, false, i);
 
-	return 0;
+	return r;
 }
 
-static int vcn_v3_0_pause_dpg_mode(struct amdgpu_device *adev,
-		   int inst_idx, struct dpg_pause_state *new_state)
+static int vcn_v3_0_pause_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				   struct dpg_pause_state *new_state)
 {
-	volatile struct amdgpu_fw_shared *fw_shared;
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	struct amdgpu_fw_shared *fw_shared;
 	struct amdgpu_ring *ring;
 	uint32_t reg_data = 0;
 	int ret_code;
@@ -1630,7 +1746,8 @@ static int vcn_v3_0_pause_dpg_mode(struct amdgpu_device *adev,
 					UVD_POWER_STATUS__STALL_DPG_POWER_UP_MASK,
 					~UVD_POWER_STATUS__STALL_DPG_POWER_UP_MASK);
 
-				if (adev->ip_versions[UVD_HWIP][0] != IP_VERSION(3, 0, 33)) {
+				if (amdgpu_ip_version(adev, UVD_HWIP, 0) !=
+				    IP_VERSION(3, 0, 33)) {
 					/* Restore */
 					fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr;
 					fw_shared->multi_queue.encode_generalpurpose_queue_mode |= cpu_to_le32(FW_QUEUE_RING_RESET);
@@ -1717,7 +1834,7 @@ static uint64_t vcn_v3_0_dec_ring_get_wptr(struct amdgpu_ring *ring)
 static void vcn_v3_0_dec_ring_set_wptr(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	volatile struct amdgpu_fw_shared *fw_shared;
+	struct amdgpu_fw_shared *fw_shared;
 
 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
 		/*whenever update RBC_RB_WPTR, we save the wptr in shared rb.wptr and scratch2 */
@@ -1767,15 +1884,19 @@ static int vcn_v3_0_limit_sched(struct amdgpu_cs_parser *p,
 				struct amdgpu_job *job)
 {
 	struct drm_gpu_scheduler **scheds;
-
-	/* The create msg must be in the first IB submitted */
-	if (atomic_read(&job->base.entity->fence_seq))
-		return -EINVAL;
+	struct dma_fence *fence;
 
 	/* if VCN0 is harvested, we can't support AV1 */
 	if (p->adev->vcn.harvest_config & AMDGPU_VCN_HARVEST_VCN0)
 		return -EINVAL;
 
+	/* wait for all jobs to finish before switching to instance 0 */
+	fence = amdgpu_ctx_get_fence(p->ctx, job->base.entity, ~0ull);
+	if (fence) {
+		dma_fence_wait(fence, false);
+		dma_fence_put(fence);
+	}
+
 	scheds = p->adev->gpu_sched[AMDGPU_HW_IP_VCN_DEC]
 		[AMDGPU_RING_PRIO_DEFAULT].sched;
 	drm_sched_entity_modify_sched(job->base.entity, scheds, 1);
@@ -1791,7 +1912,7 @@ static int vcn_v3_0_dec_msg(struct amdgpu_cs_parser *p, struct amdgpu_job *job,
 	struct amdgpu_bo *bo;
 	uint64_t start, end;
 	unsigned int i;
-	void * ptr;
+	void *ptr;
 	int r;
 
 	addr &= AMDGPU_GMC_HOLE_MASK;
@@ -1881,11 +2002,11 @@ static int vcn_v3_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p,
 		uint32_t reg = amdgpu_ib_get_value(ib, i);
 		uint32_t val = amdgpu_ib_get_value(ib, i + 1);
 
-		if (reg == PACKET0(p->adev->vcn.internal.data0, 0)) {
+		if (reg == PACKET0(p->adev->vcn.inst[ring->me].internal.data0, 0)) {
 			msg_lo = val;
-		} else if (reg == PACKET0(p->adev->vcn.internal.data1, 0)) {
+		} else if (reg == PACKET0(p->adev->vcn.inst[ring->me].internal.data1, 0)) {
 			msg_hi = val;
-		} else if (reg == PACKET0(p->adev->vcn.internal.cmd, 0) &&
+		} else if (reg == PACKET0(p->adev->vcn.inst[ring->me].internal.cmd, 0) &&
 			   val == 0) {
 			r = vcn_v3_0_dec_msg(p, job,
 					     ((u64)msg_hi) << 32 | msg_lo);
@@ -1925,6 +2046,7 @@ static const struct amdgpu_ring_funcs vcn_v3_0_dec_ring_vm_funcs = {
 	.emit_wreg = vcn_v2_0_dec_ring_emit_wreg,
 	.emit_reg_wait = vcn_v2_0_dec_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = amdgpu_vcn_ring_reset,
 };
 
 /**
@@ -2023,6 +2145,7 @@ static const struct amdgpu_ring_funcs vcn_v3_0_enc_ring_vm_funcs = {
 	.emit_wreg = vcn_v2_0_enc_ring_emit_wreg,
 	.emit_reg_wait = vcn_v2_0_enc_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = amdgpu_vcn_ring_reset,
 };
 
 static void vcn_v3_0_set_dec_ring_funcs(struct amdgpu_device *adev)
@@ -2038,8 +2161,6 @@ static void vcn_v3_0_set_dec_ring_funcs(struct amdgpu_device *adev)
 		else
 			adev->vcn.inst[i].ring_dec.funcs = &vcn_v3_0_dec_sw_ring_vm_funcs;
 		adev->vcn.inst[i].ring_dec.me = i;
-		DRM_INFO("VCN(%d) decode%s is enabled in VM mode\n", i,
-			  DEC_SW_RING_ENABLED?"(Software Ring)":"");
 	}
 }
 
@@ -2051,18 +2172,28 @@ static void vcn_v3_0_set_enc_ring_funcs(struct amdgpu_device *adev)
 		if (adev->vcn.harvest_config & (1 << i))
 			continue;
 
-		for (j = 0; j < adev->vcn.num_enc_rings; ++j) {
+		for (j = 0; j < adev->vcn.inst[i].num_enc_rings; ++j) {
 			adev->vcn.inst[i].ring_enc[j].funcs = &vcn_v3_0_enc_ring_vm_funcs;
 			adev->vcn.inst[i].ring_enc[j].me = i;
 		}
-		if (adev->vcn.num_enc_rings > 0)
-			DRM_INFO("VCN(%d) encode is enabled in VM mode\n", i);
 	}
 }
 
-static bool vcn_v3_0_is_idle(void *handle)
+static int vcn_v3_0_reset(struct amdgpu_vcn_inst *vinst)
+{
+	int r;
+
+	r = vcn_v3_0_stop(vinst);
+	if (r)
+		return r;
+	vcn_v3_0_enable_clock_gating(vinst);
+	vcn_v3_0_enable_static_power_gating(vinst);
+	return vcn_v3_0_start(vinst);
+}
+
+static bool vcn_v3_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i, ret = 1;
 
 	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
@@ -2075,9 +2206,9 @@ static bool vcn_v3_0_is_idle(void *handle)
 	return ret;
 }
 
-static int vcn_v3_0_wait_for_idle(void *handle)
+static int vcn_v3_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i, ret = 0;
 
 	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
@@ -2093,54 +2224,55 @@ static int vcn_v3_0_wait_for_idle(void *handle)
 	return ret;
 }
 
-static int vcn_v3_0_set_clockgating_state(void *handle,
+static int vcn_v3_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	bool enable = (state == AMD_CG_STATE_GATE) ? true : false;
+	struct amdgpu_device *adev = ip_block->adev;
+	bool enable = state == AMD_CG_STATE_GATE;
 	int i;
 
 	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[i];
 		if (adev->vcn.harvest_config & (1 << i))
 			continue;
 
 		if (enable) {
 			if (RREG32_SOC15(VCN, i, mmUVD_STATUS) != UVD_STATUS__IDLE)
 				return -EBUSY;
-			vcn_v3_0_enable_clock_gating(adev, i);
+			vcn_v3_0_enable_clock_gating(vinst);
 		} else {
-			vcn_v3_0_disable_clock_gating(adev, i);
+			vcn_v3_0_disable_clock_gating(vinst);
 		}
 	}
 
 	return 0;
 }
 
-static int vcn_v3_0_set_powergating_state(void *handle,
-					  enum amd_powergating_state state)
+static int vcn_v3_0_set_pg_state(struct amdgpu_vcn_inst *vinst,
+				 enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	int ret;
+	struct amdgpu_device *adev = vinst->adev;
+	int ret = 0;
 
 	/* for SRIOV, guest should not control VCN Power-gating
 	 * MMSCH FW should control Power-gating and clock-gating
 	 * guest should avoid touching CGC and PG
 	 */
 	if (amdgpu_sriov_vf(adev)) {
-		adev->vcn.cur_state = AMD_PG_STATE_UNGATE;
+		vinst->cur_state = AMD_PG_STATE_UNGATE;
 		return 0;
 	}
 
-	if(state == adev->vcn.cur_state)
+	if (state == vinst->cur_state)
 		return 0;
 
 	if (state == AMD_PG_STATE_GATE)
-		ret = vcn_v3_0_stop(adev);
+		ret = vcn_v3_0_stop(vinst);
 	else
-		ret = vcn_v3_0_start(adev);
+		ret = vcn_v3_0_start(vinst);
 
-	if(!ret)
-		adev->vcn.cur_state = state;
+	if (!ret)
+		vinst->cur_state = state;
 
 	return ret;
 }
@@ -2205,7 +2337,7 @@ static void vcn_v3_0_set_irq_funcs(struct amdgpu_device *adev)
 		if (adev->vcn.harvest_config & (1 << i))
 			continue;
 
-		adev->vcn.inst[i].irq.num_types = adev->vcn.num_enc_rings + 1;
+		adev->vcn.inst[i].irq.num_types = adev->vcn.inst[i].num_enc_rings + 1;
 		adev->vcn.inst[i].irq.funcs = &vcn_v3_0_irq_funcs;
 	}
 }
@@ -2213,7 +2345,6 @@ static void vcn_v3_0_set_irq_funcs(struct amdgpu_device *adev)
 static const struct amd_ip_funcs vcn_v3_0_ip_funcs = {
 	.name = "vcn_v3_0",
 	.early_init = vcn_v3_0_early_init,
-	.late_init = NULL,
 	.sw_init = vcn_v3_0_sw_init,
 	.sw_fini = vcn_v3_0_sw_fini,
 	.hw_init = vcn_v3_0_hw_init,
@@ -2222,16 +2353,13 @@ static const struct amd_ip_funcs vcn_v3_0_ip_funcs = {
 	.resume = vcn_v3_0_resume,
 	.is_idle = vcn_v3_0_is_idle,
 	.wait_for_idle = vcn_v3_0_wait_for_idle,
-	.check_soft_reset = NULL,
-	.pre_soft_reset = NULL,
-	.soft_reset = NULL,
-	.post_soft_reset = NULL,
 	.set_clockgating_state = vcn_v3_0_set_clockgating_state,
-	.set_powergating_state = vcn_v3_0_set_powergating_state,
+	.set_powergating_state = vcn_set_powergating_state,
+	.dump_ip_state = amdgpu_vcn_dump_ip_state,
+	.print_ip_state = amdgpu_vcn_print_ip_state,
 };
 
-const struct amdgpu_ip_block_version vcn_v3_0_ip_block =
-{
+const struct amdgpu_ip_block_version vcn_v3_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_VCN,
 	.major = 3,
 	.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c
index bf0674039598..3ae666522d57 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c
@@ -46,12 +46,49 @@
 
 #define VCN_VID_SOC_ADDRESS_2_0							0x1fb00
 #define VCN1_VID_SOC_ADDRESS_3_0						0x48300
+#define VCN1_AON_SOC_ADDRESS_3_0						0x48000
 
 #define VCN_HARVEST_MMSCH								0
 
 #define RDECODE_MSG_CREATE							0x00000000
 #define RDECODE_MESSAGE_CREATE							0x00000001
 
+static const struct amdgpu_hwip_reg_entry vcn_reg_list_4_0[] = {
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_POWER_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_CONTEXT_ID),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_CONTEXT_ID2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_GPCOM_VCPU_DATA0),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_GPCOM_VCPU_DATA1),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_GPCOM_VCPU_CMD),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_PGFSM_CONFIG),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_PGFSM_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_LMA_CTL),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_LMA_DATA),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_LMA_MASK),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_PAUSE)
+};
+
 static int amdgpu_ih_clientid_vcns[] = {
 	SOC15_IH_CLIENTID_VCN,
 	SOC15_IH_CLIENTID_VCN1
@@ -60,25 +97,25 @@ static int amdgpu_ih_clientid_vcns[] = {
 static int vcn_v4_0_start_sriov(struct amdgpu_device *adev);
 static void vcn_v4_0_set_unified_ring_funcs(struct amdgpu_device *adev);
 static void vcn_v4_0_set_irq_funcs(struct amdgpu_device *adev);
-static int vcn_v4_0_set_powergating_state(void *handle,
-        enum amd_powergating_state state);
-static int vcn_v4_0_pause_dpg_mode(struct amdgpu_device *adev,
-        int inst_idx, struct dpg_pause_state *new_state);
+static int vcn_v4_0_set_pg_state(struct amdgpu_vcn_inst *vinst,
+				 enum amd_powergating_state state);
+static int vcn_v4_0_pause_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				   struct dpg_pause_state *new_state);
 static void vcn_v4_0_unified_ring_set_wptr(struct amdgpu_ring *ring);
 static void vcn_v4_0_set_ras_funcs(struct amdgpu_device *adev);
 
 /**
  * vcn_v4_0_early_init - set function pointers and load microcode
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Set ring and irq function pointers
  * Load microcode from filesystem
  */
-static int vcn_v4_0_early_init(void *handle)
+static int vcn_v4_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	int i;
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, r;
 
 	if (amdgpu_sriov_vf(adev)) {
 		adev->vcn.harvest_config = VCN_HARVEST_MMSCH;
@@ -90,46 +127,82 @@ static int vcn_v4_0_early_init(void *handle)
 		}
 	}
 
-	/* re-use enc ring as unified ring */
-	adev->vcn.num_enc_rings = 1;
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i)
+		/* re-use enc ring as unified ring */
+		adev->vcn.inst[i].num_enc_rings = 1;
 
 	vcn_v4_0_set_unified_ring_funcs(adev);
 	vcn_v4_0_set_irq_funcs(adev);
 	vcn_v4_0_set_ras_funcs(adev);
 
-	return amdgpu_vcn_early_init(adev);
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		adev->vcn.inst[i].set_pg_state = vcn_v4_0_set_pg_state;
+
+		r = amdgpu_vcn_early_init(adev, i);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+static int vcn_v4_0_fw_shared_init(struct amdgpu_device *adev, int inst_idx)
+{
+	struct amdgpu_vcn4_fw_shared *fw_shared;
+
+	fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr;
+	fw_shared->present_flag_0 = cpu_to_le32(AMDGPU_FW_SHARED_FLAG_0_UNIFIED_QUEUE);
+	fw_shared->sq.is_enabled = 1;
+
+	fw_shared->present_flag_0 |= cpu_to_le32(AMDGPU_VCN_SMU_DPM_INTERFACE_FLAG);
+	fw_shared->smu_dpm_interface.smu_interface_type = (adev->flags & AMD_IS_APU) ?
+		AMDGPU_VCN_SMU_DPM_INTERFACE_APU : AMDGPU_VCN_SMU_DPM_INTERFACE_DGPU;
+
+	if (amdgpu_ip_version(adev, VCN_HWIP, 0) ==
+	    IP_VERSION(4, 0, 2)) {
+		fw_shared->present_flag_0 |= AMDGPU_FW_SHARED_FLAG_0_DRM_KEY_INJECT;
+		fw_shared->drm_key_wa.method =
+			AMDGPU_DRM_KEY_INJECT_WORKAROUND_VCNFW_ASD_HANDSHAKING;
+	}
+
+	if (amdgpu_vcnfw_log)
+		amdgpu_vcn_fwlog_init(&adev->vcn.inst[inst_idx]);
+
+	return 0;
 }
 
 /**
  * vcn_v4_0_sw_init - sw init for VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Load firmware and sw initialization
  */
-static int vcn_v4_0_sw_init(void *handle)
+static int vcn_v4_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i, r;
 
-	r = amdgpu_vcn_sw_init(adev);
-	if (r)
-		return r;
-
-	amdgpu_vcn_setup_ucode(adev);
-
-	r = amdgpu_vcn_resume(adev);
-	if (r)
-		return r;
-
 	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
-		volatile struct amdgpu_vcn4_fw_shared *fw_shared;
-
 		if (adev->vcn.harvest_config & (1 << i))
 			continue;
 
-		atomic_set(&adev->vcn.inst[i].sched_score, 0);
+		r = amdgpu_vcn_sw_init(adev, i);
+		if (r)
+			return r;
+
+		amdgpu_vcn_setup_ucode(adev, i);
+
+		r = amdgpu_vcn_resume(adev, i);
+		if (r)
+			return r;
+
+		/* Init instance 0 sched_score to 1, so it's scheduled after other instances */
+		if (i == 0)
+			atomic_set(&adev->vcn.inst[i].sched_score, 1);
+		else
+			atomic_set(&adev->vcn.inst[i].sched_score, 0);
 
 		/* VCN UNIFIED TRAP */
 		r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_vcns[i],
@@ -139,17 +212,18 @@ static int vcn_v4_0_sw_init(void *handle)
 
 		/* VCN POISON TRAP */
 		r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_vcns[i],
-				VCN_4_0__SRCID_UVD_POISON, &adev->vcn.inst[i].irq);
+				VCN_4_0__SRCID_UVD_POISON, &adev->vcn.inst[i].ras_poison_irq);
 		if (r)
 			return r;
 
 		ring = &adev->vcn.inst[i].ring_enc[0];
 		ring->use_doorbell = true;
 		if (amdgpu_sriov_vf(adev))
-			ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + i * (adev->vcn.num_enc_rings + 1) + 1;
+			ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + i *
+				(adev->vcn.inst[i].num_enc_rings + 1) + 1;
 		else
 			ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 2 + 8 * i;
-		ring->vm_hub = AMDGPU_MMHUB_0;
+		ring->vm_hub = AMDGPU_MMHUB0(0);
 		sprintf(ring->name, "vcn_unified_%d", i);
 
 		r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[i].irq, 0,
@@ -157,52 +231,54 @@ static int vcn_v4_0_sw_init(void *handle)
 		if (r)
 			return r;
 
-		fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
-		fw_shared->present_flag_0 = cpu_to_le32(AMDGPU_FW_SHARED_FLAG_0_UNIFIED_QUEUE);
-		fw_shared->sq.is_enabled = 1;
-
-		fw_shared->present_flag_0 |= cpu_to_le32(AMDGPU_VCN_SMU_DPM_INTERFACE_FLAG);
-		fw_shared->smu_dpm_interface.smu_interface_type = (adev->flags & AMD_IS_APU) ?
-			AMDGPU_VCN_SMU_DPM_INTERFACE_APU : AMDGPU_VCN_SMU_DPM_INTERFACE_DGPU;
+		vcn_v4_0_fw_shared_init(adev, i);
 
-		if (amdgpu_sriov_vf(adev))
-			fw_shared->present_flag_0 |= cpu_to_le32(AMDGPU_VCN_VF_RB_SETUP_FLAG);
-
-		if (amdgpu_vcnfw_log)
-			amdgpu_vcn_fwlog_init(&adev->vcn.inst[i]);
+		if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
+			adev->vcn.inst[i].pause_dpg_mode = vcn_v4_0_pause_dpg_mode;
 	}
 
+	adev->vcn.supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->vcn.inst[0].ring_enc[0]);
+	if (!amdgpu_sriov_vf(adev))
+		adev->vcn.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+
 	if (amdgpu_sriov_vf(adev)) {
 		r = amdgpu_virt_alloc_mm_table(adev);
 		if (r)
 			return r;
 	}
 
-	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
-		adev->vcn.pause_dpg_mode = vcn_v4_0_pause_dpg_mode;
 
 	r = amdgpu_vcn_ras_sw_init(adev);
 	if (r)
 		return r;
 
+	r = amdgpu_vcn_reg_dump_init(adev, vcn_reg_list_4_0, ARRAY_SIZE(vcn_reg_list_4_0));
+	if (r)
+		return r;
+
+	r = amdgpu_vcn_sysfs_reset_mask_init(adev);
+	if (r)
+		return r;
+
 	return 0;
 }
 
 /**
  * vcn_v4_0_sw_fini - sw fini for VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * VCN suspend and free up sw allocation
  */
-static int vcn_v4_0_sw_fini(void *handle)
+static int vcn_v4_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i, r, idx;
 
 	if (drm_dev_enter(adev_to_drm(adev), &idx)) {
 		for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
-			volatile struct amdgpu_vcn4_fw_shared *fw_shared;
+			struct amdgpu_vcn4_fw_shared *fw_shared;
 
 			if (adev->vcn.harvest_config & (1 << i))
 				continue;
@@ -218,32 +294,37 @@ static int vcn_v4_0_sw_fini(void *handle)
 	if (amdgpu_sriov_vf(adev))
 		amdgpu_virt_free_mm_table(adev);
 
-	r = amdgpu_vcn_suspend(adev);
-	if (r)
-		return r;
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_suspend(adev, i);
+		if (r)
+			return r;
+	}
 
-	r = amdgpu_vcn_sw_fini(adev);
+	amdgpu_vcn_sysfs_reset_mask_fini(adev);
 
-	return r;
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++)
+		amdgpu_vcn_sw_fini(adev, i);
+
+	return 0;
 }
 
 /**
  * vcn_v4_0_hw_init - start and test VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Initialize the hardware, boot up the VCPU and do some testing
  */
-static int vcn_v4_0_hw_init(void *handle)
+static int vcn_v4_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring;
 	int i, r;
 
 	if (amdgpu_sriov_vf(adev)) {
 		r = vcn_v4_0_start_sriov(adev);
 		if (r)
-			goto done;
+			return r;
 
 		for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
 			if (adev->vcn.harvest_config & (1 << i))
@@ -254,7 +335,6 @@ static int vcn_v4_0_hw_init(void *handle)
 			ring->wptr_old = 0;
 			vcn_v4_0_unified_ring_set_wptr(ring);
 			ring->sched.ready = true;
-
 		}
 	} else {
 		for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
@@ -268,45 +348,42 @@ static int vcn_v4_0_hw_init(void *handle)
 
 			r = amdgpu_ring_test_helper(ring);
 			if (r)
-				goto done;
-
+				return r;
 		}
 	}
 
-done:
-	if (!r)
-		DRM_INFO("VCN decode and encode initialized successfully(under %s).\n",
-			(adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)?"DPG Mode":"SPG Mode");
-
-	return r;
+	return 0;
 }
 
 /**
  * vcn_v4_0_hw_fini - stop the hardware block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Stop the VCN block, mark ring as not ready any more
  */
-static int vcn_v4_0_hw_fini(void *handle)
+static int vcn_v4_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
-	cancel_delayed_work_sync(&adev->vcn.idle_work);
-
 	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[i];
+
 		if (adev->vcn.harvest_config & (1 << i))
 			continue;
+
+		cancel_delayed_work_sync(&vinst->idle_work);
+
 		if (!amdgpu_sriov_vf(adev)) {
 			if ((adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) ||
-                        (adev->vcn.cur_state != AMD_PG_STATE_GATE &&
-                                RREG32_SOC15(VCN, i, regUVD_STATUS))) {
-                        vcn_v4_0_set_powergating_state(adev, AMD_PG_STATE_GATE);
+			    (vinst->cur_state != AMD_PG_STATE_GATE &&
+				 RREG32_SOC15(VCN, i, regUVD_STATUS))) {
+				vinst->set_pg_state(vinst, AMD_PG_STATE_GATE);
 			}
 		}
-
-		amdgpu_irq_put(adev, &adev->vcn.inst[i].irq, 0);
+		if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__VCN))
+			amdgpu_irq_put(adev, &vinst->ras_poison_irq, 0);
 	}
 
 	return 0;
@@ -315,41 +392,47 @@ static int vcn_v4_0_hw_fini(void *handle)
 /**
  * vcn_v4_0_suspend - suspend VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * HW fini and suspend VCN block
  */
-static int vcn_v4_0_suspend(void *handle)
+static int vcn_v4_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int r, i;
 
-	r = vcn_v4_0_hw_fini(adev);
+	r = vcn_v4_0_hw_fini(ip_block);
 	if (r)
 		return r;
 
-	r = amdgpu_vcn_suspend(adev);
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_suspend(ip_block->adev, i);
+		if (r)
+			return r;
+	}
 
-	return r;
+	return 0;
 }
 
 /**
  * vcn_v4_0_resume - resume VCN block
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Resume firmware and hw init VCN block
  */
-static int vcn_v4_0_resume(void *handle)
+static int vcn_v4_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int r, i;
 
-	r = amdgpu_vcn_resume(adev);
-	if (r)
-		return r;
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_resume(ip_block->adev, i);
+		if (r)
+			return r;
+	}
 
-	r = vcn_v4_0_hw_init(adev);
+	r = vcn_v4_0_hw_init(ip_block);
 
 	return r;
 }
@@ -357,17 +440,18 @@ static int vcn_v4_0_resume(void *handle)
 /**
  * vcn_v4_0_mc_resume - memory controller programming
  *
- * @adev: amdgpu_device pointer
- * @inst: instance number
+ * @vinst: VCN instance
  *
  * Let the VCN memory controller know it's offsets
  */
-static void vcn_v4_0_mc_resume(struct amdgpu_device *adev, int inst)
+static void vcn_v4_0_mc_resume(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
 	uint32_t offset, size;
 	const struct common_firmware_header *hdr;
 
-	hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
+	hdr = (const struct common_firmware_header *)adev->vcn.inst[inst].fw->data;
 	size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
 
 	/* cache window 0: fw */
@@ -417,17 +501,19 @@ static void vcn_v4_0_mc_resume(struct amdgpu_device *adev, int inst)
 /**
  * vcn_v4_0_mc_resume_dpg_mode - memory controller programming for dpg mode
  *
- * @adev: amdgpu_device pointer
- * @inst_idx: instance number index
+ * @vinst: VCN instance
  * @indirect: indirectly write sram
  *
  * Let the VCN memory controller know it's offsets with dpg mode
  */
-static void vcn_v4_0_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_idx, bool indirect)
+static void vcn_v4_0_mc_resume_dpg_mode(struct amdgpu_vcn_inst *vinst,
+					bool indirect)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
 	uint32_t offset, size;
 	const struct common_firmware_header *hdr;
-	hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
+	hdr = (const struct common_firmware_header *)adev->vcn.inst[inst_idx].fw->data;
 	size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
 
 	/* cache window 0: fw */
@@ -518,19 +604,21 @@ static void vcn_v4_0_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_idx
 
 	/* VCN global tiling registers */
 	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
-		VCN, 0, regUVD_GFX10_ADDR_CONFIG), adev->gfx.config.gb_addr_config, 0, indirect);
+			VCN, inst_idx, regUVD_GFX10_ADDR_CONFIG),
+			adev->gfx.config.gb_addr_config, 0, indirect);
 }
 
 /**
  * vcn_v4_0_disable_static_power_gating - disable VCN static power gating
  *
- * @adev: amdgpu_device pointer
- * @inst: instance number
+ * @vinst: VCN instance
  *
  * Disable static power gating for VCN block
  */
-static void vcn_v4_0_disable_static_power_gating(struct amdgpu_device *adev, int inst)
+static void vcn_v4_0_disable_static_power_gating(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
 	uint32_t data = 0;
 
 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN) {
@@ -589,13 +677,14 @@ static void vcn_v4_0_disable_static_power_gating(struct amdgpu_device *adev, int
 /**
  * vcn_v4_0_enable_static_power_gating - enable VCN static power gating
  *
- * @adev: amdgpu_device pointer
- * @inst: instance number
+ * @vinst: VCN instance
  *
  * Enable static power gating for VCN block
  */
-static void vcn_v4_0_enable_static_power_gating(struct amdgpu_device *adev, int inst)
+static void vcn_v4_0_enable_static_power_gating(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
 	uint32_t data;
 
 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN) {
@@ -644,13 +733,14 @@ static void vcn_v4_0_enable_static_power_gating(struct amdgpu_device *adev, int
 /**
  * vcn_v4_0_disable_clock_gating - disable VCN clock gating
  *
- * @adev: amdgpu_device pointer
- * @inst: instance number
+ * @vinst: VCN instance
  *
  * Disable clock gating for VCN block
  */
-static void vcn_v4_0_disable_clock_gating(struct amdgpu_device *adev, int inst)
+static void vcn_v4_0_disable_clock_gating(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
 	uint32_t data;
 
 	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
@@ -755,16 +845,18 @@ static void vcn_v4_0_disable_clock_gating(struct amdgpu_device *adev, int inst)
 /**
  * vcn_v4_0_disable_clock_gating_dpg_mode - disable VCN clock gating dpg mode
  *
- * @adev: amdgpu_device pointer
+ * @vinst: VCN instance
  * @sram_sel: sram select
- * @inst_idx: instance number index
  * @indirect: indirectly write sram
  *
  * Disable clock gating for VCN block with dpg mode
  */
-static void vcn_v4_0_disable_clock_gating_dpg_mode(struct amdgpu_device *adev, uint8_t sram_sel,
-      int inst_idx, uint8_t indirect)
+static void vcn_v4_0_disable_clock_gating_dpg_mode(struct amdgpu_vcn_inst *vinst,
+						   uint8_t sram_sel,
+						   uint8_t indirect)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
 	uint32_t reg_data = 0;
 
 	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
@@ -812,13 +904,14 @@ static void vcn_v4_0_disable_clock_gating_dpg_mode(struct amdgpu_device *adev, u
 /**
  * vcn_v4_0_enable_clock_gating - enable VCN clock gating
  *
- * @adev: amdgpu_device pointer
- * @inst: instance number
+ * @vinst: VCN instance
  *
  * Enable clock gating for VCN block
  */
-static void vcn_v4_0_enable_clock_gating(struct amdgpu_device *adev, int inst)
+static void vcn_v4_0_enable_clock_gating(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
 	uint32_t data;
 
 	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
@@ -866,13 +959,13 @@ static void vcn_v4_0_enable_clock_gating(struct amdgpu_device *adev, int inst)
 		| UVD_SUVD_CGC_CTRL__IME_MODE_MASK
 		| UVD_SUVD_CGC_CTRL__SITE_MODE_MASK);
 	WREG32_SOC15(VCN, inst, regUVD_SUVD_CGC_CTRL, data);
-
-	return;
 }
 
-static void vcn_v4_0_enable_ras(struct amdgpu_device *adev, int inst_idx,
+static void vcn_v4_0_enable_ras(struct amdgpu_vcn_inst *vinst,
 				bool indirect)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
 	uint32_t tmp;
 
 	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__VCN))
@@ -895,17 +988,19 @@ static void vcn_v4_0_enable_ras(struct amdgpu_device *adev, int inst_idx,
 /**
  * vcn_v4_0_start_dpg_mode - VCN start with dpg mode
  *
- * @adev: amdgpu_device pointer
- * @inst_idx: instance number index
+ * @vinst: VCN instance
  * @indirect: indirectly write sram
  *
  * Start VCN block with dpg mode
  */
-static int vcn_v4_0_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, bool indirect)
+static int vcn_v4_0_start_dpg_mode(struct amdgpu_vcn_inst *vinst, bool indirect)
 {
-	volatile struct amdgpu_vcn4_fw_shared *fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr;
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	struct amdgpu_vcn4_fw_shared *fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr;
 	struct amdgpu_ring *ring;
 	uint32_t tmp;
+	int ret;
 
 	/* disable register anti-hang mechanism */
 	WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, regUVD_POWER_STATUS), 1,
@@ -920,7 +1015,7 @@ static int vcn_v4_0_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, boo
 		adev->vcn.inst[inst_idx].dpg_sram_curr_addr = (uint32_t *)adev->vcn.inst[inst_idx].dpg_sram_cpu_addr;
 
 	/* enable clock gating */
-	vcn_v4_0_disable_clock_gating_dpg_mode(adev, 0, inst_idx, indirect);
+	vcn_v4_0_disable_clock_gating_dpg_mode(vinst, 0, indirect);
 
 	/* enable VCPU clock */
 	tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
@@ -968,7 +1063,7 @@ static int vcn_v4_0_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, boo
 		 (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
 		 (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)), 0, indirect);
 
-	vcn_v4_0_mc_resume_dpg_mode(adev, inst_idx, indirect);
+	vcn_v4_0_mc_resume_dpg_mode(vinst, indirect);
 
 	tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
 	tmp |= UVD_VCPU_CNTL__CLK_EN_MASK;
@@ -980,7 +1075,7 @@ static int vcn_v4_0_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, boo
 	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
 		VCN, inst_idx, regUVD_LMI_CTRL2), tmp, 0, indirect);
 
-	vcn_v4_0_enable_ras(adev, inst_idx, indirect);
+	vcn_v4_0_enable_ras(vinst, indirect);
 
 	/* enable master interrupt */
 	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
@@ -988,10 +1083,13 @@ static int vcn_v4_0_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, boo
 		UVD_MASTINT_EN__VCPU_EN_MASK, 0, indirect);
 
 
-	if (indirect)
-		psp_update_vcn_sram(adev, inst_idx, adev->vcn.inst[inst_idx].dpg_sram_gpu_addr,
-			(uint32_t)((uintptr_t)adev->vcn.inst[inst_idx].dpg_sram_curr_addr -
-				(uintptr_t)adev->vcn.inst[inst_idx].dpg_sram_cpu_addr));
+	if (indirect) {
+		ret = amdgpu_vcn_psp_update_sram(adev, inst_idx, 0);
+		if (ret) {
+			dev_err(adev->dev, "vcn sram load failed %d\n", ret);
+			return ret;
+		}
+	}
 
 	ring = &adev->vcn.inst[inst_idx].ring_enc[0];
 
@@ -1019,6 +1117,11 @@ static int vcn_v4_0_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, boo
 			ring->doorbell_index << VCN_RB1_DB_CTRL__OFFSET__SHIFT |
 			VCN_RB1_DB_CTRL__EN_MASK);
 
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, inst_idx, regUVD_STATUS);
+
 	return 0;
 }
 
@@ -1026,178 +1129,202 @@ static int vcn_v4_0_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, boo
 /**
  * vcn_v4_0_start - VCN start
  *
- * @adev: amdgpu_device pointer
+ * @vinst: VCN instance
  *
  * Start VCN block
  */
-static int vcn_v4_0_start(struct amdgpu_device *adev)
+static int vcn_v4_0_start(struct amdgpu_vcn_inst *vinst)
 {
-	volatile struct amdgpu_vcn4_fw_shared *fw_shared;
+	struct amdgpu_device *adev = vinst->adev;
+	int i = vinst->inst;
+	struct amdgpu_vcn4_fw_shared *fw_shared;
 	struct amdgpu_ring *ring;
 	uint32_t tmp;
-	int i, j, k, r;
+	int j, k, r;
+
+	if (adev->vcn.harvest_config & (1 << i))
+		return 0;
 
 	if (adev->pm.dpm_enabled)
-		amdgpu_dpm_enable_uvd(adev, true);
+		amdgpu_dpm_enable_vcn(adev, true, i);
 
-	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
-		fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
+	fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
 
-		if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
-			r = vcn_v4_0_start_dpg_mode(adev, i, adev->vcn.indirect_sram);
-			continue;
-		}
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
+		return vcn_v4_0_start_dpg_mode(vinst, adev->vcn.inst[i].indirect_sram);
 
-		/* disable VCN power gating */
-		vcn_v4_0_disable_static_power_gating(adev, i);
-
-		/* set VCN status busy */
-		tmp = RREG32_SOC15(VCN, i, regUVD_STATUS) | UVD_STATUS__UVD_BUSY;
-		WREG32_SOC15(VCN, i, regUVD_STATUS, tmp);
-
-		/*SW clock gating */
-		vcn_v4_0_disable_clock_gating(adev, i);
-
-		/* enable VCPU clock */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL),
-				UVD_VCPU_CNTL__CLK_EN_MASK, ~UVD_VCPU_CNTL__CLK_EN_MASK);
-
-		/* disable master interrupt */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_MASTINT_EN), 0,
-				~UVD_MASTINT_EN__VCPU_EN_MASK);
-
-		/* enable LMI MC and UMC channels */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_LMI_CTRL2), 0,
-				~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
-
-		tmp = RREG32_SOC15(VCN, i, regUVD_SOFT_RESET);
-		tmp &= ~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
-		tmp &= ~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
-		WREG32_SOC15(VCN, i, regUVD_SOFT_RESET, tmp);
-
-		/* setup regUVD_LMI_CTRL */
-		tmp = RREG32_SOC15(VCN, i, regUVD_LMI_CTRL);
-		WREG32_SOC15(VCN, i, regUVD_LMI_CTRL, tmp |
-				UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
-				UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
-				UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
-				UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK);
-
-		/* setup regUVD_MPC_CNTL */
-		tmp = RREG32_SOC15(VCN, i, regUVD_MPC_CNTL);
-		tmp &= ~UVD_MPC_CNTL__REPLACEMENT_MODE_MASK;
-		tmp |= 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT;
-		WREG32_SOC15(VCN, i, regUVD_MPC_CNTL, tmp);
-
-		/* setup UVD_MPC_SET_MUXA0 */
-		WREG32_SOC15(VCN, i, regUVD_MPC_SET_MUXA0,
-				((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) |
-				 (0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) |
-				 (0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) |
-				 (0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)));
-
-		/* setup UVD_MPC_SET_MUXB0 */
-		WREG32_SOC15(VCN, i, regUVD_MPC_SET_MUXB0,
-				((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) |
-				 (0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) |
-				 (0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) |
-				 (0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)));
-
-		/* setup UVD_MPC_SET_MUX */
-		WREG32_SOC15(VCN, i, regUVD_MPC_SET_MUX,
-				((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) |
-				 (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
-				 (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)));
-
-		vcn_v4_0_mc_resume(adev, i);
-
-		/* VCN global tiling registers */
-		WREG32_SOC15(VCN, i, regUVD_GFX10_ADDR_CONFIG,
-				adev->gfx.config.gb_addr_config);
-
-		/* unblock VCPU register access */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_RB_ARB_CTRL), 0,
-				~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
-
-		/* release VCPU reset to boot */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), 0,
-				~UVD_VCPU_CNTL__BLK_RST_MASK);
-
-		for (j = 0; j < 10; ++j) {
-			uint32_t status;
-
-			for (k = 0; k < 100; ++k) {
-				status = RREG32_SOC15(VCN, i, regUVD_STATUS);
-				if (status & 2)
-					break;
-				mdelay(10);
-				if (amdgpu_emu_mode==1)
-					msleep(1);
-			}
+	/* disable VCN power gating */
+	vcn_v4_0_disable_static_power_gating(vinst);
+
+	/* set VCN status busy */
+	tmp = RREG32_SOC15(VCN, i, regUVD_STATUS) | UVD_STATUS__UVD_BUSY;
+	WREG32_SOC15(VCN, i, regUVD_STATUS, tmp);
+
+	/*SW clock gating */
+	vcn_v4_0_disable_clock_gating(vinst);
+
+	/* enable VCPU clock */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL),
+		 UVD_VCPU_CNTL__CLK_EN_MASK, ~UVD_VCPU_CNTL__CLK_EN_MASK);
 
-			if (amdgpu_emu_mode==1) {
-				r = -1;
-				if (status & 2) {
-					r = 0;
-					break;
-				}
-			} else {
+	/* disable master interrupt */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_MASTINT_EN), 0,
+		 ~UVD_MASTINT_EN__VCPU_EN_MASK);
+
+	/* enable LMI MC and UMC channels */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_LMI_CTRL2), 0,
+		 ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
+
+	tmp = RREG32_SOC15(VCN, i, regUVD_SOFT_RESET);
+	tmp &= ~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
+	tmp &= ~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
+	WREG32_SOC15(VCN, i, regUVD_SOFT_RESET, tmp);
+
+	/* setup regUVD_LMI_CTRL */
+	tmp = RREG32_SOC15(VCN, i, regUVD_LMI_CTRL);
+	WREG32_SOC15(VCN, i, regUVD_LMI_CTRL, tmp |
+		     UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
+		     UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
+		     UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
+		     UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK);
+
+	/* setup regUVD_MPC_CNTL */
+	tmp = RREG32_SOC15(VCN, i, regUVD_MPC_CNTL);
+	tmp &= ~UVD_MPC_CNTL__REPLACEMENT_MODE_MASK;
+	tmp |= 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT;
+	WREG32_SOC15(VCN, i, regUVD_MPC_CNTL, tmp);
+
+	/* setup UVD_MPC_SET_MUXA0 */
+	WREG32_SOC15(VCN, i, regUVD_MPC_SET_MUXA0,
+		     ((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) |
+		      (0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) |
+		      (0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) |
+		      (0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)));
+
+	/* setup UVD_MPC_SET_MUXB0 */
+	WREG32_SOC15(VCN, i, regUVD_MPC_SET_MUXB0,
+		     ((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) |
+		      (0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) |
+		      (0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) |
+		      (0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)));
+
+	/* setup UVD_MPC_SET_MUX */
+	WREG32_SOC15(VCN, i, regUVD_MPC_SET_MUX,
+		     ((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) |
+		      (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
+		      (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)));
+
+	vcn_v4_0_mc_resume(vinst);
+
+	/* VCN global tiling registers */
+	WREG32_SOC15(VCN, i, regUVD_GFX10_ADDR_CONFIG,
+		     adev->gfx.config.gb_addr_config);
+
+	/* unblock VCPU register access */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_RB_ARB_CTRL), 0,
+		 ~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
+
+	/* release VCPU reset to boot */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), 0,
+		 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+
+	for (j = 0; j < 10; ++j) {
+		uint32_t status;
+
+		for (k = 0; k < 100; ++k) {
+			status = RREG32_SOC15(VCN, i, regUVD_STATUS);
+			if (status & 2)
+				break;
+			mdelay(10);
+			if (amdgpu_emu_mode == 1)
+				msleep(1);
+		}
+
+		if (amdgpu_emu_mode == 1) {
+			r = -1;
+			if (status & 2) {
 				r = 0;
-				if (status & 2)
-					break;
-
-				dev_err(adev->dev, "VCN[%d] is not responding, trying to reset the VCPU!!!\n", i);
-				WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL),
-							UVD_VCPU_CNTL__BLK_RST_MASK,
-							~UVD_VCPU_CNTL__BLK_RST_MASK);
-				mdelay(10);
-				WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), 0,
-						~UVD_VCPU_CNTL__BLK_RST_MASK);
-
-				mdelay(10);
-				r = -1;
+				break;
 			}
+		} else {
+			r = 0;
+			if (status & 2)
+				break;
+
+			dev_err(adev->dev, "VCN[%d] is not responding, trying to reset the VCPU!!!\n", i);
+			WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL),
+				 UVD_VCPU_CNTL__BLK_RST_MASK,
+				 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+			mdelay(10);
+			WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), 0,
+				 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+
+			mdelay(10);
+			r = -1;
 		}
+	}
 
-		if (r) {
-			dev_err(adev->dev, "VCN[%d] is not responding, giving up!!!\n", i);
-			return r;
-		}
+	if (r) {
+		dev_err(adev->dev, "VCN[%d] is not responding, giving up!!!\n", i);
+		return r;
+	}
 
-		/* enable master interrupt */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_MASTINT_EN),
-				UVD_MASTINT_EN__VCPU_EN_MASK,
-				~UVD_MASTINT_EN__VCPU_EN_MASK);
+	/* enable master interrupt */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_MASTINT_EN),
+		 UVD_MASTINT_EN__VCPU_EN_MASK,
+		 ~UVD_MASTINT_EN__VCPU_EN_MASK);
 
-		/* clear the busy bit of VCN_STATUS */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_STATUS), 0,
-				~(2 << UVD_STATUS__VCPU_REPORT__SHIFT));
+	/* clear the busy bit of VCN_STATUS */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_STATUS), 0,
+		 ~(2 << UVD_STATUS__VCPU_REPORT__SHIFT));
 
-		ring = &adev->vcn.inst[i].ring_enc[0];
-		WREG32_SOC15(VCN, i, regVCN_RB1_DB_CTRL,
-				ring->doorbell_index << VCN_RB1_DB_CTRL__OFFSET__SHIFT |
-				VCN_RB1_DB_CTRL__EN_MASK);
-
-		WREG32_SOC15(VCN, i, regUVD_RB_BASE_LO, ring->gpu_addr);
-		WREG32_SOC15(VCN, i, regUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
-		WREG32_SOC15(VCN, i, regUVD_RB_SIZE, ring->ring_size / 4);
-
-		tmp = RREG32_SOC15(VCN, i, regVCN_RB_ENABLE);
-		tmp &= ~(VCN_RB_ENABLE__RB1_EN_MASK);
-		WREG32_SOC15(VCN, i, regVCN_RB_ENABLE, tmp);
-		fw_shared->sq.queue_mode |= FW_QUEUE_RING_RESET;
-		WREG32_SOC15(VCN, i, regUVD_RB_RPTR, 0);
-		WREG32_SOC15(VCN, i, regUVD_RB_WPTR, 0);
-
-		tmp = RREG32_SOC15(VCN, i, regUVD_RB_RPTR);
-		WREG32_SOC15(VCN, i, regUVD_RB_WPTR, tmp);
-		ring->wptr = RREG32_SOC15(VCN, i, regUVD_RB_WPTR);
-
-		tmp = RREG32_SOC15(VCN, i, regVCN_RB_ENABLE);
-		tmp |= VCN_RB_ENABLE__RB1_EN_MASK;
-		WREG32_SOC15(VCN, i, regVCN_RB_ENABLE, tmp);
-		fw_shared->sq.queue_mode &= ~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF);
-	}
+	ring = &adev->vcn.inst[i].ring_enc[0];
+	WREG32_SOC15(VCN, i, regVCN_RB1_DB_CTRL,
+		     ring->doorbell_index << VCN_RB1_DB_CTRL__OFFSET__SHIFT |
+		     VCN_RB1_DB_CTRL__EN_MASK);
+
+	WREG32_SOC15(VCN, i, regUVD_RB_BASE_LO, ring->gpu_addr);
+	WREG32_SOC15(VCN, i, regUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
+	WREG32_SOC15(VCN, i, regUVD_RB_SIZE, ring->ring_size / 4);
+
+	tmp = RREG32_SOC15(VCN, i, regVCN_RB_ENABLE);
+	tmp &= ~(VCN_RB_ENABLE__RB1_EN_MASK);
+	WREG32_SOC15(VCN, i, regVCN_RB_ENABLE, tmp);
+	fw_shared->sq.queue_mode |= FW_QUEUE_RING_RESET;
+	WREG32_SOC15(VCN, i, regUVD_RB_RPTR, 0);
+	WREG32_SOC15(VCN, i, regUVD_RB_WPTR, 0);
+
+	tmp = RREG32_SOC15(VCN, i, regUVD_RB_RPTR);
+	WREG32_SOC15(VCN, i, regUVD_RB_WPTR, tmp);
+	ring->wptr = RREG32_SOC15(VCN, i, regUVD_RB_WPTR);
+
+	tmp = RREG32_SOC15(VCN, i, regVCN_RB_ENABLE);
+	tmp |= VCN_RB_ENABLE__RB1_EN_MASK;
+	WREG32_SOC15(VCN, i, regVCN_RB_ENABLE, tmp);
+	fw_shared->sq.queue_mode &= ~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF);
+
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, i, regUVD_STATUS);
+
+	return 0;
+}
+
+static int vcn_v4_0_init_ring_metadata(struct amdgpu_device *adev, uint32_t vcn_inst, struct amdgpu_ring *ring_enc)
+{
+	struct amdgpu_vcn_rb_metadata *rb_metadata = NULL;
+	uint8_t *rb_ptr = (uint8_t *)ring_enc->ring;
+
+	rb_ptr += ring_enc->ring_size;
+	rb_metadata = (struct amdgpu_vcn_rb_metadata *)rb_ptr;
+
+	memset(rb_metadata, 0, sizeof(struct amdgpu_vcn_rb_metadata));
+	rb_metadata->size = sizeof(struct amdgpu_vcn_rb_metadata);
+	rb_metadata->present_flag_0 |= cpu_to_le32(AMDGPU_VCN_VF_RB_SETUP_FLAG);
+	rb_metadata->present_flag_0 |= cpu_to_le32(AMDGPU_VCN_VF_RB_DECOUPLE_FLAG);
+	rb_metadata->version = 1;
+	rb_metadata->ring_id = vcn_inst & 0xFF;
 
 	return 0;
 }
@@ -1227,8 +1354,8 @@ static int vcn_v4_0_start_sriov(struct amdgpu_device *adev)
 	struct mmsch_v4_0_cmd_end end = { {0} };
 	struct mmsch_v4_0_init_header header;
 
-	volatile struct amdgpu_vcn4_fw_shared *fw_shared;
-	volatile struct amdgpu_fw_shared_rb_setup *rb_setup;
+	struct amdgpu_vcn4_fw_shared *fw_shared;
+	struct amdgpu_fw_shared_rb_setup *rb_setup;
 
 	direct_wt.cmd_header.command_type =
 		MMSCH_COMMAND__DIRECT_REG_WRITE;
@@ -1239,7 +1366,7 @@ static int vcn_v4_0_start_sriov(struct amdgpu_device *adev)
 
 	header.version = MMSCH_VERSION;
 	header.total_size = sizeof(struct mmsch_v4_0_init_header) >> 2;
-	for (i = 0; i < AMDGPU_MAX_VCN_INSTANCES; i++) {
+	for (i = 0; i < MMSCH_V4_0_VCN_INSTANCES; i++) {
 		header.inst[i].init_status = 0;
 		header.inst[i].table_offset = 0;
 		header.inst[i].table_size = 0;
@@ -1251,13 +1378,16 @@ static int vcn_v4_0_start_sriov(struct amdgpu_device *adev)
 		if (adev->vcn.harvest_config & (1 << i))
 			continue;
 
+		// Must re/init fw_shared at beginning
+		vcn_v4_0_fw_shared_init(adev, i);
+
 		table_size = 0;
 
 		MMSCH_V4_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCN, i,
 			regUVD_STATUS),
 			~UVD_STATUS__UVD_BUSY, UVD_STATUS__UVD_BUSY);
 
-		cache_size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+		cache_size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.inst[i].fw->size + 4);
 
 		if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 			MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i,
@@ -1324,11 +1454,30 @@ static int vcn_v4_0_start_sriov(struct amdgpu_device *adev)
 		rb_enc_addr = ring_enc->gpu_addr;
 
 		rb_setup->is_rb_enabled_flags |= RB_ENABLED;
-		rb_setup->rb_addr_lo = lower_32_bits(rb_enc_addr);
-		rb_setup->rb_addr_hi = upper_32_bits(rb_enc_addr);
-		rb_setup->rb_size = ring_enc->ring_size / 4;
 		fw_shared->present_flag_0 |= cpu_to_le32(AMDGPU_VCN_VF_RB_SETUP_FLAG);
 
+		if (amdgpu_sriov_is_vcn_rb_decouple(adev)) {
+			vcn_v4_0_init_ring_metadata(adev, i, ring_enc);
+
+			memset((void *)&rb_setup->rb_info, 0, sizeof(struct amdgpu_vcn_rb_setup_info) * MAX_NUM_VCN_RB_SETUP);
+			if (!(adev->vcn.harvest_config & (1 << 0))) {
+				rb_setup->rb_info[0].rb_addr_lo = lower_32_bits(adev->vcn.inst[0].ring_enc[0].gpu_addr);
+				rb_setup->rb_info[0].rb_addr_hi = upper_32_bits(adev->vcn.inst[0].ring_enc[0].gpu_addr);
+				rb_setup->rb_info[0].rb_size = adev->vcn.inst[0].ring_enc[0].ring_size / 4;
+			}
+			if (!(adev->vcn.harvest_config & (1 << 1))) {
+				rb_setup->rb_info[2].rb_addr_lo = lower_32_bits(adev->vcn.inst[1].ring_enc[0].gpu_addr);
+				rb_setup->rb_info[2].rb_addr_hi = upper_32_bits(adev->vcn.inst[1].ring_enc[0].gpu_addr);
+				rb_setup->rb_info[2].rb_size = adev->vcn.inst[1].ring_enc[0].ring_size / 4;
+			}
+			fw_shared->decouple.is_enabled = 1;
+			fw_shared->present_flag_0 |= cpu_to_le32(AMDGPU_VCN_VF_RB_DECOUPLE_FLAG);
+		} else {
+			rb_setup->rb_addr_lo = lower_32_bits(rb_enc_addr);
+			rb_setup->rb_addr_hi = upper_32_bits(rb_enc_addr);
+			rb_setup->rb_size = ring_enc->ring_size / 4;
+		}
+
 		MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i,
 			regUVD_LMI_VCPU_NC0_64BIT_BAR_LOW),
 			lower_32_bits(adev->vcn.inst[i].fw_shared.gpu_addr));
@@ -1413,15 +1562,18 @@ static int vcn_v4_0_start_sriov(struct amdgpu_device *adev)
 /**
  * vcn_v4_0_stop_dpg_mode - VCN stop with dpg mode
  *
- * @adev: amdgpu_device pointer
- * @inst_idx: instance number index
+ * @vinst: VCN instance
  *
  * Stop VCN block with dpg mode
  */
-static void vcn_v4_0_stop_dpg_mode(struct amdgpu_device *adev, int inst_idx)
+static void vcn_v4_0_stop_dpg_mode(struct amdgpu_vcn_inst *vinst)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	struct dpg_pause_state state = {.fw_based = VCN_DPG_STATE__UNPAUSE};
 	uint32_t tmp;
 
+	vcn_v4_0_pause_dpg_mode(vinst, &state);
 	/* Wait for power status to be 1 */
 	SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_POWER_STATUS, 1,
 		UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
@@ -1436,87 +1588,101 @@ static void vcn_v4_0_stop_dpg_mode(struct amdgpu_device *adev, int inst_idx)
 	/* disable dynamic power gating mode */
 	WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, regUVD_POWER_STATUS), 0,
 		~UVD_POWER_STATUS__UVD_PG_MODE_MASK);
+
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, inst_idx, regUVD_STATUS);
 }
 
 /**
  * vcn_v4_0_stop - VCN stop
  *
- * @adev: amdgpu_device pointer
+ * @vinst: VCN instance
  *
  * Stop VCN block
  */
-static int vcn_v4_0_stop(struct amdgpu_device *adev)
+static int vcn_v4_0_stop(struct amdgpu_vcn_inst *vinst)
 {
-	volatile struct amdgpu_vcn4_fw_shared *fw_shared;
+	struct amdgpu_device *adev = vinst->adev;
+	int i = vinst->inst;
+	struct amdgpu_vcn4_fw_shared *fw_shared;
 	uint32_t tmp;
-	int i, r = 0;
+	int r = 0;
 
-	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
-		fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
-		fw_shared->sq.queue_mode |= FW_QUEUE_DPG_HOLD_OFF;
+	if (adev->vcn.harvest_config & (1 << i))
+		return 0;
 
-		if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
-			vcn_v4_0_stop_dpg_mode(adev, i);
-			continue;
-		}
+	fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
+	fw_shared->sq.queue_mode |= FW_QUEUE_DPG_HOLD_OFF;
 
-		/* wait for vcn idle */
-		r = SOC15_WAIT_ON_RREG(VCN, i, regUVD_STATUS, UVD_STATUS__IDLE, 0x7);
-		if (r)
-			return r;
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
+		vcn_v4_0_stop_dpg_mode(vinst);
+		goto done;
+	}
 
-		tmp = UVD_LMI_STATUS__VCPU_LMI_WRITE_CLEAN_MASK |
-			UVD_LMI_STATUS__READ_CLEAN_MASK |
-			UVD_LMI_STATUS__WRITE_CLEAN_MASK |
-			UVD_LMI_STATUS__WRITE_CLEAN_RAW_MASK;
-		r = SOC15_WAIT_ON_RREG(VCN, i, regUVD_LMI_STATUS, tmp, tmp);
-		if (r)
-			return r;
+	/* wait for vcn idle */
+	r = SOC15_WAIT_ON_RREG(VCN, i, regUVD_STATUS, UVD_STATUS__IDLE, 0x7);
+	if (r)
+		goto done;
 
-		/* disable LMI UMC channel */
-		tmp = RREG32_SOC15(VCN, i, regUVD_LMI_CTRL2);
-		tmp |= UVD_LMI_CTRL2__STALL_ARB_UMC_MASK;
-		WREG32_SOC15(VCN, i, regUVD_LMI_CTRL2, tmp);
-		tmp = UVD_LMI_STATUS__UMC_READ_CLEAN_RAW_MASK |
-			UVD_LMI_STATUS__UMC_WRITE_CLEAN_RAW_MASK;
-		r = SOC15_WAIT_ON_RREG(VCN, i, regUVD_LMI_STATUS, tmp, tmp);
-		if (r)
-			return r;
+	tmp = UVD_LMI_STATUS__VCPU_LMI_WRITE_CLEAN_MASK |
+		UVD_LMI_STATUS__READ_CLEAN_MASK |
+		UVD_LMI_STATUS__WRITE_CLEAN_MASK |
+		UVD_LMI_STATUS__WRITE_CLEAN_RAW_MASK;
+	r = SOC15_WAIT_ON_RREG(VCN, i, regUVD_LMI_STATUS, tmp, tmp);
+	if (r)
+		goto done;
+
+	/* disable LMI UMC channel */
+	tmp = RREG32_SOC15(VCN, i, regUVD_LMI_CTRL2);
+	tmp |= UVD_LMI_CTRL2__STALL_ARB_UMC_MASK;
+	WREG32_SOC15(VCN, i, regUVD_LMI_CTRL2, tmp);
+	tmp = UVD_LMI_STATUS__UMC_READ_CLEAN_RAW_MASK |
+		UVD_LMI_STATUS__UMC_WRITE_CLEAN_RAW_MASK;
+	r = SOC15_WAIT_ON_RREG(VCN, i, regUVD_LMI_STATUS, tmp, tmp);
+	if (r)
+		goto done;
 
-		/* block VCPU register access */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_RB_ARB_CTRL),
-				UVD_RB_ARB_CTRL__VCPU_DIS_MASK,
-				~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
+	/* block VCPU register access */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_RB_ARB_CTRL),
+		 UVD_RB_ARB_CTRL__VCPU_DIS_MASK,
+		 ~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
 
-		/* reset VCPU */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL),
-				UVD_VCPU_CNTL__BLK_RST_MASK,
-				~UVD_VCPU_CNTL__BLK_RST_MASK);
+	/* reset VCPU */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL),
+		 UVD_VCPU_CNTL__BLK_RST_MASK,
+		 ~UVD_VCPU_CNTL__BLK_RST_MASK);
 
-		/* disable VCPU clock */
-		WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), 0,
-				~(UVD_VCPU_CNTL__CLK_EN_MASK));
+	/* disable VCPU clock */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), 0,
+		 ~(UVD_VCPU_CNTL__CLK_EN_MASK));
 
-		/* apply soft reset */
-		tmp = RREG32_SOC15(VCN, i, regUVD_SOFT_RESET);
-		tmp |= UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
-		WREG32_SOC15(VCN, i, regUVD_SOFT_RESET, tmp);
-		tmp = RREG32_SOC15(VCN, i, regUVD_SOFT_RESET);
-		tmp |= UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
-		WREG32_SOC15(VCN, i, regUVD_SOFT_RESET, tmp);
+	/* apply soft reset */
+	tmp = RREG32_SOC15(VCN, i, regUVD_SOFT_RESET);
+	tmp |= UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
+	WREG32_SOC15(VCN, i, regUVD_SOFT_RESET, tmp);
+	tmp = RREG32_SOC15(VCN, i, regUVD_SOFT_RESET);
+	tmp |= UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
+	WREG32_SOC15(VCN, i, regUVD_SOFT_RESET, tmp);
 
-		/* clear status */
-		WREG32_SOC15(VCN, i, regUVD_STATUS, 0);
+	/* clear status */
+	WREG32_SOC15(VCN, i, regUVD_STATUS, 0);
 
-		/* apply HW clock gating */
-		vcn_v4_0_enable_clock_gating(adev, i);
+	/* apply HW clock gating */
+	vcn_v4_0_enable_clock_gating(vinst);
 
-		/* enable VCN power gating */
-		vcn_v4_0_enable_static_power_gating(adev, i);
-	}
+	/* enable VCN power gating */
+	vcn_v4_0_enable_static_power_gating(vinst);
 
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, i, regUVD_STATUS);
+
+done:
 	if (adev->pm.dpm_enabled)
-		amdgpu_dpm_enable_uvd(adev, false);
+		amdgpu_dpm_enable_vcn(adev, false, i);
 
 	return 0;
 }
@@ -1524,15 +1690,16 @@ static int vcn_v4_0_stop(struct amdgpu_device *adev)
 /**
  * vcn_v4_0_pause_dpg_mode - VCN pause with dpg mode
  *
- * @adev: amdgpu_device pointer
- * @inst_idx: instance number index
+ * @vinst: VCN instance
  * @new_state: pause state
  *
  * Pause dpg mode for VCN block
  */
-static int vcn_v4_0_pause_dpg_mode(struct amdgpu_device *adev, int inst_idx,
-      struct dpg_pause_state *new_state)
+static int vcn_v4_0_pause_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				   struct dpg_pause_state *new_state)
 {
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
 	uint32_t reg_data = 0;
 	int ret_code;
 
@@ -1634,15 +1801,19 @@ static int vcn_v4_0_limit_sched(struct amdgpu_cs_parser *p,
 				struct amdgpu_job *job)
 {
 	struct drm_gpu_scheduler **scheds;
-
-	/* The create msg must be in the first IB submitted */
-	if (atomic_read(&job->base.entity->fence_seq))
-		return -EINVAL;
+	struct dma_fence *fence;
 
 	/* if VCN0 is harvested, we can't support AV1 */
 	if (p->adev->vcn.harvest_config & AMDGPU_VCN_HARVEST_VCN0)
 		return -EINVAL;
 
+	/* wait for all jobs to finish before switching to instance 0 */
+	fence = amdgpu_ctx_get_fence(p->ctx, job->base.entity, ~0ull);
+	if (fence) {
+		dma_fence_wait(fence, false);
+		dma_fence_put(fence);
+	}
+
 	scheds = p->adev->gpu_sched[AMDGPU_HW_IP_VCN_ENC]
 		[AMDGPU_RING_PRIO_0].sched;
 	drm_sched_entity_modify_sched(job->base.entity, scheds, 1);
@@ -1733,22 +1904,16 @@ out:
 
 #define RADEON_VCN_ENGINE_TYPE_ENCODE			(0x00000002)
 #define RADEON_VCN_ENGINE_TYPE_DECODE			(0x00000003)
-
 #define RADEON_VCN_ENGINE_INFO				(0x30000001)
-#define RADEON_VCN_ENGINE_INFO_MAX_OFFSET		16
-
 #define RENCODE_ENCODE_STANDARD_AV1			2
 #define RENCODE_IB_PARAM_SESSION_INIT			0x00000003
-#define RENCODE_IB_PARAM_SESSION_INIT_MAX_OFFSET	64
 
-/* return the offset in ib if id is found, -1 otherwise
- * to speed up the searching we only search upto max_offset
- */
-static int vcn_v4_0_enc_find_ib_param(struct amdgpu_ib *ib, uint32_t id, int max_offset)
+/* return the offset in ib if id is found, -1 otherwise */
+static int vcn_v4_0_enc_find_ib_param(struct amdgpu_ib *ib, uint32_t id, int start)
 {
 	int i;
 
-	for (i = 0; i < ib->length_dw && i < max_offset && ib->ptr[i] >= 8; i += ib->ptr[i]/4) {
+	for (i = start; i < ib->length_dw && ib->ptr[i] >= 8; i += ib->ptr[i] / 4) {
 		if (ib->ptr[i + 1] == id)
 			return i;
 	}
@@ -1763,41 +1928,56 @@ static int vcn_v4_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p,
 	struct amdgpu_vcn_decode_buffer *decode_buffer;
 	uint64_t addr;
 	uint32_t val;
-	int idx;
+	int idx = 0, sidx;
 
 	/* The first instance can decode anything */
 	if (!ring->me)
 		return 0;
 
-	/* RADEON_VCN_ENGINE_INFO is at the top of ib block */
-	idx = vcn_v4_0_enc_find_ib_param(ib, RADEON_VCN_ENGINE_INFO,
-			RADEON_VCN_ENGINE_INFO_MAX_OFFSET);
-	if (idx < 0) /* engine info is missing */
-		return 0;
-
-	val = amdgpu_ib_get_value(ib, idx + 2); /* RADEON_VCN_ENGINE_TYPE */
-	if (val == RADEON_VCN_ENGINE_TYPE_DECODE) {
-		decode_buffer = (struct amdgpu_vcn_decode_buffer *)&ib->ptr[idx + 6];
-
-		if (!(decode_buffer->valid_buf_flag  & 0x1))
-			return 0;
-
-		addr = ((u64)decode_buffer->msg_buffer_address_hi) << 32 |
-			decode_buffer->msg_buffer_address_lo;
-		return vcn_v4_0_dec_msg(p, job, addr);
-	} else if (val == RADEON_VCN_ENGINE_TYPE_ENCODE) {
-		idx = vcn_v4_0_enc_find_ib_param(ib, RENCODE_IB_PARAM_SESSION_INIT,
-			RENCODE_IB_PARAM_SESSION_INIT_MAX_OFFSET);
-		if (idx >= 0 && ib->ptr[idx + 2] == RENCODE_ENCODE_STANDARD_AV1)
-			return vcn_v4_0_limit_sched(p, job);
+	while ((idx = vcn_v4_0_enc_find_ib_param(ib, RADEON_VCN_ENGINE_INFO, idx)) >= 0) {
+		val = amdgpu_ib_get_value(ib, idx + 2); /* RADEON_VCN_ENGINE_TYPE */
+		if (val == RADEON_VCN_ENGINE_TYPE_DECODE) {
+			decode_buffer = (struct amdgpu_vcn_decode_buffer *)&ib->ptr[idx + 6];
+
+			if (!(decode_buffer->valid_buf_flag & 0x1))
+				return 0;
+
+			addr = ((u64)decode_buffer->msg_buffer_address_hi) << 32 |
+				decode_buffer->msg_buffer_address_lo;
+			return vcn_v4_0_dec_msg(p, job, addr);
+		} else if (val == RADEON_VCN_ENGINE_TYPE_ENCODE) {
+			sidx = vcn_v4_0_enc_find_ib_param(ib, RENCODE_IB_PARAM_SESSION_INIT, idx);
+			if (sidx >= 0 && ib->ptr[sidx + 2] == RENCODE_ENCODE_STANDARD_AV1)
+				return vcn_v4_0_limit_sched(p, job);
+		}
+		idx += ib->ptr[idx] / 4;
 	}
 	return 0;
 }
 
-static const struct amdgpu_ring_funcs vcn_v4_0_unified_ring_vm_funcs = {
+static int vcn_v4_0_ring_reset(struct amdgpu_ring *ring,
+			       unsigned int vmid,
+			       struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[ring->me];
+	int r;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+	r = vcn_v4_0_stop(vinst);
+	if (r)
+		return r;
+	r = vcn_v4_0_start(vinst);
+	if (r)
+		return r;
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static struct amdgpu_ring_funcs vcn_v4_0_unified_ring_vm_funcs = {
 	.type = AMDGPU_RING_TYPE_VCN_ENC,
 	.align_mask = 0x3f,
 	.nop = VCN_ENC_CMD_NO_OP,
+	.extra_bytes = sizeof(struct amdgpu_vcn_rb_metadata),
 	.get_rptr = vcn_v4_0_unified_ring_get_rptr,
 	.get_wptr = vcn_v4_0_unified_ring_get_wptr,
 	.set_wptr = vcn_v4_0_unified_ring_set_wptr,
@@ -1822,6 +2002,7 @@ static const struct amdgpu_ring_funcs vcn_v4_0_unified_ring_vm_funcs = {
 	.emit_wreg = vcn_v2_0_enc_ring_emit_wreg,
 	.emit_reg_wait = vcn_v2_0_enc_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = vcn_v4_0_ring_reset,
 };
 
 /**
@@ -1839,23 +2020,25 @@ static void vcn_v4_0_set_unified_ring_funcs(struct amdgpu_device *adev)
 		if (adev->vcn.harvest_config & (1 << i))
 			continue;
 
-		adev->vcn.inst[i].ring_enc[0].funcs = &vcn_v4_0_unified_ring_vm_funcs;
-		adev->vcn.inst[i].ring_enc[0].me = i;
+		if (amdgpu_ip_version(adev, VCN_HWIP, 0) == IP_VERSION(4, 0, 2))
+			vcn_v4_0_unified_ring_vm_funcs.secure_submission_supported = true;
 
-		DRM_INFO("VCN(%d) encode/decode are enabled in VM mode\n", i);
+		adev->vcn.inst[i].ring_enc[0].funcs =
+		       (const struct amdgpu_ring_funcs *)&vcn_v4_0_unified_ring_vm_funcs;
+		adev->vcn.inst[i].ring_enc[0].me = i;
 	}
 }
 
 /**
  * vcn_v4_0_is_idle - check VCN block is idle
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block structure
  *
  * Check whether VCN block is idle
  */
-static bool vcn_v4_0_is_idle(void *handle)
+static bool vcn_v4_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i, ret = 1;
 
 	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
@@ -1871,13 +2054,13 @@ static bool vcn_v4_0_is_idle(void *handle)
 /**
  * vcn_v4_0_wait_for_idle - wait for VCN block idle
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Wait for VCN block idle
  */
-static int vcn_v4_0_wait_for_idle(void *handle)
+static int vcn_v4_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i, ret = 0;
 
 	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
@@ -1896,81 +2079,79 @@ static int vcn_v4_0_wait_for_idle(void *handle)
 /**
  * vcn_v4_0_set_clockgating_state - set VCN block clockgating state
  *
- * @handle: amdgpu_device pointer
+ * @ip_block: amdgpu_ip_block pointer
  * @state: clock gating state
  *
  * Set VCN block clockgating state
  */
-static int vcn_v4_0_set_clockgating_state(void *handle, enum amd_clockgating_state state)
+static int vcn_v4_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	bool enable = (state == AMD_CG_STATE_GATE) ? true : false;
+	struct amdgpu_device *adev = ip_block->adev;
+	bool enable = state == AMD_CG_STATE_GATE;
 	int i;
 
 	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[i];
+
 		if (adev->vcn.harvest_config & (1 << i))
 			continue;
 
 		if (enable) {
 			if (RREG32_SOC15(VCN, i, regUVD_STATUS) != UVD_STATUS__IDLE)
 				return -EBUSY;
-			vcn_v4_0_enable_clock_gating(adev, i);
+			vcn_v4_0_enable_clock_gating(vinst);
 		} else {
-			vcn_v4_0_disable_clock_gating(adev, i);
+			vcn_v4_0_disable_clock_gating(vinst);
 		}
 	}
 
 	return 0;
 }
 
-/**
- * vcn_v4_0_set_powergating_state - set VCN block powergating state
- *
- * @handle: amdgpu_device pointer
- * @state: power gating state
- *
- * Set VCN block powergating state
- */
-static int vcn_v4_0_set_powergating_state(void *handle, enum amd_powergating_state state)
+static int vcn_v4_0_set_pg_state(struct amdgpu_vcn_inst *vinst,
+				 enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	int ret;
+	struct amdgpu_device *adev = vinst->adev;
+	int ret = 0;
 
 	/* for SRIOV, guest should not control VCN Power-gating
 	 * MMSCH FW should control Power-gating and clock-gating
 	 * guest should avoid touching CGC and PG
 	 */
 	if (amdgpu_sriov_vf(adev)) {
-		adev->vcn.cur_state = AMD_PG_STATE_UNGATE;
+		vinst->cur_state = AMD_PG_STATE_UNGATE;
 		return 0;
 	}
 
-	if(state == adev->vcn.cur_state)
+	if (state == vinst->cur_state)
 		return 0;
 
 	if (state == AMD_PG_STATE_GATE)
-		ret = vcn_v4_0_stop(adev);
+		ret = vcn_v4_0_stop(vinst);
 	else
-		ret = vcn_v4_0_start(adev);
+		ret = vcn_v4_0_start(vinst);
 
-	if(!ret)
-		adev->vcn.cur_state = state;
+	if (!ret)
+		vinst->cur_state = state;
 
 	return ret;
 }
 
 /**
- * vcn_v4_0_set_interrupt_state - set VCN block interrupt state
+ * vcn_v4_0_set_ras_interrupt_state - set VCN block RAS interrupt state
  *
  * @adev: amdgpu_device pointer
  * @source: interrupt sources
  * @type: interrupt types
  * @state: interrupt states
  *
- * Set VCN block interrupt state
+ * Set VCN block RAS interrupt state
  */
-static int vcn_v4_0_set_interrupt_state(struct amdgpu_device *adev, struct amdgpu_irq_src *source,
-      unsigned type, enum amdgpu_interrupt_state state)
+static int vcn_v4_0_set_ras_interrupt_state(struct amdgpu_device *adev,
+	struct amdgpu_irq_src *source,
+	unsigned int type,
+	enum amdgpu_interrupt_state state)
 {
 	return 0;
 }
@@ -1989,16 +2170,20 @@ static int vcn_v4_0_process_interrupt(struct amdgpu_device *adev, struct amdgpu_
 {
 	uint32_t ip_instance;
 
-	switch (entry->client_id) {
-	case SOC15_IH_CLIENTID_VCN:
-		ip_instance = 0;
-		break;
-	case SOC15_IH_CLIENTID_VCN1:
-		ip_instance = 1;
-		break;
-	default:
-		DRM_ERROR("Unhandled client id: %d\n", entry->client_id);
-		return 0;
+	if (amdgpu_sriov_is_vcn_rb_decouple(adev)) {
+		ip_instance = entry->ring_id;
+	} else {
+		switch (entry->client_id) {
+		case SOC15_IH_CLIENTID_VCN:
+			ip_instance = 0;
+			break;
+		case SOC15_IH_CLIENTID_VCN1:
+			ip_instance = 1;
+			break;
+		default:
+			DRM_ERROR("Unhandled client id: %d\n", entry->client_id);
+			return 0;
+		}
 	}
 
 	DRM_DEBUG("IH: VCN TRAP\n");
@@ -2007,9 +2192,6 @@ static int vcn_v4_0_process_interrupt(struct amdgpu_device *adev, struct amdgpu_
 	case VCN_4_0__SRCID__UVD_ENC_GENERAL_PURPOSE:
 		amdgpu_fence_process(&adev->vcn.inst[ip_instance].ring_enc[0]);
 		break;
-	case VCN_4_0__SRCID_UVD_POISON:
-		amdgpu_vcn_process_poison_irq(adev, source, entry);
-		break;
 	default:
 		DRM_ERROR("Unhandled interrupt: %d %d\n",
 			  entry->src_id, entry->src_data[0]);
@@ -2020,10 +2202,14 @@ static int vcn_v4_0_process_interrupt(struct amdgpu_device *adev, struct amdgpu_
 }
 
 static const struct amdgpu_irq_src_funcs vcn_v4_0_irq_funcs = {
-	.set = vcn_v4_0_set_interrupt_state,
 	.process = vcn_v4_0_process_interrupt,
 };
 
+static const struct amdgpu_irq_src_funcs vcn_v4_0_ras_irq_funcs = {
+	.set = vcn_v4_0_set_ras_interrupt_state,
+	.process = amdgpu_vcn_process_poison_irq,
+};
+
 /**
  * vcn_v4_0_set_irq_funcs - set VCN block interrupt irq functions
  *
@@ -2039,15 +2225,17 @@ static void vcn_v4_0_set_irq_funcs(struct amdgpu_device *adev)
 		if (adev->vcn.harvest_config & (1 << i))
 			continue;
 
-		adev->vcn.inst[i].irq.num_types = adev->vcn.num_enc_rings + 1;
+		adev->vcn.inst[i].irq.num_types = adev->vcn.inst[i].num_enc_rings + 1;
 		adev->vcn.inst[i].irq.funcs = &vcn_v4_0_irq_funcs;
+
+		adev->vcn.inst[i].ras_poison_irq.num_types = adev->vcn.inst[i].num_enc_rings + 1;
+		adev->vcn.inst[i].ras_poison_irq.funcs = &vcn_v4_0_ras_irq_funcs;
 	}
 }
 
 static const struct amd_ip_funcs vcn_v4_0_ip_funcs = {
 	.name = "vcn_v4_0",
 	.early_init = vcn_v4_0_early_init,
-	.late_init = NULL,
 	.sw_init = vcn_v4_0_sw_init,
 	.sw_fini = vcn_v4_0_sw_fini,
 	.hw_init = vcn_v4_0_hw_init,
@@ -2056,16 +2244,13 @@ static const struct amd_ip_funcs vcn_v4_0_ip_funcs = {
 	.resume = vcn_v4_0_resume,
 	.is_idle = vcn_v4_0_is_idle,
 	.wait_for_idle = vcn_v4_0_wait_for_idle,
-	.check_soft_reset = NULL,
-	.pre_soft_reset = NULL,
-	.soft_reset = NULL,
-	.post_soft_reset = NULL,
 	.set_clockgating_state = vcn_v4_0_set_clockgating_state,
-	.set_powergating_state = vcn_v4_0_set_powergating_state,
+	.set_powergating_state = vcn_set_powergating_state,
+	.dump_ip_state = amdgpu_vcn_dump_ip_state,
+	.print_ip_state = amdgpu_vcn_print_ip_state,
 };
 
-const struct amdgpu_ip_block_version vcn_v4_0_ip_block =
-{
+const struct amdgpu_ip_block_version vcn_v4_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_VCN,
 	.major = 4,
 	.minor = 0,
@@ -2114,12 +2299,13 @@ const struct amdgpu_ras_block_hw_ops vcn_v4_0_ras_hw_ops = {
 static struct amdgpu_vcn_ras vcn_v4_0_ras = {
 	.ras_block = {
 		.hw_ops = &vcn_v4_0_ras_hw_ops,
+		.ras_late_init = amdgpu_vcn_ras_late_init,
 	},
 };
 
 static void vcn_v4_0_set_ras_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[VCN_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, VCN_HWIP, 0)) {
 	case IP_VERSION(4, 0, 0):
 		adev->vcn.ras = &vcn_v4_0_ras;
 		break;
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c
new file mode 100644
index 000000000000..cb7123ec1a5d
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c
@@ -0,0 +1,2131 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/firmware.h>
+#include <drm/drm_drv.h>
+
+#include "amdgpu.h"
+#include "amdgpu_vcn.h"
+#include "amdgpu_pm.h"
+#include "soc15.h"
+#include "soc15d.h"
+#include "soc15_hw_ip.h"
+#include "vcn_v2_0.h"
+#include "vcn_v4_0_3.h"
+#include "mmsch_v4_0_3.h"
+
+#include "vcn/vcn_4_0_3_offset.h"
+#include "vcn/vcn_4_0_3_sh_mask.h"
+#include "ivsrcid/vcn/irqsrcs_vcn_4_0.h"
+
+#define mmUVD_DPG_LMA_CTL		regUVD_DPG_LMA_CTL
+#define mmUVD_DPG_LMA_CTL_BASE_IDX	regUVD_DPG_LMA_CTL_BASE_IDX
+#define mmUVD_DPG_LMA_DATA		regUVD_DPG_LMA_DATA
+#define mmUVD_DPG_LMA_DATA_BASE_IDX	regUVD_DPG_LMA_DATA_BASE_IDX
+
+#define VCN_VID_SOC_ADDRESS_2_0		0x1fb00
+#define VCN1_VID_SOC_ADDRESS_3_0	0x48300
+#define VCN1_AON_SOC_ADDRESS_3_0	0x48000
+
+static const struct amdgpu_hwip_reg_entry vcn_reg_list_4_0_3[] = {
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_POWER_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_CONTEXT_ID),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_CONTEXT_ID2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_GPCOM_VCPU_DATA0),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_GPCOM_VCPU_DATA1),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_GPCOM_VCPU_CMD),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_PGFSM_CONFIG),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_PGFSM_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_LMA_CTL),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_LMA_DATA),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_LMA_MASK),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_PAUSE)
+};
+
+#define NORMALIZE_VCN_REG_OFFSET(offset) \
+		(offset & 0x1FFFF)
+
+static int vcn_v4_0_3_start_sriov(struct amdgpu_device *adev);
+static void vcn_v4_0_3_set_unified_ring_funcs(struct amdgpu_device *adev);
+static void vcn_v4_0_3_set_irq_funcs(struct amdgpu_device *adev);
+static int vcn_v4_0_3_set_pg_state(struct amdgpu_vcn_inst *vinst,
+				   enum amd_powergating_state state);
+static int vcn_v4_0_3_pause_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				     struct dpg_pause_state *new_state);
+static void vcn_v4_0_3_unified_ring_set_wptr(struct amdgpu_ring *ring);
+static void vcn_v4_0_3_set_ras_funcs(struct amdgpu_device *adev);
+static void vcn_v4_0_3_enable_ras(struct amdgpu_device *adev,
+				  int inst_idx, bool indirect);
+
+static inline bool vcn_v4_0_3_normalizn_reqd(struct amdgpu_device *adev)
+{
+	return (adev->vcn.caps & AMDGPU_VCN_CAPS(RRMT_ENABLED)) == 0;
+}
+
+/**
+ * vcn_v4_0_3_early_init - set function pointers
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Set ring and irq function pointers
+ */
+static int vcn_v4_0_3_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, r;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i)
+		/* re-use enc ring as unified ring */
+		adev->vcn.inst[i].num_enc_rings = 1;
+
+	vcn_v4_0_3_set_unified_ring_funcs(adev);
+	vcn_v4_0_3_set_irq_funcs(adev);
+	vcn_v4_0_3_set_ras_funcs(adev);
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		adev->vcn.inst[i].set_pg_state = vcn_v4_0_3_set_pg_state;
+
+		r = amdgpu_vcn_early_init(adev, i);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+static int vcn_v4_0_3_late_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	adev->vcn.supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->vcn.inst[0].ring_enc[0]);
+
+	if (amdgpu_dpm_reset_vcn_is_supported(adev) && !amdgpu_sriov_vf(adev))
+		adev->vcn.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+
+	return 0;
+}
+
+static int vcn_v4_0_3_fw_shared_init(struct amdgpu_device *adev, int inst_idx)
+{
+	struct amdgpu_vcn4_fw_shared *fw_shared;
+
+	fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr;
+	fw_shared->present_flag_0 = cpu_to_le32(AMDGPU_FW_SHARED_FLAG_0_UNIFIED_QUEUE);
+	fw_shared->sq.is_enabled = 1;
+
+	if (amdgpu_vcnfw_log)
+		amdgpu_vcn_fwlog_init(&adev->vcn.inst[inst_idx]);
+
+	return 0;
+}
+
+/**
+ * vcn_v4_0_3_sw_init - sw init for VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Load firmware and sw initialization
+ */
+static int vcn_v4_0_3_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring;
+	int i, r, vcn_inst;
+
+	/* VCN DEC TRAP */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
+		VCN_4_0__SRCID__UVD_ENC_GENERAL_PURPOSE, &adev->vcn.inst->irq);
+	if (r)
+		return r;
+
+	/* VCN POISON TRAP */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
+		VCN_4_0__SRCID_UVD_POISON, &adev->vcn.inst->ras_poison_irq);
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+
+		r = amdgpu_vcn_sw_init(adev, i);
+		if (r)
+			return r;
+
+		amdgpu_vcn_setup_ucode(adev, i);
+
+		r = amdgpu_vcn_resume(adev, i);
+		if (r)
+			return r;
+
+		vcn_inst = GET_INST(VCN, i);
+
+		ring = &adev->vcn.inst[i].ring_enc[0];
+		ring->use_doorbell = true;
+
+		if (!amdgpu_sriov_vf(adev))
+			ring->doorbell_index =
+				(adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
+				9 * vcn_inst;
+		else
+			ring->doorbell_index =
+				(adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
+				32 * vcn_inst;
+
+		ring->vm_hub = AMDGPU_MMHUB0(adev->vcn.inst[i].aid_id);
+		sprintf(ring->name, "vcn_unified_%d", adev->vcn.inst[i].aid_id);
+
+		/* There are no per-instance irq source IDs on 4.0.3, the IH
+		 * packets use a separate field to differentiate instances.
+		 */
+		r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[0].irq, 0,
+				     AMDGPU_RING_PRIO_DEFAULT,
+				     &adev->vcn.inst[i].sched_score);
+		if (r)
+			return r;
+
+		vcn_v4_0_3_fw_shared_init(adev, i);
+
+		if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
+			adev->vcn.inst[i].pause_dpg_mode = vcn_v4_0_3_pause_dpg_mode;
+	}
+
+	if (amdgpu_sriov_vf(adev)) {
+		r = amdgpu_virt_alloc_mm_table(adev);
+		if (r)
+			return r;
+	}
+
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__VCN)) {
+		r = amdgpu_vcn_ras_sw_init(adev);
+		if (r) {
+			dev_err(adev->dev, "Failed to initialize vcn ras block!\n");
+			return r;
+		}
+	}
+
+	r = amdgpu_vcn_reg_dump_init(adev, vcn_reg_list_4_0_3, ARRAY_SIZE(vcn_reg_list_4_0_3));
+	if (r)
+		return r;
+
+	return amdgpu_vcn_sysfs_reset_mask_init(adev);
+}
+
+/**
+ * vcn_v4_0_3_sw_fini - sw fini for VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * VCN suspend and free up sw allocation
+ */
+static int vcn_v4_0_3_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, r, idx;
+
+	if (drm_dev_enter(&adev->ddev, &idx)) {
+		for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+			struct amdgpu_vcn4_fw_shared *fw_shared;
+
+			fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
+			fw_shared->present_flag_0 = 0;
+			fw_shared->sq.is_enabled = cpu_to_le32(false);
+		}
+		drm_dev_exit(idx);
+	}
+
+	if (amdgpu_sriov_vf(adev))
+		amdgpu_virt_free_mm_table(adev);
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_suspend(adev, i);
+		if (r)
+			return r;
+	}
+
+	amdgpu_vcn_sysfs_reset_mask_fini(adev);
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++)
+		amdgpu_vcn_sw_fini(adev, i);
+
+	return 0;
+}
+
+static int vcn_v4_0_3_hw_init_inst(struct amdgpu_vcn_inst *vinst)
+{
+	int vcn_inst;
+	struct amdgpu_device *adev = vinst->adev;
+	struct amdgpu_ring *ring;
+	int inst_idx = vinst->inst;
+
+	vcn_inst = GET_INST(VCN, inst_idx);
+	ring = &adev->vcn.inst[inst_idx].ring_enc[0];
+	if (ring->use_doorbell) {
+		adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell,
+			(adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 9 * vcn_inst,
+			adev->vcn.inst[inst_idx].aid_id);
+
+		WREG32_SOC15(VCN, vcn_inst, regVCN_RB1_DB_CTRL,
+			ring->doorbell_index << VCN_RB1_DB_CTRL__OFFSET__SHIFT |
+			VCN_RB1_DB_CTRL__EN_MASK);
+
+		/* Read DB_CTRL to flush the write DB_CTRL command. */
+		RREG32_SOC15(VCN, vcn_inst, regVCN_RB1_DB_CTRL);
+	}
+
+	return 0;
+}
+
+/**
+ * vcn_v4_0_3_hw_init - start and test VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Initialize the hardware, boot up the VCPU and do some testing
+ */
+static int vcn_v4_0_3_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring;
+	struct amdgpu_vcn_inst *vinst;
+	int i, r;
+
+	if (amdgpu_sriov_vf(adev)) {
+		r = vcn_v4_0_3_start_sriov(adev);
+		if (r)
+			return r;
+
+		for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+			ring = &adev->vcn.inst[i].ring_enc[0];
+			ring->wptr = 0;
+			ring->wptr_old = 0;
+			vcn_v4_0_3_unified_ring_set_wptr(ring);
+			ring->sched.ready = true;
+		}
+	} else {
+		/* This flag is not set for VF, assumed to be disabled always */
+		if (RREG32_SOC15(VCN, GET_INST(VCN, 0), regVCN_RRMT_CNTL) &
+		    0x100)
+			adev->vcn.caps |= AMDGPU_VCN_CAPS(RRMT_ENABLED);
+
+		for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+			struct amdgpu_vcn4_fw_shared *fw_shared;
+
+			ring = &adev->vcn.inst[i].ring_enc[0];
+			vinst = &adev->vcn.inst[i];
+			vcn_v4_0_3_hw_init_inst(vinst);
+
+			/* Re-init fw_shared when RAS fatal error occurred */
+			fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
+			if (!fw_shared->sq.is_enabled)
+				vcn_v4_0_3_fw_shared_init(adev, i);
+
+			r = amdgpu_ring_test_helper(ring);
+			if (r)
+				return r;
+		}
+	}
+
+	return r;
+}
+
+/**
+ * vcn_v4_0_3_hw_fini - stop the hardware block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Stop the VCN block, mark ring as not ready any more
+ */
+static int vcn_v4_0_3_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[i];
+
+		cancel_delayed_work_sync(&vinst->idle_work);
+
+		if (vinst->cur_state != AMD_PG_STATE_GATE)
+			vinst->set_pg_state(vinst, AMD_PG_STATE_GATE);
+	}
+
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__VCN) && !amdgpu_sriov_vf(adev))
+		amdgpu_irq_put(adev, &adev->vcn.inst->ras_poison_irq, 0);
+
+	return 0;
+}
+
+/**
+ * vcn_v4_0_3_suspend - suspend VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * HW fini and suspend VCN block
+ */
+static int vcn_v4_0_3_suspend(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r, i;
+
+	r = vcn_v4_0_3_hw_fini(ip_block);
+	if (r)
+		return r;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_suspend(adev, i);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+/**
+ * vcn_v4_0_3_resume - resume VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Resume firmware and hw init VCN block
+ */
+static int vcn_v4_0_3_resume(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r, i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_resume(ip_block->adev, i);
+		if (r)
+			return r;
+	}
+
+	r = vcn_v4_0_3_hw_init(ip_block);
+
+	return r;
+}
+
+/**
+ * vcn_v4_0_3_mc_resume - memory controller programming
+ *
+ * @vinst: VCN instance
+ *
+ * Let the VCN memory controller know it's offsets
+ */
+static void vcn_v4_0_3_mc_resume(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	uint32_t offset, size, vcn_inst;
+	const struct common_firmware_header *hdr;
+
+	hdr = (const struct common_firmware_header *)adev->vcn.inst[inst_idx].fw->data;
+	size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
+
+	vcn_inst = GET_INST(VCN, inst_idx);
+	/* cache window 0: fw */
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		WREG32_SOC15(
+			VCN, vcn_inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
+			(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst_idx]
+				 .tmr_mc_addr_lo));
+		WREG32_SOC15(
+			VCN, vcn_inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
+			(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst_idx]
+				 .tmr_mc_addr_hi));
+		WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_OFFSET0, 0);
+		offset = 0;
+	} else {
+		WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
+			     lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr));
+		WREG32_SOC15(VCN, vcn_inst,
+			     regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
+			     upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr));
+		offset = size;
+		WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_OFFSET0,
+			     AMDGPU_UVD_FIRMWARE_OFFSET >> 3);
+	}
+	WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_SIZE0, size);
+
+	/* cache window 1: stack */
+	WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW,
+		     lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset));
+	WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH,
+		     upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset));
+	WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_OFFSET1, 0);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_SIZE1,
+		     AMDGPU_VCN_STACK_SIZE);
+
+	/* cache window 2: context */
+	WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW,
+		     lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset +
+				   AMDGPU_VCN_STACK_SIZE));
+	WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH,
+		     upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset +
+				   AMDGPU_VCN_STACK_SIZE));
+	WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_OFFSET2, 0);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_SIZE2,
+		     AMDGPU_VCN_CONTEXT_SIZE);
+
+	/* non-cache window */
+	WREG32_SOC15(
+		VCN, vcn_inst, regUVD_LMI_VCPU_NC0_64BIT_BAR_LOW,
+		lower_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr));
+	WREG32_SOC15(
+		VCN, vcn_inst, regUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH,
+		upper_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr));
+	WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_NONCACHE_OFFSET0, 0);
+	WREG32_SOC15(
+		VCN, vcn_inst, regUVD_VCPU_NONCACHE_SIZE0,
+		AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn4_fw_shared)));
+}
+
+/**
+ * vcn_v4_0_3_mc_resume_dpg_mode - memory controller programming for dpg mode
+ *
+ * @vinst: VCN instance
+ * @indirect: indirectly write sram
+ *
+ * Let the VCN memory controller know it's offsets with dpg mode
+ */
+static void vcn_v4_0_3_mc_resume_dpg_mode(struct amdgpu_vcn_inst *vinst,
+					  bool indirect)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	uint32_t offset, size;
+	const struct common_firmware_header *hdr;
+
+	hdr = (const struct common_firmware_header *)adev->vcn.inst[inst_idx].fw->data;
+	size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
+
+	/* cache window 0: fw */
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		if (!indirect) {
+			WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+				VCN, 0, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW),
+				(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN +
+					inst_idx].tmr_mc_addr_lo), 0, indirect);
+			WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+				VCN, 0, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH),
+				(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN +
+					inst_idx].tmr_mc_addr_hi), 0, indirect);
+			WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+				VCN, 0, regUVD_VCPU_CACHE_OFFSET0), 0, 0, indirect);
+		} else {
+			WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+				VCN, 0, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW), 0, 0, indirect);
+			WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+				VCN, 0, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH), 0, 0, indirect);
+			WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+				VCN, 0, regUVD_VCPU_CACHE_OFFSET0), 0, 0, indirect);
+		}
+		offset = 0;
+	} else {
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW),
+			lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr), 0, indirect);
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH),
+			upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr), 0, indirect);
+		offset = size;
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_VCPU_CACHE_OFFSET0),
+			AMDGPU_UVD_FIRMWARE_OFFSET >> 3, 0, indirect);
+	}
+
+	if (!indirect)
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_VCPU_CACHE_SIZE0), size, 0, indirect);
+	else
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_VCPU_CACHE_SIZE0), 0, 0, indirect);
+
+	/* cache window 1: stack */
+	if (!indirect) {
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW),
+			lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset), 0, indirect);
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH),
+			upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset), 0, indirect);
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_VCPU_CACHE_OFFSET1), 0, 0, indirect);
+	} else {
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW), 0, 0, indirect);
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH), 0, 0, indirect);
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_VCPU_CACHE_OFFSET1), 0, 0, indirect);
+	}
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_VCPU_CACHE_SIZE1), AMDGPU_VCN_STACK_SIZE, 0, indirect);
+
+	/* cache window 2: context */
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW),
+			lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset +
+				AMDGPU_VCN_STACK_SIZE), 0, indirect);
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH),
+			upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset +
+				AMDGPU_VCN_STACK_SIZE), 0, indirect);
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_VCPU_CACHE_OFFSET2), 0, 0, indirect);
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_VCPU_CACHE_SIZE2), AMDGPU_VCN_CONTEXT_SIZE, 0, indirect);
+
+	/* non-cache window */
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_LMI_VCPU_NC0_64BIT_BAR_LOW),
+			lower_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr), 0, indirect);
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH),
+			upper_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr), 0, indirect);
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_VCPU_NONCACHE_OFFSET0), 0, 0, indirect);
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_VCPU_NONCACHE_SIZE0),
+			AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn4_fw_shared)), 0, indirect);
+
+	/* VCN global tiling registers */
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_GFX8_ADDR_CONFIG), adev->gfx.config.gb_addr_config, 0, indirect);
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_GFX10_ADDR_CONFIG), adev->gfx.config.gb_addr_config, 0, indirect);
+}
+
+/**
+ * vcn_v4_0_3_disable_clock_gating - disable VCN clock gating
+ *
+ * @vinst: VCN instance
+ *
+ * Disable clock gating for VCN block
+ */
+static void vcn_v4_0_3_disable_clock_gating(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	uint32_t data;
+	int vcn_inst;
+
+	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
+		return;
+
+	vcn_inst = GET_INST(VCN, inst_idx);
+
+	/* VCN disable CGC */
+	data = RREG32_SOC15(VCN, vcn_inst, regUVD_CGC_CTRL);
+	data &= ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK;
+	data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
+	data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
+	WREG32_SOC15(VCN, vcn_inst, regUVD_CGC_CTRL, data);
+
+	data = RREG32_SOC15(VCN, vcn_inst, regUVD_CGC_GATE);
+	data &= ~(UVD_CGC_GATE__SYS_MASK
+		| UVD_CGC_GATE__MPEG2_MASK
+		| UVD_CGC_GATE__REGS_MASK
+		| UVD_CGC_GATE__RBC_MASK
+		| UVD_CGC_GATE__LMI_MC_MASK
+		| UVD_CGC_GATE__LMI_UMC_MASK
+		| UVD_CGC_GATE__MPC_MASK
+		| UVD_CGC_GATE__LBSI_MASK
+		| UVD_CGC_GATE__LRBBM_MASK
+		| UVD_CGC_GATE__WCB_MASK
+		| UVD_CGC_GATE__VCPU_MASK
+		| UVD_CGC_GATE__MMSCH_MASK);
+
+	WREG32_SOC15(VCN, vcn_inst, regUVD_CGC_GATE, data);
+	SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_CGC_GATE, 0, 0xFFFFFFFF);
+
+	data = RREG32_SOC15(VCN, vcn_inst, regUVD_CGC_CTRL);
+	data &= ~(UVD_CGC_CTRL__SYS_MODE_MASK
+		| UVD_CGC_CTRL__MPEG2_MODE_MASK
+		| UVD_CGC_CTRL__REGS_MODE_MASK
+		| UVD_CGC_CTRL__RBC_MODE_MASK
+		| UVD_CGC_CTRL__LMI_MC_MODE_MASK
+		| UVD_CGC_CTRL__LMI_UMC_MODE_MASK
+		| UVD_CGC_CTRL__MPC_MODE_MASK
+		| UVD_CGC_CTRL__LBSI_MODE_MASK
+		| UVD_CGC_CTRL__LRBBM_MODE_MASK
+		| UVD_CGC_CTRL__WCB_MODE_MASK
+		| UVD_CGC_CTRL__VCPU_MODE_MASK
+		| UVD_CGC_CTRL__MMSCH_MODE_MASK);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_CGC_CTRL, data);
+
+	data = RREG32_SOC15(VCN, vcn_inst, regUVD_SUVD_CGC_GATE);
+	data |= (UVD_SUVD_CGC_GATE__SRE_MASK
+		| UVD_SUVD_CGC_GATE__SIT_MASK
+		| UVD_SUVD_CGC_GATE__SMP_MASK
+		| UVD_SUVD_CGC_GATE__SCM_MASK
+		| UVD_SUVD_CGC_GATE__SDB_MASK
+		| UVD_SUVD_CGC_GATE__SRE_H264_MASK
+		| UVD_SUVD_CGC_GATE__SRE_HEVC_MASK
+		| UVD_SUVD_CGC_GATE__SIT_H264_MASK
+		| UVD_SUVD_CGC_GATE__SIT_HEVC_MASK
+		| UVD_SUVD_CGC_GATE__SCM_H264_MASK
+		| UVD_SUVD_CGC_GATE__SCM_HEVC_MASK
+		| UVD_SUVD_CGC_GATE__SDB_H264_MASK
+		| UVD_SUVD_CGC_GATE__SDB_HEVC_MASK
+		| UVD_SUVD_CGC_GATE__ENT_MASK
+		| UVD_SUVD_CGC_GATE__SIT_HEVC_DEC_MASK
+		| UVD_SUVD_CGC_GATE__SITE_MASK
+		| UVD_SUVD_CGC_GATE__SRE_VP9_MASK
+		| UVD_SUVD_CGC_GATE__SCM_VP9_MASK
+		| UVD_SUVD_CGC_GATE__SIT_VP9_DEC_MASK
+		| UVD_SUVD_CGC_GATE__SDB_VP9_MASK
+		| UVD_SUVD_CGC_GATE__IME_HEVC_MASK);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_SUVD_CGC_GATE, data);
+
+	data = RREG32_SOC15(VCN, vcn_inst, regUVD_SUVD_CGC_CTRL);
+	data &= ~(UVD_SUVD_CGC_CTRL__SRE_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SIT_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SMP_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SCM_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SDB_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__ENT_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__IME_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SITE_MODE_MASK);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_SUVD_CGC_CTRL, data);
+}
+
+/**
+ * vcn_v4_0_3_disable_clock_gating_dpg_mode - disable VCN clock gating dpg mode
+ *
+ * @vinst: VCN instance
+ * @sram_sel: sram select
+ * @indirect: indirectly write sram
+ *
+ * Disable clock gating for VCN block with dpg mode
+ */
+static void vcn_v4_0_3_disable_clock_gating_dpg_mode(struct amdgpu_vcn_inst *vinst,
+						     uint8_t sram_sel,
+						     uint8_t indirect)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	uint32_t reg_data = 0;
+
+	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
+		return;
+
+	/* enable sw clock gating control */
+	reg_data = 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+	reg_data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
+	reg_data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
+	reg_data &= ~(UVD_CGC_CTRL__SYS_MODE_MASK |
+		 UVD_CGC_CTRL__MPEG2_MODE_MASK |
+		 UVD_CGC_CTRL__REGS_MODE_MASK |
+		 UVD_CGC_CTRL__RBC_MODE_MASK |
+		 UVD_CGC_CTRL__LMI_MC_MODE_MASK |
+		 UVD_CGC_CTRL__LMI_UMC_MODE_MASK |
+		 UVD_CGC_CTRL__IDCT_MODE_MASK |
+		 UVD_CGC_CTRL__MPRD_MODE_MASK |
+		 UVD_CGC_CTRL__MPC_MODE_MASK |
+		 UVD_CGC_CTRL__LBSI_MODE_MASK |
+		 UVD_CGC_CTRL__LRBBM_MODE_MASK |
+		 UVD_CGC_CTRL__WCB_MODE_MASK |
+		 UVD_CGC_CTRL__VCPU_MODE_MASK);
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_CGC_CTRL), reg_data, sram_sel, indirect);
+
+	/* turn off clock gating */
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_CGC_GATE), 0, sram_sel, indirect);
+
+	/* turn on SUVD clock gating */
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_SUVD_CGC_GATE), 1, sram_sel, indirect);
+
+	/* turn on sw mode in UVD_SUVD_CGC_CTRL */
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_SUVD_CGC_CTRL), 0, sram_sel, indirect);
+}
+
+/**
+ * vcn_v4_0_3_enable_clock_gating - enable VCN clock gating
+ *
+ * @vinst: VCN instance
+ *
+ * Enable clock gating for VCN block
+ */
+static void vcn_v4_0_3_enable_clock_gating(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	uint32_t data;
+	int vcn_inst;
+
+	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
+		return;
+
+	vcn_inst = GET_INST(VCN, inst_idx);
+
+	/* enable VCN CGC */
+	data = RREG32_SOC15(VCN, vcn_inst, regUVD_CGC_CTRL);
+	data |= 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+	data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
+	data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
+	WREG32_SOC15(VCN, vcn_inst, regUVD_CGC_CTRL, data);
+
+	data = RREG32_SOC15(VCN, vcn_inst, regUVD_CGC_CTRL);
+	data |= (UVD_CGC_CTRL__SYS_MODE_MASK
+		| UVD_CGC_CTRL__MPEG2_MODE_MASK
+		| UVD_CGC_CTRL__REGS_MODE_MASK
+		| UVD_CGC_CTRL__RBC_MODE_MASK
+		| UVD_CGC_CTRL__LMI_MC_MODE_MASK
+		| UVD_CGC_CTRL__LMI_UMC_MODE_MASK
+		| UVD_CGC_CTRL__MPC_MODE_MASK
+		| UVD_CGC_CTRL__LBSI_MODE_MASK
+		| UVD_CGC_CTRL__LRBBM_MODE_MASK
+		| UVD_CGC_CTRL__WCB_MODE_MASK
+		| UVD_CGC_CTRL__VCPU_MODE_MASK);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_CGC_CTRL, data);
+
+	data = RREG32_SOC15(VCN, vcn_inst, regUVD_SUVD_CGC_CTRL);
+	data |= (UVD_SUVD_CGC_CTRL__SRE_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SIT_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SMP_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SCM_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SDB_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__ENT_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__IME_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SITE_MODE_MASK);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_SUVD_CGC_CTRL, data);
+}
+
+/**
+ * vcn_v4_0_3_start_dpg_mode - VCN start with dpg mode
+ *
+ * @vinst: VCN instance
+ * @indirect: indirectly write sram
+ *
+ * Start VCN block with dpg mode
+ */
+static int vcn_v4_0_3_start_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				     bool indirect)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	struct amdgpu_vcn4_fw_shared *fw_shared =
+						adev->vcn.inst[inst_idx].fw_shared.cpu_addr;
+	struct amdgpu_ring *ring;
+	int vcn_inst, ret;
+	uint32_t tmp;
+
+	vcn_inst = GET_INST(VCN, inst_idx);
+	/* disable register anti-hang mechanism */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_POWER_STATUS), 1,
+		 ~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
+	/* enable dynamic power gating mode */
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_POWER_STATUS);
+	tmp |= UVD_POWER_STATUS__UVD_PG_MODE_MASK;
+	tmp |= UVD_POWER_STATUS__UVD_PG_EN_MASK;
+	WREG32_SOC15(VCN, vcn_inst, regUVD_POWER_STATUS, tmp);
+
+	if (indirect) {
+		DRM_DEV_DEBUG(adev->dev, "VCN %d start: on AID %d",
+			inst_idx, adev->vcn.inst[inst_idx].aid_id);
+		adev->vcn.inst[inst_idx].dpg_sram_curr_addr =
+				(uint32_t *)adev->vcn.inst[inst_idx].dpg_sram_cpu_addr;
+		/* Use dummy register 0xDEADBEEF passing AID selection to PSP FW */
+		WREG32_SOC15_DPG_MODE(inst_idx, 0xDEADBEEF,
+			adev->vcn.inst[inst_idx].aid_id, 0, true);
+	}
+
+	/* enable clock gating */
+	vcn_v4_0_3_disable_clock_gating_dpg_mode(vinst, 0, indirect);
+
+	/* enable VCPU clock */
+	tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
+	tmp |= UVD_VCPU_CNTL__CLK_EN_MASK;
+	tmp |= UVD_VCPU_CNTL__BLK_RST_MASK;
+
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_VCPU_CNTL), tmp, 0, indirect);
+
+	/* disable master interrupt */
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_MASTINT_EN), 0, 0, indirect);
+
+	/* setup regUVD_LMI_CTRL */
+	tmp = (UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
+		UVD_LMI_CTRL__REQ_MODE_MASK |
+		UVD_LMI_CTRL__CRC_RESET_MASK |
+		UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
+		UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
+		UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK |
+		(8 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) |
+		0x00100000L);
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_LMI_CTRL), tmp, 0, indirect);
+
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_MPC_CNTL),
+		0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT, 0, indirect);
+
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_MPC_SET_MUXA0),
+		((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) |
+		 (0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) |
+		 (0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) |
+		 (0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)), 0, indirect);
+
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_MPC_SET_MUXB0),
+		 ((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) |
+		 (0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) |
+		 (0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) |
+		 (0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)), 0, indirect);
+
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_MPC_SET_MUX),
+		((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) |
+		 (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
+		 (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)), 0, indirect);
+
+	vcn_v4_0_3_mc_resume_dpg_mode(vinst, indirect);
+
+	tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
+	tmp |= UVD_VCPU_CNTL__CLK_EN_MASK;
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_VCPU_CNTL), tmp, 0, indirect);
+
+	/* enable LMI MC and UMC channels */
+	tmp = 0x1f << UVD_LMI_CTRL2__RE_OFLD_MIF_WR_REQ_NUM__SHIFT;
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_LMI_CTRL2), tmp, 0, indirect);
+
+	vcn_v4_0_3_enable_ras(adev, inst_idx, indirect);
+
+	/* enable master interrupt */
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_MASTINT_EN),
+		UVD_MASTINT_EN__VCPU_EN_MASK, 0, indirect);
+
+	if (indirect) {
+		ret = amdgpu_vcn_psp_update_sram(adev, inst_idx, AMDGPU_UCODE_ID_VCN0_RAM);
+		if (ret) {
+			dev_err(adev->dev, "vcn sram load failed %d\n", ret);
+			return ret;
+		}
+	}
+
+	ring = &adev->vcn.inst[inst_idx].ring_enc[0];
+
+	/* program the RB_BASE for ring buffer */
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_BASE_LO,
+		     lower_32_bits(ring->gpu_addr));
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_BASE_HI,
+		     upper_32_bits(ring->gpu_addr));
+
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_SIZE,
+		     ring->ring_size / sizeof(uint32_t));
+
+	/* resetting ring, fw should not check RB ring */
+	tmp = RREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE);
+	tmp &= ~(VCN_RB_ENABLE__RB_EN_MASK);
+	WREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE, tmp);
+	fw_shared->sq.queue_mode |= FW_QUEUE_RING_RESET;
+
+	/* Initialize the ring buffer's read and write pointers */
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_RPTR, 0);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_WPTR, 0);
+	ring->wptr = RREG32_SOC15(VCN, vcn_inst, regUVD_RB_WPTR);
+
+	tmp = RREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE);
+	tmp |= VCN_RB_ENABLE__RB_EN_MASK;
+	WREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE, tmp);
+	fw_shared->sq.queue_mode &= ~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF);
+
+	/*resetting done, fw can check RB ring */
+	fw_shared->sq.queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET);
+
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, vcn_inst, regUVD_STATUS);
+
+	return 0;
+}
+
+static int vcn_v4_0_3_start_sriov(struct amdgpu_device *adev)
+{
+	int i, vcn_inst;
+	struct amdgpu_ring *ring_enc;
+	uint64_t cache_addr;
+	uint64_t rb_enc_addr;
+	uint64_t ctx_addr;
+	uint32_t param, resp, expected;
+	uint32_t offset, cache_size;
+	uint32_t tmp, timeout;
+
+	struct amdgpu_mm_table *table = &adev->virt.mm_table;
+	uint32_t *table_loc;
+	uint32_t table_size;
+	uint32_t size, size_dw;
+	uint32_t init_status;
+	uint32_t enabled_vcn;
+
+	struct mmsch_v4_0_cmd_direct_write
+		direct_wt = { {0} };
+	struct mmsch_v4_0_cmd_direct_read_modify_write
+		direct_rd_mod_wt = { {0} };
+	struct mmsch_v4_0_cmd_end end = { {0} };
+	struct mmsch_v4_0_3_init_header header;
+
+	struct amdgpu_vcn4_fw_shared *fw_shared;
+	struct amdgpu_fw_shared_rb_setup *rb_setup;
+
+	direct_wt.cmd_header.command_type =
+		MMSCH_COMMAND__DIRECT_REG_WRITE;
+	direct_rd_mod_wt.cmd_header.command_type =
+		MMSCH_COMMAND__DIRECT_REG_READ_MODIFY_WRITE;
+	end.cmd_header.command_type = MMSCH_COMMAND__END;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		vcn_inst = GET_INST(VCN, i);
+
+		vcn_v4_0_3_fw_shared_init(adev, vcn_inst);
+
+		memset(&header, 0, sizeof(struct mmsch_v4_0_3_init_header));
+		header.version = MMSCH_VERSION;
+		header.total_size = sizeof(struct mmsch_v4_0_3_init_header) >> 2;
+
+		table_loc = (uint32_t *)table->cpu_addr;
+		table_loc += header.total_size;
+
+		table_size = 0;
+
+		MMSCH_V4_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCN, 0, regUVD_STATUS),
+			~UVD_STATUS__UVD_BUSY, UVD_STATUS__UVD_BUSY);
+
+		cache_size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.inst[i].fw->size + 4);
+
+		if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+			MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+				regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW),
+				adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + i].tmr_mc_addr_lo);
+
+			MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+				regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH),
+				adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + i].tmr_mc_addr_hi);
+
+			offset = 0;
+			MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+				regUVD_VCPU_CACHE_OFFSET0), 0);
+		} else {
+			MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+				regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW),
+				lower_32_bits(adev->vcn.inst[i].gpu_addr));
+			MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+				regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH),
+				upper_32_bits(adev->vcn.inst[i].gpu_addr));
+			offset = cache_size;
+			MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+				regUVD_VCPU_CACHE_OFFSET0),
+				AMDGPU_UVD_FIRMWARE_OFFSET >> 3);
+		}
+
+		MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_VCPU_CACHE_SIZE0),
+			cache_size);
+
+		cache_addr = adev->vcn.inst[vcn_inst].gpu_addr + offset;
+		MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW), lower_32_bits(cache_addr));
+		MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH), upper_32_bits(cache_addr));
+		MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_VCPU_CACHE_OFFSET1), 0);
+		MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_VCPU_CACHE_SIZE1), AMDGPU_VCN_STACK_SIZE);
+
+		cache_addr = adev->vcn.inst[vcn_inst].gpu_addr + offset +
+			AMDGPU_VCN_STACK_SIZE;
+
+		MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW), lower_32_bits(cache_addr));
+
+		MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH), upper_32_bits(cache_addr));
+
+		MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_VCPU_CACHE_OFFSET2), 0);
+
+		MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_VCPU_CACHE_SIZE2), AMDGPU_VCN_CONTEXT_SIZE);
+
+		fw_shared = adev->vcn.inst[vcn_inst].fw_shared.cpu_addr;
+		rb_setup = &fw_shared->rb_setup;
+
+		ring_enc = &adev->vcn.inst[vcn_inst].ring_enc[0];
+		ring_enc->wptr = 0;
+		rb_enc_addr = ring_enc->gpu_addr;
+
+		rb_setup->is_rb_enabled_flags |= RB_ENABLED;
+		rb_setup->rb_addr_lo = lower_32_bits(rb_enc_addr);
+		rb_setup->rb_addr_hi = upper_32_bits(rb_enc_addr);
+		rb_setup->rb_size = ring_enc->ring_size / 4;
+		fw_shared->present_flag_0 |= cpu_to_le32(AMDGPU_VCN_VF_RB_SETUP_FLAG);
+
+		MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_LMI_VCPU_NC0_64BIT_BAR_LOW),
+			lower_32_bits(adev->vcn.inst[vcn_inst].fw_shared.gpu_addr));
+		MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH),
+			upper_32_bits(adev->vcn.inst[vcn_inst].fw_shared.gpu_addr));
+		MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_VCPU_NONCACHE_SIZE0),
+			AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn4_fw_shared)));
+		MMSCH_V4_0_INSERT_END();
+
+		header.vcn0.init_status = 0;
+		header.vcn0.table_offset = header.total_size;
+		header.vcn0.table_size = table_size;
+		header.total_size += table_size;
+
+		/* Send init table to mmsch */
+		size = sizeof(struct mmsch_v4_0_3_init_header);
+		table_loc = (uint32_t *)table->cpu_addr;
+		memcpy((void *)table_loc, &header, size);
+
+		ctx_addr = table->gpu_addr;
+		WREG32_SOC15(VCN, vcn_inst, regMMSCH_VF_CTX_ADDR_LO, lower_32_bits(ctx_addr));
+		WREG32_SOC15(VCN, vcn_inst, regMMSCH_VF_CTX_ADDR_HI, upper_32_bits(ctx_addr));
+
+		tmp = RREG32_SOC15(VCN, vcn_inst, regMMSCH_VF_VMID);
+		tmp &= ~MMSCH_VF_VMID__VF_CTX_VMID_MASK;
+		tmp |= (0 << MMSCH_VF_VMID__VF_CTX_VMID__SHIFT);
+		WREG32_SOC15(VCN, vcn_inst, regMMSCH_VF_VMID, tmp);
+
+		size = header.total_size;
+		WREG32_SOC15(VCN, vcn_inst, regMMSCH_VF_CTX_SIZE, size);
+
+		WREG32_SOC15(VCN, vcn_inst, regMMSCH_VF_MAILBOX_RESP, 0);
+
+		param = 0x00000001;
+		WREG32_SOC15(VCN, vcn_inst, regMMSCH_VF_MAILBOX_HOST, param);
+		tmp = 0;
+		timeout = 1000;
+		resp = 0;
+		expected = MMSCH_VF_MAILBOX_RESP__OK;
+		while (resp != expected) {
+			resp = RREG32_SOC15(VCN, vcn_inst, regMMSCH_VF_MAILBOX_RESP);
+			if (resp != 0)
+				break;
+
+			udelay(10);
+			tmp = tmp + 10;
+			if (tmp >= timeout) {
+				DRM_ERROR("failed to init MMSCH. TIME-OUT after %d usec"\
+					" waiting for regMMSCH_VF_MAILBOX_RESP "\
+					"(expected=0x%08x, readback=0x%08x)\n",
+					tmp, expected, resp);
+				return -EBUSY;
+			}
+		}
+
+		enabled_vcn = amdgpu_vcn_is_disabled_vcn(adev, VCN_DECODE_RING, 0) ? 1 : 0;
+		init_status = ((struct mmsch_v4_0_3_init_header *)(table_loc))->vcn0.init_status;
+		if (resp != expected && resp != MMSCH_VF_MAILBOX_RESP__INCOMPLETE
+					&& init_status != MMSCH_VF_ENGINE_STATUS__PASS) {
+			DRM_ERROR("MMSCH init status is incorrect! readback=0x%08x, header init "\
+				"status for VCN%x: 0x%x\n", resp, enabled_vcn, init_status);
+		}
+	}
+
+	return 0;
+}
+
+/**
+ * vcn_v4_0_3_start - VCN start
+ *
+ * @vinst: VCN instance
+ *
+ * Start VCN block
+ */
+static int vcn_v4_0_3_start(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int i = vinst->inst;
+	struct amdgpu_vcn4_fw_shared *fw_shared;
+	struct amdgpu_ring *ring;
+	int j, k, r, vcn_inst;
+	uint32_t tmp;
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
+		return vcn_v4_0_3_start_dpg_mode(vinst, adev->vcn.inst[i].indirect_sram);
+
+	vcn_inst = GET_INST(VCN, i);
+	/* set VCN status busy */
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_STATUS) |
+		UVD_STATUS__UVD_BUSY;
+	WREG32_SOC15(VCN, vcn_inst, regUVD_STATUS, tmp);
+
+	/* SW clock gating */
+	vcn_v4_0_3_disable_clock_gating(vinst);
+
+	/* enable VCPU clock */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_VCPU_CNTL),
+		 UVD_VCPU_CNTL__CLK_EN_MASK,
+		 ~UVD_VCPU_CNTL__CLK_EN_MASK);
+
+	/* disable master interrupt */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_MASTINT_EN), 0,
+		 ~UVD_MASTINT_EN__VCPU_EN_MASK);
+
+	/* enable LMI MC and UMC channels */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_LMI_CTRL2), 0,
+		 ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
+
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_SOFT_RESET);
+	tmp &= ~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
+	tmp &= ~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
+	WREG32_SOC15(VCN, vcn_inst, regUVD_SOFT_RESET, tmp);
+
+	/* setup regUVD_LMI_CTRL */
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_LMI_CTRL);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_CTRL,
+		     tmp | UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
+		     UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
+		     UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
+		     UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK);
+
+	/* setup regUVD_MPC_CNTL */
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_MPC_CNTL);
+	tmp &= ~UVD_MPC_CNTL__REPLACEMENT_MODE_MASK;
+	tmp |= 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT;
+	WREG32_SOC15(VCN, vcn_inst, regUVD_MPC_CNTL, tmp);
+
+	/* setup UVD_MPC_SET_MUXA0 */
+	WREG32_SOC15(VCN, vcn_inst, regUVD_MPC_SET_MUXA0,
+		     ((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) |
+		      (0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) |
+		      (0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) |
+		      (0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)));
+
+	/* setup UVD_MPC_SET_MUXB0 */
+	WREG32_SOC15(VCN, vcn_inst, regUVD_MPC_SET_MUXB0,
+		     ((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) |
+		      (0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) |
+		      (0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) |
+		      (0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)));
+
+	/* setup UVD_MPC_SET_MUX */
+	WREG32_SOC15(VCN, vcn_inst, regUVD_MPC_SET_MUX,
+		     ((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) |
+		      (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
+		      (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)));
+
+	vcn_v4_0_3_mc_resume(vinst);
+
+	/* VCN global tiling registers */
+	WREG32_SOC15(VCN, vcn_inst, regUVD_GFX8_ADDR_CONFIG,
+		     adev->gfx.config.gb_addr_config);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_GFX10_ADDR_CONFIG,
+		     adev->gfx.config.gb_addr_config);
+
+	/* unblock VCPU register access */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_RB_ARB_CTRL), 0,
+		 ~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
+
+	/* release VCPU reset to boot */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_VCPU_CNTL), 0,
+		 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+
+	for (j = 0; j < 10; ++j) {
+		uint32_t status;
+
+		for (k = 0; k < 100; ++k) {
+			status = RREG32_SOC15(VCN, vcn_inst,
+					      regUVD_STATUS);
+			if (status & 2)
+				break;
+			mdelay(10);
+		}
+		r = 0;
+		if (status & 2)
+			break;
+
+		DRM_DEV_ERROR(adev->dev,
+			      "VCN decode not responding, trying to reset the VCPU!!!\n");
+		WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst,
+					  regUVD_VCPU_CNTL),
+			 UVD_VCPU_CNTL__BLK_RST_MASK,
+			 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+		mdelay(10);
+		WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst,
+					  regUVD_VCPU_CNTL),
+			 0, ~UVD_VCPU_CNTL__BLK_RST_MASK);
+
+		mdelay(10);
+		r = -1;
+	}
+
+	if (r) {
+		DRM_DEV_ERROR(adev->dev, "VCN decode not responding, giving up!!!\n");
+		return r;
+	}
+
+	/* enable master interrupt */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_MASTINT_EN),
+		 UVD_MASTINT_EN__VCPU_EN_MASK,
+		 ~UVD_MASTINT_EN__VCPU_EN_MASK);
+
+	/* clear the busy bit of VCN_STATUS */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_STATUS), 0,
+		 ~(2 << UVD_STATUS__VCPU_REPORT__SHIFT));
+
+	ring = &adev->vcn.inst[i].ring_enc[0];
+	fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
+
+	/* program the RB_BASE for ring buffer */
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_BASE_LO,
+		     lower_32_bits(ring->gpu_addr));
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_BASE_HI,
+		     upper_32_bits(ring->gpu_addr));
+
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_SIZE,
+		     ring->ring_size / sizeof(uint32_t));
+
+	/* resetting ring, fw should not check RB ring */
+	tmp = RREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE);
+	tmp &= ~(VCN_RB_ENABLE__RB_EN_MASK);
+	WREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE, tmp);
+
+	/* Initialize the ring buffer's read and write pointers */
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_RPTR, 0);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_WPTR, 0);
+
+	tmp = RREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE);
+	tmp |= VCN_RB_ENABLE__RB_EN_MASK;
+	WREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE, tmp);
+
+	ring->wptr = RREG32_SOC15(VCN, vcn_inst, regUVD_RB_WPTR);
+	fw_shared->sq.queue_mode &=
+		cpu_to_le32(~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF));
+
+	return 0;
+}
+
+/**
+ * vcn_v4_0_3_stop_dpg_mode - VCN stop with dpg mode
+ *
+ * @vinst: VCN instance
+ *
+ * Stop VCN block with dpg mode
+ */
+static int vcn_v4_0_3_stop_dpg_mode(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	uint32_t tmp;
+	int vcn_inst;
+
+	vcn_inst = GET_INST(VCN, inst_idx);
+
+	/* Wait for power status to be 1 */
+	SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_POWER_STATUS, 1,
+			   UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
+
+	/* wait for read ptr to be equal to write ptr */
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_RB_WPTR);
+	SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_RB_RPTR, tmp, 0xFFFFFFFF);
+
+	SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_POWER_STATUS, 1,
+			   UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
+
+	/* disable dynamic power gating mode */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_POWER_STATUS), 0,
+		 ~UVD_POWER_STATUS__UVD_PG_MODE_MASK);
+
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, vcn_inst, regUVD_STATUS);
+
+	return 0;
+}
+
+/**
+ * vcn_v4_0_3_stop - VCN stop
+ *
+ * @vinst: VCN instance
+ *
+ * Stop VCN block
+ */
+static int vcn_v4_0_3_stop(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int i = vinst->inst;
+	struct amdgpu_vcn4_fw_shared *fw_shared;
+	int r = 0, vcn_inst;
+	uint32_t tmp;
+
+	vcn_inst = GET_INST(VCN, i);
+
+	fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
+	fw_shared->sq.queue_mode |= FW_QUEUE_DPG_HOLD_OFF;
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
+		vcn_v4_0_3_stop_dpg_mode(vinst);
+		goto Done;
+	}
+
+	/* wait for vcn idle */
+	r = SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_STATUS,
+			       UVD_STATUS__IDLE, 0x7);
+	if (r)
+		goto Done;
+
+	tmp = UVD_LMI_STATUS__VCPU_LMI_WRITE_CLEAN_MASK |
+		UVD_LMI_STATUS__READ_CLEAN_MASK |
+		UVD_LMI_STATUS__WRITE_CLEAN_MASK |
+		UVD_LMI_STATUS__WRITE_CLEAN_RAW_MASK;
+	r = SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_LMI_STATUS, tmp,
+			       tmp);
+	if (r)
+		goto Done;
+
+	/* stall UMC channel */
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_LMI_CTRL2);
+	tmp |= UVD_LMI_CTRL2__STALL_ARB_UMC_MASK;
+	WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_CTRL2, tmp);
+	tmp = UVD_LMI_STATUS__UMC_READ_CLEAN_RAW_MASK |
+		UVD_LMI_STATUS__UMC_WRITE_CLEAN_RAW_MASK;
+	r = SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_LMI_STATUS, tmp,
+			       tmp);
+	if (r)
+		goto Done;
+
+	/* Unblock VCPU Register access */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_RB_ARB_CTRL),
+		 UVD_RB_ARB_CTRL__VCPU_DIS_MASK,
+		 ~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
+
+	/* release VCPU reset to boot */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_VCPU_CNTL),
+		 UVD_VCPU_CNTL__BLK_RST_MASK,
+		 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+
+	/* disable VCPU clock */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_VCPU_CNTL), 0,
+		 ~(UVD_VCPU_CNTL__CLK_EN_MASK));
+
+	/* reset LMI UMC/LMI/VCPU */
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_SOFT_RESET);
+	tmp |= UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
+	WREG32_SOC15(VCN, vcn_inst, regUVD_SOFT_RESET, tmp);
+
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_SOFT_RESET);
+	tmp |= UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
+	WREG32_SOC15(VCN, vcn_inst, regUVD_SOFT_RESET, tmp);
+
+	/* clear VCN status */
+	WREG32_SOC15(VCN, vcn_inst, regUVD_STATUS, 0);
+
+	/* apply HW clock gating */
+	vcn_v4_0_3_enable_clock_gating(vinst);
+
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, vcn_inst, regUVD_STATUS);
+
+Done:
+	return 0;
+}
+
+/**
+ * vcn_v4_0_3_pause_dpg_mode - VCN pause with dpg mode
+ *
+ * @vinst: VCN instance
+ * @new_state: pause state
+ *
+ * Pause dpg mode for VCN block
+ */
+static int vcn_v4_0_3_pause_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				     struct dpg_pause_state *new_state)
+{
+
+	return 0;
+}
+
+/**
+ * vcn_v4_0_3_unified_ring_get_rptr - get unified read pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware unified read pointer
+ */
+static uint64_t vcn_v4_0_3_unified_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring != &adev->vcn.inst[ring->me].ring_enc[0])
+		DRM_ERROR("wrong ring id is identified in %s", __func__);
+
+	return RREG32_SOC15(VCN, GET_INST(VCN, ring->me), regUVD_RB_RPTR);
+}
+
+/**
+ * vcn_v4_0_3_unified_ring_get_wptr - get unified write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware unified write pointer
+ */
+static uint64_t vcn_v4_0_3_unified_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring != &adev->vcn.inst[ring->me].ring_enc[0])
+		DRM_ERROR("wrong ring id is identified in %s", __func__);
+
+	if (ring->use_doorbell)
+		return *ring->wptr_cpu_addr;
+	else
+		return RREG32_SOC15(VCN, GET_INST(VCN, ring->me),
+				    regUVD_RB_WPTR);
+}
+
+void vcn_v4_0_3_enc_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
+				       uint32_t val, uint32_t mask)
+{
+	/* Use normalized offsets when required */
+	if (vcn_v4_0_3_normalizn_reqd(ring->adev))
+		reg = NORMALIZE_VCN_REG_OFFSET(reg);
+
+	amdgpu_ring_write(ring, VCN_ENC_CMD_REG_WAIT);
+	amdgpu_ring_write(ring, reg << 2);
+	amdgpu_ring_write(ring, mask);
+	amdgpu_ring_write(ring, val);
+}
+
+void vcn_v4_0_3_enc_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg,
+				   uint32_t val)
+{
+	/* Use normalized offsets when required */
+	if (vcn_v4_0_3_normalizn_reqd(ring->adev))
+		reg = NORMALIZE_VCN_REG_OFFSET(reg);
+
+	amdgpu_ring_write(ring, VCN_ENC_CMD_REG_WRITE);
+	amdgpu_ring_write(ring,	reg << 2);
+	amdgpu_ring_write(ring, val);
+}
+
+void vcn_v4_0_3_enc_ring_emit_vm_flush(struct amdgpu_ring *ring,
+				       unsigned int vmid, uint64_t pd_addr)
+{
+	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
+
+	pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
+
+	/* wait for reg writes */
+	vcn_v4_0_3_enc_ring_emit_reg_wait(ring, hub->ctx0_ptb_addr_lo32 +
+					vmid * hub->ctx_addr_distance,
+					lower_32_bits(pd_addr), 0xffffffff);
+}
+
+void vcn_v4_0_3_ring_emit_hdp_flush(struct amdgpu_ring *ring)
+{
+	/* VCN engine access for HDP flush doesn't work when RRMT is enabled.
+	 * This is a workaround to avoid any HDP flush through VCN ring.
+	 */
+}
+
+/**
+ * vcn_v4_0_3_unified_ring_set_wptr - set enc write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Commits the enc write pointer to the hardware
+ */
+static void vcn_v4_0_3_unified_ring_set_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring != &adev->vcn.inst[ring->me].ring_enc[0])
+		DRM_ERROR("wrong ring id is identified in %s", __func__);
+
+	if (ring->use_doorbell) {
+		*ring->wptr_cpu_addr = lower_32_bits(ring->wptr);
+		WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr));
+	} else {
+		WREG32_SOC15(VCN, GET_INST(VCN, ring->me), regUVD_RB_WPTR,
+			     lower_32_bits(ring->wptr));
+	}
+}
+
+static int vcn_v4_0_3_ring_reset(struct amdgpu_ring *ring,
+				 unsigned int vmid,
+				 struct amdgpu_fence *timedout_fence)
+{
+	int r = 0;
+	int vcn_inst;
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[ring->me];
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	vcn_inst = GET_INST(VCN, ring->me);
+	r = amdgpu_dpm_reset_vcn(adev, 1 << vcn_inst);
+
+	if (r) {
+		DRM_DEV_ERROR(adev->dev, "VCN reset fail : %d\n", r);
+		return r;
+	}
+
+	/* This flag is not set for VF, assumed to be disabled always */
+	if (RREG32_SOC15(VCN, GET_INST(VCN, 0), regVCN_RRMT_CNTL) & 0x100)
+		adev->vcn.caps |= AMDGPU_VCN_CAPS(RRMT_ENABLED);
+	vcn_v4_0_3_hw_init_inst(vinst);
+	vcn_v4_0_3_start_dpg_mode(vinst, adev->vcn.inst[ring->me].indirect_sram);
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static const struct amdgpu_ring_funcs vcn_v4_0_3_unified_ring_vm_funcs = {
+	.type = AMDGPU_RING_TYPE_VCN_ENC,
+	.align_mask = 0x3f,
+	.nop = VCN_ENC_CMD_NO_OP,
+	.get_rptr = vcn_v4_0_3_unified_ring_get_rptr,
+	.get_wptr = vcn_v4_0_3_unified_ring_get_wptr,
+	.set_wptr = vcn_v4_0_3_unified_ring_set_wptr,
+	.emit_frame_size =
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 4 +
+		4 + /* vcn_v2_0_enc_ring_emit_vm_flush */
+		5 + 5 + /* vcn_v2_0_enc_ring_emit_fence x2 vm fence */
+		1, /* vcn_v2_0_enc_ring_insert_end */
+	.emit_ib_size = 5, /* vcn_v2_0_enc_ring_emit_ib */
+	.emit_ib = vcn_v2_0_enc_ring_emit_ib,
+	.emit_fence = vcn_v2_0_enc_ring_emit_fence,
+	.emit_vm_flush = vcn_v4_0_3_enc_ring_emit_vm_flush,
+	.emit_hdp_flush = vcn_v4_0_3_ring_emit_hdp_flush,
+	.test_ring = amdgpu_vcn_enc_ring_test_ring,
+	.test_ib = amdgpu_vcn_unified_ring_test_ib,
+	.insert_nop = amdgpu_ring_insert_nop,
+	.insert_end = vcn_v2_0_enc_ring_insert_end,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.begin_use = amdgpu_vcn_ring_begin_use,
+	.end_use = amdgpu_vcn_ring_end_use,
+	.emit_wreg = vcn_v4_0_3_enc_ring_emit_wreg,
+	.emit_reg_wait = vcn_v4_0_3_enc_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = vcn_v4_0_3_ring_reset,
+};
+
+/**
+ * vcn_v4_0_3_set_unified_ring_funcs - set unified ring functions
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set unified ring functions
+ */
+static void vcn_v4_0_3_set_unified_ring_funcs(struct amdgpu_device *adev)
+{
+	int i, vcn_inst;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		adev->vcn.inst[i].ring_enc[0].funcs = &vcn_v4_0_3_unified_ring_vm_funcs;
+		adev->vcn.inst[i].ring_enc[0].me = i;
+		vcn_inst = GET_INST(VCN, i);
+		adev->vcn.inst[i].aid_id =
+			vcn_inst / adev->vcn.num_inst_per_aid;
+	}
+}
+
+/**
+ * vcn_v4_0_3_is_idle - check VCN block is idle
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block structure
+ *
+ * Check whether VCN block is idle
+ */
+static bool vcn_v4_0_3_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, ret = 1;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		ret &= (RREG32_SOC15(VCN, GET_INST(VCN, i), regUVD_STATUS) ==
+			UVD_STATUS__IDLE);
+	}
+
+	return ret;
+}
+
+/**
+ * vcn_v4_0_3_wait_for_idle - wait for VCN block idle
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Wait for VCN block idle
+ */
+static int vcn_v4_0_3_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, ret = 0;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		ret = SOC15_WAIT_ON_RREG(VCN, GET_INST(VCN, i), regUVD_STATUS,
+					 UVD_STATUS__IDLE, UVD_STATUS__IDLE);
+		if (ret)
+			return ret;
+	}
+
+	return ret;
+}
+
+/* vcn_v4_0_3_set_clockgating_state - set VCN block clockgating state
+ *
+ * @ip_block: amdgpu_ip_block pointer
+ * @state: clock gating state
+ *
+ * Set VCN block clockgating state
+ */
+static int vcn_v4_0_3_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_clockgating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	bool enable = state == AMD_CG_STATE_GATE;
+	int i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[i];
+
+		if (enable) {
+			if (RREG32_SOC15(VCN, GET_INST(VCN, i),
+					 regUVD_STATUS) != UVD_STATUS__IDLE)
+				return -EBUSY;
+			vcn_v4_0_3_enable_clock_gating(vinst);
+		} else {
+			vcn_v4_0_3_disable_clock_gating(vinst);
+		}
+	}
+	return 0;
+}
+
+static int vcn_v4_0_3_set_pg_state(struct amdgpu_vcn_inst *vinst,
+				   enum amd_powergating_state state)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int ret = 0;
+
+	/* for SRIOV, guest should not control VCN Power-gating
+	 * MMSCH FW should control Power-gating and clock-gating
+	 * guest should avoid touching CGC and PG
+	 */
+	if (amdgpu_sriov_vf(adev)) {
+		vinst->cur_state = AMD_PG_STATE_UNGATE;
+		return 0;
+	}
+
+	if (state == vinst->cur_state)
+		return 0;
+
+	if (state == AMD_PG_STATE_GATE)
+		ret = vcn_v4_0_3_stop(vinst);
+	else
+		ret = vcn_v4_0_3_start(vinst);
+
+	if (!ret)
+		vinst->cur_state = state;
+
+	return ret;
+}
+
+/**
+ * vcn_v4_0_3_set_interrupt_state - set VCN block interrupt state
+ *
+ * @adev: amdgpu_device pointer
+ * @source: interrupt sources
+ * @type: interrupt types
+ * @state: interrupt states
+ *
+ * Set VCN block interrupt state
+ */
+static int vcn_v4_0_3_set_interrupt_state(struct amdgpu_device *adev,
+					struct amdgpu_irq_src *source,
+					unsigned int type,
+					enum amdgpu_interrupt_state state)
+{
+	return 0;
+}
+
+/**
+ * vcn_v4_0_3_process_interrupt - process VCN block interrupt
+ *
+ * @adev: amdgpu_device pointer
+ * @source: interrupt sources
+ * @entry: interrupt entry from clients and sources
+ *
+ * Process VCN block interrupt
+ */
+static int vcn_v4_0_3_process_interrupt(struct amdgpu_device *adev,
+				      struct amdgpu_irq_src *source,
+				      struct amdgpu_iv_entry *entry)
+{
+	uint32_t i, inst;
+
+	i = node_id_to_phys_map[entry->node_id];
+
+	DRM_DEV_DEBUG(adev->dev, "IH: VCN TRAP\n");
+
+	for (inst = 0; inst < adev->vcn.num_vcn_inst; ++inst)
+		if (adev->vcn.inst[inst].aid_id == i)
+			break;
+
+	if (inst >= adev->vcn.num_vcn_inst) {
+		dev_WARN_ONCE(adev->dev, 1,
+			      "Interrupt received for unknown VCN instance %d",
+			      entry->node_id);
+		return 0;
+	}
+
+	switch (entry->src_id) {
+	case VCN_4_0__SRCID__UVD_ENC_GENERAL_PURPOSE:
+		amdgpu_fence_process(&adev->vcn.inst[inst].ring_enc[0]);
+		break;
+	default:
+		DRM_DEV_ERROR(adev->dev, "Unhandled interrupt: %d %d\n",
+			  entry->src_id, entry->src_data[0]);
+		break;
+	}
+
+	return 0;
+}
+
+static int vcn_v4_0_3_set_ras_interrupt_state(struct amdgpu_device *adev,
+					struct amdgpu_irq_src *source,
+					unsigned int type,
+					enum amdgpu_interrupt_state state)
+{
+	return 0;
+}
+
+static const struct amdgpu_irq_src_funcs vcn_v4_0_3_irq_funcs = {
+	.set = vcn_v4_0_3_set_interrupt_state,
+	.process = vcn_v4_0_3_process_interrupt,
+};
+
+static const struct amdgpu_irq_src_funcs vcn_v4_0_3_ras_irq_funcs = {
+	.set = vcn_v4_0_3_set_ras_interrupt_state,
+	.process = amdgpu_vcn_process_poison_irq,
+};
+
+/**
+ * vcn_v4_0_3_set_irq_funcs - set VCN block interrupt irq functions
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set VCN block interrupt irq functions
+ */
+static void vcn_v4_0_3_set_irq_funcs(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		adev->vcn.inst->irq.num_types++;
+	}
+	adev->vcn.inst->irq.funcs = &vcn_v4_0_3_irq_funcs;
+
+	adev->vcn.inst->ras_poison_irq.num_types = 1;
+	adev->vcn.inst->ras_poison_irq.funcs = &vcn_v4_0_3_ras_irq_funcs;
+}
+
+static const struct amd_ip_funcs vcn_v4_0_3_ip_funcs = {
+	.name = "vcn_v4_0_3",
+	.early_init = vcn_v4_0_3_early_init,
+	.late_init = vcn_v4_0_3_late_init,
+	.sw_init = vcn_v4_0_3_sw_init,
+	.sw_fini = vcn_v4_0_3_sw_fini,
+	.hw_init = vcn_v4_0_3_hw_init,
+	.hw_fini = vcn_v4_0_3_hw_fini,
+	.suspend = vcn_v4_0_3_suspend,
+	.resume = vcn_v4_0_3_resume,
+	.is_idle = vcn_v4_0_3_is_idle,
+	.wait_for_idle = vcn_v4_0_3_wait_for_idle,
+	.set_clockgating_state = vcn_v4_0_3_set_clockgating_state,
+	.set_powergating_state = vcn_set_powergating_state,
+	.dump_ip_state = amdgpu_vcn_dump_ip_state,
+	.print_ip_state = amdgpu_vcn_print_ip_state,
+};
+
+const struct amdgpu_ip_block_version vcn_v4_0_3_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_VCN,
+	.major = 4,
+	.minor = 0,
+	.rev = 3,
+	.funcs = &vcn_v4_0_3_ip_funcs,
+};
+
+static const struct amdgpu_ras_err_status_reg_entry vcn_v4_0_3_ue_reg_list[] = {
+	{AMDGPU_RAS_REG_ENTRY(VCN, 0, regVCN_UE_ERR_STATUS_LO_VIDD, regVCN_UE_ERR_STATUS_HI_VIDD),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "VIDD"},
+	{AMDGPU_RAS_REG_ENTRY(VCN, 0, regVCN_UE_ERR_STATUS_LO_VIDV, regVCN_UE_ERR_STATUS_HI_VIDV),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "VIDV"},
+};
+
+static void vcn_v4_0_3_inst_query_ras_error_count(struct amdgpu_device *adev,
+						  uint32_t vcn_inst,
+						  void *ras_err_status)
+{
+	struct ras_err_data *err_data = (struct ras_err_data *)ras_err_status;
+
+	/* vcn v4_0_3 only support query uncorrectable errors */
+	amdgpu_ras_inst_query_ras_error_count(adev,
+			vcn_v4_0_3_ue_reg_list,
+			ARRAY_SIZE(vcn_v4_0_3_ue_reg_list),
+			NULL, 0, GET_INST(VCN, vcn_inst),
+			AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE,
+			&err_data->ue_count);
+}
+
+static void vcn_v4_0_3_query_ras_error_count(struct amdgpu_device *adev,
+					     void *ras_err_status)
+{
+	uint32_t i;
+
+	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__VCN)) {
+		dev_warn(adev->dev, "VCN RAS is not supported\n");
+		return;
+	}
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++)
+		vcn_v4_0_3_inst_query_ras_error_count(adev, i, ras_err_status);
+}
+
+static void vcn_v4_0_3_inst_reset_ras_error_count(struct amdgpu_device *adev,
+						  uint32_t vcn_inst)
+{
+	amdgpu_ras_inst_reset_ras_error_count(adev,
+					vcn_v4_0_3_ue_reg_list,
+					ARRAY_SIZE(vcn_v4_0_3_ue_reg_list),
+					GET_INST(VCN, vcn_inst));
+}
+
+static void vcn_v4_0_3_reset_ras_error_count(struct amdgpu_device *adev)
+{
+	uint32_t i;
+
+	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__VCN)) {
+		dev_warn(adev->dev, "VCN RAS is not supported\n");
+		return;
+	}
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++)
+		vcn_v4_0_3_inst_reset_ras_error_count(adev, i);
+}
+
+static uint32_t vcn_v4_0_3_query_poison_by_instance(struct amdgpu_device *adev,
+			uint32_t instance, uint32_t sub_block)
+{
+	uint32_t poison_stat = 0, reg_value = 0;
+
+	switch (sub_block) {
+	case AMDGPU_VCN_V4_0_3_VCPU_VCODEC:
+		reg_value = RREG32_SOC15(VCN, instance, regUVD_RAS_VCPU_VCODEC_STATUS);
+		poison_stat = REG_GET_FIELD(reg_value, UVD_RAS_VCPU_VCODEC_STATUS, POISONED_PF);
+		break;
+	default:
+		break;
+	}
+
+	if (poison_stat)
+		dev_info(adev->dev, "Poison detected in VCN%d, sub_block%d\n",
+			instance, sub_block);
+
+	return poison_stat;
+}
+
+static bool vcn_v4_0_3_query_poison_status(struct amdgpu_device *adev)
+{
+	uint32_t inst, sub;
+	uint32_t poison_stat = 0;
+
+	for (inst = 0; inst < adev->vcn.num_vcn_inst; inst++)
+		for (sub = 0; sub < AMDGPU_VCN_V4_0_3_MAX_SUB_BLOCK; sub++)
+			poison_stat +=
+			vcn_v4_0_3_query_poison_by_instance(adev, inst, sub);
+
+	return !!poison_stat;
+}
+
+static const struct amdgpu_ras_block_hw_ops vcn_v4_0_3_ras_hw_ops = {
+	.query_ras_error_count = vcn_v4_0_3_query_ras_error_count,
+	.reset_ras_error_count = vcn_v4_0_3_reset_ras_error_count,
+	.query_poison_status = vcn_v4_0_3_query_poison_status,
+};
+
+static int vcn_v4_0_3_aca_bank_parser(struct aca_handle *handle, struct aca_bank *bank,
+				      enum aca_smu_type type, void *data)
+{
+	struct aca_bank_info info;
+	u64 misc0;
+	int ret;
+
+	ret = aca_bank_info_decode(bank, &info);
+	if (ret)
+		return ret;
+
+	misc0 = bank->regs[ACA_REG_IDX_MISC0];
+	switch (type) {
+	case ACA_SMU_TYPE_UE:
+		bank->aca_err_type = ACA_ERROR_TYPE_UE;
+		ret = aca_error_cache_log_bank_error(handle, &info, ACA_ERROR_TYPE_UE,
+						     1ULL);
+		break;
+	case ACA_SMU_TYPE_CE:
+		bank->aca_err_type = ACA_ERROR_TYPE_CE;
+		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type,
+						     ACA_REG__MISC0__ERRCNT(misc0));
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return ret;
+}
+
+/* reference to smu driver if header file */
+static int vcn_v4_0_3_err_codes[] = {
+	14, 15, /* VCN */
+};
+
+static bool vcn_v4_0_3_aca_bank_is_valid(struct aca_handle *handle, struct aca_bank *bank,
+					 enum aca_smu_type type, void *data)
+{
+	u32 instlo;
+
+	instlo = ACA_REG__IPID__INSTANCEIDLO(bank->regs[ACA_REG_IDX_IPID]);
+	instlo &= GENMASK(31, 1);
+
+	if (instlo != mmSMNAID_AID0_MCA_SMU)
+		return false;
+
+	if (aca_bank_check_error_codes(handle->adev, bank,
+				       vcn_v4_0_3_err_codes,
+				       ARRAY_SIZE(vcn_v4_0_3_err_codes)))
+		return false;
+
+	return true;
+}
+
+static const struct aca_bank_ops vcn_v4_0_3_aca_bank_ops = {
+	.aca_bank_parser = vcn_v4_0_3_aca_bank_parser,
+	.aca_bank_is_valid = vcn_v4_0_3_aca_bank_is_valid,
+};
+
+static const struct aca_info vcn_v4_0_3_aca_info = {
+	.hwip = ACA_HWIP_TYPE_SMU,
+	.mask = ACA_ERROR_UE_MASK,
+	.bank_ops = &vcn_v4_0_3_aca_bank_ops,
+};
+
+static int vcn_v4_0_3_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
+{
+	int r;
+
+	r = amdgpu_ras_block_late_init(adev, ras_block);
+	if (r)
+		return r;
+
+	if (amdgpu_ras_is_supported(adev, ras_block->block) &&
+		adev->vcn.inst->ras_poison_irq.funcs) {
+		r = amdgpu_irq_get(adev, &adev->vcn.inst->ras_poison_irq, 0);
+		if (r)
+			goto late_fini;
+	}
+
+	r = amdgpu_ras_bind_aca(adev, AMDGPU_RAS_BLOCK__VCN,
+				&vcn_v4_0_3_aca_info, NULL);
+	if (r)
+		goto late_fini;
+
+	return 0;
+
+late_fini:
+	amdgpu_ras_block_late_fini(adev, ras_block);
+
+	return r;
+}
+
+static struct amdgpu_vcn_ras vcn_v4_0_3_ras = {
+	.ras_block = {
+		.hw_ops = &vcn_v4_0_3_ras_hw_ops,
+		.ras_late_init = vcn_v4_0_3_ras_late_init,
+	},
+};
+
+static void vcn_v4_0_3_set_ras_funcs(struct amdgpu_device *adev)
+{
+	adev->vcn.ras = &vcn_v4_0_3_ras;
+}
+
+static void vcn_v4_0_3_enable_ras(struct amdgpu_device *adev,
+				  int inst_idx, bool indirect)
+{
+	uint32_t tmp;
+
+	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__VCN))
+		return;
+
+	tmp = VCN_RAS_CNTL__VCPU_VCODEC_REARM_MASK |
+	      VCN_RAS_CNTL__VCPU_VCODEC_IH_EN_MASK |
+	      VCN_RAS_CNTL__VCPU_VCODEC_PMI_EN_MASK |
+	      VCN_RAS_CNTL__VCPU_VCODEC_STALL_EN_MASK;
+	WREG32_SOC15_DPG_MODE(inst_idx,
+			      SOC15_DPG_MODE_OFFSET(VCN, 0, regVCN_RAS_CNTL),
+			      tmp, 0, indirect);
+
+	tmp = UVD_VCPU_INT_EN2__RASCNTL_VCPU_VCODEC_EN_MASK;
+	WREG32_SOC15_DPG_MODE(inst_idx,
+			      SOC15_DPG_MODE_OFFSET(VCN, 0, regUVD_VCPU_INT_EN2),
+			      tmp, 0, indirect);
+
+	tmp = UVD_SYS_INT_EN__RASCNTL_VCPU_VCODEC_EN_MASK;
+	WREG32_SOC15_DPG_MODE(inst_idx,
+			      SOC15_DPG_MODE_OFFSET(VCN, 0, regUVD_SYS_INT_EN),
+			      tmp, 0, indirect);
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.h b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.h
new file mode 100644
index 000000000000..aeab89853a92
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.h
@@ -0,0 +1,44 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __VCN_V4_0_3_H__
+#define __VCN_V4_0_3_H__
+
+enum amdgpu_vcn_v4_0_3_sub_block {
+	AMDGPU_VCN_V4_0_3_VCPU_VCODEC = 0,
+
+	AMDGPU_VCN_V4_0_3_MAX_SUB_BLOCK,
+};
+
+extern const struct amdgpu_ip_block_version vcn_v4_0_3_ip_block;
+
+void vcn_v4_0_3_enc_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
+				       uint32_t val, uint32_t mask);
+
+void vcn_v4_0_3_enc_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg,
+				   uint32_t val);
+void vcn_v4_0_3_enc_ring_emit_vm_flush(struct amdgpu_ring *ring,
+				       unsigned int vmid, uint64_t pd_addr);
+void vcn_v4_0_3_ring_emit_hdp_flush(struct amdgpu_ring *ring);
+
+#endif /* __VCN_V4_0_3_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c
new file mode 100644
index 000000000000..b107ee80e472
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c
@@ -0,0 +1,1723 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/firmware.h>
+#include "amdgpu.h"
+#include "amdgpu_vcn.h"
+#include "amdgpu_pm.h"
+#include "amdgpu_cs.h"
+#include "soc15.h"
+#include "soc15d.h"
+#include "soc15_hw_ip.h"
+#include "vcn_v2_0.h"
+#include "mmsch_v4_0.h"
+#include "vcn_v4_0_5.h"
+
+#include "vcn/vcn_4_0_5_offset.h"
+#include "vcn/vcn_4_0_5_sh_mask.h"
+#include "ivsrcid/vcn/irqsrcs_vcn_4_0.h"
+
+#include <drm/drm_drv.h>
+
+#define mmUVD_DPG_LMA_CTL							regUVD_DPG_LMA_CTL
+#define mmUVD_DPG_LMA_CTL_BASE_IDX					regUVD_DPG_LMA_CTL_BASE_IDX
+#define mmUVD_DPG_LMA_DATA							regUVD_DPG_LMA_DATA
+#define mmUVD_DPG_LMA_DATA_BASE_IDX					regUVD_DPG_LMA_DATA_BASE_IDX
+
+#define VCN_VID_SOC_ADDRESS_2_0						0x1fb00
+#define VCN1_VID_SOC_ADDRESS_3_0					(0x48300 + 0x38000)
+#define VCN1_AON_SOC_ADDRESS_3_0					(0x48000 + 0x38000)
+
+#define VCN_HARVEST_MMSCH							0
+
+#define RDECODE_MSG_CREATE							0x00000000
+#define RDECODE_MESSAGE_CREATE						0x00000001
+
+static const struct amdgpu_hwip_reg_entry vcn_reg_list_4_0_5[] = {
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_POWER_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_CONTEXT_ID),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_CONTEXT_ID2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_GPCOM_VCPU_DATA0),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_GPCOM_VCPU_DATA1),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_GPCOM_VCPU_CMD),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_PGFSM_CONFIG),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_PGFSM_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_LMA_CTL),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_LMA_DATA),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_LMA_MASK),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_PAUSE)
+};
+
+static int amdgpu_ih_clientid_vcns[] = {
+	SOC15_IH_CLIENTID_VCN,
+	SOC15_IH_CLIENTID_VCN1
+};
+
+static void vcn_v4_0_5_set_unified_ring_funcs(struct amdgpu_device *adev);
+static void vcn_v4_0_5_set_irq_funcs(struct amdgpu_device *adev);
+static int vcn_v4_0_5_set_pg_state(struct amdgpu_vcn_inst *vinst,
+				   enum amd_powergating_state state);
+static int vcn_v4_0_5_pause_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				     struct dpg_pause_state *new_state);
+static void vcn_v4_0_5_unified_ring_set_wptr(struct amdgpu_ring *ring);
+
+/**
+ * vcn_v4_0_5_early_init - set function pointers and load microcode
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Set ring and irq function pointers
+ * Load microcode from filesystem
+ */
+static int vcn_v4_0_5_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, r;
+
+	if (amdgpu_ip_version(adev, UVD_HWIP, 0) == IP_VERSION(4, 0, 6))
+		adev->vcn.per_inst_fw = true;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i)
+		/* re-use enc ring as unified ring */
+		adev->vcn.inst[i].num_enc_rings = 1;
+	vcn_v4_0_5_set_unified_ring_funcs(adev);
+	vcn_v4_0_5_set_irq_funcs(adev);
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		adev->vcn.inst[i].set_pg_state = vcn_v4_0_5_set_pg_state;
+
+		r = amdgpu_vcn_early_init(adev, i);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+/**
+ * vcn_v4_0_5_sw_init - sw init for VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Load firmware and sw initialization
+ */
+static int vcn_v4_0_5_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_ring *ring;
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, r;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		struct amdgpu_vcn4_fw_shared *fw_shared;
+
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		r = amdgpu_vcn_sw_init(adev, i);
+		if (r)
+			return r;
+
+		amdgpu_vcn_setup_ucode(adev, i);
+
+		r = amdgpu_vcn_resume(adev, i);
+		if (r)
+			return r;
+
+		atomic_set(&adev->vcn.inst[i].sched_score, 0);
+
+		/* VCN UNIFIED TRAP */
+		r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_vcns[i],
+				VCN_4_0__SRCID__UVD_ENC_GENERAL_PURPOSE, &adev->vcn.inst[i].irq);
+		if (r)
+			return r;
+
+		/* VCN POISON TRAP */
+		r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_vcns[i],
+				VCN_4_0__SRCID_UVD_POISON, &adev->vcn.inst[i].irq);
+		if (r)
+			return r;
+
+		ring = &adev->vcn.inst[i].ring_enc[0];
+		ring->use_doorbell = true;
+		if (amdgpu_sriov_vf(adev))
+			ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
+						i * (adev->vcn.inst[i].num_enc_rings + 1) + 1;
+		else
+			ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
+						2 + 8 * i;
+		ring->vm_hub = AMDGPU_MMHUB0(0);
+		sprintf(ring->name, "vcn_unified_%d", i);
+
+		r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[i].irq, 0,
+				AMDGPU_RING_PRIO_0, &adev->vcn.inst[i].sched_score);
+		if (r)
+			return r;
+
+		fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
+		fw_shared->present_flag_0 = cpu_to_le32(AMDGPU_FW_SHARED_FLAG_0_UNIFIED_QUEUE);
+		fw_shared->sq.is_enabled = 1;
+
+		fw_shared->present_flag_0 |= cpu_to_le32(AMDGPU_VCN_SMU_DPM_INTERFACE_FLAG);
+		fw_shared->smu_dpm_interface.smu_interface_type = (adev->flags & AMD_IS_APU) ?
+			AMDGPU_VCN_SMU_DPM_INTERFACE_APU : AMDGPU_VCN_SMU_DPM_INTERFACE_DGPU;
+
+		if (amdgpu_sriov_vf(adev))
+			fw_shared->present_flag_0 |= cpu_to_le32(AMDGPU_VCN_VF_RB_SETUP_FLAG);
+
+		fw_shared->present_flag_0 |= AMDGPU_FW_SHARED_FLAG_0_DRM_KEY_INJECT;
+		fw_shared->drm_key_wa.method =
+			AMDGPU_DRM_KEY_INJECT_WORKAROUND_VCNFW_ASD_HANDSHAKING;
+
+		if (amdgpu_vcnfw_log)
+			amdgpu_vcn_fwlog_init(&adev->vcn.inst[i]);
+
+		if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
+			adev->vcn.inst[i].pause_dpg_mode = vcn_v4_0_5_pause_dpg_mode;
+	}
+
+	adev->vcn.supported_reset = amdgpu_get_soft_full_reset_mask(&adev->vcn.inst[0].ring_enc[0]);
+	if (!amdgpu_sriov_vf(adev))
+		adev->vcn.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+
+	r = amdgpu_vcn_sysfs_reset_mask_init(adev);
+	if (r)
+		return r;
+
+	if (amdgpu_sriov_vf(adev)) {
+		r = amdgpu_virt_alloc_mm_table(adev);
+		if (r)
+			return r;
+	}
+
+	r = amdgpu_vcn_reg_dump_init(adev, vcn_reg_list_4_0_5, ARRAY_SIZE(vcn_reg_list_4_0_5));
+
+	return r;
+}
+
+/**
+ * vcn_v4_0_5_sw_fini - sw fini for VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * VCN suspend and free up sw allocation
+ */
+static int vcn_v4_0_5_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, r, idx;
+
+	if (drm_dev_enter(adev_to_drm(adev), &idx)) {
+		for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+			struct amdgpu_vcn4_fw_shared *fw_shared;
+
+			if (adev->vcn.harvest_config & (1 << i))
+				continue;
+
+			fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
+			fw_shared->present_flag_0 = 0;
+			fw_shared->sq.is_enabled = 0;
+		}
+
+		drm_dev_exit(idx);
+	}
+
+	if (amdgpu_sriov_vf(adev))
+		amdgpu_virt_free_mm_table(adev);
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_suspend(adev, i);
+		if (r)
+			return r;
+
+		amdgpu_vcn_sw_fini(adev, i);
+	}
+
+	return 0;
+}
+
+/**
+ * vcn_v4_0_5_hw_init - start and test VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Initialize the hardware, boot up the VCPU and do some testing
+ */
+static int vcn_v4_0_5_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring;
+	int i, r;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		ring = &adev->vcn.inst[i].ring_enc[0];
+
+		adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell,
+				((adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 8 * i), i);
+
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+/**
+ * vcn_v4_0_5_hw_fini - stop the hardware block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Stop the VCN block, mark ring as not ready any more
+ */
+static int vcn_v4_0_5_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[i];
+
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		cancel_delayed_work_sync(&vinst->idle_work);
+
+		if (!amdgpu_sriov_vf(adev)) {
+			if ((adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) ||
+			    (vinst->cur_state != AMD_PG_STATE_GATE &&
+			     RREG32_SOC15(VCN, i, regUVD_STATUS))) {
+				vinst->set_pg_state(vinst, AMD_PG_STATE_GATE);
+			}
+		}
+	}
+
+	return 0;
+}
+
+/**
+ * vcn_v4_0_5_suspend - suspend VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * HW fini and suspend VCN block
+ */
+static int vcn_v4_0_5_suspend(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r, i;
+
+	r = vcn_v4_0_5_hw_fini(ip_block);
+	if (r)
+		return r;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_suspend(ip_block->adev, i);
+		if (r)
+			return r;
+	}
+
+	return r;
+}
+
+/**
+ * vcn_v4_0_5_resume - resume VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Resume firmware and hw init VCN block
+ */
+static int vcn_v4_0_5_resume(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r, i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_resume(ip_block->adev, i);
+		if (r)
+			return r;
+	}
+
+	r = vcn_v4_0_5_hw_init(ip_block);
+
+	return r;
+}
+
+/**
+ * vcn_v4_0_5_mc_resume - memory controller programming
+ *
+ * @vinst: VCN instance
+ *
+ * Let the VCN memory controller know it's offsets
+ */
+static void vcn_v4_0_5_mc_resume(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
+	uint32_t offset, size;
+	const struct common_firmware_header *hdr;
+
+	hdr = (const struct common_firmware_header *)adev->vcn.inst[inst].fw->data;
+	size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
+
+	/* cache window 0: fw */
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
+			(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst].tmr_mc_addr_lo));
+		WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
+			(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst].tmr_mc_addr_hi));
+		WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_OFFSET0, 0);
+		offset = 0;
+	} else {
+		WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
+			lower_32_bits(adev->vcn.inst[inst].gpu_addr));
+		WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
+			upper_32_bits(adev->vcn.inst[inst].gpu_addr));
+		offset = size;
+		WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_OFFSET0, AMDGPU_UVD_FIRMWARE_OFFSET >> 3);
+	}
+	WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_SIZE0, size);
+
+	/* cache window 1: stack */
+	WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW,
+		lower_32_bits(adev->vcn.inst[inst].gpu_addr + offset));
+	WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH,
+		upper_32_bits(adev->vcn.inst[inst].gpu_addr + offset));
+	WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_OFFSET1, 0);
+	WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_SIZE1, AMDGPU_VCN_STACK_SIZE);
+
+	/* cache window 2: context */
+	WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW,
+		lower_32_bits(adev->vcn.inst[inst].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE));
+	WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH,
+		upper_32_bits(adev->vcn.inst[inst].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE));
+	WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_OFFSET2, 0);
+	WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_SIZE2, AMDGPU_VCN_CONTEXT_SIZE);
+
+	/* non-cache window */
+	WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_NC0_64BIT_BAR_LOW,
+		lower_32_bits(adev->vcn.inst[inst].fw_shared.gpu_addr));
+	WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH,
+		upper_32_bits(adev->vcn.inst[inst].fw_shared.gpu_addr));
+	WREG32_SOC15(VCN, inst, regUVD_VCPU_NONCACHE_OFFSET0, 0);
+	WREG32_SOC15(VCN, inst, regUVD_VCPU_NONCACHE_SIZE0,
+		AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn4_fw_shared)));
+}
+
+/**
+ * vcn_v4_0_5_mc_resume_dpg_mode - memory controller programming for dpg mode
+ *
+ * @vinst: VCN instance
+ * @indirect: indirectly write sram
+ *
+ * Let the VCN memory controller know it's offsets with dpg mode
+ */
+static void vcn_v4_0_5_mc_resume_dpg_mode(struct amdgpu_vcn_inst *vinst,
+					  bool indirect)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	uint32_t offset, size;
+	const struct common_firmware_header *hdr;
+
+	hdr = (const struct common_firmware_header *)adev->vcn.inst[inst_idx].fw->data;
+	size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
+
+	/* cache window 0: fw */
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		if (!indirect) {
+			WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW),
+			(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst_idx].tmr_mc_addr_lo),
+			0, indirect);
+			WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH),
+			(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst_idx].tmr_mc_addr_hi),
+			0, indirect);
+			WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+				VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET0), 0, 0, indirect);
+		} else {
+			WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW), 0, 0, indirect);
+			WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH), 0, 0, indirect);
+			WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+				VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET0), 0, 0, indirect);
+		}
+		offset = 0;
+	} else {
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW),
+			lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr), 0, indirect);
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH),
+			upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr), 0, indirect);
+		offset = size;
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET0),
+			AMDGPU_UVD_FIRMWARE_OFFSET >> 3, 0, indirect);
+	}
+
+	if (!indirect)
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_VCPU_CACHE_SIZE0), size, 0, indirect);
+	else
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_VCPU_CACHE_SIZE0), 0, 0, indirect);
+
+	/* cache window 1: stack */
+	if (!indirect) {
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW),
+			lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset), 0, indirect);
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH),
+			upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset), 0, indirect);
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET1), 0, 0, indirect);
+	} else {
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW), 0, 0, indirect);
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH), 0, 0, indirect);
+		WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET1), 0, 0, indirect);
+	}
+
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_VCPU_CACHE_SIZE1), AMDGPU_VCN_STACK_SIZE, 0, indirect);
+
+	/* cache window 2: context */
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW),
+		lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE),
+		0, indirect);
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH),
+		upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE),
+		0, indirect);
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET2), 0, 0, indirect);
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_VCPU_CACHE_SIZE2), AMDGPU_VCN_CONTEXT_SIZE, 0, indirect);
+
+	/* non-cache window */
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_LMI_VCPU_NC0_64BIT_BAR_LOW),
+		lower_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr), 0, indirect);
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH),
+		upper_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr), 0, indirect);
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_VCPU_NONCACHE_OFFSET0), 0, 0, indirect);
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_VCPU_NONCACHE_SIZE0),
+		AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn4_fw_shared)), 0, indirect);
+
+	/* VCN global tiling registers */
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_GFX10_ADDR_CONFIG),
+		adev->gfx.config.gb_addr_config, 0, indirect);
+}
+
+/**
+ * vcn_v4_0_5_disable_static_power_gating - disable VCN static power gating
+ *
+ * @vinst: VCN instance
+ *
+ * Disable static power gating for VCN block
+ */
+static void vcn_v4_0_5_disable_static_power_gating(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
+	uint32_t data = 0;
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN) {
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG,
+					1 << UVD_IPX_DLDO_CONFIG__ONO2_PWR_CONFIG__SHIFT);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS, 0,
+					UVD_IPX_DLDO_STATUS__ONO2_PWR_STATUS_MASK);
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG,
+					2 << UVD_IPX_DLDO_CONFIG__ONO3_PWR_CONFIG__SHIFT);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS,
+					1 << UVD_IPX_DLDO_STATUS__ONO3_PWR_STATUS__SHIFT,
+					UVD_IPX_DLDO_STATUS__ONO3_PWR_STATUS_MASK);
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG,
+					2 << UVD_IPX_DLDO_CONFIG__ONO4_PWR_CONFIG__SHIFT);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS,
+					1 << UVD_IPX_DLDO_STATUS__ONO4_PWR_STATUS__SHIFT,
+					UVD_IPX_DLDO_STATUS__ONO4_PWR_STATUS_MASK);
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG,
+					2 << UVD_IPX_DLDO_CONFIG__ONO5_PWR_CONFIG__SHIFT);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS,
+					1 << UVD_IPX_DLDO_STATUS__ONO5_PWR_STATUS__SHIFT,
+					UVD_IPX_DLDO_STATUS__ONO5_PWR_STATUS_MASK);
+	} else {
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG,
+			1 << UVD_IPX_DLDO_CONFIG__ONO2_PWR_CONFIG__SHIFT);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS,
+			0, UVD_IPX_DLDO_STATUS__ONO2_PWR_STATUS_MASK);
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG,
+			1 << UVD_IPX_DLDO_CONFIG__ONO3_PWR_CONFIG__SHIFT);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS,
+			0, UVD_IPX_DLDO_STATUS__ONO3_PWR_STATUS_MASK);
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG,
+			1 << UVD_IPX_DLDO_CONFIG__ONO4_PWR_CONFIG__SHIFT);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS,
+			0, UVD_IPX_DLDO_STATUS__ONO4_PWR_STATUS_MASK);
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG,
+			1 << UVD_IPX_DLDO_CONFIG__ONO5_PWR_CONFIG__SHIFT);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS,
+			0, UVD_IPX_DLDO_STATUS__ONO5_PWR_STATUS_MASK);
+	}
+
+	data = RREG32_SOC15(VCN, inst, regUVD_POWER_STATUS);
+	data &= ~0x103;
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN)
+		data |= UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON |
+			UVD_POWER_STATUS__UVD_PG_EN_MASK;
+	WREG32_SOC15(VCN, inst, regUVD_POWER_STATUS, data);
+}
+
+/**
+ * vcn_v4_0_5_enable_static_power_gating - enable VCN static power gating
+ *
+ * @vinst: VCN instance
+ *
+ * Enable static power gating for VCN block
+ */
+static void vcn_v4_0_5_enable_static_power_gating(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
+	uint32_t data;
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN) {
+		/* Before power off, this indicator has to be turned on */
+		data = RREG32_SOC15(VCN, inst, regUVD_POWER_STATUS);
+		data &= ~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK;
+		data |= UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF;
+		WREG32_SOC15(VCN, inst, regUVD_POWER_STATUS, data);
+
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG,
+			2 << UVD_IPX_DLDO_CONFIG__ONO5_PWR_CONFIG__SHIFT);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS,
+			1 << UVD_IPX_DLDO_STATUS__ONO5_PWR_STATUS__SHIFT,
+			UVD_IPX_DLDO_STATUS__ONO5_PWR_STATUS_MASK);
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG,
+			2 << UVD_IPX_DLDO_CONFIG__ONO4_PWR_CONFIG__SHIFT);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS,
+			1 << UVD_IPX_DLDO_STATUS__ONO4_PWR_STATUS__SHIFT,
+			UVD_IPX_DLDO_STATUS__ONO4_PWR_STATUS_MASK);
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG,
+			2 << UVD_IPX_DLDO_CONFIG__ONO3_PWR_CONFIG__SHIFT);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS,
+			1 << UVD_IPX_DLDO_STATUS__ONO3_PWR_STATUS__SHIFT,
+			UVD_IPX_DLDO_STATUS__ONO3_PWR_STATUS_MASK);
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG,
+			2 << UVD_IPX_DLDO_CONFIG__ONO2_PWR_CONFIG__SHIFT);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS,
+			1 << UVD_IPX_DLDO_STATUS__ONO2_PWR_STATUS__SHIFT,
+			UVD_IPX_DLDO_STATUS__ONO2_PWR_STATUS_MASK);
+	}
+}
+
+/**
+ * vcn_v4_0_5_disable_clock_gating - disable VCN clock gating
+ *
+ * @vinst: VCN instance
+ *
+ * Disable clock gating for VCN block
+ */
+static void vcn_v4_0_5_disable_clock_gating(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
+	uint32_t data;
+
+	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
+		return;
+
+	/* VCN disable CGC */
+	data = RREG32_SOC15(VCN, inst, regUVD_CGC_CTRL);
+	data &= ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK;
+	data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
+	data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
+	WREG32_SOC15(VCN, inst, regUVD_CGC_CTRL, data);
+
+	data = RREG32_SOC15(VCN, inst, regUVD_CGC_GATE);
+	data &= ~(UVD_CGC_GATE__SYS_MASK
+		| UVD_CGC_GATE__UDEC_MASK
+		| UVD_CGC_GATE__MPEG2_MASK
+		| UVD_CGC_GATE__REGS_MASK
+		| UVD_CGC_GATE__RBC_MASK
+		| UVD_CGC_GATE__LMI_MC_MASK
+		| UVD_CGC_GATE__LMI_UMC_MASK
+		| UVD_CGC_GATE__IDCT_MASK
+		| UVD_CGC_GATE__MPRD_MASK
+		| UVD_CGC_GATE__MPC_MASK
+		| UVD_CGC_GATE__LBSI_MASK
+		| UVD_CGC_GATE__LRBBM_MASK
+		| UVD_CGC_GATE__UDEC_RE_MASK
+		| UVD_CGC_GATE__UDEC_CM_MASK
+		| UVD_CGC_GATE__UDEC_IT_MASK
+		| UVD_CGC_GATE__UDEC_DB_MASK
+		| UVD_CGC_GATE__UDEC_MP_MASK
+		| UVD_CGC_GATE__WCB_MASK
+		| UVD_CGC_GATE__VCPU_MASK
+		| UVD_CGC_GATE__MMSCH_MASK);
+
+	WREG32_SOC15(VCN, inst, regUVD_CGC_GATE, data);
+	SOC15_WAIT_ON_RREG(VCN, inst, regUVD_CGC_GATE, 0,  0xFFFFFFFF);
+
+	data = RREG32_SOC15(VCN, inst, regUVD_CGC_CTRL);
+	data &= ~(UVD_CGC_CTRL__UDEC_RE_MODE_MASK
+		| UVD_CGC_CTRL__UDEC_CM_MODE_MASK
+		| UVD_CGC_CTRL__UDEC_IT_MODE_MASK
+		| UVD_CGC_CTRL__UDEC_DB_MODE_MASK
+		| UVD_CGC_CTRL__UDEC_MP_MODE_MASK
+		| UVD_CGC_CTRL__SYS_MODE_MASK
+		| UVD_CGC_CTRL__UDEC_MODE_MASK
+		| UVD_CGC_CTRL__MPEG2_MODE_MASK
+		| UVD_CGC_CTRL__REGS_MODE_MASK
+		| UVD_CGC_CTRL__RBC_MODE_MASK
+		| UVD_CGC_CTRL__LMI_MC_MODE_MASK
+		| UVD_CGC_CTRL__LMI_UMC_MODE_MASK
+		| UVD_CGC_CTRL__IDCT_MODE_MASK
+		| UVD_CGC_CTRL__MPRD_MODE_MASK
+		| UVD_CGC_CTRL__MPC_MODE_MASK
+		| UVD_CGC_CTRL__LBSI_MODE_MASK
+		| UVD_CGC_CTRL__LRBBM_MODE_MASK
+		| UVD_CGC_CTRL__WCB_MODE_MASK
+		| UVD_CGC_CTRL__VCPU_MODE_MASK
+		| UVD_CGC_CTRL__MMSCH_MODE_MASK);
+	WREG32_SOC15(VCN, inst, regUVD_CGC_CTRL, data);
+
+	data = RREG32_SOC15(VCN, inst, regUVD_SUVD_CGC_GATE);
+	data |= (UVD_SUVD_CGC_GATE__SRE_MASK
+		| UVD_SUVD_CGC_GATE__SIT_MASK
+		| UVD_SUVD_CGC_GATE__SMP_MASK
+		| UVD_SUVD_CGC_GATE__SCM_MASK
+		| UVD_SUVD_CGC_GATE__SDB_MASK
+		| UVD_SUVD_CGC_GATE__SRE_H264_MASK
+		| UVD_SUVD_CGC_GATE__SRE_HEVC_MASK
+		| UVD_SUVD_CGC_GATE__SIT_H264_MASK
+		| UVD_SUVD_CGC_GATE__SIT_HEVC_MASK
+		| UVD_SUVD_CGC_GATE__SCM_H264_MASK
+		| UVD_SUVD_CGC_GATE__SCM_HEVC_MASK
+		| UVD_SUVD_CGC_GATE__SDB_H264_MASK
+		| UVD_SUVD_CGC_GATE__SDB_HEVC_MASK
+		| UVD_SUVD_CGC_GATE__SCLR_MASK
+		| UVD_SUVD_CGC_GATE__UVD_SC_MASK
+		| UVD_SUVD_CGC_GATE__ENT_MASK
+		| UVD_SUVD_CGC_GATE__SIT_HEVC_DEC_MASK
+		| UVD_SUVD_CGC_GATE__SIT_HEVC_ENC_MASK
+		| UVD_SUVD_CGC_GATE__SITE_MASK
+		| UVD_SUVD_CGC_GATE__SRE_VP9_MASK
+		| UVD_SUVD_CGC_GATE__SCM_VP9_MASK
+		| UVD_SUVD_CGC_GATE__SIT_VP9_DEC_MASK
+		| UVD_SUVD_CGC_GATE__SDB_VP9_MASK
+		| UVD_SUVD_CGC_GATE__IME_HEVC_MASK);
+	WREG32_SOC15(VCN, inst, regUVD_SUVD_CGC_GATE, data);
+
+	data = RREG32_SOC15(VCN, inst, regUVD_SUVD_CGC_CTRL);
+	data &= ~(UVD_SUVD_CGC_CTRL__SRE_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SIT_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SMP_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SCM_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SDB_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SCLR_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__UVD_SC_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__ENT_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__IME_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SITE_MODE_MASK);
+	WREG32_SOC15(VCN, inst, regUVD_SUVD_CGC_CTRL, data);
+}
+
+/**
+ * vcn_v4_0_5_disable_clock_gating_dpg_mode - disable VCN clock gating dpg mode
+ *
+ * @vinst: VCN instance
+ * @sram_sel: sram select
+ * @indirect: indirectly write sram
+ *
+ * Disable clock gating for VCN block with dpg mode
+ */
+static void vcn_v4_0_5_disable_clock_gating_dpg_mode(struct amdgpu_vcn_inst *vinst,
+						     uint8_t sram_sel,
+						     uint8_t indirect)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	uint32_t reg_data = 0;
+
+	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
+		return;
+
+	/* enable sw clock gating control */
+	reg_data = 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+	reg_data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
+	reg_data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
+	reg_data &= ~(UVD_CGC_CTRL__UDEC_RE_MODE_MASK |
+		 UVD_CGC_CTRL__UDEC_CM_MODE_MASK |
+		 UVD_CGC_CTRL__UDEC_IT_MODE_MASK |
+		 UVD_CGC_CTRL__UDEC_DB_MODE_MASK |
+		 UVD_CGC_CTRL__UDEC_MP_MODE_MASK |
+		 UVD_CGC_CTRL__SYS_MODE_MASK |
+		 UVD_CGC_CTRL__UDEC_MODE_MASK |
+		 UVD_CGC_CTRL__MPEG2_MODE_MASK |
+		 UVD_CGC_CTRL__REGS_MODE_MASK |
+		 UVD_CGC_CTRL__RBC_MODE_MASK |
+		 UVD_CGC_CTRL__LMI_MC_MODE_MASK |
+		 UVD_CGC_CTRL__LMI_UMC_MODE_MASK |
+		 UVD_CGC_CTRL__IDCT_MODE_MASK |
+		 UVD_CGC_CTRL__MPRD_MODE_MASK |
+		 UVD_CGC_CTRL__MPC_MODE_MASK |
+		 UVD_CGC_CTRL__LBSI_MODE_MASK |
+		 UVD_CGC_CTRL__LRBBM_MODE_MASK |
+		 UVD_CGC_CTRL__WCB_MODE_MASK |
+		 UVD_CGC_CTRL__VCPU_MODE_MASK);
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_CGC_CTRL), reg_data, sram_sel, indirect);
+
+	/* turn off clock gating */
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_CGC_GATE), 0, sram_sel, indirect);
+
+	/* turn on SUVD clock gating */
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_SUVD_CGC_GATE), 1, sram_sel, indirect);
+
+	/* turn on sw mode in UVD_SUVD_CGC_CTRL */
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_SUVD_CGC_CTRL), 0, sram_sel, indirect);
+}
+
+/**
+ * vcn_v4_0_5_enable_clock_gating - enable VCN clock gating
+ *
+ * @vinst: VCN instance
+ *
+ * Enable clock gating for VCN block
+ */
+static void vcn_v4_0_5_enable_clock_gating(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
+	uint32_t data;
+
+	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
+		return;
+
+	/* enable VCN CGC */
+	data = RREG32_SOC15(VCN, inst, regUVD_CGC_CTRL);
+	data |= 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+	data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
+	data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
+	WREG32_SOC15(VCN, inst, regUVD_CGC_CTRL, data);
+
+	data = RREG32_SOC15(VCN, inst, regUVD_CGC_CTRL);
+	data |= (UVD_CGC_CTRL__UDEC_RE_MODE_MASK
+		| UVD_CGC_CTRL__UDEC_CM_MODE_MASK
+		| UVD_CGC_CTRL__UDEC_IT_MODE_MASK
+		| UVD_CGC_CTRL__UDEC_DB_MODE_MASK
+		| UVD_CGC_CTRL__UDEC_MP_MODE_MASK
+		| UVD_CGC_CTRL__SYS_MODE_MASK
+		| UVD_CGC_CTRL__UDEC_MODE_MASK
+		| UVD_CGC_CTRL__MPEG2_MODE_MASK
+		| UVD_CGC_CTRL__REGS_MODE_MASK
+		| UVD_CGC_CTRL__RBC_MODE_MASK
+		| UVD_CGC_CTRL__LMI_MC_MODE_MASK
+		| UVD_CGC_CTRL__LMI_UMC_MODE_MASK
+		| UVD_CGC_CTRL__IDCT_MODE_MASK
+		| UVD_CGC_CTRL__MPRD_MODE_MASK
+		| UVD_CGC_CTRL__MPC_MODE_MASK
+		| UVD_CGC_CTRL__LBSI_MODE_MASK
+		| UVD_CGC_CTRL__LRBBM_MODE_MASK
+		| UVD_CGC_CTRL__WCB_MODE_MASK
+		| UVD_CGC_CTRL__VCPU_MODE_MASK
+		| UVD_CGC_CTRL__MMSCH_MODE_MASK);
+	WREG32_SOC15(VCN, inst, regUVD_CGC_CTRL, data);
+
+	data = RREG32_SOC15(VCN, inst, regUVD_SUVD_CGC_CTRL);
+	data |= (UVD_SUVD_CGC_CTRL__SRE_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SIT_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SMP_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SCM_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SDB_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SCLR_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__UVD_SC_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__ENT_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__IME_MODE_MASK
+		| UVD_SUVD_CGC_CTRL__SITE_MODE_MASK);
+	WREG32_SOC15(VCN, inst, regUVD_SUVD_CGC_CTRL, data);
+}
+
+/**
+ * vcn_v4_0_5_start_dpg_mode - VCN start with dpg mode
+ *
+ * @vinst: VCN instance
+ * @indirect: indirectly write sram
+ *
+ * Start VCN block with dpg mode
+ */
+static int vcn_v4_0_5_start_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				     bool indirect)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	struct amdgpu_vcn4_fw_shared *fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr;
+	struct amdgpu_ring *ring;
+	uint32_t tmp;
+	int ret;
+
+	/* disable register anti-hang mechanism */
+	WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, regUVD_POWER_STATUS), 1,
+		~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
+	/* enable dynamic power gating mode */
+	tmp = RREG32_SOC15(VCN, inst_idx, regUVD_POWER_STATUS);
+	tmp |= UVD_POWER_STATUS__UVD_PG_MODE_MASK;
+	tmp |= UVD_POWER_STATUS__UVD_PG_EN_MASK;
+	WREG32_SOC15(VCN, inst_idx, regUVD_POWER_STATUS, tmp);
+
+	if (indirect)
+		adev->vcn.inst[inst_idx].dpg_sram_curr_addr =
+					(uint32_t *)adev->vcn.inst[inst_idx].dpg_sram_cpu_addr;
+
+	/* enable clock gating */
+	vcn_v4_0_5_disable_clock_gating_dpg_mode(vinst, 0, indirect);
+
+	/* enable VCPU clock */
+	tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
+	tmp |= UVD_VCPU_CNTL__CLK_EN_MASK | UVD_VCPU_CNTL__BLK_RST_MASK;
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_VCPU_CNTL), tmp, 0, indirect);
+
+	/* disable master interrupt */
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_MASTINT_EN), 0, 0, indirect);
+
+	/* setup regUVD_LMI_CTRL */
+	tmp = (UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
+		UVD_LMI_CTRL__REQ_MODE_MASK |
+		UVD_LMI_CTRL__CRC_RESET_MASK |
+		UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
+		UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
+		UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK |
+		(8 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) |
+		0x00100000L);
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_LMI_CTRL), tmp, 0, indirect);
+
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_MPC_CNTL),
+		0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT, 0, indirect);
+
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_MPC_SET_MUXA0),
+		((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) |
+		 (0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) |
+		 (0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) |
+		 (0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)), 0, indirect);
+
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_MPC_SET_MUXB0),
+		 ((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) |
+		 (0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) |
+		 (0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) |
+		 (0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)), 0, indirect);
+
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_MPC_SET_MUX),
+		((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) |
+		 (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
+		 (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)), 0, indirect);
+
+	vcn_v4_0_5_mc_resume_dpg_mode(vinst, indirect);
+
+	tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
+	tmp |= UVD_VCPU_CNTL__CLK_EN_MASK;
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_VCPU_CNTL), tmp, 0, indirect);
+
+	/* enable LMI MC and UMC channels */
+	tmp = 0x1f << UVD_LMI_CTRL2__RE_OFLD_MIF_WR_REQ_NUM__SHIFT;
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_LMI_CTRL2), tmp, 0, indirect);
+
+	/* enable master interrupt */
+	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_MASTINT_EN),
+		UVD_MASTINT_EN__VCPU_EN_MASK, 0, indirect);
+
+	if (indirect) {
+		ret = amdgpu_vcn_psp_update_sram(adev, inst_idx, 0);
+		if (ret) {
+			dev_err(adev->dev, "vcn sram load failed %d\n", ret);
+			return ret;
+		}
+	}
+
+	ring = &adev->vcn.inst[inst_idx].ring_enc[0];
+
+	WREG32_SOC15(VCN, inst_idx, regUVD_RB_BASE_LO, ring->gpu_addr);
+	WREG32_SOC15(VCN, inst_idx, regUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
+	WREG32_SOC15(VCN, inst_idx, regUVD_RB_SIZE, ring->ring_size / 4);
+
+	tmp = RREG32_SOC15(VCN, inst_idx, regVCN_RB_ENABLE);
+	tmp &= ~(VCN_RB_ENABLE__RB1_EN_MASK);
+	WREG32_SOC15(VCN, inst_idx, regVCN_RB_ENABLE, tmp);
+	fw_shared->sq.queue_mode |= FW_QUEUE_RING_RESET;
+	WREG32_SOC15(VCN, inst_idx, regUVD_RB_RPTR, 0);
+	WREG32_SOC15(VCN, inst_idx, regUVD_RB_WPTR, 0);
+
+	tmp = RREG32_SOC15(VCN, inst_idx, regUVD_RB_RPTR);
+	WREG32_SOC15(VCN, inst_idx, regUVD_RB_WPTR, tmp);
+	ring->wptr = RREG32_SOC15(VCN, inst_idx, regUVD_RB_WPTR);
+
+	tmp = RREG32_SOC15(VCN, inst_idx, regVCN_RB_ENABLE);
+	tmp |= VCN_RB_ENABLE__RB1_EN_MASK;
+	WREG32_SOC15(VCN, inst_idx, regVCN_RB_ENABLE, tmp);
+	fw_shared->sq.queue_mode &= ~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF);
+
+	WREG32_SOC15(VCN, inst_idx, regVCN_RB1_DB_CTRL,
+			ring->doorbell_index << VCN_RB1_DB_CTRL__OFFSET__SHIFT |
+			VCN_RB1_DB_CTRL__EN_MASK);
+
+	/* Keeping one read-back to ensure all register writes are done, otherwise
+	 * it may introduce race conditions */
+	RREG32_SOC15(VCN, inst_idx, regVCN_RB1_DB_CTRL);
+
+	return 0;
+}
+
+
+/**
+ * vcn_v4_0_5_start - VCN start
+ *
+ * @vinst: VCN instance
+ *
+ * Start VCN block
+ */
+static int vcn_v4_0_5_start(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int i = vinst->inst;
+	struct amdgpu_vcn4_fw_shared *fw_shared;
+	struct amdgpu_ring *ring;
+	uint32_t tmp;
+	int j, k, r;
+
+	if (adev->vcn.harvest_config & (1 << i))
+		return 0;
+
+	if (adev->pm.dpm_enabled)
+		amdgpu_dpm_enable_vcn(adev, true, i);
+
+	fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
+		return vcn_v4_0_5_start_dpg_mode(vinst, adev->vcn.inst[i].indirect_sram);
+
+	/* disable VCN power gating */
+	vcn_v4_0_5_disable_static_power_gating(vinst);
+
+	/* set VCN status busy */
+	tmp = RREG32_SOC15(VCN, i, regUVD_STATUS) | UVD_STATUS__UVD_BUSY;
+	WREG32_SOC15(VCN, i, regUVD_STATUS, tmp);
+
+	/* SW clock gating */
+	vcn_v4_0_5_disable_clock_gating(vinst);
+
+	/* enable VCPU clock */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL),
+		 UVD_VCPU_CNTL__CLK_EN_MASK, ~UVD_VCPU_CNTL__CLK_EN_MASK);
+
+	/* disable master interrupt */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_MASTINT_EN), 0,
+		 ~UVD_MASTINT_EN__VCPU_EN_MASK);
+
+	/* enable LMI MC and UMC channels */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_LMI_CTRL2), 0,
+		 ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
+
+	tmp = RREG32_SOC15(VCN, i, regUVD_SOFT_RESET);
+	tmp &= ~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
+	tmp &= ~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
+	WREG32_SOC15(VCN, i, regUVD_SOFT_RESET, tmp);
+
+	/* setup regUVD_LMI_CTRL */
+	tmp = RREG32_SOC15(VCN, i, regUVD_LMI_CTRL);
+	WREG32_SOC15(VCN, i, regUVD_LMI_CTRL, tmp |
+		     UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
+		     UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
+		     UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
+		     UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK);
+
+	/* setup regUVD_MPC_CNTL */
+	tmp = RREG32_SOC15(VCN, i, regUVD_MPC_CNTL);
+	tmp &= ~UVD_MPC_CNTL__REPLACEMENT_MODE_MASK;
+	tmp |= 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT;
+	WREG32_SOC15(VCN, i, regUVD_MPC_CNTL, tmp);
+
+	/* setup UVD_MPC_SET_MUXA0 */
+	WREG32_SOC15(VCN, i, regUVD_MPC_SET_MUXA0,
+		     ((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) |
+		      (0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) |
+		      (0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) |
+		      (0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)));
+
+	/* setup UVD_MPC_SET_MUXB0 */
+	WREG32_SOC15(VCN, i, regUVD_MPC_SET_MUXB0,
+		     ((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) |
+		      (0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) |
+		      (0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) |
+		      (0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)));
+
+	/* setup UVD_MPC_SET_MUX */
+	WREG32_SOC15(VCN, i, regUVD_MPC_SET_MUX,
+		     ((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) |
+		      (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
+		      (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)));
+
+	vcn_v4_0_5_mc_resume(vinst);
+
+	/* VCN global tiling registers */
+	WREG32_SOC15(VCN, i, regUVD_GFX10_ADDR_CONFIG,
+		     adev->gfx.config.gb_addr_config);
+
+	/* unblock VCPU register access */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_RB_ARB_CTRL), 0,
+		 ~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
+
+	/* release VCPU reset to boot */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), 0,
+		 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+
+	for (j = 0; j < 10; ++j) {
+		uint32_t status;
+
+		for (k = 0; k < 100; ++k) {
+			status = RREG32_SOC15(VCN, i, regUVD_STATUS);
+			if (status & 2)
+				break;
+			mdelay(10);
+			if (amdgpu_emu_mode == 1)
+				msleep(1);
+		}
+
+		if (amdgpu_emu_mode == 1) {
+			r = -1;
+			if (status & 2) {
+				r = 0;
+				break;
+			}
+		} else {
+			r = 0;
+			if (status & 2)
+				break;
+
+			dev_err(adev->dev,
+				"VCN[%d] is not responding, trying to reset VCPU!!!\n", i);
+			WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL),
+				 UVD_VCPU_CNTL__BLK_RST_MASK,
+				 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+			mdelay(10);
+			WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), 0,
+				 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+
+			mdelay(10);
+			r = -1;
+		}
+	}
+
+	if (r) {
+		dev_err(adev->dev, "VCN[%d] is not responding, giving up!!!\n", i);
+		return r;
+	}
+
+	/* enable master interrupt */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_MASTINT_EN),
+		 UVD_MASTINT_EN__VCPU_EN_MASK,
+		 ~UVD_MASTINT_EN__VCPU_EN_MASK);
+
+	/* clear the busy bit of VCN_STATUS */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_STATUS), 0,
+		 ~(2 << UVD_STATUS__VCPU_REPORT__SHIFT));
+
+	ring = &adev->vcn.inst[i].ring_enc[0];
+	WREG32_SOC15(VCN, i, regVCN_RB1_DB_CTRL,
+		     ring->doorbell_index << VCN_RB1_DB_CTRL__OFFSET__SHIFT |
+		     VCN_RB1_DB_CTRL__EN_MASK);
+
+	WREG32_SOC15(VCN, i, regUVD_RB_BASE_LO, ring->gpu_addr);
+	WREG32_SOC15(VCN, i, regUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
+	WREG32_SOC15(VCN, i, regUVD_RB_SIZE, ring->ring_size / 4);
+
+	tmp = RREG32_SOC15(VCN, i, regVCN_RB_ENABLE);
+	tmp &= ~(VCN_RB_ENABLE__RB1_EN_MASK);
+	WREG32_SOC15(VCN, i, regVCN_RB_ENABLE, tmp);
+	fw_shared->sq.queue_mode |= FW_QUEUE_RING_RESET;
+	WREG32_SOC15(VCN, i, regUVD_RB_RPTR, 0);
+	WREG32_SOC15(VCN, i, regUVD_RB_WPTR, 0);
+
+	tmp = RREG32_SOC15(VCN, i, regUVD_RB_RPTR);
+	WREG32_SOC15(VCN, i, regUVD_RB_WPTR, tmp);
+	ring->wptr = RREG32_SOC15(VCN, i, regUVD_RB_WPTR);
+
+	tmp = RREG32_SOC15(VCN, i, regVCN_RB_ENABLE);
+	tmp |= VCN_RB_ENABLE__RB1_EN_MASK;
+	WREG32_SOC15(VCN, i, regVCN_RB_ENABLE, tmp);
+	fw_shared->sq.queue_mode &= ~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF);
+
+	/* Keeping one read-back to ensure all register writes are done, otherwise
+	 * it may introduce race conditions */
+	RREG32_SOC15(VCN, i, regVCN_RB_ENABLE);
+
+	return 0;
+}
+
+/**
+ * vcn_v4_0_5_stop_dpg_mode - VCN stop with dpg mode
+ *
+ * @vinst: VCN instance
+ *
+ * Stop VCN block with dpg mode
+ */
+static void vcn_v4_0_5_stop_dpg_mode(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	uint32_t tmp;
+
+	/* Wait for power status to be 1 */
+	SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_POWER_STATUS, 1,
+		UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
+
+	/* wait for read ptr to be equal to write ptr */
+	tmp = RREG32_SOC15(VCN, inst_idx, regUVD_RB_WPTR);
+	SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_RB_RPTR, tmp, 0xFFFFFFFF);
+
+	SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_POWER_STATUS, 1,
+		UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
+
+	/* disable dynamic power gating mode */
+	WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, regUVD_POWER_STATUS), 0,
+		~UVD_POWER_STATUS__UVD_PG_MODE_MASK);
+
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, inst_idx, regUVD_STATUS);
+}
+
+/**
+ * vcn_v4_0_5_stop - VCN stop
+ *
+ * @vinst: VCN instance
+ *
+ * Stop VCN block
+ */
+static int vcn_v4_0_5_stop(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int i = vinst->inst;
+	struct amdgpu_vcn4_fw_shared *fw_shared;
+	uint32_t tmp;
+	int r = 0;
+
+	if (adev->vcn.harvest_config & (1 << i))
+		return 0;
+
+	fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
+	fw_shared->sq.queue_mode |= FW_QUEUE_DPG_HOLD_OFF;
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
+		vcn_v4_0_5_stop_dpg_mode(vinst);
+		r = 0;
+		goto done;
+	}
+
+	/* wait for vcn idle */
+	r = SOC15_WAIT_ON_RREG(VCN, i, regUVD_STATUS, UVD_STATUS__IDLE, 0x7);
+	if (r)
+		goto done;
+
+	tmp = UVD_LMI_STATUS__VCPU_LMI_WRITE_CLEAN_MASK |
+		UVD_LMI_STATUS__READ_CLEAN_MASK |
+		UVD_LMI_STATUS__WRITE_CLEAN_MASK |
+		UVD_LMI_STATUS__WRITE_CLEAN_RAW_MASK;
+	r = SOC15_WAIT_ON_RREG(VCN, i, regUVD_LMI_STATUS, tmp, tmp);
+	if (r)
+		goto done;
+
+	/* disable LMI UMC channel */
+	tmp = RREG32_SOC15(VCN, i, regUVD_LMI_CTRL2);
+	tmp |= UVD_LMI_CTRL2__STALL_ARB_UMC_MASK;
+	WREG32_SOC15(VCN, i, regUVD_LMI_CTRL2, tmp);
+	tmp = UVD_LMI_STATUS__UMC_READ_CLEAN_RAW_MASK |
+		UVD_LMI_STATUS__UMC_WRITE_CLEAN_RAW_MASK;
+	r = SOC15_WAIT_ON_RREG(VCN, i, regUVD_LMI_STATUS, tmp, tmp);
+	if (r)
+		goto done;
+
+	/* block VCPU register access */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_RB_ARB_CTRL),
+		 UVD_RB_ARB_CTRL__VCPU_DIS_MASK,
+		 ~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
+
+	/* reset VCPU */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL),
+		 UVD_VCPU_CNTL__BLK_RST_MASK,
+		 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+
+	/* disable VCPU clock */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), 0,
+		 ~(UVD_VCPU_CNTL__CLK_EN_MASK));
+
+	/* apply soft reset */
+	tmp = RREG32_SOC15(VCN, i, regUVD_SOFT_RESET);
+	tmp |= UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
+	WREG32_SOC15(VCN, i, regUVD_SOFT_RESET, tmp);
+	tmp = RREG32_SOC15(VCN, i, regUVD_SOFT_RESET);
+	tmp |= UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
+	WREG32_SOC15(VCN, i, regUVD_SOFT_RESET, tmp);
+
+	/* clear status */
+	WREG32_SOC15(VCN, i, regUVD_STATUS, 0);
+
+	/* apply HW clock gating */
+	vcn_v4_0_5_enable_clock_gating(vinst);
+
+	/* enable VCN power gating */
+	vcn_v4_0_5_enable_static_power_gating(vinst);
+
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, i, regUVD_STATUS);
+
+done:
+	if (adev->pm.dpm_enabled)
+		amdgpu_dpm_enable_vcn(adev, false, i);
+
+	return r;
+}
+
+/**
+ * vcn_v4_0_5_pause_dpg_mode - VCN pause with dpg mode
+ *
+ * @vinst: VCN instance
+ * @new_state: pause state
+ *
+ * Pause dpg mode for VCN block
+ */
+static int vcn_v4_0_5_pause_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				     struct dpg_pause_state *new_state)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	uint32_t reg_data = 0;
+	int ret_code;
+
+	/* pause/unpause if state is changed */
+	if (adev->vcn.inst[inst_idx].pause_state.fw_based != new_state->fw_based) {
+		DRM_DEV_DEBUG(adev->dev, "dpg pause state changed %d -> %d",
+			adev->vcn.inst[inst_idx].pause_state.fw_based,	new_state->fw_based);
+		reg_data = RREG32_SOC15(VCN, inst_idx, regUVD_DPG_PAUSE) &
+			(~UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK);
+
+		if (new_state->fw_based == VCN_DPG_STATE__PAUSE) {
+			ret_code = SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_POWER_STATUS, 0x1,
+				UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
+
+			if (!ret_code) {
+				/* pause DPG */
+				reg_data |= UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK;
+				WREG32_SOC15(VCN, inst_idx, regUVD_DPG_PAUSE, reg_data);
+
+				/* wait for ACK */
+				SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_DPG_PAUSE,
+					UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK,
+					UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK);
+
+				SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_POWER_STATUS,
+					UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON,
+					UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
+			}
+		} else {
+			/* unpause dpg, no need to wait */
+			reg_data &= ~UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK;
+			WREG32_SOC15(VCN, inst_idx, regUVD_DPG_PAUSE, reg_data);
+		}
+		adev->vcn.inst[inst_idx].pause_state.fw_based = new_state->fw_based;
+	}
+
+	return 0;
+}
+
+/**
+ * vcn_v4_0_5_unified_ring_get_rptr - get unified read pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware unified read pointer
+ */
+static uint64_t vcn_v4_0_5_unified_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring != &adev->vcn.inst[ring->me].ring_enc[0])
+		DRM_ERROR("wrong ring id is identified in %s", __func__);
+
+	return RREG32_SOC15(VCN, ring->me, regUVD_RB_RPTR);
+}
+
+/**
+ * vcn_v4_0_5_unified_ring_get_wptr - get unified write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware unified write pointer
+ */
+static uint64_t vcn_v4_0_5_unified_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring != &adev->vcn.inst[ring->me].ring_enc[0])
+		DRM_ERROR("wrong ring id is identified in %s", __func__);
+
+	if (ring->use_doorbell)
+		return *ring->wptr_cpu_addr;
+	else
+		return RREG32_SOC15(VCN, ring->me, regUVD_RB_WPTR);
+}
+
+/**
+ * vcn_v4_0_5_unified_ring_set_wptr - set enc write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Commits the enc write pointer to the hardware
+ */
+static void vcn_v4_0_5_unified_ring_set_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring != &adev->vcn.inst[ring->me].ring_enc[0])
+		DRM_ERROR("wrong ring id is identified in %s", __func__);
+
+	if (ring->use_doorbell) {
+		*ring->wptr_cpu_addr = lower_32_bits(ring->wptr);
+		WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr));
+	} else {
+		WREG32_SOC15(VCN, ring->me, regUVD_RB_WPTR, lower_32_bits(ring->wptr));
+	}
+}
+
+static int vcn_v4_0_5_ring_reset(struct amdgpu_ring *ring,
+				 unsigned int vmid,
+				 struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[ring->me];
+	int r;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+	r = vcn_v4_0_5_stop(vinst);
+	if (r)
+		return r;
+	r = vcn_v4_0_5_start(vinst);
+	if (r)
+		return r;
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static struct amdgpu_ring_funcs vcn_v4_0_5_unified_ring_vm_funcs = {
+	.type = AMDGPU_RING_TYPE_VCN_ENC,
+	.align_mask = 0x3f,
+	.nop = VCN_ENC_CMD_NO_OP,
+	.get_rptr = vcn_v4_0_5_unified_ring_get_rptr,
+	.get_wptr = vcn_v4_0_5_unified_ring_get_wptr,
+	.set_wptr = vcn_v4_0_5_unified_ring_set_wptr,
+	.emit_frame_size =
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 4 +
+		4 + /* vcn_v2_0_enc_ring_emit_vm_flush */
+		5 + 5 + /* vcn_v2_0_enc_ring_emit_fence x2 vm fence */
+		1, /* vcn_v2_0_enc_ring_insert_end */
+	.emit_ib_size = 5, /* vcn_v2_0_enc_ring_emit_ib */
+	.emit_ib = vcn_v2_0_enc_ring_emit_ib,
+	.emit_fence = vcn_v2_0_enc_ring_emit_fence,
+	.emit_vm_flush = vcn_v2_0_enc_ring_emit_vm_flush,
+	.test_ring = amdgpu_vcn_enc_ring_test_ring,
+	.test_ib = amdgpu_vcn_unified_ring_test_ib,
+	.insert_nop = amdgpu_ring_insert_nop,
+	.insert_end = vcn_v2_0_enc_ring_insert_end,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.begin_use = amdgpu_vcn_ring_begin_use,
+	.end_use = amdgpu_vcn_ring_end_use,
+	.emit_wreg = vcn_v2_0_enc_ring_emit_wreg,
+	.emit_reg_wait = vcn_v2_0_enc_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = vcn_v4_0_5_ring_reset,
+};
+
+/**
+ * vcn_v4_0_5_set_unified_ring_funcs - set unified ring functions
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set unified ring functions
+ */
+static void vcn_v4_0_5_set_unified_ring_funcs(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		if (amdgpu_ip_version(adev, VCN_HWIP, 0) == IP_VERSION(4, 0, 5))
+			vcn_v4_0_5_unified_ring_vm_funcs.secure_submission_supported = true;
+
+		adev->vcn.inst[i].ring_enc[0].funcs = &vcn_v4_0_5_unified_ring_vm_funcs;
+		adev->vcn.inst[i].ring_enc[0].me = i;
+	}
+}
+
+/**
+ * vcn_v4_0_5_is_idle - check VCN block is idle
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block structure
+ *
+ * Check whether VCN block is idle
+ */
+static bool vcn_v4_0_5_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, ret = 1;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		ret &= (RREG32_SOC15(VCN, i, regUVD_STATUS) == UVD_STATUS__IDLE);
+	}
+
+	return ret;
+}
+
+/**
+ * vcn_v4_0_5_wait_for_idle - wait for VCN block idle
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Wait for VCN block idle
+ */
+static int vcn_v4_0_5_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, ret = 0;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		ret = SOC15_WAIT_ON_RREG(VCN, i, regUVD_STATUS, UVD_STATUS__IDLE,
+			UVD_STATUS__IDLE);
+		if (ret)
+			return ret;
+	}
+
+	return ret;
+}
+
+/**
+ * vcn_v4_0_5_set_clockgating_state - set VCN block clockgating state
+ *
+ * @ip_block: amdgpu_ip_block pointer
+ * @state: clock gating state
+ *
+ * Set VCN block clockgating state
+ */
+static int vcn_v4_0_5_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_clockgating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	bool enable = state == AMD_CG_STATE_GATE;
+	int i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[i];
+
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		if (enable) {
+			if (RREG32_SOC15(VCN, i, regUVD_STATUS) != UVD_STATUS__IDLE)
+				return -EBUSY;
+			vcn_v4_0_5_enable_clock_gating(vinst);
+		} else {
+			vcn_v4_0_5_disable_clock_gating(vinst);
+		}
+	}
+
+	return 0;
+}
+
+static int vcn_v4_0_5_set_pg_state(struct amdgpu_vcn_inst *vinst,
+				   enum amd_powergating_state state)
+{
+	int ret = 0;
+
+	if (state == vinst->cur_state)
+		return 0;
+
+	if (state == AMD_PG_STATE_GATE)
+		ret = vcn_v4_0_5_stop(vinst);
+	else
+		ret = vcn_v4_0_5_start(vinst);
+
+	if (!ret)
+		vinst->cur_state = state;
+
+	return ret;
+}
+
+/**
+ * vcn_v4_0_5_process_interrupt - process VCN block interrupt
+ *
+ * @adev: amdgpu_device pointer
+ * @source: interrupt sources
+ * @entry: interrupt entry from clients and sources
+ *
+ * Process VCN block interrupt
+ */
+static int vcn_v4_0_5_process_interrupt(struct amdgpu_device *adev, struct amdgpu_irq_src *source,
+		struct amdgpu_iv_entry *entry)
+{
+	uint32_t ip_instance;
+
+	switch (entry->client_id) {
+	case SOC15_IH_CLIENTID_VCN:
+		ip_instance = 0;
+		break;
+	case SOC15_IH_CLIENTID_VCN1:
+		ip_instance = 1;
+		break;
+	default:
+		DRM_ERROR("Unhandled client id: %d\n", entry->client_id);
+		return 0;
+	}
+
+	DRM_DEBUG("IH: VCN TRAP\n");
+
+	switch (entry->src_id) {
+	case VCN_4_0__SRCID__UVD_ENC_GENERAL_PURPOSE:
+		amdgpu_fence_process(&adev->vcn.inst[ip_instance].ring_enc[0]);
+		break;
+	case VCN_4_0__SRCID_UVD_POISON:
+		amdgpu_vcn_process_poison_irq(adev, source, entry);
+		break;
+	default:
+		DRM_ERROR("Unhandled interrupt: %d %d\n",
+			  entry->src_id, entry->src_data[0]);
+		break;
+	}
+
+	return 0;
+}
+
+static const struct amdgpu_irq_src_funcs vcn_v4_0_5_irq_funcs = {
+	.process = vcn_v4_0_5_process_interrupt,
+};
+
+/**
+ * vcn_v4_0_5_set_irq_funcs - set VCN block interrupt irq functions
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set VCN block interrupt irq functions
+ */
+static void vcn_v4_0_5_set_irq_funcs(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		adev->vcn.inst[i].irq.num_types = adev->vcn.inst[i].num_enc_rings + 1;
+		adev->vcn.inst[i].irq.funcs = &vcn_v4_0_5_irq_funcs;
+	}
+}
+
+static const struct amd_ip_funcs vcn_v4_0_5_ip_funcs = {
+	.name = "vcn_v4_0_5",
+	.early_init = vcn_v4_0_5_early_init,
+	.sw_init = vcn_v4_0_5_sw_init,
+	.sw_fini = vcn_v4_0_5_sw_fini,
+	.hw_init = vcn_v4_0_5_hw_init,
+	.hw_fini = vcn_v4_0_5_hw_fini,
+	.suspend = vcn_v4_0_5_suspend,
+	.resume = vcn_v4_0_5_resume,
+	.is_idle = vcn_v4_0_5_is_idle,
+	.wait_for_idle = vcn_v4_0_5_wait_for_idle,
+	.set_clockgating_state = vcn_v4_0_5_set_clockgating_state,
+	.set_powergating_state = vcn_set_powergating_state,
+	.dump_ip_state = amdgpu_vcn_dump_ip_state,
+	.print_ip_state = amdgpu_vcn_print_ip_state,
+};
+
+const struct amdgpu_ip_block_version vcn_v4_0_5_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_VCN,
+	.major = 4,
+	.minor = 0,
+	.rev = 5,
+	.funcs = &vcn_v4_0_5_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.h b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.h
new file mode 100644
index 000000000000..ff9b3d6f6a47
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.h
@@ -0,0 +1,35 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __VCN_V4_0_5_H__
+#define __VCN_V4_0_5_H__
+
+enum amdgpu_vcn_v4_0_5_sub_block {
+	AMDGPU_VCN_V4_0_5_VCPU_VCODEC = 0,
+
+	AMDGPU_VCN_V4_0_5_MAX_SUB_BLOCK,
+};
+
+extern const struct amdgpu_ip_block_version vcn_v4_0_5_ip_block;
+
+#endif /* __VCN_V4_0_5_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_0.c
new file mode 100644
index 000000000000..0202df5db1e1
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_0.c
@@ -0,0 +1,1442 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/firmware.h>
+#include "amdgpu.h"
+#include "amdgpu_vcn.h"
+#include "amdgpu_pm.h"
+#include "soc15.h"
+#include "soc15d.h"
+#include "soc15_hw_ip.h"
+#include "vcn_v2_0.h"
+
+#include "vcn/vcn_5_0_0_offset.h"
+#include "vcn/vcn_5_0_0_sh_mask.h"
+#include "ivsrcid/vcn/irqsrcs_vcn_5_0.h"
+#include "vcn_v5_0_0.h"
+
+#include <drm/drm_drv.h>
+
+static const struct amdgpu_hwip_reg_entry vcn_reg_list_5_0[] = {
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_POWER_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_CONTEXT_ID),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_CONTEXT_ID2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_GPCOM_VCPU_DATA0),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_GPCOM_VCPU_DATA1),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_GPCOM_VCPU_CMD),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_LMA_CTL),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_LMA_DATA),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_LMA_MASK),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_PAUSE)
+};
+
+static int amdgpu_ih_clientid_vcns[] = {
+	SOC15_IH_CLIENTID_VCN,
+	SOC15_IH_CLIENTID_VCN1
+};
+
+static void vcn_v5_0_0_set_unified_ring_funcs(struct amdgpu_device *adev);
+static void vcn_v5_0_0_set_irq_funcs(struct amdgpu_device *adev);
+static int vcn_v5_0_0_set_pg_state(struct amdgpu_vcn_inst *vinst,
+				   enum amd_powergating_state state);
+static int vcn_v5_0_0_pause_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				     struct dpg_pause_state *new_state);
+static void vcn_v5_0_0_unified_ring_set_wptr(struct amdgpu_ring *ring);
+
+/**
+ * vcn_v5_0_0_early_init - set function pointers and load microcode
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Set ring and irq function pointers
+ * Load microcode from filesystem
+ */
+static int vcn_v5_0_0_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, r;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i)
+		/* re-use enc ring as unified ring */
+		adev->vcn.inst[i].num_enc_rings = 1;
+
+	vcn_v5_0_0_set_unified_ring_funcs(adev);
+	vcn_v5_0_0_set_irq_funcs(adev);
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		adev->vcn.inst[i].set_pg_state = vcn_v5_0_0_set_pg_state;
+
+		r = amdgpu_vcn_early_init(adev, i);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+/**
+ * vcn_v5_0_0_sw_init - sw init for VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Load firmware and sw initialization
+ */
+static int vcn_v5_0_0_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_ring *ring;
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, r;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		struct amdgpu_vcn5_fw_shared *fw_shared;
+
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		r = amdgpu_vcn_sw_init(adev, i);
+		if (r)
+			return r;
+
+		amdgpu_vcn_setup_ucode(adev, i);
+
+		r = amdgpu_vcn_resume(adev, i);
+		if (r)
+			return r;
+
+		atomic_set(&adev->vcn.inst[i].sched_score, 0);
+
+		/* VCN UNIFIED TRAP */
+		r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_vcns[i],
+				VCN_5_0__SRCID__UVD_ENC_GENERAL_PURPOSE, &adev->vcn.inst[i].irq);
+		if (r)
+			return r;
+
+		/* VCN POISON TRAP */
+		r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_vcns[i],
+				VCN_5_0__SRCID_UVD_POISON, &adev->vcn.inst[i].irq);
+		if (r)
+			return r;
+
+		ring = &adev->vcn.inst[i].ring_enc[0];
+		ring->use_doorbell = true;
+		ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 2 + 8 * i;
+
+		ring->vm_hub = AMDGPU_MMHUB0(0);
+		sprintf(ring->name, "vcn_unified_%d", i);
+
+		r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[i].irq, 0,
+						AMDGPU_RING_PRIO_0, &adev->vcn.inst[i].sched_score);
+		if (r)
+			return r;
+
+		fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
+		fw_shared->present_flag_0 = cpu_to_le32(AMDGPU_FW_SHARED_FLAG_0_UNIFIED_QUEUE);
+		fw_shared->sq.is_enabled = 1;
+
+		if (amdgpu_vcnfw_log)
+			amdgpu_vcn_fwlog_init(&adev->vcn.inst[i]);
+
+		if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
+			adev->vcn.inst[i].pause_dpg_mode = vcn_v5_0_0_pause_dpg_mode;
+	}
+
+	adev->vcn.supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->vcn.inst[0].ring_enc[0]);
+	if (!amdgpu_sriov_vf(adev))
+		adev->vcn.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+
+	r = amdgpu_vcn_reg_dump_init(adev, vcn_reg_list_5_0, ARRAY_SIZE(vcn_reg_list_5_0));
+	if (r)
+		return r;
+
+	r = amdgpu_vcn_sysfs_reset_mask_init(adev);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+/**
+ * vcn_v5_0_0_sw_fini - sw fini for VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * VCN suspend and free up sw allocation
+ */
+static int vcn_v5_0_0_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, r, idx;
+
+	if (drm_dev_enter(adev_to_drm(adev), &idx)) {
+		for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+			struct amdgpu_vcn5_fw_shared *fw_shared;
+
+			if (adev->vcn.harvest_config & (1 << i))
+				continue;
+
+			fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
+			fw_shared->present_flag_0 = 0;
+			fw_shared->sq.is_enabled = 0;
+		}
+
+		drm_dev_exit(idx);
+	}
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_suspend(adev, i);
+		if (r)
+			return r;
+	}
+
+	amdgpu_vcn_sysfs_reset_mask_fini(adev);
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++)
+		amdgpu_vcn_sw_fini(adev, i);
+
+	return 0;
+}
+
+/**
+ * vcn_v5_0_0_hw_init - start and test VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Initialize the hardware, boot up the VCPU and do some testing
+ */
+static int vcn_v5_0_0_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring;
+	int i, r;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		ring = &adev->vcn.inst[i].ring_enc[0];
+
+		adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell,
+			((adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 8 * i), i);
+
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+/**
+ * vcn_v5_0_0_hw_fini - stop the hardware block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Stop the VCN block, mark ring as not ready any more
+ */
+static int vcn_v5_0_0_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[i];
+
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		cancel_delayed_work_sync(&vinst->idle_work);
+
+		if (!amdgpu_sriov_vf(adev)) {
+			if ((adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) ||
+			    (vinst->cur_state != AMD_PG_STATE_GATE &&
+			     RREG32_SOC15(VCN, i, regUVD_STATUS))) {
+				vinst->set_pg_state(vinst, AMD_PG_STATE_GATE);
+			}
+		}
+	}
+
+	return 0;
+}
+
+/**
+ * vcn_v5_0_0_suspend - suspend VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * HW fini and suspend VCN block
+ */
+static int vcn_v5_0_0_suspend(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r, i;
+
+	r = vcn_v5_0_0_hw_fini(ip_block);
+	if (r)
+		return r;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_suspend(ip_block->adev, i);
+		if (r)
+			return r;
+	}
+
+	return r;
+}
+
+/**
+ * vcn_v5_0_0_resume - resume VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Resume firmware and hw init VCN block
+ */
+static int vcn_v5_0_0_resume(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r, i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_resume(ip_block->adev, i);
+		if (r)
+			return r;
+	}
+
+	r = vcn_v5_0_0_hw_init(ip_block);
+
+	return r;
+}
+
+/**
+ * vcn_v5_0_0_mc_resume - memory controller programming
+ *
+ * @vinst: VCN instance
+ *
+ * Let the VCN memory controller know it's offsets
+ */
+static void vcn_v5_0_0_mc_resume(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
+	uint32_t offset, size;
+	const struct common_firmware_header *hdr;
+
+	hdr = (const struct common_firmware_header *)adev->vcn.inst[inst].fw->data;
+	size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
+
+	/* cache window 0: fw */
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
+			(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst].tmr_mc_addr_lo));
+		WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
+			(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst].tmr_mc_addr_hi));
+		WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_OFFSET0, 0);
+		offset = 0;
+	} else {
+		WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
+			lower_32_bits(adev->vcn.inst[inst].gpu_addr));
+		WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
+			upper_32_bits(adev->vcn.inst[inst].gpu_addr));
+		offset = size;
+		WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_OFFSET0, AMDGPU_UVD_FIRMWARE_OFFSET >> 3);
+	}
+	WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_SIZE0, size);
+
+	/* cache window 1: stack */
+	WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW,
+		lower_32_bits(adev->vcn.inst[inst].gpu_addr + offset));
+	WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH,
+		upper_32_bits(adev->vcn.inst[inst].gpu_addr + offset));
+	WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_OFFSET1, 0);
+	WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_SIZE1, AMDGPU_VCN_STACK_SIZE);
+
+	/* cache window 2: context */
+	WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW,
+		lower_32_bits(adev->vcn.inst[inst].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE));
+	WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH,
+		upper_32_bits(adev->vcn.inst[inst].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE));
+	WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_OFFSET2, 0);
+	WREG32_SOC15(VCN, inst, regUVD_VCPU_CACHE_SIZE2, AMDGPU_VCN_CONTEXT_SIZE);
+
+	/* non-cache window */
+	WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_NC0_64BIT_BAR_LOW,
+		lower_32_bits(adev->vcn.inst[inst].fw_shared.gpu_addr));
+	WREG32_SOC15(VCN, inst, regUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH,
+		upper_32_bits(adev->vcn.inst[inst].fw_shared.gpu_addr));
+	WREG32_SOC15(VCN, inst, regUVD_VCPU_NONCACHE_OFFSET0, 0);
+	WREG32_SOC15(VCN, inst, regUVD_VCPU_NONCACHE_SIZE0,
+		AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn5_fw_shared)));
+}
+
+/**
+ * vcn_v5_0_0_mc_resume_dpg_mode - memory controller programming for dpg mode
+ *
+ * @vinst: VCN instance
+ * @indirect: indirectly write sram
+ *
+ * Let the VCN memory controller know it's offsets with dpg mode
+ */
+static void vcn_v5_0_0_mc_resume_dpg_mode(struct amdgpu_vcn_inst *vinst,
+					  bool indirect)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	uint32_t offset, size;
+	const struct common_firmware_header *hdr;
+
+	hdr = (const struct common_firmware_header *)adev->vcn.inst[inst_idx].fw->data;
+	size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
+
+	/* cache window 0: fw */
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		if (!indirect) {
+			WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+				VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW),
+				(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst_idx].tmr_mc_addr_lo), 0, indirect);
+			WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+				VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH),
+				(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst_idx].tmr_mc_addr_hi), 0, indirect);
+			WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+				VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET0), 0, 0, indirect);
+		} else {
+			WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+				VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW), 0, 0, indirect);
+			WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+				VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH), 0, 0, indirect);
+			WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+				VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET0), 0, 0, indirect);
+		}
+		offset = 0;
+	} else {
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW),
+			lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr), 0, indirect);
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH),
+			upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr), 0, indirect);
+		offset = size;
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET0),
+			AMDGPU_UVD_FIRMWARE_OFFSET >> 3, 0, indirect);
+	}
+
+	if (!indirect)
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_VCPU_CACHE_SIZE0), size, 0, indirect);
+	else
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_VCPU_CACHE_SIZE0), 0, 0, indirect);
+
+	/* cache window 1: stack */
+	if (!indirect) {
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW),
+			lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset), 0, indirect);
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH),
+			upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset), 0, indirect);
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET1), 0, 0, indirect);
+	} else {
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW), 0, 0, indirect);
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH), 0, 0, indirect);
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET1), 0, 0, indirect);
+	}
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, inst_idx, regUVD_VCPU_CACHE_SIZE1), AMDGPU_VCN_STACK_SIZE, 0, indirect);
+
+	/* cache window 2: context */
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW),
+		lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE), 0, indirect);
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH),
+		upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE), 0, indirect);
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_VCPU_CACHE_OFFSET2), 0, 0, indirect);
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_VCPU_CACHE_SIZE2), AMDGPU_VCN_CONTEXT_SIZE, 0, indirect);
+
+	/* non-cache window */
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_LMI_VCPU_NC0_64BIT_BAR_LOW),
+		lower_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr), 0, indirect);
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH),
+		upper_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr), 0, indirect);
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_VCPU_NONCACHE_OFFSET0), 0, 0, indirect);
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_VCPU_NONCACHE_SIZE0),
+		AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn5_fw_shared)), 0, indirect);
+
+	/* VCN global tiling registers */
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_GFX10_ADDR_CONFIG),
+		adev->gfx.config.gb_addr_config, 0, indirect);
+
+	return;
+}
+
+/**
+ * vcn_v5_0_0_disable_static_power_gating - disable VCN static power gating
+ *
+ * @vinst: VCN instance
+ *
+ * Disable static power gating for VCN block
+ */
+static void vcn_v5_0_0_disable_static_power_gating(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
+	uint32_t data = 0;
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN) {
+		data = 1 << UVD_IPX_DLDO_CONFIG__ONO2_PWR_CONFIG__SHIFT;
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG, data);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS, 0,
+				UVD_IPX_DLDO_STATUS__ONO2_PWR_STATUS_MASK);
+
+		data = 2 << UVD_IPX_DLDO_CONFIG__ONO3_PWR_CONFIG__SHIFT;
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG, data);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS,
+				1 << UVD_IPX_DLDO_STATUS__ONO3_PWR_STATUS__SHIFT,
+				UVD_IPX_DLDO_STATUS__ONO3_PWR_STATUS_MASK);
+
+		data = 2 << UVD_IPX_DLDO_CONFIG__ONO4_PWR_CONFIG__SHIFT;
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG, data);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS,
+				1 << UVD_IPX_DLDO_STATUS__ONO4_PWR_STATUS__SHIFT,
+				UVD_IPX_DLDO_STATUS__ONO4_PWR_STATUS_MASK);
+
+		data = 2 << UVD_IPX_DLDO_CONFIG__ONO5_PWR_CONFIG__SHIFT;
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG, data);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS,
+				1 << UVD_IPX_DLDO_STATUS__ONO5_PWR_STATUS__SHIFT,
+				UVD_IPX_DLDO_STATUS__ONO5_PWR_STATUS_MASK);
+	} else {
+		data = 1 << UVD_IPX_DLDO_CONFIG__ONO2_PWR_CONFIG__SHIFT;
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG, data);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS, 0,
+				UVD_IPX_DLDO_STATUS__ONO2_PWR_STATUS_MASK);
+
+		data = 1 << UVD_IPX_DLDO_CONFIG__ONO3_PWR_CONFIG__SHIFT;
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG, data);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS, 0,
+				UVD_IPX_DLDO_STATUS__ONO3_PWR_STATUS_MASK);
+
+		data = 1 << UVD_IPX_DLDO_CONFIG__ONO4_PWR_CONFIG__SHIFT;
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG, data);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS, 0,
+				UVD_IPX_DLDO_STATUS__ONO4_PWR_STATUS_MASK);
+
+		data = 1 << UVD_IPX_DLDO_CONFIG__ONO5_PWR_CONFIG__SHIFT;
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG, data);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS, 0,
+				UVD_IPX_DLDO_STATUS__ONO5_PWR_STATUS_MASK);
+	}
+
+	data = RREG32_SOC15(VCN, inst, regUVD_POWER_STATUS);
+	data &= ~0x103;
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN)
+		data |= UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON |
+			UVD_POWER_STATUS__UVD_PG_EN_MASK;
+
+	WREG32_SOC15(VCN, inst, regUVD_POWER_STATUS, data);
+	return;
+}
+
+/**
+ * vcn_v5_0_0_enable_static_power_gating - enable VCN static power gating
+ *
+ * @vinst: VCN instance
+ *
+ * Enable static power gating for VCN block
+ */
+static void vcn_v5_0_0_enable_static_power_gating(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
+	uint32_t data;
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN) {
+		/* Before power off, this indicator has to be turned on */
+		data = RREG32_SOC15(VCN, inst, regUVD_POWER_STATUS);
+		data &= ~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK;
+		data |= UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF;
+		WREG32_SOC15(VCN, inst, regUVD_POWER_STATUS, data);
+
+		data = 2 << UVD_IPX_DLDO_CONFIG__ONO5_PWR_CONFIG__SHIFT;
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG, data);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS,
+				1 << UVD_IPX_DLDO_STATUS__ONO5_PWR_STATUS__SHIFT,
+				UVD_IPX_DLDO_STATUS__ONO5_PWR_STATUS_MASK);
+
+		data = 2 << UVD_IPX_DLDO_CONFIG__ONO4_PWR_CONFIG__SHIFT;
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG, data);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS,
+				1 << UVD_IPX_DLDO_STATUS__ONO4_PWR_STATUS__SHIFT,
+				UVD_IPX_DLDO_STATUS__ONO4_PWR_STATUS_MASK);
+
+		data = 2 << UVD_IPX_DLDO_CONFIG__ONO3_PWR_CONFIG__SHIFT;
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG, data);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS,
+				1 << UVD_IPX_DLDO_STATUS__ONO3_PWR_STATUS__SHIFT,
+				UVD_IPX_DLDO_STATUS__ONO3_PWR_STATUS_MASK);
+
+		data = 2 << UVD_IPX_DLDO_CONFIG__ONO2_PWR_CONFIG__SHIFT;
+		WREG32_SOC15(VCN, inst, regUVD_IPX_DLDO_CONFIG, data);
+		SOC15_WAIT_ON_RREG(VCN, inst, regUVD_IPX_DLDO_STATUS,
+				1 << UVD_IPX_DLDO_STATUS__ONO2_PWR_STATUS__SHIFT,
+				UVD_IPX_DLDO_STATUS__ONO2_PWR_STATUS_MASK);
+	}
+	return;
+}
+
+/**
+ * vcn_v5_0_0_disable_clock_gating - disable VCN clock gating
+ *
+ * @vinst: VCN instance
+ *
+ * Disable clock gating for VCN block
+ */
+static void vcn_v5_0_0_disable_clock_gating(struct amdgpu_vcn_inst *vinst)
+{
+	return;
+}
+
+#if 0
+/**
+ * vcn_v5_0_0_disable_clock_gating_dpg_mode - disable VCN clock gating dpg mode
+ *
+ * @vinst: VCN instance
+ * @sram_sel: sram select
+ * @indirect: indirectly write sram
+ *
+ * Disable clock gating for VCN block with dpg mode
+ */
+static void vcn_v5_0_0_disable_clock_gating_dpg_mode(struct amdgpu_vcn_inst *vinst,
+						     uint8_t sram_sel,
+						     uint8_t indirect)
+{
+	return;
+}
+#endif
+
+/**
+ * vcn_v5_0_0_enable_clock_gating - enable VCN clock gating
+ *
+ * @vinst: VCN instance
+ *
+ * Enable clock gating for VCN block
+ */
+static void vcn_v5_0_0_enable_clock_gating(struct amdgpu_vcn_inst *vinst)
+{
+	return;
+}
+
+/**
+ * vcn_v5_0_0_start_dpg_mode - VCN start with dpg mode
+ *
+ * @vinst: VCN instance
+ * @indirect: indirectly write sram
+ *
+ * Start VCN block with dpg mode
+ */
+static int vcn_v5_0_0_start_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				     bool indirect)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	struct amdgpu_vcn5_fw_shared *fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr;
+	struct amdgpu_ring *ring;
+	uint32_t tmp;
+	int ret;
+
+	/* disable register anti-hang mechanism */
+	WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, regUVD_POWER_STATUS), 1,
+		~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
+
+	/* enable dynamic power gating mode */
+	tmp = RREG32_SOC15(VCN, inst_idx, regUVD_POWER_STATUS);
+	tmp |= UVD_POWER_STATUS__UVD_PG_MODE_MASK;
+	tmp |= UVD_POWER_STATUS__UVD_PG_EN_MASK;
+	WREG32_SOC15(VCN, inst_idx, regUVD_POWER_STATUS, tmp);
+
+	if (indirect)
+		adev->vcn.inst[inst_idx].dpg_sram_curr_addr = (uint32_t *)adev->vcn.inst[inst_idx].dpg_sram_cpu_addr;
+
+	/* enable VCPU clock */
+	tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
+	tmp |= UVD_VCPU_CNTL__CLK_EN_MASK | UVD_VCPU_CNTL__BLK_RST_MASK;
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_VCPU_CNTL), tmp, 0, indirect);
+
+	/* disable master interrupt */
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_MASTINT_EN), 0, 0, indirect);
+
+	/* setup regUVD_LMI_CTRL */
+	tmp = (UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
+		UVD_LMI_CTRL__REQ_MODE_MASK |
+		UVD_LMI_CTRL__CRC_RESET_MASK |
+		UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
+		UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
+		UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK |
+		(8 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) |
+		0x00100000L);
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_LMI_CTRL), tmp, 0, indirect);
+
+	vcn_v5_0_0_mc_resume_dpg_mode(vinst, indirect);
+
+	tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
+	tmp |= UVD_VCPU_CNTL__CLK_EN_MASK;
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_VCPU_CNTL), tmp, 0, indirect);
+
+	/* enable LMI MC and UMC channels */
+	tmp = 0x1f << UVD_LMI_CTRL2__RE_OFLD_MIF_WR_REQ_NUM__SHIFT;
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_LMI_CTRL2), tmp, 0, indirect);
+
+	/* enable master interrupt */
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, inst_idx, regUVD_MASTINT_EN),
+		UVD_MASTINT_EN__VCPU_EN_MASK, 0, indirect);
+
+	if (indirect) {
+		ret = amdgpu_vcn_psp_update_sram(adev, inst_idx, 0);
+		if (ret) {
+			dev_err(adev->dev, "%s: vcn sram load failed %d\n", __func__, ret);
+			return ret;
+		}
+	}
+
+	ring = &adev->vcn.inst[inst_idx].ring_enc[0];
+
+	WREG32_SOC15(VCN, inst_idx, regUVD_RB_BASE_LO, ring->gpu_addr);
+	WREG32_SOC15(VCN, inst_idx, regUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
+	WREG32_SOC15(VCN, inst_idx, regUVD_RB_SIZE, ring->ring_size / 4);
+
+	tmp = RREG32_SOC15(VCN, inst_idx, regVCN_RB_ENABLE);
+	tmp &= ~(VCN_RB_ENABLE__RB1_EN_MASK);
+	WREG32_SOC15(VCN, inst_idx, regVCN_RB_ENABLE, tmp);
+	fw_shared->sq.queue_mode |= FW_QUEUE_RING_RESET;
+	WREG32_SOC15(VCN, inst_idx, regUVD_RB_RPTR, 0);
+	WREG32_SOC15(VCN, inst_idx, regUVD_RB_WPTR, 0);
+
+	tmp = RREG32_SOC15(VCN, inst_idx, regUVD_RB_RPTR);
+	WREG32_SOC15(VCN, inst_idx, regUVD_RB_WPTR, tmp);
+	ring->wptr = RREG32_SOC15(VCN, inst_idx, regUVD_RB_WPTR);
+
+	tmp = RREG32_SOC15(VCN, inst_idx, regVCN_RB_ENABLE);
+	tmp |= VCN_RB_ENABLE__RB1_EN_MASK;
+	WREG32_SOC15(VCN, inst_idx, regVCN_RB_ENABLE, tmp);
+	fw_shared->sq.queue_mode &= ~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF);
+
+	WREG32_SOC15(VCN, inst_idx, regVCN_RB1_DB_CTRL,
+		ring->doorbell_index << VCN_RB1_DB_CTRL__OFFSET__SHIFT |
+		VCN_RB1_DB_CTRL__EN_MASK);
+
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, inst_idx, regUVD_STATUS);
+
+	return 0;
+}
+
+/**
+ * vcn_v5_0_0_start - VCN start
+ *
+ * @vinst: VCN instance
+ *
+ * Start VCN block
+ */
+static int vcn_v5_0_0_start(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int i = vinst->inst;
+	struct amdgpu_vcn5_fw_shared *fw_shared;
+	struct amdgpu_ring *ring;
+	uint32_t tmp;
+	int j, k, r;
+
+	if (adev->vcn.harvest_config & (1 << i))
+		return 0;
+
+	if (adev->pm.dpm_enabled)
+		amdgpu_dpm_enable_vcn(adev, true, i);
+
+	fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
+		return vcn_v5_0_0_start_dpg_mode(vinst, adev->vcn.inst[i].indirect_sram);
+
+	/* disable VCN power gating */
+	vcn_v5_0_0_disable_static_power_gating(vinst);
+
+	/* set VCN status busy */
+	tmp = RREG32_SOC15(VCN, i, regUVD_STATUS) | UVD_STATUS__UVD_BUSY;
+	WREG32_SOC15(VCN, i, regUVD_STATUS, tmp);
+
+	/* enable VCPU clock */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL),
+		 UVD_VCPU_CNTL__CLK_EN_MASK, ~UVD_VCPU_CNTL__CLK_EN_MASK);
+
+	/* disable master interrupt */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_MASTINT_EN), 0,
+		 ~UVD_MASTINT_EN__VCPU_EN_MASK);
+
+	/* enable LMI MC and UMC channels */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_LMI_CTRL2), 0,
+		 ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
+
+	tmp = RREG32_SOC15(VCN, i, regUVD_SOFT_RESET);
+	tmp &= ~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
+	tmp &= ~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
+	WREG32_SOC15(VCN, i, regUVD_SOFT_RESET, tmp);
+
+	/* setup regUVD_LMI_CTRL */
+	tmp = RREG32_SOC15(VCN, i, regUVD_LMI_CTRL);
+	WREG32_SOC15(VCN, i, regUVD_LMI_CTRL, tmp |
+		     UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
+		     UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
+		     UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
+		     UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK);
+
+	vcn_v5_0_0_mc_resume(vinst);
+
+	/* VCN global tiling registers */
+	WREG32_SOC15(VCN, i, regUVD_GFX10_ADDR_CONFIG,
+		     adev->gfx.config.gb_addr_config);
+
+	/* unblock VCPU register access */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_RB_ARB_CTRL), 0,
+		 ~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
+
+	/* release VCPU reset to boot */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), 0,
+		 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+
+	for (j = 0; j < 10; ++j) {
+		uint32_t status;
+
+		for (k = 0; k < 100; ++k) {
+			status = RREG32_SOC15(VCN, i, regUVD_STATUS);
+			if (status & 2)
+				break;
+			mdelay(10);
+			if (amdgpu_emu_mode == 1)
+				msleep(1);
+		}
+
+		if (amdgpu_emu_mode == 1) {
+			r = -1;
+			if (status & 2) {
+				r = 0;
+				break;
+			}
+		} else {
+			r = 0;
+			if (status & 2)
+				break;
+
+			dev_err(adev->dev,
+				"VCN[%d] is not responding, trying to reset the VCPU!!!\n", i);
+			WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL),
+				 UVD_VCPU_CNTL__BLK_RST_MASK,
+				 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+			mdelay(10);
+			WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), 0,
+				 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+
+			mdelay(10);
+			r = -1;
+		}
+	}
+
+	if (r) {
+		dev_err(adev->dev, "VCN[%d] is not responding, giving up!!!\n", i);
+		return r;
+	}
+
+	/* enable master interrupt */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_MASTINT_EN),
+		 UVD_MASTINT_EN__VCPU_EN_MASK,
+		 ~UVD_MASTINT_EN__VCPU_EN_MASK);
+
+	/* clear the busy bit of VCN_STATUS */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_STATUS), 0,
+		 ~(2 << UVD_STATUS__VCPU_REPORT__SHIFT));
+
+	ring = &adev->vcn.inst[i].ring_enc[0];
+	WREG32_SOC15(VCN, i, regVCN_RB1_DB_CTRL,
+		     ring->doorbell_index << VCN_RB1_DB_CTRL__OFFSET__SHIFT |
+		     VCN_RB1_DB_CTRL__EN_MASK);
+
+	WREG32_SOC15(VCN, i, regUVD_RB_BASE_LO, ring->gpu_addr);
+	WREG32_SOC15(VCN, i, regUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
+	WREG32_SOC15(VCN, i, regUVD_RB_SIZE, ring->ring_size / 4);
+
+	tmp = RREG32_SOC15(VCN, i, regVCN_RB_ENABLE);
+	tmp &= ~(VCN_RB_ENABLE__RB1_EN_MASK);
+	WREG32_SOC15(VCN, i, regVCN_RB_ENABLE, tmp);
+	fw_shared->sq.queue_mode |= FW_QUEUE_RING_RESET;
+	WREG32_SOC15(VCN, i, regUVD_RB_RPTR, 0);
+	WREG32_SOC15(VCN, i, regUVD_RB_WPTR, 0);
+
+	tmp = RREG32_SOC15(VCN, i, regUVD_RB_RPTR);
+	WREG32_SOC15(VCN, i, regUVD_RB_WPTR, tmp);
+	ring->wptr = RREG32_SOC15(VCN, i, regUVD_RB_WPTR);
+
+	tmp = RREG32_SOC15(VCN, i, regVCN_RB_ENABLE);
+	tmp |= VCN_RB_ENABLE__RB1_EN_MASK;
+	WREG32_SOC15(VCN, i, regVCN_RB_ENABLE, tmp);
+	fw_shared->sq.queue_mode &= ~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF);
+
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, i, regUVD_STATUS);
+
+	return 0;
+}
+
+/**
+ * vcn_v5_0_0_stop_dpg_mode - VCN stop with dpg mode
+ *
+ * @vinst: VCN instance
+ *
+ * Stop VCN block with dpg mode
+ */
+static void vcn_v5_0_0_stop_dpg_mode(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	struct dpg_pause_state state = {.fw_based = VCN_DPG_STATE__UNPAUSE};
+	uint32_t tmp;
+
+	vcn_v5_0_0_pause_dpg_mode(vinst, &state);
+
+	/* Wait for power status to be 1 */
+	SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_POWER_STATUS, 1,
+		UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
+
+	/* wait for read ptr to be equal to write ptr */
+	tmp = RREG32_SOC15(VCN, inst_idx, regUVD_RB_WPTR);
+	SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_RB_RPTR, tmp, 0xFFFFFFFF);
+
+	/* disable dynamic power gating mode */
+	WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, regUVD_POWER_STATUS), 0,
+		~UVD_POWER_STATUS__UVD_PG_MODE_MASK);
+
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, inst_idx, regUVD_STATUS);
+
+	return;
+}
+
+/**
+ * vcn_v5_0_0_stop - VCN stop
+ *
+ * @vinst: VCN instance
+ *
+ * Stop VCN block
+ */
+static int vcn_v5_0_0_stop(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int i = vinst->inst;
+	struct amdgpu_vcn5_fw_shared *fw_shared;
+	uint32_t tmp;
+	int r = 0;
+
+	if (adev->vcn.harvest_config & (1 << i))
+		return 0;
+
+	fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
+	fw_shared->sq.queue_mode |= FW_QUEUE_DPG_HOLD_OFF;
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
+		vcn_v5_0_0_stop_dpg_mode(vinst);
+		r = 0;
+		goto done;
+	}
+
+	/* wait for vcn idle */
+	r = SOC15_WAIT_ON_RREG(VCN, i, regUVD_STATUS, UVD_STATUS__IDLE, 0x7);
+	if (r)
+		goto done;
+
+	tmp = UVD_LMI_STATUS__VCPU_LMI_WRITE_CLEAN_MASK |
+		UVD_LMI_STATUS__READ_CLEAN_MASK |
+		UVD_LMI_STATUS__WRITE_CLEAN_MASK |
+		UVD_LMI_STATUS__WRITE_CLEAN_RAW_MASK;
+	r = SOC15_WAIT_ON_RREG(VCN, i, regUVD_LMI_STATUS, tmp, tmp);
+	if (r)
+		goto done;
+
+	/* disable LMI UMC channel */
+	tmp = RREG32_SOC15(VCN, i, regUVD_LMI_CTRL2);
+	tmp |= UVD_LMI_CTRL2__STALL_ARB_UMC_MASK;
+	WREG32_SOC15(VCN, i, regUVD_LMI_CTRL2, tmp);
+	tmp = UVD_LMI_STATUS__UMC_READ_CLEAN_RAW_MASK |
+		UVD_LMI_STATUS__UMC_WRITE_CLEAN_RAW_MASK;
+	r = SOC15_WAIT_ON_RREG(VCN, i, regUVD_LMI_STATUS, tmp, tmp);
+	if (r)
+		goto done;
+
+	/* block VCPU register access */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_RB_ARB_CTRL),
+		 UVD_RB_ARB_CTRL__VCPU_DIS_MASK,
+		 ~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
+
+	/* reset VCPU */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL),
+		 UVD_VCPU_CNTL__BLK_RST_MASK,
+		 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+
+	/* disable VCPU clock */
+	WREG32_P(SOC15_REG_OFFSET(VCN, i, regUVD_VCPU_CNTL), 0,
+		 ~(UVD_VCPU_CNTL__CLK_EN_MASK));
+
+	/* apply soft reset */
+	tmp = RREG32_SOC15(VCN, i, regUVD_SOFT_RESET);
+	tmp |= UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
+	WREG32_SOC15(VCN, i, regUVD_SOFT_RESET, tmp);
+	tmp = RREG32_SOC15(VCN, i, regUVD_SOFT_RESET);
+	tmp |= UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
+	WREG32_SOC15(VCN, i, regUVD_SOFT_RESET, tmp);
+
+	/* clear status */
+	WREG32_SOC15(VCN, i, regUVD_STATUS, 0);
+
+	/* enable VCN power gating */
+	vcn_v5_0_0_enable_static_power_gating(vinst);
+
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, i, regUVD_STATUS);
+
+done:
+	if (adev->pm.dpm_enabled)
+		amdgpu_dpm_enable_vcn(adev, false, i);
+
+	return r;
+}
+
+/**
+ * vcn_v5_0_0_pause_dpg_mode - VCN pause with dpg mode
+ *
+ * @vinst: VCN instance
+ * @new_state: pause state
+ *
+ * Pause dpg mode for VCN block
+ */
+static int vcn_v5_0_0_pause_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				     struct dpg_pause_state *new_state)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	uint32_t reg_data = 0;
+	int ret_code;
+
+	/* pause/unpause if state is changed */
+	if (adev->vcn.inst[inst_idx].pause_state.fw_based != new_state->fw_based) {
+		DRM_DEV_DEBUG(adev->dev, "dpg pause state changed %d -> %d",
+			adev->vcn.inst[inst_idx].pause_state.fw_based,  new_state->fw_based);
+		reg_data = RREG32_SOC15(VCN, inst_idx, regUVD_DPG_PAUSE) &
+			(~UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK);
+
+		if (new_state->fw_based == VCN_DPG_STATE__PAUSE) {
+			ret_code = SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_POWER_STATUS, 0x1,
+					UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
+
+			if (!ret_code) {
+				/* pause DPG */
+				reg_data |= UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK;
+				WREG32_SOC15(VCN, inst_idx, regUVD_DPG_PAUSE, reg_data);
+
+				/* wait for ACK */
+				SOC15_WAIT_ON_RREG(VCN, inst_idx, regUVD_DPG_PAUSE,
+					UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK,
+					UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK);
+			}
+		} else {
+			/* unpause dpg, no need to wait */
+			reg_data &= ~UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK;
+			WREG32_SOC15(VCN, inst_idx, regUVD_DPG_PAUSE, reg_data);
+		}
+		adev->vcn.inst[inst_idx].pause_state.fw_based = new_state->fw_based;
+	}
+
+	return 0;
+}
+
+/**
+ * vcn_v5_0_0_unified_ring_get_rptr - get unified read pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware unified read pointer
+ */
+static uint64_t vcn_v5_0_0_unified_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring != &adev->vcn.inst[ring->me].ring_enc[0])
+		DRM_ERROR("wrong ring id is identified in %s", __func__);
+
+	return RREG32_SOC15(VCN, ring->me, regUVD_RB_RPTR);
+}
+
+/**
+ * vcn_v5_0_0_unified_ring_get_wptr - get unified write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware unified write pointer
+ */
+static uint64_t vcn_v5_0_0_unified_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring != &adev->vcn.inst[ring->me].ring_enc[0])
+		DRM_ERROR("wrong ring id is identified in %s", __func__);
+
+	if (ring->use_doorbell)
+		return *ring->wptr_cpu_addr;
+	else
+		return RREG32_SOC15(VCN, ring->me, regUVD_RB_WPTR);
+}
+
+/**
+ * vcn_v5_0_0_unified_ring_set_wptr - set enc write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Commits the enc write pointer to the hardware
+ */
+static void vcn_v5_0_0_unified_ring_set_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring != &adev->vcn.inst[ring->me].ring_enc[0])
+		DRM_ERROR("wrong ring id is identified in %s", __func__);
+
+	if (ring->use_doorbell) {
+		*ring->wptr_cpu_addr = lower_32_bits(ring->wptr);
+		WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr));
+	} else {
+		WREG32_SOC15(VCN, ring->me, regUVD_RB_WPTR, lower_32_bits(ring->wptr));
+	}
+}
+
+static int vcn_v5_0_0_ring_reset(struct amdgpu_ring *ring,
+				 unsigned int vmid,
+				 struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[ring->me];
+	int r;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+	r = vcn_v5_0_0_stop(vinst);
+	if (r)
+		return r;
+	r = vcn_v5_0_0_start(vinst);
+	if (r)
+		return r;
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static const struct amdgpu_ring_funcs vcn_v5_0_0_unified_ring_vm_funcs = {
+	.type = AMDGPU_RING_TYPE_VCN_ENC,
+	.align_mask = 0x3f,
+	.nop = VCN_ENC_CMD_NO_OP,
+	.get_rptr = vcn_v5_0_0_unified_ring_get_rptr,
+	.get_wptr = vcn_v5_0_0_unified_ring_get_wptr,
+	.set_wptr = vcn_v5_0_0_unified_ring_set_wptr,
+	.emit_frame_size =
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 4 +
+		4 + /* vcn_v2_0_enc_ring_emit_vm_flush */
+		5 + 5 + /* vcn_v2_0_enc_ring_emit_fence x2 vm fence */
+		1, /* vcn_v2_0_enc_ring_insert_end */
+	.emit_ib_size = 5, /* vcn_v2_0_enc_ring_emit_ib */
+	.emit_ib = vcn_v2_0_enc_ring_emit_ib,
+	.emit_fence = vcn_v2_0_enc_ring_emit_fence,
+	.emit_vm_flush = vcn_v2_0_enc_ring_emit_vm_flush,
+	.test_ring = amdgpu_vcn_enc_ring_test_ring,
+	.test_ib = amdgpu_vcn_unified_ring_test_ib,
+	.insert_nop = amdgpu_ring_insert_nop,
+	.insert_end = vcn_v2_0_enc_ring_insert_end,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.begin_use = amdgpu_vcn_ring_begin_use,
+	.end_use = amdgpu_vcn_ring_end_use,
+	.emit_wreg = vcn_v2_0_enc_ring_emit_wreg,
+	.emit_reg_wait = vcn_v2_0_enc_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = vcn_v5_0_0_ring_reset,
+};
+
+/**
+ * vcn_v5_0_0_set_unified_ring_funcs - set unified ring functions
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set unified ring functions
+ */
+static void vcn_v5_0_0_set_unified_ring_funcs(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		adev->vcn.inst[i].ring_enc[0].funcs = &vcn_v5_0_0_unified_ring_vm_funcs;
+		adev->vcn.inst[i].ring_enc[0].me = i;
+	}
+}
+
+/**
+ * vcn_v5_0_0_is_idle - check VCN block is idle
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block structure
+ *
+ * Check whether VCN block is idle
+ */
+static bool vcn_v5_0_0_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, ret = 1;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		ret &= (RREG32_SOC15(VCN, i, regUVD_STATUS) == UVD_STATUS__IDLE);
+	}
+
+	return ret;
+}
+
+/**
+ * vcn_v5_0_0_wait_for_idle - wait for VCN block idle
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Wait for VCN block idle
+ */
+static int vcn_v5_0_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, ret = 0;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		ret = SOC15_WAIT_ON_RREG(VCN, i, regUVD_STATUS, UVD_STATUS__IDLE,
+			UVD_STATUS__IDLE);
+		if (ret)
+			return ret;
+	}
+
+	return ret;
+}
+
+/**
+ * vcn_v5_0_0_set_clockgating_state - set VCN block clockgating state
+ *
+ * @ip_block: amdgpu_ip_block pointer
+ * @state: clock gating state
+ *
+ * Set VCN block clockgating state
+ */
+static int vcn_v5_0_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_clockgating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	bool enable = state == AMD_CG_STATE_GATE;
+	int i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[i];
+
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		if (enable) {
+			if (RREG32_SOC15(VCN, i, regUVD_STATUS) != UVD_STATUS__IDLE)
+				return -EBUSY;
+			vcn_v5_0_0_enable_clock_gating(vinst);
+		} else {
+			vcn_v5_0_0_disable_clock_gating(vinst);
+		}
+	}
+
+	return 0;
+}
+
+static int vcn_v5_0_0_set_pg_state(struct amdgpu_vcn_inst *vinst,
+				   enum amd_powergating_state state)
+{
+	int ret = 0;
+
+	if (state == vinst->cur_state)
+		return 0;
+
+	if (state == AMD_PG_STATE_GATE)
+		ret = vcn_v5_0_0_stop(vinst);
+	else
+		ret = vcn_v5_0_0_start(vinst);
+
+	if (!ret)
+		vinst->cur_state = state;
+
+	return ret;
+}
+
+/**
+ * vcn_v5_0_0_process_interrupt - process VCN block interrupt
+ *
+ * @adev: amdgpu_device pointer
+ * @source: interrupt sources
+ * @entry: interrupt entry from clients and sources
+ *
+ * Process VCN block interrupt
+ */
+static int vcn_v5_0_0_process_interrupt(struct amdgpu_device *adev, struct amdgpu_irq_src *source,
+	struct amdgpu_iv_entry *entry)
+{
+	uint32_t ip_instance;
+
+	switch (entry->client_id) {
+	case SOC15_IH_CLIENTID_VCN:
+		ip_instance = 0;
+		break;
+	case SOC15_IH_CLIENTID_VCN1:
+		ip_instance = 1;
+		break;
+	default:
+		DRM_ERROR("Unhandled client id: %d\n", entry->client_id);
+		return 0;
+	}
+
+	DRM_DEBUG("IH: VCN TRAP\n");
+
+	switch (entry->src_id) {
+	case VCN_5_0__SRCID__UVD_ENC_GENERAL_PURPOSE:
+		amdgpu_fence_process(&adev->vcn.inst[ip_instance].ring_enc[0]);
+		break;
+	case VCN_5_0__SRCID_UVD_POISON:
+		amdgpu_vcn_process_poison_irq(adev, source, entry);
+		break;
+	default:
+		DRM_ERROR("Unhandled interrupt: %d %d\n",
+			  entry->src_id, entry->src_data[0]);
+		break;
+	}
+
+	return 0;
+}
+
+static const struct amdgpu_irq_src_funcs vcn_v5_0_0_irq_funcs = {
+	.process = vcn_v5_0_0_process_interrupt,
+};
+
+/**
+ * vcn_v5_0_0_set_irq_funcs - set VCN block interrupt irq functions
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set VCN block interrupt irq functions
+ */
+static void vcn_v5_0_0_set_irq_funcs(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		adev->vcn.inst[i].irq.num_types = adev->vcn.inst[i].num_enc_rings + 1;
+		adev->vcn.inst[i].irq.funcs = &vcn_v5_0_0_irq_funcs;
+	}
+}
+
+static const struct amd_ip_funcs vcn_v5_0_0_ip_funcs = {
+	.name = "vcn_v5_0_0",
+	.early_init = vcn_v5_0_0_early_init,
+	.sw_init = vcn_v5_0_0_sw_init,
+	.sw_fini = vcn_v5_0_0_sw_fini,
+	.hw_init = vcn_v5_0_0_hw_init,
+	.hw_fini = vcn_v5_0_0_hw_fini,
+	.suspend = vcn_v5_0_0_suspend,
+	.resume = vcn_v5_0_0_resume,
+	.is_idle = vcn_v5_0_0_is_idle,
+	.wait_for_idle = vcn_v5_0_0_wait_for_idle,
+	.set_clockgating_state = vcn_v5_0_0_set_clockgating_state,
+	.set_powergating_state = vcn_set_powergating_state,
+	.dump_ip_state = amdgpu_vcn_dump_ip_state,
+	.print_ip_state = amdgpu_vcn_print_ip_state,
+};
+
+const struct amdgpu_ip_block_version vcn_v5_0_0_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_VCN,
+	.major = 5,
+	.minor = 0,
+	.rev = 0,
+	.funcs = &vcn_v5_0_0_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_0.h b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_0.h
new file mode 100644
index 000000000000..51bbccd4360f
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_0.h
@@ -0,0 +1,37 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __VCN_V5_0_0_H__
+#define __VCN_V5_0_0_H__
+
+#define VCN_VID_SOC_ADDRESS                        0x1FC00
+#define VCN_AON_SOC_ADDRESS                        0x1F800
+#define VCN1_VID_SOC_ADDRESS                       0x48300
+#define VCN1_AON_SOC_ADDRESS                       0x48000
+
+#define VCN_VID_IP_ADDRESS                         0x0
+#define VCN_AON_IP_ADDRESS                         0x30000
+
+extern const struct amdgpu_ip_block_version vcn_v5_0_0_ip_block;
+
+#endif /* __VCN_V5_0_0_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c
new file mode 100644
index 000000000000..8bd457dea4cf
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c
@@ -0,0 +1,1729 @@
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/firmware.h>
+#include "amdgpu.h"
+#include "amdgpu_vcn.h"
+#include "amdgpu_pm.h"
+#include "soc15.h"
+#include "soc15d.h"
+#include "soc15_hw_ip.h"
+#include "vcn_v2_0.h"
+#include "vcn_v4_0_3.h"
+#include "mmsch_v5_0.h"
+
+#include "vcn/vcn_5_0_0_offset.h"
+#include "vcn/vcn_5_0_0_sh_mask.h"
+#include "ivsrcid/vcn/irqsrcs_vcn_5_0.h"
+#include "vcn_v5_0_0.h"
+#include "vcn_v5_0_1.h"
+
+#include <drm/drm_drv.h>
+
+static const struct amdgpu_hwip_reg_entry vcn_reg_list_5_0_1[] = {
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_POWER_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_STATUS),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_CONTEXT_ID),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_CONTEXT_ID2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_GPCOM_VCPU_DATA0),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_GPCOM_VCPU_DATA1),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_GPCOM_VCPU_CMD),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_HI4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_BASE_LO4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_RPTR4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_WPTR4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE2),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE3),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_RB_SIZE4),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_LMA_CTL),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_LMA_DATA),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_LMA_MASK),
+	SOC15_REG_ENTRY_STR(VCN, 0, regUVD_DPG_PAUSE)
+};
+
+static int vcn_v5_0_1_start_sriov(struct amdgpu_device *adev);
+static void vcn_v5_0_1_set_unified_ring_funcs(struct amdgpu_device *adev);
+static void vcn_v5_0_1_set_irq_funcs(struct amdgpu_device *adev);
+static int vcn_v5_0_1_set_pg_state(struct amdgpu_vcn_inst *vinst,
+				   enum amd_powergating_state state);
+static void vcn_v5_0_1_unified_ring_set_wptr(struct amdgpu_ring *ring);
+static void vcn_v5_0_1_set_ras_funcs(struct amdgpu_device *adev);
+/**
+ * vcn_v5_0_1_early_init - set function pointers and load microcode
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Set ring and irq function pointers
+ * Load microcode from filesystem
+ */
+static int vcn_v5_0_1_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, r;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i)
+		/* re-use enc ring as unified ring */
+		adev->vcn.inst[i].num_enc_rings = 1;
+
+	vcn_v5_0_1_set_unified_ring_funcs(adev);
+	vcn_v5_0_1_set_irq_funcs(adev);
+	vcn_v5_0_1_set_ras_funcs(adev);
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		adev->vcn.inst[i].set_pg_state = vcn_v5_0_1_set_pg_state;
+
+		r = amdgpu_vcn_early_init(adev, i);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+static int vcn_v5_0_1_late_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	adev->vcn.supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->vcn.inst[0].ring_enc[0]);
+
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
+	case IP_VERSION(13, 0, 12):
+		if ((adev->psp.sos.fw_version >= 0x00450025) &&
+			amdgpu_dpm_reset_vcn_is_supported(adev) &&
+			!amdgpu_sriov_vf(adev))
+			adev->vcn.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static void vcn_v5_0_1_fw_shared_init(struct amdgpu_device *adev, int inst_idx)
+{
+	struct amdgpu_vcn5_fw_shared *fw_shared;
+
+	fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr;
+
+	if (fw_shared->sq.is_enabled)
+		return;
+	fw_shared->present_flag_0 =
+		cpu_to_le32(AMDGPU_FW_SHARED_FLAG_0_UNIFIED_QUEUE);
+	fw_shared->sq.is_enabled = 1;
+
+	if (amdgpu_vcnfw_log)
+		amdgpu_vcn_fwlog_init(&adev->vcn.inst[inst_idx]);
+}
+
+/**
+ * vcn_v5_0_1_sw_init - sw init for VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Load firmware and sw initialization
+ */
+static int vcn_v5_0_1_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring;
+	int i, r, vcn_inst;
+
+	/* VCN UNIFIED TRAP */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
+		VCN_5_0__SRCID__UVD_ENC_GENERAL_PURPOSE, &adev->vcn.inst->irq);
+	if (r)
+		return r;
+
+	/* VCN POISON TRAP */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
+		VCN_5_0__SRCID_UVD_POISON, &adev->vcn.inst->ras_poison_irq);
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		vcn_inst = GET_INST(VCN, i);
+
+		r = amdgpu_vcn_sw_init(adev, i);
+		if (r)
+			return r;
+
+		amdgpu_vcn_setup_ucode(adev, i);
+
+		r = amdgpu_vcn_resume(adev, i);
+		if (r)
+			return r;
+
+		ring = &adev->vcn.inst[i].ring_enc[0];
+		ring->use_doorbell = true;
+		if (!amdgpu_sriov_vf(adev))
+			ring->doorbell_index =
+				(adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
+				11 * vcn_inst;
+		else
+			ring->doorbell_index =
+				(adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
+				32 * vcn_inst;
+
+		ring->vm_hub = AMDGPU_MMHUB0(adev->vcn.inst[i].aid_id);
+		sprintf(ring->name, "vcn_unified_%d", adev->vcn.inst[i].aid_id);
+
+		r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[i].irq, 0,
+					AMDGPU_RING_PRIO_DEFAULT, &adev->vcn.inst[i].sched_score);
+		if (r)
+			return r;
+
+		vcn_v5_0_1_fw_shared_init(adev, i);
+	}
+
+	if (amdgpu_sriov_vf(adev)) {
+		r = amdgpu_virt_alloc_mm_table(adev);
+		if (r)
+			return r;
+	}
+
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__VCN)) {
+		r = amdgpu_vcn_ras_sw_init(adev);
+		if (r) {
+			dev_err(adev->dev, "Failed to initialize vcn ras block!\n");
+			return r;
+		}
+	}
+
+	r = amdgpu_vcn_reg_dump_init(adev, vcn_reg_list_5_0_1, ARRAY_SIZE(vcn_reg_list_5_0_1));
+	if (r)
+		return r;
+
+	return amdgpu_vcn_sysfs_reset_mask_init(adev);
+}
+
+/**
+ * vcn_v5_0_1_sw_fini - sw fini for VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * VCN suspend and free up sw allocation
+ */
+static int vcn_v5_0_1_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, r, idx;
+
+	if (drm_dev_enter(adev_to_drm(adev), &idx)) {
+		for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+			struct amdgpu_vcn5_fw_shared *fw_shared;
+
+			fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
+			fw_shared->present_flag_0 = 0;
+			fw_shared->sq.is_enabled = 0;
+		}
+
+		drm_dev_exit(idx);
+	}
+
+	if (amdgpu_sriov_vf(adev))
+		amdgpu_virt_free_mm_table(adev);
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_suspend(adev, i);
+		if (r)
+			return r;
+	}
+
+	amdgpu_vcn_sysfs_reset_mask_fini(adev);
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++)
+		amdgpu_vcn_sw_fini(adev, i);
+
+	return 0;
+}
+
+static int vcn_v5_0_1_hw_init_inst(struct amdgpu_device *adev, int i)
+{
+	struct amdgpu_ring *ring;
+	int vcn_inst;
+
+	vcn_inst = GET_INST(VCN, i);
+	ring = &adev->vcn.inst[i].ring_enc[0];
+
+	if (ring->use_doorbell)
+		adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell,
+			((adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
+			11 * vcn_inst),
+			adev->vcn.inst[i].aid_id);
+
+	return 0;
+}
+
+/**
+ * vcn_v5_0_1_hw_init - start and test VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Initialize the hardware, boot up the VCPU and do some testing
+ */
+static int vcn_v5_0_1_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_ring *ring;
+	int i, r;
+
+	if (amdgpu_sriov_vf(adev)) {
+		r = vcn_v5_0_1_start_sriov(adev);
+		if (r)
+			return r;
+
+		for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+			ring = &adev->vcn.inst[i].ring_enc[0];
+			ring->wptr = 0;
+			ring->wptr_old = 0;
+			vcn_v5_0_1_unified_ring_set_wptr(ring);
+			ring->sched.ready = true;
+		}
+	} else {
+		if (RREG32_SOC15(VCN, GET_INST(VCN, 0), regVCN_RRMT_CNTL) & 0x100)
+			adev->vcn.caps |= AMDGPU_VCN_CAPS(RRMT_ENABLED);
+		for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+			ring = &adev->vcn.inst[i].ring_enc[0];
+			vcn_v5_0_1_hw_init_inst(adev, i);
+
+			/* Re-init fw_shared, if required */
+			vcn_v5_0_1_fw_shared_init(adev, i);
+
+			r = amdgpu_ring_test_helper(ring);
+			if (r)
+				return r;
+		}
+	}
+
+	return 0;
+}
+
+/**
+ * vcn_v5_0_1_hw_fini - stop the hardware block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Stop the VCN block, mark ring as not ready any more
+ */
+static int vcn_v5_0_1_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[i];
+
+		cancel_delayed_work_sync(&adev->vcn.inst[i].idle_work);
+		if (vinst->cur_state != AMD_PG_STATE_GATE)
+			vinst->set_pg_state(vinst, AMD_PG_STATE_GATE);
+	}
+
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__VCN) && !amdgpu_sriov_vf(adev))
+		amdgpu_irq_put(adev, &adev->vcn.inst->ras_poison_irq, 0);
+
+	return 0;
+}
+
+/**
+ * vcn_v5_0_1_suspend - suspend VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * HW fini and suspend VCN block
+ */
+static int vcn_v5_0_1_suspend(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r, i;
+
+	r = vcn_v5_0_1_hw_fini(ip_block);
+	if (r)
+		return r;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		r = amdgpu_vcn_suspend(ip_block->adev, i);
+		if (r)
+			return r;
+	}
+
+	return r;
+}
+
+/**
+ * vcn_v5_0_1_resume - resume VCN block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Resume firmware and hw init VCN block
+ */
+static int vcn_v5_0_1_resume(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r, i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[i];
+
+		if (amdgpu_in_reset(adev))
+			vinst->cur_state = AMD_PG_STATE_GATE;
+
+		r = amdgpu_vcn_resume(ip_block->adev, i);
+		if (r)
+			return r;
+	}
+
+	r = vcn_v5_0_1_hw_init(ip_block);
+
+	return r;
+}
+
+/**
+ * vcn_v5_0_1_mc_resume - memory controller programming
+ *
+ * @vinst: VCN instance
+ *
+ * Let the VCN memory controller know it's offsets
+ */
+static void vcn_v5_0_1_mc_resume(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst = vinst->inst;
+	uint32_t offset, size, vcn_inst;
+	const struct common_firmware_header *hdr;
+
+	hdr = (const struct common_firmware_header *)adev->vcn.inst[inst].fw->data;
+	size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
+
+	vcn_inst = GET_INST(VCN, inst);
+	/* cache window 0: fw */
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
+			(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst].tmr_mc_addr_lo));
+		WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
+			(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + inst].tmr_mc_addr_hi));
+		WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_OFFSET0, 0);
+		offset = 0;
+	} else {
+		WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
+			lower_32_bits(adev->vcn.inst[inst].gpu_addr));
+		WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
+			upper_32_bits(adev->vcn.inst[inst].gpu_addr));
+		offset = size;
+		WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_OFFSET0,
+				AMDGPU_UVD_FIRMWARE_OFFSET >> 3);
+	}
+	WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_SIZE0, size);
+
+	/* cache window 1: stack */
+	WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW,
+		lower_32_bits(adev->vcn.inst[inst].gpu_addr + offset));
+	WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH,
+		upper_32_bits(adev->vcn.inst[inst].gpu_addr + offset));
+	WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_OFFSET1, 0);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_SIZE1, AMDGPU_VCN_STACK_SIZE);
+
+	/* cache window 2: context */
+	WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW,
+		lower_32_bits(adev->vcn.inst[inst].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE));
+	WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH,
+		upper_32_bits(adev->vcn.inst[inst].gpu_addr + offset + AMDGPU_VCN_STACK_SIZE));
+	WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_OFFSET2, 0);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_CACHE_SIZE2, AMDGPU_VCN_CONTEXT_SIZE);
+
+	/* non-cache window */
+	WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_NC0_64BIT_BAR_LOW,
+		lower_32_bits(adev->vcn.inst[inst].fw_shared.gpu_addr));
+	WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH,
+		upper_32_bits(adev->vcn.inst[inst].fw_shared.gpu_addr));
+	WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_NONCACHE_OFFSET0, 0);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_VCPU_NONCACHE_SIZE0,
+		AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn5_fw_shared)));
+}
+
+/**
+ * vcn_v5_0_1_mc_resume_dpg_mode - memory controller programming for dpg mode
+ *
+ * @vinst: VCN instance
+ * @indirect: indirectly write sram
+ *
+ * Let the VCN memory controller know it's offsets with dpg mode
+ */
+static void vcn_v5_0_1_mc_resume_dpg_mode(struct amdgpu_vcn_inst *vinst,
+					  bool indirect)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	uint32_t offset, size;
+	const struct common_firmware_header *hdr;
+
+	hdr = (const struct common_firmware_header *)adev->vcn.inst[inst_idx].fw->data;
+	size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
+
+	/* cache window 0: fw */
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		if (!indirect) {
+			WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+				VCN, 0, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW),
+				(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN +
+				 inst_idx].tmr_mc_addr_lo), 0, indirect);
+			WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+				VCN, 0, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH),
+				(adev->firmware.ucode[AMDGPU_UCODE_ID_VCN +
+				 inst_idx].tmr_mc_addr_hi), 0, indirect);
+			WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+				VCN, 0, regUVD_VCPU_CACHE_OFFSET0), 0, 0, indirect);
+		} else {
+			WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+				VCN, 0, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW), 0, 0, indirect);
+			WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+				VCN, 0, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH), 0, 0, indirect);
+			WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+				VCN, 0, regUVD_VCPU_CACHE_OFFSET0), 0, 0, indirect);
+		}
+		offset = 0;
+	} else {
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW),
+			lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr), 0, indirect);
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH),
+			upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr), 0, indirect);
+		offset = size;
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_VCPU_CACHE_OFFSET0),
+			AMDGPU_UVD_FIRMWARE_OFFSET >> 3, 0, indirect);
+	}
+
+	if (!indirect)
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_VCPU_CACHE_SIZE0), size, 0, indirect);
+	else
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_VCPU_CACHE_SIZE0), 0, 0, indirect);
+
+	/* cache window 1: stack */
+	if (!indirect) {
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW),
+			lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset), 0, indirect);
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH),
+			upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset), 0, indirect);
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_VCPU_CACHE_OFFSET1), 0, 0, indirect);
+	} else {
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW), 0, 0, indirect);
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH), 0, 0, indirect);
+		WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_VCPU_CACHE_OFFSET1), 0, 0, indirect);
+	}
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+			VCN, 0, regUVD_VCPU_CACHE_SIZE1), AMDGPU_VCN_STACK_SIZE, 0, indirect);
+
+	/* cache window 2: context */
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW),
+		lower_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset +
+			AMDGPU_VCN_STACK_SIZE), 0, indirect);
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH),
+		upper_32_bits(adev->vcn.inst[inst_idx].gpu_addr + offset +
+			AMDGPU_VCN_STACK_SIZE), 0, indirect);
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_VCPU_CACHE_OFFSET2), 0, 0, indirect);
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_VCPU_CACHE_SIZE2), AMDGPU_VCN_CONTEXT_SIZE, 0, indirect);
+
+	/* non-cache window */
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_LMI_VCPU_NC0_64BIT_BAR_LOW),
+		lower_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr), 0, indirect);
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH),
+		upper_32_bits(adev->vcn.inst[inst_idx].fw_shared.gpu_addr), 0, indirect);
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_VCPU_NONCACHE_OFFSET0), 0, 0, indirect);
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_VCPU_NONCACHE_SIZE0),
+		AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn5_fw_shared)), 0, indirect);
+
+	/* VCN global tiling registers */
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_GFX10_ADDR_CONFIG), adev->gfx.config.gb_addr_config, 0, indirect);
+}
+
+/**
+ * vcn_v5_0_1_disable_clock_gating - disable VCN clock gating
+ *
+ * @vinst: VCN instance
+ *
+ * Disable clock gating for VCN block
+ */
+static void vcn_v5_0_1_disable_clock_gating(struct amdgpu_vcn_inst *vinst)
+{
+}
+
+/**
+ * vcn_v5_0_1_enable_clock_gating - enable VCN clock gating
+ *
+ * @vinst: VCN instance
+ *
+ * Enable clock gating for VCN block
+ */
+static void vcn_v5_0_1_enable_clock_gating(struct amdgpu_vcn_inst *vinst)
+{
+}
+
+/**
+ * vcn_v5_0_1_pause_dpg_mode - VCN pause with dpg mode
+ *
+ * @vinst: VCN instance
+ * @new_state: pause state
+ *
+ * Pause dpg mode for VCN block
+ */
+static int vcn_v5_0_1_pause_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				     struct dpg_pause_state *new_state)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	uint32_t reg_data = 0;
+	int vcn_inst;
+
+	vcn_inst = GET_INST(VCN, vinst->inst);
+
+	/* pause/unpause if state is changed */
+	if (vinst->pause_state.fw_based != new_state->fw_based) {
+		DRM_DEV_DEBUG(adev->dev, "dpg pause state changed %d -> %d %s\n",
+			vinst->pause_state.fw_based, new_state->fw_based,
+			new_state->fw_based ? "VCN_DPG_STATE__PAUSE" : "VCN_DPG_STATE__UNPAUSE");
+		reg_data = RREG32_SOC15(VCN, vcn_inst, regUVD_DPG_PAUSE) &
+			(~UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK);
+
+		if (new_state->fw_based == VCN_DPG_STATE__PAUSE) {
+			/* pause DPG */
+			reg_data |= UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK;
+			WREG32_SOC15(VCN, vcn_inst, regUVD_DPG_PAUSE, reg_data);
+
+			/* wait for ACK */
+			SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_DPG_PAUSE,
+					UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK,
+					UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK);
+		} else {
+			/* unpause DPG, no need to wait */
+			reg_data &= ~UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK;
+			WREG32_SOC15(VCN, vcn_inst, regUVD_DPG_PAUSE, reg_data);
+		}
+		vinst->pause_state.fw_based = new_state->fw_based;
+	}
+
+	return 0;
+}
+
+
+/**
+ * vcn_v5_0_1_start_dpg_mode - VCN start with dpg mode
+ *
+ * @vinst: VCN instance
+ * @indirect: indirectly write sram
+ *
+ * Start VCN block with dpg mode
+ */
+static int vcn_v5_0_1_start_dpg_mode(struct amdgpu_vcn_inst *vinst,
+				     bool indirect)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	struct amdgpu_vcn5_fw_shared *fw_shared =
+		adev->vcn.inst[inst_idx].fw_shared.cpu_addr;
+	struct amdgpu_ring *ring;
+	struct dpg_pause_state state = {.fw_based = VCN_DPG_STATE__PAUSE};
+	int vcn_inst, ret;
+	uint32_t tmp;
+
+	vcn_inst = GET_INST(VCN, inst_idx);
+
+	/* disable register anti-hang mechanism */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_POWER_STATUS), 1,
+		~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
+
+	/* enable dynamic power gating mode */
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_POWER_STATUS);
+	tmp |= UVD_POWER_STATUS__UVD_PG_MODE_MASK;
+	WREG32_SOC15(VCN, vcn_inst, regUVD_POWER_STATUS, tmp);
+
+	if (indirect) {
+		adev->vcn.inst[inst_idx].dpg_sram_curr_addr =
+			(uint32_t *)adev->vcn.inst[inst_idx].dpg_sram_cpu_addr;
+		/* Use dummy register 0xDEADBEEF passing AID selection to PSP FW */
+		WREG32_SOC24_DPG_MODE(inst_idx, 0xDEADBEEF,
+				adev->vcn.inst[inst_idx].aid_id, 0, true);
+	}
+
+	/* enable VCPU clock */
+	tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
+	tmp |= UVD_VCPU_CNTL__CLK_EN_MASK | UVD_VCPU_CNTL__BLK_RST_MASK;
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_VCPU_CNTL), tmp, 0, indirect);
+
+	/* disable master interrupt */
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_MASTINT_EN), 0, 0, indirect);
+
+	/* setup regUVD_LMI_CTRL */
+	tmp = (UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
+		UVD_LMI_CTRL__REQ_MODE_MASK |
+		UVD_LMI_CTRL__CRC_RESET_MASK |
+		UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
+		UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
+		UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK |
+		(8 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) |
+		0x00100000L);
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_LMI_CTRL), tmp, 0, indirect);
+
+	vcn_v5_0_1_mc_resume_dpg_mode(vinst, indirect);
+
+	tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
+	tmp |= UVD_VCPU_CNTL__CLK_EN_MASK;
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_VCPU_CNTL), tmp, 0, indirect);
+
+	/* enable LMI MC and UMC channels */
+	tmp = 0x1f << UVD_LMI_CTRL2__RE_OFLD_MIF_WR_REQ_NUM__SHIFT;
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_LMI_CTRL2), tmp, 0, indirect);
+
+	/* enable master interrupt */
+	WREG32_SOC24_DPG_MODE(inst_idx, SOC24_DPG_MODE_OFFSET(
+		VCN, 0, regUVD_MASTINT_EN),
+		UVD_MASTINT_EN__VCPU_EN_MASK, 0, indirect);
+
+	if (indirect) {
+		ret = amdgpu_vcn_psp_update_sram(adev, inst_idx, AMDGPU_UCODE_ID_VCN0_RAM);
+		if (ret) {
+			dev_err(adev->dev, "vcn sram load failed %d\n", ret);
+			return ret;
+		}
+	}
+
+	/* resetting ring, fw should not check RB ring */
+	fw_shared->sq.queue_mode |= FW_QUEUE_RING_RESET;
+
+	/* Pause dpg */
+	vcn_v5_0_1_pause_dpg_mode(vinst, &state);
+
+	ring = &adev->vcn.inst[inst_idx].ring_enc[0];
+
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_BASE_LO, lower_32_bits(ring->gpu_addr));
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_SIZE, ring->ring_size / sizeof(uint32_t));
+
+	tmp = RREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE);
+	tmp &= ~(VCN_RB_ENABLE__RB1_EN_MASK);
+	WREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE, tmp);
+
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_RPTR, 0);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_WPTR, 0);
+
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_RB_RPTR);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_WPTR, tmp);
+	ring->wptr = RREG32_SOC15(VCN, vcn_inst, regUVD_RB_WPTR);
+
+	tmp = RREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE);
+	tmp |= VCN_RB_ENABLE__RB1_EN_MASK;
+	WREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE, tmp);
+	/* resetting done, fw can check RB ring */
+	fw_shared->sq.queue_mode &= ~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF);
+
+	WREG32_SOC15(VCN, vcn_inst, regVCN_RB1_DB_CTRL,
+		ring->doorbell_index << VCN_RB1_DB_CTRL__OFFSET__SHIFT |
+		VCN_RB1_DB_CTRL__EN_MASK);
+	/* Read DB_CTRL to flush the write DB_CTRL command. */
+	RREG32_SOC15(VCN, vcn_inst, regVCN_RB1_DB_CTRL);
+
+	return 0;
+}
+
+static int vcn_v5_0_1_start_sriov(struct amdgpu_device *adev)
+{
+	int i, vcn_inst;
+	struct amdgpu_ring *ring_enc;
+	uint64_t cache_addr;
+	uint64_t rb_enc_addr;
+	uint64_t ctx_addr;
+	uint32_t param, resp, expected;
+	uint32_t offset, cache_size;
+	uint32_t tmp, timeout;
+
+	struct amdgpu_mm_table *table = &adev->virt.mm_table;
+	uint32_t *table_loc;
+	uint32_t table_size;
+	uint32_t size, size_dw;
+	uint32_t init_status;
+	uint32_t enabled_vcn;
+
+	struct mmsch_v5_0_cmd_direct_write
+		direct_wt = { {0} };
+	struct mmsch_v5_0_cmd_direct_read_modify_write
+		direct_rd_mod_wt = { {0} };
+	struct mmsch_v5_0_cmd_end end = { {0} };
+	struct mmsch_v5_0_init_header header;
+
+	struct amdgpu_vcn5_fw_shared *fw_shared;
+	struct amdgpu_fw_shared_rb_setup *rb_setup;
+
+	direct_wt.cmd_header.command_type =
+		MMSCH_COMMAND__DIRECT_REG_WRITE;
+	direct_rd_mod_wt.cmd_header.command_type =
+		MMSCH_COMMAND__DIRECT_REG_READ_MODIFY_WRITE;
+	end.cmd_header.command_type = MMSCH_COMMAND__END;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		vcn_inst = GET_INST(VCN, i);
+
+		vcn_v5_0_1_fw_shared_init(adev, vcn_inst);
+
+		memset(&header, 0, sizeof(struct mmsch_v5_0_init_header));
+		header.version = MMSCH_VERSION;
+		header.total_size = sizeof(struct mmsch_v5_0_init_header) >> 2;
+
+		table_loc = (uint32_t *)table->cpu_addr;
+		table_loc += header.total_size;
+
+		table_size = 0;
+
+		MMSCH_V5_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCN, 0, regUVD_STATUS),
+			~UVD_STATUS__UVD_BUSY, UVD_STATUS__UVD_BUSY);
+
+		cache_size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.inst[i].fw->size + 4);
+
+		if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+			MMSCH_V5_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+				regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW),
+				adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + i].tmr_mc_addr_lo);
+
+			MMSCH_V5_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+				regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH),
+				adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + i].tmr_mc_addr_hi);
+
+			offset = 0;
+			MMSCH_V5_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+				regUVD_VCPU_CACHE_OFFSET0), 0);
+		} else {
+			MMSCH_V5_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+				regUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW),
+				lower_32_bits(adev->vcn.inst[i].gpu_addr));
+			MMSCH_V5_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+				regUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH),
+				upper_32_bits(adev->vcn.inst[i].gpu_addr));
+			offset = cache_size;
+			MMSCH_V5_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+				regUVD_VCPU_CACHE_OFFSET0),
+				AMDGPU_UVD_FIRMWARE_OFFSET >> 3);
+		}
+
+		MMSCH_V5_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_VCPU_CACHE_SIZE0),
+			cache_size);
+
+		cache_addr = adev->vcn.inst[vcn_inst].gpu_addr + offset;
+		MMSCH_V5_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW), lower_32_bits(cache_addr));
+		MMSCH_V5_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH), upper_32_bits(cache_addr));
+		MMSCH_V5_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_VCPU_CACHE_OFFSET1), 0);
+		MMSCH_V5_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_VCPU_CACHE_SIZE1), AMDGPU_VCN_STACK_SIZE);
+
+		cache_addr = adev->vcn.inst[vcn_inst].gpu_addr + offset +
+			AMDGPU_VCN_STACK_SIZE;
+
+		MMSCH_V5_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW), lower_32_bits(cache_addr));
+
+		MMSCH_V5_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH), upper_32_bits(cache_addr));
+
+		MMSCH_V5_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_VCPU_CACHE_OFFSET2), 0);
+
+		MMSCH_V5_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_VCPU_CACHE_SIZE2), AMDGPU_VCN_CONTEXT_SIZE);
+
+		fw_shared = adev->vcn.inst[vcn_inst].fw_shared.cpu_addr;
+		rb_setup = &fw_shared->rb_setup;
+
+		ring_enc = &adev->vcn.inst[vcn_inst].ring_enc[0];
+		ring_enc->wptr = 0;
+		rb_enc_addr = ring_enc->gpu_addr;
+
+		rb_setup->is_rb_enabled_flags |= RB_ENABLED;
+		rb_setup->rb_addr_lo = lower_32_bits(rb_enc_addr);
+		rb_setup->rb_addr_hi = upper_32_bits(rb_enc_addr);
+		rb_setup->rb_size = ring_enc->ring_size / 4;
+		fw_shared->present_flag_0 |= cpu_to_le32(AMDGPU_VCN_VF_RB_SETUP_FLAG);
+
+		MMSCH_V5_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_LMI_VCPU_NC0_64BIT_BAR_LOW),
+			lower_32_bits(adev->vcn.inst[vcn_inst].fw_shared.gpu_addr));
+		MMSCH_V5_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_LMI_VCPU_NC0_64BIT_BAR_HIGH),
+			upper_32_bits(adev->vcn.inst[vcn_inst].fw_shared.gpu_addr));
+		MMSCH_V5_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
+			regUVD_VCPU_NONCACHE_SIZE0),
+			AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn4_fw_shared)));
+		MMSCH_V5_0_INSERT_END();
+
+		header.vcn0.init_status = 0;
+		header.vcn0.table_offset = header.total_size;
+		header.vcn0.table_size = table_size;
+		header.total_size += table_size;
+
+		/* Send init table to mmsch */
+		size = sizeof(struct mmsch_v5_0_init_header);
+		table_loc = (uint32_t *)table->cpu_addr;
+		memcpy((void *)table_loc, &header, size);
+
+		ctx_addr = table->gpu_addr;
+		WREG32_SOC15(VCN, vcn_inst, regMMSCH_VF_CTX_ADDR_LO, lower_32_bits(ctx_addr));
+		WREG32_SOC15(VCN, vcn_inst, regMMSCH_VF_CTX_ADDR_HI, upper_32_bits(ctx_addr));
+
+		tmp = RREG32_SOC15(VCN, vcn_inst, regMMSCH_VF_VMID);
+		tmp &= ~MMSCH_VF_VMID__VF_CTX_VMID_MASK;
+		tmp |= (0 << MMSCH_VF_VMID__VF_CTX_VMID__SHIFT);
+		WREG32_SOC15(VCN, vcn_inst, regMMSCH_VF_VMID, tmp);
+
+		size = header.total_size;
+		WREG32_SOC15(VCN, vcn_inst, regMMSCH_VF_CTX_SIZE, size);
+
+		WREG32_SOC15(VCN, vcn_inst, regMMSCH_VF_MAILBOX_RESP, 0);
+
+		param = 0x00000001;
+		WREG32_SOC15(VCN, vcn_inst, regMMSCH_VF_MAILBOX_HOST, param);
+		tmp = 0;
+		timeout = 1000;
+		resp = 0;
+		expected = MMSCH_VF_MAILBOX_RESP__OK;
+		while (resp != expected) {
+			resp = RREG32_SOC15(VCN, vcn_inst, regMMSCH_VF_MAILBOX_RESP);
+			if (resp != 0)
+				break;
+
+			udelay(10);
+			tmp = tmp + 10;
+			if (tmp >= timeout) {
+				DRM_ERROR("failed to init MMSCH. TIME-OUT after %d usec"\
+					" waiting for regMMSCH_VF_MAILBOX_RESP "\
+					"(expected=0x%08x, readback=0x%08x)\n",
+					tmp, expected, resp);
+				return -EBUSY;
+			}
+		}
+
+		enabled_vcn = amdgpu_vcn_is_disabled_vcn(adev, VCN_DECODE_RING, 0) ? 1 : 0;
+		init_status = ((struct mmsch_v5_0_init_header *)(table_loc))->vcn0.init_status;
+		if (resp != expected && resp != MMSCH_VF_MAILBOX_RESP__INCOMPLETE
+					&& init_status != MMSCH_VF_ENGINE_STATUS__PASS) {
+			DRM_ERROR("MMSCH init status is incorrect! readback=0x%08x, header init "\
+				"status for VCN%x: 0x%x\n", resp, enabled_vcn, init_status);
+		}
+	}
+
+	return 0;
+}
+
+/**
+ * vcn_v5_0_1_start - VCN start
+ *
+ * @vinst: VCN instance
+ *
+ * Start VCN block
+ */
+static int vcn_v5_0_1_start(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int i = vinst->inst;
+	struct amdgpu_vcn5_fw_shared *fw_shared;
+	struct amdgpu_ring *ring;
+	uint32_t tmp;
+	int j, k, r, vcn_inst;
+
+	fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
+		return vcn_v5_0_1_start_dpg_mode(vinst, adev->vcn.inst[i].indirect_sram);
+
+	vcn_inst = GET_INST(VCN, i);
+
+	/* set VCN status busy */
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_STATUS) | UVD_STATUS__UVD_BUSY;
+	WREG32_SOC15(VCN, vcn_inst, regUVD_STATUS, tmp);
+
+	/* enable VCPU clock */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_VCPU_CNTL),
+		 UVD_VCPU_CNTL__CLK_EN_MASK, ~UVD_VCPU_CNTL__CLK_EN_MASK);
+
+	/* disable master interrupt */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_MASTINT_EN), 0,
+		 ~UVD_MASTINT_EN__VCPU_EN_MASK);
+
+	/* enable LMI MC and UMC channels */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_LMI_CTRL2), 0,
+		 ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
+
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_SOFT_RESET);
+	tmp &= ~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
+	tmp &= ~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
+	WREG32_SOC15(VCN, vcn_inst, regUVD_SOFT_RESET, tmp);
+
+	/* setup regUVD_LMI_CTRL */
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_LMI_CTRL);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_CTRL, tmp |
+		     UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
+		     UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
+		     UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
+		     UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK);
+
+	vcn_v5_0_1_mc_resume(vinst);
+
+	/* VCN global tiling registers */
+	WREG32_SOC15(VCN, vcn_inst, regUVD_GFX10_ADDR_CONFIG,
+		     adev->gfx.config.gb_addr_config);
+
+	/* unblock VCPU register access */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_RB_ARB_CTRL), 0,
+		 ~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
+
+	/* release VCPU reset to boot */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_VCPU_CNTL), 0,
+		 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+
+	for (j = 0; j < 10; ++j) {
+		uint32_t status;
+
+		for (k = 0; k < 100; ++k) {
+			status = RREG32_SOC15(VCN, vcn_inst, regUVD_STATUS);
+			if (status & 2)
+				break;
+			mdelay(100);
+			if (amdgpu_emu_mode == 1)
+				msleep(20);
+		}
+
+		if (amdgpu_emu_mode == 1) {
+			r = -1;
+			if (status & 2) {
+				r = 0;
+				break;
+			}
+		} else {
+			r = 0;
+			if (status & 2)
+				break;
+
+			dev_err(adev->dev,
+				"VCN[%d] is not responding, trying to reset the VCPU!!!\n", i);
+			WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_VCPU_CNTL),
+				 UVD_VCPU_CNTL__BLK_RST_MASK,
+				 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+			mdelay(10);
+			WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_VCPU_CNTL), 0,
+				 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+
+			mdelay(10);
+			r = -1;
+		}
+	}
+
+	if (r) {
+		dev_err(adev->dev, "VCN[%d] is not responding, giving up!!!\n", i);
+		return r;
+	}
+
+	/* enable master interrupt */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_MASTINT_EN),
+		 UVD_MASTINT_EN__VCPU_EN_MASK,
+		 ~UVD_MASTINT_EN__VCPU_EN_MASK);
+
+	/* clear the busy bit of VCN_STATUS */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_STATUS), 0,
+		 ~(2 << UVD_STATUS__VCPU_REPORT__SHIFT));
+
+	ring = &adev->vcn.inst[i].ring_enc[0];
+
+	WREG32_SOC15(VCN, vcn_inst, regVCN_RB1_DB_CTRL,
+		     ring->doorbell_index << VCN_RB1_DB_CTRL__OFFSET__SHIFT |
+		     VCN_RB1_DB_CTRL__EN_MASK);
+
+	/* Read DB_CTRL to flush the write DB_CTRL command. */
+	RREG32_SOC15(VCN, vcn_inst, regVCN_RB1_DB_CTRL);
+
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_BASE_LO, ring->gpu_addr);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_SIZE, ring->ring_size / 4);
+
+	tmp = RREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE);
+	tmp &= ~(VCN_RB_ENABLE__RB1_EN_MASK);
+	WREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE, tmp);
+	fw_shared->sq.queue_mode |= FW_QUEUE_RING_RESET;
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_RPTR, 0);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_WPTR, 0);
+
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_RB_RPTR);
+	WREG32_SOC15(VCN, vcn_inst, regUVD_RB_WPTR, tmp);
+	ring->wptr = RREG32_SOC15(VCN, vcn_inst, regUVD_RB_WPTR);
+
+	tmp = RREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE);
+	tmp |= VCN_RB_ENABLE__RB1_EN_MASK;
+	WREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE, tmp);
+	fw_shared->sq.queue_mode &= ~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF);
+
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, vcn_inst, regUVD_STATUS);
+
+	return 0;
+}
+
+/**
+ * vcn_v5_0_1_stop_dpg_mode - VCN stop with dpg mode
+ *
+ * @vinst: VCN instance
+ *
+ * Stop VCN block with dpg mode
+ */
+static void vcn_v5_0_1_stop_dpg_mode(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int inst_idx = vinst->inst;
+	uint32_t tmp;
+	int vcn_inst;
+	struct dpg_pause_state state = {.fw_based = VCN_DPG_STATE__UNPAUSE};
+
+	vcn_inst = GET_INST(VCN, inst_idx);
+
+	/* Unpause dpg */
+	vcn_v5_0_1_pause_dpg_mode(vinst, &state);
+
+	/* Wait for power status to be 1 */
+	SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_POWER_STATUS, 1,
+		UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
+
+	/* wait for read ptr to be equal to write ptr */
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_RB_WPTR);
+	SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_RB_RPTR, tmp, 0xFFFFFFFF);
+
+	/* disable dynamic power gating mode */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_POWER_STATUS), 0,
+		~UVD_POWER_STATUS__UVD_PG_MODE_MASK);
+
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, vcn_inst, regUVD_STATUS);
+}
+
+/**
+ * vcn_v5_0_1_stop - VCN stop
+ *
+ * @vinst: VCN instance
+ *
+ * Stop VCN block
+ */
+static int vcn_v5_0_1_stop(struct amdgpu_vcn_inst *vinst)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int i = vinst->inst;
+	struct amdgpu_vcn5_fw_shared *fw_shared;
+	uint32_t tmp;
+	int r = 0, vcn_inst;
+
+	vcn_inst = GET_INST(VCN, i);
+
+	fw_shared = adev->vcn.inst[i].fw_shared.cpu_addr;
+	fw_shared->sq.queue_mode |= FW_QUEUE_DPG_HOLD_OFF;
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
+		vcn_v5_0_1_stop_dpg_mode(vinst);
+		return 0;
+	}
+
+	/* wait for vcn idle */
+	r = SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_STATUS, UVD_STATUS__IDLE, 0x7);
+	if (r)
+		return r;
+
+	tmp = UVD_LMI_STATUS__VCPU_LMI_WRITE_CLEAN_MASK |
+		UVD_LMI_STATUS__READ_CLEAN_MASK |
+		UVD_LMI_STATUS__WRITE_CLEAN_MASK |
+		UVD_LMI_STATUS__WRITE_CLEAN_RAW_MASK;
+	r = SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_LMI_STATUS, tmp, tmp);
+	if (r)
+		return r;
+
+	/* disable LMI UMC channel */
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_LMI_CTRL2);
+	tmp |= UVD_LMI_CTRL2__STALL_ARB_UMC_MASK;
+	WREG32_SOC15(VCN, vcn_inst, regUVD_LMI_CTRL2, tmp);
+	tmp = UVD_LMI_STATUS__UMC_READ_CLEAN_RAW_MASK |
+		UVD_LMI_STATUS__UMC_WRITE_CLEAN_RAW_MASK;
+	r = SOC15_WAIT_ON_RREG(VCN, vcn_inst, regUVD_LMI_STATUS, tmp, tmp);
+	if (r)
+		return r;
+
+	/* block VCPU register access */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_RB_ARB_CTRL),
+		 UVD_RB_ARB_CTRL__VCPU_DIS_MASK,
+		 ~UVD_RB_ARB_CTRL__VCPU_DIS_MASK);
+
+	/* reset VCPU */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_VCPU_CNTL),
+		 UVD_VCPU_CNTL__BLK_RST_MASK,
+		 ~UVD_VCPU_CNTL__BLK_RST_MASK);
+
+	/* disable VCPU clock */
+	WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_VCPU_CNTL), 0,
+		 ~(UVD_VCPU_CNTL__CLK_EN_MASK));
+
+	/* apply soft reset */
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_SOFT_RESET);
+	tmp |= UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
+	WREG32_SOC15(VCN, vcn_inst, regUVD_SOFT_RESET, tmp);
+	tmp = RREG32_SOC15(VCN, vcn_inst, regUVD_SOFT_RESET);
+	tmp |= UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
+	WREG32_SOC15(VCN, vcn_inst, regUVD_SOFT_RESET, tmp);
+
+	/* clear status */
+	WREG32_SOC15(VCN, vcn_inst, regUVD_STATUS, 0);
+
+	/* Keeping one read-back to ensure all register writes are done,
+	 * otherwise it may introduce race conditions.
+	 */
+	RREG32_SOC15(VCN, vcn_inst, regUVD_STATUS);
+
+	return 0;
+}
+
+/**
+ * vcn_v5_0_1_unified_ring_get_rptr - get unified read pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware unified read pointer
+ */
+static uint64_t vcn_v5_0_1_unified_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring != &adev->vcn.inst[ring->me].ring_enc[0])
+		DRM_ERROR("wrong ring id is identified in %s", __func__);
+
+	return RREG32_SOC15(VCN, GET_INST(VCN, ring->me), regUVD_RB_RPTR);
+}
+
+/**
+ * vcn_v5_0_1_unified_ring_get_wptr - get unified write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware unified write pointer
+ */
+static uint64_t vcn_v5_0_1_unified_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring != &adev->vcn.inst[ring->me].ring_enc[0])
+		DRM_ERROR("wrong ring id is identified in %s", __func__);
+
+	if (ring->use_doorbell)
+		return *ring->wptr_cpu_addr;
+	else
+		return RREG32_SOC15(VCN, GET_INST(VCN, ring->me), regUVD_RB_WPTR);
+}
+
+/**
+ * vcn_v5_0_1_unified_ring_set_wptr - set enc write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Commits the enc write pointer to the hardware
+ */
+static void vcn_v5_0_1_unified_ring_set_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring != &adev->vcn.inst[ring->me].ring_enc[0])
+		DRM_ERROR("wrong ring id is identified in %s", __func__);
+
+	if (ring->use_doorbell) {
+		*ring->wptr_cpu_addr = lower_32_bits(ring->wptr);
+		WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr));
+	} else {
+		WREG32_SOC15(VCN, GET_INST(VCN, ring->me), regUVD_RB_WPTR,
+				lower_32_bits(ring->wptr));
+	}
+}
+
+static int vcn_v5_0_1_ring_reset(struct amdgpu_ring *ring,
+				 unsigned int vmid,
+				 struct amdgpu_fence *timedout_fence)
+{
+	int r = 0;
+	int vcn_inst;
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[ring->me];
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	vcn_inst = GET_INST(VCN, ring->me);
+	r = amdgpu_dpm_reset_vcn(adev, 1 << vcn_inst);
+
+	if (r) {
+		DRM_DEV_ERROR(adev->dev, "VCN reset fail : %d\n", r);
+		return r;
+	}
+
+	vcn_v5_0_1_hw_init_inst(adev, ring->me);
+	vcn_v5_0_1_start_dpg_mode(vinst, vinst->indirect_sram);
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static const struct amdgpu_ring_funcs vcn_v5_0_1_unified_ring_vm_funcs = {
+	.type = AMDGPU_RING_TYPE_VCN_ENC,
+	.align_mask = 0x3f,
+	.nop = VCN_ENC_CMD_NO_OP,
+	.get_rptr = vcn_v5_0_1_unified_ring_get_rptr,
+	.get_wptr = vcn_v5_0_1_unified_ring_get_wptr,
+	.set_wptr = vcn_v5_0_1_unified_ring_set_wptr,
+	.emit_frame_size = SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 +
+			   SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 4 +
+			   4 + /* vcn_v2_0_enc_ring_emit_vm_flush */
+			   5 +
+			   5 + /* vcn_v2_0_enc_ring_emit_fence x2 vm fence */
+			   1, /* vcn_v2_0_enc_ring_insert_end */
+	.emit_ib_size = 5, /* vcn_v2_0_enc_ring_emit_ib */
+	.emit_ib = vcn_v2_0_enc_ring_emit_ib,
+	.emit_fence = vcn_v2_0_enc_ring_emit_fence,
+	.emit_vm_flush = vcn_v4_0_3_enc_ring_emit_vm_flush,
+	.emit_hdp_flush = vcn_v4_0_3_ring_emit_hdp_flush,
+	.test_ring = amdgpu_vcn_enc_ring_test_ring,
+	.test_ib = amdgpu_vcn_unified_ring_test_ib,
+	.insert_nop = amdgpu_ring_insert_nop,
+	.insert_end = vcn_v2_0_enc_ring_insert_end,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.begin_use = amdgpu_vcn_ring_begin_use,
+	.end_use = amdgpu_vcn_ring_end_use,
+	.emit_wreg = vcn_v4_0_3_enc_ring_emit_wreg,
+	.emit_reg_wait = vcn_v4_0_3_enc_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.reset = vcn_v5_0_1_ring_reset,
+};
+
+/**
+ * vcn_v5_0_1_set_unified_ring_funcs - set unified ring functions
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set unified ring functions
+ */
+static void vcn_v5_0_1_set_unified_ring_funcs(struct amdgpu_device *adev)
+{
+	int i, vcn_inst;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		adev->vcn.inst[i].ring_enc[0].funcs = &vcn_v5_0_1_unified_ring_vm_funcs;
+		adev->vcn.inst[i].ring_enc[0].me = i;
+		vcn_inst = GET_INST(VCN, i);
+		adev->vcn.inst[i].aid_id = vcn_inst / adev->vcn.num_inst_per_aid;
+	}
+}
+
+/**
+ * vcn_v5_0_1_is_idle - check VCN block is idle
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block structure
+ *
+ * Check whether VCN block is idle
+ */
+static bool vcn_v5_0_1_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, ret = 1;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i)
+		ret &= (RREG32_SOC15(VCN, GET_INST(VCN, i), regUVD_STATUS) == UVD_STATUS__IDLE);
+
+	return ret;
+}
+
+/**
+ * vcn_v5_0_1_wait_for_idle - wait for VCN block idle
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ * Wait for VCN block idle
+ */
+static int vcn_v5_0_1_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, ret = 0;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		ret = SOC15_WAIT_ON_RREG(VCN, GET_INST(VCN, i), regUVD_STATUS, UVD_STATUS__IDLE,
+			UVD_STATUS__IDLE);
+		if (ret)
+			return ret;
+	}
+
+	return ret;
+}
+
+/**
+ * vcn_v5_0_1_set_clockgating_state - set VCN block clockgating state
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ * @state: clock gating state
+ *
+ * Set VCN block clockgating state
+ */
+static int vcn_v5_0_1_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					    enum amd_clockgating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	bool enable = state == AMD_CG_STATE_GATE;
+	int i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[i];
+
+		if (enable) {
+			if (RREG32_SOC15(VCN, GET_INST(VCN, i), regUVD_STATUS) != UVD_STATUS__IDLE)
+				return -EBUSY;
+			vcn_v5_0_1_enable_clock_gating(vinst);
+		} else {
+			vcn_v5_0_1_disable_clock_gating(vinst);
+		}
+	}
+
+	return 0;
+}
+
+static int vcn_v5_0_1_set_pg_state(struct amdgpu_vcn_inst *vinst,
+				   enum amd_powergating_state state)
+{
+	struct amdgpu_device *adev = vinst->adev;
+	int ret = 0;
+
+	/* for SRIOV, guest should not control VCN Power-gating
+	 * MMSCH FW should control Power-gating and clock-gating
+	 * guest should avoid touching CGC and PG
+	 */
+	if (amdgpu_sriov_vf(adev)) {
+		vinst->cur_state = AMD_PG_STATE_UNGATE;
+		return 0;
+	}
+
+	if (state == vinst->cur_state)
+		return 0;
+
+	if (state == AMD_PG_STATE_GATE)
+		ret = vcn_v5_0_1_stop(vinst);
+	else
+		ret = vcn_v5_0_1_start(vinst);
+
+	if (!ret)
+		vinst->cur_state = state;
+
+	return ret;
+}
+
+/**
+ * vcn_v5_0_1_process_interrupt - process VCN block interrupt
+ *
+ * @adev: amdgpu_device pointer
+ * @source: interrupt sources
+ * @entry: interrupt entry from clients and sources
+ *
+ * Process VCN block interrupt
+ */
+static int vcn_v5_0_1_process_interrupt(struct amdgpu_device *adev, struct amdgpu_irq_src *source,
+	struct amdgpu_iv_entry *entry)
+{
+	uint32_t i, inst;
+
+	i = node_id_to_phys_map[entry->node_id];
+
+	DRM_DEV_DEBUG(adev->dev, "IH: VCN TRAP\n");
+
+	for (inst = 0; inst < adev->vcn.num_vcn_inst; ++inst)
+		if (adev->vcn.inst[inst].aid_id == i)
+			break;
+	if (inst >= adev->vcn.num_vcn_inst) {
+		dev_WARN_ONCE(adev->dev, 1,
+				"Interrupt received for unknown VCN instance %d",
+				entry->node_id);
+		return 0;
+	}
+
+	switch (entry->src_id) {
+	case VCN_5_0__SRCID__UVD_ENC_GENERAL_PURPOSE:
+		amdgpu_fence_process(&adev->vcn.inst[inst].ring_enc[0]);
+		break;
+	default:
+		DRM_DEV_ERROR(adev->dev, "Unhandled interrupt: %d %d\n",
+			  entry->src_id, entry->src_data[0]);
+		break;
+	}
+
+	return 0;
+}
+
+static int vcn_v5_0_1_set_ras_interrupt_state(struct amdgpu_device *adev,
+					struct amdgpu_irq_src *source,
+					unsigned int type,
+					enum amdgpu_interrupt_state state)
+{
+	return 0;
+}
+
+static const struct amdgpu_irq_src_funcs vcn_v5_0_1_irq_funcs = {
+	.process = vcn_v5_0_1_process_interrupt,
+};
+
+static const struct amdgpu_irq_src_funcs vcn_v5_0_1_ras_irq_funcs = {
+	.set = vcn_v5_0_1_set_ras_interrupt_state,
+	.process = amdgpu_vcn_process_poison_irq,
+};
+
+
+/**
+ * vcn_v5_0_1_set_irq_funcs - set VCN block interrupt irq functions
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set VCN block interrupt irq functions
+ */
+static void vcn_v5_0_1_set_irq_funcs(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i)
+		adev->vcn.inst->irq.num_types++;
+
+	adev->vcn.inst->irq.funcs = &vcn_v5_0_1_irq_funcs;
+
+	adev->vcn.inst->ras_poison_irq.num_types = 1;
+	adev->vcn.inst->ras_poison_irq.funcs = &vcn_v5_0_1_ras_irq_funcs;
+
+}
+
+static const struct amd_ip_funcs vcn_v5_0_1_ip_funcs = {
+	.name = "vcn_v5_0_1",
+	.early_init = vcn_v5_0_1_early_init,
+	.late_init = vcn_v5_0_1_late_init,
+	.sw_init = vcn_v5_0_1_sw_init,
+	.sw_fini = vcn_v5_0_1_sw_fini,
+	.hw_init = vcn_v5_0_1_hw_init,
+	.hw_fini = vcn_v5_0_1_hw_fini,
+	.suspend = vcn_v5_0_1_suspend,
+	.resume = vcn_v5_0_1_resume,
+	.is_idle = vcn_v5_0_1_is_idle,
+	.wait_for_idle = vcn_v5_0_1_wait_for_idle,
+	.check_soft_reset = NULL,
+	.pre_soft_reset = NULL,
+	.soft_reset = NULL,
+	.post_soft_reset = NULL,
+	.set_clockgating_state = vcn_v5_0_1_set_clockgating_state,
+	.set_powergating_state = vcn_set_powergating_state,
+	.dump_ip_state = amdgpu_vcn_dump_ip_state,
+	.print_ip_state = amdgpu_vcn_print_ip_state,
+};
+
+const struct amdgpu_ip_block_version vcn_v5_0_1_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_VCN,
+	.major = 5,
+	.minor = 0,
+	.rev = 1,
+	.funcs = &vcn_v5_0_1_ip_funcs,
+};
+
+static uint32_t vcn_v5_0_1_query_poison_by_instance(struct amdgpu_device *adev,
+			uint32_t instance, uint32_t sub_block)
+{
+	uint32_t poison_stat = 0, reg_value = 0;
+
+	switch (sub_block) {
+	case AMDGPU_VCN_V5_0_1_VCPU_VCODEC:
+		reg_value = RREG32_SOC15(VCN, instance, regUVD_RAS_VCPU_VCODEC_STATUS);
+		poison_stat = REG_GET_FIELD(reg_value, UVD_RAS_VCPU_VCODEC_STATUS, POISONED_PF);
+		break;
+	default:
+		break;
+	}
+
+	if (poison_stat)
+		dev_info(adev->dev, "Poison detected in VCN%d, sub_block%d\n",
+			instance, sub_block);
+
+	return poison_stat;
+}
+
+static bool vcn_v5_0_1_query_poison_status(struct amdgpu_device *adev)
+{
+	uint32_t inst, sub;
+	uint32_t poison_stat = 0;
+
+	for (inst = 0; inst < adev->vcn.num_vcn_inst; inst++)
+		for (sub = 0; sub < AMDGPU_VCN_V5_0_1_MAX_SUB_BLOCK; sub++)
+			poison_stat +=
+			vcn_v5_0_1_query_poison_by_instance(adev, inst, sub);
+
+	return !!poison_stat;
+}
+
+static const struct amdgpu_ras_block_hw_ops vcn_v5_0_1_ras_hw_ops = {
+	.query_poison_status = vcn_v5_0_1_query_poison_status,
+};
+
+static int vcn_v5_0_1_aca_bank_parser(struct aca_handle *handle, struct aca_bank *bank,
+				      enum aca_smu_type type, void *data)
+{
+	struct aca_bank_info info;
+	u64 misc0;
+	int ret;
+
+	ret = aca_bank_info_decode(bank, &info);
+	if (ret)
+		return ret;
+
+	misc0 = bank->regs[ACA_REG_IDX_MISC0];
+	switch (type) {
+	case ACA_SMU_TYPE_UE:
+		bank->aca_err_type = ACA_ERROR_TYPE_UE;
+		ret = aca_error_cache_log_bank_error(handle, &info, ACA_ERROR_TYPE_UE,
+						     1ULL);
+		break;
+	case ACA_SMU_TYPE_CE:
+		bank->aca_err_type = ACA_ERROR_TYPE_CE;
+		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type,
+						     ACA_REG__MISC0__ERRCNT(misc0));
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return ret;
+}
+
+/* reference to smu driver if header file */
+static int vcn_v5_0_1_err_codes[] = {
+	14, 15, 47, /* VCN [D|V|S] */
+};
+
+static bool vcn_v5_0_1_aca_bank_is_valid(struct aca_handle *handle, struct aca_bank *bank,
+					 enum aca_smu_type type, void *data)
+{
+	u32 instlo;
+
+	instlo = ACA_REG__IPID__INSTANCEIDLO(bank->regs[ACA_REG_IDX_IPID]);
+	instlo &= GENMASK(31, 1);
+
+	if (instlo != mmSMNAID_AID0_MCA_SMU)
+		return false;
+
+	if (aca_bank_check_error_codes(handle->adev, bank,
+				       vcn_v5_0_1_err_codes,
+				       ARRAY_SIZE(vcn_v5_0_1_err_codes)))
+		return false;
+
+	return true;
+}
+
+static const struct aca_bank_ops vcn_v5_0_1_aca_bank_ops = {
+	.aca_bank_parser = vcn_v5_0_1_aca_bank_parser,
+	.aca_bank_is_valid = vcn_v5_0_1_aca_bank_is_valid,
+};
+
+static const struct aca_info vcn_v5_0_1_aca_info = {
+	.hwip = ACA_HWIP_TYPE_SMU,
+	.mask = ACA_ERROR_UE_MASK,
+	.bank_ops = &vcn_v5_0_1_aca_bank_ops,
+};
+
+static int vcn_v5_0_1_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
+{
+	int r;
+
+	r = amdgpu_ras_block_late_init(adev, ras_block);
+	if (r)
+		return r;
+
+	r = amdgpu_ras_bind_aca(adev, AMDGPU_RAS_BLOCK__VCN,
+				&vcn_v5_0_1_aca_info, NULL);
+	if (r)
+		goto late_fini;
+
+	if (amdgpu_ras_is_supported(adev, ras_block->block) &&
+		adev->vcn.inst->ras_poison_irq.funcs) {
+		r = amdgpu_irq_get(adev, &adev->vcn.inst->ras_poison_irq, 0);
+		if (r)
+			goto late_fini;
+	}
+
+	return 0;
+
+late_fini:
+	amdgpu_ras_block_late_fini(adev, ras_block);
+
+	return r;
+}
+
+static struct amdgpu_vcn_ras vcn_v5_0_1_ras = {
+	.ras_block = {
+		.hw_ops = &vcn_v5_0_1_ras_hw_ops,
+		.ras_late_init = vcn_v5_0_1_ras_late_init,
+	},
+};
+
+static void vcn_v5_0_1_set_ras_funcs(struct amdgpu_device *adev)
+{
+	adev->vcn.ras = &vcn_v5_0_1_ras;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.h b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.h
new file mode 100644
index 000000000000..b72e4da68317
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.h
@@ -0,0 +1,39 @@
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __VCN_v5_0_1_H__
+#define __VCN_v5_0_1_H__
+
+#define regVCN_RRMT_CNTL                          0x0940
+#define regVCN_RRMT_CNTL_BASE_IDX                 1
+
+
+enum amdgpu_vcn_v5_0_1_sub_block {
+	AMDGPU_VCN_V5_0_1_VCPU_VCODEC = 0,
+
+	AMDGPU_VCN_V5_0_1_MAX_SUB_BLOCK,
+};
+
+extern const struct amdgpu_ip_block_version vcn_v5_0_1_ip_block;
+
+#endif /* __VCN_v5_0_1_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c
index 1e83db0c5438..eb16916c6473 100644
--- a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c
@@ -364,15 +364,20 @@ static u32 vega10_ih_get_wptr(struct amdgpu_device *adev,
 	 * this should allow us to catchup.
 	 */
 	tmp = (wptr + 32) & ih->ptr_mask;
-	dev_warn(adev->dev, "IH ring buffer overflow "
-		 "(0x%08X, 0x%08X, 0x%08X)\n",
-		 wptr, ih->rptr, tmp);
+	dev_warn_ratelimited(adev->dev, "%s ring buffer overflow (0x%08X, 0x%08X, 0x%08X)\n",
+			     amdgpu_ih_ring_name(adev, ih), wptr, ih->rptr, tmp);
 	ih->rptr = tmp;
 
 	tmp = RREG32_NO_KIQ(ih_regs->ih_rb_cntl);
 	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
 	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
 
+	/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+	 * can be detected.
+	 */
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+
 out:
 	return (wptr & ih->ptr_mask);
 }
@@ -466,18 +471,18 @@ static void vega10_ih_set_self_irq_funcs(struct amdgpu_device *adev)
 	adev->irq.self_irq.funcs = &vega10_ih_self_irq_funcs;
 }
 
-static int vega10_ih_early_init(void *handle)
+static int vega10_ih_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	vega10_ih_set_interrupt_funcs(adev);
 	vega10_ih_set_self_irq_funcs(adev);
 	return 0;
 }
 
-static int vega10_ih_sw_init(void *handle)
+static int vega10_ih_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_IH, 0,
@@ -485,7 +490,7 @@ static int vega10_ih_sw_init(void *handle)
 	if (r)
 		return r;
 
-	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 256 * 1024, true);
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, IH_RING_SIZE, true);
 	if (r)
 		return r;
 
@@ -510,7 +515,7 @@ static int vega10_ih_sw_init(void *handle)
 	/* initialize ih control registers offset */
 	vega10_ih_init_register_offset(adev);
 
-	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, PAGE_SIZE, true);
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, IH_SW_RING_SIZE, true);
 	if (r)
 		return r;
 
@@ -519,58 +524,50 @@ static int vega10_ih_sw_init(void *handle)
 	return r;
 }
 
-static int vega10_ih_sw_fini(void *handle)
+static int vega10_ih_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_fini_sw(adev);
 
 	return 0;
 }
 
-static int vega10_ih_hw_init(void *handle)
+static int vega10_ih_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return vega10_ih_irq_init(adev);
+	return vega10_ih_irq_init(ip_block->adev);
 }
 
-static int vega10_ih_hw_fini(void *handle)
+static int vega10_ih_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	vega10_ih_irq_disable(adev);
+	vega10_ih_irq_disable(ip_block->adev);
 
 	return 0;
 }
 
-static int vega10_ih_suspend(void *handle)
+static int vega10_ih_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return vega10_ih_hw_fini(adev);
+	return vega10_ih_hw_fini(ip_block);
 }
 
-static int vega10_ih_resume(void *handle)
+static int vega10_ih_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return vega10_ih_hw_init(adev);
+	return vega10_ih_hw_init(ip_block);
 }
 
-static bool vega10_ih_is_idle(void *handle)
+static bool vega10_ih_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* todo */
 	return true;
 }
 
-static int vega10_ih_wait_for_idle(void *handle)
+static int vega10_ih_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* todo */
 	return -ETIMEDOUT;
 }
 
-static int vega10_ih_soft_reset(void *handle)
+static int vega10_ih_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	/* todo */
 
@@ -607,10 +604,10 @@ static void vega10_ih_update_clockgating_state(struct amdgpu_device *adev,
 	}
 }
 
-static int vega10_ih_set_clockgating_state(void *handle,
+static int vega10_ih_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	vega10_ih_update_clockgating_state(adev,
 				state == AMD_CG_STATE_GATE);
@@ -618,7 +615,7 @@ static int vega10_ih_set_clockgating_state(void *handle,
 
 }
 
-static int vega10_ih_set_powergating_state(void *handle,
+static int vega10_ih_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -627,7 +624,6 @@ static int vega10_ih_set_powergating_state(void *handle,
 const struct amd_ip_funcs vega10_ih_ip_funcs = {
 	.name = "vega10_ih",
 	.early_init = vega10_ih_early_init,
-	.late_init = NULL,
 	.sw_init = vega10_ih_sw_init,
 	.sw_fini = vega10_ih_sw_fini,
 	.hw_init = vega10_ih_hw_init,
diff --git a/drivers/gpu/drm/amd/amdgpu/vega20_ih.c b/drivers/gpu/drm/amd/amdgpu/vega20_ih.c
index 536128447b71..85846fd08ce4 100644
--- a/drivers/gpu/drm/amd/amdgpu/vega20_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/vega20_ih.c
@@ -114,6 +114,33 @@ static int vega20_ih_toggle_ring_interrupts(struct amdgpu_device *adev,
 	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_ENABLE, (enable ? 1 : 0));
 	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_GPU_TS_ENABLE, 1);
 
+	if (enable) {
+		/* Unset the CLEAR_OVERFLOW bit to make sure the next step
+		 * is switching the bit from 0 to 1
+		 */
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+		if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) {
+			if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp))
+				return -ETIMEDOUT;
+		} else {
+			WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+		}
+
+		/* Clear RB_OVERFLOW bit */
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
+		if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) {
+			if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp))
+				return -ETIMEDOUT;
+		} else {
+			WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+		}
+
+		/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+		 * can be detected.
+		 */
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+	}
+
 	/* enable_intr field is only valid in ring0 */
 	if (ih == &adev->irq.ih)
 		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, ENABLE_INTR, (enable ? 1 : 0));
@@ -291,27 +318,30 @@ static int vega20_ih_irq_init(struct amdgpu_device *adev)
 
 	adev->nbio.funcs->ih_control(adev);
 
-	if ((adev->ip_versions[OSSSYS_HWIP][0] == IP_VERSION(4, 2, 1)) &&
-	    adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
-		ih_chicken = RREG32_SOC15(OSSSYS, 0, mmIH_CHICKEN);
-		if (adev->irq.ih.use_bus_addr) {
-			ih_chicken = REG_SET_FIELD(ih_chicken, IH_CHICKEN,
-						   MC_SPACE_GPA_ENABLE, 1);
+	if (!amdgpu_sriov_vf(adev)) {
+		if ((amdgpu_ip_version(adev, OSSSYS_HWIP, 0) == IP_VERSION(4, 2, 1)) &&
+		    adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
+			ih_chicken = RREG32_SOC15(OSSSYS, 0, mmIH_CHICKEN);
+			if (adev->irq.ih.use_bus_addr) {
+				ih_chicken = REG_SET_FIELD(ih_chicken, IH_CHICKEN,
+							   MC_SPACE_GPA_ENABLE, 1);
+			}
+			WREG32_SOC15(OSSSYS, 0, mmIH_CHICKEN, ih_chicken);
 		}
-		WREG32_SOC15(OSSSYS, 0, mmIH_CHICKEN, ih_chicken);
-	}
 
-	/* psp firmware won't program IH_CHICKEN for aldebaran
-	 * driver needs to program it properly according to
-	 * MC_SPACE type in IH_RB_CNTL */
-	if ((adev->ip_versions[OSSSYS_HWIP][0] == IP_VERSION(4, 4, 0)) ||
-	    (adev->ip_versions[OSSSYS_HWIP][0] == IP_VERSION(4, 4, 2))) {
-		ih_chicken = RREG32_SOC15(OSSSYS, 0, mmIH_CHICKEN_ALDEBARAN);
-		if (adev->irq.ih.use_bus_addr) {
-			ih_chicken = REG_SET_FIELD(ih_chicken, IH_CHICKEN,
-						   MC_SPACE_GPA_ENABLE, 1);
+		/* psp firmware won't program IH_CHICKEN for aldebaran
+		 * driver needs to program it properly according to
+		 * MC_SPACE type in IH_RB_CNTL */
+		if ((amdgpu_ip_version(adev, OSSSYS_HWIP, 0) == IP_VERSION(4, 4, 0)) ||
+		    (amdgpu_ip_version(adev, OSSSYS_HWIP, 0) == IP_VERSION(4, 4, 2)) ||
+		    (amdgpu_ip_version(adev, OSSSYS_HWIP, 0) == IP_VERSION(4, 4, 5))) {
+			ih_chicken = RREG32_SOC15(OSSSYS, 0, mmIH_CHICKEN_ALDEBARAN);
+			if (adev->irq.ih.use_bus_addr) {
+				ih_chicken = REG_SET_FIELD(ih_chicken, IH_CHICKEN,
+							   MC_SPACE_GPA_ENABLE, 1);
+			}
+			WREG32_SOC15(OSSSYS, 0, mmIH_CHICKEN_ALDEBARAN, ih_chicken);
 		}
-		WREG32_SOC15(OSSSYS, 0, mmIH_CHICKEN_ALDEBARAN, ih_chicken);
 	}
 
 	for (i = 0; i < ARRAY_SIZE(ih); i++) {
@@ -320,6 +350,7 @@ static int vega20_ih_irq_init(struct amdgpu_device *adev)
 			if (ret)
 				return ret;
 		}
+		ih[i]->overflow = false;
 	}
 
 	if (!amdgpu_sriov_vf(adev))
@@ -334,7 +365,10 @@ static int vega20_ih_irq_init(struct amdgpu_device *adev)
 		vega20_setup_retry_doorbell(adev->irq.retry_cam_doorbell_index));
 
 	/* Enable IH Retry CAM */
-	if (adev->ip_versions[OSSSYS_HWIP][0] == IP_VERSION(4, 4, 0))
+	if (amdgpu_ip_version(adev, OSSSYS_HWIP, 0) == IP_VERSION(4, 4, 0) ||
+	    amdgpu_ip_version(adev, OSSSYS_HWIP, 0) == IP_VERSION(4, 4, 2) ||
+	    amdgpu_ip_version(adev, OSSSYS_HWIP, 0) == IP_VERSION(4, 4, 4) ||
+	    amdgpu_ip_version(adev, OSSSYS_HWIP, 0) == IP_VERSION(4, 4, 5))
 		WREG32_FIELD15(OSSSYS, 0, IH_RETRY_INT_CAM_CNTL_ALDEBARAN,
 			       ENABLE, 1);
 	else
@@ -404,22 +438,30 @@ static u32 vega20_ih_get_wptr(struct amdgpu_device *adev,
 	if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
 		goto out;
 
-	wptr = REG_SET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW, 0);
+	if (!amdgpu_sriov_vf(adev))
+		wptr = REG_SET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW, 0);
+	else
+		ih->overflow = true;
 
 	/* When a ring buffer overflow happen start parsing interrupt
 	 * from the last not overwritten vector (wptr + 32). Hopefully
 	 * this should allow us to catchup.
 	 */
 	tmp = (wptr + 32) & ih->ptr_mask;
-	dev_warn(adev->dev, "IH ring buffer overflow "
-		 "(0x%08X, 0x%08X, 0x%08X)\n",
-		 wptr, ih->rptr, tmp);
+	dev_warn_ratelimited(adev->dev, "%s ring buffer overflow (0x%08X, 0x%08X, 0x%08X)\n",
+			     amdgpu_ih_ring_name(adev, ih), wptr, ih->rptr, tmp);
 	ih->rptr = tmp;
 
 	tmp = RREG32_NO_KIQ(ih_regs->ih_rb_cntl);
 	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
 	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
 
+	/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+	 * can be detected.
+	 */
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+
 out:
 	return (wptr & ih->ptr_mask);
 }
@@ -499,7 +541,8 @@ static int vega20_ih_self_irq(struct amdgpu_device *adev,
 	case 2:
 		schedule_work(&adev->irq.ih2_work);
 		break;
-	default: break;
+	default:
+		break;
 	}
 	return 0;
 }
@@ -514,18 +557,19 @@ static void vega20_ih_set_self_irq_funcs(struct amdgpu_device *adev)
 	adev->irq.self_irq.funcs = &vega20_ih_self_irq_funcs;
 }
 
-static int vega20_ih_early_init(void *handle)
+static int vega20_ih_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	vega20_ih_set_interrupt_funcs(adev);
 	vega20_ih_set_self_irq_funcs(adev);
 	return 0;
 }
 
-static int vega20_ih_sw_init(void *handle)
+static int vega20_ih_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	bool use_bus_addr = true;
 	int r;
 
 	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_IH, 0,
@@ -533,21 +577,26 @@ static int vega20_ih_sw_init(void *handle)
 	if (r)
 		return r;
 
-	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 256 * 1024, true);
+	if ((adev->flags & AMD_IS_APU) &&
+	    (amdgpu_ip_version(adev, OSSSYS_HWIP, 0) == IP_VERSION(4, 4, 2)))
+		use_bus_addr = false;
+
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, IH_RING_SIZE, use_bus_addr);
 	if (r)
 		return r;
 
 	adev->irq.ih.use_doorbell = true;
 	adev->irq.ih.doorbell_index = adev->doorbell_index.ih << 1;
 
-	r = amdgpu_ih_ring_init(adev, &adev->irq.ih1, PAGE_SIZE, true);
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih1, PAGE_SIZE, use_bus_addr);
 	if (r)
 		return r;
 
 	adev->irq.ih1.use_doorbell = true;
 	adev->irq.ih1.doorbell_index = (adev->doorbell_index.ih + 1) << 1;
 
-	if (adev->ip_versions[OSSSYS_HWIP][0] != IP_VERSION(4, 4, 2)) {
+	if (amdgpu_ip_version(adev, OSSSYS_HWIP, 0) != IP_VERSION(4, 4, 2) &&
+	    amdgpu_ip_version(adev, OSSSYS_HWIP, 0) != IP_VERSION(4, 4, 5)) {
 		r = amdgpu_ih_ring_init(adev, &adev->irq.ih2, PAGE_SIZE, true);
 		if (r)
 			return r;
@@ -559,7 +608,7 @@ static int vega20_ih_sw_init(void *handle)
 	/* initialize ih control registers offset */
 	vega20_ih_init_register_offset(adev);
 
-	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, PAGE_SIZE, true);
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, IH_SW_RING_SIZE, use_bus_addr);
 	if (r)
 		return r;
 
@@ -568,19 +617,19 @@ static int vega20_ih_sw_init(void *handle)
 	return r;
 }
 
-static int vega20_ih_sw_fini(void *handle)
+static int vega20_ih_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_fini_sw(adev);
 
 	return 0;
 }
 
-static int vega20_ih_hw_init(void *handle)
+static int vega20_ih_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = vega20_ih_irq_init(adev);
 	if (r)
@@ -589,42 +638,36 @@ static int vega20_ih_hw_init(void *handle)
 	return 0;
 }
 
-static int vega20_ih_hw_fini(void *handle)
+static int vega20_ih_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	vega20_ih_irq_disable(adev);
+	vega20_ih_irq_disable(ip_block->adev);
 
 	return 0;
 }
 
-static int vega20_ih_suspend(void *handle)
+static int vega20_ih_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return vega20_ih_hw_fini(adev);
+	return vega20_ih_hw_fini(ip_block);
 }
 
-static int vega20_ih_resume(void *handle)
+static int vega20_ih_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return vega20_ih_hw_init(adev);
+	return vega20_ih_hw_init(ip_block);
 }
 
-static bool vega20_ih_is_idle(void *handle)
+static bool vega20_ih_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* todo */
 	return true;
 }
 
-static int vega20_ih_wait_for_idle(void *handle)
+static int vega20_ih_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* todo */
 	return -ETIMEDOUT;
 }
 
-static int vega20_ih_soft_reset(void *handle)
+static int vega20_ih_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	/* todo */
 
@@ -658,10 +701,10 @@ static void vega20_ih_update_clockgating_state(struct amdgpu_device *adev,
 	}
 }
 
-static int vega20_ih_set_clockgating_state(void *handle,
+static int vega20_ih_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	vega20_ih_update_clockgating_state(adev,
 				state == AMD_CG_STATE_GATE);
@@ -669,7 +712,7 @@ static int vega20_ih_set_clockgating_state(void *handle,
 
 }
 
-static int vega20_ih_set_powergating_state(void *handle,
+static int vega20_ih_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -678,7 +721,6 @@ static int vega20_ih_set_powergating_state(void *handle,
 const struct amd_ip_funcs vega20_ih_ip_funcs = {
 	.name = "vega20_ih",
 	.early_init = vega20_ih_early_init,
-	.late_init = NULL,
 	.sw_init = vega20_ih_sw_init,
 	.sw_fini = vega20_ih_sw_fini,
 	.hw_init = vega20_ih_hw_init,
@@ -704,8 +746,7 @@ static void vega20_ih_set_interrupt_funcs(struct amdgpu_device *adev)
 	adev->irq.ih_funcs = &vega20_ih_funcs;
 }
 
-const struct amdgpu_ip_block_version vega20_ih_ip_block =
-{
+const struct amdgpu_ip_block_version vega20_ih_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_IH,
 	.major = 4,
 	.minor = 2,
diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c
index 531f173ade2d..a611a7345125 100644
--- a/drivers/gpu/drm/amd/amdgpu/vi.c
+++ b/drivers/gpu/drm/amd/amdgpu/vi.c
@@ -67,7 +67,6 @@
 #include "sdma_v2_4.h"
 #include "sdma_v3_0.h"
 #include "dce_v10_0.h"
-#include "dce_v11_0.h"
 #include "iceland_ih.h"
 #include "tonga_ih.h"
 #include "cz_ih.h"
@@ -136,15 +135,15 @@ static const struct amdgpu_video_codec_info polaris_video_codecs_encode_array[]
 	{
 		.codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC,
 		.max_width = 4096,
-		.max_height = 2304,
-		.max_pixels_per_frame = 4096 * 2304,
+		.max_height = 4096,
+		.max_pixels_per_frame = 4096 * 4096,
 		.max_level = 0,
 	},
 	{
 		.codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC,
 		.max_width = 4096,
-		.max_height = 2304,
-		.max_pixels_per_frame = 4096 * 2304,
+		.max_height = 4096,
+		.max_pixels_per_frame = 4096 * 4096,
 		.max_level = 0,
 	},
 };
@@ -167,16 +166,16 @@ static const struct amdgpu_video_codec_info tonga_video_codecs_decode_array[] =
 {
 	{
 		.codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2,
-		.max_width = 4096,
-		.max_height = 4096,
-		.max_pixels_per_frame = 4096 * 4096,
+		.max_width = 1920,
+		.max_height = 1088,
+		.max_pixels_per_frame = 1920 * 1088,
 		.max_level = 3,
 	},
 	{
 		.codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4,
-		.max_width = 4096,
-		.max_height = 4096,
-		.max_pixels_per_frame = 4096 * 4096,
+		.max_width = 1920,
+		.max_height = 1088,
+		.max_pixels_per_frame = 1920 * 1088,
 		.max_level = 5,
 	},
 	{
@@ -188,9 +187,9 @@ static const struct amdgpu_video_codec_info tonga_video_codecs_decode_array[] =
 	},
 	{
 		.codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1,
-		.max_width = 4096,
-		.max_height = 4096,
-		.max_pixels_per_frame = 4096 * 4096,
+		.max_width = 1920,
+		.max_height = 1088,
+		.max_pixels_per_frame = 1920 * 1088,
 		.max_level = 4,
 	},
 };
@@ -206,16 +205,16 @@ static const struct amdgpu_video_codec_info cz_video_codecs_decode_array[] =
 {
 	{
 		.codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2,
-		.max_width = 4096,
-		.max_height = 4096,
-		.max_pixels_per_frame = 4096 * 4096,
+		.max_width = 1920,
+		.max_height = 1088,
+		.max_pixels_per_frame = 1920 * 1088,
 		.max_level = 3,
 	},
 	{
 		.codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4,
-		.max_width = 4096,
-		.max_height = 4096,
-		.max_pixels_per_frame = 4096 * 4096,
+		.max_width = 1920,
+		.max_height = 1088,
+		.max_pixels_per_frame = 1920 * 1088,
 		.max_level = 5,
 	},
 	{
@@ -227,9 +226,9 @@ static const struct amdgpu_video_codec_info cz_video_codecs_decode_array[] =
 	},
 	{
 		.codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1,
-		.max_width = 4096,
-		.max_height = 4096,
-		.max_pixels_per_frame = 4096 * 4096,
+		.max_width = 1920,
+		.max_height = 1088,
+		.max_pixels_per_frame = 1920 * 1088,
 		.max_level = 4,
 	},
 	{
@@ -542,8 +541,15 @@ static u32 vi_get_xclk(struct amdgpu_device *adev)
 	u32 reference_clock = adev->clock.spll.reference_freq;
 	u32 tmp;
 
-	if (adev->flags & AMD_IS_APU)
-		return reference_clock;
+	if (adev->flags & AMD_IS_APU) {
+		switch (adev->asic_type) {
+		case CHIP_STONEY:
+			/* vbios says 48Mhz, but the actual freq is 100Mhz */
+			return 10000;
+		default:
+			return reference_clock;
+		}
+	}
 
 	tmp = RREG32_SMC(ixCG_CLKPIN_CNTL_2);
 	if (REG_GET_FIELD(tmp, CG_CLKPIN_CNTL_2, MUX_TCLK_TO_XCLK))
@@ -580,11 +586,6 @@ void vi_srbm_select(struct amdgpu_device *adev,
 	WREG32(mmSRBM_GFX_CNTL, srbm_gfx_cntl);
 }
 
-static void vi_vga_set_state(struct amdgpu_device *adev, bool state)
-{
-	/* todo */
-}
-
 static bool vi_read_disabled_bios(struct amdgpu_device *adev)
 {
 	u32 bus_cntl;
@@ -762,12 +763,12 @@ static uint32_t vi_get_register_value(struct amdgpu_device *adev,
 
 		mutex_lock(&adev->grbm_idx_mutex);
 		if (se_num != 0xffffffff || sh_num != 0xffffffff)
-			amdgpu_gfx_select_se_sh(adev, se_num, sh_num, 0xffffffff);
+			amdgpu_gfx_select_se_sh(adev, se_num, sh_num, 0xffffffff, 0);
 
 		val = RREG32(reg_offset);
 
 		if (se_num != 0xffffffff || sh_num != 0xffffffff)
-			amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+			amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 		mutex_unlock(&adev->grbm_idx_mutex);
 		return val;
 	} else {
@@ -895,7 +896,7 @@ static int vi_asic_pci_config_reset(struct amdgpu_device *adev)
 	return r;
 }
 
-static bool vi_asic_supports_baco(struct amdgpu_device *adev)
+static int vi_asic_supports_baco(struct amdgpu_device *adev)
 {
 	switch (adev->asic_type) {
 	case CHIP_FIJI:
@@ -906,14 +907,14 @@ static bool vi_asic_supports_baco(struct amdgpu_device *adev)
 	case CHIP_TOPAZ:
 		return amdgpu_dpm_is_baco_supported(adev);
 	default:
-		return false;
+		return 0;
 	}
 }
 
 static enum amd_reset_method
 vi_asic_reset_method(struct amdgpu_device *adev)
 {
-	bool baco_reset;
+	int baco_reset;
 
 	if (amdgpu_reset_method == AMD_RESET_METHOD_LEGACY ||
 	    amdgpu_reset_method == AMD_RESET_METHOD_BACO)
@@ -933,7 +934,7 @@ vi_asic_reset_method(struct amdgpu_device *adev)
 		baco_reset = amdgpu_dpm_is_baco_supported(adev);
 		break;
 	default:
-		baco_reset = false;
+		baco_reset = 0;
 		break;
 	}
 
@@ -1122,11 +1123,10 @@ static void vi_program_aspm(struct amdgpu_device *adev)
 	bool bL1SS = false;
 	bool bClkReqSupport = true;
 
-	if (!amdgpu_device_should_use_aspm(adev) || !amdgpu_device_aspm_support_quirk())
+	if (!amdgpu_device_should_use_aspm(adev))
 		return;
 
-	if (adev->flags & AMD_IS_APU ||
-	    adev->asic_type < CHIP_POLARIS10)
+	if (adev->asic_type < CHIP_POLARIS10)
 		return;
 
 	orig = data = RREG32_PCIE(ixPCIE_LC_CNTL);
@@ -1435,7 +1435,6 @@ static const struct amdgpu_asic_funcs vi_asic_funcs =
 	.read_register = &vi_read_register,
 	.reset = &vi_asic_reset,
 	.reset_method = &vi_asic_reset_method,
-	.set_vga_state = &vi_vga_set_state,
 	.get_xclk = &vi_get_xclk,
 	.set_uvd_clocks = &vi_set_uvd_clocks,
 	.set_vce_clocks = &vi_set_vce_clocks,
@@ -1455,9 +1454,9 @@ static const struct amdgpu_asic_funcs vi_asic_funcs =
 #define CZ_REV_BRISTOL(rev)	 \
 	((rev >= 0xC8 && rev <= 0xCE) || (rev >= 0xE1 && rev <= 0xE6))
 
-static int vi_common_early_init(void *handle)
+static int vi_common_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (adev->flags & AMD_IS_APU) {
 		adev->smc_rreg = &cz_smc_rreg;
@@ -1679,9 +1678,9 @@ static int vi_common_early_init(void *handle)
 	return 0;
 }
 
-static int vi_common_late_init(void *handle)
+static int vi_common_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		xgpu_vi_mailbox_get_irq(adev);
@@ -1689,9 +1688,9 @@ static int vi_common_late_init(void *handle)
 	return 0;
 }
 
-static int vi_common_sw_init(void *handle)
+static int vi_common_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		xgpu_vi_mailbox_add_irq_id(adev);
@@ -1699,14 +1698,9 @@ static int vi_common_sw_init(void *handle)
 	return 0;
 }
 
-static int vi_common_sw_fini(void *handle)
+static int vi_common_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	return 0;
-}
-
-static int vi_common_hw_init(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* move the golden regs per IP block */
 	vi_init_golden_registers(adev);
@@ -1718,9 +1712,9 @@ static int vi_common_hw_init(void *handle)
 	return 0;
 }
 
-static int vi_common_hw_fini(void *handle)
+static int vi_common_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* enable the doorbell aperture */
 	vi_enable_doorbell_aperture(adev, false);
@@ -1731,35 +1725,21 @@ static int vi_common_hw_fini(void *handle)
 	return 0;
 }
 
-static int vi_common_suspend(void *handle)
+static int vi_common_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return vi_common_hw_fini(adev);
+	return vi_common_hw_fini(ip_block);
 }
 
-static int vi_common_resume(void *handle)
+static int vi_common_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return vi_common_hw_init(adev);
+	return vi_common_hw_init(ip_block);
 }
 
-static bool vi_common_is_idle(void *handle)
+static bool vi_common_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	return true;
 }
 
-static int vi_common_wait_for_idle(void *handle)
-{
-	return 0;
-}
-
-static int vi_common_soft_reset(void *handle)
-{
-	return 0;
-}
-
 static void vi_update_bif_medium_grain_light_sleep(struct amdgpu_device *adev,
 						   bool enable)
 {
@@ -1964,10 +1944,10 @@ static int vi_common_set_clockgating_state_by_smu(void *handle,
 	return 0;
 }
 
-static int vi_common_set_clockgating_state(void *handle,
+static int vi_common_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
@@ -2007,15 +1987,15 @@ static int vi_common_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int vi_common_set_powergating_state(void *handle,
+static int vi_common_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					    enum amd_powergating_state state)
 {
 	return 0;
 }
 
-static void vi_common_get_clockgating_state(void *handle, u64 *flags)
+static void vi_common_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	if (amdgpu_sriov_vf(adev))
@@ -2047,14 +2027,11 @@ static const struct amd_ip_funcs vi_common_ip_funcs = {
 	.early_init = vi_common_early_init,
 	.late_init = vi_common_late_init,
 	.sw_init = vi_common_sw_init,
-	.sw_fini = vi_common_sw_fini,
 	.hw_init = vi_common_hw_init,
 	.hw_fini = vi_common_hw_fini,
 	.suspend = vi_common_suspend,
 	.resume = vi_common_resume,
 	.is_idle = vi_common_is_idle,
-	.wait_for_idle = vi_common_wait_for_idle,
-	.soft_reset = vi_common_soft_reset,
 	.set_clockgating_state = vi_common_set_clockgating_state,
 	.set_powergating_state = vi_common_set_powergating_state,
 	.get_clockgating_state = vi_common_get_clockgating_state,
@@ -2146,8 +2123,6 @@ int vi_set_ip_blocks(struct amdgpu_device *adev)
 		else if (amdgpu_device_has_dc_support(adev))
 			amdgpu_device_ip_block_add(adev, &dm_ip_block);
 #endif
-		else
-			amdgpu_device_ip_block_add(adev, &dce_v11_2_ip_block);
 		amdgpu_device_ip_block_add(adev, &uvd_v6_3_ip_block);
 		amdgpu_device_ip_block_add(adev, &vce_v3_4_ip_block);
 		break;
@@ -2164,8 +2139,6 @@ int vi_set_ip_blocks(struct amdgpu_device *adev)
 		else if (amdgpu_device_has_dc_support(adev))
 			amdgpu_device_ip_block_add(adev, &dm_ip_block);
 #endif
-		else
-			amdgpu_device_ip_block_add(adev, &dce_v11_0_ip_block);
 		amdgpu_device_ip_block_add(adev, &uvd_v6_0_ip_block);
 		amdgpu_device_ip_block_add(adev, &vce_v3_1_ip_block);
 #if defined(CONFIG_DRM_AMD_ACP)
@@ -2185,8 +2158,6 @@ int vi_set_ip_blocks(struct amdgpu_device *adev)
 		else if (amdgpu_device_has_dc_support(adev))
 			amdgpu_device_ip_block_add(adev, &dm_ip_block);
 #endif
-		else
-			amdgpu_device_ip_block_add(adev, &dce_v11_0_ip_block);
 		amdgpu_device_ip_block_add(adev, &uvd_v6_2_ip_block);
 		amdgpu_device_ip_block_add(adev, &vce_v3_4_ip_block);
 #if defined(CONFIG_DRM_AMD_ACP)
diff --git a/drivers/gpu/drm/amd/amdgpu/vid.h b/drivers/gpu/drm/amd/amdgpu/vid.h
index 80ce42aacc0c..b61f6b838ec2 100644
--- a/drivers/gpu/drm/amd/amdgpu/vid.h
+++ b/drivers/gpu/drm/amd/amdgpu/vid.h
@@ -246,6 +246,7 @@
 		 * 1 - Stream
 		 * 2 - Bypass
 		 */
+#define		EOP_EXEC				(1 << 28) /* For Trailing Fence */
 #define		DATA_SEL(x)                             ((x) << 29)
 		/* 0 - discard
 		 * 1 - send low 32bit data
diff --git a/drivers/gpu/drm/amd/amdgpu/vpe_6_1_fw_if.h b/drivers/gpu/drm/amd/amdgpu/vpe_6_1_fw_if.h
new file mode 100644
index 000000000000..47534dbbd137
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/vpe_6_1_fw_if.h
@@ -0,0 +1,218 @@
+/* Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: AMD
+ *
+ */
+
+#ifndef __VPE_6_1_FW_IF_H_
+#define __VPE_6_1_FW_IF_H_
+
+/****************
+ * VPE OP Codes
+ ****************/
+enum VPE_CMD_OPCODE {
+    VPE_CMD_OPCODE_NOP          = 0x0,
+    VPE_CMD_OPCODE_VPE_DESC     = 0x1,
+    VPE_CMD_OPCODE_PLANE_CFG    = 0x2,
+    VPE_CMD_OPCODE_VPEP_CFG     = 0x3,
+    VPE_CMD_OPCODE_INDIRECT     = 0x4,
+    VPE_CMD_OPCODE_FENCE        = 0x5,
+    VPE_CMD_OPCODE_TRAP         = 0x6,
+    VPE_CMD_OPCODE_REG_WRITE    = 0x7,
+    VPE_CMD_OPCODE_POLL_REGMEM  = 0x8,
+    VPE_CMD_OPCODE_COND_EXE     = 0x9,
+    VPE_CMD_OPCODE_ATOMIC       = 0xA,
+    VPE_CMD_OPCODE_PRED_EXE     = 0xB,
+    VPE_CMD_OPCODE_COLLAB_SYNC  = 0xC,
+    VPE_CMD_OPCODE_TIMESTAMP    = 0xD
+};
+
+/** Generic Command Header
+ * Generic Commands include:
+ *  Noop, Fence, Trap,
+ *  RegisterWrite, PollRegisterWriteMemory,
+ *  SetLocalTimestamp, GetLocalTimestamp
+ *  GetGlobalGPUTimestamp */
+#define VPE_HEADER_SUB_OPCODE__SHIFT    8
+#define VPE_HEADER_SUB_OPCODE_MASK      0x0000FF00
+#define VPE_HEADER_OPCODE__SHIFT        0
+#define VPE_HEADER_OPCODE_MASK          0x000000FF
+
+#define VPE_CMD_HEADER(op, subop) \
+    (((subop << VPE_HEADER_SUB_OPCODE__SHIFT) & VPE_HEADER_SUB_OPCODE_MASK) | \
+     ((op << VPE_HEADER_OPCODE__SHIFT) & VPE_HEADER_OPCODE_MASK))
+
+
+ /***************************
+  * VPE NOP
+  ***************************/
+#define VPE_CMD_NOP_HEADER_COUNT__SHIFT    16
+#define VPE_CMD_NOP_HEADER_COUNT_MASK      0x00003FFF
+
+#define VPE_CMD_NOP_HEADER_COUNT(count) \
+     (((count) & VPE_CMD_NOP_HEADER_COUNT_MASK) << VPE_CMD_NOP_HEADER_COUNT__SHIFT)
+
+ /***************************
+  * VPE Descriptor
+  ***************************/
+#define VPE_DESC_CD__SHIFT          16
+#define VPE_DESC_CD_MASK            0x000F0000
+
+#define VPE_DESC_CMD_HEADER(cd) \
+    (VPE_CMD_HEADER(VPE_CMD_OPCODE_VPE_DESC, 0) | \
+     (((cd) << VPE_DESC_CD__SHIFT) & VPE_DESC_CD_MASK))
+
+ /***************************
+  * VPE Plane Config
+  ***************************/
+enum VPE_PLANE_CFG_SUBOP {
+    VPE_PLANE_CFG_SUBOP_1_TO_1 = 0x0,
+    VPE_PLANE_CFG_SUBOP_2_TO_1 = 0x1,
+    VPE_PLANE_CFG_SUBOP_2_TO_2 = 0x2
+};
+
+#define VPE_PLANE_CFG_ONE_PLANE     0
+#define VPE_PLANE_CFG_TWO_PLANES    1
+
+#define VPE_PLANE_CFG_NPS0__SHIFT   16
+#define VPE_PLANE_CFG_NPS0_MASK     0x00030000
+
+#define VPE_PLANE_CFG_NPD0__SHIFT   18
+#define VPE_PLANE_CFG_NPD0_MASK     0x000C0000
+
+#define VPE_PLANE_CFG_NPS1__SHIFT   20
+#define VPE_PLANE_CFG_NPS1_MASK     0x00300000
+
+#define VPE_PLANE_CFG_NPD1__SHIFT   22
+#define VPE_PLANE_CFG_NPD1_MASK     0x00C00000
+
+#define VPE_PLANE_CFG_TMZ__SHIFT    16
+#define VPE_PLANE_CFG_TMZ_MASK      0x00010000
+
+#define VPE_PLANE_CFG_SWIZZLE_MODE__SHIFT   3
+#define VPE_PLANE_CFG_SWIZZLE_MODE_MASK     0x000000F8
+
+#define VPE_PLANE_CFG_ROTATION__SHIFT       0
+#define VPE_PLANE_CFG_ROTATION_MASK         0x00000003
+
+#define VPE_PLANE_ADDR_LO__SHIFT        0
+#define VPE_PLANE_ADDR_LO_MASK          0xFFFFFF00
+
+#define VPE_PLANE_CFG_PITCH__SHIFT      0
+#define VPE_PLANE_CFG_PITCH_MASK        0x00003FFF
+
+#define VPE_PLANE_CFG_VIEWPORT_Y__SHIFT 16
+#define VPE_PLANE_CFG_VIEWPORT_Y_MASK   0x3FFF0000
+#define VPE_PLANE_CFG_VIEWPORT_X__SHIFT 0
+#define VPE_PLANE_CFG_VIEWPORT_X_MASK   0x00003FFF
+
+
+#define VPE_PLANE_CFG_VIEWPORT_HEIGHT__SHIFT 16
+#define VPE_PLANE_CFG_VIEWPORT_HEIGHT_MASK   0x1FFF0000
+#define VPE_PLANE_CFG_VIEWPORT_ELEMENT_SIZE__SHIFT  13
+#define VPE_PLANE_CFG_VIEWPORT_ELEMENT_SIZE_MASK    0x0000E000
+#define VPE_PLANE_CFG_VIEWPORT_WIDTH__SHIFT 0
+#define VPE_PLANE_CFG_VIEWPORT_WIDTH_MASK   0x00001FFF
+
+enum VPE_PLANE_CFG_ELEMENT_SIZE {
+    VPE_PLANE_CFG_ELEMENT_SIZE_8BPE     = 0,
+    VPE_PLANE_CFG_ELEMENT_SIZE_16BPE    = 1,
+    VPE_PLANE_CFG_ELEMENT_SIZE_32BPE    = 2,
+    VPE_PLANE_CFG_ELEMENT_SIZE_64BPE    = 3
+};
+
+#define VPE_PLANE_CFG_CMD_HEADER(subop, nps0, npd0, nps1, npd1) \
+    (VPE_CMD_HEADER(VPE_CMD_OPCODE_PLANE_CFG, subop) | \
+     (((nps0) << VPE_PLANE_CFG_NPS0__SHIFT) & VPE_PLANE_CFG_NPS0_MASK) | \
+     (((npd0) << VPE_PLANE_CFG_NPD0__SHIFT) & VPE_PLANE_CFG_NPD0_MASK) | \
+     (((nps1) << VPE_PLANE_CFG_NPS1__SHIFT) & VPE_PLANE_CFG_NPS1_MASK) | \
+     (((npd0) << VPE_PLANE_CFG_NPD1__SHIFT) & VPE_PLANE_CFG_NPD1_MASK))
+
+
+/************************
+ * VPEP Config
+ ************************/
+enum VPE_VPEP_CFG_SUBOP {
+    VPE_VPEP_CFG_SUBOP_DIR_CFG = 0x0,
+    VPE_VPEP_CFG_SUBOP_IND_CFG = 0x1
+};
+
+
+// Direct Config Command Header
+#define VPE_DIR_CFG_HEADER_ARRAY_SIZE__SHIFT    16
+#define VPE_DIR_CFG_HEADER_ARRAY_SIZE_MASK      0xFFFF0000
+
+#define VPE_DIR_CFG_CMD_HEADER(subop, arr_sz) \
+    (VPE_CMD_HEADER(VPE_CMD_OPCODE_VPEP_CFG, subop) | \
+     (((arr_sz) << VPE_DIR_CFG_HEADER_ARRAY_SIZE__SHIFT) & VPE_DIR_CFG_HEADER_ARRAY_SIZE_MASK))
+
+
+#define VPE_DIR_CFG_PKT_REGISTER_OFFSET__SHIFT  2
+#define VPE_DIR_CFG_PKT_REGISTER_OFFSET_MASK    0x000FFFFC
+
+#define VPE_DIR_CFG_PKT_DATA_SIZE__SHIFT        20
+#define VPE_DIR_CFG_PKT_DATA_SIZE_MASK          0xFFF00000
+
+
+// InDirect Config Command Header
+#define VPE_IND_CFG_HEADER_NUM_DST__SHIFT   28
+#define VPE_IND_CFG_HEADER_NUM_DST_MASK     0xF0000000
+
+#define VPE_IND_CFG_CMD_HEADER(subop, num_dst) \
+    (VPE_CMD_HEADER(VPE_CMD_OPCODE_VPEP_CFG, subop) | \
+     (((num_dst) << VPE_IND_CFG_HEADER_NUM_DST__SHIFT) & VPE_IND_CFG_HEADER_NUM_DST_MASK))
+
+// Indirect Buffer Command Header
+#define VPE_CMD_INDIRECT_HEADER_VMID__SHIFT   16
+#define VPE_CMD_INDIRECT_HEADER_VMID_MASK     0x0000000F
+#define VPE_CMD_INDIRECT_HEADER_VMID(vmid) \
+     (((vmid) & VPE_CMD_INDIRECT_HEADER_VMID_MASK) << VPE_CMD_INDIRECT_HEADER_VMID__SHIFT)
+
+
+/**************************
+ * Poll Reg/Mem Sub-OpCode
+ **************************/
+enum VPE_POLL_REGMEM_SUBOP {
+    VPE_POLL_REGMEM_SUBOP_REGMEM = 0x0,
+    VPE_POLL_REGMEM_SUBOP_REGMEM_WRITE = 0x1
+};
+
+#define VPE_CMD_POLL_REGMEM_HEADER_FUNC__SHIFT   28
+#define VPE_CMD_POLL_REGMEM_HEADER_FUNC_MASK     0x00000007
+#define VPE_CMD_POLL_REGMEM_HEADER_FUNC(func) \
+     (((func) & VPE_CMD_POLL_REGMEM_HEADER_FUNC_MASK) << VPE_CMD_POLL_REGMEM_HEADER_FUNC__SHIFT)
+
+#define VPE_CMD_POLL_REGMEM_HEADER_MEM__SHIFT   31
+#define VPE_CMD_POLL_REGMEM_HEADER_MEM_MASK     0x00000001
+#define VPE_CMD_POLL_REGMEM_HEADER_MEM(mem) \
+     (((mem) & VPE_CMD_POLL_REGMEM_HEADER_MEM_MASK) << VPE_CMD_POLL_REGMEM_HEADER_MEM__SHIFT)
+
+#define VPE_CMD_POLL_REGMEM_DW5_INTERVAL__SHIFT   0
+#define VPE_CMD_POLL_REGMEM_DW5_INTERVAL_MASK     0x0000FFFF
+#define VPE_CMD_POLL_REGMEM_DW5_INTERVAL(interval) \
+     (((interval) & VPE_CMD_POLL_REGMEM_DW5_INTERVAL_MASK) << VPE_CMD_POLL_REGMEM_DW5_INTERVAL__SHIFT)
+
+#define VPE_CMD_POLL_REGMEM_DW5_RETRY_COUNT__SHIFT   16
+#define VPE_CMD_POLL_REGMEM_DW5_RETRY_COUNT_MASK     0x00000FFF
+#define VPE_CMD_POLL_REGMEM_DW5_RETRY_COUNT(count) \
+     (((count) & VPE_CMD_POLL_REGMEM_DW5_RETRY_COUNT_MASK) << VPE_CMD_POLL_REGMEM_DW5_RETRY_COUNT__SHIFT)
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/vpe_v6_1.c b/drivers/gpu/drm/amd/amdgpu/vpe_v6_1.c
new file mode 100644
index 000000000000..45876883bbf3
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/vpe_v6_1.c
@@ -0,0 +1,398 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include <linux/firmware.h>
+#include <drm/drm_drv.h>
+
+#include "amdgpu.h"
+#include "amdgpu_ucode.h"
+#include "amdgpu_vpe.h"
+#include "vpe_v6_1.h"
+#include "soc15_common.h"
+#include "ivsrcid/vpe/irqsrcs_vpe_6_1.h"
+#include "vpe/vpe_6_1_0_offset.h"
+#include "vpe/vpe_6_1_0_sh_mask.h"
+
+MODULE_FIRMWARE("amdgpu/vpe_6_1_0.bin");
+MODULE_FIRMWARE("amdgpu/vpe_6_1_1.bin");
+MODULE_FIRMWARE("amdgpu/vpe_6_1_3.bin");
+
+#define VPE_THREAD1_UCODE_OFFSET	0x8000
+
+#define regVPEC_COLLABORATE_CNTL                                                0x0013
+#define regVPEC_COLLABORATE_CNTL_BASE_IDX                                       0
+#define VPEC_COLLABORATE_CNTL__COLLABORATE_MODE_EN__SHIFT                       0x0
+#define VPEC_COLLABORATE_CNTL__COLLABORATE_MODE_EN_MASK                         0x00000001L
+
+#define regVPEC_COLLABORATE_CFG                                                 0x0014
+#define regVPEC_COLLABORATE_CFG_BASE_IDX                                        0
+#define VPEC_COLLABORATE_CFG__MASTER_ID__SHIFT                                  0x0
+#define VPEC_COLLABORATE_CFG__MASTER_EN__SHIFT                                  0x3
+#define VPEC_COLLABORATE_CFG__SLAVE0_ID__SHIFT                                  0x4
+#define VPEC_COLLABORATE_CFG__SLAVE0_EN__SHIFT                                  0x7
+#define VPEC_COLLABORATE_CFG__MASTER_ID_MASK                                    0x00000007L
+#define VPEC_COLLABORATE_CFG__MASTER_EN_MASK                                    0x00000008L
+#define VPEC_COLLABORATE_CFG__SLAVE0_ID_MASK                                    0x00000070L
+#define VPEC_COLLABORATE_CFG__SLAVE0_EN_MASK                                    0x00000080L
+
+#define regVPEC_CNTL_6_1_1                                                      0x0016
+#define regVPEC_CNTL_6_1_1_BASE_IDX                                             0
+#define regVPEC_QUEUE_RESET_REQ_6_1_1                                           0x002c
+#define regVPEC_QUEUE_RESET_REQ_6_1_1_BASE_IDX                                  0
+#define regVPEC_PUB_DUMMY2_6_1_1                                                0x004c
+#define regVPEC_PUB_DUMMY2_6_1_1_BASE_IDX                                       0
+
+static uint32_t vpe_v6_1_get_reg_offset(struct amdgpu_vpe *vpe, uint32_t inst, uint32_t offset)
+{
+	uint32_t base;
+
+	base = vpe->ring.adev->reg_offset[VPE_HWIP][inst][0];
+
+	return base + offset;
+}
+
+static void vpe_v6_1_halt(struct amdgpu_vpe *vpe, bool halt)
+{
+	struct amdgpu_device *adev = vpe->ring.adev;
+	uint32_t i, f32_cntl;
+
+	for (i = 0; i < vpe->num_instances; i++) {
+		f32_cntl = RREG32(vpe_get_reg_offset(vpe, i, regVPEC_F32_CNTL));
+		f32_cntl = REG_SET_FIELD(f32_cntl, VPEC_F32_CNTL, HALT, halt ? 1 : 0);
+		f32_cntl = REG_SET_FIELD(f32_cntl, VPEC_F32_CNTL, TH1_RESET, halt ? 1 : 0);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_F32_CNTL), f32_cntl);
+	}
+}
+
+static int vpe_v6_1_irq_init(struct amdgpu_vpe *vpe)
+{
+	struct amdgpu_device *adev = container_of(vpe, struct amdgpu_device, vpe);
+	int ret;
+
+	ret = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_VPE,
+				VPE_6_1_SRCID__VPE_TRAP,
+				&adev->vpe.trap_irq);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static void vpe_v6_1_set_collaborate_mode(struct amdgpu_vpe *vpe, bool enable)
+{
+	struct amdgpu_device *adev = vpe->ring.adev;
+	uint32_t vpe_colla_cntl, vpe_colla_cfg, i;
+
+	if (!vpe->collaborate_mode)
+		return;
+
+	for (i = 0; i < vpe->num_instances; i++) {
+		vpe_colla_cntl = RREG32(vpe_get_reg_offset(vpe, i, regVPEC_COLLABORATE_CNTL));
+		vpe_colla_cntl = REG_SET_FIELD(vpe_colla_cntl, VPEC_COLLABORATE_CNTL,
+					       COLLABORATE_MODE_EN, enable ? 1 : 0);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_COLLABORATE_CNTL), vpe_colla_cntl);
+
+		vpe_colla_cfg = RREG32(vpe_get_reg_offset(vpe, i, regVPEC_COLLABORATE_CFG));
+		vpe_colla_cfg = REG_SET_FIELD(vpe_colla_cfg, VPEC_COLLABORATE_CFG, MASTER_ID, 0);
+		vpe_colla_cfg = REG_SET_FIELD(vpe_colla_cfg, VPEC_COLLABORATE_CFG, MASTER_EN, enable ? 1 : 0);
+		vpe_colla_cfg = REG_SET_FIELD(vpe_colla_cfg, VPEC_COLLABORATE_CFG, SLAVE0_ID, 1);
+		vpe_colla_cfg = REG_SET_FIELD(vpe_colla_cfg, VPEC_COLLABORATE_CFG, SLAVE0_EN, enable ? 1 : 0);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_COLLABORATE_CFG), vpe_colla_cfg);
+	}
+}
+
+static int vpe_v6_1_load_microcode(struct amdgpu_vpe *vpe)
+{
+	struct amdgpu_device *adev = vpe->ring.adev;
+	const struct vpe_firmware_header_v1_0 *vpe_hdr;
+	const __le32 *data;
+	uint32_t ucode_offset[2], ucode_size[2];
+	uint32_t i, j, size_dw;
+	uint32_t ret;
+
+	/* disable UMSCH_INT_ENABLE */
+	for (j = 0; j < vpe->num_instances; j++) {
+
+		if (amdgpu_ip_version(adev, VPE_HWIP, 0) == IP_VERSION(6, 1, 1))
+			ret = RREG32(vpe_get_reg_offset(vpe, j, regVPEC_CNTL_6_1_1));
+		else
+			ret = RREG32(vpe_get_reg_offset(vpe, j, regVPEC_CNTL));
+
+		ret = REG_SET_FIELD(ret, VPEC_CNTL, UMSCH_INT_ENABLE, 0);
+
+		if (amdgpu_ip_version(adev, VPE_HWIP, 0) == IP_VERSION(6, 1, 1))
+			WREG32(vpe_get_reg_offset(vpe, j, regVPEC_CNTL_6_1_1), ret);
+		else
+			WREG32(vpe_get_reg_offset(vpe, j, regVPEC_CNTL), ret);
+	}
+
+	/* setup collaborate mode */
+	vpe_v6_1_set_collaborate_mode(vpe, true);
+	/* setup DPM */
+	if (amdgpu_vpe_configure_dpm(vpe))
+		dev_warn(adev->dev, "VPE failed to enable DPM\n");
+
+	/*
+	 * For VPE 6.1.1, still only need to add master's offset, and psp will apply it to slave as well.
+	 * Here use instance 0 as master.
+	 */
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		uint32_t f32_offset, f32_cntl;
+
+		f32_offset = vpe_get_reg_offset(vpe, 0, regVPEC_F32_CNTL);
+		f32_cntl = RREG32(f32_offset);
+		f32_cntl = REG_SET_FIELD(f32_cntl, VPEC_F32_CNTL, HALT, 0);
+		f32_cntl = REG_SET_FIELD(f32_cntl, VPEC_F32_CNTL, TH1_RESET, 0);
+
+		adev->vpe.cmdbuf_cpu_addr[0] = f32_offset;
+		adev->vpe.cmdbuf_cpu_addr[1] = f32_cntl;
+
+		return amdgpu_vpe_psp_update_sram(adev);
+	}
+
+	vpe_hdr = (const struct vpe_firmware_header_v1_0 *)adev->vpe.fw->data;
+
+	/* Thread 0(command thread) ucode offset/size */
+	ucode_offset[0] = le32_to_cpu(vpe_hdr->header.ucode_array_offset_bytes);
+	ucode_size[0] = le32_to_cpu(vpe_hdr->ctx_ucode_size_bytes);
+	/* Thread 1(control thread) ucode offset/size */
+	ucode_offset[1] = le32_to_cpu(vpe_hdr->ctl_ucode_offset);
+	ucode_size[1] = le32_to_cpu(vpe_hdr->ctl_ucode_size_bytes);
+
+	vpe_v6_1_halt(vpe, true);
+
+	for (j = 0; j < vpe->num_instances; j++) {
+		for (i = 0; i < 2; i++) {
+			if (i > 0)
+				WREG32(vpe_get_reg_offset(vpe, j, regVPEC_UCODE_ADDR), VPE_THREAD1_UCODE_OFFSET);
+			else
+				WREG32(vpe_get_reg_offset(vpe, j, regVPEC_UCODE_ADDR), 0);
+
+			data = (const __le32 *)(adev->vpe.fw->data + ucode_offset[i]);
+			size_dw = ucode_size[i] / sizeof(__le32);
+
+			while (size_dw--) {
+				if (amdgpu_emu_mode && size_dw % 500 == 0)
+					msleep(1);
+				WREG32(vpe_get_reg_offset(vpe, j, regVPEC_UCODE_DATA), le32_to_cpup(data++));
+			}
+		}
+	}
+
+	vpe_v6_1_halt(vpe, false);
+
+	return 0;
+}
+
+static int vpe_v6_1_ring_start(struct amdgpu_vpe *vpe)
+{
+	struct amdgpu_ring *ring = &vpe->ring;
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t doorbell, doorbell_offset;
+	uint32_t rb_bufsz, rb_cntl;
+	uint32_t ib_cntl, i;
+	int ret;
+
+	for (i = 0; i < vpe->num_instances; i++) {
+		/* Set ring buffer size in dwords */
+		rb_bufsz = order_base_2(ring->ring_size / 4);
+		rb_cntl = RREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_CNTL));
+		rb_cntl = REG_SET_FIELD(rb_cntl, VPEC_QUEUE0_RB_CNTL, RB_SIZE, rb_bufsz);
+		rb_cntl = REG_SET_FIELD(rb_cntl, VPEC_QUEUE0_RB_CNTL, RB_PRIV, 1);
+		rb_cntl = REG_SET_FIELD(rb_cntl, VPEC_QUEUE0_RB_CNTL, RB_VMID, 0);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_CNTL), rb_cntl);
+
+		/* Initialize the ring buffer's read and write pointers */
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_RPTR), 0);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_RPTR_HI), 0);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_WPTR), 0);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_WPTR_HI), 0);
+
+		/* set the wb address whether it's enabled or not */
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_RPTR_ADDR_LO),
+			lower_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFC);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_RPTR_ADDR_HI),
+			upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF);
+
+		rb_cntl = REG_SET_FIELD(rb_cntl, VPEC_QUEUE0_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1);
+
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_BASE), ring->gpu_addr >> 8);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_BASE_HI), ring->gpu_addr >> 40);
+
+		ring->wptr = 0;
+
+		/* before programing wptr to a less value, need set minor_ptr_update first */
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_MINOR_PTR_UPDATE), 1);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_WPTR), lower_32_bits(ring->wptr) << 2);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_WPTR_HI), upper_32_bits(ring->wptr) << 2);
+		/* set minor_ptr_update to 0 after wptr programed */
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_MINOR_PTR_UPDATE), 0);
+
+		doorbell_offset = RREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_DOORBELL_OFFSET));
+		doorbell_offset = REG_SET_FIELD(doorbell_offset, VPEC_QUEUE0_DOORBELL_OFFSET, OFFSET, ring->doorbell_index + i*4);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_DOORBELL_OFFSET), doorbell_offset);
+
+		doorbell = RREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_DOORBELL));
+		doorbell = REG_SET_FIELD(doorbell, VPEC_QUEUE0_DOORBELL, ENABLE, ring->use_doorbell ? 1 : 0);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_DOORBELL), doorbell);
+
+		adev->nbio.funcs->vpe_doorbell_range(adev, i, ring->use_doorbell, ring->doorbell_index + i*4, 4);
+
+		rb_cntl = REG_SET_FIELD(rb_cntl, VPEC_QUEUE0_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1);
+		rb_cntl = REG_SET_FIELD(rb_cntl, VPEC_QUEUE0_RB_CNTL, RB_ENABLE, 1);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_CNTL), rb_cntl);
+
+		ib_cntl = RREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_IB_CNTL));
+		ib_cntl = REG_SET_FIELD(ib_cntl, VPEC_QUEUE0_IB_CNTL, IB_ENABLE, 1);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_IB_CNTL), ib_cntl);
+	}
+
+	ret = amdgpu_ring_test_helper(ring);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static int vpe_v_6_1_ring_stop(struct amdgpu_vpe *vpe)
+{
+	struct amdgpu_device *adev = vpe->ring.adev;
+	uint32_t queue_reset, i;
+	int ret;
+
+	for (i = 0; i < vpe->num_instances; i++) {
+		if (amdgpu_ip_version(adev, VPE_HWIP, 0) == IP_VERSION(6, 1, 1))
+			queue_reset = RREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE_RESET_REQ_6_1_1));
+		else
+			queue_reset = RREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE_RESET_REQ));
+
+		queue_reset = REG_SET_FIELD(queue_reset, VPEC_QUEUE_RESET_REQ, QUEUE0_RESET, 1);
+
+		if (amdgpu_ip_version(adev, VPE_HWIP, 0) == IP_VERSION(6, 1, 1)) {
+			WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE_RESET_REQ_6_1_1), queue_reset);
+			ret = SOC15_WAIT_ON_RREG(VPE, i, regVPEC_QUEUE_RESET_REQ_6_1_1, 0,
+						 VPEC_QUEUE_RESET_REQ__QUEUE0_RESET_MASK);
+		} else {
+			WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE_RESET_REQ), queue_reset);
+			ret = SOC15_WAIT_ON_RREG(VPE, i, regVPEC_QUEUE_RESET_REQ, 0,
+						 VPEC_QUEUE_RESET_REQ__QUEUE0_RESET_MASK);
+		}
+
+		if (ret)
+			dev_err(adev->dev, "VPE queue reset failed\n");
+	}
+
+	vpe->ring.sched.ready = false;
+
+	return ret;
+}
+
+static int vpe_v6_1_set_trap_irq_state(struct amdgpu_device *adev,
+				       struct amdgpu_irq_src *source,
+				       unsigned int type,
+				       enum amdgpu_interrupt_state state)
+{
+	struct amdgpu_vpe *vpe = &adev->vpe;
+	uint32_t vpe_cntl;
+
+	if (amdgpu_ip_version(adev, VPE_HWIP, 0) == IP_VERSION(6, 1, 1))
+		vpe_cntl = RREG32(vpe_get_reg_offset(vpe, 0, regVPEC_CNTL_6_1_1));
+	else
+		vpe_cntl = RREG32(vpe_get_reg_offset(vpe, 0, regVPEC_CNTL));
+
+	vpe_cntl = REG_SET_FIELD(vpe_cntl, VPEC_CNTL, TRAP_ENABLE,
+				 state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+
+	if (amdgpu_ip_version(adev, VPE_HWIP, 0) == IP_VERSION(6, 1, 1))
+		WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_CNTL_6_1_1), vpe_cntl);
+	else
+		WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_CNTL), vpe_cntl);
+
+	return 0;
+}
+
+static int vpe_v6_1_process_trap_irq(struct amdgpu_device *adev,
+				     struct amdgpu_irq_src *source,
+				     struct amdgpu_iv_entry *entry)
+{
+
+	dev_dbg(adev->dev, "IH: VPE trap\n");
+
+	switch (entry->client_id) {
+	case SOC21_IH_CLIENTID_VPE:
+		amdgpu_fence_process(&adev->vpe.ring);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int vpe_v6_1_set_regs(struct amdgpu_vpe *vpe)
+{
+	struct amdgpu_device *adev = container_of(vpe, struct amdgpu_device, vpe);
+
+	vpe->regs.queue0_rb_rptr_lo = regVPEC_QUEUE0_RB_RPTR;
+	vpe->regs.queue0_rb_rptr_hi = regVPEC_QUEUE0_RB_RPTR_HI;
+	vpe->regs.queue0_rb_wptr_lo = regVPEC_QUEUE0_RB_WPTR;
+	vpe->regs.queue0_rb_wptr_hi = regVPEC_QUEUE0_RB_WPTR_HI;
+	vpe->regs.queue0_preempt = regVPEC_QUEUE0_PREEMPT;
+
+	if (amdgpu_ip_version(adev, VPE_HWIP, 0) == IP_VERSION(6, 1, 1))
+		vpe->regs.dpm_enable = regVPEC_PUB_DUMMY2_6_1_1;
+	else
+		vpe->regs.dpm_enable = regVPEC_PUB_DUMMY2;
+
+	vpe->regs.dpm_pratio = regVPEC_QUEUE6_DUMMY4;
+	vpe->regs.dpm_request_interval = regVPEC_QUEUE5_DUMMY3;
+	vpe->regs.dpm_decision_threshold = regVPEC_QUEUE5_DUMMY4;
+	vpe->regs.dpm_busy_clamp_threshold = regVPEC_QUEUE7_DUMMY2;
+	vpe->regs.dpm_idle_clamp_threshold = regVPEC_QUEUE7_DUMMY3;
+	vpe->regs.dpm_request_lv = regVPEC_QUEUE7_DUMMY1;
+	vpe->regs.context_indicator = regVPEC_QUEUE6_DUMMY3;
+
+	return 0;
+}
+
+static const struct vpe_funcs vpe_v6_1_funcs = {
+	.get_reg_offset = vpe_v6_1_get_reg_offset,
+	.set_regs = vpe_v6_1_set_regs,
+	.irq_init = vpe_v6_1_irq_init,
+	.init_microcode = amdgpu_vpe_init_microcode,
+	.load_microcode = vpe_v6_1_load_microcode,
+	.ring_init = amdgpu_vpe_ring_init,
+	.ring_start = vpe_v6_1_ring_start,
+	.ring_stop = vpe_v_6_1_ring_stop,
+	.ring_fini = amdgpu_vpe_ring_fini,
+};
+
+static const struct amdgpu_irq_src_funcs vpe_v6_1_trap_irq_funcs = {
+	.set = vpe_v6_1_set_trap_irq_state,
+	.process = vpe_v6_1_process_trap_irq,
+};
+
+void vpe_v6_1_set_funcs(struct amdgpu_vpe *vpe)
+{
+	vpe->funcs = &vpe_v6_1_funcs;
+	vpe->trap_irq.funcs = &vpe_v6_1_trap_irq_funcs;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/vpe_v6_1.h b/drivers/gpu/drm/amd/amdgpu/vpe_v6_1.h
new file mode 100644
index 000000000000..a9bea7905a77
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/vpe_v6_1.h
@@ -0,0 +1,29 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+#ifndef __VPE_V6_1_H__
+#define __VPE_V6_1_H__
+
+#include "amdgpu_vpe.h"
+
+void vpe_v6_1_set_funcs(struct amdgpu_vpe *vpe);
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdkfd/Kconfig b/drivers/gpu/drm/amd/amdkfd/Kconfig
index 93bd4eda0d94..16e12c9913f9 100644
--- a/drivers/gpu/drm/amd/amdkfd/Kconfig
+++ b/drivers/gpu/drm/amd/amdkfd/Kconfig
@@ -5,8 +5,7 @@
 
 config HSA_AMD
 	bool "HSA kernel driver for AMD GPU devices"
-	depends on DRM_AMDGPU && (X86_64 || ARM64 || PPC64)
-	imply AMD_IOMMU_V2 if X86_64
+	depends on DRM_AMDGPU && (X86_64 || ARM64 || PPC64 || (RISCV && 64BIT) || (LOONGARCH && 64BIT))
 	select HMM_MIRROR
 	select MMU_NOTIFIER
 	select DRM_AMDGPU_USERPTR
diff --git a/drivers/gpu/drm/amd/amdkfd/Makefile b/drivers/gpu/drm/amd/amdkfd/Makefile
index e758c2a24cd0..0ce08113c9f0 100644
--- a/drivers/gpu/drm/amd/amdkfd/Makefile
+++ b/drivers/gpu/drm/amd/amdkfd/Makefile
@@ -27,7 +27,6 @@ AMDKFD_FILES	:= $(AMDKFD_PATH)/kfd_module.o \
 		$(AMDKFD_PATH)/kfd_device.o \
 		$(AMDKFD_PATH)/kfd_chardev.o \
 		$(AMDKFD_PATH)/kfd_topology.o \
-		$(AMDKFD_PATH)/kfd_pasid.o \
 		$(AMDKFD_PATH)/kfd_doorbell.o \
 		$(AMDKFD_PATH)/kfd_flat_memory.o \
 		$(AMDKFD_PATH)/kfd_process.o \
@@ -38,6 +37,7 @@ AMDKFD_FILES	:= $(AMDKFD_PATH)/kfd_module.o \
 		$(AMDKFD_PATH)/kfd_mqd_manager_v9.o \
 		$(AMDKFD_PATH)/kfd_mqd_manager_v10.o \
 		$(AMDKFD_PATH)/kfd_mqd_manager_v11.o \
+		$(AMDKFD_PATH)/kfd_mqd_manager_v12.o \
 		$(AMDKFD_PATH)/kfd_kernel_queue.o \
 		$(AMDKFD_PATH)/kfd_packet_manager.o \
 		$(AMDKFD_PATH)/kfd_packet_manager_vi.o \
@@ -49,17 +49,16 @@ AMDKFD_FILES	:= $(AMDKFD_PATH)/kfd_module.o \
 		$(AMDKFD_PATH)/kfd_device_queue_manager_v9.o \
 		$(AMDKFD_PATH)/kfd_device_queue_manager_v10.o \
 		$(AMDKFD_PATH)/kfd_device_queue_manager_v11.o \
+		$(AMDKFD_PATH)/kfd_device_queue_manager_v12.o \
 		$(AMDKFD_PATH)/kfd_interrupt.o \
 		$(AMDKFD_PATH)/kfd_events.o \
 		$(AMDKFD_PATH)/cik_event_interrupt.o \
 		$(AMDKFD_PATH)/kfd_int_process_v9.o \
+		$(AMDKFD_PATH)/kfd_int_process_v10.o \
 		$(AMDKFD_PATH)/kfd_int_process_v11.o \
 		$(AMDKFD_PATH)/kfd_smi_events.o \
-		$(AMDKFD_PATH)/kfd_crat.o
-
-ifneq ($(CONFIG_AMD_IOMMU_V2),)
-AMDKFD_FILES += $(AMDKFD_PATH)/kfd_iommu.o
-endif
+		$(AMDKFD_PATH)/kfd_crat.o \
+		$(AMDKFD_PATH)/kfd_debug.o
 
 ifneq ($(CONFIG_DEBUG_FS),)
 AMDKFD_FILES += $(AMDKFD_PATH)/kfd_debugfs.o
diff --git a/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c b/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c
index 5c8023cba196..73acbe0b7c21 100644
--- a/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c
+++ b/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c
@@ -26,7 +26,7 @@
 #include "amdgpu_amdkfd.h"
 #include "kfd_smi_events.h"
 
-static bool cik_event_interrupt_isr(struct kfd_dev *dev,
+static bool cik_event_interrupt_isr(struct kfd_node *dev,
 					const uint32_t *ih_ring_entry,
 					uint32_t *patched_ihre,
 					bool *patched_flag)
@@ -85,13 +85,12 @@ static bool cik_event_interrupt_isr(struct kfd_dev *dev,
 		!amdgpu_no_queue_eviction_on_vm_fault);
 }
 
-static void cik_event_interrupt_wq(struct kfd_dev *dev,
+static void cik_event_interrupt_wq(struct kfd_node *dev,
 					const uint32_t *ih_ring_entry)
 {
 	const struct cik_ih_ring_entry *ihre =
 			(const struct cik_ih_ring_entry *)ih_ring_entry;
 	uint32_t context_id = ihre->data & 0xfffffff;
-	unsigned int vmid  = (ihre->ring_id & 0x0000ff00) >> 8;
 	u32 pasid = (ihre->ring_id & 0xffff0000) >> 16;
 
 	if (pasid == 0)
@@ -107,20 +106,26 @@ static void cik_event_interrupt_wq(struct kfd_dev *dev,
 		kfd_signal_hw_exception_event(pasid);
 	else if (ihre->source_id == CIK_INTSRC_GFX_PAGE_INV_FAULT ||
 		ihre->source_id == CIK_INTSRC_GFX_MEM_PROT_FAULT) {
+		struct kfd_process_device *pdd = NULL;
 		struct kfd_vm_fault_info info;
+		struct kfd_process *p;
 
 		kfd_smi_event_update_vmfault(dev, pasid);
-		kfd_dqm_evict_pasid(dev->dqm, pasid);
+		p = kfd_lookup_process_by_pasid(pasid, &pdd);
+		if (!pdd)
+			return;
+
+		kfd_evict_process_device(pdd);
 
 		memset(&info, 0, sizeof(info));
 		amdgpu_amdkfd_gpuvm_get_vm_fault_info(dev->adev, &info);
-		if (!info.page_addr && !info.status)
+		if (!info.page_addr && !info.status) {
+			kfd_unref_process(p);
 			return;
+		}
 
-		if (info.vmid == vmid)
-			kfd_signal_vm_fault_event(dev, pasid, &info);
-		else
-			kfd_signal_vm_fault_event(dev, pasid, NULL);
+		kfd_signal_vm_fault_event(pdd, &info, NULL);
+		kfd_unref_process(p);
 	}
 }
 
diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h
index 73ca9aebf086..0320163b6e74 100644
--- a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h
+++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h
@@ -274,16 +274,16 @@ static const uint32_t cwsr_trap_gfx8_hex[] = {
 
 
 static const uint32_t cwsr_trap_gfx9_hex[] = {
-	0xbf820001, 0xbf820254,
+	0xbf820001, 0xbf820259,
 	0xb8f8f802, 0x8978ff78,
 	0x00020006, 0xb8fbf803,
 	0x866eff78, 0x00002000,
 	0xbf840009, 0x866eff6d,
 	0x00ff0000, 0xbf85001e,
 	0x866eff7b, 0x00000400,
-	0xbf850051, 0xbf8e0010,
+	0xbf850055, 0xbf8e0010,
 	0xb8fbf803, 0xbf82fffa,
-	0x866eff7b, 0x00000900,
+	0x866eff7b, 0x03c00900,
 	0xbf850015, 0x866eff7b,
 	0x000071ff, 0xbf840008,
 	0x866fff7b, 0x00007080,
@@ -294,13 +294,15 @@ static const uint32_t cwsr_trap_gfx9_hex[] = {
 	0xbf850007, 0xb8eef801,
 	0x866eff6e, 0x00000800,
 	0xbf850003, 0x866eff7b,
-	0x00000400, 0xbf850036,
+	0x00000400, 0xbf85003a,
 	0xb8faf807, 0x867aff7a,
 	0x001f8000, 0x8e7a8b7a,
 	0x8977ff77, 0xfc000000,
 	0x87777a77, 0xba7ff807,
 	0x00000000, 0xb8faf812,
 	0xb8fbf813, 0x8efa887a,
+	0xbf0d8f7b, 0xbf840002,
+	0x877bff7b, 0xffff0000,
 	0xc0031bbd, 0x00000010,
 	0xbf8cc07f, 0x8e6e976e,
 	0x8977ff77, 0x00800000,
@@ -388,141 +390,98 @@ static const uint32_t cwsr_trap_gfx9_hex[] = {
 	0xbefe007c, 0xbefc0070,
 	0xc0611c7a, 0x0000007c,
 	0xbf8cc07f, 0x80708470,
-	0xbefc007e, 0x867aff7f,
-	0x04000000, 0xbeef0080,
-	0x876f6f7a, 0xb8f02a05,
-	0x80708170, 0x8e708a70,
-	0xb8fb1605, 0x807b817b,
-	0x8e7b847b, 0x8e76827b,
-	0xbef600ff, 0x01000000,
-	0xbef20174, 0x80747074,
-	0x82758075, 0xbefc0080,
-	0xbf800000, 0xbe802b00,
-	0xbe822b02, 0xbe842b04,
-	0xbe862b06, 0xbe882b08,
-	0xbe8a2b0a, 0xbe8c2b0c,
-	0xbe8e2b0e, 0xc06b003a,
-	0x00000000, 0xbf8cc07f,
-	0xc06b013a, 0x00000010,
-	0xbf8cc07f, 0xc06b023a,
-	0x00000020, 0xbf8cc07f,
-	0xc06b033a, 0x00000030,
-	0xbf8cc07f, 0x8074c074,
-	0x82758075, 0x807c907c,
-	0xbf0a7b7c, 0xbf85ffe7,
-	0xbef40172, 0xbef00080,
-	0xbefe00c1, 0xbeff00c1,
-	0xbee80080, 0xbee90080,
-	0xbef600ff, 0x01000000,
-	0x867aff78, 0x00400000,
-	0xbf850003, 0xb8faf803,
-	0x897a7aff, 0x10000000,
-	0xbf85004d, 0xbe840080,
-	0xd2890000, 0x00000900,
-	0x80048104, 0xd2890001,
-	0x00000900, 0x80048104,
-	0xd2890002, 0x00000900,
-	0x80048104, 0xd2890003,
-	0x00000900, 0x80048104,
-	0xc069003a, 0x00000070,
-	0xbf8cc07f, 0x80709070,
-	0xbf06c004, 0xbf84ffee,
+	0xbefc007e, 0xbf108080,
+	0x867aff7f, 0x04000000,
+	0xbeef0080, 0x876f6f7a,
+	0xb8f02a05, 0x80708170,
+	0x8e708a70, 0xb8fb1605,
+	0x807b817b, 0x8e7b847b,
+	0x8e76827b, 0xbef600ff,
+	0x01000000, 0xbef20174,
+	0x80747074, 0x82758075,
+	0xbefc0080, 0xbf800000,
+	0xbe802b00, 0xbe822b02,
+	0xbe842b04, 0xbe862b06,
+	0xbe882b08, 0xbe8a2b0a,
+	0xbe8c2b0c, 0xbe8e2b0e,
+	0xc06b003a, 0x00000000,
+	0xbf8cc07f, 0xc06b013a,
+	0x00000010, 0xbf8cc07f,
+	0xc06b023a, 0x00000020,
+	0xbf8cc07f, 0xc06b033a,
+	0x00000030, 0xbf8cc07f,
+	0x8074c074, 0x82758075,
+	0x807c907c, 0xbf0a7b7c,
+	0xbf85ffe7, 0xbef40172,
+	0xbef00080, 0xbefe00c1,
+	0xbeff00c1, 0xbee80080,
+	0xbee90080, 0xbef600ff,
+	0x01000000, 0x867aff78,
+	0x00400000, 0xbf850003,
+	0xb8faf803, 0x897a7aff,
+	0x10000000, 0xbf85004d,
 	0xbe840080, 0xd2890000,
-	0x00000901, 0x80048104,
-	0xd2890001, 0x00000901,
+	0x00000900, 0x80048104,
+	0xd2890001, 0x00000900,
 	0x80048104, 0xd2890002,
-	0x00000901, 0x80048104,
-	0xd2890003, 0x00000901,
+	0x00000900, 0x80048104,
+	0xd2890003, 0x00000900,
 	0x80048104, 0xc069003a,
 	0x00000070, 0xbf8cc07f,
 	0x80709070, 0xbf06c004,
 	0xbf84ffee, 0xbe840080,
-	0xd2890000, 0x00000902,
+	0xd2890000, 0x00000901,
 	0x80048104, 0xd2890001,
-	0x00000902, 0x80048104,
-	0xd2890002, 0x00000902,
+	0x00000901, 0x80048104,
+	0xd2890002, 0x00000901,
 	0x80048104, 0xd2890003,
-	0x00000902, 0x80048104,
+	0x00000901, 0x80048104,
 	0xc069003a, 0x00000070,
 	0xbf8cc07f, 0x80709070,
 	0xbf06c004, 0xbf84ffee,
 	0xbe840080, 0xd2890000,
-	0x00000903, 0x80048104,
-	0xd2890001, 0x00000903,
+	0x00000902, 0x80048104,
+	0xd2890001, 0x00000902,
 	0x80048104, 0xd2890002,
-	0x00000903, 0x80048104,
-	0xd2890003, 0x00000903,
+	0x00000902, 0x80048104,
+	0xd2890003, 0x00000902,
 	0x80048104, 0xc069003a,
 	0x00000070, 0xbf8cc07f,
 	0x80709070, 0xbf06c004,
-	0xbf84ffee, 0xbf820008,
-	0xe0724000, 0x701d0000,
-	0xe0724100, 0x701d0100,
-	0xe0724200, 0x701d0200,
-	0xe0724300, 0x701d0300,
-	0xbefe00c1, 0xbeff00c1,
-	0xb8fb4306, 0x867bc17b,
-	0xbf840063, 0xbf8a0000,
-	0x867aff6f, 0x04000000,
-	0xbf84005f, 0x8e7b867b,
-	0x8e7b827b, 0xbef6007b,
-	0xb8f02a05, 0x80708170,
-	0x8e708a70, 0xb8fa1605,
-	0x807a817a, 0x8e7a867a,
-	0x80707a70, 0x8070ff70,
-	0x00000080, 0xbef600ff,
-	0x01000000, 0xbefc0080,
-	0xd28c0002, 0x000100c1,
-	0xd28d0003, 0x000204c1,
-	0x867aff78, 0x00400000,
-	0xbf850003, 0xb8faf803,
-	0x897a7aff, 0x10000000,
-	0xbf850030, 0x24040682,
-	0xd86e4000, 0x00000002,
-	0xbf8cc07f, 0xbe840080,
-	0xd2890000, 0x00000900,
+	0xbf84ffee, 0xbe840080,
+	0xd2890000, 0x00000903,
 	0x80048104, 0xd2890001,
-	0x00000900, 0x80048104,
-	0xd2890002, 0x00000900,
+	0x00000903, 0x80048104,
+	0xd2890002, 0x00000903,
 	0x80048104, 0xd2890003,
-	0x00000900, 0x80048104,
+	0x00000903, 0x80048104,
 	0xc069003a, 0x00000070,
 	0xbf8cc07f, 0x80709070,
 	0xbf06c004, 0xbf84ffee,
-	0xbe840080, 0xd2890000,
-	0x00000901, 0x80048104,
-	0xd2890001, 0x00000901,
-	0x80048104, 0xd2890002,
-	0x00000901, 0x80048104,
-	0xd2890003, 0x00000901,
-	0x80048104, 0xc069003a,
-	0x00000070, 0xbf8cc07f,
-	0x80709070, 0xbf06c004,
-	0xbf84ffee, 0x680404ff,
-	0x00000200, 0xd0c9006a,
-	0x0000f702, 0xbf87ffd2,
-	0xbf820015, 0xd1060002,
-	0x00011103, 0x7e0602ff,
-	0x00000200, 0xbefc00ff,
-	0x00010000, 0xbe800077,
-	0x8677ff77, 0xff7fffff,
-	0x8777ff77, 0x00058000,
-	0xd8ec0000, 0x00000002,
-	0xbf8cc07f, 0xe0765000,
-	0x701d0002, 0x68040702,
-	0xd0c9006a, 0x0000f702,
-	0xbf87fff7, 0xbef70000,
-	0xbef000ff, 0x00000400,
-	0xbefe00c1, 0xbeff00c1,
-	0xb8fb2a05, 0x807b817b,
-	0x8e7b827b, 0xbef600ff,
-	0x01000000, 0xbefc0084,
-	0xbf0a7b7c, 0xbf84006d,
-	0xbf11017c, 0x807bff7b,
-	0x00001000, 0x867aff78,
+	0xbf820008, 0xe0724000,
+	0x701d0000, 0xe0724100,
+	0x701d0100, 0xe0724200,
+	0x701d0200, 0xe0724300,
+	0x701d0300, 0xbefe00c1,
+	0xbeff00c1, 0xb8fb4306,
+	0x867bc17b, 0xbf840063,
+	0xbf8a0000, 0x867aff6f,
+	0x04000000, 0xbf84005f,
+	0x8e7b867b, 0x8e7b827b,
+	0xbef6007b, 0xb8f02a05,
+	0x80708170, 0x8e708a70,
+	0xb8fa1605, 0x807a817a,
+	0x8e7a867a, 0x80707a70,
+	0x8070ff70, 0x00000080,
+	0xbef600ff, 0x01000000,
+	0xbefc0080, 0xd28c0002,
+	0x000100c1, 0xd28d0003,
+	0x000204c1, 0x867aff78,
 	0x00400000, 0xbf850003,
 	0xb8faf803, 0x897a7aff,
-	0x10000000, 0xbf850051,
+	0x10000000, 0xbf850030,
+	0x24040682, 0xd86e4000,
+	0x00000002, 0xbf8cc07f,
 	0xbe840080, 0xd2890000,
 	0x00000900, 0x80048104,
 	0xd2890001, 0x00000900,
@@ -542,148 +501,192 @@ static const uint32_t cwsr_trap_gfx9_hex[] = {
 	0xc069003a, 0x00000070,
 	0xbf8cc07f, 0x80709070,
 	0xbf06c004, 0xbf84ffee,
+	0x680404ff, 0x00000200,
+	0xd0c9006a, 0x0000f702,
+	0xbf87ffd2, 0xbf820015,
+	0xd1060002, 0x00011103,
+	0x7e0602ff, 0x00000200,
+	0xbefc00ff, 0x00010000,
+	0xbe800077, 0x8677ff77,
+	0xff7fffff, 0x8777ff77,
+	0x00058000, 0xd8ec0000,
+	0x00000002, 0xbf8cc07f,
+	0xe0765000, 0x701d0002,
+	0x68040702, 0xd0c9006a,
+	0x0000f702, 0xbf87fff7,
+	0xbef70000, 0xbef000ff,
+	0x00000400, 0xbefe00c1,
+	0xbeff00c1, 0xb8fb2a05,
+	0x807b817b, 0x8e7b827b,
+	0xbef600ff, 0x01000000,
+	0xbefc0084, 0xbf0a7b7c,
+	0xbf84006d, 0xbf11017c,
+	0x807bff7b, 0x00001000,
+	0x867aff78, 0x00400000,
+	0xbf850003, 0xb8faf803,
+	0x897a7aff, 0x10000000,
+	0xbf850051, 0xbe840080,
+	0xd2890000, 0x00000900,
+	0x80048104, 0xd2890001,
+	0x00000900, 0x80048104,
+	0xd2890002, 0x00000900,
+	0x80048104, 0xd2890003,
+	0x00000900, 0x80048104,
+	0xc069003a, 0x00000070,
+	0xbf8cc07f, 0x80709070,
+	0xbf06c004, 0xbf84ffee,
 	0xbe840080, 0xd2890000,
-	0x00000902, 0x80048104,
-	0xd2890001, 0x00000902,
+	0x00000901, 0x80048104,
+	0xd2890001, 0x00000901,
 	0x80048104, 0xd2890002,
-	0x00000902, 0x80048104,
-	0xd2890003, 0x00000902,
+	0x00000901, 0x80048104,
+	0xd2890003, 0x00000901,
 	0x80048104, 0xc069003a,
 	0x00000070, 0xbf8cc07f,
 	0x80709070, 0xbf06c004,
 	0xbf84ffee, 0xbe840080,
-	0xd2890000, 0x00000903,
+	0xd2890000, 0x00000902,
 	0x80048104, 0xd2890001,
-	0x00000903, 0x80048104,
-	0xd2890002, 0x00000903,
+	0x00000902, 0x80048104,
+	0xd2890002, 0x00000902,
 	0x80048104, 0xd2890003,
-	0x00000903, 0x80048104,
+	0x00000902, 0x80048104,
 	0xc069003a, 0x00000070,
 	0xbf8cc07f, 0x80709070,
 	0xbf06c004, 0xbf84ffee,
-	0x807c847c, 0xbf0a7b7c,
-	0xbf85ffb1, 0xbf9c0000,
-	0xbf820012, 0x7e000300,
-	0x7e020301, 0x7e040302,
-	0x7e060303, 0xe0724000,
-	0x701d0000, 0xe0724100,
-	0x701d0100, 0xe0724200,
-	0x701d0200, 0xe0724300,
-	0x701d0300, 0x807c847c,
-	0x8070ff70, 0x00000400,
-	0xbf0a7b7c, 0xbf85ffef,
-	0xbf9c0000, 0xbf8200c7,
-	0xbef4007e, 0x8675ff7f,
-	0x0000ffff, 0x8775ff75,
-	0x00040000, 0xbef60080,
-	0xbef700ff, 0x00807fac,
-	0x866eff7f, 0x04000000,
-	0xbf84001e, 0xbefe00c1,
-	0xbeff00c1, 0xb8ef4306,
-	0x866fc16f, 0xbf840019,
-	0x8e6f866f, 0x8e6f826f,
-	0xbef6006f, 0xb8f82a05,
-	0x80788178, 0x8e788a78,
-	0xb8ee1605, 0x806e816e,
-	0x8e6e866e, 0x80786e78,
-	0x8078ff78, 0x00000080,
-	0xbef600ff, 0x01000000,
-	0xbefc0080, 0xe0510000,
-	0x781d0000, 0xe0510100,
-	0x781d0000, 0x807cff7c,
-	0x00000200, 0x8078ff78,
-	0x00000200, 0xbf0a6f7c,
-	0xbf85fff6, 0xbefe00c1,
-	0xbeff00c1, 0xbef600ff,
-	0x01000000, 0xb8ef2a05,
-	0x806f816f, 0x8e6f826f,
-	0x806fff6f, 0x00008000,
-	0xbef80080, 0xbeee0078,
-	0x8078ff78, 0x00000400,
-	0xbefc0084, 0xbf11087c,
-	0xe0524000, 0x781d0000,
-	0xe0524100, 0x781d0100,
-	0xe0524200, 0x781d0200,
-	0xe0524300, 0x781d0300,
-	0xbf8c0f70, 0x7e000300,
-	0x7e020301, 0x7e040302,
-	0x7e060303, 0x807c847c,
-	0x8078ff78, 0x00000400,
-	0xbf0a6f7c, 0xbf85ffee,
-	0xbf9c0000, 0xe0524000,
-	0x6e1d0000, 0xe0524100,
-	0x6e1d0100, 0xe0524200,
-	0x6e1d0200, 0xe0524300,
-	0x6e1d0300, 0xbf8c0f70,
+	0xbe840080, 0xd2890000,
+	0x00000903, 0x80048104,
+	0xd2890001, 0x00000903,
+	0x80048104, 0xd2890002,
+	0x00000903, 0x80048104,
+	0xd2890003, 0x00000903,
+	0x80048104, 0xc069003a,
+	0x00000070, 0xbf8cc07f,
+	0x80709070, 0xbf06c004,
+	0xbf84ffee, 0x807c847c,
+	0xbf0a7b7c, 0xbf85ffb1,
+	0xbf9c0000, 0xbf820012,
+	0x7e000300, 0x7e020301,
+	0x7e040302, 0x7e060303,
+	0xe0724000, 0x701d0000,
+	0xe0724100, 0x701d0100,
+	0xe0724200, 0x701d0200,
+	0xe0724300, 0x701d0300,
+	0x807c847c, 0x8070ff70,
+	0x00000400, 0xbf0a7b7c,
+	0xbf85ffef, 0xbf9c0000,
+	0xbf8200c7, 0xbef4007e,
+	0x8675ff7f, 0x0000ffff,
+	0x8775ff75, 0x00040000,
+	0xbef60080, 0xbef700ff,
+	0x00807fac, 0x866eff7f,
+	0x04000000, 0xbf84001e,
+	0xbefe00c1, 0xbeff00c1,
+	0xb8ef4306, 0x866fc16f,
+	0xbf840019, 0x8e6f866f,
+	0x8e6f826f, 0xbef6006f,
 	0xb8f82a05, 0x80788178,
 	0x8e788a78, 0xb8ee1605,
 	0x806e816e, 0x8e6e866e,
-	0x80786e78, 0x80f8c078,
-	0xb8ef1605, 0x806f816f,
-	0x8e6f846f, 0x8e76826f,
+	0x80786e78, 0x8078ff78,
+	0x00000080, 0xbef600ff,
+	0x01000000, 0xbefc0080,
+	0xe0510000, 0x781d0000,
+	0xe0510100, 0x781d0000,
+	0x807cff7c, 0x00000200,
+	0x8078ff78, 0x00000200,
+	0xbf0a6f7c, 0xbf85fff6,
+	0xbefe00c1, 0xbeff00c1,
 	0xbef600ff, 0x01000000,
-	0xbefc006f, 0xc031003a,
-	0x00000078, 0x80f8c078,
-	0xbf8cc07f, 0x80fc907c,
-	0xbf800000, 0xbe802d00,
-	0xbe822d02, 0xbe842d04,
-	0xbe862d06, 0xbe882d08,
-	0xbe8a2d0a, 0xbe8c2d0c,
-	0xbe8e2d0e, 0xbf06807c,
-	0xbf84fff0, 0xb8f82a05,
+	0xb8ef2a05, 0x806f816f,
+	0x8e6f826f, 0x806fff6f,
+	0x00008000, 0xbef80080,
+	0xbeee0078, 0x8078ff78,
+	0x00000400, 0xbefc0084,
+	0xbf11087c, 0xe0524000,
+	0x781d0000, 0xe0524100,
+	0x781d0100, 0xe0524200,
+	0x781d0200, 0xe0524300,
+	0x781d0300, 0xbf8c0f70,
+	0x7e000300, 0x7e020301,
+	0x7e040302, 0x7e060303,
+	0x807c847c, 0x8078ff78,
+	0x00000400, 0xbf0a6f7c,
+	0xbf85ffee, 0xbf9c0000,
+	0xe0524000, 0x6e1d0000,
+	0xe0524100, 0x6e1d0100,
+	0xe0524200, 0x6e1d0200,
+	0xe0524300, 0x6e1d0300,
+	0xbf8c0f70, 0xb8f82a05,
 	0x80788178, 0x8e788a78,
 	0xb8ee1605, 0x806e816e,
 	0x8e6e866e, 0x80786e78,
-	0xbef60084, 0xbef600ff,
-	0x01000000, 0xc0211bfa,
+	0x80f8c078, 0xb8ef1605,
+	0x806f816f, 0x8e6f846f,
+	0x8e76826f, 0xbef600ff,
+	0x01000000, 0xbefc006f,
+	0xc031003a, 0x00000078,
+	0x80f8c078, 0xbf8cc07f,
+	0x80fc907c, 0xbf800000,
+	0xbe802d00, 0xbe822d02,
+	0xbe842d04, 0xbe862d06,
+	0xbe882d08, 0xbe8a2d0a,
+	0xbe8c2d0c, 0xbe8e2d0e,
+	0xbf06807c, 0xbf84fff0,
+	0xb8f82a05, 0x80788178,
+	0x8e788a78, 0xb8ee1605,
+	0x806e816e, 0x8e6e866e,
+	0x80786e78, 0xbef60084,
+	0xbef600ff, 0x01000000,
+	0xc0211bfa, 0x00000078,
+	0x80788478, 0xc0211b3a,
 	0x00000078, 0x80788478,
-	0xc0211b3a, 0x00000078,
-	0x80788478, 0xc0211b7a,
+	0xc0211b7a, 0x00000078,
+	0x80788478, 0xc0211c3a,
 	0x00000078, 0x80788478,
-	0xc0211c3a, 0x00000078,
-	0x80788478, 0xc0211c7a,
+	0xc0211c7a, 0x00000078,
+	0x80788478, 0xc0211eba,
 	0x00000078, 0x80788478,
-	0xc0211eba, 0x00000078,
-	0x80788478, 0xc0211efa,
+	0xc0211efa, 0x00000078,
+	0x80788478, 0xc0211a3a,
 	0x00000078, 0x80788478,
-	0xc0211a3a, 0x00000078,
-	0x80788478, 0xc0211a7a,
+	0xc0211a7a, 0x00000078,
+	0x80788478, 0xc0211cfa,
 	0x00000078, 0x80788478,
-	0xc0211cfa, 0x00000078,
-	0x80788478, 0xbf8cc07f,
-	0xbefc006f, 0xbefe0070,
-	0xbeff0071, 0x866f7bff,
-	0x000003ff, 0xb96f4803,
-	0x866f7bff, 0xfffff800,
-	0x8f6f8b6f, 0xb96fa2c3,
-	0xb973f801, 0xb8ee2a05,
-	0x806e816e, 0x8e6e8a6e,
-	0xb8ef1605, 0x806f816f,
-	0x8e6f866f, 0x806e6f6e,
-	0x806e746e, 0x826f8075,
-	0x866fff6f, 0x0000ffff,
-	0xc00b1c37, 0x00000050,
-	0xc00b1d37, 0x00000060,
-	0xc0031e77, 0x00000074,
-	0xbf8cc07f, 0x8f6e8b77,
-	0x866eff6e, 0x001f8000,
-	0xb96ef807, 0x866dff6d,
-	0x0000ffff, 0x86fe7e7e,
-	0x86ea6a6a, 0x8f6e837a,
-	0xb96ee0c2, 0xbf800002,
-	0xb97a0002, 0xbf8a0000,
-	0xbe801f6c, 0xbf810000,
+	0xbf8cc07f, 0xbefc006f,
+	0xbefe0070, 0xbeff0071,
+	0x866f7bff, 0x000003ff,
+	0xb96f4803, 0x866f7bff,
+	0xfffff800, 0x8f6f8b6f,
+	0xb96fa2c3, 0xb973f801,
+	0xb8ee2a05, 0x806e816e,
+	0x8e6e8a6e, 0xb8ef1605,
+	0x806f816f, 0x8e6f866f,
+	0x806e6f6e, 0x806e746e,
+	0x826f8075, 0x866fff6f,
+	0x0000ffff, 0xc00b1c37,
+	0x00000050, 0xc00b1d37,
+	0x00000060, 0xc0031e77,
+	0x00000074, 0xbf8cc07f,
+	0x8f6e8b77, 0x866eff6e,
+	0x001f8000, 0xb96ef807,
+	0x866dff6d, 0x0000ffff,
+	0x86fe7e7e, 0x86ea6a6a,
+	0x8f6e837a, 0xb96ee0c2,
+	0xbf800002, 0xb97a0002,
+	0xbf8a0000, 0xbe801f6c,
+	0xbf9b0000, 0x00000000,
 };
 
 static const uint32_t cwsr_trap_nv1x_hex[] = {
-	0xbf820001, 0xbf8201f1,
+	0xbf820001, 0xbf820393,
 	0xb0804004, 0xb978f802,
 	0x8a78ff78, 0x00020006,
 	0xb97bf803, 0x876eff78,
 	0x00002000, 0xbf840009,
 	0x876eff6d, 0x00ff0000,
 	0xbf85001e, 0x876eff7b,
-	0x00000400, 0xbf850057,
+	0x00000400, 0xbf85005b,
 	0xbf8e0010, 0xb97bf803,
 	0xbf82fffa, 0x876eff7b,
 	0x00000900, 0xbf850015,
@@ -697,7 +700,7 @@ static const uint32_t cwsr_trap_nv1x_hex[] = {
 	0xb96ef801, 0x876eff6e,
 	0x00000800, 0xbf850003,
 	0x876eff7b, 0x00000400,
-	0xbf85003c, 0x8a77ff77,
+	0xbf850040, 0x8a77ff77,
 	0xff000000, 0xb97af807,
 	0x877bff7a, 0x02000000,
 	0x8f7b867b, 0x88777b77,
@@ -706,20 +709,22 @@ static const uint32_t cwsr_trap_nv1x_hex[] = {
 	0x8a7aff7a, 0x023f8000,
 	0xb9faf807, 0xb97af812,
 	0xb97bf813, 0x8ffa887a,
+	0xbf0d8f7b, 0xbf840002,
+	0x887bff7b, 0xffff0000,
 	0xf4011bbd, 0xfa000010,
-	0xbf8cc07f, 0x8f6e976e,
+	0xbf8c0000, 0x8f6e976e,
 	0x8a77ff77, 0x00800000,
 	0x88776e77, 0xf4051bbd,
-	0xfa000000, 0xbf8cc07f,
+	0xfa000000, 0xbf8c0000,
 	0xf4051ebd, 0xfa000008,
-	0xbf8cc07f, 0x87ee6e6e,
+	0xbf8c0000, 0x87ee6e6e,
 	0xbf840001, 0xbe80206e,
-	0x876eff6d, 0x01ff0000,
-	0xbf850005, 0x8878ff78,
-	0x00002000, 0x80ec886c,
-	0x82ed806d, 0xbf820005,
-	0x876eff6d, 0x01000000,
-	0xbf850002, 0x806c846c,
+	0x876eff6d, 0x00ff0000,
+	0xbf850008, 0x876eff6d,
+	0x01000000, 0xbf850007,
+	0x8878ff78, 0x00002000,
+	0x80ec886c, 0x82ed806d,
+	0xbf820002, 0x806c846c,
 	0x826d806d, 0x876dff6d,
 	0x0000ffff, 0x907a8977,
 	0x877bff7a, 0x003f8000,
@@ -765,13 +770,90 @@ static const uint32_t cwsr_trap_nv1x_hex[] = {
 	0x877c817c, 0xbf06817c,
 	0xbf850002, 0xbeff0380,
 	0xbf820002, 0xbeff03c1,
-	0xbf82000b, 0xbef603ff,
-	0x01000000, 0xe0704000,
-	0x705d0000, 0xe0704080,
-	0x705d0100, 0xe0704100,
-	0x705d0200, 0xe0704180,
-	0x705d0300, 0xbf82000a,
-	0xbef603ff, 0x01000000,
+	0xbf820058, 0xbef603ff,
+	0x01000000, 0xb97af803,
+	0x8a7a7aff, 0x10000000,
+	0xbf850049, 0xbe840380,
+	0xd7600000, 0x00000900,
+	0x80048104, 0xd7600001,
+	0x00000900, 0x80048104,
+	0xd7600002, 0x00000900,
+	0x80048104, 0xd7600003,
+	0x00000900, 0x80048104,
+	0xf469003a, 0xe0000000,
+	0x80709070, 0xbf06a004,
+	0xbf84ffef, 0xbe840380,
+	0xd7600000, 0x00000901,
+	0x80048104, 0xd7600001,
+	0x00000901, 0x80048104,
+	0xd7600002, 0x00000901,
+	0x80048104, 0xd7600003,
+	0x00000901, 0x80048104,
+	0xf469003a, 0xe0000000,
+	0x80709070, 0xbf06a004,
+	0xbf84ffef, 0xbe840380,
+	0xd7600000, 0x00000902,
+	0x80048104, 0xd7600001,
+	0x00000902, 0x80048104,
+	0xd7600002, 0x00000902,
+	0x80048104, 0xd7600003,
+	0x00000902, 0x80048104,
+	0xf469003a, 0xe0000000,
+	0x80709070, 0xbf06a004,
+	0xbf84ffef, 0xbe840380,
+	0xd7600000, 0x00000903,
+	0x80048104, 0xd7600001,
+	0x00000903, 0x80048104,
+	0xd7600002, 0x00000903,
+	0x80048104, 0xd7600003,
+	0x00000903, 0x80048104,
+	0xf469003a, 0xe0000000,
+	0x80709070, 0xbf06a004,
+	0xbf84ffef, 0xbf820060,
+	0xe0704000, 0x705d0000,
+	0xe0704080, 0x705d0100,
+	0xe0704100, 0x705d0200,
+	0xe0704180, 0x705d0300,
+	0xbf820057, 0xbef603ff,
+	0x01000000, 0xb97af803,
+	0x8a7a7aff, 0x10000000,
+	0xbf850049, 0xbe840380,
+	0xd7600000, 0x00000900,
+	0x80048104, 0xd7600001,
+	0x00000900, 0x80048104,
+	0xd7600002, 0x00000900,
+	0x80048104, 0xd7600003,
+	0x00000900, 0x80048104,
+	0xf469003a, 0xe0000000,
+	0x80709070, 0xbf06c004,
+	0xbf84ffef, 0xbe840380,
+	0xd7600000, 0x00000901,
+	0x80048104, 0xd7600001,
+	0x00000901, 0x80048104,
+	0xd7600002, 0x00000901,
+	0x80048104, 0xd7600003,
+	0x00000901, 0x80048104,
+	0xf469003a, 0xe0000000,
+	0x80709070, 0xbf06c004,
+	0xbf84ffef, 0xbe840380,
+	0xd7600000, 0x00000902,
+	0x80048104, 0xd7600001,
+	0x00000902, 0x80048104,
+	0xd7600002, 0x00000902,
+	0x80048104, 0xd7600003,
+	0x00000902, 0x80048104,
+	0xf469003a, 0xe0000000,
+	0x80709070, 0xbf06c004,
+	0xbf84ffef, 0xbe840380,
+	0xd7600000, 0x00000903,
+	0x80048104, 0xd7600001,
+	0x00000903, 0x80048104,
+	0xd7600002, 0x00000903,
+	0x80048104, 0xd7600003,
+	0x00000903, 0x80048104,
+	0xf469003a, 0xe0000000,
+	0x80709070, 0xbf06c004,
+	0xbf84ffef, 0xbf820008,
 	0xe0704000, 0x705d0000,
 	0xe0704100, 0x705d0100,
 	0xe0704200, 0x705d0200,
@@ -851,259 +933,386 @@ static const uint32_t cwsr_trap_nv1x_hex[] = {
 	0xbf850002, 0xbeff0380,
 	0xbf820001, 0xbeff03c1,
 	0xb97b4306, 0x877bc17b,
-	0xbf840044, 0xbf8a0000,
+	0xbf840085, 0xbf8a0000,
 	0x877aff6d, 0x80000000,
-	0xbf840040, 0x8f7b867b,
-	0x8f7b827b, 0xbef6037b,
-	0xb9703a05, 0x80708170,
-	0xbf0d9973, 0xbf850002,
-	0x8f708970, 0xbf820001,
-	0x8f708a70, 0xb97a1e06,
-	0x8f7a8a7a, 0x80707a70,
-	0x8070ff70, 0x00000200,
-	0x8070ff70, 0x00000080,
-	0xbef603ff, 0x01000000,
-	0xd7650000, 0x000100c1,
-	0xd7660000, 0x000200c1,
-	0x16000084, 0x907c9973,
-	0x877c817c, 0xbf06817c,
-	0xbefc0380, 0xbf850012,
-	0xbe8303ff, 0x00000080,
-	0xbf800000, 0xbf800000,
-	0xbf800000, 0xd8d80000,
+	0xbf840081, 0x8f7b887b,
+	0xbef6037b, 0xb9703a05,
+	0x80708170, 0xbf0d9973,
+	0xbf850002, 0x8f708970,
+	0xbf820001, 0x8f708a70,
+	0xb97a1e06, 0x8f7a8a7a,
+	0x80707a70, 0x8070ff70,
+	0x00000200, 0x8070ff70,
+	0x00000080, 0xbef603ff,
+	0x01000000, 0xd7650000,
+	0x000100c1, 0xd7660000,
+	0x000200c1, 0x16000084,
+	0x907c9973, 0x877c817c,
+	0xbf06817c, 0xbefc0380,
+	0xbf850033, 0xb97af803,
+	0x8a7a7aff, 0x10000000,
+	0xbf85001d, 0xd8d80000,
 	0x01000000, 0xbf8c0000,
-	0xe0704000, 0x705d0100,
-	0x807c037c, 0x80700370,
+	0xbe840380, 0xd7600000,
+	0x00000901, 0x80048104,
+	0xd7600001, 0x00000901,
+	0x80048104, 0xd7600002,
+	0x00000901, 0x80048104,
+	0xd7600003, 0x00000901,
+	0x80048104, 0xf469003a,
+	0xe0000000, 0x80709070,
+	0xbf06a004, 0xbf84ffef,
+	0x807cff7c, 0x00000080,
 	0xd5250000, 0x0001ff00,
 	0x00000080, 0xbf0a7b7c,
-	0xbf85fff4, 0xbf820011,
-	0xbe8303ff, 0x00000100,
+	0xbf85ffe4, 0xbf820044,
+	0xbe8303ff, 0x00000080,
 	0xbf800000, 0xbf800000,
 	0xbf800000, 0xd8d80000,
 	0x01000000, 0xbf8c0000,
 	0xe0704000, 0x705d0100,
 	0x807c037c, 0x80700370,
 	0xd5250000, 0x0001ff00,
-	0x00000100, 0xbf0a7b7c,
-	0xbf85fff4, 0xbefe03c1,
-	0x907c9973, 0x877c817c,
-	0xbf06817c, 0xbf850004,
-	0xbef003ff, 0x00000200,
-	0xbeff0380, 0xbf820003,
-	0xbef003ff, 0x00000400,
-	0xbeff03c1, 0xb97b3a05,
-	0x807b817b, 0x8f7b827b,
-	0x907c9973, 0x877c817c,
-	0xbf06817c, 0xbf850017,
+	0x00000080, 0xbf0a7b7c,
+	0xbf85fff4, 0xbf820032,
+	0xb97af803, 0x8a7a7aff,
+	0x10000000, 0xbf85001d,
+	0xd8d80000, 0x01000000,
+	0xbf8c0000, 0xbe840380,
+	0xd7600000, 0x00000901,
+	0x80048104, 0xd7600001,
+	0x00000901, 0x80048104,
+	0xd7600002, 0x00000901,
+	0x80048104, 0xd7600003,
+	0x00000901, 0x80048104,
+	0xf469003a, 0xe0000000,
+	0x80709070, 0xbf06c004,
+	0xbf84ffef, 0x807cff7c,
+	0x00000100, 0xd5250000,
+	0x0001ff00, 0x00000100,
+	0xbf0a7b7c, 0xbf85ffe4,
+	0xbf820011, 0xbe8303ff,
+	0x00000100, 0xbf800000,
+	0xbf800000, 0xbf800000,
+	0xd8d80000, 0x01000000,
+	0xbf8c0000, 0xe0704000,
+	0x705d0100, 0x807c037c,
+	0x80700370, 0xd5250000,
+	0x0001ff00, 0x00000100,
+	0xbf0a7b7c, 0xbf85fff4,
+	0xbefe03c1, 0x907c9973,
+	0x877c817c, 0xbf06817c,
+	0xbf850004, 0xbef003ff,
+	0x00000200, 0xbeff0380,
+	0xbf820003, 0xbef003ff,
+	0x00000400, 0xbeff03c1,
+	0xb97b3a05, 0x807b817b,
+	0x8f7b827b, 0x907c9973,
+	0x877c817c, 0xbf06817c,
+	0xbf85006b, 0xbef603ff,
+	0x01000000, 0xbefc0384,
+	0xbf0a7b7c, 0xbf8400fa,
+	0xb97af803, 0x8a7a7aff,
+	0x10000000, 0xbf850050,
+	0x7e008700, 0x7e028701,
+	0x7e048702, 0x7e068703,
+	0xbe840380, 0xd7600000,
+	0x00000900, 0x80048104,
+	0xd7600001, 0x00000900,
+	0x80048104, 0xd7600002,
+	0x00000900, 0x80048104,
+	0xd7600003, 0x00000900,
+	0x80048104, 0xf469003a,
+	0xe0000000, 0x80709070,
+	0xbf06a004, 0xbf84ffef,
+	0xbe840380, 0xd7600000,
+	0x00000901, 0x80048104,
+	0xd7600001, 0x00000901,
+	0x80048104, 0xd7600002,
+	0x00000901, 0x80048104,
+	0xd7600003, 0x00000901,
+	0x80048104, 0xf469003a,
+	0xe0000000, 0x80709070,
+	0xbf06a004, 0xbf84ffef,
+	0xbe840380, 0xd7600000,
+	0x00000902, 0x80048104,
+	0xd7600001, 0x00000902,
+	0x80048104, 0xd7600002,
+	0x00000902, 0x80048104,
+	0xd7600003, 0x00000902,
+	0x80048104, 0xf469003a,
+	0xe0000000, 0x80709070,
+	0xbf06a004, 0xbf84ffef,
+	0xbe840380, 0xd7600000,
+	0x00000903, 0x80048104,
+	0xd7600001, 0x00000903,
+	0x80048104, 0xd7600002,
+	0x00000903, 0x80048104,
+	0xd7600003, 0x00000903,
+	0x80048104, 0xf469003a,
+	0xe0000000, 0x80709070,
+	0xbf06a004, 0xbf84ffef,
+	0x807c847c, 0xbf0a7b7c,
+	0xbf85ffb1, 0xbf8200a6,
+	0x7e008700, 0x7e028701,
+	0x7e048702, 0x7e068703,
+	0xe0704000, 0x705d0000,
+	0xe0704080, 0x705d0100,
+	0xe0704100, 0x705d0200,
+	0xe0704180, 0x705d0300,
+	0x807c847c, 0x8070ff70,
+	0x00000200, 0xbf0a7b7c,
+	0xbf85ffef, 0xbf820094,
 	0xbef603ff, 0x01000000,
 	0xbefc0384, 0xbf0a7b7c,
-	0xbf840037, 0x7e008700,
+	0xbf840065, 0xb97af803,
+	0x8a7a7aff, 0x10000000,
+	0xbf850050, 0x7e008700,
+	0x7e028701, 0x7e048702,
+	0x7e068703, 0xbe840380,
+	0xd7600000, 0x00000900,
+	0x80048104, 0xd7600001,
+	0x00000900, 0x80048104,
+	0xd7600002, 0x00000900,
+	0x80048104, 0xd7600003,
+	0x00000900, 0x80048104,
+	0xf469003a, 0xe0000000,
+	0x80709070, 0xbf06c004,
+	0xbf84ffef, 0xbe840380,
+	0xd7600000, 0x00000901,
+	0x80048104, 0xd7600001,
+	0x00000901, 0x80048104,
+	0xd7600002, 0x00000901,
+	0x80048104, 0xd7600003,
+	0x00000901, 0x80048104,
+	0xf469003a, 0xe0000000,
+	0x80709070, 0xbf06c004,
+	0xbf84ffef, 0xbe840380,
+	0xd7600000, 0x00000902,
+	0x80048104, 0xd7600001,
+	0x00000902, 0x80048104,
+	0xd7600002, 0x00000902,
+	0x80048104, 0xd7600003,
+	0x00000902, 0x80048104,
+	0xf469003a, 0xe0000000,
+	0x80709070, 0xbf06c004,
+	0xbf84ffef, 0xbe840380,
+	0xd7600000, 0x00000903,
+	0x80048104, 0xd7600001,
+	0x00000903, 0x80048104,
+	0xd7600002, 0x00000903,
+	0x80048104, 0xd7600003,
+	0x00000903, 0x80048104,
+	0xf469003a, 0xe0000000,
+	0x80709070, 0xbf06c004,
+	0xbf84ffef, 0x807c847c,
+	0xbf0a7b7c, 0xbf85ffb1,
+	0xbf82003b, 0x7e008700,
 	0x7e028701, 0x7e048702,
 	0x7e068703, 0xe0704000,
-	0x705d0000, 0xe0704080,
-	0x705d0100, 0xe0704100,
-	0x705d0200, 0xe0704180,
+	0x705d0000, 0xe0704100,
+	0x705d0100, 0xe0704200,
+	0x705d0200, 0xe0704300,
 	0x705d0300, 0x807c847c,
-	0x8070ff70, 0x00000200,
+	0x8070ff70, 0x00000400,
 	0xbf0a7b7c, 0xbf85ffef,
-	0xbf820025, 0xbef603ff,
-	0x01000000, 0xbefc0384,
-	0xbf0a7b7c, 0xbf840011,
-	0x7e008700, 0x7e028701,
-	0x7e048702, 0x7e068703,
-	0xe0704000, 0x705d0000,
-	0xe0704100, 0x705d0100,
-	0xe0704200, 0x705d0200,
-	0xe0704300, 0x705d0300,
-	0x807c847c, 0x8070ff70,
-	0x00000400, 0xbf0a7b7c,
-	0xbf85ffef, 0xb97b1e06,
-	0x877bc17b, 0xbf84000c,
-	0x8f7b837b, 0x807b7c7b,
-	0xbefe03c1, 0xbeff0380,
+	0xb97b1e06, 0x877bc17b,
+	0xbf840027, 0x8f7b837b,
+	0x807b7c7b, 0xbefe03c1,
+	0xbeff0380, 0xb97af803,
+	0x8a7a7aff, 0x10000000,
+	0xbf850017, 0x7e008700,
+	0xbe840380, 0xd7600000,
+	0x00000900, 0x80048104,
+	0xd7600001, 0x00000900,
+	0x80048104, 0xd7600002,
+	0x00000900, 0x80048104,
+	0xd7600003, 0x00000900,
+	0x80048104, 0xf469003a,
+	0xe0000000, 0x80709070,
+	0xbf06c004, 0xbf84ffef,
+	0x807c817c, 0xbf0a7b7c,
+	0xbf85ffea, 0xbf820008,
 	0x7e008700, 0xe0704000,
 	0x705d0000, 0x807c817c,
 	0x8070ff70, 0x00000080,
 	0xbf0a7b7c, 0xbf85fff8,
-	0xbf820144, 0xbef4037e,
+	0xbf82013f, 0xbef4037e,
 	0x8775ff7f, 0x0000ffff,
 	0x8875ff75, 0x00040000,
 	0xbef60380, 0xbef703ff,
 	0x10807fac, 0xb97202dc,
 	0x8f729972, 0x876eff7f,
-	0x04000000, 0xbf840034,
+	0x04000000, 0xbf840033,
 	0xbefe03c1, 0x907c9972,
 	0x877c817c, 0xbf06817c,
 	0xbf850002, 0xbeff0380,
 	0xbf820001, 0xbeff03c1,
 	0xb96f4306, 0x876fc16f,
-	0xbf840029, 0x8f6f866f,
-	0x8f6f826f, 0xbef6036f,
-	0xb9783a05, 0x80788178,
-	0xbf0d9972, 0xbf850002,
-	0x8f788978, 0xbf820001,
-	0x8f788a78, 0xb96e1e06,
-	0x8f6e8a6e, 0x80786e78,
-	0x8078ff78, 0x00000200,
-	0x8078ff78, 0x00000080,
-	0xbef603ff, 0x01000000,
-	0x907c9972, 0x877c817c,
-	0xbf06817c, 0xbefc0380,
-	0xbf850009, 0xe0310000,
-	0x781d0000, 0x807cff7c,
-	0x00000080, 0x8078ff78,
-	0x00000080, 0xbf0a6f7c,
-	0xbf85fff8, 0xbf820008,
+	0xbf840028, 0x8f6f886f,
+	0xbef6036f, 0xb9783a05,
+	0x80788178, 0xbf0d9972,
+	0xbf850002, 0x8f788978,
+	0xbf820001, 0x8f788a78,
+	0xb96e1e06, 0x8f6e8a6e,
+	0x80786e78, 0x8078ff78,
+	0x00000200, 0x8078ff78,
+	0x00000080, 0xbef603ff,
+	0x01000000, 0x907c9972,
+	0x877c817c, 0xbf06817c,
+	0xbefc0380, 0xbf850009,
 	0xe0310000, 0x781d0000,
-	0x807cff7c, 0x00000100,
-	0x8078ff78, 0x00000100,
+	0x807cff7c, 0x00000080,
+	0x8078ff78, 0x00000080,
 	0xbf0a6f7c, 0xbf85fff8,
-	0xbef80380, 0xbefe03c1,
-	0x907c9972, 0x877c817c,
-	0xbf06817c, 0xbf850002,
-	0xbeff0380, 0xbf820001,
-	0xbeff03c1, 0xb96f3a05,
-	0x806f816f, 0x8f6f826f,
-	0x907c9972, 0x877c817c,
-	0xbf06817c, 0xbf850024,
-	0xbef603ff, 0x01000000,
-	0xbeee0378, 0x8078ff78,
-	0x00000200, 0xbefc0384,
-	0xbf0a6f7c, 0xbf840050,
-	0xe0304000, 0x785d0000,
-	0xe0304080, 0x785d0100,
-	0xe0304100, 0x785d0200,
-	0xe0304180, 0x785d0300,
-	0xbf8c3f70, 0x7e008500,
-	0x7e028501, 0x7e048502,
-	0x7e068503, 0x807c847c,
+	0xbf820008, 0xe0310000,
+	0x781d0000, 0x807cff7c,
+	0x00000100, 0x8078ff78,
+	0x00000100, 0xbf0a6f7c,
+	0xbf85fff8, 0xbef80380,
+	0xbefe03c1, 0x907c9972,
+	0x877c817c, 0xbf06817c,
+	0xbf850002, 0xbeff0380,
+	0xbf820001, 0xbeff03c1,
+	0xb96f3a05, 0x806f816f,
+	0x8f6f826f, 0x907c9972,
+	0x877c817c, 0xbf06817c,
+	0xbf850024, 0xbef603ff,
+	0x01000000, 0xbeee0378,
 	0x8078ff78, 0x00000200,
-	0xbf0a6f7c, 0xbf85ffee,
-	0xe0304000, 0x6e5d0000,
-	0xe0304080, 0x6e5d0100,
-	0xe0304100, 0x6e5d0200,
-	0xe0304180, 0x6e5d0300,
-	0xbf8c3f70, 0xbf820034,
-	0xbef603ff, 0x01000000,
-	0xbeee0378, 0x8078ff78,
-	0x00000400, 0xbefc0384,
-	0xbf0a6f7c, 0xbf840012,
-	0xe0304000, 0x785d0000,
-	0xe0304100, 0x785d0100,
-	0xe0304200, 0x785d0200,
-	0xe0304300, 0x785d0300,
-	0xbf8c3f70, 0x7e008500,
-	0x7e028501, 0x7e048502,
-	0x7e068503, 0x807c847c,
+	0xbefc0384, 0xbf0a6f7c,
+	0xbf840050, 0xe0304000,
+	0x785d0000, 0xe0304080,
+	0x785d0100, 0xe0304100,
+	0x785d0200, 0xe0304180,
+	0x785d0300, 0xbf8c0000,
+	0x7e008500, 0x7e028501,
+	0x7e048502, 0x7e068503,
+	0x807c847c, 0x8078ff78,
+	0x00000200, 0xbf0a6f7c,
+	0xbf85ffee, 0xe0304000,
+	0x6e5d0000, 0xe0304080,
+	0x6e5d0100, 0xe0304100,
+	0x6e5d0200, 0xe0304180,
+	0x6e5d0300, 0xbf8c0000,
+	0xbf820034, 0xbef603ff,
+	0x01000000, 0xbeee0378,
 	0x8078ff78, 0x00000400,
-	0xbf0a6f7c, 0xbf85ffee,
-	0xb96f1e06, 0x876fc16f,
-	0xbf84000e, 0x8f6f836f,
-	0x806f7c6f, 0xbefe03c1,
-	0xbeff0380, 0xe0304000,
-	0x785d0000, 0xbf8c3f70,
-	0x7e008500, 0x807c817c,
-	0x8078ff78, 0x00000080,
-	0xbf0a6f7c, 0xbf85fff7,
-	0xbeff03c1, 0xe0304000,
-	0x6e5d0000, 0xe0304100,
-	0x6e5d0100, 0xe0304200,
-	0x6e5d0200, 0xe0304300,
-	0x6e5d0300, 0xbf8c3f70,
+	0xbefc0384, 0xbf0a6f7c,
+	0xbf840012, 0xe0304000,
+	0x785d0000, 0xe0304100,
+	0x785d0100, 0xe0304200,
+	0x785d0200, 0xe0304300,
+	0x785d0300, 0xbf8c0000,
+	0x7e008500, 0x7e028501,
+	0x7e048502, 0x7e068503,
+	0x807c847c, 0x8078ff78,
+	0x00000400, 0xbf0a6f7c,
+	0xbf85ffee, 0xb96f1e06,
+	0x876fc16f, 0xbf84000e,
+	0x8f6f836f, 0x806f7c6f,
+	0xbefe03c1, 0xbeff0380,
+	0xe0304000, 0x785d0000,
+	0xbf8c0000, 0x7e008500,
+	0x807c817c, 0x8078ff78,
+	0x00000080, 0xbf0a6f7c,
+	0xbf85fff7, 0xbeff03c1,
+	0xe0304000, 0x6e5d0000,
+	0xe0304100, 0x6e5d0100,
+	0xe0304200, 0x6e5d0200,
+	0xe0304300, 0x6e5d0300,
+	0xbf8c0000, 0xb9783a05,
+	0x80788178, 0xbf0d9972,
+	0xbf850002, 0x8f788978,
+	0xbf820001, 0x8f788a78,
+	0xb96e1e06, 0x8f6e8a6e,
+	0x80786e78, 0x8078ff78,
+	0x00000200, 0x80f8ff78,
+	0x00000050, 0xbef603ff,
+	0x01000000, 0xbefc03ff,
+	0x0000006c, 0x80f89078,
+	0xf429003a, 0xf0000000,
+	0xbf8c0000, 0x80fc847c,
+	0xbf800000, 0xbe803100,
+	0xbe823102, 0x80f8a078,
+	0xf42d003a, 0xf0000000,
+	0xbf8c0000, 0x80fc887c,
+	0xbf800000, 0xbe803100,
+	0xbe823102, 0xbe843104,
+	0xbe863106, 0x80f8c078,
+	0xf431003a, 0xf0000000,
+	0xbf8c0000, 0x80fc907c,
+	0xbf800000, 0xbe803100,
+	0xbe823102, 0xbe843104,
+	0xbe863106, 0xbe883108,
+	0xbe8a310a, 0xbe8c310c,
+	0xbe8e310e, 0xbf06807c,
+	0xbf84fff0, 0xba80f801,
+	0x00000000, 0xbf8a0000,
 	0xb9783a05, 0x80788178,
 	0xbf0d9972, 0xbf850002,
 	0x8f788978, 0xbf820001,
 	0x8f788a78, 0xb96e1e06,
 	0x8f6e8a6e, 0x80786e78,
 	0x8078ff78, 0x00000200,
-	0x80f8ff78, 0x00000050,
 	0xbef603ff, 0x01000000,
-	0xbefc03ff, 0x0000006c,
-	0x80f89078, 0xf429003a,
-	0xf0000000, 0xbf8cc07f,
-	0x80fc847c, 0xbf800000,
-	0xbe803100, 0xbe823102,
-	0x80f8a078, 0xf42d003a,
-	0xf0000000, 0xbf8cc07f,
-	0x80fc887c, 0xbf800000,
-	0xbe803100, 0xbe823102,
-	0xbe843104, 0xbe863106,
-	0x80f8c078, 0xf431003a,
-	0xf0000000, 0xbf8cc07f,
-	0x80fc907c, 0xbf800000,
-	0xbe803100, 0xbe823102,
-	0xbe843104, 0xbe863106,
-	0xbe883108, 0xbe8a310a,
-	0xbe8c310c, 0xbe8e310e,
-	0xbf06807c, 0xbf84fff0,
-	0xba80f801, 0x00000000,
-	0xbf8a0000, 0xb9783a05,
-	0x80788178, 0xbf0d9972,
-	0xbf850002, 0x8f788978,
-	0xbf820001, 0x8f788a78,
-	0xb96e1e06, 0x8f6e8a6e,
-	0x80786e78, 0x8078ff78,
-	0x00000200, 0xbef603ff,
-	0x01000000, 0xf4211bfa,
+	0xf4211bfa, 0xf0000000,
+	0x80788478, 0xf4211b3a,
 	0xf0000000, 0x80788478,
-	0xf4211b3a, 0xf0000000,
-	0x80788478, 0xf4211b7a,
+	0xf4211b7a, 0xf0000000,
+	0x80788478, 0xf4211c3a,
 	0xf0000000, 0x80788478,
-	0xf4211c3a, 0xf0000000,
-	0x80788478, 0xf4211c7a,
+	0xf4211c7a, 0xf0000000,
+	0x80788478, 0xf4211eba,
 	0xf0000000, 0x80788478,
-	0xf4211eba, 0xf0000000,
-	0x80788478, 0xf4211efa,
+	0xf4211efa, 0xf0000000,
+	0x80788478, 0xf4211e7a,
 	0xf0000000, 0x80788478,
-	0xf4211e7a, 0xf0000000,
-	0x80788478, 0xf4211cfa,
+	0xf4211cfa, 0xf0000000,
+	0x80788478, 0xf4211bba,
 	0xf0000000, 0x80788478,
+	0xbf8c0000, 0xb9eef814,
 	0xf4211bba, 0xf0000000,
-	0x80788478, 0xbf8cc07f,
-	0xb9eef814, 0xf4211bba,
-	0xf0000000, 0x80788478,
-	0xbf8cc07f, 0xb9eef815,
-	0xbefc036f, 0xbefe0370,
-	0xbeff0371, 0x876f7bff,
-	0x000003ff, 0xb9ef4803,
-	0xb9f9f816, 0x876f7bff,
-	0xfffff800, 0x906f8b6f,
-	0xb9efa2c3, 0xb9f3f801,
-	0xb96e3a05, 0x806e816e,
-	0xbf0d9972, 0xbf850002,
-	0x8f6e896e, 0xbf820001,
-	0x8f6e8a6e, 0xb96f1e06,
-	0x8f6f8a6f, 0x806e6f6e,
-	0x806eff6e, 0x00000200,
-	0x806e746e, 0x826f8075,
-	0x876fff6f, 0x0000ffff,
-	0xf4091c37, 0xfa000050,
-	0xf4091d37, 0xfa000060,
-	0xf4011e77, 0xfa000074,
-	0xbf8cc07f, 0x906e8977,
-	0x876fff6e, 0x003f8000,
-	0x906e8677, 0x876eff6e,
-	0x02000000, 0x886e6f6e,
-	0xb9eef807, 0x876dff6d,
-	0x0000ffff, 0x87fe7e7e,
-	0x87ea6a6a, 0xb9faf802,
-	0xbe80226c, 0xbf810000,
+	0x80788478, 0xbf8c0000,
+	0xb9eef815, 0xbefc036f,
+	0xbefe0370, 0xbeff0371,
+	0xb9f9f816, 0xb9fb4803,
+	0x907b8b7b, 0xb9fba2c3,
+	0xb9f3f801, 0xb96e3a05,
+	0x806e816e, 0xbf0d9972,
+	0xbf850002, 0x8f6e896e,
+	0xbf820001, 0x8f6e8a6e,
+	0xb96f1e06, 0x8f6f8a6f,
+	0x806e6f6e, 0x806eff6e,
+	0x00000200, 0x806e746e,
+	0x826f8075, 0x876fff6f,
+	0x0000ffff, 0xf4091c37,
+	0xfa000050, 0xf4091d37,
+	0xfa000060, 0xf4011e77,
+	0xfa000074, 0xbf8c0000,
+	0x906e8977, 0x876fff6e,
+	0x003f8000, 0x906e8677,
+	0x876eff6e, 0x02000000,
+	0x886e6f6e, 0xb9eef807,
+	0x876dff6d, 0x0000ffff,
+	0x87fe7e7e, 0x87ea6a6a,
+	0xb9faf802, 0xbe80226c,
+	0xbf9b0000, 0xbf9f0000,
 	0xbf9f0000, 0xbf9f0000,
 	0xbf9f0000, 0xbf9f0000,
-	0xbf9f0000, 0x00000000,
 };
 
 static const uint32_t cwsr_trap_arcturus_hex[] = {
-	0xbf820001, 0xbf8202d0,
+	0xbf820001, 0xbf8202d5,
 	0xb8f8f802, 0x8978ff78,
 	0x00020006, 0xb8fbf803,
 	0x866eff78, 0x00002000,
 	0xbf840009, 0x866eff6d,
 	0x00ff0000, 0xbf85001e,
 	0x866eff7b, 0x00000400,
-	0xbf850051, 0xbf8e0010,
+	0xbf850055, 0xbf8e0010,
 	0xb8fbf803, 0xbf82fffa,
-	0x866eff7b, 0x00000900,
+	0x866eff7b, 0x03c00900,
 	0xbf850015, 0x866eff7b,
 	0x000071ff, 0xbf840008,
 	0x866fff7b, 0x00007080,
@@ -1114,13 +1323,15 @@ static const uint32_t cwsr_trap_arcturus_hex[] = {
 	0xbf850007, 0xb8eef801,
 	0x866eff6e, 0x00000800,
 	0xbf850003, 0x866eff7b,
-	0x00000400, 0xbf850036,
+	0x00000400, 0xbf85003a,
 	0xb8faf807, 0x867aff7a,
 	0x001f8000, 0x8e7a8b7a,
 	0x8977ff77, 0xfc000000,
 	0x87777a77, 0xba7ff807,
 	0x00000000, 0xb8faf812,
 	0xb8fbf813, 0x8efa887a,
+	0xbf0d8f7b, 0xbf840002,
+	0x877bff7b, 0xffff0000,
 	0xc0031bbd, 0x00000010,
 	0xbf8cc07f, 0x8e6e976e,
 	0x8977ff77, 0x00800000,
@@ -1209,99 +1420,37 @@ static const uint32_t cwsr_trap_arcturus_hex[] = {
 	0xbefe007c, 0xbefc0070,
 	0xc0611c7a, 0x0000007c,
 	0xbf8cc07f, 0x80708470,
-	0xbefc007e, 0x867aff7f,
-	0x04000000, 0xbeef0080,
-	0x876f6f7a, 0xb8f02a05,
-	0x80708170, 0x8e708a70,
-	0x8e708170, 0xb8fb1605,
-	0x807b817b, 0x8e7b847b,
-	0x8e76827b, 0xbef600ff,
-	0x01000000, 0xbef20174,
-	0x80747074, 0x82758075,
-	0xbefc0080, 0xbf800000,
-	0xbe802b00, 0xbe822b02,
-	0xbe842b04, 0xbe862b06,
-	0xbe882b08, 0xbe8a2b0a,
-	0xbe8c2b0c, 0xbe8e2b0e,
-	0xc06b003a, 0x00000000,
-	0xbf8cc07f, 0xc06b013a,
-	0x00000010, 0xbf8cc07f,
-	0xc06b023a, 0x00000020,
-	0xbf8cc07f, 0xc06b033a,
-	0x00000030, 0xbf8cc07f,
-	0x8074c074, 0x82758075,
-	0x807c907c, 0xbf0a7b7c,
-	0xbf85ffe7, 0xbef40172,
-	0xbef00080, 0xbefe00c1,
-	0xbeff00c1, 0xbee80080,
-	0xbee90080, 0xbef600ff,
-	0x01000000, 0x867aff78,
-	0x00400000, 0xbf850003,
-	0xb8faf803, 0x897a7aff,
-	0x10000000, 0xbf85004d,
-	0xbe840080, 0xd2890000,
-	0x00000900, 0x80048104,
-	0xd2890001, 0x00000900,
-	0x80048104, 0xd2890002,
-	0x00000900, 0x80048104,
-	0xd2890003, 0x00000900,
-	0x80048104, 0xc069003a,
-	0x00000070, 0xbf8cc07f,
-	0x80709070, 0xbf06c004,
-	0xbf84ffee, 0xbe840080,
-	0xd2890000, 0x00000901,
-	0x80048104, 0xd2890001,
-	0x00000901, 0x80048104,
-	0xd2890002, 0x00000901,
-	0x80048104, 0xd2890003,
-	0x00000901, 0x80048104,
-	0xc069003a, 0x00000070,
-	0xbf8cc07f, 0x80709070,
-	0xbf06c004, 0xbf84ffee,
-	0xbe840080, 0xd2890000,
-	0x00000902, 0x80048104,
-	0xd2890001, 0x00000902,
-	0x80048104, 0xd2890002,
-	0x00000902, 0x80048104,
-	0xd2890003, 0x00000902,
-	0x80048104, 0xc069003a,
-	0x00000070, 0xbf8cc07f,
-	0x80709070, 0xbf06c004,
-	0xbf84ffee, 0xbe840080,
-	0xd2890000, 0x00000903,
-	0x80048104, 0xd2890001,
-	0x00000903, 0x80048104,
-	0xd2890002, 0x00000903,
-	0x80048104, 0xd2890003,
-	0x00000903, 0x80048104,
-	0xc069003a, 0x00000070,
-	0xbf8cc07f, 0x80709070,
-	0xbf06c004, 0xbf84ffee,
-	0xbf820008, 0xe0724000,
-	0x701d0000, 0xe0724100,
-	0x701d0100, 0xe0724200,
-	0x701d0200, 0xe0724300,
-	0x701d0300, 0xbefe00c1,
-	0xbeff00c1, 0xb8fb4306,
-	0x867bc17b, 0xbf840064,
-	0xbf8a0000, 0x867aff6f,
-	0x04000000, 0xbf840060,
-	0x8e7b867b, 0x8e7b827b,
-	0xbef6007b, 0xb8f02a05,
-	0x80708170, 0x8e708a70,
-	0x8e708170, 0xb8fa1605,
-	0x807a817a, 0x8e7a867a,
-	0x80707a70, 0x8070ff70,
-	0x00000080, 0xbef600ff,
-	0x01000000, 0xbefc0080,
-	0xd28c0002, 0x000100c1,
-	0xd28d0003, 0x000204c1,
+	0xbefc007e, 0xbf108080,
+	0x867aff7f, 0x04000000,
+	0xbeef0080, 0x876f6f7a,
+	0xb8f02a05, 0x80708170,
+	0x8e708a70, 0x8e708170,
+	0xb8fb1605, 0x807b817b,
+	0x8e7b847b, 0x8e76827b,
+	0xbef600ff, 0x01000000,
+	0xbef20174, 0x80747074,
+	0x82758075, 0xbefc0080,
+	0xbf800000, 0xbe802b00,
+	0xbe822b02, 0xbe842b04,
+	0xbe862b06, 0xbe882b08,
+	0xbe8a2b0a, 0xbe8c2b0c,
+	0xbe8e2b0e, 0xc06b003a,
+	0x00000000, 0xbf8cc07f,
+	0xc06b013a, 0x00000010,
+	0xbf8cc07f, 0xc06b023a,
+	0x00000020, 0xbf8cc07f,
+	0xc06b033a, 0x00000030,
+	0xbf8cc07f, 0x8074c074,
+	0x82758075, 0x807c907c,
+	0xbf0a7b7c, 0xbf85ffe7,
+	0xbef40172, 0xbef00080,
+	0xbefe00c1, 0xbeff00c1,
+	0xbee80080, 0xbee90080,
+	0xbef600ff, 0x01000000,
 	0x867aff78, 0x00400000,
 	0xbf850003, 0xb8faf803,
 	0x897a7aff, 0x10000000,
-	0xbf850030, 0x24040682,
-	0xd86e4000, 0x00000002,
-	0xbf8cc07f, 0xbe840080,
+	0xbf85004d, 0xbe840080,
 	0xd2890000, 0x00000900,
 	0x80048104, 0xd2890001,
 	0x00000900, 0x80048104,
@@ -1320,31 +1469,50 @@ static const uint32_t cwsr_trap_arcturus_hex[] = {
 	0x80048104, 0xc069003a,
 	0x00000070, 0xbf8cc07f,
 	0x80709070, 0xbf06c004,
-	0xbf84ffee, 0x680404ff,
-	0x00000200, 0xd0c9006a,
-	0x0000f702, 0xbf87ffd2,
-	0xbf820015, 0xd1060002,
-	0x00011103, 0x7e0602ff,
-	0x00000200, 0xbefc00ff,
-	0x00010000, 0xbe800077,
-	0x8677ff77, 0xff7fffff,
-	0x8777ff77, 0x00058000,
-	0xd8ec0000, 0x00000002,
-	0xbf8cc07f, 0xe0765000,
-	0x701d0002, 0x68040702,
-	0xd0c9006a, 0x0000f702,
-	0xbf87fff7, 0xbef70000,
-	0xbef000ff, 0x00000400,
+	0xbf84ffee, 0xbe840080,
+	0xd2890000, 0x00000902,
+	0x80048104, 0xd2890001,
+	0x00000902, 0x80048104,
+	0xd2890002, 0x00000902,
+	0x80048104, 0xd2890003,
+	0x00000902, 0x80048104,
+	0xc069003a, 0x00000070,
+	0xbf8cc07f, 0x80709070,
+	0xbf06c004, 0xbf84ffee,
+	0xbe840080, 0xd2890000,
+	0x00000903, 0x80048104,
+	0xd2890001, 0x00000903,
+	0x80048104, 0xd2890002,
+	0x00000903, 0x80048104,
+	0xd2890003, 0x00000903,
+	0x80048104, 0xc069003a,
+	0x00000070, 0xbf8cc07f,
+	0x80709070, 0xbf06c004,
+	0xbf84ffee, 0xbf820008,
+	0xe0724000, 0x701d0000,
+	0xe0724100, 0x701d0100,
+	0xe0724200, 0x701d0200,
+	0xe0724300, 0x701d0300,
 	0xbefe00c1, 0xbeff00c1,
-	0xb8fb2a05, 0x807b817b,
-	0x8e7b827b, 0xbef600ff,
-	0x01000000, 0xbefc0084,
-	0xbf0a7b7c, 0xbf84006d,
-	0xbf11017c, 0x807bff7b,
-	0x00001000, 0x867aff78,
+	0xb8fb4306, 0x867bc17b,
+	0xbf840064, 0xbf8a0000,
+	0x867aff6f, 0x04000000,
+	0xbf840060, 0x8e7b867b,
+	0x8e7b827b, 0xbef6007b,
+	0xb8f02a05, 0x80708170,
+	0x8e708a70, 0x8e708170,
+	0xb8fa1605, 0x807a817a,
+	0x8e7a867a, 0x80707a70,
+	0x8070ff70, 0x00000080,
+	0xbef600ff, 0x01000000,
+	0xbefc0080, 0xd28c0002,
+	0x000100c1, 0xd28d0003,
+	0x000204c1, 0x867aff78,
 	0x00400000, 0xbf850003,
 	0xb8faf803, 0x897a7aff,
-	0x10000000, 0xbf850051,
+	0x10000000, 0xbf850030,
+	0x24040682, 0xd86e4000,
+	0x00000002, 0xbf8cc07f,
 	0xbe840080, 0xd2890000,
 	0x00000900, 0x80048104,
 	0xd2890001, 0x00000900,
@@ -1364,224 +1532,268 @@ static const uint32_t cwsr_trap_arcturus_hex[] = {
 	0xc069003a, 0x00000070,
 	0xbf8cc07f, 0x80709070,
 	0xbf06c004, 0xbf84ffee,
+	0x680404ff, 0x00000200,
+	0xd0c9006a, 0x0000f702,
+	0xbf87ffd2, 0xbf820015,
+	0xd1060002, 0x00011103,
+	0x7e0602ff, 0x00000200,
+	0xbefc00ff, 0x00010000,
+	0xbe800077, 0x8677ff77,
+	0xff7fffff, 0x8777ff77,
+	0x00058000, 0xd8ec0000,
+	0x00000002, 0xbf8cc07f,
+	0xe0765000, 0x701d0002,
+	0x68040702, 0xd0c9006a,
+	0x0000f702, 0xbf87fff7,
+	0xbef70000, 0xbef000ff,
+	0x00000400, 0xbefe00c1,
+	0xbeff00c1, 0xb8fb2a05,
+	0x807b817b, 0x8e7b827b,
+	0xbef600ff, 0x01000000,
+	0xbefc0084, 0xbf0a7b7c,
+	0xbf84006d, 0xbf11017c,
+	0x807bff7b, 0x00001000,
+	0x867aff78, 0x00400000,
+	0xbf850003, 0xb8faf803,
+	0x897a7aff, 0x10000000,
+	0xbf850051, 0xbe840080,
+	0xd2890000, 0x00000900,
+	0x80048104, 0xd2890001,
+	0x00000900, 0x80048104,
+	0xd2890002, 0x00000900,
+	0x80048104, 0xd2890003,
+	0x00000900, 0x80048104,
+	0xc069003a, 0x00000070,
+	0xbf8cc07f, 0x80709070,
+	0xbf06c004, 0xbf84ffee,
 	0xbe840080, 0xd2890000,
-	0x00000902, 0x80048104,
-	0xd2890001, 0x00000902,
+	0x00000901, 0x80048104,
+	0xd2890001, 0x00000901,
 	0x80048104, 0xd2890002,
-	0x00000902, 0x80048104,
-	0xd2890003, 0x00000902,
+	0x00000901, 0x80048104,
+	0xd2890003, 0x00000901,
 	0x80048104, 0xc069003a,
 	0x00000070, 0xbf8cc07f,
 	0x80709070, 0xbf06c004,
 	0xbf84ffee, 0xbe840080,
-	0xd2890000, 0x00000903,
+	0xd2890000, 0x00000902,
 	0x80048104, 0xd2890001,
-	0x00000903, 0x80048104,
-	0xd2890002, 0x00000903,
+	0x00000902, 0x80048104,
+	0xd2890002, 0x00000902,
 	0x80048104, 0xd2890003,
-	0x00000903, 0x80048104,
+	0x00000902, 0x80048104,
 	0xc069003a, 0x00000070,
 	0xbf8cc07f, 0x80709070,
 	0xbf06c004, 0xbf84ffee,
-	0x807c847c, 0xbf0a7b7c,
-	0xbf85ffb1, 0xbf9c0000,
-	0xbf820012, 0x7e000300,
-	0x7e020301, 0x7e040302,
-	0x7e060303, 0xe0724000,
-	0x701d0000, 0xe0724100,
-	0x701d0100, 0xe0724200,
-	0x701d0200, 0xe0724300,
-	0x701d0300, 0x807c847c,
-	0x8070ff70, 0x00000400,
-	0xbf0a7b7c, 0xbf85ffef,
-	0xbf9c0000, 0xbefc0080,
-	0xbf11017c, 0x867aff78,
-	0x00400000, 0xbf850003,
-	0xb8faf803, 0x897a7aff,
-	0x10000000, 0xbf850059,
-	0xd3d84000, 0x18000100,
-	0xd3d84001, 0x18000101,
-	0xd3d84002, 0x18000102,
-	0xd3d84003, 0x18000103,
 	0xbe840080, 0xd2890000,
-	0x00000900, 0x80048104,
-	0xd2890001, 0x00000900,
+	0x00000903, 0x80048104,
+	0xd2890001, 0x00000903,
 	0x80048104, 0xd2890002,
-	0x00000900, 0x80048104,
-	0xd2890003, 0x00000900,
+	0x00000903, 0x80048104,
+	0xd2890003, 0x00000903,
 	0x80048104, 0xc069003a,
 	0x00000070, 0xbf8cc07f,
 	0x80709070, 0xbf06c004,
-	0xbf84ffee, 0xbe840080,
-	0xd2890000, 0x00000901,
+	0xbf84ffee, 0x807c847c,
+	0xbf0a7b7c, 0xbf85ffb1,
+	0xbf9c0000, 0xbf820012,
+	0x7e000300, 0x7e020301,
+	0x7e040302, 0x7e060303,
+	0xe0724000, 0x701d0000,
+	0xe0724100, 0x701d0100,
+	0xe0724200, 0x701d0200,
+	0xe0724300, 0x701d0300,
+	0x807c847c, 0x8070ff70,
+	0x00000400, 0xbf0a7b7c,
+	0xbf85ffef, 0xbf9c0000,
+	0xbefc0080, 0xbf11017c,
+	0x867aff78, 0x00400000,
+	0xbf850003, 0xb8faf803,
+	0x897a7aff, 0x10000000,
+	0xbf850059, 0xd3d84000,
+	0x18000100, 0xd3d84001,
+	0x18000101, 0xd3d84002,
+	0x18000102, 0xd3d84003,
+	0x18000103, 0xbe840080,
+	0xd2890000, 0x00000900,
 	0x80048104, 0xd2890001,
-	0x00000901, 0x80048104,
-	0xd2890002, 0x00000901,
+	0x00000900, 0x80048104,
+	0xd2890002, 0x00000900,
 	0x80048104, 0xd2890003,
-	0x00000901, 0x80048104,
+	0x00000900, 0x80048104,
 	0xc069003a, 0x00000070,
 	0xbf8cc07f, 0x80709070,
 	0xbf06c004, 0xbf84ffee,
 	0xbe840080, 0xd2890000,
-	0x00000902, 0x80048104,
-	0xd2890001, 0x00000902,
+	0x00000901, 0x80048104,
+	0xd2890001, 0x00000901,
 	0x80048104, 0xd2890002,
-	0x00000902, 0x80048104,
-	0xd2890003, 0x00000902,
+	0x00000901, 0x80048104,
+	0xd2890003, 0x00000901,
 	0x80048104, 0xc069003a,
 	0x00000070, 0xbf8cc07f,
 	0x80709070, 0xbf06c004,
 	0xbf84ffee, 0xbe840080,
-	0xd2890000, 0x00000903,
+	0xd2890000, 0x00000902,
 	0x80048104, 0xd2890001,
-	0x00000903, 0x80048104,
-	0xd2890002, 0x00000903,
+	0x00000902, 0x80048104,
+	0xd2890002, 0x00000902,
 	0x80048104, 0xd2890003,
-	0x00000903, 0x80048104,
+	0x00000902, 0x80048104,
 	0xc069003a, 0x00000070,
 	0xbf8cc07f, 0x80709070,
 	0xbf06c004, 0xbf84ffee,
-	0x807c847c, 0xbf0a7b7c,
-	0xbf85ffa9, 0xbf9c0000,
-	0xbf820016, 0xd3d84000,
-	0x18000100, 0xd3d84001,
-	0x18000101, 0xd3d84002,
-	0x18000102, 0xd3d84003,
-	0x18000103, 0xe0724000,
-	0x701d0000, 0xe0724100,
-	0x701d0100, 0xe0724200,
-	0x701d0200, 0xe0724300,
-	0x701d0300, 0x807c847c,
-	0x8070ff70, 0x00000400,
-	0xbf0a7b7c, 0xbf85ffeb,
-	0xbf9c0000, 0xbf8200e3,
-	0xbef4007e, 0x8675ff7f,
-	0x0000ffff, 0x8775ff75,
-	0x00040000, 0xbef60080,
-	0xbef700ff, 0x00807fac,
-	0x866eff7f, 0x04000000,
-	0xbf84001f, 0xbefe00c1,
-	0xbeff00c1, 0xb8ef4306,
-	0x866fc16f, 0xbf84001a,
-	0x8e6f866f, 0x8e6f826f,
-	0xbef6006f, 0xb8f82a05,
-	0x80788178, 0x8e788a78,
-	0x8e788178, 0xb8ee1605,
-	0x806e816e, 0x8e6e866e,
-	0x80786e78, 0x8078ff78,
-	0x00000080, 0xbef600ff,
-	0x01000000, 0xbefc0080,
-	0xe0510000, 0x781d0000,
-	0xe0510100, 0x781d0000,
-	0x807cff7c, 0x00000200,
-	0x8078ff78, 0x00000200,
-	0xbf0a6f7c, 0xbf85fff6,
+	0xbe840080, 0xd2890000,
+	0x00000903, 0x80048104,
+	0xd2890001, 0x00000903,
+	0x80048104, 0xd2890002,
+	0x00000903, 0x80048104,
+	0xd2890003, 0x00000903,
+	0x80048104, 0xc069003a,
+	0x00000070, 0xbf8cc07f,
+	0x80709070, 0xbf06c004,
+	0xbf84ffee, 0x807c847c,
+	0xbf0a7b7c, 0xbf85ffa9,
+	0xbf9c0000, 0xbf820016,
+	0xd3d84000, 0x18000100,
+	0xd3d84001, 0x18000101,
+	0xd3d84002, 0x18000102,
+	0xd3d84003, 0x18000103,
+	0xe0724000, 0x701d0000,
+	0xe0724100, 0x701d0100,
+	0xe0724200, 0x701d0200,
+	0xe0724300, 0x701d0300,
+	0x807c847c, 0x8070ff70,
+	0x00000400, 0xbf0a7b7c,
+	0xbf85ffeb, 0xbf9c0000,
+	0xbf8200e3, 0xbef4007e,
+	0x8675ff7f, 0x0000ffff,
+	0x8775ff75, 0x00040000,
+	0xbef60080, 0xbef700ff,
+	0x00807fac, 0x866eff7f,
+	0x04000000, 0xbf84001f,
 	0xbefe00c1, 0xbeff00c1,
+	0xb8ef4306, 0x866fc16f,
+	0xbf84001a, 0x8e6f866f,
+	0x8e6f826f, 0xbef6006f,
+	0xb8f82a05, 0x80788178,
+	0x8e788a78, 0x8e788178,
+	0xb8ee1605, 0x806e816e,
+	0x8e6e866e, 0x80786e78,
+	0x8078ff78, 0x00000080,
 	0xbef600ff, 0x01000000,
-	0xb8ef2a05, 0x806f816f,
-	0x8e6f826f, 0x806fff6f,
-	0x00008000, 0xbef80080,
-	0xbeee0078, 0x8078ff78,
-	0x00000400, 0xbefc0084,
-	0xbf11087c, 0xe0524000,
-	0x781d0000, 0xe0524100,
-	0x781d0100, 0xe0524200,
-	0x781d0200, 0xe0524300,
-	0x781d0300, 0xbf8c0f70,
-	0x7e000300, 0x7e020301,
-	0x7e040302, 0x7e060303,
-	0x807c847c, 0x8078ff78,
-	0x00000400, 0xbf0a6f7c,
-	0xbf85ffee, 0xbefc0080,
-	0xbf11087c, 0xe0524000,
-	0x781d0000, 0xe0524100,
-	0x781d0100, 0xe0524200,
-	0x781d0200, 0xe0524300,
-	0x781d0300, 0xbf8c0f70,
-	0xd3d94000, 0x18000100,
-	0xd3d94001, 0x18000101,
-	0xd3d94002, 0x18000102,
-	0xd3d94003, 0x18000103,
-	0x807c847c, 0x8078ff78,
-	0x00000400, 0xbf0a6f7c,
-	0xbf85ffea, 0xbf9c0000,
-	0xe0524000, 0x6e1d0000,
-	0xe0524100, 0x6e1d0100,
-	0xe0524200, 0x6e1d0200,
-	0xe0524300, 0x6e1d0300,
-	0xbf8c0f70, 0xb8f82a05,
-	0x80788178, 0x8e788a78,
-	0x8e788178, 0xb8ee1605,
-	0x806e816e, 0x8e6e866e,
-	0x80786e78, 0x80f8c078,
-	0xb8ef1605, 0x806f816f,
-	0x8e6f846f, 0x8e76826f,
-	0xbef600ff, 0x01000000,
-	0xbefc006f, 0xc031003a,
-	0x00000078, 0x80f8c078,
-	0xbf8cc07f, 0x80fc907c,
-	0xbf800000, 0xbe802d00,
-	0xbe822d02, 0xbe842d04,
-	0xbe862d06, 0xbe882d08,
-	0xbe8a2d0a, 0xbe8c2d0c,
-	0xbe8e2d0e, 0xbf06807c,
-	0xbf84fff0, 0xb8f82a05,
-	0x80788178, 0x8e788a78,
-	0x8e788178, 0xb8ee1605,
-	0x806e816e, 0x8e6e866e,
-	0x80786e78, 0xbef60084,
-	0xbef600ff, 0x01000000,
-	0xc0211bfa, 0x00000078,
-	0x80788478, 0xc0211b3a,
+	0xbefc0080, 0xe0510000,
+	0x781d0000, 0xe0510100,
+	0x781d0000, 0x807cff7c,
+	0x00000200, 0x8078ff78,
+	0x00000200, 0xbf0a6f7c,
+	0xbf85fff6, 0xbefe00c1,
+	0xbeff00c1, 0xbef600ff,
+	0x01000000, 0xb8ef2a05,
+	0x806f816f, 0x8e6f826f,
+	0x806fff6f, 0x00008000,
+	0xbef80080, 0xbeee0078,
+	0x8078ff78, 0x00000400,
+	0xbefc0084, 0xbf11087c,
+	0xe0524000, 0x781d0000,
+	0xe0524100, 0x781d0100,
+	0xe0524200, 0x781d0200,
+	0xe0524300, 0x781d0300,
+	0xbf8c0f70, 0x7e000300,
+	0x7e020301, 0x7e040302,
+	0x7e060303, 0x807c847c,
+	0x8078ff78, 0x00000400,
+	0xbf0a6f7c, 0xbf85ffee,
+	0xbefc0080, 0xbf11087c,
+	0xe0524000, 0x781d0000,
+	0xe0524100, 0x781d0100,
+	0xe0524200, 0x781d0200,
+	0xe0524300, 0x781d0300,
+	0xbf8c0f70, 0xd3d94000,
+	0x18000100, 0xd3d94001,
+	0x18000101, 0xd3d94002,
+	0x18000102, 0xd3d94003,
+	0x18000103, 0x807c847c,
+	0x8078ff78, 0x00000400,
+	0xbf0a6f7c, 0xbf85ffea,
+	0xbf9c0000, 0xe0524000,
+	0x6e1d0000, 0xe0524100,
+	0x6e1d0100, 0xe0524200,
+	0x6e1d0200, 0xe0524300,
+	0x6e1d0300, 0xbf8c0f70,
+	0xb8f82a05, 0x80788178,
+	0x8e788a78, 0x8e788178,
+	0xb8ee1605, 0x806e816e,
+	0x8e6e866e, 0x80786e78,
+	0x80f8c078, 0xb8ef1605,
+	0x806f816f, 0x8e6f846f,
+	0x8e76826f, 0xbef600ff,
+	0x01000000, 0xbefc006f,
+	0xc031003a, 0x00000078,
+	0x80f8c078, 0xbf8cc07f,
+	0x80fc907c, 0xbf800000,
+	0xbe802d00, 0xbe822d02,
+	0xbe842d04, 0xbe862d06,
+	0xbe882d08, 0xbe8a2d0a,
+	0xbe8c2d0c, 0xbe8e2d0e,
+	0xbf06807c, 0xbf84fff0,
+	0xb8f82a05, 0x80788178,
+	0x8e788a78, 0x8e788178,
+	0xb8ee1605, 0x806e816e,
+	0x8e6e866e, 0x80786e78,
+	0xbef60084, 0xbef600ff,
+	0x01000000, 0xc0211bfa,
 	0x00000078, 0x80788478,
-	0xc0211b7a, 0x00000078,
-	0x80788478, 0xc0211c3a,
+	0xc0211b3a, 0x00000078,
+	0x80788478, 0xc0211b7a,
 	0x00000078, 0x80788478,
-	0xc0211c7a, 0x00000078,
-	0x80788478, 0xc0211eba,
+	0xc0211c3a, 0x00000078,
+	0x80788478, 0xc0211c7a,
 	0x00000078, 0x80788478,
-	0xc0211efa, 0x00000078,
-	0x80788478, 0xc0211a3a,
+	0xc0211eba, 0x00000078,
+	0x80788478, 0xc0211efa,
 	0x00000078, 0x80788478,
-	0xc0211a7a, 0x00000078,
-	0x80788478, 0xc0211cfa,
+	0xc0211a3a, 0x00000078,
+	0x80788478, 0xc0211a7a,
 	0x00000078, 0x80788478,
-	0xbf8cc07f, 0xbefc006f,
-	0xbefe0070, 0xbeff0071,
-	0x866f7bff, 0x000003ff,
-	0xb96f4803, 0x866f7bff,
-	0xfffff800, 0x8f6f8b6f,
-	0xb96fa2c3, 0xb973f801,
-	0xb8ee2a05, 0x806e816e,
-	0x8e6e8a6e, 0x8e6e816e,
-	0xb8ef1605, 0x806f816f,
-	0x8e6f866f, 0x806e6f6e,
-	0x806e746e, 0x826f8075,
-	0x866fff6f, 0x0000ffff,
-	0xc00b1c37, 0x00000050,
-	0xc00b1d37, 0x00000060,
-	0xc0031e77, 0x00000074,
-	0xbf8cc07f, 0x8f6e8b77,
-	0x866eff6e, 0x001f8000,
-	0xb96ef807, 0x866dff6d,
-	0x0000ffff, 0x86fe7e7e,
-	0x86ea6a6a, 0x8f6e837a,
-	0xb96ee0c2, 0xbf800002,
-	0xb97a0002, 0xbf8a0000,
-	0xbe801f6c, 0xbf810000,
+	0xc0211cfa, 0x00000078,
+	0x80788478, 0xbf8cc07f,
+	0xbefc006f, 0xbefe0070,
+	0xbeff0071, 0x866f7bff,
+	0x000003ff, 0xb96f4803,
+	0x866f7bff, 0xfffff800,
+	0x8f6f8b6f, 0xb96fa2c3,
+	0xb973f801, 0xb8ee2a05,
+	0x806e816e, 0x8e6e8a6e,
+	0x8e6e816e, 0xb8ef1605,
+	0x806f816f, 0x8e6f866f,
+	0x806e6f6e, 0x806e746e,
+	0x826f8075, 0x866fff6f,
+	0x0000ffff, 0xc00b1c37,
+	0x00000050, 0xc00b1d37,
+	0x00000060, 0xc0031e77,
+	0x00000074, 0xbf8cc07f,
+	0x8f6e8b77, 0x866eff6e,
+	0x001f8000, 0xb96ef807,
+	0x866dff6d, 0x0000ffff,
+	0x86fe7e7e, 0x86ea6a6a,
+	0x8f6e837a, 0xb96ee0c2,
+	0xbf800002, 0xb97a0002,
+	0xbf8a0000, 0xbe801f6c,
+	0xbf9b0000, 0x00000000,
 };
 
 static const uint32_t cwsr_trap_aldebaran_hex[] = {
-	0xbf820001, 0xbf8202db,
+	0xbf820001, 0xbf8202e0,
 	0xb8f8f802, 0x8978ff78,
 	0x00020006, 0xb8fbf803,
 	0x866eff78, 0x00002000,
 	0xbf840009, 0x866eff6d,
 	0x00ff0000, 0xbf85001e,
 	0x866eff7b, 0x00000400,
-	0xbf850051, 0xbf8e0010,
+	0xbf850055, 0xbf8e0010,
 	0xb8fbf803, 0xbf82fffa,
-	0x866eff7b, 0x00000900,
+	0x866eff7b, 0x03c00900,
 	0xbf850015, 0x866eff7b,
 	0x000071ff, 0xbf840008,
 	0x866fff7b, 0x00007080,
@@ -1592,13 +1804,15 @@ static const uint32_t cwsr_trap_aldebaran_hex[] = {
 	0xbf850007, 0xb8eef801,
 	0x866eff6e, 0x00000800,
 	0xbf850003, 0x866eff7b,
-	0x00000400, 0xbf850036,
+	0x00000400, 0xbf85003a,
 	0xb8faf807, 0x867aff7a,
 	0x001f8000, 0x8e7a8b7a,
 	0x8977ff77, 0xfc000000,
 	0x87777a77, 0xba7ff807,
 	0x00000000, 0xb8faf812,
 	0xb8fbf813, 0x8efa887a,
+	0xbf0d8f7b, 0xbf840002,
+	0x877bff7b, 0xffff0000,
 	0xc0031bbd, 0x00000010,
 	0xbf8cc07f, 0x8e6e976e,
 	0x8977ff77, 0x00800000,
@@ -1687,99 +1901,37 @@ static const uint32_t cwsr_trap_aldebaran_hex[] = {
 	0xbefe007c, 0xbefc0070,
 	0xc0611c7a, 0x0000007c,
 	0xbf8cc07f, 0x80708470,
-	0xbefc007e, 0x867aff7f,
-	0x04000000, 0xbeef0080,
-	0x876f6f7a, 0xb8f02985,
-	0x80708170, 0x8e708a70,
-	0x8e708170, 0xb8fb1605,
-	0x807b817b, 0x8e7b847b,
-	0x8e76827b, 0xbef600ff,
-	0x01000000, 0xbef20174,
-	0x80747074, 0x82758075,
-	0xbefc0080, 0xbf800000,
-	0xbe802b00, 0xbe822b02,
-	0xbe842b04, 0xbe862b06,
-	0xbe882b08, 0xbe8a2b0a,
-	0xbe8c2b0c, 0xbe8e2b0e,
-	0xc06b003a, 0x00000000,
-	0xbf8cc07f, 0xc06b013a,
-	0x00000010, 0xbf8cc07f,
-	0xc06b023a, 0x00000020,
-	0xbf8cc07f, 0xc06b033a,
-	0x00000030, 0xbf8cc07f,
-	0x8074c074, 0x82758075,
-	0x807c907c, 0xbf0a7b7c,
-	0xbf85ffe7, 0xbef40172,
-	0xbef00080, 0xbefe00c1,
-	0xbeff00c1, 0xbee80080,
-	0xbee90080, 0xbef600ff,
-	0x01000000, 0x867aff78,
-	0x00400000, 0xbf850003,
-	0xb8faf803, 0x897a7aff,
-	0x10000000, 0xbf85004d,
-	0xbe840080, 0xd2890000,
-	0x00000900, 0x80048104,
-	0xd2890001, 0x00000900,
-	0x80048104, 0xd2890002,
-	0x00000900, 0x80048104,
-	0xd2890003, 0x00000900,
-	0x80048104, 0xc069003a,
-	0x00000070, 0xbf8cc07f,
-	0x80709070, 0xbf06c004,
-	0xbf84ffee, 0xbe840080,
-	0xd2890000, 0x00000901,
-	0x80048104, 0xd2890001,
-	0x00000901, 0x80048104,
-	0xd2890002, 0x00000901,
-	0x80048104, 0xd2890003,
-	0x00000901, 0x80048104,
-	0xc069003a, 0x00000070,
-	0xbf8cc07f, 0x80709070,
-	0xbf06c004, 0xbf84ffee,
-	0xbe840080, 0xd2890000,
-	0x00000902, 0x80048104,
-	0xd2890001, 0x00000902,
-	0x80048104, 0xd2890002,
-	0x00000902, 0x80048104,
-	0xd2890003, 0x00000902,
-	0x80048104, 0xc069003a,
-	0x00000070, 0xbf8cc07f,
-	0x80709070, 0xbf06c004,
-	0xbf84ffee, 0xbe840080,
-	0xd2890000, 0x00000903,
-	0x80048104, 0xd2890001,
-	0x00000903, 0x80048104,
-	0xd2890002, 0x00000903,
-	0x80048104, 0xd2890003,
-	0x00000903, 0x80048104,
-	0xc069003a, 0x00000070,
-	0xbf8cc07f, 0x80709070,
-	0xbf06c004, 0xbf84ffee,
-	0xbf820008, 0xe0724000,
-	0x701d0000, 0xe0724100,
-	0x701d0100, 0xe0724200,
-	0x701d0200, 0xe0724300,
-	0x701d0300, 0xbefe00c1,
-	0xbeff00c1, 0xb8fb4306,
-	0x867bc17b, 0xbf840064,
-	0xbf8a0000, 0x867aff6f,
-	0x04000000, 0xbf840060,
-	0x8e7b867b, 0x8e7b827b,
-	0xbef6007b, 0xb8f02985,
-	0x80708170, 0x8e708a70,
-	0x8e708170, 0xb8fa1605,
-	0x807a817a, 0x8e7a867a,
-	0x80707a70, 0x8070ff70,
-	0x00000080, 0xbef600ff,
-	0x01000000, 0xbefc0080,
-	0xd28c0002, 0x000100c1,
-	0xd28d0003, 0x000204c1,
+	0xbefc007e, 0xbf108080,
+	0x867aff7f, 0x04000000,
+	0xbeef0080, 0x876f6f7a,
+	0xb8f02985, 0x80708170,
+	0x8e708a70, 0x8e708170,
+	0xb8fb1605, 0x807b817b,
+	0x8e7b847b, 0x8e76827b,
+	0xbef600ff, 0x01000000,
+	0xbef20174, 0x80747074,
+	0x82758075, 0xbefc0080,
+	0xbf800000, 0xbe802b00,
+	0xbe822b02, 0xbe842b04,
+	0xbe862b06, 0xbe882b08,
+	0xbe8a2b0a, 0xbe8c2b0c,
+	0xbe8e2b0e, 0xc06b003a,
+	0x00000000, 0xbf8cc07f,
+	0xc06b013a, 0x00000010,
+	0xbf8cc07f, 0xc06b023a,
+	0x00000020, 0xbf8cc07f,
+	0xc06b033a, 0x00000030,
+	0xbf8cc07f, 0x8074c074,
+	0x82758075, 0x807c907c,
+	0xbf0a7b7c, 0xbf85ffe7,
+	0xbef40172, 0xbef00080,
+	0xbefe00c1, 0xbeff00c1,
+	0xbee80080, 0xbee90080,
+	0xbef600ff, 0x01000000,
 	0x867aff78, 0x00400000,
 	0xbf850003, 0xb8faf803,
 	0x897a7aff, 0x10000000,
-	0xbf850030, 0x24040682,
-	0xd86e4000, 0x00000002,
-	0xbf8cc07f, 0xbe840080,
+	0xbf85004d, 0xbe840080,
 	0xd2890000, 0x00000900,
 	0x80048104, 0xd2890001,
 	0x00000900, 0x80048104,
@@ -1798,31 +1950,50 @@ static const uint32_t cwsr_trap_aldebaran_hex[] = {
 	0x80048104, 0xc069003a,
 	0x00000070, 0xbf8cc07f,
 	0x80709070, 0xbf06c004,
-	0xbf84ffee, 0x680404ff,
-	0x00000200, 0xd0c9006a,
-	0x0000f702, 0xbf87ffd2,
-	0xbf820015, 0xd1060002,
-	0x00011103, 0x7e0602ff,
-	0x00000200, 0xbefc00ff,
-	0x00010000, 0xbe800077,
-	0x8677ff77, 0xff7fffff,
-	0x8777ff77, 0x00058000,
-	0xd8ec0000, 0x00000002,
-	0xbf8cc07f, 0xe0765000,
-	0x701d0002, 0x68040702,
-	0xd0c9006a, 0x0000f702,
-	0xbf87fff7, 0xbef70000,
-	0xbef000ff, 0x00000400,
+	0xbf84ffee, 0xbe840080,
+	0xd2890000, 0x00000902,
+	0x80048104, 0xd2890001,
+	0x00000902, 0x80048104,
+	0xd2890002, 0x00000902,
+	0x80048104, 0xd2890003,
+	0x00000902, 0x80048104,
+	0xc069003a, 0x00000070,
+	0xbf8cc07f, 0x80709070,
+	0xbf06c004, 0xbf84ffee,
+	0xbe840080, 0xd2890000,
+	0x00000903, 0x80048104,
+	0xd2890001, 0x00000903,
+	0x80048104, 0xd2890002,
+	0x00000903, 0x80048104,
+	0xd2890003, 0x00000903,
+	0x80048104, 0xc069003a,
+	0x00000070, 0xbf8cc07f,
+	0x80709070, 0xbf06c004,
+	0xbf84ffee, 0xbf820008,
+	0xe0724000, 0x701d0000,
+	0xe0724100, 0x701d0100,
+	0xe0724200, 0x701d0200,
+	0xe0724300, 0x701d0300,
 	0xbefe00c1, 0xbeff00c1,
-	0xb8fb2b05, 0x807b817b,
-	0x8e7b827b, 0xbef600ff,
-	0x01000000, 0xbefc0084,
-	0xbf0a7b7c, 0xbf84006d,
-	0xbf11017c, 0x807bff7b,
-	0x00001000, 0x867aff78,
+	0xb8fb4306, 0x867bc17b,
+	0xbf840064, 0xbf8a0000,
+	0x867aff6f, 0x04000000,
+	0xbf840060, 0x8e7b867b,
+	0x8e7b827b, 0xbef6007b,
+	0xb8f02985, 0x80708170,
+	0x8e708a70, 0x8e708170,
+	0xb8fa1605, 0x807a817a,
+	0x8e7a867a, 0x80707a70,
+	0x8070ff70, 0x00000080,
+	0xbef600ff, 0x01000000,
+	0xbefc0080, 0xd28c0002,
+	0x000100c1, 0xd28d0003,
+	0x000204c1, 0x867aff78,
 	0x00400000, 0xbf850003,
 	0xb8faf803, 0x897a7aff,
-	0x10000000, 0xbf850051,
+	0x10000000, 0xbf850030,
+	0x24040682, 0xd86e4000,
+	0x00000002, 0xbf8cc07f,
 	0xbe840080, 0xd2890000,
 	0x00000900, 0x80048104,
 	0xd2890001, 0x00000900,
@@ -1842,51 +2013,31 @@ static const uint32_t cwsr_trap_aldebaran_hex[] = {
 	0xc069003a, 0x00000070,
 	0xbf8cc07f, 0x80709070,
 	0xbf06c004, 0xbf84ffee,
-	0xbe840080, 0xd2890000,
-	0x00000902, 0x80048104,
-	0xd2890001, 0x00000902,
-	0x80048104, 0xd2890002,
-	0x00000902, 0x80048104,
-	0xd2890003, 0x00000902,
-	0x80048104, 0xc069003a,
-	0x00000070, 0xbf8cc07f,
-	0x80709070, 0xbf06c004,
-	0xbf84ffee, 0xbe840080,
-	0xd2890000, 0x00000903,
-	0x80048104, 0xd2890001,
-	0x00000903, 0x80048104,
-	0xd2890002, 0x00000903,
-	0x80048104, 0xd2890003,
-	0x00000903, 0x80048104,
-	0xc069003a, 0x00000070,
-	0xbf8cc07f, 0x80709070,
-	0xbf06c004, 0xbf84ffee,
-	0x807c847c, 0xbf0a7b7c,
-	0xbf85ffb1, 0xbf9c0000,
-	0xbf820012, 0x7e000300,
-	0x7e020301, 0x7e040302,
-	0x7e060303, 0xe0724000,
-	0x701d0000, 0xe0724100,
-	0x701d0100, 0xe0724200,
-	0x701d0200, 0xe0724300,
-	0x701d0300, 0x807c847c,
-	0x8070ff70, 0x00000400,
-	0xbf0a7b7c, 0xbf85ffef,
-	0xbf9c0000, 0xb8fb2985,
-	0x807b817b, 0x8e7b837b,
-	0xb8fa2b05, 0x807a817a,
-	0x8e7a827a, 0x80fb7a7b,
-	0x867b7b7b, 0xbf84007a,
+	0x680404ff, 0x00000200,
+	0xd0c9006a, 0x0000f702,
+	0xbf87ffd2, 0xbf820015,
+	0xd1060002, 0x00011103,
+	0x7e0602ff, 0x00000200,
+	0xbefc00ff, 0x00010000,
+	0xbe800077, 0x8677ff77,
+	0xff7fffff, 0x8777ff77,
+	0x00058000, 0xd8ec0000,
+	0x00000002, 0xbf8cc07f,
+	0xe0765000, 0x701d0002,
+	0x68040702, 0xd0c9006a,
+	0x0000f702, 0xbf87fff7,
+	0xbef70000, 0xbef000ff,
+	0x00000400, 0xbefe00c1,
+	0xbeff00c1, 0xb8fb2b05,
+	0x807b817b, 0x8e7b827b,
+	0xbef600ff, 0x01000000,
+	0xbefc0084, 0xbf0a7b7c,
+	0xbf84006d, 0xbf11017c,
 	0x807bff7b, 0x00001000,
-	0xbefc0080, 0xbf11017c,
 	0x867aff78, 0x00400000,
 	0xbf850003, 0xb8faf803,
 	0x897a7aff, 0x10000000,
-	0xbf850059, 0xd3d84000,
-	0x18000100, 0xd3d84001,
-	0x18000101, 0xd3d84002,
-	0x18000102, 0xd3d84003,
-	0x18000103, 0xbe840080,
+	0xbf850051, 0xbe840080,
 	0xd2890000, 0x00000900,
 	0x80048104, 0xd2890001,
 	0x00000900, 0x80048104,
@@ -1925,150 +2076,214 @@ static const uint32_t cwsr_trap_aldebaran_hex[] = {
 	0x00000070, 0xbf8cc07f,
 	0x80709070, 0xbf06c004,
 	0xbf84ffee, 0x807c847c,
-	0xbf0a7b7c, 0xbf85ffa9,
-	0xbf9c0000, 0xbf820016,
-	0xd3d84000, 0x18000100,
-	0xd3d84001, 0x18000101,
-	0xd3d84002, 0x18000102,
-	0xd3d84003, 0x18000103,
+	0xbf0a7b7c, 0xbf85ffb1,
+	0xbf9c0000, 0xbf820012,
+	0x7e000300, 0x7e020301,
+	0x7e040302, 0x7e060303,
 	0xe0724000, 0x701d0000,
 	0xe0724100, 0x701d0100,
 	0xe0724200, 0x701d0200,
 	0xe0724300, 0x701d0300,
 	0x807c847c, 0x8070ff70,
 	0x00000400, 0xbf0a7b7c,
-	0xbf85ffeb, 0xbf9c0000,
-	0xbf8200ee, 0xbef4007e,
-	0x8675ff7f, 0x0000ffff,
-	0x8775ff75, 0x00040000,
-	0xbef60080, 0xbef700ff,
-	0x00807fac, 0x866eff7f,
-	0x04000000, 0xbf84001f,
+	0xbf85ffef, 0xbf9c0000,
+	0xb8fb2985, 0x807b817b,
+	0x8e7b837b, 0xb8fa2b05,
+	0x807a817a, 0x8e7a827a,
+	0x80fb7a7b, 0x867b7b7b,
+	0xbf84007a, 0x807bff7b,
+	0x00001000, 0xbefc0080,
+	0xbf11017c, 0x867aff78,
+	0x00400000, 0xbf850003,
+	0xb8faf803, 0x897a7aff,
+	0x10000000, 0xbf850059,
+	0xd3d84000, 0x18000100,
+	0xd3d84001, 0x18000101,
+	0xd3d84002, 0x18000102,
+	0xd3d84003, 0x18000103,
+	0xbe840080, 0xd2890000,
+	0x00000900, 0x80048104,
+	0xd2890001, 0x00000900,
+	0x80048104, 0xd2890002,
+	0x00000900, 0x80048104,
+	0xd2890003, 0x00000900,
+	0x80048104, 0xc069003a,
+	0x00000070, 0xbf8cc07f,
+	0x80709070, 0xbf06c004,
+	0xbf84ffee, 0xbe840080,
+	0xd2890000, 0x00000901,
+	0x80048104, 0xd2890001,
+	0x00000901, 0x80048104,
+	0xd2890002, 0x00000901,
+	0x80048104, 0xd2890003,
+	0x00000901, 0x80048104,
+	0xc069003a, 0x00000070,
+	0xbf8cc07f, 0x80709070,
+	0xbf06c004, 0xbf84ffee,
+	0xbe840080, 0xd2890000,
+	0x00000902, 0x80048104,
+	0xd2890001, 0x00000902,
+	0x80048104, 0xd2890002,
+	0x00000902, 0x80048104,
+	0xd2890003, 0x00000902,
+	0x80048104, 0xc069003a,
+	0x00000070, 0xbf8cc07f,
+	0x80709070, 0xbf06c004,
+	0xbf84ffee, 0xbe840080,
+	0xd2890000, 0x00000903,
+	0x80048104, 0xd2890001,
+	0x00000903, 0x80048104,
+	0xd2890002, 0x00000903,
+	0x80048104, 0xd2890003,
+	0x00000903, 0x80048104,
+	0xc069003a, 0x00000070,
+	0xbf8cc07f, 0x80709070,
+	0xbf06c004, 0xbf84ffee,
+	0x807c847c, 0xbf0a7b7c,
+	0xbf85ffa9, 0xbf9c0000,
+	0xbf820016, 0xd3d84000,
+	0x18000100, 0xd3d84001,
+	0x18000101, 0xd3d84002,
+	0x18000102, 0xd3d84003,
+	0x18000103, 0xe0724000,
+	0x701d0000, 0xe0724100,
+	0x701d0100, 0xe0724200,
+	0x701d0200, 0xe0724300,
+	0x701d0300, 0x807c847c,
+	0x8070ff70, 0x00000400,
+	0xbf0a7b7c, 0xbf85ffeb,
+	0xbf9c0000, 0xbf8200ee,
+	0xbef4007e, 0x8675ff7f,
+	0x0000ffff, 0x8775ff75,
+	0x00040000, 0xbef60080,
+	0xbef700ff, 0x00807fac,
+	0x866eff7f, 0x04000000,
+	0xbf84001f, 0xbefe00c1,
+	0xbeff00c1, 0xb8ef4306,
+	0x866fc16f, 0xbf84001a,
+	0x8e6f866f, 0x8e6f826f,
+	0xbef6006f, 0xb8f82985,
+	0x80788178, 0x8e788a78,
+	0x8e788178, 0xb8ee1605,
+	0x806e816e, 0x8e6e866e,
+	0x80786e78, 0x8078ff78,
+	0x00000080, 0xbef600ff,
+	0x01000000, 0xbefc0080,
+	0xe0510000, 0x781d0000,
+	0xe0510100, 0x781d0000,
+	0x807cff7c, 0x00000200,
+	0x8078ff78, 0x00000200,
+	0xbf0a6f7c, 0xbf85fff6,
 	0xbefe00c1, 0xbeff00c1,
-	0xb8ef4306, 0x866fc16f,
-	0xbf84001a, 0x8e6f866f,
-	0x8e6f826f, 0xbef6006f,
-	0xb8f82985, 0x80788178,
-	0x8e788a78, 0x8e788178,
-	0xb8ee1605, 0x806e816e,
-	0x8e6e866e, 0x80786e78,
-	0x8078ff78, 0x00000080,
 	0xbef600ff, 0x01000000,
-	0xbefc0080, 0xe0510000,
-	0x781d0000, 0xe0510100,
-	0x781d0000, 0x807cff7c,
-	0x00000200, 0x8078ff78,
-	0x00000200, 0xbf0a6f7c,
-	0xbf85fff6, 0xbefe00c1,
-	0xbeff00c1, 0xbef600ff,
-	0x01000000, 0xb8ef2b05,
-	0x806f816f, 0x8e6f826f,
-	0x806fff6f, 0x00008000,
-	0xbef80080, 0xbeee0078,
-	0x8078ff78, 0x00000400,
-	0xbefc0084, 0xbf11087c,
-	0xe0524000, 0x781d0000,
-	0xe0524100, 0x781d0100,
-	0xe0524200, 0x781d0200,
-	0xe0524300, 0x781d0300,
-	0xbf8c0f70, 0x7e000300,
-	0x7e020301, 0x7e040302,
-	0x7e060303, 0x807c847c,
-	0x8078ff78, 0x00000400,
-	0xbf0a6f7c, 0xbf85ffee,
-	0xb8ef2985, 0x806f816f,
-	0x8e6f836f, 0xb8f92b05,
-	0x80798179, 0x8e798279,
-	0x80ef796f, 0x866f6f6f,
-	0xbf84001a, 0x806fff6f,
-	0x00008000, 0xbefc0080,
+	0xb8ef2b05, 0x806f816f,
+	0x8e6f826f, 0x806fff6f,
+	0x00008000, 0xbef80080,
+	0xbeee0078, 0x8078ff78,
+	0x00000400, 0xbefc0084,
 	0xbf11087c, 0xe0524000,
 	0x781d0000, 0xe0524100,
 	0x781d0100, 0xe0524200,
 	0x781d0200, 0xe0524300,
 	0x781d0300, 0xbf8c0f70,
-	0xd3d94000, 0x18000100,
-	0xd3d94001, 0x18000101,
-	0xd3d94002, 0x18000102,
-	0xd3d94003, 0x18000103,
+	0x7e000300, 0x7e020301,
+	0x7e040302, 0x7e060303,
 	0x807c847c, 0x8078ff78,
 	0x00000400, 0xbf0a6f7c,
-	0xbf85ffea, 0xbf9c0000,
-	0xe0524000, 0x6e1d0000,
-	0xe0524100, 0x6e1d0100,
-	0xe0524200, 0x6e1d0200,
-	0xe0524300, 0x6e1d0300,
-	0xbf8c0f70, 0xb8f82985,
-	0x80788178, 0x8e788a78,
-	0x8e788178, 0xb8ee1605,
-	0x806e816e, 0x8e6e866e,
-	0x80786e78, 0x80f8c078,
-	0xb8ef1605, 0x806f816f,
-	0x8e6f846f, 0x8e76826f,
-	0xbef600ff, 0x01000000,
-	0xbefc006f, 0xc031003a,
-	0x00000078, 0x80f8c078,
-	0xbf8cc07f, 0x80fc907c,
-	0xbf800000, 0xbe802d00,
-	0xbe822d02, 0xbe842d04,
-	0xbe862d06, 0xbe882d08,
-	0xbe8a2d0a, 0xbe8c2d0c,
-	0xbe8e2d0e, 0xbf06807c,
-	0xbf84fff0, 0xb8f82985,
-	0x80788178, 0x8e788a78,
-	0x8e788178, 0xb8ee1605,
-	0x806e816e, 0x8e6e866e,
-	0x80786e78, 0xbef60084,
-	0xbef600ff, 0x01000000,
-	0xc0211bfa, 0x00000078,
-	0x80788478, 0xc0211b3a,
+	0xbf85ffee, 0xb8ef2985,
+	0x806f816f, 0x8e6f836f,
+	0xb8f92b05, 0x80798179,
+	0x8e798279, 0x80ef796f,
+	0x866f6f6f, 0xbf84001a,
+	0x806fff6f, 0x00008000,
+	0xbefc0080, 0xbf11087c,
+	0xe0524000, 0x781d0000,
+	0xe0524100, 0x781d0100,
+	0xe0524200, 0x781d0200,
+	0xe0524300, 0x781d0300,
+	0xbf8c0f70, 0xd3d94000,
+	0x18000100, 0xd3d94001,
+	0x18000101, 0xd3d94002,
+	0x18000102, 0xd3d94003,
+	0x18000103, 0x807c847c,
+	0x8078ff78, 0x00000400,
+	0xbf0a6f7c, 0xbf85ffea,
+	0xbf9c0000, 0xe0524000,
+	0x6e1d0000, 0xe0524100,
+	0x6e1d0100, 0xe0524200,
+	0x6e1d0200, 0xe0524300,
+	0x6e1d0300, 0xbf8c0f70,
+	0xb8f82985, 0x80788178,
+	0x8e788a78, 0x8e788178,
+	0xb8ee1605, 0x806e816e,
+	0x8e6e866e, 0x80786e78,
+	0x80f8c078, 0xb8ef1605,
+	0x806f816f, 0x8e6f846f,
+	0x8e76826f, 0xbef600ff,
+	0x01000000, 0xbefc006f,
+	0xc031003a, 0x00000078,
+	0x80f8c078, 0xbf8cc07f,
+	0x80fc907c, 0xbf800000,
+	0xbe802d00, 0xbe822d02,
+	0xbe842d04, 0xbe862d06,
+	0xbe882d08, 0xbe8a2d0a,
+	0xbe8c2d0c, 0xbe8e2d0e,
+	0xbf06807c, 0xbf84fff0,
+	0xb8f82985, 0x80788178,
+	0x8e788a78, 0x8e788178,
+	0xb8ee1605, 0x806e816e,
+	0x8e6e866e, 0x80786e78,
+	0xbef60084, 0xbef600ff,
+	0x01000000, 0xc0211bfa,
 	0x00000078, 0x80788478,
-	0xc0211b7a, 0x00000078,
-	0x80788478, 0xc0211c3a,
+	0xc0211b3a, 0x00000078,
+	0x80788478, 0xc0211b7a,
 	0x00000078, 0x80788478,
-	0xc0211c7a, 0x00000078,
-	0x80788478, 0xc0211eba,
+	0xc0211c3a, 0x00000078,
+	0x80788478, 0xc0211c7a,
 	0x00000078, 0x80788478,
-	0xc0211efa, 0x00000078,
-	0x80788478, 0xc0211a3a,
+	0xc0211eba, 0x00000078,
+	0x80788478, 0xc0211efa,
 	0x00000078, 0x80788478,
-	0xc0211a7a, 0x00000078,
-	0x80788478, 0xc0211cfa,
+	0xc0211a3a, 0x00000078,
+	0x80788478, 0xc0211a7a,
 	0x00000078, 0x80788478,
-	0xbf8cc07f, 0xbefc006f,
-	0xbefe0070, 0xbeff0071,
-	0x866f7bff, 0x000003ff,
-	0xb96f4803, 0x866f7bff,
-	0xfffff800, 0x8f6f8b6f,
-	0xb96fa2c3, 0xb973f801,
-	0xb8ee2985, 0x806e816e,
-	0x8e6e8a6e, 0x8e6e816e,
-	0xb8ef1605, 0x806f816f,
-	0x8e6f866f, 0x806e6f6e,
-	0x806e746e, 0x826f8075,
-	0x866fff6f, 0x0000ffff,
-	0xc00b1c37, 0x00000050,
-	0xc00b1d37, 0x00000060,
-	0xc0031e77, 0x00000074,
-	0xbf8cc07f, 0x8f6e8b77,
-	0x866eff6e, 0x001f8000,
-	0xb96ef807, 0x866dff6d,
-	0x0000ffff, 0x86fe7e7e,
-	0x86ea6a6a, 0x8f6e837a,
-	0xb96ee0c2, 0xbf800002,
-	0xb97a0002, 0xbf8a0000,
-	0xbe801f6c, 0xbf810000,
+	0xc0211cfa, 0x00000078,
+	0x80788478, 0xbf8cc07f,
+	0xbefc006f, 0xbefe0070,
+	0xbeff0071, 0x866f7bff,
+	0x000003ff, 0xb96f4803,
+	0x866f7bff, 0xfffff800,
+	0x8f6f8b6f, 0xb96fa2c3,
+	0xb973f801, 0xb8ee2985,
+	0x806e816e, 0x8e6e8a6e,
+	0x8e6e816e, 0xb8ef1605,
+	0x806f816f, 0x8e6f866f,
+	0x806e6f6e, 0x806e746e,
+	0x826f8075, 0x866fff6f,
+	0x0000ffff, 0xc00b1c37,
+	0x00000050, 0xc00b1d37,
+	0x00000060, 0xc0031e77,
+	0x00000074, 0xbf8cc07f,
+	0x8f6e8b77, 0x866eff6e,
+	0x001f8000, 0xb96ef807,
+	0x866dff6d, 0x0000ffff,
+	0x86fe7e7e, 0x86ea6a6a,
+	0x8f6e837a, 0xb96ee0c2,
+	0xbf800002, 0xb97a0002,
+	0xbf8a0000, 0xbe801f6c,
+	0xbf9b0000, 0x00000000,
 };
 
 static const uint32_t cwsr_trap_gfx10_hex[] = {
-	0xbf820001, 0xbf82021c,
+	0xbf820001, 0xbf820220,
 	0xb0804004, 0xb978f802,
 	0x8a78ff78, 0x00020006,
 	0xb97bf803, 0x876eff78,
 	0x00002000, 0xbf840009,
 	0x876eff6d, 0x00ff0000,
 	0xbf85001e, 0x876eff7b,
-	0x00000400, 0xbf850041,
+	0x00000400, 0xbf850045,
 	0xbf8e0010, 0xb97bf803,
 	0xbf82fffa, 0x876eff7b,
 	0x00000900, 0xbf850015,
@@ -2082,22 +2297,24 @@ static const uint32_t cwsr_trap_gfx10_hex[] = {
 	0xb96ef801, 0x876eff6e,
 	0x00000800, 0xbf850003,
 	0x876eff7b, 0x00000400,
-	0xbf850026, 0xb97af812,
+	0xbf85002a, 0xb97af812,
 	0xb97bf813, 0x8ffa887a,
+	0xbf0d8f7b, 0xbf840002,
+	0x887bff7b, 0xffff0000,
 	0xf4011bbd, 0xfa000010,
-	0xbf8cc07f, 0x8f6e976e,
+	0xbf8c0000, 0x8f6e976e,
 	0x8a77ff77, 0x00800000,
 	0x88776e77, 0xf4051bbd,
-	0xfa000000, 0xbf8cc07f,
+	0xfa000000, 0xbf8c0000,
 	0xf4051ebd, 0xfa000008,
-	0xbf8cc07f, 0x87ee6e6e,
+	0xbf8c0000, 0x87ee6e6e,
 	0xbf840001, 0xbe80206e,
-	0x876eff6d, 0x01ff0000,
-	0xbf850005, 0x8878ff78,
-	0x00002000, 0x80ec886c,
-	0x82ed806d, 0xbf820005,
-	0x876eff6d, 0x01000000,
-	0xbf850002, 0x806c846c,
+	0x876eff6d, 0x00ff0000,
+	0xbf850008, 0x876eff6d,
+	0x01000000, 0xbf850007,
+	0x8878ff78, 0x00002000,
+	0x80ec886c, 0x82ed806d,
+	0xbf820002, 0x806c846c,
 	0x826d806d, 0x876dff6d,
 	0x0000ffff, 0x87fe7e7e,
 	0x87ea6a6a, 0xb9f8f802,
@@ -2105,162 +2322,162 @@ static const uint32_t cwsr_trap_gfx10_hex[] = {
 	0x0000ffff, 0xbefa0380,
 	0xb9fa0283, 0xbeee037e,
 	0xbeef037f, 0xbefe0480,
-	0xbf900004, 0xbf8cc07f,
+	0xbf900004, 0xbf8c0000,
 	0x877aff7f, 0x04000000,
 	0x8f7a857a, 0x886d7a6d,
-	0xbefa037e, 0x877bff7f,
-	0x0000ffff, 0xbefe03c1,
-	0xbeff03c1, 0xdc5f8000,
-	0x007a0000, 0x7e000280,
-	0xbefe037a, 0xbeff037b,
-	0xb97b02dc, 0x8f7b997b,
-	0xb97a3a05, 0x807a817a,
-	0xbf0d997b, 0xbf850002,
-	0x8f7a897a, 0xbf820001,
-	0x8f7a8a7a, 0xb97b1e06,
-	0x8f7b8a7b, 0x807a7b7a,
+	0x7e008200, 0xbefa037e,
 	0x877bff7f, 0x0000ffff,
-	0x807aff7a, 0x00000200,
-	0x807a7e7a, 0x827b807b,
-	0xd7610000, 0x00010870,
-	0xd7610000, 0x00010a71,
-	0xd7610000, 0x00010c72,
-	0xd7610000, 0x00010e73,
-	0xd7610000, 0x00011074,
-	0xd7610000, 0x00011275,
-	0xd7610000, 0x00011476,
-	0xd7610000, 0x00011677,
-	0xd7610000, 0x00011a79,
-	0xd7610000, 0x00011c7e,
-	0xd7610000, 0x00011e7f,
-	0xbefe03ff, 0x00003fff,
-	0xbeff0380, 0xdc5f8040,
-	0x007a0000, 0xd760007a,
-	0x00011d00, 0xd760007b,
-	0x00011f00, 0xbefe037a,
-	0xbeff037b, 0xbef4037e,
-	0x8775ff7f, 0x0000ffff,
-	0x8875ff75, 0x00040000,
-	0xbef60380, 0xbef703ff,
-	0x10807fac, 0xbef1037c,
-	0xbef00380, 0xb97302dc,
-	0x8f739973, 0xbefe03c1,
-	0x907c9973, 0x877c817c,
-	0xbf06817c, 0xbf850002,
-	0xbeff0380, 0xbf820002,
-	0xbeff03c1, 0xbf820009,
+	0xbefe03c1, 0xbeff03c1,
+	0xdc5f8000, 0x007a0000,
+	0x7e000280, 0xbefe037a,
+	0xbeff037b, 0xb97b02dc,
+	0x8f7b997b, 0xb97a3a05,
+	0x807a817a, 0xbf0d997b,
+	0xbf850002, 0x8f7a897a,
+	0xbf820001, 0x8f7a8a7a,
+	0xb97b1e06, 0x8f7b8a7b,
+	0x807a7b7a, 0x877bff7f,
+	0x0000ffff, 0x807aff7a,
+	0x00000200, 0x807a7e7a,
+	0x827b807b, 0xd7610000,
+	0x00010870, 0xd7610000,
+	0x00010a71, 0xd7610000,
+	0x00010c72, 0xd7610000,
+	0x00010e73, 0xd7610000,
+	0x00011074, 0xd7610000,
+	0x00011275, 0xd7610000,
+	0x00011476, 0xd7610000,
+	0x00011677, 0xd7610000,
+	0x00011a79, 0xd7610000,
+	0x00011c7e, 0xd7610000,
+	0x00011e7f, 0xbefe03ff,
+	0x00003fff, 0xbeff0380,
+	0xdc5f8040, 0x007a0000,
+	0xd760007a, 0x00011d00,
+	0xd760007b, 0x00011f00,
+	0xbefe037a, 0xbeff037b,
+	0xbef4037e, 0x8775ff7f,
+	0x0000ffff, 0x8875ff75,
+	0x00040000, 0xbef60380,
+	0xbef703ff, 0x10807fac,
+	0xbef1037c, 0xbef00380,
+	0xb97302dc, 0x8f739973,
+	0xbefe03c1, 0x907c9973,
+	0x877c817c, 0xbf06817c,
+	0xbf850002, 0xbeff0380,
+	0xbf820002, 0xbeff03c1,
+	0xbf820009, 0xbef603ff,
+	0x01000000, 0xe0704080,
+	0x705d0100, 0xe0704100,
+	0x705d0200, 0xe0704180,
+	0x705d0300, 0xbf820008,
 	0xbef603ff, 0x01000000,
-	0xe0704080, 0x705d0100,
-	0xe0704100, 0x705d0200,
-	0xe0704180, 0x705d0300,
-	0xbf820008, 0xbef603ff,
-	0x01000000, 0xe0704100,
-	0x705d0100, 0xe0704200,
-	0x705d0200, 0xe0704300,
-	0x705d0300, 0xb9703a05,
-	0x80708170, 0xbf0d9973,
-	0xbf850002, 0x8f708970,
-	0xbf820001, 0x8f708a70,
-	0xb97a1e06, 0x8f7a8a7a,
-	0x80707a70, 0x8070ff70,
-	0x00000200, 0xbef603ff,
-	0x01000000, 0x7e000280,
-	0x7e020280, 0x7e040280,
-	0xbefc0380, 0xd7610002,
-	0x0000f871, 0x807c817c,
-	0xd7610002, 0x0000f86c,
-	0x807c817c, 0x8a7aff6d,
-	0x80000000, 0xd7610002,
-	0x0000f87a, 0x807c817c,
-	0xd7610002, 0x0000f86e,
-	0x807c817c, 0xd7610002,
-	0x0000f86f, 0x807c817c,
-	0xd7610002, 0x0000f878,
-	0x807c817c, 0xb97af803,
-	0xd7610002, 0x0000f87a,
-	0x807c817c, 0xd7610002,
-	0x0000f87b, 0x807c817c,
-	0xb971f801, 0xd7610002,
-	0x0000f871, 0x807c817c,
-	0xb971f814, 0xd7610002,
-	0x0000f871, 0x807c817c,
-	0xb971f815, 0xd7610002,
-	0x0000f871, 0x807c817c,
-	0xbefe03ff, 0x0000ffff,
-	0xbeff0380, 0xe0704000,
-	0x705d0200, 0xbefe03c1,
+	0xe0704100, 0x705d0100,
+	0xe0704200, 0x705d0200,
+	0xe0704300, 0x705d0300,
 	0xb9703a05, 0x80708170,
 	0xbf0d9973, 0xbf850002,
 	0x8f708970, 0xbf820001,
 	0x8f708a70, 0xb97a1e06,
 	0x8f7a8a7a, 0x80707a70,
+	0x8070ff70, 0x00000200,
 	0xbef603ff, 0x01000000,
-	0xbef90380, 0xbefc0380,
-	0xbf800000, 0xbe802f00,
-	0xbe822f02, 0xbe842f04,
-	0xbe862f06, 0xbe882f08,
-	0xbe8a2f0a, 0xbe8c2f0c,
-	0xbe8e2f0e, 0xd7610002,
-	0x0000f200, 0x80798179,
-	0xd7610002, 0x0000f201,
+	0x7e000280, 0x7e020280,
+	0x7e040280, 0xbefc0380,
+	0xd7610002, 0x0000f871,
+	0x807c817c, 0xd7610002,
+	0x0000f86c, 0x807c817c,
+	0x8a7aff6d, 0x80000000,
+	0xd7610002, 0x0000f87a,
+	0x807c817c, 0xd7610002,
+	0x0000f86e, 0x807c817c,
+	0xd7610002, 0x0000f86f,
+	0x807c817c, 0xd7610002,
+	0x0000f878, 0x807c817c,
+	0xb97af803, 0xd7610002,
+	0x0000f87a, 0x807c817c,
+	0xd7610002, 0x0000f87b,
+	0x807c817c, 0xb971f801,
+	0xd7610002, 0x0000f871,
+	0x807c817c, 0xb971f814,
+	0xd7610002, 0x0000f871,
+	0x807c817c, 0xb971f815,
+	0xd7610002, 0x0000f871,
+	0x807c817c, 0xbefe03ff,
+	0x0000ffff, 0xbeff0380,
+	0xe0704000, 0x705d0200,
+	0xbefe03c1, 0xb9703a05,
+	0x80708170, 0xbf0d9973,
+	0xbf850002, 0x8f708970,
+	0xbf820001, 0x8f708a70,
+	0xb97a1e06, 0x8f7a8a7a,
+	0x80707a70, 0xbef603ff,
+	0x01000000, 0xbef90380,
+	0xbefc0380, 0xbf800000,
+	0xbe802f00, 0xbe822f02,
+	0xbe842f04, 0xbe862f06,
+	0xbe882f08, 0xbe8a2f0a,
+	0xbe8c2f0c, 0xbe8e2f0e,
+	0xd7610002, 0x0000f200,
 	0x80798179, 0xd7610002,
-	0x0000f202, 0x80798179,
-	0xd7610002, 0x0000f203,
+	0x0000f201, 0x80798179,
+	0xd7610002, 0x0000f202,
 	0x80798179, 0xd7610002,
-	0x0000f204, 0x80798179,
-	0xd7610002, 0x0000f205,
+	0x0000f203, 0x80798179,
+	0xd7610002, 0x0000f204,
 	0x80798179, 0xd7610002,
-	0x0000f206, 0x80798179,
-	0xd7610002, 0x0000f207,
+	0x0000f205, 0x80798179,
+	0xd7610002, 0x0000f206,
 	0x80798179, 0xd7610002,
-	0x0000f208, 0x80798179,
-	0xd7610002, 0x0000f209,
+	0x0000f207, 0x80798179,
+	0xd7610002, 0x0000f208,
 	0x80798179, 0xd7610002,
-	0x0000f20a, 0x80798179,
-	0xd7610002, 0x0000f20b,
+	0x0000f209, 0x80798179,
+	0xd7610002, 0x0000f20a,
 	0x80798179, 0xd7610002,
-	0x0000f20c, 0x80798179,
-	0xd7610002, 0x0000f20d,
+	0x0000f20b, 0x80798179,
+	0xd7610002, 0x0000f20c,
 	0x80798179, 0xd7610002,
-	0x0000f20e, 0x80798179,
-	0xd7610002, 0x0000f20f,
-	0x80798179, 0xbf06a079,
-	0xbf840006, 0xe0704000,
-	0x705d0200, 0x8070ff70,
-	0x00000080, 0xbef90380,
-	0x7e040280, 0x807c907c,
-	0xbf0aff7c, 0x00000060,
-	0xbf85ffbc, 0xbe802f00,
-	0xbe822f02, 0xbe842f04,
-	0xbe862f06, 0xbe882f08,
-	0xbe8a2f0a, 0xd7610002,
-	0x0000f200, 0x80798179,
-	0xd7610002, 0x0000f201,
+	0x0000f20d, 0x80798179,
+	0xd7610002, 0x0000f20e,
 	0x80798179, 0xd7610002,
-	0x0000f202, 0x80798179,
-	0xd7610002, 0x0000f203,
+	0x0000f20f, 0x80798179,
+	0xbf06a079, 0xbf840006,
+	0xe0704000, 0x705d0200,
+	0x8070ff70, 0x00000080,
+	0xbef90380, 0x7e040280,
+	0x807c907c, 0xbf0aff7c,
+	0x00000060, 0xbf85ffbc,
+	0xbe802f00, 0xbe822f02,
+	0xbe842f04, 0xbe862f06,
+	0xbe882f08, 0xbe8a2f0a,
+	0xd7610002, 0x0000f200,
 	0x80798179, 0xd7610002,
-	0x0000f204, 0x80798179,
-	0xd7610002, 0x0000f205,
+	0x0000f201, 0x80798179,
+	0xd7610002, 0x0000f202,
 	0x80798179, 0xd7610002,
-	0x0000f206, 0x80798179,
-	0xd7610002, 0x0000f207,
+	0x0000f203, 0x80798179,
+	0xd7610002, 0x0000f204,
 	0x80798179, 0xd7610002,
-	0x0000f208, 0x80798179,
-	0xd7610002, 0x0000f209,
+	0x0000f205, 0x80798179,
+	0xd7610002, 0x0000f206,
 	0x80798179, 0xd7610002,
-	0x0000f20a, 0x80798179,
-	0xd7610002, 0x0000f20b,
-	0x80798179, 0xe0704000,
-	0x705d0200, 0xbefe03c1,
-	0x907c9973, 0x877c817c,
-	0xbf06817c, 0xbf850002,
-	0xbeff0380, 0xbf820001,
-	0xbeff03c1, 0xb97b4306,
-	0x877bc17b, 0xbf840044,
-	0xbf8a0000, 0x877aff6d,
-	0x80000000, 0xbf840040,
-	0x8f7b867b, 0x8f7b827b,
+	0x0000f207, 0x80798179,
+	0xd7610002, 0x0000f208,
+	0x80798179, 0xd7610002,
+	0x0000f209, 0x80798179,
+	0xd7610002, 0x0000f20a,
+	0x80798179, 0xd7610002,
+	0x0000f20b, 0x80798179,
+	0xe0704000, 0x705d0200,
+	0xbefe03c1, 0x907c9973,
+	0x877c817c, 0xbf06817c,
+	0xbf850002, 0xbeff0380,
+	0xbf820001, 0xbeff03c1,
+	0xb97b4306, 0x877bc17b,
+	0xbf840043, 0xbf8a0000,
+	0x877aff6d, 0x80000000,
+	0xbf84003f, 0x8f7b887b,
 	0xbef6037b, 0xb9703a05,
 	0x80708170, 0xbf0d9973,
 	0xbf850002, 0x8f708970,
@@ -2331,171 +2548,169 @@ static const uint32_t cwsr_trap_gfx10_hex[] = {
 	0xe0704000, 0x705d0000,
 	0x807c817c, 0x8070ff70,
 	0x00000080, 0xbf0a7b7c,
-	0xbf85fff8, 0xbf82013b,
+	0xbf85fff8, 0xbf820136,
 	0xbef4037e, 0x8775ff7f,
 	0x0000ffff, 0x8875ff75,
 	0x00040000, 0xbef60380,
 	0xbef703ff, 0x10807fac,
 	0xb97202dc, 0x8f729972,
 	0x876eff7f, 0x04000000,
-	0xbf840034, 0xbefe03c1,
+	0xbf840033, 0xbefe03c1,
 	0x907c9972, 0x877c817c,
 	0xbf06817c, 0xbf850002,
 	0xbeff0380, 0xbf820001,
 	0xbeff03c1, 0xb96f4306,
-	0x876fc16f, 0xbf840029,
-	0x8f6f866f, 0x8f6f826f,
-	0xbef6036f, 0xb9783a05,
-	0x80788178, 0xbf0d9972,
-	0xbf850002, 0x8f788978,
-	0xbf820001, 0x8f788a78,
-	0xb96e1e06, 0x8f6e8a6e,
-	0x80786e78, 0x8078ff78,
-	0x00000200, 0x8078ff78,
-	0x00000080, 0xbef603ff,
-	0x01000000, 0x907c9972,
-	0x877c817c, 0xbf06817c,
-	0xbefc0380, 0xbf850009,
-	0xe0310000, 0x781d0000,
-	0x807cff7c, 0x00000080,
+	0x876fc16f, 0xbf840028,
+	0x8f6f886f, 0xbef6036f,
+	0xb9783a05, 0x80788178,
+	0xbf0d9972, 0xbf850002,
+	0x8f788978, 0xbf820001,
+	0x8f788a78, 0xb96e1e06,
+	0x8f6e8a6e, 0x80786e78,
+	0x8078ff78, 0x00000200,
 	0x8078ff78, 0x00000080,
-	0xbf0a6f7c, 0xbf85fff8,
-	0xbf820008, 0xe0310000,
+	0xbef603ff, 0x01000000,
+	0x907c9972, 0x877c817c,
+	0xbf06817c, 0xbefc0380,
+	0xbf850009, 0xe0310000,
 	0x781d0000, 0x807cff7c,
-	0x00000100, 0x8078ff78,
-	0x00000100, 0xbf0a6f7c,
-	0xbf85fff8, 0xbef80380,
-	0xbefe03c1, 0x907c9972,
-	0x877c817c, 0xbf06817c,
-	0xbf850002, 0xbeff0380,
-	0xbf820001, 0xbeff03c1,
-	0xb96f3a05, 0x806f816f,
-	0x8f6f826f, 0x907c9972,
-	0x877c817c, 0xbf06817c,
-	0xbf850024, 0xbef603ff,
-	0x01000000, 0xbeee0378,
-	0x8078ff78, 0x00000200,
-	0xbefc0384, 0xbf0a6f7c,
-	0xbf840050, 0xe0304000,
-	0x785d0000, 0xe0304080,
-	0x785d0100, 0xe0304100,
-	0x785d0200, 0xe0304180,
-	0x785d0300, 0xbf8c3f70,
-	0x7e008500, 0x7e028501,
-	0x7e048502, 0x7e068503,
-	0x807c847c, 0x8078ff78,
-	0x00000200, 0xbf0a6f7c,
-	0xbf85ffee, 0xe0304000,
-	0x6e5d0000, 0xe0304080,
-	0x6e5d0100, 0xe0304100,
-	0x6e5d0200, 0xe0304180,
-	0x6e5d0300, 0xbf8c3f70,
-	0xbf820034, 0xbef603ff,
-	0x01000000, 0xbeee0378,
-	0x8078ff78, 0x00000400,
-	0xbefc0384, 0xbf0a6f7c,
-	0xbf840012, 0xe0304000,
-	0x785d0000, 0xe0304100,
-	0x785d0100, 0xe0304200,
-	0x785d0200, 0xe0304300,
-	0x785d0300, 0xbf8c3f70,
-	0x7e008500, 0x7e028501,
-	0x7e048502, 0x7e068503,
-	0x807c847c, 0x8078ff78,
-	0x00000400, 0xbf0a6f7c,
-	0xbf85ffee, 0xb96f1e06,
-	0x876fc16f, 0xbf84000e,
-	0x8f6f836f, 0x806f7c6f,
-	0xbefe03c1, 0xbeff0380,
-	0xe0304000, 0x785d0000,
-	0xbf8c3f70, 0x7e008500,
-	0x807c817c, 0x8078ff78,
+	0x00000080, 0x8078ff78,
 	0x00000080, 0xbf0a6f7c,
-	0xbf85fff7, 0xbeff03c1,
+	0xbf85fff8, 0xbf820008,
+	0xe0310000, 0x781d0000,
+	0x807cff7c, 0x00000100,
+	0x8078ff78, 0x00000100,
+	0xbf0a6f7c, 0xbf85fff8,
+	0xbef80380, 0xbefe03c1,
+	0x907c9972, 0x877c817c,
+	0xbf06817c, 0xbf850002,
+	0xbeff0380, 0xbf820001,
+	0xbeff03c1, 0xb96f3a05,
+	0x806f816f, 0x8f6f826f,
+	0x907c9972, 0x877c817c,
+	0xbf06817c, 0xbf850024,
+	0xbef603ff, 0x01000000,
+	0xbeee0378, 0x8078ff78,
+	0x00000200, 0xbefc0384,
+	0xbf0a6f7c, 0xbf840050,
+	0xe0304000, 0x785d0000,
+	0xe0304080, 0x785d0100,
+	0xe0304100, 0x785d0200,
+	0xe0304180, 0x785d0300,
+	0xbf8c0000, 0x7e008500,
+	0x7e028501, 0x7e048502,
+	0x7e068503, 0x807c847c,
+	0x8078ff78, 0x00000200,
+	0xbf0a6f7c, 0xbf85ffee,
 	0xe0304000, 0x6e5d0000,
-	0xe0304100, 0x6e5d0100,
-	0xe0304200, 0x6e5d0200,
-	0xe0304300, 0x6e5d0300,
-	0xbf8c3f70, 0xb9783a05,
-	0x80788178, 0xbf0d9972,
-	0xbf850002, 0x8f788978,
-	0xbf820001, 0x8f788a78,
-	0xb96e1e06, 0x8f6e8a6e,
-	0x80786e78, 0x8078ff78,
-	0x00000200, 0x80f8ff78,
-	0x00000050, 0xbef603ff,
-	0x01000000, 0xbefc03ff,
-	0x0000006c, 0x80f89078,
-	0xf429003a, 0xf0000000,
-	0xbf8cc07f, 0x80fc847c,
-	0xbf800000, 0xbe803100,
-	0xbe823102, 0x80f8a078,
-	0xf42d003a, 0xf0000000,
-	0xbf8cc07f, 0x80fc887c,
-	0xbf800000, 0xbe803100,
-	0xbe823102, 0xbe843104,
-	0xbe863106, 0x80f8c078,
-	0xf431003a, 0xf0000000,
-	0xbf8cc07f, 0x80fc907c,
-	0xbf800000, 0xbe803100,
-	0xbe823102, 0xbe843104,
-	0xbe863106, 0xbe883108,
-	0xbe8a310a, 0xbe8c310c,
-	0xbe8e310e, 0xbf06807c,
-	0xbf84fff0, 0xba80f801,
-	0x00000000, 0xbf8a0000,
+	0xe0304080, 0x6e5d0100,
+	0xe0304100, 0x6e5d0200,
+	0xe0304180, 0x6e5d0300,
+	0xbf8c0000, 0xbf820034,
+	0xbef603ff, 0x01000000,
+	0xbeee0378, 0x8078ff78,
+	0x00000400, 0xbefc0384,
+	0xbf0a6f7c, 0xbf840012,
+	0xe0304000, 0x785d0000,
+	0xe0304100, 0x785d0100,
+	0xe0304200, 0x785d0200,
+	0xe0304300, 0x785d0300,
+	0xbf8c0000, 0x7e008500,
+	0x7e028501, 0x7e048502,
+	0x7e068503, 0x807c847c,
+	0x8078ff78, 0x00000400,
+	0xbf0a6f7c, 0xbf85ffee,
+	0xb96f1e06, 0x876fc16f,
+	0xbf84000e, 0x8f6f836f,
+	0x806f7c6f, 0xbefe03c1,
+	0xbeff0380, 0xe0304000,
+	0x785d0000, 0xbf8c0000,
+	0x7e008500, 0x807c817c,
+	0x8078ff78, 0x00000080,
+	0xbf0a6f7c, 0xbf85fff7,
+	0xbeff03c1, 0xe0304000,
+	0x6e5d0000, 0xe0304100,
+	0x6e5d0100, 0xe0304200,
+	0x6e5d0200, 0xe0304300,
+	0x6e5d0300, 0xbf8c0000,
 	0xb9783a05, 0x80788178,
 	0xbf0d9972, 0xbf850002,
 	0x8f788978, 0xbf820001,
 	0x8f788a78, 0xb96e1e06,
 	0x8f6e8a6e, 0x80786e78,
 	0x8078ff78, 0x00000200,
+	0x80f8ff78, 0x00000050,
 	0xbef603ff, 0x01000000,
-	0xf4211bfa, 0xf0000000,
-	0x80788478, 0xf4211b3a,
+	0xbefc03ff, 0x0000006c,
+	0x80f89078, 0xf429003a,
+	0xf0000000, 0xbf8c0000,
+	0x80fc847c, 0xbf800000,
+	0xbe803100, 0xbe823102,
+	0x80f8a078, 0xf42d003a,
+	0xf0000000, 0xbf8c0000,
+	0x80fc887c, 0xbf800000,
+	0xbe803100, 0xbe823102,
+	0xbe843104, 0xbe863106,
+	0x80f8c078, 0xf431003a,
+	0xf0000000, 0xbf8c0000,
+	0x80fc907c, 0xbf800000,
+	0xbe803100, 0xbe823102,
+	0xbe843104, 0xbe863106,
+	0xbe883108, 0xbe8a310a,
+	0xbe8c310c, 0xbe8e310e,
+	0xbf06807c, 0xbf84fff0,
+	0xba80f801, 0x00000000,
+	0xbf8a0000, 0xb9783a05,
+	0x80788178, 0xbf0d9972,
+	0xbf850002, 0x8f788978,
+	0xbf820001, 0x8f788a78,
+	0xb96e1e06, 0x8f6e8a6e,
+	0x80786e78, 0x8078ff78,
+	0x00000200, 0xbef603ff,
+	0x01000000, 0xf4211bfa,
 	0xf0000000, 0x80788478,
-	0xf4211b7a, 0xf0000000,
-	0x80788478, 0xf4211c3a,
+	0xf4211b3a, 0xf0000000,
+	0x80788478, 0xf4211b7a,
 	0xf0000000, 0x80788478,
-	0xf4211c7a, 0xf0000000,
-	0x80788478, 0xf4211eba,
+	0xf4211c3a, 0xf0000000,
+	0x80788478, 0xf4211c7a,
 	0xf0000000, 0x80788478,
-	0xf4211efa, 0xf0000000,
-	0x80788478, 0xf4211e7a,
+	0xf4211eba, 0xf0000000,
+	0x80788478, 0xf4211efa,
 	0xf0000000, 0x80788478,
-	0xf4211cfa, 0xf0000000,
-	0x80788478, 0xf4211bba,
+	0xf4211e7a, 0xf0000000,
+	0x80788478, 0xf4211cfa,
 	0xf0000000, 0x80788478,
-	0xbf8cc07f, 0xb9eef814,
 	0xf4211bba, 0xf0000000,
-	0x80788478, 0xbf8cc07f,
-	0xb9eef815, 0xbefc036f,
-	0xbefe0370, 0xbeff0371,
-	0x876f7bff, 0x000003ff,
-	0xb9ef4803, 0x876f7bff,
-	0xfffff800, 0x906f8b6f,
-	0xb9efa2c3, 0xb9f3f801,
-	0xb96e3a05, 0x806e816e,
-	0xbf0d9972, 0xbf850002,
-	0x8f6e896e, 0xbf820001,
-	0x8f6e8a6e, 0xb96f1e06,
-	0x8f6f8a6f, 0x806e6f6e,
-	0x806eff6e, 0x00000200,
-	0x806e746e, 0x826f8075,
-	0x876fff6f, 0x0000ffff,
-	0xf4091c37, 0xfa000050,
-	0xf4091d37, 0xfa000060,
-	0xf4011e77, 0xfa000074,
-	0xbf8cc07f, 0x876dff6d,
-	0x0000ffff, 0x87fe7e7e,
-	0x87ea6a6a, 0xb9faf802,
-	0xbe80226c, 0xbf810000,
+	0x80788478, 0xbf8c0000,
+	0xb9eef814, 0xf4211bba,
+	0xf0000000, 0x80788478,
+	0xbf8c0000, 0xb9eef815,
+	0xbefc036f, 0xbefe0370,
+	0xbeff0371, 0xb9fb4803,
+	0x907b8b7b, 0xb9fba2c3,
+	0xb9f3f801, 0xb96e3a05,
+	0x806e816e, 0xbf0d9972,
+	0xbf850002, 0x8f6e896e,
+	0xbf820001, 0x8f6e8a6e,
+	0xb96f1e06, 0x8f6f8a6f,
+	0x806e6f6e, 0x806eff6e,
+	0x00000200, 0x806e746e,
+	0x826f8075, 0x876fff6f,
+	0x0000ffff, 0xf4091c37,
+	0xfa000050, 0xf4091d37,
+	0xfa000060, 0xf4011e77,
+	0xfa000074, 0xbf8c0000,
+	0x876dff6d, 0x0000ffff,
+	0x87fe7e7e, 0x87ea6a6a,
+	0xb9faf802, 0xbe80226c,
+	0xbf9b0000, 0xbf9f0000,
 	0xbf9f0000, 0xbf9f0000,
 	0xbf9f0000, 0xbf9f0000,
-	0xbf9f0000, 0x00000000,
 };
+
 static const uint32_t cwsr_trap_gfx11_hex[] = {
-	0xbfa00001, 0xbfa00221,
+	0xbfa00001, 0xbfa00227,
 	0xb0804006, 0xb8f8f802,
 	0x9178ff78, 0x00020006,
 	0xb8fbf803, 0xbf0d9e6d,
@@ -2505,9 +2720,9 @@ static const uint32_t cwsr_trap_gfx11_hex[] = {
 	0xbfa10009, 0x8b6eff6d,
 	0x00ff0000, 0xbfa2001e,
 	0x8b6eff7b, 0x00000400,
-	0xbfa20041, 0xbf830010,
+	0xbfa20045, 0xbf830010,
 	0xb8fbf803, 0xbfa0fffa,
-	0x8b6eff7b, 0x00000900,
+	0x8b6eff7b, 0x00160900,
 	0xbfa20015, 0x8b6eff7b,
 	0x000071ff, 0xbfa10008,
 	0x8b6fff7b, 0x00007080,
@@ -2518,186 +2733,189 @@ static const uint32_t cwsr_trap_gfx11_hex[] = {
 	0xbfa20007, 0xb8eef801,
 	0x8b6eff6e, 0x00000800,
 	0xbfa20003, 0x8b6eff7b,
-	0x00000400, 0xbfa20026,
-	0xbefa4d82, 0xbf89fc07,
-	0x84fa887a, 0xf4005bbd,
-	0xf8000010, 0xbf89fc07,
+	0x00000400, 0xbfa2002a,
+	0xbefa4d82, 0xbf890000,
+	0x84fa887a, 0xbf0d8f7b,
+	0xbfa10002, 0x8c7bff7b,
+	0xffff0000, 0xf4005bbd,
+	0xf8000010, 0xbf890000,
 	0x846e976e, 0x9177ff77,
 	0x00800000, 0x8c776e77,
 	0xf4045bbd, 0xf8000000,
-	0xbf89fc07, 0xf4045ebd,
-	0xf8000008, 0xbf89fc07,
+	0xbf890000, 0xf4045ebd,
+	0xf8000008, 0xbf890000,
 	0x8bee6e6e, 0xbfa10001,
 	0xbe80486e, 0x8b6eff6d,
-	0x01ff0000, 0xbfa20005,
-	0x8c78ff78, 0x00002000,
-	0x80ec886c, 0x82ed806d,
-	0xbfa00005, 0x8b6eff6d,
-	0x01000000, 0xbfa20002,
+	0x00ff0000, 0xbfa20008,
+	0x8b6eff6d, 0x01000000,
+	0xbfa20007, 0x8c78ff78,
+	0x00002000, 0x80ec886c,
+	0x82ed806d, 0xbfa00002,
 	0x806c846c, 0x826d806d,
 	0x8b6dff6d, 0x0000ffff,
 	0x8bfe7e7e, 0x8bea6a6a,
 	0xb978f802, 0xbe804a6c,
-	0x8b6dff6d, 0x0000ffff,
-	0xbefa0080, 0xb97a0283,
-	0xbeee007e, 0xbeef007f,
-	0xbefe0180, 0xbefe4d84,
-	0xbf89fc07, 0x8b7aff7f,
-	0x04000000, 0x847a857a,
-	0x8c6d7a6d, 0xbefa007e,
-	0x8b7bff7f, 0x0000ffff,
-	0xbefe00c1, 0xbeff00c1,
-	0xdca6c000, 0x007a0000,
-	0x7e000280, 0xbefe007a,
-	0xbeff007b, 0xb8fb02dc,
-	0x847b997b, 0xb8fa3b05,
-	0x807a817a, 0xbf0d997b,
-	0xbfa20002, 0x847a897a,
-	0xbfa00001, 0x847a8a7a,
-	0xb8fb1e06, 0x847b8a7b,
-	0x807a7b7a, 0x8b7bff7f,
-	0x0000ffff, 0x807aff7a,
-	0x00000200, 0x807a7e7a,
-	0x827b807b, 0xd7610000,
-	0x00010870, 0xd7610000,
-	0x00010a71, 0xd7610000,
-	0x00010c72, 0xd7610000,
-	0x00010e73, 0xd7610000,
-	0x00011074, 0xd7610000,
-	0x00011275, 0xd7610000,
-	0x00011476, 0xd7610000,
-	0x00011677, 0xd7610000,
-	0x00011a79, 0xd7610000,
-	0x00011c7e, 0xd7610000,
-	0x00011e7f, 0xbefe00ff,
-	0x00003fff, 0xbeff0080,
-	0xdca6c040, 0x007a0000,
-	0xd760007a, 0x00011d00,
-	0xd760007b, 0x00011f00,
+	0xbf0d9878, 0xbfa10001,
+	0xbfb00000, 0x8b6dff6d,
+	0x0000ffff, 0xbefa0080,
+	0xb97a0283, 0xbeee007e,
+	0xbeef007f, 0xbefe0180,
+	0xbefe4d84, 0xbf890000,
+	0x8b7aff7f, 0x04000000,
+	0x847a857a, 0x8c6d7a6d,
+	0xbefa007e, 0x8b7bff7f,
+	0x0000ffff, 0xbefe00c1,
+	0xbeff00c1, 0xdca6c000,
+	0x007a0000, 0x7e000280,
 	0xbefe007a, 0xbeff007b,
-	0xbef4007e, 0x8b75ff7f,
-	0x0000ffff, 0x8c75ff75,
-	0x00040000, 0xbef60080,
-	0xbef700ff, 0x10807fac,
-	0xbef1007d, 0xbef00080,
-	0xb8f302dc, 0x84739973,
-	0xbefe00c1, 0x857d9973,
-	0x8b7d817d, 0xbf06817d,
-	0xbfa20002, 0xbeff0080,
-	0xbfa00002, 0xbeff00c1,
-	0xbfa00009, 0xbef600ff,
-	0x01000000, 0xe0685080,
-	0x701d0100, 0xe0685100,
-	0x701d0200, 0xe0685180,
-	0x701d0300, 0xbfa00008,
+	0xb8fb02dc, 0x847b997b,
+	0xb8fa3b05, 0x807a817a,
+	0xbf0d997b, 0xbfa20002,
+	0x847a897a, 0xbfa00001,
+	0x847a8a7a, 0xb8fb1e06,
+	0x847b8a7b, 0x807a7b7a,
+	0x8b7bff7f, 0x0000ffff,
+	0x807aff7a, 0x00000200,
+	0x807a7e7a, 0x827b807b,
+	0xd7610000, 0x00010870,
+	0xd7610000, 0x00010a71,
+	0xd7610000, 0x00010c72,
+	0xd7610000, 0x00010e73,
+	0xd7610000, 0x00011074,
+	0xd7610000, 0x00011275,
+	0xd7610000, 0x00011476,
+	0xd7610000, 0x00011677,
+	0xd7610000, 0x00011a79,
+	0xd7610000, 0x00011c7e,
+	0xd7610000, 0x00011e7f,
+	0xbefe00ff, 0x00003fff,
+	0xbeff0080, 0xdca6c040,
+	0x007a0000, 0xd760007a,
+	0x00011d00, 0xd760007b,
+	0x00011f00, 0xbefe007a,
+	0xbeff007b, 0xbef4007e,
+	0x8b75ff7f, 0x0000ffff,
+	0x8c75ff75, 0x00040000,
+	0xbef60080, 0xbef700ff,
+	0x10807fac, 0xbef1007d,
+	0xbef00080, 0xb8f302dc,
+	0x84739973, 0xbefe00c1,
+	0x857d9973, 0x8b7d817d,
+	0xbf06817d, 0xbfa20002,
+	0xbeff0080, 0xbfa00002,
+	0xbeff00c1, 0xbfa00009,
 	0xbef600ff, 0x01000000,
-	0xe0685100, 0x701d0100,
-	0xe0685200, 0x701d0200,
-	0xe0685300, 0x701d0300,
+	0xe0685080, 0x701d0100,
+	0xe0685100, 0x701d0200,
+	0xe0685180, 0x701d0300,
+	0xbfa00008, 0xbef600ff,
+	0x01000000, 0xe0685100,
+	0x701d0100, 0xe0685200,
+	0x701d0200, 0xe0685300,
+	0x701d0300, 0xb8f03b05,
+	0x80708170, 0xbf0d9973,
+	0xbfa20002, 0x84708970,
+	0xbfa00001, 0x84708a70,
+	0xb8fa1e06, 0x847a8a7a,
+	0x80707a70, 0x8070ff70,
+	0x00000200, 0xbef600ff,
+	0x01000000, 0x7e000280,
+	0x7e020280, 0x7e040280,
+	0xbefd0080, 0xd7610002,
+	0x0000fa71, 0x807d817d,
+	0xd7610002, 0x0000fa6c,
+	0x807d817d, 0x917aff6d,
+	0x80000000, 0xd7610002,
+	0x0000fa7a, 0x807d817d,
+	0xd7610002, 0x0000fa6e,
+	0x807d817d, 0xd7610002,
+	0x0000fa6f, 0x807d817d,
+	0xd7610002, 0x0000fa78,
+	0x807d817d, 0xb8faf803,
+	0xd7610002, 0x0000fa7a,
+	0x807d817d, 0xd7610002,
+	0x0000fa7b, 0x807d817d,
+	0xb8f1f801, 0xd7610002,
+	0x0000fa71, 0x807d817d,
+	0xb8f1f814, 0xd7610002,
+	0x0000fa71, 0x807d817d,
+	0xb8f1f815, 0xd7610002,
+	0x0000fa71, 0x807d817d,
+	0xbefe00ff, 0x0000ffff,
+	0xbeff0080, 0xe0685000,
+	0x701d0200, 0xbefe00c1,
 	0xb8f03b05, 0x80708170,
 	0xbf0d9973, 0xbfa20002,
 	0x84708970, 0xbfa00001,
 	0x84708a70, 0xb8fa1e06,
 	0x847a8a7a, 0x80707a70,
-	0x8070ff70, 0x00000200,
 	0xbef600ff, 0x01000000,
-	0x7e000280, 0x7e020280,
-	0x7e040280, 0xbefd0080,
-	0xd7610002, 0x0000fa71,
-	0x807d817d, 0xd7610002,
-	0x0000fa6c, 0x807d817d,
-	0x917aff6d, 0x80000000,
-	0xd7610002, 0x0000fa7a,
-	0x807d817d, 0xd7610002,
-	0x0000fa6e, 0x807d817d,
-	0xd7610002, 0x0000fa6f,
-	0x807d817d, 0xd7610002,
-	0x0000fa78, 0x807d817d,
-	0xb8faf803, 0xd7610002,
-	0x0000fa7a, 0x807d817d,
-	0xd7610002, 0x0000fa7b,
-	0x807d817d, 0xb8f1f801,
-	0xd7610002, 0x0000fa71,
-	0x807d817d, 0xb8f1f814,
-	0xd7610002, 0x0000fa71,
-	0x807d817d, 0xb8f1f815,
-	0xd7610002, 0x0000fa71,
-	0x807d817d, 0xbefe00ff,
-	0x0000ffff, 0xbeff0080,
-	0xe0685000, 0x701d0200,
-	0xbefe00c1, 0xb8f03b05,
-	0x80708170, 0xbf0d9973,
-	0xbfa20002, 0x84708970,
-	0xbfa00001, 0x84708a70,
-	0xb8fa1e06, 0x847a8a7a,
-	0x80707a70, 0xbef600ff,
-	0x01000000, 0xbef90080,
-	0xbefd0080, 0xbf800000,
-	0xbe804100, 0xbe824102,
-	0xbe844104, 0xbe864106,
-	0xbe884108, 0xbe8a410a,
-	0xbe8c410c, 0xbe8e410e,
-	0xd7610002, 0x0000f200,
-	0x80798179, 0xd7610002,
-	0x0000f201, 0x80798179,
-	0xd7610002, 0x0000f202,
-	0x80798179, 0xd7610002,
-	0x0000f203, 0x80798179,
-	0xd7610002, 0x0000f204,
+	0xbef90080, 0xbefd0080,
+	0xbf800000, 0xbe804100,
+	0xbe824102, 0xbe844104,
+	0xbe864106, 0xbe884108,
+	0xbe8a410a, 0xbe8c410c,
+	0xbe8e410e, 0xd7610002,
+	0x0000f200, 0x80798179,
+	0xd7610002, 0x0000f201,
 	0x80798179, 0xd7610002,
-	0x0000f205, 0x80798179,
-	0xd7610002, 0x0000f206,
+	0x0000f202, 0x80798179,
+	0xd7610002, 0x0000f203,
 	0x80798179, 0xd7610002,
-	0x0000f207, 0x80798179,
-	0xd7610002, 0x0000f208,
+	0x0000f204, 0x80798179,
+	0xd7610002, 0x0000f205,
 	0x80798179, 0xd7610002,
-	0x0000f209, 0x80798179,
-	0xd7610002, 0x0000f20a,
+	0x0000f206, 0x80798179,
+	0xd7610002, 0x0000f207,
 	0x80798179, 0xd7610002,
-	0x0000f20b, 0x80798179,
-	0xd7610002, 0x0000f20c,
+	0x0000f208, 0x80798179,
+	0xd7610002, 0x0000f209,
 	0x80798179, 0xd7610002,
-	0x0000f20d, 0x80798179,
-	0xd7610002, 0x0000f20e,
+	0x0000f20a, 0x80798179,
+	0xd7610002, 0x0000f20b,
 	0x80798179, 0xd7610002,
-	0x0000f20f, 0x80798179,
-	0xbf06a079, 0xbfa10006,
-	0xe0685000, 0x701d0200,
-	0x8070ff70, 0x00000080,
-	0xbef90080, 0x7e040280,
-	0x807d907d, 0xbf0aff7d,
-	0x00000060, 0xbfa2ffbc,
-	0xbe804100, 0xbe824102,
-	0xbe844104, 0xbe864106,
-	0xbe884108, 0xbe8a410a,
-	0xd7610002, 0x0000f200,
+	0x0000f20c, 0x80798179,
+	0xd7610002, 0x0000f20d,
 	0x80798179, 0xd7610002,
-	0x0000f201, 0x80798179,
-	0xd7610002, 0x0000f202,
+	0x0000f20e, 0x80798179,
+	0xd7610002, 0x0000f20f,
+	0x80798179, 0xbf06a079,
+	0xbfa10006, 0xe0685000,
+	0x701d0200, 0x8070ff70,
+	0x00000080, 0xbef90080,
+	0x7e040280, 0x807d907d,
+	0xbf0aff7d, 0x00000060,
+	0xbfa2ffbc, 0xbe804100,
+	0xbe824102, 0xbe844104,
+	0xbe864106, 0xbe884108,
+	0xbe8a410a, 0xd7610002,
+	0x0000f200, 0x80798179,
+	0xd7610002, 0x0000f201,
 	0x80798179, 0xd7610002,
-	0x0000f203, 0x80798179,
-	0xd7610002, 0x0000f204,
+	0x0000f202, 0x80798179,
+	0xd7610002, 0x0000f203,
 	0x80798179, 0xd7610002,
-	0x0000f205, 0x80798179,
-	0xd7610002, 0x0000f206,
+	0x0000f204, 0x80798179,
+	0xd7610002, 0x0000f205,
 	0x80798179, 0xd7610002,
-	0x0000f207, 0x80798179,
-	0xd7610002, 0x0000f208,
+	0x0000f206, 0x80798179,
+	0xd7610002, 0x0000f207,
 	0x80798179, 0xd7610002,
-	0x0000f209, 0x80798179,
-	0xd7610002, 0x0000f20a,
+	0x0000f208, 0x80798179,
+	0xd7610002, 0x0000f209,
 	0x80798179, 0xd7610002,
-	0x0000f20b, 0x80798179,
-	0xe0685000, 0x701d0200,
-	0xbefe00c1, 0x857d9973,
-	0x8b7d817d, 0xbf06817d,
-	0xbfa20002, 0xbeff0080,
-	0xbfa00001, 0xbeff00c1,
-	0xb8fb4306, 0x8b7bc17b,
-	0xbfa10044, 0xbfbd0000,
-	0x8b7aff6d, 0x80000000,
-	0xbfa10040, 0x847b867b,
-	0x847b827b, 0xbef6007b,
+	0x0000f20a, 0x80798179,
+	0xd7610002, 0x0000f20b,
+	0x80798179, 0xe0685000,
+	0x701d0200, 0xbefe00c1,
+	0x857d9973, 0x8b7d817d,
+	0xbf06817d, 0xbfa20002,
+	0xbeff0080, 0xbfa00001,
+	0xbeff00c1, 0xb8fb4306,
+	0x8b7bc17b, 0xbfa10043,
+	0xbfbd0000, 0x8b7aff6d,
+	0x80000000, 0xbfa1003f,
+	0x847b887b, 0xbef6007b,
 	0xb8f03b05, 0x80708170,
 	0xbf0d9973, 0xbfa20002,
 	0x84708970, 0xbfa00001,
@@ -2768,94 +2986,1099 @@ static const uint32_t cwsr_trap_gfx11_hex[] = {
 	0x701d0000, 0x807d817d,
 	0x8070ff70, 0x00000080,
 	0xbf0a7b7d, 0xbfa2fff8,
-	0xbfa00146, 0xbef4007e,
+	0xbfa00143, 0xbef4007e,
 	0x8b75ff7f, 0x0000ffff,
 	0x8c75ff75, 0x00040000,
 	0xbef60080, 0xbef700ff,
 	0x10807fac, 0xb8f202dc,
 	0x84729972, 0x8b6eff7f,
-	0x04000000, 0xbfa1003a,
+	0x04000000, 0xbfa10039,
 	0xbefe00c1, 0x857d9972,
 	0x8b7d817d, 0xbf06817d,
 	0xbfa20002, 0xbeff0080,
 	0xbfa00001, 0xbeff00c1,
 	0xb8ef4306, 0x8b6fc16f,
-	0xbfa1002f, 0x846f866f,
-	0x846f826f, 0xbef6006f,
+	0xbfa1002e, 0x846f886f,
+	0xbef6006f, 0xb8f83b05,
+	0x80788178, 0xbf0d9972,
+	0xbfa20002, 0x84788978,
+	0xbfa00001, 0x84788a78,
+	0xb8ee1e06, 0x846e8a6e,
+	0x80786e78, 0x8078ff78,
+	0x00000200, 0x8078ff78,
+	0x00000080, 0xbef600ff,
+	0x01000000, 0x857d9972,
+	0x8b7d817d, 0xbf06817d,
+	0xbefd0080, 0xbfa2000c,
+	0xe0500000, 0x781d0000,
+	0xbf890000, 0xdac00000,
+	0x00000000, 0x807dff7d,
+	0x00000080, 0x8078ff78,
+	0x00000080, 0xbf0a6f7d,
+	0xbfa2fff5, 0xbfa0000b,
+	0xe0500000, 0x781d0000,
+	0xbf890000, 0xdac00000,
+	0x00000000, 0x807dff7d,
+	0x00000100, 0x8078ff78,
+	0x00000100, 0xbf0a6f7d,
+	0xbfa2fff5, 0xbef80080,
+	0xbefe00c1, 0x857d9972,
+	0x8b7d817d, 0xbf06817d,
+	0xbfa20002, 0xbeff0080,
+	0xbfa00001, 0xbeff00c1,
+	0xb8ef3b05, 0x806f816f,
+	0x846f826f, 0x857d9972,
+	0x8b7d817d, 0xbf06817d,
+	0xbfa20024, 0xbef600ff,
+	0x01000000, 0xbeee0078,
+	0x8078ff78, 0x00000200,
+	0xbefd0084, 0xbf0a6f7d,
+	0xbfa10050, 0xe0505000,
+	0x781d0000, 0xe0505080,
+	0x781d0100, 0xe0505100,
+	0x781d0200, 0xe0505180,
+	0x781d0300, 0xbf890000,
+	0x7e008500, 0x7e028501,
+	0x7e048502, 0x7e068503,
+	0x807d847d, 0x8078ff78,
+	0x00000200, 0xbf0a6f7d,
+	0xbfa2ffee, 0xe0505000,
+	0x6e1d0000, 0xe0505080,
+	0x6e1d0100, 0xe0505100,
+	0x6e1d0200, 0xe0505180,
+	0x6e1d0300, 0xbf890000,
+	0xbfa00034, 0xbef600ff,
+	0x01000000, 0xbeee0078,
+	0x8078ff78, 0x00000400,
+	0xbefd0084, 0xbf0a6f7d,
+	0xbfa10012, 0xe0505000,
+	0x781d0000, 0xe0505100,
+	0x781d0100, 0xe0505200,
+	0x781d0200, 0xe0505300,
+	0x781d0300, 0xbf890000,
+	0x7e008500, 0x7e028501,
+	0x7e048502, 0x7e068503,
+	0x807d847d, 0x8078ff78,
+	0x00000400, 0xbf0a6f7d,
+	0xbfa2ffee, 0xb8ef1e06,
+	0x8b6fc16f, 0xbfa1000e,
+	0x846f836f, 0x806f7d6f,
+	0xbefe00c1, 0xbeff0080,
+	0xe0505000, 0x781d0000,
+	0xbf890000, 0x7e008500,
+	0x807d817d, 0x8078ff78,
+	0x00000080, 0xbf0a6f7d,
+	0xbfa2fff7, 0xbeff00c1,
+	0xe0505000, 0x6e1d0000,
+	0xe0505100, 0x6e1d0100,
+	0xe0505200, 0x6e1d0200,
+	0xe0505300, 0x6e1d0300,
+	0xbf890000, 0xb8f83b05,
+	0x80788178, 0xbf0d9972,
+	0xbfa20002, 0x84788978,
+	0xbfa00001, 0x84788a78,
+	0xb8ee1e06, 0x846e8a6e,
+	0x80786e78, 0x8078ff78,
+	0x00000200, 0x80f8ff78,
+	0x00000050, 0xbef600ff,
+	0x01000000, 0xbefd00ff,
+	0x0000006c, 0x80f89078,
+	0xf428403a, 0xf0000000,
+	0xbf890000, 0x80fd847d,
+	0xbf800000, 0xbe804300,
+	0xbe824302, 0x80f8a078,
+	0xf42c403a, 0xf0000000,
+	0xbf890000, 0x80fd887d,
+	0xbf800000, 0xbe804300,
+	0xbe824302, 0xbe844304,
+	0xbe864306, 0x80f8c078,
+	0xf430403a, 0xf0000000,
+	0xbf890000, 0x80fd907d,
+	0xbf800000, 0xbe804300,
+	0xbe824302, 0xbe844304,
+	0xbe864306, 0xbe884308,
+	0xbe8a430a, 0xbe8c430c,
+	0xbe8e430e, 0xbf06807d,
+	0xbfa1fff0, 0xb980f801,
+	0x00000000, 0xbfbd0000,
 	0xb8f83b05, 0x80788178,
 	0xbf0d9972, 0xbfa20002,
 	0x84788978, 0xbfa00001,
 	0x84788a78, 0xb8ee1e06,
 	0x846e8a6e, 0x80786e78,
 	0x8078ff78, 0x00000200,
-	0x8078ff78, 0x00000080,
 	0xbef600ff, 0x01000000,
-	0x857d9972, 0x8b7d817d,
-	0xbf06817d, 0xbefd0080,
-	0xbfa2000c, 0xe0500000,
-	0x781d0000, 0xbf8903f7,
-	0xdac00000, 0x00000000,
-	0x807dff7d, 0x00000080,
-	0x8078ff78, 0x00000080,
-	0xbf0a6f7d, 0xbfa2fff5,
-	0xbfa0000b, 0xe0500000,
-	0x781d0000, 0xbf8903f7,
-	0xdac00000, 0x00000000,
-	0x807dff7d, 0x00000100,
-	0x8078ff78, 0x00000100,
-	0xbf0a6f7d, 0xbfa2fff5,
-	0xbef80080, 0xbefe00c1,
-	0x857d9972, 0x8b7d817d,
-	0xbf06817d, 0xbfa20002,
-	0xbeff0080, 0xbfa00001,
-	0xbeff00c1, 0xb8ef3b05,
-	0x806f816f, 0x846f826f,
-	0x857d9972, 0x8b7d817d,
-	0xbf06817d, 0xbfa20024,
+	0xf4205bfa, 0xf0000000,
+	0x80788478, 0xf4205b3a,
+	0xf0000000, 0x80788478,
+	0xf4205b7a, 0xf0000000,
+	0x80788478, 0xf4205c3a,
+	0xf0000000, 0x80788478,
+	0xf4205c7a, 0xf0000000,
+	0x80788478, 0xf4205eba,
+	0xf0000000, 0x80788478,
+	0xf4205efa, 0xf0000000,
+	0x80788478, 0xf4205e7a,
+	0xf0000000, 0x80788478,
+	0xf4205cfa, 0xf0000000,
+	0x80788478, 0xf4205bba,
+	0xf0000000, 0x80788478,
+	0xbf890000, 0xb96ef814,
+	0xf4205bba, 0xf0000000,
+	0x80788478, 0xbf890000,
+	0xb96ef815, 0xbefd006f,
+	0xbefe0070, 0xbeff0071,
+	0xb97b4803, 0x857b8b7b,
+	0xb97b22c3, 0x857b867b,
+	0xb97b7443, 0xb973f801,
+	0xb8ee3b05, 0x806e816e,
+	0xbf0d9972, 0xbfa20002,
+	0x846e896e, 0xbfa00001,
+	0x846e8a6e, 0xb8ef1e06,
+	0x846f8a6f, 0x806e6f6e,
+	0x806eff6e, 0x00000200,
+	0x806e746e, 0x826f8075,
+	0x8b6fff6f, 0x0000ffff,
+	0xf4085c37, 0xf8000050,
+	0xf4085d37, 0xf8000060,
+	0xf4005e77, 0xf8000074,
+	0xbf890000, 0x8b6dff6d,
+	0x0000ffff, 0x8bfe7e7e,
+	0x8bea6a6a, 0xb8eef802,
+	0xbf0d866e, 0xbfa20002,
+	0xb97af802, 0xbe80486c,
+	0xb97af802, 0xbe804a6c,
+	0xbfb10000, 0xbf9f0000,
+	0xbf9f0000, 0xbf9f0000,
+	0xbf9f0000, 0xbf9f0000,
+};
+
+static const uint32_t cwsr_trap_gfx9_4_3_hex[] = {
+	0xbf820001, 0xbf8202dc,
+	0xb8f8f802, 0x8978ff78,
+	0x00020006, 0xb8fbf803,
+	0x866eff78, 0x00002000,
+	0xbf840009, 0x866eff6d,
+	0x00ff0000, 0xbf85001a,
+	0x866eff7b, 0x00000400,
+	0xbf850051, 0xbf8e0010,
+	0xb8fbf803, 0xbf82fffa,
+	0x866eff7b, 0x03c00900,
+	0xbf850011, 0x866eff7b,
+	0x000071ff, 0xbf840008,
+	0x866fff7b, 0x00007080,
+	0xbf840001, 0xbeee1a87,
+	0xb8eff801, 0x8e6e8c6e,
+	0x866e6f6e, 0xbf850006,
+	0x866eff6d, 0x00ff0000,
+	0xbf850003, 0x866eff7b,
+	0x00000400, 0xbf85003a,
+	0xb8faf807, 0x867aff7a,
+	0x001f8000, 0x8e7a8b7a,
+	0x8979ff79, 0xfc000000,
+	0x87797a79, 0xba7ff807,
+	0x00000000, 0xb8faf812,
+	0xb8fbf813, 0x8efa887a,
+	0xbf0d8f7b, 0xbf840002,
+	0x877bff7b, 0xffff0000,
+	0xc0031bbd, 0x00000010,
+	0xbf8cc07f, 0x8e6e976e,
+	0x8979ff79, 0x00800000,
+	0x87796e79, 0xc0071bbd,
+	0x00000000, 0xbf8cc07f,
+	0xc0071ebd, 0x00000008,
+	0xbf8cc07f, 0x86ee6e6e,
+	0xbf840001, 0xbe801d6e,
+	0x866eff6d, 0x01ff0000,
+	0xbf850005, 0x8778ff78,
+	0x00002000, 0x80ec886c,
+	0x82ed806d, 0xbf820005,
+	0x866eff6d, 0x01000000,
+	0xbf850002, 0x806c846c,
+	0x826d806d, 0x866dff6d,
+	0x0000ffff, 0x8f7a8b79,
+	0x867aff7a, 0x001f8000,
+	0xb97af807, 0x86fe7e7e,
+	0x86ea6a6a, 0x8f6e8378,
+	0xb96ee0c2, 0xbf800002,
+	0xb9780002, 0xbe801f6c,
+	0x866dff6d, 0x0000ffff,
+	0xbefa0080, 0xb97a0283,
+	0xb8faf807, 0x867aff7a,
+	0x001f8000, 0x8e7a8b7a,
+	0x8979ff79, 0xfc000000,
+	0x87797a79, 0xba7ff807,
+	0x00000000, 0xbeee007e,
+	0xbeef007f, 0xbefe0180,
+	0xbf900004, 0x877a8478,
+	0xb97af802, 0xbf8e0002,
+	0xbf88fffe, 0xb8fa2985,
+	0x807a817a, 0x8e7a8a7a,
+	0x8e7a817a, 0xb8fb1605,
+	0x807b817b, 0x8e7b867b,
+	0x807a7b7a, 0x807a7e7a,
+	0x827b807f, 0x867bff7b,
+	0x0000ffff, 0xc04b1c3d,
+	0x00000050, 0xbf8cc07f,
+	0xc04b1d3d, 0x00000060,
+	0xbf8cc07f, 0xc0431e7d,
+	0x00000074, 0xbf8cc07f,
+	0xbef4007e, 0x8675ff7f,
+	0x0000ffff, 0x8775ff75,
+	0x00040000, 0xbef60080,
+	0xbef700ff, 0x00807fac,
+	0xbef1007c, 0xbef00080,
+	0xb8f02985, 0x80708170,
+	0x8e708a70, 0x8e708170,
+	0xb8fa1605, 0x807a817a,
+	0x8e7a867a, 0x80707a70,
+	0xbef60084, 0xbef600ff,
+	0x01000000, 0xbefe007c,
+	0xbefc0070, 0xc0611c7a,
+	0x0000007c, 0xbf8cc07f,
+	0x80708470, 0xbefc007e,
+	0xbefe007c, 0xbefc0070,
+	0xc0611b3a, 0x0000007c,
+	0xbf8cc07f, 0x80708470,
+	0xbefc007e, 0xbefe007c,
+	0xbefc0070, 0xc0611b7a,
+	0x0000007c, 0xbf8cc07f,
+	0x80708470, 0xbefc007e,
+	0xbefe007c, 0xbefc0070,
+	0xc0611bba, 0x0000007c,
+	0xbf8cc07f, 0x80708470,
+	0xbefc007e, 0xbefe007c,
+	0xbefc0070, 0xc0611bfa,
+	0x0000007c, 0xbf8cc07f,
+	0x80708470, 0xbefc007e,
+	0xbefe007c, 0xbefc0070,
+	0xc0611e3a, 0x0000007c,
+	0xbf8cc07f, 0x80708470,
+	0xbefc007e, 0xb8fbf803,
+	0xbefe007c, 0xbefc0070,
+	0xc0611efa, 0x0000007c,
+	0xbf8cc07f, 0x80708470,
+	0xbefc007e, 0xbefe007c,
+	0xbefc0070, 0xc0611a3a,
+	0x0000007c, 0xbf8cc07f,
+	0x80708470, 0xbefc007e,
+	0xbefe007c, 0xbefc0070,
+	0xc0611a7a, 0x0000007c,
+	0xbf8cc07f, 0x80708470,
+	0xbefc007e, 0xb8f1f801,
+	0xbefe007c, 0xbefc0070,
+	0xc0611c7a, 0x0000007c,
+	0xbf8cc07f, 0x80708470,
+	0xbefc007e, 0xbf108080,
+	0x867aff7f, 0x04000000,
+	0xbeef0080, 0x876f6f7a,
+	0xb8f02985, 0x80708170,
+	0x8e708a70, 0x8e708170,
+	0xb8fb1605, 0x807b817b,
+	0x8e7b847b, 0x8e76827b,
 	0xbef600ff, 0x01000000,
-	0xbeee0078, 0x8078ff78,
-	0x00000200, 0xbefd0084,
-	0xbf0a6f7d, 0xbfa10050,
-	0xe0505000, 0x781d0000,
-	0xe0505080, 0x781d0100,
-	0xe0505100, 0x781d0200,
-	0xe0505180, 0x781d0300,
-	0xbf8903f7, 0x7e008500,
-	0x7e028501, 0x7e048502,
-	0x7e068503, 0x807d847d,
+	0xbef20174, 0x80747074,
+	0x82758075, 0xbefc0080,
+	0xbf800000, 0xbe802b00,
+	0xbe822b02, 0xbe842b04,
+	0xbe862b06, 0xbe882b08,
+	0xbe8a2b0a, 0xbe8c2b0c,
+	0xbe8e2b0e, 0xc06b003a,
+	0x00000000, 0xbf8cc07f,
+	0xc06b013a, 0x00000010,
+	0xbf8cc07f, 0xc06b023a,
+	0x00000020, 0xbf8cc07f,
+	0xc06b033a, 0x00000030,
+	0xbf8cc07f, 0x8074c074,
+	0x82758075, 0x807c907c,
+	0xbf0a7b7c, 0xbf85ffe7,
+	0xbef40172, 0xbef00080,
+	0xbefe00c1, 0xbeff00c1,
+	0xbee80080, 0xbee90080,
+	0xbef600ff, 0x01000000,
+	0x867aff78, 0x00400000,
+	0xbf850003, 0xb8faf803,
+	0x897a7aff, 0x10000000,
+	0xbf85004d, 0xbe840080,
+	0xd2890000, 0x00000900,
+	0x80048104, 0xd2890001,
+	0x00000900, 0x80048104,
+	0xd2890002, 0x00000900,
+	0x80048104, 0xd2890003,
+	0x00000900, 0x80048104,
+	0xc069003a, 0x00000070,
+	0xbf8cc07f, 0x80709070,
+	0xbf06c004, 0xbf84ffee,
+	0xbe840080, 0xd2890000,
+	0x00000901, 0x80048104,
+	0xd2890001, 0x00000901,
+	0x80048104, 0xd2890002,
+	0x00000901, 0x80048104,
+	0xd2890003, 0x00000901,
+	0x80048104, 0xc069003a,
+	0x00000070, 0xbf8cc07f,
+	0x80709070, 0xbf06c004,
+	0xbf84ffee, 0xbe840080,
+	0xd2890000, 0x00000902,
+	0x80048104, 0xd2890001,
+	0x00000902, 0x80048104,
+	0xd2890002, 0x00000902,
+	0x80048104, 0xd2890003,
+	0x00000902, 0x80048104,
+	0xc069003a, 0x00000070,
+	0xbf8cc07f, 0x80709070,
+	0xbf06c004, 0xbf84ffee,
+	0xbe840080, 0xd2890000,
+	0x00000903, 0x80048104,
+	0xd2890001, 0x00000903,
+	0x80048104, 0xd2890002,
+	0x00000903, 0x80048104,
+	0xd2890003, 0x00000903,
+	0x80048104, 0xc069003a,
+	0x00000070, 0xbf8cc07f,
+	0x80709070, 0xbf06c004,
+	0xbf84ffee, 0xbf820008,
+	0xe0724000, 0x701d0000,
+	0xe0724100, 0x701d0100,
+	0xe0724200, 0x701d0200,
+	0xe0724300, 0x701d0300,
+	0xbefe00c1, 0xbeff00c1,
+	0xb8fb4306, 0x867bc17b,
+	0xbf840064, 0xbf8a0000,
+	0x867aff6f, 0x04000000,
+	0xbf840060, 0x8e7b867b,
+	0x8e7b827b, 0xbef6007b,
+	0xb8f02985, 0x80708170,
+	0x8e708a70, 0x8e708170,
+	0xb8fa1605, 0x807a817a,
+	0x8e7a867a, 0x80707a70,
+	0x8070ff70, 0x00000080,
+	0xbef600ff, 0x01000000,
+	0xbefc0080, 0xd28c0002,
+	0x000100c1, 0xd28d0003,
+	0x000204c1, 0x867aff78,
+	0x00400000, 0xbf850003,
+	0xb8faf803, 0x897a7aff,
+	0x10000000, 0xbf850030,
+	0x24040682, 0xd86e4000,
+	0x00000002, 0xbf8cc07f,
+	0xbe840080, 0xd2890000,
+	0x00000900, 0x80048104,
+	0xd2890001, 0x00000900,
+	0x80048104, 0xd2890002,
+	0x00000900, 0x80048104,
+	0xd2890003, 0x00000900,
+	0x80048104, 0xc069003a,
+	0x00000070, 0xbf8cc07f,
+	0x80709070, 0xbf06c004,
+	0xbf84ffee, 0xbe840080,
+	0xd2890000, 0x00000901,
+	0x80048104, 0xd2890001,
+	0x00000901, 0x80048104,
+	0xd2890002, 0x00000901,
+	0x80048104, 0xd2890003,
+	0x00000901, 0x80048104,
+	0xc069003a, 0x00000070,
+	0xbf8cc07f, 0x80709070,
+	0xbf06c004, 0xbf84ffee,
+	0x680404ff, 0x00000200,
+	0xd0c9006a, 0x0000f702,
+	0xbf87ffd2, 0xbf820015,
+	0xd1060002, 0x00011103,
+	0x7e0602ff, 0x00000200,
+	0xbefc00ff, 0x00010000,
+	0xbe800077, 0x8677ff77,
+	0xff7fffff, 0x8777ff77,
+	0x00058000, 0xd8ec0000,
+	0x00000002, 0xbf8cc07f,
+	0xe0765000, 0x701d0002,
+	0x68040702, 0xd0c9006a,
+	0x0000f702, 0xbf87fff7,
+	0xbef70000, 0xbef000ff,
+	0x00000400, 0xbefe00c1,
+	0xbeff00c1, 0xb8fb2b05,
+	0x807b817b, 0x8e7b827b,
+	0xbef600ff, 0x01000000,
+	0xbefc0084, 0xbf0a7b7c,
+	0xbf84006d, 0xbf11017c,
+	0x807bff7b, 0x00001000,
+	0x867aff78, 0x00400000,
+	0xbf850003, 0xb8faf803,
+	0x897a7aff, 0x10000000,
+	0xbf850051, 0xbe840080,
+	0xd2890000, 0x00000900,
+	0x80048104, 0xd2890001,
+	0x00000900, 0x80048104,
+	0xd2890002, 0x00000900,
+	0x80048104, 0xd2890003,
+	0x00000900, 0x80048104,
+	0xc069003a, 0x00000070,
+	0xbf8cc07f, 0x80709070,
+	0xbf06c004, 0xbf84ffee,
+	0xbe840080, 0xd2890000,
+	0x00000901, 0x80048104,
+	0xd2890001, 0x00000901,
+	0x80048104, 0xd2890002,
+	0x00000901, 0x80048104,
+	0xd2890003, 0x00000901,
+	0x80048104, 0xc069003a,
+	0x00000070, 0xbf8cc07f,
+	0x80709070, 0xbf06c004,
+	0xbf84ffee, 0xbe840080,
+	0xd2890000, 0x00000902,
+	0x80048104, 0xd2890001,
+	0x00000902, 0x80048104,
+	0xd2890002, 0x00000902,
+	0x80048104, 0xd2890003,
+	0x00000902, 0x80048104,
+	0xc069003a, 0x00000070,
+	0xbf8cc07f, 0x80709070,
+	0xbf06c004, 0xbf84ffee,
+	0xbe840080, 0xd2890000,
+	0x00000903, 0x80048104,
+	0xd2890001, 0x00000903,
+	0x80048104, 0xd2890002,
+	0x00000903, 0x80048104,
+	0xd2890003, 0x00000903,
+	0x80048104, 0xc069003a,
+	0x00000070, 0xbf8cc07f,
+	0x80709070, 0xbf06c004,
+	0xbf84ffee, 0x807c847c,
+	0xbf0a7b7c, 0xbf85ffb1,
+	0xbf9c0000, 0xbf820012,
+	0x7e000300, 0x7e020301,
+	0x7e040302, 0x7e060303,
+	0xe0724000, 0x701d0000,
+	0xe0724100, 0x701d0100,
+	0xe0724200, 0x701d0200,
+	0xe0724300, 0x701d0300,
+	0x807c847c, 0x8070ff70,
+	0x00000400, 0xbf0a7b7c,
+	0xbf85ffef, 0xbf9c0000,
+	0xb8fb2985, 0x807b817b,
+	0x8e7b837b, 0xb8fa2b05,
+	0x807a817a, 0x8e7a827a,
+	0x80fb7a7b, 0x867b7b7b,
+	0xbf84007a, 0x807bff7b,
+	0x00001000, 0xbefc0080,
+	0xbf11017c, 0x867aff78,
+	0x00400000, 0xbf850003,
+	0xb8faf803, 0x897a7aff,
+	0x10000000, 0xbf850059,
+	0xd3d84000, 0x18000100,
+	0xd3d84001, 0x18000101,
+	0xd3d84002, 0x18000102,
+	0xd3d84003, 0x18000103,
+	0xbe840080, 0xd2890000,
+	0x00000900, 0x80048104,
+	0xd2890001, 0x00000900,
+	0x80048104, 0xd2890002,
+	0x00000900, 0x80048104,
+	0xd2890003, 0x00000900,
+	0x80048104, 0xc069003a,
+	0x00000070, 0xbf8cc07f,
+	0x80709070, 0xbf06c004,
+	0xbf84ffee, 0xbe840080,
+	0xd2890000, 0x00000901,
+	0x80048104, 0xd2890001,
+	0x00000901, 0x80048104,
+	0xd2890002, 0x00000901,
+	0x80048104, 0xd2890003,
+	0x00000901, 0x80048104,
+	0xc069003a, 0x00000070,
+	0xbf8cc07f, 0x80709070,
+	0xbf06c004, 0xbf84ffee,
+	0xbe840080, 0xd2890000,
+	0x00000902, 0x80048104,
+	0xd2890001, 0x00000902,
+	0x80048104, 0xd2890002,
+	0x00000902, 0x80048104,
+	0xd2890003, 0x00000902,
+	0x80048104, 0xc069003a,
+	0x00000070, 0xbf8cc07f,
+	0x80709070, 0xbf06c004,
+	0xbf84ffee, 0xbe840080,
+	0xd2890000, 0x00000903,
+	0x80048104, 0xd2890001,
+	0x00000903, 0x80048104,
+	0xd2890002, 0x00000903,
+	0x80048104, 0xd2890003,
+	0x00000903, 0x80048104,
+	0xc069003a, 0x00000070,
+	0xbf8cc07f, 0x80709070,
+	0xbf06c004, 0xbf84ffee,
+	0x807c847c, 0xbf0a7b7c,
+	0xbf85ffa9, 0xbf9c0000,
+	0xbf820016, 0xd3d84000,
+	0x18000100, 0xd3d84001,
+	0x18000101, 0xd3d84002,
+	0x18000102, 0xd3d84003,
+	0x18000103, 0xe0724000,
+	0x701d0000, 0xe0724100,
+	0x701d0100, 0xe0724200,
+	0x701d0200, 0xe0724300,
+	0x701d0300, 0x807c847c,
+	0x8070ff70, 0x00000400,
+	0xbf0a7b7c, 0xbf85ffeb,
+	0xbf9c0000, 0xbf8200ee,
+	0xbef4007e, 0x8675ff7f,
+	0x0000ffff, 0x8775ff75,
+	0x00040000, 0xbef60080,
+	0xbef700ff, 0x00807fac,
+	0x866eff7f, 0x04000000,
+	0xbf84001f, 0xbefe00c1,
+	0xbeff00c1, 0xb8ef4306,
+	0x866fc16f, 0xbf84001a,
+	0x8e6f866f, 0x8e6f826f,
+	0xbef6006f, 0xb8f82985,
+	0x80788178, 0x8e788a78,
+	0x8e788178, 0xb8ee1605,
+	0x806e816e, 0x8e6e866e,
+	0x80786e78, 0x8078ff78,
+	0x00000080, 0xbef600ff,
+	0x01000000, 0xbefc0080,
+	0xe0510000, 0x781d0000,
+	0xe0510100, 0x781d0000,
+	0x807cff7c, 0x00000200,
 	0x8078ff78, 0x00000200,
-	0xbf0a6f7d, 0xbfa2ffee,
-	0xe0505000, 0x6e1d0000,
-	0xe0505080, 0x6e1d0100,
-	0xe0505100, 0x6e1d0200,
-	0xe0505180, 0x6e1d0300,
-	0xbf8903f7, 0xbfa00034,
+	0xbf0a6f7c, 0xbf85fff6,
+	0xbefe00c1, 0xbeff00c1,
 	0xbef600ff, 0x01000000,
+	0xb8ef2b05, 0x806f816f,
+	0x8e6f826f, 0x806fff6f,
+	0x00008000, 0xbef80080,
 	0xbeee0078, 0x8078ff78,
-	0x00000400, 0xbefd0084,
-	0xbf0a6f7d, 0xbfa10012,
-	0xe0505000, 0x781d0000,
-	0xe0505100, 0x781d0100,
-	0xe0505200, 0x781d0200,
-	0xe0505300, 0x781d0300,
-	0xbf8903f7, 0x7e008500,
-	0x7e028501, 0x7e048502,
-	0x7e068503, 0x807d847d,
+	0x00000400, 0xbefc0084,
+	0xbf11087c, 0xe0524000,
+	0x781d0000, 0xe0524100,
+	0x781d0100, 0xe0524200,
+	0x781d0200, 0xe0524300,
+	0x781d0300, 0xbf8c0f70,
+	0x7e000300, 0x7e020301,
+	0x7e040302, 0x7e060303,
+	0x807c847c, 0x8078ff78,
+	0x00000400, 0xbf0a6f7c,
+	0xbf85ffee, 0xb8ef2985,
+	0x806f816f, 0x8e6f836f,
+	0xb8f92b05, 0x80798179,
+	0x8e798279, 0x80ef796f,
+	0x866f6f6f, 0xbf84001a,
+	0x806fff6f, 0x00008000,
+	0xbefc0080, 0xbf11087c,
+	0xe0524000, 0x781d0000,
+	0xe0524100, 0x781d0100,
+	0xe0524200, 0x781d0200,
+	0xe0524300, 0x781d0300,
+	0xbf8c0f70, 0xd3d94000,
+	0x18000100, 0xd3d94001,
+	0x18000101, 0xd3d94002,
+	0x18000102, 0xd3d94003,
+	0x18000103, 0x807c847c,
 	0x8078ff78, 0x00000400,
-	0xbf0a6f7d, 0xbfa2ffee,
-	0xb8ef1e06, 0x8b6fc16f,
-	0xbfa1000e, 0x846f836f,
-	0x806f7d6f, 0xbefe00c1,
-	0xbeff0080, 0xe0505000,
-	0x781d0000, 0xbf8903f7,
+	0xbf0a6f7c, 0xbf85ffea,
+	0xbf9c0000, 0xe0524000,
+	0x6e1d0000, 0xe0524100,
+	0x6e1d0100, 0xe0524200,
+	0x6e1d0200, 0xe0524300,
+	0x6e1d0300, 0xbf8c0f70,
+	0xb8f82985, 0x80788178,
+	0x8e788a78, 0x8e788178,
+	0xb8ee1605, 0x806e816e,
+	0x8e6e866e, 0x80786e78,
+	0x80f8c078, 0xb8ef1605,
+	0x806f816f, 0x8e6f846f,
+	0x8e76826f, 0xbef600ff,
+	0x01000000, 0xbefc006f,
+	0xc031003a, 0x00000078,
+	0x80f8c078, 0xbf8cc07f,
+	0x80fc907c, 0xbf800000,
+	0xbe802d00, 0xbe822d02,
+	0xbe842d04, 0xbe862d06,
+	0xbe882d08, 0xbe8a2d0a,
+	0xbe8c2d0c, 0xbe8e2d0e,
+	0xbf06807c, 0xbf84fff0,
+	0xb8f82985, 0x80788178,
+	0x8e788a78, 0x8e788178,
+	0xb8ee1605, 0x806e816e,
+	0x8e6e866e, 0x80786e78,
+	0xbef60084, 0xbef600ff,
+	0x01000000, 0xc0211bfa,
+	0x00000078, 0x80788478,
+	0xc0211b3a, 0x00000078,
+	0x80788478, 0xc0211b7a,
+	0x00000078, 0x80788478,
+	0xc0211c3a, 0x00000078,
+	0x80788478, 0xc0211c7a,
+	0x00000078, 0x80788478,
+	0xc0211eba, 0x00000078,
+	0x80788478, 0xc0211efa,
+	0x00000078, 0x80788478,
+	0xc0211a3a, 0x00000078,
+	0x80788478, 0xc0211a7a,
+	0x00000078, 0x80788478,
+	0xc0211cfa, 0x00000078,
+	0x80788478, 0xbf8cc07f,
+	0xbefc006f, 0xbefe0070,
+	0xbeff0071, 0x866f7bff,
+	0x000003ff, 0xb96f4803,
+	0x866f7bff, 0xfffff800,
+	0x8f6f8b6f, 0xb96fa2c3,
+	0xb973f801, 0xb8ee2985,
+	0x806e816e, 0x8e6e8a6e,
+	0x8e6e816e, 0xb8ef1605,
+	0x806f816f, 0x8e6f866f,
+	0x806e6f6e, 0x806e746e,
+	0x826f8075, 0x866fff6f,
+	0x0000ffff, 0xc00b1c37,
+	0x00000050, 0xc00b1d37,
+	0x00000060, 0xc0031e77,
+	0x00000074, 0xbf8cc07f,
+	0x8f6e8b79, 0x866eff6e,
+	0x001f8000, 0xb96ef807,
+	0x866dff6d, 0x0000ffff,
+	0x86fe7e7e, 0x86ea6a6a,
+	0x8f6e837a, 0xb96ee0c2,
+	0xbf800002, 0xb97a0002,
+	0xbf8a0000, 0xbe801f6c,
+	0xbf9b0000, 0x00000000,
+};
+
+static const uint32_t cwsr_trap_gfx12_hex[] = {
+	0xbfa00001, 0xbfa002a2,
+	0xb0804009, 0xb8f8f804,
+	0x9178ff78, 0x00008c00,
+	0xb8fbf811, 0x8b6eff78,
+	0x00004000, 0xbfa10008,
+	0x8b6eff7b, 0x00000080,
+	0xbfa20018, 0x8b6ea07b,
+	0xbfa20042, 0xbf830010,
+	0xb8fbf811, 0xbfa0fffb,
+	0x8b6eff7b, 0x00000bd0,
+	0xbfa20010, 0xb8eef812,
+	0x8b6f8f7b, 0xbfa10002,
+	0x8c6eff6e, 0x00000080,
+	0xb8eff813, 0x8b6e6e6f,
+	0xbfa20008, 0x8b6eff6d,
+	0xf0000000, 0xbfa20005,
+	0x8b6fff6f, 0x00000200,
+	0xbfa20002, 0x8b6ea07b,
+	0xbfa2002c, 0xbefa4d82,
+	0xbf8a0000, 0x84fa887a,
+	0xbf0d8f7b, 0xbfa10002,
+	0x8c7bff7b, 0xffff0000,
+	0xf4601bbd, 0xf8000010,
+	0xbf8a0000, 0x846e976e,
+	0x9177ff77, 0x00800000,
+	0x8c776e77, 0xf4603bbd,
+	0xf8000000, 0xbf8a0000,
+	0xf4603ebd, 0xf8000008,
+	0xbf8a0000, 0x8bee6e6e,
+	0xbfa10001, 0xbe80486e,
+	0x8b6eff6d, 0xf0000000,
+	0xbfa20009, 0xb8eef811,
+	0x8b6eff6e, 0x00000080,
+	0xbfa20007, 0x8c78ff78,
+	0x00004000, 0x80ec886c,
+	0x82ed806d, 0xbfa00002,
+	0x806c846c, 0x826d806d,
+	0x8b6dff6d, 0x0000ffff,
+	0x8bfe7e7e, 0x8bea6a6a,
+	0x85788978, 0xb9783244,
+	0xbe804a6c, 0xb8faf802,
+	0xbf0d987a, 0xbfa10001,
+	0xbfb00000, 0x8b6dff6d,
+	0x0000ffff, 0xbefa0080,
+	0xb97a0151, 0xbeee007e,
+	0xbeef007f, 0xbefe0180,
+	0xbefe4d84, 0xbf8a0000,
+	0x8b7aff7f, 0x04000000,
+	0x847a857a, 0x8c6d7a6d,
+	0xbefa007e, 0x8b7bff7f,
+	0x0000ffff, 0xbefe00c1,
+	0xbeff00c1, 0xee0a407a,
+	0x000c0000, 0x00000000,
+	0x7e000280, 0xbefe007a,
+	0xbeff007b, 0xb8fb0742,
+	0x847b997b, 0xb8fa3b05,
+	0x807a817a, 0xbf0d997b,
+	0xbfa20002, 0x847a897a,
+	0xbfa00001, 0x847a8a7a,
+	0xb8fb1e06, 0x847b8a7b,
+	0x807a7b7a, 0x8b7bff7f,
+	0x0000ffff, 0x807aff7a,
+	0x00000200, 0x807a7e7a,
+	0x827b807b, 0xd7610000,
+	0x00010870, 0xd7610000,
+	0x00010a71, 0xd7610000,
+	0x00010c72, 0xd7610000,
+	0x00010e73, 0xd7610000,
+	0x00011074, 0xd7610000,
+	0x00011275, 0xd7610000,
+	0x00011476, 0xd7610000,
+	0x00011677, 0xd7610000,
+	0x00011a79, 0xd7610000,
+	0x00011c7e, 0xd7610000,
+	0x00011e7f, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xbefe00ff,
+	0x00003fff, 0xbeff0080,
+	0xee0a407a, 0x000c0000,
+	0x00004000, 0xd760007a,
+	0x00011d00, 0xd760007b,
+	0x00011f00, 0xbefe007a,
+	0xbeff007b, 0xbef4007e,
+	0x8b75ff7f, 0x0000ffff,
+	0x8c75ff75, 0x00040000,
+	0xbef60080, 0xbef700ff,
+	0x10807fac, 0xbef1007d,
+	0xbef00080, 0xb8f30742,
+	0x84739973, 0xbefe00c1,
+	0x857d9973, 0x8b7d817d,
+	0xbf06817d, 0xbfa20002,
+	0xbeff0080, 0xbfa00002,
+	0xbeff00c1, 0xbfa0000c,
+	0xbef600ff, 0x01000000,
+	0xc4068070, 0x008ce801,
+	0x00008000, 0xc4068070,
+	0x008ce802, 0x00010000,
+	0xc4068070, 0x008ce803,
+	0x00018000, 0xbfa0000b,
+	0xbef600ff, 0x01000000,
+	0xc4068070, 0x008ce801,
+	0x00010000, 0xc4068070,
+	0x008ce802, 0x00020000,
+	0xc4068070, 0x008ce803,
+	0x00030000, 0xb8f03b05,
+	0x80708170, 0xbf0d9973,
+	0xbfa20002, 0x84708970,
+	0xbfa00001, 0x84708a70,
+	0xb8fa1e06, 0x847a8a7a,
+	0x80707a70, 0x8070ff70,
+	0x00000200, 0xbef600ff,
+	0x01000000, 0x7e000280,
+	0x7e020280, 0x7e040280,
+	0xbe804ec2, 0xbf94fffe,
+	0xb8faf804, 0x8b7a847a,
+	0x91788478, 0x8c787a78,
+	0x917aff6d, 0x80000000,
+	0xd7610002, 0x00010071,
+	0xd7610002, 0x0001026c,
+	0xd7610002, 0x0001047a,
+	0xd7610002, 0x0001066e,
+	0xd7610002, 0x0001086f,
+	0xd7610002, 0x00010a78,
+	0xd7610002, 0x00010e7b,
+	0xd8500000, 0x00000000,
+	0xd8500000, 0x00000000,
+	0xd8500000, 0x00000000,
+	0xd8500000, 0x00000000,
+	0xd8500000, 0x00000000,
+	0xd8500000, 0x00000000,
+	0xd8500000, 0x00000000,
+	0xd8500000, 0x00000000,
+	0xb8faf811, 0xd7610002,
+	0x00010c7a, 0xb8faf801,
+	0xd7610002, 0x0001107a,
+	0xb8faf814, 0xd7610002,
+	0x0001127a, 0xb8faf815,
+	0xd7610002, 0x0001147a,
+	0xb8faf812, 0xd7610002,
+	0x0001167a, 0xb8faf813,
+	0xd7610002, 0x0001187a,
+	0xb8faf802, 0xd7610002,
+	0x00011a7a, 0xbefa50c1,
+	0xbfc70000, 0xd7610002,
+	0x00011c7a, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xbefe00ff,
+	0x0000ffff, 0xbeff0080,
+	0xc4068070, 0x008ce802,
+	0x00000000, 0xbefe00c1,
+	0xb8f03b05, 0x80708170,
+	0xbf0d9973, 0xbfa20002,
+	0x84708970, 0xbfa00001,
+	0x84708a70, 0xb8fa1e06,
+	0x847a8a7a, 0x80707a70,
+	0xbef600ff, 0x01000000,
+	0xbef90080, 0xbefd0080,
+	0xbf800000, 0xbe804100,
+	0xbe824102, 0xbe844104,
+	0xbe864106, 0xbe884108,
+	0xbe8a410a, 0xbe8c410c,
+	0xbe8e410e, 0xbf068079,
+	0xbfa10032, 0xd7610002,
+	0x00010000, 0xd7610002,
+	0x00010201, 0xd7610002,
+	0x00010402, 0xd7610002,
+	0x00010603, 0xd7610002,
+	0x00010804, 0xd7610002,
+	0x00010a05, 0xd7610002,
+	0x00010c06, 0xd7610002,
+	0x00010e07, 0xd7610002,
+	0x00011008, 0xd7610002,
+	0x00011209, 0xd7610002,
+	0x0001140a, 0xd7610002,
+	0x0001160b, 0xd7610002,
+	0x0001180c, 0xd7610002,
+	0x00011a0d, 0xd7610002,
+	0x00011c0e, 0xd7610002,
+	0x00011e0f, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0x80799079,
+	0xbfa00038, 0xd7610002,
+	0x00012000, 0xd7610002,
+	0x00012201, 0xd7610002,
+	0x00012402, 0xd7610002,
+	0x00012603, 0xd7610002,
+	0x00012804, 0xd7610002,
+	0x00012a05, 0xd7610002,
+	0x00012c06, 0xd7610002,
+	0x00012e07, 0xd7610002,
+	0x00013008, 0xd7610002,
+	0x00013209, 0xd7610002,
+	0x0001340a, 0xd7610002,
+	0x0001360b, 0xd7610002,
+	0x0001380c, 0xd7610002,
+	0x00013a0d, 0xd7610002,
+	0x00013c0e, 0xd7610002,
+	0x00013e0f, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0x80799079,
+	0xc4068070, 0x008ce802,
+	0x00000000, 0x8070ff70,
+	0x00000080, 0xbef90080,
+	0x7e040280, 0x807d907d,
+	0xbf0aff7d, 0x00000060,
+	0xbfa2ff88, 0xbe804100,
+	0xbe824102, 0xbe844104,
+	0xbe864106, 0xbe884108,
+	0xbe8a410a, 0xd7610002,
+	0x00010000, 0xd7610002,
+	0x00010201, 0xd7610002,
+	0x00010402, 0xd7610002,
+	0x00010603, 0xd7610002,
+	0x00010804, 0xd7610002,
+	0x00010a05, 0xd7610002,
+	0x00010c06, 0xd7610002,
+	0x00010e07, 0xd7610002,
+	0x00011008, 0xd7610002,
+	0x00011209, 0xd7610002,
+	0x0001140a, 0xd7610002,
+	0x0001160b, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xd8500000,
+	0x00000000, 0xc4068070,
+	0x008ce802, 0x00000000,
+	0xbefe00c1, 0x857d9973,
+	0x8b7d817d, 0xbf06817d,
+	0xbfa20002, 0xbeff0080,
+	0xbfa00001, 0xbeff00c1,
+	0xb8fb4306, 0x8b7bc17b,
+	0xbfa10044, 0x8b7aff6d,
+	0x80000000, 0xbfa10041,
+	0x847b897b, 0xbef6007b,
+	0xb8f03b05, 0x80708170,
+	0xbf0d9973, 0xbfa20002,
+	0x84708970, 0xbfa00001,
+	0x84708a70, 0xb8fa1e06,
+	0x847a8a7a, 0x80707a70,
+	0x8070ff70, 0x00000200,
+	0x8070ff70, 0x00000080,
+	0xbef600ff, 0x01000000,
+	0xd71f0000, 0x000100c1,
+	0xd7200000, 0x000200c1,
+	0x16000084, 0x857d9973,
+	0x8b7d817d, 0xbf06817d,
+	0xbefd0080, 0xbfa20013,
+	0xbe8300ff, 0x00000080,
+	0xbf800000, 0xbf800000,
+	0xbf800000, 0xd8d80000,
+	0x01000000, 0xbf8a0000,
+	0xc4068070, 0x008ce801,
+	0x00000000, 0x807d037d,
+	0x80700370, 0xd5250000,
+	0x0001ff00, 0x00000080,
+	0xbf0a7b7d, 0xbfa2fff3,
+	0xbfa00012, 0xbe8300ff,
+	0x00000100, 0xbf800000,
+	0xbf800000, 0xbf800000,
+	0xd8d80000, 0x01000000,
+	0xbf8a0000, 0xc4068070,
+	0x008ce801, 0x00000000,
+	0x807d037d, 0x80700370,
+	0xd5250000, 0x0001ff00,
+	0x00000100, 0xbf0a7b7d,
+	0xbfa2fff3, 0xbefe00c1,
+	0x857d9973, 0x8b7d817d,
+	0xbf06817d, 0xbfa20004,
+	0xbef000ff, 0x00000200,
+	0xbeff0080, 0xbfa00003,
+	0xbef000ff, 0x00000400,
+	0xbeff00c1, 0xb8fb3b05,
+	0x807b817b, 0x847b827b,
+	0x857d9973, 0x8b7d817d,
+	0xbf06817d, 0xbfa2001b,
+	0xbef600ff, 0x01000000,
+	0xbefd0084, 0xbf0a7b7d,
+	0xbfa10040, 0x7e008700,
+	0x7e028701, 0x7e048702,
+	0x7e068703, 0xc4068070,
+	0x008ce800, 0x00000000,
+	0xc4068070, 0x008ce801,
+	0x00008000, 0xc4068070,
+	0x008ce802, 0x00010000,
+	0xc4068070, 0x008ce803,
+	0x00018000, 0x807d847d,
+	0x8070ff70, 0x00000200,
+	0xbf0a7b7d, 0xbfa2ffeb,
+	0xbfa0002a, 0xbef600ff,
+	0x01000000, 0xbefd0084,
+	0xbf0a7b7d, 0xbfa10015,
+	0x7e008700, 0x7e028701,
+	0x7e048702, 0x7e068703,
+	0xc4068070, 0x008ce800,
+	0x00000000, 0xc4068070,
+	0x008ce801, 0x00010000,
+	0xc4068070, 0x008ce802,
+	0x00020000, 0xc4068070,
+	0x008ce803, 0x00030000,
+	0x807d847d, 0x8070ff70,
+	0x00000400, 0xbf0a7b7d,
+	0xbfa2ffeb, 0xb8fb1e06,
+	0x8b7bc17b, 0xbfa1000d,
+	0x847b837b, 0x807b7d7b,
+	0xbefe00c1, 0xbeff0080,
+	0x7e008700, 0xc4068070,
+	0x008ce800, 0x00000000,
+	0x807d817d, 0x8070ff70,
+	0x00000080, 0xbf0a7b7d,
+	0xbfa2fff7, 0xbfa0016e,
+	0xbef4007e, 0x8b75ff7f,
+	0x0000ffff, 0x8c75ff75,
+	0x00040000, 0xbef60080,
+	0xbef700ff, 0x10807fac,
+	0xbef1007f, 0xb8f20742,
+	0x84729972, 0x8b6eff7f,
+	0x04000000, 0xbfa1003b,
+	0xbefe00c1, 0x857d9972,
+	0x8b7d817d, 0xbf06817d,
+	0xbfa20002, 0xbeff0080,
+	0xbfa00001, 0xbeff00c1,
+	0xb8ef4306, 0x8b6fc16f,
+	0xbfa10030, 0x846f896f,
+	0xbef6006f, 0xb8f83b05,
+	0x80788178, 0xbf0d9972,
+	0xbfa20002, 0x84788978,
+	0xbfa00001, 0x84788a78,
+	0xb8ee1e06, 0x846e8a6e,
+	0x80786e78, 0x8078ff78,
+	0x00000200, 0x8078ff78,
+	0x00000080, 0xbef600ff,
+	0x01000000, 0x857d9972,
+	0x8b7d817d, 0xbf06817d,
+	0xbefd0080, 0xbfa2000d,
+	0xc4050078, 0x0080e800,
+	0x00000000, 0xbf8a0000,
+	0xdac00000, 0x00000000,
+	0x807dff7d, 0x00000080,
+	0x8078ff78, 0x00000080,
+	0xbf0a6f7d, 0xbfa2fff4,
+	0xbfa0000c, 0xc4050078,
+	0x0080e800, 0x00000000,
+	0xbf8a0000, 0xdac00000,
+	0x00000000, 0x807dff7d,
+	0x00000100, 0x8078ff78,
+	0x00000100, 0xbf0a6f7d,
+	0xbfa2fff4, 0xbef80080,
+	0xbefe00c1, 0x857d9972,
+	0x8b7d817d, 0xbf06817d,
+	0xbfa20002, 0xbeff0080,
+	0xbfa00001, 0xbeff00c1,
+	0xb8ef3b05, 0x806f816f,
+	0x846f826f, 0x857d9972,
+	0x8b7d817d, 0xbf06817d,
+	0xbfa2002c, 0xbef600ff,
+	0x01000000, 0xbeee0078,
+	0x8078ff78, 0x00000200,
+	0xbefd0084, 0xbf0a6f7d,
+	0xbfa10061, 0xc4050078,
+	0x008ce800, 0x00000000,
+	0xc4050078, 0x008ce801,
+	0x00008000, 0xc4050078,
+	0x008ce802, 0x00010000,
+	0xc4050078, 0x008ce803,
+	0x00018000, 0xbf8a0000,
+	0x7e008500, 0x7e028501,
+	0x7e048502, 0x7e068503,
+	0x807d847d, 0x8078ff78,
+	0x00000200, 0xbf0a6f7d,
+	0xbfa2ffea, 0xc405006e,
+	0x008ce800, 0x00000000,
+	0xc405006e, 0x008ce801,
+	0x00008000, 0xc405006e,
+	0x008ce802, 0x00010000,
+	0xc405006e, 0x008ce803,
+	0x00018000, 0xbf8a0000,
+	0xbfa0003d, 0xbef600ff,
+	0x01000000, 0xbeee0078,
+	0x8078ff78, 0x00000400,
+	0xbefd0084, 0xbf0a6f7d,
+	0xbfa10016, 0xc4050078,
+	0x008ce800, 0x00000000,
+	0xc4050078, 0x008ce801,
+	0x00010000, 0xc4050078,
+	0x008ce802, 0x00020000,
+	0xc4050078, 0x008ce803,
+	0x00030000, 0xbf8a0000,
+	0x7e008500, 0x7e028501,
+	0x7e048502, 0x7e068503,
+	0x807d847d, 0x8078ff78,
+	0x00000400, 0xbf0a6f7d,
+	0xbfa2ffea, 0xb8ef1e06,
+	0x8b6fc16f, 0xbfa1000f,
+	0x846f836f, 0x806f7d6f,
+	0xbefe00c1, 0xbeff0080,
+	0xc4050078, 0x008ce800,
+	0x00000000, 0xbf8a0000,
 	0x7e008500, 0x807d817d,
 	0x8078ff78, 0x00000080,
-	0xbf0a6f7d, 0xbfa2fff7,
-	0xbeff00c1, 0xe0505000,
-	0x6e1d0000, 0xe0505100,
-	0x6e1d0100, 0xe0505200,
-	0x6e1d0200, 0xe0505300,
-	0x6e1d0300, 0xbf8903f7,
+	0xbf0a6f7d, 0xbfa2fff6,
+	0xbeff00c1, 0xc405006e,
+	0x008ce800, 0x00000000,
+	0xc405006e, 0x008ce801,
+	0x00010000, 0xc405006e,
+	0x008ce802, 0x00020000,
+	0xc405006e, 0x008ce803,
+	0x00030000, 0xbf8a0000,
 	0xb8f83b05, 0x80788178,
 	0xbf0d9972, 0xbfa20002,
 	0x84788978, 0xbfa00001,
@@ -2865,17 +4088,17 @@ static const uint32_t cwsr_trap_gfx11_hex[] = {
 	0x80f8ff78, 0x00000050,
 	0xbef600ff, 0x01000000,
 	0xbefd00ff, 0x0000006c,
-	0x80f89078, 0xf428403a,
-	0xf0000000, 0xbf89fc07,
+	0x80f89078, 0xf462403a,
+	0xf0000000, 0xbf8a0000,
 	0x80fd847d, 0xbf800000,
 	0xbe804300, 0xbe824302,
-	0x80f8a078, 0xf42c403a,
-	0xf0000000, 0xbf89fc07,
+	0x80f8a078, 0xf462603a,
+	0xf0000000, 0xbf8a0000,
 	0x80fd887d, 0xbf800000,
 	0xbe804300, 0xbe824302,
 	0xbe844304, 0xbe864306,
-	0x80f8c078, 0xf430403a,
-	0xf0000000, 0xbf89fc07,
+	0x80f8c078, 0xf462803a,
+	0xf0000000, 0xbf8a0000,
 	0x80fd907d, 0xbf800000,
 	0xbe804300, 0xbe824302,
 	0xbe844304, 0xbe864306,
@@ -2883,37 +4106,49 @@ static const uint32_t cwsr_trap_gfx11_hex[] = {
 	0xbe8c430c, 0xbe8e430e,
 	0xbf06807d, 0xbfa1fff0,
 	0xb980f801, 0x00000000,
-	0xbfbd0000, 0xb8f83b05,
-	0x80788178, 0xbf0d9972,
-	0xbfa20002, 0x84788978,
-	0xbfa00001, 0x84788a78,
-	0xb8ee1e06, 0x846e8a6e,
-	0x80786e78, 0x8078ff78,
-	0x00000200, 0xbef600ff,
-	0x01000000, 0xf4205bfa,
+	0xb8f83b05, 0x80788178,
+	0xbf0d9972, 0xbfa20002,
+	0x84788978, 0xbfa00001,
+	0x84788a78, 0xb8ee1e06,
+	0x846e8a6e, 0x80786e78,
+	0x8078ff78, 0x00000200,
+	0xbef600ff, 0x01000000,
+	0xbeff0071, 0xf4621bfa,
 	0xf0000000, 0x80788478,
-	0xf4205b3a, 0xf0000000,
-	0x80788478, 0xf4205b7a,
+	0xf4621b3a, 0xf0000000,
+	0x80788478, 0xf4621b7a,
 	0xf0000000, 0x80788478,
-	0xf4205c3a, 0xf0000000,
-	0x80788478, 0xf4205c7a,
+	0xf4621c3a, 0xf0000000,
+	0x80788478, 0xf4621c7a,
 	0xf0000000, 0x80788478,
-	0xf4205eba, 0xf0000000,
-	0x80788478, 0xf4205efa,
+	0xf4621eba, 0xf0000000,
+	0x80788478, 0xf4621efa,
 	0xf0000000, 0x80788478,
-	0xf4205e7a, 0xf0000000,
-	0x80788478, 0xf4205cfa,
+	0xf4621e7a, 0xf0000000,
+	0x80788478, 0xf4621cfa,
 	0xf0000000, 0x80788478,
-	0xf4205bba, 0xf0000000,
-	0x80788478, 0xbf89fc07,
-	0xb96ef814, 0xf4205bba,
+	0xf4621bba, 0xf0000000,
+	0x80788478, 0xbf8a0000,
+	0xb96ef814, 0xf4621bba,
+	0xf0000000, 0x80788478,
+	0xbf8a0000, 0xb96ef815,
+	0xf4621bba, 0xf0000000,
+	0x80788478, 0xbf8a0000,
+	0xb96ef812, 0xf4621bba,
 	0xf0000000, 0x80788478,
-	0xbf89fc07, 0xb96ef815,
+	0xbf8a0000, 0xb96ef813,
+	0x8b6eff7f, 0x04000000,
+	0xbfa1000d, 0x80788478,
+	0xf4621bba, 0xf0000000,
+	0x80788478, 0xbf8a0000,
+	0xbf0d806e, 0xbfa10006,
+	0x856e906e, 0x8b6e6e6e,
+	0xbfa10003, 0xbe804ec1,
+	0x816ec16e, 0xbfa0fffb,
 	0xbefd006f, 0xbefe0070,
-	0xbeff0071, 0x8b6f7bff,
-	0x000003ff, 0xb96f4803,
-	0x8b6f7bff, 0xfffff800,
-	0x856f8b6f, 0xb96fa2c3,
+	0xbeff0071, 0xb97b2011,
+	0x857b867b, 0xb97b0191,
+	0x857b827b, 0xb97bba11,
 	0xb973f801, 0xb8ee3b05,
 	0x806e816e, 0xbf0d9972,
 	0xbfa20002, 0x846e896e,
@@ -2922,135 +4157,137 @@ static const uint32_t cwsr_trap_gfx11_hex[] = {
 	0x806e6f6e, 0x806eff6e,
 	0x00000200, 0x806e746e,
 	0x826f8075, 0x8b6fff6f,
-	0x0000ffff, 0xf4085c37,
-	0xf8000050, 0xf4085d37,
-	0xf8000060, 0xf4005e77,
-	0xf8000074, 0xbf89fc07,
+	0x0000ffff, 0xf4605c37,
+	0xf8000050, 0xf4605d37,
+	0xf8000060, 0xf4601e77,
+	0xf8000074, 0xbf8a0000,
 	0x8b6dff6d, 0x0000ffff,
 	0x8bfe7e7e, 0x8bea6a6a,
-	0xb8eef802, 0xbf0d866e,
-	0xbfa20002, 0xb97af802,
-	0xbe80486c, 0xb97af802,
-	0xbe804a6c, 0xbfb00000,
+	0xb97af804, 0xbe804ec2,
+	0xbf94fffe, 0xbe804a6c,
+	0xbe804ec2, 0xbf94fffe,
+	0xbfb10000, 0xbf9f0000,
 	0xbf9f0000, 0xbf9f0000,
 	0xbf9f0000, 0xbf9f0000,
-	0xbf9f0000, 0x00000000,
 };
 
-static const uint32_t cwsr_trap_gfx9_4_3_hex[] = {
-	0xbf820001, 0xbf8202d6,
-	0xb8f8f802, 0x89788678,
-	0xb8fbf803, 0x866eff78,
-	0x00002000, 0xbf840009,
-	0x866eff6d, 0x00ff0000,
-	0xbf85001a, 0x866eff7b,
-	0x00000400, 0xbf85004d,
-	0xbf8e0010, 0xb8fbf803,
-	0xbf82fffa, 0x866eff7b,
-	0x03c00900, 0xbf850011,
-	0x866eff7b, 0x000071ff,
-	0xbf840008, 0x866fff7b,
-	0x00007080, 0xbf840001,
-	0xbeee1a87, 0xb8eff801,
-	0x8e6e8c6e, 0x866e6f6e,
-	0xbf850006, 0x866eff6d,
-	0x00ff0000, 0xbf850003,
+static const uint32_t cwsr_trap_gfx9_5_0_hex[] = {
+	0xbf820001, 0xbf8202ca,
+	0xb8f8f802, 0x8978ff78,
+	0x00020006, 0xb8fbf803,
+	0x866eff78, 0x00002000,
+	0xbf840009, 0x866eff6d,
+	0x00ff0000, 0xbf85001a,
 	0x866eff7b, 0x00000400,
-	0xbf850036, 0xb8faf807,
+	0xbf850051, 0xbf8e0010,
+	0xb8fbf803, 0xbf82fffa,
+	0x866eff7b, 0x03c00900,
+	0xbf850011, 0x866eff7b,
+	0x000071ff, 0xbf840008,
+	0x866fff7b, 0x00007080,
+	0xbf840001, 0xbeee1a87,
+	0xb8eff801, 0x8e6e8c6e,
+	0x866e6f6e, 0xbf850006,
+	0x866eff6d, 0x00ff0000,
+	0xbf850003, 0x866eff7b,
+	0x00000400, 0xbf85003a,
+	0xb8faf807, 0x867aff7a,
+	0x001f8000, 0x8e7a8b7a,
+	0x8979ff79, 0xfc000000,
+	0x87797a79, 0xba7ff807,
+	0x00000000, 0xb8faf812,
+	0xb8fbf813, 0x8efa887a,
+	0xbf0d8f7b, 0xbf840002,
+	0x877bff7b, 0xffff0000,
+	0xc0031bbd, 0x00000010,
+	0xbf8cc07f, 0x8e6e976e,
+	0x8979ff79, 0x00800000,
+	0x87796e79, 0xc0071bbd,
+	0x00000000, 0xbf8cc07f,
+	0xc0071ebd, 0x00000008,
+	0xbf8cc07f, 0x86ee6e6e,
+	0xbf840001, 0xbe801d6e,
+	0x866eff6d, 0x01ff0000,
+	0xbf850005, 0x8778ff78,
+	0x00002000, 0x80ec886c,
+	0x82ed806d, 0xbf820005,
+	0x866eff6d, 0x01000000,
+	0xbf850002, 0x806c846c,
+	0x826d806d, 0x866dff6d,
+	0x0000ffff, 0x8f7a8b79,
 	0x867aff7a, 0x001f8000,
-	0x8e7a8b7a, 0x8979ff79,
-	0xfc000000, 0x87797a79,
-	0xba7ff807, 0x00000000,
-	0xb8faf812, 0xb8fbf813,
-	0x8efa887a, 0xc0031bbd,
-	0x00000010, 0xbf8cc07f,
-	0x8e6e976e, 0x8979ff79,
-	0x00800000, 0x87796e79,
-	0xc0071bbd, 0x00000000,
-	0xbf8cc07f, 0xc0071ebd,
-	0x00000008, 0xbf8cc07f,
-	0x86ee6e6e, 0xbf840001,
-	0xbe801d6e, 0x866eff6d,
-	0x01ff0000, 0xbf850005,
-	0x8778ff78, 0x00002000,
-	0x80ec886c, 0x82ed806d,
-	0xbf820005, 0x866eff6d,
-	0x01000000, 0xbf850002,
-	0x806c846c, 0x826d806d,
+	0xb97af807, 0x86fe7e7e,
+	0x86ea6a6a, 0x8f6e8378,
+	0xb96ee0c2, 0xbf800002,
+	0xb9780002, 0xbe801f6c,
 	0x866dff6d, 0x0000ffff,
-	0x8f7a8b79, 0x867aff7a,
-	0x001f8000, 0xb97af807,
-	0x86fe7e7e, 0x86ea6a6a,
-	0x8f6e8378, 0xb96ee0c2,
-	0xbf800002, 0xb9780002,
-	0xbe801f6c, 0x866dff6d,
-	0x0000ffff, 0xbefa0080,
-	0xb97a0283, 0xb8faf807,
-	0x867aff7a, 0x001f8000,
-	0x8e7a8b7a, 0x8979ff79,
-	0xfc000000, 0x87797a79,
-	0xba7ff807, 0x00000000,
-	0xbeee007e, 0xbeef007f,
-	0xbefe0180, 0xbf900004,
-	0x877a8478, 0xb97af802,
-	0xbf8e0002, 0xbf88fffe,
-	0xb8fa2985, 0x807a817a,
-	0x8e7a8a7a, 0x8e7a817a,
-	0xb8fb1605, 0x807b817b,
-	0x8e7b867b, 0x807a7b7a,
-	0x807a7e7a, 0x827b807f,
-	0x867bff7b, 0x0000ffff,
-	0xc04b1c3d, 0x00000050,
-	0xbf8cc07f, 0xc04b1d3d,
-	0x00000060, 0xbf8cc07f,
-	0xc0431e7d, 0x00000074,
-	0xbf8cc07f, 0xbef4007e,
-	0x8675ff7f, 0x0000ffff,
-	0x8775ff75, 0x00040000,
-	0xbef60080, 0xbef700ff,
-	0x00807fac, 0xbef1007c,
-	0xbef00080, 0xb8f02985,
-	0x80708170, 0x8e708a70,
-	0x8e708170, 0xb8fa1605,
-	0x807a817a, 0x8e7a867a,
-	0x80707a70, 0xbef60084,
-	0xbef600ff, 0x01000000,
-	0xbefe007c, 0xbefc0070,
-	0xc0611c7a, 0x0000007c,
-	0xbf8cc07f, 0x80708470,
-	0xbefc007e, 0xbefe007c,
-	0xbefc0070, 0xc0611b3a,
+	0xbefa0080, 0xb97a0283,
+	0xb8faf807, 0x867aff7a,
+	0x001f8000, 0x8e7a8b7a,
+	0x8979ff79, 0xfc000000,
+	0x87797a79, 0xba7ff807,
+	0x00000000, 0xbeee007e,
+	0xbeef007f, 0xbefe0180,
+	0xbf900004, 0x877a8478,
+	0xb97af802, 0xbf8e0002,
+	0xbf88fffe, 0xb8fa2985,
+	0x807a817a, 0x8e7a8a7a,
+	0x8e7a817a, 0xb8fb1605,
+	0x807b817b, 0x8e7b867b,
+	0x807a7b7a, 0x807a7e7a,
+	0x827b807f, 0x867bff7b,
+	0x0000ffff, 0xc04b1c3d,
+	0x00000050, 0xbf8cc07f,
+	0xc04b1d3d, 0x00000060,
+	0xbf8cc07f, 0xc0431e7d,
+	0x00000074, 0xbf8cc07f,
+	0xbef4007e, 0x8675ff7f,
+	0x0000ffff, 0x8775ff75,
+	0x00040000, 0xbef60080,
+	0xbef700ff, 0x00807fac,
+	0xbef1007c, 0xbef00080,
+	0xb8f02985, 0x80708170,
+	0x8e708a70, 0x8e708170,
+	0xb8fa1605, 0x807a817a,
+	0x8e7a867a, 0x80707a70,
+	0xbef60084, 0xbef600ff,
+	0x01000000, 0xbefe007c,
+	0xbefc0070, 0xc0611c7a,
 	0x0000007c, 0xbf8cc07f,
 	0x80708470, 0xbefc007e,
 	0xbefe007c, 0xbefc0070,
-	0xc0611b7a, 0x0000007c,
+	0xc0611b3a, 0x0000007c,
 	0xbf8cc07f, 0x80708470,
 	0xbefc007e, 0xbefe007c,
-	0xbefc0070, 0xc0611bba,
+	0xbefc0070, 0xc0611b7a,
 	0x0000007c, 0xbf8cc07f,
 	0x80708470, 0xbefc007e,
 	0xbefe007c, 0xbefc0070,
-	0xc0611bfa, 0x0000007c,
+	0xc0611bba, 0x0000007c,
 	0xbf8cc07f, 0x80708470,
 	0xbefc007e, 0xbefe007c,
-	0xbefc0070, 0xc0611e3a,
-	0x0000007c, 0xbf8cc07f,
-	0x80708470, 0xbefc007e,
-	0xb8fbf803, 0xbefe007c,
-	0xbefc0070, 0xc0611efa,
+	0xbefc0070, 0xc0611bfa,
 	0x0000007c, 0xbf8cc07f,
 	0x80708470, 0xbefc007e,
 	0xbefe007c, 0xbefc0070,
-	0xc0611a3a, 0x0000007c,
+	0xc0611e3a, 0x0000007c,
+	0xbf8cc07f, 0x80708470,
+	0xbefc007e, 0xb8fbf803,
+	0xbefe007c, 0xbefc0070,
+	0xc0611efa, 0x0000007c,
 	0xbf8cc07f, 0x80708470,
 	0xbefc007e, 0xbefe007c,
-	0xbefc0070, 0xc0611a7a,
-	0x0000007c, 0xbf8cc07f,
-	0x80708470, 0xbefc007e,
-	0xb8f1f801, 0xbefe007c,
-	0xbefc0070, 0xc0611c7a,
+	0xbefc0070, 0xc0611a3a,
 	0x0000007c, 0xbf8cc07f,
 	0x80708470, 0xbefc007e,
+	0xbefe007c, 0xbefc0070,
+	0xc0611a7a, 0x0000007c,
+	0xbf8cc07f, 0x80708470,
+	0xbefc007e, 0xb8f1f801,
+	0xbefe007c, 0xbefc0070,
+	0xc0611c7a, 0x0000007c,
+	0xbf8cc07f, 0x80708470,
+	0xbefc007e, 0xbf108080,
 	0x867aff7f, 0x04000000,
 	0xbeef0080, 0x876f6f7a,
 	0xb8f02985, 0x80708170,
@@ -3124,10 +4361,10 @@ static const uint32_t cwsr_trap_gfx9_4_3_hex[] = {
 	0xe0724200, 0x701d0200,
 	0xe0724300, 0x701d0300,
 	0xbefe00c1, 0xbeff00c1,
-	0xb8fb4306, 0x867bc17b,
-	0xbf840064, 0xbf8a0000,
+	0xb8fb5306, 0x867bc17b,
+	0xbf840052, 0xbf8a0000,
 	0x867aff6f, 0x04000000,
-	0xbf840060, 0x8e7b867b,
+	0xbf84004e, 0x8e7b867b,
 	0x8e7b827b, 0xbef6007b,
 	0xb8f02985, 0x80708170,
 	0x8e708a70, 0x8e708170,
@@ -3140,8 +4377,8 @@ static const uint32_t cwsr_trap_gfx9_4_3_hex[] = {
 	0x000204c1, 0x867aff78,
 	0x00400000, 0xbf850003,
 	0xb8faf803, 0x897a7aff,
-	0x10000000, 0xbf850030,
-	0x24040682, 0xd86e4000,
+	0x10000000, 0xbf85001d,
+	0x24040682, 0xd86c0000,
 	0x00000002, 0xbf8cc07f,
 	0xbe840080, 0xd2890000,
 	0x00000900, 0x80048104,
@@ -3152,29 +4389,20 @@ static const uint32_t cwsr_trap_gfx9_4_3_hex[] = {
 	0x80048104, 0xc069003a,
 	0x00000070, 0xbf8cc07f,
 	0x80709070, 0xbf06c004,
-	0xbf84ffee, 0xbe840080,
-	0xd2890000, 0x00000901,
-	0x80048104, 0xd2890001,
-	0x00000901, 0x80048104,
-	0xd2890002, 0x00000901,
-	0x80048104, 0xd2890003,
-	0x00000901, 0x80048104,
-	0xc069003a, 0x00000070,
-	0xbf8cc07f, 0x80709070,
-	0xbf06c004, 0xbf84ffee,
-	0x680404ff, 0x00000200,
+	0xbf84ffee, 0x680404ff,
+	0x00000100, 0xd0c9006a,
+	0x0000f702, 0xbf87ffe5,
+	0xbf820016, 0xd1060002,
+	0x00011103, 0x7e0602ff,
+	0x00000200, 0xbefc00ff,
+	0x00010000, 0xbe800077,
+	0x8677ff77, 0xff7fffff,
+	0x8777ff77, 0x00058000,
+	0xd8ec0000, 0x00000002,
+	0xbf8cc07f, 0xe0765000,
+	0x701d0002, 0x68040702,
 	0xd0c9006a, 0x0000f702,
-	0xbf87ffd2, 0xbf820015,
-	0xd1060002, 0x00011103,
-	0x7e0602ff, 0x00000200,
-	0xbefc00ff, 0x00010000,
-	0xbe800077, 0x8677ff77,
-	0xff7fffff, 0x8777ff77,
-	0x00058000, 0xd8ec0000,
-	0x00000002, 0xbf8cc07f,
-	0xe0765000, 0x701d0002,
-	0x68040702, 0xd0c9006a,
-	0x0000f702, 0xbf87fff7,
+	0xbefe016a, 0xbf87fff6,
 	0xbef70000, 0xbef000ff,
 	0x00000400, 0xbefe00c1,
 	0xbeff00c1, 0xb8fb2b05,
@@ -3301,15 +4529,15 @@ static const uint32_t cwsr_trap_gfx9_4_3_hex[] = {
 	0x701d0300, 0x807c847c,
 	0x8070ff70, 0x00000400,
 	0xbf0a7b7c, 0xbf85ffeb,
-	0xbf9c0000, 0xbf8200ee,
+	0xbf9c0000, 0xbf8200f4,
 	0xbef4007e, 0x8675ff7f,
 	0x0000ffff, 0x8775ff75,
 	0x00040000, 0xbef60080,
 	0xbef700ff, 0x00807fac,
 	0x866eff7f, 0x04000000,
-	0xbf84001f, 0xbefe00c1,
-	0xbeff00c1, 0xb8ef4306,
-	0x866fc16f, 0xbf84001a,
+	0xbf840025, 0xbefe00c1,
+	0xbeff00c1, 0xb8ef5306,
+	0x866fc16f, 0xbf840020,
 	0x8e6f866f, 0x8e6f826f,
 	0xbef6006f, 0xb8f82985,
 	0x80788178, 0x8e788a78,
@@ -3320,9 +4548,12 @@ static const uint32_t cwsr_trap_gfx9_4_3_hex[] = {
 	0x01000000, 0xbefc0080,
 	0xe0510000, 0x781d0000,
 	0xe0510100, 0x781d0000,
-	0x807cff7c, 0x00000200,
-	0x8078ff78, 0x00000200,
-	0xbf0a6f7c, 0xbf85fff6,
+	0xe0510200, 0x781d0000,
+	0xe0510300, 0x781d0000,
+	0xe0510400, 0x781d0000,
+	0x807cff7c, 0x00000500,
+	0x8078ff78, 0x00000500,
+	0xbf0a6f7c, 0xbf85fff0,
 	0xbefe00c1, 0xbeff00c1,
 	0xbef600ff, 0x01000000,
 	0xb8ef2b05, 0x806f816f,
@@ -3421,5 +4652,5 @@ static const uint32_t cwsr_trap_gfx9_4_3_hex[] = {
 	0x8f6e837a, 0xb96ee0c2,
 	0xbf800002, 0xb97a0002,
 	0xbf8a0000, 0xbe801f6c,
-	0xbf810000, 0x00000000,
+	0xbf9b0000, 0x00000000,
 };
diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm
index 8b92c33c2a7c..96fbb16ceb21 100644
--- a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm
+++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm
@@ -33,6 +33,7 @@
  * gfx11:
  *   cpp -DASIC_FAMILY=CHIP_PLUM_BONITO cwsr_trap_handler_gfx10.asm -P -o gfx11.sp3
  *   sp3 gfx11.sp3 -hex gfx11.hex
+ *
  */
 
 #define CHIP_NAVI10 26
@@ -43,22 +44,33 @@
 #define HAVE_XNACK (ASIC_FAMILY < CHIP_SIENNA_CICHLID)
 #define HAVE_SENDMSG_RTN (ASIC_FAMILY >= CHIP_PLUM_BONITO)
 #define HAVE_BUFFER_LDS_LOAD (ASIC_FAMILY < CHIP_PLUM_BONITO)
-#define SW_SA_TRAP (ASIC_FAMILY >= CHIP_PLUM_BONITO)
+#define SW_SA_TRAP (ASIC_FAMILY == CHIP_PLUM_BONITO)
+#define SAVE_AFTER_XNACK_ERROR (HAVE_XNACK && !NO_SQC_STORE) // workaround for TCP store failure after XNACK error when ALLOW_REPLAY=0, for debugger
+#define SINGLE_STEP_MISSED_WORKAROUND 1	//workaround for lost MODE.DEBUG_EN exception when SAVECTX raised
 
-var SINGLE_STEP_MISSED_WORKAROUND		= 1	//workaround for lost MODE.DEBUG_EN exception when SAVECTX raised
+#define S_COHERENCE glc:1
+#define V_COHERENCE slc:1 glc:1
+#define S_WAITCNT_0 s_waitcnt 0
 
 var SQ_WAVE_STATUS_SPI_PRIO_MASK		= 0x00000006
 var SQ_WAVE_STATUS_HALT_MASK			= 0x2000
 var SQ_WAVE_STATUS_ECC_ERR_MASK			= 0x20000
 var SQ_WAVE_STATUS_TRAP_EN_SHIFT		= 6
+var SQ_WAVE_IB_STS2_WAVE64_SHIFT		= 11
+var SQ_WAVE_IB_STS2_WAVE64_SIZE			= 1
+var SQ_WAVE_LDS_ALLOC_GRANULARITY		= 8
+var S_STATUS_HWREG				= HW_REG_STATUS
+var S_STATUS_ALWAYS_CLEAR_MASK			= SQ_WAVE_STATUS_SPI_PRIO_MASK|SQ_WAVE_STATUS_ECC_ERR_MASK
+var S_STATUS_HALT_MASK				= SQ_WAVE_STATUS_HALT_MASK
+var S_SAVE_PC_HI_TRAP_ID_MASK			= 0x00FF0000
+var S_SAVE_PC_HI_HT_MASK			= 0x01000000
 
+var SQ_WAVE_STATUS_NO_VGPRS_SHIFT		= 24
 var SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT		= 12
 var SQ_WAVE_LDS_ALLOC_LDS_SIZE_SIZE		= 9
 var SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SIZE		= 8
 var SQ_WAVE_LDS_ALLOC_VGPR_SHARED_SIZE_SHIFT	= 24
 var SQ_WAVE_LDS_ALLOC_VGPR_SHARED_SIZE_SIZE	= 4
-var SQ_WAVE_IB_STS2_WAVE64_SHIFT		= 11
-var SQ_WAVE_IB_STS2_WAVE64_SIZE			= 1
 
 #if ASIC_FAMILY < CHIP_PLUM_BONITO
 var SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SHIFT		= 8
@@ -73,14 +85,17 @@ var SQ_WAVE_TRAPSTS_ADDR_WATCH_MASK		= 0x80
 var SQ_WAVE_TRAPSTS_ADDR_WATCH_SHIFT		= 7
 var SQ_WAVE_TRAPSTS_MEM_VIOL_MASK		= 0x100
 var SQ_WAVE_TRAPSTS_MEM_VIOL_SHIFT		= 8
-var SQ_WAVE_TRAPSTS_PRE_SAVECTX_MASK		= 0x3FF
-var SQ_WAVE_TRAPSTS_PRE_SAVECTX_SHIFT		= 0x0
-var SQ_WAVE_TRAPSTS_PRE_SAVECTX_SIZE		= 10
-var SQ_WAVE_TRAPSTS_POST_SAVECTX_MASK		= 0xFFFFF800
-var SQ_WAVE_TRAPSTS_POST_SAVECTX_SHIFT		= 11
-var SQ_WAVE_TRAPSTS_POST_SAVECTX_SIZE		= 21
 var SQ_WAVE_TRAPSTS_ILLEGAL_INST_MASK		= 0x800
+var SQ_WAVE_TRAPSTS_ILLEGAL_INST_SHIFT		= 11
 var SQ_WAVE_TRAPSTS_EXCP_HI_MASK		= 0x7000
+#if ASIC_FAMILY >= CHIP_PLUM_BONITO
+var SQ_WAVE_TRAPSTS_HOST_TRAP_SHIFT		= 16
+var SQ_WAVE_TRAPSTS_WAVE_START_MASK		= 0x20000
+var SQ_WAVE_TRAPSTS_WAVE_START_SHIFT		= 17
+var SQ_WAVE_TRAPSTS_WAVE_END_MASK		= 0x40000
+var SQ_WAVE_TRAPSTS_TRAP_AFTER_INST_MASK	= 0x100000
+#endif
+var SQ_WAVE_TRAPSTS_XNACK_ERROR_MASK		= 0x10000000
 
 var SQ_WAVE_MODE_EXCP_EN_SHIFT			= 12
 var SQ_WAVE_MODE_EXCP_EN_ADDR_WATCH_SHIFT	= 19
@@ -92,6 +107,28 @@ var SQ_WAVE_IB_STS_RCNT_FIRST_REPLAY_MASK	= 0x003F8000
 
 var SQ_WAVE_MODE_DEBUG_EN_MASK			= 0x800
 
+var S_TRAPSTS_RESTORE_PART_1_SIZE		= SQ_WAVE_TRAPSTS_SAVECTX_SHIFT
+var S_TRAPSTS_RESTORE_PART_2_SHIFT		= SQ_WAVE_TRAPSTS_ILLEGAL_INST_SHIFT
+
+#if ASIC_FAMILY < CHIP_PLUM_BONITO
+var S_TRAPSTS_NON_MASKABLE_EXCP_MASK		= SQ_WAVE_TRAPSTS_MEM_VIOL_MASK|SQ_WAVE_TRAPSTS_ILLEGAL_INST_MASK
+var S_TRAPSTS_RESTORE_PART_2_SIZE		= 32 - S_TRAPSTS_RESTORE_PART_2_SHIFT
+var S_TRAPSTS_RESTORE_PART_3_SHIFT		= 0
+var S_TRAPSTS_RESTORE_PART_3_SIZE		= 0
+#else
+var S_TRAPSTS_NON_MASKABLE_EXCP_MASK		= SQ_WAVE_TRAPSTS_MEM_VIOL_MASK		|\
+						  SQ_WAVE_TRAPSTS_ILLEGAL_INST_MASK	|\
+						  SQ_WAVE_TRAPSTS_WAVE_START_MASK	|\
+						  SQ_WAVE_TRAPSTS_WAVE_END_MASK		|\
+						  SQ_WAVE_TRAPSTS_TRAP_AFTER_INST_MASK
+var S_TRAPSTS_RESTORE_PART_2_SIZE		= SQ_WAVE_TRAPSTS_HOST_TRAP_SHIFT - SQ_WAVE_TRAPSTS_ILLEGAL_INST_SHIFT
+var S_TRAPSTS_RESTORE_PART_3_SHIFT		= SQ_WAVE_TRAPSTS_WAVE_START_SHIFT
+var S_TRAPSTS_RESTORE_PART_3_SIZE		= 32 - S_TRAPSTS_RESTORE_PART_3_SHIFT
+#endif
+var S_TRAPSTS_HWREG				= HW_REG_TRAPSTS
+var S_TRAPSTS_SAVE_CONTEXT_MASK			= SQ_WAVE_TRAPSTS_SAVECTX_MASK
+var S_TRAPSTS_SAVE_CONTEXT_SHIFT		= SQ_WAVE_TRAPSTS_SAVECTX_SHIFT
+
 // bits [31:24] unused by SPI debug data
 var TTMP11_SAVE_REPLAY_W64H_SHIFT		= 31
 var TTMP11_SAVE_REPLAY_W64H_MASK		= 0x80000000
@@ -104,8 +141,6 @@ var TTMP11_DEBUG_TRAP_ENABLED_MASK		= 0x800000
 // when ADD_TID_ENABLE and BUF_DATA_FORMAT_32 for MTBUF), ADD_TID_ENABLE
 var S_SAVE_BUF_RSRC_WORD1_STRIDE		= 0x00040000
 var S_SAVE_BUF_RSRC_WORD3_MISC			= 0x10807FAC
-var S_SAVE_PC_HI_TRAP_ID_MASK			= 0x00FF0000
-var S_SAVE_PC_HI_HT_MASK			= 0x01000000
 var S_SAVE_SPI_INIT_FIRST_WAVE_MASK		= 0x04000000
 var S_SAVE_SPI_INIT_FIRST_WAVE_SHIFT		= 26
 
@@ -165,6 +200,7 @@ var s_restore_buf_rsrc3				= ttmp11
 var s_restore_size				= ttmp6
 var s_restore_ttmps_lo				= s_restore_tmp
 var s_restore_ttmps_hi				= s_restore_alloc_size
+var s_restore_spi_init_hi_save			= s_restore_exec_hi
 
 shader main
 	asic(DEFAULT)
@@ -177,13 +213,13 @@ L_JUMP_TO_RESTORE:
 	s_branch	L_RESTORE
 
 L_SKIP_RESTORE:
-	s_getreg_b32	s_save_status, hwreg(HW_REG_STATUS)			//save STATUS since we will change SCC
+	s_getreg_b32	s_save_status, hwreg(S_STATUS_HWREG)			//save STATUS since we will change SCC
 
 	// Clear SPI_PRIO: do not save with elevated priority.
 	// Clear ECC_ERR: prevents SQC store and triggers FATAL_HALT if setreg'd.
-	s_andn2_b32	s_save_status, s_save_status, SQ_WAVE_STATUS_SPI_PRIO_MASK|SQ_WAVE_STATUS_ECC_ERR_MASK
+	s_andn2_b32	s_save_status, s_save_status, S_STATUS_ALWAYS_CLEAR_MASK
 
-	s_getreg_b32	s_save_trapsts, hwreg(HW_REG_TRAPSTS)
+	s_getreg_b32	s_save_trapsts, hwreg(S_TRAPSTS_HWREG)
 
 #if SW_SA_TRAP
 	// If ttmp1[30] is set then issue s_barrier to unblock dependent waves.
@@ -198,7 +234,7 @@ L_TRAP_NO_BARRIER:
 	s_cbranch_scc1  L_CHECK_SAVE
 #endif
 
-	s_and_b32       ttmp2, s_save_status, SQ_WAVE_STATUS_HALT_MASK
+	s_and_b32       ttmp2, s_save_status, S_STATUS_HALT_MASK
 	s_cbranch_scc0	L_NOT_HALTED
 
 L_HALTED:
@@ -207,14 +243,14 @@ L_HALTED:
 	s_cbranch_scc1	L_FETCH_2ND_TRAP
 
 L_CHECK_SAVE:
-	s_and_b32	ttmp2, s_save_trapsts, SQ_WAVE_TRAPSTS_SAVECTX_MASK
+	s_and_b32	ttmp2, s_save_trapsts, S_TRAPSTS_SAVE_CONTEXT_MASK
 	s_cbranch_scc1	L_SAVE
 
 	// Wave is halted but neither host trap nor SAVECTX is raised.
 	// Caused by instruction fetch memory violation.
 	// Spin wait until context saved to prevent interrupt storm.
 	s_sleep		0x10
-	s_getreg_b32	s_save_trapsts, hwreg(HW_REG_TRAPSTS)
+	s_getreg_b32	s_save_trapsts, hwreg(S_TRAPSTS_HWREG)
 	s_branch	L_CHECK_SAVE
 
 L_NOT_HALTED:
@@ -224,7 +260,7 @@ L_NOT_HALTED:
 	// Check non-maskable exceptions. memory_violation, illegal_instruction
 	// and xnack_error exceptions always cause the wave to enter the trap
 	// handler.
-	s_and_b32	ttmp2, s_save_trapsts, SQ_WAVE_TRAPSTS_MEM_VIOL_MASK|SQ_WAVE_TRAPSTS_ILLEGAL_INST_MASK
+	s_and_b32	ttmp2, s_save_trapsts, S_TRAPSTS_NON_MASKABLE_EXCP_MASK
 	s_cbranch_scc1	L_FETCH_2ND_TRAP
 
 	// Check for maskable exceptions in trapsts.excp and trapsts.excp_hi.
@@ -248,15 +284,15 @@ L_CHECK_TRAP_ID:
 	s_and_b32	ttmp2, s_save_pc_hi, S_SAVE_PC_HI_TRAP_ID_MASK
 	s_cbranch_scc1	L_FETCH_2ND_TRAP
 
-if SINGLE_STEP_MISSED_WORKAROUND
+#if SINGLE_STEP_MISSED_WORKAROUND
 	// Prioritize single step exception over context save.
 	// Second-level trap will halt wave and RFE, re-entering for SAVECTX.
 	s_getreg_b32	ttmp2, hwreg(HW_REG_MODE)
 	s_and_b32	ttmp2, ttmp2, SQ_WAVE_MODE_DEBUG_EN_MASK
 	s_cbranch_scc1	L_FETCH_2ND_TRAP
-end
+#endif
 
-	s_and_b32	ttmp2, s_save_trapsts, SQ_WAVE_TRAPSTS_SAVECTX_MASK
+	s_and_b32	ttmp2, s_save_trapsts, S_TRAPSTS_SAVE_CONTEXT_MASK
 	s_cbranch_scc1	L_SAVE
 
 L_FETCH_2ND_TRAP:
@@ -269,23 +305,28 @@ L_FETCH_2ND_TRAP:
 	// ttmp12 holds SQ_WAVE_STATUS
 #if HAVE_SENDMSG_RTN
 	s_sendmsg_rtn_b64       [ttmp14, ttmp15], sendmsg(MSG_RTN_GET_TMA)
-	s_waitcnt       lgkmcnt(0)
+	S_WAITCNT_0
 #else
 	s_getreg_b32	ttmp14, hwreg(HW_REG_SHADER_TMA_LO)
 	s_getreg_b32	ttmp15, hwreg(HW_REG_SHADER_TMA_HI)
 #endif
 	s_lshl_b64	[ttmp14, ttmp15], [ttmp14, ttmp15], 0x8
 
-	s_load_dword    ttmp2, [ttmp14, ttmp15], 0x10 glc:1			// debug trap enabled flag
-	s_waitcnt       lgkmcnt(0)
+	s_bitcmp1_b32	ttmp15, 0xF
+	s_cbranch_scc0	L_NO_SIGN_EXTEND_TMA
+	s_or_b32	ttmp15, ttmp15, 0xFFFF0000
+L_NO_SIGN_EXTEND_TMA:
+
+	s_load_dword    ttmp2, [ttmp14, ttmp15], 0x10 S_COHERENCE		// debug trap enabled flag
+	S_WAITCNT_0
 	s_lshl_b32      ttmp2, ttmp2, TTMP11_DEBUG_TRAP_ENABLED_SHIFT
 	s_andn2_b32     ttmp11, ttmp11, TTMP11_DEBUG_TRAP_ENABLED_MASK
 	s_or_b32        ttmp11, ttmp11, ttmp2
 
-	s_load_dwordx2	[ttmp2, ttmp3], [ttmp14, ttmp15], 0x0 glc:1		// second-level TBA
-	s_waitcnt	lgkmcnt(0)
-	s_load_dwordx2	[ttmp14, ttmp15], [ttmp14, ttmp15], 0x8 glc:1		// second-level TMA
-	s_waitcnt	lgkmcnt(0)
+	s_load_dwordx2	[ttmp2, ttmp3], [ttmp14, ttmp15], 0x0 S_COHERENCE	// second-level TBA
+	S_WAITCNT_0
+	s_load_dwordx2	[ttmp14, ttmp15], [ttmp14, ttmp15], 0x8 S_COHERENCE	// second-level TMA
+	S_WAITCNT_0
 
 	s_and_b64	[ttmp2, ttmp3], [ttmp2, ttmp3], [ttmp2, ttmp3]
 	s_cbranch_scc0	L_NO_NEXT_TRAP						// second-level trap handler not been set
@@ -293,9 +334,13 @@ L_FETCH_2ND_TRAP:
 
 L_NO_NEXT_TRAP:
 	// If not caused by trap then halt wave to prevent re-entry.
-	s_and_b32	ttmp2, s_save_pc_hi, (S_SAVE_PC_HI_TRAP_ID_MASK|S_SAVE_PC_HI_HT_MASK)
+	s_and_b32	ttmp2, s_save_pc_hi, S_SAVE_PC_HI_TRAP_ID_MASK
 	s_cbranch_scc1	L_TRAP_CASE
-	s_or_b32	s_save_status, s_save_status, SQ_WAVE_STATUS_HALT_MASK
+
+	// Host trap will not cause trap re-entry.
+	s_and_b32	ttmp2, s_save_pc_hi, S_SAVE_PC_HI_HT_MASK
+	s_cbranch_scc1	L_EXIT_TRAP
+	s_or_b32	s_save_status, s_save_status, S_STATUS_HALT_MASK
 
 	// If the PC points to S_ENDPGM then context save will fail if STATUS.HALT is set.
 	// Rewind the PC to prevent this from occurring.
@@ -305,10 +350,6 @@ L_NO_NEXT_TRAP:
 	s_branch	L_EXIT_TRAP
 
 L_TRAP_CASE:
-	// Host trap will not cause trap re-entry.
-	s_and_b32	ttmp2, s_save_pc_hi, S_SAVE_PC_HI_HT_MASK
-	s_cbranch_scc1	L_EXIT_TRAP
-
 	// Advance past trap instruction to prevent re-entry.
 	s_add_u32	ttmp0, ttmp0, 0x4
 	s_addc_u32	ttmp1, ttmp1, 0x0
@@ -323,14 +364,22 @@ L_EXIT_TRAP:
 	// Restore SQ_WAVE_STATUS.
 	s_and_b64	exec, exec, exec					// Restore STATUS.EXECZ, not writable by s_setreg_b32
 	s_and_b64	vcc, vcc, vcc						// Restore STATUS.VCCZ, not writable by s_setreg_b32
-	s_setreg_b32	hwreg(HW_REG_STATUS), s_save_status
 
+	s_setreg_b32	hwreg(S_STATUS_HWREG), s_save_status
 	s_rfe_b64	[ttmp0, ttmp1]
 
 L_SAVE:
+	// If VGPRs have been deallocated then terminate the wavefront.
+	// It has no remaining program to run and cannot save without VGPRs.
+#if ASIC_FAMILY == CHIP_PLUM_BONITO
+	s_bitcmp1_b32	s_save_status, SQ_WAVE_STATUS_NO_VGPRS_SHIFT
+	s_cbranch_scc0	L_HAVE_VGPRS
+	s_endpgm
+L_HAVE_VGPRS:
+#endif
 	s_and_b32	s_save_pc_hi, s_save_pc_hi, 0x0000ffff			//pc[47:32]
 	s_mov_b32	s_save_tmp, 0
-	s_setreg_b32	hwreg(HW_REG_TRAPSTS, SQ_WAVE_TRAPSTS_SAVECTX_SHIFT, 1), s_save_tmp	//clear saveCtx bit
+	s_setreg_b32	hwreg(S_TRAPSTS_HWREG, S_TRAPSTS_SAVE_CONTEXT_SHIFT, 1), s_save_tmp	//clear saveCtx bit
 
 #if HAVE_XNACK
 	save_and_clear_ib_sts(s_save_tmp, s_save_trapsts)
@@ -355,7 +404,7 @@ L_SLEEP:
 	s_sleep		0x2
 	s_cbranch_execz	L_SLEEP
 #else
-	s_waitcnt	lgkmcnt(0)
+	S_WAITCNT_0
 #endif
 
 	// Save first_wave flag so we can clear high bits of save address.
@@ -364,6 +413,12 @@ L_SLEEP:
 	s_or_b32	s_save_pc_hi, s_save_pc_hi, s_save_tmp
 
 #if NO_SQC_STORE
+#if ASIC_FAMILY <= CHIP_SIENNA_CICHLID
+	// gfx10: If there was a VALU exception, the exception state must be
+	// cleared before executing the VALU instructions below.
+	v_clrexcp
+#endif
+
 	// Trap temporaries must be saved via VGPR but all VGPRs are in use.
 	// There is no ttmp space to hold the resource constant for VGPR save.
 	// Save v0 by itself since it requires only two SGPRs.
@@ -371,7 +426,7 @@ L_SLEEP:
 	s_and_b32	s_save_ttmps_hi, exec_hi, 0xFFFF
 	s_mov_b32	exec_lo, 0xFFFFFFFF
 	s_mov_b32	exec_hi, 0xFFFFFFFF
-	global_store_dword_addtid	v0, [s_save_ttmps_lo, s_save_ttmps_hi] slc:1 glc:1
+	global_store_dword_addtid	v0, [s_save_ttmps_lo, s_save_ttmps_hi] V_COHERENCE
 	v_mov_b32	v0, 0x0
 	s_mov_b32	exec_lo, s_save_ttmps_lo
 	s_mov_b32	exec_hi, s_save_ttmps_hi
@@ -379,7 +434,7 @@ L_SLEEP:
 
 	// Save trap temporaries 4-11, 13 initialized by SPI debug dispatch logic
 	// ttmp SR memory offset : size(VGPR)+size(SVGPR)+size(SGPR)+0x40
-	get_wave_size(s_save_ttmps_hi)
+	get_wave_size2(s_save_ttmps_hi)
 	get_vgpr_size_bytes(s_save_ttmps_lo, s_save_ttmps_hi)
 	get_svgpr_size_bytes(s_save_ttmps_hi)
 	s_add_u32	s_save_ttmps_lo, s_save_ttmps_lo, s_save_ttmps_hi
@@ -403,15 +458,15 @@ L_SLEEP:
 
 	s_mov_b32	exec_lo, 0x3FFF
 	s_mov_b32	exec_hi, 0x0
-	global_store_dword_addtid	v0, [s_save_ttmps_lo, s_save_ttmps_hi] inst_offset:0x40 slc:1 glc:1
+	global_store_dword_addtid	v0, [s_save_ttmps_lo, s_save_ttmps_hi] inst_offset:0x40 V_COHERENCE
 	v_readlane_b32	ttmp14, v0, 0xE
 	v_readlane_b32	ttmp15, v0, 0xF
 	s_mov_b32	exec_lo, ttmp14
 	s_mov_b32	exec_hi, ttmp15
 #else
-	s_store_dwordx4	[ttmp4, ttmp5, ttmp6, ttmp7], [s_save_ttmps_lo, s_save_ttmps_hi], 0x50 glc:1
-	s_store_dwordx4	[ttmp8, ttmp9, ttmp10, ttmp11], [s_save_ttmps_lo, s_save_ttmps_hi], 0x60 glc:1
-	s_store_dword   ttmp13, [s_save_ttmps_lo, s_save_ttmps_hi], 0x74 glc:1
+	s_store_dwordx4	[ttmp4, ttmp5, ttmp6, ttmp7], [s_save_ttmps_lo, s_save_ttmps_hi], 0x50 S_COHERENCE
+	s_store_dwordx4	[ttmp8, ttmp9, ttmp10, ttmp11], [s_save_ttmps_lo, s_save_ttmps_hi], 0x60 S_COHERENCE
+	s_store_dword   ttmp13, [s_save_ttmps_lo, s_save_ttmps_hi], 0x74 S_COHERENCE
 #endif
 
 	/* setup Resource Contants */
@@ -425,7 +480,7 @@ L_SLEEP:
 
 	/* global mem offset */
 	s_mov_b32	s_save_mem_offset, 0x0
-	get_wave_size(s_wave_size)
+	get_wave_size2(s_wave_size)
 
 #if HAVE_XNACK
 	// Save and clear vector XNACK state late to free up SGPRs.
@@ -449,12 +504,22 @@ L_SAVE_4VGPR_WAVE32:
 
 	// VGPR Allocated in 4-GPR granularity
 
+#if SAVE_AFTER_XNACK_ERROR
+	check_if_tcp_store_ok()
+	s_cbranch_scc1 L_SAVE_FIRST_VGPRS32_WITH_TCP
+
+	write_vgprs_to_mem_with_sqc_w32(v0, 4, s_save_buf_rsrc0, s_save_mem_offset)
+	s_branch L_SAVE_HWREG
+
+L_SAVE_FIRST_VGPRS32_WITH_TCP:
+#endif
+
 #if !NO_SQC_STORE
-	buffer_store_dword	v0, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1
+	buffer_store_dword	v0, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE
 #endif
-	buffer_store_dword	v1, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:128
-	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:128*2
-	buffer_store_dword	v3, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:128*3
+	buffer_store_dword	v1, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE offset:128
+	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE offset:128*2
+	buffer_store_dword	v3, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE offset:128*3
 	s_branch	L_SAVE_HWREG
 
 L_SAVE_4VGPR_WAVE64:
@@ -462,12 +527,22 @@ L_SAVE_4VGPR_WAVE64:
 
 	// VGPR Allocated in 4-GPR granularity
 
+#if  SAVE_AFTER_XNACK_ERROR
+	check_if_tcp_store_ok()
+	s_cbranch_scc1 L_SAVE_FIRST_VGPRS64_WITH_TCP
+
+	write_vgprs_to_mem_with_sqc_w64(v0, 4, s_save_buf_rsrc0, s_save_mem_offset)
+	s_branch L_SAVE_HWREG
+
+L_SAVE_FIRST_VGPRS64_WITH_TCP:
+#endif
+
 #if !NO_SQC_STORE
-	buffer_store_dword	v0, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1
+	buffer_store_dword	v0, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE
 #endif
-	buffer_store_dword	v1, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:256
-	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:256*2
-	buffer_store_dword	v3, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:256*3
+	buffer_store_dword	v1, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE offset:256
+	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE offset:256*2
+	buffer_store_dword	v3, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE offset:256*3
 
 	/* save HW registers */
 
@@ -495,7 +570,7 @@ L_SAVE_HWREG:
 	write_hwreg_to_mem(s_save_exec_hi, s_save_buf_rsrc0, s_save_mem_offset)
 	write_hwreg_to_mem(s_save_status, s_save_buf_rsrc0, s_save_mem_offset)
 
-	s_getreg_b32	s_save_tmp, hwreg(HW_REG_TRAPSTS)
+	s_getreg_b32	s_save_tmp, hwreg(S_TRAPSTS_HWREG)
 	write_hwreg_to_mem(s_save_tmp, s_save_buf_rsrc0, s_save_mem_offset)
 
 	// Not used on Sienna_Cichlid but keep layout same for debugger.
@@ -514,7 +589,7 @@ L_SAVE_HWREG:
 	// Write HWREGs with 16 VGPR lanes. TTMPs occupy space after this.
 	s_mov_b32       exec_lo, 0xFFFF
 	s_mov_b32	exec_hi, 0x0
-	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1
+	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE
 
 	// Write SGPRs with 32 VGPR lanes. This works in wave32 and wave64 mode.
 	s_mov_b32       exec_lo, 0xFFFFFFFF
@@ -557,7 +632,7 @@ L_SAVE_SGPR_LOOP:
 	s_cmp_eq_u32	ttmp13, 0x20						//have 32 VGPR lanes filled?
 	s_cbranch_scc0	L_SAVE_SGPR_SKIP_TCP_STORE
 
-	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1
+	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE
 	s_add_u32	s_save_mem_offset, s_save_mem_offset, 0x80
 	s_mov_b32	ttmp13, 0x0
 	v_mov_b32	v2, 0x0
@@ -578,7 +653,7 @@ L_SAVE_SGPR_SKIP_TCP_STORE:
 	write_12sgpr_to_mem(s0, s_save_buf_rsrc0, s_save_mem_offset)
 
 #if NO_SQC_STORE
-	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1
+	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE
 #else
 	// restore s_save_buf_rsrc0,1
 	s_mov_b32	s_save_buf_rsrc0, s_save_xnack_mask
@@ -608,8 +683,7 @@ L_SAVE_LDS_NORMAL:
 
 	// first wave do LDS save;
 
-	s_lshl_b32	s_save_alloc_size, s_save_alloc_size, 6			//LDS size in dwords = lds_size * 64dw
-	s_lshl_b32	s_save_alloc_size, s_save_alloc_size, 2			//LDS size in bytes
+	s_lshl_b32	s_save_alloc_size, s_save_alloc_size, SQ_WAVE_LDS_ALLOC_GRANULARITY
 	s_mov_b32	s_save_buf_rsrc2, s_save_alloc_size			//NUM_RECORDS in bytes
 
 	// LDS at offset: size(VGPR)+size(SVGPR)+SIZE(SGPR)+SIZE(HWREG)
@@ -634,16 +708,36 @@ L_SAVE_LDS_NORMAL:
 	s_cbranch_scc1	L_SAVE_LDS_W64
 
 L_SAVE_LDS_W32:
+#if SAVE_AFTER_XNACK_ERROR
+	check_if_tcp_store_ok()
+	s_cbranch_scc1 L_SAVE_LDS_WITH_TCP_W32
+
+L_SAVE_LDS_LOOP_SQC_W32:
+	ds_read_b32	v1, v0
+	S_WAITCNT_0
+
+	write_vgprs_to_mem_with_sqc_w32(v1, 1, s_save_buf_rsrc0, s_save_mem_offset)
+
+	s_add_u32	m0, m0, 128						//every buffer_store_lds does 128 bytes
+	v_add_nc_u32	v0, v0, 128						//mem offset increased by 128 bytes
+	s_cmp_lt_u32	m0, s_save_alloc_size					//scc=(m0 < s_save_alloc_size) ? 1 : 0
+	s_cbranch_scc1	L_SAVE_LDS_LOOP_SQC_W32					//LDS save is complete?
+
+	s_branch	L_SAVE_LDS_DONE
+
+L_SAVE_LDS_WITH_TCP_W32:
+#endif
+
 	s_mov_b32	s3, 128
 	s_nop		0
 	s_nop		0
 	s_nop		0
 L_SAVE_LDS_LOOP_W32:
 	ds_read_b32	v1, v0
-	s_waitcnt	0
-	buffer_store_dword	v1, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1
+	S_WAITCNT_0
+	buffer_store_dword	v1, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE
 
-	s_add_u32	m0, m0, s3						//every buffer_store_lds does 256 bytes
+	s_add_u32	m0, m0, s3						//every buffer_store_lds does 128 bytes
 	s_add_u32	s_save_mem_offset, s_save_mem_offset, s3
 	v_add_nc_u32	v0, v0, 128						//mem offset increased by 128 bytes
 	s_cmp_lt_u32	m0, s_save_alloc_size					//scc=(m0 < s_save_alloc_size) ? 1 : 0
@@ -652,14 +746,34 @@ L_SAVE_LDS_LOOP_W32:
 	s_branch	L_SAVE_LDS_DONE
 
 L_SAVE_LDS_W64:
+#if  SAVE_AFTER_XNACK_ERROR
+	check_if_tcp_store_ok()
+	s_cbranch_scc1 L_SAVE_LDS_WITH_TCP_W64
+
+L_SAVE_LDS_LOOP_SQC_W64:
+	ds_read_b32	v1, v0
+	S_WAITCNT_0
+
+	write_vgprs_to_mem_with_sqc_w64(v1, 1, s_save_buf_rsrc0, s_save_mem_offset)
+
+	s_add_u32	m0, m0, 256						//every buffer_store_lds does 256 bytes
+	v_add_nc_u32	v0, v0, 256						//mem offset increased by 256 bytes
+	s_cmp_lt_u32	m0, s_save_alloc_size					//scc=(m0 < s_save_alloc_size) ? 1 : 0
+	s_cbranch_scc1	L_SAVE_LDS_LOOP_SQC_W64					//LDS save is complete?
+
+	s_branch	L_SAVE_LDS_DONE
+
+L_SAVE_LDS_WITH_TCP_W64:
+#endif
+
 	s_mov_b32	s3, 256
 	s_nop		0
 	s_nop		0
 	s_nop		0
 L_SAVE_LDS_LOOP_W64:
 	ds_read_b32	v1, v0
-	s_waitcnt	0
-	buffer_store_dword	v1, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1
+	S_WAITCNT_0
+	buffer_store_dword	v1, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE
 
 	s_add_u32	m0, m0, s3						//every buffer_store_lds does 256 bytes
 	s_add_u32	s_save_mem_offset, s_save_mem_offset, s3
@@ -701,16 +815,35 @@ L_SAVE_VGPR_NORMAL:
 	s_cmp_lt_u32	m0, s_save_alloc_size
 	s_cbranch_scc0	L_SAVE_VGPR_END
 
+#if  SAVE_AFTER_XNACK_ERROR
+	check_if_tcp_store_ok()
+	s_cbranch_scc1 L_SAVE_VGPR_W32_LOOP
+
+L_SAVE_VGPR_LOOP_SQC_W32:
+	v_movrels_b32	v0, v0							//v0 = v[0+m0]
+	v_movrels_b32	v1, v1							//v1 = v[1+m0]
+	v_movrels_b32	v2, v2							//v2 = v[2+m0]
+	v_movrels_b32	v3, v3							//v3 = v[3+m0]
+
+	write_vgprs_to_mem_with_sqc_w32(v0, 4, s_save_buf_rsrc0, s_save_mem_offset)
+
+	s_add_u32 m0, m0, 4
+	s_cmp_lt_u32 m0, s_save_alloc_size
+	s_cbranch_scc1 L_SAVE_VGPR_LOOP_SQC_W32
+
+	s_branch L_SAVE_VGPR_END
+#endif
+
 L_SAVE_VGPR_W32_LOOP:
 	v_movrels_b32	v0, v0							//v0 = v[0+m0]
 	v_movrels_b32	v1, v1							//v1 = v[1+m0]
 	v_movrels_b32	v2, v2							//v2 = v[2+m0]
 	v_movrels_b32	v3, v3							//v3 = v[3+m0]
 
-	buffer_store_dword	v0, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1
-	buffer_store_dword	v1, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:128
-	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:128*2
-	buffer_store_dword	v3, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:128*3
+	buffer_store_dword	v0, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE
+	buffer_store_dword	v1, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE offset:128
+	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE offset:128*2
+	buffer_store_dword	v3, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE offset:128*3
 
 	s_add_u32	m0, m0, 4						//next vgpr index
 	s_add_u32	s_save_mem_offset, s_save_mem_offset, 128*4		//every buffer_store_dword does 128 bytes
@@ -727,16 +860,35 @@ L_SAVE_VGPR_WAVE64:
 	s_cmp_lt_u32	m0, s_save_alloc_size
 	s_cbranch_scc0	L_SAVE_SHARED_VGPR
 
+#if  SAVE_AFTER_XNACK_ERROR
+	check_if_tcp_store_ok()
+	s_cbranch_scc1 L_SAVE_VGPR_W64_LOOP
+
+L_SAVE_VGPR_LOOP_SQC_W64:
+	v_movrels_b32	v0, v0							//v0 = v[0+m0]
+	v_movrels_b32	v1, v1							//v1 = v[1+m0]
+	v_movrels_b32	v2, v2							//v2 = v[2+m0]
+	v_movrels_b32	v3, v3							//v3 = v[3+m0]
+
+	write_vgprs_to_mem_with_sqc_w64(v0, 4, s_save_buf_rsrc0, s_save_mem_offset)
+
+	s_add_u32 m0, m0, 4
+	s_cmp_lt_u32 m0, s_save_alloc_size
+	s_cbranch_scc1 L_SAVE_VGPR_LOOP_SQC_W64
+
+	s_branch L_SAVE_VGPR_END
+#endif
+
 L_SAVE_VGPR_W64_LOOP:
 	v_movrels_b32	v0, v0							//v0 = v[0+m0]
 	v_movrels_b32	v1, v1							//v1 = v[1+m0]
 	v_movrels_b32	v2, v2							//v2 = v[2+m0]
 	v_movrels_b32	v3, v3							//v3 = v[3+m0]
 
-	buffer_store_dword	v0, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1
-	buffer_store_dword	v1, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:256
-	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:256*2
-	buffer_store_dword	v3, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:256*3
+	buffer_store_dword	v0, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE
+	buffer_store_dword	v1, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE offset:256
+	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE offset:256*2
+	buffer_store_dword	v3, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE offset:256*3
 
 	s_add_u32	m0, m0, 4						//next vgpr index
 	s_add_u32	s_save_mem_offset, s_save_mem_offset, 256*4		//every buffer_store_dword does 256 bytes
@@ -754,9 +906,26 @@ L_SAVE_SHARED_VGPR:
 	s_add_u32	s_save_alloc_size, s_save_alloc_size, m0
 	s_mov_b32	exec_lo, 0xFFFFFFFF
 	s_mov_b32	exec_hi, 0x00000000
+
+#if  SAVE_AFTER_XNACK_ERROR
+	check_if_tcp_store_ok()
+	s_cbranch_scc1 L_SAVE_SHARED_VGPR_WAVE64_LOOP
+
+L_SAVE_SHARED_VGPR_WAVE64_LOOP_SQC:
+	v_movrels_b32	v0, v0
+
+	write_vgprs_to_mem_with_sqc_w64(v0, 1, s_save_buf_rsrc0, s_save_mem_offset)
+
+	s_add_u32 m0, m0, 1
+	s_cmp_lt_u32 m0, s_save_alloc_size
+	s_cbranch_scc1 L_SAVE_SHARED_VGPR_WAVE64_LOOP_SQC
+
+	s_branch L_SAVE_VGPR_END
+#endif
+
 L_SAVE_SHARED_VGPR_WAVE64_LOOP:
 	v_movrels_b32	v0, v0							//v0 = v[0+m0]
-	buffer_store_dword	v0, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1
+	buffer_store_dword	v0, v0, s_save_buf_rsrc0, s_save_mem_offset V_COHERENCE
 	s_add_u32	m0, m0, 1						//next vgpr index
 	s_add_u32	s_save_mem_offset, s_save_mem_offset, 128
 	s_cmp_lt_u32	m0, s_save_alloc_size					//scc = (m0 < s_save_alloc_size) ? 1 : 0
@@ -774,7 +943,7 @@ L_RESTORE:
 	s_mov_b32	s_restore_buf_rsrc3, S_RESTORE_BUF_RSRC_WORD3_MISC
 
 	//determine it is wave32 or wave64
-	get_wave_size(s_restore_size)
+	get_wave_size2(s_restore_size)
 
 	s_and_b32	s_restore_tmp, s_restore_spi_init_hi, S_RESTORE_SPI_INIT_FIRST_WAVE_MASK
 	s_cbranch_scc0	L_RESTORE_VGPR
@@ -794,8 +963,7 @@ L_RESTORE_LDS_NORMAL:
 	s_getreg_b32	s_restore_alloc_size, hwreg(HW_REG_LDS_ALLOC,SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT,SQ_WAVE_LDS_ALLOC_LDS_SIZE_SIZE)
 	s_and_b32	s_restore_alloc_size, s_restore_alloc_size, 0xFFFFFFFF	//lds_size is zero?
 	s_cbranch_scc0	L_RESTORE_VGPR						//no lds used? jump to L_RESTORE_VGPR
-	s_lshl_b32	s_restore_alloc_size, s_restore_alloc_size, 6		//LDS size in dwords = lds_size * 64dw
-	s_lshl_b32	s_restore_alloc_size, s_restore_alloc_size, 2		//LDS size in bytes
+	s_lshl_b32	s_restore_alloc_size, s_restore_alloc_size, SQ_WAVE_LDS_ALLOC_GRANULARITY
 	s_mov_b32	s_restore_buf_rsrc2, s_restore_alloc_size		//NUM_RECORDS in bytes
 
 	// LDS at offset: size(VGPR)+size(SVGPR)+SIZE(SGPR)+SIZE(HWREG)
@@ -819,7 +987,7 @@ L_RESTORE_LDS_LOOP_W32:
 	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset lds:1	// first 64DW
 #else
 	buffer_load_dword       v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset
-	s_waitcnt	vmcnt(0)
+	S_WAITCNT_0
 	ds_store_addtid_b32     v0
 #endif
 	s_add_u32	m0, m0, 128						// 128 DW
@@ -833,7 +1001,7 @@ L_RESTORE_LDS_LOOP_W64:
 	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset lds:1	// first 64DW
 #else
 	buffer_load_dword       v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset
-	s_waitcnt	vmcnt(0)
+	S_WAITCNT_0
 	ds_store_addtid_b32     v0
 #endif
 	s_add_u32	m0, m0, 256						// 256 DW
@@ -874,11 +1042,11 @@ L_RESTORE_VGPR_NORMAL:
 	s_cbranch_scc0	L_RESTORE_SGPR
 
 L_RESTORE_VGPR_WAVE32_LOOP:
-	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1
-	buffer_load_dword	v1, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1 offset:128
-	buffer_load_dword	v2, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1 offset:128*2
-	buffer_load_dword	v3, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1 offset:128*3
-	s_waitcnt	vmcnt(0)
+	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset V_COHERENCE
+	buffer_load_dword	v1, v0, s_restore_buf_rsrc0, s_restore_mem_offset V_COHERENCE offset:128
+	buffer_load_dword	v2, v0, s_restore_buf_rsrc0, s_restore_mem_offset V_COHERENCE offset:128*2
+	buffer_load_dword	v3, v0, s_restore_buf_rsrc0, s_restore_mem_offset V_COHERENCE offset:128*3
+	S_WAITCNT_0
 	v_movreld_b32	v0, v0							//v[0+m0] = v0
 	v_movreld_b32	v1, v1
 	v_movreld_b32	v2, v2
@@ -889,11 +1057,11 @@ L_RESTORE_VGPR_WAVE32_LOOP:
 	s_cbranch_scc1	L_RESTORE_VGPR_WAVE32_LOOP				//VGPR restore (except v0) is complete?
 
 	/* VGPR restore on v0 */
-	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save slc:1 glc:1
-	buffer_load_dword	v1, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save slc:1 glc:1 offset:128
-	buffer_load_dword	v2, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save slc:1 glc:1 offset:128*2
-	buffer_load_dword	v3, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save slc:1 glc:1 offset:128*3
-	s_waitcnt	vmcnt(0)
+	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save V_COHERENCE
+	buffer_load_dword	v1, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save V_COHERENCE offset:128
+	buffer_load_dword	v2, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save V_COHERENCE offset:128*2
+	buffer_load_dword	v3, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save V_COHERENCE offset:128*3
+	S_WAITCNT_0
 
 	s_branch	L_RESTORE_SGPR
 
@@ -908,11 +1076,11 @@ L_RESTORE_VGPR_WAVE64:
 	s_cbranch_scc0	L_RESTORE_SHARED_VGPR
 
 L_RESTORE_VGPR_WAVE64_LOOP:
-	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1
-	buffer_load_dword	v1, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1 offset:256
-	buffer_load_dword	v2, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1 offset:256*2
-	buffer_load_dword	v3, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1 offset:256*3
-	s_waitcnt	vmcnt(0)
+	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset V_COHERENCE
+	buffer_load_dword	v1, v0, s_restore_buf_rsrc0, s_restore_mem_offset V_COHERENCE offset:256
+	buffer_load_dword	v2, v0, s_restore_buf_rsrc0, s_restore_mem_offset V_COHERENCE offset:256*2
+	buffer_load_dword	v3, v0, s_restore_buf_rsrc0, s_restore_mem_offset V_COHERENCE offset:256*3
+	S_WAITCNT_0
 	v_movreld_b32	v0, v0							//v[0+m0] = v0
 	v_movreld_b32	v1, v1
 	v_movreld_b32	v2, v2
@@ -934,8 +1102,8 @@ L_RESTORE_SHARED_VGPR:
 	s_mov_b32	exec_lo, 0xFFFFFFFF
 	s_mov_b32	exec_hi, 0x00000000
 L_RESTORE_SHARED_VGPR_WAVE64_LOOP:
-	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1
-	s_waitcnt	vmcnt(0)
+	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset V_COHERENCE
+	S_WAITCNT_0
 	v_movreld_b32	v0, v0							//v[0+m0] = v0
 	s_add_u32	m0, m0, 1						//next vgpr index
 	s_add_u32	s_restore_mem_offset, s_restore_mem_offset, 128
@@ -946,11 +1114,11 @@ L_RESTORE_SHARED_VGPR_WAVE64_LOOP:
 
 	/* VGPR restore on v0 */
 L_RESTORE_V0:
-	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save slc:1 glc:1
-	buffer_load_dword	v1, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save slc:1 glc:1 offset:256
-	buffer_load_dword	v2, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save slc:1 glc:1 offset:256*2
-	buffer_load_dword	v3, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save slc:1 glc:1 offset:256*3
-	s_waitcnt	vmcnt(0)
+	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save V_COHERENCE
+	buffer_load_dword	v1, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save V_COHERENCE offset:256
+	buffer_load_dword	v2, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save V_COHERENCE offset:256*2
+	buffer_load_dword	v3, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save V_COHERENCE offset:256*3
+	S_WAITCNT_0
 
 	/* restore SGPRs */
 	//will be 2+8+16*6
@@ -967,7 +1135,7 @@ L_RESTORE_SGPR:
 	s_mov_b32	m0, s_sgpr_save_num
 
 	read_4sgpr_from_mem(s0, s_restore_buf_rsrc0, s_restore_mem_offset)
-	s_waitcnt	lgkmcnt(0)
+	S_WAITCNT_0
 
 	s_sub_u32	m0, m0, 4						// Restore from S[0] to S[104]
 	s_nop		0							// hazard SALU M0=> S_MOVREL
@@ -976,7 +1144,7 @@ L_RESTORE_SGPR:
 	s_movreld_b64	s2, s2
 
 	read_8sgpr_from_mem(s0, s_restore_buf_rsrc0, s_restore_mem_offset)
-	s_waitcnt	lgkmcnt(0)
+	S_WAITCNT_0
 
 	s_sub_u32	m0, m0, 8						// Restore from S[0] to S[96]
 	s_nop		0							// hazard SALU M0=> S_MOVREL
@@ -988,7 +1156,7 @@ L_RESTORE_SGPR:
 
  L_RESTORE_SGPR_LOOP:
 	read_16sgpr_from_mem(s0, s_restore_buf_rsrc0, s_restore_mem_offset)
-	s_waitcnt	lgkmcnt(0)
+	S_WAITCNT_0
 
 	s_sub_u32	m0, m0, 16						// Restore from S[n] to S[0]
 	s_nop		0							// hazard SALU M0=> S_MOVREL
@@ -1030,12 +1198,12 @@ L_RESTORE_HWREG:
 	read_hwreg_from_mem(s_restore_xnack_mask, s_restore_buf_rsrc0, s_restore_mem_offset)
 	read_hwreg_from_mem(s_restore_mode, s_restore_buf_rsrc0, s_restore_mem_offset)
 	read_hwreg_from_mem(s_restore_flat_scratch, s_restore_buf_rsrc0, s_restore_mem_offset)
-	s_waitcnt	lgkmcnt(0)
+	S_WAITCNT_0
 
 	s_setreg_b32	hwreg(HW_REG_SHADER_FLAT_SCRATCH_LO), s_restore_flat_scratch
 
 	read_hwreg_from_mem(s_restore_flat_scratch, s_restore_buf_rsrc0, s_restore_mem_offset)
-	s_waitcnt	lgkmcnt(0)						//from now on, it is safe to restore STATUS and IB_STS
+	S_WAITCNT_0
 
 	s_setreg_b32	hwreg(HW_REG_SHADER_FLAT_SCRATCH_HI), s_restore_flat_scratch
 
@@ -1043,16 +1211,21 @@ L_RESTORE_HWREG:
 	s_mov_b32	exec_lo, s_restore_exec_lo
 	s_mov_b32	exec_hi, s_restore_exec_hi
 
-	s_and_b32	s_restore_m0, SQ_WAVE_TRAPSTS_PRE_SAVECTX_MASK, s_restore_trapsts
-	s_setreg_b32	hwreg(HW_REG_TRAPSTS, SQ_WAVE_TRAPSTS_PRE_SAVECTX_SHIFT, SQ_WAVE_TRAPSTS_PRE_SAVECTX_SIZE), s_restore_m0
-
 #if HAVE_XNACK
 	s_setreg_b32	hwreg(HW_REG_SHADER_XNACK_MASK), s_restore_xnack_mask
 #endif
 
-	s_and_b32	s_restore_m0, SQ_WAVE_TRAPSTS_POST_SAVECTX_MASK, s_restore_trapsts
-	s_lshr_b32	s_restore_m0, s_restore_m0, SQ_WAVE_TRAPSTS_POST_SAVECTX_SHIFT
-	s_setreg_b32	hwreg(HW_REG_TRAPSTS, SQ_WAVE_TRAPSTS_POST_SAVECTX_SHIFT, SQ_WAVE_TRAPSTS_POST_SAVECTX_SIZE), s_restore_m0
+	// {TRAPSTS/EXCP_FLAG_PRIV}.SAVE_CONTEXT and HOST_TRAP may have changed.
+	// Only restore the other fields to avoid clobbering them.
+	s_setreg_b32	hwreg(S_TRAPSTS_HWREG, 0, S_TRAPSTS_RESTORE_PART_1_SIZE), s_restore_trapsts
+	s_lshr_b32	s_restore_trapsts, s_restore_trapsts, S_TRAPSTS_RESTORE_PART_2_SHIFT
+	s_setreg_b32	hwreg(S_TRAPSTS_HWREG, S_TRAPSTS_RESTORE_PART_2_SHIFT, S_TRAPSTS_RESTORE_PART_2_SIZE), s_restore_trapsts
+
+if S_TRAPSTS_RESTORE_PART_3_SIZE > 0
+	s_lshr_b32	s_restore_trapsts, s_restore_trapsts, S_TRAPSTS_RESTORE_PART_3_SHIFT - S_TRAPSTS_RESTORE_PART_2_SHIFT
+	s_setreg_b32	hwreg(S_TRAPSTS_HWREG, S_TRAPSTS_RESTORE_PART_3_SHIFT, S_TRAPSTS_RESTORE_PART_3_SIZE), s_restore_trapsts
+end
+
 	s_setreg_b32	hwreg(HW_REG_MODE), s_restore_mode
 
 	// Restore trap temporaries 4-11, 13 initialized by SPI debug dispatch logic
@@ -1064,10 +1237,10 @@ L_RESTORE_HWREG:
 	s_add_u32	s_restore_ttmps_lo, s_restore_ttmps_lo, s_restore_buf_rsrc0
 	s_addc_u32	s_restore_ttmps_hi, s_restore_buf_rsrc1, 0x0
 	s_and_b32	s_restore_ttmps_hi, s_restore_ttmps_hi, 0xFFFF
-	s_load_dwordx4	[ttmp4, ttmp5, ttmp6, ttmp7], [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x50 glc:1
-	s_load_dwordx4	[ttmp8, ttmp9, ttmp10, ttmp11], [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x60 glc:1
-	s_load_dword	ttmp13, [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x74 glc:1
-	s_waitcnt	lgkmcnt(0)
+	s_load_dwordx4	[ttmp4, ttmp5, ttmp6, ttmp7], [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x50 S_COHERENCE
+	s_load_dwordx4	[ttmp8, ttmp9, ttmp10, ttmp11], [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x60 S_COHERENCE
+	s_load_dword	ttmp13, [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x74 S_COHERENCE
+	S_WAITCNT_0
 
 #if HAVE_XNACK
 	restore_ib_sts(s_restore_tmp, s_restore_m0)
@@ -1089,11 +1262,12 @@ L_RESTORE_HWREG:
 L_RETURN_WITHOUT_PRIV:
 #endif
 
-	s_setreg_b32	hwreg(HW_REG_STATUS), s_restore_status			// SCC is included, which is changed by previous salu
+	s_setreg_b32	hwreg(S_STATUS_HWREG), s_restore_status			// SCC is included, which is changed by previous salu
+
 	s_rfe_b64	s_restore_pc_lo						//Return to the main shader program and resume execution
 
 L_END_PGM:
-	s_endpgm
+	s_endpgm_saved
 end
 
 function write_hwreg_to_mem(s, s_rsrc, s_mem_offset)
@@ -1104,7 +1278,7 @@ function write_hwreg_to_mem(s, s_rsrc, s_mem_offset)
 #else
 	s_mov_b32	exec_lo, m0
 	s_mov_b32	m0, s_mem_offset
-	s_buffer_store_dword	s, s_rsrc, m0 glc:1
+	s_buffer_store_dword	s, s_rsrc, m0 S_COHERENCE
 	s_add_u32	s_mem_offset, s_mem_offset, 4
 	s_mov_b32	m0, exec_lo
 #endif
@@ -1119,10 +1293,10 @@ function write_16sgpr_to_mem(s, s_rsrc, s_mem_offset)
 		s_add_u32	ttmp13, ttmp13, 0x1
 	end
 #else
-	s_buffer_store_dwordx4	s[0], s_rsrc, 0 glc:1
-	s_buffer_store_dwordx4	s[4], s_rsrc, 16 glc:1
-	s_buffer_store_dwordx4	s[8], s_rsrc, 32 glc:1
-	s_buffer_store_dwordx4	s[12], s_rsrc, 48 glc:1
+	s_buffer_store_dwordx4	s[0], s_rsrc, 0 S_COHERENCE
+	s_buffer_store_dwordx4	s[4], s_rsrc, 16 S_COHERENCE
+	s_buffer_store_dwordx4	s[8], s_rsrc, 32 S_COHERENCE
+	s_buffer_store_dwordx4	s[12], s_rsrc, 48 S_COHERENCE
 	s_add_u32	s_rsrc[0], s_rsrc[0], 4*16
 	s_addc_u32	s_rsrc[1], s_rsrc[1], 0x0
 #endif
@@ -1136,40 +1310,72 @@ function write_12sgpr_to_mem(s, s_rsrc, s_mem_offset)
 		s_add_u32	ttmp13, ttmp13, 0x1
 	end
 #else
-	s_buffer_store_dwordx4	s[0], s_rsrc, 0 glc:1
-	s_buffer_store_dwordx4	s[4], s_rsrc, 16 glc:1
-	s_buffer_store_dwordx4	s[8], s_rsrc, 32 glc:1
+	s_buffer_store_dwordx4	s[0], s_rsrc, 0 S_COHERENCE
+	s_buffer_store_dwordx4	s[4], s_rsrc, 16 S_COHERENCE
+	s_buffer_store_dwordx4	s[8], s_rsrc, 32 S_COHERENCE
 	s_add_u32	s_rsrc[0], s_rsrc[0], 4*12
 	s_addc_u32	s_rsrc[1], s_rsrc[1], 0x0
 #endif
 end
 
 function read_hwreg_from_mem(s, s_rsrc, s_mem_offset)
-	s_buffer_load_dword	s, s_rsrc, s_mem_offset glc:1
+	s_buffer_load_dword	s, s_rsrc, s_mem_offset S_COHERENCE
 	s_add_u32	s_mem_offset, s_mem_offset, 4
 end
 
 function read_16sgpr_from_mem(s, s_rsrc, s_mem_offset)
 	s_sub_u32	s_mem_offset, s_mem_offset, 4*16
-	s_buffer_load_dwordx16	s, s_rsrc, s_mem_offset glc:1
+	s_buffer_load_dwordx16	s, s_rsrc, s_mem_offset S_COHERENCE
 end
 
 function read_8sgpr_from_mem(s, s_rsrc, s_mem_offset)
 	s_sub_u32	s_mem_offset, s_mem_offset, 4*8
-	s_buffer_load_dwordx8	s, s_rsrc, s_mem_offset glc:1
+	s_buffer_load_dwordx8	s, s_rsrc, s_mem_offset S_COHERENCE
 end
 
 function read_4sgpr_from_mem(s, s_rsrc, s_mem_offset)
 	s_sub_u32	s_mem_offset, s_mem_offset, 4*4
-	s_buffer_load_dwordx4	s, s_rsrc, s_mem_offset glc:1
+	s_buffer_load_dwordx4	s, s_rsrc, s_mem_offset S_COHERENCE
 end
 
+#if SAVE_AFTER_XNACK_ERROR
+function check_if_tcp_store_ok
+	// If TRAPSTS.XNACK_ERROR=1 then TCP stores will fail.
+	s_getreg_b32 s_save_tmp, hwreg(HW_REG_TRAPSTS)
+	s_andn2_b32 s_save_tmp, SQ_WAVE_TRAPSTS_XNACK_ERROR_MASK, s_save_tmp
 
-function get_lds_size_bytes(s_lds_size_byte)
-	s_getreg_b32	s_lds_size_byte, hwreg(HW_REG_LDS_ALLOC, SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT, SQ_WAVE_LDS_ALLOC_LDS_SIZE_SIZE)
-	s_lshl_b32	s_lds_size_byte, s_lds_size_byte, 8			//LDS size in dwords = lds_size * 64 *4Bytes // granularity 64DW
+L_TCP_STORE_CHECK_DONE:
 end
 
+function write_vgpr_to_mem_with_sqc(vgpr, n_lanes, s_rsrc, s_mem_offset)
+	s_mov_b32 s4, 0
+
+L_WRITE_VGPR_LANE_LOOP:
+	for var lane = 0; lane < 4; ++lane
+		v_readlane_b32 s[lane], vgpr, s4
+		s_add_u32 s4, s4, 1
+	end
+
+	s_buffer_store_dwordx4 s[0:3], s_rsrc, s_mem_offset glc:1
+
+	s_add_u32 s_mem_offset, s_mem_offset, 0x10
+	s_cmp_eq_u32 s4, n_lanes
+	s_cbranch_scc0 L_WRITE_VGPR_LANE_LOOP
+end
+
+function write_vgprs_to_mem_with_sqc_w32(vgpr0, n_vgprs, s_rsrc, s_mem_offset)
+	for var vgpr = 0; vgpr < n_vgprs; ++vgpr
+		write_vgpr_to_mem_with_sqc(vgpr0[vgpr], 32, s_rsrc, s_mem_offset)
+	end
+end
+
+function write_vgprs_to_mem_with_sqc_w64(vgpr0, n_vgprs, s_rsrc, s_mem_offset)
+	for var vgpr = 0; vgpr < n_vgprs; ++vgpr
+		write_vgpr_to_mem_with_sqc(vgpr0[vgpr], 64, s_rsrc, s_mem_offset)
+	end
+end
+#endif
+
 function get_vgpr_size_bytes(s_vgpr_size_byte, s_size)
 	s_getreg_b32	s_vgpr_size_byte, hwreg(HW_REG_GPR_ALLOC,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SHIFT,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SIZE)
 	s_add_u32	s_vgpr_size_byte, s_vgpr_size_byte, 1
@@ -1195,11 +1401,12 @@ function get_hwreg_size_bytes
 	return 128
 end
 
-function get_wave_size(s_reg)
+function get_wave_size2(s_reg)
 	s_getreg_b32	s_reg, hwreg(HW_REG_IB_STS2,SQ_WAVE_IB_STS2_WAVE64_SHIFT,SQ_WAVE_IB_STS2_WAVE64_SIZE)
 	s_lshl_b32	s_reg, s_reg, S_WAVE_SIZE
 end
 
+#if HAVE_XNACK
 function save_and_clear_ib_sts(tmp1, tmp2)
 	// Preserve and clear scalar XNACK state before issuing scalar loads.
 	// Save IB_STS.REPLAY_W64H[25], RCNT[21:16], FIRST_REPLAY[15] into
@@ -1224,3 +1431,4 @@ function restore_ib_sts(tmp1, tmp2)
 	s_or_b32	tmp1, tmp1, tmp2
 	s_setreg_b32	hwreg(HW_REG_IB_STS), tmp1
 end
+#endif
diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx12.asm b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx12.asm
new file mode 100644
index 000000000000..5a1a1b1f897f
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx12.asm
@@ -0,0 +1,1136 @@
+/*
+ * Copyright 2018 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/* To compile this assembly code:
+ *
+ * gfx12:
+ *   cpp -DASIC_FAMILY=CHIP_GFX12 cwsr_trap_handler_gfx12.asm -P -o gfx12.sp3
+ *   sp3 gfx12.sp3 -hex gfx12.hex
+ */
+
+#define CHIP_GFX12 37
+
+#define SINGLE_STEP_MISSED_WORKAROUND 1	//workaround for lost TRAP_AFTER_INST exception when SAVECTX raised
+#define HAVE_VALU_SGPR_HAZARD (ASIC_FAMILY == CHIP_GFX12)
+
+var SQ_WAVE_STATE_PRIV_BARRIER_COMPLETE_MASK	= 0x4
+var SQ_WAVE_STATE_PRIV_SCC_SHIFT		= 9
+var SQ_WAVE_STATE_PRIV_SYS_PRIO_MASK		= 0xC00
+var SQ_WAVE_STATE_PRIV_HALT_MASK		= 0x4000
+var SQ_WAVE_STATE_PRIV_POISON_ERR_MASK		= 0x8000
+var SQ_WAVE_STATE_PRIV_POISON_ERR_SHIFT		= 15
+var SQ_WAVE_STATUS_WAVE64_SHIFT			= 29
+var SQ_WAVE_STATUS_WAVE64_SIZE			= 1
+var SQ_WAVE_STATUS_NO_VGPRS_SHIFT		= 24
+var SQ_WAVE_STATE_PRIV_ALWAYS_CLEAR_MASK	= SQ_WAVE_STATE_PRIV_SYS_PRIO_MASK|SQ_WAVE_STATE_PRIV_POISON_ERR_MASK
+var S_SAVE_PC_HI_TRAP_ID_MASK			= 0xF0000000
+
+var SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT		= 12
+var SQ_WAVE_LDS_ALLOC_LDS_SIZE_SIZE		= 9
+var SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SIZE		= 8
+var SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SHIFT		= 12
+var SQ_WAVE_LDS_ALLOC_VGPR_SHARED_SIZE_SHIFT	= 24
+var SQ_WAVE_LDS_ALLOC_VGPR_SHARED_SIZE_SIZE	= 4
+var SQ_WAVE_LDS_ALLOC_GRANULARITY		= 9
+
+var SQ_WAVE_EXCP_FLAG_PRIV_ADDR_WATCH_MASK	= 0xF
+var SQ_WAVE_EXCP_FLAG_PRIV_MEM_VIOL_MASK	= 0x10
+var SQ_WAVE_EXCP_FLAG_PRIV_SAVE_CONTEXT_SHIFT	= 5
+var SQ_WAVE_EXCP_FLAG_PRIV_SAVE_CONTEXT_MASK	= 0x20
+var SQ_WAVE_EXCP_FLAG_PRIV_ILLEGAL_INST_MASK	= 0x40
+var SQ_WAVE_EXCP_FLAG_PRIV_ILLEGAL_INST_SHIFT	= 6
+var SQ_WAVE_EXCP_FLAG_PRIV_HOST_TRAP_MASK	= 0x80
+var SQ_WAVE_EXCP_FLAG_PRIV_HOST_TRAP_SHIFT	= 7
+var SQ_WAVE_EXCP_FLAG_PRIV_WAVE_START_MASK	= 0x100
+var SQ_WAVE_EXCP_FLAG_PRIV_WAVE_START_SHIFT	= 8
+var SQ_WAVE_EXCP_FLAG_PRIV_WAVE_END_MASK	= 0x200
+var SQ_WAVE_EXCP_FLAG_PRIV_TRAP_AFTER_INST_MASK	= 0x800
+var SQ_WAVE_TRAP_CTRL_ADDR_WATCH_MASK		= 0x80
+var SQ_WAVE_TRAP_CTRL_TRAP_AFTER_INST_MASK	= 0x200
+
+var SQ_WAVE_EXCP_FLAG_PRIV_NON_MASKABLE_EXCP_MASK= SQ_WAVE_EXCP_FLAG_PRIV_MEM_VIOL_MASK		|\
+						  SQ_WAVE_EXCP_FLAG_PRIV_ILLEGAL_INST_MASK	|\
+						  SQ_WAVE_EXCP_FLAG_PRIV_HOST_TRAP_MASK		|\
+						  SQ_WAVE_EXCP_FLAG_PRIV_WAVE_START_MASK	|\
+						  SQ_WAVE_EXCP_FLAG_PRIV_WAVE_END_MASK		|\
+						  SQ_WAVE_EXCP_FLAG_PRIV_TRAP_AFTER_INST_MASK
+var SQ_WAVE_EXCP_FLAG_PRIV_RESTORE_PART_1_SIZE	= SQ_WAVE_EXCP_FLAG_PRIV_SAVE_CONTEXT_SHIFT
+var SQ_WAVE_EXCP_FLAG_PRIV_RESTORE_PART_2_SHIFT	= SQ_WAVE_EXCP_FLAG_PRIV_ILLEGAL_INST_SHIFT
+var SQ_WAVE_EXCP_FLAG_PRIV_RESTORE_PART_2_SIZE	= SQ_WAVE_EXCP_FLAG_PRIV_HOST_TRAP_SHIFT - SQ_WAVE_EXCP_FLAG_PRIV_ILLEGAL_INST_SHIFT
+var SQ_WAVE_EXCP_FLAG_PRIV_RESTORE_PART_3_SHIFT	= SQ_WAVE_EXCP_FLAG_PRIV_WAVE_START_SHIFT
+var SQ_WAVE_EXCP_FLAG_PRIV_RESTORE_PART_3_SIZE	= 32 - SQ_WAVE_EXCP_FLAG_PRIV_RESTORE_PART_3_SHIFT
+var BARRIER_STATE_SIGNAL_OFFSET			= 16
+var BARRIER_STATE_VALID_OFFSET			= 0
+
+var TTMP11_DEBUG_TRAP_ENABLED_SHIFT		= 23
+var TTMP11_DEBUG_TRAP_ENABLED_MASK		= 0x800000
+
+// SQ_SEL_X/Y/Z/W, BUF_NUM_FORMAT_FLOAT, (0 for MUBUF stride[17:14]
+// when ADD_TID_ENABLE and BUF_DATA_FORMAT_32 for MTBUF), ADD_TID_ENABLE
+var S_SAVE_BUF_RSRC_WORD1_STRIDE		= 0x00040000
+var S_SAVE_BUF_RSRC_WORD3_MISC			= 0x10807FAC
+var S_SAVE_SPI_INIT_FIRST_WAVE_MASK		= 0x04000000
+var S_SAVE_SPI_INIT_FIRST_WAVE_SHIFT		= 26
+
+var S_SAVE_PC_HI_FIRST_WAVE_MASK		= 0x80000000
+var S_SAVE_PC_HI_FIRST_WAVE_SHIFT		= 31
+
+var s_sgpr_save_num				= 108
+
+var s_save_spi_init_lo				= exec_lo
+var s_save_spi_init_hi				= exec_hi
+var s_save_pc_lo				= ttmp0
+var s_save_pc_hi				= ttmp1
+var s_save_exec_lo				= ttmp2
+var s_save_exec_hi				= ttmp3
+var s_save_state_priv				= ttmp12
+var s_save_excp_flag_priv			= ttmp15
+var s_save_xnack_mask				= s_save_excp_flag_priv
+var s_wave_size					= ttmp7
+var s_save_buf_rsrc0				= ttmp8
+var s_save_buf_rsrc1				= ttmp9
+var s_save_buf_rsrc2				= ttmp10
+var s_save_buf_rsrc3				= ttmp11
+var s_save_mem_offset				= ttmp4
+var s_save_alloc_size				= s_save_excp_flag_priv
+var s_save_tmp					= ttmp14
+var s_save_m0					= ttmp5
+var s_save_ttmps_lo				= s_save_tmp
+var s_save_ttmps_hi				= s_save_excp_flag_priv
+
+var S_RESTORE_BUF_RSRC_WORD1_STRIDE		= S_SAVE_BUF_RSRC_WORD1_STRIDE
+var S_RESTORE_BUF_RSRC_WORD3_MISC		= S_SAVE_BUF_RSRC_WORD3_MISC
+
+var S_RESTORE_SPI_INIT_FIRST_WAVE_MASK		= 0x04000000
+var S_RESTORE_SPI_INIT_FIRST_WAVE_SHIFT		= 26
+var S_WAVE_SIZE					= 25
+
+var s_restore_spi_init_lo			= exec_lo
+var s_restore_spi_init_hi			= exec_hi
+var s_restore_mem_offset			= ttmp12
+var s_restore_alloc_size			= ttmp3
+var s_restore_tmp				= ttmp2
+var s_restore_mem_offset_save			= s_restore_tmp
+var s_restore_m0				= s_restore_alloc_size
+var s_restore_mode				= ttmp7
+var s_restore_flat_scratch			= s_restore_tmp
+var s_restore_pc_lo				= ttmp0
+var s_restore_pc_hi				= ttmp1
+var s_restore_exec_lo				= ttmp4
+var s_restore_exec_hi				= ttmp5
+var s_restore_state_priv			= ttmp14
+var s_restore_excp_flag_priv			= ttmp15
+var s_restore_xnack_mask			= ttmp13
+var s_restore_buf_rsrc0				= ttmp8
+var s_restore_buf_rsrc1				= ttmp9
+var s_restore_buf_rsrc2				= ttmp10
+var s_restore_buf_rsrc3				= ttmp11
+var s_restore_size				= ttmp6
+var s_restore_ttmps_lo				= s_restore_tmp
+var s_restore_ttmps_hi				= s_restore_alloc_size
+var s_restore_spi_init_hi_save			= s_restore_exec_hi
+
+shader main
+	asic(DEFAULT)
+	type(CS)
+	wave_size(32)
+
+	s_branch	L_SKIP_RESTORE						//NOT restore. might be a regular trap or save
+
+L_JUMP_TO_RESTORE:
+	s_branch	L_RESTORE
+
+L_SKIP_RESTORE:
+	s_getreg_b32	s_save_state_priv, hwreg(HW_REG_WAVE_STATE_PRIV)	//save STATUS since we will change SCC
+
+	// Clear SPI_PRIO: do not save with elevated priority.
+	// Clear ECC_ERR: prevents SQC store and triggers FATAL_HALT if setreg'd.
+	s_andn2_b32	s_save_state_priv, s_save_state_priv, SQ_WAVE_STATE_PRIV_ALWAYS_CLEAR_MASK
+
+	s_getreg_b32	s_save_excp_flag_priv, hwreg(HW_REG_WAVE_EXCP_FLAG_PRIV)
+
+	s_and_b32       ttmp2, s_save_state_priv, SQ_WAVE_STATE_PRIV_HALT_MASK
+	s_cbranch_scc0	L_NOT_HALTED
+
+L_HALTED:
+	// Host trap may occur while wave is halted.
+	s_and_b32	ttmp2, s_save_excp_flag_priv, SQ_WAVE_EXCP_FLAG_PRIV_HOST_TRAP_MASK
+	s_cbranch_scc1	L_FETCH_2ND_TRAP
+
+L_CHECK_SAVE:
+	s_and_b32	ttmp2, s_save_excp_flag_priv, SQ_WAVE_EXCP_FLAG_PRIV_SAVE_CONTEXT_MASK
+	s_cbranch_scc1	L_SAVE
+
+	// Wave is halted but neither host trap nor SAVECTX is raised.
+	// Caused by instruction fetch memory violation.
+	// Spin wait until context saved to prevent interrupt storm.
+	s_sleep		0x10
+	s_getreg_b32	s_save_excp_flag_priv, hwreg(HW_REG_WAVE_EXCP_FLAG_PRIV)
+	s_branch	L_CHECK_SAVE
+
+L_NOT_HALTED:
+	// Let second-level handle non-SAVECTX exception or trap.
+	// Any concurrent SAVECTX will be handled upon re-entry once halted.
+
+	// Check non-maskable exceptions. memory_violation, illegal_instruction
+	// and xnack_error exceptions always cause the wave to enter the trap
+	// handler.
+	s_and_b32	ttmp2, s_save_excp_flag_priv, SQ_WAVE_EXCP_FLAG_PRIV_NON_MASKABLE_EXCP_MASK
+	s_cbranch_scc1	L_FETCH_2ND_TRAP
+
+	// Check for maskable exceptions in trapsts.excp and trapsts.excp_hi.
+	// Maskable exceptions only cause the wave to enter the trap handler if
+	// their respective bit in mode.excp_en is set.
+	s_getreg_b32	ttmp2, hwreg(HW_REG_WAVE_EXCP_FLAG_USER)
+	s_and_b32	ttmp3, s_save_excp_flag_priv, SQ_WAVE_EXCP_FLAG_PRIV_ADDR_WATCH_MASK
+	s_cbranch_scc0	L_NOT_ADDR_WATCH
+	s_or_b32	ttmp2, ttmp2, SQ_WAVE_TRAP_CTRL_ADDR_WATCH_MASK
+
+L_NOT_ADDR_WATCH:
+	s_getreg_b32	ttmp3, hwreg(HW_REG_WAVE_TRAP_CTRL)
+	s_and_b32	ttmp2, ttmp3, ttmp2
+	s_cbranch_scc1	L_FETCH_2ND_TRAP
+
+L_CHECK_TRAP_ID:
+	// Check trap_id != 0
+	s_and_b32	ttmp2, s_save_pc_hi, S_SAVE_PC_HI_TRAP_ID_MASK
+	s_cbranch_scc1	L_FETCH_2ND_TRAP
+
+#if SINGLE_STEP_MISSED_WORKAROUND
+	// Prioritize single step exception over context save.
+	// Second-level trap will halt wave and RFE, re-entering for SAVECTX.
+	// WAVE_TRAP_CTRL is already in ttmp3.
+	s_and_b32	ttmp3, ttmp3, SQ_WAVE_TRAP_CTRL_TRAP_AFTER_INST_MASK
+	s_cbranch_scc1	L_FETCH_2ND_TRAP
+#endif
+
+	s_and_b32	ttmp2, s_save_excp_flag_priv, SQ_WAVE_EXCP_FLAG_PRIV_SAVE_CONTEXT_MASK
+	s_cbranch_scc1	L_SAVE
+
+L_FETCH_2ND_TRAP:
+	// Read second-level TBA/TMA from first-level TMA and jump if available.
+	// ttmp[2:5] and ttmp12 can be used (others hold SPI-initialized debug data)
+	// ttmp12 holds SQ_WAVE_STATUS
+	s_sendmsg_rtn_b64       [ttmp14, ttmp15], sendmsg(MSG_RTN_GET_TMA)
+	s_wait_idle
+	s_lshl_b64	[ttmp14, ttmp15], [ttmp14, ttmp15], 0x8
+
+	s_bitcmp1_b32	ttmp15, 0xF
+	s_cbranch_scc0	L_NO_SIGN_EXTEND_TMA
+	s_or_b32	ttmp15, ttmp15, 0xFFFF0000
+L_NO_SIGN_EXTEND_TMA:
+
+	s_load_dword    ttmp2, [ttmp14, ttmp15], 0x10 scope:SCOPE_SYS		// debug trap enabled flag
+	s_wait_idle
+	s_lshl_b32      ttmp2, ttmp2, TTMP11_DEBUG_TRAP_ENABLED_SHIFT
+	s_andn2_b32     ttmp11, ttmp11, TTMP11_DEBUG_TRAP_ENABLED_MASK
+	s_or_b32        ttmp11, ttmp11, ttmp2
+
+	s_load_dwordx2	[ttmp2, ttmp3], [ttmp14, ttmp15], 0x0 scope:SCOPE_SYS	// second-level TBA
+	s_wait_idle
+	s_load_dwordx2	[ttmp14, ttmp15], [ttmp14, ttmp15], 0x8 scope:SCOPE_SYS	// second-level TMA
+	s_wait_idle
+
+	s_and_b64	[ttmp2, ttmp3], [ttmp2, ttmp3], [ttmp2, ttmp3]
+	s_cbranch_scc0	L_NO_NEXT_TRAP						// second-level trap handler not been set
+	s_setpc_b64	[ttmp2, ttmp3]						// jump to second-level trap handler
+
+L_NO_NEXT_TRAP:
+	// If not caused by trap then halt wave to prevent re-entry.
+	s_and_b32	ttmp2, s_save_pc_hi, S_SAVE_PC_HI_TRAP_ID_MASK
+	s_cbranch_scc1	L_TRAP_CASE
+
+	// Host trap will not cause trap re-entry.
+	s_getreg_b32	ttmp2, hwreg(HW_REG_WAVE_EXCP_FLAG_PRIV)
+	s_and_b32	ttmp2, ttmp2, SQ_WAVE_EXCP_FLAG_PRIV_HOST_TRAP_MASK
+	s_cbranch_scc1	L_EXIT_TRAP
+	s_or_b32	s_save_state_priv, s_save_state_priv, SQ_WAVE_STATE_PRIV_HALT_MASK
+
+	// If the PC points to S_ENDPGM then context save will fail if STATE_PRIV.HALT is set.
+	// Rewind the PC to prevent this from occurring.
+	s_sub_u32	ttmp0, ttmp0, 0x8
+	s_subb_u32	ttmp1, ttmp1, 0x0
+
+	s_branch	L_EXIT_TRAP
+
+L_TRAP_CASE:
+	// Advance past trap instruction to prevent re-entry.
+	s_add_u32	ttmp0, ttmp0, 0x4
+	s_addc_u32	ttmp1, ttmp1, 0x0
+
+L_EXIT_TRAP:
+	s_and_b32	ttmp1, ttmp1, 0xFFFF
+
+	// Restore SQ_WAVE_STATUS.
+	s_and_b64	exec, exec, exec					// Restore STATUS.EXECZ, not writable by s_setreg_b32
+	s_and_b64	vcc, vcc, vcc						// Restore STATUS.VCCZ, not writable by s_setreg_b32
+
+	// STATE_PRIV.BARRIER_COMPLETE may have changed since we read it.
+	// Only restore fields which the trap handler changes.
+	s_lshr_b32	s_save_state_priv, s_save_state_priv, SQ_WAVE_STATE_PRIV_SCC_SHIFT
+	s_setreg_b32	hwreg(HW_REG_WAVE_STATE_PRIV, SQ_WAVE_STATE_PRIV_SCC_SHIFT, \
+		SQ_WAVE_STATE_PRIV_POISON_ERR_SHIFT - SQ_WAVE_STATE_PRIV_SCC_SHIFT + 1), s_save_state_priv
+
+	s_rfe_b64	[ttmp0, ttmp1]
+
+L_SAVE:
+	// If VGPRs have been deallocated then terminate the wavefront.
+	// It has no remaining program to run and cannot save without VGPRs.
+	s_getreg_b32	s_save_tmp, hwreg(HW_REG_WAVE_STATUS)
+	s_bitcmp1_b32	s_save_tmp, SQ_WAVE_STATUS_NO_VGPRS_SHIFT
+	s_cbranch_scc0	L_HAVE_VGPRS
+	s_endpgm
+L_HAVE_VGPRS:
+
+	s_and_b32	s_save_pc_hi, s_save_pc_hi, 0x0000ffff			//pc[47:32]
+	s_mov_b32	s_save_tmp, 0
+	s_setreg_b32	hwreg(HW_REG_WAVE_EXCP_FLAG_PRIV, SQ_WAVE_EXCP_FLAG_PRIV_SAVE_CONTEXT_SHIFT, 1), s_save_tmp	//clear saveCtx bit
+
+	/* inform SPI the readiness and wait for SPI's go signal */
+	s_mov_b32	s_save_exec_lo, exec_lo					//save EXEC and use EXEC for the go signal from SPI
+	s_mov_b32	s_save_exec_hi, exec_hi
+	s_mov_b64	exec, 0x0						//clear EXEC to get ready to receive
+
+	s_sendmsg_rtn_b64       [exec_lo, exec_hi], sendmsg(MSG_RTN_SAVE_WAVE)
+	s_wait_idle
+
+	// Save first_wave flag so we can clear high bits of save address.
+	s_and_b32	s_save_tmp, s_save_spi_init_hi, S_SAVE_SPI_INIT_FIRST_WAVE_MASK
+	s_lshl_b32	s_save_tmp, s_save_tmp, (S_SAVE_PC_HI_FIRST_WAVE_SHIFT - S_SAVE_SPI_INIT_FIRST_WAVE_SHIFT)
+	s_or_b32	s_save_pc_hi, s_save_pc_hi, s_save_tmp
+
+	// Trap temporaries must be saved via VGPR but all VGPRs are in use.
+	// There is no ttmp space to hold the resource constant for VGPR save.
+	// Save v0 by itself since it requires only two SGPRs.
+	s_mov_b32	s_save_ttmps_lo, exec_lo
+	s_and_b32	s_save_ttmps_hi, exec_hi, 0xFFFF
+	s_mov_b32	exec_lo, 0xFFFFFFFF
+	s_mov_b32	exec_hi, 0xFFFFFFFF
+	global_store_dword_addtid	v0, [s_save_ttmps_lo, s_save_ttmps_hi] scope:SCOPE_SYS
+	v_mov_b32	v0, 0x0
+	s_mov_b32	exec_lo, s_save_ttmps_lo
+	s_mov_b32	exec_hi, s_save_ttmps_hi
+
+	// Save trap temporaries 4-11, 13 initialized by SPI debug dispatch logic
+	// ttmp SR memory offset : size(VGPR)+size(SVGPR)+size(SGPR)+0x40
+	get_wave_size2(s_save_ttmps_hi)
+	get_vgpr_size_bytes(s_save_ttmps_lo, s_save_ttmps_hi)
+	get_svgpr_size_bytes(s_save_ttmps_hi)
+	s_add_u32	s_save_ttmps_lo, s_save_ttmps_lo, s_save_ttmps_hi
+	s_and_b32	s_save_ttmps_hi, s_save_spi_init_hi, 0xFFFF
+	s_add_u32	s_save_ttmps_lo, s_save_ttmps_lo, get_sgpr_size_bytes()
+	s_add_u32	s_save_ttmps_lo, s_save_ttmps_lo, s_save_spi_init_lo
+	s_addc_u32	s_save_ttmps_hi, s_save_ttmps_hi, 0x0
+
+	v_writelane_b32	v0, ttmp4, 0x4
+	v_writelane_b32	v0, ttmp5, 0x5
+	v_writelane_b32	v0, ttmp6, 0x6
+	v_writelane_b32	v0, ttmp7, 0x7
+	v_writelane_b32	v0, ttmp8, 0x8
+	v_writelane_b32	v0, ttmp9, 0x9
+	v_writelane_b32	v0, ttmp10, 0xA
+	v_writelane_b32	v0, ttmp11, 0xB
+	v_writelane_b32	v0, ttmp13, 0xD
+	v_writelane_b32	v0, exec_lo, 0xE
+	v_writelane_b32	v0, exec_hi, 0xF
+	valu_sgpr_hazard()
+
+	s_mov_b32	exec_lo, 0x3FFF
+	s_mov_b32	exec_hi, 0x0
+	global_store_dword_addtid	v0, [s_save_ttmps_lo, s_save_ttmps_hi] offset:0x40 scope:SCOPE_SYS
+	v_readlane_b32	ttmp14, v0, 0xE
+	v_readlane_b32	ttmp15, v0, 0xF
+	s_mov_b32	exec_lo, ttmp14
+	s_mov_b32	exec_hi, ttmp15
+
+	/* setup Resource Contants */
+	s_mov_b32	s_save_buf_rsrc0, s_save_spi_init_lo			//base_addr_lo
+	s_and_b32	s_save_buf_rsrc1, s_save_spi_init_hi, 0x0000FFFF	//base_addr_hi
+	s_or_b32	s_save_buf_rsrc1, s_save_buf_rsrc1, S_SAVE_BUF_RSRC_WORD1_STRIDE
+	s_mov_b32	s_save_buf_rsrc2, 0					//NUM_RECORDS initial value = 0 (in bytes) although not neccessarily inited
+	s_mov_b32	s_save_buf_rsrc3, S_SAVE_BUF_RSRC_WORD3_MISC
+
+	s_mov_b32	s_save_m0, m0
+
+	/* global mem offset */
+	s_mov_b32	s_save_mem_offset, 0x0
+	get_wave_size2(s_wave_size)
+
+	/* save first 4 VGPRs, needed for SGPR save */
+	s_mov_b32	exec_lo, 0xFFFFFFFF					//need every thread from now on
+	s_lshr_b32	m0, s_wave_size, S_WAVE_SIZE
+	s_and_b32	m0, m0, 1
+	s_cmp_eq_u32	m0, 1
+	s_cbranch_scc1	L_ENABLE_SAVE_4VGPR_EXEC_HI
+	s_mov_b32	exec_hi, 0x00000000
+	s_branch	L_SAVE_4VGPR_WAVE32
+L_ENABLE_SAVE_4VGPR_EXEC_HI:
+	s_mov_b32	exec_hi, 0xFFFFFFFF
+	s_branch	L_SAVE_4VGPR_WAVE64
+L_SAVE_4VGPR_WAVE32:
+	s_mov_b32	s_save_buf_rsrc2, 0x1000000				//NUM_RECORDS in bytes
+
+	// VGPR Allocated in 4-GPR granularity
+
+	buffer_store_dword	v1, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS offset:128
+	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS offset:128*2
+	buffer_store_dword	v3, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS offset:128*3
+	s_branch	L_SAVE_HWREG
+
+L_SAVE_4VGPR_WAVE64:
+	s_mov_b32	s_save_buf_rsrc2, 0x1000000				//NUM_RECORDS in bytes
+
+	// VGPR Allocated in 4-GPR granularity
+
+	buffer_store_dword	v1, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS offset:256
+	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS offset:256*2
+	buffer_store_dword	v3, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS offset:256*3
+
+	/* save HW registers */
+
+L_SAVE_HWREG:
+	// HWREG SR memory offset : size(VGPR)+size(SVGPR)+size(SGPR)
+	get_vgpr_size_bytes(s_save_mem_offset, s_wave_size)
+	get_svgpr_size_bytes(s_save_tmp)
+	s_add_u32	s_save_mem_offset, s_save_mem_offset, s_save_tmp
+	s_add_u32	s_save_mem_offset, s_save_mem_offset, get_sgpr_size_bytes()
+
+	s_mov_b32	s_save_buf_rsrc2, 0x1000000				//NUM_RECORDS in bytes
+
+	v_mov_b32	v0, 0x0							//Offset[31:0] from buffer resource
+	v_mov_b32	v1, 0x0							//Offset[63:32] from buffer resource
+	v_mov_b32	v2, 0x0							//Set of SGPRs for TCP store
+
+	// Ensure no further changes to barrier or LDS state.
+	// STATE_PRIV.BARRIER_COMPLETE may change up to this point.
+	s_barrier_signal	-2
+	s_barrier_wait	-2
+
+	// Re-read final state of BARRIER_COMPLETE field for save.
+	s_getreg_b32	s_save_tmp, hwreg(HW_REG_WAVE_STATE_PRIV)
+	s_and_b32	s_save_tmp, s_save_tmp, SQ_WAVE_STATE_PRIV_BARRIER_COMPLETE_MASK
+	s_andn2_b32	s_save_state_priv, s_save_state_priv, SQ_WAVE_STATE_PRIV_BARRIER_COMPLETE_MASK
+	s_or_b32	s_save_state_priv, s_save_state_priv, s_save_tmp
+
+	s_andn2_b32	s_save_tmp, s_save_pc_hi, S_SAVE_PC_HI_FIRST_WAVE_MASK
+	v_writelane_b32	v2, s_save_m0, 0x0
+	v_writelane_b32	v2, s_save_pc_lo, 0x1
+	v_writelane_b32	v2, s_save_tmp, 0x2
+	v_writelane_b32	v2, s_save_exec_lo, 0x3
+	v_writelane_b32	v2, s_save_exec_hi, 0x4
+	v_writelane_b32	v2, s_save_state_priv, 0x5
+	v_writelane_b32	v2, s_save_xnack_mask, 0x7
+	valu_sgpr_hazard()
+
+	s_getreg_b32	s_save_tmp, hwreg(HW_REG_WAVE_EXCP_FLAG_PRIV)
+	v_writelane_b32	v2, s_save_tmp, 0x6
+
+	s_getreg_b32	s_save_tmp, hwreg(HW_REG_WAVE_MODE)
+	v_writelane_b32	v2, s_save_tmp, 0x8
+
+	s_getreg_b32	s_save_tmp, hwreg(HW_REG_WAVE_SCRATCH_BASE_LO)
+	v_writelane_b32	v2, s_save_tmp, 0x9
+
+	s_getreg_b32	s_save_tmp, hwreg(HW_REG_WAVE_SCRATCH_BASE_HI)
+	v_writelane_b32	v2, s_save_tmp, 0xA
+
+	s_getreg_b32	s_save_tmp, hwreg(HW_REG_WAVE_EXCP_FLAG_USER)
+	v_writelane_b32	v2, s_save_tmp, 0xB
+
+	s_getreg_b32	s_save_tmp, hwreg(HW_REG_WAVE_TRAP_CTRL)
+	v_writelane_b32	v2, s_save_tmp, 0xC
+
+	s_getreg_b32	s_save_tmp, hwreg(HW_REG_WAVE_STATUS)
+	v_writelane_b32	v2, s_save_tmp, 0xD
+
+	s_get_barrier_state s_save_tmp, -1
+	s_wait_kmcnt (0)
+	v_writelane_b32	v2, s_save_tmp, 0xE
+	valu_sgpr_hazard()
+
+	// Write HWREGs with 16 VGPR lanes. TTMPs occupy space after this.
+	s_mov_b32       exec_lo, 0xFFFF
+	s_mov_b32	exec_hi, 0x0
+	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS
+
+	// Write SGPRs with 32 VGPR lanes. This works in wave32 and wave64 mode.
+	s_mov_b32       exec_lo, 0xFFFFFFFF
+
+	/* save SGPRs */
+	// Save SGPR before LDS save, then the s0 to s4 can be used during LDS save...
+
+	// SGPR SR memory offset : size(VGPR)+size(SVGPR)
+	get_vgpr_size_bytes(s_save_mem_offset, s_wave_size)
+	get_svgpr_size_bytes(s_save_tmp)
+	s_add_u32	s_save_mem_offset, s_save_mem_offset, s_save_tmp
+	s_mov_b32	s_save_buf_rsrc2, 0x1000000				//NUM_RECORDS in bytes
+
+	s_mov_b32	ttmp13, 0x0						//next VGPR lane to copy SGPR into
+
+	s_mov_b32	m0, 0x0							//SGPR initial index value =0
+	s_nop		0x0							//Manually inserted wait states
+L_SAVE_SGPR_LOOP:
+	// SGPR is allocated in 16 SGPR granularity
+	s_movrels_b64	s0, s0							//s0 = s[0+m0], s1 = s[1+m0]
+	s_movrels_b64	s2, s2							//s2 = s[2+m0], s3 = s[3+m0]
+	s_movrels_b64	s4, s4							//s4 = s[4+m0], s5 = s[5+m0]
+	s_movrels_b64	s6, s6							//s6 = s[6+m0], s7 = s[7+m0]
+	s_movrels_b64	s8, s8							//s8 = s[8+m0], s9 = s[9+m0]
+	s_movrels_b64	s10, s10						//s10 = s[10+m0], s11 = s[11+m0]
+	s_movrels_b64	s12, s12						//s12 = s[12+m0], s13 = s[13+m0]
+	s_movrels_b64	s14, s14						//s14 = s[14+m0], s15 = s[15+m0]
+
+	s_cmp_eq_u32	ttmp13, 0x0
+	s_cbranch_scc0	L_WRITE_V2_SECOND_HALF
+	write_16sgpr_to_v2(s0, 0x0)
+	s_branch	L_SAVE_SGPR_SKIP_TCP_STORE
+L_WRITE_V2_SECOND_HALF:
+	write_16sgpr_to_v2(s0, 0x10)
+
+	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS
+	s_add_u32	s_save_mem_offset, s_save_mem_offset, 0x80
+	s_mov_b32	ttmp13, 0x0
+	v_mov_b32	v2, 0x0
+L_SAVE_SGPR_SKIP_TCP_STORE:
+
+	s_add_u32	m0, m0, 16						//next sgpr index
+	s_cmp_lt_u32	m0, 96							//scc = (m0 < first 96 SGPR) ? 1 : 0
+	s_cbranch_scc1	L_SAVE_SGPR_LOOP					//first 96 SGPR save is complete?
+
+	//save the rest 12 SGPR
+	s_movrels_b64	s0, s0							//s0 = s[0+m0], s1 = s[1+m0]
+	s_movrels_b64	s2, s2							//s2 = s[2+m0], s3 = s[3+m0]
+	s_movrels_b64	s4, s4							//s4 = s[4+m0], s5 = s[5+m0]
+	s_movrels_b64	s6, s6							//s6 = s[6+m0], s7 = s[7+m0]
+	s_movrels_b64	s8, s8							//s8 = s[8+m0], s9 = s[9+m0]
+	s_movrels_b64	s10, s10						//s10 = s[10+m0], s11 = s[11+m0]
+	write_12sgpr_to_v2(s0)
+
+	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS
+
+	/* save LDS */
+
+L_SAVE_LDS:
+	// Change EXEC to all threads...
+	s_mov_b32	exec_lo, 0xFFFFFFFF					//need every thread from now on
+	s_lshr_b32	m0, s_wave_size, S_WAVE_SIZE
+	s_and_b32	m0, m0, 1
+	s_cmp_eq_u32	m0, 1
+	s_cbranch_scc1	L_ENABLE_SAVE_LDS_EXEC_HI
+	s_mov_b32	exec_hi, 0x00000000
+	s_branch	L_SAVE_LDS_NORMAL
+L_ENABLE_SAVE_LDS_EXEC_HI:
+	s_mov_b32	exec_hi, 0xFFFFFFFF
+L_SAVE_LDS_NORMAL:
+	s_getreg_b32	s_save_alloc_size, hwreg(HW_REG_WAVE_LDS_ALLOC,SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT,SQ_WAVE_LDS_ALLOC_LDS_SIZE_SIZE)
+	s_and_b32	s_save_alloc_size, s_save_alloc_size, 0xFFFFFFFF	//lds_size is zero?
+	s_cbranch_scc0	L_SAVE_LDS_DONE						//no lds used? jump to L_SAVE_DONE
+
+	s_and_b32	s_save_tmp, s_save_pc_hi, S_SAVE_PC_HI_FIRST_WAVE_MASK
+	s_cbranch_scc0	L_SAVE_LDS_DONE
+
+	// first wave do LDS save;
+
+	s_lshl_b32	s_save_alloc_size, s_save_alloc_size, SQ_WAVE_LDS_ALLOC_GRANULARITY
+	s_mov_b32	s_save_buf_rsrc2, s_save_alloc_size			//NUM_RECORDS in bytes
+
+	// LDS at offset: size(VGPR)+size(SVGPR)+SIZE(SGPR)+SIZE(HWREG)
+	//
+	get_vgpr_size_bytes(s_save_mem_offset, s_wave_size)
+	get_svgpr_size_bytes(s_save_tmp)
+	s_add_u32	s_save_mem_offset, s_save_mem_offset, s_save_tmp
+	s_add_u32	s_save_mem_offset, s_save_mem_offset, get_sgpr_size_bytes()
+	s_add_u32	s_save_mem_offset, s_save_mem_offset, get_hwreg_size_bytes()
+
+	s_mov_b32	s_save_buf_rsrc2, 0x1000000				//NUM_RECORDS in bytes
+
+	//load 0~63*4(byte address) to vgpr v0
+	v_mbcnt_lo_u32_b32	v0, -1, 0
+	v_mbcnt_hi_u32_b32	v0, -1, v0
+	v_mul_u32_u24	v0, 4, v0
+
+	s_lshr_b32	m0, s_wave_size, S_WAVE_SIZE
+	s_and_b32	m0, m0, 1
+	s_cmp_eq_u32	m0, 1
+	s_mov_b32	m0, 0x0
+	s_cbranch_scc1	L_SAVE_LDS_W64
+
+L_SAVE_LDS_W32:
+	s_mov_b32	s3, 128
+	s_nop		0
+	s_nop		0
+	s_nop		0
+L_SAVE_LDS_LOOP_W32:
+	ds_read_b32	v1, v0
+	s_wait_idle
+	buffer_store_dword	v1, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS
+
+	s_add_u32	m0, m0, s3						//every buffer_store_lds does 128 bytes
+	s_add_u32	s_save_mem_offset, s_save_mem_offset, s3
+	v_add_nc_u32	v0, v0, 128						//mem offset increased by 128 bytes
+	s_cmp_lt_u32	m0, s_save_alloc_size					//scc=(m0 < s_save_alloc_size) ? 1 : 0
+	s_cbranch_scc1	L_SAVE_LDS_LOOP_W32					//LDS save is complete?
+
+	s_branch	L_SAVE_LDS_DONE
+
+L_SAVE_LDS_W64:
+	s_mov_b32	s3, 256
+	s_nop		0
+	s_nop		0
+	s_nop		0
+L_SAVE_LDS_LOOP_W64:
+	ds_read_b32	v1, v0
+	s_wait_idle
+	buffer_store_dword	v1, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS
+
+	s_add_u32	m0, m0, s3						//every buffer_store_lds does 256 bytes
+	s_add_u32	s_save_mem_offset, s_save_mem_offset, s3
+	v_add_nc_u32	v0, v0, 256						//mem offset increased by 256 bytes
+	s_cmp_lt_u32	m0, s_save_alloc_size					//scc=(m0 < s_save_alloc_size) ? 1 : 0
+	s_cbranch_scc1	L_SAVE_LDS_LOOP_W64					//LDS save is complete?
+
+L_SAVE_LDS_DONE:
+	/* save VGPRs  - set the Rest VGPRs */
+L_SAVE_VGPR:
+	// VGPR SR memory offset: 0
+	s_mov_b32	exec_lo, 0xFFFFFFFF					//need every thread from now on
+	s_lshr_b32	m0, s_wave_size, S_WAVE_SIZE
+	s_and_b32	m0, m0, 1
+	s_cmp_eq_u32	m0, 1
+	s_cbranch_scc1	L_ENABLE_SAVE_VGPR_EXEC_HI
+	s_mov_b32	s_save_mem_offset, (0+128*4)				// for the rest VGPRs
+	s_mov_b32	exec_hi, 0x00000000
+	s_branch	L_SAVE_VGPR_NORMAL
+L_ENABLE_SAVE_VGPR_EXEC_HI:
+	s_mov_b32	s_save_mem_offset, (0+256*4)				// for the rest VGPRs
+	s_mov_b32	exec_hi, 0xFFFFFFFF
+L_SAVE_VGPR_NORMAL:
+	s_getreg_b32	s_save_alloc_size, hwreg(HW_REG_WAVE_GPR_ALLOC,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SHIFT,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SIZE)
+	s_add_u32	s_save_alloc_size, s_save_alloc_size, 1
+	s_lshl_b32	s_save_alloc_size, s_save_alloc_size, 2			//Number of VGPRs = (vgpr_size + 1) * 4    (non-zero value)
+	//determine it is wave32 or wave64
+	s_lshr_b32	m0, s_wave_size, S_WAVE_SIZE
+	s_and_b32	m0, m0, 1
+	s_cmp_eq_u32	m0, 1
+	s_cbranch_scc1	L_SAVE_VGPR_WAVE64
+
+	s_mov_b32	s_save_buf_rsrc2, 0x1000000				//NUM_RECORDS in bytes
+
+	// VGPR Allocated in 4-GPR granularity
+
+	// VGPR store using dw burst
+	s_mov_b32	m0, 0x4							//VGPR initial index value =4
+	s_cmp_lt_u32	m0, s_save_alloc_size
+	s_cbranch_scc0	L_SAVE_VGPR_END
+
+L_SAVE_VGPR_W32_LOOP:
+	v_movrels_b32	v0, v0							//v0 = v[0+m0]
+	v_movrels_b32	v1, v1							//v1 = v[1+m0]
+	v_movrels_b32	v2, v2							//v2 = v[2+m0]
+	v_movrels_b32	v3, v3							//v3 = v[3+m0]
+
+	buffer_store_dword	v0, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS
+	buffer_store_dword	v1, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS offset:128
+	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS offset:128*2
+	buffer_store_dword	v3, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS offset:128*3
+
+	s_add_u32	m0, m0, 4						//next vgpr index
+	s_add_u32	s_save_mem_offset, s_save_mem_offset, 128*4		//every buffer_store_dword does 128 bytes
+	s_cmp_lt_u32	m0, s_save_alloc_size					//scc = (m0 < s_save_alloc_size) ? 1 : 0
+	s_cbranch_scc1	L_SAVE_VGPR_W32_LOOP					//VGPR save is complete?
+
+	s_branch	L_SAVE_VGPR_END
+
+L_SAVE_VGPR_WAVE64:
+	s_mov_b32	s_save_buf_rsrc2, 0x1000000				//NUM_RECORDS in bytes
+
+	// VGPR store using dw burst
+	s_mov_b32	m0, 0x4							//VGPR initial index value =4
+	s_cmp_lt_u32	m0, s_save_alloc_size
+	s_cbranch_scc0	L_SAVE_SHARED_VGPR
+
+L_SAVE_VGPR_W64_LOOP:
+	v_movrels_b32	v0, v0							//v0 = v[0+m0]
+	v_movrels_b32	v1, v1							//v1 = v[1+m0]
+	v_movrels_b32	v2, v2							//v2 = v[2+m0]
+	v_movrels_b32	v3, v3							//v3 = v[3+m0]
+
+	buffer_store_dword	v0, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS
+	buffer_store_dword	v1, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS offset:256
+	buffer_store_dword	v2, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS offset:256*2
+	buffer_store_dword	v3, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS offset:256*3
+
+	s_add_u32	m0, m0, 4						//next vgpr index
+	s_add_u32	s_save_mem_offset, s_save_mem_offset, 256*4		//every buffer_store_dword does 256 bytes
+	s_cmp_lt_u32	m0, s_save_alloc_size					//scc = (m0 < s_save_alloc_size) ? 1 : 0
+	s_cbranch_scc1	L_SAVE_VGPR_W64_LOOP					//VGPR save is complete?
+
+L_SAVE_SHARED_VGPR:
+	s_getreg_b32	s_save_alloc_size, hwreg(HW_REG_WAVE_LDS_ALLOC,SQ_WAVE_LDS_ALLOC_VGPR_SHARED_SIZE_SHIFT,SQ_WAVE_LDS_ALLOC_VGPR_SHARED_SIZE_SIZE)
+	s_and_b32	s_save_alloc_size, s_save_alloc_size, 0xFFFFFFFF	//shared_vgpr_size is zero?
+	s_cbranch_scc0	L_SAVE_VGPR_END						//no shared_vgpr used? jump to L_SAVE_LDS
+	s_lshl_b32	s_save_alloc_size, s_save_alloc_size, 3			//Number of SHARED_VGPRs = shared_vgpr_size * 8    (non-zero value)
+	//m0 now has the value of normal vgpr count, just add the m0 with shared_vgpr count to get the total count.
+	//save shared_vgpr will start from the index of m0
+	s_add_u32	s_save_alloc_size, s_save_alloc_size, m0
+	s_mov_b32	exec_lo, 0xFFFFFFFF
+	s_mov_b32	exec_hi, 0x00000000
+
+L_SAVE_SHARED_VGPR_WAVE64_LOOP:
+	v_movrels_b32	v0, v0							//v0 = v[0+m0]
+	buffer_store_dword	v0, v0, s_save_buf_rsrc0, s_save_mem_offset scope:SCOPE_SYS
+	s_add_u32	m0, m0, 1						//next vgpr index
+	s_add_u32	s_save_mem_offset, s_save_mem_offset, 128
+	s_cmp_lt_u32	m0, s_save_alloc_size					//scc = (m0 < s_save_alloc_size) ? 1 : 0
+	s_cbranch_scc1	L_SAVE_SHARED_VGPR_WAVE64_LOOP				//SHARED_VGPR save is complete?
+
+L_SAVE_VGPR_END:
+	s_branch	L_END_PGM
+
+L_RESTORE:
+	/* Setup Resource Contants */
+	s_mov_b32	s_restore_buf_rsrc0, s_restore_spi_init_lo		//base_addr_lo
+	s_and_b32	s_restore_buf_rsrc1, s_restore_spi_init_hi, 0x0000FFFF	//base_addr_hi
+	s_or_b32	s_restore_buf_rsrc1, s_restore_buf_rsrc1, S_RESTORE_BUF_RSRC_WORD1_STRIDE
+	s_mov_b32	s_restore_buf_rsrc2, 0					//NUM_RECORDS initial value = 0 (in bytes)
+	s_mov_b32	s_restore_buf_rsrc3, S_RESTORE_BUF_RSRC_WORD3_MISC
+
+	// Save s_restore_spi_init_hi for later use.
+	s_mov_b32 s_restore_spi_init_hi_save, s_restore_spi_init_hi
+
+	//determine it is wave32 or wave64
+	get_wave_size2(s_restore_size)
+
+	s_and_b32	s_restore_tmp, s_restore_spi_init_hi, S_RESTORE_SPI_INIT_FIRST_WAVE_MASK
+	s_cbranch_scc0	L_RESTORE_VGPR
+
+	/* restore LDS */
+L_RESTORE_LDS:
+	s_mov_b32	exec_lo, 0xFFFFFFFF					//need every thread from now on
+	s_lshr_b32	m0, s_restore_size, S_WAVE_SIZE
+	s_and_b32	m0, m0, 1
+	s_cmp_eq_u32	m0, 1
+	s_cbranch_scc1	L_ENABLE_RESTORE_LDS_EXEC_HI
+	s_mov_b32	exec_hi, 0x00000000
+	s_branch	L_RESTORE_LDS_NORMAL
+L_ENABLE_RESTORE_LDS_EXEC_HI:
+	s_mov_b32	exec_hi, 0xFFFFFFFF
+L_RESTORE_LDS_NORMAL:
+	s_getreg_b32	s_restore_alloc_size, hwreg(HW_REG_WAVE_LDS_ALLOC,SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT,SQ_WAVE_LDS_ALLOC_LDS_SIZE_SIZE)
+	s_and_b32	s_restore_alloc_size, s_restore_alloc_size, 0xFFFFFFFF	//lds_size is zero?
+	s_cbranch_scc0	L_RESTORE_VGPR						//no lds used? jump to L_RESTORE_VGPR
+	s_lshl_b32	s_restore_alloc_size, s_restore_alloc_size, SQ_WAVE_LDS_ALLOC_GRANULARITY
+	s_mov_b32	s_restore_buf_rsrc2, s_restore_alloc_size		//NUM_RECORDS in bytes
+
+	// LDS at offset: size(VGPR)+size(SVGPR)+SIZE(SGPR)+SIZE(HWREG)
+	//
+	get_vgpr_size_bytes(s_restore_mem_offset, s_restore_size)
+	get_svgpr_size_bytes(s_restore_tmp)
+	s_add_u32	s_restore_mem_offset, s_restore_mem_offset, s_restore_tmp
+	s_add_u32	s_restore_mem_offset, s_restore_mem_offset, get_sgpr_size_bytes()
+	s_add_u32	s_restore_mem_offset, s_restore_mem_offset, get_hwreg_size_bytes()
+
+	s_mov_b32	s_restore_buf_rsrc2, 0x1000000				//NUM_RECORDS in bytes
+
+	s_lshr_b32	m0, s_restore_size, S_WAVE_SIZE
+	s_and_b32	m0, m0, 1
+	s_cmp_eq_u32	m0, 1
+	s_mov_b32	m0, 0x0
+	s_cbranch_scc1	L_RESTORE_LDS_LOOP_W64
+
+L_RESTORE_LDS_LOOP_W32:
+	buffer_load_dword       v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset
+	s_wait_idle
+	ds_store_addtid_b32     v0
+	s_add_u32	m0, m0, 128						// 128 DW
+	s_add_u32	s_restore_mem_offset, s_restore_mem_offset, 128		//mem offset increased by 128DW
+	s_cmp_lt_u32	m0, s_restore_alloc_size				//scc=(m0 < s_restore_alloc_size) ? 1 : 0
+	s_cbranch_scc1	L_RESTORE_LDS_LOOP_W32					//LDS restore is complete?
+	s_branch	L_RESTORE_VGPR
+
+L_RESTORE_LDS_LOOP_W64:
+	buffer_load_dword       v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset
+	s_wait_idle
+	ds_store_addtid_b32     v0
+	s_add_u32	m0, m0, 256						// 256 DW
+	s_add_u32	s_restore_mem_offset, s_restore_mem_offset, 256		//mem offset increased by 256DW
+	s_cmp_lt_u32	m0, s_restore_alloc_size				//scc=(m0 < s_restore_alloc_size) ? 1 : 0
+	s_cbranch_scc1	L_RESTORE_LDS_LOOP_W64					//LDS restore is complete?
+
+	/* restore VGPRs */
+L_RESTORE_VGPR:
+	// VGPR SR memory offset : 0
+	s_mov_b32	s_restore_mem_offset, 0x0
+	s_mov_b32	exec_lo, 0xFFFFFFFF					//need every thread from now on
+	s_lshr_b32	m0, s_restore_size, S_WAVE_SIZE
+	s_and_b32	m0, m0, 1
+	s_cmp_eq_u32	m0, 1
+	s_cbranch_scc1	L_ENABLE_RESTORE_VGPR_EXEC_HI
+	s_mov_b32	exec_hi, 0x00000000
+	s_branch	L_RESTORE_VGPR_NORMAL
+L_ENABLE_RESTORE_VGPR_EXEC_HI:
+	s_mov_b32	exec_hi, 0xFFFFFFFF
+L_RESTORE_VGPR_NORMAL:
+	s_getreg_b32	s_restore_alloc_size, hwreg(HW_REG_WAVE_GPR_ALLOC,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SHIFT,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SIZE)
+	s_add_u32	s_restore_alloc_size, s_restore_alloc_size, 1
+	s_lshl_b32	s_restore_alloc_size, s_restore_alloc_size, 2		//Number of VGPRs = (vgpr_size + 1) * 4    (non-zero value)
+	//determine it is wave32 or wave64
+	s_lshr_b32	m0, s_restore_size, S_WAVE_SIZE
+	s_and_b32	m0, m0, 1
+	s_cmp_eq_u32	m0, 1
+	s_cbranch_scc1	L_RESTORE_VGPR_WAVE64
+
+	s_mov_b32	s_restore_buf_rsrc2, 0x1000000				//NUM_RECORDS in bytes
+
+	// VGPR load using dw burst
+	s_mov_b32	s_restore_mem_offset_save, s_restore_mem_offset		// restore start with v1, v0 will be the last
+	s_add_u32	s_restore_mem_offset, s_restore_mem_offset, 128*4
+	s_mov_b32	m0, 4							//VGPR initial index value = 4
+	s_cmp_lt_u32	m0, s_restore_alloc_size
+	s_cbranch_scc0	L_RESTORE_SGPR
+
+L_RESTORE_VGPR_WAVE32_LOOP:
+	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset scope:SCOPE_SYS
+	buffer_load_dword	v1, v0, s_restore_buf_rsrc0, s_restore_mem_offset scope:SCOPE_SYS offset:128
+	buffer_load_dword	v2, v0, s_restore_buf_rsrc0, s_restore_mem_offset scope:SCOPE_SYS offset:128*2
+	buffer_load_dword	v3, v0, s_restore_buf_rsrc0, s_restore_mem_offset scope:SCOPE_SYS offset:128*3
+	s_wait_idle
+	v_movreld_b32	v0, v0							//v[0+m0] = v0
+	v_movreld_b32	v1, v1
+	v_movreld_b32	v2, v2
+	v_movreld_b32	v3, v3
+	s_add_u32	m0, m0, 4						//next vgpr index
+	s_add_u32	s_restore_mem_offset, s_restore_mem_offset, 128*4	//every buffer_load_dword does 128 bytes
+	s_cmp_lt_u32	m0, s_restore_alloc_size				//scc = (m0 < s_restore_alloc_size) ? 1 : 0
+	s_cbranch_scc1	L_RESTORE_VGPR_WAVE32_LOOP				//VGPR restore (except v0) is complete?
+
+	/* VGPR restore on v0 */
+	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save scope:SCOPE_SYS
+	buffer_load_dword	v1, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save scope:SCOPE_SYS offset:128
+	buffer_load_dword	v2, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save scope:SCOPE_SYS offset:128*2
+	buffer_load_dword	v3, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save scope:SCOPE_SYS offset:128*3
+	s_wait_idle
+
+	s_branch	L_RESTORE_SGPR
+
+L_RESTORE_VGPR_WAVE64:
+	s_mov_b32	s_restore_buf_rsrc2, 0x1000000				//NUM_RECORDS in bytes
+
+	// VGPR load using dw burst
+	s_mov_b32	s_restore_mem_offset_save, s_restore_mem_offset		// restore start with v4, v0 will be the last
+	s_add_u32	s_restore_mem_offset, s_restore_mem_offset, 256*4
+	s_mov_b32	m0, 4							//VGPR initial index value = 4
+	s_cmp_lt_u32	m0, s_restore_alloc_size
+	s_cbranch_scc0	L_RESTORE_SHARED_VGPR
+
+L_RESTORE_VGPR_WAVE64_LOOP:
+	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset scope:SCOPE_SYS
+	buffer_load_dword	v1, v0, s_restore_buf_rsrc0, s_restore_mem_offset scope:SCOPE_SYS offset:256
+	buffer_load_dword	v2, v0, s_restore_buf_rsrc0, s_restore_mem_offset scope:SCOPE_SYS offset:256*2
+	buffer_load_dword	v3, v0, s_restore_buf_rsrc0, s_restore_mem_offset scope:SCOPE_SYS offset:256*3
+	s_wait_idle
+	v_movreld_b32	v0, v0							//v[0+m0] = v0
+	v_movreld_b32	v1, v1
+	v_movreld_b32	v2, v2
+	v_movreld_b32	v3, v3
+	s_add_u32	m0, m0, 4						//next vgpr index
+	s_add_u32	s_restore_mem_offset, s_restore_mem_offset, 256*4	//every buffer_load_dword does 256 bytes
+	s_cmp_lt_u32	m0, s_restore_alloc_size				//scc = (m0 < s_restore_alloc_size) ? 1 : 0
+	s_cbranch_scc1	L_RESTORE_VGPR_WAVE64_LOOP				//VGPR restore (except v0) is complete?
+
+L_RESTORE_SHARED_VGPR:
+	s_getreg_b32	s_restore_alloc_size, hwreg(HW_REG_WAVE_LDS_ALLOC,SQ_WAVE_LDS_ALLOC_VGPR_SHARED_SIZE_SHIFT,SQ_WAVE_LDS_ALLOC_VGPR_SHARED_SIZE_SIZE)	//shared_vgpr_size
+	s_and_b32	s_restore_alloc_size, s_restore_alloc_size, 0xFFFFFFFF	//shared_vgpr_size is zero?
+	s_cbranch_scc0	L_RESTORE_V0						//no shared_vgpr used?
+	s_lshl_b32	s_restore_alloc_size, s_restore_alloc_size, 3		//Number of SHARED_VGPRs = shared_vgpr_size * 8    (non-zero value)
+	//m0 now has the value of normal vgpr count, just add the m0 with shared_vgpr count to get the total count.
+	//restore shared_vgpr will start from the index of m0
+	s_add_u32	s_restore_alloc_size, s_restore_alloc_size, m0
+	s_mov_b32	exec_lo, 0xFFFFFFFF
+	s_mov_b32	exec_hi, 0x00000000
+L_RESTORE_SHARED_VGPR_WAVE64_LOOP:
+	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset scope:SCOPE_SYS
+	s_wait_idle
+	v_movreld_b32	v0, v0							//v[0+m0] = v0
+	s_add_u32	m0, m0, 1						//next vgpr index
+	s_add_u32	s_restore_mem_offset, s_restore_mem_offset, 128
+	s_cmp_lt_u32	m0, s_restore_alloc_size				//scc = (m0 < s_restore_alloc_size) ? 1 : 0
+	s_cbranch_scc1	L_RESTORE_SHARED_VGPR_WAVE64_LOOP			//VGPR restore (except v0) is complete?
+
+	s_mov_b32	exec_hi, 0xFFFFFFFF					//restore back exec_hi before restoring V0!!
+
+	/* VGPR restore on v0 */
+L_RESTORE_V0:
+	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save scope:SCOPE_SYS
+	buffer_load_dword	v1, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save scope:SCOPE_SYS offset:256
+	buffer_load_dword	v2, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save scope:SCOPE_SYS offset:256*2
+	buffer_load_dword	v3, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save scope:SCOPE_SYS offset:256*3
+	s_wait_idle
+
+	/* restore SGPRs */
+	//will be 2+8+16*6
+	// SGPR SR memory offset : size(VGPR)+size(SVGPR)
+L_RESTORE_SGPR:
+	get_vgpr_size_bytes(s_restore_mem_offset, s_restore_size)
+	get_svgpr_size_bytes(s_restore_tmp)
+	s_add_u32	s_restore_mem_offset, s_restore_mem_offset, s_restore_tmp
+	s_add_u32	s_restore_mem_offset, s_restore_mem_offset, get_sgpr_size_bytes()
+	s_sub_u32	s_restore_mem_offset, s_restore_mem_offset, 20*4	//s108~s127 is not saved
+
+	s_mov_b32	s_restore_buf_rsrc2, 0x1000000				//NUM_RECORDS in bytes
+
+	s_mov_b32	m0, s_sgpr_save_num
+
+	read_4sgpr_from_mem(s0, s_restore_buf_rsrc0, s_restore_mem_offset)
+	s_wait_idle
+
+	s_sub_u32	m0, m0, 4						// Restore from S[0] to S[104]
+	s_nop		0							// hazard SALU M0=> S_MOVREL
+
+	s_movreld_b64	s0, s0							//s[0+m0] = s0
+	s_movreld_b64	s2, s2
+
+	read_8sgpr_from_mem(s0, s_restore_buf_rsrc0, s_restore_mem_offset)
+	s_wait_idle
+
+	s_sub_u32	m0, m0, 8						// Restore from S[0] to S[96]
+	s_nop		0							// hazard SALU M0=> S_MOVREL
+
+	s_movreld_b64	s0, s0							//s[0+m0] = s0
+	s_movreld_b64	s2, s2
+	s_movreld_b64	s4, s4
+	s_movreld_b64	s6, s6
+
+ L_RESTORE_SGPR_LOOP:
+	read_16sgpr_from_mem(s0, s_restore_buf_rsrc0, s_restore_mem_offset)
+	s_wait_idle
+
+	s_sub_u32	m0, m0, 16						// Restore from S[n] to S[0]
+	s_nop		0							// hazard SALU M0=> S_MOVREL
+
+	s_movreld_b64	s0, s0							//s[0+m0] = s0
+	s_movreld_b64	s2, s2
+	s_movreld_b64	s4, s4
+	s_movreld_b64	s6, s6
+	s_movreld_b64	s8, s8
+	s_movreld_b64	s10, s10
+	s_movreld_b64	s12, s12
+	s_movreld_b64	s14, s14
+
+	s_cmp_eq_u32	m0, 0							//scc = (m0 < s_sgpr_save_num) ? 1 : 0
+	s_cbranch_scc0	L_RESTORE_SGPR_LOOP
+
+	// s_barrier with STATE_PRIV.TRAP_AFTER_INST=1, STATUS.PRIV=1 incorrectly asserts debug exception.
+	// Clear DEBUG_EN before and restore MODE after the barrier.
+	s_setreg_imm32_b32	hwreg(HW_REG_WAVE_MODE), 0
+
+	/* restore HW registers */
+L_RESTORE_HWREG:
+	// HWREG SR memory offset : size(VGPR)+size(SVGPR)+size(SGPR)
+	get_vgpr_size_bytes(s_restore_mem_offset, s_restore_size)
+	get_svgpr_size_bytes(s_restore_tmp)
+	s_add_u32	s_restore_mem_offset, s_restore_mem_offset, s_restore_tmp
+	s_add_u32	s_restore_mem_offset, s_restore_mem_offset, get_sgpr_size_bytes()
+
+	s_mov_b32	s_restore_buf_rsrc2, 0x1000000				//NUM_RECORDS in bytes
+
+	// Restore s_restore_spi_init_hi before the saved value gets clobbered.
+	s_mov_b32 s_restore_spi_init_hi, s_restore_spi_init_hi_save
+
+	read_hwreg_from_mem(s_restore_m0, s_restore_buf_rsrc0, s_restore_mem_offset)
+	read_hwreg_from_mem(s_restore_pc_lo, s_restore_buf_rsrc0, s_restore_mem_offset)
+	read_hwreg_from_mem(s_restore_pc_hi, s_restore_buf_rsrc0, s_restore_mem_offset)
+	read_hwreg_from_mem(s_restore_exec_lo, s_restore_buf_rsrc0, s_restore_mem_offset)
+	read_hwreg_from_mem(s_restore_exec_hi, s_restore_buf_rsrc0, s_restore_mem_offset)
+	read_hwreg_from_mem(s_restore_state_priv, s_restore_buf_rsrc0, s_restore_mem_offset)
+	read_hwreg_from_mem(s_restore_excp_flag_priv, s_restore_buf_rsrc0, s_restore_mem_offset)
+	read_hwreg_from_mem(s_restore_xnack_mask, s_restore_buf_rsrc0, s_restore_mem_offset)
+	read_hwreg_from_mem(s_restore_mode, s_restore_buf_rsrc0, s_restore_mem_offset)
+	read_hwreg_from_mem(s_restore_flat_scratch, s_restore_buf_rsrc0, s_restore_mem_offset)
+	s_wait_idle
+
+	s_setreg_b32	hwreg(HW_REG_WAVE_SCRATCH_BASE_LO), s_restore_flat_scratch
+
+	read_hwreg_from_mem(s_restore_flat_scratch, s_restore_buf_rsrc0, s_restore_mem_offset)
+	s_wait_idle
+
+	s_setreg_b32	hwreg(HW_REG_WAVE_SCRATCH_BASE_HI), s_restore_flat_scratch
+
+	read_hwreg_from_mem(s_restore_tmp, s_restore_buf_rsrc0, s_restore_mem_offset)
+	s_wait_idle
+	s_setreg_b32	hwreg(HW_REG_WAVE_EXCP_FLAG_USER), s_restore_tmp
+
+	read_hwreg_from_mem(s_restore_tmp, s_restore_buf_rsrc0, s_restore_mem_offset)
+	s_wait_idle
+	s_setreg_b32	hwreg(HW_REG_WAVE_TRAP_CTRL), s_restore_tmp
+
+	// Only the first wave needs to restore the workgroup barrier.
+	s_and_b32	s_restore_tmp, s_restore_spi_init_hi, S_RESTORE_SPI_INIT_FIRST_WAVE_MASK
+	s_cbranch_scc0	L_SKIP_BARRIER_RESTORE
+
+	// Skip over WAVE_STATUS, since there is no state to restore from it
+	s_add_u32	s_restore_mem_offset, s_restore_mem_offset, 4
+
+	read_hwreg_from_mem(s_restore_tmp, s_restore_buf_rsrc0, s_restore_mem_offset)
+	s_wait_idle
+
+	s_bitcmp1_b32	s_restore_tmp, BARRIER_STATE_VALID_OFFSET
+	s_cbranch_scc0	L_SKIP_BARRIER_RESTORE
+
+	// extract the saved signal count from s_restore_tmp
+	s_lshr_b32	s_restore_tmp, s_restore_tmp, BARRIER_STATE_SIGNAL_OFFSET
+
+	// We need to call s_barrier_signal repeatedly to restore the signal
+	// count of the work group barrier.  The member count is already
+	// initialized with the number of waves in the work group.
+L_BARRIER_RESTORE_LOOP:
+	s_and_b32	s_restore_tmp, s_restore_tmp, s_restore_tmp
+	s_cbranch_scc0	L_SKIP_BARRIER_RESTORE
+	s_barrier_signal	-1
+	s_add_i32	s_restore_tmp, s_restore_tmp, -1
+	s_branch	L_BARRIER_RESTORE_LOOP
+
+L_SKIP_BARRIER_RESTORE:
+
+	s_mov_b32	m0, s_restore_m0
+	s_mov_b32	exec_lo, s_restore_exec_lo
+	s_mov_b32	exec_hi, s_restore_exec_hi
+
+	// EXCP_FLAG_PRIV.SAVE_CONTEXT and HOST_TRAP may have changed.
+	// Only restore the other fields to avoid clobbering them.
+	s_setreg_b32	hwreg(HW_REG_WAVE_EXCP_FLAG_PRIV, 0, SQ_WAVE_EXCP_FLAG_PRIV_RESTORE_PART_1_SIZE), s_restore_excp_flag_priv
+	s_lshr_b32	s_restore_excp_flag_priv, s_restore_excp_flag_priv, SQ_WAVE_EXCP_FLAG_PRIV_RESTORE_PART_2_SHIFT
+	s_setreg_b32	hwreg(HW_REG_WAVE_EXCP_FLAG_PRIV, SQ_WAVE_EXCP_FLAG_PRIV_RESTORE_PART_2_SHIFT, SQ_WAVE_EXCP_FLAG_PRIV_RESTORE_PART_2_SIZE), s_restore_excp_flag_priv
+	s_lshr_b32	s_restore_excp_flag_priv, s_restore_excp_flag_priv, SQ_WAVE_EXCP_FLAG_PRIV_RESTORE_PART_3_SHIFT - SQ_WAVE_EXCP_FLAG_PRIV_RESTORE_PART_2_SHIFT
+	s_setreg_b32	hwreg(HW_REG_WAVE_EXCP_FLAG_PRIV, SQ_WAVE_EXCP_FLAG_PRIV_RESTORE_PART_3_SHIFT, SQ_WAVE_EXCP_FLAG_PRIV_RESTORE_PART_3_SIZE), s_restore_excp_flag_priv
+
+	s_setreg_b32	hwreg(HW_REG_WAVE_MODE), s_restore_mode
+
+	// Restore trap temporaries 4-11, 13 initialized by SPI debug dispatch logic
+	// ttmp SR memory offset : size(VGPR)+size(SVGPR)+size(SGPR)+0x40
+	get_vgpr_size_bytes(s_restore_ttmps_lo, s_restore_size)
+	get_svgpr_size_bytes(s_restore_ttmps_hi)
+	s_add_u32	s_restore_ttmps_lo, s_restore_ttmps_lo, s_restore_ttmps_hi
+	s_add_u32	s_restore_ttmps_lo, s_restore_ttmps_lo, get_sgpr_size_bytes()
+	s_add_u32	s_restore_ttmps_lo, s_restore_ttmps_lo, s_restore_buf_rsrc0
+	s_addc_u32	s_restore_ttmps_hi, s_restore_buf_rsrc1, 0x0
+	s_and_b32	s_restore_ttmps_hi, s_restore_ttmps_hi, 0xFFFF
+	s_load_dwordx4	[ttmp4, ttmp5, ttmp6, ttmp7], [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x50 scope:SCOPE_SYS
+	s_load_dwordx4	[ttmp8, ttmp9, ttmp10, ttmp11], [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x60 scope:SCOPE_SYS
+	s_load_dword	ttmp13, [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x74 scope:SCOPE_SYS
+	s_wait_idle
+
+	s_and_b32	s_restore_pc_hi, s_restore_pc_hi, 0x0000ffff		//pc[47:32] //Do it here in order not to affect STATUS
+	s_and_b64	exec, exec, exec					// Restore STATUS.EXECZ, not writable by s_setreg_b32
+	s_and_b64	vcc, vcc, vcc						// Restore STATUS.VCCZ, not writable by s_setreg_b32
+
+	s_setreg_b32	hwreg(HW_REG_WAVE_STATE_PRIV), s_restore_state_priv	// SCC is included, which is changed by previous salu
+
+	// Make barrier and LDS state visible to all waves in the group.
+	// STATE_PRIV.BARRIER_COMPLETE may change after this point.
+	s_barrier_signal	-2
+	s_barrier_wait	-2
+
+	s_rfe_b64	s_restore_pc_lo						//Return to the main shader program and resume execution
+
+L_END_PGM:
+	// Make sure that no wave of the workgroup can exit the trap handler
+	// before the workgroup barrier state is saved.
+	s_barrier_signal	-2
+	s_barrier_wait	-2
+	s_endpgm_saved
+end
+
+function write_16sgpr_to_v2(s, lane_offset)
+	// Copy into VGPR for later TCP store.
+	for var sgpr_idx = 0; sgpr_idx < 16; sgpr_idx ++
+		v_writelane_b32	v2, s[sgpr_idx], sgpr_idx + lane_offset
+	end
+	valu_sgpr_hazard()
+	s_add_u32	ttmp13, ttmp13, 0x10
+end
+
+function write_12sgpr_to_v2(s)
+	// Copy into VGPR for later TCP store.
+	for var sgpr_idx = 0; sgpr_idx < 12; sgpr_idx ++
+		v_writelane_b32	v2, s[sgpr_idx], sgpr_idx
+	end
+	valu_sgpr_hazard()
+end
+
+function read_hwreg_from_mem(s, s_rsrc, s_mem_offset)
+	s_buffer_load_dword	s, s_rsrc, s_mem_offset scope:SCOPE_SYS
+	s_add_u32	s_mem_offset, s_mem_offset, 4
+end
+
+function read_16sgpr_from_mem(s, s_rsrc, s_mem_offset)
+	s_sub_u32	s_mem_offset, s_mem_offset, 4*16
+	s_buffer_load_dwordx16	s, s_rsrc, s_mem_offset scope:SCOPE_SYS
+end
+
+function read_8sgpr_from_mem(s, s_rsrc, s_mem_offset)
+	s_sub_u32	s_mem_offset, s_mem_offset, 4*8
+	s_buffer_load_dwordx8	s, s_rsrc, s_mem_offset scope:SCOPE_SYS
+end
+
+function read_4sgpr_from_mem(s, s_rsrc, s_mem_offset)
+	s_sub_u32	s_mem_offset, s_mem_offset, 4*4
+	s_buffer_load_dwordx4	s, s_rsrc, s_mem_offset scope:SCOPE_SYS
+end
+
+function get_vgpr_size_bytes(s_vgpr_size_byte, s_size)
+	s_getreg_b32	s_vgpr_size_byte, hwreg(HW_REG_WAVE_GPR_ALLOC,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SHIFT,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SIZE)
+	s_add_u32	s_vgpr_size_byte, s_vgpr_size_byte, 1
+	s_bitcmp1_b32	s_size, S_WAVE_SIZE
+	s_cbranch_scc1	L_ENABLE_SHIFT_W64
+	s_lshl_b32	s_vgpr_size_byte, s_vgpr_size_byte, (2+7)		//Number of VGPRs = (vgpr_size + 1) * 4 * 32 * 4   (non-zero value)
+	s_branch	L_SHIFT_DONE
+L_ENABLE_SHIFT_W64:
+	s_lshl_b32	s_vgpr_size_byte, s_vgpr_size_byte, (2+8)		//Number of VGPRs = (vgpr_size + 1) * 4 * 64 * 4   (non-zero value)
+L_SHIFT_DONE:
+end
+
+function get_svgpr_size_bytes(s_svgpr_size_byte)
+	s_getreg_b32	s_svgpr_size_byte, hwreg(HW_REG_WAVE_LDS_ALLOC,SQ_WAVE_LDS_ALLOC_VGPR_SHARED_SIZE_SHIFT,SQ_WAVE_LDS_ALLOC_VGPR_SHARED_SIZE_SIZE)
+	s_lshl_b32	s_svgpr_size_byte, s_svgpr_size_byte, (3+7)
+end
+
+function get_sgpr_size_bytes
+	return 512
+end
+
+function get_hwreg_size_bytes
+	return 128
+end
+
+function get_wave_size2(s_reg)
+	s_getreg_b32	s_reg, hwreg(HW_REG_WAVE_STATUS,SQ_WAVE_STATUS_WAVE64_SHIFT,SQ_WAVE_STATUS_WAVE64_SIZE)
+	s_lshl_b32	s_reg, s_reg, S_WAVE_SIZE
+end
+
+function valu_sgpr_hazard
+#if HAVE_VALU_SGPR_HAZARD
+	for var rep = 0; rep < 8; rep ++
+		ds_nop
+	end
+#endif
+end
diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm
index f2087cc2e89d..6869e07a2fff 100644
--- a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm
+++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm
@@ -37,17 +37,28 @@
  * gc_9_4_3:
  *   cpp -DASIC_FAMILY=GC_9_4_3 cwsr_trap_handler_gfx9.asm -P -o gc_9_4_3.sp3
  *   sp3 gc_9_4_3.sp3 -hex gc_9_4_3.hex
+ *
+ * gc_9_5_0:
+ *   cpp -DASIC_FAMILY=GC_9_5_0 cwsr_trap_handler_gfx9.asm -P -o gc_9_5_0.sp3
+ *   sp3 gc_9_5_0.sp3 -hex gc_9_5_0.hex
  */
 
 #define CHIP_VEGAM 18
 #define CHIP_ARCTURUS 23
 #define CHIP_ALDEBARAN 25
 #define CHIP_GC_9_4_3 26
+#define CHIP_GC_9_5_0 27
 
 var ACK_SQC_STORE		    =	1		    //workaround for suspected SQC store bug causing incorrect stores under concurrency
 var SAVE_AFTER_XNACK_ERROR	    =	1		    //workaround for TCP store failure after XNACK error when ALLOW_REPLAY=0, for debugger
 var SINGLE_STEP_MISSED_WORKAROUND   =	(ASIC_FAMILY <= CHIP_ALDEBARAN)	//workaround for lost MODE.DEBUG_EN exception when SAVECTX raised
 
+#if ASIC_FAMILY < CHIP_GC_9_4_3
+#define VMEM_MODIFIERS slc:1 glc:1
+#else
+#define VMEM_MODIFIERS sc0:1 nt:1
+#endif
+
 /**************************************************************************/
 /*			variables					  */
 /**************************************************************************/
@@ -62,7 +73,13 @@ var SQ_WAVE_STATUS_ALLOW_REPLAY_MASK    = 0x400000
 var SQ_WAVE_STATUS_ECC_ERR_MASK         = 0x20000
 
 var SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT	= 12
+#if ASIC_FAMILY >= CHIP_GC_9_5_0
+var SQ_WAVE_LDS_ALLOC_LDS_SIZE_SIZE	= 11
+var LDS_RESTORE_GRANULARITY_BYTES	= 1280
+#else
 var SQ_WAVE_LDS_ALLOC_LDS_SIZE_SIZE	= 9
+var LDS_RESTORE_GRANULARITY_BYTES	= 512
+#endif
 var SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SIZE	= 6
 var SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SIZE	= 3			//FIXME	 sq.blk still has 4 bits at this time while SQ programming guide has 3 bits
 var SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SHIFT	= 24
@@ -283,6 +300,11 @@ L_FETCH_2ND_TRAP:
     s_getreg_b32    ttmp15, hwreg(HW_REG_SQ_SHADER_TMA_HI)
     s_lshl_b64      [ttmp14, ttmp15], [ttmp14, ttmp15], 0x8
 
+    s_bitcmp1_b32   ttmp15, 0xF
+    s_cbranch_scc0  L_NO_SIGN_EXTEND_TMA
+    s_or_b32        ttmp15, ttmp15, 0xFFFF0000
+L_NO_SIGN_EXTEND_TMA:
+
     s_load_dword    ttmp2, [ttmp14, ttmp15], 0x10 glc:1 // debug trap enabled flag
     s_waitcnt       lgkmcnt(0)
     s_lshl_b32      ttmp2, ttmp2, TTMP_DEBUG_TRAP_ENABLED_SHIFT
@@ -425,7 +447,9 @@ L_SAVE:
     s_getreg_b32    s_save_m0, hwreg(HW_REG_MODE)						    //MODE
     write_hwreg_to_mem(s_save_m0, s_save_buf_rsrc0, s_save_mem_offset)
 
-
+    // Clear VSKIP state now that MODE.VSKIP has been saved.
+    // If user shader set it then vector instructions would be skipped.
+    s_setvskip	0,0
 
     /*	    the first wave in the threadgroup	 */
     s_and_b32	    s_save_tmp, s_save_spi_init_hi, S_SAVE_SPI_INIT_FIRST_WAVE_MASK	// extract fisrt wave bit
@@ -552,12 +576,21 @@ if SAVE_AFTER_XNACK_ERROR
 
 	v_lshlrev_b32 v2, 2, v3
 L_SAVE_LDS_LOOP_SQC:
+#if ASIC_FAMILY < CHIP_GC_9_5_0
 	ds_read2_b32 v[0:1], v2 offset0:0 offset1:0x40
 	s_waitcnt lgkmcnt(0)
-
 	write_vgprs_to_mem_with_sqc(v0, 2, s_save_buf_rsrc0, s_save_mem_offset)
 
 	v_add_u32 v2, 0x200, v2
+#else
+	// gfx950 needs to save in multiple of 256 bytes.
+	ds_read_b32 v0, v2
+	s_waitcnt lgkmcnt(0)
+	write_vgprs_to_mem_with_sqc(v0, 1, s_save_buf_rsrc0, s_save_mem_offset)
+
+	v_add_u32 v2, 0x100, v2
+#endif
+
 	v_cmp_lt_u32 vcc[0:1], v2, s_save_alloc_size
 	s_cbranch_vccnz L_SAVE_LDS_LOOP_SQC
 
@@ -576,11 +609,14 @@ end
 L_SAVE_LDS_LOOP_VECTOR:
       ds_read_b64 v[0:1], v2	//x =LDS[a], byte address
       s_waitcnt lgkmcnt(0)
-      buffer_store_dwordx2  v[0:1], v2, s_save_buf_rsrc0, s_save_mem_offset offen:1  glc:1  slc:1
+      buffer_store_dwordx2  v[0:1], v2, s_save_buf_rsrc0, s_save_mem_offset VMEM_MODIFIERS offen:1
 //	s_waitcnt vmcnt(0)
 //	v_add_u32 v2, vcc[0:1], v2, v3
       v_add_u32 v2, v2, v3
       v_cmp_lt_u32 vcc[0:1], v2, s_save_alloc_size
+#if ASIC_FAMILY >= CHIP_GC_9_5_0
+      s_mov_b64 exec, vcc
+#endif
       s_cbranch_vccnz L_SAVE_LDS_LOOP_VECTOR
 
       // restore rsrc3
@@ -743,8 +779,13 @@ L_RESTORE:
   L_RESTORE_LDS_LOOP:
 	buffer_load_dword   v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset lds:1		       // first 64DW
 	buffer_load_dword   v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset lds:1 offset:256	       // second 64DW
-    s_add_u32	    m0, m0, 256*2						// 128 DW
-    s_add_u32	    s_restore_mem_offset, s_restore_mem_offset, 256*2		//mem offset increased by 128DW
+#if ASIC_FAMILY >= CHIP_GC_9_5_0
+	buffer_load_dword   v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset lds:1 offset:512	// third 64DW
+	buffer_load_dword   v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset lds:1 offset:768	// forth 64DW
+	buffer_load_dword   v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset lds:1 offset:1024	// fifth 64DW
+#endif
+    s_add_u32	    m0, m0, LDS_RESTORE_GRANULARITY_BYTES					// 128/320 DW
+    s_add_u32	    s_restore_mem_offset, s_restore_mem_offset, LDS_RESTORE_GRANULARITY_BYTES	//mem offset increased by 128/320 DW
     s_cmp_lt_u32    m0, s_restore_alloc_size					//scc=(m0 < s_restore_alloc_size) ? 1 : 0
     s_cbranch_scc1  L_RESTORE_LDS_LOOP							    //LDS restore is complete?
 
@@ -916,7 +957,7 @@ L_RESTORE:
 /*			the END						  */
 /**************************************************************************/
 L_END_PGM:
-    s_endpgm
+    s_endpgm_saved
 
 end
 
@@ -974,17 +1015,17 @@ L_TCP_STORE_CHECK_DONE:
 end
 
 function write_4vgprs_to_mem(s_rsrc, s_mem_offset)
-	buffer_store_dword v0, v0, s_rsrc, s_mem_offset slc:1 glc:1
-	buffer_store_dword v1, v0, s_rsrc, s_mem_offset slc:1 glc:1  offset:256
-	buffer_store_dword v2, v0, s_rsrc, s_mem_offset slc:1 glc:1  offset:256*2
-	buffer_store_dword v3, v0, s_rsrc, s_mem_offset slc:1 glc:1  offset:256*3
+	buffer_store_dword v0, v0, s_rsrc, s_mem_offset VMEM_MODIFIERS
+	buffer_store_dword v1, v0, s_rsrc, s_mem_offset VMEM_MODIFIERS offset:256
+	buffer_store_dword v2, v0, s_rsrc, s_mem_offset VMEM_MODIFIERS offset:256*2
+	buffer_store_dword v3, v0, s_rsrc, s_mem_offset VMEM_MODIFIERS offset:256*3
 end
 
 function read_4vgprs_from_mem(s_rsrc, s_mem_offset)
-	buffer_load_dword v0, v0, s_rsrc, s_mem_offset slc:1 glc:1
-	buffer_load_dword v1, v0, s_rsrc, s_mem_offset slc:1 glc:1 offset:256
-	buffer_load_dword v2, v0, s_rsrc, s_mem_offset slc:1 glc:1 offset:256*2
-	buffer_load_dword v3, v0, s_rsrc, s_mem_offset slc:1 glc:1 offset:256*3
+	buffer_load_dword v0, v0, s_rsrc, s_mem_offset VMEM_MODIFIERS
+	buffer_load_dword v1, v0, s_rsrc, s_mem_offset VMEM_MODIFIERS offset:256
+	buffer_load_dword v2, v0, s_rsrc, s_mem_offset VMEM_MODIFIERS offset:256*2
+	buffer_load_dword v3, v0, s_rsrc, s_mem_offset VMEM_MODIFIERS offset:256*3
 	s_waitcnt vmcnt(0)
 end
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index 81d07ecf666d..22925df6a791 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -22,7 +22,6 @@
  */
 
 #include <linux/device.h>
-#include <linux/export.h>
 #include <linux/err.h>
 #include <linux/fs.h>
 #include <linux/file.h>
@@ -36,7 +35,6 @@
 #include <linux/mman.h>
 #include <linux/ptrace.h>
 #include <linux/dma-buf.h>
-#include <linux/fdtable.h>
 #include <linux/processor.h>
 #include "kfd_priv.h"
 #include "kfd_device_queue_manager.h"
@@ -44,6 +42,7 @@
 #include "amdgpu_amdkfd.h"
 #include "kfd_smi_events.h"
 #include "amdgpu_dma_buf.h"
+#include "kfd_debug.h"
 
 static long kfd_ioctl(struct file *, unsigned int, unsigned long);
 static int kfd_open(struct inode *, struct file *);
@@ -62,8 +61,10 @@ static const struct file_operations kfd_fops = {
 };
 
 static int kfd_char_dev_major = -1;
-static struct class *kfd_class;
 struct device *kfd_device;
+static const struct class kfd_class = {
+	.name = kfd_dev_name,
+};
 
 static inline struct kfd_process_device *kfd_lock_pdd_by_id(struct kfd_process *p, __u32 gpu_id)
 {
@@ -93,14 +94,13 @@ int kfd_chardev_init(void)
 	if (err < 0)
 		goto err_register_chrdev;
 
-	kfd_class = class_create(THIS_MODULE, kfd_dev_name);
-	err = PTR_ERR(kfd_class);
-	if (IS_ERR(kfd_class))
+	err = class_register(&kfd_class);
+	if (err)
 		goto err_class_create;
 
-	kfd_device = device_create(kfd_class, NULL,
-					MKDEV(kfd_char_dev_major, 0),
-					NULL, kfd_dev_name);
+	kfd_device = device_create(&kfd_class, NULL,
+				   MKDEV(kfd_char_dev_major, 0),
+				   NULL, kfd_dev_name);
 	err = PTR_ERR(kfd_device);
 	if (IS_ERR(kfd_device))
 		goto err_device_create;
@@ -108,7 +108,7 @@ int kfd_chardev_init(void)
 	return 0;
 
 err_device_create:
-	class_destroy(kfd_class);
+	class_unregister(&kfd_class);
 err_class_create:
 	unregister_chrdev(kfd_char_dev_major, kfd_dev_name);
 err_register_chrdev:
@@ -117,8 +117,8 @@ err_register_chrdev:
 
 void kfd_chardev_exit(void)
 {
-	device_destroy(kfd_class, MKDEV(kfd_char_dev_major, 0));
-	class_destroy(kfd_class);
+	device_destroy(&kfd_class, MKDEV(kfd_char_dev_major, 0));
+	class_unregister(&kfd_class);
 	unregister_chrdev(kfd_char_dev_major, kfd_dev_name);
 	kfd_device = NULL;
 }
@@ -142,22 +142,20 @@ static int kfd_open(struct inode *inode, struct file *filep)
 		return -EPERM;
 	}
 
-	process = kfd_create_process(filep);
+	process = kfd_create_process(current);
 	if (IS_ERR(process))
 		return PTR_ERR(process);
 
-	if (kfd_is_locked()) {
-		dev_dbg(kfd_device, "kfd is locked!\n"
-				"process %d unreferenced", process->pasid);
+	if (kfd_process_init_cwsr_apu(process, filep)) {
 		kfd_unref_process(process);
-		return -EAGAIN;
+		return -EFAULT;
 	}
 
 	/* filep now owns the reference returned by kfd_create_process */
 	filep->private_data = process;
 
-	dev_dbg(kfd_device, "process %d opened, compat mode (32 bit) - %d\n",
-		process->pasid, process->is_32bit_user_mode);
+	dev_dbg(kfd_device, "process pid %d opened kfd node, compat mode (32 bit) - %d\n",
+		process->lead_thread->pid, process->is_32bit_user_mode);
 
 	return 0;
 }
@@ -186,7 +184,12 @@ static int kfd_ioctl_get_version(struct file *filep, struct kfd_process *p,
 static int set_queue_properties_from_user(struct queue_properties *q_properties,
 				struct kfd_ioctl_create_queue_args *args)
 {
-	if (args->queue_percentage > KFD_MAX_QUEUE_PERCENTAGE) {
+	/*
+	 * Repurpose queue percentage to accommodate new features:
+	 * bit 0-7: queue percentage
+	 * bit 8-15: pm4_target_xcc
+	 */
+	if ((args->queue_percentage & 0xFF) > KFD_MAX_QUEUE_PERCENTAGE) {
 		pr_err("Queue percentage must be between 0 to KFD_MAX_QUEUE_PERCENTAGE\n");
 		return -EINVAL;
 	}
@@ -208,6 +211,11 @@ static int set_queue_properties_from_user(struct queue_properties *q_properties,
 		return -EINVAL;
 	}
 
+	if (args->ring_size < KFD_MIN_QUEUE_RING_SIZE) {
+		args->ring_size = KFD_MIN_QUEUE_RING_SIZE;
+		pr_debug("Size lower. clamped to KFD_MIN_QUEUE_RING_SIZE");
+	}
+
 	if (!access_ok((const void __user *) args->read_pointer_address,
 			sizeof(uint32_t))) {
 		pr_err("Can't access read pointer\n");
@@ -236,18 +244,21 @@ static int set_queue_properties_from_user(struct queue_properties *q_properties,
 
 	q_properties->is_interop = false;
 	q_properties->is_gws = false;
-	q_properties->queue_percent = args->queue_percentage;
+	q_properties->queue_percent = args->queue_percentage & 0xFF;
+	/* bit 8-15 are repurposed to be PM4 target XCC */
+	q_properties->pm4_target_xcc = (args->queue_percentage >> 8) & 0xFF;
 	q_properties->priority = args->queue_priority;
 	q_properties->queue_address = args->ring_base_address;
 	q_properties->queue_size = args->ring_size;
-	q_properties->read_ptr = (uint32_t *) args->read_pointer_address;
-	q_properties->write_ptr = (uint32_t *) args->write_pointer_address;
+	q_properties->read_ptr = (void __user *)args->read_pointer_address;
+	q_properties->write_ptr = (void __user *)args->write_pointer_address;
 	q_properties->eop_ring_buffer_address = args->eop_buffer_address;
 	q_properties->eop_ring_buffer_size = args->eop_buffer_size;
 	q_properties->ctx_save_restore_area_address =
 			args->ctx_save_restore_address;
 	q_properties->ctx_save_restore_area_size = args->ctx_save_restore_size;
 	q_properties->ctl_stack_size = args->ctl_stack_size;
+	q_properties->sdma_engine_id = args->sdma_engine_id;
 	if (args->queue_type == KFD_IOC_QUEUE_TYPE_COMPUTE ||
 		args->queue_type == KFD_IOC_QUEUE_TYPE_COMPUTE_AQL)
 		q_properties->type = KFD_QUEUE_TYPE_COMPUTE;
@@ -255,6 +266,8 @@ static int set_queue_properties_from_user(struct queue_properties *q_properties,
 		q_properties->type = KFD_QUEUE_TYPE_SDMA;
 	else if (args->queue_type == KFD_IOC_QUEUE_TYPE_SDMA_XGMI)
 		q_properties->type = KFD_QUEUE_TYPE_SDMA_XGMI;
+	else if (args->queue_type == KFD_IOC_QUEUE_TYPE_SDMA_BY_ENG_ID)
+		q_properties->type = KFD_QUEUE_TYPE_SDMA_BY_ENG_ID;
 	else
 		return -ENOTSUPP;
 
@@ -293,13 +306,12 @@ static int kfd_ioctl_create_queue(struct file *filep, struct kfd_process *p,
 					void *data)
 {
 	struct kfd_ioctl_create_queue_args *args = data;
-	struct kfd_dev *dev;
+	struct kfd_node *dev;
 	int err = 0;
 	unsigned int queue_id;
 	struct kfd_process_device *pdd;
 	struct queue_properties q_properties;
 	uint32_t doorbell_offset_in_process = 0;
-	struct amdgpu_bo *wptr_bo = NULL;
 
 	memset(&q_properties, 0, sizeof(struct queue_properties));
 
@@ -327,54 +339,37 @@ static int kfd_ioctl_create_queue(struct file *filep, struct kfd_process *p,
 		goto err_bind_process;
 	}
 
-	if (!pdd->doorbell_index &&
-	    kfd_alloc_process_doorbells(dev, &pdd->doorbell_index) < 0) {
-		err = -ENOMEM;
-		goto err_alloc_doorbells;
-	}
-
-	/* Starting with GFX11, wptr BOs must be mapped to GART for MES to determine work
-	 * on unmapped queues for usermode queue oversubscription (no aggregated doorbell)
-	 */
-	if (dev->shared_resources.enable_mes &&
-			((dev->adev->mes.sched_version & AMDGPU_MES_API_VERSION_MASK)
-			>> AMDGPU_MES_API_VERSION_SHIFT) >= 2) {
-		struct amdgpu_bo_va_mapping *wptr_mapping;
-		struct amdgpu_vm *wptr_vm;
-
-		wptr_vm = drm_priv_to_vm(pdd->drm_priv);
-		err = amdgpu_bo_reserve(wptr_vm->root.bo, false);
-		if (err)
-			goto err_wptr_map_gart;
-
-		wptr_mapping = amdgpu_vm_bo_lookup_mapping(
-				wptr_vm, args->write_pointer_address >> PAGE_SHIFT);
-		amdgpu_bo_unreserve(wptr_vm->root.bo);
-		if (!wptr_mapping) {
-			pr_err("Failed to lookup wptr bo\n");
-			err = -EINVAL;
-			goto err_wptr_map_gart;
-		}
+	if (q_properties.type == KFD_QUEUE_TYPE_SDMA_BY_ENG_ID) {
+		int max_sdma_eng_id = kfd_get_num_sdma_engines(dev) +
+				      kfd_get_num_xgmi_sdma_engines(dev) - 1;
 
-		wptr_bo = wptr_mapping->bo_va->base.bo;
-		if (wptr_bo->tbo.base.size > PAGE_SIZE) {
-			pr_err("Requested GART mapping for wptr bo larger than one page\n");
+		if (q_properties.sdma_engine_id > max_sdma_eng_id) {
 			err = -EINVAL;
-			goto err_wptr_map_gart;
+			pr_err("sdma_engine_id %i exceeds maximum id of %i\n",
+			       q_properties.sdma_engine_id, max_sdma_eng_id);
+			goto err_sdma_engine_id;
 		}
+	}
 
-		err = amdgpu_amdkfd_map_gtt_bo_to_gart(dev->adev, wptr_bo);
+	if (!pdd->qpd.proc_doorbells) {
+		err = kfd_alloc_process_doorbells(dev->kfd, pdd);
 		if (err) {
-			pr_err("Failed to map wptr bo to GART\n");
-			goto err_wptr_map_gart;
+			pr_debug("failed to allocate process doorbells\n");
+			goto err_bind_process;
 		}
 	}
 
-	pr_debug("Creating queue for PASID 0x%x on gpu 0x%x\n",
-			p->pasid,
+	err = kfd_queue_acquire_buffers(pdd, &q_properties);
+	if (err) {
+		pr_debug("failed to acquire user queue buffers\n");
+		goto err_acquire_queue_buf;
+	}
+
+	pr_debug("Creating queue for process pid %d on gpu 0x%x\n",
+			p->lead_thread->pid,
 			dev->id);
 
-	err = pqm_create_queue(&p->pqm, dev, filep, &q_properties, &queue_id, wptr_bo,
+	err = pqm_create_queue(&p->pqm, dev, &q_properties, &queue_id,
 			NULL, NULL, NULL, &doorbell_offset_in_process);
 	if (err != 0)
 		goto err_create_queue;
@@ -404,13 +399,14 @@ static int kfd_ioctl_create_queue(struct file *filep, struct kfd_process *p,
 	pr_debug("Write ptr address   == 0x%016llX\n",
 			args->write_pointer_address);
 
+	kfd_dbg_ev_raise(KFD_EC_MASK(EC_QUEUE_NEW), p, dev, queue_id, false, NULL, 0);
 	return 0;
 
 err_create_queue:
-	if (wptr_bo)
-		amdgpu_amdkfd_free_gtt_mem(dev->adev, wptr_bo);
-err_wptr_map_gart:
-err_alloc_doorbells:
+	kfd_queue_unref_bo_vas(pdd, &q_properties);
+	kfd_queue_release_buffers(pdd, &q_properties);
+err_acquire_queue_buf:
+err_sdma_engine_id:
 err_bind_process:
 err_pdd:
 	mutex_unlock(&p->mutex);
@@ -423,9 +419,9 @@ static int kfd_ioctl_destroy_queue(struct file *filp, struct kfd_process *p,
 	int retval;
 	struct kfd_ioctl_destroy_queue_args *args = data;
 
-	pr_debug("Destroying queue id %d for pasid 0x%x\n",
+	pr_debug("Destroying queue id %d for process pid %d\n",
 				args->queue_id,
-				p->pasid);
+				p->lead_thread->pid);
 
 	mutex_lock(&p->mutex);
 
@@ -442,7 +438,12 @@ static int kfd_ioctl_update_queue(struct file *filp, struct kfd_process *p,
 	struct kfd_ioctl_update_queue_args *args = data;
 	struct queue_properties properties;
 
-	if (args->queue_percentage > KFD_MAX_QUEUE_PERCENTAGE) {
+	/*
+	 * Repurpose queue percentage to accommodate new features:
+	 * bit 0-7: queue percentage
+	 * bit 8-15: pm4_target_xcc
+	 */
+	if ((args->queue_percentage & 0xFF) > KFD_MAX_QUEUE_PERCENTAGE) {
 		pr_err("Queue percentage must be between 0 to KFD_MAX_QUEUE_PERCENTAGE\n");
 		return -EINVAL;
 	}
@@ -464,13 +465,20 @@ static int kfd_ioctl_update_queue(struct file *filp, struct kfd_process *p,
 		return -EINVAL;
 	}
 
+	if (args->ring_size < KFD_MIN_QUEUE_RING_SIZE) {
+		args->ring_size = KFD_MIN_QUEUE_RING_SIZE;
+		pr_debug("Size lower. clamped to KFD_MIN_QUEUE_RING_SIZE");
+	}
+
 	properties.queue_address = args->ring_base_address;
 	properties.queue_size = args->ring_size;
-	properties.queue_percent = args->queue_percentage;
+	properties.queue_percent = args->queue_percentage & 0xFF;
+	/* bit 8-15 are repurposed to be PM4 target XCC */
+	properties.pm4_target_xcc = (args->queue_percentage >> 8) & 0xFF;
 	properties.priority = args->queue_priority;
 
-	pr_debug("Updating queue id %d for pasid 0x%x\n",
-			args->queue_id, p->pasid);
+	pr_debug("Updating queue id %d for process pid %d\n",
+			args->queue_id, p->lead_thread->pid);
 
 	mutex_lock(&p->mutex);
 
@@ -513,26 +521,18 @@ static int kfd_ioctl_set_cu_mask(struct file *filp, struct kfd_process *p,
 		cu_mask_size = sizeof(uint32_t) * (max_num_cus/32);
 	}
 
-	minfo.cu_mask.ptr = kzalloc(cu_mask_size, GFP_KERNEL);
-	if (!minfo.cu_mask.ptr)
-		return -ENOMEM;
-
-	retval = copy_from_user(minfo.cu_mask.ptr, cu_mask_ptr, cu_mask_size);
-	if (retval) {
+	minfo.cu_mask.ptr = memdup_user(cu_mask_ptr, cu_mask_size);
+	if (IS_ERR(minfo.cu_mask.ptr)) {
 		pr_debug("Could not copy CU mask from userspace");
-		retval = -EFAULT;
-		goto out;
+		return PTR_ERR(minfo.cu_mask.ptr);
 	}
 
-	minfo.update_flag = UPDATE_FLAG_CU_MASK;
-
 	mutex_lock(&p->mutex);
 
 	retval = pqm_update_mqd(&p->pqm, args->queue_id, &minfo);
 
 	mutex_unlock(&p->mutex);
 
-out:
 	kfree(minfo.cu_mask.ptr);
 	return retval;
 }
@@ -599,7 +599,8 @@ static int kfd_ioctl_set_memory_policy(struct file *filep,
 				default_policy,
 				alternate_policy,
 				(void __user *)args->alternate_aperture_base,
-				args->alternate_aperture_size))
+				args->alternate_aperture_size,
+				args->misc_process_flag))
 		err = -EINVAL;
 
 out:
@@ -698,7 +699,7 @@ static int kfd_ioctl_get_process_apertures(struct file *filp,
 	struct kfd_process_device_apertures *pAperture;
 	int i;
 
-	dev_dbg(kfd_device, "get apertures for PASID 0x%x", p->pasid);
+	dev_dbg(kfd_device, "get apertures for process pid %d", p->lead_thread->pid);
 
 	args->num_of_nodes = 0;
 
@@ -750,7 +751,8 @@ static int kfd_ioctl_get_process_apertures_new(struct file *filp,
 	int ret;
 	int i;
 
-	dev_dbg(kfd_device, "get apertures for PASID 0x%x", p->pasid);
+	dev_dbg(kfd_device, "get apertures for process pid %d",
+			p->lead_thread->pid);
 
 	if (args->num_of_nodes == 0) {
 		/* Return number of nodes, so that user space can alloacate
@@ -765,8 +767,8 @@ static int kfd_ioctl_get_process_apertures_new(struct file *filp,
 	 * nodes, but not more than args->num_of_nodes as that is
 	 * the amount of memory allocated by user
 	 */
-	pa = kzalloc((sizeof(struct kfd_process_device_apertures) *
-				args->num_of_nodes), GFP_KERNEL);
+	pa = kcalloc(args->num_of_nodes, sizeof(struct kfd_process_device_apertures),
+		     GFP_KERNEL);
 	if (!pa)
 		return -ENOMEM;
 
@@ -887,7 +889,7 @@ static int kfd_ioctl_set_scratch_backing_va(struct file *filep,
 {
 	struct kfd_ioctl_set_scratch_backing_va_args *args = data;
 	struct kfd_process_device *pdd;
-	struct kfd_dev *dev;
+	struct kfd_node *dev;
 	long err;
 
 	mutex_lock(&p->mutex);
@@ -1006,19 +1008,23 @@ err_drm_file:
 	return ret;
 }
 
-bool kfd_dev_is_large_bar(struct kfd_dev *dev)
+bool kfd_dev_is_large_bar(struct kfd_node *dev)
 {
-	if (debug_largebar) {
+	if (dev->kfd->adev->debug_largebar) {
 		pr_debug("Simulate large-bar allocation on non large-bar machine\n");
 		return true;
 	}
 
-	if (dev->use_iommu_v2)
-		return false;
-
 	if (dev->local_mem_info.local_mem_size_private == 0 &&
-			dev->local_mem_info.local_mem_size_public > 0)
+	    dev->local_mem_info.local_mem_size_public > 0)
+		return true;
+
+	if (dev->local_mem_info.local_mem_size_public == 0 &&
+	    dev->kfd->adev->gmc.is_app_apu) {
+		pr_debug("APP APU, Consider like a large bar system\n");
 		return true;
+	}
+
 	return false;
 }
 
@@ -1030,7 +1036,8 @@ static int kfd_ioctl_get_available_memory(struct file *filep,
 
 	if (!pdd)
 		return -EINVAL;
-	args->available = amdgpu_amdkfd_get_available_memory(pdd->dev->adev);
+	args->available = amdgpu_amdkfd_get_available_memory(pdd->dev->adev,
+							pdd->dev->node_id);
 	kfd_unlock_pdd(pdd);
 	return 0;
 }
@@ -1041,7 +1048,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep,
 	struct kfd_ioctl_alloc_memory_of_gpu_args *args = data;
 	struct kfd_process_device *pdd;
 	void *mem;
-	struct kfd_dev *dev;
+	struct kfd_node *dev;
 	int idr_handle;
 	long err;
 	uint64_t offset = args->mmap_offset;
@@ -1057,7 +1064,12 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep,
 	svm_range_list_lock_and_flush_work(&p->svms, current->mm);
 	mutex_lock(&p->svms.lock);
 	mmap_write_unlock(current->mm);
-	if (interval_tree_iter_first(&p->svms.objects,
+
+	/* Skip a special case that allocates VRAM without VA,
+	 * VA will be invalid of 0.
+	 */
+	if (!(!args->va_addr && (flags & KFD_IOC_ALLOC_MEM_FLAGS_VRAM)) &&
+	    interval_tree_iter_first(&p->svms.objects,
 				     args->va_addr >> PAGE_SHIFT,
 				     (args->va_addr + args->size - 1) >> PAGE_SHIFT)) {
 		pr_err("Address: 0x%llx already allocated by SVM\n",
@@ -1105,7 +1117,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep,
 	}
 
 	if (flags & KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL) {
-		if (args->size != kfd_doorbell_process_slice(dev)) {
+		if (args->size != kfd_doorbell_process_slice(dev->kfd)) {
 			err = -EINVAL;
 			goto err_unlock;
 		}
@@ -1120,7 +1132,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep,
 			goto err_unlock;
 		}
 		offset = dev->adev->rmmio_remap.bus_addr;
-		if (!offset) {
+		if (!offset || (PAGE_SIZE > 4096)) {
 			err = -ENOMEM;
 			goto err_unlock;
 		}
@@ -1146,7 +1158,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep,
 
 		if (flags & KFD_IOC_ALLOC_MEM_FLAGS_AQL_QUEUE_MEM)
 			size >>= 1;
-		WRITE_ONCE(pdd->vram_usage, pdd->vram_usage + PAGE_ALIGN(size));
+		atomic64_add(PAGE_ALIGN(size), &pdd->vram_usage);
 	}
 
 	mutex_unlock(&p->mutex);
@@ -1217,7 +1229,7 @@ static int kfd_ioctl_free_memory_of_gpu(struct file *filep,
 		kfd_process_device_remove_obj_handle(
 			pdd, GET_IDR_HANDLE(args->handle));
 
-	WRITE_ONCE(pdd->vram_usage, pdd->vram_usage - size);
+	atomic64_sub(size, &pdd->vram_usage);
 
 err_unlock:
 err_pdd:
@@ -1231,7 +1243,7 @@ static int kfd_ioctl_map_memory_to_gpu(struct file *filep,
 	struct kfd_ioctl_map_memory_to_gpu_args *args = data;
 	struct kfd_process_device *pdd, *peer_pdd;
 	void *mem;
-	struct kfd_dev *dev;
+	struct kfd_node *dev;
 	long err = 0;
 	int i;
 	uint32_t *devices_arr = NULL;
@@ -1398,14 +1410,13 @@ static int kfd_ioctl_unmap_memory_from_gpu(struct file *filep,
 		err = amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
 			peer_pdd->dev->adev, (struct kgd_mem *)mem, peer_pdd->drm_priv);
 		if (err) {
-			pr_err("Failed to unmap from gpu %d/%d\n",
-			       i, args->n_devices);
+			pr_debug("Failed to unmap from gpu %d/%d\n", i, args->n_devices);
 			goto unmap_memory_from_gpu_failed;
 		}
 		args->n_success = i+1;
 	}
 
-	flush_tlb = kfd_flush_tlb_after_unmap(pdd->dev);
+	flush_tlb = kfd_flush_tlb_after_unmap(pdd->dev->kfd);
 	if (flush_tlb) {
 		err = amdgpu_amdkfd_gpuvm_sync_memory(pdd->dev->adev,
 				(struct kgd_mem *) mem, true);
@@ -1414,17 +1425,23 @@ static int kfd_ioctl_unmap_memory_from_gpu(struct file *filep,
 			goto sync_memory_failed;
 		}
 	}
-	mutex_unlock(&p->mutex);
 
-	if (flush_tlb) {
-		/* Flush TLBs after waiting for the page table updates to complete */
-		for (i = 0; i < args->n_devices; i++) {
-			peer_pdd = kfd_process_device_data_by_id(p, devices_arr[i]);
-			if (WARN_ON_ONCE(!peer_pdd))
-				continue;
+	/* Flush TLBs after waiting for the page table updates to complete */
+	for (i = 0; i < args->n_devices; i++) {
+		peer_pdd = kfd_process_device_data_by_id(p, devices_arr[i]);
+		if (WARN_ON_ONCE(!peer_pdd))
+			continue;
+		if (flush_tlb)
 			kfd_flush_tlb(peer_pdd, TLB_FLUSH_HEAVYWEIGHT);
-		}
+
+		/* Remove dma mapping after tlb flush to avoid IO_PAGE_FAULT */
+		err = amdgpu_amdkfd_gpuvm_dmaunmap_mem(mem, peer_pdd->drm_priv);
+		if (err)
+			goto sync_memory_failed;
 	}
+
+	mutex_unlock(&p->mutex);
+
 	kfree(devices_arr);
 
 	return 0;
@@ -1445,7 +1462,7 @@ static int kfd_ioctl_alloc_queue_gws(struct file *filep,
 	int retval;
 	struct kfd_ioctl_alloc_queue_gws_args *args = data;
 	struct queue *q;
-	struct kfd_dev *dev;
+	struct kfd_node *dev;
 
 	mutex_lock(&p->mutex);
 	q = pqm_get_user_queue(&p->pqm, args->queue_id);
@@ -1467,6 +1484,12 @@ static int kfd_ioctl_alloc_queue_gws(struct file *filep,
 		goto out_unlock;
 	}
 
+	if (p->debug_trap_enabled && (!kfd_dbg_has_gws_support(dev) ||
+				      kfd_dbg_has_cwsr_workaround(dev))) {
+		retval = -EBUSY;
+		goto out_unlock;
+	}
+
 	retval = pqm_set_gws(&p->pqm, args->queue_id, args->num_gws ? dev->gws : NULL);
 	mutex_unlock(&p->mutex);
 
@@ -1482,16 +1505,17 @@ static int kfd_ioctl_get_dmabuf_info(struct file *filep,
 		struct kfd_process *p, void *data)
 {
 	struct kfd_ioctl_get_dmabuf_info_args *args = data;
-	struct kfd_dev *dev = NULL;
+	struct kfd_node *dev = NULL;
 	struct amdgpu_device *dmabuf_adev;
 	void *metadata_buffer = NULL;
 	uint32_t flags;
+	int8_t xcp_id;
 	unsigned int i;
 	int r;
 
 	/* Find a KFD GPU device that supports the get_dmabuf_info query */
 	for (i = 0; kfd_topology_enum_kfd_devices(i, &dev) == 0; i++)
-		if (dev)
+		if (dev && !kfd_devcgroup_check_permission(dev))
 			break;
 	if (!dev)
 		return -EINVAL;
@@ -1506,17 +1530,14 @@ static int kfd_ioctl_get_dmabuf_info(struct file *filep,
 	r = amdgpu_amdkfd_get_dmabuf_info(dev->adev, args->dmabuf_fd,
 					  &dmabuf_adev, &args->size,
 					  metadata_buffer, args->metadata_size,
-					  &args->metadata_size, &flags);
+					  &args->metadata_size, &flags, &xcp_id);
 	if (r)
 		goto exit;
 
-	/* Reverse-lookup gpu_id from kgd pointer */
-	dev = kfd_device_by_adev(dmabuf_adev);
-	if (!dev) {
-		r = -EINVAL;
-		goto exit;
-	}
-	args->gpu_id = dev->id;
+	if (xcp_id >= 0)
+		args->gpu_id = dmabuf_adev->kfd.dev->nodes[xcp_id]->id;
+	else
+		args->gpu_id = dev->id;
 	args->flags = flags;
 
 	/* Copy metadata buffer to user mode */
@@ -1538,16 +1559,11 @@ static int kfd_ioctl_import_dmabuf(struct file *filep,
 {
 	struct kfd_ioctl_import_dmabuf_args *args = data;
 	struct kfd_process_device *pdd;
-	struct dma_buf *dmabuf;
 	int idr_handle;
 	uint64_t size;
 	void *mem;
 	int r;
 
-	dmabuf = dma_buf_get(args->dmabuf_fd);
-	if (IS_ERR(dmabuf))
-		return PTR_ERR(dmabuf);
-
 	mutex_lock(&p->mutex);
 	pdd = kfd_process_device_data_by_id(p, args->gpu_id);
 	if (!pdd) {
@@ -1561,10 +1577,10 @@ static int kfd_ioctl_import_dmabuf(struct file *filep,
 		goto err_unlock;
 	}
 
-	r = amdgpu_amdkfd_gpuvm_import_dmabuf(pdd->dev->adev, dmabuf,
-					      args->va_addr, pdd->drm_priv,
-					      (struct kgd_mem **)&mem, &size,
-					      NULL);
+	r = amdgpu_amdkfd_gpuvm_import_dmabuf_fd(pdd->dev->adev, args->dmabuf_fd,
+						 args->va_addr, pdd->drm_priv,
+						 (struct kgd_mem **)&mem, &size,
+						 NULL);
 	if (r)
 		goto err_unlock;
 
@@ -1575,7 +1591,6 @@ static int kfd_ioctl_import_dmabuf(struct file *filep,
 	}
 
 	mutex_unlock(&p->mutex);
-	dma_buf_put(dmabuf);
 
 	args->handle = MAKE_HANDLE(args->gpu_id, idr_handle);
 
@@ -1586,7 +1601,6 @@ err_free:
 					       pdd->drm_priv, NULL);
 err_unlock:
 	mutex_unlock(&p->mutex);
-	dma_buf_put(dmabuf);
 	return r;
 }
 
@@ -1596,7 +1610,7 @@ static int kfd_ioctl_export_dmabuf(struct file *filep,
 	struct kfd_ioctl_export_dmabuf_args *args = data;
 	struct kfd_process_device *pdd;
 	struct dma_buf *dmabuf;
-	struct kfd_dev *dev;
+	struct kfd_node *dev;
 	void *mem;
 	int ret = 0;
 
@@ -1822,33 +1836,34 @@ static uint32_t get_process_num_bos(struct kfd_process *p)
 		idr_for_each_entry(&pdd->alloc_idr, mem, id) {
 			struct kgd_mem *kgd_mem = (struct kgd_mem *)mem;
 
-			if ((uint64_t)kgd_mem->va > pdd->gpuvm_base)
+			if (!kgd_mem->va || kgd_mem->va > pdd->gpuvm_base)
 				num_of_bos++;
 		}
 	}
 	return num_of_bos;
 }
 
-static int criu_get_prime_handle(struct drm_gem_object *gobj, int flags,
-				      u32 *shared_fd)
+static int criu_get_prime_handle(struct kgd_mem *mem,
+				 int flags, u32 *shared_fd,
+				 struct file **file)
 {
 	struct dma_buf *dmabuf;
 	int ret;
 
-	dmabuf = amdgpu_gem_prime_export(gobj, flags);
-	if (IS_ERR(dmabuf)) {
-		ret = PTR_ERR(dmabuf);
+	ret = amdgpu_amdkfd_gpuvm_export_dmabuf(mem, &dmabuf);
+	if (ret) {
 		pr_err("dmabuf export failed for the BO\n");
 		return ret;
 	}
 
-	ret = dma_buf_fd(dmabuf, flags);
+	ret = get_unused_fd_flags(flags);
 	if (ret < 0) {
 		pr_err("dmabuf create fd failed, ret:%d\n", ret);
 		goto out_free_dmabuf;
 	}
 
 	*shared_fd = ret;
+	*file = dmabuf->file;
 	return 0;
 
 out_free_dmabuf:
@@ -1856,6 +1871,25 @@ out_free_dmabuf:
 	return ret;
 }
 
+static void commit_files(struct file **files,
+			 struct kfd_criu_bo_bucket *bo_buckets,
+			 unsigned int count,
+			 int err)
+{
+	while (count--) {
+		struct file *file = files[count];
+
+		if (!file)
+			continue;
+		if (err) {
+			fput(file);
+			put_unused_fd(bo_buckets[count].dmabuf_fd);
+		} else {
+			fd_install(bo_buckets[count].dmabuf_fd, file);
+		}
+	}
+}
+
 static int criu_checkpoint_bos(struct kfd_process *p,
 			       uint32_t num_bos,
 			       uint8_t __user *user_bos,
@@ -1864,6 +1898,7 @@ static int criu_checkpoint_bos(struct kfd_process *p,
 {
 	struct kfd_criu_bo_bucket *bo_buckets;
 	struct kfd_criu_bo_priv_data *bo_privs;
+	struct file **files = NULL;
 	int ret = 0, pdd_index, bo_index = 0, id;
 	void *mem;
 
@@ -1877,6 +1912,12 @@ static int criu_checkpoint_bos(struct kfd_process *p,
 		goto exit;
 	}
 
+	files = kvzalloc(num_bos * sizeof(struct file *), GFP_KERNEL);
+	if (!files) {
+		ret = -ENOMEM;
+		goto exit;
+	}
+
 	for (pdd_index = 0; pdd_index < p->n_pdds; pdd_index++) {
 		struct kfd_process_device *pdd = p->pdds[pdd_index];
 		struct amdgpu_bo *dumper_bo;
@@ -1887,15 +1928,14 @@ static int criu_checkpoint_bos(struct kfd_process *p,
 			struct kfd_criu_bo_priv_data *bo_priv;
 			int i, dev_idx = 0;
 
-			if (!mem) {
-				ret = -ENOMEM;
-				goto exit;
-			}
-
 			kgd_mem = (struct kgd_mem *)mem;
 			dumper_bo = kgd_mem->bo;
 
-			if ((uint64_t)kgd_mem->va <= pdd->gpuvm_base)
+			/* Skip checkpointing BOs that are used for Trap handler
+			 * code and state. Currently, these BOs have a VA that
+			 * is less GPUVM Base
+			 */
+			if (kgd_mem->va && kgd_mem->va <= pdd->gpuvm_base)
 				continue;
 
 			bo_bucket = &bo_buckets[bo_index];
@@ -1917,10 +1957,10 @@ static int criu_checkpoint_bos(struct kfd_process *p,
 			}
 			if (bo_bucket->alloc_flags
 			    & (KFD_IOC_ALLOC_MEM_FLAGS_VRAM | KFD_IOC_ALLOC_MEM_FLAGS_GTT)) {
-				ret = criu_get_prime_handle(&dumper_bo->tbo.base,
+				ret = criu_get_prime_handle(kgd_mem,
 						bo_bucket->alloc_flags &
 						KFD_IOC_ALLOC_MEM_FLAGS_WRITABLE ? DRM_RDWR : 0,
-						&bo_bucket->dmabuf_fd);
+						&bo_bucket->dmabuf_fd, &files[bo_index]);
 				if (ret)
 					goto exit;
 			} else {
@@ -1938,7 +1978,7 @@ static int criu_checkpoint_bos(struct kfd_process *p,
 				bo_bucket->offset = amdgpu_bo_mmap_offset(dumper_bo);
 
 			for (i = 0; i < p->n_pdds; i++) {
-				if (amdgpu_amdkfd_bo_mapped_to_dev(p->pdds[i]->dev->adev, kgd_mem))
+				if (amdgpu_amdkfd_bo_mapped_to_dev(p->pdds[i]->drm_priv, kgd_mem))
 					bo_priv->mapped_gpuids[dev_idx++] = p->pdds[i]->user_gpu_id;
 			}
 
@@ -1971,12 +2011,8 @@ static int criu_checkpoint_bos(struct kfd_process *p,
 	*priv_offset += num_bos * sizeof(*bo_privs);
 
 exit:
-	while (ret && bo_index--) {
-		if (bo_buckets[bo_index].alloc_flags
-		    & (KFD_IOC_ALLOC_MEM_FLAGS_VRAM | KFD_IOC_ALLOC_MEM_FLAGS_GTT))
-			close_fd(bo_buckets[bo_index].dmabuf_fd);
-	}
-
+	commit_files(files, bo_buckets, bo_index, ret);
+	kvfree(files);
 	kvfree(bo_buckets);
 	kvfree(bo_privs);
 	return ret;
@@ -2001,9 +2037,7 @@ static int criu_get_process_object_info(struct kfd_process *p,
 
 	num_events = kfd_get_num_events(p);
 
-	ret = svm_range_get_info(p, &num_svm_ranges, &svm_priv_data_size);
-	if (ret)
-		return ret;
+	svm_range_get_info(p, &num_svm_ranges, &svm_priv_data_size);
 
 	*num_objects = num_queues + num_events + num_svm_ranges;
 
@@ -2178,7 +2212,7 @@ static int criu_restore_devices(struct kfd_process *p,
 	}
 
 	for (i = 0; i < args->num_devices; i++) {
-		struct kfd_dev *dev;
+		struct kfd_node *dev;
 		struct kfd_process_device *pdd;
 		struct file *drm_file;
 
@@ -2239,10 +2273,10 @@ static int criu_restore_devices(struct kfd_process *p,
 			goto exit;
 		}
 
-		if (!pdd->doorbell_index &&
-		    kfd_alloc_process_doorbells(pdd->dev, &pdd->doorbell_index) < 0) {
-			ret = -ENOMEM;
-			goto exit;
+		if (!pdd->qpd.proc_doorbells) {
+			ret = kfd_alloc_process_doorbells(dev->kfd, pdd);
+			if (ret)
+				goto exit;
 		}
 	}
 
@@ -2268,7 +2302,8 @@ static int criu_restore_memory_of_gpu(struct kfd_process_device *pdd,
 	u64 offset;
 
 	if (bo_bucket->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL) {
-		if (bo_bucket->size != kfd_doorbell_process_slice(pdd->dev))
+		if (bo_bucket->size !=
+				kfd_doorbell_process_slice(pdd->dev->kfd))
 			return -EINVAL;
 
 		offset = kfd_get_process_doorbells(pdd);
@@ -2281,7 +2316,7 @@ static int criu_restore_memory_of_gpu(struct kfd_process_device *pdd,
 			return -EINVAL;
 		}
 		offset = pdd->dev->adev->rmmio_remap.bus_addr;
-		if (!offset) {
+		if (!offset || (PAGE_SIZE > 4096)) {
 			pr_err("amdgpu_amdkfd_get_mmio_remap_phys_addr failed\n");
 			return -ENOMEM;
 		}
@@ -2320,14 +2355,15 @@ static int criu_restore_memory_of_gpu(struct kfd_process_device *pdd,
 	} else if (bo_bucket->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) {
 		bo_bucket->restored_offset = offset;
 		/* Update the VRAM usage count */
-		WRITE_ONCE(pdd->vram_usage, pdd->vram_usage + bo_bucket->size);
+		atomic64_add(bo_bucket->size, &pdd->vram_usage);
 	}
 	return 0;
 }
 
 static int criu_restore_bo(struct kfd_process *p,
 			   struct kfd_criu_bo_bucket *bo_bucket,
-			   struct kfd_criu_bo_priv_data *bo_priv)
+			   struct kfd_criu_bo_priv_data *bo_priv,
+			   struct file **file)
 {
 	struct kfd_process_device *pdd;
 	struct kgd_mem *kgd_mem;
@@ -2350,7 +2386,7 @@ static int criu_restore_bo(struct kfd_process *p,
 
 	/* now map these BOs to GPU/s */
 	for (j = 0; j < p->n_pdds; j++) {
-		struct kfd_dev *peer;
+		struct kfd_node *peer;
 		struct kfd_process_device *peer_pdd;
 
 		if (!bo_priv->mapped_gpuids[j])
@@ -2378,8 +2414,8 @@ static int criu_restore_bo(struct kfd_process *p,
 	/* create the dmabuf object and export the bo */
 	if (bo_bucket->alloc_flags
 	    & (KFD_IOC_ALLOC_MEM_FLAGS_VRAM | KFD_IOC_ALLOC_MEM_FLAGS_GTT)) {
-		ret = criu_get_prime_handle(&kgd_mem->bo->tbo.base, DRM_RDWR,
-					    &bo_bucket->dmabuf_fd);
+		ret = criu_get_prime_handle(kgd_mem, DRM_RDWR,
+					    &bo_bucket->dmabuf_fd, file);
 		if (ret)
 			return ret;
 	} else {
@@ -2396,6 +2432,7 @@ static int criu_restore_bos(struct kfd_process *p,
 {
 	struct kfd_criu_bo_bucket *bo_buckets = NULL;
 	struct kfd_criu_bo_priv_data *bo_privs = NULL;
+	struct file **files = NULL;
 	int ret = 0;
 	uint32_t i = 0;
 
@@ -2409,6 +2446,12 @@ static int criu_restore_bos(struct kfd_process *p,
 	if (!bo_buckets)
 		return -ENOMEM;
 
+	files = kvzalloc(args->num_bos * sizeof(struct file *), GFP_KERNEL);
+	if (!files) {
+		ret = -ENOMEM;
+		goto exit;
+	}
+
 	ret = copy_from_user(bo_buckets, (void __user *)args->bos,
 			     args->num_bos * sizeof(*bo_buckets));
 	if (ret) {
@@ -2434,7 +2477,7 @@ static int criu_restore_bos(struct kfd_process *p,
 
 	/* Create and map new BOs */
 	for (; i < args->num_bos; i++) {
-		ret = criu_restore_bo(p, &bo_buckets[i], &bo_privs[i]);
+		ret = criu_restore_bo(p, &bo_buckets[i], &bo_privs[i], &files[i]);
 		if (ret) {
 			pr_debug("Failed to restore BO[%d] ret%d\n", i, ret);
 			goto exit;
@@ -2449,11 +2492,8 @@ static int criu_restore_bos(struct kfd_process *p,
 		ret = -EFAULT;
 
 exit:
-	while (ret && i--) {
-		if (bo_buckets[i].alloc_flags
-		   & (KFD_IOC_ALLOC_MEM_FLAGS_VRAM | KFD_IOC_ALLOC_MEM_FLAGS_GTT))
-			close_fd(bo_buckets[i].dmabuf_fd);
-	}
+	commit_files(files, bo_buckets, i, ret);
+	kvfree(files);
 	kvfree(bo_buckets);
 	kvfree(bo_privs);
 	return ret;
@@ -2525,8 +2565,8 @@ static int criu_restore(struct file *filep,
 	pr_debug("CRIU restore (num_devices:%u num_bos:%u num_objects:%u priv_data_size:%llu)\n",
 		 args->num_devices, args->num_bos, args->num_objects, args->priv_data_size);
 
-	if (!args->bos || !args->devices || !args->priv_data || !args->priv_data_size ||
-	    !args->num_devices || !args->num_bos)
+	if ((args->num_bos > 0 && !args->bos) || !args->devices || !args->priv_data ||
+	    !args->priv_data_size || !args->num_devices)
 		return -EINVAL;
 
 	mutex_lock(&p->mutex);
@@ -2715,6 +2755,372 @@ static int kfd_ioctl_criu(struct file *filep, struct kfd_process *p, void *data)
 	return ret;
 }
 
+static int runtime_enable(struct kfd_process *p, uint64_t r_debug,
+			bool enable_ttmp_setup)
+{
+	int i = 0, ret = 0;
+
+	if (p->is_runtime_retry)
+		goto retry;
+
+	if (p->runtime_info.runtime_state != DEBUG_RUNTIME_STATE_DISABLED)
+		return -EBUSY;
+
+	for (i = 0; i < p->n_pdds; i++) {
+		struct kfd_process_device *pdd = p->pdds[i];
+
+		if (pdd->qpd.queue_count)
+			return -EEXIST;
+
+		/*
+		 * Setup TTMPs by default.
+		 * Note that this call must remain here for MES ADD QUEUE to
+		 * skip_process_ctx_clear unconditionally as the first call to
+		 * SET_SHADER_DEBUGGER clears any stale process context data
+		 * saved in MES.
+		 */
+		if (pdd->dev->kfd->shared_resources.enable_mes)
+			kfd_dbg_set_mes_debug_mode(pdd, !kfd_dbg_has_cwsr_workaround(pdd->dev));
+	}
+
+	p->runtime_info.runtime_state = DEBUG_RUNTIME_STATE_ENABLED;
+	p->runtime_info.r_debug = r_debug;
+	p->runtime_info.ttmp_setup = enable_ttmp_setup;
+
+	if (p->runtime_info.ttmp_setup) {
+		for (i = 0; i < p->n_pdds; i++) {
+			struct kfd_process_device *pdd = p->pdds[i];
+
+			if (!kfd_dbg_is_rlc_restore_supported(pdd->dev)) {
+				amdgpu_gfx_off_ctrl(pdd->dev->adev, false);
+				pdd->dev->kfd2kgd->enable_debug_trap(
+						pdd->dev->adev,
+						true,
+						pdd->dev->vm_info.last_vmid_kfd);
+			} else if (kfd_dbg_is_per_vmid_supported(pdd->dev)) {
+				pdd->spi_dbg_override = pdd->dev->kfd2kgd->enable_debug_trap(
+						pdd->dev->adev,
+						false,
+						0);
+			}
+		}
+	}
+
+retry:
+	if (p->debug_trap_enabled) {
+		if (!p->is_runtime_retry) {
+			kfd_dbg_trap_activate(p);
+			kfd_dbg_ev_raise(KFD_EC_MASK(EC_PROCESS_RUNTIME),
+					p, NULL, 0, false, NULL, 0);
+		}
+
+		mutex_unlock(&p->mutex);
+		ret = down_interruptible(&p->runtime_enable_sema);
+		mutex_lock(&p->mutex);
+
+		p->is_runtime_retry = !!ret;
+	}
+
+	return ret;
+}
+
+static int runtime_disable(struct kfd_process *p)
+{
+	int i = 0, ret = 0;
+	bool was_enabled = p->runtime_info.runtime_state == DEBUG_RUNTIME_STATE_ENABLED;
+
+	p->runtime_info.runtime_state = DEBUG_RUNTIME_STATE_DISABLED;
+	p->runtime_info.r_debug = 0;
+
+	if (p->debug_trap_enabled) {
+		if (was_enabled)
+			kfd_dbg_trap_deactivate(p, false, 0);
+
+		if (!p->is_runtime_retry)
+			kfd_dbg_ev_raise(KFD_EC_MASK(EC_PROCESS_RUNTIME),
+					p, NULL, 0, false, NULL, 0);
+
+		mutex_unlock(&p->mutex);
+		ret = down_interruptible(&p->runtime_enable_sema);
+		mutex_lock(&p->mutex);
+
+		p->is_runtime_retry = !!ret;
+		if (ret)
+			return ret;
+	}
+
+	if (was_enabled && p->runtime_info.ttmp_setup) {
+		for (i = 0; i < p->n_pdds; i++) {
+			struct kfd_process_device *pdd = p->pdds[i];
+
+			if (!kfd_dbg_is_rlc_restore_supported(pdd->dev))
+				amdgpu_gfx_off_ctrl(pdd->dev->adev, true);
+		}
+	}
+
+	p->runtime_info.ttmp_setup = false;
+
+	/* disable ttmp setup */
+	for (i = 0; i < p->n_pdds; i++) {
+		struct kfd_process_device *pdd = p->pdds[i];
+		int last_err = 0;
+
+		if (kfd_dbg_is_per_vmid_supported(pdd->dev)) {
+			pdd->spi_dbg_override =
+					pdd->dev->kfd2kgd->disable_debug_trap(
+					pdd->dev->adev,
+					false,
+					pdd->dev->vm_info.last_vmid_kfd);
+
+			if (!pdd->dev->kfd->shared_resources.enable_mes)
+				last_err = debug_refresh_runlist(pdd->dev->dqm);
+			else
+				last_err = kfd_dbg_set_mes_debug_mode(pdd,
+							   !kfd_dbg_has_cwsr_workaround(pdd->dev));
+
+			if (last_err)
+				ret = last_err;
+		}
+	}
+
+	return ret;
+}
+
+static int kfd_ioctl_runtime_enable(struct file *filep, struct kfd_process *p, void *data)
+{
+	struct kfd_ioctl_runtime_enable_args *args = data;
+	int r;
+
+	mutex_lock(&p->mutex);
+
+	if (args->mode_mask & KFD_RUNTIME_ENABLE_MODE_ENABLE_MASK)
+		r = runtime_enable(p, args->r_debug,
+				!!(args->mode_mask & KFD_RUNTIME_ENABLE_MODE_TTMP_SAVE_MASK));
+	else
+		r = runtime_disable(p);
+
+	mutex_unlock(&p->mutex);
+
+	return r;
+}
+
+static int kfd_ioctl_set_debug_trap(struct file *filep, struct kfd_process *p, void *data)
+{
+	struct kfd_ioctl_dbg_trap_args *args = data;
+	struct task_struct *thread = NULL;
+	struct mm_struct *mm = NULL;
+	struct pid *pid = NULL;
+	struct kfd_process *target = NULL;
+	struct kfd_process_device *pdd = NULL;
+	int r = 0;
+
+	if (sched_policy == KFD_SCHED_POLICY_NO_HWS) {
+		pr_err("Debugging does not support sched_policy %i", sched_policy);
+		return -EINVAL;
+	}
+
+	pid = find_get_pid(args->pid);
+	if (!pid) {
+		pr_debug("Cannot find pid info for %i\n", args->pid);
+		r = -ESRCH;
+		goto out;
+	}
+
+	thread = get_pid_task(pid, PIDTYPE_PID);
+	if (!thread) {
+		r = -ESRCH;
+		goto out;
+	}
+
+	mm = get_task_mm(thread);
+	if (!mm) {
+		r = -ESRCH;
+		goto out;
+	}
+
+	if (args->op == KFD_IOC_DBG_TRAP_ENABLE) {
+		bool create_process;
+
+		rcu_read_lock();
+		create_process = thread && thread != current && ptrace_parent(thread) == current;
+		rcu_read_unlock();
+
+		target = create_process ? kfd_create_process(thread) :
+					kfd_lookup_process_by_pid(pid);
+	} else {
+		target = kfd_lookup_process_by_pid(pid);
+	}
+
+	if (IS_ERR_OR_NULL(target)) {
+		pr_debug("Cannot find process PID %i to debug\n", args->pid);
+		r = target ? PTR_ERR(target) : -ESRCH;
+		target = NULL;
+		goto out;
+	}
+
+	/* Check if target is still PTRACED. */
+	rcu_read_lock();
+	if (target != p && args->op != KFD_IOC_DBG_TRAP_DISABLE
+				&& ptrace_parent(target->lead_thread) != current) {
+		pr_err("PID %i is not PTRACED and cannot be debugged\n", args->pid);
+		r = -EPERM;
+	}
+	rcu_read_unlock();
+
+	if (r)
+		goto out;
+
+	mutex_lock(&target->mutex);
+
+	if (args->op != KFD_IOC_DBG_TRAP_ENABLE && !target->debug_trap_enabled) {
+		pr_err("PID %i not debug enabled for op %i\n", args->pid, args->op);
+		r = -EINVAL;
+		goto unlock_out;
+	}
+
+	if (target->runtime_info.runtime_state != DEBUG_RUNTIME_STATE_ENABLED &&
+			(args->op == KFD_IOC_DBG_TRAP_SET_WAVE_LAUNCH_OVERRIDE ||
+			 args->op == KFD_IOC_DBG_TRAP_SET_WAVE_LAUNCH_MODE ||
+			 args->op == KFD_IOC_DBG_TRAP_SUSPEND_QUEUES ||
+			 args->op == KFD_IOC_DBG_TRAP_RESUME_QUEUES ||
+			 args->op == KFD_IOC_DBG_TRAP_SET_NODE_ADDRESS_WATCH ||
+			 args->op == KFD_IOC_DBG_TRAP_CLEAR_NODE_ADDRESS_WATCH ||
+			 args->op == KFD_IOC_DBG_TRAP_SET_FLAGS)) {
+		r = -EPERM;
+		goto unlock_out;
+	}
+
+	if (args->op == KFD_IOC_DBG_TRAP_SET_NODE_ADDRESS_WATCH ||
+	    args->op == KFD_IOC_DBG_TRAP_CLEAR_NODE_ADDRESS_WATCH) {
+		int user_gpu_id = kfd_process_get_user_gpu_id(target,
+				args->op == KFD_IOC_DBG_TRAP_SET_NODE_ADDRESS_WATCH ?
+					args->set_node_address_watch.gpu_id :
+					args->clear_node_address_watch.gpu_id);
+
+		pdd = kfd_process_device_data_by_id(target, user_gpu_id);
+		if (user_gpu_id == -EINVAL || !pdd) {
+			r = -ENODEV;
+			goto unlock_out;
+		}
+	}
+
+	switch (args->op) {
+	case KFD_IOC_DBG_TRAP_ENABLE:
+		if (target != p)
+			target->debugger_process = p;
+
+		r = kfd_dbg_trap_enable(target,
+					args->enable.dbg_fd,
+					(void __user *)args->enable.rinfo_ptr,
+					&args->enable.rinfo_size);
+		if (!r)
+			target->exception_enable_mask = args->enable.exception_mask;
+
+		break;
+	case KFD_IOC_DBG_TRAP_DISABLE:
+		r = kfd_dbg_trap_disable(target);
+		break;
+	case KFD_IOC_DBG_TRAP_SEND_RUNTIME_EVENT:
+		r = kfd_dbg_send_exception_to_runtime(target,
+				args->send_runtime_event.gpu_id,
+				args->send_runtime_event.queue_id,
+				args->send_runtime_event.exception_mask);
+		break;
+	case KFD_IOC_DBG_TRAP_SET_EXCEPTIONS_ENABLED:
+		kfd_dbg_set_enabled_debug_exception_mask(target,
+				args->set_exceptions_enabled.exception_mask);
+		break;
+	case KFD_IOC_DBG_TRAP_SET_WAVE_LAUNCH_OVERRIDE:
+		r = kfd_dbg_trap_set_wave_launch_override(target,
+				args->launch_override.override_mode,
+				args->launch_override.enable_mask,
+				args->launch_override.support_request_mask,
+				&args->launch_override.enable_mask,
+				&args->launch_override.support_request_mask);
+		break;
+	case KFD_IOC_DBG_TRAP_SET_WAVE_LAUNCH_MODE:
+		r = kfd_dbg_trap_set_wave_launch_mode(target,
+				args->launch_mode.launch_mode);
+		break;
+	case KFD_IOC_DBG_TRAP_SUSPEND_QUEUES:
+		r = suspend_queues(target,
+				args->suspend_queues.num_queues,
+				args->suspend_queues.grace_period,
+				args->suspend_queues.exception_mask,
+				(uint32_t *)args->suspend_queues.queue_array_ptr);
+
+		break;
+	case KFD_IOC_DBG_TRAP_RESUME_QUEUES:
+		r = resume_queues(target, args->resume_queues.num_queues,
+				(uint32_t *)args->resume_queues.queue_array_ptr);
+		break;
+	case KFD_IOC_DBG_TRAP_SET_NODE_ADDRESS_WATCH:
+		r = kfd_dbg_trap_set_dev_address_watch(pdd,
+				args->set_node_address_watch.address,
+				args->set_node_address_watch.mask,
+				&args->set_node_address_watch.id,
+				args->set_node_address_watch.mode);
+		break;
+	case KFD_IOC_DBG_TRAP_CLEAR_NODE_ADDRESS_WATCH:
+		r = kfd_dbg_trap_clear_dev_address_watch(pdd,
+				args->clear_node_address_watch.id);
+		break;
+	case KFD_IOC_DBG_TRAP_SET_FLAGS:
+		r = kfd_dbg_trap_set_flags(target, &args->set_flags.flags);
+		break;
+	case KFD_IOC_DBG_TRAP_QUERY_DEBUG_EVENT:
+		r = kfd_dbg_ev_query_debug_event(target,
+				&args->query_debug_event.queue_id,
+				&args->query_debug_event.gpu_id,
+				args->query_debug_event.exception_mask,
+				&args->query_debug_event.exception_mask);
+		break;
+	case KFD_IOC_DBG_TRAP_QUERY_EXCEPTION_INFO:
+		r = kfd_dbg_trap_query_exception_info(target,
+				args->query_exception_info.source_id,
+				args->query_exception_info.exception_code,
+				args->query_exception_info.clear_exception,
+				(void __user *)args->query_exception_info.info_ptr,
+				&args->query_exception_info.info_size);
+		break;
+	case KFD_IOC_DBG_TRAP_GET_QUEUE_SNAPSHOT:
+		r = pqm_get_queue_snapshot(&target->pqm,
+				args->queue_snapshot.exception_mask,
+				(void __user *)args->queue_snapshot.snapshot_buf_ptr,
+				&args->queue_snapshot.num_queues,
+				&args->queue_snapshot.entry_size);
+		break;
+	case KFD_IOC_DBG_TRAP_GET_DEVICE_SNAPSHOT:
+		r = kfd_dbg_trap_device_snapshot(target,
+				args->device_snapshot.exception_mask,
+				(void __user *)args->device_snapshot.snapshot_buf_ptr,
+				&args->device_snapshot.num_devices,
+				&args->device_snapshot.entry_size);
+		break;
+	default:
+		pr_err("Invalid option: %i\n", args->op);
+		r = -EINVAL;
+	}
+
+unlock_out:
+	mutex_unlock(&target->mutex);
+
+out:
+	if (thread)
+		put_task_struct(thread);
+
+	if (mm)
+		mmput(mm);
+
+	if (pid)
+		put_pid(pid);
+
+	if (target)
+		kfd_unref_process(target);
+
+	return r;
+}
+
 #define AMDKFD_IOCTL_DEF(ioctl, _func, _flags) \
 	[_IOC_NR(ioctl)] = {.cmd = ioctl, .func = _func, .flags = _flags, \
 			    .cmd_drv = 0, .name = #ioctl}
@@ -2827,6 +3233,12 @@ static const struct amdkfd_ioctl_desc amdkfd_ioctls[] = {
 
 	AMDKFD_IOCTL_DEF(AMDKFD_IOC_EXPORT_DMABUF,
 				kfd_ioctl_export_dmabuf, 0),
+
+	AMDKFD_IOCTL_DEF(AMDKFD_IOC_RUNTIME_ENABLE,
+			kfd_ioctl_runtime_enable, 0),
+
+	AMDKFD_IOCTL_DEF(AMDKFD_IOC_DBG_TRAP,
+			kfd_ioctl_set_debug_trap, 0),
 };
 
 #define AMDKFD_CORE_IOCTL_COUNT	ARRAY_SIZE(amdkfd_ioctls)
@@ -2843,8 +3255,10 @@ static long kfd_ioctl(struct file *filep, unsigned int cmd, unsigned long arg)
 	int retcode = -EINVAL;
 	bool ptrace_attached = false;
 
-	if (nr >= AMDKFD_CORE_IOCTL_COUNT)
+	if (nr >= AMDKFD_CORE_IOCTL_COUNT) {
+		retcode = -ENOTTY;
 		goto err_i1;
+	}
 
 	if ((nr >= AMDKFD_COMMAND_START) && (nr < AMDKFD_COMMAND_END)) {
 		u32 amdkfd_size;
@@ -2857,8 +3271,10 @@ static long kfd_ioctl(struct file *filep, unsigned int cmd, unsigned long arg)
 			asize = amdkfd_size;
 
 		cmd = ioctl->cmd;
-	} else
+	} else {
+		retcode = -ENOTTY;
 		goto err_i1;
+	}
 
 	dev_dbg(kfd_device, "ioctl cmd 0x%x (#0x%x), arg 0x%lx\n", cmd, nr, arg);
 
@@ -2947,7 +3363,7 @@ err_i1:
 	return retcode;
 }
 
-static int kfd_mmio_mmap(struct kfd_dev *dev, struct kfd_process *process,
+static int kfd_mmio_mmap(struct kfd_node *dev, struct kfd_process *process,
 		      struct vm_area_struct *vma)
 {
 	phys_addr_t address;
@@ -2955,6 +3371,9 @@ static int kfd_mmio_mmap(struct kfd_dev *dev, struct kfd_process *process,
 	if (vma->vm_end - vma->vm_start != PAGE_SIZE)
 		return -EINVAL;
 
+	if (PAGE_SIZE > 4096)
+		return -EINVAL;
+
 	address = dev->adev->rmmio_remap.bus_addr;
 
 	vm_flags_set(vma, VM_IO | VM_DONTCOPY | VM_DONTEXPAND | VM_NORESERVE |
@@ -2962,12 +3381,12 @@ static int kfd_mmio_mmap(struct kfd_dev *dev, struct kfd_process *process,
 
 	vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
 
-	pr_debug("pasid 0x%x mapping mmio page\n"
+	pr_debug("process pid %d mapping mmio page\n"
 		 "     target user address == 0x%08llX\n"
 		 "     physical address    == 0x%08llX\n"
 		 "     vm_flags            == 0x%04lX\n"
 		 "     size                == 0x%04lX\n",
-		 process->pasid, (unsigned long long) vma->vm_start,
+		 process->lead_thread->pid, (unsigned long long) vma->vm_start,
 		 address, vma->vm_flags, PAGE_SIZE);
 
 	return io_remap_pfn_range(vma,
@@ -2981,7 +3400,7 @@ static int kfd_mmio_mmap(struct kfd_dev *dev, struct kfd_process *process,
 static int kfd_mmap(struct file *filp, struct vm_area_struct *vma)
 {
 	struct kfd_process *process;
-	struct kfd_dev *dev = NULL;
+	struct kfd_node *dev = NULL;
 	unsigned long mmap_offset;
 	unsigned int gpu_id;
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
index 475e47027354..4a7180b46b71 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
@@ -26,9 +26,9 @@
 #include "kfd_crat.h"
 #include "kfd_priv.h"
 #include "kfd_topology.h"
-#include "kfd_iommu.h"
 #include "amdgpu.h"
 #include "amdgpu_amdkfd.h"
+#include "amdgpu_xgmi.h"
 
 /* GPU Processor ID base for dGPUs for which VCRAT needs to be created.
  * GPU processor ID are expressed with Bit[31]=1.
@@ -56,6 +56,7 @@ static struct kfd_gpu_cache_info kaveri_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -65,6 +66,7 @@ static struct kfd_gpu_cache_info kaveri_cache_info[] = {
 		/* Scalar L1 Instruction Cache (in SQC module) per bank */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_INST_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -74,6 +76,7 @@ static struct kfd_gpu_cache_info kaveri_cache_info[] = {
 		/* Scalar L1 Data Cache (in SQC module) per bank */
 		.cache_size = 8,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -89,6 +92,7 @@ static struct kfd_gpu_cache_info carrizo_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -96,8 +100,9 @@ static struct kfd_gpu_cache_info carrizo_cache_info[] = {
 	},
 	{
 		/* Scalar L1 Instruction Cache (in SQC module) per bank */
-		.cache_size = 8,
+		.cache_size = 32,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_INST_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -105,8 +110,9 @@ static struct kfd_gpu_cache_info carrizo_cache_info[] = {
 	},
 	{
 		/* Scalar L1 Data Cache (in SQC module) per bank. */
-		.cache_size = 4,
+		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -136,6 +142,7 @@ static struct kfd_gpu_cache_info vega10_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -145,6 +152,7 @@ static struct kfd_gpu_cache_info vega10_cache_info[] = {
 		/* Scalar L1 Instruction Cache per SQC */
 		.cache_size = 32,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_INST_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -154,6 +162,7 @@ static struct kfd_gpu_cache_info vega10_cache_info[] = {
 		/* Scalar L1 Data Cache per SQC */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -163,6 +172,7 @@ static struct kfd_gpu_cache_info vega10_cache_info[] = {
 		/* L2 Data Cache per GPU (Total Tex Cache) */
 		.cache_size = 4096,
 		.cache_level = 2,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -175,6 +185,7 @@ static struct kfd_gpu_cache_info raven_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -184,6 +195,7 @@ static struct kfd_gpu_cache_info raven_cache_info[] = {
 		/* Scalar L1 Instruction Cache per SQC */
 		.cache_size = 32,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_INST_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -193,6 +205,7 @@ static struct kfd_gpu_cache_info raven_cache_info[] = {
 		/* Scalar L1 Data Cache per SQC */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -202,6 +215,7 @@ static struct kfd_gpu_cache_info raven_cache_info[] = {
 		/* L2 Data Cache per GPU (Total Tex Cache) */
 		.cache_size = 1024,
 		.cache_level = 2,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -214,6 +228,7 @@ static struct kfd_gpu_cache_info renoir_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -223,6 +238,7 @@ static struct kfd_gpu_cache_info renoir_cache_info[] = {
 		/* Scalar L1 Instruction Cache per SQC */
 		.cache_size = 32,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_INST_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -232,6 +248,7 @@ static struct kfd_gpu_cache_info renoir_cache_info[] = {
 		/* Scalar L1 Data Cache per SQC */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -241,6 +258,7 @@ static struct kfd_gpu_cache_info renoir_cache_info[] = {
 		/* L2 Data Cache per GPU (Total Tex Cache) */
 		.cache_size = 1024,
 		.cache_level = 2,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -253,6 +271,7 @@ static struct kfd_gpu_cache_info vega12_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -262,6 +281,7 @@ static struct kfd_gpu_cache_info vega12_cache_info[] = {
 		/* Scalar L1 Instruction Cache per SQC */
 		.cache_size = 32,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_INST_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -271,6 +291,7 @@ static struct kfd_gpu_cache_info vega12_cache_info[] = {
 		/* Scalar L1 Data Cache per SQC */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -280,6 +301,7 @@ static struct kfd_gpu_cache_info vega12_cache_info[] = {
 		/* L2 Data Cache per GPU (Total Tex Cache) */
 		.cache_size = 2048,
 		.cache_level = 2,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -292,6 +314,7 @@ static struct kfd_gpu_cache_info vega20_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -301,6 +324,7 @@ static struct kfd_gpu_cache_info vega20_cache_info[] = {
 		/* Scalar L1 Instruction Cache per SQC */
 		.cache_size = 32,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_INST_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -310,6 +334,7 @@ static struct kfd_gpu_cache_info vega20_cache_info[] = {
 		/* Scalar L1 Data Cache per SQC */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -319,6 +344,7 @@ static struct kfd_gpu_cache_info vega20_cache_info[] = {
 		/* L2 Data Cache per GPU (Total Tex Cache) */
 		.cache_size = 8192,
 		.cache_level = 2,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -331,6 +357,7 @@ static struct kfd_gpu_cache_info aldebaran_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -340,6 +367,7 @@ static struct kfd_gpu_cache_info aldebaran_cache_info[] = {
 		/* Scalar L1 Instruction Cache per SQC */
 		.cache_size = 32,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_INST_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -349,6 +377,7 @@ static struct kfd_gpu_cache_info aldebaran_cache_info[] = {
 		/* Scalar L1 Data Cache per SQC */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -358,6 +387,7 @@ static struct kfd_gpu_cache_info aldebaran_cache_info[] = {
 		/* L2 Data Cache per GPU (Total Tex Cache) */
 		.cache_size = 8192,
 		.cache_level = 2,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -370,6 +400,7 @@ static struct kfd_gpu_cache_info navi10_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -379,6 +410,7 @@ static struct kfd_gpu_cache_info navi10_cache_info[] = {
 		/* Scalar L1 Instruction Cache per SQC */
 		.cache_size = 32,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_INST_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -388,6 +420,7 @@ static struct kfd_gpu_cache_info navi10_cache_info[] = {
 		/* Scalar L1 Data Cache per SQC */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -397,6 +430,7 @@ static struct kfd_gpu_cache_info navi10_cache_info[] = {
 		/* GL1 Data Cache per SA */
 		.cache_size = 128,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -406,6 +440,7 @@ static struct kfd_gpu_cache_info navi10_cache_info[] = {
 		/* L2 Data Cache per GPU (Total Tex Cache) */
 		.cache_size = 4096,
 		.cache_level = 2,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -418,6 +453,7 @@ static struct kfd_gpu_cache_info vangogh_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -427,6 +463,7 @@ static struct kfd_gpu_cache_info vangogh_cache_info[] = {
 		/* Scalar L1 Instruction Cache per SQC */
 		.cache_size = 32,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_INST_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -436,6 +473,7 @@ static struct kfd_gpu_cache_info vangogh_cache_info[] = {
 		/* Scalar L1 Data Cache per SQC */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -445,6 +483,7 @@ static struct kfd_gpu_cache_info vangogh_cache_info[] = {
 		/* GL1 Data Cache per SA */
 		.cache_size = 128,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -454,6 +493,7 @@ static struct kfd_gpu_cache_info vangogh_cache_info[] = {
 		/* L2 Data Cache per GPU (Total Tex Cache) */
 		.cache_size = 1024,
 		.cache_level = 2,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -466,6 +506,7 @@ static struct kfd_gpu_cache_info navi14_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -475,6 +516,7 @@ static struct kfd_gpu_cache_info navi14_cache_info[] = {
 		/* Scalar L1 Instruction Cache per SQC */
 		.cache_size = 32,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_INST_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -484,6 +526,7 @@ static struct kfd_gpu_cache_info navi14_cache_info[] = {
 		/* Scalar L1 Data Cache per SQC */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -493,6 +536,7 @@ static struct kfd_gpu_cache_info navi14_cache_info[] = {
 		/* GL1 Data Cache per SA */
 		.cache_size = 128,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -502,6 +546,7 @@ static struct kfd_gpu_cache_info navi14_cache_info[] = {
 		/* L2 Data Cache per GPU (Total Tex Cache) */
 		.cache_size = 2048,
 		.cache_level = 2,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -514,6 +559,7 @@ static struct kfd_gpu_cache_info sienna_cichlid_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -523,6 +569,7 @@ static struct kfd_gpu_cache_info sienna_cichlid_cache_info[] = {
 		/* Scalar L1 Instruction Cache per SQC */
 		.cache_size = 32,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_INST_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -532,6 +579,7 @@ static struct kfd_gpu_cache_info sienna_cichlid_cache_info[] = {
 		/* Scalar L1 Data Cache per SQC */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -541,6 +589,7 @@ static struct kfd_gpu_cache_info sienna_cichlid_cache_info[] = {
 		/* GL1 Data Cache per SA */
 		.cache_size = 128,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -550,6 +599,7 @@ static struct kfd_gpu_cache_info sienna_cichlid_cache_info[] = {
 		/* L2 Data Cache per GPU (Total Tex Cache) */
 		.cache_size = 4096,
 		.cache_level = 2,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -559,6 +609,7 @@ static struct kfd_gpu_cache_info sienna_cichlid_cache_info[] = {
 		/* L3 Data Cache per GPU */
 		.cache_size = 128*1024,
 		.cache_level = 3,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -571,6 +622,7 @@ static struct kfd_gpu_cache_info navy_flounder_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -580,6 +632,7 @@ static struct kfd_gpu_cache_info navy_flounder_cache_info[] = {
 		/* Scalar L1 Instruction Cache per SQC */
 		.cache_size = 32,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_INST_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -589,6 +642,7 @@ static struct kfd_gpu_cache_info navy_flounder_cache_info[] = {
 		/* Scalar L1 Data Cache per SQC */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -598,6 +652,7 @@ static struct kfd_gpu_cache_info navy_flounder_cache_info[] = {
 		/* GL1 Data Cache per SA */
 		.cache_size = 128,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -607,6 +662,7 @@ static struct kfd_gpu_cache_info navy_flounder_cache_info[] = {
 		/* L2 Data Cache per GPU (Total Tex Cache) */
 		.cache_size = 3072,
 		.cache_level = 2,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -616,6 +672,7 @@ static struct kfd_gpu_cache_info navy_flounder_cache_info[] = {
 		/* L3 Data Cache per GPU */
 		.cache_size = 96*1024,
 		.cache_level = 3,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -628,6 +685,7 @@ static struct kfd_gpu_cache_info dimgrey_cavefish_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -637,6 +695,7 @@ static struct kfd_gpu_cache_info dimgrey_cavefish_cache_info[] = {
 		/* Scalar L1 Instruction Cache per SQC */
 		.cache_size = 32,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_INST_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -646,6 +705,7 @@ static struct kfd_gpu_cache_info dimgrey_cavefish_cache_info[] = {
 		/* Scalar L1 Data Cache per SQC */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -655,6 +715,7 @@ static struct kfd_gpu_cache_info dimgrey_cavefish_cache_info[] = {
 		/* GL1 Data Cache per SA */
 		.cache_size = 128,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -664,6 +725,7 @@ static struct kfd_gpu_cache_info dimgrey_cavefish_cache_info[] = {
 		/* L2 Data Cache per GPU (Total Tex Cache) */
 		.cache_size = 2048,
 		.cache_level = 2,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -673,6 +735,7 @@ static struct kfd_gpu_cache_info dimgrey_cavefish_cache_info[] = {
 		/* L3 Data Cache per GPU */
 		.cache_size = 32*1024,
 		.cache_level = 3,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -685,6 +748,7 @@ static struct kfd_gpu_cache_info beige_goby_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -694,6 +758,7 @@ static struct kfd_gpu_cache_info beige_goby_cache_info[] = {
 		/* Scalar L1 Instruction Cache per SQC */
 		.cache_size = 32,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_INST_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -703,6 +768,7 @@ static struct kfd_gpu_cache_info beige_goby_cache_info[] = {
 		/* Scalar L1 Data Cache per SQC */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -712,6 +778,7 @@ static struct kfd_gpu_cache_info beige_goby_cache_info[] = {
 		/* GL1 Data Cache per SA */
 		.cache_size = 128,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -721,6 +788,7 @@ static struct kfd_gpu_cache_info beige_goby_cache_info[] = {
 		/* L2 Data Cache per GPU (Total Tex Cache) */
 		.cache_size = 1024,
 		.cache_level = 2,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -730,6 +798,7 @@ static struct kfd_gpu_cache_info beige_goby_cache_info[] = {
 		/* L3 Data Cache per GPU */
 		.cache_size = 16*1024,
 		.cache_level = 3,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -742,6 +811,7 @@ static struct kfd_gpu_cache_info yellow_carp_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -751,6 +821,7 @@ static struct kfd_gpu_cache_info yellow_carp_cache_info[] = {
 		/* Scalar L1 Instruction Cache per SQC */
 		.cache_size = 32,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_INST_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -760,6 +831,7 @@ static struct kfd_gpu_cache_info yellow_carp_cache_info[] = {
 		/* Scalar L1 Data Cache per SQC */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -769,6 +841,7 @@ static struct kfd_gpu_cache_info yellow_carp_cache_info[] = {
 		/* GL1 Data Cache per SA */
 		.cache_size = 128,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -778,6 +851,7 @@ static struct kfd_gpu_cache_info yellow_carp_cache_info[] = {
 		/* L2 Data Cache per GPU (Total Tex Cache) */
 		.cache_size = 2048,
 		.cache_level = 2,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -790,6 +864,7 @@ static struct kfd_gpu_cache_info gfx1037_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -799,6 +874,7 @@ static struct kfd_gpu_cache_info gfx1037_cache_info[] = {
 		/* Scalar L1 Instruction Cache per SQC */
 		.cache_size = 32,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_INST_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -808,6 +884,7 @@ static struct kfd_gpu_cache_info gfx1037_cache_info[] = {
 		/* Scalar L1 Data Cache per SQC */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -817,6 +894,7 @@ static struct kfd_gpu_cache_info gfx1037_cache_info[] = {
 		/* GL1 Data Cache per SA */
 		.cache_size = 128,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -826,6 +904,7 @@ static struct kfd_gpu_cache_info gfx1037_cache_info[] = {
 		/* L2 Data Cache per GPU (Total Tex Cache) */
 		.cache_size = 256,
 		.cache_level = 2,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -838,6 +917,7 @@ static struct kfd_gpu_cache_info gc_10_3_6_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 			  CRAT_CACHE_FLAGS_DATA_CACHE |
 			  CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -847,6 +927,7 @@ static struct kfd_gpu_cache_info gc_10_3_6_cache_info[] = {
 		/* Scalar L1 Instruction Cache per SQC */
 		.cache_size = 32,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 			  CRAT_CACHE_FLAGS_INST_CACHE |
 			  CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -856,6 +937,7 @@ static struct kfd_gpu_cache_info gc_10_3_6_cache_info[] = {
 		/* Scalar L1 Data Cache per SQC */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 			  CRAT_CACHE_FLAGS_DATA_CACHE |
 			  CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -865,6 +947,7 @@ static struct kfd_gpu_cache_info gc_10_3_6_cache_info[] = {
 		/* GL1 Data Cache per SA */
 		.cache_size = 128,
 		.cache_level = 1,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 			  CRAT_CACHE_FLAGS_DATA_CACHE |
 			  CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -874,6 +957,7 @@ static struct kfd_gpu_cache_info gc_10_3_6_cache_info[] = {
 		/* L2 Data Cache per GPU (Total Tex Cache) */
 		.cache_size = 256,
 		.cache_level = 2,
+		.cache_line_size = 128,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 			  CRAT_CACHE_FLAGS_DATA_CACHE |
 			  CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -886,6 +970,7 @@ static struct kfd_gpu_cache_info dummy_cache_info[] = {
 		/* TCP L1 Cache per CU */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -895,6 +980,7 @@ static struct kfd_gpu_cache_info dummy_cache_info[] = {
 		/* Scalar L1 Instruction Cache per SQC */
 		.cache_size = 32,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_INST_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -904,6 +990,7 @@ static struct kfd_gpu_cache_info dummy_cache_info[] = {
 		/* Scalar L1 Data Cache per SQC */
 		.cache_size = 16,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -913,6 +1000,7 @@ static struct kfd_gpu_cache_info dummy_cache_info[] = {
 		/* GL1 Data Cache per SA */
 		.cache_size = 128,
 		.cache_level = 1,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -922,6 +1010,7 @@ static struct kfd_gpu_cache_info dummy_cache_info[] = {
 		/* L2 Data Cache per GPU (Total Tex Cache) */
 		.cache_size = 2048,
 		.cache_level = 2,
+		.cache_line_size = 64,
 		.flags = (CRAT_CACHE_FLAGS_ENABLED |
 				CRAT_CACHE_FLAGS_DATA_CACHE |
 				CRAT_CACHE_FLAGS_SIMD_CACHE),
@@ -1166,7 +1255,7 @@ static int kfd_parse_subtype_iolink(struct crat_subtype_iolink *iolink,
 			if (props->iolink_type == CRAT_IOLINK_TYPE_PCIEXPRESS)
 				props->weight = 20;
 			else if (props->iolink_type == CRAT_IOLINK_TYPE_XGMI)
-				props->weight = 15 * iolink->num_hops_xgmi;
+				props->weight = iolink->weight_xgmi;
 			else
 				props->weight = node_distance(id_from, id_to);
 
@@ -1334,6 +1423,7 @@ err:
 
 
 static int kfd_fill_gpu_cache_info_from_gfx_config(struct kfd_dev *kdev,
+						   bool cache_line_size_missing,
 						   struct kfd_gpu_cache_info *pcache_info)
 {
 	struct amdgpu_device *adev = kdev->adev;
@@ -1346,7 +1436,10 @@ static int kfd_fill_gpu_cache_info_from_gfx_config(struct kfd_dev *kdev,
 		pcache_info[i].flags = (CRAT_CACHE_FLAGS_ENABLED |
 					CRAT_CACHE_FLAGS_DATA_CACHE |
 					CRAT_CACHE_FLAGS_SIMD_CACHE);
-		pcache_info[0].num_cu_shared = adev->gfx.config.gc_num_tcp_per_wpg / 2;
+		pcache_info[i].num_cu_shared = adev->gfx.config.gc_num_tcp_per_wpg / 2;
+		pcache_info[i].cache_line_size = adev->gfx.config.gc_tcp_cache_line_size;
+		if (cache_line_size_missing && !pcache_info[i].cache_line_size)
+			pcache_info[i].cache_line_size = 128;
 		i++;
 	}
 	/* Scalar L1 Instruction Cache per SQC */
@@ -1358,6 +1451,9 @@ static int kfd_fill_gpu_cache_info_from_gfx_config(struct kfd_dev *kdev,
 					CRAT_CACHE_FLAGS_INST_CACHE |
 					CRAT_CACHE_FLAGS_SIMD_CACHE);
 		pcache_info[i].num_cu_shared = adev->gfx.config.gc_num_sqc_per_wgp * 2;
+		pcache_info[i].cache_line_size = adev->gfx.config.gc_instruction_cache_line_size;
+		if (cache_line_size_missing && !pcache_info[i].cache_line_size)
+			pcache_info[i].cache_line_size = 128;
 		i++;
 	}
 	/* Scalar L1 Data Cache per SQC */
@@ -1368,6 +1464,9 @@ static int kfd_fill_gpu_cache_info_from_gfx_config(struct kfd_dev *kdev,
 					CRAT_CACHE_FLAGS_DATA_CACHE |
 					CRAT_CACHE_FLAGS_SIMD_CACHE);
 		pcache_info[i].num_cu_shared = adev->gfx.config.gc_num_sqc_per_wgp * 2;
+		pcache_info[i].cache_line_size = adev->gfx.config.gc_scalar_data_cache_line_size;
+		if (cache_line_size_missing && !pcache_info[i].cache_line_size)
+			pcache_info[i].cache_line_size = 64;
 		i++;
 	}
 	/* GL1 Data Cache per SA */
@@ -1380,6 +1479,8 @@ static int kfd_fill_gpu_cache_info_from_gfx_config(struct kfd_dev *kdev,
 					CRAT_CACHE_FLAGS_DATA_CACHE |
 					CRAT_CACHE_FLAGS_SIMD_CACHE);
 		pcache_info[i].num_cu_shared = adev->gfx.config.max_cu_per_sh;
+		if (cache_line_size_missing)
+			pcache_info[i].cache_line_size = 128;
 		i++;
 	}
 	/* L2 Data Cache per GPU (Total Tex Cache) */
@@ -1390,12 +1491,82 @@ static int kfd_fill_gpu_cache_info_from_gfx_config(struct kfd_dev *kdev,
 					CRAT_CACHE_FLAGS_DATA_CACHE |
 					CRAT_CACHE_FLAGS_SIMD_CACHE);
 		pcache_info[i].num_cu_shared = adev->gfx.config.max_cu_per_sh;
+		pcache_info[i].cache_line_size = adev->gfx.config.gc_tcc_cache_line_size;
+		if (cache_line_size_missing && !pcache_info[i].cache_line_size)
+			pcache_info[i].cache_line_size = 128;
+		i++;
+	}
+	/* L3 Data Cache per GPU */
+	if (adev->gmc.mall_size) {
+		pcache_info[i].cache_size = adev->gmc.mall_size / 1024;
+		pcache_info[i].cache_level = 3;
+		pcache_info[i].flags = (CRAT_CACHE_FLAGS_ENABLED |
+					CRAT_CACHE_FLAGS_DATA_CACHE |
+					CRAT_CACHE_FLAGS_SIMD_CACHE);
+		pcache_info[i].num_cu_shared = adev->gfx.config.max_cu_per_sh;
+		pcache_info[i].cache_line_size = 64;
+		i++;
+	}
+	return i;
+}
+
+static int kfd_fill_gpu_cache_info_from_gfx_config_v2(struct kfd_dev *kdev,
+						   struct kfd_gpu_cache_info *pcache_info)
+{
+	struct amdgpu_device *adev = kdev->adev;
+	int i = 0;
+
+	/* TCP L1 Cache per CU */
+	if (adev->gfx.config.gc_tcp_size_per_cu) {
+		pcache_info[i].cache_size = adev->gfx.config.gc_tcp_size_per_cu;
+		pcache_info[i].cache_level = 1;
+		/* Cacheline size not available in IP discovery for gc943,gc944 */
+		pcache_info[i].cache_line_size = 128;
+		pcache_info[i].flags = (CRAT_CACHE_FLAGS_ENABLED |
+					CRAT_CACHE_FLAGS_DATA_CACHE |
+					CRAT_CACHE_FLAGS_SIMD_CACHE);
+		pcache_info[i].num_cu_shared = 1;
+		i++;
+	}
+	/* Scalar L1 Instruction Cache per SQC */
+	if (adev->gfx.config.gc_l1_instruction_cache_size_per_sqc) {
+		pcache_info[i].cache_size =
+			adev->gfx.config.gc_l1_instruction_cache_size_per_sqc;
+		pcache_info[i].cache_level = 1;
+		pcache_info[i].cache_line_size = 64;
+		pcache_info[i].flags = (CRAT_CACHE_FLAGS_ENABLED |
+					CRAT_CACHE_FLAGS_INST_CACHE |
+					CRAT_CACHE_FLAGS_SIMD_CACHE);
+		pcache_info[i].num_cu_shared = adev->gfx.config.gc_num_cu_per_sqc;
+		i++;
+	}
+	/* Scalar L1 Data Cache per SQC */
+	if (adev->gfx.config.gc_l1_data_cache_size_per_sqc) {
+		pcache_info[i].cache_size = adev->gfx.config.gc_l1_data_cache_size_per_sqc;
+		pcache_info[i].cache_level = 1;
+		pcache_info[i].cache_line_size = 64;
+		pcache_info[i].flags = (CRAT_CACHE_FLAGS_ENABLED |
+					CRAT_CACHE_FLAGS_DATA_CACHE |
+					CRAT_CACHE_FLAGS_SIMD_CACHE);
+		pcache_info[i].num_cu_shared = adev->gfx.config.gc_num_cu_per_sqc;
+		i++;
+	}
+	/* L2 Data Cache per GPU (Total Tex Cache) */
+	if (adev->gfx.config.gc_tcc_size) {
+		pcache_info[i].cache_size = adev->gfx.config.gc_tcc_size;
+		pcache_info[i].cache_level = 2;
+		pcache_info[i].cache_line_size = 128;
+		pcache_info[i].flags = (CRAT_CACHE_FLAGS_ENABLED |
+					CRAT_CACHE_FLAGS_DATA_CACHE |
+					CRAT_CACHE_FLAGS_SIMD_CACHE);
+		pcache_info[i].num_cu_shared = adev->gfx.config.max_cu_per_sh;
 		i++;
 	}
 	/* L3 Data Cache per GPU */
 	if (adev->gmc.mall_size) {
 		pcache_info[i].cache_size = adev->gmc.mall_size / 1024;
 		pcache_info[i].cache_level = 3;
+		pcache_info[i].cache_line_size = 64;
 		pcache_info[i].flags = (CRAT_CACHE_FLAGS_ENABLED |
 					CRAT_CACHE_FLAGS_DATA_CACHE |
 					CRAT_CACHE_FLAGS_SIMD_CACHE);
@@ -1405,9 +1576,10 @@ static int kfd_fill_gpu_cache_info_from_gfx_config(struct kfd_dev *kdev,
 	return i;
 }
 
-int kfd_get_gpu_cache_info(struct kfd_dev *kdev, struct kfd_gpu_cache_info **pcache_info)
+int kfd_get_gpu_cache_info(struct kfd_node *kdev, struct kfd_gpu_cache_info **pcache_info)
 {
 	int num_of_cache_types = 0;
+	bool cache_line_size_missing = false;
 
 	switch (kdev->adev->asic_type) {
 	case CHIP_KAVERI:
@@ -1462,10 +1634,16 @@ int kfd_get_gpu_cache_info(struct kfd_dev *kdev, struct kfd_gpu_cache_info **pca
 			num_of_cache_types = ARRAY_SIZE(vega20_cache_info);
 			break;
 		case IP_VERSION(9, 4, 2):
-		case IP_VERSION(9, 4, 3):
 			*pcache_info = aldebaran_cache_info;
 			num_of_cache_types = ARRAY_SIZE(aldebaran_cache_info);
 			break;
+		case IP_VERSION(9, 4, 3):
+		case IP_VERSION(9, 4, 4):
+		case IP_VERSION(9, 5, 0):
+			num_of_cache_types =
+				kfd_fill_gpu_cache_info_from_gfx_config_v2(kdev->kfd,
+									*pcache_info);
+			break;
 		case IP_VERSION(9, 1, 0):
 		case IP_VERSION(9, 2, 2):
 			*pcache_info = raven_cache_info;
@@ -1523,8 +1701,21 @@ int kfd_get_gpu_cache_info(struct kfd_dev *kdev, struct kfd_gpu_cache_info **pca
 		case IP_VERSION(11, 0, 2):
 		case IP_VERSION(11, 0, 3):
 		case IP_VERSION(11, 0, 4):
+		case IP_VERSION(11, 5, 0):
+		case IP_VERSION(11, 5, 1):
+		case IP_VERSION(11, 5, 2):
+		case IP_VERSION(11, 5, 3):
+			/* Cacheline size not available in IP discovery for gc11.
+			 * kfd_fill_gpu_cache_info_from_gfx_config to hard code it
+			 */
+			cache_line_size_missing = true;
+			fallthrough;
+		case IP_VERSION(12, 0, 0):
+		case IP_VERSION(12, 0, 1):
 			num_of_cache_types =
-				kfd_fill_gpu_cache_info_from_gfx_config(kdev, *pcache_info);
+				kfd_fill_gpu_cache_info_from_gfx_config(kdev->kfd,
+									cache_line_size_missing,
+									*pcache_info);
 			break;
 		default:
 			*pcache_info = dummy_cache_info;
@@ -1536,76 +1727,6 @@ int kfd_get_gpu_cache_info(struct kfd_dev *kdev, struct kfd_gpu_cache_info **pca
 	return num_of_cache_types;
 }
 
-static bool kfd_ignore_crat(void)
-{
-	bool ret;
-
-	if (ignore_crat)
-		return true;
-
-#ifndef KFD_SUPPORT_IOMMU_V2
-	ret = true;
-#else
-	ret = false;
-#endif
-
-	return ret;
-}
-
-/*
- * kfd_create_crat_image_acpi - Allocates memory for CRAT image and
- * copies CRAT from ACPI (if available).
- * NOTE: Call kfd_destroy_crat_image to free CRAT image memory
- *
- *	@crat_image: CRAT read from ACPI. If no CRAT in ACPI then
- *		     crat_image will be NULL
- *	@size: [OUT] size of crat_image
- *
- *	Return 0 if successful else return error code
- */
-int kfd_create_crat_image_acpi(void **crat_image, size_t *size)
-{
-	struct acpi_table_header *crat_table;
-	acpi_status status;
-	void *pcrat_image;
-	int rc = 0;
-
-	if (!crat_image)
-		return -EINVAL;
-
-	*crat_image = NULL;
-
-	if (kfd_ignore_crat()) {
-		pr_info("CRAT table disabled by module option\n");
-		return -ENODATA;
-	}
-
-	/* Fetch the CRAT table from ACPI */
-	status = acpi_get_table(CRAT_SIGNATURE, 0, &crat_table);
-	if (status == AE_NOT_FOUND) {
-		pr_info("CRAT table not found\n");
-		return -ENODATA;
-	} else if (ACPI_FAILURE(status)) {
-		const char *err = acpi_format_exception(status);
-
-		pr_err("CRAT table error: %s\n", err);
-		return -EINVAL;
-	}
-
-	pcrat_image = kvmalloc(crat_table->length, GFP_KERNEL);
-	if (!pcrat_image) {
-		rc = -ENOMEM;
-		goto out;
-	}
-
-	memcpy(pcrat_image, crat_table, crat_table->length);
-	*crat_image = pcrat_image;
-	*size = crat_table->length;
-out:
-	acpi_put_table(crat_table);
-	return rc;
-}
-
 /* Memory required to create Virtual CRAT.
  * Since there is no easy way to predict the amount of memory required, the
  * following amount is allocated for GPU Virtual CRAT. This is
@@ -1858,7 +1979,7 @@ static int kfd_create_vcrat_image_cpu(void *pcrat_image, size_t *size)
 }
 
 static int kfd_fill_gpu_memory_affinity(int *avail_size,
-		struct kfd_dev *kdev, uint8_t type, uint64_t size,
+		struct kfd_node *kdev, uint8_t type, uint64_t size,
 		struct crat_subtype_memory *sub_type_hdr,
 		uint32_t proximity_domain,
 		const struct kfd_local_mem_info *local_mem_info)
@@ -1887,7 +2008,7 @@ static int kfd_fill_gpu_memory_affinity(int *avail_size,
 }
 
 #ifdef CONFIG_ACPI_NUMA
-static void kfd_find_numa_node_in_srat(struct kfd_dev *kdev)
+static void kfd_find_numa_node_in_srat(struct kfd_node *kdev)
 {
 	struct acpi_table_header *table_header = NULL;
 	struct acpi_subtable_header *sub_header = NULL;
@@ -1972,6 +2093,9 @@ static void kfd_find_numa_node_in_srat(struct kfd_dev *kdev)
 }
 #endif
 
+#define KFD_CRAT_INTRA_SOCKET_WEIGHT	13
+#define KFD_CRAT_XGMI_WEIGHT		15
+
 /* kfd_fill_gpu_direct_io_link - Fill in direct io link from GPU
  * to its NUMA node
  *	@avail_size: Available size in the memory
@@ -1982,7 +2106,7 @@ static void kfd_find_numa_node_in_srat(struct kfd_dev *kdev)
  *	Return 0 if successful else return -ve value
  */
 static int kfd_fill_gpu_direct_io_link_to_cpu(int *avail_size,
-			struct kfd_dev *kdev,
+			struct kfd_node *kdev,
 			struct crat_subtype_iolink *sub_type_hdr,
 			uint32_t proximity_domain)
 {
@@ -2002,7 +2126,13 @@ static int kfd_fill_gpu_direct_io_link_to_cpu(int *avail_size,
 	/* Fill in IOLINK subtype.
 	 * TODO: Fill-in other fields of iolink subtype
 	 */
-	if (kdev->adev->gmc.xgmi.connected_to_cpu) {
+	if (kdev->adev->gmc.xgmi.connected_to_cpu ||
+	    (KFD_GC_VERSION(kdev) == IP_VERSION(9, 4, 3) &&
+	     kdev->adev->smuio.funcs->get_pkg_type(kdev->adev) ==
+	     AMDGPU_PKG_TYPE_APU)) {
+		bool ext_cpu = KFD_GC_VERSION(kdev) != IP_VERSION(9, 4, 3);
+		int mem_bw = 819200, weight = ext_cpu ? KFD_CRAT_XGMI_WEIGHT :
+							KFD_CRAT_INTRA_SOCKET_WEIGHT;
 		/*
 		 * with host gpu xgmi link, host can access gpu memory whether
 		 * or not pcie bar type is large, so always create bidirectional
@@ -2010,13 +2140,16 @@ static int kfd_fill_gpu_direct_io_link_to_cpu(int *avail_size,
 		 */
 		sub_type_hdr->flags |= CRAT_IOLINK_FLAGS_BI_DIRECTIONAL;
 		sub_type_hdr->io_interface_type = CRAT_IOLINK_TYPE_XGMI;
-		sub_type_hdr->num_hops_xgmi = 1;
-		if (KFD_GC_VERSION(kdev) == IP_VERSION(9, 4, 2)) {
-			sub_type_hdr->minimum_bandwidth_mbs =
-					amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(
-							kdev->adev, NULL, true);
-			sub_type_hdr->maximum_bandwidth_mbs =
-					sub_type_hdr->minimum_bandwidth_mbs;
+		sub_type_hdr->weight_xgmi = weight;
+		if (ext_cpu) {
+			amdgpu_xgmi_get_bandwidth(kdev->adev, NULL,
+						  AMDGPU_XGMI_BW_MODE_PER_LINK,
+						  AMDGPU_XGMI_BW_UNIT_MBYTES,
+						  &sub_type_hdr->minimum_bandwidth_mbs,
+						  &sub_type_hdr->maximum_bandwidth_mbs);
+		} else {
+			sub_type_hdr->minimum_bandwidth_mbs = mem_bw;
+			sub_type_hdr->maximum_bandwidth_mbs = mem_bw;
 		}
 	} else {
 		sub_type_hdr->io_interface_type = CRAT_IOLINK_TYPE_PCIEXPRESS;
@@ -2029,7 +2162,8 @@ static int kfd_fill_gpu_direct_io_link_to_cpu(int *avail_size,
 	sub_type_hdr->proximity_domain_from = proximity_domain;
 
 #ifdef CONFIG_ACPI_NUMA
-	if (kdev->adev->pdev->dev.numa_node == NUMA_NO_NODE)
+	if (kdev->adev->pdev->dev.numa_node == NUMA_NO_NODE &&
+	    num_possible_nodes() > 1)
 		kfd_find_numa_node_in_srat(kdev);
 #endif
 #ifdef CONFIG_NUMA
@@ -2044,12 +2178,14 @@ static int kfd_fill_gpu_direct_io_link_to_cpu(int *avail_size,
 }
 
 static int kfd_fill_gpu_xgmi_link_to_gpu(int *avail_size,
-			struct kfd_dev *kdev,
-			struct kfd_dev *peer_kdev,
+			struct kfd_node *kdev,
+			struct kfd_node *peer_kdev,
 			struct crat_subtype_iolink *sub_type_hdr,
 			uint32_t proximity_domain_from,
 			uint32_t proximity_domain_to)
 {
+	bool use_ta_info = kdev->kfd->num_nodes == 1;
+
 	*avail_size -= sizeof(struct crat_subtype_iolink);
 	if (*avail_size < 0)
 		return -ENOMEM;
@@ -2064,12 +2200,25 @@ static int kfd_fill_gpu_xgmi_link_to_gpu(int *avail_size,
 	sub_type_hdr->io_interface_type = CRAT_IOLINK_TYPE_XGMI;
 	sub_type_hdr->proximity_domain_from = proximity_domain_from;
 	sub_type_hdr->proximity_domain_to = proximity_domain_to;
-	sub_type_hdr->num_hops_xgmi =
-		amdgpu_amdkfd_get_xgmi_hops_count(kdev->adev, peer_kdev->adev);
-	sub_type_hdr->maximum_bandwidth_mbs =
-		amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev->adev, peer_kdev->adev, false);
-	sub_type_hdr->minimum_bandwidth_mbs = sub_type_hdr->maximum_bandwidth_mbs ?
-		amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev->adev, NULL, true) : 0;
+
+	if (use_ta_info) {
+		sub_type_hdr->weight_xgmi = KFD_CRAT_XGMI_WEIGHT *
+			amdgpu_xgmi_get_hops_count(kdev->adev, peer_kdev->adev);
+		amdgpu_xgmi_get_bandwidth(kdev->adev, peer_kdev->adev,
+					  AMDGPU_XGMI_BW_MODE_PER_PEER,
+					  AMDGPU_XGMI_BW_UNIT_MBYTES,
+					  &sub_type_hdr->minimum_bandwidth_mbs,
+					  &sub_type_hdr->maximum_bandwidth_mbs);
+	} else {
+		bool is_single_hop = kdev->kfd == peer_kdev->kfd;
+		int weight = is_single_hop ? KFD_CRAT_INTRA_SOCKET_WEIGHT :
+			(2 * KFD_CRAT_INTRA_SOCKET_WEIGHT) + KFD_CRAT_XGMI_WEIGHT;
+		int mem_bw = 819200;
+
+		sub_type_hdr->weight_xgmi = weight;
+		sub_type_hdr->maximum_bandwidth_mbs = is_single_hop ? mem_bw : 0;
+		sub_type_hdr->minimum_bandwidth_mbs = is_single_hop ? mem_bw : 0;
+	}
 
 	return 0;
 }
@@ -2081,15 +2230,16 @@ static int kfd_fill_gpu_xgmi_link_to_gpu(int *avail_size,
  *		[OUT] actual size of data filled in crat_image
  */
 static int kfd_create_vcrat_image_gpu(void *pcrat_image,
-				      size_t *size, struct kfd_dev *kdev,
+				      size_t *size, struct kfd_node *kdev,
 				      uint32_t proximity_domain)
 {
 	struct crat_header *crat_table = (struct crat_header *)pcrat_image;
+	struct amdgpu_gfx_config *gfx_info = &kdev->adev->gfx.config;
+	struct amdgpu_cu_info *cu_info = &kdev->adev->gfx.cu_info;
 	struct crat_subtype_generic *sub_type_hdr;
 	struct kfd_local_mem_info local_mem_info;
 	struct kfd_topology_device *peer_dev;
 	struct crat_subtype_computeunit *cu;
-	struct kfd_cu_info cu_info;
 	int avail_size = *size;
 	uint32_t total_num_of_cu;
 	uint32_t nid = 0;
@@ -2102,9 +2252,6 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image,
 	 * Modify length and total_entries as subunits are added.
 	 */
 	avail_size -= sizeof(struct crat_header);
-	if (avail_size < 0)
-		return -ENOMEM;
-
 	memset(crat_table, 0, sizeof(struct crat_header));
 
 	memcpy(&crat_table->signature, CRAT_SIGNATURE,
@@ -2118,9 +2265,6 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image,
 	 * First fill in the sub type header and then sub type data
 	 */
 	avail_size -= sizeof(struct crat_subtype_computeunit);
-	if (avail_size < 0)
-		return -ENOMEM;
-
 	sub_type_hdr = (struct crat_subtype_generic *)(crat_table + 1);
 	memset(sub_type_hdr, 0, sizeof(struct crat_subtype_computeunit));
 
@@ -2133,29 +2277,23 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image,
 	cu->flags |= CRAT_CU_FLAGS_GPU_PRESENT;
 	cu->proximity_domain = proximity_domain;
 
-	amdgpu_amdkfd_get_cu_info(kdev->adev, &cu_info);
-	cu->num_simd_per_cu = cu_info.simd_per_cu;
-	cu->num_simd_cores = cu_info.simd_per_cu * cu_info.cu_active_number;
-	cu->max_waves_simd = cu_info.max_waves_per_simd;
+	cu->num_simd_per_cu = cu_info->simd_per_cu;
+	cu->num_simd_cores = cu_info->simd_per_cu *
+			(cu_info->number / kdev->kfd->num_nodes);
+	cu->max_waves_simd = cu_info->max_waves_per_simd;
 
-	cu->wave_front_size = cu_info.wave_front_size;
-	cu->array_count = cu_info.num_shader_arrays_per_engine *
-		cu_info.num_shader_engines;
-	total_num_of_cu = (cu->array_count * cu_info.num_cu_per_sh);
+	cu->wave_front_size = cu_info->wave_front_size;
+	cu->array_count = gfx_info->max_sh_per_se *
+		gfx_info->max_shader_engines;
+	total_num_of_cu = (cu->array_count * gfx_info->max_cu_per_sh);
 	cu->processor_id_low = get_and_inc_gpu_processor_id(total_num_of_cu);
-	cu->num_cu_per_array = cu_info.num_cu_per_sh;
-	cu->max_slots_scatch_cu = cu_info.max_scratch_slots_per_cu;
-	cu->num_banks = cu_info.num_shader_engines;
-	cu->lds_size_in_kb = cu_info.lds_size;
+	cu->num_cu_per_array = gfx_info->max_cu_per_sh;
+	cu->max_slots_scatch_cu = cu_info->max_scratch_slots_per_cu;
+	cu->num_banks = gfx_info->max_shader_engines;
+	cu->lds_size_in_kb = cu_info->lds_size;
 
 	cu->hsa_capability = 0;
 
-	/* Check if this node supports IOMMU. During parsing this flag will
-	 * translate to HSA_CAP_ATS_PRESENT
-	 */
-	if (!kfd_iommu_check_device(kdev))
-		cu->hsa_capability |= CRAT_CU_FLAGS_IOMMU_PRESENT;
-
 	crat_table->length += sub_type_hdr->length;
 	crat_table->total_entries++;
 
@@ -2168,7 +2306,7 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image,
 	sub_type_hdr = (typeof(sub_type_hdr))((char *)sub_type_hdr +
 			sub_type_hdr->length);
 
-	if (debug_largebar)
+	if (kdev->adev->debug_largebar)
 		local_mem_info.local_mem_size_private = 0;
 
 	if (local_mem_info.local_mem_size_private == 0)
@@ -2216,12 +2354,14 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image,
 	 * (from other GPU to this GPU) will be added
 	 * in kfd_parse_subtype_iolink.
 	 */
-	if (kdev->hive_id) {
+	if (kdev->kfd->hive_id) {
 		for (nid = 0; nid < proximity_domain; ++nid) {
 			peer_dev = kfd_topology_device_by_proximity_domain_no_lock(nid);
 			if (!peer_dev->gpu)
 				continue;
-			if (peer_dev->gpu->hive_id != kdev->hive_id)
+			if (peer_dev->gpu->kfd->hive_id != kdev->kfd->hive_id)
+				continue;
+			if (!amdgpu_xgmi_get_is_sharing_enabled(kdev->adev, peer_dev->gpu->adev))
 				continue;
 			sub_type_hdr = (typeof(sub_type_hdr))(
 				(char *)sub_type_hdr +
@@ -2255,12 +2395,12 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image,
  *		(COMPUTE_UNIT_CPU | COMPUTE_UNIT_GPU) - Create VCRAT for APU
  *			-- this option is not currently implemented.
  *			The assumption is that all AMD APUs will have CRAT
- *	@kdev: Valid kfd_device required if flags contain COMPUTE_UNIT_GPU
+ *	@kdev: Valid kfd_node required if flags contain COMPUTE_UNIT_GPU
  *
  *	Return 0 if successful else return -ve value
  */
 int kfd_create_crat_image_virtual(void **crat_image, size_t *size,
-				  int flags, struct kfd_dev *kdev,
+				  int flags, struct kfd_node *kdev,
 				  uint32_t proximity_domain)
 {
 	void *pcrat_image = NULL;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.h b/drivers/gpu/drm/amd/amdkfd/kfd_crat.h
index 8d1e8ba58dee..a8ca7ecb6d27 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.h
@@ -42,8 +42,6 @@
 #define CRAT_OEMTABLEID_LENGTH	8
 #define CRAT_RESERVED_LENGTH	6
 
-#define CRAT_OEMID_64BIT_MASK ((1ULL << (CRAT_OEMID_LENGTH * 8)) - 1)
-
 /* Compute Unit flags */
 #define COMPUTE_UNIT_CPU	(1 << 0)  /* Create Virtual CRAT for CPU */
 #define COMPUTE_UNIT_GPU	(1 << 1)  /* Create Virtual CRAT for GPU */
@@ -79,6 +77,10 @@ struct crat_header {
 #define CRAT_SUBTYPE_IOLINK_AFFINITY		5
 #define CRAT_SUBTYPE_MAX			6
 
+/*
+ * Do not change the value of CRAT_SIBLINGMAP_SIZE from 32
+ * as it breaks the ABI.
+ */
 #define CRAT_SIBLINGMAP_SIZE	32
 
 /*
@@ -275,7 +277,7 @@ struct crat_subtype_iolink {
 	uint32_t	maximum_bandwidth_mbs;
 	uint32_t	recommended_transfer_size;
 	uint8_t		reserved2[CRAT_IOLINK_RESERVED_LENGTH - 1];
-	uint8_t		num_hops_xgmi;
+	uint8_t		weight_xgmi;
 };
 
 /*
@@ -293,26 +295,26 @@ struct crat_subtype_generic {
 
 #pragma pack()
 
-struct kfd_dev;
+struct kfd_node;
 
 /* Static table to describe GPU Cache information */
 struct kfd_gpu_cache_info {
 	uint32_t	cache_size;
 	uint32_t	cache_level;
+	uint32_t	cache_line_size;
 	uint32_t	flags;
 	/* Indicates how many Compute Units share this cache
 	 * within a SA. Value = 1 indicates the cache is not shared
 	 */
 	uint32_t	num_cu_shared;
 };
-int kfd_get_gpu_cache_info(struct kfd_dev *kdev, struct kfd_gpu_cache_info **pcache_info);
+int kfd_get_gpu_cache_info(struct kfd_node *kdev, struct kfd_gpu_cache_info **pcache_info);
 
-int kfd_create_crat_image_acpi(void **crat_image, size_t *size);
 void kfd_destroy_crat_image(void *crat_image);
 int kfd_parse_crat_table(void *crat_image, struct list_head *device_list,
 			 uint32_t proximity_domain);
 int kfd_create_crat_image_virtual(void **crat_image, size_t *size,
-				  int flags, struct kfd_dev *kdev,
+				  int flags, struct kfd_node *kdev,
 				  uint32_t proximity_domain);
 
 #endif /* KFD_CRAT_H_INCLUDED */
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debug.c b/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
new file mode 100644
index 000000000000..ba99e0f258ae
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_debug.c
@@ -0,0 +1,1152 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "kfd_debug.h"
+#include "kfd_device_queue_manager.h"
+#include "kfd_topology.h"
+#include <linux/file.h>
+#include <uapi/linux/kfd_ioctl.h>
+#include <uapi/linux/kfd_sysfs.h>
+
+#define MAX_WATCH_ADDRESSES	4
+
+int kfd_dbg_ev_query_debug_event(struct kfd_process *process,
+		      unsigned int *queue_id,
+		      unsigned int *gpu_id,
+		      uint64_t exception_clear_mask,
+		      uint64_t *event_status)
+{
+	struct process_queue_manager *pqm;
+	struct process_queue_node *pqn;
+	int i;
+
+	if (!(process && process->debug_trap_enabled))
+		return -ENODATA;
+
+	mutex_lock(&process->event_mutex);
+	*event_status = 0;
+	*queue_id = 0;
+	*gpu_id = 0;
+
+	/* find and report queue events */
+	pqm = &process->pqm;
+	list_for_each_entry(pqn, &pqm->queues, process_queue_list) {
+		uint64_t tmp = process->exception_enable_mask;
+
+		if (!pqn->q)
+			continue;
+
+		tmp &= pqn->q->properties.exception_status;
+
+		if (!tmp)
+			continue;
+
+		*event_status = pqn->q->properties.exception_status;
+		*queue_id = pqn->q->properties.queue_id;
+		*gpu_id = pqn->q->device->id;
+		pqn->q->properties.exception_status &= ~exception_clear_mask;
+		goto out;
+	}
+
+	/* find and report device events */
+	for (i = 0; i < process->n_pdds; i++) {
+		struct kfd_process_device *pdd = process->pdds[i];
+		uint64_t tmp = process->exception_enable_mask
+						& pdd->exception_status;
+
+		if (!tmp)
+			continue;
+
+		*event_status = pdd->exception_status;
+		*gpu_id = pdd->dev->id;
+		pdd->exception_status &= ~exception_clear_mask;
+		goto out;
+	}
+
+	/* report process events */
+	if (process->exception_enable_mask & process->exception_status) {
+		*event_status = process->exception_status;
+		process->exception_status &= ~exception_clear_mask;
+	}
+
+out:
+	mutex_unlock(&process->event_mutex);
+	return *event_status ? 0 : -EAGAIN;
+}
+
+void debug_event_write_work_handler(struct work_struct *work)
+{
+	struct kfd_process *process;
+
+	static const char write_data = '.';
+	loff_t pos = 0;
+
+	process = container_of(work,
+			struct kfd_process,
+			debug_event_workarea);
+
+	if (process->debug_trap_enabled && process->dbg_ev_file)
+		kernel_write(process->dbg_ev_file, &write_data, 1, &pos);
+}
+
+/* update process/device/queue exception status, write to descriptor
+ * only if exception_status is enabled.
+ */
+bool kfd_dbg_ev_raise(uint64_t event_mask,
+			struct kfd_process *process, struct kfd_node *dev,
+			unsigned int source_id, bool use_worker,
+			void *exception_data, size_t exception_data_size)
+{
+	struct process_queue_manager *pqm;
+	struct process_queue_node *pqn;
+	int i;
+	static const char write_data = '.';
+	loff_t pos = 0;
+	bool is_subscribed = true;
+
+	if (!(process && process->debug_trap_enabled))
+		return false;
+
+	mutex_lock(&process->event_mutex);
+
+	if (event_mask & KFD_EC_MASK_DEVICE) {
+		for (i = 0; i < process->n_pdds; i++) {
+			struct kfd_process_device *pdd = process->pdds[i];
+
+			if (pdd->dev != dev)
+				continue;
+
+			pdd->exception_status |= event_mask & KFD_EC_MASK_DEVICE;
+
+			if (event_mask & KFD_EC_MASK(EC_DEVICE_MEMORY_VIOLATION)) {
+				if (!pdd->vm_fault_exc_data) {
+					pdd->vm_fault_exc_data = kmemdup(
+							exception_data,
+							exception_data_size,
+							GFP_KERNEL);
+					if (!pdd->vm_fault_exc_data)
+						pr_debug("Failed to allocate exception data memory");
+				} else {
+					pr_debug("Debugger exception data not saved\n");
+					print_hex_dump_bytes("exception data: ",
+							DUMP_PREFIX_OFFSET,
+							exception_data,
+							exception_data_size);
+				}
+			}
+			break;
+		}
+	} else if (event_mask & KFD_EC_MASK_PROCESS) {
+		process->exception_status |= event_mask & KFD_EC_MASK_PROCESS;
+	} else {
+		pqm = &process->pqm;
+		list_for_each_entry(pqn, &pqm->queues,
+				process_queue_list) {
+			int target_id;
+
+			if (!pqn->q)
+				continue;
+
+			target_id = event_mask & KFD_EC_MASK(EC_QUEUE_NEW) ?
+					pqn->q->properties.queue_id :
+							pqn->q->doorbell_id;
+
+			if (pqn->q->device != dev || target_id != source_id)
+				continue;
+
+			pqn->q->properties.exception_status |= event_mask;
+			break;
+		}
+	}
+
+	if (process->exception_enable_mask & event_mask) {
+		if (use_worker)
+			schedule_work(&process->debug_event_workarea);
+		else
+			kernel_write(process->dbg_ev_file,
+					&write_data,
+					1,
+					&pos);
+	} else {
+		is_subscribed = false;
+	}
+
+	mutex_unlock(&process->event_mutex);
+
+	return is_subscribed;
+}
+
+/* set pending event queue entry from ring entry  */
+bool kfd_set_dbg_ev_from_interrupt(struct kfd_node *dev,
+				   unsigned int pasid,
+				   uint32_t doorbell_id,
+				   uint64_t trap_mask,
+				   void *exception_data,
+				   size_t exception_data_size)
+{
+	struct kfd_process *p;
+	struct kfd_process_device *pdd = NULL;
+	bool signaled_to_debugger_or_runtime = false;
+
+	p = kfd_lookup_process_by_pasid(pasid, &pdd);
+
+	if (!pdd)
+		return false;
+
+	if (!kfd_dbg_ev_raise(trap_mask, p, dev, doorbell_id, true,
+			      exception_data, exception_data_size)) {
+		struct process_queue_manager *pqm;
+		struct process_queue_node *pqn;
+
+		if (!!(trap_mask & KFD_EC_MASK_QUEUE) &&
+		       p->runtime_info.runtime_state == DEBUG_RUNTIME_STATE_ENABLED) {
+			mutex_lock(&p->mutex);
+
+			pqm = &p->pqm;
+			list_for_each_entry(pqn, &pqm->queues,
+							process_queue_list) {
+
+				if (!(pqn->q && pqn->q->device == dev &&
+				      pqn->q->doorbell_id == doorbell_id))
+					continue;
+
+				kfd_send_exception_to_runtime(p, pqn->q->properties.queue_id,
+							      trap_mask);
+
+				signaled_to_debugger_or_runtime = true;
+
+				break;
+			}
+
+			mutex_unlock(&p->mutex);
+		} else if (trap_mask & KFD_EC_MASK(EC_DEVICE_MEMORY_VIOLATION)) {
+			kfd_evict_process_device(pdd);
+			kfd_signal_vm_fault_event(pdd, NULL, exception_data);
+
+			signaled_to_debugger_or_runtime = true;
+		}
+	} else {
+		signaled_to_debugger_or_runtime = true;
+	}
+
+	kfd_unref_process(p);
+
+	return signaled_to_debugger_or_runtime;
+}
+
+int kfd_dbg_send_exception_to_runtime(struct kfd_process *p,
+					unsigned int dev_id,
+					unsigned int queue_id,
+					uint64_t error_reason)
+{
+	if (error_reason & KFD_EC_MASK(EC_DEVICE_MEMORY_VIOLATION)) {
+		struct kfd_process_device *pdd = NULL;
+		struct kfd_hsa_memory_exception_data *data;
+		int i;
+
+		for (i = 0; i < p->n_pdds; i++) {
+			if (p->pdds[i]->dev->id == dev_id) {
+				pdd = p->pdds[i];
+				break;
+			}
+		}
+
+		if (!pdd)
+			return -ENODEV;
+
+		data = (struct kfd_hsa_memory_exception_data *)
+						pdd->vm_fault_exc_data;
+
+		kfd_evict_process_device(pdd);
+		kfd_signal_vm_fault_event(pdd, NULL, data);
+		error_reason &= ~KFD_EC_MASK(EC_DEVICE_MEMORY_VIOLATION);
+	}
+
+	if (error_reason & (KFD_EC_MASK(EC_PROCESS_RUNTIME))) {
+		/*
+		 * block should only happen after the debugger receives runtime
+		 * enable notice.
+		 */
+		up(&p->runtime_enable_sema);
+		error_reason &= ~KFD_EC_MASK(EC_PROCESS_RUNTIME);
+	}
+
+	if (error_reason)
+		return kfd_send_exception_to_runtime(p, queue_id, error_reason);
+
+	return 0;
+}
+
+static int kfd_dbg_set_queue_workaround(struct queue *q, bool enable)
+{
+	struct mqd_update_info minfo = {0};
+	int err;
+
+	if (!q)
+		return 0;
+
+	if (!kfd_dbg_has_cwsr_workaround(q->device))
+		return 0;
+
+	if (enable && q->properties.is_user_cu_masked)
+		return -EBUSY;
+
+	minfo.update_flag = enable ? UPDATE_FLAG_DBG_WA_ENABLE : UPDATE_FLAG_DBG_WA_DISABLE;
+
+	q->properties.is_dbg_wa = enable;
+	err = q->device->dqm->ops.update_queue(q->device->dqm, q, &minfo);
+	if (err)
+		q->properties.is_dbg_wa = false;
+
+	return err;
+}
+
+static int kfd_dbg_set_workaround(struct kfd_process *target, bool enable)
+{
+	struct process_queue_manager *pqm = &target->pqm;
+	struct process_queue_node *pqn;
+	int r = 0;
+
+	list_for_each_entry(pqn, &pqm->queues, process_queue_list) {
+		r = kfd_dbg_set_queue_workaround(pqn->q, enable);
+		if (enable && r)
+			goto unwind;
+	}
+
+	return 0;
+
+unwind:
+	list_for_each_entry(pqn, &pqm->queues, process_queue_list)
+		kfd_dbg_set_queue_workaround(pqn->q, false);
+
+	if (enable)
+		target->runtime_info.runtime_state = r == -EBUSY ?
+				DEBUG_RUNTIME_STATE_ENABLED_BUSY :
+				DEBUG_RUNTIME_STATE_ENABLED_ERROR;
+
+	return r;
+}
+
+int kfd_dbg_set_mes_debug_mode(struct kfd_process_device *pdd, bool sq_trap_en)
+{
+	uint32_t spi_dbg_cntl = pdd->spi_dbg_override | pdd->spi_dbg_launch_mode;
+	uint32_t flags = pdd->process->dbg_flags;
+	struct amdgpu_device *adev = pdd->dev->adev;
+	int r;
+
+	if (!kfd_dbg_is_per_vmid_supported(pdd->dev))
+		return 0;
+
+	if (!pdd->proc_ctx_cpu_ptr) {
+		r = amdgpu_amdkfd_alloc_gtt_mem(adev,
+			AMDGPU_MES_PROC_CTX_SIZE,
+			&pdd->proc_ctx_bo,
+			&pdd->proc_ctx_gpu_addr,
+			&pdd->proc_ctx_cpu_ptr,
+			false);
+		if (r) {
+			dev_err(adev->dev,
+			"failed to allocate process context bo\n");
+			return r;
+		}
+		memset(pdd->proc_ctx_cpu_ptr, 0, AMDGPU_MES_PROC_CTX_SIZE);
+	}
+
+	return amdgpu_mes_set_shader_debugger(pdd->dev->adev, pdd->proc_ctx_gpu_addr, spi_dbg_cntl,
+						pdd->watch_points, flags, sq_trap_en);
+}
+
+#define KFD_DEBUGGER_INVALID_WATCH_POINT_ID -1
+static int kfd_dbg_get_dev_watch_id(struct kfd_process_device *pdd, int *watch_id)
+{
+	int i;
+
+	*watch_id = KFD_DEBUGGER_INVALID_WATCH_POINT_ID;
+
+	spin_lock(&pdd->dev->watch_points_lock);
+
+	for (i = 0; i < MAX_WATCH_ADDRESSES; i++) {
+		/* device watchpoint in use so skip */
+		if ((pdd->dev->alloc_watch_ids >> i) & 0x1)
+			continue;
+
+		pdd->alloc_watch_ids |= 0x1 << i;
+		pdd->dev->alloc_watch_ids |= 0x1 << i;
+		*watch_id = i;
+		spin_unlock(&pdd->dev->watch_points_lock);
+		return 0;
+	}
+
+	spin_unlock(&pdd->dev->watch_points_lock);
+
+	return -ENOMEM;
+}
+
+static void kfd_dbg_clear_dev_watch_id(struct kfd_process_device *pdd, int watch_id)
+{
+	spin_lock(&pdd->dev->watch_points_lock);
+
+	/* process owns device watch point so safe to clear */
+	if ((pdd->alloc_watch_ids >> watch_id) & 0x1) {
+		pdd->alloc_watch_ids &= ~(0x1 << watch_id);
+		pdd->dev->alloc_watch_ids &= ~(0x1 << watch_id);
+	}
+
+	spin_unlock(&pdd->dev->watch_points_lock);
+}
+
+static bool kfd_dbg_owns_dev_watch_id(struct kfd_process_device *pdd, int watch_id)
+{
+	bool owns_watch_id = false;
+
+	spin_lock(&pdd->dev->watch_points_lock);
+	owns_watch_id = watch_id < MAX_WATCH_ADDRESSES &&
+			((pdd->alloc_watch_ids >> watch_id) & 0x1);
+
+	spin_unlock(&pdd->dev->watch_points_lock);
+
+	return owns_watch_id;
+}
+
+int kfd_dbg_trap_clear_dev_address_watch(struct kfd_process_device *pdd,
+					uint32_t watch_id)
+{
+	int r;
+
+	if (!kfd_dbg_owns_dev_watch_id(pdd, watch_id))
+		return -EINVAL;
+
+	if (!pdd->dev->kfd->shared_resources.enable_mes) {
+		r = debug_lock_and_unmap(pdd->dev->dqm);
+		if (r)
+			return r;
+	}
+
+	amdgpu_gfx_off_ctrl(pdd->dev->adev, false);
+	pdd->watch_points[watch_id] = pdd->dev->kfd2kgd->clear_address_watch(
+							pdd->dev->adev,
+							watch_id);
+	amdgpu_gfx_off_ctrl(pdd->dev->adev, true);
+
+	if (!pdd->dev->kfd->shared_resources.enable_mes)
+		r = debug_map_and_unlock(pdd->dev->dqm);
+	else
+		r = kfd_dbg_set_mes_debug_mode(pdd, true);
+
+	kfd_dbg_clear_dev_watch_id(pdd, watch_id);
+
+	return r;
+}
+
+int kfd_dbg_trap_set_dev_address_watch(struct kfd_process_device *pdd,
+					uint64_t watch_address,
+					uint32_t watch_address_mask,
+					uint32_t *watch_id,
+					uint32_t watch_mode)
+{
+	int xcc_id, r = kfd_dbg_get_dev_watch_id(pdd, watch_id);
+	uint32_t xcc_mask = pdd->dev->xcc_mask;
+
+	if (r)
+		return r;
+
+	if (!pdd->dev->kfd->shared_resources.enable_mes) {
+		r = debug_lock_and_unmap(pdd->dev->dqm);
+		if (r) {
+			kfd_dbg_clear_dev_watch_id(pdd, *watch_id);
+			return r;
+		}
+	}
+
+	amdgpu_gfx_off_ctrl(pdd->dev->adev, false);
+	for_each_inst(xcc_id, xcc_mask)
+		pdd->watch_points[*watch_id] = pdd->dev->kfd2kgd->set_address_watch(
+				pdd->dev->adev,
+				watch_address,
+				watch_address_mask,
+				*watch_id,
+				watch_mode,
+				pdd->dev->vm_info.last_vmid_kfd,
+				xcc_id);
+	amdgpu_gfx_off_ctrl(pdd->dev->adev, true);
+
+	if (!pdd->dev->kfd->shared_resources.enable_mes)
+		r = debug_map_and_unlock(pdd->dev->dqm);
+	else
+		r = kfd_dbg_set_mes_debug_mode(pdd, true);
+
+	/* HWS is broken so no point in HW rollback but release the watchpoint anyways */
+	if (r)
+		kfd_dbg_clear_dev_watch_id(pdd, *watch_id);
+
+	return 0;
+}
+
+static void kfd_dbg_clear_process_address_watch(struct kfd_process *target)
+{
+	int i, j;
+
+	for (i = 0; i < target->n_pdds; i++)
+		for (j = 0; j < MAX_WATCH_ADDRESSES; j++)
+			kfd_dbg_trap_clear_dev_address_watch(target->pdds[i], j);
+}
+
+int kfd_dbg_trap_set_flags(struct kfd_process *target, uint32_t *flags)
+{
+	uint32_t prev_flags = target->dbg_flags;
+	int i, r = 0, rewind_count = 0;
+
+	for (i = 0; i < target->n_pdds; i++) {
+		struct kfd_topology_device *topo_dev =
+				kfd_topology_device_by_id(target->pdds[i]->dev->id);
+		uint32_t caps = topo_dev->node_props.capability;
+
+		if (!(caps & HSA_CAP_TRAP_DEBUG_PRECISE_MEMORY_OPERATIONS_SUPPORTED) &&
+			(*flags & KFD_DBG_TRAP_FLAG_SINGLE_MEM_OP)) {
+			*flags = prev_flags;
+			return -EACCES;
+		}
+
+		if (!(caps & HSA_CAP_TRAP_DEBUG_PRECISE_ALU_OPERATIONS_SUPPORTED) &&
+		    (*flags & KFD_DBG_TRAP_FLAG_SINGLE_ALU_OP)) {
+			*flags = prev_flags;
+			return -EACCES;
+		}
+	}
+
+	target->dbg_flags = *flags;
+	*flags = prev_flags;
+	for (i = 0; i < target->n_pdds; i++) {
+		struct kfd_process_device *pdd = target->pdds[i];
+
+		if (!kfd_dbg_is_per_vmid_supported(pdd->dev))
+			continue;
+
+		if (!pdd->dev->kfd->shared_resources.enable_mes)
+			r = debug_refresh_runlist(pdd->dev->dqm);
+		else
+			r = kfd_dbg_set_mes_debug_mode(pdd, true);
+
+		if (r) {
+			target->dbg_flags = prev_flags;
+			break;
+		}
+
+		rewind_count++;
+	}
+
+	/* Rewind flags */
+	if (r) {
+		target->dbg_flags = prev_flags;
+
+		for (i = 0; i < rewind_count; i++) {
+			struct kfd_process_device *pdd = target->pdds[i];
+
+			if (!kfd_dbg_is_per_vmid_supported(pdd->dev))
+				continue;
+
+			if (!pdd->dev->kfd->shared_resources.enable_mes)
+				debug_refresh_runlist(pdd->dev->dqm);
+			else
+				kfd_dbg_set_mes_debug_mode(pdd, true);
+		}
+	}
+
+	return r;
+}
+
+/* kfd_dbg_trap_deactivate:
+ *	target: target process
+ *	unwind: If this is unwinding a failed kfd_dbg_trap_enable()
+ *	unwind_count:
+ *		If unwind == true, how far down the pdd list we need
+ *				to unwind
+ *		else: ignored
+ */
+void kfd_dbg_trap_deactivate(struct kfd_process *target, bool unwind, int unwind_count)
+{
+	int i;
+
+	if (!unwind) {
+		uint32_t flags = 0;
+		int resume_count = resume_queues(target, 0, NULL);
+
+		if (resume_count)
+			pr_debug("Resumed %d queues\n", resume_count);
+
+		cancel_work_sync(&target->debug_event_workarea);
+		kfd_dbg_clear_process_address_watch(target);
+		kfd_dbg_trap_set_wave_launch_mode(target, 0);
+
+		kfd_dbg_trap_set_flags(target, &flags);
+	}
+
+	for (i = 0; i < target->n_pdds; i++) {
+		struct kfd_process_device *pdd = target->pdds[i];
+
+		/* If this is an unwind, and we have unwound the required
+		 * enable calls on the pdd list, we need to stop now
+		 * otherwise we may mess up another debugger session.
+		 */
+		if (unwind && i == unwind_count)
+			break;
+
+		kfd_process_set_trap_debug_flag(&pdd->qpd, false);
+
+		/* GFX off is already disabled by debug activate if not RLC restore supported. */
+		if (kfd_dbg_is_rlc_restore_supported(pdd->dev))
+			amdgpu_gfx_off_ctrl(pdd->dev->adev, false);
+		pdd->spi_dbg_override =
+				pdd->dev->kfd2kgd->disable_debug_trap(
+				pdd->dev->adev,
+				target->runtime_info.ttmp_setup,
+				pdd->dev->vm_info.last_vmid_kfd);
+		amdgpu_gfx_off_ctrl(pdd->dev->adev, true);
+
+		if (!kfd_dbg_is_per_vmid_supported(pdd->dev) &&
+				release_debug_trap_vmid(pdd->dev->dqm, &pdd->qpd))
+			pr_err("Failed to release debug vmid on [%i]\n", pdd->dev->id);
+
+		if (!pdd->dev->kfd->shared_resources.enable_mes)
+			debug_refresh_runlist(pdd->dev->dqm);
+		else
+			kfd_dbg_set_mes_debug_mode(pdd, !kfd_dbg_has_cwsr_workaround(pdd->dev));
+	}
+
+	kfd_dbg_set_workaround(target, false);
+}
+
+static void kfd_dbg_clean_exception_status(struct kfd_process *target)
+{
+	struct process_queue_manager *pqm;
+	struct process_queue_node *pqn;
+	int i;
+
+	for (i = 0; i < target->n_pdds; i++) {
+		struct kfd_process_device *pdd = target->pdds[i];
+
+		kfd_process_drain_interrupts(pdd);
+
+		pdd->exception_status = 0;
+	}
+
+	pqm = &target->pqm;
+	list_for_each_entry(pqn, &pqm->queues, process_queue_list) {
+		if (!pqn->q)
+			continue;
+
+		pqn->q->properties.exception_status = 0;
+	}
+
+	target->exception_status = 0;
+}
+
+int kfd_dbg_trap_disable(struct kfd_process *target)
+{
+	if (!target->debug_trap_enabled)
+		return 0;
+
+	/*
+	 * Defer deactivation to runtime if runtime not enabled otherwise reset
+	 * attached running target runtime state to enable for re-attach.
+	 */
+	if (target->runtime_info.runtime_state == DEBUG_RUNTIME_STATE_ENABLED)
+		kfd_dbg_trap_deactivate(target, false, 0);
+	else if (target->runtime_info.runtime_state != DEBUG_RUNTIME_STATE_DISABLED)
+		target->runtime_info.runtime_state = DEBUG_RUNTIME_STATE_ENABLED;
+
+	cancel_work_sync(&target->debug_event_workarea);
+	fput(target->dbg_ev_file);
+	target->dbg_ev_file = NULL;
+
+	if (target->debugger_process) {
+		atomic_dec(&target->debugger_process->debugged_process_count);
+		target->debugger_process = NULL;
+	}
+
+	target->debug_trap_enabled = false;
+	kfd_dbg_clean_exception_status(target);
+	kfd_unref_process(target);
+
+	return 0;
+}
+
+int kfd_dbg_trap_activate(struct kfd_process *target)
+{
+	int i, r = 0;
+
+	r = kfd_dbg_set_workaround(target, true);
+	if (r)
+		return r;
+
+	for (i = 0; i < target->n_pdds; i++) {
+		struct kfd_process_device *pdd = target->pdds[i];
+
+		if (!kfd_dbg_is_per_vmid_supported(pdd->dev)) {
+			r = reserve_debug_trap_vmid(pdd->dev->dqm, &pdd->qpd);
+
+			if (r) {
+				target->runtime_info.runtime_state = (r == -EBUSY) ?
+							DEBUG_RUNTIME_STATE_ENABLED_BUSY :
+							DEBUG_RUNTIME_STATE_ENABLED_ERROR;
+
+				goto unwind_err;
+			}
+		}
+
+		/* Disable GFX OFF to prevent garbage read/writes to debug registers.
+		 * If RLC restore of debug registers is not supported and runtime enable
+		 * hasn't done so already on ttmp setup request, restore the trap config registers.
+		 *
+		 * If RLC restore of debug registers is not supported, keep gfx off disabled for
+		 * the debug session.
+		 */
+		amdgpu_gfx_off_ctrl(pdd->dev->adev, false);
+		if (!(kfd_dbg_is_rlc_restore_supported(pdd->dev) ||
+						target->runtime_info.ttmp_setup))
+			pdd->dev->kfd2kgd->enable_debug_trap(pdd->dev->adev, true,
+								pdd->dev->vm_info.last_vmid_kfd);
+
+		pdd->spi_dbg_override = pdd->dev->kfd2kgd->enable_debug_trap(
+					pdd->dev->adev,
+					false,
+					pdd->dev->vm_info.last_vmid_kfd);
+
+		if (kfd_dbg_is_rlc_restore_supported(pdd->dev))
+			amdgpu_gfx_off_ctrl(pdd->dev->adev, true);
+
+		/*
+		 * Setting the debug flag in the trap handler requires that the TMA has been
+		 * allocated, which occurs during CWSR initialization.
+		 * In the event that CWSR has not been initialized at this point, setting the
+		 * flag will be called again during CWSR initialization if the target process
+		 * is still debug enabled.
+		 */
+		kfd_process_set_trap_debug_flag(&pdd->qpd, true);
+
+		if (!pdd->dev->kfd->shared_resources.enable_mes)
+			r = debug_refresh_runlist(pdd->dev->dqm);
+		else
+			r = kfd_dbg_set_mes_debug_mode(pdd, true);
+
+		if (r) {
+			target->runtime_info.runtime_state =
+					DEBUG_RUNTIME_STATE_ENABLED_ERROR;
+			goto unwind_err;
+		}
+	}
+
+	return 0;
+
+unwind_err:
+	/* Enabling debug failed, we need to disable on
+	 * all GPUs so the enable is all or nothing.
+	 */
+	kfd_dbg_trap_deactivate(target, true, i);
+	return r;
+}
+
+int kfd_dbg_trap_enable(struct kfd_process *target, uint32_t fd,
+			void __user *runtime_info, uint32_t *runtime_size)
+{
+	struct file *f;
+	uint32_t copy_size;
+	int i, r = 0;
+
+	if (target->debug_trap_enabled)
+		return -EALREADY;
+
+	/* Enable pre-checks */
+	for (i = 0; i < target->n_pdds; i++) {
+		struct kfd_process_device *pdd = target->pdds[i];
+
+		if (!KFD_IS_SOC15(pdd->dev))
+			return -ENODEV;
+
+		if (pdd->qpd.num_gws && (!kfd_dbg_has_gws_support(pdd->dev) ||
+					 kfd_dbg_has_cwsr_workaround(pdd->dev)))
+			return -EBUSY;
+	}
+
+	copy_size = min((size_t)(*runtime_size), sizeof(target->runtime_info));
+
+	f = fget(fd);
+	if (!f) {
+		pr_err("Failed to get file for (%i)\n", fd);
+		return -EBADF;
+	}
+
+	target->dbg_ev_file = f;
+
+	/* defer activation to runtime if not runtime enabled */
+	if (target->runtime_info.runtime_state == DEBUG_RUNTIME_STATE_ENABLED)
+		kfd_dbg_trap_activate(target);
+
+	/* We already hold the process reference but hold another one for the
+	 * debug session.
+	 */
+	kref_get(&target->ref);
+	target->debug_trap_enabled = true;
+
+	if (target->debugger_process)
+		atomic_inc(&target->debugger_process->debugged_process_count);
+
+	if (copy_to_user(runtime_info, (void *)&target->runtime_info, copy_size)) {
+		kfd_dbg_trap_deactivate(target, false, 0);
+		r = -EFAULT;
+	}
+
+	*runtime_size = sizeof(target->runtime_info);
+
+	return r;
+}
+
+static int kfd_dbg_validate_trap_override_request(struct kfd_process *p,
+						uint32_t trap_override,
+						uint32_t trap_mask_request,
+						uint32_t *trap_mask_supported)
+{
+	int i = 0;
+
+	*trap_mask_supported = 0xffffffff;
+
+	for (i = 0; i < p->n_pdds; i++) {
+		struct kfd_process_device *pdd = p->pdds[i];
+		int err = pdd->dev->kfd2kgd->validate_trap_override_request(
+								pdd->dev->adev,
+								trap_override,
+								trap_mask_supported);
+
+		if (err)
+			return err;
+	}
+
+	if (trap_mask_request & ~*trap_mask_supported)
+		return -EACCES;
+
+	return 0;
+}
+
+int kfd_dbg_trap_set_wave_launch_override(struct kfd_process *target,
+					uint32_t trap_override,
+					uint32_t trap_mask_bits,
+					uint32_t trap_mask_request,
+					uint32_t *trap_mask_prev,
+					uint32_t *trap_mask_supported)
+{
+	int r = 0, i;
+
+	r = kfd_dbg_validate_trap_override_request(target,
+						trap_override,
+						trap_mask_request,
+						trap_mask_supported);
+
+	if (r)
+		return r;
+
+	for (i = 0; i < target->n_pdds; i++) {
+		struct kfd_process_device *pdd = target->pdds[i];
+
+		amdgpu_gfx_off_ctrl(pdd->dev->adev, false);
+		pdd->spi_dbg_override = pdd->dev->kfd2kgd->set_wave_launch_trap_override(
+				pdd->dev->adev,
+				pdd->dev->vm_info.last_vmid_kfd,
+				trap_override,
+				trap_mask_bits,
+				trap_mask_request,
+				trap_mask_prev,
+				pdd->spi_dbg_override);
+		amdgpu_gfx_off_ctrl(pdd->dev->adev, true);
+
+		if (!pdd->dev->kfd->shared_resources.enable_mes)
+			r = debug_refresh_runlist(pdd->dev->dqm);
+		else
+			r = kfd_dbg_set_mes_debug_mode(pdd, true);
+
+		if (r)
+			break;
+	}
+
+	return r;
+}
+
+int kfd_dbg_trap_set_wave_launch_mode(struct kfd_process *target,
+					uint8_t wave_launch_mode)
+{
+	int r = 0, i;
+
+	if (wave_launch_mode != KFD_DBG_TRAP_WAVE_LAUNCH_MODE_NORMAL &&
+			wave_launch_mode != KFD_DBG_TRAP_WAVE_LAUNCH_MODE_HALT &&
+			wave_launch_mode != KFD_DBG_TRAP_WAVE_LAUNCH_MODE_DEBUG)
+		return -EINVAL;
+
+	for (i = 0; i < target->n_pdds; i++) {
+		struct kfd_process_device *pdd = target->pdds[i];
+
+		amdgpu_gfx_off_ctrl(pdd->dev->adev, false);
+		pdd->spi_dbg_launch_mode = pdd->dev->kfd2kgd->set_wave_launch_mode(
+				pdd->dev->adev,
+				wave_launch_mode,
+				pdd->dev->vm_info.last_vmid_kfd);
+		amdgpu_gfx_off_ctrl(pdd->dev->adev, true);
+
+		if (!pdd->dev->kfd->shared_resources.enable_mes)
+			r = debug_refresh_runlist(pdd->dev->dqm);
+		else
+			r = kfd_dbg_set_mes_debug_mode(pdd, true);
+
+		if (r)
+			break;
+	}
+
+	return r;
+}
+
+int kfd_dbg_trap_query_exception_info(struct kfd_process *target,
+		uint32_t source_id,
+		uint32_t exception_code,
+		bool clear_exception,
+		void __user *info,
+		uint32_t *info_size)
+{
+	bool found = false;
+	int r = 0;
+	uint32_t copy_size, actual_info_size = 0;
+	uint64_t *exception_status_ptr = NULL;
+
+	if (!target)
+		return -EINVAL;
+
+	if (!info || !info_size)
+		return -EINVAL;
+
+	mutex_lock(&target->event_mutex);
+
+	if (KFD_DBG_EC_TYPE_IS_QUEUE(exception_code)) {
+		/* Per queue exceptions */
+		struct queue *queue = NULL;
+		int i;
+
+		for (i = 0; i < target->n_pdds; i++) {
+			struct kfd_process_device *pdd = target->pdds[i];
+			struct qcm_process_device *qpd = &pdd->qpd;
+
+			list_for_each_entry(queue, &qpd->queues_list, list) {
+				if (!found && queue->properties.queue_id == source_id) {
+					found = true;
+					break;
+				}
+			}
+			if (found)
+				break;
+		}
+
+		if (!found) {
+			r = -EINVAL;
+			goto out;
+		}
+
+		if (!(queue->properties.exception_status & KFD_EC_MASK(exception_code))) {
+			r = -ENODATA;
+			goto out;
+		}
+		exception_status_ptr = &queue->properties.exception_status;
+	} else if (KFD_DBG_EC_TYPE_IS_DEVICE(exception_code)) {
+		/* Per device exceptions */
+		struct kfd_process_device *pdd = NULL;
+		int i;
+
+		for (i = 0; i < target->n_pdds; i++) {
+			pdd = target->pdds[i];
+			if (pdd->dev->id == source_id) {
+				found = true;
+				break;
+			}
+		}
+
+		if (!found) {
+			r = -EINVAL;
+			goto out;
+		}
+
+		if (!(pdd->exception_status & KFD_EC_MASK(exception_code))) {
+			r = -ENODATA;
+			goto out;
+		}
+
+		if (exception_code == EC_DEVICE_MEMORY_VIOLATION) {
+			copy_size = min((size_t)(*info_size), pdd->vm_fault_exc_data_size);
+
+			if (copy_to_user(info, pdd->vm_fault_exc_data, copy_size)) {
+				r = -EFAULT;
+				goto out;
+			}
+			actual_info_size = pdd->vm_fault_exc_data_size;
+			if (clear_exception) {
+				kfree(pdd->vm_fault_exc_data);
+				pdd->vm_fault_exc_data = NULL;
+				pdd->vm_fault_exc_data_size = 0;
+			}
+		}
+		exception_status_ptr = &pdd->exception_status;
+	} else if (KFD_DBG_EC_TYPE_IS_PROCESS(exception_code)) {
+		/* Per process exceptions */
+		if (!(target->exception_status & KFD_EC_MASK(exception_code))) {
+			r = -ENODATA;
+			goto out;
+		}
+
+		if (exception_code == EC_PROCESS_RUNTIME) {
+			copy_size = min((size_t)(*info_size), sizeof(target->runtime_info));
+
+			if (copy_to_user(info, (void *)&target->runtime_info, copy_size)) {
+				r = -EFAULT;
+				goto out;
+			}
+
+			actual_info_size = sizeof(target->runtime_info);
+		}
+
+		exception_status_ptr = &target->exception_status;
+	} else {
+		pr_debug("Bad exception type [%i]\n", exception_code);
+		r = -EINVAL;
+		goto out;
+	}
+
+	*info_size = actual_info_size;
+	if (clear_exception)
+		*exception_status_ptr &= ~KFD_EC_MASK(exception_code);
+out:
+	mutex_unlock(&target->event_mutex);
+	return r;
+}
+
+int kfd_dbg_trap_device_snapshot(struct kfd_process *target,
+		uint64_t exception_clear_mask,
+		void __user *user_info,
+		uint32_t *number_of_device_infos,
+		uint32_t *entry_size)
+{
+	struct kfd_dbg_device_info_entry device_info;
+	uint32_t tmp_entry_size, tmp_num_devices;
+	int i, r = 0;
+
+	if (!(target && user_info && number_of_device_infos && entry_size))
+		return -EINVAL;
+
+	tmp_entry_size = *entry_size;
+
+	tmp_num_devices = min_t(size_t, *number_of_device_infos, target->n_pdds);
+	*number_of_device_infos = target->n_pdds;
+	*entry_size = min_t(size_t, *entry_size, sizeof(device_info));
+
+	if (!tmp_num_devices)
+		return 0;
+
+	memset(&device_info, 0, sizeof(device_info));
+
+	mutex_lock(&target->event_mutex);
+
+	/* Run over all pdd of the process */
+	for (i = 0; i < tmp_num_devices; i++) {
+		struct kfd_process_device *pdd = target->pdds[i];
+		struct kfd_topology_device *topo_dev = kfd_topology_device_by_id(pdd->dev->id);
+
+		device_info.gpu_id = pdd->dev->id;
+		device_info.exception_status = pdd->exception_status;
+		device_info.lds_base = pdd->lds_base;
+		device_info.lds_limit = pdd->lds_limit;
+		device_info.scratch_base = pdd->scratch_base;
+		device_info.scratch_limit = pdd->scratch_limit;
+		device_info.gpuvm_base = pdd->gpuvm_base;
+		device_info.gpuvm_limit = pdd->gpuvm_limit;
+		device_info.location_id = topo_dev->node_props.location_id;
+		device_info.vendor_id = topo_dev->node_props.vendor_id;
+		device_info.device_id = topo_dev->node_props.device_id;
+		device_info.revision_id = pdd->dev->adev->pdev->revision;
+		device_info.subsystem_vendor_id = pdd->dev->adev->pdev->subsystem_vendor;
+		device_info.subsystem_device_id = pdd->dev->adev->pdev->subsystem_device;
+		device_info.fw_version = pdd->dev->kfd->mec_fw_version;
+		device_info.gfx_target_version =
+			topo_dev->node_props.gfx_target_version;
+		device_info.simd_count = topo_dev->node_props.simd_count;
+		device_info.max_waves_per_simd =
+			topo_dev->node_props.max_waves_per_simd;
+		device_info.array_count = topo_dev->node_props.array_count;
+		device_info.simd_arrays_per_engine =
+			topo_dev->node_props.simd_arrays_per_engine;
+		device_info.num_xcc = NUM_XCC(pdd->dev->xcc_mask);
+		device_info.capability = topo_dev->node_props.capability;
+		device_info.debug_prop = topo_dev->node_props.debug_prop;
+
+		if (exception_clear_mask)
+			pdd->exception_status &= ~exception_clear_mask;
+
+		if (copy_to_user(user_info, &device_info, *entry_size)) {
+			r = -EFAULT;
+			break;
+		}
+
+		user_info += tmp_entry_size;
+	}
+
+	mutex_unlock(&target->event_mutex);
+
+	return r;
+}
+
+void kfd_dbg_set_enabled_debug_exception_mask(struct kfd_process *target,
+					uint64_t exception_set_mask)
+{
+	uint64_t found_mask = 0;
+	struct process_queue_manager *pqm;
+	struct process_queue_node *pqn;
+	static const char write_data = '.';
+	loff_t pos = 0;
+	int i;
+
+	mutex_lock(&target->event_mutex);
+
+	found_mask |= target->exception_status;
+
+	pqm = &target->pqm;
+	list_for_each_entry(pqn, &pqm->queues, process_queue_list) {
+		if (!pqn->q)
+			continue;
+
+		found_mask |= pqn->q->properties.exception_status;
+	}
+
+	for (i = 0; i < target->n_pdds; i++) {
+		struct kfd_process_device *pdd = target->pdds[i];
+
+		found_mask |= pdd->exception_status;
+	}
+
+	if (exception_set_mask & found_mask)
+		kernel_write(target->dbg_ev_file, &write_data, 1, &pos);
+
+	target->exception_enable_mask = exception_set_mask;
+
+	mutex_unlock(&target->event_mutex);
+}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debug.h b/drivers/gpu/drm/amd/amdkfd/kfd_debug.h
new file mode 100644
index 000000000000..27aa1a5b120f
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_debug.h
@@ -0,0 +1,142 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef KFD_DEBUG_EVENTS_H_INCLUDED
+#define KFD_DEBUG_EVENTS_H_INCLUDED
+
+#include "kfd_priv.h"
+
+void kfd_dbg_trap_deactivate(struct kfd_process *target, bool unwind, int unwind_count);
+int kfd_dbg_trap_activate(struct kfd_process *target);
+int kfd_dbg_ev_query_debug_event(struct kfd_process *process,
+			unsigned int *queue_id,
+			unsigned int *gpu_id,
+			uint64_t exception_clear_mask,
+			uint64_t *event_status);
+bool kfd_set_dbg_ev_from_interrupt(struct kfd_node *dev,
+				   unsigned int pasid,
+				   uint32_t doorbell_id,
+				   uint64_t trap_mask,
+				   void *exception_data,
+				   size_t exception_data_size);
+bool kfd_dbg_ev_raise(uint64_t event_mask,
+			struct kfd_process *process, struct kfd_node *dev,
+			unsigned int source_id, bool use_worker,
+			void *exception_data,
+			size_t exception_data_size);
+int kfd_dbg_trap_disable(struct kfd_process *target);
+int kfd_dbg_trap_enable(struct kfd_process *target, uint32_t fd,
+			void __user *runtime_info,
+			uint32_t *runtime_info_size);
+int kfd_dbg_trap_set_wave_launch_override(struct kfd_process *target,
+					uint32_t trap_override,
+					uint32_t trap_mask_bits,
+					uint32_t trap_mask_request,
+					uint32_t *trap_mask_prev,
+					uint32_t *trap_mask_supported);
+int kfd_dbg_trap_set_wave_launch_mode(struct kfd_process *target,
+					uint8_t wave_launch_mode);
+int kfd_dbg_trap_clear_dev_address_watch(struct kfd_process_device *pdd,
+					uint32_t watch_id);
+int kfd_dbg_trap_set_dev_address_watch(struct kfd_process_device *pdd,
+					uint64_t watch_address,
+					uint32_t watch_address_mask,
+					uint32_t *watch_id,
+					uint32_t watch_mode);
+int kfd_dbg_trap_set_flags(struct kfd_process *target, uint32_t *flags);
+int kfd_dbg_trap_query_exception_info(struct kfd_process *target,
+		uint32_t source_id,
+		uint32_t exception_code,
+		bool clear_exception,
+		void __user *info,
+		uint32_t *info_size);
+int kfd_dbg_send_exception_to_runtime(struct kfd_process *p,
+					unsigned int dev_id,
+					unsigned int queue_id,
+					uint64_t error_reason);
+
+static inline bool kfd_dbg_is_per_vmid_supported(struct kfd_node *dev)
+{
+	return (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 2) ||
+		KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 3) ||
+		KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 4) ||
+		KFD_GC_VERSION(dev) == IP_VERSION(9, 5, 0) ||
+		KFD_GC_VERSION(dev) >= IP_VERSION(11, 0, 0));
+}
+
+void debug_event_write_work_handler(struct work_struct *work);
+int kfd_dbg_trap_device_snapshot(struct kfd_process *target,
+		uint64_t exception_clear_mask,
+		void __user *user_info,
+		uint32_t *number_of_device_infos,
+		uint32_t *entry_size);
+
+void kfd_dbg_set_enabled_debug_exception_mask(struct kfd_process *target,
+					uint64_t exception_set_mask);
+/*
+ * If GFX off is enabled, chips that do not support RLC restore for the debug
+ * registers will disable GFX off temporarily for the entire debug session.
+ * See disable_on_trap_action_entry and enable_on_trap_action_exit for details.
+ */
+static inline bool kfd_dbg_is_rlc_restore_supported(struct kfd_node *dev)
+{
+	return !(KFD_GC_VERSION(dev) == IP_VERSION(10, 1, 10) ||
+		 KFD_GC_VERSION(dev) == IP_VERSION(10, 1, 1));
+}
+
+static inline bool kfd_dbg_has_cwsr_workaround(struct kfd_node *dev)
+{
+	return KFD_GC_VERSION(dev) >= IP_VERSION(11, 0, 0) &&
+	       KFD_GC_VERSION(dev) <= IP_VERSION(11, 0, 3);
+}
+
+static inline bool kfd_dbg_has_gws_support(struct kfd_node *dev)
+{
+	if ((KFD_GC_VERSION(dev) == IP_VERSION(9, 0, 1)
+			&& dev->kfd->mec2_fw_version < 0x81b6) ||
+		(KFD_GC_VERSION(dev) >= IP_VERSION(9, 1, 0)
+			&& KFD_GC_VERSION(dev) <= IP_VERSION(9, 2, 2)
+			&& dev->kfd->mec2_fw_version < 0x1b6) ||
+		(KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 0)
+			&& dev->kfd->mec2_fw_version < 0x1b6) ||
+		(KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 1)
+			&& dev->kfd->mec2_fw_version < 0x30) ||
+		(KFD_GC_VERSION(dev) >= IP_VERSION(11, 0, 0) &&
+			KFD_GC_VERSION(dev) < IP_VERSION(12, 0, 0)))
+		return false;
+
+	/* Assume debugging and cooperative launch supported otherwise. */
+	return true;
+}
+
+int kfd_dbg_set_mes_debug_mode(struct kfd_process_device *pdd, bool sq_trap_en);
+
+static inline bool kfd_dbg_has_ttmps_always_setup(struct kfd_node *dev)
+{
+	return (KFD_GC_VERSION(dev) < IP_VERSION(11, 0, 0) &&
+			KFD_GC_VERSION(dev) != IP_VERSION(9, 4, 2)) ||
+	       (KFD_GC_VERSION(dev) >= IP_VERSION(11, 0, 0) &&
+			KFD_GC_VERSION(dev) < IP_VERSION(12, 0, 0) &&
+			(dev->adev->mes.sched_version & AMDGPU_MES_VERSION_MASK) >= 70) ||
+	       (KFD_GC_VERSION(dev) >= IP_VERSION(12, 0, 0));
+}
+#endif
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debugfs.c b/drivers/gpu/drm/amd/amdkfd/kfd_debugfs.c
index ad5a40a685ac..9bde2c64540f 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_debugfs.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_debugfs.c
@@ -27,6 +27,16 @@
 #include "kfd_priv.h"
 
 static struct dentry *debugfs_root;
+static struct dentry *debugfs_proc;
+static struct list_head procs;
+
+struct debugfs_proc_entry {
+	struct list_head list;
+	struct dentry *proc_dentry;
+	pid_t pid;
+};
+
+#define MAX_DEBUGFS_FILENAME_LEN 32
 
 static int kfd_debugfs_open(struct inode *inode, struct file *file)
 {
@@ -43,7 +53,7 @@ static int kfd_debugfs_hang_hws_read(struct seq_file *m, void *data)
 static ssize_t kfd_debugfs_hang_hws_write(struct file *file,
 	const char __user *user_buf, size_t size, loff_t *ppos)
 {
-	struct kfd_dev *dev;
+	struct kfd_node *dev;
 	char tmp[16];
 	uint32_t gpu_id;
 	int ret = -EINVAL;
@@ -92,6 +102,8 @@ static const struct file_operations kfd_debugfs_hang_hws_fops = {
 void kfd_debugfs_init(void)
 {
 	debugfs_root = debugfs_create_dir("kfd", NULL);
+	debugfs_proc = debugfs_create_dir("proc", debugfs_root);
+	INIT_LIST_HEAD(&procs);
 
 	debugfs_create_file("mqds", S_IFREG | 0444, debugfs_root,
 			    kfd_debugfs_mqds_by_process, &kfd_debugfs_fops);
@@ -107,5 +119,69 @@ void kfd_debugfs_init(void)
 
 void kfd_debugfs_fini(void)
 {
+	debugfs_remove_recursive(debugfs_proc);
 	debugfs_remove_recursive(debugfs_root);
 }
+
+static ssize_t kfd_debugfs_pasid_read(struct file *file, char __user *buf,
+				      size_t count, loff_t *ppos)
+{
+	struct kfd_process_device *pdd = file_inode(file)->i_private;
+	char tmp[32];
+	int len;
+
+	len = snprintf(tmp, sizeof(tmp), "%u\n", pdd->pasid);
+
+	return simple_read_from_buffer(buf, count, ppos, tmp, len);
+}
+
+static const struct file_operations kfd_debugfs_pasid_fops = {
+	.owner = THIS_MODULE,
+	.read = kfd_debugfs_pasid_read,
+};
+
+void kfd_debugfs_add_process(struct kfd_process *p)
+{
+	int i;
+	char name[MAX_DEBUGFS_FILENAME_LEN];
+	struct debugfs_proc_entry *entry;
+
+	entry = kzalloc(sizeof(*entry), GFP_KERNEL);
+	if (!entry)
+		return;
+
+	list_add(&entry->list, &procs);
+	entry->pid = p->lead_thread->pid;
+	snprintf(name, MAX_DEBUGFS_FILENAME_LEN, "%d",
+		 (int)entry->pid);
+	entry->proc_dentry = debugfs_create_dir(name, debugfs_proc);
+
+	/* Create debugfs files for each GPU:
+	 * - proc/<pid>/pasid_<gpuid>
+	 */
+	for (i = 0; i < p->n_pdds; i++) {
+		struct kfd_process_device *pdd = p->pdds[i];
+
+		snprintf(name, MAX_DEBUGFS_FILENAME_LEN, "pasid_%u",
+			 pdd->dev->id);
+		debugfs_create_file((const char *)name, S_IFREG | 0444,
+				    entry->proc_dentry, pdd,
+				    &kfd_debugfs_pasid_fops);
+	}
+}
+
+void kfd_debugfs_remove_process(struct kfd_process *p)
+{
+	struct debugfs_proc_entry *entry, *next;
+
+	mutex_lock(&kfd_processes_mutex);
+	list_for_each_entry_safe(entry, next, &procs, list) {
+		if (entry->pid != p->lead_thread->pid)
+			continue;
+
+		debugfs_remove_recursive(entry->proc_dentry);
+		list_del(&entry->list);
+		kfree(entry);
+	}
+	mutex_unlock(&kfd_processes_mutex);
+}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index 00f528eb9812..e9cfb80bd436 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -29,11 +29,12 @@
 #include "kfd_pm4_headers_vi.h"
 #include "kfd_pm4_headers_aldebaran.h"
 #include "cwsr_trap_handler.h"
-#include "kfd_iommu.h"
 #include "amdgpu_amdkfd.h"
 #include "kfd_smi_events.h"
+#include "kfd_svm.h"
 #include "kfd_migrate.h"
 #include "amdgpu.h"
+#include "amdgpu_xcp.h"
 
 #define MQD_SIZE_ALIGNED 768
 
@@ -42,7 +43,7 @@
  * once locked, kfd driver will stop any further GPU execution.
  * create process (open) will return -EAGAIN.
  */
-static atomic_t kfd_locked = ATOMIC_INIT(0);
+static int kfd_locked;
 
 #ifdef CONFIG_DRM_AMDGPU_CIK
 extern const struct kfd2kgd_calls gfx_v7_kfd2kgd;
@@ -51,20 +52,21 @@ extern const struct kfd2kgd_calls gfx_v8_kfd2kgd;
 extern const struct kfd2kgd_calls gfx_v9_kfd2kgd;
 extern const struct kfd2kgd_calls arcturus_kfd2kgd;
 extern const struct kfd2kgd_calls aldebaran_kfd2kgd;
+extern const struct kfd2kgd_calls gc_9_4_3_kfd2kgd;
 extern const struct kfd2kgd_calls gfx_v10_kfd2kgd;
 extern const struct kfd2kgd_calls gfx_v10_3_kfd2kgd;
 extern const struct kfd2kgd_calls gfx_v11_kfd2kgd;
+extern const struct kfd2kgd_calls gfx_v12_kfd2kgd;
 
 static int kfd_gtt_sa_init(struct kfd_dev *kfd, unsigned int buf_size,
 				unsigned int chunk_size);
 static void kfd_gtt_sa_fini(struct kfd_dev *kfd);
 
-static int kfd_resume_iommu(struct kfd_dev *kfd);
-static int kfd_resume(struct kfd_dev *kfd);
+static int kfd_resume(struct kfd_node *kfd);
 
 static void kfd_device_info_set_sdma_info(struct kfd_dev *kfd)
 {
-	uint32_t sdma_version = kfd->adev->ip_versions[SDMA0_HWIP][0];
+	uint32_t sdma_version = amdgpu_ip_version(kfd->adev, SDMA0_HWIP, 0);
 
 	switch (sdma_version) {
 	case IP_VERSION(4, 0, 0):/* VEGA10 */
@@ -81,6 +83,9 @@ static void kfd_device_info_set_sdma_info(struct kfd_dev *kfd)
 	case IP_VERSION(4, 2, 0):/* VEGA20 */
 	case IP_VERSION(4, 2, 2):/* ARCTURUS */
 	case IP_VERSION(4, 4, 0):/* ALDEBARAN */
+	case IP_VERSION(4, 4, 2):
+	case IP_VERSION(4, 4, 5):
+	case IP_VERSION(4, 4, 4):
 	case IP_VERSION(5, 0, 0):/* NAVI10 */
 	case IP_VERSION(5, 0, 1):/* CYAN_SKILLFISH */
 	case IP_VERSION(5, 0, 2):/* NAVI14 */
@@ -93,6 +98,12 @@ static void kfd_device_info_set_sdma_info(struct kfd_dev *kfd)
 	case IP_VERSION(6, 0, 1):
 	case IP_VERSION(6, 0, 2):
 	case IP_VERSION(6, 0, 3):
+	case IP_VERSION(6, 1, 0):
+	case IP_VERSION(6, 1, 1):
+	case IP_VERSION(6, 1, 2):
+	case IP_VERSION(6, 1, 3):
+	case IP_VERSION(7, 0, 0):
+	case IP_VERSION(7, 0, 1):
 		kfd->device_info.num_sdma_queues_per_engine = 8;
 		break;
 	default:
@@ -102,20 +113,25 @@ static void kfd_device_info_set_sdma_info(struct kfd_dev *kfd)
 		kfd->device_info.num_sdma_queues_per_engine = 8;
 	}
 
+	bitmap_zero(kfd->device_info.reserved_sdma_queues_bitmap, KFD_MAX_SDMA_QUEUES);
+
 	switch (sdma_version) {
 	case IP_VERSION(6, 0, 0):
+	case IP_VERSION(6, 0, 1):
 	case IP_VERSION(6, 0, 2):
 	case IP_VERSION(6, 0, 3):
+	case IP_VERSION(6, 1, 0):
+	case IP_VERSION(6, 1, 1):
+	case IP_VERSION(6, 1, 2):
+	case IP_VERSION(6, 1, 3):
+	case IP_VERSION(7, 0, 0):
+	case IP_VERSION(7, 0, 1):
 		/* Reserve 1 for paging and 1 for gfx */
 		kfd->device_info.num_reserved_sdma_queues_per_engine = 2;
 		/* BIT(0)=engine-0 queue-0; BIT(1)=engine-1 queue-0; BIT(2)=engine-0 queue-1; ... */
-		kfd->device_info.reserved_sdma_queues_bitmap = 0xFULL;
-		break;
-	case IP_VERSION(6, 0, 1):
-		/* Reserve 1 for paging and 1 for gfx */
-		kfd->device_info.num_reserved_sdma_queues_per_engine = 2;
-		/* BIT(0)=engine-0 queue-0; BIT(1)=engine-0 queue-1; ... */
-		kfd->device_info.reserved_sdma_queues_bitmap = 0x3ULL;
+		bitmap_set(kfd->device_info.reserved_sdma_queues_bitmap, 0,
+			   kfd->adev->sdma.num_instances *
+			   kfd->device_info.num_reserved_sdma_queues_per_engine);
 		break;
 	default:
 		break;
@@ -135,6 +151,14 @@ static void kfd_device_info_set_event_interrupt_class(struct kfd_dev *kfd)
 	case IP_VERSION(9, 4, 0): /* VEGA20 */
 	case IP_VERSION(9, 4, 1): /* ARCTURUS */
 	case IP_VERSION(9, 4, 2): /* ALDEBARAN */
+		kfd->device_info.event_interrupt_class = &event_interrupt_class_v9;
+		break;
+	case IP_VERSION(9, 4, 3): /* GC 9.4.3 */
+	case IP_VERSION(9, 4, 4): /* GC 9.4.4 */
+	case IP_VERSION(9, 5, 0): /* GC 9.5.0 */
+		kfd->device_info.event_interrupt_class =
+						&event_interrupt_class_v9_4_3;
+		break;
 	case IP_VERSION(10, 3, 1): /* VANGOGH */
 	case IP_VERSION(10, 3, 3): /* YELLOW_CARP */
 	case IP_VERSION(10, 3, 6): /* GC 10.3.6 */
@@ -148,13 +172,22 @@ static void kfd_device_info_set_event_interrupt_class(struct kfd_dev *kfd)
 	case IP_VERSION(10, 3, 2): /* NAVY_FLOUNDER */
 	case IP_VERSION(10, 3, 4): /* DIMGREY_CAVEFISH */
 	case IP_VERSION(10, 3, 5): /* BEIGE_GOBY */
-		kfd->device_info.event_interrupt_class = &event_interrupt_class_v9;
+		kfd->device_info.event_interrupt_class = &event_interrupt_class_v10;
 		break;
 	case IP_VERSION(11, 0, 0):
 	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 3):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
+		kfd->device_info.event_interrupt_class = &event_interrupt_class_v11;
+		break;
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		/* GFX12_TODO: Change to v12 version. */
 		kfd->device_info.event_interrupt_class = &event_interrupt_class_v11;
 		break;
 	default:
@@ -185,11 +218,6 @@ static void kfd_device_info_init(struct kfd_dev *kfd,
 
 		kfd_device_info_set_event_interrupt_class(kfd);
 
-		/* Raven */
-		if (gc_version == IP_VERSION(9, 1, 0) ||
-		    gc_version == IP_VERSION(9, 2, 2))
-			kfd->device_info.needs_iommu_device = true;
-
 		if (gc_version < IP_VERSION(11, 0, 0)) {
 			/* Navi2x+, Navi1x+ */
 			if (gc_version == IP_VERSION(10, 3, 6))
@@ -212,6 +240,11 @@ static void kfd_device_info_init(struct kfd_dev *kfd,
 			 */
 			kfd->device_info.needs_pci_atomics = true;
 			kfd->device_info.no_atomic_fw_version = kfd->adev->gfx.rs64_enable ? 509 : 0;
+		} else if (gc_version < IP_VERSION(13, 0, 0)) {
+			kfd->device_info.needs_pci_atomics = true;
+			kfd->device_info.no_atomic_fw_version = 2090;
+		} else {
+			kfd->device_info.needs_pci_atomics = true;
 		}
 	} else {
 		kfd->device_info.doorbell_size = 4;
@@ -224,10 +257,6 @@ static void kfd_device_info_init(struct kfd_dev *kfd,
 		    asic_type != CHIP_TONGA)
 			kfd->device_info.supports_cwsr = true;
 
-		if (asic_type == CHIP_KAVERI ||
-		    asic_type == CHIP_CARRIZO)
-			kfd->device_info.needs_iommu_device = true;
-
 		if (asic_type != CHIP_HAWAII && !vf)
 			kfd->device_info.needs_pci_atomics = true;
 	}
@@ -240,7 +269,6 @@ struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf)
 	uint32_t gfx_target_version = 0;
 
 	switch (adev->asic_type) {
-#ifdef KFD_SUPPORT_IOMMU_V2
 #ifdef CONFIG_DRM_AMDGPU_CIK
 	case CHIP_KAVERI:
 		gfx_target_version = 70000;
@@ -253,7 +281,6 @@ struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf)
 		if (!vf)
 			f2g = &gfx_v8_kfd2kgd;
 		break;
-#endif
 #ifdef CONFIG_DRM_AMDGPU_CIK
 	case CHIP_HAWAII:
 		gfx_target_version = 70001;
@@ -283,13 +310,12 @@ struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf)
 			f2g = &gfx_v8_kfd2kgd;
 		break;
 	default:
-		switch (adev->ip_versions[GC_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 		/* Vega 10 */
 		case IP_VERSION(9, 0, 1):
 			gfx_target_version = 90000;
 			f2g = &gfx_v9_kfd2kgd;
 			break;
-#ifdef KFD_SUPPORT_IOMMU_V2
 		/* Raven */
 		case IP_VERSION(9, 1, 0):
 		case IP_VERSION(9, 2, 2):
@@ -297,7 +323,6 @@ struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf)
 			if (!vf)
 				f2g = &gfx_v9_kfd2kgd;
 			break;
-#endif
 		/* Vega12 */
 		case IP_VERSION(9, 2, 1):
 			gfx_target_version = 90004;
@@ -327,8 +352,13 @@ struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf)
 			f2g = &aldebaran_kfd2kgd;
 			break;
 		case IP_VERSION(9, 4, 3):
-			gfx_target_version = 90400;
-			f2g = &aldebaran_kfd2kgd;
+		case IP_VERSION(9, 4, 4):
+			gfx_target_version = 90402;
+			f2g = &gc_9_4_3_kfd2kgd;
+			break;
+		case IP_VERSION(9, 5, 0):
+			gfx_target_version = 90500;
+			f2g = &gc_9_4_3_kfd2kgd;
 			break;
 		/* Navi10 */
 		case IP_VERSION(10, 1, 10):
@@ -410,6 +440,30 @@ struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf)
 			gfx_target_version = 110001;
 			f2g = &gfx_v11_kfd2kgd;
 			break;
+		case IP_VERSION(11, 5, 0):
+			gfx_target_version = 110500;
+			f2g = &gfx_v11_kfd2kgd;
+			break;
+		case IP_VERSION(11, 5, 1):
+			gfx_target_version = 110501;
+			f2g = &gfx_v11_kfd2kgd;
+			break;
+		case IP_VERSION(11, 5, 2):
+			gfx_target_version = 110502;
+			f2g = &gfx_v11_kfd2kgd;
+			break;
+		case IP_VERSION(11, 5, 3):
+			gfx_target_version = 110503;
+			f2g = &gfx_v11_kfd2kgd;
+			break;
+		case IP_VERSION(12, 0, 0):
+			gfx_target_version = 120000;
+			f2g = &gfx_v12_kfd2kgd;
+			break;
+		case IP_VERSION(12, 0, 1):
+			gfx_target_version = 120001;
+			f2g = &gfx_v12_kfd2kgd;
+			break;
 		default:
 			break;
 		}
@@ -417,11 +471,13 @@ struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf)
 	}
 
 	if (!f2g) {
-		if (adev->ip_versions[GC_HWIP][0])
-			dev_err(kfd_device, "GC IP %06x %s not supported in kfd\n",
-				adev->ip_versions[GC_HWIP][0], vf ? "VF" : "");
+		if (amdgpu_ip_version(adev, GC_HWIP, 0))
+			dev_info(kfd_device,
+				"GC IP %06x %s not supported in kfd\n",
+				amdgpu_ip_version(adev, GC_HWIP, 0),
+				vf ? "VF" : "");
 		else
-			dev_err(kfd_device, "%s %s not supported in kfd\n",
+			dev_info(kfd_device, "%s %s not supported in kfd\n",
 				amdgpu_asic_name[adev->asic_type], vf ? "VF" : "");
 		return NULL;
 	}
@@ -437,12 +493,9 @@ struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf)
 	atomic_set(&kfd->compute_profile, 0);
 
 	mutex_init(&kfd->doorbell_mutex);
-	memset(&kfd->doorbell_available_index, 0,
-		sizeof(kfd->doorbell_available_index));
-
-	atomic_set(&kfd->sram_ecc_flag, 0);
 
 	ida_init(&kfd->doorbell_ida);
+	atomic_set(&kfd->kfd_processes_count, 0);
 
 	return kfd;
 }
@@ -451,78 +504,222 @@ static void kfd_cwsr_init(struct kfd_dev *kfd)
 {
 	if (cwsr_enable && kfd->device_info.supports_cwsr) {
 		if (KFD_GC_VERSION(kfd) < IP_VERSION(9, 0, 1)) {
-			BUILD_BUG_ON(sizeof(cwsr_trap_gfx8_hex) > PAGE_SIZE);
+			BUILD_BUG_ON(sizeof(cwsr_trap_gfx8_hex)
+					     > KFD_CWSR_TMA_OFFSET);
 			kfd->cwsr_isa = cwsr_trap_gfx8_hex;
 			kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx8_hex);
 		} else if (KFD_GC_VERSION(kfd) == IP_VERSION(9, 4, 1)) {
-			BUILD_BUG_ON(sizeof(cwsr_trap_arcturus_hex) > PAGE_SIZE);
+			BUILD_BUG_ON(sizeof(cwsr_trap_arcturus_hex)
+					     > KFD_CWSR_TMA_OFFSET);
 			kfd->cwsr_isa = cwsr_trap_arcturus_hex;
 			kfd->cwsr_isa_size = sizeof(cwsr_trap_arcturus_hex);
 		} else if (KFD_GC_VERSION(kfd) == IP_VERSION(9, 4, 2)) {
-			BUILD_BUG_ON(sizeof(cwsr_trap_aldebaran_hex) > PAGE_SIZE);
+			BUILD_BUG_ON(sizeof(cwsr_trap_aldebaran_hex)
+					     > KFD_CWSR_TMA_OFFSET);
 			kfd->cwsr_isa = cwsr_trap_aldebaran_hex;
 			kfd->cwsr_isa_size = sizeof(cwsr_trap_aldebaran_hex);
-		} else if (KFD_GC_VERSION(kfd) == IP_VERSION(9, 4, 3)) {
-			BUILD_BUG_ON(sizeof(cwsr_trap_gfx9_4_3_hex) > PAGE_SIZE);
+		} else if (KFD_GC_VERSION(kfd) == IP_VERSION(9, 4, 3) ||
+			   KFD_GC_VERSION(kfd) == IP_VERSION(9, 4, 4)) {
+			BUILD_BUG_ON(sizeof(cwsr_trap_gfx9_4_3_hex)
+					     > KFD_CWSR_TMA_OFFSET);
 			kfd->cwsr_isa = cwsr_trap_gfx9_4_3_hex;
 			kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx9_4_3_hex);
+		} else if (KFD_GC_VERSION(kfd) == IP_VERSION(9, 5, 0)) {
+			BUILD_BUG_ON(sizeof(cwsr_trap_gfx9_5_0_hex) > PAGE_SIZE);
+			kfd->cwsr_isa = cwsr_trap_gfx9_5_0_hex;
+			kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx9_5_0_hex);
 		} else if (KFD_GC_VERSION(kfd) < IP_VERSION(10, 1, 1)) {
-			BUILD_BUG_ON(sizeof(cwsr_trap_gfx9_hex) > PAGE_SIZE);
+			BUILD_BUG_ON(sizeof(cwsr_trap_gfx9_hex)
+					     > KFD_CWSR_TMA_OFFSET);
 			kfd->cwsr_isa = cwsr_trap_gfx9_hex;
 			kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx9_hex);
 		} else if (KFD_GC_VERSION(kfd) < IP_VERSION(10, 3, 0)) {
-			BUILD_BUG_ON(sizeof(cwsr_trap_nv1x_hex) > PAGE_SIZE);
+			BUILD_BUG_ON(sizeof(cwsr_trap_nv1x_hex)
+					     > KFD_CWSR_TMA_OFFSET);
 			kfd->cwsr_isa = cwsr_trap_nv1x_hex;
 			kfd->cwsr_isa_size = sizeof(cwsr_trap_nv1x_hex);
 		} else if (KFD_GC_VERSION(kfd) < IP_VERSION(11, 0, 0)) {
-			BUILD_BUG_ON(sizeof(cwsr_trap_gfx10_hex) > PAGE_SIZE);
+			BUILD_BUG_ON(sizeof(cwsr_trap_gfx10_hex)
+					     > KFD_CWSR_TMA_OFFSET);
 			kfd->cwsr_isa = cwsr_trap_gfx10_hex;
 			kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx10_hex);
-		} else {
+		} else if (KFD_GC_VERSION(kfd) < IP_VERSION(12, 0, 0)) {
+			/* The gfx11 cwsr trap handler must fit inside a single
+			   page. */
 			BUILD_BUG_ON(sizeof(cwsr_trap_gfx11_hex) > PAGE_SIZE);
 			kfd->cwsr_isa = cwsr_trap_gfx11_hex;
 			kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx11_hex);
+		} else {
+			BUILD_BUG_ON(sizeof(cwsr_trap_gfx12_hex)
+					     > KFD_CWSR_TMA_OFFSET);
+			kfd->cwsr_isa = cwsr_trap_gfx12_hex;
+			kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx12_hex);
 		}
 
 		kfd->cwsr_enabled = true;
 	}
 }
 
-static int kfd_gws_init(struct kfd_dev *kfd)
+static int kfd_gws_init(struct kfd_node *node)
 {
 	int ret = 0;
+	struct kfd_dev *kfd = node->kfd;
+	uint32_t mes_rev = node->adev->mes.sched_version & AMDGPU_MES_VERSION_MASK;
 
-	if (kfd->dqm->sched_policy == KFD_SCHED_POLICY_NO_HWS)
+	if (node->dqm->sched_policy == KFD_SCHED_POLICY_NO_HWS)
 		return 0;
 
-	if (hws_gws_support || (KFD_IS_SOC15(kfd) &&
-		((KFD_GC_VERSION(kfd) == IP_VERSION(9, 0, 1)
+	if (hws_gws_support || (KFD_IS_SOC15(node) &&
+		((KFD_GC_VERSION(node) == IP_VERSION(9, 0, 1)
 			&& kfd->mec2_fw_version >= 0x81b3) ||
-		(KFD_GC_VERSION(kfd) <= IP_VERSION(9, 4, 0)
+		(KFD_GC_VERSION(node) <= IP_VERSION(9, 4, 0)
 			&& kfd->mec2_fw_version >= 0x1b3)  ||
-		(KFD_GC_VERSION(kfd) == IP_VERSION(9, 4, 1)
+		(KFD_GC_VERSION(node) == IP_VERSION(9, 4, 1)
 			&& kfd->mec2_fw_version >= 0x30)   ||
-		(KFD_GC_VERSION(kfd) == IP_VERSION(9, 4, 2)
+		(KFD_GC_VERSION(node) == IP_VERSION(9, 4, 2)
 			&& kfd->mec2_fw_version >= 0x28) ||
-		(KFD_GC_VERSION(kfd) >= IP_VERSION(10, 3, 0)
-			&& KFD_GC_VERSION(kfd) < IP_VERSION(11, 0, 0)
-			&& kfd->mec2_fw_version >= 0x6b))))
-		ret = amdgpu_amdkfd_alloc_gws(kfd->adev,
-				kfd->adev->gds.gws_size, &kfd->gws);
+		(KFD_GC_VERSION(node) == IP_VERSION(9, 4, 3) ||
+		 KFD_GC_VERSION(node) == IP_VERSION(9, 4, 4)) ||
+		(KFD_GC_VERSION(node) == IP_VERSION(9, 5, 0)) ||
+		(KFD_GC_VERSION(node) >= IP_VERSION(10, 3, 0)
+			&& KFD_GC_VERSION(node) < IP_VERSION(11, 0, 0)
+			&& kfd->mec2_fw_version >= 0x6b) ||
+		(KFD_GC_VERSION(node) >= IP_VERSION(11, 0, 0)
+			&& KFD_GC_VERSION(node) < IP_VERSION(12, 0, 0)
+			&& mes_rev >= 68) ||
+		(KFD_GC_VERSION(node) >= IP_VERSION(12, 0, 0))))) {
+		if (KFD_GC_VERSION(node) >= IP_VERSION(12, 0, 0))
+			node->adev->gds.gws_size = 64;
+		ret = amdgpu_amdkfd_alloc_gws(node->adev,
+				node->adev->gds.gws_size, &node->gws);
+	}
 
 	return ret;
 }
 
-static void kfd_smi_init(struct kfd_dev *dev)
+static void kfd_smi_init(struct kfd_node *dev)
 {
 	INIT_LIST_HEAD(&dev->smi_clients);
 	spin_lock_init(&dev->smi_lock);
 }
 
+static int kfd_init_node(struct kfd_node *node)
+{
+	int err = -1;
+
+	if (kfd_interrupt_init(node)) {
+		dev_err(kfd_device, "Error initializing interrupts\n");
+		goto kfd_interrupt_error;
+	}
+
+	node->dqm = device_queue_manager_init(node);
+	if (!node->dqm) {
+		dev_err(kfd_device, "Error initializing queue manager\n");
+		goto device_queue_manager_error;
+	}
+
+	if (kfd_gws_init(node)) {
+		dev_err(kfd_device, "Could not allocate %d gws\n",
+			node->adev->gds.gws_size);
+		goto gws_error;
+	}
+
+	if (kfd_resume(node))
+		goto kfd_resume_error;
+
+	if (kfd_topology_add_device(node)) {
+		dev_err(kfd_device, "Error adding device to topology\n");
+		goto kfd_topology_add_device_error;
+	}
+
+	kfd_smi_init(node);
+
+	return 0;
+
+kfd_topology_add_device_error:
+kfd_resume_error:
+gws_error:
+	device_queue_manager_uninit(node->dqm);
+device_queue_manager_error:
+	kfd_interrupt_exit(node);
+kfd_interrupt_error:
+	if (node->gws)
+		amdgpu_amdkfd_free_gws(node->adev, node->gws);
+
+	/* Cleanup the node memory here */
+	kfree(node);
+	return err;
+}
+
+static void kfd_cleanup_nodes(struct kfd_dev *kfd, unsigned int num_nodes)
+{
+	struct kfd_node *knode;
+	unsigned int i;
+
+	/*
+	 * flush_work ensures that there are no outstanding
+	 * work-queue items that will access interrupt_ring. New work items
+	 * can't be created because we stopped interrupt handling above.
+	 */
+	flush_workqueue(kfd->ih_wq);
+	destroy_workqueue(kfd->ih_wq);
+
+	for (i = 0; i < num_nodes; i++) {
+		knode = kfd->nodes[i];
+		device_queue_manager_uninit(knode->dqm);
+		kfd_interrupt_exit(knode);
+		kfd_topology_remove_device(knode);
+		if (knode->gws)
+			amdgpu_amdkfd_free_gws(knode->adev, knode->gws);
+		kfree(knode);
+		kfd->nodes[i] = NULL;
+	}
+}
+
+static void kfd_setup_interrupt_bitmap(struct kfd_node *node,
+				       unsigned int kfd_node_idx)
+{
+	struct amdgpu_device *adev = node->adev;
+	uint32_t xcc_mask = node->xcc_mask;
+	uint32_t xcc, mapped_xcc;
+	/*
+	 * Interrupt bitmap is setup for processing interrupts from
+	 * different XCDs and AIDs.
+	 * Interrupt bitmap is defined as follows:
+	 * 1. Bits 0-15 - correspond to the NodeId field.
+	 *    Each bit corresponds to NodeId number. For example, if
+	 *    a KFD node has interrupt bitmap set to 0x7, then this
+	 *    KFD node will process interrupts with NodeId = 0, 1 and 2
+	 *    in the IH cookie.
+	 * 2. Bits 16-31 - unused.
+	 *
+	 * Please note that the kfd_node_idx argument passed to this
+	 * function is not related to NodeId field received in the
+	 * IH cookie.
+	 *
+	 * In CPX mode, a KFD node will process an interrupt if:
+	 * - the Node Id matches the corresponding bit set in
+	 *   Bits 0-15.
+	 * - AND VMID reported in the interrupt lies within the
+	 *   VMID range of the node.
+	 */
+	for_each_inst(xcc, xcc_mask) {
+		mapped_xcc = GET_INST(GC, xcc);
+		node->interrupt_bitmap |= (mapped_xcc % 2 ? 5 : 3) << (4 * (mapped_xcc / 2));
+	}
+	dev_info(kfd_device, "Node: %d, interrupt_bitmap: %x\n", kfd_node_idx,
+							node->interrupt_bitmap);
+}
+
 bool kgd2kfd_device_init(struct kfd_dev *kfd,
 			 const struct kgd2kfd_shared_resources *gpu_resources)
 {
-	unsigned int size, map_process_packet_size;
+	unsigned int size, map_process_packet_size, i;
+	struct kfd_node *node;
+	uint32_t first_vmid_kfd, last_vmid_kfd, vmid_num_kfd;
+	unsigned int max_proc_per_quantum;
+	int partition_mode;
+	int xcp_idx;
 
 	kfd->mec_fw_version = amdgpu_amdkfd_get_fw_version(kfd->adev,
 			KGD_ENGINE_MEC1);
@@ -532,10 +729,14 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
 			KGD_ENGINE_SDMA1);
 	kfd->shared_resources = *gpu_resources;
 
-	kfd->vm_info.first_vmid_kfd = ffs(gpu_resources->compute_vmid_bitmap)-1;
-	kfd->vm_info.last_vmid_kfd = fls(gpu_resources->compute_vmid_bitmap)-1;
-	kfd->vm_info.vmid_num_kfd = kfd->vm_info.last_vmid_kfd
-			- kfd->vm_info.first_vmid_kfd + 1;
+	kfd->num_nodes = amdgpu_xcp_get_num_xcp(kfd->adev->xcp_mgr);
+
+	if (kfd->num_nodes == 0) {
+		dev_err(kfd_device,
+			"KFD num nodes cannot be 0, num_xcc_in_node: %d\n",
+			kfd->adev->gfx.num_xcc_per_xcp);
+		goto out;
+	}
 
 	/* Allow BIF to recode atomics to PCIe 3.0 AtomicOps.
 	 * 32 and 64-bit requests are possible and must be
@@ -554,11 +755,34 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
 		return false;
 	}
 
+	first_vmid_kfd = ffs(gpu_resources->compute_vmid_bitmap)-1;
+	last_vmid_kfd = fls(gpu_resources->compute_vmid_bitmap)-1;
+	vmid_num_kfd = last_vmid_kfd - first_vmid_kfd + 1;
+
+	/* For multi-partition capable GPUs, we need special handling for VMIDs
+	 * depending on partition mode.
+	 * In CPX mode, the VMID range needs to be shared between XCDs.
+	 * Additionally, there are 13 VMIDs (3-15) available for KFD. To
+	 * divide them equally, we change starting VMID to 4 and not use
+	 * VMID 3.
+	 * If the VMID range changes for multi-partition capable GPUs, then
+	 * this code MUST be revisited.
+	 */
+	if (kfd->adev->xcp_mgr) {
+		partition_mode = amdgpu_xcp_query_partition_mode(kfd->adev->xcp_mgr,
+								 AMDGPU_XCP_FL_LOCKED);
+		if (partition_mode == AMDGPU_CPX_PARTITION_MODE &&
+		    kfd->num_nodes != 1) {
+			vmid_num_kfd /= 2;
+			first_vmid_kfd = last_vmid_kfd + 1 - vmid_num_kfd*2;
+		}
+	}
+
 	/* Verify module parameters regarding mapped process number*/
 	if (hws_max_conc_proc >= 0)
-		kfd->max_proc_per_quantum = min((u32)hws_max_conc_proc, kfd->vm_info.vmid_num_kfd);
+		max_proc_per_quantum = min((u32)hws_max_conc_proc, vmid_num_kfd);
 	else
-		kfd->max_proc_per_quantum = kfd->vm_info.vmid_num_kfd;
+		max_proc_per_quantum = vmid_num_kfd;
 
 	/* calculate max size of mqds needed for queues */
 	size = max_num_of_queues_per_device *
@@ -606,81 +830,112 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
 	if (amdgpu_use_xgmi_p2p)
 		kfd->hive_id = kfd->adev->gmc.xgmi.hive_id;
 
-	kfd->noretry = kfd->adev->gmc.noretry;
+	/*
+	 * For multi-partition capable GPUs, the KFD abstracts all partitions
+	 * within a socket as xGMI connected in the topology so assign a unique
+	 * hive id per device based on the pci device location if device is in
+	 * PCIe mode.
+	 */
+	if (!kfd->hive_id && kfd->num_nodes > 1)
+		kfd->hive_id = pci_dev_id(kfd->adev->pdev);
 
-	if (kfd_interrupt_init(kfd)) {
-		dev_err(kfd_device, "Error initializing interrupts\n");
-		goto kfd_interrupt_error;
-	}
+	kfd->noretry = kfd->adev->gmc.noretry;
 
-	kfd->dqm = device_queue_manager_init(kfd);
-	if (!kfd->dqm) {
-		dev_err(kfd_device, "Error initializing queue manager\n");
-		goto device_queue_manager_error;
-	}
+	kfd_cwsr_init(kfd);
 
-	/* If supported on this device, allocate global GWS that is shared
-	 * by all KFD processes
-	 */
-	if (kfd_gws_init(kfd)) {
-		dev_err(kfd_device, "Could not allocate %d gws\n",
-			kfd->adev->gds.gws_size);
-		goto gws_error;
-	}
+	dev_info(kfd_device, "Total number of KFD nodes to be created: %d\n",
+				kfd->num_nodes);
+
+	/* Allocate the KFD nodes */
+	for (i = 0, xcp_idx = 0; i < kfd->num_nodes; i++) {
+		node = kzalloc(sizeof(struct kfd_node), GFP_KERNEL);
+		if (!node)
+			goto node_alloc_error;
+
+		node->node_id = i;
+		node->adev = kfd->adev;
+		node->kfd = kfd;
+		node->kfd2kgd = kfd->kfd2kgd;
+		node->vm_info.vmid_num_kfd = vmid_num_kfd;
+		node->xcp = amdgpu_get_next_xcp(kfd->adev->xcp_mgr, &xcp_idx);
+		/* TODO : Check if error handling is needed */
+		if (node->xcp) {
+			amdgpu_xcp_get_inst_details(node->xcp, AMDGPU_XCP_GFX,
+						    &node->xcc_mask);
+			++xcp_idx;
+		} else {
+			node->xcc_mask =
+				(1U << NUM_XCC(kfd->adev->gfx.xcc_mask)) - 1;
+		}
 
-	/* If CRAT is broken, won't set iommu enabled */
-	kfd_double_confirm_iommu_support(kfd);
+		if (node->xcp) {
+			dev_info(kfd_device, "KFD node %d partition %d size %lldM\n",
+				node->node_id, node->xcp->mem_id,
+				KFD_XCP_MEMORY_SIZE(node->adev, node->node_id) >> 20);
+		}
 
-	if (kfd_iommu_device_init(kfd)) {
-		kfd->use_iommu_v2 = false;
-		dev_err(kfd_device, "Error initializing iommuv2\n");
-		goto device_iommu_error;
-	}
+		if (partition_mode == AMDGPU_CPX_PARTITION_MODE &&
+		    kfd->num_nodes != 1) {
+			/* For multi-partition capable GPUs and CPX mode, first
+			 * XCD gets VMID range 4-9 and second XCD gets VMID
+			 * range 10-15.
+			 */
 
-	kfd_cwsr_init(kfd);
+			node->vm_info.first_vmid_kfd = (i%2 == 0) ?
+						first_vmid_kfd :
+						first_vmid_kfd+vmid_num_kfd;
+			node->vm_info.last_vmid_kfd = (i%2 == 0) ?
+						last_vmid_kfd-vmid_num_kfd :
+						last_vmid_kfd;
+			node->compute_vmid_bitmap =
+				((0x1 << (node->vm_info.last_vmid_kfd + 1)) - 1) -
+				((0x1 << (node->vm_info.first_vmid_kfd)) - 1);
+		} else {
+			node->vm_info.first_vmid_kfd = first_vmid_kfd;
+			node->vm_info.last_vmid_kfd = last_vmid_kfd;
+			node->compute_vmid_bitmap =
+				gpu_resources->compute_vmid_bitmap;
+		}
+		node->max_proc_per_quantum = max_proc_per_quantum;
+		atomic_set(&node->sram_ecc_flag, 0);
 
-	svm_migrate_init(kfd->adev);
+		amdgpu_amdkfd_get_local_mem_info(kfd->adev,
+					&node->local_mem_info, node->xcp);
 
-	if (kfd_resume_iommu(kfd))
-		goto device_iommu_error;
+		if (kfd->adev->xcp_mgr)
+			kfd_setup_interrupt_bitmap(node, i);
 
-	if (kfd_resume(kfd))
-		goto kfd_resume_error;
+		/* Initialize the KFD node */
+		if (kfd_init_node(node)) {
+			dev_err(kfd_device, "Error initializing KFD node\n");
+			goto node_init_error;
+		}
 
-	amdgpu_amdkfd_get_local_mem_info(kfd->adev, &kfd->local_mem_info);
+		spin_lock_init(&node->watch_points_lock);
 
-	if (kfd_topology_add_device(kfd)) {
-		dev_err(kfd_device, "Error adding device to topology\n");
-		goto kfd_topology_add_device_error;
+		kfd->nodes[i] = node;
 	}
 
-	kfd_smi_init(kfd);
+	svm_range_set_max_pages(kfd->adev);
 
 	kfd->init_complete = true;
 	dev_info(kfd_device, "added device %x:%x\n", kfd->adev->pdev->vendor,
 		 kfd->adev->pdev->device);
 
 	pr_debug("Starting kfd with the following scheduling policy %d\n",
-		kfd->dqm->sched_policy);
+		node->dqm->sched_policy);
 
 	goto out;
 
-kfd_topology_add_device_error:
-kfd_resume_error:
-device_iommu_error:
-gws_error:
-	device_queue_manager_uninit(kfd->dqm);
-device_queue_manager_error:
-	kfd_interrupt_exit(kfd);
-kfd_interrupt_error:
+node_init_error:
+node_alloc_error:
+	kfd_cleanup_nodes(kfd, i);
 	kfd_doorbell_fini(kfd);
 kfd_doorbell_error:
 	kfd_gtt_sa_fini(kfd);
 kfd_gtt_sa_init_error:
-	amdgpu_amdkfd_free_gtt_mem(kfd->adev, kfd->gtt_mem);
+	amdgpu_amdkfd_free_gtt_mem(kfd->adev, &kfd->gtt_mem);
 alloc_gtt_mem_failure:
-	if (kfd->gws)
-		amdgpu_amdkfd_free_gws(kfd->adev, kfd->gws);
 	dev_err(kfd_device,
 		"device %x:%x NOT added due to errors\n",
 		kfd->adev->pdev->vendor, kfd->adev->pdev->device);
@@ -691,32 +946,37 @@ out:
 void kgd2kfd_device_exit(struct kfd_dev *kfd)
 {
 	if (kfd->init_complete) {
-		device_queue_manager_uninit(kfd->dqm);
-		kfd_interrupt_exit(kfd);
-		kfd_topology_remove_device(kfd);
+		/* Cleanup KFD nodes */
+		kfd_cleanup_nodes(kfd, kfd->num_nodes);
+		/* Cleanup common/shared resources */
 		kfd_doorbell_fini(kfd);
 		ida_destroy(&kfd->doorbell_ida);
 		kfd_gtt_sa_fini(kfd);
-		amdgpu_amdkfd_free_gtt_mem(kfd->adev, kfd->gtt_mem);
-		if (kfd->gws)
-			amdgpu_amdkfd_free_gws(kfd->adev, kfd->gws);
+		amdgpu_amdkfd_free_gtt_mem(kfd->adev, &kfd->gtt_mem);
 	}
 
 	kfree(kfd);
 }
 
-int kgd2kfd_pre_reset(struct kfd_dev *kfd)
+int kgd2kfd_pre_reset(struct kfd_dev *kfd,
+		      struct amdgpu_reset_context *reset_context)
 {
+	struct kfd_node *node;
+	int i;
+
 	if (!kfd->init_complete)
 		return 0;
 
-	kfd_smi_event_update_gpu_reset(kfd, false);
+	for (i = 0; i < kfd->num_nodes; i++) {
+		node = kfd->nodes[i];
+		kfd_smi_event_update_gpu_reset(node, false, reset_context);
+	}
 
-	kfd->dqm->ops.pre_reset(kfd->dqm);
+	kgd2kfd_suspend(kfd, true);
 
-	kgd2kfd_suspend(kfd, false);
+	for (i = 0; i < kfd->num_nodes; i++)
+		kfd_signal_reset_event(kfd->nodes[i]);
 
-	kfd_signal_reset_event(kfd);
 	return 0;
 }
 
@@ -729,119 +989,141 @@ int kgd2kfd_pre_reset(struct kfd_dev *kfd)
 int kgd2kfd_post_reset(struct kfd_dev *kfd)
 {
 	int ret;
+	struct kfd_node *node;
+	int i;
 
 	if (!kfd->init_complete)
 		return 0;
 
-	ret = kfd_resume(kfd);
-	if (ret)
-		return ret;
-	atomic_dec(&kfd_locked);
+	for (i = 0; i < kfd->num_nodes; i++) {
+		ret = kfd_resume(kfd->nodes[i]);
+		if (ret)
+			return ret;
+	}
 
-	atomic_set(&kfd->sram_ecc_flag, 0);
+	mutex_lock(&kfd_processes_mutex);
+	--kfd_locked;
+	mutex_unlock(&kfd_processes_mutex);
 
-	kfd_smi_event_update_gpu_reset(kfd, true);
+	for (i = 0; i < kfd->num_nodes; i++) {
+		node = kfd->nodes[i];
+		atomic_set(&node->sram_ecc_flag, 0);
+		kfd_smi_event_update_gpu_reset(node, true, NULL);
+	}
 
 	return 0;
 }
 
-bool kfd_is_locked(void)
+bool kfd_is_locked(struct kfd_dev *kfd)
 {
-	return  (atomic_read(&kfd_locked) > 0);
+	uint8_t id  = 0;
+	struct kfd_node *dev;
+
+	lockdep_assert_held(&kfd_processes_mutex);
+
+	/* check reset/suspend lock */
+	if (kfd_locked > 0)
+		return true;
+
+	if (kfd)
+		return kfd->kfd_dev_lock > 0;
+
+	/* check lock on all cgroup accessible devices */
+	while (kfd_topology_enum_kfd_devices(id++, &dev) == 0) {
+		if (!dev || kfd_devcgroup_check_permission(dev))
+			continue;
+
+		if (dev->kfd->kfd_dev_lock > 0)
+			return true;
+	}
+
+	return false;
 }
 
-void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm)
+void kgd2kfd_suspend(struct kfd_dev *kfd, bool suspend_proc)
 {
+	struct kfd_node *node;
+	int i;
+
 	if (!kfd->init_complete)
 		return;
 
-	/* for runtime suspend, skip locking kfd */
-	if (!run_pm) {
-		/* For first KFD device suspend all the KFD processes */
-		if (atomic_inc_return(&kfd_locked) == 1)
-			kfd_suspend_all_processes();
-	}
+	if (suspend_proc)
+		kgd2kfd_suspend_process(kfd);
 
-	kfd->dqm->ops.stop(kfd->dqm);
-	kfd_iommu_suspend(kfd);
+	for (i = 0; i < kfd->num_nodes; i++) {
+		node = kfd->nodes[i];
+		node->dqm->ops.stop(node->dqm);
+	}
 }
 
-int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm)
+int kgd2kfd_resume(struct kfd_dev *kfd, bool resume_proc)
 {
-	int ret, count;
+	int ret, i;
 
 	if (!kfd->init_complete)
 		return 0;
 
-	ret = kfd_resume(kfd);
-	if (ret)
-		return ret;
-
-	/* for runtime resume, skip unlocking kfd */
-	if (!run_pm) {
-		count = atomic_dec_return(&kfd_locked);
-		WARN_ONCE(count < 0, "KFD suspend / resume ref. error");
-		if (count == 0)
-			ret = kfd_resume_all_processes();
+	for (i = 0; i < kfd->num_nodes; i++) {
+		ret = kfd_resume(kfd->nodes[i]);
+		if (ret)
+			return ret;
 	}
 
+	if (resume_proc)
+		ret = kgd2kfd_resume_process(kfd);
+
 	return ret;
 }
 
-int kgd2kfd_resume_iommu(struct kfd_dev *kfd)
+void kgd2kfd_suspend_process(struct kfd_dev *kfd)
 {
 	if (!kfd->init_complete)
-		return 0;
+		return;
 
-	return kfd_resume_iommu(kfd);
+	mutex_lock(&kfd_processes_mutex);
+	/* For first KFD device suspend all the KFD processes */
+	if (++kfd_locked == 1)
+		kfd_suspend_all_processes();
+	mutex_unlock(&kfd_processes_mutex);
 }
 
-static int kfd_resume_iommu(struct kfd_dev *kfd)
+int kgd2kfd_resume_process(struct kfd_dev *kfd)
 {
-	int err = 0;
+	int ret = 0;
 
-	err = kfd_iommu_resume(kfd);
-	if (err)
-		dev_err(kfd_device,
-			"Failed to resume IOMMU for device %x:%x\n",
-			kfd->adev->pdev->vendor, kfd->adev->pdev->device);
-	return err;
+	if (!kfd->init_complete)
+		return 0;
+
+	mutex_lock(&kfd_processes_mutex);
+	if (--kfd_locked == 0)
+		ret = kfd_resume_all_processes();
+	WARN_ONCE(kfd_locked < 0, "KFD suspend / resume ref. error");
+	mutex_unlock(&kfd_processes_mutex);
+
+	return ret;
 }
 
-static int kfd_resume(struct kfd_dev *kfd)
+static int kfd_resume(struct kfd_node *node)
 {
 	int err = 0;
 
-	err = kfd->dqm->ops.start(kfd->dqm);
+	err = node->dqm->ops.start(node->dqm);
 	if (err)
 		dev_err(kfd_device,
 			"Error starting queue manager for device %x:%x\n",
-			kfd->adev->pdev->vendor, kfd->adev->pdev->device);
+			node->adev->pdev->vendor, node->adev->pdev->device);
 
 	return err;
 }
 
-static inline void kfd_queue_work(struct workqueue_struct *wq,
-				  struct work_struct *work)
-{
-	int cpu, new_cpu;
-
-	cpu = new_cpu = smp_processor_id();
-	do {
-		new_cpu = cpumask_next(new_cpu, cpu_online_mask) % nr_cpu_ids;
-		if (cpu_to_node(new_cpu) == numa_node_id())
-			break;
-	} while (cpu != new_cpu);
-
-	queue_work_on(new_cpu, wq, work);
-}
-
 /* This is called directly from KGD at ISR. */
 void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry)
 {
-	uint32_t patched_ihre[KFD_MAX_RING_ENTRY_SIZE];
+	uint32_t patched_ihre[KFD_MAX_RING_ENTRY_SIZE], i;
 	bool is_patched = false;
 	unsigned long flags;
+	struct kfd_node *node;
 
 	if (!kfd->init_complete)
 		return;
@@ -851,16 +1133,30 @@ void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry)
 		return;
 	}
 
-	spin_lock_irqsave(&kfd->interrupt_lock, flags);
-
-	if (kfd->interrupts_active
-	    && interrupt_is_wanted(kfd, ih_ring_entry,
-				   patched_ihre, &is_patched)
-	    && enqueue_ih_ring_entry(kfd,
-				     is_patched ? patched_ihre : ih_ring_entry))
-		kfd_queue_work(kfd->ih_wq, &kfd->interrupt_work);
+	for (i = 0; i < kfd->num_nodes; i++) {
+		/* Race if another thread in b/w
+		 * kfd_cleanup_nodes and kfree(kfd),
+		 * when kfd->nodes[i] = NULL
+		 */
+		if (kfd->nodes[i])
+			node = kfd->nodes[i];
+		else
+			return;
+
+		spin_lock_irqsave(&node->interrupt_lock, flags);
+
+		if (node->interrupts_active
+		    && interrupt_is_wanted(node, ih_ring_entry,
+			    	patched_ihre, &is_patched)
+		    && enqueue_ih_ring_entry(node,
+			    	is_patched ? patched_ihre : ih_ring_entry)) {
+			queue_work(node->kfd->ih_wq, &node->interrupt_work);
+			spin_unlock_irqrestore(&node->interrupt_lock, flags);
+			return;
+		}
+		spin_unlock_irqrestore(&node->interrupt_lock, flags);
+	}
 
-	spin_unlock_irqrestore(&kfd->interrupt_lock, flags);
 }
 
 int kgd2kfd_quiesce_mm(struct mm_struct *mm, uint32_t trigger)
@@ -998,10 +1294,11 @@ static inline uint32_t *kfd_gtt_sa_calc_cpu_addr(void *start_addr,
 	return (uint32_t *) ((uint64_t) start_addr + bit_num * chunk_size);
 }
 
-int kfd_gtt_sa_allocate(struct kfd_dev *kfd, unsigned int size,
+int kfd_gtt_sa_allocate(struct kfd_node *node, unsigned int size,
 			struct kfd_mem_obj **mem_obj)
 {
 	unsigned int found, start_search, cur_size;
+	struct kfd_dev *kfd = node->kfd;
 
 	if (size == 0)
 		return -EINVAL;
@@ -1101,8 +1398,10 @@ kfd_gtt_no_free_chunk:
 	return -ENOMEM;
 }
 
-int kfd_gtt_sa_free(struct kfd_dev *kfd, struct kfd_mem_obj *mem_obj)
+int kfd_gtt_sa_free(struct kfd_node *node, struct kfd_mem_obj *mem_obj)
 {
+	struct kfd_dev *kfd = node->kfd;
+
 	/* Act like kfree when trying to free a NULL object */
 	if (!mem_obj)
 		return 0;
@@ -1124,29 +1423,47 @@ int kfd_gtt_sa_free(struct kfd_dev *kfd, struct kfd_mem_obj *mem_obj)
 
 void kgd2kfd_set_sram_ecc_flag(struct kfd_dev *kfd)
 {
+	/*
+	 * TODO: Currently update SRAM ECC flag for first node.
+	 * This needs to be updated later when we can
+	 * identify SRAM ECC error on other nodes also.
+	 */
 	if (kfd)
-		atomic_inc(&kfd->sram_ecc_flag);
+		atomic_inc(&kfd->nodes[0]->sram_ecc_flag);
 }
 
-void kfd_inc_compute_active(struct kfd_dev *kfd)
+void kfd_inc_compute_active(struct kfd_node *node)
 {
-	if (atomic_inc_return(&kfd->compute_profile) == 1)
-		amdgpu_amdkfd_set_compute_idle(kfd->adev, false);
+	if (atomic_inc_return(&node->kfd->compute_profile) == 1)
+		amdgpu_amdkfd_set_compute_idle(node->adev, false);
 }
 
-void kfd_dec_compute_active(struct kfd_dev *kfd)
+void kfd_dec_compute_active(struct kfd_node *node)
 {
-	int count = atomic_dec_return(&kfd->compute_profile);
+	int count = atomic_dec_return(&node->kfd->compute_profile);
 
 	if (count == 0)
-		amdgpu_amdkfd_set_compute_idle(kfd->adev, true);
+		amdgpu_amdkfd_set_compute_idle(node->adev, true);
 	WARN_ONCE(count < 0, "Compute profile ref. count error");
 }
 
+static bool kfd_compute_active(struct kfd_node *node)
+{
+	if (atomic_read(&node->kfd->compute_profile))
+		return true;
+	return false;
+}
+
 void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint64_t throttle_bitmask)
 {
+	/*
+	 * TODO: For now, raise the throttling event only on first node.
+	 * This will need to change after we are able to determine
+	 * which node raised the throttling event.
+	 */
 	if (kfd && kfd->init_complete)
-		kfd_smi_event_update_thermal_throttling(kfd, throttle_bitmask);
+		kfd_smi_event_update_thermal_throttling(kfd->nodes[0],
+							throttle_bitmask);
 }
 
 /* kfd_get_num_sdma_engines returns the number of PCIe optimized SDMA and
@@ -1154,19 +1471,239 @@ void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint64_t throttle_bitmask)
  * When the device has more than two engines, we reserve two for PCIe to enable
  * full-duplex and the rest are used as XGMI.
  */
-unsigned int kfd_get_num_sdma_engines(struct kfd_dev *kdev)
+unsigned int kfd_get_num_sdma_engines(struct kfd_node *node)
 {
 	/* If XGMI is not supported, all SDMA engines are PCIe */
-	if (!kdev->adev->gmc.xgmi.supported)
-		return kdev->adev->sdma.num_instances;
+	if (!node->adev->gmc.xgmi.supported)
+		return node->adev->sdma.num_instances/(int)node->kfd->num_nodes;
 
-	return min(kdev->adev->sdma.num_instances, 2);
+	return min(node->adev->sdma.num_instances/(int)node->kfd->num_nodes, 2);
 }
 
-unsigned int kfd_get_num_xgmi_sdma_engines(struct kfd_dev *kdev)
+unsigned int kfd_get_num_xgmi_sdma_engines(struct kfd_node *node)
 {
 	/* After reserved for PCIe, the rest of engines are XGMI */
-	return kdev->adev->sdma.num_instances - kfd_get_num_sdma_engines(kdev);
+	return node->adev->sdma.num_instances/(int)node->kfd->num_nodes -
+		kfd_get_num_sdma_engines(node);
+}
+
+int kgd2kfd_check_and_lock_kfd(struct kfd_dev *kfd)
+{
+	struct kfd_process *p;
+	int r = 0, temp, idx;
+
+	mutex_lock(&kfd_processes_mutex);
+
+	/* kfd_processes_count is per kfd_dev, return -EBUSY without
+	 * further check
+	 */
+	if (!!atomic_read(&kfd->kfd_processes_count)) {
+		pr_debug("process_wq_release not finished\n");
+		r = -EBUSY;
+		goto out;
+	}
+
+	if (hash_empty(kfd_processes_table) && !kfd_is_locked(kfd))
+		goto out;
+
+	/* fail under system reset/resume or kfd device is partition switching. */
+	if (kfd_is_locked(kfd)) {
+		r = -EBUSY;
+		goto out;
+	}
+
+	/*
+	 * ensure all running processes are cgroup excluded from device before mode switch.
+	 * i.e. no pdd was created on the process socket.
+	 */
+	idx = srcu_read_lock(&kfd_processes_srcu);
+	hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) {
+		int i;
+
+		for (i = 0; i < p->n_pdds; i++) {
+			if (p->pdds[i]->dev->kfd != kfd)
+				continue;
+
+			r = -EBUSY;
+			goto proc_check_unlock;
+		}
+	}
+
+proc_check_unlock:
+	srcu_read_unlock(&kfd_processes_srcu, idx);
+out:
+	if (!r)
+		++kfd->kfd_dev_lock;
+	mutex_unlock(&kfd_processes_mutex);
+
+	return r;
+}
+
+void kgd2kfd_unlock_kfd(struct kfd_dev *kfd)
+{
+	mutex_lock(&kfd_processes_mutex);
+	--kfd->kfd_dev_lock;
+	mutex_unlock(&kfd_processes_mutex);
+}
+
+int kgd2kfd_start_sched(struct kfd_dev *kfd, uint32_t node_id)
+{
+	struct kfd_node *node;
+	int ret;
+
+	if (!kfd->init_complete)
+		return 0;
+
+	if (node_id >= kfd->num_nodes) {
+		dev_warn(kfd->adev->dev, "Invalid node ID: %u exceeds %u\n",
+			 node_id, kfd->num_nodes - 1);
+		return -EINVAL;
+	}
+	node = kfd->nodes[node_id];
+
+	ret = node->dqm->ops.unhalt(node->dqm);
+	if (ret)
+		dev_err(kfd_device, "Error in starting scheduler\n");
+
+	return ret;
+}
+
+int kgd2kfd_start_sched_all_nodes(struct kfd_dev *kfd)
+{
+	struct kfd_node *node;
+	int i, r;
+
+	if (!kfd->init_complete)
+		return 0;
+
+	for (i = 0; i < kfd->num_nodes; i++) {
+		node = kfd->nodes[i];
+		r = node->dqm->ops.unhalt(node->dqm);
+		if (r) {
+			dev_err(kfd_device, "Error in starting scheduler\n");
+			return r;
+		}
+	}
+	return 0;
+}
+
+int kgd2kfd_stop_sched(struct kfd_dev *kfd, uint32_t node_id)
+{
+	struct kfd_node *node;
+
+	if (!kfd->init_complete)
+		return 0;
+
+	if (node_id >= kfd->num_nodes) {
+		dev_warn(kfd->adev->dev, "Invalid node ID: %u exceeds %u\n",
+			 node_id, kfd->num_nodes - 1);
+		return -EINVAL;
+	}
+
+	node = kfd->nodes[node_id];
+	return node->dqm->ops.halt(node->dqm);
+}
+
+int kgd2kfd_stop_sched_all_nodes(struct kfd_dev *kfd)
+{
+	struct kfd_node *node;
+	int i, r;
+
+	if (!kfd->init_complete)
+		return 0;
+
+	for (i = 0; i < kfd->num_nodes; i++) {
+		node = kfd->nodes[i];
+		r = node->dqm->ops.halt(node->dqm);
+		if (r)
+			return r;
+	}
+	return 0;
+}
+
+bool kgd2kfd_compute_active(struct kfd_dev *kfd, uint32_t node_id)
+{
+	struct kfd_node *node;
+
+	if (!kfd->init_complete)
+		return false;
+
+	if (node_id >= kfd->num_nodes) {
+		dev_warn(kfd->adev->dev, "Invalid node ID: %u exceeds %u\n",
+			 node_id, kfd->num_nodes - 1);
+		return false;
+	}
+
+	node = kfd->nodes[node_id];
+
+	return kfd_compute_active(node);
+}
+
+/**
+ * kgd2kfd_vmfault_fast_path() - KFD vm page fault interrupt handling fast path for gmc v9
+ * @adev: amdgpu device
+ * @entry: vm fault interrupt vector
+ * @retry_fault: if this is retry fault
+ *
+ * retry fault -
+ *    with CAM enabled, adev primary ring
+ *                           |  gmc_v9_0_process_interrupt()
+ *                      adev soft_ring
+ *                           |  gmc_v9_0_process_interrupt() worker failed to recover page fault
+ *                      KFD node ih_fifo
+ *                           |  KFD interrupt_wq worker
+ *                      kfd_signal_vm_fault_event
+ *
+ *    without CAM,      adev primary ring1
+ *                           |  gmc_v9_0_process_interrupt worker failed to recvoer page fault
+ *                      KFD node ih_fifo
+ *                           |  KFD interrupt_wq worker
+ *                      kfd_signal_vm_fault_event
+ *
+ * no-retry fault -
+ *                      adev primary ring
+ *                           |  gmc_v9_0_process_interrupt()
+ *                      KFD node ih_fifo
+ *                           |  KFD interrupt_wq worker
+ *                      kfd_signal_vm_fault_event
+ *
+ * fast path - After kfd_signal_vm_fault_event, gmc_v9_0_process_interrupt drop the page fault
+ *            of same process, don't copy interrupt to KFD node ih_fifo.
+ *            With gdb debugger enabled, need convert the retry fault to no-retry fault for
+ *            debugger, cannot use the fast path.
+ *
+ * Return:
+ *   true - use the fast path to handle this fault
+ *   false - use normal path to handle it
+ */
+bool kgd2kfd_vmfault_fast_path(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry,
+			       bool retry_fault)
+{
+	struct kfd_process *p;
+	u32 cam_index;
+
+	if (entry->ih == &adev->irq.ih_soft || entry->ih == &adev->irq.ih1) {
+		p = kfd_lookup_process_by_pasid(entry->pasid, NULL);
+		if (!p)
+			return true;
+
+		if (p->gpu_page_fault && !p->debug_trap_enabled) {
+			if (retry_fault && adev->irq.retry_cam_enabled) {
+				cam_index = entry->src_data[2] & 0x3ff;
+				WDOORBELL32(adev->irq.retry_cam_doorbell_index, cam_index);
+			}
+
+			kfd_unref_process(p);
+			return true;
+		}
+
+		/*
+		 * This is the first page fault, set flag and then signal user space
+		 */
+		p->gpu_page_fault = true;
+		kfd_unref_process(p);
+	}
+	return false;
 }
 
 #if defined(CONFIG_DEBUG_FS)
@@ -1174,13 +1711,18 @@ unsigned int kfd_get_num_xgmi_sdma_engines(struct kfd_dev *kdev)
 /* This function will send a package to HIQ to hang the HWS
  * which will trigger a GPU reset and bring the HWS back to normal state
  */
-int kfd_debugfs_hang_hws(struct kfd_dev *dev)
+int kfd_debugfs_hang_hws(struct kfd_node *dev)
 {
 	if (dev->dqm->sched_policy != KFD_SCHED_POLICY_HWS) {
 		pr_err("HWS is not enabled");
 		return -EINVAL;
 	}
 
+	if (dev->kfd->shared_resources.enable_mes) {
+		dev_err(dev->adev->dev, "Inducing MES hang is not supported\n");
+		return -EINVAL;
+	}
+
 	return dqm_debugfs_hang_hws(dev->dqm);
 }
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
index 7a95698d83f7..d7a2e7178ea9 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
@@ -35,21 +35,29 @@
 #include "cik_regs.h"
 #include "kfd_kernel_queue.h"
 #include "amdgpu_amdkfd.h"
-#include "mes_api_def.h"
+#include "amdgpu_reset.h"
+#include "amdgpu_sdma.h"
+#include "mes_v11_api_def.h"
+#include "kfd_debug.h"
 
 /* Size of the per-pipe EOP queue */
 #define CIK_HPD_EOP_BYTES_LOG2 11
 #define CIK_HPD_EOP_BYTES (1U << CIK_HPD_EOP_BYTES_LOG2)
+/* See unmap_queues_cpsch() */
+#define USE_DEFAULT_GRACE_PERIOD 0xffffffff
 
 static int set_pasid_vmid_mapping(struct device_queue_manager *dqm,
 				  u32 pasid, unsigned int vmid);
 
 static int execute_queues_cpsch(struct device_queue_manager *dqm,
 				enum kfd_unmap_queues_filter filter,
-				uint32_t filter_param);
+				uint32_t filter_param,
+				uint32_t grace_period);
 static int unmap_queues_cpsch(struct device_queue_manager *dqm,
 				enum kfd_unmap_queues_filter filter,
-				uint32_t filter_param, bool reset);
+				uint32_t filter_param,
+				uint32_t grace_period,
+				bool reset);
 
 static int map_queues_cpsch(struct device_queue_manager *dqm);
 
@@ -61,7 +69,8 @@ static inline void deallocate_hqd(struct device_queue_manager *dqm,
 static int allocate_hqd(struct device_queue_manager *dqm, struct queue *q);
 static int allocate_sdma_queue(struct device_queue_manager *dqm,
 				struct queue *q, const uint32_t *restore_sdma_id);
-static void kfd_process_hw_exception(struct work_struct *work);
+
+static int reset_queues_on_hws_hang(struct device_queue_manager *dqm, bool is_sdma);
 
 static inline
 enum KFD_MQD_TYPE get_mqd_type_from_queue_type(enum kfd_queue_type type)
@@ -74,31 +83,31 @@ enum KFD_MQD_TYPE get_mqd_type_from_queue_type(enum kfd_queue_type type)
 static bool is_pipe_enabled(struct device_queue_manager *dqm, int mec, int pipe)
 {
 	int i;
-	int pipe_offset = (mec * dqm->dev->shared_resources.num_pipe_per_mec
-		+ pipe) * dqm->dev->shared_resources.num_queue_per_pipe;
+	int pipe_offset = (mec * dqm->dev->kfd->shared_resources.num_pipe_per_mec
+		+ pipe) * dqm->dev->kfd->shared_resources.num_queue_per_pipe;
 
 	/* queue is available for KFD usage if bit is 1 */
-	for (i = 0; i <  dqm->dev->shared_resources.num_queue_per_pipe; ++i)
+	for (i = 0; i <  dqm->dev->kfd->shared_resources.num_queue_per_pipe; ++i)
 		if (test_bit(pipe_offset + i,
-			      dqm->dev->shared_resources.cp_queue_bitmap))
+			      dqm->dev->kfd->shared_resources.cp_queue_bitmap))
 			return true;
 	return false;
 }
 
 unsigned int get_cp_queues_num(struct device_queue_manager *dqm)
 {
-	return bitmap_weight(dqm->dev->shared_resources.cp_queue_bitmap,
-				KGD_MAX_QUEUES);
+	return bitmap_weight(dqm->dev->kfd->shared_resources.cp_queue_bitmap,
+				AMDGPU_MAX_QUEUES);
 }
 
 unsigned int get_queues_per_pipe(struct device_queue_manager *dqm)
 {
-	return dqm->dev->shared_resources.num_queue_per_pipe;
+	return dqm->dev->kfd->shared_resources.num_queue_per_pipe;
 }
 
 unsigned int get_pipes_per_mec(struct device_queue_manager *dqm)
 {
-	return dqm->dev->shared_resources.num_pipe_per_mec;
+	return dqm->dev->kfd->shared_resources.num_pipe_per_mec;
 }
 
 static unsigned int get_num_all_sdma_engines(struct device_queue_manager *dqm)
@@ -110,44 +119,62 @@ static unsigned int get_num_all_sdma_engines(struct device_queue_manager *dqm)
 unsigned int get_num_sdma_queues(struct device_queue_manager *dqm)
 {
 	return kfd_get_num_sdma_engines(dqm->dev) *
-		dqm->dev->device_info.num_sdma_queues_per_engine;
+		dqm->dev->kfd->device_info.num_sdma_queues_per_engine;
 }
 
 unsigned int get_num_xgmi_sdma_queues(struct device_queue_manager *dqm)
 {
 	return kfd_get_num_xgmi_sdma_engines(dqm->dev) *
-		dqm->dev->device_info.num_sdma_queues_per_engine;
+		dqm->dev->kfd->device_info.num_sdma_queues_per_engine;
 }
 
-static inline uint64_t get_reserved_sdma_queues_bitmap(struct device_queue_manager *dqm)
+static void init_sdma_bitmaps(struct device_queue_manager *dqm)
 {
-	return dqm->dev->device_info.reserved_sdma_queues_bitmap;
+	bitmap_zero(dqm->sdma_bitmap, KFD_MAX_SDMA_QUEUES);
+	bitmap_set(dqm->sdma_bitmap, 0, get_num_sdma_queues(dqm));
+
+	bitmap_zero(dqm->xgmi_sdma_bitmap, KFD_MAX_SDMA_QUEUES);
+	bitmap_set(dqm->xgmi_sdma_bitmap, 0, get_num_xgmi_sdma_queues(dqm));
+
+	/* Mask out the reserved queues */
+	bitmap_andnot(dqm->sdma_bitmap, dqm->sdma_bitmap,
+		      dqm->dev->kfd->device_info.reserved_sdma_queues_bitmap,
+		      KFD_MAX_SDMA_QUEUES);
 }
 
 void program_sh_mem_settings(struct device_queue_manager *dqm,
 					struct qcm_process_device *qpd)
 {
-	return dqm->dev->kfd2kgd->program_sh_mem_settings(
-						dqm->dev->adev, qpd->vmid,
-						qpd->sh_mem_config,
-						qpd->sh_mem_ape1_base,
-						qpd->sh_mem_ape1_limit,
-						qpd->sh_mem_bases);
+	uint32_t xcc_mask = dqm->dev->xcc_mask;
+	int xcc_id;
+
+	for_each_inst(xcc_id, xcc_mask)
+		dqm->dev->kfd2kgd->program_sh_mem_settings(
+			dqm->dev->adev, qpd->vmid, qpd->sh_mem_config,
+			qpd->sh_mem_ape1_base, qpd->sh_mem_ape1_limit,
+			qpd->sh_mem_bases, xcc_id);
 }
 
 static void kfd_hws_hang(struct device_queue_manager *dqm)
 {
+	struct device_process_node *cur;
+	struct qcm_process_device *qpd;
+	struct queue *q;
+
+	/* Mark all device queues as reset. */
+	list_for_each_entry(cur, &dqm->queues, list) {
+		qpd = cur->qpd;
+		list_for_each_entry(q, &qpd->queues_list, list) {
+			struct kfd_process_device *pdd = qpd_to_pdd(qpd);
+
+			pdd->has_reset_queue = true;
+		}
+	}
+
 	/*
 	 * Issue a GPU reset if HWS is unresponsive
 	 */
-	dqm->is_hws_hang = true;
-
-	/* It's possible we're detecting a HWS hang in the
-	 * middle of a GPU reset. No need to schedule another
-	 * reset in this case.
-	 */
-	if (!dqm->is_resetting)
-		schedule_work(&dqm->hw_exception_work);
+	amdgpu_amdkfd_gpu_reset(dqm->dev->adev);
 }
 
 static int convert_to_mes_queue_type(int queue_type)
@@ -179,11 +206,13 @@ static int add_queue_mes(struct device_queue_manager *dqm, struct queue *q,
 	int r, queue_type;
 	uint64_t wptr_addr_off;
 
-	if (dqm->is_hws_hang)
+	if (!dqm->sched_running || dqm->sched_halt)
+		return 0;
+	if (!down_read_trylock(&adev->reset_domain->sem))
 		return -EIO;
 
 	memset(&queue_input, 0x0, sizeof(struct mes_add_queue_input));
-	queue_input.process_id = qpd->pqm->process->pasid;
+	queue_input.process_id = pdd->pasid;
 	queue_input.page_table_base_addr =  qpd->page_table_base;
 	queue_input.process_va_start = 0;
 	queue_input.process_va_end = adev->vm_manager.max_pfn - 1;
@@ -199,10 +228,8 @@ static int add_queue_mes(struct device_queue_manager *dqm, struct queue *q,
 	queue_input.mqd_addr = q->gart_mqd_addr;
 	queue_input.wptr_addr = (uint64_t)q->properties.write_ptr;
 
-	if (q->wptr_bo) {
-		wptr_addr_off = (uint64_t)q->properties.write_ptr & (PAGE_SIZE - 1);
-		queue_input.wptr_mc_addr = ((uint64_t)q->wptr_bo->tbo.resource->start << PAGE_SHIFT) + wptr_addr_off;
-	}
+	wptr_addr_off = (uint64_t)q->properties.write_ptr & (PAGE_SIZE - 1);
+	queue_input.wptr_mc_addr = amdgpu_bo_gpu_offset(q->properties.wptr_bo) + wptr_addr_off;
 
 	queue_input.is_kfd_process = 1;
 	queue_input.is_aql_queue = (q->properties.format == KFD_QUEUE_FORMAT_AQL);
@@ -211,29 +238,33 @@ static int add_queue_mes(struct device_queue_manager *dqm, struct queue *q,
 	queue_input.paging = false;
 	queue_input.tba_addr = qpd->tba_addr;
 	queue_input.tma_addr = qpd->tma_addr;
+	queue_input.trap_en = !kfd_dbg_has_cwsr_workaround(q->device);
+	queue_input.skip_process_ctx_clear =
+		qpd->pqm->process->runtime_info.runtime_state == DEBUG_RUNTIME_STATE_ENABLED &&
+						(qpd->pqm->process->debug_trap_enabled ||
+						 kfd_dbg_has_ttmps_always_setup(q->device));
 
 	queue_type = convert_to_mes_queue_type(q->properties.type);
 	if (queue_type < 0) {
-		pr_err("Queue type not supported with MES, queue:%d\n",
-				q->properties.type);
+		dev_err(adev->dev, "Queue type not supported with MES, queue:%d\n",
+			q->properties.type);
+		up_read(&adev->reset_domain->sem);
 		return -EINVAL;
 	}
 	queue_input.queue_type = (uint32_t)queue_type;
 
-	if (q->gws) {
-		queue_input.gws_base = 0;
-		queue_input.gws_size = qpd->num_gws;
-	}
+	queue_input.exclusively_scheduled = q->properties.is_gws;
 
 	amdgpu_mes_lock(&adev->mes);
 	r = adev->mes.funcs->add_hw_queue(&adev->mes, &queue_input);
 	amdgpu_mes_unlock(&adev->mes);
+	up_read(&adev->reset_domain->sem);
 	if (r) {
-		pr_err("failed to add hardware queue to MES, doorbell=0x%x\n",
+		dev_err(adev->dev, "failed to add hardware queue to MES, doorbell=0x%x\n",
 			q->properties.doorbell_off);
-		pr_err("MES might be in unrecoverable state, issue a GPU reset\n");
+		dev_err(adev->dev, "MES might be in unrecoverable state, issue a GPU reset\n");
 		kfd_hws_hang(dqm);
-}
+	}
 
 	return r;
 }
@@ -245,7 +276,9 @@ static int remove_queue_mes(struct device_queue_manager *dqm, struct queue *q,
 	int r;
 	struct mes_remove_queue_input queue_input;
 
-	if (dqm->is_hws_hang)
+	if (!dqm->sched_running || dqm->sched_halt)
+		return 0;
+	if (!down_read_trylock(&adev->reset_domain->sem))
 		return -EIO;
 
 	memset(&queue_input, 0x0, sizeof(struct mes_remove_queue_input));
@@ -255,20 +288,22 @@ static int remove_queue_mes(struct device_queue_manager *dqm, struct queue *q,
 	amdgpu_mes_lock(&adev->mes);
 	r = adev->mes.funcs->remove_hw_queue(&adev->mes, &queue_input);
 	amdgpu_mes_unlock(&adev->mes);
+	up_read(&adev->reset_domain->sem);
 
 	if (r) {
-		pr_err("failed to remove hardware queue from MES, doorbell=0x%x\n",
+		dev_err(adev->dev, "failed to remove hardware queue from MES, doorbell=0x%x\n",
 			q->properties.doorbell_off);
-		pr_err("MES might be in unrecoverable state, issue a GPU reset\n");
+		dev_err(adev->dev, "MES might be in unrecoverable state, issue a GPU reset\n");
 		kfd_hws_hang(dqm);
 	}
 
 	return r;
 }
 
-static int remove_all_queues_mes(struct device_queue_manager *dqm)
+static int remove_all_kfd_queues_mes(struct device_queue_manager *dqm)
 {
 	struct device_process_node *cur;
+	struct device *dev = dqm->dev->adev->dev;
 	struct qcm_process_device *qpd;
 	struct queue *q;
 	int retval = 0;
@@ -279,7 +314,7 @@ static int remove_all_queues_mes(struct device_queue_manager *dqm)
 			if (q->properties.is_active) {
 				retval = remove_queue_mes(dqm, q, qpd);
 				if (retval) {
-					pr_err("%s: Failed to remove queue %d for dev %d",
+					dev_err(dev, "%s: Failed to remove queue %d for dev %d",
 						__func__,
 						q->properties.queue_id,
 						dqm->dev->id);
@@ -292,6 +327,73 @@ static int remove_all_queues_mes(struct device_queue_manager *dqm)
 	return retval;
 }
 
+static int add_all_kfd_queues_mes(struct device_queue_manager *dqm)
+{
+	struct device_process_node *cur;
+	struct device *dev = dqm->dev->adev->dev;
+	struct qcm_process_device *qpd;
+	struct queue *q;
+	int retval = 0;
+
+	list_for_each_entry(cur, &dqm->queues, list) {
+		qpd = cur->qpd;
+		list_for_each_entry(q, &qpd->queues_list, list) {
+			if (!q->properties.is_active)
+				continue;
+			retval = add_queue_mes(dqm, q, qpd);
+			if (retval) {
+				dev_err(dev, "%s: Failed to add queue %d for dev %d",
+					__func__,
+					q->properties.queue_id,
+					dqm->dev->id);
+				return retval;
+			}
+		}
+	}
+
+	return retval;
+}
+
+static int suspend_all_queues_mes(struct device_queue_manager *dqm)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)dqm->dev->adev;
+	int r = 0;
+
+	if (!down_read_trylock(&adev->reset_domain->sem))
+		return -EIO;
+
+	r = amdgpu_mes_suspend(adev);
+	up_read(&adev->reset_domain->sem);
+
+	if (r) {
+		dev_err(adev->dev, "failed to suspend gangs from MES\n");
+		dev_err(adev->dev, "MES might be in unrecoverable state, issue a GPU reset\n");
+		kfd_hws_hang(dqm);
+	}
+
+	return r;
+}
+
+static int resume_all_queues_mes(struct device_queue_manager *dqm)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)dqm->dev->adev;
+	int r = 0;
+
+	if (!down_read_trylock(&adev->reset_domain->sem))
+		return -EIO;
+
+	r = amdgpu_mes_resume(adev);
+	up_read(&adev->reset_domain->sem);
+
+	if (r) {
+		dev_err(adev->dev, "failed to resume gangs from MES\n");
+		dev_err(adev->dev, "MES might be in unrecoverable state, issue a GPU reset\n");
+		kfd_hws_hang(dqm);
+	}
+
+	return r;
+}
+
 static void increment_queue_count(struct device_queue_manager *dqm,
 				  struct qcm_process_device *qpd,
 				  struct queue *q)
@@ -330,7 +432,7 @@ static int allocate_doorbell(struct qcm_process_device *qpd,
 			     struct queue *q,
 			     uint32_t const *restore_id)
 {
-	struct kfd_dev *dev = qpd->dqm->dev;
+	struct kfd_node *dev = qpd->dqm->dev;
 
 	if (!KFD_IS_SOC15(dev)) {
 		/* On pre-SOC15 chips we need to use the queue ID to
@@ -349,8 +451,17 @@ static int allocate_doorbell(struct qcm_process_device *qpd,
 		 * for a SDMA engine is 512.
 		 */
 
-		uint32_t *idx_offset = dev->shared_resources.sdma_doorbell_idx;
-		uint32_t valid_id = idx_offset[q->properties.sdma_engine_id]
+		uint32_t *idx_offset = dev->kfd->shared_resources.sdma_doorbell_idx;
+
+		/*
+		 * q->properties.sdma_engine_id corresponds to the virtual
+		 * sdma engine number. However, for doorbell allocation,
+		 * we need the physical sdma engine id in order to get the
+		 * correct doorbell offset.
+		 */
+		uint32_t valid_id = idx_offset[qpd->dqm->dev->node_id *
+					       get_num_all_sdma_engines(qpd->dqm) +
+					       q->properties.sdma_engine_id]
 						+ (q->properties.sdma_queue_id & 1)
 						* KFD_QUEUE_DOORBELL_MIRROR_OFFSET
 						+ (q->properties.sdma_queue_id >> 1);
@@ -371,7 +482,7 @@ static int allocate_doorbell(struct qcm_process_device *qpd,
 			unsigned int found;
 
 			found = find_first_zero_bit(qpd->doorbell_bitmap,
-						KFD_MAX_NUM_OF_QUEUES_PER_PROCESS);
+						    KFD_MAX_NUM_OF_QUEUES_PER_PROCESS);
 			if (found >= KFD_MAX_NUM_OF_QUEUES_PER_PROCESS) {
 				pr_debug("No doorbells available");
 				return -EBUSY;
@@ -381,9 +492,10 @@ static int allocate_doorbell(struct qcm_process_device *qpd,
 		}
 	}
 
-	q->properties.doorbell_off =
-		kfd_get_doorbell_dw_offset_in_bar(dev, qpd_to_pdd(qpd),
-					  q->doorbell_id);
+	q->properties.doorbell_off = amdgpu_doorbell_index_on_bar(dev->adev,
+								  qpd->proc_doorbells,
+								  q->doorbell_id,
+								  dev->kfd->device_info.doorbell_size);
 	return 0;
 }
 
@@ -391,7 +503,7 @@ static void deallocate_doorbell(struct qcm_process_device *qpd,
 				struct queue *q)
 {
 	unsigned int old;
-	struct kfd_dev *dev = qpd->dqm->dev;
+	struct kfd_node *dev = qpd->dqm->dev;
 
 	if (!KFD_IS_SOC15(dev) ||
 	    q->properties.type == KFD_QUEUE_TYPE_SDMA ||
@@ -405,16 +517,22 @@ static void deallocate_doorbell(struct qcm_process_device *qpd,
 static void program_trap_handler_settings(struct device_queue_manager *dqm,
 				struct qcm_process_device *qpd)
 {
+	uint32_t xcc_mask = dqm->dev->xcc_mask;
+	int xcc_id;
+
 	if (dqm->dev->kfd2kgd->program_trap_handler_settings)
-		dqm->dev->kfd2kgd->program_trap_handler_settings(
-						dqm->dev->adev, qpd->vmid,
-						qpd->tba_addr, qpd->tma_addr);
+		for_each_inst(xcc_id, xcc_mask)
+			dqm->dev->kfd2kgd->program_trap_handler_settings(
+				dqm->dev->adev, qpd->vmid, qpd->tba_addr,
+				qpd->tma_addr, xcc_id);
 }
 
 static int allocate_vmid(struct device_queue_manager *dqm,
 			struct qcm_process_device *qpd,
 			struct queue *q)
 {
+	struct kfd_process_device *pdd = qpd_to_pdd(qpd);
+	struct device *dev = dqm->dev->adev->dev;
 	int allocated_vmid = -1, i;
 
 	for (i = dqm->dev->vm_info.first_vmid_kfd;
@@ -426,22 +544,22 @@ static int allocate_vmid(struct device_queue_manager *dqm,
 	}
 
 	if (allocated_vmid < 0) {
-		pr_err("no more vmid to allocate\n");
+		dev_err(dev, "no more vmid to allocate\n");
 		return -ENOSPC;
 	}
 
 	pr_debug("vmid allocated: %d\n", allocated_vmid);
 
-	dqm->vmid_pasid[allocated_vmid] = q->process->pasid;
+	dqm->vmid_pasid[allocated_vmid] = pdd->pasid;
 
-	set_pasid_vmid_mapping(dqm, q->process->pasid, allocated_vmid);
+	set_pasid_vmid_mapping(dqm, pdd->pasid, allocated_vmid);
 
 	qpd->vmid = allocated_vmid;
 	q->properties.vmid = allocated_vmid;
 
 	program_sh_mem_settings(dqm, qpd);
 
-	if (KFD_IS_SOC15(dqm->dev) && dqm->dev->cwsr_enabled)
+	if (KFD_IS_SOC15(dqm->dev) && dqm->dev->kfd->cwsr_enabled)
 		program_trap_handler_settings(dqm, qpd);
 
 	/* qpd->page_table_base is set earlier when register_process()
@@ -460,7 +578,7 @@ static int allocate_vmid(struct device_queue_manager *dqm,
 	return 0;
 }
 
-static int flush_texture_cache_nocpsch(struct kfd_dev *kdev,
+static int flush_texture_cache_nocpsch(struct kfd_node *kdev,
 				struct qcm_process_device *qpd)
 {
 	const struct packet_manager_funcs *pmf = qpd->dqm->packet_mgr.pmf;
@@ -482,10 +600,12 @@ static void deallocate_vmid(struct device_queue_manager *dqm,
 				struct qcm_process_device *qpd,
 				struct queue *q)
 {
+	struct device *dev = dqm->dev->adev->dev;
+
 	/* On GFX v7, CP doesn't flush TC at dequeue */
 	if (q->device->adev->asic_type == CHIP_HAWAII)
 		if (flush_texture_cache_nocpsch(q->device, qpd))
-			pr_err("Failed to flush TC\n");
+			dev_err(dev, "Failed to flush TC\n");
 
 	kfd_flush_tlb(qpd_to_pdd(qpd), TLB_FLUSH_LEGACY);
 
@@ -661,7 +781,7 @@ static inline void deallocate_hqd(struct device_queue_manager *dqm,
 #define SQ_IND_CMD_CMD_KILL		0x00000003
 #define SQ_IND_CMD_MODE_BROADCAST	0x00000001
 
-static int dbgdev_wave_reset_wavefronts(struct kfd_dev *dev, struct kfd_process *p)
+static int dbgdev_wave_reset_wavefronts(struct kfd_node *dev, struct kfd_process *p)
 {
 	int status = 0;
 	unsigned int vmid;
@@ -671,6 +791,8 @@ static int dbgdev_wave_reset_wavefronts(struct kfd_dev *dev, struct kfd_process
 	struct kfd_process_device *pdd;
 	int first_vmid_to_scan = dev->vm_info.first_vmid_kfd;
 	int last_vmid_to_scan = dev->vm_info.last_vmid_kfd;
+	uint32_t xcc_mask = dev->xcc_mask;
+	int xcc_id;
 
 	reg_sq_cmd.u32All = 0;
 	reg_gfx_index.u32All = 0;
@@ -678,10 +800,15 @@ static int dbgdev_wave_reset_wavefronts(struct kfd_dev *dev, struct kfd_process
 	pr_debug("Killing all process wavefronts\n");
 
 	if (!dev->kfd2kgd->get_atc_vmid_pasid_mapping_info) {
-		pr_err("no vmid pasid mapping supported \n");
+		dev_err(dev->adev->dev, "no vmid pasid mapping supported\n");
 		return -EOPNOTSUPP;
 	}
 
+	/* taking the VMID for that process on the safe way using PDD */
+	pdd = kfd_get_process_device_data(dev, p);
+	if (!pdd)
+		return -EFAULT;
+
 	/* Scan all registers in the range ATC_VMID8_PASID_MAPPING ..
 	 * ATC_VMID15_PASID_MAPPING
 	 * to check which VMID the current process is mapped to.
@@ -691,23 +818,19 @@ static int dbgdev_wave_reset_wavefronts(struct kfd_dev *dev, struct kfd_process
 		status = dev->kfd2kgd->get_atc_vmid_pasid_mapping_info
 				(dev->adev, vmid, &queried_pasid);
 
-		if (status && queried_pasid == p->pasid) {
-			pr_debug("Killing wave fronts of vmid %d and pasid 0x%x\n",
-					vmid, p->pasid);
+		if (status && queried_pasid == pdd->pasid) {
+			pr_debug("Killing wave fronts of vmid %d and process pid %d\n",
+					vmid, p->lead_thread->pid);
 			break;
 		}
 	}
 
 	if (vmid > last_vmid_to_scan) {
-		pr_err("Didn't find vmid for pasid 0x%x\n", p->pasid);
+		dev_err(dev->adev->dev, "Didn't find vmid for process pid %d\n",
+				p->lead_thread->pid);
 		return -EFAULT;
 	}
 
-	/* taking the VMID for that process on the safe way using PDD */
-	pdd = kfd_get_process_device_data(dev, p);
-	if (!pdd)
-		return -EFAULT;
-
 	reg_gfx_index.bits.sh_broadcast_writes = 1;
 	reg_gfx_index.bits.se_broadcast_writes = 1;
 	reg_gfx_index.bits.instance_broadcast_writes = 1;
@@ -715,9 +838,10 @@ static int dbgdev_wave_reset_wavefronts(struct kfd_dev *dev, struct kfd_process
 	reg_sq_cmd.bits.cmd = SQ_IND_CMD_CMD_KILL;
 	reg_sq_cmd.bits.vm_id = vmid;
 
-	dev->kfd2kgd->wave_control_execute(dev->adev,
-					reg_gfx_index.u32All,
-					reg_sq_cmd.u32All);
+	for_each_inst(xcc_id, xcc_mask)
+		dev->kfd2kgd->wave_control_execute(
+			dev->adev, reg_gfx_index.u32All,
+			reg_sq_cmd.u32All, xcc_id);
 
 	return 0;
 }
@@ -790,6 +914,7 @@ static int destroy_queue_nocpsch(struct device_queue_manager *dqm,
 {
 	int retval;
 	uint64_t sdma_val = 0;
+	struct device *dev = dqm->dev->adev->dev;
 	struct kfd_process_device *pdd = qpd_to_pdd(qpd);
 	struct mqd_manager *mqd_mgr =
 		dqm->mqd_mgrs[get_mqd_type_from_queue_type(q->properties.type)];
@@ -800,7 +925,7 @@ static int destroy_queue_nocpsch(struct device_queue_manager *dqm,
 		retval = read_sdma_queue_counter((uint64_t __user *)q->properties.read_ptr,
 							&sdma_val);
 		if (retval)
-			pr_err("Failed to read SDMA queue counter for queue: %d\n",
+			dev_err(dev, "Failed to read SDMA queue counter for queue: %d\n",
 				q->properties.queue_id);
 	}
 
@@ -819,6 +944,7 @@ static int update_queue(struct device_queue_manager *dqm, struct queue *q,
 			struct mqd_update_info *minfo)
 {
 	int retval = 0;
+	struct device *dev = dqm->dev->adev->dev;
 	struct mqd_manager *mqd_mgr;
 	struct kfd_process_device *pdd;
 	bool prev_active = false;
@@ -837,14 +963,20 @@ static int update_queue(struct device_queue_manager *dqm, struct queue *q,
 
 	/* Make sure the queue is unmapped before updating the MQD */
 	if (dqm->sched_policy != KFD_SCHED_POLICY_NO_HWS) {
-		if (!dqm->dev->shared_resources.enable_mes)
+		if (!dqm->dev->kfd->shared_resources.enable_mes)
 			retval = unmap_queues_cpsch(dqm,
-						    KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0, false);
+						    KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0, USE_DEFAULT_GRACE_PERIOD, false);
 		else if (prev_active)
 			retval = remove_queue_mes(dqm, q, &pdd->qpd);
 
+		/* queue is reset so inaccessable  */
+		if (pdd->has_reset_queue) {
+			retval = -EACCES;
+			goto out_unlock;
+		}
+
 		if (retval) {
-			pr_err("unmap queue failed\n");
+			dev_err(dev, "unmap queue failed\n");
 			goto out_unlock;
 		}
 	} else if (prev_active &&
@@ -858,12 +990,12 @@ static int update_queue(struct device_queue_manager *dqm, struct queue *q,
 		}
 
 		retval = mqd_mgr->destroy_mqd(mqd_mgr, q->mqd,
-				(dqm->dev->cwsr_enabled ?
+				(dqm->dev->kfd->cwsr_enabled ?
 				 KFD_PREEMPT_TYPE_WAVEFRONT_SAVE :
 				 KFD_PREEMPT_TYPE_WAVEFRONT_DRAIN),
 				KFD_UNMAP_LATENCY_MS, q->pipe, q->queue);
 		if (retval) {
-			pr_err("destroy mqd failed\n");
+			dev_err(dev, "destroy mqd failed\n");
 			goto out_unlock;
 		}
 	}
@@ -895,7 +1027,7 @@ static int update_queue(struct device_queue_manager *dqm, struct queue *q,
 	}
 
 	if (dqm->sched_policy != KFD_SCHED_POLICY_NO_HWS) {
-		if (!dqm->dev->shared_resources.enable_mes)
+		if (!dqm->dev->kfd->shared_resources.enable_mes)
 			retval = map_queues_cpsch(dqm);
 		else if (q->properties.is_active)
 			retval = add_queue_mes(dqm, q, &pdd->qpd);
@@ -917,6 +1049,92 @@ out_unlock:
 	return retval;
 }
 
+/* suspend_single_queue does not lock the dqm like the
+ * evict_process_queues_cpsch or evict_process_queues_nocpsch. You should
+ * lock the dqm before calling, and unlock after calling.
+ *
+ * The reason we don't lock the dqm is because this function may be
+ * called on multiple queues in a loop, so rather than locking/unlocking
+ * multiple times, we will just keep the dqm locked for all of the calls.
+ */
+static int suspend_single_queue(struct device_queue_manager *dqm,
+				      struct kfd_process_device *pdd,
+				      struct queue *q)
+{
+	bool is_new;
+
+	if (q->properties.is_suspended)
+		return 0;
+
+	pr_debug("Suspending process pid %d queue [%i]\n",
+			pdd->process->lead_thread->pid,
+			q->properties.queue_id);
+
+	is_new = q->properties.exception_status & KFD_EC_MASK(EC_QUEUE_NEW);
+
+	if (is_new || q->properties.is_being_destroyed) {
+		pr_debug("Suspend: skip %s queue id %i\n",
+				is_new ? "new" : "destroyed",
+				q->properties.queue_id);
+		return -EBUSY;
+	}
+
+	q->properties.is_suspended = true;
+	if (q->properties.is_active) {
+		if (dqm->dev->kfd->shared_resources.enable_mes) {
+			int r = remove_queue_mes(dqm, q, &pdd->qpd);
+
+			if (r)
+				return r;
+		}
+
+		decrement_queue_count(dqm, &pdd->qpd, q);
+		q->properties.is_active = false;
+	}
+
+	return 0;
+}
+
+/* resume_single_queue does not lock the dqm like the functions
+ * restore_process_queues_cpsch or restore_process_queues_nocpsch. You should
+ * lock the dqm before calling, and unlock after calling.
+ *
+ * The reason we don't lock the dqm is because this function may be
+ * called on multiple queues in a loop, so rather than locking/unlocking
+ * multiple times, we will just keep the dqm locked for all of the calls.
+ */
+static int resume_single_queue(struct device_queue_manager *dqm,
+				      struct qcm_process_device *qpd,
+				      struct queue *q)
+{
+	struct kfd_process_device *pdd;
+
+	if (!q->properties.is_suspended)
+		return 0;
+
+	pdd = qpd_to_pdd(qpd);
+
+	pr_debug("Restoring from suspend process pid %d queue [%i]\n",
+			    pdd->process->lead_thread->pid,
+			    q->properties.queue_id);
+
+	q->properties.is_suspended = false;
+
+	if (QUEUE_IS_ACTIVE(q->properties)) {
+		if (dqm->dev->kfd->shared_resources.enable_mes) {
+			int r = add_queue_mes(dqm, q, &pdd->qpd);
+
+			if (r)
+				return r;
+		}
+
+		q->properties.is_active = true;
+		increment_queue_count(dqm, qpd, q);
+	}
+
+	return 0;
+}
+
 static int evict_process_queues_nocpsch(struct device_queue_manager *dqm,
 					struct qcm_process_device *qpd)
 {
@@ -930,8 +1148,8 @@ static int evict_process_queues_nocpsch(struct device_queue_manager *dqm,
 		goto out;
 
 	pdd = qpd_to_pdd(qpd);
-	pr_debug_ratelimited("Evicting PASID 0x%x queues\n",
-			    pdd->process->pasid);
+	pr_debug_ratelimited("Evicting process pid %d queues\n",
+			    pdd->process->lead_thread->pid);
 
 	pdd->last_evict_timestamp = get_jiffies_64();
 	/* Mark all queues as evicted. Deactivate all active queues on
@@ -951,7 +1169,7 @@ static int evict_process_queues_nocpsch(struct device_queue_manager *dqm,
 			continue;
 
 		retval = mqd_mgr->destroy_mqd(mqd_mgr, q->mqd,
-				(dqm->dev->cwsr_enabled ?
+				(dqm->dev->kfd->cwsr_enabled ?
 				 KFD_PREEMPT_TYPE_WAVEFRONT_SAVE :
 				 KFD_PREEMPT_TYPE_WAVEFRONT_DRAIN),
 				KFD_UNMAP_LATENCY_MS, q->pipe, q->queue);
@@ -971,6 +1189,7 @@ static int evict_process_queues_cpsch(struct device_queue_manager *dqm,
 				      struct qcm_process_device *qpd)
 {
 	struct queue *q;
+	struct device *dev = dqm->dev->adev->dev;
 	struct kfd_process_device *pdd;
 	int retval = 0;
 
@@ -979,8 +1198,25 @@ static int evict_process_queues_cpsch(struct device_queue_manager *dqm,
 		goto out;
 
 	pdd = qpd_to_pdd(qpd);
-	pr_debug_ratelimited("Evicting PASID 0x%x queues\n",
-			    pdd->process->pasid);
+
+	/* The debugger creates processes that temporarily have not acquired
+	 * all VMs for all devices and has no VMs itself.
+	 * Skip queue eviction on process eviction.
+	 */
+	if (!pdd->drm_priv)
+		goto out;
+
+	pr_debug_ratelimited("Evicting process pid %d queues\n",
+			    pdd->process->lead_thread->pid);
+
+	if (dqm->dev->kfd->shared_resources.enable_mes) {
+		pdd->last_evict_timestamp = get_jiffies_64();
+		retval = suspend_all_queues_mes(dqm);
+		if (retval) {
+			dev_err(dev, "Suspending all queues failed");
+			goto out;
+		}
+	}
 
 	/* Mark all queues as evicted. Deactivate all active queues on
 	 * the qpd.
@@ -993,21 +1229,28 @@ static int evict_process_queues_cpsch(struct device_queue_manager *dqm,
 		q->properties.is_active = false;
 		decrement_queue_count(dqm, qpd, q);
 
-		if (dqm->dev->shared_resources.enable_mes) {
+		if (dqm->dev->kfd->shared_resources.enable_mes) {
 			retval = remove_queue_mes(dqm, q, qpd);
 			if (retval) {
-				pr_err("Failed to evict queue %d\n",
+				dev_err(dev, "Failed to evict queue %d\n",
 					q->properties.queue_id);
 				goto out;
 			}
 		}
 	}
-	pdd->last_evict_timestamp = get_jiffies_64();
-	if (!dqm->dev->shared_resources.enable_mes)
+
+	if (!dqm->dev->kfd->shared_resources.enable_mes) {
+		pdd->last_evict_timestamp = get_jiffies_64();
 		retval = execute_queues_cpsch(dqm,
 					      qpd->is_debug ?
 					      KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES :
-					      KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
+					      KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0,
+					      USE_DEFAULT_GRACE_PERIOD);
+	} else {
+		retval = resume_all_queues_mes(dqm);
+		if (retval)
+			dev_err(dev, "Resuming all queues failed");
+	}
 
 out:
 	dqm_unlock(dqm);
@@ -1037,8 +1280,8 @@ static int restore_process_queues_nocpsch(struct device_queue_manager *dqm,
 		goto out;
 	}
 
-	pr_debug_ratelimited("Restoring PASID 0x%x queues\n",
-			    pdd->process->pasid);
+	pr_debug_ratelimited("Restoring process pid %d queues\n",
+			    pdd->process->lead_thread->pid);
 
 	/* Update PD Base in QPD */
 	qpd->page_table_base = pd_base;
@@ -1099,14 +1342,12 @@ static int restore_process_queues_cpsch(struct device_queue_manager *dqm,
 					struct qcm_process_device *qpd)
 {
 	struct queue *q;
+	struct device *dev = dqm->dev->adev->dev;
 	struct kfd_process_device *pdd;
-	uint64_t pd_base;
 	uint64_t eviction_duration;
 	int retval = 0;
 
 	pdd = qpd_to_pdd(qpd);
-	/* Retrieve PD base */
-	pd_base = amdgpu_amdkfd_gpuvm_get_process_page_dir(pdd->drm_priv);
 
 	dqm_lock(dqm);
 	if (WARN_ON_ONCE(!qpd->evicted)) /* already restored, do nothing */
@@ -1116,12 +1357,19 @@ static int restore_process_queues_cpsch(struct device_queue_manager *dqm,
 		goto out;
 	}
 
-	pr_debug_ratelimited("Restoring PASID 0x%x queues\n",
-			    pdd->process->pasid);
+	/* The debugger creates processes that temporarily have not acquired
+	 * all VMs for all devices and has no VMs itself.
+	 * Skip queue restore on process restore.
+	 */
+	if (!pdd->drm_priv)
+		goto vm_not_acquired;
+
+	pr_debug_ratelimited("Restoring process pid %d queues\n",
+			    pdd->process->lead_thread->pid);
 
 	/* Update PD Base in QPD */
-	qpd->page_table_base = pd_base;
-	pr_debug("Updated PD address to 0x%llx\n", pd_base);
+	qpd->page_table_base = amdgpu_amdkfd_gpuvm_get_process_page_dir(pdd->drm_priv);
+	pr_debug("Updated PD address to 0x%llx\n", qpd->page_table_base);
 
 	/* activate all active queues on the qpd */
 	list_for_each_entry(q, &qpd->queues_list, list) {
@@ -1132,21 +1380,22 @@ static int restore_process_queues_cpsch(struct device_queue_manager *dqm,
 		q->properties.is_active = true;
 		increment_queue_count(dqm, &pdd->qpd, q);
 
-		if (dqm->dev->shared_resources.enable_mes) {
+		if (dqm->dev->kfd->shared_resources.enable_mes) {
 			retval = add_queue_mes(dqm, q, qpd);
 			if (retval) {
-				pr_err("Failed to restore queue %d\n",
+				dev_err(dev, "Failed to restore queue %d\n",
 					q->properties.queue_id);
 				goto out;
 			}
 		}
 	}
-	if (!dqm->dev->shared_resources.enable_mes)
+	if (!dqm->dev->kfd->shared_resources.enable_mes)
 		retval = execute_queues_cpsch(dqm,
-					      KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
-	qpd->evicted = 0;
+					      KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0, USE_DEFAULT_GRACE_PERIOD);
 	eviction_duration = get_jiffies_64() - pdd->last_evict_timestamp;
 	atomic64_add(eviction_duration, &pdd->evict_duration_counter);
+vm_not_acquired:
+	qpd->evicted = 0;
 out:
 	dqm_unlock(dqm);
 	return retval;
@@ -1229,35 +1478,32 @@ static int
 set_pasid_vmid_mapping(struct device_queue_manager *dqm, u32 pasid,
 			unsigned int vmid)
 {
-	return dqm->dev->kfd2kgd->set_pasid_vmid_mapping(
-						dqm->dev->adev, pasid, vmid);
-}
+	uint32_t xcc_mask = dqm->dev->xcc_mask;
+	int xcc_id, ret;
 
-static void init_interrupts(struct device_queue_manager *dqm)
-{
-	unsigned int i;
+	for_each_inst(xcc_id, xcc_mask) {
+		ret = dqm->dev->kfd2kgd->set_pasid_vmid_mapping(
+			dqm->dev->adev, pasid, vmid, xcc_id);
+		if (ret)
+			break;
+	}
 
-	for (i = 0 ; i < get_pipes_per_mec(dqm) ; i++)
-		if (is_pipe_enabled(dqm, 0, i))
-			dqm->dev->kfd2kgd->init_interrupts(dqm->dev->adev, i);
+	return ret;
 }
 
-static void init_sdma_bitmaps(struct device_queue_manager *dqm)
+static void init_interrupts(struct device_queue_manager *dqm)
 {
-	unsigned int num_sdma_queues =
-		min_t(unsigned int, sizeof(dqm->sdma_bitmap)*8,
-		      get_num_sdma_queues(dqm));
-	unsigned int num_xgmi_sdma_queues =
-		min_t(unsigned int, sizeof(dqm->xgmi_sdma_bitmap)*8,
-		      get_num_xgmi_sdma_queues(dqm));
+	uint32_t xcc_mask = dqm->dev->xcc_mask;
+	unsigned int i, xcc_id;
 
-	if (num_sdma_queues)
-		dqm->sdma_bitmap = GENMASK_ULL(num_sdma_queues-1, 0);
-	if (num_xgmi_sdma_queues)
-		dqm->xgmi_sdma_bitmap = GENMASK_ULL(num_xgmi_sdma_queues-1, 0);
-
-	dqm->sdma_bitmap &= ~get_reserved_sdma_queues_bitmap(dqm);
-	pr_info("sdma_bitmap: %llx\n", dqm->sdma_bitmap);
+	for_each_inst(xcc_id, xcc_mask) {
+		for (i = 0 ; i < get_pipes_per_mec(dqm) ; i++) {
+			if (is_pipe_enabled(dqm, 0, i)) {
+				dqm->dev->kfd2kgd->init_interrupts(
+					dqm->dev->adev, i, xcc_id);
+			}
+		}
+	}
 }
 
 static int initialize_nocpsch(struct device_queue_manager *dqm)
@@ -1282,7 +1528,7 @@ static int initialize_nocpsch(struct device_queue_manager *dqm)
 
 		for (queue = 0; queue < get_queues_per_pipe(dqm); queue++)
 			if (test_bit(pipe_offset + queue,
-				     dqm->dev->shared_resources.cp_queue_bitmap))
+				     dqm->dev->kfd->shared_resources.cp_queue_bitmap))
 				dqm->allocated_queues[pipe] |= 1 << queue;
 	}
 
@@ -1322,66 +1568,71 @@ static int start_nocpsch(struct device_queue_manager *dqm)
 
 static int stop_nocpsch(struct device_queue_manager *dqm)
 {
+	dqm_lock(dqm);
+	if (!dqm->sched_running) {
+		dqm_unlock(dqm);
+		return 0;
+	}
+
 	if (dqm->dev->adev->asic_type == CHIP_HAWAII)
-		pm_uninit(&dqm->packet_mgr, false);
+		pm_uninit(&dqm->packet_mgr);
 	dqm->sched_running = false;
+	dqm_unlock(dqm);
 
 	return 0;
 }
 
-static void pre_reset(struct device_queue_manager *dqm)
-{
-	dqm_lock(dqm);
-	dqm->is_resetting = true;
-	dqm_unlock(dqm);
-}
-
 static int allocate_sdma_queue(struct device_queue_manager *dqm,
 				struct queue *q, const uint32_t *restore_sdma_id)
 {
+	struct device *dev = dqm->dev->adev->dev;
 	int bit;
 
 	if (q->properties.type == KFD_QUEUE_TYPE_SDMA) {
-		if (dqm->sdma_bitmap == 0) {
-			pr_err("No more SDMA queue to allocate\n");
+		if (bitmap_empty(dqm->sdma_bitmap, get_num_sdma_queues(dqm))) {
+			dev_warn(dev, "No more SDMA queue to allocate (%d total queues)\n",
+				 get_num_sdma_queues(dqm));
 			return -ENOMEM;
 		}
 
 		if (restore_sdma_id) {
 			/* Re-use existing sdma_id */
-			if (!(dqm->sdma_bitmap & (1ULL << *restore_sdma_id))) {
-				pr_err("SDMA queue already in use\n");
+			if (!test_bit(*restore_sdma_id, dqm->sdma_bitmap)) {
+				dev_err(dev, "SDMA queue already in use\n");
 				return -EBUSY;
 			}
-			dqm->sdma_bitmap &= ~(1ULL << *restore_sdma_id);
+			clear_bit(*restore_sdma_id, dqm->sdma_bitmap);
 			q->sdma_id = *restore_sdma_id;
 		} else {
 			/* Find first available sdma_id */
-			bit = __ffs64(dqm->sdma_bitmap);
-			dqm->sdma_bitmap &= ~(1ULL << bit);
+			bit = find_first_bit(dqm->sdma_bitmap,
+					     get_num_sdma_queues(dqm));
+			clear_bit(bit, dqm->sdma_bitmap);
 			q->sdma_id = bit;
 		}
 
-		q->properties.sdma_engine_id = q->sdma_id %
-				kfd_get_num_sdma_engines(dqm->dev);
+		q->properties.sdma_engine_id =
+			q->sdma_id % kfd_get_num_sdma_engines(dqm->dev);
 		q->properties.sdma_queue_id = q->sdma_id /
 				kfd_get_num_sdma_engines(dqm->dev);
 	} else if (q->properties.type == KFD_QUEUE_TYPE_SDMA_XGMI) {
-		if (dqm->xgmi_sdma_bitmap == 0) {
-			pr_err("No more XGMI SDMA queue to allocate\n");
+		if (bitmap_empty(dqm->xgmi_sdma_bitmap, get_num_xgmi_sdma_queues(dqm))) {
+			dev_warn(dev, "No more XGMI SDMA queue to allocate (%d total queues)\n",
+				 get_num_xgmi_sdma_queues(dqm));
 			return -ENOMEM;
 		}
 		if (restore_sdma_id) {
 			/* Re-use existing sdma_id */
-			if (!(dqm->xgmi_sdma_bitmap & (1ULL << *restore_sdma_id))) {
-				pr_err("SDMA queue already in use\n");
+			if (!test_bit(*restore_sdma_id, dqm->xgmi_sdma_bitmap)) {
+				dev_err(dev, "SDMA queue already in use\n");
 				return -EBUSY;
 			}
-			dqm->xgmi_sdma_bitmap &= ~(1ULL << *restore_sdma_id);
+			clear_bit(*restore_sdma_id, dqm->xgmi_sdma_bitmap);
 			q->sdma_id = *restore_sdma_id;
 		} else {
-			bit = __ffs64(dqm->xgmi_sdma_bitmap);
-			dqm->xgmi_sdma_bitmap &= ~(1ULL << bit);
+			bit = find_first_bit(dqm->xgmi_sdma_bitmap,
+					     get_num_xgmi_sdma_queues(dqm));
+			clear_bit(bit, dqm->xgmi_sdma_bitmap);
 			q->sdma_id = bit;
 		}
 		/* sdma_engine_id is sdma id including
@@ -1395,6 +1646,41 @@ static int allocate_sdma_queue(struct device_queue_manager *dqm,
 			q->sdma_id % kfd_get_num_xgmi_sdma_engines(dqm->dev);
 		q->properties.sdma_queue_id = q->sdma_id /
 			kfd_get_num_xgmi_sdma_engines(dqm->dev);
+	} else if (q->properties.type == KFD_QUEUE_TYPE_SDMA_BY_ENG_ID) {
+		int i, num_queues, num_engines, eng_offset = 0, start_engine;
+		bool free_bit_found = false, is_xgmi = false;
+
+		if (q->properties.sdma_engine_id < kfd_get_num_sdma_engines(dqm->dev)) {
+			num_queues = get_num_sdma_queues(dqm);
+			num_engines = kfd_get_num_sdma_engines(dqm->dev);
+			q->properties.type = KFD_QUEUE_TYPE_SDMA;
+		} else {
+			num_queues = get_num_xgmi_sdma_queues(dqm);
+			num_engines = kfd_get_num_xgmi_sdma_engines(dqm->dev);
+			eng_offset = kfd_get_num_sdma_engines(dqm->dev);
+			q->properties.type = KFD_QUEUE_TYPE_SDMA_XGMI;
+			is_xgmi = true;
+		}
+
+		/* Scan available bit based on target engine ID. */
+		start_engine = q->properties.sdma_engine_id - eng_offset;
+		for (i = start_engine; i < num_queues; i += num_engines) {
+
+			if (!test_bit(i, is_xgmi ? dqm->xgmi_sdma_bitmap : dqm->sdma_bitmap))
+				continue;
+
+			clear_bit(i, is_xgmi ? dqm->xgmi_sdma_bitmap : dqm->sdma_bitmap);
+			q->sdma_id = i;
+			q->properties.sdma_queue_id = q->sdma_id / num_engines;
+			free_bit_found = true;
+			break;
+		}
+
+		if (!free_bit_found) {
+			dev_warn(dev, "No more SDMA queue to allocate for target ID %i (%d total queues)\n",
+				 q->properties.sdma_engine_id, num_queues);
+			return -ENOMEM;
+		}
 	}
 
 	pr_debug("SDMA engine id: %d\n", q->properties.sdma_engine_id);
@@ -1409,11 +1695,11 @@ static void deallocate_sdma_queue(struct device_queue_manager *dqm,
 	if (q->properties.type == KFD_QUEUE_TYPE_SDMA) {
 		if (q->sdma_id >= get_num_sdma_queues(dqm))
 			return;
-		dqm->sdma_bitmap |= (1ULL << q->sdma_id);
+		set_bit(q->sdma_id, dqm->sdma_bitmap);
 	} else if (q->properties.type == KFD_QUEUE_TYPE_SDMA_XGMI) {
 		if (q->sdma_id >= get_num_xgmi_sdma_queues(dqm))
 			return;
-		dqm->xgmi_sdma_bitmap |= (1ULL << q->sdma_id);
+		set_bit(q->sdma_id, dqm->xgmi_sdma_bitmap);
 	}
 }
 
@@ -1425,15 +1711,16 @@ static int set_sched_resources(struct device_queue_manager *dqm)
 {
 	int i, mec;
 	struct scheduling_resources res;
+	struct device *dev = dqm->dev->adev->dev;
 
-	res.vmid_mask = dqm->dev->shared_resources.compute_vmid_bitmap;
+	res.vmid_mask = dqm->dev->compute_vmid_bitmap;
 
 	res.queue_mask = 0;
-	for (i = 0; i < KGD_MAX_QUEUES; ++i) {
-		mec = (i / dqm->dev->shared_resources.num_queue_per_pipe)
-			/ dqm->dev->shared_resources.num_pipe_per_mec;
+	for (i = 0; i < AMDGPU_MAX_QUEUES; ++i) {
+		mec = (i / dqm->dev->kfd->shared_resources.num_queue_per_pipe)
+			/ dqm->dev->kfd->shared_resources.num_pipe_per_mec;
 
-		if (!test_bit(i, dqm->dev->shared_resources.cp_queue_bitmap))
+		if (!test_bit(i, dqm->dev->kfd->shared_resources.cp_queue_bitmap))
 			continue;
 
 		/* only acquire queues from the first MEC */
@@ -1445,7 +1732,7 @@ static int set_sched_resources(struct device_queue_manager *dqm)
 		 * definition of res.queue_mask needs updating
 		 */
 		if (WARN_ON(i >= (sizeof(res.queue_mask)*8))) {
-			pr_err("Invalid queue enabled by amdgpu: %d\n", i);
+			dev_err(dev, "Invalid queue enabled by amdgpu: %d\n", i);
 			break;
 		}
 
@@ -1474,22 +1761,81 @@ static int initialize_cpsch(struct device_queue_manager *dqm)
 	dqm->active_cp_queue_count = 0;
 	dqm->gws_queue_count = 0;
 	dqm->active_runlist = false;
-	INIT_WORK(&dqm->hw_exception_work, kfd_process_hw_exception);
+	dqm->trap_debug_vmid = 0;
 
 	init_sdma_bitmaps(dqm);
 
+	update_dqm_wait_times(dqm);
 	return 0;
 }
 
+/* halt_cpsch:
+ * Unmap queues so the schedule doesn't continue remaining jobs in the queue.
+ * Then set dqm->sched_halt so queues don't map to runlist until unhalt_cpsch
+ * is called.
+ */
+static int halt_cpsch(struct device_queue_manager *dqm)
+{
+	int ret = 0;
+
+	dqm_lock(dqm);
+	if (!dqm->sched_running) {
+		dqm_unlock(dqm);
+		return 0;
+	}
+
+	WARN_ONCE(dqm->sched_halt, "Scheduling is already on halt\n");
+
+	if (!dqm->is_hws_hang) {
+		if (!dqm->dev->kfd->shared_resources.enable_mes)
+			ret = unmap_queues_cpsch(dqm,
+						 KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0,
+				USE_DEFAULT_GRACE_PERIOD, false);
+		else
+			ret = remove_all_kfd_queues_mes(dqm);
+	}
+	dqm->sched_halt = true;
+	dqm_unlock(dqm);
+
+	return ret;
+}
+
+/* unhalt_cpsch
+ * Unset dqm->sched_halt and map queues back to runlist
+ */
+static int unhalt_cpsch(struct device_queue_manager *dqm)
+{
+	int ret = 0;
+
+	dqm_lock(dqm);
+	if (!dqm->sched_running || !dqm->sched_halt) {
+		WARN_ONCE(!dqm->sched_halt, "Scheduling is not on halt.\n");
+		dqm_unlock(dqm);
+		return 0;
+	}
+	dqm->sched_halt = false;
+	if (!dqm->dev->kfd->shared_resources.enable_mes)
+		ret = execute_queues_cpsch(dqm,
+					   KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES,
+			0, USE_DEFAULT_GRACE_PERIOD);
+	else
+		ret = add_all_kfd_queues_mes(dqm);
+
+	dqm_unlock(dqm);
+
+	return ret;
+}
+
 static int start_cpsch(struct device_queue_manager *dqm)
 {
-	int retval;
+	struct device *dev = dqm->dev->adev->dev;
+	int retval, num_hw_queue_slots;
 
 	retval = 0;
 
 	dqm_lock(dqm);
 
-	if (!dqm->dev->shared_resources.enable_mes) {
+	if (!dqm->dev->kfd->shared_resources.enable_mes) {
 		retval = pm_init(&dqm->packet_mgr, dqm);
 		if (retval)
 			goto fail_packet_manager_init;
@@ -1513,18 +1859,37 @@ static int start_cpsch(struct device_queue_manager *dqm)
 	init_interrupts(dqm);
 
 	/* clear hang status when driver try to start the hw scheduler */
-	dqm->is_hws_hang = false;
-	dqm->is_resetting = false;
 	dqm->sched_running = true;
-	if (!dqm->dev->shared_resources.enable_mes)
-		execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
+
+	if (!dqm->dev->kfd->shared_resources.enable_mes) {
+		if (pm_config_dequeue_wait_counts(&dqm->packet_mgr,
+				KFD_DEQUEUE_WAIT_INIT, 0 /* unused */))
+			dev_err(dev, "Setting optimized dequeue wait failed. Using default values\n");
+		execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0, USE_DEFAULT_GRACE_PERIOD);
+	}
+
+	/* setup per-queue reset detection buffer  */
+	num_hw_queue_slots =  dqm->dev->kfd->shared_resources.num_queue_per_pipe *
+			      dqm->dev->kfd->shared_resources.num_pipe_per_mec *
+			      NUM_XCC(dqm->dev->xcc_mask);
+
+	dqm->detect_hang_info_size = num_hw_queue_slots * sizeof(struct dqm_detect_hang_info);
+	dqm->detect_hang_info = kzalloc(dqm->detect_hang_info_size, GFP_KERNEL);
+
+	if (!dqm->detect_hang_info) {
+		retval = -ENOMEM;
+		goto fail_detect_hang_buffer;
+	}
+
 	dqm_unlock(dqm);
 
 	return 0;
+fail_detect_hang_buffer:
+	kfd_gtt_sa_free(dqm->dev, dqm->fence_mem);
 fail_allocate_vidmem:
 fail_set_sched_resources:
-	if (!dqm->dev->shared_resources.enable_mes)
-		pm_uninit(&dqm->packet_mgr, false);
+	if (!dqm->dev->kfd->shared_resources.enable_mes)
+		pm_uninit(&dqm->packet_mgr);
 fail_packet_manager_init:
 	dqm_unlock(dqm);
 	return retval;
@@ -1532,7 +1897,7 @@ fail_packet_manager_init:
 
 static int stop_cpsch(struct device_queue_manager *dqm)
 {
-	bool hanging;
+	int ret = 0;
 
 	dqm_lock(dqm);
 	if (!dqm->sched_running) {
@@ -1540,25 +1905,25 @@ static int stop_cpsch(struct device_queue_manager *dqm)
 		return 0;
 	}
 
-	if (!dqm->is_hws_hang) {
-		if (!dqm->dev->shared_resources.enable_mes)
-			unmap_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0, false);
-		else
-			remove_all_queues_mes(dqm);
-	}
+	if (!dqm->dev->kfd->shared_resources.enable_mes)
+		ret = unmap_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES,
+								0, USE_DEFAULT_GRACE_PERIOD, false);
+	else
+		ret = remove_all_kfd_queues_mes(dqm);
 
-	hanging = dqm->is_hws_hang || dqm->is_resetting;
 	dqm->sched_running = false;
 
-	if (!dqm->dev->shared_resources.enable_mes)
+	if (!dqm->dev->kfd->shared_resources.enable_mes)
 		pm_release_ib(&dqm->packet_mgr);
 
 	kfd_gtt_sa_free(dqm->dev, dqm->fence_mem);
-	if (!dqm->dev->shared_resources.enable_mes)
-		pm_uninit(&dqm->packet_mgr, hanging);
+	if (!dqm->dev->kfd->shared_resources.enable_mes)
+		pm_uninit(&dqm->packet_mgr);
+	kfree(dqm->detect_hang_info);
+	dqm->detect_hang_info = NULL;
 	dqm_unlock(dqm);
 
-	return 0;
+	return ret;
 }
 
 static int create_kernel_queue_cpsch(struct device_queue_manager *dqm,
@@ -1584,7 +1949,8 @@ static int create_kernel_queue_cpsch(struct device_queue_manager *dqm,
 	list_add(&kq->list, &qpd->priv_queue_list);
 	increment_queue_count(dqm, qpd, kq->queue);
 	qpd->is_debug = true;
-	execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
+	execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0,
+			USE_DEFAULT_GRACE_PERIOD);
 	dqm_unlock(dqm);
 
 	return 0;
@@ -1598,7 +1964,8 @@ static void destroy_kernel_queue_cpsch(struct device_queue_manager *dqm,
 	list_del(&kq->list);
 	decrement_queue_count(dqm, qpd, kq->queue);
 	qpd->is_debug = false;
-	execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0);
+	execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0,
+			USE_DEFAULT_GRACE_PERIOD);
 	/*
 	 * Unconditionally decrement this counter, regardless of the queue's
 	 * type.
@@ -1625,7 +1992,8 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q,
 	}
 
 	if (q->properties.type == KFD_QUEUE_TYPE_SDMA ||
-		q->properties.type == KFD_QUEUE_TYPE_SDMA_XGMI) {
+		q->properties.type == KFD_QUEUE_TYPE_SDMA_XGMI ||
+		q->properties.type == KFD_QUEUE_TYPE_SDMA_BY_ENG_ID) {
 		dqm_lock(dqm);
 		retval = allocate_sdma_queue(dqm, q, qd ? &qd->sdma_id : NULL);
 		dqm_unlock(dqm);
@@ -1658,6 +2026,8 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q,
 	 * updates the is_evicted flag but is a no-op otherwise.
 	 */
 	q->properties.is_evicted = !!qpd->evicted;
+	q->properties.is_dbg_wa = qpd->pqm->process->debug_trap_enabled &&
+				  kfd_dbg_has_cwsr_workaround(q->device);
 
 	if (qd)
 		mqd_mgr->restore_mqd(mqd_mgr, &q->mqd, q->mqd_mem_obj, &q->gart_mqd_addr,
@@ -1673,9 +2043,9 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q,
 	if (q->properties.is_active) {
 		increment_queue_count(dqm, qpd, q);
 
-		if (!dqm->dev->shared_resources.enable_mes)
+		if (!dqm->dev->kfd->shared_resources.enable_mes)
 			retval = execute_queues_cpsch(dqm,
-					KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
+					KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0, USE_DEFAULT_GRACE_PERIOD);
 		else
 			retval = add_queue_mes(dqm, q, qpd);
 		if (retval)
@@ -1714,15 +2084,22 @@ out:
 	return retval;
 }
 
-int amdkfd_fence_wait_timeout(uint64_t *fence_addr,
-				uint64_t fence_value,
-				unsigned int timeout_ms)
+int amdkfd_fence_wait_timeout(struct device_queue_manager *dqm,
+			      uint64_t fence_value,
+			      unsigned int timeout_ms)
 {
 	unsigned long end_jiffies = msecs_to_jiffies(timeout_ms) + jiffies;
+	struct device *dev = dqm->dev->adev->dev;
+	uint64_t *fence_addr = dqm->fence_addr;
 
 	while (*fence_addr != fence_value) {
+		/* Fatal err detected, this response won't come */
+		if (amdgpu_amdkfd_is_fed(dqm->dev->adev) ||
+		    amdgpu_in_reset(dqm->dev->adev))
+			return -EIO;
+
 		if (time_after(jiffies, end_jiffies)) {
-			pr_err("qcm fence wait loop timeout expired\n");
+			dev_err(dev, "qcm fence wait loop timeout expired\n");
 			/* In HWS case, this is used to halt the driver thread
 			 * in order not to mess up CP states before doing
 			 * scandumps for FW debugging.
@@ -1741,9 +2118,10 @@ int amdkfd_fence_wait_timeout(uint64_t *fence_addr,
 /* dqm->lock mutex has to be locked before calling this function */
 static int map_queues_cpsch(struct device_queue_manager *dqm)
 {
+	struct device *dev = dqm->dev->adev->dev;
 	int retval;
 
-	if (!dqm->sched_running)
+	if (!dqm->sched_running || dqm->sched_halt)
 		return 0;
 	if (dqm->active_queue_count <= 0 || dqm->processes_count <= 0)
 		return 0;
@@ -1753,7 +2131,7 @@ static int map_queues_cpsch(struct device_queue_manager *dqm)
 	retval = pm_send_runlist(&dqm->packet_mgr, &dqm->queues);
 	pr_debug("%s sent runlist\n", __func__);
 	if (retval) {
-		pr_err("failed to execute runlist\n");
+		dev_err(dev, "failed to execute runlist\n");
 		return retval;
 	}
 	dqm->active_runlist = true;
@@ -1761,35 +2139,288 @@ static int map_queues_cpsch(struct device_queue_manager *dqm)
 	return retval;
 }
 
-/* dqm->lock mutex has to be locked before calling this function */
+static void set_queue_as_reset(struct device_queue_manager *dqm, struct queue *q,
+			       struct qcm_process_device *qpd)
+{
+	struct kfd_process_device *pdd = qpd_to_pdd(qpd);
+
+	dev_err(dqm->dev->adev->dev, "queue id 0x%0x at pasid %d is reset\n",
+		q->properties.queue_id, pdd->process->lead_thread->pid);
+
+	pdd->has_reset_queue = true;
+	if (q->properties.is_active) {
+		q->properties.is_active = false;
+		decrement_queue_count(dqm, qpd, q);
+	}
+}
+
+static int detect_queue_hang(struct device_queue_manager *dqm)
+{
+	int i;
+
+	/* detect should be used only in dqm locked queue reset */
+	if (WARN_ON(dqm->detect_hang_count > 0))
+		return 0;
+
+	memset(dqm->detect_hang_info, 0, dqm->detect_hang_info_size);
+
+	for (i = 0; i < AMDGPU_MAX_QUEUES; ++i) {
+		uint32_t mec, pipe, queue;
+		int xcc_id;
+
+		mec = (i / dqm->dev->kfd->shared_resources.num_queue_per_pipe)
+			/ dqm->dev->kfd->shared_resources.num_pipe_per_mec;
+
+		if (mec || !test_bit(i, dqm->dev->kfd->shared_resources.cp_queue_bitmap))
+			continue;
+
+		amdgpu_queue_mask_bit_to_mec_queue(dqm->dev->adev, i, &mec, &pipe, &queue);
+
+		for_each_inst(xcc_id, dqm->dev->xcc_mask) {
+			uint64_t queue_addr = dqm->dev->kfd2kgd->hqd_get_pq_addr(
+						dqm->dev->adev, pipe, queue, xcc_id);
+			struct dqm_detect_hang_info hang_info;
+
+			if (!queue_addr)
+				continue;
+
+			hang_info.pipe_id = pipe;
+			hang_info.queue_id = queue;
+			hang_info.xcc_id = xcc_id;
+			hang_info.queue_address = queue_addr;
+
+			dqm->detect_hang_info[dqm->detect_hang_count] = hang_info;
+			dqm->detect_hang_count++;
+		}
+	}
+
+	return dqm->detect_hang_count;
+}
+
+static struct queue *find_queue_by_address(struct device_queue_manager *dqm, uint64_t queue_address)
+{
+	struct device_process_node *cur;
+	struct qcm_process_device *qpd;
+	struct queue *q;
+
+	list_for_each_entry(cur, &dqm->queues, list) {
+		qpd = cur->qpd;
+		list_for_each_entry(q, &qpd->queues_list, list) {
+			if (queue_address == q->properties.queue_address)
+				return q;
+		}
+	}
+
+	return NULL;
+}
+
+static int reset_hung_queues(struct device_queue_manager *dqm)
+{
+	int r = 0, reset_count = 0, i;
+
+	if (!dqm->detect_hang_info || dqm->is_hws_hang)
+		return -EIO;
+
+	/* assume dqm locked. */
+	if (!detect_queue_hang(dqm))
+		return -ENOTRECOVERABLE;
+
+	for (i = 0; i < dqm->detect_hang_count; i++) {
+		struct dqm_detect_hang_info hang_info = dqm->detect_hang_info[i];
+		struct queue *q = find_queue_by_address(dqm, hang_info.queue_address);
+		struct kfd_process_device *pdd;
+		uint64_t queue_addr = 0;
+
+		if (!q) {
+			r = -ENOTRECOVERABLE;
+			goto reset_fail;
+		}
+
+		pdd = kfd_get_process_device_data(dqm->dev, q->process);
+		if (!pdd) {
+			r = -ENOTRECOVERABLE;
+			goto reset_fail;
+		}
+
+		queue_addr = dqm->dev->kfd2kgd->hqd_reset(dqm->dev->adev,
+				hang_info.pipe_id, hang_info.queue_id, hang_info.xcc_id,
+				KFD_UNMAP_LATENCY_MS);
+
+		/* either reset failed or we reset an unexpected queue. */
+		if (queue_addr != q->properties.queue_address) {
+			r = -ENOTRECOVERABLE;
+			goto reset_fail;
+		}
+
+		set_queue_as_reset(dqm, q, &pdd->qpd);
+		reset_count++;
+	}
+
+	if (reset_count == dqm->detect_hang_count)
+		kfd_signal_reset_event(dqm->dev);
+	else
+		r = -ENOTRECOVERABLE;
+
+reset_fail:
+	dqm->detect_hang_count = 0;
+
+	return r;
+}
+
+static bool sdma_has_hang(struct device_queue_manager *dqm)
+{
+	int engine_start = dqm->dev->node_id * get_num_all_sdma_engines(dqm);
+	int engine_end = engine_start + get_num_all_sdma_engines(dqm);
+	int num_queues_per_eng =  dqm->dev->kfd->device_info.num_sdma_queues_per_engine;
+	int i, j;
+
+	for (i = engine_start; i < engine_end; i++) {
+		for (j = 0; j < num_queues_per_eng; j++) {
+			if (!dqm->dev->kfd2kgd->hqd_sdma_get_doorbell(dqm->dev->adev, i, j))
+				continue;
+
+			return true;
+		}
+	}
+
+	return false;
+}
+
+static bool set_sdma_queue_as_reset(struct device_queue_manager *dqm,
+				    uint32_t doorbell_off)
+{
+	struct device_process_node *cur;
+	struct qcm_process_device *qpd;
+	struct queue *q;
+
+	list_for_each_entry(cur, &dqm->queues, list) {
+		qpd = cur->qpd;
+		list_for_each_entry(q, &qpd->queues_list, list) {
+			if ((q->properties.type == KFD_QUEUE_TYPE_SDMA ||
+			     q->properties.type == KFD_QUEUE_TYPE_SDMA_XGMI) &&
+			     q->properties.doorbell_off == doorbell_off) {
+				set_queue_as_reset(dqm, q, qpd);
+				return true;
+			}
+		}
+	}
+
+	return false;
+}
+
+static int reset_hung_queues_sdma(struct device_queue_manager *dqm)
+{
+	int engine_start = dqm->dev->node_id * get_num_all_sdma_engines(dqm);
+	int engine_end = engine_start + get_num_all_sdma_engines(dqm);
+	int num_queues_per_eng =  dqm->dev->kfd->device_info.num_sdma_queues_per_engine;
+	int r = 0, i, j;
+
+	if (dqm->is_hws_hang)
+		return -EIO;
+
+	/* Scan for hung HW queues and reset engine. */
+	dqm->detect_hang_count = 0;
+	for (i = engine_start; i < engine_end; i++) {
+		for (j = 0; j < num_queues_per_eng; j++) {
+			uint32_t doorbell_off =
+				dqm->dev->kfd2kgd->hqd_sdma_get_doorbell(dqm->dev->adev, i, j);
+
+			if (!doorbell_off)
+				continue;
+
+			/* Reset engine and check. */
+			if (amdgpu_sdma_reset_engine(dqm->dev->adev, i, false) ||
+			    dqm->dev->kfd2kgd->hqd_sdma_get_doorbell(dqm->dev->adev, i, j) ||
+			    !set_sdma_queue_as_reset(dqm, doorbell_off)) {
+				r = -ENOTRECOVERABLE;
+				goto reset_fail;
+			}
+
+			/* Should only expect one queue active per engine */
+			dqm->detect_hang_count++;
+			break;
+		}
+	}
+
+	/* Signal process reset */
+	if (dqm->detect_hang_count)
+		kfd_signal_reset_event(dqm->dev);
+	else
+		r = -ENOTRECOVERABLE;
+
+reset_fail:
+	dqm->detect_hang_count = 0;
+
+	return r;
+}
+
+static int reset_queues_on_hws_hang(struct device_queue_manager *dqm, bool is_sdma)
+{
+	struct amdgpu_device *adev = dqm->dev->adev;
+
+	while (halt_if_hws_hang)
+		schedule();
+
+	if (adev->debug_disable_gpu_ring_reset) {
+		dev_info_once(adev->dev,
+			      "%s queue hung, but ring reset disabled",
+			      is_sdma ? "sdma" : "compute");
+
+		return -EPERM;
+	}
+	if (!amdgpu_gpu_recovery)
+		return -ENOTRECOVERABLE;
+
+	return is_sdma ? reset_hung_queues_sdma(dqm) : reset_hung_queues(dqm);
+}
+
+/* dqm->lock mutex has to be locked before calling this function
+ *
+ * @grace_period: If USE_DEFAULT_GRACE_PERIOD then default wait time
+ *   for context switch latency. Lower values are used by debugger
+ *   since context switching are triggered at high frequency.
+ *   This is configured by setting CP_IQ_WAIT_TIME2.SCH_WAVE
+ *
+ */
 static int unmap_queues_cpsch(struct device_queue_manager *dqm,
 				enum kfd_unmap_queues_filter filter,
-				uint32_t filter_param, bool reset)
+				uint32_t filter_param,
+				uint32_t grace_period,
+				bool reset)
 {
-	int retval = 0;
+	struct device *dev = dqm->dev->adev->dev;
 	struct mqd_manager *mqd_mgr;
+	int retval;
 
 	if (!dqm->sched_running)
 		return 0;
-	if (dqm->is_hws_hang || dqm->is_resetting)
-		return -EIO;
 	if (!dqm->active_runlist)
-		return retval;
+		return 0;
+	if (!down_read_trylock(&dqm->dev->adev->reset_domain->sem))
+		return -EIO;
+
+	if (grace_period != USE_DEFAULT_GRACE_PERIOD) {
+		retval = pm_config_dequeue_wait_counts(&dqm->packet_mgr,
+				KFD_DEQUEUE_WAIT_SET_SCH_WAVE, grace_period);
+		if (retval)
+			goto out;
+	}
 
 	retval = pm_send_unmap_queue(&dqm->packet_mgr, filter, filter_param, reset);
 	if (retval)
-		return retval;
+		goto out;
 
 	*dqm->fence_addr = KFD_FENCE_INIT;
+	mb();
 	pm_send_query_status(&dqm->packet_mgr, dqm->fence_gpu_addr,
 				KFD_FENCE_COMPLETED);
 	/* should be timed out */
-	retval = amdkfd_fence_wait_timeout(dqm->fence_addr, KFD_FENCE_COMPLETED,
-				queue_preemption_timeout_ms);
+	retval = amdkfd_fence_wait_timeout(dqm, KFD_FENCE_COMPLETED,
+					   queue_preemption_timeout_ms);
 	if (retval) {
-		pr_err("The cp might be in an unrecoverable state due to an unsuccessful queues preemption\n");
+		dev_err(dev, "The cp might be in an unrecoverable state due to an unsuccessful queues preemption\n");
 		kfd_hws_hang(dqm);
-		return retval;
+		goto out;
 	}
 
 	/* In the current MEC firmware implementation, if compute queue
@@ -1801,29 +2432,43 @@ static int unmap_queues_cpsch(struct device_queue_manager *dqm,
 	 * check those fields
 	 */
 	mqd_mgr = dqm->mqd_mgrs[KFD_MQD_TYPE_HIQ];
-	if (mqd_mgr->read_doorbell_id(dqm->packet_mgr.priv_queue->queue->mqd)) {
-		pr_err("HIQ MQD's queue_doorbell_id0 is not 0, Queue preemption time out\n");
-		while (halt_if_hws_hang)
-			schedule();
-		return -ETIME;
+	if (mqd_mgr->check_preemption_failed(mqd_mgr, dqm->packet_mgr.priv_queue->queue->mqd) &&
+	    reset_queues_on_hws_hang(dqm, false))
+		goto reset_fail;
+
+	/* Check for SDMA hang and attempt SDMA reset */
+	if (sdma_has_hang(dqm) && reset_queues_on_hws_hang(dqm, true))
+		goto reset_fail;
+
+	/* We need to reset the grace period value for this device */
+	if (grace_period != USE_DEFAULT_GRACE_PERIOD) {
+		if (pm_config_dequeue_wait_counts(&dqm->packet_mgr,
+				KFD_DEQUEUE_WAIT_RESET, 0 /* unused */))
+			dev_err(dev, "Failed to reset grace period\n");
 	}
 
 	pm_release_ib(&dqm->packet_mgr);
 	dqm->active_runlist = false;
-
+out:
+	up_read(&dqm->dev->adev->reset_domain->sem);
 	return retval;
+
+reset_fail:
+	dqm->is_hws_hang = true;
+	kfd_hws_hang(dqm);
+	up_read(&dqm->dev->adev->reset_domain->sem);
+	return -ETIME;
 }
 
 /* only for compute queue */
-static int reset_queues_cpsch(struct device_queue_manager *dqm,
-			uint16_t pasid)
+static int reset_queues_cpsch(struct device_queue_manager *dqm, uint16_t pasid)
 {
 	int retval;
 
 	dqm_lock(dqm);
 
 	retval = unmap_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_BY_PASID,
-			pasid, true);
+			pasid, USE_DEFAULT_GRACE_PERIOD, true);
 
 	dqm_unlock(dqm);
 	return retval;
@@ -1832,17 +2477,46 @@ static int reset_queues_cpsch(struct device_queue_manager *dqm,
 /* dqm->lock mutex has to be locked before calling this function */
 static int execute_queues_cpsch(struct device_queue_manager *dqm,
 				enum kfd_unmap_queues_filter filter,
-				uint32_t filter_param)
+				uint32_t filter_param,
+				uint32_t grace_period)
 {
 	int retval;
 
-	if (dqm->is_hws_hang)
+	if (!down_read_trylock(&dqm->dev->adev->reset_domain->sem))
 		return -EIO;
-	retval = unmap_queues_cpsch(dqm, filter, filter_param, false);
-	if (retval)
-		return retval;
+	retval = unmap_queues_cpsch(dqm, filter, filter_param, grace_period, false);
+	if (!retval)
+		retval = map_queues_cpsch(dqm);
+	up_read(&dqm->dev->adev->reset_domain->sem);
+	return retval;
+}
+
+static int wait_on_destroy_queue(struct device_queue_manager *dqm,
+				 struct queue *q)
+{
+	struct kfd_process_device *pdd = kfd_get_process_device_data(q->device,
+								q->process);
+	int ret = 0;
+
+	if (WARN_ON(!pdd))
+		return ret;
+
+	if (pdd->qpd.is_debug)
+		return ret;
+
+	q->properties.is_being_destroyed = true;
+
+	if (pdd->process->debug_trap_enabled && q->properties.is_suspended) {
+		dqm_unlock(dqm);
+		mutex_unlock(&q->process->mutex);
+		ret = wait_event_interruptible(dqm->destroy_wait,
+						!q->properties.is_suspended);
+
+		mutex_lock(&q->process->mutex);
+		dqm_lock(dqm);
+	}
 
-	return map_queues_cpsch(dqm);
+	return ret;
 }
 
 static int destroy_queue_cpsch(struct device_queue_manager *dqm,
@@ -1853,6 +2527,7 @@ static int destroy_queue_cpsch(struct device_queue_manager *dqm,
 	struct mqd_manager *mqd_mgr;
 	uint64_t sdma_val = 0;
 	struct kfd_process_device *pdd = qpd_to_pdd(qpd);
+	struct device *dev = dqm->dev->adev->dev;
 
 	/* Get the SDMA queue stats */
 	if ((q->properties.type == KFD_QUEUE_TYPE_SDMA) ||
@@ -1860,15 +2535,20 @@ static int destroy_queue_cpsch(struct device_queue_manager *dqm,
 		retval = read_sdma_queue_counter((uint64_t __user *)q->properties.read_ptr,
 							&sdma_val);
 		if (retval)
-			pr_err("Failed to read SDMA queue counter for queue: %d\n",
+			dev_err(dev, "Failed to read SDMA queue counter for queue: %d\n",
 				q->properties.queue_id);
 	}
 
-	retval = 0;
-
 	/* remove queue from list to prevent rescheduling after preemption */
 	dqm_lock(dqm);
 
+	retval = wait_on_destroy_queue(dqm, q);
+
+	if (retval) {
+		dqm_unlock(dqm);
+		return retval;
+	}
+
 	if (qpd->is_debug) {
 		/*
 		 * error, currently we do not allow to destroy a queue
@@ -1890,19 +2570,21 @@ static int destroy_queue_cpsch(struct device_queue_manager *dqm,
 		pdd->sdma_past_activity_counter += sdma_val;
 	}
 
-	list_del(&q->list);
-	qpd->queue_count--;
 	if (q->properties.is_active) {
-		if (!dqm->dev->shared_resources.enable_mes) {
-			decrement_queue_count(dqm, qpd, q);
+		decrement_queue_count(dqm, qpd, q);
+		q->properties.is_active = false;
+		if (!dqm->dev->kfd->shared_resources.enable_mes) {
 			retval = execute_queues_cpsch(dqm,
-						      KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
+						      KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0,
+						      USE_DEFAULT_GRACE_PERIOD);
 			if (retval == -ETIME)
 				qpd->reset_wavefronts = true;
 		} else {
 			retval = remove_queue_mes(dqm, q, qpd);
 		}
 	}
+	list_del(&q->list);
+	qpd->queue_count--;
 
 	/*
 	 * Unconditionally decrement this counter, regardless of the queue's
@@ -1914,7 +2596,14 @@ static int destroy_queue_cpsch(struct device_queue_manager *dqm,
 
 	dqm_unlock(dqm);
 
-	/* Do free_mqd after dqm_unlock(dqm) to avoid circular locking */
+	/*
+	 * Do free_mqd and raise delete event after dqm_unlock(dqm) to avoid
+	 * circular locking
+	 */
+	kfd_dbg_ev_raise(KFD_EC_MASK(EC_DEVICE_QUEUE_DELETE),
+				qpd->pqm->process, q->device,
+				-1, false, NULL, 0);
+
 	mqd_mgr->free_mqd(mqd_mgr, q->mqd, q->mqd_mem_obj);
 
 	return retval;
@@ -1925,20 +2614,13 @@ failed_try_destroy_debugged_queue:
 	return retval;
 }
 
-/*
- * Low bits must be 0000/FFFF as required by HW, high bits must be 0 to
- * stay in user mode.
- */
-#define APE1_FIXED_BITS_MASK 0xFFFF80000000FFFFULL
-/* APE1 limit is inclusive and 64K aligned. */
-#define APE1_LIMIT_ALIGNMENT 0xFFFF
-
 static bool set_cache_memory_policy(struct device_queue_manager *dqm,
 				   struct qcm_process_device *qpd,
 				   enum cache_policy default_policy,
 				   enum cache_policy alternate_policy,
 				   void __user *alternate_aperture_base,
-				   uint64_t alternate_aperture_size)
+				   uint64_t alternate_aperture_size,
+				   u32 misc_process_properties)
 {
 	bool retval = true;
 
@@ -1947,41 +2629,17 @@ static bool set_cache_memory_policy(struct device_queue_manager *dqm,
 
 	dqm_lock(dqm);
 
-	if (alternate_aperture_size == 0) {
-		/* base > limit disables APE1 */
-		qpd->sh_mem_ape1_base = 1;
-		qpd->sh_mem_ape1_limit = 0;
-	} else {
-		/*
-		 * In FSA64, APE1_Base[63:0] = { 16{SH_MEM_APE1_BASE[31]},
-		 *			SH_MEM_APE1_BASE[31:0], 0x0000 }
-		 * APE1_Limit[63:0] = { 16{SH_MEM_APE1_LIMIT[31]},
-		 *			SH_MEM_APE1_LIMIT[31:0], 0xFFFF }
-		 * Verify that the base and size parameters can be
-		 * represented in this format and convert them.
-		 * Additionally restrict APE1 to user-mode addresses.
-		 */
-
-		uint64_t base = (uintptr_t)alternate_aperture_base;
-		uint64_t limit = base + alternate_aperture_size - 1;
-
-		if (limit <= base || (base & APE1_FIXED_BITS_MASK) != 0 ||
-		   (limit & APE1_FIXED_BITS_MASK) != APE1_LIMIT_ALIGNMENT) {
-			retval = false;
-			goto out;
-		}
-
-		qpd->sh_mem_ape1_base = base >> 16;
-		qpd->sh_mem_ape1_limit = limit >> 16;
-	}
-
 	retval = dqm->asic_ops.set_cache_memory_policy(
 			dqm,
 			qpd,
 			default_policy,
 			alternate_policy,
 			alternate_aperture_base,
-			alternate_aperture_size);
+			alternate_aperture_size,
+			misc_process_properties);
+
+	if (retval)
+		goto out;
 
 	if ((dqm->sched_policy == KFD_SCHED_POLICY_NO_HWS) && (qpd->vmid != 0))
 		program_sh_mem_settings(dqm, qpd);
@@ -2056,7 +2714,7 @@ static int get_wave_state(struct device_queue_manager *dqm,
 	mqd_mgr = dqm->mqd_mgrs[KFD_MQD_TYPE_CP];
 
 	if (q->properties.type != KFD_QUEUE_TYPE_COMPUTE ||
-	    q->properties.is_active || !q->device->cwsr_enabled ||
+	    q->properties.is_active || !q->device->kfd->cwsr_enabled ||
 	    !mqd_mgr->get_wave_state) {
 		dqm_unlock(dqm);
 		return -EINVAL;
@@ -2069,8 +2727,8 @@ static int get_wave_state(struct device_queue_manager *dqm,
 	 * and the queue should be protected against destruction by the process
 	 * lock.
 	 */
-	return mqd_mgr->get_wave_state(mqd_mgr, q->mqd, ctl_stack,
-			ctl_stack_used_size, save_area_used_size);
+	return mqd_mgr->get_wave_state(mqd_mgr, q->mqd, &q->properties,
+			ctl_stack, ctl_stack_used_size, save_area_used_size);
 }
 
 static void get_queue_checkpoint_info(struct device_queue_manager *dqm,
@@ -2084,7 +2742,7 @@ static void get_queue_checkpoint_info(struct device_queue_manager *dqm,
 
 	dqm_lock(dqm);
 	mqd_mgr = dqm->mqd_mgrs[mqd_type];
-	*mqd_size = mqd_mgr->mqd_size;
+	*mqd_size = mqd_mgr->mqd_size * NUM_XCC(mqd_mgr->dev->xcc_mask);
 	*ctl_stack_size = 0;
 
 	if (q->properties.type == KFD_QUEUE_TYPE_COMPUTE && mqd_mgr->get_checkpoint_info)
@@ -2105,7 +2763,7 @@ static int checkpoint_mqd(struct device_queue_manager *dqm,
 
 	dqm_lock(dqm);
 
-	if (q->properties.is_active || !q->device->cwsr_enabled) {
+	if (q->properties.is_active || !q->device->kfd->cwsr_enabled) {
 		r = -EINVAL;
 		goto dqm_unlock;
 	}
@@ -2128,6 +2786,7 @@ static int process_termination_cpsch(struct device_queue_manager *dqm,
 {
 	int retval;
 	struct queue *q;
+	struct device *dev = dqm->dev->adev->dev;
 	struct kernel_queue *kq, *kq_next;
 	struct mqd_manager *mqd_mgr;
 	struct device_process_node *cur, *next_dpn;
@@ -2158,10 +2817,10 @@ static int process_termination_cpsch(struct device_queue_manager *dqm,
 		if (q->properties.is_active) {
 			decrement_queue_count(dqm, qpd, q);
 
-			if (dqm->dev->shared_resources.enable_mes) {
+			if (dqm->dev->kfd->shared_resources.enable_mes) {
 				retval = remove_queue_mes(dqm, q, qpd);
 				if (retval)
-					pr_err("Failed to remove queue %d\n",
+					dev_err(dev, "Failed to remove queue %d\n",
 						q->properties.queue_id);
 			}
 		}
@@ -2180,13 +2839,15 @@ static int process_termination_cpsch(struct device_queue_manager *dqm,
 		}
 	}
 
-	if (!dqm->dev->shared_resources.enable_mes)
-		retval = execute_queues_cpsch(dqm, filter, 0);
+	if (!dqm->dev->kfd->shared_resources.enable_mes)
+		retval = execute_queues_cpsch(dqm, filter, 0, USE_DEFAULT_GRACE_PERIOD);
 
-	if ((!dqm->is_hws_hang) && (retval || qpd->reset_wavefronts)) {
+	if ((retval || qpd->reset_wavefronts) &&
+	    down_read_trylock(&dqm->dev->adev->reset_domain->sem)) {
 		pr_warn("Resetting wave fronts (cpsch) on dev %p\n", dqm->dev);
 		dbgdev_wave_reset_wavefronts(dqm->dev, qpd->pqm->process);
 		qpd->reset_wavefronts = false;
+		up_read(&dqm->dev->adev->reset_domain->sem);
 	}
 
 	/* Lastly, free mqd resources.
@@ -2216,12 +2877,13 @@ static int process_termination_cpsch(struct device_queue_manager *dqm,
 static int init_mqd_managers(struct device_queue_manager *dqm)
 {
 	int i, j;
+	struct device *dev = dqm->dev->adev->dev;
 	struct mqd_manager *mqd_mgr;
 
 	for (i = 0; i < KFD_MQD_TYPE_MAX; i++) {
 		mqd_mgr = dqm->asic_ops.mqd_manager_init(i, dqm->dev);
 		if (!mqd_mgr) {
-			pr_err("mqd manager [%d] initialization failed\n", i);
+			dev_err(dev, "mqd manager [%d] initialization failed\n", i);
 			goto out_free;
 		}
 		dqm->mqd_mgrs[i] = mqd_mgr;
@@ -2242,12 +2904,13 @@ out_free:
 static int allocate_hiq_sdma_mqd(struct device_queue_manager *dqm)
 {
 	int retval;
-	struct kfd_dev *dev = dqm->dev;
+	struct kfd_node *dev = dqm->dev;
 	struct kfd_mem_obj *mem_obj = &dqm->hiq_sdma_mqd;
 	uint32_t size = dqm->mqd_mgrs[KFD_MQD_TYPE_SDMA]->mqd_size *
 		get_num_all_sdma_engines(dqm) *
-		dev->device_info.num_sdma_queues_per_engine +
-		dqm->mqd_mgrs[KFD_MQD_TYPE_HIQ]->mqd_size;
+		dev->kfd->device_info.num_sdma_queues_per_engine +
+		(dqm->mqd_mgrs[KFD_MQD_TYPE_HIQ]->mqd_size *
+		NUM_XCC(dqm->dev->xcc_mask));
 
 	retval = amdgpu_amdkfd_alloc_gtt_mem(dev->adev, size,
 		&(mem_obj->gtt_mem), &(mem_obj->gpu_addr),
@@ -2256,7 +2919,7 @@ static int allocate_hiq_sdma_mqd(struct device_queue_manager *dqm)
 	return retval;
 }
 
-struct device_queue_manager *device_queue_manager_init(struct kfd_dev *dev)
+struct device_queue_manager *device_queue_manager_init(struct kfd_node *dev)
 {
 	struct device_queue_manager *dqm;
 
@@ -2291,7 +2954,8 @@ struct device_queue_manager *device_queue_manager_init(struct kfd_dev *dev)
 		dqm->ops.initialize = initialize_cpsch;
 		dqm->ops.start = start_cpsch;
 		dqm->ops.stop = stop_cpsch;
-		dqm->ops.pre_reset = pre_reset;
+		dqm->ops.halt = halt_cpsch;
+		dqm->ops.unhalt = unhalt_cpsch;
 		dqm->ops.destroy_queue = destroy_queue_cpsch;
 		dqm->ops.update_queue = update_queue;
 		dqm->ops.register_process = register_process;
@@ -2312,7 +2976,6 @@ struct device_queue_manager *device_queue_manager_init(struct kfd_dev *dev)
 		/* initialize dqm for no cp scheduling */
 		dqm->ops.start = start_nocpsch;
 		dqm->ops.stop = stop_nocpsch;
-		dqm->ops.pre_reset = pre_reset;
 		dqm->ops.create_queue = create_queue_nocpsch;
 		dqm->ops.destroy_queue = destroy_queue_nocpsch;
 		dqm->ops.update_queue = update_queue;
@@ -2330,37 +2993,33 @@ struct device_queue_manager *device_queue_manager_init(struct kfd_dev *dev)
 		dqm->ops.checkpoint_mqd = checkpoint_mqd;
 		break;
 	default:
-		pr_err("Invalid scheduling policy %d\n", dqm->sched_policy);
+		dev_err(dev->adev->dev, "Invalid scheduling policy %d\n", dqm->sched_policy);
 		goto out_free;
 	}
 
 	switch (dev->adev->asic_type) {
-	case CHIP_CARRIZO:
-		device_queue_manager_init_vi(&dqm->asic_ops);
-		break;
-
 	case CHIP_KAVERI:
-		device_queue_manager_init_cik(&dqm->asic_ops);
-		break;
-
 	case CHIP_HAWAII:
-		device_queue_manager_init_cik_hawaii(&dqm->asic_ops);
+		device_queue_manager_init_cik(&dqm->asic_ops);
 		break;
 
+	case CHIP_CARRIZO:
 	case CHIP_TONGA:
 	case CHIP_FIJI:
 	case CHIP_POLARIS10:
 	case CHIP_POLARIS11:
 	case CHIP_POLARIS12:
 	case CHIP_VEGAM:
-		device_queue_manager_init_vi_tonga(&dqm->asic_ops);
+		device_queue_manager_init_vi(&dqm->asic_ops);
 		break;
 
 	default:
-		if (KFD_GC_VERSION(dev) >= IP_VERSION(11, 0, 0))
+		if (KFD_GC_VERSION(dev) >= IP_VERSION(12, 0, 0))
+			device_queue_manager_init_v12(&dqm->asic_ops);
+		else if (KFD_GC_VERSION(dev) >= IP_VERSION(11, 0, 0))
 			device_queue_manager_init_v11(&dqm->asic_ops);
 		else if (KFD_GC_VERSION(dev) >= IP_VERSION(10, 1, 1))
-			device_queue_manager_init_v10_navi10(&dqm->asic_ops);
+			device_queue_manager_init_v10(&dqm->asic_ops);
 		else if (KFD_GC_VERSION(dev) >= IP_VERSION(9, 0, 1))
 			device_queue_manager_init_v9(&dqm->asic_ops);
 		else {
@@ -2373,59 +3032,619 @@ struct device_queue_manager *device_queue_manager_init(struct kfd_dev *dev)
 	if (init_mqd_managers(dqm))
 		goto out_free;
 
-	if (!dev->shared_resources.enable_mes && allocate_hiq_sdma_mqd(dqm)) {
-		pr_err("Failed to allocate hiq sdma mqd trunk buffer\n");
+	if (!dev->kfd->shared_resources.enable_mes && allocate_hiq_sdma_mqd(dqm)) {
+		dev_err(dev->adev->dev, "Failed to allocate hiq sdma mqd trunk buffer\n");
 		goto out_free;
 	}
 
-	if (!dqm->ops.initialize(dqm))
+	if (!dqm->ops.initialize(dqm)) {
+		init_waitqueue_head(&dqm->destroy_wait);
 		return dqm;
+	}
 
 out_free:
 	kfree(dqm);
 	return NULL;
 }
 
-static void deallocate_hiq_sdma_mqd(struct kfd_dev *dev,
+static void deallocate_hiq_sdma_mqd(struct kfd_node *dev,
 				    struct kfd_mem_obj *mqd)
 {
 	WARN(!mqd, "No hiq sdma mqd trunk to free");
 
-	amdgpu_amdkfd_free_gtt_mem(dev->adev, mqd->gtt_mem);
+	amdgpu_amdkfd_free_gtt_mem(dev->adev, &mqd->gtt_mem);
 }
 
 void device_queue_manager_uninit(struct device_queue_manager *dqm)
 {
+	dqm->ops.stop(dqm);
 	dqm->ops.uninitialize(dqm);
-	if (!dqm->dev->shared_resources.enable_mes)
+	if (!dqm->dev->kfd->shared_resources.enable_mes)
 		deallocate_hiq_sdma_mqd(dqm->dev, &dqm->hiq_sdma_mqd);
 	kfree(dqm);
 }
 
-int kfd_dqm_evict_pasid(struct device_queue_manager *dqm, u32 pasid)
+int kfd_dqm_suspend_bad_queue_mes(struct kfd_node *knode, u32 pasid, u32 doorbell_id)
 {
-	struct kfd_process_device *pdd;
-	struct kfd_process *p = kfd_lookup_process_by_pasid(pasid);
+	struct kfd_process_device *pdd = NULL;
+	struct kfd_process *p = kfd_lookup_process_by_pasid(pasid, &pdd);
+	struct device_queue_manager *dqm = knode->dqm;
+	struct device *dev = dqm->dev->adev->dev;
+	struct qcm_process_device *qpd;
+	struct queue *q = NULL;
 	int ret = 0;
 
-	if (!p)
+	if (!pdd)
 		return -EINVAL;
-	WARN(debug_evictions, "Evicting pid %d", p->lead_thread->pid);
-	pdd = kfd_get_process_device_data(dqm->dev, p);
-	if (pdd)
-		ret = dqm->ops.evict_process_queues(dqm, &pdd->qpd);
-	kfd_unref_process(p);
 
+	dqm_lock(dqm);
+
+	if (pdd) {
+		qpd = &pdd->qpd;
+
+		list_for_each_entry(q, &qpd->queues_list, list) {
+			if (q->doorbell_id == doorbell_id && q->properties.is_active) {
+				ret = suspend_all_queues_mes(dqm);
+				if (ret) {
+					dev_err(dev, "Suspending all queues failed");
+					goto out;
+				}
+
+				q->properties.is_evicted = true;
+				q->properties.is_active = false;
+				decrement_queue_count(dqm, qpd, q);
+
+				ret = remove_queue_mes(dqm, q, qpd);
+				if (ret) {
+					dev_err(dev, "Removing bad queue failed");
+					goto out;
+				}
+
+				ret = resume_all_queues_mes(dqm);
+				if (ret)
+					dev_err(dev, "Resuming all queues failed");
+
+				break;
+			}
+		}
+	}
+
+out:
+	dqm_unlock(dqm);
+	kfd_unref_process(p);
 	return ret;
 }
 
-static void kfd_process_hw_exception(struct work_struct *work)
+int kfd_evict_process_device(struct kfd_process_device *pdd)
 {
-	struct device_queue_manager *dqm = container_of(work,
-			struct device_queue_manager, hw_exception_work);
-	amdgpu_amdkfd_gpu_reset(dqm->dev->adev);
+	struct device_queue_manager *dqm;
+	struct kfd_process *p;
+
+	p = pdd->process;
+	dqm = pdd->dev->dqm;
+
+	WARN(debug_evictions, "Evicting pid %d", p->lead_thread->pid);
+
+	return dqm->ops.evict_process_queues(dqm, &pdd->qpd);
+}
+
+int reserve_debug_trap_vmid(struct device_queue_manager *dqm,
+				struct qcm_process_device *qpd)
+{
+	int r;
+	struct device *dev = dqm->dev->adev->dev;
+	int updated_vmid_mask;
+
+	if (dqm->sched_policy == KFD_SCHED_POLICY_NO_HWS) {
+		dev_err(dev, "Unsupported on sched_policy: %i\n", dqm->sched_policy);
+		return -EINVAL;
+	}
+
+	dqm_lock(dqm);
+
+	if (dqm->trap_debug_vmid != 0) {
+		dev_err(dev, "Trap debug id already reserved\n");
+		r = -EBUSY;
+		goto out_unlock;
+	}
+
+	r = unmap_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0,
+			USE_DEFAULT_GRACE_PERIOD, false);
+	if (r)
+		goto out_unlock;
+
+	updated_vmid_mask = dqm->dev->kfd->shared_resources.compute_vmid_bitmap;
+	updated_vmid_mask &= ~(1 << dqm->dev->vm_info.last_vmid_kfd);
+
+	dqm->dev->kfd->shared_resources.compute_vmid_bitmap = updated_vmid_mask;
+	dqm->trap_debug_vmid = dqm->dev->vm_info.last_vmid_kfd;
+	r = set_sched_resources(dqm);
+	if (r)
+		goto out_unlock;
+
+	r = map_queues_cpsch(dqm);
+	if (r)
+		goto out_unlock;
+
+	pr_debug("Reserved VMID for trap debug: %i\n", dqm->trap_debug_vmid);
+
+out_unlock:
+	dqm_unlock(dqm);
+	return r;
+}
+
+/*
+ * Releases vmid for the trap debugger
+ */
+int release_debug_trap_vmid(struct device_queue_manager *dqm,
+			struct qcm_process_device *qpd)
+{
+	struct device *dev = dqm->dev->adev->dev;
+	int r;
+	int updated_vmid_mask;
+	uint32_t trap_debug_vmid;
+
+	if (dqm->sched_policy == KFD_SCHED_POLICY_NO_HWS) {
+		dev_err(dev, "Unsupported on sched_policy: %i\n", dqm->sched_policy);
+		return -EINVAL;
+	}
+
+	dqm_lock(dqm);
+	trap_debug_vmid = dqm->trap_debug_vmid;
+	if (dqm->trap_debug_vmid == 0) {
+		dev_err(dev, "Trap debug id is not reserved\n");
+		r = -EINVAL;
+		goto out_unlock;
+	}
+
+	r = unmap_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0,
+			USE_DEFAULT_GRACE_PERIOD, false);
+	if (r)
+		goto out_unlock;
+
+	updated_vmid_mask = dqm->dev->kfd->shared_resources.compute_vmid_bitmap;
+	updated_vmid_mask |= (1 << dqm->dev->vm_info.last_vmid_kfd);
+
+	dqm->dev->kfd->shared_resources.compute_vmid_bitmap = updated_vmid_mask;
+	dqm->trap_debug_vmid = 0;
+	r = set_sched_resources(dqm);
+	if (r)
+		goto out_unlock;
+
+	r = map_queues_cpsch(dqm);
+	if (r)
+		goto out_unlock;
+
+	pr_debug("Released VMID for trap debug: %i\n", trap_debug_vmid);
+
+out_unlock:
+	dqm_unlock(dqm);
+	return r;
 }
 
+#define QUEUE_NOT_FOUND		-1
+/* invalidate queue operation in array */
+static void q_array_invalidate(uint32_t num_queues, uint32_t *queue_ids)
+{
+	int i;
+
+	for (i = 0; i < num_queues; i++)
+		queue_ids[i] |= KFD_DBG_QUEUE_INVALID_MASK;
+}
+
+/* find queue index in array */
+static int q_array_get_index(unsigned int queue_id,
+		uint32_t num_queues,
+		uint32_t *queue_ids)
+{
+	int i;
+
+	for (i = 0; i < num_queues; i++)
+		if (queue_id == (queue_ids[i] & ~KFD_DBG_QUEUE_INVALID_MASK))
+			return i;
+
+	return QUEUE_NOT_FOUND;
+}
+
+struct copy_context_work_handler_workarea {
+	struct work_struct copy_context_work;
+	struct kfd_process *p;
+};
+
+static void copy_context_work_handler(struct work_struct *work)
+{
+	struct copy_context_work_handler_workarea *workarea;
+	struct mqd_manager *mqd_mgr;
+	struct queue *q;
+	struct mm_struct *mm;
+	struct kfd_process *p;
+	uint32_t tmp_ctl_stack_used_size, tmp_save_area_used_size;
+	int i;
+
+	workarea = container_of(work,
+			struct copy_context_work_handler_workarea,
+			copy_context_work);
+
+	p = workarea->p;
+	mm = get_task_mm(p->lead_thread);
+
+	if (!mm)
+		return;
+
+	kthread_use_mm(mm);
+	for (i = 0; i < p->n_pdds; i++) {
+		struct kfd_process_device *pdd = p->pdds[i];
+		struct device_queue_manager *dqm = pdd->dev->dqm;
+		struct qcm_process_device *qpd = &pdd->qpd;
+
+		list_for_each_entry(q, &qpd->queues_list, list) {
+			if (q->properties.type != KFD_QUEUE_TYPE_COMPUTE)
+				continue;
+
+			mqd_mgr = dqm->mqd_mgrs[KFD_MQD_TYPE_CP];
+
+			/* We ignore the return value from get_wave_state
+			 * because
+			 * i) right now, it always returns 0, and
+			 * ii) if we hit an error, we would continue to the
+			 *      next queue anyway.
+			 */
+			mqd_mgr->get_wave_state(mqd_mgr,
+					q->mqd,
+					&q->properties,
+					(void __user *)	q->properties.ctx_save_restore_area_address,
+					&tmp_ctl_stack_used_size,
+					&tmp_save_area_used_size);
+		}
+	}
+	kthread_unuse_mm(mm);
+	mmput(mm);
+}
+
+static uint32_t *get_queue_ids(uint32_t num_queues, uint32_t *usr_queue_id_array)
+{
+	size_t array_size = num_queues * sizeof(uint32_t);
+
+	if (!usr_queue_id_array)
+		return NULL;
+
+	return memdup_user(usr_queue_id_array, array_size);
+}
+
+int resume_queues(struct kfd_process *p,
+		uint32_t num_queues,
+		uint32_t *usr_queue_id_array)
+{
+	uint32_t *queue_ids = NULL;
+	int total_resumed = 0;
+	int i;
+
+	if (usr_queue_id_array) {
+		queue_ids = get_queue_ids(num_queues, usr_queue_id_array);
+
+		if (IS_ERR(queue_ids))
+			return PTR_ERR(queue_ids);
+
+		/* mask all queues as invalid.  unmask per successful request */
+		q_array_invalidate(num_queues, queue_ids);
+	}
+
+	for (i = 0; i < p->n_pdds; i++) {
+		struct kfd_process_device *pdd = p->pdds[i];
+		struct device_queue_manager *dqm = pdd->dev->dqm;
+		struct device *dev = dqm->dev->adev->dev;
+		struct qcm_process_device *qpd = &pdd->qpd;
+		struct queue *q;
+		int r, per_device_resumed = 0;
+
+		dqm_lock(dqm);
+
+		/* unmask queues that resume or already resumed as valid */
+		list_for_each_entry(q, &qpd->queues_list, list) {
+			int q_idx = QUEUE_NOT_FOUND;
+
+			if (queue_ids)
+				q_idx = q_array_get_index(
+						q->properties.queue_id,
+						num_queues,
+						queue_ids);
+
+			if (!queue_ids || q_idx != QUEUE_NOT_FOUND) {
+				int err = resume_single_queue(dqm, &pdd->qpd, q);
+
+				if (queue_ids) {
+					if (!err) {
+						queue_ids[q_idx] &=
+							~KFD_DBG_QUEUE_INVALID_MASK;
+					} else {
+						queue_ids[q_idx] |=
+							KFD_DBG_QUEUE_ERROR_MASK;
+						break;
+					}
+				}
+
+				if (dqm->dev->kfd->shared_resources.enable_mes) {
+					wake_up_all(&dqm->destroy_wait);
+					if (!err)
+						total_resumed++;
+				} else {
+					per_device_resumed++;
+				}
+			}
+		}
+
+		if (!per_device_resumed) {
+			dqm_unlock(dqm);
+			continue;
+		}
+
+		r = execute_queues_cpsch(dqm,
+					KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES,
+					0,
+					USE_DEFAULT_GRACE_PERIOD);
+		if (r) {
+			dev_err(dev, "Failed to resume process queues\n");
+			if (queue_ids) {
+				list_for_each_entry(q, &qpd->queues_list, list) {
+					int q_idx = q_array_get_index(
+							q->properties.queue_id,
+							num_queues,
+							queue_ids);
+
+					/* mask queue as error on resume fail */
+					if (q_idx != QUEUE_NOT_FOUND)
+						queue_ids[q_idx] |=
+							KFD_DBG_QUEUE_ERROR_MASK;
+				}
+			}
+		} else {
+			wake_up_all(&dqm->destroy_wait);
+			total_resumed += per_device_resumed;
+		}
+
+		dqm_unlock(dqm);
+	}
+
+	if (queue_ids) {
+		if (copy_to_user((void __user *)usr_queue_id_array, queue_ids,
+				num_queues * sizeof(uint32_t)))
+			pr_err("copy_to_user failed on queue resume\n");
+
+		kfree(queue_ids);
+	}
+
+	return total_resumed;
+}
+
+int suspend_queues(struct kfd_process *p,
+			uint32_t num_queues,
+			uint32_t grace_period,
+			uint64_t exception_clear_mask,
+			uint32_t *usr_queue_id_array)
+{
+	uint32_t *queue_ids = get_queue_ids(num_queues, usr_queue_id_array);
+	int total_suspended = 0;
+	int i;
+
+	if (IS_ERR(queue_ids))
+		return PTR_ERR(queue_ids);
+
+	/* mask all queues as invalid.  umask on successful request */
+	q_array_invalidate(num_queues, queue_ids);
+
+	for (i = 0; i < p->n_pdds; i++) {
+		struct kfd_process_device *pdd = p->pdds[i];
+		struct device_queue_manager *dqm = pdd->dev->dqm;
+		struct device *dev = dqm->dev->adev->dev;
+		struct qcm_process_device *qpd = &pdd->qpd;
+		struct queue *q;
+		int r, per_device_suspended = 0;
+
+		mutex_lock(&p->event_mutex);
+		dqm_lock(dqm);
+
+		/* unmask queues that suspend or already suspended */
+		list_for_each_entry(q, &qpd->queues_list, list) {
+			int q_idx = q_array_get_index(q->properties.queue_id,
+							num_queues,
+							queue_ids);
+
+			if (q_idx != QUEUE_NOT_FOUND) {
+				int err = suspend_single_queue(dqm, pdd, q);
+				bool is_mes = dqm->dev->kfd->shared_resources.enable_mes;
+
+				if (!err) {
+					queue_ids[q_idx] &= ~KFD_DBG_QUEUE_INVALID_MASK;
+					if (exception_clear_mask && is_mes)
+						q->properties.exception_status &=
+							~exception_clear_mask;
+
+					if (is_mes)
+						total_suspended++;
+					else
+						per_device_suspended++;
+				} else if (err != -EBUSY) {
+					r = err;
+					queue_ids[q_idx] |= KFD_DBG_QUEUE_ERROR_MASK;
+					break;
+				}
+			}
+		}
+
+		if (!per_device_suspended) {
+			dqm_unlock(dqm);
+			mutex_unlock(&p->event_mutex);
+			if (total_suspended)
+				amdgpu_amdkfd_debug_mem_fence(dqm->dev->adev);
+			continue;
+		}
+
+		r = execute_queues_cpsch(dqm,
+			KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0,
+			grace_period);
+
+		if (r)
+			dev_err(dev, "Failed to suspend process queues.\n");
+		else
+			total_suspended += per_device_suspended;
+
+		list_for_each_entry(q, &qpd->queues_list, list) {
+			int q_idx = q_array_get_index(q->properties.queue_id,
+						num_queues, queue_ids);
+
+			if (q_idx == QUEUE_NOT_FOUND)
+				continue;
+
+			/* mask queue as error on suspend fail */
+			if (r)
+				queue_ids[q_idx] |= KFD_DBG_QUEUE_ERROR_MASK;
+			else if (exception_clear_mask)
+				q->properties.exception_status &=
+							~exception_clear_mask;
+		}
+
+		dqm_unlock(dqm);
+		mutex_unlock(&p->event_mutex);
+		amdgpu_device_flush_hdp(dqm->dev->adev, NULL);
+	}
+
+	if (total_suspended) {
+		struct copy_context_work_handler_workarea copy_context_worker;
+
+		INIT_WORK_ONSTACK(
+				&copy_context_worker.copy_context_work,
+				copy_context_work_handler);
+
+		copy_context_worker.p = p;
+
+		schedule_work(&copy_context_worker.copy_context_work);
+
+
+		flush_work(&copy_context_worker.copy_context_work);
+		destroy_work_on_stack(&copy_context_worker.copy_context_work);
+	}
+
+	if (copy_to_user((void __user *)usr_queue_id_array, queue_ids,
+			num_queues * sizeof(uint32_t)))
+		pr_err("copy_to_user failed on queue suspend\n");
+
+	kfree(queue_ids);
+
+	return total_suspended;
+}
+
+static uint32_t set_queue_type_for_user(struct queue_properties *q_props)
+{
+	switch (q_props->type) {
+	case KFD_QUEUE_TYPE_COMPUTE:
+		return q_props->format == KFD_QUEUE_FORMAT_PM4
+					? KFD_IOC_QUEUE_TYPE_COMPUTE
+					: KFD_IOC_QUEUE_TYPE_COMPUTE_AQL;
+	case KFD_QUEUE_TYPE_SDMA:
+		return KFD_IOC_QUEUE_TYPE_SDMA;
+	case KFD_QUEUE_TYPE_SDMA_XGMI:
+		return KFD_IOC_QUEUE_TYPE_SDMA_XGMI;
+	default:
+		WARN_ONCE(true, "queue type not recognized!");
+		return 0xffffffff;
+	};
+}
+
+void set_queue_snapshot_entry(struct queue *q,
+			      uint64_t exception_clear_mask,
+			      struct kfd_queue_snapshot_entry *qss_entry)
+{
+	qss_entry->ring_base_address = q->properties.queue_address;
+	qss_entry->write_pointer_address = (uint64_t)q->properties.write_ptr;
+	qss_entry->read_pointer_address = (uint64_t)q->properties.read_ptr;
+	qss_entry->ctx_save_restore_address =
+				q->properties.ctx_save_restore_area_address;
+	qss_entry->ctx_save_restore_area_size =
+				q->properties.ctx_save_restore_area_size;
+	qss_entry->exception_status = q->properties.exception_status;
+	qss_entry->queue_id = q->properties.queue_id;
+	qss_entry->gpu_id = q->device->id;
+	qss_entry->ring_size = (uint32_t)q->properties.queue_size;
+	qss_entry->queue_type = set_queue_type_for_user(&q->properties);
+	q->properties.exception_status &= ~exception_clear_mask;
+}
+
+int debug_lock_and_unmap(struct device_queue_manager *dqm)
+{
+	struct device *dev = dqm->dev->adev->dev;
+	int r;
+
+	if (dqm->sched_policy == KFD_SCHED_POLICY_NO_HWS) {
+		dev_err(dev, "Unsupported on sched_policy: %i\n", dqm->sched_policy);
+		return -EINVAL;
+	}
+
+	if (!kfd_dbg_is_per_vmid_supported(dqm->dev))
+		return 0;
+
+	dqm_lock(dqm);
+
+	r = unmap_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0, 0, false);
+	if (r)
+		dqm_unlock(dqm);
+
+	return r;
+}
+
+int debug_map_and_unlock(struct device_queue_manager *dqm)
+{
+	struct device *dev = dqm->dev->adev->dev;
+	int r;
+
+	if (dqm->sched_policy == KFD_SCHED_POLICY_NO_HWS) {
+		dev_err(dev, "Unsupported on sched_policy: %i\n", dqm->sched_policy);
+		return -EINVAL;
+	}
+
+	if (!kfd_dbg_is_per_vmid_supported(dqm->dev))
+		return 0;
+
+	r = map_queues_cpsch(dqm);
+
+	dqm_unlock(dqm);
+
+	return r;
+}
+
+int debug_refresh_runlist(struct device_queue_manager *dqm)
+{
+	int r = debug_lock_and_unmap(dqm);
+
+	if (r)
+		return r;
+
+	return debug_map_and_unlock(dqm);
+}
+
+bool kfd_dqm_is_queue_in_process(struct device_queue_manager *dqm,
+				 struct qcm_process_device *qpd,
+				 int doorbell_off, u32 *queue_format)
+{
+	struct queue *q;
+	bool r = false;
+
+	if (!queue_format)
+		return r;
+
+	dqm_lock(dqm);
+
+	list_for_each_entry(q, &qpd->queues_list, list) {
+		if (q->properties.doorbell_off == doorbell_off) {
+			*queue_format = q->properties.format;
+			r = true;
+			goto out;
+		}
+	}
+
+out:
+	dqm_unlock(dqm);
+	return r;
+}
 #if defined(CONFIG_DEBUG_FS)
 
 static void seq_reg_dump(struct seq_file *m,
@@ -2452,52 +3671,66 @@ static void seq_reg_dump(struct seq_file *m,
 int dqm_debugfs_hqds(struct seq_file *m, void *data)
 {
 	struct device_queue_manager *dqm = data;
+	uint32_t xcc_mask = dqm->dev->xcc_mask;
 	uint32_t (*dump)[2], n_regs;
 	int pipe, queue;
-	int r = 0;
+	int r = 0, xcc_id;
+	uint32_t sdma_engine_start;
 
 	if (!dqm->sched_running) {
 		seq_puts(m, " Device is stopped\n");
 		return 0;
 	}
 
-	r = dqm->dev->kfd2kgd->hqd_dump(dqm->dev->adev,
-					KFD_CIK_HIQ_PIPE, KFD_CIK_HIQ_QUEUE,
-					&dump, &n_regs);
-	if (!r) {
-		seq_printf(m, "  HIQ on MEC %d Pipe %d Queue %d\n",
-			   KFD_CIK_HIQ_PIPE/get_pipes_per_mec(dqm)+1,
-			   KFD_CIK_HIQ_PIPE%get_pipes_per_mec(dqm),
-			   KFD_CIK_HIQ_QUEUE);
-		seq_reg_dump(m, dump, n_regs);
+	for_each_inst(xcc_id, xcc_mask) {
+		r = dqm->dev->kfd2kgd->hqd_dump(dqm->dev->adev,
+						KFD_CIK_HIQ_PIPE,
+						KFD_CIK_HIQ_QUEUE, &dump,
+						&n_regs, xcc_id);
+		if (!r) {
+			seq_printf(
+				m,
+				"   Inst %d, HIQ on MEC %d Pipe %d Queue %d\n",
+				xcc_id,
+				KFD_CIK_HIQ_PIPE / get_pipes_per_mec(dqm) + 1,
+				KFD_CIK_HIQ_PIPE % get_pipes_per_mec(dqm),
+				KFD_CIK_HIQ_QUEUE);
+			seq_reg_dump(m, dump, n_regs);
 
-		kfree(dump);
-	}
+			kfree(dump);
+		}
 
-	for (pipe = 0; pipe < get_pipes_per_mec(dqm); pipe++) {
-		int pipe_offset = pipe * get_queues_per_pipe(dqm);
+		for (pipe = 0; pipe < get_pipes_per_mec(dqm); pipe++) {
+			int pipe_offset = pipe * get_queues_per_pipe(dqm);
 
-		for (queue = 0; queue < get_queues_per_pipe(dqm); queue++) {
-			if (!test_bit(pipe_offset + queue,
-				      dqm->dev->shared_resources.cp_queue_bitmap))
-				continue;
+			for (queue = 0; queue < get_queues_per_pipe(dqm); queue++) {
+				if (!test_bit(pipe_offset + queue,
+				      dqm->dev->kfd->shared_resources.cp_queue_bitmap))
+					continue;
 
-			r = dqm->dev->kfd2kgd->hqd_dump(
-				dqm->dev->adev, pipe, queue, &dump, &n_regs);
-			if (r)
-				break;
+				r = dqm->dev->kfd2kgd->hqd_dump(dqm->dev->adev,
+								pipe, queue,
+								&dump, &n_regs,
+								xcc_id);
+				if (r)
+					break;
 
-			seq_printf(m, "  CP Pipe %d, Queue %d\n",
-				  pipe, queue);
-			seq_reg_dump(m, dump, n_regs);
+				seq_printf(m,
+					   " Inst %d,  CP Pipe %d, Queue %d\n",
+					   xcc_id, pipe, queue);
+				seq_reg_dump(m, dump, n_regs);
 
-			kfree(dump);
+				kfree(dump);
+			}
 		}
 	}
 
-	for (pipe = 0; pipe < get_num_all_sdma_engines(dqm); pipe++) {
+	sdma_engine_start = dqm->dev->node_id * get_num_all_sdma_engines(dqm);
+	for (pipe = sdma_engine_start;
+	     pipe < (sdma_engine_start + get_num_all_sdma_engines(dqm));
+	     pipe++) {
 		for (queue = 0;
-		     queue < dqm->dev->device_info.num_sdma_queues_per_engine;
+		     queue < dqm->dev->kfd->device_info.num_sdma_queues_per_engine;
 		     queue++) {
 			r = dqm->dev->kfd2kgd->hqd_sdma_dump(
 				dqm->dev->adev, pipe, queue, &dump, &n_regs);
@@ -2526,7 +3759,8 @@ int dqm_debugfs_hang_hws(struct device_queue_manager *dqm)
 		return r;
 	}
 	dqm->active_runlist = true;
-	r = execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0);
+	r = execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES,
+				0, USE_DEFAULT_GRACE_PERIOD);
 	dqm_unlock(dqm);
 
 	return r;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h
index a537b9ef3e16..74a61b5b2f0b 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h
@@ -105,6 +105,12 @@ union GRBM_GFX_INDEX_BITS {
  * @uninitialize: Destroys all the device queue manager resources allocated in
  * initialize routine.
  *
+ * @halt: This routine unmaps queues from runlist and set halt status to true
+ * so no more queues will be mapped to runlist until unhalt.
+ *
+ * @unhalt: This routine unset halt status to flase and maps queues back to
+ * runlist.
+ *
  * @create_kernel_queue: Creates kernel queue. Used for debug queue.
  *
  * @destroy_kernel_queue: Destroys kernel queue. Used for debug queue.
@@ -151,8 +157,9 @@ struct device_queue_manager_ops {
 	int	(*initialize)(struct device_queue_manager *dqm);
 	int	(*start)(struct device_queue_manager *dqm);
 	int	(*stop)(struct device_queue_manager *dqm);
-	void	(*pre_reset)(struct device_queue_manager *dqm);
 	void	(*uninitialize)(struct device_queue_manager *dqm);
+	int     (*halt)(struct device_queue_manager *dqm);
+	int     (*unhalt)(struct device_queue_manager *dqm);
 	int	(*create_kernel_queue)(struct device_queue_manager *dqm,
 					struct kernel_queue *kq,
 					struct qcm_process_device *qpd);
@@ -166,7 +173,8 @@ struct device_queue_manager_ops {
 					   enum cache_policy default_policy,
 					   enum cache_policy alternate_policy,
 					   void __user *alternate_aperture_base,
-					   uint64_t alternate_aperture_size);
+					   uint64_t alternate_aperture_size,
+					   u32 misc_process_properties);
 
 	int (*process_termination)(struct device_queue_manager *dqm,
 			struct qcm_process_device *qpd);
@@ -202,12 +210,20 @@ struct device_queue_manager_asic_ops {
 					   enum cache_policy default_policy,
 					   enum cache_policy alternate_policy,
 					   void __user *alternate_aperture_base,
-					   uint64_t alternate_aperture_size);
+					   uint64_t alternate_aperture_size,
+					   u32 misc_process_properties);
 	void	(*init_sdma_vm)(struct device_queue_manager *dqm,
 				struct queue *q,
 				struct qcm_process_device *qpd);
 	struct mqd_manager *	(*mqd_manager_init)(enum KFD_MQD_TYPE type,
-				 struct kfd_dev *dev);
+				 struct kfd_node *dev);
+};
+
+struct dqm_detect_hang_info {
+	int pipe_id;
+	int queue_id;
+	int xcc_id;
+	uint64_t queue_address;
 };
 
 /**
@@ -228,7 +244,7 @@ struct device_queue_manager {
 
 	struct mqd_manager	*mqd_mgrs[KFD_MQD_TYPE_MAX];
 	struct packet_manager	packet_mgr;
-	struct kfd_dev		*dev;
+	struct kfd_node		*dev;
 	struct mutex		lock_hidden; /* use dqm_lock/unlock(dqm) */
 	struct list_head	queues;
 	unsigned int		saved_flags;
@@ -239,8 +255,8 @@ struct device_queue_manager {
 	unsigned int		total_queue_count;
 	unsigned int		next_pipe_to_allocate;
 	unsigned int		*allocated_queues;
-	uint64_t		sdma_bitmap;
-	uint64_t		xgmi_sdma_bitmap;
+	DECLARE_BITMAP(sdma_bitmap, KFD_MAX_SDMA_QUEUES);
+	DECLARE_BITMAP(xgmi_sdma_bitmap, KFD_MAX_SDMA_QUEUES);
 	/* the pasid mapping for each kfd vmid */
 	uint16_t		vmid_pasid[VMID_NUM];
 	uint64_t		pipelines_addr;
@@ -249,29 +265,40 @@ struct device_queue_manager {
 	struct kfd_mem_obj	*fence_mem;
 	bool			active_runlist;
 	int			sched_policy;
+	uint32_t		trap_debug_vmid;
 
 	/* hw exception  */
 	bool			is_hws_hang;
 	bool			is_resetting;
-	struct work_struct	hw_exception_work;
 	struct kfd_mem_obj	hiq_sdma_mqd;
 	bool			sched_running;
+	bool			sched_halt;
+
+	/* used for GFX 9.4.3 only */
+	uint32_t		current_logical_xcc_start;
+
+	uint32_t		wait_times;
+
+	wait_queue_head_t	destroy_wait;
+
+	/* for per-queue reset support */
+	struct dqm_detect_hang_info *detect_hang_info;
+	size_t detect_hang_info_size;
+	int detect_hang_count;
 };
 
 void device_queue_manager_init_cik(
 		struct device_queue_manager_asic_ops *asic_ops);
-void device_queue_manager_init_cik_hawaii(
-		struct device_queue_manager_asic_ops *asic_ops);
 void device_queue_manager_init_vi(
 		struct device_queue_manager_asic_ops *asic_ops);
-void device_queue_manager_init_vi_tonga(
-		struct device_queue_manager_asic_ops *asic_ops);
 void device_queue_manager_init_v9(
 		struct device_queue_manager_asic_ops *asic_ops);
-void device_queue_manager_init_v10_navi10(
+void device_queue_manager_init_v10(
 		struct device_queue_manager_asic_ops *asic_ops);
 void device_queue_manager_init_v11(
 		struct device_queue_manager_asic_ops *asic_ops);
+void device_queue_manager_init_v12(
+		struct device_queue_manager_asic_ops *asic_ops);
 void program_sh_mem_settings(struct device_queue_manager *dqm,
 					struct qcm_process_device *qpd);
 unsigned int get_cp_queues_num(struct device_queue_manager *dqm);
@@ -279,6 +306,27 @@ unsigned int get_queues_per_pipe(struct device_queue_manager *dqm);
 unsigned int get_pipes_per_mec(struct device_queue_manager *dqm);
 unsigned int get_num_sdma_queues(struct device_queue_manager *dqm);
 unsigned int get_num_xgmi_sdma_queues(struct device_queue_manager *dqm);
+int reserve_debug_trap_vmid(struct device_queue_manager *dqm,
+			struct qcm_process_device *qpd);
+int release_debug_trap_vmid(struct device_queue_manager *dqm,
+			struct qcm_process_device *qpd);
+int suspend_queues(struct kfd_process *p,
+			uint32_t num_queues,
+			uint32_t grace_period,
+			uint64_t exception_clear_mask,
+			uint32_t *usr_queue_id_array);
+int resume_queues(struct kfd_process *p,
+		uint32_t num_queues,
+		uint32_t *usr_queue_id_array);
+void set_queue_snapshot_entry(struct queue *q,
+			      uint64_t exception_clear_mask,
+			      struct kfd_queue_snapshot_entry *qss_entry);
+int debug_lock_and_unmap(struct device_queue_manager *dqm);
+int debug_map_and_unlock(struct device_queue_manager *dqm);
+int debug_refresh_runlist(struct device_queue_manager *dqm);
+bool kfd_dqm_is_queue_in_process(struct device_queue_manager *dqm,
+				 struct qcm_process_device *qpd,
+				 int doorbell_off, u32 *queue_format);
 
 static inline unsigned int get_sh_mem_bases_32(struct kfd_process_device *pdd)
 {
@@ -311,4 +359,14 @@ static inline int read_sdma_queue_counter(uint64_t __user *q_rptr, uint64_t *val
 	/* SDMA activity counter is stored at queue's RPTR + 0x8 location. */
 	return get_user(*val, q_rptr + 1);
 }
+
+static inline void update_dqm_wait_times(struct device_queue_manager *dqm)
+{
+	if (dqm->dev->kfd2kgd->get_iq_wait_times)
+		dqm->dev->kfd2kgd->get_iq_wait_times(dqm->dev->adev,
+					&dqm->wait_times,
+					ffs(dqm->dev->xcc_mask) - 1);
+}
+
+
 #endif /* KFD_DEVICE_QUEUE_MANAGER_H_ */
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c
index b1ab5b0775e1..0508ef5a41d7 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c
@@ -27,24 +27,29 @@
 #include "oss/oss_2_4_sh_mask.h"
 #include "gca/gfx_7_2_sh_mask.h"
 
+/*
+ * Low bits must be 0000/FFFF as required by HW, high bits must be 0 to
+ * stay in user mode.
+ */
+#define APE1_FIXED_BITS_MASK 0xFFFF80000000FFFFULL
+/* APE1 limit is inclusive and 64K aligned. */
+#define APE1_LIMIT_ALIGNMENT 0xFFFF
+
 static bool set_cache_memory_policy_cik(struct device_queue_manager *dqm,
 				   struct qcm_process_device *qpd,
 				   enum cache_policy default_policy,
 				   enum cache_policy alternate_policy,
 				   void __user *alternate_aperture_base,
-				   uint64_t alternate_aperture_size);
+				   uint64_t alternate_aperture_size,
+				   u32 misc_process_properties);
 static int update_qpd_cik(struct device_queue_manager *dqm,
-					struct qcm_process_device *qpd);
-static int update_qpd_cik_hawaii(struct device_queue_manager *dqm,
-					struct qcm_process_device *qpd);
-static void init_sdma_vm(struct device_queue_manager *dqm, struct queue *q,
-				struct qcm_process_device *qpd);
-static void init_sdma_vm_hawaii(struct device_queue_manager *dqm,
-				struct queue *q,
-				struct qcm_process_device *qpd);
+			  struct qcm_process_device *qpd);
+static void init_sdma_vm(struct device_queue_manager *dqm,
+			 struct queue *q,
+			 struct qcm_process_device *qpd);
 
 void device_queue_manager_init_cik(
-		struct device_queue_manager_asic_ops *asic_ops)
+	struct device_queue_manager_asic_ops *asic_ops)
 {
 	asic_ops->set_cache_memory_policy = set_cache_memory_policy_cik;
 	asic_ops->update_qpd = update_qpd_cik;
@@ -52,15 +57,6 @@ void device_queue_manager_init_cik(
 	asic_ops->mqd_manager_init = mqd_manager_init_cik;
 }
 
-void device_queue_manager_init_cik_hawaii(
-		struct device_queue_manager_asic_ops *asic_ops)
-{
-	asic_ops->set_cache_memory_policy = set_cache_memory_policy_cik;
-	asic_ops->update_qpd = update_qpd_cik_hawaii;
-	asic_ops->init_sdma_vm = init_sdma_vm_hawaii;
-	asic_ops->mqd_manager_init = mqd_manager_init_cik_hawaii;
-}
-
 static uint32_t compute_sh_mem_bases_64bit(unsigned int top_address_nybble)
 {
 	/* In 64-bit mode, we can only control the top 3 bits of the LDS,
@@ -93,10 +89,41 @@ static bool set_cache_memory_policy_cik(struct device_queue_manager *dqm,
 				   enum cache_policy default_policy,
 				   enum cache_policy alternate_policy,
 				   void __user *alternate_aperture_base,
-				   uint64_t alternate_aperture_size)
+				   uint64_t alternate_aperture_size,
+				   u32 misc_process_properties)
 {
 	uint32_t default_mtype;
 	uint32_t ape1_mtype;
+	unsigned int temp;
+	bool retval = true;
+
+	if (alternate_aperture_size == 0) {
+		/* base > limit disables APE1 */
+		qpd->sh_mem_ape1_base = 1;
+		qpd->sh_mem_ape1_limit = 0;
+	} else {
+		/*
+		 * In FSA64, APE1_Base[63:0] = { 16{SH_MEM_APE1_BASE[31]},
+		 *			SH_MEM_APE1_BASE[31:0], 0x0000 }
+		 * APE1_Limit[63:0] = { 16{SH_MEM_APE1_LIMIT[31]},
+		 *			SH_MEM_APE1_LIMIT[31:0], 0xFFFF }
+		 * Verify that the base and size parameters can be
+		 * represented in this format and convert them.
+		 * Additionally restrict APE1 to user-mode addresses.
+		 */
+
+		uint64_t base = (uintptr_t)alternate_aperture_base;
+		uint64_t limit = base + alternate_aperture_size - 1;
+
+		if (limit <= base || (base & APE1_FIXED_BITS_MASK) != 0 ||
+		   (limit & APE1_FIXED_BITS_MASK) != APE1_LIMIT_ALIGNMENT) {
+			retval = false;
+			goto out;
+		}
+
+		qpd->sh_mem_ape1_base = base >> 16;
+		qpd->sh_mem_ape1_limit = limit >> 16;
+	}
 
 	default_mtype = (default_policy == cache_policy_coherent) ?
 			MTYPE_NONCACHED :
@@ -110,93 +137,28 @@ static bool set_cache_memory_policy_cik(struct device_queue_manager *dqm,
 			| ALIGNMENT_MODE(SH_MEM_ALIGNMENT_MODE_UNALIGNED)
 			| DEFAULT_MTYPE(default_mtype)
 			| APE1_MTYPE(ape1_mtype);
-
-	return true;
-}
-
-static int update_qpd_cik(struct device_queue_manager *dqm,
-		struct qcm_process_device *qpd)
-{
-	struct kfd_process_device *pdd;
-	unsigned int temp;
-
-	pdd = qpd_to_pdd(qpd);
-
-	/* check if sh_mem_config register already configured */
-	if (qpd->sh_mem_config == 0) {
-		qpd->sh_mem_config =
-			ALIGNMENT_MODE(SH_MEM_ALIGNMENT_MODE_UNALIGNED) |
-			DEFAULT_MTYPE(MTYPE_NONCACHED) |
-			APE1_MTYPE(MTYPE_NONCACHED);
-		qpd->sh_mem_ape1_limit = 0;
-		qpd->sh_mem_ape1_base = 0;
-	}
-
-	if (qpd->pqm->process->is_32bit_user_mode) {
-		temp = get_sh_mem_bases_32(pdd);
-		qpd->sh_mem_bases = SHARED_BASE(temp);
-		qpd->sh_mem_config |= PTR32;
-	} else {
-		temp = get_sh_mem_bases_nybble_64(pdd);
-		qpd->sh_mem_bases = compute_sh_mem_bases_64bit(temp);
-		qpd->sh_mem_config |= 1  << SH_MEM_CONFIG__PRIVATE_ATC__SHIFT;
-	}
-
-	pr_debug("is32bit process: %d sh_mem_bases nybble: 0x%X and register 0x%X\n",
-		qpd->pqm->process->is_32bit_user_mode, temp, qpd->sh_mem_bases);
-
-	return 0;
-}
-
-static int update_qpd_cik_hawaii(struct device_queue_manager *dqm,
-		struct qcm_process_device *qpd)
-{
-	struct kfd_process_device *pdd;
-	unsigned int temp;
-
-	pdd = qpd_to_pdd(qpd);
-
-	/* check if sh_mem_config register already configured */
-	if (qpd->sh_mem_config == 0) {
-		qpd->sh_mem_config =
-			ALIGNMENT_MODE(SH_MEM_ALIGNMENT_MODE_UNALIGNED) |
-			DEFAULT_MTYPE(MTYPE_NONCACHED) |
-			APE1_MTYPE(MTYPE_NONCACHED);
-		qpd->sh_mem_ape1_limit = 0;
-		qpd->sh_mem_ape1_base = 0;
-	}
-
 	/* On dGPU we're always in GPUVM64 addressing mode with 64-bit
 	 * aperture addresses.
 	 */
-	temp = get_sh_mem_bases_nybble_64(pdd);
+	temp = get_sh_mem_bases_nybble_64(qpd_to_pdd(qpd));
 	qpd->sh_mem_bases = compute_sh_mem_bases_64bit(temp);
 
 	pr_debug("is32bit process: %d sh_mem_bases nybble: 0x%X and register 0x%X\n",
 		qpd->pqm->process->is_32bit_user_mode, temp, qpd->sh_mem_bases);
 
-	return 0;
+out:
+	return retval;
 }
 
-static void init_sdma_vm(struct device_queue_manager *dqm, struct queue *q,
-				struct qcm_process_device *qpd)
+static int update_qpd_cik(struct device_queue_manager *dqm,
+			  struct qcm_process_device *qpd)
 {
-	uint32_t value = (1 << SDMA0_RLC0_VIRTUAL_ADDR__ATC__SHIFT);
-
-	if (q->process->is_32bit_user_mode)
-		value |= (1 << SDMA0_RLC0_VIRTUAL_ADDR__PTR32__SHIFT) |
-				get_sh_mem_bases_32(qpd_to_pdd(qpd));
-	else
-		value |= ((get_sh_mem_bases_nybble_64(qpd_to_pdd(qpd))) <<
-				SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE__SHIFT) &
-				SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE_MASK;
-
-	q->properties.sdma_vm_addr = value;
+	return 0;
 }
 
-static void init_sdma_vm_hawaii(struct device_queue_manager *dqm,
-				struct queue *q,
-				struct qcm_process_device *qpd)
+static void init_sdma_vm(struct device_queue_manager *dqm,
+			 struct queue *q,
+			 struct qcm_process_device *qpd)
 {
 	/* On dGPU we're always in GPUVM64 addressing mode with 64-bit
 	 * aperture addresses.
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v10.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v10.c
index f1a1f5753e65..ba6e3d747ccd 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v10.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v10.c
@@ -31,10 +31,18 @@ static int update_qpd_v10(struct device_queue_manager *dqm,
 			 struct qcm_process_device *qpd);
 static void init_sdma_vm_v10(struct device_queue_manager *dqm, struct queue *q,
 			    struct qcm_process_device *qpd);
+static bool set_cache_memory_policy_v10(struct device_queue_manager *dqm,
+				   struct qcm_process_device *qpd,
+				   enum cache_policy default_policy,
+				   enum cache_policy alternate_policy,
+				   void __user *alternate_aperture_base,
+				   uint64_t alternate_aperture_size,
+				   u32 misc_process_properties);
 
-void device_queue_manager_init_v10_navi10(
+void device_queue_manager_init_v10(
 	struct device_queue_manager_asic_ops *asic_ops)
 {
+	asic_ops->set_cache_memory_policy = set_cache_memory_policy_v10;
 	asic_ops->update_qpd = update_qpd_v10;
 	asic_ops->init_sdma_vm = init_sdma_vm_v10;
 	asic_ops->mqd_manager_init = mqd_manager_init_v10;
@@ -49,27 +57,28 @@ static uint32_t compute_sh_mem_bases_64bit(struct kfd_process_device *pdd)
 		private_base;
 }
 
-static int update_qpd_v10(struct device_queue_manager *dqm,
-			 struct qcm_process_device *qpd)
+static bool set_cache_memory_policy_v10(struct device_queue_manager *dqm,
+				   struct qcm_process_device *qpd,
+				   enum cache_policy default_policy,
+				   enum cache_policy alternate_policy,
+				   void __user *alternate_aperture_base,
+				   uint64_t alternate_aperture_size,
+				   u32 misc_process_properties)
 {
-	struct kfd_process_device *pdd;
-
-	pdd = qpd_to_pdd(qpd);
-
-	/* check if sh_mem_config register already configured */
-	if (qpd->sh_mem_config == 0) {
-		qpd->sh_mem_config =
-			(SH_MEM_ALIGNMENT_MODE_UNALIGNED <<
-				SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT) |
-			(3 << SH_MEM_CONFIG__INITIAL_INST_PREFETCH__SHIFT);
-		qpd->sh_mem_ape1_limit = 0;
-		qpd->sh_mem_ape1_base = 0;
-	}
-
-	qpd->sh_mem_bases = compute_sh_mem_bases_64bit(pdd);
+	qpd->sh_mem_config = (SH_MEM_ALIGNMENT_MODE_UNALIGNED <<
+			      SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT) |
+			      (3 << SH_MEM_CONFIG__INITIAL_INST_PREFETCH__SHIFT);
+	qpd->sh_mem_ape1_limit = 0;
+	qpd->sh_mem_ape1_base = 0;
+	qpd->sh_mem_bases = compute_sh_mem_bases_64bit(qpd_to_pdd(qpd));
 
 	pr_debug("sh_mem_bases 0x%X\n", qpd->sh_mem_bases);
+	return true;
+}
 
+static int update_qpd_v10(struct device_queue_manager *dqm,
+			 struct qcm_process_device *qpd)
+{
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v11.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v11.c
index 2e129da7acb4..8b447d04558f 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v11.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v11.c
@@ -30,10 +30,18 @@ static int update_qpd_v11(struct device_queue_manager *dqm,
 			 struct qcm_process_device *qpd);
 static void init_sdma_vm_v11(struct device_queue_manager *dqm, struct queue *q,
 			    struct qcm_process_device *qpd);
+static bool set_cache_memory_policy_v11(struct device_queue_manager *dqm,
+				   struct qcm_process_device *qpd,
+				   enum cache_policy default_policy,
+				   enum cache_policy alternate_policy,
+				   void __user *alternate_aperture_base,
+				   uint64_t alternate_aperture_size,
+				   u32 misc_process_properties);
 
 void device_queue_manager_init_v11(
 	struct device_queue_manager_asic_ops *asic_ops)
 {
+	asic_ops->set_cache_memory_policy = set_cache_memory_policy_v11;
 	asic_ops->update_qpd = update_qpd_v11;
 	asic_ops->init_sdma_vm = init_sdma_vm_v11;
 	asic_ops->mqd_manager_init = mqd_manager_init_v11;
@@ -48,28 +56,29 @@ static uint32_t compute_sh_mem_bases_64bit(struct kfd_process_device *pdd)
 		private_base;
 }
 
-static int update_qpd_v11(struct device_queue_manager *dqm,
-			 struct qcm_process_device *qpd)
+static bool set_cache_memory_policy_v11(struct device_queue_manager *dqm,
+				   struct qcm_process_device *qpd,
+				   enum cache_policy default_policy,
+				   enum cache_policy alternate_policy,
+				   void __user *alternate_aperture_base,
+				   uint64_t alternate_aperture_size,
+				   u32 misc_process_properties)
 {
-	struct kfd_process_device *pdd;
-
-	pdd = qpd_to_pdd(qpd);
-
-	/* check if sh_mem_config register already configured */
-	if (qpd->sh_mem_config == 0) {
-		qpd->sh_mem_config =
-			(SH_MEM_ALIGNMENT_MODE_UNALIGNED <<
-				SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT) |
-			(3 << SH_MEM_CONFIG__INITIAL_INST_PREFETCH__SHIFT);
-
-		qpd->sh_mem_ape1_limit = 0;
-		qpd->sh_mem_ape1_base = 0;
-	}
+	qpd->sh_mem_config = (SH_MEM_ALIGNMENT_MODE_UNALIGNED <<
+			      SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT) |
+			      (3 << SH_MEM_CONFIG__INITIAL_INST_PREFETCH__SHIFT);
 
-	qpd->sh_mem_bases = compute_sh_mem_bases_64bit(pdd);
+	qpd->sh_mem_ape1_limit = 0;
+	qpd->sh_mem_ape1_base = 0;
+	qpd->sh_mem_bases = compute_sh_mem_bases_64bit(qpd_to_pdd(qpd));
 
 	pr_debug("sh_mem_bases 0x%X\n", qpd->sh_mem_bases);
+	return true;
+}
 
+static int update_qpd_v11(struct device_queue_manager *dqm,
+			 struct qcm_process_device *qpd)
+{
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v12.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v12.c
new file mode 100644
index 000000000000..3550da3a46f9
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v12.c
@@ -0,0 +1,90 @@
+// SPDX-License-Identifier: GPL-2.0 OR MIT
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "kfd_device_queue_manager.h"
+#include "gc/gc_12_0_0_sh_mask.h"
+#include "soc24_enum.h"
+
+static int update_qpd_v12(struct device_queue_manager *dqm,
+			 struct qcm_process_device *qpd);
+static void init_sdma_vm_v12(struct device_queue_manager *dqm, struct queue *q,
+			    struct qcm_process_device *qpd);
+static bool set_cache_memory_policy_v12(struct device_queue_manager *dqm,
+				   struct qcm_process_device *qpd,
+				   enum cache_policy default_policy,
+				   enum cache_policy alternate_policy,
+				   void __user *alternate_aperture_base,
+				   uint64_t alternate_aperture_size,
+				   u32 misc_process_properties);
+
+void device_queue_manager_init_v12(
+	struct device_queue_manager_asic_ops *asic_ops)
+{
+	asic_ops->set_cache_memory_policy = set_cache_memory_policy_v12;
+	asic_ops->update_qpd = update_qpd_v12;
+	asic_ops->init_sdma_vm = init_sdma_vm_v12;
+	asic_ops->mqd_manager_init = mqd_manager_init_v12;
+}
+
+static uint32_t compute_sh_mem_bases_64bit(struct kfd_process_device *pdd)
+{
+	uint32_t shared_base = pdd->lds_base >> 48;
+	uint32_t private_base = pdd->scratch_base >> 48;
+
+	return (shared_base << SH_MEM_BASES__SHARED_BASE__SHIFT) |
+		private_base;
+}
+
+static bool set_cache_memory_policy_v12(struct device_queue_manager *dqm,
+				   struct qcm_process_device *qpd,
+				   enum cache_policy default_policy,
+				   enum cache_policy alternate_policy,
+				   void __user *alternate_aperture_base,
+				   uint64_t alternate_aperture_size,
+				   u32 misc_process_properties)
+{
+	qpd->sh_mem_config = (SH_MEM_ALIGNMENT_MODE_UNALIGNED <<
+			      SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT) |
+			      (3 << SH_MEM_CONFIG__INITIAL_INST_PREFETCH__SHIFT);
+
+	qpd->sh_mem_ape1_limit = 0;
+	qpd->sh_mem_ape1_base = 0;
+	qpd->sh_mem_bases = compute_sh_mem_bases_64bit(qpd_to_pdd(qpd));
+
+	pr_debug("sh_mem_bases 0x%X\n", qpd->sh_mem_bases);
+	return true;
+}
+
+static int update_qpd_v12(struct device_queue_manager *dqm,
+			 struct qcm_process_device *qpd)
+{
+	return 0;
+}
+
+static void init_sdma_vm_v12(struct device_queue_manager *dqm, struct queue *q,
+			    struct qcm_process_device *qpd)
+{
+	/* Not needed on SDMAv4 onwards any more */
+	q->properties.sdma_vm_addr = 0;
+}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c
index 8b2dd2670ab7..9fcc8c6e57b7 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c
@@ -24,18 +24,24 @@
 
 #include "kfd_device_queue_manager.h"
 #include "vega10_enum.h"
-#include "gc/gc_9_0_offset.h"
-#include "gc/gc_9_0_sh_mask.h"
-#include "sdma0/sdma0_4_0_sh_mask.h"
+#include "gc/gc_9_4_3_sh_mask.h"
 
 static int update_qpd_v9(struct device_queue_manager *dqm,
 			 struct qcm_process_device *qpd);
 static void init_sdma_vm_v9(struct device_queue_manager *dqm, struct queue *q,
 			    struct qcm_process_device *qpd);
+static bool set_cache_memory_policy_v9(struct device_queue_manager *dqm,
+				   struct qcm_process_device *qpd,
+				   enum cache_policy default_policy,
+				   enum cache_policy alternate_policy,
+				   void __user *alternate_aperture_base,
+				   uint64_t alternate_aperture_size,
+				   u32 misc_process_properties);
 
 void device_queue_manager_init_v9(
 	struct device_queue_manager_asic_ops *asic_ops)
 {
+	asic_ops->set_cache_memory_policy = set_cache_memory_policy_v9;
 	asic_ops->update_qpd = update_qpd_v9;
 	asic_ops->init_sdma_vm = init_sdma_vm_v9;
 	asic_ops->mqd_manager_init = mqd_manager_init_v9;
@@ -50,10 +56,42 @@ static uint32_t compute_sh_mem_bases_64bit(struct kfd_process_device *pdd)
 		private_base;
 }
 
+static bool set_cache_memory_policy_v9(struct device_queue_manager *dqm,
+				   struct qcm_process_device *qpd,
+				   enum cache_policy default_policy,
+				   enum cache_policy alternate_policy,
+				   void __user *alternate_aperture_base,
+				   uint64_t alternate_aperture_size,
+				   u32 misc_process_properties)
+{
+	qpd->sh_mem_config = SH_MEM_ALIGNMENT_MODE_UNALIGNED <<
+				SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT;
+
+	if (dqm->dev->kfd->noretry)
+		qpd->sh_mem_config |= 1 << SH_MEM_CONFIG__RETRY_DISABLE__SHIFT;
+
+	if (KFD_GC_VERSION(dqm->dev->kfd) == IP_VERSION(9, 4, 3) ||
+		KFD_GC_VERSION(dqm->dev->kfd) == IP_VERSION(9, 4, 4))
+		qpd->sh_mem_config |= (1 << SH_MEM_CONFIG__F8_MODE__SHIFT);
+
+	if (KFD_GC_VERSION(dqm->dev->kfd) == IP_VERSION(9, 5, 0)) {
+		if (misc_process_properties & KFD_PROC_FLAG_MFMA_HIGH_PRECISION)
+			qpd->sh_mem_config |= 1 << SH_MEM_CONFIG__PRECISION_MODE__SHIFT;
+	}
+
+	qpd->sh_mem_ape1_limit = 0;
+	qpd->sh_mem_ape1_base = 0;
+	qpd->sh_mem_bases = compute_sh_mem_bases_64bit(qpd_to_pdd(qpd));
+
+	pr_debug("sh_mem_bases 0x%X sh_mem_config 0x%X\n", qpd->sh_mem_bases,
+		 qpd->sh_mem_config);
+	return true;
+}
+
 static int update_qpd_v9(struct device_queue_manager *dqm,
 			 struct qcm_process_device *qpd)
 {
-	struct kfd_process_device *pdd;
+	struct kfd_process_device *pdd = qpd_to_pdd(qpd);
 
 	pdd = qpd_to_pdd(qpd);
 
@@ -62,9 +100,14 @@ static int update_qpd_v9(struct device_queue_manager *dqm,
 		qpd->sh_mem_config = SH_MEM_ALIGNMENT_MODE_UNALIGNED <<
 					SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT;
 
-		if (dqm->dev->noretry && !dqm->dev->use_iommu_v2)
+		if (dqm->dev->kfd->noretry)
 			qpd->sh_mem_config |= 1 << SH_MEM_CONFIG__RETRY_DISABLE__SHIFT;
 
+		if (KFD_GC_VERSION(dqm->dev->kfd) == IP_VERSION(9, 4, 3) ||
+		    KFD_GC_VERSION(dqm->dev->kfd) == IP_VERSION(9, 4, 4))
+			qpd->sh_mem_config |=
+				(1 << SH_MEM_CONFIG__F8_MODE__SHIFT);
+
 		qpd->sh_mem_ape1_limit = 0;
 		qpd->sh_mem_ape1_base = 0;
 	}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c
index d7d45832df0f..dad83356e976 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c
@@ -27,30 +27,29 @@
 #include "gca/gfx_8_0_sh_mask.h"
 #include "oss/oss_3_0_sh_mask.h"
 
+/*
+ * Low bits must be 0000/FFFF as required by HW, high bits must be 0 to
+ * stay in user mode.
+ */
+#define APE1_FIXED_BITS_MASK 0xFFFF80000000FFFFULL
+/* APE1 limit is inclusive and 64K aligned. */
+#define APE1_LIMIT_ALIGNMENT 0xFFFF
+
 static bool set_cache_memory_policy_vi(struct device_queue_manager *dqm,
-				   struct qcm_process_device *qpd,
-				   enum cache_policy default_policy,
-				   enum cache_policy alternate_policy,
-				   void __user *alternate_aperture_base,
-				   uint64_t alternate_aperture_size);
-static bool set_cache_memory_policy_vi_tonga(struct device_queue_manager *dqm,
-			struct qcm_process_device *qpd,
-			enum cache_policy default_policy,
-			enum cache_policy alternate_policy,
-			void __user *alternate_aperture_base,
-			uint64_t alternate_aperture_size);
+				       struct qcm_process_device *qpd,
+				       enum cache_policy default_policy,
+				       enum cache_policy alternate_policy,
+				       void __user *alternate_aperture_base,
+				       uint64_t alternate_aperture_size,
+				       u32 misc_process_properties);
 static int update_qpd_vi(struct device_queue_manager *dqm,
-					struct qcm_process_device *qpd);
-static int update_qpd_vi_tonga(struct device_queue_manager *dqm,
-			struct qcm_process_device *qpd);
-static void init_sdma_vm(struct device_queue_manager *dqm, struct queue *q,
-				struct qcm_process_device *qpd);
-static void init_sdma_vm_tonga(struct device_queue_manager *dqm,
-			struct queue *q,
-			struct qcm_process_device *qpd);
+			 struct qcm_process_device *qpd);
+static void init_sdma_vm(struct device_queue_manager *dqm,
+			 struct queue *q,
+			 struct qcm_process_device *qpd);
 
 void device_queue_manager_init_vi(
-		struct device_queue_manager_asic_ops *asic_ops)
+	struct device_queue_manager_asic_ops *asic_ops)
 {
 	asic_ops->set_cache_memory_policy = set_cache_memory_policy_vi;
 	asic_ops->update_qpd = update_qpd_vi;
@@ -58,15 +57,6 @@ void device_queue_manager_init_vi(
 	asic_ops->mqd_manager_init = mqd_manager_init_vi;
 }
 
-void device_queue_manager_init_vi_tonga(
-		struct device_queue_manager_asic_ops *asic_ops)
-{
-	asic_ops->set_cache_memory_policy = set_cache_memory_policy_vi_tonga;
-	asic_ops->update_qpd = update_qpd_vi_tonga;
-	asic_ops->init_sdma_vm = init_sdma_vm_tonga;
-	asic_ops->mqd_manager_init = mqd_manager_init_vi_tonga;
-}
-
 static uint32_t compute_sh_mem_bases_64bit(unsigned int top_address_nybble)
 {
 	/* In 64-bit mode, we can only control the top 3 bits of the LDS,
@@ -96,43 +86,45 @@ static uint32_t compute_sh_mem_bases_64bit(unsigned int top_address_nybble)
 }
 
 static bool set_cache_memory_policy_vi(struct device_queue_manager *dqm,
-				   struct qcm_process_device *qpd,
-				   enum cache_policy default_policy,
-				   enum cache_policy alternate_policy,
-				   void __user *alternate_aperture_base,
-				   uint64_t alternate_aperture_size)
-{
-	uint32_t default_mtype;
-	uint32_t ape1_mtype;
-
-	default_mtype = (default_policy == cache_policy_coherent) ?
-			MTYPE_CC :
-			MTYPE_NC;
-
-	ape1_mtype = (alternate_policy == cache_policy_coherent) ?
-			MTYPE_CC :
-			MTYPE_NC;
-
-	qpd->sh_mem_config = (qpd->sh_mem_config &
-			SH_MEM_CONFIG__ADDRESS_MODE_MASK) |
-		SH_MEM_ALIGNMENT_MODE_UNALIGNED <<
-				SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT |
-		default_mtype << SH_MEM_CONFIG__DEFAULT_MTYPE__SHIFT |
-		ape1_mtype << SH_MEM_CONFIG__APE1_MTYPE__SHIFT |
-		SH_MEM_CONFIG__PRIVATE_ATC_MASK;
-
-	return true;
-}
-
-static bool set_cache_memory_policy_vi_tonga(struct device_queue_manager *dqm,
 		struct qcm_process_device *qpd,
 		enum cache_policy default_policy,
 		enum cache_policy alternate_policy,
 		void __user *alternate_aperture_base,
-		uint64_t alternate_aperture_size)
+		uint64_t alternate_aperture_size,
+		u32 misc_process_properties)
 {
 	uint32_t default_mtype;
 	uint32_t ape1_mtype;
+	unsigned int temp;
+	bool retval = true;
+
+	if (alternate_aperture_size == 0) {
+		/* base > limit disables APE1 */
+		qpd->sh_mem_ape1_base = 1;
+		qpd->sh_mem_ape1_limit = 0;
+	} else {
+		/*
+		 * In FSA64, APE1_Base[63:0] = { 16{SH_MEM_APE1_BASE[31]},
+		 *			SH_MEM_APE1_BASE[31:0], 0x0000 }
+		 * APE1_Limit[63:0] = { 16{SH_MEM_APE1_LIMIT[31]},
+		 *			SH_MEM_APE1_LIMIT[31:0], 0xFFFF }
+		 * Verify that the base and size parameters can be
+		 * represented in this format and convert them.
+		 * Additionally restrict APE1 to user-mode addresses.
+		 */
+
+		uint64_t base = (uintptr_t)alternate_aperture_base;
+		uint64_t limit = base + alternate_aperture_size - 1;
+
+		if (limit <= base || (base & APE1_FIXED_BITS_MASK) != 0 ||
+		   (limit & APE1_FIXED_BITS_MASK) != APE1_LIMIT_ALIGNMENT) {
+			retval = false;
+			goto out;
+		}
+
+		qpd->sh_mem_ape1_base = base >> 16;
+		qpd->sh_mem_ape1_limit = limit >> 16;
+	}
 
 	default_mtype = (default_policy == cache_policy_coherent) ?
 			MTYPE_UC :
@@ -148,103 +140,27 @@ static bool set_cache_memory_policy_vi_tonga(struct device_queue_manager *dqm,
 			default_mtype << SH_MEM_CONFIG__DEFAULT_MTYPE__SHIFT |
 			ape1_mtype << SH_MEM_CONFIG__APE1_MTYPE__SHIFT;
 
-	return true;
-}
-
-static int update_qpd_vi(struct device_queue_manager *dqm,
-					struct qcm_process_device *qpd)
-{
-	struct kfd_process_device *pdd;
-	unsigned int temp;
-
-	pdd = qpd_to_pdd(qpd);
-
-	/* check if sh_mem_config register already configured */
-	if (qpd->sh_mem_config == 0) {
-		qpd->sh_mem_config =
-			SH_MEM_ALIGNMENT_MODE_UNALIGNED <<
-				SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT |
-			MTYPE_CC << SH_MEM_CONFIG__DEFAULT_MTYPE__SHIFT |
-			MTYPE_CC << SH_MEM_CONFIG__APE1_MTYPE__SHIFT |
-			SH_MEM_CONFIG__PRIVATE_ATC_MASK;
-
-		qpd->sh_mem_ape1_limit = 0;
-		qpd->sh_mem_ape1_base = 0;
-	}
-
-	if (qpd->pqm->process->is_32bit_user_mode) {
-		temp = get_sh_mem_bases_32(pdd);
-		qpd->sh_mem_bases = temp << SH_MEM_BASES__SHARED_BASE__SHIFT;
-		qpd->sh_mem_config |= SH_MEM_ADDRESS_MODE_HSA32 <<
-					SH_MEM_CONFIG__ADDRESS_MODE__SHIFT;
-	} else {
-		temp = get_sh_mem_bases_nybble_64(pdd);
-		qpd->sh_mem_bases = compute_sh_mem_bases_64bit(temp);
-		qpd->sh_mem_config |= SH_MEM_ADDRESS_MODE_HSA64 <<
-			SH_MEM_CONFIG__ADDRESS_MODE__SHIFT;
-		qpd->sh_mem_config |= 1  <<
-			SH_MEM_CONFIG__PRIVATE_ATC__SHIFT;
-	}
-
-	pr_debug("is32bit process: %d sh_mem_bases nybble: 0x%X and register 0x%X\n",
-		qpd->pqm->process->is_32bit_user_mode, temp, qpd->sh_mem_bases);
-
-	return 0;
-}
-
-static int update_qpd_vi_tonga(struct device_queue_manager *dqm,
-			struct qcm_process_device *qpd)
-{
-	struct kfd_process_device *pdd;
-	unsigned int temp;
-
-	pdd = qpd_to_pdd(qpd);
-
-	/* check if sh_mem_config register already configured */
-	if (qpd->sh_mem_config == 0) {
-		qpd->sh_mem_config =
-				SH_MEM_ALIGNMENT_MODE_UNALIGNED <<
-					SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT |
-				MTYPE_UC <<
-					SH_MEM_CONFIG__DEFAULT_MTYPE__SHIFT |
-				MTYPE_UC <<
-					SH_MEM_CONFIG__APE1_MTYPE__SHIFT;
-
-		qpd->sh_mem_ape1_limit = 0;
-		qpd->sh_mem_ape1_base = 0;
-	}
-
 	/* On dGPU we're always in GPUVM64 addressing mode with 64-bit
 	 * aperture addresses.
 	 */
-	temp = get_sh_mem_bases_nybble_64(pdd);
+	temp = get_sh_mem_bases_nybble_64(qpd_to_pdd(qpd));
 	qpd->sh_mem_bases = compute_sh_mem_bases_64bit(temp);
 
 	pr_debug("sh_mem_bases nybble: 0x%X and register 0x%X\n",
 		temp, qpd->sh_mem_bases);
-
-	return 0;
+out:
+	return retval;
 }
 
-static void init_sdma_vm(struct device_queue_manager *dqm, struct queue *q,
-				struct qcm_process_device *qpd)
+static int update_qpd_vi(struct device_queue_manager *dqm,
+			 struct qcm_process_device *qpd)
 {
-	uint32_t value = (1 << SDMA0_RLC0_VIRTUAL_ADDR__ATC__SHIFT);
-
-	if (q->process->is_32bit_user_mode)
-		value |= (1 << SDMA0_RLC0_VIRTUAL_ADDR__PTR32__SHIFT) |
-				get_sh_mem_bases_32(qpd_to_pdd(qpd));
-	else
-		value |= ((get_sh_mem_bases_nybble_64(qpd_to_pdd(qpd))) <<
-				SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE__SHIFT) &
-				SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE_MASK;
-
-	q->properties.sdma_vm_addr = value;
+	return 0;
 }
 
-static void init_sdma_vm_tonga(struct device_queue_manager *dqm,
-			struct queue *q,
-			struct qcm_process_device *qpd)
+static void init_sdma_vm(struct device_queue_manager *dqm,
+			 struct queue *q,
+			 struct qcm_process_device *qpd)
 {
 	/* On dGPU we're always in GPUVM64 addressing mode with 64-bit
 	 * aperture addresses.
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
index 38c9e1ca6691..05c74887fd6f 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
@@ -61,84 +61,49 @@ size_t kfd_doorbell_process_slice(struct kfd_dev *kfd)
 /* Doorbell calculations for device init. */
 int kfd_doorbell_init(struct kfd_dev *kfd)
 {
-	size_t doorbell_start_offset;
-	size_t doorbell_aperture_size;
-	size_t doorbell_process_limit;
+	int size = PAGE_SIZE;
+	int r;
 
 	/*
-	 * With MES enabled, just set the doorbell base as it is needed
-	 * to calculate doorbell physical address.
-	 */
-	if (kfd->shared_resources.enable_mes) {
-		kfd->doorbell_base =
-			kfd->shared_resources.doorbell_physical_address;
-		return 0;
-	}
-
-	/*
-	 * We start with calculations in bytes because the input data might
-	 * only be byte-aligned.
-	 * Only after we have done the rounding can we assume any alignment.
+	 * Todo: KFD kernel level operations need only one doorbell for
+	 * ring test/HWS. So instead of reserving a whole page here for
+	 * kernel, reserve and consume a doorbell from existing KGD kernel
+	 * doorbell page.
 	 */
 
-	doorbell_start_offset =
-			roundup(kfd->shared_resources.doorbell_start_offset,
-					kfd_doorbell_process_slice(kfd));
-
-	doorbell_aperture_size =
-			rounddown(kfd->shared_resources.doorbell_aperture_size,
-					kfd_doorbell_process_slice(kfd));
-
-	if (doorbell_aperture_size > doorbell_start_offset)
-		doorbell_process_limit =
-			(doorbell_aperture_size - doorbell_start_offset) /
-						kfd_doorbell_process_slice(kfd);
-	else
-		return -ENOSPC;
-
-	if (!kfd->max_doorbell_slices ||
-	    doorbell_process_limit < kfd->max_doorbell_slices)
-		kfd->max_doorbell_slices = doorbell_process_limit;
-
-	kfd->doorbell_base = kfd->shared_resources.doorbell_physical_address +
-				doorbell_start_offset;
-
-	kfd->doorbell_base_dw_offset = doorbell_start_offset / sizeof(u32);
-
-	kfd->doorbell_kernel_ptr = ioremap(kfd->doorbell_base,
-					   kfd_doorbell_process_slice(kfd));
-
-	if (!kfd->doorbell_kernel_ptr)
+	/* Bitmap to dynamically allocate doorbells from kernel page */
+	kfd->doorbell_bitmap = bitmap_zalloc(size / sizeof(u32), GFP_KERNEL);
+	if (!kfd->doorbell_bitmap) {
+		DRM_ERROR("Failed to allocate kernel doorbell bitmap\n");
 		return -ENOMEM;
+	}
 
-	pr_debug("Doorbell initialization:\n");
-	pr_debug("doorbell base           == 0x%08lX\n",
-			(uintptr_t)kfd->doorbell_base);
-
-	pr_debug("doorbell_base_dw_offset      == 0x%08lX\n",
-			kfd->doorbell_base_dw_offset);
-
-	pr_debug("doorbell_process_limit  == 0x%08lX\n",
-			doorbell_process_limit);
-
-	pr_debug("doorbell_kernel_offset  == 0x%08lX\n",
-			(uintptr_t)kfd->doorbell_base);
-
-	pr_debug("doorbell aperture size  == 0x%08lX\n",
-			kfd->shared_resources.doorbell_aperture_size);
-
-	pr_debug("doorbell kernel address == %p\n", kfd->doorbell_kernel_ptr);
+	/* Alloc a doorbell page for KFD kernel usages */
+	r = amdgpu_bo_create_kernel(kfd->adev,
+				    size,
+				    PAGE_SIZE,
+				    AMDGPU_GEM_DOMAIN_DOORBELL,
+				    &kfd->doorbells,
+				    NULL,
+				    (void **)&kfd->doorbell_kernel_ptr);
+	if (r) {
+		pr_err("failed to allocate kernel doorbells\n");
+		bitmap_free(kfd->doorbell_bitmap);
+		return r;
+	}
 
+	pr_debug("Doorbell kernel address == %p\n", kfd->doorbell_kernel_ptr);
 	return 0;
 }
 
 void kfd_doorbell_fini(struct kfd_dev *kfd)
 {
-	if (kfd->doorbell_kernel_ptr)
-		iounmap(kfd->doorbell_kernel_ptr);
+	bitmap_free(kfd->doorbell_bitmap);
+	amdgpu_bo_free_kernel(&kfd->doorbells, NULL,
+			     (void **)&kfd->doorbell_kernel_ptr);
 }
 
-int kfd_doorbell_mmap(struct kfd_dev *dev, struct kfd_process *process,
+int kfd_doorbell_mmap(struct kfd_node *dev, struct kfd_process *process,
 		      struct vm_area_struct *vma)
 {
 	phys_addr_t address;
@@ -148,7 +113,7 @@ int kfd_doorbell_mmap(struct kfd_dev *dev, struct kfd_process *process,
 	 * For simplicitly we only allow mapping of the entire doorbell
 	 * allocation of a single device & process.
 	 */
-	if (vma->vm_end - vma->vm_start != kfd_doorbell_process_slice(dev))
+	if (vma->vm_end - vma->vm_start != kfd_doorbell_process_slice(dev->kfd))
 		return -EINVAL;
 
 	pdd = kfd_get_process_device_data(dev, process);
@@ -170,13 +135,13 @@ int kfd_doorbell_mmap(struct kfd_dev *dev, struct kfd_process *process,
 		 "     vm_flags            == 0x%04lX\n"
 		 "     size                == 0x%04lX\n",
 		 (unsigned long long) vma->vm_start, address, vma->vm_flags,
-		 kfd_doorbell_process_slice(dev));
+		 kfd_doorbell_process_slice(dev->kfd));
 
 
 	return io_remap_pfn_range(vma,
 				vma->vm_start,
 				address >> PAGE_SHIFT,
-				kfd_doorbell_process_slice(dev),
+				kfd_doorbell_process_slice(dev->kfd),
 				vma->vm_page_prot);
 }
 
@@ -188,22 +153,19 @@ void __iomem *kfd_get_kernel_doorbell(struct kfd_dev *kfd,
 	u32 inx;
 
 	mutex_lock(&kfd->doorbell_mutex);
-	inx = find_first_zero_bit(kfd->doorbell_available_index,
-					KFD_MAX_NUM_OF_QUEUES_PER_PROCESS);
+	inx = find_first_zero_bit(kfd->doorbell_bitmap, PAGE_SIZE / sizeof(u32));
 
-	__set_bit(inx, kfd->doorbell_available_index);
+	__set_bit(inx, kfd->doorbell_bitmap);
 	mutex_unlock(&kfd->doorbell_mutex);
 
 	if (inx >= KFD_MAX_NUM_OF_QUEUES_PER_PROCESS)
 		return NULL;
 
-	inx *= kfd->device_info.doorbell_size / sizeof(u32);
-
-	/*
-	 * Calculating the kernel doorbell offset using the first
-	 * doorbell page.
-	 */
-	*doorbell_off = kfd->doorbell_base_dw_offset + inx;
+	*doorbell_off = amdgpu_doorbell_index_on_bar(kfd->adev,
+						     kfd->doorbells,
+						     inx,
+						     kfd->device_info.doorbell_size);
+	inx *= 2;
 
 	pr_debug("Get kernel queue doorbell\n"
 			"     doorbell offset   == 0x%08X\n"
@@ -217,11 +179,11 @@ void kfd_release_kernel_doorbell(struct kfd_dev *kfd, u32 __iomem *db_addr)
 {
 	unsigned int inx;
 
-	inx = (unsigned int)(db_addr - kfd->doorbell_kernel_ptr)
-		* sizeof(u32) / kfd->device_info.doorbell_size;
+	inx = (unsigned int)(db_addr - kfd->doorbell_kernel_ptr);
+	inx /= 2;
 
 	mutex_lock(&kfd->doorbell_mutex);
-	__clear_bit(inx, kfd->doorbell_available_index);
+	__clear_bit(inx, kfd->doorbell_bitmap);
 	mutex_unlock(&kfd->doorbell_mutex);
 }
 
@@ -243,80 +205,99 @@ void write_kernel_doorbell64(void __iomem *db, u64 value)
 	}
 }
 
-unsigned int kfd_get_doorbell_dw_offset_in_bar(struct kfd_dev *kfd,
-					struct kfd_process_device *pdd,
-					unsigned int doorbell_id)
+static int init_doorbell_bitmap(struct qcm_process_device *qpd,
+				struct kfd_dev *dev)
 {
-	/*
-	 * doorbell_base_dw_offset accounts for doorbells taken by KGD.
-	 * index * kfd_doorbell_process_slice/sizeof(u32) adjusts to
-	 * the process's doorbells. The offset returned is in dword
-	 * units regardless of the ASIC-dependent doorbell size.
-	 */
-	if (!kfd->shared_resources.enable_mes)
-		return kfd->doorbell_base_dw_offset +
-			pdd->doorbell_index
-			* kfd_doorbell_process_slice(kfd) / sizeof(u32) +
-			doorbell_id *
-			kfd->device_info.doorbell_size / sizeof(u32);
-	else
-		return amdgpu_mes_get_doorbell_dw_offset_in_bar(
-				(struct amdgpu_device *)kfd->adev,
-				pdd->doorbell_index, doorbell_id);
-}
+	unsigned int i;
+	int range_start = dev->shared_resources.non_cp_doorbells_start;
+	int range_end = dev->shared_resources.non_cp_doorbells_end;
 
-uint64_t kfd_get_number_elems(struct kfd_dev *kfd)
-{
-	uint64_t num_of_elems = (kfd->shared_resources.doorbell_aperture_size -
-				kfd->shared_resources.doorbell_start_offset) /
-					kfd_doorbell_process_slice(kfd) + 1;
+	if (!KFD_IS_SOC15(dev))
+		return 0;
 
-	return num_of_elems;
+	/* Mask out doorbells reserved for SDMA, IH, and VCN on SOC15. */
+	pr_debug("reserved doorbell 0x%03x - 0x%03x\n", range_start, range_end);
+	pr_debug("reserved doorbell 0x%03x - 0x%03x\n",
+			range_start + KFD_QUEUE_DOORBELL_MIRROR_OFFSET,
+			range_end + KFD_QUEUE_DOORBELL_MIRROR_OFFSET);
+
+	for (i = 0; i < KFD_MAX_NUM_OF_QUEUES_PER_PROCESS / 2; i++) {
+		if (i >= range_start && i <= range_end) {
+			__set_bit(i, qpd->doorbell_bitmap);
+			__set_bit(i + KFD_QUEUE_DOORBELL_MIRROR_OFFSET,
+				  qpd->doorbell_bitmap);
+		}
+	}
 
+	return 0;
 }
 
 phys_addr_t kfd_get_process_doorbells(struct kfd_process_device *pdd)
 {
-	if (!pdd->doorbell_index) {
-		int r = kfd_alloc_process_doorbells(pdd->dev,
-						    &pdd->doorbell_index);
-		if (r < 0)
+	struct amdgpu_device *adev = pdd->dev->adev;
+	uint32_t first_db_index;
+
+	if (!pdd->qpd.proc_doorbells) {
+		if (kfd_alloc_process_doorbells(pdd->dev->kfd, pdd))
+			/* phys_addr_t 0 is error */
 			return 0;
 	}
 
-	return pdd->dev->doorbell_base +
-		pdd->doorbell_index * kfd_doorbell_process_slice(pdd->dev);
+	first_db_index = amdgpu_doorbell_index_on_bar(adev,
+						      pdd->qpd.proc_doorbells,
+						      0,
+						      pdd->dev->kfd->device_info.doorbell_size);
+	return adev->doorbell.base + first_db_index * sizeof(uint32_t);
 }
 
-int kfd_alloc_process_doorbells(struct kfd_dev *kfd, unsigned int *doorbell_index)
+int kfd_alloc_process_doorbells(struct kfd_dev *kfd, struct kfd_process_device *pdd)
 {
-	int r = 0;
-
-	if (!kfd->shared_resources.enable_mes)
-		r = ida_simple_get(&kfd->doorbell_ida, 1,
-				   kfd->max_doorbell_slices, GFP_KERNEL);
-	else
-		r = amdgpu_mes_alloc_process_doorbells(
-				(struct amdgpu_device *)kfd->adev,
-				doorbell_index);
+	int r;
+	struct qcm_process_device *qpd = &pdd->qpd;
+
+	/* Allocate bitmap for dynamic doorbell allocation */
+	qpd->doorbell_bitmap = bitmap_zalloc(KFD_MAX_NUM_OF_QUEUES_PER_PROCESS,
+					     GFP_KERNEL);
+	if (!qpd->doorbell_bitmap) {
+		DRM_ERROR("Failed to allocate process doorbell bitmap\n");
+		return -ENOMEM;
+	}
 
-	if (r > 0)
-		*doorbell_index = r;
+	r = init_doorbell_bitmap(&pdd->qpd, kfd);
+	if (r) {
+		DRM_ERROR("Failed to initialize process doorbells\n");
+		r = -ENOMEM;
+		goto err;
+	}
 
-	if (r < 0)
-		pr_err("Failed to allocate process doorbells\n");
+	/* Allocate doorbells for this process */
+	r = amdgpu_bo_create_kernel(kfd->adev,
+				    kfd_doorbell_process_slice(kfd),
+				    PAGE_SIZE,
+				    AMDGPU_GEM_DOMAIN_DOORBELL,
+				    &qpd->proc_doorbells,
+				    NULL,
+				    NULL);
+	if (r) {
+		DRM_ERROR("Failed to allocate process doorbells\n");
+		goto err;
+	}
+	return 0;
 
+err:
+	bitmap_free(qpd->doorbell_bitmap);
+	qpd->doorbell_bitmap = NULL;
 	return r;
 }
 
-void kfd_free_process_doorbells(struct kfd_dev *kfd, unsigned int doorbell_index)
+void kfd_free_process_doorbells(struct kfd_dev *kfd, struct kfd_process_device *pdd)
 {
-	if (doorbell_index) {
-		if (!kfd->shared_resources.enable_mes)
-			ida_simple_remove(&kfd->doorbell_ida, doorbell_index);
-		else
-			amdgpu_mes_free_process_doorbells(
-					(struct amdgpu_device *)kfd->adev,
-					doorbell_index);
+	struct qcm_process_device *qpd = &pdd->qpd;
+
+	if (qpd->doorbell_bitmap) {
+		bitmap_free(qpd->doorbell_bitmap);
+		qpd->doorbell_bitmap = NULL;
 	}
+
+	amdgpu_bo_free_kernel(&qpd->proc_doorbells, NULL, NULL);
 }
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_events.c
index c894cf8f7c50..5a190dd6be4e 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_events.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_events.c
@@ -31,7 +31,7 @@
 #include <linux/memory.h>
 #include "kfd_priv.h"
 #include "kfd_events.h"
-#include "kfd_iommu.h"
+#include "kfd_device_queue_manager.h"
 #include <linux/device.h>
 
 /*
@@ -41,6 +41,7 @@ struct kfd_event_waiter {
 	wait_queue_entry_t wait;
 	struct kfd_event *event; /* Event to wait for */
 	bool activated;		 /* Becomes true when event is signaled */
+	bool event_age_enabled;  /* set to true when last_event_age is non-zero */
 };
 
 /*
@@ -348,7 +349,7 @@ static int kfd_event_page_set(struct kfd_process *p, void *kernel_address,
 
 int kfd_kmap_event_page(struct kfd_process *p, uint64_t event_page_offset)
 {
-	struct kfd_dev *kfd;
+	struct kfd_node *kfd;
 	struct kfd_process_device *pdd;
 	void *mem, *kern_addr;
 	uint64_t size;
@@ -431,6 +432,7 @@ int kfd_event_create(struct file *devkfd, struct kfd_process *p,
 	if (!ret) {
 		*event_id = ev->event_id;
 		*event_trigger_data = ev->event_id;
+		ev->event_age = 1;
 	} else {
 		kfree(ev);
 	}
@@ -629,6 +631,11 @@ static void set_event(struct kfd_event *ev)
 	 * updating the wait queues in kfd_wait_on_events.
 	 */
 	ev->signaled = !ev->auto_reset || !waitqueue_active(&ev->wq);
+	if (!(++ev->event_age)) {
+		/* Never wrap back to reserved/default event age 0/1 */
+		ev->event_age = 2;
+		WARN_ONCE(1, "event_age wrap back!");
+	}
 
 	list_for_each_entry(waiter, &ev->wq.head, wait.entry)
 		WRITE_ONCE(waiter->activated, true);
@@ -720,7 +727,7 @@ void kfd_signal_event_interrupt(u32 pasid, uint32_t partial_id,
 	 * to process context, kfd_process could attempt to exit while we are
 	 * running so the lookup function increments the process ref count.
 	 */
-	struct kfd_process *p = kfd_lookup_process_by_pasid(pasid);
+	struct kfd_process *p = kfd_lookup_process_by_pasid(pasid, NULL);
 
 	if (!p)
 		return; /* Presumably process exited. */
@@ -791,9 +798,9 @@ static struct kfd_event_waiter *alloc_event_waiters(uint32_t num_events)
 
 static int init_event_waiter(struct kfd_process *p,
 		struct kfd_event_waiter *waiter,
-		uint32_t event_id)
+		struct kfd_event_data *event_data)
 {
-	struct kfd_event *ev = lookup_event_by_id(p, event_id);
+	struct kfd_event *ev = lookup_event_by_id(p, event_data->event_id);
 
 	if (!ev)
 		return -EINVAL;
@@ -802,6 +809,15 @@ static int init_event_waiter(struct kfd_process *p,
 	waiter->event = ev;
 	waiter->activated = ev->signaled;
 	ev->signaled = ev->signaled && !ev->auto_reset;
+
+	/* last_event_age = 0 reserved for backward compatible */
+	if (waiter->event->type == KFD_EVENT_TYPE_SIGNAL &&
+		event_data->signal_event_data.last_event_age) {
+		waiter->event_age_enabled = true;
+		if (ev->event_age != event_data->signal_event_data.last_event_age)
+			waiter->activated = true;
+	}
+
 	if (!waiter->activated)
 		add_wait_queue(&ev->wq, &waiter->wait);
 	spin_unlock(&ev->lock);
@@ -849,22 +865,33 @@ static int copy_signaled_event_data(uint32_t num_events,
 		struct kfd_event_waiter *event_waiters,
 		struct kfd_event_data __user *data)
 {
-	struct kfd_hsa_memory_exception_data *src;
-	struct kfd_hsa_memory_exception_data __user *dst;
+	void *src;
+	void __user *dst;
 	struct kfd_event_waiter *waiter;
 	struct kfd_event *event;
-	uint32_t i;
+	uint32_t i, size = 0;
 
 	for (i = 0; i < num_events; i++) {
 		waiter = &event_waiters[i];
 		event = waiter->event;
 		if (!event)
 			return -EINVAL; /* event was destroyed */
-		if (waiter->activated && event->type == KFD_EVENT_TYPE_MEMORY) {
-			dst = &data[i].memory_exception_data;
-			src = &event->memory_exception_data;
-			if (copy_to_user(dst, src,
-				sizeof(struct kfd_hsa_memory_exception_data)))
+		if (waiter->activated) {
+			if (event->type == KFD_EVENT_TYPE_MEMORY) {
+				dst = &data[i].memory_exception_data;
+				src = &event->memory_exception_data;
+				size = sizeof(struct kfd_hsa_memory_exception_data);
+			} else if (event->type == KFD_EVENT_TYPE_HW_EXCEPTION) {
+				dst = &data[i].memory_exception_data;
+				src = &event->hw_exception_data;
+				size = sizeof(struct kfd_hsa_hw_exception_data);
+			} else if (event->type == KFD_EVENT_TYPE_SIGNAL &&
+				waiter->event_age_enabled) {
+				dst = &data[i].signal_event_data.last_event_age;
+				src = &event->event_age;
+				size = sizeof(u64);
+			}
+			if (size && copy_to_user(dst, src, size))
 				return -EFAULT;
 		}
 	}
@@ -942,8 +969,7 @@ int kfd_wait_on_events(struct kfd_process *p,
 			goto out_unlock;
 		}
 
-		ret = init_event_waiter(p, &event_waiters[i],
-					event_data.event_id);
+		ret = init_event_waiter(p, &event_waiters[i], &event_data);
 		if (ret)
 			goto out_unlock;
 	}
@@ -1102,8 +1128,8 @@ static void lookup_events_by_type_and_signal(struct kfd_process *p,
 
 	if (type == KFD_EVENT_TYPE_MEMORY) {
 		dev_warn(kfd_device,
-			"Sending SIGSEGV to process %d (pasid 0x%x)",
-				p->lead_thread->pid, p->pasid);
+			"Sending SIGSEGV to process pid %d",
+				p->lead_thread->pid);
 		send_sig(SIGSEGV, p->lead_thread, 0);
 	}
 
@@ -1111,147 +1137,89 @@ static void lookup_events_by_type_and_signal(struct kfd_process *p,
 	if (send_signal) {
 		if (send_sigterm) {
 			dev_warn(kfd_device,
-				"Sending SIGTERM to process %d (pasid 0x%x)",
-					p->lead_thread->pid, p->pasid);
+				"Sending SIGTERM to process pid %d",
+					p->lead_thread->pid);
 			send_sig(SIGTERM, p->lead_thread, 0);
 		} else {
 			dev_err(kfd_device,
-				"Process %d (pasid 0x%x) got unhandled exception",
-				p->lead_thread->pid, p->pasid);
+				"Process pid %d got unhandled exception",
+				p->lead_thread->pid);
 		}
 	}
 
 	rcu_read_unlock();
 }
 
-#ifdef KFD_SUPPORT_IOMMU_V2
-void kfd_signal_iommu_event(struct kfd_dev *dev, u32 pasid,
-		unsigned long address, bool is_write_requested,
-		bool is_execute_requested)
+void kfd_signal_hw_exception_event(u32 pasid)
 {
-	struct kfd_hsa_memory_exception_data memory_exception_data;
-	struct vm_area_struct *vma;
-	int user_gpu_id;
-
 	/*
 	 * Because we are called from arbitrary context (workqueue) as opposed
 	 * to process context, kfd_process could attempt to exit while we are
 	 * running so the lookup function increments the process ref count.
 	 */
-	struct kfd_process *p = kfd_lookup_process_by_pasid(pasid);
-	struct mm_struct *mm;
+	struct kfd_process *p = kfd_lookup_process_by_pasid(pasid, NULL);
 
 	if (!p)
 		return; /* Presumably process exited. */
 
-	/* Take a safe reference to the mm_struct, which may otherwise
-	 * disappear even while the kfd_process is still referenced.
-	 */
-	mm = get_task_mm(p->lead_thread);
-	if (!mm) {
-		kfd_unref_process(p);
-		return; /* Process is exiting */
-	}
-
-	user_gpu_id = kfd_process_get_user_gpu_id(p, dev->id);
-	if (unlikely(user_gpu_id == -EINVAL)) {
-		WARN_ONCE(1, "Could not get user_gpu_id from dev->id:%x\n", dev->id);
-		return;
-	}
-	memset(&memory_exception_data, 0, sizeof(memory_exception_data));
-
-	mmap_read_lock(mm);
-	vma = find_vma(mm, address);
-
-	memory_exception_data.gpu_id = user_gpu_id;
-	memory_exception_data.va = address;
-	/* Set failure reason */
-	memory_exception_data.failure.NotPresent = 1;
-	memory_exception_data.failure.NoExecute = 0;
-	memory_exception_data.failure.ReadOnly = 0;
-	if (vma && address >= vma->vm_start) {
-		memory_exception_data.failure.NotPresent = 0;
-
-		if (is_write_requested && !(vma->vm_flags & VM_WRITE))
-			memory_exception_data.failure.ReadOnly = 1;
-		else
-			memory_exception_data.failure.ReadOnly = 0;
-
-		if (is_execute_requested && !(vma->vm_flags & VM_EXEC))
-			memory_exception_data.failure.NoExecute = 1;
-		else
-			memory_exception_data.failure.NoExecute = 0;
-	}
-
-	mmap_read_unlock(mm);
-	mmput(mm);
-
-	pr_debug("notpresent %d, noexecute %d, readonly %d\n",
-			memory_exception_data.failure.NotPresent,
-			memory_exception_data.failure.NoExecute,
-			memory_exception_data.failure.ReadOnly);
-
-	/* Workaround on Raven to not kill the process when memory is freed
-	 * before IOMMU is able to finish processing all the excessive PPRs
-	 */
-
-	if (KFD_GC_VERSION(dev) != IP_VERSION(9, 1, 0) &&
-	    KFD_GC_VERSION(dev) != IP_VERSION(9, 2, 2) &&
-	    KFD_GC_VERSION(dev) != IP_VERSION(9, 3, 0))
-		lookup_events_by_type_and_signal(p, KFD_EVENT_TYPE_MEMORY,
-				&memory_exception_data);
-
+	lookup_events_by_type_and_signal(p, KFD_EVENT_TYPE_HW_EXCEPTION, NULL);
 	kfd_unref_process(p);
 }
-#endif /* KFD_SUPPORT_IOMMU_V2 */
 
-void kfd_signal_hw_exception_event(u32 pasid)
+void kfd_signal_vm_fault_event_with_userptr(struct kfd_process *p, uint64_t gpu_va)
 {
-	/*
-	 * Because we are called from arbitrary context (workqueue) as opposed
-	 * to process context, kfd_process could attempt to exit while we are
-	 * running so the lookup function increments the process ref count.
-	 */
-	struct kfd_process *p = kfd_lookup_process_by_pasid(pasid);
-
-	if (!p)
-		return; /* Presumably process exited. */
-
-	lookup_events_by_type_and_signal(p, KFD_EVENT_TYPE_HW_EXCEPTION, NULL);
-	kfd_unref_process(p);
+	struct kfd_process_device *pdd;
+	struct kfd_hsa_memory_exception_data exception_data;
+	int i;
+
+	memset(&exception_data, 0, sizeof(exception_data));
+	exception_data.va = gpu_va;
+	exception_data.failure.NotPresent = 1;
+
+	// Send VM seg fault to all kfd process device
+	for (i = 0; i < p->n_pdds; i++) {
+		pdd = p->pdds[i];
+		exception_data.gpu_id = pdd->user_gpu_id;
+		kfd_evict_process_device(pdd);
+		kfd_signal_vm_fault_event(pdd, NULL, &exception_data);
+	}
 }
 
-void kfd_signal_vm_fault_event(struct kfd_dev *dev, u32 pasid,
-				struct kfd_vm_fault_info *info)
+void kfd_signal_vm_fault_event(struct kfd_process_device *pdd,
+				struct kfd_vm_fault_info *info,
+				struct kfd_hsa_memory_exception_data *data)
 {
 	struct kfd_event *ev;
 	uint32_t id;
-	struct kfd_process *p = kfd_lookup_process_by_pasid(pasid);
+	struct kfd_process *p = pdd->process;
 	struct kfd_hsa_memory_exception_data memory_exception_data;
 	int user_gpu_id;
 
-	if (!p)
-		return; /* Presumably process exited. */
-
-	user_gpu_id = kfd_process_get_user_gpu_id(p, dev->id);
+	user_gpu_id = kfd_process_get_user_gpu_id(p, pdd->dev->id);
 	if (unlikely(user_gpu_id == -EINVAL)) {
-		WARN_ONCE(1, "Could not get user_gpu_id from dev->id:%x\n", dev->id);
+		WARN_ONCE(1, "Could not get user_gpu_id from dev->id:%x\n",
+			  pdd->dev->id);
 		return;
 	}
 
-	memset(&memory_exception_data, 0, sizeof(memory_exception_data));
-	memory_exception_data.gpu_id = user_gpu_id;
-	memory_exception_data.failure.imprecise = true;
-	/* Set failure reason */
-	if (info) {
-		memory_exception_data.va = (info->page_addr) << PAGE_SHIFT;
-		memory_exception_data.failure.NotPresent =
-			info->prot_valid ? 1 : 0;
-		memory_exception_data.failure.NoExecute =
-			info->prot_exec ? 1 : 0;
-		memory_exception_data.failure.ReadOnly =
-			info->prot_write ? 1 : 0;
-		memory_exception_data.failure.imprecise = 0;
+	/* SoC15 chips and onwards will pass in data from now on. */
+	if (!data) {
+		memset(&memory_exception_data, 0, sizeof(memory_exception_data));
+		memory_exception_data.gpu_id = user_gpu_id;
+		memory_exception_data.failure.imprecise = true;
+
+		/* Set failure reason */
+		if (info) {
+			memory_exception_data.va = (info->page_addr) <<
+								PAGE_SHIFT;
+			memory_exception_data.failure.NotPresent =
+				info->prot_valid ? 1 : 0;
+			memory_exception_data.failure.NoExecute =
+				info->prot_exec ? 1 : 0;
+			memory_exception_data.failure.ReadOnly =
+				info->prot_write ? 1 : 0;
+			memory_exception_data.failure.imprecise = 0;
+		}
 	}
 
 	rcu_read_lock();
@@ -1260,16 +1228,16 @@ void kfd_signal_vm_fault_event(struct kfd_dev *dev, u32 pasid,
 	idr_for_each_entry_continue(&p->event_idr, ev, id)
 		if (ev->type == KFD_EVENT_TYPE_MEMORY) {
 			spin_lock(&ev->lock);
-			ev->memory_exception_data = memory_exception_data;
+			ev->memory_exception_data = data ? *data :
+							memory_exception_data;
 			set_event(ev);
 			spin_unlock(&ev->lock);
 		}
 
 	rcu_read_unlock();
-	kfd_unref_process(p);
 }
 
-void kfd_signal_reset_event(struct kfd_dev *dev)
+void kfd_signal_reset_event(struct kfd_node *dev)
 {
 	struct kfd_hsa_hw_exception_data hw_exception_data;
 	struct kfd_hsa_memory_exception_data memory_exception_data;
@@ -1293,12 +1261,41 @@ void kfd_signal_reset_event(struct kfd_dev *dev)
 	idx = srcu_read_lock(&kfd_processes_srcu);
 	hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) {
 		int user_gpu_id = kfd_process_get_user_gpu_id(p, dev->id);
+		struct kfd_process_device *pdd = kfd_get_process_device_data(dev, p);
 
 		if (unlikely(user_gpu_id == -EINVAL)) {
 			WARN_ONCE(1, "Could not get user_gpu_id from dev->id:%x\n", dev->id);
 			continue;
 		}
 
+		if (unlikely(!pdd)) {
+			WARN_ONCE(1, "Could not get device data from process pid:%d\n",
+				  p->lead_thread->pid);
+			continue;
+		}
+
+		if (dev->dqm->detect_hang_count && !pdd->has_reset_queue)
+			continue;
+
+		if (dev->dqm->detect_hang_count) {
+			struct amdgpu_task_info *ti;
+			struct amdgpu_fpriv *drv_priv;
+
+			if (unlikely(amdgpu_file_to_fpriv(pdd->drm_file, &drv_priv))) {
+				WARN_ONCE(1, "Could not get vm for device %x from pid:%d\n",
+					  dev->id, p->lead_thread->pid);
+				continue;
+			}
+
+			ti = amdgpu_vm_get_task_info_vm(&drv_priv->vm);
+			if (ti) {
+				dev_err(dev->adev->dev,
+					"Queues reset on process %s tid %d thread %s pid %d\n",
+					ti->process_name, ti->tgid, ti->task.comm, ti->task.pid);
+				amdgpu_vm_put_task_info(ti);
+			}
+		}
+
 		rcu_read_lock();
 
 		id = KFD_FIRST_NONSIGNAL_EVENT_ID;
@@ -1325,21 +1322,24 @@ void kfd_signal_reset_event(struct kfd_dev *dev)
 	srcu_read_unlock(&kfd_processes_srcu, idx);
 }
 
-void kfd_signal_poison_consumed_event(struct kfd_dev *dev, u32 pasid)
+void kfd_signal_poison_consumed_event(struct kfd_node *dev, u32 pasid)
 {
-	struct kfd_process *p = kfd_lookup_process_by_pasid(pasid);
+	struct kfd_process *p = kfd_lookup_process_by_pasid(pasid, NULL);
 	struct kfd_hsa_memory_exception_data memory_exception_data;
 	struct kfd_hsa_hw_exception_data hw_exception_data;
 	struct kfd_event *ev;
 	uint32_t id = KFD_FIRST_NONSIGNAL_EVENT_ID;
 	int user_gpu_id;
 
-	if (!p)
+	if (!p) {
+		dev_warn(dev->adev->dev, "Not find process with pasid:%d\n", pasid);
 		return; /* Presumably process exited. */
+	}
 
 	user_gpu_id = kfd_process_get_user_gpu_id(p, dev->id);
 	if (unlikely(user_gpu_id == -EINVAL)) {
 		WARN_ONCE(1, "Could not get user_gpu_id from dev->id:%x\n", dev->id);
+		kfd_unref_process(p);
 		return;
 	}
 
@@ -1371,6 +1371,8 @@ void kfd_signal_poison_consumed_event(struct kfd_dev *dev, u32 pasid)
 		}
 	}
 
+	dev_warn(dev->adev->dev, "Send SIGBUS to process %s(pasid:%d)\n",
+		p->lead_thread->comm, pasid);
 	rcu_read_unlock();
 
 	/* user application will handle SIGBUS signal */
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_events.h b/drivers/gpu/drm/amd/amdkfd/kfd_events.h
index 1c62c8dd6460..52ccfd397c2b 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_events.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_events.h
@@ -53,6 +53,7 @@ struct signal_page;
 
 struct kfd_event {
 	u32 event_id;
+	u64 event_age;
 
 	bool signaled;
 	bool auto_reset;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c
index 8aebe408c544..1d170dc50df3 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c
@@ -23,7 +23,6 @@
  */
 
 #include <linux/device.h>
-#include <linux/export.h>
 #include <linux/err.h>
 #include <linux/fs.h>
 #include <linux/sched.h>
@@ -36,6 +35,7 @@
 #include <linux/mm.h>
 #include <linux/mman.h>
 #include <linux/processor.h>
+#include "amdgpu_vm.h"
 
 /*
  * The primary memory I/O features being added for revisions of gfxip
@@ -322,22 +322,19 @@ static void kfd_init_apertures_vi(struct kfd_process_device *pdd, uint8_t id)
 	pdd->lds_base = MAKE_LDS_APP_BASE_VI();
 	pdd->lds_limit = MAKE_LDS_APP_LIMIT(pdd->lds_base);
 
-	if (!pdd->dev->use_iommu_v2) {
-		/* dGPUs: SVM aperture starting at 0
-		 * with small reserved space for kernel.
-		 * Set them to CANONICAL addresses.
-		 */
-		pdd->gpuvm_base = SVM_USER_BASE;
-		pdd->gpuvm_limit =
-			pdd->dev->shared_resources.gpuvm_size - 1;
-	} else {
-		/* set them to non CANONICAL addresses, and no SVM is
-		 * allocated.
-		 */
-		pdd->gpuvm_base = MAKE_GPUVM_APP_BASE_VI(id + 1);
-		pdd->gpuvm_limit = MAKE_GPUVM_APP_LIMIT(pdd->gpuvm_base,
-				pdd->dev->shared_resources.gpuvm_size);
-	}
+	/* dGPUs: SVM aperture starting at 0
+	 * with small reserved space for kernel.
+	 * Set them to CANONICAL addresses.
+	 */
+	pdd->gpuvm_base = max(SVM_USER_BASE, AMDGPU_VA_RESERVED_BOTTOM);
+	pdd->gpuvm_limit =
+		pdd->dev->kfd->shared_resources.gpuvm_size - 1;
+
+	/* dGPUs: the reserved space for kernel
+	 * before SVM
+	 */
+	pdd->qpd.cwsr_base = SVM_CWSR_BASE;
+	pdd->qpd.ib_base = SVM_IB_BASE;
 
 	pdd->scratch_base = MAKE_SCRATCH_APP_BASE_VI();
 	pdd->scratch_limit = MAKE_SCRATCH_APP_LIMIT(pdd->scratch_base);
@@ -348,24 +345,24 @@ static void kfd_init_apertures_v9(struct kfd_process_device *pdd, uint8_t id)
 	pdd->lds_base = MAKE_LDS_APP_BASE_V9();
 	pdd->lds_limit = MAKE_LDS_APP_LIMIT(pdd->lds_base);
 
-	/* Raven needs SVM to support graphic handle, etc. Leave the small
-	 * reserved space before SVM on Raven as well, even though we don't
-	 * have to.
-	 * Set gpuvm_base and gpuvm_limit to CANONICAL addresses so that they
-	 * are used in Thunk to reserve SVM.
-	 */
-	pdd->gpuvm_base = SVM_USER_BASE;
+	pdd->gpuvm_base = AMDGPU_VA_RESERVED_BOTTOM;
 	pdd->gpuvm_limit =
-		pdd->dev->shared_resources.gpuvm_size - 1;
+		pdd->dev->kfd->shared_resources.gpuvm_size - 1;
 
 	pdd->scratch_base = MAKE_SCRATCH_APP_BASE_V9();
 	pdd->scratch_limit = MAKE_SCRATCH_APP_LIMIT(pdd->scratch_base);
+
+	/*
+	 * Place TBA/TMA on opposite side of VM hole to prevent
+	 * stray faults from triggering SVM on these pages.
+	 */
+	pdd->qpd.cwsr_base = AMDGPU_VA_RESERVED_TRAP_START(pdd->dev->adev);
 }
 
 int kfd_init_apertures(struct kfd_process *process)
 {
 	uint8_t id  = 0;
-	struct kfd_dev *dev;
+	struct kfd_node *dev;
 	struct kfd_process_device *pdd;
 
 	/*Iterating over all devices*/
@@ -382,7 +379,8 @@ int kfd_init_apertures(struct kfd_process *process)
 
 		pdd = kfd_create_process_device_data(dev, process);
 		if (!pdd) {
-			pr_err("Failed to create process device data\n");
+			dev_err(dev->adev->dev,
+				"Failed to create process device data\n");
 			return -ENOMEM;
 		}
 		/*
@@ -416,14 +414,6 @@ int kfd_init_apertures(struct kfd_process *process)
 					return -EINVAL;
 				}
 			}
-
-			if (!dev->use_iommu_v2) {
-				/* dGPUs: the reserved space for kernel
-				 * before SVM
-				 */
-				pdd->qpd.cwsr_base = SVM_CWSR_BASE;
-				pdd->qpd.ib_base = SVM_IB_BASE;
-			}
 		}
 
 		dev_dbg(kfd_device, "node id %u\n", id);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v10.c b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v10.c
new file mode 100644
index 000000000000..3e1ad8974797
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v10.c
@@ -0,0 +1,386 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "kfd_events.h"
+#include "kfd_debug.h"
+#include "soc15_int.h"
+#include "kfd_device_queue_manager.h"
+
+/*
+ * GFX10 SQ Interrupts
+ *
+ * There are 3 encoding types of interrupts sourced from SQ sent as a 44-bit
+ * packet to the Interrupt Handler:
+ * Auto - Generated by the SQG (various cmd overflows, timestamps etc)
+ * Wave - Generated by S_SENDMSG through a shader program
+ * Error - HW generated errors (Illegal instructions, Memviols, EDC etc)
+ *
+ * The 44-bit packet is mapped as {context_id1[7:0],context_id0[31:0]} plus
+ * 4-bits for VMID (SOC15_VMID_FROM_IH_ENTRY) as such:
+ *
+ * - context_id1[7:6]
+ * Encoding type (0 = Auto, 1 = Wave, 2 = Error)
+ *
+ * - context_id0[24]
+ * PRIV bit indicates that Wave S_SEND or error occurred within trap
+ *
+ * - context_id0[22:0]
+ * 23-bit data with the following layout per encoding type:
+ * Auto - only context_id0[8:0] is used, which reports various interrupts
+ * generated by SQG.  The rest is 0.
+ * Wave - user data sent from m0 via S_SENDMSG
+ * Error - Error type (context_id0[22:19]), Error Details (rest of bits)
+ *
+ * The other context_id bits show coordinates (SE/SH/CU/SIMD/WGP) for wave
+ * S_SENDMSG and Errors.  These are 0 for Auto.
+ */
+
+enum SQ_INTERRUPT_WORD_ENCODING {
+	SQ_INTERRUPT_WORD_ENCODING_AUTO = 0x0,
+	SQ_INTERRUPT_WORD_ENCODING_INST,
+	SQ_INTERRUPT_WORD_ENCODING_ERROR,
+};
+
+enum SQ_INTERRUPT_ERROR_TYPE {
+	SQ_INTERRUPT_ERROR_TYPE_EDC_FUE = 0x0,
+	SQ_INTERRUPT_ERROR_TYPE_ILLEGAL_INST,
+	SQ_INTERRUPT_ERROR_TYPE_MEMVIOL,
+	SQ_INTERRUPT_ERROR_TYPE_EDC_FED,
+};
+
+/* SQ_INTERRUPT_WORD_AUTO_CTXID */
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__THREAD_TRACE__SHIFT 0
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__WLT__SHIFT 1
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__THREAD_TRACE_BUF0_FULL__SHIFT 2
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__THREAD_TRACE_BUF1_FULL__SHIFT 3
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__THREAD_TRACE_UTC_ERROR__SHIFT 7
+#define SQ_INTERRUPT_WORD_AUTO_CTXID1__SE_ID__SHIFT 4
+#define SQ_INTERRUPT_WORD_AUTO_CTXID1__ENCODING__SHIFT 6
+
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__THREAD_TRACE_MASK 0x00000001
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__WLT_MASK 0x00000002
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__THREAD_TRACE_BUF0_FULL_MASK 0x00000004
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__THREAD_TRACE_BUF1_FULL_MASK 0x00000008
+#define SQ_INTERRUPT_WORD_AUTO_CTXID0__THREAD_TRACE_UTC_ERROR_MASK 0x00000080
+#define SQ_INTERRUPT_WORD_AUTO_CTXID1__SE_ID_MASK 0x030
+#define SQ_INTERRUPT_WORD_AUTO_CTXID1__ENCODING_MASK 0x0c0
+
+/* SQ_INTERRUPT_WORD_WAVE_CTXID */
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__DATA__SHIFT 0
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__SA_ID__SHIFT 23
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__PRIV__SHIFT 24
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__WAVE_ID__SHIFT 25
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__SIMD_ID__SHIFT 30
+#define SQ_INTERRUPT_WORD_WAVE_CTXID1__WGP_ID__SHIFT 0
+#define SQ_INTERRUPT_WORD_WAVE_CTXID1__SE_ID__SHIFT 4
+#define SQ_INTERRUPT_WORD_WAVE_CTXID1__ENCODING__SHIFT 6
+
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__DATA_MASK 0x000007fffff
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__SA_ID_MASK 0x0000800000
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__PRIV_MASK 0x00001000000
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__WAVE_ID_MASK 0x0003e000000
+#define SQ_INTERRUPT_WORD_WAVE_CTXID0__SIMD_ID_MASK 0x000c0000000
+#define SQ_INTERRUPT_WORD_WAVE_CTXID1__WGP_ID_MASK 0x00f
+#define SQ_INTERRUPT_WORD_WAVE_CTXID1__SE_ID_MASK 0x030
+#define SQ_INTERRUPT_WORD_WAVE_CTXID1__ENCODING_MASK 0x0c0
+
+#define KFD_CTXID0__ERR_TYPE_MASK 0x780000
+#define KFD_CTXID0__ERR_TYPE__SHIFT 19
+
+/* GFX10 SQ interrupt ENC type bit (context_id1[7:6]) for wave s_sendmsg */
+#define KFD_CONTEXT_ID1_ENC_TYPE_WAVE_MASK	0x40
+/* GFX10 SQ interrupt PRIV bit (context_id0[24]) for s_sendmsg inside trap */
+#define KFD_CONTEXT_ID0_PRIV_MASK		0x1000000
+/*
+ * The debugger will send user data(m0) with PRIV=1 to indicate it requires
+ * notification from the KFD with the following queue id (DOORBELL_ID) and
+ * trap code (TRAP_CODE).
+ */
+#define KFD_CONTEXT_ID0_DEBUG_DOORBELL_MASK	0x0003ff
+#define KFD_CONTEXT_ID0_DEBUG_TRAP_CODE_SHIFT	10
+#define KFD_CONTEXT_ID0_DEBUG_TRAP_CODE_MASK	0x07fc00
+#define KFD_DEBUG_DOORBELL_ID(ctxid0)	((ctxid0) &	\
+				KFD_CONTEXT_ID0_DEBUG_DOORBELL_MASK)
+#define KFD_DEBUG_TRAP_CODE(ctxid0)	(((ctxid0) &	\
+				KFD_CONTEXT_ID0_DEBUG_TRAP_CODE_MASK)	\
+				>> KFD_CONTEXT_ID0_DEBUG_TRAP_CODE_SHIFT)
+#define KFD_DEBUG_CP_BAD_OP_ECODE_MASK		0x3fffc00
+#define KFD_DEBUG_CP_BAD_OP_ECODE_SHIFT		10
+#define KFD_DEBUG_CP_BAD_OP_ECODE(ctxid0) (((ctxid0) &			\
+				KFD_DEBUG_CP_BAD_OP_ECODE_MASK)		\
+				>> KFD_DEBUG_CP_BAD_OP_ECODE_SHIFT)
+
+static bool event_interrupt_isr_v10(struct kfd_node *dev,
+					const uint32_t *ih_ring_entry,
+					uint32_t *patched_ihre,
+					bool *patched_flag)
+{
+	uint16_t source_id, client_id, pasid, vmid;
+	const uint32_t *data = ih_ring_entry;
+
+	source_id = SOC15_SOURCE_ID_FROM_IH_ENTRY(ih_ring_entry);
+	client_id = SOC15_CLIENT_ID_FROM_IH_ENTRY(ih_ring_entry);
+
+	/* Only handle interrupts from KFD VMIDs */
+	vmid = SOC15_VMID_FROM_IH_ENTRY(ih_ring_entry);
+	if (!KFD_IRQ_IS_FENCE(client_id, source_id) &&
+	   (vmid < dev->vm_info.first_vmid_kfd ||
+	    vmid > dev->vm_info.last_vmid_kfd))
+		return false;
+
+	pasid = SOC15_PASID_FROM_IH_ENTRY(ih_ring_entry);
+
+	/* Only handle clients we care about */
+	if (client_id != SOC15_IH_CLIENTID_GRBM_CP &&
+	    client_id != SOC15_IH_CLIENTID_SDMA0 &&
+	    client_id != SOC15_IH_CLIENTID_SDMA1 &&
+	    client_id != SOC15_IH_CLIENTID_SDMA2 &&
+	    client_id != SOC15_IH_CLIENTID_SDMA3 &&
+	    client_id != SOC15_IH_CLIENTID_SDMA4 &&
+	    client_id != SOC15_IH_CLIENTID_SDMA5 &&
+	    client_id != SOC15_IH_CLIENTID_SDMA6 &&
+	    client_id != SOC15_IH_CLIENTID_SDMA7 &&
+	    client_id != SOC15_IH_CLIENTID_VMC &&
+	    client_id != SOC15_IH_CLIENTID_VMC1 &&
+	    client_id != SOC15_IH_CLIENTID_UTCL2 &&
+	    client_id != SOC15_IH_CLIENTID_SE0SH &&
+	    client_id != SOC15_IH_CLIENTID_SE1SH &&
+	    client_id != SOC15_IH_CLIENTID_SE2SH &&
+	    client_id != SOC15_IH_CLIENTID_SE3SH)
+		return false;
+
+	dev_dbg(dev->adev->dev,
+		"client id 0x%x, source id %d, vmid %d, pasid 0x%x. raw data:\n",
+		client_id, source_id, vmid, pasid);
+	dev_dbg(dev->adev->dev, "%8X, %8X, %8X, %8X, %8X, %8X, %8X, %8X.\n",
+		data[0], data[1], data[2], data[3], data[4], data[5], data[6],
+		data[7]);
+
+	if (pasid == 0)
+		return 0;
+
+	/* Interrupt types we care about: various signals and faults.
+	 * They will be forwarded to a work queue (see below).
+	 */
+	return source_id == SOC15_INTSRC_CP_END_OF_PIPE ||
+		source_id == SOC15_INTSRC_SDMA_TRAP ||
+		source_id == SOC15_INTSRC_SQ_INTERRUPT_MSG ||
+		source_id == SOC15_INTSRC_CP_BAD_OPCODE ||
+		client_id == SOC15_IH_CLIENTID_VMC ||
+		client_id == SOC15_IH_CLIENTID_VMC1 ||
+		client_id == SOC15_IH_CLIENTID_UTCL2 ||
+		KFD_IRQ_IS_FENCE(client_id, source_id);
+}
+
+static void event_interrupt_wq_v10(struct kfd_node *dev,
+					const uint32_t *ih_ring_entry)
+{
+	uint16_t source_id, client_id, pasid, vmid;
+	uint32_t context_id0, context_id1;
+	uint32_t encoding, sq_intr_err_type;
+
+	source_id = SOC15_SOURCE_ID_FROM_IH_ENTRY(ih_ring_entry);
+	client_id = SOC15_CLIENT_ID_FROM_IH_ENTRY(ih_ring_entry);
+	pasid = SOC15_PASID_FROM_IH_ENTRY(ih_ring_entry);
+	vmid = SOC15_VMID_FROM_IH_ENTRY(ih_ring_entry);
+	context_id0 = SOC15_CONTEXT_ID0_FROM_IH_ENTRY(ih_ring_entry);
+	context_id1 = SOC15_CONTEXT_ID1_FROM_IH_ENTRY(ih_ring_entry);
+
+	if (client_id == SOC15_IH_CLIENTID_GRBM_CP ||
+	    client_id == SOC15_IH_CLIENTID_SE0SH ||
+	    client_id == SOC15_IH_CLIENTID_SE1SH ||
+	    client_id == SOC15_IH_CLIENTID_SE2SH ||
+	    client_id == SOC15_IH_CLIENTID_SE3SH) {
+		if (source_id == SOC15_INTSRC_CP_END_OF_PIPE)
+			kfd_signal_event_interrupt(pasid, context_id0, 32);
+		else if (source_id == SOC15_INTSRC_SQ_INTERRUPT_MSG) {
+			encoding = REG_GET_FIELD(context_id1,
+						SQ_INTERRUPT_WORD_WAVE_CTXID1, ENCODING);
+			switch (encoding) {
+			case SQ_INTERRUPT_WORD_ENCODING_AUTO:
+				dev_dbg_ratelimited(
+					dev->adev->dev,
+					"sq_intr: auto, se %d, ttrace %d, wlt %d, ttrac_buf0_full %d, ttrac_buf1_full %d, ttrace_utc_err %d\n",
+					REG_GET_FIELD(
+						context_id1,
+						SQ_INTERRUPT_WORD_AUTO_CTXID1,
+						SE_ID),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_AUTO_CTXID0,
+						THREAD_TRACE),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_AUTO_CTXID0,
+						WLT),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_AUTO_CTXID0,
+						THREAD_TRACE_BUF0_FULL),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_AUTO_CTXID0,
+						THREAD_TRACE_BUF1_FULL),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_AUTO_CTXID0,
+						THREAD_TRACE_UTC_ERROR));
+				break;
+			case SQ_INTERRUPT_WORD_ENCODING_INST:
+				dev_dbg_ratelimited(
+					dev->adev->dev,
+					"sq_intr: inst, se %d, data 0x%x, sa %d, priv %d, wave_id %d, simd_id %d, wgp_id %d\n",
+					REG_GET_FIELD(
+						context_id1,
+						SQ_INTERRUPT_WORD_WAVE_CTXID1,
+						SE_ID),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID0,
+						DATA),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID0,
+						SA_ID),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID0,
+						PRIV),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID0,
+						WAVE_ID),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID0,
+						SIMD_ID),
+					REG_GET_FIELD(
+						context_id1,
+						SQ_INTERRUPT_WORD_WAVE_CTXID1,
+						WGP_ID));
+				if (context_id0 & SQ_INTERRUPT_WORD_WAVE_CTXID0__PRIV_MASK) {
+					if (kfd_set_dbg_ev_from_interrupt(dev, pasid,
+							KFD_DEBUG_DOORBELL_ID(context_id0),
+							KFD_DEBUG_TRAP_CODE(context_id0),
+							NULL, 0))
+						return;
+				}
+				break;
+			case SQ_INTERRUPT_WORD_ENCODING_ERROR:
+				sq_intr_err_type = REG_GET_FIELD(context_id0, KFD_CTXID0,
+								ERR_TYPE);
+				dev_warn_ratelimited(
+					dev->adev->dev,
+					"sq_intr: error, se %d, data 0x%x, sa %d, priv %d, wave_id %d, simd_id %d, wgp_id %d, err_type %d\n",
+					REG_GET_FIELD(
+						context_id1,
+						SQ_INTERRUPT_WORD_WAVE_CTXID1,
+						SE_ID),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID0,
+						DATA),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID0,
+						SA_ID),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID0,
+						PRIV),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID0,
+						WAVE_ID),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID0,
+						SIMD_ID),
+					REG_GET_FIELD(
+						context_id1,
+						SQ_INTERRUPT_WORD_WAVE_CTXID1,
+						WGP_ID),
+					sq_intr_err_type);
+				break;
+			default:
+				break;
+			}
+			kfd_signal_event_interrupt(pasid, context_id0 & 0x7fffff, 23);
+		} else if (source_id == SOC15_INTSRC_CP_BAD_OPCODE &&
+			   KFD_DBG_EC_TYPE_IS_PACKET(KFD_DEBUG_CP_BAD_OP_ECODE(context_id0))) {
+			kfd_set_dbg_ev_from_interrupt(dev, pasid,
+				KFD_DEBUG_DOORBELL_ID(context_id0),
+				KFD_EC_MASK(KFD_DEBUG_CP_BAD_OP_ECODE(context_id0)),
+				NULL,
+				0);
+		}
+	} else if (client_id == SOC15_IH_CLIENTID_SDMA0 ||
+		   client_id == SOC15_IH_CLIENTID_SDMA1 ||
+		   client_id == SOC15_IH_CLIENTID_SDMA2 ||
+		   client_id == SOC15_IH_CLIENTID_SDMA3 ||
+		   (client_id == SOC15_IH_CLIENTID_SDMA3_Sienna_Cichlid &&
+		    KFD_GC_VERSION(dev) == IP_VERSION(10, 3, 0)) ||
+		   client_id == SOC15_IH_CLIENTID_SDMA4 ||
+		   client_id == SOC15_IH_CLIENTID_SDMA5 ||
+		   client_id == SOC15_IH_CLIENTID_SDMA6 ||
+		   client_id == SOC15_IH_CLIENTID_SDMA7) {
+		if (source_id == SOC15_INTSRC_SDMA_TRAP) {
+			kfd_signal_event_interrupt(pasid, context_id0 & 0xfffffff, 28);
+		}
+	} else if (client_id == SOC15_IH_CLIENTID_VMC ||
+		   client_id == SOC15_IH_CLIENTID_VMC1 ||
+		   client_id == SOC15_IH_CLIENTID_UTCL2) {
+		struct kfd_vm_fault_info info = {0};
+		uint16_t ring_id = SOC15_RING_ID_FROM_IH_ENTRY(ih_ring_entry);
+		struct kfd_hsa_memory_exception_data exception_data;
+
+		info.vmid = vmid;
+		info.mc_id = client_id;
+		info.page_addr = ih_ring_entry[4] |
+			(uint64_t)(ih_ring_entry[5] & 0xf) << 32;
+		info.prot_valid = ring_id & 0x08;
+		info.prot_read  = ring_id & 0x10;
+		info.prot_write = ring_id & 0x20;
+
+		memset(&exception_data, 0, sizeof(exception_data));
+		exception_data.gpu_id = dev->id;
+		exception_data.va = (info.page_addr) << PAGE_SHIFT;
+		exception_data.failure.NotPresent = info.prot_valid ? 1 : 0;
+		exception_data.failure.NoExecute = info.prot_exec ? 1 : 0;
+		exception_data.failure.ReadOnly = info.prot_write ? 1 : 0;
+		exception_data.failure.imprecise = 0;
+
+		kfd_set_dbg_ev_from_interrupt(dev,
+						pasid,
+						-1,
+						KFD_EC_MASK(EC_DEVICE_MEMORY_VIOLATION),
+						&exception_data,
+						sizeof(exception_data));
+	} else if (KFD_IRQ_IS_FENCE(client_id, source_id)) {
+		kfd_process_close_interrupt_drain(pasid);
+	}
+}
+
+const struct kfd_event_interrupt_class event_interrupt_class_v10 = {
+	.interrupt_isr = event_interrupt_isr_v10,
+	.interrupt_wq = event_interrupt_wq_v10,
+};
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v11.c b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v11.c
index 0d53f6067422..2788a52714d1 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v11.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v11.c
@@ -26,6 +26,7 @@
 #include "kfd_device_queue_manager.h"
 #include "ivsrcid/vmc/irqsrcs_vmc_1_0.h"
 #include "kfd_smi_events.h"
+#include "kfd_debug.h"
 
 /*
  * GFX11 SQ Interrupts
@@ -147,51 +148,78 @@ enum SQ_INTERRUPT_ERROR_TYPE {
 #define KFD_CTXID0_DOORBELL_ID(ctxid0)		((ctxid0) & \
 				KFD_CTXID0_DOORBELL_ID_MASK)
 
-static void print_sq_intr_info_auto(uint32_t context_id0, uint32_t context_id1)
+static void print_sq_intr_info_auto(struct kfd_node *dev, uint32_t context_id0,
+				    uint32_t context_id1)
 {
-	pr_debug(
+	dev_dbg_ratelimited(
+		dev->adev->dev,
 		"sq_intr: auto, ttrace %d, wlt %d, ttrace_buf_full %d, reg_tms %d, cmd_tms %d, host_cmd_ovf %d, host_reg_ovf %d, immed_ovf %d, ttrace_utc_err %d\n",
-		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0, THREAD_TRACE),
+		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0,
+			      THREAD_TRACE),
 		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0, WLT),
-		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0, THREAD_TRACE_BUF_FULL),
-		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0, REG_TIMESTAMP),
-		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0, CMD_TIMESTAMP),
-		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0, HOST_CMD_OVERFLOW),
-		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0, HOST_REG_OVERFLOW),
-		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0, IMMED_OVERFLOW),
-		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0, THREAD_TRACE_UTC_ERROR));
+		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0,
+			      THREAD_TRACE_BUF_FULL),
+		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0,
+			      REG_TIMESTAMP),
+		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0,
+			      CMD_TIMESTAMP),
+		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0,
+			      HOST_CMD_OVERFLOW),
+		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0,
+			      HOST_REG_OVERFLOW),
+		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0,
+			      IMMED_OVERFLOW),
+		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID0,
+			      THREAD_TRACE_UTC_ERROR));
 }
 
-static void print_sq_intr_info_inst(uint32_t context_id0, uint32_t context_id1)
+static void print_sq_intr_info_inst(struct kfd_node *dev, uint32_t context_id0,
+				    uint32_t context_id1)
 {
-	pr_debug(
+	dev_dbg_ratelimited(
+		dev->adev->dev,
 		"sq_intr: inst, data 0x%08x, sh %d, priv %d, wave_id %d, simd_id %d, wgp_id %d\n",
 		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID0, DATA),
-		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID0, SH_ID),
+		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID0,
+			      SH_ID),
 		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID0, PRIV),
-		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID0, WAVE_ID),
-		REG_GET_FIELD(context_id1, SQ_INTERRUPT_WORD_WAVE_CTXID1, SIMD_ID),
-		REG_GET_FIELD(context_id1, SQ_INTERRUPT_WORD_WAVE_CTXID1, WGP_ID));
+		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID0,
+			      WAVE_ID),
+		REG_GET_FIELD(context_id1, SQ_INTERRUPT_WORD_WAVE_CTXID1,
+			      SIMD_ID),
+		REG_GET_FIELD(context_id1, SQ_INTERRUPT_WORD_WAVE_CTXID1,
+			      WGP_ID));
 }
 
-static void print_sq_intr_info_error(uint32_t context_id0, uint32_t context_id1)
+static void print_sq_intr_info_error(struct kfd_node *dev, uint32_t context_id0,
+				     uint32_t context_id1)
 {
-	pr_warn(
+	dev_warn_ratelimited(
+		dev->adev->dev,
 		"sq_intr: error, detail 0x%08x, type %d, sh %d, priv %d, wave_id %d, simd_id %d, wgp_id %d\n",
-		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_ERROR_CTXID0, DETAIL),
-		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_ERROR_CTXID0, TYPE),
-		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_ERROR_CTXID0, SH_ID),
-		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_ERROR_CTXID0, PRIV),
-		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_ERROR_CTXID0, WAVE_ID),
-		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_ERROR_CTXID1, SIMD_ID),
-		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_ERROR_CTXID1, WGP_ID));
+		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_ERROR_CTXID0,
+			      DETAIL),
+		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_ERROR_CTXID0,
+			      TYPE),
+		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_ERROR_CTXID0,
+			      SH_ID),
+		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_ERROR_CTXID0,
+			      PRIV),
+		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_ERROR_CTXID0,
+			      WAVE_ID),
+		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_ERROR_CTXID1,
+			      SIMD_ID),
+		REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_ERROR_CTXID1,
+			      WGP_ID));
 }
 
-static void event_interrupt_poison_consumption_v11(struct kfd_dev *dev,
+static void event_interrupt_poison_consumption_v11(struct kfd_node *dev,
 				uint16_t pasid, uint16_t source_id)
 {
+	enum amdgpu_ras_block block = 0;
 	int ret = -EINVAL;
-	struct kfd_process *p = kfd_lookup_process_by_pasid(pasid);
+	uint32_t reset = 0;
+	struct kfd_process *p = kfd_lookup_process_by_pasid(pasid, NULL);
 
 	if (!p)
 		return;
@@ -209,9 +237,14 @@ static void event_interrupt_poison_consumption_v11(struct kfd_dev *dev,
 	case SOC15_INTSRC_SQ_INTERRUPT_MSG:
 		if (dev->dqm->ops.reset_queues)
 			ret = dev->dqm->ops.reset_queues(dev->dqm, pasid);
+		block = AMDGPU_RAS_BLOCK__GFX;
+		if (ret)
+			reset = AMDGPU_RAS_GPU_RESET_MODE2_RESET;
 		break;
 	case SOC21_INTSRC_SDMA_ECC:
 	default:
+		block = AMDGPU_RAS_BLOCK__GFX;
+		reset = AMDGPU_RAS_GPU_RESET_MODE2_RESET;
 		break;
 	}
 
@@ -219,13 +252,10 @@ static void event_interrupt_poison_consumption_v11(struct kfd_dev *dev,
 
 	/* resetting queue passes, do page retirement without gpu reset
 	   resetting queue fails, fallback to gpu reset solution */
-	if (!ret)
-		amdgpu_amdkfd_ras_poison_consumption_handler(dev->adev, false);
-	else
-		amdgpu_amdkfd_ras_poison_consumption_handler(dev->adev, true);
+	amdgpu_amdkfd_ras_poison_consumption_handler(dev->adev, block, reset);
 }
 
-static bool event_interrupt_isr_v11(struct kfd_dev *dev,
+static bool event_interrupt_isr_v11(struct kfd_node *dev,
 					const uint32_t *ih_ring_entry,
 					uint32_t *patched_ihre,
 					bool *patched_flag)
@@ -238,7 +268,7 @@ static bool event_interrupt_isr_v11(struct kfd_dev *dev,
 	client_id = SOC15_CLIENT_ID_FROM_IH_ENTRY(ih_ring_entry);
 	/* Only handle interrupts from KFD VMIDs */
 	vmid = SOC15_VMID_FROM_IH_ENTRY(ih_ring_entry);
-	if (/*!KFD_IRQ_IS_FENCE(client_id, source_id) &&*/
+	if (!KFD_IRQ_IS_FENCE(client_id, source_id) &&
 	    (vmid < dev->vm_info.first_vmid_kfd ||
 	    vmid > dev->vm_info.last_vmid_kfd))
 		return false;
@@ -250,14 +280,14 @@ static bool event_interrupt_isr_v11(struct kfd_dev *dev,
 	    (context_id0 & AMDGPU_FENCE_MES_QUEUE_FLAG))
 		return false;
 
-	pr_debug("client id 0x%x, source id %d, vmid %d, pasid 0x%x. raw data:\n",
-		 client_id, source_id, vmid, pasid);
-	pr_debug("%8X, %8X, %8X, %8X, %8X, %8X, %8X, %8X.\n",
-		 data[0], data[1], data[2], data[3],
-		 data[4], data[5], data[6], data[7]);
+	dev_dbg(dev->adev->dev,
+		"client id 0x%x, source id %d, vmid %d, pasid 0x%x. raw data:\n",
+		client_id, source_id, vmid, pasid);
+	dev_dbg(dev->adev->dev, "%8X, %8X, %8X, %8X, %8X, %8X, %8X, %8X.\n",
+		data[0], data[1], data[2], data[3], data[4], data[5], data[6],
+		data[7]);
 
-	/* If there is no valid PASID, it's likely a bug */
-	if (WARN_ONCE(pasid == 0, "Bug: No PASID in KFD interrupt"))
+	if (pasid == 0)
 		return false;
 
 	/* Interrupt types we care about: various signals and faults.
@@ -267,19 +297,19 @@ static bool event_interrupt_isr_v11(struct kfd_dev *dev,
 		source_id == SOC15_INTSRC_SQ_INTERRUPT_MSG ||
 		source_id == SOC15_INTSRC_CP_BAD_OPCODE ||
 		source_id == SOC21_INTSRC_SDMA_TRAP ||
-		/* KFD_IRQ_IS_FENCE(client_id, source_id) || */
+		KFD_IRQ_IS_FENCE(client_id, source_id) ||
 		(((client_id == SOC21_IH_CLIENTID_VMC) ||
 		 ((client_id == SOC21_IH_CLIENTID_GFX) &&
 		  (source_id == UTCL2_1_0__SRCID__FAULT))) &&
 		  !amdgpu_no_queue_eviction_on_vm_fault);
 }
 
-static void event_interrupt_wq_v11(struct kfd_dev *dev,
+static void event_interrupt_wq_v11(struct kfd_node *dev,
 					const uint32_t *ih_ring_entry)
 {
 	uint16_t source_id, client_id, ring_id, pasid, vmid;
 	uint32_t context_id0, context_id1;
-	uint8_t sq_int_enc, sq_int_errtype, sq_int_priv;
+	uint8_t sq_int_enc, sq_int_priv, sq_int_errtype;
 	struct kfd_vm_fault_info info = {0};
 	struct kfd_hsa_memory_exception_data exception_data;
 
@@ -312,9 +342,9 @@ static void event_interrupt_wq_v11(struct kfd_dev *dev,
 		exception_data.failure.ReadOnly = info.prot_write ? 1 : 0;
 		exception_data.failure.imprecise = 0;
 
-		/*kfd_set_dbg_ev_from_interrupt(dev, pasid, -1,
+		kfd_set_dbg_ev_from_interrupt(dev, pasid, -1,
 					      KFD_EC_MASK(EC_DEVICE_MEMORY_VIOLATION),
-					      &exception_data, sizeof(exception_data));*/
+					      &exception_data, sizeof(exception_data));
 		kfd_smi_event_update_vmfault(dev, pasid);
 
 	/* GRBM, SDMA, SE, PMM */
@@ -324,11 +354,15 @@ static void event_interrupt_wq_v11(struct kfd_dev *dev,
 		/* CP */
 		if (source_id == SOC15_INTSRC_CP_END_OF_PIPE)
 			kfd_signal_event_interrupt(pasid, context_id0, 32);
-		/*else if (source_id == SOC15_INTSRC_CP_BAD_OPCODE)
-			kfd_set_dbg_ev_from_interrupt(dev, pasid,
-				KFD_CTXID0_DOORBELL_ID(context_id0),
+		else if (source_id == SOC15_INTSRC_CP_BAD_OPCODE &&
+			 KFD_DBG_EC_TYPE_IS_PACKET(KFD_CTXID0_CP_BAD_OP_ECODE(context_id0))) {
+			u32 doorbell_id = KFD_CTXID0_DOORBELL_ID(context_id0);
+
+			kfd_set_dbg_ev_from_interrupt(dev, pasid, doorbell_id,
 				KFD_EC_MASK(KFD_CTXID0_CP_BAD_OP_ECODE(context_id0)),
-				NULL, 0);*/
+				NULL, 0);
+			kfd_dqm_suspend_bad_queue_mes(dev, pasid, doorbell_id);
+		}
 
 		/* SDMA */
 		else if (source_id == SOC21_INTSRC_SDMA_TRAP)
@@ -344,20 +378,20 @@ static void event_interrupt_wq_v11(struct kfd_dev *dev,
 					SQ_INTERRUPT_WORD_WAVE_CTXID1, ENCODING);
 			switch (sq_int_enc) {
 			case SQ_INTERRUPT_WORD_ENCODING_AUTO:
-				print_sq_intr_info_auto(context_id0, context_id1);
+				print_sq_intr_info_auto(dev, context_id0, context_id1);
 				break;
 			case SQ_INTERRUPT_WORD_ENCODING_INST:
-				print_sq_intr_info_inst(context_id0, context_id1);
+				print_sq_intr_info_inst(dev, context_id0, context_id1);
 				sq_int_priv = REG_GET_FIELD(context_id0,
 						SQ_INTERRUPT_WORD_WAVE_CTXID0, PRIV);
-				/*if (sq_int_priv && (kfd_set_dbg_ev_from_interrupt(dev, pasid,
+				if (sq_int_priv && (kfd_set_dbg_ev_from_interrupt(dev, pasid,
 						KFD_CTXID0_DOORBELL_ID(context_id0),
 						KFD_CTXID0_TRAP_CODE(context_id0),
 						NULL, 0)))
-					return;*/
+					return;
 				break;
 			case SQ_INTERRUPT_WORD_ENCODING_ERROR:
-				print_sq_intr_info_error(context_id0, context_id1);
+				print_sq_intr_info_error(dev, context_id0, context_id1);
 				sq_int_errtype = REG_GET_FIELD(context_id0,
 						SQ_INTERRUPT_WORD_ERROR_CTXID0, TYPE);
 				if (sq_int_errtype != SQ_INTERRUPT_ERROR_TYPE_ILLEGAL_INST &&
@@ -373,8 +407,8 @@ static void event_interrupt_wq_v11(struct kfd_dev *dev,
 			kfd_signal_event_interrupt(pasid, context_id0 & 0xffffff, 24);
 		}
 
-	/*} else if (KFD_IRQ_IS_FENCE(client_id, source_id)) {
-		kfd_process_close_interrupt_drain(pasid);*/
+	} else if (KFD_IRQ_IS_FENCE(client_id, source_id)) {
+		kfd_process_close_interrupt_drain(pasid);
 	}
 }
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c
index 0b75a37b689b..d76fb61869c7 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c
@@ -23,9 +23,41 @@
 
 #include "kfd_priv.h"
 #include "kfd_events.h"
+#include "kfd_debug.h"
 #include "soc15_int.h"
 #include "kfd_device_queue_manager.h"
 #include "kfd_smi_events.h"
+#include "amdgpu_ras.h"
+#include "amdgpu_ras_mgr.h"
+
+/*
+ * GFX9 SQ Interrupts
+ *
+ * There are 3 encoding types of interrupts sourced from SQ sent as a 44-bit
+ * packet to the Interrupt Handler:
+ * Auto - Generated by the SQG (various cmd overflows, timestamps etc)
+ * Wave - Generated by S_SENDMSG through a shader program
+ * Error - HW generated errors (Illegal instructions, Memviols, EDC etc)
+ *
+ * The 44-bit packet is mapped as {context_id1[7:0],context_id0[31:0]} plus
+ * 4-bits for VMID (SOC15_VMID_FROM_IH_ENTRY) as such:
+ *
+ * - context_id0[27:26]
+ * Encoding type (0 = Auto, 1 = Wave, 2 = Error)
+ *
+ * - context_id0[13]
+ * PRIV bit indicates that Wave S_SEND or error occurred within trap
+ *
+ * - {context_id1[7:0],context_id0[31:28],context_id0[11:0]}
+ * 24-bit data with the following layout per encoding type:
+ * Auto - only context_id0[8:0] is used, which reports various interrupts
+ * generated by SQG.  The rest is 0.
+ * Wave - user data sent from m0 via S_SENDMSG
+ * Error - Error type (context_id1[7:4]), Error Details (rest of bits)
+ *
+ * The other context_id bits show coordinates (SE/SH/CU/SIMD/WAVE) for wave
+ * S_SENDMSG and Errors.  These are 0 for Auto.
+ */
 
 enum SQ_INTERRUPT_WORD_ENCODING {
 	SQ_INTERRUPT_WORD_ENCODING_AUTO = 0x0,
@@ -84,17 +116,41 @@ enum SQ_INTERRUPT_ERROR_TYPE {
 #define SQ_INTERRUPT_WORD_WAVE_CTXID__SE_ID_MASK 0x03000000
 #define SQ_INTERRUPT_WORD_WAVE_CTXID__ENCODING_MASK 0x0c000000
 
+/* GFX9 SQ interrupt 24-bit data from context_id<0,1> */
 #define KFD_CONTEXT_ID_GET_SQ_INT_DATA(ctx0, ctx1)                             \
 	((ctx0 & 0xfff) | ((ctx0 >> 16) & 0xf000) | ((ctx1 << 16) & 0xff0000))
 
 #define KFD_SQ_INT_DATA__ERR_TYPE_MASK 0xF00000
 #define KFD_SQ_INT_DATA__ERR_TYPE__SHIFT 20
 
-static void event_interrupt_poison_consumption_v9(struct kfd_dev *dev,
+/*
+ * The debugger will send user data(m0) with PRIV=1 to indicate it requires
+ * notification from the KFD with the following queue id (DOORBELL_ID) and
+ * trap code (TRAP_CODE).
+ */
+#define KFD_INT_DATA_DEBUG_DOORBELL_MASK	0x0003ff
+#define KFD_INT_DATA_DEBUG_TRAP_CODE_SHIFT	10
+#define KFD_INT_DATA_DEBUG_TRAP_CODE_MASK	0x07fc00
+#define KFD_DEBUG_DOORBELL_ID(sq_int_data)	((sq_int_data) &	\
+				KFD_INT_DATA_DEBUG_DOORBELL_MASK)
+#define KFD_DEBUG_TRAP_CODE(sq_int_data)	(((sq_int_data) &	\
+				KFD_INT_DATA_DEBUG_TRAP_CODE_MASK)	\
+				>> KFD_INT_DATA_DEBUG_TRAP_CODE_SHIFT)
+#define KFD_DEBUG_CP_BAD_OP_ECODE_MASK		0x3fffc00
+#define KFD_DEBUG_CP_BAD_OP_ECODE_SHIFT		10
+#define KFD_DEBUG_CP_BAD_OP_ECODE(ctxid0)	(((ctxid0) &		\
+				KFD_DEBUG_CP_BAD_OP_ECODE_MASK)		\
+				>> KFD_DEBUG_CP_BAD_OP_ECODE_SHIFT)
+
+static void event_interrupt_poison_consumption_v9(struct kfd_node *dev,
 				uint16_t pasid, uint16_t client_id)
 {
-	int old_poison, ret = -EINVAL;
-	struct kfd_process *p = kfd_lookup_process_by_pasid(pasid);
+	enum amdgpu_ras_block block = 0;
+	uint32_t reset = 0;
+	struct kfd_process *p = kfd_lookup_process_by_pasid(pasid, NULL);
+	enum ras_event_type type = RAS_EVENT_TYPE_POISON_CONSUMPTION;
+	u64 event_id;
+	int old_poison, ret;
 
 	if (!p)
 		return;
@@ -111,34 +167,79 @@ static void event_interrupt_poison_consumption_v9(struct kfd_dev *dev,
 	case SOC15_IH_CLIENTID_SE2SH:
 	case SOC15_IH_CLIENTID_SE3SH:
 	case SOC15_IH_CLIENTID_UTCL2:
-		ret = kfd_dqm_evict_pasid(dev->dqm, pasid);
+		block = AMDGPU_RAS_BLOCK__GFX;
+		if (amdgpu_ip_version(dev->adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3)) {
+			/* driver mode-2 for gfx poison is only supported by
+			 * pmfw 0x00557300 and onwards */
+			if (dev->adev->pm.fw_version < 0x00557300)
+				reset = AMDGPU_RAS_GPU_RESET_MODE1_RESET;
+			else
+				reset = AMDGPU_RAS_GPU_RESET_MODE2_RESET;
+		} else if (amdgpu_ip_version(dev->adev, GC_HWIP, 0) == IP_VERSION(9, 4, 4)) {
+			/* driver mode-2 for gfx poison is only supported by
+			 * pmfw 0x05550C00 and onwards */
+			if (dev->adev->pm.fw_version < 0x05550C00)
+				reset = AMDGPU_RAS_GPU_RESET_MODE1_RESET;
+			else
+				reset = AMDGPU_RAS_GPU_RESET_MODE2_RESET;
+		} else {
+			reset = AMDGPU_RAS_GPU_RESET_MODE2_RESET;
+		}
+		amdgpu_ras_set_err_poison(dev->adev, AMDGPU_RAS_BLOCK__GFX);
+		break;
+	case SOC15_IH_CLIENTID_VMC:
+	case SOC15_IH_CLIENTID_VMC1:
+		block = AMDGPU_RAS_BLOCK__MMHUB;
+		reset = AMDGPU_RAS_GPU_RESET_MODE1_RESET;
 		break;
 	case SOC15_IH_CLIENTID_SDMA0:
 	case SOC15_IH_CLIENTID_SDMA1:
 	case SOC15_IH_CLIENTID_SDMA2:
 	case SOC15_IH_CLIENTID_SDMA3:
 	case SOC15_IH_CLIENTID_SDMA4:
+		block = AMDGPU_RAS_BLOCK__SDMA;
+		if (amdgpu_ip_version(dev->adev, SDMA0_HWIP, 0) == IP_VERSION(4, 4, 2)) {
+			/* driver mode-2 for gfx poison is only supported by
+			 * pmfw 0x00557300 and onwards */
+			if (dev->adev->pm.fw_version < 0x00557300)
+				reset = AMDGPU_RAS_GPU_RESET_MODE1_RESET;
+			else
+				reset = AMDGPU_RAS_GPU_RESET_MODE2_RESET;
+		} else if (amdgpu_ip_version(dev->adev, SDMA0_HWIP, 0) == IP_VERSION(4, 4, 5)) {
+			/* driver mode-2 for gfx poison is only supported by
+			 * pmfw 0x05550C00 and onwards */
+			if (dev->adev->pm.fw_version < 0x05550C00)
+				reset = AMDGPU_RAS_GPU_RESET_MODE1_RESET;
+			else
+				reset = AMDGPU_RAS_GPU_RESET_MODE2_RESET;
+		} else {
+			reset = AMDGPU_RAS_GPU_RESET_MODE2_RESET;
+		}
+		amdgpu_ras_set_err_poison(dev->adev, AMDGPU_RAS_BLOCK__SDMA);
 		break;
 	default:
-		break;
+		dev_warn(dev->adev->dev,
+			 "client %d does not support poison consumption\n", client_id);
+		return;
 	}
 
+	ret = amdgpu_ras_mark_ras_event(dev->adev, type);
+	if (ret)
+		return;
+
 	kfd_signal_poison_consumed_event(dev, pasid);
 
-	/* resetting queue passes, do page retirement without gpu reset
-	 * resetting queue fails, fallback to gpu reset solution
-	 */
-	if (!ret) {
-		dev_warn(dev->adev->dev,
-			"RAS poison consumption, unmap queue flow succeeded: client id %d\n",
-			client_id);
-		amdgpu_amdkfd_ras_poison_consumption_handler(dev->adev, false);
-	} else {
-		dev_warn(dev->adev->dev,
-			"RAS poison consumption, fall back to gpu reset flow: client id %d\n",
-			client_id);
-		amdgpu_amdkfd_ras_poison_consumption_handler(dev->adev, true);
-	}
+	if (amdgpu_uniras_enabled(dev->adev))
+		event_id = amdgpu_ras_mgr_gen_ras_event_seqno(dev->adev,
+					RAS_SEQNO_TYPE_POISON_CONSUMPTION);
+	else
+		event_id = amdgpu_ras_acquire_event_id(dev->adev, type);
+
+	RAS_EVENT_LOG(dev->adev, event_id,
+		      "poison is consumed by client %d, kick off gpu reset flow\n", client_id);
+
+	amdgpu_amdkfd_ras_pasid_poison_consumption_handler(dev->adev,
+		block, pasid, NULL, NULL, reset);
 }
 
 static bool context_id_expected(struct kfd_dev *dev)
@@ -160,7 +261,7 @@ static bool context_id_expected(struct kfd_dev *dev)
 	}
 }
 
-static bool event_interrupt_isr_v9(struct kfd_dev *dev,
+static bool event_interrupt_isr_v9(struct kfd_node *dev,
 					const uint32_t *ih_ring_entry,
 					uint32_t *patched_ihre,
 					bool *patched_flag)
@@ -168,14 +269,16 @@ static bool event_interrupt_isr_v9(struct kfd_dev *dev,
 	uint16_t source_id, client_id, pasid, vmid;
 	const uint32_t *data = ih_ring_entry;
 
+	source_id = SOC15_SOURCE_ID_FROM_IH_ENTRY(ih_ring_entry);
+	client_id = SOC15_CLIENT_ID_FROM_IH_ENTRY(ih_ring_entry);
+
 	/* Only handle interrupts from KFD VMIDs */
 	vmid = SOC15_VMID_FROM_IH_ENTRY(ih_ring_entry);
-	if (vmid < dev->vm_info.first_vmid_kfd ||
-	    vmid > dev->vm_info.last_vmid_kfd)
+	if (!KFD_IRQ_IS_FENCE(client_id, source_id) &&
+	   (vmid < dev->vm_info.first_vmid_kfd ||
+	    vmid > dev->vm_info.last_vmid_kfd))
 		return false;
 
-	source_id = SOC15_SOURCE_ID_FROM_IH_ENTRY(ih_ring_entry);
-	client_id = SOC15_CLIENT_ID_FROM_IH_ENTRY(ih_ring_entry);
 	pasid = SOC15_PASID_FROM_IH_ENTRY(ih_ring_entry);
 
 	/* Only handle clients we care about */
@@ -194,7 +297,8 @@ static bool event_interrupt_isr_v9(struct kfd_dev *dev,
 	    client_id != SOC15_IH_CLIENTID_SE0SH &&
 	    client_id != SOC15_IH_CLIENTID_SE1SH &&
 	    client_id != SOC15_IH_CLIENTID_SE2SH &&
-	    client_id != SOC15_IH_CLIENTID_SE3SH)
+	    client_id != SOC15_IH_CLIENTID_SE3SH &&
+	    !KFD_IRQ_IS_FENCE(client_id, source_id))
 		return false;
 
 	/* This is a known issue for gfx9. Under non HWS, pasid is not set
@@ -206,7 +310,7 @@ static bool event_interrupt_isr_v9(struct kfd_dev *dev,
 
 		*patched_flag = true;
 		memcpy(patched_ihre, ih_ring_entry,
-				dev->device_info.ih_ring_entry_size);
+				dev->kfd->device_info.ih_ring_entry_size);
 
 		pasid = dev->dqm->vmid_pasid[vmid];
 
@@ -215,11 +319,12 @@ static bool event_interrupt_isr_v9(struct kfd_dev *dev,
 					& ~pasid_mask) | pasid);
 	}
 
-	pr_debug("client id 0x%x, source id %d, vmid %d, pasid 0x%x. raw data:\n",
-		 client_id, source_id, vmid, pasid);
-	pr_debug("%8X, %8X, %8X, %8X, %8X, %8X, %8X, %8X.\n",
-		 data[0], data[1], data[2], data[3],
-		 data[4], data[5], data[6], data[7]);
+	dev_dbg(dev->adev->dev,
+		"client id 0x%x, source id %d, vmid %d, pasid 0x%x. raw data:\n",
+		client_id, source_id, vmid, pasid);
+	dev_dbg(dev->adev->dev, "%8X, %8X, %8X, %8X, %8X, %8X, %8X, %8X.\n",
+		data[0], data[1], data[2], data[3], data[4], data[5], data[6],
+		data[7]);
 
 	/* If there is no valid PASID, it's likely a bug */
 	if (WARN_ONCE(pasid == 0, "Bug: No PASID in KFD interrupt"))
@@ -235,7 +340,7 @@ static bool event_interrupt_isr_v9(struct kfd_dev *dev,
 		uint32_t context_id =
 			SOC15_CONTEXT_ID0_FROM_IH_ENTRY(ih_ring_entry);
 
-		if (context_id == 0 && context_id_expected(dev))
+		if (context_id == 0 && context_id_expected(dev->kfd))
 			return false;
 	}
 
@@ -247,13 +352,14 @@ static bool event_interrupt_isr_v9(struct kfd_dev *dev,
 		source_id == SOC15_INTSRC_SDMA_ECC ||
 		source_id == SOC15_INTSRC_SQ_INTERRUPT_MSG ||
 		source_id == SOC15_INTSRC_CP_BAD_OPCODE ||
+		KFD_IRQ_IS_FENCE(client_id, source_id) ||
 		((client_id == SOC15_IH_CLIENTID_VMC ||
 		client_id == SOC15_IH_CLIENTID_VMC1 ||
 		client_id == SOC15_IH_CLIENTID_UTCL2) &&
 		!amdgpu_no_queue_eviction_on_vm_fault);
 }
 
-static void event_interrupt_wq_v9(struct kfd_dev *dev,
+static void event_interrupt_wq_v9(struct kfd_node *dev,
 					const uint32_t *ih_ring_entry)
 {
 	uint16_t source_id, client_id, pasid, vmid;
@@ -279,40 +385,124 @@ static void event_interrupt_wq_v9(struct kfd_dev *dev,
 			encoding = REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID, ENCODING);
 			switch (encoding) {
 			case SQ_INTERRUPT_WORD_ENCODING_AUTO:
-				pr_debug(
+				dev_dbg_ratelimited(
+					dev->adev->dev,
 					"sq_intr: auto, se %d, ttrace %d, wlt %d, ttrac_buf_full %d, reg_tms %d, cmd_tms %d, host_cmd_ovf %d, host_reg_ovf %d, immed_ovf %d, ttrace_utc_err %d\n",
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID, SE_ID),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID, THREAD_TRACE),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID, WLT),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID, THREAD_TRACE_BUF_FULL),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID, REG_TIMESTAMP),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID, CMD_TIMESTAMP),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID, HOST_CMD_OVERFLOW),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID, HOST_REG_OVERFLOW),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID, IMMED_OVERFLOW),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_AUTO_CTXID, THREAD_TRACE_UTC_ERROR));
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_AUTO_CTXID,
+						SE_ID),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_AUTO_CTXID,
+						THREAD_TRACE),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_AUTO_CTXID,
+						WLT),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_AUTO_CTXID,
+						THREAD_TRACE_BUF_FULL),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_AUTO_CTXID,
+						REG_TIMESTAMP),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_AUTO_CTXID,
+						CMD_TIMESTAMP),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_AUTO_CTXID,
+						HOST_CMD_OVERFLOW),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_AUTO_CTXID,
+						HOST_REG_OVERFLOW),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_AUTO_CTXID,
+						IMMED_OVERFLOW),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_AUTO_CTXID,
+						THREAD_TRACE_UTC_ERROR));
 				break;
 			case SQ_INTERRUPT_WORD_ENCODING_INST:
-				pr_debug("sq_intr: inst, se %d, data 0x%x, sh %d, priv %d, wave_id %d, simd_id %d, cu_id %d, intr_data 0x%x\n",
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID, SE_ID),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID, DATA),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID, SH_ID),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID, PRIV),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID, WAVE_ID),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID, SIMD_ID),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID, CU_ID),
+				dev_dbg_ratelimited(
+					dev->adev->dev,
+					"sq_intr: inst, se %d, data 0x%x, sh %d, priv %d, wave_id %d, simd_id %d, cu_id %d, intr_data 0x%x\n",
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID,
+						SE_ID),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID,
+						DATA),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID,
+						SH_ID),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID,
+						PRIV),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID,
+						WAVE_ID),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID,
+						SIMD_ID),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID,
+						CU_ID),
 					sq_int_data);
+				if (context_id0 & SQ_INTERRUPT_WORD_WAVE_CTXID__PRIV_MASK) {
+					if (kfd_set_dbg_ev_from_interrupt(dev, pasid,
+							KFD_DEBUG_DOORBELL_ID(sq_int_data),
+							KFD_DEBUG_TRAP_CODE(sq_int_data),
+							NULL, 0))
+						return;
+				}
 				break;
 			case SQ_INTERRUPT_WORD_ENCODING_ERROR:
 				sq_intr_err = REG_GET_FIELD(sq_int_data, KFD_SQ_INT_DATA, ERR_TYPE);
-				pr_warn("sq_intr: error, se %d, data 0x%x, sh %d, priv %d, wave_id %d, simd_id %d, cu_id %d, err_type %d\n",
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID, SE_ID),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID, DATA),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID, SH_ID),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID, PRIV),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID, WAVE_ID),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID, SIMD_ID),
-					REG_GET_FIELD(context_id0, SQ_INTERRUPT_WORD_WAVE_CTXID, CU_ID),
+				dev_warn_ratelimited(
+					dev->adev->dev,
+					"sq_intr: error, se %d, data 0x%x, sh %d, priv %d, wave_id %d, simd_id %d, cu_id %d, err_type %d\n",
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID,
+						SE_ID),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID,
+						DATA),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID,
+						SH_ID),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID,
+						PRIV),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID,
+						WAVE_ID),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID,
+						SIMD_ID),
+					REG_GET_FIELD(
+						context_id0,
+						SQ_INTERRUPT_WORD_WAVE_CTXID,
+						CU_ID),
 					sq_intr_err);
 				if (sq_intr_err != SQ_INTERRUPT_ERROR_TYPE_ILLEGAL_INST &&
 					sq_intr_err != SQ_INTERRUPT_ERROR_TYPE_MEMVIOL) {
@@ -323,9 +513,14 @@ static void event_interrupt_wq_v9(struct kfd_dev *dev,
 			default:
 				break;
 			}
-			kfd_signal_event_interrupt(pasid, context_id0 & 0xffffff, 24);
-		} else if (source_id == SOC15_INTSRC_CP_BAD_OPCODE)
-			kfd_signal_hw_exception_event(pasid);
+			kfd_signal_event_interrupt(pasid, sq_int_data, 24);
+		} else if (source_id == SOC15_INTSRC_CP_BAD_OPCODE &&
+			   KFD_DBG_EC_TYPE_IS_PACKET(KFD_DEBUG_CP_BAD_OP_ECODE(context_id0))) {
+			kfd_set_dbg_ev_from_interrupt(dev, pasid,
+				KFD_DEBUG_DOORBELL_ID(context_id0),
+				KFD_EC_MASK(KFD_DEBUG_CP_BAD_OP_ECODE(context_id0)),
+				NULL, 0);
+		}
 	} else if (client_id == SOC15_IH_CLIENTID_SDMA0 ||
 		   client_id == SOC15_IH_CLIENTID_SDMA1 ||
 		   client_id == SOC15_IH_CLIENTID_SDMA2 ||
@@ -345,9 +540,9 @@ static void event_interrupt_wq_v9(struct kfd_dev *dev,
 		   client_id == SOC15_IH_CLIENTID_UTCL2) {
 		struct kfd_vm_fault_info info = {0};
 		uint16_t ring_id = SOC15_RING_ID_FROM_IH_ENTRY(ih_ring_entry);
+		struct kfd_hsa_memory_exception_data exception_data;
 
-		if (client_id == SOC15_IH_CLIENTID_UTCL2 &&
-		    amdgpu_amdkfd_ras_query_utcl2_poison_status(dev->adev)) {
+		if (source_id == SOC15_INTSRC_VMC_UTCL2_POISON) {
 			event_interrupt_poison_consumption_v9(dev, pasid, client_id);
 			return;
 		}
@@ -360,13 +555,56 @@ static void event_interrupt_wq_v9(struct kfd_dev *dev,
 		info.prot_read  = ring_id & 0x10;
 		info.prot_write = ring_id & 0x20;
 
+		memset(&exception_data, 0, sizeof(exception_data));
+		exception_data.gpu_id = dev->id;
+		exception_data.va = (info.page_addr) << PAGE_SHIFT;
+		exception_data.failure.NotPresent = info.prot_valid ? 1 : 0;
+		exception_data.failure.NoExecute = info.prot_exec ? 1 : 0;
+		exception_data.failure.ReadOnly = info.prot_write ? 1 : 0;
+		exception_data.failure.imprecise = 0;
+
+		kfd_set_dbg_ev_from_interrupt(dev,
+						pasid,
+						-1,
+						KFD_EC_MASK(EC_DEVICE_MEMORY_VIOLATION),
+						&exception_data,
+						sizeof(exception_data));
 		kfd_smi_event_update_vmfault(dev, pasid);
-		kfd_dqm_evict_pasid(dev->dqm, pasid);
-		kfd_signal_vm_fault_event(dev, pasid, &info);
+	} else if (KFD_IRQ_IS_FENCE(client_id, source_id)) {
+		kfd_process_close_interrupt_drain(pasid);
 	}
 }
 
+static bool event_interrupt_isr_v9_4_3(struct kfd_node *node,
+				const uint32_t *ih_ring_entry,
+				uint32_t *patched_ihre,
+				bool *patched_flag)
+{
+	uint16_t node_id, vmid;
+
+	/*
+	 * For GFX 9.4.3, process the interrupt if:
+	 * - NodeID field in IH entry matches the corresponding bit
+	 *   set in interrupt_bitmap Bits 0-15.
+	 *   OR
+	 * - If partition mode is CPX and interrupt came from
+	 *   Node_id 0,4,8,12, then check if the Bit (16 + client id)
+	 *   is set in interrupt bitmap Bits 16-31.
+	 */
+	node_id = SOC15_NODEID_FROM_IH_ENTRY(ih_ring_entry);
+	vmid = SOC15_VMID_FROM_IH_ENTRY(ih_ring_entry);
+	if (kfd_irq_is_from_node(node, node_id, vmid))
+		return event_interrupt_isr_v9(node, ih_ring_entry,
+					patched_ihre, patched_flag);
+	return false;
+}
+
 const struct kfd_event_interrupt_class event_interrupt_class_v9 = {
 	.interrupt_isr = event_interrupt_isr_v9,
 	.interrupt_wq = event_interrupt_wq_v9,
 };
+
+const struct kfd_event_interrupt_class event_interrupt_class_v9_4_3 = {
+	.interrupt_isr = event_interrupt_isr_v9_4_3,
+	.interrupt_wq = event_interrupt_wq_v9,
+};
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c b/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c
index 34772fe74296..783c2f5a04e4 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c
@@ -46,33 +46,36 @@
 #include <linux/kfifo.h>
 #include "kfd_priv.h"
 
-#define KFD_IH_NUM_ENTRIES 8192
+#define KFD_IH_NUM_ENTRIES 16384
 
 static void interrupt_wq(struct work_struct *);
 
-int kfd_interrupt_init(struct kfd_dev *kfd)
+int kfd_interrupt_init(struct kfd_node *node)
 {
 	int r;
 
-	r = kfifo_alloc(&kfd->ih_fifo,
-		KFD_IH_NUM_ENTRIES * kfd->device_info.ih_ring_entry_size,
+	r = kfifo_alloc(&node->ih_fifo,
+		KFD_IH_NUM_ENTRIES * node->kfd->device_info.ih_ring_entry_size,
 		GFP_KERNEL);
 	if (r) {
-		dev_err(kfd->adev->dev, "Failed to allocate IH fifo\n");
+		dev_err(node->adev->dev, "Failed to allocate IH fifo\n");
 		return r;
 	}
 
-	kfd->ih_wq = alloc_workqueue("KFD IH", WQ_HIGHPRI, 1);
-	if (unlikely(!kfd->ih_wq)) {
-		kfifo_free(&kfd->ih_fifo);
-		dev_err(kfd->adev->dev, "Failed to allocate KFD IH workqueue\n");
-		return -ENOMEM;
+	if (!node->kfd->ih_wq) {
+		node->kfd->ih_wq = alloc_workqueue("KFD IH", WQ_HIGHPRI | WQ_UNBOUND,
+						   node->kfd->num_nodes);
+		if (unlikely(!node->kfd->ih_wq)) {
+			kfifo_free(&node->ih_fifo);
+			dev_err(node->adev->dev, "Failed to allocate KFD IH workqueue\n");
+			return -ENOMEM;
+		}
 	}
-	spin_lock_init(&kfd->interrupt_lock);
+	spin_lock_init(&node->interrupt_lock);
 
-	INIT_WORK(&kfd->interrupt_work, interrupt_wq);
+	INIT_WORK(&node->interrupt_work, interrupt_wq);
 
-	kfd->interrupts_active = true;
+	node->interrupts_active = true;
 
 	/*
 	 * After this function returns, the interrupt will be enabled. This
@@ -84,7 +87,7 @@ int kfd_interrupt_init(struct kfd_dev *kfd)
 	return 0;
 }
 
-void kfd_interrupt_exit(struct kfd_dev *kfd)
+void kfd_interrupt_exit(struct kfd_node *node)
 {
 	/*
 	 * Stop the interrupt handler from writing to the ring and scheduling
@@ -93,87 +96,72 @@ void kfd_interrupt_exit(struct kfd_dev *kfd)
 	 */
 	unsigned long flags;
 
-	spin_lock_irqsave(&kfd->interrupt_lock, flags);
-	kfd->interrupts_active = false;
-	spin_unlock_irqrestore(&kfd->interrupt_lock, flags);
-
-	/*
-	 * flush_work ensures that there are no outstanding
-	 * work-queue items that will access interrupt_ring. New work items
-	 * can't be created because we stopped interrupt handling above.
-	 */
-	flush_workqueue(kfd->ih_wq);
-
-	kfifo_free(&kfd->ih_fifo);
+	spin_lock_irqsave(&node->interrupt_lock, flags);
+	node->interrupts_active = false;
+	spin_unlock_irqrestore(&node->interrupt_lock, flags);
+	kfifo_free(&node->ih_fifo);
 }
 
 /*
  * Assumption: single reader/writer. This function is not re-entrant
  */
-bool enqueue_ih_ring_entry(struct kfd_dev *kfd,	const void *ih_ring_entry)
+bool enqueue_ih_ring_entry(struct kfd_node *node, const void *ih_ring_entry)
 {
-	int count;
-
-	count = kfifo_in(&kfd->ih_fifo, ih_ring_entry,
-				kfd->device_info.ih_ring_entry_size);
-	if (count != kfd->device_info.ih_ring_entry_size) {
-		dev_dbg_ratelimited(kfd->adev->dev,
-			"Interrupt ring overflow, dropping interrupt %d\n",
-			count);
+	if (kfifo_is_full(&node->ih_fifo)) {
+		dev_warn_ratelimited(node->adev->dev, "KFD node %d ih_fifo overflow\n",
+				     node->node_id);
 		return false;
 	}
 
+	kfifo_in(&node->ih_fifo, ih_ring_entry, node->kfd->device_info.ih_ring_entry_size);
 	return true;
 }
 
 /*
  * Assumption: single reader/writer. This function is not re-entrant
  */
-static bool dequeue_ih_ring_entry(struct kfd_dev *kfd, void *ih_ring_entry)
+static bool dequeue_ih_ring_entry(struct kfd_node *node, u32 **ih_ring_entry)
 {
 	int count;
 
-	count = kfifo_out(&kfd->ih_fifo, ih_ring_entry,
-				kfd->device_info.ih_ring_entry_size);
-
-	WARN_ON(count && count != kfd->device_info.ih_ring_entry_size);
+	if (kfifo_is_empty(&node->ih_fifo))
+		return false;
 
-	return count == kfd->device_info.ih_ring_entry_size;
+	count = kfifo_out_linear_ptr(&node->ih_fifo, ih_ring_entry,
+				     node->kfd->device_info.ih_ring_entry_size);
+	WARN_ON(count != node->kfd->device_info.ih_ring_entry_size);
+	return count == node->kfd->device_info.ih_ring_entry_size;
 }
 
 static void interrupt_wq(struct work_struct *work)
 {
-	struct kfd_dev *dev = container_of(work, struct kfd_dev,
-						interrupt_work);
-	uint32_t ih_ring_entry[KFD_MAX_RING_ENTRY_SIZE];
+	struct kfd_node *dev = container_of(work, struct kfd_node, interrupt_work);
+	uint32_t *ih_ring_entry;
 	unsigned long start_jiffies = jiffies;
 
-	if (dev->device_info.ih_ring_entry_size > sizeof(ih_ring_entry)) {
-		dev_err_once(dev->adev->dev, "Ring entry too small\n");
-		return;
-	}
-
-	while (dequeue_ih_ring_entry(dev, ih_ring_entry)) {
-		dev->device_info.event_interrupt_class->interrupt_wq(dev,
+	while (dequeue_ih_ring_entry(dev, &ih_ring_entry)) {
+		dev->kfd->device_info.event_interrupt_class->interrupt_wq(dev,
 								ih_ring_entry);
+		kfifo_skip_count(&dev->ih_fifo, dev->kfd->device_info.ih_ring_entry_size);
+
 		if (time_is_before_jiffies(start_jiffies + HZ)) {
 			/* If we spent more than a second processing signals,
 			 * reschedule the worker to avoid soft-lockup warnings
 			 */
-			queue_work(dev->ih_wq, &dev->interrupt_work);
+			queue_work(dev->kfd->ih_wq, &dev->interrupt_work);
 			break;
 		}
 	}
 }
 
-bool interrupt_is_wanted(struct kfd_dev *dev,
+bool interrupt_is_wanted(struct kfd_node *dev,
 			const uint32_t *ih_ring_entry,
 			uint32_t *patched_ihre, bool *flag)
 {
 	/* integer and bitwise OR so there is no boolean short-circuiting */
 	unsigned int wanted = 0;
 
-	wanted |= dev->device_info.event_interrupt_class->interrupt_isr(dev,
+	wanted |= dev->kfd->device_info.event_interrupt_class->interrupt_isr(dev,
 					 ih_ring_entry, patched_ihre, flag);
 
 	return wanted != 0;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_iommu.c b/drivers/gpu/drm/amd/amdkfd/kfd_iommu.c
deleted file mode 100644
index ec1bf611624e..000000000000
--- a/drivers/gpu/drm/amd/amdkfd/kfd_iommu.c
+++ /dev/null
@@ -1,349 +0,0 @@
-// SPDX-License-Identifier: GPL-2.0 OR MIT
-/*
- * Copyright 2018-2022 Advanced Micro Devices, Inc.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
- * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
- * OTHER DEALINGS IN THE SOFTWARE.
- */
-
-#include <linux/kconfig.h>
-
-#if IS_REACHABLE(CONFIG_AMD_IOMMU_V2)
-
-#include <linux/printk.h>
-#include <linux/device.h>
-#include <linux/slab.h>
-#include <linux/pci.h>
-#include <linux/amd-iommu.h>
-#include "kfd_priv.h"
-#include "kfd_topology.h"
-#include "kfd_iommu.h"
-
-static const u32 required_iommu_flags = AMD_IOMMU_DEVICE_FLAG_ATS_SUP |
-					AMD_IOMMU_DEVICE_FLAG_PRI_SUP |
-					AMD_IOMMU_DEVICE_FLAG_PASID_SUP;
-
-/** kfd_iommu_check_device - Check whether IOMMU is available for device
- */
-int kfd_iommu_check_device(struct kfd_dev *kfd)
-{
-	struct amd_iommu_device_info iommu_info;
-	int err;
-
-	if (!kfd->use_iommu_v2)
-		return -ENODEV;
-
-	iommu_info.flags = 0;
-	err = amd_iommu_device_info(kfd->adev->pdev, &iommu_info);
-	if (err)
-		return err;
-
-	if ((iommu_info.flags & required_iommu_flags) != required_iommu_flags)
-		return -ENODEV;
-
-	return 0;
-}
-
-/** kfd_iommu_device_init - Initialize IOMMU for device
- */
-int kfd_iommu_device_init(struct kfd_dev *kfd)
-{
-	struct amd_iommu_device_info iommu_info;
-	unsigned int pasid_limit;
-	int err;
-
-	if (!kfd->use_iommu_v2)
-		return 0;
-
-	iommu_info.flags = 0;
-	err = amd_iommu_device_info(kfd->adev->pdev, &iommu_info);
-	if (err < 0) {
-		dev_err(kfd_device,
-			"error getting iommu info. is the iommu enabled?\n");
-		return -ENODEV;
-	}
-
-	if ((iommu_info.flags & required_iommu_flags) != required_iommu_flags) {
-		dev_err(kfd_device,
-			"error required iommu flags ats %i, pri %i, pasid %i\n",
-		       (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_ATS_SUP) != 0,
-		       (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_PRI_SUP) != 0,
-		       (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_PASID_SUP)
-									!= 0);
-		return -ENODEV;
-	}
-
-	pasid_limit = min_t(unsigned int,
-			(unsigned int)(1 << kfd->device_info.max_pasid_bits),
-			iommu_info.max_pasids);
-
-	if (!kfd_set_pasid_limit(pasid_limit)) {
-		dev_err(kfd_device, "error setting pasid limit\n");
-		return -EBUSY;
-	}
-
-	return 0;
-}
-
-/** kfd_iommu_bind_process_to_device - Have the IOMMU bind a process
- *
- * Binds the given process to the given device using its PASID. This
- * enables IOMMUv2 address translation for the process on the device.
- *
- * This function assumes that the process mutex is held.
- */
-int kfd_iommu_bind_process_to_device(struct kfd_process_device *pdd)
-{
-	struct kfd_dev *dev = pdd->dev;
-	struct kfd_process *p = pdd->process;
-	int err;
-
-	if (!dev->use_iommu_v2 || pdd->bound == PDD_BOUND)
-		return 0;
-
-	if (unlikely(pdd->bound == PDD_BOUND_SUSPENDED)) {
-		pr_err("Binding PDD_BOUND_SUSPENDED pdd is unexpected!\n");
-		return -EINVAL;
-	}
-
-	err = amd_iommu_bind_pasid(dev->adev->pdev, p->pasid, p->lead_thread);
-	if (!err)
-		pdd->bound = PDD_BOUND;
-
-	return err;
-}
-
-/** kfd_iommu_unbind_process - Unbind process from all devices
- *
- * This removes all IOMMU device bindings of the process. To be used
- * before process termination.
- */
-void kfd_iommu_unbind_process(struct kfd_process *p)
-{
-	int i;
-
-	for (i = 0; i < p->n_pdds; i++)
-		if (p->pdds[i]->bound == PDD_BOUND)
-			amd_iommu_unbind_pasid(p->pdds[i]->dev->adev->pdev,
-					       p->pasid);
-}
-
-/* Callback for process shutdown invoked by the IOMMU driver */
-static void iommu_pasid_shutdown_callback(struct pci_dev *pdev, u32 pasid)
-{
-	struct kfd_dev *dev = kfd_device_by_pci_dev(pdev);
-	struct kfd_process *p;
-	struct kfd_process_device *pdd;
-
-	if (!dev)
-		return;
-
-	/*
-	 * Look for the process that matches the pasid. If there is no such
-	 * process, we either released it in amdkfd's own notifier, or there
-	 * is a bug. Unfortunately, there is no way to tell...
-	 */
-	p = kfd_lookup_process_by_pasid(pasid);
-	if (!p)
-		return;
-
-	pr_debug("Unbinding process 0x%x from IOMMU\n", pasid);
-
-	mutex_lock(&p->mutex);
-
-	pdd = kfd_get_process_device_data(dev, p);
-	if (pdd)
-		/* For GPU relying on IOMMU, we need to dequeue here
-		 * when PASID is still bound.
-		 */
-		kfd_process_dequeue_from_device(pdd);
-
-	mutex_unlock(&p->mutex);
-
-	kfd_unref_process(p);
-}
-
-/* This function called by IOMMU driver on PPR failure */
-static int iommu_invalid_ppr_cb(struct pci_dev *pdev, u32 pasid,
-				unsigned long address, u16 flags)
-{
-	struct kfd_dev *dev;
-
-	dev_warn_ratelimited(kfd_device,
-			"Invalid PPR device %x:%x.%x pasid 0x%x address 0x%lX flags 0x%X",
-			pdev->bus->number,
-			PCI_SLOT(pdev->devfn),
-			PCI_FUNC(pdev->devfn),
-			pasid,
-			address,
-			flags);
-
-	dev = kfd_device_by_pci_dev(pdev);
-	if (!WARN_ON(!dev))
-		kfd_signal_iommu_event(dev, pasid, address,
-			flags & PPR_FAULT_WRITE, flags & PPR_FAULT_EXEC);
-
-	return AMD_IOMMU_INV_PRI_RSP_INVALID;
-}
-
-/*
- * Bind processes do the device that have been temporarily unbound
- * (PDD_BOUND_SUSPENDED) in kfd_unbind_processes_from_device.
- */
-static int kfd_bind_processes_to_device(struct kfd_dev *kfd)
-{
-	struct kfd_process_device *pdd;
-	struct kfd_process *p;
-	unsigned int temp;
-	int err = 0;
-
-	int idx = srcu_read_lock(&kfd_processes_srcu);
-
-	hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) {
-		mutex_lock(&p->mutex);
-		pdd = kfd_get_process_device_data(kfd, p);
-
-		if (WARN_ON(!pdd) || pdd->bound != PDD_BOUND_SUSPENDED) {
-			mutex_unlock(&p->mutex);
-			continue;
-		}
-
-		err = amd_iommu_bind_pasid(kfd->adev->pdev, p->pasid,
-				p->lead_thread);
-		if (err < 0) {
-			pr_err("Unexpected pasid 0x%x binding failure\n",
-					p->pasid);
-			mutex_unlock(&p->mutex);
-			break;
-		}
-
-		pdd->bound = PDD_BOUND;
-		mutex_unlock(&p->mutex);
-	}
-
-	srcu_read_unlock(&kfd_processes_srcu, idx);
-
-	return err;
-}
-
-/*
- * Mark currently bound processes as PDD_BOUND_SUSPENDED. These
- * processes will be restored to PDD_BOUND state in
- * kfd_bind_processes_to_device.
- */
-static void kfd_unbind_processes_from_device(struct kfd_dev *kfd)
-{
-	struct kfd_process_device *pdd;
-	struct kfd_process *p;
-	unsigned int temp;
-
-	int idx = srcu_read_lock(&kfd_processes_srcu);
-
-	hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) {
-		mutex_lock(&p->mutex);
-		pdd = kfd_get_process_device_data(kfd, p);
-
-		if (WARN_ON(!pdd)) {
-			mutex_unlock(&p->mutex);
-			continue;
-		}
-
-		if (pdd->bound == PDD_BOUND)
-			pdd->bound = PDD_BOUND_SUSPENDED;
-		mutex_unlock(&p->mutex);
-	}
-
-	srcu_read_unlock(&kfd_processes_srcu, idx);
-}
-
-/** kfd_iommu_suspend - Prepare IOMMU for suspend
- *
- * This unbinds processes from the device and disables the IOMMU for
- * the device.
- */
-void kfd_iommu_suspend(struct kfd_dev *kfd)
-{
-	if (!kfd->use_iommu_v2)
-		return;
-
-	kfd_unbind_processes_from_device(kfd);
-
-	amd_iommu_set_invalidate_ctx_cb(kfd->adev->pdev, NULL);
-	amd_iommu_set_invalid_ppr_cb(kfd->adev->pdev, NULL);
-	amd_iommu_free_device(kfd->adev->pdev);
-}
-
-/** kfd_iommu_resume - Restore IOMMU after resume
- *
- * This reinitializes the IOMMU for the device and re-binds previously
- * suspended processes to the device.
- */
-int kfd_iommu_resume(struct kfd_dev *kfd)
-{
-	unsigned int pasid_limit;
-	int err;
-
-	if (!kfd->use_iommu_v2)
-		return 0;
-
-	pasid_limit = kfd_get_pasid_limit();
-
-	err = amd_iommu_init_device(kfd->adev->pdev, pasid_limit);
-	if (err)
-		return -ENXIO;
-
-	amd_iommu_set_invalidate_ctx_cb(kfd->adev->pdev,
-					iommu_pasid_shutdown_callback);
-	amd_iommu_set_invalid_ppr_cb(kfd->adev->pdev,
-				     iommu_invalid_ppr_cb);
-
-	err = kfd_bind_processes_to_device(kfd);
-	if (err) {
-		amd_iommu_set_invalidate_ctx_cb(kfd->adev->pdev, NULL);
-		amd_iommu_set_invalid_ppr_cb(kfd->adev->pdev, NULL);
-		amd_iommu_free_device(kfd->adev->pdev);
-		return err;
-	}
-
-	return 0;
-}
-
-/** kfd_iommu_add_perf_counters - Add IOMMU performance counters to topology
- */
-int kfd_iommu_add_perf_counters(struct kfd_topology_device *kdev)
-{
-	struct kfd_perf_properties *props;
-
-	if (!(kdev->node_props.capability & HSA_CAP_ATS_PRESENT))
-		return 0;
-
-	if (!amd_iommu_pc_supported())
-		return 0;
-
-	props = kfd_alloc_struct(props);
-	if (!props)
-		return -ENOMEM;
-	strcpy(props->block_name, "iommu");
-	props->max_concurrent = amd_iommu_pc_get_max_banks(0) *
-		amd_iommu_pc_get_max_counters(0); /* assume one iommu */
-	list_add_tail(&props->list, &kdev->perf_props);
-
-	return 0;
-}
-
-#endif
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_iommu.h b/drivers/gpu/drm/amd/amdkfd/kfd_iommu.h
deleted file mode 100644
index 8cf0fcbe87c2..000000000000
--- a/drivers/gpu/drm/amd/amdkfd/kfd_iommu.h
+++ /dev/null
@@ -1,84 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0 OR MIT */
-/*
- * Copyright 2018-2022 Advanced Micro Devices, Inc.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
- * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
- * OTHER DEALINGS IN THE SOFTWARE.
- */
-
-#ifndef __KFD_IOMMU_H__
-#define __KFD_IOMMU_H__
-
-#include <linux/kconfig.h>
-
-#if IS_REACHABLE(CONFIG_AMD_IOMMU_V2)
-
-#define KFD_SUPPORT_IOMMU_V2
-
-int kfd_iommu_check_device(struct kfd_dev *kfd);
-int kfd_iommu_device_init(struct kfd_dev *kfd);
-
-int kfd_iommu_bind_process_to_device(struct kfd_process_device *pdd);
-void kfd_iommu_unbind_process(struct kfd_process *p);
-
-void kfd_iommu_suspend(struct kfd_dev *kfd);
-int kfd_iommu_resume(struct kfd_dev *kfd);
-
-int kfd_iommu_add_perf_counters(struct kfd_topology_device *kdev);
-
-#else
-
-static inline int kfd_iommu_check_device(struct kfd_dev *kfd)
-{
-	return -ENODEV;
-}
-static inline int kfd_iommu_device_init(struct kfd_dev *kfd)
-{
-#if IS_MODULE(CONFIG_AMD_IOMMU_V2)
-	WARN_ONCE(1, "iommu_v2 module is not usable by built-in KFD");
-#endif
-	return 0;
-}
-
-static inline int kfd_iommu_bind_process_to_device(
-	struct kfd_process_device *pdd)
-{
-	return 0;
-}
-static inline void kfd_iommu_unbind_process(struct kfd_process *p)
-{
-	/* empty */
-}
-
-static inline void kfd_iommu_suspend(struct kfd_dev *kfd)
-{
-	/* empty */
-}
-static inline int kfd_iommu_resume(struct kfd_dev *kfd)
-{
-	return 0;
-}
-
-static inline int kfd_iommu_add_perf_counters(struct kfd_topology_device *kdev)
-{
-	return 0;
-}
-
-#endif /* IS_REACHABLE(CONFIG_AMD_IOMMU_V2) */
-
-#endif /* __KFD_IOMMU_H__ */
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c
index bcf7bc3302c9..fb3129883a4c 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c
@@ -32,24 +32,21 @@
 #include "kfd_device_queue_manager.h"
 #include "kfd_pm4_headers.h"
 #include "kfd_pm4_opcodes.h"
+#include "amdgpu_reset.h"
 
 #define PM4_COUNT_ZERO (((1 << 15) - 1) << 16)
 
 /* Initialize a kernel queue, including allocations of GART memory
  * needed for the queue.
  */
-static bool kq_initialize(struct kernel_queue *kq, struct kfd_dev *dev,
+static bool kq_initialize(struct kernel_queue *kq, struct kfd_node *dev,
 		enum kfd_queue_type type, unsigned int queue_size)
 {
 	struct queue_properties prop;
 	int retval;
 	union PM4_MES_TYPE_3_HEADER nop;
 
-	if (WARN_ON(type != KFD_QUEUE_TYPE_DIQ && type != KFD_QUEUE_TYPE_HIQ))
-		return false;
-
-	pr_debug("Initializing queue type %d size %d\n", KFD_QUEUE_TYPE_HIQ,
-			queue_size);
+	pr_debug("Initializing queue type %d size %d\n", type, queue_size);
 
 	memset(&prop, 0, sizeof(prop));
 	memset(&nop, 0, sizeof(nop));
@@ -68,23 +65,25 @@ static bool kq_initialize(struct kernel_queue *kq, struct kfd_dev *dev,
 		kq->mqd_mgr = dev->dqm->mqd_mgrs[KFD_MQD_TYPE_HIQ];
 		break;
 	default:
-		pr_err("Invalid queue type %d\n", type);
+		WARN(1, "Invalid queue type %d\n", type);
+		dev_err(dev->adev->dev, "Invalid queue type %d\n", type);
 		return false;
 	}
 
 	if (!kq->mqd_mgr)
 		return false;
 
-	prop.doorbell_ptr = kfd_get_kernel_doorbell(dev, &prop.doorbell_off);
+	prop.doorbell_ptr = kfd_get_kernel_doorbell(dev->kfd, &prop.doorbell_off);
 
 	if (!prop.doorbell_ptr) {
-		pr_err("Failed to initialize doorbell");
+		dev_err(dev->adev->dev, "Failed to initialize doorbell");
 		goto err_get_kernel_doorbell;
 	}
 
 	retval = kfd_gtt_sa_allocate(dev, queue_size, &kq->pq);
 	if (retval != 0) {
-		pr_err("Failed to init pq queues size %d\n", queue_size);
+		dev_err(dev->adev->dev, "Failed to init pq queues size %d\n",
+			queue_size);
 		goto err_pq_allocate_vidmem;
 	}
 
@@ -112,7 +111,7 @@ static bool kq_initialize(struct kernel_queue *kq, struct kfd_dev *dev,
 	kq->rptr_kernel = kq->rptr_mem->cpu_ptr;
 	kq->rptr_gpu_addr = kq->rptr_mem->gpu_addr;
 
-	retval = kfd_gtt_sa_allocate(dev, dev->device_info.doorbell_size,
+	retval = kfd_gtt_sa_allocate(dev, dev->kfd->device_info.doorbell_size,
 					&kq->wptr_mem);
 
 	if (retval != 0)
@@ -123,7 +122,7 @@ static bool kq_initialize(struct kernel_queue *kq, struct kfd_dev *dev,
 
 	memset(kq->pq_kernel_addr, 0, queue_size);
 	memset(kq->rptr_kernel, 0, sizeof(*kq->rptr_kernel));
-	memset(kq->wptr_kernel, 0, sizeof(*kq->wptr_kernel));
+	memset(kq->wptr_kernel, 0, dev->kfd->device_info.doorbell_size);
 
 	prop.queue_size = queue_size;
 	prop.is_interop = false;
@@ -189,22 +188,24 @@ err_rptr_allocate_vidmem:
 err_eop_allocate_vidmem:
 	kfd_gtt_sa_free(dev, kq->pq);
 err_pq_allocate_vidmem:
-	kfd_release_kernel_doorbell(dev, prop.doorbell_ptr);
+	kfd_release_kernel_doorbell(dev->kfd, prop.doorbell_ptr);
 err_get_kernel_doorbell:
 	return false;
 
 }
 
 /* Uninitialize a kernel queue and free all its memory usages. */
-static void kq_uninitialize(struct kernel_queue *kq, bool hanging)
+static void kq_uninitialize(struct kernel_queue *kq)
 {
-	if (kq->queue->properties.type == KFD_QUEUE_TYPE_HIQ && !hanging)
+	if (kq->queue->properties.type == KFD_QUEUE_TYPE_HIQ && down_read_trylock(&kq->dev->adev->reset_domain->sem)) {
 		kq->mqd_mgr->destroy_mqd(kq->mqd_mgr,
 					kq->queue->mqd,
 					KFD_PREEMPT_TYPE_WAVEFRONT_RESET,
 					KFD_UNMAP_LATENCY_MS,
 					kq->queue->pipe,
 					kq->queue->queue);
+		up_read(&kq->dev->adev->reset_domain->sem);
+	}
 	else if (kq->queue->properties.type == KFD_QUEUE_TYPE_DIQ)
 		kfd_gtt_sa_free(kq->dev, kq->fence_mem_obj);
 
@@ -220,7 +221,7 @@ static void kq_uninitialize(struct kernel_queue *kq, bool hanging)
 	kfd_gtt_sa_free(kq->dev, kq->eop_mem);
 
 	kfd_gtt_sa_free(kq->dev, kq->pq);
-	kfd_release_kernel_doorbell(kq->dev,
+	kfd_release_kernel_doorbell(kq->dev->kfd,
 					kq->queue->properties.doorbell_ptr);
 	uninit_queue(kq->queue);
 }
@@ -286,7 +287,7 @@ err_no_space:
 	return -ENOMEM;
 }
 
-void kq_submit_packet(struct kernel_queue *kq)
+int kq_submit_packet(struct kernel_queue *kq)
 {
 #ifdef DEBUG
 	int i;
@@ -298,20 +299,31 @@ void kq_submit_packet(struct kernel_queue *kq)
 	}
 	pr_debug("\n");
 #endif
-	if (kq->dev->device_info.doorbell_size == 8) {
+	/* Fatal err detected, packet submission won't go through */
+	if (amdgpu_amdkfd_is_fed(kq->dev->adev))
+		return -EIO;
+
+	/* Make sure ring buffer is updated before wptr updated */
+	mb();
+
+	if (kq->dev->kfd->device_info.doorbell_size == 8) {
 		*kq->wptr64_kernel = kq->pending_wptr64;
+		mb(); /* Make sure wptr updated before ring doorbell */
 		write_kernel_doorbell64(kq->queue->properties.doorbell_ptr,
 					kq->pending_wptr64);
 	} else {
 		*kq->wptr_kernel = kq->pending_wptr;
+		mb(); /* Make sure wptr updated before ring doorbell */
 		write_kernel_doorbell(kq->queue->properties.doorbell_ptr,
 					kq->pending_wptr);
 	}
+
+	return 0;
 }
 
 void kq_rollback_packet(struct kernel_queue *kq)
 {
-	if (kq->dev->device_info.doorbell_size == 8) {
+	if (kq->dev->kfd->device_info.doorbell_size == 8) {
 		kq->pending_wptr64 = *kq->wptr64_kernel;
 		kq->pending_wptr = *kq->wptr_kernel %
 			(kq->queue->properties.queue_size / 4);
@@ -320,7 +332,7 @@ void kq_rollback_packet(struct kernel_queue *kq)
 	}
 }
 
-struct kernel_queue *kernel_queue_init(struct kfd_dev *dev,
+struct kernel_queue *kernel_queue_init(struct kfd_node *dev,
 					enum kfd_queue_type type)
 {
 	struct kernel_queue *kq;
@@ -332,45 +344,45 @@ struct kernel_queue *kernel_queue_init(struct kfd_dev *dev,
 	if (kq_initialize(kq, dev, type, KFD_KERNEL_QUEUE_SIZE))
 		return kq;
 
-	pr_err("Failed to init kernel queue\n");
+	dev_err(dev->adev->dev, "Failed to init kernel queue\n");
 
 	kfree(kq);
 	return NULL;
 }
 
-void kernel_queue_uninit(struct kernel_queue *kq, bool hanging)
+void kernel_queue_uninit(struct kernel_queue *kq)
 {
-	kq_uninitialize(kq, hanging);
+	kq_uninitialize(kq);
 	kfree(kq);
 }
 
 /* FIXME: Can this test be removed? */
-static __attribute__((unused)) void test_kq(struct kfd_dev *dev)
+static __attribute__((unused)) void test_kq(struct kfd_node *dev)
 {
 	struct kernel_queue *kq;
 	uint32_t *buffer, i;
 	int retval;
 
-	pr_err("Starting kernel queue test\n");
+	dev_err(dev->adev->dev, "Starting kernel queue test\n");
 
 	kq = kernel_queue_init(dev, KFD_QUEUE_TYPE_HIQ);
 	if (unlikely(!kq)) {
-		pr_err("  Failed to initialize HIQ\n");
-		pr_err("Kernel queue test failed\n");
+		dev_err(dev->adev->dev, "  Failed to initialize HIQ\n");
+		dev_err(dev->adev->dev, "Kernel queue test failed\n");
 		return;
 	}
 
 	retval = kq_acquire_packet_buffer(kq, 5, &buffer);
 	if (unlikely(retval != 0)) {
-		pr_err("  Failed to acquire packet buffer\n");
-		pr_err("Kernel queue test failed\n");
+		dev_err(dev->adev->dev, "  Failed to acquire packet buffer\n");
+		dev_err(dev->adev->dev, "Kernel queue test failed\n");
 		return;
 	}
 	for (i = 0; i < 5; i++)
 		buffer[i] = kq->nop_packet;
 	kq_submit_packet(kq);
 
-	pr_err("Ending kernel queue test\n");
+	dev_err(dev->adev->dev, "Ending kernel queue test\n");
 }
 
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h
index 383202fd1ea2..e24ee50acdf0 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h
@@ -47,13 +47,13 @@
 int kq_acquire_packet_buffer(struct kernel_queue *kq,
 				size_t packet_size_in_dwords,
 				unsigned int **buffer_ptr);
-void kq_submit_packet(struct kernel_queue *kq);
+int kq_submit_packet(struct kernel_queue *kq);
 void kq_rollback_packet(struct kernel_queue *kq);
 
 
 struct kernel_queue {
 	/* data */
-	struct kfd_dev		*dev;
+	struct kfd_node		*dev;
 	struct mqd_manager	*mqd_mgr;
 	struct queue		*queue;
 	uint64_t		pending_wptr64;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
index 54933903bcb8..af53e796ea1b 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
@@ -21,7 +21,6 @@
  * OTHER DEALINGS IN THE SOFTWARE.
  */
 #include <linux/types.h>
-#include <linux/hmm.h>
 #include <linux/dma-direction.h>
 #include <linux/dma-mapping.h>
 #include <linux/migrate.h>
@@ -39,22 +38,22 @@
 #endif
 #define dev_fmt(fmt) "kfd_migrate: " fmt
 
-static uint64_t
-svm_migrate_direct_mapping_addr(struct amdgpu_device *adev, uint64_t addr)
+static u64
+svm_migrate_direct_mapping_addr(struct amdgpu_device *adev, u64 addr)
 {
 	return addr + amdgpu_ttm_domain_start(adev, TTM_PL_VRAM);
 }
 
 static int
-svm_migrate_gart_map(struct amdgpu_ring *ring, uint64_t npages,
-		     dma_addr_t *addr, uint64_t *gart_addr, uint64_t flags)
+svm_migrate_gart_map(struct amdgpu_ring *ring, u64 npages,
+		     dma_addr_t *addr, u64 *gart_addr, u64 flags)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct amdgpu_job *job;
 	unsigned int num_dw, num_bytes;
 	struct dma_fence *fence;
-	uint64_t src_addr, dst_addr;
-	uint64_t pte_flags;
+	u64 src_addr, dst_addr;
+	u64 pte_flags;
 	void *cpu_addr;
 	int r;
 
@@ -64,11 +63,12 @@ svm_migrate_gart_map(struct amdgpu_ring *ring, uint64_t npages,
 	num_dw = ALIGN(adev->mman.buffer_funcs->copy_num_dw, 8);
 	num_bytes = npages * 8;
 
-	r = amdgpu_job_alloc_with_ib(adev, &adev->mman.entity,
+	r = amdgpu_job_alloc_with_ib(adev, &adev->mman.high_pr,
 				     AMDGPU_FENCE_OWNER_UNDEFINED,
 				     num_dw * 4 + num_bytes,
 				     AMDGPU_IB_POOL_DELAYED,
-				     &job);
+				     &job,
+				     AMDGPU_KERNEL_JOB_ID_KFD_GART_MAP);
 	if (r)
 		return r;
 
@@ -77,7 +77,7 @@ svm_migrate_gart_map(struct amdgpu_ring *ring, uint64_t npages,
 
 	dst_addr = amdgpu_bo_gpu_offset(adev->gart.bo);
 	amdgpu_emit_copy_buffer(adev, &job->ibs[0], src_addr,
-				dst_addr, num_bytes, false);
+				dst_addr, num_bytes, 0);
 
 	amdgpu_ring_pad_ib(ring, &job->ibs[0]);
 	WARN_ON(job->ibs[0].length_dw > num_dw);
@@ -122,15 +122,15 @@ svm_migrate_gart_map(struct amdgpu_ring *ring, uint64_t npages,
 
 static int
 svm_migrate_copy_memory_gart(struct amdgpu_device *adev, dma_addr_t *sys,
-			     uint64_t *vram, uint64_t npages,
+			     u64 *vram, u64 npages,
 			     enum MIGRATION_COPY_DIR direction,
 			     struct dma_fence **mfence)
 {
-	const uint64_t GTT_MAX_PAGES = AMDGPU_GTT_MAX_TRANSFER_SIZE;
+	const u64 GTT_MAX_PAGES = AMDGPU_GTT_MAX_TRANSFER_SIZE;
 	struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
-	uint64_t gart_s, gart_d;
+	u64 gart_s, gart_d;
 	struct dma_fence *next;
-	uint64_t size;
+	u64 size;
 	int r;
 
 	mutex_lock(&adev->mman.gtt_window_lock);
@@ -153,7 +153,7 @@ svm_migrate_copy_memory_gart(struct amdgpu_device *adev, dma_addr_t *sys,
 		}
 
 		r = amdgpu_copy_buffer(ring, gart_s, gart_d, size * PAGE_SIZE,
-				       NULL, &next, false, true, false);
+				       NULL, &next, false, true, 0);
 		if (r) {
 			dev_err(adev->dev, "fail %d to copy memory\n", r);
 			goto out_unlock;
@@ -206,7 +206,7 @@ svm_migrate_copy_done(struct amdgpu_device *adev, struct dma_fence *mfence)
 unsigned long
 svm_migrate_addr_to_pfn(struct amdgpu_device *adev, unsigned long addr)
 {
-	return (addr + adev->kfd.dev->pgmap.range.start) >> PAGE_SHIFT;
+	return (addr + adev->kfd.pgmap.range.start) >> PAGE_SHIFT;
 }
 
 static void
@@ -217,7 +217,7 @@ svm_migrate_get_vram_page(struct svm_range *prange, unsigned long pfn)
 	page = pfn_to_page(pfn);
 	svm_range_bo_ref(prange->svm_bo);
 	page->zone_device_data = prange->svm_bo;
-	zone_device_page_init(page);
+	zone_device_page_init(page, 0);
 }
 
 static void
@@ -236,7 +236,7 @@ svm_migrate_addr(struct amdgpu_device *adev, struct page *page)
 	unsigned long addr;
 
 	addr = page_to_pfn(page) << PAGE_SHIFT;
-	return (addr - adev->kfd.dev->pgmap.range.start);
+	return (addr - adev->kfd.pgmap.range.start);
 }
 
 static struct page *
@@ -262,66 +262,57 @@ static void svm_migrate_put_sys_page(unsigned long addr)
 
 static unsigned long svm_migrate_successful_pages(struct migrate_vma *migrate)
 {
-	unsigned long cpages = 0;
+	unsigned long mpages = 0;
 	unsigned long i;
 
 	for (i = 0; i < migrate->npages; i++) {
-		if (migrate->src[i] & MIGRATE_PFN_VALID &&
+		if (migrate->dst[i] & MIGRATE_PFN_VALID &&
 		    migrate->src[i] & MIGRATE_PFN_MIGRATE)
-			cpages++;
+			mpages++;
 	}
-	return cpages;
-}
-
-static unsigned long svm_migrate_unsuccessful_pages(struct migrate_vma *migrate)
-{
-	unsigned long upages = 0;
-	unsigned long i;
-
-	for (i = 0; i < migrate->npages; i++) {
-		if (migrate->src[i] & MIGRATE_PFN_VALID &&
-		    !(migrate->src[i] & MIGRATE_PFN_MIGRATE))
-			upages++;
-	}
-	return upages;
+	return mpages;
 }
 
 static int
-svm_migrate_copy_to_vram(struct amdgpu_device *adev, struct svm_range *prange,
+svm_migrate_copy_to_vram(struct kfd_node *node, struct svm_range *prange,
 			 struct migrate_vma *migrate, struct dma_fence **mfence,
-			 dma_addr_t *scratch, uint64_t ttm_res_offset)
+			 dma_addr_t *scratch, u64 ttm_res_offset)
 {
-	uint64_t npages = migrate->npages;
+	u64 npages = migrate->npages;
+	struct amdgpu_device *adev = node->adev;
 	struct device *dev = adev->dev;
 	struct amdgpu_res_cursor cursor;
+	u64 mpages = 0;
 	dma_addr_t *src;
-	uint64_t *dst;
-	uint64_t i, j;
+	u64 *dst;
+	u64 i, j;
 	int r;
 
 	pr_debug("svms 0x%p [0x%lx 0x%lx 0x%llx]\n", prange->svms, prange->start,
 		 prange->last, ttm_res_offset);
 
 	src = scratch;
-	dst = (uint64_t *)(scratch + npages);
+	dst = (u64 *)(scratch + npages);
 
 	amdgpu_res_first(prange->ttm_res, ttm_res_offset,
 			 npages << PAGE_SHIFT, &cursor);
-	for (i = j = 0; i < npages; i++) {
+	for (i = j = 0; (i < npages) && (mpages < migrate->cpages); i++) {
 		struct page *spage;
 
-		dst[i] = cursor.start + (j << PAGE_SHIFT);
-		migrate->dst[i] = svm_migrate_addr_to_pfn(adev, dst[i]);
-		svm_migrate_get_vram_page(prange, migrate->dst[i]);
-		migrate->dst[i] = migrate_pfn(migrate->dst[i]);
-
+		if (migrate->src[i] & MIGRATE_PFN_MIGRATE) {
+			dst[i] = cursor.start + (j << PAGE_SHIFT);
+			migrate->dst[i] = svm_migrate_addr_to_pfn(adev, dst[i]);
+			svm_migrate_get_vram_page(prange, migrate->dst[i]);
+			migrate->dst[i] = migrate_pfn(migrate->dst[i]);
+			mpages++;
+		}
 		spage = migrate_pfn_to_page(migrate->src[i]);
 		if (spage && !is_zone_device_page(spage)) {
 			src[i] = dma_map_page(dev, spage, 0, PAGE_SIZE,
-					      DMA_TO_DEVICE);
+					      DMA_BIDIRECTIONAL);
 			r = dma_mapping_error(dev, src[i]);
 			if (r) {
-				dev_err(adev->dev, "%s: fail %d dma_map_page\n",
+				dev_err(dev, "%s: fail %d dma_map_page\n",
 					__func__, r);
 				goto out_free_vram_pages;
 			}
@@ -365,9 +356,12 @@ svm_migrate_copy_to_vram(struct amdgpu_device *adev, struct svm_range *prange,
 out_free_vram_pages:
 	if (r) {
 		pr_debug("failed %d to copy memory to vram\n", r);
-		while (i--) {
+		for (i = 0; i < npages && mpages; i++) {
+			if (!dst[i])
+				continue;
 			svm_migrate_put_vram_page(adev, dst[i]);
 			migrate->dst[i] = 0;
+			mpages--;
 		}
 	}
 
@@ -390,16 +384,18 @@ out_free_vram_pages:
 }
 
 static long
-svm_migrate_vma_to_vram(struct amdgpu_device *adev, struct svm_range *prange,
-			struct vm_area_struct *vma, uint64_t start,
-			uint64_t end, uint32_t trigger, uint64_t ttm_res_offset)
+svm_migrate_vma_to_vram(struct kfd_node *node, struct svm_range *prange,
+			struct vm_area_struct *vma, u64 start,
+			u64 end, uint32_t trigger, u64 ttm_res_offset)
 {
 	struct kfd_process *p = container_of(prange->svms, struct kfd_process, svms);
-	uint64_t npages = (end - start) >> PAGE_SHIFT;
+	u64 npages = (end - start) >> PAGE_SHIFT;
+	struct amdgpu_device *adev = node->adev;
 	struct kfd_process_device *pdd;
 	struct dma_fence *mfence = NULL;
 	struct migrate_vma migrate = { 0 };
 	unsigned long cpages = 0;
+	unsigned long mpages = 0;
 	dma_addr_t *scratch;
 	void *buf;
 	int r = -ENOMEM;
@@ -412,7 +408,7 @@ svm_migrate_vma_to_vram(struct amdgpu_device *adev, struct svm_range *prange,
 	migrate.pgmap_owner = SVM_ADEV_PGMAP_OWNER(adev);
 
 	buf = kvcalloc(npages,
-		       2 * sizeof(*migrate.src) + sizeof(uint64_t) + sizeof(dma_addr_t),
+		       2 * sizeof(*migrate.src) + sizeof(u64) + sizeof(dma_addr_t),
 		       GFP_KERNEL);
 	if (!buf)
 		goto out;
@@ -421,9 +417,9 @@ svm_migrate_vma_to_vram(struct amdgpu_device *adev, struct svm_range *prange,
 	migrate.dst = migrate.src + npages;
 	scratch = (dma_addr_t *)(migrate.dst + npages);
 
-	kfd_smi_event_migration_start(adev->kfd.dev, p->lead_thread->pid,
+	kfd_smi_event_migration_start(node, p->lead_thread->pid,
 				      start >> PAGE_SHIFT, end >> PAGE_SHIFT,
-				      0, adev->kfd.dev->id, prange->prefetch_loc,
+				      0, node->id, prange->prefetch_loc,
 				      prange->preferred_loc, trigger);
 
 	r = migrate_vma_setup(&migrate);
@@ -440,36 +436,35 @@ svm_migrate_vma_to_vram(struct amdgpu_device *adev, struct svm_range *prange,
 		goto out_free;
 	}
 	if (cpages != npages)
-		pr_debug("partial migration, 0x%lx/0x%llx pages migrated\n",
+		pr_debug("partial migration, 0x%lx/0x%llx pages collected\n",
 			 cpages, npages);
 	else
-		pr_debug("0x%lx pages migrated\n", cpages);
+		pr_debug("0x%lx pages collected\n", cpages);
 
-	r = svm_migrate_copy_to_vram(adev, prange, &migrate, &mfence, scratch, ttm_res_offset);
+	r = svm_migrate_copy_to_vram(node, prange, &migrate, &mfence, scratch, ttm_res_offset);
 	migrate_vma_pages(&migrate);
 
-	pr_debug("successful/cpages/npages 0x%lx/0x%lx/0x%lx\n",
-		svm_migrate_successful_pages(&migrate), cpages, migrate.npages);
-
 	svm_migrate_copy_done(adev, mfence);
 	migrate_vma_finalize(&migrate);
 
-	kfd_smi_event_migration_end(adev->kfd.dev, p->lead_thread->pid,
-				    start >> PAGE_SHIFT, end >> PAGE_SHIFT,
-				    0, adev->kfd.dev->id, trigger);
+	mpages = svm_migrate_successful_pages(&migrate);
+	pr_debug("migrated/collected/requested 0x%lx/0x%lx/0x%lx\n",
+		 mpages, cpages, migrate.npages);
 
-	svm_range_dma_unmap(adev->dev, scratch, 0, npages);
-	svm_range_free_dma_mappings(prange);
+	svm_range_dma_unmap_dev(adev->dev, scratch, 0, npages);
 
 out_free:
 	kvfree(buf);
+	kfd_smi_event_migration_end(node, p->lead_thread->pid,
+				    start >> PAGE_SHIFT, end >> PAGE_SHIFT,
+				    0, node->id, trigger, r);
 out:
-	if (!r && cpages) {
-		pdd = svm_range_get_pdd_by_adev(prange, adev);
+	if (!r && mpages) {
+		pdd = svm_range_get_pdd_by_node(prange, node);
 		if (pdd)
-			WRITE_ONCE(pdd->page_in, pdd->page_in + cpages);
+			WRITE_ONCE(pdd->page_in, pdd->page_in + mpages);
 
-		return cpages;
+		return mpages;
 	}
 	return r;
 }
@@ -478,6 +473,8 @@ out:
  * svm_migrate_ram_to_vram - migrate svm range from system to device
  * @prange: range structure
  * @best_loc: the device to migrate to
+ * @start_mgr: start page to migrate
+ * @last_mgr: last page to migrate
  * @mm: the process mm structure
  * @trigger: reason of migration
  *
@@ -488,39 +485,50 @@ out:
  */
 static int
 svm_migrate_ram_to_vram(struct svm_range *prange, uint32_t best_loc,
+			unsigned long start_mgr, unsigned long last_mgr,
 			struct mm_struct *mm, uint32_t trigger)
 {
 	unsigned long addr, start, end;
 	struct vm_area_struct *vma;
-	struct amdgpu_device *adev;
-	uint64_t ttm_res_offset;
-	unsigned long cpages = 0;
+	u64 ttm_res_offset;
+	struct kfd_node *node;
+	unsigned long mpages = 0;
 	long r = 0;
 
-	if (prange->actual_loc == best_loc) {
-		pr_debug("svms 0x%p [0x%lx 0x%lx] already on best_loc 0x%x\n",
-			 prange->svms, prange->start, prange->last, best_loc);
-		return 0;
+	if (start_mgr < prange->start || last_mgr > prange->last) {
+		pr_debug("range [0x%lx 0x%lx] out prange [0x%lx 0x%lx]\n",
+			 start_mgr, last_mgr, prange->start, prange->last);
+		return -EFAULT;
 	}
 
-	adev = svm_range_get_adev_by_id(prange, best_loc);
-	if (!adev) {
-		pr_debug("failed to get device by id 0x%x\n", best_loc);
+	node = svm_range_get_node_by_id(prange, best_loc);
+	if (!node) {
+		pr_debug("failed to get kfd node by id 0x%x\n", best_loc);
 		return -ENODEV;
 	}
 
-	pr_debug("svms 0x%p [0x%lx 0x%lx] to gpu 0x%x\n", prange->svms,
-		 prange->start, prange->last, best_loc);
+	pr_debug("svms 0x%p [0x%lx 0x%lx] in [0x%lx 0x%lx] to gpu 0x%x\n",
+		prange->svms, start_mgr, last_mgr, prange->start, prange->last,
+		best_loc);
 
-	start = prange->start << PAGE_SHIFT;
-	end = (prange->last + 1) << PAGE_SHIFT;
+	start = start_mgr << PAGE_SHIFT;
+	end = (last_mgr + 1) << PAGE_SHIFT;
 
-	r = svm_range_vram_node_new(adev, prange, true);
+	r = amdgpu_amdkfd_reserve_mem_limit(node->adev,
+					prange->npages * PAGE_SIZE,
+					KFD_IOC_ALLOC_MEM_FLAGS_VRAM,
+					node->xcp ? node->xcp->id : 0);
 	if (r) {
-		dev_dbg(adev->dev, "fail %ld to alloc vram\n", r);
-		return r;
+		dev_dbg(node->adev->dev, "failed to reserve VRAM, r: %ld\n", r);
+		return -ENOSPC;
+	}
+
+	r = svm_range_vram_node_new(node, prange, true);
+	if (r) {
+		dev_dbg(node->adev->dev, "fail %ld to alloc vram\n", r);
+		goto out;
 	}
-	ttm_res_offset = prange->offset << PAGE_SHIFT;
+	ttm_res_offset = (start_mgr - prange->start + prange->offset) << PAGE_SHIFT;
 
 	for (addr = start; addr < end;) {
 		unsigned long next;
@@ -530,27 +538,38 @@ svm_migrate_ram_to_vram(struct svm_range *prange, uint32_t best_loc,
 			break;
 
 		next = min(vma->vm_end, end);
-		r = svm_migrate_vma_to_vram(adev, prange, vma, addr, next, trigger, ttm_res_offset);
+		r = svm_migrate_vma_to_vram(node, prange, vma, addr, next, trigger, ttm_res_offset);
 		if (r < 0) {
 			pr_debug("failed %ld to migrate\n", r);
 			break;
 		} else {
-			cpages += r;
+			mpages += r;
 		}
 		ttm_res_offset += next - addr;
 		addr = next;
 	}
 
-	if (cpages)
+	if (mpages) {
 		prange->actual_loc = best_loc;
-	else
+		prange->vram_pages += mpages;
+	} else if (!prange->actual_loc) {
+		/* if no page migrated and all pages from prange are at
+		 * sys ram drop svm_bo got from svm_range_vram_node_new
+		 */
 		svm_range_vram_node_free(prange);
+	}
 
+out:
+	amdgpu_amdkfd_unreserve_mem_limit(node->adev,
+					prange->npages * PAGE_SIZE,
+					KFD_IOC_ALLOC_MEM_FLAGS_VRAM,
+					node->xcp ? node->xcp->id : 0);
 	return r < 0 ? r : 0;
 }
 
-static void svm_migrate_page_free(struct page *page)
+static void svm_migrate_folio_free(struct folio *folio)
 {
+	struct page *page = &folio->page;
 	struct svm_range_bo *svm_bo = page->zone_device_data;
 
 	if (svm_bo) {
@@ -562,22 +581,22 @@ static void svm_migrate_page_free(struct page *page)
 static int
 svm_migrate_copy_to_ram(struct amdgpu_device *adev, struct svm_range *prange,
 			struct migrate_vma *migrate, struct dma_fence **mfence,
-			dma_addr_t *scratch, uint64_t npages)
+			dma_addr_t *scratch, u64 npages)
 {
 	struct device *dev = adev->dev;
-	uint64_t *src;
+	u64 *src;
 	dma_addr_t *dst;
 	struct page *dpage;
-	uint64_t i = 0, j;
-	uint64_t addr;
+	u64 i = 0, j;
+	u64 addr;
 	int r = 0;
 
 	pr_debug("svms 0x%p [0x%lx 0x%lx]\n", prange->svms, prange->start,
 		 prange->last);
 
-	addr = prange->start << PAGE_SHIFT;
+	addr = migrate->start;
 
-	src = (uint64_t *)(scratch + npages);
+	src = (u64 *)(scratch + npages);
 	dst = scratch;
 
 	for (i = 0, j = 0; i < npages; i++, addr += PAGE_SIZE) {
@@ -617,7 +636,7 @@ svm_migrate_copy_to_ram(struct amdgpu_device *adev, struct svm_range *prange,
 			goto out_oom;
 		}
 
-		dst[i] = dma_map_page(dev, dpage, 0, PAGE_SIZE, DMA_FROM_DEVICE);
+		dst[i] = dma_map_page(dev, dpage, 0, PAGE_SIZE, DMA_BIDIRECTIONAL);
 		r = dma_mapping_error(dev, dst[i]);
 		if (r) {
 			dev_err(adev->dev, "%s: fail %d dma_map_page\n", __func__, r);
@@ -649,28 +668,30 @@ out_oom:
 /**
  * svm_migrate_vma_to_ram - migrate range inside one vma from device to system
  *
- * @adev: amdgpu device to migrate from
  * @prange: svm range structure
  * @vma: vm_area_struct that range [start, end] belongs to
  * @start: range start virtual address in pages
  * @end: range end virtual address in pages
+ * @node: kfd node device to migrate from
+ * @trigger: reason of migration
+ * @fault_page: is from vmf->page, svm_migrate_to_ram(), this is CPU page fault callback
  *
  * Context: Process context, caller hold mmap read lock, prange->migrate_mutex
  *
  * Return:
- *   0 - success with all pages migrated
  *   negative values - indicate error
- *   positive values - partial migration, number of pages not migrated
+ *   positive values or zero - number of pages got migrated
  */
 static long
-svm_migrate_vma_to_ram(struct amdgpu_device *adev, struct svm_range *prange,
-		       struct vm_area_struct *vma, uint64_t start, uint64_t end,
+svm_migrate_vma_to_ram(struct kfd_node *node, struct svm_range *prange,
+		       struct vm_area_struct *vma, u64 start, u64 end,
 		       uint32_t trigger, struct page *fault_page)
 {
 	struct kfd_process *p = container_of(prange->svms, struct kfd_process, svms);
-	uint64_t npages = (end - start) >> PAGE_SHIFT;
-	unsigned long upages = npages;
+	u64 npages = (end - start) >> PAGE_SHIFT;
 	unsigned long cpages = 0;
+	unsigned long mpages = 0;
+	struct amdgpu_device *adev = node->adev;
 	struct kfd_process_device *pdd;
 	struct dma_fence *mfence = NULL;
 	struct migrate_vma migrate = { 0 };
@@ -689,7 +710,7 @@ svm_migrate_vma_to_ram(struct amdgpu_device *adev, struct svm_range *prange,
 		migrate.flags = MIGRATE_VMA_SELECT_DEVICE_PRIVATE;
 
 	buf = kvcalloc(npages,
-		       2 * sizeof(*migrate.src) + sizeof(uint64_t) + sizeof(dma_addr_t),
+		       2 * sizeof(*migrate.src) + sizeof(u64) + sizeof(dma_addr_t),
 		       GFP_KERNEL);
 	if (!buf)
 		goto out;
@@ -699,9 +720,9 @@ svm_migrate_vma_to_ram(struct amdgpu_device *adev, struct svm_range *prange,
 	migrate.fault_page = fault_page;
 	scratch = (dma_addr_t *)(migrate.dst + npages);
 
-	kfd_smi_event_migration_start(adev->kfd.dev, p->lead_thread->pid,
+	kfd_smi_event_migration_start(node, p->lead_thread->pid,
 				      start >> PAGE_SHIFT, end >> PAGE_SHIFT,
-				      adev->kfd.dev->id, 0, prange->prefetch_loc,
+				      node->id, 0, prange->prefetch_loc,
 				      prange->preferred_loc, trigger);
 
 	r = migrate_vma_setup(&migrate);
@@ -715,48 +736,50 @@ svm_migrate_vma_to_ram(struct amdgpu_device *adev, struct svm_range *prange,
 	if (!cpages) {
 		pr_debug("failed collect migrate device pages [0x%lx 0x%lx]\n",
 			 prange->start, prange->last);
-		upages = svm_migrate_unsuccessful_pages(&migrate);
 		goto out_free;
 	}
 	if (cpages != npages)
-		pr_debug("partial migration, 0x%lx/0x%llx pages migrated\n",
+		pr_debug("partial migration, 0x%lx/0x%llx pages collected\n",
 			 cpages, npages);
 	else
-		pr_debug("0x%lx pages migrated\n", cpages);
+		pr_debug("0x%lx pages collected\n", cpages);
 
 	r = svm_migrate_copy_to_ram(adev, prange, &migrate, &mfence,
 				    scratch, npages);
 	migrate_vma_pages(&migrate);
 
-	upages = svm_migrate_unsuccessful_pages(&migrate);
-	pr_debug("unsuccessful/cpages/npages 0x%lx/0x%lx/0x%lx\n",
-		 upages, cpages, migrate.npages);
+	mpages = svm_migrate_successful_pages(&migrate);
+	pr_debug("migrated/collected/requested 0x%lx/0x%lx/0x%lx\n",
+		 mpages, cpages, migrate.npages);
 
 	svm_migrate_copy_done(adev, mfence);
 	migrate_vma_finalize(&migrate);
 
-	kfd_smi_event_migration_end(adev->kfd.dev, p->lead_thread->pid,
-				    start >> PAGE_SHIFT, end >> PAGE_SHIFT,
-				    adev->kfd.dev->id, 0, trigger);
-
-	svm_range_dma_unmap(adev->dev, scratch, 0, npages);
+	svm_range_dma_unmap_dev(adev->dev, scratch, 0, npages);
 
 out_free:
 	kvfree(buf);
+	kfd_smi_event_migration_end(node, p->lead_thread->pid,
+				    start >> PAGE_SHIFT, end >> PAGE_SHIFT,
+				    node->id, 0, trigger, r);
 out:
-	if (!r && cpages) {
-		pdd = svm_range_get_pdd_by_adev(prange, adev);
+	if (!r && mpages) {
+		pdd = svm_range_get_pdd_by_node(prange, node);
 		if (pdd)
-			WRITE_ONCE(pdd->page_out, pdd->page_out + cpages);
+			WRITE_ONCE(pdd->page_out, pdd->page_out + mpages);
 	}
-	return r ? r : upages;
+
+	return r ? r : mpages;
 }
 
 /**
  * svm_migrate_vram_to_ram - migrate svm range from device to system
  * @prange: range structure
  * @mm: process mm, use current->mm if NULL
+ * @start_mgr: start page need be migrated to sys ram
+ * @last_mgr: last page need be migrated to sys ram
  * @trigger: reason of migration
+ * @fault_page: is from vmf->page, svm_migrate_to_ram(), this is CPU page fault callback
  *
  * Context: Process context, caller hold mmap read lock, prange->migrate_mutex
  *
@@ -764,35 +787,41 @@ out:
  * 0 - OK, otherwise error code
  */
 int svm_migrate_vram_to_ram(struct svm_range *prange, struct mm_struct *mm,
+			    unsigned long start_mgr, unsigned long last_mgr,
 			    uint32_t trigger, struct page *fault_page)
 {
-	struct amdgpu_device *adev;
+	struct kfd_node *node;
 	struct vm_area_struct *vma;
 	unsigned long addr;
 	unsigned long start;
 	unsigned long end;
-	unsigned long upages = 0;
+	unsigned long mpages = 0;
 	long r = 0;
 
+	/* this pragne has no any vram page to migrate to sys ram */
 	if (!prange->actual_loc) {
 		pr_debug("[0x%lx 0x%lx] already migrated to ram\n",
 			 prange->start, prange->last);
 		return 0;
 	}
 
-	adev = svm_range_get_adev_by_id(prange, prange->actual_loc);
-	if (!adev) {
-		pr_debug("failed to get device by id 0x%x\n",
-			 prange->actual_loc);
-		return -ENODEV;
+	if (start_mgr < prange->start || last_mgr > prange->last) {
+		pr_debug("range [0x%lx 0x%lx] out prange [0x%lx 0x%lx]\n",
+			 start_mgr, last_mgr, prange->start, prange->last);
+		return -EFAULT;
 	}
 
+	node = svm_range_get_node_by_id(prange, prange->actual_loc);
+	if (!node) {
+		pr_debug("failed to get kfd node by id 0x%x\n", prange->actual_loc);
+		return -ENODEV;
+	}
 	pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx] from gpu 0x%x to ram\n",
-		 prange->svms, prange, prange->start, prange->last,
+		 prange->svms, prange, start_mgr, last_mgr,
 		 prange->actual_loc);
 
-	start = prange->start << PAGE_SHIFT;
-	end = (prange->last + 1) << PAGE_SHIFT;
+	start = start_mgr << PAGE_SHIFT;
+	end = (last_mgr + 1) << PAGE_SHIFT;
 
 	for (addr = start; addr < end;) {
 		unsigned long next;
@@ -805,20 +834,30 @@ int svm_migrate_vram_to_ram(struct svm_range *prange, struct mm_struct *mm,
 		}
 
 		next = min(vma->vm_end, end);
-		r = svm_migrate_vma_to_ram(adev, prange, vma, addr, next, trigger,
+		r = svm_migrate_vma_to_ram(node, prange, vma, addr, next, trigger,
 			fault_page);
 		if (r < 0) {
 			pr_debug("failed %ld to migrate prange %p\n", r, prange);
 			break;
 		} else {
-			upages += r;
+			mpages += r;
 		}
 		addr = next;
 	}
 
-	if (r >= 0 && !upages) {
-		svm_range_vram_node_free(prange);
-		prange->actual_loc = 0;
+	if (r >= 0) {
+		WARN_ONCE(prange->vram_pages < mpages,
+			  "Recorded vram pages(0x%llx) should not be less than migration pages(0x%lx).",
+			  prange->vram_pages, mpages);
+		prange->vram_pages -= mpages;
+
+		/* prange does not have vram page set its actual_loc to system
+		 * and drop its svm_bo ref
+		 */
+		if (prange->vram_pages == 0 && prange->ttm_res) {
+			prange->actual_loc = 0;
+			svm_range_vram_node_free(prange);
+		}
 	}
 
 	return r < 0 ? r : 0;
@@ -828,17 +867,23 @@ int svm_migrate_vram_to_ram(struct svm_range *prange, struct mm_struct *mm,
  * svm_migrate_vram_to_vram - migrate svm range from device to device
  * @prange: range structure
  * @best_loc: the device to migrate to
+ * @start: start page need be migrated to sys ram
+ * @last: last page need be migrated to sys ram
  * @mm: process mm, use current->mm if NULL
  * @trigger: reason of migration
  *
  * Context: Process context, caller hold mmap read lock, svms lock, prange lock
  *
+ * migrate all vram pages in prange to sys ram, then migrate
+ * [start, last] pages from sys ram to gpu node best_loc.
+ *
  * Return:
  * 0 - OK, otherwise error code
  */
 static int
 svm_migrate_vram_to_vram(struct svm_range *prange, uint32_t best_loc,
-			 struct mm_struct *mm, uint32_t trigger)
+			unsigned long start, unsigned long last,
+			struct mm_struct *mm, uint32_t trigger)
 {
 	int r, retries = 3;
 
@@ -850,7 +895,8 @@ svm_migrate_vram_to_vram(struct svm_range *prange, uint32_t best_loc,
 	pr_debug("from gpu 0x%x to gpu 0x%x\n", prange->actual_loc, best_loc);
 
 	do {
-		r = svm_migrate_vram_to_ram(prange, mm, trigger, NULL);
+		r = svm_migrate_vram_to_ram(prange, mm, prange->start, prange->last,
+					    trigger, NULL);
 		if (r)
 			return r;
 	} while (prange->actual_loc && --retries);
@@ -858,17 +904,21 @@ svm_migrate_vram_to_vram(struct svm_range *prange, uint32_t best_loc,
 	if (prange->actual_loc)
 		return -EDEADLK;
 
-	return svm_migrate_ram_to_vram(prange, best_loc, mm, trigger);
+	return svm_migrate_ram_to_vram(prange, best_loc, start, last, mm, trigger);
 }
 
 int
 svm_migrate_to_vram(struct svm_range *prange, uint32_t best_loc,
+		    unsigned long start, unsigned long last,
 		    struct mm_struct *mm, uint32_t trigger)
 {
-	if  (!prange->actual_loc)
-		return svm_migrate_ram_to_vram(prange, best_loc, mm, trigger);
+	if  (!prange->actual_loc || prange->actual_loc == best_loc)
+		return svm_migrate_ram_to_vram(prange, best_loc, start, last,
+					       mm, trigger);
+
 	else
-		return svm_migrate_vram_to_vram(prange, best_loc, mm, trigger);
+		return svm_migrate_vram_to_vram(prange, best_loc, start, last,
+						mm, trigger);
 
 }
 
@@ -884,10 +934,9 @@ svm_migrate_to_vram(struct svm_range *prange, uint32_t best_loc,
  */
 static vm_fault_t svm_migrate_to_ram(struct vm_fault *vmf)
 {
+	unsigned long start, last, size;
 	unsigned long addr = vmf->address;
 	struct svm_range_bo *svm_bo;
-	enum svm_work_list_ops op;
-	struct svm_range *parent;
 	struct svm_range *prange;
 	struct kfd_process *p;
 	struct mm_struct *mm;
@@ -924,51 +973,31 @@ static vm_fault_t svm_migrate_to_ram(struct vm_fault *vmf)
 
 	mutex_lock(&p->svms.lock);
 
-	prange = svm_range_from_addr(&p->svms, addr, &parent);
+	prange = svm_range_from_addr(&p->svms, addr, NULL);
 	if (!prange) {
 		pr_debug("failed get range svms 0x%p addr 0x%lx\n", &p->svms, addr);
 		r = -EFAULT;
 		goto out_unlock_svms;
 	}
 
-	mutex_lock(&parent->migrate_mutex);
-	if (prange != parent)
-		mutex_lock_nested(&prange->migrate_mutex, 1);
+	mutex_lock(&prange->migrate_mutex);
 
 	if (!prange->actual_loc)
 		goto out_unlock_prange;
 
-	svm_range_lock(parent);
-	if (prange != parent)
-		mutex_lock_nested(&prange->lock, 1);
-	r = svm_range_split_by_granularity(p, mm, addr, parent, prange);
-	if (prange != parent)
-		mutex_unlock(&prange->lock);
-	svm_range_unlock(parent);
-	if (r) {
-		pr_debug("failed %d to split range by granularity\n", r);
-		goto out_unlock_prange;
-	}
+	/* Align migration range start and size to granularity size */
+	size = 1UL << prange->granularity;
+	start = max(ALIGN_DOWN(addr, size), prange->start);
+	last = min(ALIGN(addr + 1, size) - 1, prange->last);
 
-	r = svm_migrate_vram_to_ram(prange, vmf->vma->vm_mm,
-				    KFD_MIGRATE_TRIGGER_PAGEFAULT_CPU,
-				    vmf->page);
+	r = svm_migrate_vram_to_ram(prange, vmf->vma->vm_mm, start, last,
+				    KFD_MIGRATE_TRIGGER_PAGEFAULT_CPU, vmf->page);
 	if (r)
 		pr_debug("failed %d migrate svms 0x%p range 0x%p [0x%lx 0x%lx]\n",
-			 r, prange->svms, prange, prange->start, prange->last);
-
-	/* xnack on, update mapping on GPUs with ACCESS_IN_PLACE */
-	if (p->xnack_enabled && parent == prange)
-		op = SVM_OP_UPDATE_RANGE_NOTIFIER_AND_MAP;
-	else
-		op = SVM_OP_UPDATE_RANGE_NOTIFIER;
-	svm_range_add_list_work(&p->svms, parent, mm, op);
-	schedule_deferred_list_work(&p->svms);
+			r, prange->svms, prange, start, last);
 
 out_unlock_prange:
-	if (prange != parent)
-		mutex_unlock(&prange->migrate_mutex);
-	mutex_unlock(&parent->migrate_mutex);
+	mutex_unlock(&prange->migrate_mutex);
 out_unlock_svms:
 	mutex_unlock(&p->svms.lock);
 out_unref_process:
@@ -980,25 +1009,28 @@ out_mmput:
 }
 
 static const struct dev_pagemap_ops svm_migrate_pgmap_ops = {
-	.page_free		= svm_migrate_page_free,
+	.folio_free		= svm_migrate_folio_free,
 	.migrate_to_ram		= svm_migrate_to_ram,
 };
 
 /* Each VRAM page uses sizeof(struct page) on system memory */
 #define SVM_HMM_PAGE_STRUCT_SIZE(size) ((size)/PAGE_SIZE * sizeof(struct page))
 
-int svm_migrate_init(struct amdgpu_device *adev)
+int kgd2kfd_init_zone_device(struct amdgpu_device *adev)
 {
-	struct kfd_dev *kfddev = adev->kfd.dev;
+	struct amdgpu_kfd_dev *kfddev = &adev->kfd;
 	struct dev_pagemap *pgmap;
 	struct resource *res = NULL;
 	unsigned long size;
 	void *r;
 
-	/* Page migration works on Vega10 or newer */
-	if (!KFD_IS_SOC15(kfddev))
+	/* Page migration works on gfx9 or newer */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(9, 0, 1))
 		return -EINVAL;
 
+	if (adev->apu_prefer_gtt)
+		return 0;
+
 	pgmap = &kfddev->pgmap;
 	memset(pgmap, 0, sizeof(*pgmap));
 
@@ -1013,7 +1045,7 @@ int svm_migrate_init(struct amdgpu_device *adev)
 	} else {
 		res = devm_request_free_mem_region(adev->dev, &iomem_resource, size);
 		if (IS_ERR(res))
-			return -ENOMEM;
+			return PTR_ERR(res);
 		pgmap->range.start = res->start;
 		pgmap->range.end = res->end;
 		pgmap->type = MEMORY_DEVICE_PRIVATE;
@@ -1029,10 +1061,10 @@ int svm_migrate_init(struct amdgpu_device *adev)
 	r = devm_memremap_pages(adev->dev, pgmap);
 	if (IS_ERR(r)) {
 		pr_err("failed to register HMM device memory\n");
-		/* Disable SVM support capability */
-		pgmap->type = 0;
 		if (pgmap->type == MEMORY_DEVICE_PRIVATE)
 			devm_release_mem_region(adev->dev, res->start, resource_size(res));
+		/* Disable SVM support capability */
+		pgmap->type = 0;
 		return PTR_ERR(r);
 	}
 
@@ -1041,8 +1073,6 @@ int svm_migrate_init(struct amdgpu_device *adev)
 
 	amdgpu_amdkfd_reserve_system_mem(SVM_HMM_PAGE_STRUCT_SIZE(size));
 
-	svm_range_set_max_pages(adev);
-
 	pr_info("HMM registered %ldMB device memory\n", size >> 20);
 
 	return 0;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.h b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.h
index a5d7e6d22264..2b7fd442d29c 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.h
@@ -31,7 +31,6 @@
 #include <linux/list.h>
 #include <linux/mutex.h>
 #include <linux/sched/mm.h>
-#include <linux/hmm.h>
 #include "kfd_priv.h"
 #include "kfd_svm.h"
 
@@ -41,21 +40,16 @@ enum MIGRATION_COPY_DIR {
 };
 
 int svm_migrate_to_vram(struct svm_range *prange,  uint32_t best_loc,
+			unsigned long start, unsigned long last,
 			struct mm_struct *mm, uint32_t trigger);
+
 int svm_migrate_vram_to_ram(struct svm_range *prange, struct mm_struct *mm,
+			    unsigned long start, unsigned long last,
 			    uint32_t trigger, struct page *fault_page);
+
 unsigned long
 svm_migrate_addr_to_pfn(struct amdgpu_device *adev, unsigned long addr);
 
-int svm_migrate_init(struct amdgpu_device *adev);
-
-#else
-
-static inline int svm_migrate_init(struct amdgpu_device *adev)
-{
-	return 0;
-}
-
 #endif /* IS_ENABLED(CONFIG_HSA_AMD_SVM) */
 
 #endif /* KFD_MIGRATE_H_ */
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_module.c b/drivers/gpu/drm/amd/amdkfd/kfd_module.c
index aee2212e52f6..33aa23450b3f 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_module.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_module.c
@@ -78,8 +78,8 @@ err_ioctl:
 static void kfd_exit(void)
 {
 	kfd_cleanup_processes();
-	kfd_debugfs_fini();
 	kfd_process_destroy_wq();
+	kfd_debugfs_fini();
 	kfd_procfs_shutdown();
 	kfd_topology_shutdown();
 	kfd_chardev_exit();
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
index 623ccd227b7d..d9ae854b6908 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
@@ -46,9 +46,9 @@ int pipe_priority_map[] = {
 	KFD_PIPE_PRIORITY_CS_HIGH
 };
 
-struct kfd_mem_obj *allocate_hiq_mqd(struct kfd_dev *dev, struct queue_properties *q)
+struct kfd_mem_obj *allocate_hiq_mqd(struct kfd_node *dev, struct queue_properties *q)
 {
-	struct kfd_mem_obj *mqd_mem_obj = NULL;
+	struct kfd_mem_obj *mqd_mem_obj;
 
 	mqd_mem_obj = kzalloc(sizeof(struct kfd_mem_obj), GFP_KERNEL);
 	if (!mqd_mem_obj)
@@ -61,10 +61,10 @@ struct kfd_mem_obj *allocate_hiq_mqd(struct kfd_dev *dev, struct queue_propertie
 	return mqd_mem_obj;
 }
 
-struct kfd_mem_obj *allocate_sdma_mqd(struct kfd_dev *dev,
+struct kfd_mem_obj *allocate_sdma_mqd(struct kfd_node *dev,
 					struct queue_properties *q)
 {
-	struct kfd_mem_obj *mqd_mem_obj = NULL;
+	struct kfd_mem_obj *mqd_mem_obj;
 	uint64_t offset;
 
 	mqd_mem_obj = kzalloc(sizeof(struct kfd_mem_obj), GFP_KERNEL);
@@ -72,11 +72,12 @@ struct kfd_mem_obj *allocate_sdma_mqd(struct kfd_dev *dev,
 		return NULL;
 
 	offset = (q->sdma_engine_id *
-		dev->device_info.num_sdma_queues_per_engine +
+		dev->kfd->device_info.num_sdma_queues_per_engine +
 		q->sdma_queue_id) *
 		dev->dqm->mqd_mgrs[KFD_MQD_TYPE_SDMA]->mqd_size;
 
-	offset += dev->dqm->mqd_mgrs[KFD_MQD_TYPE_HIQ]->mqd_size;
+	offset += dev->dqm->mqd_mgrs[KFD_MQD_TYPE_HIQ]->mqd_size *
+		  NUM_XCC(dev->xcc_mask);
 
 	mqd_mem_obj->gtt_mem = (void *)((uint64_t)dev->dqm->hiq_sdma_mqd.gtt_mem
 				+ offset);
@@ -96,35 +97,41 @@ void free_mqd_hiq_sdma(struct mqd_manager *mm, void *mqd,
 
 void mqd_symmetrically_map_cu_mask(struct mqd_manager *mm,
 		const uint32_t *cu_mask, uint32_t cu_mask_count,
-		uint32_t *se_mask)
+		uint32_t *se_mask, uint32_t inst)
 {
-	struct kfd_cu_info cu_info;
+	struct amdgpu_cu_info *cu_info = &mm->dev->adev->gfx.cu_info;
+	struct amdgpu_gfx_config *gfx_info = &mm->dev->adev->gfx.config;
 	uint32_t cu_per_sh[KFD_MAX_NUM_SE][KFD_MAX_NUM_SH_PER_SE] = {0};
 	bool wgp_mode_req = KFD_GC_VERSION(mm->dev) >= IP_VERSION(10, 0, 0);
 	uint32_t en_mask = wgp_mode_req ? 0x3 : 0x1;
-	int i, se, sh, cu, cu_bitmap_sh_mul, inc = wgp_mode_req ? 2 : 1;
+	int i, se, sh, cu, cu_bitmap_sh_mul, cu_inc = wgp_mode_req ? 2 : 1;
+	uint32_t cu_active_per_node;
+	int inc = cu_inc * NUM_XCC(mm->dev->xcc_mask);
+	int xcc_inst = inst + ffs(mm->dev->xcc_mask) - 1;
 
-	amdgpu_amdkfd_get_cu_info(mm->dev->adev, &cu_info);
-
-	if (cu_mask_count > cu_info.cu_active_number)
-		cu_mask_count = cu_info.cu_active_number;
+	cu_active_per_node = cu_info->number / mm->dev->kfd->num_nodes;
+	if (cu_mask_count > cu_active_per_node)
+		cu_mask_count = cu_active_per_node;
 
 	/* Exceeding these bounds corrupts the stack and indicates a coding error.
 	 * Returning with no CU's enabled will hang the queue, which should be
 	 * attention grabbing.
 	 */
-	if (cu_info.num_shader_engines > KFD_MAX_NUM_SE) {
-		pr_err("Exceeded KFD_MAX_NUM_SE, chip reports %d\n", cu_info.num_shader_engines);
+	if (gfx_info->max_shader_engines > KFD_MAX_NUM_SE) {
+		dev_err(mm->dev->adev->dev,
+			"Exceeded KFD_MAX_NUM_SE, chip reports %d\n",
+			gfx_info->max_shader_engines);
 		return;
 	}
-	if (cu_info.num_shader_arrays_per_engine > KFD_MAX_NUM_SH_PER_SE) {
-		pr_err("Exceeded KFD_MAX_NUM_SH, chip reports %d\n",
-			cu_info.num_shader_arrays_per_engine * cu_info.num_shader_engines);
+	if (gfx_info->max_sh_per_se > KFD_MAX_NUM_SH_PER_SE) {
+		dev_err(mm->dev->adev->dev,
+			"Exceeded KFD_MAX_NUM_SH, chip reports %d\n",
+			gfx_info->max_sh_per_se * gfx_info->max_shader_engines);
 		return;
 	}
 
 	cu_bitmap_sh_mul = (KFD_GC_VERSION(mm->dev) >= IP_VERSION(11, 0, 0) &&
-			    KFD_GC_VERSION(mm->dev) < IP_VERSION(12, 0, 0)) ? 2 : 1;
+			    KFD_GC_VERSION(mm->dev) < IP_VERSION(13, 0, 0)) ? 2 : 1;
 
 	/* Count active CUs per SH.
 	 *
@@ -137,10 +144,11 @@ void mqd_symmetrically_map_cu_mask(struct mqd_manager *mm,
 	 * See note on Arcturus cu_bitmap layout in gfx_v9_0_get_cu_info.
 	 * See note on GFX11 cu_bitmap layout in gfx_v11_0_get_cu_info.
 	 */
-	for (se = 0; se < cu_info.num_shader_engines; se++)
-		for (sh = 0; sh < cu_info.num_shader_arrays_per_engine; sh++)
+	for (se = 0; se < gfx_info->max_shader_engines; se++)
+		for (sh = 0; sh < gfx_info->max_sh_per_se; sh++)
 			cu_per_sh[se][sh] = hweight32(
-				cu_info.cu_bitmap[se % 4][sh + (se / 4) * cu_bitmap_sh_mul]);
+				cu_info->bitmap[xcc_inst][se % 4][sh + (se / 4) *
+				cu_bitmap_sh_mul]);
 
 	/* Symmetrically map cu_mask to all SEs & SHs:
 	 * se_mask programs up to 2 SH in the upper and lower 16 bits.
@@ -163,20 +171,33 @@ void mqd_symmetrically_map_cu_mask(struct mqd_manager *mm,
 	 * cu_mask[0] bit8 -> se_mask[0] bit1 (SE0,SH0,CU1)
 	 * ...
 	 *
+	 * For GFX 9.4.3, the following code only looks at a
+	 * subset of the cu_mask corresponding to the inst parameter.
+	 * If we have n XCCs under one GPU node
+	 * cu_mask[0] bit0 -> XCC0 se_mask[0] bit0 (XCC0,SE0,SH0,CU0)
+	 * cu_mask[0] bit1 -> XCC1 se_mask[0] bit0 (XCC1,SE0,SH0,CU0)
+	 * ..
+	 * cu_mask[0] bitn -> XCCn se_mask[0] bit0 (XCCn,SE0,SH0,CU0)
+	 * cu_mask[0] bit n+1 -> XCC0 se_mask[1] bit0 (XCC0,SE1,SH0,CU0)
+	 *
+	 * For example, if there are 6 XCCs under 1 KFD node, this code
+	 * running for each inst, will look at the bits as:
+	 * inst, inst + 6, inst + 12...
+	 *
 	 * First ensure all CUs are disabled, then enable user specified CUs.
 	 */
-	for (i = 0; i < cu_info.num_shader_engines; i++)
+	for (i = 0; i < gfx_info->max_shader_engines; i++)
 		se_mask[i] = 0;
 
-	i = 0;
-	for (cu = 0; cu < 16; cu += inc) {
-		for (sh = 0; sh < cu_info.num_shader_arrays_per_engine; sh++) {
-			for (se = 0; se < cu_info.num_shader_engines; se++) {
+	i = inst;
+	for (cu = 0; cu < 16; cu += cu_inc) {
+		for (sh = 0; sh < gfx_info->max_sh_per_se; sh++) {
+			for (se = 0; se < gfx_info->max_shader_engines; se++) {
 				if (cu_per_sh[se][sh] > cu) {
 					if (cu_mask[i / 32] & (en_mask << (i % 32)))
 						se_mask[se] |= en_mask << (cu + sh * 16);
 					i += inc;
-					if (i == cu_mask_count)
+					if (i >= cu_mask_count)
 						return;
 				}
 			}
@@ -189,7 +210,7 @@ int kfd_hiq_load_mqd_kiq(struct mqd_manager *mm, void *mqd,
 		     struct queue_properties *p, struct mm_struct *mms)
 {
 	return mm->dev->kfd2kgd->hiq_mqd_load(mm->dev->adev, mqd, pipe_id,
-					      queue_id, p->doorbell_off);
+					      queue_id, p->doorbell_off, 0);
 }
 
 int kfd_destroy_mqd_cp(struct mqd_manager *mm, void *mqd,
@@ -197,14 +218,14 @@ int kfd_destroy_mqd_cp(struct mqd_manager *mm, void *mqd,
 		uint32_t pipe_id, uint32_t queue_id)
 {
 	return mm->dev->kfd2kgd->hqd_destroy(mm->dev->adev, mqd, type, timeout,
-						pipe_id, queue_id);
+						pipe_id, queue_id, 0);
 }
 
 void kfd_free_mqd_cp(struct mqd_manager *mm, void *mqd,
 	      struct kfd_mem_obj *mqd_mem_obj)
 {
 	if (mqd_mem_obj->gtt_mem) {
-		amdgpu_amdkfd_free_gtt_mem(mm->dev->adev, mqd_mem_obj->gtt_mem);
+		amdgpu_amdkfd_free_gtt_mem(mm->dev->adev, &mqd_mem_obj->gtt_mem);
 		kfree(mqd_mem_obj);
 	} else {
 		kfd_gtt_sa_free(mm->dev, mqd_mem_obj);
@@ -216,7 +237,7 @@ bool kfd_is_occupied_cp(struct mqd_manager *mm, void *mqd,
 		 uint32_t queue_id)
 {
 	return mm->dev->kfd2kgd->hqd_is_occupied(mm->dev->adev, queue_address,
-						pipe_id, queue_id);
+						pipe_id, queue_id, 0);
 }
 
 int kfd_load_mqd_sdma(struct mqd_manager *mm, void *mqd,
@@ -246,3 +267,46 @@ bool kfd_is_occupied_sdma(struct mqd_manager *mm, void *mqd,
 {
 	return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->adev, mqd);
 }
+
+uint64_t kfd_hiq_mqd_stride(struct kfd_node *dev)
+{
+	return dev->dqm->mqd_mgrs[KFD_MQD_TYPE_HIQ]->mqd_size;
+}
+
+void kfd_get_hiq_xcc_mqd(struct kfd_node *dev, struct kfd_mem_obj *mqd_mem_obj,
+		     uint32_t virtual_xcc_id)
+{
+	uint64_t offset;
+
+	offset = kfd_hiq_mqd_stride(dev) * virtual_xcc_id;
+
+	mqd_mem_obj->gtt_mem = (virtual_xcc_id == 0) ?
+			dev->dqm->hiq_sdma_mqd.gtt_mem : NULL;
+	mqd_mem_obj->gpu_addr = dev->dqm->hiq_sdma_mqd.gpu_addr + offset;
+	mqd_mem_obj->cpu_ptr = (uint32_t *)((uintptr_t)
+				dev->dqm->hiq_sdma_mqd.cpu_ptr + offset);
+}
+
+uint64_t kfd_mqd_stride(struct mqd_manager *mm,
+			struct queue_properties *q)
+{
+	return mm->mqd_size;
+}
+
+bool kfd_check_hiq_mqd_doorbell_id(struct kfd_node *node, uint32_t doorbell_id,
+				   uint32_t inst)
+{
+	if (doorbell_id) {
+		struct device *dev = node->adev->dev;
+
+		if (node->adev->xcp_mgr && node->adev->xcp_mgr->num_xcps > 0)
+			dev_err(dev, "XCC %d: Queue preemption failed for queue with doorbell_id: %x\n",
+							inst, doorbell_id);
+		else
+			dev_err(dev, "Queue preemption failed for queue with doorbell_id: %x\n",
+							doorbell_id);
+		return true;
+	}
+
+	return false;
+}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h
index 57f900ccaa10..17cc1f25c8d0 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h
@@ -68,7 +68,7 @@
  */
 extern int pipe_priority_map[];
 struct mqd_manager {
-	struct kfd_mem_obj*	(*allocate_mqd)(struct kfd_dev *kfd,
+	struct kfd_mem_obj*	(*allocate_mqd)(struct kfd_node *kfd,
 		struct queue_properties *q);
 
 	void	(*init_mqd)(struct mqd_manager *mm, void **mqd,
@@ -97,6 +97,7 @@ struct mqd_manager {
 				uint32_t queue_id);
 
 	int	(*get_wave_state)(struct mqd_manager *mm, void *mqd,
+				  struct queue_properties *q,
 				  void __user *ctl_stack,
 				  u32 *ctl_stack_used_size,
 				  u32 *save_area_used_size);
@@ -118,24 +119,51 @@ struct mqd_manager {
 #if defined(CONFIG_DEBUG_FS)
 	int	(*debugfs_show_mqd)(struct seq_file *m, void *data);
 #endif
-	uint32_t (*read_doorbell_id)(void *mqd);
+	bool (*check_preemption_failed)(struct mqd_manager *mm, void *mqd);
+	uint64_t (*mqd_stride)(struct mqd_manager *mm,
+				struct queue_properties *p);
 
 	struct mutex	mqd_mutex;
-	struct kfd_dev	*dev;
+	struct kfd_node	*dev;
 	uint32_t mqd_size;
 };
 
-struct kfd_mem_obj *allocate_hiq_mqd(struct kfd_dev *dev,
+struct mqd_user_context_save_area_header {
+	/* Byte offset from start of user context
+	 * save area to the last saved top (lowest
+	 * address) of control stack data. Must be
+	 * 4 byte aligned.
+	 */
+	uint32_t control_stack_offset;
+
+	/* Byte size of the last saved control stack
+	 * data. Must be 4 byte aligned.
+	 */
+	uint32_t control_stack_size;
+
+	/* Byte offset from start of user context save
+	 * area to the last saved base (lowest address)
+	 * of wave state data. Must be 4 byte aligned.
+	 */
+	uint32_t wave_state_offset;
+
+	/* Byte size of the last saved wave state data.
+	 * Must be 4 byte aligned.
+	 */
+	uint32_t wave_state_size;
+};
+
+struct kfd_mem_obj *allocate_hiq_mqd(struct kfd_node *dev,
 				struct queue_properties *q);
 
-struct kfd_mem_obj *allocate_sdma_mqd(struct kfd_dev *dev,
+struct kfd_mem_obj *allocate_sdma_mqd(struct kfd_node *dev,
 					struct queue_properties *q);
 void free_mqd_hiq_sdma(struct mqd_manager *mm, void *mqd,
 				struct kfd_mem_obj *mqd_mem_obj);
 
 void mqd_symmetrically_map_cu_mask(struct mqd_manager *mm,
 		const uint32_t *cu_mask, uint32_t cu_mask_count,
-		uint32_t *se_mask);
+		uint32_t *se_mask, uint32_t inst);
 
 int kfd_hiq_load_mqd_kiq(struct mqd_manager *mm, void *mqd,
 		uint32_t pipe_id, uint32_t queue_id,
@@ -164,4 +192,12 @@ bool kfd_is_occupied_sdma(struct mqd_manager *mm, void *mqd,
 		uint64_t queue_address, uint32_t pipe_id,
 		uint32_t queue_id);
 
+void kfd_get_hiq_xcc_mqd(struct kfd_node *dev,
+		struct kfd_mem_obj *mqd_mem_obj, uint32_t virtual_xcc_id);
+
+uint64_t kfd_hiq_mqd_stride(struct kfd_node *dev);
+uint64_t kfd_mqd_stride(struct mqd_manager *mm,
+			struct queue_properties *q);
+bool kfd_check_hiq_mqd_doorbell_id(struct kfd_node *node, uint32_t doorbell_id,
+				   uint32_t inst);
 #endif /* KFD_MQD_MANAGER_H_ */
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c
index 4889865c725c..05f3ac2eaef9 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c
@@ -48,12 +48,11 @@ static void update_cu_mask(struct mqd_manager *mm, void *mqd,
 	struct cik_mqd *m;
 	uint32_t se_mask[4] = {0}; /* 4 is the max # of SEs */
 
-	if (!minfo || (minfo->update_flag != UPDATE_FLAG_CU_MASK) ||
-	    !minfo->cu_mask.ptr)
+	if (!minfo || !minfo->cu_mask.ptr)
 		return;
 
 	mqd_symmetrically_map_cu_mask(mm,
-		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask);
+		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask, 0);
 
 	m = get_mqd(mqd);
 	m->compute_static_thread_mgmt_se0 = se_mask[0];
@@ -74,7 +73,7 @@ static void set_priority(struct cik_mqd *m, struct queue_properties *q)
 	m->cp_hqd_queue_priority = q->priority;
 }
 
-static struct kfd_mem_obj *allocate_mqd(struct kfd_dev *kfd,
+static struct kfd_mem_obj *allocate_mqd(struct kfd_node *kfd,
 					struct queue_properties *q)
 {
 	struct kfd_mem_obj *mqd_mem_obj;
@@ -167,7 +166,7 @@ static int load_mqd(struct mqd_manager *mm, void *mqd, uint32_t pipe_id,
 
 	return mm->dev->kfd2kgd->hqd_load(mm->dev->adev, mqd, pipe_id, queue_id,
 					  (uint32_t __user *)p->write_ptr,
-					  wptr_shift, wptr_mask, mms);
+					  wptr_shift, wptr_mask, mms, 0);
 }
 
 static void __update_mqd(struct mqd_manager *mm, void *mqd,
@@ -207,23 +206,16 @@ static void __update_mqd(struct mqd_manager *mm, void *mqd,
 	q->is_active = QUEUE_IS_ACTIVE(*q);
 }
 
-static void update_mqd(struct mqd_manager *mm, void *mqd,
-			struct queue_properties *q,
-			struct mqd_update_info *minfo)
-{
-	__update_mqd(mm, mqd, q, minfo, 1);
-}
-
-static uint32_t read_doorbell_id(void *mqd)
+static bool check_preemption_failed(struct mqd_manager *mm, void *mqd)
 {
 	struct cik_mqd *m = (struct cik_mqd *)mqd;
 
-	return m->queue_doorbell_id0;
+	return kfd_check_hiq_mqd_doorbell_id(mm->dev, m->queue_doorbell_id0, 0);
 }
 
-static void update_mqd_hawaii(struct mqd_manager *mm, void *mqd,
-			struct queue_properties *q,
-			struct mqd_update_info *minfo)
+static void update_mqd(struct mqd_manager *mm, void *mqd,
+		       struct queue_properties *q,
+		       struct mqd_update_info *minfo)
 {
 	__update_mqd(mm, mqd, q, minfo, 0);
 }
@@ -388,9 +380,8 @@ static int debugfs_show_mqd_sdma(struct seq_file *m, void *data)
 
 #endif
 
-
 struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type,
-		struct kfd_dev *dev)
+		struct kfd_node *dev)
 {
 	struct mqd_manager *mqd;
 
@@ -428,10 +419,11 @@ struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type,
 		mqd->destroy_mqd = kfd_destroy_mqd_cp;
 		mqd->is_occupied = kfd_is_occupied_cp;
 		mqd->mqd_size = sizeof(struct cik_mqd);
+		mqd->mqd_stride = kfd_mqd_stride;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd;
 #endif
-		mqd->read_doorbell_id = read_doorbell_id;
+		mqd->check_preemption_failed = check_preemption_failed;
 		break;
 	case KFD_MQD_TYPE_DIQ:
 		mqd->allocate_mqd = allocate_mqd;
@@ -442,6 +434,7 @@ struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type,
 		mqd->destroy_mqd = kfd_destroy_mqd_cp;
 		mqd->is_occupied = kfd_is_occupied_cp;
 		mqd->mqd_size = sizeof(struct cik_mqd);
+		mqd->mqd_stride = kfd_mqd_stride;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd;
 #endif
@@ -457,6 +450,7 @@ struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type,
 		mqd->checkpoint_mqd = checkpoint_mqd_sdma;
 		mqd->restore_mqd = restore_mqd_sdma;
 		mqd->mqd_size = sizeof(struct cik_sdma_rlc_registers);
+		mqd->mqd_stride = kfd_mqd_stride;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd_sdma;
 #endif
@@ -468,16 +462,3 @@ struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type,
 
 	return mqd;
 }
-
-struct mqd_manager *mqd_manager_init_cik_hawaii(enum KFD_MQD_TYPE type,
-			struct kfd_dev *dev)
-{
-	struct mqd_manager *mqd;
-
-	mqd = mqd_manager_init_cik(type, dev);
-	if (!mqd)
-		return NULL;
-	if (type == KFD_MQD_TYPE_CP)
-		mqd->update_mqd = update_mqd_hawaii;
-	return mqd;
-}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c
index d3e2b6a599a4..1695dd78ede8 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c
@@ -48,12 +48,11 @@ static void update_cu_mask(struct mqd_manager *mm, void *mqd,
 	struct v10_compute_mqd *m;
 	uint32_t se_mask[4] = {0}; /* 4 is the max # of SEs */
 
-	if (!minfo || (minfo->update_flag != UPDATE_FLAG_CU_MASK) ||
-	    !minfo->cu_mask.ptr)
+	if (!minfo || !minfo->cu_mask.ptr)
 		return;
 
 	mqd_symmetrically_map_cu_mask(mm,
-		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask);
+		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask, 0);
 
 	m = get_mqd(mqd);
 	m->compute_static_thread_mgmt_se0 = se_mask[0];
@@ -74,7 +73,7 @@ static void set_priority(struct v10_compute_mqd *m, struct queue_properties *q)
 	m->cp_hqd_queue_priority = q->priority;
 }
 
-static struct kfd_mem_obj *allocate_mqd(struct kfd_dev *kfd,
+static struct kfd_mem_obj *allocate_mqd(struct kfd_node *kfd,
 		struct queue_properties *q)
 {
 	struct kfd_mem_obj *mqd_mem_obj;
@@ -108,6 +107,8 @@ static void init_mqd(struct mqd_manager *mm, void **mqd,
 	m->cp_hqd_persistent_state = CP_HQD_PERSISTENT_STATE__PRELOAD_REQ_MASK |
 			0x53 << CP_HQD_PERSISTENT_STATE__PRELOAD_SIZE__SHIFT;
 
+	m->cp_hqd_pq_control = 5 << CP_HQD_PQ_CONTROL__RPTR_BLOCK_SIZE__SHIFT;
+	m->cp_hqd_pq_control |= CP_HQD_PQ_CONTROL__UNORD_DISPATCH_MASK;
 	m->cp_mqd_control = 1 << CP_MQD_CONTROL__PRIV_STATE__SHIFT;
 
 	m->cp_mqd_base_addr_lo        = lower_32_bits(addr);
@@ -117,12 +118,17 @@ static void init_mqd(struct mqd_manager *mm, void **mqd,
 			1 << CP_HQD_QUANTUM__QUANTUM_SCALE__SHIFT |
 			1 << CP_HQD_QUANTUM__QUANTUM_DURATION__SHIFT;
 
+	/* Set cp_hqd_hq_scheduler0 bit 14 to 1 to have the CP set up the
+	 * DISPATCH_PTR.  This is required for the kfd debugger
+	 */
+	m->cp_hqd_hq_scheduler0 = 1 << 14;
+
 	if (q->format == KFD_QUEUE_FORMAT_AQL) {
 		m->cp_hqd_aql_control =
 			1 << CP_HQD_AQL_CONTROL__CONTROL0__SHIFT;
 	}
 
-	if (mm->dev->cwsr_enabled) {
+	if (mm->dev->kfd->cwsr_enabled) {
 		m->cp_hqd_persistent_state |=
 			(1 << CP_HQD_PERSISTENT_STATE__QSWITCH_MODE__SHIFT);
 		m->cp_hqd_ctx_save_base_addr_lo =
@@ -151,7 +157,7 @@ static int load_mqd(struct mqd_manager *mm, void *mqd,
 
 	r = mm->dev->kfd2kgd->hqd_load(mm->dev->adev, mqd, pipe_id, queue_id,
 					  (uint32_t __user *)p->write_ptr,
-					  wptr_shift, 0, mms);
+					  wptr_shift, 0, mms, 0);
 	return r;
 }
 
@@ -163,9 +169,10 @@ static void update_mqd(struct mqd_manager *mm, void *mqd,
 
 	m = get_mqd(mqd);
 
-	m->cp_hqd_pq_control = 5 << CP_HQD_PQ_CONTROL__RPTR_BLOCK_SIZE__SHIFT;
+	m->cp_hqd_pq_control &= ~CP_HQD_PQ_CONTROL__QUEUE_SIZE_MASK;
 	m->cp_hqd_pq_control |=
 			ffs(q->queue_size / sizeof(unsigned int)) - 1 - 1;
+
 	pr_debug("cp_hqd_pq_control 0x%x\n", m->cp_hqd_pq_control);
 
 	m->cp_hqd_pq_base_lo = lower_32_bits((uint64_t)q->queue_address >> 8);
@@ -210,7 +217,7 @@ static void update_mqd(struct mqd_manager *mm, void *mqd,
 		m->cp_hqd_pq_doorbell_control |=
 			1 << CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_BIF_DROP__SHIFT;
 	}
-	if (mm->dev->cwsr_enabled)
+	if (mm->dev->kfd->cwsr_enabled)
 		m->cp_hqd_ctx_save_control = 0;
 
 	update_cu_mask(mm, mqd, minfo);
@@ -219,19 +226,21 @@ static void update_mqd(struct mqd_manager *mm, void *mqd,
 	q->is_active = QUEUE_IS_ACTIVE(*q);
 }
 
-static uint32_t read_doorbell_id(void *mqd)
+static bool check_preemption_failed(struct mqd_manager *mm, void *mqd)
 {
 	struct v10_compute_mqd *m = (struct v10_compute_mqd *)mqd;
 
-	return m->queue_doorbell_id0;
+	return kfd_check_hiq_mqd_doorbell_id(mm->dev, m->queue_doorbell_id0, 0);
 }
 
 static int get_wave_state(struct mqd_manager *mm, void *mqd,
+			  struct queue_properties *q,
 			  void __user *ctl_stack,
 			  u32 *ctl_stack_used_size,
 			  u32 *save_area_used_size)
 {
 	struct v10_compute_mqd *m;
+	struct kfd_context_save_area_header header;
 
 	m = get_mqd(mqd);
 
@@ -250,6 +259,15 @@ static int get_wave_state(struct mqd_manager *mm, void *mqd,
 	 * accessible to user mode
 	 */
 
+	header.wave_state.control_stack_size = *ctl_stack_used_size;
+	header.wave_state.wave_state_size = *save_area_used_size;
+
+	header.wave_state.wave_state_offset = m->cp_hqd_wg_state_offset;
+	header.wave_state.control_stack_offset = m->cp_hqd_cntl_stack_offset;
+
+	if (copy_to_user(ctl_stack, &header, sizeof(header.wave_state)))
+		return -EFAULT;
+
 	return 0;
 }
 
@@ -303,6 +321,26 @@ static void init_mqd_hiq(struct mqd_manager *mm, void **mqd,
 			1 << CP_HQD_PQ_CONTROL__KMD_QUEUE__SHIFT;
 }
 
+static int destroy_hiq_mqd(struct mqd_manager *mm, void *mqd,
+			enum kfd_preempt_type type, unsigned int timeout,
+			uint32_t pipe_id, uint32_t queue_id)
+{
+	int err;
+	struct v10_compute_mqd *m;
+	u32 doorbell_off;
+
+	m = get_mqd(mqd);
+
+	doorbell_off = m->cp_hqd_pq_doorbell_control >>
+			CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET__SHIFT;
+
+	err = amdgpu_amdkfd_unmap_hiq(mm->dev->adev, doorbell_off, 0);
+	if (err)
+		pr_debug("Destroy HIQ MQD failed: %d\n", err);
+
+	return err;
+}
+
 static void init_mqd_sdma(struct mqd_manager *mm, void **mqd,
 		struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr,
 		struct queue_properties *q)
@@ -405,7 +443,7 @@ static int debugfs_show_mqd_sdma(struct seq_file *m, void *data)
 #endif
 
 struct mqd_manager *mqd_manager_init_v10(enum KFD_MQD_TYPE type,
-		struct kfd_dev *dev)
+		struct kfd_node *dev)
 {
 	struct mqd_manager *mqd;
 
@@ -432,6 +470,7 @@ struct mqd_manager *mqd_manager_init_v10(enum KFD_MQD_TYPE type,
 		mqd->get_wave_state = get_wave_state;
 		mqd->checkpoint_mqd = checkpoint_mqd;
 		mqd->restore_mqd = restore_mqd;
+		mqd->mqd_stride = kfd_mqd_stride;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd;
 #endif
@@ -444,13 +483,14 @@ struct mqd_manager *mqd_manager_init_v10(enum KFD_MQD_TYPE type,
 		mqd->free_mqd = free_mqd_hiq_sdma;
 		mqd->load_mqd = kfd_hiq_load_mqd_kiq;
 		mqd->update_mqd = update_mqd;
-		mqd->destroy_mqd = kfd_destroy_mqd_cp;
+		mqd->destroy_mqd = destroy_hiq_mqd;
 		mqd->is_occupied = kfd_is_occupied_cp;
 		mqd->mqd_size = sizeof(struct v10_compute_mqd);
+		mqd->mqd_stride = kfd_mqd_stride;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd;
 #endif
-		mqd->read_doorbell_id = read_doorbell_id;
+		mqd->check_preemption_failed = check_preemption_failed;
 		pr_debug("%s@%i\n", __func__, __LINE__);
 		break;
 	case KFD_MQD_TYPE_DIQ:
@@ -478,6 +518,7 @@ struct mqd_manager *mqd_manager_init_v10(enum KFD_MQD_TYPE type,
 		mqd->checkpoint_mqd = checkpoint_mqd_sdma;
 		mqd->restore_mqd = restore_mqd_sdma;
 		mqd->mqd_size = sizeof(struct v10_sdma_mqd);
+		mqd->mqd_stride = kfd_mqd_stride;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd_sdma;
 #endif
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c
index 5aa75f72caa1..3c0ae28c5923 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c
@@ -46,15 +46,33 @@ static void update_cu_mask(struct mqd_manager *mm, void *mqd,
 {
 	struct v11_compute_mqd *m;
 	uint32_t se_mask[KFD_MAX_NUM_SE] = {0};
+	bool has_wa_flag = minfo && (minfo->update_flag & (UPDATE_FLAG_DBG_WA_ENABLE |
+			UPDATE_FLAG_DBG_WA_DISABLE));
 
-	if (!minfo || (minfo->update_flag != UPDATE_FLAG_CU_MASK) ||
-	    !minfo->cu_mask.ptr)
+	if (!minfo || !(has_wa_flag || minfo->cu_mask.ptr))
 		return;
 
+	m = get_mqd(mqd);
+
+	if (has_wa_flag) {
+		uint32_t wa_mask =
+			(minfo->update_flag & UPDATE_FLAG_DBG_WA_ENABLE) ? 0xffff : 0xffffffff;
+
+		m->compute_static_thread_mgmt_se0 = wa_mask;
+		m->compute_static_thread_mgmt_se1 = wa_mask;
+		m->compute_static_thread_mgmt_se2 = wa_mask;
+		m->compute_static_thread_mgmt_se3 = wa_mask;
+		m->compute_static_thread_mgmt_se4 = wa_mask;
+		m->compute_static_thread_mgmt_se5 = wa_mask;
+		m->compute_static_thread_mgmt_se6 = wa_mask;
+		m->compute_static_thread_mgmt_se7 = wa_mask;
+
+		return;
+	}
+
 	mqd_symmetrically_map_cu_mask(mm,
-		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask);
+		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask, 0);
 
-	m = get_mqd(mqd);
 	m->compute_static_thread_mgmt_se0 = se_mask[0];
 	m->compute_static_thread_mgmt_se1 = se_mask[1];
 	m->compute_static_thread_mgmt_se2 = se_mask[2];
@@ -81,7 +99,7 @@ static void set_priority(struct v11_compute_mqd *m, struct queue_properties *q)
 	m->cp_hqd_queue_priority = q->priority;
 }
 
-static struct kfd_mem_obj *allocate_mqd(struct kfd_dev *kfd,
+static struct kfd_mem_obj *allocate_mqd(struct kfd_node *node,
 		struct queue_properties *q)
 {
 	struct kfd_mem_obj *mqd_mem_obj;
@@ -91,12 +109,12 @@ static struct kfd_mem_obj *allocate_mqd(struct kfd_dev *kfd,
 	 * MES write to areas beyond MQD size. So allocate
 	 * 1 PAGE_SIZE memory for MQD is MES is enabled.
 	 */
-	if (kfd->shared_resources.enable_mes)
+	if (node->kfd->shared_resources.enable_mes)
 		size = PAGE_SIZE;
 	else
 		size = sizeof(struct v11_compute_mqd);
 
-	if (kfd_gtt_sa_allocate(kfd, size, &mqd_mem_obj))
+	if (kfd_gtt_sa_allocate(node, size, &mqd_mem_obj))
 		return NULL;
 
 	return mqd_mem_obj;
@@ -109,11 +127,12 @@ static void init_mqd(struct mqd_manager *mm, void **mqd,
 	uint64_t addr;
 	struct v11_compute_mqd *m;
 	int size;
+	uint32_t wa_mask = q->is_dbg_wa ? 0xffff : 0xffffffff;
 
 	m = (struct v11_compute_mqd *) mqd_mem_obj->cpu_ptr;
 	addr = mqd_mem_obj->gpu_addr;
 
-	if (mm->dev->shared_resources.enable_mes)
+	if (mm->dev->kfd->shared_resources.enable_mes)
 		size = PAGE_SIZE;
 	else
 		size = sizeof(struct v11_compute_mqd);
@@ -122,18 +141,21 @@ static void init_mqd(struct mqd_manager *mm, void **mqd,
 
 	m->header = 0xC0310800;
 	m->compute_pipelinestat_enable = 1;
-	m->compute_static_thread_mgmt_se0 = 0xFFFFFFFF;
-	m->compute_static_thread_mgmt_se1 = 0xFFFFFFFF;
-	m->compute_static_thread_mgmt_se2 = 0xFFFFFFFF;
-	m->compute_static_thread_mgmt_se3 = 0xFFFFFFFF;
-	m->compute_static_thread_mgmt_se4 = 0xFFFFFFFF;
-	m->compute_static_thread_mgmt_se5 = 0xFFFFFFFF;
-	m->compute_static_thread_mgmt_se6 = 0xFFFFFFFF;
-	m->compute_static_thread_mgmt_se7 = 0xFFFFFFFF;
+
+	m->compute_static_thread_mgmt_se0 = wa_mask;
+	m->compute_static_thread_mgmt_se1 = wa_mask;
+	m->compute_static_thread_mgmt_se2 = wa_mask;
+	m->compute_static_thread_mgmt_se3 = wa_mask;
+	m->compute_static_thread_mgmt_se4 = wa_mask;
+	m->compute_static_thread_mgmt_se5 = wa_mask;
+	m->compute_static_thread_mgmt_se6 = wa_mask;
+	m->compute_static_thread_mgmt_se7 = wa_mask;
 
 	m->cp_hqd_persistent_state = CP_HQD_PERSISTENT_STATE__PRELOAD_REQ_MASK |
 			0x55 << CP_HQD_PERSISTENT_STATE__PRELOAD_SIZE__SHIFT;
 
+	m->cp_hqd_pq_control = 5 << CP_HQD_PQ_CONTROL__RPTR_BLOCK_SIZE__SHIFT;
+	m->cp_hqd_pq_control |= CP_HQD_PQ_CONTROL__UNORD_DISPATCH_MASK;
 	m->cp_mqd_control = 1 << CP_MQD_CONTROL__PRIV_STATE__SHIFT;
 
 	m->cp_mqd_base_addr_lo        = lower_32_bits(addr);
@@ -143,6 +165,11 @@ static void init_mqd(struct mqd_manager *mm, void **mqd,
 			1 << CP_HQD_QUANTUM__QUANTUM_SCALE__SHIFT |
 			1 << CP_HQD_QUANTUM__QUANTUM_DURATION__SHIFT;
 
+	/* Set cp_hqd_hq_scheduler0 bit 14 to 1 to have the CP set up the
+	 * DISPATCH_PTR.  This is required for the kfd debugger
+	 */
+	m->cp_hqd_hq_status0 = 1 << 14;
+
 	/*
 	 * GFX11 RS64 CPFW version >= 509 supports PCIe atomics support
 	 * acknowledgment.
@@ -155,7 +182,7 @@ static void init_mqd(struct mqd_manager *mm, void **mqd,
 			1 << CP_HQD_AQL_CONTROL__CONTROL0__SHIFT;
 	}
 
-	if (mm->dev->cwsr_enabled) {
+	if (mm->dev->kfd->cwsr_enabled) {
 		m->cp_hqd_persistent_state |=
 			(1 << CP_HQD_PERSISTENT_STATE__QSWITCH_MODE__SHIFT);
 		m->cp_hqd_ctx_save_base_addr_lo =
@@ -184,7 +211,7 @@ static int load_mqd(struct mqd_manager *mm, void *mqd,
 
 	r = mm->dev->kfd2kgd->hqd_load(mm->dev->adev, mqd, pipe_id, queue_id,
 					  (uint32_t __user *)p->write_ptr,
-					  wptr_shift, 0, mms);
+					  wptr_shift, 0, mms, 0);
 	return r;
 }
 
@@ -196,7 +223,7 @@ static void update_mqd(struct mqd_manager *mm, void *mqd,
 
 	m = get_mqd(mqd);
 
-	m->cp_hqd_pq_control = 5 << CP_HQD_PQ_CONTROL__RPTR_BLOCK_SIZE__SHIFT;
+	m->cp_hqd_pq_control &= ~CP_HQD_PQ_CONTROL__QUEUE_SIZE_MASK;
 	m->cp_hqd_pq_control |=
 			ffs(q->queue_size / sizeof(unsigned int)) - 1 - 1;
 	pr_debug("cp_hqd_pq_control 0x%x\n", m->cp_hqd_pq_control);
@@ -243,7 +270,7 @@ static void update_mqd(struct mqd_manager *mm, void *mqd,
 		m->cp_hqd_pq_doorbell_control |=
 			1 << CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_BIF_DROP__SHIFT;
 	}
-	if (mm->dev->cwsr_enabled)
+	if (mm->dev->kfd->cwsr_enabled)
 		m->cp_hqd_ctx_save_control = 0;
 
 	update_cu_mask(mm, mqd, minfo);
@@ -252,20 +279,21 @@ static void update_mqd(struct mqd_manager *mm, void *mqd,
 	q->is_active = QUEUE_IS_ACTIVE(*q);
 }
 
-static uint32_t read_doorbell_id(void *mqd)
+static bool check_preemption_failed(struct mqd_manager *mm, void *mqd)
 {
 	struct v11_compute_mqd *m = (struct v11_compute_mqd *)mqd;
 
-	return m->queue_doorbell_id0;
+	return kfd_check_hiq_mqd_doorbell_id(mm->dev, m->queue_doorbell_id0, 0);
 }
 
 static int get_wave_state(struct mqd_manager *mm, void *mqd,
+			  struct queue_properties *q,
 			  void __user *ctl_stack,
 			  u32 *ctl_stack_used_size,
 			  u32 *save_area_used_size)
 {
 	struct v11_compute_mqd *m;
-	/*struct mqd_user_context_save_area_header header;*/
+	struct kfd_context_save_area_header header;
 
 	m = get_mqd(mqd);
 
@@ -283,19 +311,55 @@ static int get_wave_state(struct mqd_manager *mm, void *mqd,
 	 * it's part of the context save area that is already
 	 * accessible to user mode
 	 */
-/*
-	header.control_stack_size = *ctl_stack_used_size;
-	header.wave_state_size = *save_area_used_size;
+	header.wave_state.control_stack_size = *ctl_stack_used_size;
+	header.wave_state.wave_state_size = *save_area_used_size;
 
-	header.wave_state_offset = m->cp_hqd_wg_state_offset;
-	header.control_stack_offset = m->cp_hqd_cntl_stack_offset;
+	header.wave_state.wave_state_offset = m->cp_hqd_wg_state_offset;
+	header.wave_state.control_stack_offset = m->cp_hqd_cntl_stack_offset;
 
-	if (copy_to_user(ctl_stack, &header, sizeof(header)))
+	if (copy_to_user(ctl_stack, &header, sizeof(header.wave_state)))
 		return -EFAULT;
-*/
+
 	return 0;
 }
 
+static void checkpoint_mqd(struct mqd_manager *mm, void *mqd, void *mqd_dst, void *ctl_stack_dst)
+{
+	struct v11_compute_mqd *m;
+
+	m = get_mqd(mqd);
+
+	memcpy(mqd_dst, m, sizeof(struct v11_compute_mqd));
+}
+
+static void restore_mqd(struct mqd_manager *mm, void **mqd,
+			struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr,
+			struct queue_properties *qp,
+			const void *mqd_src,
+			const void *ctl_stack_src, const u32 ctl_stack_size)
+{
+	uint64_t addr;
+	struct v11_compute_mqd *m;
+
+	m = (struct v11_compute_mqd *) mqd_mem_obj->cpu_ptr;
+	addr = mqd_mem_obj->gpu_addr;
+
+	memcpy(m, mqd_src, sizeof(*m));
+
+	*mqd = m;
+	if (gart_addr)
+		*gart_addr = addr;
+
+	m->cp_hqd_pq_doorbell_control =
+		qp->doorbell_off <<
+			CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET__SHIFT;
+	pr_debug("cp_hqd_pq_doorbell_control 0x%x\n",
+			m->cp_hqd_pq_doorbell_control);
+
+	qp->is_active = 0;
+}
+
+
 static void init_mqd_hiq(struct mqd_manager *mm, void **mqd,
 			struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr,
 			struct queue_properties *q)
@@ -310,6 +374,26 @@ static void init_mqd_hiq(struct mqd_manager *mm, void **mqd,
 			1 << CP_HQD_PQ_CONTROL__KMD_QUEUE__SHIFT;
 }
 
+static int destroy_hiq_mqd(struct mqd_manager *mm, void *mqd,
+			enum kfd_preempt_type type, unsigned int timeout,
+			uint32_t pipe_id, uint32_t queue_id)
+{
+	int err;
+	struct v11_compute_mqd *m;
+	u32 doorbell_off;
+
+	m = get_mqd(mqd);
+
+	doorbell_off = m->cp_hqd_pq_doorbell_control >>
+			CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET__SHIFT;
+
+	err = amdgpu_amdkfd_unmap_hiq(mm->dev->adev, doorbell_off, 0);
+	if (err)
+		pr_debug("Destroy HIQ MQD failed: %d\n", err);
+
+	return err;
+}
+
 static void init_mqd_sdma(struct mqd_manager *mm, void **mqd,
 		struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr,
 		struct queue_properties *q)
@@ -319,7 +403,7 @@ static void init_mqd_sdma(struct mqd_manager *mm, void **mqd,
 
 	m = (struct v11_sdma_mqd *) mqd_mem_obj->cpu_ptr;
 
-	if (mm->dev->shared_resources.enable_mes)
+	if (mm->dev->kfd->shared_resources.enable_mes)
 		size = PAGE_SIZE;
 	else
 		size = sizeof(struct v11_sdma_mqd);
@@ -387,7 +471,7 @@ static int debugfs_show_mqd_sdma(struct seq_file *m, void *data)
 #endif
 
 struct mqd_manager *mqd_manager_init_v11(enum KFD_MQD_TYPE type,
-		struct kfd_dev *dev)
+		struct kfd_node *dev)
 {
 	struct mqd_manager *mqd;
 
@@ -412,6 +496,9 @@ struct mqd_manager *mqd_manager_init_v11(enum KFD_MQD_TYPE type,
 		mqd->is_occupied = kfd_is_occupied_cp;
 		mqd->mqd_size = sizeof(struct v11_compute_mqd);
 		mqd->get_wave_state = get_wave_state;
+		mqd->mqd_stride = kfd_mqd_stride;
+		mqd->checkpoint_mqd = checkpoint_mqd;
+		mqd->restore_mqd = restore_mqd;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd;
 #endif
@@ -424,13 +511,14 @@ struct mqd_manager *mqd_manager_init_v11(enum KFD_MQD_TYPE type,
 		mqd->free_mqd = free_mqd_hiq_sdma;
 		mqd->load_mqd = kfd_hiq_load_mqd_kiq;
 		mqd->update_mqd = update_mqd;
-		mqd->destroy_mqd = kfd_destroy_mqd_cp;
+		mqd->destroy_mqd = destroy_hiq_mqd;
 		mqd->is_occupied = kfd_is_occupied_cp;
 		mqd->mqd_size = sizeof(struct v11_compute_mqd);
+		mqd->mqd_stride = kfd_mqd_stride;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd;
 #endif
-		mqd->read_doorbell_id = read_doorbell_id;
+		mqd->check_preemption_failed = check_preemption_failed;
 		pr_debug("%s@%i\n", __func__, __LINE__);
 		break;
 	case KFD_MQD_TYPE_DIQ:
@@ -455,7 +543,10 @@ struct mqd_manager *mqd_manager_init_v11(enum KFD_MQD_TYPE type,
 		mqd->update_mqd = update_mqd_sdma;
 		mqd->destroy_mqd = kfd_destroy_mqd_sdma;
 		mqd->is_occupied = kfd_is_occupied_sdma;
+		mqd->checkpoint_mqd = checkpoint_mqd;
+		mqd->restore_mqd = restore_mqd;
 		mqd->mqd_size = sizeof(struct v11_sdma_mqd);
+		mqd->mqd_stride = kfd_mqd_stride;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd_sdma;
 #endif
@@ -463,7 +554,7 @@ struct mqd_manager *mqd_manager_init_v11(enum KFD_MQD_TYPE type,
 		 * To allocate SDMA MQDs by generic functions
 		 * when MES is enabled.
 		 */
-		if (dev->shared_resources.enable_mes) {
+		if (dev->kfd->shared_resources.enable_mes) {
 			mqd->allocate_mqd = allocate_mqd;
 			mqd->free_mqd = kfd_free_mqd_cp;
 		}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v12.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v12.c
new file mode 100644
index 000000000000..565858b9044d
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v12.c
@@ -0,0 +1,459 @@
+// SPDX-License-Identifier: GPL-2.0 OR MIT
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/printk.h>
+#include <linux/slab.h>
+#include <linux/uaccess.h>
+#include "kfd_priv.h"
+#include "kfd_mqd_manager.h"
+#include "v12_structs.h"
+#include "gc/gc_12_0_0_sh_mask.h"
+#include "amdgpu_amdkfd.h"
+
+static inline struct v12_compute_mqd *get_mqd(void *mqd)
+{
+	return (struct v12_compute_mqd *)mqd;
+}
+
+static inline struct v12_sdma_mqd *get_sdma_mqd(void *mqd)
+{
+	return (struct v12_sdma_mqd *)mqd;
+}
+
+static void update_cu_mask(struct mqd_manager *mm, void *mqd,
+			   struct mqd_update_info *minfo)
+{
+	struct v12_compute_mqd *m;
+	uint32_t se_mask[KFD_MAX_NUM_SE] = {0};
+
+	if (!minfo || !minfo->cu_mask.ptr)
+		return;
+
+	mqd_symmetrically_map_cu_mask(mm,
+		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask, 0);
+
+	m = get_mqd(mqd);
+	m->compute_static_thread_mgmt_se0 = se_mask[0];
+	m->compute_static_thread_mgmt_se1 = se_mask[1];
+	m->compute_static_thread_mgmt_se2 = se_mask[2];
+	m->compute_static_thread_mgmt_se3 = se_mask[3];
+	m->compute_static_thread_mgmt_se4 = se_mask[4];
+	m->compute_static_thread_mgmt_se5 = se_mask[5];
+	m->compute_static_thread_mgmt_se6 = se_mask[6];
+	m->compute_static_thread_mgmt_se7 = se_mask[7];
+
+	pr_debug("update cu mask to %#x %#x %#x %#x %#x %#x %#x %#x\n",
+		m->compute_static_thread_mgmt_se0,
+		m->compute_static_thread_mgmt_se1,
+		m->compute_static_thread_mgmt_se2,
+		m->compute_static_thread_mgmt_se3,
+		m->compute_static_thread_mgmt_se4,
+		m->compute_static_thread_mgmt_se5,
+		m->compute_static_thread_mgmt_se6,
+		m->compute_static_thread_mgmt_se7);
+}
+
+static void set_priority(struct v12_compute_mqd *m, struct queue_properties *q)
+{
+	m->cp_hqd_pipe_priority = pipe_priority_map[q->priority];
+	m->cp_hqd_queue_priority = q->priority;
+}
+
+static struct kfd_mem_obj *allocate_mqd(struct kfd_node *node,
+		struct queue_properties *q)
+{
+	struct kfd_mem_obj *mqd_mem_obj;
+
+	/*
+	 * Allocate one PAGE_SIZE memory for MQD as MES writes to areas beyond
+	 * struct MQD size.
+	 */
+	if (kfd_gtt_sa_allocate(node, PAGE_SIZE, &mqd_mem_obj))
+		return NULL;
+
+	return mqd_mem_obj;
+}
+
+static void init_mqd(struct mqd_manager *mm, void **mqd,
+			struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr,
+			struct queue_properties *q)
+{
+	uint64_t addr;
+	struct v12_compute_mqd *m;
+
+	m = (struct v12_compute_mqd *) mqd_mem_obj->cpu_ptr;
+	addr = mqd_mem_obj->gpu_addr;
+
+	memset(m, 0, PAGE_SIZE);
+
+	m->header = 0xC0310800;
+	m->compute_pipelinestat_enable = 1;
+	m->compute_static_thread_mgmt_se0 = 0xFFFFFFFF;
+	m->compute_static_thread_mgmt_se1 = 0xFFFFFFFF;
+	m->compute_static_thread_mgmt_se2 = 0xFFFFFFFF;
+	m->compute_static_thread_mgmt_se3 = 0xFFFFFFFF;
+	m->compute_static_thread_mgmt_se4 = 0xFFFFFFFF;
+	m->compute_static_thread_mgmt_se5 = 0xFFFFFFFF;
+	m->compute_static_thread_mgmt_se6 = 0xFFFFFFFF;
+	m->compute_static_thread_mgmt_se7 = 0xFFFFFFFF;
+
+	m->cp_hqd_persistent_state = CP_HQD_PERSISTENT_STATE__PRELOAD_REQ_MASK |
+			0x55 << CP_HQD_PERSISTENT_STATE__PRELOAD_SIZE__SHIFT;
+
+	m->cp_hqd_pq_control = 5 << CP_HQD_PQ_CONTROL__RPTR_BLOCK_SIZE__SHIFT;
+	m->cp_hqd_pq_control |= CP_HQD_PQ_CONTROL__UNORD_DISPATCH_MASK;
+	m->cp_mqd_control = 1 << CP_MQD_CONTROL__PRIV_STATE__SHIFT;
+
+	m->cp_mqd_base_addr_lo        = lower_32_bits(addr);
+	m->cp_mqd_base_addr_hi        = upper_32_bits(addr);
+
+	m->cp_hqd_quantum = 1 << CP_HQD_QUANTUM__QUANTUM_EN__SHIFT |
+			1 << CP_HQD_QUANTUM__QUANTUM_SCALE__SHIFT |
+			1 << CP_HQD_QUANTUM__QUANTUM_DURATION__SHIFT;
+
+	/* Set cp_hqd_hq_status0.c_queue_debug_en to 1 to have the CP set up the
+	 * DISPATCH_PTR.  This is required for the kfd debugger
+	 */
+	m->cp_hqd_hq_status0 = 1 << 14;
+
+	if (amdgpu_amdkfd_have_atomics_support(mm->dev->adev))
+		m->cp_hqd_hq_status0 |= 1 << 29;
+
+	if (q->format == KFD_QUEUE_FORMAT_AQL) {
+		m->cp_hqd_aql_control =
+			1 << CP_HQD_AQL_CONTROL__CONTROL0__SHIFT;
+	}
+
+	if (mm->dev->kfd->cwsr_enabled) {
+		m->cp_hqd_persistent_state |=
+			(1 << CP_HQD_PERSISTENT_STATE__QSWITCH_MODE__SHIFT);
+		m->cp_hqd_ctx_save_base_addr_lo =
+			lower_32_bits(q->ctx_save_restore_area_address);
+		m->cp_hqd_ctx_save_base_addr_hi =
+			upper_32_bits(q->ctx_save_restore_area_address);
+		m->cp_hqd_ctx_save_size = q->ctx_save_restore_area_size;
+		m->cp_hqd_cntl_stack_size = q->ctl_stack_size;
+		m->cp_hqd_cntl_stack_offset = q->ctl_stack_size;
+		m->cp_hqd_wg_state_offset = q->ctl_stack_size;
+	}
+
+	*mqd = m;
+	if (gart_addr)
+		*gart_addr = addr;
+	mm->update_mqd(mm, m, q, NULL);
+}
+
+static int load_mqd(struct mqd_manager *mm, void *mqd,
+			uint32_t pipe_id, uint32_t queue_id,
+			struct queue_properties *p, struct mm_struct *mms)
+{
+	int r = 0;
+	/* AQL write pointer counts in 64B packets, PM4/CP counts in dwords. */
+	uint32_t wptr_shift = (p->format == KFD_QUEUE_FORMAT_AQL ? 4 : 0);
+
+	r = mm->dev->kfd2kgd->hqd_load(mm->dev->adev, mqd, pipe_id, queue_id,
+					  (uint32_t __user *)p->write_ptr,
+					  wptr_shift, 0, mms, 0);
+	return r;
+}
+
+static void update_mqd(struct mqd_manager *mm, void *mqd,
+		       struct queue_properties *q,
+		       struct mqd_update_info *minfo)
+{
+	struct v12_compute_mqd *m;
+
+	m = get_mqd(mqd);
+
+	m->cp_hqd_pq_control &= ~CP_HQD_PQ_CONTROL__QUEUE_SIZE_MASK;
+	m->cp_hqd_pq_control |=
+			ffs(q->queue_size / sizeof(unsigned int)) - 1 - 1;
+	pr_debug("cp_hqd_pq_control 0x%x\n", m->cp_hqd_pq_control);
+
+	m->cp_hqd_pq_base_lo = lower_32_bits((uint64_t)q->queue_address >> 8);
+	m->cp_hqd_pq_base_hi = upper_32_bits((uint64_t)q->queue_address >> 8);
+
+	m->cp_hqd_pq_rptr_report_addr_lo = lower_32_bits((uint64_t)q->read_ptr);
+	m->cp_hqd_pq_rptr_report_addr_hi = upper_32_bits((uint64_t)q->read_ptr);
+	m->cp_hqd_pq_wptr_poll_addr_lo = lower_32_bits((uint64_t)q->write_ptr);
+	m->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits((uint64_t)q->write_ptr);
+
+	m->cp_hqd_pq_doorbell_control =
+		q->doorbell_off <<
+			CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET__SHIFT;
+	pr_debug("cp_hqd_pq_doorbell_control 0x%x\n",
+			m->cp_hqd_pq_doorbell_control);
+
+	m->cp_hqd_ib_control = 3 << CP_HQD_IB_CONTROL__MIN_IB_AVAIL_SIZE__SHIFT;
+
+	/*
+	 * HW does not clamp this field correctly. Maximum EOP queue size
+	 * is constrained by per-SE EOP done signal count, which is 8-bit.
+	 * Limit is 0xFF EOP entries (= 0x7F8 dwords). CP will not submit
+	 * more than (EOP entry count - 1) so a queue size of 0x800 dwords
+	 * is safe, giving a maximum field value of 0xA.
+	 */
+	m->cp_hqd_eop_control = min(0xA,
+		ffs(q->eop_ring_buffer_size / sizeof(unsigned int)) - 1 - 1);
+	m->cp_hqd_eop_base_addr_lo =
+			lower_32_bits(q->eop_ring_buffer_address >> 8);
+	m->cp_hqd_eop_base_addr_hi =
+			upper_32_bits(q->eop_ring_buffer_address >> 8);
+
+	m->cp_hqd_iq_timer = 0;
+
+	m->cp_hqd_vmid = q->vmid;
+
+	if (q->format == KFD_QUEUE_FORMAT_AQL) {
+		/* GC 10 removed WPP_CLAMP from PQ Control */
+		m->cp_hqd_pq_control |= CP_HQD_PQ_CONTROL__NO_UPDATE_RPTR_MASK |
+				2 << CP_HQD_PQ_CONTROL__SLOT_BASED_WPTR__SHIFT |
+				1 << CP_HQD_PQ_CONTROL__QUEUE_FULL_EN__SHIFT;
+		m->cp_hqd_pq_doorbell_control |=
+			1 << CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_BIF_DROP__SHIFT;
+	}
+	if (mm->dev->kfd->cwsr_enabled)
+		m->cp_hqd_ctx_save_control = 0;
+
+	update_cu_mask(mm, mqd, minfo);
+	set_priority(m, q);
+
+	q->is_active = QUEUE_IS_ACTIVE(*q);
+}
+
+static bool check_preemption_failed(struct mqd_manager *mm, void *mqd)
+{
+	struct v12_compute_mqd *m = (struct v12_compute_mqd *)mqd;
+
+	return kfd_check_hiq_mqd_doorbell_id(mm->dev, m->queue_doorbell_id0, 0);
+}
+
+static int get_wave_state(struct mqd_manager *mm, void *mqd,
+			  struct queue_properties *q,
+			  void __user *ctl_stack,
+			  u32 *ctl_stack_used_size,
+			  u32 *save_area_used_size)
+{
+	struct v12_compute_mqd *m;
+	struct mqd_user_context_save_area_header header;
+
+	m = get_mqd(mqd);
+
+	/* Control stack is written backwards, while workgroup context data
+	 * is written forwards. Both starts from m->cp_hqd_cntl_stack_size.
+	 * Current position is at m->cp_hqd_cntl_stack_offset and
+	 * m->cp_hqd_wg_state_offset, respectively.
+	 */
+	*ctl_stack_used_size = m->cp_hqd_cntl_stack_size -
+		m->cp_hqd_cntl_stack_offset;
+	*save_area_used_size = m->cp_hqd_wg_state_offset -
+		m->cp_hqd_cntl_stack_size;
+
+	/* Control stack is not copied to user mode for GFXv12 because
+	 * it's part of the context save area that is already
+	 * accessible to user mode
+	 */
+	header.control_stack_size = *ctl_stack_used_size;
+	header.wave_state_size = *save_area_used_size;
+
+	header.wave_state_offset = m->cp_hqd_wg_state_offset;
+	header.control_stack_offset = m->cp_hqd_cntl_stack_offset;
+
+	if (copy_to_user(ctl_stack, &header, sizeof(header)))
+		return -EFAULT;
+
+	return 0;
+}
+
+static void init_mqd_hiq(struct mqd_manager *mm, void **mqd,
+			struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr,
+			struct queue_properties *q)
+{
+	struct v12_compute_mqd *m;
+
+	init_mqd(mm, mqd, mqd_mem_obj, gart_addr, q);
+
+	m = get_mqd(*mqd);
+
+	m->cp_hqd_pq_control |= 1 << CP_HQD_PQ_CONTROL__PRIV_STATE__SHIFT |
+			1 << CP_HQD_PQ_CONTROL__KMD_QUEUE__SHIFT;
+}
+
+static void init_mqd_sdma(struct mqd_manager *mm, void **mqd,
+		struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr,
+		struct queue_properties *q)
+{
+	struct v12_sdma_mqd *m;
+
+	m = (struct v12_sdma_mqd *) mqd_mem_obj->cpu_ptr;
+
+	memset(m, 0, sizeof(struct v12_sdma_mqd));
+
+	*mqd = m;
+	if (gart_addr)
+		*gart_addr = mqd_mem_obj->gpu_addr;
+
+	mm->update_mqd(mm, m, q, NULL);
+}
+
+#define SDMA_RLC_DUMMY_DEFAULT 0xf
+
+static void update_mqd_sdma(struct mqd_manager *mm, void *mqd,
+		struct queue_properties *q,
+		struct mqd_update_info *minfo)
+{
+	struct v12_sdma_mqd *m;
+
+	m = get_sdma_mqd(mqd);
+	m->sdmax_rlcx_rb_cntl = (ffs(q->queue_size / sizeof(unsigned int)) - 1)
+		<< SDMA0_QUEUE0_RB_CNTL__RB_SIZE__SHIFT |
+		q->vmid << SDMA0_QUEUE0_RB_CNTL__RB_VMID__SHIFT |
+		1 << SDMA0_QUEUE0_RB_CNTL__RPTR_WRITEBACK_ENABLE__SHIFT |
+		6 << SDMA0_QUEUE0_RB_CNTL__RPTR_WRITEBACK_TIMER__SHIFT |
+		1 << SDMA0_QUEUE0_RB_CNTL__MCU_WPTR_POLL_ENABLE__SHIFT;
+
+	m->sdmax_rlcx_rb_base = lower_32_bits(q->queue_address >> 8);
+	m->sdmax_rlcx_rb_base_hi = upper_32_bits(q->queue_address >> 8);
+	m->sdmax_rlcx_rb_rptr_addr_lo = lower_32_bits((uint64_t)q->read_ptr);
+	m->sdmax_rlcx_rb_rptr_addr_hi = upper_32_bits((uint64_t)q->read_ptr);
+	m->sdmax_rlcx_rb_wptr_poll_addr_lo = lower_32_bits((uint64_t)q->write_ptr);
+	m->sdmax_rlcx_rb_wptr_poll_addr_hi = upper_32_bits((uint64_t)q->write_ptr);
+	m->sdmax_rlcx_doorbell_offset =
+		q->doorbell_off << SDMA0_QUEUE0_DOORBELL_OFFSET__OFFSET__SHIFT;
+
+	m->sdmax_rlcx_sched_cntl = (amdgpu_sdma_phase_quantum
+		<< SDMA0_QUEUE0_SCHEDULE_CNTL__CONTEXT_QUANTUM__SHIFT)
+		 & SDMA0_QUEUE0_SCHEDULE_CNTL__CONTEXT_QUANTUM_MASK;
+
+	m->sdma_engine_id = q->sdma_engine_id;
+	m->sdma_queue_id = q->sdma_queue_id;
+
+	m->sdmax_rlcx_dummy_reg = SDMA_RLC_DUMMY_DEFAULT;
+
+	q->is_active = QUEUE_IS_ACTIVE(*q);
+}
+
+#if defined(CONFIG_DEBUG_FS)
+
+static int debugfs_show_mqd(struct seq_file *m, void *data)
+{
+	seq_hex_dump(m, "    ", DUMP_PREFIX_OFFSET, 32, 4,
+		     data, sizeof(struct v12_compute_mqd), false);
+	return 0;
+}
+
+static int debugfs_show_mqd_sdma(struct seq_file *m, void *data)
+{
+	seq_hex_dump(m, "    ", DUMP_PREFIX_OFFSET, 32, 4,
+		     data, sizeof(struct v12_sdma_mqd), false);
+	return 0;
+}
+
+#endif
+
+struct mqd_manager *mqd_manager_init_v12(enum KFD_MQD_TYPE type,
+		struct kfd_node *dev)
+{
+	struct mqd_manager *mqd;
+
+	if (WARN_ON(type >= KFD_MQD_TYPE_MAX))
+		return NULL;
+
+	mqd = kzalloc(sizeof(*mqd), GFP_KERNEL);
+	if (!mqd)
+		return NULL;
+
+	mqd->dev = dev;
+
+	switch (type) {
+	case KFD_MQD_TYPE_CP:
+		pr_debug("%s@%i\n", __func__, __LINE__);
+		mqd->allocate_mqd = allocate_mqd;
+		mqd->init_mqd = init_mqd;
+		mqd->free_mqd = kfd_free_mqd_cp;
+		mqd->load_mqd = load_mqd;
+		mqd->update_mqd = update_mqd;
+		mqd->destroy_mqd = kfd_destroy_mqd_cp;
+		mqd->is_occupied = kfd_is_occupied_cp;
+		mqd->mqd_size = sizeof(struct v12_compute_mqd);
+		mqd->get_wave_state = get_wave_state;
+		mqd->mqd_stride = kfd_mqd_stride;
+#if defined(CONFIG_DEBUG_FS)
+		mqd->debugfs_show_mqd = debugfs_show_mqd;
+#endif
+		pr_debug("%s@%i\n", __func__, __LINE__);
+		break;
+	case KFD_MQD_TYPE_HIQ:
+		pr_debug("%s@%i\n", __func__, __LINE__);
+		mqd->allocate_mqd = allocate_hiq_mqd;
+		mqd->init_mqd = init_mqd_hiq;
+		mqd->free_mqd = free_mqd_hiq_sdma;
+		mqd->load_mqd = kfd_hiq_load_mqd_kiq;
+		mqd->update_mqd = update_mqd;
+		mqd->destroy_mqd = kfd_destroy_mqd_cp;
+		mqd->is_occupied = kfd_is_occupied_cp;
+		mqd->mqd_size = sizeof(struct v12_compute_mqd);
+		mqd->mqd_stride = kfd_mqd_stride;
+#if defined(CONFIG_DEBUG_FS)
+		mqd->debugfs_show_mqd = debugfs_show_mqd;
+#endif
+		mqd->check_preemption_failed = check_preemption_failed;
+		pr_debug("%s@%i\n", __func__, __LINE__);
+		break;
+	case KFD_MQD_TYPE_DIQ:
+		mqd->allocate_mqd = allocate_mqd;
+		mqd->init_mqd = init_mqd_hiq;
+		mqd->free_mqd = kfd_free_mqd_cp;
+		mqd->load_mqd = load_mqd;
+		mqd->update_mqd = update_mqd;
+		mqd->destroy_mqd = kfd_destroy_mqd_cp;
+		mqd->is_occupied = kfd_is_occupied_cp;
+		mqd->mqd_size = sizeof(struct v12_compute_mqd);
+#if defined(CONFIG_DEBUG_FS)
+		mqd->debugfs_show_mqd = debugfs_show_mqd;
+#endif
+		break;
+	case KFD_MQD_TYPE_SDMA:
+		pr_debug("%s@%i\n", __func__, __LINE__);
+		mqd->allocate_mqd = allocate_mqd;
+		mqd->init_mqd = init_mqd_sdma;
+		mqd->free_mqd = kfd_free_mqd_cp;
+		mqd->load_mqd = kfd_load_mqd_sdma;
+		mqd->update_mqd = update_mqd_sdma;
+		mqd->destroy_mqd = kfd_destroy_mqd_sdma;
+		mqd->is_occupied = kfd_is_occupied_sdma;
+		mqd->mqd_size = sizeof(struct v12_sdma_mqd);
+		mqd->mqd_stride = kfd_mqd_stride;
+#if defined(CONFIG_DEBUG_FS)
+		mqd->debugfs_show_mqd = debugfs_show_mqd_sdma;
+#endif
+		pr_debug("%s@%i\n", __func__, __LINE__);
+		break;
+	default:
+		kfree(mqd);
+		return NULL;
+	}
+
+	return mqd;
+}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
index fdbfd725841f..f2dee320fada 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
@@ -32,6 +32,22 @@
 #include "gc/gc_9_0_sh_mask.h"
 #include "sdma0/sdma0_4_0_sh_mask.h"
 #include "amdgpu_amdkfd.h"
+#include "kfd_device_queue_manager.h"
+
+static void update_mqd(struct mqd_manager *mm, void *mqd,
+		       struct queue_properties *q,
+		       struct mqd_update_info *minfo);
+
+static uint64_t mqd_stride_v9(struct mqd_manager *mm,
+				struct queue_properties *q)
+{
+	if (mm->dev->kfd->cwsr_enabled &&
+	    q->type == KFD_QUEUE_TYPE_COMPUTE)
+		return ALIGN(q->ctl_stack_size, PAGE_SIZE) +
+			ALIGN(sizeof(struct v9_mqd), PAGE_SIZE);
+
+	return mm->mqd_size;
+}
 
 static inline struct v9_mqd *get_mqd(void *mqd)
 {
@@ -44,37 +60,47 @@ static inline struct v9_sdma_mqd *get_sdma_mqd(void *mqd)
 }
 
 static void update_cu_mask(struct mqd_manager *mm, void *mqd,
-			struct mqd_update_info *minfo)
+			struct mqd_update_info *minfo, uint32_t inst)
 {
 	struct v9_mqd *m;
 	uint32_t se_mask[KFD_MAX_NUM_SE] = {0};
 
-	if (!minfo || (minfo->update_flag != UPDATE_FLAG_CU_MASK) ||
-	    !minfo->cu_mask.ptr)
+	if (!minfo || !minfo->cu_mask.ptr)
 		return;
 
 	mqd_symmetrically_map_cu_mask(mm,
-		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask);
+		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask, inst);
 
 	m = get_mqd(mqd);
+
 	m->compute_static_thread_mgmt_se0 = se_mask[0];
 	m->compute_static_thread_mgmt_se1 = se_mask[1];
 	m->compute_static_thread_mgmt_se2 = se_mask[2];
 	m->compute_static_thread_mgmt_se3 = se_mask[3];
-	m->compute_static_thread_mgmt_se4 = se_mask[4];
-	m->compute_static_thread_mgmt_se5 = se_mask[5];
-	m->compute_static_thread_mgmt_se6 = se_mask[6];
-	m->compute_static_thread_mgmt_se7 = se_mask[7];
-
-	pr_debug("update cu mask to %#x %#x %#x %#x %#x %#x %#x %#x\n",
-		m->compute_static_thread_mgmt_se0,
-		m->compute_static_thread_mgmt_se1,
-		m->compute_static_thread_mgmt_se2,
-		m->compute_static_thread_mgmt_se3,
-		m->compute_static_thread_mgmt_se4,
-		m->compute_static_thread_mgmt_se5,
-		m->compute_static_thread_mgmt_se6,
-		m->compute_static_thread_mgmt_se7);
+	if (KFD_GC_VERSION(mm->dev) != IP_VERSION(9, 4, 3) &&
+	    KFD_GC_VERSION(mm->dev) != IP_VERSION(9, 4, 4) &&
+	    KFD_GC_VERSION(mm->dev) != IP_VERSION(9, 5, 0)) {
+		m->compute_static_thread_mgmt_se4 = se_mask[4];
+		m->compute_static_thread_mgmt_se5 = se_mask[5];
+		m->compute_static_thread_mgmt_se6 = se_mask[6];
+		m->compute_static_thread_mgmt_se7 = se_mask[7];
+
+		pr_debug("update cu mask to %#x %#x %#x %#x %#x %#x %#x %#x\n",
+			m->compute_static_thread_mgmt_se0,
+			m->compute_static_thread_mgmt_se1,
+			m->compute_static_thread_mgmt_se2,
+			m->compute_static_thread_mgmt_se3,
+			m->compute_static_thread_mgmt_se4,
+			m->compute_static_thread_mgmt_se5,
+			m->compute_static_thread_mgmt_se6,
+			m->compute_static_thread_mgmt_se7);
+	} else {
+		pr_debug("inst: %u, update cu mask to %#x %#x %#x %#x\n",
+			inst, m->compute_static_thread_mgmt_se0,
+			m->compute_static_thread_mgmt_se1,
+			m->compute_static_thread_mgmt_se2,
+			m->compute_static_thread_mgmt_se3);
+	}
 }
 
 static void set_priority(struct v9_mqd *m, struct queue_properties *q)
@@ -83,7 +109,7 @@ static void set_priority(struct v9_mqd *m, struct queue_properties *q)
 	m->cp_hqd_queue_priority = q->priority;
 }
 
-static struct kfd_mem_obj *allocate_mqd(struct kfd_dev *kfd,
+static struct kfd_mem_obj *allocate_mqd(struct kfd_node *node,
 		struct queue_properties *q)
 {
 	int retval;
@@ -105,28 +131,30 @@ static struct kfd_mem_obj *allocate_mqd(struct kfd_dev *kfd,
 	 * pass a special bo flag AMDGPU_GEM_CREATE_CP_MQD_GFX9 to instruct
 	 * amdgpu memory functions to do so.
 	 */
-	if (kfd->cwsr_enabled && (q->type == KFD_QUEUE_TYPE_COMPUTE)) {
+	if (node->kfd->cwsr_enabled && (q->type == KFD_QUEUE_TYPE_COMPUTE)) {
 		mqd_mem_obj = kzalloc(sizeof(struct kfd_mem_obj), GFP_KERNEL);
 		if (!mqd_mem_obj)
 			return NULL;
-		retval = amdgpu_amdkfd_alloc_gtt_mem(kfd->adev,
-			ALIGN(q->ctl_stack_size, PAGE_SIZE) +
-				ALIGN(sizeof(struct v9_mqd), PAGE_SIZE),
+		retval = amdgpu_amdkfd_alloc_gtt_mem(node->adev,
+			(ALIGN(q->ctl_stack_size, PAGE_SIZE) +
+			ALIGN(sizeof(struct v9_mqd), PAGE_SIZE)) *
+			NUM_XCC(node->xcc_mask),
 			&(mqd_mem_obj->gtt_mem),
 			&(mqd_mem_obj->gpu_addr),
 			(void *)&(mqd_mem_obj->cpu_ptr), true);
+
+		if (retval) {
+			kfree(mqd_mem_obj);
+			return NULL;
+		}
 	} else {
-		retval = kfd_gtt_sa_allocate(kfd, sizeof(struct v9_mqd),
+		retval = kfd_gtt_sa_allocate(node, sizeof(struct v9_mqd),
 				&mqd_mem_obj);
-	}
-
-	if (retval) {
-		kfree(mqd_mem_obj);
-		return NULL;
+		if (retval)
+			return NULL;
 	}
 
 	return mqd_mem_obj;
-
 }
 
 static void init_mqd(struct mqd_manager *mm, void **mqd,
@@ -135,7 +163,6 @@ static void init_mqd(struct mqd_manager *mm, void **mqd,
 {
 	uint64_t addr;
 	struct v9_mqd *m;
-	struct amdgpu_device *adev = (struct amdgpu_device *)mm->dev->adev;
 
 	m = (struct v9_mqd *) mqd_mem_obj->cpu_ptr;
 	addr = mqd_mem_obj->gpu_addr;
@@ -156,6 +183,9 @@ static void init_mqd(struct mqd_manager *mm, void **mqd,
 	m->cp_hqd_persistent_state = CP_HQD_PERSISTENT_STATE__PRELOAD_REQ_MASK |
 			0x53 << CP_HQD_PERSISTENT_STATE__PRELOAD_SIZE__SHIFT;
 
+	m->cp_hqd_pq_control = 5 << CP_HQD_PQ_CONTROL__RPTR_BLOCK_SIZE__SHIFT;
+	m->cp_hqd_pq_control |= CP_HQD_PQ_CONTROL__UNORD_DISPATCH_MASK;
+
 	m->cp_mqd_control = 1 << CP_MQD_CONTROL__PRIV_STATE__SHIFT;
 
 	m->cp_mqd_base_addr_lo        = lower_32_bits(addr);
@@ -165,31 +195,21 @@ static void init_mqd(struct mqd_manager *mm, void **mqd,
 			1 << CP_HQD_QUANTUM__QUANTUM_SCALE__SHIFT |
 			1 << CP_HQD_QUANTUM__QUANTUM_DURATION__SHIFT;
 
-	if (q->format == KFD_QUEUE_FORMAT_AQL) {
+	/* Set cp_hqd_hq_scheduler0 bit 14 to 1 to have the CP set up the
+	 * DISPATCH_PTR.  This is required for the kfd debugger
+	 */
+	m->cp_hqd_hq_status0 = 1 << 14;
+
+	if (q->format == KFD_QUEUE_FORMAT_AQL)
 		m->cp_hqd_aql_control =
 			1 << CP_HQD_AQL_CONTROL__CONTROL0__SHIFT;
-		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3)) {
-			/* On GC 9.4.3, DW 41 is re-purposed as
-			 * compute_tg_chunk_size.
-			 * TODO: review this setting when active CUs in the
-			 * partition play a role
-			 */
-			m->compute_static_thread_mgmt_se6 = 1;
-		}
-	} else {
-		/* PM4 queue */
-		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3)) {
-			m->compute_static_thread_mgmt_se6 = 0;
-			/* TODO: program pm4_target_xcc */
-		}
-	}
 
 	if (q->tba_addr) {
 		m->compute_pgm_rsrc2 |=
 			(1 << COMPUTE_PGM_RSRC2__TRAP_PRESENT__SHIFT);
 	}
 
-	if (mm->dev->cwsr_enabled && q->ctx_save_restore_area_address) {
+	if (mm->dev->kfd->cwsr_enabled && q->ctx_save_restore_area_address) {
 		m->cp_hqd_persistent_state |=
 			(1 << CP_HQD_PERSISTENT_STATE__QSWITCH_MODE__SHIFT);
 		m->cp_hqd_ctx_save_base_addr_lo =
@@ -205,7 +225,7 @@ static void init_mqd(struct mqd_manager *mm, void **mqd,
 	*mqd = m;
 	if (gart_addr)
 		*gart_addr = addr;
-	mm->update_mqd(mm, m, q, NULL);
+	update_mqd(mm, m, q, NULL);
 }
 
 static int load_mqd(struct mqd_manager *mm, void *mqd,
@@ -217,19 +237,18 @@ static int load_mqd(struct mqd_manager *mm, void *mqd,
 
 	return mm->dev->kfd2kgd->hqd_load(mm->dev->adev, mqd, pipe_id, queue_id,
 					  (uint32_t __user *)p->write_ptr,
-					  wptr_shift, 0, mms);
+					  wptr_shift, 0, mms, 0);
 }
 
 static void update_mqd(struct mqd_manager *mm, void *mqd,
 			struct queue_properties *q,
 			struct mqd_update_info *minfo)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)mm->dev->adev;
 	struct v9_mqd *m;
 
 	m = get_mqd(mqd);
 
-	m->cp_hqd_pq_control = 5 << CP_HQD_PQ_CONTROL__RPTR_BLOCK_SIZE__SHIFT;
+	m->cp_hqd_pq_control &= ~CP_HQD_PQ_CONTROL__QUEUE_SIZE_MASK;
 	m->cp_hqd_pq_control |= order_base_2(q->queue_size / 4) - 1;
 	pr_debug("cp_hqd_pq_control 0x%x\n", m->cp_hqd_pq_control);
 
@@ -257,9 +276,14 @@ static void update_mqd(struct mqd_manager *mm, void *mqd,
 	 * Limit is 0xFF EOP entries (= 0x7F8 dwords). CP will not submit
 	 * more than (EOP entry count - 1) so a queue size of 0x800 dwords
 	 * is safe, giving a maximum field value of 0xA.
+	 *
+	 * Also, do calculation only if EOP is used (size > 0), otherwise
+	 * the order_base_2 calculation provides incorrect result.
+	 *
 	 */
-	m->cp_hqd_eop_control = min(0xA,
-		order_base_2(q->eop_ring_buffer_size / 4) - 1);
+	m->cp_hqd_eop_control = q->eop_ring_buffer_size ?
+		min(0xA, order_base_2(q->eop_ring_buffer_size / 4) - 1) : 0;
+
 	m->cp_hqd_eop_base_addr_lo =
 			lower_32_bits(q->eop_ring_buffer_address >> 8);
 	m->cp_hqd_eop_base_addr_hi =
@@ -270,39 +294,53 @@ static void update_mqd(struct mqd_manager *mm, void *mqd,
 	m->cp_hqd_vmid = q->vmid;
 
 	if (q->format == KFD_QUEUE_FORMAT_AQL) {
-		m->cp_hqd_pq_control |=
+		m->cp_hqd_pq_control |= CP_HQD_PQ_CONTROL__NO_UPDATE_RPTR_MASK |
 				2 << CP_HQD_PQ_CONTROL__SLOT_BASED_WPTR__SHIFT |
 				1 << CP_HQD_PQ_CONTROL__QUEUE_FULL_EN__SHIFT |
 				1 << CP_HQD_PQ_CONTROL__WPP_CLAMP_EN__SHIFT;
-		if (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 3))
-			m->cp_hqd_pq_control |=
-				 CP_HQD_PQ_CONTROL__NO_UPDATE_RPTR_MASK;
 		m->cp_hqd_pq_doorbell_control |= 1 <<
 			CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_BIF_DROP__SHIFT;
 	}
-	if (mm->dev->cwsr_enabled && q->ctx_save_restore_area_address)
+	if (mm->dev->kfd->cwsr_enabled && q->ctx_save_restore_area_address)
 		m->cp_hqd_ctx_save_control = 0;
 
-	update_cu_mask(mm, mqd, minfo);
+	if (KFD_GC_VERSION(mm->dev) != IP_VERSION(9, 4, 3) &&
+	    KFD_GC_VERSION(mm->dev) != IP_VERSION(9, 4, 4) &&
+	    KFD_GC_VERSION(mm->dev) != IP_VERSION(9, 5, 0))
+		update_cu_mask(mm, mqd, minfo, 0);
 	set_priority(m, q);
 
+	if (minfo && KFD_GC_VERSION(mm->dev) >= IP_VERSION(9, 4, 2)) {
+		if (minfo->update_flag & UPDATE_FLAG_IS_GWS)
+			m->compute_resource_limits |=
+				COMPUTE_RESOURCE_LIMITS__FORCE_SIMD_DIST_MASK;
+		else
+			m->compute_resource_limits &=
+				~COMPUTE_RESOURCE_LIMITS__FORCE_SIMD_DIST_MASK;
+	}
+
 	q->is_active = QUEUE_IS_ACTIVE(*q);
 }
 
 
-static uint32_t read_doorbell_id(void *mqd)
+static bool check_preemption_failed(struct mqd_manager *mm, void *mqd)
 {
 	struct v9_mqd *m = (struct v9_mqd *)mqd;
+	uint32_t doorbell_id = m->queue_doorbell_id0;
+
+	m->queue_doorbell_id0 = 0;
 
-	return m->queue_doorbell_id0;
+	return kfd_check_hiq_mqd_doorbell_id(mm->dev, doorbell_id, 0);
 }
 
 static int get_wave_state(struct mqd_manager *mm, void *mqd,
+			  struct queue_properties *q,
 			  void __user *ctl_stack,
 			  u32 *ctl_stack_used_size,
 			  u32 *save_area_used_size)
 {
 	struct v9_mqd *m;
+	struct kfd_context_save_area_header header;
 
 	/* Control stack is located one page after MQD. */
 	void *mqd_ctl_stack = (void *)((uintptr_t)mqd + PAGE_SIZE);
@@ -314,7 +352,18 @@ static int get_wave_state(struct mqd_manager *mm, void *mqd,
 	*save_area_used_size = m->cp_hqd_wg_state_offset -
 		m->cp_hqd_cntl_stack_size;
 
-	if (copy_to_user(ctl_stack, mqd_ctl_stack, m->cp_hqd_cntl_stack_size))
+	header.wave_state.control_stack_size = *ctl_stack_used_size;
+	header.wave_state.wave_state_size = *save_area_used_size;
+
+	header.wave_state.wave_state_offset = m->cp_hqd_wg_state_offset;
+	header.wave_state.control_stack_offset = m->cp_hqd_cntl_stack_offset;
+
+	if (copy_to_user(ctl_stack, &header, sizeof(header.wave_state)))
+		return -EFAULT;
+
+	if (copy_to_user(ctl_stack + m->cp_hqd_cntl_stack_offset,
+				mqd_ctl_stack + m->cp_hqd_cntl_stack_offset,
+				*ctl_stack_used_size))
 		return -EFAULT;
 
 	return 0;
@@ -324,7 +373,7 @@ static void get_checkpoint_info(struct mqd_manager *mm, void *mqd, u32 *ctl_stac
 {
 	struct v9_mqd *m = get_mqd(mqd);
 
-	*ctl_stack_size = m->cp_hqd_cntl_stack_size;
+	*ctl_stack_size = m->cp_hqd_cntl_stack_size * NUM_XCC(mm->dev->xcc_mask);
 }
 
 static void checkpoint_mqd(struct mqd_manager *mm, void *mqd, void *mqd_dst, void *ctl_stack_dst)
@@ -339,6 +388,24 @@ static void checkpoint_mqd(struct mqd_manager *mm, void *mqd, void *mqd_dst, voi
 	memcpy(ctl_stack_dst, ctl_stack, m->cp_hqd_cntl_stack_size);
 }
 
+static void checkpoint_mqd_v9_4_3(struct mqd_manager *mm,
+								  void *mqd,
+								  void *mqd_dst,
+								  void *ctl_stack_dst)
+{
+	struct v9_mqd *m;
+	int xcc;
+	uint64_t size = get_mqd(mqd)->cp_mqd_stride_size;
+
+	for (xcc = 0; xcc < NUM_XCC(mm->dev->xcc_mask); xcc++) {
+		m = get_mqd(mqd + size * xcc);
+
+		checkpoint_mqd(mm, m,
+				(uint8_t *)mqd_dst + sizeof(*m) * xcc,
+				(uint8_t *)ctl_stack_dst + m->cp_hqd_cntl_stack_size * xcc);
+	}
+}
+
 static void restore_mqd(struct mqd_manager *mm, void **mqd,
 			struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr,
 			struct queue_properties *qp,
@@ -385,6 +452,25 @@ static void init_mqd_hiq(struct mqd_manager *mm, void **mqd,
 			1 << CP_HQD_PQ_CONTROL__KMD_QUEUE__SHIFT;
 }
 
+static int destroy_hiq_mqd(struct mqd_manager *mm, void *mqd,
+			enum kfd_preempt_type type, unsigned int timeout,
+			uint32_t pipe_id, uint32_t queue_id)
+{
+	int err;
+	struct v9_mqd *m;
+	u32 doorbell_off;
+
+	m = get_mqd(mqd);
+
+	doorbell_off = m->cp_hqd_pq_doorbell_control >>
+			CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET__SHIFT;
+	err = amdgpu_amdkfd_unmap_hiq(mm->dev->adev, doorbell_off, 0);
+	if (err)
+		pr_debug("Destroy HIQ MQD failed: %d\n", err);
+
+	return err;
+}
+
 static void init_mqd_sdma(struct mqd_manager *mm, void **mqd,
 		struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr,
 		struct queue_properties *q)
@@ -427,6 +513,10 @@ static void update_mqd_sdma(struct mqd_manager *mm, void *mqd,
 	m->sdma_engine_id = q->sdma_engine_id;
 	m->sdma_queue_id = q->sdma_queue_id;
 	m->sdmax_rlcx_dummy_reg = SDMA_RLC_DUMMY_DEFAULT;
+	/* Allow context switch so we don't cross-process starve with a massive
+	 * command buffer of long-running SDMA commands
+	 */
+	m->sdmax_rlcx_ib_cntl |= SDMA0_GFX_IB_CNTL__SWITCH_INSIDE_IB_MASK;
 
 	q->is_active = QUEUE_IS_ACTIVE(*q);
 }
@@ -467,6 +557,357 @@ static void restore_mqd_sdma(struct mqd_manager *mm, void **mqd,
 	qp->is_active = 0;
 }
 
+static void init_mqd_hiq_v9_4_3(struct mqd_manager *mm, void **mqd,
+			struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr,
+			struct queue_properties *q)
+{
+	struct v9_mqd *m;
+	int xcc = 0;
+	struct kfd_mem_obj xcc_mqd_mem_obj;
+	uint64_t xcc_gart_addr = 0;
+
+	memset(&xcc_mqd_mem_obj, 0x0, sizeof(struct kfd_mem_obj));
+
+	for (xcc = 0; xcc < NUM_XCC(mm->dev->xcc_mask); xcc++) {
+		kfd_get_hiq_xcc_mqd(mm->dev, &xcc_mqd_mem_obj, xcc);
+
+		init_mqd(mm, (void **)&m, &xcc_mqd_mem_obj, &xcc_gart_addr, q);
+
+		m->cp_hqd_pq_control |= CP_HQD_PQ_CONTROL__NO_UPDATE_RPTR_MASK |
+					1 << CP_HQD_PQ_CONTROL__PRIV_STATE__SHIFT |
+					1 << CP_HQD_PQ_CONTROL__KMD_QUEUE__SHIFT;
+		if (amdgpu_sriov_multi_vf_mode(mm->dev->adev))
+			m->cp_hqd_pq_doorbell_control |= 1 <<
+				CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_MODE__SHIFT;
+		m->cp_mqd_stride_size = kfd_hiq_mqd_stride(mm->dev);
+		if (xcc == 0) {
+			/* Set no_update_rptr = 0 in Master XCC */
+			m->cp_hqd_pq_control &= ~CP_HQD_PQ_CONTROL__NO_UPDATE_RPTR_MASK;
+
+			/* Set the MQD pointer and gart address to XCC0 MQD */
+			*mqd = m;
+			*gart_addr = xcc_gart_addr;
+		}
+	}
+}
+
+static int hiq_load_mqd_kiq_v9_4_3(struct mqd_manager *mm, void *mqd,
+			uint32_t pipe_id, uint32_t queue_id,
+			struct queue_properties *p, struct mm_struct *mms)
+{
+	uint32_t xcc_mask = mm->dev->xcc_mask;
+	int xcc_id, err, inst = 0;
+	void *xcc_mqd;
+	uint64_t hiq_mqd_size = kfd_hiq_mqd_stride(mm->dev);
+
+	for_each_inst(xcc_id, xcc_mask) {
+		xcc_mqd = mqd + hiq_mqd_size * inst;
+		err = mm->dev->kfd2kgd->hiq_mqd_load(mm->dev->adev, xcc_mqd,
+						     pipe_id, queue_id,
+						     p->doorbell_off, xcc_id);
+		if (err) {
+			pr_debug("Failed to load HIQ MQD for XCC: %d\n", inst);
+			break;
+		}
+		++inst;
+	}
+
+	return err;
+}
+
+static int destroy_hiq_mqd_v9_4_3(struct mqd_manager *mm, void *mqd,
+			enum kfd_preempt_type type, unsigned int timeout,
+			uint32_t pipe_id, uint32_t queue_id)
+{
+	uint32_t xcc_mask = mm->dev->xcc_mask;
+	int xcc_id, err, inst = 0;
+	uint64_t hiq_mqd_size = kfd_hiq_mqd_stride(mm->dev);
+	struct v9_mqd *m;
+	u32 doorbell_off;
+
+	for_each_inst(xcc_id, xcc_mask) {
+		m = get_mqd(mqd + hiq_mqd_size * inst);
+
+		doorbell_off = m->cp_hqd_pq_doorbell_control >>
+				CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET__SHIFT;
+
+		err = amdgpu_amdkfd_unmap_hiq(mm->dev->adev, doorbell_off, xcc_id);
+		if (err) {
+			pr_debug("Destroy HIQ MQD failed for xcc: %d\n", inst);
+			break;
+		}
+		++inst;
+	}
+
+	return err;
+}
+
+static bool check_preemption_failed_v9_4_3(struct mqd_manager *mm, void *mqd)
+{
+	uint64_t hiq_mqd_size = kfd_hiq_mqd_stride(mm->dev);
+	uint32_t xcc_mask = mm->dev->xcc_mask;
+	int inst = 0, xcc_id;
+	struct v9_mqd *m;
+	bool ret = false;
+
+	for_each_inst(xcc_id, xcc_mask) {
+		m = get_mqd(mqd + hiq_mqd_size * inst);
+		ret |= kfd_check_hiq_mqd_doorbell_id(mm->dev,
+					m->queue_doorbell_id0, inst);
+		m->queue_doorbell_id0 = 0;
+		++inst;
+	}
+
+	return ret;
+}
+
+static void get_xcc_mqd(struct kfd_mem_obj *mqd_mem_obj,
+			       struct kfd_mem_obj *xcc_mqd_mem_obj,
+			       uint64_t offset)
+{
+	xcc_mqd_mem_obj->gtt_mem = (offset == 0) ?
+					mqd_mem_obj->gtt_mem : NULL;
+	xcc_mqd_mem_obj->gpu_addr = mqd_mem_obj->gpu_addr + offset;
+	xcc_mqd_mem_obj->cpu_ptr = (uint32_t *)((uintptr_t)mqd_mem_obj->cpu_ptr
+						+ offset);
+}
+
+static void init_mqd_v9_4_3(struct mqd_manager *mm, void **mqd,
+			struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr,
+			struct queue_properties *q)
+{
+	struct v9_mqd *m;
+	int xcc = 0;
+	struct kfd_mem_obj xcc_mqd_mem_obj;
+	uint64_t xcc_gart_addr = 0;
+	uint64_t xcc_ctx_save_restore_area_address;
+	uint64_t offset = mm->mqd_stride(mm, q);
+	uint32_t local_xcc_start = mm->dev->dqm->current_logical_xcc_start++;
+
+	memset(&xcc_mqd_mem_obj, 0x0, sizeof(struct kfd_mem_obj));
+	for (xcc = 0; xcc < NUM_XCC(mm->dev->xcc_mask); xcc++) {
+		get_xcc_mqd(mqd_mem_obj, &xcc_mqd_mem_obj, offset*xcc);
+
+		init_mqd(mm, (void **)&m, &xcc_mqd_mem_obj, &xcc_gart_addr, q);
+		if (amdgpu_sriov_multi_vf_mode(mm->dev->adev))
+				m->cp_hqd_pq_doorbell_control |= 1 <<
+					CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_MODE__SHIFT;
+		m->cp_mqd_stride_size = offset;
+
+		/*
+		 * Update the CWSR address for each XCC if CWSR is enabled
+		 * and CWSR area is allocated in thunk
+		 */
+		if (mm->dev->kfd->cwsr_enabled &&
+		    q->ctx_save_restore_area_address) {
+			xcc_ctx_save_restore_area_address =
+				q->ctx_save_restore_area_address +
+				(xcc * q->ctx_save_restore_area_size);
+
+			m->cp_hqd_ctx_save_base_addr_lo =
+				lower_32_bits(xcc_ctx_save_restore_area_address);
+			m->cp_hqd_ctx_save_base_addr_hi =
+				upper_32_bits(xcc_ctx_save_restore_area_address);
+		}
+
+		if (q->format == KFD_QUEUE_FORMAT_AQL) {
+			m->compute_tg_chunk_size = 1;
+			m->compute_current_logic_xcc_id =
+					(local_xcc_start + xcc) %
+					NUM_XCC(mm->dev->xcc_mask);
+
+			switch (xcc) {
+			case 0:
+				/* Master XCC */
+				m->cp_hqd_pq_control &=
+					~CP_HQD_PQ_CONTROL__NO_UPDATE_RPTR_MASK;
+				break;
+			default:
+				break;
+			}
+		} else {
+			/* PM4 Queue */
+			m->compute_current_logic_xcc_id = 0;
+			m->compute_tg_chunk_size = 0;
+			m->pm4_target_xcc_in_xcp = q->pm4_target_xcc;
+		}
+
+		if (xcc == 0) {
+			/* Set the MQD pointer and gart address to XCC0 MQD */
+			*mqd = m;
+			*gart_addr = xcc_gart_addr;
+		}
+	}
+}
+
+static void update_mqd_v9_4_3(struct mqd_manager *mm, void *mqd,
+		      struct queue_properties *q, struct mqd_update_info *minfo)
+{
+	struct v9_mqd *m;
+	int xcc = 0;
+	uint64_t size = mm->mqd_stride(mm, q);
+
+	for (xcc = 0; xcc < NUM_XCC(mm->dev->xcc_mask); xcc++) {
+		m = get_mqd(mqd + size * xcc);
+		update_mqd(mm, m, q, minfo);
+
+		if (amdgpu_sriov_multi_vf_mode(mm->dev->adev))
+				m->cp_hqd_pq_doorbell_control |= 1 <<
+					CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_MODE__SHIFT;
+		update_cu_mask(mm, m, minfo, xcc);
+
+		if (q->format == KFD_QUEUE_FORMAT_AQL) {
+			switch (xcc) {
+			case 0:
+				/* Master XCC */
+				m->cp_hqd_pq_control &=
+					~CP_HQD_PQ_CONTROL__NO_UPDATE_RPTR_MASK;
+				break;
+			default:
+				break;
+			}
+			m->compute_tg_chunk_size = 1;
+		} else {
+			/* PM4 Queue */
+			m->compute_current_logic_xcc_id = 0;
+			m->compute_tg_chunk_size = 0;
+			m->pm4_target_xcc_in_xcp = q->pm4_target_xcc;
+		}
+	}
+}
+
+static void restore_mqd_v9_4_3(struct mqd_manager *mm, void **mqd,
+			struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr,
+			struct queue_properties *qp,
+			const void *mqd_src,
+			const void *ctl_stack_src, u32 ctl_stack_size)
+{
+	struct kfd_mem_obj xcc_mqd_mem_obj;
+	u32 mqd_ctl_stack_size;
+	struct v9_mqd *m;
+	u32 num_xcc;
+	int xcc;
+
+	uint64_t offset = mm->mqd_stride(mm, qp);
+
+	mm->dev->dqm->current_logical_xcc_start++;
+
+	num_xcc = NUM_XCC(mm->dev->xcc_mask);
+	mqd_ctl_stack_size = ctl_stack_size / num_xcc;
+
+	memset(&xcc_mqd_mem_obj, 0x0, sizeof(struct kfd_mem_obj));
+
+	/* Set the MQD pointer and gart address to XCC0 MQD */
+	*mqd = mqd_mem_obj->cpu_ptr;
+	if (gart_addr)
+		*gart_addr = mqd_mem_obj->gpu_addr;
+
+	for (xcc = 0; xcc < num_xcc; xcc++) {
+		get_xcc_mqd(mqd_mem_obj, &xcc_mqd_mem_obj, offset * xcc);
+		restore_mqd(mm, (void **)&m,
+					&xcc_mqd_mem_obj,
+					NULL,
+					qp,
+					(uint8_t *)mqd_src + xcc * sizeof(*m),
+					(uint8_t *)ctl_stack_src + xcc *  mqd_ctl_stack_size,
+					mqd_ctl_stack_size);
+	}
+}
+static int destroy_mqd_v9_4_3(struct mqd_manager *mm, void *mqd,
+		   enum kfd_preempt_type type, unsigned int timeout,
+		   uint32_t pipe_id, uint32_t queue_id)
+{
+	uint32_t xcc_mask = mm->dev->xcc_mask;
+	int xcc_id, err, inst = 0;
+	void *xcc_mqd;
+	struct v9_mqd *m;
+	uint64_t mqd_offset;
+
+	m = get_mqd(mqd);
+	mqd_offset = m->cp_mqd_stride_size;
+
+	for_each_inst(xcc_id, xcc_mask) {
+		xcc_mqd = mqd + mqd_offset * inst;
+		err = mm->dev->kfd2kgd->hqd_destroy(mm->dev->adev, xcc_mqd,
+						    type, timeout, pipe_id,
+						    queue_id, xcc_id);
+		if (err) {
+			pr_debug("Destroy MQD failed for xcc: %d\n", inst);
+			break;
+		}
+		++inst;
+	}
+
+	return err;
+}
+
+static int load_mqd_v9_4_3(struct mqd_manager *mm, void *mqd,
+			uint32_t pipe_id, uint32_t queue_id,
+			struct queue_properties *p, struct mm_struct *mms)
+{
+	/* AQL write pointer counts in 64B packets, PM4/CP counts in dwords. */
+	uint32_t wptr_shift = (p->format == KFD_QUEUE_FORMAT_AQL ? 4 : 0);
+	uint32_t xcc_mask = mm->dev->xcc_mask;
+	int xcc_id, err, inst = 0;
+	void *xcc_mqd;
+	uint64_t mqd_stride_size = mm->mqd_stride(mm, p);
+
+	for_each_inst(xcc_id, xcc_mask) {
+		xcc_mqd = mqd + mqd_stride_size * inst;
+		err = mm->dev->kfd2kgd->hqd_load(
+			mm->dev->adev, xcc_mqd, pipe_id, queue_id,
+			(uint32_t __user *)p->write_ptr, wptr_shift, 0, mms,
+			xcc_id);
+		if (err) {
+			pr_debug("Load MQD failed for xcc: %d\n", inst);
+			break;
+		}
+		++inst;
+	}
+
+	return err;
+}
+
+static int get_wave_state_v9_4_3(struct mqd_manager *mm, void *mqd,
+				 struct queue_properties *q,
+				 void __user *ctl_stack,
+				 u32 *ctl_stack_used_size,
+				 u32 *save_area_used_size)
+{
+	int xcc, err = 0;
+	void *xcc_mqd;
+	void __user *xcc_ctl_stack;
+	uint64_t mqd_stride_size = mm->mqd_stride(mm, q);
+	u32 tmp_ctl_stack_used_size = 0, tmp_save_area_used_size = 0;
+
+	for (xcc = 0; xcc < NUM_XCC(mm->dev->xcc_mask); xcc++) {
+		xcc_mqd = mqd + mqd_stride_size * xcc;
+		xcc_ctl_stack = (void __user *)((uintptr_t)ctl_stack +
+					q->ctx_save_restore_area_size * xcc);
+
+		err = get_wave_state(mm, xcc_mqd, q, xcc_ctl_stack,
+				     &tmp_ctl_stack_used_size,
+				     &tmp_save_area_used_size);
+		if (err)
+			break;
+
+		/*
+		 * Set the ctl_stack_used_size and save_area_used_size to
+		 * ctl_stack_used_size and save_area_used_size of XCC 0 when
+		 * passing the info the user-space.
+		 * For multi XCC, user-space would have to look at the header
+		 * info of each Control stack area to determine the control
+		 * stack size and save area used.
+		 */
+		if (xcc == 0) {
+			*ctl_stack_used_size = tmp_ctl_stack_used_size;
+			*save_area_used_size = tmp_save_area_used_size;
+		}
+	}
+
+	return err;
+}
+
 #if defined(CONFIG_DEBUG_FS)
 
 static int debugfs_show_mqd(struct seq_file *m, void *data)
@@ -486,7 +927,7 @@ static int debugfs_show_mqd_sdma(struct seq_file *m, void *data)
 #endif
 
 struct mqd_manager *mqd_manager_init_v9(enum KFD_MQD_TYPE type,
-		struct kfd_dev *dev)
+		struct kfd_node *dev)
 {
 	struct mqd_manager *mqd;
 
@@ -502,34 +943,58 @@ struct mqd_manager *mqd_manager_init_v9(enum KFD_MQD_TYPE type,
 	switch (type) {
 	case KFD_MQD_TYPE_CP:
 		mqd->allocate_mqd = allocate_mqd;
-		mqd->init_mqd = init_mqd;
 		mqd->free_mqd = kfd_free_mqd_cp;
-		mqd->load_mqd = load_mqd;
-		mqd->update_mqd = update_mqd;
-		mqd->destroy_mqd = kfd_destroy_mqd_cp;
 		mqd->is_occupied = kfd_is_occupied_cp;
-		mqd->get_wave_state = get_wave_state;
 		mqd->get_checkpoint_info = get_checkpoint_info;
-		mqd->checkpoint_mqd = checkpoint_mqd;
-		mqd->restore_mqd = restore_mqd;
 		mqd->mqd_size = sizeof(struct v9_mqd);
+		mqd->mqd_stride = mqd_stride_v9;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd;
 #endif
+		if (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 3) ||
+		    KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 4) ||
+		    KFD_GC_VERSION(dev) == IP_VERSION(9, 5, 0)) {
+			mqd->init_mqd = init_mqd_v9_4_3;
+			mqd->load_mqd = load_mqd_v9_4_3;
+			mqd->update_mqd = update_mqd_v9_4_3;
+			mqd->destroy_mqd = destroy_mqd_v9_4_3;
+			mqd->get_wave_state = get_wave_state_v9_4_3;
+			mqd->checkpoint_mqd = checkpoint_mqd_v9_4_3;
+			mqd->restore_mqd = restore_mqd_v9_4_3;
+		} else {
+			mqd->init_mqd = init_mqd;
+			mqd->load_mqd = load_mqd;
+			mqd->update_mqd = update_mqd;
+			mqd->destroy_mqd = kfd_destroy_mqd_cp;
+			mqd->get_wave_state = get_wave_state;
+			mqd->checkpoint_mqd = checkpoint_mqd;
+			mqd->restore_mqd = restore_mqd;
+		}
 		break;
 	case KFD_MQD_TYPE_HIQ:
 		mqd->allocate_mqd = allocate_hiq_mqd;
-		mqd->init_mqd = init_mqd_hiq;
 		mqd->free_mqd = free_mqd_hiq_sdma;
-		mqd->load_mqd = kfd_hiq_load_mqd_kiq;
 		mqd->update_mqd = update_mqd;
-		mqd->destroy_mqd = kfd_destroy_mqd_cp;
 		mqd->is_occupied = kfd_is_occupied_cp;
 		mqd->mqd_size = sizeof(struct v9_mqd);
+		mqd->mqd_stride = kfd_mqd_stride;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd;
 #endif
-		mqd->read_doorbell_id = read_doorbell_id;
+		mqd->check_preemption_failed = check_preemption_failed;
+		if (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 3) ||
+		    KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 4) ||
+		    KFD_GC_VERSION(dev) == IP_VERSION(9, 5, 0)) {
+			mqd->init_mqd = init_mqd_hiq_v9_4_3;
+			mqd->load_mqd = hiq_load_mqd_kiq_v9_4_3;
+			mqd->destroy_mqd = destroy_hiq_mqd_v9_4_3;
+			mqd->check_preemption_failed = check_preemption_failed_v9_4_3;
+		} else {
+			mqd->init_mqd = init_mqd_hiq;
+			mqd->load_mqd = kfd_hiq_load_mqd_kiq;
+			mqd->destroy_mqd = destroy_hiq_mqd;
+			mqd->check_preemption_failed = check_preemption_failed;
+		}
 		break;
 	case KFD_MQD_TYPE_DIQ:
 		mqd->allocate_mqd = allocate_mqd;
@@ -555,6 +1020,7 @@ struct mqd_manager *mqd_manager_init_v9(enum KFD_MQD_TYPE type,
 		mqd->checkpoint_mqd = checkpoint_mqd_sdma;
 		mqd->restore_mqd = restore_mqd_sdma;
 		mqd->mqd_size = sizeof(struct v9_sdma_mqd);
+		mqd->mqd_stride = kfd_mqd_stride;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd_sdma;
 #endif
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c
index 530ba6f5b57e..c1fafc502515 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c
@@ -51,12 +51,11 @@ static void update_cu_mask(struct mqd_manager *mm, void *mqd,
 	struct vi_mqd *m;
 	uint32_t se_mask[4] = {0}; /* 4 is the max # of SEs */
 
-	if (!minfo || (minfo->update_flag != UPDATE_FLAG_CU_MASK) ||
-	    !minfo->cu_mask.ptr)
+	if (!minfo || !minfo->cu_mask.ptr)
 		return;
 
 	mqd_symmetrically_map_cu_mask(mm,
-		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask);
+		minfo->cu_mask.ptr, minfo->cu_mask.count, se_mask, 0);
 
 	m = get_mqd(mqd);
 	m->compute_static_thread_mgmt_se0 = se_mask[0];
@@ -77,7 +76,7 @@ static void set_priority(struct vi_mqd *m, struct queue_properties *q)
 	m->cp_hqd_queue_priority = q->priority;
 }
 
-static struct kfd_mem_obj *allocate_mqd(struct kfd_dev *kfd,
+static struct kfd_mem_obj *allocate_mqd(struct kfd_node *kfd,
 					struct queue_properties *q)
 {
 	struct kfd_mem_obj *mqd_mem_obj;
@@ -136,7 +135,7 @@ static void init_mqd(struct mqd_manager *mm, void **mqd,
 			(1 << COMPUTE_PGM_RSRC2__TRAP_PRESENT__SHIFT);
 	}
 
-	if (mm->dev->cwsr_enabled && q->ctx_save_restore_area_address) {
+	if (mm->dev->kfd->cwsr_enabled && q->ctx_save_restore_area_address) {
 		m->cp_hqd_persistent_state |=
 			(1 << CP_HQD_PERSISTENT_STATE__QSWITCH_MODE__SHIFT);
 		m->cp_hqd_ctx_save_base_addr_lo =
@@ -165,7 +164,7 @@ static int load_mqd(struct mqd_manager *mm, void *mqd,
 
 	return mm->dev->kfd2kgd->hqd_load(mm->dev->adev, mqd, pipe_id, queue_id,
 					  (uint32_t __user *)p->write_ptr,
-					  wptr_shift, wptr_mask, mms);
+					  wptr_shift, wptr_mask, mms, 0);
 }
 
 static void __update_mqd(struct mqd_manager *mm, void *mqd,
@@ -227,7 +226,7 @@ static void __update_mqd(struct mqd_manager *mm, void *mqd,
 				2 << CP_HQD_PQ_CONTROL__SLOT_BASED_WPTR__SHIFT;
 	}
 
-	if (mm->dev->cwsr_enabled && q->ctx_save_restore_area_address)
+	if (mm->dev->kfd->cwsr_enabled && q->ctx_save_restore_area_address)
 		m->cp_hqd_ctx_save_control =
 			atc_bit << CP_HQD_CTX_SAVE_CONTROL__ATC__SHIFT |
 			mtype << CP_HQD_CTX_SAVE_CONTROL__MTYPE__SHIFT;
@@ -238,29 +237,22 @@ static void __update_mqd(struct mqd_manager *mm, void *mqd,
 	q->is_active = QUEUE_IS_ACTIVE(*q);
 }
 
-
-static void update_mqd(struct mqd_manager *mm, void *mqd,
-			struct queue_properties *q,
-			struct mqd_update_info *minfo)
-{
-	__update_mqd(mm, mqd, q, minfo, MTYPE_CC, 1);
-}
-
-static uint32_t read_doorbell_id(void *mqd)
+static bool check_preemption_failed(struct mqd_manager *mm, void *mqd)
 {
 	struct vi_mqd *m = (struct vi_mqd *)mqd;
 
-	return m->queue_doorbell_id0;
+	return kfd_check_hiq_mqd_doorbell_id(mm->dev, m->queue_doorbell_id0, 0);
 }
 
-static void update_mqd_tonga(struct mqd_manager *mm, void *mqd,
-			struct queue_properties *q,
-			struct mqd_update_info *minfo)
+static void update_mqd(struct mqd_manager *mm, void *mqd,
+		       struct queue_properties *q,
+		       struct mqd_update_info *minfo)
 {
 	__update_mqd(mm, mqd, q, minfo, MTYPE_UC, 0);
 }
 
 static int get_wave_state(struct mqd_manager *mm, void *mqd,
+			  struct queue_properties *q,
 			  void __user *ctl_stack,
 			  u32 *ctl_stack_used_size,
 			  u32 *save_area_used_size)
@@ -446,7 +438,7 @@ static int debugfs_show_mqd_sdma(struct seq_file *m, void *data)
 #endif
 
 struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type,
-		struct kfd_dev *dev)
+		struct kfd_node *dev)
 {
 	struct mqd_manager *mqd;
 
@@ -486,10 +478,11 @@ struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type,
 		mqd->destroy_mqd = kfd_destroy_mqd_cp;
 		mqd->is_occupied = kfd_is_occupied_cp;
 		mqd->mqd_size = sizeof(struct vi_mqd);
+		mqd->mqd_stride = kfd_mqd_stride;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd;
 #endif
-		mqd->read_doorbell_id = read_doorbell_id;
+		mqd->check_preemption_failed = check_preemption_failed;
 		break;
 	case KFD_MQD_TYPE_DIQ:
 		mqd->allocate_mqd = allocate_mqd;
@@ -500,6 +493,7 @@ struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type,
 		mqd->destroy_mqd = kfd_destroy_mqd_cp;
 		mqd->is_occupied = kfd_is_occupied_cp;
 		mqd->mqd_size = sizeof(struct vi_mqd);
+		mqd->mqd_stride = kfd_mqd_stride;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd;
 #endif
@@ -515,6 +509,7 @@ struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type,
 		mqd->checkpoint_mqd = checkpoint_mqd_sdma;
 		mqd->restore_mqd = restore_mqd_sdma;
 		mqd->mqd_size = sizeof(struct vi_sdma_mqd);
+		mqd->mqd_stride = kfd_mqd_stride;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd_sdma;
 #endif
@@ -526,16 +521,3 @@ struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type,
 
 	return mqd;
 }
-
-struct mqd_manager *mqd_manager_init_vi_tonga(enum KFD_MQD_TYPE type,
-			struct kfd_dev *dev)
-{
-	struct mqd_manager *mqd;
-
-	mqd = mqd_manager_init_vi(type, dev);
-	if (!mqd)
-		return NULL;
-	if (type == KFD_MQD_TYPE_CP)
-		mqd->update_mqd = update_mqd_tonga;
-	return mqd;
-}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
index f612325241aa..b1a6eb349bb3 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
@@ -28,6 +28,11 @@
 #include "kfd_kernel_queue.h"
 #include "kfd_priv.h"
 
+#define OVER_SUBSCRIPTION_PROCESS_COUNT (1 << 0)
+#define OVER_SUBSCRIPTION_COMPUTE_QUEUE_COUNT (1 << 1)
+#define OVER_SUBSCRIPTION_GWS_QUEUE_COUNT (1 << 2)
+#define OVER_SUBSCRIPTION_XNACK_CONFLICT (1 << 3)
+
 static inline void inc_wptr(unsigned int *wptr, unsigned int increment_bytes,
 				unsigned int buffer_size_bytes)
 {
@@ -40,12 +45,14 @@ static inline void inc_wptr(unsigned int *wptr, unsigned int increment_bytes,
 
 static void pm_calc_rlib_size(struct packet_manager *pm,
 				unsigned int *rlib_size,
-				bool *over_subscription)
+				int *over_subscription,
+				int xnack_conflict)
 {
 	unsigned int process_count, queue_count, compute_queue_count, gws_queue_count;
 	unsigned int map_queue_size;
 	unsigned int max_proc_per_quantum = 1;
-	struct kfd_dev *dev = pm->dqm->dev;
+	struct kfd_node *node = pm->dqm->dev;
+	struct device *dev = node->adev->dev;
 
 	process_count = pm->dqm->processes_count;
 	queue_count = pm->dqm->active_queue_count;
@@ -57,17 +64,22 @@ static void pm_calc_rlib_size(struct packet_manager *pm,
 	 * hws_max_conc_proc has been done in
 	 * kgd2kfd_device_init().
 	 */
-	*over_subscription = false;
+	*over_subscription = 0;
 
-	if (dev->max_proc_per_quantum > 1)
-		max_proc_per_quantum = dev->max_proc_per_quantum;
+	if (node->max_proc_per_quantum > 1)
+		max_proc_per_quantum = node->max_proc_per_quantum;
 
-	if ((process_count > max_proc_per_quantum) ||
-	    compute_queue_count > get_cp_queues_num(pm->dqm) ||
-	    gws_queue_count > 1) {
-		*over_subscription = true;
-		pr_debug("Over subscribed runlist\n");
-	}
+	if (process_count > max_proc_per_quantum)
+		*over_subscription |= OVER_SUBSCRIPTION_PROCESS_COUNT;
+	if (compute_queue_count > get_cp_queues_num(pm->dqm))
+		*over_subscription |= OVER_SUBSCRIPTION_COMPUTE_QUEUE_COUNT;
+	if (gws_queue_count > 1)
+		*over_subscription |= OVER_SUBSCRIPTION_GWS_QUEUE_COUNT;
+	if (xnack_conflict && (node->adev->gmc.xnack_flags & AMDGPU_GMC_XNACK_FLAG_CHAIN))
+		*over_subscription |= OVER_SUBSCRIPTION_XNACK_CONFLICT;
+
+	if (*over_subscription)
+		dev_dbg(dev, "Over subscribed runlist\n");
 
 	map_queue_size = pm->pmf->map_queues_size;
 	/* calculate run list ib allocation size */
@@ -81,29 +93,32 @@ static void pm_calc_rlib_size(struct packet_manager *pm,
 	if (*over_subscription)
 		*rlib_size += pm->pmf->runlist_size;
 
-	pr_debug("runlist ib size %d\n", *rlib_size);
+	dev_dbg(dev, "runlist ib size %d\n", *rlib_size);
 }
 
 static int pm_allocate_runlist_ib(struct packet_manager *pm,
 				unsigned int **rl_buffer,
 				uint64_t *rl_gpu_buffer,
 				unsigned int *rl_buffer_size,
-				bool *is_over_subscription)
+				int *is_over_subscription,
+				int xnack_conflict)
 {
+	struct kfd_node *node = pm->dqm->dev;
+	struct device *dev = node->adev->dev;
 	int retval;
 
 	if (WARN_ON(pm->allocated))
 		return -EINVAL;
 
-	pm_calc_rlib_size(pm, rl_buffer_size, is_over_subscription);
+	pm_calc_rlib_size(pm, rl_buffer_size, is_over_subscription,
+				xnack_conflict);
 
 	mutex_lock(&pm->lock);
 
-	retval = kfd_gtt_sa_allocate(pm->dqm->dev, *rl_buffer_size,
-					&pm->ib_buffer_obj);
+	retval = kfd_gtt_sa_allocate(node, *rl_buffer_size, &pm->ib_buffer_obj);
 
 	if (retval) {
-		pr_err("Failed to allocate runlist IB\n");
+		dev_err(dev, "Failed to allocate runlist IB\n");
 		goto out;
 	}
 
@@ -125,32 +140,54 @@ static int pm_create_runlist_ib(struct packet_manager *pm,
 {
 	unsigned int alloc_size_bytes;
 	unsigned int *rl_buffer, rl_wptr, i;
+	struct kfd_node *node = pm->dqm->dev;
+	struct device *dev = node->adev->dev;
 	int retval, processes_mapped;
 	struct device_process_node *cur;
 	struct qcm_process_device *qpd;
 	struct queue *q;
 	struct kernel_queue *kq;
-	bool is_over_subscription;
+	int is_over_subscription;
+	int xnack_enabled = -1;
+	bool xnack_conflict = 0;
 
 	rl_wptr = retval = processes_mapped = 0;
 
+	/* Check if processes set different xnack modes */
+	list_for_each_entry(cur, queues, list) {
+		qpd = cur->qpd;
+		if (xnack_enabled < 0)
+			/* First process */
+			xnack_enabled = qpd->pqm->process->xnack_enabled;
+		else if (qpd->pqm->process->xnack_enabled != xnack_enabled) {
+			/* Found a process with a different xnack mode */
+			xnack_conflict = 1;
+			break;
+		}
+	}
+
 	retval = pm_allocate_runlist_ib(pm, &rl_buffer, rl_gpu_addr,
-				&alloc_size_bytes, &is_over_subscription);
+				&alloc_size_bytes, &is_over_subscription,
+				xnack_conflict);
 	if (retval)
 		return retval;
 
 	*rl_size_bytes = alloc_size_bytes;
 	pm->ib_size_bytes = alloc_size_bytes;
 
-	pr_debug("Building runlist ib process count: %d queues count %d\n",
+	dev_dbg(dev, "Building runlist ib process count: %d queues count %d\n",
 		pm->dqm->processes_count, pm->dqm->active_queue_count);
 
+build_runlist_ib:
 	/* build the run list ib packet */
 	list_for_each_entry(cur, queues, list) {
 		qpd = cur->qpd;
+		/* group processes with the same xnack mode together */
+		if (qpd->pqm->process->xnack_enabled != xnack_enabled)
+			continue;
 		/* build map process packet */
 		if (processes_mapped >= pm->dqm->processes_count) {
-			pr_debug("Not enough space left in runlist IB\n");
+			dev_dbg(dev, "Not enough space left in runlist IB\n");
 			pm_release_ib(pm);
 			return -ENOMEM;
 		}
@@ -167,7 +204,8 @@ static int pm_create_runlist_ib(struct packet_manager *pm,
 			if (!kq->queue->properties.is_active)
 				continue;
 
-			pr_debug("static_queue, mapping kernel q %d, is debug status %d\n",
+			dev_dbg(dev,
+				"static_queue, mapping kernel q %d, is debug status %d\n",
 				kq->queue->queue, qpd->is_debug);
 
 			retval = pm->pmf->map_queues(pm,
@@ -186,7 +224,8 @@ static int pm_create_runlist_ib(struct packet_manager *pm,
 			if (!q->properties.is_active)
 				continue;
 
-			pr_debug("static_queue, mapping user queue %d, is debug status %d\n",
+			dev_dbg(dev,
+				"static_queue, mapping user queue %d, is debug status %d\n",
 				q->queue, qpd->is_debug);
 
 			retval = pm->pmf->map_queues(pm,
@@ -202,18 +241,33 @@ static int pm_create_runlist_ib(struct packet_manager *pm,
 				alloc_size_bytes);
 		}
 	}
+	if (xnack_conflict) {
+		/* pick up processes with the other xnack mode */
+		xnack_enabled = !xnack_enabled;
+		xnack_conflict = 0;
+		goto build_runlist_ib;
+	}
 
-	pr_debug("Finished map process and queues to runlist\n");
+	dev_dbg(dev, "Finished map process and queues to runlist\n");
 
 	if (is_over_subscription) {
 		if (!pm->is_over_subscription)
-			pr_warn("Runlist is getting oversubscribed. Expect reduced ROCm performance.\n");
+			dev_warn(dev, "Runlist is getting oversubscribed due to%s%s%s%s. Expect reduced ROCm performance.\n",
+				is_over_subscription & OVER_SUBSCRIPTION_PROCESS_COUNT ?
+				" too many processes" : "",
+				is_over_subscription & OVER_SUBSCRIPTION_COMPUTE_QUEUE_COUNT ?
+				" too many queues" : "",
+				is_over_subscription & OVER_SUBSCRIPTION_GWS_QUEUE_COUNT ?
+				" multiple processes using cooperative launch" : "",
+				is_over_subscription & OVER_SUBSCRIPTION_XNACK_CONFLICT ?
+				" xnack on/off processes mixed on gfx9" : "");
+
 		retval = pm->pmf->runlist(pm, &rl_buffer[rl_wptr],
 					*rl_gpu_addr,
 					alloc_size_bytes / sizeof(uint32_t),
 					true);
 	}
-	pm->is_over_subscription = is_over_subscription;
+	pm->is_over_subscription = !!is_over_subscription;
 
 	for (i = 0; i < alloc_size_bytes / sizeof(uint32_t); i++)
 		pr_debug("0x%2X ", rl_buffer[i]);
@@ -239,7 +293,9 @@ int pm_init(struct packet_manager *pm, struct device_queue_manager *dqm)
 		break;
 	default:
 		if (KFD_GC_VERSION(dqm->dev) == IP_VERSION(9, 4, 2) ||
-		    KFD_GC_VERSION(dqm->dev) == IP_VERSION(9, 4, 3))
+		    KFD_GC_VERSION(dqm->dev) == IP_VERSION(9, 4, 3) ||
+		    KFD_GC_VERSION(dqm->dev) == IP_VERSION(9, 4, 4) ||
+		    KFD_GC_VERSION(dqm->dev) == IP_VERSION(9, 5, 0))
 			pm->pmf = &kfd_aldebaran_pm_funcs;
 		else if (KFD_GC_VERSION(dqm->dev) >= IP_VERSION(9, 0, 1))
 			pm->pmf = &kfd_v9_pm_funcs;
@@ -262,16 +318,18 @@ int pm_init(struct packet_manager *pm, struct device_queue_manager *dqm)
 	return 0;
 }
 
-void pm_uninit(struct packet_manager *pm, bool hanging)
+void pm_uninit(struct packet_manager *pm)
 {
 	mutex_destroy(&pm->lock);
-	kernel_queue_uninit(pm->priv_queue, hanging);
+	kernel_queue_uninit(pm->priv_queue);
 	pm->priv_queue = NULL;
 }
 
 int pm_send_set_resources(struct packet_manager *pm,
 				struct scheduling_resources *res)
 {
+	struct kfd_node *node = pm->dqm->dev;
+	struct device *dev = node->adev->dev;
 	uint32_t *buffer, size;
 	int retval = 0;
 
@@ -281,14 +339,14 @@ int pm_send_set_resources(struct packet_manager *pm,
 					size / sizeof(uint32_t),
 					(unsigned int **)&buffer);
 	if (!buffer) {
-		pr_err("Failed to allocate buffer on kernel queue\n");
+		dev_err(dev, "Failed to allocate buffer on kernel queue\n");
 		retval = -ENOMEM;
 		goto out;
 	}
 
 	retval = pm->pmf->set_resources(pm, buffer, res);
 	if (!retval)
-		kq_submit_packet(pm->priv_queue);
+		retval = kq_submit_packet(pm->priv_queue);
 	else
 		kq_rollback_packet(pm->priv_queue);
 
@@ -325,7 +383,7 @@ int pm_send_runlist(struct packet_manager *pm, struct list_head *dqm_queues)
 	if (retval)
 		goto fail_create_runlist;
 
-	kq_submit_packet(pm->priv_queue);
+	retval = kq_submit_packet(pm->priv_queue);
 
 	mutex_unlock(&pm->lock);
 
@@ -343,6 +401,8 @@ fail_create_runlist_ib:
 int pm_send_query_status(struct packet_manager *pm, uint64_t fence_address,
 			uint64_t fence_value)
 {
+	struct kfd_node *node = pm->dqm->dev;
+	struct device *dev = node->adev->dev;
 	uint32_t *buffer, size;
 	int retval = 0;
 
@@ -354,14 +414,14 @@ int pm_send_query_status(struct packet_manager *pm, uint64_t fence_address,
 	kq_acquire_packet_buffer(pm->priv_queue,
 			size / sizeof(uint32_t), (unsigned int **)&buffer);
 	if (!buffer) {
-		pr_err("Failed to allocate buffer on kernel queue\n");
+		dev_err(dev, "Failed to allocate buffer on kernel queue\n");
 		retval = -ENOMEM;
 		goto out;
 	}
 
 	retval = pm->pmf->query_status(pm, buffer, fence_address, fence_value);
 	if (!retval)
-		kq_submit_packet(pm->priv_queue);
+		retval = kq_submit_packet(pm->priv_queue);
 	else
 		kq_rollback_packet(pm->priv_queue);
 
@@ -370,10 +430,71 @@ out:
 	return retval;
 }
 
+/* pm_config_dequeue_wait_counts: Configure dequeue timer Wait Counts
+ *  by writing to CP_IQ_WAIT_TIME2 registers.
+ *
+ *  @cmd: See emum kfd_config_dequeue_wait_counts_cmd definition
+ *  @value: Depends on the cmd. This parameter is unused for
+ *    KFD_DEQUEUE_WAIT_INIT and KFD_DEQUEUE_WAIT_RESET. For
+ *    KFD_DEQUEUE_WAIT_SET_SCH_WAVE it holds value to be set
+ *
+ */
+int pm_config_dequeue_wait_counts(struct packet_manager *pm,
+		enum kfd_config_dequeue_wait_counts_cmd cmd,
+		uint32_t value)
+{
+	struct kfd_node *node = pm->dqm->dev;
+	struct device *dev = node->adev->dev;
+	int retval = 0;
+	uint32_t *buffer, size;
+
+	if (!pm->pmf->config_dequeue_wait_counts ||
+	    !pm->pmf->config_dequeue_wait_counts_size)
+		return 0;
+
+	if (cmd == KFD_DEQUEUE_WAIT_INIT && (KFD_GC_VERSION(pm->dqm->dev) < IP_VERSION(9, 4, 1) ||
+	   KFD_GC_VERSION(pm->dqm->dev) >= IP_VERSION(10, 0, 0)))
+		return 0;
+
+	size = pm->pmf->config_dequeue_wait_counts_size;
+
+	mutex_lock(&pm->lock);
+
+	if (size) {
+		kq_acquire_packet_buffer(pm->priv_queue,
+			size / sizeof(uint32_t),
+			(unsigned int **)&buffer);
+
+		if (!buffer) {
+			dev_err(dev,
+				"Failed to allocate buffer on kernel queue\n");
+			retval = -ENOMEM;
+			goto out;
+		}
+
+		retval = pm->pmf->config_dequeue_wait_counts(pm, buffer,
+							     cmd, value);
+		if (!retval) {
+			retval = kq_submit_packet(pm->priv_queue);
+
+			/* If default value is modified, cache that in dqm->wait_times */
+			if (!retval && cmd == KFD_DEQUEUE_WAIT_INIT)
+				update_dqm_wait_times(pm->dqm);
+		} else {
+			kq_rollback_packet(pm->priv_queue);
+		}
+	}
+out:
+	mutex_unlock(&pm->lock);
+	return retval;
+}
+
 int pm_send_unmap_queue(struct packet_manager *pm,
 			enum kfd_unmap_queues_filter filter,
 			uint32_t filter_param, bool reset)
 {
+	struct kfd_node *node = pm->dqm->dev;
+	struct device *dev = node->adev->dev;
 	uint32_t *buffer, size;
 	int retval = 0;
 
@@ -382,14 +503,14 @@ int pm_send_unmap_queue(struct packet_manager *pm,
 	kq_acquire_packet_buffer(pm->priv_queue,
 			size / sizeof(uint32_t), (unsigned int **)&buffer);
 	if (!buffer) {
-		pr_err("Failed to allocate buffer on kernel queue\n");
+		dev_err(dev, "Failed to allocate buffer on kernel queue\n");
 		retval = -ENOMEM;
 		goto out;
 	}
 
 	retval = pm->pmf->unmap_queues(pm, buffer, filter, filter_param, reset);
 	if (!retval)
-		kq_submit_packet(pm->priv_queue);
+		retval = kq_submit_packet(pm->priv_queue);
 	else
 		kq_rollback_packet(pm->priv_queue);
 
@@ -431,6 +552,8 @@ out:
 
 int pm_debugfs_hang_hws(struct packet_manager *pm)
 {
+	struct kfd_node *node = pm->dqm->dev;
+	struct device *dev = node->adev->dev;
 	uint32_t *buffer, size;
 	int r = 0;
 
@@ -442,16 +565,16 @@ int pm_debugfs_hang_hws(struct packet_manager *pm)
 	kq_acquire_packet_buffer(pm->priv_queue,
 			size / sizeof(uint32_t), (unsigned int **)&buffer);
 	if (!buffer) {
-		pr_err("Failed to allocate buffer on kernel queue\n");
+		dev_err(dev, "Failed to allocate buffer on kernel queue\n");
 		r = -ENOMEM;
 		goto out;
 	}
 	memset(buffer, 0x55, size);
 	kq_submit_packet(pm->priv_queue);
 
-	pr_info("Submitting %x %x %x %x %x %x %x to HIQ to hang the HWS.",
-		buffer[0], buffer[1], buffer[2], buffer[3],
-		buffer[4], buffer[5], buffer[6]);
+	dev_info(dev, "Submitting %x %x %x %x %x %x %x to HIQ to hang the HWS.",
+		 buffer[0], buffer[1], buffer[2], buffer[3], buffer[4],
+		 buffer[5], buffer[6]);
 out:
 	mutex_unlock(&pm->lock);
 	return r;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c
index 18250845a989..505036968a77 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c
@@ -34,14 +34,20 @@ static int pm_map_process_v9(struct packet_manager *pm,
 {
 	struct pm4_mes_map_process *packet;
 	uint64_t vm_page_table_base_addr = qpd->page_table_base;
+	struct kfd_node *kfd = pm->dqm->dev;
+	struct kfd_process_device *pdd =
+			container_of(qpd, struct kfd_process_device, qpd);
+	struct amdgpu_device *adev = kfd->adev;
 
 	packet = (struct pm4_mes_map_process *)buffer;
 	memset(buffer, 0, sizeof(struct pm4_mes_map_process));
 	packet->header.u32All = pm_build_pm4_header(IT_MAP_PROCESS,
 					sizeof(struct pm4_mes_map_process));
+	if (adev->enforce_isolation[kfd->node_id] == AMDGPU_ENFORCE_ISOLATION_ENABLE)
+		packet->bitfields2.exec_cleaner_shader = 1;
 	packet->bitfields2.diq_enable = (qpd->is_debug) ? 1 : 0;
 	packet->bitfields2.process_quantum = 10;
-	packet->bitfields2.pasid = qpd->pqm->process->pasid;
+	packet->bitfields2.pasid = pdd->pasid;
 	packet->bitfields14.gds_size = qpd->gds_size & 0x3F;
 	packet->bitfields14.gds_size_hi = (qpd->gds_size >> 6) & 0xF;
 	packet->bitfields14.num_gws = (qpd->mapped_gws_queue) ? qpd->num_gws : 0;
@@ -49,6 +55,12 @@ static int pm_map_process_v9(struct packet_manager *pm,
 	packet->bitfields14.sdma_enable = 1;
 	packet->bitfields14.num_queues = (qpd->is_debug) ? 0 : qpd->queue_count;
 
+	if (kfd->dqm->trap_debug_vmid && pdd->process->debug_trap_enabled &&
+			pdd->process->runtime_info.runtime_state == DEBUG_RUNTIME_STATE_ENABLED) {
+		packet->bitfields2.debug_vmid = kfd->dqm->trap_debug_vmid;
+		packet->bitfields2.new_debug = 1;
+	}
+
 	packet->sh_mem_config = qpd->sh_mem_config;
 	packet->sh_mem_bases = qpd->sh_mem_bases;
 	if (qpd->tba_addr) {
@@ -79,25 +91,45 @@ static int pm_map_process_aldebaran(struct packet_manager *pm,
 {
 	struct pm4_mes_map_process_aldebaran *packet;
 	uint64_t vm_page_table_base_addr = qpd->page_table_base;
+	struct kfd_dev *kfd = pm->dqm->dev->kfd;
+	struct kfd_node *knode = pm->dqm->dev;
+	struct kfd_process_device *pdd =
+			container_of(qpd, struct kfd_process_device, qpd);
+	int i;
+	struct amdgpu_device *adev = kfd->adev;
 
 	packet = (struct pm4_mes_map_process_aldebaran *)buffer;
 	memset(buffer, 0, sizeof(struct pm4_mes_map_process_aldebaran));
 	packet->header.u32All = pm_build_pm4_header(IT_MAP_PROCESS,
 			sizeof(struct pm4_mes_map_process_aldebaran));
+	if (adev->enforce_isolation[knode->node_id] ==
+	    AMDGPU_ENFORCE_ISOLATION_ENABLE)
+		packet->bitfields2.exec_cleaner_shader = 1;
 	packet->bitfields2.diq_enable = (qpd->is_debug) ? 1 : 0;
 	packet->bitfields2.process_quantum = 10;
-	packet->bitfields2.pasid = qpd->pqm->process->pasid;
+	packet->bitfields2.pasid = pdd->pasid;
 	packet->bitfields14.gds_size = qpd->gds_size & 0x3F;
 	packet->bitfields14.gds_size_hi = (qpd->gds_size >> 6) & 0xF;
 	packet->bitfields14.num_gws = (qpd->mapped_gws_queue) ? qpd->num_gws : 0;
 	packet->bitfields14.num_oac = qpd->num_oac;
 	packet->bitfields14.sdma_enable = 1;
 	packet->bitfields14.num_queues = (qpd->is_debug) ? 0 : qpd->queue_count;
+	packet->spi_gdbg_per_vmid_cntl = pdd->spi_dbg_override |
+						pdd->spi_dbg_launch_mode;
+
+	if (pdd->process->debug_trap_enabled) {
+		for (i = 0; i < kfd->device_info.num_of_watch_points; i++)
+			packet->tcp_watch_cntl[i] = pdd->watch_points[i];
+
+		packet->bitfields2.single_memops =
+				!!(pdd->process->dbg_flags & KFD_DBG_TRAP_FLAG_SINGLE_MEM_OP);
+	}
 
 	packet->sh_mem_config = qpd->sh_mem_config;
 	packet->sh_mem_bases = qpd->sh_mem_bases;
 	if (qpd->tba_addr) {
 		packet->sq_shader_tba_lo = lower_32_bits(qpd->tba_addr >> 8);
+		packet->sq_shader_tba_hi = upper_32_bits(qpd->tba_addr >> 8);
 		packet->sq_shader_tma_lo = lower_32_bits(qpd->tma_addr >> 8);
 		packet->sq_shader_tma_hi = upper_32_bits(qpd->tma_addr >> 8);
 	}
@@ -119,19 +151,24 @@ static int pm_runlist_v9(struct packet_manager *pm, uint32_t *buffer,
 	struct pm4_mes_runlist *packet;
 
 	int concurrent_proc_cnt = 0;
-	struct kfd_dev *kfd = pm->dqm->dev;
+	struct kfd_node *kfd = pm->dqm->dev;
+	struct amdgpu_device *adev = kfd->adev;
 
 	/* Determine the number of processes to map together to HW:
 	 * it can not exceed the number of VMIDs available to the
 	 * scheduler, and it is determined by the smaller of the number
 	 * of processes in the runlist and kfd module parameter
 	 * hws_max_conc_proc.
+	 * However, if enforce_isolation is set (toggle LDS/VGPRs/SGPRs
+	 * cleaner between process switch), enable single-process mode
+	 * in HWS.
 	 * Note: the arbitration between the number of VMIDs and
 	 * hws_max_conc_proc has been done in
 	 * kgd2kfd_device_init().
 	 */
-	concurrent_proc_cnt = min(pm->dqm->processes_count,
-			kfd->max_proc_per_quantum);
+	concurrent_proc_cnt = (adev->enforce_isolation[kfd->node_id] ==
+			       AMDGPU_ENFORCE_ISOLATION_ENABLE) ?
+		1 : min(pm->dqm->processes_count, kfd->max_proc_per_quantum);
 
 	packet = (struct pm4_mes_runlist *)buffer;
 
@@ -166,6 +203,8 @@ static int pm_set_resources_v9(struct packet_manager *pm, uint32_t *buffer,
 			queue_type__mes_set_resources__hsa_interface_queue_hiq;
 	packet->bitfields2.vmid_mask = res->vmid_mask;
 	packet->bitfields2.unmap_latency = KFD_UNMAP_LATENCY_MS / 100;
+	if (pm->dqm->dev->adev->gmc.xnack_flags & AMDGPU_GMC_XNACK_FLAG_CHAIN)
+		packet->bitfields2.enb_xnack_retry_disable_check = 1;
 	packet->bitfields7.oac_mask = res->oac_mask;
 	packet->bitfields8.gds_heap_base = res->gds_heap_base;
 	packet->bitfields8.gds_heap_size = res->gds_heap_size;
@@ -181,14 +220,14 @@ static int pm_set_resources_v9(struct packet_manager *pm, uint32_t *buffer,
 
 static inline bool pm_use_ext_eng(struct kfd_dev *dev)
 {
-	return dev->adev->ip_versions[SDMA0_HWIP][0] >= IP_VERSION(5, 2, 0);
+	return amdgpu_ip_version(dev->adev, SDMA0_HWIP, 0) >=
+	       IP_VERSION(5, 2, 0);
 }
 
 static int pm_map_queues_v9(struct packet_manager *pm, uint32_t *buffer,
 		struct queue *q, bool is_static)
 {
 	struct pm4_mes_map_queues *packet;
-	bool use_static = is_static;
 
 	packet = (struct pm4_mes_map_queues *)buffer;
 	memset(buffer, 0, sizeof(struct pm4_mes_map_queues));
@@ -201,7 +240,7 @@ static int pm_map_queues_v9(struct packet_manager *pm, uint32_t *buffer,
 
 	packet->bitfields2.engine_sel =
 		engine_sel__mes_map_queues__compute_vi;
-	packet->bitfields2.gws_control_queue = q->gws ? 1 : 0;
+	packet->bitfields2.gws_control_queue = q->properties.is_gws ? 1 : 0;
 	packet->bitfields2.extended_engine_sel =
 		extended_engine_sel__mes_map_queues__legacy_engine_sel;
 	packet->bitfields2.queue_type =
@@ -209,7 +248,7 @@ static int pm_map_queues_v9(struct packet_manager *pm, uint32_t *buffer,
 
 	switch (q->properties.type) {
 	case KFD_QUEUE_TYPE_COMPUTE:
-		if (use_static)
+		if (is_static)
 			packet->bitfields2.queue_type =
 		queue_type__mes_map_queues__normal_latency_static_queue_vi;
 		break;
@@ -219,14 +258,24 @@ static int pm_map_queues_v9(struct packet_manager *pm, uint32_t *buffer,
 		break;
 	case KFD_QUEUE_TYPE_SDMA:
 	case KFD_QUEUE_TYPE_SDMA_XGMI:
-		use_static = false; /* no static queues under SDMA */
-		if (q->properties.sdma_engine_id < 2 && !pm_use_ext_eng(q->device))
+		if (q->properties.sdma_engine_id < 2 &&
+		    !pm_use_ext_eng(q->device->kfd))
 			packet->bitfields2.engine_sel = q->properties.sdma_engine_id +
 				engine_sel__mes_map_queues__sdma0_vi;
 		else {
-			packet->bitfields2.extended_engine_sel =
-				extended_engine_sel__mes_map_queues__sdma0_to_7_sel;
-			packet->bitfields2.engine_sel = q->properties.sdma_engine_id;
+			/*
+			 * For GFX9.4.3, SDMA engine id can be greater than 8.
+			 * For such cases, set extended_engine_sel to 2 and
+			 * ensure engine_sel lies between 0-7.
+			 */
+			if (q->properties.sdma_engine_id >= 8)
+				packet->bitfields2.extended_engine_sel =
+					extended_engine_sel__mes_map_queues__sdma8_to_15_sel;
+			else
+				packet->bitfields2.extended_engine_sel =
+					extended_engine_sel__mes_map_queues__sdma0_to_7_sel;
+
+			packet->bitfields2.engine_sel = q->properties.sdma_engine_id % 8;
 		}
 		break;
 	default:
@@ -251,6 +300,97 @@ static int pm_map_queues_v9(struct packet_manager *pm, uint32_t *buffer,
 	return 0;
 }
 
+static inline void pm_build_dequeue_wait_counts_packet_info(struct packet_manager *pm,
+			uint32_t sch_value, uint32_t que_sleep, uint32_t *reg_offset,
+			uint32_t *reg_data)
+{
+	pm->dqm->dev->kfd2kgd->build_dequeue_wait_counts_packet_info(
+		pm->dqm->dev->adev,
+		pm->dqm->wait_times,
+		sch_value,
+		que_sleep,
+		reg_offset,
+		reg_data);
+}
+
+/* pm_config_dequeue_wait_counts_v9: Builds WRITE_DATA packet with
+ *    register/value for configuring dequeue wait counts
+ *
+ * @return: -ve for failure and 0 for success and buffer is
+ *  filled in with packet
+ *
+ **/
+static int pm_config_dequeue_wait_counts_v9(struct packet_manager *pm,
+		uint32_t *buffer,
+		enum kfd_config_dequeue_wait_counts_cmd cmd,
+		uint32_t value)
+{
+	struct pm4_mec_write_data_mmio *packet;
+	uint32_t reg_offset = 0;
+	uint32_t reg_data = 0;
+
+	switch (cmd) {
+	case KFD_DEQUEUE_WAIT_INIT: {
+		uint32_t sch_wave = 0, que_sleep = 1;
+
+		/* For all gfx9 ASICs > gfx941,
+		 * Reduce CP_IQ_WAIT_TIME2.QUE_SLEEP to 0x1 from default 0x40.
+		 * On a 1GHz machine this is roughly 1 microsecond, which is
+		 * about how long it takes to load data out of memory during
+		 * queue connect
+		 * QUE_SLEEP: Wait Count for Dequeue Retry.
+		 *
+		 * Set CWSR grace period to 1x1000 cycle for GFX9.4.3 APU
+		 */
+		if (KFD_GC_VERSION(pm->dqm->dev) < IP_VERSION(9, 4, 1) ||
+		    KFD_GC_VERSION(pm->dqm->dev) >= IP_VERSION(10, 0, 0))
+			return -EPERM;
+
+		if (amdgpu_emu_mode == 0 && pm->dqm->dev->adev->gmc.is_app_apu &&
+		    (KFD_GC_VERSION(pm->dqm->dev) == IP_VERSION(9, 4, 3)))
+			sch_wave = 1;
+
+		pm_build_dequeue_wait_counts_packet_info(pm, sch_wave, que_sleep,
+			&reg_offset, &reg_data);
+
+		break;
+	}
+	case KFD_DEQUEUE_WAIT_RESET:
+		/* reg_data would be set to dqm->wait_times */
+		pm_build_dequeue_wait_counts_packet_info(pm, 0, 0, &reg_offset, &reg_data);
+		break;
+
+	case KFD_DEQUEUE_WAIT_SET_SCH_WAVE:
+		/* The CP cannot handle value 0 and it will result in
+		 * an infinite grace period being set so set to 1 to prevent this. Also
+		 * avoid debugger API breakage as it sets 0 and expects a low value.
+		 */
+		if (!value)
+			value = 1;
+		pm_build_dequeue_wait_counts_packet_info(pm, value, 0, &reg_offset, &reg_data);
+		break;
+	default:
+		pr_err("Invalid dequeue wait cmd\n");
+		return -EINVAL;
+	}
+
+	packet = (struct pm4_mec_write_data_mmio *)buffer;
+	memset(buffer, 0, sizeof(struct pm4_mec_write_data_mmio));
+
+	packet->header.u32All = pm_build_pm4_header(IT_WRITE_DATA,
+					sizeof(struct pm4_mec_write_data_mmio));
+
+	packet->bitfields2.dst_sel  = dst_sel___write_data__mem_mapped_register;
+	packet->bitfields2.addr_incr =
+			addr_incr___write_data__do_not_increment_address;
+
+	packet->bitfields3.dst_mmreg_addr = reg_offset;
+
+	packet->data = reg_data;
+
+	return 0;
+}
+
 static int pm_unmap_queues_v9(struct packet_manager *pm, uint32_t *buffer,
 			enum kfd_unmap_queues_filter filter,
 			uint32_t filter_param, bool reset)
@@ -263,7 +403,8 @@ static int pm_unmap_queues_v9(struct packet_manager *pm, uint32_t *buffer,
 	packet->header.u32All = pm_build_pm4_header(IT_UNMAP_QUEUES,
 					sizeof(struct pm4_mes_unmap_queues));
 
-	packet->bitfields2.extended_engine_sel = pm_use_ext_eng(pm->dqm->dev) ?
+	packet->bitfields2.extended_engine_sel =
+				pm_use_ext_eng(pm->dqm->dev->kfd) ?
 		extended_engine_sel__mes_unmap_queues__sdma0_to_7_sel :
 		extended_engine_sel__mes_unmap_queues__legacy_engine_sel;
 
@@ -333,6 +474,7 @@ const struct packet_manager_funcs kfd_v9_pm_funcs = {
 	.set_resources		= pm_set_resources_v9,
 	.map_queues		= pm_map_queues_v9,
 	.unmap_queues		= pm_unmap_queues_v9,
+	.config_dequeue_wait_counts = pm_config_dequeue_wait_counts_v9,
 	.query_status		= pm_query_status_v9,
 	.release_mem		= NULL,
 	.map_process_size	= sizeof(struct pm4_mes_map_process),
@@ -340,6 +482,7 @@ const struct packet_manager_funcs kfd_v9_pm_funcs = {
 	.set_resources_size	= sizeof(struct pm4_mes_set_resources),
 	.map_queues_size	= sizeof(struct pm4_mes_map_queues),
 	.unmap_queues_size	= sizeof(struct pm4_mes_unmap_queues),
+	.config_dequeue_wait_counts_size  = sizeof(struct pm4_mec_write_data_mmio),
 	.query_status_size	= sizeof(struct pm4_mes_query_status),
 	.release_mem_size	= 0,
 };
@@ -350,6 +493,7 @@ const struct packet_manager_funcs kfd_aldebaran_pm_funcs = {
 	.set_resources		= pm_set_resources_v9,
 	.map_queues		= pm_map_queues_v9,
 	.unmap_queues		= pm_unmap_queues_v9,
+	.config_dequeue_wait_counts = pm_config_dequeue_wait_counts_v9,
 	.query_status		= pm_query_status_v9,
 	.release_mem		= NULL,
 	.map_process_size	= sizeof(struct pm4_mes_map_process_aldebaran),
@@ -357,6 +501,7 @@ const struct packet_manager_funcs kfd_aldebaran_pm_funcs = {
 	.set_resources_size	= sizeof(struct pm4_mes_set_resources),
 	.map_queues_size	= sizeof(struct pm4_mes_map_queues),
 	.unmap_queues_size	= sizeof(struct pm4_mes_unmap_queues),
+	.config_dequeue_wait_counts_size  = sizeof(struct pm4_mec_write_data_mmio),
 	.query_status_size	= sizeof(struct pm4_mes_query_status),
 	.release_mem_size	= 0,
 };
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_vi.c
index 4f951eaa6ee8..a1de5d7e173a 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_vi.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_vi.c
@@ -42,6 +42,7 @@ unsigned int pm_build_pm4_header(unsigned int opcode, size_t packet_size)
 static int pm_map_process_vi(struct packet_manager *pm, uint32_t *buffer,
 				struct qcm_process_device *qpd)
 {
+	struct kfd_process_device *pdd = qpd_to_pdd(qpd);
 	struct pm4_mes_map_process *packet;
 
 	packet = (struct pm4_mes_map_process *)buffer;
@@ -52,7 +53,7 @@ static int pm_map_process_vi(struct packet_manager *pm, uint32_t *buffer,
 					sizeof(struct pm4_mes_map_process));
 	packet->bitfields2.diq_enable = (qpd->is_debug) ? 1 : 0;
 	packet->bitfields2.process_quantum = 10;
-	packet->bitfields2.pasid = qpd->pqm->process->pasid;
+	packet->bitfields2.pasid = pdd->pasid;
 	packet->bitfields3.page_table_base = qpd->page_table_base;
 	packet->bitfields10.gds_size = qpd->gds_size;
 	packet->bitfields10.num_gws = qpd->num_gws;
@@ -77,7 +78,7 @@ static int pm_runlist_vi(struct packet_manager *pm, uint32_t *buffer,
 {
 	struct pm4_mes_runlist *packet;
 	int concurrent_proc_cnt = 0;
-	struct kfd_dev *kfd = pm->dqm->dev;
+	struct kfd_node *kfd = pm->dqm->dev;
 
 	if (WARN_ON(!ib))
 		return -EFAULT;
@@ -303,6 +304,7 @@ const struct packet_manager_funcs kfd_vi_pm_funcs = {
 	.set_resources		= pm_set_resources_vi,
 	.map_queues		= pm_map_queues_vi,
 	.unmap_queues		= pm_unmap_queues_vi,
+	.config_dequeue_wait_counts	= NULL,
 	.query_status		= pm_query_status_vi,
 	.release_mem		= pm_release_mem_vi,
 	.map_process_size	= sizeof(struct pm4_mes_map_process),
@@ -310,6 +312,7 @@ const struct packet_manager_funcs kfd_vi_pm_funcs = {
 	.set_resources_size	= sizeof(struct pm4_mes_set_resources),
 	.map_queues_size	= sizeof(struct pm4_mes_map_queues),
 	.unmap_queues_size	= sizeof(struct pm4_mes_unmap_queues),
+	.config_dequeue_wait_counts_size	= 0,
 	.query_status_size	= sizeof(struct pm4_mes_query_status),
 	.release_mem_size	= sizeof(struct pm4_mec_release_mem)
 };
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h
index a666710ed403..e356a207d03c 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h
@@ -63,7 +63,8 @@ struct pm4_mes_set_resources {
 		struct {
 			uint32_t vmid_mask:16;
 			uint32_t unmap_latency:8;
-			uint32_t reserved1:5;
+			uint32_t reserved1:4;
+			uint32_t enb_xnack_retry_disable_check:1;
 			enum mes_set_resources_queue_type_enum queue_type:3;
 		} bitfields2;
 		uint32_t ordinal2;
@@ -145,8 +146,12 @@ struct pm4_mes_map_process {
 
 	union {
 		struct {
-			uint32_t pasid:16;
-			uint32_t reserved1:8;
+			uint32_t pasid:16;		/* 0 - 15  */
+			uint32_t reserved1:1;		/* 16      */
+			uint32_t exec_cleaner_shader:1;	/* 17      */
+			uint32_t debug_vmid:4;
+			uint32_t new_debug:1;
+			uint32_t reserved2:1;
 			uint32_t diq_enable:1;
 			uint32_t process_quantum:7;
 		} bitfields2;
@@ -263,7 +268,8 @@ enum mes_map_queues_engine_sel_enum {
 
 enum mes_map_queues_extended_engine_sel_enum {
 	extended_engine_sel__mes_map_queues__legacy_engine_sel = 0,
-	extended_engine_sel__mes_map_queues__sdma0_to_7_sel = 1
+	extended_engine_sel__mes_map_queues__sdma0_to_7_sel  = 1,
+	extended_engine_sel__mes_map_queues__sdma8_to_15_sel = 2
 };
 
 struct pm4_mes_map_queues {
@@ -583,6 +589,71 @@ struct pm4_mec_release_mem {
 
 #endif
 
+#ifndef PM4_MEC_WRITE_DATA_DEFINED
+#define PM4_MEC_WRITE_DATA_DEFINED
+
+enum WRITE_DATA_dst_sel_enum {
+	dst_sel___write_data__mem_mapped_register = 0,
+	dst_sel___write_data__tc_l2 = 2,
+	dst_sel___write_data__gds = 3,
+	dst_sel___write_data__memory = 5,
+	dst_sel___write_data__memory_mapped_adc_persistent_state = 6,
+};
+
+enum WRITE_DATA_addr_incr_enum {
+	addr_incr___write_data__increment_address = 0,
+	addr_incr___write_data__do_not_increment_address = 1
+};
+
+enum WRITE_DATA_wr_confirm_enum {
+	wr_confirm___write_data__do_not_wait_for_write_confirmation = 0,
+	wr_confirm___write_data__wait_for_write_confirmation = 1
+};
+
+enum WRITE_DATA_cache_policy_enum {
+	cache_policy___write_data__lru = 0,
+	cache_policy___write_data__stream = 1
+};
+
+
+struct pm4_mec_write_data_mmio {
+	union {
+		union PM4_MES_TYPE_3_HEADER header;     /*header */
+		unsigned int ordinal1;
+	};
+
+	union {
+		struct {
+			unsigned int reserved1:8;
+			unsigned int dst_sel:4;
+			unsigned int reserved2:4;
+			unsigned int addr_incr:1;
+			unsigned int reserved3:2;
+			unsigned int resume_vf:1;
+			unsigned int wr_confirm:1;
+			unsigned int reserved4:4;
+			unsigned int cache_policy:2;
+			unsigned int reserved5:5;
+		} bitfields2;
+		unsigned int ordinal2;
+	};
+
+	union {
+		struct {
+			unsigned int dst_mmreg_addr:18;
+			unsigned int reserved6:14;
+		} bitfields3;
+		unsigned int ordinal3;
+	};
+
+	uint32_t reserved7;
+
+	uint32_t data;
+
+};
+
+#endif
+
 enum {
 	CACHE_FLUSH_AND_INV_TS_EVENT = 0x00000014
 };
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_aldebaran.h b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_aldebaran.h
index 38f5cb6a222a..e0ed62c4ade0 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_aldebaran.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_aldebaran.h
@@ -37,7 +37,7 @@ struct pm4_mes_map_process_aldebaran {
 		struct {
 			uint32_t pasid:16;	    /* 0 - 15  */
 			uint32_t single_memops:1;   /* 16      */
-			uint32_t reserved1:1;	    /* 17      */
+			uint32_t exec_cleaner_shader:1;	    /* 17      */
 			uint32_t debug_vmid:4;	    /* 18 - 21 */
 			uint32_t new_debug:1;	    /* 22      */
 			uint32_t tmz:1;		    /* 23      */
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 94a438956868..70ef051511bb 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -32,7 +32,7 @@
 #include <linux/atomic.h>
 #include <linux/workqueue.h>
 #include <linux/spinlock.h>
-#include <linux/kfd_ioctl.h>
+#include <uapi/linux/kfd_ioctl.h>
 #include <linux/idr.h>
 #include <linux/kfifo.h>
 #include <linux/seq_file.h>
@@ -99,11 +99,11 @@
 /*
  * Size of the per-process TBA+TMA buffer: 2 pages
  *
- * The first page is the TBA used for the CWSR ISA code. The second
- * page is used as TMA for user-mode trap handler setup in daisy-chain mode.
+ * The first chunk is the TBA used for the CWSR ISA code. The second
+ * chunk is used as TMA for user-mode trap handler setup in daisy-chain mode.
  */
 #define KFD_CWSR_TBA_TMA_SIZE (PAGE_SIZE * 2)
-#define KFD_CWSR_TMA_OFFSET PAGE_SIZE
+#define KFD_CWSR_TMA_OFFSET (PAGE_SIZE + 2048)
 
 #define KFD_MAX_NUM_OF_QUEUES_PER_DEVICE		\
 	(KFD_MAX_NUM_OF_PROCESSES *			\
@@ -111,7 +111,16 @@
 
 #define KFD_KERNEL_QUEUE_SIZE 2048
 
-#define KFD_UNMAP_LATENCY_MS	(4000)
+/*  KFD_UNMAP_LATENCY_MS is the timeout CP waiting for SDMA preemption. One XCC
+ *  can be associated to 2 SDMA engines. queue_preemption_timeout_ms is the time
+ *  driver waiting for CP returning the UNMAP_QUEUE fence. Thus the math is
+ *  queue_preemption_timeout_ms = sdma_preemption_time * 2 + cp workload
+ *  The format here makes CP workload 10% of total timeout
+ */
+#define KFD_UNMAP_LATENCY_MS	\
+	((queue_preemption_timeout_ms - queue_preemption_timeout_ms / 10) >> 1)
+
+#define KFD_MAX_SDMA_QUEUES	128
 
 /*
  * 512 = 0x200
@@ -173,12 +182,6 @@ extern int send_sigterm;
  */
 extern int debug_largebar;
 
-/*
- * Ignore CRAT table during KFD initialization, can be used to work around
- * broken CRAT tables on some AMD systems
- */
-extern int ignore_crat;
-
 /* Set sh_mem_config.retry_disable on GFX v9 */
 extern int amdgpu_noretry;
 
@@ -199,22 +202,28 @@ extern int amdgpu_no_queue_eviction_on_vm_fault;
 /* Enable eviction debug messages */
 extern bool debug_evictions;
 
+extern struct mutex kfd_processes_mutex;
+
 enum cache_policy {
 	cache_policy_coherent,
 	cache_policy_noncoherent
 };
 
-#define KFD_GC_VERSION(dev) ((dev)->adev->ip_versions[GC_HWIP][0])
+#define KFD_GC_VERSION(dev) (amdgpu_ip_version((dev)->adev, GC_HWIP, 0))
 #define KFD_IS_SOC15(dev)   ((KFD_GC_VERSION(dev)) >= (IP_VERSION(9, 0, 1)))
 #define KFD_SUPPORT_XNACK_PER_PROCESS(dev)\
 	((KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 2)) ||	\
-	 (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 3)))
+	 (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 3)) ||	\
+	 (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 4)) ||	\
+	 (KFD_GC_VERSION(dev) == IP_VERSION(9, 5, 0)))
+
+struct kfd_node;
 
 struct kfd_event_interrupt_class {
-	bool (*interrupt_isr)(struct kfd_dev *dev,
+	bool (*interrupt_isr)(struct kfd_node *dev,
 			const uint32_t *ih_ring_entry, uint32_t *patched_ihre,
 			bool *patched_flag);
-	void (*interrupt_wq)(struct kfd_dev *dev,
+	void (*interrupt_wq)(struct kfd_node *dev,
 			const uint32_t *ih_ring_entry);
 };
 
@@ -228,16 +237,15 @@ struct kfd_device_info {
 	uint8_t num_of_watch_points;
 	uint16_t mqd_size_aligned;
 	bool supports_cwsr;
-	bool needs_iommu_device;
 	bool needs_pci_atomics;
 	uint32_t no_atomic_fw_version;
 	unsigned int num_sdma_queues_per_engine;
 	unsigned int num_reserved_sdma_queues_per_engine;
-	uint64_t reserved_sdma_queues_bitmap;
+	DECLARE_BITMAP(reserved_sdma_queues_bitmap, KFD_MAX_SDMA_QUEUES);
 };
 
-unsigned int kfd_get_num_sdma_engines(struct kfd_dev *kdev);
-unsigned int kfd_get_num_xgmi_sdma_engines(struct kfd_dev *kdev);
+unsigned int kfd_get_num_sdma_engines(struct kfd_node *kdev);
+unsigned int kfd_get_num_xgmi_sdma_engines(struct kfd_node *kdev);
 
 struct kfd_mem_obj {
 	uint32_t range_start;
@@ -253,34 +261,80 @@ struct kfd_vmid_info {
 	uint32_t vmid_num_kfd;
 };
 
+#define MAX_KFD_NODES	8
+
+struct kfd_dev;
+
+struct kfd_node {
+	unsigned int node_id;
+	struct amdgpu_device *adev;     /* Duplicated here along with keeping
+					 * a copy in kfd_dev to save a hop
+					 */
+	const struct kfd2kgd_calls *kfd2kgd; /* Duplicated here along with
+					      * keeping a copy in kfd_dev to
+					      * save a hop
+					      */
+	struct kfd_vmid_info vm_info;
+	unsigned int id;                /* topology stub index */
+	uint32_t xcc_mask; /* Instance mask of XCCs present */
+	struct amdgpu_xcp *xcp;
+
+	/* Interrupts */
+	struct kfifo ih_fifo;
+	struct work_struct interrupt_work;
+	spinlock_t interrupt_lock;
+
+	/*
+	 * Interrupts of interest to KFD are copied
+	 * from the HW ring into a SW ring.
+	 */
+	bool interrupts_active;
+	uint32_t interrupt_bitmap; /* Only used for GFX 9.4.3 */
+
+	/* QCM Device instance */
+	struct device_queue_manager *dqm;
+
+	/* Global GWS resource shared between processes */
+	void *gws;
+
+	/* Clients watching SMI events */
+	struct list_head smi_clients;
+	spinlock_t smi_lock;
+	uint32_t reset_seq_num;
+
+	/* SRAM ECC flag */
+	atomic_t sram_ecc_flag;
+
+	/*spm process id */
+	unsigned int spm_pasid;
+
+	/* Maximum process number mapped to HW scheduler */
+	unsigned int max_proc_per_quantum;
+
+	unsigned int compute_vmid_bitmap;
+
+	struct kfd_local_mem_info local_mem_info;
+
+	struct kfd_dev *kfd;
+
+	/* Track per device allocated watch points */
+	uint32_t alloc_watch_ids;
+	spinlock_t watch_points_lock;
+};
+
 struct kfd_dev {
 	struct amdgpu_device *adev;
 
 	struct kfd_device_info device_info;
 
-	unsigned int id;		/* topology stub index */
-
-	phys_addr_t doorbell_base;	/* Start of actual doorbells used by
-					 * KFD. It is aligned for mapping
-					 * into user mode
-					 */
-	size_t doorbell_base_dw_offset;	/* Offset from the start of the PCI
-					 * doorbell BAR to the first KFD
-					 * doorbell in dwords. GFX reserves
-					 * the segment before this offset.
-					 */
 	u32 __iomem *doorbell_kernel_ptr; /* This is a pointer for a doorbells
 					   * page used by kernel queue
 					   */
 
 	struct kgd2kfd_shared_resources shared_resources;
-	struct kfd_vmid_info vm_info;
-	struct kfd_local_mem_info local_mem_info;
 
 	const struct kfd2kgd_calls *kfd2kgd;
 	struct mutex doorbell_mutex;
-	DECLARE_BITMAP(doorbell_available_index,
-			KFD_MAX_NUM_OF_QUEUES_PER_PROCESS);
 
 	void *gtt_mem;
 	uint64_t gtt_start_gpu_addr;
@@ -290,30 +344,13 @@ struct kfd_dev {
 	unsigned int gtt_sa_chunk_size;
 	unsigned int gtt_sa_num_of_chunks;
 
-	/* Interrupts */
-	struct kfifo ih_fifo;
-	struct workqueue_struct *ih_wq;
-	struct work_struct interrupt_work;
-	spinlock_t interrupt_lock;
-
-	/* QCM Device instance */
-	struct device_queue_manager *dqm;
-
 	bool init_complete;
-	/*
-	 * Interrupts of interest to KFD are copied
-	 * from the HW ring into a SW ring.
-	 */
-	bool interrupts_active;
 
 	/* Firmware versions */
 	uint16_t mec_fw_version;
 	uint16_t mec2_fw_version;
 	uint16_t sdma_fw_version;
 
-	/* Maximum process number mapped to HW scheduler */
-	unsigned int max_proc_per_quantum;
-
 	/* CWSR */
 	bool cwsr_enabled;
 	const void *cwsr_isa;
@@ -324,31 +361,29 @@ struct kfd_dev {
 
 	bool pci_atomic_requested;
 
-	/* Use IOMMU v2 flag */
-	bool use_iommu_v2;
-
-	/* SRAM ECC flag */
-	atomic_t sram_ecc_flag;
-
 	/* Compute Profile ref. count */
 	atomic_t compute_profile;
 
-	/* Global GWS resource shared between processes */
-	void *gws;
-
-	/* Clients watching SMI events */
-	struct list_head smi_clients;
-	spinlock_t smi_lock;
-
-	uint32_t reset_seq_num;
-
 	struct ida doorbell_ida;
 	unsigned int max_doorbell_slices;
 
 	int noretry;
 
-	/* HMM page migration MEMORY_DEVICE_PRIVATE mapping */
-	struct dev_pagemap pgmap;
+	struct kfd_node *nodes[MAX_KFD_NODES];
+	unsigned int num_nodes;
+
+	struct workqueue_struct *ih_wq;
+
+	/* Kernel doorbells for KFD device */
+	struct amdgpu_bo *doorbells;
+
+	/* bitmap for dynamic doorbell allocation from doorbell object */
+	unsigned long *doorbell_bitmap;
+
+	/* for dynamic partitioning */
+	int kfd_dev_lock;
+
+	atomic_t kfd_processes_count;
 };
 
 enum kfd_mempool {
@@ -392,13 +427,16 @@ enum kfd_unmap_queues_filter {
  * @KFD_QUEUE_TYPE_DIQ: DIQ queue type.
  *
  * @KFD_QUEUE_TYPE_SDMA_XGMI: Special SDMA queue for XGMI interface.
+ *
+ * @KFD_QUEUE_TYPE_SDMA_BY_ENG_ID:  SDMA user mode queue with target SDMA engine ID.
  */
 enum kfd_queue_type  {
 	KFD_QUEUE_TYPE_COMPUTE,
 	KFD_QUEUE_TYPE_SDMA,
 	KFD_QUEUE_TYPE_HIQ,
 	KFD_QUEUE_TYPE_DIQ,
-	KFD_QUEUE_TYPE_SDMA_XGMI
+	KFD_QUEUE_TYPE_SDMA_XGMI,
+	KFD_QUEUE_TYPE_SDMA_BY_ENG_ID
 };
 
 enum kfd_queue_format {
@@ -472,14 +510,19 @@ struct queue_properties {
 	uint64_t  queue_size;
 	uint32_t priority;
 	uint32_t queue_percent;
-	uint32_t *read_ptr;
-	uint32_t *write_ptr;
+	void __user *read_ptr;
+	void __user *write_ptr;
 	void __iomem *doorbell_ptr;
 	uint32_t doorbell_off;
 	bool is_interop;
 	bool is_evicted;
+	bool is_suspended;
+	bool is_being_destroyed;
 	bool is_active;
 	bool is_gws;
+	uint32_t pm4_target_xcc;
+	bool is_dbg_wa;
+	bool is_user_cu_masked;
 	/* Not relevant for user mode queues in cp scheduling */
 	unsigned int vmid;
 	/* Relevant only for sdma queues*/
@@ -494,15 +537,25 @@ struct queue_properties {
 	uint32_t ctl_stack_size;
 	uint64_t tba_addr;
 	uint64_t tma_addr;
+	uint64_t exception_status;
+
+	struct amdgpu_bo *wptr_bo;
+	struct amdgpu_bo *rptr_bo;
+	struct amdgpu_bo *ring_bo;
+	struct amdgpu_bo *eop_buf_bo;
+	struct amdgpu_bo *cwsr_bo;
 };
 
 #define QUEUE_IS_ACTIVE(q) ((q).queue_size > 0 &&	\
 			    (q).queue_address != 0 &&	\
 			    (q).queue_percent > 0 &&	\
-			    !(q).is_evicted)
+			    !(q).is_evicted &&		\
+			    !(q).is_suspended)
 
 enum mqd_update_flag {
-	UPDATE_FLAG_CU_MASK = 0,
+	UPDATE_FLAG_DBG_WA_ENABLE = 1,
+	UPDATE_FLAG_DBG_WA_DISABLE = 2,
+	UPDATE_FLAG_IS_GWS = 4, /* quirk for gfx9 IP */
 };
 
 struct mqd_update_info {
@@ -563,7 +616,7 @@ struct queue {
 	unsigned int doorbell_id;
 
 	struct kfd_process	*process;
-	struct kfd_dev		*device;
+	struct kfd_node		*device;
 	void *gws;
 
 	/* procfs */
@@ -573,7 +626,7 @@ struct queue {
 	uint64_t gang_ctx_gpu_addr;
 	void *gang_ctx_cpu_ptr;
 
-	struct amdgpu_bo *wptr_bo;
+	struct amdgpu_bo *wptr_bo_gart;
 };
 
 enum KFD_MQD_TYPE {
@@ -658,7 +711,10 @@ struct qcm_process_device {
 	uint64_t ib_base;
 	void *ib_kaddr;
 
-	/* doorbell resources per process per device */
+	/* doorbells for kfd process */
+	struct amdgpu_bo *proc_doorbells;
+
+	/* bitmap for dynamic doorbell allocation from the bo */
 	unsigned long *doorbell_bitmap;
 };
 
@@ -697,7 +753,7 @@ enum kfd_pdd_bound {
 /* Data that is per-process-per device. */
 struct kfd_process_device {
 	/* The device that owns this data. */
-	struct kfd_dev *dev;
+	struct kfd_node *dev;
 
 	/* The process that owns this kfd_process_device. */
 	struct kfd_process *process;
@@ -716,7 +772,6 @@ struct kfd_process_device {
 	/* VM context for GPUVM allocations */
 	struct file *drm_file;
 	void *drm_priv;
-	atomic64_t tlb_seq;
 
 	/* GPUVM allocations storage */
 	struct idr alloc_idr;
@@ -733,7 +788,7 @@ struct kfd_process_device {
 	enum kfd_pdd_bound bound;
 
 	/* VRAM usage */
-	uint64_t vram_usage;
+	atomic64_t vram_usage;
 	struct attribute attr_vram;
 	char vram_filename[MAX_SYSFS_FILENAME_LEN];
 
@@ -748,7 +803,6 @@ struct kfd_process_device {
 	struct attribute attr_evict;
 
 	struct kobject *kobj_stats;
-	unsigned int doorbell_index;
 
 	/*
 	 * @cu_occupancy: Reports occupancy of Compute Units (CU) of a process
@@ -783,6 +837,18 @@ struct kfd_process_device {
 	uint64_t faults;
 	uint64_t page_in;
 	uint64_t page_out;
+
+	/* Exception code status*/
+	uint64_t exception_status;
+	void *vm_fault_exc_data;
+	size_t vm_fault_exc_data_size;
+
+	/* Tracks debug per-vmid request settings */
+	uint32_t spi_dbg_override;
+	uint32_t spi_dbg_launch_mode;
+	uint32_t watch_points[4];
+	uint32_t alloc_watch_ids;
+
 	/*
 	 * If this process has been checkpointed before, then the user
 	 * application will use the original gpu_id on the
@@ -793,6 +859,11 @@ struct kfd_process_device {
 	void *proc_ctx_bo;
 	uint64_t proc_ctx_gpu_addr;
 	void *proc_ctx_cpu_ptr;
+
+	/* Tracks queue reset status */
+	bool has_reset_queue;
+
+	u32 pasid;
 };
 
 #define qpd_to_pdd(x) container_of(x, struct kfd_process_device, qpd)
@@ -810,6 +881,14 @@ struct svm_range_list {
 	struct delayed_work		restore_work;
 	DECLARE_BITMAP(bitmap_supported, MAX_GPU_INSTANCE);
 	struct task_struct		*faulting_task;
+	/* check point ts decides if page fault recovery need be dropped */
+	uint64_t			checkpoint_ts[MAX_GPU_INSTANCE];
+
+	/* Default granularity to use in buffer migration
+	 * and restoration of backing memory while handling
+	 * recoverable page faults
+	 */
+	uint8_t default_granularity;
 };
 
 /* Process data */
@@ -844,8 +923,6 @@ struct kfd_process {
 	/* We want to receive a notification when the mm_struct is destroyed */
 	struct mmu_notifier mmu_notifier;
 
-	u32 pasid;
-
 	/*
 	 * Array of kfd_process_device pointers,
 	 * one for each device the process is using.
@@ -875,7 +952,7 @@ struct kfd_process {
 	 * fence will be triggered during eviction and new one will be created
 	 * during restore
 	 */
-	struct dma_fence *ef;
+	struct dma_fence __rcu *ef;
 
 	/* Work items for evicting and restoring BOs */
 	struct delayed_work eviction_work;
@@ -887,19 +964,60 @@ struct kfd_process {
 	 */
 	unsigned long last_restore_timestamp;
 
+	/* Indicates device process is debug attached with reserved vmid. */
+	bool debug_trap_enabled;
+
+	/* per-process-per device debug event fd file */
+	struct file *dbg_ev_file;
+
+	/* If the process is a kfd debugger, we need to know so we can clean
+	 * up at exit time.  If a process enables debugging on itself, it does
+	 * its own clean-up, so we don't set the flag here.  We track this by
+	 * counting the number of processes this process is debugging.
+	 */
+	atomic_t debugged_process_count;
+
+	/* If the process is a debugged, this is the debugger process */
+	struct kfd_process *debugger_process;
+
 	/* Kobj for our procfs */
 	struct kobject *kobj;
 	struct kobject *kobj_queues;
 	struct attribute attr_pasid;
 
+	/* Keep track cwsr init */
+	bool has_cwsr;
+
+	/* Exception code enable mask and status */
+	uint64_t exception_enable_mask;
+	uint64_t exception_status;
+
+	/* Used to drain stale interrupts */
+	wait_queue_head_t wait_irq_drain;
+	bool irq_drain_is_open;
+
 	/* shared virtual memory registered by this process */
 	struct svm_range_list svms;
 
 	bool xnack_enabled;
 
+	/* Work area for debugger event writer worker. */
+	struct work_struct debug_event_workarea;
+
+	/* Tracks debug per-vmid request for debug flags */
+	u32 dbg_flags;
+
 	atomic_t poison;
 	/* Queues are in paused stated because we are in the process of doing a CRIU checkpoint */
 	bool queues_paused;
+
+	/* Tracks runtime enable status */
+	struct semaphore runtime_enable_sema;
+	bool is_runtime_retry;
+	struct kfd_runtime_info runtime_info;
+
+	/* if gpu page fault sent to KFD */
+	bool gpu_page_fault;
 };
 
 #define KFD_PROCESS_TABLE_SIZE 5 /* bits: 32 entries */
@@ -925,20 +1043,20 @@ struct amdkfd_ioctl_desc {
 	unsigned int cmd_drv;
 	const char *name;
 };
-bool kfd_dev_is_large_bar(struct kfd_dev *dev);
+bool kfd_dev_is_large_bar(struct kfd_node *dev);
 
 int kfd_process_create_wq(void);
 void kfd_process_destroy_wq(void);
 void kfd_cleanup_processes(void);
-struct kfd_process *kfd_create_process(struct file *filep);
+struct kfd_process *kfd_create_process(struct task_struct *thread);
 struct kfd_process *kfd_get_process(const struct task_struct *task);
-struct kfd_process *kfd_lookup_process_by_pasid(u32 pasid);
+struct kfd_process *kfd_lookup_process_by_pasid(u32 pasid,
+						 struct kfd_process_device **pdd);
 struct kfd_process *kfd_lookup_process_by_mm(const struct mm_struct *mm);
 
 int kfd_process_gpuidx_from_gpuid(struct kfd_process *p, uint32_t gpu_id);
-int kfd_process_gpuid_from_adev(struct kfd_process *p,
-			       struct amdgpu_device *adev, uint32_t *gpuid,
-			       uint32_t *gpuidx);
+int kfd_process_gpuid_from_node(struct kfd_process *p, struct kfd_node *node,
+				uint32_t *gpuid, uint32_t *gpuidx);
 static inline int kfd_process_gpuid_from_gpuidx(struct kfd_process *p,
 				uint32_t gpuidx, uint32_t *gpuid) {
 	return gpuidx < p->n_pdds ? p->pdds[gpuidx]->dev->id : -EINVAL;
@@ -961,16 +1079,16 @@ int kfd_process_get_user_gpu_id(struct kfd_process *p, uint32_t actual_gpu_id);
 
 int kfd_process_device_init_vm(struct kfd_process_device *pdd,
 			       struct file *drm_file);
-struct kfd_process_device *kfd_bind_process_to_device(struct kfd_dev *dev,
+struct kfd_process_device *kfd_bind_process_to_device(struct kfd_node *dev,
 						struct kfd_process *p);
-struct kfd_process_device *kfd_get_process_device_data(struct kfd_dev *dev,
+struct kfd_process_device *kfd_get_process_device_data(struct kfd_node *dev,
 							struct kfd_process *p);
-struct kfd_process_device *kfd_create_process_device_data(struct kfd_dev *dev,
+struct kfd_process_device *kfd_create_process_device_data(struct kfd_node *dev,
 							struct kfd_process *p);
 
 bool kfd_process_xnack_mode(struct kfd_process *p, bool supported);
 
-int kfd_reserved_mem_mmap(struct kfd_dev *dev, struct kfd_process *process,
+int kfd_reserved_mem_mmap(struct kfd_node *dev, struct kfd_process *process,
 			  struct vm_area_struct *vma);
 
 /* KFD process API for creating and translating handles */
@@ -985,8 +1103,6 @@ struct kfd_process *kfd_lookup_process_by_pid(struct pid *pid);
 /* PASIDs */
 int kfd_pasid_init(void);
 void kfd_pasid_exit(void);
-bool kfd_set_pasid_limit(unsigned int new_limit);
-unsigned int kfd_get_pasid_limit(void);
 u32 kfd_pasid_alloc(void);
 void kfd_pasid_free(u32 pasid);
 
@@ -994,7 +1110,7 @@ void kfd_pasid_free(u32 pasid);
 size_t kfd_doorbell_process_slice(struct kfd_dev *kfd);
 int kfd_doorbell_init(struct kfd_dev *kfd);
 void kfd_doorbell_fini(struct kfd_dev *kfd);
-int kfd_doorbell_mmap(struct kfd_dev *dev, struct kfd_process *process,
+int kfd_doorbell_mmap(struct kfd_node *dev, struct kfd_process *process,
 		      struct vm_area_struct *vma);
 void __iomem *kfd_get_kernel_doorbell(struct kfd_dev *kfd,
 					unsigned int *doorbell_off);
@@ -1007,15 +1123,15 @@ unsigned int kfd_get_doorbell_dw_offset_in_bar(struct kfd_dev *kfd,
 					unsigned int doorbell_id);
 phys_addr_t kfd_get_process_doorbells(struct kfd_process_device *pdd);
 int kfd_alloc_process_doorbells(struct kfd_dev *kfd,
-				unsigned int *doorbell_index);
+				struct kfd_process_device *pdd);
 void kfd_free_process_doorbells(struct kfd_dev *kfd,
-				unsigned int doorbell_index);
+				struct kfd_process_device *pdd);
 /* GTT Sub-Allocator */
 
-int kfd_gtt_sa_allocate(struct kfd_dev *kfd, unsigned int size,
+int kfd_gtt_sa_allocate(struct kfd_node *node, unsigned int size,
 			struct kfd_mem_obj **mem_obj);
 
-int kfd_gtt_sa_free(struct kfd_dev *kfd, struct kfd_mem_obj *mem_obj);
+int kfd_gtt_sa_free(struct kfd_node *node, struct kfd_mem_obj *mem_obj);
 
 extern struct device *kfd_device;
 
@@ -1028,27 +1144,53 @@ void kfd_procfs_del_queue(struct queue *q);
 /* Topology */
 int kfd_topology_init(void);
 void kfd_topology_shutdown(void);
-int kfd_topology_add_device(struct kfd_dev *gpu);
-int kfd_topology_remove_device(struct kfd_dev *gpu);
+int kfd_topology_add_device(struct kfd_node *gpu);
+int kfd_topology_remove_device(struct kfd_node *gpu);
 struct kfd_topology_device *kfd_topology_device_by_proximity_domain(
 						uint32_t proximity_domain);
 struct kfd_topology_device *kfd_topology_device_by_proximity_domain_no_lock(
 						uint32_t proximity_domain);
 struct kfd_topology_device *kfd_topology_device_by_id(uint32_t gpu_id);
-struct kfd_dev *kfd_device_by_id(uint32_t gpu_id);
-struct kfd_dev *kfd_device_by_pci_dev(const struct pci_dev *pdev);
-struct kfd_dev *kfd_device_by_adev(const struct amdgpu_device *adev);
-int kfd_topology_enum_kfd_devices(uint8_t idx, struct kfd_dev **kdev);
+struct kfd_node *kfd_device_by_id(uint32_t gpu_id);
+static inline bool kfd_irq_is_from_node(struct kfd_node *node, uint32_t node_id,
+					uint32_t vmid)
+{
+	return (node->interrupt_bitmap & (1 << node_id)) != 0 &&
+	       (node->compute_vmid_bitmap & (1 << vmid)) != 0;
+}
+static inline struct kfd_node *kfd_node_by_irq_ids(struct amdgpu_device *adev,
+					uint32_t node_id, uint32_t vmid) {
+	struct kfd_dev *dev = adev->kfd.dev;
+	uint32_t i;
+
+	if (KFD_GC_VERSION(dev) != IP_VERSION(9, 4, 3) &&
+	    KFD_GC_VERSION(dev) != IP_VERSION(9, 4, 4) &&
+	    KFD_GC_VERSION(dev) != IP_VERSION(9, 5, 0))
+		return dev->nodes[0];
+
+	for (i = 0; i < dev->num_nodes; i++)
+		if (kfd_irq_is_from_node(dev->nodes[i], node_id, vmid))
+			return dev->nodes[i];
+
+	return NULL;
+}
+int kfd_topology_enum_kfd_devices(uint8_t idx, struct kfd_node **kdev);
 int kfd_numa_node_to_apic_id(int numa_node_id);
-void kfd_double_confirm_iommu_support(struct kfd_dev *gpu);
 
 /* Interrupts */
-int kfd_interrupt_init(struct kfd_dev *dev);
-void kfd_interrupt_exit(struct kfd_dev *dev);
-bool enqueue_ih_ring_entry(struct kfd_dev *kfd,	const void *ih_ring_entry);
-bool interrupt_is_wanted(struct kfd_dev *dev,
+#define	KFD_IRQ_FENCE_CLIENTID	0xff
+#define	KFD_IRQ_FENCE_SOURCEID	0xff
+#define	KFD_IRQ_IS_FENCE(client, source)				\
+				((client) == KFD_IRQ_FENCE_CLIENTID &&	\
+				(source) == KFD_IRQ_FENCE_SOURCEID)
+int kfd_interrupt_init(struct kfd_node *dev);
+void kfd_interrupt_exit(struct kfd_node *dev);
+bool enqueue_ih_ring_entry(struct kfd_node *kfd, const void *ih_ring_entry);
+bool interrupt_is_wanted(struct kfd_node *dev,
 				const uint32_t *ih_ring_entry,
 				uint32_t *patched_ihre, bool *flag);
+int kfd_process_drain_interrupts(struct kfd_process_device *pdd);
+void kfd_process_close_interrupt_drain(unsigned int pasid);
 
 /* amdkfd Apertures */
 int kfd_init_apertures(struct kfd_process *process);
@@ -1056,6 +1198,11 @@ int kfd_init_apertures(struct kfd_process *process);
 void kfd_process_set_trap_handler(struct qcm_process_device *qpd,
 				  uint64_t tba_addr,
 				  uint64_t tma_addr);
+void kfd_process_set_trap_debug_flag(struct qcm_process_device *qpd,
+				     bool enabled);
+
+/* CWSR initialization */
+int kfd_process_init_cwsr_apu(struct kfd_process *process, struct file *filep);
 
 /* CRIU */
 /*
@@ -1172,27 +1319,35 @@ int init_queue(struct queue **q, const struct queue_properties *properties);
 void uninit_queue(struct queue *q);
 void print_queue_properties(struct queue_properties *q);
 void print_queue(struct queue *q);
+int kfd_queue_buffer_get(struct amdgpu_vm *vm, void __user *addr, struct amdgpu_bo **pbo,
+			 u64 expected_size);
+void kfd_queue_buffer_put(struct amdgpu_bo **bo);
+int kfd_queue_acquire_buffers(struct kfd_process_device *pdd, struct queue_properties *properties);
+int kfd_queue_release_buffers(struct kfd_process_device *pdd, struct queue_properties *properties);
+void kfd_queue_unref_bo_va(struct amdgpu_vm *vm, struct amdgpu_bo **bo);
+int kfd_queue_unref_bo_vas(struct kfd_process_device *pdd,
+			   struct queue_properties *properties);
+void kfd_queue_ctx_save_restore_size(struct kfd_topology_device *dev);
 
 struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type,
-		struct kfd_dev *dev);
-struct mqd_manager *mqd_manager_init_cik_hawaii(enum KFD_MQD_TYPE type,
-		struct kfd_dev *dev);
+		struct kfd_node *dev);
 struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type,
-		struct kfd_dev *dev);
-struct mqd_manager *mqd_manager_init_vi_tonga(enum KFD_MQD_TYPE type,
-		struct kfd_dev *dev);
+		struct kfd_node *dev);
 struct mqd_manager *mqd_manager_init_v9(enum KFD_MQD_TYPE type,
-		struct kfd_dev *dev);
+		struct kfd_node *dev);
 struct mqd_manager *mqd_manager_init_v10(enum KFD_MQD_TYPE type,
-		struct kfd_dev *dev);
+		struct kfd_node *dev);
 struct mqd_manager *mqd_manager_init_v11(enum KFD_MQD_TYPE type,
-		struct kfd_dev *dev);
-struct device_queue_manager *device_queue_manager_init(struct kfd_dev *dev);
+		struct kfd_node *dev);
+struct mqd_manager *mqd_manager_init_v12(enum KFD_MQD_TYPE type,
+		struct kfd_node *dev);
+struct device_queue_manager *device_queue_manager_init(struct kfd_node *dev);
 void device_queue_manager_uninit(struct device_queue_manager *dqm);
-struct kernel_queue *kernel_queue_init(struct kfd_dev *dev,
+struct kernel_queue *kernel_queue_init(struct kfd_node *dev,
 					enum kfd_queue_type type);
-void kernel_queue_uninit(struct kernel_queue *kq, bool hanging);
-int kfd_dqm_evict_pasid(struct device_queue_manager *dqm, u32 pasid);
+void kernel_queue_uninit(struct kernel_queue *kq);
+int kfd_evict_process_device(struct kfd_process_device *pdd);
+int kfd_dqm_suspend_bad_queue_mes(struct kfd_node *knode, u32 pasid, u32 doorbell_id);
 
 /* Process Queue Manager */
 struct process_queue_node {
@@ -1206,11 +1361,9 @@ void kfd_process_dequeue_from_all_devices(struct kfd_process *p);
 int pqm_init(struct process_queue_manager *pqm, struct kfd_process *p);
 void pqm_uninit(struct process_queue_manager *pqm);
 int pqm_create_queue(struct process_queue_manager *pqm,
-			    struct kfd_dev *dev,
-			    struct file *f,
+			    struct kfd_node *dev,
 			    struct queue_properties *properties,
 			    unsigned int *qid,
-			    struct amdgpu_bo *wptr_bo,
 			    const struct kfd_criu_queue_priv_data *q_data,
 			    const void *restore_mqd,
 			    const void *restore_ctl_stack,
@@ -1222,8 +1375,6 @@ int pqm_update_mqd(struct process_queue_manager *pqm, unsigned int qid,
 			struct mqd_update_info *minfo);
 int pqm_set_gws(struct process_queue_manager *pqm, unsigned int qid,
 			void *gws);
-struct kernel_queue *pqm_get_kernel_queue(struct process_queue_manager *pqm,
-						unsigned int qid);
 struct queue *pqm_get_user_queue(struct process_queue_manager *pqm,
 						unsigned int qid);
 int pqm_get_wave_state(struct process_queue_manager *pqm,
@@ -1231,8 +1382,13 @@ int pqm_get_wave_state(struct process_queue_manager *pqm,
 		       void __user *ctl_stack,
 		       u32 *ctl_stack_used_size,
 		       u32 *save_area_used_size);
+int pqm_get_queue_snapshot(struct process_queue_manager *pqm,
+			   uint64_t exception_clear_mask,
+			   void __user *buf,
+			   int *num_qss_entries,
+			   uint32_t *entry_size);
 
-int amdkfd_fence_wait_timeout(uint64_t *fence_addr,
+int amdkfd_fence_wait_timeout(struct device_queue_manager *dqm,
 			      uint64_t fence_value,
 			      unsigned int timeout_ms);
 
@@ -1245,6 +1401,24 @@ int pqm_get_queue_checkpoint_info(struct process_queue_manager *pqm,
 #define KFD_FENCE_COMPLETED (100)
 #define KFD_FENCE_INIT   (10)
 
+/**
+ * enum kfd_config_dequeue_wait_counts_cmd - Command for configuring
+ *  dequeue wait counts.
+ *
+ * @KFD_DEQUEUE_WAIT_INIT: Set optimized dequeue wait counts for a
+ *	certain ASICs. For these ASICs, this is default value used by RESET
+ * @KFD_DEQUEUE_WAIT_RESET: Reset dequeue wait counts to the optimized value
+ *	for certain ASICs. For others set it to default hardware reset value
+ * @KFD_DEQUEUE_WAIT_SET_SCH_WAVE: Set context switch latency wait
+ *
+ */
+enum kfd_config_dequeue_wait_counts_cmd {
+	KFD_DEQUEUE_WAIT_INIT = 1,
+	KFD_DEQUEUE_WAIT_RESET = 2,
+	KFD_DEQUEUE_WAIT_SET_SCH_WAVE = 3
+};
+
+
 struct packet_manager {
 	struct device_queue_manager *dqm;
 	struct kernel_queue *priv_queue;
@@ -1270,6 +1444,8 @@ struct packet_manager_funcs {
 	int (*unmap_queues)(struct packet_manager *pm, uint32_t *buffer,
 			enum kfd_unmap_queues_filter mode,
 			uint32_t filter_param, bool reset);
+	int (*config_dequeue_wait_counts)(struct packet_manager *pm, uint32_t *buffer,
+			enum kfd_config_dequeue_wait_counts_cmd cmd, uint32_t value);
 	int (*query_status)(struct packet_manager *pm, uint32_t *buffer,
 			uint64_t fence_address,	uint64_t fence_value);
 	int (*release_mem)(uint64_t gpu_addr, uint32_t *buffer);
@@ -1280,6 +1456,7 @@ struct packet_manager_funcs {
 	int set_resources_size;
 	int map_queues_size;
 	int unmap_queues_size;
+	int config_dequeue_wait_counts_size;
 	int query_status_size;
 	int release_mem_size;
 };
@@ -1289,7 +1466,7 @@ extern const struct packet_manager_funcs kfd_v9_pm_funcs;
 extern const struct packet_manager_funcs kfd_aldebaran_pm_funcs;
 
 int pm_init(struct packet_manager *pm, struct device_queue_manager *dqm);
-void pm_uninit(struct packet_manager *pm, bool hanging);
+void pm_uninit(struct packet_manager *pm);
 int pm_send_set_resources(struct packet_manager *pm,
 				struct scheduling_resources *res);
 int pm_send_runlist(struct packet_manager *pm, struct list_head *dqm_queues);
@@ -1302,6 +1479,10 @@ int pm_send_unmap_queue(struct packet_manager *pm,
 
 void pm_release_ib(struct packet_manager *pm);
 
+int pm_config_dequeue_wait_counts(struct packet_manager *pm,
+			enum kfd_config_dequeue_wait_counts_cmd cmd,
+			uint32_t wait_counts_config);
+
 /* Following PM funcs can be shared among VI and AI */
 unsigned int pm_build_pm4_header(unsigned int opcode, size_t packet_size);
 
@@ -1310,6 +1491,8 @@ uint64_t kfd_get_number_elems(struct kfd_dev *kfd);
 /* Events */
 extern const struct kfd_event_interrupt_class event_interrupt_class_cik;
 extern const struct kfd_event_interrupt_class event_interrupt_class_v9;
+extern const struct kfd_event_interrupt_class event_interrupt_class_v9_4_3;
+extern const struct kfd_event_interrupt_class event_interrupt_class_v10;
 extern const struct kfd_event_interrupt_class event_interrupt_class_v11;
 
 extern const struct kfd_device_global_init_class device_global_init_class_cik;
@@ -1323,9 +1506,6 @@ int kfd_wait_on_events(struct kfd_process *p,
 		       uint32_t *wait_result);
 void kfd_signal_event_interrupt(u32 pasid, uint32_t partial_id,
 				uint32_t valid_id_bits);
-void kfd_signal_iommu_event(struct kfd_dev *dev,
-			    u32 pasid, unsigned long address,
-			    bool is_write_requested, bool is_execute_requested);
 void kfd_signal_hw_exception_event(u32 pasid);
 int kfd_set_event(struct kfd_process *p, uint32_t event_id);
 int kfd_reset_event(struct kfd_process *p, uint32_t event_id);
@@ -1339,35 +1519,52 @@ int kfd_event_create(struct file *devkfd, struct kfd_process *p,
 int kfd_get_num_events(struct kfd_process *p);
 int kfd_event_destroy(struct kfd_process *p, uint32_t event_id);
 
-void kfd_signal_vm_fault_event(struct kfd_dev *dev, u32 pasid,
-				struct kfd_vm_fault_info *info);
+void kfd_signal_vm_fault_event_with_userptr(struct kfd_process *p, uint64_t gpu_va);
 
-void kfd_signal_reset_event(struct kfd_dev *dev);
+void kfd_signal_vm_fault_event(struct kfd_process_device *pdd,
+				struct kfd_vm_fault_info *info,
+				struct kfd_hsa_memory_exception_data *data);
 
-void kfd_signal_poison_consumed_event(struct kfd_dev *dev, u32 pasid);
+void kfd_signal_reset_event(struct kfd_node *dev);
 
-void kfd_flush_tlb(struct kfd_process_device *pdd, enum TLB_FLUSH_TYPE type);
+void kfd_signal_poison_consumed_event(struct kfd_node *dev, u32 pasid);
+
+static inline void kfd_flush_tlb(struct kfd_process_device *pdd,
+				 enum TLB_FLUSH_TYPE type)
+{
+	struct amdgpu_device *adev = pdd->dev->adev;
+	struct amdgpu_vm *vm = drm_priv_to_vm(pdd->drm_priv);
+
+	amdgpu_vm_flush_compute_tlb(adev, vm, type, pdd->dev->xcc_mask);
+}
 
 static inline bool kfd_flush_tlb_after_unmap(struct kfd_dev *dev)
 {
-	return KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 2) ||
-	       (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 1) &&
-	       dev->adev->sdma.instance[0].fw_version >= 18) ||
+	return KFD_GC_VERSION(dev) >= IP_VERSION(9, 4, 2) ||
+	       (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 1) && dev->sdma_fw_version >= 18) ||
 	       KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 0);
 }
 
-bool kfd_is_locked(void);
+int kfd_send_exception_to_runtime(struct kfd_process *p,
+				unsigned int queue_id,
+				uint64_t error_reason);
+bool kfd_is_locked(struct kfd_dev *kfd);
 
 /* Compute profile */
-void kfd_inc_compute_active(struct kfd_dev *dev);
-void kfd_dec_compute_active(struct kfd_dev *dev);
+void kfd_inc_compute_active(struct kfd_node *dev);
+void kfd_dec_compute_active(struct kfd_node *dev);
 
 /* Cgroup Support */
 /* Check with device cgroup if @kfd device is accessible */
-static inline int kfd_devcgroup_check_permission(struct kfd_dev *kfd)
+static inline int kfd_devcgroup_check_permission(struct kfd_node *node)
 {
 #if defined(CONFIG_CGROUP_DEVICE) || defined(CONFIG_CGROUP_BPF)
-	struct drm_device *ddev = adev_to_drm(kfd->adev);
+	struct drm_device *ddev;
+
+	if (node->xcp)
+		ddev = node->xcp->ddev;
+	else
+		ddev = adev_to_drm(node->adev);
 
 	return devcgroup_check_permission(DEVCG_DEV_CHAR, DRM_MAJOR,
 					  ddev->render->index,
@@ -1377,6 +1574,11 @@ static inline int kfd_devcgroup_check_permission(struct kfd_dev *kfd)
 #endif
 }
 
+static inline bool kfd_is_first_node(struct kfd_node *node)
+{
+	return (node == node->kfd->nodes[0]);
+}
+
 /* Debugfs */
 #if defined(CONFIG_DEBUG_FS)
 
@@ -1389,14 +1591,19 @@ int dqm_debugfs_hqds(struct seq_file *m, void *data);
 int kfd_debugfs_rls_by_device(struct seq_file *m, void *data);
 int pm_debugfs_runlist(struct seq_file *m, void *data);
 
-int kfd_debugfs_hang_hws(struct kfd_dev *dev);
+int kfd_debugfs_hang_hws(struct kfd_node *dev);
 int pm_debugfs_hang_hws(struct packet_manager *pm);
 int dqm_debugfs_hang_hws(struct device_queue_manager *dqm);
 
+void kfd_debugfs_add_process(struct kfd_process *p);
+void kfd_debugfs_remove_process(struct kfd_process *p);
+
 #else
 
 static inline void kfd_debugfs_init(void) {}
 static inline void kfd_debugfs_fini(void) {}
+static inline void kfd_debugfs_add_process(struct kfd_process *p) {}
+static inline void kfd_debugfs_remove_process(struct kfd_process *p) {}
 
 #endif
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 07a9eaf9b7d8..a085faac9fe1 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -28,7 +28,6 @@
 #include <linux/sched/task.h>
 #include <linux/mmu_context.h>
 #include <linux/slab.h>
-#include <linux/amd-iommu.h>
 #include <linux/notifier.h>
 #include <linux/compat.h>
 #include <linux/mman.h>
@@ -36,21 +35,22 @@
 #include <linux/pm_runtime.h>
 #include "amdgpu_amdkfd.h"
 #include "amdgpu.h"
+#include "amdgpu_reset.h"
 
 struct mm_struct;
 
 #include "kfd_priv.h"
 #include "kfd_device_queue_manager.h"
-#include "kfd_iommu.h"
 #include "kfd_svm.h"
 #include "kfd_smi_events.h"
+#include "kfd_debug.h"
 
 /*
  * List of struct kfd_process (field kfd_process).
  * Unique/indexed by mm_struct*
  */
 DEFINE_HASHTABLE(kfd_processes_table, KFD_PROCESS_TABLE_SIZE);
-static DEFINE_MUTEX(kfd_processes_mutex);
+DEFINE_MUTEX(kfd_processes_mutex);
 
 DEFINE_SRCU(kfd_processes_srcu);
 
@@ -69,7 +69,6 @@ static struct kfd_process *find_process(const struct task_struct *thread,
 					bool ref);
 static void kfd_process_ref_release(struct kref *ref);
 static struct kfd_process *create_process(const struct task_struct *thread);
-static int kfd_process_init_cwsr_apu(struct kfd_process *p, struct file *filep);
 
 static void evict_process_worker(struct work_struct *work);
 static void restore_process_worker(struct work_struct *work);
@@ -269,9 +268,12 @@ static int kfd_get_cu_occupancy(struct attribute *attr, char *buffer)
 	int cu_cnt;
 	int wave_cnt;
 	int max_waves_per_cu;
-	struct kfd_dev *dev = NULL;
+	struct kfd_node *dev = NULL;
 	struct kfd_process *proc = NULL;
 	struct kfd_process_device *pdd = NULL;
+	int i;
+	struct kfd_cu_occupancy *cu_occupancy;
+	u32 queue_format;
 
 	pdd = container_of(attr, struct kfd_process_device, attr_cu_occupancy);
 	dev = pdd->dev;
@@ -281,41 +283,64 @@ static int kfd_get_cu_occupancy(struct attribute *attr, char *buffer)
 	cu_cnt = 0;
 	proc = pdd->process;
 	if (pdd->qpd.queue_count == 0) {
-		pr_debug("Gpu-Id: %d has no active queues for process %d\n",
-			 dev->id, proc->pasid);
+		pr_debug("Gpu-Id: %d has no active queues for process pid %d\n",
+			 dev->id, (int)proc->lead_thread->pid);
 		return snprintf(buffer, PAGE_SIZE, "%d\n", cu_cnt);
 	}
 
 	/* Collect wave count from device if it supports */
 	wave_cnt = 0;
 	max_waves_per_cu = 0;
-	dev->kfd2kgd->get_cu_occupancy(dev->adev, proc->pasid, &wave_cnt,
-			&max_waves_per_cu);
+
+	cu_occupancy = kcalloc(AMDGPU_MAX_QUEUES, sizeof(*cu_occupancy), GFP_KERNEL);
+	if (!cu_occupancy)
+		return -ENOMEM;
+
+	/*
+	 * For GFX 9.4.3, fetch the CU occupancy from the first XCC in the partition.
+	 * For AQL queues, because of cooperative dispatch we multiply the wave count
+	 * by number of XCCs in the partition to get the total wave counts across all
+	 * XCCs in the partition.
+	 * For PM4 queues, there is no cooperative dispatch so wave_cnt stay as it is.
+	 */
+	dev->kfd2kgd->get_cu_occupancy(dev->adev, cu_occupancy,
+			&max_waves_per_cu, ffs(dev->xcc_mask) - 1);
+
+	for (i = 0; i < AMDGPU_MAX_QUEUES; i++) {
+		if (cu_occupancy[i].wave_cnt != 0 &&
+		    kfd_dqm_is_queue_in_process(dev->dqm, &pdd->qpd,
+						cu_occupancy[i].doorbell_off,
+						&queue_format)) {
+			if (unlikely(queue_format == KFD_QUEUE_FORMAT_PM4))
+				wave_cnt += cu_occupancy[i].wave_cnt;
+			else
+				wave_cnt += (NUM_XCC(dev->xcc_mask) *
+						cu_occupancy[i].wave_cnt);
+		}
+	}
 
 	/* Translate wave count to number of compute units */
 	cu_cnt = (wave_cnt + (max_waves_per_cu - 1)) / max_waves_per_cu;
+	kfree(cu_occupancy);
 	return snprintf(buffer, PAGE_SIZE, "%d\n", cu_cnt);
 }
 
 static ssize_t kfd_procfs_show(struct kobject *kobj, struct attribute *attr,
 			       char *buffer)
 {
-	if (strcmp(attr->name, "pasid") == 0) {
-		struct kfd_process *p = container_of(attr, struct kfd_process,
-						     attr_pasid);
-
-		return snprintf(buffer, PAGE_SIZE, "%d\n", p->pasid);
-	} else if (strncmp(attr->name, "vram_", 5) == 0) {
+	if (strcmp(attr->name, "pasid") == 0)
+		return snprintf(buffer, PAGE_SIZE, "%d\n", 0);
+	else if (strncmp(attr->name, "vram_", 5) == 0) {
 		struct kfd_process_device *pdd = container_of(attr, struct kfd_process_device,
 							      attr_vram);
-		return snprintf(buffer, PAGE_SIZE, "%llu\n", READ_ONCE(pdd->vram_usage));
+		return snprintf(buffer, PAGE_SIZE, "%llu\n", atomic64_read(&pdd->vram_usage));
 	} else if (strncmp(attr->name, "sdma_", 5) == 0) {
 		struct kfd_process_device *pdd = container_of(attr, struct kfd_process_device,
 							      attr_sdma);
 		struct kfd_sdma_activity_handler_workarea sdma_activity_work_handler;
 
-		INIT_WORK(&sdma_activity_work_handler.sdma_activity_work,
-					kfd_sdma_activity_worker);
+		INIT_WORK_ONSTACK(&sdma_activity_work_handler.sdma_activity_work,
+				  kfd_sdma_activity_worker);
 
 		sdma_activity_work_handler.pdd = pdd;
 		sdma_activity_work_handler.sdma_activity_counter = 0;
@@ -323,6 +348,7 @@ static ssize_t kfd_procfs_show(struct kobject *kobj, struct attribute *attr,
 		schedule_work(&sdma_activity_work_handler.sdma_activity_work);
 
 		flush_work(&sdma_activity_work_handler.sdma_activity_work);
+		destroy_work_on_stack(&sdma_activity_work_handler.sdma_activity_work);
 
 		return snprintf(buffer, PAGE_SIZE, "%llu\n",
 				(sdma_activity_work_handler.sdma_activity_counter)/
@@ -666,7 +692,8 @@ int kfd_process_create_wq(void)
 	if (!kfd_process_wq)
 		kfd_process_wq = alloc_workqueue("kfd_process_wq", 0, 0);
 	if (!kfd_restore_wq)
-		kfd_restore_wq = alloc_ordered_workqueue("kfd_restore_wq", 0);
+		kfd_restore_wq = alloc_ordered_workqueue("kfd_restore_wq",
+							 WQ_FREEZABLE);
 
 	if (!kfd_process_wq || !kfd_restore_wq) {
 		kfd_process_destroy_wq();
@@ -691,7 +718,7 @@ void kfd_process_destroy_wq(void)
 static void kfd_process_free_gpuvm(struct kgd_mem *mem,
 			struct kfd_process_device *pdd, void **kptr)
 {
-	struct kfd_dev *dev = pdd->dev;
+	struct kfd_node *dev = pdd->dev;
 
 	if (kptr && *kptr) {
 		amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(mem);
@@ -713,7 +740,7 @@ static int kfd_process_alloc_gpuvm(struct kfd_process_device *pdd,
 				   uint64_t gpu_va, uint32_t size,
 				   uint32_t flags, struct kgd_mem **mem, void **kptr)
 {
-	struct kfd_dev *kdev = pdd->dev;
+	struct kfd_node *kdev = pdd->dev;
 	int err;
 
 	err = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(kdev->adev, gpu_va, size,
@@ -798,18 +825,27 @@ static void kfd_process_device_destroy_ib_mem(struct kfd_process_device *pdd)
 	kfd_process_free_gpuvm(qpd->ib_mem, pdd, &qpd->ib_kaddr);
 }
 
-struct kfd_process *kfd_create_process(struct file *filep)
+struct kfd_process *kfd_create_process(struct task_struct *thread)
 {
 	struct kfd_process *process;
-	struct task_struct *thread = current;
 	int ret;
 
-	if (!thread->mm)
+	if (!(thread->mm && mmget_not_zero(thread->mm)))
 		return ERR_PTR(-EINVAL);
 
 	/* Only the pthreads threading model is supported. */
-	if (thread->group_leader->mm != thread->mm)
+	if (thread->group_leader->mm != thread->mm) {
+		mmput(thread->mm);
 		return ERR_PTR(-EINVAL);
+	}
+
+	/* If the process just called exec(3), it is possible that the
+	 * cleanup of the kfd_process (following the release of the mm
+	 * of the old process image) is still in the cleanup work queue.
+	 * Make sure to drain any job before trying to recreate any
+	 * resource for this process.
+	 */
+	flush_workqueue(kfd_process_wq);
 
 	/*
 	 * take kfd processes mutex before starting of process creation
@@ -818,8 +854,16 @@ struct kfd_process *kfd_create_process(struct file *filep)
 	 */
 	mutex_lock(&kfd_processes_mutex);
 
-	/* A prior open of /dev/kfd could have already created the process. */
-	process = find_process(thread, false);
+	if (kfd_is_locked(NULL)) {
+		pr_debug("KFD is locked! Cannot create process");
+		process = ERR_PTR(-EINVAL);
+		goto out;
+	}
+
+	/* A prior open of /dev/kfd could have already created the process.
+	 * find_process will increase process kref in this case
+	 */
+	process = find_process(thread, true);
 	if (process) {
 		pr_debug("Process already found\n");
 	} else {
@@ -827,10 +871,6 @@ struct kfd_process *kfd_create_process(struct file *filep)
 		if (IS_ERR(process))
 			goto out;
 
-		ret = kfd_process_init_cwsr_apu(process, filep);
-		if (ret)
-			goto out_destroy;
-
 		if (!procfs.kobj)
 			goto out;
 
@@ -859,21 +899,16 @@ struct kfd_process *kfd_create_process(struct file *filep)
 		kfd_procfs_add_sysfs_stats(process);
 		kfd_procfs_add_sysfs_files(process);
 		kfd_procfs_add_sysfs_counters(process);
+
+		kfd_debugfs_add_process(process);
+
+		init_waitqueue_head(&process->wait_irq_drain);
 	}
 out:
-	if (!IS_ERR(process))
-		kref_get(&process->ref);
 	mutex_unlock(&kfd_processes_mutex);
+	mmput(thread->mm);
 
 	return process;
-
-out_destroy:
-	hash_del_rcu(&process->kfd_processes);
-	mutex_unlock(&kfd_processes_mutex);
-	synchronize_srcu(&kfd_processes_srcu);
-	/* kfd_process_free_notifier will trigger the cleanup */
-	mmu_notifier_put(&process->mmu_notifier);
-	return ERR_PTR(ret);
 }
 
 struct kfd_process *kfd_get_process(const struct task_struct *thread)
@@ -982,7 +1017,7 @@ static void kfd_process_device_free_bos(struct kfd_process_device *pdd)
 static void kfd_process_kunmap_signal_bo(struct kfd_process *p)
 {
 	struct kfd_process_device *pdd;
-	struct kfd_dev *kdev;
+	struct kfd_node *kdev;
 	void *mem;
 
 	kdev = kfd_device_by_id(GET_GPU_ID(p->signal_handle));
@@ -1021,40 +1056,39 @@ static void kfd_process_destroy_pdds(struct kfd_process *p)
 	for (i = 0; i < p->n_pdds; i++) {
 		struct kfd_process_device *pdd = p->pdds[i];
 
-		pr_debug("Releasing pdd (topology id %d) for process (pasid 0x%x)\n",
-				pdd->dev->id, p->pasid);
+		kfd_smi_event_process(pdd, false);
 
+		pr_debug("Releasing pdd (topology id %d, for pid %d)\n",
+			pdd->dev->id, p->lead_thread->pid);
 		kfd_process_device_destroy_cwsr_dgpu(pdd);
 		kfd_process_device_destroy_ib_mem(pdd);
 
-		if (pdd->drm_file) {
-			amdgpu_amdkfd_gpuvm_release_process_vm(
-					pdd->dev->adev, pdd->drm_priv);
+		if (pdd->drm_file)
 			fput(pdd->drm_file);
-		}
 
 		if (pdd->qpd.cwsr_kaddr && !pdd->qpd.cwsr_base)
 			free_pages((unsigned long)pdd->qpd.cwsr_kaddr,
 				get_order(KFD_CWSR_TBA_TMA_SIZE));
 
-		bitmap_free(pdd->qpd.doorbell_bitmap);
 		idr_destroy(&pdd->alloc_idr);
 
-		kfd_free_process_doorbells(pdd->dev, pdd->doorbell_index);
+		kfd_free_process_doorbells(pdd->dev->kfd, pdd);
 
-		if (pdd->dev->shared_resources.enable_mes)
+		if (pdd->dev->kfd->shared_resources.enable_mes &&
+			pdd->proc_ctx_cpu_ptr)
 			amdgpu_amdkfd_free_gtt_mem(pdd->dev->adev,
-						   pdd->proc_ctx_bo);
+						   &pdd->proc_ctx_bo);
 		/*
 		 * before destroying pdd, make sure to report availability
 		 * for auto suspend
 		 */
 		if (pdd->runtime_inuse) {
-			pm_runtime_mark_last_busy(adev_to_drm(pdd->dev->adev)->dev);
 			pm_runtime_put_autosuspend(adev_to_drm(pdd->dev->adev)->dev);
 			pdd->runtime_inuse = false;
 		}
 
+		atomic_dec(&pdd->dev->kfd->kfd_processes_count);
+
 		kfree(pdd);
 		p->pdds[i] = NULL;
 	}
@@ -1105,6 +1139,17 @@ static void kfd_process_remove_sysfs(struct kfd_process *p)
 	p->kobj = NULL;
 }
 
+/*
+ * If any GPU is ongoing reset, wait for reset complete.
+ */
+static void kfd_process_wait_gpu_reset_complete(struct kfd_process *p)
+{
+	int i;
+
+	for (i = 0; i < p->n_pdds; i++)
+		flush_workqueue(p->pdds[i]->dev->adev->reset_domain->wq);
+}
+
 /* No process locking is needed in this function, because the process
  * is not findable any more. We must assume that no other thread is
  * using it any more, otherwise we couldn't safely free the process
@@ -1114,29 +1159,34 @@ static void kfd_process_wq_release(struct work_struct *work)
 {
 	struct kfd_process *p = container_of(work, struct kfd_process,
 					     release_work);
+	struct dma_fence *ef;
 
-	kfd_process_dequeue_from_all_devices(p);
-	pqm_uninit(&p->pqm);
+	/*
+	 * If GPU in reset, user queues may still running, wait for reset complete.
+	 */
+	kfd_process_wait_gpu_reset_complete(p);
 
 	/* Signal the eviction fence after user mode queues are
 	 * destroyed. This allows any BOs to be freed without
 	 * triggering pointless evictions or waiting for fences.
 	 */
-	dma_fence_signal(p->ef);
+	synchronize_rcu();
+	ef = rcu_access_pointer(p->ef);
+	if (ef)
+		dma_fence_signal(ef);
 
 	kfd_process_remove_sysfs(p);
-	kfd_iommu_unbind_process(p);
+	kfd_debugfs_remove_process(p);
 
 	kfd_process_kunmap_signal_bo(p);
 	kfd_process_free_outstanding_kfd_bos(p);
 	svm_range_list_fini(p);
 
 	kfd_process_destroy_pdds(p);
-	dma_fence_put(p->ef);
+	dma_fence_put(ef);
 
 	kfd_event_free_process(p);
 
-	kfd_pasid_free(p->pasid);
 	mutex_destroy(&p->mutex);
 
 	put_task_struct(p->lead_thread);
@@ -1154,10 +1204,8 @@ static void kfd_process_ref_release(struct kref *ref)
 
 static struct mmu_notifier *kfd_process_alloc_notifier(struct mm_struct *mm)
 {
-	int idx = srcu_read_lock(&kfd_processes_srcu);
-	struct kfd_process *p = find_process_by_mm(mm);
-
-	srcu_read_unlock(&kfd_processes_srcu, idx);
+	/* This increments p->ref counter if kfd process p exists */
+	struct kfd_process *p = kfd_lookup_process_by_mm(mm);
 
 	return p ? &p->mmu_notifier : ERR_PTR(-ESRCH);
 }
@@ -1169,11 +1217,48 @@ static void kfd_process_free_notifier(struct mmu_notifier *mn)
 
 static void kfd_process_notifier_release_internal(struct kfd_process *p)
 {
+	int i;
+
 	cancel_delayed_work_sync(&p->eviction_work);
 	cancel_delayed_work_sync(&p->restore_work);
 
+	/*
+	 * Dequeue and destroy user queues, it is not safe for GPU to access
+	 * system memory after mmu release notifier callback returns because
+	 * exit_mmap free process memory afterwards.
+	 */
+	kfd_process_dequeue_from_all_devices(p);
+	pqm_uninit(&p->pqm);
+
+	for (i = 0; i < p->n_pdds; i++) {
+		struct kfd_process_device *pdd = p->pdds[i];
+
+		/* re-enable GFX OFF since runtime enable with ttmp setup disabled it. */
+		if (!kfd_dbg_is_rlc_restore_supported(pdd->dev) && p->runtime_info.ttmp_setup)
+			amdgpu_gfx_off_ctrl(pdd->dev->adev, true);
+	}
+
 	/* Indicate to other users that MM is no longer valid */
 	p->mm = NULL;
+	kfd_dbg_trap_disable(p);
+
+	if (atomic_read(&p->debugged_process_count) > 0) {
+		struct kfd_process *target;
+		unsigned int temp;
+		int idx = srcu_read_lock(&kfd_processes_srcu);
+
+		hash_for_each_rcu(kfd_processes_table, temp, target, kfd_processes) {
+			if (target->debugger_process && target->debugger_process == p) {
+				mutex_lock_nested(&target->mutex, 1);
+				kfd_dbg_trap_disable(target);
+				mutex_unlock(&target->mutex);
+				if (atomic_read(&p->debugged_process_count) == 0)
+					break;
+			}
+		}
+
+		srcu_read_unlock(&kfd_processes_srcu, idx);
+	}
 
 	mmu_notifier_put(&p->mmu_notifier);
 }
@@ -1253,16 +1338,19 @@ void kfd_cleanup_processes(void)
 	mmu_notifier_synchronize();
 }
 
-static int kfd_process_init_cwsr_apu(struct kfd_process *p, struct file *filep)
+int kfd_process_init_cwsr_apu(struct kfd_process *p, struct file *filep)
 {
 	unsigned long  offset;
 	int i;
 
+	if (p->has_cwsr)
+		return 0;
+
 	for (i = 0; i < p->n_pdds; i++) {
-		struct kfd_dev *dev = p->pdds[i]->dev;
+		struct kfd_node *dev = p->pdds[i]->dev;
 		struct qcm_process_device *qpd = &p->pdds[i]->qpd;
 
-		if (!dev->cwsr_enabled || qpd->cwsr_kaddr || qpd->cwsr_base)
+		if (!dev->kfd->cwsr_enabled || qpd->cwsr_kaddr || qpd->cwsr_base)
 			continue;
 
 		offset = KFD_MMAP_TYPE_RESERVED_MEM | KFD_MMAP_GPU_ID(dev->id);
@@ -1273,25 +1361,30 @@ static int kfd_process_init_cwsr_apu(struct kfd_process *p, struct file *filep)
 		if (IS_ERR_VALUE(qpd->tba_addr)) {
 			int err = qpd->tba_addr;
 
-			pr_err("Failure to set tba address. error %d.\n", err);
+			dev_err(dev->adev->dev,
+				"Failure to set tba address. error %d.\n", err);
 			qpd->tba_addr = 0;
 			qpd->cwsr_kaddr = NULL;
 			return err;
 		}
 
-		memcpy(qpd->cwsr_kaddr, dev->cwsr_isa, dev->cwsr_isa_size);
+		memcpy(qpd->cwsr_kaddr, dev->kfd->cwsr_isa, dev->kfd->cwsr_isa_size);
+
+		kfd_process_set_trap_debug_flag(qpd, p->debug_trap_enabled);
 
 		qpd->tma_addr = qpd->tba_addr + KFD_CWSR_TMA_OFFSET;
 		pr_debug("set tba :0x%llx, tma:0x%llx, cwsr_kaddr:%p for pqm.\n",
 			qpd->tba_addr, qpd->tma_addr, qpd->cwsr_kaddr);
 	}
 
+	p->has_cwsr = true;
+
 	return 0;
 }
 
 static int kfd_process_device_init_cwsr_dgpu(struct kfd_process_device *pdd)
 {
-	struct kfd_dev *dev = pdd->dev;
+	struct kfd_node *dev = pdd->dev;
 	struct qcm_process_device *qpd = &pdd->qpd;
 	uint32_t flags = KFD_IOC_ALLOC_MEM_FLAGS_GTT
 			| KFD_IOC_ALLOC_MEM_FLAGS_NO_SUBSTITUTE
@@ -1300,7 +1393,7 @@ static int kfd_process_device_init_cwsr_dgpu(struct kfd_process_device *pdd)
 	void *kaddr;
 	int ret;
 
-	if (!dev->cwsr_enabled || qpd->cwsr_kaddr || !qpd->cwsr_base)
+	if (!dev->kfd->cwsr_enabled || qpd->cwsr_kaddr || !qpd->cwsr_base)
 		return 0;
 
 	/* cwsr_base is only set for dGPU */
@@ -1313,7 +1406,10 @@ static int kfd_process_device_init_cwsr_dgpu(struct kfd_process_device *pdd)
 	qpd->cwsr_kaddr = kaddr;
 	qpd->tba_addr = qpd->cwsr_base;
 
-	memcpy(qpd->cwsr_kaddr, dev->cwsr_isa, dev->cwsr_isa_size);
+	memcpy(qpd->cwsr_kaddr, dev->kfd->cwsr_isa, dev->kfd->cwsr_isa_size);
+
+	kfd_process_set_trap_debug_flag(&pdd->qpd,
+					pdd->process->debug_trap_enabled);
 
 	qpd->tma_addr = qpd->tba_addr + KFD_CWSR_TMA_OFFSET;
 	pr_debug("set tba :0x%llx, tma:0x%llx, cwsr_kaddr:%p for pqm.\n",
@@ -1324,10 +1420,10 @@ static int kfd_process_device_init_cwsr_dgpu(struct kfd_process_device *pdd)
 
 static void kfd_process_device_destroy_cwsr_dgpu(struct kfd_process_device *pdd)
 {
-	struct kfd_dev *dev = pdd->dev;
+	struct kfd_node *dev = pdd->dev;
 	struct qcm_process_device *qpd = &pdd->qpd;
 
-	if (!dev->cwsr_enabled || !qpd->cwsr_kaddr || !qpd->cwsr_base)
+	if (!dev->kfd->cwsr_enabled || !qpd->cwsr_kaddr || !qpd->cwsr_base)
 		return;
 
 	kfd_process_free_gpuvm(qpd->cwsr_mem, pdd, &qpd->cwsr_kaddr);
@@ -1371,7 +1467,7 @@ bool kfd_process_xnack_mode(struct kfd_process *p, bool supported)
 	 * support retry.
 	 */
 	for (i = 0; i < p->n_pdds; i++) {
-		struct kfd_dev *dev = p->pdds[i]->dev;
+		struct kfd_node *dev = p->pdds[i]->dev;
 
 		/* Only consider GFXv9 and higher GPUs. Older GPUs don't
 		 * support the SVM APIs and don't need to be considered
@@ -1383,8 +1479,13 @@ bool kfd_process_xnack_mode(struct kfd_process *p, bool supported)
 		 * per-process XNACK mode selection. But let the dev->noretry
 		 * setting still influence the default XNACK mode.
 		 */
-		if (supported && KFD_SUPPORT_XNACK_PER_PROCESS(dev))
+		if (supported && KFD_SUPPORT_XNACK_PER_PROCESS(dev)) {
+			if (!amdgpu_sriov_xnack_support(dev->kfd->adev)) {
+				pr_debug("SRIOV platform xnack not supported\n");
+				return false;
+			}
 			continue;
+		}
 
 		/* GFXv10 and later GPUs do not support shader preemption
 		 * during page faults. This can lead to poor QoS for queue
@@ -1394,13 +1495,23 @@ bool kfd_process_xnack_mode(struct kfd_process *p, bool supported)
 		if (KFD_GC_VERSION(dev) >= IP_VERSION(10, 1, 1))
 			return false;
 
-		if (dev->noretry)
+		if (dev->kfd->noretry)
 			return false;
 	}
 
 	return true;
 }
 
+void kfd_process_set_trap_debug_flag(struct qcm_process_device *qpd,
+				     bool enabled)
+{
+	if (qpd->cwsr_kaddr) {
+		uint64_t *tma =
+			(uint64_t *)(qpd->cwsr_kaddr + KFD_CWSR_TMA_OFFSET);
+		tma[2] = enabled;
+	}
+}
+
 /*
  * On return the kfd_process is fully operational and will be freed when the
  * mm is released
@@ -1428,12 +1539,11 @@ static struct kfd_process *create_process(const struct task_struct *thread)
 	if (err)
 		goto err_event_init;
 	process->is_32bit_user_mode = in_compat_syscall();
-
-	process->pasid = kfd_pasid_alloc();
-	if (process->pasid == 0) {
-		err = -ENOSPC;
-		goto err_alloc_pasid;
-	}
+	process->debug_trap_enabled = false;
+	process->debugger_process = NULL;
+	process->exception_enable_mask = 0;
+	atomic_set(&process->debugged_process_count, 0);
+	sema_init(&process->runtime_enable_sema, 0);
 
 	err = pqm_init(&process->pqm, process);
 	if (err != 0)
@@ -1475,6 +1585,8 @@ static struct kfd_process *create_process(const struct task_struct *thread)
 	kfd_unref_process(process);
 	get_task_struct(process->lead_thread);
 
+	INIT_WORK(&process->debug_event_workarea, debug_event_write_work_handler);
+
 	return process;
 
 err_register_notifier:
@@ -1486,8 +1598,6 @@ err_init_svm_range_list:
 err_init_apertures:
 	pqm_uninit(&process->pqm);
 err_process_pqm_init:
-	kfd_pasid_free(process->pasid);
-err_alloc_pasid:
 	kfd_event_free_process(process);
 err_event_init:
 	mutex_destroy(&process->mutex);
@@ -1496,39 +1606,7 @@ err_alloc_process:
 	return ERR_PTR(err);
 }
 
-static int init_doorbell_bitmap(struct qcm_process_device *qpd,
-			struct kfd_dev *dev)
-{
-	unsigned int i;
-	int range_start = dev->shared_resources.non_cp_doorbells_start;
-	int range_end = dev->shared_resources.non_cp_doorbells_end;
-
-	if (!KFD_IS_SOC15(dev))
-		return 0;
-
-	qpd->doorbell_bitmap = bitmap_zalloc(KFD_MAX_NUM_OF_QUEUES_PER_PROCESS,
-					     GFP_KERNEL);
-	if (!qpd->doorbell_bitmap)
-		return -ENOMEM;
-
-	/* Mask out doorbells reserved for SDMA, IH, and VCN on SOC15. */
-	pr_debug("reserved doorbell 0x%03x - 0x%03x\n", range_start, range_end);
-	pr_debug("reserved doorbell 0x%03x - 0x%03x\n",
-			range_start + KFD_QUEUE_DOORBELL_MIRROR_OFFSET,
-			range_end + KFD_QUEUE_DOORBELL_MIRROR_OFFSET);
-
-	for (i = 0; i < KFD_MAX_NUM_OF_QUEUES_PER_PROCESS / 2; i++) {
-		if (i >= range_start && i <= range_end) {
-			__set_bit(i, qpd->doorbell_bitmap);
-			__set_bit(i + KFD_QUEUE_DOORBELL_MIRROR_OFFSET,
-				  qpd->doorbell_bitmap);
-		}
-	}
-
-	return 0;
-}
-
-struct kfd_process_device *kfd_get_process_device_data(struct kfd_dev *dev,
+struct kfd_process_device *kfd_get_process_device_data(struct kfd_node *dev,
 							struct kfd_process *p)
 {
 	int i;
@@ -1540,11 +1618,10 @@ struct kfd_process_device *kfd_get_process_device_data(struct kfd_dev *dev,
 	return NULL;
 }
 
-struct kfd_process_device *kfd_create_process_device_data(struct kfd_dev *dev,
+struct kfd_process_device *kfd_create_process_device_data(struct kfd_node *dev,
 							struct kfd_process *p)
 {
 	struct kfd_process_device *pdd = NULL;
-	int retval = 0;
 
 	if (WARN_ON_ONCE(p->n_pdds >= MAX_GPU_INSTANCE))
 		return NULL;
@@ -1552,11 +1629,6 @@ struct kfd_process_device *kfd_create_process_device_data(struct kfd_dev *dev,
 	if (!pdd)
 		return NULL;
 
-	if (init_doorbell_bitmap(&pdd->qpd, dev)) {
-		pr_err("Failed to init doorbell for process\n");
-		goto err_free_pdd;
-	}
-
 	pdd->dev = dev;
 	INIT_LIST_HEAD(&pdd->qpd.queues_list);
 	INIT_LIST_HEAD(&pdd->qpd.priv_queue_list);
@@ -1568,35 +1640,24 @@ struct kfd_process_device *kfd_create_process_device_data(struct kfd_dev *dev,
 	pdd->bound = PDD_UNBOUND;
 	pdd->already_dequeued = false;
 	pdd->runtime_inuse = false;
-	pdd->vram_usage = 0;
+	atomic64_set(&pdd->vram_usage, 0);
 	pdd->sdma_past_activity_counter = 0;
 	pdd->user_gpu_id = dev->id;
 	atomic64_set(&pdd->evict_duration_counter, 0);
 
-	if (dev->shared_resources.enable_mes) {
-		retval = amdgpu_amdkfd_alloc_gtt_mem(dev->adev,
-						AMDGPU_MES_PROC_CTX_SIZE,
-						&pdd->proc_ctx_bo,
-						&pdd->proc_ctx_gpu_addr,
-						&pdd->proc_ctx_cpu_ptr,
-						false);
-		if (retval) {
-			pr_err("failed to allocate process context bo\n");
-			goto err_free_pdd;
-		}
-		memset(pdd->proc_ctx_cpu_ptr, 0, AMDGPU_MES_PROC_CTX_SIZE);
-	}
-
 	p->pdds[p->n_pdds++] = pdd;
+	if (kfd_dbg_is_per_vmid_supported(pdd->dev))
+		pdd->spi_dbg_override = pdd->dev->kfd2kgd->disable_debug_trap(
+							pdd->dev->adev,
+							false,
+							0);
 
 	/* Init idr used for memory handle translation */
 	idr_init(&pdd->alloc_idr);
 
-	return pdd;
+	atomic_inc(&dev->kfd->kfd_processes_count);
 
-err_free_pdd:
-	kfree(pdd);
-	return NULL;
+	return pdd;
 }
 
 /**
@@ -1619,7 +1680,8 @@ int kfd_process_device_init_vm(struct kfd_process_device *pdd,
 	struct amdgpu_fpriv *drv_priv;
 	struct amdgpu_vm *avm;
 	struct kfd_process *p;
-	struct kfd_dev *dev;
+	struct dma_fence *ef;
+	struct kfd_node *dev;
 	int ret;
 
 	if (!drm_file)
@@ -1638,13 +1700,16 @@ int kfd_process_device_init_vm(struct kfd_process_device *pdd,
 
 	ret = amdgpu_amdkfd_gpuvm_acquire_process_vm(dev->adev, avm,
 						     &p->kgd_process_info,
-						     &p->ef);
+						     p->ef ? NULL : &ef);
 	if (ret) {
-		pr_err("Failed to create process VM object\n");
+		dev_err(dev->adev->dev, "Failed to create process VM object\n");
 		return ret;
 	}
+
+	if (!p->ef)
+		RCU_INIT_POINTER(p->ef, ef);
+
 	pdd->drm_priv = drm_file->private_data;
-	atomic64_set(&pdd->tlb_seq, 0);
 
 	ret = kfd_process_device_reserve_ib_mem(pdd);
 	if (ret)
@@ -1653,15 +1718,21 @@ int kfd_process_device_init_vm(struct kfd_process_device *pdd,
 	if (ret)
 		goto err_init_cwsr;
 
-	ret = amdgpu_amdkfd_gpuvm_set_vm_pasid(dev->adev, avm, p->pasid);
-	if (ret)
-		goto err_set_pasid;
+	if (unlikely(!avm->pasid)) {
+		dev_warn(pdd->dev->adev->dev, "WARN: vm %p has no pasid associated",
+				 avm);
+		ret = -EINVAL;
+		goto err_get_pasid;
+	}
 
+	pdd->pasid = avm->pasid;
 	pdd->drm_file = drm_file;
 
+	kfd_smi_event_process(pdd, true);
+
 	return 0;
 
-err_set_pasid:
+err_get_pasid:
 	kfd_process_device_destroy_cwsr_dgpu(pdd);
 err_init_cwsr:
 	kfd_process_device_destroy_ib_mem(pdd);
@@ -1679,7 +1750,7 @@ err_reserve_ib_mem:
  *
  * Assumes that the process lock is held.
  */
-struct kfd_process_device *kfd_bind_process_to_device(struct kfd_dev *dev,
+struct kfd_process_device *kfd_bind_process_to_device(struct kfd_node *dev,
 							struct kfd_process *p)
 {
 	struct kfd_process_device *pdd;
@@ -1687,7 +1758,7 @@ struct kfd_process_device *kfd_bind_process_to_device(struct kfd_dev *dev,
 
 	pdd = kfd_get_process_device_data(dev, p);
 	if (!pdd) {
-		pr_err("Process device data doesn't exist\n");
+		dev_err(dev->adev->dev, "Process device data doesn't exist\n");
 		return ERR_PTR(-ENOMEM);
 	}
 
@@ -1707,10 +1778,6 @@ struct kfd_process_device *kfd_bind_process_to_device(struct kfd_dev *dev,
 		}
 	}
 
-	err = kfd_iommu_bind_process_to_device(pdd);
-	if (err)
-		goto out;
-
 	/*
 	 * make sure that runtime_usage counter is incremented just once
 	 * per pdd
@@ -1718,15 +1785,6 @@ struct kfd_process_device *kfd_bind_process_to_device(struct kfd_dev *dev,
 	pdd->runtime_inuse = true;
 
 	return pdd;
-
-out:
-	/* balance runpm reference count and exit with error */
-	if (!pdd->runtime_inuse) {
-		pm_runtime_mark_last_busy(adev_to_drm(dev->adev)->dev);
-		pm_runtime_put_autosuspend(adev_to_drm(dev->adev)->dev);
-	}
-
-	return ERR_PTR(err);
 }
 
 /* Create specific handle mapped to mem from process local memory idr
@@ -1760,25 +1818,50 @@ void kfd_process_device_remove_obj_handle(struct kfd_process_device *pdd,
 		idr_remove(&pdd->alloc_idr, handle);
 }
 
-/* This increments the process->ref counter. */
-struct kfd_process *kfd_lookup_process_by_pasid(u32 pasid)
+static struct kfd_process_device *kfd_lookup_process_device_by_pasid(u32 pasid)
 {
-	struct kfd_process *p, *ret_p = NULL;
+	struct kfd_process_device *ret_p = NULL;
+	struct kfd_process *p;
 	unsigned int temp;
-
-	int idx = srcu_read_lock(&kfd_processes_srcu);
+	int i;
 
 	hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) {
-		if (p->pasid == pasid) {
-			kref_get(&p->ref);
-			ret_p = p;
-			break;
+		for (i = 0; i < p->n_pdds; i++) {
+			if (p->pdds[i]->pasid == pasid) {
+				ret_p = p->pdds[i];
+				break;
+			}
 		}
+		if (ret_p)
+			break;
+	}
+	return ret_p;
+}
+
+/* This increments the process->ref counter. */
+struct kfd_process *kfd_lookup_process_by_pasid(u32 pasid,
+						struct kfd_process_device **pdd)
+{
+	struct kfd_process_device *ret_p;
+
+	int idx = srcu_read_lock(&kfd_processes_srcu);
+
+	ret_p = kfd_lookup_process_device_by_pasid(pasid);
+	if (ret_p) {
+		if (pdd)
+			*pdd = ret_p;
+		kref_get(&ret_p->process->ref);
+
+		srcu_read_unlock(&kfd_processes_srcu, idx);
+		return ret_p->process;
 	}
 
 	srcu_read_unlock(&kfd_processes_srcu, idx);
 
-	return ret_p;
+	if (pdd)
+		*pdd = NULL;
+
+	return NULL;
 }
 
 /* This increments the process->ref counter. */
@@ -1810,6 +1893,7 @@ int kfd_process_evict_queues(struct kfd_process *p, uint32_t trigger)
 
 	for (i = 0; i < p->n_pdds; i++) {
 		struct kfd_process_device *pdd = p->pdds[i];
+		struct device *dev = pdd->dev->adev->dev;
 
 		kfd_smi_event_queue_eviction(pdd->dev, p->lead_thread->pid,
 					     trigger);
@@ -1821,10 +1905,12 @@ int kfd_process_evict_queues(struct kfd_process *p, uint32_t trigger)
 		 * them been add back since they actually not be saved right now.
 		 */
 		if (r && r != -EIO) {
-			pr_err("Failed to evict process queues\n");
+			dev_err(dev, "Failed to evict process queues\n");
 			goto fail;
 		}
 		n_evicted++;
+
+		pdd->dev->dqm->is_hws_hang = false;
 	}
 
 	return r;
@@ -1843,7 +1929,8 @@ fail:
 
 		if (pdd->dev->dqm->ops.restore_process_queues(pdd->dev->dqm,
 							      &pdd->qpd))
-			pr_err("Failed to restore queues\n");
+			dev_err(pdd->dev->adev->dev,
+				"Failed to restore queues\n");
 
 		n_evicted--;
 	}
@@ -1859,13 +1946,14 @@ int kfd_process_restore_queues(struct kfd_process *p)
 
 	for (i = 0; i < p->n_pdds; i++) {
 		struct kfd_process_device *pdd = p->pdds[i];
+		struct device *dev = pdd->dev->adev->dev;
 
 		kfd_smi_event_queue_restore(pdd->dev, p->lead_thread->pid);
 
 		r = pdd->dev->dqm->ops.restore_process_queues(pdd->dev->dqm,
 							      &pdd->qpd);
 		if (r) {
-			pr_err("Failed to restore process queues\n");
+			dev_err(dev, "Failed to restore process queues\n");
 			if (!ret)
 				ret = r;
 		}
@@ -1885,13 +1973,13 @@ int kfd_process_gpuidx_from_gpuid(struct kfd_process *p, uint32_t gpu_id)
 }
 
 int
-kfd_process_gpuid_from_adev(struct kfd_process *p, struct amdgpu_device *adev,
-			   uint32_t *gpuid, uint32_t *gpuidx)
+kfd_process_gpuid_from_node(struct kfd_process *p, struct kfd_node *node,
+			    uint32_t *gpuid, uint32_t *gpuidx)
 {
 	int i;
 
 	for (i = 0; i < p->n_pdds; i++)
-		if (p->pdds[i] && p->pdds[i]->dev->adev == adev) {
+		if (p->pdds[i] && p->pdds[i]->dev == node) {
 			*gpuid = p->pdds[i]->user_gpu_id;
 			*gpuidx = i;
 			return 0;
@@ -1899,6 +1987,23 @@ kfd_process_gpuid_from_adev(struct kfd_process *p, struct amdgpu_device *adev,
 	return -EINVAL;
 }
 
+static int signal_eviction_fence(struct kfd_process *p)
+{
+	struct dma_fence *ef;
+	int ret;
+
+	rcu_read_lock();
+	ef = dma_fence_get_rcu_safe(&p->ef);
+	rcu_read_unlock();
+	if (!ef)
+		return -EINVAL;
+
+	ret = dma_fence_signal(ef);
+	dma_fence_put(ef);
+
+	return ret;
+}
+
 static void evict_process_worker(struct work_struct *work)
 {
 	int ret;
@@ -1911,29 +2016,45 @@ static void evict_process_worker(struct work_struct *work)
 	 * lifetime of this thread, kfd_process p will be valid
 	 */
 	p = container_of(dwork, struct kfd_process, eviction_work);
-	WARN_ONCE(p->last_eviction_seqno != p->ef->seqno,
-		  "Eviction fence mismatch\n");
-
-	/* Narrow window of overlap between restore and evict work
-	 * item is possible. Once amdgpu_amdkfd_gpuvm_restore_process_bos
-	 * unreserves KFD BOs, it is possible to evicted again. But
-	 * restore has few more steps of finish. So lets wait for any
-	 * previous restore work to complete
-	 */
-	flush_delayed_work(&p->restore_work);
 
-	pr_debug("Started evicting pasid 0x%x\n", p->pasid);
+	pr_debug("Started evicting process pid %d\n", p->lead_thread->pid);
 	ret = kfd_process_evict_queues(p, KFD_QUEUE_EVICTION_TRIGGER_TTM);
 	if (!ret) {
-		dma_fence_signal(p->ef);
-		dma_fence_put(p->ef);
-		p->ef = NULL;
-		queue_delayed_work(kfd_restore_wq, &p->restore_work,
-				msecs_to_jiffies(PROCESS_RESTORE_TIME_MS));
+		/* If another thread already signaled the eviction fence,
+		 * they are responsible stopping the queues and scheduling
+		 * the restore work.
+		 */
+		if (signal_eviction_fence(p) ||
+		    mod_delayed_work(kfd_restore_wq, &p->restore_work,
+				     msecs_to_jiffies(PROCESS_RESTORE_TIME_MS)))
+			kfd_process_restore_queues(p);
 
-		pr_debug("Finished evicting pasid 0x%x\n", p->pasid);
+		pr_debug("Finished evicting process pid %d\n", p->lead_thread->pid);
 	} else
-		pr_err("Failed to evict queues of pasid 0x%x\n", p->pasid);
+		pr_err("Failed to evict queues of process pid %d\n", p->lead_thread->pid);
+}
+
+static int restore_process_helper(struct kfd_process *p)
+{
+	int ret = 0;
+
+	/* VMs may not have been acquired yet during debugging. */
+	if (p->kgd_process_info) {
+		ret = amdgpu_amdkfd_gpuvm_restore_process_bos(
+			p->kgd_process_info, &p->ef);
+		if (ret)
+			return ret;
+	}
+
+	ret = kfd_process_restore_queues(p);
+	if (!ret)
+		pr_debug("Finished restoring process pid %d\n",
+			p->lead_thread->pid);
+	else
+		pr_err("Failed to restore queues of process pid %d\n",
+		      p->lead_thread->pid);
+
+	return ret;
 }
 
 static void restore_process_worker(struct work_struct *work)
@@ -1948,7 +2069,7 @@ static void restore_process_worker(struct work_struct *work)
 	 * lifetime of this thread, kfd_process p will be valid
 	 */
 	p = container_of(dwork, struct kfd_process, restore_work);
-	pr_debug("Started restoring pasid 0x%x\n", p->pasid);
+	pr_debug("Started restoring process pasid %d\n", (int)p->lead_thread->pid);
 
 	/* Setting last_restore_timestamp before successful restoration.
 	 * Otherwise this would have to be set by KGD (restore_process_bos)
@@ -1961,22 +2082,15 @@ static void restore_process_worker(struct work_struct *work)
 	 */
 
 	p->last_restore_timestamp = get_jiffies_64();
-	ret = amdgpu_amdkfd_gpuvm_restore_process_bos(p->kgd_process_info,
-						     &p->ef);
+
+	ret = restore_process_helper(p);
 	if (ret) {
-		pr_debug("Failed to restore BOs of pasid 0x%x, retry after %d ms\n",
-			 p->pasid, PROCESS_BACK_OFF_TIME_MS);
-		ret = queue_delayed_work(kfd_restore_wq, &p->restore_work,
-				msecs_to_jiffies(PROCESS_BACK_OFF_TIME_MS));
-		WARN(!ret, "reschedule restore work failed\n");
-		return;
+		pr_debug("Failed to restore BOs of process pid %d, retry after %d ms\n",
+			 p->lead_thread->pid, PROCESS_BACK_OFF_TIME_MS);
+		if (mod_delayed_work(kfd_restore_wq, &p->restore_work,
+				     msecs_to_jiffies(PROCESS_RESTORE_TIME_MS)))
+			kfd_process_restore_queues(p);
 	}
-
-	ret = kfd_process_restore_queues(p);
-	if (!ret)
-		pr_debug("Finished restoring pasid 0x%x\n", p->pasid);
-	else
-		pr_err("Failed to restore queues of pasid 0x%x\n", p->pasid);
 }
 
 void kfd_suspend_all_processes(void)
@@ -1987,14 +2101,9 @@ void kfd_suspend_all_processes(void)
 
 	WARN(debug_evictions, "Evicting all processes");
 	hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) {
-		cancel_delayed_work_sync(&p->eviction_work);
-		cancel_delayed_work_sync(&p->restore_work);
-
 		if (kfd_process_evict_queues(p, KFD_QUEUE_EVICTION_TRIGGER_SUSPEND))
-			pr_err("Failed to suspend process 0x%x\n", p->pasid);
-		dma_fence_signal(p->ef);
-		dma_fence_put(p->ef);
-		p->ef = NULL;
+			pr_err("Failed to suspend process pid %d\n", p->lead_thread->pid);
+		signal_eviction_fence(p);
 	}
 	srcu_read_unlock(&kfd_processes_srcu, idx);
 }
@@ -2006,9 +2115,9 @@ int kfd_resume_all_processes(void)
 	int ret = 0, idx = srcu_read_lock(&kfd_processes_srcu);
 
 	hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) {
-		if (!queue_delayed_work(kfd_restore_wq, &p->restore_work, 0)) {
-			pr_err("Restore process %d failed during resume\n",
-			       p->pasid);
+		if (restore_process_helper(p)) {
+			pr_err("Restore process pid %d failed during resume\n",
+			      p->lead_thread->pid);
 			ret = -EFAULT;
 		}
 	}
@@ -2016,14 +2125,14 @@ int kfd_resume_all_processes(void)
 	return ret;
 }
 
-int kfd_reserved_mem_mmap(struct kfd_dev *dev, struct kfd_process *process,
+int kfd_reserved_mem_mmap(struct kfd_node *dev, struct kfd_process *process,
 			  struct vm_area_struct *vma)
 {
 	struct kfd_process_device *pdd;
 	struct qcm_process_device *qpd;
 
 	if ((vma->vm_end - vma->vm_start) != KFD_CWSR_TBA_TMA_SIZE) {
-		pr_err("Incorrect CWSR mapping size.\n");
+		dev_err(dev->adev->dev, "Incorrect CWSR mapping size.\n");
 		return -EINVAL;
 	}
 
@@ -2035,7 +2144,8 @@ int kfd_reserved_mem_mmap(struct kfd_dev *dev, struct kfd_process *process,
 	qpd->cwsr_kaddr = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO,
 					get_order(KFD_CWSR_TBA_TMA_SIZE));
 	if (!qpd->cwsr_kaddr) {
-		pr_err("Error allocating per process CWSR buffer.\n");
+		dev_err(dev->adev->dev,
+			"Error allocating per process CWSR buffer.\n");
 		return -ENOMEM;
 	}
 
@@ -2047,31 +2157,129 @@ int kfd_reserved_mem_mmap(struct kfd_dev *dev, struct kfd_process *process,
 			       KFD_CWSR_TBA_TMA_SIZE, vma->vm_page_prot);
 }
 
-void kfd_flush_tlb(struct kfd_process_device *pdd, enum TLB_FLUSH_TYPE type)
+/* assumes caller holds process lock. */
+int kfd_process_drain_interrupts(struct kfd_process_device *pdd)
 {
-	struct amdgpu_vm *vm = drm_priv_to_vm(pdd->drm_priv);
-	uint64_t tlb_seq = amdgpu_vm_tlb_seq(vm);
-	struct kfd_dev *dev = pdd->dev;
+	uint32_t irq_drain_fence[8];
+	uint8_t node_id = 0;
+	int r = 0;
+
+	if (!KFD_IS_SOC15(pdd->dev))
+		return 0;
+
+	pdd->process->irq_drain_is_open = true;
+
+	memset(irq_drain_fence, 0, sizeof(irq_drain_fence));
+	irq_drain_fence[0] = (KFD_IRQ_FENCE_SOURCEID << 8) |
+							KFD_IRQ_FENCE_CLIENTID;
+	irq_drain_fence[3] = pdd->pasid;
 
 	/*
-	 * It can be that we race and lose here, but that is extremely unlikely
-	 * and the worst thing which could happen is that we flush the changes
-	 * into the TLB once more which is harmless.
+	 * For GFX 9.4.3/9.5.0, send the NodeId also in IH cookie DW[3]
 	 */
-	if (atomic64_xchg(&pdd->tlb_seq, tlb_seq) == tlb_seq)
-		return;
+	if (KFD_GC_VERSION(pdd->dev->kfd) == IP_VERSION(9, 4, 3) ||
+	    KFD_GC_VERSION(pdd->dev->kfd) == IP_VERSION(9, 4, 4) ||
+	    KFD_GC_VERSION(pdd->dev->kfd) == IP_VERSION(9, 5, 0)) {
+		node_id = ffs(pdd->dev->interrupt_bitmap) - 1;
+		irq_drain_fence[3] |= node_id << 16;
+	}
 
-	if (dev->dqm->sched_policy == KFD_SCHED_POLICY_NO_HWS) {
-		/* Nothing to flush until a VMID is assigned, which
-		 * only happens when the first queue is created.
-		 */
-		if (pdd->qpd.vmid)
-			amdgpu_amdkfd_flush_gpu_tlb_vmid(dev->adev,
-							pdd->qpd.vmid);
-	} else {
-		amdgpu_amdkfd_flush_gpu_tlb_pasid(dev->adev,
-					pdd->process->pasid, type);
+	/* ensure stale irqs scheduled KFD interrupts and send drain fence. */
+	if (amdgpu_amdkfd_send_close_event_drain_irq(pdd->dev->adev,
+						     irq_drain_fence)) {
+		pdd->process->irq_drain_is_open = false;
+		return 0;
 	}
+
+	r = wait_event_interruptible(pdd->process->wait_irq_drain,
+				     !READ_ONCE(pdd->process->irq_drain_is_open));
+	if (r)
+		pdd->process->irq_drain_is_open = false;
+
+	return r;
+}
+
+void kfd_process_close_interrupt_drain(unsigned int pasid)
+{
+	struct kfd_process *p;
+
+	p = kfd_lookup_process_by_pasid(pasid, NULL);
+
+	if (!p)
+		return;
+
+	WRITE_ONCE(p->irq_drain_is_open, false);
+	wake_up_all(&p->wait_irq_drain);
+	kfd_unref_process(p);
+}
+
+struct send_exception_work_handler_workarea {
+	struct work_struct work;
+	struct kfd_process *p;
+	unsigned int queue_id;
+	uint64_t error_reason;
+};
+
+static void send_exception_work_handler(struct work_struct *work)
+{
+	struct send_exception_work_handler_workarea *workarea;
+	struct kfd_process *p;
+	struct queue *q;
+	struct mm_struct *mm;
+	struct kfd_context_save_area_header __user *csa_header;
+	uint64_t __user *err_payload_ptr;
+	uint64_t cur_err;
+	uint32_t ev_id;
+
+	workarea = container_of(work,
+				struct send_exception_work_handler_workarea,
+				work);
+	p = workarea->p;
+
+	mm = get_task_mm(p->lead_thread);
+
+	if (!mm)
+		return;
+
+	kthread_use_mm(mm);
+
+	q = pqm_get_user_queue(&p->pqm, workarea->queue_id);
+
+	if (!q)
+		goto out;
+
+	csa_header = (void __user *)q->properties.ctx_save_restore_area_address;
+
+	get_user(err_payload_ptr, (uint64_t __user **)&csa_header->err_payload_addr);
+	get_user(cur_err, err_payload_ptr);
+	cur_err |= workarea->error_reason;
+	put_user(cur_err, err_payload_ptr);
+	get_user(ev_id, &csa_header->err_event_id);
+
+	kfd_set_event(p, ev_id);
+
+out:
+	kthread_unuse_mm(mm);
+	mmput(mm);
+}
+
+int kfd_send_exception_to_runtime(struct kfd_process *p,
+			unsigned int queue_id,
+			uint64_t error_reason)
+{
+	struct send_exception_work_handler_workarea worker;
+
+	INIT_WORK_ONSTACK(&worker.work, send_exception_work_handler);
+
+	worker.p = p;
+	worker.queue_id = queue_id;
+	worker.error_reason = error_reason;
+
+	schedule_work(&worker.work);
+	flush_work(&worker.work);
+	destroy_work_on_stack(&worker.work);
+
+	return 0;
 }
 
 struct kfd_process_device *kfd_process_device_data_by_id(struct kfd_process *p, uint32_t gpu_id)
@@ -2116,8 +2324,8 @@ int kfd_debugfs_mqds_by_process(struct seq_file *m, void *data)
 	int idx = srcu_read_lock(&kfd_processes_srcu);
 
 	hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) {
-		seq_printf(m, "Process %d PASID 0x%x:\n",
-			   p->lead_thread->tgid, p->pasid);
+		seq_printf(m, "Process %d PASID %d:\n",
+			   p->lead_thread->tgid, p->lead_thread->pid);
 
 		mutex_lock(&p->mutex);
 		r = pqm_debugfs_mqds(m, &p->pqm);
@@ -2133,4 +2341,3 @@ int kfd_debugfs_mqds_by_process(struct seq_file *m, void *data)
 }
 
 #endif
-
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
index 4236539d9f93..7fbb5c274ccc 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
@@ -28,6 +28,7 @@
 #include "kfd_priv.h"
 #include "kfd_kernel_queue.h"
 #include "amdgpu_amdkfd.h"
+#include "amdgpu_reset.h"
 
 static inline struct process_queue_node *get_queue_by_qid(
 			struct process_queue_manager *pqm, unsigned int qid)
@@ -68,8 +69,8 @@ static int find_available_queue_slot(struct process_queue_manager *pqm,
 	pr_debug("The new slot id %lu\n", found);
 
 	if (found >= KFD_MAX_NUM_OF_QUEUES_PER_PROCESS) {
-		pr_info("Cannot open more queues for process with pasid 0x%x\n",
-				pqm->process->pasid);
+		pr_info("Cannot open more queues for process with pid %d\n",
+			pqm->process->lead_thread->pid);
 		return -ENOMEM;
 	}
 
@@ -81,19 +82,29 @@ static int find_available_queue_slot(struct process_queue_manager *pqm,
 
 void kfd_process_dequeue_from_device(struct kfd_process_device *pdd)
 {
-	struct kfd_dev *dev = pdd->dev;
+	struct kfd_node *dev = pdd->dev;
 
 	if (pdd->already_dequeued)
 		return;
-
+	/* The MES context flush needs to filter out the case which the
+	 * KFD process is created without setting up the MES context and
+	 * queue for creating a compute queue.
+	 */
 	dev->dqm->ops.process_termination(dev->dqm, &pdd->qpd);
+	if (dev->kfd->shared_resources.enable_mes && !!pdd->proc_ctx_gpu_addr &&
+	    down_read_trylock(&dev->adev->reset_domain->sem)) {
+		amdgpu_mes_flush_shader_debugger(dev->adev,
+						 pdd->proc_ctx_gpu_addr);
+		up_read(&dev->adev->reset_domain->sem);
+	}
 	pdd->already_dequeued = true;
 }
 
 int pqm_set_gws(struct process_queue_manager *pqm, unsigned int qid,
 			void *gws)
 {
-	struct kfd_dev *dev = NULL;
+	struct mqd_update_info minfo = {0};
+	struct kfd_node *dev = NULL;
 	struct process_queue_node *pqn;
 	struct kfd_process_device *pdd;
 	struct kgd_mem *mem = NULL;
@@ -123,20 +134,34 @@ int pqm_set_gws(struct process_queue_manager *pqm, unsigned int qid,
 	if (!gws && pdd->qpd.num_gws == 0)
 		return -EINVAL;
 
-	if (gws)
-		ret = amdgpu_amdkfd_add_gws_to_process(pdd->process->kgd_process_info,
-			gws, &mem);
-	else
-		ret = amdgpu_amdkfd_remove_gws_from_process(pdd->process->kgd_process_info,
-			pqn->q->gws);
-	if (unlikely(ret))
-		return ret;
+	if ((KFD_GC_VERSION(dev) != IP_VERSION(9, 4, 3) &&
+	     KFD_GC_VERSION(dev) != IP_VERSION(9, 4, 4) &&
+	     KFD_GC_VERSION(dev) != IP_VERSION(9, 5, 0)) &&
+	    !dev->kfd->shared_resources.enable_mes) {
+		if (gws)
+			ret = amdgpu_amdkfd_add_gws_to_process(pdd->process->kgd_process_info,
+				gws, &mem);
+		else
+			ret = amdgpu_amdkfd_remove_gws_from_process(pdd->process->kgd_process_info,
+				pqn->q->gws);
+		if (unlikely(ret))
+			return ret;
+		pqn->q->gws = mem;
+	} else {
+		/*
+		 * Intentionally set GWS to a non-NULL value
+		 * for devices that do not use GWS for global wave
+		 * synchronization but require the formality
+		 * of setting GWS for cooperative groups.
+		 */
+		pqn->q->gws = gws ? ERR_PTR(-ENOMEM) : NULL;
+	}
 
-	pqn->q->gws = mem;
 	pdd->qpd.num_gws = gws ? dev->adev->gds.gws_size : 0;
+	minfo.update_flag = gws ? UPDATE_FLAG_IS_GWS : 0;
 
 	return pqn->q->device->dqm->ops.update_queue(pqn->q->device->dqm,
-							pqn->q, NULL);
+							pqn->q, &minfo);
 }
 
 void kfd_process_dequeue_from_all_devices(struct kfd_process *p)
@@ -159,14 +184,53 @@ int pqm_init(struct process_queue_manager *pqm, struct kfd_process *p)
 	return 0;
 }
 
+static void pqm_clean_queue_resource(struct process_queue_manager *pqm,
+				     struct process_queue_node *pqn)
+{
+	struct kfd_node *dev;
+	struct kfd_process_device *pdd;
+
+	dev = pqn->q->device;
+
+	pdd = kfd_get_process_device_data(dev, pqm->process);
+	if (!pdd) {
+		pr_err("Process device data doesn't exist\n");
+		return;
+	}
+
+	if (pqn->q->gws) {
+		if (KFD_GC_VERSION(pqn->q->device) != IP_VERSION(9, 4, 3) &&
+		    KFD_GC_VERSION(pqn->q->device) != IP_VERSION(9, 4, 4) &&
+		    KFD_GC_VERSION(pqn->q->device) != IP_VERSION(9, 5, 0) &&
+		    !dev->kfd->shared_resources.enable_mes)
+			amdgpu_amdkfd_remove_gws_from_process(
+				pqm->process->kgd_process_info, pqn->q->gws);
+		pdd->qpd.num_gws = 0;
+	}
+
+	if (dev->kfd->shared_resources.enable_mes) {
+		amdgpu_amdkfd_free_gtt_mem(dev->adev, &pqn->q->gang_ctx_bo);
+		amdgpu_amdkfd_free_gtt_mem(dev->adev, (void **)&pqn->q->wptr_bo_gart);
+	}
+}
+
 void pqm_uninit(struct process_queue_manager *pqm)
 {
 	struct process_queue_node *pqn, *next;
 
 	list_for_each_entry_safe(pqn, next, &pqm->queues, process_queue_list) {
-		if (pqn->q && pqn->q->gws)
-			amdgpu_amdkfd_remove_gws_from_process(pqm->process->kgd_process_info,
-				pqn->q->gws);
+		if (pqn->q) {
+			struct kfd_process_device *pdd = kfd_get_process_device_data(pqn->q->device,
+										     pqm->process);
+			if (pdd) {
+				kfd_queue_unref_bo_vas(pdd, &pqn->q->properties);
+				kfd_queue_release_buffers(pdd, &pqn->q->properties);
+			} else {
+				WARN_ON(!pdd);
+			}
+			pqm_clean_queue_resource(pqm, pqn);
+		}
+
 		kfd_procfs_del_queue(pqn->q);
 		uninit_queue(pqn->q);
 		list_del(&pqn->process_queue_list);
@@ -178,15 +242,15 @@ void pqm_uninit(struct process_queue_manager *pqm)
 }
 
 static int init_user_queue(struct process_queue_manager *pqm,
-				struct kfd_dev *dev, struct queue **q,
+				struct kfd_node *dev, struct queue **q,
 				struct queue_properties *q_properties,
-				struct file *f, struct amdgpu_bo *wptr_bo,
 				unsigned int qid)
 {
 	int retval;
 
 	/* Doorbell initialized in user space*/
 	q_properties->doorbell_ptr = NULL;
+	q_properties->exception_status = KFD_EC_MASK(EC_QUEUE_NEW);
 
 	/* let DQM handle it*/
 	q_properties->vmid = 0;
@@ -199,7 +263,7 @@ static int init_user_queue(struct process_queue_manager *pqm,
 	(*q)->device = dev;
 	(*q)->process = pqm->process;
 
-	if (dev->shared_resources.enable_mes) {
+	if (dev->kfd->shared_resources.enable_mes) {
 		retval = amdgpu_amdkfd_alloc_gtt_mem(dev->adev,
 						AMDGPU_MES_GANG_CTX_SIZE,
 						&(*q)->gang_ctx_bo,
@@ -211,12 +275,29 @@ static int init_user_queue(struct process_queue_manager *pqm,
 			goto cleanup;
 		}
 		memset((*q)->gang_ctx_cpu_ptr, 0, AMDGPU_MES_GANG_CTX_SIZE);
-		(*q)->wptr_bo = wptr_bo;
+
+		/* Starting with GFX11, wptr BOs must be mapped to GART for MES to determine work
+		 * on unmapped queues for usermode queue oversubscription (no aggregated doorbell)
+		 */
+		if (dev->adev != amdgpu_ttm_adev(q_properties->wptr_bo->tbo.bdev)) {
+			pr_err("Queue memory allocated to wrong device\n");
+			retval = -EINVAL;
+			goto free_gang_ctx_bo;
+		}
+
+		retval = amdgpu_amdkfd_map_gtt_bo_to_gart(q_properties->wptr_bo,
+							  &(*q)->wptr_bo_gart);
+		if (retval) {
+			pr_err("Failed to map wptr bo to GART\n");
+			goto free_gang_ctx_bo;
+		}
 	}
 
 	pr_debug("PQM After init queue");
 	return 0;
 
+free_gang_ctx_bo:
+	amdgpu_amdkfd_free_gtt_mem(dev->adev, &(*q)->gang_ctx_bo);
 cleanup:
 	uninit_queue(*q);
 	*q = NULL;
@@ -224,11 +305,9 @@ cleanup:
 }
 
 int pqm_create_queue(struct process_queue_manager *pqm,
-			    struct kfd_dev *dev,
-			    struct file *f,
+			    struct kfd_node *dev,
 			    struct queue_properties *properties,
 			    unsigned int *qid,
-			    struct amdgpu_bo *wptr_bo,
 			    const struct kfd_criu_queue_priv_data *q_data,
 			    const void *restore_mqd,
 			    const void *restore_ctl_stack,
@@ -242,6 +321,15 @@ int pqm_create_queue(struct process_queue_manager *pqm,
 	enum kfd_queue_type type = properties->type;
 	unsigned int max_queues = 127; /* HWS limit */
 
+	/*
+	 * On GFX 9.4.3/9.5.0, increase the number of queues that
+	 * can be created to 255. No HWS limit on GFX 9.4.3/9.5.0.
+	 */
+	if (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 3) ||
+	    KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 4) ||
+	    KFD_GC_VERSION(dev) == IP_VERSION(9, 5, 0))
+		max_queues = 255;
+
 	q = NULL;
 	kq = NULL;
 
@@ -258,7 +346,7 @@ int pqm_create_queue(struct process_queue_manager *pqm,
 	 * Hence we also check the type as well
 	 */
 	if ((pdd->qpd.is_debug) || (type == KFD_QUEUE_TYPE_DIQ))
-		max_queues = dev->device_info.max_no_of_hqd/2;
+		max_queues = dev->kfd->device_info.max_no_of_hqd/2;
 
 	if (pdd->qpd.queue_count >= max_queues)
 		return -ENOSPC;
@@ -272,10 +360,26 @@ int pqm_create_queue(struct process_queue_manager *pqm,
 	if (retval != 0)
 		return retval;
 
+	/* Register process if this is the first queue */
 	if (list_empty(&pdd->qpd.queues_list) &&
 	    list_empty(&pdd->qpd.priv_queue_list))
 		dev->dqm->ops.register_process(dev->dqm, &pdd->qpd);
 
+	/* Allocate proc_ctx_bo only if MES is enabled and this is the first queue */
+	if (!pdd->proc_ctx_cpu_ptr && dev->kfd->shared_resources.enable_mes) {
+		retval = amdgpu_amdkfd_alloc_gtt_mem(dev->adev,
+						     AMDGPU_MES_PROC_CTX_SIZE,
+						     &pdd->proc_ctx_bo,
+						     &pdd->proc_ctx_gpu_addr,
+						     &pdd->proc_ctx_cpu_ptr,
+						     false);
+		if (retval) {
+			dev_err(dev->adev->dev, "failed to allocate process context bo\n");
+			return retval;
+		}
+		memset(pdd->proc_ctx_cpu_ptr, 0, AMDGPU_MES_PROC_CTX_SIZE);
+	}
+
 	pqn = kzalloc(sizeof(*pqn), GFP_KERNEL);
 	if (!pqn) {
 		retval = -ENOMEM;
@@ -285,13 +389,14 @@ int pqm_create_queue(struct process_queue_manager *pqm,
 	switch (type) {
 	case KFD_QUEUE_TYPE_SDMA:
 	case KFD_QUEUE_TYPE_SDMA_XGMI:
+	case KFD_QUEUE_TYPE_SDMA_BY_ENG_ID:
 		/* SDMA queues are always allocated statically no matter
 		 * which scheduler mode is used. We also do not need to
 		 * check whether a SDMA queue can be allocated here, because
 		 * allocate_sdma_queue() in create_queue() has the
 		 * corresponding check logic.
 		 */
-		retval = init_user_queue(pqm, dev, &q, properties, f, wptr_bo, *qid);
+		retval = init_user_queue(pqm, dev, &q, properties, *qid);
 		if (retval != 0)
 			goto err_create_queue;
 		pqn->q = q;
@@ -312,7 +417,7 @@ int pqm_create_queue(struct process_queue_manager *pqm,
 			goto err_create_queue;
 		}
 
-		retval = init_user_queue(pqm, dev, &q, properties, f, wptr_bo, *qid);
+		retval = init_user_queue(pqm, dev, &q, properties, *qid);
 		if (retval != 0)
 			goto err_create_queue;
 		pqn->q = q;
@@ -330,6 +435,10 @@ int pqm_create_queue(struct process_queue_manager *pqm,
 		kq->queue->properties.queue_id = *qid;
 		pqn->kq = kq;
 		pqn->q = NULL;
+		retval = kfd_process_drain_interrupts(pdd);
+		if (retval)
+			break;
+
 		retval = dev->dqm->ops.create_kernel_queue(dev->dqm,
 							kq, &pdd->qpd);
 		break;
@@ -339,22 +448,33 @@ int pqm_create_queue(struct process_queue_manager *pqm,
 	}
 
 	if (retval != 0) {
-		pr_err("Pasid 0x%x DQM create queue type %d failed. ret %d\n",
-			pqm->process->pasid, type, retval);
+		if ((type == KFD_QUEUE_TYPE_SDMA ||
+		    type == KFD_QUEUE_TYPE_SDMA_XGMI ||
+		    type == KFD_QUEUE_TYPE_SDMA_BY_ENG_ID) &&
+		    retval == -ENOMEM)
+			pr_warn("process pid %d DQM create queue type %d failed. ret %d\n",
+				pqm->process->lead_thread->pid, type, retval);
+		else
+			pr_err("process pid %d DQM create queue type %d failed. ret %d\n",
+				pqm->process->lead_thread->pid, type, retval);
 		goto err_create_queue;
 	}
 
-	if (q && p_doorbell_offset_in_process)
+	if (q && p_doorbell_offset_in_process) {
 		/* Return the doorbell offset within the doorbell page
 		 * to the caller so it can be passed up to user mode
 		 * (in bytes).
-		 * There are always 1024 doorbells per process, so in case
-		 * of 8-byte doorbells, there are two doorbell pages per
-		 * process.
+		 * relative doorbell index = Absolute doorbell index -
+		 * absolute index of first doorbell in the page.
 		 */
-		*p_doorbell_offset_in_process =
-			(q->properties.doorbell_off * sizeof(uint32_t)) &
-			(kfd_doorbell_process_slice(dev) - 1);
+		uint32_t first_db_index = amdgpu_doorbell_index_on_bar(pdd->dev->adev,
+								       pdd->qpd.proc_doorbells,
+								       0,
+								       pdd->dev->kfd->device_info.doorbell_size);
+
+		*p_doorbell_offset_in_process = (q->properties.doorbell_off
+						- first_db_index) * sizeof(uint32_t);
+	}
 
 	pr_debug("PQM After DQM create queue\n");
 
@@ -371,7 +491,7 @@ int pqm_create_queue(struct process_queue_manager *pqm,
 err_create_queue:
 	uninit_queue(q);
 	if (kq)
-		kernel_queue_uninit(kq, false);
+		kernel_queue_uninit(kq);
 	kfree(pqn);
 err_allocate_pqn:
 	/* check if queues list is empty unregister process from device */
@@ -387,7 +507,7 @@ int pqm_destroy_queue(struct process_queue_manager *pqm, unsigned int qid)
 	struct process_queue_node *pqn;
 	struct kfd_process_device *pdd;
 	struct device_queue_manager *dqm;
-	struct kfd_dev *dev;
+	struct kfd_node *dev;
 	int retval;
 
 	dqm = NULL;
@@ -418,34 +538,26 @@ int pqm_destroy_queue(struct process_queue_manager *pqm, unsigned int qid)
 		/* destroy kernel queue (DIQ) */
 		dqm = pqn->kq->dev->dqm;
 		dqm->ops.destroy_kernel_queue(dqm, pqn->kq, &pdd->qpd);
-		kernel_queue_uninit(pqn->kq, false);
+		kernel_queue_uninit(pqn->kq);
 	}
 
 	if (pqn->q) {
-		kfd_procfs_del_queue(pqn->q);
+		retval = kfd_queue_unref_bo_vas(pdd, &pqn->q->properties);
+		if (retval)
+			goto err_destroy_queue;
+
 		dqm = pqn->q->device->dqm;
 		retval = dqm->ops.destroy_queue(dqm, &pdd->qpd, pqn->q);
 		if (retval) {
 			pr_err("Pasid 0x%x destroy queue %d failed, ret %d\n",
-				pqm->process->pasid,
+				pdd->pasid,
 				pqn->q->properties.queue_id, retval);
-			if (retval != -ETIME)
+			if (retval != -ETIME && retval != -EIO)
 				goto err_destroy_queue;
 		}
-
-		if (pqn->q->gws) {
-			amdgpu_amdkfd_remove_gws_from_process(pqm->process->kgd_process_info,
-				pqn->q->gws);
-			pdd->qpd.num_gws = 0;
-		}
-
-		if (dev->shared_resources.enable_mes) {
-			amdgpu_amdkfd_free_gtt_mem(dev->adev,
-						   pqn->q->gang_ctx_bo);
-			if (pqn->q->wptr_bo)
-				amdgpu_amdkfd_free_gtt_mem(dev->adev, pqn->q->wptr_bo);
-
-		}
+		kfd_procfs_del_queue(pqn->q);
+		kfd_queue_release_buffers(pdd, &pqn->q->properties);
+		pqm_clean_queue_resource(pqm, pqn);
 		uninit_queue(pqn->q);
 	}
 
@@ -468,15 +580,47 @@ int pqm_update_queue_properties(struct process_queue_manager *pqm,
 	struct process_queue_node *pqn;
 
 	pqn = get_queue_by_qid(pqm, qid);
-	if (!pqn) {
+	if (!pqn || !pqn->q) {
 		pr_debug("No queue %d exists for update operation\n", qid);
 		return -EFAULT;
 	}
 
+	/*
+	 * Update with NULL ring address is used to disable the queue
+	 */
+	if (p->queue_address && p->queue_size) {
+		struct kfd_process_device *pdd;
+		struct amdgpu_vm *vm;
+		struct queue *q = pqn->q;
+		int err;
+
+		pdd = kfd_get_process_device_data(q->device, q->process);
+		if (!pdd)
+			return -ENODEV;
+		vm = drm_priv_to_vm(pdd->drm_priv);
+		err = amdgpu_bo_reserve(vm->root.bo, false);
+		if (err)
+			return err;
+
+		if (kfd_queue_buffer_get(vm, (void *)p->queue_address, &p->ring_bo,
+					 p->queue_size)) {
+			pr_debug("ring buf 0x%llx size 0x%llx not mapped on GPU\n",
+				 p->queue_address, p->queue_size);
+			return -EFAULT;
+		}
+
+		kfd_queue_unref_bo_va(vm, &pqn->q->properties.ring_bo);
+		kfd_queue_buffer_put(&pqn->q->properties.ring_bo);
+		amdgpu_bo_unreserve(vm->root.bo);
+
+		pqn->q->properties.ring_bo = p->ring_bo;
+	}
+
 	pqn->q->properties.queue_address = p->queue_address;
 	pqn->q->properties.queue_size = p->queue_size;
 	pqn->q->properties.queue_percent = p->queue_percent;
 	pqn->q->properties.priority = p->priority;
+	pqn->q->properties.pm4_target_xcc = p->pm4_target_xcc;
 
 	retval = pqn->q->device->dqm->ops.update_queue(pqn->q->device->dqm,
 							pqn->q, NULL);
@@ -498,8 +642,12 @@ int pqm_update_mqd(struct process_queue_manager *pqm,
 		return -EFAULT;
 	}
 
+	/* CUs are masked for debugger requirements so deny user mask  */
+	if (pqn->q->properties.is_dbg_wa && minfo && minfo->cu_mask.ptr)
+		return -EBUSY;
+
 	/* ASICs that have WGPs must enforce pairwise enabled mask checks. */
-	if (minfo && minfo->update_flag == UPDATE_FLAG_CU_MASK && minfo->cu_mask.ptr &&
+	if (minfo && minfo->cu_mask.ptr &&
 			KFD_GC_VERSION(pqn->q->device) >= IP_VERSION(10, 0, 0)) {
 		int i;
 
@@ -518,20 +666,10 @@ int pqm_update_mqd(struct process_queue_manager *pqm,
 	if (retval != 0)
 		return retval;
 
-	return 0;
-}
-
-struct kernel_queue *pqm_get_kernel_queue(
-					struct process_queue_manager *pqm,
-					unsigned int qid)
-{
-	struct process_queue_node *pqn;
+	if (minfo && minfo->cu_mask.ptr)
+		pqn->q->properties.is_user_cu_masked = true;
 
-	pqn = get_queue_by_qid(pqm, qid);
-	if (pqn && pqn->kq)
-		return pqn->kq;
-
-	return NULL;
+	return 0;
 }
 
 struct queue *pqm_get_user_queue(struct process_queue_manager *pqm,
@@ -565,6 +703,46 @@ int pqm_get_wave_state(struct process_queue_manager *pqm,
 						       save_area_used_size);
 }
 
+int pqm_get_queue_snapshot(struct process_queue_manager *pqm,
+			   uint64_t exception_clear_mask,
+			   void __user *buf,
+			   int *num_qss_entries,
+			   uint32_t *entry_size)
+{
+	struct process_queue_node *pqn;
+	struct kfd_queue_snapshot_entry src;
+	uint32_t tmp_entry_size = *entry_size, tmp_qss_entries = *num_qss_entries;
+	int r = 0;
+
+	*num_qss_entries = 0;
+	if (!(*entry_size))
+		return -EINVAL;
+
+	*entry_size = min_t(size_t, *entry_size, sizeof(struct kfd_queue_snapshot_entry));
+	mutex_lock(&pqm->process->event_mutex);
+
+	memset(&src, 0, sizeof(src));
+
+	list_for_each_entry(pqn, &pqm->queues, process_queue_list) {
+		if (!pqn->q)
+			continue;
+
+		if (*num_qss_entries < tmp_qss_entries) {
+			set_queue_snapshot_entry(pqn->q, exception_clear_mask, &src);
+
+			if (copy_to_user(buf, &src, *entry_size)) {
+				r = -EFAULT;
+				break;
+			}
+			buf += tmp_entry_size;
+		}
+		*num_qss_entries += 1;
+	}
+
+	mutex_unlock(&pqm->process->event_mutex);
+	return r;
+}
+
 static int get_queue_data_sizes(struct kfd_process_device *pdd,
 				struct queue *q,
 				uint32_t *mqd_size,
@@ -736,7 +914,10 @@ static int criu_checkpoint_queues_device(struct kfd_process_device *pdd,
 
 		q_data = (struct kfd_criu_queue_priv_data *)q_private_data;
 
-		/* data stored in this order: priv_data, mqd, ctl_stack */
+		/*
+		 * data stored in this order:
+		 * priv_data, mqd[xcc0], mqd[xcc1],..., ctl_stack[xcc0], ctl_stack[xcc1]...
+		 */
 		q_data->mqd_size = mqd_size;
 		q_data->ctl_stack_size = ctl_stack_size;
 
@@ -785,7 +966,7 @@ int kfd_criu_checkpoint_queues(struct kfd_process *p,
 }
 
 static void set_queue_properties_from_criu(struct queue_properties *qp,
-					  struct kfd_criu_queue_priv_data *q_data)
+					  struct kfd_criu_queue_priv_data *q_data, uint32_t num_xcc)
 {
 	qp->is_interop = false;
 	qp->queue_percent = q_data->q_percent;
@@ -798,7 +979,11 @@ static void set_queue_properties_from_criu(struct queue_properties *qp,
 	qp->eop_ring_buffer_size = q_data->eop_ring_buffer_size;
 	qp->ctx_save_restore_area_address = q_data->ctx_save_restore_area_address;
 	qp->ctx_save_restore_area_size = q_data->ctx_save_restore_area_size;
-	qp->ctl_stack_size = q_data->ctl_stack_size;
+	if (q_data->type == KFD_QUEUE_TYPE_COMPUTE)
+		qp->ctl_stack_size = q_data->ctl_stack_size / num_xcc;
+	else
+		qp->ctl_stack_size = q_data->ctl_stack_size;
+
 	qp->type = q_data->type;
 	qp->format = q_data->format;
 }
@@ -858,23 +1043,19 @@ int kfd_criu_restore_queue(struct kfd_process *p,
 		goto exit;
 	}
 
-	if (!pdd->doorbell_index &&
-	    kfd_alloc_process_doorbells(pdd->dev, &pdd->doorbell_index) < 0) {
-		ret = -ENOMEM;
-		goto exit;
-	}
-
-	/* data stored in this order: mqd, ctl_stack */
+	/*
+	 * data stored in this order:
+	 * mqd[xcc0], mqd[xcc1],..., ctl_stack[xcc0], ctl_stack[xcc1]...
+	 */
 	mqd = q_extra_data;
 	ctl_stack = mqd + q_data->mqd_size;
 
 	memset(&qp, 0, sizeof(qp));
-	set_queue_properties_from_criu(&qp, q_data);
+	set_queue_properties_from_criu(&qp, q_data, NUM_XCC(pdd->dev->adev->gfx.xcc_mask));
 
 	print_queue_properties(&qp);
 
-	ret = pqm_create_queue(&p->pqm, pdd->dev, NULL, &qp, &queue_id, NULL, q_data, mqd, ctl_stack,
-				NULL);
+	ret = pqm_create_queue(&p->pqm, pdd->dev, &qp, &queue_id, q_data, mqd, ctl_stack, NULL);
 	if (ret) {
 		pr_err("Failed to create new queue err:%d\n", ret);
 		goto exit;
@@ -890,6 +1071,7 @@ exit:
 		pr_debug("Queue id %d was restored successfully\n", queue_id);
 
 	kfree(q_data);
+	kfree(q_extra_data);
 
 	return ret;
 }
@@ -927,7 +1109,9 @@ int pqm_debugfs_mqds(struct seq_file *m, void *data)
 	struct queue *q;
 	enum KFD_MQD_TYPE mqd_type;
 	struct mqd_manager *mqd_mgr;
-	int r = 0;
+	int r = 0, xcc, num_xccs = 1;
+	void *mqd;
+	uint64_t size = 0;
 
 	list_for_each_entry(pqn, &pqm->queues, process_queue_list) {
 		if (pqn->q) {
@@ -943,6 +1127,7 @@ int pqm_debugfs_mqds(struct seq_file *m, void *data)
 				seq_printf(m, "  Compute queue on device %x\n",
 					   q->device->id);
 				mqd_type = KFD_MQD_TYPE_CP;
+				num_xccs = NUM_XCC(q->device->xcc_mask);
 				break;
 			default:
 				seq_printf(m,
@@ -951,6 +1136,8 @@ int pqm_debugfs_mqds(struct seq_file *m, void *data)
 				continue;
 			}
 			mqd_mgr = q->device->dqm->mqd_mgrs[mqd_type];
+			size = mqd_mgr->mqd_stride(mqd_mgr,
+							&q->properties);
 		} else if (pqn->kq) {
 			q = pqn->kq->queue;
 			mqd_mgr = pqn->kq->mqd_mgr;
@@ -972,9 +1159,12 @@ int pqm_debugfs_mqds(struct seq_file *m, void *data)
 			continue;
 		}
 
-		r = mqd_mgr->debugfs_show_mqd(m, q->mqd);
-		if (r != 0)
-			break;
+		for (xcc = 0; xcc < num_xccs; xcc++) {
+			mqd = q->mqd + size * xcc;
+			r = mqd_mgr->debugfs_show_mqd(m, mqd);
+			if (r != 0)
+				break;
+		}
 	}
 
 	return r;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_queue.c b/drivers/gpu/drm/amd/amdkfd/kfd_queue.c
index 0f6992b1895c..f1e7583650c4 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_queue.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_queue.c
@@ -24,6 +24,8 @@
 
 #include <linux/slab.h>
 #include "kfd_priv.h"
+#include "kfd_topology.h"
+#include "kfd_svm.h"
 
 void print_queue_properties(struct queue_properties *q)
 {
@@ -82,3 +84,386 @@ void uninit_queue(struct queue *q)
 {
 	kfree(q);
 }
+
+#if IS_ENABLED(CONFIG_HSA_AMD_SVM)
+
+static int kfd_queue_buffer_svm_get(struct kfd_process_device *pdd, u64 addr, u64 size)
+{
+	struct kfd_process *p = pdd->process;
+	struct list_head update_list;
+	struct svm_range *prange;
+	int ret = -EINVAL;
+
+	INIT_LIST_HEAD(&update_list);
+	addr >>= PAGE_SHIFT;
+	size >>= PAGE_SHIFT;
+
+	mutex_lock(&p->svms.lock);
+
+	/*
+	 * range may split to multiple svm pranges aligned to granularity boundaery.
+	 */
+	while (size) {
+		uint32_t gpuid, gpuidx;
+		int r;
+
+		prange = svm_range_from_addr(&p->svms, addr, NULL);
+		if (!prange)
+			break;
+
+		if (!prange->mapped_to_gpu)
+			break;
+
+		r = kfd_process_gpuid_from_node(p, pdd->dev, &gpuid, &gpuidx);
+		if (r < 0)
+			break;
+		if (!test_bit(gpuidx, prange->bitmap_access) &&
+		    !test_bit(gpuidx, prange->bitmap_aip))
+			break;
+
+		if (!(prange->flags & KFD_IOCTL_SVM_FLAG_GPU_ALWAYS_MAPPED))
+			break;
+
+		list_add(&prange->update_list, &update_list);
+
+		if (prange->last - prange->start + 1 >= size) {
+			size = 0;
+			break;
+		}
+
+		size -= prange->last - prange->start + 1;
+		addr += prange->last - prange->start + 1;
+	}
+	if (size) {
+		pr_debug("[0x%llx 0x%llx] not registered\n", addr, addr + size - 1);
+		goto out_unlock;
+	}
+
+	list_for_each_entry(prange, &update_list, update_list)
+		atomic_inc(&prange->queue_refcount);
+	ret = 0;
+
+out_unlock:
+	mutex_unlock(&p->svms.lock);
+	return ret;
+}
+
+static void kfd_queue_buffer_svm_put(struct kfd_process_device *pdd, u64 addr, u64 size)
+{
+	struct kfd_process *p = pdd->process;
+	struct svm_range *prange, *pchild;
+	struct interval_tree_node *node;
+	unsigned long last;
+
+	addr >>= PAGE_SHIFT;
+	last = addr + (size >> PAGE_SHIFT) - 1;
+
+	mutex_lock(&p->svms.lock);
+
+	node = interval_tree_iter_first(&p->svms.objects, addr, last);
+	while (node) {
+		struct interval_tree_node *next_node;
+		unsigned long next_start;
+
+		prange = container_of(node, struct svm_range, it_node);
+		next_node = interval_tree_iter_next(node, addr, last);
+		next_start = min(node->last, last) + 1;
+
+		if (atomic_add_unless(&prange->queue_refcount, -1, 0)) {
+			list_for_each_entry(pchild, &prange->child_list, child_list)
+				atomic_add_unless(&pchild->queue_refcount, -1, 0);
+		}
+
+		node = next_node;
+		addr = next_start;
+	}
+
+	mutex_unlock(&p->svms.lock);
+}
+#else
+
+static int kfd_queue_buffer_svm_get(struct kfd_process_device *pdd, u64 addr, u64 size)
+{
+	return -EINVAL;
+}
+
+static void kfd_queue_buffer_svm_put(struct kfd_process_device *pdd, u64 addr, u64 size)
+{
+}
+
+#endif
+
+int kfd_queue_buffer_get(struct amdgpu_vm *vm, void __user *addr, struct amdgpu_bo **pbo,
+			 u64 expected_size)
+{
+	struct amdgpu_bo_va_mapping *mapping;
+	u64 user_addr;
+	u64 size;
+
+	user_addr = (u64)addr >> AMDGPU_GPU_PAGE_SHIFT;
+	size = expected_size >> AMDGPU_GPU_PAGE_SHIFT;
+
+	mapping = amdgpu_vm_bo_lookup_mapping(vm, user_addr);
+	if (!mapping)
+		goto out_err;
+
+	if (user_addr != mapping->start ||
+	    (size != 0 && user_addr + size - 1 != mapping->last)) {
+		pr_debug("expected size 0x%llx not equal to mapping addr 0x%llx size 0x%llx\n",
+			expected_size, mapping->start << AMDGPU_GPU_PAGE_SHIFT,
+			(mapping->last - mapping->start + 1) << AMDGPU_GPU_PAGE_SHIFT);
+		goto out_err;
+	}
+
+	*pbo = amdgpu_bo_ref(mapping->bo_va->base.bo);
+	mapping->bo_va->queue_refcount++;
+	return 0;
+
+out_err:
+	*pbo = NULL;
+	return -EINVAL;
+}
+
+/* FIXME: remove this function, just call amdgpu_bo_unref directly */
+void kfd_queue_buffer_put(struct amdgpu_bo **bo)
+{
+	amdgpu_bo_unref(bo);
+}
+
+int kfd_queue_acquire_buffers(struct kfd_process_device *pdd, struct queue_properties *properties)
+{
+	struct kfd_topology_device *topo_dev;
+	u64 expected_queue_size;
+	struct amdgpu_vm *vm;
+	u32 total_cwsr_size;
+	int err;
+
+	topo_dev = kfd_topology_device_by_id(pdd->dev->id);
+	if (!topo_dev)
+		return -EINVAL;
+
+	/* AQL queues on GFX7 and GFX8 appear twice their actual size */
+	if (properties->type == KFD_QUEUE_TYPE_COMPUTE &&
+	    properties->format == KFD_QUEUE_FORMAT_AQL &&
+	    topo_dev->node_props.gfx_target_version >= 70000 &&
+	    topo_dev->node_props.gfx_target_version < 90000)
+		expected_queue_size = properties->queue_size / 2;
+	else
+		expected_queue_size = properties->queue_size;
+
+	vm = drm_priv_to_vm(pdd->drm_priv);
+	err = amdgpu_bo_reserve(vm->root.bo, false);
+	if (err)
+		return err;
+
+	err = kfd_queue_buffer_get(vm, properties->write_ptr, &properties->wptr_bo, PAGE_SIZE);
+	if (err)
+		goto out_err_unreserve;
+
+	err = kfd_queue_buffer_get(vm, properties->read_ptr, &properties->rptr_bo, PAGE_SIZE);
+	if (err)
+		goto out_err_unreserve;
+
+	err = kfd_queue_buffer_get(vm, (void *)properties->queue_address,
+				   &properties->ring_bo, expected_queue_size);
+	if (err)
+		goto out_err_unreserve;
+
+	/* only compute queue requires EOP buffer and CWSR area */
+	if (properties->type != KFD_QUEUE_TYPE_COMPUTE)
+		goto out_unreserve;
+
+	/* EOP buffer is not required for all ASICs */
+	if (properties->eop_ring_buffer_address) {
+		if (properties->eop_ring_buffer_size != topo_dev->node_props.eop_buffer_size) {
+			pr_debug("queue eop bo size 0x%x not equal to node eop buf size 0x%x\n",
+				properties->eop_ring_buffer_size,
+				topo_dev->node_props.eop_buffer_size);
+			err = -EINVAL;
+			goto out_err_unreserve;
+		}
+		err = kfd_queue_buffer_get(vm, (void *)properties->eop_ring_buffer_address,
+					   &properties->eop_buf_bo,
+					   properties->eop_ring_buffer_size);
+		if (err)
+			goto out_err_unreserve;
+	}
+
+	if (properties->ctl_stack_size != topo_dev->node_props.ctl_stack_size) {
+		pr_debug("queue ctl stack size 0x%x not equal to node ctl stack size 0x%x\n",
+			properties->ctl_stack_size,
+			topo_dev->node_props.ctl_stack_size);
+		err = -EINVAL;
+		goto out_err_unreserve;
+	}
+
+	if (properties->ctx_save_restore_area_size < topo_dev->node_props.cwsr_size) {
+		pr_debug("queue cwsr size 0x%x not sufficient for node cwsr size 0x%x\n",
+			properties->ctx_save_restore_area_size,
+			topo_dev->node_props.cwsr_size);
+		err = -EINVAL;
+		goto out_err_unreserve;
+	}
+
+	total_cwsr_size = (properties->ctx_save_restore_area_size +
+			   topo_dev->node_props.debug_memory_size) * NUM_XCC(pdd->dev->xcc_mask);
+	total_cwsr_size = ALIGN(total_cwsr_size, PAGE_SIZE);
+
+	err = kfd_queue_buffer_get(vm, (void *)properties->ctx_save_restore_area_address,
+				   &properties->cwsr_bo, total_cwsr_size);
+	if (!err)
+		goto out_unreserve;
+
+	amdgpu_bo_unreserve(vm->root.bo);
+
+	err = kfd_queue_buffer_svm_get(pdd, properties->ctx_save_restore_area_address,
+				       total_cwsr_size);
+	if (err)
+		goto out_err_release;
+
+	return 0;
+
+out_unreserve:
+	amdgpu_bo_unreserve(vm->root.bo);
+	return 0;
+
+out_err_unreserve:
+	amdgpu_bo_unreserve(vm->root.bo);
+out_err_release:
+	/* FIXME: make a _locked version of this that can be called before
+	 * dropping the VM reservation.
+	 */
+	kfd_queue_unref_bo_vas(pdd, properties);
+	kfd_queue_release_buffers(pdd, properties);
+	return err;
+}
+
+int kfd_queue_release_buffers(struct kfd_process_device *pdd, struct queue_properties *properties)
+{
+	struct kfd_topology_device *topo_dev;
+	u32 total_cwsr_size;
+
+	kfd_queue_buffer_put(&properties->wptr_bo);
+	kfd_queue_buffer_put(&properties->rptr_bo);
+	kfd_queue_buffer_put(&properties->ring_bo);
+	kfd_queue_buffer_put(&properties->eop_buf_bo);
+	kfd_queue_buffer_put(&properties->cwsr_bo);
+
+	topo_dev = kfd_topology_device_by_id(pdd->dev->id);
+	if (!topo_dev)
+		return -EINVAL;
+	total_cwsr_size = (properties->ctx_save_restore_area_size +
+			   topo_dev->node_props.debug_memory_size) * NUM_XCC(pdd->dev->xcc_mask);
+	total_cwsr_size = ALIGN(total_cwsr_size, PAGE_SIZE);
+
+	kfd_queue_buffer_svm_put(pdd, properties->ctx_save_restore_area_address, total_cwsr_size);
+	return 0;
+}
+
+void kfd_queue_unref_bo_va(struct amdgpu_vm *vm, struct amdgpu_bo **bo)
+{
+	if (*bo) {
+		struct amdgpu_bo_va *bo_va;
+
+		bo_va = amdgpu_vm_bo_find(vm, *bo);
+		if (bo_va && bo_va->queue_refcount)
+			bo_va->queue_refcount--;
+	}
+}
+
+int kfd_queue_unref_bo_vas(struct kfd_process_device *pdd,
+			   struct queue_properties *properties)
+{
+	struct amdgpu_vm *vm;
+	int err;
+
+	vm = drm_priv_to_vm(pdd->drm_priv);
+	err = amdgpu_bo_reserve(vm->root.bo, false);
+	if (err)
+		return err;
+
+	kfd_queue_unref_bo_va(vm, &properties->wptr_bo);
+	kfd_queue_unref_bo_va(vm, &properties->rptr_bo);
+	kfd_queue_unref_bo_va(vm, &properties->ring_bo);
+	kfd_queue_unref_bo_va(vm, &properties->eop_buf_bo);
+	kfd_queue_unref_bo_va(vm, &properties->cwsr_bo);
+
+	amdgpu_bo_unreserve(vm->root.bo);
+	return 0;
+}
+
+#define SGPR_SIZE_PER_CU	0x4000
+#define LDS_SIZE_PER_CU		0x10000
+#define HWREG_SIZE_PER_CU	0x1000
+#define DEBUGGER_BYTES_ALIGN	64
+#define DEBUGGER_BYTES_PER_WAVE	32
+
+static u32 kfd_get_vgpr_size_per_cu(u32 gfxv)
+{
+	u32 vgpr_size = 0x40000;
+
+	if (gfxv == 90402 ||			/* GFX_VERSION_AQUA_VANJARAM */
+	    gfxv == 90010 ||			/* GFX_VERSION_ALDEBARAN */
+	    gfxv == 90008 ||			/* GFX_VERSION_ARCTURUS */
+	    gfxv == 90500)
+		vgpr_size = 0x80000;
+	else if (gfxv == 110000 ||		/* GFX_VERSION_PLUM_BONITO */
+		 gfxv == 110001 ||		/* GFX_VERSION_WHEAT_NAS */
+		 gfxv == 120000 ||		/* GFX_VERSION_GFX1200 */
+		 gfxv == 120001)		/* GFX_VERSION_GFX1201 */
+		vgpr_size = 0x60000;
+
+	return vgpr_size;
+}
+
+#define WG_CONTEXT_DATA_SIZE_PER_CU(gfxv, props)	\
+	(kfd_get_vgpr_size_per_cu(gfxv) + SGPR_SIZE_PER_CU +\
+	 (((gfxv) == 90500) ? (props->lds_size_in_kb << 10) : LDS_SIZE_PER_CU) +\
+	 HWREG_SIZE_PER_CU)
+
+#define CNTL_STACK_BYTES_PER_WAVE(gfxv)	\
+	((gfxv) >= 100100 ? 12 : 8)	/* GFX_VERSION_NAVI10*/
+
+#define SIZEOF_HSA_USER_CONTEXT_SAVE_AREA_HEADER 40
+
+void kfd_queue_ctx_save_restore_size(struct kfd_topology_device *dev)
+{
+	struct kfd_node_properties *props = &dev->node_props;
+	u32 gfxv = props->gfx_target_version;
+	u32 ctl_stack_size;
+	u32 wg_data_size;
+	u32 wave_num;
+	u32 cu_num;
+
+	if (gfxv < 80001)	/* GFX_VERSION_CARRIZO */
+		return;
+
+	cu_num = props->simd_count / props->simd_per_cu / NUM_XCC(dev->gpu->xcc_mask);
+	wave_num = (gfxv < 100100) ?	/* GFX_VERSION_NAVI10 */
+		    min(cu_num * 40, props->array_count / props->simd_arrays_per_engine * 512)
+		    : cu_num * 32;
+
+	wg_data_size = ALIGN(cu_num * WG_CONTEXT_DATA_SIZE_PER_CU(gfxv, props), PAGE_SIZE);
+	ctl_stack_size = wave_num * CNTL_STACK_BYTES_PER_WAVE(gfxv) + 8;
+	ctl_stack_size = ALIGN(SIZEOF_HSA_USER_CONTEXT_SAVE_AREA_HEADER + ctl_stack_size,
+			       PAGE_SIZE);
+
+	if ((gfxv / 10000 * 10000) == 100000) {
+		/* HW design limits control stack size to 0x7000.
+		 * This is insufficient for theoretical PM4 cases
+		 * but sufficient for AQL, limited by SPI events.
+		 */
+		ctl_stack_size = min(ctl_stack_size, 0x7000);
+	}
+
+	props->ctl_stack_size = ctl_stack_size;
+	props->debug_memory_size = ALIGN(wave_num * DEBUGGER_BYTES_PER_WAVE, DEBUGGER_BYTES_ALIGN);
+	props->cwsr_size = ctl_stack_size + wg_data_size;
+
+	if (gfxv == 80002)	/* GFX_VERSION_TONGA */
+		props->eop_buffer_size = 0x8000;
+	else if (gfxv == 90402)	/* GFX_VERSION_AQUA_VANJARAM */
+		props->eop_buffer_size = 4096;
+	else if (gfxv >= 80000)
+		props->eop_buffer_size = 4096;
+}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
index 0472b56de245..a499449fcb06 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
@@ -29,6 +29,7 @@
 #include "amdgpu_vm.h"
 #include "kfd_priv.h"
 #include "kfd_smi_events.h"
+#include "amdgpu_reset.h"
 
 struct kfd_smi_client {
 	struct list_head list;
@@ -36,14 +37,14 @@ struct kfd_smi_client {
 	wait_queue_head_t wait_queue;
 	/* events enabled */
 	uint64_t events;
-	struct kfd_dev *dev;
+	struct kfd_node *dev;
 	spinlock_t lock;
 	struct rcu_head rcu;
 	pid_t pid;
 	bool suser;
 };
 
-#define MAX_KFIFO_SIZE	1024
+#define KFD_MAX_KFIFO_SIZE	8192
 
 static __poll_t kfd_smi_ev_poll(struct file *, struct poll_table_struct *);
 static ssize_t kfd_smi_ev_read(struct file *, char __user *, size_t, loff_t *);
@@ -85,7 +86,7 @@ static ssize_t kfd_smi_ev_read(struct file *filep, char __user *user,
 	struct kfd_smi_client *client = filep->private_data;
 	unsigned char *buf;
 
-	size = min_t(size_t, size, MAX_KFIFO_SIZE);
+	size = min_t(size_t, size, KFD_MAX_KFIFO_SIZE);
 	buf = kmalloc(size, GFP_KERNEL);
 	if (!buf)
 		return -ENOMEM;
@@ -149,7 +150,7 @@ static void kfd_smi_ev_client_free(struct rcu_head *p)
 static int kfd_smi_ev_release(struct inode *inode, struct file *filep)
 {
 	struct kfd_smi_client *client = filep->private_data;
-	struct kfd_dev *dev = client->dev;
+	struct kfd_node *dev = client->dev;
 
 	spin_lock(&dev->smi_lock);
 	list_del_rcu(&client->list);
@@ -162,16 +163,15 @@ static int kfd_smi_ev_release(struct inode *inode, struct file *filep)
 static bool kfd_smi_ev_enabled(pid_t pid, struct kfd_smi_client *client,
 			       unsigned int event)
 {
-	uint64_t all = KFD_SMI_EVENT_MASK_FROM_INDEX(KFD_SMI_EVENT_ALL_PROCESS);
 	uint64_t events = READ_ONCE(client->events);
 
-	if (pid && client->pid != pid && !(client->suser && (events & all)))
+	if (pid && client->pid != pid && !client->suser)
 		return false;
 
 	return events & KFD_SMI_EVENT_MASK_FROM_INDEX(event);
 }
 
-static void add_event_to_kfifo(pid_t pid, struct kfd_dev *dev,
+static void add_event_to_kfifo(pid_t pid, struct kfd_node *dev,
 			       unsigned int smi_event, char *event_msg, int len)
 {
 	struct kfd_smi_client *client;
@@ -196,7 +196,7 @@ static void add_event_to_kfifo(pid_t pid, struct kfd_dev *dev,
 }
 
 __printf(4, 5)
-static void kfd_smi_event_add(pid_t pid, struct kfd_dev *dev,
+static void kfd_smi_event_add(pid_t pid, struct kfd_node *dev,
 			      unsigned int event, char *fmt, ...)
 {
 	char fifo_in[KFD_SMI_EVENT_MSG_SIZE];
@@ -215,9 +215,11 @@ static void kfd_smi_event_add(pid_t pid, struct kfd_dev *dev,
 	add_event_to_kfifo(pid, dev, event, fifo_in, len);
 }
 
-void kfd_smi_event_update_gpu_reset(struct kfd_dev *dev, bool post_reset)
+void kfd_smi_event_update_gpu_reset(struct kfd_node *dev, bool post_reset,
+				    struct amdgpu_reset_context *reset_context)
 {
 	unsigned int event;
+	char reset_cause[64];
 
 	if (post_reset) {
 		event = KFD_SMI_EVENT_GPU_POST_RESET;
@@ -225,83 +227,92 @@ void kfd_smi_event_update_gpu_reset(struct kfd_dev *dev, bool post_reset)
 		event = KFD_SMI_EVENT_GPU_PRE_RESET;
 		++(dev->reset_seq_num);
 	}
-	kfd_smi_event_add(0, dev, event, "%x\n", dev->reset_seq_num);
+
+	memset(reset_cause, 0, sizeof(reset_cause));
+
+	if (reset_context)
+		amdgpu_reset_get_desc(reset_context, reset_cause,
+				      sizeof(reset_cause));
+
+	kfd_smi_event_add(0, dev, event, KFD_EVENT_FMT_UPDATE_GPU_RESET(
+			  dev->reset_seq_num, reset_cause));
 }
 
-void kfd_smi_event_update_thermal_throttling(struct kfd_dev *dev,
+void kfd_smi_event_update_thermal_throttling(struct kfd_node *dev,
 					     uint64_t throttle_bitmask)
 {
-	kfd_smi_event_add(0, dev, KFD_SMI_EVENT_THERMAL_THROTTLE, "%llx:%llx\n",
+	kfd_smi_event_add(0, dev, KFD_SMI_EVENT_THERMAL_THROTTLE, KFD_EVENT_FMT_THERMAL_THROTTLING(
 			  throttle_bitmask,
-			  amdgpu_dpm_get_thermal_throttling_counter(dev->adev));
+			  amdgpu_dpm_get_thermal_throttling_counter(dev->adev)));
 }
 
-void kfd_smi_event_update_vmfault(struct kfd_dev *dev, uint16_t pasid)
+void kfd_smi_event_update_vmfault(struct kfd_node *dev, uint16_t pasid)
 {
-	struct amdgpu_task_info task_info;
-
-	memset(&task_info, 0, sizeof(struct amdgpu_task_info));
-	amdgpu_vm_get_task_info(dev->adev, pasid, &task_info);
-	/* Report VM faults from user applications, not retry from kernel */
-	if (!task_info.pid)
-		return;
-
-	kfd_smi_event_add(0, dev, KFD_SMI_EVENT_VMFAULT, "%x:%s\n",
-			  task_info.pid, task_info.task_name);
+	struct amdgpu_task_info *task_info;
+
+	task_info = amdgpu_vm_get_task_info_pasid(dev->adev, pasid);
+	if (task_info) {
+		/* Report VM faults from user applications, not retry from kernel */
+		if (task_info->task.pid)
+			kfd_smi_event_add(0, dev, KFD_SMI_EVENT_VMFAULT, KFD_EVENT_FMT_VMFAULT(
+					  task_info->task.pid, task_info->task.comm));
+		amdgpu_vm_put_task_info(task_info);
+	}
 }
 
-void kfd_smi_event_page_fault_start(struct kfd_dev *dev, pid_t pid,
+void kfd_smi_event_page_fault_start(struct kfd_node *node, pid_t pid,
 				    unsigned long address, bool write_fault,
 				    ktime_t ts)
 {
-	kfd_smi_event_add(pid, dev, KFD_SMI_EVENT_PAGE_FAULT_START,
-			  "%lld -%d @%lx(%x) %c\n", ktime_to_ns(ts), pid,
-			  address, dev->id, write_fault ? 'W' : 'R');
+	kfd_smi_event_add(pid, node, KFD_SMI_EVENT_PAGE_FAULT_START,
+			  KFD_EVENT_FMT_PAGEFAULT_START(ktime_to_ns(ts), pid,
+			  address, node->id, write_fault ? 'W' : 'R'));
 }
 
-void kfd_smi_event_page_fault_end(struct kfd_dev *dev, pid_t pid,
+void kfd_smi_event_page_fault_end(struct kfd_node *node, pid_t pid,
 				  unsigned long address, bool migration)
 {
-	kfd_smi_event_add(pid, dev, KFD_SMI_EVENT_PAGE_FAULT_END,
-			  "%lld -%d @%lx(%x) %c\n", ktime_get_boottime_ns(),
-			  pid, address, dev->id, migration ? 'M' : 'U');
+	kfd_smi_event_add(pid, node, KFD_SMI_EVENT_PAGE_FAULT_END,
+			  KFD_EVENT_FMT_PAGEFAULT_END(ktime_get_boottime_ns(),
+			  pid, address, node->id, migration ? 'M' : 'U'));
 }
 
-void kfd_smi_event_migration_start(struct kfd_dev *dev, pid_t pid,
+void kfd_smi_event_migration_start(struct kfd_node *node, pid_t pid,
 				   unsigned long start, unsigned long end,
 				   uint32_t from, uint32_t to,
 				   uint32_t prefetch_loc, uint32_t preferred_loc,
 				   uint32_t trigger)
 {
-	kfd_smi_event_add(pid, dev, KFD_SMI_EVENT_MIGRATE_START,
-			  "%lld -%d @%lx(%lx) %x->%x %x:%x %d\n",
+	kfd_smi_event_add(pid, node, KFD_SMI_EVENT_MIGRATE_START,
+			  KFD_EVENT_FMT_MIGRATE_START(
 			  ktime_get_boottime_ns(), pid, start, end - start,
-			  from, to, prefetch_loc, preferred_loc, trigger);
+			  from, to, prefetch_loc, preferred_loc, trigger));
 }
 
-void kfd_smi_event_migration_end(struct kfd_dev *dev, pid_t pid,
+void kfd_smi_event_migration_end(struct kfd_node *node, pid_t pid,
 				 unsigned long start, unsigned long end,
-				 uint32_t from, uint32_t to, uint32_t trigger)
+				 uint32_t from, uint32_t to, uint32_t trigger,
+				 int error_code)
 {
-	kfd_smi_event_add(pid, dev, KFD_SMI_EVENT_MIGRATE_END,
-			  "%lld -%d @%lx(%lx) %x->%x %d\n",
+	kfd_smi_event_add(pid, node, KFD_SMI_EVENT_MIGRATE_END,
+			  KFD_EVENT_FMT_MIGRATE_END(
 			  ktime_get_boottime_ns(), pid, start, end - start,
-			  from, to, trigger);
+			  from, to, trigger, error_code));
 }
 
-void kfd_smi_event_queue_eviction(struct kfd_dev *dev, pid_t pid,
+void kfd_smi_event_queue_eviction(struct kfd_node *node, pid_t pid,
 				  uint32_t trigger)
 {
-	kfd_smi_event_add(pid, dev, KFD_SMI_EVENT_QUEUE_EVICTION,
-			  "%lld -%d %x %d\n", ktime_get_boottime_ns(), pid,
-			  dev->id, trigger);
+	kfd_smi_event_add(pid, node, KFD_SMI_EVENT_QUEUE_EVICTION,
+			  KFD_EVENT_FMT_QUEUE_EVICTION(ktime_get_boottime_ns(), pid,
+			  node->id, trigger));
 }
 
-void kfd_smi_event_queue_restore(struct kfd_dev *dev, pid_t pid)
+void kfd_smi_event_queue_restore(struct kfd_node *node, pid_t pid)
 {
-	kfd_smi_event_add(pid, dev, KFD_SMI_EVENT_QUEUE_RESTORE,
-			  "%lld -%d %x\n", ktime_get_boottime_ns(), pid,
-			  dev->id);
+	kfd_smi_event_add(pid, node, KFD_SMI_EVENT_QUEUE_RESTORE,
+			  KFD_EVENT_FMT_QUEUE_RESTORE(ktime_get_boottime_ns(), pid,
+			  node->id, 0));
 }
 
 void kfd_smi_event_queue_restore_rescheduled(struct mm_struct *mm)
@@ -318,22 +329,43 @@ void kfd_smi_event_queue_restore_rescheduled(struct mm_struct *mm)
 
 		kfd_smi_event_add(p->lead_thread->pid, pdd->dev,
 				  KFD_SMI_EVENT_QUEUE_RESTORE,
-				  "%lld -%d %x %c\n", ktime_get_boottime_ns(),
-				  p->lead_thread->pid, pdd->dev->id, 'R');
+				  KFD_EVENT_FMT_QUEUE_RESTORE(ktime_get_boottime_ns(),
+				  p->lead_thread->pid, pdd->dev->id, 'R'));
 	}
 	kfd_unref_process(p);
 }
 
-void kfd_smi_event_unmap_from_gpu(struct kfd_dev *dev, pid_t pid,
+void kfd_smi_event_unmap_from_gpu(struct kfd_node *node, pid_t pid,
 				  unsigned long address, unsigned long last,
 				  uint32_t trigger)
 {
-	kfd_smi_event_add(pid, dev, KFD_SMI_EVENT_UNMAP_FROM_GPU,
-			  "%lld -%d @%lx(%lx) %x %d\n", ktime_get_boottime_ns(),
-			  pid, address, last - address + 1, dev->id, trigger);
+	kfd_smi_event_add(pid, node, KFD_SMI_EVENT_UNMAP_FROM_GPU,
+			  KFD_EVENT_FMT_UNMAP_FROM_GPU(ktime_get_boottime_ns(),
+			  pid, address, last - address + 1, node->id, trigger));
+}
+
+void kfd_smi_event_process(struct kfd_process_device *pdd, bool start)
+{
+	struct amdgpu_task_info *task_info;
+	struct amdgpu_vm *avm;
+
+	if (!pdd->drm_priv)
+		return;
+
+	avm = drm_priv_to_vm(pdd->drm_priv);
+	task_info = amdgpu_vm_get_task_info_vm(avm);
+
+	if (task_info) {
+		kfd_smi_event_add(0, pdd->dev,
+				  start ? KFD_SMI_EVENT_PROCESS_START :
+				  KFD_SMI_EVENT_PROCESS_END,
+				  KFD_EVENT_FMT_PROCESS(task_info->task.pid,
+				  task_info->task.comm));
+		amdgpu_vm_put_task_info(task_info);
+	}
 }
 
-int kfd_smi_event_open(struct kfd_dev *dev, uint32_t *fd)
+int kfd_smi_event_open(struct kfd_node *dev, uint32_t *fd)
 {
 	struct kfd_smi_client *client;
 	int ret;
@@ -343,7 +375,7 @@ int kfd_smi_event_open(struct kfd_dev *dev, uint32_t *fd)
 		return -ENOMEM;
 	INIT_LIST_HEAD(&client->list);
 
-	ret = kfifo_alloc(&client->fifo, MAX_KFIFO_SIZE, GFP_KERNEL);
+	ret = kfifo_alloc(&client->fifo, KFD_MAX_KFIFO_SIZE, GFP_KERNEL);
 	if (ret) {
 		kfree(client);
 		return ret;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.h b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.h
index 76fe4e0ec2d2..bb4d72b57387 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.h
@@ -24,29 +24,34 @@
 #ifndef KFD_SMI_EVENTS_H_INCLUDED
 #define KFD_SMI_EVENTS_H_INCLUDED
 
-int kfd_smi_event_open(struct kfd_dev *dev, uint32_t *fd);
-void kfd_smi_event_update_vmfault(struct kfd_dev *dev, uint16_t pasid);
-void kfd_smi_event_update_thermal_throttling(struct kfd_dev *dev,
+struct amdgpu_reset_context;
+
+int kfd_smi_event_open(struct kfd_node *dev, uint32_t *fd);
+void kfd_smi_event_update_vmfault(struct kfd_node *dev, uint16_t pasid);
+void kfd_smi_event_update_thermal_throttling(struct kfd_node *dev,
 					     uint64_t throttle_bitmask);
-void kfd_smi_event_update_gpu_reset(struct kfd_dev *dev, bool post_reset);
-void kfd_smi_event_page_fault_start(struct kfd_dev *dev, pid_t pid,
+void kfd_smi_event_update_gpu_reset(struct kfd_node *dev, bool post_reset,
+				    struct amdgpu_reset_context *reset_context);
+void kfd_smi_event_page_fault_start(struct kfd_node *node, pid_t pid,
 				    unsigned long address, bool write_fault,
 				    ktime_t ts);
-void kfd_smi_event_page_fault_end(struct kfd_dev *dev, pid_t pid,
+void kfd_smi_event_page_fault_end(struct kfd_node *node, pid_t pid,
 				  unsigned long address, bool migration);
-void kfd_smi_event_migration_start(struct kfd_dev *dev, pid_t pid,
+void kfd_smi_event_migration_start(struct kfd_node *node, pid_t pid,
 			     unsigned long start, unsigned long end,
 			     uint32_t from, uint32_t to,
 			     uint32_t prefetch_loc, uint32_t preferred_loc,
 			     uint32_t trigger);
-void kfd_smi_event_migration_end(struct kfd_dev *dev, pid_t pid,
+void kfd_smi_event_migration_end(struct kfd_node *node, pid_t pid,
 			     unsigned long start, unsigned long end,
-			     uint32_t from, uint32_t to, uint32_t trigger);
-void kfd_smi_event_queue_eviction(struct kfd_dev *dev, pid_t pid,
+			     uint32_t from, uint32_t to, uint32_t trigger,
+			     int error_code);
+void kfd_smi_event_queue_eviction(struct kfd_node *node, pid_t pid,
 				  uint32_t trigger);
-void kfd_smi_event_queue_restore(struct kfd_dev *dev, pid_t pid);
+void kfd_smi_event_queue_restore(struct kfd_node *node, pid_t pid);
 void kfd_smi_event_queue_restore_rescheduled(struct mm_struct *mm);
-void kfd_smi_event_unmap_from_gpu(struct kfd_dev *dev, pid_t pid,
+void kfd_smi_event_unmap_from_gpu(struct kfd_node *node, pid_t pid,
 				  unsigned long address, unsigned long last,
 				  uint32_t trigger);
+void kfd_smi_event_process(struct kfd_process_device *pdd, bool start);
 #endif
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
index 96a138a39515..97c2270f278f 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
@@ -23,7 +23,10 @@
 
 #include <linux/types.h>
 #include <linux/sched/task.h>
+#include <linux/dynamic_debug.h>
 #include <drm/ttm/ttm_tt.h>
+#include <drm/drm_exec.h>
+
 #include "amdgpu_sync.h"
 #include "amdgpu_object.h"
 #include "amdgpu_vm.h"
@@ -46,6 +49,13 @@
  * page table is updated.
  */
 #define AMDGPU_SVM_RANGE_RETRY_FAULT_PENDING	(2UL * NSEC_PER_MSEC)
+#if IS_ENABLED(CONFIG_DYNAMIC_DEBUG)
+#define dynamic_svm_range_dump(svms) \
+	_dynamic_func_call_no_desc("svm_range_dump", svm_range_debug_dump, svms)
+#else
+#define dynamic_svm_range_dump(svms) \
+	do { if (0) svm_range_debug_dump(svms); } while (0)
+#endif
 
 /* Giant svm range split into smaller ranges based on this, it is decided using
  * minimum of all dGPU/APU 1/32 VRAM size, between 2MB to 1GB and alignment to
@@ -170,12 +180,11 @@ svm_range_dma_map_dev(struct amdgpu_device *adev, struct svm_range *prange,
 
 		page = hmm_pfn_to_page(hmm_pfns[i]);
 		if (is_zone_device_page(page)) {
-			struct amdgpu_device *bo_adev =
-					amdgpu_ttm_adev(prange->svm_bo->bo->tbo.bdev);
+			struct amdgpu_device *bo_adev = prange->svm_bo->node->adev;
 
 			addr[i] = (hmm_pfns[i] << PAGE_SHIFT) +
 				   bo_adev->vm_manager.vram_base_offset -
-				   bo_adev->kfd.dev->pgmap.range.start;
+				   bo_adev->kfd.pgmap.range.start;
 			addr[i] |= SVM_RANGE_VRAM_DOMAIN;
 			pr_debug_ratelimited("vram address: 0x%llx\n", addr[i]);
 			continue;
@@ -189,6 +198,7 @@ svm_range_dma_map_dev(struct amdgpu_device *adev, struct svm_range *prange,
 		pr_debug_ratelimited("dma mapping 0x%llx for page addr 0x%lx\n",
 				     addr[i] >> PAGE_SHIFT, page_to_pfn(page));
 	}
+
 	return 0;
 }
 
@@ -222,7 +232,7 @@ svm_range_dma_map(struct svm_range *prange, unsigned long *bitmap,
 	return r;
 }
 
-void svm_range_dma_unmap(struct device *dev, dma_addr_t *dma_addr,
+void svm_range_dma_unmap_dev(struct device *dev, dma_addr_t *dma_addr,
 			 unsigned long offset, unsigned long npages)
 {
 	enum dma_data_direction dir = DMA_BIDIRECTIONAL;
@@ -240,7 +250,7 @@ void svm_range_dma_unmap(struct device *dev, dma_addr_t *dma_addr,
 	}
 }
 
-void svm_range_free_dma_mappings(struct svm_range *prange)
+void svm_range_dma_unmap(struct svm_range *prange)
 {
 	struct kfd_process_device *pdd;
 	dma_addr_t *dma_addr;
@@ -261,40 +271,51 @@ void svm_range_free_dma_mappings(struct svm_range *prange)
 			continue;
 		}
 		dev = &pdd->dev->adev->pdev->dev;
-		svm_range_dma_unmap(dev, dma_addr, 0, prange->npages);
-		kvfree(dma_addr);
-		prange->dma_addr[gpuidx] = NULL;
+
+		svm_range_dma_unmap_dev(dev, dma_addr, 0, prange->npages);
 	}
 }
 
-static void svm_range_free(struct svm_range *prange, bool update_mem_usage)
+static void svm_range_free(struct svm_range *prange, bool do_unmap)
 {
 	uint64_t size = (prange->last - prange->start + 1) << PAGE_SHIFT;
 	struct kfd_process *p = container_of(prange->svms, struct kfd_process, svms);
+	uint32_t gpuidx;
 
 	pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx]\n", prange->svms, prange,
 		 prange->start, prange->last);
 
 	svm_range_vram_node_free(prange);
-	svm_range_free_dma_mappings(prange);
+	if (do_unmap)
+		svm_range_dma_unmap(prange);
 
-	if (update_mem_usage && !p->xnack_enabled) {
+	if (do_unmap && !p->xnack_enabled) {
 		pr_debug("unreserve prange 0x%p size: 0x%llx\n", prange, size);
 		amdgpu_amdkfd_unreserve_mem_limit(NULL, size,
-					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR);
+					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0);
+	}
+
+	/* free dma_addr array for each gpu */
+	for (gpuidx = 0; gpuidx < MAX_GPU_INSTANCE; gpuidx++) {
+		if (prange->dma_addr[gpuidx]) {
+			kvfree(prange->dma_addr[gpuidx]);
+			prange->dma_addr[gpuidx] = NULL;
+		}
 	}
+
 	mutex_destroy(&prange->lock);
 	mutex_destroy(&prange->migrate_mutex);
 	kfree(prange);
 }
 
 static void
-svm_range_set_default_attributes(int32_t *location, int32_t *prefetch_loc,
-				 uint8_t *granularity, uint32_t *flags)
+svm_range_set_default_attributes(struct svm_range_list *svms, int32_t *location,
+				 int32_t *prefetch_loc, uint8_t *granularity,
+				 uint32_t *flags)
 {
 	*location = KFD_IOCTL_SVM_LOCATION_UNDEFINED;
 	*prefetch_loc = KFD_IOCTL_SVM_LOCATION_UNDEFINED;
-	*granularity = 9;
+	*granularity = svms->default_granularity;
 	*flags =
 		KFD_IOCTL_SVM_FLAG_HOST_ACCESS | KFD_IOCTL_SVM_FLAG_COHERENT;
 }
@@ -314,7 +335,7 @@ svm_range *svm_range_new(struct svm_range_list *svms, uint64_t start,
 	p = container_of(svms, struct kfd_process, svms);
 	if (!p->xnack_enabled && update_mem_usage &&
 	    amdgpu_amdkfd_reserve_mem_limit(NULL, size << PAGE_SHIFT,
-					    KFD_IOC_ALLOC_MEM_FLAGS_USERPTR)) {
+				    KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0)) {
 		pr_info("SVM mapping failed, exceeds resident system memory limit\n");
 		kfree(prange);
 		return NULL;
@@ -330,6 +351,7 @@ svm_range *svm_range_new(struct svm_range_list *svms, uint64_t start,
 	INIT_LIST_HEAD(&prange->child_list);
 	atomic_set(&prange->invalid, 0);
 	prange->validate_timestamp = 0;
+	prange->vram_pages = 0;
 	mutex_init(&prange->migrate_mutex);
 	mutex_init(&prange->lock);
 
@@ -337,7 +359,7 @@ svm_range *svm_range_new(struct svm_range_list *svms, uint64_t start,
 		bitmap_copy(prange->bitmap_access, svms->bitmap_supported,
 			    MAX_GPU_INSTANCE);
 
-	svm_range_set_default_attributes(&prange->preferred_loc,
+	svm_range_set_default_attributes(svms, &prange->preferred_loc,
 					 &prange->prefetch_loc,
 					 &prange->granularity, &prange->flags);
 
@@ -376,19 +398,37 @@ static void svm_range_bo_release(struct kref *kref)
 			 prange->start, prange->last);
 		mutex_lock(&prange->lock);
 		prange->svm_bo = NULL;
+		/* prange should not hold vram page now */
+		WARN_ONCE(prange->actual_loc, "prange should not hold vram page");
 		mutex_unlock(&prange->lock);
 
 		spin_lock(&svm_bo->list_lock);
 	}
 	spin_unlock(&svm_bo->list_lock);
-	if (!dma_fence_is_signaled(&svm_bo->eviction_fence->base)) {
-		/* We're not in the eviction worker.
-		 * Signal the fence and synchronize with any
-		 * pending eviction work.
+
+	if (mmget_not_zero(svm_bo->eviction_fence->mm)) {
+		struct kfd_process_device *pdd;
+		struct kfd_process *p;
+		struct mm_struct *mm;
+
+		mm = svm_bo->eviction_fence->mm;
+		/*
+		 * The forked child process takes svm_bo device pages ref, svm_bo could be
+		 * released after parent process is gone.
 		 */
-		dma_fence_signal(&svm_bo->eviction_fence->base);
-		cancel_work_sync(&svm_bo->eviction_work);
+		p = kfd_lookup_process_by_mm(mm);
+		if (p) {
+			pdd = kfd_get_process_device_data(svm_bo->node, p);
+			if (pdd)
+				atomic64_sub(amdgpu_bo_size(svm_bo->bo), &pdd->vram_usage);
+			kfd_unref_process(p);
+		}
+		mmput(mm);
 	}
+
+	if (!dma_fence_is_signaled(&svm_bo->eviction_fence->base))
+		/* We're not in the eviction worker. Signal the fence. */
+		dma_fence_signal(&svm_bo->eviction_fence->base);
 	dma_fence_put(&svm_bo->eviction_fence->base);
 	amdgpu_bo_unref(&svm_bo->bo);
 	kfree(svm_bo);
@@ -424,10 +464,8 @@ static void svm_range_bo_unref(struct svm_range_bo *svm_bo)
 }
 
 static bool
-svm_range_validate_svm_bo(struct amdgpu_device *adev, struct svm_range *prange)
+svm_range_validate_svm_bo(struct kfd_node *node, struct svm_range *prange)
 {
-	struct amdgpu_device *bo_adev;
-
 	mutex_lock(&prange->lock);
 	if (!prange->svm_bo) {
 		mutex_unlock(&prange->lock);
@@ -440,12 +478,11 @@ svm_range_validate_svm_bo(struct amdgpu_device *adev, struct svm_range *prange)
 	}
 	if (svm_bo_ref_unless_zero(prange->svm_bo)) {
 		/*
-		 * Migrate from GPU to GPU, remove range from source bo_adev
-		 * svm_bo range list, and return false to allocate svm_bo from
-		 * destination adev.
+		 * Migrate from GPU to GPU, remove range from source svm_bo->node
+		 * range list, and return false to allocate svm_bo from destination
+		 * node.
 		 */
-		bo_adev = amdgpu_ttm_adev(prange->svm_bo->bo->tbo.bdev);
-		if (bo_adev != adev) {
+		if (prange->svm_bo->node != node) {
 			mutex_unlock(&prange->lock);
 
 			spin_lock(&prange->svm_bo->list_lock);
@@ -488,11 +525,11 @@ svm_range_validate_svm_bo(struct amdgpu_device *adev, struct svm_range *prange)
 
 	/* We need a new svm_bo. Spin-loop to wait for concurrent
 	 * svm_range_bo_release to finish removing this range from
-	 * its range list. After this, it is safe to reuse the
-	 * svm_bo pointer and svm_bo_list head.
+	 * its range list and set prange->svm_bo to null. After this,
+	 * it is safe to reuse the svm_bo pointer and svm_bo_list head.
 	 */
-	while (!list_empty_careful(&prange->svm_bo_list))
-		;
+	while (!list_empty_careful(&prange->svm_bo_list) || prange->svm_bo)
+		cond_resched();
 
 	return false;
 }
@@ -513,9 +550,10 @@ static struct svm_range_bo *svm_range_bo_new(void)
 }
 
 int
-svm_range_vram_node_new(struct amdgpu_device *adev, struct svm_range *prange,
+svm_range_vram_node_new(struct kfd_node *node, struct svm_range *prange,
 			bool clear)
 {
+	struct kfd_process_device *pdd;
 	struct amdgpu_bo_param bp;
 	struct svm_range_bo *svm_bo;
 	struct amdgpu_bo_user *ubo;
@@ -525,10 +563,11 @@ svm_range_vram_node_new(struct amdgpu_device *adev, struct svm_range *prange,
 	int r;
 
 	p = container_of(prange->svms, struct kfd_process, svms);
-	pr_debug("pasid: %x svms 0x%p [0x%lx 0x%lx]\n", p->pasid, prange->svms,
+	pr_debug("process pid: %d svms 0x%p [0x%lx 0x%lx]\n",
+		 p->lead_thread->pid, prange->svms,
 		 prange->start, prange->last);
 
-	if (svm_range_validate_svm_bo(adev, prange))
+	if (svm_range_validate_svm_bo(node, prange))
 		return 0;
 
 	svm_bo = svm_range_bo_new();
@@ -542,6 +581,7 @@ svm_range_vram_node_new(struct amdgpu_device *adev, struct svm_range *prange,
 		kfree(svm_bo);
 		return -ESRCH;
 	}
+	svm_bo->node = node;
 	svm_bo->eviction_fence =
 		amdgpu_amdkfd_fence_create(dma_fence_context_alloc(1),
 					   mm,
@@ -558,13 +598,20 @@ svm_range_vram_node_new(struct amdgpu_device *adev, struct svm_range *prange,
 	bp.flags |= AMDGPU_GEM_CREATE_DISCARDABLE;
 	bp.type = ttm_bo_type_device;
 	bp.resv = NULL;
+	if (node->xcp)
+		bp.xcp_id_plus1 = node->xcp->id + 1;
 
-	r = amdgpu_bo_create_user(adev, &bp, &ubo);
+	r = amdgpu_bo_create_user(node->adev, &bp, &ubo);
 	if (r) {
 		pr_debug("failed %d to create bo\n", r);
 		goto create_bo_failed;
 	}
 	bo = &ubo->bo;
+
+	pr_debug("alloc bo at offset 0x%lx size 0x%lx on partition %d\n",
+		 bo->tbo.resource->start << PAGE_SHIFT, bp.size,
+		 bp.xcp_id_plus1 - 1);
+
 	r = amdgpu_bo_reserve(bo, true);
 	if (r) {
 		pr_debug("failed %d to reserve bo\n", r);
@@ -599,6 +646,10 @@ svm_range_vram_node_new(struct amdgpu_device *adev, struct svm_range *prange,
 	list_add(&prange->svm_bo_list, &svm_bo->range_list);
 	spin_unlock(&svm_bo->list_lock);
 
+	pdd = svm_range_get_pdd_by_node(prange, node);
+	if (pdd)
+		atomic64_add(amdgpu_bo_size(bo), &pdd->vram_usage);
+
 	return 0;
 
 reserve_bo_failed:
@@ -613,49 +664,41 @@ create_bo_failed:
 
 void svm_range_vram_node_free(struct svm_range *prange)
 {
-	svm_range_bo_unref(prange->svm_bo);
-	prange->ttm_res = NULL;
+	/* serialize prange->svm_bo unref */
+	mutex_lock(&prange->lock);
+	/* prange->svm_bo has not been unref */
+	if (prange->ttm_res) {
+		prange->ttm_res = NULL;
+		mutex_unlock(&prange->lock);
+		svm_range_bo_unref(prange->svm_bo);
+	} else
+		mutex_unlock(&prange->lock);
 }
 
-struct amdgpu_device *
-svm_range_get_adev_by_id(struct svm_range *prange, uint32_t gpu_id)
+struct kfd_node *
+svm_range_get_node_by_id(struct svm_range *prange, uint32_t gpu_id)
 {
-	struct kfd_process_device *pdd;
 	struct kfd_process *p;
-	int32_t gpu_idx;
+	struct kfd_process_device *pdd;
 
 	p = container_of(prange->svms, struct kfd_process, svms);
-
-	gpu_idx = kfd_process_gpuidx_from_gpuid(p, gpu_id);
-	if (gpu_idx < 0) {
-		pr_debug("failed to get device by id 0x%x\n", gpu_id);
-		return NULL;
-	}
-	pdd = kfd_process_device_from_gpuidx(p, gpu_idx);
+	pdd = kfd_process_device_data_by_id(p, gpu_id);
 	if (!pdd) {
-		pr_debug("failed to get device by idx 0x%x\n", gpu_idx);
+		pr_debug("failed to get kfd process device by id 0x%x\n", gpu_id);
 		return NULL;
 	}
 
-	return pdd->dev->adev;
+	return pdd->dev;
 }
 
 struct kfd_process_device *
-svm_range_get_pdd_by_adev(struct svm_range *prange, struct amdgpu_device *adev)
+svm_range_get_pdd_by_node(struct svm_range *prange, struct kfd_node *node)
 {
 	struct kfd_process *p;
-	int32_t gpu_idx, gpuid;
-	int r;
 
 	p = container_of(prange->svms, struct kfd_process, svms);
 
-	r = kfd_process_gpuid_from_adev(p, adev, &gpuid, &gpu_idx);
-	if (r) {
-		pr_debug("failed to get device id by adev %p\n", adev);
-		return NULL;
-	}
-
-	return kfd_process_device_from_gpuidx(p, gpu_idx);
+	return kfd_get_process_device_data(node, p);
 }
 
 static int svm_range_bo_validate(void *param, struct amdgpu_bo *bo)
@@ -735,7 +778,9 @@ svm_range_apply_attrs(struct kfd_process *p, struct svm_range *prange,
 		case KFD_IOCTL_SVM_ATTR_ACCESS:
 		case KFD_IOCTL_SVM_ATTR_ACCESS_IN_PLACE:
 		case KFD_IOCTL_SVM_ATTR_NO_ACCESS:
-			*update_mapping = true;
+			if (!p->xnack_enabled)
+				*update_mapping = true;
+
 			gpuidx = kfd_process_gpuidx_from_gpuid(p,
 							       attrs[i].value);
 			if (attrs[i].type == KFD_IOCTL_SVM_ATTR_NO_ACCESS) {
@@ -758,7 +803,7 @@ svm_range_apply_attrs(struct kfd_process *p, struct svm_range *prange,
 			prange->flags &= ~attrs[i].value;
 			break;
 		case KFD_IOCTL_SVM_ATTR_GRANULARITY:
-			prange->granularity = attrs[i].value;
+			prange->granularity = min_t(uint32_t, attrs[i].value, 0x3F);
 			break;
 		default:
 			WARN_ONCE(1, "svm_range_check_attrs wasn't called?");
@@ -858,10 +903,56 @@ static void svm_range_debug_dump(struct svm_range_list *svms)
 	}
 }
 
+static void *
+svm_range_copy_array(void *psrc, size_t size, uint64_t num_elements,
+		     uint64_t offset, uint64_t *vram_pages)
+{
+	unsigned char *src = (unsigned char *)psrc + offset;
+	unsigned char *dst;
+	uint64_t i;
+
+	dst = kvmalloc_array(num_elements, size, GFP_KERNEL);
+	if (!dst)
+		return NULL;
+
+	if (!vram_pages) {
+		memcpy(dst, src, num_elements * size);
+		return (void *)dst;
+	}
+
+	*vram_pages = 0;
+	for (i = 0; i < num_elements; i++) {
+		dma_addr_t *temp;
+		temp = (dma_addr_t *)dst + i;
+		*temp = *((dma_addr_t *)src + i);
+		if (*temp&SVM_RANGE_VRAM_DOMAIN)
+			(*vram_pages)++;
+	}
+
+	return (void *)dst;
+}
+
+static int
+svm_range_copy_dma_addrs(struct svm_range *dst, struct svm_range *src)
+{
+	int i;
+
+	for (i = 0; i < MAX_GPU_INSTANCE; i++) {
+		if (!src->dma_addr[i])
+			continue;
+		dst->dma_addr[i] = svm_range_copy_array(src->dma_addr[i],
+					sizeof(*src->dma_addr[i]), src->npages, 0, NULL);
+		if (!dst->dma_addr[i])
+			return -ENOMEM;
+	}
+
+	return 0;
+}
+
 static int
 svm_range_split_array(void *ppnew, void *ppold, size_t size,
 		      uint64_t old_start, uint64_t old_n,
-		      uint64_t new_start, uint64_t new_n)
+		      uint64_t new_start, uint64_t new_n, uint64_t *new_vram_pages)
 {
 	unsigned char *new, *old, *pold;
 	uint64_t d;
@@ -872,22 +963,17 @@ svm_range_split_array(void *ppnew, void *ppold, size_t size,
 	if (!pold)
 		return 0;
 
-	new = kvmalloc_array(new_n, size, GFP_KERNEL);
+	d = (new_start - old_start) * size;
+	/* get dma addr array for new range and calculte its vram page number */
+	new = svm_range_copy_array(pold, size, new_n, d, new_vram_pages);
 	if (!new)
 		return -ENOMEM;
-
-	d = (new_start - old_start) * size;
-	memcpy(new, pold + d, new_n * size);
-
-	old = kvmalloc_array(old_n, size, GFP_KERNEL);
+	d = (new_start == old_start) ? new_n * size : 0;
+	old = svm_range_copy_array(pold, size, old_n, d, NULL);
 	if (!old) {
 		kvfree(new);
 		return -ENOMEM;
 	}
-
-	d = (new_start == old_start) ? new_n * size : 0;
-	memcpy(old, pold + d, old_n * size);
-
 	kvfree(pold);
 	*(void **)ppold = old;
 	*(void **)ppnew = new;
@@ -905,10 +991,13 @@ svm_range_split_pages(struct svm_range *new, struct svm_range *old,
 	for (i = 0; i < MAX_GPU_INSTANCE; i++) {
 		r = svm_range_split_array(&new->dma_addr[i], &old->dma_addr[i],
 					  sizeof(*old->dma_addr[i]), old->start,
-					  npages, new->start, new->npages);
+					  npages, new->start, new->npages,
+					  old->actual_loc ? &new->vram_pages : NULL);
 		if (r)
 			return r;
 	}
+	if (old->actual_loc)
+		old->vram_pages -= new->vram_pages;
 
 	return 0;
 }
@@ -990,6 +1079,7 @@ svm_range_split_adjust(struct svm_range *new, struct svm_range *old,
 	new->mapped_to_gpu = old->mapped_to_gpu;
 	bitmap_copy(new->bitmap_access, old->bitmap_access, MAX_GPU_INSTANCE);
 	bitmap_copy(new->bitmap_aip, old->bitmap_aip, MAX_GPU_INSTANCE);
+	atomic_set(&new->queue_refcount, atomic_read(&old->queue_refcount));
 
 	return 0;
 }
@@ -1051,126 +1141,80 @@ svm_range_split(struct svm_range *prange, uint64_t start, uint64_t last,
 }
 
 static int
-svm_range_split_tail(struct svm_range *prange,
-		     uint64_t new_last, struct list_head *insert_list)
+svm_range_split_tail(struct svm_range *prange, uint64_t new_last,
+		     struct list_head *insert_list, struct list_head *remap_list)
 {
-	struct svm_range *tail;
+	struct svm_range *tail = NULL;
 	int r = svm_range_split(prange, prange->start, new_last, &tail);
 
-	if (!r)
+	if (!r) {
 		list_add(&tail->list, insert_list);
+		if (!IS_ALIGNED(new_last + 1, 1UL << prange->granularity))
+			list_add(&tail->update_list, remap_list);
+	}
 	return r;
 }
 
 static int
-svm_range_split_head(struct svm_range *prange,
-		     uint64_t new_start, struct list_head *insert_list)
+svm_range_split_head(struct svm_range *prange, uint64_t new_start,
+		     struct list_head *insert_list, struct list_head *remap_list)
 {
-	struct svm_range *head;
+	struct svm_range *head = NULL;
 	int r = svm_range_split(prange, new_start, prange->last, &head);
 
-	if (!r)
+	if (!r) {
 		list_add(&head->list, insert_list);
+		if (!IS_ALIGNED(new_start, 1UL << prange->granularity))
+			list_add(&head->update_list, remap_list);
+	}
 	return r;
 }
 
 static void
-svm_range_add_child(struct svm_range *prange, struct mm_struct *mm,
-		    struct svm_range *pchild, enum svm_work_list_ops op)
+svm_range_add_child(struct svm_range *prange, struct svm_range *pchild, enum svm_work_list_ops op)
 {
 	pr_debug("add child 0x%p [0x%lx 0x%lx] to prange 0x%p child list %d\n",
 		 pchild, pchild->start, pchild->last, prange, op);
 
-	pchild->work_item.mm = mm;
+	pchild->work_item.mm = NULL;
 	pchild->work_item.op = op;
 	list_add_tail(&pchild->child_list, &prange->child_list);
 }
 
-/**
- * svm_range_split_by_granularity - collect ranges within granularity boundary
- *
- * @p: the process with svms list
- * @mm: mm structure
- * @addr: the vm fault address in pages, to split the prange
- * @parent: parent range if prange is from child list
- * @prange: prange to split
- *
- * Trims @prange to be a single aligned block of prange->granularity if
- * possible. The head and tail are added to the child_list in @parent.
- *
- * Context: caller must hold mmap_read_lock and prange->lock
- *
- * Return:
- * 0 - OK, otherwise error code
- */
-int
-svm_range_split_by_granularity(struct kfd_process *p, struct mm_struct *mm,
-			       unsigned long addr, struct svm_range *parent,
-			       struct svm_range *prange)
+static bool
+svm_nodes_in_same_hive(struct kfd_node *node_a, struct kfd_node *node_b)
 {
-	struct svm_range *head, *tail;
-	unsigned long start, last, size;
-	int r;
-
-	/* Align splited range start and size to granularity size, then a single
-	 * PTE will be used for whole range, this reduces the number of PTE
-	 * updated and the L1 TLB space used for translation.
-	 */
-	size = 1UL << prange->granularity;
-	start = ALIGN_DOWN(addr, size);
-	last = ALIGN(addr + 1, size) - 1;
-
-	pr_debug("svms 0x%p split [0x%lx 0x%lx] to [0x%lx 0x%lx] size 0x%lx\n",
-		 prange->svms, prange->start, prange->last, start, last, size);
-
-	if (start > prange->start) {
-		r = svm_range_split(prange, start, prange->last, &head);
-		if (r)
-			return r;
-		svm_range_add_child(parent, mm, head, SVM_OP_ADD_RANGE);
-	}
-
-	if (last < prange->last) {
-		r = svm_range_split(prange, prange->start, last, &tail);
-		if (r)
-			return r;
-		svm_range_add_child(parent, mm, tail, SVM_OP_ADD_RANGE);
-	}
-
-	/* xnack on, update mapping on GPUs with ACCESS_IN_PLACE */
-	if (p->xnack_enabled && prange->work_item.op == SVM_OP_ADD_RANGE) {
-		prange->work_item.op = SVM_OP_ADD_RANGE_AND_MAP;
-		pr_debug("change prange 0x%p [0x%lx 0x%lx] op %d\n",
-			 prange, prange->start, prange->last,
-			 SVM_OP_ADD_RANGE_AND_MAP);
-	}
-	return 0;
+	return (node_a->adev == node_b->adev ||
+		amdgpu_xgmi_same_hive(node_a->adev, node_b->adev));
 }
 
 static uint64_t
-svm_range_get_pte_flags(struct amdgpu_device *adev, struct svm_range *prange,
-			int domain)
+svm_range_get_pte_flags(struct kfd_node *node, struct amdgpu_vm *vm,
+			struct svm_range *prange, int domain)
 {
-	struct amdgpu_device *bo_adev;
+	struct kfd_node *bo_node;
 	uint32_t flags = prange->flags;
 	uint32_t mapping_flags = 0;
+	uint32_t gc_ip_version = KFD_GC_VERSION(node);
 	uint64_t pte_flags;
 	bool snoop = (domain != SVM_RANGE_VRAM_DOMAIN);
-	bool coherent = flags & KFD_IOCTL_SVM_FLAG_COHERENT;
+	bool coherent = flags & (KFD_IOCTL_SVM_FLAG_COHERENT | KFD_IOCTL_SVM_FLAG_EXT_COHERENT);
+	bool ext_coherent = flags & KFD_IOCTL_SVM_FLAG_EXT_COHERENT;
+	unsigned int mtype_local;
 
 	if (domain == SVM_RANGE_VRAM_DOMAIN)
-		bo_adev = amdgpu_ttm_adev(prange->svm_bo->bo->tbo.bdev);
+		bo_node = prange->svm_bo->node;
 
-	switch (KFD_GC_VERSION(adev->kfd.dev)) {
+	switch (gc_ip_version) {
 	case IP_VERSION(9, 4, 1):
 		if (domain == SVM_RANGE_VRAM_DOMAIN) {
-			if (bo_adev == adev) {
+			if (bo_node == node) {
 				mapping_flags |= coherent ?
 					AMDGPU_VM_MTYPE_CC : AMDGPU_VM_MTYPE_RW;
 			} else {
 				mapping_flags |= coherent ?
 					AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC;
-				if (amdgpu_xgmi_same_hive(adev, bo_adev))
+				if (svm_nodes_in_same_hive(node, bo_node))
 					snoop = true;
 			}
 		} else {
@@ -1180,15 +1224,15 @@ svm_range_get_pte_flags(struct amdgpu_device *adev, struct svm_range *prange,
 		break;
 	case IP_VERSION(9, 4, 2):
 		if (domain == SVM_RANGE_VRAM_DOMAIN) {
-			if (bo_adev == adev) {
+			if (bo_node == node) {
 				mapping_flags |= coherent ?
 					AMDGPU_VM_MTYPE_CC : AMDGPU_VM_MTYPE_RW;
-				if (adev->gmc.xgmi.connected_to_cpu)
+				if (node->adev->gmc.xgmi.connected_to_cpu)
 					snoop = true;
 			} else {
 				mapping_flags |= coherent ?
 					AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC;
-				if (amdgpu_xgmi_same_hive(adev, bo_adev))
+				if (svm_nodes_in_same_hive(node, bo_node))
 					snoop = true;
 			}
 		} else {
@@ -1196,23 +1240,71 @@ svm_range_get_pte_flags(struct amdgpu_device *adev, struct svm_range *prange,
 				AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC;
 		}
 		break;
+	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+	case IP_VERSION(9, 5, 0):
+		if (ext_coherent)
+			mtype_local = AMDGPU_VM_MTYPE_CC;
+		else
+			mtype_local = amdgpu_mtype_local == 1 ? AMDGPU_VM_MTYPE_NC :
+				amdgpu_mtype_local == 2 ? AMDGPU_VM_MTYPE_CC : AMDGPU_VM_MTYPE_RW;
+		snoop = true;
+		if (domain == SVM_RANGE_VRAM_DOMAIN) {
+			/* local HBM region close to partition */
+			if (bo_node->adev == node->adev &&
+			    (!bo_node->xcp || !node->xcp || bo_node->xcp->mem_id == node->xcp->mem_id))
+				mapping_flags |= mtype_local;
+			/* local HBM region far from partition or remote XGMI GPU
+			 * with regular system scope coherence
+			 */
+			else if (svm_nodes_in_same_hive(bo_node, node) && !ext_coherent)
+				mapping_flags |= AMDGPU_VM_MTYPE_NC;
+			/* PCIe P2P on GPUs pre-9.5.0 */
+			else if (gc_ip_version < IP_VERSION(9, 5, 0) &&
+				 !svm_nodes_in_same_hive(bo_node, node))
+				mapping_flags |= AMDGPU_VM_MTYPE_UC;
+			/* Other remote memory */
+			else
+				mapping_flags |= ext_coherent ? AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC;
+		/* system memory accessed by the APU */
+		} else if (node->adev->flags & AMD_IS_APU) {
+			/* On NUMA systems, locality is determined per-page
+			 * in amdgpu_gmc_override_vm_pte_flags
+			 */
+			if (num_possible_nodes() <= 1)
+				mapping_flags |= mtype_local;
+			else
+				mapping_flags |= ext_coherent ? AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC;
+		/* system memory accessed by the dGPU */
+		} else {
+			if (gc_ip_version < IP_VERSION(9, 5, 0) || ext_coherent)
+				mapping_flags |= AMDGPU_VM_MTYPE_UC;
+			else
+				mapping_flags |= AMDGPU_VM_MTYPE_NC;
+		}
+		break;
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		mapping_flags |= AMDGPU_VM_MTYPE_NC;
+		break;
 	default:
 		mapping_flags |= coherent ?
 			AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC;
 	}
 
-	mapping_flags |= AMDGPU_VM_PAGE_READABLE | AMDGPU_VM_PAGE_WRITEABLE;
-
-	if (flags & KFD_IOCTL_SVM_FLAG_GPU_RO)
-		mapping_flags &= ~AMDGPU_VM_PAGE_WRITEABLE;
 	if (flags & KFD_IOCTL_SVM_FLAG_GPU_EXEC)
 		mapping_flags |= AMDGPU_VM_PAGE_EXECUTABLE;
 
 	pte_flags = AMDGPU_PTE_VALID;
 	pte_flags |= (domain == SVM_RANGE_VRAM_DOMAIN) ? 0 : AMDGPU_PTE_SYSTEM;
 	pte_flags |= snoop ? AMDGPU_PTE_SNOOPED : 0;
+	if (gc_ip_version >= IP_VERSION(12, 0, 0))
+		pte_flags |= AMDGPU_PTE_IS_PTE;
 
-	pte_flags |= amdgpu_gem_va_map_flags(adev, mapping_flags);
+	amdgpu_gmc_get_vm_pte(node->adev, vm, NULL, mapping_flags, &pte_flags);
+	pte_flags |= AMDGPU_PTE_READABLE;
+	if (!(flags & KFD_IOCTL_SVM_FLAG_GPU_RO))
+		pte_flags |= AMDGPU_PTE_WRITEABLE;
 	return pte_flags;
 }
 
@@ -1225,7 +1317,7 @@ svm_range_unmap_from_gpu(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 
 	pr_debug("[0x%llx 0x%llx]\n", start, last);
 
-	return amdgpu_vm_update_range(adev, vm, false, true, true, NULL, start,
+	return amdgpu_vm_update_range(adev, vm, false, true, true, false, NULL, start,
 				      last, init_pte_value, 0, 0, NULL, NULL,
 				      fence);
 }
@@ -1319,7 +1411,7 @@ svm_range_map_to_gpu(struct kfd_process_device *pdd, struct svm_range *prange,
 		pr_debug("Mapping range [0x%lx 0x%llx] on domain: %s\n",
 			 last_start, prange->start + i, last_domain ? "GPU" : "CPU");
 
-		pte_flags = svm_range_get_pte_flags(adev, prange, last_domain);
+		pte_flags = svm_range_get_pte_flags(pdd->dev, vm, prange, last_domain);
 		if (readonly)
 			pte_flags &= ~AMDGPU_PTE_WRITEABLE;
 
@@ -1328,8 +1420,12 @@ svm_range_map_to_gpu(struct kfd_process_device *pdd, struct svm_range *prange,
 			 (last_domain == SVM_RANGE_VRAM_DOMAIN) ? 1 : 0,
 			 pte_flags);
 
-		r = amdgpu_vm_update_range(adev, vm, false, false, flush_tlb, NULL,
-					   last_start, prange->start + i,
+		/* For dGPU mode, we use same vm_manager to allocate VRAM for
+		 * different memory partition based on fpfn/lpfn, we should use
+		 * same vm_manager.vram_base_offset regardless memory partition.
+		 */
+		r = amdgpu_vm_update_range(adev, vm, false, false, flush_tlb, true,
+					   NULL, last_start, prange->start + i,
 					   pte_flags,
 					   (last_start - prange->start) << PAGE_SHIFT,
 					   bo_adev ? bo_adev->vm_manager.vram_base_offset : 0,
@@ -1365,16 +1461,14 @@ svm_range_map_to_gpus(struct svm_range *prange, unsigned long offset,
 		      unsigned long *bitmap, bool wait, bool flush_tlb)
 {
 	struct kfd_process_device *pdd;
-	struct amdgpu_device *bo_adev;
+	struct amdgpu_device *bo_adev = NULL;
 	struct kfd_process *p;
 	struct dma_fence *fence = NULL;
 	uint32_t gpuidx;
 	int r = 0;
 
 	if (prange->svm_bo && prange->ttm_res)
-		bo_adev = amdgpu_ttm_adev(prange->svm_bo->bo->tbo.bdev);
-	else
-		bo_adev = NULL;
+		bo_adev = prange->svm_bo->node->adev;
 
 	p = container_of(prange->svms, struct kfd_process, svms);
 	for_each_set_bit(gpuidx, bitmap, MAX_GPU_INSTANCE) {
@@ -1423,37 +1517,34 @@ struct svm_validate_context {
 	struct svm_range *prange;
 	bool intr;
 	DECLARE_BITMAP(bitmap, MAX_GPU_INSTANCE);
-	struct ttm_validate_buffer tv[MAX_GPU_INSTANCE];
-	struct list_head validate_list;
-	struct ww_acquire_ctx ticket;
+	struct drm_exec exec;
 };
 
-static int svm_range_reserve_bos(struct svm_validate_context *ctx)
+static int svm_range_reserve_bos(struct svm_validate_context *ctx, bool intr)
 {
 	struct kfd_process_device *pdd;
 	struct amdgpu_vm *vm;
 	uint32_t gpuidx;
 	int r;
 
-	INIT_LIST_HEAD(&ctx->validate_list);
-	for_each_set_bit(gpuidx, ctx->bitmap, MAX_GPU_INSTANCE) {
-		pdd = kfd_process_device_from_gpuidx(ctx->process, gpuidx);
-		if (!pdd) {
-			pr_debug("failed to find device idx %d\n", gpuidx);
-			return -EINVAL;
-		}
-		vm = drm_priv_to_vm(pdd->drm_priv);
-
-		ctx->tv[gpuidx].bo = &vm->root.bo->tbo;
-		ctx->tv[gpuidx].num_shared = 4;
-		list_add(&ctx->tv[gpuidx].head, &ctx->validate_list);
-	}
+	drm_exec_init(&ctx->exec, intr ? DRM_EXEC_INTERRUPTIBLE_WAIT: 0, 0);
+	drm_exec_until_all_locked(&ctx->exec) {
+		for_each_set_bit(gpuidx, ctx->bitmap, MAX_GPU_INSTANCE) {
+			pdd = kfd_process_device_from_gpuidx(ctx->process, gpuidx);
+			if (!pdd) {
+				pr_debug("failed to find device idx %d\n", gpuidx);
+				r = -EINVAL;
+				goto unreserve_out;
+			}
+			vm = drm_priv_to_vm(pdd->drm_priv);
 
-	r = ttm_eu_reserve_buffers(&ctx->ticket, &ctx->validate_list,
-				   ctx->intr, NULL);
-	if (r) {
-		pr_debug("failed %d to reserve bo\n", r);
-		return r;
+			r = amdgpu_vm_lock_pd(vm, &ctx->exec, 2);
+			drm_exec_retry_on_contention(&ctx->exec);
+			if (unlikely(r)) {
+				pr_debug("failed %d to reserve bo\n", r);
+				goto unreserve_out;
+			}
+		}
 	}
 
 	for_each_set_bit(gpuidx, ctx->bitmap, MAX_GPU_INSTANCE) {
@@ -1464,9 +1555,9 @@ static int svm_range_reserve_bos(struct svm_validate_context *ctx)
 			goto unreserve_out;
 		}
 
-		r = amdgpu_vm_validate_pt_bos(pdd->dev->adev,
-					      drm_priv_to_vm(pdd->drm_priv),
-					      svm_range_bo_validate, NULL);
+		r = amdgpu_vm_validate(pdd->dev->adev,
+				       drm_priv_to_vm(pdd->drm_priv), NULL,
+				       svm_range_bo_validate, NULL);
 		if (r) {
 			pr_debug("failed %d validate pt bos\n", r);
 			goto unreserve_out;
@@ -1476,13 +1567,13 @@ static int svm_range_reserve_bos(struct svm_validate_context *ctx)
 	return 0;
 
 unreserve_out:
-	ttm_eu_backoff_reservation(&ctx->ticket, &ctx->validate_list);
+	drm_exec_fini(&ctx->exec);
 	return r;
 }
 
 static void svm_range_unreserve_bos(struct svm_validate_context *ctx)
 {
-	ttm_eu_backoff_reservation(&ctx->ticket, &ctx->validate_list);
+	drm_exec_fini(&ctx->exec);
 }
 
 static void *kfd_svm_page_owner(struct kfd_process *p, int32_t gpuidx)
@@ -1490,6 +1581,8 @@ static void *kfd_svm_page_owner(struct kfd_process *p, int32_t gpuidx)
 	struct kfd_process_device *pdd;
 
 	pdd = kfd_process_device_from_gpuidx(p, gpuidx);
+	if (!pdd)
+		return NULL;
 
 	return SVM_ADEV_PGMAP_OWNER(pdd->dev->adev);
 }
@@ -1519,51 +1612,64 @@ static void *kfd_svm_page_owner(struct kfd_process *p, int32_t gpuidx)
  * 5. Release page table (and SVM BO) reservation
  */
 static int svm_range_validate_and_map(struct mm_struct *mm,
+				      unsigned long map_start, unsigned long map_last,
 				      struct svm_range *prange, int32_t gpuidx,
 				      bool intr, bool wait, bool flush_tlb)
 {
-	struct svm_validate_context ctx;
+	struct svm_validate_context *ctx;
 	unsigned long start, end, addr;
 	struct kfd_process *p;
 	void *owner;
 	int32_t idx;
 	int r = 0;
 
-	ctx.process = container_of(prange->svms, struct kfd_process, svms);
-	ctx.prange = prange;
-	ctx.intr = intr;
+	ctx = kzalloc(sizeof(struct svm_validate_context), GFP_KERNEL);
+	if (!ctx)
+		return -ENOMEM;
+	ctx->process = container_of(prange->svms, struct kfd_process, svms);
+	ctx->prange = prange;
+	ctx->intr = intr;
 
 	if (gpuidx < MAX_GPU_INSTANCE) {
-		bitmap_zero(ctx.bitmap, MAX_GPU_INSTANCE);
-		bitmap_set(ctx.bitmap, gpuidx, 1);
-	} else if (ctx.process->xnack_enabled) {
-		bitmap_copy(ctx.bitmap, prange->bitmap_aip, MAX_GPU_INSTANCE);
+		bitmap_zero(ctx->bitmap, MAX_GPU_INSTANCE);
+		bitmap_set(ctx->bitmap, gpuidx, 1);
+	} else if (ctx->process->xnack_enabled) {
+		bitmap_copy(ctx->bitmap, prange->bitmap_aip, MAX_GPU_INSTANCE);
 
 		/* If prefetch range to GPU, or GPU retry fault migrate range to
 		 * GPU, which has ACCESS attribute to the range, create mapping
 		 * on that GPU.
 		 */
 		if (prange->actual_loc) {
-			gpuidx = kfd_process_gpuidx_from_gpuid(ctx.process,
+			gpuidx = kfd_process_gpuidx_from_gpuid(ctx->process,
 							prange->actual_loc);
 			if (gpuidx < 0) {
 				WARN_ONCE(1, "failed get device by id 0x%x\n",
 					 prange->actual_loc);
-				return -EINVAL;
+				r = -EINVAL;
+				goto free_ctx;
 			}
 			if (test_bit(gpuidx, prange->bitmap_access))
-				bitmap_set(ctx.bitmap, gpuidx, 1);
+				bitmap_set(ctx->bitmap, gpuidx, 1);
+		}
+
+		/*
+		 * If prange is already mapped or with always mapped flag,
+		 * update mapping on GPUs with ACCESS attribute
+		 */
+		if (bitmap_empty(ctx->bitmap, MAX_GPU_INSTANCE)) {
+			if (prange->mapped_to_gpu ||
+			    prange->flags & KFD_IOCTL_SVM_FLAG_GPU_ALWAYS_MAPPED)
+				bitmap_copy(ctx->bitmap, prange->bitmap_access, MAX_GPU_INSTANCE);
 		}
 	} else {
-		bitmap_or(ctx.bitmap, prange->bitmap_access,
+		bitmap_or(ctx->bitmap, prange->bitmap_access,
 			  prange->bitmap_aip, MAX_GPU_INSTANCE);
 	}
 
-	if (bitmap_empty(ctx.bitmap, MAX_GPU_INSTANCE)) {
-		if (!prange->mapped_to_gpu)
-			return 0;
-
-		bitmap_copy(ctx.bitmap, prange->bitmap_access, MAX_GPU_INSTANCE);
+	if (bitmap_empty(ctx->bitmap, MAX_GPU_INSTANCE)) {
+		r = 0;
+		goto free_ctx;
 	}
 
 	if (prange->actual_loc && !prange->ttm_res) {
@@ -1571,90 +1677,132 @@ static int svm_range_validate_and_map(struct mm_struct *mm,
 		 * svm_migrate_ram_to_vram after allocating a BO.
 		 */
 		WARN_ONCE(1, "VRAM BO missing during validation\n");
-		return -EINVAL;
+		r = -EINVAL;
+		goto free_ctx;
 	}
 
-	svm_range_reserve_bos(&ctx);
+	r = svm_range_reserve_bos(ctx, intr);
+	if (r)
+		goto free_ctx;
 
 	p = container_of(prange->svms, struct kfd_process, svms);
-	owner = kfd_svm_page_owner(p, find_first_bit(ctx.bitmap,
+	owner = kfd_svm_page_owner(p, find_first_bit(ctx->bitmap,
 						MAX_GPU_INSTANCE));
-	for_each_set_bit(idx, ctx.bitmap, MAX_GPU_INSTANCE) {
+	for_each_set_bit(idx, ctx->bitmap, MAX_GPU_INSTANCE) {
 		if (kfd_svm_page_owner(p, idx) != owner) {
 			owner = NULL;
 			break;
 		}
 	}
 
-	start = prange->start << PAGE_SHIFT;
-	end = (prange->last + 1) << PAGE_SHIFT;
-	for (addr = start; addr < end && !r; ) {
-		struct hmm_range *hmm_range;
+	start = map_start << PAGE_SHIFT;
+	end = (map_last + 1) << PAGE_SHIFT;
+	for (addr = start; !r && addr < end; ) {
+		struct amdgpu_hmm_range *range = NULL;
+		unsigned long map_start_vma;
+		unsigned long map_last_vma;
 		struct vm_area_struct *vma;
-		unsigned long next;
+		unsigned long next = 0;
 		unsigned long offset;
 		unsigned long npages;
 		bool readonly;
 
 		vma = vma_lookup(mm, addr);
-		if (!vma) {
-			r = -EFAULT;
-			goto unreserve_out;
-		}
-		readonly = !(vma->vm_flags & VM_WRITE);
+		if (vma) {
+			readonly = !(vma->vm_flags & VM_WRITE);
+
+			next = min(vma->vm_end, end);
+			npages = (next - addr) >> PAGE_SHIFT;
+			/* HMM requires at least READ permissions. If provided with PROT_NONE,
+			 * unmap the memory. If it's not already mapped, this is a no-op
+			 * If PROT_WRITE is provided without READ, warn first then unmap
+			 */
+			if (!(vma->vm_flags & VM_READ)) {
+				unsigned long e, s;
+
+				svm_range_lock(prange);
+				if (vma->vm_flags & VM_WRITE)
+					pr_debug("VM_WRITE without VM_READ is not supported");
+				s = max(start, prange->start);
+				e = min(end, prange->last);
+				if (e >= s)
+					r = svm_range_unmap_from_gpus(prange, s, e,
+						       KFD_SVM_UNMAP_TRIGGER_UNMAP_FROM_CPU);
+				svm_range_unlock(prange);
+				/* If unmap returns non-zero, we'll bail on the next for loop
+				 * iteration, so just leave r and continue
+				 */
+				addr = next;
+				continue;
+			}
 
-		next = min(vma->vm_end, end);
-		npages = (next - addr) >> PAGE_SHIFT;
-		WRITE_ONCE(p->svms.faulting_task, current);
-		r = amdgpu_hmm_range_get_pages(&prange->notifier, addr, npages,
-					       readonly, owner, NULL,
-					       &hmm_range);
-		WRITE_ONCE(p->svms.faulting_task, NULL);
-		if (r) {
-			pr_debug("failed %d to get svm range pages\n", r);
-			goto unreserve_out;
+			WRITE_ONCE(p->svms.faulting_task, current);
+			range = amdgpu_hmm_range_alloc(NULL);
+			if (likely(range))
+				r = amdgpu_hmm_range_get_pages(&prange->notifier, addr, npages,
+							       readonly, owner, range);
+			else
+				r = -ENOMEM;
+			WRITE_ONCE(p->svms.faulting_task, NULL);
+			if (r)
+				pr_debug("failed %d to get svm range pages\n", r);
+		} else {
+			r = -EFAULT;
 		}
 
-		offset = (addr - start) >> PAGE_SHIFT;
-		r = svm_range_dma_map(prange, ctx.bitmap, offset, npages,
-				      hmm_range->hmm_pfns);
-		if (r) {
-			pr_debug("failed %d to dma map range\n", r);
-			goto unreserve_out;
+		if (!r) {
+			offset = (addr >> PAGE_SHIFT) - prange->start;
+			r = svm_range_dma_map(prange, ctx->bitmap, offset, npages,
+					      range->hmm_range.hmm_pfns);
+			if (r)
+				pr_debug("failed %d to dma map range\n", r);
 		}
 
 		svm_range_lock(prange);
-		if (amdgpu_hmm_range_get_pages_done(hmm_range)) {
+
+		/* Free backing memory of hmm_range if it was initialized
+		 * Override return value to TRY AGAIN only if prior returns
+		 * were successful
+		 */
+		if (range && !amdgpu_hmm_range_valid(range) && !r) {
 			pr_debug("hmm update the range, need validate again\n");
 			r = -EAGAIN;
-			goto unlock_out;
 		}
-		if (!list_empty(&prange->child_list)) {
+
+		/* Free the hmm range */
+		amdgpu_hmm_range_free(range);
+
+		if (!r && !list_empty(&prange->child_list)) {
 			pr_debug("range split by unmap in parallel, validate again\n");
 			r = -EAGAIN;
-			goto unlock_out;
 		}
 
-		r = svm_range_map_to_gpus(prange, offset, npages, readonly,
-					  ctx.bitmap, wait, flush_tlb);
+		if (!r) {
+			map_start_vma = max(map_start, prange->start + offset);
+			map_last_vma = min(map_last, prange->start + offset + npages - 1);
+			if (map_start_vma <= map_last_vma) {
+				offset = map_start_vma - prange->start;
+				npages = map_last_vma - map_start_vma + 1;
+				r = svm_range_map_to_gpus(prange, offset, npages, readonly,
+							  ctx->bitmap, wait, flush_tlb);
+			}
+		}
+
+		if (!r && next == end)
+			prange->mapped_to_gpu = true;
 
-unlock_out:
 		svm_range_unlock(prange);
 
 		addr = next;
 	}
 
-	if (addr == end) {
-		prange->validated_once = true;
-		prange->mapped_to_gpu = true;
-	}
-
-unreserve_out:
-	svm_range_unreserve_bos(&ctx);
-
+	svm_range_unreserve_bos(ctx);
 	if (!r)
 		prange->validate_timestamp = ktime_get_boottime();
 
+free_ctx:
+	kfree(ctx);
+
 	return r;
 }
 
@@ -1731,8 +1879,8 @@ static void svm_range_restore_work(struct work_struct *work)
 		 */
 		mutex_lock(&prange->migrate_mutex);
 
-		r = svm_range_validate_and_map(mm, prange, MAX_GPU_INSTANCE,
-					       false, true, false);
+		r = svm_range_validate_and_map(mm, prange->start, prange->last, prange,
+					       MAX_GPU_INSTANCE, false, true, false);
 		if (r)
 			pr_debug("failed %d to map 0x%lx to gpus\n", r,
 				 prange->start);
@@ -1769,7 +1917,7 @@ out_reschedule:
 	/* If validation failed, reschedule another attempt */
 	if (evicted_ranges) {
 		pr_debug("reschedule to restore svm range\n");
-		schedule_delayed_work(&svms->restore_work,
+		queue_delayed_work(system_freezable_wq, &svms->restore_work,
 			msecs_to_jiffies(AMDGPU_SVM_RANGE_RESTORE_DELAY_MS));
 
 		kfd_smi_event_queue_restore_rescheduled(mm);
@@ -1783,6 +1931,7 @@ out_reschedule:
  * @mm: current process mm_struct
  * @start: starting process queue number
  * @last: last process queue number
+ * @event: mmu notifier event when range is evicted or migrated
  *
  * Stop all queues of the process to ensure GPU doesn't access the memory, then
  * return to let CPU evict the buffer and proceed CPU pagetable update.
@@ -1844,7 +1993,7 @@ svm_range_evict(struct svm_range *prange, struct mm_struct *mm,
 			pr_debug("failed to quiesce KFD\n");
 
 		pr_debug("schedule to restore svm %p ranges\n", svms);
-		schedule_delayed_work(&svms->restore_work,
+		queue_delayed_work(system_freezable_wq, &svms->restore_work,
 			msecs_to_jiffies(AMDGPU_SVM_RANGE_RESTORE_DELAY_MS));
 	} else {
 		unsigned long s, l;
@@ -1881,7 +2030,10 @@ static struct svm_range *svm_range_clone(struct svm_range *old)
 	new = svm_range_new(old->svms, old->start, old->last, false);
 	if (!new)
 		return NULL;
-
+	if (svm_range_copy_dma_addrs(new, old)) {
+		svm_range_free(new, false);
+		return NULL;
+	}
 	if (old->svm_bo) {
 		new->ttm_res = old->ttm_res;
 		new->offset = old->offset;
@@ -1896,8 +2048,10 @@ static struct svm_range *svm_range_clone(struct svm_range *old)
 	new->actual_loc = old->actual_loc;
 	new->granularity = old->granularity;
 	new->mapped_to_gpu = old->mapped_to_gpu;
+	new->vram_pages = old->vram_pages;
 	bitmap_copy(new->bitmap_access, old->bitmap_access, MAX_GPU_INSTANCE);
 	bitmap_copy(new->bitmap_aip, old->bitmap_aip, MAX_GPU_INSTANCE);
+	atomic_set(&new->queue_refcount, atomic_read(&old->queue_refcount));
 
 	return new;
 }
@@ -1906,14 +2060,23 @@ void svm_range_set_max_pages(struct amdgpu_device *adev)
 {
 	uint64_t max_pages;
 	uint64_t pages, _pages;
+	uint64_t min_pages = 0;
+	int i, id;
+
+	for (i = 0; i < adev->kfd.dev->num_nodes; i++) {
+		if (adev->kfd.dev->nodes[i]->xcp)
+			id = adev->kfd.dev->nodes[i]->xcp->id;
+		else
+			id = -1;
+		pages = KFD_XCP_MEMORY_SIZE(adev, id) >> 17;
+		pages = clamp(pages, 1ULL << 9, 1ULL << 18);
+		pages = rounddown_pow_of_two(pages);
+		min_pages = min_not_zero(min_pages, pages);
+	}
 
-	/* 1/32 VRAM size in pages */
-	pages = adev->gmc.real_vram_size >> 17;
-	pages = clamp(pages, 1ULL << 9, 1ULL << 18);
-	pages = rounddown_pow_of_two(pages);
 	do {
 		max_pages = READ_ONCE(max_svm_range_pages);
-		_pages = min_not_zero(max_pages, pages);
+		_pages = min_not_zero(max_pages, min_pages);
 	} while (cmpxchg(&max_svm_range_pages, max_pages, _pages) != max_pages);
 }
 
@@ -1952,6 +2115,7 @@ svm_range_split_new(struct svm_range_list *svms, uint64_t start, uint64_t last,
  * @update_list: output, the ranges need validate and update GPU mapping
  * @insert_list: output, the ranges need insert to svms
  * @remove_list: output, the ranges are replaced and need remove from svms
+ * @remap_list: output, remap unaligned svm ranges
  *
  * Check if the virtual address range has overlap with any existing ranges,
  * split partly overlapping ranges and add new ranges in the gaps. All changes
@@ -1975,7 +2139,7 @@ static int
 svm_range_add(struct kfd_process *p, uint64_t start, uint64_t size,
 	      uint32_t nattr, struct kfd_ioctl_svm_attribute *attrs,
 	      struct list_head *update_list, struct list_head *insert_list,
-	      struct list_head *remove_list)
+	      struct list_head *remove_list, struct list_head *remap_list)
 {
 	unsigned long last = start + size - 1UL;
 	struct svm_range_list *svms = &p->svms;
@@ -1991,6 +2155,7 @@ svm_range_add(struct kfd_process *p, uint64_t start, uint64_t size,
 	INIT_LIST_HEAD(insert_list);
 	INIT_LIST_HEAD(remove_list);
 	INIT_LIST_HEAD(&new_list);
+	INIT_LIST_HEAD(remap_list);
 
 	node = interval_tree_iter_first(&svms->objects, start, last);
 	while (node) {
@@ -2004,7 +2169,8 @@ svm_range_add(struct kfd_process *p, uint64_t start, uint64_t size,
 		next = interval_tree_iter_next(node, start, last);
 		next_start = min(node->last, last) + 1;
 
-		if (svm_range_is_same_attrs(p, prange, nattr, attrs)) {
+		if (svm_range_is_same_attrs(p, prange, nattr, attrs) &&
+		    prange->mapped_to_gpu) {
 			/* nothing to do */
 		} else if (node->start < start || node->last > last) {
 			/* node intersects the update range and its attributes
@@ -2026,14 +2192,14 @@ svm_range_add(struct kfd_process *p, uint64_t start, uint64_t size,
 			if (node->start < start) {
 				pr_debug("change old range start\n");
 				r = svm_range_split_head(prange, start,
-							 insert_list);
+							 insert_list, remap_list);
 				if (r)
 					goto out;
 			}
 			if (node->last > last) {
 				pr_debug("change old range last\n");
 				r = svm_range_split_tail(prange, last,
-							 insert_list);
+							 insert_list, remap_list);
 				if (r)
 					goto out;
 			}
@@ -2154,16 +2320,10 @@ static void svm_range_drain_retry_fault(struct svm_range_list *svms)
 {
 	struct kfd_process_device *pdd;
 	struct kfd_process *p;
-	int drain;
 	uint32_t i;
 
 	p = container_of(svms, struct kfd_process, svms);
 
-restart:
-	drain = atomic_read(&svms->drain_pagefaults);
-	if (!drain)
-		return;
-
 	for_each_set_bit(i, svms->bitmap_supported, p->n_pdds) {
 		pdd = p->pdds[i];
 		if (!pdd)
@@ -2183,8 +2343,6 @@ restart:
 
 		pr_debug("drain retry fault gpu %d svms 0x%p done\n", i, svms);
 	}
-	if (atomic_cmpxchg(&svms->drain_pagefaults, drain, 0) != drain)
-		goto restart;
 }
 
 static void svm_range_deferred_list_work(struct work_struct *work)
@@ -2206,17 +2364,8 @@ static void svm_range_deferred_list_work(struct work_struct *work)
 			 prange->start, prange->last, prange->work_item.op);
 
 		mm = prange->work_item.mm;
-retry:
-		mmap_write_lock(mm);
 
-		/* Checking for the need to drain retry faults must be inside
-		 * mmap write lock to serialize with munmap notifiers.
-		 */
-		if (unlikely(atomic_read(&svms->drain_pagefaults))) {
-			mmap_write_unlock(mm);
-			svm_range_drain_retry_fault(svms);
-			goto retry;
-		}
+		mmap_write_lock(mm);
 
 		/* Remove from deferred_list must be inside mmap write lock, for
 		 * two race cases:
@@ -2248,8 +2397,10 @@ retry:
 		mutex_unlock(&svms->lock);
 		mmap_write_unlock(mm);
 
-		/* Pairs with mmget in svm_range_add_list_work */
-		mmput(mm);
+		/* Pairs with mmget in svm_range_add_list_work. If dropping the
+		 * last mm refcount, schedule release work to avoid circular locking
+		 */
+		mmput_async(mm);
 
 		spin_lock(&svms->deferred_list_lock);
 	}
@@ -2270,15 +2421,17 @@ svm_range_add_list_work(struct svm_range_list *svms, struct svm_range *prange,
 		    prange->work_item.op != SVM_OP_UNMAP_RANGE)
 			prange->work_item.op = op;
 	} else {
-		prange->work_item.op = op;
-
-		/* Pairs with mmput in deferred_list_work */
-		mmget(mm);
-		prange->work_item.mm = mm;
-		list_add_tail(&prange->deferred_list,
-			      &prange->svms->deferred_range_list);
-		pr_debug("add prange 0x%p [0x%lx 0x%lx] to work list op %d\n",
-			 prange, prange->start, prange->last, op);
+		/* Pairs with mmput in deferred_list_work.
+		 * If process is exiting and mm is gone, don't update mmu notifier.
+		 */
+		if (mmget_not_zero(mm)) {
+			prange->work_item.mm = mm;
+			prange->work_item.op = op;
+			list_add_tail(&prange->deferred_list,
+				      &prange->svms->deferred_range_list);
+			pr_debug("add prange 0x%p [0x%lx 0x%lx] to work list op %d\n",
+				 prange, prange->start, prange->last, op);
+		}
 	}
 	spin_unlock(&svms->deferred_list_lock);
 }
@@ -2292,8 +2445,7 @@ void schedule_deferred_list_work(struct svm_range_list *svms)
 }
 
 static void
-svm_range_unmap_split(struct mm_struct *mm, struct svm_range *parent,
-		      struct svm_range *prange, unsigned long start,
+svm_range_unmap_split(struct svm_range *parent, struct svm_range *prange, unsigned long start,
 		      unsigned long last)
 {
 	struct svm_range *head;
@@ -2314,12 +2466,12 @@ svm_range_unmap_split(struct mm_struct *mm, struct svm_range *parent,
 		svm_range_split(tail, last + 1, tail->last, &head);
 
 	if (head != prange && tail != prange) {
-		svm_range_add_child(parent, mm, head, SVM_OP_UNMAP_RANGE);
-		svm_range_add_child(parent, mm, tail, SVM_OP_ADD_RANGE);
+		svm_range_add_child(parent, head, SVM_OP_UNMAP_RANGE);
+		svm_range_add_child(parent, tail, SVM_OP_ADD_RANGE);
 	} else if (tail != prange) {
-		svm_range_add_child(parent, mm, tail, SVM_OP_UNMAP_RANGE);
+		svm_range_add_child(parent, tail, SVM_OP_UNMAP_RANGE);
 	} else if (head != prange) {
-		svm_range_add_child(parent, mm, head, SVM_OP_UNMAP_RANGE);
+		svm_range_add_child(parent, head, SVM_OP_UNMAP_RANGE);
 	} else if (parent != prange) {
 		prange->work_item.op = SVM_OP_UNMAP_RANGE;
 	}
@@ -2335,6 +2487,17 @@ svm_range_unmap_from_cpu(struct mm_struct *mm, struct svm_range *prange,
 	struct kfd_process *p;
 	unsigned long s, l;
 	bool unmap_parent;
+	uint32_t i;
+
+	if (atomic_read(&prange->queue_refcount)) {
+		int r;
+
+		pr_warn("Freeing queue vital buffer 0x%lx, queue evicted\n",
+			prange->start << PAGE_SHIFT);
+		r = kgd2kfd_quiesce_mm(mm, KFD_QUEUE_EVICTION_TRIGGER_SVM);
+		if (r)
+			pr_debug("failed %d to quiesce KFD queues\n", r);
+	}
 
 	p = kfd_lookup_process_by_mm(mm);
 	if (!p)
@@ -2344,11 +2507,38 @@ svm_range_unmap_from_cpu(struct mm_struct *mm, struct svm_range *prange,
 	pr_debug("svms 0x%p prange 0x%p [0x%lx 0x%lx] [0x%lx 0x%lx]\n", svms,
 		 prange, prange->start, prange->last, start, last);
 
-	/* Make sure pending page faults are drained in the deferred worker
-	 * before the range is freed to avoid straggler interrupts on
-	 * unmapped memory causing "phantom faults".
+	/* calculate time stamps that are used to decide which page faults need be
+	 * dropped or handled before unmap pages from gpu vm
 	 */
-	atomic_inc(&svms->drain_pagefaults);
+	for_each_set_bit(i, svms->bitmap_supported, p->n_pdds) {
+		struct kfd_process_device *pdd;
+		struct amdgpu_device *adev;
+		struct amdgpu_ih_ring *ih;
+		uint32_t checkpoint_wptr;
+
+		pdd = p->pdds[i];
+		if (!pdd)
+			continue;
+
+		adev = pdd->dev->adev;
+
+		/* Check and drain ih1 ring if cam not available */
+		if (adev->irq.ih1.ring_size) {
+			ih = &adev->irq.ih1;
+			checkpoint_wptr = amdgpu_ih_get_wptr(adev, ih);
+			if (ih->rptr != checkpoint_wptr) {
+				svms->checkpoint_ts[i] =
+					amdgpu_ih_decode_iv_ts(adev, ih, checkpoint_wptr, -1);
+				continue;
+			}
+		}
+
+		/* check if dev->irq.ih_soft is not empty */
+		ih = &adev->irq.ih_soft;
+		checkpoint_wptr = amdgpu_ih_get_wptr(adev, ih);
+		if (ih->rptr != checkpoint_wptr)
+			svms->checkpoint_ts[i] = amdgpu_ih_decode_iv_ts(adev, ih, checkpoint_wptr, -1);
+	}
 
 	unmap_parent = start <= prange->start && last >= prange->last;
 
@@ -2358,14 +2548,14 @@ svm_range_unmap_from_cpu(struct mm_struct *mm, struct svm_range *prange,
 		l = min(last, pchild->last);
 		if (l >= s)
 			svm_range_unmap_from_gpus(pchild, s, l, trigger);
-		svm_range_unmap_split(mm, prange, pchild, start, last);
+		svm_range_unmap_split(prange, pchild, start, last);
 		mutex_unlock(&pchild->lock);
 	}
 	s = max(start, prange->start);
 	l = min(last, prange->last);
 	if (l >= s)
 		svm_range_unmap_from_gpus(prange, s, l, trigger);
-	svm_range_unmap_split(mm, prange, prange, start, last);
+	svm_range_unmap_split(prange, prange, start, last);
 
 	if (unmap_parent)
 		svm_range_add_list_work(svms, prange, mm, SVM_OP_UNMAP_RANGE);
@@ -2408,8 +2598,6 @@ svm_range_cpu_invalidate_pagetables(struct mmu_interval_notifier *mni,
 
 	if (range->event == MMU_NOTIFY_RELEASE)
 		return true;
-	if (!mmget_not_zero(mni->mm))
-		return true;
 
 	start = mni->interval_tree.start;
 	last = mni->interval_tree.last;
@@ -2436,7 +2624,6 @@ svm_range_cpu_invalidate_pagetables(struct mmu_interval_notifier *mni,
 	}
 
 	svm_range_unlock(prange);
-	mmput(mni->mm);
 
 	return true;
 }
@@ -2507,29 +2694,31 @@ svm_range_from_addr(struct svm_range_list *svms, unsigned long addr,
  */
 static int32_t
 svm_range_best_restore_location(struct svm_range *prange,
-				struct amdgpu_device *adev,
+				struct kfd_node *node,
 				int32_t *gpuidx)
 {
-	struct amdgpu_device *bo_adev, *preferred_adev;
+	struct kfd_node *bo_node, *preferred_node;
 	struct kfd_process *p;
 	uint32_t gpuid;
 	int r;
 
 	p = container_of(prange->svms, struct kfd_process, svms);
 
-	r = kfd_process_gpuid_from_adev(p, adev, &gpuid, gpuidx);
+	r = kfd_process_gpuid_from_node(p, node, &gpuid, gpuidx);
 	if (r < 0) {
 		pr_debug("failed to get gpuid from kgd\n");
 		return -1;
 	}
 
+	if (node->adev->apu_prefer_gtt)
+		return 0;
+
 	if (prange->preferred_loc == gpuid ||
 	    prange->preferred_loc == KFD_IOCTL_SVM_LOCATION_SYSMEM) {
 		return prange->preferred_loc;
 	} else if (prange->preferred_loc != KFD_IOCTL_SVM_LOCATION_UNDEFINED) {
-		preferred_adev = svm_range_get_adev_by_id(prange,
-							prange->preferred_loc);
-		if (amdgpu_xgmi_same_hive(adev, preferred_adev))
+		preferred_node = svm_range_get_node_by_id(prange, prange->preferred_loc);
+		if (preferred_node && svm_nodes_in_same_hive(node, preferred_node))
 			return prange->preferred_loc;
 		/* fall through */
 	}
@@ -2541,8 +2730,8 @@ svm_range_best_restore_location(struct svm_range *prange,
 		if (!prange->actual_loc)
 			return 0;
 
-		bo_adev = svm_range_get_adev_by_id(prange, prange->actual_loc);
-		if (amdgpu_xgmi_same_hive(adev, bo_adev))
+		bo_node = svm_range_get_node_by_id(prange, prange->actual_loc);
+		if (bo_node && svm_nodes_in_same_hive(node, bo_node))
 			return prange->actual_loc;
 		else
 			return 0;
@@ -2558,6 +2747,7 @@ svm_range_get_range_boundaries(struct kfd_process *p, int64_t addr,
 {
 	struct vm_area_struct *vma;
 	struct interval_tree_node *node;
+	struct rb_node *rb_node;
 	unsigned long start_limit, end_limit;
 
 	vma = vma_lookup(p->mm, addr << PAGE_SHIFT);
@@ -2566,30 +2756,27 @@ svm_range_get_range_boundaries(struct kfd_process *p, int64_t addr,
 		return -EFAULT;
 	}
 
-	*is_heap_stack = (vma->vm_start <= vma->vm_mm->brk &&
-			  vma->vm_end >= vma->vm_mm->start_brk) ||
-			 (vma->vm_start <= vma->vm_mm->start_stack &&
-			  vma->vm_end >= vma->vm_mm->start_stack);
+	*is_heap_stack = vma_is_initial_heap(vma) || vma_is_initial_stack(vma);
 
 	start_limit = max(vma->vm_start >> PAGE_SHIFT,
-		      (unsigned long)ALIGN_DOWN(addr, 2UL << 8));
+		      (unsigned long)ALIGN_DOWN(addr, 1UL << p->svms.default_granularity));
 	end_limit = min(vma->vm_end >> PAGE_SHIFT,
-		    (unsigned long)ALIGN(addr + 1, 2UL << 8));
+		    (unsigned long)ALIGN(addr + 1, 1UL << p->svms.default_granularity));
+
 	/* First range that starts after the fault address */
 	node = interval_tree_iter_first(&p->svms.objects, addr + 1, ULONG_MAX);
 	if (node) {
 		end_limit = min(end_limit, node->start);
 		/* Last range that ends before the fault address */
-		node = container_of(rb_prev(&node->rb),
-				    struct interval_tree_node, rb);
+		rb_node = rb_prev(&node->rb);
 	} else {
 		/* Last range must end before addr because
 		 * there was no range after addr
 		 */
-		node = container_of(rb_last(&p->svms.objects.rb_root),
-				    struct interval_tree_node, rb);
+		rb_node = rb_last(&p->svms.objects.rb_root);
 	}
-	if (node) {
+	if (rb_node) {
+		node = container_of(rb_node, struct interval_tree_node, rb);
 		if (node->last >= addr) {
 			WARN(1, "Overlap with prev node and page fault addr\n");
 			return -EFAULT;
@@ -2659,7 +2846,7 @@ svm_range_check_vm_userptr(struct kfd_process *p, uint64_t start, uint64_t last,
 }
 
 static struct
-svm_range *svm_range_create_unregistered_range(struct amdgpu_device *adev,
+svm_range *svm_range_create_unregistered_range(struct kfd_node *node,
 						struct kfd_process *p,
 						struct mm_struct *mm,
 						int64_t addr)
@@ -2694,7 +2881,7 @@ svm_range *svm_range_create_unregistered_range(struct amdgpu_device *adev,
 		pr_debug("Failed to create prange in address [0x%llx]\n", addr);
 		return NULL;
 	}
-	if (kfd_process_gpuid_from_adev(p, adev, &gpuid, &gpuidx)) {
+	if (kfd_process_gpuid_from_node(p, node, &gpuid, &gpuidx)) {
 		pr_debug("failed to get gpuid from kgd\n");
 		svm_range_free(prange, true);
 		return NULL;
@@ -2748,7 +2935,7 @@ static bool svm_range_skip_recover(struct svm_range *prange)
 }
 
 static void
-svm_range_count_fault(struct amdgpu_device *adev, struct kfd_process *p,
+svm_range_count_fault(struct kfd_node *node, struct kfd_process *p,
 		      int32_t gpuidx)
 {
 	struct kfd_process_device *pdd;
@@ -2761,7 +2948,7 @@ svm_range_count_fault(struct amdgpu_device *adev, struct kfd_process *p,
 		uint32_t gpuid;
 		int r;
 
-		r = kfd_process_gpuid_from_adev(p, adev, &gpuid, &gpuidx);
+		r = kfd_process_gpuid_from_node(p, node, &gpuid, &gpuidx);
 		if (r < 0)
 			return;
 	}
@@ -2789,26 +2976,29 @@ svm_fault_allowed(struct vm_area_struct *vma, bool write_fault)
 
 int
 svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid,
-			uint64_t addr, bool write_fault)
+			uint32_t vmid, uint32_t node_id,
+			uint64_t addr, uint64_t ts, bool write_fault)
 {
+	unsigned long start, last, size;
 	struct mm_struct *mm = NULL;
 	struct svm_range_list *svms;
 	struct svm_range *prange;
 	struct kfd_process *p;
 	ktime_t timestamp = ktime_get_boottime();
+	struct kfd_node *node;
 	int32_t best_loc;
-	int32_t gpuidx = MAX_GPU_INSTANCE;
+	int32_t gpuid, gpuidx = MAX_GPU_INSTANCE;
 	bool write_locked = false;
 	struct vm_area_struct *vma;
 	bool migration = false;
 	int r = 0;
 
-	if (!KFD_IS_SVM_API_SUPPORTED(adev->kfd.dev)) {
+	if (!KFD_IS_SVM_API_SUPPORTED(adev)) {
 		pr_debug("device does not support SVM\n");
 		return -EFAULT;
 	}
 
-	p = kfd_lookup_process_by_pasid(pasid);
+	p = kfd_lookup_process_by_pasid(pasid, NULL);
 	if (!p) {
 		pr_debug("kfd process not founded pasid 0x%x\n", pasid);
 		return 0;
@@ -2818,11 +3008,25 @@ svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid,
 	pr_debug("restoring svms 0x%p fault address 0x%llx\n", svms, addr);
 
 	if (atomic_read(&svms->drain_pagefaults)) {
-		pr_debug("draining retry fault, drop fault 0x%llx\n", addr);
+		pr_debug("page fault handling disabled, drop fault 0x%llx\n", addr);
 		r = 0;
 		goto out;
 	}
 
+	node = kfd_node_by_irq_ids(adev, node_id, vmid);
+	if (!node) {
+		pr_debug("kfd node does not exist node_id: %d, vmid: %d\n", node_id,
+			 vmid);
+		r = -EFAULT;
+		goto out;
+	}
+
+	if (kfd_process_gpuid_from_node(p, node, &gpuid, &gpuidx)) {
+		pr_debug("failed to get gpuid/gpuidex for node_id: %d\n", node_id);
+		r = -EFAULT;
+		goto out;
+	}
+
 	if (!p->xnack_enabled) {
 		pr_debug("XNACK not enabled for pasid 0x%x\n", pasid);
 		r = -EFAULT;
@@ -2842,6 +3046,23 @@ svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid,
 	mmap_read_lock(mm);
 retry_write_locked:
 	mutex_lock(&svms->lock);
+
+	/* check if this page fault time stamp is before svms->checkpoint_ts */
+	if (svms->checkpoint_ts[gpuidx] != 0) {
+		if (amdgpu_ih_ts_after_or_equal(ts,  svms->checkpoint_ts[gpuidx])) {
+			pr_debug("draining retry fault, drop fault 0x%llx\n", addr);
+			if (write_locked)
+				mmap_write_downgrade(mm);
+			r = -EAGAIN;
+			goto out_unlock_svms;
+		} else {
+			/* ts is after svms->checkpoint_ts now, reset svms->checkpoint_ts
+			 * to zero to avoid following ts wrap around give wrong comparing
+			 */
+			svms->checkpoint_ts[gpuidx] = 0;
+		}
+	}
+
 	prange = svm_range_from_addr(svms, addr, NULL);
 	if (!prange) {
 		pr_debug("failed to find prange svms 0x%p address [0x%llx]\n",
@@ -2857,7 +3078,7 @@ retry_write_locked:
 			write_locked = true;
 			goto retry_write_locked;
 		}
-		prange = svm_range_create_unregistered_range(adev, p, mm, addr);
+		prange = svm_range_create_unregistered_range(node, p, mm, addr);
 		if (!prange) {
 			pr_debug("failed to create unregistered range svms 0x%p address [0x%llx]\n",
 				 svms, addr);
@@ -2872,7 +3093,7 @@ retry_write_locked:
 	mutex_lock(&prange->migrate_mutex);
 
 	if (svm_range_skip_recover(prange)) {
-		amdgpu_gmc_filter_faults_remove(adev, addr, pasid);
+		amdgpu_gmc_filter_faults_remove(node->adev, addr, pasid);
 		r = 0;
 		goto out_unlock_range;
 	}
@@ -2903,7 +3124,7 @@ retry_write_locked:
 		goto out_unlock_range;
 	}
 
-	best_loc = svm_range_best_restore_location(prange, adev, &gpuidx);
+	best_loc = svm_range_best_restore_location(prange, node, &gpuidx);
 	if (best_loc == -1) {
 		pr_debug("svms %p failed get best restore loc [0x%lx 0x%lx]\n",
 			 svms, prange->start, prange->last);
@@ -2915,45 +3136,50 @@ retry_write_locked:
 		 svms, prange->start, prange->last, best_loc,
 		 prange->actual_loc);
 
-	kfd_smi_event_page_fault_start(adev->kfd.dev, p->lead_thread->pid, addr,
+	kfd_smi_event_page_fault_start(node, p->lead_thread->pid, addr,
 				       write_fault, timestamp);
 
-	if (prange->actual_loc != best_loc) {
-		migration = true;
+	/* Align migration range start and size to granularity size */
+	size = 1UL << prange->granularity;
+	start = max_t(unsigned long, ALIGN_DOWN(addr, size), prange->start);
+	last = min_t(unsigned long, ALIGN(addr + 1, size) - 1, prange->last);
+	if (prange->actual_loc != 0 || best_loc != 0) {
 		if (best_loc) {
-			r = svm_migrate_to_vram(prange, best_loc, mm,
-					KFD_MIGRATE_TRIGGER_PAGEFAULT_GPU);
+			r = svm_migrate_to_vram(prange, best_loc, start, last,
+					mm, KFD_MIGRATE_TRIGGER_PAGEFAULT_GPU);
 			if (r) {
 				pr_debug("svm_migrate_to_vram failed (%d) at %llx, falling back to system memory\n",
 					 r, addr);
 				/* Fallback to system memory if migration to
 				 * VRAM failed
 				 */
-				if (prange->actual_loc)
-					r = svm_migrate_vram_to_ram(prange, mm,
-					   KFD_MIGRATE_TRIGGER_PAGEFAULT_GPU,
-					   NULL);
+				if (prange->actual_loc && prange->actual_loc != best_loc)
+					r = svm_migrate_vram_to_ram(prange, mm, start, last,
+						KFD_MIGRATE_TRIGGER_PAGEFAULT_GPU, NULL);
 				else
 					r = 0;
 			}
 		} else {
-			r = svm_migrate_vram_to_ram(prange, mm,
-					KFD_MIGRATE_TRIGGER_PAGEFAULT_GPU,
-					NULL);
+			r = svm_migrate_vram_to_ram(prange, mm, start, last,
+					KFD_MIGRATE_TRIGGER_PAGEFAULT_GPU, NULL);
 		}
 		if (r) {
 			pr_debug("failed %d to migrate svms %p [0x%lx 0x%lx]\n",
-				 r, svms, prange->start, prange->last);
-			goto out_unlock_range;
+				 r, svms, start, last);
+			goto out_migrate_fail;
+		} else {
+			migration = true;
 		}
 	}
 
-	r = svm_range_validate_and_map(mm, prange, gpuidx, false, false, false);
+	r = svm_range_validate_and_map(mm, start, last, prange, gpuidx, false,
+				       false, false);
 	if (r)
 		pr_debug("failed %d to map svms 0x%p [0x%lx 0x%lx] to gpus\n",
-			 r, svms, prange->start, prange->last);
+			 r, svms, start, last);
 
-	kfd_smi_event_page_fault_end(adev->kfd.dev, p->lead_thread->pid, addr,
+out_migrate_fail:
+	kfd_smi_event_page_fault_end(node, p->lead_thread->pid, addr,
 				     migration);
 
 out_unlock_range:
@@ -2962,7 +3188,8 @@ out_unlock_svms:
 	mutex_unlock(&svms->lock);
 	mmap_read_unlock(mm);
 
-	svm_range_count_fault(adev, p, gpuidx);
+	if (r != -EAGAIN)
+		svm_range_count_fault(node, p, gpuidx);
 
 	mmput(mm);
 out:
@@ -2970,7 +3197,7 @@ out:
 
 	if (r == -EAGAIN) {
 		pr_debug("recover vm fault later\n");
-		amdgpu_gmc_filter_faults_remove(adev, addr, pasid);
+		amdgpu_gmc_filter_faults_remove(node->adev, addr, pasid);
 		r = 0;
 	}
 	return r;
@@ -2994,10 +3221,10 @@ svm_range_switch_xnack_reserve_mem(struct kfd_process *p, bool xnack_enabled)
 			size = (pchild->last - pchild->start + 1) << PAGE_SHIFT;
 			if (xnack_enabled) {
 				amdgpu_amdkfd_unreserve_mem_limit(NULL, size,
-						KFD_IOC_ALLOC_MEM_FLAGS_USERPTR);
+					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0);
 			} else {
 				r = amdgpu_amdkfd_reserve_mem_limit(NULL, size,
-						KFD_IOC_ALLOC_MEM_FLAGS_USERPTR);
+					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0);
 				if (r)
 					goto out_unlock;
 				reserved_size += size;
@@ -3007,10 +3234,10 @@ svm_range_switch_xnack_reserve_mem(struct kfd_process *p, bool xnack_enabled)
 		size = (prange->last - prange->start + 1) << PAGE_SHIFT;
 		if (xnack_enabled) {
 			amdgpu_amdkfd_unreserve_mem_limit(NULL, size,
-						KFD_IOC_ALLOC_MEM_FLAGS_USERPTR);
+					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0);
 		} else {
 			r = amdgpu_amdkfd_reserve_mem_limit(NULL, size,
-						KFD_IOC_ALLOC_MEM_FLAGS_USERPTR);
+					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0);
 			if (r)
 				goto out_unlock;
 			reserved_size += size;
@@ -3023,7 +3250,7 @@ out_unlock:
 
 	if (r)
 		amdgpu_amdkfd_unreserve_mem_limit(NULL, reserved_size,
-						KFD_IOC_ALLOC_MEM_FLAGS_USERPTR);
+					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0);
 	else
 		/* Change xnack mode must be inside svms lock, to avoid race with
 		 * svm_range_deferred_list_work unreserve memory in parallel.
@@ -3039,7 +3266,8 @@ void svm_range_list_fini(struct kfd_process *p)
 	struct svm_range *prange;
 	struct svm_range *next;
 
-	pr_debug("pasid 0x%x svms 0x%p\n", p->pasid, &p->svms);
+	pr_debug("process pid %d svms 0x%p\n", p->lead_thread->pid,
+		 &p->svms);
 
 	cancel_delayed_work_sync(&p->svms.restore_work);
 
@@ -3049,8 +3277,9 @@ void svm_range_list_fini(struct kfd_process *p)
 	/*
 	 * Ensure no retry fault comes in afterwards, as page fault handler will
 	 * not find kfd process and take mm lock to recover fault.
+	 * stop kfd page fault handing, then wait pending page faults got drained
 	 */
-	atomic_inc(&p->svms.drain_pagefaults);
+	atomic_set(&p->svms.drain_pagefaults, 1);
 	svm_range_drain_retry_fault(&p->svms);
 
 	list_for_each_entry_safe(prange, next, &p->svms.list, list) {
@@ -3061,7 +3290,8 @@ void svm_range_list_fini(struct kfd_process *p)
 
 	mutex_destroy(&p->svms.lock);
 
-	pr_debug("pasid 0x%x svms 0x%p done\n", p->pasid, &p->svms);
+	pr_debug("process pid %d svms 0x%p done\n",
+		p->lead_thread->pid, &p->svms);
 }
 
 int svm_range_list_init(struct kfd_process *p)
@@ -3081,9 +3311,15 @@ int svm_range_list_init(struct kfd_process *p)
 	spin_lock_init(&svms->deferred_list_lock);
 
 	for (i = 0; i < p->n_pdds; i++)
-		if (KFD_IS_SVM_API_SUPPORTED(p->pdds[i]->dev))
+		if (KFD_IS_SVM_API_SUPPORTED(p->pdds[i]->dev->adev))
 			bitmap_set(svms->bitmap_supported, i, 1);
 
+	 /* Value of default granularity cannot exceed 0x1B, the
+	  * number of pages supported by a 4-level paging table
+	  */
+	svms->default_granularity = min_t(u8, amdgpu_svm_default_granularity, 0x1B);
+	pr_debug("Default SVM Granularity to use: %d\n", svms->default_granularity);
+
 	return 0;
 }
 
@@ -3212,7 +3448,7 @@ svm_range_best_prefetch_location(struct svm_range *prange)
 	DECLARE_BITMAP(bitmap, MAX_GPU_INSTANCE);
 	uint32_t best_loc = prange->prefetch_loc;
 	struct kfd_process_device *pdd;
-	struct amdgpu_device *bo_adev;
+	struct kfd_node *bo_node;
 	struct kfd_process *p;
 	uint32_t gpuidx;
 
@@ -3221,9 +3457,14 @@ svm_range_best_prefetch_location(struct svm_range *prange)
 	if (!best_loc || best_loc == KFD_IOCTL_SVM_LOCATION_UNDEFINED)
 		goto out;
 
-	bo_adev = svm_range_get_adev_by_id(prange, best_loc);
-	if (!bo_adev) {
-		WARN_ONCE(1, "failed to get device by id 0x%x\n", best_loc);
+	bo_node = svm_range_get_node_by_id(prange, best_loc);
+	if (!bo_node) {
+		WARN_ONCE(1, "failed to get valid kfd node at id%x\n", best_loc);
+		best_loc = 0;
+		goto out;
+	}
+
+	if (bo_node->adev->apu_prefer_gtt) {
 		best_loc = 0;
 		goto out;
 	}
@@ -3241,10 +3482,10 @@ svm_range_best_prefetch_location(struct svm_range *prange)
 			continue;
 		}
 
-		if (pdd->dev->adev == bo_adev)
+		if (pdd->dev->adev == bo_node->adev)
 			continue;
 
-		if (!amdgpu_xgmi_same_hive(pdd->dev->adev, bo_adev)) {
+		if (!svm_nodes_in_same_hive(pdd->dev, bo_node)) {
 			best_loc = 0;
 			break;
 		}
@@ -3292,32 +3533,39 @@ svm_range_trigger_migration(struct mm_struct *mm, struct svm_range *prange,
 	*migrated = false;
 	best_loc = svm_range_best_prefetch_location(prange);
 
-	if (best_loc == KFD_IOCTL_SVM_LOCATION_UNDEFINED ||
-	    best_loc == prange->actual_loc)
+	/* when best_loc is a gpu node and same as prange->actual_loc
+	 * we still need do migration as prange->actual_loc !=0 does
+	 * not mean all pages in prange are vram. hmm migrate will pick
+	 * up right pages during migration.
+	 */
+	if ((best_loc == KFD_IOCTL_SVM_LOCATION_UNDEFINED) ||
+	    (best_loc == 0 && prange->actual_loc == 0))
 		return 0;
 
 	if (!best_loc) {
-		r = svm_migrate_vram_to_ram(prange, mm,
+		r = svm_migrate_vram_to_ram(prange, mm, prange->start, prange->last,
 					KFD_MIGRATE_TRIGGER_PREFETCH, NULL);
 		*migrated = !r;
 		return r;
 	}
 
-	r = svm_migrate_to_vram(prange, best_loc, mm, KFD_MIGRATE_TRIGGER_PREFETCH);
+	r = svm_migrate_to_vram(prange, best_loc, prange->start, prange->last,
+				mm, KFD_MIGRATE_TRIGGER_PREFETCH);
 	*migrated = !r;
 
-	return r;
+	return 0;
 }
 
 int svm_range_schedule_evict_svm_bo(struct amdgpu_amdkfd_fence *fence)
 {
-	if (!fence)
-		return -EINVAL;
-
-	if (dma_fence_is_signaled(&fence->base))
-		return 0;
-
-	if (fence->svm_bo) {
+	/* Dereferencing fence->svm_bo is safe here because the fence hasn't
+	 * signaled yet and we're under the protection of the fence->lock.
+	 * After the fence is signaled in svm_range_bo_release, we cannot get
+	 * here any more.
+	 *
+	 * Reference is dropped in svm_range_evict_svm_bo_worker.
+	 */
+	if (svm_bo_ref_unless_zero(fence->svm_bo)) {
 		WRITE_ONCE(fence->svm_bo->evicting, 1);
 		schedule_work(&fence->svm_bo->eviction_work);
 	}
@@ -3332,8 +3580,6 @@ static void svm_range_evict_svm_bo_worker(struct work_struct *work)
 	int r = 0;
 
 	svm_bo = container_of(work, struct svm_range_bo, eviction_work);
-	if (!svm_bo_ref_unless_zero(svm_bo))
-		return; /* svm_bo was freed while eviction was pending */
 
 	if (mmget_not_zero(svm_bo->eviction_fence->mm)) {
 		mm = svm_bo->eviction_fence->mm;
@@ -3358,7 +3604,11 @@ static void svm_range_evict_svm_bo_worker(struct work_struct *work)
 
 		mutex_lock(&prange->migrate_mutex);
 		do {
+			/* migrate all vram pages in this prange to sys ram
+			 * after that prange->actual_loc should be zero
+			 */
 			r = svm_migrate_vram_to_ram(prange, mm,
+					prange->start, prange->last,
 					KFD_MIGRATE_TRIGGER_TTM_EVICTION, NULL);
 		} while (!r && prange->actual_loc && --retries);
 
@@ -3396,15 +3646,16 @@ svm_range_set_attr(struct kfd_process *p, struct mm_struct *mm,
 	struct list_head update_list;
 	struct list_head insert_list;
 	struct list_head remove_list;
+	struct list_head remap_list;
 	struct svm_range_list *svms;
 	struct svm_range *prange;
 	struct svm_range *next;
 	bool update_mapping = false;
 	bool flush_tlb;
-	int r = 0;
+	int r, ret = 0;
 
-	pr_debug("pasid 0x%x svms 0x%p [0x%llx 0x%llx] pages 0x%llx\n",
-		 p->pasid, &p->svms, start, start + size - 1, size);
+	pr_debug("process pid %d svms 0x%p [0x%llx 0x%llx] pages 0x%llx\n",
+		 p->lead_thread->pid, &p->svms, start, start + size - 1, size);
 
 	r = svm_range_check_attr(p, nattr, attrs);
 	if (r)
@@ -3427,7 +3678,7 @@ svm_range_set_attr(struct kfd_process *p, struct mm_struct *mm,
 
 	/* Add new range and split existing ranges as needed */
 	r = svm_range_add(p, start, size, nattr, attrs, &update_list,
-			  &insert_list, &remove_list);
+			  &insert_list, &remove_list, &remap_list);
 	if (r) {
 		mutex_unlock(&svms->lock);
 		mmap_write_unlock(mm);
@@ -3442,6 +3693,8 @@ svm_range_set_attr(struct kfd_process *p, struct mm_struct *mm,
 		svm_range_apply_attrs(p, prange, nattr, attrs, &update_mapping);
 		/* TODO: unmap ranges from GPU that lost access */
 	}
+	update_mapping |= !p->xnack_enabled && !list_empty(&remap_list);
+
 	list_for_each_entry_safe(prange, next, &remove_list, update_list) {
 		pr_debug("unlink old 0x%p prange 0x%p [0x%lx 0x%lx]\n",
 			 prange->svms, prange, prange->start,
@@ -3481,28 +3734,41 @@ svm_range_set_attr(struct kfd_process *p, struct mm_struct *mm,
 
 		flush_tlb = !migrated && update_mapping && prange->mapped_to_gpu;
 
-		r = svm_range_validate_and_map(mm, prange, MAX_GPU_INSTANCE,
-					       true, true, flush_tlb);
+		r = svm_range_validate_and_map(mm, prange->start, prange->last, prange,
+					       MAX_GPU_INSTANCE, true, true, flush_tlb);
 		if (r)
 			pr_debug("failed %d to map svm range\n", r);
 
 out_unlock_range:
 		mutex_unlock(&prange->migrate_mutex);
 		if (r)
-			break;
+			ret = r;
+	}
+
+	list_for_each_entry(prange, &remap_list, update_list) {
+		pr_debug("Remapping prange 0x%p [0x%lx 0x%lx]\n",
+			 prange, prange->start, prange->last);
+		mutex_lock(&prange->migrate_mutex);
+		r = svm_range_validate_and_map(mm,  prange->start, prange->last, prange,
+					       MAX_GPU_INSTANCE, true, true, prange->mapped_to_gpu);
+		if (r)
+			pr_debug("failed %d on remap svm range\n", r);
+		mutex_unlock(&prange->migrate_mutex);
+		if (r)
+			ret = r;
 	}
 
-	svm_range_debug_dump(svms);
+	dynamic_svm_range_dump(svms);
 
 	mutex_unlock(&svms->lock);
 	mmap_read_unlock(mm);
 out:
 	mutex_unlock(&process_info->lock);
 
-	pr_debug("pasid 0x%x svms 0x%p [0x%llx 0x%llx] done, r=%d\n", p->pasid,
-		 &p->svms, start, start + size - 1, r);
+	pr_debug("process pid %d svms 0x%p [0x%llx 0x%llx] done, r=%d\n",
+		 p->lead_thread->pid, &p->svms, start, start + size - 1, r);
 
-	return r;
+	return ret ? ret : r;
 }
 
 static int
@@ -3583,7 +3849,7 @@ svm_range_get_attr(struct kfd_process *p, struct mm_struct *mm,
 	node = interval_tree_iter_first(&svms->objects, start, last);
 	if (!node) {
 		pr_debug("range attrs not found return default values\n");
-		svm_range_set_default_attributes(&location, &prefetch_loc,
+		svm_range_set_default_attributes(svms, &location, &prefetch_loc,
 						 &granularity, &flags_and);
 		flags_or = flags_and;
 		if (p->xnack_enabled)
@@ -3838,8 +4104,8 @@ exit:
 	return ret;
 }
 
-int svm_range_get_info(struct kfd_process *p, uint32_t *num_svm_ranges,
-		       uint64_t *svm_priv_data_size)
+void svm_range_get_info(struct kfd_process *p, uint32_t *num_svm_ranges,
+			uint64_t *svm_priv_data_size)
 {
 	uint64_t total_size, accessibility_size, common_attr_size;
 	int nattr_common = 4, nattr_accessibility = 1;
@@ -3851,8 +4117,6 @@ int svm_range_get_info(struct kfd_process *p, uint32_t *num_svm_ranges,
 	*svm_priv_data_size = 0;
 
 	svms = &p->svms;
-	if (!svms)
-		return -EINVAL;
 
 	mutex_lock(&svms->lock);
 	list_for_each_entry(prange, &svms->list, list) {
@@ -3894,7 +4158,6 @@ int svm_range_get_info(struct kfd_process *p, uint32_t *num_svm_ranges,
 
 	pr_debug("num_svm_ranges %u total_priv_size %llu\n", *num_svm_ranges,
 		 *svm_priv_data_size);
-	return 0;
 }
 
 int kfd_criu_checkpoint_svm(struct kfd_process *p,
@@ -3911,8 +4174,6 @@ int kfd_criu_checkpoint_svm(struct kfd_process *p,
 	struct mm_struct *mm;
 
 	svms = &p->svms;
-	if (!svms)
-		return -EINVAL;
 
 	mm = get_task_mm(p->lead_thread);
 	if (!mm) {
@@ -4010,7 +4271,7 @@ svm_ioctl(struct kfd_process *p, enum kfd_ioctl_svm_op op, uint64_t start,
 		r = svm_range_get_attr(p, mm, start, size, nattrs, attrs);
 		break;
 	default:
-		r = EINVAL;
+		r = -EINVAL;
 		break;
 	}
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.h b/drivers/gpu/drm/amd/amdkfd/kfd_svm.h
index 7a33b93f9df6..a63dfc95b602 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.h
@@ -31,7 +31,6 @@
 #include <linux/list.h>
 #include <linux/mutex.h>
 #include <linux/sched/mm.h>
-#include <linux/hmm.h>
 #include "amdgpu.h"
 #include "kfd_priv.h"
 
@@ -48,6 +47,7 @@ struct svm_range_bo {
 	struct work_struct		eviction_work;
 	uint32_t			evicting;
 	struct work_struct		release_work;
+	struct kfd_node			*node;
 };
 
 enum svm_work_list_ops {
@@ -77,6 +77,7 @@ struct svm_work_list_item {
  * @update_list:link list node used to add to update_list
  * @mapping:    bo_va mapping structure to create and update GPU page table
  * @npages:     number of pages
+ * @vram_pages: vram pages number in this svm_range
  * @dma_addr:   dma mapping address on each GPU for system memory physical page
  * @ttm_res:    vram ttm resource map
  * @offset:     range start offset within mm_nodes
@@ -87,7 +88,9 @@ struct svm_work_list_item {
  * @flags:      flags defined as KFD_IOCTL_SVM_FLAG_*
  * @perferred_loc: perferred location, 0 for CPU, or GPU id
  * @perfetch_loc: last prefetch location, 0 for CPU, or GPU id
- * @actual_loc: the actual location, 0 for CPU, or GPU id
+ * @actual_loc: this svm_range location. 0: all pages are from sys ram;
+ *              GPU id: this svm_range may include vram pages from GPU with
+ *              id actual_loc.
  * @granularity:migration granularity, log2 num pages
  * @invalid:    not 0 means cpu page table is invalidated
  * @validate_timestamp: system timestamp when range is validated
@@ -111,6 +114,7 @@ struct svm_range {
 	struct list_head		list;
 	struct list_head		update_list;
 	uint64_t			npages;
+	uint64_t			vram_pages;
 	dma_addr_t			*dma_addr[MAX_GPU_INSTANCE];
 	struct ttm_resource		*ttm_res;
 	uint64_t			offset;
@@ -131,8 +135,8 @@ struct svm_range {
 	struct list_head		child_list;
 	DECLARE_BITMAP(bitmap_access, MAX_GPU_INSTANCE);
 	DECLARE_BITMAP(bitmap_aip, MAX_GPU_INSTANCE);
-	bool				validated_once;
 	bool				mapped_to_gpu;
+	atomic_t			queue_refcount;
 };
 
 static inline void svm_range_lock(struct svm_range *prange)
@@ -163,26 +167,24 @@ int svm_ioctl(struct kfd_process *p, enum kfd_ioctl_svm_op op, uint64_t start,
 struct svm_range *svm_range_from_addr(struct svm_range_list *svms,
 				      unsigned long addr,
 				      struct svm_range **parent);
-struct amdgpu_device *svm_range_get_adev_by_id(struct svm_range *prange,
-					       uint32_t id);
-int svm_range_vram_node_new(struct amdgpu_device *adev,
-			    struct svm_range *prange, bool clear);
+struct kfd_node *svm_range_get_node_by_id(struct svm_range *prange,
+					  uint32_t gpu_id);
+int svm_range_vram_node_new(struct kfd_node *node, struct svm_range *prange,
+			    bool clear);
 void svm_range_vram_node_free(struct svm_range *prange);
-int svm_range_split_by_granularity(struct kfd_process *p, struct mm_struct *mm,
-			       unsigned long addr, struct svm_range *parent,
-			       struct svm_range *prange);
-int svm_range_restore_pages(struct amdgpu_device *adev,
-			    unsigned int pasid, uint64_t addr, bool write_fault);
+int svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid,
+			    uint32_t vmid, uint32_t node_id, uint64_t addr, uint64_t ts,
+			    bool write_fault);
 int svm_range_schedule_evict_svm_bo(struct amdgpu_amdkfd_fence *fence);
 void svm_range_add_list_work(struct svm_range_list *svms,
 			     struct svm_range *prange, struct mm_struct *mm,
 			     enum svm_work_list_ops op);
 void schedule_deferred_list_work(struct svm_range_list *svms);
-void svm_range_dma_unmap(struct device *dev, dma_addr_t *dma_addr,
+void svm_range_dma_unmap_dev(struct device *dev, dma_addr_t *dma_addr,
 			 unsigned long offset, unsigned long npages);
-void svm_range_free_dma_mappings(struct svm_range *prange);
-int svm_range_get_info(struct kfd_process *p, uint32_t *num_svm_ranges,
-		       uint64_t *svm_priv_data_size);
+void svm_range_dma_unmap(struct svm_range *prange);
+void svm_range_get_info(struct kfd_process *p, uint32_t *num_svm_ranges,
+			uint64_t *svm_priv_data_size);
 int kfd_criu_checkpoint_svm(struct kfd_process *p,
 			    uint8_t __user *user_priv_data,
 			    uint64_t *priv_offset);
@@ -192,13 +194,14 @@ int kfd_criu_restore_svm(struct kfd_process *p,
 			 uint64_t max_priv_data_size);
 int kfd_criu_resume_svm(struct kfd_process *p);
 struct kfd_process_device *
-svm_range_get_pdd_by_adev(struct svm_range *prange, struct amdgpu_device *adev);
+svm_range_get_pdd_by_node(struct svm_range *prange, struct kfd_node *node);
 void svm_range_list_lock_and_flush_work(struct svm_range_list *svms, struct mm_struct *mm);
 
 /* SVM API and HMM page migration work together, device memory type
  * is initialized to not 0 when page migration register device memory.
  */
-#define KFD_IS_SVM_API_SUPPORTED(dev) ((dev)->pgmap.type != 0)
+#define KFD_IS_SVM_API_SUPPORTED(adev) ((adev)->kfd.pgmap.type != 0 ||\
+					((adev)->apu_prefer_gtt))
 
 void svm_range_bo_unref_async(struct svm_range_bo *svm_bo);
 
@@ -219,8 +222,9 @@ static inline void svm_range_list_fini(struct kfd_process *p)
 }
 
 static inline int svm_range_restore_pages(struct amdgpu_device *adev,
-					  unsigned int pasid, uint64_t addr,
-					  bool write_fault)
+					  unsigned int pasid,
+					  uint32_t client_id, uint32_t node_id,
+					  uint64_t addr, uint64_t ts, bool write_fault)
 {
 	return -EFAULT;
 }
@@ -232,13 +236,12 @@ static inline int svm_range_schedule_evict_svm_bo(
 	return -EINVAL;
 }
 
-static inline int svm_range_get_info(struct kfd_process *p,
-				     uint32_t *num_svm_ranges,
-				     uint64_t *svm_priv_data_size)
+static inline void svm_range_get_info(struct kfd_process *p,
+				      uint32_t *num_svm_ranges,
+				      uint64_t *svm_priv_data_size)
 {
 	*num_svm_ranges = 0;
 	*svm_priv_data_size = 0;
-	return 0;
 }
 
 static inline int kfd_criu_checkpoint_svm(struct kfd_process *p,
@@ -261,6 +264,10 @@ static inline int kfd_criu_resume_svm(struct kfd_process *p)
 	return 0;
 }
 
+static inline void svm_range_set_max_pages(struct amdgpu_device *adev)
+{
+}
+
 #define KFD_IS_SVM_API_SUPPORTED(dev) false
 
 #endif /* IS_ENABLED(CONFIG_HSA_AMD_SVM) */
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
index 8e4124dcb6e4..811636af14ea 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
@@ -31,13 +31,14 @@
 #include <linux/log2.h>
 #include <linux/dmi.h>
 #include <linux/atomic.h>
+#include <linux/crc16.h>
 
 #include "kfd_priv.h"
 #include "kfd_crat.h"
 #include "kfd_topology.h"
 #include "kfd_device_queue_manager.h"
-#include "kfd_iommu.h"
 #include "kfd_svm.h"
+#include "kfd_debug.h"
 #include "amdgpu_amdkfd.h"
 #include "amdgpu_ras.h"
 #include "amdgpu.h"
@@ -96,7 +97,7 @@ struct kfd_topology_device *kfd_topology_device_by_id(uint32_t gpu_id)
 	return ret;
 }
 
-struct kfd_dev *kfd_device_by_id(uint32_t gpu_id)
+struct kfd_node *kfd_device_by_id(uint32_t gpu_id)
 {
 	struct kfd_topology_device *top_dev;
 
@@ -107,42 +108,6 @@ struct kfd_dev *kfd_device_by_id(uint32_t gpu_id)
 	return top_dev->gpu;
 }
 
-struct kfd_dev *kfd_device_by_pci_dev(const struct pci_dev *pdev)
-{
-	struct kfd_topology_device *top_dev;
-	struct kfd_dev *device = NULL;
-
-	down_read(&topology_lock);
-
-	list_for_each_entry(top_dev, &topology_device_list, list)
-		if (top_dev->gpu && top_dev->gpu->adev->pdev == pdev) {
-			device = top_dev->gpu;
-			break;
-		}
-
-	up_read(&topology_lock);
-
-	return device;
-}
-
-struct kfd_dev *kfd_device_by_adev(const struct amdgpu_device *adev)
-{
-	struct kfd_topology_device *top_dev;
-	struct kfd_dev *device = NULL;
-
-	down_read(&topology_lock);
-
-	list_for_each_entry(top_dev, &topology_device_list, list)
-		if (top_dev->gpu && top_dev->gpu->adev == adev) {
-			device = top_dev->gpu;
-			break;
-		}
-
-	up_read(&topology_lock);
-
-	return device;
-}
-
 /* Called with write topology_lock acquired */
 static void kfd_release_topology_device(struct kfd_topology_device *dev)
 {
@@ -309,6 +274,8 @@ static ssize_t iolink_show(struct kobject *kobj, struct attribute *attr,
 			      iolink->max_bandwidth);
 	sysfs_show_32bit_prop(buffer, offs, "recommended_transfer_size",
 			      iolink->rec_transfer_size);
+	sysfs_show_32bit_prop(buffer, offs, "recommended_sdma_engine_id_mask",
+			      iolink->rec_sdma_eng_id_mask);
 	sysfs_show_32bit_prop(buffer, offs, "flags", iolink->flags);
 
 	return offs;
@@ -492,7 +459,8 @@ static ssize_t node_show(struct kobject *kobj, struct attribute *attr,
 	sysfs_show_32bit_prop(buffer, offs, "wave_front_size",
 			      dev->node_props.wave_front_size);
 	sysfs_show_32bit_prop(buffer, offs, "array_count",
-			      dev->node_props.array_count);
+			      dev->gpu ? (dev->node_props.array_count *
+					  NUM_XCC(dev->gpu->xcc_mask)) : 0);
 	sysfs_show_32bit_prop(buffer, offs, "simd_arrays_per_engine",
 			      dev->node_props.simd_arrays_per_engine);
 	sysfs_show_32bit_prop(buffer, offs, "cu_per_simd_array",
@@ -526,7 +494,7 @@ static ssize_t node_show(struct kobject *kobj, struct attribute *attr,
 
 	if (dev->gpu) {
 		log_max_watch_addr =
-			__ilog2_u32(dev->gpu->device_info.num_of_watch_points);
+			__ilog2_u32(dev->gpu->kfd->device_info.num_of_watch_points);
 
 		if (log_max_watch_addr) {
 			dev->node_props.capability |=
@@ -542,20 +510,33 @@ static ssize_t node_show(struct kobject *kobj, struct attribute *attr,
 			dev->node_props.capability |=
 					HSA_CAP_AQL_QUEUE_DOUBLE_MAP;
 
+		if (KFD_GC_VERSION(dev->gpu) < IP_VERSION(10, 0, 0) &&
+			(dev->gpu->adev->sdma.supported_reset & AMDGPU_RESET_TYPE_PER_QUEUE))
+				dev->node_props.capability2 |= HSA_CAP2_PER_SDMA_QUEUE_RESET_SUPPORTED;
+
 		sysfs_show_32bit_prop(buffer, offs, "max_engine_clk_fcompute",
 			dev->node_props.max_engine_clk_fcompute);
 
 		sysfs_show_64bit_prop(buffer, offs, "local_mem_size", 0ULL);
 
 		sysfs_show_32bit_prop(buffer, offs, "fw_version",
-				      dev->gpu->mec_fw_version);
+				      dev->gpu->kfd->mec_fw_version);
 		sysfs_show_32bit_prop(buffer, offs, "capability",
 				      dev->node_props.capability);
+		sysfs_show_32bit_prop(buffer, offs, "capability2",
+				      dev->node_props.capability2);
+		sysfs_show_64bit_prop(buffer, offs, "debug_prop",
+				      dev->node_props.debug_prop);
 		sysfs_show_32bit_prop(buffer, offs, "sdma_fw_version",
-				      dev->gpu->sdma_fw_version);
+				      dev->gpu->kfd->sdma_fw_version);
 		sysfs_show_64bit_prop(buffer, offs, "unique_id",
+				      dev->gpu->xcp &&
+				      (dev->gpu->xcp->xcp_mgr->mode !=
+				       AMDGPU_SPX_PARTITION_MODE) ?
+				      dev->gpu->xcp->unique_id :
 				      dev->gpu->adev->unique_id);
-
+		sysfs_show_32bit_prop(buffer, offs, "num_xcc",
+				      NUM_XCC(dev->gpu->xcc_mask));
 	}
 
 	return sysfs_show_32bit_prop(buffer, offs, "max_engine_clk_ccompute",
@@ -972,44 +953,31 @@ static void kfd_update_system_properties(void)
 	dev = list_last_entry(&topology_device_list,
 			struct kfd_topology_device, list);
 	if (dev) {
-		sys_props.platform_id =
-			(*((uint64_t *)dev->oem_id)) & CRAT_OEMID_64BIT_MASK;
+		sys_props.platform_id = dev->oem_id64;
 		sys_props.platform_oem = *((uint64_t *)dev->oem_table_id);
 		sys_props.platform_rev = dev->oem_revision;
 	}
 	up_read(&topology_lock);
 }
 
-static void find_system_memory(const struct dmi_header *dm,
-	void *private)
+static void find_system_memory(const struct dmi_header *dm, void *private)
 {
+	struct dmi_mem_device *memdev = container_of(dm, struct dmi_mem_device, header);
 	struct kfd_mem_properties *mem;
-	u16 mem_width, mem_clock;
 	struct kfd_topology_device *kdev =
 		(struct kfd_topology_device *)private;
-	const u8 *dmi_data = (const u8 *)(dm + 1);
-
-	if (dm->type == DMI_ENTRY_MEM_DEVICE && dm->length >= 0x15) {
-		mem_width = (u16)(*(const u16 *)(dmi_data + 0x6));
-		mem_clock = (u16)(*(const u16 *)(dmi_data + 0x11));
-		list_for_each_entry(mem, &kdev->mem_props, list) {
-			if (mem_width != 0xFFFF && mem_width != 0)
-				mem->width = mem_width;
-			if (mem_clock != 0)
-				mem->mem_clk_max = mem_clock;
-		}
-	}
-}
 
-/*
- * Performance counters information is not part of CRAT but we would like to
- * put them in the sysfs under topology directory for Thunk to get the data.
- * This function is called before updating the sysfs.
- */
-static int kfd_add_perf_to_topology(struct kfd_topology_device *kdev)
-{
-	/* These are the only counters supported so far */
-	return kfd_iommu_add_perf_counters(kdev);
+	if (memdev->header.type != DMI_ENTRY_MEM_DEVICE)
+		return;
+	if (memdev->header.length < sizeof(struct dmi_mem_device))
+		return;
+
+	list_for_each_entry(mem, &kdev->mem_props, list) {
+		if (memdev->total_width != 0xFFFF && memdev->total_width != 0)
+			mem->width = memdev->total_width;
+		if (memdev->speed != 0)
+			mem->mem_clk_max = memdev->speed;
+	}
 }
 
 /* kfd_add_non_crat_information - Add information that is not currently
@@ -1026,25 +994,6 @@ static void kfd_add_non_crat_information(struct kfd_topology_device *kdev)
 	/* TODO: For GPU node, rearrange code from kfd_topology_add_device */
 }
 
-/* kfd_is_acpi_crat_invalid - CRAT from ACPI is valid only for AMD APU devices.
- *	Ignore CRAT for all other devices. AMD APU is identified if both CPU
- *	and GPU cores are present.
- * @device_list - topology device list created by parsing ACPI CRAT table.
- * @return - TRUE if invalid, FALSE is valid.
- */
-static bool kfd_is_acpi_crat_invalid(struct list_head *device_list)
-{
-	struct kfd_topology_device *dev;
-
-	list_for_each_entry(dev, device_list, list) {
-		if (dev->node_props.cpu_cores_count &&
-			dev->node_props.simd_count)
-			return false;
-	}
-	pr_info("Ignoring ACPI CRAT on non-APU system\n");
-	return true;
-}
-
 int kfd_topology_init(void)
 {
 	void *crat_image = NULL;
@@ -1075,48 +1024,25 @@ int kfd_topology_init(void)
 	 */
 	proximity_domain = 0;
 
-	/*
-	 * Get the CRAT image from the ACPI. If ACPI doesn't have one
-	 * or if ACPI CRAT is invalid create a virtual CRAT.
-	 * NOTE: The current implementation expects all AMD APUs to have
-	 *	CRAT. If no CRAT is available, it is assumed to be a CPU
-	 */
-	ret = kfd_create_crat_image_acpi(&crat_image, &image_size);
-	if (!ret) {
-		ret = kfd_parse_crat_table(crat_image,
-					   &temp_topology_device_list,
-					   proximity_domain);
-		if (ret ||
-		    kfd_is_acpi_crat_invalid(&temp_topology_device_list)) {
-			kfd_release_topology_device_list(
-				&temp_topology_device_list);
-			kfd_destroy_crat_image(crat_image);
-			crat_image = NULL;
-		}
+	ret = kfd_create_crat_image_virtual(&crat_image, &image_size,
+					    COMPUTE_UNIT_CPU, NULL,
+					    proximity_domain);
+	cpu_only_node = 1;
+	if (ret) {
+		pr_err("Error creating VCRAT table for CPU\n");
+		return ret;
 	}
 
-	if (!crat_image) {
-		ret = kfd_create_crat_image_virtual(&crat_image, &image_size,
-						    COMPUTE_UNIT_CPU, NULL,
-						    proximity_domain);
-		cpu_only_node = 1;
-		if (ret) {
-			pr_err("Error creating VCRAT table for CPU\n");
-			return ret;
-		}
-
-		ret = kfd_parse_crat_table(crat_image,
-					   &temp_topology_device_list,
-					   proximity_domain);
-		if (ret) {
-			pr_err("Error parsing VCRAT table for CPU\n");
-			goto err;
-		}
+	ret = kfd_parse_crat_table(crat_image,
+				   &temp_topology_device_list,
+				   proximity_domain);
+	if (ret) {
+		pr_err("Error parsing VCRAT table for CPU\n");
+		goto err;
 	}
 
 	kdev = list_first_entry(&temp_topology_device_list,
 				struct kfd_topology_device, list);
-	kfd_add_perf_to_topology(kdev);
 
 	down_write(&topology_lock);
 	kfd_topology_update_device_list(&temp_topology_device_list,
@@ -1157,16 +1083,19 @@ void kfd_topology_shutdown(void)
 	up_write(&topology_lock);
 }
 
-static uint32_t kfd_generate_gpu_id(struct kfd_dev *gpu)
+static uint32_t kfd_generate_gpu_id(struct kfd_node *gpu)
 {
-	uint32_t hashout;
-	uint32_t buf[7];
+	uint32_t gpu_id;
+	uint32_t buf[8];
 	uint64_t local_mem_size;
-	int i;
+	struct kfd_topology_device *dev;
+	bool is_unique;
+	uint8_t *crc_buf;
 
 	if (!gpu)
 		return 0;
 
+	crc_buf = (uint8_t *)&buf;
 	local_mem_size = gpu->local_mem_info.local_mem_size_private +
 			gpu->local_mem_info.local_mem_size_public;
 	buf[0] = gpu->adev->pdev->devfn;
@@ -1177,18 +1106,43 @@ static uint32_t kfd_generate_gpu_id(struct kfd_dev *gpu)
 	buf[4] = gpu->adev->pdev->bus->number;
 	buf[5] = lower_32_bits(local_mem_size);
 	buf[6] = upper_32_bits(local_mem_size);
+	buf[7] = (ffs(gpu->xcc_mask) - 1) | (NUM_XCC(gpu->xcc_mask) << 16);
 
-	for (i = 0, hashout = 0; i < 7; i++)
-		hashout ^= hash_32(buf[i], KFD_GPU_ID_HASH_WIDTH);
+	gpu_id = crc16(0, crc_buf, sizeof(buf)) &
+		 ((1 << KFD_GPU_ID_HASH_WIDTH) - 1);
 
-	return hashout;
+	/* There is a very small possibility when generating a
+	 * 16 (KFD_GPU_ID_HASH_WIDTH) bit value from 8 word buffer
+	 * that the value could be 0 or non-unique. So, check if
+	 * it is unique and non-zero. If not unique increment till
+	 * unique one is found. In case of overflow, restart from 1
+	 */
+
+	down_read(&topology_lock);
+	do {
+		is_unique = true;
+		if (!gpu_id)
+			gpu_id = 1;
+		list_for_each_entry(dev, &topology_device_list, list) {
+			if (dev->gpu && dev->gpu_id == gpu_id) {
+				is_unique = false;
+				break;
+			}
+		}
+		if (unlikely(!is_unique))
+			gpu_id = (gpu_id + 1) &
+				  ((1 << KFD_GPU_ID_HASH_WIDTH) - 1);
+	} while (!is_unique);
+	up_read(&topology_lock);
+
+	return gpu_id;
 }
 /* kfd_assign_gpu - Attach @gpu to the correct kfd topology device. If
  *		the GPU device is not already present in the topology device
  *		list then return NULL. This means a new topology device has to
  *		be created for this GPU.
  */
-static struct kfd_topology_device *kfd_assign_gpu(struct kfd_dev *gpu)
+static struct kfd_topology_device *kfd_assign_gpu(struct kfd_node *gpu)
 {
 	struct kfd_topology_device *dev;
 	struct kfd_topology_device *out_dev = NULL;
@@ -1201,8 +1155,7 @@ static struct kfd_topology_device *kfd_assign_gpu(struct kfd_dev *gpu)
 		/* Discrete GPUs need their own topology device list
 		 * entries. Don't assign them to CPU/APU nodes.
 		 */
-		if (!gpu->use_iommu_v2 &&
-		    dev->node_props.cpu_cores_count)
+		if (dev->node_props.cpu_cores_count)
 			continue;
 
 		if (!dev->gpu && (dev->node_props.simd_count > 0)) {
@@ -1248,7 +1201,8 @@ static void kfd_fill_mem_clk_max_info(struct kfd_topology_device *dev)
 	 * for APUs - If CRAT from ACPI reports more than one bank, then
 	 *	all the banks will report the same mem_clk_max information
 	 */
-	amdgpu_amdkfd_get_local_mem_info(dev->gpu->adev, &local_mem_info);
+	amdgpu_amdkfd_get_local_mem_info(dev->gpu->adev, &local_mem_info,
+					 dev->gpu->xcp);
 
 	list_for_each_entry(mem, &dev->mem_props, list)
 		mem->mem_clk_max = local_mem_info.mem_clk_max;
@@ -1275,7 +1229,7 @@ static void kfd_set_iolink_no_atomics(struct kfd_topology_device *dev,
 				CRAT_IOLINK_FLAGS_NO_ATOMICS_64_BIT;
 	/* set gpu (dev) flags. */
 	} else {
-		if (!dev->gpu->pci_atomic_requested ||
+		if (!dev->gpu->kfd->pci_atomic_requested ||
 				dev->gpu->adev->asic_type == CHIP_HAWAII)
 			link->flags |= CRAT_IOLINK_FLAGS_NO_ATOMICS_32_BIT |
 				CRAT_IOLINK_FLAGS_NO_ATOMICS_64_BIT;
@@ -1304,6 +1258,61 @@ static void kfd_set_iolink_non_coherent(struct kfd_topology_device *to_dev,
 	}
 }
 
+#define REC_SDMA_NUM_GPU	8
+static const int rec_sdma_eng_map[REC_SDMA_NUM_GPU][REC_SDMA_NUM_GPU] = {
+							{ -1, 14, 12, 2, 4, 8, 10, 6 },
+							{ 14, -1, 2, 10, 8, 4, 6, 12 },
+							{ 10, 2, -1, 12, 14, 6, 4, 8 },
+							{ 2, 12, 10, -1, 6, 14, 8, 4 },
+							{ 4, 8, 14, 6, -1, 10, 12, 2 },
+							{ 8, 4, 6, 14, 12, -1, 2, 10 },
+							{ 10, 6, 4, 8, 12, 2, -1, 14 },
+							{ 6, 12, 8, 4, 2, 10, 14, -1 }};
+
+static void kfd_set_recommended_sdma_engines(struct kfd_topology_device *to_dev,
+					     struct kfd_iolink_properties *outbound_link,
+					     struct kfd_iolink_properties *inbound_link)
+{
+	struct kfd_node *gpu = outbound_link->gpu;
+	struct amdgpu_device *adev = gpu->adev;
+	unsigned int num_xgmi_nodes = adev->gmc.xgmi.num_physical_nodes;
+	unsigned int num_xgmi_sdma_engines = kfd_get_num_xgmi_sdma_engines(gpu);
+	unsigned int num_sdma_engines = kfd_get_num_sdma_engines(gpu);
+	uint32_t sdma_eng_id_mask = (1 << num_sdma_engines) - 1;
+	uint32_t xgmi_sdma_eng_id_mask =
+			((1 << num_xgmi_sdma_engines) - 1) << num_sdma_engines;
+
+	bool support_rec_eng = !amdgpu_sriov_vf(adev) && to_dev->gpu &&
+		adev->aid_mask && num_xgmi_nodes && gpu->kfd->num_nodes == 1 &&
+		num_xgmi_sdma_engines >= 6 && (!(adev->flags & AMD_IS_APU) &&
+		num_xgmi_nodes == 8);
+
+	if (support_rec_eng) {
+		int src_socket_id = adev->gmc.xgmi.physical_node_id;
+		int dst_socket_id = to_dev->gpu->adev->gmc.xgmi.physical_node_id;
+		unsigned int reshift = num_xgmi_sdma_engines == 6 ? 1 : 0;
+
+		outbound_link->rec_sdma_eng_id_mask =
+			1 << (rec_sdma_eng_map[src_socket_id][dst_socket_id] >> reshift);
+		inbound_link->rec_sdma_eng_id_mask =
+			1 << (rec_sdma_eng_map[dst_socket_id][src_socket_id] >> reshift);
+
+		/* If recommended engine is out of range, need to reset the mask */
+		if (outbound_link->rec_sdma_eng_id_mask & sdma_eng_id_mask)
+			outbound_link->rec_sdma_eng_id_mask = xgmi_sdma_eng_id_mask;
+		if (inbound_link->rec_sdma_eng_id_mask & sdma_eng_id_mask)
+			inbound_link->rec_sdma_eng_id_mask = xgmi_sdma_eng_id_mask;
+
+	} else {
+		uint32_t engine_mask = (outbound_link->iolink_type == CRAT_IOLINK_TYPE_XGMI &&
+				num_xgmi_sdma_engines && to_dev->gpu) ? xgmi_sdma_eng_id_mask :
+				sdma_eng_id_mask;
+
+		outbound_link->rec_sdma_eng_id_mask = engine_mask;
+		inbound_link->rec_sdma_eng_id_mask = engine_mask;
+	}
+}
+
 static void kfd_fill_iolink_non_crat_info(struct kfd_topology_device *dev)
 {
 	struct kfd_iolink_properties *link, *inbound_link;
@@ -1323,10 +1332,16 @@ static void kfd_fill_iolink_non_crat_info(struct kfd_topology_device *dev)
 			continue;
 
 		/* Include the CPU peer in GPU hive if connected over xGMI. */
-		if (!peer_dev->gpu && !peer_dev->node_props.hive_id &&
-				dev->node_props.hive_id &&
-				dev->gpu->adev->gmc.xgmi.connected_to_cpu)
+		if (!peer_dev->gpu &&
+		    link->iolink_type == CRAT_IOLINK_TYPE_XGMI) {
+			/*
+			 * If the GPU is not part of a GPU hive, use its pci
+			 * device location as the hive ID to bind with the CPU.
+			 */
+			if (!dev->node_props.hive_id)
+				dev->node_props.hive_id = pci_dev_id(dev->gpu->adev->pdev);
 			peer_dev->node_props.hive_id = dev->node_props.hive_id;
+		}
 
 		list_for_each_entry(inbound_link, &peer_dev->io_link_props,
 									list) {
@@ -1336,6 +1351,7 @@ static void kfd_fill_iolink_non_crat_info(struct kfd_topology_device *dev)
 			inbound_link->flags = CRAT_IOLINK_FLAGS_ENABLED;
 			kfd_set_iolink_no_atomics(peer_dev, dev, inbound_link);
 			kfd_set_iolink_non_coherent(peer_dev, link, inbound_link);
+			kfd_set_recommended_sdma_engines(peer_dev, link, inbound_link);
 		}
 	}
 
@@ -1402,10 +1418,11 @@ static int kfd_create_indirect_link_prop(struct kfd_topology_device *kdev, int g
 		num_cpu++;
 	}
 
+	if (list_empty(&kdev->io_link_props))
+		return -ENODATA;
+
 	gpu_link = list_first_entry(&kdev->io_link_props,
-					struct kfd_iolink_properties, list);
-	if (!gpu_link)
-		return -ENOMEM;
+				    struct kfd_iolink_properties, list);
 
 	for (i = 0; i < num_cpu; i++) {
 		/* CPU <--> GPU */
@@ -1483,15 +1500,17 @@ static int kfd_add_peer_prop(struct kfd_topology_device *kdev,
 				peer->gpu->adev))
 		return ret;
 
+	if (list_empty(&kdev->io_link_props))
+		return -ENODATA;
+
 	iolink1 = list_first_entry(&kdev->io_link_props,
-							struct kfd_iolink_properties, list);
-	if (!iolink1)
-		return -ENOMEM;
+				   struct kfd_iolink_properties, list);
+
+	if (list_empty(&peer->io_link_props))
+		return -ENODATA;
 
 	iolink2 = list_first_entry(&peer->io_link_props,
-							struct kfd_iolink_properties, list);
-	if (!iolink2)
-		return -ENOMEM;
+				   struct kfd_iolink_properties, list);
 
 	props = kfd_alloc_struct(props);
 	if (!props)
@@ -1509,17 +1528,19 @@ static int kfd_add_peer_prop(struct kfd_topology_device *kdev,
 		/* CPU->CPU  link*/
 		cpu_dev = kfd_topology_device_by_proximity_domain(iolink1->node_to);
 		if (cpu_dev) {
-			list_for_each_entry(iolink3, &cpu_dev->io_link_props, list)
-				if (iolink3->node_to == iolink2->node_to)
-					break;
-
-			props->weight += iolink3->weight;
-			props->min_latency += iolink3->min_latency;
-			props->max_latency += iolink3->max_latency;
-			props->min_bandwidth = min(props->min_bandwidth,
-							iolink3->min_bandwidth);
-			props->max_bandwidth = min(props->max_bandwidth,
-							iolink3->max_bandwidth);
+			list_for_each_entry(iolink3, &cpu_dev->io_link_props, list) {
+				if (iolink3->node_to != iolink2->node_to)
+					continue;
+
+				props->weight += iolink3->weight;
+				props->min_latency += iolink3->min_latency;
+				props->max_latency += iolink3->max_latency;
+				props->min_bandwidth = min(props->min_bandwidth,
+							   iolink3->min_bandwidth);
+				props->max_bandwidth = min(props->max_bandwidth,
+							   iolink3->max_bandwidth);
+				break;
+			}
 		} else {
 			WARN(1, "CPU node not found");
 		}
@@ -1569,8 +1590,9 @@ static int kfd_dev_create_p2p_links(void)
 		if (dev == new_dev)
 			break;
 		if (!dev->gpu || !dev->gpu->adev ||
-		    (dev->gpu->hive_id &&
-		     dev->gpu->hive_id == new_dev->gpu->hive_id))
+		    (dev->gpu->kfd->hive_id &&
+		     dev->gpu->kfd->hive_id == new_dev->gpu->kfd->hive_id &&
+		     amdgpu_xgmi_get_is_sharing_enabled(dev->gpu->adev, new_dev->gpu->adev)))
 			goto next;
 
 		/* check if node(s) is/are peer accessible in one direction or bi-direction */
@@ -1590,11 +1612,9 @@ out:
 	return ret;
 }
 
-
 /* Helper function. See kfd_fill_gpu_cache_info for parameter description */
 static int fill_in_l1_pcache(struct kfd_cache_properties **props_ext,
 				struct kfd_gpu_cache_info *pcache_info,
-				struct kfd_cu_info *cu_info,
 				int cu_bitmask,
 				int cache_type, unsigned int cu_processor_id,
 				int cu_block)
@@ -1621,6 +1641,7 @@ static int fill_in_l1_pcache(struct kfd_cache_properties **props_ext,
 		pcache->processor_id_low = cu_processor_id + (first_active_cu - 1);
 		pcache->cache_level = pcache_info[cache_type].cache_level;
 		pcache->cache_size = pcache_info[cache_type].cache_size;
+		pcache->cacheline_size = pcache_info[cache_type].cache_line_size;
 
 		if (pcache_info[cache_type].flags & CRAT_CACHE_FLAGS_DATA_CACHE)
 			pcache->cache_type |= HSA_CACHE_TYPE_DATA;
@@ -1656,15 +1677,37 @@ static int fill_in_l1_pcache(struct kfd_cache_properties **props_ext,
 /* Helper function. See kfd_fill_gpu_cache_info for parameter description */
 static int fill_in_l2_l3_pcache(struct kfd_cache_properties **props_ext,
 				struct kfd_gpu_cache_info *pcache_info,
-				struct kfd_cu_info *cu_info,
-				int cache_type, unsigned int cu_processor_id)
+				struct amdgpu_cu_info *cu_info,
+				struct amdgpu_gfx_config *gfx_info,
+				int cache_type, unsigned int cu_processor_id,
+				struct kfd_node *knode)
 {
-	unsigned int cu_sibling_map_mask;
+	unsigned int cu_sibling_map_mask = 0;
 	int first_active_cu;
-	int i, j, k;
+	int i, j, k, xcc, start, end;
+	int num_xcc = NUM_XCC(knode->xcc_mask);
 	struct kfd_cache_properties *pcache = NULL;
+	enum amdgpu_memory_partition mode;
+	struct amdgpu_device *adev = knode->adev;
+	bool found = false;
+
+	start = ffs(knode->xcc_mask) - 1;
+	end = start + num_xcc;
+
+	/* To find the bitmap in the first active cu in the first
+	 * xcc, it is based on the assumption that evrey xcc must
+	 * have at least one active cu.
+	 */
+	for (i = 0; i < gfx_info->max_shader_engines && !found; i++) {
+		for (j = 0; j < gfx_info->max_sh_per_se && !found; j++) {
+			if (cu_info->bitmap[start][i % 4][j % 4]) {
+				cu_sibling_map_mask =
+					cu_info->bitmap[start][i % 4][j % 4];
+				found = true;
+			}
+		}
+	}
 
-	cu_sibling_map_mask = cu_info->cu_bitmap[0][0];
 	cu_sibling_map_mask &=
 		((1 << pcache_info[cache_type].num_cu_shared) - 1);
 	first_active_cu = ffs(cu_sibling_map_mask);
@@ -1682,7 +1725,19 @@ static int fill_in_l2_l3_pcache(struct kfd_cache_properties **props_ext,
 		pcache->processor_id_low = cu_processor_id
 					+ (first_active_cu - 1);
 		pcache->cache_level = pcache_info[cache_type].cache_level;
+		pcache->cacheline_size = pcache_info[cache_type].cache_line_size;
+
+		if (KFD_GC_VERSION(knode) == IP_VERSION(9, 4, 3) ||
+		    KFD_GC_VERSION(knode) == IP_VERSION(9, 4, 4) ||
+		    KFD_GC_VERSION(knode) == IP_VERSION(9, 5, 0))
+			mode = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
+		else
+			mode = UNKNOWN_MEMORY_PARTITION_MODE;
+
 		pcache->cache_size = pcache_info[cache_type].cache_size;
+		/* Partition mode only affects L3 cache size */
+		if (mode && pcache->cache_level == 3)
+			pcache->cache_size /= mode;
 
 		if (pcache_info[cache_type].flags & CRAT_CACHE_FLAGS_DATA_CACHE)
 			pcache->cache_type |= HSA_CACHE_TYPE_DATA;
@@ -1699,16 +1754,18 @@ static int fill_in_l2_l3_pcache(struct kfd_cache_properties **props_ext,
 		cu_sibling_map_mask = cu_sibling_map_mask >> (first_active_cu - 1);
 		k = 0;
 
-		for (i = 0; i < cu_info->num_shader_engines; i++) {
-			for (j = 0; j < cu_info->num_shader_arrays_per_engine; j++) {
-				pcache->sibling_map[k] = (uint8_t)(cu_sibling_map_mask & 0xFF);
-				pcache->sibling_map[k+1] = (uint8_t)((cu_sibling_map_mask >> 8) & 0xFF);
-				pcache->sibling_map[k+2] = (uint8_t)((cu_sibling_map_mask >> 16) & 0xFF);
-				pcache->sibling_map[k+3] = (uint8_t)((cu_sibling_map_mask >> 24) & 0xFF);
-				k += 4;
-
-				cu_sibling_map_mask = cu_info->cu_bitmap[i % 4][j + i / 4];
-				cu_sibling_map_mask &= ((1 << pcache_info[cache_type].num_cu_shared) - 1);
+		for (xcc = start; xcc < end; xcc++) {
+			for (i = 0; i < gfx_info->max_shader_engines; i++) {
+				for (j = 0; j < gfx_info->max_sh_per_se; j++) {
+					pcache->sibling_map[k] = (uint8_t)(cu_sibling_map_mask & 0xFF);
+					pcache->sibling_map[k+1] = (uint8_t)((cu_sibling_map_mask >> 8) & 0xFF);
+					pcache->sibling_map[k+2] = (uint8_t)((cu_sibling_map_mask >> 16) & 0xFF);
+					pcache->sibling_map[k+3] = (uint8_t)((cu_sibling_map_mask >> 24) & 0xFF);
+					k += 4;
+
+					cu_sibling_map_mask = cu_info->bitmap[xcc][i % 4][j + i / 4];
+					cu_sibling_map_mask &= ((1 << pcache_info[cache_type].num_cu_shared) - 1);
+				}
 			}
 		}
 		pcache->sibling_map_size = k;
@@ -1723,27 +1780,26 @@ static int fill_in_l2_l3_pcache(struct kfd_cache_properties **props_ext,
 /* kfd_fill_cache_non_crat_info - Fill GPU cache info using kfd_gpu_cache_info
  * tables
  */
-static void kfd_fill_cache_non_crat_info(struct kfd_topology_device *dev, struct kfd_dev *kdev)
+static void kfd_fill_cache_non_crat_info(struct kfd_topology_device *dev, struct kfd_node *kdev)
 {
 	struct kfd_gpu_cache_info *pcache_info = NULL;
-	int i, j, k;
+	int i, j, k, xcc, start, end;
 	int ct = 0;
 	unsigned int cu_processor_id;
 	int ret;
 	unsigned int num_cu_shared;
-	struct kfd_cu_info cu_info;
-	struct kfd_cu_info *pcu_info;
+	struct amdgpu_cu_info *cu_info = &kdev->adev->gfx.cu_info;
+	struct amdgpu_gfx_config *gfx_info = &kdev->adev->gfx.config;
 	int gpu_processor_id;
-	struct kfd_cache_properties *props_ext;
+	struct kfd_cache_properties *props_ext = NULL;
 	int num_of_entries = 0;
 	int num_of_cache_types = 0;
 	struct kfd_gpu_cache_info cache_info[KFD_MAX_CACHE_TYPES];
 
-	amdgpu_amdkfd_get_cu_info(kdev->adev, &cu_info);
-	pcu_info = &cu_info;
 
 	gpu_processor_id = dev->node_props.simd_id_base;
 
+	memset(cache_info, 0, sizeof(cache_info));
 	pcache_info = cache_info;
 	num_of_cache_types = kfd_get_gpu_cache_info(kdev, &pcache_info);
 	if (!num_of_cache_types) {
@@ -1760,37 +1816,42 @@ static void kfd_fill_cache_non_crat_info(struct kfd_topology_device *dev, struct
 	 *			then it will consider only one CU from
 	 *			the shared unit
 	 */
+	start = ffs(kdev->xcc_mask) - 1;
+	end = start + NUM_XCC(kdev->xcc_mask);
+
 	for (ct = 0; ct < num_of_cache_types; ct++) {
 		cu_processor_id = gpu_processor_id;
 		if (pcache_info[ct].cache_level == 1) {
-			for (i = 0; i < pcu_info->num_shader_engines; i++) {
-				for (j = 0; j < pcu_info->num_shader_arrays_per_engine; j++) {
-					for (k = 0; k < pcu_info->num_cu_per_sh; k += pcache_info[ct].num_cu_shared) {
+			for (xcc = start; xcc < end; xcc++) {
+				for (i = 0; i < gfx_info->max_shader_engines; i++) {
+					for (j = 0; j < gfx_info->max_sh_per_se; j++) {
+						for (k = 0; k < gfx_info->max_cu_per_sh; k += pcache_info[ct].num_cu_shared) {
 
-						ret = fill_in_l1_pcache(&props_ext, pcache_info, pcu_info,
-										pcu_info->cu_bitmap[i % 4][j + i / 4], ct,
+							ret = fill_in_l1_pcache(&props_ext, pcache_info,
+										cu_info->bitmap[xcc][i % 4][j + i / 4], ct,
 										cu_processor_id, k);
 
-						if (ret < 0)
-							break;
+							if (ret < 0)
+								break;
 
-						if (!ret) {
-							num_of_entries++;
-							list_add_tail(&props_ext->list, &dev->cache_props);
-						}
+							if (!ret) {
+								num_of_entries++;
+								list_add_tail(&props_ext->list, &dev->cache_props);
+							}
 
-						/* Move to next CU block */
-						num_cu_shared = ((k + pcache_info[ct].num_cu_shared) <=
-							pcu_info->num_cu_per_sh) ?
-							pcache_info[ct].num_cu_shared :
-							(pcu_info->num_cu_per_sh - k);
-						cu_processor_id += num_cu_shared;
+							/* Move to next CU block */
+							num_cu_shared = ((k + pcache_info[ct].num_cu_shared) <=
+								gfx_info->max_cu_per_sh) ?
+								pcache_info[ct].num_cu_shared :
+								(gfx_info->max_cu_per_sh - k);
+							cu_processor_id += num_cu_shared;
+						}
 					}
 				}
 			}
 		} else {
 			ret = fill_in_l2_l3_pcache(&props_ext, pcache_info,
-								pcu_info, ct, cu_processor_id);
+						   cu_info, gfx_info, ct, cu_processor_id, kdev);
 
 			if (ret < 0)
 				break;
@@ -1805,7 +1866,7 @@ static void kfd_fill_cache_non_crat_info(struct kfd_topology_device *dev, struct
 	pr_debug("Added [%d] GPU cache entries\n", num_of_entries);
 }
 
-static int kfd_topology_add_device_locked(struct kfd_dev *gpu, uint32_t gpu_id,
+static int kfd_topology_add_device_locked(struct kfd_node *gpu,
 					  struct kfd_topology_device **dev)
 {
 	int proximity_domain = ++topology_crat_proximity_domain;
@@ -1818,8 +1879,7 @@ static int kfd_topology_add_device_locked(struct kfd_dev *gpu, uint32_t gpu_id,
 					    COMPUTE_UNIT_GPU, gpu,
 					    proximity_domain);
 	if (res) {
-		pr_err("Error creating VCRAT for GPU (ID: 0x%x)\n",
-		       gpu_id);
+		dev_err(gpu->adev->dev, "Error creating VCRAT\n");
 		topology_crat_proximity_domain--;
 		goto err;
 	}
@@ -1830,8 +1890,7 @@ static int kfd_topology_add_device_locked(struct kfd_dev *gpu, uint32_t gpu_id,
 				   &temp_topology_device_list,
 				   proximity_domain);
 	if (res) {
-		pr_err("Error parsing VCRAT for GPU (ID: 0x%x)\n",
-		       gpu_id);
+		dev_err(gpu->adev->dev, "Error parsing VCRAT\n");
 		topology_crat_proximity_domain--;
 		goto err;
 	}
@@ -1857,25 +1916,136 @@ static int kfd_topology_add_device_locked(struct kfd_dev *gpu, uint32_t gpu_id,
 	if (!res)
 		sys_props.generation_count++;
 	else
-		pr_err("Failed to update GPU (ID: 0x%x) to sysfs topology. res=%d\n",
-		       gpu_id, res);
+		dev_err(gpu->adev->dev, "Failed to update GPU to sysfs topology. res=%d\n",
+			res);
 
 err:
 	kfd_destroy_crat_image(crat_image);
 	return res;
 }
 
-int kfd_topology_add_device(struct kfd_dev *gpu)
+static void kfd_topology_set_dbg_firmware_support(struct kfd_topology_device *dev)
+{
+	bool firmware_supported = true;
+
+	if (KFD_GC_VERSION(dev->gpu) >= IP_VERSION(11, 0, 0) &&
+			KFD_GC_VERSION(dev->gpu) < IP_VERSION(12, 0, 0)) {
+		uint32_t mes_api_rev = (dev->gpu->adev->mes.sched_version &
+						AMDGPU_MES_API_VERSION_MASK) >>
+						AMDGPU_MES_API_VERSION_SHIFT;
+		uint32_t mes_rev = dev->gpu->adev->mes.sched_version &
+						AMDGPU_MES_VERSION_MASK;
+
+		firmware_supported = (mes_api_rev >= 14) && (mes_rev >= 64);
+		goto out;
+	}
+
+	/*
+	 * Note: Any unlisted devices here are assumed to support exception handling.
+	 * Add additional checks here as needed.
+	 */
+	switch (KFD_GC_VERSION(dev->gpu)) {
+	case IP_VERSION(9, 0, 1):
+		firmware_supported = dev->gpu->kfd->mec_fw_version >= 459 + 32768;
+		break;
+	case IP_VERSION(9, 1, 0):
+	case IP_VERSION(9, 2, 1):
+	case IP_VERSION(9, 2, 2):
+	case IP_VERSION(9, 3, 0):
+	case IP_VERSION(9, 4, 0):
+		firmware_supported = dev->gpu->kfd->mec_fw_version >= 459;
+		break;
+	case IP_VERSION(9, 4, 1):
+		firmware_supported = dev->gpu->kfd->mec_fw_version >= 60;
+		break;
+	case IP_VERSION(9, 4, 2):
+		firmware_supported = dev->gpu->kfd->mec_fw_version >= 51;
+		break;
+	case IP_VERSION(10, 1, 10):
+	case IP_VERSION(10, 1, 2):
+	case IP_VERSION(10, 1, 1):
+		firmware_supported = dev->gpu->kfd->mec_fw_version >= 144;
+		break;
+	case IP_VERSION(10, 3, 0):
+	case IP_VERSION(10, 3, 2):
+	case IP_VERSION(10, 3, 1):
+	case IP_VERSION(10, 3, 4):
+	case IP_VERSION(10, 3, 5):
+		firmware_supported = dev->gpu->kfd->mec_fw_version >= 89;
+		break;
+	case IP_VERSION(10, 1, 3):
+	case IP_VERSION(10, 3, 3):
+		firmware_supported = false;
+		break;
+	default:
+		break;
+	}
+
+out:
+	if (firmware_supported)
+		dev->node_props.capability |= HSA_CAP_TRAP_DEBUG_FIRMWARE_SUPPORTED;
+}
+
+static void kfd_topology_set_capabilities(struct kfd_topology_device *dev)
+{
+	dev->node_props.capability |= ((HSA_CAP_DOORBELL_TYPE_2_0 <<
+				HSA_CAP_DOORBELL_TYPE_TOTALBITS_SHIFT) &
+				HSA_CAP_DOORBELL_TYPE_TOTALBITS_MASK);
+
+	dev->node_props.capability |= HSA_CAP_TRAP_DEBUG_SUPPORT |
+			HSA_CAP_TRAP_DEBUG_WAVE_LAUNCH_TRAP_OVERRIDE_SUPPORTED |
+			HSA_CAP_TRAP_DEBUG_WAVE_LAUNCH_MODE_SUPPORTED;
+
+	if (kfd_dbg_has_ttmps_always_setup(dev->gpu))
+		dev->node_props.debug_prop |= HSA_DBG_DISPATCH_INFO_ALWAYS_VALID;
+
+	if (KFD_GC_VERSION(dev->gpu) < IP_VERSION(10, 0, 0)) {
+		if (KFD_GC_VERSION(dev->gpu) == IP_VERSION(9, 4, 3) ||
+		    KFD_GC_VERSION(dev->gpu) == IP_VERSION(9, 4, 4))
+			dev->node_props.debug_prop |=
+				HSA_DBG_WATCH_ADDR_MASK_LO_BIT_GFX9_4_3 |
+				HSA_DBG_WATCH_ADDR_MASK_HI_BIT_GFX9_4_3;
+		else
+			dev->node_props.debug_prop |=
+				HSA_DBG_WATCH_ADDR_MASK_LO_BIT_GFX9 |
+				HSA_DBG_WATCH_ADDR_MASK_HI_BIT;
+
+		if (KFD_GC_VERSION(dev->gpu) >= IP_VERSION(9, 4, 2))
+			dev->node_props.capability |=
+				HSA_CAP_TRAP_DEBUG_PRECISE_MEMORY_OPERATIONS_SUPPORTED;
+
+		if (!amdgpu_sriov_vf(dev->gpu->adev))
+			dev->node_props.capability |= HSA_CAP_PER_QUEUE_RESET_SUPPORTED;
+
+	} else {
+		dev->node_props.debug_prop |= HSA_DBG_WATCH_ADDR_MASK_LO_BIT_GFX10 |
+					HSA_DBG_WATCH_ADDR_MASK_HI_BIT;
+
+		if (KFD_GC_VERSION(dev->gpu) >= IP_VERSION(12, 0, 0))
+			dev->node_props.capability |=
+				HSA_CAP_TRAP_DEBUG_PRECISE_ALU_OPERATIONS_SUPPORTED;
+	}
+
+	kfd_topology_set_dbg_firmware_support(dev);
+}
+
+int kfd_topology_add_device(struct kfd_node *gpu)
 {
 	uint32_t gpu_id;
 	struct kfd_topology_device *dev;
-	struct kfd_cu_info cu_info;
 	int res = 0;
 	int i;
 	const char *asic_name = amdgpu_asic_name[gpu->adev->asic_type];
+	struct amdgpu_gfx_config *gfx_info = &gpu->adev->gfx.config;
+	struct amdgpu_cu_info *cu_info = &gpu->adev->gfx.cu_info;
 
-	gpu_id = kfd_generate_gpu_id(gpu);
-	pr_debug("Adding new GPU (ID: 0x%x) to topology\n", gpu_id);
+	if (gpu->xcp && !gpu->xcp->ddev) {
+		dev_warn(gpu->adev->dev,
+			 "Won't add GPU to topology since it has no drm node assigned.");
+		return 0;
+	} else {
+		dev_dbg(gpu->adev->dev, "Adding new GPU to topology\n");
+	}
 
 	/* Check to see if this gpu device exists in the topology_device_list.
 	 * If so, assign the gpu to that device,
@@ -1886,11 +2056,12 @@ int kfd_topology_add_device(struct kfd_dev *gpu)
 	down_write(&topology_lock);
 	dev = kfd_assign_gpu(gpu);
 	if (!dev)
-		res = kfd_topology_add_device_locked(gpu, gpu_id, &dev);
+		res = kfd_topology_add_device_locked(gpu, &dev);
 	up_write(&topology_lock);
 	if (res)
 		return res;
 
+	gpu_id = kfd_generate_gpu_id(gpu);
 	dev->gpu_id = gpu_id;
 	gpu->id = gpu_id;
 
@@ -1903,9 +2074,6 @@ int kfd_topology_add_device(struct kfd_dev *gpu)
 	/* Fill-in additional information that is not available in CRAT but
 	 * needed for the topology
 	 */
-
-	amdgpu_amdkfd_get_cu_info(dev->gpu->adev, &cu_info);
-
 	for (i = 0; i < KFD_TOPOLOGY_PUBLIC_NAME_SIZE-1; i++) {
 		dev->node_props.name[i] = __tolower(asic_name[i]);
 		if (asic_name[i] == '\0')
@@ -1914,30 +2082,39 @@ int kfd_topology_add_device(struct kfd_dev *gpu)
 	dev->node_props.name[i] = '\0';
 
 	dev->node_props.simd_arrays_per_engine =
-		cu_info.num_shader_arrays_per_engine;
+		gfx_info->max_sh_per_se;
 
-	dev->node_props.gfx_target_version = gpu->device_info.gfx_target_version;
+	dev->node_props.gfx_target_version =
+				gpu->kfd->device_info.gfx_target_version;
 	dev->node_props.vendor_id = gpu->adev->pdev->vendor;
 	dev->node_props.device_id = gpu->adev->pdev->device;
 	dev->node_props.capability |=
 		((dev->gpu->adev->rev_id << HSA_CAP_ASIC_REVISION_SHIFT) &
 			HSA_CAP_ASIC_REVISION_MASK);
+
 	dev->node_props.location_id = pci_dev_id(gpu->adev->pdev);
+	if (gpu->kfd->num_nodes > 1)
+		dev->node_props.location_id |= dev->gpu->node_id;
+
 	dev->node_props.domain = pci_domain_nr(gpu->adev->pdev->bus);
 	dev->node_props.max_engine_clk_fcompute =
 		amdgpu_amdkfd_get_max_engine_clock_in_mhz(dev->gpu->adev);
 	dev->node_props.max_engine_clk_ccompute =
 		cpufreq_quick_get_max(0) / 1000;
-	dev->node_props.drm_render_minor =
-		gpu->shared_resources.drm_render_minor;
 
-	dev->node_props.hive_id = gpu->hive_id;
+	if (gpu->xcp)
+		dev->node_props.drm_render_minor = gpu->xcp->ddev->render->index;
+	else
+		dev->node_props.drm_render_minor =
+				gpu->kfd->shared_resources.drm_render_minor;
+
+	dev->node_props.hive_id = gpu->kfd->hive_id;
 	dev->node_props.num_sdma_engines = kfd_get_num_sdma_engines(gpu);
 	dev->node_props.num_sdma_xgmi_engines =
 					kfd_get_num_xgmi_sdma_engines(gpu);
 	dev->node_props.num_sdma_queues_per_engine =
-				gpu->device_info.num_sdma_queues_per_engine -
-				gpu->device_info.num_reserved_sdma_queues_per_engine;
+				gpu->kfd->device_info.num_sdma_queues_per_engine -
+				gpu->kfd->device_info.num_reserved_sdma_queues_per_engine;
 	dev->node_props.num_gws = (dev->gpu->gws &&
 		dev->gpu->dqm->sched_policy != KFD_SCHED_POLICY_NO_HWS) ?
 		dev->gpu->adev->gds.gws_size : 0;
@@ -1966,23 +2143,18 @@ int kfd_topology_add_device(struct kfd_dev *gpu)
 			HSA_CAP_DOORBELL_TYPE_TOTALBITS_MASK);
 		break;
 	default:
-		if (KFD_GC_VERSION(dev->gpu) >= IP_VERSION(9, 0, 1))
-			dev->node_props.capability |= ((HSA_CAP_DOORBELL_TYPE_2_0 <<
-				HSA_CAP_DOORBELL_TYPE_TOTALBITS_SHIFT) &
-				HSA_CAP_DOORBELL_TYPE_TOTALBITS_MASK);
-		else
+		if (KFD_GC_VERSION(dev->gpu) < IP_VERSION(9, 0, 1))
 			WARN(1, "Unexpected ASIC family %u",
 			     dev->gpu->adev->asic_type);
+		else
+			kfd_topology_set_capabilities(dev);
 	}
 
 	/*
 	 * Overwrite ATS capability according to needs_iommu_device to fix
 	 * potential missing corresponding bit in CRAT of BIOS.
 	 */
-	if (dev->gpu->use_iommu_v2)
-		dev->node_props.capability |= HSA_CAP_ATS_PRESENT;
-	else
-		dev->node_props.capability &= ~HSA_CAP_ATS_PRESENT;
+	dev->node_props.capability &= ~HSA_CAP_ATS_PRESENT;
 
 	/* Fix errors in CZ CRAT.
 	 * simd_count: Carrizo CRAT reports wrong simd_count, probably
@@ -1991,7 +2163,7 @@ int kfd_topology_add_device(struct kfd_dev *gpu)
 	 */
 	if (dev->gpu->adev->asic_type == CHIP_CARRIZO) {
 		dev->node_props.simd_count =
-			cu_info.simd_per_cu * cu_info.cu_active_number;
+			cu_info->simd_per_cu * cu_info->number;
 		dev->node_props.max_waves_per_simd = 10;
 	}
 
@@ -2007,9 +2179,15 @@ int kfd_topology_add_device(struct kfd_dev *gpu)
 		dev->node_props.capability |= (dev->gpu->adev->ras_enabled != 0) ?
 			HSA_CAP_RASEVENTNOTIFY : 0;
 
-	if (KFD_IS_SVM_API_SUPPORTED(dev->gpu->adev->kfd.dev))
+	if (KFD_IS_SVM_API_SUPPORTED(dev->gpu->adev))
 		dev->node_props.capability |= HSA_CAP_SVMAPI_SUPPORTED;
 
+	if (dev->gpu->adev->gmc.is_app_apu ||
+		dev->gpu->adev->gmc.xgmi.connected_to_cpu)
+		dev->node_props.capability |= HSA_CAP_FLAGS_COHERENTHOSTACCESS;
+
+	kfd_queue_ctx_save_restore_size(dev);
+
 	kfd_debug_print_topology();
 
 	kfd_notify_gpu_change(gpu_id, 1);
@@ -2079,7 +2257,7 @@ static void kfd_topology_update_io_links(int proximity_domain)
 	}
 }
 
-int kfd_topology_remove_device(struct kfd_dev *gpu)
+int kfd_topology_remove_device(struct kfd_node *gpu)
 {
 	struct kfd_topology_device *dev, *tmp;
 	uint32_t gpu_id;
@@ -2119,7 +2297,7 @@ int kfd_topology_remove_device(struct kfd_dev *gpu)
  * Return -	0: On success (@kdev will be NULL for non GPU nodes)
  *		-1: If end of list
  */
-int kfd_topology_enum_kfd_devices(uint8_t idx, struct kfd_dev **kdev)
+int kfd_topology_enum_kfd_devices(uint8_t idx, struct kfd_node **kdev)
 {
 
 	struct kfd_topology_device *top_dev;
@@ -2154,7 +2332,7 @@ static int kfd_cpumask_to_apic_id(const struct cpumask *cpumask)
 	if (first_cpu_of_numa_node >= nr_cpu_ids)
 		return -1;
 #ifdef CONFIG_X86_64
-	return cpu_data(first_cpu_of_numa_node).apicid;
+	return cpu_data(first_cpu_of_numa_node).topo.apicid;
 #else
 	return first_cpu_of_numa_node;
 #endif
@@ -2173,29 +2351,6 @@ int kfd_numa_node_to_apic_id(int numa_node_id)
 	return kfd_cpumask_to_apic_id(cpumask_of_node(numa_node_id));
 }
 
-void kfd_double_confirm_iommu_support(struct kfd_dev *gpu)
-{
-	struct kfd_topology_device *dev;
-
-	gpu->use_iommu_v2 = false;
-
-	if (!gpu->device_info.needs_iommu_device)
-		return;
-
-	down_read(&topology_lock);
-
-	/* Only use IOMMUv2 if there is an APU topology node with no GPU
-	 * assigned yet. This GPU will be assigned to it.
-	 */
-	list_for_each_entry(dev, &topology_device_list, list)
-		if (dev->node_props.cpu_cores_count &&
-		    dev->node_props.simd_count &&
-		    !dev->gpu)
-			gpu->use_iommu_v2 = true;
-
-	up_read(&topology_lock);
-}
-
 #if defined(CONFIG_DEBUG_FS)
 
 int kfd_debugfs_hqds_by_device(struct seq_file *m, void *data)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.h b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h
index fca30d00a9bb..3de8ec0043bb 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h
@@ -24,6 +24,7 @@
 #ifndef __KFD_TOPOLOGY_H__
 #define __KFD_TOPOLOGY_H__
 
+#include <linux/dmi.h>
 #include <linux/types.h>
 #include <linux/list.h>
 #include <linux/kfd_sysfs.h>
@@ -31,6 +32,14 @@
 
 #define KFD_TOPOLOGY_PUBLIC_NAME_SIZE 32
 
+#define HSA_DBG_WATCH_ADDR_MASK_LO_BIT_GFX9	6
+#define HSA_DBG_WATCH_ADDR_MASK_LO_BIT_GFX9_4_3 7
+#define HSA_DBG_WATCH_ADDR_MASK_LO_BIT_GFX10	7
+#define HSA_DBG_WATCH_ADDR_MASK_HI_BIT  \
+			(29 << HSA_DBG_WATCH_ADDR_MASK_HI_BIT_SHIFT)
+#define HSA_DBG_WATCH_ADDR_MASK_HI_BIT_GFX9_4_3 \
+			(30 << HSA_DBG_WATCH_ADDR_MASK_HI_BIT_SHIFT)
+
 struct kfd_node_properties {
 	uint64_t hive_id;
 	uint32_t cpu_cores_count;
@@ -42,6 +51,8 @@ struct kfd_node_properties {
 	uint32_t cpu_core_id_base;
 	uint32_t simd_id_base;
 	uint32_t capability;
+	uint32_t capability2;
+	uint64_t debug_prop;
 	uint32_t max_waves_per_simd;
 	uint32_t lds_size_in_kb;
 	uint32_t gds_size_in_kb;
@@ -65,6 +76,10 @@ struct kfd_node_properties {
 	uint32_t num_sdma_xgmi_engines;
 	uint32_t num_sdma_queues_per_engine;
 	uint32_t num_cp_queues;
+	uint32_t cwsr_size;
+	uint32_t ctl_stack_size;
+	uint32_t eop_buffer_size;
+	uint32_t debug_memory_size;
 	char name[KFD_TOPOLOGY_PUBLIC_NAME_SIZE];
 };
 
@@ -75,12 +90,12 @@ struct kfd_mem_properties {
 	uint32_t		flags;
 	uint32_t		width;
 	uint32_t		mem_clk_max;
-	struct kfd_dev		*gpu;
+	struct kfd_node		*gpu;
 	struct kobject		*kobj;
 	struct attribute	attr;
 };
 
-#define CACHE_SIBLINGMAP_SIZE 64
+#define CACHE_SIBLINGMAP_SIZE 128
 
 struct kfd_cache_properties {
 	struct list_head	list;
@@ -93,7 +108,7 @@ struct kfd_cache_properties {
 	uint32_t		cache_latency;
 	uint32_t		cache_type;
 	uint8_t			sibling_map[CACHE_SIBLINGMAP_SIZE];
-	struct kfd_dev		*gpu;
+	struct kfd_node		*gpu;
 	struct kobject		*kobj;
 	struct attribute	attr;
 	uint32_t		sibling_map_size;
@@ -112,8 +127,9 @@ struct kfd_iolink_properties {
 	uint32_t		min_bandwidth;
 	uint32_t		max_bandwidth;
 	uint32_t		rec_transfer_size;
+	uint32_t		rec_sdma_eng_id_mask;
 	uint32_t		flags;
-	struct kfd_dev		*gpu;
+	struct kfd_node		*gpu;
 	struct kobject		*kobj;
 	struct attribute	attr;
 };
@@ -135,7 +151,7 @@ struct kfd_topology_device {
 	struct list_head		io_link_props;
 	struct list_head		p2p_link_props;
 	struct list_head		perf_props;
-	struct kfd_dev			*gpu;
+	struct kfd_node			*gpu;
 	struct kobject			*kobj_node;
 	struct kobject			*kobj_mem;
 	struct kobject			*kobj_cache;
@@ -145,7 +161,10 @@ struct kfd_topology_device {
 	struct attribute		attr_gpuid;
 	struct attribute		attr_name;
 	struct attribute		attr_props;
-	uint8_t				oem_id[CRAT_OEMID_LENGTH];
+	union {
+		uint8_t				oem_id[CRAT_OEMID_LENGTH];
+		uint64_t			oem_id64;
+	};
 	uint8_t				oem_table_id[CRAT_OEMTABLEID_LENGTH];
 	uint32_t			oem_revision;
 };
@@ -162,6 +181,22 @@ struct kfd_system_properties {
 	struct attribute	attr_props;
 };
 
+struct dmi_mem_device {
+	struct dmi_header header;
+	u16 physical_handle;
+	u16 error_handle;
+	u16 total_width;
+	u16 data_width;
+	u16 size;
+	u8 form_factor;
+	u8 device_set;
+	u8 device_locator;
+	u8 bank_locator;
+	u8 memory_type;
+	u16 type_detail;
+	u16 speed;
+} __packed;
+
 struct kfd_topology_device *kfd_create_topology_device(
 		struct list_head *device_list);
 void kfd_release_topology_device_list(struct list_head *device_list);
diff --git a/drivers/gpu/drm/amd/amdkfd/soc15_int.h b/drivers/gpu/drm/amd/amdkfd/soc15_int.h
index e3f3b0b93a59..e5c0205f2618 100644
--- a/drivers/gpu/drm/amd/amdkfd/soc15_int.h
+++ b/drivers/gpu/drm/amd/amdkfd/soc15_int.h
@@ -29,6 +29,7 @@
 #define SOC15_INTSRC_CP_BAD_OPCODE	183
 #define SOC15_INTSRC_SQ_INTERRUPT_MSG	239
 #define SOC15_INTSRC_VMC_FAULT		0
+#define SOC15_INTSRC_VMC_UTCL2_POISON	1
 #define SOC15_INTSRC_SDMA_TRAP		224
 #define SOC15_INTSRC_SDMA_ECC		220
 #define SOC21_INTSRC_SDMA_TRAP		49
@@ -40,6 +41,7 @@
 #define SOC15_VMID_FROM_IH_ENTRY(entry) (le32_to_cpu(entry[0]) >> 24 & 0xf)
 #define SOC15_VMID_TYPE_FROM_IH_ENTRY(entry) (le32_to_cpu(entry[0]) >> 31 & 0x1)
 #define SOC15_PASID_FROM_IH_ENTRY(entry) (le32_to_cpu(entry[3]) & 0xffff)
+#define SOC15_NODEID_FROM_IH_ENTRY(entry) (le32_to_cpu(entry[3]) >> 16 & 0xff)
 #define SOC15_CONTEXT_ID0_FROM_IH_ENTRY(entry) (le32_to_cpu(entry[4]))
 #define SOC15_CONTEXT_ID1_FROM_IH_ENTRY(entry) (le32_to_cpu(entry[5]))
 #define SOC15_CONTEXT_ID2_FROM_IH_ENTRY(entry) (le32_to_cpu(entry[6]))
diff --git a/drivers/gpu/drm/amd/display/dc/dcn316/Makefile b/drivers/gpu/drm/amd/amdxcp/Makefile
index 819d44a9439b..870501a4bb8c 100644
--- a/drivers/gpu/drm/amd/display/dc/dcn316/Makefile
+++ b/drivers/gpu/drm/amd/amdxcp/Makefile
@@ -1,5 +1,5 @@
 #
-# Copyright 2021 Advanced Micro Devices, Inc.
+# Copyright 2023 Advanced Micro Devices, Inc.
 #
 # Permission is hereby granted, free of charge, to any person obtaining a
 # copy of this software and associated documentation files (the "Software"),
@@ -19,12 +19,7 @@
 # ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
 # OTHER DEALINGS IN THE SOFTWARE.
 #
-# Authors: AMD
-#
-# Makefile for dcn316.
-
-DCN316 = dcn316_resource.o
 
-AMD_DAL_DCN316 = $(addprefix $(AMDDALPATH)/dc/dcn316/,$(DCN316))
+amdxcp-y := amdgpu_xcp_drv.o
 
-AMD_DISPLAY_FILES += $(AMD_DAL_DCN316)
+obj-$(CONFIG_DRM_AMDGPU) += amdxcp.o
diff --git a/drivers/gpu/drm/amd/amdxcp/amdgpu_xcp_drv.c b/drivers/gpu/drm/amd/amdxcp/amdgpu_xcp_drv.c
new file mode 100644
index 000000000000..44009aa8216e
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdxcp/amdgpu_xcp_drv.c
@@ -0,0 +1,152 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/export.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+
+#include <drm/drm_drv.h>
+
+#include "amdgpu_xcp_drv.h"
+
+#define MAX_XCP_PLATFORM_DEVICE 64
+
+struct xcp_device {
+	struct drm_device drm;
+	struct platform_device *pdev;
+};
+
+static const struct drm_driver amdgpu_xcp_driver = {
+	.driver_features = DRIVER_GEM | DRIVER_RENDER,
+	.name = "amdgpu_xcp_drv",
+	.major = 1,
+	.minor = 0,
+};
+
+static int8_t pdev_num;
+static struct xcp_device *xcp_dev[MAX_XCP_PLATFORM_DEVICE];
+static DEFINE_MUTEX(xcp_mutex);
+
+int amdgpu_xcp_drm_dev_alloc(struct drm_device **ddev)
+{
+	struct platform_device *pdev;
+	struct xcp_device *pxcp_dev;
+	char dev_name[20];
+	int ret, i;
+
+	guard(mutex)(&xcp_mutex);
+
+	if (pdev_num >= MAX_XCP_PLATFORM_DEVICE)
+		return -ENODEV;
+
+	for (i = 0; i < MAX_XCP_PLATFORM_DEVICE; i++) {
+		if (!xcp_dev[i])
+			break;
+	}
+
+	if (i >= MAX_XCP_PLATFORM_DEVICE)
+		return -ENODEV;
+
+	snprintf(dev_name, sizeof(dev_name), "amdgpu_xcp_%d", i);
+	pdev = platform_device_register_simple(dev_name, -1, NULL, 0);
+	if (IS_ERR(pdev))
+		return PTR_ERR(pdev);
+
+	if (!devres_open_group(&pdev->dev, NULL, GFP_KERNEL)) {
+		ret = -ENOMEM;
+		goto out_unregister;
+	}
+
+	pxcp_dev = devm_drm_dev_alloc(&pdev->dev, &amdgpu_xcp_driver, struct xcp_device, drm);
+	if (IS_ERR(pxcp_dev)) {
+		ret = PTR_ERR(pxcp_dev);
+		goto out_devres;
+	}
+
+	xcp_dev[i] = pxcp_dev;
+	xcp_dev[i]->pdev = pdev;
+	*ddev = &pxcp_dev->drm;
+	pdev_num++;
+
+	return 0;
+
+out_devres:
+	devres_release_group(&pdev->dev, NULL);
+out_unregister:
+	platform_device_unregister(pdev);
+
+	return ret;
+}
+EXPORT_SYMBOL(amdgpu_xcp_drm_dev_alloc);
+
+static void free_xcp_dev(int8_t index)
+{
+	if ((index < MAX_XCP_PLATFORM_DEVICE) && (xcp_dev[index])) {
+		struct platform_device *pdev = xcp_dev[index]->pdev;
+
+		devres_release_group(&pdev->dev, NULL);
+		platform_device_unregister(pdev);
+
+		xcp_dev[index] = NULL;
+		pdev_num--;
+	}
+}
+
+void amdgpu_xcp_drm_dev_free(struct drm_device *ddev)
+{
+	int8_t i;
+
+	guard(mutex)(&xcp_mutex);
+
+	for (i = 0; i < MAX_XCP_PLATFORM_DEVICE; i++) {
+		if ((xcp_dev[i]) && (&xcp_dev[i]->drm == ddev)) {
+			free_xcp_dev(i);
+			break;
+		}
+	}
+}
+EXPORT_SYMBOL(amdgpu_xcp_drm_dev_free);
+
+void amdgpu_xcp_drv_release(void)
+{
+	int8_t i;
+
+	guard(mutex)(&xcp_mutex);
+
+	for (i = 0; pdev_num && i < MAX_XCP_PLATFORM_DEVICE; i++) {
+		free_xcp_dev(i);
+	}
+}
+EXPORT_SYMBOL(amdgpu_xcp_drv_release);
+
+static void __exit amdgpu_xcp_drv_exit(void)
+{
+	amdgpu_xcp_drv_release();
+}
+
+module_exit(amdgpu_xcp_drv_exit);
+
+MODULE_AUTHOR("AMD linux driver team");
+MODULE_DESCRIPTION("AMD XCP PLATFORM DEVICES");
+MODULE_LICENSE("GPL and additional rights");
diff --git a/drivers/gpu/drm/amd/amdxcp/amdgpu_xcp_drv.h b/drivers/gpu/drm/amd/amdxcp/amdgpu_xcp_drv.h
new file mode 100644
index 000000000000..580a1602c8e3
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdxcp/amdgpu_xcp_drv.h
@@ -0,0 +1,30 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef _AMDGPU_XCP_DRV_H_
+#define _AMDGPU_XCP_DRV_H_
+
+int amdgpu_xcp_drm_dev_alloc(struct drm_device **ddev);
+void amdgpu_xcp_drm_dev_free(struct drm_device *ddev);
+void amdgpu_xcp_drv_release(void);
+#endif /* _AMDGPU_XCP_DRV_H_ */
diff --git a/drivers/gpu/drm/amd/display/Kconfig b/drivers/gpu/drm/amd/display/Kconfig
index 06b438217c61..abd3b6564373 100644
--- a/drivers/gpu/drm/amd/display/Kconfig
+++ b/drivers/gpu/drm/amd/display/Kconfig
@@ -1,24 +1,29 @@
 # SPDX-License-Identifier: MIT
+# Copyright © 2019-2024 Advanced Micro Devices, Inc. All rights reserved.
+
 menu "Display Engine Configuration"
 	depends on DRM && DRM_AMDGPU
 
 config DRM_AMD_DC
 	bool "AMD DC - Enable new display engine"
 	default y
-	depends on BROKEN || !CC_IS_CLANG || X86_64 || SPARC64 || ARM64
+	depends on BROKEN || !CC_IS_CLANG || ARM64 || LOONGARCH || RISCV || SPARC64 || X86_64
+	select CEC_CORE
+	select CEC_NOTIFIER
 	select SND_HDA_COMPONENT if SND_HDA_CORE
 	# !CC_IS_CLANG: https://github.com/ClangBuiltLinux/linux/issues/1752
-	select DRM_AMD_DC_FP if (X86 || PPC64 || (ARM64 && KERNEL_MODE_NEON && !CC_IS_CLANG))
+	select DRM_AMD_DC_FP if ARCH_HAS_KERNEL_FPU_SUPPORT && !(CC_IS_CLANG && (ARM64 || LOONGARCH || RISCV))
 	help
 	  Choose this option if you want to use the new display engine
 	  support for AMDGPU. This adds required support for Vega and
 	  Raven ASICs.
 
-	  calculate_bandwidth() is presently broken on all !(X86_64 || SPARC64 || ARM64)
-	  architectures built with Clang (all released versions), whereby the stack
-	  frame gets blown up to well over 5k.  This would cause an immediate kernel
-	  panic on most architectures.  We'll revert this when the following bug report
-	  has been resolved: https://github.com/llvm/llvm-project/issues/41896.
+	  calculate_bandwidth() is presently broken on all !(X86_64 || SPARC64 ||
+	  ARM64 || LOONGARCH || RISCV) architectures built with Clang (all released
+	  versions), whereby the stack frame gets blown up to well over 5k.  This
+	  would cause an immediate kernel panic on most architectures.  We'll revert
+	  this when the following bug report has been resolved:
+	  https://github.com/llvm/llvm-project/issues/41896.
 
 config DRM_AMD_DC_FP
 	def_bool n
@@ -42,16 +47,13 @@ config DEBUG_KERNEL_DC
 	  Choose this option if you want to hit kdgb_break in assert.
 
 config DRM_AMD_SECURE_DISPLAY
-        bool "Enable secure display support"
-        depends on DEBUG_FS
-        depends on DRM_AMD_DC_FP
-        help
-            Choose this option if you want to
-            support secure display
-
-            This option enables the calculation
-            of crc of specific region via debugfs.
-            Cooperate with specific DMCU FW.
+	bool "Enable secure display support"
+	depends on DEBUG_FS
+	depends on DRM_AMD_DC_FP
+	help
+	  Choose this option if you want to support secure display
 
+	  This option enables the calculation of crc of specific region via
+	  debugfs. Cooperate with specific DMCU FW.
 
 endmenu
diff --git a/drivers/gpu/drm/amd/display/Makefile b/drivers/gpu/drm/amd/display/Makefile
index 0d610cb376bb..0084a8d55254 100644
--- a/drivers/gpu/drm/amd/display/Makefile
+++ b/drivers/gpu/drm/amd/display/Makefile
@@ -29,6 +29,22 @@ AMDDALPATH = $(RELATIVE_AMD_DISPLAY_PATH)
 subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/inc/
 subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/inc/hw
 subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/clk_mgr
+subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/hwss
+subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/resource
+subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/dsc
+subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/optc
+subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/dpp
+subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/hubbub
+subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/dccg
+subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/hubp
+subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/dio
+subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/dwb
+subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/hpo
+subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/mmhubbub
+subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/mpc
+subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/opp
+subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/pg
+subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc/soc_and_ip_translator
 subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/modules/inc
 subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/modules/freesync
 subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/modules/color
diff --git a/drivers/gpu/drm/amd/display/TODO b/drivers/gpu/drm/amd/display/TODO
deleted file mode 100644
index a8a6c106e8c7..000000000000
--- a/drivers/gpu/drm/amd/display/TODO
+++ /dev/null
@@ -1,110 +0,0 @@
-===============================================================================
-TODOs
-===============================================================================
-
-1. Base this on drm-next - WIP
-
-
-2. Cleanup commit history
-
-
-3. WIP - Drop page flip helper and use DRM's version
-
-
-4. DONE - Flatten all DC objects
-    * dc_stream/core_stream/stream should just be dc_stream
-    * Same for other DC objects
-
-    "Is there any major reason to keep all those abstractions?
-
-    Could you collapse everything into struct dc_stream?
-
-    I haven't looked recently but I didn't get the impression there was a
-    lot of design around what was public/protected, more whatever needed
-    to be used by someone else was in public."
-    ~ Dave Airlie
-
-
-5. DONE - Rename DC objects to align more with DRM
-    * dc_surface -> dc_plane_state
-    * dc_stream -> dc_stream_state
-
-
-6. DONE - Per-plane and per-stream validation
-
-
-7. WIP - Per-plane and per-stream commit
-
-
-8. WIP - Split pipe_ctx into plane and stream resource structs
-
-
-9. Attach plane and stream reources to state object instead of validate_context
-
-
-10. Remove dc_edid_caps and drm_helpers_parse_edid_caps
-    * Use drm_display_info instead
-    * Remove DC's edid quirks and rely on DRM's quirks (add quirks if needed)
-
-    "Making sure you use the sink-specific helper libraries and kernel
-    subsystems, since there's really no good reason to have 2nd
-    implementation of those in the kernel. Looks likes that's done for mst
-    and edid parsing. There's still a bit a midlayer feeling to the edid
-    parsing side (e.g. dc_edid_caps and dm_helpers_parse_edid_caps, I
-    think it'd be much better if you convert that over to reading stuff
-    from drm_display_info and if needed, push stuff into the core). Also,
-    I can't come up with a good reason why DC needs all this (except to
-    reimplement half of our edid quirk table, which really isn't a good
-    idea). Might be good if you put this onto the list of things to fix
-    long-term, but imo not a blocker. Definitely make sure new stuff
-    doesn't slip in (i.e. if you start adding edid quirks to DC instead of
-    the drm core, refactoring to use the core edid stuff was pointless)."
-    ~ Daniel Vetter
-
-
-11. Remove dc/i2caux. This folder can be somewhat misleading. It's basically an
-overy complicated HW programming function for sendind and receiving i2c/aux
-commands. We can greatly simplify that and move it into dc/dceXYZ like other
-HW blocks.
-
-12. drm_modeset_lock in MST should no longer be needed in recent kernels
-    * Adopt appropriate locking scheme
-
-13. get_modes and best_encoder callbacks look a bit funny. Can probably rip out
-a few indirections, and consider removing entirely and using the
-drm_atomic_helper_best_encoder default behaviour.
-
-14. core/dc_debug.c, consider switching to the atomic state debug helpers and
-moving all your driver state printing into the various atomic_print_state
-callbacks. There's also plans to expose this stuff in a standard way across all
-drivers, to make debugging userspace compositors easier across different hw.
-
-15. Move DP/HDMI dual mode adaptors to drm_dp_dual_mode_helper.c. See
-dal_ddc_service_i2c_query_dp_dual_mode_adaptor.
-
-16. Move to core SCDC helpers (I think those are new since initial DC review).
-
-17. There's still a pretty massive layer cake around dp aux and DPCD handling,
-with like 3 levels of abstraction and using your own structures instead of the
-stuff in drm_dp_helper.h. drm_dp_helper.h isn't really great and already has 2
-incompatible styles, just means more reasons not to add a third (or well third
-one gets to do the cleanup refactor).
-
-18. There's a pile of sink handling code, both for DP and HDMI where I didn't
-immediately recognize the standard. I think long term it'd be best for the drm
-subsystem if we try to move as much of that into helpers/core as possible, and
-share it with drivers. But that's a very long term goal, and by far not just an
-issue with DC - other drivers, especially around DP sink handling, are equally
-guilty.
-
-19. DONE - The DC logger is still a rather sore thing, but I know that the
-DRM_DEBUG stuff just isn't up to the challenges either. We need to figure out
-something that integrates better with DRM and linux debug printing, while not
-being useless with filtering output. dynamic debug printing might be an option.
-
-20. Use kernel i2c device to program HDMI retimer. Some boards have an HDMI
-retimer that we need to program to pass PHY compliance. Currently that's
-bypassing the i2c device and goes directly to HW. This should be changed.
-
-21. Remove vector.c from dc/basics. It's used in DDC code which can probably
-be simplified enough to no longer need a vector implementation.
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/Makefile b/drivers/gpu/drm/amd/display/amdgpu_dm/Makefile
index 249b073f6a23..8e949fe77312 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/Makefile
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/Makefile
@@ -25,22 +25,27 @@
 
 
 
+ifneq ($(CONFIG_DRM_AMD_DC),)
 AMDGPUDM = \
 	amdgpu_dm.o \
 	amdgpu_dm_plane.o \
 	amdgpu_dm_crtc.o \
 	amdgpu_dm_irq.o \
 	amdgpu_dm_mst_types.o \
-	amdgpu_dm_color.o
+	amdgpu_dm_color.o \
+	amdgpu_dm_services.o \
+	amdgpu_dm_helpers.o \
+	amdgpu_dm_pp_smu.o \
+	amdgpu_dm_psr.o \
+	amdgpu_dm_replay.o \
+	amdgpu_dm_quirks.o \
+	amdgpu_dm_wb.o \
+	amdgpu_dm_colorop.o
 
 ifdef CONFIG_DRM_AMD_DC_FP
 AMDGPUDM += dc_fpu.o
 endif
 
-ifneq ($(CONFIG_DRM_AMD_DC),)
-AMDGPUDM += amdgpu_dm_services.o amdgpu_dm_helpers.o amdgpu_dm_pp_smu.o amdgpu_dm_psr.o
-endif
-
 AMDGPUDM += amdgpu_dm_hdcp.o
 
 ifneq ($(CONFIG_DEBUG_FS),)
@@ -52,3 +57,4 @@ subdir-ccflags-y += -I$(FULL_AMD_DISPLAY_PATH)/dc
 AMDGPU_DM = $(addprefix $(AMDDALPATH)/amdgpu_dm/,$(AMDGPUDM))
 
 AMD_DISPLAY_FILES += $(AMDGPU_DM)
+endif
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
index c4aa87086a4e..740711ac1037 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
@@ -1,3 +1,4 @@
+// SPDX-License-Identifier: MIT
 /*
  * Copyright 2015 Advanced Micro Devices, Inc.
  *
@@ -37,14 +38,13 @@
 #include "dc/dc_dmub_srv.h"
 #include "dc/dc_edid_parser.h"
 #include "dc/dc_stat.h"
+#include "dc/dc_state.h"
 #include "amdgpu_dm_trace.h"
-#include "dpcd_defs.h"
 #include "link/protocols/link_dpcd.h"
 #include "link_service_types.h"
 #include "link/protocols/link_dp_capability.h"
 #include "link/protocols/link_ddc.h"
 
-#include "vid.h"
 #include "amdgpu.h"
 #include "amdgpu_display.h"
 #include "amdgpu_ucode.h"
@@ -54,7 +54,7 @@
 #include "amdgpu_dm_crtc.h"
 #include "amdgpu_dm_hdcp.h"
 #include <drm/display/drm_hdcp_helper.h>
-#include "amdgpu_pm.h"
+#include "amdgpu_dm_wb.h"
 #include "amdgpu_atombios.h"
 
 #include "amd_shared.h"
@@ -65,6 +65,7 @@
 #include "amdgpu_dm_debugfs.h"
 #endif
 #include "amdgpu_dm_psr.h"
+#include "amdgpu_dm_replay.h"
 
 #include "ivsrcid/ivsrcid_vislands30.h"
 
@@ -74,39 +75,37 @@
 #include <linux/types.h>
 #include <linux/pm_runtime.h>
 #include <linux/pci.h>
+#include <linux/power_supply.h>
 #include <linux/firmware.h>
 #include <linux/component.h>
-#include <linux/dmi.h>
+#include <linux/sort.h>
 
+#include <drm/drm_privacy_screen_consumer.h>
 #include <drm/display/drm_dp_mst_helper.h>
 #include <drm/display/drm_hdmi_helper.h>
 #include <drm/drm_atomic.h>
 #include <drm/drm_atomic_uapi.h>
 #include <drm/drm_atomic_helper.h>
 #include <drm/drm_blend.h>
+#include <drm/drm_fixed.h>
 #include <drm/drm_fourcc.h>
 #include <drm/drm_edid.h>
+#include <drm/drm_eld.h>
+#include <drm/drm_utils.h>
 #include <drm/drm_vblank.h>
 #include <drm/drm_audio_component.h>
 #include <drm/drm_gem_atomic_helper.h>
-#include <drm/drm_plane_helper.h>
 
+#include <media/cec-notifier.h>
 #include <acpi/video.h>
 
 #include "ivsrcid/dcn/irqsrcs_dcn_1_0.h"
 
-#include "dcn/dcn_1_0_offset.h"
-#include "dcn/dcn_1_0_sh_mask.h"
-#include "soc15_hw_ip.h"
-#include "soc15_common.h"
-#include "vega10_ip_offset.h"
-
-#include "gc/gc_11_0_0_offset.h"
-#include "gc/gc_11_0_0_sh_mask.h"
-
 #include "modules/inc/mod_freesync.h"
 #include "modules/power/power_helpers.h"
 
+static_assert(AMDGPU_DMUB_NOTIFICATION_MAX == DMUB_NOTIFICATION_MAX, "AMDGPU_DMUB_NOTIFICATION_MAX mismatch");
+
 #define FIRMWARE_RENOIR_DMUB "amdgpu/renoir_dmcub.bin"
 MODULE_FIRMWARE(FIRMWARE_RENOIR_DMUB);
 #define FIRMWARE_SIENNA_CICHLID_DMUB "amdgpu/sienna_cichlid_dmcub.bin"
@@ -141,6 +140,18 @@ MODULE_FIRMWARE(FIRMWARE_RAVEN_DMCU);
 #define FIRMWARE_NAVI12_DMCU            "amdgpu/navi12_dmcu.bin"
 MODULE_FIRMWARE(FIRMWARE_NAVI12_DMCU);
 
+#define FIRMWARE_DCN_35_DMUB "amdgpu/dcn_3_5_dmcub.bin"
+MODULE_FIRMWARE(FIRMWARE_DCN_35_DMUB);
+
+#define FIRMWARE_DCN_351_DMUB "amdgpu/dcn_3_5_1_dmcub.bin"
+MODULE_FIRMWARE(FIRMWARE_DCN_351_DMUB);
+
+#define FIRMWARE_DCN_36_DMUB "amdgpu/dcn_3_6_dmcub.bin"
+MODULE_FIRMWARE(FIRMWARE_DCN_36_DMUB);
+
+#define FIRMWARE_DCN_401_DMUB "amdgpu/dcn_4_0_1_dmcub.bin"
+MODULE_FIRMWARE(FIRMWARE_DCN_401_DMUB);
+
 /* Number of bytes in PSP header for firmware. */
 #define PSP_HEADER_BYTES 0x100
 
@@ -161,6 +172,9 @@ MODULE_FIRMWARE(FIRMWARE_NAVI12_DMCU);
 static int amdgpu_dm_init(struct amdgpu_device *adev);
 static void amdgpu_dm_fini(struct amdgpu_device *adev);
 static bool is_freesync_video_mode(const struct drm_display_mode *mode, struct amdgpu_dm_connector *aconnector);
+static void reset_freesync_config_for_crtc(struct dm_crtc_state *new_crtc_state);
+static struct amdgpu_i2c_adapter *
+create_i2c(struct ddc_service *ddc_service, bool oem);
 
 static enum drm_mode_subconnector get_subconnector_type(struct dc_link *link)
 {
@@ -219,6 +233,7 @@ static int amdgpu_dm_encoder_init(struct drm_device *dev,
 
 static int amdgpu_dm_connector_get_modes(struct drm_connector *connector);
 
+static int amdgpu_dm_atomic_setup_commit(struct drm_atomic_state *state);
 static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state);
 
 static int amdgpu_dm_atomic_check(struct drm_device *dev,
@@ -227,6 +242,10 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 static void handle_hpd_irq_helper(struct amdgpu_dm_connector *aconnector);
 static void handle_hpd_rx_irq(void *param);
 
+static void amdgpu_dm_backlight_set_level(struct amdgpu_display_manager *dm,
+					 int bl_idx,
+					 u32 user_brightness);
+
 static bool
 is_timing_unchanged_for_freesync(struct drm_crtc_state *old_crtc_state,
 				 struct drm_crtc_state *new_crtc_state);
@@ -245,72 +264,77 @@ is_timing_unchanged_for_freesync(struct drm_crtc_state *old_crtc_state,
  */
 static u32 dm_vblank_get_counter(struct amdgpu_device *adev, int crtc)
 {
+	struct amdgpu_crtc *acrtc = NULL;
+
 	if (crtc >= adev->mode_info.num_crtc)
 		return 0;
-	else {
-		struct amdgpu_crtc *acrtc = adev->mode_info.crtcs[crtc];
 
-		if (acrtc->dm_irq_params.stream == NULL) {
-			DRM_ERROR("dc_stream_state is NULL for crtc '%d'!\n",
-				  crtc);
-			return 0;
-		}
+	acrtc = adev->mode_info.crtcs[crtc];
 
-		return dc_stream_get_vblank_counter(acrtc->dm_irq_params.stream);
+	if (!acrtc->dm_irq_params.stream) {
+		drm_err(adev_to_drm(adev), "dc_stream_state is NULL for crtc '%d'!\n",
+			  crtc);
+		return 0;
 	}
+
+	return dc_stream_get_vblank_counter(acrtc->dm_irq_params.stream);
 }
 
 static int dm_crtc_get_scanoutpos(struct amdgpu_device *adev, int crtc,
 				  u32 *vbl, u32 *position)
 {
-	u32 v_blank_start, v_blank_end, h_position, v_position;
+	u32 v_blank_start = 0, v_blank_end = 0, h_position = 0, v_position = 0;
+	struct amdgpu_crtc *acrtc = NULL;
+	struct dc *dc = adev->dm.dc;
 
 	if ((crtc < 0) || (crtc >= adev->mode_info.num_crtc))
 		return -EINVAL;
-	else {
-		struct amdgpu_crtc *acrtc = adev->mode_info.crtcs[crtc];
 
-		if (acrtc->dm_irq_params.stream ==  NULL) {
-			DRM_ERROR("dc_stream_state is NULL for crtc '%d'!\n",
-				  crtc);
-			return 0;
-		}
+	acrtc = adev->mode_info.crtcs[crtc];
 
-		/*
-		 * TODO rework base driver to use values directly.
-		 * for now parse it back into reg-format
-		 */
-		dc_stream_get_scanoutpos(acrtc->dm_irq_params.stream,
-					 &v_blank_start,
-					 &v_blank_end,
-					 &h_position,
-					 &v_position);
-
-		*position = v_position | (h_position << 16);
-		*vbl = v_blank_start | (v_blank_end << 16);
+	if (!acrtc->dm_irq_params.stream) {
+		drm_err(adev_to_drm(adev), "dc_stream_state is NULL for crtc '%d'!\n",
+			  crtc);
+		return 0;
 	}
 
+	if (dc && dc->caps.ips_support && dc->idle_optimizations_allowed)
+		dc_allow_idle_optimizations(dc, false);
+
+	/*
+	 * TODO rework base driver to use values directly.
+	 * for now parse it back into reg-format
+	 */
+	dc_stream_get_scanoutpos(acrtc->dm_irq_params.stream,
+				 &v_blank_start,
+				 &v_blank_end,
+				 &h_position,
+				 &v_position);
+
+	*position = v_position | (h_position << 16);
+	*vbl = v_blank_start | (v_blank_end << 16);
+
 	return 0;
 }
 
-static bool dm_is_idle(void *handle)
+static bool dm_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* XXX todo */
 	return true;
 }
 
-static int dm_wait_for_idle(void *handle)
+static int dm_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* XXX todo */
 	return 0;
 }
 
-static bool dm_check_soft_reset(void *handle)
+static bool dm_check_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	return false;
 }
 
-static int dm_soft_reset(void *handle)
+static int dm_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	/* XXX todo */
 	return 0;
@@ -340,6 +364,8 @@ get_crtc_by_otg_inst(struct amdgpu_device *adev,
 static inline bool is_dc_timing_adjust_needed(struct dm_crtc_state *old_state,
 					      struct dm_crtc_state *new_state)
 {
+	if (new_state->stream->adjust.timing_adjust_pending)
+		return true;
 	if (new_state->freesync_config.state ==  VRR_STATE_ACTIVE_FIXED)
 		return true;
 	else if (amdgpu_dm_crtc_vrr_active(old_state) != amdgpu_dm_crtc_vrr_active(new_state))
@@ -348,13 +374,18 @@ static inline bool is_dc_timing_adjust_needed(struct dm_crtc_state *old_state,
 		return false;
 }
 
-static inline void reverse_planes_order(struct dc_surface_update *array_of_surface_update,
-					int planes_count)
+/*
+ * DC will program planes with their z-order determined by their ordering
+ * in the dc_surface_updates array. This comparator is used to sort them
+ * by descending zpos.
+ */
+static int dm_plane_layer_index_cmp(const void *a, const void *b)
 {
-	int i, j;
+	const struct dc_surface_update *sa = (struct dc_surface_update *)a;
+	const struct dc_surface_update *sb = (struct dc_surface_update *)b;
 
-	for (i = 0, j = planes_count - 1; i < j; i++, j--)
-		swap(array_of_surface_update[i], array_of_surface_update[j]);
+	/* Sort by descending dc_plane layer_index (i.e. normalized_zpos) */
+	return sb->surface->layer_index - sa->surface->layer_index;
 }
 
 /**
@@ -365,6 +396,14 @@ static inline void reverse_planes_order(struct dc_surface_update *array_of_surfa
  * adjustments and preparation before calling it. This function is a wrapper
  * for the dc_update_planes_and_stream that does any required configuration
  * before passing control to DC.
+ *
+ * @dc: Display Core control structure
+ * @update_type: specify whether it is FULL/MEDIUM/FAST update
+ * @planes_count: planes count to update
+ * @stream: stream state
+ * @stream_update: stream update
+ * @array_of_surface_update: dc surface update pointer
+ *
  */
 static inline bool update_planes_and_stream_adapter(struct dc *dc,
 						    int update_type,
@@ -373,13 +412,13 @@ static inline bool update_planes_and_stream_adapter(struct dc *dc,
 						    struct dc_stream_update *stream_update,
 						    struct dc_surface_update *array_of_surface_update)
 {
-	reverse_planes_order(array_of_surface_update, planes_count);
+	sort(array_of_surface_update, planes_count,
+	     sizeof(*array_of_surface_update), dm_plane_layer_index_cmp, NULL);
 
 	/*
 	 * Previous frame finished and HW is ready for optimization.
 	 */
-	if (update_type == UPDATE_TYPE_FAST)
-		dc_post_update_surfaces_to_stream(dc);
+	dc_post_update_surfaces_to_stream(dc);
 
 	return dc_update_planes_and_stream(dc,
 					   array_of_surface_update,
@@ -400,6 +439,7 @@ static void dm_pflip_high_irq(void *interrupt_params)
 	struct amdgpu_crtc *amdgpu_crtc;
 	struct common_irq_params *irq_params = interrupt_params;
 	struct amdgpu_device *adev = irq_params->adev;
+	struct drm_device *dev = adev_to_drm(adev);
 	unsigned long flags;
 	struct drm_pending_vblank_event *e;
 	u32 vpos, hpos, v_blank_start, v_blank_end;
@@ -410,18 +450,17 @@ static void dm_pflip_high_irq(void *interrupt_params)
 	/* IRQ could occur when in initial stage */
 	/* TODO work and BO cleanup */
 	if (amdgpu_crtc == NULL) {
-		DC_LOG_PFLIP("CRTC is null, returning.\n");
+		drm_dbg_state(dev, "CRTC is null, returning.\n");
 		return;
 	}
 
 	spin_lock_irqsave(&adev_to_drm(adev)->event_lock, flags);
 
-	if (amdgpu_crtc->pflip_status != AMDGPU_FLIP_SUBMITTED){
-		DC_LOG_PFLIP("amdgpu_crtc->pflip_status = %d !=AMDGPU_FLIP_SUBMITTED(%d) on crtc:%d[%p] \n",
-						 amdgpu_crtc->pflip_status,
-						 AMDGPU_FLIP_SUBMITTED,
-						 amdgpu_crtc->crtc_id,
-						 amdgpu_crtc);
+	if (amdgpu_crtc->pflip_status != AMDGPU_FLIP_SUBMITTED) {
+		drm_dbg_state(dev,
+			      "amdgpu_crtc->pflip_status = %d != AMDGPU_FLIP_SUBMITTED(%d) on crtc:%d[%p]\n",
+			      amdgpu_crtc->pflip_status, AMDGPU_FLIP_SUBMITTED,
+			      amdgpu_crtc->crtc_id, amdgpu_crtc);
 		spin_unlock_irqrestore(&adev_to_drm(adev)->event_lock, flags);
 		return;
 	}
@@ -487,9 +526,53 @@ static void dm_pflip_high_irq(void *interrupt_params)
 	amdgpu_crtc->pflip_status = AMDGPU_FLIP_NONE;
 	spin_unlock_irqrestore(&adev_to_drm(adev)->event_lock, flags);
 
-	DC_LOG_PFLIP("crtc:%d[%p], pflip_stat:AMDGPU_FLIP_NONE, vrr[%d]-fp %d\n",
-		     amdgpu_crtc->crtc_id, amdgpu_crtc,
-		     vrr_active, (int) !e);
+	drm_dbg_state(dev,
+		      "crtc:%d[%p], pflip_stat:AMDGPU_FLIP_NONE, vrr[%d]-fp %d\n",
+		      amdgpu_crtc->crtc_id, amdgpu_crtc, vrr_active, (int)!e);
+}
+
+static void dm_handle_vmin_vmax_update(struct work_struct *offload_work)
+{
+	struct vupdate_offload_work *work = container_of(offload_work, struct vupdate_offload_work, work);
+	struct amdgpu_device *adev = work->adev;
+	struct dc_stream_state *stream = work->stream;
+	struct dc_crtc_timing_adjust *adjust = work->adjust;
+
+	mutex_lock(&adev->dm.dc_lock);
+	dc_stream_adjust_vmin_vmax(adev->dm.dc, stream, adjust);
+	mutex_unlock(&adev->dm.dc_lock);
+
+	dc_stream_release(stream);
+	kfree(work->adjust);
+	kfree(work);
+}
+
+static void schedule_dc_vmin_vmax(struct amdgpu_device *adev,
+	struct dc_stream_state *stream,
+	struct dc_crtc_timing_adjust *adjust)
+{
+	struct vupdate_offload_work *offload_work = kzalloc(sizeof(*offload_work), GFP_NOWAIT);
+	if (!offload_work) {
+		drm_dbg_driver(adev_to_drm(adev), "Failed to allocate vupdate_offload_work\n");
+		return;
+	}
+
+	struct dc_crtc_timing_adjust *adjust_copy = kzalloc(sizeof(*adjust_copy), GFP_NOWAIT);
+	if (!adjust_copy) {
+		drm_dbg_driver(adev_to_drm(adev), "Failed to allocate adjust_copy\n");
+		kfree(offload_work);
+		return;
+	}
+
+	dc_stream_retain(stream);
+	memcpy(adjust_copy, adjust, sizeof(*adjust_copy));
+
+	INIT_WORK(&offload_work->work, dm_handle_vmin_vmax_update);
+	offload_work->adev = adev;
+	offload_work->stream = stream;
+	offload_work->adjust = adjust_copy;
+
+	queue_work(system_wq, &offload_work->work);
 }
 
 static void dm_vupdate_high_irq(void *interrupt_params)
@@ -508,7 +591,7 @@ static void dm_vupdate_high_irq(void *interrupt_params)
 	if (acrtc) {
 		vrr_active = amdgpu_dm_crtc_vrr_active_irq(acrtc);
 		drm_dev = acrtc->base.dev;
-		vblank = &drm_dev->vblank[acrtc->base.index];
+		vblank = drm_crtc_vblank_crtc(&acrtc->base);
 		previous_timestamp = atomic64_read(&irq_params->previous_timestamp);
 		frame_duration_ns = vblank->time - previous_timestamp;
 
@@ -519,9 +602,9 @@ static void dm_vupdate_high_irq(void *interrupt_params)
 			atomic64_set(&irq_params->previous_timestamp, vblank->time);
 		}
 
-		DC_LOG_VBLANK("crtc:%d, vupdate-vrr:%d\n",
-			      acrtc->crtc_id,
-			      vrr_active);
+		drm_dbg_vbl(drm_dev,
+			    "crtc:%d, vupdate-vrr:%d\n", acrtc->crtc_id,
+			    vrr_active);
 
 		/* Core vblank handling is done here after end of front-porch in
 		 * vrr mode, as vblank timestamping will give valid results
@@ -529,22 +612,27 @@ static void dm_vupdate_high_irq(void *interrupt_params)
 		 * page-flip completion events that have been queued to us
 		 * if a pageflip happened inside front-porch.
 		 */
-		if (vrr_active) {
+		if (vrr_active && acrtc->dm_irq_params.stream) {
+			bool replay_en = acrtc->dm_irq_params.stream->link->replay_settings.replay_feature_enabled;
+			bool psr_en = acrtc->dm_irq_params.stream->link->psr_settings.psr_feature_enabled;
+			bool fs_active_var_en = acrtc->dm_irq_params.freesync_config.state
+				== VRR_STATE_ACTIVE_VARIABLE;
+
 			amdgpu_dm_crtc_handle_vblank(acrtc);
 
 			/* BTR processing for pre-DCE12 ASICs */
-			if (acrtc->dm_irq_params.stream &&
-			    adev->family < AMDGPU_FAMILY_AI) {
+			if (adev->family < AMDGPU_FAMILY_AI) {
 				spin_lock_irqsave(&adev_to_drm(adev)->event_lock, flags);
 				mod_freesync_handle_v_update(
 				    adev->dm.freesync_module,
 				    acrtc->dm_irq_params.stream,
 				    &acrtc->dm_irq_params.vrr_params);
 
-				dc_stream_adjust_vmin_vmax(
-				    adev->dm.dc,
-				    acrtc->dm_irq_params.stream,
-				    &acrtc->dm_irq_params.vrr_params.adjust);
+				if (fs_active_var_en || (!fs_active_var_en && !replay_en && !psr_en)) {
+					schedule_dc_vmin_vmax(adev,
+						acrtc->dm_irq_params.stream,
+						&acrtc->dm_irq_params.vrr_params.adjust);
+				}
 				spin_unlock_irqrestore(&adev_to_drm(adev)->event_lock, flags);
 			}
 		}
@@ -562,6 +650,7 @@ static void dm_crtc_high_irq(void *interrupt_params)
 {
 	struct common_irq_params *irq_params = interrupt_params;
 	struct amdgpu_device *adev = irq_params->adev;
+	struct drm_writeback_job *job;
 	struct amdgpu_crtc *acrtc;
 	unsigned long flags;
 	int vrr_active;
@@ -570,10 +659,39 @@ static void dm_crtc_high_irq(void *interrupt_params)
 	if (!acrtc)
 		return;
 
+	if (acrtc->wb_conn) {
+		spin_lock_irqsave(&acrtc->wb_conn->job_lock, flags);
+
+		if (acrtc->wb_pending) {
+			job = list_first_entry_or_null(&acrtc->wb_conn->job_queue,
+						       struct drm_writeback_job,
+						       list_entry);
+			acrtc->wb_pending = false;
+			spin_unlock_irqrestore(&acrtc->wb_conn->job_lock, flags);
+
+			if (job) {
+				unsigned int v_total, refresh_hz;
+				struct dc_stream_state *stream = acrtc->dm_irq_params.stream;
+
+				v_total = stream->adjust.v_total_max ?
+					  stream->adjust.v_total_max : stream->timing.v_total;
+				refresh_hz = div_u64((uint64_t) stream->timing.pix_clk_100hz *
+					     100LL, (v_total * stream->timing.h_total));
+				mdelay(1000 / refresh_hz);
+
+				drm_writeback_signal_completion(acrtc->wb_conn, 0);
+				dc_stream_fc_disable_writeback(adev->dm.dc,
+							       acrtc->dm_irq_params.stream, 0);
+			}
+		} else
+			spin_unlock_irqrestore(&acrtc->wb_conn->job_lock, flags);
+	}
+
 	vrr_active = amdgpu_dm_crtc_vrr_active_irq(acrtc);
 
-	DC_LOG_VBLANK("crtc:%d, vupdate-vrr:%d, planes:%d\n", acrtc->crtc_id,
-		      vrr_active, acrtc->dm_irq_params.active_planes);
+	drm_dbg_vbl(adev_to_drm(adev),
+		    "crtc:%d, vupdate-vrr:%d, planes:%d\n", acrtc->crtc_id,
+		    vrr_active, acrtc->dm_irq_params.active_planes);
 
 	/**
 	 * Core vblank handling at start of front-porch is only possible
@@ -597,15 +715,20 @@ static void dm_crtc_high_irq(void *interrupt_params)
 	spin_lock_irqsave(&adev_to_drm(adev)->event_lock, flags);
 
 	if (acrtc->dm_irq_params.stream &&
-	    acrtc->dm_irq_params.vrr_params.supported &&
-	    acrtc->dm_irq_params.freesync_config.state ==
-		    VRR_STATE_ACTIVE_VARIABLE) {
+		acrtc->dm_irq_params.vrr_params.supported) {
+		bool replay_en = acrtc->dm_irq_params.stream->link->replay_settings.replay_feature_enabled;
+		bool psr_en = acrtc->dm_irq_params.stream->link->psr_settings.psr_feature_enabled;
+		bool fs_active_var_en = acrtc->dm_irq_params.freesync_config.state == VRR_STATE_ACTIVE_VARIABLE;
+
 		mod_freesync_handle_v_update(adev->dm.freesync_module,
 					     acrtc->dm_irq_params.stream,
 					     &acrtc->dm_irq_params.vrr_params);
 
-		dc_stream_adjust_vmin_vmax(adev->dm.dc, acrtc->dm_irq_params.stream,
-					   &acrtc->dm_irq_params.vrr_params.adjust);
+		/* update vmin_vmax only if freesync is enabled, or only if PSR and REPLAY are disabled */
+		if (fs_active_var_en || (!fs_active_var_en && !replay_en && !psr_en)) {
+			schedule_dc_vmin_vmax(adev, acrtc->dm_irq_params.stream,
+					&acrtc->dm_irq_params.vrr_params.adjust);
+		}
 	}
 
 	/*
@@ -673,6 +796,29 @@ static void dmub_aux_setconfig_callback(struct amdgpu_device *adev,
 		complete(&adev->dm.dmub_aux_transfer_done);
 }
 
+static void dmub_aux_fused_io_callback(struct amdgpu_device *adev,
+					struct dmub_notification *notify)
+{
+	if (!adev || !notify) {
+		ASSERT(false);
+		return;
+	}
+
+	const struct dmub_cmd_fused_request *req = &notify->fused_request;
+	const uint8_t ddc_line = req->u.aux.ddc_line;
+
+	if (ddc_line >= ARRAY_SIZE(adev->dm.fused_io)) {
+		ASSERT(false);
+		return;
+	}
+
+	struct fused_io_sync *sync = &adev->dm.fused_io[ddc_line];
+
+	static_assert(sizeof(*req) <= sizeof(sync->reply_data), "Size mismatch");
+	memcpy(sync->reply_data, req, sizeof(*req));
+	complete(&sync->replied);
+}
+
 /**
  * dmub_hpd_callback - DMUB HPD interrupt processing callback.
  * @adev: amdgpu_device pointer
@@ -696,12 +842,18 @@ static void dmub_hpd_callback(struct amdgpu_device *adev,
 		return;
 
 	if (notify == NULL) {
-		DRM_ERROR("DMUB HPD callback notification was NULL");
+		drm_err(adev_to_drm(adev), "DMUB HPD callback notification was NULL");
 		return;
 	}
 
 	if (notify->link_index > adev->dm.dc->link_count) {
-		DRM_ERROR("DMUB HPD index (%u)is abnormal", notify->link_index);
+		drm_err(adev_to_drm(adev), "DMUB HPD index (%u)is abnormal", notify->link_index);
+		return;
+	}
+
+	/* Skip DMUB HPD IRQ in suspend/resume. We will probe them later. */
+	if (notify->type == DMUB_NOTIFICATION_HPD && adev->in_suspend) {
+		drm_info(adev_to_drm(adev), "Skip DMUB HPD IRQ callback in suspend/resume\n");
 		return;
 	}
 
@@ -711,14 +863,18 @@ static void dmub_hpd_callback(struct amdgpu_device *adev,
 
 	drm_connector_list_iter_begin(dev, &iter);
 	drm_for_each_connector_iter(connector, &iter) {
+
+		if (connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+			continue;
+
 		aconnector = to_amdgpu_dm_connector(connector);
 		if (link && aconnector->dc_link == link) {
 			if (notify->type == DMUB_NOTIFICATION_HPD)
-				DRM_INFO("DMUB HPD callback: link_index=%u\n", link_index);
+				drm_info(adev_to_drm(adev), "DMUB HPD IRQ callback: link_index=%u\n", link_index);
 			else if (notify->type == DMUB_NOTIFICATION_HPD_IRQ)
-				DRM_INFO("DMUB HPD IRQ callback: link_index=%u\n", link_index);
+				drm_info(adev_to_drm(adev), "DMUB HPD RX IRQ callback: link_index=%u\n", link_index);
 			else
-				DRM_WARN("DMUB Unknown HPD callback type %d, link_index=%u\n",
+				drm_warn(adev_to_drm(adev), "DMUB Unknown HPD callback type %d, link_index=%u\n",
 						notify->type, link_index);
 
 			hpd_aconnector = aconnector;
@@ -728,14 +884,31 @@ static void dmub_hpd_callback(struct amdgpu_device *adev,
 	drm_connector_list_iter_end(&iter);
 
 	if (hpd_aconnector) {
-		if (notify->type == DMUB_NOTIFICATION_HPD)
+		if (notify->type == DMUB_NOTIFICATION_HPD) {
+			if (hpd_aconnector->dc_link->hpd_status == (notify->hpd_status == DP_HPD_PLUG))
+				drm_warn(adev_to_drm(adev), "DMUB reported hpd status unchanged. link_index=%u\n", link_index);
 			handle_hpd_irq_helper(hpd_aconnector);
-		else if (notify->type == DMUB_NOTIFICATION_HPD_IRQ)
+		} else if (notify->type == DMUB_NOTIFICATION_HPD_IRQ) {
 			handle_hpd_rx_irq(hpd_aconnector);
+		}
 	}
 }
 
 /**
+ * dmub_hpd_sense_callback - DMUB HPD sense processing callback.
+ * @adev: amdgpu_device pointer
+ * @notify: dmub notification structure
+ *
+ * HPD sense changes can occur during low power states and need to be
+ * notified from firmware to driver.
+ */
+static void dmub_hpd_sense_callback(struct amdgpu_device *adev,
+			      struct dmub_notification *notify)
+{
+	drm_dbg_driver(adev_to_drm(adev), "DMUB HPD SENSE callback.\n");
+}
+
+/**
  * register_dmub_notify_callback - Sets callback for DMUB notify
  * @adev: amdgpu_device pointer
  * @type: Type of dmub notification
@@ -768,7 +941,7 @@ static void dm_handle_hpd_work(struct work_struct *work)
 	dmub_hpd_wrk = container_of(work, struct dmub_hpd_work, handle_hpd_work);
 
 	if (!dmub_hpd_wrk->dmub_notify) {
-		DRM_ERROR("dmub_hpd_wrk dmub_notify is NULL");
+		drm_err(adev_to_drm(dmub_hpd_wrk->adev), "dmub_hpd_wrk dmub_notify is NULL");
 		return;
 	}
 
@@ -782,6 +955,30 @@ static void dm_handle_hpd_work(struct work_struct *work)
 
 }
 
+static const char *dmub_notification_type_str(enum dmub_notification_type e)
+{
+	switch (e) {
+	case DMUB_NOTIFICATION_NO_DATA:
+		return "NO_DATA";
+	case DMUB_NOTIFICATION_AUX_REPLY:
+		return "AUX_REPLY";
+	case DMUB_NOTIFICATION_HPD:
+		return "HPD";
+	case DMUB_NOTIFICATION_HPD_IRQ:
+		return "HPD_IRQ";
+	case DMUB_NOTIFICATION_SET_CONFIG_REPLY:
+		return "SET_CONFIG_REPLY";
+	case DMUB_NOTIFICATION_DPIA_NOTIFICATION:
+		return "DPIA_NOTIFICATION";
+	case DMUB_NOTIFICATION_HPD_SENSE_NOTIFY:
+		return "HPD_SENSE_NOTIFY";
+	case DMUB_NOTIFICATION_FUSED_IO:
+		return "FUSED_IO";
+	default:
+		return "<unknown>";
+	}
+}
+
 #define DMUB_TRACE_MAX_READ 64
 /**
  * dm_dmub_outbox1_low_irq() - Handles Outbox interrupt
@@ -792,14 +989,30 @@ static void dm_handle_hpd_work(struct work_struct *work)
  */
 static void dm_dmub_outbox1_low_irq(void *interrupt_params)
 {
-	struct dmub_notification notify;
+	struct dmub_notification notify = {0};
 	struct common_irq_params *irq_params = interrupt_params;
 	struct amdgpu_device *adev = irq_params->adev;
 	struct amdgpu_display_manager *dm = &adev->dm;
 	struct dmcub_trace_buf_entry entry = { 0 };
 	u32 count = 0;
 	struct dmub_hpd_work *dmub_hpd_wrk;
-	struct dc_link *plink = NULL;
+
+	do {
+		if (dc_dmub_srv_get_dmub_outbox0_msg(dm->dc, &entry)) {
+			trace_amdgpu_dmub_trace_high_irq(entry.trace_code, entry.tick_count,
+							entry.param0, entry.param1);
+
+			drm_dbg_driver(adev_to_drm(adev), "trace_code:%u, tick_count:%u, param0:%u, param1:%u\n",
+				 entry.trace_code, entry.tick_count, entry.param0, entry.param1);
+		} else
+			break;
+
+		count++;
+
+	} while (count <= DMUB_TRACE_MAX_READ);
+
+	if (count > DMUB_TRACE_MAX_READ)
+		drm_dbg_driver(adev_to_drm(adev), "Warning : count > DMUB_TRACE_MAX_READ");
 
 	if (dc_enable_dmub_notifications(adev->dm.dc) &&
 		irq_params->irq_src == DC_IRQ_SOURCE_DMCUB_OUTBOX) {
@@ -807,81 +1020,56 @@ static void dm_dmub_outbox1_low_irq(void *interrupt_params)
 		do {
 			dc_stat_get_dmub_notification(adev->dm.dc, &notify);
 			if (notify.type >= ARRAY_SIZE(dm->dmub_thread_offload)) {
-				DRM_ERROR("DM: notify type %d invalid!", notify.type);
+				drm_err(adev_to_drm(adev), "DM: notify type %d invalid!", notify.type);
 				continue;
 			}
 			if (!dm->dmub_callback[notify.type]) {
-				DRM_DEBUG_DRIVER("DMUB notification skipped, no handler: type=%d\n", notify.type);
+				drm_warn(adev_to_drm(adev), "DMUB notification skipped due to no handler: type=%s\n",
+					dmub_notification_type_str(notify.type));
 				continue;
 			}
 			if (dm->dmub_thread_offload[notify.type] == true) {
 				dmub_hpd_wrk = kzalloc(sizeof(*dmub_hpd_wrk), GFP_ATOMIC);
 				if (!dmub_hpd_wrk) {
-					DRM_ERROR("Failed to allocate dmub_hpd_wrk");
+					drm_err(adev_to_drm(adev), "Failed to allocate dmub_hpd_wrk");
 					return;
 				}
 				dmub_hpd_wrk->dmub_notify = kmemdup(&notify, sizeof(struct dmub_notification),
 								    GFP_ATOMIC);
 				if (!dmub_hpd_wrk->dmub_notify) {
 					kfree(dmub_hpd_wrk);
-					DRM_ERROR("Failed to allocate dmub_hpd_wrk->dmub_notify");
+					drm_err(adev_to_drm(adev), "Failed to allocate dmub_hpd_wrk->dmub_notify");
 					return;
 				}
 				INIT_WORK(&dmub_hpd_wrk->handle_hpd_work, dm_handle_hpd_work);
 				dmub_hpd_wrk->adev = adev;
-				if (notify.type == DMUB_NOTIFICATION_HPD) {
-					plink = adev->dm.dc->links[notify.link_index];
-					if (plink) {
-						plink->hpd_status =
-							notify.hpd_status == DP_HPD_PLUG;
-					}
-				}
 				queue_work(adev->dm.delayed_hpd_wq, &dmub_hpd_wrk->handle_hpd_work);
 			} else {
 				dm->dmub_callback[notify.type](adev, &notify);
 			}
 		} while (notify.pending_notification);
 	}
-
-
-	do {
-		if (dc_dmub_srv_get_dmub_outbox0_msg(dm->dc, &entry)) {
-			trace_amdgpu_dmub_trace_high_irq(entry.trace_code, entry.tick_count,
-							entry.param0, entry.param1);
-
-			DRM_DEBUG_DRIVER("trace_code:%u, tick_count:%u, param0:%u, param1:%u\n",
-				 entry.trace_code, entry.tick_count, entry.param0, entry.param1);
-		} else
-			break;
-
-		count++;
-
-	} while (count <= DMUB_TRACE_MAX_READ);
-
-	if (count > DMUB_TRACE_MAX_READ)
-		DRM_DEBUG_DRIVER("Warning : count > DMUB_TRACE_MAX_READ");
 }
 
-static int dm_set_clockgating_state(void *handle,
+static int dm_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 		  enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int dm_set_powergating_state(void *handle,
+static int dm_set_powergating_state(struct amdgpu_ip_block *ip_block,
 		  enum amd_powergating_state state)
 {
 	return 0;
 }
 
 /* Prototypes of private functions */
-static int dm_early_init(void* handle);
+static int dm_early_init(struct amdgpu_ip_block *ip_block);
 
 /* Allocate memory for FBC compressed data  */
 static void amdgpu_dm_fbc_init(struct drm_connector *connector)
 {
-	struct drm_device *dev = connector->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
+	struct amdgpu_device *adev = drm_to_adev(connector->dev);
 	struct dm_compressor_info *compressor = &adev->dm.compressor;
 	struct amdgpu_dm_connector *aconn = to_amdgpu_dm_connector(connector);
 	struct drm_display_mode *mode;
@@ -898,8 +1086,8 @@ static void amdgpu_dm_fbc_init(struct drm_connector *connector)
 
 
 	list_for_each_entry(mode, &connector->modes, head) {
-		if (max_size < mode->htotal * mode->vtotal)
-			max_size = mode->htotal * mode->vtotal;
+		if (max_size < (unsigned long) mode->htotal * mode->vtotal)
+			max_size = (unsigned long) mode->htotal * mode->vtotal;
 	}
 
 	if (max_size) {
@@ -908,10 +1096,10 @@ static void amdgpu_dm_fbc_init(struct drm_connector *connector)
 			    &compressor->gpu_addr, &compressor->cpu_addr);
 
 		if (r)
-			DRM_ERROR("DM: Failed to initialize FBC\n");
+			drm_err(adev_to_drm(adev), "DM: Failed to initialize FBC\n");
 		else {
 			adev->dm.dc->ctx->fbc_gpu_addr = compressor->gpu_addr;
-			DRM_INFO("DM: FBC alloc %lu\n", max_size*4);
+			drm_info(adev_to_drm(adev), "DM: FBC alloc %lu\n", max_size*4);
 		}
 
 	}
@@ -935,13 +1123,19 @@ static int amdgpu_dm_audio_component_get_eld(struct device *kdev, int port,
 
 	drm_connector_list_iter_begin(dev, &conn_iter);
 	drm_for_each_connector_iter(connector, &conn_iter) {
+
+		if (connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+			continue;
+
 		aconnector = to_amdgpu_dm_connector(connector);
 		if (aconnector->audio_inst != port)
 			continue;
 
 		*enabled = true;
+		mutex_lock(&connector->eld_mutex);
 		ret = drm_eld_size(connector->eld);
 		memcpy(buf, connector->eld, min(max_bytes, ret));
+		mutex_unlock(&connector->eld_mutex);
 
 		break;
 	}
@@ -975,8 +1169,7 @@ static int amdgpu_dm_audio_component_bind(struct device *kdev,
 static void amdgpu_dm_audio_component_unbind(struct device *kdev,
 					  struct device *hda_kdev, void *data)
 {
-	struct drm_device *dev = dev_get_drvdata(kdev);
-	struct amdgpu_device *adev = drm_to_adev(dev);
+	struct amdgpu_device *adev = drm_to_adev(dev_get_drvdata(kdev));
 	struct drm_audio_component *acomp = data;
 
 	acomp->ops = NULL;
@@ -1059,6 +1252,7 @@ static int dm_dmub_hw_init(struct amdgpu_device *adev)
 	const struct firmware *dmub_fw = adev->dm.dmub_fw;
 	struct dmcu *dmcu = adev->dm.dc->res_pool->dmcu;
 	struct abm *abm = adev->dm.dc->res_pool->abm;
+	struct dc_context *ctx = adev->dm.dc->ctx;
 	struct dmub_srv_hw_params hw_params;
 	enum dmub_status status;
 	const unsigned char *fw_inst_const, *fw_bss_data;
@@ -1070,31 +1264,35 @@ static int dm_dmub_hw_init(struct amdgpu_device *adev)
 		return 0;
 
 	if (!fb_info) {
-		DRM_ERROR("No framebuffer info for DMUB service.\n");
+		drm_err(adev_to_drm(adev), "No framebuffer info for DMUB service.\n");
 		return -EINVAL;
 	}
 
 	if (!dmub_fw) {
 		/* Firmware required for DMUB support. */
-		DRM_ERROR("No firmware provided for DMUB.\n");
+		drm_err(adev_to_drm(adev), "No firmware provided for DMUB.\n");
 		return -EINVAL;
 	}
 
+	/* initialize register offsets for ASICs with runtime initialization available */
+	if (dmub_srv->hw_funcs.init_reg_offsets)
+		dmub_srv->hw_funcs.init_reg_offsets(dmub_srv, ctx);
+
 	status = dmub_srv_has_hw_support(dmub_srv, &has_hw_support);
 	if (status != DMUB_STATUS_OK) {
-		DRM_ERROR("Error checking HW support for DMUB: %d\n", status);
+		drm_err(adev_to_drm(adev), "Error checking HW support for DMUB: %d\n", status);
 		return -EINVAL;
 	}
 
 	if (!has_hw_support) {
-		DRM_INFO("DMUB unsupported on ASIC\n");
+		drm_info(adev_to_drm(adev), "DMUB unsupported on ASIC\n");
 		return 0;
 	}
 
 	/* Reset DMCUB if it was previously running - before we overwrite its memory. */
 	status = dmub_srv_hw_reset(dmub_srv);
 	if (status != DMUB_STATUS_OK)
-		DRM_WARN("Error resetting DMUB HW: %d\n", status);
+		drm_warn(adev_to_drm(adev), "Error resetting DMUB HW: %d\n", status);
 
 	hdr = (const struct dmcub_firmware_header_v1_0 *)dmub_fw->data;
 
@@ -1140,6 +1338,9 @@ static int dm_dmub_hw_init(struct amdgpu_device *adev)
 	memset(fb_info->fb[DMUB_WINDOW_6_FW_STATE].cpu_addr, 0,
 	       fb_info->fb[DMUB_WINDOW_6_FW_STATE].size);
 
+	memset(fb_info->fb[DMUB_WINDOW_SHARED_STATE].cpu_addr, 0,
+	       fb_info->fb[DMUB_WINDOW_SHARED_STATE].size);
+
 	/* Initialize hardware. */
 	memset(&hw_params, 0, sizeof(hw_params));
 	hw_params.fb_base = adev->gmc.fb_start;
@@ -1155,9 +1356,13 @@ static int dm_dmub_hw_init(struct amdgpu_device *adev)
 	for (i = 0; i < fb_info->num_fb; ++i)
 		hw_params.fb[i] = &fb_info->fb[i];
 
-	switch (adev->ip_versions[DCE_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
 	case IP_VERSION(3, 1, 3):
 	case IP_VERSION(3, 1, 4):
+	case IP_VERSION(3, 5, 0):
+	case IP_VERSION(3, 5, 1):
+	case IP_VERSION(3, 6, 0):
+	case IP_VERSION(4, 0, 1):
 		hw_params.dpia_supported = true;
 		hw_params.disable_dpia = adev->dm.dc->debug.dpia_debug.bits.disable_dpia;
 		break;
@@ -1165,16 +1370,27 @@ static int dm_dmub_hw_init(struct amdgpu_device *adev)
 		break;
 	}
 
+	switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
+	case IP_VERSION(3, 5, 0):
+	case IP_VERSION(3, 5, 1):
+	case IP_VERSION(3, 6, 0):
+		hw_params.ips_sequential_ono = adev->external_rev_id > 0x10;
+		hw_params.lower_hbr3_phy_ssc = true;
+		break;
+	default:
+		break;
+	}
+
 	status = dmub_srv_hw_init(dmub_srv, &hw_params);
 	if (status != DMUB_STATUS_OK) {
-		DRM_ERROR("Error initializing DMUB HW: %d\n", status);
+		drm_err(adev_to_drm(adev), "Error initializing DMUB HW: %d\n", status);
 		return -EINVAL;
 	}
 
 	/* Wait for firmware load to finish. */
 	status = dmub_srv_wait_for_auto_load(dmub_srv, 100000);
 	if (status != DMUB_STATUS_OK)
-		DRM_WARN("Wait for DMUB auto-load failed: %d\n", status);
+		drm_warn(adev_to_drm(adev), "Wait for DMUB auto-load failed: %d\n", status);
 
 	/* Init DMCU and ABM if available. */
 	if (dmcu && abm) {
@@ -1185,13 +1401,36 @@ static int dm_dmub_hw_init(struct amdgpu_device *adev)
 	if (!adev->dm.dc->ctx->dmub_srv)
 		adev->dm.dc->ctx->dmub_srv = dc_dmub_srv_create(adev->dm.dc, dmub_srv);
 	if (!adev->dm.dc->ctx->dmub_srv) {
-		DRM_ERROR("Couldn't allocate DC DMUB server!\n");
+		drm_err(adev_to_drm(adev), "Couldn't allocate DC DMUB server!\n");
 		return -ENOMEM;
 	}
 
-	DRM_INFO("DMUB hardware initialized: version=0x%08X\n",
+	drm_info(adev_to_drm(adev), "DMUB hardware initialized: version=0x%08X\n",
 		 adev->dm.dmcub_fw_version);
 
+	/* Keeping sanity checks off if
+	 * DCN31 >= 4.0.59.0
+	 * DCN314 >= 8.0.16.0
+	 * Otherwise, turn on sanity checks
+	 */
+	switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
+	case IP_VERSION(3, 1, 2):
+	case IP_VERSION(3, 1, 3):
+		if (adev->dm.dmcub_fw_version &&
+			adev->dm.dmcub_fw_version >= DMUB_FW_VERSION(4, 0, 0) &&
+			adev->dm.dmcub_fw_version < DMUB_FW_VERSION(4, 0, 59))
+				adev->dm.dc->debug.sanity_checks = true;
+		break;
+	case IP_VERSION(3, 1, 4):
+		if (adev->dm.dmcub_fw_version &&
+			adev->dm.dmcub_fw_version >= DMUB_FW_VERSION(4, 0, 0) &&
+			adev->dm.dmcub_fw_version < DMUB_FW_VERSION(8, 0, 16))
+				adev->dm.dc->debug.sanity_checks = true;
+		break;
+	default:
+		break;
+	}
+
 	return 0;
 }
 
@@ -1200,6 +1439,7 @@ static void dm_dmub_hw_resume(struct amdgpu_device *adev)
 	struct dmub_srv *dmub_srv = adev->dm.dmub_srv;
 	enum dmub_status status;
 	bool init;
+	int r;
 
 	if (!dmub_srv) {
 		/* DMUB isn't supported on the ASIC. */
@@ -1208,16 +1448,18 @@ static void dm_dmub_hw_resume(struct amdgpu_device *adev)
 
 	status = dmub_srv_is_hw_init(dmub_srv, &init);
 	if (status != DMUB_STATUS_OK)
-		DRM_WARN("DMUB hardware init check failed: %d\n", status);
+		drm_warn(adev_to_drm(adev), "DMUB hardware init check failed: %d\n", status);
 
 	if (status == DMUB_STATUS_OK && init) {
 		/* Wait for firmware load to finish. */
 		status = dmub_srv_wait_for_auto_load(dmub_srv, 100000);
 		if (status != DMUB_STATUS_OK)
-			DRM_WARN("Wait for DMUB auto-load failed: %d\n", status);
+			drm_warn(adev_to_drm(adev), "Wait for DMUB auto-load failed: %d\n", status);
 	} else {
 		/* Perform the full hardware initialization. */
-		dm_dmub_hw_init(adev);
+		r = dm_dmub_hw_init(adev);
+		if (r)
+			drm_err(adev_to_drm(adev), "DMUB interface failed to initialize: status=%d\n", r);
 	}
 }
 
@@ -1236,9 +1478,11 @@ static void mmhub_read_system_context(struct amdgpu_device *adev, struct dc_phy_
 	agp_top = adev->gmc.agp_end >> 24;
 
 	/* AGP aperture is disabled */
-	if (agp_bot == agp_top) {
+	if (agp_bot > agp_top) {
 		logical_addr_low = adev->gmc.fb_start >> 18;
-		if (adev->apu_flags & AMD_APU_IS_RAVEN2)
+		if (adev->apu_flags & (AMD_APU_IS_RAVEN2 |
+				       AMD_APU_IS_RENOIR |
+				       AMD_APU_IS_GREEN_SARDINE))
 			/*
 			 * Raven2 has a HW issue that it is unable to use the vram which
 			 * is out of MC_VM_SYSTEM_APERTURE_HIGH_ADDR. So here is the
@@ -1250,7 +1494,9 @@ static void mmhub_read_system_context(struct amdgpu_device *adev, struct dc_phy_
 			logical_addr_high = adev->gmc.fb_end >> 18;
 	} else {
 		logical_addr_low = min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18;
-		if (adev->apu_flags & AMD_APU_IS_RAVEN2)
+		if (adev->apu_flags & (AMD_APU_IS_RAVEN2 |
+				       AMD_APU_IS_RENOIR |
+				       AMD_APU_IS_GREEN_SARDINE))
 			/*
 			 * Raven2 has a HW issue that it is unable to use the vram which
 			 * is out of MC_VM_SYSTEM_APERTURE_HIGH_ADDR. So here is the
@@ -1264,17 +1510,21 @@ static void mmhub_read_system_context(struct amdgpu_device *adev, struct dc_phy_
 
 	pt_base = amdgpu_gmc_pd_addr(adev->gart.bo);
 
-	page_table_start.high_part = (u32)(adev->gmc.gart_start >> 44) & 0xF;
-	page_table_start.low_part = (u32)(adev->gmc.gart_start >> 12);
-	page_table_end.high_part = (u32)(adev->gmc.gart_end >> 44) & 0xF;
-	page_table_end.low_part = (u32)(adev->gmc.gart_end >> 12);
-	page_table_base.high_part = upper_32_bits(pt_base) & 0xF;
+	page_table_start.high_part = upper_32_bits(adev->gmc.gart_start >>
+						   AMDGPU_GPU_PAGE_SHIFT);
+	page_table_start.low_part = lower_32_bits(adev->gmc.gart_start >>
+						  AMDGPU_GPU_PAGE_SHIFT);
+	page_table_end.high_part = upper_32_bits(adev->gmc.gart_end >>
+						 AMDGPU_GPU_PAGE_SHIFT);
+	page_table_end.low_part = lower_32_bits(adev->gmc.gart_end >>
+						AMDGPU_GPU_PAGE_SHIFT);
+	page_table_base.high_part = upper_32_bits(pt_base);
 	page_table_base.low_part = lower_32_bits(pt_base);
 
 	pa_config->system_aperture.start_addr = (uint64_t)logical_addr_low << 18;
 	pa_config->system_aperture.end_addr = (uint64_t)logical_addr_high << 18;
 
-	pa_config->system_aperture.agp_base = (uint64_t)agp_base << 24 ;
+	pa_config->system_aperture.agp_base = (uint64_t)agp_base << 24;
 	pa_config->system_aperture.agp_bot = (uint64_t)agp_bot << 24;
 	pa_config->system_aperture.agp_top = (uint64_t)agp_top << 24;
 
@@ -1319,18 +1569,18 @@ static void dm_handle_hpd_rx_offload_work(struct work_struct *work)
 
 	offload_work = container_of(work, struct hpd_rx_irq_offload_work, work);
 	aconnector = offload_work->offload_wq->aconnector;
+	adev = offload_work->adev;
 
 	if (!aconnector) {
-		DRM_ERROR("Can't retrieve aconnector in hpd_rx_irq_offload_work");
+		drm_err(adev_to_drm(adev), "Can't retrieve aconnector in hpd_rx_irq_offload_work");
 		goto skip;
 	}
 
-	adev = drm_to_adev(aconnector->base.dev);
 	dc_link = aconnector->dc_link;
 
 	mutex_lock(&aconnector->hpd_lock);
 	if (!dc_link_detect_connection_type(dc_link, &new_connection_type))
-		DRM_ERROR("KMS: Failed to detect connector\n");
+		drm_err(adev_to_drm(adev), "KMS: Failed to detect connector\n");
 	mutex_unlock(&aconnector->hpd_lock);
 
 	if (new_connection_type == dc_connection_none)
@@ -1339,6 +1589,15 @@ static void dm_handle_hpd_rx_offload_work(struct work_struct *work)
 	if (amdgpu_in_reset(adev))
 		goto skip;
 
+	if (offload_work->data.bytes.device_service_irq.bits.UP_REQ_MSG_RDY ||
+		offload_work->data.bytes.device_service_irq.bits.DOWN_REP_MSG_RDY) {
+		dm_handle_mst_sideband_msg_ready_event(&aconnector->mst_mgr, DOWN_OR_UP_MSG_RDY_EVENT);
+		spin_lock_irqsave(&offload_work->offload_wq->offload_lock, flags);
+		offload_work->offload_wq->is_handling_mst_msg_rdy_event = false;
+		spin_unlock_irqrestore(&offload_work->offload_wq->offload_lock, flags);
+		goto skip;
+	}
+
 	mutex_lock(&adev->dm.dc_lock);
 	if (offload_work->data.bytes.device_service_irq.bits.AUTOMATED_TEST) {
 		dc_link_dp_handle_automated_test(dc_link);
@@ -1357,8 +1616,7 @@ static void dm_handle_hpd_rx_offload_work(struct work_struct *work)
 		DP_TEST_RESPONSE,
 		&test_response.raw,
 		sizeof(test_response));
-	}
-	else if ((dc_link->connector_signal != SIGNAL_TYPE_EDP) &&
+	} else if ((dc_link->connector_signal != SIGNAL_TYPE_EDP) &&
 			dc_link_check_link_loss_status(dc_link, &offload_work->data) &&
 			dc_link_dp_allow_hpd_rx_irq(dc_link)) {
 		/* offload_work->data is from handle_hpd_rx_irq->
@@ -1391,8 +1649,9 @@ skip:
 
 }
 
-static struct hpd_rx_irq_offload_work_queue *hpd_rx_irq_create_workqueue(struct dc *dc)
+static struct hpd_rx_irq_offload_work_queue *hpd_rx_irq_create_workqueue(struct amdgpu_device *adev)
 {
+	struct dc *dc = adev->dm.dc;
 	int max_caps = dc->caps.max_links;
 	int i = 0;
 	struct hpd_rx_irq_offload_work_queue *hpd_rx_offload_wq = NULL;
@@ -1408,7 +1667,7 @@ static struct hpd_rx_irq_offload_work_queue *hpd_rx_irq_create_workqueue(struct
 				    create_singlethread_workqueue("amdgpu_dm_hpd_rx_offload_wq");
 
 		if (hpd_rx_offload_wq[i].wq == NULL) {
-			DRM_ERROR("create amdgpu_dm_hpd_rx_offload_wq fail!");
+			drm_err(adev_to_drm(adev), "create amdgpu_dm_hpd_rx_offload_wq fail!");
 			goto out_err;
 		}
 
@@ -1457,76 +1716,163 @@ static bool dm_should_disable_stutter(struct pci_dev *pdev)
 	return false;
 }
 
-static const struct dmi_system_id hpd_disconnect_quirk_table[] = {
-	{
-		.matches = {
-			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
-			DMI_MATCH(DMI_PRODUCT_NAME, "Precision 3660"),
-		},
-	},
-	{
-		.matches = {
-			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
-			DMI_MATCH(DMI_PRODUCT_NAME, "Precision 3260"),
-		},
-	},
-	{
-		.matches = {
-			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
-			DMI_MATCH(DMI_PRODUCT_NAME, "Precision 3460"),
-		},
-	},
-	{
-		.matches = {
-			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
-			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex Tower Plus 7010"),
-		},
-	},
-	{
-		.matches = {
-			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
-			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex Tower 7010"),
-		},
-	},
-	{
-		.matches = {
-			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
-			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex SFF Plus 7010"),
-		},
-	},
-	{
-		.matches = {
-			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
-			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex SFF 7010"),
-		},
-	},
-	{
-		.matches = {
-			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
-			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex Micro Plus 7010"),
-		},
-	},
-	{
-		.matches = {
-			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
-			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex Micro 7010"),
-		},
-	},
-	{}
-	/* TODO: refactor this from a fixed table to a dynamic option */
-};
 
-static void retrieve_dmi_info(struct amdgpu_display_manager *dm)
+void*
+dm_allocate_gpu_mem(
+		struct amdgpu_device *adev,
+		enum dc_gpu_mem_alloc_type type,
+		size_t size,
+		long long *addr)
+{
+	struct dal_allocation *da;
+	u32 domain = (type == DC_MEM_ALLOC_TYPE_GART) ?
+		AMDGPU_GEM_DOMAIN_GTT : AMDGPU_GEM_DOMAIN_VRAM;
+	int ret;
+
+	da = kzalloc(sizeof(struct dal_allocation), GFP_KERNEL);
+	if (!da)
+		return NULL;
+
+	ret = amdgpu_bo_create_kernel(adev, size, PAGE_SIZE,
+				      domain, &da->bo,
+				      &da->gpu_addr, &da->cpu_ptr);
+
+	*addr = da->gpu_addr;
+
+	if (ret) {
+		kfree(da);
+		return NULL;
+	}
+
+	/* add da to list in dm */
+	list_add(&da->list, &adev->dm.da_list);
+
+	return da->cpu_ptr;
+}
+
+void
+dm_free_gpu_mem(
+		struct amdgpu_device *adev,
+		enum dc_gpu_mem_alloc_type type,
+		void *pvMem)
 {
-	const struct dmi_system_id *dmi_id;
+	struct dal_allocation *da;
+
+	/* walk the da list in DM */
+	list_for_each_entry(da, &adev->dm.da_list, list) {
+		if (pvMem == da->cpu_ptr) {
+			amdgpu_bo_free_kernel(&da->bo, &da->gpu_addr, &da->cpu_ptr);
+			list_del(&da->list);
+			kfree(da);
+			break;
+		}
+	}
+
+}
+
+static enum dmub_status
+dm_dmub_send_vbios_gpint_command(struct amdgpu_device *adev,
+				 enum dmub_gpint_command command_code,
+				 uint16_t param,
+				 uint32_t timeout_us)
+{
+	union dmub_gpint_data_register reg, test;
+	uint32_t i;
+
+	/* Assume that VBIOS DMUB is ready to take commands */
+
+	reg.bits.status = 1;
+	reg.bits.command_code = command_code;
+	reg.bits.param = param;
+
+	cgs_write_register(adev->dm.cgs_device, 0x34c0 + 0x01f8, reg.all);
 
-	dm->aux_hpd_discon_quirk = false;
+	for (i = 0; i < timeout_us; ++i) {
+		udelay(1);
 
-	dmi_id = dmi_first_match(hpd_disconnect_quirk_table);
-	if (dmi_id) {
-		dm->aux_hpd_discon_quirk = true;
-		DRM_INFO("aux_hpd_discon_quirk attached\n");
+		/* Check if our GPINT got acked */
+		reg.bits.status = 0;
+		test = (union dmub_gpint_data_register)
+			cgs_read_register(adev->dm.cgs_device, 0x34c0 + 0x01f8);
+
+		if (test.all == reg.all)
+			return DMUB_STATUS_OK;
 	}
+
+	return DMUB_STATUS_TIMEOUT;
+}
+
+static void *dm_dmub_get_vbios_bounding_box(struct amdgpu_device *adev)
+{
+	void *bb;
+	long long addr;
+	unsigned int bb_size;
+	int i = 0;
+	uint16_t chunk;
+	enum dmub_gpint_command send_addrs[] = {
+		DMUB_GPINT__SET_BB_ADDR_WORD0,
+		DMUB_GPINT__SET_BB_ADDR_WORD1,
+		DMUB_GPINT__SET_BB_ADDR_WORD2,
+		DMUB_GPINT__SET_BB_ADDR_WORD3,
+	};
+	enum dmub_status ret;
+
+	switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
+	case IP_VERSION(4, 0, 1):
+		bb_size = sizeof(struct dml2_soc_bb);
+		break;
+	default:
+		return NULL;
+	}
+
+	bb =  dm_allocate_gpu_mem(adev,
+				  DC_MEM_ALLOC_TYPE_GART,
+				  bb_size,
+				  &addr);
+	if (!bb)
+		return NULL;
+
+	for (i = 0; i < 4; i++) {
+		/* Extract 16-bit chunk */
+		chunk = ((uint64_t) addr >> (i * 16)) & 0xFFFF;
+		/* Send the chunk */
+		ret = dm_dmub_send_vbios_gpint_command(adev, send_addrs[i], chunk, 30000);
+		if (ret != DMUB_STATUS_OK)
+			goto free_bb;
+	}
+
+	/* Now ask DMUB to copy the bb */
+	ret = dm_dmub_send_vbios_gpint_command(adev, DMUB_GPINT__BB_COPY, 1, 200000);
+	if (ret != DMUB_STATUS_OK)
+		goto free_bb;
+
+	return bb;
+
+free_bb:
+	dm_free_gpu_mem(adev, DC_MEM_ALLOC_TYPE_GART, (void *) bb);
+	return NULL;
+
+}
+
+static enum dmub_ips_disable_type dm_get_default_ips_mode(
+	struct amdgpu_device *adev)
+{
+	enum dmub_ips_disable_type ret = DMUB_IPS_ENABLE;
+
+	switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
+	case IP_VERSION(3, 5, 0):
+	case IP_VERSION(3, 6, 0):
+	case IP_VERSION(3, 5, 1):
+		ret =  DMUB_IPS_RCG_IN_ACTIVE_IPS2_IN_OFF;
+		break;
+	default:
+		/* ASICs older than DCN35 do not have IPSs */
+		if (amdgpu_ip_version(adev, DCE_HWIP, 0) < IP_VERSION(3, 5, 0))
+			ret = DMUB_IPS_DISABLE_ALL;
+		break;
+	}
+
+	return ret;
 }
 
 static int amdgpu_dm_init(struct amdgpu_device *adev)
@@ -1546,8 +1892,8 @@ static int amdgpu_dm_init(struct amdgpu_device *adev)
 	mutex_init(&adev->dm.dc_lock);
 	mutex_init(&adev->dm.audio_lock);
 
-	if(amdgpu_dm_irq_init(adev)) {
-		DRM_ERROR("amdgpu: failed to initialize DM IRQ support.\n");
+	if (amdgpu_dm_irq_init(adev)) {
+		drm_err(adev_to_drm(adev), "failed to initialize DM IRQ support.\n");
 		goto error;
 	}
 
@@ -1564,18 +1910,12 @@ static int amdgpu_dm_init(struct amdgpu_device *adev)
 
 	init_data.driver = adev;
 
-	adev->dm.cgs_device = amdgpu_cgs_create_device(adev);
-
-	if (!adev->dm.cgs_device) {
-		DRM_ERROR("amdgpu: failed to create cgs device.\n");
-		goto error;
-	}
-
+	/* cgs_device was created in dm_sw_init() */
 	init_data.cgs_device = adev->dm.cgs_device;
 
 	init_data.dce_environment = DCE_ENV_PRODUCTION_DRV;
 
-	switch (adev->ip_versions[DCE_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
 	case IP_VERSION(2, 1, 0):
 		switch (adev->dm.dmcub_fw_version) {
 		case 0: /* development */
@@ -1594,40 +1934,27 @@ static int amdgpu_dm_init(struct amdgpu_device *adev)
 		break;
 	}
 
-	switch (adev->asic_type) {
-	case CHIP_CARRIZO:
-	case CHIP_STONEY:
-		init_data.flags.gpu_vm_support = true;
-		break;
-	default:
-		switch (adev->ip_versions[DCE_HWIP][0]) {
-		case IP_VERSION(1, 0, 0):
-		case IP_VERSION(1, 0, 1):
-			/* enable S/G on PCO and RV2 */
-			if ((adev->apu_flags & AMD_APU_IS_RAVEN2) ||
-			    (adev->apu_flags & AMD_APU_IS_PICASSO))
-				init_data.flags.gpu_vm_support = true;
-			break;
-		case IP_VERSION(2, 1, 0):
-		case IP_VERSION(3, 0, 1):
-		case IP_VERSION(3, 1, 2):
-		case IP_VERSION(3, 1, 3):
-		case IP_VERSION(3, 1, 4):
-		case IP_VERSION(3, 1, 5):
-		case IP_VERSION(3, 1, 6):
-			init_data.flags.gpu_vm_support = true;
-			break;
-		default:
-			break;
-		}
-		break;
-	}
-	if (init_data.flags.gpu_vm_support &&
-	    (amdgpu_sg_display == 0))
+	/* APU support S/G display by default except:
+	 * ASICs before Carrizo,
+	 * RAVEN1 (Users reported stability issue)
+	 */
+
+	if (adev->asic_type < CHIP_CARRIZO) {
 		init_data.flags.gpu_vm_support = false;
+	} else if (adev->asic_type == CHIP_RAVEN) {
+		if (adev->apu_flags & AMD_APU_IS_RAVEN)
+			init_data.flags.gpu_vm_support = false;
+		else
+			init_data.flags.gpu_vm_support = (amdgpu_sg_display != 0);
+	} else {
+		if (amdgpu_ip_version(adev, DCE_HWIP, 0) == IP_VERSION(2, 0, 3))
+			init_data.flags.gpu_vm_support = (amdgpu_sg_display == 1);
+		else
+			init_data.flags.gpu_vm_support =
+				(amdgpu_sg_display != 0) && (adev->flags & AMD_IS_APU);
+	}
 
-	if (init_data.flags.gpu_vm_support)
-		adev->mode_info.gpu_vm_support = true;
+	adev->mode_info.gpu_vm_support = init_data.flags.gpu_vm_support;
 
 	if (amdgpu_dc_feature_mask & DC_FBC_MASK)
 		init_data.flags.fbc_support = true;
@@ -1646,35 +1973,58 @@ static int amdgpu_dm_init(struct amdgpu_device *adev)
 	if (amdgpu_dc_feature_mask & DC_DISABLE_LTTPR_DP2_0)
 		init_data.flags.allow_lttpr_non_transparent_mode.bits.DP2_0 = true;
 
-	/* Disable SubVP + DRR config by default */
-	init_data.flags.disable_subvp_drr = true;
-	if (amdgpu_dc_feature_mask & DC_ENABLE_SUBVP_DRR)
-		init_data.flags.disable_subvp_drr = false;
-
 	init_data.flags.seamless_boot_edp_requested = false;
 
-	if (check_seamless_boot_capability(adev)) {
+	if (amdgpu_device_seamless_boot_supported(adev)) {
 		init_data.flags.seamless_boot_edp_requested = true;
 		init_data.flags.allow_seamless_boot_optimization = true;
-		DRM_INFO("Seamless boot condition check passed\n");
+		drm_dbg(adev->dm.ddev, "Seamless boot requested\n");
 	}
 
 	init_data.flags.enable_mipi_converter_optimization = true;
 
 	init_data.dcn_reg_offsets = adev->reg_offset[DCE_HWIP][0];
 	init_data.nbio_reg_offsets = adev->reg_offset[NBIO_HWIP][0];
+	init_data.clk_reg_offsets = adev->reg_offset[CLK_HWIP][0];
+
+	if (amdgpu_dc_debug_mask & DC_DISABLE_IPS)
+		init_data.flags.disable_ips = DMUB_IPS_DISABLE_ALL;
+	else if (amdgpu_dc_debug_mask & DC_DISABLE_IPS_DYNAMIC)
+		init_data.flags.disable_ips = DMUB_IPS_DISABLE_DYNAMIC;
+	else if (amdgpu_dc_debug_mask & DC_DISABLE_IPS2_DYNAMIC)
+		init_data.flags.disable_ips = DMUB_IPS_RCG_IN_ACTIVE_IPS2_IN_OFF;
+	else if (amdgpu_dc_debug_mask & DC_FORCE_IPS_ENABLE)
+		init_data.flags.disable_ips = DMUB_IPS_ENABLE;
+	else
+		init_data.flags.disable_ips = dm_get_default_ips_mode(adev);
 
-	INIT_LIST_HEAD(&adev->dm.da_list);
+	init_data.flags.disable_ips_in_vpb = 0;
+
+	/* DCN35 and above supports dynamic DTBCLK switch */
+	if (amdgpu_ip_version(adev, DCE_HWIP, 0) >= IP_VERSION(3, 5, 0))
+		init_data.flags.allow_0_dtb_clk = true;
+
+	/* Enable DWB for tested platforms only */
+	if (amdgpu_ip_version(adev, DCE_HWIP, 0) >= IP_VERSION(3, 0, 0))
+		init_data.num_virtual_links = 1;
 
 	retrieve_dmi_info(&adev->dm);
+	if (adev->dm.edp0_on_dp1_quirk)
+		init_data.flags.support_edp0_on_dp1 = true;
+
+	if (adev->dm.bb_from_dmub)
+		init_data.bb_from_dmub = adev->dm.bb_from_dmub;
+	else
+		init_data.bb_from_dmub = NULL;
 
 	/* Display Core create. */
 	adev->dm.dc = dc_create(&init_data);
 
 	if (adev->dm.dc) {
-		DRM_INFO("Display Core initialized with v%s!\n", DC_VER);
+		drm_info(adev_to_drm(adev), "Display Core v%s initialized on %s\n", DC_VER,
+			 dce_version_to_string(adev->dm.dc->ctx->dce_version));
 	} else {
-		DRM_INFO("Display Core failed to initialize with v%s!\n", DC_VER);
+		drm_info(adev_to_drm(adev), "Display Core failed to initialize with v%s!\n", DC_VER);
 		goto error;
 	}
 
@@ -1691,9 +2041,8 @@ static int amdgpu_dm_init(struct amdgpu_device *adev)
 	if (amdgpu_dc_debug_mask & DC_DISABLE_STUTTER)
 		adev->dm.dc->debug.disable_stutter = true;
 
-	if (amdgpu_dc_debug_mask & DC_DISABLE_DSC) {
+	if (amdgpu_dc_debug_mask & DC_DISABLE_DSC)
 		adev->dm.dc->debug.disable_dsc = true;
-	}
 
 	if (amdgpu_dc_debug_mask & DC_DISABLE_CLOCK_GATING)
 		adev->dm.dc->debug.disable_clock_gate = true;
@@ -1701,42 +2050,44 @@ static int amdgpu_dm_init(struct amdgpu_device *adev)
 	if (amdgpu_dc_debug_mask & DC_FORCE_SUBVP_MCLK_SWITCH)
 		adev->dm.dc->debug.force_subvp_mclk_switch = true;
 
+	if (amdgpu_dc_debug_mask & DC_DISABLE_SUBVP_FAMS) {
+		adev->dm.dc->debug.force_disable_subvp = true;
+		adev->dm.dc->debug.fams2_config.bits.enable = false;
+	}
+
+	if (amdgpu_dc_debug_mask & DC_ENABLE_DML2) {
+		adev->dm.dc->debug.using_dml2 = true;
+		adev->dm.dc->debug.using_dml21 = true;
+	}
+
+	if (amdgpu_dc_debug_mask & DC_HDCP_LC_FORCE_FW_ENABLE)
+		adev->dm.dc->debug.hdcp_lc_force_fw_enable = true;
+
+	if (amdgpu_dc_debug_mask & DC_HDCP_LC_ENABLE_SW_FALLBACK)
+		adev->dm.dc->debug.hdcp_lc_enable_sw_fallback = true;
+
+	if (amdgpu_dc_debug_mask & DC_SKIP_DETECTION_LT)
+		adev->dm.dc->debug.skip_detection_link_training = true;
+
 	adev->dm.dc->debug.visual_confirm = amdgpu_dc_visual_confirm;
 
 	/* TODO: Remove after DP2 receiver gets proper support of Cable ID feature */
 	adev->dm.dc->debug.ignore_cable_id = true;
 
-	/* TODO: There is a new drm mst change where the freedom of
-	 * vc_next_start_slot update is revoked/moved into drm, instead of in
-	 * driver. This forces us to make sure to get vc_next_start_slot updated
-	 * in drm function each time without considering if mst_state is active
-	 * or not. Otherwise, next time hotplug will give wrong start_slot
-	 * number. We are implementing a temporary solution to even notify drm
-	 * mst deallocation when link is no longer of MST type when uncommitting
-	 * the stream so we will have more time to work on a proper solution.
-	 * Ideally when dm_helpers_dp_mst_stop_top_mgr message is triggered, we
-	 * should notify drm to do a complete "reset" of its states and stop
-	 * calling further drm mst functions when link is no longer of an MST
-	 * type. This could happen when we unplug an MST hubs/displays. When
-	 * uncommit stream comes later after unplug, we should just reset
-	 * hardware states only.
-	 */
-	adev->dm.dc->debug.temp_mst_deallocation_sequence = true;
-
 	if (adev->dm.dc->caps.dp_hdmi21_pcon_support)
-		DRM_INFO("DP-HDMI FRL PCON supported\n");
+		drm_info(adev_to_drm(adev), "DP-HDMI FRL PCON supported\n");
 
 	r = dm_dmub_hw_init(adev);
 	if (r) {
-		DRM_ERROR("DMUB interface failed to initialize: status=%d\n", r);
+		drm_err(adev_to_drm(adev), "DMUB interface failed to initialize: status=%d\n", r);
 		goto error;
 	}
 
 	dc_hardware_init(adev->dm.dc);
 
-	adev->dm.hpd_rx_offload_wq = hpd_rx_irq_create_workqueue(adev->dm.dc);
+	adev->dm.hpd_rx_offload_wq = hpd_rx_irq_create_workqueue(adev);
 	if (!adev->dm.hpd_rx_offload_wq) {
-		DRM_ERROR("amdgpu: failed to create hpd rx offload workqueue.\n");
+		drm_err(adev_to_drm(adev), "failed to create hpd rx offload workqueue.\n");
 		goto error;
 	}
 
@@ -1751,10 +2102,10 @@ static int amdgpu_dm_init(struct amdgpu_device *adev)
 
 	adev->dm.freesync_module = mod_freesync_create(adev->dm.dc);
 	if (!adev->dm.freesync_module) {
-		DRM_ERROR(
-		"amdgpu: failed to initialize freesync_module.\n");
+		drm_err(adev_to_drm(adev),
+		"failed to initialize freesync_module.\n");
 	} else
-		DRM_DEBUG_DRIVER("amdgpu: freesync_module init done %p.\n",
+		drm_dbg_driver(adev_to_drm(adev), "amdgpu: freesync_module init done %p.\n",
 				adev->dm.freesync_module);
 
 	amdgpu_dm_init_color_mod();
@@ -1763,65 +2114,68 @@ static int amdgpu_dm_init(struct amdgpu_device *adev)
 		adev->dm.vblank_control_workqueue =
 			create_singlethread_workqueue("dm_vblank_control_workqueue");
 		if (!adev->dm.vblank_control_workqueue)
-			DRM_ERROR("amdgpu: failed to initialize vblank_workqueue.\n");
+			drm_err(adev_to_drm(adev), "failed to initialize vblank_workqueue.\n");
 	}
 
+	if (adev->dm.dc->caps.ips_support &&
+	    adev->dm.dc->config.disable_ips != DMUB_IPS_DISABLE_ALL)
+		adev->dm.idle_workqueue = idle_create_workqueue(adev);
+
 	if (adev->dm.dc->caps.max_links > 0 && adev->family >= AMDGPU_FAMILY_RV) {
 		adev->dm.hdcp_workqueue = hdcp_create_workqueue(adev, &init_params.cp_psp, adev->dm.dc);
 
 		if (!adev->dm.hdcp_workqueue)
-			DRM_ERROR("amdgpu: failed to initialize hdcp_workqueue.\n");
+			drm_err(adev_to_drm(adev), "failed to initialize hdcp_workqueue.\n");
 		else
-			DRM_DEBUG_DRIVER("amdgpu: hdcp_workqueue init done %p.\n", adev->dm.hdcp_workqueue);
+			drm_dbg_driver(adev_to_drm(adev), "amdgpu: hdcp_workqueue init done %p.\n", adev->dm.hdcp_workqueue);
 
 		dc_init_callbacks(adev->dm.dc, &init_params);
 	}
-#if defined(CONFIG_DRM_AMD_SECURE_DISPLAY)
-	adev->dm.secure_display_ctxs = amdgpu_dm_crtc_secure_display_create_contexts(adev);
-	if (!adev->dm.secure_display_ctxs) {
-		DRM_ERROR("amdgpu: failed to initialize secure_display_ctxs.\n");
-	}
-#endif
 	if (dc_is_dmub_outbox_supported(adev->dm.dc)) {
 		init_completion(&adev->dm.dmub_aux_transfer_done);
 		adev->dm.dmub_notify = kzalloc(sizeof(struct dmub_notification), GFP_KERNEL);
 		if (!adev->dm.dmub_notify) {
-			DRM_INFO("amdgpu: fail to allocate adev->dm.dmub_notify");
+			drm_info(adev_to_drm(adev), "fail to allocate adev->dm.dmub_notify");
 			goto error;
 		}
 
 		adev->dm.delayed_hpd_wq = create_singlethread_workqueue("amdgpu_dm_hpd_wq");
 		if (!adev->dm.delayed_hpd_wq) {
-			DRM_ERROR("amdgpu: failed to create hpd offload workqueue.\n");
+			drm_err(adev_to_drm(adev), "failed to create hpd offload workqueue.\n");
 			goto error;
 		}
 
 		amdgpu_dm_outbox_init(adev);
 		if (!register_dmub_notify_callback(adev, DMUB_NOTIFICATION_AUX_REPLY,
 			dmub_aux_setconfig_callback, false)) {
-			DRM_ERROR("amdgpu: fail to register dmub aux callback");
+			drm_err(adev_to_drm(adev), "fail to register dmub aux callback");
 			goto error;
 		}
-		if (!register_dmub_notify_callback(adev, DMUB_NOTIFICATION_HPD, dmub_hpd_callback, true)) {
-			DRM_ERROR("amdgpu: fail to register dmub hpd callback");
-			goto error;
-		}
-		if (!register_dmub_notify_callback(adev, DMUB_NOTIFICATION_HPD_IRQ, dmub_hpd_callback, true)) {
-			DRM_ERROR("amdgpu: fail to register dmub hpd callback");
+
+		for (size_t i = 0; i < ARRAY_SIZE(adev->dm.fused_io); i++)
+			init_completion(&adev->dm.fused_io[i].replied);
+
+		if (!register_dmub_notify_callback(adev, DMUB_NOTIFICATION_FUSED_IO,
+			dmub_aux_fused_io_callback, false)) {
+			drm_err(adev_to_drm(adev), "fail to register dmub fused io callback");
 			goto error;
 		}
-	}
-
-	/* Enable outbox notification only after IRQ handlers are registered and DMUB is alive.
-	 * It is expected that DMUB will resend any pending notifications at this point, for
-	 * example HPD from DPIA.
-	 */
-	if (dc_is_dmub_outbox_supported(adev->dm.dc))
+		/* Enable outbox notification only after IRQ handlers are registered and DMUB is alive.
+		 * It is expected that DMUB will resend any pending notifications at this point. Note
+		 * that hpd and hpd_irq handler registration are deferred to register_hpd_handlers() to
+		 * align legacy interface initialization sequence. Connection status will be proactivly
+		 * detected once in the amdgpu_dm_initialize_drm_device.
+		 */
 		dc_enable_dmub_outbox(adev->dm.dc);
 
+		/* DPIA trace goes to dmesg logs only if outbox is enabled */
+		if (amdgpu_dc_debug_mask & DC_ENABLE_DPIA_TRACE)
+			dc_dmub_srv_enable_dpia_trace(adev->dm.dc);
+	}
+
 	if (amdgpu_dm_initialize_drm_device(adev)) {
-		DRM_ERROR(
-		"amdgpu: failed to initialize sw for display support.\n");
+		drm_err(adev_to_drm(adev),
+		"failed to initialize sw for display support.\n");
 		goto error;
 	}
 
@@ -1835,13 +2189,22 @@ static int amdgpu_dm_init(struct amdgpu_device *adev)
 	adev_to_drm(adev)->mode_config.cursor_height = adev->dm.dc->caps.max_cursor_size;
 
 	if (drm_vblank_init(adev_to_drm(adev), adev->dm.display_indexes_num)) {
-		DRM_ERROR(
-		"amdgpu: failed to initialize sw for display support.\n");
+		drm_err(adev_to_drm(adev),
+		"failed to initialize vblank for display support.\n");
 		goto error;
 	}
 
+#if defined(CONFIG_DRM_AMD_SECURE_DISPLAY)
+	amdgpu_dm_crtc_secure_display_create_contexts(adev);
+	if (!adev->dm.secure_display_ctx.crtc_ctx)
+		drm_err(adev_to_drm(adev), "failed to initialize secure display contexts.\n");
+
+	if (amdgpu_ip_version(adev, DCE_HWIP, 0) >= IP_VERSION(4, 0, 1))
+		adev->dm.secure_display_ctx.support_mul_roi = true;
+
+#endif
 
-	DRM_DEBUG_DRIVER("KMS initialized.\n");
+	drm_dbg_driver(adev_to_drm(adev), "KMS initialized.\n");
 
 	return 0;
 error:
@@ -1850,9 +2213,9 @@ error:
 	return -EINVAL;
 }
 
-static int amdgpu_dm_early_fini(void *handle)
+static int amdgpu_dm_early_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_dm_audio_fini(adev);
 
@@ -1868,18 +2231,28 @@ static void amdgpu_dm_fini(struct amdgpu_device *adev)
 		adev->dm.vblank_control_workqueue = NULL;
 	}
 
+	if (adev->dm.idle_workqueue) {
+		if (adev->dm.idle_workqueue->running) {
+			adev->dm.idle_workqueue->enable = false;
+			flush_work(&adev->dm.idle_workqueue->work);
+		}
+
+		kfree(adev->dm.idle_workqueue);
+		adev->dm.idle_workqueue = NULL;
+	}
+
 	amdgpu_dm_destroy_drm_device(&adev->dm);
 
 #if defined(CONFIG_DRM_AMD_SECURE_DISPLAY)
-	if (adev->dm.secure_display_ctxs) {
+	if (adev->dm.secure_display_ctx.crtc_ctx) {
 		for (i = 0; i < adev->mode_info.num_crtc; i++) {
-			if (adev->dm.secure_display_ctxs[i].crtc) {
-				flush_work(&adev->dm.secure_display_ctxs[i].notify_ta_work);
-				flush_work(&adev->dm.secure_display_ctxs[i].forward_roi_work);
+			if (adev->dm.secure_display_ctx.crtc_ctx[i].crtc) {
+				flush_work(&adev->dm.secure_display_ctx.crtc_ctx[i].notify_ta_work);
+				flush_work(&adev->dm.secure_display_ctx.crtc_ctx[i].forward_roi_work);
 			}
 		}
-		kfree(adev->dm.secure_display_ctxs);
-		adev->dm.secure_display_ctxs = NULL;
+		kfree(adev->dm.secure_display_ctx.crtc_ctx);
+		adev->dm.secure_display_ctx.crtc_ctx = NULL;
 	}
 #endif
 	if (adev->dm.hdcp_workqueue) {
@@ -1887,17 +2260,15 @@ static void amdgpu_dm_fini(struct amdgpu_device *adev)
 		adev->dm.hdcp_workqueue = NULL;
 	}
 
-	if (adev->dm.dc)
+	if (adev->dm.dc) {
 		dc_deinit_callbacks(adev->dm.dc);
-
-	if (adev->dm.dc)
 		dc_dmub_srv_destroy(&adev->dm.dc->ctx->dmub_srv);
-
-	if (dc_enable_dmub_notifications(adev->dm.dc)) {
-		kfree(adev->dm.dmub_notify);
-		adev->dm.dmub_notify = NULL;
-		destroy_workqueue(adev->dm.delayed_hpd_wq);
-		adev->dm.delayed_hpd_wq = NULL;
+		if (dc_enable_dmub_notifications(adev->dm.dc)) {
+			kfree(adev->dm.dmub_notify);
+			adev->dm.dmub_notify = NULL;
+			destroy_workqueue(adev->dm.delayed_hpd_wq);
+			adev->dm.delayed_hpd_wq = NULL;
+		}
 	}
 
 	if (adev->dm.dmub_bo)
@@ -1905,7 +2276,7 @@ static void amdgpu_dm_fini(struct amdgpu_device *adev)
 				      &adev->dm.dmub_bo_gpu_addr,
 				      &adev->dm.dmub_bo_cpu_addr);
 
-	if (adev->dm.hpd_rx_offload_wq) {
+	if (adev->dm.hpd_rx_offload_wq && adev->dm.dc) {
 		for (i = 0; i < adev->dm.dc->caps.max_links; i++) {
 			if (adev->dm.hpd_rx_offload_wq[i].wq) {
 				destroy_workqueue(adev->dm.hpd_rx_offload_wq[i].wq);
@@ -1938,8 +2309,6 @@ static void amdgpu_dm_fini(struct amdgpu_device *adev)
 	mutex_destroy(&adev->dm.audio_lock);
 	mutex_destroy(&adev->dm.dc_lock);
 	mutex_destroy(&adev->dm.dpia_aux_lock);
-
-	return;
 }
 
 static int load_dmcu_fw(struct amdgpu_device *adev)
@@ -1948,7 +2317,7 @@ static int load_dmcu_fw(struct amdgpu_device *adev)
 	int r;
 	const struct dmcu_firmware_header_v1_0 *hdr;
 
-	switch(adev->asic_type) {
+	switch (adev->asic_type) {
 #if defined(CONFIG_DRM_AMD_DC_SI)
 	case CHIP_TAHITI:
 	case CHIP_PITCAIRN:
@@ -1984,7 +2353,7 @@ static int load_dmcu_fw(struct amdgpu_device *adev)
 			return 0;
 		break;
 	default:
-		switch (adev->ip_versions[DCE_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
 		case IP_VERSION(2, 0, 2):
 		case IP_VERSION(2, 0, 3):
 		case IP_VERSION(2, 0, 0):
@@ -2000,11 +2369,15 @@ static int load_dmcu_fw(struct amdgpu_device *adev)
 		case IP_VERSION(3, 1, 6):
 		case IP_VERSION(3, 2, 0):
 		case IP_VERSION(3, 2, 1):
+		case IP_VERSION(3, 5, 0):
+		case IP_VERSION(3, 5, 1):
+		case IP_VERSION(3, 6, 0):
+		case IP_VERSION(4, 0, 1):
 			return 0;
 		default:
 			break;
 		}
-		DRM_ERROR("Unsupported ASIC type: 0x%X\n", adev->asic_type);
+		drm_err(adev_to_drm(adev), "Unsupported ASIC type: 0x%X\n", adev->asic_type);
 		return -EINVAL;
 	}
 
@@ -2013,7 +2386,8 @@ static int load_dmcu_fw(struct amdgpu_device *adev)
 		return 0;
 	}
 
-	r = amdgpu_ucode_request(adev, &adev->dm.fw_dmcu, fw_name_dmcu);
+	r = amdgpu_ucode_request(adev, &adev->dm.fw_dmcu, AMDGPU_UCODE_REQUIRED,
+				 "%s", fw_name_dmcu);
 	if (r == -ENODEV) {
 		/* DMCU firmware is not necessary, so don't raise a fuss if it's missing */
 		DRM_DEBUG_KMS("dm: DMCU firmware not found\n");
@@ -2021,7 +2395,7 @@ static int load_dmcu_fw(struct amdgpu_device *adev)
 		return 0;
 	}
 	if (r) {
-		dev_err(adev->dev, "amdgpu_dm: Can't validate firmware \"%s\"\n",
+		drm_err(adev_to_drm(adev), "amdgpu_dm: Can't validate firmware \"%s\"\n",
 			fw_name_dmcu);
 		amdgpu_ucode_release(&adev->dm.fw_dmcu);
 		return r;
@@ -2065,15 +2439,27 @@ static int dm_dmub_sw_init(struct amdgpu_device *adev)
 	struct dmub_srv_create_params create_params;
 	struct dmub_srv_region_params region_params;
 	struct dmub_srv_region_info region_info;
-	struct dmub_srv_fb_params fb_params;
+	struct dmub_srv_memory_params memory_params;
 	struct dmub_srv_fb_info *fb_info;
 	struct dmub_srv *dmub_srv;
 	const struct dmcub_firmware_header_v1_0 *hdr;
 	enum dmub_asic dmub_asic;
 	enum dmub_status status;
+	static enum dmub_window_memory_type window_memory_type[DMUB_WINDOW_TOTAL] = {
+		DMUB_WINDOW_MEMORY_TYPE_FB,		//DMUB_WINDOW_0_INST_CONST
+		DMUB_WINDOW_MEMORY_TYPE_FB,		//DMUB_WINDOW_1_STACK
+		DMUB_WINDOW_MEMORY_TYPE_FB,		//DMUB_WINDOW_2_BSS_DATA
+		DMUB_WINDOW_MEMORY_TYPE_FB,		//DMUB_WINDOW_3_VBIOS
+		DMUB_WINDOW_MEMORY_TYPE_FB,		//DMUB_WINDOW_4_MAILBOX
+		DMUB_WINDOW_MEMORY_TYPE_FB,		//DMUB_WINDOW_5_TRACEBUFF
+		DMUB_WINDOW_MEMORY_TYPE_FB,		//DMUB_WINDOW_6_FW_STATE
+		DMUB_WINDOW_MEMORY_TYPE_FB,		//DMUB_WINDOW_7_SCRATCH_MEM
+		DMUB_WINDOW_MEMORY_TYPE_FB,		//DMUB_WINDOW_IB_MEM
+		DMUB_WINDOW_MEMORY_TYPE_FB,		//DMUB_WINDOW_SHARED_STATE
+	};
 	int r;
 
-	switch (adev->ip_versions[DCE_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
 	case IP_VERSION(2, 1, 0):
 		dmub_asic = DMUB_ASIC_DCN21;
 		break;
@@ -2108,6 +2494,17 @@ static int dm_dmub_sw_init(struct amdgpu_device *adev)
 	case IP_VERSION(3, 2, 1):
 		dmub_asic = DMUB_ASIC_DCN321;
 		break;
+	case IP_VERSION(3, 5, 0):
+	case IP_VERSION(3, 5, 1):
+		dmub_asic = DMUB_ASIC_DCN35;
+		break;
+	case IP_VERSION(3, 6, 0):
+		dmub_asic = DMUB_ASIC_DCN36;
+		break;
+	case IP_VERSION(4, 0, 1):
+		dmub_asic = DMUB_ASIC_DCN401;
+		break;
+
 	default:
 		/* ASIC doesn't support DMUB. */
 		return 0;
@@ -2124,7 +2521,7 @@ static int dm_dmub_sw_init(struct amdgpu_device *adev)
 		adev->firmware.fw_size +=
 			ALIGN(le32_to_cpu(hdr->inst_const_bytes), PAGE_SIZE);
 
-		DRM_INFO("Loading DMUB firmware via PSP: version=0x%08X\n",
+		drm_info(adev_to_drm(adev), "Loading DMUB firmware via PSP: version=0x%08X\n",
 			 adev->dm.dmcub_fw_version);
 	}
 
@@ -2133,7 +2530,7 @@ static int dm_dmub_sw_init(struct amdgpu_device *adev)
 	dmub_srv = adev->dm.dmub_srv;
 
 	if (!dmub_srv) {
-		DRM_ERROR("Failed to allocate DMUB service!\n");
+		drm_err(adev_to_drm(adev), "Failed to allocate DMUB service!\n");
 		return -ENOMEM;
 	}
 
@@ -2146,7 +2543,7 @@ static int dm_dmub_sw_init(struct amdgpu_device *adev)
 	/* Create the DMUB service. */
 	status = dmub_srv_create(dmub_srv, &create_params);
 	if (status != DMUB_STATUS_OK) {
-		DRM_ERROR("Error creating DMUB service: %d\n", status);
+		drm_err(adev_to_drm(adev), "Error creating DMUB service: %d\n", status);
 		return -EINVAL;
 	}
 
@@ -2165,12 +2562,13 @@ static int dm_dmub_sw_init(struct amdgpu_device *adev)
 		adev->dm.dmub_fw->data +
 		le32_to_cpu(hdr->header.ucode_array_offset_bytes) +
 		PSP_HEADER_BYTES;
+	region_params.window_memory_type = window_memory_type;
 
 	status = dmub_srv_calc_region_info(dmub_srv, &region_params,
 					   &region_info);
 
 	if (status != DMUB_STATUS_OK) {
-		DRM_ERROR("Error calculating DMUB region info: %d\n", status);
+		drm_err(adev_to_drm(adev), "Error calculating DMUB region info: %d\n", status);
 		return -EINVAL;
 	}
 
@@ -2188,35 +2586,48 @@ static int dm_dmub_sw_init(struct amdgpu_device *adev)
 		return r;
 
 	/* Rebase the regions on the framebuffer address. */
-	memset(&fb_params, 0, sizeof(fb_params));
-	fb_params.cpu_addr = adev->dm.dmub_bo_cpu_addr;
-	fb_params.gpu_addr = adev->dm.dmub_bo_gpu_addr;
-	fb_params.region_info = &region_info;
+	memset(&memory_params, 0, sizeof(memory_params));
+	memory_params.cpu_fb_addr = adev->dm.dmub_bo_cpu_addr;
+	memory_params.gpu_fb_addr = adev->dm.dmub_bo_gpu_addr;
+	memory_params.region_info = &region_info;
+	memory_params.window_memory_type = window_memory_type;
 
 	adev->dm.dmub_fb_info =
 		kzalloc(sizeof(*adev->dm.dmub_fb_info), GFP_KERNEL);
 	fb_info = adev->dm.dmub_fb_info;
 
 	if (!fb_info) {
-		DRM_ERROR(
+		drm_err(adev_to_drm(adev),
 			"Failed to allocate framebuffer info for DMUB service!\n");
 		return -ENOMEM;
 	}
 
-	status = dmub_srv_calc_fb_info(dmub_srv, &fb_params, fb_info);
+	status = dmub_srv_calc_mem_info(dmub_srv, &memory_params, fb_info);
 	if (status != DMUB_STATUS_OK) {
-		DRM_ERROR("Error calculating DMUB FB info: %d\n", status);
+		drm_err(adev_to_drm(adev), "Error calculating DMUB FB info: %d\n", status);
 		return -EINVAL;
 	}
 
+	adev->dm.bb_from_dmub = dm_dmub_get_vbios_bounding_box(adev);
+
 	return 0;
 }
 
-static int dm_sw_init(void *handle)
+static int dm_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
+	adev->dm.cgs_device = amdgpu_cgs_create_device(adev);
+
+	if (!adev->dm.cgs_device) {
+		drm_err(adev_to_drm(adev), "failed to create cgs device.\n");
+		return -EINVAL;
+	}
+
+	/* Moved from dm init since we need to use allocations for storing bounding box data */
+	INIT_LIST_HEAD(&adev->dm.da_list);
+
 	r = dm_dmub_sw_init(adev);
 	if (r)
 		return r;
@@ -2224,15 +2635,28 @@ static int dm_sw_init(void *handle)
 	return load_dmcu_fw(adev);
 }
 
-static int dm_sw_fini(void *handle)
+static int dm_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	struct dal_allocation *da;
+
+	list_for_each_entry(da, &adev->dm.da_list, list) {
+		if (adev->dm.bb_from_dmub == (void *) da->cpu_ptr) {
+			amdgpu_bo_free_kernel(&da->bo, &da->gpu_addr, &da->cpu_ptr);
+			list_del(&da->list);
+			kfree(da);
+			adev->dm.bb_from_dmub = NULL;
+			break;
+		}
+	}
+
 
 	kfree(adev->dm.dmub_fb_info);
 	adev->dm.dmub_fb_info = NULL;
 
 	if (adev->dm.dmub_srv) {
 		dmub_srv_destroy(adev->dm.dmub_srv);
+		kfree(adev->dm.dmub_srv);
 		adev->dm.dmub_srv = NULL;
 	}
 
@@ -2251,16 +2675,20 @@ static int detect_mst_link_for_all_connectors(struct drm_device *dev)
 
 	drm_connector_list_iter_begin(dev, &iter);
 	drm_for_each_connector_iter(connector, &iter) {
+
+		if (connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+			continue;
+
 		aconnector = to_amdgpu_dm_connector(connector);
 		if (aconnector->dc_link->type == dc_connection_mst_branch &&
 		    aconnector->mst_mgr.aux) {
-			DRM_DEBUG_DRIVER("DM_MST: starting TM on aconnector: %p [id: %d]\n",
+			drm_dbg_kms(dev, "DM_MST: starting TM on aconnector: %p [id: %d]\n",
 					 aconnector,
 					 aconnector->base.base.id);
 
 			ret = drm_dp_mst_topology_mgr_set_mst(&aconnector->mst_mgr, true);
 			if (ret < 0) {
-				DRM_ERROR("DM_MST: Failed to start MST\n");
+				drm_err(dev, "DM_MST: Failed to start MST\n");
 				aconnector->dc_link->type =
 					dc_connection_single;
 				ret = dm_helpers_dp_mst_stop_top_mgr(aconnector->dc_link->ctx,
@@ -2274,9 +2702,9 @@ static int detect_mst_link_for_all_connectors(struct drm_device *dev)
 	return ret;
 }
 
-static int dm_late_init(void *handle)
+static int dm_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	struct dmcu_iram_parameters params;
 	unsigned int linear_lut[16];
@@ -2320,17 +2748,112 @@ static int dm_late_init(void *handle)
 	return detect_mst_link_for_all_connectors(adev_to_drm(adev));
 }
 
+static void resume_mst_branch_status(struct drm_dp_mst_topology_mgr *mgr)
+{
+	u8 buf[UUID_SIZE];
+	guid_t guid;
+	int ret;
+
+	mutex_lock(&mgr->lock);
+	if (!mgr->mst_primary)
+		goto out_fail;
+
+	if (drm_dp_read_dpcd_caps(mgr->aux, mgr->dpcd) < 0) {
+		drm_dbg_kms(mgr->dev, "dpcd read failed - undocked during suspend?\n");
+		goto out_fail;
+	}
+
+	ret = drm_dp_dpcd_writeb(mgr->aux, DP_MSTM_CTRL,
+				 DP_MST_EN |
+				 DP_UP_REQ_EN |
+				 DP_UPSTREAM_IS_SRC);
+	if (ret < 0) {
+		drm_dbg_kms(mgr->dev, "mst write failed - undocked during suspend?\n");
+		goto out_fail;
+	}
+
+	/* Some hubs forget their guids after they resume */
+	ret = drm_dp_dpcd_read(mgr->aux, DP_GUID, buf, sizeof(buf));
+	if (ret != sizeof(buf)) {
+		drm_dbg_kms(mgr->dev, "dpcd read failed - undocked during suspend?\n");
+		goto out_fail;
+	}
+
+	import_guid(&guid, buf);
+
+	if (guid_is_null(&guid)) {
+		guid_gen(&guid);
+		export_guid(buf, &guid);
+
+		ret = drm_dp_dpcd_write(mgr->aux, DP_GUID, buf, sizeof(buf));
+
+		if (ret != sizeof(buf)) {
+			drm_dbg_kms(mgr->dev, "check mstb guid failed - undocked during suspend?\n");
+			goto out_fail;
+		}
+	}
+
+	guid_copy(&mgr->mst_primary->guid, &guid);
+
+out_fail:
+	mutex_unlock(&mgr->lock);
+}
+
+void hdmi_cec_unset_edid(struct amdgpu_dm_connector *aconnector)
+{
+	struct cec_notifier *n = aconnector->notifier;
+
+	if (!n)
+		return;
+
+	cec_notifier_phys_addr_invalidate(n);
+}
+
+void hdmi_cec_set_edid(struct amdgpu_dm_connector *aconnector)
+{
+	struct drm_connector *connector = &aconnector->base;
+	struct cec_notifier *n = aconnector->notifier;
+
+	if (!n)
+		return;
+
+	cec_notifier_set_phys_addr(n,
+				   connector->display_info.source_physical_address);
+}
+
+static void s3_handle_hdmi_cec(struct drm_device *ddev, bool suspend)
+{
+	struct amdgpu_dm_connector *aconnector;
+	struct drm_connector *connector;
+	struct drm_connector_list_iter conn_iter;
+
+	drm_connector_list_iter_begin(ddev, &conn_iter);
+	drm_for_each_connector_iter(connector, &conn_iter) {
+		if (connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+			continue;
+
+		aconnector = to_amdgpu_dm_connector(connector);
+		if (suspend)
+			hdmi_cec_unset_edid(aconnector);
+		else
+			hdmi_cec_set_edid(aconnector);
+	}
+	drm_connector_list_iter_end(&conn_iter);
+}
+
 static void s3_handle_mst(struct drm_device *dev, bool suspend)
 {
 	struct amdgpu_dm_connector *aconnector;
 	struct drm_connector *connector;
 	struct drm_connector_list_iter iter;
 	struct drm_dp_mst_topology_mgr *mgr;
-	int ret;
-	bool need_hotplug = false;
 
 	drm_connector_list_iter_begin(dev, &iter);
 	drm_for_each_connector_iter(connector, &iter) {
+
+		if (connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+			continue;
+
 		aconnector = to_amdgpu_dm_connector(connector);
 		if (aconnector->dc_link->type != dc_connection_mst_branch ||
 		    aconnector->mst_root)
@@ -2349,18 +2872,15 @@ static void s3_handle_mst(struct drm_device *dev, bool suspend)
 			if (!dp_is_lttpr_present(aconnector->dc_link))
 				try_to_configure_aux_timeout(aconnector->dc_link->ddc, LINK_AUX_DEFAULT_TIMEOUT_PERIOD);
 
-			ret = drm_dp_mst_topology_mgr_resume(mgr, true);
-			if (ret < 0) {
-				dm_helpers_dp_mst_stop_top_mgr(aconnector->dc_link->ctx,
-					aconnector->dc_link);
-				need_hotplug = true;
-			}
+			/* TODO: move resume_mst_branch_status() into drm mst resume again
+			 * once topology probing work is pulled out from mst resume into mst
+			 * resume 2nd step. mst resume 2nd step should be called after old
+			 * state getting restored (i.e. drm_atomic_helper_resume()).
+			 */
+			resume_mst_branch_status(mgr);
 		}
 	}
 	drm_connector_list_iter_end(&iter);
-
-	if (need_hotplug)
-		drm_kms_helper_hotplug_event(dev);
 }
 
 static int amdgpu_dm_smu_write_watermarks_table(struct amdgpu_device *adev)
@@ -2397,7 +2917,7 @@ static int amdgpu_dm_smu_write_watermarks_table(struct amdgpu_device *adev)
 	 * therefore, this function apply to navi10/12/14 but not Renoir
 	 * *
 	 */
-	switch (adev->ip_versions[DCE_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
 	case IP_VERSION(2, 0, 2):
 	case IP_VERSION(2, 0, 0):
 		break;
@@ -2407,16 +2927,43 @@ static int amdgpu_dm_smu_write_watermarks_table(struct amdgpu_device *adev)
 
 	ret = amdgpu_dpm_write_watermarks_table(adev);
 	if (ret) {
-		DRM_ERROR("Failed to update WMTABLE!\n");
+		drm_err(adev_to_drm(adev), "Failed to update WMTABLE!\n");
 		return ret;
 	}
 
 	return 0;
 }
 
+static int dm_oem_i2c_hw_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_display_manager *dm = &adev->dm;
+	struct amdgpu_i2c_adapter *oem_i2c;
+	struct ddc_service *oem_ddc_service;
+	int r;
+
+	oem_ddc_service = dc_get_oem_i2c_device(adev->dm.dc);
+	if (oem_ddc_service) {
+		oem_i2c = create_i2c(oem_ddc_service, true);
+		if (!oem_i2c) {
+			drm_info(adev_to_drm(adev), "Failed to create oem i2c adapter data\n");
+			return -ENOMEM;
+		}
+
+		r = devm_i2c_add_adapter(adev->dev, &oem_i2c->base);
+		if (r) {
+			drm_info(adev_to_drm(adev), "Failed to register oem i2c\n");
+			kfree(oem_i2c);
+			return r;
+		}
+		dm->oem_i2c = oem_i2c;
+	}
+
+	return 0;
+}
+
 /**
  * dm_hw_init() - Initialize DC device
- * @handle: The base driver device containing the amdgpu_dm device.
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Initialize the &struct amdgpu_display_manager device. This involves calling
  * the initializers of each DM component, then populating the struct with them.
@@ -2434,27 +2981,35 @@ static int amdgpu_dm_smu_write_watermarks_table(struct amdgpu_device *adev)
  * - Vblank support
  * - Debug FS entries, if enabled
  */
-static int dm_hw_init(void *handle)
+static int dm_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
 	/* Create DAL display manager */
-	amdgpu_dm_init(adev);
+	r = amdgpu_dm_init(adev);
+	if (r)
+		return r;
 	amdgpu_dm_hpd_init(adev);
 
+	r = dm_oem_i2c_hw_init(adev);
+	if (r)
+		drm_info(adev_to_drm(adev), "Failed to add OEM i2c bus\n");
+
 	return 0;
 }
 
 /**
  * dm_hw_fini() - Teardown DC device
- * @handle: The base driver device containing the amdgpu_dm device.
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  * Teardown components within &struct amdgpu_display_manager that require
  * cleanup. This involves cleaning up the DRM device, DC, and any modules that
  * were loaded. Also flush IRQ workqueues and disable them.
  */
-static int dm_hw_fini(void *handle)
+static int dm_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_dm_hpd_fini(adev);
 
@@ -2479,40 +3034,51 @@ static void dm_gpureset_toggle_interrupts(struct amdgpu_device *adev,
 		if (acrtc && state->stream_status[i].plane_count != 0) {
 			irq_source = IRQ_TYPE_PFLIP + acrtc->otg_inst;
 			rc = dc_interrupt_set(adev->dm.dc, irq_source, enable) ? 0 : -EBUSY;
-			DRM_DEBUG_VBL("crtc %d - vupdate irq %sabling: r=%d\n",
-				      acrtc->crtc_id, enable ? "en" : "dis", rc);
 			if (rc)
-				DRM_WARN("Failed to %s pflip interrupts\n",
+				drm_warn(adev_to_drm(adev), "Failed to %s pflip interrupts\n",
 					 enable ? "enable" : "disable");
 
-			if (enable) {
-				rc = amdgpu_dm_crtc_enable_vblank(&acrtc->base);
+			if (dc_supports_vrr(adev->dm.dc->ctx->dce_version)) {
+				if (enable) {
+					if (amdgpu_dm_crtc_vrr_active(
+							to_dm_crtc_state(acrtc->base.state)))
+						rc = amdgpu_dm_crtc_set_vupdate_irq(
+							&acrtc->base, true);
+				} else
+					rc = amdgpu_dm_crtc_set_vupdate_irq(
+							&acrtc->base, false);
+
 				if (rc)
-					DRM_WARN("Failed to enable vblank interrupts\n");
-			} else {
-				amdgpu_dm_crtc_disable_vblank(&acrtc->base);
+					drm_warn(adev_to_drm(adev), "Failed to %sable vupdate interrupt\n",
+						enable ? "en" : "dis");
 			}
 
+			irq_source = IRQ_TYPE_VBLANK + acrtc->otg_inst;
+			/* During gpu-reset we disable and then enable vblank irq, so
+			 * don't use amdgpu_irq_get/put() to avoid refcount change.
+			 */
+			if (!dc_interrupt_set(adev->dm.dc, irq_source, enable))
+				drm_warn(adev_to_drm(adev), "Failed to %sable vblank interrupt\n", enable ? "en" : "dis");
 		}
 	}
 
 }
 
+DEFINE_FREE(state_release, struct dc_state *, if (_T) dc_state_release(_T))
+
 static enum dc_status amdgpu_dm_commit_zero_streams(struct dc *dc)
 {
-	struct dc_state *context = NULL;
-	enum dc_status res = DC_ERROR_UNEXPECTED;
+	struct dc_state *context __free(state_release) = NULL;
 	int i;
 	struct dc_stream_state *del_streams[MAX_PIPES];
 	int del_streams_count = 0;
+	struct dc_commit_streams_params params = {};
 
 	memset(del_streams, 0, sizeof(del_streams));
 
-	context = dc_create_state(dc);
+	context = dc_state_create_current_copy(dc);
 	if (context == NULL)
-		goto context_alloc_fail;
-
-	dc_resource_state_copy_construct_current(dc, context);
+		return DC_ERROR_UNEXPECTED;
 
 	/* First remove from context all streams */
 	for (i = 0; i < context->stream_count; i++) {
@@ -2523,23 +3089,20 @@ static enum dc_status amdgpu_dm_commit_zero_streams(struct dc *dc)
 
 	/* Remove all planes for removed streams and then remove the streams */
 	for (i = 0; i < del_streams_count; i++) {
-		if (!dc_rem_all_planes_for_stream(dc, del_streams[i], context)) {
-			res = DC_FAIL_DETACH_SURFACES;
-			goto fail;
-		}
+		enum dc_status res;
 
-		res = dc_remove_stream_from_ctx(dc, context, del_streams[i]);
+		if (!dc_state_rem_all_planes_for_stream(dc, del_streams[i], context))
+			return DC_FAIL_DETACH_SURFACES;
+
+		res = dc_state_remove_stream(dc, context, del_streams[i]);
 		if (res != DC_OK)
-			goto fail;
+			return res;
 	}
 
-	res = dc_commit_streams(dc, context->streams, context->stream_count);
+	params.streams = context->streams;
+	params.stream_count = context->stream_count;
 
-fail:
-	dc_release_state(context);
-
-context_alloc_fail:
-	return res;
+	return dc_commit_streams(dc, &params);
 }
 
 static void hpd_rx_irq_work_suspend(struct amdgpu_display_manager *dm)
@@ -2552,32 +3115,108 @@ static void hpd_rx_irq_work_suspend(struct amdgpu_display_manager *dm)
 	}
 }
 
-static int dm_suspend(void *handle)
+static int dm_cache_state(struct amdgpu_device *adev)
+{
+	int r;
+
+	adev->dm.cached_state = drm_atomic_helper_suspend(adev_to_drm(adev));
+	if (IS_ERR(adev->dm.cached_state)) {
+		r = PTR_ERR(adev->dm.cached_state);
+		adev->dm.cached_state = NULL;
+	}
+
+	return adev->dm.cached_state ? 0 : r;
+}
+
+static void dm_destroy_cached_state(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = handle;
 	struct amdgpu_display_manager *dm = &adev->dm;
-	int ret = 0;
+	struct drm_device *ddev = adev_to_drm(adev);
+	struct dm_plane_state *dm_new_plane_state;
+	struct drm_plane_state *new_plane_state;
+	struct dm_crtc_state *dm_new_crtc_state;
+	struct drm_crtc_state *new_crtc_state;
+	struct drm_plane *plane;
+	struct drm_crtc *crtc;
+	int i;
+
+	if (!dm->cached_state)
+		return;
+
+	/* Force mode set in atomic commit */
+	for_each_new_crtc_in_state(dm->cached_state, crtc, new_crtc_state, i) {
+		new_crtc_state->active_changed = true;
+		dm_new_crtc_state = to_dm_crtc_state(new_crtc_state);
+		reset_freesync_config_for_crtc(dm_new_crtc_state);
+	}
+
+	/*
+	 * atomic_check is expected to create the dc states. We need to release
+	 * them here, since they were duplicated as part of the suspend
+	 * procedure.
+	 */
+	for_each_new_crtc_in_state(dm->cached_state, crtc, new_crtc_state, i) {
+		dm_new_crtc_state = to_dm_crtc_state(new_crtc_state);
+		if (dm_new_crtc_state->stream) {
+			WARN_ON(kref_read(&dm_new_crtc_state->stream->refcount) > 1);
+			dc_stream_release(dm_new_crtc_state->stream);
+			dm_new_crtc_state->stream = NULL;
+		}
+		dm_new_crtc_state->base.color_mgmt_changed = true;
+	}
+
+	for_each_new_plane_in_state(dm->cached_state, plane, new_plane_state, i) {
+		dm_new_plane_state = to_dm_plane_state(new_plane_state);
+		if (dm_new_plane_state->dc_state) {
+			WARN_ON(kref_read(&dm_new_plane_state->dc_state->refcount) > 1);
+			dc_plane_state_release(dm_new_plane_state->dc_state);
+			dm_new_plane_state->dc_state = NULL;
+		}
+	}
+
+	drm_atomic_helper_resume(ddev, dm->cached_state);
+
+	dm->cached_state = NULL;
+}
+
+static int dm_suspend(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_display_manager *dm = &adev->dm;
 
 	if (amdgpu_in_reset(adev)) {
+		enum dc_status res;
+
 		mutex_lock(&dm->dc_lock);
 
 		dc_allow_idle_optimizations(adev->dm.dc, false);
 
-		dm->cached_dc_state = dc_copy_state(dm->dc->current_state);
+		dm->cached_dc_state = dc_state_create_copy(dm->dc->current_state);
 
-		dm_gpureset_toggle_interrupts(adev, dm->cached_dc_state, false);
+		if (dm->cached_dc_state)
+			dm_gpureset_toggle_interrupts(adev, dm->cached_dc_state, false);
 
-		amdgpu_dm_commit_zero_streams(dm->dc);
+		res = amdgpu_dm_commit_zero_streams(dm->dc);
+		if (res != DC_OK) {
+			drm_err(adev_to_drm(adev), "Failed to commit zero streams: %d\n", res);
+			return -EINVAL;
+		}
 
 		amdgpu_dm_irq_suspend(adev);
 
 		hpd_rx_irq_work_suspend(dm);
 
-		return ret;
+		return 0;
 	}
 
-	WARN_ON(adev->dm.cached_state);
-	adev->dm.cached_state = drm_atomic_helper_suspend(adev_to_drm(adev));
+	if (!adev->dm.cached_state) {
+		int r = dm_cache_state(adev);
+
+		if (r)
+			return r;
+	}
+
+	s3_handle_hdmi_cec(adev_to_drm(adev), true);
 
 	s3_handle_mst(adev_to_drm(adev), true);
 
@@ -2587,10 +3226,15 @@ static int dm_suspend(void *handle)
 
 	dc_set_power_state(dm->dc, DC_ACPI_CM_POWER_STATE_D3);
 
+	if (dm->dc->caps.ips_support && adev->in_s0ix)
+		dc_allow_idle_optimizations(dm->dc, true);
+
+	dc_dmub_srv_set_power_state(dm->dc->ctx->dmub_srv, DC_ACPI_CM_POWER_STATE_D3);
+
 	return 0;
 }
 
-struct amdgpu_dm_connector *
+struct drm_connector *
 amdgpu_dm_find_first_crtc_matching_connector(struct drm_atomic_state *state,
 					     struct drm_crtc *crtc)
 {
@@ -2603,7 +3247,7 @@ amdgpu_dm_find_first_crtc_matching_connector(struct drm_atomic_state *state,
 		crtc_from_state = new_con_state->crtc;
 
 		if (crtc_from_state == crtc)
-			return to_amdgpu_dm_connector(connector);
+			return connector;
 	}
 
 	return NULL;
@@ -2615,6 +3259,7 @@ static void emulated_link_detect(struct dc_link *link)
 	struct display_sink_capability sink_caps = { 0 };
 	enum dc_edid_status edid_status;
 	struct dc_context *dc_ctx = link->ctx;
+	struct drm_device *dev = adev_to_drm(dc_ctx->driver_context);
 	struct dc_sink *sink = NULL;
 	struct dc_sink *prev_sink = NULL;
 
@@ -2664,7 +3309,7 @@ static void emulated_link_detect(struct dc_link *link)
 	}
 
 	default:
-		DC_ERROR("Invalid connector type! signal:%d\n",
+		drm_err(dev, "Invalid connector type! signal:%d\n",
 			link->connector_signal);
 		return;
 	}
@@ -2674,7 +3319,7 @@ static void emulated_link_detect(struct dc_link *link)
 
 	sink = dc_sink_create(&sink_init_data);
 	if (!sink) {
-		DC_ERROR("Failed to create sink!\n");
+		drm_err(dev, "Failed to create sink!\n");
 		return;
 	}
 
@@ -2687,7 +3332,7 @@ static void emulated_link_detect(struct dc_link *link)
 			sink);
 
 	if (edid_status != EDID_OK)
-		DC_ERROR("Failed to read EDID");
+		drm_err(dev, "Failed to read EDID\n");
 
 }
 
@@ -2700,58 +3345,131 @@ static void dm_gpureset_commit_state(struct dc_state *dc_state,
 		struct dc_scaling_info scaling_infos[MAX_SURFACES];
 		struct dc_flip_addrs flip_addrs[MAX_SURFACES];
 		struct dc_stream_update stream_update;
-	} * bundle;
+	} *bundle __free(kfree);
 	int k, m;
 
 	bundle = kzalloc(sizeof(*bundle), GFP_KERNEL);
 
 	if (!bundle) {
-		dm_error("Failed to allocate update bundle\n");
-		goto cleanup;
+		drm_err(dm->ddev, "Failed to allocate update bundle\n");
+		return;
 	}
 
 	for (k = 0; k < dc_state->stream_count; k++) {
 		bundle->stream_update.stream = dc_state->streams[k];
 
-		for (m = 0; m < dc_state->stream_status->plane_count; m++) {
+		for (m = 0; m < dc_state->stream_status[k].plane_count; m++) {
 			bundle->surface_updates[m].surface =
-				dc_state->stream_status->plane_states[m];
+				dc_state->stream_status[k].plane_states[m];
 			bundle->surface_updates[m].surface->force_full_update =
 				true;
 		}
 
 		update_planes_and_stream_adapter(dm->dc,
 					 UPDATE_TYPE_FULL,
-					 dc_state->stream_status->plane_count,
+					 dc_state->stream_status[k].plane_count,
 					 dc_state->streams[k],
 					 &bundle->stream_update,
 					 bundle->surface_updates);
 	}
+}
 
-cleanup:
-	kfree(bundle);
+static void apply_delay_after_dpcd_poweroff(struct amdgpu_device *adev,
+					    struct dc_sink *sink)
+{
+	struct dc_panel_patch *ppatch = NULL;
 
-	return;
+	if (!sink)
+		return;
+
+	ppatch = &sink->edid_caps.panel_patch;
+	if (ppatch->wait_after_dpcd_poweroff_ms) {
+		msleep(ppatch->wait_after_dpcd_poweroff_ms);
+		drm_dbg_driver(adev_to_drm(adev),
+			       "%s: adding a %ds delay as w/a for panel\n",
+			       __func__,
+			       ppatch->wait_after_dpcd_poweroff_ms / 1000);
+	}
 }
 
-static int dm_resume(void *handle)
+/**
+ * amdgpu_dm_dump_links_and_sinks - Debug dump of all DC links and their sinks
+ * @adev: amdgpu device pointer
+ *
+ * Iterates through all DC links and dumps information about local and remote
+ * (MST) sinks. Should be called after connector detection is complete to see
+ * the final state of all links.
+ */
+static void amdgpu_dm_dump_links_and_sinks(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = handle;
+	struct dc *dc = adev->dm.dc;
+	struct drm_device *dev = adev_to_drm(adev);
+	int li;
+
+	if (!dc)
+		return;
+
+	for (li = 0; li < dc->link_count; li++) {
+		struct dc_link *l = dc->links[li];
+		const char *name = NULL;
+		int rs;
+
+		if (!l)
+			continue;
+		if (l->local_sink && l->local_sink->edid_caps.display_name[0])
+			name = l->local_sink->edid_caps.display_name;
+		else
+			name = "n/a";
+
+		drm_dbg_kms(dev,
+			"LINK_DUMP[%d]: local_sink=%p type=%d sink_signal=%d sink_count=%u edid_name=%s mst_capable=%d mst_alloc_streams=%d\n",
+			li,
+			l->local_sink,
+			l->type,
+			l->local_sink ? l->local_sink->sink_signal : SIGNAL_TYPE_NONE,
+			l->sink_count,
+			name,
+			l->dpcd_caps.is_mst_capable,
+			l->mst_stream_alloc_table.stream_count);
+
+		/* Dump remote (MST) sinks if any */
+		for (rs = 0; rs < l->sink_count; rs++) {
+			struct dc_sink *rsink = l->remote_sinks[rs];
+			const char *rname = NULL;
+
+			if (!rsink)
+				continue;
+			if (rsink->edid_caps.display_name[0])
+				rname = rsink->edid_caps.display_name;
+			else
+				rname = "n/a";
+			drm_dbg_kms(dev,
+				"  REMOTE_SINK[%d:%d]: sink=%p signal=%d edid_name=%s\n",
+				li, rs,
+				rsink,
+				rsink->sink_signal,
+				rname);
+		}
+	}
+}
+
+static int dm_resume(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
 	struct drm_device *ddev = adev_to_drm(adev);
 	struct amdgpu_display_manager *dm = &adev->dm;
 	struct amdgpu_dm_connector *aconnector;
 	struct drm_connector *connector;
 	struct drm_connector_list_iter iter;
-	struct drm_crtc *crtc;
-	struct drm_crtc_state *new_crtc_state;
-	struct dm_crtc_state *dm_new_crtc_state;
-	struct drm_plane *plane;
-	struct drm_plane_state *new_plane_state;
-	struct dm_plane_state *dm_new_plane_state;
 	struct dm_atomic_state *dm_state = to_dm_atomic_state(dm->atomic_obj.state);
 	enum dc_connection_type new_connection_type = dc_connection_none;
 	struct dc_state *dc_state;
 	int i, r, j;
+	struct dc_commit_streams_params commit_params = {};
+
+	if (dm->dc->caps.ips_support) {
+		dc_dmub_srv_apply_idle_power_optimizations(dm->dc, false);
+	}
 
 	if (amdgpu_in_reset(adev)) {
 		dc_state = dm->cached_dc_state;
@@ -2774,10 +3492,14 @@ static int dm_resume(void *handle)
 		link_enc_cfg_copy(adev->dm.dc->current_state, dc_state);
 
 		r = dm_dmub_hw_init(adev);
-		if (r)
-			DRM_ERROR("DMUB interface failed to initialize: status=%d\n", r);
+		if (r) {
+			drm_err(adev_to_drm(adev), "DMUB interface failed to initialize: status=%d\n", r);
+			return r;
+		}
 
+		dc_dmub_srv_set_power_state(dm->dc->ctx->dmub_srv, DC_ACPI_CM_POWER_STATE_D0);
 		dc_set_power_state(dm->dc, DC_ACPI_CM_POWER_STATE_D0);
+
 		dc_resume(dm->dc);
 
 		amdgpu_dm_irq_resume_early(adev);
@@ -2795,26 +3517,34 @@ static int dm_resume(void *handle)
 			dc_enable_dmub_outbox(adev->dm.dc);
 		}
 
-		WARN_ON(!dc_commit_streams(dm->dc, dc_state->streams, dc_state->stream_count));
+		commit_params.streams = dc_state->streams;
+		commit_params.stream_count = dc_state->stream_count;
+		dc_exit_ips_for_hw_access(dm->dc);
+		WARN_ON(!dc_commit_streams(dm->dc, &commit_params));
 
 		dm_gpureset_commit_state(dm->cached_dc_state, dm);
 
 		dm_gpureset_toggle_interrupts(adev, dm->cached_dc_state, true);
 
-		dc_release_state(dm->cached_dc_state);
+		dc_state_release(dm->cached_dc_state);
 		dm->cached_dc_state = NULL;
 
 		amdgpu_dm_irq_resume_late(adev);
 
 		mutex_unlock(&dm->dc_lock);
 
+		/* set the backlight after a reset */
+		for (i = 0; i < dm->num_of_edps; i++) {
+			if (dm->backlight_dev[i])
+				amdgpu_dm_backlight_set_level(dm, i, dm->brightness[i]);
+		}
+
 		return 0;
 	}
 	/* Recreate dc_state - DC invalidates it when setting power state to S3. */
-	dc_release_state(dm_state->context);
-	dm_state->context = dc_create_state(dm->dc);
+	dc_state_release(dm_state->context);
+	dm_state->context = dc_state_create(dm->dc, NULL);
 	/* TODO: Remove dc_state->dccg, use dc->dccg directly. */
-	dc_resource_state_construct(dm->dc, dm_state->context);
 
 	/* Before powering on DC we need to re-initialize DMUB. */
 	dm_dmub_hw_resume(adev);
@@ -2826,6 +3556,7 @@ static int dm_resume(void *handle)
 	}
 
 	/* power on hardware */
+	dc_dmub_srv_set_power_state(dm->dc->ctx->dmub_srv, DC_ACPI_CM_POWER_STATE_D0);
 	dc_set_power_state(dm->dc, DC_ACPI_CM_POWER_STATE_D0);
 
 	/* program HPD filter */
@@ -2837,34 +3568,45 @@ static int dm_resume(void *handle)
 	 */
 	amdgpu_dm_irq_resume_early(adev);
 
+	s3_handle_hdmi_cec(ddev, false);
+
 	/* On resume we need to rewrite the MSTM control bits to enable MST*/
 	s3_handle_mst(ddev, false);
 
 	/* Do detection*/
 	drm_connector_list_iter_begin(ddev, &iter);
 	drm_for_each_connector_iter(connector, &iter) {
+		bool ret;
+
+		if (connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+			continue;
+
 		aconnector = to_amdgpu_dm_connector(connector);
 
 		if (!aconnector->dc_link)
 			continue;
 
 		/*
-		 * this is the case when traversing through already created
+		 * this is the case when traversing through already created end sink
 		 * MST connectors, should be skipped
 		 */
-		if (aconnector->dc_link->type == dc_connection_mst_branch)
+		if (aconnector->mst_root)
 			continue;
 
-		mutex_lock(&aconnector->hpd_lock);
+		guard(mutex)(&aconnector->hpd_lock);
 		if (!dc_link_detect_connection_type(aconnector->dc_link, &new_connection_type))
-			DRM_ERROR("KMS: Failed to detect connector\n");
+			drm_err(adev_to_drm(adev), "KMS: Failed to detect connector\n");
 
 		if (aconnector->base.force && new_connection_type == dc_connection_none) {
 			emulated_link_detect(aconnector->dc_link);
 		} else {
-			mutex_lock(&dm->dc_lock);
-			dc_link_detect(aconnector->dc_link, DETECT_REASON_HPD);
-			mutex_unlock(&dm->dc_lock);
+			guard(mutex)(&dm->dc_lock);
+			dc_exit_ips_for_hw_access(dm->dc);
+			ret = dc_link_detect(aconnector->dc_link, DETECT_REASON_RESUMEFROMS3S4);
+			if (ret) {
+				/* w/a delay for certain panels */
+				apply_delay_after_dpcd_poweroff(adev, aconnector->dc_sink);
+			}
 		}
 
 		if (aconnector->fake_enable && aconnector->dc_link->local_sink)
@@ -2874,45 +3616,47 @@ static int dm_resume(void *handle)
 			dc_sink_release(aconnector->dc_sink);
 		aconnector->dc_sink = NULL;
 		amdgpu_dm_update_connector_after_detect(aconnector);
-		mutex_unlock(&aconnector->hpd_lock);
 	}
 	drm_connector_list_iter_end(&iter);
 
-	/* Force mode set in atomic commit */
-	for_each_new_crtc_in_state(dm->cached_state, crtc, new_crtc_state, i)
-		new_crtc_state->active_changed = true;
+	dm_destroy_cached_state(adev);
 
-	/*
-	 * atomic_check is expected to create the dc states. We need to release
-	 * them here, since they were duplicated as part of the suspend
-	 * procedure.
-	 */
-	for_each_new_crtc_in_state(dm->cached_state, crtc, new_crtc_state, i) {
-		dm_new_crtc_state = to_dm_crtc_state(new_crtc_state);
-		if (dm_new_crtc_state->stream) {
-			WARN_ON(kref_read(&dm_new_crtc_state->stream->refcount) > 1);
-			dc_stream_release(dm_new_crtc_state->stream);
-			dm_new_crtc_state->stream = NULL;
-		}
-	}
+	/* Do mst topology probing after resuming cached state*/
+	drm_connector_list_iter_begin(ddev, &iter);
+	drm_for_each_connector_iter(connector, &iter) {
+		bool init = false;
 
-	for_each_new_plane_in_state(dm->cached_state, plane, new_plane_state, i) {
-		dm_new_plane_state = to_dm_plane_state(new_plane_state);
-		if (dm_new_plane_state->dc_state) {
-			WARN_ON(kref_read(&dm_new_plane_state->dc_state->refcount) > 1);
-			dc_plane_state_release(dm_new_plane_state->dc_state);
-			dm_new_plane_state->dc_state = NULL;
+		if (connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+			continue;
+
+		aconnector = to_amdgpu_dm_connector(connector);
+		if (aconnector->dc_link->type != dc_connection_mst_branch ||
+		    aconnector->mst_root)
+			continue;
+
+		scoped_guard(mutex, &aconnector->mst_mgr.lock) {
+			init = !aconnector->mst_mgr.mst_primary;
 		}
+		if (init)
+			dm_helpers_dp_mst_start_top_mgr(aconnector->dc_link->ctx,
+				aconnector->dc_link, false);
+		else
+			drm_dp_mst_topology_queue_probe(&aconnector->mst_mgr);
 	}
+	drm_connector_list_iter_end(&iter);
 
-	drm_atomic_helper_resume(ddev, dm->cached_state);
-
-	dm->cached_state = NULL;
+	/* Debug dump: list all DC links and their associated sinks after detection
+	 * is complete for all connectors. This provides a comprehensive view of the
+	 * final state without repeating the dump for each connector.
+	 */
+	amdgpu_dm_dump_links_and_sinks(adev);
 
 	amdgpu_dm_irq_resume_late(adev);
 
 	amdgpu_dm_smu_write_watermarks_table(adev);
 
+	drm_kms_helper_hotplug_event(ddev);
+
 	return 0;
 }
 
@@ -2945,8 +3689,7 @@ static const struct amd_ip_funcs amdgpu_dm_funcs = {
 	.set_powergating_state = dm_set_powergating_state,
 };
 
-const struct amdgpu_ip_block_version dm_ip_block =
-{
+const struct amdgpu_ip_block_version dm_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_DCE,
 	.major = 1,
 	.minor = 0,
@@ -2970,54 +3713,85 @@ static const struct drm_mode_config_funcs amdgpu_dm_mode_funcs = {
 
 static struct drm_mode_config_helper_funcs amdgpu_dm_mode_config_helperfuncs = {
 	.atomic_commit_tail = amdgpu_dm_atomic_commit_tail,
-	.atomic_commit_setup = drm_dp_mst_atomic_setup_commit,
+	.atomic_commit_setup = amdgpu_dm_atomic_setup_commit,
 };
 
 static void update_connector_ext_caps(struct amdgpu_dm_connector *aconnector)
 {
+	const struct drm_panel_backlight_quirk *panel_backlight_quirk;
 	struct amdgpu_dm_backlight_caps *caps;
 	struct drm_connector *conn_base;
 	struct amdgpu_device *adev;
 	struct drm_luminance_range_info *luminance_range;
+	struct drm_device *drm;
 
 	if (aconnector->bl_idx == -1 ||
 	    aconnector->dc_link->connector_signal != SIGNAL_TYPE_EDP)
 		return;
 
 	conn_base = &aconnector->base;
-	adev = drm_to_adev(conn_base->dev);
+	drm = conn_base->dev;
+	adev = drm_to_adev(drm);
 
 	caps = &adev->dm.backlight_caps[aconnector->bl_idx];
 	caps->ext_caps = &aconnector->dc_link->dpcd_sink_ext_caps;
 	caps->aux_support = false;
 
-	if (caps->ext_caps->bits.oled == 1 /*||
-	    caps->ext_caps->bits.sdr_aux_backlight_control == 1 ||
-	    caps->ext_caps->bits.hdr_aux_backlight_control == 1*/)
+	if (caps->ext_caps->bits.oled == 1
+	    /*
+	     * ||
+	     * caps->ext_caps->bits.sdr_aux_backlight_control == 1 ||
+	     * caps->ext_caps->bits.hdr_aux_backlight_control == 1
+	     */)
 		caps->aux_support = true;
 
 	if (amdgpu_backlight == 0)
 		caps->aux_support = false;
 	else if (amdgpu_backlight == 1)
 		caps->aux_support = true;
+	if (caps->aux_support)
+		aconnector->dc_link->backlight_control_type = BACKLIGHT_CONTROL_AMD_AUX;
 
 	luminance_range = &conn_base->display_info.luminance_range;
 
-	if (luminance_range->max_luminance) {
-		caps->aux_min_input_signal = luminance_range->min_luminance;
+	if (luminance_range->max_luminance)
 		caps->aux_max_input_signal = luminance_range->max_luminance;
-	} else {
-		caps->aux_min_input_signal = 0;
+	else
 		caps->aux_max_input_signal = 512;
+
+	if (luminance_range->min_luminance)
+		caps->aux_min_input_signal = luminance_range->min_luminance;
+	else
+		caps->aux_min_input_signal = 1;
+
+	panel_backlight_quirk =
+		drm_get_panel_backlight_quirk(aconnector->drm_edid);
+	if (!IS_ERR_OR_NULL(panel_backlight_quirk)) {
+		if (panel_backlight_quirk->min_brightness) {
+			caps->min_input_signal =
+				panel_backlight_quirk->min_brightness - 1;
+			drm_info(drm,
+				 "Applying panel backlight quirk, min_brightness: %d\n",
+				 caps->min_input_signal);
+		}
+		if (panel_backlight_quirk->brightness_mask) {
+			drm_info(drm,
+				 "Applying panel backlight quirk, brightness_mask: 0x%X\n",
+				 panel_backlight_quirk->brightness_mask);
+			caps->brightness_mask =
+				panel_backlight_quirk->brightness_mask;
+		}
 	}
 }
 
+DEFINE_FREE(sink_release, struct dc_sink *, if (_T) dc_sink_release(_T))
+
 void amdgpu_dm_update_connector_after_detect(
 		struct amdgpu_dm_connector *aconnector)
 {
 	struct drm_connector *connector = &aconnector->base;
+	struct dc_sink *sink __free(sink_release) = NULL;
 	struct drm_device *dev = connector->dev;
-	struct dc_sink *sink;
 
 	/* MST handled by drm_mst framework */
 	if (aconnector->mst_mgr.mst_state == true)
@@ -3039,7 +3813,7 @@ void amdgpu_dm_update_connector_after_detect(
 		 * For S3 resume with headless use eml_sink to fake stream
 		 * because on resume connector->sink is set to NULL
 		 */
-		mutex_lock(&dev->mode_config.mutex);
+		guard(mutex)(&dev->mode_config.mutex);
 
 		if (sink) {
 			if (aconnector->dc_sink) {
@@ -3055,7 +3829,7 @@ void amdgpu_dm_update_connector_after_detect(
 			aconnector->dc_sink = sink;
 			dc_sink_retain(aconnector->dc_sink);
 			amdgpu_dm_update_freesync_caps(connector,
-					aconnector->edid);
+					aconnector->drm_edid);
 		} else {
 			amdgpu_dm_update_freesync_caps(connector, NULL);
 			if (!aconnector->dc_sink) {
@@ -3064,10 +3838,6 @@ void amdgpu_dm_update_connector_after_detect(
 			}
 		}
 
-		mutex_unlock(&dev->mode_config.mutex);
-
-		if (sink)
-			dc_sink_release(sink);
 		return;
 	}
 
@@ -3075,27 +3845,25 @@ void amdgpu_dm_update_connector_after_detect(
 	 * TODO: temporary guard to look for proper fix
 	 * if this sink is MST sink, we should not do anything
 	 */
-	if (sink && sink->sink_signal == SIGNAL_TYPE_DISPLAY_PORT_MST) {
-		dc_sink_release(sink);
+	if (sink && sink->sink_signal == SIGNAL_TYPE_DISPLAY_PORT_MST)
 		return;
-	}
 
 	if (aconnector->dc_sink == sink) {
 		/*
 		 * We got a DP short pulse (Link Loss, DP CTS, etc...).
 		 * Do nothing!!
 		 */
-		DRM_DEBUG_DRIVER("DCHPD: connector_id=%d: dc_sink didn't change.\n",
-				aconnector->connector_id);
-		if (sink)
-			dc_sink_release(sink);
+		drm_dbg_kms(dev, "DCHPD: connector_id=%d: dc_sink didn't change.\n",
+				 aconnector->connector_id);
 		return;
 	}
 
-	DRM_DEBUG_DRIVER("DCHPD: connector_id=%d: Old sink=%p New sink=%p\n",
-		aconnector->connector_id, aconnector->dc_sink, sink);
+	drm_dbg_kms(dev, "DCHPD: connector_id=%d: Old sink=%p New sink=%p\n",
+		    aconnector->connector_id, aconnector->dc_sink, sink);
 
-	mutex_lock(&dev->mode_config.mutex);
+	/* When polling, DRM has already locked the mutex for us. */
+	if (!drm_kms_helper_is_poll_worker())
+		mutex_lock(&dev->mode_config.mutex);
 
 	/*
 	 * 1. Update status of the drm connector
@@ -3114,35 +3882,42 @@ void amdgpu_dm_update_connector_after_detect(
 		aconnector->dc_sink = sink;
 		dc_sink_retain(aconnector->dc_sink);
 		if (sink->dc_edid.length == 0) {
-			aconnector->edid = NULL;
+			aconnector->drm_edid = NULL;
+			hdmi_cec_unset_edid(aconnector);
 			if (aconnector->dc_link->aux_mode) {
-				drm_dp_cec_unset_edid(
-					&aconnector->dm_dp_aux.aux);
+				drm_dp_cec_unset_edid(&aconnector->dm_dp_aux.aux);
 			}
 		} else {
-			aconnector->edid =
-				(struct edid *)sink->dc_edid.raw_edid;
+			const struct edid *edid = (const struct edid *)sink->dc_edid.raw_edid;
+
+			aconnector->drm_edid = drm_edid_alloc(edid, sink->dc_edid.length);
+			drm_edid_connector_update(connector, aconnector->drm_edid);
 
+			hdmi_cec_set_edid(aconnector);
 			if (aconnector->dc_link->aux_mode)
-				drm_dp_cec_set_edid(&aconnector->dm_dp_aux.aux,
-						    aconnector->edid);
+				drm_dp_cec_attach(&aconnector->dm_dp_aux.aux,
+						  connector->display_info.source_physical_address);
 		}
 
-		aconnector->timing_requested = kzalloc(sizeof(struct dc_crtc_timing), GFP_KERNEL);
-		if (!aconnector->timing_requested)
-			dm_error("%s: failed to create aconnector->requested_timing\n", __func__);
+		if (!aconnector->timing_requested) {
+			aconnector->timing_requested =
+				kzalloc(sizeof(struct dc_crtc_timing), GFP_KERNEL);
+			if (!aconnector->timing_requested)
+				drm_err(dev,
+					"failed to create aconnector->requested_timing\n");
+		}
 
-		drm_connector_update_edid_property(connector, aconnector->edid);
-		amdgpu_dm_update_freesync_caps(connector, aconnector->edid);
+		amdgpu_dm_update_freesync_caps(connector, aconnector->drm_edid);
 		update_connector_ext_caps(aconnector);
 	} else {
+		hdmi_cec_unset_edid(aconnector);
 		drm_dp_cec_unset_edid(&aconnector->dm_dp_aux.aux);
 		amdgpu_dm_update_freesync_caps(connector, NULL);
-		drm_connector_update_edid_property(connector, NULL);
 		aconnector->num_modes = 0;
 		dc_sink_release(aconnector->dc_sink);
 		aconnector->dc_sink = NULL;
-		aconnector->edid = NULL;
+		drm_edid_free(aconnector->drm_edid);
+		aconnector->drm_edid = NULL;
 		kfree(aconnector->timing_requested);
 		aconnector->timing_requested = NULL;
 		/* Set CP to DESIRED if it was ENABLED, so we can re-enable it again on hotplug */
@@ -3150,12 +3925,102 @@ void amdgpu_dm_update_connector_after_detect(
 			connector->state->content_protection = DRM_MODE_CONTENT_PROTECTION_DESIRED;
 	}
 
-	mutex_unlock(&dev->mode_config.mutex);
-
 	update_subconnector_property(aconnector);
 
-	if (sink)
-		dc_sink_release(sink);
+	/* When polling, the mutex will be unlocked for us by DRM. */
+	if (!drm_kms_helper_is_poll_worker())
+		mutex_unlock(&dev->mode_config.mutex);
+}
+
+static bool are_sinks_equal(const struct dc_sink *sink1, const struct dc_sink *sink2)
+{
+	if (!sink1 || !sink2)
+		return false;
+	if (sink1->sink_signal != sink2->sink_signal)
+		return false;
+
+	if (sink1->dc_edid.length != sink2->dc_edid.length)
+		return false;
+
+	if (memcmp(sink1->dc_edid.raw_edid, sink2->dc_edid.raw_edid,
+		   sink1->dc_edid.length) != 0)
+		return false;
+	return true;
+}
+
+
+/**
+ * DOC: hdmi_hpd_debounce_work
+ *
+ * HDMI HPD debounce delay in milliseconds. When an HDMI display toggles HPD
+ * (such as during power save transitions), this delay determines how long to
+ * wait before processing the HPD event. This allows distinguishing between a
+ * physical unplug (>hdmi_hpd_debounce_delay)
+ * and a spontaneous RX HPD toggle (<hdmi_hpd_debounce_delay).
+ *
+ * If the toggle is less than this delay, the driver compares sink capabilities
+ * and permits a hotplug event if they changed.
+ *
+ * The default value of 1500ms was chosen based on experimental testing with
+ * various monitors that exhibit spontaneous HPD toggling behavior.
+ */
+static void hdmi_hpd_debounce_work(struct work_struct *work)
+{
+	struct amdgpu_dm_connector *aconnector =
+		container_of(to_delayed_work(work), struct amdgpu_dm_connector,
+			     hdmi_hpd_debounce_work);
+	struct drm_connector *connector = &aconnector->base;
+	struct drm_device *dev = connector->dev;
+	struct amdgpu_device *adev = drm_to_adev(dev);
+	struct dc *dc = aconnector->dc_link->ctx->dc;
+	bool fake_reconnect = false;
+	bool reallow_idle = false;
+	bool ret = false;
+	guard(mutex)(&aconnector->hpd_lock);
+
+	/* Re-detect the display */
+	scoped_guard(mutex, &adev->dm.dc_lock) {
+		if (dc->caps.ips_support && dc->ctx->dmub_srv->idle_allowed) {
+			dc_allow_idle_optimizations(dc, false);
+			reallow_idle = true;
+		}
+		ret = dc_link_detect(aconnector->dc_link, DETECT_REASON_HPD);
+	}
+
+	if (ret) {
+		/* Apply workaround delay for certain panels */
+		apply_delay_after_dpcd_poweroff(adev, aconnector->dc_sink);
+		/* Compare sinks to determine if this was a spontaneous HPD toggle */
+		if (are_sinks_equal(aconnector->dc_link->local_sink, aconnector->hdmi_prev_sink)) {
+			/*
+			* Sinks match - this was a spontaneous HDMI HPD toggle.
+			*/
+			drm_dbg_kms(dev, "HDMI HPD: Sink unchanged after debounce, internal re-enable\n");
+			fake_reconnect = true;
+		}
+
+		/* Update connector state */
+		amdgpu_dm_update_connector_after_detect(aconnector);
+
+		drm_modeset_lock_all(dev);
+		dm_restore_drm_connector_state(dev, connector);
+		drm_modeset_unlock_all(dev);
+
+		/* Only notify OS if sink actually changed */
+		if (!fake_reconnect && aconnector->base.force == DRM_FORCE_UNSPECIFIED)
+			drm_kms_helper_hotplug_event(dev);
+	}
+
+	/* Release the cached sink reference */
+	if (aconnector->hdmi_prev_sink) {
+		dc_sink_release(aconnector->hdmi_prev_sink);
+		aconnector->hdmi_prev_sink = NULL;
+	}
+
+	scoped_guard(mutex, &adev->dm.dc_lock) {
+		if (reallow_idle && dc->caps.ips_support)
+			dc_allow_idle_optimizations(dc, true);
+	}
 }
 
 static void handle_hpd_irq_helper(struct amdgpu_dm_connector *aconnector)
@@ -3165,7 +4030,9 @@ static void handle_hpd_irq_helper(struct amdgpu_dm_connector *aconnector)
 	enum dc_connection_type new_connection_type = dc_connection_none;
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct dm_connector_state *dm_con_state = to_dm_connector_state(connector->state);
+	struct dc *dc = aconnector->dc_link->ctx->dc;
 	bool ret = false;
+	bool debounce_required = false;
 
 	if (adev->dm.disable_hpd_irq)
 		return;
@@ -3174,7 +4041,7 @@ static void handle_hpd_irq_helper(struct amdgpu_dm_connector *aconnector)
 	 * In case of failure or MST no need to update connector status or notify the OS
 	 * since (for MST case) MST does this in its own context.
 	 */
-	mutex_lock(&aconnector->hpd_lock);
+	guard(mutex)(&aconnector->hpd_lock);
 
 	if (adev->dm.hdcp_workqueue) {
 		hdcp_reset_display(adev->dm.hdcp_workqueue, aconnector->dc_link->link_index);
@@ -3186,7 +4053,15 @@ static void handle_hpd_irq_helper(struct amdgpu_dm_connector *aconnector)
 	aconnector->timing_changed = false;
 
 	if (!dc_link_detect_connection_type(aconnector->dc_link, &new_connection_type))
-		DRM_ERROR("KMS: Failed to detect connector\n");
+		drm_err(adev_to_drm(adev), "KMS: Failed to detect connector\n");
+
+	/*
+	 * Check for HDMI disconnect with debounce enabled.
+	 */
+	debounce_required = (aconnector->hdmi_hpd_debounce_delay_ms > 0 &&
+			      dc_is_hdmi_signal(aconnector->dc_link->connector_signal) &&
+			      new_connection_type == dc_connection_none &&
+			      aconnector->dc_link->local_sink != NULL);
 
 	if (aconnector->base.force && new_connection_type == dc_connection_none) {
 		emulated_link_detect(aconnector->dc_link);
@@ -3197,11 +4072,41 @@ static void handle_hpd_irq_helper(struct amdgpu_dm_connector *aconnector)
 
 		if (aconnector->base.force == DRM_FORCE_UNSPECIFIED)
 			drm_kms_helper_connector_hotplug_event(connector);
+	} else if (debounce_required) {
+		/*
+		 * HDMI disconnect detected - schedule delayed work instead of
+		 * processing immediately. This allows us to coalesce spurious
+		 * HDMI signals from physical unplugs.
+		 */
+		drm_dbg_kms(dev, "HDMI HPD: Disconnect detected, scheduling debounce work (%u ms)\n",
+			    aconnector->hdmi_hpd_debounce_delay_ms);
+
+		/* Cache the current sink for later comparison */
+		if (aconnector->hdmi_prev_sink)
+			dc_sink_release(aconnector->hdmi_prev_sink);
+		aconnector->hdmi_prev_sink = aconnector->dc_link->local_sink;
+		if (aconnector->hdmi_prev_sink)
+			dc_sink_retain(aconnector->hdmi_prev_sink);
+
+		/* Schedule delayed detection. */
+		if (mod_delayed_work(system_wq,
+				 &aconnector->hdmi_hpd_debounce_work,
+				 msecs_to_jiffies(aconnector->hdmi_hpd_debounce_delay_ms)))
+			drm_dbg_kms(dev, "HDMI HPD: Re-scheduled debounce work\n");
+
 	} else {
-		mutex_lock(&adev->dm.dc_lock);
-		ret = dc_link_detect(aconnector->dc_link, DETECT_REASON_HPD);
-		mutex_unlock(&adev->dm.dc_lock);
+
+		/* If the aconnector->hdmi_hpd_debounce_work is scheduled, exit early */
+		if (delayed_work_pending(&aconnector->hdmi_hpd_debounce_work))
+			return;
+
+		scoped_guard(mutex, &adev->dm.dc_lock) {
+			dc_exit_ips_for_hw_access(dc);
+			ret = dc_link_detect(aconnector->dc_link, DETECT_REASON_HPD);
+		}
 		if (ret) {
+			/* w/a delay for certain panels */
+			apply_delay_after_dpcd_poweroff(adev, aconnector->dc_sink);
 			amdgpu_dm_update_connector_after_detect(aconnector);
 
 			drm_modeset_lock_all(dev);
@@ -3212,8 +4117,6 @@ static void handle_hpd_irq_helper(struct amdgpu_dm_connector *aconnector)
 				drm_kms_helper_connector_hotplug_event(connector);
 		}
 	}
-	mutex_unlock(&aconnector->hpd_lock);
-
 }
 
 static void handle_hpd_irq(void *param)
@@ -3224,98 +4127,21 @@ static void handle_hpd_irq(void *param)
 
 }
 
-static void dm_handle_mst_sideband_msg(struct amdgpu_dm_connector *aconnector)
-{
-	u8 esi[DP_PSR_ERROR_STATUS - DP_SINK_COUNT_ESI] = { 0 };
-	u8 dret;
-	bool new_irq_handled = false;
-	int dpcd_addr;
-	int dpcd_bytes_to_read;
-
-	const int max_process_count = 30;
-	int process_count = 0;
-
-	const struct dc_link_status *link_status = dc_link_get_status(aconnector->dc_link);
-
-	if (link_status->dpcd_caps->dpcd_rev.raw < 0x12) {
-		dpcd_bytes_to_read = DP_LANE0_1_STATUS - DP_SINK_COUNT;
-		/* DPCD 0x200 - 0x201 for downstream IRQ */
-		dpcd_addr = DP_SINK_COUNT;
-	} else {
-		dpcd_bytes_to_read = DP_PSR_ERROR_STATUS - DP_SINK_COUNT_ESI;
-		/* DPCD 0x2002 - 0x2005 for downstream IRQ */
-		dpcd_addr = DP_SINK_COUNT_ESI;
-	}
-
-	dret = drm_dp_dpcd_read(
-		&aconnector->dm_dp_aux.aux,
-		dpcd_addr,
-		esi,
-		dpcd_bytes_to_read);
-
-	while (dret == dpcd_bytes_to_read &&
-		process_count < max_process_count) {
-		u8 retry;
-		dret = 0;
-
-		process_count++;
-
-		DRM_DEBUG_DRIVER("ESI %02x %02x %02x\n", esi[0], esi[1], esi[2]);
-		/* handle HPD short pulse irq */
-		if (aconnector->mst_mgr.mst_state)
-			drm_dp_mst_hpd_irq(
-				&aconnector->mst_mgr,
-				esi,
-				&new_irq_handled);
-
-		if (new_irq_handled) {
-			/* ACK at DPCD to notify down stream */
-			const int ack_dpcd_bytes_to_write =
-				dpcd_bytes_to_read - 1;
-
-			for (retry = 0; retry < 3; retry++) {
-				u8 wret;
-
-				wret = drm_dp_dpcd_write(
-					&aconnector->dm_dp_aux.aux,
-					dpcd_addr + 1,
-					&esi[1],
-					ack_dpcd_bytes_to_write);
-				if (wret == ack_dpcd_bytes_to_write)
-					break;
-			}
-
-			/* check if there is new irq to be handled */
-			dret = drm_dp_dpcd_read(
-				&aconnector->dm_dp_aux.aux,
-				dpcd_addr,
-				esi,
-				dpcd_bytes_to_read);
-
-			new_irq_handled = false;
-		} else {
-			break;
-		}
-	}
-
-	if (process_count == max_process_count)
-		DRM_DEBUG_DRIVER("Loop exceeded max iterations\n");
-}
-
-static void schedule_hpd_rx_offload_work(struct hpd_rx_irq_offload_work_queue *offload_wq,
+static void schedule_hpd_rx_offload_work(struct amdgpu_device *adev, struct hpd_rx_irq_offload_work_queue *offload_wq,
 							union hpd_irq_data hpd_irq_data)
 {
 	struct hpd_rx_irq_offload_work *offload_work =
 				kzalloc(sizeof(*offload_work), GFP_KERNEL);
 
 	if (!offload_work) {
-		DRM_ERROR("Failed to allocate hpd_rx_irq_offload_work.\n");
+		drm_err(adev_to_drm(adev), "Failed to allocate hpd_rx_irq_offload_work.\n");
 		return;
 	}
 
 	INIT_WORK(&offload_work->work, dm_handle_hpd_rx_offload_work);
 	offload_work->data = hpd_irq_data;
 	offload_work->offload_wq = offload_wq;
+	offload_work->adev = adev;
 
 	queue_work(offload_wq->wq, &offload_work->work);
 	DRM_DEBUG_KMS("queue work to handle hpd_rx offload work");
@@ -3336,6 +4162,7 @@ static void handle_hpd_rx_irq(void *param)
 	bool has_left_work = false;
 	int idx = dc_link->link_index;
 	struct hpd_rx_irq_offload_work_queue *offload_wq = &adev->dm.hpd_rx_offload_wq[idx];
+	struct dc *dc = aconnector->dc_link->ctx->dc;
 
 	memset(&hpd_irq_data, 0, sizeof(hpd_irq_data));
 
@@ -3356,14 +4183,30 @@ static void handle_hpd_rx_irq(void *param)
 		goto out;
 
 	if (hpd_irq_data.bytes.device_service_irq.bits.AUTOMATED_TEST) {
-		schedule_hpd_rx_offload_work(offload_wq, hpd_irq_data);
+		schedule_hpd_rx_offload_work(adev, offload_wq, hpd_irq_data);
 		goto out;
 	}
 
 	if (dc_link_dp_allow_hpd_rx_irq(dc_link)) {
 		if (hpd_irq_data.bytes.device_service_irq.bits.UP_REQ_MSG_RDY ||
 			hpd_irq_data.bytes.device_service_irq.bits.DOWN_REP_MSG_RDY) {
-			dm_handle_mst_sideband_msg(aconnector);
+			bool skip = false;
+
+			/*
+			 * DOWN_REP_MSG_RDY is also handled by polling method
+			 * mgr->cbs->poll_hpd_irq()
+			 */
+			spin_lock(&offload_wq->offload_lock);
+			skip = offload_wq->is_handling_mst_msg_rdy_event;
+
+			if (!skip)
+				offload_wq->is_handling_mst_msg_rdy_event = true;
+
+			spin_unlock(&offload_wq->offload_lock);
+
+			if (!skip)
+				schedule_hpd_rx_offload_work(adev, offload_wq, hpd_irq_data);
+
 			goto out;
 		}
 
@@ -3379,7 +4222,7 @@ static void handle_hpd_rx_irq(void *param)
 			spin_unlock(&offload_wq->offload_lock);
 
 			if (!skip)
-				schedule_hpd_rx_offload_work(offload_wq, hpd_irq_data);
+				schedule_hpd_rx_offload_work(adev, offload_wq, hpd_irq_data);
 
 			goto out;
 		}
@@ -3389,7 +4232,7 @@ out:
 	if (result && !is_mst_root_connector) {
 		/* Downstream Port status changed. */
 		if (!dc_link_detect_connection_type(dc_link, &new_connection_type))
-			DRM_ERROR("KMS: Failed to detect connector\n");
+			drm_err(adev_to_drm(adev), "KMS: Failed to detect connector\n");
 
 		if (aconnector->base.force && new_connection_type == dc_connection_none) {
 			emulated_link_detect(dc_link);
@@ -3409,6 +4252,7 @@ out:
 			bool ret = false;
 
 			mutex_lock(&adev->dm.dc_lock);
+			dc_exit_ips_for_hw_access(dc);
 			ret = dc_link_detect(dc_link, DETECT_REASON_HPDRX);
 			mutex_unlock(&adev->dm.dc_lock);
 
@@ -3437,7 +4281,7 @@ out:
 	mutex_unlock(&aconnector->hpd_lock);
 }
 
-static void register_hpd_handlers(struct amdgpu_device *adev)
+static int register_hpd_handlers(struct amdgpu_device *adev)
 {
 	struct drm_device *dev = adev_to_drm(adev);
 	struct drm_connector *connector;
@@ -3448,36 +4292,70 @@ static void register_hpd_handlers(struct amdgpu_device *adev)
 	int_params.requested_polarity = INTERRUPT_POLARITY_DEFAULT;
 	int_params.current_polarity = INTERRUPT_POLARITY_DEFAULT;
 
+	if (dc_is_dmub_outbox_supported(adev->dm.dc)) {
+		if (!register_dmub_notify_callback(adev, DMUB_NOTIFICATION_HPD,
+			dmub_hpd_callback, true)) {
+			drm_err(adev_to_drm(adev), "fail to register dmub hpd callback");
+			return -EINVAL;
+		}
+
+		if (!register_dmub_notify_callback(adev, DMUB_NOTIFICATION_HPD_IRQ,
+			dmub_hpd_callback, true)) {
+			drm_err(adev_to_drm(adev), "fail to register dmub hpd callback");
+			return -EINVAL;
+		}
+
+		if (!register_dmub_notify_callback(adev, DMUB_NOTIFICATION_HPD_SENSE_NOTIFY,
+			dmub_hpd_sense_callback, true)) {
+			drm_err(adev_to_drm(adev), "fail to register dmub hpd sense callback");
+			return -EINVAL;
+		}
+	}
+
 	list_for_each_entry(connector,
 			&dev->mode_config.connector_list, head)	{
 
+		if (connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+			continue;
+
 		aconnector = to_amdgpu_dm_connector(connector);
 		dc_link = aconnector->dc_link;
 
-		if (DC_IRQ_SOURCE_INVALID != dc_link->irq_source_hpd) {
+		if (dc_link->irq_source_hpd != DC_IRQ_SOURCE_INVALID) {
 			int_params.int_context = INTERRUPT_LOW_IRQ_CONTEXT;
 			int_params.irq_source = dc_link->irq_source_hpd;
 
-			amdgpu_dm_irq_register_interrupt(adev, &int_params,
-					handle_hpd_irq,
-					(void *) aconnector);
+			if (int_params.irq_source == DC_IRQ_SOURCE_INVALID ||
+				int_params.irq_source  < DC_IRQ_SOURCE_HPD1 ||
+				int_params.irq_source  > DC_IRQ_SOURCE_HPD6) {
+				drm_err(adev_to_drm(adev), "Failed to register hpd irq!\n");
+				return -EINVAL;
+			}
+
+			if (!amdgpu_dm_irq_register_interrupt(adev, &int_params,
+				handle_hpd_irq, (void *) aconnector))
+				return -ENOMEM;
 		}
 
-		if (DC_IRQ_SOURCE_INVALID != dc_link->irq_source_hpd_rx) {
+		if (dc_link->irq_source_hpd_rx != DC_IRQ_SOURCE_INVALID) {
 
 			/* Also register for DP short pulse (hpd_rx). */
 			int_params.int_context = INTERRUPT_LOW_IRQ_CONTEXT;
 			int_params.irq_source =	dc_link->irq_source_hpd_rx;
 
-			amdgpu_dm_irq_register_interrupt(adev, &int_params,
-					handle_hpd_rx_irq,
-					(void *) aconnector);
+			if (int_params.irq_source == DC_IRQ_SOURCE_INVALID ||
+				int_params.irq_source  < DC_IRQ_SOURCE_HPD1RX ||
+				int_params.irq_source  > DC_IRQ_SOURCE_HPD6RX) {
+				drm_err(adev_to_drm(adev), "Failed to register hpd rx irq!\n");
+				return -EINVAL;
+			}
 
-			if (adev->dm.hpd_rx_offload_wq)
-				adev->dm.hpd_rx_offload_wq[dc_link->link_index].aconnector =
-					aconnector;
+			if (!amdgpu_dm_irq_register_interrupt(adev, &int_params,
+				handle_hpd_rx_irq, (void *) aconnector))
+				return -ENOMEM;
 		}
 	}
+	return 0;
 }
 
 #if defined(CONFIG_DRM_AMD_DC_SI)
@@ -3489,7 +4367,7 @@ static int dce60_register_irq_handlers(struct amdgpu_device *adev)
 	struct dc_interrupt_params int_params = {0};
 	int r;
 	int i;
-	unsigned client_id = AMDGPU_IRQ_CLIENTID_LEGACY;
+	unsigned int client_id = AMDGPU_IRQ_CLIENTID_LEGACY;
 
 	int_params.requested_polarity = INTERRUPT_POLARITY_DEFAULT;
 	int_params.current_polarity = INTERRUPT_POLARITY_DEFAULT;
@@ -3503,27 +4381,36 @@ static int dce60_register_irq_handlers(struct amdgpu_device *adev)
 	 *    Base driver will call amdgpu_dm_irq_handler() for ALL interrupts
 	 *    coming from DC hardware.
 	 *    amdgpu_dm_irq_handler() will re-direct the interrupt to DC
-	 *    for acknowledging and handling. */
+	 *    for acknowledging and handling.
+	 */
 
 	/* Use VBLANK interrupt */
 	for (i = 0; i < adev->mode_info.num_crtc; i++) {
-		r = amdgpu_irq_add_id(adev, client_id, i+1 , &adev->crtc_irq);
+		r = amdgpu_irq_add_id(adev, client_id, i + 1, &adev->crtc_irq);
 		if (r) {
-			DRM_ERROR("Failed to add crtc irq id!\n");
+			drm_err(adev_to_drm(adev), "Failed to add crtc irq id!\n");
 			return r;
 		}
 
 		int_params.int_context = INTERRUPT_HIGH_IRQ_CONTEXT;
 		int_params.irq_source =
-			dc_interrupt_to_irq_source(dc, i+1 , 0);
+			dc_interrupt_to_irq_source(dc, i + 1, 0);
+
+		if (int_params.irq_source == DC_IRQ_SOURCE_INVALID ||
+			int_params.irq_source  < DC_IRQ_SOURCE_VBLANK1 ||
+			int_params.irq_source  > DC_IRQ_SOURCE_VBLANK6) {
+			drm_err(adev_to_drm(adev), "Failed to register vblank irq!\n");
+			return -EINVAL;
+		}
 
 		c_irq_params = &adev->dm.vblank_params[int_params.irq_source - DC_IRQ_SOURCE_VBLANK1];
 
 		c_irq_params->adev = adev;
 		c_irq_params->irq_src = int_params.irq_source;
 
-		amdgpu_dm_irq_register_interrupt(adev, &int_params,
-				dm_crtc_high_irq, c_irq_params);
+		if (!amdgpu_dm_irq_register_interrupt(adev, &int_params,
+			dm_crtc_high_irq, c_irq_params))
+			return -ENOMEM;
 	}
 
 	/* Use GRPH_PFLIP interrupt */
@@ -3531,7 +4418,7 @@ static int dce60_register_irq_handlers(struct amdgpu_device *adev)
 			i <= VISLANDS30_IV_SRCID_D6_GRPH_PFLIP; i += 2) {
 		r = amdgpu_irq_add_id(adev, client_id, i, &adev->pageflip_irq);
 		if (r) {
-			DRM_ERROR("Failed to add page flip irq id!\n");
+			drm_err(adev_to_drm(adev), "Failed to add page flip irq id!\n");
 			return r;
 		}
 
@@ -3539,27 +4426,34 @@ static int dce60_register_irq_handlers(struct amdgpu_device *adev)
 		int_params.irq_source =
 			dc_interrupt_to_irq_source(dc, i, 0);
 
+		if (int_params.irq_source == DC_IRQ_SOURCE_INVALID ||
+			int_params.irq_source  < DC_IRQ_SOURCE_PFLIP_FIRST ||
+			int_params.irq_source  > DC_IRQ_SOURCE_PFLIP_LAST) {
+			drm_err(adev_to_drm(adev), "Failed to register pflip irq!\n");
+			return -EINVAL;
+		}
+
 		c_irq_params = &adev->dm.pflip_params[int_params.irq_source - DC_IRQ_SOURCE_PFLIP_FIRST];
 
 		c_irq_params->adev = adev;
 		c_irq_params->irq_src = int_params.irq_source;
 
-		amdgpu_dm_irq_register_interrupt(adev, &int_params,
-				dm_pflip_high_irq, c_irq_params);
-
+		if (!amdgpu_dm_irq_register_interrupt(adev, &int_params,
+			dm_pflip_high_irq, c_irq_params))
+			return -ENOMEM;
 	}
 
 	/* HPD */
 	r = amdgpu_irq_add_id(adev, client_id,
 			VISLANDS30_IV_SRCID_HOTPLUG_DETECT_A, &adev->hpd_irq);
 	if (r) {
-		DRM_ERROR("Failed to add hpd irq id!\n");
+		drm_err(adev_to_drm(adev), "Failed to add hpd irq id!\n");
 		return r;
 	}
 
-	register_hpd_handlers(adev);
+	r = register_hpd_handlers(adev);
 
-	return 0;
+	return r;
 }
 #endif
 
@@ -3571,7 +4465,7 @@ static int dce110_register_irq_handlers(struct amdgpu_device *adev)
 	struct dc_interrupt_params int_params = {0};
 	int r;
 	int i;
-	unsigned client_id = AMDGPU_IRQ_CLIENTID_LEGACY;
+	unsigned int client_id = AMDGPU_IRQ_CLIENTID_LEGACY;
 
 	if (adev->family >= AMDGPU_FAMILY_AI)
 		client_id = SOC15_IH_CLIENTID_DCE;
@@ -3588,13 +4482,14 @@ static int dce110_register_irq_handlers(struct amdgpu_device *adev)
 	 *    Base driver will call amdgpu_dm_irq_handler() for ALL interrupts
 	 *    coming from DC hardware.
 	 *    amdgpu_dm_irq_handler() will re-direct the interrupt to DC
-	 *    for acknowledging and handling. */
+	 *    for acknowledging and handling.
+	 */
 
 	/* Use VBLANK interrupt */
 	for (i = VISLANDS30_IV_SRCID_D1_VERTICAL_INTERRUPT0; i <= VISLANDS30_IV_SRCID_D6_VERTICAL_INTERRUPT0; i++) {
 		r = amdgpu_irq_add_id(adev, client_id, i, &adev->crtc_irq);
 		if (r) {
-			DRM_ERROR("Failed to add crtc irq id!\n");
+			drm_err(adev_to_drm(adev), "Failed to add crtc irq id!\n");
 			return r;
 		}
 
@@ -3602,20 +4497,28 @@ static int dce110_register_irq_handlers(struct amdgpu_device *adev)
 		int_params.irq_source =
 			dc_interrupt_to_irq_source(dc, i, 0);
 
+		if (int_params.irq_source == DC_IRQ_SOURCE_INVALID ||
+			int_params.irq_source  < DC_IRQ_SOURCE_VBLANK1 ||
+			int_params.irq_source  > DC_IRQ_SOURCE_VBLANK6) {
+			drm_err(adev_to_drm(adev), "Failed to register vblank irq!\n");
+			return -EINVAL;
+		}
+
 		c_irq_params = &adev->dm.vblank_params[int_params.irq_source - DC_IRQ_SOURCE_VBLANK1];
 
 		c_irq_params->adev = adev;
 		c_irq_params->irq_src = int_params.irq_source;
 
-		amdgpu_dm_irq_register_interrupt(adev, &int_params,
-				dm_crtc_high_irq, c_irq_params);
+		if (!amdgpu_dm_irq_register_interrupt(adev, &int_params,
+			dm_crtc_high_irq, c_irq_params))
+			return -ENOMEM;
 	}
 
 	/* Use VUPDATE interrupt */
 	for (i = VISLANDS30_IV_SRCID_D1_V_UPDATE_INT; i <= VISLANDS30_IV_SRCID_D6_V_UPDATE_INT; i += 2) {
 		r = amdgpu_irq_add_id(adev, client_id, i, &adev->vupdate_irq);
 		if (r) {
-			DRM_ERROR("Failed to add vupdate irq id!\n");
+			drm_err(adev_to_drm(adev), "Failed to add vupdate irq id!\n");
 			return r;
 		}
 
@@ -3623,13 +4526,21 @@ static int dce110_register_irq_handlers(struct amdgpu_device *adev)
 		int_params.irq_source =
 			dc_interrupt_to_irq_source(dc, i, 0);
 
+		if (int_params.irq_source == DC_IRQ_SOURCE_INVALID ||
+			int_params.irq_source  < DC_IRQ_SOURCE_VUPDATE1 ||
+			int_params.irq_source  > DC_IRQ_SOURCE_VUPDATE6) {
+			drm_err(adev_to_drm(adev), "Failed to register vupdate irq!\n");
+			return -EINVAL;
+		}
+
 		c_irq_params = &adev->dm.vupdate_params[int_params.irq_source - DC_IRQ_SOURCE_VUPDATE1];
 
 		c_irq_params->adev = adev;
 		c_irq_params->irq_src = int_params.irq_source;
 
-		amdgpu_dm_irq_register_interrupt(adev, &int_params,
-				dm_vupdate_high_irq, c_irq_params);
+		if (!amdgpu_dm_irq_register_interrupt(adev, &int_params,
+			dm_vupdate_high_irq, c_irq_params))
+			return -ENOMEM;
 	}
 
 	/* Use GRPH_PFLIP interrupt */
@@ -3637,7 +4548,7 @@ static int dce110_register_irq_handlers(struct amdgpu_device *adev)
 			i <= VISLANDS30_IV_SRCID_D6_GRPH_PFLIP; i += 2) {
 		r = amdgpu_irq_add_id(adev, client_id, i, &adev->pageflip_irq);
 		if (r) {
-			DRM_ERROR("Failed to add page flip irq id!\n");
+			drm_err(adev_to_drm(adev), "Failed to add page flip irq id!\n");
 			return r;
 		}
 
@@ -3645,27 +4556,34 @@ static int dce110_register_irq_handlers(struct amdgpu_device *adev)
 		int_params.irq_source =
 			dc_interrupt_to_irq_source(dc, i, 0);
 
+		if (int_params.irq_source == DC_IRQ_SOURCE_INVALID ||
+			int_params.irq_source  < DC_IRQ_SOURCE_PFLIP_FIRST ||
+			int_params.irq_source  > DC_IRQ_SOURCE_PFLIP_LAST) {
+			drm_err(adev_to_drm(adev), "Failed to register pflip irq!\n");
+			return -EINVAL;
+		}
+
 		c_irq_params = &adev->dm.pflip_params[int_params.irq_source - DC_IRQ_SOURCE_PFLIP_FIRST];
 
 		c_irq_params->adev = adev;
 		c_irq_params->irq_src = int_params.irq_source;
 
-		amdgpu_dm_irq_register_interrupt(adev, &int_params,
-				dm_pflip_high_irq, c_irq_params);
-
+		if (!amdgpu_dm_irq_register_interrupt(adev, &int_params,
+			dm_pflip_high_irq, c_irq_params))
+			return -ENOMEM;
 	}
 
 	/* HPD */
 	r = amdgpu_irq_add_id(adev, client_id,
 			VISLANDS30_IV_SRCID_HOTPLUG_DETECT_A, &adev->hpd_irq);
 	if (r) {
-		DRM_ERROR("Failed to add hpd irq id!\n");
+		drm_err(adev_to_drm(adev), "Failed to add hpd irq id!\n");
 		return r;
 	}
 
-	register_hpd_handlers(adev);
+	r = register_hpd_handlers(adev);
 
-	return 0;
+	return r;
 }
 
 /* Register IRQ sources and initialize IRQ callbacks */
@@ -3709,7 +4627,7 @@ static int dcn10_register_irq_handlers(struct amdgpu_device *adev)
 		r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_DCE, i, &adev->crtc_irq);
 
 		if (r) {
-			DRM_ERROR("Failed to add crtc irq id!\n");
+			drm_err(adev_to_drm(adev), "Failed to add crtc irq id!\n");
 			return r;
 		}
 
@@ -3717,13 +4635,21 @@ static int dcn10_register_irq_handlers(struct amdgpu_device *adev)
 		int_params.irq_source =
 			dc_interrupt_to_irq_source(dc, i, 0);
 
+		if (int_params.irq_source == DC_IRQ_SOURCE_INVALID ||
+			int_params.irq_source  < DC_IRQ_SOURCE_VBLANK1 ||
+			int_params.irq_source  > DC_IRQ_SOURCE_VBLANK6) {
+			drm_err(adev_to_drm(adev), "Failed to register vblank irq!\n");
+			return -EINVAL;
+		}
+
 		c_irq_params = &adev->dm.vblank_params[int_params.irq_source - DC_IRQ_SOURCE_VBLANK1];
 
 		c_irq_params->adev = adev;
 		c_irq_params->irq_src = int_params.irq_source;
 
-		amdgpu_dm_irq_register_interrupt(
-			adev, &int_params, dm_crtc_high_irq, c_irq_params);
+		if (!amdgpu_dm_irq_register_interrupt(adev, &int_params,
+			dm_crtc_high_irq, c_irq_params))
+			return -ENOMEM;
 	}
 
 	/* Use otg vertical line interrupt */
@@ -3733,7 +4659,7 @@ static int dcn10_register_irq_handlers(struct amdgpu_device *adev)
 				vrtl_int_srcid[i], &adev->vline0_irq);
 
 		if (r) {
-			DRM_ERROR("Failed to add vline0 irq id!\n");
+			drm_err(adev_to_drm(adev), "Failed to add vline0 irq id!\n");
 			return r;
 		}
 
@@ -3741,9 +4667,11 @@ static int dcn10_register_irq_handlers(struct amdgpu_device *adev)
 		int_params.irq_source =
 			dc_interrupt_to_irq_source(dc, vrtl_int_srcid[i], 0);
 
-		if (int_params.irq_source == DC_IRQ_SOURCE_INVALID) {
-			DRM_ERROR("Failed to register vline0 irq %d!\n", vrtl_int_srcid[i]);
-			break;
+		if (int_params.irq_source == DC_IRQ_SOURCE_INVALID ||
+			int_params.irq_source < DC_IRQ_SOURCE_DC1_VLINE0 ||
+			int_params.irq_source > DC_IRQ_SOURCE_DC6_VLINE0) {
+			drm_err(adev_to_drm(adev), "Failed to register vline0 irq!\n");
+			return -EINVAL;
 		}
 
 		c_irq_params = &adev->dm.vline0_params[int_params.irq_source
@@ -3752,8 +4680,10 @@ static int dcn10_register_irq_handlers(struct amdgpu_device *adev)
 		c_irq_params->adev = adev;
 		c_irq_params->irq_src = int_params.irq_source;
 
-		amdgpu_dm_irq_register_interrupt(adev, &int_params,
-				dm_dcn_vertical_interrupt0_high_irq, c_irq_params);
+		if (!amdgpu_dm_irq_register_interrupt(adev, &int_params,
+			dm_dcn_vertical_interrupt0_high_irq,
+			c_irq_params))
+			return -ENOMEM;
 	}
 #endif
 
@@ -3768,7 +4698,7 @@ static int dcn10_register_irq_handlers(struct amdgpu_device *adev)
 		r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_DCE, i, &adev->vupdate_irq);
 
 		if (r) {
-			DRM_ERROR("Failed to add vupdate irq id!\n");
+			drm_err(adev_to_drm(adev), "Failed to add vupdate irq id!\n");
 			return r;
 		}
 
@@ -3776,13 +4706,21 @@ static int dcn10_register_irq_handlers(struct amdgpu_device *adev)
 		int_params.irq_source =
 			dc_interrupt_to_irq_source(dc, i, 0);
 
+		if (int_params.irq_source == DC_IRQ_SOURCE_INVALID ||
+			int_params.irq_source  < DC_IRQ_SOURCE_VUPDATE1 ||
+			int_params.irq_source  > DC_IRQ_SOURCE_VUPDATE6) {
+			drm_err(adev_to_drm(adev), "Failed to register vupdate irq!\n");
+			return -EINVAL;
+		}
+
 		c_irq_params = &adev->dm.vupdate_params[int_params.irq_source - DC_IRQ_SOURCE_VUPDATE1];
 
 		c_irq_params->adev = adev;
 		c_irq_params->irq_src = int_params.irq_source;
 
-		amdgpu_dm_irq_register_interrupt(adev, &int_params,
-				dm_vupdate_high_irq, c_irq_params);
+		if (!amdgpu_dm_irq_register_interrupt(adev, &int_params,
+			dm_vupdate_high_irq, c_irq_params))
+			return -ENOMEM;
 	}
 
 	/* Use GRPH_PFLIP interrupt */
@@ -3791,7 +4729,7 @@ static int dcn10_register_irq_handlers(struct amdgpu_device *adev)
 			i++) {
 		r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_DCE, i, &adev->pageflip_irq);
 		if (r) {
-			DRM_ERROR("Failed to add page flip irq id!\n");
+			drm_err(adev_to_drm(adev), "Failed to add page flip irq id!\n");
 			return r;
 		}
 
@@ -3799,27 +4737,34 @@ static int dcn10_register_irq_handlers(struct amdgpu_device *adev)
 		int_params.irq_source =
 			dc_interrupt_to_irq_source(dc, i, 0);
 
+		if (int_params.irq_source == DC_IRQ_SOURCE_INVALID ||
+			int_params.irq_source  < DC_IRQ_SOURCE_PFLIP_FIRST ||
+			int_params.irq_source  > DC_IRQ_SOURCE_PFLIP_LAST) {
+			drm_err(adev_to_drm(adev), "Failed to register pflip irq!\n");
+			return -EINVAL;
+		}
+
 		c_irq_params = &adev->dm.pflip_params[int_params.irq_source - DC_IRQ_SOURCE_PFLIP_FIRST];
 
 		c_irq_params->adev = adev;
 		c_irq_params->irq_src = int_params.irq_source;
 
-		amdgpu_dm_irq_register_interrupt(adev, &int_params,
-				dm_pflip_high_irq, c_irq_params);
-
+		if (!amdgpu_dm_irq_register_interrupt(adev, &int_params,
+			dm_pflip_high_irq, c_irq_params))
+			return -ENOMEM;
 	}
 
 	/* HPD */
 	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_DCE, DCN_1_0__SRCID__DC_HPD1_INT,
 			&adev->hpd_irq);
 	if (r) {
-		DRM_ERROR("Failed to add hpd irq id!\n");
+		drm_err(adev_to_drm(adev), "Failed to add hpd irq id!\n");
 		return r;
 	}
 
-	register_hpd_handlers(adev);
+	r = register_hpd_handlers(adev);
 
-	return 0;
+	return r;
 }
 /* Register Outbox IRQ sources and initialize IRQ callbacks */
 static int register_outbox_irq_handlers(struct amdgpu_device *adev)
@@ -3835,7 +4780,7 @@ static int register_outbox_irq_handlers(struct amdgpu_device *adev)
 	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_DCE, DCN_1_0__SRCID__DMCUB_OUTBOX_LOW_PRIORITY_READY_INT,
 			&adev->dmub_outbox_irq);
 	if (r) {
-		DRM_ERROR("Failed to add outbox irq id!\n");
+		drm_err(adev_to_drm(adev), "Failed to add outbox irq id!\n");
 		return r;
 	}
 
@@ -3850,8 +4795,9 @@ static int register_outbox_irq_handlers(struct amdgpu_device *adev)
 		c_irq_params->adev = adev;
 		c_irq_params->irq_src = int_params.irq_source;
 
-		amdgpu_dm_irq_register_interrupt(adev, &int_params,
-				dm_dmub_outbox1_low_irq, c_irq_params);
+		if (!amdgpu_dm_irq_register_interrupt(adev, &int_params,
+			dm_dmub_outbox1_low_irq, c_irq_params))
+			return -ENOMEM;
 	}
 
 	return 0;
@@ -3915,7 +4861,7 @@ dm_atomic_duplicate_state(struct drm_private_obj *obj)
 	old_state = to_dm_atomic_state(obj->state);
 
 	if (old_state && old_state->context)
-		new_state->context = dc_copy_state(old_state->context);
+		new_state->context = dc_state_create_copy(old_state->context);
 
 	if (!new_state->context) {
 		kfree(new_state);
@@ -3931,7 +4877,7 @@ static void dm_atomic_destroy_state(struct drm_private_obj *obj,
 	struct dm_atomic_state *dm_state = to_dm_atomic_state(state);
 
 	if (dm_state && dm_state->context)
-		dc_release_state(dm_state->context);
+		dc_state_release(dm_state->context);
 
 	kfree(dm_state);
 }
@@ -3967,14 +4913,12 @@ static int amdgpu_dm_mode_config_init(struct amdgpu_device *adev)
 	if (!state)
 		return -ENOMEM;
 
-	state->context = dc_create_state(adev->dm.dc);
+	state->context = dc_state_create_current_copy(adev->dm.dc);
 	if (!state->context) {
 		kfree(state);
 		return -ENOMEM;
 	}
 
-	dc_resource_state_copy_construct_current(adev->dm.dc, state->context);
-
 	drm_atomic_private_obj_init(adev_to_drm(adev),
 				    &adev->dm.atomic_obj,
 				    &state->base,
@@ -3982,14 +4926,22 @@ static int amdgpu_dm_mode_config_init(struct amdgpu_device *adev)
 
 	r = amdgpu_display_modeset_create_props(adev);
 	if (r) {
-		dc_release_state(state->context);
+		dc_state_release(state->context);
 		kfree(state);
 		return r;
 	}
 
+#ifdef AMD_PRIVATE_COLOR
+	if (amdgpu_dm_create_color_properties(adev)) {
+		dc_state_release(state->context);
+		kfree(state);
+		return -ENOMEM;
+	}
+#endif
+
 	r = amdgpu_dm_audio_init(adev);
 	if (r) {
-		dc_release_state(state->context);
+		dc_state_release(state->context);
 		kfree(state);
 		return r;
 	}
@@ -3999,43 +4951,51 @@ static int amdgpu_dm_mode_config_init(struct amdgpu_device *adev)
 
 #define AMDGPU_DM_DEFAULT_MIN_BACKLIGHT 12
 #define AMDGPU_DM_DEFAULT_MAX_BACKLIGHT 255
+#define AMDGPU_DM_MIN_SPREAD ((AMDGPU_DM_DEFAULT_MAX_BACKLIGHT - AMDGPU_DM_DEFAULT_MIN_BACKLIGHT) / 2)
 #define AUX_BL_DEFAULT_TRANSITION_TIME_MS 50
 
 static void amdgpu_dm_update_backlight_caps(struct amdgpu_display_manager *dm,
 					    int bl_idx)
 {
+	struct amdgpu_dm_backlight_caps *caps = &dm->backlight_caps[bl_idx];
+
+	if (caps->caps_valid)
+		return;
+
 #if defined(CONFIG_ACPI)
-	struct amdgpu_dm_backlight_caps caps;
+	amdgpu_acpi_get_backlight_caps(caps);
 
-	memset(&caps, 0, sizeof(caps));
+	/* validate the firmware value is sane */
+	if (caps->caps_valid) {
+		int spread = caps->max_input_signal - caps->min_input_signal;
 
-	if (dm->backlight_caps[bl_idx].caps_valid)
-		return;
+		if (caps->max_input_signal > AMDGPU_DM_DEFAULT_MAX_BACKLIGHT ||
+		    caps->min_input_signal < 0 ||
+		    spread > AMDGPU_DM_DEFAULT_MAX_BACKLIGHT ||
+		    spread < AMDGPU_DM_MIN_SPREAD) {
+			DRM_DEBUG_KMS("DM: Invalid backlight caps: min=%d, max=%d\n",
+				      caps->min_input_signal, caps->max_input_signal);
+			caps->caps_valid = false;
+		}
+	}
 
-	amdgpu_acpi_get_backlight_caps(&caps);
-	if (caps.caps_valid) {
-		dm->backlight_caps[bl_idx].caps_valid = true;
-		if (caps.aux_support)
-			return;
-		dm->backlight_caps[bl_idx].min_input_signal = caps.min_input_signal;
-		dm->backlight_caps[bl_idx].max_input_signal = caps.max_input_signal;
-	} else {
-		dm->backlight_caps[bl_idx].min_input_signal =
-				AMDGPU_DM_DEFAULT_MIN_BACKLIGHT;
-		dm->backlight_caps[bl_idx].max_input_signal =
-				AMDGPU_DM_DEFAULT_MAX_BACKLIGHT;
+	if (!caps->caps_valid) {
+		caps->min_input_signal = AMDGPU_DM_DEFAULT_MIN_BACKLIGHT;
+		caps->max_input_signal = AMDGPU_DM_DEFAULT_MAX_BACKLIGHT;
+		caps->caps_valid = true;
 	}
 #else
-	if (dm->backlight_caps[bl_idx].aux_support)
+	if (caps->aux_support)
 		return;
 
-	dm->backlight_caps[bl_idx].min_input_signal = AMDGPU_DM_DEFAULT_MIN_BACKLIGHT;
-	dm->backlight_caps[bl_idx].max_input_signal = AMDGPU_DM_DEFAULT_MAX_BACKLIGHT;
+	caps->min_input_signal = AMDGPU_DM_DEFAULT_MIN_BACKLIGHT;
+	caps->max_input_signal = AMDGPU_DM_DEFAULT_MAX_BACKLIGHT;
+	caps->caps_valid = true;
 #endif
 }
 
 static int get_brightness_range(const struct amdgpu_dm_backlight_caps *caps,
-				unsigned *min, unsigned *max)
+				unsigned int *min, unsigned int *max)
 {
 	if (!caps)
 		return 0;
@@ -4052,31 +5012,108 @@ static int get_brightness_range(const struct amdgpu_dm_backlight_caps *caps,
 	return 1;
 }
 
+/* Rescale from [min..max] to [0..AMDGPU_MAX_BL_LEVEL] */
+static inline u32 scale_input_to_fw(int min, int max, u64 input)
+{
+	return DIV_ROUND_CLOSEST_ULL(input * AMDGPU_MAX_BL_LEVEL, max - min);
+}
+
+/* Rescale from [0..AMDGPU_MAX_BL_LEVEL] to [min..max] */
+static inline u32 scale_fw_to_input(int min, int max, u64 input)
+{
+	return min + DIV_ROUND_CLOSEST_ULL(input * (max - min), AMDGPU_MAX_BL_LEVEL);
+}
+
+static void convert_custom_brightness(const struct amdgpu_dm_backlight_caps *caps,
+				      unsigned int min, unsigned int max,
+				      uint32_t *user_brightness)
+{
+	u32 brightness = scale_input_to_fw(min, max, *user_brightness);
+	u8 lower_signal, upper_signal, upper_lum, lower_lum, lum;
+	int left, right;
+
+	if (amdgpu_dc_debug_mask & DC_DISABLE_CUSTOM_BRIGHTNESS_CURVE)
+		return;
+
+	if (!caps->data_points)
+		return;
+
+	/*
+	 * Handle the case where brightness is below the first data point
+	 * Interpolate between (0,0) and (first_signal, first_lum)
+	 */
+	if (brightness < caps->luminance_data[0].input_signal) {
+		lum = DIV_ROUND_CLOSEST(caps->luminance_data[0].luminance * brightness,
+					caps->luminance_data[0].input_signal);
+		goto scale;
+	}
+
+	left = 0;
+	right = caps->data_points - 1;
+	while (left <= right) {
+		int mid = left + (right - left) / 2;
+		u8 signal = caps->luminance_data[mid].input_signal;
+
+		/* Exact match found */
+		if (signal == brightness) {
+			lum = caps->luminance_data[mid].luminance;
+			goto scale;
+		}
+
+		if (signal < brightness)
+			left = mid + 1;
+		else
+			right = mid - 1;
+	}
+
+	/* verify bound */
+	if (left >= caps->data_points)
+		left = caps->data_points - 1;
+
+	/* At this point, left > right */
+	lower_signal = caps->luminance_data[right].input_signal;
+	upper_signal = caps->luminance_data[left].input_signal;
+	lower_lum = caps->luminance_data[right].luminance;
+	upper_lum = caps->luminance_data[left].luminance;
+
+	/* interpolate */
+	if (right == left || !lower_lum)
+		lum = upper_lum;
+	else
+		lum = lower_lum + DIV_ROUND_CLOSEST((upper_lum - lower_lum) *
+						    (brightness - lower_signal),
+						    upper_signal - lower_signal);
+scale:
+	*user_brightness = scale_fw_to_input(min, max,
+					     DIV_ROUND_CLOSEST(lum * brightness, 101));
+}
+
 static u32 convert_brightness_from_user(const struct amdgpu_dm_backlight_caps *caps,
 					uint32_t brightness)
 {
-	unsigned min, max;
+	unsigned int min, max;
 
 	if (!get_brightness_range(caps, &min, &max))
 		return brightness;
 
-	// Rescale 0..255 to min..max
-	return min + DIV_ROUND_CLOSEST((max - min) * brightness,
-				       AMDGPU_MAX_BL_LEVEL);
+	convert_custom_brightness(caps, min, max, &brightness);
+
+	// Rescale 0..max to min..max
+	return min + DIV_ROUND_CLOSEST_ULL((u64)(max - min) * brightness, max);
 }
 
 static u32 convert_brightness_to_user(const struct amdgpu_dm_backlight_caps *caps,
 				      uint32_t brightness)
 {
-	unsigned min, max;
+	unsigned int min, max;
 
 	if (!get_brightness_range(caps, &min, &max))
 		return brightness;
 
 	if (brightness < min)
 		return 0;
-	// Rescale min..max to 0..255
-	return DIV_ROUND_CLOSEST(AMDGPU_MAX_BL_LEVEL * (brightness - min),
+	// Rescale min..max to 0..max
+	return DIV_ROUND_CLOSEST_ULL((u64)max * (brightness - min),
 				 max - min);
 }
 
@@ -4084,33 +5121,76 @@ static void amdgpu_dm_backlight_set_level(struct amdgpu_display_manager *dm,
 					 int bl_idx,
 					 u32 user_brightness)
 {
-	struct amdgpu_dm_backlight_caps caps;
+	struct amdgpu_dm_backlight_caps *caps;
 	struct dc_link *link;
 	u32 brightness;
-	bool rc;
+	bool rc, reallow_idle = false;
+	struct drm_connector *connector;
+
+	list_for_each_entry(connector, &dm->ddev->mode_config.connector_list, head) {
+		struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(connector);
+
+		if (aconnector->bl_idx != bl_idx)
+			continue;
+
+		/* if connector is off, save the brightness for next time it's on */
+		if (!aconnector->base.encoder) {
+			dm->brightness[bl_idx] = user_brightness;
+			dm->actual_brightness[bl_idx] = 0;
+			return;
+		}
+	}
 
 	amdgpu_dm_update_backlight_caps(dm, bl_idx);
-	caps = dm->backlight_caps[bl_idx];
+	caps = &dm->backlight_caps[bl_idx];
 
 	dm->brightness[bl_idx] = user_brightness;
 	/* update scratch register */
 	if (bl_idx == 0)
 		amdgpu_atombios_scratch_regs_set_backlight_level(dm->adev, dm->brightness[bl_idx]);
-	brightness = convert_brightness_from_user(&caps, dm->brightness[bl_idx]);
+	brightness = convert_brightness_from_user(caps, dm->brightness[bl_idx]);
 	link = (struct dc_link *)dm->backlight_link[bl_idx];
 
+	/* Apply brightness quirk */
+	if (caps->brightness_mask)
+		brightness |= caps->brightness_mask;
+
 	/* Change brightness based on AUX property */
-	if (caps.aux_support) {
+	mutex_lock(&dm->dc_lock);
+	if (dm->dc->caps.ips_support && dm->dc->ctx->dmub_srv->idle_allowed) {
+		dc_allow_idle_optimizations(dm->dc, false);
+		reallow_idle = true;
+	}
+
+	if (trace_amdgpu_dm_brightness_enabled()) {
+		trace_amdgpu_dm_brightness(__builtin_return_address(0),
+					   user_brightness,
+					   brightness,
+					   caps->aux_support,
+					   power_supply_is_system_supplied() > 0);
+	}
+
+	if (caps->aux_support) {
 		rc = dc_link_set_backlight_level_nits(link, true, brightness,
 						      AUX_BL_DEFAULT_TRANSITION_TIME_MS);
 		if (!rc)
 			DRM_DEBUG("DM: Failed to update backlight via AUX on eDP[%d]\n", bl_idx);
 	} else {
-		rc = dc_link_set_backlight_level(link, brightness, 0);
+		struct set_backlight_level_params backlight_level_params = { 0 };
+
+		backlight_level_params.backlight_pwm_u16_16 = brightness;
+		backlight_level_params.transition_time_in_ms = 0;
+
+		rc = dc_link_set_backlight_level(link, &backlight_level_params);
 		if (!rc)
 			DRM_DEBUG("DM: Failed to update backlight on eDP[%d]\n", bl_idx);
 	}
 
+	if (dm->dc->caps.ips_support && reallow_idle)
+		dc_allow_idle_optimizations(dm->dc, true);
+
+	mutex_unlock(&dm->dc_lock);
+
 	if (rc)
 		dm->actual_brightness[bl_idx] = user_brightness;
 }
@@ -4134,6 +5214,7 @@ static int amdgpu_dm_backlight_update_status(struct backlight_device *bd)
 static u32 amdgpu_dm_backlight_get_level(struct amdgpu_display_manager *dm,
 					 int bl_idx)
 {
+	int ret;
 	struct amdgpu_dm_backlight_caps caps;
 	struct dc_link *link = (struct dc_link *)dm->backlight_link[bl_idx];
 
@@ -4142,19 +5223,18 @@ static u32 amdgpu_dm_backlight_get_level(struct amdgpu_display_manager *dm,
 
 	if (caps.aux_support) {
 		u32 avg, peak;
-		bool rc;
 
-		rc = dc_link_get_backlight_level_nits(link, &avg, &peak);
-		if (!rc)
+		if (!dc_link_get_backlight_level_nits(link, &avg, &peak))
 			return dm->brightness[bl_idx];
 		return convert_brightness_to_user(&caps, avg);
-	} else {
-		int ret = dc_link_get_backlight_level(link);
-
-		if (ret == DC_ERROR_UNEXPECTED)
-			return dm->brightness[bl_idx];
-		return convert_brightness_to_user(&caps, ret);
 	}
+
+	ret = dc_link_get_backlight_level(link);
+
+	if (ret == DC_ERROR_UNEXPECTED)
+		return dm->brightness[bl_idx];
+
+	return convert_brightness_to_user(&caps, ret);
 }
 
 static int amdgpu_dm_backlight_get_brightness(struct backlight_device *bd)
@@ -4183,7 +5263,9 @@ amdgpu_dm_register_backlight_device(struct amdgpu_dm_connector *aconnector)
 	struct drm_device *drm = aconnector->base.dev;
 	struct amdgpu_display_manager *dm = &drm_to_adev(drm)->dm;
 	struct backlight_properties props = { 0 };
+	struct amdgpu_dm_backlight_caps *caps;
 	char bl_name[16];
+	int min, max;
 
 	if (aconnector->bl_idx == -1)
 		return;
@@ -4195,8 +5277,24 @@ amdgpu_dm_register_backlight_device(struct amdgpu_dm_connector *aconnector)
 		return;
 	}
 
-	props.max_brightness = AMDGPU_MAX_BL_LEVEL;
-	props.brightness = AMDGPU_MAX_BL_LEVEL;
+	caps = &dm->backlight_caps[aconnector->bl_idx];
+	if (get_brightness_range(caps, &min, &max)) {
+		if (power_supply_is_system_supplied() > 0)
+			props.brightness = DIV_ROUND_CLOSEST((max - min) * caps->ac_level, 100);
+		else
+			props.brightness = DIV_ROUND_CLOSEST((max - min) * caps->dc_level, 100);
+		/* min is zero, so max needs to be adjusted */
+		props.max_brightness = max - min;
+		drm_dbg(drm, "Backlight caps: min: %d, max: %d, ac %d, dc %d\n", min, max,
+			caps->ac_level, caps->dc_level);
+	} else
+		props.brightness = props.max_brightness = MAX_BACKLIGHT_LEVEL;
+
+	if (caps->data_points && !(amdgpu_dc_debug_mask & DC_DISABLE_CUSTOM_BRIGHTNESS_CURVE)) {
+		drm_info(drm, "Using custom brightness curve\n");
+		props.scale = BACKLIGHT_SCALE_NON_LINEAR;
+	} else
+		props.scale = BACKLIGHT_SCALE_LINEAR;
 	props.type = BACKLIGHT_RAW;
 
 	snprintf(bl_name, sizeof(bl_name), "amdgpu_bl%d",
@@ -4205,12 +5303,13 @@ amdgpu_dm_register_backlight_device(struct amdgpu_dm_connector *aconnector)
 	dm->backlight_dev[aconnector->bl_idx] =
 		backlight_device_register(bl_name, aconnector->base.kdev, dm,
 					  &amdgpu_dm_backlight_ops, &props);
+	dm->brightness[aconnector->bl_idx] = props.brightness;
 
 	if (IS_ERR(dm->backlight_dev[aconnector->bl_idx])) {
-		DRM_ERROR("DM: Backlight registration failed!\n");
+		drm_err(drm, "DM: Backlight registration failed!\n");
 		dm->backlight_dev[aconnector->bl_idx] = NULL;
 	} else
-		DRM_DEBUG_DRIVER("DM: Registered Backlight device: %s\n", bl_name);
+		drm_dbg_driver(drm, "DM: Registered Backlight device: %s\n", bl_name);
 }
 
 static int initialize_plane(struct amdgpu_display_manager *dm,
@@ -4224,7 +5323,7 @@ static int initialize_plane(struct amdgpu_display_manager *dm,
 
 	plane = kzalloc(sizeof(struct drm_plane), GFP_KERNEL);
 	if (!plane) {
-		DRM_ERROR("KMS: Failed to allocate plane\n");
+		drm_err(adev_to_drm(dm->adev), "KMS: Failed to allocate plane\n");
 		return -ENOMEM;
 	}
 	plane->type = plane_type;
@@ -4242,7 +5341,7 @@ static int initialize_plane(struct amdgpu_display_manager *dm,
 	ret = amdgpu_dm_plane_init(dm, plane, possible_crtcs, plane_cap);
 
 	if (ret) {
-		DRM_ERROR("KMS: Failed to initialize plane\n");
+		drm_err(adev_to_drm(dm->adev), "KMS: Failed to initialize plane\n");
 		kfree(plane);
 		return ret;
 	}
@@ -4257,6 +5356,7 @@ static int initialize_plane(struct amdgpu_display_manager *dm,
 static void setup_backlight_device(struct amdgpu_display_manager *dm,
 				   struct amdgpu_dm_connector *aconnector)
 {
+	struct amdgpu_dm_backlight_caps *caps;
 	struct dc_link *link = aconnector->dc_link;
 	int bl_idx = dm->num_of_edps;
 
@@ -4272,11 +5372,17 @@ static void setup_backlight_device(struct amdgpu_display_manager *dm,
 	aconnector->bl_idx = bl_idx;
 
 	amdgpu_dm_update_backlight_caps(dm, bl_idx);
-	dm->brightness[bl_idx] = AMDGPU_MAX_BL_LEVEL;
 	dm->backlight_link[bl_idx] = link;
 	dm->num_of_edps++;
 
 	update_connector_ext_caps(aconnector);
+	caps = &dm->backlight_caps[aconnector->bl_idx];
+
+	/* Only offer ABM property when non-OLED and user didn't turn off by module parameter */
+	if (!caps->ext_caps->bits.oled && amdgpu_dm_abm_level < 0)
+		drm_object_attach_property(&aconnector->base.base,
+					   dm->adev->mode_info.abm_level_property,
+					   ABM_SYSFS_CONTROL);
 }
 
 static void amdgpu_set_panel_orientation(struct drm_connector *connector);
@@ -4301,6 +5407,7 @@ static int amdgpu_dm_initialize_drm_device(struct amdgpu_device *adev)
 	enum dc_connection_type new_connection_type = dc_connection_none;
 	const struct dc_plane_cap *plane;
 	bool psr_feature_enabled = false;
+	bool replay_feature_enabled = false;
 	int max_overlay = dm->dc->caps.max_slave_planes;
 
 	dm->display_indexes_num = dm->dc->caps.max_streams;
@@ -4311,13 +5418,16 @@ static int amdgpu_dm_initialize_drm_device(struct amdgpu_device *adev)
 
 	link_cnt = dm->dc->caps.max_links;
 	if (amdgpu_dm_mode_config_init(dm->adev)) {
-		DRM_ERROR("DM: Failed to initialize mode config\n");
+		drm_err(adev_to_drm(adev), "DM: Failed to initialize mode config\n");
 		return -EINVAL;
 	}
 
 	/* There is one primary plane per CRTC */
 	primary_planes = dm->dc->caps.max_streams;
-	ASSERT(primary_planes <= AMDGPU_MAX_PLANES);
+	if (primary_planes > AMDGPU_MAX_PLANES) {
+		drm_err(adev_to_drm(adev), "DM: Plane nums out of 6 planes\n");
+		return -EINVAL;
+	}
 
 	/*
 	 * Initialize primary planes, implicit planes for legacy IOCTLS.
@@ -4328,7 +5438,7 @@ static int amdgpu_dm_initialize_drm_device(struct amdgpu_device *adev)
 
 		if (initialize_plane(dm, mode_info, i,
 				     DRM_PLANE_TYPE_PRIMARY, plane)) {
-			DRM_ERROR("KMS: Failed to initialize primary plane\n");
+			drm_err(adev_to_drm(adev), "KMS: Failed to initialize primary plane\n");
 			goto fail;
 		}
 	}
@@ -4360,19 +5470,19 @@ static int amdgpu_dm_initialize_drm_device(struct amdgpu_device *adev)
 
 		if (initialize_plane(dm, NULL, primary_planes + i,
 				     DRM_PLANE_TYPE_OVERLAY, plane)) {
-			DRM_ERROR("KMS: Failed to initialize overlay plane\n");
+			drm_err(adev_to_drm(adev), "KMS: Failed to initialize overlay plane\n");
 			goto fail;
 		}
 	}
 
 	for (i = 0; i < dm->dc->caps.max_streams; i++)
 		if (amdgpu_dm_crtc_init(dm, mode_info->planes[i], i)) {
-			DRM_ERROR("KMS: Failed to initialize crtc\n");
+			drm_err(adev_to_drm(adev), "KMS: Failed to initialize crtc\n");
 			goto fail;
 		}
 
 	/* Use Outbox interrupt */
-	switch (adev->ip_versions[DCE_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
 	case IP_VERSION(3, 0, 0):
 	case IP_VERSION(3, 1, 2):
 	case IP_VERSION(3, 1, 3):
@@ -4382,19 +5492,23 @@ static int amdgpu_dm_initialize_drm_device(struct amdgpu_device *adev)
 	case IP_VERSION(3, 2, 0):
 	case IP_VERSION(3, 2, 1):
 	case IP_VERSION(2, 1, 0):
+	case IP_VERSION(3, 5, 0):
+	case IP_VERSION(3, 5, 1):
+	case IP_VERSION(3, 6, 0):
+	case IP_VERSION(4, 0, 1):
 		if (register_outbox_irq_handlers(dm->adev)) {
-			DRM_ERROR("DM: Failed to initialize IRQ\n");
+			drm_err(adev_to_drm(adev), "DM: Failed to initialize IRQ\n");
 			goto fail;
 		}
 		break;
 	default:
 		DRM_DEBUG_KMS("Unsupported DCN IP version for outbox: 0x%X\n",
-			      adev->ip_versions[DCE_HWIP][0]);
+			      amdgpu_ip_version(adev, DCE_HWIP, 0));
 	}
 
 	/* Determine whether to enable PSR support by default. */
 	if (!(amdgpu_dc_debug_mask & DC_DISABLE_PSR)) {
-		switch (adev->ip_versions[DCE_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
 		case IP_VERSION(3, 1, 2):
 		case IP_VERSION(3, 1, 3):
 		case IP_VERSION(3, 1, 4):
@@ -4402,6 +5516,10 @@ static int amdgpu_dm_initialize_drm_device(struct amdgpu_device *adev)
 		case IP_VERSION(3, 1, 6):
 		case IP_VERSION(3, 2, 0):
 		case IP_VERSION(3, 2, 1):
+		case IP_VERSION(3, 5, 0):
+		case IP_VERSION(3, 5, 1):
+		case IP_VERSION(3, 6, 0):
+		case IP_VERSION(4, 0, 1):
 			psr_feature_enabled = true;
 			break;
 		default:
@@ -4410,14 +5528,54 @@ static int amdgpu_dm_initialize_drm_device(struct amdgpu_device *adev)
 		}
 	}
 
+	/* Determine whether to enable Replay support by default. */
+	if (!(amdgpu_dc_debug_mask & DC_DISABLE_REPLAY)) {
+		switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
+		case IP_VERSION(3, 1, 4):
+		case IP_VERSION(3, 2, 0):
+		case IP_VERSION(3, 2, 1):
+		case IP_VERSION(3, 5, 0):
+		case IP_VERSION(3, 5, 1):
+		case IP_VERSION(3, 6, 0):
+			replay_feature_enabled = true;
+			break;
+
+		default:
+			replay_feature_enabled = amdgpu_dc_feature_mask & DC_REPLAY_MASK;
+			break;
+		}
+	}
+
+	if (link_cnt > MAX_LINKS) {
+		drm_err(adev_to_drm(adev),
+			"KMS: Cannot support more than %d display indexes\n",
+				MAX_LINKS);
+		goto fail;
+	}
+
 	/* loops over all connectors on the board */
 	for (i = 0; i < link_cnt; i++) {
 		struct dc_link *link = NULL;
 
-		if (i > AMDGPU_DM_MAX_DISPLAY_INDEX) {
-			DRM_ERROR(
-				"KMS: Cannot support more than %d display indexes\n",
-					AMDGPU_DM_MAX_DISPLAY_INDEX);
+		link = dc_get_link_at_index(dm->dc, i);
+
+		if (link->connector_signal == SIGNAL_TYPE_VIRTUAL) {
+			struct amdgpu_dm_wb_connector *wbcon = kzalloc(sizeof(*wbcon), GFP_KERNEL);
+
+			if (!wbcon) {
+				drm_err(adev_to_drm(adev), "KMS: Failed to allocate writeback connector\n");
+				continue;
+			}
+
+			if (amdgpu_dm_wb_connector_init(dm, wbcon, i)) {
+				drm_err(adev_to_drm(adev), "KMS: Failed to initialize writeback connector\n");
+				kfree(wbcon);
+				continue;
+			}
+
+			link->psr_settings.psr_feature_enabled = false;
+			link->psr_settings.psr_version = DC_PSR_VERSION_UNSUPPORTED;
+
 			continue;
 		}
 
@@ -4430,19 +5588,21 @@ static int amdgpu_dm_initialize_drm_device(struct amdgpu_device *adev)
 			goto fail;
 
 		if (amdgpu_dm_encoder_init(dm->ddev, aencoder, i)) {
-			DRM_ERROR("KMS: Failed to initialize encoder\n");
+			drm_err(adev_to_drm(adev), "KMS: Failed to initialize encoder\n");
 			goto fail;
 		}
 
 		if (amdgpu_dm_connector_init(dm, aconnector, i, aencoder)) {
-			DRM_ERROR("KMS: Failed to initialize connector\n");
+			drm_err(adev_to_drm(adev), "KMS: Failed to initialize connector\n");
 			goto fail;
 		}
 
-		link = dc_get_link_at_index(dm->dc, i);
+		if (dm->hpd_rx_offload_wq)
+			dm->hpd_rx_offload_wq[aconnector->base.index].aconnector =
+				aconnector;
 
 		if (!dc_link_detect_connection_type(link, &new_connection_type))
-			DRM_ERROR("KMS: Failed to detect connector\n");
+			drm_err(adev_to_drm(adev), "KMS: Failed to detect connector\n");
 
 		if (aconnector->base.force && new_connection_type == dc_connection_none) {
 			emulated_link_detect(link);
@@ -4451,6 +5611,7 @@ static int amdgpu_dm_initialize_drm_device(struct amdgpu_device *adev)
 			bool ret = false;
 
 			mutex_lock(&dm->dc_lock);
+			dc_exit_ips_for_hw_access(dm->dc);
 			ret = dc_link_detect(link, DETECT_REASON_BOOT);
 			mutex_unlock(&dm->dc_lock);
 
@@ -4458,19 +5619,31 @@ static int amdgpu_dm_initialize_drm_device(struct amdgpu_device *adev)
 				amdgpu_dm_update_connector_after_detect(aconnector);
 				setup_backlight_device(dm, aconnector);
 
-				if (psr_feature_enabled)
-					amdgpu_dm_set_psr_caps(link);
+				/* Disable PSR if Replay can be enabled */
+				if (replay_feature_enabled)
+					if (amdgpu_dm_set_replay_caps(link, aconnector))
+						psr_feature_enabled = false;
 
-				/* TODO: Fix vblank control helpers to delay PSR entry to allow this when
-				 * PSR is also supported.
-				 */
-				if (link->psr_settings.psr_feature_enabled)
-					adev_to_drm(adev)->vblank_disable_immediate = false;
+				if (psr_feature_enabled) {
+					amdgpu_dm_set_psr_caps(link);
+					drm_info(adev_to_drm(adev), "PSR support %d, DC PSR ver %d, sink PSR ver %d DPCD caps 0x%x su_y_granularity %d\n",
+						 link->psr_settings.psr_feature_enabled,
+						 link->psr_settings.psr_version,
+						 link->dpcd_caps.psr_info.psr_version,
+						 link->dpcd_caps.psr_info.psr_dpcd_caps.raw,
+						 link->dpcd_caps.psr_info.psr2_su_y_granularity_cap);
+				}
 			}
 		}
 		amdgpu_set_panel_orientation(&aconnector->base);
 	}
 
+	/* Debug dump: list all DC links and their associated sinks after detection
+	 * is complete for all connectors. This provides a comprehensive view of the
+	 * final state without repeating the dump for each connector.
+	 */
+	amdgpu_dm_dump_links_and_sinks(adev);
+
 	/* Software is initialized. Now we can register interrupt handlers. */
 	switch (adev->asic_type) {
 #if defined(CONFIG_DRM_AMD_DC_SI)
@@ -4479,7 +5652,7 @@ static int amdgpu_dm_initialize_drm_device(struct amdgpu_device *adev)
 	case CHIP_VERDE:
 	case CHIP_OLAND:
 		if (dce60_register_irq_handlers(dm->adev)) {
-			DRM_ERROR("DM: Failed to initialize IRQ\n");
+			drm_err(adev_to_drm(adev), "DM: Failed to initialize IRQ\n");
 			goto fail;
 		}
 		break;
@@ -4501,12 +5674,12 @@ static int amdgpu_dm_initialize_drm_device(struct amdgpu_device *adev)
 	case CHIP_VEGA12:
 	case CHIP_VEGA20:
 		if (dce110_register_irq_handlers(dm->adev)) {
-			DRM_ERROR("DM: Failed to initialize IRQ\n");
+			drm_err(adev_to_drm(adev), "DM: Failed to initialize IRQ\n");
 			goto fail;
 		}
 		break;
 	default:
-		switch (adev->ip_versions[DCE_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
 		case IP_VERSION(1, 0, 0):
 		case IP_VERSION(1, 0, 1):
 		case IP_VERSION(2, 0, 2):
@@ -4524,14 +5697,18 @@ static int amdgpu_dm_initialize_drm_device(struct amdgpu_device *adev)
 		case IP_VERSION(3, 1, 6):
 		case IP_VERSION(3, 2, 0):
 		case IP_VERSION(3, 2, 1):
+		case IP_VERSION(3, 5, 0):
+		case IP_VERSION(3, 5, 1):
+		case IP_VERSION(3, 6, 0):
+		case IP_VERSION(4, 0, 1):
 			if (dcn10_register_irq_handlers(dm->adev)) {
-				DRM_ERROR("DM: Failed to initialize IRQ\n");
+				drm_err(adev_to_drm(adev), "DM: Failed to initialize IRQ\n");
 				goto fail;
 			}
 			break;
 		default:
-			DRM_ERROR("Unsupported DCE IP versions: 0x%X\n",
-					adev->ip_versions[DCE_HWIP][0]);
+			drm_err(adev_to_drm(adev), "Unsupported DCE IP versions: 0x%X\n",
+					amdgpu_ip_version(adev, DCE_HWIP, 0));
 			goto fail;
 		}
 		break;
@@ -4547,8 +5724,8 @@ fail:
 
 static void amdgpu_dm_destroy_drm_device(struct amdgpu_display_manager *dm)
 {
-	drm_atomic_private_obj_fini(&dm->atomic_obj);
-	return;
+	if (dm->atomic_obj.state)
+		drm_atomic_private_obj_fini(&dm->atomic_obj);
 }
 
 /******************************************************************************
@@ -4592,15 +5769,20 @@ static ssize_t s3_debug_store(struct device *device,
 	int s3_state;
 	struct drm_device *drm_dev = dev_get_drvdata(device);
 	struct amdgpu_device *adev = drm_to_adev(drm_dev);
+	struct amdgpu_ip_block *ip_block;
+
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_DCE);
+	if (!ip_block)
+		return -EINVAL;
 
 	ret = kstrtoint(buf, 0, &s3_state);
 
 	if (ret == 0) {
 		if (s3_state) {
-			dm_resume(adev);
+			dm_resume(ip_block);
 			drm_kms_helper_hotplug_event(adev_to_drm(adev));
 		} else
-			dm_suspend(adev);
+			dm_suspend(ip_block);
 	}
 
 	return ret == 0 ? count : 0;
@@ -4615,14 +5797,14 @@ static int dm_init_microcode(struct amdgpu_device *adev)
 	char *fw_name_dmub;
 	int r;
 
-	switch (adev->ip_versions[DCE_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
 	case IP_VERSION(2, 1, 0):
 		fw_name_dmub = FIRMWARE_RENOIR_DMUB;
 		if (ASICREV_IS_GREEN_SARDINE(adev->external_rev_id))
 			fw_name_dmub = FIRMWARE_GREEN_SARDINE_DMUB;
 		break;
 	case IP_VERSION(3, 0, 0):
-		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 0))
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 3, 0))
 			fw_name_dmub = FIRMWARE_SIENNA_CICHLID_DMUB;
 		else
 			fw_name_dmub = FIRMWARE_NAVY_FLOUNDER_DMUB;
@@ -4655,19 +5837,30 @@ static int dm_init_microcode(struct amdgpu_device *adev)
 	case IP_VERSION(3, 2, 1):
 		fw_name_dmub = FIRMWARE_DCN_V3_2_1_DMCUB;
 		break;
+	case IP_VERSION(3, 5, 0):
+		fw_name_dmub = FIRMWARE_DCN_35_DMUB;
+		break;
+	case IP_VERSION(3, 5, 1):
+		fw_name_dmub = FIRMWARE_DCN_351_DMUB;
+		break;
+	case IP_VERSION(3, 6, 0):
+		fw_name_dmub = FIRMWARE_DCN_36_DMUB;
+		break;
+	case IP_VERSION(4, 0, 1):
+		fw_name_dmub = FIRMWARE_DCN_401_DMUB;
+		break;
 	default:
 		/* ASIC doesn't support DMUB. */
 		return 0;
 	}
-	r = amdgpu_ucode_request(adev, &adev->dm.dmub_fw, fw_name_dmub);
-	if (r)
-		DRM_ERROR("DMUB firmware loading failed: %d\n", r);
+	r = amdgpu_ucode_request(adev, &adev->dm.dmub_fw, AMDGPU_UCODE_REQUIRED,
+				 "%s", fw_name_dmub);
 	return r;
 }
 
-static int dm_early_init(void *handle)
+static int dm_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_mode_info *mode_info = &adev->mode_info;
 	struct atom_context *ctx = mode_info->atom_context;
 	int index = GetIndexIntoMasterTable(DATA, Object_Header);
@@ -4676,7 +5869,7 @@ static int dm_early_init(void *handle)
 	/* if there is no object header, skip DM */
 	if (!amdgpu_atom_parse_data_header(ctx, index, NULL, NULL, NULL, &data_offset)) {
 		adev->harvest_ip_mask |= AMD_HARVEST_IP_DMU_MASK;
-		dev_info(adev->dev, "No object header, skipping DM\n");
+		drm_info(adev_to_drm(adev), "No object header, skipping DM\n");
 		return -ENOENT;
 	}
 
@@ -4749,7 +5942,7 @@ static int dm_early_init(void *handle)
 		break;
 	default:
 
-		switch (adev->ip_versions[DCE_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
 		case IP_VERSION(2, 0, 2):
 		case IP_VERSION(3, 0, 0):
 			adev->mode_info.num_crtc = 6;
@@ -4779,13 +5972,17 @@ static int dm_early_init(void *handle)
 		case IP_VERSION(3, 1, 6):
 		case IP_VERSION(3, 2, 0):
 		case IP_VERSION(3, 2, 1):
+		case IP_VERSION(3, 5, 0):
+		case IP_VERSION(3, 5, 1):
+		case IP_VERSION(3, 6, 0):
+		case IP_VERSION(4, 0, 1):
 			adev->mode_info.num_crtc = 4;
 			adev->mode_info.num_hpd = 4;
 			adev->mode_info.num_dig = 4;
 			break;
 		default:
-			DRM_ERROR("Unsupported DCE IP versions: 0x%x\n",
-					adev->ip_versions[DCE_HWIP][0]);
+			drm_err(adev_to_drm(adev), "Unsupported DCE IP versions: 0x%x\n",
+					amdgpu_ip_version(adev, DCE_HWIP, 0));
 			return -EINVAL;
 		}
 		break;
@@ -4833,6 +6030,10 @@ fill_plane_color_attributes(const struct drm_plane_state *plane_state,
 
 	*color_space = COLOR_SPACE_SRGB;
 
+	/* Ignore properties when DRM_CLIENT_CAP_PLANE_COLOR_PIPELINE is set */
+	if (plane_state->state && plane_state->state->plane_color_pipeline)
+		return 0;
+
 	/* DRM color properties only affect non-RGB formats. */
 	if (format < SURFACE_PIXEL_FORMAT_VIDEO_BEGIN)
 		return 0;
@@ -4856,9 +6057,9 @@ fill_plane_color_attributes(const struct drm_plane_state *plane_state,
 
 	case DRM_COLOR_YCBCR_BT2020:
 		if (full_range)
-			*color_space = COLOR_SPACE_2020_YCBCR;
+			*color_space = COLOR_SPACE_2020_YCBCR_FULL;
 		else
-			return -EINVAL;
+			*color_space = COLOR_SPACE_2020_YCBCR_LIMITED;
 		break;
 
 	default:
@@ -4874,8 +6075,7 @@ fill_dc_plane_info_and_addr(struct amdgpu_device *adev,
 			    const u64 tiling_flags,
 			    struct dc_plane_info *plane_info,
 			    struct dc_plane_address *address,
-			    bool tmz_surface,
-			    bool force_disable_dcc)
+			    bool tmz_surface)
 {
 	const struct drm_framebuffer *fb = plane_state->fb;
 	const struct amdgpu_framebuffer *afb =
@@ -4934,7 +6134,7 @@ fill_dc_plane_info_and_addr(struct amdgpu_device *adev,
 		plane_info->format = SURFACE_PIXEL_FORMAT_GRPH_ABGR16161616;
 		break;
 	default:
-		DRM_ERROR(
+		drm_err(adev_to_drm(adev),
 			"Unsupported screen format %p4cc\n",
 			&fb->format->format);
 		return -EINVAL;
@@ -4974,7 +6174,7 @@ fill_dc_plane_info_and_addr(struct amdgpu_device *adev,
 					   &plane_info->tiling_info,
 					   &plane_info->plane_size,
 					   &plane_info->dcc, address,
-					   tmz_surface, force_disable_dcc);
+					   tmz_surface);
 	if (ret)
 		return ret;
 
@@ -4995,7 +6195,6 @@ static int fill_dc_plane_attributes(struct amdgpu_device *adev,
 	struct dc_scaling_info scaling_info;
 	struct dc_plane_info plane_info;
 	int ret;
-	bool force_disable_dcc = false;
 
 	ret = amdgpu_dm_plane_fill_dc_scaling_info(adev, plane_state, &scaling_info);
 	if (ret)
@@ -5006,13 +6205,11 @@ static int fill_dc_plane_attributes(struct amdgpu_device *adev,
 	dc_plane_state->clip_rect = scaling_info.clip_rect;
 	dc_plane_state->scaling_quality = scaling_info.scaling_quality;
 
-	force_disable_dcc = adev->asic_type == CHIP_RAVEN && adev->in_suspend;
 	ret = fill_dc_plane_info_and_addr(adev, plane_state,
 					  afb->tiling_flags,
 					  &plane_info,
 					  &dc_plane_state->address,
-					  afb->tmz_surface,
-					  force_disable_dcc);
+					  afb->tmz_surface);
 	if (ret)
 		return ret;
 
@@ -5037,7 +6234,9 @@ static int fill_dc_plane_attributes(struct amdgpu_device *adev,
 	 * Always set input transfer function, since plane state is refreshed
 	 * every time.
 	 */
-	ret = amdgpu_dm_update_plane_color_mgmt(dm_crtc_state, dc_plane_state);
+	ret = amdgpu_dm_update_plane_color_mgmt(dm_crtc_state,
+						plane_state,
+						dc_plane_state);
 	if (ret)
 		return ret;
 
@@ -5049,11 +6248,7 @@ static inline void fill_dc_dirty_rect(struct drm_plane *plane,
 				      s32 y, s32 width, s32 height,
 				      int *i, bool ffu)
 {
-	if (*i > DC_MAX_DIRTY_RECTS)
-		return;
-
-	if (*i == DC_MAX_DIRTY_RECTS)
-		goto out;
+	WARN_ON(*i >= DC_MAX_DIRTY_RECTS);
 
 	dirty_rect->x = x;
 	dirty_rect->y = y;
@@ -5069,7 +6264,6 @@ static inline void fill_dc_dirty_rect(struct drm_plane *plane,
 			"[PLANE:%d] PSR SU dirty rect at (%d, %d) size (%d, %d)",
 			plane->base.id, x, y, width, height);
 
-out:
 	(*i)++;
 }
 
@@ -5082,6 +6276,10 @@ out:
  * @new_plane_state: New state of @plane
  * @crtc_state: New state of CRTC connected to the @plane
  * @flip_addrs: DC flip tracking struct, which also tracts dirty rects
+ * @is_psr_su: Flag indicating whether Panel Self Refresh Selective Update (PSR SU) is enabled.
+ *             If PSR SU is enabled and damage clips are available, only the regions of the screen
+ *             that have changed will be updated. If PSR SU is not enabled,
+ *             or if damage clips are not available, the entire screen will be updated.
  * @dirty_regions_changed: dirty regions changed
  *
  * For PSR SU, DC informs the DMUB uController of dirty rectangle regions
@@ -5100,6 +6298,7 @@ static void fill_dc_dirty_rects(struct drm_plane *plane,
 				struct drm_plane_state *new_plane_state,
 				struct drm_crtc_state *crtc_state,
 				struct dc_flip_addrs *flip_addrs,
+				bool is_psr_su,
 				bool *dirty_regions_changed)
 {
 	struct dm_crtc_state *dm_crtc_state = to_dm_crtc_state(crtc_state);
@@ -5118,9 +6317,16 @@ static void fill_dc_dirty_rects(struct drm_plane *plane,
 	if (plane->type == DRM_PLANE_TYPE_CURSOR)
 		return;
 
+	if (new_plane_state->rotation != DRM_MODE_ROTATE_0)
+		goto ffu;
+
 	num_clips = drm_plane_get_damage_clips_count(new_plane_state);
 	clips = drm_plane_get_damage_clips(new_plane_state);
 
+	if (num_clips && (!amdgpu_damage_clips || (amdgpu_damage_clips < 0 &&
+						   is_psr_su)))
+		goto ffu;
+
 	if (!dm_crtc_state->mpo_requested) {
 		if (!num_clips || num_clips > DC_MAX_DIRTY_RECTS)
 			goto ffu;
@@ -5156,6 +6362,9 @@ static void fill_dc_dirty_rects(struct drm_plane *plane,
 
 	*dirty_regions_changed = bb_changed;
 
+	if ((num_clips + (bb_changed ? 2 : 0)) > DC_MAX_DIRTY_RECTS)
+		goto ffu;
+
 	if (bb_changed) {
 		fill_dc_dirty_rect(new_plane_state->plane, &dirty_rects[i],
 				   new_plane_state->crtc_x,
@@ -5185,9 +6394,6 @@ static void fill_dc_dirty_rects(struct drm_plane *plane,
 				   new_plane_state->crtc_h, &i, false);
 	}
 
-	if (i > DC_MAX_DIRTY_RECTS)
-		goto ffu;
-
 	flip_addrs->dirty_rect_count = i;
 	return;
 
@@ -5323,21 +6529,46 @@ get_aspect_ratio(const struct drm_display_mode *mode_in)
 }
 
 static enum dc_color_space
-get_output_color_space(const struct dc_crtc_timing *dc_crtc_timing)
+get_output_color_space(const struct dc_crtc_timing *dc_crtc_timing,
+		       const struct drm_connector_state *connector_state)
 {
 	enum dc_color_space color_space = COLOR_SPACE_SRGB;
 
-	switch (dc_crtc_timing->pixel_encoding)	{
-	case PIXEL_ENCODING_YCBCR422:
-	case PIXEL_ENCODING_YCBCR444:
-	case PIXEL_ENCODING_YCBCR420:
-	{
+	switch (connector_state->colorspace) {
+	case DRM_MODE_COLORIMETRY_BT601_YCC:
+		if (dc_crtc_timing->flags.Y_ONLY)
+			color_space = COLOR_SPACE_YCBCR601_LIMITED;
+		else
+			color_space = COLOR_SPACE_YCBCR601;
+		break;
+	case DRM_MODE_COLORIMETRY_BT709_YCC:
+		if (dc_crtc_timing->flags.Y_ONLY)
+			color_space = COLOR_SPACE_YCBCR709_LIMITED;
+		else
+			color_space = COLOR_SPACE_YCBCR709;
+		break;
+	case DRM_MODE_COLORIMETRY_OPRGB:
+		color_space = COLOR_SPACE_ADOBERGB;
+		break;
+	case DRM_MODE_COLORIMETRY_BT2020_RGB:
+	case DRM_MODE_COLORIMETRY_BT2020_YCC:
+		if (dc_crtc_timing->pixel_encoding == PIXEL_ENCODING_RGB)
+			color_space = COLOR_SPACE_2020_RGB_FULLRANGE;
+		else
+			color_space = COLOR_SPACE_2020_YCBCR_LIMITED;
+		break;
+	case DRM_MODE_COLORIMETRY_DEFAULT: // ITU601
+	default:
+		if (dc_crtc_timing->pixel_encoding == PIXEL_ENCODING_RGB) {
+			color_space = COLOR_SPACE_SRGB;
+			if (connector_state->hdmi.broadcast_rgb == DRM_HDMI_BROADCAST_RGB_LIMITED)
+				color_space = COLOR_SPACE_SRGB_LIMITED;
 		/*
 		 * 27030khz is the separation point between HDTV and SDTV
 		 * according to HDMI spec, we use YCbCr709 and YCbCr601
 		 * respectively
 		 */
-		if (dc_crtc_timing->pix_clk_100hz > 270300) {
+		} else if (dc_crtc_timing->pix_clk_100hz > 270300) {
 			if (dc_crtc_timing->flags.Y_ONLY)
 				color_space =
 					COLOR_SPACE_YCBCR709_LIMITED;
@@ -5350,27 +6581,37 @@ get_output_color_space(const struct dc_crtc_timing *dc_crtc_timing)
 			else
 				color_space = COLOR_SPACE_YCBCR601;
 		}
-
-	}
-	break;
-	case PIXEL_ENCODING_RGB:
-		color_space = COLOR_SPACE_SRGB;
-		break;
-
-	default:
-		WARN_ON(1);
 		break;
 	}
 
 	return color_space;
 }
 
+static enum display_content_type
+get_output_content_type(const struct drm_connector_state *connector_state)
+{
+	switch (connector_state->content_type) {
+	default:
+	case DRM_MODE_CONTENT_TYPE_NO_DATA:
+		return DISPLAY_CONTENT_TYPE_NO_DATA;
+	case DRM_MODE_CONTENT_TYPE_GRAPHICS:
+		return DISPLAY_CONTENT_TYPE_GRAPHICS;
+	case DRM_MODE_CONTENT_TYPE_PHOTO:
+		return DISPLAY_CONTENT_TYPE_PHOTO;
+	case DRM_MODE_CONTENT_TYPE_CINEMA:
+		return DISPLAY_CONTENT_TYPE_CINEMA;
+	case DRM_MODE_CONTENT_TYPE_GAME:
+		return DISPLAY_CONTENT_TYPE_GAME;
+	}
+}
+
 static bool adjust_colour_depth_from_display_info(
 	struct dc_crtc_timing *timing_out,
 	const struct drm_display_info *info)
 {
 	enum dc_color_depth depth = timing_out->display_color_depth;
 	int normalized_clk;
+
 	do {
 		normalized_clk = timing_out->pix_clk_100hz / 10;
 		/* YCbCr 4:2:0 requires additional adjustment of 1/2 */
@@ -5411,9 +6652,13 @@ static void fill_stream_properties_from_drm_display_mode(
 {
 	struct dc_crtc_timing *timing_out = &stream->timing;
 	const struct drm_display_info *info = &connector->display_info;
-	struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(connector);
+	struct amdgpu_dm_connector *aconnector = NULL;
 	struct hdmi_vendor_infoframe hv_frame;
 	struct hdmi_avi_infoframe avi_frame;
+	ssize_t err;
+
+	if (connector->connector_type != DRM_MODE_CONNECTOR_WRITEBACK)
+		aconnector = to_amdgpu_dm_connector(connector);
 
 	memset(&hv_frame, 0, sizeof(hv_frame));
 	memset(&avi_frame, 0, sizeof(avi_frame));
@@ -5427,8 +6672,13 @@ static void fill_stream_properties_from_drm_display_mode(
 			&& stream->signal == SIGNAL_TYPE_HDMI_TYPE_A)
 		timing_out->pixel_encoding = PIXEL_ENCODING_YCBCR420;
 	else if (drm_mode_is_420_also(info, mode_in)
+			&& aconnector
 			&& aconnector->force_yuv420_output)
 		timing_out->pixel_encoding = PIXEL_ENCODING_YCBCR420;
+	else if ((connector->display_info.color_formats & DRM_COLOR_FORMAT_YCBCR422)
+			&& aconnector
+			&& aconnector->force_yuv422_output)
+		timing_out->pixel_encoding = PIXEL_ENCODING_YCBCR422;
 	else if ((connector->display_info.color_formats & DRM_COLOR_FORMAT_YCBCR444)
 			&& stream->signal == SIGNAL_TYPE_HDMI_TYPE_A)
 		timing_out->pixel_encoding = PIXEL_ENCODING_YCBCR444;
@@ -5456,13 +6706,23 @@ static void fill_stream_properties_from_drm_display_mode(
 	}
 
 	if (stream->signal == SIGNAL_TYPE_HDMI_TYPE_A) {
-		drm_hdmi_avi_infoframe_from_display_mode(&avi_frame, (struct drm_connector *)connector, mode_in);
+		err = drm_hdmi_avi_infoframe_from_display_mode(&avi_frame,
+							       (struct drm_connector *)connector,
+							       mode_in);
+		if (err < 0)
+			drm_warn_once(connector->dev, "Failed to setup avi infoframe on connector %s: %zd\n",
+				      connector->name, err);
 		timing_out->vic = avi_frame.video_code;
-		drm_hdmi_vendor_infoframe_from_display_mode(&hv_frame, (struct drm_connector *)connector, mode_in);
+		err = drm_hdmi_vendor_infoframe_from_display_mode(&hv_frame,
+								  (struct drm_connector *)connector,
+								  mode_in);
+		if (err < 0)
+			drm_warn_once(connector->dev, "Failed to setup vendor infoframe on connector %s: %zd\n",
+				      connector->name, err);
 		timing_out->hdmi_vic = hv_frame.vic;
 	}
 
-	if (is_freesync_video_mode(mode_in, aconnector)) {
+	if (aconnector && is_freesync_video_mode(mode_in, aconnector)) {
 		timing_out->h_addressable = mode_in->hdisplay;
 		timing_out->h_total = mode_in->htotal;
 		timing_out->h_sync_width = mode_in->hsync_end - mode_in->hsync_start;
@@ -5486,8 +6746,8 @@ static void fill_stream_properties_from_drm_display_mode(
 
 	timing_out->aspect_ratio = get_aspect_ratio(mode_in);
 
-	stream->out_transfer_func->type = TF_TYPE_PREDEFINED;
-	stream->out_transfer_func->tf = TRANSFER_FUNCTION_SRGB;
+	stream->out_transfer_func.type = TF_TYPE_PREDEFINED;
+	stream->out_transfer_func.tf = TRANSFER_FUNCTION_SRGB;
 	if (stream->signal == SIGNAL_TYPE_HDMI_TYPE_A) {
 		if (!adjust_colour_depth_from_display_info(timing_out, info) &&
 		    drm_mode_is_420_also(info, mode_in) &&
@@ -5497,7 +6757,8 @@ static void fill_stream_properties_from_drm_display_mode(
 		}
 	}
 
-	stream->output_color_space = get_output_color_space(timing_out);
+	stream->output_color_space = get_output_color_space(timing_out, connector_state);
+	stream->content_type = get_output_content_type(connector_state);
 }
 
 static void fill_audio_info(struct audio_info *audio_info,
@@ -5570,28 +6831,29 @@ decide_crtc_timing_for_drm_display_mode(struct drm_display_mode *drm_mode,
 					const struct drm_display_mode *native_mode,
 					bool scale_enabled)
 {
-	if (scale_enabled) {
-		copy_crtc_timing_for_drm_display_mode(native_mode, drm_mode);
-	} else if (native_mode->clock == drm_mode->clock &&
-			native_mode->htotal == drm_mode->htotal &&
-			native_mode->vtotal == drm_mode->vtotal) {
-		copy_crtc_timing_for_drm_display_mode(native_mode, drm_mode);
+	if (scale_enabled || (
+	    native_mode->clock == drm_mode->clock &&
+	    native_mode->htotal == drm_mode->htotal &&
+	    native_mode->vtotal == drm_mode->vtotal)) {
+		if (native_mode->crtc_clock)
+			copy_crtc_timing_for_drm_display_mode(native_mode, drm_mode);
 	} else {
 		/* no scaling nor amdgpu inserted, no need to patch */
 	}
 }
 
 static struct dc_sink *
-create_fake_sink(struct amdgpu_dm_connector *aconnector)
+create_fake_sink(struct drm_device *dev, struct dc_link *link)
 {
 	struct dc_sink_init_data sink_init_data = { 0 };
 	struct dc_sink *sink = NULL;
-	sink_init_data.link = aconnector->dc_link;
-	sink_init_data.sink_signal = aconnector->dc_link->connector_signal;
+
+	sink_init_data.link = link;
+	sink_init_data.sink_signal = link->connector_signal;
 
 	sink = dc_sink_create(&sink_init_data);
 	if (!sink) {
-		DRM_ERROR("Failed to create sink!\n");
+		drm_err(dev, "Failed to create sink!\n");
 		return NULL;
 	}
 	sink->sink_signal = SIGNAL_TYPE_VIRTUAL;
@@ -5705,11 +6967,14 @@ get_highest_refresh_rate_mode(struct amdgpu_dm_connector *aconnector,
 		&aconnector->base.probed_modes :
 		&aconnector->base.modes;
 
+	if (aconnector->base.connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+		return NULL;
+
 	if (aconnector->freesync_vid_base.clock != 0)
 		return &aconnector->freesync_vid_base;
 
 	/* Find the preferred mode */
-	list_for_each_entry (m, list_head, head) {
+	list_for_each_entry(m, list_head, head) {
 		if (m->type & DRM_MODE_TYPE_PREFERRED) {
 			m_pref = m;
 			break;
@@ -5721,7 +6986,7 @@ get_highest_refresh_rate_mode(struct amdgpu_dm_connector *aconnector,
 		m_pref = list_first_entry_or_null(
 				&aconnector->base.modes, struct drm_display_mode, head);
 		if (!m_pref) {
-			DRM_DEBUG_DRIVER("No preferred mode found in EDID\n");
+			drm_dbg_driver(aconnector->base.dev, "No preferred mode found in EDID\n");
 			return NULL;
 		}
 	}
@@ -5733,7 +6998,7 @@ get_highest_refresh_rate_mode(struct amdgpu_dm_connector *aconnector,
 	 * For some monitors, preferred mode is not the mode with highest
 	 * supported refresh rate.
 	 */
-	list_for_each_entry (m, list_head, head) {
+	list_for_each_entry(m, list_head, head) {
 		current_refresh  = drm_mode_vrefresh(m);
 
 		if (m->hdisplay == m_pref->hdisplay &&
@@ -5775,6 +7040,7 @@ static bool is_freesync_video_mode(const struct drm_display_mode *mode,
 		return true;
 }
 
+#if defined(CONFIG_DRM_AMD_DC_FP)
 static void update_dsc_caps(struct amdgpu_dm_connector *aconnector,
 			    struct dc_sink *sink, struct dc_stream_state *stream,
 			    struct dsc_dec_dpcd_caps *dsc_caps)
@@ -5793,7 +7059,6 @@ static void update_dsc_caps(struct amdgpu_dm_connector *aconnector,
 	}
 }
 
-
 static void apply_dsc_policy_for_edp(struct amdgpu_dm_connector *aconnector,
 				    struct dc_sink *sink, struct dc_stream_state *stream,
 				    struct dsc_dec_dpcd_caps *dsc_caps,
@@ -5826,6 +7091,7 @@ static void apply_dsc_policy_for_edp(struct amdgpu_dm_connector *aconnector,
 				edp_min_bpp_x16, edp_max_bpp_x16,
 				dsc_caps,
 				&stream->timing,
+				dc_link_get_highest_encoding_format(aconnector->dc_link),
 				&bw_range)) {
 
 		if (bw_range.max_kbps < link_bw_in_kbps) {
@@ -5834,6 +7100,7 @@ static void apply_dsc_policy_for_edp(struct amdgpu_dm_connector *aconnector,
 					&dsc_options,
 					0,
 					&stream->timing,
+					dc_link_get_highest_encoding_format(aconnector->dc_link),
 					&dsc_cfg)) {
 				stream->timing.dsc_cfg = dsc_cfg;
 				stream->timing.flags.DSC = 1;
@@ -5848,13 +7115,13 @@ static void apply_dsc_policy_for_edp(struct amdgpu_dm_connector *aconnector,
 				&dsc_options,
 				link_bw_in_kbps,
 				&stream->timing,
+				dc_link_get_highest_encoding_format(aconnector->dc_link),
 				&dsc_cfg)) {
 		stream->timing.dsc_cfg = dsc_cfg;
 		stream->timing.flags.DSC = 1;
 	}
 }
 
-
 static void apply_dsc_policy_for_stream(struct amdgpu_dm_connector *aconnector,
 					struct dc_sink *sink, struct dc_stream_state *stream,
 					struct dsc_dec_dpcd_caps *dsc_caps)
@@ -5878,25 +7145,28 @@ static void apply_dsc_policy_for_stream(struct amdgpu_dm_connector *aconnector,
 	dc_dsc_policy_set_enable_dsc_when_not_needed(
 		aconnector->dsc_settings.dsc_force_enable == DSC_CLK_FORCE_ENABLE);
 
-	if (aconnector->dc_link && sink->sink_signal == SIGNAL_TYPE_EDP &&
+	if (sink->sink_signal == SIGNAL_TYPE_EDP &&
 	    !aconnector->dc_link->panel_config.dsc.disable_dsc_edp &&
 	    dc->caps.edp_dsc_support && aconnector->dsc_settings.dsc_force_enable != DSC_CLK_FORCE_DISABLE) {
 
 		apply_dsc_policy_for_edp(aconnector, sink, stream, dsc_caps, max_dsc_target_bpp_limit_override);
 
-	} else if (aconnector->dc_link && sink->sink_signal == SIGNAL_TYPE_DISPLAY_PORT) {
+	} else if (sink->sink_signal == SIGNAL_TYPE_DISPLAY_PORT) {
 		if (sink->link->dpcd_caps.dongle_type == DISPLAY_DONGLE_NONE) {
 			if (dc_dsc_compute_config(aconnector->dc_link->ctx->dc->res_pool->dscs[0],
 						dsc_caps,
 						&dsc_options,
 						link_bandwidth_kbps,
 						&stream->timing,
+						dc_link_get_highest_encoding_format(aconnector->dc_link),
 						&stream->timing.dsc_cfg)) {
 				stream->timing.flags.DSC = 1;
-				DRM_DEBUG_DRIVER("%s: [%s] DSC is selected from SST RX\n", __func__, drm_connector->name);
+				drm_dbg_driver(drm_connector->dev, "%s: SST_DSC [%s] DSC is selected from SST RX\n",
+							__func__, drm_connector->name);
 			}
 		} else if (sink->link->dpcd_caps.dongle_type == DISPLAY_DONGLE_DP_HDMI_CONVERTER) {
-			timing_bw_in_kbps = dc_bandwidth_in_kbps_from_timing(&stream->timing);
+			timing_bw_in_kbps = dc_bandwidth_in_kbps_from_timing(&stream->timing,
+					dc_link_get_highest_encoding_format(aconnector->dc_link));
 			max_supported_bw_in_kbps = link_bandwidth_kbps;
 			dsc_max_supported_bw_in_kbps = link_bandwidth_kbps;
 
@@ -5908,9 +7178,10 @@ static void apply_dsc_policy_for_stream(struct amdgpu_dm_connector *aconnector,
 						&dsc_options,
 						dsc_max_supported_bw_in_kbps,
 						&stream->timing,
+						dc_link_get_highest_encoding_format(aconnector->dc_link),
 						&stream->timing.dsc_cfg)) {
 					stream->timing.flags.DSC = 1;
-					DRM_DEBUG_DRIVER("%s: [%s] DSC is selected from DP-HDMI PCON\n",
+					drm_dbg_driver(drm_connector->dev, "%s: SST_DSC [%s] DSC is selected from DP-HDMI PCON\n",
 									 __func__, drm_connector->name);
 				}
 		}
@@ -5929,46 +7200,56 @@ static void apply_dsc_policy_for_stream(struct amdgpu_dm_connector *aconnector,
 	if (stream->timing.flags.DSC && aconnector->dsc_settings.dsc_bits_per_pixel)
 		stream->timing.dsc_cfg.bits_per_pixel = aconnector->dsc_settings.dsc_bits_per_pixel;
 }
+#endif
 
 static struct dc_stream_state *
-create_stream_for_sink(struct amdgpu_dm_connector *aconnector,
+create_stream_for_sink(struct drm_connector *connector,
 		       const struct drm_display_mode *drm_mode,
 		       const struct dm_connector_state *dm_state,
 		       const struct dc_stream_state *old_stream,
 		       int requested_bpc)
 {
+	struct drm_device *dev = connector->dev;
+	struct amdgpu_dm_connector *aconnector = NULL;
 	struct drm_display_mode *preferred_mode = NULL;
-	struct drm_connector *drm_connector;
-	const struct drm_connector_state *con_state =
-		dm_state ? &dm_state->base : NULL;
+	const struct drm_connector_state *con_state = &dm_state->base;
 	struct dc_stream_state *stream = NULL;
 	struct drm_display_mode mode;
 	struct drm_display_mode saved_mode;
 	struct drm_display_mode *freesync_mode = NULL;
 	bool native_mode_found = false;
 	bool recalculate_timing = false;
-	bool scale = dm_state ? (dm_state->scaling != RMX_OFF) : false;
+	bool scale = dm_state->scaling != RMX_OFF;
 	int mode_refresh;
 	int preferred_refresh = 0;
 	enum color_transfer_func tf = TRANSFER_FUNC_UNKNOWN;
+#if defined(CONFIG_DRM_AMD_DC_FP)
 	struct dsc_dec_dpcd_caps dsc_caps;
-
+#endif
+	struct dc_link *link = NULL;
 	struct dc_sink *sink = NULL;
 
 	drm_mode_init(&mode, drm_mode);
 	memset(&saved_mode, 0, sizeof(saved_mode));
 
-	if (aconnector == NULL) {
-		DRM_ERROR("aconnector is NULL!\n");
-		return stream;
-	}
+	if (connector->connector_type != DRM_MODE_CONNECTOR_WRITEBACK) {
+		aconnector = NULL;
+		aconnector = to_amdgpu_dm_connector(connector);
+		link = aconnector->dc_link;
+	} else {
+		struct drm_writeback_connector *wbcon = NULL;
+		struct amdgpu_dm_wb_connector *dm_wbcon = NULL;
 
-	drm_connector = &aconnector->base;
+		wbcon = drm_connector_to_writeback(connector);
+		dm_wbcon = to_amdgpu_dm_wb_connector(wbcon);
+		link = dm_wbcon->link;
+	}
 
-	if (!aconnector->dc_sink) {
-		sink = create_fake_sink(aconnector);
+	if (!aconnector || !aconnector->dc_sink) {
+		sink = create_fake_sink(dev, link);
 		if (!sink)
 			return stream;
+
 	} else {
 		sink = aconnector->dc_sink;
 		dc_sink_retain(sink);
@@ -5977,16 +7258,17 @@ create_stream_for_sink(struct amdgpu_dm_connector *aconnector,
 	stream = dc_create_stream_for_sink(sink);
 
 	if (stream == NULL) {
-		DRM_ERROR("Failed to create stream for sink!\n");
+		drm_err(dev, "Failed to create stream for sink!\n");
 		goto finish;
 	}
 
+	/* We leave this NULL for writeback connectors */
 	stream->dm_stream_context = aconnector;
 
 	stream->timing.flags.LTE_340MCSC_SCRAMBLE =
-		drm_connector->display_info.hdmi.scdc.scrambling.low_rates;
+		connector->display_info.hdmi.scdc.scrambling.low_rates;
 
-	list_for_each_entry(preferred_mode, &aconnector->base.modes, head) {
+	list_for_each_entry(preferred_mode, &connector->modes, head) {
 		/* Search for preferred mode */
 		if (preferred_mode->type & DRM_MODE_TYPE_PREFERRED) {
 			native_mode_found = true;
@@ -5995,7 +7277,7 @@ create_stream_for_sink(struct amdgpu_dm_connector *aconnector,
 	}
 	if (!native_mode_found)
 		preferred_mode = list_first_entry_or_null(
-				&aconnector->base.modes,
+				&connector->modes,
 				struct drm_display_mode,
 				head);
 
@@ -6006,16 +7288,18 @@ create_stream_for_sink(struct amdgpu_dm_connector *aconnector,
 		 * This may not be an error, the use case is when we have no
 		 * usermode calls to reset and set mode upon hotplug. In this
 		 * case, we call set mode ourselves to restore the previous mode
-		 * and the modelist may not be filled in in time.
+		 * and the modelist may not be filled in time.
 		 */
-		DRM_DEBUG_DRIVER("No preferred mode found\n");
-	} else {
+		drm_dbg_driver(dev, "No preferred mode found\n");
+	} else if (aconnector) {
 		recalculate_timing = amdgpu_freesync_vid_mode &&
 				 is_freesync_video_mode(&mode, aconnector);
 		if (recalculate_timing) {
 			freesync_mode = get_highest_refresh_rate_mode(aconnector, false);
 			drm_mode_copy(&saved_mode, &mode);
+			saved_mode.picture_aspect_ratio = mode.picture_aspect_ratio;
 			drm_mode_copy(&mode, freesync_mode);
+			mode.picture_aspect_ratio = saved_mode.picture_aspect_ratio;
 		} else {
 			decide_crtc_timing_for_drm_display_mode(
 					&mode, preferred_mode, scale);
@@ -6026,40 +7310,44 @@ create_stream_for_sink(struct amdgpu_dm_connector *aconnector,
 
 	if (recalculate_timing)
 		drm_mode_set_crtcinfo(&saved_mode, 0);
-	else if (!dm_state)
-		drm_mode_set_crtcinfo(&mode, 0);
 
 	/*
-	* If scaling is enabled and refresh rate didn't change
-	* we copy the vic and polarities of the old timings
-	*/
+	 * If scaling is enabled and refresh rate didn't change
+	 * we copy the vic and polarities of the old timings
+	 */
 	if (!scale || mode_refresh != preferred_refresh)
 		fill_stream_properties_from_drm_display_mode(
-			stream, &mode, &aconnector->base, con_state, NULL,
+			stream, &mode, connector, con_state, NULL,
 			requested_bpc);
 	else
 		fill_stream_properties_from_drm_display_mode(
-			stream, &mode, &aconnector->base, con_state, old_stream,
+			stream, &mode, connector, con_state, old_stream,
 			requested_bpc);
 
+	/* The rest isn't needed for writeback connectors */
+	if (!aconnector)
+		goto finish;
+
 	if (aconnector->timing_changed) {
-		DC_LOG_DEBUG("%s: overriding timing for automated test, bpc %d, changing to %d\n",
-				__func__,
-				stream->timing.display_color_depth,
-				aconnector->timing_requested->display_color_depth);
+		drm_dbg(aconnector->base.dev,
+			"overriding timing for automated test, bpc %d, changing to %d\n",
+			stream->timing.display_color_depth,
+			aconnector->timing_requested->display_color_depth);
 		stream->timing = *aconnector->timing_requested;
 	}
 
+#if defined(CONFIG_DRM_AMD_DC_FP)
 	/* SST DSC determination policy */
 	update_dsc_caps(aconnector, sink, stream, &dsc_caps);
 	if (aconnector->dsc_settings.dsc_force_enable != DSC_CLK_FORCE_DISABLE && dsc_caps.is_dsc_supported)
 		apply_dsc_policy_for_stream(aconnector, sink, stream, &dsc_caps);
+#endif
 
 	update_stream_scaling_settings(&mode, dm_state, stream);
 
 	fill_audio_info(
 		&stream->audio_info,
-		drm_connector,
+		connector,
 		sink);
 
 	update_stream_signal(stream, sink);
@@ -6067,23 +7355,29 @@ create_stream_for_sink(struct amdgpu_dm_connector *aconnector,
 	if (stream->signal == SIGNAL_TYPE_HDMI_TYPE_A)
 		mod_build_hf_vsif_infopacket(stream, &stream->vsp_infopacket);
 
-	if (stream->link->psr_settings.psr_feature_enabled) {
+	if (stream->signal == SIGNAL_TYPE_DISPLAY_PORT ||
+	    stream->signal == SIGNAL_TYPE_DISPLAY_PORT_MST ||
+	    stream->signal == SIGNAL_TYPE_EDP) {
+		const struct dc_edid_caps *edid_caps;
+		unsigned int disable_colorimetry = 0;
+
+		if (aconnector->dc_sink) {
+			edid_caps = &aconnector->dc_sink->edid_caps;
+			disable_colorimetry = edid_caps->panel_patch.disable_colorimetry;
+		}
+
 		//
 		// should decide stream support vsc sdp colorimetry capability
 		// before building vsc info packet
 		//
-		stream->use_vsc_sdp_for_colorimetry = false;
-		if (aconnector->dc_sink->sink_signal == SIGNAL_TYPE_DISPLAY_PORT_MST) {
-			stream->use_vsc_sdp_for_colorimetry =
-				aconnector->dc_sink->is_vsc_sdp_colorimetry_supported;
-		} else {
-			if (stream->link->dpcd_caps.dprx_feature.bits.VSC_SDP_COLORIMETRY_SUPPORTED)
-				stream->use_vsc_sdp_for_colorimetry = true;
-		}
-		if (stream->out_transfer_func->tf == TRANSFER_FUNCTION_GAMMA22)
+		stream->use_vsc_sdp_for_colorimetry = stream->link->dpcd_caps.dpcd_rev.raw >= 0x14 &&
+						      stream->link->dpcd_caps.dprx_feature.bits.VSC_SDP_COLORIMETRY_SUPPORTED &&
+						      !disable_colorimetry;
+
+		if (stream->out_transfer_func.tf == TRANSFER_FUNCTION_GAMMA22)
 			tf = TRANSFER_FUNC_GAMMA_22;
 		mod_build_vsc_infopacket(stream, &stream->vsc_infopacket, stream->output_color_space, tf);
-		aconnector->psr_skip_count = AMDGPU_DM_PSR_ENTRY_DELAY;
+		aconnector->sr_skip_count = AMDGPU_DM_PSR_ENTRY_DELAY;
 
 	}
 finish:
@@ -6092,29 +7386,117 @@ finish:
 	return stream;
 }
 
+/**
+ * amdgpu_dm_connector_poll - Poll a connector to see if it's connected to a display
+ * @aconnector: DM connector to poll (owns @base drm_connector and @dc_link)
+ * @force: if true, force polling even when DAC load detection was used
+ *
+ * Used for connectors that don't support HPD (hotplug detection) to
+ * periodically check whether the connector is connected to a display.
+ *
+ * When connection was determined via DAC load detection, we avoid
+ * re-running it on normal polls to prevent visible glitches, unless
+ * @force is set.
+ *
+ * Return: The probed connector status (connected/disconnected/unknown).
+ */
 static enum drm_connector_status
-amdgpu_dm_connector_detect(struct drm_connector *connector, bool force)
+amdgpu_dm_connector_poll(struct amdgpu_dm_connector *aconnector, bool force)
 {
-	bool connected;
-	struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(connector);
+	struct drm_connector *connector = &aconnector->base;
+	struct drm_device *dev = connector->dev;
+	struct amdgpu_device *adev = drm_to_adev(dev);
+	struct dc_link *link = aconnector->dc_link;
+	enum dc_connection_type conn_type = dc_connection_none;
+	enum drm_connector_status status = connector_status_disconnected;
 
-	/*
-	 * Notes:
-	 * 1. This interface is NOT called in context of HPD irq.
-	 * 2. This interface *is called* in context of user-mode ioctl. Which
-	 * makes it a bad place for *any* MST-related activity.
+	/* When we determined the connection using DAC load detection,
+	 * do NOT poll the connector do detect disconnect because
+	 * that would run DAC load detection again which can cause
+	 * visible visual glitches.
+	 *
+	 * Only allow to poll such a connector again when forcing.
 	 */
+	if (!force && link->local_sink && link->type == dc_connection_dac_load)
+		return connector->status;
 
-	if (aconnector->base.force == DRM_FORCE_UNSPECIFIED &&
-	    !aconnector->fake_enable)
-		connected = (aconnector->dc_sink != NULL);
-	else
-		connected = (aconnector->base.force == DRM_FORCE_ON ||
-				aconnector->base.force == DRM_FORCE_ON_DIGITAL);
+	mutex_lock(&aconnector->hpd_lock);
+
+	if (dc_link_detect_connection_type(aconnector->dc_link, &conn_type) &&
+	    conn_type != dc_connection_none) {
+		mutex_lock(&adev->dm.dc_lock);
+
+		/* Only call full link detection when a sink isn't created yet,
+		 * ie. just when the display is plugged in, otherwise we risk flickering.
+		 */
+		if (link->local_sink ||
+			dc_link_detect(link, DETECT_REASON_HPD))
+			status = connector_status_connected;
+
+		mutex_unlock(&adev->dm.dc_lock);
+	}
+
+	if (connector->status != status) {
+		if (status == connector_status_disconnected) {
+			if (link->local_sink)
+				dc_sink_release(link->local_sink);
+
+			link->local_sink = NULL;
+			link->dpcd_sink_count = 0;
+			link->type = dc_connection_none;
+		}
+
+		amdgpu_dm_update_connector_after_detect(aconnector);
+	}
+
+	mutex_unlock(&aconnector->hpd_lock);
+	return status;
+}
+
+/**
+ * amdgpu_dm_connector_detect() - Detect whether a DRM connector is connected to a display
+ *
+ * A connector is considered connected when it has a sink that is not NULL.
+ * For connectors that support HPD (hotplug detection), the connection is
+ * handled in the HPD interrupt.
+ * For connectors that may not support HPD, such as analog connectors,
+ * DRM will call this function repeatedly to poll them.
+ *
+ * Notes:
+ * 1. This interface is NOT called in context of HPD irq.
+ * 2. This interface *is called* in context of user-mode ioctl. Which
+ *    makes it a bad place for *any* MST-related activity.
+ *
+ * @connector: The DRM connector we are checking. We convert it to
+ *             amdgpu_dm_connector so we can read the DC link and state.
+ * @force:     If true, do a full detect again. This is used even when
+ *             a lighter check would normally be used to avoid flicker.
+ *
+ * Return: The connector status (connected, disconnected, or unknown).
+ *
+ */
+static enum drm_connector_status
+amdgpu_dm_connector_detect(struct drm_connector *connector, bool force)
+{
+	struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(connector);
 
 	update_subconnector_property(aconnector);
 
-	return (connected ? connector_status_connected :
+	if (aconnector->base.force == DRM_FORCE_ON ||
+		aconnector->base.force == DRM_FORCE_ON_DIGITAL)
+		return connector_status_connected;
+	else if (aconnector->base.force == DRM_FORCE_OFF)
+		return connector_status_disconnected;
+
+	/* Poll analog connectors and only when either
+	 * disconnected or connected to an analog display.
+	 */
+	if (drm_kms_helper_is_poll_worker() &&
+		dc_connector_supports_analog(aconnector->dc_link->link_id.id) &&
+		(!aconnector->dc_sink || aconnector->dc_sink->edid_caps.analog))
+		return amdgpu_dm_connector_poll(aconnector, force);
+
+	return (aconnector->dc_sink ? connector_status_connected :
 			connector_status_disconnected);
 }
 
@@ -6166,7 +7548,18 @@ int amdgpu_dm_connector_atomic_set_property(struct drm_connector *connector,
 		dm_new_state->underscan_enable = val;
 		ret = 0;
 	} else if (property == adev->mode_info.abm_level_property) {
-		dm_new_state->abm_level = val;
+		switch (val) {
+		case ABM_SYSFS_CONTROL:
+			dm_new_state->abm_sysfs_forbidden = false;
+			break;
+		case ABM_LEVEL_OFF:
+			dm_new_state->abm_sysfs_forbidden = true;
+			dm_new_state->abm_level = ABM_LEVEL_IMMEDIATE_DISABLE;
+			break;
+		default:
+			dm_new_state->abm_sysfs_forbidden = true;
+			dm_new_state->abm_level = val;
+		}
 		ret = 0;
 	}
 
@@ -6211,17 +7604,122 @@ int amdgpu_dm_connector_atomic_get_property(struct drm_connector *connector,
 		*val = dm_state->underscan_enable;
 		ret = 0;
 	} else if (property == adev->mode_info.abm_level_property) {
-		*val = dm_state->abm_level;
+		if (!dm_state->abm_sysfs_forbidden)
+			*val = ABM_SYSFS_CONTROL;
+		else
+			*val = (dm_state->abm_level != ABM_LEVEL_IMMEDIATE_DISABLE) ?
+				dm_state->abm_level : 0;
 		ret = 0;
 	}
 
 	return ret;
 }
 
+/**
+ * DOC: panel power savings
+ *
+ * The display manager allows you to set your desired **panel power savings**
+ * level (between 0-4, with 0 representing off), e.g. using the following::
+ *
+ *   # echo 3 > /sys/class/drm/card0-eDP-1/amdgpu/panel_power_savings
+ *
+ * Modifying this value can have implications on color accuracy, so tread
+ * carefully.
+ */
+
+static ssize_t panel_power_savings_show(struct device *device,
+					struct device_attribute *attr,
+					char *buf)
+{
+	struct drm_connector *connector = dev_get_drvdata(device);
+	struct drm_device *dev = connector->dev;
+	u8 val;
+
+	drm_modeset_lock(&dev->mode_config.connection_mutex, NULL);
+	val = to_dm_connector_state(connector->state)->abm_level ==
+		ABM_LEVEL_IMMEDIATE_DISABLE ? 0 :
+		to_dm_connector_state(connector->state)->abm_level;
+	drm_modeset_unlock(&dev->mode_config.connection_mutex);
+
+	return sysfs_emit(buf, "%u\n", val);
+}
+
+static ssize_t panel_power_savings_store(struct device *device,
+					 struct device_attribute *attr,
+					 const char *buf, size_t count)
+{
+	struct drm_connector *connector = dev_get_drvdata(device);
+	struct drm_device *dev = connector->dev;
+	long val;
+	int ret;
+
+	ret = kstrtol(buf, 0, &val);
+
+	if (ret)
+		return ret;
+
+	if (val < 0 || val > 4)
+		return -EINVAL;
+
+	drm_modeset_lock(&dev->mode_config.connection_mutex, NULL);
+	if (to_dm_connector_state(connector->state)->abm_sysfs_forbidden)
+		ret = -EBUSY;
+	else
+		to_dm_connector_state(connector->state)->abm_level = val ?:
+			ABM_LEVEL_IMMEDIATE_DISABLE;
+	drm_modeset_unlock(&dev->mode_config.connection_mutex);
+
+	if (ret)
+		return ret;
+
+	drm_kms_helper_hotplug_event(dev);
+
+	return count;
+}
+
+static DEVICE_ATTR_RW(panel_power_savings);
+
+static struct attribute *amdgpu_attrs[] = {
+	&dev_attr_panel_power_savings.attr,
+	NULL
+};
+
+static const struct attribute_group amdgpu_group = {
+	.name = "amdgpu",
+	.attrs = amdgpu_attrs
+};
+
+static bool
+amdgpu_dm_should_create_sysfs(struct amdgpu_dm_connector *amdgpu_dm_connector)
+{
+	if (amdgpu_dm_abm_level >= 0)
+		return false;
+
+	if (amdgpu_dm_connector->base.connector_type != DRM_MODE_CONNECTOR_eDP)
+		return false;
+
+	/* check for OLED panels */
+	if (amdgpu_dm_connector->bl_idx >= 0) {
+		struct drm_device *drm = amdgpu_dm_connector->base.dev;
+		struct amdgpu_display_manager *dm = &drm_to_adev(drm)->dm;
+		struct amdgpu_dm_backlight_caps *caps;
+
+		caps = &dm->backlight_caps[amdgpu_dm_connector->bl_idx];
+		if (caps->aux_support)
+			return false;
+	}
+
+	return true;
+}
+
 static void amdgpu_dm_connector_unregister(struct drm_connector *connector)
 {
 	struct amdgpu_dm_connector *amdgpu_dm_connector = to_amdgpu_dm_connector(connector);
 
+	if (amdgpu_dm_should_create_sysfs(amdgpu_dm_connector))
+		sysfs_remove_group(&connector->kdev->kobj, &amdgpu_group);
+
+	cec_notifier_conn_unregister(amdgpu_dm_connector->notifier);
 	drm_dp_aux_unregister(&amdgpu_dm_connector->dm_dp_aux.aux);
 }
 
@@ -6238,6 +7736,13 @@ static void amdgpu_dm_connector_destroy(struct drm_connector *connector)
 	if (aconnector->mst_mgr.dev)
 		drm_dp_mst_topology_mgr_destroy(&aconnector->mst_mgr);
 
+	/* Cancel and flush any pending HDMI HPD debounce work */
+	cancel_delayed_work_sync(&aconnector->hdmi_hpd_debounce_work);
+	if (aconnector->hdmi_prev_sink) {
+		dc_sink_release(aconnector->hdmi_prev_sink);
+		aconnector->hdmi_prev_sink = NULL;
+	}
+
 	if (aconnector->bl_idx != -1) {
 		backlight_device_unregister(dm->backlight_dev[aconnector->bl_idx]);
 		dm->backlight_dev[aconnector->bl_idx] = NULL;
@@ -6253,10 +7758,6 @@ static void amdgpu_dm_connector_destroy(struct drm_connector *connector)
 	drm_dp_cec_unregister_connector(&aconnector->dm_dp_aux.aux);
 	drm_connector_unregister(connector);
 	drm_connector_cleanup(connector);
-	if (aconnector->i2c) {
-		i2c_del_adapter(&aconnector->i2c->base);
-		kfree(aconnector->i2c);
-	}
 	kfree(aconnector->dm_dp_aux.aux.name);
 
 	kfree(connector);
@@ -6283,8 +7784,12 @@ void amdgpu_dm_connector_funcs_reset(struct drm_connector *connector)
 		state->vcpi_slots = 0;
 		state->pbn = 0;
 
-		if (connector->connector_type == DRM_MODE_CONNECTOR_eDP)
-			state->abm_level = amdgpu_dm_abm_level;
+		if (connector->connector_type == DRM_MODE_CONNECTOR_eDP) {
+			if (amdgpu_dm_abm_level <= 0)
+				state->abm_level = ABM_LEVEL_IMMEDIATE_DISABLE;
+			else
+				state->abm_level = amdgpu_dm_abm_level;
+		}
 
 		__drm_atomic_helper_connector_reset(connector, &state->base);
 	}
@@ -6322,6 +7827,13 @@ amdgpu_dm_connector_late_register(struct drm_connector *connector)
 		to_amdgpu_dm_connector(connector);
 	int r;
 
+	if (amdgpu_dm_should_create_sysfs(amdgpu_dm_connector)) {
+		r = sysfs_create_group(&connector->kdev->kobj,
+				       &amdgpu_group);
+		if (r)
+			return r;
+	}
+
 	amdgpu_dm_register_backlight_device(amdgpu_dm_connector);
 
 	if ((connector->connector_type == DRM_MODE_CONNECTOR_DisplayPort) ||
@@ -6339,6 +7851,43 @@ amdgpu_dm_connector_late_register(struct drm_connector *connector)
 	return 0;
 }
 
+static void amdgpu_dm_connector_funcs_force(struct drm_connector *connector)
+{
+	struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(connector);
+	struct dc_link *dc_link = aconnector->dc_link;
+	struct dc_sink *dc_em_sink = aconnector->dc_em_sink;
+	const struct drm_edid *drm_edid;
+	struct i2c_adapter *ddc;
+	struct drm_device *dev = connector->dev;
+
+	if (dc_link && dc_link->aux_mode)
+		ddc = &aconnector->dm_dp_aux.aux.ddc;
+	else
+		ddc = &aconnector->i2c->base;
+
+	drm_edid = drm_edid_read_ddc(connector, ddc);
+	drm_edid_connector_update(connector, drm_edid);
+	if (!drm_edid) {
+		drm_err(dev, "No EDID found on connector: %s.\n", connector->name);
+		return;
+	}
+
+	aconnector->drm_edid = drm_edid;
+	/* Update emulated (virtual) sink's EDID */
+	if (dc_em_sink && dc_link) {
+		// FIXME: Get rid of drm_edid_raw()
+		const struct edid *edid = drm_edid_raw(drm_edid);
+
+		memset(&dc_em_sink->edid_caps, 0, sizeof(struct dc_edid_caps));
+		memmove(dc_em_sink->dc_edid.raw_edid, edid,
+			(edid->extensions + 1) * EDID_LENGTH);
+		dm_helpers_parse_edid_caps(
+			dc_link,
+			&dc_em_sink->dc_edid,
+			&dc_em_sink->edid_caps);
+	}
+}
+
 static const struct drm_connector_funcs amdgpu_dm_connector_funcs = {
 	.reset = amdgpu_dm_connector_funcs_reset,
 	.detect = amdgpu_dm_connector_detect,
@@ -6349,7 +7898,8 @@ static const struct drm_connector_funcs amdgpu_dm_connector_funcs = {
 	.atomic_set_property = amdgpu_dm_connector_atomic_set_property,
 	.atomic_get_property = amdgpu_dm_connector_atomic_get_property,
 	.late_register = amdgpu_dm_connector_late_register,
-	.early_unregister = amdgpu_dm_connector_unregister
+	.early_unregister = amdgpu_dm_connector_unregister,
+	.force = amdgpu_dm_connector_funcs_force
 };
 
 static int get_modes(struct drm_connector *connector)
@@ -6359,24 +7909,34 @@ static int get_modes(struct drm_connector *connector)
 
 static void create_eml_sink(struct amdgpu_dm_connector *aconnector)
 {
+	struct drm_connector *connector = &aconnector->base;
+	struct dc_link *dc_link = aconnector->dc_link;
 	struct dc_sink_init_data init_params = {
 			.link = aconnector->dc_link,
 			.sink_signal = SIGNAL_TYPE_VIRTUAL
 	};
-	struct edid *edid;
+	const struct drm_edid *drm_edid;
+	const struct edid *edid;
+	struct i2c_adapter *ddc;
 
-	if (!aconnector->base.edid_blob_ptr) {
-		DRM_ERROR("No EDID firmware found on connector: %s ,forcing to OFF!\n",
-				aconnector->base.name);
+	if (dc_link && dc_link->aux_mode)
+		ddc = &aconnector->dm_dp_aux.aux.ddc;
+	else
+		ddc = &aconnector->i2c->base;
 
-		aconnector->base.force = DRM_FORCE_OFF;
+	drm_edid = drm_edid_read_ddc(connector, ddc);
+	drm_edid_connector_update(connector, drm_edid);
+	if (!drm_edid) {
+		drm_err(connector->dev, "No EDID found on connector: %s.\n", connector->name);
 		return;
 	}
 
-	edid = (struct edid *) aconnector->base.edid_blob_ptr->data;
+	if (connector->display_info.is_hdmi)
+		init_params.sink_signal = SIGNAL_TYPE_HDMI_TYPE_A;
 
-	aconnector->edid = edid;
+	aconnector->drm_edid = drm_edid;
 
+	edid = drm_edid_raw(drm_edid); // FIXME: Get rid of drm_edid_raw()
 	aconnector->dc_em_sink = dc_link_add_remote_sink(
 		aconnector->dc_link,
 		(uint8_t *)edid,
@@ -6387,7 +7947,8 @@ static void create_eml_sink(struct amdgpu_dm_connector *aconnector)
 		aconnector->dc_sink = aconnector->dc_link->local_sink ?
 		aconnector->dc_link->local_sink :
 		aconnector->dc_em_sink;
-		dc_sink_retain(aconnector->dc_sink);
+		if (aconnector->dc_sink)
+			dc_sink_retain(aconnector->dc_sink);
 	}
 }
 
@@ -6421,7 +7982,7 @@ static enum dc_status dm_validate_stream_and_context(struct dc *dc,
 	if (!dc_plane_state)
 		goto cleanup;
 
-	dc_state = dc_create_state(dc);
+	dc_state = dc_state_create(dc, NULL);
 	if (!dc_state)
 		goto cleanup;
 
@@ -6448,9 +8009,9 @@ static enum dc_status dm_validate_stream_and_context(struct dc *dc,
 		dc_result = dc_validate_plane(dc, dc_plane_state);
 
 	if (dc_result == DC_OK)
-		dc_result = dc_add_stream_to_ctx(dc, dc_state, stream);
+		dc_result = dc_state_add_stream(dc, dc_state, stream);
 
-	if (dc_result == DC_OK && !dc_add_plane_to_context(
+	if (dc_result == DC_OK && !dc_state_add_plane(
 						dc,
 						stream,
 						dc_plane_state,
@@ -6458,11 +8019,11 @@ static enum dc_status dm_validate_stream_and_context(struct dc *dc,
 		dc_result = DC_FAIL_ATTACH_SURFACES;
 
 	if (dc_result == DC_OK)
-		dc_result = dc_validate_global_state(dc, dc_state, true);
+		dc_result = dc_validate_global_state(dc, dc_state, DC_VALIDATE_MODE_ONLY);
 
 cleanup:
 	if (dc_state)
-		dc_release_state(dc_state);
+		dc_state_release(dc_state);
 
 	if (dc_plane_state)
 		dc_plane_state_release(dc_plane_state);
@@ -6471,28 +8032,45 @@ cleanup:
 }
 
 struct dc_stream_state *
-create_validate_stream_for_sink(struct amdgpu_dm_connector *aconnector,
+create_validate_stream_for_sink(struct drm_connector *connector,
 				const struct drm_display_mode *drm_mode,
 				const struct dm_connector_state *dm_state,
 				const struct dc_stream_state *old_stream)
 {
-	struct drm_connector *connector = &aconnector->base;
+	struct amdgpu_dm_connector *aconnector = NULL;
 	struct amdgpu_device *adev = drm_to_adev(connector->dev);
 	struct dc_stream_state *stream;
 	const struct drm_connector_state *drm_state = dm_state ? &dm_state->base : NULL;
 	int requested_bpc = drm_state ? drm_state->max_requested_bpc : 8;
 	enum dc_status dc_result = DC_OK;
+	uint8_t bpc_limit = 6;
+
+	if (!dm_state)
+		return NULL;
+
+	if (connector->connector_type != DRM_MODE_CONNECTOR_WRITEBACK)
+		aconnector = to_amdgpu_dm_connector(connector);
+
+	if (aconnector &&
+	    (aconnector->dc_link->connector_signal == SIGNAL_TYPE_HDMI_TYPE_A ||
+	     aconnector->dc_link->dpcd_caps.dongle_type == DISPLAY_DONGLE_DP_HDMI_CONVERTER))
+		bpc_limit = 8;
 
 	do {
-		stream = create_stream_for_sink(aconnector, drm_mode,
+		drm_dbg_kms(connector->dev, "Trying with %d bpc\n", requested_bpc);
+		stream = create_stream_for_sink(connector, drm_mode,
 						dm_state, old_stream,
 						requested_bpc);
 		if (stream == NULL) {
-			DRM_ERROR("Failed to create stream for sink!\n");
+			drm_err(adev_to_drm(adev), "Failed to create stream for sink!\n");
 			break;
 		}
 
 		dc_result = dc_validate_stream(adev->dm.dc, stream);
+
+		if (!aconnector) /* writeback connector */
+			return stream;
+
 		if (dc_result == DC_OK && stream->signal == SIGNAL_TYPE_DISPLAY_PORT_MST)
 			dc_result = dm_dp_mst_is_port_support_mode(aconnector, stream);
 
@@ -6500,11 +8078,12 @@ create_validate_stream_for_sink(struct amdgpu_dm_connector *aconnector,
 			dc_result = dm_validate_stream_and_context(adev->dm.dc, stream);
 
 		if (dc_result != DC_OK) {
-			DRM_DEBUG_KMS("Mode %dx%d (clk %d) failed DC validation with error %d (%s)\n",
+			DRM_DEBUG_KMS("Pruned mode %d x %d (clk %d) %s %s -- %s\n",
 				      drm_mode->hdisplay,
 				      drm_mode->vdisplay,
 				      drm_mode->clock,
-				      dc_result,
+				      dc_pixel_encoding_to_str(stream->timing.pixel_encoding),
+				      dc_color_depth_to_str(stream->timing.display_color_depth),
 				      dc_status_to_str(dc_result));
 
 			dc_stream_release(stream);
@@ -6512,27 +8091,59 @@ create_validate_stream_for_sink(struct amdgpu_dm_connector *aconnector,
 			requested_bpc -= 2; /* lower bpc to retry validation */
 		}
 
-	} while (stream == NULL && requested_bpc >= 6);
-
-	if (dc_result == DC_FAIL_ENC_VALIDATE && !aconnector->force_yuv420_output) {
-		DRM_DEBUG_KMS("Retry forcing YCbCr420 encoding\n");
+	} while (stream == NULL && requested_bpc >= bpc_limit);
 
-		aconnector->force_yuv420_output = true;
-		stream = create_validate_stream_for_sink(aconnector, drm_mode,
-						dm_state, old_stream);
+	switch (dc_result) {
+	/*
+	 * If we failed to validate DP bandwidth stream with the requested RGB color depth,
+	 * we try to fallback and configure in order:
+	 * YUV422 (8bpc, 6bpc)
+	 * YUV420 (8bpc, 6bpc)
+	 */
+	case DC_FAIL_ENC_VALIDATE:
+	case DC_EXCEED_DONGLE_CAP:
+	case DC_NO_DP_LINK_BANDWIDTH:
+		/* recursively entered twice and already tried both YUV422 and YUV420 */
+		if (aconnector->force_yuv422_output && aconnector->force_yuv420_output)
+			break;
+		/* first failure; try YUV422 */
+		if (!aconnector->force_yuv422_output) {
+			drm_dbg_kms(connector->dev, "%s:%d Validation failed with %d, retrying w/ YUV422\n",
+				    __func__, __LINE__, dc_result);
+			aconnector->force_yuv422_output = true;
+		/* recursively entered and YUV422 failed, try YUV420 */
+		} else if (!aconnector->force_yuv420_output) {
+			drm_dbg_kms(connector->dev, "%s:%d Validation failed with %d, retrying w/ YUV420\n",
+				    __func__, __LINE__, dc_result);
+			aconnector->force_yuv420_output = true;
+		}
+		stream = create_validate_stream_for_sink(connector, drm_mode,
+							 dm_state, old_stream);
+		aconnector->force_yuv422_output = false;
 		aconnector->force_yuv420_output = false;
+		break;
+	case DC_OK:
+		break;
+	default:
+		drm_dbg_kms(connector->dev, "%s:%d Unhandled validation failure %d\n",
+			    __func__, __LINE__, dc_result);
+		break;
 	}
 
 	return stream;
 }
 
 enum drm_mode_status amdgpu_dm_connector_mode_valid(struct drm_connector *connector,
-				   struct drm_display_mode *mode)
+				   const struct drm_display_mode *mode)
 {
 	int result = MODE_ERROR;
 	struct dc_sink *dc_sink;
+	struct drm_display_mode *test_mode;
 	/* TODO: Unhardcode stream count */
 	struct dc_stream_state *stream;
+	/* we always have an amdgpu_dm_connector here since we got
+	 * here via the amdgpu_dm_connector_helper_funcs
+	 */
 	struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(connector);
 
 	if ((mode->flags & DRM_MODE_FLAG_INTERLACE) ||
@@ -6551,11 +8162,20 @@ enum drm_mode_status amdgpu_dm_connector_mode_valid(struct drm_connector *connec
 
 	if (dc_sink == NULL && aconnector->base.force != DRM_FORCE_ON_DIGITAL &&
 				aconnector->base.force != DRM_FORCE_ON) {
-		DRM_ERROR("dc_sink is NULL!\n");
+		drm_err(connector->dev, "dc_sink is NULL!\n");
 		goto fail;
 	}
 
-	stream = create_validate_stream_for_sink(aconnector, mode, NULL, NULL);
+	test_mode = drm_mode_duplicate(connector->dev, mode);
+	if (!test_mode)
+		goto fail;
+
+	drm_mode_set_crtcinfo(test_mode, 0);
+
+	stream = create_validate_stream_for_sink(connector, test_mode,
+						 to_dm_connector_state(connector->state),
+						 NULL);
+	drm_mode_destroy(connector->dev, test_mode);
 	if (stream) {
 		dc_stream_release(stream);
 		result = MODE_OK;
@@ -6638,6 +8258,9 @@ amdgpu_dm_connector_atomic_check(struct drm_connector *conn,
 	struct amdgpu_dm_connector *aconn = to_amdgpu_dm_connector(conn);
 	int ret;
 
+	if (WARN_ON(unlikely(!old_con_state || !new_con_state)))
+		return -EINVAL;
+
 	trace_amdgpu_dm_connector_atomic_check(new_con_state);
 
 	if (conn->connector_type == DRM_MODE_CONNECTOR_DisplayPort) {
@@ -6649,6 +8272,30 @@ amdgpu_dm_connector_atomic_check(struct drm_connector *conn,
 	if (!crtc)
 		return 0;
 
+	if (new_con_state->privacy_screen_sw_state != old_con_state->privacy_screen_sw_state) {
+		new_crtc_state = drm_atomic_get_crtc_state(state, crtc);
+		if (IS_ERR(new_crtc_state))
+			return PTR_ERR(new_crtc_state);
+
+		new_crtc_state->mode_changed = true;
+	}
+
+	if (new_con_state->colorspace != old_con_state->colorspace) {
+		new_crtc_state = drm_atomic_get_crtc_state(state, crtc);
+		if (IS_ERR(new_crtc_state))
+			return PTR_ERR(new_crtc_state);
+
+		new_crtc_state->mode_changed = true;
+	}
+
+	if (new_con_state->content_type != old_con_state->content_type) {
+		new_crtc_state = drm_atomic_get_crtc_state(state, crtc);
+		if (IS_ERR(new_crtc_state))
+			return PTR_ERR(new_crtc_state);
+
+		new_crtc_state->mode_changed = true;
+	}
+
 	if (!drm_connector_atomic_hdr_metadata_equal(old_con_state, new_con_state)) {
 		struct dc_info_packet hdr_infopacket;
 
@@ -6671,7 +8318,7 @@ amdgpu_dm_connector_atomic_check(struct drm_connector *conn,
 		 * set is permissible, however. So only force a
 		 * modeset if we're entering or exiting HDR.
 		 */
-		new_crtc_state->mode_changed =
+		new_crtc_state->mode_changed = new_crtc_state->mode_changed ||
 			!old_con_state->hdr_output_metadata ||
 			!new_con_state->hdr_output_metadata;
 	}
@@ -6734,7 +8381,24 @@ static int dm_encoder_helper_atomic_check(struct drm_encoder *encoder,
 	int clock, bpp = 0;
 	bool is_y420 = false;
 
-	if (!aconnector->mst_output_port || !aconnector->dc_sink)
+	if ((connector->connector_type == DRM_MODE_CONNECTOR_eDP) ||
+	    (connector->connector_type == DRM_MODE_CONNECTOR_LVDS)) {
+		struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
+		struct drm_display_mode *native_mode = &amdgpu_encoder->native_mode;
+		enum drm_mode_status result;
+
+		result = drm_crtc_helper_mode_valid_fixed(encoder->crtc, adjusted_mode, native_mode);
+		if (result != MODE_OK && dm_new_connector_state->scaling == RMX_OFF) {
+			drm_dbg_driver(encoder->dev,
+				       "mode %dx%d@%dHz is not native, enabling scaling\n",
+				       adjusted_mode->hdisplay, adjusted_mode->vdisplay,
+				       drm_mode_vrefresh(adjusted_mode));
+			dm_new_connector_state->scaling = RMX_ASPECT;
+		}
+		return 0;
+	}
+
+	if (!aconnector->mst_output_port)
 		return 0;
 
 	mst_port = aconnector->mst_output_port;
@@ -6747,11 +8411,11 @@ static int dm_encoder_helper_atomic_check(struct drm_encoder *encoder,
 	if (IS_ERR(mst_state))
 		return PTR_ERR(mst_state);
 
-	if (!mst_state->pbn_div)
-		mst_state->pbn_div = dm_mst_get_pbn_divider(aconnector->mst_root->dc_link);
+	mst_state->pbn_div.full = dm_mst_get_pbn_divider(aconnector->mst_root->dc_link);
 
 	if (!state->duplicated) {
 		int max_bpc = conn_state->max_requested_bpc;
+
 		is_y420 = drm_mode_is_420_also(&connector->display_info, adjusted_mode) &&
 			  aconnector->force_yuv420_output;
 		color_depth = convert_color_depth_from_display_info(connector,
@@ -6759,7 +8423,7 @@ static int dm_encoder_helper_atomic_check(struct drm_encoder *encoder,
 								    max_bpc);
 		bpp = convert_dc_color_depth_into_bpc(color_depth) * 3;
 		clock = adjusted_mode->clock;
-		dm_new_connector_state->pbn = drm_dp_calc_pbn_mode(clock, bpp, false);
+		dm_new_connector_state->pbn = drm_dp_calc_pbn_mode(clock, bpp << 4);
 	}
 
 	dm_new_connector_state->vcpi_slots =
@@ -6787,10 +8451,13 @@ static int dm_update_mst_vcpi_slots_for_dsc(struct drm_atomic_state *state,
 	struct amdgpu_dm_connector *aconnector;
 	struct dm_connector_state *dm_conn_state;
 	int i, j, ret;
-	int vcpi, pbn_div, pbn, slot_num = 0;
+	int vcpi, pbn_div, pbn = 0, slot_num = 0;
 
 	for_each_new_connector_in_state(state, connector, new_con_state, i) {
 
+		if (connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+			continue;
+
 		aconnector = to_amdgpu_dm_connector(connector);
 
 		if (!aconnector->mst_output_port)
@@ -6824,7 +8491,7 @@ static int dm_update_mst_vcpi_slots_for_dsc(struct drm_atomic_state *state,
 			}
 		}
 
-		if (j == dc_state->stream_count)
+		if (j == dc_state->stream_count || pbn_div == 0)
 			continue;
 
 		slot_num = DIV_ROUND_UP(pbn, pbn_div);
@@ -6851,7 +8518,7 @@ static int dm_update_mst_vcpi_slots_for_dsc(struct drm_atomic_state *state,
 	return 0;
 }
 
-static int to_drm_connector_type(enum signal_type st)
+static int to_drm_connector_type(enum signal_type st, uint32_t connector_id)
 {
 	switch (st) {
 	case SIGNAL_TYPE_HDMI_TYPE_A:
@@ -6867,6 +8534,10 @@ static int to_drm_connector_type(enum signal_type st)
 		return DRM_MODE_CONNECTOR_DisplayPort;
 	case SIGNAL_TYPE_DVI_DUAL_LINK:
 	case SIGNAL_TYPE_DVI_SINGLE_LINK:
+		if (connector_id == CONNECTOR_ID_SINGLE_LINK_DVII ||
+			connector_id == CONNECTOR_ID_DUAL_LINK_DVII)
+			return DRM_MODE_CONNECTOR_DVII;
+
 		return DRM_MODE_CONNECTOR_DVID;
 	case SIGNAL_TYPE_VIRTUAL:
 		return DRM_MODE_CONNECTOR_VIRTUAL;
@@ -6918,7 +8589,7 @@ static void amdgpu_dm_get_native_mode(struct drm_connector *connector)
 
 static struct drm_display_mode *
 amdgpu_dm_create_common_mode(struct drm_encoder *encoder,
-			     char *name,
+			     const char *name,
 			     int hdisplay, int vdisplay)
 {
 	struct drm_device *dev = encoder->dev;
@@ -6940,6 +8611,24 @@ amdgpu_dm_create_common_mode(struct drm_encoder *encoder,
 
 }
 
+static const struct amdgpu_dm_mode_size {
+	char name[DRM_DISPLAY_MODE_LEN];
+	int w;
+	int h;
+} common_modes[] = {
+	{  "640x480",  640,  480},
+	{  "800x600",  800,  600},
+	{ "1024x768", 1024,  768},
+	{ "1280x720", 1280,  720},
+	{ "1280x800", 1280,  800},
+	{"1280x1024", 1280, 1024},
+	{ "1440x900", 1440,  900},
+	{"1680x1050", 1680, 1050},
+	{"1600x1200", 1600, 1200},
+	{"1920x1080", 1920, 1080},
+	{"1920x1200", 1920, 1200}
+};
+
 static void amdgpu_dm_connector_add_common_modes(struct drm_encoder *encoder,
 						 struct drm_connector *connector)
 {
@@ -6950,23 +8639,10 @@ static void amdgpu_dm_connector_add_common_modes(struct drm_encoder *encoder,
 				to_amdgpu_dm_connector(connector);
 	int i;
 	int n;
-	struct mode_size {
-		char name[DRM_DISPLAY_MODE_LEN];
-		int w;
-		int h;
-	} common_modes[] = {
-		{  "640x480",  640,  480},
-		{  "800x600",  800,  600},
-		{ "1024x768", 1024,  768},
-		{ "1280x720", 1280,  720},
-		{ "1280x800", 1280,  800},
-		{"1280x1024", 1280, 1024},
-		{ "1440x900", 1440,  900},
-		{"1680x1050", 1680, 1050},
-		{"1600x1200", 1600, 1200},
-		{"1920x1080", 1920, 1080},
-		{"1920x1200", 1920, 1200}
-	};
+
+	if ((connector->connector_type != DRM_MODE_CONNECTOR_eDP) &&
+	    (connector->connector_type != DRM_MODE_CONNECTOR_LVDS))
+		return;
 
 	n = ARRAY_SIZE(common_modes);
 
@@ -7033,16 +8709,16 @@ static void amdgpu_set_panel_orientation(struct drm_connector *connector)
 }
 
 static void amdgpu_dm_connector_ddc_get_modes(struct drm_connector *connector,
-					      struct edid *edid)
+					      const struct drm_edid *drm_edid)
 {
 	struct amdgpu_dm_connector *amdgpu_dm_connector =
 			to_amdgpu_dm_connector(connector);
 
-	if (edid) {
+	if (drm_edid) {
 		/* empty probed_modes */
 		INIT_LIST_HEAD(&connector->probed_modes);
 		amdgpu_dm_connector->num_modes =
-				drm_add_edid_modes(connector, edid);
+				drm_edid_connector_add_modes(connector);
 
 		/* sorting the probed modes before calling function
 		 * amdgpu_dm_get_native_mode() since EDID can have
@@ -7056,10 +8732,10 @@ static void amdgpu_dm_connector_ddc_get_modes(struct drm_connector *connector,
 		amdgpu_dm_get_native_mode(connector);
 
 		/* Freesync capabilities are reset by calling
-		 * drm_add_edid_modes() and need to be
+		 * drm_edid_connector_add_modes() and need to be
 		 * restored here.
 		 */
-		amdgpu_dm_update_freesync_caps(connector, edid);
+		amdgpu_dm_update_freesync_caps(connector, drm_edid);
 	} else {
 		amdgpu_dm_connector->num_modes = 0;
 	}
@@ -7070,7 +8746,7 @@ static bool is_duplicate_mode(struct amdgpu_dm_connector *aconnector,
 {
 	struct drm_display_mode *m;
 
-	list_for_each_entry (m, &aconnector->base.probed_modes, head) {
+	list_for_each_entry(m, &aconnector->base.probed_modes, head) {
 		if (drm_mode_equal(m, mode))
 			return true;
 	}
@@ -7155,12 +8831,22 @@ static uint add_fs_modes(struct amdgpu_dm_connector *aconnector)
 }
 
 static void amdgpu_dm_connector_add_freesync_modes(struct drm_connector *connector,
-						   struct edid *edid)
+						   const struct drm_edid *drm_edid)
 {
 	struct amdgpu_dm_connector *amdgpu_dm_connector =
 		to_amdgpu_dm_connector(connector);
 
-	if (!(amdgpu_freesync_vid_mode && edid))
+	if (!(amdgpu_freesync_vid_mode && drm_edid))
+		return;
+
+	if (!amdgpu_dm_connector->dc_sink || !amdgpu_dm_connector->dc_link)
+		return;
+
+	if (!dc_supports_vrr(amdgpu_dm_connector->dc_sink->ctx->dce_version))
+		return;
+
+	if (dc_connector_supports_analog(amdgpu_dm_connector->dc_link->link_id.id) &&
+	    amdgpu_dm_connector->dc_sink->edid_caps.analog)
 		return;
 
 	if (amdgpu_dm_connector->max_vfreq - amdgpu_dm_connector->min_vfreq > 10)
@@ -7172,30 +8858,48 @@ static int amdgpu_dm_connector_get_modes(struct drm_connector *connector)
 {
 	struct amdgpu_dm_connector *amdgpu_dm_connector =
 			to_amdgpu_dm_connector(connector);
+	struct dc_link *dc_link = amdgpu_dm_connector->dc_link;
 	struct drm_encoder *encoder;
-	struct edid *edid = amdgpu_dm_connector->edid;
-	struct dc_link_settings *verified_link_cap =
-			&amdgpu_dm_connector->dc_link->verified_link_cap;
-	const struct dc *dc = amdgpu_dm_connector->dc_link->dc;
+	const struct drm_edid *drm_edid = amdgpu_dm_connector->drm_edid;
+	struct dc_link_settings *verified_link_cap = &dc_link->verified_link_cap;
+	const struct dc *dc = dc_link->dc;
 
 	encoder = amdgpu_dm_connector_to_encoder(connector);
 
-	if (!drm_edid_is_valid(edid)) {
+	if (!drm_edid) {
 		amdgpu_dm_connector->num_modes =
 				drm_add_modes_noedid(connector, 640, 480);
 		if (dc->link_srv->dp_get_encoding_format(verified_link_cap) == DP_128b_132b_ENCODING)
 			amdgpu_dm_connector->num_modes +=
 				drm_add_modes_noedid(connector, 1920, 1080);
+
+		if (amdgpu_dm_connector->dc_sink &&
+		    amdgpu_dm_connector->dc_sink->edid_caps.analog &&
+		    dc_connector_supports_analog(dc_link->link_id.id)) {
+			/* Analog monitor connected by DAC load detection.
+			 * Add common modes. It will be up to the user to select one that works.
+			 */
+			for (int i = 0; i < ARRAY_SIZE(common_modes); i++)
+				amdgpu_dm_connector->num_modes += drm_add_modes_noedid(
+					connector, common_modes[i].w, common_modes[i].h);
+		}
 	} else {
-		amdgpu_dm_connector_ddc_get_modes(connector, edid);
-		amdgpu_dm_connector_add_common_modes(encoder, connector);
-		amdgpu_dm_connector_add_freesync_modes(connector, edid);
+		amdgpu_dm_connector_ddc_get_modes(connector, drm_edid);
+		if (encoder)
+			amdgpu_dm_connector_add_common_modes(encoder, connector);
+		amdgpu_dm_connector_add_freesync_modes(connector, drm_edid);
 	}
 	amdgpu_dm_fbc_init(connector);
 
 	return amdgpu_dm_connector->num_modes;
 }
 
+static const u32 supported_colorspaces =
+	BIT(DRM_MODE_COLORIMETRY_BT709_YCC) |
+	BIT(DRM_MODE_COLORIMETRY_OPRGB) |
+	BIT(DRM_MODE_COLORIMETRY_BT2020_RGB) |
+	BIT(DRM_MODE_COLORIMETRY_BT2020_YCC);
+
 void amdgpu_dm_connector_init_helper(struct amdgpu_display_manager *dm,
 				     struct amdgpu_dm_connector *aconnector,
 				     int connector_type,
@@ -7224,6 +8928,11 @@ void amdgpu_dm_connector_init_helper(struct amdgpu_display_manager *dm,
 	aconnector->as_type = ADAPTIVE_SYNC_TYPE_NONE;
 	memset(&aconnector->vsdb_info, 0, sizeof(aconnector->vsdb_info));
 	mutex_init(&aconnector->hpd_lock);
+	mutex_init(&aconnector->handle_mst_msg_ready);
+
+	aconnector->hdmi_hpd_debounce_delay_ms = AMDGPU_DM_HDMI_HPD_DEBOUNCE_MS;
+	INIT_DELAYED_WORK(&aconnector->hdmi_hpd_debounce_work, hdmi_hpd_debounce_work);
+	aconnector->hdmi_prev_sink = NULL;
 
 	/*
 	 * configure support HPD hot plug connector_>polled default value is 0
@@ -7246,6 +8955,11 @@ void amdgpu_dm_connector_init_helper(struct amdgpu_display_manager *dm,
 	case DRM_MODE_CONNECTOR_DVID:
 		aconnector->base.polled = DRM_CONNECTOR_POLL_HPD;
 		break;
+	case DRM_MODE_CONNECTOR_DVII:
+	case DRM_MODE_CONNECTOR_VGA:
+		aconnector->base.polled =
+			DRM_CONNECTOR_POLL_CONNECT | DRM_CONNECTOR_POLL_DISCONNECT;
+		break;
 	default:
 		break;
 	}
@@ -7254,6 +8968,10 @@ void amdgpu_dm_connector_init_helper(struct amdgpu_display_manager *dm,
 				dm->ddev->mode_config.scaling_mode_property,
 				DRM_MODE_SCALE_NONE);
 
+	if (connector_type == DRM_MODE_CONNECTOR_HDMIA
+		|| (connector_type == DRM_MODE_CONNECTOR_DisplayPort && !aconnector->mst_root))
+		drm_connector_attach_broadcast_rgb_property(&aconnector->base);
+
 	drm_object_attach_property(&aconnector->base.base,
 				adev->mode_info.underscan_property,
 				UNDERSCAN_OFF);
@@ -7270,10 +8988,18 @@ void amdgpu_dm_connector_init_helper(struct amdgpu_display_manager *dm,
 	aconnector->base.state->max_bpc = 16;
 	aconnector->base.state->max_requested_bpc = aconnector->base.state->max_bpc;
 
-	if (connector_type == DRM_MODE_CONNECTOR_eDP &&
-	    (dc_is_dmcu_initialized(adev->dm.dc) || adev->dm.dc->ctx->dmub_srv)) {
-		drm_object_attach_property(&aconnector->base.base,
-				adev->mode_info.abm_level_property, 0);
+	if (connector_type == DRM_MODE_CONNECTOR_HDMIA) {
+		/* Content Type is currently only implemented for HDMI. */
+		drm_connector_attach_content_type_property(&aconnector->base);
+	}
+
+	if (connector_type == DRM_MODE_CONNECTOR_HDMIA) {
+		if (!drm_mode_create_hdmi_colorspace_property(&aconnector->base, supported_colorspaces))
+			drm_connector_attach_colorspace_property(&aconnector->base);
+	} else if ((connector_type == DRM_MODE_CONNECTOR_DisplayPort && !aconnector->mst_root) ||
+		   connector_type == DRM_MODE_CONNECTOR_eDP) {
+		if (!drm_mode_create_dp_colorspace_property(&aconnector->base, supported_colorspaces))
+			drm_connector_attach_colorspace_property(&aconnector->base);
 	}
 
 	if (connector_type == DRM_MODE_CONNECTOR_HDMIA ||
@@ -7287,6 +9013,18 @@ void amdgpu_dm_connector_init_helper(struct amdgpu_display_manager *dm,
 		if (adev->dm.hdcp_workqueue)
 			drm_connector_attach_content_protection_property(&aconnector->base, true);
 	}
+
+	if (connector_type == DRM_MODE_CONNECTOR_eDP) {
+		struct drm_privacy_screen *privacy_screen;
+
+		privacy_screen = drm_privacy_screen_get(adev_to_drm(adev)->dev, NULL);
+		if (!IS_ERR(privacy_screen)) {
+			drm_connector_attach_privacy_screen_provider(&aconnector->base,
+								     privacy_screen);
+		} else if (PTR_ERR(privacy_screen) != -ENODEV) {
+			drm_warn(adev_to_drm(adev), "Error getting privacy-screen\n");
+		}
+	}
 }
 
 static int amdgpu_dm_i2c_xfer(struct i2c_adapter *i2c_adap,
@@ -7298,6 +9036,9 @@ static int amdgpu_dm_i2c_xfer(struct i2c_adapter *i2c_adap,
 	int i;
 	int result = -EIO;
 
+	if (!ddc_service->ddc_pin)
+		return result;
+
 	cmd.payloads = kcalloc(num, sizeof(struct i2c_payload), GFP_KERNEL);
 
 	if (!cmd.payloads)
@@ -7314,11 +9055,18 @@ static int amdgpu_dm_i2c_xfer(struct i2c_adapter *i2c_adap,
 		cmd.payloads[i].data = msgs[i].buf;
 	}
 
-	if (dc_submit_i2c(
-			ddc_service->ctx->dc,
-			ddc_service->link->link_index,
-			&cmd))
-		result = num;
+	if (i2c->oem) {
+		if (dc_submit_i2c_oem(
+			    ddc_service->ctx->dc,
+			    &cmd))
+			result = num;
+	} else {
+		if (dc_submit_i2c(
+			    ddc_service->ctx->dc,
+			    ddc_service->link->link_index,
+			    &cmd))
+			result = num;
+	}
 
 	kfree(cmd.payloads);
 	return result;
@@ -7335,9 +9083,7 @@ static const struct i2c_algorithm amdgpu_dm_i2c_algo = {
 };
 
 static struct amdgpu_i2c_adapter *
-create_i2c(struct ddc_service *ddc_service,
-	   int link_index,
-	   int *res)
+create_i2c(struct ddc_service *ddc_service, bool oem)
 {
 	struct amdgpu_device *adev = ddc_service->ctx->driver_context;
 	struct amdgpu_i2c_adapter *i2c;
@@ -7346,16 +9092,41 @@ create_i2c(struct ddc_service *ddc_service,
 	if (!i2c)
 		return NULL;
 	i2c->base.owner = THIS_MODULE;
-	i2c->base.class = I2C_CLASS_DDC;
 	i2c->base.dev.parent = &adev->pdev->dev;
 	i2c->base.algo = &amdgpu_dm_i2c_algo;
-	snprintf(i2c->base.name, sizeof(i2c->base.name), "AMDGPU DM i2c hw bus %d", link_index);
+	if (oem)
+		snprintf(i2c->base.name, sizeof(i2c->base.name), "AMDGPU DM i2c OEM bus");
+	else
+		snprintf(i2c->base.name, sizeof(i2c->base.name), "AMDGPU DM i2c hw bus %d",
+			 ddc_service->link->link_index);
 	i2c_set_adapdata(&i2c->base, i2c);
 	i2c->ddc_service = ddc_service;
+	i2c->oem = oem;
 
 	return i2c;
 }
 
+int amdgpu_dm_initialize_hdmi_connector(struct amdgpu_dm_connector *aconnector)
+{
+	struct cec_connector_info conn_info;
+	struct drm_device *ddev = aconnector->base.dev;
+	struct device *hdmi_dev = ddev->dev;
+
+	if (amdgpu_dc_debug_mask & DC_DISABLE_HDMI_CEC) {
+		drm_info(ddev, "HDMI-CEC feature masked\n");
+		return -EINVAL;
+	}
+
+	cec_fill_conn_info_from_drm(&conn_info, &aconnector->base);
+	aconnector->notifier =
+		cec_notifier_conn_register(hdmi_dev, NULL, &conn_info);
+	if (!aconnector->notifier) {
+		drm_err(ddev, "Failed to create cec notifier\n");
+		return -ENOMEM;
+	}
+
+	return 0;
+}
 
 /*
  * Note: this function assumes that dc_link_detect() was called for the
@@ -7372,25 +9143,25 @@ static int amdgpu_dm_connector_init(struct amdgpu_display_manager *dm,
 	struct dc_link *link = dc_get_link_at_index(dc, link_index);
 	struct amdgpu_i2c_adapter *i2c;
 
+	/* Not needed for writeback connector */
 	link->priv = aconnector;
 
-	DRM_DEBUG_DRIVER("%s()\n", __func__);
 
-	i2c = create_i2c(link->ddc, link->link_index, &res);
+	i2c = create_i2c(link->ddc, false);
 	if (!i2c) {
-		DRM_ERROR("Failed to create i2c adapter data\n");
+		drm_err(adev_to_drm(dm->adev), "Failed to create i2c adapter data\n");
 		return -ENOMEM;
 	}
 
 	aconnector->i2c = i2c;
-	res = i2c_add_adapter(&i2c->base);
+	res = devm_i2c_add_adapter(dm->adev->dev, &i2c->base);
 
 	if (res) {
-		DRM_ERROR("Failed to register hw i2c %d\n", link->link_index);
+		drm_err(adev_to_drm(dm->adev), "Failed to register hw i2c %d\n", link->link_index);
 		goto out_free;
 	}
 
-	connector_type = to_drm_connector_type(link->connector_signal);
+	connector_type = to_drm_connector_type(link->connector_signal, link->link_id.id);
 
 	res = drm_connector_init_with_ddc(
 			dm->ddev,
@@ -7400,7 +9171,7 @@ static int amdgpu_dm_connector_init(struct amdgpu_display_manager *dm,
 			&i2c->base);
 
 	if (res) {
-		DRM_ERROR("connector_init failed\n");
+		drm_err(adev_to_drm(dm->adev), "connector_init failed\n");
 		aconnector->connector_id = -1;
 		goto out_free;
 	}
@@ -7419,6 +9190,10 @@ static int amdgpu_dm_connector_init(struct amdgpu_display_manager *dm,
 	drm_connector_attach_encoder(
 		&aconnector->base, &aencoder->base);
 
+	if (connector_type == DRM_MODE_CONNECTOR_HDMIA ||
+	    connector_type == DRM_MODE_CONNECTOR_HDMIB)
+		amdgpu_dm_initialize_hdmi_connector(aconnector);
+
 	if (connector_type == DRM_MODE_CONNECTOR_DisplayPort
 		|| connector_type == DRM_MODE_CONNECTOR_eDP)
 		amdgpu_dm_initialize_dp_connector(dm, aconnector, link->link_index);
@@ -7476,42 +9251,98 @@ static int amdgpu_dm_encoder_init(struct drm_device *dev,
 
 static void manage_dm_interrupts(struct amdgpu_device *adev,
 				 struct amdgpu_crtc *acrtc,
-				 bool enable)
-{
-	/*
-	 * We have no guarantee that the frontend index maps to the same
+				 struct dm_crtc_state *acrtc_state)
+{	/*
+	 * We cannot be sure that the frontend index maps to the same
 	 * backend index - some even map to more than one.
-	 *
-	 * TODO: Use a different interrupt or check DC itself for the mapping.
+	 * So we have to go through the CRTC to find the right IRQ.
 	 */
-	int irq_type =
-		amdgpu_display_crtc_idx_to_irq_type(
+	int irq_type = amdgpu_display_crtc_idx_to_irq_type(
 			adev,
 			acrtc->crtc_id);
+	struct drm_device *dev = adev_to_drm(adev);
 
-	if (enable) {
-		drm_crtc_vblank_on(&acrtc->base);
-		amdgpu_irq_get(
-			adev,
-			&adev->pageflip_irq,
-			irq_type);
+	struct drm_vblank_crtc_config config = {0};
+	struct dc_crtc_timing *timing;
+	int offdelay;
+
+	if (acrtc_state) {
+		timing = &acrtc_state->stream->timing;
+
+		/*
+		 * Depending on when the HW latching event of double-buffered
+		 * registers happen relative to the PSR SDP deadline, and how
+		 * bad the Panel clock has drifted since the last ALPM off
+		 * event, there can be up to 3 frames of delay between sending
+		 * the PSR exit cmd to DMUB fw, and when the panel starts
+		 * displaying live frames.
+		 *
+		 * We can set:
+		 *
+		 * 20/100 * offdelay_ms = 3_frames_ms
+		 * => offdelay_ms = 5 * 3_frames_ms
+		 *
+		 * This ensures that `3_frames_ms` will only be experienced as a
+		 * 20% delay on top how long the display has been static, and
+		 * thus make the delay less perceivable.
+		 */
+		if (acrtc_state->stream->link->psr_settings.psr_version <
+		    DC_PSR_VERSION_UNSUPPORTED) {
+			offdelay = DIV64_U64_ROUND_UP((u64)5 * 3 * 10 *
+						      timing->v_total *
+						      timing->h_total,
+						      timing->pix_clk_100hz);
+			config.offdelay_ms = offdelay ?: 30;
+		} else if (amdgpu_ip_version(adev, DCE_HWIP, 0) <
+			   IP_VERSION(3, 5, 0) ||
+			   !(adev->flags & AMD_IS_APU)) {
+			/*
+			 * Older HW and DGPU have issues with instant off;
+			 * use a 2 frame offdelay.
+			 */
+			offdelay = DIV64_U64_ROUND_UP((u64)20 *
+						      timing->v_total *
+						      timing->h_total,
+						      timing->pix_clk_100hz);
+
+			config.offdelay_ms = offdelay ?: 30;
+		} else {
+			/* offdelay_ms = 0 will never disable vblank */
+			config.offdelay_ms = 1;
+			config.disable_immediate = true;
+		}
+
+		drm_crtc_vblank_on_config(&acrtc->base,
+					  &config);
+		/* Allow RX6xxx, RX7700, RX7800 GPUs to call amdgpu_irq_get.*/
+		switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
+		case IP_VERSION(3, 0, 0):
+		case IP_VERSION(3, 0, 2):
+		case IP_VERSION(3, 0, 3):
+		case IP_VERSION(3, 2, 0):
+			if (amdgpu_irq_get(adev, &adev->pageflip_irq, irq_type))
+				drm_err(dev, "DM_IRQ: Cannot get pageflip irq!\n");
 #if defined(CONFIG_DRM_AMD_SECURE_DISPLAY)
-		amdgpu_irq_get(
-			adev,
-			&adev->vline0_irq,
-			irq_type);
+			if (amdgpu_irq_get(adev, &adev->vline0_irq, irq_type))
+				drm_err(dev, "DM_IRQ: Cannot get vline0 irq!\n");
 #endif
+		}
+
 	} else {
+		/* Allow RX6xxx, RX7700, RX7800 GPUs to call amdgpu_irq_put.*/
+		switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
+		case IP_VERSION(3, 0, 0):
+		case IP_VERSION(3, 0, 2):
+		case IP_VERSION(3, 0, 3):
+		case IP_VERSION(3, 2, 0):
 #if defined(CONFIG_DRM_AMD_SECURE_DISPLAY)
-		amdgpu_irq_put(
-			adev,
-			&adev->vline0_irq,
-			irq_type);
+			if (amdgpu_irq_put(adev, &adev->vline0_irq, irq_type))
+				drm_err(dev, "DM_IRQ: Cannot put vline0 irq!\n");
 #endif
-		amdgpu_irq_put(
-			adev,
-			&adev->pageflip_irq,
-			irq_type);
+			if (amdgpu_irq_put(adev, &adev->pageflip_irq, irq_type))
+				drm_err(dev, "DM_IRQ: Cannot put pageflip irq!\n");
+		}
+
 		drm_crtc_vblank_off(&acrtc->base);
 	}
 }
@@ -7691,8 +9522,9 @@ static void prepare_flip_isr(struct amdgpu_crtc *acrtc)
 	/* Mark this event as consumed */
 	acrtc->base.state->event = NULL;
 
-	DC_LOG_PFLIP("crtc:%d, pflip_stat:AMDGPU_FLIP_SUBMITTED\n",
-		     acrtc->crtc_id);
+	drm_dbg_state(acrtc->base.dev,
+		      "crtc:%d, pflip_stat:AMDGPU_FLIP_SUBMITTED\n",
+		      acrtc->crtc_id);
 }
 
 static void update_freesync_state_on_stream(
@@ -7747,7 +9579,7 @@ static void update_freesync_state_on_stream(
 
 	aconn = (struct amdgpu_dm_connector *)new_stream->dm_stream_context;
 
-	if (aconn && aconn->as_type == FREESYNC_TYPE_PCON_IN_WHITELIST) {
+	if (aconn && (aconn->as_type == FREESYNC_TYPE_PCON_IN_WHITELIST || aconn->vsdb_info.replay_mode)) {
 		pack_sdp_v1_3 = aconn->pack_sdp_v1_3;
 
 		if (aconn->vsdb_info.amd_vsdb_version == 1)
@@ -7866,7 +9698,7 @@ static void amdgpu_dm_handle_vrr_transition(struct dm_crtc_state *old_state,
 		 */
 		WARN_ON(amdgpu_dm_crtc_set_vupdate_irq(new_state->base.crtc, true) != 0);
 		WARN_ON(drm_crtc_vblank_get(new_state->base.crtc) != 0);
-		DRM_DEBUG_DRIVER("%s: crtc=%u VRR off->on: Get vblank ref\n",
+		drm_dbg_driver(new_state->base.crtc->dev, "%s: crtc=%u VRR off->on: Get vblank ref\n",
 				 __func__, new_state->base.crtc->base.id);
 	} else if (old_vrr_active && !new_vrr_active) {
 		/* Transition VRR active -> inactive:
@@ -7874,7 +9706,7 @@ static void amdgpu_dm_handle_vrr_transition(struct dm_crtc_state *old_state,
 		 */
 		WARN_ON(amdgpu_dm_crtc_set_vupdate_irq(new_state->base.crtc, false) != 0);
 		drm_crtc_vblank_put(new_state->base.crtc);
-		DRM_DEBUG_DRIVER("%s: crtc=%u VRR on->off: Drop vblank ref\n",
+		drm_dbg_driver(new_state->base.crtc->dev, "%s: crtc=%u VRR on->off: Drop vblank ref\n",
 				 __func__, new_state->base.crtc->base.id);
 	}
 }
@@ -7894,8 +9726,138 @@ static void amdgpu_dm_commit_cursors(struct drm_atomic_state *state)
 			amdgpu_dm_plane_handle_cursor_update(plane, old_plane_state);
 }
 
+static inline uint32_t get_mem_type(struct drm_framebuffer *fb)
+{
+	struct amdgpu_bo *abo = gem_to_amdgpu_bo(fb->obj[0]);
+
+	return abo->tbo.resource ? abo->tbo.resource->mem_type : 0;
+}
+
+static void amdgpu_dm_update_cursor(struct drm_plane *plane,
+				    struct drm_plane_state *old_plane_state,
+				    struct dc_stream_update *update)
+{
+	struct amdgpu_device *adev = drm_to_adev(plane->dev);
+	struct amdgpu_framebuffer *afb = to_amdgpu_framebuffer(plane->state->fb);
+	struct drm_crtc *crtc = afb ? plane->state->crtc : old_plane_state->crtc;
+	struct dm_crtc_state *crtc_state = crtc ? to_dm_crtc_state(crtc->state) : NULL;
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+	uint64_t address = afb ? afb->address : 0;
+	struct dc_cursor_position position = {0};
+	struct dc_cursor_attributes attributes;
+	int ret;
+
+	if (!plane->state->fb && !old_plane_state->fb)
+		return;
+
+	drm_dbg_atomic(plane->dev, "crtc_id=%d with size %d to %d\n",
+		       amdgpu_crtc->crtc_id, plane->state->crtc_w,
+		       plane->state->crtc_h);
+
+	ret = amdgpu_dm_plane_get_cursor_position(plane, crtc, &position);
+	if (ret)
+		return;
+
+	if (!position.enable) {
+		/* turn off cursor */
+		if (crtc_state && crtc_state->stream) {
+			dc_stream_set_cursor_position(crtc_state->stream,
+						      &position);
+			update->cursor_position = &crtc_state->stream->cursor_position;
+		}
+		return;
+	}
+
+	amdgpu_crtc->cursor_width = plane->state->crtc_w;
+	amdgpu_crtc->cursor_height = plane->state->crtc_h;
+
+	memset(&attributes, 0, sizeof(attributes));
+	attributes.address.high_part = upper_32_bits(address);
+	attributes.address.low_part  = lower_32_bits(address);
+	attributes.width             = plane->state->crtc_w;
+	attributes.height            = plane->state->crtc_h;
+	attributes.color_format      = CURSOR_MODE_COLOR_PRE_MULTIPLIED_ALPHA;
+	attributes.rotation_angle    = 0;
+	attributes.attribute_flags.value = 0;
+
+	/* Enable cursor degamma ROM on DCN3+ for implicit sRGB degamma in DRM
+	 * legacy gamma setup.
+	 */
+	if (crtc_state->cm_is_degamma_srgb &&
+	    adev->dm.dc->caps.color.dpp.gamma_corr)
+		attributes.attribute_flags.bits.ENABLE_CURSOR_DEGAMMA = 1;
+
+	if (afb)
+		attributes.pitch = afb->base.pitches[0] / afb->base.format->cpp[0];
+
+	if (crtc_state->stream) {
+		if (!dc_stream_set_cursor_attributes(crtc_state->stream,
+						     &attributes))
+			drm_err(adev_to_drm(adev), "DC failed to set cursor attributes\n");
+
+		update->cursor_attributes = &crtc_state->stream->cursor_attributes;
+
+		if (!dc_stream_set_cursor_position(crtc_state->stream,
+						   &position))
+			drm_err(adev_to_drm(adev), "DC failed to set cursor position\n");
+
+		update->cursor_position = &crtc_state->stream->cursor_position;
+	}
+}
+
+static void amdgpu_dm_enable_self_refresh(struct amdgpu_crtc *acrtc_attach,
+					  const struct dm_crtc_state *acrtc_state,
+					  const u64 current_ts)
+{
+	struct psr_settings *psr = &acrtc_state->stream->link->psr_settings;
+	struct replay_settings *pr = &acrtc_state->stream->link->replay_settings;
+	struct amdgpu_dm_connector *aconn =
+		(struct amdgpu_dm_connector *)acrtc_state->stream->dm_stream_context;
+	bool vrr_active = amdgpu_dm_crtc_vrr_active(acrtc_state);
+
+	if (acrtc_state->update_type > UPDATE_TYPE_FAST) {
+		if (pr->config.replay_supported && !pr->replay_feature_enabled)
+			amdgpu_dm_link_setup_replay(acrtc_state->stream->link, aconn);
+		else if (psr->psr_version != DC_PSR_VERSION_UNSUPPORTED &&
+			     !psr->psr_feature_enabled)
+			if (!aconn->disallow_edp_enter_psr)
+				amdgpu_dm_link_setup_psr(acrtc_state->stream);
+	}
+
+	/* Decrement skip count when SR is enabled and we're doing fast updates. */
+	if (acrtc_state->update_type == UPDATE_TYPE_FAST &&
+	    (psr->psr_feature_enabled || pr->config.replay_supported)) {
+		if (aconn->sr_skip_count > 0)
+			aconn->sr_skip_count--;
+
+		/* Allow SR when skip count is 0. */
+		acrtc_attach->dm_irq_params.allow_sr_entry = !aconn->sr_skip_count;
+
+		/*
+		 * If sink supports PSR SU/Panel Replay, there is no need to rely on
+		 * a vblank event disable request to enable PSR/RP. PSR SU/RP
+		 * can be enabled immediately once OS demonstrates an
+		 * adequate number of fast atomic commits to notify KMD
+		 * of update events. See `vblank_control_worker()`.
+		 */
+		if (!vrr_active &&
+		    acrtc_attach->dm_irq_params.allow_sr_entry &&
+#ifdef CONFIG_DRM_AMD_SECURE_DISPLAY
+		    !amdgpu_dm_crc_window_is_activated(acrtc_state->base.crtc) &&
+#endif
+		    (current_ts - psr->psr_dirty_rects_change_timestamp_ns) > 500000000) {
+			if (pr->replay_feature_enabled && !pr->replay_allow_active)
+				amdgpu_dm_replay_enable(acrtc_state->stream, true);
+			if (psr->psr_version == DC_PSR_VERSION_SU_1 &&
+			    !psr->psr_allow_active && !aconn->disallow_edp_enter_psr)
+				amdgpu_dm_psr_enable(acrtc_state->stream);
+		}
+	} else {
+		acrtc_attach->dm_irq_params.allow_sr_entry = false;
+	}
+}
+
 static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
-				    struct dc_state *dc_state,
 				    struct drm_device *dev,
 				    struct amdgpu_display_manager *dm,
 				    struct drm_crtc *pcrtc,
@@ -7918,6 +9880,7 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
 	bool cursor_update = false;
 	bool pflip_present = false;
 	bool dirty_rects_changed = false;
+	bool updated_planes_and_streams = false;
 	struct {
 		struct dc_surface_update surface_updates[MAX_SURFACES];
 		struct dc_plane_info plane_infos[MAX_SURFACES];
@@ -7929,7 +9892,7 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
 	bundle = kzalloc(sizeof(*bundle), GFP_KERNEL);
 
 	if (!bundle) {
-		dm_error("Failed to allocate update bundle\n");
+		drm_err(dev, "Failed to allocate update bundle\n");
 		goto cleanup;
 	}
 
@@ -7937,8 +9900,24 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
 	 * Disable the cursor first if we're disabling all the planes.
 	 * It'll remain on the screen after the planes are re-enabled
 	 * if we don't.
+	 *
+	 * If the cursor is transitioning from native to overlay mode, the
+	 * native cursor needs to be disabled first.
 	 */
-	if (acrtc_state->active_planes == 0)
+	if (acrtc_state->cursor_mode == DM_CURSOR_OVERLAY_MODE &&
+	    dm_old_crtc_state->cursor_mode == DM_CURSOR_NATIVE_MODE) {
+		struct dc_cursor_position cursor_position = {0};
+
+		if (!dc_stream_set_cursor_position(acrtc_state->stream,
+						   &cursor_position))
+			drm_err(dev, "DC failed to disable native cursor\n");
+
+		bundle->stream_update.cursor_position =
+				&acrtc_state->stream->cursor_position;
+	}
+
+	if (acrtc_state->active_planes == 0 &&
+	    dm_old_crtc_state->cursor_mode == DM_CURSOR_NATIVE_MODE)
 		amdgpu_dm_commit_cursors(state);
 
 	/* update planes when needed */
@@ -7952,10 +9931,14 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
 		struct dm_plane_state *dm_new_plane_state = to_dm_plane_state(new_plane_state);
 
 		/* Cursor plane is handled after stream updates */
-		if (plane->type == DRM_PLANE_TYPE_CURSOR) {
+		if (plane->type == DRM_PLANE_TYPE_CURSOR &&
+		    acrtc_state->cursor_mode == DM_CURSOR_NATIVE_MODE) {
 			if ((fb && crtc == pcrtc) ||
-			    (old_plane_state->fb && old_plane_state->crtc == pcrtc))
+			    (old_plane_state->fb && old_plane_state->crtc == pcrtc)) {
 				cursor_update = true;
+				if (amdgpu_ip_version(dm->adev, DCE_HWIP, 0) != 0)
+					amdgpu_dm_update_cursor(plane, old_plane_state, &bundle->stream_update);
+			}
 
 			continue;
 		}
@@ -7968,12 +9951,18 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
 			continue;
 
 		dc_plane = dm_new_plane_state->dc_state;
+		if (!dc_plane)
+			continue;
 
 		bundle->surface_updates[planes_count].surface = dc_plane;
 		if (new_pcrtc_state->color_mgmt_changed) {
-			bundle->surface_updates[planes_count].gamma = dc_plane->gamma_correction;
-			bundle->surface_updates[planes_count].in_transfer_func = dc_plane->in_transfer_func;
+			bundle->surface_updates[planes_count].gamma = &dc_plane->gamma_correction;
+			bundle->surface_updates[planes_count].in_transfer_func = &dc_plane->in_transfer_func;
 			bundle->surface_updates[planes_count].gamut_remap_matrix = &dc_plane->gamut_remap_matrix;
+			bundle->surface_updates[planes_count].hdr_mult = dc_plane->hdr_mult;
+			bundle->surface_updates[planes_count].func_shaper = &dc_plane->in_shaper_func;
+			bundle->surface_updates[planes_count].lut3d_func = &dc_plane->lut3d_func;
+			bundle->surface_updates[planes_count].blend_tf = &dc_plane->blend_tf;
 		}
 
 		amdgpu_dm_plane_fill_dc_scaling_info(dm->adev, new_plane_state,
@@ -7996,7 +9985,7 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
 			afb->tiling_flags,
 			&bundle->plane_infos[planes_count],
 			&bundle->flip_addrs[planes_count].address,
-			afb->tmz_surface, false);
+			afb->tmz_surface);
 
 		drm_dbg_state(state->dev, "plane: id=%d dcc_en=%d\n",
 				 new_plane_state->plane->index,
@@ -8005,10 +9994,13 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
 		bundle->surface_updates[planes_count].plane_info =
 			&bundle->plane_infos[planes_count];
 
-		if (acrtc_state->stream->link->psr_settings.psr_feature_enabled) {
+		if (acrtc_state->stream->link->psr_settings.psr_feature_enabled ||
+		    acrtc_state->stream->link->replay_settings.replay_feature_enabled) {
 			fill_dc_dirty_rects(plane, old_plane_state,
 					    new_plane_state, new_crtc_state,
 					    &bundle->flip_addrs[planes_count],
+					    acrtc_state->stream->link->psr_settings.psr_version ==
+					    DC_PSR_VERSION_SU_1,
 					    &dirty_rects_changed);
 
 			/*
@@ -8018,7 +10010,7 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
 			 * during the PSR-SU was disabled.
 			 */
 			if (acrtc_state->stream->link->psr_settings.psr_version >= DC_PSR_VERSION_SU_1 &&
-			    acrtc_attach->dm_irq_params.allow_psr_entry &&
+			    acrtc_attach->dm_irq_params.allow_sr_entry &&
 #ifdef CONFIG_DRM_AMD_SECURE_DISPLAY
 			    !amdgpu_dm_crc_window_is_activated(acrtc_state->base.crtc) &&
 #endif
@@ -8027,18 +10019,30 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
 				acrtc_state->stream->link->psr_settings.psr_dirty_rects_change_timestamp_ns =
 				timestamp_ns;
 				if (acrtc_state->stream->link->psr_settings.psr_allow_active)
-					amdgpu_dm_psr_disable(acrtc_state->stream);
+					amdgpu_dm_psr_disable(acrtc_state->stream, true);
 				mutex_unlock(&dm->dc_lock);
 			}
 		}
 
 		/*
 		 * Only allow immediate flips for fast updates that don't
-		 * change FB pitch, DCC state, rotation or mirroing.
+		 * change memory domain, FB pitch, DCC state, rotation or
+		 * mirroring.
+		 *
+		 * dm_crtc_helper_atomic_check() only accepts async flips with
+		 * fast updates.
 		 */
+		if (crtc->state->async_flip &&
+		    (acrtc_state->update_type != UPDATE_TYPE_FAST ||
+		     get_mem_type(old_plane_state->fb) != get_mem_type(fb)))
+			drm_warn_once(state->dev,
+				      "[PLANE:%d:%s] async flip with non-fast update\n",
+				      plane->base.id, plane->name);
+
 		bundle->flip_addrs[planes_count].flip_immediate =
 			crtc->state->async_flip &&
-			acrtc_state->update_type == UPDATE_TYPE_FAST;
+			acrtc_state->update_type == UPDATE_TYPE_FAST &&
+			get_mem_type(old_plane_state->fb) == get_mem_type(fb);
 
 		timestamp_ns = ktime_get_ns();
 		bundle->flip_addrs[planes_count].flip_timestamp_in_us = div_u64(timestamp_ns, 1000);
@@ -8046,7 +10050,7 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
 		bundle->surface_updates[planes_count].surface = dc_plane;
 
 		if (!bundle->surface_updates[planes_count].surface) {
-			DRM_ERROR("No surface for CRTC: id=%d\n",
+			drm_err(dev, "No surface for CRTC: id=%d\n",
 					acrtc_attach->crtc_id);
 			continue;
 		}
@@ -8077,8 +10081,7 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
 			 * DRI3/Present extension with defined target_msc.
 			 */
 			last_flip_vblank = amdgpu_get_vblank_counter_kms(pcrtc);
-		}
-		else {
+		} else {
 			/* For variable refresh rate mode only:
 			 * Get vblank of last completed flip to avoid > 1 vrr
 			 * flips per video frame by use of throttling, but allow
@@ -8134,15 +10137,13 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
 				bundle->stream_update.vrr_infopacket =
 					&acrtc_state->stream->vrr_infopacket;
 		}
-	} else if (cursor_update && acrtc_state->active_planes > 0 &&
-		   acrtc_attach->base.state->event) {
-		drm_crtc_vblank_get(pcrtc);
-
+	} else if (cursor_update && acrtc_state->active_planes > 0) {
 		spin_lock_irqsave(&pcrtc->dev->event_lock, flags);
-
-		acrtc_attach->event = acrtc_attach->base.state->event;
-		acrtc_attach->base.state->event = NULL;
-
+		if (acrtc_attach->base.state->event) {
+			drm_crtc_vblank_get(pcrtc);
+			acrtc_attach->event = acrtc_attach->base.state->event;
+			acrtc_attach->base.state->event = NULL;
+		}
 		spin_unlock_irqrestore(&pcrtc->dev->event_lock, flags);
 	}
 
@@ -8172,13 +10173,26 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
 			bundle->stream_update.output_csc_transform =
 				&acrtc_state->stream->csc_color_matrix;
 			bundle->stream_update.out_transfer_func =
-				acrtc_state->stream->out_transfer_func;
+				&acrtc_state->stream->out_transfer_func;
+			bundle->stream_update.lut3d_func =
+				(struct dc_3dlut *) acrtc_state->stream->lut3d_func;
+			bundle->stream_update.func_shaper =
+				(struct dc_transfer_func *) acrtc_state->stream->func_shaper;
 		}
 
 		acrtc_state->stream->abm_level = acrtc_state->abm_level;
 		if (acrtc_state->abm_level != dm_old_crtc_state->abm_level)
 			bundle->stream_update.abm_level = &acrtc_state->abm_level;
 
+		mutex_lock(&dm->dc_lock);
+		if ((acrtc_state->update_type > UPDATE_TYPE_FAST) || vrr_active) {
+			if (acrtc_state->stream->link->replay_settings.replay_allow_active)
+				amdgpu_dm_replay_disable(acrtc_state->stream);
+			if (acrtc_state->stream->link->psr_settings.psr_allow_active)
+				amdgpu_dm_psr_disable(acrtc_state->stream, true);
+		}
+		mutex_unlock(&dm->dc_lock);
+
 		/*
 		 * If FreeSync state on the stream has changed then we need to
 		 * re-adjust the min/max bounds now that DC doesn't handle this
@@ -8192,16 +10206,13 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
 			spin_unlock_irqrestore(&pcrtc->dev->event_lock, flags);
 		}
 		mutex_lock(&dm->dc_lock);
-		if ((acrtc_state->update_type > UPDATE_TYPE_FAST) &&
-				acrtc_state->stream->link->psr_settings.psr_allow_active)
-			amdgpu_dm_psr_disable(acrtc_state->stream);
-
 		update_planes_and_stream_adapter(dm->dc,
 					 acrtc_state->update_type,
 					 planes_count,
 					 acrtc_state->stream,
 					 &bundle->stream_update,
 					 bundle->surface_updates);
+		updated_planes_and_streams = true;
 
 		/**
 		 * Enable or disable the interrupts on the backend.
@@ -8220,44 +10231,7 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
 			dm_update_pflip_irq_state(drm_to_adev(dev),
 						  acrtc_attach);
 
-		if ((acrtc_state->update_type > UPDATE_TYPE_FAST) &&
-				acrtc_state->stream->link->psr_settings.psr_version != DC_PSR_VERSION_UNSUPPORTED &&
-				!acrtc_state->stream->link->psr_settings.psr_feature_enabled)
-			amdgpu_dm_link_setup_psr(acrtc_state->stream);
-
-		/* Decrement skip count when PSR is enabled and we're doing fast updates. */
-		if (acrtc_state->update_type == UPDATE_TYPE_FAST &&
-		    acrtc_state->stream->link->psr_settings.psr_feature_enabled) {
-			struct amdgpu_dm_connector *aconn =
-				(struct amdgpu_dm_connector *)acrtc_state->stream->dm_stream_context;
-
-			if (aconn->psr_skip_count > 0)
-				aconn->psr_skip_count--;
-
-			/* Allow PSR when skip count is 0. */
-			acrtc_attach->dm_irq_params.allow_psr_entry = !aconn->psr_skip_count;
-
-			/*
-			 * If sink supports PSR SU, there is no need to rely on
-			 * a vblank event disable request to enable PSR. PSR SU
-			 * can be enabled immediately once OS demonstrates an
-			 * adequate number of fast atomic commits to notify KMD
-			 * of update events. See `vblank_control_worker()`.
-			 */
-			if (acrtc_state->stream->link->psr_settings.psr_version >= DC_PSR_VERSION_SU_1 &&
-			    acrtc_attach->dm_irq_params.allow_psr_entry &&
-#ifdef CONFIG_DRM_AMD_SECURE_DISPLAY
-			    !amdgpu_dm_crc_window_is_activated(acrtc_state->base.crtc) &&
-#endif
-			    !acrtc_state->stream->link->psr_settings.psr_allow_active &&
-			    (timestamp_ns -
-			    acrtc_state->stream->link->psr_settings.psr_dirty_rects_change_timestamp_ns) >
-			    500000000)
-				amdgpu_dm_psr_enable(acrtc_state->stream);
-		} else {
-			acrtc_attach->dm_irq_params.allow_psr_entry = false;
-		}
-
+		amdgpu_dm_enable_self_refresh(acrtc_attach, acrtc_state, timestamp_ns);
 		mutex_unlock(&dm->dc_lock);
 	}
 
@@ -8266,7 +10240,9 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
 	 * This avoids redundant programming in the case where we're going
 	 * to be disabling a single plane - those pipes are being disabled.
 	 */
-	if (acrtc_state->active_planes)
+	if (acrtc_state->active_planes &&
+	    (!updated_planes_and_streams || amdgpu_ip_version(dm->adev, DCE_HWIP, 0) == 0) &&
+	    acrtc_state->cursor_mode == DM_CURSOR_NATIVE_MODE)
 		amdgpu_dm_commit_cursors(state);
 
 cleanup:
@@ -8305,6 +10281,9 @@ static void amdgpu_dm_commit_audio(struct drm_device *dev,
 			continue;
 
 notify:
+		if (connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+			continue;
+
 		aconnector = to_amdgpu_dm_connector(connector);
 
 		mutex_lock(&adev->dm.audio_lock);
@@ -8337,6 +10316,9 @@ notify:
 		if (!status)
 			continue;
 
+		if (connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+			continue;
+
 		aconnector = to_amdgpu_dm_connector(connector);
 
 		mutex_lock(&adev->dm.audio_lock);
@@ -8362,55 +10344,57 @@ static void amdgpu_dm_crtc_copy_transient_flags(struct drm_crtc_state *crtc_stat
 	stream_state->mode_changed = drm_atomic_crtc_needs_modeset(crtc_state);
 }
 
-/**
- * amdgpu_dm_atomic_commit_tail() - AMDgpu DM's commit tail implementation.
- * @state: The atomic state to commit
- *
- * This will tell DC to commit the constructed DC state from atomic_check,
- * programming the hardware. Any failures here implies a hardware failure, since
- * atomic check should have filtered anything non-kosher.
- */
-static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
+static void dm_clear_writeback(struct amdgpu_display_manager *dm,
+			      struct dm_crtc_state *crtc_state)
+{
+	dc_stream_remove_writeback(dm->dc, crtc_state->stream, 0);
+}
+
+static void amdgpu_dm_commit_streams(struct drm_atomic_state *state,
+					struct dc_state *dc_state)
 {
 	struct drm_device *dev = state->dev;
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct amdgpu_display_manager *dm = &adev->dm;
-	struct dm_atomic_state *dm_state;
-	struct dc_state *dc_state = NULL, *dc_state_temp = NULL;
-	u32 i, j;
 	struct drm_crtc *crtc;
 	struct drm_crtc_state *old_crtc_state, *new_crtc_state;
-	unsigned long flags;
-	bool wait_for_vblank = true;
-	struct drm_connector *connector;
-	struct drm_connector_state *old_con_state, *new_con_state;
 	struct dm_crtc_state *dm_old_crtc_state, *dm_new_crtc_state;
-	int crtc_disable_count = 0;
+	struct drm_connector_state *old_con_state;
+	struct drm_connector *connector;
 	bool mode_set_reset_required = false;
-	int r;
+	u32 i;
+	struct dc_commit_streams_params params = {dc_state->streams, dc_state->stream_count};
+	bool set_backlight_level = false;
 
-	trace_amdgpu_dm_atomic_commit_tail_begin(state);
+	/* Disable writeback */
+	for_each_old_connector_in_state(state, connector, old_con_state, i) {
+		struct dm_connector_state *dm_old_con_state;
+		struct amdgpu_crtc *acrtc;
 
-	r = drm_atomic_helper_wait_for_fences(dev, state, false);
-	if (unlikely(r))
-		DRM_ERROR("Waiting for fences timed out!");
+		if (connector->connector_type != DRM_MODE_CONNECTOR_WRITEBACK)
+			continue;
 
-	drm_atomic_helper_update_legacy_modeset_state(dev, state);
-	drm_dp_mst_atomic_wait_for_dependencies(state);
+		old_crtc_state = NULL;
 
-	dm_state = dm_atomic_get_new_state(state);
-	if (dm_state && dm_state->context) {
-		dc_state = dm_state->context;
-	} else {
-		/* No state changes, retain current state. */
-		dc_state_temp = dc_create_state(dm->dc);
-		ASSERT(dc_state_temp);
-		dc_state = dc_state_temp;
-		dc_resource_state_copy_construct_current(dm->dc, dc_state);
+		dm_old_con_state = to_dm_connector_state(old_con_state);
+		if (!dm_old_con_state->base.crtc)
+			continue;
+
+		acrtc = to_amdgpu_crtc(dm_old_con_state->base.crtc);
+		if (acrtc)
+			old_crtc_state = drm_atomic_get_old_crtc_state(state, &acrtc->base);
+
+		if (!acrtc || !acrtc->wb_enabled)
+			continue;
+
+		dm_old_crtc_state = to_dm_crtc_state(old_crtc_state);
+
+		dm_clear_writeback(dm, dm_old_crtc_state);
+		acrtc->wb_enabled = false;
 	}
 
-	for_each_oldnew_crtc_in_state (state, crtc, old_crtc_state,
-				       new_crtc_state, i) {
+	for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state,
+				      new_crtc_state, i) {
 		struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
 
 		dm_old_crtc_state = to_dm_crtc_state(old_crtc_state);
@@ -8418,7 +10402,7 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 		if (old_crtc_state->active &&
 		    (!new_crtc_state->active ||
 		     drm_atomic_crtc_needs_modeset(new_crtc_state))) {
-			manage_dm_interrupts(adev, acrtc, false);
+			manage_dm_interrupts(adev, acrtc, NULL);
 			dc_stream_release(dm_old_crtc_state->stream);
 		}
 	}
@@ -8433,9 +10417,7 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 		dm_old_crtc_state = to_dm_crtc_state(old_crtc_state);
 
 		drm_dbg_state(state->dev,
-			"amdgpu_crtc id:%d crtc_state_flags: enable:%d, active:%d, "
-			"planes_changed:%d, mode_changed:%d,active_changed:%d,"
-			"connectors_changed:%d\n",
+			"amdgpu_crtc id:%d crtc_state_flags: enable:%d, active:%d, planes_changed:%d, mode_changed:%d,active_changed:%d,connectors_changed:%d\n",
 			acrtc->crtc_id,
 			new_crtc_state->enable,
 			new_crtc_state->active,
@@ -8450,7 +10432,8 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 
 			memset(&position, 0, sizeof(position));
 			mutex_lock(&dm->dc_lock);
-			dc_stream_set_cursor_position(dm_old_crtc_state->stream, &position);
+			dc_exit_ips_for_hw_access(dm->dc);
+			dc_stream_program_cursor_position(dm_old_crtc_state->stream, &position);
 			mutex_unlock(&dm->dc_lock);
 		}
 
@@ -8466,7 +10449,9 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 
 		if (amdgpu_dm_crtc_modeset_required(new_crtc_state, dm_new_crtc_state->stream, dm_old_crtc_state->stream)) {
 
-			DRM_DEBUG_ATOMIC("Atomic commit: SET crtc id %d: [%p]\n", acrtc->crtc_id, acrtc);
+			drm_dbg_atomic(dev,
+				       "Atomic commit: SET crtc id %d: [%p]\n",
+				       acrtc->crtc_id, acrtc);
 
 			if (!dm_new_crtc_state->stream) {
 				/*
@@ -8484,8 +10469,9 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 				 * have a sink to keep the pipe running so that
 				 * hw state is consistent with the sw state
 				 */
-				DRM_DEBUG_DRIVER("%s: Failed to create new stream for crtc %d\n",
-						__func__, acrtc->base.base.id);
+				drm_dbg_atomic(dev,
+					       "Failed to create new stream for crtc %d\n",
+						acrtc->base.base.id);
 				continue;
 			}
 
@@ -8498,8 +10484,11 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 			acrtc->hw_mode = new_crtc_state->mode;
 			crtc->hwmode = new_crtc_state->mode;
 			mode_set_reset_required = true;
+			set_backlight_level = true;
 		} else if (modereset_required(new_crtc_state)) {
-			DRM_DEBUG_ATOMIC("Atomic commit: RESET. crtc id %d:[%p]\n", acrtc->crtc_id, acrtc);
+			drm_dbg_atomic(dev,
+				       "Atomic commit: RESET. crtc id %d:[%p]\n",
+				       acrtc->crtc_id, acrtc);
 			/* i.e. reset mode */
 			if (dm_old_crtc_state->stream)
 				remove_stream(adev, acrtc, dm_old_crtc_state->stream);
@@ -8508,24 +10497,24 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 		}
 	} /* for_each_crtc_in_state() */
 
-	if (dc_state) {
-		/* if there mode set or reset, disable eDP PSR */
-		if (mode_set_reset_required) {
-			if (dm->vblank_control_workqueue)
-				flush_workqueue(dm->vblank_control_workqueue);
+	/* if there mode set or reset, disable eDP PSR, Replay */
+	if (mode_set_reset_required) {
+		if (dm->vblank_control_workqueue)
+			flush_workqueue(dm->vblank_control_workqueue);
 
-			amdgpu_dm_psr_disable_all(dm);
-		}
+		amdgpu_dm_replay_disable_all(dm);
+		amdgpu_dm_psr_disable_all(dm);
+	}
 
-		dm_enable_per_frame_crtc_master_sync(dc_state);
-		mutex_lock(&dm->dc_lock);
-		WARN_ON(!dc_commit_streams(dm->dc, dc_state->streams, dc_state->stream_count));
+	dm_enable_per_frame_crtc_master_sync(dc_state);
+	mutex_lock(&dm->dc_lock);
+	dc_exit_ips_for_hw_access(dm->dc);
+	WARN_ON(!dc_commit_streams(dm->dc, &params));
 
-		/* Allow idle optimization when vblank count is 0 for display off */
-		if (dm->active_vblank_irq_count == 0)
-			dc_allow_idle_optimizations(dm->dc, true);
-		mutex_unlock(&dm->dc_lock);
-	}
+	/* Allow idle optimization when vblank count is 0 for display off */
+	if ((dm->active_vblank_irq_count == 0) && amdgpu_dm_is_headless(dm->adev))
+		dc_allow_idle_optimizations(dm->dc, true);
+	mutex_unlock(&dm->dc_lock);
 
 	for_each_new_crtc_in_state(state, crtc, new_crtc_state, i) {
 		struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
@@ -8537,33 +10526,164 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 					dc_stream_get_status(dm_new_crtc_state->stream);
 
 			if (!status)
-				status = dc_stream_get_status_from_state(dc_state,
+				status = dc_state_get_stream_status(dc_state,
 									 dm_new_crtc_state->stream);
 			if (!status)
-				DC_ERR("got no status for stream %p on acrtc%p\n", dm_new_crtc_state->stream, acrtc);
+				drm_err(dev,
+					"got no status for stream %p on acrtc%p\n",
+					dm_new_crtc_state->stream, acrtc);
 			else
 				acrtc->otg_inst = status->primary_otg_inst;
 		}
 	}
+
+	/* During boot up and resume the DC layer will reset the panel brightness
+	 * to fix a flicker issue.
+	 * It will cause the dm->actual_brightness is not the current panel brightness
+	 * level. (the dm->brightness is the correct panel level)
+	 * So we set the backlight level with dm->brightness value after set mode
+	 */
+	if (set_backlight_level) {
+		for (i = 0; i < dm->num_of_edps; i++) {
+			if (dm->backlight_dev[i])
+				amdgpu_dm_backlight_set_level(dm, i, dm->brightness[i]);
+		}
+	}
+}
+
+static void dm_set_writeback(struct amdgpu_display_manager *dm,
+			      struct dm_crtc_state *crtc_state,
+			      struct drm_connector *connector,
+			      struct drm_connector_state *new_con_state)
+{
+	struct drm_writeback_connector *wb_conn = drm_connector_to_writeback(connector);
+	struct amdgpu_device *adev = dm->adev;
+	struct amdgpu_crtc *acrtc;
+	struct dc_writeback_info *wb_info;
+	struct pipe_ctx *pipe = NULL;
+	struct amdgpu_framebuffer *afb;
+	int i = 0;
+
+	wb_info = kzalloc(sizeof(*wb_info), GFP_KERNEL);
+	if (!wb_info) {
+		drm_err(adev_to_drm(adev), "Failed to allocate wb_info\n");
+		return;
+	}
+
+	acrtc = to_amdgpu_crtc(wb_conn->encoder.crtc);
+	if (!acrtc) {
+		drm_err(adev_to_drm(adev), "no amdgpu_crtc found\n");
+		kfree(wb_info);
+		return;
+	}
+
+	afb = to_amdgpu_framebuffer(new_con_state->writeback_job->fb);
+	if (!afb) {
+		drm_err(adev_to_drm(adev), "No amdgpu_framebuffer found\n");
+		kfree(wb_info);
+		return;
+	}
+
+	for (i = 0; i < MAX_PIPES; i++) {
+		if (dm->dc->current_state->res_ctx.pipe_ctx[i].stream == crtc_state->stream) {
+			pipe = &dm->dc->current_state->res_ctx.pipe_ctx[i];
+			break;
+		}
+	}
+
+	/* fill in wb_info */
+	wb_info->wb_enabled = true;
+
+	wb_info->dwb_pipe_inst = 0;
+	wb_info->dwb_params.dwbscl_black_color = 0;
+	wb_info->dwb_params.hdr_mult = 0x1F000;
+	wb_info->dwb_params.csc_params.gamut_adjust_type = CM_GAMUT_ADJUST_TYPE_BYPASS;
+	wb_info->dwb_params.csc_params.gamut_coef_format = CM_GAMUT_REMAP_COEF_FORMAT_S2_13;
+	wb_info->dwb_params.output_depth = DWB_OUTPUT_PIXEL_DEPTH_10BPC;
+	wb_info->dwb_params.cnv_params.cnv_out_bpc = DWB_CNV_OUT_BPC_10BPC;
+
+	/* width & height from crtc */
+	wb_info->dwb_params.cnv_params.src_width = acrtc->base.mode.crtc_hdisplay;
+	wb_info->dwb_params.cnv_params.src_height = acrtc->base.mode.crtc_vdisplay;
+	wb_info->dwb_params.dest_width = acrtc->base.mode.crtc_hdisplay;
+	wb_info->dwb_params.dest_height = acrtc->base.mode.crtc_vdisplay;
+
+	wb_info->dwb_params.cnv_params.crop_en = false;
+	wb_info->dwb_params.stereo_params.stereo_enabled = false;
+
+	wb_info->dwb_params.cnv_params.out_max_pix_val = 0x3ff;	// 10 bits
+	wb_info->dwb_params.cnv_params.out_min_pix_val = 0;
+	wb_info->dwb_params.cnv_params.fc_out_format = DWB_OUT_FORMAT_32BPP_ARGB;
+	wb_info->dwb_params.cnv_params.out_denorm_mode = DWB_OUT_DENORM_BYPASS;
+
+	wb_info->dwb_params.out_format = dwb_scaler_mode_bypass444;
+
+	wb_info->dwb_params.capture_rate = dwb_capture_rate_0;
+
+	wb_info->dwb_params.scaler_taps.h_taps = 4;
+	wb_info->dwb_params.scaler_taps.v_taps = 4;
+	wb_info->dwb_params.scaler_taps.h_taps_c = 2;
+	wb_info->dwb_params.scaler_taps.v_taps_c = 2;
+	wb_info->dwb_params.subsample_position = DWB_INTERSTITIAL_SUBSAMPLING;
+
+	wb_info->mcif_buf_params.luma_pitch = afb->base.pitches[0];
+	wb_info->mcif_buf_params.chroma_pitch = afb->base.pitches[1];
+
+	for (i = 0; i < DWB_MCIF_BUF_COUNT; i++) {
+		wb_info->mcif_buf_params.luma_address[i] = afb->address;
+		wb_info->mcif_buf_params.chroma_address[i] = 0;
+	}
+
+	wb_info->mcif_buf_params.p_vmid = 1;
+	if (amdgpu_ip_version(adev, DCE_HWIP, 0) >= IP_VERSION(3, 0, 0)) {
+		wb_info->mcif_warmup_params.start_address.quad_part = afb->address;
+		wb_info->mcif_warmup_params.region_size =
+			wb_info->mcif_buf_params.luma_pitch * wb_info->dwb_params.dest_height;
+	}
+	wb_info->mcif_warmup_params.p_vmid = 1;
+	wb_info->writeback_source_plane = pipe->plane_state;
+
+	dc_stream_add_writeback(dm->dc, crtc_state->stream, wb_info);
+
+	acrtc->wb_pending = true;
+	acrtc->wb_conn = wb_conn;
+	drm_writeback_queue_job(wb_conn, new_con_state);
+}
+
+static void amdgpu_dm_update_hdcp(struct drm_atomic_state *state)
+{
+	struct drm_connector_state *old_con_state, *new_con_state;
+	struct drm_device *dev = state->dev;
+	struct drm_connector *connector;
+	struct amdgpu_device *adev = drm_to_adev(dev);
+	int i;
+
+	if (!adev->dm.hdcp_workqueue)
+		return;
+
 	for_each_oldnew_connector_in_state(state, connector, old_con_state, new_con_state, i) {
 		struct dm_connector_state *dm_new_con_state = to_dm_connector_state(new_con_state);
 		struct amdgpu_crtc *acrtc = to_amdgpu_crtc(dm_new_con_state->base.crtc);
-		struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(connector);
-
-		pr_debug("[HDCP_DM] -------------- i : %x ----------\n", i);
+		struct drm_crtc_state *old_crtc_state, *new_crtc_state;
+		struct dm_crtc_state *dm_new_crtc_state;
+		struct amdgpu_dm_connector *aconnector;
 
-		if (!connector)
+		if (!connector || connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
 			continue;
 
-		pr_debug("[HDCP_DM] connector->index: %x connect_status: %x dpms: %x\n",
+		aconnector = to_amdgpu_dm_connector(connector);
+
+		drm_dbg(dev, "[HDCP_DM] -------------- i : %x ----------\n", i);
+
+		drm_dbg(dev, "[HDCP_DM] connector->index: %x connect_status: %x dpms: %x\n",
 			connector->index, connector->status, connector->dpms);
-		pr_debug("[HDCP_DM] state protection old: %x new: %x\n",
+		drm_dbg(dev, "[HDCP_DM] state protection old: %x new: %x\n",
 			old_con_state->content_protection, new_con_state->content_protection);
 
 		if (aconnector->dc_sink) {
 			if (aconnector->dc_sink->sink_signal != SIGNAL_TYPE_VIRTUAL &&
 				aconnector->dc_sink->sink_signal != SIGNAL_TYPE_NONE) {
-				pr_debug("[HDCP_DM] pipe_ctx dispname=%s\n",
+				drm_dbg(dev, "[HDCP_DM] pipe_ctx dispname=%s\n",
 				aconnector->dc_sink->edid_caps.display_name);
 			}
 		}
@@ -8577,7 +10697,7 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 		}
 
 		if (old_crtc_state)
-			pr_debug("old crtc en: %x a: %x m: %x a-chg: %x c-chg: %x\n",
+			drm_dbg(dev, "old crtc en: %x a: %x m: %x a-chg: %x c-chg: %x\n",
 			old_crtc_state->enable,
 			old_crtc_state->active,
 			old_crtc_state->mode_changed,
@@ -8585,26 +10705,13 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 			old_crtc_state->connectors_changed);
 
 		if (new_crtc_state)
-			pr_debug("NEW crtc en: %x a: %x m: %x a-chg: %x c-chg: %x\n",
+			drm_dbg(dev, "NEW crtc en: %x a: %x m: %x a-chg: %x c-chg: %x\n",
 			new_crtc_state->enable,
 			new_crtc_state->active,
 			new_crtc_state->mode_changed,
 			new_crtc_state->active_changed,
 			new_crtc_state->connectors_changed);
-	}
 
-	for_each_oldnew_connector_in_state(state, connector, old_con_state, new_con_state, i) {
-		struct dm_connector_state *dm_new_con_state = to_dm_connector_state(new_con_state);
-		struct amdgpu_crtc *acrtc = to_amdgpu_crtc(dm_new_con_state->base.crtc);
-		struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(connector);
-
-		new_crtc_state = NULL;
-		old_crtc_state = NULL;
-
-		if (acrtc) {
-			new_crtc_state = drm_atomic_get_new_crtc_state(state, &acrtc->base);
-			old_crtc_state = drm_atomic_get_old_crtc_state(state, &acrtc->base);
-		}
 
 		dm_new_crtc_state = to_dm_crtc_state(new_crtc_state);
 
@@ -8648,26 +10755,98 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 				new_con_state->content_protection >= DRM_MODE_CONTENT_PROTECTION_DESIRED)
 				enable_encryption = true;
 
-			DRM_INFO("[HDCP_DM] hdcp_update_display enable_encryption = %x\n", enable_encryption);
+			drm_info(dev, "[HDCP_DM] hdcp_update_display enable_encryption = %x\n", enable_encryption);
 
-			hdcp_update_display(
-				adev->dm.hdcp_workqueue, aconnector->dc_link->link_index, aconnector,
-				new_con_state->hdcp_content_type, enable_encryption);
+			if (aconnector->dc_link)
+				hdcp_update_display(
+					adev->dm.hdcp_workqueue, aconnector->dc_link->link_index, aconnector,
+					new_con_state->hdcp_content_type, enable_encryption);
+		}
+	}
+}
+
+static int amdgpu_dm_atomic_setup_commit(struct drm_atomic_state *state)
+{
+	struct drm_crtc *crtc;
+	struct drm_crtc_state *old_crtc_state, *new_crtc_state;
+	struct dm_crtc_state *dm_old_crtc_state, *dm_new_crtc_state;
+	int i, ret;
+
+	ret = drm_dp_mst_atomic_setup_commit(state);
+	if (ret)
+		return ret;
+
+	for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i) {
+		dm_old_crtc_state = to_dm_crtc_state(old_crtc_state);
+		dm_new_crtc_state = to_dm_crtc_state(new_crtc_state);
+		/*
+		 * Color management settings. We also update color properties
+		 * when a modeset is needed, to ensure it gets reprogrammed.
+		 */
+		if (dm_new_crtc_state->base.active && dm_new_crtc_state->stream &&
+		    (dm_new_crtc_state->base.color_mgmt_changed ||
+		     dm_old_crtc_state->regamma_tf != dm_new_crtc_state->regamma_tf ||
+		     drm_atomic_crtc_needs_modeset(new_crtc_state))) {
+			ret = amdgpu_dm_update_crtc_color_mgmt(dm_new_crtc_state);
+			if (ret) {
+				drm_dbg_atomic(state->dev, "Failed to update color state\n");
+				return ret;
+			}
 		}
 	}
 
+	return 0;
+}
+
+/**
+ * amdgpu_dm_atomic_commit_tail() - AMDgpu DM's commit tail implementation.
+ * @state: The atomic state to commit
+ *
+ * This will tell DC to commit the constructed DC state from atomic_check,
+ * programming the hardware. Any failures here implies a hardware failure, since
+ * atomic check should have filtered anything non-kosher.
+ */
+static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
+{
+	struct drm_device *dev = state->dev;
+	struct amdgpu_device *adev = drm_to_adev(dev);
+	struct amdgpu_display_manager *dm = &adev->dm;
+	struct dm_atomic_state *dm_state;
+	struct dc_state *dc_state = NULL;
+	u32 i, j;
+	struct drm_crtc *crtc;
+	struct drm_crtc_state *old_crtc_state, *new_crtc_state;
+	unsigned long flags;
+	bool wait_for_vblank = true;
+	struct drm_connector *connector;
+	struct drm_connector_state *old_con_state = NULL, *new_con_state = NULL;
+	struct dm_crtc_state *dm_old_crtc_state, *dm_new_crtc_state;
+	int crtc_disable_count = 0;
+
+	trace_amdgpu_dm_atomic_commit_tail_begin(state);
+
+	drm_atomic_helper_update_legacy_modeset_state(dev, state);
+	drm_dp_mst_atomic_wait_for_dependencies(state);
+
+	dm_state = dm_atomic_get_new_state(state);
+	if (dm_state && dm_state->context) {
+		dc_state = dm_state->context;
+		amdgpu_dm_commit_streams(state, dc_state);
+	}
+
+	amdgpu_dm_update_hdcp(state);
+
 	/* Handle connector state changes */
 	for_each_oldnew_connector_in_state(state, connector, old_con_state, new_con_state, i) {
 		struct dm_connector_state *dm_new_con_state = to_dm_connector_state(new_con_state);
 		struct dm_connector_state *dm_old_con_state = to_dm_connector_state(old_con_state);
 		struct amdgpu_crtc *acrtc = to_amdgpu_crtc(dm_new_con_state->base.crtc);
-		struct dc_surface_update dummy_updates[MAX_SURFACES];
+		struct dc_surface_update *dummy_updates;
 		struct dc_stream_update stream_update;
 		struct dc_info_packet hdr_packet;
 		struct dc_stream_status *status = NULL;
-		bool abm_changed, hdr_changed, scaling_changed;
+		bool abm_changed, hdr_changed, scaling_changed, output_color_space_changed = false;
 
-		memset(&dummy_updates, 0, sizeof(dummy_updates));
 		memset(&stream_update, 0, sizeof(stream_update));
 
 		if (acrtc) {
@@ -8685,13 +10864,18 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 		scaling_changed = is_scaling_state_different(dm_new_con_state,
 							     dm_old_con_state);
 
+		if ((new_con_state->hdmi.broadcast_rgb != old_con_state->hdmi.broadcast_rgb) &&
+			(dm_old_crtc_state->stream->output_color_space !=
+				get_output_color_space(&dm_new_crtc_state->stream->timing, new_con_state)))
+			output_color_space_changed = true;
+
 		abm_changed = dm_new_crtc_state->abm_level !=
 			      dm_old_crtc_state->abm_level;
 
 		hdr_changed =
 			!drm_connector_atomic_hdr_metadata_equal(old_con_state, new_con_state);
 
-		if (!scaling_changed && !abm_changed && !hdr_changed)
+		if (!scaling_changed && !abm_changed && !hdr_changed && !output_color_space_changed)
 			continue;
 
 		stream_update.stream = dm_new_crtc_state->stream;
@@ -8703,6 +10887,13 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 			stream_update.dst = dm_new_crtc_state->stream->dst;
 		}
 
+		if (output_color_space_changed) {
+			dm_new_crtc_state->stream->output_color_space
+				= get_output_color_space(&dm_new_crtc_state->stream->timing, new_con_state);
+
+			stream_update.output_color_space = &dm_new_crtc_state->stream->output_color_space;
+		}
+
 		if (abm_changed) {
 			dm_new_crtc_state->stream->abm_level = dm_new_crtc_state->abm_level;
 
@@ -8726,17 +10917,28 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 		 * Here we create an empty update on each plane.
 		 * To fix this, DC should permit updating only stream properties.
 		 */
+		dummy_updates = kzalloc(sizeof(struct dc_surface_update) * MAX_SURFACES, GFP_KERNEL);
+		if (!dummy_updates) {
+			drm_err(adev_to_drm(adev), "Failed to allocate memory for dummy_updates.\n");
+			continue;
+		}
 		for (j = 0; j < status->plane_count; j++)
 			dummy_updates[j].surface = status->plane_states[0];
 
+		sort(dummy_updates, status->plane_count,
+		     sizeof(*dummy_updates), dm_plane_layer_index_cmp, NULL);
 
 		mutex_lock(&dm->dc_lock);
+		dc_exit_ips_for_hw_access(dm->dc);
 		dc_update_planes_and_stream(dm->dc,
 					    dummy_updates,
 					    status->plane_count,
 					    dm_new_crtc_state->stream,
 					    &stream_update);
 		mutex_unlock(&dm->dc_lock);
+		kfree(dummy_updates);
+
+		drm_connector_update_privacy_screen(new_con_state);
 	}
 
 	/**
@@ -8771,7 +10973,7 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 		     drm_atomic_crtc_needs_modeset(new_crtc_state))) {
 			dc_stream_retain(dm_new_crtc_state->stream);
 			acrtc->dm_irq_params.stream = dm_new_crtc_state->stream;
-			manage_dm_interrupts(adev, acrtc, true);
+			manage_dm_interrupts(adev, acrtc, dm_new_crtc_state);
 		}
 		/* Handle vrr on->off / off->on transitions */
 		amdgpu_dm_handle_vrr_transition(dm_old_crtc_state, dm_new_crtc_state);
@@ -8787,20 +10989,26 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 			if (amdgpu_dm_is_valid_crc_source(cur_crc_src)) {
 #if defined(CONFIG_DRM_AMD_SECURE_DISPLAY)
 				if (amdgpu_dm_crc_window_is_activated(crtc)) {
-					spin_lock_irqsave(&adev_to_drm(adev)->event_lock, flags);
-					acrtc->dm_irq_params.window_param.update_win = true;
+					uint8_t cnt;
 
-					/**
-					 * It takes 2 frames for HW to stably generate CRC when
-					 * resuming from suspend, so we set skip_frame_cnt 2.
-					 */
-					acrtc->dm_irq_params.window_param.skip_frame_cnt = 2;
+					spin_lock_irqsave(&adev_to_drm(adev)->event_lock, flags);
+					for (cnt = 0; cnt < MAX_CRC_WINDOW_NUM; cnt++) {
+						if (acrtc->dm_irq_params.window_param[cnt].enable) {
+							acrtc->dm_irq_params.window_param[cnt].update_win = true;
+
+							/**
+							 * It takes 2 frames for HW to stably generate CRC when
+							 * resuming from suspend, so we set skip_frame_cnt 2.
+							 */
+							acrtc->dm_irq_params.window_param[cnt].skip_frame_cnt = 2;
+						}
+					}
 					spin_unlock_irqrestore(&adev_to_drm(adev)->event_lock, flags);
 				}
 #endif
 				if (amdgpu_dm_crtc_configure_crc_source(
 					crtc, dm_new_crtc_state, cur_crc_src))
-					DRM_DEBUG_DRIVER("Failed to configure crc source");
+					drm_dbg_atomic(dev, "Failed to configure crc source");
 			}
 		}
 #endif
@@ -8815,8 +11023,32 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 		dm_new_crtc_state = to_dm_crtc_state(new_crtc_state);
 
 		if (dm_new_crtc_state->stream)
-			amdgpu_dm_commit_planes(state, dc_state, dev,
-						dm, crtc, wait_for_vblank);
+			amdgpu_dm_commit_planes(state, dev, dm, crtc, wait_for_vblank);
+	}
+
+	/* Enable writeback */
+	for_each_new_connector_in_state(state, connector, new_con_state, i) {
+		struct dm_connector_state *dm_new_con_state = to_dm_connector_state(new_con_state);
+		struct amdgpu_crtc *acrtc = to_amdgpu_crtc(dm_new_con_state->base.crtc);
+
+		if (connector->connector_type != DRM_MODE_CONNECTOR_WRITEBACK)
+			continue;
+
+		if (!new_con_state->writeback_job)
+			continue;
+
+		new_crtc_state = drm_atomic_get_new_crtc_state(state, &acrtc->base);
+
+		if (!new_crtc_state)
+			continue;
+
+		if (acrtc->wb_enabled)
+			continue;
+
+		dm_new_crtc_state = to_dm_crtc_state(new_crtc_state);
+
+		dm_set_writeback(dm, dm_new_crtc_state, connector, new_con_state);
+		acrtc->wb_enabled = true;
 	}
 
 	/* Update audio instances for each connector. */
@@ -8851,10 +11083,17 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 
 	drm_atomic_helper_cleanup_planes(dev, state);
 
-	/* return the stolen vga memory back to VRAM */
-	if (!adev->mman.keep_stolen_vga_memory)
-		amdgpu_bo_free_kernel(&adev->mman.stolen_vga_memory, NULL, NULL);
-	amdgpu_bo_free_kernel(&adev->mman.stolen_extended_memory, NULL, NULL);
+	/* Don't free the memory if we are hitting this as part of suspend.
+	 * This way we don't free any memory during suspend; see
+	 * amdgpu_bo_free_kernel().  The memory will be freed in the first
+	 * non-suspend modeset or when the driver is torn down.
+	 */
+	if (!adev->in_suspend) {
+		/* return the stolen vga memory back to VRAM */
+		if (!adev->mman.keep_stolen_vga_memory)
+			amdgpu_bo_free_kernel(&adev->mman.stolen_vga_memory, NULL, NULL);
+		amdgpu_bo_free_kernel(&adev->mman.stolen_extended_memory, NULL, NULL);
+	}
 
 	/*
 	 * Finally, drop a runtime PM reference for each newly disabled CRTC,
@@ -8865,8 +11104,7 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state)
 		pm_runtime_put_autosuspend(dev->dev);
 	pm_runtime_mark_last_busy(dev->dev);
 
-	if (dc_state_temp)
-		dc_release_state(dc_state_temp);
+	trace_amdgpu_dm_atomic_commit_tail_finish(state);
 }
 
 static int dm_force_atomic_commit(struct drm_connector *connector)
@@ -8892,16 +11130,20 @@ static int dm_force_atomic_commit(struct drm_connector *connector)
 	 */
 	conn_state = drm_atomic_get_connector_state(state, connector);
 
-	ret = PTR_ERR_OR_ZERO(conn_state);
-	if (ret)
+	/* Check for error in getting connector state */
+	if (IS_ERR(conn_state)) {
+		ret = PTR_ERR(conn_state);
 		goto out;
+	}
 
 	/* Attach crtc to drm_atomic_state*/
 	crtc_state = drm_atomic_get_crtc_state(state, &disconnected_acrtc->base);
 
-	ret = PTR_ERR_OR_ZERO(crtc_state);
-	if (ret)
+	/* Check for error in getting crtc state */
+	if (IS_ERR(crtc_state)) {
+		ret = PTR_ERR(crtc_state);
 		goto out;
+	}
 
 	/* force a restore */
 	crtc_state->mode_changed = true;
@@ -8909,9 +11151,11 @@ static int dm_force_atomic_commit(struct drm_connector *connector)
 	/* Attach plane to drm_atomic_state */
 	plane_state = drm_atomic_get_plane_state(state, plane);
 
-	ret = PTR_ERR_OR_ZERO(plane_state);
-	if (ret)
+	/* Check for error in getting plane state */
+	if (IS_ERR(plane_state)) {
+		ret = PTR_ERR(plane_state);
 		goto out;
+	}
 
 	/* Call commit internally with the state we just constructed */
 	ret = drm_atomic_commit(state);
@@ -8919,7 +11163,7 @@ static int dm_force_atomic_commit(struct drm_connector *connector)
 out:
 	drm_atomic_state_put(state);
 	if (ret)
-		DRM_ERROR("Restoring old state failed with %i\n", ret);
+		drm_err(ddev, "Restoring old state failed with %i\n", ret);
 
 	return ret;
 }
@@ -8932,10 +11176,15 @@ out:
 void dm_restore_drm_connector_state(struct drm_device *dev,
 				    struct drm_connector *connector)
 {
-	struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(connector);
+	struct amdgpu_dm_connector *aconnector;
 	struct amdgpu_crtc *disconnected_acrtc;
 	struct dm_crtc_state *acrtc_state;
 
+	if (connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+		return;
+
+	aconnector = to_amdgpu_dm_connector(connector);
+
 	if (!aconnector->dc_sink || !connector->state || !connector->encoder)
 		return;
 
@@ -8998,8 +11247,8 @@ static int do_aquire_global_lock(struct drm_device *dev,
 					&commit->flip_done, 10*HZ);
 
 		if (ret == 0)
-			DRM_ERROR("[CRTC:%d:%s] hw_done or flip_done "
-				  "timed out\n", crtc->base.id, crtc->name);
+			drm_err(dev, "[CRTC:%d:%s] hw_done or flip_done timed out\n",
+				  crtc->base.id, crtc->name);
 
 		drm_crtc_commit_put(commit);
 	}
@@ -9012,12 +11261,16 @@ static void get_freesync_config_for_crtc(
 	struct dm_connector_state *new_con_state)
 {
 	struct mod_freesync_config config = {0};
-	struct amdgpu_dm_connector *aconnector =
-			to_amdgpu_dm_connector(new_con_state->base.connector);
+	struct amdgpu_dm_connector *aconnector;
 	struct drm_display_mode *mode = &new_crtc_state->base.mode;
 	int vrefresh = drm_mode_vrefresh(mode);
 	bool fs_vid_mode = false;
 
+	if (new_con_state->base.connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+		return;
+
+	aconnector = to_amdgpu_dm_connector(new_con_state->base.connector);
+
 	new_crtc_state->vrr_supported = new_con_state->freesync_capable &&
 					vrefresh >= aconnector->min_vfreq &&
 					vrefresh <= aconnector->max_vfreq;
@@ -9040,6 +11293,8 @@ static void get_freesync_config_for_crtc(
 		} else {
 			config.state = VRR_STATE_INACTIVE;
 		}
+	} else {
+		config.state = VRR_STATE_UNSUPPORTED;
 	}
 out:
 	new_crtc_state->freesync_config = config;
@@ -9084,7 +11339,8 @@ is_timing_unchanged_for_freesync(struct drm_crtc_state *old_crtc_state,
 	return false;
 }
 
-static void set_freesync_fixed_config(struct dm_crtc_state *dm_new_crtc_state) {
+static void set_freesync_fixed_config(struct dm_crtc_state *dm_new_crtc_state)
+{
 	u64 num, den, res;
 	struct drm_crtc_state *new_crtc_state = &dm_new_crtc_state->base;
 
@@ -9109,6 +11365,7 @@ static int dm_update_crtc_state(struct amdgpu_display_manager *dm,
 	struct dm_atomic_state *dm_state = NULL;
 	struct dm_crtc_state *dm_old_crtc_state, *dm_new_crtc_state;
 	struct dc_stream_state *new_stream;
+	struct amdgpu_device *adev = dm->adev;
 	int ret = 0;
 
 	/*
@@ -9116,6 +11373,7 @@ static int dm_update_crtc_state(struct amdgpu_display_manager *dm,
 	 * update changed items
 	 */
 	struct amdgpu_crtc *acrtc = NULL;
+	struct drm_connector *connector = NULL;
 	struct amdgpu_dm_connector *aconnector = NULL;
 	struct drm_connector_state *drm_new_conn_state = NULL, *drm_old_conn_state = NULL;
 	struct dm_connector_state *dm_new_conn_state = NULL, *dm_old_conn_state = NULL;
@@ -9125,18 +11383,20 @@ static int dm_update_crtc_state(struct amdgpu_display_manager *dm,
 	dm_old_crtc_state = to_dm_crtc_state(old_crtc_state);
 	dm_new_crtc_state = to_dm_crtc_state(new_crtc_state);
 	acrtc = to_amdgpu_crtc(crtc);
-	aconnector = amdgpu_dm_find_first_crtc_matching_connector(state, crtc);
+	connector = amdgpu_dm_find_first_crtc_matching_connector(state, crtc);
+	if (connector)
+		aconnector = to_amdgpu_dm_connector(connector);
 
 	/* TODO This hack should go away */
-	if (aconnector && enable) {
+	if (connector && enable) {
 		/* Make sure fake sink is created in plug-in scenario */
 		drm_new_conn_state = drm_atomic_get_new_connector_state(state,
-							    &aconnector->base);
+									connector);
 		drm_old_conn_state = drm_atomic_get_old_connector_state(state,
-							    &aconnector->base);
+									connector);
 
-		if (IS_ERR(drm_new_conn_state)) {
-			ret = PTR_ERR_OR_ZERO(drm_new_conn_state);
+		if (WARN_ON(!drm_new_conn_state)) {
+			ret = -EINVAL;
 			goto fail;
 		}
 
@@ -9146,7 +11406,7 @@ static int dm_update_crtc_state(struct amdgpu_display_manager *dm,
 		if (!drm_atomic_crtc_needs_modeset(new_crtc_state))
 			goto skip_modeset;
 
-		new_stream = create_validate_stream_for_sink(aconnector,
+		new_stream = create_validate_stream_for_sink(connector,
 							     &new_crtc_state->mode,
 							     dm_new_conn_state,
 							     dm_old_crtc_state->stream);
@@ -9159,7 +11419,7 @@ static int dm_update_crtc_state(struct amdgpu_display_manager *dm,
 		 */
 
 		if (!new_stream) {
-			DRM_DEBUG_DRIVER("%s: Failed to create new stream for crtc %d\n",
+			drm_dbg_driver(adev_to_drm(adev), "%s: Failed to create new stream for crtc %d\n",
 					__func__, acrtc->base.base.id);
 			ret = -ENOMEM;
 			goto fail;
@@ -9197,7 +11457,7 @@ static int dm_update_crtc_state(struct amdgpu_display_manager *dm,
 		    dc_is_stream_unchanged(new_stream, dm_old_crtc_state->stream) &&
 		    dc_is_stream_scaling_unchanged(new_stream, dm_old_crtc_state->stream)) {
 			new_crtc_state->mode_changed = false;
-			DRM_DEBUG_DRIVER("Mode change not required, setting mode_changed to %d",
+			drm_dbg_driver(adev_to_drm(adev), "Mode change not required, setting mode_changed to %d",
 					 new_crtc_state->mode_changed);
 		}
 	}
@@ -9207,9 +11467,7 @@ static int dm_update_crtc_state(struct amdgpu_display_manager *dm,
 		goto skip_modeset;
 
 	drm_dbg_state(state->dev,
-		"amdgpu_crtc id:%d crtc_state_flags: enable:%d, active:%d, "
-		"planes_changed:%d, mode_changed:%d,active_changed:%d,"
-		"connectors_changed:%d\n",
+		"amdgpu_crtc id:%d crtc_state_flags: enable:%d, active:%d, planes_changed:%d, mode_changed:%d,active_changed:%d,connectors_changed:%d\n",
 		acrtc->crtc_id,
 		new_crtc_state->enable,
 		new_crtc_state->active,
@@ -9232,12 +11490,13 @@ static int dm_update_crtc_state(struct amdgpu_display_manager *dm,
 
 		/* Now check if we should set freesync video mode */
 		if (amdgpu_freesync_vid_mode && dm_new_crtc_state->stream &&
+		    dc_is_stream_unchanged(new_stream, dm_old_crtc_state->stream) &&
+		    dc_is_stream_scaling_unchanged(new_stream, dm_old_crtc_state->stream) &&
 		    is_timing_unchanged_for_freesync(new_crtc_state,
 						     old_crtc_state)) {
 			new_crtc_state->mode_changed = false;
-			DRM_DEBUG_DRIVER(
-				"Mode change not required for front porch change, "
-				"setting mode_changed to %d",
+			drm_dbg_driver(adev_to_drm(adev),
+				"Mode change not required for front porch change, setting mode_changed to %d",
 				new_crtc_state->mode_changed);
 
 			set_freesync_fixed_config(dm_new_crtc_state);
@@ -9249,20 +11508,19 @@ static int dm_update_crtc_state(struct amdgpu_display_manager *dm,
 			struct drm_display_mode *high_mode;
 
 			high_mode = get_highest_refresh_rate_mode(aconnector, false);
-			if (!drm_mode_equal(&new_crtc_state->mode, high_mode)) {
+			if (!drm_mode_equal(&new_crtc_state->mode, high_mode))
 				set_freesync_fixed_config(dm_new_crtc_state);
-			}
 		}
 
 		ret = dm_atomic_get_state(state, &dm_state);
 		if (ret)
 			goto fail;
 
-		DRM_DEBUG_DRIVER("Disabling DRM crtc: %d\n",
+		drm_dbg_driver(adev_to_drm(adev), "Disabling DRM crtc: %d\n",
 				crtc->base.id);
 
 		/* i.e. reset mode */
-		if (dc_remove_stream_from_ctx(
+		if (dc_state_remove_stream(
 				dm->dc,
 				dm_state->context,
 				dm_old_crtc_state->stream) != DC_OK) {
@@ -9283,7 +11541,7 @@ static int dm_update_crtc_state(struct amdgpu_display_manager *dm,
 		 * added MST connectors not found in existing crtc_state in the chained mode
 		 * TODO: need to dig out the root cause of that
 		 */
-		if (!aconnector)
+		if (!connector)
 			goto skip_modeset;
 
 		if (modereset_required(new_crtc_state))
@@ -9305,7 +11563,7 @@ static int dm_update_crtc_state(struct amdgpu_display_manager *dm,
 			DRM_DEBUG_ATOMIC("Enabling DRM crtc: %d\n",
 					 crtc->base.id);
 
-			if (dc_add_stream_to_ctx(
+			if (dc_state_add_stream(
 					dm->dc,
 					dm_state->context,
 					dm_new_crtc_state->stream) != DC_OK) {
@@ -9326,7 +11584,7 @@ skip_modeset:
 	 * We want to do dc stream updates that do not require a
 	 * full modeset below.
 	 */
-	if (!(enable && aconnector && new_crtc_state->active))
+	if (!(enable && connector && new_crtc_state->active))
 		return 0;
 	/*
 	 * Given above conditions, the dc state cannot be NULL because:
@@ -9352,8 +11610,9 @@ skip_modeset:
 	 * when a modeset is needed, to ensure it gets reprogrammed.
 	 */
 	if (dm_new_crtc_state->base.color_mgmt_changed ||
+	    dm_old_crtc_state->regamma_tf != dm_new_crtc_state->regamma_tf ||
 	    drm_atomic_crtc_needs_modeset(new_crtc_state)) {
-		ret = amdgpu_dm_update_crtc_color_mgmt(dm_new_crtc_state);
+		ret = amdgpu_dm_check_crtc_color_mgmt(dm_new_crtc_state, true);
 		if (ret)
 			goto fail;
 	}
@@ -9377,15 +11636,20 @@ static bool should_reset_plane(struct drm_atomic_state *state,
 {
 	struct drm_plane *other;
 	struct drm_plane_state *old_other_state, *new_other_state;
-	struct drm_crtc_state *new_crtc_state;
+	struct drm_crtc_state *old_crtc_state, *new_crtc_state;
+	struct dm_crtc_state *old_dm_crtc_state, *new_dm_crtc_state;
+	struct amdgpu_device *adev = drm_to_adev(plane->dev);
 	int i;
 
 	/*
-	 * TODO: Remove this hack once the checks below are sufficient
-	 * enough to determine when we need to reset all the planes on
-	 * the stream.
+	 * TODO: Remove this hack for all asics once it proves that the
+	 * fast updates works fine on DCN3.2+.
 	 */
-	if (state->allow_modeset)
+	if (amdgpu_ip_version(adev, DCE_HWIP, 0) < IP_VERSION(3, 2, 0) &&
+	    state->allow_modeset)
+		return true;
+
+	if (amdgpu_in_reset(adev) && state->allow_modeset)
 		return true;
 
 	/* Exit early if we know that we're adding or removing the plane. */
@@ -9398,14 +11662,38 @@ static bool should_reset_plane(struct drm_atomic_state *state,
 
 	new_crtc_state =
 		drm_atomic_get_new_crtc_state(state, new_plane_state->crtc);
+	old_crtc_state =
+		drm_atomic_get_old_crtc_state(state, old_plane_state->crtc);
 
 	if (!new_crtc_state)
 		return true;
 
+	/*
+	 * A change in cursor mode means a new dc pipe needs to be acquired or
+	 * released from the state
+	 */
+	old_dm_crtc_state = to_dm_crtc_state(old_crtc_state);
+	new_dm_crtc_state = to_dm_crtc_state(new_crtc_state);
+	if (plane->type == DRM_PLANE_TYPE_CURSOR &&
+	    old_dm_crtc_state != NULL &&
+	    old_dm_crtc_state->cursor_mode != new_dm_crtc_state->cursor_mode) {
+		return true;
+	}
+
 	/* CRTC Degamma changes currently require us to recreate planes. */
 	if (new_crtc_state->color_mgmt_changed)
 		return true;
 
+	/*
+	 * On zpos change, planes need to be reordered by removing and re-adding
+	 * them one by one to the dc state, in order of descending zpos.
+	 *
+	 * TODO: We can likely skip bandwidth validation if the only thing that
+	 * changed about the plane was it'z z-ordering.
+	 */
+	if (old_plane_state->normalized_zpos != new_plane_state->normalized_zpos)
+		return true;
+
 	if (drm_atomic_crtc_needs_modeset(new_crtc_state))
 		return true;
 
@@ -9419,6 +11707,11 @@ static bool should_reset_plane(struct drm_atomic_state *state,
 	 */
 	for_each_oldnew_plane_in_state(state, other, old_other_state, new_other_state, i) {
 		struct amdgpu_framebuffer *old_afb, *new_afb;
+		struct dm_plane_state *dm_new_other_state, *dm_old_other_state;
+
+		dm_new_other_state = to_dm_plane_state(new_other_state);
+		dm_old_other_state = to_dm_plane_state(old_other_state);
+
 		if (other->type == DRM_PLANE_TYPE_CURSOR)
 			continue;
 
@@ -9454,6 +11747,18 @@ static bool should_reset_plane(struct drm_atomic_state *state,
 		    old_other_state->color_encoding != new_other_state->color_encoding)
 			return true;
 
+		/* HDR/Transfer Function changes. */
+		if (dm_old_other_state->degamma_tf != dm_new_other_state->degamma_tf ||
+		    dm_old_other_state->degamma_lut != dm_new_other_state->degamma_lut ||
+		    dm_old_other_state->hdr_mult != dm_new_other_state->hdr_mult ||
+		    dm_old_other_state->ctm != dm_new_other_state->ctm ||
+		    dm_old_other_state->shaper_lut != dm_new_other_state->shaper_lut ||
+		    dm_old_other_state->shaper_tf != dm_new_other_state->shaper_tf ||
+		    dm_old_other_state->lut3d != dm_new_other_state->lut3d ||
+		    dm_old_other_state->blend_lut != dm_new_other_state->blend_lut ||
+		    dm_old_other_state->blend_tf != dm_new_other_state->blend_tf)
+			return true;
+
 		/* Framebuffer checks fall at the end. */
 		if (!old_other_state->fb || !new_other_state->fb)
 			continue;
@@ -9517,14 +11822,17 @@ static int dm_check_cursor_fb(struct amdgpu_crtc *new_acrtc,
 	}
 
 	/* Core DRM takes care of checking FB modifiers, so we only need to
-	 * check tiling flags when the FB doesn't have a modifier. */
+	 * check tiling flags when the FB doesn't have a modifier.
+	 */
 	if (!(fb->flags & DRM_MODE_FB_MODIFIERS)) {
-		if (adev->family < AMDGPU_FAMILY_AI) {
+		if (adev->family >= AMDGPU_FAMILY_GC_12_0_0) {
+			linear = AMDGPU_TILING_GET(afb->tiling_flags, GFX12_SWIZZLE_MODE) == 0;
+		} else if (adev->family >= AMDGPU_FAMILY_AI) {
+			linear = AMDGPU_TILING_GET(afb->tiling_flags, SWIZZLE_MODE) == 0;
+		} else {
 			linear = AMDGPU_TILING_GET(afb->tiling_flags, ARRAY_MODE) != DC_ARRAY_2D_TILED_THIN1 &&
-			         AMDGPU_TILING_GET(afb->tiling_flags, ARRAY_MODE) != DC_ARRAY_1D_TILED_THIN1 &&
+				 AMDGPU_TILING_GET(afb->tiling_flags, ARRAY_MODE) != DC_ARRAY_1D_TILED_THIN1 &&
 				 AMDGPU_TILING_GET(afb->tiling_flags, MICRO_TILE_MODE) == 0;
-		} else {
-			linear = AMDGPU_TILING_GET(afb->tiling_flags, SWIZZLE_MODE) == 0;
 		}
 		if (!linear) {
 			DRM_DEBUG_ATOMIC("Cursor FB not linear");
@@ -9535,6 +11843,68 @@ static int dm_check_cursor_fb(struct amdgpu_crtc *new_acrtc,
 	return 0;
 }
 
+/*
+ * Helper function for checking the cursor in native mode
+ */
+static int dm_check_native_cursor_state(struct drm_crtc *new_plane_crtc,
+					struct drm_plane *plane,
+					struct drm_plane_state *new_plane_state,
+					bool enable)
+{
+
+	struct amdgpu_crtc *new_acrtc;
+	int ret;
+
+	if (!enable || !new_plane_crtc ||
+	    drm_atomic_plane_disabling(plane->state, new_plane_state))
+		return 0;
+
+	new_acrtc = to_amdgpu_crtc(new_plane_crtc);
+
+	if (new_plane_state->src_x != 0 || new_plane_state->src_y != 0) {
+		DRM_DEBUG_ATOMIC("Cropping not supported for cursor plane\n");
+		return -EINVAL;
+	}
+
+	if (new_plane_state->fb) {
+		ret = dm_check_cursor_fb(new_acrtc, new_plane_state,
+						new_plane_state->fb);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+static bool dm_should_update_native_cursor(struct drm_atomic_state *state,
+					   struct drm_crtc *old_plane_crtc,
+					   struct drm_crtc *new_plane_crtc,
+					   bool enable)
+{
+	struct drm_crtc_state *old_crtc_state, *new_crtc_state;
+	struct dm_crtc_state *dm_old_crtc_state, *dm_new_crtc_state;
+
+	if (!enable) {
+		if (old_plane_crtc == NULL)
+			return true;
+
+		old_crtc_state = drm_atomic_get_old_crtc_state(
+			state, old_plane_crtc);
+		dm_old_crtc_state = to_dm_crtc_state(old_crtc_state);
+
+		return dm_old_crtc_state->cursor_mode == DM_CURSOR_NATIVE_MODE;
+	} else {
+		if (new_plane_crtc == NULL)
+			return true;
+
+		new_crtc_state = drm_atomic_get_new_crtc_state(
+			state, new_plane_crtc);
+		dm_new_crtc_state = to_dm_crtc_state(new_crtc_state);
+
+		return dm_new_crtc_state->cursor_mode == DM_CURSOR_NATIVE_MODE;
+	}
+}
+
 static int dm_update_plane_state(struct dc *dc,
 				 struct drm_atomic_state *state,
 				 struct drm_plane *plane,
@@ -9550,8 +11920,7 @@ static int dm_update_plane_state(struct dc *dc,
 	struct drm_crtc_state *old_crtc_state, *new_crtc_state;
 	struct dm_crtc_state *dm_new_crtc_state, *dm_old_crtc_state;
 	struct dm_plane_state *dm_new_plane_state, *dm_old_plane_state;
-	struct amdgpu_crtc *new_acrtc;
-	bool needs_reset;
+	bool needs_reset, update_native_cursor;
 	int ret = 0;
 
 
@@ -9560,24 +11929,16 @@ static int dm_update_plane_state(struct dc *dc,
 	dm_new_plane_state = to_dm_plane_state(new_plane_state);
 	dm_old_plane_state = to_dm_plane_state(old_plane_state);
 
-	if (plane->type == DRM_PLANE_TYPE_CURSOR) {
-		if (!enable || !new_plane_crtc ||
-			drm_atomic_plane_disabling(plane->state, new_plane_state))
-			return 0;
-
-		new_acrtc = to_amdgpu_crtc(new_plane_crtc);
+	update_native_cursor = dm_should_update_native_cursor(state,
+							      old_plane_crtc,
+							      new_plane_crtc,
+							      enable);
 
-		if (new_plane_state->src_x != 0 || new_plane_state->src_y != 0) {
-			DRM_DEBUG_ATOMIC("Cropping not supported for cursor plane\n");
-			return -EINVAL;
-		}
-
-		if (new_plane_state->fb) {
-			ret = dm_check_cursor_fb(new_acrtc, new_plane_state,
-						 new_plane_state->fb);
-			if (ret)
-				return ret;
-		}
+	if (plane->type == DRM_PLANE_TYPE_CURSOR && update_native_cursor) {
+		ret = dm_check_native_cursor_state(new_plane_crtc, plane,
+						    new_plane_state, enable);
+		if (ret)
+			return ret;
 
 		return 0;
 	}
@@ -9607,7 +11968,7 @@ static int dm_update_plane_state(struct dc *dc,
 		if (ret)
 			return ret;
 
-		if (!dc_remove_plane_from_context(
+		if (!dc_state_remove_plane(
 				dc,
 				dm_old_crtc_state->stream,
 				dm_old_plane_state->dc_state,
@@ -9616,8 +11977,9 @@ static int dm_update_plane_state(struct dc *dc,
 			return -EINVAL;
 		}
 
+		if (dm_old_plane_state->dc_state)
+			dc_plane_state_release(dm_old_plane_state->dc_state);
 
-		dc_plane_state_release(dm_old_plane_state->dc_state);
 		dm_new_plane_state->dc_state = NULL;
 
 		*lock_and_validation_needed = true;
@@ -9642,20 +12004,14 @@ static int dm_update_plane_state(struct dc *dc,
 
 		ret = amdgpu_dm_plane_helper_check_state(new_plane_state, new_crtc_state);
 		if (ret)
-			return ret;
+			goto out;
 
 		WARN_ON(dm_new_plane_state->dc_state);
 
 		dc_new_plane_state = dc_create_plane_state(dc);
-		if (!dc_new_plane_state)
-			return -ENOMEM;
-
-		/* Block top most plane from being a video plane */
-		if (plane->type == DRM_PLANE_TYPE_OVERLAY) {
-			if (is_video_format(new_plane_state->fb->format->format) && *is_top_most_overlay)
-				return -EINVAL;
-			else
-				*is_top_most_overlay = false;
+		if (!dc_new_plane_state) {
+			ret = -ENOMEM;
+			goto out;
 		}
 
 		DRM_DEBUG_ATOMIC("Enabling DRM plane: %d on DRM crtc %d\n",
@@ -9668,13 +12024,13 @@ static int dm_update_plane_state(struct dc *dc,
 			new_crtc_state);
 		if (ret) {
 			dc_plane_state_release(dc_new_plane_state);
-			return ret;
+			goto out;
 		}
 
 		ret = dm_atomic_get_state(state, &dm_state);
 		if (ret) {
 			dc_plane_state_release(dc_new_plane_state);
-			return ret;
+			goto out;
 		}
 
 		/*
@@ -9684,14 +12040,15 @@ static int dm_update_plane_state(struct dc *dc,
 		 * state. It'll be released when the atomic state is
 		 * cleaned.
 		 */
-		if (!dc_add_plane_to_context(
+		if (!dc_state_add_plane(
 				dc,
 				dm_new_crtc_state->stream,
 				dc_new_plane_state,
 				dm_state->context)) {
 
 			dc_plane_state_release(dc_new_plane_state);
-			return -EINVAL;
+			ret = -EINVAL;
+			goto out;
 		}
 
 		dm_new_plane_state->dc_state = dc_new_plane_state;
@@ -9706,6 +12063,16 @@ static int dm_update_plane_state(struct dc *dc,
 		*lock_and_validation_needed = true;
 	}
 
+out:
+	/* If enabling cursor overlay failed, attempt fallback to native mode */
+	if (enable && ret == -EINVAL && plane->type == DRM_PLANE_TYPE_CURSOR) {
+		ret = dm_check_native_cursor_state(new_plane_crtc, plane,
+						    new_plane_state, enable);
+		if (ret)
+			return ret;
+
+		dm_new_crtc_state->cursor_mode = DM_CURSOR_NATIVE_MODE;
+	}
 
 	return ret;
 }
@@ -9728,70 +12095,82 @@ static void dm_get_oriented_plane_size(struct drm_plane_state *plane_state,
 	}
 }
 
-static int dm_check_crtc_cursor(struct drm_atomic_state *state,
-				struct drm_crtc *crtc,
-				struct drm_crtc_state *new_crtc_state)
+static void
+dm_get_plane_scale(struct drm_plane_state *plane_state,
+		   int *out_plane_scale_w, int *out_plane_scale_h)
 {
-	struct drm_plane *cursor = crtc->cursor, *underlying;
-	struct drm_plane_state *new_cursor_state, *new_underlying_state;
-	int i;
-	int cursor_scale_w, cursor_scale_h, underlying_scale_w, underlying_scale_h;
-	int cursor_src_w, cursor_src_h;
-	int underlying_src_w, underlying_src_h;
+	int plane_src_w, plane_src_h;
 
-	/* On DCE and DCN there is no dedicated hardware cursor plane. We get a
-	 * cursor per pipe but it's going to inherit the scaling and
-	 * positioning from the underlying pipe. Check the cursor plane's
-	 * blending properties match the underlying planes'. */
-
-	new_cursor_state = drm_atomic_get_new_plane_state(state, cursor);
-	if (!new_cursor_state || !new_cursor_state->fb) {
-		return 0;
-	}
+	dm_get_oriented_plane_size(plane_state, &plane_src_w, &plane_src_h);
+	*out_plane_scale_w = plane_src_w ? plane_state->crtc_w * 1000 / plane_src_w : 0;
+	*out_plane_scale_h = plane_src_h ? plane_state->crtc_h * 1000 / plane_src_h : 0;
+}
 
-	dm_get_oriented_plane_size(new_cursor_state, &cursor_src_w, &cursor_src_h);
-	cursor_scale_w = new_cursor_state->crtc_w * 1000 / cursor_src_w;
-	cursor_scale_h = new_cursor_state->crtc_h * 1000 / cursor_src_h;
+/*
+ * The normalized_zpos value cannot be used by this iterator directly. It's only
+ * calculated for enabled planes, potentially causing normalized_zpos collisions
+ * between enabled/disabled planes in the atomic state. We need a unique value
+ * so that the iterator will not generate the same object twice, or loop
+ * indefinitely.
+ */
+static inline struct __drm_planes_state *__get_next_zpos(
+	struct drm_atomic_state *state,
+	struct __drm_planes_state *prev)
+{
+	unsigned int highest_zpos = 0, prev_zpos = 256;
+	uint32_t highest_id = 0, prev_id = UINT_MAX;
+	struct drm_plane_state *new_plane_state;
+	struct drm_plane *plane;
+	int i, highest_i = -1;
 
-	for_each_new_plane_in_state_reverse(state, underlying, new_underlying_state, i) {
-		/* Narrow down to non-cursor planes on the same CRTC as the cursor */
-		if (new_underlying_state->crtc != crtc || underlying == crtc->cursor)
-			continue;
+	if (prev != NULL) {
+		prev_zpos = prev->new_state->zpos;
+		prev_id = prev->ptr->base.id;
+	}
 
-		/* Ignore disabled planes */
-		if (!new_underlying_state->fb)
+	for_each_new_plane_in_state(state, plane, new_plane_state, i) {
+		/* Skip planes with higher zpos than the previously returned */
+		if (new_plane_state->zpos > prev_zpos ||
+		    (new_plane_state->zpos == prev_zpos &&
+		     plane->base.id >= prev_id))
 			continue;
 
-		dm_get_oriented_plane_size(new_underlying_state,
-					   &underlying_src_w, &underlying_src_h);
-		underlying_scale_w = new_underlying_state->crtc_w * 1000 / underlying_src_w;
-		underlying_scale_h = new_underlying_state->crtc_h * 1000 / underlying_src_h;
-
-		if (cursor_scale_w != underlying_scale_w ||
-		    cursor_scale_h != underlying_scale_h) {
-			drm_dbg_atomic(crtc->dev,
-				       "Cursor [PLANE:%d:%s] scaling doesn't match underlying [PLANE:%d:%s]\n",
-				       cursor->base.id, cursor->name, underlying->base.id, underlying->name);
-			return -EINVAL;
+		/* Save the index of the plane with highest zpos */
+		if (new_plane_state->zpos > highest_zpos ||
+		    (new_plane_state->zpos == highest_zpos &&
+		     plane->base.id > highest_id)) {
+			highest_zpos = new_plane_state->zpos;
+			highest_id = plane->base.id;
+			highest_i = i;
 		}
-
-		/* If this plane covers the whole CRTC, no need to check planes underneath */
-		if (new_underlying_state->crtc_x <= 0 &&
-		    new_underlying_state->crtc_y <= 0 &&
-		    new_underlying_state->crtc_x + new_underlying_state->crtc_w >= new_crtc_state->mode.hdisplay &&
-		    new_underlying_state->crtc_y + new_underlying_state->crtc_h >= new_crtc_state->mode.vdisplay)
-			break;
 	}
 
-	return 0;
+	if (highest_i < 0)
+		return NULL;
+
+	return &state->planes[highest_i];
 }
 
+/*
+ * Use the uniqueness of the plane's (zpos, drm obj ID) combination to iterate
+ * by descending zpos, as read from the new plane state. This is the same
+ * ordering as defined by drm_atomic_normalize_zpos().
+ */
+#define for_each_oldnew_plane_in_descending_zpos(__state, plane, old_plane_state, new_plane_state) \
+	for (struct __drm_planes_state *__i = __get_next_zpos((__state), NULL); \
+	     __i != NULL; __i = __get_next_zpos((__state), __i))		\
+		for_each_if(((plane) = __i->ptr,				\
+			     (void)(plane) /* Only to avoid unused-but-set-variable warning */, \
+			     (old_plane_state) = __i->old_state,		\
+			     (new_plane_state) = __i->new_state, 1))
+
 static int add_affected_mst_dsc_crtcs(struct drm_atomic_state *state, struct drm_crtc *crtc)
 {
 	struct drm_connector *connector;
 	struct drm_connector_state *conn_state, *old_conn_state;
 	struct amdgpu_dm_connector *aconnector = NULL;
 	int i;
+
 	for_each_oldnew_connector_in_state(state, connector, old_conn_state, conn_state, i) {
 		if (!conn_state->crtc)
 			conn_state = old_conn_state;
@@ -9799,6 +12178,9 @@ static int add_affected_mst_dsc_crtcs(struct drm_atomic_state *state, struct drm
 		if (conn_state->crtc != crtc)
 			continue;
 
+		if (connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+			continue;
+
 		aconnector = to_amdgpu_dm_connector(connector);
 		if (!aconnector->mst_output_port || !aconnector->mst_root)
 			aconnector = NULL;
@@ -9813,6 +12195,193 @@ static int add_affected_mst_dsc_crtcs(struct drm_atomic_state *state, struct drm
 }
 
 /**
+ * DOC: Cursor Modes - Native vs Overlay
+ *
+ * In native mode, the cursor uses a integrated cursor pipe within each DCN hw
+ * plane. It does not require a dedicated hw plane to enable, but it is
+ * subjected to the same z-order and scaling as the hw plane. It also has format
+ * restrictions, a RGB cursor in native mode cannot be enabled within a non-RGB
+ * hw plane.
+ *
+ * In overlay mode, the cursor uses a separate DCN hw plane, and thus has its
+ * own scaling and z-pos. It also has no blending restrictions. It lends to a
+ * cursor behavior more akin to a DRM client's expectations. However, it does
+ * occupy an extra DCN plane, and therefore will only be used if a DCN plane is
+ * available.
+ */
+
+/**
+ * dm_crtc_get_cursor_mode() - Determine the required cursor mode on crtc
+ * @adev: amdgpu device
+ * @state: DRM atomic state
+ * @dm_crtc_state: amdgpu state for the CRTC containing the cursor
+ * @cursor_mode: Returns the required cursor mode on dm_crtc_state
+ *
+ * Get whether the cursor should be enabled in native mode, or overlay mode, on
+ * the dm_crtc_state.
+ *
+ * The cursor should be enabled in overlay mode if there exists an underlying
+ * plane - on which the cursor may be blended - that is either YUV formatted, or
+ * scaled differently from the cursor.
+ *
+ * Since zpos info is required, drm_atomic_normalize_zpos must be called before
+ * calling this function.
+ *
+ * Return: 0 on success, or an error code if getting the cursor plane state
+ * failed.
+ */
+static int dm_crtc_get_cursor_mode(struct amdgpu_device *adev,
+				   struct drm_atomic_state *state,
+				   struct dm_crtc_state *dm_crtc_state,
+				   enum amdgpu_dm_cursor_mode *cursor_mode)
+{
+	struct drm_plane_state *old_plane_state, *plane_state, *cursor_state;
+	struct drm_crtc_state *crtc_state = &dm_crtc_state->base;
+	struct drm_plane *plane;
+	bool consider_mode_change = false;
+	bool entire_crtc_covered = false;
+	bool cursor_changed = false;
+	int underlying_scale_w, underlying_scale_h;
+	int cursor_scale_w, cursor_scale_h;
+	int i;
+
+	/* Overlay cursor not supported on HW before DCN
+	 * DCN401 does not have the cursor-on-scaled-plane or cursor-on-yuv-plane restrictions
+	 * as previous DCN generations, so enable native mode on DCN401 in addition to DCE
+	 */
+	if (amdgpu_ip_version(adev, DCE_HWIP, 0) == 0 ||
+	    amdgpu_ip_version(adev, DCE_HWIP, 0) == IP_VERSION(4, 0, 1)) {
+		*cursor_mode = DM_CURSOR_NATIVE_MODE;
+		return 0;
+	}
+
+	/* Init cursor_mode to be the same as current */
+	*cursor_mode = dm_crtc_state->cursor_mode;
+
+	/*
+	 * Cursor mode can change if a plane's format changes, scale changes, is
+	 * enabled/disabled, or z-order changes.
+	 */
+	for_each_oldnew_plane_in_state(state, plane, old_plane_state, plane_state, i) {
+		int new_scale_w, new_scale_h, old_scale_w, old_scale_h;
+
+		/* Only care about planes on this CRTC */
+		if ((drm_plane_mask(plane) & crtc_state->plane_mask) == 0)
+			continue;
+
+		if (plane->type == DRM_PLANE_TYPE_CURSOR)
+			cursor_changed = true;
+
+		if (drm_atomic_plane_enabling(old_plane_state, plane_state) ||
+		    drm_atomic_plane_disabling(old_plane_state, plane_state) ||
+		    old_plane_state->fb->format != plane_state->fb->format) {
+			consider_mode_change = true;
+			break;
+		}
+
+		dm_get_plane_scale(plane_state, &new_scale_w, &new_scale_h);
+		dm_get_plane_scale(old_plane_state, &old_scale_w, &old_scale_h);
+		if (new_scale_w != old_scale_w || new_scale_h != old_scale_h) {
+			consider_mode_change = true;
+			break;
+		}
+	}
+
+	if (!consider_mode_change && !crtc_state->zpos_changed)
+		return 0;
+
+	/*
+	 * If no cursor change on this CRTC, and not enabled on this CRTC, then
+	 * no need to set cursor mode. This avoids needlessly locking the cursor
+	 * state.
+	 */
+	if (!cursor_changed &&
+	    !(drm_plane_mask(crtc_state->crtc->cursor) & crtc_state->plane_mask)) {
+		return 0;
+	}
+
+	cursor_state = drm_atomic_get_plane_state(state,
+						  crtc_state->crtc->cursor);
+	if (IS_ERR(cursor_state))
+		return PTR_ERR(cursor_state);
+
+	/* Cursor is disabled */
+	if (!cursor_state->fb)
+		return 0;
+
+	/* For all planes in descending z-order (all of which are below cursor
+	 * as per zpos definitions), check their scaling and format
+	 */
+	for_each_oldnew_plane_in_descending_zpos(state, plane, old_plane_state, plane_state) {
+
+		/* Only care about non-cursor planes on this CRTC */
+		if ((drm_plane_mask(plane) & crtc_state->plane_mask) == 0 ||
+		    plane->type == DRM_PLANE_TYPE_CURSOR)
+			continue;
+
+		/* Underlying plane is YUV format - use overlay cursor */
+		if (amdgpu_dm_plane_is_video_format(plane_state->fb->format->format)) {
+			*cursor_mode = DM_CURSOR_OVERLAY_MODE;
+			return 0;
+		}
+
+		dm_get_plane_scale(plane_state,
+				   &underlying_scale_w, &underlying_scale_h);
+		dm_get_plane_scale(cursor_state,
+				   &cursor_scale_w, &cursor_scale_h);
+
+		/* Underlying plane has different scale - use overlay cursor */
+		if (cursor_scale_w != underlying_scale_w &&
+		    cursor_scale_h != underlying_scale_h) {
+			*cursor_mode = DM_CURSOR_OVERLAY_MODE;
+			return 0;
+		}
+
+		/* If this plane covers the whole CRTC, no need to check planes underneath */
+		if (plane_state->crtc_x <= 0 && plane_state->crtc_y <= 0 &&
+		    plane_state->crtc_x + plane_state->crtc_w >= crtc_state->mode.hdisplay &&
+		    plane_state->crtc_y + plane_state->crtc_h >= crtc_state->mode.vdisplay) {
+			entire_crtc_covered = true;
+			break;
+		}
+	}
+
+	/* If planes do not cover the entire CRTC, use overlay mode to enable
+	 * cursor over holes
+	 */
+	if (entire_crtc_covered)
+		*cursor_mode = DM_CURSOR_NATIVE_MODE;
+	else
+		*cursor_mode = DM_CURSOR_OVERLAY_MODE;
+
+	return 0;
+}
+
+static bool amdgpu_dm_crtc_mem_type_changed(struct drm_device *dev,
+					    struct drm_atomic_state *state,
+					    struct drm_crtc_state *crtc_state)
+{
+	struct drm_plane *plane;
+	struct drm_plane_state *new_plane_state, *old_plane_state;
+
+	drm_for_each_plane_mask(plane, dev, crtc_state->plane_mask) {
+		new_plane_state = drm_atomic_get_plane_state(state, plane);
+		old_plane_state = drm_atomic_get_plane_state(state, plane);
+
+		if (IS_ERR(new_plane_state) || IS_ERR(old_plane_state)) {
+			drm_err(dev, "Failed to get plane state for plane %s\n", plane->name);
+			return false;
+		}
+
+		if (old_plane_state->fb && new_plane_state->fb &&
+		    get_mem_type(old_plane_state->fb) != get_mem_type(new_plane_state->fb))
+			return true;
+	}
+
+	return false;
+}
+
+/**
  * amdgpu_dm_atomic_check() - Atomic check implementation for AMDgpu DM.
  *
  * @dev: The DRM device
@@ -9848,7 +12417,7 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 	struct drm_crtc *crtc;
 	struct drm_crtc_state *old_crtc_state, *new_crtc_state;
 	struct drm_plane *plane;
-	struct drm_plane_state *old_plane_state, *new_plane_state;
+	struct drm_plane_state *old_plane_state, *new_plane_state, *new_cursor_state;
 	enum dc_status status;
 	int ret, i;
 	bool lock_and_validation_needed = false;
@@ -9856,13 +12425,13 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 	struct dm_crtc_state *dm_old_crtc_state, *dm_new_crtc_state;
 	struct drm_dp_mst_topology_mgr *mgr;
 	struct drm_dp_mst_topology_state *mst_state;
-	struct dsc_mst_fairness_vars vars[MAX_PIPES];
+	struct dsc_mst_fairness_vars vars[MAX_PIPES] = {0};
 
 	trace_amdgpu_dm_atomic_check_begin(state);
 
 	ret = drm_atomic_helper_check_modeset(dev, state);
 	if (ret) {
-		DRM_DEBUG_DRIVER("drm_atomic_helper_check_modeset() failed\n");
+		drm_dbg_atomic(dev, "drm_atomic_helper_check_modeset() failed\n");
 		goto fail;
 	}
 
@@ -9877,7 +12446,7 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 
 		new_crtc_state = drm_atomic_get_crtc_state(state, new_con_state->crtc);
 		if (IS_ERR(new_crtc_state)) {
-			DRM_DEBUG_DRIVER("drm_atomic_get_crtc_state() failed\n");
+			drm_dbg_atomic(dev, "drm_atomic_get_crtc_state() failed\n");
 			ret = PTR_ERR(new_crtc_state);
 			goto fail;
 		}
@@ -9892,7 +12461,7 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 			if (drm_atomic_crtc_needs_modeset(new_crtc_state)) {
 				ret = add_affected_mst_dsc_crtcs(state, crtc);
 				if (ret) {
-					DRM_DEBUG_DRIVER("add_affected_mst_dsc_crtcs() failed\n");
+					drm_dbg_atomic(dev, "add_affected_mst_dsc_crtcs() failed\n");
 					goto fail;
 				}
 			}
@@ -9909,7 +12478,7 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 
 		ret = amdgpu_dm_verify_lut_sizes(new_crtc_state);
 		if (ret) {
-			DRM_DEBUG_DRIVER("amdgpu_dm_verify_lut_sizes() failed\n");
+			drm_dbg_atomic(dev, "amdgpu_dm_verify_lut_sizes() failed\n");
 			goto fail;
 		}
 
@@ -9918,13 +12487,13 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 
 		ret = drm_atomic_add_affected_connectors(state, crtc);
 		if (ret) {
-			DRM_DEBUG_DRIVER("drm_atomic_add_affected_connectors() failed\n");
+			drm_dbg_atomic(dev, "drm_atomic_add_affected_connectors() failed\n");
 			goto fail;
 		}
 
 		ret = drm_atomic_add_affected_planes(state, crtc);
 		if (ret) {
-			DRM_DEBUG_DRIVER("drm_atomic_add_affected_planes() failed\n");
+			drm_dbg_atomic(dev, "drm_atomic_add_affected_planes() failed\n");
 			goto fail;
 		}
 
@@ -9963,7 +12532,7 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 
 			if (IS_ERR(new_plane_state)) {
 				ret = PTR_ERR(new_plane_state);
-				DRM_DEBUG_DRIVER("new_plane_state is BAD\n");
+				drm_dbg_atomic(dev, "new_plane_state is BAD\n");
 				goto fail;
 			}
 		}
@@ -9981,8 +12550,35 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 		goto fail;
 	}
 
+	/*
+	 * Determine whether cursors on each CRTC should be enabled in native or
+	 * overlay mode.
+	 */
+	for_each_new_crtc_in_state(state, crtc, new_crtc_state, i) {
+		dm_new_crtc_state = to_dm_crtc_state(new_crtc_state);
+
+		ret = dm_crtc_get_cursor_mode(adev, state, dm_new_crtc_state,
+					      &dm_new_crtc_state->cursor_mode);
+		if (ret) {
+			drm_dbg(dev, "Failed to determine cursor mode\n");
+			goto fail;
+		}
+
+		/*
+		 * If overlay cursor is needed, DC cannot go through the
+		 * native cursor update path. All enabled planes on the CRTC
+		 * need to be added for DC to not disable a plane by mistake
+		 */
+		if (dm_new_crtc_state->cursor_mode == DM_CURSOR_OVERLAY_MODE) {
+			ret = drm_atomic_add_affected_planes(state, crtc);
+			if (ret)
+				goto fail;
+		}
+	}
+
 	/* Remove exiting planes if they are modified */
-	for_each_oldnew_plane_in_state_reverse(state, plane, old_plane_state, new_plane_state, i) {
+	for_each_oldnew_plane_in_descending_zpos(state, plane, old_plane_state, new_plane_state) {
+
 		ret = dm_update_plane_state(dc, state, plane,
 					    old_plane_state,
 					    new_plane_state,
@@ -9990,7 +12586,7 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 					    &lock_and_validation_needed,
 					    &is_top_most_overlay);
 		if (ret) {
-			DRM_DEBUG_DRIVER("dm_update_plane_state() failed\n");
+			drm_dbg_atomic(dev, "dm_update_plane_state() failed\n");
 			goto fail;
 		}
 	}
@@ -10003,7 +12599,7 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 					   false,
 					   &lock_and_validation_needed);
 		if (ret) {
-			DRM_DEBUG_DRIVER("DISABLE: dm_update_crtc_state() failed\n");
+			drm_dbg_atomic(dev, "DISABLE: dm_update_crtc_state() failed\n");
 			goto fail;
 		}
 	}
@@ -10016,13 +12612,13 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 					   true,
 					   &lock_and_validation_needed);
 		if (ret) {
-			DRM_DEBUG_DRIVER("ENABLE: dm_update_crtc_state() failed\n");
+			drm_dbg_atomic(dev, "ENABLE: dm_update_crtc_state() failed\n");
 			goto fail;
 		}
 	}
 
 	/* Add new/modified planes */
-	for_each_oldnew_plane_in_state_reverse(state, plane, old_plane_state, new_plane_state, i) {
+	for_each_oldnew_plane_in_descending_zpos(state, plane, old_plane_state, new_plane_state) {
 		ret = dm_update_plane_state(dc, state, plane,
 					    old_plane_state,
 					    new_plane_state,
@@ -10030,35 +12626,75 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 					    &lock_and_validation_needed,
 					    &is_top_most_overlay);
 		if (ret) {
-			DRM_DEBUG_DRIVER("dm_update_plane_state() failed\n");
+			drm_dbg_atomic(dev, "dm_update_plane_state() failed\n");
 			goto fail;
 		}
 	}
 
+#if defined(CONFIG_DRM_AMD_DC_FP)
 	if (dc_resource_is_dsc_encoding_supported(dc)) {
 		ret = pre_validate_dsc(state, &dm_state, vars);
 		if (ret != 0)
 			goto fail;
 	}
+#endif
 
 	/* Run this here since we want to validate the streams we created */
 	ret = drm_atomic_helper_check_planes(dev, state);
 	if (ret) {
-		DRM_DEBUG_DRIVER("drm_atomic_helper_check_planes() failed\n");
+		drm_dbg_atomic(dev, "drm_atomic_helper_check_planes() failed\n");
 		goto fail;
 	}
 
 	for_each_new_crtc_in_state(state, crtc, new_crtc_state, i) {
 		dm_new_crtc_state = to_dm_crtc_state(new_crtc_state);
 		if (dm_new_crtc_state->mpo_requested)
-			DRM_DEBUG_DRIVER("MPO enablement requested on crtc:[%p]\n", crtc);
+			drm_dbg_atomic(dev, "MPO enablement requested on crtc:[%p]\n", crtc);
 	}
 
-	/* Check cursor planes scaling */
+	/* Check cursor restrictions */
 	for_each_new_crtc_in_state(state, crtc, new_crtc_state, i) {
-		ret = dm_check_crtc_cursor(state, crtc, new_crtc_state);
+		enum amdgpu_dm_cursor_mode required_cursor_mode;
+		int is_rotated, is_scaled;
+
+		/* Overlay cusor not subject to native cursor restrictions */
+		dm_new_crtc_state = to_dm_crtc_state(new_crtc_state);
+		if (dm_new_crtc_state->cursor_mode == DM_CURSOR_OVERLAY_MODE)
+			continue;
+
+		/* Check if rotation or scaling is enabled on DCN401 */
+		if ((drm_plane_mask(crtc->cursor) & new_crtc_state->plane_mask) &&
+		    amdgpu_ip_version(adev, DCE_HWIP, 0) == IP_VERSION(4, 0, 1)) {
+			new_cursor_state = drm_atomic_get_new_plane_state(state, crtc->cursor);
+
+			is_rotated = new_cursor_state &&
+				((new_cursor_state->rotation & DRM_MODE_ROTATE_MASK) != DRM_MODE_ROTATE_0);
+			is_scaled = new_cursor_state && ((new_cursor_state->src_w >> 16 != new_cursor_state->crtc_w) ||
+				(new_cursor_state->src_h >> 16 != new_cursor_state->crtc_h));
+
+			if (is_rotated || is_scaled) {
+				drm_dbg_driver(
+					crtc->dev,
+					"[CRTC:%d:%s] cannot enable hardware cursor due to rotation/scaling\n",
+					crtc->base.id, crtc->name);
+				ret = -EINVAL;
+				goto fail;
+			}
+		}
+
+		/* If HW can only do native cursor, check restrictions again */
+		ret = dm_crtc_get_cursor_mode(adev, state, dm_new_crtc_state,
+					      &required_cursor_mode);
 		if (ret) {
-			DRM_DEBUG_DRIVER("dm_check_crtc_cursor() failed\n");
+			drm_dbg_driver(crtc->dev,
+				       "[CRTC:%d:%s] Checking cursor mode failed\n",
+				       crtc->base.id, crtc->name);
+			goto fail;
+		} else if (required_cursor_mode == DM_CURSOR_OVERLAY_MODE) {
+			drm_dbg_driver(crtc->dev,
+				       "[CRTC:%d:%s] Cannot enable native cursor due to scaling or YUV restrictions\n",
+				       crtc->base.id, crtc->name);
+			ret = -EINVAL;
 			goto fail;
 		}
 	}
@@ -10141,25 +12777,30 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 	if (lock_and_validation_needed) {
 		ret = dm_atomic_get_state(state, &dm_state);
 		if (ret) {
-			DRM_DEBUG_DRIVER("dm_atomic_get_state() failed\n");
+			drm_dbg_atomic(dev, "dm_atomic_get_state() failed\n");
 			goto fail;
 		}
 
 		ret = do_aquire_global_lock(dev, state);
 		if (ret) {
-			DRM_DEBUG_DRIVER("do_aquire_global_lock() failed\n");
+			drm_dbg_atomic(dev, "do_aquire_global_lock() failed\n");
 			goto fail;
 		}
 
-		ret = compute_mst_dsc_configs_for_state(state, dm_state->context, vars);
-		if (ret) {
-			DRM_DEBUG_DRIVER("compute_mst_dsc_configs_for_state() failed\n");
-			goto fail;
+#if defined(CONFIG_DRM_AMD_DC_FP)
+		if (dc_resource_is_dsc_encoding_supported(dc)) {
+			ret = compute_mst_dsc_configs_for_state(state, dm_state->context, vars);
+			if (ret) {
+				drm_dbg_atomic(dev, "MST_DSC compute_mst_dsc_configs_for_state() failed\n");
+				ret = -EINVAL;
+				goto fail;
+			}
 		}
+#endif
 
 		ret = dm_update_mst_vcpi_slots_for_dsc(state, dm_state->context, vars);
 		if (ret) {
-			DRM_DEBUG_DRIVER("dm_update_mst_vcpi_slots_for_dsc() failed\n");
+			drm_dbg_atomic(dev, "dm_update_mst_vcpi_slots_for_dsc() failed\n");
 			goto fail;
 		}
 
@@ -10171,12 +12812,12 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 		 */
 		ret = drm_dp_mst_atomic_check(state);
 		if (ret) {
-			DRM_DEBUG_DRIVER("drm_dp_mst_atomic_check() failed\n");
+			drm_dbg_atomic(dev, "MST drm_dp_mst_atomic_check() failed\n");
 			goto fail;
 		}
-		status = dc_validate_global_state(dc, dm_state->context, true);
+		status = dc_validate_global_state(dc, dm_state->context, DC_VALIDATE_MODE_ONLY);
 		if (status != DC_OK) {
-			DRM_DEBUG_DRIVER("DC global validation failure: %s (%d)",
+			drm_dbg_atomic(dev, "DC global validation failure: %s (%d)",
 				       dc_status_to_str(status), status);
 			ret = -EINVAL;
 			goto fail;
@@ -10203,7 +12844,7 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 				int j = state->num_private_objs-1;
 
 				dm_atomic_destroy_state(obj,
-						state->private_objs[i].state);
+						state->private_objs[i].state_to_destroy);
 
 				/* If i is not at the end of the array then the
 				 * last element needs to be moved to where i was
@@ -10214,7 +12855,7 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 						state->private_objs[j];
 
 				state->private_objs[j].ptr = NULL;
-				state->private_objs[j].state = NULL;
+				state->private_objs[j].state_to_destroy = NULL;
 				state->private_objs[j].old_state = NULL;
 				state->private_objs[j].new_state = NULL;
 
@@ -10225,13 +12866,26 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 	}
 
 	/* Store the overall update type for use later in atomic check. */
-	for_each_new_crtc_in_state (state, crtc, new_crtc_state, i) {
+	for_each_new_crtc_in_state(state, crtc, new_crtc_state, i) {
 		struct dm_crtc_state *dm_new_crtc_state =
 			to_dm_crtc_state(new_crtc_state);
 
+		/*
+		 * Only allow async flips for fast updates that don't change
+		 * the FB pitch, the DCC state, rotation, mem_type, etc.
+		 */
+		if (new_crtc_state->async_flip &&
+		    (lock_and_validation_needed ||
+		     amdgpu_dm_crtc_mem_type_changed(dev, state, new_crtc_state))) {
+			drm_dbg_atomic(crtc->dev,
+				       "[CRTC:%d:%s] async flips are only supported for fast updates\n",
+				       crtc->base.id, crtc->name);
+			ret = -EINVAL;
+			goto fail;
+		}
+
 		dm_new_crtc_state->update_type = lock_and_validation_needed ?
-							 UPDATE_TYPE_FULL :
-							 UPDATE_TYPE_FAST;
+			UPDATE_TYPE_FULL : UPDATE_TYPE_FAST;
 	}
 
 	/* Must be success */
@@ -10243,36 +12897,17 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev,
 
 fail:
 	if (ret == -EDEADLK)
-		DRM_DEBUG_DRIVER("Atomic check stopped to avoid deadlock.\n");
+		drm_dbg_atomic(dev, "Atomic check stopped to avoid deadlock.\n");
 	else if (ret == -EINTR || ret == -EAGAIN || ret == -ERESTARTSYS)
-		DRM_DEBUG_DRIVER("Atomic check stopped due to signal.\n");
+		drm_dbg_atomic(dev, "Atomic check stopped due to signal.\n");
 	else
-		DRM_DEBUG_DRIVER("Atomic check failed with err: %d \n", ret);
+		drm_dbg_atomic(dev, "Atomic check failed with err: %d\n", ret);
 
 	trace_amdgpu_dm_atomic_check_finish(state, ret);
 
 	return ret;
 }
 
-static bool is_dp_capable_without_timing_msa(struct dc *dc,
-					     struct amdgpu_dm_connector *amdgpu_dm_connector)
-{
-	u8 dpcd_data;
-	bool capable = false;
-
-	if (amdgpu_dm_connector->dc_link &&
-		dm_helpers_dp_read_dpcd(
-				NULL,
-				amdgpu_dm_connector->dc_link,
-				DP_DOWN_STREAM_PORT_COUNT,
-				&dpcd_data,
-				sizeof(dpcd_data))) {
-		capable = (dpcd_data & DP_MSA_TIMING_PAR_IGNORED) ? true:false;
-	}
-
-	return capable;
-}
-
 static bool dm_edid_parser_send_cea(struct amdgpu_display_manager *dm,
 		unsigned int offset,
 		unsigned int total_length,
@@ -10301,9 +12936,9 @@ static bool dm_edid_parser_send_cea(struct amdgpu_display_manager *dm,
 	input->cea_total_length = total_length;
 	memcpy(input->payload, data, length);
 
-	res = dc_dmub_srv_cmd_with_reply_data(dm->dc->ctx->dmub_srv, &cmd);
+	res = dc_wake_and_execute_dmub_cmd(dm->dc->ctx, &cmd, DM_DMUB_WAIT_TYPE_WAIT_WITH_REPLY);
 	if (!res) {
-		DRM_ERROR("EDID CEA parser failed\n");
+		drm_err(adev_to_drm(dm->adev), "EDID CEA parser failed\n");
 		return false;
 	}
 
@@ -10311,7 +12946,7 @@ static bool dm_edid_parser_send_cea(struct amdgpu_display_manager *dm,
 
 	if (output->type == DMUB_CMD__EDID_CEA_ACK) {
 		if (!output->ack.success) {
-			DRM_ERROR("EDID CEA ack failed at offset %d\n",
+			drm_err(adev_to_drm(dm->adev), "EDID CEA ack failed at offset %d\n",
 					output->ack.offset);
 		}
 	} else if (output->type == DMUB_CMD__EDID_CEA_AMD_VSDB) {
@@ -10323,7 +12958,7 @@ static bool dm_edid_parser_send_cea(struct amdgpu_display_manager *dm,
 		vsdb->min_refresh_rate_hz = output->amd_vsdb.min_frame_rate;
 		vsdb->max_refresh_rate_hz = output->amd_vsdb.max_frame_rate;
 	} else {
-		DRM_WARN("Unknown EDID CEA parser results\n");
+		drm_warn(adev_to_drm(dm->adev), "Unknown EDID CEA parser results\n");
 		return false;
 	}
 
@@ -10403,8 +13038,87 @@ static bool parse_edid_cea(struct amdgpu_dm_connector *aconnector,
 	return ret;
 }
 
+static void parse_edid_displayid_vrr(struct drm_connector *connector,
+				     const struct edid *edid)
+{
+	u8 *edid_ext = NULL;
+	int i;
+	int j = 0;
+	u16 min_vfreq;
+	u16 max_vfreq;
+
+	if (edid == NULL || edid->extensions == 0)
+		return;
+
+	/* Find DisplayID extension */
+	for (i = 0; i < edid->extensions; i++) {
+		edid_ext = (void *)(edid + (i + 1));
+		if (edid_ext[0] == DISPLAYID_EXT)
+			break;
+	}
+
+	if (edid_ext == NULL)
+		return;
+
+	while (j < EDID_LENGTH) {
+		/* Get dynamic video timing range from DisplayID if available */
+		if (EDID_LENGTH - j > 13 && edid_ext[j] == 0x25	&&
+		    (edid_ext[j+1] & 0xFE) == 0 && (edid_ext[j+2] == 9)) {
+			min_vfreq = edid_ext[j+9];
+			if (edid_ext[j+1] & 7)
+				max_vfreq = edid_ext[j+10] + ((edid_ext[j+11] & 3) << 8);
+			else
+				max_vfreq = edid_ext[j+10];
+
+			if (max_vfreq && min_vfreq) {
+				connector->display_info.monitor_range.max_vfreq = max_vfreq;
+				connector->display_info.monitor_range.min_vfreq = min_vfreq;
+
+				return;
+			}
+		}
+		j++;
+	}
+}
+
+static int parse_amd_vsdb(struct amdgpu_dm_connector *aconnector,
+			  const struct edid *edid, struct amdgpu_hdmi_vsdb_info *vsdb_info)
+{
+	u8 *edid_ext = NULL;
+	int i;
+	int j = 0;
+
+	if (edid == NULL || edid->extensions == 0)
+		return -ENODEV;
+
+	/* Find DisplayID extension */
+	for (i = 0; i < edid->extensions; i++) {
+		edid_ext = (void *)(edid + (i + 1));
+		if (edid_ext[0] == DISPLAYID_EXT)
+			break;
+	}
+
+	while (j < EDID_LENGTH - sizeof(struct amd_vsdb_block)) {
+		struct amd_vsdb_block *amd_vsdb = (struct amd_vsdb_block *)&edid_ext[j];
+		unsigned int ieeeId = (amd_vsdb->ieee_id[2] << 16) | (amd_vsdb->ieee_id[1] << 8) | (amd_vsdb->ieee_id[0]);
+
+		if (ieeeId == HDMI_AMD_VENDOR_SPECIFIC_DATA_BLOCK_IEEE_REGISTRATION_ID &&
+				amd_vsdb->version == HDMI_AMD_VENDOR_SPECIFIC_DATA_BLOCK_VERSION_3) {
+			vsdb_info->replay_mode = (amd_vsdb->feature_caps & AMD_VSDB_VERSION_3_FEATURECAP_REPLAYMODE) ? true : false;
+			vsdb_info->amd_vsdb_version = HDMI_AMD_VENDOR_SPECIFIC_DATA_BLOCK_VERSION_3;
+			DRM_DEBUG_KMS("Panel supports Replay Mode: %d\n", vsdb_info->replay_mode);
+
+			return true;
+		}
+		j++;
+	}
+
+	return false;
+}
+
 static int parse_hdmi_amd_vsdb(struct amdgpu_dm_connector *aconnector,
-		struct edid *edid, struct amdgpu_hdmi_vsdb_info *vsdb_info)
+			       const struct edid *edid,
+			       struct amdgpu_hdmi_vsdb_info *vsdb_info)
 {
 	u8 *edid_ext = NULL;
 	int i;
@@ -10438,7 +13152,7 @@ static int parse_hdmi_amd_vsdb(struct amdgpu_dm_connector *aconnector,
  * amdgpu_dm_update_freesync_caps - Update Freesync capabilities
  *
  * @connector: Connector to query.
- * @edid: EDID from monitor
+ * @drm_edid: DRM EDID from monitor
  *
  * Amdgpu supports Freesync in DP and HDMI displays, and it is required to keep
  * track of some of the display information in the internal data struct used by
@@ -10446,25 +13160,21 @@ static int parse_hdmi_amd_vsdb(struct amdgpu_dm_connector *aconnector,
  * FreeSync parameters.
  */
 void amdgpu_dm_update_freesync_caps(struct drm_connector *connector,
-				    struct edid *edid)
+				    const struct drm_edid *drm_edid)
 {
 	int i = 0;
-	struct detailed_timing *timing;
-	struct detailed_non_pixel *data;
-	struct detailed_data_monitor_range *range;
 	struct amdgpu_dm_connector *amdgpu_dm_connector =
 			to_amdgpu_dm_connector(connector);
 	struct dm_connector_state *dm_con_state = NULL;
 	struct dc_sink *sink;
-
-	struct drm_device *dev = connector->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
+	struct amdgpu_device *adev = drm_to_adev(connector->dev);
 	struct amdgpu_hdmi_vsdb_info vsdb_info = {0};
+	const struct edid *edid;
 	bool freesync_capable = false;
 	enum adaptive_sync_type as_type = ADAPTIVE_SYNC_TYPE_NONE;
 
 	if (!connector->state) {
-		DRM_ERROR("%s - Connector has no state", __func__);
+		drm_err(adev_to_drm(adev), "%s - Connector has no state", __func__);
 		goto update;
 	}
 
@@ -10472,14 +13182,13 @@ void amdgpu_dm_update_freesync_caps(struct drm_connector *connector,
 		amdgpu_dm_connector->dc_sink :
 		amdgpu_dm_connector->dc_em_sink;
 
-	if (!edid || !sink) {
+	drm_edid_connector_update(connector, drm_edid);
+
+	if (!drm_edid || !sink) {
 		dm_con_state = to_dm_connector_state(connector->state);
 
 		amdgpu_dm_connector->min_vfreq = 0;
 		amdgpu_dm_connector->max_vfreq = 0;
-		amdgpu_dm_connector->pixel_clock_mhz = 0;
-		connector->display_info.monitor_range.min_vfreq = 0;
-		connector->display_info.monitor_range.max_vfreq = 0;
 		freesync_capable = false;
 
 		goto update;
@@ -10487,63 +13196,37 @@ void amdgpu_dm_update_freesync_caps(struct drm_connector *connector,
 
 	dm_con_state = to_dm_connector_state(connector->state);
 
-	if (!adev->dm.freesync_module)
+	if (!adev->dm.freesync_module || !dc_supports_vrr(sink->ctx->dce_version))
 		goto update;
 
-	if (sink->sink_signal == SIGNAL_TYPE_DISPLAY_PORT
-		|| sink->sink_signal == SIGNAL_TYPE_EDP) {
-		bool edid_check_required = false;
+	edid = drm_edid_raw(drm_edid); // FIXME: Get rid of drm_edid_raw()
 
-		if (edid) {
-			edid_check_required = is_dp_capable_without_timing_msa(
-						adev->dm.dc,
-						amdgpu_dm_connector);
-		}
-
-		if (edid_check_required == true && (edid->version > 1 ||
-		   (edid->version == 1 && edid->revision > 1))) {
-			for (i = 0; i < 4; i++) {
+	/* Some eDP panels only have the refresh rate range info in DisplayID */
+	if ((connector->display_info.monitor_range.min_vfreq == 0 ||
+	     connector->display_info.monitor_range.max_vfreq == 0))
+		parse_edid_displayid_vrr(connector, edid);
 
-				timing	= &edid->detailed_timings[i];
-				data	= &timing->data.other_data;
-				range	= &data->data.range;
-				/*
-				 * Check if monitor has continuous frequency mode
-				 */
-				if (data->type != EDID_DETAIL_MONITOR_RANGE)
-					continue;
-				/*
-				 * Check for flag range limits only. If flag == 1 then
-				 * no additional timing information provided.
-				 * Default GTF, GTF Secondary curve and CVT are not
-				 * supported
-				 */
-				if (range->flags != 1)
-					continue;
-
-				amdgpu_dm_connector->min_vfreq = range->min_vfreq;
-				amdgpu_dm_connector->max_vfreq = range->max_vfreq;
-				amdgpu_dm_connector->pixel_clock_mhz =
-					range->pixel_clock_mhz * 10;
-
-				connector->display_info.monitor_range.min_vfreq = range->min_vfreq;
-				connector->display_info.monitor_range.max_vfreq = range->max_vfreq;
-
-				break;
-			}
+	if (edid && (sink->sink_signal == SIGNAL_TYPE_DISPLAY_PORT ||
+		     sink->sink_signal == SIGNAL_TYPE_EDP)) {
+		if (amdgpu_dm_connector->dc_link &&
+		    amdgpu_dm_connector->dc_link->dpcd_caps.allow_invalid_MSA_timing_param) {
+			amdgpu_dm_connector->min_vfreq = connector->display_info.monitor_range.min_vfreq;
+			amdgpu_dm_connector->max_vfreq = connector->display_info.monitor_range.max_vfreq;
+			if (amdgpu_dm_connector->max_vfreq - amdgpu_dm_connector->min_vfreq > 10)
+				freesync_capable = true;
+		}
 
-			if (amdgpu_dm_connector->max_vfreq -
-			    amdgpu_dm_connector->min_vfreq > 10) {
+		parse_amd_vsdb(amdgpu_dm_connector, edid, &vsdb_info);
 
-				freesync_capable = true;
-			}
+		if (vsdb_info.replay_mode) {
+			amdgpu_dm_connector->vsdb_info.replay_mode = vsdb_info.replay_mode;
+			amdgpu_dm_connector->vsdb_info.amd_vsdb_version = vsdb_info.amd_vsdb_version;
+			amdgpu_dm_connector->as_type = ADAPTIVE_SYNC_TYPE_EDP;
 		}
-	} else if (edid && sink->sink_signal == SIGNAL_TYPE_HDMI_TYPE_A) {
+
+	} else if (drm_edid && sink->sink_signal == SIGNAL_TYPE_HDMI_TYPE_A) {
 		i = parse_hdmi_amd_vsdb(amdgpu_dm_connector, edid, &vsdb_info);
 		if (i >= 0 && vsdb_info.freesync_supported) {
-			timing  = &edid->detailed_timings[i];
-			data    = &timing->data.other_data;
-
 			amdgpu_dm_connector->min_vfreq = vsdb_info.min_refresh_rate_hz;
 			amdgpu_dm_connector->max_vfreq = vsdb_info.max_refresh_rate_hz;
 			if (amdgpu_dm_connector->max_vfreq - amdgpu_dm_connector->min_vfreq > 10)
@@ -10554,7 +13237,8 @@ void amdgpu_dm_update_freesync_caps(struct drm_connector *connector,
 		}
 	}
 
-	as_type = dm_get_adaptive_sync_support_type(amdgpu_dm_connector->dc_link);
+	if (amdgpu_dm_connector->dc_link)
+		as_type = dm_get_adaptive_sync_support_type(amdgpu_dm_connector->dc_link);
 
 	if (as_type == FREESYNC_TYPE_PCON_IN_WHITELIST) {
 		i = parse_hdmi_amd_vsdb(amdgpu_dm_connector, edid, &vsdb_info);
@@ -10578,6 +13262,12 @@ update:
 	if (dm_con_state)
 		dm_con_state->freesync_capable = freesync_capable;
 
+	if (connector->state && amdgpu_dm_connector->dc_link && !freesync_capable &&
+	    amdgpu_dm_connector->dc_link->replay_settings.config.replay_supported) {
+		amdgpu_dm_connector->dc_link->replay_settings.config.replay_supported = false;
+		amdgpu_dm_connector->dc_link->replay_settings.replay_feature_enabled = false;
+	}
+
 	if (connector->vrr_capable_property)
 		drm_connector_set_vrr_capable_property(connector,
 						       freesync_capable);
@@ -10602,15 +13292,24 @@ void amdgpu_dm_trigger_timing_sync(struct drm_device *dev)
 	mutex_unlock(&adev->dm.dc_lock);
 }
 
+static inline void amdgpu_dm_exit_ips_for_hw_access(struct dc *dc)
+{
+	if (dc->ctx->dmub_srv && !dc->ctx->dmub_srv->idle_exit_counter)
+		dc_exit_ips_for_hw_access(dc);
+}
+
 void dm_write_reg_func(const struct dc_context *ctx, uint32_t address,
 		       u32 value, const char *func_name)
 {
 #ifdef DM_CHECK_ADDR_0
 	if (address == 0) {
-		DC_ERR("invalid register write. address = 0");
+		drm_err(adev_to_drm(ctx->driver_context),
+			"invalid register write. address = 0");
 		return;
 	}
 #endif
+
+	amdgpu_dm_exit_ips_for_hw_access(ctx->dc);
 	cgs_write_register(ctx->cgs_device, address, value);
 	trace_amdgpu_dc_wreg(&ctx->perf_trace->write_count, address, value);
 }
@@ -10621,7 +13320,8 @@ uint32_t dm_read_reg_func(const struct dc_context *ctx, uint32_t address,
 	u32 value;
 #ifdef DM_CHECK_ADDR_0
 	if (address == 0) {
-		DC_ERR("invalid register read; address = 0\n");
+		drm_err(adev_to_drm(ctx->driver_context),
+			"invalid register read; address = 0\n");
 		return 0;
 	}
 #endif
@@ -10633,6 +13333,8 @@ uint32_t dm_read_reg_func(const struct dc_context *ctx, uint32_t address,
 		return 0;
 	}
 
+	amdgpu_dm_exit_ips_for_hw_access(ctx->dc);
+
 	value = cgs_read_register(ctx->cgs_device, address);
 
 	trace_amdgpu_dc_rreg(&ctx->perf_trace->read_count, address, value);
@@ -10657,7 +13359,7 @@ int amdgpu_dm_process_dmub_aux_transfer_sync(
 	}
 
 	if (!wait_for_completion_timeout(&adev->dm.dmub_aux_transfer_done, 10 * HZ)) {
-		DRM_ERROR("wait_for_completion_timeout timeout!");
+		drm_err(adev_to_drm(adev), "wait_for_completion_timeout timeout!");
 		*operation_result = AUX_RET_ERROR_TIMEOUT;
 		goto out;
 	}
@@ -10667,31 +13369,24 @@ int amdgpu_dm_process_dmub_aux_transfer_sync(
 		 * Transient states before tunneling is enabled could
 		 * lead to this error. We can ignore this for now.
 		 */
-		if (p_notify->result != AUX_RET_ERROR_PROTOCOL_ERROR) {
-			DRM_WARN("DPIA AUX failed on 0x%x(%d), error %d\n",
+		if (p_notify->result == AUX_RET_ERROR_PROTOCOL_ERROR) {
+			drm_warn(adev_to_drm(adev), "DPIA AUX failed on 0x%x(%d), error %d\n",
 					payload->address, payload->length,
 					p_notify->result);
 		}
-		*operation_result = AUX_RET_ERROR_INVALID_REPLY;
+		*operation_result = p_notify->result;
 		goto out;
 	}
 
+	payload->reply[0] = adev->dm.dmub_notify->aux_reply.command & 0xF;
+	if (adev->dm.dmub_notify->aux_reply.command & 0xF0)
+		/* The reply is stored in the top nibble of the command. */
+		payload->reply[0] = (adev->dm.dmub_notify->aux_reply.command >> 4) & 0xF;
 
-	payload->reply[0] = adev->dm.dmub_notify->aux_reply.command;
-	if (!payload->write && p_notify->aux_reply.length &&
-			(payload->reply[0] == AUX_TRANSACTION_REPLY_AUX_ACK)) {
-
-		if (payload->length != p_notify->aux_reply.length) {
-			DRM_WARN("invalid read length %d from DPIA AUX 0x%x(%d)!\n",
-				p_notify->aux_reply.length,
-					payload->address, payload->length);
-			*operation_result = AUX_RET_ERROR_INVALID_REPLY;
-			goto out;
-		}
-
+	/*write req may receive a byte indicating partially written number as well*/
+	if (p_notify->aux_reply.length)
 		memcpy(payload->data, p_notify->aux_reply.data,
 				p_notify->aux_reply.length);
-	}
 
 	/* success */
 	ret = p_notify->aux_reply.length;
@@ -10702,6 +13397,79 @@ out:
 	return ret;
 }
 
+static void abort_fused_io(
+		struct dc_context *ctx,
+		const struct dmub_cmd_fused_request *request
+)
+{
+	union dmub_rb_cmd command = { 0 };
+	struct dmub_rb_cmd_fused_io *io = &command.fused_io;
+
+	io->header.type = DMUB_CMD__FUSED_IO;
+	io->header.sub_type = DMUB_CMD__FUSED_IO_ABORT;
+	io->header.payload_bytes = sizeof(*io) - sizeof(io->header);
+	io->request = *request;
+	dm_execute_dmub_cmd(ctx, &command, DM_DMUB_WAIT_TYPE_NO_WAIT);
+}
+
+static bool execute_fused_io(
+		struct amdgpu_device *dev,
+		struct dc_context *ctx,
+		union dmub_rb_cmd *commands,
+		uint8_t count,
+		uint32_t timeout_us
+)
+{
+	const uint8_t ddc_line = commands[0].fused_io.request.u.aux.ddc_line;
+
+	if (ddc_line >= ARRAY_SIZE(dev->dm.fused_io))
+		return false;
+
+	struct fused_io_sync *sync = &dev->dm.fused_io[ddc_line];
+	struct dmub_rb_cmd_fused_io *first = &commands[0].fused_io;
+	const bool result = dm_execute_dmub_cmd_list(ctx, count, commands, DM_DMUB_WAIT_TYPE_WAIT_WITH_REPLY)
+			&& first->header.ret_status
+			&& first->request.status == FUSED_REQUEST_STATUS_SUCCESS;
+
+	if (!result)
+		return false;
+
+	while (wait_for_completion_timeout(&sync->replied, usecs_to_jiffies(timeout_us))) {
+		reinit_completion(&sync->replied);
+
+		struct dmub_cmd_fused_request *reply = (struct dmub_cmd_fused_request *) sync->reply_data;
+
+		static_assert(sizeof(*reply) <= sizeof(sync->reply_data), "Size mismatch");
+
+		if (reply->identifier == first->request.identifier) {
+			first->request = *reply;
+			return true;
+		}
+	}
+
+	reinit_completion(&sync->replied);
+	first->request.status = FUSED_REQUEST_STATUS_TIMEOUT;
+	abort_fused_io(ctx, &first->request);
+	return false;
+}
+
+bool amdgpu_dm_execute_fused_io(
+		struct amdgpu_device *dev,
+		struct dc_link *link,
+		union dmub_rb_cmd *commands,
+		uint8_t count,
+		uint32_t timeout_us)
+{
+	struct amdgpu_display_manager *dm = &dev->dm;
+
+	mutex_lock(&dm->dpia_aux_lock);
+
+	const bool result = execute_fused_io(dev, link->ctx, commands, count, timeout_us);
+
+	mutex_unlock(&dm->dpia_aux_lock);
+	return result;
+}
+
 int amdgpu_dm_process_dmub_set_config_sync(
 		struct dc_context *ctx,
 		unsigned int link_index,
@@ -10720,7 +13488,7 @@ int amdgpu_dm_process_dmub_set_config_sync(
 		ret = 0;
 		*operation_result = adev->dm.dmub_notify->sc_status;
 	} else {
-		DRM_ERROR("wait_for_completion_timeout timeout!");
+		drm_err(adev_to_drm(adev), "wait_for_completion_timeout timeout!");
 		ret = -1;
 		*operation_result = SET_CONFIG_UNKNOWN_ERROR;
 	}
@@ -10731,23 +13499,19 @@ int amdgpu_dm_process_dmub_set_config_sync(
 	return ret;
 }
 
-/*
- * Check whether seamless boot is supported.
- *
- * So far we only support seamless boot on CHIP_VANGOGH.
- * If everything goes well, we may consider expanding
- * seamless boot to other ASICs.
- */
-bool check_seamless_boot_capability(struct amdgpu_device *adev)
+bool dm_execute_dmub_cmd(const struct dc_context *ctx, union dmub_rb_cmd *cmd, enum dm_dmub_wait_type wait_type)
 {
-	switch (adev->ip_versions[DCE_HWIP][0]) {
-	case IP_VERSION(3, 0, 1):
-		if (!adev->mman.keep_stolen_vga_memory)
-			return true;
-		break;
-	default:
-		break;
-	}
+	return dc_dmub_srv_cmd_run(ctx->dmub_srv, cmd, wait_type);
+}
 
-	return false;
+bool dm_execute_dmub_cmd_list(const struct dc_context *ctx, unsigned int count, union dmub_rb_cmd *cmd, enum dm_dmub_wait_type wait_type)
+{
+	return dc_dmub_srv_cmd_run_list(ctx->dmub_srv, count, cmd, wait_type);
+}
+
+void dm_acpi_process_phy_transition_interlock(
+	const struct dc_context *ctx,
+	struct dm_process_phy_transition_init_params process_phy_transition_init_params)
+{
+	// Not yet implemented
 }
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h
index 2e2413fd73a4..ef97cede9926 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h
@@ -1,3 +1,4 @@
+/* SPDX-License-Identifier: MIT */
 /*
  * Copyright (C) 2015-2020 Advanced Micro Devices, Inc. All rights reserved.
  *
@@ -32,6 +33,7 @@
 #include <drm/drm_crtc.h>
 #include <drm/drm_plane.h>
 #include "link_service_types.h"
+#include <drm/drm_writeback.h>
 
 /*
  * This file contains the definition for amdgpu_display_manager
@@ -49,8 +51,15 @@
 
 #define AMDGPU_DM_MAX_NUM_EDP 2
 
-#define AMDGPU_DMUB_NOTIFICATION_MAX 5
+#define AMDGPU_DMUB_NOTIFICATION_MAX 8
 
+#define HDMI_AMD_VENDOR_SPECIFIC_DATA_BLOCK_IEEE_REGISTRATION_ID 0x00001A
+#define AMD_VSDB_VERSION_3_FEATURECAP_REPLAYMODE 0x40
+#define HDMI_AMD_VENDOR_SPECIFIC_DATA_BLOCK_VERSION_3 0x3
+
+#define AMDGPU_HDR_MULT_DEFAULT (0x100000000LL)
+
+#define AMDGPU_DM_HDMI_HPD_DEBOUNCE_MS 1500
 /*
 #include "include/amdgpu_dal_power_if.h"
 #include "amdgpu_dm_irq.h"
@@ -74,6 +83,13 @@ struct amdgpu_bo;
 struct dmub_srv;
 struct dc_plane_state;
 struct dmub_notification;
+struct dmub_cmd_fused_request;
+
+struct amd_vsdb_block {
+	unsigned char ieee_id[3];
+	unsigned char version;
+	unsigned char feature_caps;
+};
 
 struct common_irq_params {
 	struct amdgpu_device *adev;
@@ -125,6 +141,46 @@ struct vblank_control_work {
 };
 
 /**
+ * struct idle_workqueue - Work data for periodic action in idle
+ * @work: Kernel work data for the work event
+ * @dm: amdgpu display manager device
+ * @enable: true if idle worker is enabled
+ * @running: true if idle worker is running
+ */
+struct idle_workqueue {
+	struct work_struct work;
+	struct amdgpu_display_manager *dm;
+	bool enable;
+	bool running;
+};
+
+/**
+ * struct vupdate_offload_work - Work data for offloading task from vupdate handler
+ * @work: Kernel work data for the work event
+ * @adev: amdgpu_device back pointer
+ * @stream: DC stream associated with the crtc
+ * @adjust: DC CRTC timing adjust to be applied to the crtc
+ */
+struct vupdate_offload_work {
+	struct work_struct work;
+	struct amdgpu_device *adev;
+	struct dc_stream_state *stream;
+	struct dc_crtc_timing_adjust *adjust;
+};
+
+#define MAX_LUMINANCE_DATA_POINTS 99
+
+/**
+ * struct amdgpu_dm_luminance_data - Custom luminance data
+ * @luminance: Luminance in percent
+ * @input_signal: Input signal in range 0-255
+ */
+struct amdgpu_dm_luminance_data {
+	u8 luminance;
+	u8 input_signal;
+} __packed;
+
+/**
  * struct amdgpu_dm_backlight_caps - Information about backlight
  *
  * Describe the backlight support for ACPI or eDP AUX.
@@ -160,6 +216,27 @@ struct amdgpu_dm_backlight_caps {
 	 * @aux_support: Describes if the display supports AUX backlight.
 	 */
 	bool aux_support;
+	/**
+	 * @brightness_mask: After deriving brightness, OR it with this mask.
+	 * Workaround for panels with issues with certain brightness values.
+	 */
+	u32 brightness_mask;
+	/**
+	 * @ac_level: the default brightness if booted on AC
+	 */
+	u8 ac_level;
+	/**
+	 * @dc_level: the default brightness if booted on DC
+	 */
+	u8 dc_level;
+	/**
+	 * @data_points: the number of custom luminance data points
+	 */
+	u8 data_points;
+	/**
+	 * @luminance_data: custom luminance data
+	 */
+	struct amdgpu_dm_luminance_data luminance_data[MAX_LUMINANCE_DATA_POINTS];
 };
 
 /**
@@ -195,6 +272,11 @@ struct hpd_rx_irq_offload_work_queue {
 	 */
 	bool is_handling_link_loss;
 	/**
+	 * @is_handling_mst_msg_rdy_event: Used to prevent inserting mst message
+	 * ready event when we're already handling mst message ready event
+	 */
+	bool is_handling_mst_msg_rdy_event;
+	/**
 	 * @aconnector: The aconnector that this work queue is attached to
 	 */
 	struct amdgpu_dm_connector *aconnector;
@@ -216,6 +298,10 @@ struct hpd_rx_irq_offload_work {
 	 * @offload_wq: offload work queue that this work is queued to
 	 */
 	struct hpd_rx_irq_offload_work_queue *offload_wq;
+	/**
+	 * @adev: amdgpu_device pointer
+	 */
+	struct amdgpu_device *adev;
 };
 
 /**
@@ -470,6 +556,13 @@ struct amdgpu_display_manager {
 	 */
 	struct workqueue_struct *vblank_control_workqueue;
 
+	/**
+	 * @idle_workqueue:
+	 *
+	 * Periodic work for idle events.
+	 */
+	struct idle_workqueue *idle_workqueue;
+
 	struct drm_atomic_state *cached_state;
 	struct dc_state *cached_dc_state;
 
@@ -494,12 +587,12 @@ struct amdgpu_display_manager {
 
 #if defined(CONFIG_DRM_AMD_SECURE_DISPLAY)
 	/**
-	 * @secure_display_ctxs:
+	 * @secure_display_ctx:
 	 *
-	 * Store the ROI information and the work_struct to command dmub and psp for
-	 * all crtcs.
+	 * Store secure display relevant info. e.g. the ROI information
+	 * , the work_struct to command dmub, etc.
 	 */
-	struct secure_display_context *secure_display_ctxs;
+	struct secure_display_context secure_display_ctx;
 #endif
 	/**
 	 * @hpd_rx_offload_wq:
@@ -547,11 +640,43 @@ struct amdgpu_display_manager {
 	bool aux_hpd_discon_quirk;
 
 	/**
+	 * @edp0_on_dp1_quirk:
+	 *
+	 * quirk for platforms that put edp0 on DP1.
+	 */
+	bool edp0_on_dp1_quirk;
+
+	/**
 	 * @dpia_aux_lock:
 	 *
 	 * Guards access to DPIA AUX
 	 */
 	struct mutex dpia_aux_lock;
+
+	/**
+	 * @bb_from_dmub:
+	 *
+	 * Bounding box data read from dmub during early initialization for DCN4+
+	 * Data is stored as a byte array that should be casted to the appropriate bb struct
+	 */
+	void *bb_from_dmub;
+
+	/**
+	 * @oem_i2c:
+	 *
+	 * OEM i2c bus
+	 */
+	struct amdgpu_i2c_adapter *oem_i2c;
+
+	/**
+	 * @fused_io:
+	 *
+	 * dmub fused io interface
+	 */
+	struct fused_io_sync {
+		struct completion replied;
+		char reply_data[0x40];  // Cannot include dmub_cmd here
+	} fused_io[8];
 };
 
 enum dsc_clock_force_state {
@@ -604,6 +729,11 @@ struct amdgpu_hdmi_vsdb_info {
 	 * @max_refresh_rate_hz: FreeSync Maximum Refresh Rate in Hz
 	 */
 	unsigned int max_refresh_rate_hz;
+
+	/**
+	 * @replay_mode: Replay supported
+	 */
+	bool replay_mode;
 };
 
 struct amdgpu_dm_connector {
@@ -612,9 +742,11 @@ struct amdgpu_dm_connector {
 	uint32_t connector_id;
 	int bl_idx;
 
+	struct cec_notifier *notifier;
+
 	/* we need to mind the EDID between detect
 	   and get modes due to analog/digital/tvencoder */
-	struct edid *edid;
+	const struct drm_edid *drm_edid;
 
 	/* shared with amdgpu */
 	struct amdgpu_hpd hpd;
@@ -638,6 +770,13 @@ struct amdgpu_dm_connector {
 	struct drm_dp_mst_port *mst_output_port;
 	struct amdgpu_dm_connector *mst_root;
 	struct drm_dp_aux *dsc_aux;
+	uint32_t mst_local_bw;
+	uint16_t vc_full_pbn;
+	struct mutex handle_mst_msg_ready;
+
+	/* branch device specific data */
+	uint32_t branch_ieee_oui;
+
 	/* TODO see if we can merge with ddc_bus or make a dm_connector */
 	struct amdgpu_i2c_adapter *i2c;
 
@@ -653,7 +792,6 @@ struct amdgpu_dm_connector {
 	 * value is set to zero when there is no FreeSync support.
 	 */
 	int max_vfreq ;
-	int pixel_clock_mhz;
 
 	/* Audio instance - protected by audio_lock. */
 	int audio_inst;
@@ -661,17 +799,15 @@ struct amdgpu_dm_connector {
 	struct mutex hpd_lock;
 
 	bool fake_enable;
-#ifdef CONFIG_DEBUG_FS
-	uint32_t debugfs_dpcd_address;
-	uint32_t debugfs_dpcd_size;
-#endif
 	bool force_yuv420_output;
+	bool force_yuv422_output;
 	struct dsc_preferred_settings dsc_settings;
 	union dp_downstream_port_present mst_downstream_port_present;
 	/* Cached display modes */
 	struct drm_display_mode freesync_vid_base;
 
-	int psr_skip_count;
+	int sr_skip_count;
+	bool disallow_edp_enter_psr;
 
 	/* Record progress status of mst*/
 	uint8_t mst_status;
@@ -684,6 +820,11 @@ struct amdgpu_dm_connector {
 	bool pack_sdp_v1_3;
 	enum adaptive_sync_type as_type;
 	struct amdgpu_hdmi_vsdb_info vsdb_info;
+
+	/* HDMI HPD debounce support */
+	unsigned int hdmi_hpd_debounce_delay_ms;
+	struct delayed_work hdmi_hpd_debounce_work;
+	struct dc_sink *hdmi_prev_sink;
 };
 
 static inline void amdgpu_dm_set_mst_status(uint8_t *status,
@@ -697,11 +838,112 @@ static inline void amdgpu_dm_set_mst_status(uint8_t *status,
 
 #define to_amdgpu_dm_connector(x) container_of(x, struct amdgpu_dm_connector, base)
 
+struct amdgpu_dm_wb_connector {
+	struct drm_writeback_connector base;
+	struct dc_link *link;
+};
+
+#define to_amdgpu_dm_wb_connector(x) container_of(x, struct amdgpu_dm_wb_connector, base)
+
 extern const struct amdgpu_ip_block_version dm_ip_block;
 
+/* enum amdgpu_transfer_function: pre-defined transfer function supported by AMD.
+ *
+ * It includes standardized transfer functions and pure power functions. The
+ * transfer function coefficients are available at modules/color/color_gamma.c
+ */
+enum amdgpu_transfer_function {
+	AMDGPU_TRANSFER_FUNCTION_DEFAULT,
+	AMDGPU_TRANSFER_FUNCTION_SRGB_EOTF,
+	AMDGPU_TRANSFER_FUNCTION_BT709_INV_OETF,
+	AMDGPU_TRANSFER_FUNCTION_PQ_EOTF,
+	AMDGPU_TRANSFER_FUNCTION_IDENTITY,
+	AMDGPU_TRANSFER_FUNCTION_GAMMA22_EOTF,
+	AMDGPU_TRANSFER_FUNCTION_GAMMA24_EOTF,
+	AMDGPU_TRANSFER_FUNCTION_GAMMA26_EOTF,
+	AMDGPU_TRANSFER_FUNCTION_SRGB_INV_EOTF,
+	AMDGPU_TRANSFER_FUNCTION_BT709_OETF,
+	AMDGPU_TRANSFER_FUNCTION_PQ_INV_EOTF,
+	AMDGPU_TRANSFER_FUNCTION_GAMMA22_INV_EOTF,
+	AMDGPU_TRANSFER_FUNCTION_GAMMA24_INV_EOTF,
+	AMDGPU_TRANSFER_FUNCTION_GAMMA26_INV_EOTF,
+	AMDGPU_TRANSFER_FUNCTION_COUNT
+};
+
 struct dm_plane_state {
 	struct drm_plane_state base;
 	struct dc_plane_state *dc_state;
+
+	/* Plane color mgmt */
+	/**
+	 * @degamma_lut:
+	 *
+	 * 1D LUT for mapping framebuffer/plane pixel data before sampling or
+	 * blending operations. It's usually applied to linearize input space.
+	 * The blob (if not NULL) is an array of &struct drm_color_lut.
+	 */
+	struct drm_property_blob *degamma_lut;
+	/**
+	 * @degamma_tf:
+	 *
+	 * Predefined transfer function to tell DC driver the input space to
+	 * linearize.
+	 */
+	enum amdgpu_transfer_function degamma_tf;
+	/**
+	 * @hdr_mult:
+	 *
+	 * Multiplier to 'gain' the plane.  When PQ is decoded using the fixed
+	 * func transfer function to the internal FP16 fb, 1.0 -> 80 nits (on
+	 * AMD at least). When sRGB is decoded, 1.0 -> 1.0, obviously.
+	 * Therefore, 1.0 multiplier = 80 nits for SDR content.  So if you
+	 * want, 203 nits for SDR content, pass in (203.0 / 80.0).  Format is
+	 * S31.32 sign-magnitude.
+	 *
+	 * HDR multiplier can wide range beyond [0.0, 1.0]. This means that PQ
+	 * TF is needed for any subsequent linear-to-non-linear transforms.
+	 */
+	__u64 hdr_mult;
+	/**
+	 * @ctm:
+	 *
+	 * Color transformation matrix. The blob (if not NULL) is a &struct
+	 * drm_color_ctm_3x4.
+	 */
+	struct drm_property_blob *ctm;
+	/**
+	 * @shaper_lut: shaper lookup table blob. The blob (if not NULL) is an
+	 * array of &struct drm_color_lut.
+	 */
+	struct drm_property_blob *shaper_lut;
+	/**
+	 * @shaper_tf:
+	 *
+	 * Predefined transfer function to delinearize color space.
+	 */
+	enum amdgpu_transfer_function shaper_tf;
+	/**
+	 * @lut3d: 3D lookup table blob. The blob (if not NULL) is an array of
+	 * &struct drm_color_lut.
+	 */
+	struct drm_property_blob *lut3d;
+	/**
+	 * @blend_lut: blend lut lookup table blob. The blob (if not NULL) is an
+	 * array of &struct drm_color_lut.
+	 */
+	struct drm_property_blob *blend_lut;
+	/**
+	 * @blend_tf:
+	 *
+	 * Pre-defined transfer function for converting plane pixel data before
+	 * applying blend LUT.
+	 */
+	enum amdgpu_transfer_function blend_tf;
+};
+
+enum amdgpu_dm_cursor_mode {
+	DM_CURSOR_NATIVE_MODE = 0,
+	DM_CURSOR_OVERLAY_MODE,
 };
 
 struct dm_crtc_state {
@@ -726,6 +968,16 @@ struct dm_crtc_state {
 	struct dc_info_packet vrr_infopacket;
 
 	int abm_level;
+
+	/**
+	 * @regamma_tf:
+	 *
+	 * Pre-defined transfer function for converting internal FB -> wire
+	 * encoding.
+	 */
+	enum amdgpu_transfer_function regamma_tf;
+
+	enum amdgpu_dm_cursor_mode cursor_mode;
 };
 
 #define to_dm_crtc_state(x) container_of(x, struct dm_crtc_state, base)
@@ -747,6 +999,7 @@ struct dm_connector_state {
 	bool underscan_enable;
 	bool freesync_capable;
 	bool update_hdcp;
+	bool abm_sysfs_forbidden;
 	uint8_t abm_level;
 	int vcpi_slots;
 	uint64_t pbn;
@@ -777,24 +1030,34 @@ void amdgpu_dm_connector_init_helper(struct amdgpu_display_manager *dm,
 				     int link_index);
 
 enum drm_mode_status amdgpu_dm_connector_mode_valid(struct drm_connector *connector,
-				   struct drm_display_mode *mode);
+				   const struct drm_display_mode *mode);
 
 void dm_restore_drm_connector_state(struct drm_device *dev,
 				    struct drm_connector *connector);
 
 void amdgpu_dm_update_freesync_caps(struct drm_connector *connector,
-					struct edid *edid);
+				    const struct drm_edid *drm_edid);
 
 void amdgpu_dm_trigger_timing_sync(struct drm_device *dev);
 
+/* 3D LUT max size is 17x17x17 (4913 entries) */
+#define MAX_COLOR_3DLUT_SIZE 17
+#define MAX_COLOR_3DLUT_BITDEPTH 12
+int amdgpu_dm_verify_lut3d_size(struct amdgpu_device *adev,
+				struct drm_plane_state *plane_state);
+/* 1D LUT size */
 #define MAX_COLOR_LUT_ENTRIES 4096
 /* Legacy gamm LUT users such as X doesn't like large LUT sizes */
 #define MAX_COLOR_LEGACY_LUT_ENTRIES 256
 
 void amdgpu_dm_init_color_mod(void);
+int amdgpu_dm_create_color_properties(struct amdgpu_device *adev);
 int amdgpu_dm_verify_lut_sizes(const struct drm_crtc_state *crtc_state);
 int amdgpu_dm_update_crtc_color_mgmt(struct dm_crtc_state *crtc);
+int amdgpu_dm_check_crtc_color_mgmt(struct dm_crtc_state *crtc,
+				    bool check_only);
 int amdgpu_dm_update_plane_color_mgmt(struct dm_crtc_state *crtc,
+				      struct drm_plane_state *plane_state,
 				      struct dc_plane_state *dc_plane_state);
 
 void amdgpu_dm_update_connector_after_detect(
@@ -805,13 +1068,19 @@ extern const struct drm_encoder_helper_funcs amdgpu_dm_encoder_helper_funcs;
 int amdgpu_dm_process_dmub_aux_transfer_sync(struct dc_context *ctx, unsigned int link_index,
 					struct aux_payload *payload, enum aux_return_code_type *operation_result);
 
+bool amdgpu_dm_execute_fused_io(
+		struct amdgpu_device *dev,
+		struct dc_link *link,
+		union dmub_rb_cmd *commands,
+		uint8_t count,
+		uint32_t timeout_us
+);
+
 int amdgpu_dm_process_dmub_set_config_sync(struct dc_context *ctx, unsigned int link_index,
 					struct set_config_cmd_payload *payload, enum set_config_status *operation_result);
 
-bool check_seamless_boot_capability(struct amdgpu_device *adev);
-
 struct dc_stream_state *
-	create_validate_stream_for_sink(struct amdgpu_dm_connector *aconnector,
+	create_validate_stream_for_sink(struct drm_connector *connector,
 					const struct drm_display_mode *drm_mode,
 					const struct dm_connector_state *dm_state,
 					const struct dc_stream_state *old_stream);
@@ -819,9 +1088,27 @@ struct dc_stream_state *
 int dm_atomic_get_state(struct drm_atomic_state *state,
 			struct dm_atomic_state **dm_state);
 
-struct amdgpu_dm_connector *
+struct drm_connector *
 amdgpu_dm_find_first_crtc_matching_connector(struct drm_atomic_state *state,
 					     struct drm_crtc *crtc);
 
 int convert_dc_color_depth_into_bpc(enum dc_color_depth display_color_depth);
+struct idle_workqueue *idle_create_workqueue(struct amdgpu_device *adev);
+
+void *dm_allocate_gpu_mem(struct amdgpu_device *adev,
+						  enum dc_gpu_mem_alloc_type type,
+						  size_t size,
+						  long long *addr);
+void dm_free_gpu_mem(struct amdgpu_device *adev,
+						  enum dc_gpu_mem_alloc_type type,
+						  void *addr);
+
+bool amdgpu_dm_is_headless(struct amdgpu_device *adev);
+
+void hdmi_cec_set_edid(struct amdgpu_dm_connector *aconnector);
+void hdmi_cec_unset_edid(struct amdgpu_dm_connector *aconnector);
+int amdgpu_dm_initialize_hdmi_connector(struct amdgpu_dm_connector *aconnector);
+
+void retrieve_dmi_info(struct amdgpu_display_manager *dm);
+
 #endif /* __AMDGPU_DM_H__ */
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_color.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_color.c
index a4cb23d059bd..1dcc79b35225 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_color.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_color.c
@@ -1,3 +1,4 @@
+// SPDX-License-Identifier: MIT
 /*
  * Copyright 2018 Advanced Micro Devices, Inc.
  *
@@ -25,13 +26,39 @@
 #include "amdgpu.h"
 #include "amdgpu_mode.h"
 #include "amdgpu_dm.h"
+#include "amdgpu_dm_colorop.h"
 #include "dc.h"
 #include "modules/color/color_gamma.h"
-#include "basics/conversion.h"
 
 /**
  * DOC: overview
  *
+ * We have three types of color management in the AMD display driver.
+ * 1. the legacy &drm_crtc DEGAMMA, CTM, and GAMMA properties
+ * 2. AMD driver private color management on &drm_plane and &drm_crtc
+ * 3. AMD plane color pipeline
+ *
+ * The CRTC properties are the original color management. When they were
+ * implemented per-plane color management was not a thing yet. Because
+ * of that we could get away with plumbing the DEGAMMA and CTM
+ * properties to pre-blending HW functions. This is incompatible with
+ * per-plane color management, such as via the AMD private properties or
+ * the new drm_plane color pipeline. The only compatible CRTC property
+ * with per-plane color management is the GAMMA property as it is
+ * applied post-blending.
+ *
+ * The AMD driver private color management properties are only exposed
+ * when the kernel is built explicitly with -DAMD_PRIVATE_COLOR. They
+ * are temporary building blocks on the path to full-fledged &drm_plane
+ * and &drm_crtc color pipelines and lay the driver's groundwork for the
+ * color pipelines.
+ *
+ * The AMD plane color pipeline describes AMD's &drm_colorops via the
+ * &drm_plane's COLOR_PIPELINE property.
+ *
+ * drm_crtc Properties
+ * -------------------
+ *
  * The DC interface to HW gives us the following color management blocks
  * per pipe (surface):
  *
@@ -42,36 +69,94 @@
  * - Surface regamma LUT (normalized)
  * - Output CSC (normalized)
  *
- * But these aren't a direct mapping to DRM color properties. The current DRM
- * interface exposes CRTC degamma, CRTC CTM and CRTC regamma while our hardware
- * is essentially giving:
+ * But these aren't a direct mapping to DRM color properties. The
+ * current DRM interface exposes CRTC degamma, CRTC CTM and CRTC regamma
+ * while our hardware is essentially giving:
  *
  * Plane CTM -> Plane degamma -> Plane CTM -> Plane regamma -> Plane CTM
  *
- * The input gamma LUT block isn't really applicable here since it operates
- * on the actual input data itself rather than the HW fp representation. The
- * input and output CSC blocks are technically available to use as part of
- * the DC interface but are typically used internally by DC for conversions
- * between color spaces. These could be blended together with user
- * adjustments in the future but for now these should remain untouched.
+ * The input gamma LUT block isn't really applicable here since it
+ * operates on the actual input data itself rather than the HW fp
+ * representation. The input and output CSC blocks are technically
+ * available to use as part of the DC interface but are typically used
+ * internally by DC for conversions between color spaces. These could be
+ * blended together with user adjustments in the future but for now
+ * these should remain untouched.
+ *
+ * The pipe blending also happens after these blocks so we don't
+ * actually support any CRTC props with correct blending with multiple
+ * planes - but we can still support CRTC color management properties in
+ * DM in most single plane cases correctly with clever management of the
+ * DC interface in DM.
+ *
+ * As per DRM documentation, blocks should be in hardware bypass when
+ * their respective property is set to NULL. A linear DGM/RGM LUT should
+ * also considered as putting the respective block into bypass mode.
+ *
+ * This means that the following configuration is assumed to be the
+ * default:
+ *
+ * Plane DGM Bypass -> Plane CTM Bypass -> Plane RGM Bypass -> ... CRTC
+ * DGM Bypass -> CRTC CTM Bypass -> CRTC RGM Bypass
+ *
+ * AMD Private Color Management on drm_plane
+ * -----------------------------------------
+ *
+ * The AMD private color management properties on a &drm_plane are:
+ *
+ * - AMD_PLANE_DEGAMMA_LUT
+ * - AMD_PLANE_DEGAMMA_LUT_SIZE
+ * - AMD_PLANE_DEGAMMA_TF
+ * - AMD_PLANE_HDR_MULT
+ * - AMD_PLANE_CTM
+ * - AMD_PLANE_SHAPER_LUT
+ * - AMD_PLANE_SHAPER_LUT_SIZE
+ * - AMD_PLANE_SHAPER_TF
+ * - AMD_PLANE_LUT3D
+ * - AMD_PLANE_LUT3D_SIZE
+ * - AMD_PLANE_BLEND_LUT
+ * - AMD_PLANE_BLEND_LUT_SIZE
+ * - AMD_PLANE_BLEND_TF
+ *
+ * The AMD private color management property on a &drm_crtc is:
+ *
+ * - AMD_CRTC_REGAMMA_TF
+ *
+ * Use of these properties is discouraged.
+ *
+ * AMD plane color pipeline
+ * ------------------------
+ *
+ * The AMD &drm_plane color pipeline is advertised for DCN generations
+ * 3.0 and newer. It exposes these elements in this order:
+ *
+ * 1. 1D curve colorop
+ * 2. Multiplier
+ * 3. 3x4 CTM
+ * 4. 1D curve colorop
+ * 5. 1D LUT
+ * 6. 3D LUT
+ * 7. 1D curve colorop
+ * 8. 1D LUT
  *
- * The pipe blending also happens after these blocks so we don't actually
- * support any CRTC props with correct blending with multiple planes - but we
- * can still support CRTC color management properties in DM in most single
- * plane cases correctly with clever management of the DC interface in DM.
+ * The multiplier (#2) is a simple multiplier that is applied to all
+ * channels.
  *
- * As per DRM documentation, blocks should be in hardware bypass when their
- * respective property is set to NULL. A linear DGM/RGM LUT should also
- * considered as putting the respective block into bypass mode.
+ * The 3x4 CTM (#3) is a simple 3x4 matrix.
  *
- * This means that the following
- * configuration is assumed to be the default:
+ * #1, and #7 are non-linear to linear curves. #4 is a linear to
+ * non-linear curve. They support sRGB, PQ, and BT.709/BT.2020 EOTFs or
+ * their inverse.
+ *
+ * The 1D LUTs (#5 and #8) are plain 4096 entry LUTs.
+ *
+ * The 3DLUT (#6) is a tetrahedrally interpolated 17 cube LUT.
  *
- * Plane DGM Bypass -> Plane CTM Bypass -> Plane RGM Bypass -> ...
- * CRTC DGM Bypass -> CRTC CTM Bypass -> CRTC RGM Bypass
  */
 
 #define MAX_DRM_LUT_VALUE 0xFFFF
+#define MAX_DRM_LUT32_VALUE 0xFFFFFFFF
+#define SDR_WHITE_LEVEL_INIT_VALUE 80
 
 /**
  * amdgpu_dm_init_color_mod - Initialize the color module.
@@ -84,6 +169,247 @@ void amdgpu_dm_init_color_mod(void)
 	setup_x_points_distribution();
 }
 
+static inline struct fixed31_32 amdgpu_dm_fixpt_from_s3132(__u64 x)
+{
+	struct fixed31_32 val;
+
+	/* If negative, convert to 2's complement. */
+	if (x & (1ULL << 63))
+		x = -(x & ~(1ULL << 63));
+
+	val.value = x;
+	return val;
+}
+
+#ifdef AMD_PRIVATE_COLOR
+/* Pre-defined Transfer Functions (TF)
+ *
+ * AMD driver supports pre-defined mathematical functions for transferring
+ * between encoded values and optical/linear space. Depending on HW color caps,
+ * ROMs and curves built by the AMD color module support these transforms.
+ *
+ * The driver-specific color implementation exposes properties for pre-blending
+ * degamma TF, shaper TF (before 3D LUT), and blend(dpp.ogam) TF and
+ * post-blending regamma (mpc.ogam) TF. However, only pre-blending degamma
+ * supports ROM curves. AMD color module uses pre-defined coefficients to build
+ * curves for the other blocks. What can be done by each color block is
+ * described by struct dpp_color_capsand struct mpc_color_caps.
+ *
+ * AMD driver-specific color API exposes the following pre-defined transfer
+ * functions:
+ *
+ * - Identity: linear/identity relationship between pixel value and
+ *   luminance value;
+ * - Gamma 2.2, Gamma 2.4, Gamma 2.6: pure power functions;
+ * - sRGB: 2.4: The piece-wise transfer function from IEC 61966-2-1:1999;
+ * - BT.709: has a linear segment in the bottom part and then a power function
+ *   with a 0.45 (~1/2.22) gamma for the rest of the range; standardized by
+ *   ITU-R BT.709-6;
+ * - PQ (Perceptual Quantizer): used for HDR display, allows luminance range
+ *   capability of 0 to 10,000 nits; standardized by SMPTE ST 2084.
+ *
+ * The AMD color model is designed with an assumption that SDR (sRGB, BT.709,
+ * Gamma 2.2, etc.) peak white maps (normalized to 1.0 FP) to 80 nits in the PQ
+ * system. This has the implication that PQ EOTF (non-linear to linear) maps to
+ * [0.0..125.0] where 125.0 = 10,000 nits / 80 nits.
+ *
+ * Non-linear and linear forms are described in the table below:
+ *
+ * ┌───────────┬─────────────────────┬──────────────────────┐
+ * │           │     Non-linear      │   Linear             │
+ * ├───────────┼─────────────────────┼──────────────────────┤
+ * │      sRGB │ UNORM or [0.0, 1.0] │ [0.0, 1.0]           │
+ * ├───────────┼─────────────────────┼──────────────────────┤
+ * │     BT709 │ UNORM or [0.0, 1.0] │ [0.0, 1.0]           │
+ * ├───────────┼─────────────────────┼──────────────────────┤
+ * │ Gamma 2.x │ UNORM or [0.0, 1.0] │ [0.0, 1.0]           │
+ * ├───────────┼─────────────────────┼──────────────────────┤
+ * │        PQ │ UNORM or FP16 CCCS* │ [0.0, 125.0]         │
+ * ├───────────┼─────────────────────┼──────────────────────┤
+ * │  Identity │ UNORM or FP16 CCCS* │ [0.0, 1.0] or CCCS** │
+ * └───────────┴─────────────────────┴──────────────────────┘
+ * * CCCS: Windows canonical composition color space
+ * ** Respectively
+ *
+ * In the driver-specific API, color block names attached to TF properties
+ * suggest the intention regarding non-linear encoding pixel's luminance
+ * values. As some newer encodings don't use gamma curve, we make encoding and
+ * decoding explicit by defining an enum list of transfer functions supported
+ * in terms of EOTF and inverse EOTF, where:
+ *
+ * - EOTF (electro-optical transfer function): is the transfer function to go
+ *   from the encoded value to an optical (linear) value. De-gamma functions
+ *   traditionally do this.
+ * - Inverse EOTF (simply the inverse of the EOTF): is usually intended to go
+ *   from an optical/linear space (which might have been used for blending)
+ *   back to the encoded values. Gamma functions traditionally do this.
+ */
+static const char * const
+amdgpu_transfer_function_names[] = {
+	[AMDGPU_TRANSFER_FUNCTION_DEFAULT]		= "Default",
+	[AMDGPU_TRANSFER_FUNCTION_IDENTITY]		= "Identity",
+	[AMDGPU_TRANSFER_FUNCTION_SRGB_EOTF]		= "sRGB EOTF",
+	[AMDGPU_TRANSFER_FUNCTION_BT709_INV_OETF]	= "BT.709 inv_OETF",
+	[AMDGPU_TRANSFER_FUNCTION_PQ_EOTF]		= "PQ EOTF",
+	[AMDGPU_TRANSFER_FUNCTION_GAMMA22_EOTF]		= "Gamma 2.2 EOTF",
+	[AMDGPU_TRANSFER_FUNCTION_GAMMA24_EOTF]		= "Gamma 2.4 EOTF",
+	[AMDGPU_TRANSFER_FUNCTION_GAMMA26_EOTF]		= "Gamma 2.6 EOTF",
+	[AMDGPU_TRANSFER_FUNCTION_SRGB_INV_EOTF]	= "sRGB inv_EOTF",
+	[AMDGPU_TRANSFER_FUNCTION_BT709_OETF]		= "BT.709 OETF",
+	[AMDGPU_TRANSFER_FUNCTION_PQ_INV_EOTF]		= "PQ inv_EOTF",
+	[AMDGPU_TRANSFER_FUNCTION_GAMMA22_INV_EOTF]	= "Gamma 2.2 inv_EOTF",
+	[AMDGPU_TRANSFER_FUNCTION_GAMMA24_INV_EOTF]	= "Gamma 2.4 inv_EOTF",
+	[AMDGPU_TRANSFER_FUNCTION_GAMMA26_INV_EOTF]	= "Gamma 2.6 inv_EOTF",
+};
+
+static const u32 amdgpu_eotf =
+	BIT(AMDGPU_TRANSFER_FUNCTION_SRGB_EOTF) |
+	BIT(AMDGPU_TRANSFER_FUNCTION_BT709_INV_OETF) |
+	BIT(AMDGPU_TRANSFER_FUNCTION_PQ_EOTF) |
+	BIT(AMDGPU_TRANSFER_FUNCTION_GAMMA22_EOTF) |
+	BIT(AMDGPU_TRANSFER_FUNCTION_GAMMA24_EOTF) |
+	BIT(AMDGPU_TRANSFER_FUNCTION_GAMMA26_EOTF);
+
+static const u32 amdgpu_inv_eotf =
+	BIT(AMDGPU_TRANSFER_FUNCTION_SRGB_INV_EOTF) |
+	BIT(AMDGPU_TRANSFER_FUNCTION_BT709_OETF) |
+	BIT(AMDGPU_TRANSFER_FUNCTION_PQ_INV_EOTF) |
+	BIT(AMDGPU_TRANSFER_FUNCTION_GAMMA22_INV_EOTF) |
+	BIT(AMDGPU_TRANSFER_FUNCTION_GAMMA24_INV_EOTF) |
+	BIT(AMDGPU_TRANSFER_FUNCTION_GAMMA26_INV_EOTF);
+
+static struct drm_property *
+amdgpu_create_tf_property(struct drm_device *dev,
+			  const char *name,
+			  u32 supported_tf)
+{
+	u32 transfer_functions = supported_tf |
+				 BIT(AMDGPU_TRANSFER_FUNCTION_DEFAULT) |
+				 BIT(AMDGPU_TRANSFER_FUNCTION_IDENTITY);
+	struct drm_prop_enum_list enum_list[AMDGPU_TRANSFER_FUNCTION_COUNT];
+	int i, len;
+
+	len = 0;
+	for (i = 0; i < AMDGPU_TRANSFER_FUNCTION_COUNT; i++) {
+		if ((transfer_functions & BIT(i)) == 0)
+			continue;
+
+		enum_list[len].type = i;
+		enum_list[len].name = amdgpu_transfer_function_names[i];
+		len++;
+	}
+
+	return drm_property_create_enum(dev, DRM_MODE_PROP_ENUM,
+					name, enum_list, len);
+}
+
+int
+amdgpu_dm_create_color_properties(struct amdgpu_device *adev)
+{
+	struct drm_property *prop;
+
+	prop = drm_property_create(adev_to_drm(adev),
+				   DRM_MODE_PROP_BLOB,
+				   "AMD_PLANE_DEGAMMA_LUT", 0);
+	if (!prop)
+		return -ENOMEM;
+	adev->mode_info.plane_degamma_lut_property = prop;
+
+	prop = drm_property_create_range(adev_to_drm(adev),
+					 DRM_MODE_PROP_IMMUTABLE,
+					 "AMD_PLANE_DEGAMMA_LUT_SIZE",
+					 0, UINT_MAX);
+	if (!prop)
+		return -ENOMEM;
+	adev->mode_info.plane_degamma_lut_size_property = prop;
+
+	prop = amdgpu_create_tf_property(adev_to_drm(adev),
+					 "AMD_PLANE_DEGAMMA_TF",
+					 amdgpu_eotf);
+	if (!prop)
+		return -ENOMEM;
+	adev->mode_info.plane_degamma_tf_property = prop;
+
+	prop = drm_property_create_range(adev_to_drm(adev),
+					 0, "AMD_PLANE_HDR_MULT", 0, U64_MAX);
+	if (!prop)
+		return -ENOMEM;
+	adev->mode_info.plane_hdr_mult_property = prop;
+
+	prop = drm_property_create(adev_to_drm(adev),
+				   DRM_MODE_PROP_BLOB,
+				   "AMD_PLANE_CTM", 0);
+	if (!prop)
+		return -ENOMEM;
+	adev->mode_info.plane_ctm_property = prop;
+
+	prop = drm_property_create(adev_to_drm(adev),
+				   DRM_MODE_PROP_BLOB,
+				   "AMD_PLANE_SHAPER_LUT", 0);
+	if (!prop)
+		return -ENOMEM;
+	adev->mode_info.plane_shaper_lut_property = prop;
+
+	prop = drm_property_create_range(adev_to_drm(adev),
+					 DRM_MODE_PROP_IMMUTABLE,
+					 "AMD_PLANE_SHAPER_LUT_SIZE", 0, UINT_MAX);
+	if (!prop)
+		return -ENOMEM;
+	adev->mode_info.plane_shaper_lut_size_property = prop;
+
+	prop = amdgpu_create_tf_property(adev_to_drm(adev),
+					 "AMD_PLANE_SHAPER_TF",
+					 amdgpu_inv_eotf);
+	if (!prop)
+		return -ENOMEM;
+	adev->mode_info.plane_shaper_tf_property = prop;
+
+	prop = drm_property_create(adev_to_drm(adev),
+				   DRM_MODE_PROP_BLOB,
+				   "AMD_PLANE_LUT3D", 0);
+	if (!prop)
+		return -ENOMEM;
+	adev->mode_info.plane_lut3d_property = prop;
+
+	prop = drm_property_create_range(adev_to_drm(adev),
+					 DRM_MODE_PROP_IMMUTABLE,
+					 "AMD_PLANE_LUT3D_SIZE", 0, UINT_MAX);
+	if (!prop)
+		return -ENOMEM;
+	adev->mode_info.plane_lut3d_size_property = prop;
+
+	prop = drm_property_create(adev_to_drm(adev),
+				   DRM_MODE_PROP_BLOB,
+				   "AMD_PLANE_BLEND_LUT", 0);
+	if (!prop)
+		return -ENOMEM;
+	adev->mode_info.plane_blend_lut_property = prop;
+
+	prop = drm_property_create_range(adev_to_drm(adev),
+					 DRM_MODE_PROP_IMMUTABLE,
+					 "AMD_PLANE_BLEND_LUT_SIZE", 0, UINT_MAX);
+	if (!prop)
+		return -ENOMEM;
+	adev->mode_info.plane_blend_lut_size_property = prop;
+
+	prop = amdgpu_create_tf_property(adev_to_drm(adev),
+					 "AMD_PLANE_BLEND_TF",
+					 amdgpu_eotf);
+	if (!prop)
+		return -ENOMEM;
+	adev->mode_info.plane_blend_tf_property = prop;
+
+	prop = amdgpu_create_tf_property(adev_to_drm(adev),
+					 "AMD_CRTC_REGAMMA_TF",
+					 amdgpu_inv_eotf);
+	if (!prop)
+		return -ENOMEM;
+	adev->mode_info.regamma_tf_property = prop;
+
+	return 0;
+}
+#endif
+
 /**
  * __extract_blob_lut - Extracts the DRM lut and lut size from a blob.
  * @blob: DRM color mgmt property blob
@@ -100,6 +426,21 @@ __extract_blob_lut(const struct drm_property_blob *blob, uint32_t *size)
 }
 
 /**
+ * __extract_blob_lut32 - Extracts the DRM lut and lut size from a blob.
+ * @blob: DRM color mgmt property blob
+ * @size: lut size
+ *
+ * Returns:
+ * DRM LUT or NULL
+ */
+static const struct drm_color_lut32 *
+__extract_blob_lut32(const struct drm_property_blob *blob, uint32_t *size)
+{
+	*size = blob ? drm_color_lut32_size(blob) : 0;
+	return blob ? (struct drm_color_lut32 *)blob->data : NULL;
+}
+
+/**
  * __is_lut_linear - check if the given lut is a linear mapping of values
  * @lut: given lut to check values
  * @size: lut size
@@ -173,6 +514,24 @@ static void __drm_lut_to_dc_gamma(const struct drm_color_lut *lut,
 }
 
 /**
+ * __drm_lut32_to_dc_gamma - convert the drm_color_lut to dc_gamma.
+ * @lut: DRM lookup table for color conversion
+ * @gamma: DC gamma to set entries
+ *
+ * The conversion depends on the size of the lut - whether or not it's legacy.
+ */
+static void __drm_lut32_to_dc_gamma(const struct drm_color_lut32 *lut, struct dc_gamma *gamma)
+{
+	int i;
+
+	for (i = 0; i < MAX_COLOR_LUT_ENTRIES; i++) {
+		gamma->entries.red[i] = dc_fixpt_from_fraction(lut[i].red, MAX_DRM_LUT32_VALUE);
+		gamma->entries.green[i] = dc_fixpt_from_fraction(lut[i].green, MAX_DRM_LUT32_VALUE);
+		gamma->entries.blue[i] = dc_fixpt_from_fraction(lut[i].blue, MAX_DRM_LUT32_VALUE);
+	}
+}
+
+/**
  * __drm_ctm_to_dc_matrix - converts a DRM CTM to a DC CSC float matrix
  * @ctm: DRM color transformation matrix
  * @matrix: DC CSC float matrix
@@ -182,7 +541,6 @@ static void __drm_lut_to_dc_gamma(const struct drm_color_lut *lut,
 static void __drm_ctm_to_dc_matrix(const struct drm_color_ctm *ctm,
 				   struct fixed31_32 *matrix)
 {
-	int64_t val;
 	int i;
 
 	/*
@@ -201,12 +559,29 @@ static void __drm_ctm_to_dc_matrix(const struct drm_color_ctm *ctm,
 		}
 
 		/* gamut_remap_matrix[i] = ctm[i - floor(i/4)] */
-		val = ctm->matrix[i - (i / 4)];
-		/* If negative, convert to 2's complement. */
-		if (val & (1ULL << 63))
-			val = -(val & ~(1ULL << 63));
+		matrix[i] = amdgpu_dm_fixpt_from_s3132(ctm->matrix[i - (i / 4)]);
+	}
+}
+
+/**
+ * __drm_ctm_3x4_to_dc_matrix - converts a DRM CTM 3x4 to a DC CSC float matrix
+ * @ctm: DRM color transformation matrix with 3x4 dimensions
+ * @matrix: DC CSC float matrix
+ *
+ * The matrix needs to be a 3x4 (12 entry) matrix.
+ */
+static void __drm_ctm_3x4_to_dc_matrix(const struct drm_color_ctm_3x4 *ctm,
+				       struct fixed31_32 *matrix)
+{
+	int i;
 
-		matrix[i].value = val;
+	/* The format provided is S31.32, using signed-magnitude representation.
+	 * Our fixed31_32 is also S31.32, but is using 2's complement. We have
+	 * to convert from signed-magnitude to 2's complement.
+	 */
+	for (i = 0; i < 12; i++) {
+		/* gamut_remap_matrix[i] = ctm[i - floor(i/4)] */
+		matrix[i] = amdgpu_dm_fixpt_from_s3132(ctm->matrix[i]);
 	}
 }
 
@@ -268,16 +643,74 @@ static int __set_output_tf(struct dc_transfer_func *func,
 	struct calculate_buffer cal_buffer = {0};
 	bool res;
 
-	ASSERT(lut && lut_size == MAX_COLOR_LUT_ENTRIES);
+	cal_buffer.buffer_index = -1;
+
+	if (lut_size) {
+		ASSERT(lut && lut_size == MAX_COLOR_LUT_ENTRIES);
+
+		gamma = dc_create_gamma();
+		if (!gamma)
+			return -ENOMEM;
+
+		gamma->num_entries = lut_size;
+		__drm_lut_to_dc_gamma(lut, gamma, false);
+	}
+
+	if (func->tf == TRANSFER_FUNCTION_LINEAR) {
+		/*
+		 * Color module doesn't like calculating regamma params
+		 * on top of a linear input. But degamma params can be used
+		 * instead to simulate this.
+		 */
+		if (gamma)
+			gamma->type = GAMMA_CUSTOM;
+		res = mod_color_calculate_degamma_params(NULL, func,
+							 gamma, gamma != NULL);
+	} else {
+		/*
+		 * Assume sRGB. The actual mapping will depend on whether the
+		 * input was legacy or not.
+		 */
+		if (gamma)
+			gamma->type = GAMMA_CS_TFM_1D;
+		res = mod_color_calculate_regamma_params(func, gamma, gamma != NULL,
+							 has_rom, NULL, &cal_buffer);
+	}
+
+	if (gamma)
+		dc_gamma_release(&gamma);
+
+	return res ? 0 : -ENOMEM;
+}
+
+/**
+ * __set_output_tf_32 - calculates the output transfer function based on expected input space.
+ * @func: transfer function
+ * @lut: lookup table that defines the color space
+ * @lut_size: size of respective lut
+ * @has_rom: if ROM can be used for hardcoded curve
+ *
+ * Returns:
+ * 0 in case of success. -ENOMEM if fails.
+ */
+static int __set_output_tf_32(struct dc_transfer_func *func,
+			      const struct drm_color_lut32 *lut, uint32_t lut_size,
+			      bool has_rom)
+{
+	struct dc_gamma *gamma = NULL;
+	struct calculate_buffer cal_buffer = {0};
+	bool res;
 
 	cal_buffer.buffer_index = -1;
 
-	gamma = dc_create_gamma();
-	if (!gamma)
-		return -ENOMEM;
+	if (lut_size) {
+		gamma = dc_create_gamma();
+		if (!gamma)
+			return -ENOMEM;
 
-	gamma->num_entries = lut_size;
-	__drm_lut_to_dc_gamma(lut, gamma, false);
+		gamma->num_entries = lut_size;
+		__drm_lut32_to_dc_gamma(lut, gamma);
+	}
 
 	if (func->tf == TRANSFER_FUNCTION_LINEAR) {
 		/*
@@ -285,27 +718,68 @@ static int __set_output_tf(struct dc_transfer_func *func,
 		 * on top of a linear input. But degamma params can be used
 		 * instead to simulate this.
 		 */
-		gamma->type = GAMMA_CUSTOM;
+		if (gamma)
+			gamma->type = GAMMA_CUSTOM;
 		res = mod_color_calculate_degamma_params(NULL, func,
-							gamma, true);
+							 gamma, gamma != NULL);
 	} else {
 		/*
 		 * Assume sRGB. The actual mapping will depend on whether the
 		 * input was legacy or not.
 		 */
-		gamma->type = GAMMA_CS_TFM_1D;
-		res = mod_color_calculate_regamma_params(func, gamma, false,
+		if (gamma)
+			gamma->type = GAMMA_CS_TFM_1D;
+		res = mod_color_calculate_regamma_params(func, gamma, gamma != NULL,
 							 has_rom, NULL, &cal_buffer);
 	}
 
-	dc_gamma_release(&gamma);
+	if (gamma)
+		dc_gamma_release(&gamma);
 
 	return res ? 0 : -ENOMEM;
 }
 
+
+static int amdgpu_dm_set_atomic_regamma(struct dc_transfer_func *out_tf,
+					const struct drm_color_lut *regamma_lut,
+					uint32_t regamma_size, bool has_rom,
+					enum dc_transfer_func_predefined tf)
+{
+	int ret = 0;
+
+	if (regamma_size || tf != TRANSFER_FUNCTION_LINEAR) {
+		/*
+		 * CRTC RGM goes into RGM LUT.
+		 *
+		 * Note: there is no implicit sRGB regamma here. We are using
+		 * degamma calculation from color module to calculate the curve
+		 * from a linear base if gamma TF is not set. However, if gamma
+		 * TF (!= Linear) and LUT are set at the same time, we will use
+		 * regamma calculation, and the color module will combine the
+		 * pre-defined TF and the custom LUT values into the LUT that's
+		 * actually programmed.
+		 */
+		out_tf->type = TF_TYPE_DISTRIBUTED_POINTS;
+		out_tf->tf = tf;
+		out_tf->sdr_ref_white_level = SDR_WHITE_LEVEL_INIT_VALUE;
+
+		ret = __set_output_tf(out_tf, regamma_lut, regamma_size, has_rom);
+	} else {
+		/*
+		 * No CRTC RGM means we can just put the block into bypass
+		 * since we don't have any plane level adjustments using it.
+		 */
+		out_tf->type = TF_TYPE_BYPASS;
+		out_tf->tf = TRANSFER_FUNCTION_LINEAR;
+	}
+
+	return ret;
+}
+
 /**
  * __set_input_tf - calculates the input transfer function based on expected
  * input space.
+ * @caps: dc color capabilities
  * @func: transfer function
  * @lut: lookup table that defines the color space
  * @lut_size: size of respective lut.
@@ -313,27 +787,350 @@ static int __set_output_tf(struct dc_transfer_func *func,
  * Returns:
  * 0 in case of success. -ENOMEM if fails.
  */
-static int __set_input_tf(struct dc_transfer_func *func,
+static int __set_input_tf(struct dc_color_caps *caps, struct dc_transfer_func *func,
 			  const struct drm_color_lut *lut, uint32_t lut_size)
 {
 	struct dc_gamma *gamma = NULL;
 	bool res;
 
-	gamma = dc_create_gamma();
-	if (!gamma)
-		return -ENOMEM;
+	if (lut_size) {
+		gamma = dc_create_gamma();
+		if (!gamma)
+			return -ENOMEM;
 
-	gamma->type = GAMMA_CUSTOM;
-	gamma->num_entries = lut_size;
+		gamma->type = GAMMA_CUSTOM;
+		gamma->num_entries = lut_size;
 
-	__drm_lut_to_dc_gamma(lut, gamma, false);
+		__drm_lut_to_dc_gamma(lut, gamma, false);
+	}
 
-	res = mod_color_calculate_degamma_params(NULL, func, gamma, true);
-	dc_gamma_release(&gamma);
+	res = mod_color_calculate_degamma_params(caps, func, gamma, gamma != NULL);
+
+	if (gamma)
+		dc_gamma_release(&gamma);
+
+	return res ? 0 : -ENOMEM;
+}
+
+/**
+ * __set_input_tf_32 - calculates the input transfer function based on expected
+ * input space.
+ * @caps: dc color capabilities
+ * @func: transfer function
+ * @lut: lookup table that defines the color space
+ * @lut_size: size of respective lut.
+ *
+ * Returns:
+ * 0 in case of success. -ENOMEM if fails.
+ */
+static int __set_input_tf_32(struct dc_color_caps *caps, struct dc_transfer_func *func,
+			     const struct drm_color_lut32 *lut, uint32_t lut_size)
+{
+	struct dc_gamma *gamma = NULL;
+	bool res;
+
+	if (lut_size) {
+		gamma = dc_create_gamma();
+		if (!gamma)
+			return -ENOMEM;
+
+		gamma->type = GAMMA_CUSTOM;
+		gamma->num_entries = lut_size;
+
+		__drm_lut32_to_dc_gamma(lut, gamma);
+	}
+
+	res = mod_color_calculate_degamma_params(caps, func, gamma, gamma != NULL);
+
+	if (gamma)
+		dc_gamma_release(&gamma);
 
 	return res ? 0 : -ENOMEM;
 }
 
+static enum dc_transfer_func_predefined
+amdgpu_tf_to_dc_tf(enum amdgpu_transfer_function tf)
+{
+	switch (tf) {
+	default:
+	case AMDGPU_TRANSFER_FUNCTION_DEFAULT:
+	case AMDGPU_TRANSFER_FUNCTION_IDENTITY:
+		return TRANSFER_FUNCTION_LINEAR;
+	case AMDGPU_TRANSFER_FUNCTION_SRGB_EOTF:
+	case AMDGPU_TRANSFER_FUNCTION_SRGB_INV_EOTF:
+		return TRANSFER_FUNCTION_SRGB;
+	case AMDGPU_TRANSFER_FUNCTION_BT709_OETF:
+	case AMDGPU_TRANSFER_FUNCTION_BT709_INV_OETF:
+		return TRANSFER_FUNCTION_BT709;
+	case AMDGPU_TRANSFER_FUNCTION_PQ_EOTF:
+	case AMDGPU_TRANSFER_FUNCTION_PQ_INV_EOTF:
+		return TRANSFER_FUNCTION_PQ;
+	case AMDGPU_TRANSFER_FUNCTION_GAMMA22_EOTF:
+	case AMDGPU_TRANSFER_FUNCTION_GAMMA22_INV_EOTF:
+		return TRANSFER_FUNCTION_GAMMA22;
+	case AMDGPU_TRANSFER_FUNCTION_GAMMA24_EOTF:
+	case AMDGPU_TRANSFER_FUNCTION_GAMMA24_INV_EOTF:
+		return TRANSFER_FUNCTION_GAMMA24;
+	case AMDGPU_TRANSFER_FUNCTION_GAMMA26_EOTF:
+	case AMDGPU_TRANSFER_FUNCTION_GAMMA26_INV_EOTF:
+		return TRANSFER_FUNCTION_GAMMA26;
+	}
+}
+
+static enum dc_transfer_func_predefined
+amdgpu_colorop_tf_to_dc_tf(enum drm_colorop_curve_1d_type tf)
+{
+	switch (tf) {
+	case DRM_COLOROP_1D_CURVE_SRGB_EOTF:
+	case DRM_COLOROP_1D_CURVE_SRGB_INV_EOTF:
+		return TRANSFER_FUNCTION_SRGB;
+	case DRM_COLOROP_1D_CURVE_PQ_125_EOTF:
+	case DRM_COLOROP_1D_CURVE_PQ_125_INV_EOTF:
+		return TRANSFER_FUNCTION_PQ;
+	case DRM_COLOROP_1D_CURVE_BT2020_INV_OETF:
+	case DRM_COLOROP_1D_CURVE_BT2020_OETF:
+		return TRANSFER_FUNCTION_BT709;
+	case DRM_COLOROP_1D_CURVE_GAMMA22:
+	case DRM_COLOROP_1D_CURVE_GAMMA22_INV:
+		return TRANSFER_FUNCTION_GAMMA22;
+	default:
+		return TRANSFER_FUNCTION_LINEAR;
+	}
+}
+
+static void __to_dc_lut3d_color(struct dc_rgb *rgb,
+				const struct drm_color_lut lut,
+				int bit_precision)
+{
+	rgb->red = drm_color_lut_extract(lut.red, bit_precision);
+	rgb->green = drm_color_lut_extract(lut.green, bit_precision);
+	rgb->blue  = drm_color_lut_extract(lut.blue, bit_precision);
+}
+
+static void __drm_3dlut_to_dc_3dlut(const struct drm_color_lut *lut,
+				    uint32_t lut3d_size,
+				    struct tetrahedral_params *params,
+				    bool use_tetrahedral_9,
+				    int bit_depth)
+{
+	struct dc_rgb *lut0;
+	struct dc_rgb *lut1;
+	struct dc_rgb *lut2;
+	struct dc_rgb *lut3;
+	int lut_i, i;
+
+
+	if (use_tetrahedral_9) {
+		lut0 = params->tetrahedral_9.lut0;
+		lut1 = params->tetrahedral_9.lut1;
+		lut2 = params->tetrahedral_9.lut2;
+		lut3 = params->tetrahedral_9.lut3;
+	} else {
+		lut0 = params->tetrahedral_17.lut0;
+		lut1 = params->tetrahedral_17.lut1;
+		lut2 = params->tetrahedral_17.lut2;
+		lut3 = params->tetrahedral_17.lut3;
+	}
+
+	for (lut_i = 0, i = 0; i < lut3d_size - 4; lut_i++, i += 4) {
+		/*
+		 * We should consider the 3D LUT RGB values are distributed
+		 * along four arrays lut0-3 where the first sizes 1229 and the
+		 * other 1228. The bit depth supported for 3dlut channel is
+		 * 12-bit, but DC also supports 10-bit.
+		 *
+		 * TODO: improve color pipeline API to enable the userspace set
+		 * bit depth and 3D LUT size/stride, as specified by VA-API.
+		 */
+		__to_dc_lut3d_color(&lut0[lut_i], lut[i], bit_depth);
+		__to_dc_lut3d_color(&lut1[lut_i], lut[i + 1], bit_depth);
+		__to_dc_lut3d_color(&lut2[lut_i], lut[i + 2], bit_depth);
+		__to_dc_lut3d_color(&lut3[lut_i], lut[i + 3], bit_depth);
+	}
+	/* lut0 has 1229 points (lut_size/4 + 1) */
+	__to_dc_lut3d_color(&lut0[lut_i], lut[i], bit_depth);
+}
+
+static void __to_dc_lut3d_32_color(struct dc_rgb *rgb,
+				   const struct drm_color_lut32 lut,
+				   int bit_precision)
+{
+	rgb->red = drm_color_lut32_extract(lut.red, bit_precision);
+	rgb->green = drm_color_lut32_extract(lut.green, bit_precision);
+	rgb->blue  = drm_color_lut32_extract(lut.blue, bit_precision);
+}
+
+static void __drm_3dlut32_to_dc_3dlut(const struct drm_color_lut32 *lut,
+				       uint32_t lut3d_size,
+				       struct tetrahedral_params *params,
+				       bool use_tetrahedral_9,
+				       int bit_depth)
+{
+	struct dc_rgb *lut0;
+	struct dc_rgb *lut1;
+	struct dc_rgb *lut2;
+	struct dc_rgb *lut3;
+	int lut_i, i;
+
+
+	if (use_tetrahedral_9) {
+		lut0 = params->tetrahedral_9.lut0;
+		lut1 = params->tetrahedral_9.lut1;
+		lut2 = params->tetrahedral_9.lut2;
+		lut3 = params->tetrahedral_9.lut3;
+	} else {
+		lut0 = params->tetrahedral_17.lut0;
+		lut1 = params->tetrahedral_17.lut1;
+		lut2 = params->tetrahedral_17.lut2;
+		lut3 = params->tetrahedral_17.lut3;
+	}
+
+	for (lut_i = 0, i = 0; i < lut3d_size - 4; lut_i++, i += 4) {
+		/*
+		 * We should consider the 3D LUT RGB values are distributed
+		 * along four arrays lut0-3 where the first sizes 1229 and the
+		 * other 1228. The bit depth supported for 3dlut channel is
+		 * 12-bit, but DC also supports 10-bit.
+		 *
+		 * TODO: improve color pipeline API to enable the userspace set
+		 * bit depth and 3D LUT size/stride, as specified by VA-API.
+		 */
+		__to_dc_lut3d_32_color(&lut0[lut_i], lut[i], bit_depth);
+		__to_dc_lut3d_32_color(&lut1[lut_i], lut[i + 1], bit_depth);
+		__to_dc_lut3d_32_color(&lut2[lut_i], lut[i + 2], bit_depth);
+		__to_dc_lut3d_32_color(&lut3[lut_i], lut[i + 3], bit_depth);
+	}
+	/* lut0 has 1229 points (lut_size/4 + 1) */
+	__to_dc_lut3d_32_color(&lut0[lut_i], lut[i], bit_depth);
+}
+
+/* amdgpu_dm_atomic_lut3d - set DRM 3D LUT to DC stream
+ * @drm_lut3d: user 3D LUT
+ * @drm_lut3d_size: size of 3D LUT
+ * @lut3d: DC 3D LUT
+ *
+ * Map user 3D LUT data to DC 3D LUT and all necessary bits to program it
+ * on DCN accordingly.
+ */
+static void amdgpu_dm_atomic_lut3d(const struct drm_color_lut *drm_lut3d,
+				   uint32_t drm_lut3d_size,
+				   struct dc_3dlut *lut)
+{
+	if (!drm_lut3d_size) {
+		lut->state.bits.initialized = 0;
+	} else {
+		/* Stride and bit depth are not programmable by API yet.
+		 * Therefore, only supports 17x17x17 3D LUT (12-bit).
+		 */
+		lut->lut_3d.use_tetrahedral_9 = false;
+		lut->lut_3d.use_12bits = true;
+		lut->state.bits.initialized = 1;
+		__drm_3dlut_to_dc_3dlut(drm_lut3d, drm_lut3d_size, &lut->lut_3d,
+					lut->lut_3d.use_tetrahedral_9,
+					MAX_COLOR_3DLUT_BITDEPTH);
+	}
+}
+
+static int amdgpu_dm_atomic_shaper_lut(const struct drm_color_lut *shaper_lut,
+				       bool has_rom,
+				       enum dc_transfer_func_predefined tf,
+				       uint32_t shaper_size,
+				       struct dc_transfer_func *func_shaper)
+{
+	int ret = 0;
+
+	if (shaper_size || tf != TRANSFER_FUNCTION_LINEAR) {
+		/*
+		 * If user shaper LUT is set, we assume a linear color space
+		 * (linearized by degamma 1D LUT or not).
+		 */
+		func_shaper->type = TF_TYPE_DISTRIBUTED_POINTS;
+		func_shaper->tf = tf;
+		func_shaper->sdr_ref_white_level = SDR_WHITE_LEVEL_INIT_VALUE;
+
+		ret = __set_output_tf(func_shaper, shaper_lut, shaper_size, has_rom);
+	} else {
+		func_shaper->type = TF_TYPE_BYPASS;
+		func_shaper->tf = TRANSFER_FUNCTION_LINEAR;
+	}
+
+	return ret;
+}
+
+static int amdgpu_dm_atomic_blend_lut(const struct drm_color_lut *blend_lut,
+				       bool has_rom,
+				       enum dc_transfer_func_predefined tf,
+				       uint32_t blend_size,
+				       struct dc_transfer_func *func_blend)
+{
+	int ret = 0;
+
+	if (blend_size || tf != TRANSFER_FUNCTION_LINEAR) {
+		/*
+		 * DRM plane gamma LUT or TF means we are linearizing color
+		 * space before blending (similar to degamma programming). As
+		 * we don't have hardcoded curve support, or we use AMD color
+		 * module to fill the parameters that will be translated to HW
+		 * points.
+		 */
+		func_blend->type = TF_TYPE_DISTRIBUTED_POINTS;
+		func_blend->tf = tf;
+		func_blend->sdr_ref_white_level = SDR_WHITE_LEVEL_INIT_VALUE;
+
+		ret = __set_input_tf(NULL, func_blend, blend_lut, blend_size);
+	} else {
+		func_blend->type = TF_TYPE_BYPASS;
+		func_blend->tf = TRANSFER_FUNCTION_LINEAR;
+	}
+
+	return ret;
+}
+
+/**
+ * amdgpu_dm_verify_lut3d_size - verifies if 3D LUT is supported and if user
+ * shaper and 3D LUTs match the hw supported size
+ * @adev: amdgpu device
+ * @plane_state: the DRM plane state
+ *
+ * Verifies if pre-blending (DPP) 3D LUT is supported by the HW (DCN 2.0 or
+ * newer) and if the user shaper and 3D LUTs match the supported size.
+ *
+ * Returns:
+ * 0 on success. -EINVAL if lut size are invalid.
+ */
+int amdgpu_dm_verify_lut3d_size(struct amdgpu_device *adev,
+				struct drm_plane_state *plane_state)
+{
+	struct dm_plane_state *dm_plane_state = to_dm_plane_state(plane_state);
+	const struct drm_color_lut *shaper = NULL, *lut3d = NULL;
+	uint32_t exp_size, size, dim_size = MAX_COLOR_3DLUT_SIZE;
+	bool has_3dlut = adev->dm.dc->caps.color.dpp.hw_3d_lut || adev->dm.dc->caps.color.mpc.preblend;
+
+	/* shaper LUT is only available if 3D LUT color caps */
+	exp_size = has_3dlut ? MAX_COLOR_LUT_ENTRIES : 0;
+	shaper = __extract_blob_lut(dm_plane_state->shaper_lut, &size);
+
+	if (shaper && size != exp_size) {
+		drm_dbg(&adev->ddev,
+			"Invalid Shaper LUT size. Should be %u but got %u.\n",
+			exp_size, size);
+		return -EINVAL;
+	}
+
+	/* The number of 3D LUT entries is the dimension size cubed */
+	exp_size = has_3dlut ? dim_size * dim_size * dim_size : 0;
+	lut3d = __extract_blob_lut(dm_plane_state->lut3d, &size);
+
+	if (lut3d && size != exp_size) {
+		drm_dbg(&adev->ddev,
+			"Invalid 3D LUT size. Should be %u but got %u.\n",
+			exp_size, size);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
 /**
  * amdgpu_dm_verify_lut_sizes - verifies if DRM luts match the hw supported sizes
  * @crtc_state: the DRM CRTC state
@@ -371,39 +1168,42 @@ int amdgpu_dm_verify_lut_sizes(const struct drm_crtc_state *crtc_state)
 }
 
 /**
- * amdgpu_dm_update_crtc_color_mgmt: Maps DRM color management to DC stream.
+ * amdgpu_dm_check_crtc_color_mgmt: Check if DRM color props are programmable by DC.
  * @crtc: amdgpu_dm crtc state
+ * @check_only: only check color state without update dc stream
  *
- * With no plane level color management properties we're free to use any
- * of the HW blocks as long as the CRTC CTM always comes before the
- * CRTC RGM and after the CRTC DGM.
- *
- * - The CRTC RGM block will be placed in the RGM LUT block if it is non-linear.
- * - The CRTC DGM block will be placed in the DGM LUT block if it is non-linear.
- * - The CRTC CTM will be placed in the gamut remap block if it is non-linear.
+ * This function just verifies CRTC LUT sizes, if there is enough space for
+ * output transfer function and if its parameters can be calculated by AMD
+ * color module. It also adjusts some settings for programming CRTC degamma at
+ * plane stage, using plane DGM block.
  *
  * The RGM block is typically more fully featured and accurate across
  * all ASICs - DCE can't support a custom non-linear CRTC DGM.
  *
  * For supporting both plane level color management and CRTC level color
- * management at once we have to either restrict the usage of CRTC properties
- * or blend adjustments together.
+ * management at once we have to either restrict the usage of some CRTC
+ * properties or blend adjustments together.
  *
  * Returns:
- * 0 on success. Error code if setup fails.
+ * 0 on success. Error code if validation fails.
  */
-int amdgpu_dm_update_crtc_color_mgmt(struct dm_crtc_state *crtc)
+
+int amdgpu_dm_check_crtc_color_mgmt(struct dm_crtc_state *crtc,
+				    bool check_only)
 {
 	struct dc_stream_state *stream = crtc->stream;
 	struct amdgpu_device *adev = drm_to_adev(crtc->base.state->dev);
 	bool has_rom = adev->asic_type <= CHIP_RAVEN;
-	struct drm_color_ctm *ctm = NULL;
+	struct dc_transfer_func *out_tf;
 	const struct drm_color_lut *degamma_lut, *regamma_lut;
 	uint32_t degamma_size, regamma_size;
 	bool has_regamma, has_degamma;
+	enum dc_transfer_func_predefined tf = TRANSFER_FUNCTION_LINEAR;
 	bool is_legacy;
 	int r;
 
+	tf = amdgpu_tf_to_dc_tf(crtc->regamma_tf);
+
 	r = amdgpu_dm_verify_lut_sizes(&crtc->base);
 	if (r)
 		return r;
@@ -423,6 +1223,14 @@ int amdgpu_dm_update_crtc_color_mgmt(struct dm_crtc_state *crtc)
 	crtc->cm_has_degamma = false;
 	crtc->cm_is_degamma_srgb = false;
 
+	if (check_only) {
+		out_tf = kvzalloc(sizeof(*out_tf), GFP_KERNEL);
+		if (!out_tf)
+			return -ENOMEM;
+	} else {
+		out_tf = &stream->out_transfer_func;
+	}
+
 	/* Setup regamma and degamma. */
 	if (is_legacy) {
 		/*
@@ -437,29 +1245,21 @@ int amdgpu_dm_update_crtc_color_mgmt(struct dm_crtc_state *crtc)
 		 * inverse color ramp in legacy userspace.
 		 */
 		crtc->cm_is_degamma_srgb = true;
-		stream->out_transfer_func->type = TF_TYPE_DISTRIBUTED_POINTS;
-		stream->out_transfer_func->tf = TRANSFER_FUNCTION_SRGB;
-
-		r = __set_legacy_tf(stream->out_transfer_func, regamma_lut,
-				    regamma_size, has_rom);
-		if (r)
-			return r;
-	} else if (has_regamma) {
-		/* If atomic regamma, CRTC RGM goes into RGM LUT. */
-		stream->out_transfer_func->type = TF_TYPE_DISTRIBUTED_POINTS;
-		stream->out_transfer_func->tf = TRANSFER_FUNCTION_LINEAR;
-
-		r = __set_output_tf(stream->out_transfer_func, regamma_lut,
-				    regamma_size, has_rom);
-		if (r)
-			return r;
-	} else {
+		out_tf->type = TF_TYPE_DISTRIBUTED_POINTS;
+		out_tf->tf = TRANSFER_FUNCTION_SRGB;
 		/*
-		 * No CRTC RGM means we can just put the block into bypass
-		 * since we don't have any plane level adjustments using it.
+		 * Note: although we pass has_rom as parameter here, we never
+		 * actually use ROM because the color module only takes the ROM
+		 * path if transfer_func->type == PREDEFINED.
+		 *
+		 * See more in mod_color_calculate_regamma_params()
 		 */
-		stream->out_transfer_func->type = TF_TYPE_BYPASS;
-		stream->out_transfer_func->tf = TRANSFER_FUNCTION_LINEAR;
+		r = __set_legacy_tf(out_tf, regamma_lut,
+				    regamma_size, has_rom);
+	} else {
+		regamma_size = has_regamma ? regamma_size : 0;
+		r = amdgpu_dm_set_atomic_regamma(out_tf, regamma_lut,
+						 regamma_size, has_rom, tf);
 	}
 
 	/*
@@ -468,6 +1268,43 @@ int amdgpu_dm_update_crtc_color_mgmt(struct dm_crtc_state *crtc)
 	 * have to place the CTM in the OCSC in that case.
 	 */
 	crtc->cm_has_degamma = has_degamma;
+	if (check_only)
+		kvfree(out_tf);
+
+	return r;
+}
+
+/**
+ * amdgpu_dm_update_crtc_color_mgmt: Maps DRM color management to DC stream.
+ * @crtc: amdgpu_dm crtc state
+ *
+ * With no plane level color management properties we're free to use any
+ * of the HW blocks as long as the CRTC CTM always comes before the
+ * CRTC RGM and after the CRTC DGM.
+ *
+ * - The CRTC RGM block will be placed in the RGM LUT block if it is non-linear.
+ * - The CRTC DGM block will be placed in the DGM LUT block if it is non-linear.
+ * - The CRTC CTM will be placed in the gamut remap block if it is non-linear.
+ *
+ * The RGM block is typically more fully featured and accurate across
+ * all ASICs - DCE can't support a custom non-linear CRTC DGM.
+ *
+ * For supporting both plane level color management and CRTC level color
+ * management at once we have to either restrict the usage of CRTC properties
+ * or blend adjustments together.
+ *
+ * Returns:
+ * 0 on success. Error code if setup fails.
+ */
+int amdgpu_dm_update_crtc_color_mgmt(struct dm_crtc_state *crtc)
+{
+	struct dc_stream_state *stream = crtc->stream;
+	struct drm_color_ctm *ctm = NULL;
+	int ret;
+
+	ret = amdgpu_dm_check_crtc_color_mgmt(crtc, false);
+	if (ret)
+		return ret;
 
 	/* Setup CRTC CTM. */
 	if (crtc->base.ctm) {
@@ -495,20 +1332,10 @@ int amdgpu_dm_update_crtc_color_mgmt(struct dm_crtc_state *crtc)
 	return 0;
 }
 
-/**
- * amdgpu_dm_update_plane_color_mgmt: Maps DRM color management to DC plane.
- * @crtc: amdgpu_dm crtc state
- * @dc_plane_state: target DC surface
- *
- * Update the underlying dc_stream_state's input transfer function (ITF) in
- * preparation for hardware commit. The transfer function used depends on
- * the preparation done on the stream for color management.
- *
- * Returns:
- * 0 on success. -ENOMEM if mem allocation fails.
- */
-int amdgpu_dm_update_plane_color_mgmt(struct dm_crtc_state *crtc,
-				      struct dc_plane_state *dc_plane_state)
+static int
+map_crtc_degamma_to_dc_plane(struct dm_crtc_state *crtc,
+			     struct dc_plane_state *dc_plane_state,
+			     struct dc_color_caps *caps)
 {
 	const struct drm_color_lut *degamma_lut;
 	enum dc_transfer_func_predefined tf = TRANSFER_FUNCTION_SRGB;
@@ -531,8 +1358,7 @@ int amdgpu_dm_update_plane_color_mgmt(struct dm_crtc_state *crtc,
 						 &degamma_size);
 		ASSERT(degamma_size == MAX_COLOR_LUT_ENTRIES);
 
-		dc_plane_state->in_transfer_func->type =
-			TF_TYPE_DISTRIBUTED_POINTS;
+		dc_plane_state->in_transfer_func.type = TF_TYPE_DISTRIBUTED_POINTS;
 
 		/*
 		 * This case isn't fully correct, but also fairly
@@ -559,32 +1385,671 @@ int amdgpu_dm_update_plane_color_mgmt(struct dm_crtc_state *crtc,
 		 * map these to the atomic one instead.
 		 */
 		if (crtc->cm_is_degamma_srgb)
-			dc_plane_state->in_transfer_func->tf = tf;
+			dc_plane_state->in_transfer_func.tf = tf;
 		else
-			dc_plane_state->in_transfer_func->tf =
+			dc_plane_state->in_transfer_func.tf =
 				TRANSFER_FUNCTION_LINEAR;
 
-		r = __set_input_tf(dc_plane_state->in_transfer_func,
+		r = __set_input_tf(caps, &dc_plane_state->in_transfer_func,
 				   degamma_lut, degamma_size);
 		if (r)
 			return r;
-	} else if (crtc->cm_is_degamma_srgb) {
+	} else {
 		/*
 		 * For legacy gamma support we need the regamma input
 		 * in linear space. Assume that the input is sRGB.
 		 */
-		dc_plane_state->in_transfer_func->type = TF_TYPE_PREDEFINED;
-		dc_plane_state->in_transfer_func->tf = tf;
+		dc_plane_state->in_transfer_func.type = TF_TYPE_PREDEFINED;
+		dc_plane_state->in_transfer_func.tf = tf;
 
 		if (tf != TRANSFER_FUNCTION_SRGB &&
-		    !mod_color_calculate_degamma_params(NULL,
-			    dc_plane_state->in_transfer_func, NULL, false))
+		    !mod_color_calculate_degamma_params(caps,
+							&dc_plane_state->in_transfer_func,
+							NULL, false))
 			return -ENOMEM;
-	} else {
-		/* ...Otherwise we can just bypass the DGM block. */
-		dc_plane_state->in_transfer_func->type = TF_TYPE_BYPASS;
-		dc_plane_state->in_transfer_func->tf = TRANSFER_FUNCTION_LINEAR;
 	}
 
 	return 0;
 }
+
+static int
+__set_dm_plane_degamma(struct drm_plane_state *plane_state,
+		       struct dc_plane_state *dc_plane_state,
+		       struct dc_color_caps *color_caps)
+{
+	struct dm_plane_state *dm_plane_state = to_dm_plane_state(plane_state);
+	const struct drm_color_lut *degamma_lut;
+	enum amdgpu_transfer_function tf = AMDGPU_TRANSFER_FUNCTION_DEFAULT;
+	uint32_t degamma_size;
+	bool has_degamma_lut;
+	int ret;
+
+	degamma_lut = __extract_blob_lut(dm_plane_state->degamma_lut,
+					 &degamma_size);
+
+	has_degamma_lut = degamma_lut &&
+			  !__is_lut_linear(degamma_lut, degamma_size);
+
+	tf = dm_plane_state->degamma_tf;
+
+	/* If we don't have plane degamma LUT nor TF to set on DC, we have
+	 * nothing to do here, return.
+	 */
+	if (!has_degamma_lut && tf == AMDGPU_TRANSFER_FUNCTION_DEFAULT)
+		return -EINVAL;
+
+	dc_plane_state->in_transfer_func.tf = amdgpu_tf_to_dc_tf(tf);
+
+	if (has_degamma_lut) {
+		ASSERT(degamma_size == MAX_COLOR_LUT_ENTRIES);
+
+		dc_plane_state->in_transfer_func.type =
+			TF_TYPE_DISTRIBUTED_POINTS;
+
+		ret = __set_input_tf(color_caps, &dc_plane_state->in_transfer_func,
+				     degamma_lut, degamma_size);
+		if (ret)
+			return ret;
+       } else {
+		dc_plane_state->in_transfer_func.type =
+			TF_TYPE_PREDEFINED;
+
+		if (!mod_color_calculate_degamma_params(color_caps,
+		    &dc_plane_state->in_transfer_func, NULL, false))
+			return -ENOMEM;
+	}
+	return 0;
+}
+
+static int
+__set_colorop_in_tf_1d_curve(struct dc_plane_state *dc_plane_state,
+			     struct drm_colorop_state *colorop_state)
+{
+	struct dc_transfer_func *tf = &dc_plane_state->in_transfer_func;
+	struct drm_colorop *colorop = colorop_state->colorop;
+	struct drm_device *drm = colorop->dev;
+
+	if (colorop->type != DRM_COLOROP_1D_CURVE)
+		return -EINVAL;
+
+	if (!(BIT(colorop_state->curve_1d_type) & amdgpu_dm_supported_degam_tfs))
+		return -EINVAL;
+
+	if (colorop_state->bypass) {
+		tf->type = TF_TYPE_BYPASS;
+		tf->tf = TRANSFER_FUNCTION_LINEAR;
+		return 0;
+	}
+
+	drm_dbg(drm, "Degamma colorop with ID: %d\n", colorop->base.id);
+
+	tf->type = TF_TYPE_PREDEFINED;
+	tf->tf = amdgpu_colorop_tf_to_dc_tf(colorop_state->curve_1d_type);
+
+	return 0;
+}
+
+static int
+__set_dm_plane_colorop_degamma(struct drm_plane_state *plane_state,
+			       struct dc_plane_state *dc_plane_state,
+			       struct drm_colorop *colorop)
+{
+	struct drm_colorop *old_colorop;
+	struct drm_colorop_state *colorop_state = NULL, *new_colorop_state;
+	struct drm_atomic_state *state = plane_state->state;
+	int i = 0;
+
+	old_colorop = colorop;
+
+	/* 1st op: 1d curve - degamma */
+	for_each_new_colorop_in_state(state, colorop, new_colorop_state, i) {
+		if (new_colorop_state->colorop == old_colorop &&
+		    (BIT(new_colorop_state->curve_1d_type) & amdgpu_dm_supported_degam_tfs)) {
+			colorop_state = new_colorop_state;
+			break;
+		}
+	}
+
+	if (!colorop_state)
+		return -EINVAL;
+
+	return __set_colorop_in_tf_1d_curve(dc_plane_state, colorop_state);
+}
+
+static int
+__set_dm_plane_colorop_3x4_matrix(struct drm_plane_state *plane_state,
+				  struct dc_plane_state *dc_plane_state,
+				  struct drm_colorop *colorop)
+{
+	struct drm_colorop *old_colorop;
+	struct drm_colorop_state *colorop_state = NULL, *new_colorop_state;
+	struct drm_atomic_state *state = plane_state->state;
+	const struct drm_device *dev = colorop->dev;
+	const struct drm_property_blob *blob;
+	struct drm_color_ctm_3x4 *ctm = NULL;
+	int i = 0;
+
+	/* 3x4 matrix */
+	old_colorop = colorop;
+	for_each_new_colorop_in_state(state, colorop, new_colorop_state, i) {
+		if (new_colorop_state->colorop == old_colorop &&
+		    new_colorop_state->colorop->type == DRM_COLOROP_CTM_3X4) {
+			colorop_state = new_colorop_state;
+			break;
+		}
+	}
+
+	if (colorop_state && !colorop_state->bypass && colorop->type == DRM_COLOROP_CTM_3X4) {
+		drm_dbg(dev, "3x4 matrix colorop with ID: %d\n", colorop->base.id);
+		blob = colorop_state->data;
+		if (blob->length == sizeof(struct drm_color_ctm_3x4)) {
+			ctm = (struct drm_color_ctm_3x4 *) blob->data;
+			__drm_ctm_3x4_to_dc_matrix(ctm, dc_plane_state->gamut_remap_matrix.matrix);
+			dc_plane_state->gamut_remap_matrix.enable_remap = true;
+			dc_plane_state->input_csc_color_matrix.enable_adjustment = false;
+		} else {
+			drm_warn(dev, "blob->length (%zu) isn't equal to drm_color_ctm_3x4 (%zu)\n",
+				 blob->length, sizeof(struct drm_color_ctm_3x4));
+			return -EINVAL;
+		}
+	}
+
+	return 0;
+}
+
+static int
+__set_dm_plane_colorop_multiplier(struct drm_plane_state *plane_state,
+				  struct dc_plane_state *dc_plane_state,
+				  struct drm_colorop *colorop)
+{
+	struct drm_colorop *old_colorop;
+	struct drm_colorop_state *colorop_state = NULL, *new_colorop_state;
+	struct drm_atomic_state *state = plane_state->state;
+	const struct drm_device *dev = colorop->dev;
+	int i = 0;
+
+	/* Multiplier */
+	old_colorop = colorop;
+	for_each_new_colorop_in_state(state, colorop, new_colorop_state, i) {
+		if (new_colorop_state->colorop == old_colorop &&
+		    new_colorop_state->colorop->type == DRM_COLOROP_MULTIPLIER) {
+			colorop_state = new_colorop_state;
+			break;
+		}
+	}
+
+	if (colorop_state && !colorop_state->bypass && colorop->type == DRM_COLOROP_MULTIPLIER) {
+		drm_dbg(dev, "Multiplier colorop with ID: %d\n", colorop->base.id);
+		dc_plane_state->hdr_mult = amdgpu_dm_fixpt_from_s3132(colorop_state->multiplier);
+	}
+
+	return 0;
+}
+
+static int
+__set_dm_plane_colorop_shaper(struct drm_plane_state *plane_state,
+			      struct dc_plane_state *dc_plane_state,
+			      struct drm_colorop *colorop)
+{
+	struct drm_colorop *old_colorop;
+	struct drm_colorop_state *colorop_state = NULL, *new_colorop_state;
+	struct drm_atomic_state *state = plane_state->state;
+	enum dc_transfer_func_predefined default_tf = TRANSFER_FUNCTION_LINEAR;
+	struct dc_transfer_func *tf = &dc_plane_state->in_shaper_func;
+	const struct drm_color_lut32 *shaper_lut;
+	struct drm_device *dev = colorop->dev;
+	bool enabled = false;
+	u32 shaper_size;
+	int i = 0, ret = 0;
+
+	/* 1D Curve - SHAPER TF */
+	old_colorop = colorop;
+	for_each_new_colorop_in_state(state, colorop, new_colorop_state, i) {
+		if (new_colorop_state->colorop == old_colorop &&
+		    (BIT(new_colorop_state->curve_1d_type) & amdgpu_dm_supported_shaper_tfs)) {
+			colorop_state = new_colorop_state;
+			break;
+		}
+	}
+
+	if (colorop_state && !colorop_state->bypass && colorop->type == DRM_COLOROP_1D_CURVE) {
+		drm_dbg(dev, "Shaper TF colorop with ID: %d\n", colorop->base.id);
+		tf->type = TF_TYPE_DISTRIBUTED_POINTS;
+		tf->tf = default_tf = amdgpu_colorop_tf_to_dc_tf(colorop_state->curve_1d_type);
+		tf->sdr_ref_white_level = SDR_WHITE_LEVEL_INIT_VALUE;
+		ret = __set_output_tf(tf, 0, 0, false);
+		if (ret)
+			return ret;
+		enabled = true;
+	}
+
+	/* 1D LUT - SHAPER LUT */
+	colorop = old_colorop->next;
+	if (!colorop) {
+		drm_dbg(dev, "no Shaper LUT colorop found\n");
+		return -EINVAL;
+	}
+
+	old_colorop = colorop;
+	for_each_new_colorop_in_state(state, colorop, new_colorop_state, i) {
+		if (new_colorop_state->colorop == old_colorop &&
+		    new_colorop_state->colorop->type == DRM_COLOROP_1D_LUT) {
+			colorop_state = new_colorop_state;
+			break;
+		}
+	}
+
+	if (colorop_state && !colorop_state->bypass && colorop->type == DRM_COLOROP_1D_LUT) {
+		drm_dbg(dev, "Shaper LUT colorop with ID: %d\n", colorop->base.id);
+		tf->type = TF_TYPE_DISTRIBUTED_POINTS;
+		tf->tf = default_tf;
+		tf->sdr_ref_white_level = SDR_WHITE_LEVEL_INIT_VALUE;
+		shaper_lut = __extract_blob_lut32(colorop_state->data, &shaper_size);
+		shaper_size = shaper_lut != NULL ? shaper_size : 0;
+
+		/* Custom LUT size must be the same as supported size */
+		if (shaper_size == colorop->size) {
+			ret = __set_output_tf_32(tf, shaper_lut, shaper_size, false);
+			if (ret)
+				return ret;
+			enabled = true;
+		}
+	}
+
+	if (!enabled)
+		tf->type = TF_TYPE_BYPASS;
+
+	return 0;
+}
+
+/* __set_colorop_3dlut - set DRM 3D LUT to DC stream
+ * @drm_lut3d: user 3D LUT
+ * @drm_lut3d_size: size of 3D LUT
+ * @lut3d: DC 3D LUT
+ *
+ * Map user 3D LUT data to DC 3D LUT and all necessary bits to program it
+ * on DCN accordingly.
+ *
+ * Returns:
+ * 0 on success. -EINVAL if drm_lut3d_size is zero.
+ */
+static int __set_colorop_3dlut(const struct drm_color_lut32 *drm_lut3d,
+				uint32_t drm_lut3d_size,
+				struct dc_3dlut *lut)
+{
+	if (!drm_lut3d_size) {
+		lut->state.bits.initialized = 0;
+		return -EINVAL;
+	}
+
+	/* Only supports 17x17x17 3D LUT (12-bit) now */
+	lut->lut_3d.use_12bits = true;
+	lut->lut_3d.use_tetrahedral_9 = false;
+
+	lut->state.bits.initialized = 1;
+	__drm_3dlut32_to_dc_3dlut(drm_lut3d, drm_lut3d_size, &lut->lut_3d,
+				   lut->lut_3d.use_tetrahedral_9, 12);
+
+	return 0;
+}
+
+static int
+__set_dm_plane_colorop_3dlut(struct drm_plane_state *plane_state,
+			     struct dc_plane_state *dc_plane_state,
+			     struct drm_colorop *colorop)
+{
+	struct drm_colorop *old_colorop;
+	struct drm_colorop_state *colorop_state = NULL, *new_colorop_state;
+	struct dc_transfer_func *tf = &dc_plane_state->in_shaper_func;
+	struct drm_atomic_state *state = plane_state->state;
+	const struct amdgpu_device *adev = drm_to_adev(colorop->dev);
+	const struct drm_device *dev = colorop->dev;
+	const struct drm_color_lut32 *lut3d;
+	uint32_t lut3d_size;
+	int i = 0, ret = 0;
+
+	/* 3D LUT */
+	old_colorop = colorop;
+	for_each_new_colorop_in_state(state, colorop, new_colorop_state, i) {
+		if (new_colorop_state->colorop == old_colorop &&
+		    new_colorop_state->colorop->type == DRM_COLOROP_3D_LUT) {
+			colorop_state = new_colorop_state;
+			break;
+		}
+	}
+
+	if (colorop_state && !colorop_state->bypass && colorop->type == DRM_COLOROP_3D_LUT) {
+		if (!adev->dm.dc->caps.color.dpp.hw_3d_lut) {
+			drm_dbg(dev, "3D LUT is not supported by hardware\n");
+			return -EINVAL;
+		}
+
+		drm_dbg(dev, "3D LUT colorop with ID: %d\n", colorop->base.id);
+		lut3d = __extract_blob_lut32(colorop_state->data, &lut3d_size);
+		lut3d_size = lut3d != NULL ? lut3d_size : 0;
+		ret = __set_colorop_3dlut(lut3d, lut3d_size, &dc_plane_state->lut3d_func);
+		if (ret) {
+			drm_dbg(dev, "3D LUT colorop with ID: %d has LUT size = %d\n",
+				colorop->base.id, lut3d_size);
+			return ret;
+		}
+
+		/* 3D LUT requires shaper. If shaper colorop is bypassed, enable shaper curve
+		 * with TRANSFER_FUNCTION_LINEAR
+		 */
+		if (tf->type == TF_TYPE_BYPASS) {
+			tf->type = TF_TYPE_DISTRIBUTED_POINTS;
+			tf->tf = TRANSFER_FUNCTION_LINEAR;
+			tf->sdr_ref_white_level = SDR_WHITE_LEVEL_INIT_VALUE;
+			ret = __set_output_tf_32(tf, NULL, 0, false);
+		}
+	}
+
+	return ret;
+}
+
+static int
+__set_dm_plane_colorop_blend(struct drm_plane_state *plane_state,
+			     struct dc_plane_state *dc_plane_state,
+			     struct drm_colorop *colorop)
+{
+	struct drm_colorop *old_colorop;
+	struct drm_colorop_state *colorop_state = NULL, *new_colorop_state;
+	struct drm_atomic_state *state = plane_state->state;
+	enum dc_transfer_func_predefined default_tf = TRANSFER_FUNCTION_LINEAR;
+	struct dc_transfer_func *tf = &dc_plane_state->blend_tf;
+	const struct drm_color_lut32 *blend_lut = NULL;
+	struct drm_device *dev = colorop->dev;
+	uint32_t blend_size = 0;
+	int i = 0;
+
+	/* 1D Curve - BLND TF */
+	old_colorop = colorop;
+	for_each_new_colorop_in_state(state, colorop, new_colorop_state, i) {
+		if (new_colorop_state->colorop == old_colorop &&
+		    (BIT(new_colorop_state->curve_1d_type) & amdgpu_dm_supported_blnd_tfs)) {
+			colorop_state = new_colorop_state;
+			break;
+		}
+	}
+
+	if (colorop_state && !colorop_state->bypass && colorop->type == DRM_COLOROP_1D_CURVE &&
+	    (BIT(colorop_state->curve_1d_type) & amdgpu_dm_supported_blnd_tfs)) {
+		drm_dbg(dev, "Blend TF colorop with ID: %d\n", colorop->base.id);
+		tf->type = TF_TYPE_DISTRIBUTED_POINTS;
+		tf->tf = default_tf = amdgpu_colorop_tf_to_dc_tf(colorop_state->curve_1d_type);
+		tf->sdr_ref_white_level = SDR_WHITE_LEVEL_INIT_VALUE;
+		__set_input_tf_32(NULL, tf, blend_lut, blend_size);
+	}
+
+	/* 1D Curve - BLND LUT */
+	colorop = old_colorop->next;
+	if (!colorop) {
+		drm_dbg(dev, "no Blend LUT colorop found\n");
+		return -EINVAL;
+	}
+
+	old_colorop = colorop;
+	for_each_new_colorop_in_state(state, colorop, new_colorop_state, i) {
+		if (new_colorop_state->colorop == old_colorop &&
+		    new_colorop_state->colorop->type == DRM_COLOROP_1D_LUT) {
+			colorop_state = new_colorop_state;
+			break;
+		}
+	}
+
+	if (colorop_state && !colorop_state->bypass && colorop->type == DRM_COLOROP_1D_LUT &&
+	    (BIT(colorop_state->curve_1d_type) & amdgpu_dm_supported_blnd_tfs)) {
+		drm_dbg(dev, "Blend LUT colorop with ID: %d\n", colorop->base.id);
+		tf->type = TF_TYPE_DISTRIBUTED_POINTS;
+		tf->tf = default_tf;
+		tf->sdr_ref_white_level = SDR_WHITE_LEVEL_INIT_VALUE;
+		blend_lut = __extract_blob_lut32(colorop_state->data, &blend_size);
+		blend_size = blend_lut != NULL ? blend_size : 0;
+
+		/* Custom LUT size must be the same as supported size */
+		if (blend_size == colorop->size)
+			__set_input_tf_32(NULL, tf, blend_lut, blend_size);
+	}
+
+	return 0;
+}
+
+static int
+amdgpu_dm_plane_set_color_properties(struct drm_plane_state *plane_state,
+				     struct dc_plane_state *dc_plane_state)
+{
+	struct dm_plane_state *dm_plane_state = to_dm_plane_state(plane_state);
+	enum amdgpu_transfer_function shaper_tf = AMDGPU_TRANSFER_FUNCTION_DEFAULT;
+	enum amdgpu_transfer_function blend_tf = AMDGPU_TRANSFER_FUNCTION_DEFAULT;
+	const struct drm_color_lut *shaper_lut, *lut3d, *blend_lut;
+	uint32_t shaper_size, lut3d_size, blend_size;
+	int ret;
+
+	dc_plane_state->hdr_mult = amdgpu_dm_fixpt_from_s3132(dm_plane_state->hdr_mult);
+
+	shaper_lut = __extract_blob_lut(dm_plane_state->shaper_lut, &shaper_size);
+	shaper_size = shaper_lut != NULL ? shaper_size : 0;
+	shaper_tf = dm_plane_state->shaper_tf;
+	lut3d = __extract_blob_lut(dm_plane_state->lut3d, &lut3d_size);
+	lut3d_size = lut3d != NULL ? lut3d_size : 0;
+
+	amdgpu_dm_atomic_lut3d(lut3d, lut3d_size, &dc_plane_state->lut3d_func);
+	ret = amdgpu_dm_atomic_shaper_lut(shaper_lut, false,
+					  amdgpu_tf_to_dc_tf(shaper_tf),
+					  shaper_size,
+					  &dc_plane_state->in_shaper_func);
+	if (ret) {
+		drm_dbg_kms(plane_state->plane->dev,
+			    "setting plane %d shaper LUT failed.\n",
+			    plane_state->plane->index);
+
+		return ret;
+	}
+
+	blend_tf = dm_plane_state->blend_tf;
+	blend_lut = __extract_blob_lut(dm_plane_state->blend_lut, &blend_size);
+	blend_size = blend_lut != NULL ? blend_size : 0;
+
+	ret = amdgpu_dm_atomic_blend_lut(blend_lut, false,
+					 amdgpu_tf_to_dc_tf(blend_tf),
+					 blend_size, &dc_plane_state->blend_tf);
+	if (ret) {
+		drm_dbg_kms(plane_state->plane->dev,
+			    "setting plane %d gamma lut failed.\n",
+			    plane_state->plane->index);
+
+		return ret;
+	}
+
+	return 0;
+}
+
+static int
+amdgpu_dm_plane_set_colorop_properties(struct drm_plane_state *plane_state,
+				       struct dc_plane_state *dc_plane_state)
+{
+	struct drm_colorop *colorop = plane_state->color_pipeline;
+	struct drm_device *dev = plane_state->plane->dev;
+	struct amdgpu_device *adev = drm_to_adev(dev);
+	int ret;
+
+	/* 1D Curve - DEGAM TF */
+	if (!colorop)
+		return -EINVAL;
+
+	ret = __set_dm_plane_colorop_degamma(plane_state, dc_plane_state, colorop);
+	if (ret)
+		return ret;
+
+	/* Multiplier */
+	colorop = colorop->next;
+	if (!colorop) {
+		drm_dbg(dev, "no multiplier colorop found\n");
+		return -EINVAL;
+	}
+
+	ret = __set_dm_plane_colorop_multiplier(plane_state, dc_plane_state, colorop);
+	if (ret)
+		return ret;
+
+	/* 3x4 matrix */
+	colorop = colorop->next;
+	if (!colorop) {
+		drm_dbg(dev, "no 3x4 matrix colorop found\n");
+		return -EINVAL;
+	}
+
+	ret = __set_dm_plane_colorop_3x4_matrix(plane_state, dc_plane_state, colorop);
+	if (ret)
+		return ret;
+
+	if (adev->dm.dc->caps.color.dpp.hw_3d_lut) {
+		/* 1D Curve & LUT - SHAPER TF & LUT */
+		colorop = colorop->next;
+		if (!colorop) {
+			drm_dbg(dev, "no Shaper TF colorop found\n");
+			return -EINVAL;
+		}
+
+		ret = __set_dm_plane_colorop_shaper(plane_state, dc_plane_state, colorop);
+		if (ret)
+			return ret;
+
+		/* Shaper LUT colorop is already handled, just skip here */
+		colorop = colorop->next;
+		if (!colorop)
+			return -EINVAL;
+
+		/* 3D LUT */
+		colorop = colorop->next;
+		if (!colorop) {
+			drm_dbg(dev, "no 3D LUT colorop found\n");
+			return -EINVAL;
+		}
+
+		ret = __set_dm_plane_colorop_3dlut(plane_state, dc_plane_state, colorop);
+		if (ret)
+			return ret;
+	}
+
+	/* 1D Curve & LUT - BLND TF & LUT */
+	colorop = colorop->next;
+	if (!colorop) {
+		drm_dbg(dev, "no Blend TF colorop found\n");
+		return -EINVAL;
+	}
+
+	ret = __set_dm_plane_colorop_blend(plane_state, dc_plane_state, colorop);
+	if (ret)
+		return ret;
+
+	/* BLND LUT colorop is already handled, just skip here */
+	colorop = colorop->next;
+	if (!colorop)
+		return -EINVAL;
+
+	return 0;
+}
+
+/**
+ * amdgpu_dm_update_plane_color_mgmt: Maps DRM color management to DC plane.
+ * @crtc: amdgpu_dm crtc state
+ * @plane_state: DRM plane state
+ * @dc_plane_state: target DC surface
+ *
+ * Update the underlying dc_stream_state's input transfer function (ITF) in
+ * preparation for hardware commit. The transfer function used depends on
+ * the preparation done on the stream for color management.
+ *
+ * Returns:
+ * 0 on success. -ENOMEM if mem allocation fails.
+ */
+int amdgpu_dm_update_plane_color_mgmt(struct dm_crtc_state *crtc,
+				      struct drm_plane_state *plane_state,
+				      struct dc_plane_state *dc_plane_state)
+{
+	struct amdgpu_device *adev = drm_to_adev(crtc->base.state->dev);
+	struct dm_plane_state *dm_plane_state = to_dm_plane_state(plane_state);
+	struct drm_color_ctm_3x4 *ctm = NULL;
+	struct dc_color_caps *color_caps = NULL;
+	bool has_crtc_cm_degamma;
+	int ret;
+
+	ret = amdgpu_dm_verify_lut3d_size(adev, plane_state);
+	if (ret) {
+		drm_dbg_driver(&adev->ddev, "amdgpu_dm_verify_lut3d_size() failed\n");
+		return ret;
+	}
+
+	if (dc_plane_state->ctx && dc_plane_state->ctx->dc)
+		color_caps = &dc_plane_state->ctx->dc->caps.color;
+
+	/* Initially, we can just bypass the DGM block. */
+	dc_plane_state->in_transfer_func.type = TF_TYPE_BYPASS;
+	dc_plane_state->in_transfer_func.tf = TRANSFER_FUNCTION_LINEAR;
+
+	/* After, we start to update values according to color props */
+	has_crtc_cm_degamma = (crtc->cm_has_degamma || crtc->cm_is_degamma_srgb);
+
+	ret = __set_dm_plane_degamma(plane_state, dc_plane_state, color_caps);
+	if (ret == -ENOMEM)
+		return ret;
+
+	/* We only have one degamma block available (pre-blending) for the
+	 * whole color correction pipeline, so that we can't actually perform
+	 * plane and CRTC degamma at the same time. Explicitly reject atomic
+	 * updates when userspace sets both plane and CRTC degamma properties.
+	 */
+	if (has_crtc_cm_degamma && ret != -EINVAL) {
+		drm_dbg_kms(crtc->base.crtc->dev,
+			    "doesn't support plane and CRTC degamma at the same time\n");
+		return -EINVAL;
+	}
+
+	/* If we are here, it means we don't have plane degamma settings, check
+	 * if we have CRTC degamma waiting for mapping to pre-blending degamma
+	 * block
+	 */
+	if (has_crtc_cm_degamma) {
+		/*
+		 * AMD HW doesn't have post-blending degamma caps. When DRM
+		 * CRTC atomic degamma is set, we maps it to DPP degamma block
+		 * (pre-blending) or, on legacy gamma, we use DPP degamma to
+		 * linearize (implicit degamma) from sRGB/BT709 according to
+		 * the input space.
+		 */
+		ret = map_crtc_degamma_to_dc_plane(crtc, dc_plane_state, color_caps);
+		if (ret)
+			return ret;
+	}
+
+	/* Setup CRTC CTM. */
+	if (dm_plane_state->ctm) {
+		ctm = (struct drm_color_ctm_3x4 *)dm_plane_state->ctm->data;
+		/*
+		 * DCN2 and older don't support both pre-blending and
+		 * post-blending gamut remap. For this HW family, if we have
+		 * the plane and CRTC CTMs simultaneously, CRTC CTM takes
+		 * priority, and we discard plane CTM, as implemented in
+		 * dcn10_program_gamut_remap(). However, DCN3+ has DPP
+		 * (pre-blending) and MPC (post-blending) `gamut remap` blocks;
+		 * therefore, we can program plane and CRTC CTMs together by
+		 * mapping CRTC CTM to MPC and keeping plane CTM setup at DPP,
+		 * as it's done by dcn30_program_gamut_remap().
+		 */
+		__drm_ctm_3x4_to_dc_matrix(ctm, dc_plane_state->gamut_remap_matrix.matrix);
+
+		dc_plane_state->gamut_remap_matrix.enable_remap = true;
+		dc_plane_state->input_csc_color_matrix.enable_adjustment = false;
+	} else {
+		/* Bypass CTM. */
+		dc_plane_state->gamut_remap_matrix.enable_remap = false;
+		dc_plane_state->input_csc_color_matrix.enable_adjustment = false;
+	}
+
+	if (!amdgpu_dm_plane_set_colorop_properties(plane_state, dc_plane_state))
+		return 0;
+
+	return amdgpu_dm_plane_set_color_properties(plane_state, dc_plane_state);
+}
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_colorop.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_colorop.c
new file mode 100644
index 000000000000..d585618b8064
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_colorop.c
@@ -0,0 +1,209 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: AMD
+ *
+ */
+
+#include <drm/drm_print.h>
+#include <drm/drm_plane.h>
+#include <drm/drm_property.h>
+#include <drm/drm_colorop.h>
+
+#include "amdgpu.h"
+#include "amdgpu_dm_colorop.h"
+#include "dc.h"
+
+const u64 amdgpu_dm_supported_degam_tfs =
+	BIT(DRM_COLOROP_1D_CURVE_SRGB_EOTF) |
+	BIT(DRM_COLOROP_1D_CURVE_PQ_125_EOTF) |
+	BIT(DRM_COLOROP_1D_CURVE_BT2020_INV_OETF) |
+	BIT(DRM_COLOROP_1D_CURVE_GAMMA22_INV);
+
+const u64 amdgpu_dm_supported_shaper_tfs =
+	BIT(DRM_COLOROP_1D_CURVE_SRGB_INV_EOTF) |
+	BIT(DRM_COLOROP_1D_CURVE_PQ_125_INV_EOTF) |
+	BIT(DRM_COLOROP_1D_CURVE_BT2020_OETF) |
+	BIT(DRM_COLOROP_1D_CURVE_GAMMA22);
+
+const u64 amdgpu_dm_supported_blnd_tfs =
+	BIT(DRM_COLOROP_1D_CURVE_SRGB_EOTF) |
+	BIT(DRM_COLOROP_1D_CURVE_PQ_125_EOTF) |
+	BIT(DRM_COLOROP_1D_CURVE_BT2020_INV_OETF) |
+	BIT(DRM_COLOROP_1D_CURVE_GAMMA22_INV);
+
+#define MAX_COLOR_PIPELINE_OPS 10
+
+#define LUT3D_SIZE		17
+
+int amdgpu_dm_initialize_default_pipeline(struct drm_plane *plane, struct drm_prop_enum_list *list)
+{
+	struct drm_colorop *ops[MAX_COLOR_PIPELINE_OPS];
+	struct drm_device *dev = plane->dev;
+	struct amdgpu_device *adev = drm_to_adev(dev);
+	int ret;
+	int i = 0;
+
+	memset(ops, 0, sizeof(ops));
+
+	/* 1D curve - DEGAM TF */
+	ops[i] = kzalloc(sizeof(*ops[0]), GFP_KERNEL);
+	if (!ops[i]) {
+		ret = -ENOMEM;
+		goto cleanup;
+	}
+
+	ret = drm_plane_colorop_curve_1d_init(dev, ops[i], plane,
+					      amdgpu_dm_supported_degam_tfs,
+					      DRM_COLOROP_FLAG_ALLOW_BYPASS);
+	if (ret)
+		goto cleanup;
+
+	list->type = ops[i]->base.id;
+	list->name = kasprintf(GFP_KERNEL, "Color Pipeline %d", ops[i]->base.id);
+
+	i++;
+
+	/* Multiplier */
+	ops[i] = kzalloc(sizeof(struct drm_colorop), GFP_KERNEL);
+	if (!ops[i]) {
+		ret = -ENOMEM;
+		goto cleanup;
+	}
+
+	ret = drm_plane_colorop_mult_init(dev, ops[i], plane, DRM_COLOROP_FLAG_ALLOW_BYPASS);
+	if (ret)
+		goto cleanup;
+
+	drm_colorop_set_next_property(ops[i-1], ops[i]);
+
+	i++;
+
+	/* 3x4 matrix */
+	ops[i] = kzalloc(sizeof(struct drm_colorop), GFP_KERNEL);
+	if (!ops[i]) {
+		ret = -ENOMEM;
+		goto cleanup;
+	}
+
+	ret = drm_plane_colorop_ctm_3x4_init(dev, ops[i], plane, DRM_COLOROP_FLAG_ALLOW_BYPASS);
+	if (ret)
+		goto cleanup;
+
+	drm_colorop_set_next_property(ops[i-1], ops[i]);
+
+	i++;
+
+	if (adev->dm.dc->caps.color.dpp.hw_3d_lut) {
+		/* 1D curve - SHAPER TF */
+		ops[i] = kzalloc(sizeof(*ops[0]), GFP_KERNEL);
+		if (!ops[i]) {
+			ret = -ENOMEM;
+			goto cleanup;
+		}
+
+		ret = drm_plane_colorop_curve_1d_init(dev, ops[i], plane,
+						amdgpu_dm_supported_shaper_tfs,
+						DRM_COLOROP_FLAG_ALLOW_BYPASS);
+		if (ret)
+			goto cleanup;
+
+		drm_colorop_set_next_property(ops[i-1], ops[i]);
+
+		i++;
+
+		/* 1D LUT - SHAPER LUT */
+		ops[i] = kzalloc(sizeof(*ops[0]), GFP_KERNEL);
+		if (!ops[i]) {
+			ret = -ENOMEM;
+			goto cleanup;
+		}
+
+		ret = drm_plane_colorop_curve_1d_lut_init(dev, ops[i], plane, MAX_COLOR_LUT_ENTRIES,
+							DRM_COLOROP_LUT1D_INTERPOLATION_LINEAR,
+							DRM_COLOROP_FLAG_ALLOW_BYPASS);
+		if (ret)
+			goto cleanup;
+
+		drm_colorop_set_next_property(ops[i-1], ops[i]);
+
+		i++;
+
+		/* 3D LUT */
+		ops[i] = kzalloc(sizeof(*ops[0]), GFP_KERNEL);
+		if (!ops[i]) {
+			ret = -ENOMEM;
+			goto cleanup;
+		}
+
+		ret = drm_plane_colorop_3dlut_init(dev, ops[i], plane, LUT3D_SIZE,
+					DRM_COLOROP_LUT3D_INTERPOLATION_TETRAHEDRAL,
+					DRM_COLOROP_FLAG_ALLOW_BYPASS);
+		if (ret)
+			goto cleanup;
+
+		drm_colorop_set_next_property(ops[i-1], ops[i]);
+
+		i++;
+	}
+
+	/* 1D curve - BLND TF */
+	ops[i] = kzalloc(sizeof(*ops[0]), GFP_KERNEL);
+	if (!ops[i]) {
+		ret = -ENOMEM;
+		goto cleanup;
+	}
+
+	ret = drm_plane_colorop_curve_1d_init(dev, ops[i], plane,
+					      amdgpu_dm_supported_blnd_tfs,
+					      DRM_COLOROP_FLAG_ALLOW_BYPASS);
+	if (ret)
+		goto cleanup;
+
+	drm_colorop_set_next_property(ops[i - 1], ops[i]);
+
+	i++;
+
+	/* 1D LUT - BLND LUT */
+	ops[i] = kzalloc(sizeof(struct drm_colorop), GFP_KERNEL);
+	if (!ops[i]) {
+		ret = -ENOMEM;
+		goto cleanup;
+	}
+
+	ret = drm_plane_colorop_curve_1d_lut_init(dev, ops[i], plane, MAX_COLOR_LUT_ENTRIES,
+						  DRM_COLOROP_LUT1D_INTERPOLATION_LINEAR,
+						  DRM_COLOROP_FLAG_ALLOW_BYPASS);
+	if (ret)
+		goto cleanup;
+
+	drm_colorop_set_next_property(ops[i-1], ops[i]);
+	return 0;
+
+cleanup:
+	if (ret == -ENOMEM)
+		drm_err(plane->dev, "KMS: Failed to allocate colorop\n");
+
+	drm_colorop_pipeline_destroy(dev);
+
+	return ret;
+}
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_colorop.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_colorop.h
new file mode 100644
index 000000000000..2e1617ffc8ee
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_colorop.h
@@ -0,0 +1,36 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: AMD
+ *
+ */
+
+#ifndef __AMDGPU_DM_COLOROP_H__
+#define __AMDGPU_DM_COLOROP_H__
+
+extern const u64 amdgpu_dm_supported_degam_tfs;
+extern const u64 amdgpu_dm_supported_shaper_tfs;
+extern const u64 amdgpu_dm_supported_blnd_tfs;
+
+int amdgpu_dm_initialize_default_pipeline(struct drm_plane *plane, struct drm_prop_enum_list *list);
+
+#endif /* __AMDGPU_DM_COLOROP_H__*/
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.c
index 27711743c22c..e20aa7438066 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.c
@@ -1,3 +1,4 @@
+// SPDX-License-Identifier: MIT
 /*
  * Copyright 2015 Advanced Micro Devices, Inc.
  *
@@ -30,6 +31,7 @@
 #include "amdgpu_dm.h"
 #include "dc.h"
 #include "amdgpu_securedisplay.h"
+#include "amdgpu_dm_psr.h"
 
 static const char *const pipe_crc_sources[] = {
 	"none",
@@ -83,38 +85,277 @@ const char *const *amdgpu_dm_crtc_get_crc_sources(struct drm_crtc *crtc,
 }
 
 #ifdef CONFIG_DRM_AMD_SECURE_DISPLAY
-static void amdgpu_dm_set_crc_window_default(struct drm_crtc *crtc)
+static void update_phy_id_mapping(struct amdgpu_device *adev)
+{
+	struct drm_device *ddev = adev_to_drm(adev);
+	struct amdgpu_display_manager *dm = &adev->dm;
+	struct drm_connector *connector;
+	struct amdgpu_dm_connector *aconnector;
+	struct amdgpu_dm_connector *sort_connector[AMDGPU_DM_MAX_CRTC] = {NULL};
+	struct drm_connector_list_iter iter;
+	uint8_t idx = 0, idx_2 = 0, connector_cnt = 0;
+
+	dm->secure_display_ctx.phy_mapping_updated = false;
+
+	mutex_lock(&ddev->mode_config.mutex);
+	drm_connector_list_iter_begin(ddev, &iter);
+	drm_for_each_connector_iter(connector, &iter) {
+
+		if (connector->status != connector_status_connected)
+			continue;
+
+		if (idx >= AMDGPU_DM_MAX_CRTC) {
+			DRM_WARN("%s connected connectors exceed max crtc\n", __func__);
+			mutex_unlock(&ddev->mode_config.mutex);
+			return;
+		}
+
+		aconnector = to_amdgpu_dm_connector(connector);
+
+		sort_connector[idx] = aconnector;
+		idx++;
+		connector_cnt++;
+	}
+	drm_connector_list_iter_end(&iter);
+
+	/* sort connectors by link_enc_hw_instance first */
+	for (idx = connector_cnt; idx > 1 ; idx--) {
+		for (idx_2 = 0; idx_2 < (idx - 1); idx_2++) {
+			if (sort_connector[idx_2]->dc_link->link_enc_hw_inst >
+			    sort_connector[idx_2 + 1]->dc_link->link_enc_hw_inst)
+				swap(sort_connector[idx_2], sort_connector[idx_2 + 1]);
+		}
+	}
+
+	/*
+	 * Sort mst connectors by RAD. mst connectors with the same enc_hw_instance are already
+	 * sorted together above.
+	 */
+	for (idx = 0; idx < connector_cnt; /*Do nothing*/) {
+		if (sort_connector[idx]->mst_root) {
+			uint8_t i, j, k;
+			uint8_t mst_con_cnt = 1;
+
+			for (idx_2 = (idx + 1); idx_2 < connector_cnt; idx_2++) {
+				if (sort_connector[idx_2]->mst_root == sort_connector[idx]->mst_root)
+					mst_con_cnt++;
+				else
+					break;
+			}
+
+			for (i = mst_con_cnt; i > 1; i--) {
+				for (j = idx; j < (idx + i - 2); j++) {
+					int mstb_lct = sort_connector[j]->mst_output_port->parent->lct;
+					int next_mstb_lct = sort_connector[j + 1]->mst_output_port->parent->lct;
+					u8 *rad;
+					u8 *next_rad;
+					bool swap = false;
+
+					/* Sort by mst tree depth first. Then compare RAD if depth is the same*/
+					if (mstb_lct > next_mstb_lct) {
+						swap = true;
+					} else if (mstb_lct == next_mstb_lct) {
+						if (mstb_lct == 1) {
+							if (sort_connector[j]->mst_output_port->port_num > sort_connector[j + 1]->mst_output_port->port_num)
+								swap = true;
+						} else if (mstb_lct > 1) {
+							rad = sort_connector[j]->mst_output_port->parent->rad;
+							next_rad = sort_connector[j + 1]->mst_output_port->parent->rad;
+
+							for (k = 0; k < mstb_lct - 1; k++) {
+								int shift = (k % 2) ? 0 : 4;
+								int port_num = (rad[k / 2] >> shift) & 0xf;
+								int next_port_num = (next_rad[k / 2] >> shift) & 0xf;
+
+								if (port_num > next_port_num) {
+									swap = true;
+									break;
+								}
+							}
+						} else {
+							DRM_ERROR("MST LCT shouldn't be set as < 1");
+							mutex_unlock(&ddev->mode_config.mutex);
+							return;
+						}
+					}
+
+					if (swap)
+						swap(sort_connector[j], sort_connector[j + 1]);
+				}
+			}
+
+			idx += mst_con_cnt;
+		} else {
+			idx++;
+		}
+	}
+
+	/* Complete sorting. Assign relavant result to dm->secure_display_ctx.phy_id_mapping[]*/
+	memset(dm->secure_display_ctx.phy_id_mapping, 0, sizeof(dm->secure_display_ctx.phy_id_mapping));
+	for (idx = 0; idx < connector_cnt; idx++) {
+		aconnector = sort_connector[idx];
+
+		dm->secure_display_ctx.phy_id_mapping[idx].assigned = true;
+		dm->secure_display_ctx.phy_id_mapping[idx].is_mst = false;
+		dm->secure_display_ctx.phy_id_mapping[idx].enc_hw_inst = aconnector->dc_link->link_enc_hw_inst;
+
+		if (sort_connector[idx]->mst_root) {
+			dm->secure_display_ctx.phy_id_mapping[idx].is_mst = true;
+			dm->secure_display_ctx.phy_id_mapping[idx].lct = aconnector->mst_output_port->parent->lct;
+			dm->secure_display_ctx.phy_id_mapping[idx].port_num = aconnector->mst_output_port->port_num;
+			memcpy(dm->secure_display_ctx.phy_id_mapping[idx].rad,
+				aconnector->mst_output_port->parent->rad, sizeof(aconnector->mst_output_port->parent->rad));
+		}
+	}
+	mutex_unlock(&ddev->mode_config.mutex);
+
+	dm->secure_display_ctx.phy_id_mapping_cnt = connector_cnt;
+	dm->secure_display_ctx.phy_mapping_updated = true;
+}
+
+static bool get_phy_id(struct amdgpu_display_manager *dm,
+			struct amdgpu_dm_connector *aconnector, uint8_t *phy_id)
+{
+	int idx, idx_2;
+	bool found = false;
+
+	/*
+	 * Assume secure display start after all connectors are probed. The connection
+	 * config is static as well
+	 */
+	if (!dm->secure_display_ctx.phy_mapping_updated) {
+		DRM_WARN("%s Should update the phy id table before get it's value", __func__);
+		return false;
+	}
+
+	for (idx = 0; idx < dm->secure_display_ctx.phy_id_mapping_cnt; idx++) {
+		if (!dm->secure_display_ctx.phy_id_mapping[idx].assigned) {
+			DRM_ERROR("phy_id_mapping[%d] should be assigned", idx);
+			return false;
+		}
+
+		if (aconnector->dc_link->link_enc_hw_inst ==
+				dm->secure_display_ctx.phy_id_mapping[idx].enc_hw_inst) {
+			if (!dm->secure_display_ctx.phy_id_mapping[idx].is_mst) {
+				found = true;
+				goto out;
+			} else {
+				/* Could caused by wrongly pass mst root connector */
+				if (!aconnector->mst_output_port) {
+					DRM_ERROR("%s Check mst case but connector without a port assigned", __func__);
+					return false;
+				}
+
+				if (aconnector->mst_root &&
+					aconnector->mst_root->mst_mgr.mst_primary == NULL) {
+					DRM_WARN("%s pass in a stale mst connector", __func__);
+				}
+
+				if (aconnector->mst_output_port->parent->lct == dm->secure_display_ctx.phy_id_mapping[idx].lct &&
+					aconnector->mst_output_port->port_num == dm->secure_display_ctx.phy_id_mapping[idx].port_num) {
+					if (aconnector->mst_output_port->parent->lct == 1) {
+						found = true;
+						goto out;
+					} else if (aconnector->mst_output_port->parent->lct > 1) {
+						/* Check RAD */
+						for (idx_2 = 0; idx_2 < aconnector->mst_output_port->parent->lct - 1; idx_2++) {
+							int shift = (idx_2 % 2) ? 0 : 4;
+							int port_num = (aconnector->mst_output_port->parent->rad[idx_2 / 2] >> shift) & 0xf;
+							int port_num2 = (dm->secure_display_ctx.phy_id_mapping[idx].rad[idx_2 / 2] >> shift) & 0xf;
+
+							if (port_num != port_num2)
+								break;
+						}
+
+						if (idx_2 == aconnector->mst_output_port->parent->lct - 1) {
+							found = true;
+							goto out;
+						}
+					} else {
+						DRM_ERROR("lCT should be >= 1");
+						return false;
+					}
+				}
+			}
+		}
+	}
+
+out:
+	if (found) {
+		DRM_DEBUG_DRIVER("Associated secure display PHY ID as %d", idx);
+		*phy_id = idx;
+	} else {
+		DRM_WARN("Can't find associated phy ID");
+		return false;
+	}
+
+	return true;
+}
+
+static void amdgpu_dm_set_crc_window_default(struct drm_crtc *crtc, struct dc_stream_state *stream)
 {
 	struct drm_device *drm_dev = crtc->dev;
+	struct amdgpu_display_manager *dm = &drm_to_adev(drm_dev)->dm;
 	struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
+	struct amdgpu_dm_connector *aconnector;
+	bool was_activated;
+	uint8_t phy_id;
+	unsigned long flags;
+	int i;
 
-	spin_lock_irq(&drm_dev->event_lock);
-	acrtc->dm_irq_params.window_param.x_start = 0;
-	acrtc->dm_irq_params.window_param.y_start = 0;
-	acrtc->dm_irq_params.window_param.x_end = 0;
-	acrtc->dm_irq_params.window_param.y_end = 0;
-	acrtc->dm_irq_params.window_param.activated = false;
-	acrtc->dm_irq_params.window_param.update_win = false;
-	acrtc->dm_irq_params.window_param.skip_frame_cnt = 0;
-	spin_unlock_irq(&drm_dev->event_lock);
+	spin_lock_irqsave(&drm_dev->event_lock, flags);
+	was_activated = acrtc->dm_irq_params.crc_window_activated;
+	for (i = 0; i < MAX_CRC_WINDOW_NUM; i++) {
+		acrtc->dm_irq_params.window_param[i].x_start = 0;
+		acrtc->dm_irq_params.window_param[i].y_start = 0;
+		acrtc->dm_irq_params.window_param[i].x_end = 0;
+		acrtc->dm_irq_params.window_param[i].y_end = 0;
+		acrtc->dm_irq_params.window_param[i].enable = false;
+		acrtc->dm_irq_params.window_param[i].update_win = false;
+		acrtc->dm_irq_params.window_param[i].skip_frame_cnt = 0;
+	}
+	acrtc->dm_irq_params.crc_window_activated = false;
+	spin_unlock_irqrestore(&drm_dev->event_lock, flags);
+
+	/* Disable secure_display if it was enabled */
+	if (was_activated && dm->secure_display_ctx.op_mode == LEGACY_MODE) {
+		/* stop ROI update on this crtc */
+		flush_work(&dm->secure_display_ctx.crtc_ctx[crtc->index].notify_ta_work);
+		flush_work(&dm->secure_display_ctx.crtc_ctx[crtc->index].forward_roi_work);
+		aconnector = (struct amdgpu_dm_connector *)stream->dm_stream_context;
+
+		if (aconnector && get_phy_id(dm, aconnector, &phy_id)) {
+			if (dm->secure_display_ctx.support_mul_roi)
+				dc_stream_forward_multiple_crc_window(stream, NULL, phy_id, true);
+			else
+				dc_stream_forward_crc_window(stream, NULL, phy_id, true);
+		} else {
+			DRM_DEBUG_DRIVER("%s Can't find matching phy id", __func__);
+		}
+	}
 }
 
 static void amdgpu_dm_crtc_notify_ta_to_read(struct work_struct *work)
 {
-	struct secure_display_context *secure_display_ctx;
+	struct secure_display_crtc_context *crtc_ctx;
 	struct psp_context *psp;
 	struct ta_securedisplay_cmd *securedisplay_cmd;
 	struct drm_crtc *crtc;
 	struct dc_stream_state *stream;
+	struct amdgpu_dm_connector *aconnector;
 	uint8_t phy_inst;
+	struct amdgpu_display_manager *dm;
+	struct crc_data crc_cpy[MAX_CRC_WINDOW_NUM];
+	unsigned long flags;
+	uint8_t roi_idx = 0;
 	int ret;
+	int i;
 
-	secure_display_ctx = container_of(work, struct secure_display_context, notify_ta_work);
-	crtc = secure_display_ctx->crtc;
+	crtc_ctx = container_of(work, struct secure_display_crtc_context, notify_ta_work);
+	crtc = crtc_ctx->crtc;
 
-	if (!crtc) {
+	if (!crtc)
 		return;
-	}
 
 	psp = &drm_to_adev(crtc->dev)->psp;
 
@@ -123,26 +364,54 @@ static void amdgpu_dm_crtc_notify_ta_to_read(struct work_struct *work)
 		return;
 	}
 
+	dm = &drm_to_adev(crtc->dev)->dm;
 	stream = to_amdgpu_crtc(crtc)->dm_irq_params.stream;
-	phy_inst = stream->link->link_enc_hw_inst;
-
-	/* need lock for multiple crtcs to use the command buffer */
-	mutex_lock(&psp->securedisplay_context.mutex);
+	aconnector = (struct amdgpu_dm_connector *)stream->dm_stream_context;
+	if (!aconnector)
+		return;
 
-	psp_prep_securedisplay_cmd_buf(psp, &securedisplay_cmd,
-						TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC);
+	mutex_lock(&crtc->dev->mode_config.mutex);
+	if (!get_phy_id(dm, aconnector, &phy_inst)) {
+		DRM_WARN("%s Can't find mapping phy id!", __func__);
+		mutex_unlock(&crtc->dev->mode_config.mutex);
+		return;
+	}
+	mutex_unlock(&crtc->dev->mode_config.mutex);
 
-	securedisplay_cmd->securedisplay_in_message.send_roi_crc.phy_id = phy_inst;
+	spin_lock_irqsave(&crtc->dev->event_lock, flags);
+	memcpy(crc_cpy, crtc_ctx->crc_info.crc, sizeof(struct crc_data) * MAX_CRC_WINDOW_NUM);
+	spin_unlock_irqrestore(&crtc->dev->event_lock, flags);
 
+	/* need lock for multiple crtcs to use the command buffer */
+	mutex_lock(&psp->securedisplay_context.mutex);
 	/* PSP TA is expected to finish data transmission over I2C within current frame,
 	 * even there are up to 4 crtcs request to send in this frame.
 	 */
-	ret = psp_securedisplay_invoke(psp, TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC);
+	if (dm->secure_display_ctx.support_mul_roi) {
+		psp_prep_securedisplay_cmd_buf(psp, &securedisplay_cmd,
+							TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC_V2);
+
+		securedisplay_cmd->securedisplay_in_message.send_roi_crc_v2.phy_id = phy_inst;
+
+		for (i = 0; i < MAX_CRC_WINDOW_NUM; i++) {
+			if (crc_cpy[i].crc_ready)
+				roi_idx |= 1 << i;
+		}
+		securedisplay_cmd->securedisplay_in_message.send_roi_crc_v2.roi_idx = roi_idx;
+
+		ret = psp_securedisplay_invoke(psp, TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC_V2);
+	} else {
+		psp_prep_securedisplay_cmd_buf(psp, &securedisplay_cmd,
+							TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC);
+
+		securedisplay_cmd->securedisplay_in_message.send_roi_crc.phy_id = phy_inst;
+
+		ret = psp_securedisplay_invoke(psp, TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC);
+	}
 
 	if (!ret) {
-		if (securedisplay_cmd->status != TA_SECUREDISPLAY_STATUS__SUCCESS) {
+		if (securedisplay_cmd->status != TA_SECUREDISPLAY_STATUS__SUCCESS)
 			psp_securedisplay_parse_resp_status(psp, securedisplay_cmd->status);
-		}
 	}
 
 	mutex_unlock(&psp->securedisplay_context.mutex);
@@ -151,22 +420,47 @@ static void amdgpu_dm_crtc_notify_ta_to_read(struct work_struct *work)
 static void
 amdgpu_dm_forward_crc_window(struct work_struct *work)
 {
-	struct secure_display_context *secure_display_ctx;
+	struct secure_display_crtc_context *crtc_ctx;
 	struct amdgpu_display_manager *dm;
 	struct drm_crtc *crtc;
 	struct dc_stream_state *stream;
+	struct amdgpu_dm_connector *aconnector;
+	struct crc_window roi_cpy[MAX_CRC_WINDOW_NUM];
+	unsigned long flags;
+	uint8_t phy_id;
 
-	secure_display_ctx = container_of(work, struct secure_display_context, forward_roi_work);
-	crtc = secure_display_ctx->crtc;
+	crtc_ctx = container_of(work, struct secure_display_crtc_context, forward_roi_work);
+	crtc = crtc_ctx->crtc;
 
 	if (!crtc)
 		return;
 
 	dm = &drm_to_adev(crtc->dev)->dm;
 	stream = to_amdgpu_crtc(crtc)->dm_irq_params.stream;
+	aconnector = (struct amdgpu_dm_connector *)stream->dm_stream_context;
+
+	if (!aconnector)
+		return;
+
+	mutex_lock(&crtc->dev->mode_config.mutex);
+	if (!get_phy_id(dm, aconnector, &phy_id)) {
+		DRM_WARN("%s Can't find mapping phy id!", __func__);
+		mutex_unlock(&crtc->dev->mode_config.mutex);
+		return;
+	}
+	mutex_unlock(&crtc->dev->mode_config.mutex);
+
+	spin_lock_irqsave(&crtc->dev->event_lock, flags);
+	memcpy(roi_cpy, crtc_ctx->roi, sizeof(struct crc_window) * MAX_CRC_WINDOW_NUM);
+	spin_unlock_irqrestore(&crtc->dev->event_lock, flags);
 
 	mutex_lock(&dm->dc_lock);
-	dc_stream_forward_crc_window(stream, &secure_display_ctx->rect, false);
+	if (dm->secure_display_ctx.support_mul_roi)
+		dc_stream_forward_multiple_crc_window(stream, roi_cpy,
+			phy_id, false);
+	else
+		dc_stream_forward_crc_window(stream, &roi_cpy[0].rect,
+			phy_id, false);
 	mutex_unlock(&dm->dc_lock);
 }
 
@@ -177,7 +471,7 @@ bool amdgpu_dm_crc_window_is_activated(struct drm_crtc *crtc)
 	bool ret = false;
 
 	spin_lock_irq(&drm_dev->event_lock);
-	ret = acrtc->dm_irq_params.window_param.activated;
+	ret = acrtc->dm_irq_params.crc_window_activated;
 	spin_unlock_irq(&drm_dev->event_lock);
 
 	return ret;
@@ -204,9 +498,6 @@ int amdgpu_dm_crtc_configure_crc_source(struct drm_crtc *crtc,
 					struct dm_crtc_state *dm_crtc_state,
 					enum amdgpu_dm_pipe_crc_source source)
 {
-#if defined(CONFIG_DRM_AMD_SECURE_DISPLAY)
-	int i;
-#endif
 	struct amdgpu_device *adev = drm_to_adev(crtc->dev);
 	struct dc_stream_state *stream_state = dm_crtc_state->stream;
 	bool enable = amdgpu_dm_is_valid_crc_source(source);
@@ -218,23 +509,14 @@ int amdgpu_dm_crtc_configure_crc_source(struct drm_crtc *crtc,
 
 	mutex_lock(&adev->dm.dc_lock);
 
+	/* For PSR1, check that the panel has exited PSR */
+	if (stream_state->link->psr_settings.psr_version < DC_PSR_VERSION_SU_1)
+		amdgpu_dm_psr_wait_disable(stream_state);
+
 	/* Enable or disable CRTC CRC generation */
 	if (dm_is_crc_source_crtc(source) || source == AMDGPU_DM_PIPE_CRC_SOURCE_NONE) {
-#if defined(CONFIG_DRM_AMD_SECURE_DISPLAY)
-		/* Disable secure_display if it was enabled */
-		if (!enable) {
-			for (i = 0; i < adev->mode_info.num_crtc; i++) {
-				if (adev->dm.secure_display_ctxs[i].crtc == crtc) {
-					/* stop ROI update on this crtc */
-					flush_work(&adev->dm.secure_display_ctxs[i].notify_ta_work);
-					flush_work(&adev->dm.secure_display_ctxs[i].forward_roi_work);
-					dc_stream_forward_crc_window(stream_state, NULL, true);
-				}
-			}
-		}
-#endif
 		if (!dc_stream_configure_crc(stream_state->ctx->dc,
-					     stream_state, NULL, enable, enable)) {
+					     stream_state, NULL, enable, enable, 0, true)) {
 			ret = -EINVAL;
 			goto unlock;
 		}
@@ -265,6 +547,10 @@ int amdgpu_dm_crtc_set_crc_source(struct drm_crtc *crtc, const char *src_name)
 	struct drm_crtc_commit *commit;
 	struct dm_crtc_state *crtc_state;
 	struct drm_device *drm_dev = crtc->dev;
+#if defined(CONFIG_DRM_AMD_SECURE_DISPLAY)
+	struct amdgpu_device *adev = drm_to_adev(drm_dev);
+	struct amdgpu_display_manager *dm = &adev->dm;
+#endif
 	struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
 	struct drm_dp_aux *aux = NULL;
 	bool enable = false;
@@ -333,6 +619,9 @@ int amdgpu_dm_crtc_set_crc_source(struct drm_crtc *crtc, const char *src_name)
 			if (!connector->state || connector->state->crtc != crtc)
 				continue;
 
+			if (connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+				continue;
+
 			aconn = to_amdgpu_dm_connector(connector);
 			break;
 		}
@@ -361,16 +650,6 @@ int amdgpu_dm_crtc_set_crc_source(struct drm_crtc *crtc, const char *src_name)
 
 	}
 
-#if defined(CONFIG_DRM_AMD_SECURE_DISPLAY)
-	/* Reset secure_display when we change crc source from debugfs */
-	amdgpu_dm_set_crc_window_default(crtc);
-#endif
-
-	if (amdgpu_dm_crtc_configure_crc_source(crtc, crtc_state, source)) {
-		ret = -EINVAL;
-		goto cleanup;
-	}
-
 	/*
 	 * Reading the CRC requires the vblank interrupt handler to be
 	 * enabled. Keep a reference until CRC capture stops.
@@ -380,7 +659,19 @@ int amdgpu_dm_crtc_set_crc_source(struct drm_crtc *crtc, const char *src_name)
 		ret = drm_crtc_vblank_get(crtc);
 		if (ret)
 			goto cleanup;
+	}
+
+#if defined(CONFIG_DRM_AMD_SECURE_DISPLAY)
+	/* Reset secure_display when we change crc source from debugfs */
+	amdgpu_dm_set_crc_window_default(crtc, crtc_state->stream);
+#endif
 
+	if (amdgpu_dm_crtc_configure_crc_source(crtc, crtc_state, source)) {
+		ret = -EINVAL;
+		goto cleanup;
+	}
+
+	if (!enabled && enable) {
 		if (dm_is_crc_source_dprx(source)) {
 			if (drm_dp_start_crc(aux, crtc)) {
 				DRM_DEBUG_DRIVER("dp start crc failed\n");
@@ -406,6 +697,13 @@ int amdgpu_dm_crtc_set_crc_source(struct drm_crtc *crtc, const char *src_name)
 	/* Reset crc_skipped on dm state */
 	crtc_state->crc_skip_count = 0;
 
+#if defined(CONFIG_DRM_AMD_SECURE_DISPLAY)
+	/* Initialize phy id mapping table for secure display*/
+	if (dm->secure_display_ctx.op_mode == LEGACY_MODE &&
+		!dm->secure_display_ctx.phy_mapping_updated)
+		update_phy_id_mapping(adev);
+#endif
+
 cleanup:
 	if (commit)
 		drm_crtc_commit_put(commit);
@@ -460,7 +758,7 @@ void amdgpu_dm_crtc_handle_crc_irq(struct drm_crtc *crtc)
 	}
 
 	if (dm_is_crc_source_crtc(cur_crc_src)) {
-		if (!dc_stream_get_crc(stream_state->ctx->dc, stream_state,
+		if (!dc_stream_get_crc(stream_state->ctx->dc, stream_state, 0,
 				       &crcs[0], &crcs[1], &crcs[2]))
 			return;
 
@@ -476,8 +774,17 @@ void amdgpu_dm_crtc_handle_crc_window_irq(struct drm_crtc *crtc)
 	enum amdgpu_dm_pipe_crc_source cur_crc_src;
 	struct amdgpu_crtc *acrtc = NULL;
 	struct amdgpu_device *adev = NULL;
-	struct secure_display_context *secure_display_ctx = NULL;
+	struct secure_display_crtc_context *crtc_ctx = NULL;
+	bool reset_crc_frame_count[MAX_CRC_WINDOW_NUM] = {false};
+	uint32_t crc_r[MAX_CRC_WINDOW_NUM] = {0};
+	uint32_t crc_g[MAX_CRC_WINDOW_NUM] = {0};
+	uint32_t crc_b[MAX_CRC_WINDOW_NUM] = {0};
 	unsigned long flags1;
+	bool forward_roi_change = false;
+	bool notify_ta = false;
+	bool all_crc_ready = true;
+	struct dc_stream_state *stream_state;
+	int i;
 
 	if (crtc == NULL)
 		return;
@@ -485,78 +792,160 @@ void amdgpu_dm_crtc_handle_crc_window_irq(struct drm_crtc *crtc)
 	acrtc = to_amdgpu_crtc(crtc);
 	adev = drm_to_adev(crtc->dev);
 	drm_dev = crtc->dev;
+	stream_state = to_dm_crtc_state(crtc->state)->stream;
 
 	spin_lock_irqsave(&drm_dev->event_lock, flags1);
 	cur_crc_src = acrtc->dm_irq_params.crc_src;
 
 	/* Early return if CRC capture is not enabled. */
 	if (!amdgpu_dm_is_valid_crc_source(cur_crc_src) ||
-		!dm_is_crc_source_crtc(cur_crc_src))
-		goto cleanup;
-
-	if (!acrtc->dm_irq_params.window_param.activated)
-		goto cleanup;
+	    !dm_is_crc_source_crtc(cur_crc_src)) {
+		spin_unlock_irqrestore(&drm_dev->event_lock, flags1);
+		return;
+	}
 
-	if (acrtc->dm_irq_params.window_param.skip_frame_cnt) {
-		acrtc->dm_irq_params.window_param.skip_frame_cnt -= 1;
-		goto cleanup;
+	if (!acrtc->dm_irq_params.crc_window_activated) {
+		spin_unlock_irqrestore(&drm_dev->event_lock, flags1);
+		return;
 	}
 
-	secure_display_ctx = &adev->dm.secure_display_ctxs[acrtc->crtc_id];
-	if (WARN_ON(secure_display_ctx->crtc != crtc)) {
-		/* We have set the crtc when creating secure_display_context,
+	crtc_ctx = &adev->dm.secure_display_ctx.crtc_ctx[acrtc->crtc_id];
+	if (WARN_ON(crtc_ctx->crtc != crtc)) {
+		/* We have set the crtc when creating secure_display_crtc_context,
 		 * don't expect it to be changed here.
 		 */
-		secure_display_ctx->crtc = crtc;
+		crtc_ctx->crtc = crtc;
 	}
 
-	if (acrtc->dm_irq_params.window_param.update_win) {
-		/* prepare work for dmub to update ROI */
-		secure_display_ctx->rect.x = acrtc->dm_irq_params.window_param.x_start;
-		secure_display_ctx->rect.y = acrtc->dm_irq_params.window_param.y_start;
-		secure_display_ctx->rect.width = acrtc->dm_irq_params.window_param.x_end -
-								acrtc->dm_irq_params.window_param.x_start;
-		secure_display_ctx->rect.height = acrtc->dm_irq_params.window_param.y_end -
-								acrtc->dm_irq_params.window_param.y_start;
-		schedule_work(&secure_display_ctx->forward_roi_work);
-
-		acrtc->dm_irq_params.window_param.update_win = false;
+	for (i = 0; i < MAX_CRC_WINDOW_NUM; i++) {
+		struct crc_params crc_window = {
+			.windowa_x_start = acrtc->dm_irq_params.window_param[i].x_start,
+			.windowa_y_start = acrtc->dm_irq_params.window_param[i].y_start,
+			.windowa_x_end = acrtc->dm_irq_params.window_param[i].x_end,
+			.windowa_y_end = acrtc->dm_irq_params.window_param[i].y_end,
+			.windowb_x_start = acrtc->dm_irq_params.window_param[i].x_start,
+			.windowb_y_start = acrtc->dm_irq_params.window_param[i].y_start,
+			.windowb_x_end = acrtc->dm_irq_params.window_param[i].x_end,
+			.windowb_y_end = acrtc->dm_irq_params.window_param[i].y_end,
+		};
+
+		crtc_ctx->roi[i].enable = acrtc->dm_irq_params.window_param[i].enable;
+
+		if (!acrtc->dm_irq_params.window_param[i].enable) {
+			crtc_ctx->crc_info.crc[i].crc_ready = false;
+			continue;
+		}
 
-		/* Statically skip 1 frame, because we may need to wait below things
-		 * before sending ROI to dmub:
-		 * 1. We defer the work by using system workqueue.
-		 * 2. We may need to wait for dc_lock before accessing dmub.
-		 */
-		acrtc->dm_irq_params.window_param.skip_frame_cnt = 1;
+		if (acrtc->dm_irq_params.window_param[i].skip_frame_cnt) {
+			acrtc->dm_irq_params.window_param[i].skip_frame_cnt -= 1;
+			crtc_ctx->crc_info.crc[i].crc_ready = false;
+			continue;
+		}
 
-	} else {
-		/* prepare work for psp to read ROI/CRC and send to I2C */
-		schedule_work(&secure_display_ctx->notify_ta_work);
+		if (acrtc->dm_irq_params.window_param[i].update_win) {
+			crtc_ctx->roi[i].rect.x = crc_window.windowa_x_start;
+			crtc_ctx->roi[i].rect.y = crc_window.windowa_y_start;
+			crtc_ctx->roi[i].rect.width = crc_window.windowa_x_end -
+						crc_window.windowa_x_start;
+			crtc_ctx->roi[i].rect.height = crc_window.windowa_y_end -
+						crc_window.windowa_y_start;
+
+			if (adev->dm.secure_display_ctx.op_mode == LEGACY_MODE)
+				/* forward task to dmub to update ROI */
+				forward_roi_change = true;
+			else if (adev->dm.secure_display_ctx.op_mode == DISPLAY_CRC_MODE)
+				/* update ROI via dm*/
+				dc_stream_configure_crc(stream_state->ctx->dc, stream_state,
+					&crc_window, true, true, i, false);
+
+			reset_crc_frame_count[i] = true;
+
+			acrtc->dm_irq_params.window_param[i].update_win = false;
+
+			/* Statically skip 1 frame, because we may need to wait below things
+			 * before sending ROI to dmub:
+			 * 1. We defer the work by using system workqueue.
+			 * 2. We may need to wait for dc_lock before accessing dmub.
+			 */
+			acrtc->dm_irq_params.window_param[i].skip_frame_cnt = 1;
+			crtc_ctx->crc_info.crc[i].crc_ready = false;
+		} else {
+			if (!dc_stream_get_crc(stream_state->ctx->dc, stream_state, i,
+						&crc_r[i], &crc_g[i], &crc_b[i]))
+				DRM_ERROR("Secure Display: fail to get crc from engine %d\n", i);
+
+			if (adev->dm.secure_display_ctx.op_mode == LEGACY_MODE)
+				/* forward task to psp to read ROI/CRC and output via I2C */
+				notify_ta = true;
+			else if (adev->dm.secure_display_ctx.op_mode == DISPLAY_CRC_MODE)
+				/* Avoid ROI window get changed, keep overwriting. */
+				dc_stream_configure_crc(stream_state->ctx->dc, stream_state,
+						&crc_window, true, true, i, false);
+
+			/* crc ready for psp to read out */
+			crtc_ctx->crc_info.crc[i].crc_ready = true;
+		}
 	}
 
-cleanup:
 	spin_unlock_irqrestore(&drm_dev->event_lock, flags1);
+
+	if (forward_roi_change)
+		schedule_work(&crtc_ctx->forward_roi_work);
+
+	if (notify_ta)
+		schedule_work(&crtc_ctx->notify_ta_work);
+
+	spin_lock_irqsave(&crtc_ctx->crc_info.lock, flags1);
+	for (i = 0; i < MAX_CRC_WINDOW_NUM; i++) {
+		crtc_ctx->crc_info.crc[i].crc_R = crc_r[i];
+		crtc_ctx->crc_info.crc[i].crc_G = crc_g[i];
+		crtc_ctx->crc_info.crc[i].crc_B = crc_b[i];
+
+		if (!crtc_ctx->roi[i].enable) {
+			crtc_ctx->crc_info.crc[i].frame_count = 0;
+			continue;
+		}
+
+		if (!crtc_ctx->crc_info.crc[i].crc_ready)
+			all_crc_ready = false;
+
+		if (reset_crc_frame_count[i] || crtc_ctx->crc_info.crc[i].frame_count == UINT_MAX)
+			/* Reset the reference frame count after user update the ROI
+			 * or it reaches the maximum value.
+			 */
+			crtc_ctx->crc_info.crc[i].frame_count = 0;
+		else
+			crtc_ctx->crc_info.crc[i].frame_count += 1;
+	}
+	spin_unlock_irqrestore(&crtc_ctx->crc_info.lock, flags1);
+
+	if (all_crc_ready)
+		complete_all(&crtc_ctx->crc_info.completion);
 }
 
-struct secure_display_context *
-amdgpu_dm_crtc_secure_display_create_contexts(struct amdgpu_device *adev)
+void amdgpu_dm_crtc_secure_display_create_contexts(struct amdgpu_device *adev)
 {
-	struct secure_display_context *secure_display_ctxs = NULL;
+	struct secure_display_crtc_context *crtc_ctx = NULL;
 	int i;
 
-	secure_display_ctxs = kcalloc(adev->mode_info.num_crtc,
-				      sizeof(struct secure_display_context),
+	crtc_ctx = kcalloc(adev->mode_info.num_crtc,
+				      sizeof(struct secure_display_crtc_context),
 				      GFP_KERNEL);
 
-	if (!secure_display_ctxs)
-		return NULL;
+	if (!crtc_ctx) {
+		adev->dm.secure_display_ctx.crtc_ctx = NULL;
+		return;
+	}
 
 	for (i = 0; i < adev->mode_info.num_crtc; i++) {
-		INIT_WORK(&secure_display_ctxs[i].forward_roi_work, amdgpu_dm_forward_crc_window);
-		INIT_WORK(&secure_display_ctxs[i].notify_ta_work, amdgpu_dm_crtc_notify_ta_to_read);
-		secure_display_ctxs[i].crtc = &adev->mode_info.crtcs[i]->base;
+		INIT_WORK(&crtc_ctx[i].forward_roi_work, amdgpu_dm_forward_crc_window);
+		INIT_WORK(&crtc_ctx[i].notify_ta_work, amdgpu_dm_crtc_notify_ta_to_read);
+		crtc_ctx[i].crtc = &adev->mode_info.crtcs[i]->base;
+		spin_lock_init(&crtc_ctx[i].crc_info.lock);
 	}
 
-	return secure_display_ctxs;
+	adev->dm.secure_display_ctx.crtc_ctx = crtc_ctx;
+
+	adev->dm.secure_display_ctx.op_mode = DISPLAY_CRC_MODE;
 }
 #endif
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.h
index 935adca6f048..95bdb8699d7f 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.h
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.h
@@ -1,3 +1,4 @@
+/* SPDX-License-Identifier: MIT */
 /*
  * Copyright 2019 Advanced Micro Devices, Inc.
  *
@@ -40,20 +41,53 @@ enum amdgpu_dm_pipe_crc_source {
 };
 
 #ifdef CONFIG_DRM_AMD_SECURE_DISPLAY
+#define MAX_CRTC 6
+
+enum secure_display_mode {
+	/* via dmub + psp */
+	LEGACY_MODE = 0,
+	/* driver directly */
+	DISPLAY_CRC_MODE,
+	SECURE_DISPLAY_MODE_MAX,
+};
+
+struct phy_id_mapping {
+	bool assigned;
+	bool is_mst;
+	uint8_t enc_hw_inst;
+	u8 lct;
+	u8 port_num;
+	u8 rad[8];
+};
+
+struct crc_data {
+	uint32_t crc_R;
+	uint32_t crc_G;
+	uint32_t crc_B;
+	uint32_t frame_count;
+	bool crc_ready;
+};
+
+struct crc_info {
+	struct crc_data crc[MAX_CRC_WINDOW_NUM];
+	struct completion completion;
+	spinlock_t lock;
+};
+
 struct crc_window_param {
 	uint16_t x_start;
 	uint16_t y_start;
 	uint16_t x_end;
 	uint16_t y_end;
 	/* CRC window is activated or not*/
-	bool activated;
+	bool enable;
 	/* Update crc window during vertical blank or not */
 	bool update_win;
 	/* skip reading/writing for few frames */
 	int skip_frame_cnt;
 };
 
-struct secure_display_context {
+struct secure_display_crtc_context {
 	/* work to notify PSP TA*/
 	struct work_struct notify_ta_work;
 
@@ -63,7 +97,20 @@ struct secure_display_context {
 	struct drm_crtc *crtc;
 
 	/* Region of Interest (ROI) */
-	struct rect rect;
+	struct crc_window roi[MAX_CRC_WINDOW_NUM];
+
+	struct crc_info crc_info;
+};
+
+struct secure_display_context {
+
+	struct secure_display_crtc_context *crtc_ctx;
+    /* Whether dmub support multiple ROI setting */
+	bool support_mul_roi;
+	enum secure_display_mode op_mode;
+	bool phy_mapping_updated;
+	int phy_id_mapping_cnt;
+	struct phy_id_mapping phy_id_mapping[MAX_CRTC];
 };
 #endif
 
@@ -95,12 +142,11 @@ void amdgpu_dm_crtc_handle_crc_irq(struct drm_crtc *crtc);
 #ifdef CONFIG_DRM_AMD_SECURE_DISPLAY
 bool amdgpu_dm_crc_window_is_activated(struct drm_crtc *crtc);
 void amdgpu_dm_crtc_handle_crc_window_irq(struct drm_crtc *crtc);
-struct secure_display_context *amdgpu_dm_crtc_secure_display_create_contexts(
-						struct amdgpu_device *adev);
+void amdgpu_dm_crtc_secure_display_create_contexts(struct amdgpu_device *adev);
 #else
 #define amdgpu_dm_crc_window_is_activated(x)
 #define amdgpu_dm_crtc_handle_crc_window_irq(x)
-#define amdgpu_dm_crtc_secure_display_create_contexts()
+#define amdgpu_dm_crtc_secure_display_create_contexts(x)
 #endif
 
 #endif /* AMD_DAL_DEV_AMDGPU_DM_AMDGPU_DM_CRC_H_ */
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crtc.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crtc.c
index e3762e806617..697e232acebf 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crtc.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crtc.c
@@ -29,11 +29,15 @@
 #include "dc.h"
 #include "amdgpu.h"
 #include "amdgpu_dm_psr.h"
+#include "amdgpu_dm_replay.h"
 #include "amdgpu_dm_crtc.h"
 #include "amdgpu_dm_plane.h"
 #include "amdgpu_dm_trace.h"
 #include "amdgpu_dm_debugfs.h"
 
+#define HPD_DETECTION_PERIOD_uS 2000000
+#define HPD_DETECTION_TIME_uS 100000
+
 void amdgpu_dm_crtc_handle_vblank(struct amdgpu_crtc *acrtc)
 {
 	struct drm_crtc *crtc = &acrtc->base;
@@ -89,17 +93,163 @@ int amdgpu_dm_crtc_set_vupdate_irq(struct drm_crtc *crtc, bool enable)
 	return rc;
 }
 
-bool amdgpu_dm_crtc_vrr_active(struct dm_crtc_state *dm_state)
+bool amdgpu_dm_crtc_vrr_active(const struct dm_crtc_state *dm_state)
 {
 	return dm_state->freesync_config.state == VRR_STATE_ACTIVE_VARIABLE ||
 	       dm_state->freesync_config.state == VRR_STATE_ACTIVE_FIXED;
 }
 
-static void vblank_control_worker(struct work_struct *work)
+/**
+ * amdgpu_dm_crtc_set_panel_sr_feature() - Manage panel self-refresh features.
+ *
+ * @vblank_work:    is a pointer to a struct vblank_control_work object.
+ * @vblank_enabled: indicates whether the DRM vblank counter is currently
+ *                  enabled (true) or disabled (false).
+ * @allow_sr_entry: represents whether entry into the self-refresh mode is
+ *                  allowed (true) or not allowed (false).
+ *
+ * The DRM vblank counter enable/disable action is used as the trigger to enable
+ * or disable various panel self-refresh features:
+ *
+ * Panel Replay and PSR SU
+ * - Enable when:
+ *      - VRR is disabled
+ *      - vblank counter is disabled
+ *      - entry is allowed: usermode demonstrates an adequate number of fast
+ *        commits)
+ *     - CRC capture window isn't active
+ * - Keep enabled even when vblank counter gets enabled
+ *
+ * PSR1
+ * - Enable condition same as above
+ * - Disable when vblank counter is enabled
+ */
+static void amdgpu_dm_crtc_set_panel_sr_feature(
+	struct vblank_control_work *vblank_work,
+	bool vblank_enabled, bool allow_sr_entry)
+{
+	struct dc_link *link = vblank_work->stream->link;
+	bool is_sr_active = (link->replay_settings.replay_allow_active ||
+				 link->psr_settings.psr_allow_active);
+	bool is_crc_window_active = false;
+	bool vrr_active = amdgpu_dm_crtc_vrr_active_irq(vblank_work->acrtc);
+
+#ifdef CONFIG_DRM_AMD_SECURE_DISPLAY
+	is_crc_window_active =
+		amdgpu_dm_crc_window_is_activated(&vblank_work->acrtc->base);
+#endif
+
+	if (link->replay_settings.replay_feature_enabled && !vrr_active &&
+		allow_sr_entry && !is_sr_active && !is_crc_window_active) {
+		amdgpu_dm_replay_enable(vblank_work->stream, true);
+	} else if (vblank_enabled) {
+		if (link->psr_settings.psr_version < DC_PSR_VERSION_SU_1 && is_sr_active)
+			amdgpu_dm_psr_disable(vblank_work->stream, false);
+	} else if (link->psr_settings.psr_feature_enabled && !vrr_active &&
+		allow_sr_entry && !is_sr_active && !is_crc_window_active) {
+
+		struct amdgpu_dm_connector *aconn =
+			(struct amdgpu_dm_connector *) vblank_work->stream->dm_stream_context;
+
+		if (!aconn->disallow_edp_enter_psr) {
+			struct amdgpu_display_manager *dm = vblank_work->dm;
+
+			amdgpu_dm_psr_enable(vblank_work->stream);
+			if (dm->idle_workqueue &&
+			    (dm->dc->config.disable_ips == DMUB_IPS_ENABLE) &&
+			    dm->dc->idle_optimizations_allowed &&
+			    dm->idle_workqueue->enable &&
+			    !dm->idle_workqueue->running)
+				schedule_work(&dm->idle_workqueue->work);
+		}
+	}
+}
+
+bool amdgpu_dm_is_headless(struct amdgpu_device *adev)
+{
+	struct drm_connector *connector;
+	struct drm_connector_list_iter iter;
+	struct drm_device *dev;
+	bool is_headless = true;
+
+	if (adev == NULL)
+		return true;
+
+	dev = adev->dm.ddev;
+
+	drm_connector_list_iter_begin(dev, &iter);
+	drm_for_each_connector_iter(connector, &iter) {
+
+		if (connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+			continue;
+
+		if (connector->status == connector_status_connected) {
+			is_headless = false;
+			break;
+		}
+	}
+	drm_connector_list_iter_end(&iter);
+	return is_headless;
+}
+
+static void amdgpu_dm_idle_worker(struct work_struct *work)
+{
+	struct idle_workqueue *idle_work;
+
+	idle_work = container_of(work, struct idle_workqueue, work);
+	idle_work->dm->idle_workqueue->running = true;
+
+	while (idle_work->enable) {
+		fsleep(HPD_DETECTION_PERIOD_uS);
+		mutex_lock(&idle_work->dm->dc_lock);
+		if (!idle_work->dm->dc->idle_optimizations_allowed) {
+			mutex_unlock(&idle_work->dm->dc_lock);
+			break;
+		}
+		dc_allow_idle_optimizations(idle_work->dm->dc, false);
+
+		mutex_unlock(&idle_work->dm->dc_lock);
+		fsleep(HPD_DETECTION_TIME_uS);
+		mutex_lock(&idle_work->dm->dc_lock);
+
+		if (!amdgpu_dm_is_headless(idle_work->dm->adev) &&
+		    !amdgpu_dm_psr_is_active_allowed(idle_work->dm)) {
+			mutex_unlock(&idle_work->dm->dc_lock);
+			break;
+		}
+
+		if (idle_work->enable) {
+			dc_post_update_surfaces_to_stream(idle_work->dm->dc);
+			dc_allow_idle_optimizations(idle_work->dm->dc, true);
+		}
+		mutex_unlock(&idle_work->dm->dc_lock);
+	}
+	idle_work->dm->idle_workqueue->running = false;
+}
+
+struct idle_workqueue *idle_create_workqueue(struct amdgpu_device *adev)
+{
+	struct idle_workqueue *idle_work;
+
+	idle_work = kzalloc(sizeof(*idle_work), GFP_KERNEL);
+	if (ZERO_OR_NULL_PTR(idle_work))
+		return NULL;
+
+	idle_work->dm = &adev->dm;
+	idle_work->enable = false;
+	idle_work->running = false;
+	INIT_WORK(&idle_work->work, amdgpu_dm_idle_worker);
+
+	return idle_work;
+}
+
+static void amdgpu_dm_crtc_vblank_control_worker(struct work_struct *work)
 {
 	struct vblank_control_work *vblank_work =
 		container_of(work, struct vblank_control_work, work);
 	struct amdgpu_display_manager *dm = vblank_work->dm;
+	struct amdgpu_device *adev = drm_to_adev(dm->ddev);
+	int r;
 
 	mutex_lock(&dm->dc_lock);
 
@@ -108,9 +258,8 @@ static void vblank_control_worker(struct work_struct *work)
 	else if (dm->active_vblank_irq_count)
 		dm->active_vblank_irq_count--;
 
-	dc_allow_idle_optimizations(dm->dc, dm->active_vblank_irq_count == 0);
-
-	DRM_DEBUG_KMS("Allow idle optimizations (MALL): %d\n", dm->active_vblank_irq_count == 0);
+	if (dm->active_vblank_irq_count > 0)
+		dc_allow_idle_optimizations(dm->dc, false);
 
 	/*
 	 * Control PSR based on vblank requirements from OS
@@ -122,19 +271,24 @@ static void vblank_control_worker(struct work_struct *work)
 	 * where the SU region is the full hactive*vactive region. See
 	 * fill_dc_dirty_rects().
 	 */
-	if (vblank_work->stream && vblank_work->stream->link) {
-		if (vblank_work->enable) {
-			if (vblank_work->stream->link->psr_settings.psr_version < DC_PSR_VERSION_SU_1 &&
-			    vblank_work->stream->link->psr_settings.psr_allow_active)
-				amdgpu_dm_psr_disable(vblank_work->stream);
-		} else if (vblank_work->stream->link->psr_settings.psr_feature_enabled &&
-			   !vblank_work->stream->link->psr_settings.psr_allow_active &&
-#ifdef CONFIG_DRM_AMD_SECURE_DISPLAY
-			   !amdgpu_dm_crc_window_is_activated(&vblank_work->acrtc->base) &&
-#endif
-			   vblank_work->acrtc->dm_irq_params.allow_psr_entry) {
-			amdgpu_dm_psr_enable(vblank_work->stream);
-		}
+	if (vblank_work->stream && vblank_work->stream->link && vblank_work->acrtc) {
+		amdgpu_dm_crtc_set_panel_sr_feature(
+			vblank_work, vblank_work->enable,
+			vblank_work->acrtc->dm_irq_params.allow_sr_entry);
+	}
+
+	if (dm->active_vblank_irq_count == 0) {
+		dc_post_update_surfaces_to_stream(dm->dc);
+
+		r = amdgpu_dpm_pause_power_profile(adev, true);
+		if (r)
+			dev_warn(adev->dev, "failed to set default power profile mode\n");
+
+		dc_allow_idle_optimizations(dm->dc, true);
+
+		r = amdgpu_dpm_pause_power_profile(adev, false);
+		if (r)
+			dev_warn(adev->dev, "failed to restore the power profile mode\n");
 	}
 
 	mutex_unlock(&dm->dc_lock);
@@ -144,48 +298,107 @@ static void vblank_control_worker(struct work_struct *work)
 	kfree(vblank_work);
 }
 
-static inline int dm_set_vblank(struct drm_crtc *crtc, bool enable)
+static inline int amdgpu_dm_crtc_set_vblank(struct drm_crtc *crtc, bool enable)
 {
-	enum dc_irq_source irq_source;
 	struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
 	struct amdgpu_device *adev = drm_to_adev(crtc->dev);
 	struct dm_crtc_state *acrtc_state = to_dm_crtc_state(crtc->state);
 	struct amdgpu_display_manager *dm = &adev->dm;
 	struct vblank_control_work *work;
+	int irq_type;
 	int rc = 0;
 
-	if (acrtc->otg_inst == -1)
-		goto skip;
+	if (enable && !acrtc->base.enabled) {
+		drm_dbg_vbl(crtc->dev,
+				"Reject vblank enable on unconfigured CRTC %d (enabled=%d)\n",
+				acrtc->crtc_id, acrtc->base.enabled);
+		return -EINVAL;
+	}
+
+	irq_type = amdgpu_display_crtc_idx_to_irq_type(adev, acrtc->crtc_id);
+
+	if (enable) {
+		struct dc *dc = adev->dm.dc;
+		struct drm_vblank_crtc *vblank = drm_crtc_vblank_crtc(crtc);
+		struct psr_settings *psr = &acrtc_state->stream->link->psr_settings;
+		struct replay_settings *pr = &acrtc_state->stream->link->replay_settings;
+		bool sr_supported = (psr->psr_version != DC_PSR_VERSION_UNSUPPORTED) ||
+								pr->config.replay_supported;
+
+		/*
+		 * IPS & self-refresh feature can cause vblank counter resets between
+		 * vblank disable and enable.
+		 * It may cause system stuck due to waiting for the vblank counter.
+		 * Call this function to estimate missed vblanks by using timestamps and
+		 * update the vblank counter in DRM.
+		 */
+		if (dc->caps.ips_support &&
+			dc->config.disable_ips != DMUB_IPS_DISABLE_ALL &&
+			sr_supported && vblank->config.disable_immediate)
+			drm_crtc_vblank_restore(crtc);
+	}
+
+	if (dc_supports_vrr(dm->dc->ctx->dce_version)) {
+		if (enable) {
+			/* vblank irq on -> Only need vupdate irq in vrr mode */
+			if (amdgpu_dm_crtc_vrr_active(acrtc_state))
+				rc = amdgpu_dm_crtc_set_vupdate_irq(crtc, true);
+		} else {
+			/* vblank irq off -> vupdate irq off */
+			rc = amdgpu_dm_crtc_set_vupdate_irq(crtc, false);
+		}
+	}
+
+	if (rc)
+		return rc;
 
+	/* crtc vblank or vstartup interrupt */
 	if (enable) {
-		/* vblank irq on -> Only need vupdate irq in vrr mode */
-		if (amdgpu_dm_crtc_vrr_active(acrtc_state))
-			rc = amdgpu_dm_crtc_set_vupdate_irq(crtc, true);
+		rc = amdgpu_irq_get(adev, &adev->crtc_irq, irq_type);
+		drm_dbg_vbl(crtc->dev, "Get crtc_irq ret=%d\n", rc);
 	} else {
-		/* vblank irq off -> vupdate irq off */
-		rc = amdgpu_dm_crtc_set_vupdate_irq(crtc, false);
+		rc = amdgpu_irq_put(adev, &adev->crtc_irq, irq_type);
+		drm_dbg_vbl(crtc->dev, "Put crtc_irq ret=%d\n", rc);
 	}
 
 	if (rc)
 		return rc;
 
-	if (amdgpu_in_reset(adev)) {
-		irq_source = IRQ_TYPE_VBLANK + acrtc->otg_inst;
-		/* During gpu-reset we disable and then enable vblank irq, so
-		 * don't use amdgpu_irq_get/put() to avoid refcount change.
-		 */
-		if (!dc_interrupt_set(adev->dm.dc, irq_source, enable))
-			rc = -EBUSY;
+	/*
+	 * hubp surface flip interrupt
+	 *
+	 * We have no guarantee that the frontend index maps to the same
+	 * backend index - some even map to more than one.
+	 *
+	 * TODO: Use a different interrupt or check DC itself for the mapping.
+	 */
+	if (enable) {
+		rc = amdgpu_irq_get(adev, &adev->pageflip_irq, irq_type);
+		drm_dbg_vbl(crtc->dev, "Get pageflip_irq ret=%d\n", rc);
 	} else {
-		rc = (enable)
-			? amdgpu_irq_get(adev, &adev->crtc_irq, acrtc->crtc_id)
-			: amdgpu_irq_put(adev, &adev->crtc_irq, acrtc->crtc_id);
+		rc = amdgpu_irq_put(adev, &adev->pageflip_irq, irq_type);
+		drm_dbg_vbl(crtc->dev, "Put pageflip_irq ret=%d\n", rc);
 	}
 
 	if (rc)
 		return rc;
 
-skip:
+#if defined(CONFIG_DRM_AMD_SECURE_DISPLAY)
+	/* crtc vline0 interrupt, only available on DCN+ */
+	if (amdgpu_ip_version(adev, DCE_HWIP, 0) != 0) {
+		if (enable) {
+			rc = amdgpu_irq_get(adev, &adev->vline0_irq, irq_type);
+			drm_dbg_vbl(crtc->dev, "Get vline0_irq ret=%d\n", rc);
+		} else {
+			rc = amdgpu_irq_put(adev, &adev->vline0_irq, irq_type);
+			drm_dbg_vbl(crtc->dev, "Put vline0_irq ret=%d\n", rc);
+		}
+
+		if (rc)
+			return rc;
+	}
+#endif
+
 	if (amdgpu_in_reset(adev))
 		return 0;
 
@@ -194,7 +407,7 @@ skip:
 		if (!work)
 			return -ENOMEM;
 
-		INIT_WORK(&work->work, vblank_control_worker);
+		INIT_WORK(&work->work, amdgpu_dm_crtc_vblank_control_worker);
 		work->dm = dm;
 		work->acrtc = acrtc;
 		work->enable = enable;
@@ -212,15 +425,15 @@ skip:
 
 int amdgpu_dm_crtc_enable_vblank(struct drm_crtc *crtc)
 {
-	return dm_set_vblank(crtc, true);
+	return amdgpu_dm_crtc_set_vblank(crtc, true);
 }
 
 void amdgpu_dm_crtc_disable_vblank(struct drm_crtc *crtc)
 {
-	dm_set_vblank(crtc, false);
+	amdgpu_dm_crtc_set_vblank(crtc, false);
 }
 
-static void dm_crtc_destroy_state(struct drm_crtc *crtc,
+static void amdgpu_dm_crtc_destroy_state(struct drm_crtc *crtc,
 				  struct drm_crtc_state *state)
 {
 	struct dm_crtc_state *cur = to_dm_crtc_state(state);
@@ -236,7 +449,7 @@ static void dm_crtc_destroy_state(struct drm_crtc *crtc,
 	kfree(state);
 }
 
-static struct drm_crtc_state *dm_crtc_duplicate_state(struct drm_crtc *crtc)
+static struct drm_crtc_state *amdgpu_dm_crtc_duplicate_state(struct drm_crtc *crtc)
 {
 	struct dm_crtc_state *state, *cur;
 
@@ -263,8 +476,10 @@ static struct drm_crtc_state *dm_crtc_duplicate_state(struct drm_crtc *crtc)
 	state->freesync_config = cur->freesync_config;
 	state->cm_has_degamma = cur->cm_has_degamma;
 	state->cm_is_degamma_srgb = cur->cm_is_degamma_srgb;
+	state->regamma_tf = cur->regamma_tf;
 	state->crc_skip_count = cur->crc_skip_count;
 	state->mpo_requested = cur->mpo_requested;
+	state->cursor_mode = cur->cursor_mode;
 	/* TODO Duplicate dc_stream after objects are stream object is flattened */
 
 	return &state->base;
@@ -276,12 +491,12 @@ static void amdgpu_dm_crtc_destroy(struct drm_crtc *crtc)
 	kfree(crtc);
 }
 
-static void dm_crtc_reset_state(struct drm_crtc *crtc)
+static void amdgpu_dm_crtc_reset_state(struct drm_crtc *crtc)
 {
 	struct dm_crtc_state *state;
 
 	if (crtc->state)
-		dm_crtc_destroy_state(crtc, crtc->state);
+		amdgpu_dm_crtc_destroy_state(crtc, crtc->state);
 
 	state = kzalloc(sizeof(*state), GFP_KERNEL);
 	if (WARN_ON(!state))
@@ -299,14 +514,78 @@ static int amdgpu_dm_crtc_late_register(struct drm_crtc *crtc)
 }
 #endif
 
+#ifdef AMD_PRIVATE_COLOR
+/**
+ * dm_crtc_additional_color_mgmt - enable additional color properties
+ * @crtc: DRM CRTC
+ *
+ * This function lets the driver enable post-blending CRTC regamma transfer
+ * function property in addition to DRM CRTC gamma LUT. Default value means
+ * linear transfer function, which is the default CRTC gamma LUT behaviour
+ * without this property.
+ */
+static void
+dm_crtc_additional_color_mgmt(struct drm_crtc *crtc)
+{
+	struct amdgpu_device *adev = drm_to_adev(crtc->dev);
+
+	if (adev->dm.dc->caps.color.mpc.ogam_ram)
+		drm_object_attach_property(&crtc->base,
+					   adev->mode_info.regamma_tf_property,
+					   AMDGPU_TRANSFER_FUNCTION_DEFAULT);
+}
+
+static int
+amdgpu_dm_atomic_crtc_set_property(struct drm_crtc *crtc,
+				   struct drm_crtc_state *state,
+				   struct drm_property *property,
+				   uint64_t val)
+{
+	struct amdgpu_device *adev = drm_to_adev(crtc->dev);
+	struct dm_crtc_state *acrtc_state = to_dm_crtc_state(state);
+
+	if (property == adev->mode_info.regamma_tf_property) {
+		if (acrtc_state->regamma_tf != val) {
+			acrtc_state->regamma_tf = val;
+			acrtc_state->base.color_mgmt_changed |= 1;
+		}
+	} else {
+		drm_dbg_atomic(crtc->dev,
+			       "[CRTC:%d:%s] unknown property [PROP:%d:%s]]\n",
+			       crtc->base.id, crtc->name,
+			       property->base.id, property->name);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int
+amdgpu_dm_atomic_crtc_get_property(struct drm_crtc *crtc,
+				   const struct drm_crtc_state *state,
+				   struct drm_property *property,
+				   uint64_t *val)
+{
+	struct amdgpu_device *adev = drm_to_adev(crtc->dev);
+	struct dm_crtc_state *acrtc_state = to_dm_crtc_state(state);
+
+	if (property == adev->mode_info.regamma_tf_property)
+		*val = acrtc_state->regamma_tf;
+	else
+		return -EINVAL;
+
+	return 0;
+}
+#endif
+
 /* Implemented only the options currently available for the driver */
 static const struct drm_crtc_funcs amdgpu_dm_crtc_funcs = {
-	.reset = dm_crtc_reset_state,
+	.reset = amdgpu_dm_crtc_reset_state,
 	.destroy = amdgpu_dm_crtc_destroy,
 	.set_config = drm_atomic_helper_set_config,
 	.page_flip = drm_atomic_helper_page_flip,
-	.atomic_duplicate_state = dm_crtc_duplicate_state,
-	.atomic_destroy_state = dm_crtc_destroy_state,
+	.atomic_duplicate_state = amdgpu_dm_crtc_duplicate_state,
+	.atomic_destroy_state = amdgpu_dm_crtc_destroy_state,
 	.set_crc_source = amdgpu_dm_crtc_set_crc_source,
 	.verify_crc_source = amdgpu_dm_crtc_verify_crc_source,
 	.get_crc_sources = amdgpu_dm_crtc_get_crc_sources,
@@ -317,13 +596,17 @@ static const struct drm_crtc_funcs amdgpu_dm_crtc_funcs = {
 #if defined(CONFIG_DEBUG_FS)
 	.late_register = amdgpu_dm_crtc_late_register,
 #endif
+#ifdef AMD_PRIVATE_COLOR
+	.atomic_set_property = amdgpu_dm_atomic_crtc_set_property,
+	.atomic_get_property = amdgpu_dm_atomic_crtc_get_property,
+#endif
 };
 
-static void dm_crtc_helper_disable(struct drm_crtc *crtc)
+static void amdgpu_dm_crtc_helper_disable(struct drm_crtc *crtc)
 {
 }
 
-static int count_crtc_active_planes(struct drm_crtc_state *new_crtc_state)
+static int amdgpu_dm_crtc_count_crtc_active_planes(struct drm_crtc_state *new_crtc_state)
 {
 	struct drm_atomic_state *state = new_crtc_state->state;
 	struct drm_plane *plane;
@@ -355,8 +638,8 @@ static int count_crtc_active_planes(struct drm_crtc_state *new_crtc_state)
 	return num_active;
 }
 
-static void dm_update_crtc_active_planes(struct drm_crtc *crtc,
-					 struct drm_crtc_state *new_crtc_state)
+static void amdgpu_dm_crtc_update_crtc_active_planes(struct drm_crtc *crtc,
+						     struct drm_crtc_state *new_crtc_state)
 {
 	struct dm_crtc_state *dm_new_crtc_state =
 		to_dm_crtc_state(new_crtc_state);
@@ -367,18 +650,18 @@ static void dm_update_crtc_active_planes(struct drm_crtc *crtc,
 		return;
 
 	dm_new_crtc_state->active_planes =
-		count_crtc_active_planes(new_crtc_state);
+		amdgpu_dm_crtc_count_crtc_active_planes(new_crtc_state);
 }
 
-static bool dm_crtc_helper_mode_fixup(struct drm_crtc *crtc,
+static bool amdgpu_dm_crtc_helper_mode_fixup(struct drm_crtc *crtc,
 				      const struct drm_display_mode *mode,
 				      struct drm_display_mode *adjusted_mode)
 {
 	return true;
 }
 
-static int dm_crtc_helper_atomic_check(struct drm_crtc *crtc,
-				      struct drm_atomic_state *state)
+static int amdgpu_dm_crtc_helper_atomic_check(struct drm_crtc *crtc,
+					      struct drm_atomic_state *state)
 {
 	struct drm_crtc_state *crtc_state = drm_atomic_get_new_crtc_state(state,
 										crtc);
@@ -389,7 +672,7 @@ static int dm_crtc_helper_atomic_check(struct drm_crtc *crtc,
 
 	trace_amdgpu_dm_crtc_atomic_check(crtc_state);
 
-	dm_update_crtc_active_planes(crtc, crtc_state);
+	amdgpu_dm_crtc_update_crtc_active_planes(crtc, crtc_state);
 
 	if (WARN_ON(unlikely(!dm_crtc_state->stream &&
 			amdgpu_dm_crtc_modeset_required(crtc_state, NULL, dm_crtc_state->stream)))) {
@@ -408,6 +691,27 @@ static int dm_crtc_helper_atomic_check(struct drm_crtc *crtc,
 		return -EINVAL;
 	}
 
+	/*
+	 * Only allow async flips for fast updates that don't change the FB
+	 * pitch, the DCC state, rotation, etc.
+	 */
+	if (crtc_state->async_flip &&
+	    dm_crtc_state->update_type != UPDATE_TYPE_FAST) {
+		drm_dbg_atomic(crtc->dev,
+			       "[CRTC:%d:%s] async flips are only supported for fast updates\n",
+			       crtc->base.id, crtc->name);
+		return -EINVAL;
+	}
+
+	if (!state->legacy_cursor_update && amdgpu_dm_crtc_vrr_active(dm_crtc_state)) {
+		struct drm_plane_state *primary_state;
+
+		/* Pull in primary plane for correct VRR handling */
+		primary_state = drm_atomic_get_plane_state(state, crtc->primary);
+		if (IS_ERR(primary_state))
+			return PTR_ERR(primary_state);
+	}
+
 	/* In some use cases, like reset, no stream is attached */
 	if (!dm_crtc_state->stream)
 		return 0;
@@ -420,9 +724,9 @@ static int dm_crtc_helper_atomic_check(struct drm_crtc *crtc,
 }
 
 static const struct drm_crtc_helper_funcs amdgpu_dm_crtc_helper_funcs = {
-	.disable = dm_crtc_helper_disable,
-	.atomic_check = dm_crtc_helper_atomic_check,
-	.mode_fixup = dm_crtc_helper_mode_fixup,
+	.disable = amdgpu_dm_crtc_helper_disable,
+	.atomic_check = amdgpu_dm_crtc_helper_atomic_check,
+	.mode_fixup = amdgpu_dm_crtc_helper_mode_fixup,
 	.get_scanout_position = amdgpu_crtc_get_scanout_position,
 };
 
@@ -432,7 +736,7 @@ int amdgpu_dm_crtc_init(struct amdgpu_display_manager *dm,
 {
 	struct amdgpu_crtc *acrtc = NULL;
 	struct drm_plane *cursor_plane;
-	bool is_dcn;
+	bool has_degamma;
 	int res = -ENOMEM;
 
 	cursor_plane = kzalloc(sizeof(*cursor_plane), GFP_KERNEL);
@@ -471,15 +775,25 @@ int amdgpu_dm_crtc_init(struct amdgpu_display_manager *dm,
 
 	dm->adev->mode_info.crtcs[crtc_index] = acrtc;
 
-	/* Don't enable DRM CRTC degamma property for DCE since it doesn't
-	 * support programmable degamma anywhere.
+	/* Don't enable DRM CRTC degamma property for
+	 * 1. Degamma is replaced by color pipeline.
+	 * 2. DCE since it doesn't support programmable degamma anywhere.
+	 * 3. DCN401 since pre-blending degamma LUT doesn't apply to cursor.
 	 */
-	is_dcn = dm->adev->dm.dc->caps.color.dpp.dcn_arch;
-	drm_crtc_enable_color_mgmt(&acrtc->base, is_dcn ? MAX_COLOR_LUT_ENTRIES : 0,
+	if (plane->color_pipeline_property)
+		has_degamma = false;
+	else
+		has_degamma = dm->adev->dm.dc->caps.color.dpp.dcn_arch &&
+			      dm->adev->dm.dc->ctx->dce_version != DCN_VERSION_4_01;
+
+	drm_crtc_enable_color_mgmt(&acrtc->base, has_degamma ? MAX_COLOR_LUT_ENTRIES : 0,
 				   true, MAX_COLOR_LUT_ENTRIES);
 
 	drm_mode_crtc_set_gamma_size(&acrtc->base, MAX_COLOR_LEGACY_LUT_ENTRIES);
 
+#ifdef AMD_PRIVATE_COLOR
+	dm_crtc_additional_color_mgmt(&acrtc->base);
+#endif
 	return 0;
 
 fail:
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crtc.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crtc.h
index 17e948753f59..c1212947a77b 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crtc.h
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crtc.h
@@ -37,7 +37,7 @@ int amdgpu_dm_crtc_set_vupdate_irq(struct drm_crtc *crtc, bool enable);
 
 bool amdgpu_dm_crtc_vrr_active_irq(struct amdgpu_crtc *acrtc);
 
-bool amdgpu_dm_crtc_vrr_active(struct dm_crtc_state *dm_state);
+bool amdgpu_dm_crtc_vrr_active(const struct dm_crtc_state *dm_state);
 
 int amdgpu_dm_crtc_enable_vblank(struct drm_crtc *crtc);
 
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_debugfs.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_debugfs.c
index 827fcb4fb3b3..a9839485f2a2 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_debugfs.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_debugfs.c
@@ -1,3 +1,4 @@
+// SPDX-License-Identifier: MIT
 /*
  * Copyright 2018 Advanced Micro Devices, Inc.
  *
@@ -25,11 +26,13 @@
 
 #include <linux/string_helpers.h>
 #include <linux/uaccess.h>
+#include <media/cec-notifier.h>
 
 #include "dc.h"
 #include "amdgpu.h"
 #include "amdgpu_dm.h"
 #include "amdgpu_dm_debugfs.h"
+#include "amdgpu_dm_replay.h"
 #include "dm_helpers.h"
 #include "dmub/dmub_srv.h"
 #include "resource.h"
@@ -37,6 +40,7 @@
 #include "link_hwss.h"
 #include "dc/dc_dmub_srv.h"
 #include "link/protocols/link_dp_capability.h"
+#include "inc/hw/dchubbub.h"
 
 #ifdef CONFIG_DRM_AMD_SECURE_DISPLAY
 #include "amdgpu_dm_psr.h"
@@ -256,7 +260,7 @@ static ssize_t dp_link_settings_write(struct file *f, const char __user *buf,
 	struct dc_link *link = connector->dc_link;
 	struct amdgpu_device *adev = drm_to_adev(connector->base.dev);
 	struct dc *dc = (struct dc *)link->dc;
-	struct dc_link_settings prefer_link_settings;
+	struct dc_link_settings prefer_link_settings = {0};
 	char *wr_buf = NULL;
 	const uint32_t wr_buf_size = 40;
 	/* 0: lane_count; 1: link_rate */
@@ -336,6 +340,153 @@ static ssize_t dp_link_settings_write(struct file *f, const char __user *buf,
 	return size;
 }
 
+static bool dp_mst_is_end_device(struct amdgpu_dm_connector *aconnector)
+{
+	bool is_end_device = false;
+	struct drm_dp_mst_topology_mgr *mgr = NULL;
+	struct drm_dp_mst_port *port = NULL;
+
+	if (aconnector->mst_root && aconnector->mst_root->mst_mgr.mst_state) {
+		mgr = &aconnector->mst_root->mst_mgr;
+		port = aconnector->mst_output_port;
+
+		drm_modeset_lock(&mgr->base.lock, NULL);
+		if (port->pdt == DP_PEER_DEVICE_SST_SINK ||
+			port->pdt == DP_PEER_DEVICE_DP_LEGACY_CONV)
+			is_end_device = true;
+		drm_modeset_unlock(&mgr->base.lock);
+	}
+
+	return is_end_device;
+}
+
+/* Change MST link setting
+ *
+ * valid lane count value: 1, 2, 4
+ * valid link rate value:
+ * 06h = 1.62Gbps per lane
+ * 0Ah = 2.7Gbps per lane
+ * 0Ch = 3.24Gbps per lane
+ * 14h = 5.4Gbps per lane
+ * 1Eh = 8.1Gbps per lane
+ * 3E8h = 10.0Gbps per lane
+ * 546h = 13.5Gbps per lane
+ * 7D0h = 20.0Gbps per lane
+ *
+ * debugfs is located at /sys/kernel/debug/dri/0/DP-x/mst_link_settings
+ *
+ * for example, to force to  2 lane, 10.0GHz,
+ * echo 2 0x3e8 > /sys/kernel/debug/dri/0/DP-x/mst_link_settings
+ *
+ * Valid input will trigger hotplug event to get new link setting applied
+ * Invalid input will trigger training setting reset
+ *
+ * The usage can be referred to link_settings entry
+ *
+ */
+static ssize_t dp_mst_link_setting(struct file *f, const char __user *buf,
+				 size_t size, loff_t *pos)
+{
+	struct amdgpu_dm_connector *aconnector = file_inode(f)->i_private;
+	struct dc_link *link = aconnector->dc_link;
+	struct amdgpu_device *adev = drm_to_adev(aconnector->base.dev);
+	struct dc *dc = (struct dc *)link->dc;
+	struct dc_link_settings prefer_link_settings = {0};
+	char *wr_buf = NULL;
+	const uint32_t wr_buf_size = 40;
+	/* 0: lane_count; 1: link_rate */
+	int max_param_num = 2;
+	uint8_t param_nums = 0;
+	long param[2];
+	bool valid_input = true;
+
+	if (!dp_mst_is_end_device(aconnector))
+		return -EINVAL;
+
+	if (size == 0)
+		return -EINVAL;
+
+	wr_buf = kcalloc(wr_buf_size, sizeof(char), GFP_KERNEL);
+	if (!wr_buf)
+		return -ENOSPC;
+
+	if (parse_write_buffer_into_params(wr_buf, wr_buf_size,
+					   (long *)param, buf,
+					   max_param_num,
+					   &param_nums)) {
+		kfree(wr_buf);
+		return -EINVAL;
+	}
+
+	if (param_nums <= 0) {
+		kfree(wr_buf);
+		DRM_DEBUG_DRIVER("user data not be read\n");
+		return -EINVAL;
+	}
+
+	switch (param[0]) {
+	case LANE_COUNT_ONE:
+	case LANE_COUNT_TWO:
+	case LANE_COUNT_FOUR:
+		break;
+	default:
+		valid_input = false;
+		break;
+	}
+
+	switch (param[1]) {
+	case LINK_RATE_LOW:
+	case LINK_RATE_HIGH:
+	case LINK_RATE_RBR2:
+	case LINK_RATE_HIGH2:
+	case LINK_RATE_HIGH3:
+	case LINK_RATE_UHBR10:
+	case LINK_RATE_UHBR13_5:
+	case LINK_RATE_UHBR20:
+		break;
+	default:
+		valid_input = false;
+		break;
+	}
+
+	if (!valid_input) {
+		kfree(wr_buf);
+		DRM_DEBUG_DRIVER("Invalid Input value No HW will be programmed\n");
+		mutex_lock(&adev->dm.dc_lock);
+		dc_link_set_preferred_training_settings(dc, NULL, NULL, link, false);
+		mutex_unlock(&adev->dm.dc_lock);
+		return -EINVAL;
+	}
+
+	/* save user force lane_count, link_rate to preferred settings
+	 * spread spectrum will not be changed
+	 */
+	prefer_link_settings.link_spread = link->cur_link_settings.link_spread;
+	prefer_link_settings.use_link_rate_set = false;
+	prefer_link_settings.lane_count = param[0];
+	prefer_link_settings.link_rate = param[1];
+
+	/* skip immediate retrain, and train to new link setting after hotplug event triggered */
+	mutex_lock(&adev->dm.dc_lock);
+	dc_link_set_preferred_training_settings(dc, &prefer_link_settings, NULL, link, true);
+	mutex_unlock(&adev->dm.dc_lock);
+
+	mutex_lock(&aconnector->base.dev->mode_config.mutex);
+	aconnector->base.force = DRM_FORCE_OFF;
+	mutex_unlock(&aconnector->base.dev->mode_config.mutex);
+	drm_kms_helper_hotplug_event(aconnector->base.dev);
+
+	msleep(100);
+
+	mutex_lock(&aconnector->base.dev->mode_config.mutex);
+	aconnector->base.force = DRM_FORCE_UNSPECIFIED;
+	mutex_unlock(&aconnector->base.dev->mode_config.mutex);
+	drm_kms_helper_hotplug_event(aconnector->base.dev);
+
+	kfree(wr_buf);
+	return size;
+}
+
 /* function: get current DP PHY settings: voltage swing, pre-emphasis,
  * post-cursor2 (defined by VESA DP specification)
  *
@@ -464,7 +615,7 @@ static ssize_t dp_phy_settings_write(struct file *f, const char __user *buf,
 	uint32_t wr_buf_size = 40;
 	long param[3];
 	bool use_prefer_link_setting;
-	struct link_training_settings link_lane_settings;
+	struct link_training_settings link_lane_settings = {0};
 	int max_param_num = 3;
 	uint8_t param_nums = 0;
 	int r = 0;
@@ -608,6 +759,7 @@ static ssize_t dp_phy_test_pattern_debugfs_write(struct file *f, const char __us
 	int max_param_num = 11;
 	enum dp_test_pattern test_pattern = DP_TEST_PATTERN_UNSUPPORTED;
 	bool disable_hpd = false;
+	bool supports_hpd = link->irq_source_hpd != DC_IRQ_SOURCE_INVALID;
 	bool valid_test_pattern = false;
 	uint8_t param_nums = 0;
 	/* init with default 80bit custom pattern */
@@ -619,7 +771,7 @@ static ssize_t dp_phy_test_pattern_debugfs_write(struct file *f, const char __us
 			LINK_RATE_UNKNOWN, LINK_SPREAD_DISABLED};
 	struct dc_link_settings cur_link_settings = {LANE_COUNT_UNKNOWN,
 			LINK_RATE_UNKNOWN, LINK_SPREAD_DISABLED};
-	struct link_training_settings link_training_settings;
+	struct link_training_settings link_training_settings = {0};
 	int i;
 
 	if (size == 0)
@@ -699,7 +851,7 @@ static ssize_t dp_phy_test_pattern_debugfs_write(struct file *f, const char __us
 	 * because it might have been disabled after a test pattern was set.
 	 * AUX depends on HPD * sequence dependent, do not move!
 	 */
-	if (!disable_hpd)
+	if (supports_hpd && !disable_hpd)
 		dc_link_enable_hpd(link);
 
 	prefer_link_settings.lane_count = link->verified_link_cap.lane_count;
@@ -737,7 +889,7 @@ static ssize_t dp_phy_test_pattern_debugfs_write(struct file *f, const char __us
 	 * Need disable interrupt to avoid SW driver disable DP output. This is
 	 * done after the test pattern is set.
 	 */
-	if (valid_test_pattern && disable_hpd)
+	if (valid_test_pattern && supports_hpd && disable_hpd)
 		dc_link_disable_hpd(link);
 
 	kfree(wr_buf);
@@ -753,9 +905,10 @@ static int dmub_tracebuffer_show(struct seq_file *m, void *data)
 {
 	struct amdgpu_device *adev = m->private;
 	struct dmub_srv_fb_info *fb_info = adev->dm.dmub_fb_info;
+	struct dmub_fw_meta_info *fw_meta_info = NULL;
 	struct dmub_debugfs_trace_entry *entries;
 	uint8_t *tbuf_base;
-	uint32_t tbuf_size, max_entries, num_entries, i;
+	uint32_t tbuf_size, max_entries, num_entries, first_entry, i;
 
 	if (!fb_info)
 		return 0;
@@ -764,20 +917,42 @@ static int dmub_tracebuffer_show(struct seq_file *m, void *data)
 	if (!tbuf_base)
 		return 0;
 
-	tbuf_size = fb_info->fb[DMUB_WINDOW_5_TRACEBUFF].size;
+	if (adev->dm.dmub_srv)
+		fw_meta_info = &adev->dm.dmub_srv->meta_info;
+
+	tbuf_size = fw_meta_info ? fw_meta_info->trace_buffer_size :
+				   DMUB_TRACE_BUFFER_SIZE;
 	max_entries = (tbuf_size - sizeof(struct dmub_debugfs_trace_header)) /
 		      sizeof(struct dmub_debugfs_trace_entry);
 
 	num_entries =
 		((struct dmub_debugfs_trace_header *)tbuf_base)->entry_count;
 
+	/* DMCUB tracebuffer is a ring. If it rolled over, print a hint that
+	 * entries are being overwritten.
+	 */
+	if (num_entries > max_entries)
+		seq_printf(m, "...\n");
+
+	first_entry = num_entries % max_entries;
 	num_entries = min(num_entries, max_entries);
 
 	entries = (struct dmub_debugfs_trace_entry
 			   *)(tbuf_base +
 			      sizeof(struct dmub_debugfs_trace_header));
 
-	for (i = 0; i < num_entries; ++i) {
+	/* To print entries chronologically, start from the first entry till the
+	 * top of buffer, then from base of buffer to first entry.
+	 */
+	for (i = first_entry; i < num_entries; ++i) {
+		struct dmub_debugfs_trace_entry *entry = &entries[i];
+
+		seq_printf(m,
+			   "trace_code=%u tick_count=%u param0=%u param1=%u\n",
+			   entry->trace_code, entry->tick_count, entry->param0,
+			   entry->param1);
+	}
+	for (i = 0; i < first_entry; ++i) {
 		struct dmub_debugfs_trace_entry *entry = &entries[i];
 
 		seq_printf(m,
@@ -812,6 +987,58 @@ static int dmub_fw_state_show(struct seq_file *m, void *data)
 	return seq_write(m, state_base, state_size);
 }
 
+/* replay_capability_show() - show eDP panel replay capability
+ *
+ * The read function: replay_capability_show
+ * Shows if sink and driver has Replay capability or not.
+ *
+ *	cat /sys/kernel/debug/dri/0/eDP-X/replay_capability
+ *
+ * Expected output:
+ * "Sink support: no\n" - if panel doesn't support Replay
+ * "Sink support: yes\n" - if panel supports Replay
+ * "Driver support: no\n" - if driver doesn't support Replay
+ * "Driver support: yes\n" - if driver supports Replay
+ */
+static int replay_capability_show(struct seq_file *m, void *data)
+{
+	struct drm_connector *connector = m->private;
+	struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(connector);
+	struct dc_link *link = aconnector->dc_link;
+	bool sink_support_replay = false;
+	bool driver_support_replay = false;
+
+	if (!link)
+		return -ENODEV;
+
+	if (link->type == dc_connection_none)
+		return -ENODEV;
+
+	if (!(link->connector_signal & SIGNAL_TYPE_EDP))
+		return -ENODEV;
+
+	/* If Replay is already set to support, skip the checks */
+	if (link->replay_settings.config.replay_supported) {
+		sink_support_replay = true;
+		driver_support_replay = true;
+	} else if ((amdgpu_dc_debug_mask & DC_DISABLE_REPLAY)) {
+		sink_support_replay = amdgpu_dm_link_supports_replay(link, aconnector);
+	} else {
+		struct dc *dc = link->ctx->dc;
+
+		sink_support_replay = amdgpu_dm_link_supports_replay(link, aconnector);
+		if (dc->ctx->dmub_srv && dc->ctx->dmub_srv->dmub)
+			driver_support_replay =
+				(bool)dc->ctx->dmub_srv->dmub->feature_caps.replay_supported;
+	}
+
+	seq_printf(m, "Sink support: %s\n", str_yes_no(sink_support_replay));
+	seq_printf(m, "Driver support: %s\n", str_yes_no(driver_support_replay));
+	seq_printf(m, "Config support: %s\n", str_yes_no(link->replay_settings.config.replay_supported));
+
+	return 0;
+}
+
 /* psr_capability_show() - show eDP panel PSR capability
  *
  * The read function: sink_psr_capability_show
@@ -907,6 +1134,61 @@ unlock:
 DEFINE_SHOW_ATTRIBUTE(amdgpu_current_bpc);
 
 /*
+ * Returns the current colorspace for the crtc.
+ * Example usage: cat /sys/kernel/debug/dri/0/crtc-0/amdgpu_current_colorspace
+ */
+static int amdgpu_current_colorspace_show(struct seq_file *m, void *data)
+{
+	struct drm_crtc *crtc = m->private;
+	struct drm_device *dev = crtc->dev;
+	struct dm_crtc_state *dm_crtc_state = NULL;
+	int res = -ENODEV;
+
+	mutex_lock(&dev->mode_config.mutex);
+	drm_modeset_lock(&crtc->mutex, NULL);
+	if (crtc->state == NULL)
+		goto unlock;
+
+	dm_crtc_state = to_dm_crtc_state(crtc->state);
+	if (dm_crtc_state->stream == NULL)
+		goto unlock;
+
+	switch (dm_crtc_state->stream->output_color_space) {
+	case COLOR_SPACE_SRGB:
+		seq_puts(m, "sRGB");
+		break;
+	case COLOR_SPACE_YCBCR601:
+	case COLOR_SPACE_YCBCR601_LIMITED:
+		seq_puts(m, "BT601_YCC");
+		break;
+	case COLOR_SPACE_YCBCR709:
+	case COLOR_SPACE_YCBCR709_LIMITED:
+		seq_puts(m, "BT709_YCC");
+		break;
+	case COLOR_SPACE_ADOBERGB:
+		seq_puts(m, "opRGB");
+		break;
+	case COLOR_SPACE_2020_RGB_FULLRANGE:
+		seq_puts(m, "BT2020_RGB");
+		break;
+	case COLOR_SPACE_2020_YCBCR_LIMITED:
+		seq_puts(m, "BT2020_YCC");
+		break;
+	default:
+		goto unlock;
+	}
+	res = 0;
+
+unlock:
+	drm_modeset_unlock(&crtc->mutex);
+	mutex_unlock(&dev->mode_config.mutex);
+
+	return res;
+}
+DEFINE_SHOW_ATTRIBUTE(amdgpu_current_colorspace);
+
+
+/*
  * Example usage:
  * Disable dsc passthrough, i.e.,: have dsc decoding at converver, not external RX
  *   echo 1 /sys/kernel/debug/dri/0/DP-1/dsc_disable_passthrough
@@ -999,6 +1281,36 @@ static int internal_display_show(struct seq_file *m, void *data)
 	return 0;
 }
 
+/*
+ * Returns the number of segments used if ODM Combine mode is enabled.
+ * Example usage: cat /sys/kernel/debug/dri/0/DP-1/odm_combine_segments
+ */
+static int odm_combine_segments_show(struct seq_file *m, void *unused)
+{
+	struct drm_connector *connector = m->private;
+	struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(connector);
+	struct dc_link *link = aconnector->dc_link;
+	struct pipe_ctx *pipe_ctx = NULL;
+	int i, segments = -EOPNOTSUPP;
+
+	for (i = 0; i < MAX_PIPES; i++) {
+		pipe_ctx = &link->dc->current_state->res_ctx.pipe_ctx[i];
+		if (pipe_ctx->stream &&
+		    pipe_ctx->stream->link == link)
+			break;
+	}
+
+	if (connector->status != connector_status_connected)
+		return -ENODEV;
+
+	if (pipe_ctx && pipe_ctx->stream_res.tg &&
+	    pipe_ctx->stream_res.tg->funcs->get_odm_combine_segments)
+		pipe_ctx->stream_res.tg->funcs->get_odm_combine_segments(pipe_ctx->stream_res.tg, &segments);
+
+	seq_printf(m, "%d\n", segments);
+	return 0;
+}
+
 /* function description
  *
  * generic SDP message access for testing
@@ -1017,7 +1329,7 @@ static ssize_t dp_sdp_message_debugfs_write(struct file *f, const char __user *b
 				 size_t size, loff_t *pos)
 {
 	int r;
-	uint8_t data[36];
+	uint8_t data[36] = {0};
 	struct amdgpu_dm_connector *connector = file_inode(f)->i_private;
 	struct dm_crtc_state *acrtc_state;
 	uint32_t write_size = 36;
@@ -1039,88 +1351,6 @@ static ssize_t dp_sdp_message_debugfs_write(struct file *f, const char __user *b
 	return write_size;
 }
 
-static ssize_t dp_dpcd_address_write(struct file *f, const char __user *buf,
-				 size_t size, loff_t *pos)
-{
-	int r;
-	struct amdgpu_dm_connector *connector = file_inode(f)->i_private;
-
-	if (size < sizeof(connector->debugfs_dpcd_address))
-		return -EINVAL;
-
-	r = copy_from_user(&connector->debugfs_dpcd_address,
-			buf, sizeof(connector->debugfs_dpcd_address));
-
-	return size - r;
-}
-
-static ssize_t dp_dpcd_size_write(struct file *f, const char __user *buf,
-				 size_t size, loff_t *pos)
-{
-	int r;
-	struct amdgpu_dm_connector *connector = file_inode(f)->i_private;
-
-	if (size < sizeof(connector->debugfs_dpcd_size))
-		return -EINVAL;
-
-	r = copy_from_user(&connector->debugfs_dpcd_size,
-			buf, sizeof(connector->debugfs_dpcd_size));
-
-	if (connector->debugfs_dpcd_size > 256)
-		connector->debugfs_dpcd_size = 0;
-
-	return size - r;
-}
-
-static ssize_t dp_dpcd_data_write(struct file *f, const char __user *buf,
-				 size_t size, loff_t *pos)
-{
-	int r;
-	char *data;
-	struct amdgpu_dm_connector *connector = file_inode(f)->i_private;
-	struct dc_link *link = connector->dc_link;
-	uint32_t write_size = connector->debugfs_dpcd_size;
-
-	if (!write_size || size < write_size)
-		return -EINVAL;
-
-	data = kzalloc(write_size, GFP_KERNEL);
-	if (!data)
-		return 0;
-
-	r = copy_from_user(data, buf, write_size);
-
-	dm_helpers_dp_write_dpcd(link->ctx, link,
-			connector->debugfs_dpcd_address, data, write_size - r);
-	kfree(data);
-	return write_size - r;
-}
-
-static ssize_t dp_dpcd_data_read(struct file *f, char __user *buf,
-				 size_t size, loff_t *pos)
-{
-	int r;
-	char *data;
-	struct amdgpu_dm_connector *connector = file_inode(f)->i_private;
-	struct dc_link *link = connector->dc_link;
-	uint32_t read_size = connector->debugfs_dpcd_size;
-
-	if (!read_size || size < read_size)
-		return 0;
-
-	data = kzalloc(read_size, GFP_KERNEL);
-	if (!data)
-		return 0;
-
-	dm_helpers_dp_read_dpcd(link->ctx, link,
-			connector->debugfs_dpcd_address, data, read_size);
-
-	r = copy_to_user(buf, data, read_size);
-
-	kfree(data);
-	return read_size - r;
-}
-
 /* function: Read link's DSC & FEC capabilities
  *
  *
@@ -1217,7 +1447,7 @@ static ssize_t trigger_hotplug(struct file *f, const char __user *buf,
 	uint8_t param_nums = 0;
 	bool ret = false;
 
-	if (!aconnector || !aconnector->dc_link)
+	if (!aconnector->dc_link)
 		return -EINVAL;
 
 	if (size == 0)
@@ -1326,26 +1556,25 @@ static ssize_t dp_dsc_clock_en_read(struct file *f, char __user *buf,
 				    size_t size, loff_t *pos)
 {
 	char *rd_buf = NULL;
-	char *rd_buf_ptr = NULL;
 	struct amdgpu_dm_connector *aconnector = file_inode(f)->i_private;
 	struct display_stream_compressor *dsc;
 	struct dcn_dsc_state dsc_state = {0};
 	const uint32_t rd_buf_size = 10;
 	struct pipe_ctx *pipe_ctx;
 	ssize_t result = 0;
-	int i, r, str_len = 30;
+	int i, r, str_len = 10;
 
 	rd_buf = kcalloc(rd_buf_size, sizeof(char), GFP_KERNEL);
 
 	if (!rd_buf)
 		return -ENOMEM;
 
-	rd_buf_ptr = rd_buf;
-
 	for (i = 0; i < MAX_PIPES; i++) {
 		pipe_ctx = &aconnector->dc_link->dc->current_state->res_ctx.pipe_ctx[i];
 		if (pipe_ctx->stream &&
-		    pipe_ctx->stream->link == aconnector->dc_link)
+		    pipe_ctx->stream->link == aconnector->dc_link &&
+		    pipe_ctx->stream->sink &&
+		    pipe_ctx->stream->sink == aconnector->dc_sink)
 			break;
 	}
 
@@ -1353,10 +1582,9 @@ static ssize_t dp_dsc_clock_en_read(struct file *f, char __user *buf,
 	if (dsc)
 		dsc->funcs->dsc_read_state(dsc, &dsc_state);
 
-	snprintf(rd_buf_ptr, str_len,
+	snprintf(rd_buf, str_len,
 		"%d\n",
 		dsc_state.dsc_clock_en);
-	rd_buf_ptr += str_len;
 
 	while (size) {
 		if (*pos >= rd_buf_size)
@@ -1446,7 +1674,9 @@ static ssize_t dp_dsc_clock_en_write(struct file *f, const char __user *buf,
 	for (i = 0; i < MAX_PIPES; i++) {
 		pipe_ctx = &aconnector->dc_link->dc->current_state->res_ctx.pipe_ctx[i];
 		if (pipe_ctx->stream &&
-		    pipe_ctx->stream->link == aconnector->dc_link)
+		    pipe_ctx->stream->link == aconnector->dc_link &&
+		    pipe_ctx->stream->sink &&
+		    pipe_ctx->stream->sink == aconnector->dc_sink)
 			break;
 	}
 
@@ -1512,7 +1742,6 @@ static ssize_t dp_dsc_slice_width_read(struct file *f, char __user *buf,
 				    size_t size, loff_t *pos)
 {
 	char *rd_buf = NULL;
-	char *rd_buf_ptr = NULL;
 	struct amdgpu_dm_connector *aconnector = file_inode(f)->i_private;
 	struct display_stream_compressor *dsc;
 	struct dcn_dsc_state dsc_state = {0};
@@ -1526,12 +1755,12 @@ static ssize_t dp_dsc_slice_width_read(struct file *f, char __user *buf,
 	if (!rd_buf)
 		return -ENOMEM;
 
-	rd_buf_ptr = rd_buf;
-
 	for (i = 0; i < MAX_PIPES; i++) {
 		pipe_ctx = &aconnector->dc_link->dc->current_state->res_ctx.pipe_ctx[i];
 		if (pipe_ctx->stream &&
-		    pipe_ctx->stream->link == aconnector->dc_link)
+		    pipe_ctx->stream->link == aconnector->dc_link &&
+		    pipe_ctx->stream->sink &&
+		    pipe_ctx->stream->sink == aconnector->dc_sink)
 			break;
 	}
 
@@ -1539,10 +1768,9 @@ static ssize_t dp_dsc_slice_width_read(struct file *f, char __user *buf,
 	if (dsc)
 		dsc->funcs->dsc_read_state(dsc, &dsc_state);
 
-	snprintf(rd_buf_ptr, str_len,
+	snprintf(rd_buf, str_len,
 		"%d\n",
 		dsc_state.dsc_slice_width);
-	rd_buf_ptr += str_len;
 
 	while (size) {
 		if (*pos >= rd_buf_size)
@@ -1630,7 +1858,9 @@ static ssize_t dp_dsc_slice_width_write(struct file *f, const char __user *buf,
 	for (i = 0; i < MAX_PIPES; i++) {
 		pipe_ctx = &aconnector->dc_link->dc->current_state->res_ctx.pipe_ctx[i];
 		if (pipe_ctx->stream &&
-		    pipe_ctx->stream->link == aconnector->dc_link)
+		    pipe_ctx->stream->link == aconnector->dc_link &&
+		    pipe_ctx->stream->sink &&
+		    pipe_ctx->stream->sink == aconnector->dc_sink)
 			break;
 	}
 
@@ -1696,7 +1926,6 @@ static ssize_t dp_dsc_slice_height_read(struct file *f, char __user *buf,
 				    size_t size, loff_t *pos)
 {
 	char *rd_buf = NULL;
-	char *rd_buf_ptr = NULL;
 	struct amdgpu_dm_connector *aconnector = file_inode(f)->i_private;
 	struct display_stream_compressor *dsc;
 	struct dcn_dsc_state dsc_state = {0};
@@ -1710,12 +1939,12 @@ static ssize_t dp_dsc_slice_height_read(struct file *f, char __user *buf,
 	if (!rd_buf)
 		return -ENOMEM;
 
-	rd_buf_ptr = rd_buf;
-
 	for (i = 0; i < MAX_PIPES; i++) {
 		pipe_ctx = &aconnector->dc_link->dc->current_state->res_ctx.pipe_ctx[i];
 		if (pipe_ctx->stream &&
-		    pipe_ctx->stream->link == aconnector->dc_link)
+		    pipe_ctx->stream->link == aconnector->dc_link &&
+		    pipe_ctx->stream->sink &&
+		    pipe_ctx->stream->sink == aconnector->dc_sink)
 			break;
 	}
 
@@ -1723,10 +1952,9 @@ static ssize_t dp_dsc_slice_height_read(struct file *f, char __user *buf,
 	if (dsc)
 		dsc->funcs->dsc_read_state(dsc, &dsc_state);
 
-	snprintf(rd_buf_ptr, str_len,
+	snprintf(rd_buf, str_len,
 		"%d\n",
 		dsc_state.dsc_slice_height);
-	rd_buf_ptr += str_len;
 
 	while (size) {
 		if (*pos >= rd_buf_size)
@@ -1814,7 +2042,9 @@ static ssize_t dp_dsc_slice_height_write(struct file *f, const char __user *buf,
 	for (i = 0; i < MAX_PIPES; i++) {
 		pipe_ctx = &aconnector->dc_link->dc->current_state->res_ctx.pipe_ctx[i];
 		if (pipe_ctx->stream &&
-		    pipe_ctx->stream->link == aconnector->dc_link)
+		    pipe_ctx->stream->link == aconnector->dc_link &&
+		    pipe_ctx->stream->sink &&
+		    pipe_ctx->stream->sink == aconnector->dc_sink)
 			break;
 	}
 
@@ -1876,7 +2106,6 @@ static ssize_t dp_dsc_bits_per_pixel_read(struct file *f, char __user *buf,
 				    size_t size, loff_t *pos)
 {
 	char *rd_buf = NULL;
-	char *rd_buf_ptr = NULL;
 	struct amdgpu_dm_connector *aconnector = file_inode(f)->i_private;
 	struct display_stream_compressor *dsc;
 	struct dcn_dsc_state dsc_state = {0};
@@ -1890,12 +2119,12 @@ static ssize_t dp_dsc_bits_per_pixel_read(struct file *f, char __user *buf,
 	if (!rd_buf)
 		return -ENOMEM;
 
-	rd_buf_ptr = rd_buf;
-
 	for (i = 0; i < MAX_PIPES; i++) {
 		pipe_ctx = &aconnector->dc_link->dc->current_state->res_ctx.pipe_ctx[i];
 		if (pipe_ctx->stream &&
-		    pipe_ctx->stream->link == aconnector->dc_link)
+		    pipe_ctx->stream->link == aconnector->dc_link &&
+		    pipe_ctx->stream->sink &&
+		    pipe_ctx->stream->sink == aconnector->dc_sink)
 			break;
 	}
 
@@ -1903,10 +2132,9 @@ static ssize_t dp_dsc_bits_per_pixel_read(struct file *f, char __user *buf,
 	if (dsc)
 		dsc->funcs->dsc_read_state(dsc, &dsc_state);
 
-	snprintf(rd_buf_ptr, str_len,
+	snprintf(rd_buf, str_len,
 		"%d\n",
 		dsc_state.dsc_bits_per_pixel);
-	rd_buf_ptr += str_len;
 
 	while (size) {
 		if (*pos >= rd_buf_size)
@@ -1991,7 +2219,9 @@ static ssize_t dp_dsc_bits_per_pixel_write(struct file *f, const char __user *bu
 	for (i = 0; i < MAX_PIPES; i++) {
 		pipe_ctx = &aconnector->dc_link->dc->current_state->res_ctx.pipe_ctx[i];
 		if (pipe_ctx->stream &&
-		    pipe_ctx->stream->link == aconnector->dc_link)
+		    pipe_ctx->stream->link == aconnector->dc_link &&
+		    pipe_ctx->stream->sink &&
+		    pipe_ctx->stream->sink == aconnector->dc_sink)
 			break;
 	}
 
@@ -2051,7 +2281,6 @@ static ssize_t dp_dsc_pic_width_read(struct file *f, char __user *buf,
 				    size_t size, loff_t *pos)
 {
 	char *rd_buf = NULL;
-	char *rd_buf_ptr = NULL;
 	struct amdgpu_dm_connector *aconnector = file_inode(f)->i_private;
 	struct display_stream_compressor *dsc;
 	struct dcn_dsc_state dsc_state = {0};
@@ -2065,12 +2294,12 @@ static ssize_t dp_dsc_pic_width_read(struct file *f, char __user *buf,
 	if (!rd_buf)
 		return -ENOMEM;
 
-	rd_buf_ptr = rd_buf;
-
 	for (i = 0; i < MAX_PIPES; i++) {
 		pipe_ctx = &aconnector->dc_link->dc->current_state->res_ctx.pipe_ctx[i];
 		if (pipe_ctx->stream &&
-		    pipe_ctx->stream->link == aconnector->dc_link)
+		    pipe_ctx->stream->link == aconnector->dc_link &&
+		    pipe_ctx->stream->sink &&
+		    pipe_ctx->stream->sink == aconnector->dc_sink)
 			break;
 	}
 
@@ -2078,10 +2307,9 @@ static ssize_t dp_dsc_pic_width_read(struct file *f, char __user *buf,
 	if (dsc)
 		dsc->funcs->dsc_read_state(dsc, &dsc_state);
 
-	snprintf(rd_buf_ptr, str_len,
+	snprintf(rd_buf, str_len,
 		"%d\n",
 		dsc_state.dsc_pic_width);
-	rd_buf_ptr += str_len;
 
 	while (size) {
 		if (*pos >= rd_buf_size)
@@ -2107,7 +2335,6 @@ static ssize_t dp_dsc_pic_height_read(struct file *f, char __user *buf,
 				    size_t size, loff_t *pos)
 {
 	char *rd_buf = NULL;
-	char *rd_buf_ptr = NULL;
 	struct amdgpu_dm_connector *aconnector = file_inode(f)->i_private;
 	struct display_stream_compressor *dsc;
 	struct dcn_dsc_state dsc_state = {0};
@@ -2121,12 +2348,12 @@ static ssize_t dp_dsc_pic_height_read(struct file *f, char __user *buf,
 	if (!rd_buf)
 		return -ENOMEM;
 
-	rd_buf_ptr = rd_buf;
-
 	for (i = 0; i < MAX_PIPES; i++) {
 		pipe_ctx = &aconnector->dc_link->dc->current_state->res_ctx.pipe_ctx[i];
 		if (pipe_ctx->stream &&
-		    pipe_ctx->stream->link == aconnector->dc_link)
+		    pipe_ctx->stream->link == aconnector->dc_link &&
+		    pipe_ctx->stream->sink &&
+		    pipe_ctx->stream->sink == aconnector->dc_sink)
 			break;
 	}
 
@@ -2134,10 +2361,9 @@ static ssize_t dp_dsc_pic_height_read(struct file *f, char __user *buf,
 	if (dsc)
 		dsc->funcs->dsc_read_state(dsc, &dsc_state);
 
-	snprintf(rd_buf_ptr, str_len,
+	snprintf(rd_buf, str_len,
 		"%d\n",
 		dsc_state.dsc_pic_height);
-	rd_buf_ptr += str_len;
 
 	while (size) {
 		if (*pos >= rd_buf_size)
@@ -2178,7 +2404,6 @@ static ssize_t dp_dsc_chunk_size_read(struct file *f, char __user *buf,
 				    size_t size, loff_t *pos)
 {
 	char *rd_buf = NULL;
-	char *rd_buf_ptr = NULL;
 	struct amdgpu_dm_connector *aconnector = file_inode(f)->i_private;
 	struct display_stream_compressor *dsc;
 	struct dcn_dsc_state dsc_state = {0};
@@ -2192,12 +2417,12 @@ static ssize_t dp_dsc_chunk_size_read(struct file *f, char __user *buf,
 	if (!rd_buf)
 		return -ENOMEM;
 
-	rd_buf_ptr = rd_buf;
-
 	for (i = 0; i < MAX_PIPES; i++) {
 		pipe_ctx = &aconnector->dc_link->dc->current_state->res_ctx.pipe_ctx[i];
 		if (pipe_ctx->stream &&
-		    pipe_ctx->stream->link == aconnector->dc_link)
+		    pipe_ctx->stream->link == aconnector->dc_link &&
+		    pipe_ctx->stream->sink &&
+		    pipe_ctx->stream->sink == aconnector->dc_sink)
 			break;
 	}
 
@@ -2205,10 +2430,9 @@ static ssize_t dp_dsc_chunk_size_read(struct file *f, char __user *buf,
 	if (dsc)
 		dsc->funcs->dsc_read_state(dsc, &dsc_state);
 
-	snprintf(rd_buf_ptr, str_len,
+	snprintf(rd_buf, str_len,
 		"%d\n",
 		dsc_state.dsc_chunk_size);
-	rd_buf_ptr += str_len;
 
 	while (size) {
 		if (*pos >= rd_buf_size)
@@ -2249,7 +2473,6 @@ static ssize_t dp_dsc_slice_bpg_offset_read(struct file *f, char __user *buf,
 				    size_t size, loff_t *pos)
 {
 	char *rd_buf = NULL;
-	char *rd_buf_ptr = NULL;
 	struct amdgpu_dm_connector *aconnector = file_inode(f)->i_private;
 	struct display_stream_compressor *dsc;
 	struct dcn_dsc_state dsc_state = {0};
@@ -2263,12 +2486,12 @@ static ssize_t dp_dsc_slice_bpg_offset_read(struct file *f, char __user *buf,
 	if (!rd_buf)
 		return -ENOMEM;
 
-	rd_buf_ptr = rd_buf;
-
 	for (i = 0; i < MAX_PIPES; i++) {
 		pipe_ctx = &aconnector->dc_link->dc->current_state->res_ctx.pipe_ctx[i];
 		if (pipe_ctx->stream &&
-		    pipe_ctx->stream->link == aconnector->dc_link)
+		    pipe_ctx->stream->link == aconnector->dc_link &&
+		    pipe_ctx->stream->sink &&
+		    pipe_ctx->stream->sink == aconnector->dc_sink)
 			break;
 	}
 
@@ -2276,10 +2499,9 @@ static ssize_t dp_dsc_slice_bpg_offset_read(struct file *f, char __user *buf,
 	if (dsc)
 		dsc->funcs->dsc_read_state(dsc, &dsc_state);
 
-	snprintf(rd_buf_ptr, str_len,
+	snprintf(rd_buf, str_len,
 		"%d\n",
 		dsc_state.dsc_slice_bpg_offset);
-	rd_buf_ptr += str_len;
 
 	while (size) {
 		if (*pos >= rd_buf_size)
@@ -2445,6 +2667,49 @@ unlock:
 }
 
 /*
+ * IPS status.  Read only.
+ *
+ * Example usage: cat /sys/kernel/debug/dri/0/amdgpu_dm_ips_status
+ */
+static int ips_status_show(struct seq_file *m, void *unused)
+{
+	struct amdgpu_device *adev = m->private;
+	struct dc *dc = adev->dm.dc;
+	struct dc_dmub_srv *dc_dmub_srv;
+
+	seq_printf(m, "IPS config: %d\n", dc->config.disable_ips);
+	seq_printf(m, "Idle optimization: %d\n", dc->idle_optimizations_allowed);
+
+	if (adev->dm.idle_workqueue) {
+		seq_printf(m, "Idle workqueue - enabled: %d\n", adev->dm.idle_workqueue->enable);
+		seq_printf(m, "Idle workqueue - running: %d\n", adev->dm.idle_workqueue->running);
+	}
+
+	dc_dmub_srv = dc->ctx->dmub_srv;
+	if (dc_dmub_srv && dc_dmub_srv->dmub) {
+		uint32_t rcg_count, ips1_count, ips2_count;
+		volatile const struct dmub_shared_state_ips_fw *ips_fw =
+			&dc_dmub_srv->dmub->shared_state[DMUB_SHARED_SHARE_FEATURE__IPS_FW].data.ips_fw;
+		rcg_count = ips_fw->rcg_entry_count;
+		ips1_count = ips_fw->ips1_entry_count;
+		ips2_count = ips_fw->ips2_entry_count;
+		seq_printf(m, "entry counts: rcg=%u ips1=%u ips2=%u\n",
+			   rcg_count,
+			   ips1_count,
+			   ips2_count);
+		rcg_count = ips_fw->rcg_exit_count;
+		ips1_count = ips_fw->ips1_exit_count;
+		ips2_count = ips_fw->ips2_exit_count;
+		seq_printf(m, "exit counts: rcg=%u ips1=%u ips2=%u",
+			   rcg_count,
+			   ips1_count,
+			   ips2_count);
+		seq_puts(m, "\n");
+	}
+	return 0;
+}
+
+/*
  * Backlight at this moment.  Read only.
  * As written to display, taking ABM and backlight lut into account.
  * Ranges from 0x0 to 0x10000 (= 100% PWM)
@@ -2587,16 +2852,80 @@ static int is_dpia_link_show(struct seq_file *m, void *data)
 	return 0;
 }
 
+/**
+ * hdmi_cec_state_show - Read out the HDMI-CEC feature status
+ * @m: sequence file.
+ * @data: unused.
+ *
+ * Return 0 on success
+ */
+static int hdmi_cec_state_show(struct seq_file *m, void *data)
+{
+	struct drm_connector *connector = m->private;
+	struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(connector);
+
+	seq_printf(m, "%s:%d\n", connector->name, connector->base.id);
+	seq_printf(m, "HDMI-CEC status: %d\n", aconnector->notifier ? 1 : 0);
+
+	return 0;
+}
+
+/**
+ * hdmi_cec_state_write - Enable/Disable HDMI-CEC feature from driver side
+ * @f: file structure.
+ * @buf: userspace buffer. set to '1' to enable; '0' to disable cec feature.
+ * @size: size of buffer from userpsace.
+ * @pos: unused.
+ *
+ * Return size on success, error code on failure
+ */
+static ssize_t hdmi_cec_state_write(struct file *f, const char __user *buf,
+				    size_t size, loff_t *pos)
+{
+	int ret;
+	bool enable;
+	struct amdgpu_dm_connector *aconnector = file_inode(f)->i_private;
+	struct drm_device *ddev = aconnector->base.dev;
+
+	if (size == 0)
+		return -EINVAL;
+
+	ret = kstrtobool_from_user(buf, size, &enable);
+	if (ret) {
+		drm_dbg_driver(ddev, "invalid user data !\n");
+		return ret;
+	}
+
+	if (enable) {
+		if (aconnector->notifier)
+			return -EINVAL;
+		ret = amdgpu_dm_initialize_hdmi_connector(aconnector);
+		if (ret)
+			return ret;
+		hdmi_cec_set_edid(aconnector);
+	} else {
+		if (!aconnector->notifier)
+			return -EINVAL;
+		cec_notifier_conn_unregister(aconnector->notifier);
+		aconnector->notifier = NULL;
+	}
+
+	return size;
+}
+
 DEFINE_SHOW_ATTRIBUTE(dp_dsc_fec_support);
 DEFINE_SHOW_ATTRIBUTE(dmub_fw_state);
 DEFINE_SHOW_ATTRIBUTE(dmub_tracebuffer);
 DEFINE_SHOW_ATTRIBUTE(dp_lttpr_status);
 DEFINE_SHOW_ATTRIBUTE(hdcp_sink_capability);
 DEFINE_SHOW_ATTRIBUTE(internal_display);
+DEFINE_SHOW_ATTRIBUTE(odm_combine_segments);
+DEFINE_SHOW_ATTRIBUTE(replay_capability);
 DEFINE_SHOW_ATTRIBUTE(psr_capability);
 DEFINE_SHOW_ATTRIBUTE(dp_is_mst_connector);
 DEFINE_SHOW_ATTRIBUTE(dp_mst_progress_status);
 DEFINE_SHOW_ATTRIBUTE(is_dpia_link);
+DEFINE_SHOW_STORE_ATTRIBUTE(hdmi_cec_state);
 
 static const struct file_operations dp_dsc_clock_en_debugfs_fops = {
 	.owner = THIS_MODULE,
@@ -2682,25 +3011,6 @@ static const struct file_operations sdp_message_fops = {
 	.llseek = default_llseek
 };
 
-static const struct file_operations dp_dpcd_address_debugfs_fops = {
-	.owner = THIS_MODULE,
-	.write = dp_dpcd_address_write,
-	.llseek = default_llseek
-};
-
-static const struct file_operations dp_dpcd_size_debugfs_fops = {
-	.owner = THIS_MODULE,
-	.write = dp_dpcd_size_write,
-	.llseek = default_llseek
-};
-
-static const struct file_operations dp_dpcd_data_debugfs_fops = {
-	.owner = THIS_MODULE,
-	.read = dp_dpcd_data_read,
-	.write = dp_dpcd_data_write,
-	.llseek = default_llseek
-};
-
 static const struct file_operations dp_max_bpc_debugfs_fops = {
 	.owner = THIS_MODULE,
 	.read = dp_max_bpc_read,
@@ -2714,6 +3024,12 @@ static const struct file_operations dp_dsc_disable_passthrough_debugfs_fops = {
 	.llseek = default_llseek
 };
 
+static const struct file_operations dp_mst_link_settings_debugfs_fops = {
+	.owner = THIS_MODULE,
+	.write = dp_mst_link_setting,
+	.llseek = default_llseek
+};
+
 static const struct {
 	char *name;
 	const struct file_operations *fops;
@@ -2724,9 +3040,6 @@ static const struct {
 		{"test_pattern", &dp_phy_test_pattern_fops},
 		{"hdcp_sink_capability", &hdcp_sink_capability_fops},
 		{"sdp_message", &sdp_message_fops},
-		{"aux_dpcd_address", &dp_dpcd_address_debugfs_fops},
-		{"aux_dpcd_size", &dp_dpcd_size_debugfs_fops},
-		{"aux_dpcd_data", &dp_dpcd_data_debugfs_fops},
 		{"dsc_clock_en", &dp_dsc_clock_en_debugfs_fops},
 		{"dsc_slice_width", &dp_dsc_slice_width_debugfs_fops},
 		{"dsc_slice_height", &dp_dsc_slice_height_debugfs_fops},
@@ -2740,14 +3053,16 @@ static const struct {
 		{"dsc_disable_passthrough", &dp_dsc_disable_passthrough_debugfs_fops},
 		{"is_mst_connector", &dp_is_mst_connector_fops},
 		{"mst_progress_status", &dp_mst_progress_status_fops},
-		{"is_dpia_link", &is_dpia_link_fops}
+		{"is_dpia_link", &is_dpia_link_fops},
+		{"mst_link_settings", &dp_mst_link_settings_debugfs_fops}
 };
 
 static const struct {
 	char *name;
 	const struct file_operations *fops;
 } hdmi_debugfs_entries[] = {
-		{"hdcp_sink_capability", &hdcp_sink_capability_fops}
+		{"hdcp_sink_capability", &hdcp_sink_capability_fops},
+		{"hdmi_cec_state", &hdmi_cec_state_fops}
 };
 
 /*
@@ -2778,6 +3093,51 @@ DEFINE_DEBUGFS_ATTRIBUTE(force_yuv420_output_fops, force_yuv420_output_get,
 			 force_yuv420_output_set, "%llu\n");
 
 /*
+ *  Read Replay state
+ */
+static int replay_get_state(void *data, u64 *val)
+{
+	struct amdgpu_dm_connector *connector = data;
+	struct dc_link *link = connector->dc_link;
+	uint64_t state = REPLAY_STATE_INVALID;
+
+	dc_link_get_replay_state(link, &state);
+
+	*val = state;
+
+	return 0;
+}
+
+/*
+ *  Start / Stop capture Replay residency
+ */
+static int replay_set_residency(void *data, u64 val)
+{
+	struct amdgpu_dm_connector *connector = data;
+	struct dc_link *link = connector->dc_link;
+	bool is_start = (val != 0);
+	u32 residency = 0;
+
+	link->dc->link_srv->edp_replay_residency(link, &residency, is_start, PR_RESIDENCY_MODE_PHY);
+	return 0;
+}
+
+/*
+ *  Read Replay residency
+ */
+static int replay_get_residency(void *data, u64 *val)
+{
+	struct amdgpu_dm_connector *connector = data;
+	struct dc_link *link = connector->dc_link;
+	u32 residency = 0;
+
+	link->dc->link_srv->edp_replay_residency(link, &residency, false, PR_RESIDENCY_MODE_PHY);
+	*val = (u64)residency;
+
+	return 0;
+}
+
+/*
  *  Read PSR state
  */
 static int psr_get(void *data, u64 *val)
@@ -2800,15 +3160,167 @@ static int psr_read_residency(void *data, u64 *val)
 {
 	struct amdgpu_dm_connector *connector = data;
 	struct dc_link *link = connector->dc_link;
-	u32 residency;
+	u32 residency = 0;
 
-	link->dc->link_srv->edp_get_psr_residency(link, &residency);
+	link->dc->link_srv->edp_get_psr_residency(link, &residency, PSR_RESIDENCY_MODE_PHY);
 
 	*val = (u64)residency;
 
 	return 0;
 }
 
+/* read allow_edp_hotplug_detection */
+static int allow_edp_hotplug_detection_get(void *data, u64 *val)
+{
+	struct amdgpu_dm_connector *aconnector = data;
+	struct drm_connector *connector = &aconnector->base;
+	struct drm_device *dev = connector->dev;
+	struct amdgpu_device *adev = drm_to_adev(dev);
+
+	*val = adev->dm.dc->config.allow_edp_hotplug_detection;
+
+	return 0;
+}
+
+/* set allow_edp_hotplug_detection */
+static int allow_edp_hotplug_detection_set(void *data, u64 val)
+{
+	struct amdgpu_dm_connector *aconnector = data;
+	struct drm_connector *connector = &aconnector->base;
+	struct drm_device *dev = connector->dev;
+	struct amdgpu_device *adev = drm_to_adev(dev);
+
+	adev->dm.dc->config.allow_edp_hotplug_detection = (uint32_t) val;
+
+	return 0;
+}
+
+/* check if kernel disallow eDP enter psr state
+ * cat /sys/kernel/debug/dri/0/eDP-X/disallow_edp_enter_psr
+ * 0: allow edp enter psr; 1: disallow
+ */
+static int disallow_edp_enter_psr_get(void *data, u64 *val)
+{
+	struct amdgpu_dm_connector *aconnector = data;
+
+	*val = (u64) aconnector->disallow_edp_enter_psr;
+	return 0;
+}
+
+/* set kernel disallow eDP enter psr state
+ * echo 0x0 /sys/kernel/debug/dri/0/eDP-X/disallow_edp_enter_psr
+ * 0: allow edp enter psr; 1: disallow
+ *
+ * usage: test app read crc from PSR eDP rx.
+ *
+ * during kernel boot up, kernel write dpcd 0x170 = 5.
+ * this notify eDP rx psr enable and let rx check crc.
+ * rx fw will start checking crc for rx internal logic.
+ * crc read count within dpcd 0x246 is not updated and
+ * value is 0. when eDP tx driver wants to read rx crc
+ * from dpcd 0x246, 0x270, read count 0 lead tx driver
+ * timeout.
+ *
+ * to avoid this, we add this debugfs to let test app to disbable
+ * rx crc checking for rx internal logic. then test app can read
+ * non-zero crc read count.
+ *
+ * expected app sequence is as below:
+ * 1. disable eDP PHY and notify eDP rx with dpcd 0x600 = 2.
+ * 2. echo 0x1 /sys/kernel/debug/dri/0/eDP-X/disallow_edp_enter_psr
+ * 3. enable eDP PHY and notify eDP rx with dpcd 0x600 = 1 but
+ *    without dpcd 0x170 = 5.
+ * 4. read crc from rx dpcd 0x270, 0x246, etc.
+ * 5. echo 0x0 /sys/kernel/debug/dri/0/eDP-X/disallow_edp_enter_psr.
+ *    this will let eDP back to normal with psr setup dpcd 0x170 = 5.
+ */
+static int disallow_edp_enter_psr_set(void *data, u64 val)
+{
+	struct amdgpu_dm_connector *aconnector = data;
+
+	aconnector->disallow_edp_enter_psr = val ? true : false;
+	return 0;
+}
+
+static int dmub_trace_mask_set(void *data, u64 val)
+{
+	struct amdgpu_device *adev = data;
+	struct dmub_srv *srv = adev->dm.dc->ctx->dmub_srv->dmub;
+	enum dmub_gpint_command cmd;
+	u64 mask = 0xffff;
+	u8 shift = 0;
+	u32 res;
+	int i;
+
+	if (!srv->fw_version)
+		return -EINVAL;
+
+	for (i = 0;  i < 4; i++) {
+		res = (val & mask) >> shift;
+
+		switch (i) {
+		case 0:
+			cmd = DMUB_GPINT__SET_TRACE_BUFFER_MASK_WORD0;
+			break;
+		case 1:
+			cmd = DMUB_GPINT__SET_TRACE_BUFFER_MASK_WORD1;
+			break;
+		case 2:
+			cmd = DMUB_GPINT__SET_TRACE_BUFFER_MASK_WORD2;
+			break;
+		case 3:
+			cmd = DMUB_GPINT__SET_TRACE_BUFFER_MASK_WORD3;
+			break;
+		}
+
+		if (!dc_wake_and_execute_gpint(adev->dm.dc->ctx, cmd, res, NULL, DM_DMUB_WAIT_TYPE_WAIT))
+			return -EIO;
+
+		usleep_range(100, 1000);
+
+		mask <<= 16;
+		shift += 16;
+	}
+
+	return 0;
+}
+
+static int dmub_trace_mask_show(void *data, u64 *val)
+{
+	enum dmub_gpint_command cmd = DMUB_GPINT__GET_TRACE_BUFFER_MASK_WORD0;
+	struct amdgpu_device *adev = data;
+	struct dmub_srv *srv = adev->dm.dc->ctx->dmub_srv->dmub;
+	u8 shift = 0;
+	u64 raw = 0;
+	u64 res = 0;
+	int i = 0;
+
+	if (!srv->fw_version)
+		return -EINVAL;
+
+	while (i < 4) {
+		uint32_t response;
+
+		if (!dc_wake_and_execute_gpint(adev->dm.dc->ctx, cmd, 0, &response, DM_DMUB_WAIT_TYPE_WAIT_WITH_REPLY))
+			return -EIO;
+
+		raw = response;
+		usleep_range(100, 1000);
+
+		cmd++;
+		res |= (raw << shift);
+		shift += 16;
+		i++;
+	}
+
+	*val = res;
+
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(dmub_trace_mask_fops, dmub_trace_mask_show,
+			 dmub_trace_mask_set, "0x%llx\n");
+
 /*
  * Set dmcub trace event IRQ enable or disable.
  * Usage to enable dmcub trace event IRQ: echo 1 > /sys/kernel/debug/dri/0/amdgpu_dm_dmcub_trace_event_en
@@ -2843,12 +3355,24 @@ static int dmcub_trace_event_state_get(void *data, u64 *val)
 DEFINE_DEBUGFS_ATTRIBUTE(dmcub_trace_event_state_fops, dmcub_trace_event_state_get,
 			 dmcub_trace_event_state_set, "%llu\n");
 
+DEFINE_DEBUGFS_ATTRIBUTE(replay_state_fops, replay_get_state, NULL, "%llu\n");
+DEFINE_DEBUGFS_ATTRIBUTE(replay_residency_fops, replay_get_residency, replay_set_residency,
+			 "%llu\n");
 DEFINE_DEBUGFS_ATTRIBUTE(psr_fops, psr_get, NULL, "%llu\n");
 DEFINE_DEBUGFS_ATTRIBUTE(psr_residency_fops, psr_read_residency, NULL,
 			 "%llu\n");
 
+DEFINE_DEBUGFS_ATTRIBUTE(allow_edp_hotplug_detection_fops,
+			allow_edp_hotplug_detection_get,
+			allow_edp_hotplug_detection_set, "%llu\n");
+
+DEFINE_DEBUGFS_ATTRIBUTE(disallow_edp_enter_psr_fops,
+			disallow_edp_enter_psr_get,
+			disallow_edp_enter_psr_set, "%llu\n");
+
 DEFINE_SHOW_ATTRIBUTE(current_backlight);
 DEFINE_SHOW_ATTRIBUTE(target_backlight);
+DEFINE_SHOW_ATTRIBUTE(ips_status);
 
 static const struct {
 	char *name;
@@ -2856,7 +3380,8 @@ static const struct {
 } connector_debugfs_entries[] = {
 		{"force_yuv420_output", &force_yuv420_output_fops},
 		{"trigger_hotplug", &trigger_hotplug_debugfs_fops},
-		{"internal_display", &internal_display_fops}
+		{"internal_display", &internal_display_fops},
+		{"odm_combine_segments", &odm_combine_segments_fops}
 };
 
 /*
@@ -2887,7 +3412,7 @@ static int edp_ilr_show(struct seq_file *m, void *unused)
 			seq_printf(m, "[%d] %d kHz\n", entry/2, link_rate_in_khz);
 		}
 	} else {
-		seq_printf(m, "ILR is not supported by this eDP panel.\n");
+		seq_puts(m, "ILR is not supported by this eDP panel.\n");
 	}
 
 	return 0;
@@ -3007,6 +3532,11 @@ void connector_debugfs_init(struct amdgpu_dm_connector *connector)
 		}
 	}
 	if (connector->base.connector_type == DRM_MODE_CONNECTOR_eDP) {
+		debugfs_create_file("replay_capability", 0444, dir, connector,
+					&replay_capability_fops);
+		debugfs_create_file("replay_state", 0444, dir, connector, &replay_state_fops);
+		debugfs_create_file_unsafe("replay_residency", 0444, dir,
+					   connector, &replay_residency_fops);
 		debugfs_create_file_unsafe("psr_capability", 0444, dir, connector, &psr_capability_fops);
 		debugfs_create_file_unsafe("psr_state", 0444, dir, connector, &psr_fops);
 		debugfs_create_file_unsafe("psr_residency", 0444, dir,
@@ -3017,6 +3547,10 @@ void connector_debugfs_init(struct amdgpu_dm_connector *connector)
 				    &target_backlight_fops);
 		debugfs_create_file("ilr_setting", 0644, dir, connector,
 					&edp_ilr_debugfs_fops);
+		debugfs_create_file("allow_edp_hotplug_detection", 0644, dir, connector,
+					&allow_edp_hotplug_detection_fops);
+		debugfs_create_file("disallow_edp_enter_psr", 0644, dir, connector,
+					&disallow_edp_enter_psr_fops);
 	}
 
 	for (i = 0; i < ARRAY_SIZE(connector_debugfs_entries); i++) {
@@ -3025,9 +3559,6 @@ void connector_debugfs_init(struct amdgpu_dm_connector *connector)
 				    connector_debugfs_entries[i].fops);
 	}
 
-	connector->debugfs_dpcd_address = 0;
-	connector->debugfs_dpcd_size = 0;
-
 	if (connector->base.connector_type == DRM_MODE_CONNECTOR_HDMIA) {
 		for (i = 0; i < ARRAY_SIZE(hdmi_debugfs_entries); i++) {
 			debugfs_create_file(hdmi_debugfs_entries[i].name,
@@ -3048,8 +3579,8 @@ static int crc_win_x_start_set(void *data, u64 val)
 	struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
 
 	spin_lock_irq(&drm_dev->event_lock);
-	acrtc->dm_irq_params.window_param.x_start = (uint16_t) val;
-	acrtc->dm_irq_params.window_param.update_win = false;
+	acrtc->dm_irq_params.window_param[0].x_start = (uint16_t) val;
+	acrtc->dm_irq_params.window_param[0].update_win = false;
 	spin_unlock_irq(&drm_dev->event_lock);
 
 	return 0;
@@ -3065,7 +3596,7 @@ static int crc_win_x_start_get(void *data, u64 *val)
 	struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
 
 	spin_lock_irq(&drm_dev->event_lock);
-	*val = acrtc->dm_irq_params.window_param.x_start;
+	*val = acrtc->dm_irq_params.window_param[0].x_start;
 	spin_unlock_irq(&drm_dev->event_lock);
 
 	return 0;
@@ -3085,8 +3616,8 @@ static int crc_win_y_start_set(void *data, u64 val)
 	struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
 
 	spin_lock_irq(&drm_dev->event_lock);
-	acrtc->dm_irq_params.window_param.y_start = (uint16_t) val;
-	acrtc->dm_irq_params.window_param.update_win = false;
+	acrtc->dm_irq_params.window_param[0].y_start = (uint16_t) val;
+	acrtc->dm_irq_params.window_param[0].update_win = false;
 	spin_unlock_irq(&drm_dev->event_lock);
 
 	return 0;
@@ -3102,7 +3633,7 @@ static int crc_win_y_start_get(void *data, u64 *val)
 	struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
 
 	spin_lock_irq(&drm_dev->event_lock);
-	*val = acrtc->dm_irq_params.window_param.y_start;
+	*val = acrtc->dm_irq_params.window_param[0].y_start;
 	spin_unlock_irq(&drm_dev->event_lock);
 
 	return 0;
@@ -3121,8 +3652,8 @@ static int crc_win_x_end_set(void *data, u64 val)
 	struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
 
 	spin_lock_irq(&drm_dev->event_lock);
-	acrtc->dm_irq_params.window_param.x_end = (uint16_t) val;
-	acrtc->dm_irq_params.window_param.update_win = false;
+	acrtc->dm_irq_params.window_param[0].x_end = (uint16_t) val;
+	acrtc->dm_irq_params.window_param[0].update_win = false;
 	spin_unlock_irq(&drm_dev->event_lock);
 
 	return 0;
@@ -3138,7 +3669,7 @@ static int crc_win_x_end_get(void *data, u64 *val)
 	struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
 
 	spin_lock_irq(&drm_dev->event_lock);
-	*val = acrtc->dm_irq_params.window_param.x_end;
+	*val = acrtc->dm_irq_params.window_param[0].x_end;
 	spin_unlock_irq(&drm_dev->event_lock);
 
 	return 0;
@@ -3157,8 +3688,8 @@ static int crc_win_y_end_set(void *data, u64 val)
 	struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
 
 	spin_lock_irq(&drm_dev->event_lock);
-	acrtc->dm_irq_params.window_param.y_end = (uint16_t) val;
-	acrtc->dm_irq_params.window_param.update_win = false;
+	acrtc->dm_irq_params.window_param[0].y_end = (uint16_t) val;
+	acrtc->dm_irq_params.window_param[0].update_win = false;
 	spin_unlock_irq(&drm_dev->event_lock);
 
 	return 0;
@@ -3174,7 +3705,7 @@ static int crc_win_y_end_get(void *data, u64 *val)
 	struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
 
 	spin_lock_irq(&drm_dev->event_lock);
-	*val = acrtc->dm_irq_params.window_param.y_end;
+	*val = acrtc->dm_irq_params.window_param[0].y_end;
 	spin_unlock_irq(&drm_dev->event_lock);
 
 	return 0;
@@ -3197,13 +3728,14 @@ static int crc_win_update_set(void *data, u64 val)
 		/* PSR may write to OTG CRC window control register,
 		 * so close it before starting secure_display.
 		 */
-		amdgpu_dm_psr_disable(acrtc->dm_irq_params.stream);
+		amdgpu_dm_psr_disable(acrtc->dm_irq_params.stream, true);
 
 		spin_lock_irq(&adev_to_drm(adev)->event_lock);
 
-		acrtc->dm_irq_params.window_param.activated = true;
-		acrtc->dm_irq_params.window_param.update_win = true;
-		acrtc->dm_irq_params.window_param.skip_frame_cnt = 0;
+		acrtc->dm_irq_params.window_param[0].enable = true;
+		acrtc->dm_irq_params.window_param[0].update_win = true;
+		acrtc->dm_irq_params.window_param[0].skip_frame_cnt = 0;
+		acrtc->dm_irq_params.crc_window_activated = true;
 
 		spin_unlock_irq(&adev_to_drm(adev)->event_lock);
 		mutex_unlock(&adev->dm.dc_lock);
@@ -3246,6 +3778,8 @@ void crtc_debugfs_init(struct drm_crtc *crtc)
 #endif
 	debugfs_create_file("amdgpu_current_bpc", 0644, crtc->debugfs_entry,
 			    crtc, &amdgpu_current_bpc_fops);
+	debugfs_create_file("amdgpu_current_colorspace", 0644, crtc->debugfs_entry,
+			    crtc, &amdgpu_current_colorspace_fops);
 }
 
 /*
@@ -3352,6 +3886,7 @@ static int trigger_hpd_mst_set(void *data, u64 val)
 	struct amdgpu_dm_connector *aconnector;
 	struct drm_connector *connector;
 	struct dc_link *link = NULL;
+	int ret;
 
 	if (val == 1) {
 		drm_connector_list_iter_begin(dev, &iter);
@@ -3360,10 +3895,15 @@ static int trigger_hpd_mst_set(void *data, u64 val)
 			if (aconnector->dc_link->type == dc_connection_mst_branch &&
 			    aconnector->mst_mgr.aux) {
 				mutex_lock(&adev->dm.dc_lock);
-				dc_link_detect(aconnector->dc_link, DETECT_REASON_HPD);
+				ret = dc_link_detect(aconnector->dc_link, DETECT_REASON_HPD);
 				mutex_unlock(&adev->dm.dc_lock);
 
-				drm_dp_mst_topology_mgr_set_mst(&aconnector->mst_mgr, true);
+				if (!ret)
+					DRM_ERROR("DM_MST: Failed to detect dc link!");
+
+				ret = drm_dp_mst_topology_mgr_set_mst(&aconnector->mst_mgr, true);
+				if (ret < 0)
+					DRM_ERROR("DM_MST: Failed to set the device into MST mode!");
 			}
 		}
 	} else if (val == 0) {
@@ -3470,6 +4010,36 @@ DEFINE_DEBUGFS_ATTRIBUTE(disable_hpd_ops, disable_hpd_get,
 			 disable_hpd_set, "%llu\n");
 
 /*
+ * Prints hardware capabilities. These are used for IGT testing.
+ */
+static int capabilities_show(struct seq_file *m, void *unused)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct dc *dc = adev->dm.dc;
+	bool mall_supported = dc->caps.mall_size_total;
+	bool subvp_supported = dc->caps.subvp_fw_processing_delay_us;
+	unsigned int mall_in_use = false;
+	unsigned int subvp_in_use = false;
+
+	struct hubbub *hubbub = dc->res_pool->hubbub;
+
+	if (hubbub && hubbub->funcs->get_mall_en)
+		hubbub->funcs->get_mall_en(hubbub, &mall_in_use);
+
+	if (dc->cap_funcs.get_subvp_en)
+		subvp_in_use = dc->cap_funcs.get_subvp_en(dc, dc->current_state);
+
+	seq_printf(m, "mall supported: %s, enabled: %s\n",
+			   mall_supported ? "yes" : "no", mall_in_use ? "yes" : "no");
+	seq_printf(m, "sub-viewport supported: %s, enabled: %s\n",
+			   subvp_supported ? "yes" : "no", subvp_in_use ? "yes" : "no");
+
+	return 0;
+}
+
+DEFINE_SHOW_ATTRIBUTE(capabilities);
+
+/*
  * Temporary w/a to force sst sequence in M42D DP2 mst receiver
  * Example usage: echo 1 > /sys/kernel/debug/dri/0/amdgpu_dm_dp_set_mst_en_for_sst
  */
@@ -3662,6 +4232,8 @@ void dtn_debugfs_init(struct amdgpu_device *adev)
 
 	debugfs_create_file("amdgpu_mst_topology", 0444, root,
 			    adev, &mst_topo_fops);
+	debugfs_create_file("amdgpu_dm_capabilities", 0444, root,
+			    adev, &capabilities_fops);
 	debugfs_create_file("amdgpu_dm_dtn_log", 0644, root, adev,
 			    &dtn_log_fops);
 	debugfs_create_file("amdgpu_dm_dp_set_mst_en_for_sst", 0644, root, adev,
@@ -3684,6 +4256,9 @@ void dtn_debugfs_init(struct amdgpu_device *adev)
 	debugfs_create_file_unsafe("amdgpu_dm_force_timing_sync", 0644, root,
 				   adev, &force_timing_sync_ops);
 
+	debugfs_create_file_unsafe("amdgpu_dm_dmub_trace_mask", 0644, root,
+				   adev, &dmub_trace_mask_fops);
+
 	debugfs_create_file_unsafe("amdgpu_dm_dmcub_trace_event_en", 0644, root,
 				   adev, &dmcub_trace_event_state_fops);
 
@@ -3696,4 +4271,7 @@ void dtn_debugfs_init(struct amdgpu_device *adev)
 	debugfs_create_file_unsafe("amdgpu_dm_disable_hpd", 0644, root, adev,
 				   &disable_hpd_ops);
 
+	if (adev->dm.dc->caps.ips_support)
+		debugfs_create_file_unsafe("amdgpu_dm_ips_status", 0644, root, adev,
+					   &ips_status_fops);
 }
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_debugfs.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_debugfs.h
index 071200473c27..122cdc124b3b 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_debugfs.h
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_debugfs.h
@@ -1,3 +1,4 @@
+/* SPDX-License-Identifier: MIT */
 /*
  * Copyright 2018 Advanced Micro Devices, Inc.
  *
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_hdcp.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_hdcp.c
index 5536d17306d0..85ce558cefc5 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_hdcp.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_hdcp.c
@@ -1,3 +1,4 @@
+// SPDX-License-Identifier: MIT
 /*
  * Copyright 2019 Advanced Micro Devices, Inc.
  *
@@ -26,6 +27,7 @@
 #include "amdgpu_dm_hdcp.h"
 #include "amdgpu.h"
 #include "amdgpu_dm.h"
+#include "dc_fused_io.h"
 #include "dm_helpers.h"
 #include <drm/display/drm_hdcp_helper.h>
 #include "hdcp_psp.h"
@@ -39,10 +41,10 @@
 static bool
 lp_write_i2c(void *handle, uint32_t address, const uint8_t *data, uint32_t size)
 {
-
 	struct dc_link *link = handle;
 	struct i2c_payload i2c_payloads[] = {{true, address, size, (void *)data} };
-	struct i2c_command cmd = {i2c_payloads, 1, I2C_COMMAND_ENGINE_HW, link->dc->caps.i2c_speed_in_khz};
+	struct i2c_command cmd = {i2c_payloads, 1, I2C_COMMAND_ENGINE_HW,
+				  link->dc->caps.i2c_speed_in_khz};
 
 	return dm_helpers_submit_i2c(link->ctx, link, &cmd);
 }
@@ -52,8 +54,10 @@ lp_read_i2c(void *handle, uint32_t address, uint8_t offset, uint8_t *data, uint3
 {
 	struct dc_link *link = handle;
 
-	struct i2c_payload i2c_payloads[] = {{true, address, 1, &offset}, {false, address, size, data} };
-	struct i2c_command cmd = {i2c_payloads, 2, I2C_COMMAND_ENGINE_HW, link->dc->caps.i2c_speed_in_khz};
+	struct i2c_payload i2c_payloads[] = {{true, address, 1, &offset},
+					     {false, address, size, data} };
+	struct i2c_command cmd = {i2c_payloads, 2, I2C_COMMAND_ENGINE_HW,
+				  link->dc->caps.i2c_speed_in_khz};
 
 	return dm_helpers_submit_i2c(link->ctx, link, &cmd);
 }
@@ -74,9 +78,36 @@ lp_read_dpcd(void *handle, uint32_t address, uint8_t *data, uint32_t size)
 	return dm_helpers_dp_read_dpcd(link->ctx, link, address, data, size);
 }
 
-static uint8_t *psp_get_srm(struct psp_context *psp, uint32_t *srm_version, uint32_t *srm_size)
+static bool lp_atomic_write_poll_read_i2c(
+		void *handle,
+		const struct mod_hdcp_atomic_op_i2c *write,
+		const struct mod_hdcp_atomic_op_i2c *poll,
+		struct mod_hdcp_atomic_op_i2c *read,
+		uint32_t poll_timeout_us,
+		uint8_t poll_mask_msb
+)
 {
+	struct dc_link *link = handle;
+
+	return dm_atomic_write_poll_read_i2c(link, write, poll, read, poll_timeout_us, poll_mask_msb);
+}
+
+static bool lp_atomic_write_poll_read_aux(
+		void *handle,
+		const struct mod_hdcp_atomic_op_aux *write,
+		const struct mod_hdcp_atomic_op_aux *poll,
+		struct mod_hdcp_atomic_op_aux *read,
+		uint32_t poll_timeout_us,
+		uint8_t poll_mask_msb
+)
+{
+	struct dc_link *link = handle;
+
+	return dm_atomic_write_poll_read_aux(link, write, poll, read, poll_timeout_us, poll_mask_msb);
+}
 
+static uint8_t *psp_get_srm(struct psp_context *psp, uint32_t *srm_version, uint32_t *srm_size)
+{
 	struct ta_hdcp_shared_memory *hdcp_cmd;
 
 	if (!psp->hdcp_context.context.initialized) {
@@ -96,13 +127,12 @@ static uint8_t *psp_get_srm(struct psp_context *psp, uint32_t *srm_version, uint
 	*srm_version = hdcp_cmd->out_msg.hdcp_get_srm.srm_version;
 	*srm_size = hdcp_cmd->out_msg.hdcp_get_srm.srm_buf_size;
 
-
 	return hdcp_cmd->out_msg.hdcp_get_srm.srm_buf;
 }
 
-static int psp_set_srm(struct psp_context *psp, uint8_t *srm, uint32_t srm_size, uint32_t *srm_version)
+static int psp_set_srm(struct psp_context *psp,
+		       u8 *srm, uint32_t srm_size, uint32_t *srm_version)
 {
-
 	struct ta_hdcp_shared_memory *hdcp_cmd;
 
 	if (!psp->hdcp_context.context.initialized) {
@@ -119,7 +149,8 @@ static int psp_set_srm(struct psp_context *psp, uint8_t *srm, uint32_t srm_size,
 
 	psp_hdcp_invoke(psp, hdcp_cmd->cmd_id);
 
-	if (hdcp_cmd->hdcp_status != TA_HDCP_STATUS__SUCCESS || hdcp_cmd->out_msg.hdcp_set_srm.valid_signature != 1 ||
+	if (hdcp_cmd->hdcp_status != TA_HDCP_STATUS__SUCCESS ||
+	    hdcp_cmd->out_msg.hdcp_set_srm.valid_signature != 1 ||
 	    hdcp_cmd->out_msg.hdcp_set_srm.srm_version == PSP_SRM_VERSION_MAX)
 		return -EINVAL;
 
@@ -150,7 +181,6 @@ static void process_output(struct hdcp_workqueue *hdcp_work)
 
 static void link_lock(struct hdcp_workqueue *work, bool lock)
 {
-
 	int i = 0;
 
 	for (i = 0; i < work->max_link; i++) {
@@ -160,74 +190,74 @@ static void link_lock(struct hdcp_workqueue *work, bool lock)
 			mutex_unlock(&work[i].mutex);
 	}
 }
+
 void hdcp_update_display(struct hdcp_workqueue *hdcp_work,
 			 unsigned int link_index,
 			 struct amdgpu_dm_connector *aconnector,
-			 uint8_t content_type,
+			 u8 content_type,
 			 bool enable_encryption)
 {
 	struct hdcp_workqueue *hdcp_w = &hdcp_work[link_index];
-	struct mod_hdcp_display *display = &hdcp_work[link_index].display;
-	struct mod_hdcp_link *link = &hdcp_work[link_index].link;
-	struct mod_hdcp_display_query query;
+	struct mod_hdcp_link_adjustment link_adjust;
+	struct mod_hdcp_display_adjustment display_adjust;
 	unsigned int conn_index = aconnector->base.index;
+	const struct dc *dc = aconnector->dc_link->dc;
 
-	mutex_lock(&hdcp_w->mutex);
+	guard(mutex)(&hdcp_w->mutex);
+	drm_connector_get(&aconnector->base);
+	if (hdcp_w->aconnector[conn_index])
+		drm_connector_put(&hdcp_w->aconnector[conn_index]->base);
 	hdcp_w->aconnector[conn_index] = aconnector;
 
-	query.display = NULL;
-	mod_hdcp_query_display(&hdcp_w->hdcp, aconnector->base.index, &query);
-
-	if (query.display != NULL) {
-		memcpy(display, query.display, sizeof(struct mod_hdcp_display));
-		mod_hdcp_remove_display(&hdcp_w->hdcp, aconnector->base.index, &hdcp_w->output);
-
-		hdcp_w->link.adjust.hdcp2.force_type = MOD_HDCP_FORCE_TYPE_0;
-
-		if (enable_encryption) {
-			/* Explicitly set the saved SRM as sysfs call will be after we already enabled hdcp
-			 * (s3 resume case)
-			 */
-			if (hdcp_work->srm_size > 0)
-				psp_set_srm(hdcp_work->hdcp.config.psp.handle, hdcp_work->srm, hdcp_work->srm_size,
-					    &hdcp_work->srm_version);
-
-			display->adjust.disable = MOD_HDCP_DISPLAY_NOT_DISABLE;
-			if (content_type == DRM_MODE_HDCP_CONTENT_TYPE0) {
-				hdcp_w->link.adjust.hdcp1.disable = 0;
-				hdcp_w->link.adjust.hdcp2.force_type = MOD_HDCP_FORCE_TYPE_0;
-			} else if (content_type == DRM_MODE_HDCP_CONTENT_TYPE1) {
-				hdcp_w->link.adjust.hdcp1.disable = 1;
-				hdcp_w->link.adjust.hdcp2.force_type = MOD_HDCP_FORCE_TYPE_1;
-			}
+	memset(&link_adjust, 0, sizeof(link_adjust));
+	memset(&display_adjust, 0, sizeof(display_adjust));
 
-			schedule_delayed_work(&hdcp_w->property_validate_dwork,
-					      msecs_to_jiffies(DRM_HDCP_CHECK_PERIOD_MS));
-		} else {
-			display->adjust.disable = MOD_HDCP_DISPLAY_DISABLE_AUTHENTICATION;
-			hdcp_w->encryption_status[conn_index] = MOD_HDCP_ENCRYPTION_STATUS_HDCP_OFF;
-			cancel_delayed_work(&hdcp_w->property_validate_dwork);
-		}
+	if (enable_encryption) {
+		/* Explicitly set the saved SRM as sysfs call will be after we already enabled hdcp
+		 * (s3 resume case)
+		 */
+		if (hdcp_work->srm_size > 0)
+			psp_set_srm(hdcp_work->hdcp.config.psp.handle, hdcp_work->srm,
+				    hdcp_work->srm_size,
+				    &hdcp_work->srm_version);
+
+		display_adjust.disable = MOD_HDCP_DISPLAY_NOT_DISABLE;
 
-		display->state = MOD_HDCP_DISPLAY_ACTIVE;
+		link_adjust.auth_delay = 2;
+		link_adjust.retry_limit = MAX_NUM_OF_ATTEMPTS;
+
+		if (content_type == DRM_MODE_HDCP_CONTENT_TYPE0) {
+			link_adjust.hdcp2.force_type = MOD_HDCP_FORCE_TYPE_0;
+		} else if (content_type == DRM_MODE_HDCP_CONTENT_TYPE1) {
+			link_adjust.hdcp1.disable = 1;
+			link_adjust.hdcp2.force_type = MOD_HDCP_FORCE_TYPE_1;
+		}
+		link_adjust.hdcp2.use_fw_locality_check =
+				(dc->caps.fused_io_supported || dc->debug.hdcp_lc_force_fw_enable);
+		link_adjust.hdcp2.use_sw_locality_fallback = dc->debug.hdcp_lc_enable_sw_fallback;
+
+		schedule_delayed_work(&hdcp_w->property_validate_dwork,
+				      msecs_to_jiffies(DRM_HDCP_CHECK_PERIOD_MS));
+	} else {
+		display_adjust.disable = MOD_HDCP_DISPLAY_DISABLE_AUTHENTICATION;
+		hdcp_w->encryption_status[conn_index] = MOD_HDCP_ENCRYPTION_STATUS_HDCP_OFF;
+		cancel_delayed_work(&hdcp_w->property_validate_dwork);
 	}
 
-	mod_hdcp_add_display(&hdcp_w->hdcp, link, display, &hdcp_w->output);
+	mod_hdcp_update_display(&hdcp_w->hdcp, conn_index, &link_adjust, &display_adjust, &hdcp_w->output);
 
 	process_output(hdcp_w);
-	mutex_unlock(&hdcp_w->mutex);
 }
 
 static void hdcp_remove_display(struct hdcp_workqueue *hdcp_work,
-			 unsigned int link_index,
+				unsigned int link_index,
 			 struct amdgpu_dm_connector *aconnector)
 {
 	struct hdcp_workqueue *hdcp_w = &hdcp_work[link_index];
 	struct drm_connector_state *conn_state = aconnector->base.state;
 	unsigned int conn_index = aconnector->base.index;
 
-	mutex_lock(&hdcp_w->mutex);
-	hdcp_w->aconnector[conn_index] = aconnector;
+	guard(mutex)(&hdcp_w->mutex);
 
 	/* the removal of display will invoke auth reset -> hdcp destroy and
 	 * we'd expect the Content Protection (CP) property changed back to
@@ -238,20 +268,24 @@ static void hdcp_remove_display(struct hdcp_workqueue *hdcp_work,
 		conn_state->content_protection = DRM_MODE_CONTENT_PROTECTION_DESIRED;
 
 		DRM_DEBUG_DRIVER("[HDCP_DM] display %d, CP 2 -> 1, type %u, DPMS %u\n",
-			 aconnector->base.index, conn_state->hdcp_content_type, aconnector->base.dpms);
+				 aconnector->base.index, conn_state->hdcp_content_type,
+				 aconnector->base.dpms);
 	}
 
 	mod_hdcp_remove_display(&hdcp_w->hdcp, aconnector->base.index, &hdcp_w->output);
-
+	if (hdcp_w->aconnector[conn_index]) {
+		drm_connector_put(&hdcp_w->aconnector[conn_index]->base);
+		hdcp_w->aconnector[conn_index] = NULL;
+	}
 	process_output(hdcp_w);
-	mutex_unlock(&hdcp_w->mutex);
 }
+
 void hdcp_reset_display(struct hdcp_workqueue *hdcp_work, unsigned int link_index)
 {
 	struct hdcp_workqueue *hdcp_w = &hdcp_work[link_index];
 	unsigned int conn_index;
 
-	mutex_lock(&hdcp_w->mutex);
+	guard(mutex)(&hdcp_w->mutex);
 
 	mod_hdcp_reset_connection(&hdcp_w->hdcp,  &hdcp_w->output);
 
@@ -260,11 +294,13 @@ void hdcp_reset_display(struct hdcp_workqueue *hdcp_work, unsigned int link_inde
 	for (conn_index = 0; conn_index < AMDGPU_DM_MAX_DISPLAY_INDEX; conn_index++) {
 		hdcp_w->encryption_status[conn_index] =
 			MOD_HDCP_ENCRYPTION_STATUS_HDCP_OFF;
+		if (hdcp_w->aconnector[conn_index]) {
+			drm_connector_put(&hdcp_w->aconnector[conn_index]->base);
+			hdcp_w->aconnector[conn_index] = NULL;
+		}
 	}
 
 	process_output(hdcp_w);
-
-	mutex_unlock(&hdcp_w->mutex);
 }
 
 void hdcp_handle_cpirq(struct hdcp_workqueue *hdcp_work, unsigned int link_index)
@@ -274,17 +310,14 @@ void hdcp_handle_cpirq(struct hdcp_workqueue *hdcp_work, unsigned int link_index
 	schedule_work(&hdcp_w->cpirq_work);
 }
 
-
-
-
 static void event_callback(struct work_struct *work)
 {
 	struct hdcp_workqueue *hdcp_work;
 
 	hdcp_work = container_of(to_delayed_work(work), struct hdcp_workqueue,
-				      callback_dwork);
+				 callback_dwork);
 
-	mutex_lock(&hdcp_work->mutex);
+	guard(mutex)(&hdcp_work->mutex);
 
 	cancel_delayed_work(&hdcp_work->callback_dwork);
 
@@ -292,15 +325,12 @@ static void event_callback(struct work_struct *work)
 			       &hdcp_work->output);
 
 	process_output(hdcp_work);
-
-	mutex_unlock(&hdcp_work->mutex);
-
-
 }
 
 static void event_property_update(struct work_struct *work)
 {
-	struct hdcp_workqueue *hdcp_work = container_of(work, struct hdcp_workqueue, property_update_work);
+	struct hdcp_workqueue *hdcp_work = container_of(work, struct hdcp_workqueue,
+							property_update_work);
 	struct amdgpu_dm_connector *aconnector = NULL;
 	struct drm_device *dev;
 	long ret;
@@ -331,14 +361,13 @@ static void event_property_update(struct work_struct *work)
 			continue;
 
 		drm_modeset_lock(&dev->mode_config.connection_mutex, NULL);
-		mutex_lock(&hdcp_work->mutex);
+		guard(mutex)(&hdcp_work->mutex);
 
 		if (conn_state->commit) {
-			ret = wait_for_completion_interruptible_timeout(
-				&conn_state->commit->hw_done, 10 * HZ);
+			ret = wait_for_completion_interruptible_timeout(&conn_state->commit->hw_done,
+									10 * HZ);
 			if (ret == 0) {
-				DRM_ERROR(
-					"HDCP state unknown! Setting it to DESIRED");
+				DRM_ERROR("HDCP state unknown! Setting it to DESIRED\n");
 				hdcp_work->encryption_status[conn_index] =
 					MOD_HDCP_ENCRYPTION_STATUS_HDCP_OFF;
 			}
@@ -349,26 +378,21 @@ static void event_property_update(struct work_struct *work)
 				DRM_MODE_HDCP_CONTENT_TYPE0 &&
 				hdcp_work->encryption_status[conn_index] <=
 				MOD_HDCP_ENCRYPTION_STATUS_HDCP2_TYPE0_ON) {
-
 				DRM_DEBUG_DRIVER("[HDCP_DM] DRM_MODE_CONTENT_PROTECTION_ENABLED\n");
-				drm_hdcp_update_content_protection(
-					connector,
-					DRM_MODE_CONTENT_PROTECTION_ENABLED);
+				drm_hdcp_update_content_protection(connector,
+								   DRM_MODE_CONTENT_PROTECTION_ENABLED);
 			} else if (conn_state->hdcp_content_type ==
 					DRM_MODE_HDCP_CONTENT_TYPE1 &&
 					hdcp_work->encryption_status[conn_index] ==
 					MOD_HDCP_ENCRYPTION_STATUS_HDCP2_TYPE1_ON) {
-				drm_hdcp_update_content_protection(
-					connector,
-					DRM_MODE_CONTENT_PROTECTION_ENABLED);
+				drm_hdcp_update_content_protection(connector,
+								   DRM_MODE_CONTENT_PROTECTION_ENABLED);
 			}
 		} else {
 			DRM_DEBUG_DRIVER("[HDCP_DM] DRM_MODE_CONTENT_PROTECTION_DESIRED\n");
-			drm_hdcp_update_content_protection(
-				connector, DRM_MODE_CONTENT_PROTECTION_DESIRED);
-
+			drm_hdcp_update_content_protection(connector,
+							   DRM_MODE_CONTENT_PROTECTION_DESIRED);
 		}
-		mutex_unlock(&hdcp_work->mutex);
 		drm_modeset_unlock(&dev->mode_config.connection_mutex);
 	}
 }
@@ -381,7 +405,7 @@ static void event_property_validate(struct work_struct *work)
 	struct amdgpu_dm_connector *aconnector;
 	unsigned int conn_index;
 
-	mutex_lock(&hdcp_work->mutex);
+	guard(mutex)(&hdcp_work->mutex);
 
 	for (conn_index = 0; conn_index < AMDGPU_DM_MAX_DISPLAY_INDEX;
 	     conn_index++) {
@@ -402,7 +426,7 @@ static void event_property_validate(struct work_struct *work)
 				       &query);
 
 		DRM_DEBUG_DRIVER("[HDCP_DM] disp %d, connector->CP %u, (query, work): (%d, %d)\n",
-			aconnector->base.index,
+				 aconnector->base.index,
 			aconnector->base.state->content_protection,
 			query.encryption_status,
 			hdcp_work->encryption_status[conn_index]);
@@ -410,7 +434,8 @@ static void event_property_validate(struct work_struct *work)
 		if (query.encryption_status !=
 		    hdcp_work->encryption_status[conn_index]) {
 			DRM_DEBUG_DRIVER("[HDCP_DM] encryption_status change from %x to %x\n",
-				hdcp_work->encryption_status[conn_index], query.encryption_status);
+					 hdcp_work->encryption_status[conn_index],
+					 query.encryption_status);
 
 			hdcp_work->encryption_status[conn_index] =
 				query.encryption_status;
@@ -420,8 +445,6 @@ static void event_property_validate(struct work_struct *work)
 			schedule_work(&hdcp_work->property_update_work);
 		}
 	}
-
-	mutex_unlock(&hdcp_work->mutex);
 }
 
 static void event_watchdog_timer(struct work_struct *work)
@@ -429,10 +452,10 @@ static void event_watchdog_timer(struct work_struct *work)
 	struct hdcp_workqueue *hdcp_work;
 
 	hdcp_work = container_of(to_delayed_work(work),
-				      struct hdcp_workqueue,
+				 struct hdcp_workqueue,
 				      watchdog_timer_dwork);
 
-	mutex_lock(&hdcp_work->mutex);
+	guard(mutex)(&hdcp_work->mutex);
 
 	cancel_delayed_work(&hdcp_work->watchdog_timer_dwork);
 
@@ -441,9 +464,6 @@ static void event_watchdog_timer(struct work_struct *work)
 			       &hdcp_work->output);
 
 	process_output(hdcp_work);
-
-	mutex_unlock(&hdcp_work->mutex);
-
 }
 
 static void event_cpirq(struct work_struct *work)
@@ -452,17 +472,13 @@ static void event_cpirq(struct work_struct *work)
 
 	hdcp_work = container_of(work, struct hdcp_workqueue, cpirq_work);
 
-	mutex_lock(&hdcp_work->mutex);
+	guard(mutex)(&hdcp_work->mutex);
 
 	mod_hdcp_process_event(&hdcp_work->hdcp, MOD_HDCP_EVENT_CPIRQ, &hdcp_work->output);
 
 	process_output(hdcp_work);
-
-	mutex_unlock(&hdcp_work->mutex);
-
 }
 
-
 void hdcp_destroy(struct kobject *kobj, struct hdcp_workqueue *hdcp_work)
 {
 	int i = 0;
@@ -470,6 +486,7 @@ void hdcp_destroy(struct kobject *kobj, struct hdcp_workqueue *hdcp_work)
 	for (i = 0; i < hdcp_work->max_link; i++) {
 		cancel_delayed_work_sync(&hdcp_work[i].callback_dwork);
 		cancel_delayed_work_sync(&hdcp_work[i].watchdog_timer_dwork);
+		cancel_delayed_work_sync(&hdcp_work[i].property_validate_dwork);
 	}
 
 	sysfs_remove_bin_file(kobj, &hdcp_work[0].attr);
@@ -478,15 +495,12 @@ void hdcp_destroy(struct kobject *kobj, struct hdcp_workqueue *hdcp_work)
 	kfree(hdcp_work);
 }
 
-
 static bool enable_assr(void *handle, struct dc_link *link)
 {
-
 	struct hdcp_workqueue *hdcp_work = handle;
 	struct mod_hdcp hdcp = hdcp_work->hdcp;
 	struct psp_context *psp = hdcp.config.psp.handle;
 	struct ta_dtm_shared_memory *dtm_cmd;
-	bool res = true;
 
 	if (!psp->dtm_context.context.initialized) {
 		DRM_INFO("Failed to enable ASSR, DTM TA is not initialized.");
@@ -495,23 +509,22 @@ static bool enable_assr(void *handle, struct dc_link *link)
 
 	dtm_cmd = (struct ta_dtm_shared_memory *)psp->dtm_context.context.mem_context.shared_buf;
 
-	mutex_lock(&psp->dtm_context.mutex);
+	guard(mutex)(&psp->dtm_context.mutex);
 	memset(dtm_cmd, 0, sizeof(struct ta_dtm_shared_memory));
 
 	dtm_cmd->cmd_id = TA_DTM_COMMAND__TOPOLOGY_ASSR_ENABLE;
-	dtm_cmd->dtm_in_message.topology_assr_enable.display_topology_dig_be_index = link->link_enc_hw_inst;
+	dtm_cmd->dtm_in_message.topology_assr_enable.display_topology_dig_be_index =
+		link->link_enc_hw_inst;
 	dtm_cmd->dtm_status = TA_DTM_STATUS__GENERIC_FAILURE;
 
 	psp_dtm_invoke(psp, dtm_cmd->cmd_id);
 
 	if (dtm_cmd->dtm_status != TA_DTM_STATUS__SUCCESS) {
 		DRM_INFO("Failed to enable ASSR");
-		res = false;
+		return false;
 	}
 
-	mutex_unlock(&psp->dtm_context.mutex);
-
-	return res;
+	return true;
 }
 
 static void update_config(void *handle, struct cp_psp_stream_config *config)
@@ -519,11 +532,13 @@ static void update_config(void *handle, struct cp_psp_stream_config *config)
 	struct hdcp_workqueue *hdcp_work = handle;
 	struct amdgpu_dm_connector *aconnector = config->dm_stream_ctx;
 	int link_index = aconnector->dc_link->link_index;
+	unsigned int conn_index = aconnector->base.index;
 	struct mod_hdcp_display *display = &hdcp_work[link_index].display;
 	struct mod_hdcp_link *link = &hdcp_work[link_index].link;
-	struct drm_connector_state *conn_state;
+	struct hdcp_workqueue *hdcp_w = &hdcp_work[link_index];
 	struct dc_sink *sink = NULL;
 	bool link_is_hdcp14 = false;
+	const struct dc *dc = aconnector->dc_link->dc;
 
 	if (config->dpms_off) {
 		hdcp_remove_display(hdcp_work, link_index, aconnector);
@@ -541,7 +556,7 @@ static void update_config(void *handle, struct cp_psp_stream_config *config)
 	else if (aconnector->dc_em_sink)
 		sink = aconnector->dc_em_sink;
 
-	if (sink != NULL)
+	if (sink)
 		link->mode = mod_hdcp_signal_type_to_operation_mode(sink->sink_signal);
 
 	display->controller = CONTROLLER_ID_D0 + config->otg_inst;
@@ -563,20 +578,32 @@ static void update_config(void *handle, struct cp_psp_stream_config *config)
 	link->dp.usb4_enabled = config->usb4_enabled;
 	display->adjust.disable = MOD_HDCP_DISPLAY_DISABLE_AUTHENTICATION;
 	link->adjust.auth_delay = 2;
+	link->adjust.retry_limit = MAX_NUM_OF_ATTEMPTS;
 	link->adjust.hdcp1.disable = 0;
-	conn_state = aconnector->base.state;
+	link->adjust.hdcp2.use_fw_locality_check = (dc->caps.fused_io_supported || dc->debug.hdcp_lc_force_fw_enable);
+	link->adjust.hdcp2.use_sw_locality_fallback = dc->debug.hdcp_lc_enable_sw_fallback;
+	hdcp_w->encryption_status[display->index] = MOD_HDCP_ENCRYPTION_STATUS_HDCP_OFF;
 
 	DRM_DEBUG_DRIVER("[HDCP_DM] display %d, CP %d, type %d\n", aconnector->base.index,
-			(!!aconnector->base.state) ? aconnector->base.state->content_protection : -1,
-			(!!aconnector->base.state) ? aconnector->base.state->hdcp_content_type : -1);
+			 (!!aconnector->base.state) ?
+			 aconnector->base.state->content_protection : -1,
+			 (!!aconnector->base.state) ?
+			 aconnector->base.state->hdcp_content_type : -1);
 
-	if (conn_state)
-		hdcp_update_display(hdcp_work, link_index, aconnector,
-			conn_state->hdcp_content_type, false);
-}
+	guard(mutex)(&hdcp_w->mutex);
 
+	mod_hdcp_add_display(&hdcp_w->hdcp, link, display, &hdcp_w->output);
+	drm_connector_get(&aconnector->base);
+	if (hdcp_w->aconnector[conn_index])
+		drm_connector_put(&hdcp_w->aconnector[conn_index]->base);
+	hdcp_w->aconnector[conn_index] = aconnector;
+	process_output(hdcp_w);
+}
 
-/* NOTE: From the usermodes prospective you only need to call write *ONCE*, the kernel
+/**
+ * DOC: Add sysfs interface for set/get srm
+ *
+ * NOTE: From the usermodes prospective you only need to call write *ONCE*, the kernel
  *      will automatically call once or twice depending on the size
  *
  * call: "cat file > /sys/class/drm/card0/device/hdcp_srm" from usermode no matter what the size is
@@ -587,23 +614,23 @@ static void update_config(void *handle, struct cp_psp_stream_config *config)
  * sysfs interface doesn't tell us the size we will get so we are sending partial SRMs to psp and on
  * the last call we will send the full SRM. PSP will fail on every call before the last.
  *
- * This means we don't know if the SRM is good until the last call. And because of this limitation we
- * cannot throw errors early as it will stop the kernel from writing to sysfs
+ * This means we don't know if the SRM is good until the last call. And because of this
+ * limitation we cannot throw errors early as it will stop the kernel from writing to sysfs
  *
  * Example 1:
- * 	Good SRM size = 5096
- * 	first call to write 4096 -> PSP fails
- * 	Second call to write 1000 -> PSP Pass -> SRM is set
+ *	Good SRM size = 5096
+ *	first call to write 4096 -> PSP fails
+ *	Second call to write 1000 -> PSP Pass -> SRM is set
  *
  * Example 2:
- * 	Bad SRM size = 4096
- * 	first call to write 4096 -> PSP fails (This is the same as above, but we don't know if this
- * 	is the last call)
+ *	Bad SRM size = 4096
+ *	first call to write 4096 -> PSP fails (This is the same as above, but we don't know if this
+ *	is the last call)
  *
  * Solution?:
- * 	1: Parse the SRM? -> It is signed so we don't know the EOF
- * 	2: We can have another sysfs that passes the size before calling set. -> simpler solution
- * 	below
+ *	1: Parse the SRM? -> It is signed so we don't know the EOF
+ *	2: We can have another sysfs that passes the size before calling set. -> simpler solution
+ *	below
  *
  * Easy Solution:
  * Always call get after Set to verify if set was successful.
@@ -612,20 +639,21 @@ static void update_config(void *handle, struct cp_psp_stream_config *config)
  * +----------------------+
  * PSP will only update its srm if its older than the one we are trying to load.
  * Always do set first than get.
- * 	-if we try to "1. SET" a older version PSP will reject it and we can "2. GET" the newer
- * 	version and save it
+ *	-if we try to "1. SET" a older version PSP will reject it and we can "2. GET" the newer
+ *	version and save it
  *
- * 	-if we try to "1. SET" a newer version PSP will accept it and we can "2. GET" the
- * 	same(newer) version back and save it
+ *	-if we try to "1. SET" a newer version PSP will accept it and we can "2. GET" the
+ *	same(newer) version back and save it
  *
- * 	-if we try to "1. SET" a newer version and PSP rejects it. That means the format is
- * 	incorrect/corrupted and we should correct our SRM by getting it from PSP
+ *	-if we try to "1. SET" a newer version and PSP rejects it. That means the format is
+ *	incorrect/corrupted and we should correct our SRM by getting it from PSP
  */
-static ssize_t srm_data_write(struct file *filp, struct kobject *kobj, struct bin_attribute *bin_attr, char *buffer,
+static ssize_t srm_data_write(struct file *filp, struct kobject *kobj,
+			      const struct bin_attribute *bin_attr, char *buffer,
 			      loff_t pos, size_t count)
 {
 	struct hdcp_workqueue *work;
-	uint32_t srm_version = 0;
+	u32 srm_version = 0;
 
 	work = container_of(bin_attr, struct hdcp_workqueue, attr);
 	link_lock(work, true);
@@ -639,19 +667,19 @@ static ssize_t srm_data_write(struct file *filp, struct kobject *kobj, struct bi
 		work->srm_version = srm_version;
 	}
 
-
 	link_lock(work, false);
 
 	return count;
 }
 
-static ssize_t srm_data_read(struct file *filp, struct kobject *kobj, struct bin_attribute *bin_attr, char *buffer,
+static ssize_t srm_data_read(struct file *filp, struct kobject *kobj,
+			     const struct bin_attribute *bin_attr, char *buffer,
 			     loff_t pos, size_t count)
 {
 	struct hdcp_workqueue *work;
-	uint8_t *srm = NULL;
-	uint32_t srm_version;
-	uint32_t srm_size;
+	u8 *srm = NULL;
+	u32 srm_version;
+	u32 srm_size;
 	size_t ret = count;
 
 	work = container_of(bin_attr, struct hdcp_workqueue, attr);
@@ -684,12 +712,12 @@ ret:
 /* From the hdcp spec (5.Renewability) SRM needs to be stored in a non-volatile memory.
  *
  * For example,
- * 	if Application "A" sets the SRM (ver 2) and we reboot/suspend and later when Application "B"
- * 	needs to use HDCP, the version in PSP should be SRM(ver 2). So SRM should be persistent
- * 	across boot/reboots/suspend/resume/shutdown
+ *	if Application "A" sets the SRM (ver 2) and we reboot/suspend and later when Application "B"
+ *	needs to use HDCP, the version in PSP should be SRM(ver 2). So SRM should be persistent
+ *	across boot/reboots/suspend/resume/shutdown
  *
- * Currently when the system goes down (suspend/shutdown) the SRM is cleared from PSP. For HDCP we need
- * to make the SRM persistent.
+ * Currently when the system goes down (suspend/shutdown) the SRM is cleared from PSP. For HDCP
+ * we need to make the SRM persistent.
  *
  * -PSP owns the checking of SRM but doesn't have the ability to store it in a non-volatile memory.
  * -The kernel cannot write to the file systems.
@@ -699,8 +727,8 @@ ret:
  *
  * Usermode can read/write to/from PSP using the sysfs interface
  * For example:
- * 	to save SRM from PSP to storage : cat /sys/class/drm/card0/device/hdcp_srm > srmfile
- * 	to load from storage to PSP: cat srmfile > /sys/class/drm/card0/device/hdcp_srm
+ *	to save SRM from PSP to storage : cat /sys/class/drm/card0/device/hdcp_srm > srmfile
+ *	to load from storage to PSP: cat srmfile > /sys/class/drm/card0/device/hdcp_srm
  */
 static const struct bin_attribute data_attr = {
 	.attr = {.name = "hdcp_srm", .mode = 0664},
@@ -709,10 +737,9 @@ static const struct bin_attribute data_attr = {
 	.read = srm_data_read,
 };
 
-
-struct hdcp_workqueue *hdcp_create_workqueue(struct amdgpu_device *adev, struct cp_psp *cp_psp, struct dc *dc)
+struct hdcp_workqueue *hdcp_create_workqueue(struct amdgpu_device *adev,
+					     struct cp_psp *cp_psp, struct dc *dc)
 {
-
 	int max_caps = dc->caps.max_links;
 	struct hdcp_workqueue *hdcp_work;
 	int i = 0;
@@ -721,14 +748,16 @@ struct hdcp_workqueue *hdcp_create_workqueue(struct amdgpu_device *adev, struct
 	if (ZERO_OR_NULL_PTR(hdcp_work))
 		return NULL;
 
-	hdcp_work->srm = kcalloc(PSP_HDCP_SRM_FIRST_GEN_MAX_SIZE, sizeof(*hdcp_work->srm), GFP_KERNEL);
+	hdcp_work->srm = kcalloc(PSP_HDCP_SRM_FIRST_GEN_MAX_SIZE,
+				 sizeof(*hdcp_work->srm), GFP_KERNEL);
 
-	if (hdcp_work->srm == NULL)
+	if (!hdcp_work->srm)
 		goto fail_alloc_context;
 
-	hdcp_work->srm_temp = kcalloc(PSP_HDCP_SRM_FIRST_GEN_MAX_SIZE, sizeof(*hdcp_work->srm_temp), GFP_KERNEL);
+	hdcp_work->srm_temp = kcalloc(PSP_HDCP_SRM_FIRST_GEN_MAX_SIZE,
+				      sizeof(*hdcp_work->srm_temp), GFP_KERNEL);
 
-	if (hdcp_work->srm_temp == NULL)
+	if (!hdcp_work->srm_temp)
 		goto fail_alloc_context;
 
 	hdcp_work->max_link = max_caps;
@@ -742,17 +771,30 @@ struct hdcp_workqueue *hdcp_create_workqueue(struct amdgpu_device *adev, struct
 		INIT_DELAYED_WORK(&hdcp_work[i].watchdog_timer_dwork, event_watchdog_timer);
 		INIT_DELAYED_WORK(&hdcp_work[i].property_validate_dwork, event_property_validate);
 
-		hdcp_work[i].hdcp.config.psp.handle = &adev->psp;
-		if (dc->ctx->dce_version == DCN_VERSION_3_1 ||
+		struct mod_hdcp_config *config = &hdcp_work[i].hdcp.config;
+		struct mod_hdcp_ddc_funcs *ddc_funcs = &config->ddc.funcs;
+
+		config->psp.handle = &adev->psp;
+		if (dc->ctx->dce_version == DCN_VERSION_3_1  ||
 		    dc->ctx->dce_version == DCN_VERSION_3_14 ||
 		    dc->ctx->dce_version == DCN_VERSION_3_15 ||
-		    dc->ctx->dce_version == DCN_VERSION_3_16)
-			hdcp_work[i].hdcp.config.psp.caps.dtm_v3_supported = 1;
-		hdcp_work[i].hdcp.config.ddc.handle = dc_get_link_at_index(dc, i);
-		hdcp_work[i].hdcp.config.ddc.funcs.write_i2c = lp_write_i2c;
-		hdcp_work[i].hdcp.config.ddc.funcs.read_i2c = lp_read_i2c;
-		hdcp_work[i].hdcp.config.ddc.funcs.write_dpcd = lp_write_dpcd;
-		hdcp_work[i].hdcp.config.ddc.funcs.read_dpcd = lp_read_dpcd;
+		    dc->ctx->dce_version == DCN_VERSION_3_16 ||
+		    dc->ctx->dce_version == DCN_VERSION_3_2  ||
+		    dc->ctx->dce_version == DCN_VERSION_3_21 ||
+		    dc->ctx->dce_version == DCN_VERSION_3_5  ||
+		    dc->ctx->dce_version == DCN_VERSION_3_51 ||
+		    dc->ctx->dce_version == DCN_VERSION_3_6  ||
+		    dc->ctx->dce_version == DCN_VERSION_4_01)
+			config->psp.caps.dtm_v3_supported = 1;
+
+		config->ddc.handle = dc_get_link_at_index(dc, i);
+
+		ddc_funcs->write_i2c = lp_write_i2c;
+		ddc_funcs->read_i2c = lp_read_i2c;
+		ddc_funcs->write_dpcd = lp_write_dpcd;
+		ddc_funcs->read_dpcd = lp_read_dpcd;
+		ddc_funcs->atomic_write_poll_read_i2c = lp_atomic_write_poll_read_i2c;
+		ddc_funcs->atomic_write_poll_read_aux = lp_atomic_write_poll_read_aux;
 
 		memset(hdcp_work[i].aconnector, 0,
 		       sizeof(struct amdgpu_dm_connector *) *
@@ -781,10 +823,5 @@ fail_alloc_context:
 	kfree(hdcp_work);
 
 	return NULL;
-
-
-
 }
 
-
-
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_hdcp.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_hdcp.h
index 69b445b011c8..4faa344f196e 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_hdcp.h
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_hdcp.h
@@ -1,3 +1,4 @@
+/* SPDX-License-Identifier: MIT */
 /*
  * Copyright 2019 Advanced Micro Devices, Inc.
  *
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c
index 330ab036c830..ac98c746c3de 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c
@@ -1,3 +1,4 @@
+// SPDX-License-Identifier: MIT
 /*
  * Copyright 2015 Advanced Micro Devices, Inc.
  *
@@ -23,6 +24,8 @@
  *
  */
 
+#include <acpi/video.h>
+
 #include <linux/string.h>
 #include <linux/acpi.h>
 #include <linux/i2c.h>
@@ -31,6 +34,7 @@
 #include <drm/drm_probe_helper.h>
 #include <drm/amdgpu_drm.h>
 #include <drm/drm_edid.h>
+#include <drm/drm_fixed.h>
 
 #include "dm_services.h"
 #include "amdgpu.h"
@@ -43,19 +47,60 @@
 
 #include "dm_helpers.h"
 #include "ddc_service_types.h"
+#include "clk_mgr.h"
 
-/* MST Dock */
-static const uint8_t SYNAPTICS_DEVICE_ID[] = "SYNA";
+static u32 edid_extract_panel_id(struct edid *edid)
+{
+	return (u32)edid->mfg_id[0] << 24   |
+	       (u32)edid->mfg_id[1] << 16   |
+	       (u32)EDID_PRODUCT_ID(edid);
+}
 
-/* dm_helpers_parse_edid_caps
- *
- * Parse edid caps
+static void apply_edid_quirks(struct drm_device *dev, struct edid *edid, struct dc_edid_caps *edid_caps)
+{
+	uint32_t panel_id = edid_extract_panel_id(edid);
+
+	switch (panel_id) {
+	/* Workaround for monitors that need a delay after detecting the link */
+	case drm_edid_encode_panel_id('G', 'B', 'T', 0x3215):
+		drm_dbg_driver(dev, "Add 10s delay for link detection for panel id %X\n", panel_id);
+		edid_caps->panel_patch.wait_after_dpcd_poweroff_ms = 10000;
+		break;
+	/* Workaround for some monitors which does not work well with FAMS */
+	case drm_edid_encode_panel_id('S', 'A', 'M', 0x0E5E):
+	case drm_edid_encode_panel_id('S', 'A', 'M', 0x7053):
+	case drm_edid_encode_panel_id('S', 'A', 'M', 0x71AC):
+		drm_dbg_driver(dev, "Disabling FAMS on monitor with panel id %X\n", panel_id);
+		edid_caps->panel_patch.disable_fams = true;
+		break;
+	/* Workaround for some monitors that do not clear DPCD 0x317 if FreeSync is unsupported */
+	case drm_edid_encode_panel_id('A', 'U', 'O', 0xA7AB):
+	case drm_edid_encode_panel_id('A', 'U', 'O', 0xE69B):
+	case drm_edid_encode_panel_id('B', 'O', 'E', 0x092A):
+	case drm_edid_encode_panel_id('L', 'G', 'D', 0x06D1):
+	case drm_edid_encode_panel_id('M', 'S', 'F', 0x1003):
+		drm_dbg_driver(dev, "Clearing DPCD 0x317 on monitor with panel id %X\n", panel_id);
+		edid_caps->panel_patch.remove_sink_ext_caps = true;
+		break;
+	case drm_edid_encode_panel_id('S', 'D', 'C', 0x4154):
+	case drm_edid_encode_panel_id('S', 'D', 'C', 0x4171):
+		drm_dbg_driver(dev, "Disabling VSC on monitor with panel id %X\n", panel_id);
+		edid_caps->panel_patch.disable_colorimetry = true;
+		break;
+	default:
+		return;
+	}
+}
+
+/**
+ * dm_helpers_parse_edid_caps() - Parse edid caps
  *
+ * @link: current detected link
  * @edid:	[in] pointer to edid
- *  edid_caps:	[in] pointer to edid caps
- * @return
- *	void
- * */
+ * @edid_caps:	[in] pointer to edid caps
+ *
+ * Return: void
+ */
 enum dc_edid_status dm_helpers_parse_edid_caps(
 		struct dc_link *link,
 		const struct dc_edid *edid,
@@ -63,6 +108,7 @@ enum dc_edid_status dm_helpers_parse_edid_caps(
 {
 	struct amdgpu_dm_connector *aconnector = link->priv;
 	struct drm_connector *connector = &aconnector->base;
+	struct drm_device *dev = connector->dev;
 	struct edid *edid_buf = edid ? (struct edid *) edid->raw_edid : NULL;
 	struct cea_sad *sads;
 	int sad_count = -1;
@@ -85,6 +131,7 @@ enum dc_edid_status dm_helpers_parse_edid_caps(
 	edid_caps->serial_number = edid_buf->serial;
 	edid_caps->manufacture_week = edid_buf->mfg_week;
 	edid_caps->manufacture_year = edid_buf->mfg_year;
+	edid_caps->analog = !(edid_buf->input & DRM_EDID_INPUT_DIGITAL);
 
 	drm_edid_get_monitor_name(edid_buf,
 				  edid_caps->display_name,
@@ -92,11 +139,13 @@ enum dc_edid_status dm_helpers_parse_edid_caps(
 
 	edid_caps->edid_hdmi = connector->display_info.is_hdmi;
 
+	apply_edid_quirks(dev, edid_buf, edid_caps);
+
 	sad_count = drm_edid_to_sad((struct edid *) edid->raw_edid, &sads);
 	if (sad_count <= 0)
 		return result;
 
-	edid_caps->audio_mode_count = sad_count < DC_MAX_AUDIO_DESC_COUNT ? sad_count : DC_MAX_AUDIO_DESC_COUNT;
+	edid_caps->audio_mode_count = min(sad_count, DC_MAX_AUDIO_DESC_COUNT);
 	for (i = 0; i < edid_caps->audio_mode_count; ++i) {
 		struct cea_sad *sad = &sads[i];
 
@@ -181,37 +230,35 @@ void dm_helpers_dp_update_branch_info(
 {}
 
 static void dm_helpers_construct_old_payload(
-			struct dc_link *link,
-			int pbn_per_slot,
+			struct drm_dp_mst_topology_mgr *mgr,
+			struct drm_dp_mst_topology_state *mst_state,
 			struct drm_dp_mst_atomic_payload *new_payload,
 			struct drm_dp_mst_atomic_payload *old_payload)
 {
-	struct link_mst_stream_allocation_table current_link_table =
-									link->mst_stream_alloc_table;
-	struct link_mst_stream_allocation *dc_alloc;
-	int i;
+	struct drm_dp_mst_atomic_payload *pos;
+	int pbn_per_slot = dfixed_trunc(mst_state->pbn_div);
+	u8 next_payload_vc_start = mgr->next_start_slot;
+	u8 payload_vc_start = new_payload->vc_start_slot;
+	u8 allocated_time_slots;
 
 	*old_payload = *new_payload;
 
 	/* Set correct time_slots/PBN of old payload.
 	 * other fields (delete & dsc_enabled) in
 	 * struct drm_dp_mst_atomic_payload are don't care fields
-	 * while calling drm_dp_remove_payload()
+	 * while calling drm_dp_remove_payload_part2()
 	 */
-	for (i = 0; i < current_link_table.stream_count; i++) {
-		dc_alloc =
-			&current_link_table.stream_allocations[i];
-
-		if (dc_alloc->vcp_id == new_payload->vcpi) {
-			old_payload->time_slots = dc_alloc->slot_count;
-			old_payload->pbn = dc_alloc->slot_count * pbn_per_slot;
-			break;
-		}
+	list_for_each_entry(pos, &mst_state->payloads, next) {
+		if (pos != new_payload &&
+		    pos->vc_start_slot > payload_vc_start &&
+		    pos->vc_start_slot < next_payload_vc_start)
+			next_payload_vc_start = pos->vc_start_slot;
 	}
 
-	/* make sure there is an old payload*/
-	ASSERT(i != current_link_table.stream_count);
+	allocated_time_slots = next_payload_vc_start - payload_vc_start;
 
+	old_payload->time_slots = allocated_time_slots;
+	old_payload->pbn = allocated_time_slots * pbn_per_slot;
 }
 
 /*
@@ -232,34 +279,35 @@ bool dm_helpers_dp_mst_write_payload_allocation_table(
 	/* Accessing the connector state is required for vcpi_slots allocation
 	 * and directly relies on behaviour in commit check
 	 * that blocks before commit guaranteeing that the state
-	 * is not gonna be swapped while still in use in commit tail */
+	 * is not gonna be swapped while still in use in commit tail
+	 */
 
 	if (!aconnector || !aconnector->mst_root)
 		return false;
 
 	mst_mgr = &aconnector->mst_root->mst_mgr;
 	mst_state = to_drm_dp_mst_topology_state(mst_mgr->base.state);
-
-	/* It's OK for this to fail */
 	new_payload = drm_atomic_get_mst_payload_state(mst_state, aconnector->mst_output_port);
 
 	if (enable) {
 		target_payload = new_payload;
 
+		/* It's OK for this to fail */
 		drm_dp_add_payload_part1(mst_mgr, mst_state, new_payload);
 	} else {
 		/* construct old payload by VCPI*/
-		dm_helpers_construct_old_payload(stream->link, mst_state->pbn_div,
-						new_payload, &old_payload);
+		dm_helpers_construct_old_payload(mst_mgr, mst_state,
+						 new_payload, &old_payload);
 		target_payload = &old_payload;
 
-		drm_dp_remove_payload(mst_mgr, mst_state, &old_payload, new_payload);
+		drm_dp_remove_payload_part1(mst_mgr, mst_state, new_payload);
 	}
 
 	/* mst_mgr->->payloads are VC payload notify MST branch using DPCD or
 	 * AUX message. The sequence is slot 1-63 allocated sequence for each
 	 * stream. AMD ASIC stream slot allocation should follow the same
-	 * sequence. copy DRM MST allocation to dc */
+	 * sequence. copy DRM MST allocation to dc
+	 */
 	fill_dc_mst_payload_table_from_drm(stream->link, enable, target_payload, proposed_table);
 
 	return true;
@@ -311,15 +359,14 @@ enum act_return_status dm_helpers_dp_mst_poll_for_allocation_change_trigger(
 	return ACT_SUCCESS;
 }
 
-bool dm_helpers_dp_mst_send_payload_allocation(
+void dm_helpers_dp_mst_send_payload_allocation(
 		struct dc_context *ctx,
-		const struct dc_stream_state *stream,
-		bool enable)
+		const struct dc_stream_state *stream)
 {
 	struct amdgpu_dm_connector *aconnector;
 	struct drm_dp_mst_topology_state *mst_state;
 	struct drm_dp_mst_topology_mgr *mst_mgr;
-	struct drm_dp_mst_atomic_payload *payload;
+	struct drm_dp_mst_atomic_payload *new_payload;
 	enum mst_progress_status set_flag = MST_ALLOCATE_NEW_PAYLOAD;
 	enum mst_progress_status clr_flag = MST_CLEAR_ALLOCATED_PAYLOAD;
 	int ret = 0;
@@ -327,20 +374,13 @@ bool dm_helpers_dp_mst_send_payload_allocation(
 	aconnector = (struct amdgpu_dm_connector *)stream->dm_stream_context;
 
 	if (!aconnector || !aconnector->mst_root)
-		return false;
+		return;
 
 	mst_mgr = &aconnector->mst_root->mst_mgr;
 	mst_state = to_drm_dp_mst_topology_state(mst_mgr->base.state);
+	new_payload = drm_atomic_get_mst_payload_state(mst_state, aconnector->mst_output_port);
 
-	payload = drm_atomic_get_mst_payload_state(mst_state, aconnector->mst_output_port);
-
-	if (!enable) {
-		set_flag = MST_CLEAR_ALLOCATED_PAYLOAD;
-		clr_flag = MST_ALLOCATE_NEW_PAYLOAD;
-	}
-
-	if (enable)
-		ret = drm_dp_add_payload_part2(mst_mgr, mst_state->base.state, payload);
+	ret = drm_dp_add_payload_part2(mst_mgr, new_payload);
 
 	if (ret) {
 		amdgpu_dm_set_mst_status(&aconnector->mst_status,
@@ -351,10 +391,36 @@ bool dm_helpers_dp_mst_send_payload_allocation(
 		amdgpu_dm_set_mst_status(&aconnector->mst_status,
 			clr_flag, false);
 	}
-
-	return true;
 }
 
+void dm_helpers_dp_mst_update_mst_mgr_for_deallocation(
+		struct dc_context *ctx,
+		const struct dc_stream_state *stream)
+{
+	struct amdgpu_dm_connector *aconnector;
+	struct drm_dp_mst_topology_state *mst_state;
+	struct drm_dp_mst_topology_mgr *mst_mgr;
+	struct drm_dp_mst_atomic_payload *new_payload, old_payload;
+	enum mst_progress_status set_flag = MST_CLEAR_ALLOCATED_PAYLOAD;
+	enum mst_progress_status clr_flag = MST_ALLOCATE_NEW_PAYLOAD;
+
+	aconnector = (struct amdgpu_dm_connector *)stream->dm_stream_context;
+
+	if (!aconnector || !aconnector->mst_root)
+		return;
+
+	mst_mgr = &aconnector->mst_root->mst_mgr;
+	mst_state = to_drm_dp_mst_topology_state(mst_mgr->base.state);
+	new_payload = drm_atomic_get_mst_payload_state(mst_state, aconnector->mst_output_port);
+	dm_helpers_construct_old_payload(mst_mgr, mst_state,
+					 new_payload, &old_payload);
+
+	drm_dp_remove_payload_part2(mst_mgr, mst_state, &old_payload, new_payload);
+
+	amdgpu_dm_set_mst_status(&aconnector->mst_status, set_flag, true);
+	amdgpu_dm_set_mst_status(&aconnector->mst_status, clr_flag, false);
+ }
+
 void dm_dtn_log_begin(struct dc_context *ctx,
 	struct dc_log_buffer_ctx *log_ctx)
 {
@@ -403,7 +469,7 @@ void dm_dtn_log_append_v(struct dc_context *ctx,
 	total = log_ctx->pos + n + 1;
 
 	if (total > log_ctx->size) {
-		char *buf = (char *)kvcalloc(total, sizeof(char), GFP_KERNEL);
+		char *buf = kvcalloc(total, sizeof(char), GFP_KERNEL);
 
 		if (buf) {
 			memcpy(buf, log_ctx->buf, log_ctx->pos);
@@ -509,10 +575,8 @@ bool dm_helpers_dp_read_dpcd(
 
 	struct amdgpu_dm_connector *aconnector = link->priv;
 
-	if (!aconnector) {
-		DC_LOG_DC("Failed to find connector for link!\n");
+	if (!aconnector)
 		return false;
-	}
 
 	return drm_dp_dpcd_read(&aconnector->dm_dp_aux.aux, address, data,
 				size) == size;
@@ -527,10 +591,8 @@ bool dm_helpers_dp_write_dpcd(
 {
 	struct amdgpu_dm_connector *aconnector = link->priv;
 
-	if (!aconnector) {
-		DRM_ERROR("Failed to find connector for link!");
+	if (!aconnector)
 		return false;
-	}
 
 	return drm_dp_dpcd_write(&aconnector->dm_dp_aux.aux,
 			address, (uint8_t *)data, size) > 0;
@@ -571,6 +633,19 @@ bool dm_helpers_submit_i2c(
 	return result;
 }
 
+bool dm_helpers_execute_fused_io(
+		struct dc_context *ctx,
+		struct dc_link *link,
+		union dmub_rb_cmd *commands,
+		uint8_t count,
+		uint32_t timeout_us
+)
+{
+	struct amdgpu_device *dev = ctx->driver_context;
+
+	return amdgpu_dm_execute_fused_io(dev, link, commands, count, timeout_us);
+}
+
 static bool execute_synaptics_rc_command(struct drm_dp_aux *aux,
 		bool is_write_cmd,
 		unsigned char cmd,
@@ -591,6 +666,8 @@ static bool execute_synaptics_rc_command(struct drm_dp_aux *aux,
 		// write rc data
 		memmove(rc_data, data, length);
 		ret = drm_dp_dpcd_write(aux, SYNAPTICS_RC_DATA, rc_data, sizeof(rc_data));
+		if (ret < 0)
+			goto err;
 	}
 
 	// write rc offset
@@ -599,20 +676,21 @@ static bool execute_synaptics_rc_command(struct drm_dp_aux *aux,
 	rc_offset[2] = (unsigned char) (offset >> 16) & 0xFF;
 	rc_offset[3] = (unsigned char) (offset >> 24) & 0xFF;
 	ret = drm_dp_dpcd_write(aux, SYNAPTICS_RC_OFFSET, rc_offset, sizeof(rc_offset));
+	if (ret < 0)
+		goto err;
 
 	// write rc length
 	rc_length[0] = (unsigned char) length & 0xFF;
 	rc_length[1] = (unsigned char) (length >> 8) & 0xFF;
 	ret = drm_dp_dpcd_write(aux, SYNAPTICS_RC_LENGTH, rc_length, sizeof(rc_length));
+	if (ret < 0)
+		goto err;
 
 	// write rc cmd
 	rc_cmd = cmd | 0x80;
 	ret = drm_dp_dpcd_write(aux, SYNAPTICS_RC_COMMAND, &rc_cmd, sizeof(rc_cmd));
-
-	if (ret < 0) {
-		DRM_ERROR("	execute_synaptics_rc_command - write cmd ..., err = %d\n", ret);
-		return false;
-	}
+	if (ret < 0)
+		goto err;
 
 	// poll until active is 0
 	for (i = 0; i < 10; i++) {
@@ -632,16 +710,20 @@ static bool execute_synaptics_rc_command(struct drm_dp_aux *aux,
 		drm_dp_dpcd_read(aux, SYNAPTICS_RC_DATA, data, length);
 	}
 
-	DC_LOG_DC("	execute_synaptics_rc_command - success = %d\n", success);
+	drm_dbg_dp(aux->drm_dev, "success = %d\n", success);
 
 	return success;
+
+err:
+	DRM_ERROR("%s: write cmd ..., err = %d\n",  __func__, ret);
+	return false;
 }
 
 static void apply_synaptics_fifo_reset_wa(struct drm_dp_aux *aux)
 {
 	unsigned char data[16] = {0};
 
-	DC_LOG_DC("Start apply_synaptics_fifo_reset_wa\n");
+	drm_dbg_dp(aux->drm_dev, "Start\n");
 
 	// Step 2
 	data[0] = 'P';
@@ -687,7 +769,6 @@ static void apply_synaptics_fifo_reset_wa(struct drm_dp_aux *aux)
 		return;
 
 	data[0] |= (1 << 1); // set bit 1 to 1
-		return;
 
 	if (!execute_synaptics_rc_command(aux, false, 0x31, 4, 0x221198, data))
 		return;
@@ -700,9 +781,12 @@ static void apply_synaptics_fifo_reset_wa(struct drm_dp_aux *aux)
 	if (!execute_synaptics_rc_command(aux, true, 0x02, 0, 0, NULL))
 		return;
 
-	DC_LOG_DC("Done apply_synaptics_fifo_reset_wa\n");
+	drm_dbg_dp(aux->drm_dev, "Done\n");
 }
 
+/* MST Dock */
+static const uint8_t SYNAPTICS_DEVICE_ID[] = "SYNA";
+
 static uint8_t write_dsc_enable_synaptics_non_virtual_dpcd_mst(
 		struct drm_dp_aux *aux,
 		const struct dc_stream_state *stream,
@@ -710,7 +794,8 @@ static uint8_t write_dsc_enable_synaptics_non_virtual_dpcd_mst(
 {
 	uint8_t ret = 0;
 
-	DC_LOG_DC("Configure DSC to non-virtual dpcd synaptics\n");
+	drm_dbg_dp(aux->drm_dev,
+		   "MST_DSC Configure DSC to non-virtual dpcd synaptics\n");
 
 	if (enable) {
 		/* When DSC is enabled on previous boot and reboot with the hub,
@@ -723,7 +808,7 @@ static uint8_t write_dsc_enable_synaptics_non_virtual_dpcd_mst(
 			apply_synaptics_fifo_reset_wa(aux);
 
 		ret = drm_dp_dpcd_write(aux, DP_DSC_ENABLE, &enable, 1);
-		DRM_INFO("Send DSC enable to synaptics\n");
+		DRM_INFO("MST_DSC Send DSC enable to synaptics\n");
 
 	} else {
 		/* Synaptics hub not support virtual dpcd,
@@ -732,7 +817,7 @@ static uint8_t write_dsc_enable_synaptics_non_virtual_dpcd_mst(
 		 */
 		if (!stream->link->link_status.link_active) {
 			ret = drm_dp_dpcd_write(aux, DP_DSC_ENABLE, &enable, 1);
-			DRM_INFO("Send DSC disable to synaptics\n");
+			DRM_INFO("MST_DSC Send DSC disable to synaptics\n");
 		}
 	}
 
@@ -748,18 +833,15 @@ bool dm_helpers_dp_write_dsc_enable(
 	static const uint8_t DSC_DECODING = 0x01;
 	static const uint8_t DSC_PASSTHROUGH = 0x02;
 
-	struct amdgpu_dm_connector *aconnector;
+	struct amdgpu_dm_connector *aconnector =
+		(struct amdgpu_dm_connector *)stream->dm_stream_context;
+	struct drm_device *dev = aconnector->base.dev;
 	struct drm_dp_mst_port *port;
 	uint8_t enable_dsc = enable ? DSC_DECODING : DSC_DISABLE;
 	uint8_t enable_passthrough = enable ? DSC_PASSTHROUGH : DSC_DISABLE;
 	uint8_t ret = 0;
 
-	if (!stream)
-		return false;
-
 	if (stream->signal == SIGNAL_TYPE_DISPLAY_PORT_MST) {
-		aconnector = (struct amdgpu_dm_connector *)stream->dm_stream_context;
-
 		if (!aconnector->dsc_aux)
 			return false;
 
@@ -776,30 +858,34 @@ bool dm_helpers_dp_write_dsc_enable(
 				ret = drm_dp_dpcd_write(port->passthrough_aux,
 							DP_DSC_ENABLE,
 							&enable_passthrough, 1);
-				DC_LOG_DC("Sent DSC pass-through enable to virtual dpcd port, ret = %u\n",
-					  ret);
+				drm_dbg_dp(dev,
+					   "MST_DSC Sent DSC pass-through enable to virtual dpcd port, ret = %u\n",
+					   ret);
 			}
 
 			ret = drm_dp_dpcd_write(aconnector->dsc_aux,
 						DP_DSC_ENABLE, &enable_dsc, 1);
-			DC_LOG_DC("Sent DSC decoding enable to %s port, ret = %u\n",
-				  (port->passthrough_aux) ? "remote RX" :
-				  "virtual dpcd",
-				  ret);
+			drm_dbg_dp(dev,
+				   "MST_DSC Sent DSC decoding enable to %s port, ret = %u\n",
+				   (port->passthrough_aux) ? "remote RX" :
+				   "virtual dpcd",
+				   ret);
 		} else {
 			ret = drm_dp_dpcd_write(aconnector->dsc_aux,
 						DP_DSC_ENABLE, &enable_dsc, 1);
-			DC_LOG_DC("Sent DSC decoding disable to %s port, ret = %u\n",
-				  (port->passthrough_aux) ? "remote RX" :
-				  "virtual dpcd",
-				  ret);
+			drm_dbg_dp(dev,
+				   "MST_DSC Sent DSC decoding disable to %s port, ret = %u\n",
+				   (port->passthrough_aux) ? "remote RX" :
+				   "virtual dpcd",
+				   ret);
 
 			if (port->passthrough_aux) {
 				ret = drm_dp_dpcd_write(port->passthrough_aux,
 							DP_DSC_ENABLE,
 							&enable_passthrough, 1);
-				DC_LOG_DC("Sent DSC pass-through disable to virtual dpcd port, ret = %u\n",
-					  ret);
+				drm_dbg_dp(dev,
+					   "MST_DSC Sent DSC pass-through disable to virtual dpcd port, ret = %u\n",
+					   ret);
 			}
 		}
 	}
@@ -807,16 +893,26 @@ bool dm_helpers_dp_write_dsc_enable(
 	if (stream->signal == SIGNAL_TYPE_DISPLAY_PORT || stream->signal == SIGNAL_TYPE_EDP) {
 		if (stream->sink->link->dpcd_caps.dongle_type == DISPLAY_DONGLE_NONE) {
 			ret = dm_helpers_dp_write_dpcd(ctx, stream->link, DP_DSC_ENABLE, &enable_dsc, 1);
-			DC_LOG_DC("Send DSC %s to SST RX\n", enable_dsc ? "enable" : "disable");
+			drm_dbg_dp(dev,
+				   "SST_DSC Send DSC %s to SST RX\n",
+				   enable_dsc ? "enable" : "disable");
 		} else if (stream->sink->link->dpcd_caps.dongle_type == DISPLAY_DONGLE_DP_HDMI_CONVERTER) {
 			ret = dm_helpers_dp_write_dpcd(ctx, stream->link, DP_DSC_ENABLE, &enable_dsc, 1);
-			DC_LOG_DC("Send DSC %s to DP-HDMI PCON\n", enable_dsc ? "enable" : "disable");
+			drm_dbg_dp(dev,
+				   "SST_DSC Send DSC %s to DP-HDMI PCON\n",
+				   enable_dsc ? "enable" : "disable");
 		}
 	}
 
 	return ret;
 }
 
+bool dm_helpers_dp_write_hblank_reduction(struct dc_context *ctx, const struct dc_stream_state *stream)
+{
+	// TODO
+	return false;
+}
+
 bool dm_helpers_is_dp_sink_present(struct dc_link *link)
 {
 	bool dp_sink_present;
@@ -833,6 +929,67 @@ bool dm_helpers_is_dp_sink_present(struct dc_link *link)
 	return dp_sink_present;
 }
 
+static int
+dm_helpers_probe_acpi_edid(void *data, u8 *buf, unsigned int block, size_t len)
+{
+	struct drm_connector *connector = data;
+	struct acpi_device *acpidev = ACPI_COMPANION(connector->dev->dev);
+	unsigned short start = block * EDID_LENGTH;
+	struct edid *edid;
+	int r;
+
+	if (!acpidev)
+		return -ENODEV;
+
+	/* fetch the entire edid from BIOS */
+	r = acpi_video_get_edid(acpidev, ACPI_VIDEO_DISPLAY_LCD, -1, (void *)&edid);
+	if (r < 0) {
+		drm_dbg(connector->dev, "Failed to get EDID from ACPI: %d\n", r);
+		return r;
+	}
+	if (len > r || start > r || start + len > r) {
+		r = -EINVAL;
+		goto cleanup;
+	}
+
+	/* sanity check */
+	if (edid->revision < 4 || !(edid->input & DRM_EDID_INPUT_DIGITAL) ||
+	    (edid->input & DRM_EDID_DIGITAL_TYPE_MASK) == DRM_EDID_DIGITAL_TYPE_UNDEF) {
+		r = -EINVAL;
+		goto cleanup;
+	}
+
+	memcpy(buf, (void *)edid + start, len);
+	r = 0;
+
+cleanup:
+	kfree(edid);
+
+	return r;
+}
+
+static const struct drm_edid *
+dm_helpers_read_acpi_edid(struct amdgpu_dm_connector *aconnector)
+{
+	struct drm_connector *connector = &aconnector->base;
+
+	if (amdgpu_dc_debug_mask & DC_DISABLE_ACPI_EDID)
+		return NULL;
+
+	switch (connector->connector_type) {
+	case DRM_MODE_CONNECTOR_LVDS:
+	case DRM_MODE_CONNECTOR_eDP:
+		break;
+	default:
+		return NULL;
+	}
+
+	if (connector->force == DRM_FORCE_OFF)
+		return NULL;
+
+	return drm_edid_read_custom(connector, dm_helpers_probe_acpi_edid, connector);
+}
+
 enum dc_edid_status dm_helpers_read_local_edid(
 		struct dc_context *ctx,
 		struct dc_link *link,
@@ -841,9 +998,10 @@ enum dc_edid_status dm_helpers_read_local_edid(
 	struct amdgpu_dm_connector *aconnector = link->priv;
 	struct drm_connector *connector = &aconnector->base;
 	struct i2c_adapter *ddc;
-	int retry = 3;
-	enum dc_edid_status edid_status;
-	struct edid *edid;
+	int retry = 25;
+	enum dc_edid_status edid_status = EDID_NO_RESPONSE;
+	const struct drm_edid *drm_edid;
+	const struct edid *edid;
 
 	if (link->aux_mode)
 		ddc = &aconnector->dm_dp_aux.aux.ddc;
@@ -854,42 +1012,75 @@ enum dc_edid_status dm_helpers_read_local_edid(
 	 * do check sum and retry to make sure read correct edid.
 	 */
 	do {
-
-		edid = drm_get_edid(&aconnector->base, ddc);
+		drm_edid = dm_helpers_read_acpi_edid(aconnector);
+		if (drm_edid)
+			drm_info(connector->dev, "Using ACPI provided EDID for %s\n", connector->name);
+		else
+			drm_edid = drm_edid_read_ddc(connector, ddc);
+		drm_edid_connector_update(connector, drm_edid);
 
 		/* DP Compliance Test 4.2.2.6 */
 		if (link->aux_mode && connector->edid_corrupt)
 			drm_dp_send_real_edid_checksum(&aconnector->dm_dp_aux.aux, connector->real_edid_checksum);
 
-		if (!edid && connector->edid_corrupt) {
+		if (!drm_edid && connector->edid_corrupt) {
 			connector->edid_corrupt = false;
 			return EDID_BAD_CHECKSUM;
 		}
 
-		if (!edid)
-			return EDID_NO_RESPONSE;
+		if (!drm_edid)
+			continue;
+
+		edid = drm_edid_raw(drm_edid); // FIXME: Get rid of drm_edid_raw()
+		if (!edid ||
+		    edid->extensions >= sizeof(sink->dc_edid.raw_edid) / EDID_LENGTH)
+			return EDID_BAD_INPUT;
 
 		sink->dc_edid.length = EDID_LENGTH * (edid->extensions + 1);
 		memmove(sink->dc_edid.raw_edid, (uint8_t *)edid, sink->dc_edid.length);
 
 		/* We don't need the original edid anymore */
-		kfree(edid);
+		drm_edid_free(drm_edid);
 
 		edid_status = dm_helpers_parse_edid_caps(
 						link,
 						&sink->dc_edid,
 						&sink->edid_caps);
 
-	} while (edid_status == EDID_BAD_CHECKSUM && --retry > 0);
+	} while ((edid_status == EDID_BAD_CHECKSUM || edid_status == EDID_NO_RESPONSE) && --retry > 0);
 
 	if (edid_status != EDID_OK)
 		DRM_ERROR("EDID err: %d, on connector: %s",
 				edid_status,
 				aconnector->base.name);
+	if (link->aux_mode) {
+		union test_request test_request = {0};
+		union test_response test_response = {0};
 
-	/* DP Compliance Test 4.2.2.3 */
-	if (link->aux_mode)
-		drm_dp_send_real_edid_checksum(&aconnector->dm_dp_aux.aux, sink->dc_edid.raw_edid[sink->dc_edid.length-1]);
+		dm_helpers_dp_read_dpcd(ctx,
+					link,
+					DP_TEST_REQUEST,
+					&test_request.raw,
+					sizeof(union test_request));
+
+		if (!test_request.bits.EDID_READ)
+			return edid_status;
+
+		test_response.bits.EDID_CHECKSUM_WRITE = 1;
+
+		dm_helpers_dp_write_dpcd(ctx,
+					link,
+					DP_TEST_EDID_CHECKSUM,
+					&sink->dc_edid.raw_edid[sink->dc_edid.length-1],
+					1);
+
+		dm_helpers_dp_write_dpcd(ctx,
+					link,
+					DP_TEST_RESPONSE,
+					&test_response.raw,
+					sizeof(test_response));
+
+	}
 
 	return edid_status;
 }
@@ -899,6 +1090,11 @@ int dm_helper_dmub_aux_transfer_sync(
 		struct aux_payload *payload,
 		enum aux_return_code_type *operation_result)
 {
+	if (!link->hpd_status) {
+		*operation_result = AUX_RET_ERROR_HPD_DISCON;
+		return -1;
+	}
+
 	return amdgpu_dm_process_dmub_aux_transfer_sync(ctx, link->link_index, payload,
 			operation_result);
 }
@@ -946,9 +1142,8 @@ void dm_helpers_override_panel_settings(
 	struct dc_panel_config *panel_config)
 {
 	// Feature DSC
-	if (amdgpu_dc_debug_mask & DC_DISABLE_DSC) {
+	if (amdgpu_dc_debug_mask & DC_DISABLE_DSC)
 		panel_config->dsc.disable_dsc_edp = true;
-	}
 }
 
 void *dm_helpers_allocate_gpu_mem(
@@ -958,30 +1153,8 @@ void *dm_helpers_allocate_gpu_mem(
 		long long *addr)
 {
 	struct amdgpu_device *adev = ctx->driver_context;
-	struct dal_allocation *da;
-	u32 domain = (type == DC_MEM_ALLOC_TYPE_GART) ?
-		AMDGPU_GEM_DOMAIN_GTT : AMDGPU_GEM_DOMAIN_VRAM;
-	int ret;
-
-	da = kzalloc(sizeof(struct dal_allocation), GFP_KERNEL);
-	if (!da)
-		return NULL;
-
-	ret = amdgpu_bo_create_kernel(adev, size, PAGE_SIZE,
-				      domain, &da->bo,
-				      &da->gpu_addr, &da->cpu_ptr);
-
-	*addr = da->gpu_addr;
-
-	if (ret) {
-		kfree(da);
-		return NULL;
-	}
-
-	/* add da to list in dm */
-	list_add(&da->list, &adev->dm.da_list);
 
-	return da->cpu_ptr;
+	return dm_allocate_gpu_mem(adev, type, size, addr);
 }
 
 void dm_helpers_free_gpu_mem(
@@ -990,17 +1163,8 @@ void dm_helpers_free_gpu_mem(
 		void *pvMem)
 {
 	struct amdgpu_device *adev = ctx->driver_context;
-	struct dal_allocation *da;
-
-	/* walk the da list in DM */
-	list_for_each_entry(da, &adev->dm.da_list, list) {
-		if (pvMem == da->cpu_ptr) {
-			amdgpu_bo_free_kernel(&da->bo, &da->gpu_addr, &da->cpu_ptr);
-			list_del(&da->list);
-			kfree(da);
-			break;
-		}
-	}
+
+	dm_free_gpu_mem(adev, type, pvMem);
 }
 
 bool dm_helpers_dmub_outbox_interrupt_control(struct dc_context *ctx, bool enable)
@@ -1056,6 +1220,9 @@ bool dm_helpers_dp_handle_test_pattern_request(
 	struct pipe_ctx *pipes = link->dc->current_state->res_ctx.pipe_ctx;
 	struct pipe_ctx *pipe_ctx = NULL;
 	struct amdgpu_dm_connector *aconnector = link->priv;
+	struct drm_device *dev = aconnector->base.dev;
+	struct dc_state *dc_state = ctx->dc->current_state;
+	struct clk_mgr *clk_mgr = ctx->dc->clk_mgr;
 	int i;
 
 	for (i = 0; i < MAX_PIPES; i++) {
@@ -1133,12 +1300,12 @@ bool dm_helpers_dp_handle_test_pattern_request(
 		&& pipe_ctx->stream->timing.display_color_depth != requestColorDepth)
 		|| (requestPixelEncoding != PIXEL_ENCODING_UNDEFINED
 		&& pipe_ctx->stream->timing.pixel_encoding != requestPixelEncoding)) {
-		DC_LOG_DEBUG("%s: original bpc %d pix encoding %d, changing to %d  %d\n",
-				__func__,
-				pipe_ctx->stream->timing.display_color_depth,
-				pipe_ctx->stream->timing.pixel_encoding,
-				requestColorDepth,
-				requestPixelEncoding);
+		drm_dbg(dev,
+			"original bpc %d pix encoding %d, changing to %d  %d\n",
+			pipe_ctx->stream->timing.display_color_depth,
+			pipe_ctx->stream->timing.pixel_encoding,
+			requestColorDepth,
+			requestPixelEncoding);
 		pipe_ctx->stream->timing.display_color_depth = requestColorDepth;
 		pipe_ctx->stream->timing.pixel_encoding = requestPixelEncoding;
 
@@ -1149,10 +1316,23 @@ bool dm_helpers_dp_handle_test_pattern_request(
 		if (aconnector->timing_requested)
 			*aconnector->timing_requested = pipe_ctx->stream->timing;
 		else
-			DC_LOG_ERROR("%s: timing storage failed\n", __func__);
+			drm_err(dev, "timing storage failed\n");
 
 	}
 
+	pipe_ctx->stream->test_pattern.type = test_pattern;
+	pipe_ctx->stream->test_pattern.color_space = test_pattern_color_space;
+
+	/* Temp W/A for compliance test failure */
+	dc_state->bw_ctx.bw.dcn.clk.p_state_change_support = false;
+	dc_state->bw_ctx.bw.dcn.clk.dramclk_khz = clk_mgr->dc_mode_softmax_enabled ?
+		clk_mgr->bw_params->dc_mode_softmax_memclk : clk_mgr->bw_params->max_memclk_mhz;
+	dc_state->bw_ctx.bw.dcn.clk.idle_dramclk_khz = dc_state->bw_ctx.bw.dcn.clk.dramclk_khz;
+	ctx->dc->clk_mgr->funcs->update_clocks(
+			ctx->dc->clk_mgr,
+			dc_state,
+			false);
+
 	dc_link_dp_set_test_pattern(
 		(struct dc_link *) link,
 		test_pattern,
@@ -1171,7 +1351,13 @@ void dm_set_phyd32clk(struct dc_context *ctx, int freq_khz)
 
 void dm_helpers_enable_periodic_detection(struct dc_context *ctx, bool enable)
 {
-	/* TODO: add periodic detection implementation */
+	struct amdgpu_device *adev = ctx->driver_context;
+
+	if (adev->dm.idle_workqueue) {
+		adev->dm.idle_workqueue->enable = enable;
+		if (enable && !adev->dm.idle_workqueue->running && amdgpu_dm_is_headless(adev))
+			schedule_work(&adev->dm.idle_workqueue->work);
+	}
 }
 
 void dm_helpers_dp_mst_update_branch_bandwidth(
@@ -1216,3 +1402,15 @@ enum adaptive_sync_type dm_get_adaptive_sync_support_type(struct dc_link *link)
 
 	return as_type;
 }
+
+bool dm_helpers_is_fullscreen(struct dc_context *ctx, struct dc_stream_state *stream)
+{
+	// TODO
+	return false;
+}
+
+bool dm_helpers_is_hdr_on(struct dc_context *ctx, struct dc_stream_state *stream)
+{
+	// TODO
+	return false;
+}
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.c
index 19f543ba7205..0a2a3f233a0e 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.c
@@ -1,3 +1,4 @@
+// SPDX-License-Identifier: MIT
 /*
  * Copyright 2015 Advanced Micro Devices, Inc.
  *
@@ -120,7 +121,8 @@ static void dm_irq_work_func(struct work_struct *work)
 
 	/* Call a DAL subcomponent which registered for interrupt notification
 	 * at INTERRUPT_LOW_IRQ_CONTEXT.
-	 * (The most common use is HPD interrupt) */
+	 * (The most common use is HPD interrupt)
+	 */
 }
 
 /*
@@ -172,7 +174,8 @@ static struct list_head *remove_irq_handler(struct amdgpu_device *adev,
 
 	if (handler_removed == false) {
 		/* Not necessarily an error - caller may not
-		 * know the context. */
+		 * know the context.
+		 */
 		return NULL;
 	}
 
@@ -261,7 +264,7 @@ validate_irq_registration_params(struct dc_interrupt_params *int_params,
 static bool validate_irq_unregistration_params(enum dc_irq_source irq_source,
 					       irq_handler_idx handler_idx)
 {
-	if (DAL_INVALID_IRQ_HANDLER_IDX == handler_idx) {
+	if (handler_idx == DAL_INVALID_IRQ_HANDLER_IDX) {
 		DRM_ERROR("DM_IRQ: invalid handler_idx==NULL!\n");
 		return false;
 	}
@@ -343,7 +346,8 @@ void *amdgpu_dm_irq_register_interrupt(struct amdgpu_device *adev,
 	/* This pointer will be stored by code which requested interrupt
 	 * registration.
 	 * The same pointer will be needed in order to unregister the
-	 * interrupt. */
+	 * interrupt.
+	 */
 
 	DRM_DEBUG_KMS(
 		"DM_IRQ: added irq handler: %p for: dal_src=%d, irq context=%d\n",
@@ -390,7 +394,8 @@ void amdgpu_dm_irq_unregister_interrupt(struct amdgpu_device *adev,
 
 	if (handler_list == NULL) {
 		/* If we got here, it means we searched all irq contexts
-		 * for this irq source, but the handler was not found. */
+		 * for this irq source, but the handler was not found.
+		 */
 		DRM_ERROR(
 		"DM_IRQ: failed to find irq handler:%p for irq_source:%d!\n",
 			ih, irq_source);
@@ -450,7 +455,8 @@ void amdgpu_dm_irq_fini(struct amdgpu_device *adev)
 		DM_IRQ_TABLE_LOCK(adev, irq_table_flags);
 		/* The handler was removed from the table,
 		 * it means it is safe to flush all the 'work'
-		 * (because no code can schedule a new one). */
+		 * (because no code can schedule a new one).
+		 */
 		lh = &adev->dm.irq_handler_list_low_tab[src];
 		DM_IRQ_TABLE_UNLOCK(adev, irq_table_flags);
 
@@ -468,8 +474,9 @@ void amdgpu_dm_irq_fini(struct amdgpu_device *adev)
 	unregister_all_irq_handlers(adev);
 }
 
-int amdgpu_dm_irq_suspend(struct amdgpu_device *adev)
+void amdgpu_dm_irq_suspend(struct amdgpu_device *adev)
 {
+	struct drm_device *dev = adev_to_drm(adev);
 	int src;
 	struct list_head *hnd_list_h;
 	struct list_head *hnd_list_l;
@@ -494,7 +501,7 @@ int amdgpu_dm_irq_suspend(struct amdgpu_device *adev)
 		DM_IRQ_TABLE_UNLOCK(adev, irq_table_flags);
 
 		if (!list_empty(hnd_list_l)) {
-			list_for_each_safe (entry, tmp, hnd_list_l) {
+			list_for_each_safe(entry, tmp, hnd_list_l) {
 				handler = list_entry(
 					entry,
 					struct amdgpu_dm_irq_handler_data,
@@ -506,10 +513,12 @@ int amdgpu_dm_irq_suspend(struct amdgpu_device *adev)
 	}
 
 	DM_IRQ_TABLE_UNLOCK(adev, irq_table_flags);
-	return 0;
+
+	if (dev->mode_config.poll_enabled)
+		drm_kms_helper_poll_disable(dev);
 }
 
-int amdgpu_dm_irq_resume_early(struct amdgpu_device *adev)
+void amdgpu_dm_irq_resume_early(struct amdgpu_device *adev)
 {
 	int src;
 	struct list_head *hnd_list_h, *hnd_list_l;
@@ -517,7 +526,7 @@ int amdgpu_dm_irq_resume_early(struct amdgpu_device *adev)
 
 	DM_IRQ_TABLE_LOCK(adev, irq_table_flags);
 
-	DRM_DEBUG_KMS("DM_IRQ: early resume\n");
+	drm_dbg(adev_to_drm(adev), "DM_IRQ: early resume\n");
 
 	/* re-enable short pulse interrupts HW interrupt */
 	for (src = DC_IRQ_SOURCE_HPD1RX; src <= DC_IRQ_SOURCE_HPD6RX; src++) {
@@ -528,19 +537,18 @@ int amdgpu_dm_irq_resume_early(struct amdgpu_device *adev)
 	}
 
 	DM_IRQ_TABLE_UNLOCK(adev, irq_table_flags);
-
-	return 0;
 }
 
-int amdgpu_dm_irq_resume_late(struct amdgpu_device *adev)
+void amdgpu_dm_irq_resume_late(struct amdgpu_device *adev)
 {
+	struct drm_device *dev = adev_to_drm(adev);
 	int src;
 	struct list_head *hnd_list_h, *hnd_list_l;
 	unsigned long irq_table_flags;
 
 	DM_IRQ_TABLE_LOCK(adev, irq_table_flags);
 
-	DRM_DEBUG_KMS("DM_IRQ: resume\n");
+	drm_dbg(adev_to_drm(adev), "DM_IRQ: resume\n");
 
 	/**
 	 * Renable HW interrupt  for HPD and only since FLIP and VBLANK
@@ -554,7 +562,9 @@ int amdgpu_dm_irq_resume_late(struct amdgpu_device *adev)
 	}
 
 	DM_IRQ_TABLE_UNLOCK(adev, irq_table_flags);
-	return 0;
+
+	if (dev->mode_config.poll_enabled)
+		drm_kms_helper_poll_enable(dev);
 }
 
 /*
@@ -571,7 +581,7 @@ static void amdgpu_dm_irq_schedule_work(struct amdgpu_device *adev,
 	if (list_empty(handler_list))
 		return;
 
-	list_for_each_entry (handler_data, handler_list, list) {
+	list_for_each_entry(handler_data, handler_list, list) {
 		if (queue_work(system_highpri_wq, &handler_data->work)) {
 			work_queued = true;
 			break;
@@ -627,7 +637,8 @@ static void amdgpu_dm_irq_immediate_work(struct amdgpu_device *adev,
 			    &adev->dm.irq_handler_list_high_tab[irq_source],
 			    list) {
 		/* Call a subcomponent which registered for immediate
-		 * interrupt notification */
+		 * interrupt notification
+		 */
 		handler_data->handler(handler_data->handler_arg);
 	}
 
@@ -664,7 +675,7 @@ static int amdgpu_dm_irq_handler(struct amdgpu_device *adev,
 	return 0;
 }
 
-static enum dc_irq_source amdgpu_dm_hpd_to_dal_irq_source(unsigned type)
+static enum dc_irq_source amdgpu_dm_hpd_to_dal_irq_source(unsigned int type)
 {
 	switch (type) {
 	case AMDGPU_HPD_1:
@@ -686,7 +697,7 @@ static enum dc_irq_source amdgpu_dm_hpd_to_dal_irq_source(unsigned type)
 
 static int amdgpu_dm_set_hpd_irq_state(struct amdgpu_device *adev,
 				       struct amdgpu_irq_src *source,
-				       unsigned type,
+				       unsigned int type,
 				       enum amdgpu_interrupt_state state)
 {
 	enum dc_irq_source src = amdgpu_dm_hpd_to_dal_irq_source(type);
@@ -698,14 +709,14 @@ static int amdgpu_dm_set_hpd_irq_state(struct amdgpu_device *adev,
 
 static inline int dm_irq_state(struct amdgpu_device *adev,
 			       struct amdgpu_irq_src *source,
-			       unsigned crtc_id,
+			       unsigned int crtc_id,
 			       enum amdgpu_interrupt_state state,
 			       const enum irq_type dal_irq_type,
 			       const char *func)
 {
 	bool st;
 	enum dc_irq_source irq_source;
-
+	struct dc *dc = adev->dm.dc;
 	struct amdgpu_crtc *acrtc = adev->mode_info.crtcs[crtc_id];
 
 	if (!acrtc) {
@@ -723,13 +734,16 @@ static inline int dm_irq_state(struct amdgpu_device *adev,
 
 	st = (state == AMDGPU_IRQ_STATE_ENABLE);
 
+	if (dc && dc->caps.ips_support && dc->idle_optimizations_allowed)
+		dc_allow_idle_optimizations(dc, false);
+
 	dc_interrupt_set(adev->dm.dc, irq_source, st);
 	return 0;
 }
 
 static int amdgpu_dm_set_pflip_irq_state(struct amdgpu_device *adev,
 					 struct amdgpu_irq_src *source,
-					 unsigned crtc_id,
+					 unsigned int crtc_id,
 					 enum amdgpu_interrupt_state state)
 {
 	return dm_irq_state(
@@ -743,7 +757,7 @@ static int amdgpu_dm_set_pflip_irq_state(struct amdgpu_device *adev,
 
 static int amdgpu_dm_set_crtc_irq_state(struct amdgpu_device *adev,
 					struct amdgpu_irq_src *source,
-					unsigned crtc_id,
+					unsigned int crtc_id,
 					enum amdgpu_interrupt_state state)
 {
 	return dm_irq_state(
@@ -885,27 +899,68 @@ void amdgpu_dm_hpd_init(struct amdgpu_device *adev)
 	struct drm_device *dev = adev_to_drm(adev);
 	struct drm_connector *connector;
 	struct drm_connector_list_iter iter;
+	int irq_type;
+	int i;
+	bool use_polling = false;
+
+	/* First, clear all hpd and hpdrx interrupts */
+	for (i = DC_IRQ_SOURCE_HPD1; i <= DC_IRQ_SOURCE_HPD6RX; i++) {
+		if (!dc_interrupt_set(adev->dm.dc, i, false))
+			drm_err(dev, "Failed to clear hpd(rx) source=%d on init\n",
+				i);
+	}
 
 	drm_connector_list_iter_begin(dev, &iter);
 	drm_for_each_connector_iter(connector, &iter) {
-		struct amdgpu_dm_connector *amdgpu_dm_connector =
-				to_amdgpu_dm_connector(connector);
+		struct amdgpu_dm_connector *amdgpu_dm_connector;
+		const struct dc_link *dc_link;
 
-		const struct dc_link *dc_link = amdgpu_dm_connector->dc_link;
+		use_polling |= connector->polled != DRM_CONNECTOR_POLL_HPD;
 
-		if (DC_IRQ_SOURCE_INVALID != dc_link->irq_source_hpd) {
-			dc_interrupt_set(adev->dm.dc,
-					dc_link->irq_source_hpd,
-					true);
+		if (connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+			continue;
+
+		amdgpu_dm_connector = to_amdgpu_dm_connector(connector);
+
+		dc_link = amdgpu_dm_connector->dc_link;
+
+		/*
+		 * Get a base driver irq reference for hpd ints for the lifetime
+		 * of dm. Note that only hpd interrupt types are registered with
+		 * base driver; hpd_rx types aren't. IOW, amdgpu_irq_get/put on
+		 * hpd_rx isn't available. DM currently controls hpd_rx
+		 * explicitly with dc_interrupt_set()
+		 */
+		if (dc_link->irq_source_hpd != DC_IRQ_SOURCE_INVALID) {
+			irq_type = dc_link->irq_source_hpd - DC_IRQ_SOURCE_HPD1;
+			/*
+			 * TODO: There's a mismatch between mode_info.num_hpd
+			 * and what bios reports as the # of connectors with hpd
+			 * sources. Since the # of hpd source types registered
+			 * with base driver == mode_info.num_hpd, we have to
+			 * fallback to dc_interrupt_set for the remaining types.
+			 */
+			if (irq_type < adev->mode_info.num_hpd) {
+				if (amdgpu_irq_get(adev, &adev->hpd_irq, irq_type))
+					drm_err(dev, "DM_IRQ: Failed get HPD for source=%d)!\n",
+						dc_link->irq_source_hpd);
+			} else {
+				dc_interrupt_set(adev->dm.dc,
+						 dc_link->irq_source_hpd,
+						 true);
+			}
 		}
 
-		if (DC_IRQ_SOURCE_INVALID != dc_link->irq_source_hpd_rx) {
+		if (dc_link->irq_source_hpd_rx != DC_IRQ_SOURCE_INVALID) {
 			dc_interrupt_set(adev->dm.dc,
 					dc_link->irq_source_hpd_rx,
 					true);
 		}
 	}
 	drm_connector_list_iter_end(&iter);
+
+	if (use_polling)
+		drm_kms_helper_poll_init(dev);
 }
 
 /**
@@ -921,24 +976,42 @@ void amdgpu_dm_hpd_fini(struct amdgpu_device *adev)
 	struct drm_device *dev = adev_to_drm(adev);
 	struct drm_connector *connector;
 	struct drm_connector_list_iter iter;
+	int irq_type;
 
 	drm_connector_list_iter_begin(dev, &iter);
 	drm_for_each_connector_iter(connector, &iter) {
-		struct amdgpu_dm_connector *amdgpu_dm_connector =
-				to_amdgpu_dm_connector(connector);
-		const struct dc_link *dc_link = amdgpu_dm_connector->dc_link;
+		struct amdgpu_dm_connector *amdgpu_dm_connector;
+		const struct dc_link *dc_link;
 
-		if (DC_IRQ_SOURCE_INVALID != dc_link->irq_source_hpd) {
-			dc_interrupt_set(adev->dm.dc,
-					dc_link->irq_source_hpd,
-					false);
+		if (connector->connector_type == DRM_MODE_CONNECTOR_WRITEBACK)
+			continue;
+
+		amdgpu_dm_connector = to_amdgpu_dm_connector(connector);
+		dc_link = amdgpu_dm_connector->dc_link;
+
+		if (dc_link->irq_source_hpd != DC_IRQ_SOURCE_INVALID) {
+			irq_type = dc_link->irq_source_hpd - DC_IRQ_SOURCE_HPD1;
+
+			/* TODO: See same TODO in amdgpu_dm_hpd_init() */
+			if (irq_type < adev->mode_info.num_hpd) {
+				if (amdgpu_irq_put(adev, &adev->hpd_irq, irq_type))
+					drm_err(dev, "DM_IRQ: Failed put HPD for source=%d!\n",
+						dc_link->irq_source_hpd);
+			} else {
+				dc_interrupt_set(adev->dm.dc,
+						 dc_link->irq_source_hpd,
+						 false);
+			}
 		}
 
-		if (DC_IRQ_SOURCE_INVALID != dc_link->irq_source_hpd_rx) {
+		if (dc_link->irq_source_hpd_rx != DC_IRQ_SOURCE_INVALID) {
 			dc_interrupt_set(adev->dm.dc,
 					dc_link->irq_source_hpd_rx,
 					false);
 		}
 	}
 	drm_connector_list_iter_end(&iter);
+
+	if (dev->mode_config.poll_enabled)
+		drm_kms_helper_poll_fini(dev);
 }
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.h
index 2349238a626b..4f6b58f4f90d 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.h
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq.h
@@ -1,3 +1,4 @@
+/* SPDX-License-Identifier: MIT */
 /*
  * Copyright 2015 Advanced Micro Devices, Inc.
  *
@@ -90,14 +91,14 @@ void amdgpu_dm_hpd_fini(struct amdgpu_device *adev);
  * amdgpu_dm_irq_suspend - disable ASIC interrupt during suspend.
  *
  */
-int amdgpu_dm_irq_suspend(struct amdgpu_device *adev);
+void amdgpu_dm_irq_suspend(struct amdgpu_device *adev);
 
 /**
  * amdgpu_dm_irq_resume_early - enable HPDRX ASIC interrupts during resume.
  * amdgpu_dm_irq_resume - enable ASIC interrupt during resume.
  *
  */
-int amdgpu_dm_irq_resume_early(struct amdgpu_device *adev);
-int amdgpu_dm_irq_resume_late(struct amdgpu_device *adev);
+void amdgpu_dm_irq_resume_early(struct amdgpu_device *adev);
+void amdgpu_dm_irq_resume_late(struct amdgpu_device *adev);
 
 #endif /* __AMDGPU_DM_IRQ_H__ */
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq_params.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq_params.h
index 5c9303241aeb..3c9995275cbd 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq_params.h
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq_params.h
@@ -1,3 +1,4 @@
+/* SPDX-License-Identifier: MIT */
 /*
  * Copyright 2020 Advanced Micro Devices, Inc.
  *
@@ -33,13 +34,15 @@ struct dm_irq_params {
 	struct mod_vrr_params vrr_params;
 	struct dc_stream_state *stream;
 	int active_planes;
-	bool allow_psr_entry;
+	bool allow_sr_entry;
 	struct mod_freesync_config freesync_config;
 
 #ifdef CONFIG_DEBUG_FS
 	enum amdgpu_dm_pipe_crc_source crc_src;
 #ifdef CONFIG_DRM_AMD_SECURE_DISPLAY
-	struct crc_window_param window_param;
+	struct crc_window_param window_param[MAX_CRC_WINDOW_NUM];
+	/* At least one CRC window is activated or not*/
+	bool crc_window_activated;
 #endif
 #endif
 };
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c
index 994ba426ca66..dbd1da4d85d3 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c
@@ -1,3 +1,4 @@
+// SPDX-License-Identifier: MIT
 /*
  * Copyright 2012-15 Advanced Micro Devices, Inc.
  *
@@ -23,10 +24,13 @@
  *
  */
 
+#include <linux/vmalloc.h>
 #include <drm/display/drm_dp_helper.h>
 #include <drm/display/drm_dp_mst_helper.h>
 #include <drm/drm_atomic.h>
 #include <drm/drm_atomic_helper.h>
+#include <drm/drm_fixed.h>
+#include <drm/drm_edid.h>
 #include "dm_services.h"
 #include "amdgpu.h"
 #include "amdgpu_dm.h"
@@ -44,11 +48,13 @@
 #include "amdgpu_dm_debugfs.h"
 #endif
 
-#include "dc/dcn20/dcn20_resource.h"
-bool is_timing_changed(struct dc_stream_state *cur_stream,
-		       struct dc_stream_state *new_stream);
+#include "dc/resource/dcn20/dcn20_resource.h"
+
 #define PEAK_FACTOR_X1000 1006
 
+/*
+ * This function handles both native AUX and I2C-Over-AUX transactions.
+ */
 static ssize_t dm_dp_aux_transfer(struct drm_dp_aux *aux,
 				  struct drm_dp_aux_msg *msg)
 {
@@ -57,6 +63,7 @@ static ssize_t dm_dp_aux_transfer(struct drm_dp_aux *aux,
 	enum aux_return_code_type operation_result;
 	struct amdgpu_device *adev;
 	struct ddc_service *ddc;
+	uint8_t copy[16];
 
 	if (WARN_ON(msg->size > 16))
 		return -E2BIG;
@@ -72,6 +79,11 @@ static ssize_t dm_dp_aux_transfer(struct drm_dp_aux *aux,
 			(msg->request & DP_AUX_I2C_WRITE_STATUS_UPDATE) != 0;
 	payload.defer_delay = 0;
 
+	if (payload.write) {
+		memcpy(copy, msg->buffer, msg->size);
+		payload.data = copy;
+	}
+
 	result = dc_link_aux_transfer_raw(TO_DM_AUX(aux)->ddc_service, &payload,
 				      &operation_result);
 
@@ -85,15 +97,25 @@ static ssize_t dm_dp_aux_transfer(struct drm_dp_aux *aux,
 	if (adev->dm.aux_hpd_discon_quirk) {
 		if (msg->address == DP_SIDEBAND_MSG_DOWN_REQ_BASE &&
 			operation_result == AUX_RET_ERROR_HPD_DISCON) {
-			result = 0;
+			result = msg->size;
 			operation_result = AUX_RET_SUCCESS;
 		}
 	}
 
-	if (payload.write && result >= 0)
-		result = msg->size;
+	/*
+	 * result equals to 0 includes the cases of AUX_DEFER/I2C_DEFER
+	 */
+	if (payload.write && result >= 0) {
+		if (result) {
+			/*one byte indicating partially written bytes*/
+			drm_dbg_dp(adev_to_drm(adev), "AUX partially written\n");
+			result = payload.data[0];
+		} else if (!payload.reply[0])
+			/*I2C_ACK|AUX_ACK*/
+			result = msg->size;
+	}
 
-	if (result < 0)
+	if (result < 0) {
 		switch (operation_result) {
 		case AUX_RET_SUCCESS:
 			break;
@@ -112,6 +134,13 @@ static ssize_t dm_dp_aux_transfer(struct drm_dp_aux *aux,
 			break;
 		}
 
+		drm_dbg_dp(adev_to_drm(adev), "DP AUX transfer fail:%d\n", operation_result);
+	}
+
+	if (payload.reply[0])
+		drm_dbg_dp(adev_to_drm(adev), "AUX reply command not ACK: 0x%02x.",
+			payload.reply[0]);
+
 	return result;
 }
 
@@ -127,7 +156,7 @@ dm_dp_mst_connector_destroy(struct drm_connector *connector)
 		dc_sink_release(aconnector->dc_sink);
 	}
 
-	kfree(aconnector->edid);
+	drm_edid_free(aconnector->drm_edid);
 
 	drm_connector_cleanup(connector);
 	drm_dp_mst_put_port_malloc(aconnector->mst_output_port);
@@ -153,6 +182,17 @@ amdgpu_dm_mst_connector_late_register(struct drm_connector *connector)
 	return 0;
 }
 
+
+static inline void
+amdgpu_dm_mst_reset_mst_connector_setting(struct amdgpu_dm_connector *aconnector)
+{
+	aconnector->drm_edid = NULL;
+	aconnector->dsc_aux = NULL;
+	aconnector->mst_output_port->passthrough_aux = NULL;
+	aconnector->mst_local_bw = 0;
+	aconnector->vc_full_pbn = 0;
+}
+
 static void
 amdgpu_dm_mst_connector_early_unregister(struct drm_connector *connector)
 {
@@ -174,12 +214,13 @@ amdgpu_dm_mst_connector_early_unregister(struct drm_connector *connector)
 		if (dc_link->sink_count)
 			dc_link_remove_remote_sink(dc_link, dc_sink);
 
-		DC_LOG_MST("DM_MST: remove remote sink 0x%p, %d remaining\n",
-			dc_sink, dc_link->sink_count);
+		drm_dbg_dp(connector->dev,
+			   "DM_MST: remove remote sink 0x%p, %d remaining\n",
+			   dc_sink, dc_link->sink_count);
 
 		dc_sink_release(dc_sink);
 		aconnector->dc_sink = NULL;
-		aconnector->edid = NULL;
+		amdgpu_dm_mst_reset_mst_connector_setting(aconnector);
 	}
 
 	aconnector->mst_status = MST_STATUS_DEFAULT;
@@ -208,6 +249,7 @@ bool needs_dsc_aux_workaround(struct dc_link *link)
 	return false;
 }
 
+#if defined(CONFIG_DRM_AMD_DC_FP)
 static bool is_synaptics_cascaded_panamera(struct dc_link *link, struct drm_dp_mst_port *port)
 {
 	u8 branch_vendor_data[4] = { 0 }; // Vendor data 0x50C ~ 0x50F
@@ -247,7 +289,7 @@ static bool validate_dsc_caps_on_connector(struct amdgpu_dm_connector *aconnecto
 		aconnector->dsc_aux = &aconnector->mst_root->dm_dp_aux.aux;
 
 	/* synaptics cascaded MST hub case */
-	if (!aconnector->dsc_aux && is_synaptics_cascaded_panamera(aconnector->dc_link, port))
+	if (is_synaptics_cascaded_panamera(aconnector->dc_link, port))
 		aconnector->dsc_aux = port->mgr->aux;
 
 	if (!aconnector->dsc_aux)
@@ -267,6 +309,7 @@ static bool validate_dsc_caps_on_connector(struct amdgpu_dm_connector *aconnecto
 
 	return true;
 }
+#endif
 
 static bool retrieve_downstream_port_device(struct amdgpu_dm_connector *aconnector)
 {
@@ -287,6 +330,34 @@ static bool retrieve_downstream_port_device(struct amdgpu_dm_connector *aconnect
 	return true;
 }
 
+static bool retrieve_branch_specific_data(struct amdgpu_dm_connector *aconnector)
+{
+	struct drm_connector *connector = &aconnector->base;
+	struct drm_dp_mst_port *port = aconnector->mst_output_port;
+	struct drm_dp_mst_port *port_parent;
+	struct drm_dp_aux *immediate_upstream_aux;
+	struct drm_dp_desc branch_desc;
+
+	if (!port->parent)
+		return false;
+
+	port_parent = port->parent->port_parent;
+
+	immediate_upstream_aux = port_parent ? &port_parent->aux : port->mgr->aux;
+
+	if (drm_dp_read_desc(immediate_upstream_aux, &branch_desc, true))
+		return false;
+
+	aconnector->branch_ieee_oui = (branch_desc.ident.oui[0] << 16) +
+				      (branch_desc.ident.oui[1] << 8) +
+				      (branch_desc.ident.oui[2]);
+
+	drm_dbg_dp(port->aux.drm_dev, "MST branch oui 0x%x detected at %s\n",
+		   aconnector->branch_ieee_oui, connector->name);
+
+	return true;
+}
+
 static int dm_dp_mst_get_modes(struct drm_connector *connector)
 {
 	struct amdgpu_dm_connector *aconnector = to_amdgpu_dm_connector(connector);
@@ -295,15 +366,18 @@ static int dm_dp_mst_get_modes(struct drm_connector *connector)
 	if (!aconnector)
 		return drm_add_edid_modes(connector, NULL);
 
-	if (!aconnector->edid) {
-		struct edid *edid;
-		edid = drm_dp_mst_get_edid(connector, &aconnector->mst_root->mst_mgr, aconnector->mst_output_port);
+	if (!aconnector->drm_edid) {
+		const struct drm_edid *drm_edid;
 
-		if (!edid) {
+		drm_edid = drm_dp_mst_edid_read(connector,
+						&aconnector->mst_root->mst_mgr,
+						aconnector->mst_output_port);
+
+		if (!drm_edid) {
 			amdgpu_dm_set_mst_status(&aconnector->mst_status,
 			MST_REMOTE_EDID, false);
 
-			drm_connector_update_edid_property(
+			drm_edid_connector_update(
 				&aconnector->base,
 				NULL);
 
@@ -325,8 +399,10 @@ static int dm_dp_mst_get_modes(struct drm_connector *connector)
 					return 0;
 				}
 
-				DC_LOG_MST("DM_MST: add remote sink 0x%p, %d remaining\n",
-					dc_sink, aconnector->dc_link->sink_count);
+				drm_dbg_dp(connector->dev,
+					   "DM_MST: add remote sink 0x%p, %d remaining\n",
+					   dc_sink,
+					   aconnector->dc_link->sink_count);
 
 				dc_sink->priv = aconnector;
 				aconnector->dc_sink = dc_sink;
@@ -335,7 +411,7 @@ static int dm_dp_mst_get_modes(struct drm_connector *connector)
 			return ret;
 		}
 
-		aconnector->edid = edid;
+		aconnector->drm_edid = drm_edid;
 		amdgpu_dm_set_mst_status(&aconnector->mst_status,
 			MST_REMOTE_EDID, true);
 	}
@@ -350,10 +426,13 @@ static int dm_dp_mst_get_modes(struct drm_connector *connector)
 		struct dc_sink_init_data init_params = {
 				.link = aconnector->dc_link,
 				.sink_signal = SIGNAL_TYPE_DISPLAY_PORT_MST };
+		const struct edid *edid;
+
+		edid = drm_edid_raw(aconnector->drm_edid); // FIXME: Get rid of drm_edid_raw()
 		dc_sink = dc_link_add_remote_sink(
 			aconnector->dc_link,
-			(uint8_t *)aconnector->edid,
-			(aconnector->edid->extensions + 1) * EDID_LENGTH,
+			(uint8_t *)edid,
+			(edid->extensions + 1) * EDID_LENGTH,
 			&init_params);
 
 		if (!dc_sink) {
@@ -361,8 +440,9 @@ static int dm_dp_mst_get_modes(struct drm_connector *connector)
 			return 0;
 		}
 
-		DC_LOG_MST("DM_MST: add remote sink 0x%p, %d remaining\n",
-			dc_sink, aconnector->dc_link->sink_count);
+		drm_dbg_dp(connector->dev,
+			   "DM_MST: add remote sink 0x%p, %d remaining\n",
+			   dc_sink, aconnector->dc_link->sink_count);
 
 		dc_sink->priv = aconnector;
 		/* dc_link_add_remote_sink returns a new reference */
@@ -379,22 +459,28 @@ static int dm_dp_mst_get_modes(struct drm_connector *connector)
 		if (aconnector->dc_sink && connector->state) {
 			struct drm_device *dev = connector->dev;
 			struct amdgpu_device *adev = drm_to_adev(dev);
-			struct hdcp_workqueue *hdcp_work = adev->dm.hdcp_workqueue;
-			struct hdcp_workqueue *hdcp_w = &hdcp_work[aconnector->dc_link->link_index];
 
-			connector->state->hdcp_content_type =
-			hdcp_w->hdcp_content_type[connector->index];
-			connector->state->content_protection =
-			hdcp_w->content_protection[connector->index];
+			if (adev->dm.hdcp_workqueue) {
+				struct hdcp_workqueue *hdcp_work = adev->dm.hdcp_workqueue;
+				struct hdcp_workqueue *hdcp_w =
+					&hdcp_work[aconnector->dc_link->link_index];
+
+				connector->state->hdcp_content_type =
+				hdcp_w->hdcp_content_type[connector->index];
+				connector->state->content_protection =
+				hdcp_w->content_protection[connector->index];
+			}
 		}
 
 		if (aconnector->dc_sink) {
 			amdgpu_dm_update_freesync_caps(
-					connector, aconnector->edid);
+					connector, aconnector->drm_edid);
 
+#if defined(CONFIG_DRM_AMD_DC_FP)
 			if (!validate_dsc_caps_on_connector(aconnector))
 				memset(&aconnector->dc_sink->dsc_caps,
 				       0, sizeof(aconnector->dc_sink->dsc_caps));
+#endif
 
 			if (!retrieve_downstream_port_device(aconnector))
 				memset(&aconnector->mst_downstream_port_present,
@@ -402,10 +488,9 @@ static int dm_dp_mst_get_modes(struct drm_connector *connector)
 		}
 	}
 
-	drm_connector_update_edid_property(
-					&aconnector->base, aconnector->edid);
+	drm_edid_connector_update(&aconnector->base, aconnector->drm_edid);
 
-	ret = drm_add_edid_modes(connector, aconnector->edid);
+	ret = drm_edid_connector_add_modes(connector);
 
 	return ret;
 }
@@ -416,8 +501,7 @@ dm_mst_atomic_best_encoder(struct drm_connector *connector,
 {
 	struct drm_connector_state *connector_state = drm_atomic_get_new_connector_state(state,
 											 connector);
-	struct drm_device *dev = connector->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
+	struct amdgpu_device *adev = drm_to_adev(connector->dev);
 	struct amdgpu_crtc *acrtc = to_amdgpu_crtc(connector_state->crtc);
 
 	return &adev->dm.mst_encoders[acrtc->crtc_id].base;
@@ -477,12 +561,14 @@ dm_dp_mst_detect(struct drm_connector *connector,
 		if (aconnector->dc_link->sink_count)
 			dc_link_remove_remote_sink(aconnector->dc_link, aconnector->dc_sink);
 
-		DC_LOG_MST("DM_MST: remove remote sink 0x%p, %d remaining\n",
-			aconnector->dc_link, aconnector->dc_link->sink_count);
+		drm_dbg_dp(connector->dev,
+			   "DM_MST: remove remote sink 0x%p, %d remaining\n",
+			   aconnector->dc_link,
+			   aconnector->dc_link->sink_count);
 
 		dc_sink_release(aconnector->dc_sink);
 		aconnector->dc_sink = NULL;
-		aconnector->edid = NULL;
+		amdgpu_dm_mst_reset_mst_connector_setting(aconnector);
 
 		amdgpu_dm_set_mst_status(&aconnector->mst_status,
 			MST_REMOTE_EDID | MST_ALLOCATE_NEW_PAYLOAD | MST_CLEAR_ALLOCATED_PAYLOAD,
@@ -558,17 +644,20 @@ dm_dp_add_mst_connector(struct drm_dp_mst_topology_mgr *mgr,
 	if (!aconnector)
 		return NULL;
 
+	DRM_DEBUG_DRIVER("%s: Create aconnector 0x%p for port 0x%p\n", __func__, aconnector, port);
+
 	connector = &aconnector->base;
 	aconnector->mst_output_port = port;
 	aconnector->mst_root = master;
 	amdgpu_dm_set_mst_status(&aconnector->mst_status,
 			MST_PROBE, true);
 
-	if (drm_connector_init(
+	if (drm_connector_dynamic_init(
 		dev,
 		connector,
 		&dm_dp_mst_connector_funcs,
-		DRM_MODE_CONNECTOR_DisplayPort)) {
+		DRM_MODE_CONNECTOR_DisplayPort,
+		NULL)) {
 		kfree(aconnector);
 		return NULL;
 	}
@@ -602,9 +691,15 @@ dm_dp_add_mst_connector(struct drm_dp_mst_topology_mgr *mgr,
 		&connector->base,
 		dev->mode_config.tile_property,
 		0);
+	connector->colorspace_property = master->base.colorspace_property;
+	if (connector->colorspace_property)
+		drm_connector_attach_colorspace_property(connector);
 
 	drm_connector_set_path_property(connector, pathprop);
 
+	if (!retrieve_branch_specific_data(aconnector))
+		aconnector->branch_ieee_oui = 0;
+
 	/*
 	 * Initialize connector state before adding the connectror to drm and
 	 * framebuffer lists
@@ -616,8 +711,118 @@ dm_dp_add_mst_connector(struct drm_dp_mst_topology_mgr *mgr,
 	return connector;
 }
 
+void dm_handle_mst_sideband_msg_ready_event(
+	struct drm_dp_mst_topology_mgr *mgr,
+	enum mst_msg_ready_type msg_rdy_type)
+{
+	uint8_t esi[DP_PSR_ERROR_STATUS - DP_SINK_COUNT_ESI] = { 0 };
+	uint8_t dret;
+	bool new_irq_handled = false;
+	int dpcd_addr;
+	uint8_t dpcd_bytes_to_read;
+	const uint8_t max_process_count = 30;
+	uint8_t process_count = 0;
+	u8 retry;
+	struct amdgpu_dm_connector *aconnector =
+			container_of(mgr, struct amdgpu_dm_connector, mst_mgr);
+
+
+	const struct dc_link_status *link_status = dc_link_get_status(aconnector->dc_link);
+
+	if (link_status->dpcd_caps->dpcd_rev.raw < 0x12) {
+		dpcd_bytes_to_read = DP_LANE0_1_STATUS - DP_SINK_COUNT;
+		/* DPCD 0x200 - 0x201 for downstream IRQ */
+		dpcd_addr = DP_SINK_COUNT;
+	} else {
+		dpcd_bytes_to_read = DP_PSR_ERROR_STATUS - DP_SINK_COUNT_ESI;
+		/* DPCD 0x2002 - 0x2005 for downstream IRQ */
+		dpcd_addr = DP_SINK_COUNT_ESI;
+	}
+
+	mutex_lock(&aconnector->handle_mst_msg_ready);
+
+	while (process_count < max_process_count) {
+		u8 ack[DP_PSR_ERROR_STATUS - DP_SINK_COUNT_ESI] = {};
+
+		process_count++;
+
+		dret = drm_dp_dpcd_read(
+			&aconnector->dm_dp_aux.aux,
+			dpcd_addr,
+			esi,
+			dpcd_bytes_to_read);
+
+		if (dret != dpcd_bytes_to_read) {
+			DRM_DEBUG_KMS("DPCD read and acked number is not as expected!");
+			break;
+		}
+
+		DRM_DEBUG_DRIVER("ESI %02x %02x %02x\n", esi[0], esi[1], esi[2]);
+
+		switch (msg_rdy_type) {
+		case DOWN_REP_MSG_RDY_EVENT:
+			/* Only handle DOWN_REP_MSG_RDY case*/
+			esi[1] &= DP_DOWN_REP_MSG_RDY;
+			break;
+		case UP_REQ_MSG_RDY_EVENT:
+			/* Only handle UP_REQ_MSG_RDY case*/
+			esi[1] &= DP_UP_REQ_MSG_RDY;
+			break;
+		default:
+			/* Handle both cases*/
+			esi[1] &= (DP_DOWN_REP_MSG_RDY | DP_UP_REQ_MSG_RDY);
+			break;
+		}
+
+		if (!esi[1])
+			break;
+
+		/* handle MST irq */
+		if (aconnector->mst_mgr.mst_state)
+			drm_dp_mst_hpd_irq_handle_event(&aconnector->mst_mgr,
+						 esi,
+						 ack,
+						 &new_irq_handled);
+
+		if (new_irq_handled) {
+			/* ACK at DPCD to notify down stream */
+			for (retry = 0; retry < 3; retry++) {
+				ssize_t wret;
+
+				wret = drm_dp_dpcd_writeb(&aconnector->dm_dp_aux.aux,
+							  dpcd_addr + 1,
+							  ack[1]);
+				if (wret == 1)
+					break;
+			}
+
+			if (retry == 3) {
+				DRM_ERROR("Failed to ack MST event.\n");
+				break;
+			}
+
+			drm_dp_mst_hpd_irq_send_new_request(&aconnector->mst_mgr);
+
+			new_irq_handled = false;
+		} else {
+			break;
+		}
+	}
+
+	mutex_unlock(&aconnector->handle_mst_msg_ready);
+
+	if (process_count == max_process_count)
+		DRM_DEBUG_DRIVER("Loop exceeded max iterations\n");
+}
+
+static void dm_handle_mst_down_rep_msg_ready(struct drm_dp_mst_topology_mgr *mgr)
+{
+	dm_handle_mst_sideband_msg_ready_event(mgr, DOWN_REP_MSG_RDY_EVENT);
+}
+
 static const struct drm_dp_mst_topology_cbs dm_mst_cbs = {
 	.add_connector = dm_dp_add_mst_connector,
+	.poll_hpd_irq = dm_handle_mst_down_rep_msg_ready,
 };
 
 void amdgpu_dm_initialize_dp_connector(struct amdgpu_display_manager *dm,
@@ -636,6 +841,7 @@ void amdgpu_dm_initialize_dp_connector(struct amdgpu_display_manager *dm,
 	drm_dp_aux_init(&aconnector->dm_dp_aux.aux);
 	drm_dp_cec_register_connector(&aconnector->dm_dp_aux.aux,
 				      &aconnector->base);
+	drm_dp_dpcd_set_probe(&aconnector->dm_dp_aux.aux, false);
 
 	if (aconnector->base.connector_type == DRM_MODE_CONNECTOR_eDP)
 		return;
@@ -648,13 +854,20 @@ void amdgpu_dm_initialize_dp_connector(struct amdgpu_display_manager *dm,
 	drm_connector_attach_dp_subconnector_property(&aconnector->base);
 }
 
-int dm_mst_get_pbn_divider(struct dc_link *link)
+uint32_t dm_mst_get_pbn_divider(struct dc_link *link)
 {
+	uint32_t pbn_div_x100;
+	uint64_t dividend, divisor;
+
 	if (!link)
 		return 0;
 
-	return dc_link_bandwidth_kbps(link,
-			dc_link_get_link_cap(link)) / (8 * 1000 * 54);
+	dividend = (uint64_t)dc_link_bandwidth_kbps(link, dc_link_get_link_cap(link)) * 100;
+	divisor = 8 * 1000 * 54;
+
+	pbn_div_x100 = div64_u64(dividend, divisor);
+
+	return dfixed_const(pbn_div_x100) / 100;
 }
 
 struct dsc_mst_fairness_params {
@@ -670,26 +883,29 @@ struct dsc_mst_fairness_params {
 	struct amdgpu_dm_connector *aconnector;
 };
 
-static uint16_t get_fec_overhead_multiplier(struct dc_link *dc_link)
+#if defined(CONFIG_DRM_AMD_DC_FP)
+static uint64_t kbps_to_pbn(int kbps, bool is_peak_pbn)
 {
-	u8 link_coding_cap;
-	uint16_t fec_overhead_multiplier_x1000 = PBN_FEC_OVERHEAD_MULTIPLIER_8B_10B;
+	uint64_t effective_kbps = (uint64_t)kbps;
 
-	link_coding_cap = dc_link_dp_mst_decide_link_encoding_format(dc_link);
-	if (link_coding_cap == DP_128b_132b_ENCODING)
-		fec_overhead_multiplier_x1000 = PBN_FEC_OVERHEAD_MULTIPLIER_128B_132B;
+	if (is_peak_pbn) {	// add 0.6% (1006/1000) overhead into effective kbps
+		effective_kbps *= 1006;
+		effective_kbps = div_u64(effective_kbps, 1000);
+	}
 
-	return fec_overhead_multiplier_x1000;
+	return (uint64_t) DIV64_U64_ROUND_UP(effective_kbps * 64, (54 * 8 * 1000));
 }
 
-static int kbps_to_peak_pbn(int kbps, uint16_t fec_overhead_multiplier_x1000)
+static uint32_t pbn_to_kbps(unsigned int pbn, bool with_margin)
 {
-	u64 peak_kbps = kbps;
+	uint64_t pbn_effective = (uint64_t)pbn;
+
+	if (with_margin)	// deduct 0.6% (994/1000) overhead from effective pbn
+		pbn_effective *= (1000000 / PEAK_FACTOR_X1000);
+	else
+		pbn_effective *= 1000;
 
-	peak_kbps *= 1006;
-	peak_kbps *= fec_overhead_multiplier_x1000;
-	peak_kbps = div_u64(peak_kbps, 1000 * 1000);
-	return (int) DIV64_U64_ROUND_UP(peak_kbps * 64, (54 * 8 * 1000));
+	return DIV_U64_ROUND_UP(pbn_effective * 8 * 54, 64);
 }
 
 static void set_dsc_configs_from_fairness_vars(struct dsc_mst_fairness_params *params,
@@ -714,6 +930,7 @@ static void set_dsc_configs_from_fairness_vars(struct dsc_mst_fairness_params *p
 					&dsc_options,
 					0,
 					params[i].timing,
+					dc_link_get_highest_encoding_format(params[i].aconnector->dc_link),
 					&params[i].timing->dsc_cfg)) {
 			params[i].timing->flags.DSC = 1;
 
@@ -737,11 +954,11 @@ static void set_dsc_configs_from_fairness_vars(struct dsc_mst_fairness_params *p
 		if (params[i].sink) {
 			if (params[i].sink->sink_signal != SIGNAL_TYPE_VIRTUAL &&
 				params[i].sink->sink_signal != SIGNAL_TYPE_NONE)
-				DRM_DEBUG_DRIVER("%s i=%d dispname=%s\n", __func__, i,
+				DRM_DEBUG_DRIVER("MST_DSC %s i=%d dispname=%s\n", __func__, i,
 					params[i].sink->edid_caps.display_name);
 		}
 
-		DRM_DEBUG_DRIVER("dsc=%d bits_per_pixel=%d pbn=%d\n",
+		DRM_DEBUG_DRIVER("MST_DSC dsc=%d bits_per_pixel=%d pbn=%d\n",
 			params[i].timing->flags.DSC,
 			params[i].timing->dsc_cfg.bits_per_pixel,
 			vars[i + k].pbn);
@@ -759,12 +976,14 @@ static int bpp_x16_from_pbn(struct dsc_mst_fairness_params param, int pbn)
 	dc_dsc_get_default_config_option(param.sink->ctx->dc, &dsc_options);
 	dsc_options.max_target_bpp_limit_override_x16 = drm_connector->display_info.max_dsc_bpp * 16;
 
-	kbps = div_u64((u64)pbn * 994 * 8 * 54, 64);
+	kbps = pbn_to_kbps(pbn, false);
 	dc_dsc_compute_config(
 			param.sink->ctx->dc->res_pool->dscs[0],
 			&param.sink->dsc_caps.dsc_dec_caps,
 			&dsc_options,
-			(int) kbps, param.timing, &dsc_config);
+			(int) kbps, param.timing,
+			dc_link_get_highest_encoding_format(param.aconnector->dc_link),
+			&dsc_config);
 
 	return dsc_config.bits_per_pixel;
 }
@@ -786,12 +1005,11 @@ static int increase_dsc_bpp(struct drm_atomic_state *state,
 	int link_timeslots_used;
 	int fair_pbn_alloc;
 	int ret = 0;
-	uint16_t fec_overhead_multiplier_x1000 = get_fec_overhead_multiplier(dc_link);
 
 	for (i = 0; i < count; i++) {
 		if (vars[i + k].dsc_enabled) {
 			initial_slack[i] =
-			kbps_to_peak_pbn(params[i].bw_range.max_kbps, fec_overhead_multiplier_x1000) - vars[i + k].pbn;
+			kbps_to_pbn(params[i].bw_range.max_kbps, false) - vars[i + k].pbn;
 			bpp_increased[i] = false;
 			remaining_to_increase += 1;
 		} else {
@@ -818,10 +1036,10 @@ static int increase_dsc_bpp(struct drm_atomic_state *state,
 		link_timeslots_used = 0;
 
 		for (i = 0; i < count; i++)
-			link_timeslots_used += DIV_ROUND_UP(vars[i + k].pbn, mst_state->pbn_div);
+			link_timeslots_used += DIV_ROUND_UP(vars[i + k].pbn, dfixed_trunc(mst_state->pbn_div));
 
 		fair_pbn_alloc =
-			(63 - link_timeslots_used) / remaining_to_increase * mst_state->pbn_div;
+			(63 - link_timeslots_used) / remaining_to_increase * dfixed_trunc(mst_state->pbn_div);
 
 		if (initial_slack[next_index] > fair_pbn_alloc) {
 			vars[next_index].pbn += fair_pbn_alloc;
@@ -887,7 +1105,7 @@ static int try_disable_dsc(struct drm_atomic_state *state,
 	int next_index;
 	int remaining_to_try = 0;
 	int ret;
-	uint16_t fec_overhead_multiplier_x1000 = get_fec_overhead_multiplier(dc_link);
+	int var_pbn;
 
 	for (i = 0; i < count; i++) {
 		if (vars[i + k].dsc_enabled
@@ -917,26 +1135,37 @@ static int try_disable_dsc(struct drm_atomic_state *state,
 		if (next_index == -1)
 			break;
 
-		vars[next_index].pbn = kbps_to_peak_pbn(params[next_index].bw_range.stream_kbps, fec_overhead_multiplier_x1000);
+		DRM_DEBUG_DRIVER("MST_DSC index #%d, try no compression\n", next_index);
+		var_pbn = vars[next_index].pbn;
+		vars[next_index].pbn = kbps_to_pbn(params[next_index].bw_range.stream_kbps, true);
 		ret = drm_dp_atomic_find_time_slots(state,
 						    params[next_index].port->mgr,
 						    params[next_index].port,
 						    vars[next_index].pbn);
-		if (ret < 0)
+		if (ret < 0) {
+			DRM_DEBUG_DRIVER("%s:%d MST_DSC index #%d, failed to set pbn to the state, %d\n",
+						__func__, __LINE__, next_index, ret);
+			vars[next_index].pbn = var_pbn;
 			return ret;
+		}
 
 		ret = drm_dp_mst_atomic_check(state);
 		if (ret == 0) {
+			DRM_DEBUG_DRIVER("MST_DSC index #%d, greedily disable dsc\n", next_index);
 			vars[next_index].dsc_enabled = false;
 			vars[next_index].bpp_x16 = 0;
 		} else {
-			vars[next_index].pbn = kbps_to_peak_pbn(params[next_index].bw_range.max_kbps, fec_overhead_multiplier_x1000);
+			DRM_DEBUG_DRIVER("MST_DSC index #%d, restore optimized pbn value\n", next_index);
+			vars[next_index].pbn = var_pbn;
 			ret = drm_dp_atomic_find_time_slots(state,
 							    params[next_index].port->mgr,
 							    params[next_index].port,
 							    vars[next_index].pbn);
-			if (ret < 0)
+			if (ret < 0) {
+				DRM_DEBUG_DRIVER("%s:%d MST_DSC index #%d, failed to set pbn to the state, %d\n",
+							__func__, __LINE__, next_index, ret);
 				return ret;
+			}
 		}
 
 		tried[next_index] = true;
@@ -945,6 +1174,15 @@ static int try_disable_dsc(struct drm_atomic_state *state,
 	return 0;
 }
 
+static void log_dsc_params(int count, struct dsc_mst_fairness_vars *vars, int k)
+{
+	int i;
+
+	for (i = 0; i < count; i++)
+		DRM_DEBUG_DRIVER("MST_DSC DSC params: stream #%d --- dsc_enabled = %d, bpp_x16 = %d, pbn = %d\n",
+				 i, vars[i + k].dsc_enabled, vars[i + k].bpp_x16, vars[i + k].pbn);
+}
+
 static int compute_mst_dsc_configs_for_link(struct drm_atomic_state *state,
 					    struct dc_state *dc_state,
 					    struct dc_link *dc_link,
@@ -959,7 +1197,7 @@ static int compute_mst_dsc_configs_for_link(struct drm_atomic_state *state,
 	int count = 0;
 	int i, k, ret;
 	bool debugfs_overwrite = false;
-	uint16_t fec_overhead_multiplier_x1000 = get_fec_overhead_multiplier(dc_link);
+	struct drm_connector_state *new_conn_state;
 
 	memset(params, 0, sizeof(params));
 
@@ -967,6 +1205,7 @@ static int compute_mst_dsc_configs_for_link(struct drm_atomic_state *state,
 		return PTR_ERR(mst_state);
 
 	/* Set up params */
+	DRM_DEBUG_DRIVER("%s: MST_DSC Try to set up params from %d streams\n", __func__, dc_state->stream_count);
 	for (i = 0; i < dc_state->stream_count; i++) {
 		struct dc_dsc_policy dsc_policy = {0};
 
@@ -982,6 +1221,14 @@ static int compute_mst_dsc_configs_for_link(struct drm_atomic_state *state,
 		if (!aconnector->mst_output_port)
 			continue;
 
+		new_conn_state = drm_atomic_get_new_connector_state(state, &aconnector->base);
+
+		if (!new_conn_state) {
+			DRM_DEBUG_DRIVER("%s:%d MST_DSC Skip the stream 0x%p with invalid new_conn_state\n",
+					__func__, __LINE__, stream);
+			continue;
+		}
+
 		stream->timing.flags.DSC = 0;
 
 		params[count].timing = &stream->timing;
@@ -995,19 +1242,27 @@ static int compute_mst_dsc_configs_for_link(struct drm_atomic_state *state,
 		params[count].num_slices_v = aconnector->dsc_settings.dsc_num_slices_v;
 		params[count].bpp_overwrite = aconnector->dsc_settings.dsc_bits_per_pixel;
 		params[count].compression_possible = stream->sink->dsc_caps.dsc_dec_caps.is_dsc_supported;
-		dc_dsc_get_policy_for_timing(params[count].timing, 0, &dsc_policy);
+		dc_dsc_get_policy_for_timing(params[count].timing, 0, &dsc_policy, dc_link_get_highest_encoding_format(stream->link));
 		if (!dc_dsc_compute_bandwidth_range(
 				stream->sink->ctx->dc->res_pool->dscs[0],
 				stream->sink->ctx->dc->debug.dsc_min_slice_height_override,
 				dsc_policy.min_target_bpp * 16,
 				dsc_policy.max_target_bpp * 16,
 				&stream->sink->dsc_caps.dsc_dec_caps,
-				&stream->timing, &params[count].bw_range))
-			params[count].bw_range.stream_kbps = dc_bandwidth_in_kbps_from_timing(&stream->timing);
-
+				&stream->timing,
+				dc_link_get_highest_encoding_format(dc_link),
+				&params[count].bw_range))
+			params[count].bw_range.stream_kbps = dc_bandwidth_in_kbps_from_timing(&stream->timing,
+					dc_link_get_highest_encoding_format(dc_link));
+
+		DRM_DEBUG_DRIVER("MST_DSC #%d stream 0x%p - max_kbps = %u, min_kbps = %u, uncompressed_kbps = %u\n",
+			count, stream, params[count].bw_range.max_kbps, params[count].bw_range.min_kbps,
+			params[count].bw_range.stream_kbps);
 		count++;
 	}
 
+	DRM_DEBUG_DRIVER("%s: MST_DSC Params set up for %d streams\n", __func__, count);
+
 	if (count == 0) {
 		ASSERT(0);
 		return 0;
@@ -1019,9 +1274,10 @@ static int compute_mst_dsc_configs_for_link(struct drm_atomic_state *state,
 	*link_vars_start_index += count;
 
 	/* Try no compression */
+	DRM_DEBUG_DRIVER("MST_DSC Try no compression\n");
 	for (i = 0; i < count; i++) {
 		vars[i + k].aconnector = params[i].aconnector;
-		vars[i + k].pbn = kbps_to_peak_pbn(params[i].bw_range.stream_kbps, fec_overhead_multiplier_x1000);
+		vars[i + k].pbn = kbps_to_pbn(params[i].bw_range.stream_kbps, false);
 		vars[i + k].dsc_enabled = false;
 		vars[i + k].bpp_x16 = 0;
 		ret = drm_dp_atomic_find_time_slots(state, params[i].port->mgr, params[i].port,
@@ -1037,10 +1293,13 @@ static int compute_mst_dsc_configs_for_link(struct drm_atomic_state *state,
 		return ret;
 	}
 
+	log_dsc_params(count, vars, k);
+
 	/* Try max compression */
+	DRM_DEBUG_DRIVER("MST_DSC Try max compression\n");
 	for (i = 0; i < count; i++) {
 		if (params[i].compression_possible && params[i].clock_force_enable != DSC_CLK_FORCE_DISABLE) {
-			vars[i + k].pbn = kbps_to_peak_pbn(params[i].bw_range.min_kbps, fec_overhead_multiplier_x1000);
+			vars[i + k].pbn = kbps_to_pbn(params[i].bw_range.min_kbps, false);
 			vars[i + k].dsc_enabled = true;
 			vars[i + k].bpp_x16 = params[i].bw_range.min_target_bpp_x16;
 			ret = drm_dp_atomic_find_time_slots(state, params[i].port->mgr,
@@ -1048,7 +1307,7 @@ static int compute_mst_dsc_configs_for_link(struct drm_atomic_state *state,
 			if (ret < 0)
 				return ret;
 		} else {
-			vars[i + k].pbn = kbps_to_peak_pbn(params[i].bw_range.stream_kbps, fec_overhead_multiplier_x1000);
+			vars[i + k].pbn = kbps_to_pbn(params[i].bw_range.stream_kbps, false);
 			vars[i + k].dsc_enabled = false;
 			vars[i + k].bpp_x16 = 0;
 			ret = drm_dp_atomic_find_time_slots(state, params[i].port->mgr,
@@ -1061,14 +1320,26 @@ static int compute_mst_dsc_configs_for_link(struct drm_atomic_state *state,
 	if (ret != 0)
 		return ret;
 
+	log_dsc_params(count, vars, k);
+
 	/* Optimize degree of compression */
+	DRM_DEBUG_DRIVER("MST_DSC Try optimize compression\n");
 	ret = increase_dsc_bpp(state, mst_state, dc_link, params, vars, count, k);
-	if (ret < 0)
+	if (ret < 0) {
+		DRM_DEBUG_DRIVER("MST_DSC Failed to optimize compression\n");
 		return ret;
+	}
+
+	log_dsc_params(count, vars, k);
 
+	DRM_DEBUG_DRIVER("MST_DSC Try disable compression\n");
 	ret = try_disable_dsc(state, dc_link, params, vars, count, k);
-	if (ret < 0)
+	if (ret < 0) {
+		DRM_DEBUG_DRIVER("MST_DSC Failed to disable compression\n");
 		return ret;
+	}
+
+	log_dsc_params(count, vars, k);
 
 	set_dsc_configs_from_fairness_vars(params, vars, count, k);
 
@@ -1090,15 +1361,19 @@ static bool is_dsc_need_re_compute(
 
 	/* only check phy used by dsc mst branch */
 	if (dc_link->type != dc_connection_mst_branch)
-		return false;
+		goto out;
 
-	if (!(dc_link->dpcd_caps.dsc_caps.dsc_basic_caps.fields.dsc_support.DSC_SUPPORT ||
-		dc_link->dpcd_caps.dsc_caps.dsc_basic_caps.fields.dsc_support.DSC_PASSTHROUGH_SUPPORT))
-		return false;
+	/* add a check for older MST DSC with no virtual DPCDs */
+	if (needs_dsc_aux_workaround(dc_link)  &&
+		(!(dc_link->dpcd_caps.dsc_caps.dsc_basic_caps.fields.dsc_support.DSC_SUPPORT ||
+		dc_link->dpcd_caps.dsc_caps.dsc_basic_caps.fields.dsc_support.DSC_PASSTHROUGH_SUPPORT)))
+		goto out;
 
 	for (i = 0; i < MAX_PIPES; i++)
 		stream_on_link[i] = NULL;
 
+	DRM_DEBUG_DRIVER("%s: MST_DSC check on %d streams in new dc_state\n", __func__, dc_state->stream_count);
+
 	/* check if there is mode change in new request */
 	for (i = 0; i < dc_state->stream_count; i++) {
 		struct drm_crtc_state *new_crtc_state;
@@ -1108,6 +1383,8 @@ static bool is_dsc_need_re_compute(
 		if (!stream)
 			continue;
 
+		DRM_DEBUG_DRIVER("%s:%d MST_DSC checking #%d stream 0x%p\n", __func__, __LINE__, i, stream);
+
 		/* check if stream using the same link for mst */
 		if (stream->link != dc_link)
 			continue;
@@ -1120,8 +1397,11 @@ static bool is_dsc_need_re_compute(
 		new_stream_on_link_num++;
 
 		new_conn_state = drm_atomic_get_new_connector_state(state, &aconnector->base);
-		if (!new_conn_state)
+		if (!new_conn_state) {
+			DRM_DEBUG_DRIVER("%s:%d MST_DSC no new_conn_state for stream 0x%p, aconnector 0x%p\n",
+					 __func__, __LINE__, stream, aconnector);
 			continue;
+		}
 
 		if (IS_ERR(new_conn_state))
 			continue;
@@ -1130,19 +1410,37 @@ static bool is_dsc_need_re_compute(
 			continue;
 
 		new_crtc_state = drm_atomic_get_new_crtc_state(state, new_conn_state->crtc);
-		if (!new_crtc_state)
+		if (!new_crtc_state) {
+			DRM_DEBUG_DRIVER("%s:%d MST_DSC no new_crtc_state for crtc of stream 0x%p, aconnector 0x%p\n",
+						__func__, __LINE__, stream, aconnector);
 			continue;
+		}
 
 		if (IS_ERR(new_crtc_state))
 			continue;
 
 		if (new_crtc_state->enable && new_crtc_state->active) {
 			if (new_crtc_state->mode_changed || new_crtc_state->active_changed ||
-				new_crtc_state->connectors_changed)
-				return true;
+					new_crtc_state->connectors_changed) {
+				DRM_DEBUG_DRIVER("%s:%d MST_DSC dsc recompute required."
+						 "stream 0x%p in new dc_state\n",
+						 __func__, __LINE__, stream);
+				is_dsc_need_re_compute = true;
+				goto out;
+			}
 		}
 	}
 
+	if (new_stream_on_link_num == 0) {
+		DRM_DEBUG_DRIVER("%s:%d MST_DSC no mode change request for streams in new dc_state\n",
+				 __func__, __LINE__);
+		is_dsc_need_re_compute = false;
+		goto out;
+	}
+
+	DRM_DEBUG_DRIVER("%s: MST_DSC check on %d streams in current dc_state\n",
+			 __func__, dc->current_state->stream_count);
+
 	/* check current_state if there stream on link but it is not in
 	 * new request state
 	 */
@@ -1165,11 +1463,18 @@ static bool is_dsc_need_re_compute(
 
 		if (j == new_stream_on_link_num) {
 			/* not in new state */
+			DRM_DEBUG_DRIVER("%s:%d MST_DSC dsc recompute required."
+					 "stream 0x%p in current dc_state but not in new dc_state\n",
+						__func__, __LINE__, stream);
 			is_dsc_need_re_compute = true;
 			break;
 		}
 	}
 
+out:
+	DRM_DEBUG_DRIVER("%s: MST_DSC dsc recompute %s\n",
+			 __func__, is_dsc_need_re_compute ? "required" : "not required");
+
 	return is_dsc_need_re_compute;
 }
 
@@ -1198,6 +1503,9 @@ int compute_mst_dsc_configs_for_state(struct drm_atomic_state *state,
 
 		aconnector = (struct amdgpu_dm_connector *)stream->dm_stream_context;
 
+		DRM_DEBUG_DRIVER("%s: MST_DSC compute mst dsc configs for stream 0x%p, aconnector 0x%p\n",
+				__func__, stream, aconnector);
+
 		if (!aconnector || !aconnector->dc_sink || !aconnector->mst_output_port)
 			continue;
 
@@ -1207,7 +1515,7 @@ int compute_mst_dsc_configs_for_state(struct drm_atomic_state *state,
 		if (computed_streams[i])
 			continue;
 
-		if (!res_pool->funcs->remove_stream_from_ctx ||
+		if (res_pool->funcs->remove_stream_from_ctx &&
 		    res_pool->funcs->remove_stream_from_ctx(stream->ctx->dc, dc_state, stream) != DC_OK)
 			return -EINVAL;
 
@@ -1230,8 +1538,11 @@ int compute_mst_dsc_configs_for_state(struct drm_atomic_state *state,
 		stream = dc_state->streams[i];
 
 		if (stream->timing.flags.DSC == 1)
-			if (dc_stream_add_dsc_to_resource(stream->ctx->dc, dc_state, stream) != DC_OK)
+			if (dc_stream_add_dsc_to_resource(stream->ctx->dc, dc_state, stream) != DC_OK) {
+				DRM_DEBUG_DRIVER("%s:%d MST_DSC Failed to request dsc hw resource for stream 0x%p\n",
+							__func__, __LINE__, stream);
 				return -EINVAL;
+			}
 	}
 
 	return ret;
@@ -1260,6 +1571,9 @@ static int pre_compute_mst_dsc_configs_for_state(struct drm_atomic_state *state,
 
 		aconnector = (struct amdgpu_dm_connector *)stream->dm_stream_context;
 
+		DRM_DEBUG_DRIVER("MST_DSC pre compute mst dsc configs for #%d stream 0x%p, aconnector 0x%p\n",
+					i, stream, aconnector);
+
 		if (!aconnector || !aconnector->dc_sink || !aconnector->mst_output_port)
 			continue;
 
@@ -1349,12 +1663,12 @@ int pre_validate_dsc(struct drm_atomic_state *state,
 	int ret = 0;
 
 	if (!is_dsc_precompute_needed(state)) {
-		DRM_INFO_ONCE("DSC precompute is not needed.\n");
+		DRM_INFO_ONCE("%s:%d MST_DSC dsc precompute is not needed\n", __func__, __LINE__);
 		return 0;
 	}
 	ret = dm_atomic_get_state(state, dm_state_ptr);
 	if (ret != 0) {
-		DRM_INFO_ONCE("dm_atomic_get_state() failed\n");
+		DRM_INFO_ONCE("%s:%d MST_DSC dm_atomic_get_state() failed\n", __func__, __LINE__);
 		return ret;
 	}
 	dm_state = *dm_state_ptr;
@@ -1365,31 +1679,35 @@ int pre_validate_dsc(struct drm_atomic_state *state,
 	 * from dm_state->context.
 	 */
 
-	local_dc_state = kmemdup(dm_state->context, sizeof(struct dc_state), GFP_KERNEL);
+	local_dc_state = vmalloc(sizeof(struct dc_state));
 	if (!local_dc_state)
 		return -ENOMEM;
+	memcpy(local_dc_state, dm_state->context, sizeof(struct dc_state));
 
 	for (i = 0; i < local_dc_state->stream_count; i++) {
 		struct dc_stream_state *stream = dm_state->context->streams[i];
 		int ind = find_crtc_index_in_state_by_stream(state, stream);
 
 		if (ind >= 0) {
-			struct amdgpu_dm_connector *aconnector;
+			struct drm_connector *connector;
 			struct drm_connector_state *drm_new_conn_state;
 			struct dm_connector_state *dm_new_conn_state;
 			struct dm_crtc_state *dm_old_crtc_state;
 
-			aconnector =
+			connector =
 				amdgpu_dm_find_first_crtc_matching_connector(state,
 									     state->crtcs[ind].ptr);
+			if (!connector)
+				continue;
+
 			drm_new_conn_state =
 				drm_atomic_get_new_connector_state(state,
-								   &aconnector->base);
+								   connector);
 			dm_new_conn_state = to_dm_connector_state(drm_new_conn_state);
 			dm_old_crtc_state = to_dm_crtc_state(state->crtcs[ind].old_state);
 
 			local_dc_state->streams[i] =
-				create_validate_stream_for_sink(aconnector,
+				create_validate_stream_for_sink(connector,
 								&state->crtcs[ind].new_state->mode,
 								dm_new_conn_state,
 								dm_old_crtc_state->stream);
@@ -1405,7 +1723,9 @@ int pre_validate_dsc(struct drm_atomic_state *state,
 
 	ret = pre_compute_mst_dsc_configs_for_state(state, local_dc_state, vars);
 	if (ret != 0) {
-		DRM_INFO_ONCE("pre_compute_mst_dsc_configs_for_state() failed\n");
+		DRM_INFO_ONCE("%s:%d MST_DSC dsc pre_compute_mst_dsc_configs_for_state() failed\n",
+				__func__, __LINE__);
+		ret = -EINVAL;
 		goto clean_exit;
 	}
 
@@ -1417,13 +1737,16 @@ int pre_validate_dsc(struct drm_atomic_state *state,
 		struct dc_stream_state *stream = dm_state->context->streams[i];
 
 		if (local_dc_state->streams[i] &&
-		    is_timing_changed(stream, local_dc_state->streams[i])) {
-			DRM_INFO_ONCE("crtc[%d] needs mode_changed\n", i);
+		    dc_is_timing_changed(stream, local_dc_state->streams[i])) {
+			DRM_INFO_ONCE("%s:%d MST_DSC crtc[%d] needs mode_change\n", __func__, __LINE__, i);
 		} else {
 			int ind = find_crtc_index_in_state_by_stream(state, stream);
 
-			if (ind >= 0)
+			if (ind >= 0) {
+				DRM_INFO_ONCE("%s:%d MST_DSC no mode changed for stream 0x%p\n",
+						__func__, __LINE__, stream);
 				state->crtcs[ind].new_state->mode_changed = 0;
+			}
 		}
 	}
 clean_exit:
@@ -1434,112 +1757,218 @@ clean_exit:
 			dc_stream_release(local_dc_state->streams[i]);
 	}
 
-	kfree(local_dc_state);
+	vfree(local_dc_state);
 
 	return ret;
 }
 
-static unsigned int kbps_from_pbn(unsigned int pbn)
+static bool is_dsc_common_config_possible(struct dc_stream_state *stream,
+					  struct dc_dsc_bw_range *bw_range)
 {
-	unsigned int kbps = pbn;
+	struct dc_dsc_policy dsc_policy = {0};
+	bool is_dsc_possible;
 
-	kbps *= (1000000 / PEAK_FACTOR_X1000);
-	kbps *= 8;
-	kbps *= 54;
-	kbps /= 64;
+	dc_dsc_get_policy_for_timing(&stream->timing, 0, &dsc_policy, dc_link_get_highest_encoding_format(stream->link));
+	is_dsc_possible = dc_dsc_compute_bandwidth_range(stream->sink->ctx->dc->res_pool->dscs[0],
+							 stream->sink->ctx->dc->debug.dsc_min_slice_height_override,
+							 dsc_policy.min_target_bpp * 16,
+							 dsc_policy.max_target_bpp * 16,
+							 &stream->sink->dsc_caps.dsc_dec_caps,
+							 &stream->timing, dc_link_get_highest_encoding_format(stream->link), bw_range);
 
-	return kbps;
+	return is_dsc_possible;
 }
+#endif
 
-static bool is_dsc_common_config_possible(struct dc_stream_state *stream,
-					  struct dc_dsc_bw_range *bw_range)
+#if defined(CONFIG_DRM_AMD_DC_FP)
+static bool dp_get_link_current_set_bw(struct drm_dp_aux *aux, uint32_t *cur_link_bw)
 {
-	struct dc_dsc_policy dsc_policy = {0};
+	uint32_t total_data_bw_efficiency_x10000 = 0;
+	uint32_t link_rate_per_lane_kbps = 0;
+	enum dc_link_rate link_rate;
+	union lane_count_set lane_count;
+	u8 dp_link_encoding;
+	u8 link_bw_set = 0;
+	u8 data[16] = {0};
+
+	*cur_link_bw = 0;
+
+	if (drm_dp_dpcd_read(aux, DP_LINK_BW_SET, data, 16) != 16)
+		return false;
+
+	dp_link_encoding = data[DP_MAIN_LINK_CHANNEL_CODING_SET - DP_LINK_BW_SET];
+	link_bw_set = data[DP_LINK_BW_SET - DP_LINK_BW_SET];
+	lane_count.raw = data[DP_LANE_COUNT_SET - DP_LINK_BW_SET];
+
+	drm_dbg_dp(aux->drm_dev, "MST_DSC downlink setting: %d, 0x%x x %d\n",
+		   dp_link_encoding, link_bw_set, lane_count.bits.LANE_COUNT_SET);
 
-	dc_dsc_get_policy_for_timing(&stream->timing, 0, &dsc_policy);
-	dc_dsc_compute_bandwidth_range(stream->sink->ctx->dc->res_pool->dscs[0],
-				       stream->sink->ctx->dc->debug.dsc_min_slice_height_override,
-				       dsc_policy.min_target_bpp * 16,
-				       dsc_policy.max_target_bpp * 16,
-				       &stream->sink->dsc_caps.dsc_dec_caps,
-				       &stream->timing, bw_range);
+	switch (dp_link_encoding) {
+	case DP_8b_10b_ENCODING:
+		link_rate = link_bw_set;
+		link_rate_per_lane_kbps = link_rate * LINK_RATE_REF_FREQ_IN_KHZ * BITS_PER_DP_BYTE;
+		total_data_bw_efficiency_x10000 = DATA_EFFICIENCY_8b_10b_x10000;
+		total_data_bw_efficiency_x10000 /= 100;
+		total_data_bw_efficiency_x10000 *= DATA_EFFICIENCY_8b_10b_FEC_EFFICIENCY_x100;
+		break;
+	case DP_128b_132b_ENCODING:
+		switch (link_bw_set) {
+		case DP_LINK_BW_10:
+			link_rate = LINK_RATE_UHBR10;
+			break;
+		case DP_LINK_BW_13_5:
+			link_rate = LINK_RATE_UHBR13_5;
+			break;
+		case DP_LINK_BW_20:
+			link_rate = LINK_RATE_UHBR20;
+			break;
+		default:
+			return false;
+		}
 
-	return bw_range->max_target_bpp_x16 && bw_range->min_target_bpp_x16;
+		link_rate_per_lane_kbps = link_rate * 10000;
+		total_data_bw_efficiency_x10000 = DATA_EFFICIENCY_128b_132b_x10000;
+		break;
+	default:
+		return false;
+	}
+
+	*cur_link_bw = link_rate_per_lane_kbps * lane_count.bits.LANE_COUNT_SET / 10000 * total_data_bw_efficiency_x10000;
+	return true;
 }
+#endif
 
 enum dc_status dm_dp_mst_is_port_support_mode(
 	struct amdgpu_dm_connector *aconnector,
 	struct dc_stream_state *stream)
 {
-	int bpp, pbn, branch_max_throughput_mps = 0;
+#if defined(CONFIG_DRM_AMD_DC_FP)
+	int branch_max_throughput_mps = 0;
 	struct dc_link_settings cur_link_settings;
-	unsigned int end_to_end_bw_in_kbps = 0;
-	unsigned int upper_link_bw_in_kbps = 0, down_link_bw_in_kbps = 0;
-	unsigned int max_compressed_bw_in_kbps = 0;
+	uint32_t end_to_end_bw_in_kbps = 0;
+	uint32_t root_link_bw_in_kbps = 0;
+	uint32_t virtual_channel_bw_in_kbps = 0;
 	struct dc_dsc_bw_range bw_range = {0};
-	struct drm_dp_mst_topology_mgr *mst_mgr;
+	struct dc_dsc_config_options dsc_options = {0};
+	uint32_t stream_kbps;
 
-	/*
-	 * check if the mode could be supported if DSC pass-through is supported
-	 * AND check if there enough bandwidth available to support the mode
-	 * with DSC enabled.
+	/* DSC unnecessary case
+	 * Check if timing could be supported within end-to-end BW
 	 */
-	if (is_dsc_common_config_possible(stream, &bw_range) &&
-	    aconnector->mst_output_port->passthrough_aux) {
-		mst_mgr = aconnector->mst_output_port->mgr;
-		mutex_lock(&mst_mgr->lock);
+	stream_kbps =
+		dc_bandwidth_in_kbps_from_timing(&stream->timing,
+			dc_link_get_highest_encoding_format(stream->link));
+	cur_link_settings = stream->link->verified_link_cap;
+	root_link_bw_in_kbps = dc_link_bandwidth_kbps(aconnector->dc_link, &cur_link_settings);
+	virtual_channel_bw_in_kbps = pbn_to_kbps(aconnector->mst_output_port->full_pbn, true);
+
+	/* pick the end to end bw bottleneck */
+	end_to_end_bw_in_kbps = min(root_link_bw_in_kbps, virtual_channel_bw_in_kbps);
+
+	if (stream_kbps <= end_to_end_bw_in_kbps) {
+		DRM_DEBUG_DRIVER("MST_DSC no dsc required. End-to-end bw sufficient\n");
+		return DC_OK;
+	}
 
-		cur_link_settings = stream->link->verified_link_cap;
+	/*DSC necessary case*/
+	if (!aconnector->dsc_aux)
+		return DC_FAIL_BANDWIDTH_VALIDATE;
 
-		upper_link_bw_in_kbps = dc_link_bandwidth_kbps(aconnector->dc_link,
-							       &cur_link_settings
-							       );
-		down_link_bw_in_kbps = kbps_from_pbn(aconnector->mst_output_port->full_pbn);
+	if (is_dsc_common_config_possible(stream, &bw_range)) {
 
-		/* pick the bottleneck */
-		end_to_end_bw_in_kbps = min(upper_link_bw_in_kbps,
-					    down_link_bw_in_kbps);
+		/*capable of dsc passthough. dsc bitstream along the entire path*/
+		if (aconnector->mst_output_port->passthrough_aux) {
+			if (bw_range.min_kbps > end_to_end_bw_in_kbps) {
+				DRM_DEBUG_DRIVER("MST_DSC dsc passthrough and decode at endpoint"
+						 "Max dsc compression bw can't fit into end-to-end bw\n");
+				return DC_FAIL_BANDWIDTH_VALIDATE;
+			}
+		} else {
+			/*dsc bitstream decoded at the dp last link*/
+			struct drm_dp_mst_port *immediate_upstream_port = NULL;
+			uint32_t end_link_bw = 0;
+
+			/*Get last DP link BW capability. Mode shall be supported by Legacy peer*/
+			if (aconnector->mst_output_port->pdt != DP_PEER_DEVICE_DP_LEGACY_CONV &&
+				aconnector->mst_output_port->pdt != DP_PEER_DEVICE_NONE) {
+				if (aconnector->vc_full_pbn != aconnector->mst_output_port->full_pbn) {
+					dp_get_link_current_set_bw(&aconnector->mst_output_port->aux, &end_link_bw);
+					aconnector->vc_full_pbn = aconnector->mst_output_port->full_pbn;
+					aconnector->mst_local_bw = end_link_bw;
+				} else {
+					end_link_bw = aconnector->mst_local_bw;
+				}
 
-		mutex_unlock(&mst_mgr->lock);
+				if (end_link_bw > 0 &&
+				    stream_kbps > end_link_bw &&
+				    aconnector->branch_ieee_oui != DP_BRANCH_DEVICE_ID_90CC24) {
+					DRM_DEBUG_DRIVER("MST_DSC dsc decode at last link. "
+							 "Mode required bw can't fit into last link\n");
+					return DC_FAIL_BANDWIDTH_VALIDATE;
+				}
+			}
 
-		/*
-		 * use the maximum dsc compression bandwidth as the required
-		 * bandwidth for the mode
-		 */
-		max_compressed_bw_in_kbps = bw_range.min_kbps;
+			/*Get virtual channel bandwidth between source and the link before the last link*/
+			if (aconnector->mst_output_port->parent->port_parent)
+				immediate_upstream_port = aconnector->mst_output_port->parent->port_parent;
 
-		if (end_to_end_bw_in_kbps < max_compressed_bw_in_kbps) {
-			DRM_DEBUG_DRIVER("Mode does not fit into DSC pass-through bandwidth validation\n");
-			return DC_FAIL_BANDWIDTH_VALIDATE;
+			if (immediate_upstream_port) {
+				virtual_channel_bw_in_kbps = pbn_to_kbps(immediate_upstream_port->full_pbn, true);
+				virtual_channel_bw_in_kbps = min(root_link_bw_in_kbps, virtual_channel_bw_in_kbps);
+			} else {
+				/* For topology LCT 1 case - only one mstb*/
+				virtual_channel_bw_in_kbps = root_link_bw_in_kbps;
+			}
+
+			if (bw_range.min_kbps > virtual_channel_bw_in_kbps) {
+				DRM_DEBUG_DRIVER("MST_DSC dsc decode at last link."
+						 "Max dsc compression can't fit into MST available bw\n");
+				return DC_FAIL_BANDWIDTH_VALIDATE;
+			}
 		}
-	} else {
-		/* check if mode could be supported within full_pbn */
-		bpp = convert_dc_color_depth_into_bpc(stream->timing.display_color_depth) * 3;
-		pbn = drm_dp_calc_pbn_mode(stream->timing.pix_clk_100hz / 10, bpp, false);
 
-		if (pbn > aconnector->mst_output_port->full_pbn)
+		/*Confirm if we can obtain dsc config*/
+		dc_dsc_get_default_config_option(stream->link->dc, &dsc_options);
+		dsc_options.max_target_bpp_limit_override_x16 = aconnector->base.display_info.max_dsc_bpp * 16;
+		if (dc_dsc_compute_config(stream->sink->ctx->dc->res_pool->dscs[0],
+				&stream->sink->dsc_caps.dsc_dec_caps,
+				&dsc_options,
+				end_to_end_bw_in_kbps,
+				&stream->timing,
+				dc_link_get_highest_encoding_format(stream->link),
+				&stream->timing.dsc_cfg)) {
+			stream->timing.flags.DSC = 1;
+			DRM_DEBUG_DRIVER("MST_DSC require dsc and dsc config found\n");
+		} else {
+			DRM_DEBUG_DRIVER("MST_DSC require dsc but can't find appropriate dsc config\n");
 			return DC_FAIL_BANDWIDTH_VALIDATE;
-	}
+		}
 
-	/* check is mst dsc output bandwidth branch_overall_throughput_0_mps */
-	switch (stream->timing.pixel_encoding) {
-	case PIXEL_ENCODING_RGB:
-	case PIXEL_ENCODING_YCBCR444:
-		branch_max_throughput_mps =
-			aconnector->dc_sink->dsc_caps.dsc_dec_caps.branch_overall_throughput_0_mps;
-		break;
-	case PIXEL_ENCODING_YCBCR422:
-	case PIXEL_ENCODING_YCBCR420:
-		branch_max_throughput_mps =
-			aconnector->dc_sink->dsc_caps.dsc_dec_caps.branch_overall_throughput_1_mps;
-		break;
-	default:
-		break;
-	}
+		/* check is mst dsc output bandwidth branch_overall_throughput_0_mps */
+		switch (stream->timing.pixel_encoding) {
+		case PIXEL_ENCODING_RGB:
+		case PIXEL_ENCODING_YCBCR444:
+			branch_max_throughput_mps =
+				aconnector->dc_sink->dsc_caps.dsc_dec_caps.branch_overall_throughput_0_mps;
+			break;
+		case PIXEL_ENCODING_YCBCR422:
+		case PIXEL_ENCODING_YCBCR420:
+			branch_max_throughput_mps =
+				aconnector->dc_sink->dsc_caps.dsc_dec_caps.branch_overall_throughput_1_mps;
+			break;
+		default:
+			break;
+		}
 
-	if (branch_max_throughput_mps != 0 &&
-		((stream->timing.pix_clk_100hz / 10) >  branch_max_throughput_mps * 1000))
+		if (branch_max_throughput_mps != 0 &&
+			((stream->timing.pix_clk_100hz / 10) >  branch_max_throughput_mps * 1000)) {
+			DRM_DEBUG_DRIVER("MST_DSC require dsc but max throughput mps fails\n");
+			return DC_FAIL_BANDWIDTH_VALIDATE;
+		}
+	} else {
+		DRM_DEBUG_DRIVER("MST_DSC require dsc but can't find common dsc config\n");
 		return DC_FAIL_BANDWIDTH_VALIDATE;
-
+	}
+#endif
 	return DC_OK;
 }
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.h
index 1e4ede1e57ab..6f7ea684b555 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.h
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.h
@@ -1,3 +1,4 @@
+/* SPDX-License-Identifier: MIT */
 /*
  * Copyright 2012-15 Advanced Micro Devices, Inc.
  *
@@ -46,13 +47,20 @@
 #define SYNAPTICS_CASCADED_HUB_ID  0x5A
 #define IS_SYNAPTICS_CASCADED_PANAMERA(devName, data) ((IS_SYNAPTICS_PANAMERA(devName) && ((int)data[2] == SYNAPTICS_CASCADED_HUB_ID)) ? 1 : 0)
 
-#define PBN_FEC_OVERHEAD_MULTIPLIER_8B_10B	1031
-#define PBN_FEC_OVERHEAD_MULTIPLIER_128B_132B	1000
+#define PBN_FEC_OVERHEAD_MULTIPLIER_8B_10B     1031
+#define PBN_FEC_OVERHEAD_MULTIPLIER_128B_132B  1000
+
+enum mst_msg_ready_type {
+	NONE_MSG_RDY_EVENT = 0,
+	DOWN_REP_MSG_RDY_EVENT = 1,
+	UP_REQ_MSG_RDY_EVENT = 2,
+	DOWN_OR_UP_MSG_RDY_EVENT = 3
+};
 
 struct amdgpu_display_manager;
 struct amdgpu_dm_connector;
 
-int dm_mst_get_pbn_divider(struct dc_link *link);
+uint32_t dm_mst_get_pbn_divider(struct dc_link *link);
 
 void amdgpu_dm_initialize_dp_connector(struct amdgpu_display_manager *dm,
 				       struct amdgpu_dm_connector *aconnector,
@@ -61,6 +69,10 @@ void amdgpu_dm_initialize_dp_connector(struct amdgpu_display_manager *dm,
 void
 dm_dp_create_fake_mst_encoders(struct amdgpu_device *adev);
 
+void dm_handle_mst_sideband_msg_ready_event(
+	struct drm_dp_mst_topology_mgr *mgr,
+	enum mst_msg_ready_type msg_rdy_type);
+
 struct dsc_mst_fairness_vars {
 	int pbn;
 	bool dsc_enabled;
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c
index 322668973747..2e3ee78999d9 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c
@@ -26,8 +26,10 @@
 
 #include <drm/drm_atomic_helper.h>
 #include <drm/drm_blend.h>
+#include "drm/drm_framebuffer.h"
 #include <drm/drm_gem_atomic_helper.h>
 #include <drm/drm_plane_helper.h>
+#include <drm/drm_gem_framebuffer_helper.h>
 #include <drm/drm_fourcc.h>
 
 #include "amdgpu.h"
@@ -35,6 +37,7 @@
 #include "amdgpu_display.h"
 #include "amdgpu_dm_trace.h"
 #include "amdgpu_dm_plane.h"
+#include "amdgpu_dm_colorop.h"
 #include "gc/gc_11_0_0_offset.h"
 #include "gc/gc_11_0_0_sh_mask.h"
 
@@ -90,9 +93,9 @@ enum dm_micro_swizzle {
 	MICRO_SWIZZLE_R = 3
 };
 
-const struct drm_format_info *amdgpu_dm_plane_get_format_info(const struct drm_mode_fb_cmd2 *cmd)
+const struct drm_format_info *amdgpu_dm_plane_get_format_info(u32 pixel_format, u64 modifier)
 {
-	return amdgpu_lookup_format_info(cmd->pixel_format, cmd->modifier[0]);
+	return amdgpu_lookup_format_info(pixel_format, modifier);
 }
 
 void amdgpu_dm_plane_fill_blending_from_plane_state(const struct drm_plane_state *plane_state,
@@ -104,8 +107,6 @@ void amdgpu_dm_plane_fill_blending_from_plane_state(const struct drm_plane_state
 	*global_alpha = false;
 	*global_alpha_value = 0xff;
 
-	if (plane_state->plane->type != DRM_PLANE_TYPE_OVERLAY)
-		return;
 
 	if (plane_state->pixel_blend_mode == DRM_MODE_BLEND_PREMULTI ||
 		plane_state->pixel_blend_mode == DRM_MODE_BLEND_COVERAGE) {
@@ -113,6 +114,11 @@ void amdgpu_dm_plane_fill_blending_from_plane_state(const struct drm_plane_state
 			DRM_FORMAT_ARGB8888,
 			DRM_FORMAT_RGBA8888,
 			DRM_FORMAT_ABGR8888,
+			DRM_FORMAT_ARGB2101010,
+			DRM_FORMAT_ABGR2101010,
+			DRM_FORMAT_ARGB16161616,
+			DRM_FORMAT_ABGR16161616,
+			DRM_FORMAT_ARGB16161616F,
 		};
 		uint32_t format = plane_state->fb->format->format;
 		unsigned int i;
@@ -134,14 +140,14 @@ void amdgpu_dm_plane_fill_blending_from_plane_state(const struct drm_plane_state
 	}
 }
 
-static void add_modifier(uint64_t **mods, uint64_t *size, uint64_t *cap, uint64_t mod)
+static void amdgpu_dm_plane_add_modifier(uint64_t **mods, uint64_t *size, uint64_t *cap, uint64_t mod)
 {
 	if (!*mods)
 		return;
 
 	if (*cap - *size < 1) {
 		uint64_t new_cap = *cap * 2;
-		uint64_t *new_mods = kmalloc(new_cap * sizeof(uint64_t), GFP_KERNEL);
+		uint64_t *new_mods = kmalloc_array(new_cap, sizeof(uint64_t), GFP_KERNEL);
 
 		if (!new_mods) {
 			kfree(*mods);
@@ -159,12 +165,12 @@ static void add_modifier(uint64_t **mods, uint64_t *size, uint64_t *cap, uint64_
 	*size += 1;
 }
 
-static bool modifier_has_dcc(uint64_t modifier)
+static bool amdgpu_dm_plane_modifier_has_dcc(uint64_t modifier)
 {
 	return IS_AMD_FMT_MOD(modifier) && AMD_FMT_MOD_GET(DCC, modifier);
 }
 
-static unsigned modifier_gfx9_swizzle_mode(uint64_t modifier)
+static unsigned int amdgpu_dm_plane_modifier_gfx9_swizzle_mode(uint64_t modifier)
 {
 	if (modifier == DRM_FORMAT_MOD_LINEAR)
 		return 0;
@@ -172,8 +178,8 @@ static unsigned modifier_gfx9_swizzle_mode(uint64_t modifier)
 	return AMD_FMT_MOD_GET(TILE, modifier);
 }
 
-static void fill_gfx8_tiling_info_from_flags(union dc_tiling_info *tiling_info,
-				 uint64_t tiling_flags)
+static void amdgpu_dm_plane_fill_gfx8_tiling_info_from_flags(struct dc_tiling_info *tiling_info,
+							     uint64_t tiling_flags)
 {
 	/* Fill GFX8 params */
 	if (AMDGPU_TILING_GET(tiling_flags, ARRAY_MODE) == DC_ARRAY_2D_TILED_THIN1) {
@@ -185,6 +191,7 @@ static void fill_gfx8_tiling_info_from_flags(union dc_tiling_info *tiling_info,
 		tile_split = AMDGPU_TILING_GET(tiling_flags, TILE_SPLIT);
 		num_banks = AMDGPU_TILING_GET(tiling_flags, NUM_BANKS);
 
+		tiling_info->gfxversion = DcGfxVersion8;
 		/* XXX fix me for VI */
 		tiling_info->gfx8.num_banks = num_banks;
 		tiling_info->gfx8.array_mode =
@@ -204,8 +211,8 @@ static void fill_gfx8_tiling_info_from_flags(union dc_tiling_info *tiling_info,
 			AMDGPU_TILING_GET(tiling_flags, PIPE_CONFIG);
 }
 
-static void fill_gfx9_tiling_info_from_device(const struct amdgpu_device *adev,
-				  union dc_tiling_info *tiling_info)
+static void amdgpu_dm_plane_fill_gfx9_tiling_info_from_device(const struct amdgpu_device *adev,
+							      struct dc_tiling_info *tiling_info)
 {
 	/* Fill GFX9 params */
 	tiling_info->gfx9.num_pipes =
@@ -221,13 +228,13 @@ static void fill_gfx9_tiling_info_from_device(const struct amdgpu_device *adev,
 	tiling_info->gfx9.num_rb_per_se =
 		adev->gfx.config.gb_addr_config_fields.num_rb_per_se;
 	tiling_info->gfx9.shaderEnable = 1;
-	if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 3, 0))
 		tiling_info->gfx9.num_pkrs = adev->gfx.config.gb_addr_config_fields.num_pkrs;
 }
 
-static void fill_gfx9_tiling_info_from_modifier(const struct amdgpu_device *adev,
-				    union dc_tiling_info *tiling_info,
-				    uint64_t modifier)
+static void amdgpu_dm_plane_fill_gfx9_tiling_info_from_modifier(const struct amdgpu_device *adev,
+								struct dc_tiling_info *tiling_info,
+								uint64_t modifier)
 {
 	unsigned int mod_bank_xor_bits = AMD_FMT_MOD_GET(BANK_XOR_BITS, modifier);
 	unsigned int mod_pipe_xor_bits = AMD_FMT_MOD_GET(PIPE_XOR_BITS, modifier);
@@ -236,7 +243,7 @@ static void fill_gfx9_tiling_info_from_modifier(const struct amdgpu_device *adev
 
 	pipes_log2 = min(5u, mod_pipe_xor_bits);
 
-	fill_gfx9_tiling_info_from_device(adev, tiling_info);
+	amdgpu_dm_plane_fill_gfx9_tiling_info_from_device(adev, tiling_info);
 
 	if (!IS_AMD_FMT_MOD(modifier))
 		return;
@@ -253,13 +260,13 @@ static void fill_gfx9_tiling_info_from_modifier(const struct amdgpu_device *adev
 	}
 }
 
-static int validate_dcc(struct amdgpu_device *adev,
-	     const enum surface_pixel_format format,
-	     const enum dc_rotation_angle rotation,
-	     const union dc_tiling_info *tiling_info,
-	     const struct dc_plane_dcc_param *dcc,
-	     const struct dc_plane_address *address,
-	     const struct plane_size *plane_size)
+static int amdgpu_dm_plane_validate_dcc(struct amdgpu_device *adev,
+					const enum surface_pixel_format format,
+					const enum dc_rotation_angle rotation,
+					const struct dc_tiling_info *tiling_info,
+					const struct dc_plane_dcc_param *dcc,
+					const struct dc_plane_address *address,
+					const struct plane_size *plane_size)
 {
 	struct dc *dc = adev->dm.dc;
 	struct dc_dcc_surface_param input;
@@ -271,8 +278,11 @@ static int validate_dcc(struct amdgpu_device *adev,
 	if (!dcc->enable)
 		return 0;
 
-	if (format >= SURFACE_PIXEL_FORMAT_VIDEO_BEGIN ||
-	    !dc->cap_funcs.get_dcc_compression_cap)
+	if (adev->family < AMDGPU_FAMILY_GC_12_0_0 &&
+	    format >= SURFACE_PIXEL_FORMAT_VIDEO_BEGIN)
+		return -EINVAL;
+
+	if (!dc->cap_funcs.get_dcc_compression_cap)
 		return -EINVAL;
 
 	input.format = format;
@@ -298,23 +308,23 @@ static int validate_dcc(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int fill_gfx9_plane_attributes_from_modifiers(struct amdgpu_device *adev,
-					  const struct amdgpu_framebuffer *afb,
-					  const enum surface_pixel_format format,
-					  const enum dc_rotation_angle rotation,
-					  const struct plane_size *plane_size,
-					  union dc_tiling_info *tiling_info,
-					  struct dc_plane_dcc_param *dcc,
-					  struct dc_plane_address *address,
-					  const bool force_disable_dcc)
+static int amdgpu_dm_plane_fill_gfx9_plane_attributes_from_modifiers(struct amdgpu_device *adev,
+								     const struct amdgpu_framebuffer *afb,
+								     const enum surface_pixel_format format,
+								     const enum dc_rotation_angle rotation,
+								     const struct plane_size *plane_size,
+								     struct dc_tiling_info *tiling_info,
+								     struct dc_plane_dcc_param *dcc,
+								     struct dc_plane_address *address)
 {
 	const uint64_t modifier = afb->base.modifier;
 	int ret = 0;
 
-	fill_gfx9_tiling_info_from_modifier(adev, tiling_info, modifier);
-	tiling_info->gfx9.swizzle = modifier_gfx9_swizzle_mode(modifier);
+	amdgpu_dm_plane_fill_gfx9_tiling_info_from_modifier(adev, tiling_info, modifier);
+	tiling_info->gfx9.swizzle = amdgpu_dm_plane_modifier_gfx9_swizzle_mode(modifier);
+	tiling_info->gfxversion = DcGfxVersion9;
 
-	if (modifier_has_dcc(modifier) && !force_disable_dcc) {
+	if (amdgpu_dm_plane_modifier_has_dcc(modifier)) {
 		uint64_t dcc_address = afb->address + afb->base.offsets[1];
 		bool independent_64b_blks = AMD_FMT_MOD_GET(DCC_INDEPENDENT_64B, modifier);
 		bool independent_128b_blks = AMD_FMT_MOD_GET(DCC_INDEPENDENT_128B, modifier);
@@ -342,60 +352,104 @@ static int fill_gfx9_plane_attributes_from_modifiers(struct amdgpu_device *adev,
 		address->grph.meta_addr.high_part = upper_32_bits(dcc_address);
 	}
 
-	ret = validate_dcc(adev, format, rotation, tiling_info, dcc, address, plane_size);
+	ret = amdgpu_dm_plane_validate_dcc(adev, format, rotation, tiling_info, dcc, address, plane_size);
 	if (ret)
-		drm_dbg_kms(adev_to_drm(adev), "validate_dcc: returned error: %d\n", ret);
+		drm_dbg_kms(adev_to_drm(adev), "amdgpu_dm_plane_validate_dcc: returned error: %d\n", ret);
 
 	return ret;
 }
 
-static void add_gfx10_1_modifiers(const struct amdgpu_device *adev,
-		      uint64_t **mods, uint64_t *size, uint64_t *capacity)
+static int amdgpu_dm_plane_fill_gfx12_plane_attributes_from_modifiers(struct amdgpu_device *adev,
+								      const struct amdgpu_framebuffer *afb,
+								      const enum surface_pixel_format format,
+								      const enum dc_rotation_angle rotation,
+								      const struct plane_size *plane_size,
+								      struct dc_tiling_info *tiling_info,
+								      struct dc_plane_dcc_param *dcc,
+								      struct dc_plane_address *address)
+{
+	const uint64_t modifier = afb->base.modifier;
+	int ret = 0;
+
+	/* TODO: Most of this function shouldn't be needed on GFX12. */
+	amdgpu_dm_plane_fill_gfx9_tiling_info_from_device(adev, tiling_info);
+
+	tiling_info->gfx9.swizzle = amdgpu_dm_plane_modifier_gfx9_swizzle_mode(modifier);
+	tiling_info->gfxversion = DcGfxAddr3;
+
+	if (amdgpu_dm_plane_modifier_has_dcc(modifier)) {
+		int max_compressed_block = AMD_FMT_MOD_GET(DCC_MAX_COMPRESSED_BLOCK, modifier);
+
+		dcc->enable = 1;
+		dcc->independent_64b_blks = max_compressed_block == 0;
+
+		if (max_compressed_block == 0)
+			dcc->dcc_ind_blk = hubp_ind_block_64b;
+		else if (max_compressed_block == 1)
+			dcc->dcc_ind_blk = hubp_ind_block_128b;
+		else
+			dcc->dcc_ind_blk = hubp_ind_block_unconstrained;
+	}
+
+	/* TODO: This seems wrong because there is no DCC plane on GFX12. */
+	ret = amdgpu_dm_plane_validate_dcc(adev, format, rotation, tiling_info, dcc, address, plane_size);
+	if (ret)
+		drm_dbg_kms(adev_to_drm(adev), "amdgpu_dm_plane_validate_dcc: returned error: %d\n", ret);
+
+	return ret;
+}
+
+static void amdgpu_dm_plane_add_gfx10_1_modifiers(const struct amdgpu_device *adev,
+						  uint64_t **mods,
+						  uint64_t *size,
+						  uint64_t *capacity)
 {
 	int pipe_xor_bits = ilog2(adev->gfx.config.gb_addr_config_fields.num_pipes);
 
-	add_modifier(mods, size, capacity, AMD_FMT_MOD |
-		    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_R_X) |
-		    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10) |
-		    AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
-		    AMD_FMT_MOD_SET(DCC, 1) |
-		    AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 1) |
-		    AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 1) |
-		    AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B));
-
-	add_modifier(mods, size, capacity, AMD_FMT_MOD |
-		    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_R_X) |
-		    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10) |
-		    AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
-		    AMD_FMT_MOD_SET(DCC, 1) |
-		    AMD_FMT_MOD_SET(DCC_RETILE, 1) |
-		    AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 1) |
-		    AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 1) |
-		    AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B));
-
-	add_modifier(mods, size, capacity, AMD_FMT_MOD |
-		    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_R_X) |
-		    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10) |
-		    AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits));
-
-	add_modifier(mods, size, capacity, AMD_FMT_MOD |
-		    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S_X) |
-		    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10) |
-		    AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits));
-
-
-	/* Only supported for 64bpp, will be filtered in dm_plane_format_mod_supported */
-	add_modifier(mods, size, capacity, AMD_FMT_MOD |
-		    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_D) |
-		    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9));
-
-	add_modifier(mods, size, capacity, AMD_FMT_MOD |
-		    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S) |
-		    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9));
+	amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+				     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_R_X) |
+				     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10) |
+				     AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
+				     AMD_FMT_MOD_SET(DCC, 1) |
+				     AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 1) |
+				     AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 1) |
+				     AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B));
+
+	amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+				     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_R_X) |
+				     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10) |
+				     AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
+				     AMD_FMT_MOD_SET(DCC, 1) |
+				     AMD_FMT_MOD_SET(DCC_RETILE, 1) |
+				     AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 1) |
+				     AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 1) |
+				     AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B));
+
+	amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+				     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_R_X) |
+				     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10) |
+				     AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits));
+
+	amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+				     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S_X) |
+				     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10) |
+				     AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits));
+
+
+	/* Only supported for 64bpp, will be filtered in amdgpu_dm_plane_format_mod_supported */
+	amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+				     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_D) |
+				     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9));
+
+	amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+				     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S) |
+				     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9));
 }
 
-static void add_gfx9_modifiers(const struct amdgpu_device *adev,
-		   uint64_t **mods, uint64_t *size, uint64_t *capacity)
+static void amdgpu_dm_plane_add_gfx9_modifiers(const struct amdgpu_device *adev,
+					       uint64_t **mods,
+					       uint64_t *size,
+					       uint64_t *capacity)
 {
 	int pipes = ilog2(adev->gfx.config.gb_addr_config_fields.num_pipes);
 	int pipe_xor_bits = min(8, pipes +
@@ -416,163 +470,164 @@ static void add_gfx9_modifiers(const struct amdgpu_device *adev,
 		 */
 
 		if (has_constant_encode) {
-			add_modifier(mods, size, capacity, AMD_FMT_MOD |
-				    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S_X) |
-				    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9) |
-				    AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
-				    AMD_FMT_MOD_SET(BANK_XOR_BITS, bank_xor_bits) |
-				    AMD_FMT_MOD_SET(DCC, 1) |
-				    AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 1) |
-				    AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B) |
-				    AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 1));
+			amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+						     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S_X) |
+						     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9) |
+						     AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
+						     AMD_FMT_MOD_SET(BANK_XOR_BITS, bank_xor_bits) |
+						     AMD_FMT_MOD_SET(DCC, 1) |
+						     AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 1) |
+						     AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B) |
+						     AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 1));
 		}
 
-		add_modifier(mods, size, capacity, AMD_FMT_MOD |
-			    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S_X) |
-			    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9) |
-			    AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
-			    AMD_FMT_MOD_SET(BANK_XOR_BITS, bank_xor_bits) |
-			    AMD_FMT_MOD_SET(DCC, 1) |
-			    AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 1) |
-			    AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B) |
-			    AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 0));
+		amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+					     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S_X) |
+					     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9) |
+					     AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
+					     AMD_FMT_MOD_SET(BANK_XOR_BITS, bank_xor_bits) |
+					     AMD_FMT_MOD_SET(DCC, 1) |
+					     AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 1) |
+					     AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B) |
+					     AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 0));
 
 		if (has_constant_encode) {
-			add_modifier(mods, size, capacity, AMD_FMT_MOD |
-				    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S_X) |
-				    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9) |
-				    AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
-				    AMD_FMT_MOD_SET(BANK_XOR_BITS, bank_xor_bits) |
-				    AMD_FMT_MOD_SET(DCC, 1) |
-				    AMD_FMT_MOD_SET(DCC_RETILE, 1) |
-				    AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 1) |
-				    AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B) |
-
-				    AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 1) |
-				    AMD_FMT_MOD_SET(RB, rb) |
-				    AMD_FMT_MOD_SET(PIPE, pipes));
+			amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+						     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S_X) |
+						     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9) |
+						     AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
+						     AMD_FMT_MOD_SET(BANK_XOR_BITS, bank_xor_bits) |
+						     AMD_FMT_MOD_SET(DCC, 1) |
+						     AMD_FMT_MOD_SET(DCC_RETILE, 1) |
+						     AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 1) |
+						     AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B) |
+						     AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 1) |
+						     AMD_FMT_MOD_SET(RB, rb) |
+						     AMD_FMT_MOD_SET(PIPE, pipes));
 		}
 
-		add_modifier(mods, size, capacity, AMD_FMT_MOD |
-			    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S_X) |
-			    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9) |
-			    AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
-			    AMD_FMT_MOD_SET(BANK_XOR_BITS, bank_xor_bits) |
-			    AMD_FMT_MOD_SET(DCC, 1) |
-			    AMD_FMT_MOD_SET(DCC_RETILE, 1) |
-			    AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 1) |
-			    AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B) |
-			    AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 0) |
-			    AMD_FMT_MOD_SET(RB, rb) |
-			    AMD_FMT_MOD_SET(PIPE, pipes));
+		amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+					     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S_X) |
+					     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9) |
+					     AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
+					     AMD_FMT_MOD_SET(BANK_XOR_BITS, bank_xor_bits) |
+					     AMD_FMT_MOD_SET(DCC, 1) |
+					     AMD_FMT_MOD_SET(DCC_RETILE, 1) |
+					     AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 1) |
+					     AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B) |
+					     AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 0) |
+					     AMD_FMT_MOD_SET(RB, rb) |
+					     AMD_FMT_MOD_SET(PIPE, pipes));
 	}
 
 	/*
 	 * Only supported for 64bpp on Raven, will be filtered on format in
-	 * dm_plane_format_mod_supported.
+	 * amdgpu_dm_plane_format_mod_supported.
 	 */
-	add_modifier(mods, size, capacity, AMD_FMT_MOD |
-		    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_D_X) |
-		    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9) |
-		    AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
-		    AMD_FMT_MOD_SET(BANK_XOR_BITS, bank_xor_bits));
+	amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+				     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_D_X) |
+				     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9) |
+				     AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
+				     AMD_FMT_MOD_SET(BANK_XOR_BITS, bank_xor_bits));
 
 	if (adev->family == AMDGPU_FAMILY_RV) {
-		add_modifier(mods, size, capacity, AMD_FMT_MOD |
-			    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S_X) |
-			    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9) |
-			    AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
-			    AMD_FMT_MOD_SET(BANK_XOR_BITS, bank_xor_bits));
+		amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+					     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S_X) |
+					     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9) |
+					     AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
+					     AMD_FMT_MOD_SET(BANK_XOR_BITS, bank_xor_bits));
 	}
 
 	/*
 	 * Only supported for 64bpp on Raven, will be filtered on format in
-	 * dm_plane_format_mod_supported.
+	 * amdgpu_dm_plane_format_mod_supported.
 	 */
-	add_modifier(mods, size, capacity, AMD_FMT_MOD |
-		    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_D) |
-		    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9));
+	amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+				     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_D) |
+				     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9));
 
 	if (adev->family == AMDGPU_FAMILY_RV) {
-		add_modifier(mods, size, capacity, AMD_FMT_MOD |
-			    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S) |
-			    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9));
+		amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+					     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S) |
+					     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9));
 	}
 }
 
-static void add_gfx10_3_modifiers(const struct amdgpu_device *adev,
-		      uint64_t **mods, uint64_t *size, uint64_t *capacity)
+static void amdgpu_dm_plane_add_gfx10_3_modifiers(const struct amdgpu_device *adev,
+						  uint64_t **mods,
+						  uint64_t *size,
+						  uint64_t *capacity)
 {
 	int pipe_xor_bits = ilog2(adev->gfx.config.gb_addr_config_fields.num_pipes);
 	int pkrs = ilog2(adev->gfx.config.gb_addr_config_fields.num_pkrs);
 
-	add_modifier(mods, size, capacity, AMD_FMT_MOD |
-		    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_R_X) |
-		    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS) |
-		    AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
-		    AMD_FMT_MOD_SET(PACKERS, pkrs) |
-		    AMD_FMT_MOD_SET(DCC, 1) |
-		    AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 1) |
-		    AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 1) |
-		    AMD_FMT_MOD_SET(DCC_INDEPENDENT_128B, 1) |
-		    AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B));
-
-	add_modifier(mods, size, capacity, AMD_FMT_MOD |
-		    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_R_X) |
-		    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS) |
-		    AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
-		    AMD_FMT_MOD_SET(PACKERS, pkrs) |
-		    AMD_FMT_MOD_SET(DCC, 1) |
-		    AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 1) |
-		    AMD_FMT_MOD_SET(DCC_INDEPENDENT_128B, 1) |
-		    AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_128B));
-
-	add_modifier(mods, size, capacity, AMD_FMT_MOD |
-		    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_R_X) |
-		    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS) |
-		    AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
-		    AMD_FMT_MOD_SET(PACKERS, pkrs) |
-		    AMD_FMT_MOD_SET(DCC, 1) |
-		    AMD_FMT_MOD_SET(DCC_RETILE, 1) |
-		    AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 1) |
-		    AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 1) |
-		    AMD_FMT_MOD_SET(DCC_INDEPENDENT_128B, 1) |
-		    AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B));
-
-	add_modifier(mods, size, capacity, AMD_FMT_MOD |
-		    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_R_X) |
-		    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS) |
-		    AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
-		    AMD_FMT_MOD_SET(PACKERS, pkrs) |
-		    AMD_FMT_MOD_SET(DCC, 1) |
-		    AMD_FMT_MOD_SET(DCC_RETILE, 1) |
-		    AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 1) |
-		    AMD_FMT_MOD_SET(DCC_INDEPENDENT_128B, 1) |
-		    AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_128B));
-
-	add_modifier(mods, size, capacity, AMD_FMT_MOD |
-		    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_R_X) |
-		    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS) |
-		    AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
-		    AMD_FMT_MOD_SET(PACKERS, pkrs));
-
-	add_modifier(mods, size, capacity, AMD_FMT_MOD |
-		    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S_X) |
-		    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS) |
-		    AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
-		    AMD_FMT_MOD_SET(PACKERS, pkrs));
-
-	/* Only supported for 64bpp, will be filtered in dm_plane_format_mod_supported */
-	add_modifier(mods, size, capacity, AMD_FMT_MOD |
-		    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_D) |
-		    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9));
-
-	add_modifier(mods, size, capacity, AMD_FMT_MOD |
-		    AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S) |
-		    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9));
+	amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+				     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_R_X) |
+				     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS) |
+				     AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
+				     AMD_FMT_MOD_SET(PACKERS, pkrs) |
+				     AMD_FMT_MOD_SET(DCC, 1) |
+				     AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 1) |
+				     AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 1) |
+				     AMD_FMT_MOD_SET(DCC_INDEPENDENT_128B, 1) |
+				     AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B));
+
+	amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+				     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_R_X) |
+				     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS) |
+				     AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
+				     AMD_FMT_MOD_SET(PACKERS, pkrs) |
+				     AMD_FMT_MOD_SET(DCC, 1) |
+				     AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 1) |
+				     AMD_FMT_MOD_SET(DCC_INDEPENDENT_128B, 1) |
+				     AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_128B));
+
+	amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+				     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_R_X) |
+				     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS) |
+				     AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
+				     AMD_FMT_MOD_SET(PACKERS, pkrs) |
+				     AMD_FMT_MOD_SET(DCC, 1) |
+				     AMD_FMT_MOD_SET(DCC_RETILE, 1) |
+				     AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 1) |
+				     AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 1) |
+				     AMD_FMT_MOD_SET(DCC_INDEPENDENT_128B, 1) |
+				     AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B));
+
+	amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+				     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_R_X) |
+				     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS) |
+				     AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
+				     AMD_FMT_MOD_SET(PACKERS, pkrs) |
+				     AMD_FMT_MOD_SET(DCC, 1) |
+				     AMD_FMT_MOD_SET(DCC_RETILE, 1) |
+				     AMD_FMT_MOD_SET(DCC_CONSTANT_ENCODE, 1) |
+				     AMD_FMT_MOD_SET(DCC_INDEPENDENT_128B, 1) |
+				     AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_128B));
+
+	amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+				     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_R_X) |
+				     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS) |
+				     AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
+				     AMD_FMT_MOD_SET(PACKERS, pkrs));
+
+	amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+				     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S_X) |
+				     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS) |
+				     AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
+				     AMD_FMT_MOD_SET(PACKERS, pkrs));
+
+	/* Only supported for 64bpp, will be filtered in amdgpu_dm_plane_format_mod_supported */
+	amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+				     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_D) |
+				     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9));
+
+	amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+				     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_S) |
+				     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9));
 }
 
-static void add_gfx11_modifiers(struct amdgpu_device *adev,
+static void amdgpu_dm_plane_add_gfx11_modifiers(struct amdgpu_device *adev,
 		      uint64_t **mods, uint64_t *size, uint64_t *capacity)
 {
 	int num_pipes = 0;
@@ -581,7 +636,7 @@ static void add_gfx11_modifiers(struct amdgpu_device *adev,
 	int pkrs = 0;
 	u32 gb_addr_config;
 	u8 i = 0;
-	unsigned swizzle_r_x;
+	unsigned int swizzle_r_x;
 	uint64_t modifier_r_x;
 	uint64_t modifier_dcc_best;
 	uint64_t modifier_dcc_4k;
@@ -623,21 +678,53 @@ static void add_gfx11_modifiers(struct amdgpu_device *adev,
 				  AMD_FMT_MOD_SET(DCC_INDEPENDENT_128B, 1) |
 				  AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B);
 
-		add_modifier(mods, size, capacity, modifier_dcc_best);
-		add_modifier(mods, size, capacity, modifier_dcc_4k);
+		amdgpu_dm_plane_add_modifier(mods, size, capacity, modifier_dcc_best);
+		amdgpu_dm_plane_add_modifier(mods, size, capacity, modifier_dcc_4k);
 
-		add_modifier(mods, size, capacity, modifier_dcc_best | AMD_FMT_MOD_SET(DCC_RETILE, 1));
-		add_modifier(mods, size, capacity, modifier_dcc_4k | AMD_FMT_MOD_SET(DCC_RETILE, 1));
+		amdgpu_dm_plane_add_modifier(mods, size, capacity, modifier_dcc_best | AMD_FMT_MOD_SET(DCC_RETILE, 1));
+		amdgpu_dm_plane_add_modifier(mods, size, capacity, modifier_dcc_4k | AMD_FMT_MOD_SET(DCC_RETILE, 1));
 
-		add_modifier(mods, size, capacity, modifier_r_x);
+		amdgpu_dm_plane_add_modifier(mods, size, capacity, modifier_r_x);
 	}
 
-	add_modifier(mods, size, capacity, AMD_FMT_MOD |
-			AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX11) |
-			AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_D));
+	amdgpu_dm_plane_add_modifier(mods, size, capacity, AMD_FMT_MOD |
+				     AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX11) |
+				     AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_D));
+}
+
+static void amdgpu_dm_plane_add_gfx12_modifiers(struct amdgpu_device *adev,
+		      uint64_t **mods, uint64_t *size, uint64_t *capacity)
+{
+	uint64_t ver = AMD_FMT_MOD | AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX12);
+	uint64_t mod_256k = ver | AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX12_256K_2D);
+	uint64_t mod_64k = ver | AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX12_64K_2D);
+	uint64_t mod_4k = ver | AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX12_4K_2D);
+	uint64_t mod_256b = ver | AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX12_256B_2D);
+	uint64_t dcc = ver | AMD_FMT_MOD_SET(DCC, 1);
+	uint8_t max_comp_block[] = {2, 1, 0};
+	uint64_t max_comp_block_mod[ARRAY_SIZE(max_comp_block)] = {0};
+	uint8_t i = 0, j = 0;
+	uint64_t gfx12_modifiers[] = {mod_256k, mod_64k, mod_4k, mod_256b, DRM_FORMAT_MOD_LINEAR};
+
+	for (i = 0; i < ARRAY_SIZE(max_comp_block); i++)
+		max_comp_block_mod[i] = AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, max_comp_block[i]);
+
+	/* With DCC: Best choice should be kept first. Hence, add all 256k modifiers of different
+	 * max compressed blocks first and then move on to the next smaller sized layouts.
+	 * Do not add the linear modifier here, and hence the condition of size-1 for the loop
+	 */
+	for (j = 0; j < ARRAY_SIZE(gfx12_modifiers) - 1; j++)
+		for (i = 0; i < ARRAY_SIZE(max_comp_block); i++)
+			amdgpu_dm_plane_add_modifier(mods, size, capacity,
+						     ver | dcc | max_comp_block_mod[i] | gfx12_modifiers[j]);
+
+	/* Without DCC. Add all modifiers including linear at the end */
+	for (i = 0; i < ARRAY_SIZE(gfx12_modifiers); i++)
+		amdgpu_dm_plane_add_modifier(mods, size, capacity, gfx12_modifiers[i]);
+
 }
 
-static int get_plane_modifiers(struct amdgpu_device *adev, unsigned int plane_type, uint64_t **mods)
+static int amdgpu_dm_plane_get_plane_modifiers(struct amdgpu_device *adev, unsigned int plane_type, uint64_t **mods)
 {
 	uint64_t size = 0, capacity = 128;
 	*mods = NULL;
@@ -646,39 +733,43 @@ static int get_plane_modifiers(struct amdgpu_device *adev, unsigned int plane_ty
 	if (adev->family < AMDGPU_FAMILY_AI)
 		return 0;
 
-	*mods = kmalloc(capacity * sizeof(uint64_t), GFP_KERNEL);
+	*mods = kmalloc_array(capacity, sizeof(uint64_t), GFP_KERNEL);
 
 	if (plane_type == DRM_PLANE_TYPE_CURSOR) {
-		add_modifier(mods, &size, &capacity, DRM_FORMAT_MOD_LINEAR);
-		add_modifier(mods, &size, &capacity, DRM_FORMAT_MOD_INVALID);
+		amdgpu_dm_plane_add_modifier(mods, &size, &capacity, DRM_FORMAT_MOD_LINEAR);
+		amdgpu_dm_plane_add_modifier(mods, &size, &capacity, DRM_FORMAT_MOD_INVALID);
 		return *mods ? 0 : -ENOMEM;
 	}
 
 	switch (adev->family) {
 	case AMDGPU_FAMILY_AI:
 	case AMDGPU_FAMILY_RV:
-		add_gfx9_modifiers(adev, mods, &size, &capacity);
+		amdgpu_dm_plane_add_gfx9_modifiers(adev, mods, &size, &capacity);
 		break;
 	case AMDGPU_FAMILY_NV:
 	case AMDGPU_FAMILY_VGH:
 	case AMDGPU_FAMILY_YC:
 	case AMDGPU_FAMILY_GC_10_3_6:
 	case AMDGPU_FAMILY_GC_10_3_7:
-		if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0))
-			add_gfx10_3_modifiers(adev, mods, &size, &capacity);
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 3, 0))
+			amdgpu_dm_plane_add_gfx10_3_modifiers(adev, mods, &size, &capacity);
 		else
-			add_gfx10_1_modifiers(adev, mods, &size, &capacity);
+			amdgpu_dm_plane_add_gfx10_1_modifiers(adev, mods, &size, &capacity);
 		break;
 	case AMDGPU_FAMILY_GC_11_0_0:
 	case AMDGPU_FAMILY_GC_11_0_1:
-		add_gfx11_modifiers(adev, mods, &size, &capacity);
+	case AMDGPU_FAMILY_GC_11_5_0:
+		amdgpu_dm_plane_add_gfx11_modifiers(adev, mods, &size, &capacity);
+		break;
+	case AMDGPU_FAMILY_GC_12_0_0:
+		amdgpu_dm_plane_add_gfx12_modifiers(adev, mods, &size, &capacity);
 		break;
 	}
 
-	add_modifier(mods, &size, &capacity, DRM_FORMAT_MOD_LINEAR);
+	amdgpu_dm_plane_add_modifier(mods, &size, &capacity, DRM_FORMAT_MOD_LINEAR);
 
 	/* INVALID marks the end of the list. */
-	add_modifier(mods, &size, &capacity, DRM_FORMAT_MOD_INVALID);
+	amdgpu_dm_plane_add_modifier(mods, &size, &capacity, DRM_FORMAT_MOD_INVALID);
 
 	if (!*mods)
 		return -ENOMEM;
@@ -686,9 +777,9 @@ static int get_plane_modifiers(struct amdgpu_device *adev, unsigned int plane_ty
 	return 0;
 }
 
-static int get_plane_formats(const struct drm_plane *plane,
-			     const struct dc_plane_cap *plane_cap,
-			     uint32_t *formats, int max_formats)
+static int amdgpu_dm_plane_get_plane_formats(const struct drm_plane *plane,
+					     const struct dc_plane_cap *plane_cap,
+					     uint32_t *formats, int max_formats)
 {
 	int i, num_formats = 0;
 
@@ -698,8 +789,8 @@ static int get_plane_formats(const struct drm_plane *plane,
 	 * caps list.
 	 */
 
-	switch (plane->type) {
-	case DRM_PLANE_TYPE_PRIMARY:
+	if (plane->type == DRM_PLANE_TYPE_PRIMARY ||
+		(plane_cap && plane_cap->type == DC_PLANE_TYPE_DCN_UNIVERSAL && plane->type != DRM_PLANE_TYPE_CURSOR)) {
 		for (i = 0; i < ARRAY_SIZE(rgb_formats); ++i) {
 			if (num_formats >= max_formats)
 				break;
@@ -717,25 +808,29 @@ static int get_plane_formats(const struct drm_plane *plane,
 			formats[num_formats++] = DRM_FORMAT_XBGR16161616F;
 			formats[num_formats++] = DRM_FORMAT_ABGR16161616F;
 		}
-		break;
+	} else {
+		switch (plane->type) {
+		case DRM_PLANE_TYPE_OVERLAY:
+			for (i = 0; i < ARRAY_SIZE(overlay_formats); ++i) {
+				if (num_formats >= max_formats)
+					break;
 
-	case DRM_PLANE_TYPE_OVERLAY:
-		for (i = 0; i < ARRAY_SIZE(overlay_formats); ++i) {
-			if (num_formats >= max_formats)
-				break;
+				formats[num_formats++] = overlay_formats[i];
+			}
+			break;
 
-			formats[num_formats++] = overlay_formats[i];
-		}
-		break;
+		case DRM_PLANE_TYPE_CURSOR:
+			for (i = 0; i < ARRAY_SIZE(cursor_formats); ++i) {
+				if (num_formats >= max_formats)
+					break;
 
-	case DRM_PLANE_TYPE_CURSOR:
-		for (i = 0; i < ARRAY_SIZE(cursor_formats); ++i) {
-			if (num_formats >= max_formats)
-				break;
+				formats[num_formats++] = cursor_formats[i];
+			}
+			break;
 
-			formats[num_formats++] = cursor_formats[i];
+		default:
+			break;
 		}
-		break;
 	}
 
 	return num_formats;
@@ -746,12 +841,11 @@ int amdgpu_dm_plane_fill_plane_buffer_attributes(struct amdgpu_device *adev,
 			     const enum surface_pixel_format format,
 			     const enum dc_rotation_angle rotation,
 			     const uint64_t tiling_flags,
-			     union dc_tiling_info *tiling_info,
+			     struct dc_tiling_info *tiling_info,
 			     struct plane_size *plane_size,
 			     struct dc_plane_dcc_param *dcc,
 			     struct dc_plane_address *address,
-			     bool tmz_surface,
-			     bool force_disable_dcc)
+			     bool tmz_surface)
 {
 	const struct drm_framebuffer *fb = &afb->base;
 	int ret;
@@ -807,23 +901,29 @@ int amdgpu_dm_plane_fill_plane_buffer_attributes(struct amdgpu_device *adev,
 			upper_32_bits(chroma_addr);
 	}
 
-	if (adev->family >= AMDGPU_FAMILY_AI) {
-		ret = fill_gfx9_plane_attributes_from_modifiers(adev, afb, format,
-								rotation, plane_size,
-								tiling_info, dcc,
-								address,
-								force_disable_dcc);
+	if (adev->family >= AMDGPU_FAMILY_GC_12_0_0) {
+		ret = amdgpu_dm_plane_fill_gfx12_plane_attributes_from_modifiers(adev, afb, format,
+										 rotation, plane_size,
+										 tiling_info, dcc,
+										 address);
+		if (ret)
+			return ret;
+	} else if (adev->family >= AMDGPU_FAMILY_AI) {
+		ret = amdgpu_dm_plane_fill_gfx9_plane_attributes_from_modifiers(adev, afb, format,
+										rotation, plane_size,
+										tiling_info, dcc,
+										address);
 		if (ret)
 			return ret;
 	} else {
-		fill_gfx8_tiling_info_from_flags(tiling_info, tiling_flags);
+		amdgpu_dm_plane_fill_gfx8_tiling_info_from_flags(tiling_info, tiling_flags);
 	}
 
 	return 0;
 }
 
-static int dm_plane_helper_prepare_fb(struct drm_plane *plane,
-				      struct drm_plane_state *new_state)
+static int amdgpu_dm_plane_helper_prepare_fb(struct drm_plane *plane,
+					     struct drm_plane_state *new_state)
 {
 	struct amdgpu_framebuffer *afb;
 	struct drm_gem_object *obj;
@@ -839,19 +939,23 @@ static int dm_plane_helper_prepare_fb(struct drm_plane *plane,
 	}
 
 	afb = to_amdgpu_framebuffer(new_state->fb);
-	obj = new_state->fb->obj[0];
+	obj = drm_gem_fb_get_obj(new_state->fb, 0);
+	if (!obj) {
+		DRM_ERROR("Failed to get obj from framebuffer\n");
+		return -EINVAL;
+	}
+
 	rbo = gem_to_amdgpu_bo(obj);
 	adev = amdgpu_ttm_adev(rbo->tbo.bdev);
-
 	r = amdgpu_bo_reserve(rbo, true);
 	if (r) {
-		dev_err(adev->dev, "fail to reserve bo (%d)\n", r);
+		drm_err(adev_to_drm(adev), "fail to reserve bo (%d)\n", r);
 		return r;
 	}
 
 	r = dma_resv_reserve_fences(rbo->tbo.base.resv, 1);
 	if (r) {
-		dev_err(adev->dev, "reserving fence slot failed (%d)\n", r);
+		drm_err(adev_to_drm(adev), "reserving fence slot failed (%d)\n", r);
 		goto error_unlock;
 	}
 
@@ -860,6 +964,7 @@ static int dm_plane_helper_prepare_fb(struct drm_plane *plane,
 	else
 		domain = AMDGPU_GEM_DOMAIN_VRAM;
 
+	rbo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 	r = amdgpu_bo_pin(rbo, domain);
 	if (unlikely(r != 0)) {
 		if (r != -ERESTARTSYS)
@@ -898,14 +1003,13 @@ static int dm_plane_helper_prepare_fb(struct drm_plane *plane,
 	    dm_plane_state_old->dc_state != dm_plane_state_new->dc_state) {
 		struct dc_plane_state *plane_state =
 			dm_plane_state_new->dc_state;
-		bool force_disable_dcc = !plane_state->dcc.enable;
 
 		amdgpu_dm_plane_fill_plane_buffer_attributes(
 			adev, afb, plane_state->format, plane_state->rotation,
 			afb->tiling_flags,
 			&plane_state->tiling_info, &plane_state->plane_size,
 			&plane_state->dcc, &plane_state->address,
-			afb->tmz_surface, force_disable_dcc);
+			afb->tmz_surface);
 	}
 
 	return 0;
@@ -918,8 +1022,8 @@ error_unlock:
 	return r;
 }
 
-static void dm_plane_helper_cleanup_fb(struct drm_plane *plane,
-				       struct drm_plane_state *old_state)
+static void amdgpu_dm_plane_helper_cleanup_fb(struct drm_plane *plane,
+					      struct drm_plane_state *old_state)
 {
 	struct amdgpu_bo *rbo;
 	int r;
@@ -939,7 +1043,7 @@ static void dm_plane_helper_cleanup_fb(struct drm_plane *plane,
 	amdgpu_bo_unref(&rbo);
 }
 
-static void get_min_max_dc_plane_scaling(struct drm_device *dev,
+static void amdgpu_dm_plane_get_min_max_dc_plane_scaling(struct drm_device *dev,
 					 struct drm_framebuffer *fb,
 					 int *min_downscale, int *max_upscale)
 {
@@ -1020,8 +1124,8 @@ int amdgpu_dm_plane_helper_check_state(struct drm_plane_state *state,
 		}
 
 		/* Get min/max allowed scaling factors from plane caps. */
-		get_min_max_dc_plane_scaling(state->crtc->dev, fb,
-					     &min_downscale, &max_upscale);
+		amdgpu_dm_plane_get_min_max_dc_plane_scaling(state->crtc->dev, fb,
+							     &min_downscale, &max_upscale);
 		/*
 		 * Convert to drm convention: 16.16 fixed point, instead of dc's
 		 * 1.0 == 1000. Also drm scaling is src/dst instead of dc's
@@ -1059,8 +1163,8 @@ int amdgpu_dm_plane_fill_dc_scaling_info(struct amdgpu_device *adev,
 	 * is to gesture the YouTube Android app into full screen
 	 * on ChromeOS.
 	 */
-	if (((adev->ip_versions[DCE_HWIP][0] == IP_VERSION(1, 0, 0)) ||
-	    (adev->ip_versions[DCE_HWIP][0] == IP_VERSION(1, 0, 1))) &&
+	if (((amdgpu_ip_version(adev, DCE_HWIP, 0) == IP_VERSION(1, 0, 0)) ||
+	    (amdgpu_ip_version(adev, DCE_HWIP, 0) == IP_VERSION(1, 0, 1))) &&
 	    (state->fb && state->fb->format->format == DRM_FORMAT_NV12 &&
 	    (scaling_info->src_rect.x != 0 || scaling_info->src_rect.y != 0)))
 		return -EINVAL;
@@ -1091,8 +1195,8 @@ int amdgpu_dm_plane_fill_dc_scaling_info(struct amdgpu_device *adev,
 
 	/* Validate scaling per-format with DC plane caps */
 	if (state->plane && state->plane->dev && state->fb) {
-		get_min_max_dc_plane_scaling(state->plane->dev, state->fb,
-					     &min_downscale, &max_upscale);
+		amdgpu_dm_plane_get_min_max_dc_plane_scaling(state->plane->dev, state->fb,
+							     &min_downscale, &max_upscale);
 	} else {
 		min_downscale = 250;
 		max_upscale = 16000;
@@ -1118,8 +1222,8 @@ int amdgpu_dm_plane_fill_dc_scaling_info(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int dm_plane_atomic_check(struct drm_plane *plane,
-				 struct drm_atomic_state *state)
+static int amdgpu_dm_plane_atomic_check(struct drm_plane *plane,
+					struct drm_atomic_state *state)
 {
 	struct drm_plane_state *new_plane_state = drm_atomic_get_new_plane_state(state,
 										 plane);
@@ -1157,20 +1261,35 @@ static int dm_plane_atomic_check(struct drm_plane *plane,
 	return -EINVAL;
 }
 
-static int dm_plane_atomic_async_check(struct drm_plane *plane,
-				       struct drm_atomic_state *state)
+static int amdgpu_dm_plane_atomic_async_check(struct drm_plane *plane,
+					      struct drm_atomic_state *state, bool flip)
 {
-	/* Only support async updates on cursor planes. */
-	if (plane->type != DRM_PLANE_TYPE_CURSOR)
+	struct drm_crtc_state *new_crtc_state;
+	struct drm_plane_state *new_plane_state;
+	struct dm_crtc_state *dm_new_crtc_state;
+
+	if (flip) {
+		if (plane->type != DRM_PLANE_TYPE_OVERLAY)
+			return -EINVAL;
+	} else if (plane->type != DRM_PLANE_TYPE_CURSOR) {
+		return -EINVAL;
+	}
+
+	new_plane_state = drm_atomic_get_new_plane_state(state, plane);
+	new_crtc_state = drm_atomic_get_new_crtc_state(state, new_plane_state->crtc);
+	dm_new_crtc_state = to_dm_crtc_state(new_crtc_state);
+	/* Reject overlay cursors for now*/
+	if (!flip && dm_new_crtc_state->cursor_mode == DM_CURSOR_OVERLAY_MODE)
 		return -EINVAL;
 
 	return 0;
 }
 
-static int get_cursor_position(struct drm_plane *plane, struct drm_crtc *crtc,
-			       struct dc_cursor_position *position)
+int amdgpu_dm_plane_get_cursor_position(struct drm_plane *plane, struct drm_crtc *crtc,
+					struct dc_cursor_position *position)
 {
 	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+	struct amdgpu_device *adev = drm_to_adev(plane->dev);
 	int x, y;
 	int xorigin = 0, yorigin = 0;
 
@@ -1202,12 +1321,14 @@ static int get_cursor_position(struct drm_plane *plane, struct drm_crtc *crtc,
 		y = 0;
 	}
 	position->enable = true;
-	position->translate_by_source = true;
 	position->x = x;
 	position->y = y;
 	position->x_hotspot = xorigin;
 	position->y_hotspot = yorigin;
 
+	if (amdgpu_ip_version(adev, DCE_HWIP, 0) < IP_VERSION(4, 0, 1))
+		position->translate_by_source = true;
+
 	return 0;
 }
 
@@ -1227,13 +1348,11 @@ void amdgpu_dm_plane_handle_cursor_update(struct drm_plane *plane,
 	if (!plane->state->fb && !old_plane_state->fb)
 		return;
 
-	DC_LOG_CURSOR("%s: crtc_id=%d with size %d to %d\n",
-		      __func__,
-		      amdgpu_crtc->crtc_id,
-		      plane->state->crtc_w,
-		      plane->state->crtc_h);
+	drm_dbg_atomic(plane->dev, "crtc_id=%d with size %d to %d\n",
+		       amdgpu_crtc->crtc_id, plane->state->crtc_w,
+		       plane->state->crtc_h);
 
-	ret = get_cursor_position(plane, crtc, &position);
+	ret = amdgpu_dm_plane_get_cursor_position(plane, crtc, &position);
 	if (ret)
 		return;
 
@@ -1241,7 +1360,7 @@ void amdgpu_dm_plane_handle_cursor_update(struct drm_plane *plane,
 		/* turn off cursor */
 		if (crtc_state && crtc_state->stream) {
 			mutex_lock(&adev->dm.dc_lock);
-			dc_stream_set_cursor_position(crtc_state->stream,
+			dc_stream_program_cursor_position(crtc_state->stream,
 						      &position);
 			mutex_unlock(&adev->dm.dc_lock);
 		}
@@ -1260,23 +1379,31 @@ void amdgpu_dm_plane_handle_cursor_update(struct drm_plane *plane,
 	attributes.rotation_angle    = 0;
 	attributes.attribute_flags.value = 0;
 
-	attributes.pitch = afb->base.pitches[0] / afb->base.format->cpp[0];
+	/* Enable cursor degamma ROM on DCN3+ for implicit sRGB degamma in DRM
+	 * legacy gamma setup.
+	 */
+	if (crtc_state->cm_is_degamma_srgb &&
+	    adev->dm.dc->caps.color.dpp.gamma_corr)
+		attributes.attribute_flags.bits.ENABLE_CURSOR_DEGAMMA = 1;
+
+	if (afb)
+		attributes.pitch = afb->base.pitches[0] / afb->base.format->cpp[0];
 
 	if (crtc_state->stream) {
 		mutex_lock(&adev->dm.dc_lock);
-		if (!dc_stream_set_cursor_attributes(crtc_state->stream,
+		if (!dc_stream_program_cursor_attributes(crtc_state->stream,
 							 &attributes))
 			DRM_ERROR("DC failed to set cursor attributes\n");
 
-		if (!dc_stream_set_cursor_position(crtc_state->stream,
+		if (!dc_stream_program_cursor_position(crtc_state->stream,
 						   &position))
 			DRM_ERROR("DC failed to set cursor position\n");
 		mutex_unlock(&adev->dm.dc_lock);
 	}
 }
 
-static void dm_plane_atomic_async_update(struct drm_plane *plane,
-					 struct drm_atomic_state *state)
+static void amdgpu_dm_plane_atomic_async_update(struct drm_plane *plane,
+						struct drm_atomic_state *state)
 {
 	struct drm_plane_state *new_state = drm_atomic_get_new_plane_state(state,
 									   plane);
@@ -1299,15 +1426,39 @@ static void dm_plane_atomic_async_update(struct drm_plane *plane,
 	amdgpu_dm_plane_handle_cursor_update(plane, old_state);
 }
 
+static void amdgpu_dm_plane_panic_flush(struct drm_plane *plane)
+{
+	struct dm_plane_state *dm_plane_state = to_dm_plane_state(plane->state);
+	struct drm_framebuffer *fb = plane->state->fb;
+	struct dc_plane_state *dc_plane_state;
+
+	if (!dm_plane_state || !dm_plane_state->dc_state)
+		return;
+
+	dc_plane_state = dm_plane_state->dc_state;
+
+	dc_plane_force_dcc_and_tiling_disable(dc_plane_state, fb->modifier ? true : false);
+}
+
 static const struct drm_plane_helper_funcs dm_plane_helper_funcs = {
-	.prepare_fb = dm_plane_helper_prepare_fb,
-	.cleanup_fb = dm_plane_helper_cleanup_fb,
-	.atomic_check = dm_plane_atomic_check,
-	.atomic_async_check = dm_plane_atomic_async_check,
-	.atomic_async_update = dm_plane_atomic_async_update
+	.prepare_fb = amdgpu_dm_plane_helper_prepare_fb,
+	.cleanup_fb = amdgpu_dm_plane_helper_cleanup_fb,
+	.atomic_check = amdgpu_dm_plane_atomic_check,
+	.atomic_async_check = amdgpu_dm_plane_atomic_async_check,
+	.atomic_async_update = amdgpu_dm_plane_atomic_async_update
+};
+
+static const struct drm_plane_helper_funcs dm_primary_plane_helper_funcs = {
+	.prepare_fb = amdgpu_dm_plane_helper_prepare_fb,
+	.cleanup_fb = amdgpu_dm_plane_helper_cleanup_fb,
+	.atomic_check = amdgpu_dm_plane_atomic_check,
+	.atomic_async_check = amdgpu_dm_plane_atomic_async_check,
+	.atomic_async_update = amdgpu_dm_plane_atomic_async_update,
+	.get_scanout_buffer = amdgpu_display_get_scanout_buffer,
+	.panic_flush = amdgpu_dm_plane_panic_flush,
 };
 
-static void dm_drm_plane_reset(struct drm_plane *plane)
+static void amdgpu_dm_plane_drm_plane_reset(struct drm_plane *plane)
 {
 	struct dm_plane_state *amdgpu_state = NULL;
 
@@ -1317,12 +1468,17 @@ static void dm_drm_plane_reset(struct drm_plane *plane)
 	amdgpu_state = kzalloc(sizeof(*amdgpu_state), GFP_KERNEL);
 	WARN_ON(amdgpu_state == NULL);
 
-	if (amdgpu_state)
-		__drm_atomic_helper_plane_reset(plane, &amdgpu_state->base);
+	if (!amdgpu_state)
+		return;
+
+	__drm_atomic_helper_plane_reset(plane, &amdgpu_state->base);
+	amdgpu_state->degamma_tf = AMDGPU_TRANSFER_FUNCTION_DEFAULT;
+	amdgpu_state->hdr_mult = AMDGPU_HDR_MULT_DEFAULT;
+	amdgpu_state->shaper_tf = AMDGPU_TRANSFER_FUNCTION_DEFAULT;
+	amdgpu_state->blend_tf = AMDGPU_TRANSFER_FUNCTION_DEFAULT;
 }
 
-static struct drm_plane_state *
-dm_drm_plane_duplicate_state(struct drm_plane *plane)
+static struct drm_plane_state *amdgpu_dm_plane_drm_plane_duplicate_state(struct drm_plane *plane)
 {
 	struct dm_plane_state *dm_plane_state, *old_dm_plane_state;
 
@@ -1338,19 +1494,38 @@ dm_drm_plane_duplicate_state(struct drm_plane *plane)
 		dc_plane_state_retain(dm_plane_state->dc_state);
 	}
 
+	if (old_dm_plane_state->degamma_lut)
+		dm_plane_state->degamma_lut =
+			drm_property_blob_get(old_dm_plane_state->degamma_lut);
+	if (old_dm_plane_state->ctm)
+		dm_plane_state->ctm =
+			drm_property_blob_get(old_dm_plane_state->ctm);
+	if (old_dm_plane_state->shaper_lut)
+		dm_plane_state->shaper_lut =
+			drm_property_blob_get(old_dm_plane_state->shaper_lut);
+	if (old_dm_plane_state->lut3d)
+		dm_plane_state->lut3d =
+			drm_property_blob_get(old_dm_plane_state->lut3d);
+	if (old_dm_plane_state->blend_lut)
+		dm_plane_state->blend_lut =
+			drm_property_blob_get(old_dm_plane_state->blend_lut);
+
+	dm_plane_state->degamma_tf = old_dm_plane_state->degamma_tf;
+	dm_plane_state->hdr_mult = old_dm_plane_state->hdr_mult;
+	dm_plane_state->shaper_tf = old_dm_plane_state->shaper_tf;
+	dm_plane_state->blend_tf = old_dm_plane_state->blend_tf;
+
 	return &dm_plane_state->base;
 }
 
-static bool dm_plane_format_mod_supported(struct drm_plane *plane,
-					  uint32_t format,
-					  uint64_t modifier)
+static bool amdgpu_dm_plane_format_mod_supported(struct drm_plane *plane,
+						 uint32_t format,
+						 uint64_t modifier)
 {
 	struct amdgpu_device *adev = drm_to_adev(plane->dev);
 	const struct drm_format_info *info = drm_format_info(format);
 	int i;
 
-	enum dm_micro_swizzle microtile = modifier_gfx9_swizzle_mode(modifier) & 3;
-
 	if (!info)
 		return false;
 
@@ -1372,53 +1547,289 @@ static bool dm_plane_format_mod_supported(struct drm_plane *plane,
 	if (i == plane->modifier_count)
 		return false;
 
-	/*
-	 * For D swizzle the canonical modifier depends on the bpp, so check
-	 * it here.
-	 */
-	if (AMD_FMT_MOD_GET(TILE_VERSION, modifier) == AMD_FMT_MOD_TILE_VER_GFX9 &&
-	    adev->family >= AMDGPU_FAMILY_NV) {
-		if (microtile == MICRO_SWIZZLE_D && info->cpp[0] == 4)
-			return false;
-	}
+	/* GFX12 doesn't have these limitations. */
+	if (AMD_FMT_MOD_GET(TILE_VERSION, modifier) <= AMD_FMT_MOD_TILE_VER_GFX11) {
+		enum dm_micro_swizzle microtile = amdgpu_dm_plane_modifier_gfx9_swizzle_mode(modifier) & 3;
 
-	if (adev->family >= AMDGPU_FAMILY_RV && microtile == MICRO_SWIZZLE_D &&
-	    info->cpp[0] < 8)
-		return false;
-
-	if (modifier_has_dcc(modifier)) {
-		/* Per radeonsi comments 16/64 bpp are more complicated. */
-		if (info->cpp[0] != 4)
-			return false;
-		/* We support multi-planar formats, but not when combined with
-		 * additional DCC metadata planes.
+		/*
+		 * For D swizzle the canonical modifier depends on the bpp, so check
+		 * it here.
 		 */
-		if (info->num_planes > 1)
+		if (AMD_FMT_MOD_GET(TILE_VERSION, modifier) == AMD_FMT_MOD_TILE_VER_GFX9 &&
+		    adev->family >= AMDGPU_FAMILY_NV) {
+			if (microtile == MICRO_SWIZZLE_D && info->cpp[0] == 4)
+				return false;
+		}
+
+		if (adev->family >= AMDGPU_FAMILY_RV && microtile == MICRO_SWIZZLE_D &&
+		    info->cpp[0] < 8)
 			return false;
+
+		if (amdgpu_dm_plane_modifier_has_dcc(modifier)) {
+			/* Per radeonsi comments 16/64 bpp are more complicated. */
+			if (info->cpp[0] != 4)
+				return false;
+			/* We support multi-planar formats, but not when combined with
+			 * additional DCC metadata planes.
+			 */
+			if (info->num_planes > 1)
+				return false;
+		}
 	}
 
 	return true;
 }
 
-static void dm_drm_plane_destroy_state(struct drm_plane *plane,
-				struct drm_plane_state *state)
+static void amdgpu_dm_plane_drm_plane_destroy_state(struct drm_plane *plane,
+						    struct drm_plane_state *state)
 {
 	struct dm_plane_state *dm_plane_state = to_dm_plane_state(state);
 
+	if (dm_plane_state->degamma_lut)
+		drm_property_blob_put(dm_plane_state->degamma_lut);
+	if (dm_plane_state->ctm)
+		drm_property_blob_put(dm_plane_state->ctm);
+	if (dm_plane_state->lut3d)
+		drm_property_blob_put(dm_plane_state->lut3d);
+	if (dm_plane_state->shaper_lut)
+		drm_property_blob_put(dm_plane_state->shaper_lut);
+	if (dm_plane_state->blend_lut)
+		drm_property_blob_put(dm_plane_state->blend_lut);
+
 	if (dm_plane_state->dc_state)
 		dc_plane_state_release(dm_plane_state->dc_state);
 
 	drm_atomic_helper_plane_destroy_state(plane, state);
 }
 
+#ifdef AMD_PRIVATE_COLOR
+static void
+dm_atomic_plane_attach_color_mgmt_properties(struct amdgpu_display_manager *dm,
+					     struct drm_plane *plane)
+{
+	struct amdgpu_mode_info mode_info = dm->adev->mode_info;
+	struct dpp_color_caps dpp_color_caps = dm->dc->caps.color.dpp;
+
+	/* Check HW color pipeline capabilities on DPP block (pre-blending)
+	 * before exposing related properties.
+	 */
+	if (dpp_color_caps.dgam_ram || dpp_color_caps.gamma_corr) {
+		drm_object_attach_property(&plane->base,
+					   mode_info.plane_degamma_lut_property,
+					   0);
+		drm_object_attach_property(&plane->base,
+					   mode_info.plane_degamma_lut_size_property,
+					   MAX_COLOR_LUT_ENTRIES);
+		drm_object_attach_property(&plane->base,
+					   dm->adev->mode_info.plane_degamma_tf_property,
+					   AMDGPU_TRANSFER_FUNCTION_DEFAULT);
+	}
+	/* HDR MULT is always available */
+	drm_object_attach_property(&plane->base,
+				   dm->adev->mode_info.plane_hdr_mult_property,
+				   AMDGPU_HDR_MULT_DEFAULT);
+
+	/* Only enable plane CTM if both DPP and MPC gamut remap is available. */
+	if (dm->dc->caps.color.mpc.gamut_remap)
+		drm_object_attach_property(&plane->base,
+					   dm->adev->mode_info.plane_ctm_property, 0);
+
+	if (dpp_color_caps.hw_3d_lut || dm->dc->caps.color.mpc.preblend) {
+		drm_object_attach_property(&plane->base,
+					   mode_info.plane_shaper_lut_property, 0);
+		drm_object_attach_property(&plane->base,
+					   mode_info.plane_shaper_lut_size_property,
+					   MAX_COLOR_LUT_ENTRIES);
+		drm_object_attach_property(&plane->base,
+					   mode_info.plane_shaper_tf_property,
+					   AMDGPU_TRANSFER_FUNCTION_DEFAULT);
+		drm_object_attach_property(&plane->base,
+					   mode_info.plane_lut3d_property, 0);
+		drm_object_attach_property(&plane->base,
+					   mode_info.plane_lut3d_size_property,
+					   MAX_COLOR_3DLUT_SIZE);
+	}
+
+	if (dpp_color_caps.ogam_ram) {
+		drm_object_attach_property(&plane->base,
+					   mode_info.plane_blend_lut_property, 0);
+		drm_object_attach_property(&plane->base,
+					   mode_info.plane_blend_lut_size_property,
+					   MAX_COLOR_LUT_ENTRIES);
+		drm_object_attach_property(&plane->base,
+					   mode_info.plane_blend_tf_property,
+					   AMDGPU_TRANSFER_FUNCTION_DEFAULT);
+	}
+}
+
+static int
+dm_atomic_plane_set_property(struct drm_plane *plane,
+			     struct drm_plane_state *state,
+			     struct drm_property *property,
+			     uint64_t val)
+{
+	struct dm_plane_state *dm_plane_state = to_dm_plane_state(state);
+	struct amdgpu_device *adev = drm_to_adev(plane->dev);
+	bool replaced = false;
+	int ret;
+
+	if (property == adev->mode_info.plane_degamma_lut_property) {
+		ret = drm_property_replace_blob_from_id(plane->dev,
+							&dm_plane_state->degamma_lut,
+							val, -1,
+							sizeof(struct drm_color_lut),
+							&replaced);
+		dm_plane_state->base.color_mgmt_changed |= replaced;
+		return ret;
+	} else if (property == adev->mode_info.plane_degamma_tf_property) {
+		if (dm_plane_state->degamma_tf != val) {
+			dm_plane_state->degamma_tf = val;
+			dm_plane_state->base.color_mgmt_changed = 1;
+		}
+	} else if (property == adev->mode_info.plane_hdr_mult_property) {
+		if (dm_plane_state->hdr_mult != val) {
+			dm_plane_state->hdr_mult = val;
+			dm_plane_state->base.color_mgmt_changed = 1;
+		}
+	} else if (property == adev->mode_info.plane_ctm_property) {
+		ret = drm_property_replace_blob_from_id(plane->dev,
+							&dm_plane_state->ctm,
+							val,
+							sizeof(struct drm_color_ctm_3x4), -1,
+							&replaced);
+		dm_plane_state->base.color_mgmt_changed |= replaced;
+		return ret;
+	} else if (property == adev->mode_info.plane_shaper_lut_property) {
+		ret = drm_property_replace_blob_from_id(plane->dev,
+							&dm_plane_state->shaper_lut,
+							val, -1,
+							sizeof(struct drm_color_lut),
+							&replaced);
+		dm_plane_state->base.color_mgmt_changed |= replaced;
+		return ret;
+	} else if (property == adev->mode_info.plane_shaper_tf_property) {
+		if (dm_plane_state->shaper_tf != val) {
+			dm_plane_state->shaper_tf = val;
+			dm_plane_state->base.color_mgmt_changed = 1;
+		}
+	} else if (property == adev->mode_info.plane_lut3d_property) {
+		ret = drm_property_replace_blob_from_id(plane->dev,
+							&dm_plane_state->lut3d,
+							val, -1,
+							sizeof(struct drm_color_lut),
+							&replaced);
+		dm_plane_state->base.color_mgmt_changed |= replaced;
+		return ret;
+	} else if (property == adev->mode_info.plane_blend_lut_property) {
+		ret = drm_property_replace_blob_from_id(plane->dev,
+							&dm_plane_state->blend_lut,
+							val, -1,
+							sizeof(struct drm_color_lut),
+							&replaced);
+		dm_plane_state->base.color_mgmt_changed |= replaced;
+		return ret;
+	} else if (property == adev->mode_info.plane_blend_tf_property) {
+		if (dm_plane_state->blend_tf != val) {
+			dm_plane_state->blend_tf = val;
+			dm_plane_state->base.color_mgmt_changed = 1;
+		}
+	} else {
+		drm_dbg_atomic(plane->dev,
+			       "[PLANE:%d:%s] unknown property [PROP:%d:%s]]\n",
+			       plane->base.id, plane->name,
+			       property->base.id, property->name);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int
+dm_atomic_plane_get_property(struct drm_plane *plane,
+			     const struct drm_plane_state *state,
+			     struct drm_property *property,
+			     uint64_t *val)
+{
+	struct dm_plane_state *dm_plane_state = to_dm_plane_state(state);
+	struct amdgpu_device *adev = drm_to_adev(plane->dev);
+
+	if (property == adev->mode_info.plane_degamma_lut_property) {
+		*val = (dm_plane_state->degamma_lut) ?
+			dm_plane_state->degamma_lut->base.id : 0;
+	} else if (property == adev->mode_info.plane_degamma_tf_property) {
+		*val = dm_plane_state->degamma_tf;
+	} else if (property == adev->mode_info.plane_hdr_mult_property) {
+		*val = dm_plane_state->hdr_mult;
+	} else if (property == adev->mode_info.plane_ctm_property) {
+		*val = (dm_plane_state->ctm) ?
+			dm_plane_state->ctm->base.id : 0;
+	} else 	if (property == adev->mode_info.plane_shaper_lut_property) {
+		*val = (dm_plane_state->shaper_lut) ?
+			dm_plane_state->shaper_lut->base.id : 0;
+	} else if (property == adev->mode_info.plane_shaper_tf_property) {
+		*val = dm_plane_state->shaper_tf;
+	} else 	if (property == adev->mode_info.plane_lut3d_property) {
+		*val = (dm_plane_state->lut3d) ?
+			dm_plane_state->lut3d->base.id : 0;
+	} else 	if (property == adev->mode_info.plane_blend_lut_property) {
+		*val = (dm_plane_state->blend_lut) ?
+			dm_plane_state->blend_lut->base.id : 0;
+	} else if (property == adev->mode_info.plane_blend_tf_property) {
+		*val = dm_plane_state->blend_tf;
+
+	} else {
+		return -EINVAL;
+	}
+
+	return 0;
+}
+#else
+
+#define MAX_COLOR_PIPELINES 5
+
+static int
+dm_plane_init_colorops(struct drm_plane *plane)
+{
+	struct drm_prop_enum_list pipelines[MAX_COLOR_PIPELINES];
+	struct drm_device *dev = plane->dev;
+	struct amdgpu_device *adev = drm_to_adev(dev);
+	struct dc *dc = adev->dm.dc;
+	int len = 0;
+	int ret;
+
+	if (plane->type == DRM_PLANE_TYPE_CURSOR)
+		return 0;
+
+	/* initialize pipeline */
+	if (dc->ctx->dce_version >= DCN_VERSION_3_0) {
+		ret = amdgpu_dm_initialize_default_pipeline(plane, &pipelines[len]);
+		if (ret) {
+			drm_err(plane->dev, "Failed to create color pipeline for plane %d: %d\n",
+				plane->base.id, ret);
+			return ret;
+		}
+		len++;
+
+		/* Create COLOR_PIPELINE property and attach */
+		drm_plane_create_color_pipeline_property(plane, pipelines, len);
+	}
+
+	return 0;
+}
+#endif
+
 static const struct drm_plane_funcs dm_plane_funcs = {
 	.update_plane	= drm_atomic_helper_update_plane,
 	.disable_plane	= drm_atomic_helper_disable_plane,
 	.destroy	= drm_plane_helper_destroy,
-	.reset = dm_drm_plane_reset,
-	.atomic_duplicate_state = dm_drm_plane_duplicate_state,
-	.atomic_destroy_state = dm_drm_plane_destroy_state,
-	.format_mod_supported = dm_plane_format_mod_supported,
+	.reset = amdgpu_dm_plane_drm_plane_reset,
+	.atomic_duplicate_state = amdgpu_dm_plane_drm_plane_duplicate_state,
+	.atomic_destroy_state = amdgpu_dm_plane_drm_plane_destroy_state,
+	.format_mod_supported = amdgpu_dm_plane_format_mod_supported,
+#ifdef AMD_PRIVATE_COLOR
+	.atomic_set_property = dm_atomic_plane_set_property,
+	.atomic_get_property = dm_atomic_plane_get_property,
+#endif
 };
 
 int amdgpu_dm_plane_init(struct amdgpu_display_manager *dm,
@@ -1431,11 +1842,12 @@ int amdgpu_dm_plane_init(struct amdgpu_display_manager *dm,
 	int res = -EPERM;
 	unsigned int supported_rotations;
 	uint64_t *modifiers = NULL;
+	unsigned int primary_zpos = dm->dc->caps.max_slave_planes;
 
-	num_formats = get_plane_formats(plane, plane_cap, formats,
-					ARRAY_SIZE(formats));
+	num_formats = amdgpu_dm_plane_get_plane_formats(plane, plane_cap, formats,
+							ARRAY_SIZE(formats));
 
-	res = get_plane_modifiers(dm->adev, plane->type, &modifiers);
+	res = amdgpu_dm_plane_get_plane_modifiers(dm->adev, plane->type, &modifiers);
 	if (res)
 		return res;
 
@@ -1459,6 +1871,24 @@ int amdgpu_dm_plane_init(struct amdgpu_display_manager *dm,
 		drm_plane_create_blend_mode_property(plane, blend_caps);
 	}
 
+	if (plane->type == DRM_PLANE_TYPE_PRIMARY) {
+		/*
+		 * Allow OVERLAY planes to be used as underlays by assigning an
+		 * immutable zpos = # of OVERLAY planes to the PRIMARY plane.
+		 */
+		drm_plane_create_zpos_immutable_property(plane, primary_zpos);
+	} else if (plane->type == DRM_PLANE_TYPE_OVERLAY) {
+		/*
+		 * OVERLAY planes can be below or above the PRIMARY, but cannot
+		 * be above the CURSOR plane.
+		 */
+		unsigned int zpos = primary_zpos + 1 + drm_plane_index(plane);
+
+		drm_plane_create_zpos_property(plane, zpos, 0, 254);
+	} else if (plane->type == DRM_PLANE_TYPE_CURSOR) {
+		drm_plane_create_zpos_immutable_property(plane, 255);
+	}
+
 	if (plane->type == DRM_PLANE_TYPE_PRIMARY &&
 	    plane_cap &&
 	    (plane_cap->pixel_format_support.nv12 ||
@@ -1483,11 +1913,22 @@ int amdgpu_dm_plane_init(struct amdgpu_display_manager *dm,
 		drm_plane_create_rotation_property(plane, DRM_MODE_ROTATE_0,
 						   supported_rotations);
 
-	if (dm->adev->ip_versions[DCE_HWIP][0] > IP_VERSION(3, 0, 1) &&
+	if (amdgpu_ip_version(dm->adev, DCE_HWIP, 0) > IP_VERSION(3, 0, 1) &&
 	    plane->type != DRM_PLANE_TYPE_CURSOR)
 		drm_plane_enable_fb_damage_clips(plane);
 
-	drm_plane_helper_add(plane, &dm_plane_helper_funcs);
+	if (plane->type == DRM_PLANE_TYPE_PRIMARY)
+		drm_plane_helper_add(plane, &dm_primary_plane_helper_funcs);
+	else
+		drm_plane_helper_add(plane, &dm_plane_helper_funcs);
+
+#ifdef AMD_PRIVATE_COLOR
+	dm_atomic_plane_attach_color_mgmt_properties(dm, plane);
+#else
+	res = dm_plane_init_colorops(plane);
+	if (res)
+		return res;
+#endif
 
 	/* Create (reset) the plane state */
 	if (plane->funcs->reset)
@@ -1496,7 +1937,7 @@ int amdgpu_dm_plane_init(struct amdgpu_display_manager *dm,
 	return 0;
 }
 
-bool is_video_format(uint32_t format)
+bool amdgpu_dm_plane_is_video_format(uint32_t format)
 {
 	int i;
 
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.h
index 930f1572f898..ea2619b507db 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.h
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.h
@@ -29,6 +29,9 @@
 
 #include "dc.h"
 
+int amdgpu_dm_plane_get_cursor_position(struct drm_plane *plane, struct drm_crtc *crtc,
+					struct dc_cursor_position *position);
+
 void amdgpu_dm_plane_handle_cursor_update(struct drm_plane *plane,
 			  struct drm_plane_state *old_plane_state);
 
@@ -44,23 +47,22 @@ int amdgpu_dm_plane_fill_plane_buffer_attributes(struct amdgpu_device *adev,
 				 const enum surface_pixel_format format,
 				 const enum dc_rotation_angle rotation,
 				 const uint64_t tiling_flags,
-				 union dc_tiling_info *tiling_info,
+				 struct dc_tiling_info *tiling_info,
 				 struct plane_size *plane_size,
 				 struct dc_plane_dcc_param *dcc,
 				 struct dc_plane_address *address,
-				 bool tmz_surface,
-				 bool force_disable_dcc);
+				 bool tmz_surface);
 
 int amdgpu_dm_plane_init(struct amdgpu_display_manager *dm,
 			 struct drm_plane *plane,
 			 unsigned long possible_crtcs,
 			 const struct dc_plane_cap *plane_cap);
 
-const struct drm_format_info *amdgpu_dm_plane_get_format_info(const struct drm_mode_fb_cmd2 *cmd);
+const struct drm_format_info *amdgpu_dm_plane_get_format_info(u32 pixel_format, u64 modifier);
 
 void amdgpu_dm_plane_fill_blending_from_plane_state(const struct drm_plane_state *plane_state,
 				    bool *per_pixel_alpha, bool *pre_multiplied_alpha,
 				    bool *global_alpha, int *global_alpha_value);
 
-bool is_video_format(uint32_t format);
+bool amdgpu_dm_plane_is_video_format(uint32_t format);
 #endif
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_pp_smu.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_pp_smu.c
index 75284e2cec74..11b2ea6edf95 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_pp_smu.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_pp_smu.c
@@ -1,3 +1,4 @@
+// SPDX-License-Identifier: MIT
 /*
  * Copyright 2018 Advanced Micro Devices, Inc.
  *
@@ -97,6 +98,7 @@ bool dm_pp_apply_display_requirements(
 			const struct dm_pp_single_disp_config *dc_cfg =
 						&pp_display_cfg->disp_configs[i];
 			adev->pm.pm_display_cfg.displays[i].controller_id = dc_cfg->pipe_idx + 1;
+			adev->pm.pm_display_cfg.displays[i].pixel_clock = dc_cfg->pixel_clock;
 		}
 
 		amdgpu_dpm_display_configuration_change(adev, &adev->pm.pm_display_cfg);
@@ -334,7 +336,8 @@ bool dm_pp_get_clock_levels_by_type(
 			if (dc_clks->clocks_in_khz[i] > validation_clks.engine_max_clock) {
 				/* This clock is higher the validation clock.
 				 * Than means the previous one is the highest
-				 * non-boosted one. */
+				 * non-boosted one.
+				 */
 				DRM_INFO("DM_PPLIB: reducing engine clock level from %d to %d\n",
 						dc_clks->num_levels, i);
 				dc_clks->num_levels = i > 0 ? i : 1;
@@ -406,10 +409,10 @@ bool dm_pp_notify_wm_clock_changes(
 	 * TODO: expand this to other ASICs
 	 */
 	if ((adev->asic_type >= CHIP_POLARIS10) &&
-	     (adev->asic_type <= CHIP_VEGAM) &&
-	     !amdgpu_dpm_set_watermarks_for_clocks_ranges(adev,
-						(void *)wm_with_clock_ranges))
-			return true;
+	    (adev->asic_type <= CHIP_VEGAM) &&
+	    !amdgpu_dpm_set_watermarks_for_clocks_ranges(adev,
+							 (void *)wm_with_clock_ranges))
+		return true;
 
 	return false;
 }
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_psr.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_psr.c
index d647f68fd563..fd491b7a3cd7 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_psr.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_psr.c
@@ -1,3 +1,4 @@
+// SPDX-License-Identifier: MIT
 /*
  * Copyright 2021 Advanced Micro Devices, Inc.
  *
@@ -24,8 +25,8 @@
  */
 
 #include "amdgpu_dm_psr.h"
+#include "dc_dmub_srv.h"
 #include "dc.h"
-#include "dm_helpers.h"
 #include "amdgpu_dm.h"
 #include "modules/power/power_helpers.h"
 
@@ -50,7 +51,11 @@ static bool link_supports_psrsu(struct dc_link *link)
 	    !link->dpcd_caps.psr_info.psr2_su_y_granularity_cap)
 		return false;
 
-	return true;
+	if (amdgpu_dc_debug_mask & DC_DISABLE_PSR_SU)
+		return false;
+
+	/* Temporarily disable PSR-SU to avoid glitches */
+	return false;
 }
 
 /*
@@ -82,14 +87,6 @@ void amdgpu_dm_set_psr_caps(struct dc_link *link)
 
 		link->psr_settings.psr_feature_enabled = true;
 	}
-
-	DRM_INFO("PSR support %d, DC PSR ver %d, sink PSR ver %d DPCD caps 0x%x su_y_granularity %d\n",
-		link->psr_settings.psr_feature_enabled,
-		link->psr_settings.psr_version,
-		link->dpcd_caps.psr_info.psr_version,
-		link->dpcd_caps.psr_info.psr_dpcd_caps.raw,
-		link->dpcd_caps.psr_info.psr2_su_y_granularity_cap);
-
 }
 
 /*
@@ -122,8 +119,10 @@ bool amdgpu_dm_link_setup_psr(struct dc_stream_state *stream)
 		psr_config.allow_multi_disp_optimizations =
 			(amdgpu_dc_feature_mask & DC_PSR_ALLOW_MULTI_DISP_OPT);
 
-		if (!psr_su_set_dsc_slice_height(dc, link, stream, &psr_config))
-			return false;
+		if (link->psr_settings.psr_version == DC_PSR_VERSION_SU_1) {
+			if (!psr_su_set_dsc_slice_height(dc, link, stream, &psr_config))
+				return false;
+		}
 
 		ret = dc_link_setup_psr(link, stream, &psr_config, &psr_context);
 
@@ -137,9 +136,8 @@ bool amdgpu_dm_link_setup_psr(struct dc_stream_state *stream)
  * amdgpu_dm_psr_enable() - enable psr f/w
  * @stream: stream state
  *
- * Return: true if success
  */
-bool amdgpu_dm_psr_enable(struct dc_stream_state *stream)
+void amdgpu_dm_psr_enable(struct dc_stream_state *stream)
 {
 	struct dc_link *link = stream->link;
 	unsigned int vsync_rate_hz = 0;
@@ -155,7 +153,7 @@ bool amdgpu_dm_psr_enable(struct dc_stream_state *stream)
 	DRM_DEBUG_DRIVER("Enabling psr...\n");
 
 	vsync_rate_hz = div64_u64(div64_u64((
-			stream->timing.pix_clk_100hz * 100),
+			stream->timing.pix_clk_100hz * (uint64_t)100),
 			stream->timing.v_total),
 			stream->timing.h_total);
 
@@ -165,6 +163,7 @@ bool amdgpu_dm_psr_enable(struct dc_stream_state *stream)
 	 */
 	if (vsync_rate_hz != 0) {
 		unsigned int frame_time_microsec = 1000000 / vsync_rate_hz;
+
 		num_frames_static = (30000 / frame_time_microsec) + 1;
 	}
 
@@ -185,7 +184,10 @@ bool amdgpu_dm_psr_enable(struct dc_stream_state *stream)
 	if (link->psr_settings.psr_version < DC_PSR_VERSION_SU_1)
 		power_opt |= psr_power_opt_z10_static_screen;
 
-	return dc_link_set_psr_allow_active(link, &psr_enable, false, false, &power_opt);
+	dc_link_set_psr_allow_active(link, &psr_enable, false, false, &power_opt);
+
+	if (link->ctx->dc->caps.ips_support)
+		dc_allow_idle_optimizations(link->ctx->dc, true);
 }
 
 /*
@@ -194,18 +196,17 @@ bool amdgpu_dm_psr_enable(struct dc_stream_state *stream)
  *
  * Return: true if success
  */
-bool amdgpu_dm_psr_disable(struct dc_stream_state *stream)
+bool amdgpu_dm_psr_disable(struct dc_stream_state *stream, bool wait)
 {
-	unsigned int power_opt = 0;
 	bool psr_enable = false;
 
 	DRM_DEBUG_DRIVER("Disabling psr...\n");
 
-	return dc_link_set_psr_allow_active(stream->link, &psr_enable, true, false, &power_opt);
+	return dc_link_set_psr_allow_active(stream->link, &psr_enable, wait, false, NULL);
 }
 
 /*
- * amdgpu_dm_psr_disable() - disable psr f/w
+ * amdgpu_dm_psr_disable_all() - disable psr f/w for all streams
  * if psr is enabled on any stream
  *
  * Return: true if success
@@ -216,3 +217,61 @@ bool amdgpu_dm_psr_disable_all(struct amdgpu_display_manager *dm)
 	return dc_set_psr_allow_active(dm->dc, false);
 }
 
+/*
+ * amdgpu_dm_psr_is_active_allowed() - check if psr is allowed on any stream
+ * @dm:  pointer to amdgpu_display_manager
+ *
+ * Return: true if allowed
+ */
+
+bool amdgpu_dm_psr_is_active_allowed(struct amdgpu_display_manager *dm)
+{
+	unsigned int i;
+	bool allow_active = false;
+
+	for (i = 0; i < dm->dc->current_state->stream_count ; i++) {
+		struct dc_link *link;
+		struct dc_stream_state *stream = dm->dc->current_state->streams[i];
+
+		link = stream->link;
+		if (!link)
+			continue;
+		if (link->psr_settings.psr_feature_enabled &&
+		    link->psr_settings.psr_allow_active) {
+			allow_active = true;
+			break;
+		}
+	}
+
+	return allow_active;
+}
+
+/**
+ * amdgpu_dm_psr_wait_disable() - Wait for eDP panel to exit PSR
+ * @stream: stream state attached to the eDP link
+ *
+ * Waits for a max of 500ms for the eDP panel to exit PSR.
+ *
+ * Return: true if panel exited PSR, false otherwise.
+ */
+bool amdgpu_dm_psr_wait_disable(struct dc_stream_state *stream)
+{
+	enum dc_psr_state psr_state = PSR_STATE0;
+	struct dc_link *link = stream->link;
+	int retry_count;
+
+	if (link == NULL)
+		return false;
+
+	for (retry_count = 0; retry_count <= 1000; retry_count++) {
+		dc_link_get_psr_state(link, &psr_state);
+		if (psr_state == PSR_STATE0)
+			break;
+		udelay(500);
+	}
+
+	if (retry_count == 1000)
+		return false;
+
+	return true;
+}
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_psr.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_psr.h
index 6806b3c9c84b..4fb8626913cf 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_psr.h
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_psr.h
@@ -1,3 +1,4 @@
+/* SPDX-License-Identifier: MIT */
 /*
  * Copyright 2021 Advanced Micro Devices, Inc.
  *
@@ -32,9 +33,11 @@
 #define AMDGPU_DM_PSR_ENTRY_DELAY 5
 
 void amdgpu_dm_set_psr_caps(struct dc_link *link);
-bool amdgpu_dm_psr_enable(struct dc_stream_state *stream);
+void amdgpu_dm_psr_enable(struct dc_stream_state *stream);
 bool amdgpu_dm_link_setup_psr(struct dc_stream_state *stream);
-bool amdgpu_dm_psr_disable(struct dc_stream_state *stream);
+bool amdgpu_dm_psr_disable(struct dc_stream_state *stream, bool wait);
 bool amdgpu_dm_psr_disable_all(struct amdgpu_display_manager *dm);
+bool amdgpu_dm_psr_is_active_allowed(struct amdgpu_display_manager *dm);
+bool amdgpu_dm_psr_wait_disable(struct dc_stream_state *stream);
 
 #endif /* AMDGPU_DM_AMDGPU_DM_PSR_H_ */
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_quirks.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_quirks.c
new file mode 100644
index 000000000000..1da07ebf9217
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_quirks.c
@@ -0,0 +1,178 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: AMD
+ *
+ */
+
+#include <linux/dmi.h>
+
+#include "amdgpu.h"
+#include "amdgpu_dm.h"
+
+struct amdgpu_dm_quirks {
+	bool aux_hpd_discon;
+	bool support_edp0_on_dp1;
+};
+
+static struct amdgpu_dm_quirks quirk_entries = {
+	.aux_hpd_discon = false,
+	.support_edp0_on_dp1 = false
+};
+
+static int edp0_on_dp1_callback(const struct dmi_system_id *id)
+{
+	quirk_entries.support_edp0_on_dp1 = true;
+	return 0;
+}
+
+static int aux_hpd_discon_callback(const struct dmi_system_id *id)
+{
+	quirk_entries.aux_hpd_discon = true;
+	return 0;
+}
+
+static const struct dmi_system_id dmi_quirk_table[] = {
+	{
+		.callback = aux_hpd_discon_callback,
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
+			DMI_MATCH(DMI_PRODUCT_NAME, "Precision 3660"),
+		},
+	},
+	{
+		.callback = aux_hpd_discon_callback,
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
+			DMI_MATCH(DMI_PRODUCT_NAME, "Precision 3260"),
+		},
+	},
+	{
+		.callback = aux_hpd_discon_callback,
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
+			DMI_MATCH(DMI_PRODUCT_NAME, "Precision 3460"),
+		},
+	},
+	{
+		.callback = aux_hpd_discon_callback,
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
+			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex Tower Plus 7010"),
+		},
+	},
+	{
+		.callback = aux_hpd_discon_callback,
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
+			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex Tower 7010"),
+		},
+	},
+	{
+		.callback = aux_hpd_discon_callback,
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
+			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex SFF Plus 7010"),
+		},
+	},
+	{
+		.callback = aux_hpd_discon_callback,
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
+			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex SFF 7010"),
+		},
+	},
+	{
+		.callback = aux_hpd_discon_callback,
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
+			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex Micro Plus 7010"),
+		},
+	},
+	{
+		.callback = aux_hpd_discon_callback,
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
+			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex Micro 7010"),
+		},
+	},
+	{
+		.callback = edp0_on_dp1_callback,
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "HP"),
+			DMI_MATCH(DMI_PRODUCT_NAME, "HP Elite mt645 G8 Mobile Thin Client"),
+		},
+	},
+	{
+		.callback = edp0_on_dp1_callback,
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "HP"),
+			DMI_MATCH(DMI_PRODUCT_NAME, "HP EliteBook 645 14 inch G11 Notebook PC"),
+		},
+	},
+	{
+		.callback = edp0_on_dp1_callback,
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "HP"),
+			DMI_MATCH(DMI_PRODUCT_NAME, "HP EliteBook 665 16 inch G11 Notebook PC"),
+		},
+	},
+	{
+		.callback = edp0_on_dp1_callback,
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "HP"),
+			DMI_MATCH(DMI_PRODUCT_NAME, "HP ProBook 445 14 inch G11 Notebook PC"),
+		},
+	},
+	{
+		.callback = edp0_on_dp1_callback,
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "HP"),
+			DMI_MATCH(DMI_PRODUCT_NAME, "HP ProBook 465 16 inch G11 Notebook PC"),
+		},
+	},
+	{}
+	/* TODO: refactor this from a fixed table to a dynamic option */
+};
+
+void retrieve_dmi_info(struct amdgpu_display_manager *dm)
+{
+	struct drm_device *dev = dm->ddev;
+	int dmi_id;
+
+	dm->aux_hpd_discon_quirk = false;
+	dm->edp0_on_dp1_quirk = false;
+
+	dmi_id = dmi_check_system(dmi_quirk_table);
+
+	if (!dmi_id)
+		return;
+
+	if (quirk_entries.aux_hpd_discon) {
+		dm->aux_hpd_discon_quirk = true;
+		drm_info(dev, "aux_hpd_discon_quirk attached\n");
+	}
+	if (quirk_entries.support_edp0_on_dp1) {
+		dm->edp0_on_dp1_quirk = true;
+		drm_info(dev, "support_edp0_on_dp1 attached\n");
+	}
+}
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_replay.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_replay.c
new file mode 100644
index 000000000000..da94e3544b65
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_replay.c
@@ -0,0 +1,209 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: AMD
+ *
+ */
+
+#include "amdgpu_dm_replay.h"
+#include "dc_dmub_srv.h"
+#include "dc.h"
+#include "dm_helpers.h"
+#include "amdgpu_dm.h"
+#include "modules/power/power_helpers.h"
+#include "dmub/inc/dmub_cmd.h"
+#include "dc/inc/link_service.h"
+
+/*
+ * amdgpu_dm_link_supports_replay() - check if the link supports replay
+ * @link: link
+ * @aconnector: aconnector
+ *
+ */
+bool amdgpu_dm_link_supports_replay(struct dc_link *link, struct amdgpu_dm_connector *aconnector)
+{
+	struct dm_connector_state *state = to_dm_connector_state(aconnector->base.state);
+	struct dpcd_caps *dpcd_caps = &link->dpcd_caps;
+	struct adaptive_sync_caps *as_caps = &link->dpcd_caps.adaptive_sync_caps;
+
+	if (!state->freesync_capable)
+		return false;
+
+	if (!aconnector->vsdb_info.replay_mode)
+		return false;
+
+	// Check the eDP version
+	if (dpcd_caps->edp_rev < EDP_REVISION_13)
+		return false;
+
+	if (!dpcd_caps->alpm_caps.bits.AUX_WAKE_ALPM_CAP)
+		return false;
+
+	// Check adaptive sync support cap
+	if (!as_caps->dp_adap_sync_caps.bits.ADAPTIVE_SYNC_SDP_SUPPORT)
+		return false;
+
+	// Sink shall populate line deviation information
+	if (dpcd_caps->pr_info.pixel_deviation_per_line == 0 ||
+		dpcd_caps->pr_info.max_deviation_line == 0)
+		return false;
+
+	return true;
+}
+
+/*
+ * amdgpu_dm_set_replay_caps() - setup Replay capabilities
+ * @link: link
+ * @aconnector: aconnector
+ *
+ */
+bool amdgpu_dm_set_replay_caps(struct dc_link *link, struct amdgpu_dm_connector *aconnector)
+{
+	struct replay_config pr_config = { 0 };
+	union replay_debug_flags *debug_flags = NULL;
+	struct dc *dc = link->ctx->dc;
+
+	// If Replay is already set to support, return true to skip checks
+	if (link->replay_settings.config.replay_supported)
+		return true;
+
+	if (!dc_is_embedded_signal(link->connector_signal))
+		return false;
+
+	if (link->panel_config.psr.disallow_replay)
+		return false;
+
+	if (!amdgpu_dm_link_supports_replay(link, aconnector))
+		return false;
+
+	if (!dc->ctx->dmub_srv || !dc->ctx->dmub_srv->dmub ||
+		!dc->ctx->dmub_srv->dmub->feature_caps.replay_supported)
+		return false;
+
+	// Mark Replay is supported in pr_config
+	pr_config.replay_supported = true;
+
+	debug_flags = (union replay_debug_flags *)&pr_config.debug_flags;
+	debug_flags->u32All = 0;
+	debug_flags->bitfields.visual_confirm =
+		link->ctx->dc->debug.visual_confirm == VISUAL_CONFIRM_REPLAY;
+
+	init_replay_config(link, &pr_config);
+
+	return true;
+}
+
+/*
+ * amdgpu_dm_link_setup_replay() - configure replay link
+ * @link: link
+ * @aconnector: aconnector
+ *
+ */
+bool amdgpu_dm_link_setup_replay(struct dc_link *link, struct amdgpu_dm_connector *aconnector)
+{
+	struct replay_config *pr_config;
+
+	if (link == NULL || aconnector == NULL)
+		return false;
+
+	pr_config = &link->replay_settings.config;
+
+	if (!pr_config->replay_supported)
+		return false;
+
+	pr_config->replay_power_opt_supported = 0x11;
+	pr_config->replay_smu_opt_supported = false;
+	pr_config->replay_enable_option |= pr_enable_option_static_screen;
+	pr_config->replay_support_fast_resync_in_ultra_sleep_mode = aconnector->max_vfreq >= 2 * aconnector->min_vfreq;
+	pr_config->replay_timing_sync_supported = false;
+
+	if (!pr_config->replay_timing_sync_supported)
+		pr_config->replay_enable_option &= ~pr_enable_option_general_ui;
+
+	link->replay_settings.replay_feature_enabled = true;
+
+	return true;
+}
+
+/*
+ * amdgpu_dm_replay_enable() - enable replay f/w
+ * @stream: stream state
+ *
+ * Return: true if success
+ */
+bool amdgpu_dm_replay_enable(struct dc_stream_state *stream, bool wait)
+{
+	bool replay_active = true;
+	struct dc_link *link = NULL;
+
+	if (stream == NULL)
+		return false;
+
+	link = stream->link;
+
+	if (link) {
+		link->dc->link_srv->edp_setup_replay(link, stream);
+		link->dc->link_srv->edp_set_coasting_vtotal(link, stream->timing.v_total, 0);
+		DRM_DEBUG_DRIVER("Enabling replay...\n");
+		link->dc->link_srv->edp_set_replay_allow_active(link, &replay_active, wait, false, NULL);
+		return true;
+	}
+
+	return false;
+}
+
+/*
+ * amdgpu_dm_replay_disable() - disable replay f/w
+ * @stream:  stream state
+ *
+ * Return: true if success
+ */
+bool amdgpu_dm_replay_disable(struct dc_stream_state *stream)
+{
+	bool replay_active = false;
+	struct dc_link *link = NULL;
+
+	if (stream == NULL)
+		return false;
+
+	link = stream->link;
+
+	if (link) {
+		DRM_DEBUG_DRIVER("Disabling replay...\n");
+		link->dc->link_srv->edp_set_replay_allow_active(stream->link, &replay_active, true, false, NULL);
+		return true;
+	}
+
+	return false;
+}
+
+/*
+ * amdgpu_dm_replay_disable_all() - disable replay f/w
+ * if replay is enabled on any stream
+ *
+ * Return: true if success
+ */
+bool amdgpu_dm_replay_disable_all(struct amdgpu_display_manager *dm)
+{
+	DRM_DEBUG_DRIVER("Disabling replay if replay is enabled on any stream\n");
+	return dc_set_replay_allow_active(dm->dc, false);
+}
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_replay.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_replay.h
new file mode 100644
index 000000000000..73b6c67ae5e7
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_replay.h
@@ -0,0 +1,49 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2021 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: AMD
+ *
+ */
+
+#ifndef AMDGPU_DM_AMDGPU_DM_REPLAY_H_
+#define AMDGPU_DM_AMDGPU_DM_REPLAY_H_
+
+#include "amdgpu.h"
+
+enum replay_enable_option {
+	pr_enable_option_static_screen = 0x1,
+	pr_enable_option_mpo_video = 0x2,
+	pr_enable_option_full_screen_video = 0x4,
+	pr_enable_option_general_ui = 0x8,
+	pr_enable_option_static_screen_coasting = 0x10000,
+	pr_enable_option_mpo_video_coasting = 0x20000,
+	pr_enable_option_full_screen_video_coasting = 0x40000,
+};
+
+bool amdgpu_dm_link_supports_replay(struct dc_link *link, struct amdgpu_dm_connector *aconnector);
+bool amdgpu_dm_replay_enable(struct dc_stream_state *stream, bool enable);
+bool amdgpu_dm_set_replay_caps(struct dc_link *link, struct amdgpu_dm_connector *aconnector);
+bool amdgpu_dm_link_setup_replay(struct dc_link *link, struct amdgpu_dm_connector *aconnector);
+bool amdgpu_dm_replay_disable(struct dc_stream_state *stream);
+bool amdgpu_dm_replay_disable_all(struct amdgpu_display_manager *dm);
+
+#endif /* AMDGPU_DM_AMDGPU_DM_REPLAY_H_ */
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_services.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_services.c
index d9e33c6bccd9..8550d5e8b753 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_services.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_services.c
@@ -1,3 +1,4 @@
+//SPDX-License-Identifier: MIT
 /*
  * Copyright 2015 Advanced Micro Devices, Inc.
  *
@@ -52,4 +53,12 @@ void dm_perf_trace_timestamp(const char *func_name, unsigned int line, struct dc
 				    func_name, line);
 }
 
+void dm_trace_smu_enter(uint32_t msg_id, uint32_t param_in, unsigned int delay, struct dc_context *ctx)
+{
+}
+
+void dm_trace_smu_exit(bool success, uint32_t response, struct dc_context *ctx)
+{
+}
+
 /**** power component interfaces ****/
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_trace.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_trace.h
index 0f580ea37576..aa56fd6d56c3 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_trace.h
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_trace.h
@@ -1,3 +1,4 @@
+//SPDX-License-Identifier: MIT
 /*
  * Copyright 2018 Advanced Micro Devices, Inc.
  *
@@ -37,7 +38,7 @@
 #include <drm/drm_framebuffer.h>
 #include <drm/drm_encoder.h>
 #include <drm/drm_atomic.h>
-#include "dcn10/dcn10_optc.h"
+#include "dc/inc/hw/optc.h"
 
 #include "dc/inc/core_types.h"
 
@@ -87,7 +88,7 @@ TRACE_EVENT(amdgpu_dc_performance,
 			__entry->writes = write_count;
 			__entry->read_delta = read_count - *last_read;
 			__entry->write_delta = write_count - *last_write;
-			__assign_str(func, func);
+			__assign_str(func);
 			__entry->line = line;
 			*last_read = read_count;
 			*last_write = write_count;
@@ -726,6 +727,32 @@ TRACE_EVENT(dcn_optc_lock_unlock_state,
 	    )
 );
 
+TRACE_EVENT(amdgpu_dm_brightness,
+	TP_PROTO(void *function, u32 user_brightness, u32 converted_brightness, bool aux, bool ac),
+	TP_ARGS(function, user_brightness, converted_brightness, aux, ac),
+	TP_STRUCT__entry(
+		__field(void *, function)
+		__field(u32, user_brightness)
+		__field(u32, converted_brightness)
+		__field(bool, aux)
+		__field(bool, ac)
+	),
+	TP_fast_assign(
+		__entry->function = function;
+		__entry->user_brightness = user_brightness;
+		__entry->converted_brightness = converted_brightness;
+		__entry->aux = aux;
+		__entry->ac = ac;
+	),
+	TP_printk("%ps: brightness requested=%u converted=%u aux=%s power=%s",
+		  (void *)__entry->function,
+		  (u32)__entry->user_brightness,
+		  (u32)__entry->converted_brightness,
+		  (__entry->aux) ? "true" : "false",
+		  (__entry->ac) ? "AC" : "DC"
+	)
+);
+
 #endif /* _AMDGPU_DM_TRACE_H_ */
 
 #undef TRACE_INCLUDE_PATH
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_wb.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_wb.c
new file mode 100644
index 000000000000..d9527c05fc87
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_wb.c
@@ -0,0 +1,215 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: AMD
+ *
+ */
+
+#include "dm_services_types.h"
+
+#include "amdgpu.h"
+#include "amdgpu_dm.h"
+#include "amdgpu_dm_wb.h"
+#include "amdgpu_display.h"
+#include "dc.h"
+
+#include <drm/drm_edid.h>
+#include <drm/drm_atomic_state_helper.h>
+#include <drm/drm_modeset_helper_vtables.h>
+
+static const u32 amdgpu_dm_wb_formats[] = {
+	DRM_FORMAT_XRGB2101010,
+};
+
+static int amdgpu_dm_wb_encoder_atomic_check(struct drm_encoder *encoder,
+					struct drm_crtc_state *crtc_state,
+					struct drm_connector_state *conn_state)
+{
+	struct drm_framebuffer *fb;
+	const struct drm_display_mode *mode = &crtc_state->mode;
+	bool found = false;
+	uint8_t i;
+
+	if (!conn_state->writeback_job || !conn_state->writeback_job->fb)
+		return 0;
+
+	fb = conn_state->writeback_job->fb;
+	if (fb->width != mode->hdisplay || fb->height != mode->vdisplay) {
+		DRM_DEBUG_KMS("Invalid framebuffer size %ux%u\n",
+			      fb->width, fb->height);
+		return -EINVAL;
+	}
+
+	for (i = 0; i < sizeof(amdgpu_dm_wb_formats) / sizeof(u32); i++) {
+		if (fb->format->format == amdgpu_dm_wb_formats[i])
+			found = true;
+	}
+
+	if (!found) {
+		DRM_DEBUG_KMS("Invalid pixel format %p4cc\n",
+			      &fb->format->format);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+
+static int amdgpu_dm_wb_connector_get_modes(struct drm_connector *connector)
+{
+	/* Maximum resolution supported by DWB */
+	return drm_add_modes_noedid(connector, 3840, 2160);
+}
+
+static int amdgpu_dm_wb_prepare_job(struct drm_writeback_connector *wb_connector,
+			       struct drm_writeback_job *job)
+{
+	struct amdgpu_framebuffer *afb;
+	struct drm_gem_object *obj;
+	struct amdgpu_device *adev;
+	struct amdgpu_bo *rbo;
+	uint32_t domain;
+	int r;
+
+	if (!job->fb) {
+		DRM_DEBUG_KMS("No FB bound\n");
+		return 0;
+	}
+
+	afb = to_amdgpu_framebuffer(job->fb);
+	obj = job->fb->obj[0];
+	rbo = gem_to_amdgpu_bo(obj);
+	adev = amdgpu_ttm_adev(rbo->tbo.bdev);
+
+	r = amdgpu_bo_reserve(rbo, true);
+	if (r) {
+		drm_err(adev_to_drm(adev), "fail to reserve bo (%d)\n", r);
+		return r;
+	}
+
+	r = dma_resv_reserve_fences(rbo->tbo.base.resv, 1);
+	if (r) {
+		drm_err(adev_to_drm(adev), "reserving fence slot failed (%d)\n", r);
+		goto error_unlock;
+	}
+
+	domain = amdgpu_display_supported_domains(adev, rbo->flags);
+
+	rbo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
+	r = amdgpu_bo_pin(rbo, domain);
+	if (unlikely(r != 0)) {
+		if (r != -ERESTARTSYS)
+			DRM_ERROR("Failed to pin framebuffer with error %d\n", r);
+		goto error_unlock;
+	}
+
+	r = amdgpu_ttm_alloc_gart(&rbo->tbo);
+	if (unlikely(r != 0)) {
+		DRM_ERROR("%p bind failed\n", rbo);
+		goto error_unpin;
+	}
+
+	amdgpu_bo_unreserve(rbo);
+
+	afb->address = amdgpu_bo_gpu_offset(rbo);
+
+	amdgpu_bo_ref(rbo);
+
+	return 0;
+
+error_unpin:
+	amdgpu_bo_unpin(rbo);
+
+error_unlock:
+	amdgpu_bo_unreserve(rbo);
+	return r;
+}
+
+static void amdgpu_dm_wb_cleanup_job(struct drm_writeback_connector *connector,
+				struct drm_writeback_job *job)
+{
+	struct amdgpu_bo *rbo;
+	int r;
+
+	if (!job->fb)
+		return;
+
+	rbo = gem_to_amdgpu_bo(job->fb->obj[0]);
+	r = amdgpu_bo_reserve(rbo, false);
+	if (unlikely(r)) {
+		DRM_ERROR("failed to reserve rbo before unpin\n");
+		return;
+	}
+
+	amdgpu_bo_unpin(rbo);
+	amdgpu_bo_unreserve(rbo);
+	amdgpu_bo_unref(&rbo);
+}
+
+static const struct drm_encoder_helper_funcs amdgpu_dm_wb_encoder_helper_funcs = {
+	.atomic_check = amdgpu_dm_wb_encoder_atomic_check,
+};
+
+static const struct drm_connector_funcs amdgpu_dm_wb_connector_funcs = {
+	.fill_modes = drm_helper_probe_single_connector_modes,
+	.destroy = drm_connector_cleanup,
+	.reset = amdgpu_dm_connector_funcs_reset,
+	.atomic_duplicate_state = amdgpu_dm_connector_atomic_duplicate_state,
+	.atomic_destroy_state = drm_atomic_helper_connector_destroy_state,
+};
+
+static const struct drm_connector_helper_funcs amdgpu_dm_wb_conn_helper_funcs = {
+	.get_modes = amdgpu_dm_wb_connector_get_modes,
+	.prepare_writeback_job = amdgpu_dm_wb_prepare_job,
+	.cleanup_writeback_job = amdgpu_dm_wb_cleanup_job,
+};
+
+int amdgpu_dm_wb_connector_init(struct amdgpu_display_manager *dm,
+				struct amdgpu_dm_wb_connector *wbcon,
+				uint32_t link_index)
+{
+	struct dc *dc = dm->dc;
+	struct dc_link *link = dc_get_link_at_index(dc, link_index);
+	int res = 0;
+
+	wbcon->link = link;
+
+	drm_connector_helper_add(&wbcon->base.base, &amdgpu_dm_wb_conn_helper_funcs);
+
+	res = drm_writeback_connector_init(&dm->adev->ddev, &wbcon->base,
+					    &amdgpu_dm_wb_connector_funcs,
+					    &amdgpu_dm_wb_encoder_helper_funcs,
+					    amdgpu_dm_wb_formats,
+					    ARRAY_SIZE(amdgpu_dm_wb_formats),
+					    amdgpu_dm_get_encoder_crtc_mask(dm->adev));
+
+	if (res)
+		return res;
+	/*
+	 * Some of the properties below require access to state, like bpc.
+	 * Allocate some default initial connector state with our reset helper.
+	 */
+	if (wbcon->base.base.funcs->reset)
+		wbcon->base.base.funcs->reset(&wbcon->base.base);
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_wb.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_wb.h
new file mode 100644
index 000000000000..13d31c857dee
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_wb.h
@@ -0,0 +1,36 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: AMD
+ *
+ */
+
+#ifndef __AMDGPU_DM_WB_H__
+#define __AMDGPU_DM_WB_H__
+
+#include <drm/drm_writeback.h>
+
+int amdgpu_dm_wb_connector_init(struct amdgpu_display_manager *dm,
+				struct amdgpu_dm_wb_connector *dm_wbcon,
+				uint32_t link_index);
+
+#endif
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/dc_fpu.c b/drivers/gpu/drm/amd/display/amdgpu_dm/dc_fpu.c
index 1743ca0a3641..e46f8ce41d87 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/dc_fpu.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/dc_fpu.c
@@ -26,14 +26,7 @@
 
 #include "dc_trace.h"
 
-#if defined(CONFIG_X86)
-#include <asm/fpu/api.h>
-#elif defined(CONFIG_PPC64)
-#include <asm/switch_to.h>
-#include <asm/cputable.h>
-#elif defined(CONFIG_ARM64)
-#include <asm/neon.h>
-#endif
+#include <linux/fpu.h>
 
 /**
  * DOC: DC FPU manipulation overview
@@ -58,11 +51,9 @@ static DEFINE_PER_CPU(int, fpu_recursion_depth);
  */
 inline void dc_assert_fp_enabled(void)
 {
-	int *pcpu, depth = 0;
+	int depth;
 
-	pcpu = get_cpu_ptr(&fpu_recursion_depth);
-	depth = *pcpu;
-	put_cpu_ptr(&fpu_recursion_depth);
+	depth = __this_cpu_read(fpu_recursion_depth);
 
 	ASSERT(depth >= 1);
 }
@@ -82,32 +73,17 @@ inline void dc_assert_fp_enabled(void)
  */
 void dc_fpu_begin(const char *function_name, const int line)
 {
-	int *pcpu;
+	int depth;
 
-	pcpu = get_cpu_ptr(&fpu_recursion_depth);
-	*pcpu += 1;
-
-	if (*pcpu == 1) {
-#if defined(CONFIG_X86)
+	WARN_ON_ONCE(!in_task());
+	preempt_disable();
+	depth = __this_cpu_inc_return(fpu_recursion_depth);
+	if (depth == 1) {
+		BUG_ON(!kernel_fpu_available());
 		kernel_fpu_begin();
-#elif defined(CONFIG_PPC64)
-		if (cpu_has_feature(CPU_FTR_VSX_COMP)) {
-			preempt_disable();
-			enable_kernel_vsx();
-		} else if (cpu_has_feature(CPU_FTR_ALTIVEC_COMP)) {
-			preempt_disable();
-			enable_kernel_altivec();
-		} else if (!cpu_has_feature(CPU_FTR_FPU_UNAVAILABLE)) {
-			preempt_disable();
-			enable_kernel_fp();
-		}
-#elif defined(CONFIG_ARM64)
-		kernel_neon_begin();
-#endif
 	}
 
-	TRACE_DCN_FPU(true, function_name, line, *pcpu);
-	put_cpu_ptr(&fpu_recursion_depth);
+	TRACE_DCN_FPU(true, function_name, line, depth);
 }
 
 /**
@@ -122,29 +98,15 @@ void dc_fpu_begin(const char *function_name, const int line)
  */
 void dc_fpu_end(const char *function_name, const int line)
 {
-	int *pcpu;
+	int depth;
 
-	pcpu = get_cpu_ptr(&fpu_recursion_depth);
-	*pcpu -= 1;
-	if (*pcpu <= 0) {
-#if defined(CONFIG_X86)
+	depth = __this_cpu_dec_return(fpu_recursion_depth);
+	if (depth == 0) {
 		kernel_fpu_end();
-#elif defined(CONFIG_PPC64)
-		if (cpu_has_feature(CPU_FTR_VSX_COMP)) {
-			disable_kernel_vsx();
-			preempt_enable();
-		} else if (cpu_has_feature(CPU_FTR_ALTIVEC_COMP)) {
-			disable_kernel_altivec();
-			preempt_enable();
-		} else if (!cpu_has_feature(CPU_FTR_FPU_UNAVAILABLE)) {
-			disable_kernel_fp();
-			preempt_enable();
-		}
-#elif defined(CONFIG_ARM64)
-		kernel_neon_end();
-#endif
+	} else {
+		WARN_ON_ONCE(depth < 0);
 	}
 
-	TRACE_DCN_FPU(false, function_name, line, *pcpu);
-	put_cpu_ptr(&fpu_recursion_depth);
+	TRACE_DCN_FPU(false, function_name, line, depth);
+	preempt_enable();
 }
diff --git a/drivers/gpu/drm/amd/display/dc/Makefile b/drivers/gpu/drm/amd/display/dc/Makefile
index 69ffd4424dc7..7277ed21552f 100644
--- a/drivers/gpu/drm/amd/display/dc/Makefile
+++ b/drivers/gpu/drm/amd/display/dc/Makefile
@@ -22,33 +22,28 @@
 #
 # Makefile for Display Core (dc) component.
 
-DC_LIBS = basics bios dml clk_mgr dce gpio irq link virtual dsc
+DC_LIBS = basics bios dml clk_mgr dce gpio hwss irq link virtual dsc resource optc dpp hubbub dccg hubp dio dwb hpo mmhubbub mpc opp pg
 
 ifdef CONFIG_DRM_AMD_DC_FP
 
 KCOV_INSTRUMENT := n
 
-DC_LIBS += dcn20
 DC_LIBS += dcn10
+DC_LIBS += dcn20
 DC_LIBS += dcn21
 DC_LIBS += dcn201
 DC_LIBS += dcn30
 DC_LIBS += dcn301
-DC_LIBS += dcn302
-DC_LIBS += dcn303
 DC_LIBS += dcn31
-DC_LIBS += dcn314
-DC_LIBS += dcn315
-DC_LIBS += dcn316
-DC_LIBS += dcn32
-DC_LIBS += dcn321
+DC_LIBS += dml
+DC_LIBS += dml2_0
+DC_LIBS += soc_and_ip_translator
 endif
 
 DC_LIBS += dce120
 
 DC_LIBS += dce112
 DC_LIBS += dce110
-DC_LIBS += dce100
 DC_LIBS += dce80
 
 ifdef CONFIG_DRM_AMD_DC_SI
@@ -57,24 +52,32 @@ endif
 
 DC_LIBS += hdcp
 
+ifdef CONFIG_DRM_AMD_DC_FP
+DC_LIBS += sspl
+AMD_DISPLAY_FILES += $(addprefix $(AMDDALPATH)/dc/, dc_spl_translate.o)
+endif
+
 AMD_DC = $(addsuffix /Makefile, $(addprefix $(FULL_AMD_DISPLAY_PATH)/dc/,$(DC_LIBS)))
 
 include $(AMD_DC)
 
-DISPLAY_CORE = dc.o dc_stat.o dc_resource.o dc_hw_sequencer.o dc_sink.o \
-dc_surface.o dc_debug.o dc_stream.o dc_link_enc_cfg.o dc_link_exports.o
-
-DISPLAY_CORE += dc_vm_helper.o
-
-AMD_DISPLAY_CORE = $(addprefix $(AMDDALPATH)/dc/core/,$(DISPLAY_CORE))
-
-AMD_DM_REG_UPDATE = $(addprefix $(AMDDALPATH)/dc/,dc_helper.o)
-
-AMD_DISPLAY_FILES += $(AMD_DISPLAY_CORE)
-AMD_DISPLAY_FILES += $(AMD_DM_REG_UPDATE)
+FILES =
+FILES += dc_dmub_srv.o
+FILES += dc_edid_parser.o
+FILES += dc_fused_io.o
+FILES += dc_helper.o
+FILES += core/dc.o
+FILES += core/dc_debug.o
+FILES += core/dc_hw_sequencer.o
+FILES += core/dc_link_enc_cfg.o
+FILES += core/dc_link_exports.o
+FILES += core/dc_resource.o
+FILES += core/dc_sink.o
+FILES += core/dc_stat.o
+FILES += core/dc_state.o
+FILES += core/dc_stream.o
+FILES += core/dc_surface.o
+FILES += core/dc_vm_helper.o
+
+AMD_DISPLAY_FILES += $(addprefix $(AMDDALPATH)/dc/, $(FILES))
 
-DC_DMUB += dc_dmub_srv.o
-DC_EDID += dc_edid_parser.o
-AMD_DISPLAY_DMUB = $(addprefix $(AMDDALPATH)/dc/,$(DC_DMUB))
-AMD_DISPLAY_EDID = $(addprefix $(AMDDALPATH)/dc/,$(DC_EDID))
-AMD_DISPLAY_FILES += $(AMD_DISPLAY_DMUB) $(AMD_DISPLAY_EDID)
diff --git a/drivers/gpu/drm/amd/display/dc/basics/Makefile b/drivers/gpu/drm/amd/display/dc/basics/Makefile
index 01b99e0d788e..aabcebf69049 100644
--- a/drivers/gpu/drm/amd/display/dc/basics/Makefile
+++ b/drivers/gpu/drm/amd/display/dc/basics/Makefile
@@ -24,7 +24,14 @@
 # It provides the general basic services required by other DAL
 # subcomponents.
 
-BASICS = conversion.o fixpt31_32.o vector.o dc_common.o
+BASICS := \
+	conversion.o \
+	fixpt31_32.o \
+	vector.o \
+	dc_common.o \
+	dce_calcs.o \
+	custom_float.o \
+	bw_fixed.o
 
 AMD_DAL_BASICS = $(addprefix $(AMDDALPATH)/dc/basics/,$(BASICS))
 
diff --git a/drivers/gpu/drm/amd/display/dc/dml/calcs/bw_fixed.c b/drivers/gpu/drm/amd/display/dc/basics/bw_fixed.c
index 3aa8dd0acd5e..c8cb89e0d4d0 100644
--- a/drivers/gpu/drm/amd/display/dc/dml/calcs/bw_fixed.c
+++ b/drivers/gpu/drm/amd/display/dc/basics/bw_fixed.c
@@ -1,5 +1,6 @@
+// SPDX-License-Identifier: MIT
 /*
- * Copyright 2015 Advanced Micro Devices, Inc.
+ * Copyright 2023 Advanced Micro Devices, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -106,9 +107,8 @@ struct bw_fixed bw_frc_to_fixed(int64_t numerator, int64_t denominator)
 	return res;
 }
 
-struct bw_fixed bw_floor2(
-	const struct bw_fixed arg,
-	const struct bw_fixed significance)
+struct bw_fixed bw_floor2(const struct bw_fixed arg,
+			  const struct bw_fixed significance)
 {
 	struct bw_fixed result;
 	int64_t multiplicand;
@@ -119,9 +119,8 @@ struct bw_fixed bw_floor2(
 	return result;
 }
 
-struct bw_fixed bw_ceil2(
-	const struct bw_fixed arg,
-	const struct bw_fixed significance)
+struct bw_fixed bw_ceil2(const struct bw_fixed arg,
+			 const struct bw_fixed significance)
 {
 	struct bw_fixed result;
 	int64_t multiplicand;
diff --git a/drivers/gpu/drm/amd/display/dc/dml/calcs/calcs_logger.h b/drivers/gpu/drm/amd/display/dc/basics/calcs_logger.h
index 62435bfc274d..62435bfc274d 100644
--- a/drivers/gpu/drm/amd/display/dc/dml/calcs/calcs_logger.h
+++ b/drivers/gpu/drm/amd/display/dc/basics/calcs_logger.h
diff --git a/drivers/gpu/drm/amd/display/dc/basics/conversion.c b/drivers/gpu/drm/amd/display/dc/basics/conversion.c
index 352e9afb85c6..bd1f60ecaba4 100644
--- a/drivers/gpu/drm/amd/display/dc/basics/conversion.c
+++ b/drivers/gpu/drm/amd/display/dc/basics/conversion.c
@@ -24,7 +24,7 @@
  */
 
 #include "dm_services.h"
-#include "conversion.h"
+#include "basics/conversion.h"
 
 #define DIVIDER 10000
 
@@ -101,9 +101,44 @@ void convert_float_matrix(
 	}
 }
 
+static struct fixed31_32 int_frac_to_fixed_point(uint16_t arg,
+						 uint8_t integer_bits,
+						 uint8_t fractional_bits)
+{
+	struct fixed31_32 result;
+	uint16_t sign_mask = 1 << (fractional_bits + integer_bits);
+	uint16_t value_mask = sign_mask - 1;
+
+	result.value = (long long)(arg & value_mask) <<
+		       (FIXED31_32_BITS_PER_FRACTIONAL_PART - fractional_bits);
+
+	if (arg & sign_mask)
+		result = dc_fixpt_neg(result);
+
+	return result;
+}
+
+/**
+ * convert_hw_matrix - converts HW values into fixed31_32 matrix.
+ * @matrix: fixed point 31.32 matrix
+ * @reg: array of register values
+ * @buffer_size: size of the array of register values
+ *
+ * Converts HW register spec defined format S2D13 into a fixed-point 31.32
+ * matrix.
+ */
+void convert_hw_matrix(struct fixed31_32 *matrix,
+		       uint16_t *reg,
+		       uint32_t buffer_size)
+{
+	for (int i = 0; i < buffer_size; ++i)
+		matrix[i] = int_frac_to_fixed_point(reg[i], 2, 13);
+}
+
 static uint32_t find_gcd(uint32_t a, uint32_t b)
 {
-	uint32_t remainder = 0;
+	uint32_t remainder;
+
 	while (b != 0) {
 		remainder = a % b;
 		a = b;
diff --git a/drivers/gpu/drm/amd/display/dc/basics/conversion.h b/drivers/gpu/drm/amd/display/dc/basics/conversion.h
index 81da4e6f7a1a..a433cef78496 100644
--- a/drivers/gpu/drm/amd/display/dc/basics/conversion.h
+++ b/drivers/gpu/drm/amd/display/dc/basics/conversion.h
@@ -41,6 +41,10 @@ void convert_float_matrix(
 void reduce_fraction(uint32_t num, uint32_t den,
 		uint32_t *out_num, uint32_t *out_den);
 
+void convert_hw_matrix(struct fixed31_32 *matrix,
+		       uint16_t *reg,
+		       uint32_t buffer_size);
+
 static inline unsigned int log_2(unsigned int num)
 {
 	return ilog2(num);
diff --git a/drivers/gpu/drm/amd/display/dc/dml/calcs/custom_float.c b/drivers/gpu/drm/amd/display/dc/basics/custom_float.c
index 31d167bc548f..ae05ded9a7f3 100644
--- a/drivers/gpu/drm/amd/display/dc/dml/calcs/custom_float.c
+++ b/drivers/gpu/drm/amd/display/dc/basics/custom_float.c
@@ -1,5 +1,6 @@
+// SPDX-License-Identifier: MIT
 /*
- * Copyright 2017 Advanced Micro Devices, Inc.
+ * Copyright 2023 Advanced Micro Devices, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -25,52 +26,41 @@
 #include "dm_services.h"
 #include "custom_float.h"
 
-
-static bool build_custom_float(
-	struct fixed31_32 value,
-	const struct custom_float_format *format,
-	bool *negative,
-	uint32_t *mantissa,
-	uint32_t *exponenta)
+static bool build_custom_float(struct fixed31_32 value,
+			       const struct custom_float_format *format,
+			       bool *negative,
+			       uint32_t *mantissa,
+			       uint32_t *exponenta)
 {
 	uint32_t exp_offset = (1 << (format->exponenta_bits - 1)) - 1;
 
 	const struct fixed31_32 mantissa_constant_plus_max_fraction =
-		dc_fixpt_from_fraction(
-			(1LL << (format->mantissa_bits + 1)) - 1,
-			1LL << format->mantissa_bits);
+		dc_fixpt_from_fraction((1LL << (format->mantissa_bits + 1)) - 1,
+				       1LL << format->mantissa_bits);
 
 	struct fixed31_32 mantiss;
 
-	if (dc_fixpt_eq(
-		value,
-		dc_fixpt_zero)) {
+	if (dc_fixpt_eq(value, dc_fixpt_zero)) {
 		*negative = false;
 		*mantissa = 0;
 		*exponenta = 0;
 		return true;
 	}
 
-	if (dc_fixpt_lt(
-		value,
-		dc_fixpt_zero)) {
+	if (dc_fixpt_lt(value, dc_fixpt_zero)) {
 		*negative = format->sign;
 		value = dc_fixpt_neg(value);
 	} else {
 		*negative = false;
 	}
 
-	if (dc_fixpt_lt(
-		value,
-		dc_fixpt_one)) {
+	if (dc_fixpt_lt(value, dc_fixpt_one)) {
 		uint32_t i = 1;
 
 		do {
 			value = dc_fixpt_shl(value, 1);
 			++i;
-		} while (dc_fixpt_lt(
-			value,
-			dc_fixpt_one));
+		} while (dc_fixpt_lt(value, dc_fixpt_one));
 
 		--i;
 
@@ -81,54 +71,40 @@ static bool build_custom_float(
 		}
 
 		*exponenta = exp_offset - i;
-	} else if (dc_fixpt_le(
-		mantissa_constant_plus_max_fraction,
-		value)) {
+	} else if (dc_fixpt_le(mantissa_constant_plus_max_fraction, value)) {
 		uint32_t i = 1;
 
 		do {
 			value = dc_fixpt_shr(value, 1);
 			++i;
-		} while (dc_fixpt_lt(
-			mantissa_constant_plus_max_fraction,
-			value));
+		} while (dc_fixpt_lt(mantissa_constant_plus_max_fraction, value));
 
 		*exponenta = exp_offset + i - 1;
 	} else {
 		*exponenta = exp_offset;
 	}
 
-	mantiss = dc_fixpt_sub(
-		value,
-		dc_fixpt_one);
+	mantiss = dc_fixpt_sub(value, dc_fixpt_one);
 
-	if (dc_fixpt_lt(
-			mantiss,
-			dc_fixpt_zero) ||
-		dc_fixpt_lt(
-			dc_fixpt_one,
-			mantiss))
+	if (dc_fixpt_lt(mantiss, dc_fixpt_zero) ||
+	    dc_fixpt_lt(dc_fixpt_one, mantiss))
 		mantiss = dc_fixpt_zero;
 	else
-		mantiss = dc_fixpt_shl(
-			mantiss,
-			format->mantissa_bits);
+		mantiss = dc_fixpt_shl(mantiss, format->mantissa_bits);
 
 	*mantissa = dc_fixpt_floor(mantiss);
 
 	return true;
 }
 
-static bool setup_custom_float(
-	const struct custom_float_format *format,
-	bool negative,
-	uint32_t mantissa,
-	uint32_t exponenta,
-	uint32_t *result)
+static bool setup_custom_float(const struct custom_float_format *format,
+			       bool negative,
+			       uint32_t mantissa,
+			       uint32_t exponenta,
+			       uint32_t *result)
 {
 	uint32_t i = 0;
 	uint32_t j = 0;
-
 	uint32_t value = 0;
 
 	/* verification code:
@@ -179,19 +155,19 @@ static bool setup_custom_float(
 	return true;
 }
 
-bool convert_to_custom_float_format(
-	struct fixed31_32 value,
-	const struct custom_float_format *format,
-	uint32_t *result)
+bool convert_to_custom_float_format(struct fixed31_32 value,
+				    const struct custom_float_format *format,
+				    uint32_t *result)
 {
 	uint32_t mantissa;
 	uint32_t exponenta;
 	bool negative;
 
-	return build_custom_float(
-		value, format, &negative, &mantissa, &exponenta) &&
-	setup_custom_float(
-		format, negative, mantissa, exponenta, result);
+	return build_custom_float(value, format, &negative, &mantissa, &exponenta) &&
+				  setup_custom_float(format,
+						     negative,
+						     mantissa,
+						     exponenta,
+						     result);
 }
 
-
diff --git a/drivers/gpu/drm/amd/display/dc/basics/dc_common.c b/drivers/gpu/drm/amd/display/dc/basics/dc_common.c
index b2fc4f8e6482..a51c2701da24 100644
--- a/drivers/gpu/drm/amd/display/dc/basics/dc_common.c
+++ b/drivers/gpu/drm/amd/display/dc/basics/dc_common.c
@@ -40,7 +40,8 @@ bool is_rgb_cspace(enum dc_color_space output_color_space)
 	case COLOR_SPACE_YCBCR709:
 	case COLOR_SPACE_YCBCR601_LIMITED:
 	case COLOR_SPACE_YCBCR709_LIMITED:
-	case COLOR_SPACE_2020_YCBCR:
+	case COLOR_SPACE_2020_YCBCR_LIMITED:
+	case COLOR_SPACE_2020_YCBCR_FULL:
 		return false;
 	default:
 		/* Add a case to switch */
diff --git a/drivers/gpu/drm/amd/display/dc/dml/calcs/dce_calcs.c b/drivers/gpu/drm/amd/display/dc/basics/dce_calcs.c
index 0100a6053ab6..4da5adab799c 100644
--- a/drivers/gpu/drm/amd/display/dc/dml/calcs/dce_calcs.c
+++ b/drivers/gpu/drm/amd/display/dc/basics/dce_calcs.c
@@ -23,8 +23,6 @@
  *
  */
 
-#include <linux/slab.h>
-
 #include "resource.h"
 #include "dm_services.h"
 #include "dce_calcs.h"
@@ -94,7 +92,7 @@ static void calculate_bandwidth(
 	const uint32_t s_high = 7;
 	const uint32_t dmif_chunk_buff_margin = 1;
 
-	uint32_t max_chunks_fbc_mode;
+	uint32_t max_chunks_fbc_mode = 0;
 	int32_t num_cursor_lines;
 
 	int32_t i, j, k;
@@ -571,7 +569,7 @@ static void calculate_bandwidth(
 				break;
 			}
 			data->lb_partitions[i] = bw_floor2(bw_div(data->lb_size_per_component[i], data->lb_line_pitch), bw_int_to_fixed(1));
-			/*clamp the partitions to the maxium number supported by the lb*/
+			/* clamp the partitions to the maximum number supported by the lb */
 			if ((surface_type[i] != bw_def_graphics || dceip->graphics_lb_nodownscaling_multi_line_prefetching == 1)) {
 				data->lb_partitions_max[i] = bw_int_to_fixed(10);
 			}
@@ -1138,7 +1136,7 @@ static void calculate_bandwidth(
 			}
 		}
 	}
-	data->total_dmifmc_urgent_trips = bw_ceil2(bw_div(data->total_requests_for_adjusted_dmif_size, (bw_add(dceip->dmif_request_buffer_size, bw_int_to_fixed(vbios->number_of_request_slots_gmc_reserves_for_dmif_per_channel * data->number_of_dram_channels)))), bw_int_to_fixed(1));
+	data->total_dmifmc_urgent_trips = bw_ceil2(bw_div(data->total_requests_for_adjusted_dmif_size, (bw_add(dceip->dmif_request_buffer_size, bw_int_to_fixed((uint64_t)vbios->number_of_request_slots_gmc_reserves_for_dmif_per_channel * data->number_of_dram_channels)))), bw_int_to_fixed(1));
 	data->total_dmifmc_urgent_latency = bw_mul(vbios->dmifmc_urgent_latency, data->total_dmifmc_urgent_trips);
 	data->total_display_reads_required_data = bw_int_to_fixed(0);
 	data->total_display_reads_required_dram_access_data = bw_int_to_fixed(0);
@@ -1395,7 +1393,7 @@ static void calculate_bandwidth(
 						if ((bw_mtn(data->dram_speed_change_margin, bw_int_to_fixed(0)) && bw_ltn(data->dram_speed_change_margin, bw_int_to_fixed(9999)))) {
 							/*determine the minimum dram clock change margin for each set of clock frequencies*/
 							data->min_dram_speed_change_margin[i][j] = bw_min2(data->min_dram_speed_change_margin[i][j], data->dram_speed_change_margin);
-							/*compute the maximum clock frequuency required for the dram clock change at each set of clock frequencies*/
+							/*compute the maximum clock frequency required for the dram clock change at each set of clock frequencies*/
 							data->dispclk_required_for_dram_speed_change_pipe[i][j] = bw_max2(bw_div(bw_div(bw_mul(data->src_pixels_for_first_output_pixel[k], dceip->display_pipe_throughput_factor), dceip->lb_write_pixels_per_dispclk), (bw_sub(bw_sub(bw_sub(data->maximum_latency_hiding_with_cursor[k], vbios->nbp_state_change_latency), data->dmif_burst_time[i][j]), data->dram_speed_change_line_source_transfer_time[k][i][j]))), bw_div(bw_div(bw_mul(data->src_pixels_for_last_output_pixel[k], dceip->display_pipe_throughput_factor), dceip->lb_write_pixels_per_dispclk), (bw_add(bw_sub(bw_sub(bw_sub(data->maximum_latency_hiding_with_cursor[k], vbios->nbp_state_change_latency), data->dmif_burst_time[i][j]), data->dram_speed_change_line_source_transfer_time[k][i][j]), data->active_time[k]))));
 							if ((bw_ltn(data->dispclk_required_for_dram_speed_change_pipe[i][j], vbios->high_voltage_max_dispclk))) {
 								data->display_pstate_change_enable[k] = 1;
@@ -1409,7 +1407,7 @@ static void calculate_bandwidth(
 						if ((bw_mtn(data->dram_speed_change_margin, bw_int_to_fixed(0)) && bw_ltn(data->dram_speed_change_margin, bw_int_to_fixed(9999)))) {
 							/*determine the minimum dram clock change margin for each display pipe*/
 							data->min_dram_speed_change_margin[i][j] = bw_min2(data->min_dram_speed_change_margin[i][j], data->dram_speed_change_margin);
-							/*compute the maximum clock frequuency required for the dram clock change at each set of clock frequencies*/
+							/*compute the maximum clock frequency required for the dram clock change at each set of clock frequencies*/
 							data->dispclk_required_for_dram_speed_change_pipe[i][j] = bw_max2(bw_div(bw_div(bw_mul(data->src_pixels_for_first_output_pixel[k], dceip->display_pipe_throughput_factor), dceip->lb_write_pixels_per_dispclk), (bw_sub(bw_sub(bw_sub(bw_sub(data->maximum_latency_hiding_with_cursor[k], vbios->nbp_state_change_latency), data->dmif_burst_time[i][j]), data->dram_speed_change_line_source_transfer_time[k][i][j]), data->mcifwr_burst_time[i][j]))), bw_div(bw_div(bw_mul(data->src_pixels_for_last_output_pixel[k], dceip->display_pipe_throughput_factor), dceip->lb_write_pixels_per_dispclk), (bw_add(bw_sub(bw_sub(bw_sub(bw_sub(data->maximum_latency_hiding_with_cursor[k], vbios->nbp_state_change_latency), data->dmif_burst_time[i][j]), data->dram_speed_change_line_source_transfer_time[k][i][j]), data->mcifwr_burst_time[i][j]), data->active_time[k]))));
 							if ((bw_ltn(data->dispclk_required_for_dram_speed_change_pipe[i][j], vbios->high_voltage_max_dispclk))) {
 								data->display_pstate_change_enable[k] = 1;
@@ -1855,7 +1853,7 @@ static void calculate_bandwidth(
 	/*compute total time to request one chunk from each active display pipe*/
 	for (i = 0; i <= maximum_number_of_surfaces - 1; i++) {
 		if (data->enable[i]) {
-			data->chunk_request_time = bw_add(data->chunk_request_time, (bw_div((bw_div(bw_int_to_fixed(pixels_per_chunk * data->bytes_per_pixel[i]), data->useful_bytes_per_request[i])), bw_min2(sclk[data->sclk_level], bw_div(data->dispclk, bw_int_to_fixed(2))))));
+			data->chunk_request_time = bw_add(data->chunk_request_time, (bw_div((bw_div(bw_int_to_fixed(pixels_per_chunk * (int64_t)data->bytes_per_pixel[i]), data->useful_bytes_per_request[i])), bw_min2(sclk[data->sclk_level], bw_div(data->dispclk, bw_int_to_fixed(2))))));
 		}
 	}
 	/*compute total time to request cursor data*/
@@ -3015,7 +3013,7 @@ static bool all_displays_in_sync(const struct pipe_ctx pipe[],
 	int i, num_active_pipes = 0;
 
 	for (i = 0; i < pipe_count; i++) {
-		if (!pipe[i].stream || pipe[i].top_pipe)
+		if (!resource_is_pipe_type(&pipe[i], OPP_HEAD))
 			continue;
 
 		active_pipes[num_active_pipes++] = &pipe[i];
diff --git a/drivers/gpu/drm/amd/display/dc/basics/fixpt31_32.c b/drivers/gpu/drm/amd/display/dc/basics/fixpt31_32.c
index 1726bdf89bae..6073cadde76c 100644
--- a/drivers/gpu/drm/amd/display/dc/basics/fixpt31_32.c
+++ b/drivers/gpu/drm/amd/display/dc/basics/fixpt31_32.c
@@ -51,8 +51,6 @@ static inline unsigned long long complete_integer_division_u64(
 {
 	unsigned long long result;
 
-	ASSERT(divisor);
-
 	result = div64_u64_rem(dividend, divisor, remainder);
 
 	return result;
@@ -140,8 +138,6 @@ struct fixed31_32 dc_fixpt_mul(struct fixed31_32 arg1, struct fixed31_32 arg2)
 
 	res.value = arg1_int * arg2_int;
 
-	ASSERT(res.value <= LONG_MAX);
-
 	res.value <<= FIXED31_32_BITS_PER_FRACTIONAL_PART;
 
 	tmp = arg1_int * arg2_fra;
@@ -185,8 +181,6 @@ struct fixed31_32 dc_fixpt_sqr(struct fixed31_32 arg)
 
 	res.value = arg_int * arg_int;
 
-	ASSERT(res.value <= LONG_MAX);
-
 	res.value <<= FIXED31_32_BITS_PER_FRACTIONAL_PART;
 
 	tmp = arg_int * arg_fra;
@@ -217,9 +211,6 @@ struct fixed31_32 dc_fixpt_recip(struct fixed31_32 arg)
 	 * @note
 	 * Good idea to use Newton's method
 	 */
-
-	ASSERT(arg.value);
-
 	return dc_fixpt_from_fraction(
 		dc_fixpt_one.value,
 		arg.value);
@@ -293,7 +284,7 @@ struct fixed31_32 dc_fixpt_cos(struct fixed31_32 arg)
 				dc_fixpt_mul(
 					square,
 					res),
-				n * (n - 1)));
+				(long long)n * (n - 1)));
 
 		n -= 2;
 	} while (n != 0);
@@ -490,3 +481,30 @@ int dc_fixpt_s4d19(struct fixed31_32 arg)
 	else
 		return ux_dy(arg.value, 4, 19);
 }
+
+struct fixed31_32 dc_fixpt_from_ux_dy(unsigned int value,
+	unsigned int integer_bits,
+	unsigned int fractional_bits)
+{
+	struct fixed31_32 fixpt_value = dc_fixpt_zero;
+	struct fixed31_32 fixpt_int_value = dc_fixpt_zero;
+	long long frac_mask = ((long long)1 << (long long)integer_bits) - 1;
+
+	fixpt_value.value = (long long)value << (FIXED31_32_BITS_PER_FRACTIONAL_PART - fractional_bits);
+	frac_mask = frac_mask << fractional_bits;
+	fixpt_int_value.value = value & frac_mask;
+	fixpt_int_value.value <<= (FIXED31_32_BITS_PER_FRACTIONAL_PART - fractional_bits);
+	fixpt_value.value |= fixpt_int_value.value;
+	return fixpt_value;
+}
+
+struct fixed31_32 dc_fixpt_from_int_dy(unsigned int int_value,
+	unsigned int frac_value,
+	unsigned int integer_bits,
+	unsigned int fractional_bits)
+{
+	struct fixed31_32 fixpt_value = dc_fixpt_from_int(int_value);
+
+	fixpt_value.value |= (long long)frac_value << (FIXED31_32_BITS_PER_FRACTIONAL_PART - fractional_bits);
+	return fixpt_value;
+}
diff --git a/drivers/gpu/drm/amd/display/dc/basics/vector.c b/drivers/gpu/drm/amd/display/dc/basics/vector.c
index 84aeccf36b4b..b413a672c2c0 100644
--- a/drivers/gpu/drm/amd/display/dc/basics/vector.c
+++ b/drivers/gpu/drm/amd/display/dc/basics/vector.c
@@ -50,12 +50,11 @@ bool dal_vector_construct(
 	return true;
 }
 
-static bool dal_vector_presized_costruct(
-	struct vector *vector,
-	struct dc_context *ctx,
-	uint32_t count,
-	void *initial_value,
-	uint32_t struct_size)
+static bool dal_vector_presized_costruct(struct vector *vector,
+					 struct dc_context *ctx,
+					 uint32_t count,
+					 void *initial_value,
+					 uint32_t struct_size)
 {
 	uint32_t i;
 
@@ -171,7 +170,7 @@ bool dal_vector_remove_at_index(
 		memmove(
 			vector->container + (index * vector->struct_size),
 			vector->container + ((index + 1) * vector->struct_size),
-			(vector->count - index - 1) * vector->struct_size);
+			(size_t)(vector->count - index - 1) * vector->struct_size);
 	vector->count -= 1;
 
 	return true;
@@ -220,7 +219,7 @@ bool dal_vector_insert_at(
 		memmove(
 			insert_address + vector->struct_size,
 			insert_address,
-			vector->struct_size * (vector->count - position));
+			(size_t)vector->struct_size * (vector->count - position));
 
 	memmove(
 		insert_address,
@@ -272,7 +271,7 @@ struct vector *dal_vector_clone(
 
 	/* copy vector's data */
 	memmove(vec_cloned->container, vector->container,
-			vec_cloned->struct_size * vec_cloned->capacity);
+			(size_t)vec_cloned->struct_size * vec_cloned->capacity);
 
 	return vec_cloned;
 }
diff --git a/drivers/gpu/drm/amd/display/dc/bios/bios_parser.c b/drivers/gpu/drm/amd/display/dc/bios/bios_parser.c
index 27af9d3c2b73..d1471f34e419 100644
--- a/drivers/gpu/drm/amd/display/dc/bios/bios_parser.c
+++ b/drivers/gpu/drm/amd/display/dc/bios/bios_parser.c
@@ -44,8 +44,6 @@
 
 #include "bios_parser_common.h"
 
-#include "dc.h"
-
 #define THREE_PERCENT_OF_10000 300
 
 #define LAST_RECORD_TYPE 0xff
@@ -69,7 +67,9 @@ static ATOM_HPD_INT_RECORD *get_hpd_record(struct bios_parser *bp,
 	ATOM_OBJECT *object);
 static struct device_id device_type_from_device_id(uint16_t device_id);
 static uint32_t signal_to_ss_id(enum as_signal_type signal);
-static uint32_t get_support_mask_for_device_id(struct device_id device_id);
+static uint32_t get_support_mask_for_device_id(
+	enum dal_device_type device_type,
+	uint32_t enum_id);
 static ATOM_ENCODER_CAP_RECORD_V2 *get_encoder_cap_record(
 	struct bios_parser *bp,
 	ATOM_OBJECT *object);
@@ -96,7 +96,7 @@ struct dc_bios *bios_parser_create(
 	struct bp_init_data *init,
 	enum dce_version dce_version)
 {
-	struct bios_parser *bp = NULL;
+	struct bios_parser *bp;
 
 	bp = kzalloc(sizeof(struct bios_parser), GFP_KERNEL);
 	if (!bp)
@@ -176,11 +176,8 @@ static struct graphics_object_id bios_parser_get_connector_id(
 		return object_id;
 	}
 
-	if (tbl->ucNumberOfObjects <= i) {
-		dm_error("Can't find connector id %d in connector table of size %d.\n",
-			 i, tbl->ucNumberOfObjects);
+	if (tbl->ucNumberOfObjects <= i)
 		return object_id;
-	}
 
 	id = le16_to_cpu(tbl->asObjects[i].usObjectID);
 	object_id = object_id_from_bios_object_id(id);
@@ -446,6 +443,7 @@ static enum bp_result get_firmware_info_v1_4(
 		le32_to_cpu(firmware_info->ulMinPixelClockPLL_Output) * 10;
 	info->pll_info.max_output_pxl_clk_pll_frequency =
 		le32_to_cpu(firmware_info->ulMaxPixelClockPLL_Output) * 10;
+	info->max_pixel_clock = le16_to_cpu(firmware_info->usMaxPixelClock) * 10;
 
 	if (firmware_info->usFirmwareCapability.sbfAccess.MemoryClockSS_Support)
 		/* Since there is no information on the SS, report conservative
@@ -502,6 +500,7 @@ static enum bp_result get_firmware_info_v2_1(
 	info->external_clock_source_frequency_for_dp =
 		le16_to_cpu(firmwareInfo->usUniphyDPModeExtClkFreq) * 10;
 	info->min_allowed_bl_level = firmwareInfo->ucMinAllowedBL_Level;
+	info->max_pixel_clock = le16_to_cpu(firmwareInfo->usMaxPixelClock) * 10;
 
 	/* There should be only one entry in the SS info table for Memory Clock
 	 */
@@ -667,6 +666,9 @@ static enum bp_result get_ss_info_v3_1(
 	ss_table_header_include = ((ATOM_ASIC_INTERNAL_SS_INFO_V3 *) bios_get_image(&bp->base,
 				DATA_TABLES(ASIC_InternalSS_Info),
 				struct_size(ss_table_header_include, asSpreadSpectrum, 1)));
+	if (!ss_table_header_include)
+		return BP_RESULT_UNSUPPORTED;
+
 	table_size =
 		(le16_to_cpu(ss_table_header_include->sHeader.usStructureSize)
 				- sizeof(ATOM_COMMON_TABLE_HEADER))
@@ -738,18 +740,94 @@ static enum bp_result bios_parser_transmitter_control(
 	return bp->cmd_tbl.transmitter_control(bp, cntl);
 }
 
+static enum bp_result bios_parser_select_crtc_source(
+	struct dc_bios *dcb,
+	struct bp_crtc_source_select *bp_params)
+{
+	struct bios_parser *bp = BP_FROM_DCB(dcb);
+
+	if (!bp->cmd_tbl.select_crtc_source)
+		return BP_RESULT_FAILURE;
+
+	return bp->cmd_tbl.select_crtc_source(bp, bp_params);
+}
+
 static enum bp_result bios_parser_encoder_control(
 	struct dc_bios *dcb,
 	struct bp_encoder_control *cntl)
 {
 	struct bios_parser *bp = BP_FROM_DCB(dcb);
 
+	if (cntl->engine_id == ENGINE_ID_DACA) {
+		if (!bp->cmd_tbl.dac1_encoder_control)
+			return BP_RESULT_FAILURE;
+
+		return bp->cmd_tbl.dac1_encoder_control(
+			bp, cntl->action == ENCODER_CONTROL_ENABLE,
+			cntl->pixel_clock, ATOM_DAC1_PS2);
+	} else if (cntl->engine_id == ENGINE_ID_DACB) {
+		if (!bp->cmd_tbl.dac2_encoder_control)
+			return BP_RESULT_FAILURE;
+
+		return bp->cmd_tbl.dac2_encoder_control(
+			bp, cntl->action == ENCODER_CONTROL_ENABLE,
+			cntl->pixel_clock, ATOM_DAC1_PS2);
+	}
+
 	if (!bp->cmd_tbl.dig_encoder_control)
 		return BP_RESULT_FAILURE;
 
 	return bp->cmd_tbl.dig_encoder_control(bp, cntl);
 }
 
+static enum bp_result bios_parser_dac_load_detection(
+	struct dc_bios *dcb,
+	enum engine_id engine_id,
+	enum dal_device_type device_type,
+	uint32_t enum_id)
+{
+	struct bios_parser *bp = BP_FROM_DCB(dcb);
+	struct dc_context *ctx = dcb->ctx;
+	struct bp_load_detection_parameters bp_params = {0};
+	enum bp_result bp_result;
+	uint32_t bios_0_scratch;
+	uint32_t device_id_mask = 0;
+
+	bp_params.engine_id = engine_id;
+	bp_params.device_id = get_support_mask_for_device_id(device_type, enum_id);
+
+	if (engine_id != ENGINE_ID_DACA &&
+	    engine_id != ENGINE_ID_DACB)
+		return BP_RESULT_UNSUPPORTED;
+
+	if (!bp->cmd_tbl.dac_load_detection)
+		return BP_RESULT_UNSUPPORTED;
+
+	if (bp_params.device_id == ATOM_DEVICE_CRT1_SUPPORT)
+		device_id_mask = ATOM_S0_CRT1_MASK;
+	else if (bp_params.device_id == ATOM_DEVICE_CRT2_SUPPORT)
+		device_id_mask = ATOM_S0_CRT2_MASK;
+	else
+		return BP_RESULT_UNSUPPORTED;
+
+	/* BIOS will write the detected devices to BIOS_SCRATCH_0, clear corresponding bit */
+	bios_0_scratch = dm_read_reg(ctx, bp->base.regs->BIOS_SCRATCH_0);
+	bios_0_scratch &= ~device_id_mask;
+	dm_write_reg(ctx, bp->base.regs->BIOS_SCRATCH_0, bios_0_scratch);
+
+	bp_result = bp->cmd_tbl.dac_load_detection(bp, &bp_params);
+
+	if (bp_result != BP_RESULT_OK)
+		return bp_result;
+
+	bios_0_scratch = dm_read_reg(ctx, bp->base.regs->BIOS_SCRATCH_0);
+
+	if (bios_0_scratch & device_id_mask)
+		return BP_RESULT_OK;
+
+	return BP_RESULT_FAILURE;
+}
+
 static enum bp_result bios_parser_adjust_pixel_clock(
 	struct dc_bios *dcb,
 	struct bp_adjust_pixel_clock_parameters *bp_params)
@@ -860,7 +938,7 @@ static bool bios_parser_is_device_id_supported(
 {
 	struct bios_parser *bp = BP_FROM_DCB(dcb);
 
-	uint32_t mask = get_support_mask_for_device_id(id);
+	uint32_t mask = get_support_mask_for_device_id(id.device_type, id.enum_id);
 
 	return (le16_to_cpu(bp->object_info_tbl.v1_1->usDeviceSupport) & mask) != 0;
 }
@@ -1036,6 +1114,8 @@ static enum bp_result get_ss_info_from_internal_ss_info_tbl_V2_1(
 				&bp->base,
 				DATA_TABLES(ASIC_InternalSS_Info),
 				struct_size(header, asSpreadSpectrum, 1)));
+	if (!header)
+		return result;
 
 	memset(info, 0, sizeof(struct spread_spectrum_info));
 
@@ -1109,6 +1189,8 @@ static enum bp_result get_ss_info_from_ss_info_table(
 	get_atom_data_table_revision(header, &revision);
 
 	tbl = GET_IMAGE(ATOM_SPREAD_SPECTRUM_INFO, DATA_TABLES(SS_Info));
+	if (!tbl)
+		return result;
 
 	if (1 != revision.major || 2 > revision.minor)
 		return result;
@@ -1636,6 +1718,8 @@ static uint32_t get_ss_entry_number_from_ss_info_tbl(
 
 	tbl = GET_IMAGE(ATOM_SPREAD_SPECTRUM_INFO,
 			DATA_TABLES(SS_Info));
+	if (!tbl)
+		return number;
 
 	if (1 != revision.major || 2 > revision.minor)
 		return number;
@@ -1718,6 +1802,8 @@ static uint32_t get_ss_entry_number_from_internal_ss_info_tbl_v2_1(
 				&bp->base,
 				DATA_TABLES(ASIC_InternalSS_Info),
 				struct_size(header_include, asSpreadSpectrum, 1)));
+	if (!header_include)
+		return 0;
 
 	size = (le16_to_cpu(header_include->sHeader.usStructureSize)
 			- sizeof(ATOM_COMMON_TABLE_HEADER))
@@ -1731,6 +1817,7 @@ static uint32_t get_ss_entry_number_from_internal_ss_info_tbl_v2_1(
 
 	return 0;
 }
+
 /**
  * get_ss_entry_number_from_internal_ss_info_tbl_V3_1
  * Get Number of SpreadSpectrum Entry from the ASIC_InternalSS_Info table of
@@ -1756,6 +1843,9 @@ static uint32_t get_ss_entry_number_from_internal_ss_info_tbl_V3_1(
 	header_include = ((ATOM_ASIC_INTERNAL_SS_INFO_V3 *) bios_get_image(&bp->base,
 				DATA_TABLES(ASIC_InternalSS_Info),
 				struct_size(header_include, asSpreadSpectrum, 1)));
+	if (!header_include)
+		return number;
+
 	size = (le16_to_cpu(header_include->sHeader.usStructureSize) -
 			sizeof(ATOM_COMMON_TABLE_HEADER)) /
 					sizeof(ATOM_ASIC_SS_ASSIGNMENT_V3);
@@ -2139,11 +2229,10 @@ static uint32_t signal_to_ss_id(enum as_signal_type signal)
 	return clk_id_ss;
 }
 
-static uint32_t get_support_mask_for_device_id(struct device_id device_id)
+static uint32_t get_support_mask_for_device_id(
+	enum dal_device_type device_type,
+	uint32_t enum_id)
 {
-	enum dal_device_type device_type = device_id.device_type;
-	uint32_t enum_id = device_id.enum_id;
-
 	switch (device_type) {
 	case DEVICE_TYPE_LCD:
 		switch (enum_id) {
@@ -2371,10 +2460,10 @@ static enum bp_result get_integrated_info_v8(
 }
 
 /*
- * get_integrated_info_v8
+ * get_integrated_info_v9
  *
  * @brief
- * Get V8 integrated BIOS information
+ * Get V9 integrated BIOS information
  *
  * @param
  * bios_parser *bp - [in]BIOS parser handler to get master data table
@@ -2552,8 +2641,8 @@ static enum bp_result construct_integrated_info(
 
 	/* Sort voltage table from low to high*/
 	if (result == BP_RESULT_OK) {
-		uint32_t i;
-		uint32_t j;
+		int32_t i;
+		int32_t j;
 
 		for (i = 1; i < NUMBER_OF_DISP_CLK_VOLTAGE; ++i) {
 			for (j = i; j > 0; --j) {
@@ -2576,7 +2665,7 @@ static struct integrated_info *bios_parser_create_integrated_info(
 	struct dc_bios *dcb)
 {
 	struct bios_parser *bp = BP_FROM_DCB(dcb);
-	struct integrated_info *info = NULL;
+	struct integrated_info *info;
 
 	info = kzalloc(sizeof(struct integrated_info), GFP_KERNEL);
 
@@ -2593,11 +2682,10 @@ static struct integrated_info *bios_parser_create_integrated_info(
 	return NULL;
 }
 
-static enum bp_result update_slot_layout_info(
-	struct dc_bios *dcb,
-	unsigned int i,
-	struct slot_layout_info *slot_layout_info,
-	unsigned int record_offset)
+static enum bp_result update_slot_layout_info(struct dc_bios *dcb,
+					      unsigned int i,
+					      struct slot_layout_info *slot_layout_info,
+					      unsigned int record_offset)
 {
 	unsigned int j;
 	struct bios_parser *bp;
@@ -2696,10 +2784,9 @@ static enum bp_result update_slot_layout_info(
 }
 
 
-static enum bp_result get_bracket_layout_record(
-	struct dc_bios *dcb,
-	unsigned int bracket_layout_id,
-	struct slot_layout_info *slot_layout_info)
+static enum bp_result get_bracket_layout_record(struct dc_bios *dcb,
+						unsigned int bracket_layout_id,
+						struct slot_layout_info *slot_layout_info)
 {
 	unsigned int i;
 	unsigned int record_offset;
@@ -2748,6 +2835,7 @@ static enum bp_result bios_get_board_layout_info(
 	struct board_layout_info *board_layout_info)
 {
 	unsigned int i;
+	struct bios_parser *bp;
 	enum bp_result record_result;
 
 	const unsigned int slot_index_to_vbios_id[MAX_BOARD_SLOTS] = {
@@ -2756,6 +2844,8 @@ static enum bp_result bios_get_board_layout_info(
 		0, 0
 	};
 
+	bp = BP_FROM_DCB(dcb);
+
 	if (board_layout_info == NULL) {
 		DC_LOG_DETECTION_EDID_PARSER("Invalid board_layout_info\n");
 		return BP_RESULT_BADINPUT;
@@ -2818,8 +2908,12 @@ static const struct dc_vbios_funcs vbios_funcs = {
 	.is_device_id_supported = bios_parser_is_device_id_supported,
 
 	/* COMMANDS */
+	.select_crtc_source = bios_parser_select_crtc_source,
+
 	.encoder_control = bios_parser_encoder_control,
 
+	.dac_load_detection = bios_parser_dac_load_detection,
+
 	.transmitter_control = bios_parser_transmitter_control,
 
 	.enable_crtc = bios_parser_enable_crtc,
diff --git a/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c b/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c
index cce47d3f1a13..550a9f1d03f8 100644
--- a/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c
+++ b/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c
@@ -75,6 +75,10 @@ static enum bp_result get_firmware_info_v3_4(
 	struct bios_parser *bp,
 	struct dc_firmware_info *info);
 
+static enum bp_result get_firmware_info_v3_5(
+	struct bios_parser *bp,
+	struct dc_firmware_info *info);
+
 static struct atom_hpd_int_record *get_hpd_record(struct bios_parser *bp,
 		struct atom_display_object_path_v2 *object);
 
@@ -340,9 +344,8 @@ static struct atom_display_object_path_v2 *get_bios_object(
 }
 
 /* from graphics_object_id, find display path which includes the object_id */
-static struct atom_display_object_path_v3 *get_bios_object_from_path_v3(
-	struct bios_parser *bp,
-	struct graphics_object_id id)
+static struct atom_display_object_path_v3 *get_bios_object_from_path_v3(struct bios_parser *bp,
+									struct graphics_object_id id)
 {
 	unsigned int i;
 	struct graphics_object_id obj_id = {0};
@@ -356,7 +359,7 @@ static struct atom_display_object_path_v3 *get_bios_object_from_path_v3(
 					&& id.enum_id == obj_id.enum_id)
 				return &bp->object_info_tbl.v1_5->display_path[i];
 		}
-        break;
+	break;
 
 	case OBJECT_TYPE_CONNECTOR:
 	case OBJECT_TYPE_GENERIC:
@@ -370,7 +373,7 @@ static struct atom_display_object_path_v3 *get_bios_object_from_path_v3(
 					&& id.enum_id == obj_id.enum_id)
 				return &bp->object_info_tbl.v1_5->display_path[i];
 		}
-        break;
+	break;
 
 	default:
 		return NULL;
@@ -406,16 +409,16 @@ static enum bp_result bios_parser_get_i2c_info(struct dc_bios *dcb,
 	}
 
 	switch (bp->object_info_tbl.revision.minor) {
-	    case 4:
-	    default:
-	        object = get_bios_object(bp, id);
+	case 4:
+	default:
+		object = get_bios_object(bp, id);
 
-	        if (!object)
-				return BP_RESULT_BADINPUT;
+		if (!object)
+			return BP_RESULT_BADINPUT;
 
-	        offset = object->disp_recordoffset + bp->object_info_tbl_offset;
-	        break;
-	    case 5:
+		offset = object->disp_recordoffset + bp->object_info_tbl_offset;
+		break;
+	case 5:
 		object_path_v3 = get_bios_object_from_path_v3(bp, id);
 
 		if (!object_path_v3)
@@ -521,9 +524,8 @@ static enum bp_result get_gpio_i2c_info(
 	return BP_RESULT_OK;
 }
 
-static struct atom_hpd_int_record *get_hpd_record_for_path_v3(
-	struct bios_parser *bp,
-	struct atom_display_object_path_v3 *object)
+static struct atom_hpd_int_record *get_hpd_record_for_path_v3(struct bios_parser *bp,
+							      struct atom_display_object_path_v3 *object)
 {
 	struct atom_common_record_header *header;
 	uint32_t offset;
@@ -570,17 +572,16 @@ static enum bp_result bios_parser_get_hpd_info(
 		return BP_RESULT_BADINPUT;
 
 	switch (bp->object_info_tbl.revision.minor) {
-	    case 4:
-	    default:
-	        object = get_bios_object(bp, id);
+	case 4:
+	default:
+		object = get_bios_object(bp, id);
 
 		if (!object)
 			return BP_RESULT_BADINPUT;
 
-	        record = get_hpd_record(bp, object);
-
-	        break;
-	    case 5:
+		record = get_hpd_record(bp, object);
+		break;
+	case 5:
 		object_path_v3 = get_bios_object_from_path_v3(bp, id);
 
 		if (!object_path_v3)
@@ -774,20 +775,20 @@ static enum bp_result bios_parser_get_device_tag(
 		return BP_RESULT_BADINPUT;
 
 	switch (bp->object_info_tbl.revision.minor) {
-	    case 4:
-	    default:
+	case 4:
+	default:
 	        /* getBiosObject will return MXM object */
-	        object = get_bios_object(bp, connector_object_id);
+		object = get_bios_object(bp, connector_object_id);
 
 		if (!object) {
 			BREAK_TO_DEBUGGER(); /* Invalid object id */
 			return BP_RESULT_BADINPUT;
 		}
 
-	        info->acpi_device = 0; /* BIOS no longer provides this */
-	        info->dev_id = device_type_from_device_id(object->device_tag);
-	        break;
-	    case 5:
+		info->acpi_device = 0; /* BIOS no longer provides this */
+		info->dev_id = device_type_from_device_id(object->device_tag);
+		break;
+	case 5:
 		object_path_v3 = get_bios_object_from_path_v3(bp, connector_object_id);
 
 		if (!object_path_v3) {
@@ -1017,13 +1018,20 @@ static enum bp_result get_ss_info_v4_5(
 		DC_LOG_BIOS("AS_SIGNAL_TYPE_HDMI ss_percentage: %d\n", ss_info->spread_spectrum_percentage);
 		break;
 	case AS_SIGNAL_TYPE_DISPLAY_PORT:
-		ss_info->spread_spectrum_percentage =
+		if (bp->base.integrated_info) {
+			DC_LOG_BIOS("gpuclk_ss_percentage (unit of 0.001 percent): %d\n", bp->base.integrated_info->gpuclk_ss_percentage);
+			ss_info->spread_spectrum_percentage =
+					bp->base.integrated_info->gpuclk_ss_percentage;
+			ss_info->type.CENTER_MODE =
+					bp->base.integrated_info->gpuclk_ss_type;
+		} else {
+			ss_info->spread_spectrum_percentage =
 				disp_cntl_tbl->dp_ss_percentage;
-		ss_info->spread_spectrum_range =
+			ss_info->spread_spectrum_range =
 				disp_cntl_tbl->dp_ss_rate_10hz * 10;
-		if (disp_cntl_tbl->dp_ss_mode & ATOM_SS_CENTRE_SPREAD_MODE)
-			ss_info->type.CENTER_MODE = true;
-
+			if (disp_cntl_tbl->dp_ss_mode & ATOM_SS_CENTRE_SPREAD_MODE)
+				ss_info->type.CENTER_MODE = true;
+		}
 		DC_LOG_BIOS("AS_SIGNAL_TYPE_DISPLAY_PORT ss_percentage: %d\n", ss_info->spread_spectrum_percentage);
 		break;
 	case AS_SIGNAL_TYPE_GPU_PLL:
@@ -1472,10 +1480,10 @@ static enum bp_result get_embedded_panel_info_v2_1(
 	/* not provided by VBIOS */
 	info->lcd_timing.misc_info.HORIZONTAL_CUT_OFF = 0;
 
-	info->lcd_timing.misc_info.H_SYNC_POLARITY = ~(uint32_t) (lvds->lcd_timing.miscinfo
-			& ATOM_HSYNC_POLARITY);
-	info->lcd_timing.misc_info.V_SYNC_POLARITY = ~(uint32_t) (lvds->lcd_timing.miscinfo
-			& ATOM_VSYNC_POLARITY);
+	info->lcd_timing.misc_info.H_SYNC_POLARITY = !(lvds->lcd_timing.miscinfo &
+						       ATOM_HSYNC_POLARITY);
+	info->lcd_timing.misc_info.V_SYNC_POLARITY = !(lvds->lcd_timing.miscinfo &
+						       ATOM_VSYNC_POLARITY);
 
 	/* not provided by VBIOS */
 	info->lcd_timing.misc_info.VERTICAL_CUT_OFF = 0;
@@ -1582,16 +1590,14 @@ static bool bios_parser_is_device_id_supported(
 	uint32_t mask = get_support_mask_for_device_id(id);
 
 	switch (bp->object_info_tbl.revision.minor) {
-	    case 4:
-	    default:
-	        return (le16_to_cpu(bp->object_info_tbl.v1_4->supporteddevices) & mask) != 0;
-			break;
-	    case 5:
-			return (le16_to_cpu(bp->object_info_tbl.v1_5->supporteddevices) & mask) != 0;
-			break;
+	case 4:
+	default:
+		return (le16_to_cpu(bp->object_info_tbl.v1_4->supporteddevices) & mask) != 0;
+		break;
+	case 5:
+		return (le16_to_cpu(bp->object_info_tbl.v1_5->supporteddevices) & mask) != 0;
+		break;
 	}
-
-	return false;
 }
 
 static uint32_t bios_parser_get_ss_entry_number(
@@ -1694,7 +1700,7 @@ static enum bp_result bios_parser_enable_disp_power_gating(
 static enum bp_result bios_parser_enable_lvtma_control(
 	struct dc_bios *dcb,
 	uint8_t uc_pwr_on,
-	uint8_t panel_instance,
+	uint8_t pwrseq_instance,
 	uint8_t bypass_panel_control_wait)
 {
 	struct bios_parser *bp = BP_FROM_DCB(dcb);
@@ -1702,7 +1708,7 @@ static enum bp_result bios_parser_enable_lvtma_control(
 	if (!bp->cmd_tbl.enable_lvtma_control)
 		return BP_RESULT_FAILURE;
 
-	return bp->cmd_tbl.enable_lvtma_control(bp, uc_pwr_on, panel_instance, bypass_panel_control_wait);
+	return bp->cmd_tbl.enable_lvtma_control(bp, uc_pwr_on, pwrseq_instance, bypass_panel_control_wait);
 }
 
 static bool bios_parser_is_accelerated_mode(
@@ -1725,15 +1731,6 @@ static void bios_parser_set_scratch_critical_state(
 	bios_set_scratch_critical_state(dcb, state);
 }
 
-struct atom_dig_transmitter_info_header_v5_3 {
-    struct atom_common_table_header table_header;
-    uint16_t dpphy_hdmi_settings_offset;
-    uint16_t dpphy_dvi_settings_offset;
-    uint16_t dpphy_dp_setting_table_offset;
-    uint16_t uniphy_xbar_settings_v2_table_offset;
-    uint16_t dpphy_internal_reg_overide_offset;
-};
-
 static enum bp_result bios_parser_get_firmware_info(
 	struct dc_bios *dcb,
 	struct dc_firmware_info *info)
@@ -1757,10 +1754,13 @@ static enum bp_result bios_parser_get_firmware_info(
 			case 2:
 			case 3:
 				result = get_firmware_info_v3_2(bp, info);
-                                break;
+				break;
 			case 4:
 				result = get_firmware_info_v3_4(bp, info);
 				break;
+			case 5:
+				result = get_firmware_info_v3_5(bp, info);
+				break;
 			default:
 				break;
 			}
@@ -1778,6 +1778,7 @@ static enum bp_result get_firmware_info_v3_1(
 	struct dc_firmware_info *info)
 {
 	struct atom_firmware_info_v3_1 *firmware_info;
+	struct atom_firmware_info_v3_2 *firmware_info32;
 	struct atom_display_controller_info_v4_1 *dce_info = NULL;
 
 	if (!info)
@@ -1785,11 +1786,13 @@ static enum bp_result get_firmware_info_v3_1(
 
 	firmware_info = GET_IMAGE(struct atom_firmware_info_v3_1,
 			DATA_TABLES(firmwareinfo));
+	firmware_info32 = GET_IMAGE(struct atom_firmware_info_v3_2,
+			DATA_TABLES(firmwareinfo));
 
 	dce_info = GET_IMAGE(struct atom_display_controller_info_v4_1,
 			DATA_TABLES(dce_info));
 
-	if (!firmware_info || !dce_info)
+	if (!firmware_info || !firmware_info32 || !dce_info)
 		return BP_RESULT_BADBIOSTABLE;
 
 	memset(info, 0, sizeof(*info));
@@ -1817,7 +1820,15 @@ static enum bp_result get_firmware_info_v3_1(
 				bp->cmd_tbl.get_smu_clock_info(bp, SMU9_SYSPLL0_ID) * 10;
 	}
 
-	info->oem_i2c_present = false;
+	/* These fields are marked as reserved in v3_1, but they appear to be populated
+	 * properly.
+	 */
+	if (firmware_info32 && firmware_info32->board_i2c_feature_id == 0x2) {
+		info->oem_i2c_present = true;
+		info->oem_i2c_obj_id = firmware_info32->board_i2c_feature_gpio_id;
+	} else {
+		info->oem_i2c_present = false;
+	}
 
 	return BP_RESULT_OK;
 }
@@ -1855,19 +1866,21 @@ static enum bp_result get_firmware_info_v3_2(
 		/* Vega12 */
 		smu_info_v3_2 = GET_IMAGE(struct atom_smu_info_v3_2,
 							DATA_TABLES(smu_info));
-		DC_LOG_BIOS("gpuclk_ss_percentage (unit of 0.001 percent): %d\n", smu_info_v3_2->gpuclk_ss_percentage);
 		if (!smu_info_v3_2)
 			return BP_RESULT_BADBIOSTABLE;
 
+		DC_LOG_BIOS("gpuclk_ss_percentage (unit of 0.001 percent): %d\n", smu_info_v3_2->gpuclk_ss_percentage);
+
 		info->default_engine_clk = smu_info_v3_2->bootup_dcefclk_10khz * 10;
 	} else if (revision.minor == 3) {
 		/* Vega20 */
 		smu_info_v3_3 = GET_IMAGE(struct atom_smu_info_v3_3,
 							DATA_TABLES(smu_info));
-		DC_LOG_BIOS("gpuclk_ss_percentage (unit of 0.001 percent): %d\n", smu_info_v3_3->gpuclk_ss_percentage);
 		if (!smu_info_v3_3)
 			return BP_RESULT_BADBIOSTABLE;
 
+		DC_LOG_BIOS("gpuclk_ss_percentage (unit of 0.001 percent): %d\n", smu_info_v3_3->gpuclk_ss_percentage);
+
 		info->default_engine_clk = smu_info_v3_3->bootup_dcefclk_10khz * 10;
 	}
 
@@ -2049,6 +2062,63 @@ static enum bp_result get_firmware_info_v3_4(
 	return BP_RESULT_OK;
 }
 
+static enum bp_result get_firmware_info_v3_5(
+	struct bios_parser *bp,
+	struct dc_firmware_info *info)
+{
+	struct atom_firmware_info_v3_5 *firmware_info;
+	struct atom_common_table_header *header;
+	struct atom_data_revision revision;
+	struct atom_display_controller_info_v4_5 *dce_info_v4_5 = NULL;
+
+	if (!info)
+		return BP_RESULT_BADINPUT;
+
+	firmware_info = GET_IMAGE(struct atom_firmware_info_v3_5,
+			DATA_TABLES(firmwareinfo));
+
+	if (!firmware_info)
+		return BP_RESULT_BADBIOSTABLE;
+
+	memset(info, 0, sizeof(*info));
+
+	if (firmware_info->board_i2c_feature_id == 0x2) {
+		info->oem_i2c_present = true;
+		info->oem_i2c_obj_id = firmware_info->board_i2c_feature_gpio_id;
+	} else {
+		info->oem_i2c_present = false;
+	}
+
+	header = GET_IMAGE(struct atom_common_table_header,
+					DATA_TABLES(dce_info));
+
+	get_atom_data_table_revision(header, &revision);
+
+	switch (revision.major) {
+	case 4:
+		switch (revision.minor) {
+		case 5:
+			dce_info_v4_5 = GET_IMAGE(struct atom_display_controller_info_v4_5,
+							DATA_TABLES(dce_info));
+
+			if (!dce_info_v4_5)
+				return BP_RESULT_BADBIOSTABLE;
+
+			 /* 100MHz expected */
+			info->pll_info.crystal_frequency = dce_info_v4_5->dce_refclk_10khz * 10;
+			break;
+		default:
+			break;
+		}
+		break;
+	default:
+		break;
+	}
+
+
+	return BP_RESULT_OK;
+}
+
 static enum bp_result bios_parser_get_encoder_cap_info(
 	struct dc_bios *dcb,
 	struct graphics_object_id object_id,
@@ -2175,9 +2245,8 @@ static struct atom_disp_connector_caps_record *get_disp_connector_caps_record(
 	return NULL;
 }
 
-static struct atom_connector_caps_record *get_connector_caps_record(
-	struct bios_parser *bp,
-	struct atom_display_object_path_v3 *object)
+static struct atom_connector_caps_record *get_connector_caps_record(struct bios_parser *bp,
+								    struct atom_display_object_path_v3 *object)
 {
 	struct atom_common_record_header *header;
 	uint32_t offset;
@@ -2218,33 +2287,31 @@ static enum bp_result bios_parser_get_disp_connector_caps_info(
 {
 	struct bios_parser *bp = BP_FROM_DCB(dcb);
 	struct atom_display_object_path_v2 *object;
-
 	struct atom_display_object_path_v3 *object_path_v3;
 	struct atom_connector_caps_record *record_path_v3;
-
 	struct atom_disp_connector_caps_record *record = NULL;
 
 	if (!info)
 		return BP_RESULT_BADINPUT;
 
 	switch (bp->object_info_tbl.revision.minor) {
-	    case 4:
-	    default:
-		    object = get_bios_object(bp, object_id);
-
-		    if (!object)
-			    return BP_RESULT_BADINPUT;
-
-		    record = get_disp_connector_caps_record(bp, object);
-		    if (!record)
-			    return BP_RESULT_NORECORD;
-
-		    info->INTERNAL_DISPLAY =
-			    (record->connectcaps & ATOM_CONNECTOR_CAP_INTERNAL_DISPLAY) ? 1 : 0;
-		    info->INTERNAL_DISPLAY_BL =
-			    (record->connectcaps & ATOM_CONNECTOR_CAP_INTERNAL_DISPLAY_BL) ? 1 : 0;
-		    break;
-	    case 5:
+	case 4:
+		default:
+			object = get_bios_object(bp, object_id);
+
+			if (!object)
+				return BP_RESULT_BADINPUT;
+
+			record = get_disp_connector_caps_record(bp, object);
+			if (!record)
+				return BP_RESULT_NORECORD;
+
+			info->INTERNAL_DISPLAY =
+				(record->connectcaps & ATOM_CONNECTOR_CAP_INTERNAL_DISPLAY) ? 1 : 0;
+			info->INTERNAL_DISPLAY_BL =
+				(record->connectcaps & ATOM_CONNECTOR_CAP_INTERNAL_DISPLAY_BL) ? 1 : 0;
+			break;
+	case 5:
 		object_path_v3 = get_bios_object_from_path_v3(bp, object_id);
 
 		if (!object_path_v3)
@@ -2264,9 +2331,8 @@ static enum bp_result bios_parser_get_disp_connector_caps_info(
 	return BP_RESULT_OK;
 }
 
-static struct atom_connector_speed_record *get_connector_speed_cap_record(
-	struct bios_parser *bp,
-	struct atom_display_object_path_v3 *object)
+static struct atom_connector_speed_record *get_connector_speed_cap_record(struct bios_parser *bp,
+									  struct atom_display_object_path_v3 *object)
 {
 	struct atom_common_record_header *header;
 	uint32_t offset;
@@ -2407,6 +2473,24 @@ static enum bp_result get_vram_info_v30(
 	return result;
 }
 
+static enum bp_result get_vram_info_from_umc_info_v40(
+		struct bios_parser *bp,
+		struct dc_vram_info *info)
+{
+	struct atom_umc_info_v4_0 *info_v40;
+	enum bp_result result = BP_RESULT_OK;
+
+	info_v40 = GET_IMAGE(struct atom_umc_info_v4_0,
+						DATA_TABLES(umc_info));
+
+	if (info_v40 == NULL)
+		return BP_RESULT_BADBIOSTABLE;
+
+	info->num_chans = info_v40->channel_num;
+	info->dram_channel_width_bytes = (1 << info_v40->channel_width) / 8;
+
+	return result;
+}
 
 /*
  * get_integrated_info_v11
@@ -2432,10 +2516,11 @@ static enum bp_result get_integrated_info_v11(
 	info_v11 = GET_IMAGE(struct atom_integrated_system_info_v1_11,
 					DATA_TABLES(integratedsysteminfo));
 
-	DC_LOG_BIOS("gpuclk_ss_percentage (unit of 0.001 percent): %d\n", info_v11->gpuclk_ss_percentage);
 	if (info_v11 == NULL)
 		return BP_RESULT_BADBIOSTABLE;
 
+	DC_LOG_BIOS("gpuclk_ss_percentage (unit of 0.001 percent): %d\n", info_v11->gpuclk_ss_percentage);
+
 	info->gpu_cap_info =
 	le32_to_cpu(info_v11->gpucapinfo);
 	/*
@@ -2647,11 +2732,12 @@ static enum bp_result get_integrated_info_v2_1(
 
 	info_v2_1 = GET_IMAGE(struct atom_integrated_system_info_v2_1,
 					DATA_TABLES(integratedsysteminfo));
-	DC_LOG_BIOS("gpuclk_ss_percentage (unit of 0.001 percent): %d\n", info_v2_1->gpuclk_ss_percentage);
 
 	if (info_v2_1 == NULL)
 		return BP_RESULT_BADBIOSTABLE;
 
+	DC_LOG_BIOS("gpuclk_ss_percentage (unit of 0.001 percent): %d\n", info_v2_1->gpuclk_ss_percentage);
+
 	info->gpu_cap_info =
 	le32_to_cpu(info_v2_1->gpucapinfo);
 	/*
@@ -2809,11 +2895,11 @@ static enum bp_result get_integrated_info_v2_2(
 	info_v2_2 = GET_IMAGE(struct atom_integrated_system_info_v2_2,
 					DATA_TABLES(integratedsysteminfo));
 
-	DC_LOG_BIOS("gpuclk_ss_percentage (unit of 0.001 percent): %d\n", info_v2_2->gpuclk_ss_percentage);
-
 	if (info_v2_2 == NULL)
 		return BP_RESULT_BADBIOSTABLE;
 
+	DC_LOG_BIOS("gpuclk_ss_percentage (unit of 0.001 percent): %d\n", info_v2_2->gpuclk_ss_percentage);
+
 	info->gpu_cap_info =
 	le32_to_cpu(info_v2_2->gpucapinfo);
 	/*
@@ -2830,6 +2916,8 @@ static enum bp_result get_integrated_info_v2_2(
 	info->ma_channel_number = info_v2_2->umachannelnumber;
 	info->dp_ss_control =
 		le16_to_cpu(info_v2_2->reserved1);
+	info->gpuclk_ss_percentage = info_v2_2->gpuclk_ss_percentage;
+	info->gpuclk_ss_type = info_v2_2->gpuclk_ss_type;
 
 	for (i = 0; i < NUMBER_OF_UCHAR_FOR_GUID; ++i) {
 		info->ext_disp_conn_info.gu_id[i] =
@@ -2926,8 +3014,11 @@ static enum bp_result construct_integrated_info(
 	struct atom_common_table_header *header;
 	struct atom_data_revision revision;
 
-	uint32_t i;
-	uint32_t j;
+	int32_t i;
+	int32_t j;
+
+	if (!info)
+		return result;
 
 	if (info && DATA_TABLES(integratedsysteminfo)) {
 		header = GET_IMAGE(struct atom_common_table_header,
@@ -2952,6 +3043,7 @@ static enum bp_result construct_integrated_info(
 				result = get_integrated_info_v2_1(bp, info);
 				break;
 			case 2:
+			case 3:
 				result = get_integrated_info_v2_2(bp, info);
 				break;
 			default:
@@ -3007,11 +3099,12 @@ static enum bp_result construct_integrated_info(
 						info->ext_disp_conn_info.path[i].ext_encoder_obj_id.id,
 						info->ext_disp_conn_info.path[i].caps
 						);
-			if (info->ext_disp_conn_info.path[i].caps & EXT_DISPLAY_PATH_CAPS__DP_FIXED_VS_EN)
-				DC_LOG_BIOS("BIOS EXT_DISPLAY_PATH_CAPS__DP_FIXED_VS_EN on path %d\n", i);
+			if ((info->ext_disp_conn_info.path[i].caps & AMD_EXT_DISPLAY_PATH_CAPS__EXT_CHIP_MASK) == AMD_EXT_DISPLAY_PATH_CAPS__DP_FIXED_VS_EN)
+				DC_LOG_BIOS("BIOS AMD_EXT_DISPLAY_PATH_CAPS__DP_FIXED_VS_EN on path %d\n", i);
 			else if (bp->base.ctx->dc->config.force_bios_fixed_vs) {
-				info->ext_disp_conn_info.path[i].caps |= EXT_DISPLAY_PATH_CAPS__DP_FIXED_VS_EN;
-				DC_LOG_BIOS("driver forced EXT_DISPLAY_PATH_CAPS__DP_FIXED_VS_EN on path %d\n", i);
+				info->ext_disp_conn_info.path[i].caps &= ~AMD_EXT_DISPLAY_PATH_CAPS__EXT_CHIP_MASK;
+				info->ext_disp_conn_info.path[i].caps |= AMD_EXT_DISPLAY_PATH_CAPS__DP_FIXED_VS_EN;
+				DC_LOG_BIOS("driver forced AMD_EXT_DISPLAY_PATH_CAPS__DP_FIXED_VS_EN on path %d\n", i);
 			}
 		}
 		// Log the Checksum and Voltage Swing
@@ -3041,11 +3134,33 @@ static enum bp_result bios_parser_get_vram_info(
 		struct dc_vram_info *info)
 {
 	struct bios_parser *bp = BP_FROM_DCB(dcb);
-	static enum bp_result result = BP_RESULT_BADBIOSTABLE;
+	enum bp_result result = BP_RESULT_BADBIOSTABLE;
 	struct atom_common_table_header *header;
 	struct atom_data_revision revision;
 
-	if (info && DATA_TABLES(vram_info)) {
+	// vram info moved to umc_info for DCN4x
+	if (info && DATA_TABLES(umc_info)) {
+		header = GET_IMAGE(struct atom_common_table_header,
+					DATA_TABLES(umc_info));
+
+		get_atom_data_table_revision(header, &revision);
+
+		switch (revision.major) {
+		case 4:
+			switch (revision.minor) {
+			case 0:
+				result = get_vram_info_from_umc_info_v40(bp, info);
+				break;
+			default:
+				break;
+			}
+			break;
+		default:
+			break;
+		}
+	}
+
+	if (result != BP_RESULT_OK && info && DATA_TABLES(vram_info)) {
 		header = GET_IMAGE(struct atom_common_table_header,
 					DATA_TABLES(vram_info));
 
@@ -3090,7 +3205,7 @@ static struct integrated_info *bios_parser_create_integrated_info(
 	struct dc_bios *dcb)
 {
 	struct bios_parser *bp = BP_FROM_DCB(dcb);
-	struct integrated_info *info = NULL;
+	struct integrated_info *info;
 
 	info = kzalloc(sizeof(struct integrated_info), GFP_KERNEL);
 
@@ -3339,27 +3454,28 @@ static enum bp_result get_bracket_layout_record(
 		DC_LOG_DETECTION_EDID_PARSER("Invalid slot_layout_info\n");
 		return BP_RESULT_BADINPUT;
 	}
+
 	tbl = &bp->object_info_tbl;
 	v1_4 = tbl->v1_4;
 	v1_5 = tbl->v1_5;
 
 	result = BP_RESULT_NORECORD;
 	switch (bp->object_info_tbl.revision.minor) {
-		case 4:
-		default:
-			for (i = 0; i < v1_4->number_of_path; ++i)	{
-				if (bracket_layout_id ==
-					v1_4->display_path[i].display_objid) {
-					result = update_slot_layout_info(dcb, i, slot_layout_info);
-					break;
-				}
+	case 4:
+	default:
+		for (i = 0; i < v1_4->number_of_path; ++i) {
+			if (bracket_layout_id == v1_4->display_path[i].display_objid) {
+				result = update_slot_layout_info(dcb, i, slot_layout_info);
+				break;
 			}
-		    break;
-		case 5:
-			for (i = 0; i < v1_5->number_of_path; ++i)
-				result = update_slot_layout_info_v2(dcb, i, slot_layout_info);
-			break;
+		}
+		break;
+	case 5:
+		for (i = 0; i < v1_5->number_of_path; ++i)
+			result = update_slot_layout_info_v2(dcb, i, slot_layout_info);
+		break;
 	}
+
 	return result;
 }
 
@@ -3368,9 +3484,7 @@ static enum bp_result bios_get_board_layout_info(
 	struct board_layout_info *board_layout_info)
 {
 	unsigned int i;
-
 	struct bios_parser *bp;
-
 	static enum bp_result record_result;
 	unsigned int max_slots;
 
@@ -3380,7 +3494,6 @@ static enum bp_result bios_get_board_layout_info(
 		0, 0
 	};
 
-
 	bp = BP_FROM_DCB(dcb);
 
 	if (board_layout_info == NULL) {
@@ -3561,7 +3674,6 @@ static const struct dc_vbios_funcs vbios_funcs = {
 	.bios_parser_destroy = firmware_parser_destroy,
 
 	.get_board_layout_info = bios_get_board_layout_info,
-	/* TODO: use this fn in hw init?*/
 	.pack_data_tables = bios_parser_pack_data_tables,
 
 	.get_atom_dc_golden_table = bios_get_atom_dc_golden_table,
@@ -3671,7 +3783,7 @@ static bool bios_parser2_construct(
 	bp->base.integrated_info = bios_parser_create_integrated_info(&bp->base);
 	bp->base.fw_info_valid = bios_parser_get_firmware_info(&bp->base, &bp->base.fw_info) == BP_RESULT_OK;
 	bios_parser_get_vram_info(&bp->base, &bp->base.vram_info);
-
+	bios_parser_get_soc_bb_info(&bp->base, &bp->base.bb_info);
 	return true;
 }
 
@@ -3679,7 +3791,7 @@ struct dc_bios *firmware_parser_create(
 	struct bp_init_data *init,
 	enum dce_version dce_version)
 {
-	struct bios_parser *bp = NULL;
+	struct bios_parser *bp;
 
 	bp = kzalloc(sizeof(struct bios_parser), GFP_KERNEL);
 	if (!bp)
diff --git a/drivers/gpu/drm/amd/display/dc/bios/bios_parser_helper.c b/drivers/gpu/drm/amd/display/dc/bios/bios_parser_helper.c
index adc710fe4a45..8d2cf95ae739 100644
--- a/drivers/gpu/drm/amd/display/dc/bios/bios_parser_helper.c
+++ b/drivers/gpu/drm/amd/display/dc/bios/bios_parser_helper.c
@@ -78,10 +78,3 @@ void bios_set_scratch_critical_state(
 	uint32_t critial_state = state ? 1 : 0;
 	REG_UPDATE(BIOS_SCRATCH_6, S6_CRITICAL_STATE, critial_state);
 }
-
-uint32_t bios_get_vga_enabled_displays(
-	struct dc_bios *bios)
-{
-	return REG_READ(BIOS_SCRATCH_3) & 0XFFFF;
-}
-
diff --git a/drivers/gpu/drm/amd/display/dc/bios/bios_parser_helper.h b/drivers/gpu/drm/amd/display/dc/bios/bios_parser_helper.h
index e1b4a40a353d..ab162f2fe577 100644
--- a/drivers/gpu/drm/amd/display/dc/bios/bios_parser_helper.h
+++ b/drivers/gpu/drm/amd/display/dc/bios/bios_parser_helper.h
@@ -34,7 +34,6 @@ uint8_t *bios_get_image(struct dc_bios *bp, uint32_t offset,
 bool bios_is_accelerated_mode(struct dc_bios *bios);
 void bios_set_scratch_acc_mode_change(struct dc_bios *bios, uint32_t state);
 void bios_set_scratch_critical_state(struct dc_bios *bios, bool state);
-uint32_t bios_get_vga_enabled_displays(struct dc_bios *bios);
 
 #define GET_IMAGE(type, offset) ((type *) bios_get_image(&bp->base, offset, sizeof(type)))
 
diff --git a/drivers/gpu/drm/amd/display/dc/bios/command_table.c b/drivers/gpu/drm/amd/display/dc/bios/command_table.c
index 818a529cacc3..22457f417e65 100644
--- a/drivers/gpu/drm/amd/display/dc/bios/command_table.c
+++ b/drivers/gpu/drm/amd/display/dc/bios/command_table.c
@@ -37,7 +37,7 @@
 #define EXEC_BIOS_CMD_TABLE(command, params)\
 	(amdgpu_atom_execute_table(((struct amdgpu_device *)bp->base.ctx->driver_context)->mode_info.atom_context, \
 		GetIndexIntoMasterTable(COMMAND, command), \
-		(uint32_t *)&params) == 0)
+		(uint32_t *)&params, sizeof(params)) == 0)
 
 #define BIOS_CMD_TABLE_REVISION(command, frev, crev)\
 	amdgpu_atom_parse_cmd_header(((struct amdgpu_device *)bp->base.ctx->driver_context)->mode_info.atom_context, \
@@ -52,7 +52,9 @@ static void init_transmitter_control(struct bios_parser *bp);
 static void init_set_pixel_clock(struct bios_parser *bp);
 static void init_enable_spread_spectrum_on_ppll(struct bios_parser *bp);
 static void init_adjust_display_pll(struct bios_parser *bp);
+static void init_select_crtc_source(struct bios_parser *bp);
 static void init_dac_encoder_control(struct bios_parser *bp);
+static void init_dac_load_detection(struct bios_parser *bp);
 static void init_dac_output_control(struct bios_parser *bp);
 static void init_set_crtc_timing(struct bios_parser *bp);
 static void init_enable_crtc(struct bios_parser *bp);
@@ -69,7 +71,9 @@ void dal_bios_parser_init_cmd_tbl(struct bios_parser *bp)
 	init_set_pixel_clock(bp);
 	init_enable_spread_spectrum_on_ppll(bp);
 	init_adjust_display_pll(bp);
+	init_select_crtc_source(bp);
 	init_dac_encoder_control(bp);
+	init_dac_load_detection(bp);
 	init_dac_output_control(bp);
 	init_set_crtc_timing(bp);
 	init_enable_crtc(bp);
@@ -399,7 +403,7 @@ static enum bp_result transmitter_control_v1_6(
 static void init_transmitter_control(struct bios_parser *bp)
 {
 	uint8_t frev;
-	uint8_t crev;
+	uint8_t crev = 0;
 
 	if (BIOS_CMD_TABLE_REVISION(UNIPHYTransmitterControl,
 			frev, crev) == false)
@@ -993,7 +997,7 @@ static enum bp_result set_pixel_clock_v3(
 	allocation.sPCLKInput.usFbDiv =
 			cpu_to_le16((uint16_t)bp_params->feedback_divider);
 	allocation.sPCLKInput.ucFracFbDiv =
-			(uint8_t)bp_params->fractional_feedback_divider;
+			(uint8_t)(bp_params->fractional_feedback_divider / 100000);
 	allocation.sPCLKInput.ucPostDiv =
 			(uint8_t)bp_params->pixel_clock_post_divider;
 
@@ -1612,6 +1616,198 @@ static enum bp_result adjust_display_pll_v3(
 /*******************************************************************************
  ********************************************************************************
  **
+ **                  SELECT CRTC SOURCE
+ **
+ ********************************************************************************
+ *******************************************************************************/
+
+static enum bp_result select_crtc_source_v1(
+	struct bios_parser *bp,
+	struct bp_crtc_source_select *bp_params);
+static enum bp_result select_crtc_source_v2(
+	struct bios_parser *bp,
+	struct bp_crtc_source_select *bp_params);
+static enum bp_result select_crtc_source_v3(
+	struct bios_parser *bp,
+	struct bp_crtc_source_select *bp_params);
+
+static void init_select_crtc_source(struct bios_parser *bp)
+{
+	switch (BIOS_CMD_TABLE_PARA_REVISION(SelectCRTC_Source)) {
+	case 1:
+		bp->cmd_tbl.select_crtc_source = select_crtc_source_v1;
+		break;
+	case 2:
+		bp->cmd_tbl.select_crtc_source = select_crtc_source_v2;
+		break;
+	case 3:
+		bp->cmd_tbl.select_crtc_source = select_crtc_source_v3;
+		break;
+	default:
+		bp->cmd_tbl.select_crtc_source = NULL;
+		break;
+	}
+}
+
+static enum bp_result select_crtc_source_v1(
+	struct bios_parser *bp,
+	struct bp_crtc_source_select *bp_params)
+{
+	enum bp_result result = BP_RESULT_FAILURE;
+	SELECT_CRTC_SOURCE_PS_ALLOCATION params;
+
+	if (!bp->cmd_helper->controller_id_to_atom(bp_params->controller_id, &params.ucCRTC))
+		return BP_RESULT_BADINPUT;
+
+	switch (bp_params->engine_id) {
+	case ENGINE_ID_DACA:
+		params.ucDevice = ATOM_DEVICE_CRT1_INDEX;
+		break;
+	case ENGINE_ID_DACB:
+		params.ucDevice = ATOM_DEVICE_CRT2_INDEX;
+		break;
+	default:
+		return BP_RESULT_BADINPUT;
+	}
+
+	if (EXEC_BIOS_CMD_TABLE(SelectCRTC_Source, params))
+		result = BP_RESULT_OK;
+
+	return result;
+}
+
+static bool select_crtc_source_v2_encoder_id(
+	enum engine_id engine_id, uint8_t *out_encoder_id)
+{
+	uint8_t encoder_id = 0;
+
+	switch (engine_id) {
+	case ENGINE_ID_DIGA:
+		encoder_id = ASIC_INT_DIG1_ENCODER_ID;
+		break;
+	case ENGINE_ID_DIGB:
+		encoder_id = ASIC_INT_DIG2_ENCODER_ID;
+		break;
+	case ENGINE_ID_DIGC:
+		encoder_id = ASIC_INT_DIG3_ENCODER_ID;
+		break;
+	case ENGINE_ID_DIGD:
+		encoder_id = ASIC_INT_DIG4_ENCODER_ID;
+		break;
+	case ENGINE_ID_DIGE:
+		encoder_id = ASIC_INT_DIG5_ENCODER_ID;
+		break;
+	case ENGINE_ID_DIGF:
+		encoder_id = ASIC_INT_DIG6_ENCODER_ID;
+		break;
+	case ENGINE_ID_DIGG:
+		encoder_id = ASIC_INT_DIG7_ENCODER_ID;
+		break;
+	case ENGINE_ID_DACA:
+		encoder_id = ASIC_INT_DAC1_ENCODER_ID;
+		break;
+	case ENGINE_ID_DACB:
+		encoder_id = ASIC_INT_DAC2_ENCODER_ID;
+		break;
+	default:
+		return false;
+	}
+
+	*out_encoder_id = encoder_id;
+	return true;
+}
+
+static bool select_crtc_source_v2_encoder_mode(
+	enum signal_type signal_type, uint8_t *out_encoder_mode)
+{
+	uint8_t encoder_mode = 0;
+
+	switch (signal_type) {
+	case SIGNAL_TYPE_DVI_SINGLE_LINK:
+	case SIGNAL_TYPE_DVI_DUAL_LINK:
+		encoder_mode = ATOM_ENCODER_MODE_DVI;
+		break;
+	case SIGNAL_TYPE_HDMI_TYPE_A:
+		encoder_mode = ATOM_ENCODER_MODE_HDMI;
+		break;
+	case SIGNAL_TYPE_LVDS:
+		encoder_mode = ATOM_ENCODER_MODE_LVDS;
+		break;
+	case SIGNAL_TYPE_RGB:
+		encoder_mode = ATOM_ENCODER_MODE_CRT;
+		break;
+	case SIGNAL_TYPE_DISPLAY_PORT:
+		encoder_mode = ATOM_ENCODER_MODE_DP;
+		break;
+	case SIGNAL_TYPE_DISPLAY_PORT_MST:
+		encoder_mode = ATOM_ENCODER_MODE_DP_MST;
+		break;
+	case SIGNAL_TYPE_EDP:
+		encoder_mode = ATOM_ENCODER_MODE_DP;
+		break;
+	default:
+		return false;
+	}
+
+	*out_encoder_mode = encoder_mode;
+	return true;
+}
+
+static enum bp_result select_crtc_source_v2(
+	struct bios_parser *bp,
+	struct bp_crtc_source_select *bp_params)
+{
+	enum bp_result result = BP_RESULT_FAILURE;
+	SELECT_CRTC_SOURCE_PARAMETERS_V3 params;
+
+	if (!bp->cmd_helper->controller_id_to_atom(bp_params->controller_id, &params.ucCRTC))
+		return BP_RESULT_BADINPUT;
+
+	if (!select_crtc_source_v2_encoder_id(
+		bp_params->engine_id,
+		&params.ucEncoderID))
+		return BP_RESULT_BADINPUT;
+	if (!select_crtc_source_v2_encoder_mode(
+		bp_params->sink_signal,
+		&params.ucEncodeMode))
+		return BP_RESULT_BADINPUT;
+
+	if (EXEC_BIOS_CMD_TABLE(SelectCRTC_Source, params))
+		result = BP_RESULT_OK;
+
+	return result;
+}
+
+static enum bp_result select_crtc_source_v3(
+	struct bios_parser *bp,
+	struct bp_crtc_source_select *bp_params)
+{
+	enum bp_result result = BP_RESULT_FAILURE;
+	SELECT_CRTC_SOURCE_PARAMETERS_V3 params;
+
+	if (!bp->cmd_helper->controller_id_to_atom(bp_params->controller_id, &params.ucCRTC))
+		return BP_RESULT_BADINPUT;
+
+	if (!select_crtc_source_v2_encoder_id(
+		bp_params->engine_id,
+		&params.ucEncoderID))
+		return BP_RESULT_BADINPUT;
+	if (!select_crtc_source_v2_encoder_mode(
+		bp_params->sink_signal,
+		&params.ucEncodeMode))
+		return BP_RESULT_BADINPUT;
+
+	params.ucDstBpc = bp_params->bit_depth;
+
+	if (EXEC_BIOS_CMD_TABLE(SelectCRTC_Source, params))
+		result = BP_RESULT_OK;
+
+	return result;
+}
+
+/*******************************************************************************
+ ********************************************************************************
+ **
  **                  DAC ENCODER CONTROL
  **
  ********************************************************************************
@@ -1711,6 +1907,96 @@ static enum bp_result dac2_encoder_control_v1(
 /*******************************************************************************
  ********************************************************************************
  **
+ **                  DAC LOAD DETECTION
+ **
+ ********************************************************************************
+ *******************************************************************************/
+
+static enum bp_result dac_load_detection_v1(
+	struct bios_parser *bp,
+	struct bp_load_detection_parameters *bp_params);
+
+static enum bp_result dac_load_detection_v3(
+	struct bios_parser *bp,
+	struct bp_load_detection_parameters *bp_params);
+
+static void init_dac_load_detection(struct bios_parser *bp)
+{
+	switch (BIOS_CMD_TABLE_PARA_REVISION(DAC_LoadDetection)) {
+	case 1:
+	case 2:
+		bp->cmd_tbl.dac_load_detection = dac_load_detection_v1;
+		break;
+	case 3:
+	default:
+		bp->cmd_tbl.dac_load_detection = dac_load_detection_v3;
+		break;
+	}
+}
+
+static void dac_load_detect_prepare_params(
+	struct _DAC_LOAD_DETECTION_PS_ALLOCATION *params,
+	enum engine_id engine_id,
+	uint16_t device_id,
+	uint8_t misc)
+{
+	uint8_t dac_type = ENGINE_ID_DACA;
+
+	if (engine_id == ENGINE_ID_DACB)
+		dac_type = ATOM_DAC_B;
+
+	params->sDacload.usDeviceID = cpu_to_le16(device_id);
+	params->sDacload.ucDacType = dac_type;
+	params->sDacload.ucMisc = misc;
+}
+
+static enum bp_result dac_load_detection_v1(
+	struct bios_parser *bp,
+	struct bp_load_detection_parameters *bp_params)
+{
+	enum bp_result result = BP_RESULT_FAILURE;
+	DAC_LOAD_DETECTION_PS_ALLOCATION params;
+
+	dac_load_detect_prepare_params(
+		&params,
+		bp_params->engine_id,
+		bp_params->device_id,
+		0);
+
+	if (EXEC_BIOS_CMD_TABLE(DAC_LoadDetection, params))
+		result = BP_RESULT_OK;
+
+	return result;
+}
+
+static enum bp_result dac_load_detection_v3(
+	struct bios_parser *bp,
+	struct bp_load_detection_parameters *bp_params)
+{
+	enum bp_result result = BP_RESULT_FAILURE;
+	DAC_LOAD_DETECTION_PS_ALLOCATION params;
+
+	uint8_t misc = 0;
+
+	if (bp_params->device_id == ATOM_DEVICE_CV_SUPPORT ||
+	    bp_params->device_id == ATOM_DEVICE_TV1_SUPPORT)
+		misc = DAC_LOAD_MISC_YPrPb;
+
+	dac_load_detect_prepare_params(
+		&params,
+		bp_params->engine_id,
+		bp_params->device_id,
+		misc);
+
+	if (EXEC_BIOS_CMD_TABLE(DAC_LoadDetection, params))
+		result = BP_RESULT_OK;
+
+	return result;
+}
+
+/*******************************************************************************
+ ********************************************************************************
+ **
  **                  DAC OUTPUT CONTROL
  **
  ********************************************************************************
diff --git a/drivers/gpu/drm/amd/display/dc/bios/command_table.h b/drivers/gpu/drm/amd/display/dc/bios/command_table.h
index ad533775e724..e89b1ba0048b 100644
--- a/drivers/gpu/drm/amd/display/dc/bios/command_table.h
+++ b/drivers/gpu/drm/amd/display/dc/bios/command_table.h
@@ -52,6 +52,9 @@ struct cmd_tbl {
 	enum bp_result (*adjust_display_pll)(
 		struct bios_parser *bp,
 		struct bp_adjust_pixel_clock_parameters *bp_params);
+	enum bp_result (*select_crtc_source)(
+		struct bios_parser *bp,
+		struct bp_crtc_source_select *bp_params);
 	enum bp_result (*dac1_encoder_control)(
 		struct bios_parser *bp,
 		bool enable,
@@ -68,6 +71,9 @@ struct cmd_tbl {
 	enum bp_result (*dac2_output_control)(
 		struct bios_parser *bp,
 		bool enable);
+	enum bp_result (*dac_load_detection)(
+		struct bios_parser *bp,
+		struct bp_load_detection_parameters *bp_params);
 	enum bp_result (*set_crtc_timing)(
 		struct bios_parser *bp,
 		struct bp_hw_crtc_timing_parameters *bp_params);
diff --git a/drivers/gpu/drm/amd/display/dc/bios/command_table2.c b/drivers/gpu/drm/amd/display/dc/bios/command_table2.c
index 1ef9e4053bb7..f2b1720a6a66 100644
--- a/drivers/gpu/drm/amd/display/dc/bios/command_table2.c
+++ b/drivers/gpu/drm/amd/display/dc/bios/command_table2.c
@@ -49,7 +49,7 @@
 #define EXEC_BIOS_CMD_TABLE(fname, params)\
 	(amdgpu_atom_execute_table(((struct amdgpu_device *)bp->base.ctx->driver_context)->mode_info.atom_context, \
 		GET_INDEX_INTO_MASTER_TABLE(command, fname), \
-		(uint32_t *)&params) == 0)
+		(uint32_t *)&params, sizeof(params)) == 0)
 
 #define BIOS_CMD_TABLE_REVISION(fname, frev, crev)\
 	amdgpu_atom_parse_cmd_header(((struct amdgpu_device *)bp->base.ctx->driver_context)->mode_info.atom_context, \
@@ -101,7 +101,6 @@ static void init_dig_encoder_control(struct bios_parser *bp)
 		bp->cmd_tbl.dig_encoder_control = encoder_control_digx_v1_5;
 		break;
 	default:
-		dm_output_to_console("Don't have dig_encoder_control for v%d\n", version);
 		bp->cmd_tbl.dig_encoder_control = encoder_control_fallback;
 		break;
 	}
@@ -123,9 +122,7 @@ static void encoder_control_dmcub(
 		sizeof(cmd.digx_encoder_control.header);
 	cmd.digx_encoder_control.encoder_control.dig.stream_param = *dig;
 
-	dc_dmub_srv_cmd_queue(dmcub, &cmd);
-	dc_dmub_srv_cmd_execute(dmcub);
-	dc_dmub_srv_wait_idle(dmcub);
+	dc_wake_and_execute_dmub_cmd(dmcub->ctx, &cmd, DM_DMUB_WAIT_TYPE_WAIT);
 }
 
 static enum bp_result encoder_control_digx_v1_5(
@@ -212,6 +209,7 @@ static enum bp_result encoder_control_fallback(
  ******************************************************************************
  *****************************************************************************/
 
+
 static enum bp_result transmitter_control_v1_6(
 	struct bios_parser *bp,
 	struct bp_transmitter_control *cntl);
@@ -227,9 +225,10 @@ static enum bp_result transmitter_control_fallback(
 static void init_transmitter_control(struct bios_parser *bp)
 {
 	uint8_t frev;
-	uint8_t crev;
+	uint8_t crev = 0;
 
-	BIOS_CMD_TABLE_REVISION(dig1transmittercontrol, frev, crev);
+	if (!BIOS_CMD_TABLE_REVISION(dig1transmittercontrol, frev, crev) && (bp->base.ctx->dc->ctx->dce_version <= DCN_VERSION_2_0))
+		BREAK_TO_DEBUGGER();
 
 	switch (crev) {
 	case 6:
@@ -239,7 +238,6 @@ static void init_transmitter_control(struct bios_parser *bp)
 		bp->cmd_tbl.transmitter_control = transmitter_control_v1_7;
 		break;
 	default:
-		dm_output_to_console("Don't have transmitter_control for v%d\n", crev);
 		bp->cmd_tbl.transmitter_control = transmitter_control_fallback;
 		break;
 	}
@@ -261,9 +259,7 @@ static void transmitter_control_dmcub(
 		sizeof(cmd.dig1_transmitter_control.header);
 	cmd.dig1_transmitter_control.transmitter_control.dig = *dig;
 
-	dc_dmub_srv_cmd_queue(dmcub, &cmd);
-	dc_dmub_srv_cmd_execute(dmcub);
-	dc_dmub_srv_wait_idle(dmcub);
+	dc_wake_and_execute_dmub_cmd(dmcub->ctx, &cmd, DM_DMUB_WAIT_TYPE_WAIT);
 }
 
 static enum bp_result transmitter_control_v1_6(
@@ -325,9 +321,22 @@ static void transmitter_control_dmcub_v1_7(
 		sizeof(cmd.dig1_transmitter_control.header);
 	cmd.dig1_transmitter_control.transmitter_control.dig_v1_7 = *dig;
 
-	dc_dmub_srv_cmd_queue(dmcub, &cmd);
-	dc_dmub_srv_cmd_execute(dmcub);
-	dc_dmub_srv_wait_idle(dmcub);
+	dc_wake_and_execute_dmub_cmd(dmcub->ctx, &cmd, DM_DMUB_WAIT_TYPE_WAIT);
+}
+
+static struct dc_link *get_link_by_phy_id(struct dc *p_dc, uint32_t phy_id)
+{
+	struct dc_link *link = NULL;
+
+	// Get Transition Bitmask from dc_link structure associated with PHY
+	for (uint8_t link_id = 0; link_id < MAX_LINKS; link_id++) {
+		if (phy_id == p_dc->links[link_id]->link_enc->transmitter) {
+			link = p_dc->links[link_id];
+			break;
+		}
+	}
+
+	return link;
 }
 
 static enum bp_result transmitter_control_v1_7(
@@ -368,7 +377,38 @@ static enum bp_result transmitter_control_v1_7(
 
 	if (bp->base.ctx->dc->ctx->dmub_srv &&
 		bp->base.ctx->dc->debug.dmub_command_table) {
+		struct dm_process_phy_transition_init_params process_phy_transition_init_params = {0};
+		struct dc_link *link = get_link_by_phy_id(bp->base.ctx->dc, dig_v1_7.phyid);
+		bool is_phy_transition_interlock_allowed = false;
+		uint8_t action = dig_v1_7.action;
+
+		if (link) {
+			if (link->phy_transition_bitmask &&
+				(action == TRANSMITTER_CONTROL_ENABLE || action == TRANSMITTER_CONTROL_DISABLE)) {
+				is_phy_transition_interlock_allowed = true;
+
+				// Prepare input parameters for processing ACPI retimers
+				process_phy_transition_init_params.action                   = action;
+				process_phy_transition_init_params.display_port_lanes_count = cntl->lanes_number;
+				process_phy_transition_init_params.phy_id                   = dig_v1_7.phyid;
+				process_phy_transition_init_params.signal                   = cntl->signal;
+				process_phy_transition_init_params.sym_clock_10khz          = dig_v1_7.symclk_units.symclk_10khz;
+				process_phy_transition_init_params.display_port_link_rate   = link->cur_link_settings.link_rate;
+				process_phy_transition_init_params.transition_bitmask       = link->phy_transition_bitmask;
+			}
+			dig_v1_7.skip_phy_ssc_reduction = link->wa_flags.skip_phy_ssc_reduction;
+		}
+
+		// Handle PRE_OFF_TO_ON: Process ACPI PHY Transition Interlock
+		if (is_phy_transition_interlock_allowed && action == TRANSMITTER_CONTROL_ENABLE)
+			dm_acpi_process_phy_transition_interlock(bp->base.ctx, process_phy_transition_init_params);
+
 		transmitter_control_dmcub_v1_7(bp->base.ctx->dmub_srv, &dig_v1_7);
+
+		// Handle POST_ON_TO_OFF: Process ACPI PHY Transition Interlock
+		if (is_phy_transition_interlock_allowed && action == TRANSMITTER_CONTROL_DISABLE)
+			dm_acpi_process_phy_transition_interlock(bp->base.ctx, process_phy_transition_init_params);
+
 		return BP_RESULT_OK;
 	}
 
@@ -413,8 +453,6 @@ static void init_set_pixel_clock(struct bios_parser *bp)
 		bp->cmd_tbl.set_pixel_clock = set_pixel_clock_v7;
 		break;
 	default:
-		dm_output_to_console("Don't have set_pixel_clock for v%d\n",
-			 BIOS_CMD_TABLE_PARA_REVISION(setpixelclock));
 		bp->cmd_tbl.set_pixel_clock = set_pixel_clock_fallback;
 		break;
 	}
@@ -435,9 +473,7 @@ static void set_pixel_clock_dmcub(
 		sizeof(cmd.set_pixel_clock.header);
 	cmd.set_pixel_clock.pixel_clock.clk = *clk;
 
-	dc_dmub_srv_cmd_queue(dmcub, &cmd);
-	dc_dmub_srv_cmd_execute(dmcub);
-	dc_dmub_srv_wait_idle(dmcub);
+	dc_wake_and_execute_dmub_cmd(dmcub->ctx, &cmd, DM_DMUB_WAIT_TYPE_WAIT);
 }
 
 static enum bp_result set_pixel_clock_v7(
@@ -561,7 +597,6 @@ static void init_set_crtc_timing(struct bios_parser *bp)
 			set_crtc_using_dtd_timing_v3;
 		break;
 	default:
-		dm_output_to_console("Don't have set_crtc_timing for v%d\n", dtd_version);
 		bp->cmd_tbl.set_crtc_timing = NULL;
 		break;
 	}
@@ -678,8 +713,6 @@ static void init_enable_crtc(struct bios_parser *bp)
 		bp->cmd_tbl.enable_crtc = enable_crtc_v1;
 		break;
 	default:
-		dm_output_to_console("Don't have enable_crtc for v%d\n",
-			 BIOS_CMD_TABLE_PARA_REVISION(enablecrtc));
 		bp->cmd_tbl.enable_crtc = NULL;
 		break;
 	}
@@ -804,9 +837,7 @@ static void enable_disp_power_gating_dmcub(
 		sizeof(cmd.enable_disp_power_gating.header);
 	cmd.enable_disp_power_gating.power_gating.pwr = *pwr;
 
-	dc_dmub_srv_cmd_queue(dmcub, &cmd);
-	dc_dmub_srv_cmd_execute(dmcub);
-	dc_dmub_srv_wait_idle(dmcub);
+	dc_wake_and_execute_dmub_cmd(dmcub->ctx, &cmd, DM_DMUB_WAIT_TYPE_WAIT);
 }
 
 static enum bp_result enable_disp_power_gating_v2_1(
@@ -873,8 +904,6 @@ static void init_set_dce_clock(struct bios_parser *bp)
 		bp->cmd_tbl.set_dce_clock = set_dce_clock_v2_1;
 		break;
 	default:
-		dm_output_to_console("Don't have set_dce_clock for v%d\n",
-			 BIOS_CMD_TABLE_PARA_REVISION(setdceclock));
 		bp->cmd_tbl.set_dce_clock = NULL;
 		break;
 	}
@@ -986,7 +1015,7 @@ static unsigned int get_smu_clock_info_v3_1(struct bios_parser *bp, uint8_t id)
 static enum bp_result enable_lvtma_control(
 	struct bios_parser *bp,
 	uint8_t uc_pwr_on,
-	uint8_t panel_instance,
+	uint8_t pwrseq_instance,
 	uint8_t bypass_panel_control_wait);
 
 static void init_enable_lvtma_control(struct bios_parser *bp)
@@ -999,7 +1028,7 @@ static void init_enable_lvtma_control(struct bios_parser *bp)
 static void enable_lvtma_control_dmcub(
 	struct dc_dmub_srv *dmcub,
 	uint8_t uc_pwr_on,
-	uint8_t panel_instance,
+	uint8_t pwrseq_instance,
 	uint8_t bypass_panel_control_wait)
 {
 
@@ -1012,20 +1041,17 @@ static void enable_lvtma_control_dmcub(
 			DMUB_CMD__VBIOS_LVTMA_CONTROL;
 	cmd.lvtma_control.data.uc_pwr_action =
 			uc_pwr_on;
-	cmd.lvtma_control.data.panel_inst =
-			panel_instance;
+	cmd.lvtma_control.data.pwrseq_inst =
+			pwrseq_instance;
 	cmd.lvtma_control.data.bypass_panel_control_wait =
 			bypass_panel_control_wait;
-	dc_dmub_srv_cmd_queue(dmcub, &cmd);
-	dc_dmub_srv_cmd_execute(dmcub);
-	dc_dmub_srv_wait_idle(dmcub);
-
+	dc_wake_and_execute_dmub_cmd(dmcub->ctx, &cmd, DM_DMUB_WAIT_TYPE_WAIT);
 }
 
 static enum bp_result enable_lvtma_control(
 	struct bios_parser *bp,
 	uint8_t uc_pwr_on,
-	uint8_t panel_instance,
+	uint8_t pwrseq_instance,
 	uint8_t bypass_panel_control_wait)
 {
 	enum bp_result result = BP_RESULT_FAILURE;
@@ -1034,7 +1060,7 @@ static enum bp_result enable_lvtma_control(
 	    bp->base.ctx->dc->debug.dmub_command_table) {
 		enable_lvtma_control_dmcub(bp->base.ctx->dmub_srv,
 				uc_pwr_on,
-				panel_instance,
+				pwrseq_instance,
 				bypass_panel_control_wait);
 		return BP_RESULT_OK;
 	}
@@ -1058,3 +1084,4 @@ void dal_firmware_parser_init_cmd_tbl(struct bios_parser *bp)
 
 	init_enable_lvtma_control(bp);
 }
+
diff --git a/drivers/gpu/drm/amd/display/dc/bios/command_table2.h b/drivers/gpu/drm/amd/display/dc/bios/command_table2.h
index b6d09bf6cf72..41c8c014397f 100644
--- a/drivers/gpu/drm/amd/display/dc/bios/command_table2.h
+++ b/drivers/gpu/drm/amd/display/dc/bios/command_table2.h
@@ -96,7 +96,7 @@ struct cmd_tbl {
 			struct bios_parser *bp, uint8_t id);
 	enum bp_result (*enable_lvtma_control)(struct bios_parser *bp,
 			uint8_t uc_pwr_on,
-			uint8_t panel_instance,
+			uint8_t pwrseq_instance,
 			uint8_t bypass_panel_control_wait);
 };
 
diff --git a/drivers/gpu/drm/amd/display/dc/bios/command_table_helper.c b/drivers/gpu/drm/amd/display/dc/bios/command_table_helper.c
index e317a3615147..91bc8a06e2cf 100644
--- a/drivers/gpu/drm/amd/display/dc/bios/command_table_helper.c
+++ b/drivers/gpu/drm/amd/display/dc/bios/command_table_helper.c
@@ -293,3 +293,107 @@ uint8_t dal_cmd_table_helper_encoder_id_to_atom(
 		return ENCODER_OBJECT_ID_NONE;
 	}
 }
+
+uint8_t phy_id_to_atom(enum transmitter t)
+{
+	uint8_t atom_phy_id;
+
+	switch (t) {
+	case TRANSMITTER_UNIPHY_A:
+		atom_phy_id = ATOM_PHY_ID_UNIPHYA;
+		break;
+	case TRANSMITTER_UNIPHY_B:
+		atom_phy_id = ATOM_PHY_ID_UNIPHYB;
+		break;
+	case TRANSMITTER_UNIPHY_C:
+		atom_phy_id = ATOM_PHY_ID_UNIPHYC;
+		break;
+	case TRANSMITTER_UNIPHY_D:
+		atom_phy_id = ATOM_PHY_ID_UNIPHYD;
+		break;
+	case TRANSMITTER_UNIPHY_E:
+		atom_phy_id = ATOM_PHY_ID_UNIPHYE;
+		break;
+	case TRANSMITTER_UNIPHY_F:
+		atom_phy_id = ATOM_PHY_ID_UNIPHYF;
+		break;
+	case TRANSMITTER_UNIPHY_G:
+		atom_phy_id = ATOM_PHY_ID_UNIPHYG;
+		break;
+	default:
+		atom_phy_id = ATOM_PHY_ID_UNIPHYA;
+		break;
+	}
+	return atom_phy_id;
+}
+
+uint8_t clock_source_id_to_atom_phy_clk_src_id(
+		enum clock_source_id id)
+{
+	uint8_t atom_phy_clk_src_id = 0;
+
+	switch (id) {
+	case CLOCK_SOURCE_ID_PLL0:
+		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P0PLL;
+		break;
+	case CLOCK_SOURCE_ID_PLL1:
+		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P1PLL;
+		break;
+	case CLOCK_SOURCE_ID_PLL2:
+		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P2PLL;
+		break;
+	case CLOCK_SOURCE_ID_EXTERNAL:
+		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_REFCLK_SRC_EXT;
+		break;
+	default:
+		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P1PLL;
+		break;
+	}
+
+	return atom_phy_clk_src_id >> 2;
+}
+
+bool engine_bp_to_atom(enum engine_id id, uint32_t *atom_engine_id)
+{
+	bool result = false;
+
+	if (atom_engine_id != NULL)
+		switch (id) {
+		case ENGINE_ID_DIGA:
+			*atom_engine_id = ASIC_INT_DIG1_ENCODER_ID;
+			result = true;
+			break;
+		case ENGINE_ID_DIGB:
+			*atom_engine_id = ASIC_INT_DIG2_ENCODER_ID;
+			result = true;
+			break;
+		case ENGINE_ID_DIGC:
+			*atom_engine_id = ASIC_INT_DIG3_ENCODER_ID;
+			result = true;
+			break;
+		case ENGINE_ID_DIGD:
+			*atom_engine_id = ASIC_INT_DIG4_ENCODER_ID;
+			result = true;
+			break;
+		case ENGINE_ID_DIGE:
+			*atom_engine_id = ASIC_INT_DIG5_ENCODER_ID;
+			result = true;
+			break;
+		case ENGINE_ID_DIGF:
+			*atom_engine_id = ASIC_INT_DIG6_ENCODER_ID;
+			result = true;
+			break;
+		case ENGINE_ID_DIGG:
+			*atom_engine_id = ASIC_INT_DIG7_ENCODER_ID;
+			result = true;
+			break;
+		case ENGINE_ID_DACA:
+			*atom_engine_id = ASIC_INT_DAC1_ENCODER_ID;
+			result = true;
+			break;
+		default:
+			break;
+		}
+
+	return result;
+}
diff --git a/drivers/gpu/drm/amd/display/dc/bios/command_table_helper.h b/drivers/gpu/drm/amd/display/dc/bios/command_table_helper.h
index dfd30aaf4032..547700e119a6 100644
--- a/drivers/gpu/drm/amd/display/dc/bios/command_table_helper.h
+++ b/drivers/gpu/drm/amd/display/dc/bios/command_table_helper.h
@@ -59,4 +59,12 @@ uint8_t dal_cmd_table_helper_transmitter_bp_to_atom(
 
 uint8_t dal_cmd_table_helper_encoder_id_to_atom(
 	enum encoder_id id);
+
+uint8_t phy_id_to_atom(enum transmitter t);
+
+uint8_t clock_source_id_to_atom_phy_clk_src_id(
+		enum clock_source_id id);
+
+bool engine_bp_to_atom(enum engine_id id, uint32_t *atom_engine_id);
+
 #endif
diff --git a/drivers/gpu/drm/amd/display/dc/bios/command_table_helper2.c b/drivers/gpu/drm/amd/display/dc/bios/command_table_helper2.c
index 8538f13e01bf..268e2414b34f 100644
--- a/drivers/gpu/drm/amd/display/dc/bios/command_table_helper2.c
+++ b/drivers/gpu/drm/amd/display/dc/bios/command_table_helper2.c
@@ -80,12 +80,15 @@ bool dal_bios_parser_init_cmd_tbl_helper2(
 	case DCN_VERSION_3_16:
 	case DCN_VERSION_3_2:
 	case DCN_VERSION_3_21:
+	case DCN_VERSION_3_5:
+	case DCN_VERSION_3_51:
+	case DCN_VERSION_3_6:
+	case DCN_VERSION_4_01:
 		*h = dal_cmd_tbl_helper_dce112_get_table2();
 		return true;
 
 	default:
-		/* Unsupported DCE */
-		BREAK_TO_DEBUGGER();
+		*h = dal_cmd_tbl_helper_dce112_get_table2();
 		return false;
 	}
 }
diff --git a/drivers/gpu/drm/amd/display/dc/bios/dce110/command_table_helper_dce110.c b/drivers/gpu/drm/amd/display/dc/bios/dce110/command_table_helper_dce110.c
index 11bf247bb180..3099128223df 100644
--- a/drivers/gpu/drm/amd/display/dc/bios/dce110/command_table_helper_dce110.c
+++ b/drivers/gpu/drm/amd/display/dc/bios/dce110/command_table_helper_dce110.c
@@ -31,39 +31,6 @@
 
 #include "../command_table_helper.h"
 
-static uint8_t phy_id_to_atom(enum transmitter t)
-{
-	uint8_t atom_phy_id;
-
-	switch (t) {
-	case TRANSMITTER_UNIPHY_A:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYA;
-		break;
-	case TRANSMITTER_UNIPHY_B:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYB;
-		break;
-	case TRANSMITTER_UNIPHY_C:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYC;
-		break;
-	case TRANSMITTER_UNIPHY_D:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYD;
-		break;
-	case TRANSMITTER_UNIPHY_E:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYE;
-		break;
-	case TRANSMITTER_UNIPHY_F:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYF;
-		break;
-	case TRANSMITTER_UNIPHY_G:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYG;
-		break;
-	default:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYA;
-		break;
-	}
-	return atom_phy_id;
-}
-
 static uint8_t signal_type_to_atom_dig_mode(enum signal_type s)
 {
 	uint8_t atom_dig_mode = ATOM_TRANSMITTER_DIGMODE_V5_DP;
@@ -94,32 +61,6 @@ static uint8_t signal_type_to_atom_dig_mode(enum signal_type s)
 	return atom_dig_mode;
 }
 
-static uint8_t clock_source_id_to_atom_phy_clk_src_id(
-		enum clock_source_id id)
-{
-	uint8_t atom_phy_clk_src_id = 0;
-
-	switch (id) {
-	case CLOCK_SOURCE_ID_PLL0:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P0PLL;
-		break;
-	case CLOCK_SOURCE_ID_PLL1:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P1PLL;
-		break;
-	case CLOCK_SOURCE_ID_PLL2:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P2PLL;
-		break;
-	case CLOCK_SOURCE_ID_EXTERNAL:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_REFCLK_SRC_EXT;
-		break;
-	default:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P1PLL;
-		break;
-	}
-
-	return atom_phy_clk_src_id >> 2;
-}
-
 static uint8_t hpd_sel_to_atom(enum hpd_source_id id)
 {
 	uint8_t atom_hpd_sel = 0;
@@ -207,51 +148,6 @@ static bool clock_source_id_to_atom(
 	return result;
 }
 
-static bool engine_bp_to_atom(enum engine_id id, uint32_t *atom_engine_id)
-{
-	bool result = false;
-
-	if (atom_engine_id != NULL)
-		switch (id) {
-		case ENGINE_ID_DIGA:
-			*atom_engine_id = ASIC_INT_DIG1_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGB:
-			*atom_engine_id = ASIC_INT_DIG2_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGC:
-			*atom_engine_id = ASIC_INT_DIG3_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGD:
-			*atom_engine_id = ASIC_INT_DIG4_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGE:
-			*atom_engine_id = ASIC_INT_DIG5_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGF:
-			*atom_engine_id = ASIC_INT_DIG6_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGG:
-			*atom_engine_id = ASIC_INT_DIG7_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DACA:
-			*atom_engine_id = ASIC_INT_DAC1_ENCODER_ID;
-			result = true;
-			break;
-		default:
-			break;
-		}
-
-	return result;
-}
-
 static uint8_t encoder_action_to_atom(enum bp_encoder_control_action action)
 {
 	uint8_t atom_action = 0;
diff --git a/drivers/gpu/drm/amd/display/dc/bios/dce112/command_table_helper2_dce112.c b/drivers/gpu/drm/amd/display/dc/bios/dce112/command_table_helper2_dce112.c
index 755b6e33140a..349f0e5d5856 100644
--- a/drivers/gpu/drm/amd/display/dc/bios/dce112/command_table_helper2_dce112.c
+++ b/drivers/gpu/drm/amd/display/dc/bios/dce112/command_table_helper2_dce112.c
@@ -29,40 +29,9 @@
 
 #include "include/bios_parser_types.h"
 
-#include "../command_table_helper2.h"
-
-static uint8_t phy_id_to_atom(enum transmitter t)
-{
-	uint8_t atom_phy_id;
+#include "../command_table_helper.h"
 
-	switch (t) {
-	case TRANSMITTER_UNIPHY_A:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYA;
-		break;
-	case TRANSMITTER_UNIPHY_B:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYB;
-		break;
-	case TRANSMITTER_UNIPHY_C:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYC;
-		break;
-	case TRANSMITTER_UNIPHY_D:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYD;
-		break;
-	case TRANSMITTER_UNIPHY_E:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYE;
-		break;
-	case TRANSMITTER_UNIPHY_F:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYF;
-		break;
-	case TRANSMITTER_UNIPHY_G:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYG;
-		break;
-	default:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYA;
-		break;
-	}
-	return atom_phy_id;
-}
+#include "../command_table_helper2.h"
 
 static uint8_t signal_type_to_atom_dig_mode(enum signal_type s)
 {
@@ -91,32 +60,6 @@ static uint8_t signal_type_to_atom_dig_mode(enum signal_type s)
 	return atom_dig_mode;
 }
 
-static uint8_t clock_source_id_to_atom_phy_clk_src_id(
-		enum clock_source_id id)
-{
-	uint8_t atom_phy_clk_src_id = 0;
-
-	switch (id) {
-	case CLOCK_SOURCE_ID_PLL0:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P0PLL;
-		break;
-	case CLOCK_SOURCE_ID_PLL1:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P1PLL;
-		break;
-	case CLOCK_SOURCE_ID_PLL2:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P2PLL;
-		break;
-	case CLOCK_SOURCE_ID_EXTERNAL:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_REFCLK_SRC_EXT;
-		break;
-	default:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P1PLL;
-		break;
-	}
-
-	return atom_phy_clk_src_id >> 2;
-}
-
 static uint8_t hpd_sel_to_atom(enum hpd_source_id id)
 {
 	uint8_t atom_hpd_sel = 0;
@@ -209,51 +152,6 @@ static bool clock_source_id_to_atom(
 	return result;
 }
 
-static bool engine_bp_to_atom(enum engine_id id, uint32_t *atom_engine_id)
-{
-	bool result = false;
-
-	if (atom_engine_id != NULL)
-		switch (id) {
-		case ENGINE_ID_DIGA:
-			*atom_engine_id = ASIC_INT_DIG1_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGB:
-			*atom_engine_id = ASIC_INT_DIG2_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGC:
-			*atom_engine_id = ASIC_INT_DIG3_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGD:
-			*atom_engine_id = ASIC_INT_DIG4_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGE:
-			*atom_engine_id = ASIC_INT_DIG5_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGF:
-			*atom_engine_id = ASIC_INT_DIG6_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGG:
-			*atom_engine_id = ASIC_INT_DIG7_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DACA:
-			*atom_engine_id = ASIC_INT_DAC1_ENCODER_ID;
-			result = true;
-			break;
-		default:
-			break;
-		}
-
-	return result;
-}
-
 static uint8_t encoder_action_to_atom(enum bp_encoder_control_action action)
 {
 	uint8_t atom_action = 0;
diff --git a/drivers/gpu/drm/amd/display/dc/bios/dce112/command_table_helper_dce112.c b/drivers/gpu/drm/amd/display/dc/bios/dce112/command_table_helper_dce112.c
index 06b4f7fa4a50..1a5fefcde8af 100644
--- a/drivers/gpu/drm/amd/display/dc/bios/dce112/command_table_helper_dce112.c
+++ b/drivers/gpu/drm/amd/display/dc/bios/dce112/command_table_helper_dce112.c
@@ -31,39 +31,6 @@
 
 #include "../command_table_helper.h"
 
-static uint8_t phy_id_to_atom(enum transmitter t)
-{
-	uint8_t atom_phy_id;
-
-	switch (t) {
-	case TRANSMITTER_UNIPHY_A:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYA;
-		break;
-	case TRANSMITTER_UNIPHY_B:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYB;
-		break;
-	case TRANSMITTER_UNIPHY_C:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYC;
-		break;
-	case TRANSMITTER_UNIPHY_D:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYD;
-		break;
-	case TRANSMITTER_UNIPHY_E:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYE;
-		break;
-	case TRANSMITTER_UNIPHY_F:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYF;
-		break;
-	case TRANSMITTER_UNIPHY_G:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYG;
-		break;
-	default:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYA;
-		break;
-	}
-	return atom_phy_id;
-}
-
 static uint8_t signal_type_to_atom_dig_mode(enum signal_type s)
 {
 	uint8_t atom_dig_mode = ATOM_TRANSMITTER_DIGMODE_V6_DP;
@@ -91,32 +58,6 @@ static uint8_t signal_type_to_atom_dig_mode(enum signal_type s)
 	return atom_dig_mode;
 }
 
-static uint8_t clock_source_id_to_atom_phy_clk_src_id(
-		enum clock_source_id id)
-{
-	uint8_t atom_phy_clk_src_id = 0;
-
-	switch (id) {
-	case CLOCK_SOURCE_ID_PLL0:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P0PLL;
-		break;
-	case CLOCK_SOURCE_ID_PLL1:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P1PLL;
-		break;
-	case CLOCK_SOURCE_ID_PLL2:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P2PLL;
-		break;
-	case CLOCK_SOURCE_ID_EXTERNAL:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_REFCLK_SRC_EXT;
-		break;
-	default:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P1PLL;
-		break;
-	}
-
-	return atom_phy_clk_src_id >> 2;
-}
-
 static uint8_t hpd_sel_to_atom(enum hpd_source_id id)
 {
 	uint8_t atom_hpd_sel = 0;
@@ -209,51 +150,6 @@ static bool clock_source_id_to_atom(
 	return result;
 }
 
-static bool engine_bp_to_atom(enum engine_id id, uint32_t *atom_engine_id)
-{
-	bool result = false;
-
-	if (atom_engine_id != NULL)
-		switch (id) {
-		case ENGINE_ID_DIGA:
-			*atom_engine_id = ASIC_INT_DIG1_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGB:
-			*atom_engine_id = ASIC_INT_DIG2_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGC:
-			*atom_engine_id = ASIC_INT_DIG3_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGD:
-			*atom_engine_id = ASIC_INT_DIG4_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGE:
-			*atom_engine_id = ASIC_INT_DIG5_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGF:
-			*atom_engine_id = ASIC_INT_DIG6_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGG:
-			*atom_engine_id = ASIC_INT_DIG7_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DACA:
-			*atom_engine_id = ASIC_INT_DAC1_ENCODER_ID;
-			result = true;
-			break;
-		default:
-			break;
-		}
-
-	return result;
-}
-
 static uint8_t encoder_action_to_atom(enum bp_encoder_control_action action)
 {
 	uint8_t atom_action = 0;
diff --git a/drivers/gpu/drm/amd/display/dc/bios/dce60/command_table_helper_dce60.c b/drivers/gpu/drm/amd/display/dc/bios/dce60/command_table_helper_dce60.c
index 710221b4f5c5..01ccc803040c 100644
--- a/drivers/gpu/drm/amd/display/dc/bios/dce60/command_table_helper_dce60.c
+++ b/drivers/gpu/drm/amd/display/dc/bios/dce60/command_table_helper_dce60.c
@@ -58,51 +58,6 @@ static uint8_t encoder_action_to_atom(enum bp_encoder_control_action action)
 	return atom_action;
 }
 
-static bool engine_bp_to_atom(enum engine_id id, uint32_t *atom_engine_id)
-{
-	bool result = false;
-
-	if (atom_engine_id != NULL)
-		switch (id) {
-		case ENGINE_ID_DIGA:
-			*atom_engine_id = ASIC_INT_DIG1_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGB:
-			*atom_engine_id = ASIC_INT_DIG2_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGC:
-			*atom_engine_id = ASIC_INT_DIG3_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGD:
-			*atom_engine_id = ASIC_INT_DIG4_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGE:
-			*atom_engine_id = ASIC_INT_DIG5_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGF:
-			*atom_engine_id = ASIC_INT_DIG6_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGG:
-			*atom_engine_id = ASIC_INT_DIG7_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DACA:
-			*atom_engine_id = ASIC_INT_DAC1_ENCODER_ID;
-			result = true;
-			break;
-		default:
-			break;
-		}
-
-	return result;
-}
-
 static bool clock_source_id_to_atom(
 	enum clock_source_id id,
 	uint32_t *atom_pll_id)
@@ -149,32 +104,6 @@ static bool clock_source_id_to_atom(
 	return result;
 }
 
-static uint8_t clock_source_id_to_atom_phy_clk_src_id(
-		enum clock_source_id id)
-{
-	uint8_t atom_phy_clk_src_id = 0;
-
-	switch (id) {
-	case CLOCK_SOURCE_ID_PLL0:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P0PLL;
-		break;
-	case CLOCK_SOURCE_ID_PLL1:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P1PLL;
-		break;
-	case CLOCK_SOURCE_ID_PLL2:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P2PLL;
-		break;
-	case CLOCK_SOURCE_ID_EXTERNAL:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_REFCLK_SRC_EXT;
-		break;
-	default:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P1PLL;
-		break;
-	}
-
-	return atom_phy_clk_src_id >> 2;
-}
-
 static uint8_t signal_type_to_atom_dig_mode(enum signal_type s)
 {
 	uint8_t atom_dig_mode = ATOM_TRANSMITTER_DIGMODE_V5_DP;
@@ -270,39 +199,6 @@ static uint8_t dig_encoder_sel_to_atom(enum engine_id id)
 	return atom_dig_encoder_sel;
 }
 
-static uint8_t phy_id_to_atom(enum transmitter t)
-{
-	uint8_t atom_phy_id;
-
-	switch (t) {
-	case TRANSMITTER_UNIPHY_A:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYA;
-		break;
-	case TRANSMITTER_UNIPHY_B:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYB;
-		break;
-	case TRANSMITTER_UNIPHY_C:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYC;
-		break;
-	case TRANSMITTER_UNIPHY_D:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYD;
-		break;
-	case TRANSMITTER_UNIPHY_E:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYE;
-		break;
-	case TRANSMITTER_UNIPHY_F:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYF;
-		break;
-	case TRANSMITTER_UNIPHY_G:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYG;
-		break;
-	default:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYA;
-		break;
-	}
-	return atom_phy_id;
-}
-
 static uint8_t disp_power_gating_action_to_atom(
 	enum bp_pipe_control_action action)
 {
diff --git a/drivers/gpu/drm/amd/display/dc/bios/dce80/command_table_helper_dce80.c b/drivers/gpu/drm/amd/display/dc/bios/dce80/command_table_helper_dce80.c
index 8b30b558cf1f..2ec5264536c7 100644
--- a/drivers/gpu/drm/amd/display/dc/bios/dce80/command_table_helper_dce80.c
+++ b/drivers/gpu/drm/amd/display/dc/bios/dce80/command_table_helper_dce80.c
@@ -58,51 +58,6 @@ static uint8_t encoder_action_to_atom(enum bp_encoder_control_action action)
 	return atom_action;
 }
 
-static bool engine_bp_to_atom(enum engine_id id, uint32_t *atom_engine_id)
-{
-	bool result = false;
-
-	if (atom_engine_id != NULL)
-		switch (id) {
-		case ENGINE_ID_DIGA:
-			*atom_engine_id = ASIC_INT_DIG1_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGB:
-			*atom_engine_id = ASIC_INT_DIG2_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGC:
-			*atom_engine_id = ASIC_INT_DIG3_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGD:
-			*atom_engine_id = ASIC_INT_DIG4_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGE:
-			*atom_engine_id = ASIC_INT_DIG5_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGF:
-			*atom_engine_id = ASIC_INT_DIG6_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DIGG:
-			*atom_engine_id = ASIC_INT_DIG7_ENCODER_ID;
-			result = true;
-			break;
-		case ENGINE_ID_DACA:
-			*atom_engine_id = ASIC_INT_DAC1_ENCODER_ID;
-			result = true;
-			break;
-		default:
-			break;
-		}
-
-	return result;
-}
-
 static bool clock_source_id_to_atom(
 	enum clock_source_id id,
 	uint32_t *atom_pll_id)
@@ -149,32 +104,6 @@ static bool clock_source_id_to_atom(
 	return result;
 }
 
-static uint8_t clock_source_id_to_atom_phy_clk_src_id(
-		enum clock_source_id id)
-{
-	uint8_t atom_phy_clk_src_id = 0;
-
-	switch (id) {
-	case CLOCK_SOURCE_ID_PLL0:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P0PLL;
-		break;
-	case CLOCK_SOURCE_ID_PLL1:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P1PLL;
-		break;
-	case CLOCK_SOURCE_ID_PLL2:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P2PLL;
-		break;
-	case CLOCK_SOURCE_ID_EXTERNAL:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_REFCLK_SRC_EXT;
-		break;
-	default:
-		atom_phy_clk_src_id = ATOM_TRANSMITTER_CONFIG_V5_P1PLL;
-		break;
-	}
-
-	return atom_phy_clk_src_id >> 2;
-}
-
 static uint8_t signal_type_to_atom_dig_mode(enum signal_type s)
 {
 	uint8_t atom_dig_mode = ATOM_TRANSMITTER_DIGMODE_V5_DP;
@@ -270,39 +199,6 @@ static uint8_t dig_encoder_sel_to_atom(enum engine_id id)
 	return atom_dig_encoder_sel;
 }
 
-static uint8_t phy_id_to_atom(enum transmitter t)
-{
-	uint8_t atom_phy_id;
-
-	switch (t) {
-	case TRANSMITTER_UNIPHY_A:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYA;
-		break;
-	case TRANSMITTER_UNIPHY_B:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYB;
-		break;
-	case TRANSMITTER_UNIPHY_C:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYC;
-		break;
-	case TRANSMITTER_UNIPHY_D:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYD;
-		break;
-	case TRANSMITTER_UNIPHY_E:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYE;
-		break;
-	case TRANSMITTER_UNIPHY_F:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYF;
-		break;
-	case TRANSMITTER_UNIPHY_G:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYG;
-		break;
-	default:
-		atom_phy_id = ATOM_PHY_ID_UNIPHYA;
-		break;
-	}
-	return atom_phy_id;
-}
-
 static uint8_t disp_power_gating_action_to_atom(
 	enum bp_pipe_control_action action)
 {
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/Makefile b/drivers/gpu/drm/amd/display/dc/clk_mgr/Makefile
index ad390e4cd0a9..60021671b386 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/Makefile
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/Makefile
@@ -83,7 +83,6 @@ CLK_MGR_DCN10 = rv1_clk_mgr.o rv1_clk_mgr_vbios_smu.o rv2_clk_mgr.o
 AMD_DAL_CLK_MGR_DCN10 = $(addprefix $(AMDDALPATH)/dc/clk_mgr/dcn10/,$(CLK_MGR_DCN10))
 
 AMD_DISPLAY_FILES += $(AMD_DAL_CLK_MGR_DCN10)
-
 ###############################################################################
 # DCN20
 ###############################################################################
@@ -113,7 +112,7 @@ AMD_DISPLAY_FILES += $(AMD_DAL_CLK_MGR_DCN21)
 ###############################################################################
 # DCN30
 ###############################################################################
-CLK_MGR_DCN30 = dcn30_clk_mgr.o dcn30_clk_mgr_smu_msg.o
+CLK_MGR_DCN30 = dcn30_clk_mgr.o dcn30_clk_mgr_smu_msg.o dcn30m_clk_mgr.o dcn30m_clk_mgr_smu_msg.o
 
 AMD_DAL_CLK_MGR_DCN30 = $(addprefix $(AMDDALPATH)/dc/clk_mgr/dcn30/,$(CLK_MGR_DCN30))
 
@@ -172,4 +171,21 @@ AMD_DAL_CLK_MGR_DCN32 = $(addprefix $(AMDDALPATH)/dc/clk_mgr/dcn32/,$(CLK_MGR_DC
 
 AMD_DISPLAY_FILES += $(AMD_DAL_CLK_MGR_DCN32)
 
+###############################################################################
+# DCN35
+###############################################################################
+CLK_MGR_DCN35 = dcn35_smu.o dcn351_clk_mgr.o dcn35_clk_mgr.o
+
+AMD_DAL_CLK_MGR_DCN35 = $(addprefix $(AMDDALPATH)/dc/clk_mgr/dcn35/,$(CLK_MGR_DCN35))
+
+AMD_DISPLAY_FILES += $(AMD_DAL_CLK_MGR_DCN35)
+
+###############################################################################
+# DCN401
+###############################################################################
+CLK_MGR_DCN401 = dcn401_clk_mgr.o dcn401_clk_mgr_smu_msg.o
+
+AMD_DAL_CLK_MGR_DCN401 = $(addprefix $(AMDDALPATH)/dc/clk_mgr/dcn401/,$(CLK_MGR_DCN401))
+
+AMD_DISPLAY_FILES += $(AMD_DAL_CLK_MGR_DCN401)
 endif
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/clk_mgr.c
index 6127d6045336..15cf13ec5302 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/clk_mgr.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/clk_mgr.c
@@ -23,13 +23,12 @@
  *
  */
 
-#include <linux/slab.h>
-
 #include "dal_asic_id.h"
 #include "dc_types.h"
 #include "dccg.h"
 #include "clk_mgr_internal.h"
-#include "link.h"
+#include "dc_state_priv.h"
+#include "link_service.h"
 
 #include "dce100/dce_clk_mgr.h"
 #include "dce110/dce110_clk_mgr.h"
@@ -48,6 +47,8 @@
 #include "dcn315/dcn315_clk_mgr.h"
 #include "dcn316/dcn316_clk_mgr.h"
 #include "dcn32/dcn32_clk_mgr.h"
+#include "dcn35/dcn35_clk_mgr.h"
+#include "dcn401/dcn401_clk_mgr.h"
 
 int clk_mgr_helper_get_active_display_cnt(
 		struct dc *dc,
@@ -58,20 +59,15 @@ int clk_mgr_helper_get_active_display_cnt(
 	display_count = 0;
 	for (i = 0; i < context->stream_count; i++) {
 		const struct dc_stream_state *stream = context->streams[i];
+		const struct dc_stream_status *stream_status = &context->stream_status[i];
 
 		/* Don't count SubVP phantom pipes as part of active
 		 * display count
 		 */
-		if (stream->mall_stream_config.type == SUBVP_PHANTOM)
+		if (dc_state_get_stream_subvp_type(context, stream) == SUBVP_PHANTOM)
 			continue;
 
-		/*
-		 * Only notify active stream or virtual stream.
-		 * Need to notify virtual stream to work around
-		 * headless case. HPD does not fire when system is in
-		 * S0i2.
-		 */
-		if (!stream->dpms_off || stream->signal == SIGNAL_TYPE_VIRTUAL)
+		if (!stream->dpms_off || dc->is_switch_in_progress_dest || (stream_status && stream_status->plane_count))
 			display_count++;
 	}
 
@@ -117,6 +113,7 @@ void clk_mgr_exit_optimized_pwr_state(const struct dc *dc, struct clk_mgr *clk_m
 				continue;
 			clk_mgr->psr_allow_active_cache = edp_link->psr_settings.psr_allow_active;
 			dc->link_srv->edp_set_psr_allow_active(edp_link, &allow_active, false, false, NULL);
+			dc->link_srv->edp_set_replay_allow_active(edp_link, &allow_active, false, false, NULL);
 		}
 	}
 
@@ -137,6 +134,8 @@ void clk_mgr_optimize_pwr_state(const struct dc *dc, struct clk_mgr *clk_mgr)
 				continue;
 			dc->link_srv->edp_set_psr_allow_active(edp_link,
 					&clk_mgr->psr_allow_active_cache, false, false, NULL);
+			dc->link_srv->edp_set_replay_allow_active(edp_link,
+					&clk_mgr->psr_allow_active_cache, false, false, NULL);
 		}
 	}
 
@@ -159,7 +158,6 @@ struct clk_mgr *dc_clk_mgr_create(struct dc_context *ctx, struct pp_smu_funcs *p
 			return NULL;
 		}
 		dce60_clk_mgr_construct(ctx, clk_mgr);
-		dce_clk_mgr_construct(ctx, clk_mgr);
 		return &clk_mgr->base;
 	}
 #endif
@@ -269,7 +267,7 @@ struct clk_mgr *dc_clk_mgr_create(struct dc_context *ctx, struct pp_smu_funcs *p
 			dcn3_clk_mgr_construct(ctx, clk_mgr, pp_smu, dccg);
 			return &clk_mgr->base;
 		}
-		if (asic_id.chip_id == DEVICE_ID_NV_13FE) {
+		if (ctx->dce_version == DCN_VERSION_2_01) {
 			dcn201_clk_mgr_construct(ctx, clk_mgr, pp_smu, dccg);
 			return &clk_mgr->base;
 		}
@@ -326,16 +324,14 @@ struct clk_mgr *dc_clk_mgr_create(struct dc_context *ctx, struct pp_smu_funcs *p
 	}
 		break;
 	case AMDGPU_FAMILY_GC_11_0_0: {
-	    struct clk_mgr_internal *clk_mgr = kzalloc(sizeof(*clk_mgr), GFP_KERNEL);
-
-	    if (clk_mgr == NULL) {
-		BREAK_TO_DEBUGGER();
-		return NULL;
-	    }
+		struct clk_mgr_internal *clk_mgr = kzalloc(sizeof(*clk_mgr), GFP_KERNEL);
 
-	    dcn32_clk_mgr_construct(ctx, clk_mgr, pp_smu, dccg);
-	    return &clk_mgr->base;
-	    break;
+		if (clk_mgr == NULL) {
+			BREAK_TO_DEBUGGER();
+			return NULL;
+		}
+		dcn32_clk_mgr_construct(ctx, clk_mgr, pp_smu, dccg);
+		return &clk_mgr->base;
 	}
 
 	case AMDGPU_FAMILY_GC_11_0_1: {
@@ -351,7 +347,34 @@ struct clk_mgr *dc_clk_mgr_create(struct dc_context *ctx, struct pp_smu_funcs *p
 	}
 	break;
 
-#endif /* CONFIG_DRM_AMD_DC_FP - Family RV */
+	case AMDGPU_FAMILY_GC_11_5_0: {
+		struct clk_mgr_dcn35 *clk_mgr = kzalloc(sizeof(*clk_mgr), GFP_KERNEL);
+
+		if (clk_mgr == NULL) {
+			BREAK_TO_DEBUGGER();
+			return NULL;
+		}
+		if (ctx->dce_version == DCN_VERSION_3_51)
+			dcn351_clk_mgr_construct(ctx, clk_mgr, pp_smu, dccg);
+		else
+			dcn35_clk_mgr_construct(ctx, clk_mgr, pp_smu, dccg);
+
+		return &clk_mgr->base.base;
+	}
+	break;
+
+	case AMDGPU_FAMILY_GC_12_0_0: {
+		struct clk_mgr_internal *clk_mgr = dcn401_clk_mgr_construct(ctx, dccg);
+
+		if (clk_mgr == NULL) {
+			BREAK_TO_DEBUGGER();
+			return NULL;
+		}
+
+		return &clk_mgr->base;
+	}
+	break;
+#endif	/* CONFIG_DRM_AMD_DC_FP */
 	default:
 		ASSERT(0); /* Unknown Asic */
 		break;
@@ -402,6 +425,13 @@ void dc_destroy_clk_mgr(struct clk_mgr *clk_mgr_base)
 		dcn314_clk_mgr_destroy(clk_mgr);
 		break;
 
+	case AMDGPU_FAMILY_GC_11_5_0:
+		dcn35_clk_mgr_destroy(clk_mgr);
+		break;
+	case AMDGPU_FAMILY_GC_12_0_0:
+		dcn401_clk_mgr_destroy(clk_mgr);
+		break;
+
 	default:
 		break;
 	}
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dce100/dce_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dce100/dce_clk_mgr.c
index 26feefbb8990..6131ede2db7a 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dce100/dce_clk_mgr.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dce100/dce_clk_mgr.c
@@ -72,9 +72,9 @@ static const struct state_dependent_clocks dce80_max_clks_by_state[] = {
 /* ClocksStateLow */
 { .display_clk_khz = 352000, .pixel_clk_khz = 330000},
 /* ClocksStateNominal */
-{ .display_clk_khz = 600000, .pixel_clk_khz = 400000 },
+{ .display_clk_khz = 625000, .pixel_clk_khz = 400000 },
 /* ClocksStatePerformance */
-{ .display_clk_khz = 600000, .pixel_clk_khz = 400000 } };
+{ .display_clk_khz = 625000, .pixel_clk_khz = 400000 } };
 
 int dentist_get_divider_from_did(int did)
 {
@@ -245,6 +245,11 @@ int dce_set_clock(
 	pxl_clk_params.target_pixel_clock_100hz = requested_clk_khz * 10;
 	pxl_clk_params.pll_id = CLOCK_SOURCE_ID_DFS;
 
+	/* DCE 6.0, DCE 6.4: engine clock is the same as PLL0 */
+	if (clk_mgr_base->ctx->dce_version == DCE_VERSION_6_0 ||
+	    clk_mgr_base->ctx->dce_version == DCE_VERSION_6_4)
+		pxl_clk_params.pll_id = CLOCK_SOURCE_ID_PLL0;
+
 	if (clk_mgr_dce->dfs_bypass_active)
 		pxl_clk_params.flags.SET_DISPCLK_DFS_BYPASS = true;
 
@@ -386,8 +391,6 @@ static void dce_pplib_apply_display_requirements(
 {
 	struct dm_pp_display_configuration *pp_display_cfg = &context->pp_display_cfg;
 
-	pp_display_cfg->avail_mclk_switch_time_us = dce110_get_min_vblank_time_us(context);
-
 	dce110_fill_display_configs(context, pp_display_cfg);
 
 	if (memcmp(&dc->current_state->pp_display_cfg, pp_display_cfg, sizeof(*pp_display_cfg)) !=  0)
@@ -400,11 +403,9 @@ static void dce_update_clocks(struct clk_mgr *clk_mgr_base,
 {
 	struct clk_mgr_internal *clk_mgr_dce = TO_CLK_MGR_INTERNAL(clk_mgr_base);
 	struct dm_pp_power_level_change_request level_change_req;
-	int patched_disp_clk = context->bw_ctx.bw.dce.dispclk_khz;
-
-	/*TODO: W/A for dal3 linux, investigate why this works */
-	if (!clk_mgr_dce->dfs_bypass_active)
-		patched_disp_clk = patched_disp_clk * 115 / 100;
+	const int max_disp_clk =
+		clk_mgr_dce->max_clks_by_state[DM_PP_CLOCKS_STATE_PERFORMANCE].display_clk_khz;
+	int patched_disp_clk = MIN(max_disp_clk, context->bw_ctx.bw.dce.dispclk_khz);
 
 	level_change_req.power_level = dce_get_required_clocks_state(clk_mgr_base, context);
 	/* get max clock state from PPLIB */
@@ -462,6 +463,9 @@ void dce_clk_mgr_construct(
 		clk_mgr->max_clks_state = DM_PP_CLOCKS_STATE_NOMINAL;
 	clk_mgr->cur_min_clks_state = DM_PP_CLOCKS_STATE_INVALID;
 
+	base->clks.max_supported_dispclk_khz =
+		clk_mgr->max_clks_by_state[DM_PP_CLOCKS_STATE_PERFORMANCE].display_clk_khz;
+
 	dce_clock_read_integrated_info(clk_mgr);
 	dce_clock_read_ss_info(clk_mgr);
 }
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dce110/dce110_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dce110/dce110_clk_mgr.c
index 78df96882d6e..d50b9440210e 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dce110/dce110_clk_mgr.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dce110/dce110_clk_mgr.c
@@ -120,9 +120,15 @@ void dce110_fill_display_configs(
 	const struct dc_state *context,
 	struct dm_pp_display_configuration *pp_display_cfg)
 {
+	struct dc *dc = context->clk_mgr->ctx->dc;
 	int j;
 	int num_cfgs = 0;
 
+	pp_display_cfg->avail_mclk_switch_time_us = dce110_get_min_vblank_time_us(context);
+	pp_display_cfg->disp_clk_khz = dc->clk_mgr->clks.dispclk_khz;
+	pp_display_cfg->avail_mclk_switch_time_in_disp_active_us = 0;
+	pp_display_cfg->crtc_index = dc->res_pool->res_cap->num_timing_generator;
+
 	for (j = 0; j < context->stream_count; j++) {
 		int k;
 
@@ -158,12 +164,29 @@ void dce110_fill_display_configs(
 			stream->link->cur_link_settings.link_rate;
 		cfg->link_settings.link_spread =
 			stream->link->cur_link_settings.link_spread;
-		cfg->sym_clock = stream->phy_pix_clk;
+		cfg->pixel_clock = stream->phy_pix_clk;
 		/* Round v_refresh*/
 		cfg->v_refresh = stream->timing.pix_clk_100hz * 100;
 		cfg->v_refresh /= stream->timing.h_total;
 		cfg->v_refresh = (cfg->v_refresh + stream->timing.v_total / 2)
 							/ stream->timing.v_total;
+
+		/* Find first CRTC index and calculate its line time.
+		 * This is necessary for DPM on SI GPUs.
+		 */
+		if (cfg->pipe_idx < pp_display_cfg->crtc_index) {
+			const struct dc_crtc_timing *timing =
+				&context->streams[0]->timing;
+
+			pp_display_cfg->crtc_index = cfg->pipe_idx;
+			pp_display_cfg->line_time_in_us =
+				timing->h_total * 10000 / timing->pix_clk_100hz;
+		}
+	}
+
+	if (!num_cfgs) {
+		pp_display_cfg->crtc_index = 0;
+		pp_display_cfg->line_time_in_us = 0;
 	}
 
 	pp_display_cfg->display_count = num_cfgs;
@@ -195,7 +218,7 @@ void dce11_pplib_apply_display_requirements(
 	 * , then change minimum memory clock based on real-time bandwidth
 	 * limitation.
 	 */
-	if ((dc->ctx->asic_id.chip_family == FAMILY_AI) &&
+	if (dc->bw_vbios && (dc->ctx->asic_id.chip_family == FAMILY_AI) &&
 	     ASICREV_IS_VEGA20_P(dc->ctx->asic_id.hw_internal_rev) && (context->stream_count >= 2)) {
 		pp_display_cfg->min_memory_clock_khz = max(pp_display_cfg->min_memory_clock_khz,
 							   (uint32_t) div64_s64(
@@ -223,25 +246,8 @@ void dce11_pplib_apply_display_requirements(
 	pp_display_cfg->min_engine_clock_deep_sleep_khz
 			= context->bw_ctx.bw.dce.sclk_deep_sleep_khz;
 
-	pp_display_cfg->avail_mclk_switch_time_us =
-						dce110_get_min_vblank_time_us(context);
-	/* TODO: dce11.2*/
-	pp_display_cfg->avail_mclk_switch_time_in_disp_active_us = 0;
-
-	pp_display_cfg->disp_clk_khz = dc->clk_mgr->clks.dispclk_khz;
-
 	dce110_fill_display_configs(context, pp_display_cfg);
 
-	/* TODO: is this still applicable?*/
-	if (pp_display_cfg->display_count == 1) {
-		const struct dc_crtc_timing *timing =
-			&context->streams[0]->timing;
-
-		pp_display_cfg->crtc_index =
-			pp_display_cfg->disp_configs[0].pipe_idx;
-		pp_display_cfg->line_time_in_us = timing->h_total * 10000 / timing->pix_clk_100hz;
-	}
-
 	if (memcmp(&dc->current_state->pp_display_cfg, pp_display_cfg, sizeof(*pp_display_cfg)) !=  0)
 		dm_pp_apply_display_requirements(dc->ctx, pp_display_cfg);
 }
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dce112/dce112_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dce112/dce112_clk_mgr.c
index 934e6423dc1a..1f36ad8a7de4 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dce112/dce112_clk_mgr.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dce112/dce112_clk_mgr.c
@@ -111,12 +111,10 @@ int dce112_set_clock(struct clk_mgr *clk_mgr_base, int requested_clk_khz)
 
 	bp->funcs->set_dce_clock(bp, &dce_clk_params);
 
-	if (!IS_FPGA_MAXIMUS_DC(dc->ctx->dce_environment)) {
-		if (dmcu && dmcu->funcs->is_dmcu_initialized(dmcu)) {
-			if (clk_mgr_dce->dfs_bypass_disp_clk != actual_clock)
-				dmcu->funcs->set_psr_wait_loop(dmcu,
-						actual_clock / 1000 / 7);
-		}
+	if (dmcu && dmcu->funcs->is_dmcu_initialized(dmcu)) {
+		if (clk_mgr_dce->dfs_bypass_disp_clk != actual_clock)
+			dmcu->funcs->set_psr_wait_loop(dmcu,
+					actual_clock / 1000 / 7);
 	}
 
 	clk_mgr_dce->dfs_bypass_disp_clk = actual_clock;
@@ -153,12 +151,10 @@ int dce112_set_dispclk(struct clk_mgr_internal *clk_mgr, int requested_clk_khz)
 		clk_mgr->cur_min_clks_state = DM_PP_CLOCKS_STATE_NOMINAL;
 
 
-	if (!IS_FPGA_MAXIMUS_DC(dc->ctx->dce_environment)) {
-		if (dmcu && dmcu->funcs->is_dmcu_initialized(dmcu)) {
-			if (clk_mgr->dfs_bypass_disp_clk != actual_clock)
-				dmcu->funcs->set_psr_wait_loop(dmcu,
-						actual_clock / 1000 / 7);
-		}
+	if (dmcu && dmcu->funcs->is_dmcu_initialized(dmcu)) {
+		if (clk_mgr->dfs_bypass_disp_clk != actual_clock)
+			dmcu->funcs->set_psr_wait_loop(dmcu,
+					actual_clock / 1000 / 7);
 	}
 
 	clk_mgr->dfs_bypass_disp_clk = actual_clock;
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dce120/dce120_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dce120/dce120_clk_mgr.c
index 5399b8cf6b75..c9ba7b3fd2c3 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dce120/dce120_clk_mgr.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dce120/dce120_clk_mgr.c
@@ -30,7 +30,7 @@
 #include "dce110/dce110_clk_mgr.h"
 #include "dce120_clk_mgr.h"
 #include "dce100/dce_clk_mgr.h"
-#include "dce120/dce120_hw_sequencer.h"
+#include "dce120/dce120_hwseq.h"
 
 static const struct state_dependent_clocks dce120_max_clks_by_state[] = {
 /*ClocksStateInvalid - should not be used*/
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dce60/dce60_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dce60/dce60_clk_mgr.c
index 0267644717b2..69dd80d9f738 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dce60/dce60_clk_mgr.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dce60/dce60_clk_mgr.c
@@ -83,22 +83,13 @@ static const struct state_dependent_clocks dce60_max_clks_by_state[] = {
 static int dce60_get_dp_ref_freq_khz(struct clk_mgr *clk_mgr_base)
 {
 	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
-	int dprefclk_wdivider;
-	int dp_ref_clk_khz;
-	int target_div;
+	struct dc_context *ctx = clk_mgr_base->ctx;
+	int dp_ref_clk_khz = 0;
 
-	/* DCE6 has no DPREFCLK_CNTL to read DP Reference Clock source */
-
-	/* Read the mmDENTIST_DISPCLK_CNTL to get the currently
-	 * programmed DID DENTIST_DPREFCLK_WDIVIDER*/
-	REG_GET(DENTIST_DISPCLK_CNTL, DENTIST_DPREFCLK_WDIVIDER, &dprefclk_wdivider);
-
-	/* Convert DENTIST_DPREFCLK_WDIVIDERto actual divider*/
-	target_div = dentist_get_divider_from_did(dprefclk_wdivider);
-
-	/* Calculate the current DFS clock, in kHz.*/
-	dp_ref_clk_khz = (DENTIST_DIVIDER_RANGE_SCALE_FACTOR
-		* clk_mgr->base.dentist_vco_freq_khz) / target_div;
+	if (ASIC_REV_IS_TAHITI_P(ctx->asic_id.hw_internal_rev))
+		dp_ref_clk_khz = ctx->dc_bios->fw_info.default_display_engine_pll_frequency;
+	else
+		dp_ref_clk_khz = clk_mgr_base->clks.dispclk_khz;
 
 	return dce_adjust_dp_ref_freq_for_ss(clk_mgr, dp_ref_clk_khz);
 }
@@ -109,8 +100,6 @@ static void dce60_pplib_apply_display_requirements(
 {
 	struct dm_pp_display_configuration *pp_display_cfg = &context->pp_display_cfg;
 
-	pp_display_cfg->avail_mclk_switch_time_us = dce110_get_min_vblank_time_us(context);
-
 	dce110_fill_display_configs(context, pp_display_cfg);
 
 	if (memcmp(&dc->current_state->pp_display_cfg, pp_display_cfg, sizeof(*pp_display_cfg)) !=  0)
@@ -123,11 +112,9 @@ static void dce60_update_clocks(struct clk_mgr *clk_mgr_base,
 {
 	struct clk_mgr_internal *clk_mgr_dce = TO_CLK_MGR_INTERNAL(clk_mgr_base);
 	struct dm_pp_power_level_change_request level_change_req;
-	int patched_disp_clk = context->bw_ctx.bw.dce.dispclk_khz;
-
-	/*TODO: W/A for dal3 linux, investigate why this works */
-	if (!clk_mgr_dce->dfs_bypass_active)
-		patched_disp_clk = patched_disp_clk * 115 / 100;
+	const int max_disp_clk =
+		clk_mgr_dce->max_clks_by_state[DM_PP_CLOCKS_STATE_PERFORMANCE].display_clk_khz;
+	int patched_disp_clk = MIN(max_disp_clk, context->bw_ctx.bw.dce.dispclk_khz);
 
 	level_change_req.power_level = dce_get_required_clocks_state(clk_mgr_base, context);
 	/* get max clock state from PPLIB */
@@ -160,6 +147,8 @@ void dce60_clk_mgr_construct(
 		struct dc_context *ctx,
 		struct clk_mgr_internal *clk_mgr)
 {
+	struct clk_mgr *base = &clk_mgr->base;
+
 	dce_clk_mgr_construct(ctx, clk_mgr);
 
 	memcpy(clk_mgr->max_clks_by_state,
@@ -170,5 +159,8 @@ void dce60_clk_mgr_construct(
 	clk_mgr->clk_mgr_shift = &disp_clk_shift;
 	clk_mgr->clk_mgr_mask = &disp_clk_mask;
 	clk_mgr->base.funcs = &dce60_funcs;
+
+	base->clks.max_supported_dispclk_khz =
+		clk_mgr->max_clks_by_state[DM_PP_CLOCKS_STATE_PERFORMANCE].display_clk_khz;
 }
 
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn10/rv1_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn10/rv1_clk_mgr.c
index 60761ff3cbf1..369421e46c52 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn10/rv1_clk_mgr.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn10/rv1_clk_mgr.c
@@ -23,9 +23,6 @@
  *
  */
 
-#include <linux/slab.h>
-
-#include "reg_helper.h"
 #include "core_types.h"
 #include "clk_mgr_internal.h"
 #include "rv1_clk_mgr.h"
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn10/rv1_clk_mgr_clk.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn10/rv1_clk_mgr_clk.c
deleted file mode 100644
index 61dd12198a3c..000000000000
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn10/rv1_clk_mgr_clk.c
+++ /dev/null
@@ -1,79 +0,0 @@
-/*
- * Copyright 2012-16 Advanced Micro Devices, Inc.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
- * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
- * OTHER DEALINGS IN THE SOFTWARE.
- *
- * Authors: AMD
- *
- */
-
-#include "reg_helper.h"
-#include "clk_mgr_internal.h"
-#include "rv1_clk_mgr_clk.h"
-
-#include "ip/Discovery/hwid.h"
-#include "ip/Discovery/v1/ip_offset_1.h"
-#include "ip/CLK/clk_10_0_default.h"
-#include "ip/CLK/clk_10_0_offset.h"
-#include "ip/CLK/clk_10_0_reg.h"
-#include "ip/CLK/clk_10_0_sh_mask.h"
-
-#include "dce100/dce_clk_mgr.h"
-
-#define CLK_BASE_INNER(inst) \
-	CLK_BASE__INST ## inst ## _SEG0
-
-
-#define CLK_REG(reg_name, block, inst)\
-	CLK_BASE(mm ## block ## _ ## inst ## _ ## reg_name ## _BASE_IDX) + \
-					mm ## block ## _ ## inst ## _ ## reg_name
-
-#define REG(reg_name) \
-	CLK_REG(reg_name, CLK0, 0)
-
-
-/* Only used by testing framework*/
-void rv1_dump_clk_registers(struct clk_state_registers *regs, struct clk_bypass *bypass, struct clk_mgr *clk_mgr_base)
-{
-	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
-
-		regs->CLK0_CLK8_CURRENT_CNT = REG_READ(CLK0_CLK8_CURRENT_CNT) / 10; //dcf clk
-
-		bypass->dcfclk_bypass = REG_READ(CLK0_CLK8_BYPASS_CNTL) & 0x0007;
-		if (bypass->dcfclk_bypass < 0 || bypass->dcfclk_bypass > 4)
-			bypass->dcfclk_bypass = 0;
-
-
-		regs->CLK0_CLK8_DS_CNTL = REG_READ(CLK0_CLK8_DS_CNTL) / 10;	//dcf deep sleep divider
-
-		regs->CLK0_CLK8_ALLOW_DS = REG_READ(CLK0_CLK8_ALLOW_DS); //dcf deep sleep allow
-
-		regs->CLK0_CLK10_CURRENT_CNT = REG_READ(CLK0_CLK10_CURRENT_CNT) / 10; //dpref clk
-
-		bypass->dispclk_pypass = REG_READ(CLK0_CLK10_BYPASS_CNTL) & 0x0007;
-		if (bypass->dispclk_pypass < 0 || bypass->dispclk_pypass > 4)
-			bypass->dispclk_pypass = 0;
-
-		regs->CLK0_CLK11_CURRENT_CNT = REG_READ(CLK0_CLK11_CURRENT_CNT) / 10; //disp clk
-
-		bypass->dprefclk_bypass = REG_READ(CLK0_CLK11_BYPASS_CNTL) & 0x0007;
-		if (bypass->dprefclk_bypass < 0 || bypass->dprefclk_bypass > 4)
-			bypass->dprefclk_bypass = 0;
-
-}
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn10/rv1_clk_mgr_vbios_smu.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn10/rv1_clk_mgr_vbios_smu.c
index 450eaead4f20..d82a52319088 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn10/rv1_clk_mgr_vbios_smu.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn10/rv1_clk_mgr_vbios_smu.c
@@ -26,7 +26,6 @@
 #include "core_types.h"
 #include "clk_mgr_internal.h"
 #include "reg_helper.h"
-#include <linux/delay.h>
 
 #include "rv1_clk_mgr_vbios_smu.h"
 
@@ -135,27 +134,11 @@ int rv1_vbios_smu_set_dispclk(struct clk_mgr_internal *clk_mgr, int requested_di
 			VBIOSSMC_MSG_SetDispclkFreq,
 			khz_to_mhz_ceil(requested_dispclk_khz));
 
-	if (!IS_FPGA_MAXIMUS_DC(dc->ctx->dce_environment)) {
-		if (dmcu && dmcu->funcs->is_dmcu_initialized(dmcu)) {
-			if (clk_mgr->dfs_bypass_disp_clk != actual_dispclk_set_mhz)
-				dmcu->funcs->set_psr_wait_loop(dmcu,
-						actual_dispclk_set_mhz / 7);
-		}
+	if (dmcu && dmcu->funcs->is_dmcu_initialized(dmcu)) {
+		if (clk_mgr->dfs_bypass_disp_clk != actual_dispclk_set_mhz)
+			dmcu->funcs->set_psr_wait_loop(dmcu,
+					actual_dispclk_set_mhz / 7);
 	}
 
 	return actual_dispclk_set_mhz * 1000;
 }
-
-int rv1_vbios_smu_set_dprefclk(struct clk_mgr_internal *clk_mgr)
-{
-	int actual_dprefclk_set_mhz = -1;
-
-	actual_dprefclk_set_mhz = rv1_vbios_smu_send_msg_with_param(
-			clk_mgr,
-			VBIOSSMC_MSG_SetDprefclkFreq,
-			khz_to_mhz_ceil(clk_mgr->base.dprefclk_khz));
-
-	/* TODO: add code for programing DP DTO, currently this is down by command table */
-
-	return actual_dprefclk_set_mhz * 1000;
-}
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn10/rv1_clk_mgr_vbios_smu.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn10/rv1_clk_mgr_vbios_smu.h
index 083cb3158859..81d7c912549c 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn10/rv1_clk_mgr_vbios_smu.h
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn10/rv1_clk_mgr_vbios_smu.h
@@ -27,6 +27,5 @@
 #define DAL_DC_DCN10_RV1_CLK_MGR_VBIOS_SMU_H_
 
 int rv1_vbios_smu_set_dispclk(struct clk_mgr_internal *clk_mgr, int requested_dispclk_khz);
-int rv1_vbios_smu_set_dprefclk(struct clk_mgr_internal *clk_mgr);
 
 #endif /* DAL_DC_DCN10_RV1_CLK_MGR_VBIOS_SMU_H_ */
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn20/dcn20_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn20/dcn20_clk_mgr.c
index 650f3b4b562e..bb4f3bd7532e 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn20/dcn20_clk_mgr.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn20/dcn20_clk_mgr.c
@@ -157,7 +157,7 @@ void dcn20_update_clocks_update_dentist(struct clk_mgr_internal *clk_mgr, struct
 			int32_t N;
 			int32_t j;
 
-			if (!pipe_ctx->stream)
+			if (!resource_is_pipe_type(pipe_ctx, OTG_MASTER))
 				continue;
 			/* Virtual encoders don't have this function */
 			if (!stream_enc->funcs->get_fifo_cal_average_level)
@@ -188,7 +188,7 @@ void dcn20_update_clocks_update_dentist(struct clk_mgr_internal *clk_mgr, struct
 			int32_t N;
 			int32_t j;
 
-			if (!pipe_ctx->stream)
+			if (!resource_is_pipe_type(pipe_ctx, OTG_MASTER))
 				continue;
 			/* Virtual encoders don't have this function */
 			if (!stream_enc->funcs->get_fifo_cal_average_level)
@@ -503,7 +503,7 @@ static void dcn2_notify_link_rate_change(struct clk_mgr *clk_mgr_base, struct dc
 
 	clk_mgr->cur_phyclk_req_table[link->link_index] = link->cur_link_settings.link_rate * LINK_RATE_REF_FREQ_IN_KHZ;
 
-	for (i = 0; i < MAX_PIPES * 2; i++) {
+	for (i = 0; i < MAX_LINKS; i++) {
 		if (clk_mgr->cur_phyclk_req_table[i] > max_phyclk_req)
 			max_phyclk_req = clk_mgr->cur_phyclk_req_table[i];
 	}
@@ -531,6 +531,11 @@ void dcn20_clk_mgr_construct(
 		struct pp_smu_funcs *pp_smu,
 		struct dccg *dccg)
 {
+	int dprefclk_did;
+	int target_div;
+	uint32_t pll_req_reg;
+	struct fixed31_32 pll_req;
+
 	clk_mgr->base.ctx = ctx;
 	clk_mgr->pp_smu = pp_smu;
 	clk_mgr->base.funcs = &dcn2_funcs;
@@ -547,42 +552,34 @@ void dcn20_clk_mgr_construct(
 
 	clk_mgr->base.dprefclk_khz = 700000; // 700 MHz planned if VCO is 3.85 GHz, will be retrieved
 
-	if (IS_FPGA_MAXIMUS_DC(ctx->dce_environment)) {
-		dcn2_funcs.update_clocks = dcn2_update_clocks_fpga;
-		clk_mgr->base.dentist_vco_freq_khz = 3850000;
+	/* DFS Slice 2 should be used for DPREFCLK */
+	dprefclk_did = REG_READ(CLK3_CLK2_DFS_CNTL);
+	/* Convert DPREFCLK DFS Slice DID to actual divider */
+	target_div = dentist_get_divider_from_did(dprefclk_did);
+	/* get FbMult value */
+	pll_req_reg = REG_READ(CLK3_CLK_PLL_REQ);
 
-	} else {
-		/* DFS Slice 2 should be used for DPREFCLK */
-		int dprefclk_did = REG_READ(CLK3_CLK2_DFS_CNTL);
-		/* Convert DPREFCLK DFS Slice DID to actual divider*/
-		int target_div = dentist_get_divider_from_did(dprefclk_did);
-
-		/* get FbMult value */
-		uint32_t pll_req_reg = REG_READ(CLK3_CLK_PLL_REQ);
-		struct fixed31_32 pll_req;
-
-		/* set up a fixed-point number
-		 * this works because the int part is on the right edge of the register
-		 * and the frac part is on the left edge
-		 */
+	/* set up a fixed-point number
+	 * this works because the int part is on the right edge of the register
+	 * and the frac part is on the left edge
+	 */
 
-		pll_req = dc_fixpt_from_int(pll_req_reg & clk_mgr->clk_mgr_mask->FbMult_int);
-		pll_req.value |= pll_req_reg & clk_mgr->clk_mgr_mask->FbMult_frac;
+	pll_req = dc_fixpt_from_int(pll_req_reg & clk_mgr->clk_mgr_mask->FbMult_int);
+	pll_req.value |= pll_req_reg & clk_mgr->clk_mgr_mask->FbMult_frac;
 
-		/* multiply by REFCLK period */
-		pll_req = dc_fixpt_mul_int(pll_req, 100000);
+	/* multiply by REFCLK period */
+	pll_req = dc_fixpt_mul_int(pll_req, 100000);
 
-		/* integer part is now VCO frequency in kHz */
-		clk_mgr->base.dentist_vco_freq_khz = dc_fixpt_floor(pll_req);
+	/* integer part is now VCO frequency in kHz */
+	clk_mgr->base.dentist_vco_freq_khz = dc_fixpt_floor(pll_req);
 
-		/* in case we don't get a value from the register, use default */
-		if (clk_mgr->base.dentist_vco_freq_khz == 0)
-			clk_mgr->base.dentist_vco_freq_khz = 3850000;
+	/* in case we don't get a value from the register, use default */
+	if (clk_mgr->base.dentist_vco_freq_khz == 0)
+		clk_mgr->base.dentist_vco_freq_khz = 3850000;
 
-		/* Calculate the DPREFCLK in kHz.*/
-		clk_mgr->base.dprefclk_khz = (DENTIST_DIVIDER_RANGE_SCALE_FACTOR
-			* clk_mgr->base.dentist_vco_freq_khz) / target_div;
-	}
+	/* Calculate the DPREFCLK in kHz.*/
+	clk_mgr->base.dprefclk_khz = (DENTIST_DIVIDER_RANGE_SCALE_FACTOR
+		* clk_mgr->base.dentist_vco_freq_khz) / target_div;
 	//Integrated_info table does not exist on dGPU projects so should not be referenced
 	//anywhere in code for dGPUs.
 	//Also there is no plan for now that DFS BYPASS will be used on NV10/12/14.
@@ -590,4 +587,3 @@ void dcn20_clk_mgr_construct(
 
 	dce_clock_read_ss_info(clk_mgr);
 }
-
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn201/dcn201_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn201/dcn201_clk_mgr.c
index 811720749faf..76c612ecfe3c 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn201/dcn201_clk_mgr.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn201/dcn201_clk_mgr.c
@@ -34,8 +34,8 @@
 #include "dm_services.h"
 
 #include "cyan_skillfish_ip_offset.h"
-#include "dcn/dcn_2_0_3_offset.h"
-#include "dcn/dcn_2_0_3_sh_mask.h"
+#include "dcn/dcn_2_0_1_offset.h"
+#include "dcn/dcn_2_0_1_sh_mask.h"
 #include "clk/clk_11_0_1_offset.h"
 #include "clk/clk_11_0_1_sh_mask.h"
 
@@ -59,8 +59,6 @@
 
 #define CTX \
 	clk_mgr->base.ctx
-#define DC_LOGGER \
-	clk_mgr->base.ctx->logger
 
 static const struct clk_mgr_registers clk_mgr_regs = {
 		CLK_COMMON_REG_LIST_DCN_201()
@@ -102,13 +100,19 @@ static void dcn201_update_clocks(struct clk_mgr *clk_mgr_base,
 
 	if (clk_mgr_base->clks.dispclk_khz == 0 ||
 	    dc->debug.force_clock_mode & 0x1) {
+		/* this is from resume or boot up, if forced_clock cfg option
+		 * used, we bypass program dispclk and DPPCLK, but need set them
+		 * for S3.
+		 */
+
 		force_reset = true;
+		/* force_clock_mode 0x1:  force reset the clock even it is the
+		 * same clock as long as it is in Passive level.
+		 */
 
 		dcn2_read_clocks_from_hw_dentist(clk_mgr_base);
 	}
 
-	clk_mgr_helper_get_active_display_cnt(dc, context);
-
 	if (should_set_clock(safe_to_lower, new_clocks->phyclk_khz, clk_mgr_base->clks.phyclk_khz))
 		clk_mgr_base->clks.phyclk_khz = new_clocks->phyclk_khz;
 
@@ -152,11 +156,14 @@ static void dcn201_update_clocks(struct clk_mgr *clk_mgr_base,
 
 	if (dc->config.forced_clocks == false || (force_reset && safe_to_lower)) {
 		if (dpp_clock_lowered) {
+			// if clock is being lowered, increase DTO before lowering refclk
 			dcn20_update_clocks_update_dpp_dto(clk_mgr, context, safe_to_lower);
 			dcn20_update_clocks_update_dentist(clk_mgr, context);
 		} else {
+			// if clock is being raised, increase refclk before lowering DTO
 			if (update_dppclk || update_dispclk)
 				dcn20_update_clocks_update_dentist(clk_mgr, context);
+			// always update dtos unless clock is lowered and not safe to lower
 			dcn20_update_clocks_update_dpp_dto(clk_mgr, context, safe_to_lower);
 		}
 	}
@@ -190,23 +197,17 @@ void dcn201_clk_mgr_construct(struct dc_context *ctx,
 	clk_mgr->dprefclk_ss_divider = 1000;
 	clk_mgr->ss_on_dprefclk = false;
 
-	if (IS_FPGA_MAXIMUS_DC(ctx->dce_environment)) {
-		dcn201_funcs.update_clocks = dcn2_update_clocks_fpga;
-		clk_mgr->base.dprefclk_khz = 600000;
-		clk_mgr->base.dentist_vco_freq_khz = 3000000;
-	} else {
-		clk_mgr->base.dprefclk_khz = REG_READ(CLK4_CLK2_CURRENT_CNT);
-		clk_mgr->base.dprefclk_khz *= 100;
+	clk_mgr->base.dprefclk_khz = REG_READ(CLK4_CLK2_CURRENT_CNT);
+	clk_mgr->base.dprefclk_khz *= 100;
 
-		if (clk_mgr->base.dprefclk_khz == 0)
-			clk_mgr->base.dprefclk_khz = 600000;
+	if (clk_mgr->base.dprefclk_khz == 0)
+		clk_mgr->base.dprefclk_khz = 600000;
 
-		REG_GET(CLK4_CLK_PLL_REQ, FbMult_int, &clk_mgr->base.dentist_vco_freq_khz);
-		clk_mgr->base.dentist_vco_freq_khz *= 100000;
+	REG_GET(CLK4_CLK_PLL_REQ, FbMult_int, &clk_mgr->base.dentist_vco_freq_khz);
+	clk_mgr->base.dentist_vco_freq_khz *= 100000;
 
-		if (clk_mgr->base.dentist_vco_freq_khz == 0)
-			clk_mgr->base.dentist_vco_freq_khz = 3000000;
-	}
+	if (clk_mgr->base.dentist_vco_freq_khz == 0)
+		clk_mgr->base.dentist_vco_freq_khz = 3000000;
 
 	if (!debug->disable_dfs_bypass && bp->integrated_info)
 		if (bp->integrated_info->gpu_cap_info & DFS_BYPASS_ENABLE)
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn21/rn_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn21/rn_clk_mgr.c
index bd9fd0b54f46..e18097f82091 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn21/rn_clk_mgr.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn21/rn_clk_mgr.c
@@ -320,16 +320,16 @@ static void rn_dump_clk_registers(struct clk_state_registers_and_bypass *regs_an
 	regs_and_bypass->dppclk = internal.CLK1_CLK1_CURRENT_CNT / 10;
 
 	regs_and_bypass->dppclk_bypass = internal.CLK1_CLK1_BYPASS_CNTL & 0x0007;
-	if (regs_and_bypass->dppclk_bypass < 0 || regs_and_bypass->dppclk_bypass > 4)
+	if (regs_and_bypass->dppclk_bypass > 4)
 		regs_and_bypass->dppclk_bypass = 0;
 	regs_and_bypass->dcfclk_bypass = internal.CLK1_CLK3_BYPASS_CNTL & 0x0007;
-	if (regs_and_bypass->dcfclk_bypass < 0 || regs_and_bypass->dcfclk_bypass > 4)
+	if (regs_and_bypass->dcfclk_bypass > 4)
 		regs_and_bypass->dcfclk_bypass = 0;
 	regs_and_bypass->dispclk_bypass = internal.CLK1_CLK0_BYPASS_CNTL & 0x0007;
-	if (regs_and_bypass->dispclk_bypass < 0 || regs_and_bypass->dispclk_bypass > 4)
+	if (regs_and_bypass->dispclk_bypass > 4)
 		regs_and_bypass->dispclk_bypass = 0;
 	regs_and_bypass->dprefclk_bypass = internal.CLK1_CLK2_BYPASS_CNTL & 0x0007;
-	if (regs_and_bypass->dprefclk_bypass < 0 || regs_and_bypass->dprefclk_bypass > 4)
+	if (regs_and_bypass->dprefclk_bypass > 4)
 		regs_and_bypass->dprefclk_bypass = 0;
 
 	if (log_info->enabled) {
@@ -484,7 +484,8 @@ static void build_watermark_ranges(struct clk_bw_params *bw_params, struct pp_sm
 			ranges->reader_wm_sets[num_valid_sets].max_fill_clk_mhz = PP_SMU_WM_SET_RANGE_CLK_UNCONSTRAINED_MAX;
 
 			/* Modify previous watermark range to cover up to max */
-			ranges->reader_wm_sets[num_valid_sets - 1].max_fill_clk_mhz = PP_SMU_WM_SET_RANGE_CLK_UNCONSTRAINED_MAX;
+			if (num_valid_sets > 0)
+				ranges->reader_wm_sets[num_valid_sets - 1].max_fill_clk_mhz = PP_SMU_WM_SET_RANGE_CLK_UNCONSTRAINED_MAX;
 		}
 		num_valid_sets++;
 	}
@@ -548,7 +549,7 @@ static void rn_notify_link_rate_change(struct clk_mgr *clk_mgr_base, struct dc_l
 
 	clk_mgr->cur_phyclk_req_table[link->link_index] = link->cur_link_settings.link_rate * LINK_RATE_REF_FREQ_IN_KHZ;
 
-	for (i = 0; i < MAX_PIPES * 2; i++) {
+	for (i = 0; i < MAX_LINKS; i++) {
 		if (clk_mgr->cur_phyclk_req_table[i] > max_phyclk_req)
 			max_phyclk_req = clk_mgr->cur_phyclk_req_table[i];
 	}
@@ -642,7 +643,8 @@ static void rn_clk_mgr_helper_populate_bw_params(struct clk_bw_params *bw_params
 
 	j = -1;
 
-	ASSERT(PP_SMU_NUM_FCLK_DPM_LEVELS <= MAX_NUM_DPM_LVL);
+	static_assert(PP_SMU_NUM_FCLK_DPM_LEVELS <= MAX_NUM_DPM_LVL,
+		"number of reported FCLK DPM levels exceed maximum");
 
 	/* Find lowest DPM, FCLK is filled in reverse order*/
 
@@ -705,10 +707,9 @@ void rn_clk_mgr_construct(
 	struct dpm_clocks clock_table = { 0 };
 	enum pp_smu_status status = 0;
 	int is_green_sardine = 0;
+	struct clk_log_info log_info = {0};
 
-#if defined(CONFIG_DRM_AMD_DC_FP)
 	is_green_sardine = ASICREV_IS_GREEN_SARDINE(ctx->asic_id.hw_internal_rev);
-#endif
 
 	clk_mgr->base.ctx = ctx;
 	clk_mgr->base.funcs = &dcn21_funcs;
@@ -725,48 +726,41 @@ void rn_clk_mgr_construct(
 
 	clk_mgr->smu_ver = rn_vbios_smu_get_smu_version(clk_mgr);
 
-	if (IS_FPGA_MAXIMUS_DC(ctx->dce_environment)) {
-		dcn21_funcs.update_clocks = dcn2_update_clocks_fpga;
+	clk_mgr->periodic_retraining_disabled = rn_vbios_smu_is_periodic_retraining_disabled(clk_mgr);
+
+	/* SMU Version 55.51.0 and up no longer have an issue
+	 * that needs to limit minimum dispclk */
+	if (clk_mgr->smu_ver >= SMU_VER_55_51_0)
+		debug->min_disp_clk_khz = 0;
+
+	/* TODO: Check we get what we expect during bringup */
+	clk_mgr->base.dentist_vco_freq_khz = get_vco_frequency_from_reg(clk_mgr);
+
+	/* in case we don't get a value from the register, use default */
+	if (clk_mgr->base.dentist_vco_freq_khz == 0)
 		clk_mgr->base.dentist_vco_freq_khz = 3600000;
-	} else {
-		struct clk_log_info log_info = {0};
-
-		clk_mgr->periodic_retraining_disabled = rn_vbios_smu_is_periodic_retraining_disabled(clk_mgr);
-
-		/* SMU Version 55.51.0 and up no longer have an issue
-		 * that needs to limit minimum dispclk */
-		if (clk_mgr->smu_ver >= SMU_VER_55_51_0)
-			debug->min_disp_clk_khz = 0;
-
-		/* TODO: Check we get what we expect during bringup */
-		clk_mgr->base.dentist_vco_freq_khz = get_vco_frequency_from_reg(clk_mgr);
-
-		/* in case we don't get a value from the register, use default */
-		if (clk_mgr->base.dentist_vco_freq_khz == 0)
-			clk_mgr->base.dentist_vco_freq_khz = 3600000;
-
-		if (ctx->dc_bios->integrated_info->memory_type == LpDdr4MemType) {
-			if (clk_mgr->periodic_retraining_disabled) {
-				rn_bw_params.wm_table = lpddr4_wm_table_with_disabled_ppt;
-			} else {
-				if (is_green_sardine)
-					rn_bw_params.wm_table = lpddr4_wm_table_gs;
-				else
-					rn_bw_params.wm_table = lpddr4_wm_table_rn;
-			}
+
+	if (ctx->dc_bios->integrated_info->memory_type == LpDdr4MemType) {
+		if (clk_mgr->periodic_retraining_disabled) {
+			rn_bw_params.wm_table = lpddr4_wm_table_with_disabled_ppt;
 		} else {
 			if (is_green_sardine)
-				rn_bw_params.wm_table = ddr4_wm_table_gs;
-			else {
-				if (ctx->dc->config.is_single_rank_dimm)
-					rn_bw_params.wm_table = ddr4_1R_wm_table_rn;
-				else
-					rn_bw_params.wm_table = ddr4_wm_table_rn;
-			}
+				rn_bw_params.wm_table = lpddr4_wm_table_gs;
+			else
+				rn_bw_params.wm_table = lpddr4_wm_table_rn;
+		}
+	} else {
+		if (is_green_sardine)
+			rn_bw_params.wm_table = ddr4_wm_table_gs;
+		else {
+			if (ctx->dc->config.is_single_rank_dimm)
+				rn_bw_params.wm_table = ddr4_1R_wm_table_rn;
+			else
+				rn_bw_params.wm_table = ddr4_wm_table_rn;
 		}
-		/* Saved clocks configured at boot for debug purposes */
-		rn_dump_clk_registers(&clk_mgr->base.boot_snapshot, &clk_mgr->base, &log_info);
 	}
+	/* Saved clocks configured at boot for debug purposes */
+	rn_dump_clk_registers(&clk_mgr->base.boot_snapshot, &clk_mgr->base, &log_info);
 
 	clk_mgr->base.dprefclk_khz = 600000;
 	dce_clock_read_ss_info(clk_mgr);
@@ -778,7 +772,7 @@ void rn_clk_mgr_construct(
 		status = pp_smu->rn_funcs.get_dpm_clock_table(&pp_smu->rn_funcs.pp_smu, &clock_table);
 
 		if (status == PP_SMU_RESULT_OK &&
-		    ctx->dc_bios && ctx->dc_bios->integrated_info) {
+		    ctx->dc_bios->integrated_info) {
 			rn_clk_mgr_helper_populate_bw_params (clk_mgr->base.bw_params, &clock_table, ctx->dc_bios->integrated_info);
 			/* treat memory config as single channel if memory is asymmetrics. */
 			if (ctx->dc->config.is_asymmetric_memory)
@@ -786,9 +780,8 @@ void rn_clk_mgr_construct(
 		}
 	}
 
-	if (!IS_FPGA_MAXIMUS_DC(ctx->dce_environment) && clk_mgr->smu_ver >= 0x00371500) {
-		/* enable powerfeatures when displaycount goes to 0 */
+	/* enable powerfeatures when displaycount goes to 0 */
+	if (clk_mgr->smu_ver >= 0x00371500)
 		rn_vbios_smu_enable_48mhz_tmdp_refclk_pwrdwn(clk_mgr, !debug->disable_48mhz_pwrdwn);
-	}
 }
 
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn21/rn_clk_mgr_vbios_smu.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn21/rn_clk_mgr_vbios_smu.c
index 27fbe906682f..5a633333dbb5 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn21/rn_clk_mgr_vbios_smu.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn21/rn_clk_mgr_vbios_smu.c
@@ -26,6 +26,10 @@
 #include "core_types.h"
 #include "clk_mgr_internal.h"
 #include "reg_helper.h"
+#include "dm_helpers.h"
+
+#include "rn_clk_mgr_vbios_smu.h"
+
 #include <linux/delay.h>
 
 #include "renoir_ip_offset.h"
@@ -33,8 +37,6 @@
 #include "mp/mp_12_0_0_offset.h"
 #include "mp/mp_12_0_0_sh_mask.h"
 
-#include "rn_clk_mgr_vbios_smu.h"
-
 #define REG(reg_name) \
 	(MP0_BASE.instance[0].segment[mm ## reg_name ## _BASE_IDX] + mm ## reg_name)
 
@@ -120,7 +122,10 @@ static int rn_vbios_smu_send_msg_with_param(struct clk_mgr_internal *clk_mgr,
 
 	result = rn_smu_wait_for_response(clk_mgr, 10, 200000);
 
-	ASSERT(result == VBIOSSMC_Result_OK || result == VBIOSSMC_Result_UnknownCmd);
+	if (IS_SMU_TIMEOUT(result)) {
+		ASSERT(0);
+		dm_helpers_smu_timeout(CTX, msg_id, param, 10 * 200000);
+	}
 
 	/* Actual dispclk set is returned in the parameter register */
 	return REG_READ(MP1_SMN_C2PMSG_83);
@@ -147,35 +152,18 @@ int rn_vbios_smu_set_dispclk(struct clk_mgr_internal *clk_mgr, int requested_dis
 			VBIOSSMC_MSG_SetDispclkFreq,
 			khz_to_mhz_ceil(requested_dispclk_khz));
 
-	if (!IS_FPGA_MAXIMUS_DC(dc->ctx->dce_environment)) {
-		if (dmcu && dmcu->funcs->is_dmcu_initialized(dmcu)) {
-			if (clk_mgr->dfs_bypass_disp_clk != actual_dispclk_set_mhz)
-				dmcu->funcs->set_psr_wait_loop(dmcu,
-						actual_dispclk_set_mhz / 7);
-		}
+	if (dmcu && dmcu->funcs->is_dmcu_initialized(dmcu)) {
+		if (clk_mgr->dfs_bypass_disp_clk != actual_dispclk_set_mhz)
+			dmcu->funcs->set_psr_wait_loop(dmcu,
+					actual_dispclk_set_mhz / 7);
 	}
 
 	// pmfw always set clock more than or equal requested clock
-	if (!IS_DIAG_DC(dc->ctx->dce_environment))
-		ASSERT(actual_dispclk_set_mhz >= khz_to_mhz_ceil(requested_dispclk_khz));
+	ASSERT(actual_dispclk_set_mhz >= khz_to_mhz_ceil(requested_dispclk_khz));
 
 	return actual_dispclk_set_mhz * 1000;
 }
 
-int rn_vbios_smu_set_dprefclk(struct clk_mgr_internal *clk_mgr)
-{
-	int actual_dprefclk_set_mhz = -1;
-
-	actual_dprefclk_set_mhz = rn_vbios_smu_send_msg_with_param(
-			clk_mgr,
-			VBIOSSMC_MSG_SetDprefclkFreq,
-			khz_to_mhz_ceil(clk_mgr->base.dprefclk_khz));
-
-	/* TODO: add code for programing DP DTO, currently this is down by command table */
-
-	return actual_dprefclk_set_mhz * 1000;
-}
-
 int rn_vbios_smu_set_hard_min_dcfclk(struct clk_mgr_internal *clk_mgr, int requested_dcfclk_khz)
 {
 	int actual_dcfclk_set_mhz = -1;
@@ -188,10 +176,6 @@ int rn_vbios_smu_set_hard_min_dcfclk(struct clk_mgr_internal *clk_mgr, int reque
 			VBIOSSMC_MSG_SetHardMinDcfclkByFreq,
 			khz_to_mhz_ceil(requested_dcfclk_khz));
 
-#ifdef DBG
-	smu_print("actual_dcfclk_set_mhz %d is set to : %d\n", actual_dcfclk_set_mhz, actual_dcfclk_set_mhz * 1000);
-#endif
-
 	return actual_dcfclk_set_mhz * 1000;
 }
 
@@ -221,15 +205,13 @@ void rn_vbios_smu_set_phyclk(struct clk_mgr_internal *clk_mgr, int requested_phy
 int rn_vbios_smu_set_dppclk(struct clk_mgr_internal *clk_mgr, int requested_dpp_khz)
 {
 	int actual_dppclk_set_mhz = -1;
-	struct dc *dc = clk_mgr->base.ctx->dc;
 
 	actual_dppclk_set_mhz = rn_vbios_smu_send_msg_with_param(
 			clk_mgr,
 			VBIOSSMC_MSG_SetDppclkFreq,
 			khz_to_mhz_ceil(requested_dpp_khz));
 
-	if (!IS_DIAG_DC(dc->ctx->dce_environment))
-		ASSERT(actual_dppclk_set_mhz >= khz_to_mhz_ceil(requested_dpp_khz));
+	ASSERT(actual_dppclk_set_mhz >= khz_to_mhz_ceil(requested_dpp_khz));
 
 	return actual_dppclk_set_mhz * 1000;
 }
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn21/rn_clk_mgr_vbios_smu.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn21/rn_clk_mgr_vbios_smu.h
index 1ce19d875358..f76fad87f0e1 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn21/rn_clk_mgr_vbios_smu.h
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn21/rn_clk_mgr_vbios_smu.h
@@ -30,7 +30,6 @@ enum dcn_pwr_state;
 
 int rn_vbios_smu_get_smu_version(struct clk_mgr_internal *clk_mgr);
 int rn_vbios_smu_set_dispclk(struct clk_mgr_internal *clk_mgr, int requested_dispclk_khz);
-int rn_vbios_smu_set_dprefclk(struct clk_mgr_internal *clk_mgr);
 int rn_vbios_smu_set_hard_min_dcfclk(struct clk_mgr_internal *clk_mgr, int requested_dcfclk_khz);
 int rn_vbios_smu_set_min_deep_sleep_dcfclk(struct clk_mgr_internal *clk_mgr, int requested_min_ds_dcfclk_khz);
 void rn_vbios_smu_set_phyclk(struct clk_mgr_internal *clk_mgr, int requested_phyclk_khz);
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dalsmc.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dalsmc.h
index fa09c594fd36..06da34676965 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dalsmc.h
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dalsmc.h
@@ -56,6 +56,7 @@
 #define DALSMC_MSG_SetDisplayRefreshFromMall      0xF
 #define DALSMC_MSG_SetExternalClientDfCstateAllow 0x10
 #define DALSMC_MSG_BacoAudioD3PME                 0x11
-#define DALSMC_Message_Count                      0x12
+#define DALSMC_MSG_SmartAccess                    0x12
+#define DALSMC_Message_Count                      0x13
 
 #endif
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30_clk_mgr.c
index 694a9d3d92ae..ef77fcd164ed 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30_clk_mgr.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30_clk_mgr.c
@@ -30,6 +30,7 @@
 #include "dce100/dce_clk_mgr.h"
 #include "dcn30/dcn30_clk_mgr.h"
 #include "dml/dcn30/dcn30_fpu.h"
+#include "dcn30/dcn30m_clk_mgr.h"
 #include "reg_helper.h"
 #include "core_types.h"
 #include "dm_helpers.h"
@@ -206,7 +207,6 @@ static void dcn3_update_clocks(struct clk_mgr *clk_mgr_base,
 	bool force_reset = false;
 	bool update_uclk = false;
 	bool p_state_change_support;
-	int total_plane_count;
 
 	if (dc->work_arounds.skip_clock_update || !clk_mgr->smu_present)
 		return;
@@ -247,8 +247,7 @@ static void dcn3_update_clocks(struct clk_mgr *clk_mgr_base,
 		clk_mgr_base->clks.socclk_khz = new_clocks->socclk_khz;
 
 	clk_mgr_base->clks.prev_p_state_change_support = clk_mgr_base->clks.p_state_change_support;
-	total_plane_count = clk_mgr_helper_get_active_plane_cnt(dc, context);
-	p_state_change_support = new_clocks->p_state_change_support || (total_plane_count == 0);
+	p_state_change_support = new_clocks->p_state_change_support;
 
 	// invalidate the current P-State forced min in certain dc_mode_softmax situations
 	if (dc->clk_mgr->dc_mode_softmax_enabled && safe_to_lower && !p_state_change_support) {
@@ -476,7 +475,7 @@ static void dcn30_notify_link_rate_change(struct clk_mgr *clk_mgr_base, struct d
 
 	clk_mgr->cur_phyclk_req_table[link->link_index] = link->cur_link_settings.link_rate * LINK_RATE_REF_FREQ_IN_KHZ;
 
-	for (i = 0; i < MAX_PIPES * 2; i++) {
+	for (i = 0; i < MAX_LINKS; i++) {
 		if (clk_mgr->cur_phyclk_req_table[i] > max_phyclk_req)
 			max_phyclk_req = clk_mgr->cur_phyclk_req_table[i];
 	}
@@ -500,7 +499,8 @@ static struct clk_mgr_funcs dcn3_funcs = {
 		.are_clock_states_equal = dcn3_are_clock_states_equal,
 		.enable_pme_wa = dcn3_enable_pme_wa,
 		.notify_link_rate_change = dcn30_notify_link_rate_change,
-		.is_smu_present = dcn3_is_smu_present
+		.is_smu_present = dcn3_is_smu_present,
+		.set_smartmux_switch = dcn30m_set_smartmux_switch
 };
 
 static void dcn3_init_clocks_fpga(struct clk_mgr *clk_mgr)
@@ -523,6 +523,8 @@ void dcn3_clk_mgr_construct(
 		struct pp_smu_funcs *pp_smu,
 		struct dccg *dccg)
 {
+	struct clk_state_registers_and_bypass s = { 0 };
+
 	clk_mgr->base.ctx = ctx;
 	clk_mgr->base.funcs = &dcn3_funcs;
 	clk_mgr->regs = &clk_mgr_regs;
@@ -539,27 +541,19 @@ void dcn3_clk_mgr_construct(
 
 	clk_mgr->base.dprefclk_khz = 730000; // 700 MHz planned if VCO is 3.85 GHz, will be retrieved
 
-	if (IS_FPGA_MAXIMUS_DC(ctx->dce_environment)) {
-		clk_mgr->base.funcs  = &dcn3_fpga_funcs;
-		clk_mgr->base.dentist_vco_freq_khz = 3650000;
-
-	} else {
-		struct clk_state_registers_and_bypass s = { 0 };
+	/* integer part is now VCO frequency in kHz */
+	clk_mgr->base.dentist_vco_freq_khz = dcn30_get_vco_frequency_from_reg(clk_mgr);
 
-		/* integer part is now VCO frequency in kHz */
-		clk_mgr->base.dentist_vco_freq_khz = dcn30_get_vco_frequency_from_reg(clk_mgr);
+	/* in case we don't get a value from the register, use default */
+	if (clk_mgr->base.dentist_vco_freq_khz == 0)
+		clk_mgr->base.dentist_vco_freq_khz = 3650000;
+	/* Convert dprefclk units from MHz to KHz */
+	/* Value already divided by 10, some resolution lost */
 
-		/* in case we don't get a value from the register, use default */
-		if (clk_mgr->base.dentist_vco_freq_khz == 0)
-			clk_mgr->base.dentist_vco_freq_khz = 3650000;
-		/* Convert dprefclk units from MHz to KHz */
-		/* Value already divided by 10, some resolution lost */
-
-		/*TODO: uncomment assert once dcn3_dump_clk_registers is implemented */
-		//ASSERT(s.dprefclk != 0);
-		if (s.dprefclk != 0)
-			clk_mgr->base.dprefclk_khz = s.dprefclk * 1000;
-	}
+	/*TODO: uncomment assert once dcn3_dump_clk_registers is implemented */
+	//ASSERT(s.dprefclk != 0);
+	if (s.dprefclk != 0)
+		clk_mgr->base.dprefclk_khz = s.dprefclk * 1000;
 
 	clk_mgr->dfs_bypass_enabled = false;
 
@@ -568,11 +562,19 @@ void dcn3_clk_mgr_construct(
 	dce_clock_read_ss_info(clk_mgr);
 
 	clk_mgr->base.bw_params = kzalloc(sizeof(*clk_mgr->base.bw_params), GFP_KERNEL);
+	if (!clk_mgr->base.bw_params) {
+		BREAK_TO_DEBUGGER();
+		return;
+	}
 
 	/* need physical address of table to give to PMFW */
 	clk_mgr->wm_range_table = dm_helpers_allocate_gpu_mem(clk_mgr->base.ctx,
 			DC_MEM_ALLOC_TYPE_GART, sizeof(WatermarksExternal_t),
 			&clk_mgr->wm_range_table_addr);
+	if (!clk_mgr->wm_range_table) {
+		BREAK_TO_DEBUGGER();
+		return;
+	}
 }
 
 void dcn3_clk_mgr_destroy(struct clk_mgr_internal *clk_mgr)
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30_clk_mgr_smu_msg.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30_clk_mgr_smu_msg.c
index 1fbf1c105dc1..827bc2431d5d 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30_clk_mgr_smu_msg.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30_clk_mgr_smu_msg.c
@@ -23,7 +23,6 @@
  *
  */
 
-#include <linux/delay.h>
 #include "dcn30_clk_mgr_smu_msg.h"
 
 #include "clk_mgr_internal.h"
@@ -54,6 +53,7 @@
  */
 static uint32_t dcn30_smu_wait_for_response(struct clk_mgr_internal *clk_mgr, unsigned int delay_us, unsigned int max_retries)
 {
+	const uint32_t initial_max_retries = max_retries;
 	uint32_t reg = 0;
 
 	do {
@@ -69,7 +69,7 @@ static uint32_t dcn30_smu_wait_for_response(struct clk_mgr_internal *clk_mgr, un
 
 	/* handle DALSMC_Result_CmdRejectedBusy? */
 
-	/* Log? */
+	TRACE_SMU_MSG_DELAY(0, 0, delay_us * (initial_max_retries - max_retries), clk_mgr->base.ctx);
 
 	return reg;
 }
@@ -89,6 +89,8 @@ static bool dcn30_smu_send_msg_with_param(struct clk_mgr_internal *clk_mgr, uint
 	/* Trigger the message transaction by writing the message ID */
 	REG_WRITE(DAL_MSG_REG, msg_id);
 
+	TRACE_SMU_MSG(msg_id, param_in, clk_mgr->base.ctx);
+
 	result = dcn30_smu_wait_for_response(clk_mgr, 10, 200000);
 
 	if (IS_SMU_TIMEOUT(result)) {
@@ -312,6 +314,9 @@ void dcn30_smu_set_display_refresh_from_mall(struct clk_mgr_internal *clk_mgr, b
 	/* bits 8:7 for cache timer scale, bits 6:1 for cache timer delay, bit 0 = 1 for enable, = 0 for disable */
 	uint32_t param = (cache_timer_scale << 7) | (cache_timer_delay << 1) | (enable ? 1 : 0);
 
+	smu_print("SMU Set display refresh from mall: enable = %d, cache_timer_delay = %d, cache_timer_scale = %d\n",
+		enable, cache_timer_delay, cache_timer_scale);
+
 	dcn30_smu_send_msg_with_param(clk_mgr,
 			DALSMC_MSG_SetDisplayRefreshFromMall, param, NULL);
 }
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30_smu11_driver_if.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30_smu11_driver_if.h
index 61bb1d86182e..1bfd6f66f035 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30_smu11_driver_if.h
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30_smu11_driver_if.h
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// This is a stripped-down version of the smu11_driver_if.h file for the relevant DAL interfaces.
+/* Copyright © 2022-2024 Advanced Micro Devices, Inc. All rights reserved. */
 
 #define SMU11_DRIVER_IF_VERSION 0x40
 
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30m_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30m_clk_mgr.c
new file mode 100644
index 000000000000..8e8a11c7437e
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30m_clk_mgr.c
@@ -0,0 +1,36 @@
+/*
+ * Copyright 2019 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: AMD
+ *
+ */
+
+#include "clk_mgr_internal.h"
+#include "dcn30/dcn30m_clk_mgr.h"
+#include "dcn30m_clk_mgr_smu_msg.h"
+
+
+uint32_t dcn30m_set_smartmux_switch(struct clk_mgr *clk_mgr_base, uint32_t pins_to_set)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+
+	return dcn30m_smu_set_smart_mux_switch(clk_mgr, pins_to_set);
+}
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30m_clk_mgr.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30m_clk_mgr.h
new file mode 100644
index 000000000000..757985b2eadc
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30m_clk_mgr.h
@@ -0,0 +1,31 @@
+/*
+ * Copyright 2019 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: AMD
+ *
+ */
+
+#ifndef __DCN30M_CLK_MGR_H__
+#define __DCN30M_CLK_MGR_H__
+
+uint32_t dcn30m_set_smartmux_switch(struct clk_mgr *clk_mgr_base, uint32_t pins_to_set);
+
+#endif //__DCN30M_CLK_MGR_H__
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30m_clk_mgr_smu_msg.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30m_clk_mgr_smu_msg.c
new file mode 100644
index 000000000000..0dd0583ff21e
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30m_clk_mgr_smu_msg.c
@@ -0,0 +1,118 @@
+/*
+ * Copyright 2020 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: AMD
+ *
+ */
+
+#include "dcn30m_clk_mgr_smu_msg.h"
+
+#include "clk_mgr_internal.h"
+#include "reg_helper.h"
+#include "dm_helpers.h"
+
+#include "dalsmc.h"
+
+#define mmDAL_MSG_REG  0x1628A
+#define mmDAL_ARG_REG  0x16273
+#define mmDAL_RESP_REG 0x16274
+
+#define REG(reg_name) \
+	mm ## reg_name
+
+#include "logger_types.h"
+#undef DC_LOGGER
+#define DC_LOGGER \
+	CTX->logger
+#define smu_print(str, ...) {DC_LOG_SMU(str, ##__VA_ARGS__); }
+
+
+/*
+ * Function to be used instead of REG_WAIT macro because the wait ends when
+ * the register is NOT EQUAL to zero, and because the translation in msg_if.h
+ * won't work with REG_WAIT.
+ */
+static uint32_t dcn30m_smu_wait_for_response(struct clk_mgr_internal *clk_mgr,
+	unsigned int delay_us, unsigned int max_retries)
+{
+	uint32_t reg = 0;
+
+	do {
+		reg = REG_READ(DAL_RESP_REG);
+		if (reg)
+			break;
+
+		if (delay_us >= 1000)
+			msleep(delay_us/1000);
+		else if (delay_us > 0)
+			udelay(delay_us);
+	} while (max_retries--);
+
+	/* handle DALSMC_Result_CmdRejectedBusy? */
+
+	/* Log? */
+
+	return reg;
+}
+
+static bool dcn30m_smu_send_msg_with_param(struct clk_mgr_internal *clk_mgr,
+	uint32_t msg_id, uint32_t param_in, uint32_t *param_out)
+{
+	uint32_t result;
+	/* Wait for response register to be ready */
+	dcn30m_smu_wait_for_response(clk_mgr, 10, 200000);
+
+	/* Clear response register */
+	REG_WRITE(DAL_RESP_REG, 0);
+
+	/* Set the parameter register for the SMU message */
+	REG_WRITE(DAL_ARG_REG, param_in);
+
+	/* Trigger the message transaction by writing the message ID */
+	REG_WRITE(DAL_MSG_REG, msg_id);
+
+	result = dcn30m_smu_wait_for_response(clk_mgr, 10, 200000);
+
+	if (IS_SMU_TIMEOUT(result))
+		dm_helpers_smu_timeout(CTX, msg_id, param_in, 10 * 200000);
+
+	/* Wait for response */
+	if (result == DALSMC_Result_OK) {
+		if (param_out)
+			*param_out = REG_READ(DAL_ARG_REG);
+
+		return true;
+	}
+
+	return false;
+}
+
+uint32_t dcn30m_smu_set_smart_mux_switch(struct clk_mgr_internal *clk_mgr, uint32_t pins_to_set)
+{
+	uint32_t response = 0;
+
+	smu_print("SMU Set SmartMux Switch: switch_dgpu = %d\n", pins_to_set);
+
+	dcn30m_smu_send_msg_with_param(clk_mgr,
+			DALSMC_MSG_SmartAccess, pins_to_set, &response);
+
+	return response;
+}
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30m_clk_mgr_smu_msg.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30m_clk_mgr_smu_msg.h
new file mode 100644
index 000000000000..8a59a473fc5e
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30m_clk_mgr_smu_msg.h
@@ -0,0 +1,34 @@
+/*
+ * Copyright 2020 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: AMD
+ *
+ */
+
+#ifndef DAL_DC_DCN30M_CLK_MGR_SMU_MSG_H_
+#define DAL_DC_DCN30M_CLK_MGR_SMU_MSG_H_
+
+#include "core_types.h"
+
+struct clk_mgr_internal;
+
+uint32_t     dcn30m_smu_set_smart_mux_switch(struct clk_mgr_internal *clk_mgr, uint32_t pins_to_set);
+#endif /* DAL_DC_DCN30M_CLK_MGR_SMU_MSG_H_ */
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn301/dcn301_smu.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn301/dcn301_smu.c
index e4f96b6fd79d..b4fb17b7a096 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn301/dcn301_smu.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn301/dcn301_smu.c
@@ -29,6 +29,7 @@
 #include <linux/delay.h>
 
 #include "dcn301_smu.h"
+#include "dm_helpers.h"
 
 #include "vangogh_ip_offset.h"
 
@@ -120,7 +121,10 @@ static int dcn301_smu_send_msg_with_param(struct clk_mgr_internal *clk_mgr,
 
 	result = dcn301_smu_wait_for_response(clk_mgr, 10, 200000);
 
-	ASSERT(result == VBIOSSMC_Result_OK);
+	if (IS_SMU_TIMEOUT(result)) {
+		ASSERT(0);
+		dm_helpers_smu_timeout(CTX, msg_id, param, 10 * 200000);
+	}
 
 	/* Actual dispclk set is returned in the parameter register */
 	return REG_READ(MP1_SMN_C2PMSG_83);
@@ -180,10 +184,6 @@ int dcn301_smu_set_hard_min_dcfclk(struct clk_mgr_internal *clk_mgr, int request
 			VBIOSSMC_MSG_SetHardMinDcfclkByFreq,
 			khz_to_mhz_ceil(requested_dcfclk_khz));
 
-#ifdef DBG
-	smu_print("actual_dcfclk_set_mhz %d is set to : %d\n", actual_dcfclk_set_mhz, actual_dcfclk_set_mhz * 1000);
-#endif
-
 	return actual_dcfclk_set_mhz * 1000;
 }
 
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn301/vg_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn301/vg_clk_mgr.c
index 01383aac6b41..7aee02d56292 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn301/vg_clk_mgr.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn301/vg_clk_mgr.c
@@ -117,7 +117,7 @@ static void vg_update_clocks(struct clk_mgr *clk_mgr_base,
 
 			display_count = vg_get_active_display_cnt_wa(dc, context);
 			/* if we can go lower, go lower */
-			if (display_count == 0 && !IS_DIAG_DC(dc->ctx->dce_environment)) {
+			if (display_count == 0) {
 				union display_idle_optimization_u idle_info = { 0 };
 
 				idle_info.idle_info.df_request_disabled = 1;
@@ -151,10 +151,8 @@ static void vg_update_clocks(struct clk_mgr *clk_mgr_base,
 	}
 
 	// workaround: Limit dppclk to 100Mhz to avoid lower eDP panel switch to plus 4K monitor underflow.
-	if (!IS_DIAG_DC(dc->ctx->dce_environment)) {
-		if (new_clocks->dppclk_khz < 100000)
-			new_clocks->dppclk_khz = 100000;
-	}
+	if (new_clocks->dppclk_khz < 100000)
+		new_clocks->dppclk_khz = 100000;
 
 	if (should_set_clock(safe_to_lower, new_clocks->dppclk_khz, clk_mgr->base.clks.dppclk_khz)) {
 		if (clk_mgr->base.clks.dppclk_khz > new_clocks->dppclk_khz)
@@ -254,16 +252,16 @@ static void vg_dump_clk_registers(struct clk_state_registers_and_bypass *regs_an
 	regs_and_bypass->dppclk = internal.CLK1_CLK1_CURRENT_CNT / 10;
 
 	regs_and_bypass->dppclk_bypass = internal.CLK1_CLK1_BYPASS_CNTL & 0x0007;
-	if (regs_and_bypass->dppclk_bypass < 0 || regs_and_bypass->dppclk_bypass > 4)
+	if (regs_and_bypass->dppclk_bypass > 4)
 		regs_and_bypass->dppclk_bypass = 0;
 	regs_and_bypass->dcfclk_bypass = internal.CLK1_CLK3_BYPASS_CNTL & 0x0007;
-	if (regs_and_bypass->dcfclk_bypass < 0 || regs_and_bypass->dcfclk_bypass > 4)
+	if (regs_and_bypass->dcfclk_bypass > 4)
 		regs_and_bypass->dcfclk_bypass = 0;
 	regs_and_bypass->dispclk_bypass = internal.CLK1_CLK0_BYPASS_CNTL & 0x0007;
-	if (regs_and_bypass->dispclk_bypass < 0 || regs_and_bypass->dispclk_bypass > 4)
+	if (regs_and_bypass->dispclk_bypass > 4)
 		regs_and_bypass->dispclk_bypass = 0;
 	regs_and_bypass->dprefclk_bypass = internal.CLK1_CLK2_BYPASS_CNTL & 0x0007;
-	if (regs_and_bypass->dprefclk_bypass < 0 || regs_and_bypass->dprefclk_bypass > 4)
+	if (regs_and_bypass->dprefclk_bypass > 4)
 		regs_and_bypass->dprefclk_bypass = 0;
 
 	if (log_info->enabled) {
@@ -548,6 +546,8 @@ static unsigned int find_dcfclk_for_voltage(const struct vg_dpm_clocks *clock_ta
 	int i;
 
 	for (i = 0; i < VG_NUM_SOC_VOLTAGE_LEVELS; i++) {
+		if (i >= VG_NUM_DCFCLK_DPM_LEVELS)
+			break;
 		if (clock_table->SocVoltage[i] == voltage)
 			return clock_table->DcfClocks[i];
 	}
@@ -563,10 +563,12 @@ static void vg_clk_mgr_helper_populate_bw_params(
 {
 	int i, j;
 	struct clk_bw_params *bw_params = clk_mgr->base.bw_params;
+	uint32_t max_dispclk = 0, max_dppclk = 0;
 
 	j = -1;
 
-	ASSERT(VG_NUM_FCLK_DPM_LEVELS <= MAX_NUM_DPM_LVL);
+	static_assert(VG_NUM_FCLK_DPM_LEVELS <= MAX_NUM_DPM_LVL,
+		"number of reported FCLK DPM levels exceeds maximum");
 
 	/* Find lowest DPM, FCLK is filled in reverse order*/
 
@@ -583,6 +585,15 @@ static void vg_clk_mgr_helper_populate_bw_params(
 		return;
 	}
 
+	/* dispclk and dppclk can be max at any voltage, same number of levels for both */
+	if (clock_table->NumDispClkLevelsEnabled <= VG_NUM_DISPCLK_DPM_LEVELS &&
+	    clock_table->NumDispClkLevelsEnabled <= VG_NUM_DPPCLK_DPM_LEVELS) {
+		max_dispclk = find_max_clk_value(clock_table->DispClocks, clock_table->NumDispClkLevelsEnabled);
+		max_dppclk = find_max_clk_value(clock_table->DppClocks, clock_table->NumDispClkLevelsEnabled);
+	} else {
+		ASSERT(0);
+	}
+
 	bw_params->clk_table.num_entries = j + 1;
 
 	for (i = 0; i < bw_params->clk_table.num_entries - 1; i++, j--) {
@@ -590,11 +601,17 @@ static void vg_clk_mgr_helper_populate_bw_params(
 		bw_params->clk_table.entries[i].memclk_mhz = clock_table->DfPstateTable[j].memclk;
 		bw_params->clk_table.entries[i].voltage = clock_table->DfPstateTable[j].voltage;
 		bw_params->clk_table.entries[i].dcfclk_mhz = find_dcfclk_for_voltage(clock_table, clock_table->DfPstateTable[j].voltage);
+
+		/* Now update clocks we do read */
+		bw_params->clk_table.entries[i].dispclk_mhz = max_dispclk;
+		bw_params->clk_table.entries[i].dppclk_mhz = max_dppclk;
 	}
 	bw_params->clk_table.entries[i].fclk_mhz = clock_table->DfPstateTable[j].fclk;
 	bw_params->clk_table.entries[i].memclk_mhz = clock_table->DfPstateTable[j].memclk;
 	bw_params->clk_table.entries[i].voltage = clock_table->DfPstateTable[j].voltage;
 	bw_params->clk_table.entries[i].dcfclk_mhz = find_max_clk_value(clock_table->DcfClocks, VG_NUM_DCFCLK_DPM_LEVELS);
+	bw_params->clk_table.entries[i].dispclk_mhz = find_max_clk_value(clock_table->DispClocks, VG_NUM_DISPCLK_DPM_LEVELS);
+	bw_params->clk_table.entries[i].dppclk_mhz = find_max_clk_value(clock_table->DppClocks, VG_NUM_DPPCLK_DPM_LEVELS);
 
 	bw_params->vram_type = bios_info->memory_type;
 	bw_params->num_channels = bios_info->ma_channel_number;
@@ -664,6 +681,7 @@ void vg_clk_mgr_construct(
 		struct dccg *dccg)
 {
 	struct smu_dpm_clks smu_dpm_clks = { 0 };
+	struct clk_log_info log_info = {0};
 
 	clk_mgr->base.base.ctx = ctx;
 	clk_mgr->base.base.funcs = &vg_funcs;
@@ -703,32 +721,25 @@ void vg_clk_mgr_construct(
 
 	ASSERT(smu_dpm_clks.dpm_clks);
 
-	if (IS_FPGA_MAXIMUS_DC(ctx->dce_environment)) {
-		vg_funcs.update_clocks = dcn2_update_clocks_fpga;
-		clk_mgr->base.base.dentist_vco_freq_khz = 3600000;
-	} else {
-		struct clk_log_info log_info = {0};
-
-		clk_mgr->base.smu_ver = dcn301_smu_get_smu_version(&clk_mgr->base);
+	clk_mgr->base.smu_ver = dcn301_smu_get_smu_version(&clk_mgr->base);
 
-		if (clk_mgr->base.smu_ver)
-			clk_mgr->base.smu_present = true;
+	if (clk_mgr->base.smu_ver)
+		clk_mgr->base.smu_present = true;
 
-		/* TODO: Check we get what we expect during bringup */
-		clk_mgr->base.base.dentist_vco_freq_khz = get_vco_frequency_from_reg(&clk_mgr->base);
+	/* TODO: Check we get what we expect during bringup */
+	clk_mgr->base.base.dentist_vco_freq_khz = get_vco_frequency_from_reg(&clk_mgr->base);
 
-		/* in case we don't get a value from the register, use default */
-		if (clk_mgr->base.base.dentist_vco_freq_khz == 0)
-			clk_mgr->base.base.dentist_vco_freq_khz = 3600000;
+	/* in case we don't get a value from the register, use default */
+	if (clk_mgr->base.base.dentist_vco_freq_khz == 0)
+		clk_mgr->base.base.dentist_vco_freq_khz = 3600000;
 
-		if (ctx->dc_bios->integrated_info->memory_type == LpDdr5MemType) {
-			vg_bw_params.wm_table = lpddr5_wm_table;
-		} else {
-			vg_bw_params.wm_table = ddr4_wm_table;
-		}
-		/* Saved clocks configured at boot for debug purposes */
-		vg_dump_clk_registers(&clk_mgr->base.base.boot_snapshot, &clk_mgr->base.base, &log_info);
+	if (ctx->dc_bios->integrated_info->memory_type == LpDdr5MemType) {
+		vg_bw_params.wm_table = lpddr5_wm_table;
+	} else {
+		vg_bw_params.wm_table = ddr4_wm_table;
 	}
+	/* Saved clocks configured at boot for debug purposes */
+	vg_dump_clk_registers(&clk_mgr->base.base.boot_snapshot, &clk_mgr->base.base, &log_info);
 
 	clk_mgr->base.base.dprefclk_khz = 600000;
 	dce_clock_read_ss_info(&clk_mgr->base);
@@ -736,7 +747,7 @@ void vg_clk_mgr_construct(
 	clk_mgr->base.base.bw_params = &vg_bw_params;
 
 	vg_get_dpm_table_from_smu(&clk_mgr->base, &smu_dpm_clks);
-	if (ctx->dc_bios && ctx->dc_bios->integrated_info) {
+	if (ctx->dc_bios->integrated_info) {
 		vg_clk_mgr_helper_populate_bw_params(
 				&clk_mgr->base,
 				ctx->dc_bios->integrated_info,
@@ -746,12 +757,6 @@ void vg_clk_mgr_construct(
 	if (smu_dpm_clks.dpm_clks && smu_dpm_clks.mc_address.quad_part != 0)
 		dm_helpers_free_gpu_mem(clk_mgr->base.base.ctx, DC_MEM_ALLOC_TYPE_FRAME_BUFFER,
 				smu_dpm_clks.dpm_clks);
-/*
-	if (!IS_FPGA_MAXIMUS_DC(ctx->dce_environment) && clk_mgr->base.smu_ver) {
-		 enable powerfeatures when displaycount goes to 0
-		dcn301_smu_enable_phy_refclk_pwrdwn(clk_mgr, !debug->disable_48mhz_pwrdwn);
-	}
-*/
 }
 
 void vg_clk_mgr_destroy(struct clk_mgr_internal *clk_mgr_int)
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn31/dcn31_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn31/dcn31_clk_mgr.c
index f9e2e0c3095e..051052bd10c9 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn31/dcn31_clk_mgr.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn31/dcn31_clk_mgr.c
@@ -47,15 +47,16 @@
 #include "dcn30/dcn30_clk_mgr.h"
 
 #include "dc_dmub_srv.h"
-#include "link.h"
+#include "link_service.h"
 
 #include "logger_types.h"
+
+
+#include "yellow_carp_offset.h"
 #undef DC_LOGGER
 #define DC_LOGGER \
 	clk_mgr->base.base.ctx->logger
 
-#include "yellow_carp_offset.h"
-
 #define regCLK1_CLK_PLL_REQ			0x0237
 #define regCLK1_CLK_PLL_REQ_BASE_IDX		0
 
@@ -87,6 +88,11 @@ static int dcn31_get_active_display_cnt_wa(
 				stream->signal == SIGNAL_TYPE_DVI_SINGLE_LINK ||
 				stream->signal == SIGNAL_TYPE_DVI_DUAL_LINK)
 			tmds_present = true;
+
+		/* Checking stream / link detection ensuring that PHY is active*/
+		if (dc_is_dp_signal(stream->signal) && !stream->dpms_off)
+			display_count++;
+
 	}
 
 	for (i = 0; i < dc->link_count; i++) {
@@ -205,10 +211,8 @@ void dcn31_update_clocks(struct clk_mgr *clk_mgr_base,
 	}
 
 	// workaround: Limit dppclk to 100Mhz to avoid lower eDP panel switch to plus 4K monitor underflow.
-	if (!IS_DIAG_DC(dc->ctx->dce_environment)) {
-		if (new_clocks->dppclk_khz < 100000)
-			new_clocks->dppclk_khz = 100000;
-	}
+	if (new_clocks->dppclk_khz < 100000)
+		new_clocks->dppclk_khz = 100000;
 
 	if (should_set_clock(safe_to_lower, new_clocks->dppclk_khz, clk_mgr->base.clks.dppclk_khz)) {
 		if (clk_mgr->base.clks.dppclk_khz > new_clocks->dppclk_khz)
@@ -250,9 +254,7 @@ void dcn31_update_clocks(struct clk_mgr *clk_mgr_base,
 	cmd.notify_clocks.clocks.dispclk_khz = clk_mgr_base->clks.dispclk_khz;
 	cmd.notify_clocks.clocks.dppclk_khz = clk_mgr_base->clks.dppclk_khz;
 
-	dc_dmub_srv_cmd_queue(dc->ctx->dmub_srv, &cmd);
-	dc_dmub_srv_cmd_execute(dc->ctx->dmub_srv);
-	dc_dmub_srv_wait_idle(dc->ctx->dmub_srv);
+	dc_wake_and_execute_dmub_cmd(dc->ctx, &cmd, DM_DMUB_WAIT_TYPE_WAIT);
 }
 
 static int get_vco_frequency_from_reg(struct clk_mgr_internal *clk_mgr)
@@ -561,7 +563,8 @@ static void dcn31_clk_mgr_helper_populate_bw_params(struct clk_mgr_internal *clk
 
 	j = -1;
 
-	ASSERT(NUM_DF_PSTATE_LEVELS <= MAX_NUM_DPM_LVL);
+	static_assert(NUM_DF_PSTATE_LEVELS <= MAX_NUM_DPM_LVL,
+		"number of reported pstate levels exceeds maximum");
 
 	/* Find lowest DPM, FCLK is filled in reverse order*/
 
@@ -674,6 +677,7 @@ void dcn31_clk_mgr_construct(
 		struct dccg *dccg)
 {
 	struct dcn31_smu_dpm_clks smu_dpm_clks = { 0 };
+	struct clk_log_info log_info = {0};
 
 	clk_mgr->base.base.ctx = ctx;
 	clk_mgr->base.base.funcs = &dcn31_funcs;
@@ -713,29 +717,22 @@ void dcn31_clk_mgr_construct(
 
 	ASSERT(smu_dpm_clks.dpm_clks);
 
-	if (IS_FPGA_MAXIMUS_DC(ctx->dce_environment)) {
-		clk_mgr->base.base.funcs = &dcn3_fpga_funcs;
-	} else {
-		struct clk_log_info log_info = {0};
-
-		clk_mgr->base.smu_ver = dcn31_smu_get_smu_version(&clk_mgr->base);
+	clk_mgr->base.smu_ver = dcn31_smu_get_smu_version(&clk_mgr->base);
 
-		if (clk_mgr->base.smu_ver)
-			clk_mgr->base.smu_present = true;
+	if (clk_mgr->base.smu_ver)
+		clk_mgr->base.smu_present = true;
 
-		/* TODO: Check we get what we expect during bringup */
-		clk_mgr->base.base.dentist_vco_freq_khz = get_vco_frequency_from_reg(&clk_mgr->base);
-
-		if (ctx->dc_bios->integrated_info->memory_type == LpDdr5MemType) {
-			dcn31_bw_params.wm_table = lpddr5_wm_table;
-		} else {
-			dcn31_bw_params.wm_table = ddr5_wm_table;
-		}
-		/* Saved clocks configured at boot for debug purposes */
-		dcn31_dump_clk_registers(&clk_mgr->base.base.boot_snapshot,
-					 &clk_mgr->base.base, &log_info);
+	/* TODO: Check we get what we expect during bringup */
+	clk_mgr->base.base.dentist_vco_freq_khz = get_vco_frequency_from_reg(&clk_mgr->base);
 
+	if (ctx->dc_bios->integrated_info->memory_type == LpDdr5MemType) {
+		dcn31_bw_params.wm_table = lpddr5_wm_table;
+	} else {
+		dcn31_bw_params.wm_table = ddr5_wm_table;
 	}
+	/* Saved clocks configured at boot for debug purposes */
+	dcn31_dump_clk_registers(&clk_mgr->base.base.boot_snapshot,
+				 &clk_mgr->base.base, &log_info);
 
 	clk_mgr->base.base.dprefclk_khz = 600000;
 	clk_mgr->base.base.clks.ref_dtbclk_khz = 600000;
@@ -789,7 +786,7 @@ void dcn31_clk_mgr_construct(
 					   i, smu_dpm_clks.dpm_clks->DfPstateTable[i].MemClk,
 					   i, smu_dpm_clks.dpm_clks->DfPstateTable[i].Voltage);
 		}
-		if (ctx->dc_bios && ctx->dc_bios->integrated_info) {
+		if (ctx->dc_bios->integrated_info) {
 			dcn31_clk_mgr_helper_populate_bw_params(
 					&clk_mgr->base,
 					ctx->dc_bios->integrated_info,
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn31/dcn31_smu.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn31/dcn31_smu.c
index 0827c7df2855..f201628e4e98 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn31/dcn31_smu.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn31/dcn31_smu.c
@@ -23,7 +23,6 @@
  *
  */
 
-#include <linux/delay.h>
 #include "core_types.h"
 #include "clk_mgr_internal.h"
 #include "reg_helper.h"
@@ -130,7 +129,7 @@ static int dcn31_smu_send_msg_with_param(struct clk_mgr_internal *clk_mgr,
 	if (result == VBIOSSMC_Result_Failed) {
 		if (msg_id == VBIOSSMC_MSG_TransferTableDram2Smu &&
 		    param == TABLE_WATERMARKS)
-			DC_LOG_WARNING("Watermarks table not configured properly by SMU");
+			DC_LOG_DEBUG("Watermarks table not configured properly by SMU");
 		else
 			ASSERT(0);
 		REG_WRITE(MP1_SMN_C2PMSG_91, VBIOSSMC_Result_OK);
@@ -202,10 +201,6 @@ int dcn31_smu_set_hard_min_dcfclk(struct clk_mgr_internal *clk_mgr, int requeste
 			VBIOSSMC_MSG_SetHardMinDcfclkByFreq,
 			khz_to_mhz_ceil(requested_dcfclk_khz));
 
-#ifdef DBG
-	smu_print("actual_dcfclk_set_mhz %d is set to : %d\n", actual_dcfclk_set_mhz, actual_dcfclk_set_mhz * 1000);
-#endif
-
 	return actual_dcfclk_set_mhz * 1000;
 }
 
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn314/dcn314_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn314/dcn314_clk_mgr.c
index 5cb44f838bde..db687a13174d 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn314/dcn314_clk_mgr.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn314/dcn314_clk_mgr.c
@@ -48,14 +48,11 @@
 #include "dcn31/dcn31_clk_mgr.h"
 
 #include "dc_dmub_srv.h"
-#include "link.h"
+#include "link_service.h"
 #include "dcn314_smu.h"
 
 
 #include "logger_types.h"
-#undef DC_LOGGER
-#define DC_LOGGER \
-	clk_mgr->base.base.ctx->logger
 
 
 #define MAX_INSTANCE                                        7
@@ -77,6 +74,10 @@ static const struct IP_BASE CLK_BASE = { { { { 0x00016C00, 0x02401800, 0, 0, 0,
 					{ { 0x0001B200, 0x0242DC00, 0, 0, 0, 0, 0, 0 } },
 					{ { 0x0001B400, 0x0242E000, 0, 0, 0, 0, 0, 0 } } } };
 
+#undef DC_LOGGER
+#define DC_LOGGER \
+	clk_mgr->base.base.ctx->logger
+
 #define regCLK1_CLK_PLL_REQ			0x0237
 #define regCLK1_CLK_PLL_REQ_BASE_IDX		0
 
@@ -87,6 +88,82 @@ static const struct IP_BASE CLK_BASE = { { { { 0x00016C00, 0x02401800, 0, 0, 0,
 #define CLK1_CLK_PLL_REQ__PllSpineDiv_MASK	0x0000F000L
 #define CLK1_CLK_PLL_REQ__FbMult_frac_MASK	0xFFFF0000L
 
+#define regCLK1_CLK0_DFS_CNTL				0x0269
+#define regCLK1_CLK0_DFS_CNTL_BASE_IDX		0
+#define regCLK1_CLK1_DFS_CNTL				0x026c
+#define regCLK1_CLK1_DFS_CNTL_BASE_IDX		0
+#define regCLK1_CLK2_DFS_CNTL				0x026f
+#define regCLK1_CLK2_DFS_CNTL_BASE_IDX		0
+#define regCLK1_CLK3_DFS_CNTL				0x0272
+#define regCLK1_CLK3_DFS_CNTL_BASE_IDX		0
+#define regCLK1_CLK4_DFS_CNTL				0x0275
+#define regCLK1_CLK4_DFS_CNTL_BASE_IDX		0
+#define regCLK1_CLK5_DFS_CNTL				0x0278
+#define regCLK1_CLK5_DFS_CNTL_BASE_IDX		0
+
+#define regCLK1_CLK0_CURRENT_CNT			0x02fb
+#define regCLK1_CLK0_CURRENT_CNT_BASE_IDX	0
+#define regCLK1_CLK1_CURRENT_CNT			0x02fc
+#define regCLK1_CLK1_CURRENT_CNT_BASE_IDX	0
+#define regCLK1_CLK2_CURRENT_CNT			0x02fd
+#define regCLK1_CLK2_CURRENT_CNT_BASE_IDX	0
+#define regCLK1_CLK3_CURRENT_CNT			0x02fe
+#define regCLK1_CLK3_CURRENT_CNT_BASE_IDX	0
+#define regCLK1_CLK4_CURRENT_CNT			0x02ff
+#define regCLK1_CLK4_CURRENT_CNT_BASE_IDX	0
+#define regCLK1_CLK5_CURRENT_CNT			0x0300
+#define regCLK1_CLK5_CURRENT_CNT_BASE_IDX	0
+
+#define regCLK1_CLK0_BYPASS_CNTL			0x028a
+#define regCLK1_CLK0_BYPASS_CNTL_BASE_IDX	0
+#define regCLK1_CLK1_BYPASS_CNTL			0x0293
+#define regCLK1_CLK1_BYPASS_CNTL_BASE_IDX	0
+#define regCLK1_CLK2_BYPASS_CNTL			0x029c
+#define regCLK1_CLK2_BYPASS_CNTL_BASE_IDX	0
+#define regCLK1_CLK3_BYPASS_CNTL			0x02a5
+#define regCLK1_CLK3_BYPASS_CNTL_BASE_IDX	0
+#define regCLK1_CLK4_BYPASS_CNTL			0x02ae
+#define regCLK1_CLK4_BYPASS_CNTL_BASE_IDX	0
+#define regCLK1_CLK5_BYPASS_CNTL			0x02b7
+#define regCLK1_CLK5_BYPASS_CNTL_BASE_IDX	0
+
+#define regCLK1_CLK0_DS_CNTL				0x0283
+#define regCLK1_CLK0_DS_CNTL_BASE_IDX		0
+#define regCLK1_CLK1_DS_CNTL				0x028c
+#define regCLK1_CLK1_DS_CNTL_BASE_IDX		0
+#define regCLK1_CLK2_DS_CNTL				0x0295
+#define regCLK1_CLK2_DS_CNTL_BASE_IDX		0
+#define regCLK1_CLK3_DS_CNTL				0x029e
+#define regCLK1_CLK3_DS_CNTL_BASE_IDX		0
+#define regCLK1_CLK4_DS_CNTL				0x02a7
+#define regCLK1_CLK4_DS_CNTL_BASE_IDX		0
+#define regCLK1_CLK5_DS_CNTL				0x02b0
+#define regCLK1_CLK5_DS_CNTL_BASE_IDX		0
+
+#define regCLK1_CLK0_ALLOW_DS				0x0284
+#define regCLK1_CLK0_ALLOW_DS_BASE_IDX		0
+#define regCLK1_CLK1_ALLOW_DS				0x028d
+#define regCLK1_CLK1_ALLOW_DS_BASE_IDX		0
+#define regCLK1_CLK2_ALLOW_DS				0x0296
+#define regCLK1_CLK2_ALLOW_DS_BASE_IDX		0
+#define regCLK1_CLK3_ALLOW_DS				0x029f
+#define regCLK1_CLK3_ALLOW_DS_BASE_IDX		0
+#define regCLK1_CLK4_ALLOW_DS				0x02a8
+#define regCLK1_CLK4_ALLOW_DS_BASE_IDX		0
+#define regCLK1_CLK5_ALLOW_DS				0x02b1
+#define regCLK1_CLK5_ALLOW_DS_BASE_IDX		0
+
+#define CLK1_CLK2_BYPASS_CNTL__CLK2_BYPASS_SEL__SHIFT	0x0
+#define CLK1_CLK2_BYPASS_CNTL__CLK2_BYPASS_DIV__SHIFT	0x10
+#define CLK1_CLK2_BYPASS_CNTL__CLK2_BYPASS_SEL_MASK		0x00000007L
+#define CLK1_CLK2_BYPASS_CNTL__CLK2_BYPASS_DIV_MASK		0x000F0000L
+
+#define regCLK6_0_CLK6_spll_field_8				0x464b
+#define regCLK6_0_CLK6_spll_field_8_BASE_IDX	0
+
+#define CLK6_0_CLK6_spll_field_8__spll_ssc_en__SHIFT	0xd
+#define CLK6_0_CLK6_spll_field_8__spll_ssc_en_MASK		0x00002000L
+
 #define REG(reg_name) \
 	(CLK_BASE.instance[0].segment[reg ## reg_name ## _BASE_IDX] + reg ## reg_name)
 
@@ -131,35 +208,68 @@ static int dcn314_get_active_display_cnt_wa(
 	return display_count;
 }
 
-static void dcn314_disable_otg_wa(struct clk_mgr *clk_mgr_base, struct dc_state *context, bool disable)
+static void dcn314_disable_otg_wa(struct clk_mgr *clk_mgr_base, struct dc_state *context,
+				  bool safe_to_lower, bool disable)
 {
 	struct dc *dc = clk_mgr_base->ctx->dc;
 	int i;
 
 	for (i = 0; i < dc->res_pool->pipe_count; ++i) {
-		struct pipe_ctx *pipe = &dc->current_state->res_ctx.pipe_ctx[i];
+		struct pipe_ctx *pipe = safe_to_lower
+			? &context->res_ctx.pipe_ctx[i]
+			: &dc->current_state->res_ctx.pipe_ctx[i];
 
 		if (pipe->top_pipe || pipe->prev_odm_pipe)
 			continue;
 		if (pipe->stream && (pipe->stream->dpms_off || dc_is_virtual_signal(pipe->stream->signal))) {
-			struct stream_encoder *stream_enc = pipe->stream_res.stream_enc;
-
 			if (disable) {
-				if (stream_enc && stream_enc->funcs->disable_fifo)
-					pipe->stream_res.stream_enc->funcs->disable_fifo(stream_enc);
+				if (pipe->stream_res.tg && pipe->stream_res.tg->funcs->immediate_disable_crtc)
+					pipe->stream_res.tg->funcs->immediate_disable_crtc(pipe->stream_res.tg);
 
-				pipe->stream_res.tg->funcs->immediate_disable_crtc(pipe->stream_res.tg);
 				reset_sync_context_for_pipe(dc, context, i);
 			} else {
 				pipe->stream_res.tg->funcs->enable_crtc(pipe->stream_res.tg);
-
-				if (stream_enc && stream_enc->funcs->enable_fifo)
-					pipe->stream_res.stream_enc->funcs->enable_fifo(stream_enc);
 			}
 		}
 	}
 }
 
+bool dcn314_is_spll_ssc_enabled(struct clk_mgr *clk_mgr_base)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+	uint32_t ssc_enable;
+
+	REG_GET(CLK6_0_CLK6_spll_field_8, spll_ssc_en, &ssc_enable);
+
+	return ssc_enable == 1;
+}
+
+void dcn314_init_clocks(struct clk_mgr *clk_mgr)
+{
+	struct clk_mgr_internal *clk_mgr_int = TO_CLK_MGR_INTERNAL(clk_mgr);
+	uint32_t ref_dtbclk = clk_mgr->clks.ref_dtbclk_khz;
+	struct clk_mgr_dcn314 *clk_mgr_dcn314 = TO_CLK_MGR_DCN314(clk_mgr_int);
+	struct clk_log_info log_info = {0};
+
+	memset(&(clk_mgr->clks), 0, sizeof(struct dc_clocks));
+	// Assumption is that boot state always supports pstate
+	clk_mgr->clks.ref_dtbclk_khz = ref_dtbclk;	// restore ref_dtbclk
+	clk_mgr->clks.p_state_change_support = true;
+	clk_mgr->clks.prev_p_state_change_support = true;
+	clk_mgr->clks.pwr_state = DCN_PWR_STATE_UNKNOWN;
+	clk_mgr->clks.zstate_support = DCN_ZSTATE_SUPPORT_UNKNOWN;
+
+	// to adjust dp_dto reference clock if ssc is enable otherwise to apply dprefclk
+	if (dcn314_is_spll_ssc_enabled(clk_mgr))
+		clk_mgr->dp_dto_source_clock_in_khz =
+			dce_adjust_dp_ref_freq_for_ss(clk_mgr_int, clk_mgr->dprefclk_khz);
+	else
+		clk_mgr->dp_dto_source_clock_in_khz = clk_mgr->dprefclk_khz;
+
+	dcn314_dump_clk_registers(&clk_mgr->boot_snapshot, &clk_mgr_dcn314->base.base, &log_info);
+	clk_mgr->clks.dispclk_khz =  clk_mgr->boot_snapshot.dispclk * 1000;
+}
+
 void dcn314_update_clocks(struct clk_mgr *clk_mgr_base,
 			struct dc_state *context,
 			bool safe_to_lower)
@@ -176,6 +286,8 @@ void dcn314_update_clocks(struct clk_mgr *clk_mgr_base,
 	if (dc->work_arounds.skip_clock_update)
 		return;
 
+	display_count = dcn314_get_active_display_cnt_wa(dc, context);
+
 	/*
 	 * if it is safe to lower, but we are already in the lower state, we don't have to do anything
 	 * also if safe to lower is false, we just go in the higher state
@@ -194,7 +306,6 @@ void dcn314_update_clocks(struct clk_mgr *clk_mgr_base,
 		}
 		/* check that we're not already in lower */
 		if (clk_mgr_base->clks.pwr_state != DCN_PWR_STATE_LOW_POWER) {
-			display_count = dcn314_get_active_display_cnt_wa(dc, context);
 			/* if we can go lower, go lower */
 			if (display_count == 0) {
 				union display_idle_optimization_u idle_info = { 0 };
@@ -241,10 +352,8 @@ void dcn314_update_clocks(struct clk_mgr *clk_mgr_base,
 	}
 
 	// workaround: Limit dppclk to 100Mhz to avoid lower eDP panel switch to plus 4K monitor underflow.
-	if (!IS_DIAG_DC(dc->ctx->dce_environment)) {
-		if (new_clocks->dppclk_khz < 100000)
-			new_clocks->dppclk_khz = 100000;
-	}
+	if (new_clocks->dppclk_khz < 100000)
+		new_clocks->dppclk_khz = 100000;
 
 	if (should_set_clock(safe_to_lower, new_clocks->dppclk_khz, clk_mgr->base.clks.dppclk_khz)) {
 		if (clk_mgr->base.clks.dppclk_khz > new_clocks->dppclk_khz)
@@ -253,12 +362,20 @@ void dcn314_update_clocks(struct clk_mgr *clk_mgr_base,
 		update_dppclk = true;
 	}
 
-	if (should_set_clock(safe_to_lower, new_clocks->dispclk_khz, clk_mgr_base->clks.dispclk_khz)) {
-		dcn314_disable_otg_wa(clk_mgr_base, context, true);
+	if (should_set_clock(safe_to_lower, new_clocks->dispclk_khz, clk_mgr_base->clks.dispclk_khz) &&
+	    (new_clocks->dispclk_khz > 0 || (safe_to_lower && display_count == 0))) {
+		int requested_dispclk_khz = new_clocks->dispclk_khz;
+
+		dcn314_disable_otg_wa(clk_mgr_base, context, safe_to_lower, true);
+
+		/* Clamp the requested clock to PMFW based on their limit. */
+		if (dc->debug.min_disp_clk_khz > 0 && requested_dispclk_khz < dc->debug.min_disp_clk_khz)
+			requested_dispclk_khz = dc->debug.min_disp_clk_khz;
 
+		dcn314_smu_set_dispclk(clk_mgr, requested_dispclk_khz);
 		clk_mgr_base->clks.dispclk_khz = new_clocks->dispclk_khz;
-		dcn314_smu_set_dispclk(clk_mgr, clk_mgr_base->clks.dispclk_khz);
-		dcn314_disable_otg_wa(clk_mgr_base, context, false);
+
+		dcn314_disable_otg_wa(clk_mgr_base, context, safe_to_lower, false);
 
 		update_dispclk = true;
 	}
@@ -286,9 +403,7 @@ void dcn314_update_clocks(struct clk_mgr *clk_mgr_base,
 	cmd.notify_clocks.clocks.dispclk_khz = clk_mgr_base->clks.dispclk_khz;
 	cmd.notify_clocks.clocks.dppclk_khz = clk_mgr_base->clks.dppclk_khz;
 
-	dc_dmub_srv_cmd_queue(dc->ctx->dmub_srv, &cmd);
-	dc_dmub_srv_cmd_execute(dc->ctx->dmub_srv);
-	dc_dmub_srv_wait_idle(dc->ctx->dmub_srv);
+	dc_wake_and_execute_dmub_cmd(dc->ctx, &cmd, DM_DMUB_WAIT_TYPE_WAIT);
 }
 
 static int get_vco_frequency_from_reg(struct clk_mgr_internal *clk_mgr)
@@ -347,10 +462,65 @@ bool dcn314_are_clock_states_equal(struct dc_clocks *a,
 	return true;
 }
 
-static void dcn314_dump_clk_registers(struct clk_state_registers_and_bypass *regs_and_bypass,
+
+static void dcn314_dump_clk_registers_internal(struct dcn35_clk_internal *internal, struct clk_mgr *clk_mgr_base)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+
+	// read dtbclk
+	internal->CLK1_CLK4_CURRENT_CNT = REG_READ(CLK1_CLK4_CURRENT_CNT);
+	internal->CLK1_CLK4_BYPASS_CNTL = REG_READ(CLK1_CLK4_BYPASS_CNTL);
+
+	// read dcfclk
+	internal->CLK1_CLK3_CURRENT_CNT = REG_READ(CLK1_CLK3_CURRENT_CNT);
+	internal->CLK1_CLK3_BYPASS_CNTL = REG_READ(CLK1_CLK3_BYPASS_CNTL);
+
+	// read dcf deep sleep divider
+	internal->CLK1_CLK3_DS_CNTL = REG_READ(CLK1_CLK3_DS_CNTL);
+	internal->CLK1_CLK3_ALLOW_DS = REG_READ(CLK1_CLK3_ALLOW_DS);
+
+	// read dppclk
+	internal->CLK1_CLK1_CURRENT_CNT = REG_READ(CLK1_CLK1_CURRENT_CNT);
+	internal->CLK1_CLK1_BYPASS_CNTL = REG_READ(CLK1_CLK1_BYPASS_CNTL);
+
+	// read dprefclk
+	internal->CLK1_CLK2_CURRENT_CNT = REG_READ(CLK1_CLK2_CURRENT_CNT);
+	internal->CLK1_CLK2_BYPASS_CNTL = REG_READ(CLK1_CLK2_BYPASS_CNTL);
+
+	// read dispclk
+	internal->CLK1_CLK0_CURRENT_CNT = REG_READ(CLK1_CLK0_CURRENT_CNT);
+	internal->CLK1_CLK0_BYPASS_CNTL = REG_READ(CLK1_CLK0_BYPASS_CNTL);
+}
+
+void dcn314_dump_clk_registers(struct clk_state_registers_and_bypass *regs_and_bypass,
 		struct clk_mgr *clk_mgr_base, struct clk_log_info *log_info)
 {
-	return;
+
+	struct dcn35_clk_internal internal = {0};
+
+	dcn314_dump_clk_registers_internal(&internal, clk_mgr_base);
+
+	regs_and_bypass->dcfclk = internal.CLK1_CLK3_CURRENT_CNT / 10;
+	regs_and_bypass->dcf_deep_sleep_divider = internal.CLK1_CLK3_DS_CNTL / 10;
+	regs_and_bypass->dcf_deep_sleep_allow = internal.CLK1_CLK3_ALLOW_DS;
+	regs_and_bypass->dprefclk = internal.CLK1_CLK2_CURRENT_CNT / 10;
+	regs_and_bypass->dispclk = internal.CLK1_CLK0_CURRENT_CNT / 10;
+	regs_and_bypass->dppclk = internal.CLK1_CLK1_CURRENT_CNT / 10;
+	regs_and_bypass->dtbclk = internal.CLK1_CLK4_CURRENT_CNT / 10;
+
+	regs_and_bypass->dppclk_bypass = internal.CLK1_CLK1_BYPASS_CNTL & 0x0007;
+	if (regs_and_bypass->dppclk_bypass > 4)
+		regs_and_bypass->dppclk_bypass = 0;
+	regs_and_bypass->dcfclk_bypass = internal.CLK1_CLK3_BYPASS_CNTL & 0x0007;
+	if (regs_and_bypass->dcfclk_bypass > 4)
+		regs_and_bypass->dcfclk_bypass = 0;
+	regs_and_bypass->dispclk_bypass = internal.CLK1_CLK0_BYPASS_CNTL & 0x0007;
+	if (regs_and_bypass->dispclk_bypass > 4)
+		regs_and_bypass->dispclk_bypass = 0;
+	regs_and_bypass->dprefclk_bypass = internal.CLK1_CLK2_BYPASS_CNTL & 0x0007;
+	if (regs_and_bypass->dprefclk_bypass > 4)
+		regs_and_bypass->dprefclk_bypass = 0;
+
 }
 
 static struct clk_bw_params dcn314_bw_params = {
@@ -405,32 +575,32 @@ static struct wm_table lpddr5_wm_table = {
 			.wm_inst = WM_A,
 			.wm_type = WM_TYPE_PSTATE_CHG,
 			.pstate_latency_us = 11.65333,
-			.sr_exit_time_us = 16.5,
-			.sr_enter_plus_exit_time_us = 18.5,
+			.sr_exit_time_us = 30.0,
+			.sr_enter_plus_exit_time_us = 32.0,
 			.valid = true,
 		},
 		{
 			.wm_inst = WM_B,
 			.wm_type = WM_TYPE_PSTATE_CHG,
 			.pstate_latency_us = 11.65333,
-			.sr_exit_time_us = 16.5,
-			.sr_enter_plus_exit_time_us = 18.5,
+			.sr_exit_time_us = 30.0,
+			.sr_enter_plus_exit_time_us = 32.0,
 			.valid = true,
 		},
 		{
 			.wm_inst = WM_C,
 			.wm_type = WM_TYPE_PSTATE_CHG,
 			.pstate_latency_us = 11.65333,
-			.sr_exit_time_us = 16.5,
-			.sr_enter_plus_exit_time_us = 18.5,
+			.sr_exit_time_us = 30.0,
+			.sr_enter_plus_exit_time_us = 32.0,
 			.valid = true,
 		},
 		{
 			.wm_inst = WM_D,
 			.wm_type = WM_TYPE_PSTATE_CHG,
 			.pstate_latency_us = 11.65333,
-			.sr_exit_time_us = 16.5,
-			.sr_enter_plus_exit_time_us = 18.5,
+			.sr_exit_time_us = 30.0,
+			.sr_enter_plus_exit_time_us = 32.0,
 			.valid = true,
 		},
 	}
@@ -440,6 +610,11 @@ static DpmClocks314_t dummy_clocks;
 
 static struct dcn314_watermarks dummy_wms = { 0 };
 
+static struct dcn314_ss_info_table ss_info_table = {
+	.ss_divider = 1000,
+	.ss_percentage = {0, 0, 375, 375, 375}
+};
+
 static void dcn314_build_watermark_ranges(struct clk_bw_params *bw_params, struct dcn314_watermarks *table)
 {
 	int i, num_valid_sets;
@@ -712,13 +887,31 @@ static struct clk_mgr_funcs dcn314_funcs = {
 	.get_dp_ref_clk_frequency = dce12_get_dp_ref_freq_khz,
 	.get_dtb_ref_clk_frequency = dcn31_get_dtb_ref_freq_khz,
 	.update_clocks = dcn314_update_clocks,
-	.init_clocks = dcn31_init_clocks,
+	.init_clocks = dcn314_init_clocks,
 	.enable_pme_wa = dcn314_enable_pme_wa,
 	.are_clock_states_equal = dcn314_are_clock_states_equal,
 	.notify_wm_ranges = dcn314_notify_wm_ranges
 };
 extern struct clk_mgr_funcs dcn3_fpga_funcs;
 
+static void dcn314_read_ss_info_from_lut(struct clk_mgr_internal *clk_mgr)
+{
+	uint32_t clock_source;
+	//uint32_t ssc_enable;
+
+	REG_GET(CLK1_CLK2_BYPASS_CNTL, CLK2_BYPASS_SEL, &clock_source);
+	//REG_GET(CLK6_0_CLK6_spll_field_8, spll_ssc_en, &ssc_enable);
+
+	if (dcn314_is_spll_ssc_enabled(&clk_mgr->base) && (clock_source < ARRAY_SIZE(ss_info_table.ss_percentage))) {
+		clk_mgr->dprefclk_ss_percentage = ss_info_table.ss_percentage[clock_source];
+
+		if (clk_mgr->dprefclk_ss_percentage != 0) {
+			clk_mgr->ss_on_dprefclk = true;
+			clk_mgr->dprefclk_ss_divider = ss_info_table.ss_divider;
+		}
+	}
+}
+
 void dcn314_clk_mgr_construct(
 		struct dc_context *ctx,
 		struct clk_mgr_dcn314 *clk_mgr,
@@ -726,6 +919,7 @@ void dcn314_clk_mgr_construct(
 		struct dccg *dccg)
 {
 	struct dcn314_smu_dpm_clks smu_dpm_clks = { 0 };
+	struct clk_log_info log_info = {0};
 
 	clk_mgr->base.base.ctx = ctx;
 	clk_mgr->base.base.funcs = &dcn314_funcs;
@@ -765,35 +959,28 @@ void dcn314_clk_mgr_construct(
 
 	ASSERT(smu_dpm_clks.dpm_clks);
 
-	if (IS_FPGA_MAXIMUS_DC(ctx->dce_environment)) {
-		clk_mgr->base.base.funcs = &dcn3_fpga_funcs;
-	} else {
-		struct clk_log_info log_info = {0};
-
-		clk_mgr->base.smu_ver = dcn314_smu_get_smu_version(&clk_mgr->base);
-
-		if (clk_mgr->base.smu_ver)
-			clk_mgr->base.smu_present = true;
+	clk_mgr->base.smu_ver = dcn314_smu_get_smu_version(&clk_mgr->base);
 
-		/* TODO: Check we get what we expect during bringup */
-		clk_mgr->base.base.dentist_vco_freq_khz = get_vco_frequency_from_reg(&clk_mgr->base);
+	if (clk_mgr->base.smu_ver)
+		clk_mgr->base.smu_present = true;
 
-		if (ctx->dc_bios->integrated_info->memory_type == LpDdr5MemType)
-			dcn314_bw_params.wm_table = lpddr5_wm_table;
-		else
-			dcn314_bw_params.wm_table = ddr5_wm_table;
+	/* TODO: Check we get what we expect during bringup */
+	clk_mgr->base.base.dentist_vco_freq_khz = get_vco_frequency_from_reg(&clk_mgr->base);
 
-		/* Saved clocks configured at boot for debug purposes */
-		dcn314_dump_clk_registers(&clk_mgr->base.base.boot_snapshot,
-					  &clk_mgr->base.base, &log_info);
+	if (ctx->dc_bios->integrated_info->memory_type == LpDdr5MemType)
+		dcn314_bw_params.wm_table = lpddr5_wm_table;
+	else
+		dcn314_bw_params.wm_table = ddr5_wm_table;
 
-	}
+	/* Saved clocks configured at boot for debug purposes */
+	dcn314_dump_clk_registers(&clk_mgr->base.base.boot_snapshot,
+				  &clk_mgr->base.base, &log_info);
 
 	clk_mgr->base.base.dprefclk_khz = 600000;
 	clk_mgr->base.base.clks.ref_dtbclk_khz = 600000;
 	dce_clock_read_ss_info(&clk_mgr->base);
+	dcn314_read_ss_info_from_lut(&clk_mgr->base);
 	/*if bios enabled SS, driver needs to adjust dtb clock, only enable with correct bios*/
-	//clk_mgr->base.dccg->ref_dtbclk_khz = dce_adjust_dp_ref_freq_for_ss(clk_mgr_internal, clk_mgr->base.base.dprefclk_khz);
 
 	clk_mgr->base.base.bw_params = &dcn314_bw_params;
 
@@ -841,7 +1028,7 @@ void dcn314_clk_mgr_construct(
 					   i, smu_dpm_clks.dpm_clks->DfPstateTable[i].Voltage);
 		}
 
-		if (ctx->dc_bios && ctx->dc_bios->integrated_info && ctx->dc->config.use_default_clock_table == false) {
+		if (ctx->dc_bios->integrated_info && ctx->dc->config.use_default_clock_table == false) {
 			dcn314_clk_mgr_helper_populate_bw_params(
 					&clk_mgr->base,
 					ctx->dc_bios->integrated_info,
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn314/dcn314_clk_mgr.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn314/dcn314_clk_mgr.h
index 171f84340eb2..0577eb527bc3 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn314/dcn314_clk_mgr.h
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn314/dcn314_clk_mgr.h
@@ -28,6 +28,8 @@
 #define __DCN314_CLK_MGR_H__
 #include "clk_mgr_internal.h"
 
+#define DCN314_NUM_CLOCK_SOURCES   5
+
 struct dcn314_watermarks;
 
 struct dcn314_smu_watermark_set {
@@ -40,9 +42,18 @@ struct clk_mgr_dcn314 {
 	struct dcn314_smu_watermark_set smu_wm_set;
 };
 
+struct dcn314_ss_info_table {
+	uint32_t ss_divider;
+	uint32_t ss_percentage[DCN314_NUM_CLOCK_SOURCES];
+};
+
 bool dcn314_are_clock_states_equal(struct dc_clocks *a,
 		struct dc_clocks *b);
 
+bool dcn314_is_spll_ssc_enabled(struct clk_mgr *clk_mgr_base);
+
+void dcn314_init_clocks(struct clk_mgr *clk_mgr);
+
 void dcn314_update_clocks(struct clk_mgr *clk_mgr_base,
 			struct dc_state *context,
 			bool safe_to_lower);
@@ -54,4 +65,9 @@ void dcn314_clk_mgr_construct(struct dc_context *ctx,
 
 void dcn314_clk_mgr_destroy(struct clk_mgr_internal *clk_mgr_int);
 
+
+void dcn314_dump_clk_registers(struct clk_state_registers_and_bypass *regs_and_bypass,
+		struct clk_mgr *clk_mgr_base, struct clk_log_info *log_info);
+
+
 #endif //__DCN314_CLK_MGR_H__
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn314/dcn314_smu.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn314/dcn314_smu.c
index 0765334f0825..c4af406146b7 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn314/dcn314_smu.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn314/dcn314_smu.c
@@ -145,7 +145,7 @@ static int dcn314_smu_send_msg_with_param(struct clk_mgr_internal *clk_mgr,
 	if (result == VBIOSSMC_Result_Failed) {
 		if (msg_id == VBIOSSMC_MSG_TransferTableDram2Smu &&
 		    param == TABLE_WATERMARKS)
-			DC_LOG_WARNING("Watermarks table not configured properly by SMU");
+			DC_LOG_DEBUG("Watermarks table not configured properly by SMU");
 		else if (msg_id == VBIOSSMC_MSG_SetHardMinDcfclkByFreq ||
 			 msg_id == VBIOSSMC_MSG_SetMinDeepSleepDcfclk)
 			DC_LOG_WARNING("DCFCLK_DPM is not enabled by BIOS");
@@ -220,12 +220,6 @@ int dcn314_smu_set_hard_min_dcfclk(struct clk_mgr_internal *clk_mgr, int request
 			VBIOSSMC_MSG_SetHardMinDcfclkByFreq,
 			khz_to_mhz_ceil(requested_dcfclk_khz));
 
-#ifdef DBG
-	smu_print("actual_dcfclk_set_mhz %d is set to : %d\n",
-			actual_dcfclk_set_mhz,
-			actual_dcfclk_set_mhz * 1000);
-#endif
-
 	return actual_dcfclk_set_mhz * 1000;
 }
 
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn314/dcn314_smu.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn314/dcn314_smu.h
index 047d19ea919c..78ca1e5c5e9e 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn314/dcn314_smu.h
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn314/dcn314_smu.h
@@ -37,34 +37,34 @@ typedef enum {
 } WCK_RATIO_e;
 
 typedef struct {
-  uint32_t FClk;
-  uint32_t MemClk;
-  uint32_t Voltage;
-  uint8_t  WckRatio;
-  uint8_t  Spare[3];
+	uint32_t FClk;
+	uint32_t MemClk;
+	uint32_t Voltage;
+	uint8_t  WckRatio;
+	uint8_t  Spare[3];
 } DfPstateTable314_t;
 
 //Freq in MHz
 //Voltage in milli volts with 2 fractional bits
 typedef struct {
-  uint32_t DcfClocks[NUM_DCFCLK_DPM_LEVELS];
-  uint32_t DispClocks[NUM_DISPCLK_DPM_LEVELS];
-  uint32_t DppClocks[NUM_DPPCLK_DPM_LEVELS];
-  uint32_t SocClocks[NUM_SOCCLK_DPM_LEVELS];
-  uint32_t VClocks[NUM_VCN_DPM_LEVELS];
-  uint32_t DClocks[NUM_VCN_DPM_LEVELS];
-  uint32_t SocVoltage[NUM_SOC_VOLTAGE_LEVELS];
-  DfPstateTable314_t DfPstateTable[NUM_DF_PSTATE_LEVELS];
+	uint32_t DcfClocks[NUM_DCFCLK_DPM_LEVELS];
+	uint32_t DispClocks[NUM_DISPCLK_DPM_LEVELS];
+	uint32_t DppClocks[NUM_DPPCLK_DPM_LEVELS];
+	uint32_t SocClocks[NUM_SOCCLK_DPM_LEVELS];
+	uint32_t VClocks[NUM_VCN_DPM_LEVELS];
+	uint32_t DClocks[NUM_VCN_DPM_LEVELS];
+	uint32_t SocVoltage[NUM_SOC_VOLTAGE_LEVELS];
+	DfPstateTable314_t DfPstateTable[NUM_DF_PSTATE_LEVELS];
 
-  uint8_t  NumDcfClkLevelsEnabled;
-  uint8_t  NumDispClkLevelsEnabled; //Applies to both Dispclk and Dppclk
-  uint8_t  NumSocClkLevelsEnabled;
-  uint8_t  VcnClkLevelsEnabled;     //Applies to both Vclk and Dclk
-  uint8_t  NumDfPstatesEnabled;
-  uint8_t  spare[3];
+	uint8_t  NumDcfClkLevelsEnabled;
+	uint8_t  NumDispClkLevelsEnabled; //Applies to both Dispclk and Dppclk
+	uint8_t  NumSocClkLevelsEnabled;
+	uint8_t  VcnClkLevelsEnabled;     //Applies to both Vclk and Dclk
+	uint8_t  NumDfPstatesEnabled;
+	uint8_t  spare[3];
 
-  uint32_t MinGfxClk;
-  uint32_t MaxGfxClk;
+	uint32_t MinGfxClk;
+	uint32_t MaxGfxClk;
 } DpmClocks314_t;
 
 struct dcn314_watermarks {
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn315/dcn315_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn315/dcn315_clk_mgr.c
index b737cbc468f5..3a881451e9da 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn315/dcn315_clk_mgr.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn315/dcn315_clk_mgr.c
@@ -40,17 +40,51 @@
 #include "dm_helpers.h"
 
 #include "dc_dmub_srv.h"
-
+#include "reg_helper.h"
 #include "logger_types.h"
 #undef DC_LOGGER
 #define DC_LOGGER \
 	clk_mgr->base.base.ctx->logger
 
-#include "link.h"
+#include "link_service.h"
+
+#define MAX_INSTANCE                                        7
+#define MAX_SEGMENT                                         8
+
+struct IP_BASE_INSTANCE {
+	unsigned int segment[MAX_SEGMENT];
+};
+
+struct IP_BASE {
+	struct IP_BASE_INSTANCE instance[MAX_INSTANCE];
+};
+
+static const struct IP_BASE CLK_BASE = { { { { 0x00016C00, 0x02401800, 0, 0, 0, 0, 0, 0 } },
+					{ { 0x00016E00, 0x02401C00, 0, 0, 0, 0, 0, 0 } },
+					{ { 0x00017000, 0x02402000, 0, 0, 0, 0, 0, 0 } },
+					{ { 0x00017200, 0x02402400, 0, 0, 0, 0, 0, 0 } },
+					{ { 0x0001B000, 0x0242D800, 0, 0, 0, 0, 0, 0 } },
+					{ { 0x0001B200, 0x0242DC00, 0, 0, 0, 0, 0, 0 } } } };
+
+#define regCLK1_CLK0_CURRENT_CNT			0x0314
+#define regCLK1_CLK0_CURRENT_CNT_BASE_IDX	0
+#define regCLK1_CLK1_CURRENT_CNT			0x0315
+#define regCLK1_CLK1_CURRENT_CNT_BASE_IDX	0
+#define regCLK1_CLK2_CURRENT_CNT			0x0316
+#define regCLK1_CLK2_CURRENT_CNT_BASE_IDX	0
+#define regCLK1_CLK3_CURRENT_CNT			0x0317
+#define regCLK1_CLK3_CURRENT_CNT_BASE_IDX	0
+#define regCLK1_CLK4_CURRENT_CNT			0x0318
+#define regCLK1_CLK4_CURRENT_CNT_BASE_IDX	0
+#define regCLK1_CLK5_CURRENT_CNT			0x0319
+#define regCLK1_CLK5_CURRENT_CNT_BASE_IDX	0
 
 #define TO_CLK_MGR_DCN315(clk_mgr)\
 	container_of(clk_mgr, struct clk_mgr_dcn315, base)
 
+#define REG(reg_name) \
+	(CLK_BASE.instance[0].segment[reg ## reg_name ## _BASE_IDX] + reg ## reg_name)
+
 #define UNSUPPORTED_DCFCLK 10000000
 #define MIN_DPP_DISP_CLK     100000
 
@@ -130,7 +164,7 @@ static void dcn315_update_clocks(struct clk_mgr *clk_mgr_base,
 	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
 	struct dc_clocks *new_clocks = &context->bw_ctx.bw.dcn.clk;
 	struct dc *dc = clk_mgr_base->ctx->dc;
-	int display_count;
+	int display_count = 0;
 	bool update_dppclk = false;
 	bool update_dispclk = false;
 	bool dpp_clock_lowered = false;
@@ -145,6 +179,10 @@ static void dcn315_update_clocks(struct clk_mgr *clk_mgr_base,
 	 */
 	clk_mgr_base->clks.zstate_support = new_clocks->zstate_support;
 	if (safe_to_lower) {
+		if (clk_mgr_base->clks.dtbclk_en && !new_clocks->dtbclk_en) {
+			dcn315_smu_set_dtbclk(clk_mgr, false);
+			clk_mgr_base->clks.dtbclk_en = new_clocks->dtbclk_en;
+		}
 		/* check that we're not already in lower */
 		if (clk_mgr_base->clks.pwr_state != DCN_PWR_STATE_LOW_POWER) {
 			display_count = dcn315_get_active_display_cnt_wa(dc, context);
@@ -160,6 +198,10 @@ static void dcn315_update_clocks(struct clk_mgr *clk_mgr_base,
 			}
 		}
 	} else {
+		if (!clk_mgr_base->clks.dtbclk_en && new_clocks->dtbclk_en) {
+			dcn315_smu_set_dtbclk(clk_mgr, true);
+			clk_mgr_base->clks.dtbclk_en = new_clocks->dtbclk_en;
+		}
 		/* check that we're not already in D0 */
 		if (clk_mgr_base->clks.pwr_state != DCN_PWR_STATE_MISSION_MODE) {
 			union display_idle_optimization_u idle_info = { 0 };
@@ -184,12 +226,8 @@ static void dcn315_update_clocks(struct clk_mgr *clk_mgr_base,
 	}
 
 	// workaround: Limit dppclk to 100Mhz to avoid lower eDP panel switch to plus 4K monitor underflow.
-	if (!IS_DIAG_DC(dc->ctx->dce_environment)) {
-		if (new_clocks->dppclk_khz < MIN_DPP_DISP_CLK)
-			new_clocks->dppclk_khz = MIN_DPP_DISP_CLK;
-		if (new_clocks->dispclk_khz < MIN_DPP_DISP_CLK)
-			new_clocks->dispclk_khz = MIN_DPP_DISP_CLK;
-	}
+	if (new_clocks->dppclk_khz < MIN_DPP_DISP_CLK)
+		new_clocks->dppclk_khz = MIN_DPP_DISP_CLK;
 
 	if (should_set_clock(safe_to_lower, new_clocks->dppclk_khz, clk_mgr->base.clks.dppclk_khz)) {
 		if (clk_mgr->base.clks.dppclk_khz > new_clocks->dppclk_khz)
@@ -198,15 +236,19 @@ static void dcn315_update_clocks(struct clk_mgr *clk_mgr_base,
 		update_dppclk = true;
 	}
 
-	if (should_set_clock(safe_to_lower, new_clocks->dispclk_khz, clk_mgr_base->clks.dispclk_khz)) {
-		/* No need to apply the w/a if we haven't taken over from bios yet */
-		if (clk_mgr_base->clks.dispclk_khz)
-			dcn315_disable_otg_wa(clk_mgr_base, context, true);
+	if (should_set_clock(safe_to_lower, new_clocks->dispclk_khz, clk_mgr_base->clks.dispclk_khz) &&
+	    (new_clocks->dispclk_khz > 0 || (safe_to_lower && display_count == 0))) {
+		int requested_dispclk_khz = new_clocks->dispclk_khz;
+
+		dcn315_disable_otg_wa(clk_mgr_base, context, true);
+
+		/* Clamp the requested clock to PMFW based on their limit. */
+		if (dc->debug.min_disp_clk_khz > 0 && requested_dispclk_khz < dc->debug.min_disp_clk_khz)
+			requested_dispclk_khz = dc->debug.min_disp_clk_khz;
 
+		dcn315_smu_set_dispclk(clk_mgr, requested_dispclk_khz);
 		clk_mgr_base->clks.dispclk_khz = new_clocks->dispclk_khz;
-		dcn315_smu_set_dispclk(clk_mgr, clk_mgr_base->clks.dispclk_khz);
-		if (clk_mgr_base->clks.dispclk_khz)
-			dcn315_disable_otg_wa(clk_mgr_base, context, false);
+		dcn315_disable_otg_wa(clk_mgr_base, context, false);
 
 		update_dispclk = true;
 	}
@@ -234,14 +276,41 @@ static void dcn315_update_clocks(struct clk_mgr *clk_mgr_base,
 	cmd.notify_clocks.clocks.dispclk_khz = clk_mgr_base->clks.dispclk_khz;
 	cmd.notify_clocks.clocks.dppclk_khz = clk_mgr_base->clks.dppclk_khz;
 
-	dc_dmub_srv_cmd_queue(dc->ctx->dmub_srv, &cmd);
-	dc_dmub_srv_cmd_execute(dc->ctx->dmub_srv);
-	dc_dmub_srv_wait_idle(dc->ctx->dmub_srv);
+	dc_wake_and_execute_dmub_cmd(dc->ctx, &cmd, DM_DMUB_WAIT_TYPE_WAIT);
+}
+
+static void dcn315_dump_clk_registers_internal(struct dcn35_clk_internal *internal, struct clk_mgr *clk_mgr_base)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+
+	// read dtbclk
+	internal->CLK1_CLK4_CURRENT_CNT = REG_READ(CLK1_CLK4_CURRENT_CNT);
+
+	// read dcfclk
+	internal->CLK1_CLK3_CURRENT_CNT = REG_READ(CLK1_CLK3_CURRENT_CNT);
+
+	// read dppclk
+	internal->CLK1_CLK1_CURRENT_CNT = REG_READ(CLK1_CLK1_CURRENT_CNT);
+
+	// read dprefclk
+	internal->CLK1_CLK2_CURRENT_CNT = REG_READ(CLK1_CLK2_CURRENT_CNT);
+
+	// read dispclk
+	internal->CLK1_CLK0_CURRENT_CNT = REG_READ(CLK1_CLK0_CURRENT_CNT);
 }
 
 static void dcn315_dump_clk_registers(struct clk_state_registers_and_bypass *regs_and_bypass,
 		struct clk_mgr *clk_mgr_base, struct clk_log_info *log_info)
 {
+	struct dcn35_clk_internal internal = {0};
+
+	dcn315_dump_clk_registers_internal(&internal, clk_mgr_base);
+
+	regs_and_bypass->dcfclk = internal.CLK1_CLK3_CURRENT_CNT / 10;
+	regs_and_bypass->dprefclk = internal.CLK1_CLK2_CURRENT_CNT / 10;
+	regs_and_bypass->dispclk = internal.CLK1_CLK0_CURRENT_CNT / 10;
+	regs_and_bypass->dppclk = internal.CLK1_CLK1_CURRENT_CNT / 10;
+	regs_and_bypass->dtbclk = internal.CLK1_CLK4_CURRENT_CNT / 10;
 	return;
 }
 
@@ -338,7 +407,7 @@ static struct wm_table lpddr5_wm_table = {
 		{
 			.wm_inst = WM_A,
 			.wm_type = WM_TYPE_PSTATE_CHG,
-			.pstate_latency_us = 11.65333,
+			.pstate_latency_us = 129.0,
 			.sr_exit_time_us = 11.5,
 			.sr_enter_plus_exit_time_us = 14.5,
 			.valid = true,
@@ -346,7 +415,7 @@ static struct wm_table lpddr5_wm_table = {
 		{
 			.wm_inst = WM_B,
 			.wm_type = WM_TYPE_PSTATE_CHG,
-			.pstate_latency_us = 11.65333,
+			.pstate_latency_us = 129.0,
 			.sr_exit_time_us = 11.5,
 			.sr_enter_plus_exit_time_us = 14.5,
 			.valid = true,
@@ -354,7 +423,7 @@ static struct wm_table lpddr5_wm_table = {
 		{
 			.wm_inst = WM_C,
 			.wm_type = WM_TYPE_PSTATE_CHG,
-			.pstate_latency_us = 11.65333,
+			.pstate_latency_us = 129.0,
 			.sr_exit_time_us = 11.5,
 			.sr_enter_plus_exit_time_us = 14.5,
 			.valid = true,
@@ -362,7 +431,7 @@ static struct wm_table lpddr5_wm_table = {
 		{
 			.wm_inst = WM_D,
 			.wm_type = WM_TYPE_PSTATE_CHG,
-			.pstate_latency_us = 11.65333,
+			.pstate_latency_us = 129.0,
 			.sr_exit_time_us = 11.5,
 			.sr_enter_plus_exit_time_us = 14.5,
 			.valid = true,
@@ -588,13 +657,32 @@ static struct clk_mgr_funcs dcn315_funcs = {
 	.get_dp_ref_clk_frequency = dce12_get_dp_ref_freq_khz,
 	.get_dtb_ref_clk_frequency = dcn31_get_dtb_ref_freq_khz,
 	.update_clocks = dcn315_update_clocks,
-	.init_clocks = dcn31_init_clocks,
+	.init_clocks = dcn315_init_clocks,
 	.enable_pme_wa = dcn315_enable_pme_wa,
 	.are_clock_states_equal = dcn31_are_clock_states_equal,
 	.notify_wm_ranges = dcn315_notify_wm_ranges
 };
 extern struct clk_mgr_funcs dcn3_fpga_funcs;
 
+void dcn315_init_clocks(struct clk_mgr *clk_mgr)
+{
+	struct clk_mgr_internal *clk_mgr_int = TO_CLK_MGR_INTERNAL(clk_mgr);
+	uint32_t ref_dtbclk = clk_mgr->clks.ref_dtbclk_khz;
+	struct clk_mgr_dcn315 *clk_mgr_dcn315 = TO_CLK_MGR_DCN315(clk_mgr_int);
+	struct clk_log_info log_info = {0};
+
+	memset(&(clk_mgr->clks), 0, sizeof(struct dc_clocks));
+	// Assumption is that boot state always supports pstate
+	clk_mgr->clks.ref_dtbclk_khz = ref_dtbclk;	// restore ref_dtbclk
+	clk_mgr->clks.p_state_change_support = true;
+	clk_mgr->clks.prev_p_state_change_support = true;
+	clk_mgr->clks.pwr_state = DCN_PWR_STATE_UNKNOWN;
+	clk_mgr->clks.zstate_support = DCN_ZSTATE_SUPPORT_UNKNOWN;
+
+	dcn315_dump_clk_registers(&clk_mgr->boot_snapshot, &clk_mgr_dcn315->base.base, &log_info);
+	clk_mgr->clks.dispclk_khz =  clk_mgr->boot_snapshot.dispclk * 1000;
+}
+
 void dcn315_clk_mgr_construct(
 		struct dc_context *ctx,
 		struct clk_mgr_dcn315 *clk_mgr,
@@ -602,6 +690,7 @@ void dcn315_clk_mgr_construct(
 		struct dccg *dccg)
 {
 	struct dcn315_smu_dpm_clks smu_dpm_clks = { 0 };
+	struct clk_log_info log_info = {0};
 
 	clk_mgr->base.base.ctx = ctx;
 	clk_mgr->base.base.funcs = &dcn315_funcs;
@@ -641,26 +730,20 @@ void dcn315_clk_mgr_construct(
 
 	ASSERT(smu_dpm_clks.dpm_clks);
 
-	if (IS_FPGA_MAXIMUS_DC(ctx->dce_environment)) {
-		clk_mgr->base.base.funcs = &dcn3_fpga_funcs;
-	} else {
-		struct clk_log_info log_info = {0};
-
-		clk_mgr->base.smu_ver = dcn315_smu_get_smu_version(&clk_mgr->base);
+	clk_mgr->base.smu_ver = dcn315_smu_get_smu_version(&clk_mgr->base);
 
-		if (clk_mgr->base.smu_ver > 0)
-			clk_mgr->base.smu_present = true;
-
-		if (ctx->dc_bios->integrated_info->memory_type == LpDdr5MemType) {
-			dcn315_bw_params.wm_table = lpddr5_wm_table;
-		} else {
-			dcn315_bw_params.wm_table = ddr5_wm_table;
-		}
-		/* Saved clocks configured at boot for debug purposes */
-		dcn315_dump_clk_registers(&clk_mgr->base.base.boot_snapshot,
-					  &clk_mgr->base.base, &log_info);
+	if (clk_mgr->base.smu_ver > 0)
+		clk_mgr->base.smu_present = true;
 
+	if (ctx->dc_bios->integrated_info->memory_type == LpDdr5MemType) {
+		dcn315_bw_params.wm_table = lpddr5_wm_table;
+	} else {
+		dcn315_bw_params.wm_table = ddr5_wm_table;
 	}
+	/* Saved clocks configured at boot for debug purposes */
+	dcn315_dump_clk_registers(&clk_mgr->base.base.boot_snapshot,
+				  &clk_mgr->base.base, &log_info);
+	clk_mgr->base.base.clks.dispclk_khz =  clk_mgr->base.base.boot_snapshot.dispclk * 1000;
 
 	clk_mgr->base.base.dprefclk_khz = 600000;
 	clk_mgr->base.base.dprefclk_khz = dcn315_smu_get_dpref_clk(&clk_mgr->base);
@@ -714,7 +797,7 @@ void dcn315_clk_mgr_construct(
 					   i, smu_dpm_clks.dpm_clks->DfPstateTable[i].Voltage);
 		}
 
-		if (ctx->dc_bios && ctx->dc_bios->integrated_info) {
+		if (ctx->dc_bios->integrated_info) {
 			dcn315_clk_mgr_helper_populate_bw_params(
 					&clk_mgr->base,
 					ctx->dc_bios->integrated_info,
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn315/dcn315_clk_mgr.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn315/dcn315_clk_mgr.h
index ac36ddf5dd1a..642ae3d4a790 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn315/dcn315_clk_mgr.h
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn315/dcn315_clk_mgr.h
@@ -44,6 +44,7 @@ void dcn315_clk_mgr_construct(struct dc_context *ctx,
 		struct pp_smu_funcs *pp_smu,
 		struct dccg *dccg);
 
+void dcn315_init_clocks(struct clk_mgr *clk_mgr);
 void dcn315_clk_mgr_destroy(struct clk_mgr_internal *clk_mgr_int);
 
 #endif //__DCN315_CLK_MGR_H__
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn315/dcn315_smu.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn315/dcn315_smu.c
index 925d6e13620e..478b4d6a3544 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn315/dcn315_smu.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn315/dcn315_smu.c
@@ -29,32 +29,29 @@
 #include "dm_helpers.h"
 #include "dcn315_smu.h"
 #include "mp/mp_13_0_5_offset.h"
+#include "logger_types.h"
 
 #define MAX_INSTANCE                                        6
 #define MAX_SEGMENT                                         6
+#define SMU_REGISTER_WRITE_RETRY_COUNT                      5
 
-struct IP_BASE_INSTANCE
-{
+struct IP_BASE_INSTANCE {
     unsigned int segment[MAX_SEGMENT];
 };
 
-struct IP_BASE
-{
+struct IP_BASE {
     struct IP_BASE_INSTANCE instance[MAX_INSTANCE];
 };
 
 static const struct IP_BASE MP0_BASE = { { { { 0x00016000, 0x00DC0000, 0x00E00000, 0x00E40000, 0x0243FC00, 0 } },
-                                        { { 0, 0, 0, 0, 0, 0 } },
-                                        { { 0, 0, 0, 0, 0, 0 } },
-                                        { { 0, 0, 0, 0, 0, 0 } },
-                                        { { 0, 0, 0, 0, 0, 0 } },
-                                        { { 0, 0, 0, 0, 0, 0 } } } };
-static const struct IP_BASE NBIO_BASE = { { { { 0x00000000, 0x00000014, 0x00000D20, 0x00010400, 0x0241B000, 0x04040000 } },
-                                        { { 0, 0, 0, 0, 0, 0 } },
-                                        { { 0, 0, 0, 0, 0, 0 } },
-                                        { { 0, 0, 0, 0, 0, 0 } },
-                                        { { 0, 0, 0, 0, 0, 0 } },
-                                        { { 0, 0, 0, 0, 0, 0 } } } };
+					{ { 0, 0, 0, 0, 0, 0 } },
+					{ { 0, 0, 0, 0, 0, 0 } },
+					{ { 0, 0, 0, 0, 0, 0 } },
+					{ { 0, 0, 0, 0, 0, 0 } },
+					{ { 0, 0, 0, 0, 0, 0 } } } };
+
+#define CTX clk_mgr->base.ctx
+#define IND_REG(offset)	offset
 
 #define regBIF_BX_PF2_RSMU_INDEX                                                                        0x0000
 #define regBIF_BX_PF2_RSMU_INDEX_BASE_IDX                                                               1
@@ -67,10 +64,6 @@ static const struct IP_BASE NBIO_BASE = { { { { 0x00000000, 0x00000014, 0x00000D
 #define FN(reg_name, field) \
 	FD(reg_name##__##field)
 
-#define REG_NBIO(reg_name) \
-	(NBIO_BASE.instance[0].segment[regBIF_BX_PF2_ ## reg_name ## _BASE_IDX] + regBIF_BX_PF2_ ## reg_name)
-
-#include "logger_types.h"
 #undef DC_LOGGER
 #define DC_LOGGER \
 	CTX->logger
@@ -78,6 +71,13 @@ static const struct IP_BASE NBIO_BASE = { { { { 0x00000000, 0x00000014, 0x00000D
 
 #define mmMP1_C2PMSG_3                            0x3B1050C
 
+#define reg__MP1_C2PMSG_3_MASK					(0xFFFFFFFF)
+#define reg__MP1_C2PMSG_3__SHIFT					(0)
+
+
+#define data_reg_name__MP1_C2PMSG_3_MASK		(0xFFFFFFFF)
+#define data_reg_name__MP1_C2PMSG_3__SHIFT		(0)
+
 #define VBIOSSMC_MSG_TestMessage                  0x01 ///< To check if PMFW is alive and responding. Requirement specified by PMFW team
 #define VBIOSSMC_MSG_GetPmfwVersion               0x02 ///< Get PMFW version
 #define VBIOSSMC_MSG_Spare0                       0x03 ///< Spare0
@@ -134,6 +134,8 @@ static int dcn315_smu_send_msg_with_param(
 		unsigned int msg_id, unsigned int param)
 {
 	uint32_t result;
+	uint32_t i = 0;
+	uint32_t read_back_data;
 
 	result = dcn315_smu_wait_for_response(clk_mgr, 10, 200000);
 
@@ -150,10 +152,17 @@ static int dcn315_smu_send_msg_with_param(
 	/* Set the parameter register for the SMU message, unit is Mhz */
 	REG_WRITE(MP1_SMN_C2PMSG_37, param);
 
-	/* Trigger the message transaction by writing the message ID */
-	generic_write_indirect_reg(CTX,
-		REG_NBIO(RSMU_INDEX), REG_NBIO(RSMU_DATA),
-		mmMP1_C2PMSG_3, msg_id);
+	for (i = 0; i < SMU_REGISTER_WRITE_RETRY_COUNT; i++) {
+		/* Trigger the message transaction by writing the message ID */
+		IX_REG_SET_SYNC(mmMP1_C2PMSG_3, 0,
+			MP1_C2PMSG_3, msg_id);
+		IX_REG_GET_SYNC(mmMP1_C2PMSG_3,
+			MP1_C2PMSG_3, &read_back_data);
+		if (read_back_data == msg_id)
+			break;
+		udelay(2);
+		smu_print("SMU msg id write fail %x times. \n", i + 1);
+	}
 
 	result = dcn315_smu_wait_for_response(clk_mgr, 10, 200000);
 
@@ -205,10 +214,6 @@ int dcn315_smu_set_hard_min_dcfclk(struct clk_mgr_internal *clk_mgr, int request
 			VBIOSSMC_MSG_SetHardMinDcfclkByFreq,
 			khz_to_mhz_ceil(requested_dcfclk_khz));
 
-#ifdef DBG
-	smu_print("actual_dcfclk_set_mhz %d is set to : %d\n", actual_dcfclk_set_mhz, actual_dcfclk_set_mhz * 1000);
-#endif
-
 	return actual_dcfclk_set_mhz * 1000;
 }
 
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn316/dcn316_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn316/dcn316_clk_mgr.c
index 93db4dbee713..1769b1f26e75 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn316/dcn316_clk_mgr.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn316/dcn316_clk_mgr.c
@@ -39,30 +39,20 @@
 #include "dcn316_smu.h"
 #include "dm_helpers.h"
 #include "dc_dmub_srv.h"
-#include "link.h"
+#include "link_service.h"
 
 // DCN316 this is CLK1 instance
 #define MAX_INSTANCE                                        7
 #define MAX_SEGMENT                                         6
 
-struct IP_BASE_INSTANCE
-{
+struct IP_BASE_INSTANCE {
     unsigned int segment[MAX_SEGMENT];
 };
 
-struct IP_BASE
-{
+struct IP_BASE {
     struct IP_BASE_INSTANCE instance[MAX_INSTANCE];
 };
 
-static const struct IP_BASE CLK_BASE = { { { { 0x00016C00, 0x02401800, 0, 0, 0, 0 } },
-                                        { { 0x00016E00, 0x02401C00, 0, 0, 0, 0 } },
-                                        { { 0x00017000, 0x02402000, 0, 0, 0, 0 } },
-                                        { { 0x00017200, 0x02402400, 0, 0, 0, 0 } },
-                                        { { 0x0001B000, 0x0242D800, 0, 0, 0, 0 } },
-                                        { { 0x0001B200, 0x0242DC00, 0, 0, 0, 0 } },
-                                        { { 0x0001B400, 0x0242E000, 0, 0, 0, 0 } } } };
-
 #define regCLK1_CLK_PLL_REQ						0x0237
 #define regCLK1_CLK_PLL_REQ_BASE_IDX			0
 
@@ -73,9 +63,6 @@ static const struct IP_BASE CLK_BASE = { { { { 0x00016C00, 0x02401800, 0, 0, 0,
 #define CLK1_CLK_PLL_REQ__PllSpineDiv_MASK		0x0000F000L
 #define CLK1_CLK_PLL_REQ__FbMult_frac_MASK		0xFFFF0000L
 
-#define REG(reg_name) \
-	(CLK_BASE.instance[0].segment[reg ## reg_name ## _BASE_IDX] + reg ## reg_name)
-
 #define TO_CLK_MGR_DCN316(clk_mgr)\
 	container_of(clk_mgr, struct clk_mgr_dcn316, base)
 
@@ -112,20 +99,25 @@ static int dcn316_get_active_display_cnt_wa(
 	return display_count;
 }
 
-static void dcn316_disable_otg_wa(struct clk_mgr *clk_mgr_base, struct dc_state *context, bool disable)
+static void dcn316_disable_otg_wa(struct clk_mgr *clk_mgr_base, struct dc_state *context,
+		bool safe_to_lower, bool disable)
 {
 	struct dc *dc = clk_mgr_base->ctx->dc;
 	int i;
 
 	for (i = 0; i < dc->res_pool->pipe_count; ++i) {
-		struct pipe_ctx *pipe = &dc->current_state->res_ctx.pipe_ctx[i];
+		struct pipe_ctx *pipe = safe_to_lower
+			? &context->res_ctx.pipe_ctx[i]
+			: &dc->current_state->res_ctx.pipe_ctx[i];
 
 		if (pipe->top_pipe || pipe->prev_odm_pipe)
 			continue;
-		if (pipe->stream && (pipe->stream->dpms_off || pipe->plane_state == NULL ||
-				     dc_is_virtual_signal(pipe->stream->signal))) {
+		if (pipe->stream && (pipe->stream->dpms_off || dc_is_virtual_signal(pipe->stream->signal) ||
+				     !pipe->stream->link_enc)) {
 			if (disable) {
-				pipe->stream_res.tg->funcs->immediate_disable_crtc(pipe->stream_res.tg);
+				if (pipe->stream_res.tg && pipe->stream_res.tg->funcs->immediate_disable_crtc)
+					pipe->stream_res.tg->funcs->immediate_disable_crtc(pipe->stream_res.tg);
+
 				reset_sync_context_for_pipe(dc, context, i);
 			} else
 				pipe->stream_res.tg->funcs->enable_crtc(pipe->stream_res.tg);
@@ -148,7 +140,7 @@ static void dcn316_update_clocks(struct clk_mgr *clk_mgr_base,
 	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
 	struct dc_clocks *new_clocks = &context->bw_ctx.bw.dcn.clk;
 	struct dc *dc = clk_mgr_base->ctx->dc;
-	int display_count;
+	int display_count = 0;
 	bool update_dppclk = false;
 	bool update_dispclk = false;
 	bool dpp_clock_lowered = false;
@@ -207,12 +199,8 @@ static void dcn316_update_clocks(struct clk_mgr *clk_mgr_base,
 	}
 
 	// workaround: Limit dppclk to 100Mhz to avoid lower eDP panel switch to plus 4K monitor underflow.
-	if (!IS_DIAG_DC(dc->ctx->dce_environment)) {
-		if (new_clocks->dppclk_khz < 100000)
-			new_clocks->dppclk_khz = 100000;
-		if (new_clocks->dispclk_khz < 100000)
-			new_clocks->dispclk_khz = 100000;
-	}
+	if (new_clocks->dppclk_khz < 100000)
+		new_clocks->dppclk_khz = 100000;
 
 	if (should_set_clock(safe_to_lower, new_clocks->dppclk_khz, clk_mgr->base.clks.dppclk_khz)) {
 		if (clk_mgr->base.clks.dppclk_khz > new_clocks->dppclk_khz)
@@ -221,12 +209,19 @@ static void dcn316_update_clocks(struct clk_mgr *clk_mgr_base,
 		update_dppclk = true;
 	}
 
-	if (should_set_clock(safe_to_lower, new_clocks->dispclk_khz, clk_mgr_base->clks.dispclk_khz)) {
-		dcn316_disable_otg_wa(clk_mgr_base, context, true);
+	if (should_set_clock(safe_to_lower, new_clocks->dispclk_khz, clk_mgr_base->clks.dispclk_khz) &&
+	    (new_clocks->dispclk_khz > 0 || (safe_to_lower && display_count == 0))) {
+		int requested_dispclk_khz = new_clocks->dispclk_khz;
+
+		dcn316_disable_otg_wa(clk_mgr_base, context, safe_to_lower, true);
 
+		/* Clamp the requested clock to PMFW based on their limit. */
+		if (dc->debug.min_disp_clk_khz > 0 && requested_dispclk_khz < dc->debug.min_disp_clk_khz)
+			requested_dispclk_khz = dc->debug.min_disp_clk_khz;
+
+		dcn316_smu_set_dispclk(clk_mgr, requested_dispclk_khz);
 		clk_mgr_base->clks.dispclk_khz = new_clocks->dispclk_khz;
-		dcn316_smu_set_dispclk(clk_mgr, clk_mgr_base->clks.dispclk_khz);
-		dcn316_disable_otg_wa(clk_mgr_base, context, false);
+		dcn316_disable_otg_wa(clk_mgr_base, context, safe_to_lower, false);
 
 		update_dispclk = true;
 	}
@@ -254,9 +249,7 @@ static void dcn316_update_clocks(struct clk_mgr *clk_mgr_base,
 	cmd.notify_clocks.clocks.dispclk_khz = clk_mgr_base->clks.dispclk_khz;
 	cmd.notify_clocks.clocks.dppclk_khz = clk_mgr_base->clks.dppclk_khz;
 
-	dc_dmub_srv_cmd_queue(dc->ctx->dmub_srv, &cmd);
-	dc_dmub_srv_cmd_execute(dc->ctx->dmub_srv);
-	dc_dmub_srv_wait_idle(dc->ctx->dmub_srv);
+	dc_wake_and_execute_dmub_cmd(dc->ctx, &cmd, DM_DMUB_WAIT_TYPE_WAIT);
 }
 
 static void dcn316_dump_clk_registers(struct clk_state_registers_and_bypass *regs_and_bypass,
@@ -497,7 +490,8 @@ static void dcn316_clk_mgr_helper_populate_bw_params(
 
 	j = -1;
 
-	ASSERT(NUM_DF_PSTATE_LEVELS <= MAX_NUM_DPM_LVL);
+	static_assert(NUM_DF_PSTATE_LEVELS <= MAX_NUM_DPM_LVL,
+		"number of reported pstate levels exceeds maximum");
 
 	/* Find lowest DPM, FCLK is filled in reverse order*/
 
@@ -581,36 +575,6 @@ static struct clk_mgr_funcs dcn316_funcs = {
 };
 extern struct clk_mgr_funcs dcn3_fpga_funcs;
 
-static int get_vco_frequency_from_reg(struct clk_mgr_internal *clk_mgr)
-{
-	/* get FbMult value */
-	struct fixed31_32 pll_req;
-	unsigned int fbmult_frac_val = 0;
-	unsigned int fbmult_int_val = 0;
-
-	/*
-	 * Register value of fbmult is in 8.16 format, we are converting to 31.32
-	 * to leverage the fix point operations available in driver
-	 */
-
-	REG_GET(CLK1_CLK_PLL_REQ, FbMult_frac, &fbmult_frac_val); /* 16 bit fractional part*/
-	REG_GET(CLK1_CLK_PLL_REQ, FbMult_int, &fbmult_int_val); /* 8 bit integer part */
-
-	pll_req = dc_fixpt_from_int(fbmult_int_val);
-
-	/*
-	 * since fractional part is only 16 bit in register definition but is 32 bit
-	 * in our fix point definiton, need to shift left by 16 to obtain correct value
-	 */
-	pll_req.value |= fbmult_frac_val << 16;
-
-	/* multiply by REFCLK period */
-	pll_req = dc_fixpt_mul_int(pll_req, clk_mgr->dfs_ref_freq_khz);
-
-	/* integer part is now VCO frequency in kHz */
-	return dc_fixpt_floor(pll_req);
-}
-
 void dcn316_clk_mgr_construct(
 		struct dc_context *ctx,
 		struct clk_mgr_dcn316 *clk_mgr,
@@ -618,6 +582,7 @@ void dcn316_clk_mgr_construct(
 		struct dccg *dccg)
 {
 	struct dcn316_smu_dpm_clks smu_dpm_clks = { 0 };
+	struct clk_log_info log_info = {0};
 
 	clk_mgr->base.base.ctx = ctx;
 	clk_mgr->base.base.funcs = &dcn316_funcs;
@@ -657,35 +622,28 @@ void dcn316_clk_mgr_construct(
 
 	ASSERT(smu_dpm_clks.dpm_clks);
 
-	if (IS_FPGA_MAXIMUS_DC(ctx->dce_environment)) {
-		clk_mgr->base.base.funcs = &dcn3_fpga_funcs;
-		clk_mgr->base.base.dentist_vco_freq_khz = 2500000;
-	} else {
-		struct clk_log_info log_info = {0};
-
-		clk_mgr->base.smu_ver = dcn316_smu_get_smu_version(&clk_mgr->base);
-
-		if (clk_mgr->base.smu_ver > 0)
-			clk_mgr->base.smu_present = true;
+	clk_mgr->base.smu_ver = dcn316_smu_get_smu_version(&clk_mgr->base);
 
-		// Skip this for now as it did not work on DCN315, renable during bring up
-		clk_mgr->base.base.dentist_vco_freq_khz = get_vco_frequency_from_reg(&clk_mgr->base);
+	if (clk_mgr->base.smu_ver > 0)
+		clk_mgr->base.smu_present = true;
 
-		/* in case we don't get a value from the register, use default */
-		if (clk_mgr->base.base.dentist_vco_freq_khz == 0)
-			clk_mgr->base.base.dentist_vco_freq_khz = 2500000; /* 2400MHz */
+	// Skip this for now as it did not work on DCN315, renable during bring up
+	//clk_mgr->base.base.dentist_vco_freq_khz = get_vco_frequency_from_reg(&clk_mgr->base);
+	clk_mgr->base.base.dentist_vco_freq_khz = 2500000;
 
+	/* in case we don't get a value from the register, use default */
+	if (clk_mgr->base.base.dentist_vco_freq_khz == 0)
+		clk_mgr->base.base.dentist_vco_freq_khz = 2500000; /* 2400MHz */
 
-		if (ctx->dc_bios->integrated_info->memory_type == LpDdr5MemType) {
-			dcn316_bw_params.wm_table = lpddr5_wm_table;
-		} else {
-			dcn316_bw_params.wm_table = ddr4_wm_table;
-		}
-		/* Saved clocks configured at boot for debug purposes */
-		dcn316_dump_clk_registers(&clk_mgr->base.base.boot_snapshot,
-					  &clk_mgr->base.base, &log_info);
 
+	if (ctx->dc_bios->integrated_info->memory_type == LpDdr5MemType) {
+		dcn316_bw_params.wm_table = lpddr5_wm_table;
+	} else {
+		dcn316_bw_params.wm_table = ddr4_wm_table;
 	}
+	/* Saved clocks configured at boot for debug purposes */
+	dcn316_dump_clk_registers(&clk_mgr->base.base.boot_snapshot,
+				  &clk_mgr->base.base, &log_info);
 
 	clk_mgr->base.base.dprefclk_khz = 600000;
 	clk_mgr->base.base.dprefclk_khz = dcn316_smu_get_dpref_clk(&clk_mgr->base);
@@ -699,7 +657,7 @@ void dcn316_clk_mgr_construct(
 	if (clk_mgr->base.base.ctx->dc->debug.pstate_enabled) {
 		dcn316_get_dpm_table_from_smu(&clk_mgr->base, &smu_dpm_clks);
 
-		if (ctx->dc_bios && ctx->dc_bios->integrated_info) {
+		if (ctx->dc_bios->integrated_info) {
 			dcn316_clk_mgr_helper_populate_bw_params(
 					&clk_mgr->base,
 					ctx->dc_bios->integrated_info,
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn316/dcn316_smu.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn316/dcn316_smu.c
index 457a9254ae1c..8b82092b91cd 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn316/dcn316_smu.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn316/dcn316_smu.c
@@ -34,23 +34,21 @@
 #define MAX_INSTANCE                                        7
 #define MAX_SEGMENT                                         6
 
-struct IP_BASE_INSTANCE
-{
+struct IP_BASE_INSTANCE {
     unsigned int segment[MAX_SEGMENT];
 };
 
-struct IP_BASE
-{
+struct IP_BASE {
     struct IP_BASE_INSTANCE instance[MAX_INSTANCE];
 };
 
 static const struct IP_BASE MP0_BASE = { { { { 0x00016000, 0x00DC0000, 0x00E00000, 0x00E40000, 0x0243FC00, 0 } },
-                                        { { 0, 0, 0, 0, 0, 0 } },
-                                        { { 0, 0, 0, 0, 0, 0 } },
-                                        { { 0, 0, 0, 0, 0, 0 } },
-                                        { { 0, 0, 0, 0, 0, 0 } },
-                                        { { 0, 0, 0, 0, 0, 0 } },
-                                        { { 0, 0, 0, 0, 0, 0 } } } };
+					{ { 0, 0, 0, 0, 0, 0 } },
+					{ { 0, 0, 0, 0, 0, 0 } },
+					{ { 0, 0, 0, 0, 0, 0 } },
+					{ { 0, 0, 0, 0, 0, 0 } },
+					{ { 0, 0, 0, 0, 0, 0 } },
+					{ { 0, 0, 0, 0, 0, 0 } } } };
 
 #define REG(reg_name) \
 	(MP0_BASE.instance[0].segment[reg ## reg_name ## _BASE_IDX] + reg ## reg_name)
@@ -191,10 +189,6 @@ int dcn316_smu_set_hard_min_dcfclk(struct clk_mgr_internal *clk_mgr, int request
 			VBIOSSMC_MSG_SetHardMinDcfclkByFreq,
 			khz_to_mhz_ceil(requested_dcfclk_khz));
 
-#ifdef DBG
-	smu_print("actual_dcfclk_set_mhz %d is set to : %d\n", actual_dcfclk_set_mhz, actual_dcfclk_set_mhz * 1000);
-#endif
-
 	return actual_dcfclk_set_mhz * 1000;
 }
 
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dalsmc.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dalsmc.h
index c427be6add8a..724a508b0adb 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dalsmc.h
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dalsmc.h
@@ -55,7 +55,16 @@
 #define DALSMC_MSG_SetFclkSwitchAllow             0x11
 #define DALSMC_MSG_SetCabForUclkPstate            0x12
 #define DALSMC_MSG_SetWorstCaseUclkLatency        0x13
-#define DALSMC_Message_Count                      0x14
+#define DALSMC_MSG_SetAlwaysWaitDmcubResp         0x14
+#define DALSMC_MSG_ReturnHardMinStatus            0x15
+#define DALSMC_Message_Count                      0x16
+
+#define CHECK_HARD_MIN_CLK_DISPCLK                0x1
+#define CHECK_HARD_MIN_CLK_DPPCLK                 0x2
+#define CHECK_HARD_MIN_CLK_DPREFCLK               0x4
+#define CHECK_HARD_MIN_CLK_DCFCLK                 0x8
+#define CHECK_HARD_MIN_CLK_DTBCLK                 0x10
+#define CHECK_HARD_MIN_CLK_UCLK                   0x20
 
 typedef enum {
 	FCLK_SWITCH_DISALLOW,
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dcn32_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dcn32_clk_mgr.c
index ea753f8fa175..7da7b41bd092 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dcn32_clk_mgr.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dcn32_clk_mgr.c
@@ -25,23 +25,22 @@
 
 #include "dccg.h"
 #include "clk_mgr_internal.h"
-
 #include "dcn32/dcn32_clk_mgr_smu_msg.h"
 #include "dcn20/dcn20_clk_mgr.h"
 #include "dce100/dce_clk_mgr.h"
 #include "dcn31/dcn31_clk_mgr.h"
+#include "dcn32/dcn32_clk_mgr.h"
 #include "reg_helper.h"
 #include "core_types.h"
 #include "dm_helpers.h"
-#include "link.h"
-
+#include "link_service.h"
+#include "dc_state_priv.h"
 #include "atomfirmware.h"
-#include "smu13_driver_if.h"
+#include "dcn32_smu13_driver_if.h"
 
 #include "dcn/dcn_3_2_0_offset.h"
 #include "dcn/dcn_3_2_0_sh_mask.h"
 
-#include "dcn32/dcn32_clk_mgr.h"
 #include "dml/dcn32/dcn32_fpu.h"
 
 #define DCN_BASE__INST0_SEG1                       0x000000C0
@@ -53,6 +52,14 @@
 #define mmCLK1_CLK3_DFS_CNTL                            0x16E72
 #define mmCLK1_CLK4_DFS_CNTL                            0x16E75
 
+#define mmCLK1_CLK0_CURRENT_CNT                         0x16EE7
+#define mmCLK1_CLK1_CURRENT_CNT                         0x16EE8
+#define mmCLK1_CLK2_CURRENT_CNT                         0x16EE9
+#define mmCLK1_CLK3_CURRENT_CNT                         0x16EEA
+#define mmCLK1_CLK4_CURRENT_CNT                         0x16EEB
+
+#define mmCLK4_CLK0_CURRENT_CNT                         0x1B0C9
+
 #define CLK1_CLK_PLL_REQ__FbMult_int_MASK               0x000001ffUL
 #define CLK1_CLK_PLL_REQ__PllSpineDiv_MASK              0x0000f000UL
 #define CLK1_CLK_PLL_REQ__FbMult_frac_MASK              0xffff0000UL
@@ -156,9 +163,14 @@ void dcn32_init_clocks(struct clk_mgr *clk_mgr_base)
 {
 	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
 	unsigned int num_levels;
-	struct clk_limit_num_entries *num_entries_per_clk = &clk_mgr_base->bw_params->clk_table.num_entries_per_clk;
+	struct clk_limit_num_entries *num_entries_per_clk;
 	unsigned int i;
 
+	if (!clk_mgr_base->bw_params)
+		return;
+
+	num_entries_per_clk = &clk_mgr_base->bw_params->clk_table.num_entries_per_clk;
+
 	memset(&(clk_mgr_base->clks), 0, sizeof(struct dc_clocks));
 	clk_mgr_base->clks.p_state_change_support = true;
 	clk_mgr_base->clks.prev_p_state_change_support = true;
@@ -166,9 +178,6 @@ void dcn32_init_clocks(struct clk_mgr *clk_mgr_base)
 	clk_mgr->smu_present = false;
 	clk_mgr->dpm_present = false;
 
-	if (!clk_mgr_base->bw_params)
-		return;
-
 	if (!clk_mgr_base->force_smu_not_present && dcn30_smu_get_smu_version(clk_mgr, &clk_mgr->smu_ver))
 		clk_mgr->smu_present = true;
 
@@ -182,23 +191,42 @@ void dcn32_init_clocks(struct clk_mgr *clk_mgr_base)
 	dcn32_init_single_clock(clk_mgr, PPCLK_DCFCLK,
 			&clk_mgr_base->bw_params->clk_table.entries[0].dcfclk_mhz,
 			&num_entries_per_clk->num_dcfclk_levels);
+	clk_mgr_base->bw_params->dc_mode_limit.dcfclk_mhz = dcn30_smu_get_dc_mode_max_dpm_freq(clk_mgr, PPCLK_DCFCLK);
 
 	/* SOCCLK */
 	dcn32_init_single_clock(clk_mgr, PPCLK_SOCCLK,
 					&clk_mgr_base->bw_params->clk_table.entries[0].socclk_mhz,
 					&num_entries_per_clk->num_socclk_levels);
+	clk_mgr_base->bw_params->dc_mode_limit.socclk_mhz = dcn30_smu_get_dc_mode_max_dpm_freq(clk_mgr, PPCLK_SOCCLK);
 
 	/* DTBCLK */
-	if (!clk_mgr->base.ctx->dc->debug.disable_dtb_ref_clk_switch)
+	if (!clk_mgr->base.ctx->dc->debug.disable_dtb_ref_clk_switch) {
 		dcn32_init_single_clock(clk_mgr, PPCLK_DTBCLK,
 				&clk_mgr_base->bw_params->clk_table.entries[0].dtbclk_mhz,
 				&num_entries_per_clk->num_dtbclk_levels);
+		clk_mgr_base->bw_params->dc_mode_limit.dtbclk_mhz =
+				dcn30_smu_get_dc_mode_max_dpm_freq(clk_mgr, PPCLK_DTBCLK);
+	}
 
 	/* DISPCLK */
 	dcn32_init_single_clock(clk_mgr, PPCLK_DISPCLK,
 			&clk_mgr_base->bw_params->clk_table.entries[0].dispclk_mhz,
 			&num_entries_per_clk->num_dispclk_levels);
 	num_levels = num_entries_per_clk->num_dispclk_levels;
+	clk_mgr_base->bw_params->dc_mode_limit.dispclk_mhz = dcn30_smu_get_dc_mode_max_dpm_freq(clk_mgr, PPCLK_DISPCLK);
+	//HW recommends limit of 1950 MHz in display clock for all DCN3.2.x
+	if (clk_mgr_base->bw_params->dc_mode_limit.dispclk_mhz > 1950)
+		clk_mgr_base->bw_params->dc_mode_limit.dispclk_mhz = 1950;
+
+	/* DPPCLK */
+	dcn32_init_single_clock(clk_mgr, PPCLK_DPPCLK,
+			&clk_mgr_base->bw_params->clk_table.entries[0].dppclk_mhz,
+			&num_entries_per_clk->num_dppclk_levels);
+	num_levels = num_entries_per_clk->num_dppclk_levels;
+	clk_mgr_base->bw_params->dc_mode_limit.dppclk_mhz = dcn30_smu_get_dc_mode_max_dpm_freq(clk_mgr, PPCLK_DPPCLK);
+	//HW recommends limit of 1950 MHz in display clock for all DCN3.2.x
+	if (clk_mgr_base->bw_params->dc_mode_limit.dppclk_mhz > 1950)
+		clk_mgr_base->bw_params->dc_mode_limit.dppclk_mhz = 1950;
 
 	if (num_entries_per_clk->num_dcfclk_levels &&
 			num_entries_per_clk->num_dtbclk_levels &&
@@ -224,13 +252,41 @@ void dcn32_init_clocks(struct clk_mgr *clk_mgr_base)
 					= khz_to_mhz_ceil(clk_mgr_base->ctx->dc->debug.min_dpp_clk_khz);
 	}
 
+	for (i = 0; i < num_levels; i++)
+		if (clk_mgr_base->bw_params->clk_table.entries[i].dppclk_mhz > 1950)
+			clk_mgr_base->bw_params->clk_table.entries[i].dppclk_mhz = 1950;
+
 	/* Get UCLK, update bounding box */
 	clk_mgr_base->funcs->get_memclk_states_from_smu(clk_mgr_base);
 
-	DC_FP_START();
 	/* WM range table */
 	dcn32_build_wm_range_table(clk_mgr);
-	DC_FP_END();
+}
+
+static void dcn32_update_clocks_update_dtb_dto(struct clk_mgr_internal *clk_mgr,
+			struct dc_state *context,
+			int ref_dtbclk_khz)
+{
+	struct dccg *dccg = clk_mgr->dccg;
+	uint32_t tg_mask = 0;
+	int i;
+
+	for (i = 0; i < clk_mgr->base.ctx->dc->res_pool->pipe_count; i++) {
+		struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i];
+		struct dtbclk_dto_params dto_params = {0};
+
+		/* use mask to program DTO once per tg */
+		if (pipe_ctx->stream_res.tg &&
+				!(tg_mask & (1 << pipe_ctx->stream_res.tg->inst))) {
+			tg_mask |= (1 << pipe_ctx->stream_res.tg->inst);
+
+			dto_params.otg_inst = pipe_ctx->stream_res.tg->inst;
+			dto_params.ref_dtbclk_khz = ref_dtbclk_khz;
+
+			dccg->funcs->set_dtbclk_dto(clk_mgr->dccg, &dto_params);
+			//dccg->funcs->set_audio_dtbclk_dto(clk_mgr->dccg, &dto_params);
+		}
+	}
 }
 
 /* Since DPPCLK request to PMFW needs to be exact (due to DPP DTO programming),
@@ -262,7 +318,7 @@ void dcn32_update_clocks_update_dpp_dto(struct clk_mgr_internal *clk_mgr,
 
 	clk_mgr->dccg->ref_dppclk = clk_mgr->base.clks.dppclk_khz;
 	for (i = 0; i < clk_mgr->base.ctx->dc->res_pool->pipe_count; i++) {
-		int dpp_inst, dppclk_khz, prev_dppclk_khz;
+		int dpp_inst = 0, dppclk_khz, prev_dppclk_khz;
 
 		dppclk_khz = context->res_ctx.pipe_ctx[i].plane_res.bw.dppclk_khz;
 
@@ -320,7 +376,7 @@ static void dcn32_update_clocks_update_dentist(
 			int32_t N;
 			int32_t j;
 
-			if (!pipe_ctx->stream)
+			if (!resource_is_pipe_type(pipe_ctx, OTG_MASTER))
 				continue;
 			/* Virtual encoders don't have this function */
 			if (!stream_enc->funcs->get_fifo_cal_average_level)
@@ -345,7 +401,15 @@ static void dcn32_update_clocks_update_dentist(
 		uint32_t temp_dispclk_khz = (DENTIST_DIVIDER_RANGE_SCALE_FACTOR * clk_mgr->base.dentist_vco_freq_khz) / temp_disp_divider;
 
 		if (clk_mgr->smu_present)
-			dcn32_smu_set_hard_min_by_freq(clk_mgr, PPCLK_DISPCLK, khz_to_mhz_ceil(temp_dispclk_khz));
+			/*
+			 * SMU uses discrete dispclk presets. We applied
+			 * the same formula to increase our dppclk_khz
+			 * to the next matching discrete value. By
+			 * contract, we should use the preset dispclk
+			 * floored in Mhz to describe the intended clock.
+			 */
+			dcn32_smu_set_hard_min_by_freq(clk_mgr, PPCLK_DISPCLK,
+					khz_to_mhz_floor(temp_dispclk_khz));
 
 		if (dc->debug.override_dispclk_programming) {
 			REG_GET(DENTIST_DISPCLK_CNTL,
@@ -366,7 +430,7 @@ static void dcn32_update_clocks_update_dentist(
 			int32_t N;
 			int32_t j;
 
-			if (!pipe_ctx->stream)
+			if (!resource_is_pipe_type(pipe_ctx, OTG_MASTER))
 				continue;
 			/* Virtual encoders don't have this function */
 			if (!stream_enc->funcs->get_fifo_cal_average_level)
@@ -384,7 +448,15 @@ static void dcn32_update_clocks_update_dentist(
 
 	/* do requested DISPCLK updates*/
 	if (clk_mgr->smu_present)
-		dcn32_smu_set_hard_min_by_freq(clk_mgr, PPCLK_DISPCLK, khz_to_mhz_ceil(clk_mgr->base.clks.dispclk_khz));
+		/*
+		 * SMU uses discrete dispclk presets. We applied
+		 * the same formula to increase our dppclk_khz
+		 * to the next matching discrete value. By
+		 * contract, we should use the preset dispclk
+		 * floored in Mhz to describe the intended clock.
+		 */
+		dcn32_smu_set_hard_min_by_freq(clk_mgr, PPCLK_DISPCLK,
+				khz_to_mhz_floor(clk_mgr->base.clks.dispclk_khz));
 
 	if (dc->debug.override_dispclk_programming) {
 		REG_GET(DENTIST_DISPCLK_CNTL,
@@ -415,6 +487,136 @@ static int dcn32_get_dispclk_from_dentist(struct clk_mgr *clk_mgr_base)
 	return 0;
 }
 
+static bool dcn32_check_native_scaling(struct pipe_ctx *pipe)
+{
+	bool is_native_scaling = false;
+	int width = pipe->plane_state->src_rect.width;
+	int height = pipe->plane_state->src_rect.height;
+
+	if (pipe->stream->timing.h_addressable == width &&
+			pipe->stream->timing.v_addressable == height &&
+			pipe->plane_state->dst_rect.width == width &&
+			pipe->plane_state->dst_rect.height == height)
+		is_native_scaling = true;
+
+	return is_native_scaling;
+}
+
+static void dcn32_auto_dpm_test_log(
+		struct dc_clocks *new_clocks,
+		struct clk_mgr_internal *clk_mgr,
+		struct dc_state *context)
+{
+	unsigned int dispclk_khz_reg, dppclk_khz_reg, dprefclk_khz_reg, dcfclk_khz_reg, dtbclk_khz_reg,
+				 fclk_khz_reg, mall_ss_size_bytes;
+	int dramclk_khz_override, fclk_khz_override, num_fclk_levels;
+
+	struct pipe_ctx *pipe_ctx_list[MAX_PIPES];
+	int active_pipe_count = 0;
+
+	for (int i = 0; i < MAX_PIPES; i++) {
+		struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i];
+
+		if (pipe_ctx->stream && dc_state_get_pipe_subvp_type(context, pipe_ctx) != SUBVP_PHANTOM) {
+			pipe_ctx_list[active_pipe_count] = pipe_ctx;
+			active_pipe_count++;
+		}
+	}
+
+	msleep(5);
+
+	mall_ss_size_bytes = context->bw_ctx.bw.dcn.mall_ss_size_bytes;
+
+    dispclk_khz_reg    = REG_READ(CLK1_CLK0_CURRENT_CNT); // DISPCLK
+    dppclk_khz_reg     = REG_READ(CLK1_CLK1_CURRENT_CNT); // DPPCLK
+    dprefclk_khz_reg   = REG_READ(CLK1_CLK2_CURRENT_CNT); // DPREFCLK
+    dcfclk_khz_reg     = REG_READ(CLK1_CLK3_CURRENT_CNT); // DCFCLK
+    dtbclk_khz_reg     = REG_READ(CLK1_CLK4_CURRENT_CNT); // DTBCLK
+    fclk_khz_reg       = REG_READ(CLK4_CLK0_CURRENT_CNT); // FCLK
+
+    // Overrides for these clocks in case there is no p_state change support
+    dramclk_khz_override = new_clocks->dramclk_khz;
+    fclk_khz_override = new_clocks->fclk_khz;
+
+    num_fclk_levels = clk_mgr->base.bw_params->clk_table.num_entries_per_clk.num_fclk_levels - 1;
+
+    if (!new_clocks->p_state_change_support) {
+	    dramclk_khz_override = clk_mgr->base.bw_params->max_memclk_mhz * 1000;
+    }
+    if (!new_clocks->fclk_p_state_change_support) {
+	    fclk_khz_override = clk_mgr->base.bw_params->clk_table.entries[num_fclk_levels].fclk_mhz * 1000;
+    }
+
+	////////////////////////////////////////////////////////////////////////////
+	//	IMPORTANT: 	When adding more clocks to these logs, do NOT put a newline
+	//	 			anywhere other than at the very end of the string.
+	//
+	//	Formatting example (make sure to have " - " between each entry):
+	//
+	//				AutoDPMTest: clk1:%d - clk2:%d - clk3:%d - clk4:%d\n"
+	////////////////////////////////////////////////////////////////////////////
+	if (active_pipe_count > 0 &&
+		new_clocks->dramclk_khz > 0 &&
+		new_clocks->fclk_khz > 0 &&
+		new_clocks->dcfclk_khz > 0 &&
+		new_clocks->dppclk_khz > 0) {
+
+		uint32_t pix_clk_list[MAX_PIPES] = {0};
+		int p_state_list[MAX_PIPES] = {0};
+		int disp_src_width_list[MAX_PIPES] = {0};
+		int disp_src_height_list[MAX_PIPES] = {0};
+		uint64_t disp_src_refresh_list[MAX_PIPES] = {0};
+		bool is_scaled_list[MAX_PIPES] = {0};
+
+		for (int i = 0; i < active_pipe_count; i++) {
+			struct pipe_ctx *curr_pipe_ctx = pipe_ctx_list[i];
+			uint64_t refresh_rate;
+
+			pix_clk_list[i] = curr_pipe_ctx->stream->timing.pix_clk_100hz;
+			p_state_list[i] = curr_pipe_ctx->p_state_type;
+
+			refresh_rate = (curr_pipe_ctx->stream->timing.pix_clk_100hz * (uint64_t)100 +
+				curr_pipe_ctx->stream->timing.v_total * (uint64_t)curr_pipe_ctx->stream->timing.h_total - (uint64_t)1);
+			refresh_rate = div_u64(refresh_rate, curr_pipe_ctx->stream->timing.v_total);
+			refresh_rate = div_u64(refresh_rate, curr_pipe_ctx->stream->timing.h_total);
+			disp_src_refresh_list[i] = refresh_rate;
+
+			if (curr_pipe_ctx->plane_state) {
+				is_scaled_list[i] = !(dcn32_check_native_scaling(curr_pipe_ctx));
+				disp_src_width_list[i] = curr_pipe_ctx->plane_state->src_rect.width;
+				disp_src_height_list[i] = curr_pipe_ctx->plane_state->src_rect.height;
+			}
+		}
+
+		DC_LOG_AUTO_DPM_TEST("AutoDPMTest: dramclk:%d - fclk:%d - "
+			"dcfclk:%d - dppclk:%d - dispclk_hw:%d - "
+			"dppclk_hw:%d - dprefclk_hw:%d - dcfclk_hw:%d - "
+			"dtbclk_hw:%d - fclk_hw:%d - pix_clk_0:%d - pix_clk_1:%d - "
+			"pix_clk_2:%d - pix_clk_3:%d - mall_ss_size:%d - p_state_type_0:%d - "
+			"p_state_type_1:%d - p_state_type_2:%d - p_state_type_3:%d - "
+			"pix_width_0:%d - pix_height_0:%d - refresh_rate_0:%lld - is_scaled_0:%d - "
+			"pix_width_1:%d - pix_height_1:%d - refresh_rate_1:%lld - is_scaled_1:%d - "
+			"pix_width_2:%d - pix_height_2:%d - refresh_rate_2:%lld - is_scaled_2:%d - "
+			"pix_width_3:%d - pix_height_3:%d - refresh_rate_3:%lld - is_scaled_3:%d - LOG_END\n",
+			dramclk_khz_override,
+			fclk_khz_override,
+			new_clocks->dcfclk_khz,
+			new_clocks->dppclk_khz,
+			dispclk_khz_reg,
+			dppclk_khz_reg,
+			dprefclk_khz_reg,
+			dcfclk_khz_reg,
+			dtbclk_khz_reg,
+			fclk_khz_reg,
+			pix_clk_list[0], pix_clk_list[1], pix_clk_list[3], pix_clk_list[2],
+			mall_ss_size_bytes,
+			p_state_list[0], p_state_list[1], p_state_list[2], p_state_list[3],
+			disp_src_width_list[0], disp_src_height_list[0], disp_src_refresh_list[0], is_scaled_list[0],
+			disp_src_width_list[1], disp_src_height_list[1], disp_src_refresh_list[1], is_scaled_list[1],
+			disp_src_width_list[2], disp_src_height_list[2], disp_src_refresh_list[2], is_scaled_list[2],
+			disp_src_width_list[3], disp_src_height_list[3], disp_src_refresh_list[3], is_scaled_list[3]);
+	}
+}
 
 static void dcn32_update_clocks(struct clk_mgr *clk_mgr_base,
 			struct dc_state *context,
@@ -433,10 +635,6 @@ static void dcn32_update_clocks(struct clk_mgr *clk_mgr_base,
 	bool update_uclk = false, update_fclk = false;
 	bool p_state_change_support;
 	bool fclk_p_state_change_support;
-	int total_plane_count;
-
-	if (dc->work_arounds.skip_clock_update)
-		return;
 
 	if (clk_mgr_base->clks.dispclk_khz == 0 ||
 			(dc->debug.force_clock_mode & 0x1)) {
@@ -462,10 +660,10 @@ static void dcn32_update_clocks(struct clk_mgr *clk_mgr_base,
 
 		clk_mgr_base->clks.fclk_prev_p_state_change_support = clk_mgr_base->clks.fclk_p_state_change_support;
 
-		total_plane_count = clk_mgr_helper_get_active_plane_cnt(dc, context);
-		fclk_p_state_change_support = new_clocks->fclk_p_state_change_support || (total_plane_count == 0);
+		fclk_p_state_change_support = new_clocks->fclk_p_state_change_support;
 
-		if (should_update_pstate_support(safe_to_lower, fclk_p_state_change_support, clk_mgr_base->clks.fclk_p_state_change_support)) {
+		if (should_update_pstate_support(safe_to_lower, fclk_p_state_change_support, clk_mgr_base->clks.fclk_p_state_change_support) &&
+				!dc->work_arounds.clock_update_disable_mask.fclk) {
 			clk_mgr_base->clks.fclk_p_state_change_support = fclk_p_state_change_support;
 
 			/* To enable FCLK P-state switching, send FCLK_PSTATE_SUPPORTED message to PMFW */
@@ -479,12 +677,14 @@ static void dcn32_update_clocks(struct clk_mgr *clk_mgr_base,
 			new_clocks->dcfclk_khz = (new_clocks->dcfclk_khz > (dc->debug.force_min_dcfclk_mhz * 1000)) ?
 					new_clocks->dcfclk_khz : (dc->debug.force_min_dcfclk_mhz * 1000);
 
-		if (should_set_clock(safe_to_lower, new_clocks->dcfclk_khz, clk_mgr_base->clks.dcfclk_khz)) {
+		if (should_set_clock(safe_to_lower, new_clocks->dcfclk_khz, clk_mgr_base->clks.dcfclk_khz) &&
+				!dc->work_arounds.clock_update_disable_mask.dcfclk) {
 			clk_mgr_base->clks.dcfclk_khz = new_clocks->dcfclk_khz;
 			dcn32_smu_set_hard_min_by_freq(clk_mgr, PPCLK_DCFCLK, khz_to_mhz_ceil(clk_mgr_base->clks.dcfclk_khz));
 		}
 
-		if (should_set_clock(safe_to_lower, new_clocks->dcfclk_deep_sleep_khz, clk_mgr_base->clks.dcfclk_deep_sleep_khz)) {
+		if (should_set_clock(safe_to_lower, new_clocks->dcfclk_deep_sleep_khz, clk_mgr_base->clks.dcfclk_deep_sleep_khz) &&
+				!dc->work_arounds.clock_update_disable_mask.dcfclk_ds) {
 			clk_mgr_base->clks.dcfclk_deep_sleep_khz = new_clocks->dcfclk_deep_sleep_khz;
 			dcn30_smu_set_min_deep_sleep_dcef_clk(clk_mgr, khz_to_mhz_ceil(clk_mgr_base->clks.dcfclk_deep_sleep_khz));
 		}
@@ -502,37 +702,63 @@ static void dcn32_update_clocks(struct clk_mgr *clk_mgr_base,
 			dcn32_smu_send_cab_for_uclk_message(clk_mgr, clk_mgr_base->clks.num_ways);
 		}
 
-
-		p_state_change_support = new_clocks->p_state_change_support || (total_plane_count == 0);
-		if (should_update_pstate_support(safe_to_lower, p_state_change_support, clk_mgr_base->clks.p_state_change_support)) {
+		p_state_change_support = new_clocks->p_state_change_support;
+		if (should_update_pstate_support(safe_to_lower, p_state_change_support, clk_mgr_base->clks.p_state_change_support) &&
+				!dc->work_arounds.clock_update_disable_mask.uclk) {
 			clk_mgr_base->clks.p_state_change_support = p_state_change_support;
 
 			/* to disable P-State switching, set UCLK min = max */
-			if (!clk_mgr_base->clks.p_state_change_support)
-				dcn32_smu_set_hard_min_by_freq(clk_mgr, PPCLK_UCLK,
-						clk_mgr_base->bw_params->clk_table.entries[clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_memclk_levels - 1].memclk_mhz);
+			if (!clk_mgr_base->clks.p_state_change_support) {
+				if (dc->clk_mgr->dc_mode_softmax_enabled) {
+					/* On DCN32x we will never have the functional UCLK min above the softmax
+					 * since we calculate mode support based on softmax being the max UCLK
+					 * frequency.
+					 */
+					if (dc->debug.disable_dc_mode_overwrite) {
+						dcn30_smu_set_hard_max_by_freq(clk_mgr, PPCLK_UCLK, dc->clk_mgr->bw_params->max_memclk_mhz);
+						dcn32_smu_set_hard_min_by_freq(clk_mgr, PPCLK_UCLK, dc->clk_mgr->bw_params->max_memclk_mhz);
+					} else
+						dcn32_smu_set_hard_min_by_freq(clk_mgr, PPCLK_UCLK,
+								dc->clk_mgr->bw_params->dc_mode_softmax_memclk);
+				} else {
+					dcn32_smu_set_hard_min_by_freq(clk_mgr, PPCLK_UCLK, dc->clk_mgr->bw_params->max_memclk_mhz);
+				}
+			}
 		}
 
+		if (context->bw_ctx.bw.dcn.clk.fw_based_mclk_switching)
+			dcn32_smu_wait_for_dmub_ack_mclk(clk_mgr, true);
+		else
+			dcn32_smu_wait_for_dmub_ack_mclk(clk_mgr, false);
+
 		/* Always update saved value, even if new value not set due to P-State switching unsupported. Also check safe_to_lower for FCLK */
 		if (safe_to_lower && (clk_mgr_base->clks.fclk_p_state_change_support != clk_mgr_base->clks.fclk_prev_p_state_change_support)) {
 			update_fclk = true;
 		}
 
-		if (clk_mgr_base->ctx->dce_version != DCN_VERSION_3_21 && !clk_mgr_base->clks.fclk_p_state_change_support && update_fclk) {
+		if (clk_mgr_base->ctx->dce_version != DCN_VERSION_3_21 && !clk_mgr_base->clks.fclk_p_state_change_support && update_fclk &&
+				!dc->work_arounds.clock_update_disable_mask.fclk) {
 			/* Handle code for sending a message to PMFW that FCLK P-state change is not supported */
 			dcn32_smu_send_fclk_pstate_message(clk_mgr, FCLK_PSTATE_NOTSUPPORTED);
 		}
 
 		/* Always update saved value, even if new value not set due to P-State switching unsupported */
-		if (should_set_clock(safe_to_lower, new_clocks->dramclk_khz, clk_mgr_base->clks.dramclk_khz)) {
+		if (should_set_clock(safe_to_lower, new_clocks->dramclk_khz, clk_mgr_base->clks.dramclk_khz) &&
+				!dc->work_arounds.clock_update_disable_mask.uclk) {
 			clk_mgr_base->clks.dramclk_khz = new_clocks->dramclk_khz;
 			update_uclk = true;
 		}
 
 		/* set UCLK to requested value if P-State switching is supported, or to re-enable P-State switching */
 		if (clk_mgr_base->clks.p_state_change_support &&
-				(update_uclk || !clk_mgr_base->clks.prev_p_state_change_support))
+				(update_uclk || !clk_mgr_base->clks.prev_p_state_change_support) &&
+				!dc->work_arounds.clock_update_disable_mask.uclk) {
+			if (dc->clk_mgr->dc_mode_softmax_enabled && dc->debug.disable_dc_mode_overwrite)
+				dcn30_smu_set_hard_max_by_freq(clk_mgr, PPCLK_UCLK,
+						max((int)dc->clk_mgr->bw_params->dc_mode_softmax_memclk, khz_to_mhz_ceil(clk_mgr_base->clks.dramclk_khz)));
+
 			dcn32_smu_set_hard_min_by_freq(clk_mgr, PPCLK_UCLK, khz_to_mhz_ceil(clk_mgr_base->clks.dramclk_khz));
+		}
 
 		if (clk_mgr_base->clks.num_ways != new_clocks->num_ways &&
 				clk_mgr_base->clks.num_ways > new_clocks->num_ways) {
@@ -549,7 +775,15 @@ static void dcn32_update_clocks(struct clk_mgr *clk_mgr_base,
 		clk_mgr_base->clks.dppclk_khz = new_clocks->dppclk_khz;
 
 		if (clk_mgr->smu_present && !dpp_clock_lowered)
-			dcn32_smu_set_hard_min_by_freq(clk_mgr, PPCLK_DPPCLK, khz_to_mhz_ceil(clk_mgr_base->clks.dppclk_khz));
+			/*
+			 * SMU uses discrete dppclk presets. We applied
+			 * the same formula to increase our dppclk_khz
+			 * to the next matching discrete value. By
+			 * contract, we should use the preset dppclk
+			 * floored in Mhz to describe the intended clock.
+			 */
+			dcn32_smu_set_hard_min_by_freq(clk_mgr, PPCLK_DPPCLK,
+					khz_to_mhz_floor(clk_mgr_base->clks.dppclk_khz));
 
 		update_dppclk = true;
 	}
@@ -570,6 +804,8 @@ static void dcn32_update_clocks(struct clk_mgr *clk_mgr_base,
 		/* DCCG requires KHz precision for DTBCLK */
 		clk_mgr_base->clks.ref_dtbclk_khz =
 				dcn32_smu_set_hard_min_by_freq(clk_mgr, PPCLK_DTBCLK, khz_to_mhz_ceil(new_clocks->ref_dtbclk_khz));
+
+		dcn32_update_clocks_update_dtb_dto(clk_mgr, context, clk_mgr_base->clks.ref_dtbclk_khz);
 	}
 
 	if (dc->config.forced_clocks == false || (force_reset && safe_to_lower)) {
@@ -578,7 +814,15 @@ static void dcn32_update_clocks(struct clk_mgr *clk_mgr_base,
 			dcn32_update_clocks_update_dpp_dto(clk_mgr, context, safe_to_lower);
 			dcn32_update_clocks_update_dentist(clk_mgr, context);
 			if (clk_mgr->smu_present)
-				dcn32_smu_set_hard_min_by_freq(clk_mgr, PPCLK_DPPCLK, khz_to_mhz_ceil(clk_mgr_base->clks.dppclk_khz));
+				/*
+				 * SMU uses discrete dppclk presets. We applied
+				 * the same formula to increase our dppclk_khz
+				 * to the next matching discrete value. By
+				 * contract, we should use the preset dppclk
+				 * floored in Mhz to describe the intended clock.
+				 */
+				dcn32_smu_set_hard_min_by_freq(clk_mgr, PPCLK_DPPCLK,
+						khz_to_mhz_floor(clk_mgr_base->clks.dppclk_khz));
 		} else {
 			/* if clock is being raised, increase refclk before lowering DTO */
 			if (update_dppclk || update_dispclk)
@@ -595,6 +839,10 @@ static void dcn32_update_clocks(struct clk_mgr *clk_mgr_base,
 		/*update dmcu for wait_loop count*/
 		dmcu->funcs->set_psr_wait_loop(dmcu,
 				clk_mgr_base->clks.dispclk_khz / 1000 / 7);
+
+	if (dc->config.enable_auto_dpm_test_logs) {
+	    dcn32_auto_dpm_test_log(new_clocks, clk_mgr, context);
+	}
 }
 
 static uint32_t dcn32_get_vco_frequency_from_reg(struct clk_mgr_internal *clk_mgr)
@@ -756,7 +1004,7 @@ static void dcn32_set_hard_min_memclk(struct clk_mgr *clk_mgr_base, bool current
 					khz_to_mhz_ceil(clk_mgr_base->clks.dramclk_khz));
 		else
 			dcn32_smu_set_hard_min_by_freq(clk_mgr, PPCLK_UCLK,
-					clk_mgr_base->bw_params->clk_table.entries[clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_memclk_levels - 1].memclk_mhz);
+					clk_mgr_base->bw_params->max_memclk_mhz);
 	} else {
 		dcn32_smu_set_hard_min_by_freq(clk_mgr, PPCLK_UCLK,
 				clk_mgr_base->bw_params->clk_table.entries[0].memclk_mhz);
@@ -771,8 +1019,7 @@ static void dcn32_set_hard_max_memclk(struct clk_mgr *clk_mgr_base)
 	if (!clk_mgr->smu_present)
 		return;
 
-	dcn30_smu_set_hard_max_by_freq(clk_mgr, PPCLK_UCLK,
-			clk_mgr_base->bw_params->clk_table.entries[clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_memclk_levels - 1].memclk_mhz);
+	dcn30_smu_set_hard_max_by_freq(clk_mgr, PPCLK_UCLK, clk_mgr_base->bw_params->max_memclk_mhz);
 }
 
 /* Get current memclk states, update bounding box */
@@ -789,6 +1036,8 @@ static void dcn32_get_memclk_states_from_smu(struct clk_mgr *clk_mgr_base)
 	dcn32_init_single_clock(clk_mgr, PPCLK_UCLK,
 			&clk_mgr_base->bw_params->clk_table.entries[0].memclk_mhz,
 			&num_entries_per_clk->num_memclk_levels);
+	clk_mgr_base->bw_params->dc_mode_limit.memclk_mhz = dcn30_smu_get_dc_mode_max_dpm_freq(clk_mgr, PPCLK_UCLK);
+	clk_mgr_base->bw_params->dc_mode_softmax_memclk = clk_mgr_base->bw_params->dc_mode_limit.memclk_mhz;
 
 	/* memclk must have at least one level */
 	num_entries_per_clk->num_memclk_levels = num_entries_per_clk->num_memclk_levels ? num_entries_per_clk->num_memclk_levels : 1;
@@ -796,13 +1045,12 @@ static void dcn32_get_memclk_states_from_smu(struct clk_mgr *clk_mgr_base)
 	dcn32_init_single_clock(clk_mgr, PPCLK_FCLK,
 			&clk_mgr_base->bw_params->clk_table.entries[0].fclk_mhz,
 			&num_entries_per_clk->num_fclk_levels);
+	clk_mgr_base->bw_params->dc_mode_limit.fclk_mhz = dcn30_smu_get_dc_mode_max_dpm_freq(clk_mgr, PPCLK_FCLK);
 
-	if (num_entries_per_clk->num_memclk_levels >= num_entries_per_clk->num_fclk_levels) {
-		num_levels = num_entries_per_clk->num_memclk_levels;
-	} else {
-		num_levels = num_entries_per_clk->num_fclk_levels;
-	}
+	num_levels = max(num_entries_per_clk->num_memclk_levels, num_entries_per_clk->num_fclk_levels);
 
+	clk_mgr_base->bw_params->max_memclk_mhz =
+			clk_mgr_base->bw_params->clk_table.entries[num_entries_per_clk->num_memclk_levels - 1].memclk_mhz;
 	clk_mgr_base->bw_params->clk_table.num_entries = num_levels ? num_levels : 1;
 
 	if (clk_mgr->dpm_present && !num_levels)
@@ -855,6 +1103,25 @@ static bool dcn32_is_smu_present(struct clk_mgr *clk_mgr_base)
 	return clk_mgr->smu_present;
 }
 
+static void dcn32_set_max_memclk(struct clk_mgr *clk_mgr_base, unsigned int memclk_mhz)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+
+	if (!clk_mgr->smu_present)
+		return;
+
+	dcn30_smu_set_hard_max_by_freq(clk_mgr, PPCLK_UCLK, memclk_mhz);
+}
+
+static void dcn32_set_min_memclk(struct clk_mgr *clk_mgr_base, unsigned int memclk_mhz)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+
+	if (!clk_mgr->smu_present)
+		return;
+
+	dcn32_smu_set_hard_min_by_freq(clk_mgr, PPCLK_UCLK, memclk_mhz);
+}
 
 static struct clk_mgr_funcs dcn32_funcs = {
 		.get_dp_ref_clk_frequency = dce12_get_dp_ref_freq_khz,
@@ -865,6 +1132,8 @@ static struct clk_mgr_funcs dcn32_funcs = {
 		.notify_wm_ranges = dcn32_notify_wm_ranges,
 		.set_hard_min_memclk = dcn32_set_hard_min_memclk,
 		.set_hard_max_memclk = dcn32_set_hard_max_memclk,
+		.set_max_memclk = dcn32_set_max_memclk,
+		.set_min_memclk = dcn32_set_min_memclk,
 		.get_memclk_states_from_smu = dcn32_get_memclk_states_from_smu,
 		.are_clock_states_equal = dcn32_are_clock_states_equal,
 		.enable_pme_wa = dcn32_enable_pme_wa,
@@ -878,6 +1147,8 @@ void dcn32_clk_mgr_construct(
 		struct pp_smu_funcs *pp_smu,
 		struct dccg *dccg)
 {
+	struct clk_log_info log_info = {0};
+
 	clk_mgr->base.ctx = ctx;
 	clk_mgr->base.funcs = &dcn32_funcs;
 	if (ASICREV_IS_GC_11_0_2(clk_mgr->base.ctx->asic_id.hw_internal_rev)) {
@@ -911,6 +1182,7 @@ void dcn32_clk_mgr_construct(
 			clk_mgr->base.clks.ref_dtbclk_khz = 268750;
 	}
 
+
 	/* integer part is now VCO frequency in kHz */
 	clk_mgr->base.dentist_vco_freq_khz = dcn32_get_vco_frequency_from_reg(clk_mgr);
 
@@ -918,6 +1190,8 @@ void dcn32_clk_mgr_construct(
 	if (clk_mgr->base.dentist_vco_freq_khz == 0)
 		clk_mgr->base.dentist_vco_freq_khz = 4300000; /* Updated as per HW docs */
 
+	dcn32_dump_clk_registers(&clk_mgr->base.boot_snapshot, &clk_mgr->base, &log_info);
+
 	if (ctx->dc->debug.disable_dtb_ref_clk_switch &&
 			clk_mgr->base.clks.ref_dtbclk_khz != clk_mgr->base.boot_snapshot.dtbclk) {
 		clk_mgr->base.clks.ref_dtbclk_khz = clk_mgr->base.boot_snapshot.dtbclk;
@@ -933,11 +1207,19 @@ void dcn32_clk_mgr_construct(
 	clk_mgr->smu_present = false;
 
 	clk_mgr->base.bw_params = kzalloc(sizeof(*clk_mgr->base.bw_params), GFP_KERNEL);
+	if (!clk_mgr->base.bw_params) {
+		BREAK_TO_DEBUGGER();
+		return;
+	}
 
 	/* need physical address of table to give to PMFW */
 	clk_mgr->wm_range_table = dm_helpers_allocate_gpu_mem(clk_mgr->base.ctx,
 			DC_MEM_ALLOC_TYPE_GART, sizeof(WatermarksExternal_t),
 			&clk_mgr->wm_range_table_addr);
+	if (!clk_mgr->wm_range_table) {
+		BREAK_TO_DEBUGGER();
+		return;
+	}
 }
 
 void dcn32_clk_mgr_destroy(struct clk_mgr_internal *clk_mgr)
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dcn32_clk_mgr_smu_msg.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dcn32_clk_mgr_smu_msg.c
index fb524fe4ab26..5d80fdf63ffc 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dcn32_clk_mgr_smu_msg.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dcn32_clk_mgr_smu_msg.c
@@ -28,7 +28,7 @@
 #include "clk_mgr_internal.h"
 #include "reg_helper.h"
 #include "dalsmc.h"
-#include "smu13_driver_if.h"
+#include "dcn32_smu13_driver_if.h"
 
 #define mmDAL_MSG_REG  0x1628A
 #define mmDAL_ARG_REG  0x16273
@@ -49,6 +49,7 @@
  */
 static uint32_t dcn32_smu_wait_for_response(struct clk_mgr_internal *clk_mgr, unsigned int delay_us, unsigned int max_retries)
 {
+	const uint32_t initial_max_retries = max_retries;
 	uint32_t reg = 0;
 
 	do {
@@ -62,6 +63,9 @@ static uint32_t dcn32_smu_wait_for_response(struct clk_mgr_internal *clk_mgr, un
 			udelay(delay_us);
 	} while (max_retries--);
 
+	TRACE_SMU_MSG_DELAY(0, 0, delay_us * (initial_max_retries - max_retries), clk_mgr->base.ctx);
+
+
 	return reg;
 }
 
@@ -79,6 +83,8 @@ static bool dcn32_smu_send_msg_with_param(struct clk_mgr_internal *clk_mgr, uint
 	/* Trigger the message transaction by writing the message ID */
 	REG_WRITE(DAL_MSG_REG, msg_id);
 
+	TRACE_SMU_MSG(msg_id, param_in, clk_mgr->base.ctx);
+
 	/* Wait for response */
 	if (dcn32_smu_wait_for_response(clk_mgr, 10, 200000) == DALSMC_Result_OK) {
 		if (param_out)
@@ -90,6 +96,68 @@ static bool dcn32_smu_send_msg_with_param(struct clk_mgr_internal *clk_mgr, uint
 	return false;
 }
 
+/*
+ * Use these functions to return back delay information so we can aggregate the total
+ *  delay when requesting hardmin clk
+ *
+ * dcn32_smu_wait_for_response_delay
+ * dcn32_smu_send_msg_with_param_delay
+ *
+ */
+static uint32_t dcn32_smu_wait_for_response_delay(struct clk_mgr_internal *clk_mgr, unsigned int delay_us, unsigned int max_retries, unsigned int *total_delay_us)
+{
+	uint32_t reg = 0;
+	*total_delay_us = 0;
+
+	do {
+		reg = REG_READ(DAL_RESP_REG);
+		if (reg)
+			break;
+
+		if (delay_us >= 1000)
+			msleep(delay_us/1000);
+		else if (delay_us > 0)
+			udelay(delay_us);
+		*total_delay_us += delay_us;
+	} while (max_retries--);
+
+	TRACE_SMU_MSG_DELAY(0, 0, *total_delay_us, clk_mgr->base.ctx);
+
+	return reg;
+}
+
+static bool dcn32_smu_send_msg_with_param_delay(struct clk_mgr_internal *clk_mgr, uint32_t msg_id, uint32_t param_in, uint32_t *param_out, unsigned int *total_delay_us)
+{
+	unsigned int delay1_us, delay2_us;
+	*total_delay_us = 0;
+
+	/* Wait for response register to be ready */
+	dcn32_smu_wait_for_response_delay(clk_mgr, 10, 200000, &delay1_us);
+
+	/* Clear response register */
+	REG_WRITE(DAL_RESP_REG, 0);
+
+	/* Set the parameter register for the SMU message */
+	REG_WRITE(DAL_ARG_REG, param_in);
+
+	/* Trigger the message transaction by writing the message ID */
+	REG_WRITE(DAL_MSG_REG, msg_id);
+
+	TRACE_SMU_MSG(msg_id, param_in, clk_mgr->base.ctx);
+
+	/* Wait for response */
+	if (dcn32_smu_wait_for_response_delay(clk_mgr, 10, 200000, &delay2_us) == DALSMC_Result_OK) {
+		if (param_out)
+			*param_out = REG_READ(DAL_ARG_REG);
+
+		*total_delay_us = delay1_us + delay2_us;
+		return true;
+	}
+
+	*total_delay_us = delay1_us + 2000000;
+	return false;
+}
+
 void dcn32_smu_send_fclk_pstate_message(struct clk_mgr_internal *clk_mgr, bool enable)
 {
 	smu_print("FCLK P-state support value is : %d\n", enable);
@@ -122,10 +190,98 @@ void dcn32_smu_set_pme_workaround(struct clk_mgr_internal *clk_mgr)
 		DALSMC_MSG_BacoAudioD3PME, 0, NULL);
 }
 
+/* Check PMFW version if it supports ReturnHardMinStatus message */
+static bool dcn32_get_hard_min_status_supported(struct clk_mgr_internal *clk_mgr)
+{
+	if (ASICREV_IS_GC_11_0_0(clk_mgr->base.ctx->asic_id.hw_internal_rev)) {
+		if (clk_mgr->smu_ver >= 0x4e6a00)
+			return true;
+	} else if (ASICREV_IS_GC_11_0_2(clk_mgr->base.ctx->asic_id.hw_internal_rev)) {
+		if (clk_mgr->smu_ver >= 0x524e00)
+			return true;
+	} else { /* ASICREV_IS_GC_11_0_3 */
+		if (clk_mgr->smu_ver >= 0x503900)
+			return true;
+	}
+	return false;
+}
+
+/* Returns the clocks which were fulfilled by the DAL hard min arbiter in PMFW */
+static unsigned int dcn32_smu_get_hard_min_status(struct clk_mgr_internal *clk_mgr, bool *no_timeout, unsigned int *total_delay_us)
+{
+	uint32_t response = 0;
+
+	/* bits 23:16 for clock type, lower 16 bits for frequency in MHz */
+	uint32_t param = 0;
+
+	*no_timeout = dcn32_smu_send_msg_with_param_delay(clk_mgr,
+			DALSMC_MSG_ReturnHardMinStatus, param, &response, total_delay_us);
+
+	smu_print("SMU Get hard min status: no_timeout %d delay %d us clk bits %x\n",
+		*no_timeout, *total_delay_us, response);
+
+	return response;
+}
+
+static bool dcn32_smu_wait_get_hard_min_status(struct clk_mgr_internal *clk_mgr,
+	uint32_t clk)
+{
+	int readDalHardMinClkBits, checkDalHardMinClkBits;
+	unsigned int total_delay_us, read_total_delay_us;
+	bool no_timeout, hard_min_done;
+
+	static unsigned int cur_wait_get_hard_min_max_us;
+	static unsigned int cur_wait_get_hard_min_max_timeouts;
+
+	checkDalHardMinClkBits = CHECK_HARD_MIN_CLK_DPREFCLK;
+	if (clk == PPCLK_DISPCLK)
+		checkDalHardMinClkBits |= CHECK_HARD_MIN_CLK_DISPCLK;
+	if (clk == PPCLK_DPPCLK)
+		checkDalHardMinClkBits |= CHECK_HARD_MIN_CLK_DPPCLK;
+	if (clk == PPCLK_DCFCLK)
+		checkDalHardMinClkBits |= CHECK_HARD_MIN_CLK_DCFCLK;
+	if (clk == PPCLK_DTBCLK)
+		checkDalHardMinClkBits |= CHECK_HARD_MIN_CLK_DTBCLK;
+	if (clk == PPCLK_UCLK)
+		checkDalHardMinClkBits |= CHECK_HARD_MIN_CLK_UCLK;
+
+	if (checkDalHardMinClkBits == CHECK_HARD_MIN_CLK_DPREFCLK)
+		return 0;
+
+	total_delay_us = 0;
+	hard_min_done = false;
+	while (1) {
+		readDalHardMinClkBits = dcn32_smu_get_hard_min_status(clk_mgr, &no_timeout, &read_total_delay_us);
+		total_delay_us += read_total_delay_us;
+		if (checkDalHardMinClkBits == (readDalHardMinClkBits & checkDalHardMinClkBits)) {
+			hard_min_done = true;
+			break;
+		}
+
+
+		if (total_delay_us >= 2000000) {
+			cur_wait_get_hard_min_max_timeouts++;
+			smu_print("SMU Wait get hard min status: %d timeouts\n", cur_wait_get_hard_min_max_timeouts);
+			break;
+		}
+		msleep(1);
+		total_delay_us += 1000;
+	}
+
+	if (total_delay_us > cur_wait_get_hard_min_max_us)
+		cur_wait_get_hard_min_max_us = total_delay_us;
+
+	smu_print("SMU Wait get hard min status: no_timeout %d, delay %d us, max %d us, read %x, check %x\n",
+		no_timeout, total_delay_us, cur_wait_get_hard_min_max_us, readDalHardMinClkBits, checkDalHardMinClkBits);
+
+	return hard_min_done;
+}
+
 /* Returns the actual frequency that was set in MHz, 0 on failure */
 unsigned int dcn32_smu_set_hard_min_by_freq(struct clk_mgr_internal *clk_mgr, uint32_t clk, uint16_t freq_mhz)
 {
 	uint32_t response = 0;
+	bool hard_min_done = false;
 
 	/* bits 23:16 for clock type, lower 16 bits for frequency in MHz */
 	uint32_t param = (clk << 16) | freq_mhz;
@@ -133,9 +289,20 @@ unsigned int dcn32_smu_set_hard_min_by_freq(struct clk_mgr_internal *clk_mgr, ui
 	smu_print("SMU Set hard min by freq: clk = %d, freq_mhz = %d MHz\n", clk, freq_mhz);
 
 	dcn32_smu_send_msg_with_param(clk_mgr,
-			DALSMC_MSG_SetHardMinByFreq, param, &response);
+		DALSMC_MSG_SetHardMinByFreq, param, &response);
 
-	smu_print("SMU Frequency set = %d KHz\n", response);
+	if (dcn32_get_hard_min_status_supported(clk_mgr)) {
+		hard_min_done = dcn32_smu_wait_get_hard_min_status(clk_mgr, clk);
+		smu_print("SMU Frequency set = %d KHz hard_min_done %d\n", response, hard_min_done);
+	} else
+		smu_print("SMU Frequency set = %d KHz\n", response);
 
 	return response;
 }
+
+void dcn32_smu_wait_for_dmub_ack_mclk(struct clk_mgr_internal *clk_mgr, bool enable)
+{
+	smu_print("PMFW to wait for DMCUB ack for MCLK : %d\n", enable);
+
+	dcn32_smu_send_msg_with_param(clk_mgr, 0x14, enable ? 1 : 0, NULL);
+}
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dcn32_clk_mgr_smu_msg.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dcn32_clk_mgr_smu_msg.h
index a68038a41972..5c44ab0e8667 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dcn32_clk_mgr_smu_msg.h
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dcn32_clk_mgr_smu_msg.h
@@ -36,12 +36,11 @@
 #define DALSMC_MSG_SetCabForUclkPstate	0x12
 #define DALSMC_Result_OK				0x1
 
-void
-dcn32_smu_send_fclk_pstate_message(struct clk_mgr_internal *clk_mgr, bool enable);
-void dcn32_smu_transfer_wm_table_dram_2_smu(struct clk_mgr_internal *clk_mgr);
-void dcn32_smu_set_pme_workaround(struct clk_mgr_internal *clk_mgr);
+void dcn32_smu_send_fclk_pstate_message(struct clk_mgr_internal *clk_mgr, bool enable);
 void dcn32_smu_send_cab_for_uclk_message(struct clk_mgr_internal *clk_mgr, unsigned int num_ways);
 void dcn32_smu_transfer_wm_table_dram_2_smu(struct clk_mgr_internal *clk_mgr);
+void dcn32_smu_set_pme_workaround(struct clk_mgr_internal *clk_mgr);
 unsigned int dcn32_smu_set_hard_min_by_freq(struct clk_mgr_internal *clk_mgr, uint32_t clk, uint16_t freq_mhz);
+void dcn32_smu_wait_for_dmub_ack_mclk(struct clk_mgr_internal *clk_mgr, bool enable);
 
 #endif /* __DCN32_CLK_MGR_SMU_MSG_H_ */
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dcn32_smu13_driver_if.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dcn32_smu13_driver_if.h
index d3d5a8caccf8..8d54865bbd5d 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dcn32_smu13_driver_if.h
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/dcn32_smu13_driver_if.h
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// This is a stripped-down version of the smu13_driver_if.h file for the relevant DAL interfaces.
+/* Copyright © 2022-2024 Advanced Micro Devices, Inc. All rights reserved. */
 
 #define SMU13_DRIVER_IF_VERSION  0x18
 
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/smu13_driver_if.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/smu13_driver_if.h
deleted file mode 100644
index deeb85047e7b..000000000000
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn32/smu13_driver_if.h
+++ /dev/null
@@ -1,108 +0,0 @@
-/*
- * Copyright 2021 Advanced Micro Devices, Inc.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
- * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
- * OTHER DEALINGS IN THE SOFTWARE.
- *
- * Authors: AMD
- *
- */
-#ifndef SMU13_DRIVER_IF_DCN32_H
-#define SMU13_DRIVER_IF_DCN32_H
-
-// *** IMPORTANT ***
-// PMFW TEAM: Always increment the interface version on any change to this file
-#define SMU13_DRIVER_IF_VERSION  0x18
-
-//Only Clks that have DPM descriptors are listed here
-typedef enum {
-  PPCLK_GFXCLK = 0,
-  PPCLK_SOCCLK,
-  PPCLK_UCLK,
-  PPCLK_FCLK,
-  PPCLK_DCLK_0,
-  PPCLK_VCLK_0,
-  PPCLK_DCLK_1,
-  PPCLK_VCLK_1,
-  PPCLK_DISPCLK,
-  PPCLK_DPPCLK,
-  PPCLK_DPREFCLK,
-  PPCLK_DCFCLK,
-  PPCLK_DTBCLK,
-  PPCLK_COUNT,
-} PPCLK_e;
-
-typedef enum {
-  UCLK_DIV_BY_1 = 0,
-  UCLK_DIV_BY_2,
-  UCLK_DIV_BY_4,
-  UCLK_DIV_BY_8,
-} UCLK_DIV_e;
-
-typedef struct {
-  uint8_t  WmSetting;
-  uint8_t  Flags;
-  uint8_t  Padding[2];
-
-} WatermarkRowGeneric_t;
-
-#define NUM_WM_RANGES 4
-
-typedef enum {
-  WATERMARKS_CLOCK_RANGE = 0,
-  WATERMARKS_DUMMY_PSTATE,
-  WATERMARKS_MALL,
-  WATERMARKS_COUNT,
-} WATERMARKS_FLAGS_e;
-
-typedef struct {
-  // Watermarks
-  WatermarkRowGeneric_t WatermarkRow[NUM_WM_RANGES];
-} Watermarks_t;
-
-typedef struct {
-  Watermarks_t Watermarks;
-  uint32_t  Spare[16];
-
-  uint32_t     MmHubPadding[8]; // SMU internal use
-} WatermarksExternal_t;
-
-// These defines are used with the following messages:
-// SMC_MSG_TransferTableDram2Smu
-// SMC_MSG_TransferTableSmu2Dram
-
-// Table transfer status
-#define TABLE_TRANSFER_OK         0x0
-#define TABLE_TRANSFER_FAILED     0xFF
-#define TABLE_TRANSFER_PENDING    0xAB
-
-// Table types
-#define TABLE_PMFW_PPTABLE            0
-#define TABLE_COMBO_PPTABLE           1
-#define TABLE_WATERMARKS              2
-#define TABLE_AVFS_PSM_DEBUG          3
-#define TABLE_PMSTATUSLOG             4
-#define TABLE_SMU_METRICS             5
-#define TABLE_DRIVER_SMU_CONFIG       6
-#define TABLE_ACTIVITY_MONITOR_COEFF  7
-#define TABLE_OVERDRIVE               8
-#define TABLE_I2C_COMMANDS            9
-#define TABLE_DRIVER_INFO             10
-#define TABLE_COUNT                   11
-
-#endif
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn351_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn351_clk_mgr.c
new file mode 100644
index 000000000000..4607eff07253
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn351_clk_mgr.c
@@ -0,0 +1,141 @@
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: AMD
+ *
+ */
+
+#include "core_types.h"
+#include "dcn35_clk_mgr.h"
+
+#define DCN_BASE__INST0_SEG1 0x000000C0
+#define mmCLK1_CLK_PLL_REQ 0x16E37
+
+#define mmCLK1_CLK0_DFS_CNTL 0x16E69
+#define mmCLK1_CLK1_DFS_CNTL 0x16E6C
+#define mmCLK1_CLK2_DFS_CNTL 0x16E6F
+#define mmCLK1_CLK3_DFS_CNTL 0x16E72
+#define mmCLK1_CLK4_DFS_CNTL 0x16E75
+#define mmCLK1_CLK5_DFS_CNTL 0x16E78
+
+#define mmCLK1_CLK0_CURRENT_CNT 0x16EFC
+#define mmCLK1_CLK1_CURRENT_CNT 0x16EFD
+#define mmCLK1_CLK2_CURRENT_CNT 0x16EFE
+#define mmCLK1_CLK3_CURRENT_CNT 0x16EFF
+#define mmCLK1_CLK4_CURRENT_CNT 0x16F00
+#define mmCLK1_CLK5_CURRENT_CNT 0x16F01
+
+#define mmCLK1_CLK0_BYPASS_CNTL 0x16E8A
+#define mmCLK1_CLK1_BYPASS_CNTL 0x16E93
+#define mmCLK1_CLK2_BYPASS_CNTL 0x16E9C
+#define mmCLK1_CLK3_BYPASS_CNTL 0x16EA5
+#define mmCLK1_CLK4_BYPASS_CNTL 0x16EAE
+#define mmCLK1_CLK5_BYPASS_CNTL 0x16EB7
+
+#define mmCLK1_CLK0_DS_CNTL 0x16E83
+#define mmCLK1_CLK1_DS_CNTL 0x16E8C
+#define mmCLK1_CLK2_DS_CNTL 0x16E95
+#define mmCLK1_CLK3_DS_CNTL 0x16E9E
+#define mmCLK1_CLK4_DS_CNTL 0x16EA7
+#define mmCLK1_CLK5_DS_CNTL 0x16EB0
+
+#define mmCLK1_CLK0_ALLOW_DS 0x16E84
+#define mmCLK1_CLK1_ALLOW_DS 0x16E8D
+#define mmCLK1_CLK2_ALLOW_DS 0x16E96
+#define mmCLK1_CLK3_ALLOW_DS 0x16E9F
+#define mmCLK1_CLK4_ALLOW_DS 0x16EA8
+#define mmCLK1_CLK5_ALLOW_DS 0x16EB1
+
+#define mmCLK5_spll_field_8 0x1B04B
+#define mmCLK6_spll_field_8 0x1B24B
+#define mmDENTIST_DISPCLK_CNTL 0x0124
+#define regDENTIST_DISPCLK_CNTL 0x0064
+#define regDENTIST_DISPCLK_CNTL_BASE_IDX 1
+
+#define CLK1_CLK_PLL_REQ__FbMult_int__SHIFT 0x0
+#define CLK1_CLK_PLL_REQ__PllSpineDiv__SHIFT 0xc
+#define CLK1_CLK_PLL_REQ__FbMult_frac__SHIFT 0x10
+#define CLK1_CLK_PLL_REQ__FbMult_int_MASK 0x000001FFL
+#define CLK1_CLK_PLL_REQ__PllSpineDiv_MASK 0x0000F000L
+#define CLK1_CLK_PLL_REQ__FbMult_frac_MASK 0xFFFF0000L
+
+#define CLK1_CLK2_BYPASS_CNTL__CLK2_BYPASS_SEL_MASK 0x00000007L
+
+// DENTIST_DISPCLK_CNTL
+#define DENTIST_DISPCLK_CNTL__DENTIST_DISPCLK_WDIVIDER__SHIFT 0x0
+#define DENTIST_DISPCLK_CNTL__DENTIST_DISPCLK_RDIVIDER__SHIFT 0x8
+#define DENTIST_DISPCLK_CNTL__DENTIST_DISPCLK_CHG_DONE__SHIFT 0x13
+#define DENTIST_DISPCLK_CNTL__DENTIST_DPPCLK_CHG_DONE__SHIFT 0x14
+#define DENTIST_DISPCLK_CNTL__DENTIST_DPPCLK_WDIVIDER__SHIFT 0x18
+#define DENTIST_DISPCLK_CNTL__DENTIST_DISPCLK_WDIVIDER_MASK 0x0000007FL
+#define DENTIST_DISPCLK_CNTL__DENTIST_DISPCLK_RDIVIDER_MASK 0x00007F00L
+#define DENTIST_DISPCLK_CNTL__DENTIST_DISPCLK_CHG_DONE_MASK 0x00080000L
+#define DENTIST_DISPCLK_CNTL__DENTIST_DPPCLK_CHG_DONE_MASK 0x00100000L
+#define DENTIST_DISPCLK_CNTL__DENTIST_DPPCLK_WDIVIDER_MASK 0x7F000000L
+
+#define CLK5_spll_field_8__spll_ssc_en_MASK 0x00002000L
+
+#define REG(reg) \
+	(clk_mgr->regs->reg)
+
+#define BASE_INNER(seg) DCN_BASE__INST0_SEG ## seg
+
+#define BASE(seg) BASE_INNER(seg)
+
+#define SR(reg_name)\
+		.reg_name = BASE(reg ## reg_name ## _BASE_IDX) +  \
+					reg ## reg_name
+
+#define CLK_SR_DCN35(reg_name)\
+	.reg_name = mm ## reg_name
+
+static const struct clk_mgr_registers clk_mgr_regs_dcn351 = {
+	CLK_REG_LIST_DCN35()
+};
+
+static const struct clk_mgr_shift clk_mgr_shift_dcn351 = {
+	CLK_COMMON_MASK_SH_LIST_DCN32(__SHIFT)
+};
+
+static const struct clk_mgr_mask clk_mgr_mask_dcn351 = {
+	CLK_COMMON_MASK_SH_LIST_DCN32(_MASK)
+};
+
+#define TO_CLK_MGR_DCN35(clk_mgr)\
+	container_of(clk_mgr, struct clk_mgr_dcn35, base)
+
+
+void dcn351_clk_mgr_construct(
+		struct dc_context *ctx,
+		struct clk_mgr_dcn35 *clk_mgr,
+		struct pp_smu_funcs *pp_smu,
+		struct dccg *dccg)
+{
+	/*register offset changed*/
+	clk_mgr->base.regs = &clk_mgr_regs_dcn351;
+	clk_mgr->base.clk_mgr_shift = &clk_mgr_shift_dcn351;
+	clk_mgr->base.clk_mgr_mask = &clk_mgr_mask_dcn351;
+
+	dcn35_clk_mgr_construct(ctx,  clk_mgr, pp_smu, dccg);
+
+}
+
+
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn35_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn35_clk_mgr.c
new file mode 100644
index 000000000000..dfd0c9505af0
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn35_clk_mgr.c
@@ -0,0 +1,1588 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: AMD
+ *
+ */
+
+
+#include "dcn35_clk_mgr.h"
+
+#include "dccg.h"
+#include "clk_mgr_internal.h"
+
+// For dce12_get_dp_ref_freq_khz
+#include "dce100/dce_clk_mgr.h"
+
+// For dcn20_update_clocks_update_dpp_dto
+#include "dcn20/dcn20_clk_mgr.h"
+
+
+#include "reg_helper.h"
+#include "core_types.h"
+#include "dcn35_smu.h"
+#include "dm_helpers.h"
+
+#include "dcn31/dcn31_clk_mgr.h"
+
+#include "dc_dmub_srv.h"
+#include "link_service.h"
+#include "logger_types.h"
+
+#undef DC_LOGGER
+#define DC_LOGGER \
+	clk_mgr->base.base.ctx->logger
+
+#define DCN_BASE__INST0_SEG1 0x000000C0
+#define mmCLK1_CLK_PLL_REQ 0x16E37
+
+#define mmCLK1_CLK0_DFS_CNTL 0x16E69
+#define mmCLK1_CLK1_DFS_CNTL 0x16E6C
+#define mmCLK1_CLK2_DFS_CNTL 0x16E6F
+#define mmCLK1_CLK3_DFS_CNTL 0x16E72
+#define mmCLK1_CLK4_DFS_CNTL 0x16E75
+#define mmCLK1_CLK5_DFS_CNTL 0x16E78
+
+#define mmCLK1_CLK0_CURRENT_CNT 0x16EFB
+#define mmCLK1_CLK1_CURRENT_CNT 0x16EFC
+#define mmCLK1_CLK2_CURRENT_CNT 0x16EFD
+#define mmCLK1_CLK3_CURRENT_CNT 0x16EFE
+#define mmCLK1_CLK4_CURRENT_CNT 0x16EFF
+#define mmCLK1_CLK5_CURRENT_CNT 0x16F00
+
+#define mmCLK1_CLK0_BYPASS_CNTL 0x16E8A
+#define mmCLK1_CLK1_BYPASS_CNTL 0x16E93
+#define mmCLK1_CLK2_BYPASS_CNTL 0x16E9C
+#define mmCLK1_CLK3_BYPASS_CNTL 0x16EA5
+#define mmCLK1_CLK4_BYPASS_CNTL 0x16EAE
+#define mmCLK1_CLK5_BYPASS_CNTL 0x16EB7
+
+#define mmCLK1_CLK0_DS_CNTL 0x16E83
+#define mmCLK1_CLK1_DS_CNTL 0x16E8C
+#define mmCLK1_CLK2_DS_CNTL 0x16E95
+#define mmCLK1_CLK3_DS_CNTL 0x16E9E
+#define mmCLK1_CLK4_DS_CNTL 0x16EA7
+#define mmCLK1_CLK5_DS_CNTL 0x16EB0
+
+#define mmCLK1_CLK0_ALLOW_DS 0x16E84
+#define mmCLK1_CLK1_ALLOW_DS 0x16E8D
+#define mmCLK1_CLK2_ALLOW_DS 0x16E96
+#define mmCLK1_CLK3_ALLOW_DS 0x16E9F
+#define mmCLK1_CLK4_ALLOW_DS 0x16EA8
+#define mmCLK1_CLK5_ALLOW_DS 0x16EB1
+
+#define mmCLK5_spll_field_8 0x1B24B
+#define mmCLK6_spll_field_8 0x1B24B
+#define mmDENTIST_DISPCLK_CNTL 0x0124
+#define regDENTIST_DISPCLK_CNTL 0x0064
+#define regDENTIST_DISPCLK_CNTL_BASE_IDX 1
+
+#define CLK1_CLK_PLL_REQ__FbMult_int__SHIFT 0x0
+#define CLK1_CLK_PLL_REQ__PllSpineDiv__SHIFT 0xc
+#define CLK1_CLK_PLL_REQ__FbMult_frac__SHIFT 0x10
+#define CLK1_CLK_PLL_REQ__FbMult_int_MASK 0x000001FFL
+#define CLK1_CLK_PLL_REQ__PllSpineDiv_MASK 0x0000F000L
+#define CLK1_CLK_PLL_REQ__FbMult_frac_MASK 0xFFFF0000L
+
+#define CLK1_CLK2_BYPASS_CNTL__CLK2_BYPASS_SEL_MASK 0x00000007L
+#define CLK1_CLK2_BYPASS_CNTL__CLK2_BYPASS_DIV_MASK 0x000F0000L
+// DENTIST_DISPCLK_CNTL
+#define DENTIST_DISPCLK_CNTL__DENTIST_DISPCLK_WDIVIDER__SHIFT 0x0
+#define DENTIST_DISPCLK_CNTL__DENTIST_DISPCLK_RDIVIDER__SHIFT 0x8
+#define DENTIST_DISPCLK_CNTL__DENTIST_DISPCLK_CHG_DONE__SHIFT 0x13
+#define DENTIST_DISPCLK_CNTL__DENTIST_DPPCLK_CHG_DONE__SHIFT 0x14
+#define DENTIST_DISPCLK_CNTL__DENTIST_DPPCLK_WDIVIDER__SHIFT 0x18
+#define DENTIST_DISPCLK_CNTL__DENTIST_DISPCLK_WDIVIDER_MASK 0x0000007FL
+#define DENTIST_DISPCLK_CNTL__DENTIST_DISPCLK_RDIVIDER_MASK 0x00007F00L
+#define DENTIST_DISPCLK_CNTL__DENTIST_DISPCLK_CHG_DONE_MASK 0x00080000L
+#define DENTIST_DISPCLK_CNTL__DENTIST_DPPCLK_CHG_DONE_MASK 0x00100000L
+#define DENTIST_DISPCLK_CNTL__DENTIST_DPPCLK_WDIVIDER_MASK 0x7F000000L
+
+#define CLK5_spll_field_8__spll_ssc_en_MASK 0x00002000L
+#define CLK6_spll_field_8__spll_ssc_en_MASK 0x00002000L
+
+#define SMU_VER_THRESHOLD 0x5D4A00 //93.74.0
+#undef FN
+#define FN(reg_name, field_name) \
+	clk_mgr->clk_mgr_shift->field_name, clk_mgr->clk_mgr_mask->field_name
+
+#define REG(reg) \
+	(clk_mgr->regs->reg)
+
+#define BASE_INNER(seg) DCN_BASE__INST0_SEG ## seg
+
+#define BASE(seg) BASE_INNER(seg)
+
+#define SR(reg_name)\
+		.reg_name = BASE(reg ## reg_name ## _BASE_IDX) +  \
+					reg ## reg_name
+
+#define CLK_SR_DCN35(reg_name)\
+	.reg_name = mm ## reg_name
+
+static const struct clk_mgr_registers clk_mgr_regs_dcn35 = {
+	CLK_REG_LIST_DCN35()
+};
+
+static const struct clk_mgr_shift clk_mgr_shift_dcn35 = {
+	CLK_COMMON_MASK_SH_LIST_DCN32(__SHIFT)
+};
+
+static const struct clk_mgr_mask clk_mgr_mask_dcn35 = {
+	CLK_COMMON_MASK_SH_LIST_DCN32(_MASK)
+};
+
+#define TO_CLK_MGR_DCN35(clk_mgr)\
+	container_of(clk_mgr, struct clk_mgr_dcn35, base)
+
+static int dcn35_get_active_display_cnt_wa(
+		struct dc *dc,
+		struct dc_state *context,
+		int *all_active_disps)
+{
+	int i, display_count = 0;
+	bool tmds_present = false;
+
+	for (i = 0; i < context->stream_count; i++) {
+		const struct dc_stream_state *stream = context->streams[i];
+
+		if (stream->signal == SIGNAL_TYPE_HDMI_TYPE_A ||
+				stream->signal == SIGNAL_TYPE_DVI_SINGLE_LINK ||
+				stream->signal == SIGNAL_TYPE_DVI_DUAL_LINK)
+			tmds_present = true;
+	}
+
+	for (i = 0; i < dc->link_count; i++) {
+		const struct dc_link *link = dc->links[i];
+
+		/* abusing the fact that the dig and phy are coupled to see if the phy is enabled */
+		if (link->link_enc && link->link_enc->funcs->is_dig_enabled &&
+				link->link_enc->funcs->is_dig_enabled(link->link_enc))
+			display_count++;
+	}
+	if (all_active_disps != NULL)
+		*all_active_disps = display_count;
+	/* WA for hang on HDMI after display off back on*/
+	if (display_count == 0 && tmds_present)
+		display_count = 1;
+
+	return display_count;
+}
+static void dcn35_disable_otg_wa(struct clk_mgr *clk_mgr_base, struct dc_state *context,
+		bool safe_to_lower, bool disable)
+{
+	struct dc *dc = clk_mgr_base->ctx->dc;
+	int i;
+
+	if (dc->ctx->dce_environment == DCE_ENV_DIAG)
+		return;
+
+	for (i = 0; i < dc->res_pool->pipe_count; ++i) {
+		struct pipe_ctx *old_pipe = &dc->current_state->res_ctx.pipe_ctx[i];
+		struct pipe_ctx *new_pipe = &context->res_ctx.pipe_ctx[i];
+		struct clk_mgr_internal *clk_mgr_internal = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+		struct dccg *dccg = clk_mgr_internal->dccg;
+		struct pipe_ctx *pipe = safe_to_lower
+			? &context->res_ctx.pipe_ctx[i]
+			: &dc->current_state->res_ctx.pipe_ctx[i];
+		struct link_encoder *new_pipe_link_enc = new_pipe->link_res.dio_link_enc;
+		struct link_encoder *pipe_link_enc = pipe->link_res.dio_link_enc;
+		bool stream_changed_otg_dig_on = false;
+		bool has_active_hpo = false;
+
+		if (pipe->top_pipe || pipe->prev_odm_pipe)
+			continue;
+
+		if (!dc->config.unify_link_enc_assignment) {
+			if (new_pipe->stream)
+				new_pipe_link_enc = new_pipe->stream->link_enc;
+			if (pipe->stream)
+				pipe_link_enc = pipe->stream->link_enc;
+		}
+
+		stream_changed_otg_dig_on = old_pipe->stream && new_pipe->stream &&
+		old_pipe->stream != new_pipe->stream &&
+		old_pipe->stream_res.tg == new_pipe->stream_res.tg &&
+		new_pipe_link_enc && !new_pipe->stream->dpms_off &&
+		new_pipe_link_enc->funcs->is_dig_enabled &&
+		new_pipe_link_enc->funcs->is_dig_enabled(
+		new_pipe_link_enc) &&
+		new_pipe->stream_res.stream_enc &&
+		new_pipe->stream_res.stream_enc->funcs->is_fifo_enabled &&
+		new_pipe->stream_res.stream_enc->funcs->is_fifo_enabled(new_pipe->stream_res.stream_enc);
+
+		if (old_pipe->stream && new_pipe->stream && old_pipe->stream == new_pipe->stream) {
+			has_active_hpo =  dccg->ctx->dc->link_srv->dp_is_128b_132b_signal(old_pipe) &&
+			dccg->ctx->dc->link_srv->dp_is_128b_132b_signal(new_pipe);
+
+		}
+
+		if (!has_active_hpo && !stream_changed_otg_dig_on && pipe->stream &&
+		    (pipe->stream->dpms_off || dc_is_virtual_signal(pipe->stream->signal) || !pipe_link_enc) &&
+		    !dccg->ctx->dc->link_srv->dp_is_128b_132b_signal(pipe)) {
+			/* This w/a should not trigger when we have a dig active */
+			if (disable) {
+				if (pipe->stream_res.tg && pipe->stream_res.tg->funcs->immediate_disable_crtc)
+					pipe->stream_res.tg->funcs->immediate_disable_crtc(pipe->stream_res.tg);
+
+				reset_sync_context_for_pipe(dc, context, i);
+			} else {
+				pipe->stream_res.tg->funcs->enable_crtc(pipe->stream_res.tg);
+			}
+		}
+	}
+}
+
+static void dcn35_update_clocks_update_dtb_dto(struct clk_mgr_internal *clk_mgr,
+			struct dc_state *context,
+			int ref_dtbclk_khz)
+{
+	struct dccg *dccg = clk_mgr->dccg;
+	uint32_t tg_mask = 0;
+	int i;
+
+	for (i = 0; i < clk_mgr->base.ctx->dc->res_pool->pipe_count; i++) {
+		struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i];
+		struct dtbclk_dto_params dto_params = {0};
+
+		/* use mask to program DTO once per tg */
+		if (pipe_ctx->stream_res.tg &&
+				!(tg_mask & (1 << pipe_ctx->stream_res.tg->inst))) {
+			tg_mask |= (1 << pipe_ctx->stream_res.tg->inst);
+
+			dto_params.otg_inst = pipe_ctx->stream_res.tg->inst;
+			dto_params.ref_dtbclk_khz = ref_dtbclk_khz;
+
+			dccg->funcs->set_dtbclk_dto(clk_mgr->dccg, &dto_params);
+			//dccg->funcs->set_audio_dtbclk_dto(clk_mgr->dccg, &dto_params);
+		}
+	}
+}
+
+static void dcn35_update_clocks_update_dpp_dto(struct clk_mgr_internal *clk_mgr,
+		struct dc_state *context, bool safe_to_lower)
+{
+	int i;
+	bool dppclk_active[MAX_PIPES] = {0};
+
+
+	clk_mgr->dccg->ref_dppclk = clk_mgr->base.clks.dppclk_khz;
+	for (i = 0; i < clk_mgr->base.ctx->dc->res_pool->pipe_count; i++) {
+		int dpp_inst = 0, dppclk_khz, prev_dppclk_khz;
+
+		dppclk_khz = context->res_ctx.pipe_ctx[i].plane_res.bw.dppclk_khz;
+
+		if (context->res_ctx.pipe_ctx[i].plane_res.dpp)
+			dpp_inst = context->res_ctx.pipe_ctx[i].plane_res.dpp->inst;
+		else if (!context->res_ctx.pipe_ctx[i].plane_res.dpp && dppclk_khz == 0) {
+			/* dpp == NULL && dppclk_khz == 0 is valid because of pipe harvesting.
+			 * In this case just continue in loop
+			 */
+			continue;
+		} else if (!context->res_ctx.pipe_ctx[i].plane_res.dpp && dppclk_khz > 0) {
+			/* The software state is not valid if dpp resource is NULL and
+			 * dppclk_khz > 0.
+			 */
+			ASSERT(false);
+			continue;
+		}
+
+		prev_dppclk_khz = clk_mgr->dccg->pipe_dppclk_khz[i];
+
+		if (safe_to_lower || prev_dppclk_khz < dppclk_khz)
+			clk_mgr->dccg->funcs->update_dpp_dto(
+							clk_mgr->dccg, dpp_inst, dppclk_khz);
+		dppclk_active[dpp_inst] = true;
+	}
+	if (safe_to_lower)
+		for (i = 0; i < clk_mgr->base.ctx->dc->res_pool->pipe_count; i++) {
+			struct dpp *old_dpp = clk_mgr->base.ctx->dc->current_state->res_ctx.pipe_ctx[i].plane_res.dpp;
+
+			if (old_dpp && !dppclk_active[old_dpp->inst])
+				clk_mgr->dccg->funcs->update_dpp_dto(clk_mgr->dccg, old_dpp->inst, 0);
+		}
+}
+
+static uint8_t get_lowest_dpia_index(const struct dc_link *link)
+{
+	const struct dc *dc_struct = link->dc;
+	uint8_t idx = 0xFF;
+	int i;
+
+	for (i = 0; i < MAX_PIPES * 2; ++i) {
+		if (!dc_struct->links[i] || dc_struct->links[i]->ep_type != DISPLAY_ENDPOINT_USB4_DPIA)
+			continue;
+
+		if (idx > dc_struct->links[i]->link_index)
+			idx = dc_struct->links[i]->link_index;
+	}
+
+	return idx;
+}
+
+static void dcn35_notify_host_router_bw(struct clk_mgr *clk_mgr_base, struct dc_state *context,
+					bool safe_to_lower)
+{
+	struct dc_clocks *new_clocks = &context->bw_ctx.bw.dcn.clk;
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+	uint32_t host_router_bw_kbps[MAX_HOST_ROUTERS_NUM] = { 0 };
+	int i;
+	for (i = 0; i < context->stream_count; ++i) {
+		const struct dc_stream_state *stream = context->streams[i];
+		const struct dc_link *link = stream->link;
+		uint8_t lowest_dpia_index = 0;
+		unsigned int hr_index = 0;
+
+		if (!link)
+			continue;
+
+		lowest_dpia_index = get_lowest_dpia_index(link);
+		if (link->link_index < lowest_dpia_index)
+			continue;
+
+		hr_index = (link->link_index - lowest_dpia_index) / 2;
+		if (hr_index >= MAX_HOST_ROUTERS_NUM)
+			continue;
+		host_router_bw_kbps[hr_index] += dc_bandwidth_in_kbps_from_timing(
+			&stream->timing, dc_link_get_highest_encoding_format(link));
+	}
+
+	for (i = 0; i < MAX_HOST_ROUTERS_NUM; ++i) {
+		new_clocks->host_router_bw_kbps[i] = host_router_bw_kbps[i];
+		if (should_set_clock(safe_to_lower, new_clocks->host_router_bw_kbps[i], clk_mgr_base->clks.host_router_bw_kbps[i])) {
+			clk_mgr_base->clks.host_router_bw_kbps[i] = new_clocks->host_router_bw_kbps[i];
+			dcn35_smu_notify_host_router_bw(clk_mgr, i, new_clocks->host_router_bw_kbps[i]);
+		}
+	}
+}
+
+void dcn35_update_clocks(struct clk_mgr *clk_mgr_base,
+			struct dc_state *context,
+			bool safe_to_lower)
+{
+	union dmub_rb_cmd cmd;
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+	struct dc_clocks *new_clocks = &context->bw_ctx.bw.dcn.clk;
+	struct dc *dc = clk_mgr_base->ctx->dc;
+	int display_count = 0;
+	bool update_dppclk = false;
+	bool update_dispclk = false;
+	bool dpp_clock_lowered = false;
+	int all_active_disps = 0;
+
+	if (dc->work_arounds.skip_clock_update)
+		return;
+
+	display_count = dcn35_get_active_display_cnt_wa(dc, context, &all_active_disps);
+	if (new_clocks->dtbclk_en && !new_clocks->ref_dtbclk_khz)
+		new_clocks->ref_dtbclk_khz = 600000;
+	else if (!new_clocks->dtbclk_en && new_clocks->ref_dtbclk_khz > 590000)
+		new_clocks->ref_dtbclk_khz = 0;
+
+	/*
+	 * if it is safe to lower, but we are already in the lower state, we don't have to do anything
+	 * also if safe to lower is false, we just go in the higher state
+	 */
+	if (safe_to_lower) {
+		if (new_clocks->zstate_support != DCN_ZSTATE_SUPPORT_DISALLOW &&
+				new_clocks->zstate_support != clk_mgr_base->clks.zstate_support) {
+			dcn35_smu_set_zstate_support(clk_mgr, new_clocks->zstate_support);
+			dm_helpers_enable_periodic_detection(clk_mgr_base->ctx, true);
+			clk_mgr_base->clks.zstate_support = new_clocks->zstate_support;
+		}
+
+		if (clk_mgr_base->clks.dtbclk_en && !new_clocks->dtbclk_en) {
+			if (clk_mgr->base.ctx->dc->config.allow_0_dtb_clk)
+				dcn35_smu_set_dtbclk(clk_mgr, false);
+
+			clk_mgr_base->clks.dtbclk_en = new_clocks->dtbclk_en;
+		}
+		/* check that we're not already in lower */
+		if (clk_mgr_base->clks.pwr_state != DCN_PWR_STATE_LOW_POWER) {
+			/* if we can go lower, go lower */
+			if (display_count == 0)
+				clk_mgr_base->clks.pwr_state = DCN_PWR_STATE_LOW_POWER;
+		}
+	} else {
+		if (new_clocks->zstate_support == DCN_ZSTATE_SUPPORT_DISALLOW &&
+				new_clocks->zstate_support != clk_mgr_base->clks.zstate_support) {
+			dcn35_smu_set_zstate_support(clk_mgr, DCN_ZSTATE_SUPPORT_DISALLOW);
+			dm_helpers_enable_periodic_detection(clk_mgr_base->ctx, false);
+			clk_mgr_base->clks.zstate_support = new_clocks->zstate_support;
+		}
+
+		if (!clk_mgr_base->clks.dtbclk_en && new_clocks->dtbclk_en) {
+			int actual_dtbclk = 0;
+
+			dcn35_update_clocks_update_dtb_dto(clk_mgr, context, new_clocks->ref_dtbclk_khz);
+			dcn35_smu_set_dtbclk(clk_mgr, true);
+
+			actual_dtbclk = REG_READ(CLK1_CLK4_CURRENT_CNT);
+
+			if (actual_dtbclk > 590000) {
+				clk_mgr_base->clks.ref_dtbclk_khz = new_clocks->ref_dtbclk_khz;
+				clk_mgr_base->clks.dtbclk_en = new_clocks->dtbclk_en;
+			}
+		}
+
+		/* check that we're not already in D0 */
+		if (clk_mgr_base->clks.pwr_state != DCN_PWR_STATE_MISSION_MODE) {
+			union display_idle_optimization_u idle_info = { 0 };
+
+			dcn35_smu_set_display_idle_optimization(clk_mgr, idle_info.data);
+			/* update power state */
+			clk_mgr_base->clks.pwr_state = DCN_PWR_STATE_MISSION_MODE;
+		}
+	}
+	if (dc->debug.force_min_dcfclk_mhz > 0)
+		new_clocks->dcfclk_khz = (new_clocks->dcfclk_khz > (dc->debug.force_min_dcfclk_mhz * 1000)) ?
+				new_clocks->dcfclk_khz : (dc->debug.force_min_dcfclk_mhz * 1000);
+
+	if (should_set_clock(safe_to_lower, new_clocks->dcfclk_khz, clk_mgr_base->clks.dcfclk_khz)) {
+		clk_mgr_base->clks.dcfclk_khz = new_clocks->dcfclk_khz;
+		dcn35_smu_set_hard_min_dcfclk(clk_mgr, clk_mgr_base->clks.dcfclk_khz);
+	}
+
+	if (should_set_clock(safe_to_lower,
+			new_clocks->dcfclk_deep_sleep_khz, clk_mgr_base->clks.dcfclk_deep_sleep_khz)) {
+		clk_mgr_base->clks.dcfclk_deep_sleep_khz = new_clocks->dcfclk_deep_sleep_khz;
+		dcn35_smu_set_min_deep_sleep_dcfclk(clk_mgr, clk_mgr_base->clks.dcfclk_deep_sleep_khz);
+	}
+
+	// workaround: Limit dppclk to 100Mhz to avoid lower eDP panel switch to plus 4K monitor underflow.
+	if (new_clocks->dppclk_khz < 100000)
+		new_clocks->dppclk_khz = 100000;
+
+	if (should_set_clock(safe_to_lower, new_clocks->dppclk_khz, clk_mgr->base.clks.dppclk_khz)) {
+		if (clk_mgr->base.clks.dppclk_khz > new_clocks->dppclk_khz)
+			dpp_clock_lowered = true;
+		clk_mgr_base->clks.dppclk_khz = new_clocks->dppclk_khz;
+		update_dppclk = true;
+	}
+
+	if (should_set_clock(safe_to_lower, new_clocks->dispclk_khz, clk_mgr_base->clks.dispclk_khz) &&
+	    (new_clocks->dispclk_khz > 0 || (safe_to_lower && display_count == 0))) {
+		int requested_dispclk_khz = new_clocks->dispclk_khz;
+
+		dcn35_disable_otg_wa(clk_mgr_base, context, safe_to_lower, true);
+
+		/* Clamp the requested clock to PMFW based on their limit. */
+		if (dc->debug.min_disp_clk_khz > 0 && requested_dispclk_khz < dc->debug.min_disp_clk_khz)
+			requested_dispclk_khz = dc->debug.min_disp_clk_khz;
+
+		dcn35_smu_set_dispclk(clk_mgr, requested_dispclk_khz);
+		clk_mgr_base->clks.dispclk_khz = new_clocks->dispclk_khz;
+
+		dcn35_disable_otg_wa(clk_mgr_base, context, safe_to_lower, false);
+
+		update_dispclk = true;
+	}
+
+	/* clock limits are received with MHz precision, divide by 1000 to prevent setting clocks at every call */
+	if (!dc->debug.disable_dtb_ref_clk_switch &&
+	    should_set_clock(safe_to_lower, new_clocks->ref_dtbclk_khz / 1000,
+			     clk_mgr_base->clks.ref_dtbclk_khz / 1000)) {
+		dcn35_update_clocks_update_dtb_dto(clk_mgr, context, new_clocks->ref_dtbclk_khz);
+		clk_mgr_base->clks.ref_dtbclk_khz = new_clocks->ref_dtbclk_khz;
+	}
+
+	if (dpp_clock_lowered) {
+		// increase per DPP DTO before lowering global dppclk
+		dcn35_update_clocks_update_dpp_dto(clk_mgr, context, safe_to_lower);
+		dcn35_smu_set_dppclk(clk_mgr, clk_mgr_base->clks.dppclk_khz);
+	} else {
+		// increase global DPPCLK before lowering per DPP DTO
+		if (update_dppclk || update_dispclk)
+			dcn35_smu_set_dppclk(clk_mgr, clk_mgr_base->clks.dppclk_khz);
+		dcn35_update_clocks_update_dpp_dto(clk_mgr, context, safe_to_lower);
+	}
+
+	// notify PMFW of bandwidth per DPIA tunnel
+	if (dc->debug.notify_dpia_hr_bw)
+		dcn35_notify_host_router_bw(clk_mgr_base, context, safe_to_lower);
+
+	// notify DMCUB of latest clocks
+	memset(&cmd, 0, sizeof(cmd));
+	cmd.notify_clocks.header.type = DMUB_CMD__CLK_MGR;
+	cmd.notify_clocks.header.sub_type = DMUB_CMD__CLK_MGR_NOTIFY_CLOCKS;
+	cmd.notify_clocks.clocks.dcfclk_khz = clk_mgr_base->clks.dcfclk_khz;
+	cmd.notify_clocks.clocks.dcfclk_deep_sleep_khz =
+		clk_mgr_base->clks.dcfclk_deep_sleep_khz;
+	cmd.notify_clocks.clocks.dispclk_khz = clk_mgr_base->clks.dispclk_khz;
+	cmd.notify_clocks.clocks.dppclk_khz = clk_mgr_base->clks.dppclk_khz;
+
+	dc_wake_and_execute_dmub_cmd(dc->ctx, &cmd, DM_DMUB_WAIT_TYPE_WAIT);
+}
+
+static int get_vco_frequency_from_reg(struct clk_mgr_internal *clk_mgr)
+{
+	/* get FbMult value */
+	struct fixed31_32 pll_req;
+	unsigned int fbmult_frac_val = 0;
+	unsigned int fbmult_int_val = 0;
+
+	/*
+	 * Register value of fbmult is in 8.16 format, we are converting to 314.32
+	 * to leverage the fix point operations available in driver
+	 */
+
+	REG_GET(CLK1_CLK_PLL_REQ, FbMult_frac, &fbmult_frac_val); /* 16 bit fractional part*/
+	REG_GET(CLK1_CLK_PLL_REQ, FbMult_int, &fbmult_int_val); /* 8 bit integer part */
+
+	pll_req = dc_fixpt_from_int(fbmult_int_val);
+
+	/*
+	 * since fractional part is only 16 bit in register definition but is 32 bit
+	 * in our fix point definiton, need to shift left by 16 to obtain correct value
+	 */
+	pll_req.value |= fbmult_frac_val << 16;
+
+	/* multiply by REFCLK period */
+	pll_req = dc_fixpt_mul_int(pll_req, clk_mgr->dfs_ref_freq_khz);
+
+	/* integer part is now VCO frequency in kHz */
+	return dc_fixpt_floor(pll_req);
+}
+
+static void dcn35_enable_pme_wa(struct clk_mgr *clk_mgr_base)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+
+	dcn35_smu_enable_pme_wa(clk_mgr);
+}
+
+
+bool dcn35_are_clock_states_equal(struct dc_clocks *a,
+		struct dc_clocks *b)
+{
+	if (a->dispclk_khz != b->dispclk_khz)
+		return false;
+	else if (a->dppclk_khz != b->dppclk_khz)
+		return false;
+	else if (a->dcfclk_khz != b->dcfclk_khz)
+		return false;
+	else if (a->dcfclk_deep_sleep_khz != b->dcfclk_deep_sleep_khz)
+		return false;
+	else if (a->zstate_support != b->zstate_support)
+		return false;
+	else if (a->dtbclk_en != b->dtbclk_en)
+		return false;
+
+	return true;
+}
+
+static void dcn35_save_clk_registers_internal(struct dcn35_clk_internal *internal, struct clk_mgr *clk_mgr_base)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+
+	// read dtbclk
+	internal->CLK1_CLK4_CURRENT_CNT = REG_READ(CLK1_CLK4_CURRENT_CNT);
+	internal->CLK1_CLK4_BYPASS_CNTL = REG_READ(CLK1_CLK4_BYPASS_CNTL);
+
+	// read dcfclk
+	internal->CLK1_CLK3_CURRENT_CNT = REG_READ(CLK1_CLK3_CURRENT_CNT);
+	internal->CLK1_CLK3_BYPASS_CNTL = REG_READ(CLK1_CLK3_BYPASS_CNTL);
+
+	// read dcf deep sleep divider
+	internal->CLK1_CLK3_DS_CNTL = REG_READ(CLK1_CLK3_DS_CNTL);
+	internal->CLK1_CLK3_ALLOW_DS = REG_READ(CLK1_CLK3_ALLOW_DS);
+
+	// read dppclk
+	internal->CLK1_CLK1_CURRENT_CNT = REG_READ(CLK1_CLK1_CURRENT_CNT);
+	internal->CLK1_CLK1_BYPASS_CNTL = REG_READ(CLK1_CLK1_BYPASS_CNTL);
+
+	// read dprefclk
+	internal->CLK1_CLK2_CURRENT_CNT = REG_READ(CLK1_CLK2_CURRENT_CNT);
+	internal->CLK1_CLK2_BYPASS_CNTL = REG_READ(CLK1_CLK2_BYPASS_CNTL);
+
+	// read dispclk
+	internal->CLK1_CLK0_CURRENT_CNT = REG_READ(CLK1_CLK0_CURRENT_CNT);
+	internal->CLK1_CLK0_BYPASS_CNTL = REG_READ(CLK1_CLK0_BYPASS_CNTL);
+}
+
+static void dcn35_save_clk_registers(struct clk_state_registers_and_bypass *regs_and_bypass,
+		struct clk_mgr_dcn35 *clk_mgr)
+{
+	struct dcn35_clk_internal internal = {0};
+	char *bypass_clks[5] = {"0x0 DFS", "0x1 REFCLK", "0x2 ERROR", "0x3 400 FCH", "0x4 600 FCH"};
+
+	dcn35_save_clk_registers_internal(&internal, &clk_mgr->base.base);
+
+	regs_and_bypass->dcfclk = internal.CLK1_CLK3_CURRENT_CNT / 10;
+	regs_and_bypass->dcf_deep_sleep_divider = internal.CLK1_CLK3_DS_CNTL / 10;
+	regs_and_bypass->dcf_deep_sleep_allow = internal.CLK1_CLK3_ALLOW_DS;
+	regs_and_bypass->dprefclk = internal.CLK1_CLK2_CURRENT_CNT / 10;
+	regs_and_bypass->dispclk = internal.CLK1_CLK0_CURRENT_CNT / 10;
+	regs_and_bypass->dppclk = internal.CLK1_CLK1_CURRENT_CNT / 10;
+	regs_and_bypass->dtbclk = internal.CLK1_CLK4_CURRENT_CNT / 10;
+
+	regs_and_bypass->dppclk_bypass = internal.CLK1_CLK1_BYPASS_CNTL & 0x0007;
+	if (regs_and_bypass->dppclk_bypass > 4)
+		regs_and_bypass->dppclk_bypass = 0;
+	regs_and_bypass->dcfclk_bypass = internal.CLK1_CLK3_BYPASS_CNTL & 0x0007;
+	if (regs_and_bypass->dcfclk_bypass > 4)
+		regs_and_bypass->dcfclk_bypass = 0;
+	regs_and_bypass->dispclk_bypass = internal.CLK1_CLK0_BYPASS_CNTL & 0x0007;
+	if (regs_and_bypass->dispclk_bypass > 4)
+		regs_and_bypass->dispclk_bypass = 0;
+	regs_and_bypass->dprefclk_bypass = internal.CLK1_CLK2_BYPASS_CNTL & 0x0007;
+	if (regs_and_bypass->dprefclk_bypass > 4)
+		regs_and_bypass->dprefclk_bypass = 0;
+
+	if (clk_mgr->base.base.ctx->dc->debug.pstate_enabled) {
+		DC_LOG_SMU("clk_type,clk_value,deepsleep_cntl,deepsleep_allow,bypass\n");
+
+		DC_LOG_SMU("dcfclk,%d,%d,%d,%s\n",
+				   regs_and_bypass->dcfclk,
+				   regs_and_bypass->dcf_deep_sleep_divider,
+				   regs_and_bypass->dcf_deep_sleep_allow,
+				   bypass_clks[(int) regs_and_bypass->dcfclk_bypass]);
+
+		DC_LOG_SMU("dprefclk,%d,N/A,N/A,%s\n",
+			regs_and_bypass->dprefclk,
+			bypass_clks[(int) regs_and_bypass->dprefclk_bypass]);
+
+		DC_LOG_SMU("dispclk,%d,N/A,N/A,%s\n",
+			regs_and_bypass->dispclk,
+			bypass_clks[(int) regs_and_bypass->dispclk_bypass]);
+
+		// REGISTER VALUES
+		DC_LOG_SMU("reg_name,value,clk_type");
+
+		DC_LOG_SMU("CLK1_CLK3_CURRENT_CNT,%d,dcfclk",
+				internal.CLK1_CLK3_CURRENT_CNT);
+
+		DC_LOG_SMU("CLK1_CLK4_CURRENT_CNT,%d,dtbclk",
+					internal.CLK1_CLK4_CURRENT_CNT);
+
+		DC_LOG_SMU("CLK1_CLK3_DS_CNTL,%d,dcf_deep_sleep_divider",
+					internal.CLK1_CLK3_DS_CNTL);
+
+		DC_LOG_SMU("CLK1_CLK3_ALLOW_DS,%d,dcf_deep_sleep_allow",
+					internal.CLK1_CLK3_ALLOW_DS);
+
+		DC_LOG_SMU("CLK1_CLK2_CURRENT_CNT,%d,dprefclk",
+					internal.CLK1_CLK2_CURRENT_CNT);
+
+		DC_LOG_SMU("CLK1_CLK0_CURRENT_CNT,%d,dispclk",
+					internal.CLK1_CLK0_CURRENT_CNT);
+
+		DC_LOG_SMU("CLK1_CLK1_CURRENT_CNT,%d,dppclk",
+					internal.CLK1_CLK1_CURRENT_CNT);
+
+		DC_LOG_SMU("CLK1_CLK3_BYPASS_CNTL,%d,dcfclk_bypass",
+					internal.CLK1_CLK3_BYPASS_CNTL);
+
+		DC_LOG_SMU("CLK1_CLK2_BYPASS_CNTL,%d,dprefclk_bypass",
+					internal.CLK1_CLK2_BYPASS_CNTL);
+
+		DC_LOG_SMU("CLK1_CLK0_BYPASS_CNTL,%d,dispclk_bypass",
+					internal.CLK1_CLK0_BYPASS_CNTL);
+
+		DC_LOG_SMU("CLK1_CLK1_BYPASS_CNTL,%d,dppclk_bypass",
+					internal.CLK1_CLK1_BYPASS_CNTL);
+
+	}
+}
+
+static bool dcn35_is_spll_ssc_enabled(struct clk_mgr *clk_mgr_base)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+
+	uint32_t ssc_enable;
+
+	if (clk_mgr_base->ctx->dce_version == DCN_VERSION_3_51) {
+		ssc_enable = REG_READ(CLK6_spll_field_8) & CLK6_spll_field_8__spll_ssc_en_MASK;
+	} else {
+		ssc_enable = REG_READ(CLK5_spll_field_8) & CLK5_spll_field_8__spll_ssc_en_MASK;
+	}
+
+	return ssc_enable != 0;
+}
+
+static void init_clk_states(struct clk_mgr *clk_mgr)
+{
+	uint32_t ref_dtbclk = clk_mgr->clks.ref_dtbclk_khz;
+
+	memset(&(clk_mgr->clks), 0, sizeof(struct dc_clocks));
+
+	clk_mgr->clks.ref_dtbclk_khz = ref_dtbclk;	// restore ref_dtbclk
+	clk_mgr->clks.p_state_change_support = true;
+	clk_mgr->clks.prev_p_state_change_support = true;
+	clk_mgr->clks.pwr_state = DCN_PWR_STATE_UNKNOWN;
+	clk_mgr->clks.zstate_support = DCN_ZSTATE_SUPPORT_UNKNOWN;
+}
+
+void dcn35_init_clocks(struct clk_mgr *clk_mgr)
+{
+	struct clk_mgr_internal *clk_mgr_int = TO_CLK_MGR_INTERNAL(clk_mgr);
+	struct clk_mgr_dcn35 *clk_mgr_dcn35 = TO_CLK_MGR_DCN35(clk_mgr_int);
+
+	init_clk_states(clk_mgr);
+
+	// to adjust dp_dto reference clock if ssc is enable otherwise to apply dprefclk
+	if (dcn35_is_spll_ssc_enabled(clk_mgr))
+		clk_mgr->dp_dto_source_clock_in_khz =
+			dce_adjust_dp_ref_freq_for_ss(clk_mgr_int, clk_mgr->dprefclk_khz);
+	else
+		clk_mgr->dp_dto_source_clock_in_khz = clk_mgr->dprefclk_khz;
+
+	dcn35_save_clk_registers(&clk_mgr->boot_snapshot, clk_mgr_dcn35);
+
+	clk_mgr->clks.ref_dtbclk_khz =  clk_mgr->boot_snapshot.dtbclk * 10;
+	if (clk_mgr->boot_snapshot.dtbclk > 59000) {
+		/*dtbclk enabled based on */
+		clk_mgr->clks.dtbclk_en = true;
+	}
+}
+static struct clk_bw_params dcn35_bw_params = {
+	.vram_type = Ddr4MemType,
+	.num_channels = 1,
+	.clk_table = {
+		.num_entries = 4,
+	},
+
+};
+
+static struct wm_table ddr5_wm_table = {
+	.entries = {
+		{
+			.wm_inst = WM_A,
+			.wm_type = WM_TYPE_PSTATE_CHG,
+			.pstate_latency_us = 11.72,
+			.sr_exit_time_us = 28.0,
+			.sr_enter_plus_exit_time_us = 30.0,
+			.valid = true,
+		},
+		{
+			.wm_inst = WM_B,
+			.wm_type = WM_TYPE_PSTATE_CHG,
+			.pstate_latency_us = 11.72,
+			.sr_exit_time_us = 28.0,
+			.sr_enter_plus_exit_time_us = 30.0,
+			.valid = true,
+		},
+		{
+			.wm_inst = WM_C,
+			.wm_type = WM_TYPE_PSTATE_CHG,
+			.pstate_latency_us = 11.72,
+			.sr_exit_time_us = 28.0,
+			.sr_enter_plus_exit_time_us = 30.0,
+			.valid = true,
+		},
+		{
+			.wm_inst = WM_D,
+			.wm_type = WM_TYPE_PSTATE_CHG,
+			.pstate_latency_us = 11.72,
+			.sr_exit_time_us = 28.0,
+			.sr_enter_plus_exit_time_us = 30.0,
+			.valid = true,
+		},
+	}
+};
+
+static struct wm_table lpddr5_wm_table = {
+	.entries = {
+		{
+			.wm_inst = WM_A,
+			.wm_type = WM_TYPE_PSTATE_CHG,
+			.pstate_latency_us = 11.65333,
+			.sr_exit_time_us = 28.0,
+			.sr_enter_plus_exit_time_us = 30.0,
+			.valid = true,
+		},
+		{
+			.wm_inst = WM_B,
+			.wm_type = WM_TYPE_PSTATE_CHG,
+			.pstate_latency_us = 11.65333,
+			.sr_exit_time_us = 28.0,
+			.sr_enter_plus_exit_time_us = 30.0,
+			.valid = true,
+		},
+		{
+			.wm_inst = WM_C,
+			.wm_type = WM_TYPE_PSTATE_CHG,
+			.pstate_latency_us = 11.65333,
+			.sr_exit_time_us = 28.0,
+			.sr_enter_plus_exit_time_us = 30.0,
+			.valid = true,
+		},
+		{
+			.wm_inst = WM_D,
+			.wm_type = WM_TYPE_PSTATE_CHG,
+			.pstate_latency_us = 11.65333,
+			.sr_exit_time_us = 28.0,
+			.sr_enter_plus_exit_time_us = 30.0,
+			.valid = true,
+		},
+	}
+};
+
+static DpmClocks_t_dcn35 dummy_clocks;
+static DpmClocks_t_dcn351 dummy_clocks_dcn351;
+
+static struct dcn35_watermarks dummy_wms = { 0 };
+
+static struct dcn35_ss_info_table ss_info_table = {
+	.ss_divider = 1000,
+	.ss_percentage = {0, 0, 375, 375, 375}
+};
+
+static void dcn35_read_ss_info_from_lut(struct clk_mgr_internal *clk_mgr)
+{
+	uint32_t clock_source = 0;
+
+	clock_source = REG_READ(CLK1_CLK2_BYPASS_CNTL) & CLK1_CLK2_BYPASS_CNTL__CLK2_BYPASS_SEL_MASK;
+
+	// If it's DFS mode, clock_source is 0.
+	if (dcn35_is_spll_ssc_enabled(&clk_mgr->base) && (clock_source < ARRAY_SIZE(ss_info_table.ss_percentage))) {
+		clk_mgr->dprefclk_ss_percentage = ss_info_table.ss_percentage[clock_source];
+
+		if (clk_mgr->dprefclk_ss_percentage != 0) {
+			clk_mgr->ss_on_dprefclk = true;
+			clk_mgr->dprefclk_ss_divider = ss_info_table.ss_divider;
+		}
+	}
+}
+
+static void dcn35_build_watermark_ranges(struct clk_bw_params *bw_params, struct dcn35_watermarks *table)
+{
+	int i, num_valid_sets;
+
+	num_valid_sets = 0;
+
+	for (i = 0; i < WM_SET_COUNT; i++) {
+		/* skip empty entries, the smu array has no holes*/
+		if (!bw_params->wm_table.entries[i].valid)
+			continue;
+
+		table->WatermarkRow[WM_DCFCLK][num_valid_sets].WmSetting = bw_params->wm_table.entries[i].wm_inst;
+		table->WatermarkRow[WM_DCFCLK][num_valid_sets].WmType = bw_params->wm_table.entries[i].wm_type;
+		/* We will not select WM based on fclk, so leave it as unconstrained */
+		table->WatermarkRow[WM_DCFCLK][num_valid_sets].MinClock = 0;
+		table->WatermarkRow[WM_DCFCLK][num_valid_sets].MaxClock = 0xFFFF;
+
+		if (table->WatermarkRow[WM_DCFCLK][num_valid_sets].WmType == WM_TYPE_PSTATE_CHG) {
+			if (i == 0)
+				table->WatermarkRow[WM_DCFCLK][num_valid_sets].MinMclk = 0;
+			else {
+				/* add 1 to make it non-overlapping with next lvl */
+				table->WatermarkRow[WM_DCFCLK][num_valid_sets].MinMclk =
+						bw_params->clk_table.entries[i - 1].dcfclk_mhz + 1;
+			}
+			table->WatermarkRow[WM_DCFCLK][num_valid_sets].MaxMclk =
+					bw_params->clk_table.entries[i].dcfclk_mhz;
+
+		} else {
+			/* unconstrained for memory retraining */
+			table->WatermarkRow[WM_DCFCLK][num_valid_sets].MinClock = 0;
+			table->WatermarkRow[WM_DCFCLK][num_valid_sets].MaxClock = 0xFFFF;
+
+			/* Modify previous watermark range to cover up to max */
+			table->WatermarkRow[WM_DCFCLK][num_valid_sets - 1].MaxClock = 0xFFFF;
+		}
+		num_valid_sets++;
+	}
+
+	ASSERT(num_valid_sets != 0); /* Must have at least one set of valid watermarks */
+
+	/* modify the min and max to make sure we cover the whole range*/
+	table->WatermarkRow[WM_DCFCLK][0].MinMclk = 0;
+	table->WatermarkRow[WM_DCFCLK][0].MinClock = 0;
+	table->WatermarkRow[WM_DCFCLK][num_valid_sets - 1].MaxMclk = 0xFFFF;
+	table->WatermarkRow[WM_DCFCLK][num_valid_sets - 1].MaxClock = 0xFFFF;
+
+	/* This is for writeback only, does not matter currently as no writeback support*/
+	table->WatermarkRow[WM_SOCCLK][0].WmSetting = WM_A;
+	table->WatermarkRow[WM_SOCCLK][0].MinClock = 0;
+	table->WatermarkRow[WM_SOCCLK][0].MaxClock = 0xFFFF;
+	table->WatermarkRow[WM_SOCCLK][0].MinMclk = 0;
+	table->WatermarkRow[WM_SOCCLK][0].MaxMclk = 0xFFFF;
+}
+
+static void dcn35_notify_wm_ranges(struct clk_mgr *clk_mgr_base)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+	struct clk_mgr_dcn35 *clk_mgr_dcn35 = TO_CLK_MGR_DCN35(clk_mgr);
+	struct dcn35_watermarks *table = clk_mgr_dcn35->smu_wm_set.wm_set;
+
+	if (!clk_mgr->smu_ver)
+		return;
+
+	if (!table || clk_mgr_dcn35->smu_wm_set.mc_address.quad_part == 0)
+		return;
+
+	memset(table, 0, sizeof(*table));
+
+	dcn35_build_watermark_ranges(clk_mgr_base->bw_params, table);
+
+	dcn35_smu_set_dram_addr_high(clk_mgr,
+			clk_mgr_dcn35->smu_wm_set.mc_address.high_part);
+	dcn35_smu_set_dram_addr_low(clk_mgr,
+			clk_mgr_dcn35->smu_wm_set.mc_address.low_part);
+	dcn35_smu_transfer_wm_table_dram_2_smu(clk_mgr);
+}
+
+static void dcn35_get_dpm_table_from_smu(struct clk_mgr_internal *clk_mgr,
+		struct dcn35_smu_dpm_clks *smu_dpm_clks)
+{
+	DpmClocks_t_dcn35 *table = smu_dpm_clks->dpm_clks;
+
+	if (!clk_mgr->smu_ver)
+		return;
+
+	if (!table || smu_dpm_clks->mc_address.quad_part == 0)
+		return;
+
+	memset(table, 0, sizeof(*table));
+
+	dcn35_smu_set_dram_addr_high(clk_mgr,
+			smu_dpm_clks->mc_address.high_part);
+	dcn35_smu_set_dram_addr_low(clk_mgr,
+			smu_dpm_clks->mc_address.low_part);
+	dcn35_smu_transfer_dpm_table_smu_2_dram(clk_mgr);
+}
+
+static void dcn351_get_dpm_table_from_smu(struct clk_mgr_internal *clk_mgr,
+		struct dcn351_smu_dpm_clks *smu_dpm_clks)
+{
+	DpmClocks_t_dcn351 *table = smu_dpm_clks->dpm_clks;
+
+	if (!clk_mgr->smu_ver)
+		return;
+	if (!table || smu_dpm_clks->mc_address.quad_part == 0)
+		return;
+	memset(table, 0, sizeof(*table));
+	dcn35_smu_set_dram_addr_high(clk_mgr,
+			smu_dpm_clks->mc_address.high_part);
+	dcn35_smu_set_dram_addr_low(clk_mgr,
+			smu_dpm_clks->mc_address.low_part);
+	dcn35_smu_transfer_dpm_table_smu_2_dram(clk_mgr);
+}
+static uint32_t find_max_clk_value(const uint32_t clocks[], uint32_t num_clocks)
+{
+	uint32_t max = 0;
+	int i;
+
+	for (i = 0; i < num_clocks; ++i) {
+		if (clocks[i] > max)
+			max = clocks[i];
+	}
+
+	return max;
+}
+
+static inline bool is_valid_clock_value(uint32_t clock_value)
+{
+	return clock_value > 1 && clock_value < 100000;
+}
+
+static unsigned int convert_wck_ratio(uint8_t wck_ratio)
+{
+	switch (wck_ratio) {
+	case WCK_RATIO_1_2:
+		return 2;
+
+	case WCK_RATIO_1_4:
+		return 4;
+	/* Find lowest DPM, FCLK is filled in reverse order*/
+
+	default:
+			break;
+	}
+
+	return 1;
+}
+
+static inline uint32_t calc_dram_speed_mts(const MemPstateTable_t *entry)
+{
+	return entry->UClk * convert_wck_ratio(entry->WckRatio) * 2;
+}
+
+static void dcn35_clk_mgr_helper_populate_bw_params(struct clk_mgr_internal *clk_mgr,
+						    struct integrated_info *bios_info,
+						    DpmClocks_t_dcn35 *clock_table)
+{
+	struct clk_bw_params *bw_params = clk_mgr->base.bw_params;
+	struct clk_limit_table_entry def_max = bw_params->clk_table.entries[bw_params->clk_table.num_entries - 1];
+	uint32_t max_fclk = 0, min_pstate = 0, max_dispclk = 0, max_dppclk = 0;
+	uint32_t max_pstate = 0, max_dram_speed_mts = 0, min_dram_speed_mts = 0;
+	uint32_t num_memps, num_fclk, num_dcfclk;
+	int i;
+
+	/* Determine min/max p-state values. */
+	num_memps = (clock_table->NumMemPstatesEnabled > NUM_MEM_PSTATE_LEVELS) ? NUM_MEM_PSTATE_LEVELS :
+		clock_table->NumMemPstatesEnabled;
+	for (i = 0; i < num_memps; i++) {
+		uint32_t dram_speed_mts = calc_dram_speed_mts(&clock_table->MemPstateTable[i]);
+
+		if (is_valid_clock_value(dram_speed_mts) && dram_speed_mts > max_dram_speed_mts) {
+			max_dram_speed_mts = dram_speed_mts;
+			max_pstate = i;
+		}
+	}
+
+	min_dram_speed_mts = max_dram_speed_mts;
+	min_pstate = max_pstate;
+
+	for (i = 0; i < num_memps; i++) {
+		uint32_t dram_speed_mts = calc_dram_speed_mts(&clock_table->MemPstateTable[i]);
+
+		if (is_valid_clock_value(dram_speed_mts) && dram_speed_mts < min_dram_speed_mts) {
+			min_dram_speed_mts = dram_speed_mts;
+			min_pstate = i;
+		}
+	}
+
+	/* We expect the table to contain at least one valid P-state entry. */
+	ASSERT(clock_table->NumMemPstatesEnabled &&
+	       is_valid_clock_value(max_dram_speed_mts) &&
+	       is_valid_clock_value(min_dram_speed_mts));
+
+	/* dispclk and dppclk can be max at any voltage, same number of levels for both */
+	if (clock_table->NumDispClkLevelsEnabled <= NUM_DISPCLK_DPM_LEVELS &&
+	    clock_table->NumDispClkLevelsEnabled <= NUM_DPPCLK_DPM_LEVELS) {
+		max_dispclk = find_max_clk_value(clock_table->DispClocks,
+			clock_table->NumDispClkLevelsEnabled);
+		max_dppclk = find_max_clk_value(clock_table->DppClocks,
+			clock_table->NumDispClkLevelsEnabled);
+	} else {
+		/* Invalid number of entries in the table from PMFW. */
+		ASSERT(0);
+	}
+
+	/* Base the clock table on dcfclk, need at least one entry regardless of pmfw table */
+	ASSERT(clock_table->NumDcfClkLevelsEnabled > 0);
+
+	num_fclk = (clock_table->NumFclkLevelsEnabled > NUM_FCLK_DPM_LEVELS) ? NUM_FCLK_DPM_LEVELS :
+		clock_table->NumFclkLevelsEnabled;
+	max_fclk = find_max_clk_value(clock_table->FclkClocks_Freq, num_fclk);
+
+	num_dcfclk = (clock_table->NumDcfClkLevelsEnabled > NUM_DCFCLK_DPM_LEVELS) ? NUM_DCFCLK_DPM_LEVELS :
+		clock_table->NumDcfClkLevelsEnabled;
+	for (i = 0; i < num_dcfclk; i++) {
+		int j;
+
+		/* First search defaults for the clocks we don't read using closest lower or equal default dcfclk */
+		for (j = bw_params->clk_table.num_entries - 1; j > 0; j--)
+			if (bw_params->clk_table.entries[j].dcfclk_mhz <= clock_table->DcfClocks[i])
+				break;
+
+		bw_params->clk_table.entries[i].phyclk_mhz = bw_params->clk_table.entries[j].phyclk_mhz;
+		bw_params->clk_table.entries[i].phyclk_d18_mhz = bw_params->clk_table.entries[j].phyclk_d18_mhz;
+		bw_params->clk_table.entries[i].dtbclk_mhz = bw_params->clk_table.entries[j].dtbclk_mhz;
+
+		/* Now update clocks we do read */
+		bw_params->clk_table.entries[i].memclk_mhz = clock_table->MemPstateTable[min_pstate].MemClk;
+		bw_params->clk_table.entries[i].voltage = clock_table->MemPstateTable[min_pstate].Voltage;
+		bw_params->clk_table.entries[i].dcfclk_mhz = clock_table->DcfClocks[i];
+		bw_params->clk_table.entries[i].socclk_mhz = clock_table->SocClocks[i];
+		bw_params->clk_table.entries[i].dispclk_mhz = max_dispclk;
+		bw_params->clk_table.entries[i].dppclk_mhz = max_dppclk;
+		bw_params->clk_table.entries[i].wck_ratio =
+			convert_wck_ratio(clock_table->MemPstateTable[min_pstate].WckRatio);
+
+		/* Dcfclk and Fclk are tied, but at a different ratio */
+		bw_params->clk_table.entries[i].fclk_mhz = min(max_fclk, 2 * clock_table->DcfClocks[i]);
+	}
+
+	/* Make sure to include at least one entry at highest pstate */
+	if (max_pstate != min_pstate || i == 0) {
+		if (i > MAX_NUM_DPM_LVL - 1)
+			i = MAX_NUM_DPM_LVL - 1;
+
+		bw_params->clk_table.entries[i].fclk_mhz = max_fclk;
+		bw_params->clk_table.entries[i].memclk_mhz = clock_table->MemPstateTable[max_pstate].MemClk;
+		bw_params->clk_table.entries[i].voltage = clock_table->MemPstateTable[max_pstate].Voltage;
+		bw_params->clk_table.entries[i].dcfclk_mhz =
+			find_max_clk_value(clock_table->DcfClocks, NUM_DCFCLK_DPM_LEVELS);
+		bw_params->clk_table.entries[i].socclk_mhz =
+			find_max_clk_value(clock_table->SocClocks, NUM_SOCCLK_DPM_LEVELS);
+		bw_params->clk_table.entries[i].dispclk_mhz = max_dispclk;
+		bw_params->clk_table.entries[i].dppclk_mhz = max_dppclk;
+		bw_params->clk_table.entries[i].wck_ratio = convert_wck_ratio(
+			clock_table->MemPstateTable[max_pstate].WckRatio);
+		i++;
+	}
+	bw_params->clk_table.num_entries = i--;
+
+	/* Make sure all highest clocks are included*/
+	bw_params->clk_table.entries[i].socclk_mhz =
+		find_max_clk_value(clock_table->SocClocks, NUM_SOCCLK_DPM_LEVELS);
+	bw_params->clk_table.entries[i].dispclk_mhz =
+		find_max_clk_value(clock_table->DispClocks, NUM_DISPCLK_DPM_LEVELS);
+	bw_params->clk_table.entries[i].dppclk_mhz =
+		find_max_clk_value(clock_table->DppClocks, NUM_DPPCLK_DPM_LEVELS);
+	bw_params->clk_table.entries[i].fclk_mhz =
+		find_max_clk_value(clock_table->FclkClocks_Freq, NUM_FCLK_DPM_LEVELS);
+	ASSERT(clock_table->DcfClocks[i] == find_max_clk_value(clock_table->DcfClocks, NUM_DCFCLK_DPM_LEVELS));
+	bw_params->clk_table.entries[i].phyclk_mhz = def_max.phyclk_mhz;
+	bw_params->clk_table.entries[i].phyclk_d18_mhz = def_max.phyclk_d18_mhz;
+	bw_params->clk_table.entries[i].dtbclk_mhz = def_max.dtbclk_mhz;
+	bw_params->clk_table.num_entries_per_clk.num_dcfclk_levels = clock_table->NumDcfClkLevelsEnabled;
+	bw_params->clk_table.num_entries_per_clk.num_dispclk_levels = clock_table->NumDispClkLevelsEnabled;
+	bw_params->clk_table.num_entries_per_clk.num_dppclk_levels = clock_table->NumDispClkLevelsEnabled;
+	bw_params->clk_table.num_entries_per_clk.num_fclk_levels = clock_table->NumFclkLevelsEnabled;
+	bw_params->clk_table.num_entries_per_clk.num_memclk_levels = clock_table->NumMemPstatesEnabled;
+	bw_params->clk_table.num_entries_per_clk.num_socclk_levels = clock_table->NumSocClkLevelsEnabled;
+
+	/*
+	 * Set any 0 clocks to max default setting. Not an issue for
+	 * power since we aren't doing switching in such case anyway
+	 */
+	for (i = 0; i < bw_params->clk_table.num_entries; i++) {
+		if (!bw_params->clk_table.entries[i].fclk_mhz) {
+			bw_params->clk_table.entries[i].fclk_mhz = def_max.fclk_mhz;
+			bw_params->clk_table.entries[i].memclk_mhz = def_max.memclk_mhz;
+			bw_params->clk_table.entries[i].voltage = def_max.voltage;
+		}
+		if (!bw_params->clk_table.entries[i].dcfclk_mhz)
+			bw_params->clk_table.entries[i].dcfclk_mhz = def_max.dcfclk_mhz;
+		if (!bw_params->clk_table.entries[i].socclk_mhz)
+			bw_params->clk_table.entries[i].socclk_mhz = def_max.socclk_mhz;
+		if (!bw_params->clk_table.entries[i].dispclk_mhz)
+			bw_params->clk_table.entries[i].dispclk_mhz = def_max.dispclk_mhz;
+		if (!bw_params->clk_table.entries[i].dppclk_mhz)
+			bw_params->clk_table.entries[i].dppclk_mhz = def_max.dppclk_mhz;
+		if (!bw_params->clk_table.entries[i].fclk_mhz)
+			bw_params->clk_table.entries[i].fclk_mhz = def_max.fclk_mhz;
+		if (!bw_params->clk_table.entries[i].phyclk_mhz)
+			bw_params->clk_table.entries[i].phyclk_mhz = def_max.phyclk_mhz;
+		if (!bw_params->clk_table.entries[i].phyclk_d18_mhz)
+			bw_params->clk_table.entries[i].phyclk_d18_mhz = def_max.phyclk_d18_mhz;
+		if (!bw_params->clk_table.entries[i].dtbclk_mhz)
+			bw_params->clk_table.entries[i].dtbclk_mhz = def_max.dtbclk_mhz;
+	}
+	ASSERT(bw_params->clk_table.entries[i-1].dcfclk_mhz);
+	bw_params->vram_type = bios_info->memory_type;
+	bw_params->dram_channel_width_bytes = bios_info->memory_type == 0x22 ? 8 : 4;
+	bw_params->num_channels = bios_info->ma_channel_number ? bios_info->ma_channel_number : 4;
+
+	for (i = 0; i < WM_SET_COUNT; i++) {
+		bw_params->wm_table.entries[i].wm_inst = i;
+
+		if (i >= bw_params->clk_table.num_entries) {
+			bw_params->wm_table.entries[i].valid = false;
+			continue;
+		}
+
+		bw_params->wm_table.entries[i].wm_type = WM_TYPE_PSTATE_CHG;
+		bw_params->wm_table.entries[i].valid = true;
+	}
+}
+
+static void dcn35_set_low_power_state(struct clk_mgr *clk_mgr_base)
+{
+	int display_count;
+	struct dc *dc = clk_mgr_base->ctx->dc;
+	struct dc_state *context = dc->current_state;
+
+	if (clk_mgr_base->clks.pwr_state != DCN_PWR_STATE_LOW_POWER) {
+		display_count = dcn35_get_active_display_cnt_wa(dc, context, NULL);
+		/* if we can go lower, go lower */
+		if (display_count == 0)
+			clk_mgr_base->clks.pwr_state = DCN_PWR_STATE_LOW_POWER;
+	}
+}
+
+static void dcn35_exit_low_power_state(struct clk_mgr *clk_mgr_base)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+
+	//SMU optimization is performed part of low power state exit.
+	dcn35_smu_exit_low_power_state(clk_mgr);
+
+}
+
+static bool dcn35_is_ips_supported(struct clk_mgr *clk_mgr_base)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+
+	return dcn35_smu_get_ips_supported(clk_mgr) ? true : false;
+}
+
+static void dcn35_init_clocks_fpga(struct clk_mgr *clk_mgr)
+{
+	init_clk_states(clk_mgr);
+
+/* TODO: Implement the functions and remove the ifndef guard */
+}
+
+static void dcn35_update_clocks_fpga(struct clk_mgr *clk_mgr,
+		struct dc_state *context,
+		bool safe_to_lower)
+{
+	struct clk_mgr_internal *clk_mgr_int = TO_CLK_MGR_INTERNAL(clk_mgr);
+	struct dc_clocks *new_clocks = &context->bw_ctx.bw.dcn.clk;
+	int fclk_adj = new_clocks->fclk_khz;
+
+	/* TODO: remove this after correctly set by DML */
+	new_clocks->dcfclk_khz = 400000;
+	new_clocks->socclk_khz = 400000;
+
+	/* Min fclk = 1.2GHz since all the extra scemi logic seems to run off of it */
+	//int fclk_adj = new_clocks->fclk_khz > 1200000 ? new_clocks->fclk_khz : 1200000;
+	new_clocks->fclk_khz = 4320000;
+
+	if (should_set_clock(safe_to_lower, new_clocks->phyclk_khz, clk_mgr->clks.phyclk_khz)) {
+		clk_mgr->clks.phyclk_khz = new_clocks->phyclk_khz;
+	}
+
+	if (should_set_clock(safe_to_lower, new_clocks->dcfclk_khz, clk_mgr->clks.dcfclk_khz)) {
+		clk_mgr->clks.dcfclk_khz = new_clocks->dcfclk_khz;
+	}
+
+	if (should_set_clock(safe_to_lower,
+			new_clocks->dcfclk_deep_sleep_khz, clk_mgr->clks.dcfclk_deep_sleep_khz)) {
+		clk_mgr->clks.dcfclk_deep_sleep_khz = new_clocks->dcfclk_deep_sleep_khz;
+	}
+
+	if (should_set_clock(safe_to_lower, new_clocks->socclk_khz, clk_mgr->clks.socclk_khz)) {
+		clk_mgr->clks.socclk_khz = new_clocks->socclk_khz;
+	}
+
+	if (should_set_clock(safe_to_lower, new_clocks->dramclk_khz, clk_mgr->clks.dramclk_khz)) {
+		clk_mgr->clks.dramclk_khz = new_clocks->dramclk_khz;
+	}
+
+	if (should_set_clock(safe_to_lower, new_clocks->dppclk_khz, clk_mgr->clks.dppclk_khz)) {
+		clk_mgr->clks.dppclk_khz = new_clocks->dppclk_khz;
+	}
+
+	if (should_set_clock(safe_to_lower, fclk_adj, clk_mgr->clks.fclk_khz)) {
+		clk_mgr->clks.fclk_khz = fclk_adj;
+	}
+
+	if (should_set_clock(safe_to_lower, new_clocks->dispclk_khz, clk_mgr->clks.dispclk_khz)) {
+		clk_mgr->clks.dispclk_khz = new_clocks->dispclk_khz;
+	}
+
+	/* Both fclk and ref_dppclk run on the same scemi clock.
+	 * So take the higher value since the DPP DTO is typically programmed
+	 * such that max dppclk is 1:1 with ref_dppclk.
+	 */
+	if (clk_mgr->clks.fclk_khz > clk_mgr->clks.dppclk_khz)
+		clk_mgr->clks.dppclk_khz = clk_mgr->clks.fclk_khz;
+	if (clk_mgr->clks.dppclk_khz > clk_mgr->clks.fclk_khz)
+		clk_mgr->clks.fclk_khz = clk_mgr->clks.dppclk_khz;
+
+	// Both fclk and ref_dppclk run on the same scemi clock.
+	clk_mgr_int->dccg->ref_dppclk = clk_mgr->clks.fclk_khz;
+
+	/* TODO: set dtbclk in correct place */
+	clk_mgr->clks.dtbclk_en = true;
+	dm_set_dcn_clocks(clk_mgr->ctx, &clk_mgr->clks);
+	dcn35_update_clocks_update_dpp_dto(clk_mgr_int, context, safe_to_lower);
+
+	dcn35_update_clocks_update_dtb_dto(clk_mgr_int, context, clk_mgr->clks.ref_dtbclk_khz);
+}
+
+static unsigned int dcn35_get_max_clock_khz(struct clk_mgr *clk_mgr_base, enum clk_type clk_type)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+
+	unsigned int num_clk_levels;
+
+	switch (clk_type) {
+	case CLK_TYPE_DISPCLK:
+		num_clk_levels = clk_mgr->base.bw_params->clk_table.num_entries_per_clk.num_dispclk_levels;
+		return num_clk_levels ?
+				clk_mgr->base.bw_params->clk_table.entries[num_clk_levels - 1].dispclk_mhz * 1000 :
+				clk_mgr->base.boot_snapshot.dispclk;
+	case CLK_TYPE_DPPCLK:
+		num_clk_levels = clk_mgr->base.bw_params->clk_table.num_entries_per_clk.num_dppclk_levels;
+		return num_clk_levels ?
+				clk_mgr->base.bw_params->clk_table.entries[num_clk_levels - 1].dppclk_mhz * 1000 :
+				clk_mgr->base.boot_snapshot.dppclk;
+	case CLK_TYPE_DSCCLK:
+		num_clk_levels = clk_mgr->base.bw_params->clk_table.num_entries_per_clk.num_dispclk_levels;
+		return num_clk_levels ?
+				clk_mgr->base.bw_params->clk_table.entries[num_clk_levels - 1].dispclk_mhz * 1000 / 3 :
+				clk_mgr->base.boot_snapshot.dispclk / 3;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static struct clk_mgr_funcs dcn35_funcs = {
+	.get_dp_ref_clk_frequency = dce12_get_dp_ref_freq_khz,
+	.get_dtb_ref_clk_frequency = dcn31_get_dtb_ref_freq_khz,
+	.update_clocks = dcn35_update_clocks,
+	.init_clocks = dcn35_init_clocks,
+	.enable_pme_wa = dcn35_enable_pme_wa,
+	.are_clock_states_equal = dcn35_are_clock_states_equal,
+	.notify_wm_ranges = dcn35_notify_wm_ranges,
+	.set_low_power_state = dcn35_set_low_power_state,
+	.exit_low_power_state = dcn35_exit_low_power_state,
+	.is_ips_supported = dcn35_is_ips_supported,
+	.get_max_clock_khz = dcn35_get_max_clock_khz,
+};
+
+struct clk_mgr_funcs dcn35_fpga_funcs = {
+	.get_dp_ref_clk_frequency = dce12_get_dp_ref_freq_khz,
+	.update_clocks = dcn35_update_clocks_fpga,
+	.init_clocks = dcn35_init_clocks_fpga,
+	.get_dtb_ref_clk_frequency = dcn31_get_dtb_ref_freq_khz,
+};
+
+static void translate_to_DpmClocks_t_dcn35(struct dcn351_smu_dpm_clks *smu_dpm_clks_a,
+		struct dcn35_smu_dpm_clks *smu_dpm_clks_b)
+{
+	/*translate two structures and only take need clock tables*/
+	uint8_t i;
+
+	if (smu_dpm_clks_a == NULL || smu_dpm_clks_b == NULL ||
+		smu_dpm_clks_a->dpm_clks == NULL || smu_dpm_clks_b->dpm_clks == NULL)
+		return;
+
+	for (i = 0; i < NUM_DCFCLK_DPM_LEVELS; i++)
+		smu_dpm_clks_b->dpm_clks->DcfClocks[i] = smu_dpm_clks_a->dpm_clks->DcfClocks[i];
+
+	for (i = 0; i < NUM_DISPCLK_DPM_LEVELS; i++)
+		smu_dpm_clks_b->dpm_clks->DispClocks[i] = smu_dpm_clks_a->dpm_clks->DispClocks[i];
+
+	for (i = 0; i < NUM_DPPCLK_DPM_LEVELS; i++)
+		smu_dpm_clks_b->dpm_clks->DppClocks[i] = smu_dpm_clks_a->dpm_clks->DppClocks[i];
+
+	for (i = 0; i < NUM_FCLK_DPM_LEVELS; i++) {
+		smu_dpm_clks_b->dpm_clks->FclkClocks_Freq[i] = smu_dpm_clks_a->dpm_clks->FclkClocks_Freq[i];
+		smu_dpm_clks_b->dpm_clks->FclkClocks_Voltage[i] = smu_dpm_clks_a->dpm_clks->FclkClocks_Voltage[i];
+	}
+	for (i = 0; i < NUM_MEM_PSTATE_LEVELS; i++) {
+		smu_dpm_clks_b->dpm_clks->MemPstateTable[i].MemClk =
+			smu_dpm_clks_a->dpm_clks->MemPstateTable[i].MemClk;
+		smu_dpm_clks_b->dpm_clks->MemPstateTable[i].UClk =
+			smu_dpm_clks_a->dpm_clks->MemPstateTable[i].UClk;
+		smu_dpm_clks_b->dpm_clks->MemPstateTable[i].Voltage =
+			smu_dpm_clks_a->dpm_clks->MemPstateTable[i].Voltage;
+		smu_dpm_clks_b->dpm_clks->MemPstateTable[i].WckRatio =
+			smu_dpm_clks_a->dpm_clks->MemPstateTable[i].WckRatio;
+	}
+	smu_dpm_clks_b->dpm_clks->MaxGfxClk = smu_dpm_clks_a->dpm_clks->MaxGfxClk;
+	smu_dpm_clks_b->dpm_clks->MinGfxClk = smu_dpm_clks_a->dpm_clks->MinGfxClk;
+	smu_dpm_clks_b->dpm_clks->NumDcfClkLevelsEnabled =
+		smu_dpm_clks_a->dpm_clks->NumDcfClkLevelsEnabled;
+	smu_dpm_clks_b->dpm_clks->NumDispClkLevelsEnabled =
+		smu_dpm_clks_a->dpm_clks->NumDispClkLevelsEnabled;
+	smu_dpm_clks_b->dpm_clks->NumFclkLevelsEnabled =
+		smu_dpm_clks_a->dpm_clks->NumFclkLevelsEnabled;
+	smu_dpm_clks_b->dpm_clks->NumMemPstatesEnabled =
+		smu_dpm_clks_a->dpm_clks->NumMemPstatesEnabled;
+	smu_dpm_clks_b->dpm_clks->NumSocClkLevelsEnabled =
+		smu_dpm_clks_a->dpm_clks->NumSocClkLevelsEnabled;
+
+	for (i = 0; i < NUM_SOC_VOLTAGE_LEVELS; i++) {
+		smu_dpm_clks_b->dpm_clks->SocClocks[i] = smu_dpm_clks_a->dpm_clks->SocClocks[i];
+		smu_dpm_clks_b->dpm_clks->SocVoltage[i] = smu_dpm_clks_a->dpm_clks->SocVoltage[i];
+	}
+}
+void dcn35_clk_mgr_construct(
+		struct dc_context *ctx,
+		struct clk_mgr_dcn35 *clk_mgr,
+		struct pp_smu_funcs *pp_smu,
+		struct dccg *dccg)
+{
+	struct dcn35_smu_dpm_clks smu_dpm_clks = { 0 };
+	struct dcn351_smu_dpm_clks smu_dpm_clks_dcn351 = { 0 };
+	clk_mgr->base.base.ctx = ctx;
+	clk_mgr->base.base.funcs = &dcn35_funcs;
+
+	clk_mgr->base.pp_smu = pp_smu;
+
+	clk_mgr->base.dccg = dccg;
+	clk_mgr->base.dfs_bypass_disp_clk = 0;
+
+	clk_mgr->base.dprefclk_ss_percentage = 0;
+	clk_mgr->base.dprefclk_ss_divider = 1000;
+	clk_mgr->base.ss_on_dprefclk = false;
+	clk_mgr->base.dfs_ref_freq_khz = 48000;
+	if (ctx->dce_version != DCN_VERSION_3_51) {
+		clk_mgr->base.regs = &clk_mgr_regs_dcn35;
+		clk_mgr->base.clk_mgr_shift = &clk_mgr_shift_dcn35;
+		clk_mgr->base.clk_mgr_mask = &clk_mgr_mask_dcn35;
+	}
+
+
+	clk_mgr->smu_wm_set.wm_set = (struct dcn35_watermarks *)dm_helpers_allocate_gpu_mem(
+				clk_mgr->base.base.ctx,
+				DC_MEM_ALLOC_TYPE_GART,
+				sizeof(struct dcn35_watermarks),
+				&clk_mgr->smu_wm_set.mc_address.quad_part);
+
+	if (!clk_mgr->smu_wm_set.wm_set) {
+		clk_mgr->smu_wm_set.wm_set = &dummy_wms;
+		clk_mgr->smu_wm_set.mc_address.quad_part = 0;
+	}
+	ASSERT(clk_mgr->smu_wm_set.wm_set);
+
+	smu_dpm_clks.dpm_clks = (DpmClocks_t_dcn35 *)dm_helpers_allocate_gpu_mem(
+				clk_mgr->base.base.ctx,
+				DC_MEM_ALLOC_TYPE_GART,
+				sizeof(DpmClocks_t_dcn35),
+				&smu_dpm_clks.mc_address.quad_part);
+	if (smu_dpm_clks.dpm_clks == NULL) {
+		smu_dpm_clks.dpm_clks = &dummy_clocks;
+		smu_dpm_clks.mc_address.quad_part = 0;
+	}
+	ASSERT(smu_dpm_clks.dpm_clks);
+
+	if (ctx->dce_version == DCN_VERSION_3_51) {
+		smu_dpm_clks_dcn351.dpm_clks = (DpmClocks_t_dcn351 *)dm_helpers_allocate_gpu_mem(
+				clk_mgr->base.base.ctx,
+				DC_MEM_ALLOC_TYPE_GART,
+				sizeof(DpmClocks_t_dcn351),
+				&smu_dpm_clks_dcn351.mc_address.quad_part);
+		if (smu_dpm_clks_dcn351.dpm_clks == NULL) {
+			smu_dpm_clks_dcn351.dpm_clks = &dummy_clocks_dcn351;
+			smu_dpm_clks_dcn351.mc_address.quad_part = 0;
+		}
+	}
+
+	clk_mgr->base.smu_ver = dcn35_smu_get_smu_version(&clk_mgr->base);
+
+	if (clk_mgr->base.smu_ver)
+		clk_mgr->base.smu_present = true;
+
+	/* TODO: Check we get what we expect during bringup */
+	clk_mgr->base.base.dentist_vco_freq_khz = get_vco_frequency_from_reg(&clk_mgr->base);
+
+	if (ctx->dc_bios->integrated_info->memory_type == LpDdr5MemType) {
+		dcn35_bw_params.wm_table = lpddr5_wm_table;
+	} else {
+		dcn35_bw_params.wm_table = ddr5_wm_table;
+	}
+	/* Saved clocks configured at boot for debug purposes */
+	dcn35_save_clk_registers(&clk_mgr->base.base.boot_snapshot, clk_mgr);
+
+	clk_mgr->base.base.dprefclk_khz = dcn35_smu_get_dprefclk(&clk_mgr->base);
+	clk_mgr->base.base.clks.ref_dtbclk_khz = 600000;
+
+	dce_clock_read_ss_info(&clk_mgr->base);
+	/*when clk src is from FCH, it could have ss, same clock src as DPREF clk*/
+
+	dcn35_read_ss_info_from_lut(&clk_mgr->base);
+
+	clk_mgr->base.base.bw_params = &dcn35_bw_params;
+
+	if (clk_mgr->base.base.ctx->dc->debug.pstate_enabled) {
+		int i;
+		if (ctx->dce_version == DCN_VERSION_3_51) {
+			dcn351_get_dpm_table_from_smu(&clk_mgr->base, &smu_dpm_clks_dcn351);
+			translate_to_DpmClocks_t_dcn35(&smu_dpm_clks_dcn351, &smu_dpm_clks);
+		} else
+			dcn35_get_dpm_table_from_smu(&clk_mgr->base, &smu_dpm_clks);
+		DC_LOG_SMU("NumDcfClkLevelsEnabled: %d\n"
+				   "NumDispClkLevelsEnabled: %d\n"
+				   "NumSocClkLevelsEnabled: %d\n"
+				   "VcnClkLevelsEnabled: %d\n"
+				   "FClkLevelsEnabled: %d\n"
+				   "NumMemPstatesEnabled: %d\n"
+				   "MinGfxClk: %d\n"
+				   "MaxGfxClk: %d\n",
+				   smu_dpm_clks.dpm_clks->NumDcfClkLevelsEnabled,
+				   smu_dpm_clks.dpm_clks->NumDispClkLevelsEnabled,
+				   smu_dpm_clks.dpm_clks->NumSocClkLevelsEnabled,
+				   smu_dpm_clks.dpm_clks->VcnClkLevelsEnabled,
+				   smu_dpm_clks.dpm_clks->NumFclkLevelsEnabled,
+				   smu_dpm_clks.dpm_clks->NumMemPstatesEnabled,
+				   smu_dpm_clks.dpm_clks->MinGfxClk,
+				   smu_dpm_clks.dpm_clks->MaxGfxClk);
+		for (i = 0; i < smu_dpm_clks.dpm_clks->NumDcfClkLevelsEnabled; i++) {
+			DC_LOG_SMU("smu_dpm_clks.dpm_clks->DcfClocks[%d] = %d\n",
+					   i,
+					   smu_dpm_clks.dpm_clks->DcfClocks[i]);
+		}
+		for (i = 0; i < smu_dpm_clks.dpm_clks->NumDispClkLevelsEnabled; i++) {
+			DC_LOG_SMU("smu_dpm_clks.dpm_clks->DispClocks[%d] = %d\n",
+					   i, smu_dpm_clks.dpm_clks->DispClocks[i]);
+		}
+		for (i = 0; i < smu_dpm_clks.dpm_clks->NumSocClkLevelsEnabled; i++) {
+			DC_LOG_SMU("smu_dpm_clks.dpm_clks->SocClocks[%d] = %d\n",
+					   i, smu_dpm_clks.dpm_clks->SocClocks[i]);
+		}
+		for (i = 0; i < smu_dpm_clks.dpm_clks->NumFclkLevelsEnabled; i++) {
+			DC_LOG_SMU("smu_dpm_clks.dpm_clks->FclkClocks_Freq[%d] = %d\n",
+					   i, smu_dpm_clks.dpm_clks->FclkClocks_Freq[i]);
+			DC_LOG_SMU("smu_dpm_clks.dpm_clks->FclkClocks_Voltage[%d] = %d\n",
+					   i, smu_dpm_clks.dpm_clks->FclkClocks_Voltage[i]);
+		}
+		for (i = 0; i < smu_dpm_clks.dpm_clks->NumSocClkLevelsEnabled; i++)
+			DC_LOG_SMU("smu_dpm_clks.dpm_clks->SocVoltage[%d] = %d\n",
+					   i, smu_dpm_clks.dpm_clks->SocVoltage[i]);
+
+		for (i = 0; i < smu_dpm_clks.dpm_clks->NumMemPstatesEnabled; i++) {
+			DC_LOG_SMU("smu_dpm_clks.dpm_clks.MemPstateTable[%d].UClk = %d\n"
+					   "smu_dpm_clks.dpm_clks->MemPstateTable[%d].MemClk= %d\n"
+					   "smu_dpm_clks.dpm_clks->MemPstateTable[%d].Voltage = %d\n",
+					   i, smu_dpm_clks.dpm_clks->MemPstateTable[i].UClk,
+					   i, smu_dpm_clks.dpm_clks->MemPstateTable[i].MemClk,
+					   i, smu_dpm_clks.dpm_clks->MemPstateTable[i].Voltage);
+		}
+
+		if (ctx->dc_bios->integrated_info && ctx->dc->config.use_default_clock_table == false) {
+			dcn35_clk_mgr_helper_populate_bw_params(
+					&clk_mgr->base,
+					ctx->dc_bios->integrated_info,
+					smu_dpm_clks.dpm_clks);
+		}
+	}
+
+	if (smu_dpm_clks.dpm_clks && smu_dpm_clks.mc_address.quad_part != 0)
+		dm_helpers_free_gpu_mem(clk_mgr->base.base.ctx, DC_MEM_ALLOC_TYPE_GART,
+				smu_dpm_clks.dpm_clks);
+
+	if (smu_dpm_clks_dcn351.dpm_clks && smu_dpm_clks_dcn351.mc_address.quad_part != 0)
+		dm_helpers_free_gpu_mem(clk_mgr->base.base.ctx, DC_MEM_ALLOC_TYPE_GART,
+				smu_dpm_clks_dcn351.dpm_clks);
+
+	if (ctx->dc->config.disable_ips != DMUB_IPS_DISABLE_ALL) {
+		bool ips_support = false;
+
+		/*avoid call pmfw at init*/
+		ips_support = dcn35_smu_get_ips_supported(&clk_mgr->base);
+		if (ips_support) {
+			ctx->dc->debug.ignore_pg = false;
+			ctx->dc->debug.disable_dpp_power_gate = false;
+			ctx->dc->debug.disable_hubp_power_gate = false;
+			ctx->dc->debug.disable_dsc_power_gate = false;
+
+			/* Disable dynamic IPS2 in older PMFW (93.12) for Z8 interop. */
+			if (ctx->dc->config.disable_ips == DMUB_IPS_ENABLE &&
+			    ctx->dce_version != DCN_VERSION_3_51 &&
+			    ((clk_mgr->base.smu_ver & 0x00FFFFFF) <= 0x005d0c00))
+				ctx->dc->config.disable_ips = DMUB_IPS_RCG_IN_ACTIVE_IPS2_IN_OFF;
+		} else {
+			/*let's reset the config control flag*/
+			ctx->dc->config.disable_ips = DMUB_IPS_DISABLE_ALL; /*pmfw not support it, disable it all*/
+		}
+	}
+}
+
+void dcn35_clk_mgr_destroy(struct clk_mgr_internal *clk_mgr_int)
+{
+	struct clk_mgr_dcn35 *clk_mgr = TO_CLK_MGR_DCN35(clk_mgr_int);
+
+	if (clk_mgr->smu_wm_set.wm_set && clk_mgr->smu_wm_set.mc_address.quad_part != 0)
+		dm_helpers_free_gpu_mem(clk_mgr_int->base.ctx, DC_MEM_ALLOC_TYPE_FRAME_BUFFER,
+				clk_mgr->smu_wm_set.wm_set);
+}
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn35_clk_mgr.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn35_clk_mgr.h
new file mode 100644
index 000000000000..a12a9bf90806
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn35_clk_mgr.h
@@ -0,0 +1,67 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: AMD
+ *
+ */
+
+#ifndef __DCN35_CLK_MGR_H__
+#define __DCN35_CLK_MGR_H__
+#include "clk_mgr_internal.h"
+
+#define NUM_CLOCK_SOURCES   5
+
+struct dcn35_watermarks;
+
+struct dcn35_smu_watermark_set {
+	struct dcn35_watermarks *wm_set;
+	union large_integer mc_address;
+};
+
+struct dcn35_ss_info_table {
+	uint32_t ss_divider;
+	uint32_t ss_percentage[NUM_CLOCK_SOURCES];
+};
+
+struct clk_mgr_dcn35 {
+	struct clk_mgr_internal base;
+	struct dcn35_smu_watermark_set smu_wm_set;
+};
+
+bool dcn35_are_clock_states_equal(struct dc_clocks *a,
+		struct dc_clocks *b);
+void dcn35_init_clocks(struct clk_mgr *clk_mgr);
+void dcn35_update_clocks(struct clk_mgr *clk_mgr_base,
+			struct dc_state *context,
+			bool safe_to_lower);
+
+void dcn35_clk_mgr_construct(struct dc_context *ctx,
+		struct clk_mgr_dcn35 *clk_mgr,
+		struct pp_smu_funcs *pp_smu,
+		struct dccg *dccg);
+
+void dcn35_clk_mgr_destroy(struct clk_mgr_internal *clk_mgr_int);
+
+void dcn351_clk_mgr_construct(struct dc_context *ctx,
+		struct clk_mgr_dcn35 *clk_mgr,
+		struct pp_smu_funcs *pp_smu,
+		struct dccg *dccg);
+#endif //__DCN35_CLK_MGR_H__
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn35_smu.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn35_smu.c
new file mode 100644
index 000000000000..604d256cb47a
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn35_smu.c
@@ -0,0 +1,508 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: AMD
+ *
+ */
+
+
+
+#include "core_types.h"
+#include "clk_mgr_internal.h"
+#include "reg_helper.h"
+#include "dm_helpers.h"
+#include "dcn35_smu.h"
+
+#include "mp/mp_14_0_0_offset.h"
+#include "mp/mp_14_0_0_sh_mask.h"
+
+/* TODO: Use the real headers when they're correct */
+#define MP1_BASE__INST0_SEG0                       0x00016000
+#define MP1_BASE__INST0_SEG1                       0x0243FC00
+#define MP1_BASE__INST0_SEG2                       0x00DC0000
+#define MP1_BASE__INST0_SEG3                       0x00E00000
+#define MP1_BASE__INST0_SEG4                       0x00E40000
+#define MP1_BASE__INST0_SEG5                       0
+
+#ifdef BASE_INNER
+#undef BASE_INNER
+#endif
+
+#define BASE_INNER(seg) MP1_BASE__INST0_SEG ## seg
+
+#define BASE(seg) BASE_INNER(seg)
+
+#define REG(reg_name) (BASE(reg##reg_name##_BASE_IDX) + reg##reg_name)
+
+#define FN(reg_name, field) \
+	FD(reg_name##__##field)
+
+#include "logger_types.h"
+#undef DC_LOGGER
+#define DC_LOGGER \
+	CTX->logger
+#define smu_print(str, ...) {DC_LOG_SMU(str, ##__VA_ARGS__); }
+
+#define VBIOSSMC_MSG_TestMessage                  0x1
+#define VBIOSSMC_MSG_GetSmuVersion                0x2
+#define VBIOSSMC_MSG_PowerUpGfx                   0x3
+#define VBIOSSMC_MSG_SetDispclkFreq               0x4
+#define VBIOSSMC_MSG_SetDprefclkFreq              0x5   //Not used. DPRef is constant
+#define VBIOSSMC_MSG_SetDppclkFreq                0x6
+#define VBIOSSMC_MSG_SetHardMinDcfclkByFreq       0x7
+#define VBIOSSMC_MSG_SetMinDeepSleepDcfclk        0x8
+#define VBIOSSMC_MSG_SetPhyclkVoltageByFreq       0x9	//Keep it in case VMIN dees not support phy clk
+#define VBIOSSMC_MSG_GetFclkFrequency             0xA
+#define VBIOSSMC_MSG_SetDisplayCount              0xB   //Not used anymore
+#define VBIOSSMC_MSG_EnableTmdp48MHzRefclkPwrDown 0xC   //To ask PMFW turn off TMDP 48MHz refclk during display off to save power
+#define VBIOSSMC_MSG_UpdatePmeRestore             0xD
+#define VBIOSSMC_MSG_SetVbiosDramAddrHigh         0xE   //Used for WM table txfr
+#define VBIOSSMC_MSG_SetVbiosDramAddrLow          0xF
+#define VBIOSSMC_MSG_TransferTableSmu2Dram        0x10
+#define VBIOSSMC_MSG_TransferTableDram2Smu        0x11
+#define VBIOSSMC_MSG_SetDisplayIdleOptimizations  0x12
+#define VBIOSSMC_MSG_GetDprefclkFreq              0x13
+#define VBIOSSMC_MSG_GetDtbclkFreq                0x14
+#define VBIOSSMC_MSG_AllowZstatesEntry            0x15
+#define VBIOSSMC_MSG_DisallowZstatesEntry     	  0x16
+#define VBIOSSMC_MSG_SetDtbClk                    0x17
+#define VBIOSSMC_MSG_DispIPS2Entry                0x18 ///< Display IPS2 entry, DMU
+#define VBIOSSMC_MSG_DispIPS2Exit                 0x19 ///< Display IPS2 exit, DMU
+#define VBIOSSMC_MSG_DisableLSdma                 0x1A ///< Disable LSDMA; only sent by VBIOS
+#define VBIOSSMC_MSG_DpControllerPhyStatus        0x1B ///< Inform PMFW about the pre conditions for turning SLDO2 on/off . bit[0]==1 precondition is met, bit[1-2] are for DPPHY number
+#define VBIOSSMC_MSG_QueryIPS2Support             0x1C ///< Return 1: support; else not supported
+#define VBIOSSMC_MSG_NotifyHostRouterBW           0x1D
+#define VBIOSSMC_Message_Count                    0x1E
+
+#define VBIOSSMC_Status_BUSY                      0x0
+#define VBIOSSMC_Result_OK                        0x1
+#define VBIOSSMC_Result_Failed                    0xFF
+#define VBIOSSMC_Result_UnknownCmd                0xFE
+#define VBIOSSMC_Result_CmdRejectedPrereq         0xFD
+#define VBIOSSMC_Result_CmdRejectedBusy           0xFC
+
+union dcn35_dpia_host_router_bw {
+	struct {
+		uint32_t hr_id : 16;
+		uint32_t bw_mbps : 16;
+	} bits;
+	uint32_t all;
+};
+
+/*
+ * Function to be used instead of REG_WAIT macro because the wait ends when
+ * the register is NOT EQUAL to zero, and because `the translation in msg_if.h
+ * won't work with REG_WAIT.
+ */
+static uint32_t dcn35_smu_wait_for_response(struct clk_mgr_internal *clk_mgr, unsigned int delay_us, unsigned int max_retries)
+{
+	uint32_t res_val = VBIOSSMC_Status_BUSY;
+
+	do {
+		res_val = REG_READ(MP1_SMN_C2PMSG_91);
+		if (res_val != VBIOSSMC_Status_BUSY)
+			break;
+
+		if (delay_us >= 1000)
+			msleep(delay_us/1000);
+		else if (delay_us > 0)
+			udelay(delay_us);
+
+		if (clk_mgr->base.ctx->dc->debug.disable_timeout)
+			max_retries++;
+	} while (max_retries--);
+
+	return res_val;
+}
+
+static int dcn35_smu_send_msg_with_param(struct clk_mgr_internal *clk_mgr,
+					 unsigned int msg_id,
+					 unsigned int param)
+{
+	uint32_t result;
+
+	result = dcn35_smu_wait_for_response(clk_mgr, 10, 2000000);
+	ASSERT(result == VBIOSSMC_Result_OK);
+
+	if (result != VBIOSSMC_Result_OK) {
+		DC_LOG_WARNING("SMU response after wait: %d, msg id = %d\n", result, msg_id);
+
+		if (result == VBIOSSMC_Status_BUSY)
+			return -1;
+	}
+
+	/* First clear response register */
+	REG_WRITE(MP1_SMN_C2PMSG_91, VBIOSSMC_Status_BUSY);
+
+	/* Set the parameter register for the SMU message, unit is Mhz */
+	REG_WRITE(MP1_SMN_C2PMSG_83, param);
+
+	/* Trigger the message transaction by writing the message ID */
+	REG_WRITE(MP1_SMN_C2PMSG_67, msg_id);
+
+	result = dcn35_smu_wait_for_response(clk_mgr, 10, 2000000);
+
+	if (result == VBIOSSMC_Result_Failed) {
+		if (msg_id == VBIOSSMC_MSG_TransferTableDram2Smu &&
+		    param == TABLE_WATERMARKS)
+			DC_LOG_WARNING("Watermarks table not configured properly by SMU");
+		else
+			ASSERT(0);
+		REG_WRITE(MP1_SMN_C2PMSG_91, VBIOSSMC_Result_OK);
+		DC_LOG_WARNING("SMU response after wait: %d, msg id = %d\n", result, msg_id);
+		return -1;
+	}
+
+	if (IS_SMU_TIMEOUT(result)) {
+		ASSERT(0);
+		result = dcn35_smu_wait_for_response(clk_mgr, 10, 2000000);
+		//dm_helpers_smu_timeout(CTX, msg_id, param, 10 * 200000);
+		DC_LOG_WARNING("SMU response after wait: %d, msg id = %d\n", result, msg_id);
+	}
+
+	return REG_READ(MP1_SMN_C2PMSG_83);
+}
+
+int dcn35_smu_get_smu_version(struct clk_mgr_internal *clk_mgr)
+{
+	return dcn35_smu_send_msg_with_param(
+			clk_mgr,
+			VBIOSSMC_MSG_GetSmuVersion,
+			0);
+}
+
+
+int dcn35_smu_set_dispclk(struct clk_mgr_internal *clk_mgr, int requested_dispclk_khz)
+{
+	int actual_dispclk_set_mhz = -1;
+
+	if (!clk_mgr->smu_present)
+		return requested_dispclk_khz;
+
+	/*  Unit of SMU msg parameter is Mhz */
+	actual_dispclk_set_mhz = dcn35_smu_send_msg_with_param(
+			clk_mgr,
+			VBIOSSMC_MSG_SetDispclkFreq,
+			khz_to_mhz_ceil(requested_dispclk_khz));
+
+	smu_print("requested_dispclk_khz = %d, actual_dispclk_set_mhz: %d\n", requested_dispclk_khz, actual_dispclk_set_mhz);
+	return actual_dispclk_set_mhz * 1000;
+}
+
+int dcn35_smu_set_dprefclk(struct clk_mgr_internal *clk_mgr)
+{
+	int actual_dprefclk_set_mhz = -1;
+
+	if (!clk_mgr->smu_present)
+		return clk_mgr->base.dprefclk_khz;
+
+	actual_dprefclk_set_mhz = dcn35_smu_send_msg_with_param(
+			clk_mgr,
+			VBIOSSMC_MSG_SetDprefclkFreq,
+			khz_to_mhz_ceil(clk_mgr->base.dprefclk_khz));
+
+	/* TODO: add code for programing DP DTO, currently this is down by command table */
+
+	return actual_dprefclk_set_mhz * 1000;
+}
+
+int dcn35_smu_set_hard_min_dcfclk(struct clk_mgr_internal *clk_mgr, int requested_dcfclk_khz)
+{
+	int actual_dcfclk_set_mhz = -1;
+
+	if (!clk_mgr->smu_present)
+		return requested_dcfclk_khz;
+
+	actual_dcfclk_set_mhz = dcn35_smu_send_msg_with_param(
+			clk_mgr,
+			VBIOSSMC_MSG_SetHardMinDcfclkByFreq,
+			khz_to_mhz_ceil(requested_dcfclk_khz));
+
+	smu_print("requested_dcfclk_khz = %d, actual_dcfclk_set_mhz: %d\n", requested_dcfclk_khz, actual_dcfclk_set_mhz);
+
+	return actual_dcfclk_set_mhz * 1000;
+}
+
+int dcn35_smu_set_min_deep_sleep_dcfclk(struct clk_mgr_internal *clk_mgr, int requested_min_ds_dcfclk_khz)
+{
+	int actual_min_ds_dcfclk_mhz = -1;
+
+	if (!clk_mgr->smu_present)
+		return requested_min_ds_dcfclk_khz;
+
+	actual_min_ds_dcfclk_mhz = dcn35_smu_send_msg_with_param(
+			clk_mgr,
+			VBIOSSMC_MSG_SetMinDeepSleepDcfclk,
+			khz_to_mhz_ceil(requested_min_ds_dcfclk_khz));
+
+	smu_print("requested_min_ds_dcfclk_khz = %d, actual_min_ds_dcfclk_mhz: %d\n", requested_min_ds_dcfclk_khz, actual_min_ds_dcfclk_mhz);
+
+	return actual_min_ds_dcfclk_mhz * 1000;
+}
+
+int dcn35_smu_set_dppclk(struct clk_mgr_internal *clk_mgr, int requested_dpp_khz)
+{
+	int actual_dppclk_set_mhz = -1;
+
+	if (!clk_mgr->smu_present)
+		return requested_dpp_khz;
+
+	actual_dppclk_set_mhz = dcn35_smu_send_msg_with_param(
+			clk_mgr,
+			VBIOSSMC_MSG_SetDppclkFreq,
+			khz_to_mhz_ceil(requested_dpp_khz));
+
+	smu_print("requested_dpp_khz = %d, actual_dppclk_set_mhz: %d\n", requested_dpp_khz, actual_dppclk_set_mhz);
+
+	return actual_dppclk_set_mhz * 1000;
+}
+
+void dcn35_smu_set_display_idle_optimization(struct clk_mgr_internal *clk_mgr, uint32_t idle_info)
+{
+	if (!clk_mgr->base.ctx->dc->debug.pstate_enabled)
+		return;
+
+	if (!clk_mgr->smu_present)
+		return;
+
+	//TODO: Work with smu team to define optimization options.
+	dcn35_smu_send_msg_with_param(
+		clk_mgr,
+		VBIOSSMC_MSG_SetDisplayIdleOptimizations,
+		idle_info);
+	smu_print("%s: VBIOSSMC_MSG_SetDisplayIdleOptimizations idle_info  = %x\n", __func__, idle_info);
+}
+
+void dcn35_smu_enable_phy_refclk_pwrdwn(struct clk_mgr_internal *clk_mgr, bool enable)
+{
+	union display_idle_optimization_u idle_info = { 0 };
+
+	if (!clk_mgr->smu_present)
+		return;
+
+	if (enable) {
+		idle_info.idle_info.df_request_disabled = 1;
+		idle_info.idle_info.phy_ref_clk_off = 1;
+	}
+
+	dcn35_smu_send_msg_with_param(
+			clk_mgr,
+			VBIOSSMC_MSG_SetDisplayIdleOptimizations,
+			idle_info.data);
+	smu_print("%s smu_enable_phy_refclk_pwrdwn  = %d\n", __func__, enable ? 1 : 0);
+}
+
+void dcn35_smu_enable_pme_wa(struct clk_mgr_internal *clk_mgr)
+{
+	if (!clk_mgr->smu_present)
+		return;
+
+	dcn35_smu_send_msg_with_param(
+			clk_mgr,
+			VBIOSSMC_MSG_UpdatePmeRestore,
+			0);
+	smu_print("%s: SMC_MSG_UpdatePmeRestore\n", __func__);
+}
+
+void dcn35_smu_set_dram_addr_high(struct clk_mgr_internal *clk_mgr, uint32_t addr_high)
+{
+	if (!clk_mgr->smu_present)
+		return;
+
+	dcn35_smu_send_msg_with_param(clk_mgr,
+			VBIOSSMC_MSG_SetVbiosDramAddrHigh, addr_high);
+}
+
+void dcn35_smu_set_dram_addr_low(struct clk_mgr_internal *clk_mgr, uint32_t addr_low)
+{
+	if (!clk_mgr->smu_present)
+		return;
+
+	dcn35_smu_send_msg_with_param(clk_mgr,
+			VBIOSSMC_MSG_SetVbiosDramAddrLow, addr_low);
+}
+
+void dcn35_smu_transfer_dpm_table_smu_2_dram(struct clk_mgr_internal *clk_mgr)
+{
+	if (!clk_mgr->smu_present)
+		return;
+
+	dcn35_smu_send_msg_with_param(clk_mgr,
+			VBIOSSMC_MSG_TransferTableSmu2Dram, TABLE_DPMCLOCKS);
+}
+
+void dcn35_smu_transfer_wm_table_dram_2_smu(struct clk_mgr_internal *clk_mgr)
+{
+	if (!clk_mgr->smu_present)
+		return;
+
+	dcn35_smu_send_msg_with_param(clk_mgr,
+			VBIOSSMC_MSG_TransferTableDram2Smu, TABLE_WATERMARKS);
+}
+
+void dcn35_smu_set_zstate_support(struct clk_mgr_internal *clk_mgr, enum dcn_zstate_support_state support)
+{
+	unsigned int msg_id, param, retv;
+
+	if (!clk_mgr->smu_present)
+		return;
+
+	switch (support) {
+
+	case DCN_ZSTATE_SUPPORT_ALLOW:
+		msg_id = VBIOSSMC_MSG_AllowZstatesEntry;
+		param = (1 << 10) | (1 << 9) | (1 << 8);
+		smu_print("%s: SMC_MSG_AllowZstatesEntry msg = ALLOW, param = 0x%x\n", __func__, param);
+		break;
+
+	case DCN_ZSTATE_SUPPORT_DISALLOW:
+		msg_id = VBIOSSMC_MSG_AllowZstatesEntry;
+		param = 0;
+		smu_print("%s: SMC_MSG_AllowZstatesEntry msg_id = DISALLOW, param = 0x%x\n",  __func__, param);
+		break;
+
+
+	case DCN_ZSTATE_SUPPORT_ALLOW_Z10_ONLY:
+		msg_id = VBIOSSMC_MSG_AllowZstatesEntry;
+		param = (1 << 10);
+		smu_print("%s: SMC_MSG_AllowZstatesEntry msg = ALLOW_Z10_ONLY, param = 0x%x\n", __func__, param);
+		break;
+
+	case DCN_ZSTATE_SUPPORT_ALLOW_Z8_Z10_ONLY:
+		msg_id = VBIOSSMC_MSG_AllowZstatesEntry;
+		param = (1 << 10) | (1 << 8);
+		smu_print("%s: SMC_MSG_AllowZstatesEntry msg = ALLOW_Z8_Z10_ONLY, param = 0x%x\n", __func__, param);
+		break;
+
+	case DCN_ZSTATE_SUPPORT_ALLOW_Z8_ONLY:
+		msg_id = VBIOSSMC_MSG_AllowZstatesEntry;
+		param = (1 << 8);
+		smu_print("%s: SMC_MSG_AllowZstatesEntry msg = ALLOW_Z8_ONLY, param = 0x%x\n", __func__, param);
+		break;
+
+	default: //DCN_ZSTATE_SUPPORT_UNKNOWN
+		msg_id = VBIOSSMC_MSG_AllowZstatesEntry;
+		param = 0;
+		break;
+	}
+
+
+	retv = dcn35_smu_send_msg_with_param(
+		clk_mgr,
+		msg_id,
+		param);
+	smu_print("%s:  msg_id = %d, param = 0x%x, return = 0x%x\n", __func__, msg_id, param, retv);
+}
+
+int dcn35_smu_get_dprefclk(struct clk_mgr_internal *clk_mgr)
+{
+	int dprefclk;
+
+	if (!clk_mgr->smu_present)
+		return 0;
+
+	dprefclk = dcn35_smu_send_msg_with_param(clk_mgr,
+						 VBIOSSMC_MSG_GetDprefclkFreq,
+						 0);
+
+	smu_print("%s:  SMU DPREF clk  = %d mhz\n",  __func__, dprefclk);
+	return dprefclk * 1000;
+}
+
+int dcn35_smu_get_dtbclk(struct clk_mgr_internal *clk_mgr)
+{
+	int dtbclk;
+
+	if (!clk_mgr->smu_present)
+		return 0;
+
+	dtbclk = dcn35_smu_send_msg_with_param(clk_mgr,
+					       VBIOSSMC_MSG_GetDtbclkFreq,
+					       0);
+
+	smu_print("%s: get_dtbclk  = %dmhz\n", __func__, dtbclk);
+	return dtbclk * 1000;
+}
+/* Arg = 1: Turn DTB on; 0: Turn DTB CLK OFF. when it is on, it is 600MHZ */
+void dcn35_smu_set_dtbclk(struct clk_mgr_internal *clk_mgr, bool enable)
+{
+	if (!clk_mgr->smu_present)
+		return;
+
+	dcn35_smu_send_msg_with_param(
+			clk_mgr,
+			VBIOSSMC_MSG_SetDtbClk,
+			enable);
+	smu_print("%s: smu_set_dtbclk = %d\n", __func__, enable ? 1 : 0);
+}
+
+void dcn35_vbios_smu_enable_48mhz_tmdp_refclk_pwrdwn(struct clk_mgr_internal *clk_mgr, bool enable)
+{
+	if (!clk_mgr->smu_present)
+		return;
+
+	dcn35_smu_send_msg_with_param(
+			clk_mgr,
+			VBIOSSMC_MSG_EnableTmdp48MHzRefclkPwrDown,
+			enable);
+	smu_print("%s: smu_enable_48mhz_tmdp_refclk_pwrdwn = %d\n", __func__, enable ? 1 : 0);
+}
+
+int dcn35_smu_exit_low_power_state(struct clk_mgr_internal *clk_mgr)
+{
+	int retv;
+
+	if (!clk_mgr->smu_present)
+		return 0;
+
+	retv = dcn35_smu_send_msg_with_param(
+		clk_mgr,
+		VBIOSSMC_MSG_DispIPS2Exit,
+		0);
+	smu_print("%s: smu_exit_low_power_state return = %d\n", __func__, retv);
+	return retv;
+}
+
+int dcn35_smu_get_ips_supported(struct clk_mgr_internal *clk_mgr)
+{
+	int retv;
+
+	if (!clk_mgr->smu_present)
+		return 0;
+
+	retv = dcn35_smu_send_msg_with_param(
+			clk_mgr,
+			VBIOSSMC_MSG_QueryIPS2Support,
+			0);
+
+	//smu_print("%s: VBIOSSMC_MSG_QueryIPS2Support return = %x\n", __func__, retv);
+	return retv;
+}
+
+void dcn35_smu_notify_host_router_bw(struct clk_mgr_internal *clk_mgr, uint32_t hr_id, uint32_t bw_kbps)
+{
+	union dcn35_dpia_host_router_bw msg_data = { 0 };
+
+	msg_data.bits.hr_id = hr_id;
+	msg_data.bits.bw_mbps = bw_kbps / 1000;
+
+	dcn35_smu_send_msg_with_param(clk_mgr, VBIOSSMC_MSG_NotifyHostRouterBW, msg_data.all);
+}
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn35_smu.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn35_smu.h
new file mode 100644
index 000000000000..ab9d21ba0c43
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn35_smu.h
@@ -0,0 +1,220 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: AMD
+ *
+ */
+
+#ifndef DAL_DC_35_SMU_H_
+#define DAL_DC_35_SMU_H_
+
+#include "os_types.h"
+
+#ifndef PMFW_DRIVER_IF_H
+#define PMFW_DRIVER_IF_H
+#define PMFW_DRIVER_IF_VERSION 4
+
+typedef enum {
+  DSPCLK_DCFCLK = 0,
+  DSPCLK_DISPCLK,
+  DSPCLK_PIXCLK,
+  DSPCLK_PHYCLK,
+  DSPCLK_COUNT,
+} DSPCLK_e;
+
+typedef struct {
+  uint16_t Freq; // in MHz
+  uint16_t Vid;  // min voltage in SVI3 VID
+} DisplayClockTable_t;
+
+typedef struct {
+  uint16_t MinClock; // This is either DCFCLK or SOCCLK (in MHz)
+  uint16_t MaxClock; // This is either DCFCLK or SOCCLK (in MHz)
+  uint16_t MinMclk;
+  uint16_t MaxMclk;
+
+  uint8_t  WmSetting;
+  uint8_t  WmType;  // Used for normal pstate change or memory retraining
+  uint8_t  Padding[2];
+} WatermarkRowGeneric_t;
+
+#define NUM_WM_RANGES 4
+#define WM_PSTATE_CHG 0
+#define WM_RETRAINING 1
+
+typedef enum {
+  WM_SOCCLK = 0,
+  WM_DCFCLK,
+  WM_COUNT,
+} WM_CLOCK_e;
+
+typedef struct {
+  // Watermarks
+  WatermarkRowGeneric_t WatermarkRow[WM_COUNT][NUM_WM_RANGES];
+
+  uint32_t MmHubPadding[7]; // SMU internal use
+} Watermarks_t;
+
+#define NUM_DCFCLK_DPM_LEVELS   8
+#define NUM_DISPCLK_DPM_LEVELS  8
+#define NUM_DPPCLK_DPM_LEVELS   8
+#define NUM_SOCCLK_DPM_LEVELS   8
+#define NUM_VCN_DPM_LEVELS      8
+#define NUM_SOC_VOLTAGE_LEVELS  8
+#define NUM_VPE_DPM_LEVELS        8
+#define NUM_FCLK_DPM_LEVELS       8
+#define NUM_MEM_PSTATE_LEVELS     4
+
+typedef enum{
+  WCK_RATIO_1_1 = 0,  // DDR5, Wck:ck is always 1:1;
+  WCK_RATIO_1_2,
+  WCK_RATIO_1_4,
+  WCK_RATIO_MAX
+} WCK_RATIO_e;
+
+typedef struct {
+  uint32_t UClk;
+  uint32_t MemClk;
+  uint32_t Voltage;
+  uint8_t  WckRatio;
+  uint8_t  Spare[3];
+} MemPstateTable_t;
+
+//Freq in MHz
+//Voltage in milli volts with 2 fractional bits
+typedef struct {
+  uint32_t DcfClocks[NUM_DCFCLK_DPM_LEVELS];
+  uint32_t DispClocks[NUM_DISPCLK_DPM_LEVELS];
+  uint32_t DppClocks[NUM_DPPCLK_DPM_LEVELS];
+  uint32_t SocClocks[NUM_SOCCLK_DPM_LEVELS];
+  uint32_t VClocks[NUM_VCN_DPM_LEVELS];
+  uint32_t DClocks[NUM_VCN_DPM_LEVELS];
+  uint32_t VPEClocks[NUM_VPE_DPM_LEVELS];
+  uint32_t FclkClocks_Freq[NUM_FCLK_DPM_LEVELS];
+  uint32_t FclkClocks_Voltage[NUM_FCLK_DPM_LEVELS];
+  uint32_t SocVoltage[NUM_SOC_VOLTAGE_LEVELS];
+  MemPstateTable_t MemPstateTable[NUM_MEM_PSTATE_LEVELS];
+
+  uint8_t  NumDcfClkLevelsEnabled;
+  uint8_t  NumDispClkLevelsEnabled; //Applies to both Dispclk and Dppclk
+  uint8_t  NumSocClkLevelsEnabled;
+  uint8_t  VcnClkLevelsEnabled;     //Applies to both Vclk and Dclk
+  uint8_t  VpeClkLevelsEnabled;
+  uint8_t  NumMemPstatesEnabled;
+  uint8_t  NumFclkLevelsEnabled;
+  uint8_t  spare[2];
+
+  uint32_t MinGfxClk;
+  uint32_t MaxGfxClk;
+} DpmClocks_t_dcn35;
+
+typedef struct {
+	uint32_t DcfClocks[NUM_DCFCLK_DPM_LEVELS];
+	uint32_t DispClocks[NUM_DISPCLK_DPM_LEVELS];
+	uint32_t DppClocks[NUM_DPPCLK_DPM_LEVELS];
+	uint32_t SocClocks[NUM_SOCCLK_DPM_LEVELS];
+	uint32_t VClocks0[NUM_VCN_DPM_LEVELS];
+	uint32_t VClocks1[NUM_VCN_DPM_LEVELS];
+	uint32_t DClocks0[NUM_VCN_DPM_LEVELS];
+	uint32_t DClocks1[NUM_VCN_DPM_LEVELS];
+	uint32_t VPEClocks[NUM_VPE_DPM_LEVELS];
+	uint32_t FclkClocks_Freq[NUM_FCLK_DPM_LEVELS];
+	uint32_t FclkClocks_Voltage[NUM_FCLK_DPM_LEVELS];
+	uint32_t SocVoltage[NUM_SOC_VOLTAGE_LEVELS];
+	MemPstateTable_t MemPstateTable[NUM_MEM_PSTATE_LEVELS];
+	uint8_t NumDcfClkLevelsEnabled;
+	uint8_t NumDispClkLevelsEnabled; // Applies to both Dispclk and Dppclk
+	uint8_t NumSocClkLevelsEnabled;
+	uint8_t Vcn0ClkLevelsEnabled; // Applies to both Vclk0 and Dclk0
+	uint8_t Vcn1ClkLevelsEnabled; // Applies to both Vclk1 and Dclk1
+	uint8_t VpeClkLevelsEnabled;
+	uint8_t NumMemPstatesEnabled;
+	uint8_t NumFclkLevelsEnabled;
+	uint32_t MinGfxClk;
+	uint32_t MaxGfxClk;
+} DpmClocks_t_dcn351;
+
+#define TABLE_BIOS_IF            0 // Called by BIOS
+#define TABLE_WATERMARKS         1 // Called by DAL through VBIOS
+#define TABLE_CUSTOM_DPM         2 // Called by Driver
+#define TABLE_SPARE1             3
+#define TABLE_DPMCLOCKS          4 // Called by Driver
+#define TABLE_MOMENTARY_PM       5 // Called by Tools
+#define TABLE_MODERN_STDBY       6 // Called by Tools for Modern Standby Log
+#define TABLE_SMU_METRICS        7 // Called by Driver
+#define TABLE_COUNT              8
+
+#endif
+
+struct dcn35_watermarks {
+  // Watermarks
+  WatermarkRowGeneric_t WatermarkRow[WM_COUNT][NUM_WM_RANGES];
+
+  uint32_t MmHubPadding[7]; // SMU internal use
+};
+
+struct dcn35_smu_dpm_clks {
+	DpmClocks_t_dcn35 *dpm_clks;
+	union large_integer mc_address;
+};
+
+struct dcn351_smu_dpm_clks {
+	DpmClocks_t_dcn351 *dpm_clks;
+	union large_integer mc_address;
+};
+/* TODO: taken from vgh, may not be correct */
+struct display_idle_optimization {
+	unsigned int df_request_disabled : 1;
+	unsigned int phy_ref_clk_off     : 1;
+	unsigned int s0i2_rdy            : 1;
+	unsigned int reserved            : 29;
+};
+
+union display_idle_optimization_u {
+	struct display_idle_optimization idle_info;
+	uint32_t data;
+};
+
+int dcn35_smu_get_smu_version(struct clk_mgr_internal *clk_mgr);
+int dcn35_smu_set_dispclk(struct clk_mgr_internal *clk_mgr, int requested_dispclk_khz);
+int dcn35_smu_set_dprefclk(struct clk_mgr_internal *clk_mgr);
+int dcn35_smu_set_hard_min_dcfclk(struct clk_mgr_internal *clk_mgr, int requested_dcfclk_khz);
+int dcn35_smu_set_min_deep_sleep_dcfclk(struct clk_mgr_internal *clk_mgr, int requested_min_ds_dcfclk_khz);
+int dcn35_smu_set_dppclk(struct clk_mgr_internal *clk_mgr, int requested_dpp_khz);
+void dcn35_smu_set_display_idle_optimization(struct clk_mgr_internal *clk_mgr, uint32_t idle_info);
+void dcn35_smu_enable_phy_refclk_pwrdwn(struct clk_mgr_internal *clk_mgr, bool enable);
+void dcn35_smu_enable_pme_wa(struct clk_mgr_internal *clk_mgr);
+void dcn35_smu_set_dram_addr_high(struct clk_mgr_internal *clk_mgr, uint32_t addr_high);
+void dcn35_smu_set_dram_addr_low(struct clk_mgr_internal *clk_mgr, uint32_t addr_low);
+void dcn35_smu_transfer_dpm_table_smu_2_dram(struct clk_mgr_internal *clk_mgr);
+void dcn35_smu_transfer_wm_table_dram_2_smu(struct clk_mgr_internal *clk_mgr);
+
+void dcn35_smu_set_zstate_support(struct clk_mgr_internal *clk_mgr, enum dcn_zstate_support_state support);
+void dcn35_smu_set_dtbclk(struct clk_mgr_internal *clk_mgr, bool enable);
+void dcn35_vbios_smu_enable_48mhz_tmdp_refclk_pwrdwn(struct clk_mgr_internal *clk_mgr, bool enable);
+
+int dcn35_smu_exit_low_power_state(struct clk_mgr_internal *clk_mgr);
+int dcn35_smu_get_ips_supported(struct clk_mgr_internal *clk_mgr);
+int dcn35_smu_get_dtbclk(struct clk_mgr_internal *clk_mgr);
+int dcn35_smu_get_dprefclk(struct clk_mgr_internal *clk_mgr);
+void dcn35_smu_notify_host_router_bw(struct clk_mgr_internal *clk_mgr, uint32_t hr_id, uint32_t bw_kbps);
+
+#endif /* DAL_DC_35_SMU_H_ */
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn401/dalsmc.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn401/dalsmc.h
new file mode 100644
index 000000000000..2e0d34fd7512
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn401/dalsmc.h
@@ -0,0 +1,55 @@
+// SPDX-License-Identifier: MIT
+//
+// Copyright 2024 Advanced Micro Devices, Inc.
+
+#ifndef DALSMC_H
+#define DALSMC_H
+
+#define DALSMC_VERSION 0x1
+
+// SMU Response Codes:
+#define DALSMC_Result_OK                   0x1
+#define DALSMC_Result_Failed               0xFF
+#define DALSMC_Result_UnknownCmd           0xFE
+#define DALSMC_Result_CmdRejectedPrereq    0xFD
+#define DALSMC_Result_CmdRejectedBusy      0xFC
+
+
+
+// Message Definitions:
+#define DALSMC_MSG_TestMessage                    0x1
+#define DALSMC_MSG_GetSmuVersion                  0x2
+#define DALSMC_MSG_GetDriverIfVersion             0x3
+#define DALSMC_MSG_GetMsgHeaderVersion            0x4
+#define DALSMC_MSG_SetDalDramAddrHigh             0x5
+#define DALSMC_MSG_SetDalDramAddrLow              0x6
+#define DALSMC_MSG_TransferTableSmu2Dram          0x7
+#define DALSMC_MSG_TransferTableDram2Smu          0x8
+#define DALSMC_MSG_SetHardMinByFreq               0x9
+#define DALSMC_MSG_SetHardMaxByFreq               0xA
+#define DALSMC_MSG_GetDpmFreqByIndex              0xB
+#define DALSMC_MSG_GetDcModeMaxDpmFreq            0xC
+#define DALSMC_MSG_SetMinDeepSleepDcfclk          0xD
+#define DALSMC_MSG_NumOfDisplays                  0xE
+#define DALSMC_MSG_SetExternalClientDfCstateAllow 0xF
+#define DALSMC_MSG_BacoAudioD3PME                 0x10
+#define DALSMC_MSG_SetFclkSwitchAllow             0x11
+#define DALSMC_MSG_SetCabForUclkPstate            0x12
+#define DALSMC_MSG_SetWorstCaseUclkLatency        0x13
+#define DALSMC_MSG_DcnExitReset                   0x14
+#define DALSMC_MSG_ReturnHardMinStatus            0x15
+#define DALSMC_MSG_SetAlwaysWaitDmcubResp         0x16
+#define DALSMC_MSG_IndicateDrrStatus              0x17  // PMFW 15811
+#define DALSMC_MSG_ActiveUclkFclk                 0x18
+#define DALSMC_MSG_IdleUclkFclk                   0x19
+#define DALSMC_MSG_SetUclkPstateAllow             0x1A
+#define DALSMC_MSG_SubvpUclkFclk                  0x1B
+#define DALSMC_MSG_GetNumUmcChannels              0x1C
+#define DALSMC_Message_Count                      0x1D
+
+typedef enum {
+  FCLK_SWITCH_DISALLOW,
+  FCLK_SWITCH_ALLOW,
+} FclkSwitchAllow_e;
+
+#endif
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn401/dcn401_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn401/dcn401_clk_mgr.c
new file mode 100644
index 000000000000..306016c1f109
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn401/dcn401_clk_mgr.c
@@ -0,0 +1,1631 @@
+// SPDX-License-Identifier: MIT
+//
+// Copyright 2024 Advanced Micro Devices, Inc.
+
+#include "dccg.h"
+#include "clk_mgr_internal.h"
+#include "dcn401/dcn401_clk_mgr_smu_msg.h"
+#include "dcn20/dcn20_clk_mgr.h"
+#include "dce100/dce_clk_mgr.h"
+#include "dcn31/dcn31_clk_mgr.h"
+#include "dcn32/dcn32_clk_mgr.h"
+#include "dcn401/dcn401_clk_mgr.h"
+#include "reg_helper.h"
+#include "core_types.h"
+#include "dm_helpers.h"
+#include "link_service.h"
+#include "dc_state_priv.h"
+#include "atomfirmware.h"
+
+#include "dcn401_smu14_driver_if.h"
+
+#include "dcn/dcn_4_1_0_offset.h"
+#include "dcn/dcn_4_1_0_sh_mask.h"
+
+#define DCN_BASE__INST0_SEG1                       0x000000C0
+
+#define mmCLK01_CLK0_CLK_PLL_REQ                        0x16E37
+#define mmCLK01_CLK0_CLK0_DFS_CNTL                      0x16E69
+#define mmCLK01_CLK0_CLK1_DFS_CNTL                      0x16E6C
+#define mmCLK01_CLK0_CLK2_DFS_CNTL                      0x16E6F
+#define mmCLK01_CLK0_CLK3_DFS_CNTL                      0x16E72
+#define mmCLK01_CLK0_CLK4_DFS_CNTL                      0x16E75
+#define mmCLK20_CLK2_CLK2_DFS_CNTL                      0x1B051
+
+#define CLK0_CLK_PLL_REQ__FbMult_int_MASK                  0x000001ffUL
+#define CLK0_CLK_PLL_REQ__PllSpineDiv_MASK                 0x0000f000UL
+#define CLK0_CLK_PLL_REQ__FbMult_frac_MASK                 0xffff0000UL
+#define CLK0_CLK_PLL_REQ__FbMult_int__SHIFT                0x00000000
+#define CLK0_CLK_PLL_REQ__PllSpineDiv__SHIFT               0x0000000c
+#define CLK0_CLK_PLL_REQ__FbMult_frac__SHIFT               0x00000010
+
+#undef FN
+#define FN(reg_name, field_name) \
+	clk_mgr->clk_mgr_shift->field_name, clk_mgr->clk_mgr_mask->field_name
+
+#define REG(reg) \
+	(clk_mgr->regs->reg)
+
+#define BASE_INNER(seg) DCN_BASE__INST0_SEG ## seg
+
+#define BASE(seg) BASE_INNER(seg)
+
+#define SR(reg_name)\
+		.reg_name = BASE(reg ## reg_name ## _BASE_IDX) +  \
+					reg ## reg_name
+
+#define CLK_SR_DCN401(reg_name, block, inst)\
+	.reg_name = mm ## block ## _ ## reg_name
+
+static const struct clk_mgr_registers clk_mgr_regs_dcn401 = {
+	CLK_REG_LIST_DCN401()
+};
+
+static const struct clk_mgr_shift clk_mgr_shift_dcn401 = {
+	CLK_COMMON_MASK_SH_LIST_DCN401(__SHIFT)
+};
+
+static const struct clk_mgr_mask clk_mgr_mask_dcn401 = {
+	CLK_COMMON_MASK_SH_LIST_DCN401(_MASK)
+};
+
+#define TO_DCN401_CLK_MGR(clk_mgr)\
+	container_of(clk_mgr, struct dcn401_clk_mgr, base)
+
+static bool dcn401_is_ppclk_dpm_enabled(struct clk_mgr_internal *clk_mgr, PPCLK_e clk)
+{
+	bool ppclk_dpm_enabled = false;
+
+	switch (clk) {
+	case PPCLK_SOCCLK:
+		ppclk_dpm_enabled =
+				clk_mgr->base.bw_params->clk_table.num_entries_per_clk.num_socclk_levels > 1;
+		break;
+	case PPCLK_UCLK:
+		ppclk_dpm_enabled =
+				clk_mgr->base.bw_params->clk_table.num_entries_per_clk.num_memclk_levels > 1;
+		break;
+	case PPCLK_FCLK:
+		ppclk_dpm_enabled =
+				clk_mgr->base.bw_params->clk_table.num_entries_per_clk.num_fclk_levels > 1;
+		break;
+	case PPCLK_DISPCLK:
+		ppclk_dpm_enabled =
+				clk_mgr->base.bw_params->clk_table.num_entries_per_clk.num_dispclk_levels > 1;
+		break;
+	case PPCLK_DPPCLK:
+		ppclk_dpm_enabled =
+				clk_mgr->base.bw_params->clk_table.num_entries_per_clk.num_dppclk_levels > 1;
+		break;
+	case PPCLK_DPREFCLK:
+		ppclk_dpm_enabled = false;
+		break;
+	case PPCLK_DCFCLK:
+		ppclk_dpm_enabled =
+				clk_mgr->base.bw_params->clk_table.num_entries_per_clk.num_dcfclk_levels > 1;
+		break;
+	case PPCLK_DTBCLK:
+		ppclk_dpm_enabled =
+				clk_mgr->base.bw_params->clk_table.num_entries_per_clk.num_dtbclk_levels > 1;
+		break;
+	default:
+		ppclk_dpm_enabled = false;
+	}
+
+	ppclk_dpm_enabled &= clk_mgr->smu_present;
+
+	return ppclk_dpm_enabled;
+}
+
+static bool dcn401_is_ppclk_idle_dpm_enabled(struct clk_mgr_internal *clk_mgr, PPCLK_e clk)
+{
+	bool ppclk_idle_dpm_enabled = false;
+
+	switch (clk) {
+	case PPCLK_UCLK:
+	case PPCLK_FCLK:
+		if (ASICREV_IS_GC_12_0_0_A0(clk_mgr->base.ctx->asic_id.hw_internal_rev) &&
+				clk_mgr->smu_ver >= 0x681800) {
+			ppclk_idle_dpm_enabled = true;
+		} else if (ASICREV_IS_GC_12_0_1_A0(clk_mgr->base.ctx->asic_id.hw_internal_rev) &&
+				clk_mgr->smu_ver >= 0x661300) {
+			ppclk_idle_dpm_enabled = true;
+		}
+		break;
+	default:
+		ppclk_idle_dpm_enabled = false;
+	}
+
+	ppclk_idle_dpm_enabled &= clk_mgr->smu_present;
+
+	return ppclk_idle_dpm_enabled;
+}
+
+static bool dcn401_is_df_throttle_opt_enabled(struct clk_mgr_internal *clk_mgr)
+{
+	bool is_df_throttle_opt_enabled = false;
+
+	if (ASICREV_IS_GC_12_0_1_A0(clk_mgr->base.ctx->asic_id.hw_internal_rev) &&
+			clk_mgr->smu_ver >= 0x663500) {
+		is_df_throttle_opt_enabled = !clk_mgr->base.ctx->dc->debug.force_subvp_df_throttle;
+	}
+
+	is_df_throttle_opt_enabled &= clk_mgr->smu_present;
+
+	return is_df_throttle_opt_enabled;
+}
+
+/* Query SMU for all clock states for a particular clock */
+static void dcn401_init_single_clock(struct clk_mgr_internal *clk_mgr, PPCLK_e clk, unsigned int *entry_0,
+		unsigned int *num_levels)
+{
+	unsigned int i;
+	char *entry_i = (char *)entry_0;
+
+	uint32_t ret = dcn401_smu_get_dpm_freq_by_index(clk_mgr, clk, 0xFF);
+
+	if (ret & (1 << 31))
+		/* fine-grained, only min and max */
+		*num_levels = 2;
+	else
+		/* discrete, a number of fixed states */
+		/* will set num_levels to 0 on failure */
+		*num_levels = ret & 0xFF;
+
+	/* if the initial message failed, num_levels will be 0 */
+	for (i = 0; i < *num_levels && i < ARRAY_SIZE(clk_mgr->base.bw_params->clk_table.entries); i++) {
+		*((unsigned int *)entry_i) = (dcn401_smu_get_dpm_freq_by_index(clk_mgr, clk, i) & 0xFFFF);
+		entry_i += sizeof(clk_mgr->base.bw_params->clk_table.entries[0]);
+	}
+}
+
+static void dcn401_build_wm_range_table(struct clk_mgr *clk_mgr)
+{
+	/* For min clocks use as reported by PM FW and report those as min */
+	uint16_t min_uclk_mhz = clk_mgr->bw_params->clk_table.entries[0].memclk_mhz;
+	uint16_t min_dcfclk_mhz	= clk_mgr->bw_params->clk_table.entries[0].dcfclk_mhz;
+
+	/* Set A - Normal - default values */
+	clk_mgr->bw_params->wm_table.nv_entries[WM_A].valid = true;
+	clk_mgr->bw_params->wm_table.nv_entries[WM_A].pmfw_breakdown.wm_type = WATERMARKS_CLOCK_RANGE;
+	clk_mgr->bw_params->wm_table.nv_entries[WM_A].pmfw_breakdown.min_dcfclk = min_dcfclk_mhz;
+	clk_mgr->bw_params->wm_table.nv_entries[WM_A].pmfw_breakdown.max_dcfclk = 0xFFFF;
+	clk_mgr->bw_params->wm_table.nv_entries[WM_A].pmfw_breakdown.min_uclk = min_uclk_mhz;
+	clk_mgr->bw_params->wm_table.nv_entries[WM_A].pmfw_breakdown.max_uclk = 0xFFFF;
+
+	/* Set B - Unused on dcn4 */
+	clk_mgr->bw_params->wm_table.nv_entries[WM_B].valid = false;
+
+	/* Set 1A - Dummy P-State - P-State latency set to "dummy p-state" value */
+	/* 'DalDummyClockChangeLatencyNs' registry key option set to 0x7FFFFFFF can be used to disable Set C for dummy p-state */
+	if (clk_mgr->ctx->dc->bb_overrides.dummy_clock_change_latency_ns != 0x7FFFFFFF) {
+		clk_mgr->bw_params->wm_table.nv_entries[WM_1A].valid = true;
+		clk_mgr->bw_params->wm_table.nv_entries[WM_1A].pmfw_breakdown.wm_type = WATERMARKS_DUMMY_PSTATE;
+		clk_mgr->bw_params->wm_table.nv_entries[WM_1A].pmfw_breakdown.min_dcfclk = min_dcfclk_mhz;
+		clk_mgr->bw_params->wm_table.nv_entries[WM_1A].pmfw_breakdown.max_dcfclk = 0xFFFF;
+		clk_mgr->bw_params->wm_table.nv_entries[WM_1A].pmfw_breakdown.min_uclk = min_uclk_mhz;
+		clk_mgr->bw_params->wm_table.nv_entries[WM_1A].pmfw_breakdown.max_uclk = 0xFFFF;
+	} else {
+		clk_mgr->bw_params->wm_table.nv_entries[WM_1A].valid = false;
+	}
+
+	/* Set 1B - Unused on dcn4 */
+	clk_mgr->bw_params->wm_table.nv_entries[WM_1B].valid = false;
+}
+
+void dcn401_init_clocks(struct clk_mgr *clk_mgr_base)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+	struct clk_limit_num_entries *num_entries_per_clk;
+	unsigned int i;
+
+	if (!clk_mgr_base->bw_params)
+		return;
+
+	num_entries_per_clk = &clk_mgr_base->bw_params->clk_table.num_entries_per_clk;
+
+	memset(&(clk_mgr_base->clks), 0, sizeof(struct dc_clocks));
+	clk_mgr_base->clks.p_state_change_support = true;
+	clk_mgr_base->clks.prev_p_state_change_support = true;
+	clk_mgr_base->clks.fclk_prev_p_state_change_support = true;
+	clk_mgr->smu_present = false;
+	clk_mgr->dpm_present = false;
+
+	if (!clk_mgr_base->force_smu_not_present && dcn401_smu_get_smu_version(clk_mgr, &clk_mgr->smu_ver))
+		clk_mgr->smu_present = true;
+
+	if (!clk_mgr->smu_present)
+		return;
+
+	dcn401_smu_check_driver_if_version(clk_mgr);
+	dcn401_smu_check_msg_header_version(clk_mgr);
+
+	/* DCFCLK */
+	dcn401_init_single_clock(clk_mgr, PPCLK_DCFCLK,
+			&clk_mgr_base->bw_params->clk_table.entries[0].dcfclk_mhz,
+			&num_entries_per_clk->num_dcfclk_levels);
+	clk_mgr_base->bw_params->dc_mode_limit.dcfclk_mhz = dcn401_smu_get_dc_mode_max_dpm_freq(clk_mgr, PPCLK_DCFCLK);
+	if (num_entries_per_clk->num_dcfclk_levels && clk_mgr_base->bw_params->dc_mode_limit.dcfclk_mhz ==
+			clk_mgr_base->bw_params->clk_table.entries[num_entries_per_clk->num_dcfclk_levels - 1].dcfclk_mhz)
+		clk_mgr_base->bw_params->dc_mode_limit.dcfclk_mhz = 0;
+
+	/* SOCCLK */
+	dcn401_init_single_clock(clk_mgr, PPCLK_SOCCLK,
+					&clk_mgr_base->bw_params->clk_table.entries[0].socclk_mhz,
+					&num_entries_per_clk->num_socclk_levels);
+	clk_mgr_base->bw_params->dc_mode_limit.socclk_mhz = dcn401_smu_get_dc_mode_max_dpm_freq(clk_mgr, PPCLK_SOCCLK);
+	if (num_entries_per_clk->num_socclk_levels && clk_mgr_base->bw_params->dc_mode_limit.socclk_mhz ==
+			clk_mgr_base->bw_params->clk_table.entries[num_entries_per_clk->num_socclk_levels - 1].socclk_mhz)
+		clk_mgr_base->bw_params->dc_mode_limit.socclk_mhz = 0;
+
+	/* DTBCLK */
+	if (!clk_mgr->base.ctx->dc->debug.disable_dtb_ref_clk_switch) {
+		dcn401_init_single_clock(clk_mgr, PPCLK_DTBCLK,
+				&clk_mgr_base->bw_params->clk_table.entries[0].dtbclk_mhz,
+				&num_entries_per_clk->num_dtbclk_levels);
+		clk_mgr_base->bw_params->dc_mode_limit.dtbclk_mhz = dcn401_smu_get_dc_mode_max_dpm_freq(clk_mgr, PPCLK_DTBCLK);
+		if (num_entries_per_clk->num_dtbclk_levels && clk_mgr_base->bw_params->dc_mode_limit.dtbclk_mhz ==
+				clk_mgr_base->bw_params->clk_table.entries[num_entries_per_clk->num_dtbclk_levels - 1].dtbclk_mhz)
+			clk_mgr_base->bw_params->dc_mode_limit.dtbclk_mhz = 0;
+	}
+
+	/* DISPCLK */
+	dcn401_init_single_clock(clk_mgr, PPCLK_DISPCLK,
+			&clk_mgr_base->bw_params->clk_table.entries[0].dispclk_mhz,
+			&num_entries_per_clk->num_dispclk_levels);
+	clk_mgr_base->bw_params->dc_mode_limit.dispclk_mhz = dcn401_smu_get_dc_mode_max_dpm_freq(clk_mgr, PPCLK_DISPCLK);
+	if (num_entries_per_clk->num_dispclk_levels && clk_mgr_base->bw_params->dc_mode_limit.dispclk_mhz ==
+			clk_mgr_base->bw_params->clk_table.entries[num_entries_per_clk->num_dispclk_levels - 1].dispclk_mhz)
+		clk_mgr_base->bw_params->dc_mode_limit.dispclk_mhz = 0;
+
+	/* DPPCLK */
+	dcn401_init_single_clock(clk_mgr, PPCLK_DPPCLK,
+			&clk_mgr_base->bw_params->clk_table.entries[0].dppclk_mhz,
+			&num_entries_per_clk->num_dppclk_levels);
+
+	if (num_entries_per_clk->num_dcfclk_levels &&
+			num_entries_per_clk->num_dtbclk_levels &&
+			num_entries_per_clk->num_dispclk_levels)
+		clk_mgr->dpm_present = true;
+
+	if (clk_mgr_base->ctx->dc->debug.min_disp_clk_khz) {
+		for (i = 0; i < num_entries_per_clk->num_dispclk_levels; i++)
+			if (clk_mgr_base->bw_params->clk_table.entries[i].dispclk_mhz
+					< khz_to_mhz_ceil(clk_mgr_base->ctx->dc->debug.min_disp_clk_khz))
+				clk_mgr_base->bw_params->clk_table.entries[i].dispclk_mhz
+					= khz_to_mhz_ceil(clk_mgr_base->ctx->dc->debug.min_disp_clk_khz);
+	}
+
+	if (clk_mgr_base->ctx->dc->debug.min_dpp_clk_khz) {
+		for (i = 0; i < num_entries_per_clk->num_dppclk_levels; i++)
+			if (clk_mgr_base->bw_params->clk_table.entries[i].dppclk_mhz
+					< khz_to_mhz_ceil(clk_mgr_base->ctx->dc->debug.min_dpp_clk_khz))
+				clk_mgr_base->bw_params->clk_table.entries[i].dppclk_mhz
+					= khz_to_mhz_ceil(clk_mgr_base->ctx->dc->debug.min_dpp_clk_khz);
+	}
+
+	/* Get UCLK, update bounding box */
+	clk_mgr_base->funcs->get_memclk_states_from_smu(clk_mgr_base);
+
+	/* WM range table */
+	dcn401_build_wm_range_table(clk_mgr_base);
+}
+
+bool dcn401_is_dc_mode_present(struct clk_mgr *clk_mgr_base)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+
+	return clk_mgr->smu_present && clk_mgr->dpm_present &&
+			((clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_dcfclk_levels &&
+			clk_mgr_base->bw_params->dc_mode_limit.dcfclk_mhz) ||
+			(clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_dispclk_levels &&
+			clk_mgr_base->bw_params->dc_mode_limit.dispclk_mhz) ||
+			(clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_dtbclk_levels &&
+			clk_mgr_base->bw_params->dc_mode_limit.dtbclk_mhz) ||
+			(clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_fclk_levels &&
+			clk_mgr_base->bw_params->dc_mode_limit.fclk_mhz) ||
+			(clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_memclk_levels &&
+			clk_mgr_base->bw_params->dc_mode_limit.memclk_mhz) ||
+			(clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_socclk_levels &&
+			clk_mgr_base->bw_params->dc_mode_limit.socclk_mhz));
+}
+
+static void dcn401_dump_clk_registers(struct clk_state_registers_and_bypass *regs_and_bypass,
+		struct clk_mgr *clk_mgr_base, struct clk_log_info *log_info)
+{
+		struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+		uint32_t dprefclk_did = 0;
+		uint32_t dcfclk_did = 0;
+		uint32_t dtbclk_did = 0;
+		uint32_t dispclk_did = 0;
+		uint32_t dppclk_did = 0;
+		uint32_t fclk_did = 0;
+		uint32_t target_div = 0;
+
+		/* DFS Slice 0 is used for DISPCLK */
+		dispclk_did = REG_READ(CLK0_CLK0_DFS_CNTL);
+		/* DFS Slice 1 is used for DPPCLK */
+		dppclk_did = REG_READ(CLK0_CLK1_DFS_CNTL);
+		/* DFS Slice 2 is used for DPREFCLK */
+		dprefclk_did = REG_READ(CLK0_CLK2_DFS_CNTL);
+		/* DFS Slice 3 is used for DCFCLK */
+		dcfclk_did = REG_READ(CLK0_CLK3_DFS_CNTL);
+		/* DFS Slice 4 is used for DTBCLK */
+		dtbclk_did = REG_READ(CLK0_CLK4_DFS_CNTL);
+		/* DFS Slice _ is used for FCLK */
+		fclk_did = REG_READ(CLK2_CLK2_DFS_CNTL);
+
+		/* Convert DISPCLK DFS Slice DID to divider*/
+		target_div = dentist_get_divider_from_did(dispclk_did);
+		//Get dispclk in khz
+		regs_and_bypass->dispclk = (DENTIST_DIVIDER_RANGE_SCALE_FACTOR
+				* clk_mgr->base.dentist_vco_freq_khz) / target_div;
+
+		/* Convert DISPCLK DFS Slice DID to divider*/
+		target_div = dentist_get_divider_from_did(dppclk_did);
+		//Get dppclk in khz
+		regs_and_bypass->dppclk = (DENTIST_DIVIDER_RANGE_SCALE_FACTOR
+				* clk_mgr->base.dentist_vco_freq_khz) / target_div;
+
+		/* Convert DPREFCLK DFS Slice DID to divider*/
+		target_div = dentist_get_divider_from_did(dprefclk_did);
+		//Get dprefclk in khz
+		regs_and_bypass->dprefclk = (DENTIST_DIVIDER_RANGE_SCALE_FACTOR
+				* clk_mgr->base.dentist_vco_freq_khz) / target_div;
+
+		/* Convert DCFCLK DFS Slice DID to divider*/
+		target_div = dentist_get_divider_from_did(dcfclk_did);
+		//Get dcfclk in khz
+		regs_and_bypass->dcfclk = (DENTIST_DIVIDER_RANGE_SCALE_FACTOR
+				* clk_mgr->base.dentist_vco_freq_khz) / target_div;
+
+		/* Convert DTBCLK DFS Slice DID to divider*/
+		target_div = dentist_get_divider_from_did(dtbclk_did);
+		//Get dtbclk in khz
+		regs_and_bypass->dtbclk = (DENTIST_DIVIDER_RANGE_SCALE_FACTOR
+				* clk_mgr->base.dentist_vco_freq_khz) / target_div;
+
+		/* Convert DTBCLK DFS Slice DID to divider*/
+		target_div = dentist_get_divider_from_did(fclk_did);
+		//Get fclk in khz
+		regs_and_bypass->fclk = (DENTIST_DIVIDER_RANGE_SCALE_FACTOR
+				* clk_mgr->base.dentist_vco_freq_khz) / target_div;
+}
+
+static bool dcn401_check_native_scaling(struct pipe_ctx *pipe)
+{
+	bool is_native_scaling = false;
+	int width = pipe->plane_state->src_rect.width;
+	int height = pipe->plane_state->src_rect.height;
+
+	if (pipe->stream->timing.h_addressable == width &&
+			pipe->stream->timing.v_addressable == height &&
+			pipe->plane_state->dst_rect.width == width &&
+			pipe->plane_state->dst_rect.height == height)
+		is_native_scaling = true;
+
+	return is_native_scaling;
+}
+
+static void dcn401_auto_dpm_test_log(
+		struct dc_clocks *new_clocks,
+		struct clk_mgr_internal *clk_mgr,
+		struct dc_state *context)
+{
+	unsigned int mall_ss_size_bytes;
+	int dramclk_khz_override, fclk_khz_override, num_fclk_levels;
+
+	struct pipe_ctx *pipe_ctx_list[MAX_PIPES];
+	int active_pipe_count = 0;
+
+	for (int i = 0; i < MAX_PIPES; i++) {
+		struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i];
+
+		if (pipe_ctx->stream && dc_state_get_pipe_subvp_type(context, pipe_ctx) != SUBVP_PHANTOM) {
+			pipe_ctx_list[active_pipe_count] = pipe_ctx;
+			active_pipe_count++;
+		}
+	}
+
+	msleep(5);
+
+	mall_ss_size_bytes = context->bw_ctx.bw.dcn.mall_ss_size_bytes;
+
+	struct clk_log_info log_info = {0};
+	struct clk_state_registers_and_bypass clk_register_dump;
+
+	dcn401_dump_clk_registers(&clk_register_dump, &clk_mgr->base, &log_info);
+
+	// Overrides for these clocks in case there is no p_state change support
+	dramclk_khz_override = new_clocks->dramclk_khz;
+	fclk_khz_override = new_clocks->fclk_khz;
+
+	num_fclk_levels = clk_mgr->base.bw_params->clk_table.num_entries_per_clk.num_fclk_levels - 1;
+
+	if (!new_clocks->p_state_change_support)
+		dramclk_khz_override = clk_mgr->base.bw_params->max_memclk_mhz * 1000;
+
+	if (!new_clocks->fclk_p_state_change_support)
+		fclk_khz_override = clk_mgr->base.bw_params->clk_table.entries[num_fclk_levels].fclk_mhz * 1000;
+
+
+	////////////////////////////////////////////////////////////////////////////
+	//	IMPORTANT: 	When adding more clocks to these logs, do NOT put a newline
+	//	 			anywhere other than at the very end of the string.
+	//
+	//	Formatting example (make sure to have " - " between each entry):
+	//
+	//				AutoDPMTest: clk1:%d - clk2:%d - clk3:%d - clk4:%d\n"
+	////////////////////////////////////////////////////////////////////////////
+	if (active_pipe_count > 0 &&
+		new_clocks->dramclk_khz > 0 &&
+		new_clocks->fclk_khz > 0 &&
+		new_clocks->dcfclk_khz > 0 &&
+		new_clocks->dppclk_khz > 0) {
+
+		uint32_t pix_clk_list[MAX_PIPES] = {0};
+		int p_state_list[MAX_PIPES] = {0};
+		int disp_src_width_list[MAX_PIPES] = {0};
+		int disp_src_height_list[MAX_PIPES] = {0};
+		uint64_t disp_src_refresh_list[MAX_PIPES] = {0};
+		bool is_scaled_list[MAX_PIPES] = {0};
+
+		for (int i = 0; i < active_pipe_count; i++) {
+			struct pipe_ctx *curr_pipe_ctx = pipe_ctx_list[i];
+			uint64_t refresh_rate;
+
+			pix_clk_list[i] = curr_pipe_ctx->stream->timing.pix_clk_100hz;
+			p_state_list[i] = curr_pipe_ctx->p_state_type;
+
+			refresh_rate = (curr_pipe_ctx->stream->timing.pix_clk_100hz * (uint64_t)100 +
+				curr_pipe_ctx->stream->timing.v_total
+				* (uint64_t) curr_pipe_ctx->stream->timing.h_total - (uint64_t)1);
+			refresh_rate = div_u64(refresh_rate, curr_pipe_ctx->stream->timing.v_total);
+			refresh_rate = div_u64(refresh_rate, curr_pipe_ctx->stream->timing.h_total);
+			disp_src_refresh_list[i] = refresh_rate;
+
+			if (curr_pipe_ctx->plane_state) {
+				is_scaled_list[i] = !(dcn401_check_native_scaling(curr_pipe_ctx));
+				disp_src_width_list[i] = curr_pipe_ctx->plane_state->src_rect.width;
+				disp_src_height_list[i] = curr_pipe_ctx->plane_state->src_rect.height;
+			}
+		}
+
+		DC_LOG_AUTO_DPM_TEST("AutoDPMTest: dramclk:%d - fclk:%d - "
+			"dcfclk:%d - dppclk:%d - dispclk_hw:%d - "
+			"dppclk_hw:%d - dprefclk_hw:%d - dcfclk_hw:%d - "
+			"dtbclk_hw:%d - fclk_hw:%d - pix_clk_0:%d - pix_clk_1:%d - "
+			"pix_clk_2:%d - pix_clk_3:%d - mall_ss_size:%d - p_state_type_0:%d - "
+			"p_state_type_1:%d - p_state_type_2:%d - p_state_type_3:%d - "
+			"pix_width_0:%d - pix_height_0:%d - refresh_rate_0:%lld - is_scaled_0:%d - "
+			"pix_width_1:%d - pix_height_1:%d - refresh_rate_1:%lld - is_scaled_1:%d - "
+			"pix_width_2:%d - pix_height_2:%d - refresh_rate_2:%lld - is_scaled_2:%d - "
+			"pix_width_3:%d - pix_height_3:%d - refresh_rate_3:%lld - is_scaled_3:%d - LOG_END\n",
+			dramclk_khz_override,
+			fclk_khz_override,
+			new_clocks->dcfclk_khz,
+			new_clocks->dppclk_khz,
+			clk_register_dump.dispclk,
+			clk_register_dump.dppclk,
+			clk_register_dump.dprefclk,
+			clk_register_dump.dcfclk,
+			clk_register_dump.dtbclk,
+			clk_register_dump.fclk,
+			pix_clk_list[0], pix_clk_list[1], pix_clk_list[3], pix_clk_list[2],
+			mall_ss_size_bytes,
+			p_state_list[0], p_state_list[1], p_state_list[2], p_state_list[3],
+			disp_src_width_list[0], disp_src_height_list[0], disp_src_refresh_list[0], is_scaled_list[0],
+			disp_src_width_list[1], disp_src_height_list[1], disp_src_refresh_list[1], is_scaled_list[1],
+			disp_src_width_list[2], disp_src_height_list[2], disp_src_refresh_list[2], is_scaled_list[2],
+			disp_src_width_list[3], disp_src_height_list[3], disp_src_refresh_list[3], is_scaled_list[3]);
+	}
+}
+
+static void dcn401_update_clocks_update_dtb_dto(struct clk_mgr_internal *clk_mgr,
+			struct dc_state *context,
+			int ref_dtbclk_khz)
+{
+	int i;
+	struct dccg *dccg = clk_mgr->dccg;
+	struct pipe_ctx *otg_master;
+	bool use_hpo_encoder;
+
+
+	for (i = 0; i < context->stream_count; i++) {
+		otg_master = resource_get_otg_master_for_stream(
+				&context->res_ctx, context->streams[i]);
+		ASSERT(otg_master);
+		ASSERT(otg_master->clock_source);
+		ASSERT(otg_master->clock_source->funcs->program_pix_clk);
+		ASSERT(otg_master->stream_res.pix_clk_params.controller_id >= CONTROLLER_ID_D0);
+
+		use_hpo_encoder = dccg->ctx->dc->link_srv->dp_is_128b_132b_signal(otg_master);
+		if (!use_hpo_encoder)
+			continue;
+
+		if (otg_master->stream_res.pix_clk_params.controller_id > CONTROLLER_ID_UNDEFINED)
+			otg_master->clock_source->funcs->program_pix_clk(
+				otg_master->clock_source,
+				&otg_master->stream_res.pix_clk_params,
+				dccg->ctx->dc->link_srv->dp_get_encoding_format(
+					&otg_master->link_config.dp_link_settings),
+				&otg_master->pll_settings);
+	}
+}
+
+static void dcn401_update_clocks_update_dpp_dto(struct clk_mgr_internal *clk_mgr,
+		struct dc_state *context, bool safe_to_lower, int ref_dppclk_khz)
+{
+	int i;
+
+	clk_mgr->dccg->ref_dppclk = ref_dppclk_khz;
+	for (i = 0; i < clk_mgr->base.ctx->dc->res_pool->pipe_count; i++) {
+		int dpp_inst = 0, dppclk_khz, prev_dppclk_khz;
+
+		dppclk_khz = context->res_ctx.pipe_ctx[i].plane_res.bw.dppclk_khz;
+
+		if (context->res_ctx.pipe_ctx[i].plane_res.dpp)
+			dpp_inst = context->res_ctx.pipe_ctx[i].plane_res.dpp->inst;
+		else if (!context->res_ctx.pipe_ctx[i].plane_res.dpp && dppclk_khz == 0) {
+			/* dpp == NULL && dppclk_khz == 0 is valid because of pipe harvesting.
+			 * In this case just continue in loop
+			 */
+			continue;
+		} else if (!context->res_ctx.pipe_ctx[i].plane_res.dpp && dppclk_khz > 0) {
+			/* The software state is not valid if dpp resource is NULL and
+			 * dppclk_khz > 0.
+			 */
+			ASSERT(false);
+			continue;
+		}
+
+		prev_dppclk_khz = clk_mgr->dccg->pipe_dppclk_khz[i];
+
+		if (safe_to_lower || prev_dppclk_khz < dppclk_khz)
+			clk_mgr->dccg->funcs->update_dpp_dto(
+							clk_mgr->dccg, dpp_inst, dppclk_khz);
+	}
+}
+
+static int dcn401_set_hard_min_by_freq_optimized(struct clk_mgr_internal *clk_mgr, PPCLK_e clk, int requested_clk_khz)
+{
+	if (!clk_mgr->smu_present || !dcn401_is_ppclk_dpm_enabled(clk_mgr, clk))
+		return 0;
+
+	/*
+	 * SMU set hard min interface takes requested clock in mhz and return
+	 * actual clock configured in khz. If we floor requested clk to mhz,
+	 * there is a chance that the actual clock configured in khz is less
+	 * than requested. If we ceil it to mhz, there is a chance that it
+	 * unnecessarily dumps up to a higher dpm level, which burns more power.
+	 * The solution is to set by flooring it to mhz first. If the actual
+	 * clock returned is less than requested, then we will ceil the
+	 * requested value to mhz and call it again.
+	 */
+	int actual_clk_khz = dcn401_smu_set_hard_min_by_freq(clk_mgr, clk, khz_to_mhz_floor(requested_clk_khz));
+
+	if (actual_clk_khz < requested_clk_khz)
+		actual_clk_khz = dcn401_smu_set_hard_min_by_freq(clk_mgr, clk, khz_to_mhz_ceil(requested_clk_khz));
+
+	return actual_clk_khz;
+}
+
+static void dcn401_update_clocks_update_dentist(
+		struct clk_mgr_internal *clk_mgr,
+		struct dc_state *context)
+{
+	uint32_t new_disp_divider = 0;
+	uint32_t new_dispclk_wdivider = 0;
+	uint32_t dentist_dispclk_wdivider_readback = 0;
+	struct dc *dc = clk_mgr->base.ctx->dc;
+
+	if (clk_mgr->base.clks.dispclk_khz == 0)
+		return;
+
+	new_disp_divider = DENTIST_DIVIDER_RANGE_SCALE_FACTOR
+			* clk_mgr->base.dentist_vco_freq_khz / clk_mgr->base.clks.dispclk_khz;
+
+	new_dispclk_wdivider = dentist_get_did_from_divider(new_disp_divider);
+
+	if (dc->debug.override_dispclk_programming) {
+		REG_GET(DENTIST_DISPCLK_CNTL,
+				DENTIST_DISPCLK_WDIVIDER, &dentist_dispclk_wdivider_readback);
+
+		if (dentist_dispclk_wdivider_readback > new_dispclk_wdivider) {
+			REG_UPDATE(DENTIST_DISPCLK_CNTL,
+					DENTIST_DISPCLK_WDIVIDER, new_dispclk_wdivider);
+			REG_WAIT(DENTIST_DISPCLK_CNTL, DENTIST_DISPCLK_CHG_DONE, 1, 50, 2000);
+		}
+	}
+
+}
+
+static void dcn401_execute_block_sequence(struct clk_mgr *clk_mgr_base, unsigned int num_steps)
+{
+	struct clk_mgr_internal *clk_mgr_internal = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+	struct dcn401_clk_mgr *clk_mgr401 = TO_DCN401_CLK_MGR(clk_mgr_internal);
+
+	unsigned int i;
+	union dcn401_clk_mgr_block_sequence_params *params;
+
+	/* execute sequence */
+	for (i = 0; i < num_steps; i++) {
+		params = &clk_mgr401->block_sequence[i].params;
+
+		switch (clk_mgr401->block_sequence[i].func) {
+		case CLK_MGR401_READ_CLOCKS_FROM_DENTIST:
+			dcn2_read_clocks_from_hw_dentist(clk_mgr_base);
+			break;
+		case CLK_MGR401_UPDATE_NUM_DISPLAYS:
+			dcn401_smu_set_num_of_displays(clk_mgr_internal,
+					params->update_num_displays_params.num_displays);
+			break;
+		case CLK_MGR401_UPDATE_HARDMIN_PPCLK:
+			if (params->update_hardmin_params.response)
+				*params->update_hardmin_params.response = dcn401_smu_set_hard_min_by_freq(
+						clk_mgr_internal,
+						params->update_hardmin_params.ppclk,
+						params->update_hardmin_params.freq_mhz);
+			else
+				dcn401_smu_set_hard_min_by_freq(clk_mgr_internal,
+						params->update_hardmin_params.ppclk,
+						params->update_hardmin_params.freq_mhz);
+			break;
+		case CLK_MGR401_UPDATE_HARDMIN_PPCLK_OPTIMIZED:
+			if (params->update_hardmin_optimized_params.response)
+				*params->update_hardmin_optimized_params.response = dcn401_set_hard_min_by_freq_optimized(
+						clk_mgr_internal,
+						params->update_hardmin_optimized_params.ppclk,
+						params->update_hardmin_optimized_params.freq_khz);
+			else
+				dcn401_set_hard_min_by_freq_optimized(clk_mgr_internal,
+						params->update_hardmin_optimized_params.ppclk,
+						params->update_hardmin_optimized_params.freq_khz);
+			break;
+		case CLK_MGR401_UPDATE_ACTIVE_HARDMINS:
+			dcn401_smu_set_active_uclk_fclk_hardmin(
+					clk_mgr_internal,
+					params->update_idle_hardmin_params.uclk_mhz,
+					params->update_idle_hardmin_params.fclk_mhz);
+			break;
+		case CLK_MGR401_UPDATE_IDLE_HARDMINS:
+			dcn401_smu_set_idle_uclk_fclk_hardmin(
+					clk_mgr_internal,
+					params->update_idle_hardmin_params.uclk_mhz,
+					params->update_idle_hardmin_params.fclk_mhz);
+			break;
+		case CLK_MGR401_UPDATE_SUBVP_HARDMINS:
+			dcn401_smu_set_subvp_uclk_fclk_hardmin(
+					clk_mgr_internal,
+					params->update_idle_hardmin_params.uclk_mhz,
+					params->update_idle_hardmin_params.fclk_mhz);
+			break;
+		case CLK_MGR401_UPDATE_DEEP_SLEEP_DCFCLK:
+			dcn401_smu_set_min_deep_sleep_dcef_clk(
+					clk_mgr_internal,
+					params->update_deep_sleep_dcfclk_params.freq_mhz);
+			break;
+		case CLK_MGR401_UPDATE_FCLK_PSTATE_SUPPORT:
+			dcn401_smu_send_fclk_pstate_message(
+					clk_mgr_internal,
+					params->update_pstate_support_params.support);
+			break;
+		case CLK_MGR401_UPDATE_UCLK_PSTATE_SUPPORT:
+			dcn401_smu_send_uclk_pstate_message(
+					clk_mgr_internal,
+					params->update_pstate_support_params.support);
+			break;
+		case CLK_MGR401_UPDATE_CAB_FOR_UCLK:
+			dcn401_smu_send_cab_for_uclk_message(
+				clk_mgr_internal,
+				params->update_cab_for_uclk_params.num_ways);
+			break;
+		case CLK_MGR401_UPDATE_WAIT_FOR_DMUB_ACK:
+			dcn401_smu_wait_for_dmub_ack_mclk(
+					clk_mgr_internal,
+					params->update_wait_for_dmub_ack_params.enable);
+			break;
+		case CLK_MGR401_INDICATE_DRR_STATUS:
+			dcn401_smu_indicate_drr_status(
+					clk_mgr_internal,
+					params->indicate_drr_status_params.mod_drr_for_pstate);
+			break;
+		case CLK_MGR401_UPDATE_DPPCLK_DTO:
+			dcn401_update_clocks_update_dpp_dto(
+					clk_mgr_internal,
+					params->update_dppclk_dto_params.context,
+					params->update_dppclk_dto_params.safe_to_lower,
+					*params->update_dppclk_dto_params.ref_dppclk_khz);
+			break;
+		case CLK_MGR401_UPDATE_DTBCLK_DTO:
+			dcn401_update_clocks_update_dtb_dto(
+					clk_mgr_internal,
+					params->update_dtbclk_dto_params.context,
+					*params->update_dtbclk_dto_params.ref_dtbclk_khz);
+			break;
+		case CLK_MGR401_UPDATE_DENTIST:
+			dcn401_update_clocks_update_dentist(
+					clk_mgr_internal,
+					params->update_dentist_params.context);
+			break;
+		case CLK_MGR401_UPDATE_PSR_WAIT_LOOP:
+			params->update_psr_wait_loop_params.dmcu->funcs->set_psr_wait_loop(
+					params->update_psr_wait_loop_params.dmcu,
+					params->update_psr_wait_loop_params.wait);
+			break;
+		default:
+			/* this should never happen */
+			BREAK_TO_DEBUGGER();
+			break;
+		}
+	}
+}
+
+static unsigned int dcn401_build_update_bandwidth_clocks_sequence(
+		struct clk_mgr *clk_mgr_base,
+		struct dc_state *context,
+		struct dc_clocks *new_clocks,
+		bool safe_to_lower)
+{
+	struct clk_mgr_internal *clk_mgr_internal = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+	struct dcn401_clk_mgr *clk_mgr401 = TO_DCN401_CLK_MGR(clk_mgr_internal);
+	struct dc *dc = clk_mgr_base->ctx->dc;
+	struct dcn401_clk_mgr_block_sequence *block_sequence = clk_mgr401->block_sequence;
+	bool enter_display_off = false;
+	bool update_active_fclk = false;
+	bool update_active_uclk = false;
+	bool update_idle_fclk = false;
+	bool update_idle_uclk = false;
+	bool update_subvp_prefetch_dramclk = false;
+	bool update_subvp_prefetch_fclk = false;
+	bool is_idle_dpm_enabled = dcn401_is_ppclk_dpm_enabled(clk_mgr_internal, PPCLK_UCLK) &&
+			dcn401_is_ppclk_dpm_enabled(clk_mgr_internal, PPCLK_FCLK) &&
+			dcn401_is_ppclk_idle_dpm_enabled(clk_mgr_internal, PPCLK_UCLK) &&
+			dcn401_is_ppclk_idle_dpm_enabled(clk_mgr_internal, PPCLK_FCLK);
+	bool is_df_throttle_opt_enabled = is_idle_dpm_enabled &&
+		dcn401_is_df_throttle_opt_enabled(clk_mgr_internal);
+	int total_plane_count = clk_mgr_helper_get_active_plane_cnt(dc, context);
+	int active_uclk_mhz = khz_to_mhz_ceil(clk_mgr_base->clks.dramclk_khz);
+	int active_fclk_mhz = khz_to_mhz_ceil(clk_mgr_base->clks.fclk_khz);
+	int idle_uclk_mhz = khz_to_mhz_ceil(clk_mgr_base->clks.idle_dramclk_khz);
+	int idle_fclk_mhz = khz_to_mhz_ceil(clk_mgr_base->clks.idle_fclk_khz);
+	int subvp_prefetch_dramclk_mhz = khz_to_mhz_ceil(clk_mgr_base->clks.subvp_prefetch_dramclk_khz);
+	int subvp_prefetch_fclk_mhz = khz_to_mhz_ceil(clk_mgr_base->clks.subvp_prefetch_fclk_khz);
+
+	unsigned int num_steps = 0;
+
+	int display_count;
+	bool fclk_p_state_change_support, uclk_p_state_change_support;
+
+	/* CLK_MGR401_UPDATE_NUM_DISPLAYS */
+	if (clk_mgr_internal->smu_present) {
+		display_count = clk_mgr_helper_get_active_display_cnt(dc, context);
+
+		if (display_count == 0)
+			enter_display_off = true;
+
+		if (enter_display_off == safe_to_lower) {
+			block_sequence[num_steps].params.update_num_displays_params.num_displays = display_count;
+			block_sequence[num_steps].func = CLK_MGR401_UPDATE_NUM_DISPLAYS;
+			num_steps++;
+		}
+	}
+
+	/* CLK_MGR401_UPDATE_FCLK_PSTATE_SUPPORT */
+	clk_mgr_base->clks.fclk_prev_p_state_change_support = clk_mgr_base->clks.fclk_p_state_change_support;
+	fclk_p_state_change_support = new_clocks->fclk_p_state_change_support || (total_plane_count == 0);
+	if (should_update_pstate_support(safe_to_lower, fclk_p_state_change_support, clk_mgr_base->clks.fclk_prev_p_state_change_support)) {
+		clk_mgr_base->clks.fclk_p_state_change_support = fclk_p_state_change_support;
+		update_active_fclk = true;
+		update_idle_fclk = true;
+
+		/* To enable FCLK P-state switching, send PSTATE_SUPPORTED message to PMFW (message not supported on DCN401)*/
+		// if (clk_mgr_base->clks.fclk_p_state_change_support) {
+		// 	/* Handle the code for sending a message to PMFW that FCLK P-state change is supported */
+		// 	if (dcn401_is_ppclk_dpm_enabled(clk_mgr_internal, PPCLK_FCLK)) {
+		// 		block_sequence[num_steps].params.update_pstate_support_params.support = true;
+		// 		block_sequence[num_steps].func = CLK_MGR401_UPDATE_FCLK_PSTATE_SUPPORT;
+		// 		num_steps++;
+		// 	}
+		// }
+	}
+
+	if (!clk_mgr_base->clks.fclk_p_state_change_support && dcn401_is_ppclk_dpm_enabled(clk_mgr_internal, PPCLK_FCLK)) {
+		/* when P-State switching disabled, set UCLK min = max */
+		idle_fclk_mhz =
+				clk_mgr_base->bw_params->clk_table.entries[clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_fclk_levels - 1].fclk_mhz;
+		active_fclk_mhz = idle_fclk_mhz;
+	}
+
+	/* UPDATE DCFCLK */
+	if (dc->debug.force_min_dcfclk_mhz > 0)
+		new_clocks->dcfclk_khz = (new_clocks->dcfclk_khz > (dc->debug.force_min_dcfclk_mhz * 1000)) ?
+				new_clocks->dcfclk_khz : (dc->debug.force_min_dcfclk_mhz * 1000);
+
+	if (should_set_clock(safe_to_lower, new_clocks->dcfclk_khz, clk_mgr_base->clks.dcfclk_khz)) {
+		clk_mgr_base->clks.dcfclk_khz = new_clocks->dcfclk_khz;
+		if (dcn401_is_ppclk_dpm_enabled(clk_mgr_internal, PPCLK_DCFCLK)) {
+			block_sequence[num_steps].params.update_hardmin_params.ppclk = PPCLK_DCFCLK;
+			block_sequence[num_steps].params.update_hardmin_params.freq_mhz = khz_to_mhz_ceil(clk_mgr_base->clks.dcfclk_khz);
+			block_sequence[num_steps].params.update_hardmin_params.response = NULL;
+			block_sequence[num_steps].func = CLK_MGR401_UPDATE_HARDMIN_PPCLK;
+			num_steps++;
+		}
+	}
+
+	/* CLK_MGR401_UPDATE_DEEP_SLEEP_DCFCLK */
+	if (should_set_clock(safe_to_lower, new_clocks->dcfclk_deep_sleep_khz, clk_mgr_base->clks.dcfclk_deep_sleep_khz)) {
+		clk_mgr_base->clks.dcfclk_deep_sleep_khz = new_clocks->dcfclk_deep_sleep_khz;
+		if (dcn401_is_ppclk_dpm_enabled(clk_mgr_internal, PPCLK_DCFCLK)) {
+			block_sequence[num_steps].params.update_deep_sleep_dcfclk_params.freq_mhz = khz_to_mhz_ceil(clk_mgr_base->clks.dcfclk_deep_sleep_khz);
+			block_sequence[num_steps].func = CLK_MGR401_UPDATE_DEEP_SLEEP_DCFCLK;
+			num_steps++;
+		}
+	}
+
+	/* SOCCLK */
+	if (should_set_clock(safe_to_lower, new_clocks->socclk_khz, clk_mgr_base->clks.socclk_khz))
+		/* We don't actually care about socclk, don't notify SMU of hard min */
+		clk_mgr_base->clks.socclk_khz = new_clocks->socclk_khz;
+
+	/* UCLK */
+	if (new_clocks->fw_based_mclk_switching != clk_mgr_base->clks.fw_based_mclk_switching &&
+			new_clocks->fw_based_mclk_switching) {
+		/* enable FAMS features */
+		clk_mgr_base->clks.fw_based_mclk_switching = new_clocks->fw_based_mclk_switching;
+
+		block_sequence[num_steps].params.update_wait_for_dmub_ack_params.enable = clk_mgr_base->clks.fw_based_mclk_switching;
+		block_sequence[num_steps].func = CLK_MGR401_UPDATE_WAIT_FOR_DMUB_ACK;
+		num_steps++;
+
+		block_sequence[num_steps].params.indicate_drr_status_params.mod_drr_for_pstate = clk_mgr_base->clks.fw_based_mclk_switching;
+		block_sequence[num_steps].func = CLK_MGR401_INDICATE_DRR_STATUS;
+		num_steps++;
+	}
+
+	/* CLK_MGR401_UPDATE_CAB_FOR_UCLK */
+	clk_mgr_base->clks.prev_num_ways = clk_mgr_base->clks.num_ways;
+	if (clk_mgr_base->clks.num_ways != new_clocks->num_ways &&
+			clk_mgr_base->clks.num_ways < new_clocks->num_ways) {
+		/* increase num ways for subvp */
+		clk_mgr_base->clks.num_ways = new_clocks->num_ways;
+		if (dcn401_is_ppclk_dpm_enabled(clk_mgr_internal, PPCLK_UCLK)) {
+			block_sequence[num_steps].params.update_cab_for_uclk_params.num_ways = clk_mgr_base->clks.num_ways;
+			block_sequence[num_steps].func = CLK_MGR401_UPDATE_CAB_FOR_UCLK;
+			num_steps++;
+		}
+	}
+
+	clk_mgr_base->clks.prev_p_state_change_support = clk_mgr_base->clks.p_state_change_support;
+	uclk_p_state_change_support = new_clocks->p_state_change_support || (total_plane_count == 0);
+	if (should_update_pstate_support(safe_to_lower, uclk_p_state_change_support, clk_mgr_base->clks.prev_p_state_change_support)) {
+		clk_mgr_base->clks.p_state_change_support = uclk_p_state_change_support;
+		update_active_uclk = true;
+		update_idle_uclk = true;
+
+		if (clk_mgr_base->clks.p_state_change_support) {
+			/* enable UCLK switching  */
+			if (dcn401_is_ppclk_dpm_enabled(clk_mgr_internal, PPCLK_UCLK)) {
+				block_sequence[num_steps].params.update_pstate_support_params.support = true;
+				block_sequence[num_steps].func = CLK_MGR401_UPDATE_UCLK_PSTATE_SUPPORT;
+				num_steps++;
+			}
+		}
+	}
+
+	if (!clk_mgr_base->clks.p_state_change_support && dcn401_is_ppclk_dpm_enabled(clk_mgr_internal, PPCLK_UCLK)) {
+		/* when P-State switching disabled, set UCLK min = max */
+		if (dc->clk_mgr->dc_mode_softmax_enabled) {
+			/* will never have the functional UCLK min above the softmax
+			* since we calculate mode support based on softmax being the max UCLK
+			* frequency.
+			*/
+			active_uclk_mhz = clk_mgr_base->bw_params->dc_mode_softmax_memclk;
+		} else {
+			active_uclk_mhz = clk_mgr_base->bw_params->max_memclk_mhz;
+		}
+		idle_uclk_mhz = active_uclk_mhz;
+	}
+
+	/* Always update saved value, even if new value not set due to P-State switching unsupported */
+	if (should_set_clock(safe_to_lower, new_clocks->dramclk_khz, clk_mgr_base->clks.dramclk_khz)) {
+		clk_mgr_base->clks.dramclk_khz = new_clocks->dramclk_khz;
+
+		if (clk_mgr_base->clks.p_state_change_support) {
+			update_active_uclk = true;
+			active_uclk_mhz = khz_to_mhz_ceil(clk_mgr_base->clks.dramclk_khz);
+		}
+	}
+
+	if (should_set_clock(safe_to_lower, new_clocks->idle_dramclk_khz, clk_mgr_base->clks.idle_dramclk_khz)) {
+		clk_mgr_base->clks.idle_dramclk_khz = new_clocks->idle_dramclk_khz;
+
+		if (clk_mgr_base->clks.p_state_change_support) {
+			update_idle_uclk = true;
+			idle_uclk_mhz = khz_to_mhz_ceil(clk_mgr_base->clks.idle_dramclk_khz);
+		}
+	}
+
+	if (should_set_clock(safe_to_lower, new_clocks->subvp_prefetch_dramclk_khz, clk_mgr_base->clks.subvp_prefetch_dramclk_khz)) {
+		clk_mgr_base->clks.subvp_prefetch_dramclk_khz = new_clocks->subvp_prefetch_dramclk_khz;
+		update_subvp_prefetch_dramclk = true;
+		subvp_prefetch_dramclk_mhz = khz_to_mhz_ceil(clk_mgr_base->clks.subvp_prefetch_dramclk_khz);
+	}
+
+	/* FCLK */
+	/* Always update saved value, even if new value not set due to P-State switching unsupported */
+	if (should_set_clock(safe_to_lower, new_clocks->fclk_khz, clk_mgr_base->clks.fclk_khz)) {
+		clk_mgr_base->clks.fclk_khz = new_clocks->fclk_khz;
+
+		if (clk_mgr_base->clks.fclk_p_state_change_support) {
+			update_active_fclk = true;
+			active_fclk_mhz = khz_to_mhz_ceil(clk_mgr_base->clks.fclk_khz);
+		}
+	}
+
+	if (should_set_clock(safe_to_lower, new_clocks->idle_fclk_khz, clk_mgr_base->clks.idle_fclk_khz)) {
+		clk_mgr_base->clks.idle_fclk_khz = new_clocks->idle_fclk_khz;
+
+		if (clk_mgr_base->clks.fclk_p_state_change_support) {
+			update_idle_fclk = true;
+			idle_fclk_mhz = khz_to_mhz_ceil(clk_mgr_base->clks.idle_fclk_khz);
+		}
+	}
+
+	if (should_set_clock(safe_to_lower, new_clocks->subvp_prefetch_fclk_khz, clk_mgr_base->clks.subvp_prefetch_fclk_khz)) {
+		clk_mgr_base->clks.subvp_prefetch_fclk_khz = new_clocks->subvp_prefetch_fclk_khz;
+		update_subvp_prefetch_fclk = true;
+		subvp_prefetch_fclk_mhz = khz_to_mhz_ceil(clk_mgr_base->clks.subvp_prefetch_fclk_khz);
+	}
+
+	/* When idle DPM is enabled, need to send active and idle hardmins separately */
+	/* CLK_MGR401_UPDATE_ACTIVE_HARDMINS */
+	if ((update_active_uclk || update_active_fclk) && is_idle_dpm_enabled) {
+		block_sequence[num_steps].params.update_idle_hardmin_params.uclk_mhz = active_uclk_mhz;
+		block_sequence[num_steps].params.update_idle_hardmin_params.fclk_mhz = active_fclk_mhz;
+		block_sequence[num_steps].func = CLK_MGR401_UPDATE_ACTIVE_HARDMINS;
+		num_steps++;
+	}
+
+	/* CLK_MGR401_UPDATE_IDLE_HARDMINS */
+	if ((update_idle_uclk || update_idle_fclk) && is_idle_dpm_enabled) {
+		block_sequence[num_steps].params.update_idle_hardmin_params.uclk_mhz = idle_uclk_mhz;
+		block_sequence[num_steps].params.update_idle_hardmin_params.fclk_mhz = idle_fclk_mhz;
+		block_sequence[num_steps].func = CLK_MGR401_UPDATE_IDLE_HARDMINS;
+		num_steps++;
+	}
+
+	/* CLK_MGR401_UPDATE_SUBVP_HARDMINS */
+	if ((update_subvp_prefetch_dramclk || update_subvp_prefetch_fclk) && is_df_throttle_opt_enabled) {
+		block_sequence[num_steps].params.update_idle_hardmin_params.uclk_mhz = subvp_prefetch_dramclk_mhz;
+		block_sequence[num_steps].params.update_idle_hardmin_params.fclk_mhz = subvp_prefetch_fclk_mhz;
+		block_sequence[num_steps].func = CLK_MGR401_UPDATE_SUBVP_HARDMINS;
+		num_steps++;
+	}
+
+	/* set UCLK to requested value if P-State switching is supported, or to re-enable P-State switching */
+	if (update_active_uclk || update_idle_uclk) {
+		if (!is_idle_dpm_enabled) {
+			block_sequence[num_steps].params.update_hardmin_params.ppclk = PPCLK_UCLK;
+			block_sequence[num_steps].params.update_hardmin_params.freq_mhz = active_uclk_mhz;
+			block_sequence[num_steps].params.update_hardmin_params.response = NULL;
+			block_sequence[num_steps].func = CLK_MGR401_UPDATE_HARDMIN_PPCLK;
+			num_steps++;
+		}
+
+		/* disable UCLK P-State support if needed */
+		if (!uclk_p_state_change_support &&
+				should_update_pstate_support(safe_to_lower, uclk_p_state_change_support, clk_mgr_base->clks.prev_p_state_change_support) &&
+				dcn401_is_ppclk_dpm_enabled(clk_mgr_internal, PPCLK_UCLK)) {
+			block_sequence[num_steps].params.update_pstate_support_params.support = false;
+			block_sequence[num_steps].func = CLK_MGR401_UPDATE_UCLK_PSTATE_SUPPORT;
+			num_steps++;
+		}
+	}
+
+	/* set FCLK to requested value if P-State switching is supported, or to re-enable P-State switching */
+	if (update_active_fclk || update_idle_fclk) {
+		/* No need to send active FCLK hardmin, automatically set based on DCFCLK */
+		// if (!is_idle_dpm_enabled) {
+		// 	block_sequence[*num_steps].update_hardmin_params.clk_mgr = clk_mgr;
+		// 	block_sequence[*num_steps].update_hardmin_params.ppclk = PPCLK_FCLK;
+		// 	block_sequence[*num_steps].update_hardmin_params.freq_mhz = active_fclk_mhz;
+		// 	block_sequence[*num_steps].update_hardmin_params.response = NULL;
+		// 	block_sequence[*num_steps].func = CLK_MGR401_UPDATE_HARDMIN_PPCLK;
+		// 	(*num_steps)++;
+		// }
+
+		/* disable FCLK P-State support if needed (message not supported on DCN401)*/
+		// if (!fclk_p_state_change_support &&
+		// 		should_update_pstate_support(safe_to_lower, fclk_p_state_change_support, clk_mgr_base->clks.fclk_prev_p_state_change_support) &&
+		// 		dcn401_is_ppclk_dpm_enabled(clk_mgr_internal, PPCLK_FCLK)) {
+		// 	block_sequence[num_steps].params.update_pstate_support_params.support = false;
+		// 	block_sequence[num_steps].func = CLK_MGR401_UPDATE_FCLK_PSTATE_SUPPORT;
+		// 	num_steps++;
+		// }
+	}
+
+	if (new_clocks->fw_based_mclk_switching != clk_mgr_base->clks.fw_based_mclk_switching &&
+			safe_to_lower && !new_clocks->fw_based_mclk_switching) {
+		/* disable FAMS features */
+		clk_mgr_base->clks.fw_based_mclk_switching = new_clocks->fw_based_mclk_switching;
+
+		block_sequence[num_steps].params.update_wait_for_dmub_ack_params.enable = clk_mgr_base->clks.fw_based_mclk_switching;
+		block_sequence[num_steps].func = CLK_MGR401_UPDATE_WAIT_FOR_DMUB_ACK;
+		num_steps++;
+
+		block_sequence[num_steps].params.indicate_drr_status_params.mod_drr_for_pstate = clk_mgr_base->clks.fw_based_mclk_switching;
+		block_sequence[num_steps].func = CLK_MGR401_INDICATE_DRR_STATUS;
+		num_steps++;
+	}
+
+	/* CLK_MGR401_UPDATE_CAB_FOR_UCLK */
+	if (clk_mgr_base->clks.num_ways != new_clocks->num_ways &&
+			safe_to_lower && clk_mgr_base->clks.num_ways > new_clocks->num_ways) {
+		/* decrease num ways for subvp */
+		clk_mgr_base->clks.num_ways = new_clocks->num_ways;
+		if (dcn401_is_ppclk_dpm_enabled(clk_mgr_internal, PPCLK_UCLK)) {
+			block_sequence[num_steps].params.update_cab_for_uclk_params.num_ways = clk_mgr_base->clks.num_ways;
+			block_sequence[num_steps].func = CLK_MGR401_UPDATE_CAB_FOR_UCLK;
+			num_steps++;
+		}
+	}
+
+	return num_steps;
+}
+
+static unsigned int dcn401_build_update_display_clocks_sequence(
+		struct clk_mgr *clk_mgr_base,
+		struct dc_state *context,
+		struct dc_clocks *new_clocks,
+		bool safe_to_lower)
+{
+	struct clk_mgr_internal *clk_mgr_internal = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+	struct dcn401_clk_mgr *clk_mgr401 = TO_DCN401_CLK_MGR(clk_mgr_internal);
+	struct dc *dc = clk_mgr_base->ctx->dc;
+	struct dmcu *dmcu = clk_mgr_base->ctx->dc->res_pool->dmcu;
+	struct dcn401_clk_mgr_block_sequence *block_sequence = clk_mgr401->block_sequence;
+	bool force_reset = false;
+	bool update_dispclk = false;
+	bool update_dppclk = false;
+	bool dppclk_lowered = false;
+
+	unsigned int num_steps = 0;
+
+	/* CLK_MGR401_READ_CLOCKS_FROM_DENTIST */
+	if (clk_mgr_base->clks.dispclk_khz == 0 ||
+			(dc->debug.force_clock_mode & 0x1)) {
+		/* This is from resume or boot up, if forced_clock cfg option used,
+		 * we bypass program dispclk and DPPCLK, but need set them for S3.
+		 * Force_clock_mode 0x1:  force reset the clock even it is the same clock
+		 * as long as it is in Passive level.
+		 */
+		force_reset = true;
+
+		clk_mgr_base->clks.dispclk_khz = clk_mgr_base->boot_snapshot.dispclk;
+		clk_mgr_base->clks.actual_dispclk_khz = clk_mgr_base->clks.dispclk_khz;
+
+		clk_mgr_base->clks.dppclk_khz = clk_mgr_base->boot_snapshot.dppclk;
+		clk_mgr_base->clks.actual_dppclk_khz = clk_mgr_base->clks.dppclk_khz;
+	}
+
+	/* DTBCLK */
+	if (!new_clocks->dtbclk_en && dcn401_is_ppclk_dpm_enabled(clk_mgr_internal, PPCLK_DTBCLK)) {
+		new_clocks->ref_dtbclk_khz = clk_mgr_base->bw_params->clk_table.entries[0].dtbclk_mhz * 1000;
+	}
+
+	/* clock limits are received with MHz precision, divide by 1000 to prevent setting clocks at every call */
+	if (!dc->debug.disable_dtb_ref_clk_switch &&
+			should_set_clock(safe_to_lower, new_clocks->ref_dtbclk_khz / 1000, clk_mgr_base->clks.ref_dtbclk_khz / 1000) && //TODO these should be ceiled
+			dcn401_is_ppclk_dpm_enabled(clk_mgr_internal, PPCLK_DTBCLK)) {
+		/* DCCG requires KHz precision for DTBCLK */
+		block_sequence[num_steps].params.update_hardmin_params.ppclk = PPCLK_DTBCLK;
+		block_sequence[num_steps].params.update_hardmin_params.freq_mhz = khz_to_mhz_ceil(new_clocks->ref_dtbclk_khz);
+		block_sequence[num_steps].params.update_hardmin_params.response = &clk_mgr_base->clks.ref_dtbclk_khz;
+		block_sequence[num_steps].func = CLK_MGR401_UPDATE_HARDMIN_PPCLK;
+		num_steps++;
+
+		/* Update DTO in DCCG */
+		block_sequence[num_steps].params.update_dtbclk_dto_params.context = context;
+		block_sequence[num_steps].params.update_dtbclk_dto_params.ref_dtbclk_khz = &clk_mgr_base->clks.ref_dtbclk_khz;
+		block_sequence[num_steps].func = CLK_MGR401_UPDATE_DTBCLK_DTO;
+		num_steps++;
+	}
+
+	if (should_set_clock(safe_to_lower, new_clocks->dppclk_khz, clk_mgr_base->clks.dppclk_khz)) {
+		if (clk_mgr_base->clks.dppclk_khz > new_clocks->dppclk_khz)
+			dppclk_lowered = true;
+
+		clk_mgr_base->clks.dppclk_khz = new_clocks->dppclk_khz;
+		clk_mgr_base->clks.actual_dppclk_khz = new_clocks->dppclk_khz;
+
+		update_dppclk = true;
+	}
+
+	if (should_set_clock(safe_to_lower, new_clocks->dispclk_khz, clk_mgr_base->clks.dispclk_khz)) {
+		clk_mgr_base->clks.dispclk_khz = new_clocks->dispclk_khz;
+
+		block_sequence[num_steps].params.update_hardmin_optimized_params.ppclk = PPCLK_DISPCLK;
+		block_sequence[num_steps].params.update_hardmin_optimized_params.freq_khz = clk_mgr_base->clks.dispclk_khz;
+		block_sequence[num_steps].params.update_hardmin_optimized_params.response = &clk_mgr_base->clks.actual_dispclk_khz;
+		block_sequence[num_steps].func = CLK_MGR401_UPDATE_HARDMIN_PPCLK_OPTIMIZED;
+		num_steps++;
+
+		update_dispclk = true;
+	}
+
+	if (dc->config.forced_clocks == false || (force_reset && safe_to_lower)) {
+		if (dppclk_lowered) {
+			/* if clock is being lowered, increase DTO before lowering refclk */
+			block_sequence[num_steps].params.update_dppclk_dto_params.context = context;
+			block_sequence[num_steps].params.update_dppclk_dto_params.ref_dppclk_khz = &clk_mgr_base->clks.dppclk_khz;
+			block_sequence[num_steps].params.update_dppclk_dto_params.safe_to_lower = safe_to_lower;
+			block_sequence[num_steps].func = CLK_MGR401_UPDATE_DPPCLK_DTO;
+			num_steps++;
+
+			block_sequence[num_steps].params.update_dentist_params.context = context;
+			block_sequence[num_steps].func = CLK_MGR401_UPDATE_DENTIST;
+			num_steps++;
+
+			if (dcn401_is_ppclk_dpm_enabled(clk_mgr_internal, PPCLK_DPPCLK)) {
+				block_sequence[num_steps].params.update_hardmin_optimized_params.ppclk = PPCLK_DPPCLK;
+				block_sequence[num_steps].params.update_hardmin_optimized_params.freq_khz = clk_mgr_base->clks.dppclk_khz;
+				block_sequence[num_steps].params.update_hardmin_optimized_params.response = &clk_mgr_base->clks.actual_dppclk_khz;
+				block_sequence[num_steps].func = CLK_MGR401_UPDATE_HARDMIN_PPCLK_OPTIMIZED;
+				num_steps++;
+
+				block_sequence[num_steps].params.update_dppclk_dto_params.context = context;
+				block_sequence[num_steps].params.update_dppclk_dto_params.ref_dppclk_khz = &clk_mgr_base->clks.actual_dppclk_khz;
+				block_sequence[num_steps].params.update_dppclk_dto_params.safe_to_lower = safe_to_lower;
+				block_sequence[num_steps].func = CLK_MGR401_UPDATE_DPPCLK_DTO;
+				num_steps++;
+			}
+		} else {
+			/* if clock is being raised, increase refclk before lowering DTO */
+			if (update_dppclk && dcn401_is_ppclk_dpm_enabled(clk_mgr_internal, PPCLK_DPPCLK)) {
+				block_sequence[num_steps].params.update_hardmin_optimized_params.ppclk = PPCLK_DPPCLK;
+				block_sequence[num_steps].params.update_hardmin_optimized_params.freq_khz = clk_mgr_base->clks.dppclk_khz;
+				block_sequence[num_steps].params.update_hardmin_optimized_params.response = &clk_mgr_base->clks.actual_dppclk_khz;
+				block_sequence[num_steps].func = CLK_MGR401_UPDATE_HARDMIN_PPCLK_OPTIMIZED;
+				num_steps++;
+			}
+
+			if (update_dppclk || update_dispclk) {
+				block_sequence[num_steps].params.update_dentist_params.context = context;
+				block_sequence[num_steps].func = CLK_MGR401_UPDATE_DENTIST;
+				num_steps++;
+			}
+
+			block_sequence[num_steps].params.update_dppclk_dto_params.context = context;
+			block_sequence[num_steps].params.update_dppclk_dto_params.ref_dppclk_khz = &clk_mgr_base->clks.actual_dppclk_khz;
+			block_sequence[num_steps].params.update_dppclk_dto_params.safe_to_lower = safe_to_lower;
+			block_sequence[num_steps].func = CLK_MGR401_UPDATE_DPPCLK_DTO;
+			num_steps++;
+		}
+	}
+
+	if (update_dispclk && dmcu && dmcu->funcs->is_dmcu_initialized(dmcu)) {
+		/*update dmcu for wait_loop count*/
+		block_sequence[num_steps].params.update_psr_wait_loop_params.dmcu = dmcu;
+		block_sequence[num_steps].params.update_psr_wait_loop_params.wait = clk_mgr_base->clks.dispclk_khz / 1000 / 7;
+		block_sequence[num_steps].func = CLK_MGR401_UPDATE_PSR_WAIT_LOOP;
+		num_steps++;
+	}
+
+	return num_steps;
+}
+
+static void dcn401_update_clocks(struct clk_mgr *clk_mgr_base,
+		struct dc_state *context,
+		bool safe_to_lower)
+{
+	struct dc *dc = clk_mgr_base->ctx->dc;
+
+	unsigned int num_steps = 0;
+
+	/* build bandwidth related clocks update sequence */
+	num_steps = dcn401_build_update_bandwidth_clocks_sequence(clk_mgr_base,
+			context,
+			&context->bw_ctx.bw.dcn.clk,
+			safe_to_lower);
+
+	/* execute sequence */
+	dcn401_execute_block_sequence(clk_mgr_base,	num_steps);
+
+	/* build display related clocks update sequence */
+	num_steps = dcn401_build_update_display_clocks_sequence(clk_mgr_base,
+			context,
+			&context->bw_ctx.bw.dcn.clk,
+			safe_to_lower);
+
+	/* execute sequence */
+	dcn401_execute_block_sequence(clk_mgr_base,	num_steps);
+
+	if (dc->config.enable_auto_dpm_test_logs)
+		dcn401_auto_dpm_test_log(&context->bw_ctx.bw.dcn.clk, TO_CLK_MGR_INTERNAL(clk_mgr_base), context);
+
+}
+
+
+static uint32_t dcn401_get_vco_frequency_from_reg(struct clk_mgr_internal *clk_mgr)
+{
+		struct fixed31_32 pll_req;
+		uint32_t pll_req_reg = 0;
+
+		/* get FbMult value */
+		pll_req_reg = REG_READ(CLK0_CLK_PLL_REQ);
+
+		/* set up a fixed-point number
+		 * this works because the int part is on the right edge of the register
+		 * and the frac part is on the left edge
+		 */
+		pll_req = dc_fixpt_from_int(pll_req_reg & clk_mgr->clk_mgr_mask->FbMult_int);
+		pll_req.value |= pll_req_reg & clk_mgr->clk_mgr_mask->FbMult_frac;
+
+		/* multiply by REFCLK period */
+		pll_req = dc_fixpt_mul_int(pll_req, clk_mgr->dfs_ref_freq_khz);
+
+		return dc_fixpt_floor(pll_req);
+}
+
+static void dcn401_clock_read_ss_info(struct clk_mgr_internal *clk_mgr)
+{
+	struct dc_bios *bp = clk_mgr->base.ctx->dc_bios;
+	int ss_info_num = bp->funcs->get_ss_entry_number(
+			bp, AS_SIGNAL_TYPE_GPU_PLL);
+
+	if (ss_info_num) {
+		struct spread_spectrum_info info = { { 0 } };
+		enum bp_result result = bp->funcs->get_spread_spectrum_info(
+				bp, AS_SIGNAL_TYPE_GPU_PLL, 0, &info);
+
+		/* SSInfo.spreadSpectrumPercentage !=0 would be sign
+		 * that SS is enabled
+		 */
+		if (result == BP_RESULT_OK &&
+				info.spread_spectrum_percentage != 0) {
+			clk_mgr->ss_on_dprefclk = true;
+			clk_mgr->dprefclk_ss_divider = info.spread_percentage_divider;
+
+			if (info.type.CENTER_MODE == 0) {
+				/* Currently for DP Reference clock we
+				 * need only SS percentage for
+				 * downspread
+				 */
+				clk_mgr->dprefclk_ss_percentage =
+						info.spread_spectrum_percentage;
+			}
+		}
+	}
+}
+static void dcn401_notify_wm_ranges(struct clk_mgr *clk_mgr_base)
+{
+	unsigned int i;
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+	WatermarksExternal_t *table = (WatermarksExternal_t *) clk_mgr->wm_range_table;
+
+	if (!clk_mgr->smu_present)
+		return;
+
+	if (!table)
+		return;
+
+	memset(table, 0, sizeof(*table));
+
+	/* collect valid ranges, place in pmfw table */
+	for (i = 0; i < WM_SET_COUNT; i++)
+		if (clk_mgr->base.bw_params->wm_table.nv_entries[i].valid) {
+			table->Watermarks.WatermarkRow[i].WmSetting = i;
+			table->Watermarks.WatermarkRow[i].Flags = clk_mgr->base.bw_params->wm_table.nv_entries[i].pmfw_breakdown.wm_type;
+		}
+	dcn401_smu_set_dram_addr_high(clk_mgr, clk_mgr->wm_range_table_addr >> 32);
+	dcn401_smu_set_dram_addr_low(clk_mgr, clk_mgr->wm_range_table_addr & 0xFFFFFFFF);
+	dcn401_smu_transfer_wm_table_dram_2_smu(clk_mgr);
+}
+
+/* Set min memclk to minimum, either constrained by the current mode or DPM0 */
+static void dcn401_set_hard_min_memclk(struct clk_mgr *clk_mgr_base, bool current_mode)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+	const struct dc *dc = clk_mgr->base.ctx->dc;
+	struct dc_state *context = dc->current_state;
+	struct dc_clocks new_clocks;
+	int num_steps;
+
+	if (!clk_mgr->smu_present || !dcn401_is_ppclk_dpm_enabled(clk_mgr, PPCLK_UCLK))
+		return;
+
+	/* build clock update */
+	memcpy(&new_clocks, &clk_mgr_base->clks, sizeof(struct dc_clocks));
+
+	if (current_mode) {
+		new_clocks.dramclk_khz = context->bw_ctx.bw.dcn.clk.dramclk_khz;
+		new_clocks.idle_dramclk_khz = context->bw_ctx.bw.dcn.clk.idle_dramclk_khz;
+		new_clocks.p_state_change_support = context->bw_ctx.bw.dcn.clk.p_state_change_support;
+	} else {
+		new_clocks.dramclk_khz = clk_mgr_base->bw_params->clk_table.entries[0].memclk_mhz * 1000;
+		new_clocks.idle_dramclk_khz = new_clocks.dramclk_khz;
+		new_clocks.p_state_change_support = true;
+	}
+
+	num_steps = dcn401_build_update_bandwidth_clocks_sequence(clk_mgr_base,
+			context,
+			&new_clocks,
+			true);
+
+	/* execute sequence */
+	dcn401_execute_block_sequence(clk_mgr_base,	num_steps);
+}
+
+static int dcn401_get_hard_min_memclk(struct clk_mgr *clk_mgr_base)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+
+	return clk_mgr->base.ctx->dc->current_state->bw_ctx.bw.dcn.clk.dramclk_khz;
+}
+
+static int dcn401_get_hard_min_fclk(struct clk_mgr *clk_mgr_base)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+
+	return clk_mgr->base.ctx->dc->current_state->bw_ctx.bw.dcn.clk.fclk_khz;
+}
+
+/* Get current memclk states, update bounding box */
+static void dcn401_get_memclk_states_from_smu(struct clk_mgr *clk_mgr_base)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+	struct clk_limit_num_entries *num_entries_per_clk = &clk_mgr_base->bw_params->clk_table.num_entries_per_clk;
+	unsigned int num_levels;
+
+	if (!clk_mgr->smu_present)
+		return;
+
+	/* Refresh memclk and fclk states */
+	dcn401_init_single_clock(clk_mgr, PPCLK_UCLK,
+			&clk_mgr_base->bw_params->clk_table.entries[0].memclk_mhz,
+			&num_entries_per_clk->num_memclk_levels);
+	if (num_entries_per_clk->num_memclk_levels) {
+		clk_mgr_base->bw_params->max_memclk_mhz =
+				clk_mgr_base->bw_params->clk_table.entries[num_entries_per_clk->num_memclk_levels - 1].memclk_mhz;
+	}
+
+	clk_mgr_base->bw_params->dc_mode_limit.memclk_mhz = dcn401_smu_get_dc_mode_max_dpm_freq(clk_mgr, PPCLK_UCLK);
+	if (num_entries_per_clk->num_memclk_levels && clk_mgr_base->bw_params->dc_mode_limit.memclk_mhz ==
+			clk_mgr_base->bw_params->clk_table.entries[num_entries_per_clk->num_memclk_levels - 1].memclk_mhz)
+		clk_mgr_base->bw_params->dc_mode_limit.memclk_mhz = 0;
+	clk_mgr_base->bw_params->dc_mode_softmax_memclk = clk_mgr_base->bw_params->dc_mode_limit.memclk_mhz;
+
+	dcn401_init_single_clock(clk_mgr, PPCLK_FCLK,
+			&clk_mgr_base->bw_params->clk_table.entries[0].fclk_mhz,
+			&num_entries_per_clk->num_fclk_levels);
+	clk_mgr_base->bw_params->dc_mode_limit.fclk_mhz = dcn401_smu_get_dc_mode_max_dpm_freq(clk_mgr, PPCLK_FCLK);
+	if (num_entries_per_clk->num_fclk_levels && clk_mgr_base->bw_params->dc_mode_limit.fclk_mhz ==
+			clk_mgr_base->bw_params->clk_table.entries[num_entries_per_clk->num_fclk_levels - 1].fclk_mhz)
+		clk_mgr_base->bw_params->dc_mode_limit.fclk_mhz = 0;
+
+	num_levels = max(num_entries_per_clk->num_memclk_levels, num_entries_per_clk->num_fclk_levels);
+
+	clk_mgr_base->bw_params->clk_table.num_entries = num_levels ? num_levels : 1;
+
+	if (clk_mgr->dpm_present && !num_levels)
+		clk_mgr->dpm_present = false;
+
+	clk_mgr_base->bw_params->num_channels = dcn401_smu_get_num_of_umc_channels(clk_mgr);
+	if (clk_mgr_base->ctx->dc_bios) {
+		/* use BIOS values if none provided by PMFW */
+		if (clk_mgr_base->bw_params->num_channels == 0) {
+			clk_mgr_base->bw_params->num_channels = clk_mgr_base->ctx->dc_bios->vram_info.num_chans;
+		}
+		clk_mgr_base->bw_params->dram_channel_width_bytes = clk_mgr_base->ctx->dc_bios->vram_info.dram_channel_width_bytes;
+	}
+
+	/* Refresh bounding box */
+	clk_mgr_base->ctx->dc->res_pool->funcs->update_bw_bounding_box(
+			clk_mgr->base.ctx->dc, clk_mgr_base->bw_params);
+}
+
+static bool dcn401_are_clock_states_equal(struct dc_clocks *a,
+					struct dc_clocks *b)
+{
+	if (a->dispclk_khz != b->dispclk_khz)
+		return false;
+	else if (a->dppclk_khz != b->dppclk_khz)
+		return false;
+	else if (a->dcfclk_khz != b->dcfclk_khz)
+		return false;
+	else if (a->dcfclk_deep_sleep_khz != b->dcfclk_deep_sleep_khz)
+		return false;
+	else if (a->dramclk_khz != b->dramclk_khz)
+		return false;
+	else if (a->p_state_change_support != b->p_state_change_support)
+		return false;
+	else if (a->fclk_p_state_change_support != b->fclk_p_state_change_support)
+		return false;
+
+	return true;
+}
+
+static void dcn401_enable_pme_wa(struct clk_mgr *clk_mgr_base)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+
+	if (!clk_mgr->smu_present)
+		return;
+
+	dcn401_smu_set_pme_workaround(clk_mgr);
+}
+
+static bool dcn401_is_smu_present(struct clk_mgr *clk_mgr_base)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+	return clk_mgr->smu_present;
+}
+
+
+static int dcn401_get_dtb_ref_freq_khz(struct clk_mgr *clk_mgr_base)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+
+	int dtb_ref_clk_khz = 0;
+
+	if (clk_mgr->smu_present && dcn401_is_ppclk_dpm_enabled(clk_mgr, PPCLK_DTBCLK)) {
+		/* DPM enabled, use currently set value */
+		dtb_ref_clk_khz = clk_mgr_base->clks.ref_dtbclk_khz;
+	} else {
+		/* DPM disabled, so use boot snapshot */
+		dtb_ref_clk_khz = clk_mgr_base->boot_snapshot.dtbclk;
+	}
+
+	return dtb_ref_clk_khz;
+}
+
+static int dcn401_get_dispclk_from_dentist(struct clk_mgr *clk_mgr_base)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+	uint32_t dispclk_wdivider;
+	int disp_divider;
+
+	REG_GET(DENTIST_DISPCLK_CNTL, DENTIST_DISPCLK_WDIVIDER, &dispclk_wdivider);
+	disp_divider = dentist_get_divider_from_did(dispclk_wdivider);
+
+	/* Return DISPCLK freq in Khz */
+	if (disp_divider)
+		return (DENTIST_DIVIDER_RANGE_SCALE_FACTOR * clk_mgr->base.dentist_vco_freq_khz) / disp_divider;
+
+	return 0;
+}
+
+unsigned int dcn401_get_max_clock_khz(struct clk_mgr *clk_mgr_base, enum clk_type clk_type)
+{
+	struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base);
+
+	unsigned int num_clk_levels;
+
+	switch (clk_type) {
+	case CLK_TYPE_DISPCLK:
+		num_clk_levels = clk_mgr->base.bw_params->clk_table.num_entries_per_clk.num_dispclk_levels;
+		return dcn401_is_ppclk_dpm_enabled(clk_mgr, PPCLK_DISPCLK) ?
+				clk_mgr->base.bw_params->clk_table.entries[num_clk_levels - 1].dispclk_mhz * 1000 :
+				clk_mgr->base.boot_snapshot.dispclk;
+	case CLK_TYPE_DPPCLK:
+		num_clk_levels = clk_mgr->base.bw_params->clk_table.num_entries_per_clk.num_dppclk_levels;
+		return dcn401_is_ppclk_dpm_enabled(clk_mgr, PPCLK_DPPCLK) ?
+				clk_mgr->base.bw_params->clk_table.entries[num_clk_levels - 1].dppclk_mhz * 1000 :
+				clk_mgr->base.boot_snapshot.dppclk;
+	case CLK_TYPE_DSCCLK:
+		num_clk_levels = clk_mgr->base.bw_params->clk_table.num_entries_per_clk.num_dispclk_levels;
+		return dcn401_is_ppclk_dpm_enabled(clk_mgr, PPCLK_DISPCLK) ?
+				clk_mgr->base.bw_params->clk_table.entries[num_clk_levels - 1].dispclk_mhz * 1000 / 3 :
+				clk_mgr->base.boot_snapshot.dispclk / 3;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static struct clk_mgr_funcs dcn401_funcs = {
+		.get_dp_ref_clk_frequency = dce12_get_dp_ref_freq_khz,
+		.get_dtb_ref_clk_frequency = dcn401_get_dtb_ref_freq_khz,
+		.update_clocks = dcn401_update_clocks,
+		.dump_clk_registers = dcn401_dump_clk_registers,
+		.init_clocks = dcn401_init_clocks,
+		.notify_wm_ranges = dcn401_notify_wm_ranges,
+		.set_hard_min_memclk = dcn401_set_hard_min_memclk,
+		.get_memclk_states_from_smu = dcn401_get_memclk_states_from_smu,
+		.are_clock_states_equal = dcn401_are_clock_states_equal,
+		.enable_pme_wa = dcn401_enable_pme_wa,
+		.is_smu_present = dcn401_is_smu_present,
+		.get_dispclk_from_dentist = dcn401_get_dispclk_from_dentist,
+		.get_hard_min_memclk = dcn401_get_hard_min_memclk,
+		.get_hard_min_fclk = dcn401_get_hard_min_fclk,
+		.is_dc_mode_present = dcn401_is_dc_mode_present,
+		.get_max_clock_khz = dcn401_get_max_clock_khz,
+};
+
+struct clk_mgr_internal *dcn401_clk_mgr_construct(
+		struct dc_context *ctx,
+		struct dccg *dccg)
+{
+	struct clk_log_info log_info = {0};
+	struct dcn401_clk_mgr *clk_mgr401 = kzalloc(sizeof(struct dcn401_clk_mgr), GFP_KERNEL);
+	struct clk_mgr_internal *clk_mgr;
+
+	if (!clk_mgr401)
+		return NULL;
+
+	clk_mgr = &clk_mgr401->base;
+	clk_mgr->base.ctx = ctx;
+	clk_mgr->base.funcs = &dcn401_funcs;
+	clk_mgr->regs = &clk_mgr_regs_dcn401;
+	clk_mgr->clk_mgr_shift = &clk_mgr_shift_dcn401;
+	clk_mgr->clk_mgr_mask = &clk_mgr_mask_dcn401;
+
+	clk_mgr->dccg = dccg;
+	clk_mgr->dfs_bypass_disp_clk = 0;
+
+	clk_mgr->dprefclk_ss_percentage = 0;
+	clk_mgr->dprefclk_ss_divider = 1000;
+	clk_mgr->ss_on_dprefclk = false;
+	clk_mgr->dfs_ref_freq_khz = 100000;
+
+	/* Changed from DCN3.2_clock_frequency doc to match
+	 * dcn401_dump_clk_registers from 4 * dentist_vco_freq_khz /
+	 * dprefclk DID divider
+	 */
+	clk_mgr->base.dprefclk_khz = 720000; //TODO update from VBIOS
+
+	/* integer part is now VCO frequency in kHz */
+	clk_mgr->base.dentist_vco_freq_khz = dcn401_get_vco_frequency_from_reg(clk_mgr);
+
+	/* in case we don't get a value from the register, use default */
+	if (clk_mgr->base.dentist_vco_freq_khz == 0)
+		clk_mgr->base.dentist_vco_freq_khz = 4500000; //TODO Update from VBIOS
+
+	dcn401_dump_clk_registers(&clk_mgr->base.boot_snapshot, &clk_mgr->base, &log_info);
+
+	if (ctx->dc->debug.disable_dtb_ref_clk_switch &&
+			clk_mgr->base.clks.ref_dtbclk_khz != clk_mgr->base.boot_snapshot.dtbclk) {
+		clk_mgr->base.clks.ref_dtbclk_khz = clk_mgr->base.boot_snapshot.dtbclk;
+	}
+
+	if (clk_mgr->base.boot_snapshot.dprefclk != 0) {
+		clk_mgr->base.dprefclk_khz = clk_mgr->base.boot_snapshot.dprefclk;
+	}
+	dcn401_clock_read_ss_info(clk_mgr);
+
+	clk_mgr->dfs_bypass_enabled = false;
+
+	clk_mgr->smu_present = false;
+
+	clk_mgr->base.bw_params = kzalloc(sizeof(*clk_mgr->base.bw_params), GFP_KERNEL);
+	if (!clk_mgr->base.bw_params) {
+		BREAK_TO_DEBUGGER();
+		kfree(clk_mgr401);
+		return NULL;
+	}
+
+	/* need physical address of table to give to PMFW */
+	clk_mgr->wm_range_table = dm_helpers_allocate_gpu_mem(clk_mgr->base.ctx,
+			DC_MEM_ALLOC_TYPE_GART, sizeof(WatermarksExternal_t),
+			&clk_mgr->wm_range_table_addr);
+	if (!clk_mgr->wm_range_table) {
+		BREAK_TO_DEBUGGER();
+		kfree(clk_mgr->base.bw_params);
+		kfree(clk_mgr401);
+		return NULL;
+	}
+
+	return &clk_mgr401->base;
+}
+
+void dcn401_clk_mgr_destroy(struct clk_mgr_internal *clk_mgr)
+{
+	kfree(clk_mgr->base.bw_params);
+
+	if (clk_mgr->wm_range_table)
+		dm_helpers_free_gpu_mem(clk_mgr->base.ctx, DC_MEM_ALLOC_TYPE_GART,
+				clk_mgr->wm_range_table);
+}
+
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn401/dcn401_clk_mgr.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn401/dcn401_clk_mgr.h
new file mode 100644
index 000000000000..97a1ce1e8a9e
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn401/dcn401_clk_mgr.h
@@ -0,0 +1,117 @@
+// SPDX-License-Identifier: MIT
+//
+// Copyright 2024 Advanced Micro Devices, Inc.
+
+#ifndef __DCN401_CLK_MGR_H_
+#define __DCN401_CLK_MGR_H_
+
+#define DCN401_CLK_MGR_MAX_SEQUENCE_SIZE 30
+
+union dcn401_clk_mgr_block_sequence_params {
+	struct {
+		/* inputs */
+		uint32_t num_displays;
+	} update_num_displays_params;
+	struct {
+		/* inputs */
+		uint32_t ppclk;
+		uint16_t freq_mhz;
+		/* outputs */
+		uint32_t *response;
+	} update_hardmin_params;
+	struct {
+		/* inputs */
+		uint32_t ppclk;
+		int freq_khz;
+		/* outputs */
+		uint32_t *response;
+	} update_hardmin_optimized_params;
+	struct {
+		/* inputs */
+		uint16_t uclk_mhz;
+		uint16_t fclk_mhz;
+	} update_idle_hardmin_params;
+	struct {
+		/* inputs */
+		uint16_t freq_mhz;
+	} update_deep_sleep_dcfclk_params;
+	struct {
+		/* inputs */
+		bool support;
+	} update_pstate_support_params;
+	struct {
+		/* inputs */
+		unsigned int num_ways;
+	} update_cab_for_uclk_params;
+	struct {
+		/* inputs */
+		bool enable;
+	} update_wait_for_dmub_ack_params;
+	struct {
+		/* inputs */
+		bool mod_drr_for_pstate;
+	} indicate_drr_status_params;
+	struct {
+		/* inputs */
+		struct dc_state *context;
+		int *ref_dppclk_khz;
+		bool safe_to_lower;
+	} update_dppclk_dto_params;
+	struct {
+		/* inputs */
+		struct dc_state *context;
+		int *ref_dtbclk_khz;
+	} update_dtbclk_dto_params;
+	struct {
+		/* inputs */
+		struct dc_state *context;
+	} update_dentist_params;
+	struct {
+		/* inputs */
+		struct dmcu *dmcu;
+		unsigned int wait;
+	} update_psr_wait_loop_params;
+};
+
+enum dcn401_clk_mgr_block_sequence_func {
+	CLK_MGR401_READ_CLOCKS_FROM_DENTIST,
+	CLK_MGR401_UPDATE_NUM_DISPLAYS,
+	CLK_MGR401_UPDATE_HARDMIN_PPCLK,
+	CLK_MGR401_UPDATE_HARDMIN_PPCLK_OPTIMIZED,
+	CLK_MGR401_UPDATE_ACTIVE_HARDMINS,
+	CLK_MGR401_UPDATE_IDLE_HARDMINS,
+	CLK_MGR401_UPDATE_DEEP_SLEEP_DCFCLK,
+	CLK_MGR401_UPDATE_FCLK_PSTATE_SUPPORT,
+	CLK_MGR401_UPDATE_UCLK_PSTATE_SUPPORT,
+	CLK_MGR401_UPDATE_CAB_FOR_UCLK,
+	CLK_MGR401_UPDATE_WAIT_FOR_DMUB_ACK,
+	CLK_MGR401_INDICATE_DRR_STATUS,
+	CLK_MGR401_UPDATE_DPPCLK_DTO,
+	CLK_MGR401_UPDATE_DTBCLK_DTO,
+	CLK_MGR401_UPDATE_DENTIST,
+	CLK_MGR401_UPDATE_PSR_WAIT_LOOP,
+	CLK_MGR401_UPDATE_SUBVP_HARDMINS,
+};
+
+struct dcn401_clk_mgr_block_sequence {
+	union dcn401_clk_mgr_block_sequence_params params;
+	enum dcn401_clk_mgr_block_sequence_func func;
+};
+
+struct dcn401_clk_mgr {
+	struct clk_mgr_internal base;
+
+	struct dcn401_clk_mgr_block_sequence block_sequence[DCN401_CLK_MGR_MAX_SEQUENCE_SIZE];
+};
+
+void dcn401_init_clocks(struct clk_mgr *clk_mgr_base);
+bool dcn401_is_dc_mode_present(struct clk_mgr *clk_mgr_base);
+
+struct clk_mgr_internal *dcn401_clk_mgr_construct(struct dc_context *ctx,
+		struct dccg *dccg);
+
+void dcn401_clk_mgr_destroy(struct clk_mgr_internal *clk_mgr);
+
+unsigned int dcn401_get_max_clock_khz(struct clk_mgr *clk_mgr_base, enum clk_type clk_type);
+
+#endif /* __DCN401_CLK_MGR_H_ */
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn401/dcn401_clk_mgr_smu_msg.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn401/dcn401_clk_mgr_smu_msg.c
new file mode 100644
index 000000000000..3a263840893e
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn401/dcn401_clk_mgr_smu_msg.c
@@ -0,0 +1,472 @@
+// SPDX-License-Identifier: MIT
+//
+// Copyright 2024 Advanced Micro Devices, Inc.
+
+#include "dcn401_clk_mgr_smu_msg.h"
+
+#include "clk_mgr_internal.h"
+#include "reg_helper.h"
+
+#include "dalsmc.h"
+#include "dcn401_smu14_driver_if.h"
+
+#define mmDAL_MSG_REG  0x1628A
+#define mmDAL_ARG_REG  0x16273
+#define mmDAL_RESP_REG 0x16274
+
+#define REG(reg_name) \
+	mm ## reg_name
+
+#include "logger_types.h"
+
+#define smu_print(str, ...) {DC_LOG_SMU(str, ##__VA_ARGS__); }
+
+/* temporary define */
+#ifndef DALSMC_MSG_SubvpUclkFclk
+#define DALSMC_MSG_SubvpUclkFclk 0x1B
+#endif
+#ifndef DALSMC_MSG_GetNumUmcChannels
+#define DALSMC_MSG_GetNumUmcChannels 0x1C
+#endif
+
+/*
+ * Function to be used instead of REG_WAIT macro because the wait ends when
+ * the register is NOT EQUAL to zero, and because the translation in msg_if.h
+ * won't work with REG_WAIT.
+ */
+static uint32_t dcn401_smu_wait_for_response(struct clk_mgr_internal *clk_mgr, unsigned int delay_us, unsigned int max_retries)
+{
+	uint32_t reg = 0;
+
+	do {
+		reg = REG_READ(DAL_RESP_REG);
+		if (reg)
+			break;
+
+		if (delay_us >= 1000)
+			msleep(delay_us/1000);
+		else if (delay_us > 0)
+			udelay(delay_us);
+	} while (max_retries--);
+
+	return reg;
+}
+
+static bool dcn401_smu_send_msg_with_param(struct clk_mgr_internal *clk_mgr, uint32_t msg_id, uint32_t param_in, uint32_t *param_out)
+{
+	/* Wait for response register to be ready */
+	dcn401_smu_wait_for_response(clk_mgr, 10, 200000);
+
+	TRACE_SMU_MSG_ENTER(msg_id, param_in, clk_mgr->base.ctx);
+
+	/* Clear response register */
+	REG_WRITE(DAL_RESP_REG, 0);
+
+	/* Set the parameter register for the SMU message */
+	REG_WRITE(DAL_ARG_REG, param_in);
+
+	/* Trigger the message transaction by writing the message ID */
+	REG_WRITE(DAL_MSG_REG, msg_id);
+
+	/* Wait for response */
+	if (dcn401_smu_wait_for_response(clk_mgr, 10, 200000) == DALSMC_Result_OK) {
+		if (param_out)
+			*param_out = REG_READ(DAL_ARG_REG);
+
+		TRACE_SMU_MSG_EXIT(true, param_out ? *param_out : 0, clk_mgr->base.ctx);
+		return true;
+	}
+
+	TRACE_SMU_MSG_EXIT(false, 0, clk_mgr->base.ctx);
+	return false;
+}
+
+/*
+ * Use these functions to return back delay information so we can aggregate the total
+ *  delay when requesting hardmin clk
+ *
+ * dcn401_smu_wait_for_response_delay
+ * dcn401_smu_send_msg_with_param_delay
+ *
+ */
+static uint32_t dcn401_smu_wait_for_response_delay(struct clk_mgr_internal *clk_mgr, unsigned int delay_us, unsigned int max_retries, unsigned int *total_delay_us)
+{
+	uint32_t reg = 0;
+	*total_delay_us = 0;
+
+	do {
+		reg = REG_READ(DAL_RESP_REG);
+		if (reg)
+			break;
+
+		if (delay_us >= 1000)
+			msleep(delay_us/1000);
+		else if (delay_us > 0)
+			udelay(delay_us);
+		*total_delay_us += delay_us;
+	} while (max_retries--);
+
+	return reg;
+}
+
+static bool dcn401_smu_send_msg_with_param_delay(struct clk_mgr_internal *clk_mgr, uint32_t msg_id, uint32_t param_in, uint32_t *param_out, unsigned int *total_delay_us)
+{
+	unsigned int delay1_us, delay2_us;
+	*total_delay_us = 0;
+
+	/* Wait for response register to be ready */
+	dcn401_smu_wait_for_response_delay(clk_mgr, 10, 200000, &delay1_us);
+
+	TRACE_SMU_MSG_ENTER(msg_id, param_in, clk_mgr->base.ctx);
+
+	/* Clear response register */
+	REG_WRITE(DAL_RESP_REG, 0);
+
+	/* Set the parameter register for the SMU message */
+	REG_WRITE(DAL_ARG_REG, param_in);
+
+	/* Trigger the message transaction by writing the message ID */
+	REG_WRITE(DAL_MSG_REG, msg_id);
+
+	/* Wait for response */
+	if (dcn401_smu_wait_for_response_delay(clk_mgr, 10, 200000, &delay2_us) == DALSMC_Result_OK) {
+		if (param_out)
+			*param_out = REG_READ(DAL_ARG_REG);
+
+		*total_delay_us = delay1_us + delay2_us;
+		TRACE_SMU_MSG_EXIT(true, param_out ? *param_out : 0, clk_mgr->base.ctx);
+		return true;
+	}
+
+	*total_delay_us = delay1_us + 2000000;
+	TRACE_SMU_MSG_EXIT(false, 0, clk_mgr->base.ctx);
+	return false;
+}
+
+bool dcn401_smu_get_smu_version(struct clk_mgr_internal *clk_mgr, unsigned int *version)
+{
+	smu_print("SMU Get SMU version\n");
+
+	if (dcn401_smu_send_msg_with_param(clk_mgr,
+			DALSMC_MSG_GetSmuVersion, 0, version)) {
+
+		smu_print("SMU version: %d\n", *version);
+
+		return true;
+	}
+
+	return false;
+}
+
+/* Message output should match SMU11_DRIVER_IF_VERSION in smu11_driver_if.h */
+bool dcn401_smu_check_driver_if_version(struct clk_mgr_internal *clk_mgr)
+{
+	uint32_t response = 0;
+
+	smu_print("SMU Check driver if version\n");
+
+	if (dcn401_smu_send_msg_with_param(clk_mgr,
+			DALSMC_MSG_GetDriverIfVersion, 0, &response)) {
+
+		smu_print("SMU driver if version: %d\n", response);
+
+		if (response == SMU14_DRIVER_IF_VERSION)
+			return true;
+	}
+
+	return false;
+}
+
+/* Message output should match DALSMC_VERSION in dalsmc.h */
+bool dcn401_smu_check_msg_header_version(struct clk_mgr_internal *clk_mgr)
+{
+	uint32_t response = 0;
+
+	smu_print("SMU Check msg header version\n");
+
+	if (dcn401_smu_send_msg_with_param(clk_mgr,
+			DALSMC_MSG_GetMsgHeaderVersion, 0, &response)) {
+
+		smu_print("SMU msg header version: %d\n", response);
+
+		if (response == DALSMC_VERSION)
+			return true;
+	}
+
+	return false;
+}
+
+void dcn401_smu_send_fclk_pstate_message(struct clk_mgr_internal *clk_mgr, bool support)
+{
+	smu_print("FCLK P-state support value is : %d\n", support);
+
+	dcn401_smu_send_msg_with_param(clk_mgr,
+			DALSMC_MSG_SetFclkSwitchAllow, support, NULL);
+}
+
+void dcn401_smu_send_uclk_pstate_message(struct clk_mgr_internal *clk_mgr, bool support)
+{
+	smu_print("UCLK P-state support value is : %d\n", support);
+
+	dcn401_smu_send_msg_with_param(clk_mgr,
+			DALSMC_MSG_SetUclkPstateAllow, support, NULL);
+}
+
+void dcn401_smu_send_cab_for_uclk_message(struct clk_mgr_internal *clk_mgr, unsigned int num_ways)
+{
+	uint32_t param = (num_ways << 1) | (num_ways > 0);
+
+	dcn401_smu_send_msg_with_param(clk_mgr, DALSMC_MSG_SetCabForUclkPstate, param, NULL);
+	smu_print("Numways for SubVP : %d\n", num_ways);
+}
+
+void dcn401_smu_set_dram_addr_high(struct clk_mgr_internal *clk_mgr, uint32_t addr_high)
+{
+	smu_print("SMU Set DRAM addr high: %d\n", addr_high);
+
+	dcn401_smu_send_msg_with_param(clk_mgr,
+			DALSMC_MSG_SetDalDramAddrHigh, addr_high, NULL);
+}
+
+void dcn401_smu_set_dram_addr_low(struct clk_mgr_internal *clk_mgr, uint32_t addr_low)
+{
+	smu_print("SMU Set DRAM addr low: %d\n", addr_low);
+
+	dcn401_smu_send_msg_with_param(clk_mgr,
+			DALSMC_MSG_SetDalDramAddrLow, addr_low, NULL);
+}
+
+void dcn401_smu_transfer_wm_table_dram_2_smu(struct clk_mgr_internal *clk_mgr)
+{
+	smu_print("SMU Transfer WM table DRAM 2 SMU\n");
+
+	dcn401_smu_send_msg_with_param(clk_mgr,
+			DALSMC_MSG_TransferTableDram2Smu, TABLE_WATERMARKS, NULL);
+}
+
+void dcn401_smu_set_pme_workaround(struct clk_mgr_internal *clk_mgr)
+{
+	smu_print("SMU Set PME workaround\n");
+
+	dcn401_smu_send_msg_with_param(clk_mgr,
+		DALSMC_MSG_BacoAudioD3PME, 0, NULL);
+}
+
+static unsigned int dcn401_smu_get_hard_min_status(struct clk_mgr_internal *clk_mgr, bool *no_timeout, unsigned int *total_delay_us)
+{
+	uint32_t response = 0;
+
+	/* bits 23:16 for clock type, lower 16 bits for frequency in MHz */
+	uint32_t param = 0;
+
+	*no_timeout = dcn401_smu_send_msg_with_param_delay(clk_mgr,
+			DALSMC_MSG_ReturnHardMinStatus, param, &response, total_delay_us);
+
+	smu_print("SMU Get hard min status: no_timeout %d delay %d us clk bits %x\n",
+		*no_timeout, *total_delay_us, response);
+
+	return response;
+}
+
+static bool dcn401_smu_wait_hard_min_status(struct clk_mgr_internal *clk_mgr, uint32_t ppclk)
+{
+	const unsigned int max_delay_us = 1000000;
+
+	unsigned int hardmin_status_mask = (1 << ppclk);
+	unsigned int total_delay_us = 0;
+	bool hardmin_done = false;
+
+	while (!hardmin_done && total_delay_us < max_delay_us) {
+		unsigned int hardmin_status;
+		unsigned int read_total_delay_us;
+		bool no_timeout;
+
+		if (!hardmin_done && total_delay_us > 0) {
+			/* hardmin not yet fulfilled, wait 500us and retry*/
+			udelay(500);
+			total_delay_us += 500;
+
+			smu_print("SMU Wait hard min status for %d us\n", total_delay_us);
+		}
+
+		hardmin_status = dcn401_smu_get_hard_min_status(clk_mgr, &no_timeout, &read_total_delay_us);
+		total_delay_us += read_total_delay_us;
+		hardmin_done = hardmin_status & hardmin_status_mask;
+	}
+
+	return hardmin_done;
+}
+
+/* Returns the actual frequency that was set in MHz, 0 on failure */
+unsigned int dcn401_smu_set_hard_min_by_freq(struct clk_mgr_internal *clk_mgr, uint32_t clk, uint16_t freq_mhz)
+{
+	uint32_t response = 0;
+	bool hard_min_done = false;
+
+	/* bits 23:16 for clock type, lower 16 bits for frequency in MHz */
+	uint32_t param = (clk << 16) | freq_mhz;
+
+	smu_print("SMU Set hard min by freq: clk = %d, freq_mhz = %d MHz\n", clk, freq_mhz);
+
+	dcn401_smu_send_msg_with_param(clk_mgr,
+			DALSMC_MSG_SetHardMinByFreq, param, &response);
+
+	/* wait until hardmin acknowledged */
+	hard_min_done = dcn401_smu_wait_hard_min_status(clk_mgr, clk);
+	smu_print("SMU Frequency set = %d KHz hard_min_done %d\n", response, hard_min_done);
+
+	return response;
+}
+
+void dcn401_smu_wait_for_dmub_ack_mclk(struct clk_mgr_internal *clk_mgr, bool enable)
+{
+	smu_print("SMU to wait for DMCUB ack for MCLK : %d\n", enable);
+
+	dcn401_smu_send_msg_with_param(clk_mgr, DALSMC_MSG_SetAlwaysWaitDmcubResp, enable ? 1 : 0, NULL);
+}
+
+void dcn401_smu_indicate_drr_status(struct clk_mgr_internal *clk_mgr, bool mod_drr_for_pstate)
+{
+	smu_print("SMU Set indicate drr status = %d\n", mod_drr_for_pstate);
+
+	dcn401_smu_send_msg_with_param(clk_mgr,
+			DALSMC_MSG_IndicateDrrStatus, mod_drr_for_pstate ? 1 : 0, NULL);
+}
+
+bool dcn401_smu_set_idle_uclk_fclk_hardmin(struct clk_mgr_internal *clk_mgr,
+		uint16_t uclk_freq_mhz,
+		uint16_t fclk_freq_mhz)
+{
+	uint32_t response = 0;
+	bool success;
+
+	/* 15:0 for uclk, 32:16 for fclk */
+	uint32_t param = (fclk_freq_mhz << 16) | uclk_freq_mhz;
+
+	smu_print("SMU Set idle hardmin by freq: uclk_freq_mhz = %d MHz, fclk_freq_mhz = %d MHz\n", uclk_freq_mhz, fclk_freq_mhz);
+
+	success = dcn401_smu_send_msg_with_param(clk_mgr,
+			DALSMC_MSG_IdleUclkFclk, param, &response);
+
+	/* wait until hardmin acknowledged */
+	success &= dcn401_smu_wait_hard_min_status(clk_mgr, PPCLK_UCLK);
+	smu_print("SMU hard_min_done %d\n", success);
+
+	return success;
+}
+
+bool dcn401_smu_set_active_uclk_fclk_hardmin(struct clk_mgr_internal *clk_mgr,
+		uint16_t uclk_freq_mhz,
+		uint16_t fclk_freq_mhz)
+{
+	uint32_t response = 0;
+	bool success;
+
+	/* 15:0 for uclk, 32:16 for fclk */
+	uint32_t param = (fclk_freq_mhz << 16) | uclk_freq_mhz;
+
+	smu_print("SMU Set active hardmin by freq: uclk_freq_mhz = %d MHz, fclk_freq_mhz = %d MHz\n", uclk_freq_mhz, fclk_freq_mhz);
+
+	success = dcn401_smu_send_msg_with_param(clk_mgr,
+			DALSMC_MSG_ActiveUclkFclk, param, &response);
+
+	/* wait until hardmin acknowledged */
+	success &= dcn401_smu_wait_hard_min_status(clk_mgr, PPCLK_UCLK);
+	smu_print("SMU hard_min_done %d\n", success);
+
+	return success;
+}
+
+bool dcn401_smu_set_subvp_uclk_fclk_hardmin(struct clk_mgr_internal *clk_mgr,
+		uint16_t uclk_freq_mhz,
+		uint16_t fclk_freq_mhz)
+{
+	uint32_t response = 0;
+	bool success;
+
+	/* 15:0 for uclk, 32:16 for fclk */
+	uint32_t param = (fclk_freq_mhz << 16) | uclk_freq_mhz;
+
+	smu_print("SMU Set active hardmin by freq: uclk_freq_mhz = %d MHz, fclk_freq_mhz = %d MHz\n", uclk_freq_mhz, fclk_freq_mhz);
+
+	success = dcn401_smu_send_msg_with_param(clk_mgr,
+			DALSMC_MSG_SubvpUclkFclk, param, &response);
+
+	return success;
+}
+
+void dcn401_smu_set_min_deep_sleep_dcef_clk(struct clk_mgr_internal *clk_mgr, uint32_t freq_mhz)
+{
+	smu_print("SMU Set min deep sleep dcef clk: freq_mhz = %d MHz\n", freq_mhz);
+
+	dcn401_smu_send_msg_with_param(clk_mgr,
+			DALSMC_MSG_SetMinDeepSleepDcfclk, freq_mhz, NULL);
+}
+
+void dcn401_smu_set_num_of_displays(struct clk_mgr_internal *clk_mgr, uint32_t num_displays)
+{
+	smu_print("SMU Set num of displays: num_displays = %d\n", num_displays);
+
+	dcn401_smu_send_msg_with_param(clk_mgr,
+			DALSMC_MSG_NumOfDisplays, num_displays, NULL);
+}
+
+unsigned int dcn401_smu_get_num_of_umc_channels(struct clk_mgr_internal *clk_mgr)
+{
+	unsigned int response = 0;
+
+	dcn401_smu_send_msg_with_param(clk_mgr, DALSMC_MSG_GetNumUmcChannels, 0, &response);
+
+	smu_print("SMU Get Num UMC Channels: num_umc_channels = %d\n", response);
+
+	return response;
+}
+
+/*
+ * Frequency in MHz returned in lower 16 bits for valid DPM level
+ *
+ * Call with dpm_level = 0xFF to query features, return value will be:
+ *     Bits 7:0 - number of DPM levels
+ *     Bit   28 - 1 = auto DPM on
+ *     Bit   29 - 1 = sweep DPM on
+ *     Bit   30 - 1 = forced DPM on
+ *     Bit   31 - 0 = discrete, 1 = fine-grained
+ *
+ * With fine-grained DPM, only min and max frequencies will be reported
+ *
+ * Returns 0 on failure
+ */
+unsigned int dcn401_smu_get_dpm_freq_by_index(struct clk_mgr_internal *clk_mgr, uint32_t clk, uint8_t dpm_level)
+{
+	uint32_t response = 0;
+
+	/* bits 23:16 for clock type, lower 8 bits for DPM level */
+	uint32_t param = (clk << 16) | dpm_level;
+
+	smu_print("SMU Get dpm freq by index: clk = %d, dpm_level = %d\n", clk, dpm_level);
+
+	dcn401_smu_send_msg_with_param(clk_mgr,
+			DALSMC_MSG_GetDpmFreqByIndex, param, &response);
+
+	smu_print("SMU dpm freq: %d MHz\n", response);
+
+	return response;
+}
+
+/* Returns the max DPM frequency in DC mode in MHz, 0 on failure */
+unsigned int dcn401_smu_get_dc_mode_max_dpm_freq(struct clk_mgr_internal *clk_mgr, uint32_t clk)
+{
+	uint32_t response = 0;
+
+	/* bits 23:16 for clock type */
+	uint32_t param = clk << 16;
+
+	smu_print("SMU Get DC mode max DPM freq: clk = %d\n", clk);
+
+	dcn401_smu_send_msg_with_param(clk_mgr,
+			DALSMC_MSG_GetDcModeMaxDpmFreq, param, &response);
+
+	smu_print("SMU DC mode max DMP freq: %d MHz\n", response);
+
+	return response;
+}
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn401/dcn401_clk_mgr_smu_msg.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn401/dcn401_clk_mgr_smu_msg.h
new file mode 100644
index 000000000000..4f5ac603e822
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn401/dcn401_clk_mgr_smu_msg.h
@@ -0,0 +1,41 @@
+// SPDX-License-Identifier: MIT
+//
+// Copyright 2024 Advanced Micro Devices, Inc.
+
+#ifndef __DCN401_CLK_MGR_SMU_MSG_H_
+#define __DCN401_CLK_MGR_SMU_MSG_H_
+
+#include "os_types.h"
+#include "core_types.h"
+
+struct clk_mgr_internal;
+
+bool dcn401_smu_get_smu_version(struct clk_mgr_internal *clk_mgr, unsigned int *version);
+bool dcn401_smu_check_driver_if_version(struct clk_mgr_internal *clk_mgr);
+bool dcn401_smu_check_msg_header_version(struct clk_mgr_internal *clk_mgr);
+void dcn401_smu_send_fclk_pstate_message(struct clk_mgr_internal *clk_mgr, bool support);
+void dcn401_smu_send_uclk_pstate_message(struct clk_mgr_internal *clk_mgr, bool support);
+void dcn401_smu_send_cab_for_uclk_message(struct clk_mgr_internal *clk_mgr, unsigned int num_ways);
+void dcn401_smu_set_dram_addr_high(struct clk_mgr_internal *clk_mgr, uint32_t addr_high);
+void dcn401_smu_set_dram_addr_low(struct clk_mgr_internal *clk_mgr, uint32_t addr_low);
+void dcn401_smu_transfer_wm_table_dram_2_smu(struct clk_mgr_internal *clk_mgr);
+void dcn401_smu_set_pme_workaround(struct clk_mgr_internal *clk_mgr);
+unsigned int dcn401_smu_set_hard_min_by_freq(struct clk_mgr_internal *clk_mgr, uint32_t clk, uint16_t freq_mhz);
+void dcn401_smu_wait_for_dmub_ack_mclk(struct clk_mgr_internal *clk_mgr, bool enable);
+void dcn401_smu_indicate_drr_status(struct clk_mgr_internal *clk_mgr, bool mod_drr_for_pstate);
+bool dcn401_smu_set_idle_uclk_fclk_hardmin(struct clk_mgr_internal *clk_mgr,
+		uint16_t uclk_freq_mhz,
+		uint16_t fclk_freq_mhz);
+bool dcn401_smu_set_active_uclk_fclk_hardmin(struct clk_mgr_internal *clk_mgr,
+		uint16_t uclk_freq_mhz,
+		uint16_t fclk_freq_mhz);
+bool dcn401_smu_set_subvp_uclk_fclk_hardmin(struct clk_mgr_internal *clk_mgr,
+		uint16_t uclk_freq_mhz,
+		uint16_t fclk_freq_mhz);
+void dcn401_smu_set_min_deep_sleep_dcef_clk(struct clk_mgr_internal *clk_mgr, uint32_t freq_mhz);
+void dcn401_smu_set_num_of_displays(struct clk_mgr_internal *clk_mgr, uint32_t num_displays);
+unsigned int dcn401_smu_get_num_of_umc_channels(struct clk_mgr_internal *clk_mgr);
+unsigned int dcn401_smu_get_dc_mode_max_dpm_freq(struct clk_mgr_internal *clk_mgr, uint32_t clk);
+unsigned int dcn401_smu_get_dpm_freq_by_index(struct clk_mgr_internal *clk_mgr, uint32_t clk, uint8_t dpm_level);
+
+#endif /* __DCN401_CLK_MGR_SMU_MSG_H_ */
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn401/dcn401_smu14_driver_if.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn401/dcn401_smu14_driver_if.h
new file mode 100644
index 000000000000..36034b32870c
--- /dev/null
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn401/dcn401_smu14_driver_if.h
@@ -0,0 +1,66 @@
+// SPDX-License-Identifier: MIT
+//
+// Copyright 2024 Advanced Micro Devices, Inc.
+//
+// This is a stripped-down version of the smu13_driver_if.h file for the relevant DAL interfaces.
+
+#define SMU14_DRIVER_IF_VERSION  0x1
+
+//Only Clks that have DPM descriptors are listed here
+typedef enum {
+  PPCLK_GFXCLK = 0,
+  PPCLK_SOCCLK,
+  PPCLK_UCLK,
+  PPCLK_FCLK,
+  PPCLK_DCLK_0,
+  PPCLK_VCLK_0,
+  PPCLK_DISPCLK,
+  PPCLK_DPPCLK,
+  PPCLK_DPREFCLK,
+  PPCLK_DCFCLK,
+  PPCLK_DTBCLK,
+  PPCLK_COUNT,
+} PPCLK_e;
+
+typedef struct {
+  uint8_t  WmSetting;
+  uint8_t  Flags;
+  uint8_t  Padding[2];
+
+} WatermarkRowGeneric_t;
+
+#define NUM_WM_RANGES 4
+
+typedef enum {
+  WATERMARKS_CLOCK_RANGE = 0,
+  WATERMARKS_DUMMY_PSTATE,
+  WATERMARKS_MALL,
+  WATERMARKS_COUNT,
+} WATERMARKS_FLAGS_e;
+
+typedef struct {
+  // Watermarks
+  WatermarkRowGeneric_t WatermarkRow[NUM_WM_RANGES];
+} Watermarks_t;
+
+typedef struct {
+  Watermarks_t Watermarks;
+  uint32_t  Spare[16];
+
+  uint32_t     MmHubPadding[8]; // SMU internal use
+} WatermarksExternal_t;
+
+// Table types
+#define TABLE_PMFW_PPTABLE            0
+#define TABLE_COMBO_PPTABLE           1
+#define TABLE_WATERMARKS              2
+#define TABLE_AVFS_PSM_DEBUG          3
+#define TABLE_PMSTATUSLOG             4
+#define TABLE_SMU_METRICS             5
+#define TABLE_DRIVER_SMU_CONFIG       6
+#define TABLE_ACTIVITY_MONITOR_COEFF  7
+#define TABLE_OVERDRIVE               8
+#define TABLE_I2C_COMMANDS            9
+#define TABLE_DRIVER_INFO             10
+#define TABLE_ECCINFO                 11
+#define TABLE_COUNT                   12
diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c
index 52564b93f7eb..8be9cbd43e18 100644
--- a/drivers/gpu/drm/amd/display/dc/core/dc.c
+++ b/drivers/gpu/drm/amd/display/dc/core/dc.c